2007-05-09 01:00:38 +00:00
|
|
|
/*
|
|
|
|
|
* Copyright (c) 2004, 2005 Topspin Communications. All rights reserved.
|
|
|
|
|
* Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved.
|
2008-07-25 17:32:52 +00:00
|
|
|
* Copyright (c) 2005, 2006, 2007, 2008 Mellanox Technologies. All rights reserved.
|
2007-05-09 01:00:38 +00:00
|
|
|
* Copyright (c) 2006, 2007 Cisco Systems, Inc. All rights reserved.
|
|
|
|
|
*
|
|
|
|
|
* This software is available to you under a choice of one of two
|
|
|
|
|
* licenses. You may choose to be licensed under the terms of the GNU
|
|
|
|
|
* General Public License (GPL) Version 2, available from the file
|
|
|
|
|
* COPYING in the main directory of this source tree, or the
|
|
|
|
|
* OpenIB.org BSD license below:
|
|
|
|
|
*
|
|
|
|
|
* Redistribution and use in source and binary forms, with or
|
|
|
|
|
* without modification, are permitted provided that the following
|
|
|
|
|
* conditions are met:
|
|
|
|
|
*
|
|
|
|
|
* - Redistributions of source code must retain the above
|
|
|
|
|
* copyright notice, this list of conditions and the following
|
|
|
|
|
* disclaimer.
|
|
|
|
|
*
|
|
|
|
|
* - Redistributions in binary form must reproduce the above
|
|
|
|
|
* copyright notice, this list of conditions and the following
|
|
|
|
|
* disclaimer in the documentation and/or other materials
|
|
|
|
|
* provided with the distribution.
|
|
|
|
|
*
|
|
|
|
|
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
|
|
|
|
|
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
|
|
|
|
|
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
|
|
|
|
|
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
|
|
|
|
|
* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
|
|
|
|
|
* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
|
|
|
|
|
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
|
|
|
|
* SOFTWARE.
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
#include <linux/module.h>
|
|
|
|
|
#include <linux/init.h>
|
|
|
|
|
#include <linux/errno.h>
|
|
|
|
|
#include <linux/pci.h>
|
|
|
|
|
#include <linux/dma-mapping.h>
|
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 08:04:11 +00:00
|
|
|
#include <linux/slab.h>
|
2011-03-22 22:38:41 +00:00
|
|
|
#include <linux/io-mapping.h>
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
#include <linux/delay.h>
|
2013-10-15 14:55:24 +00:00
|
|
|
#include <linux/kmod.h>
|
2016-02-26 16:32:24 +00:00
|
|
|
#include <net/devlink.h>
|
2007-05-09 01:00:38 +00:00
|
|
|
|
|
|
|
|
#include <linux/mlx4/device.h>
|
|
|
|
|
#include <linux/mlx4/doorbell.h>
|
|
|
|
|
|
|
|
|
|
#include "mlx4.h"
|
|
|
|
|
#include "fw.h"
|
|
|
|
|
#include "icm.h"
|
|
|
|
|
|
|
|
|
|
MODULE_AUTHOR("Roland Dreier");
|
|
|
|
|
MODULE_DESCRIPTION("Mellanox ConnectX HCA low-level driver");
|
|
|
|
|
MODULE_LICENSE("Dual BSD/GPL");
|
|
|
|
|
MODULE_VERSION(DRV_VERSION);
|
|
|
|
|
|
2009-03-19 02:45:11 +00:00
|
|
|
struct workqueue_struct *mlx4_wq;
|
|
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
#ifdef CONFIG_MLX4_DEBUG
|
|
|
|
|
|
|
|
|
|
int mlx4_debug_level = 0;
|
|
|
|
|
module_param_named(debug_level, mlx4_debug_level, int, 0644);
|
|
|
|
|
MODULE_PARM_DESC(debug_level, "Enable debug tracing if > 0");
|
|
|
|
|
|
|
|
|
|
#endif /* CONFIG_MLX4_DEBUG */
|
|
|
|
|
|
|
|
|
|
#ifdef CONFIG_PCI_MSI
|
|
|
|
|
|
2007-08-07 13:08:28 +00:00
|
|
|
static int msi_x = 1;
|
2007-05-09 01:00:38 +00:00
|
|
|
module_param(msi_x, int, 0444);
|
|
|
|
|
MODULE_PARM_DESC(msi_x, "attempt to use MSI-X if nonzero");
|
|
|
|
|
|
|
|
|
|
#else /* CONFIG_PCI_MSI */
|
|
|
|
|
|
|
|
|
|
#define msi_x (0)
|
|
|
|
|
|
|
|
|
|
#endif /* CONFIG_PCI_MSI */
|
|
|
|
|
|
2014-03-19 16:11:53 +00:00
|
|
|
static uint8_t num_vfs[3] = {0, 0, 0};
|
2014-09-23 13:05:59 +00:00
|
|
|
static int num_vfs_argc;
|
2014-03-19 16:11:53 +00:00
|
|
|
module_param_array(num_vfs, byte , &num_vfs_argc, 0444);
|
|
|
|
|
MODULE_PARM_DESC(num_vfs, "enable #num_vfs functions if num_vfs > 0\n"
|
|
|
|
|
"num_vfs=port1,port2,port1+2");
|
|
|
|
|
|
|
|
|
|
static uint8_t probe_vf[3] = {0, 0, 0};
|
2014-09-23 13:05:59 +00:00
|
|
|
static int probe_vfs_argc;
|
2014-03-19 16:11:53 +00:00
|
|
|
module_param_array(probe_vf, byte, &probe_vfs_argc, 0444);
|
|
|
|
|
MODULE_PARM_DESC(probe_vf, "number of vfs to probe by pf driver (num_vfs > 0)\n"
|
|
|
|
|
"probe_vf=port1,port2,port1+2");
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
|
2012-12-06 17:12:00 +00:00
|
|
|
int mlx4_log_num_mgm_entry_size = MLX4_DEFAULT_MGM_LOG_ENTRY_SIZE;
|
2011-12-13 04:16:02 +00:00
|
|
|
module_param_named(log_num_mgm_entry_size,
|
|
|
|
|
mlx4_log_num_mgm_entry_size, int, 0444);
|
|
|
|
|
MODULE_PARM_DESC(log_num_mgm_entry_size, "log mgm size, that defines the num"
|
|
|
|
|
" of qp per mcg, for example:"
|
2012-12-06 17:12:00 +00:00
|
|
|
" 10 gives 248.range: 7 <="
|
{NET, IB}/mlx4: Add device managed flow steering firmware API
The driver is modified to support three operation modes.
If supported by firmware use the device managed flow steering
API, that which we call device managed steering mode. Else, if
the firmware supports the B0 steering mode use it, and finally,
if none of the above, use the A0 steering mode.
When the steering mode is device managed, the code is modified
such that L2 based rules set by the mlx4_en driver for Ethernet
unicast and multicast, and the IB stack multicast attach calls
done through the mlx4_ib driver are all routed to use the device
managed API.
When attaching rule using device managed flow steering API,
the firmware returns a 64 bit registration id, which is to be
provided during detach.
Currently the firmware is always programmed during HCA initialization
to use standard L2 hashing. Future work should be done to allow
configuring the flow-steering hash function with common, non
proprietary means.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-07-05 04:03:46 +00:00
|
|
|
" log_num_mgm_entry_size <= 12."
|
2012-12-06 17:12:00 +00:00
|
|
|
" To activate device managed"
|
|
|
|
|
" flow steering when available, set to -1");
|
2011-12-13 04:16:02 +00:00
|
|
|
|
2013-12-19 19:20:15 +00:00
|
|
|
static bool enable_64b_cqe_eqe = true;
|
2012-10-21 14:59:24 +00:00
|
|
|
module_param(enable_64b_cqe_eqe, bool, 0444);
|
|
|
|
|
MODULE_PARM_DESC(enable_64b_cqe_eqe,
|
2013-12-19 19:20:15 +00:00
|
|
|
"Enable 64 byte CQEs/EQEs when the FW supports this (default: True)");
|
2012-10-21 14:59:24 +00:00
|
|
|
|
2016-03-17 16:49:42 +00:00
|
|
|
static bool enable_4k_uar;
|
|
|
|
|
module_param(enable_4k_uar, bool, 0444);
|
|
|
|
|
MODULE_PARM_DESC(enable_4k_uar,
|
|
|
|
|
"Enable using 4K UAR. Should not be enabled if have VFs which do not support 4K UARs (default: false)");
|
|
|
|
|
|
2014-09-18 08:50:59 +00:00
|
|
|
#define PF_CONTEXT_BEHAVIOUR_MASK (MLX4_FUNC_CAP_64B_EQE_CQE | \
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 08:58:00 +00:00
|
|
|
MLX4_FUNC_CAP_EQE_CQE_STRIDE | \
|
|
|
|
|
MLX4_FUNC_CAP_DMFS_A0_STATIC)
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
|
2015-01-25 14:59:42 +00:00
|
|
|
#define RESET_PERSIST_MASK_FLAGS (MLX4_FLAG_SRIOV)
|
|
|
|
|
|
2012-12-03 14:23:15 +00:00
|
|
|
static char mlx4_version[] =
|
2007-05-09 01:00:38 +00:00
|
|
|
DRV_NAME ": Mellanox ConnectX core driver v"
|
|
|
|
|
DRV_VERSION " (" DRV_RELDATE ")\n";
|
|
|
|
|
|
|
|
|
|
static struct mlx4_profile default_profile = {
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
.num_qp = 1 << 18,
|
2007-05-09 01:00:38 +00:00
|
|
|
.num_srq = 1 << 16,
|
2007-07-17 10:11:43 +00:00
|
|
|
.rdmarc_per_qp = 1 << 4,
|
2007-05-09 01:00:38 +00:00
|
|
|
.num_cq = 1 << 16,
|
|
|
|
|
.num_mcg = 1 << 13,
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
.num_mpt = 1 << 19,
|
2012-01-19 09:45:31 +00:00
|
|
|
.num_mtt = 1 << 20, /* It is really num mtt segements */
|
2007-05-09 01:00:38 +00:00
|
|
|
};
|
|
|
|
|
|
2014-07-22 12:44:11 +00:00
|
|
|
static struct mlx4_profile low_mem_profile = {
|
|
|
|
|
.num_qp = 1 << 17,
|
|
|
|
|
.num_srq = 1 << 6,
|
|
|
|
|
.rdmarc_per_qp = 1 << 4,
|
|
|
|
|
.num_cq = 1 << 8,
|
|
|
|
|
.num_mcg = 1 << 8,
|
|
|
|
|
.num_mpt = 1 << 9,
|
|
|
|
|
.num_mtt = 1 << 7,
|
|
|
|
|
};
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
static int log_num_mac = 7;
|
2008-10-22 17:25:29 +00:00
|
|
|
module_param_named(log_num_mac, log_num_mac, int, 0444);
|
|
|
|
|
MODULE_PARM_DESC(log_num_mac, "Log2 max number of MACs per ETH port (1-7)");
|
|
|
|
|
|
|
|
|
|
static int log_num_vlan;
|
|
|
|
|
module_param_named(log_num_vlan, log_num_vlan, int, 0444);
|
|
|
|
|
MODULE_PARM_DESC(log_num_vlan, "Log2 max number of VLANs per ETH port (0-7)");
|
2011-10-16 08:26:21 +00:00
|
|
|
/* Log2 max number of VLANs per ETH port (0-7) */
|
|
|
|
|
#define MLX4_LOG_NUM_VLANS 7
|
2014-07-22 12:44:11 +00:00
|
|
|
#define MLX4_MIN_LOG_NUM_VLANS 0
|
|
|
|
|
#define MLX4_MIN_LOG_NUM_MAC 1
|
2008-10-22 17:25:29 +00:00
|
|
|
|
2011-12-19 14:08:01 +00:00
|
|
|
static bool use_prio;
|
2008-10-22 17:25:29 +00:00
|
|
|
module_param_named(use_prio, use_prio, bool, 0444);
|
2014-05-22 12:55:39 +00:00
|
|
|
MODULE_PARM_DESC(use_prio, "Enable steering by VLAN priority on ETH ports (deprecated)");
|
2008-10-22 17:25:29 +00:00
|
|
|
|
2011-12-13 04:16:56 +00:00
|
|
|
int log_mtts_per_seg = ilog2(MLX4_MTT_ENTRY_PER_SEG);
|
2009-05-27 21:38:34 +00:00
|
|
|
module_param_named(log_mtts_per_seg, log_mtts_per_seg, int, 0444);
|
2010-09-20 06:42:38 +00:00
|
|
|
MODULE_PARM_DESC(log_mtts_per_seg, "Log2 number of MTT entries per segment (1-7)");
|
2009-05-27 21:38:34 +00:00
|
|
|
|
2011-12-19 04:00:34 +00:00
|
|
|
static int port_type_array[2] = {MLX4_PORT_TYPE_NONE, MLX4_PORT_TYPE_NONE};
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
static int arr_argc = 2;
|
|
|
|
|
module_param_array(port_type_array, int, &arr_argc, 0444);
|
2011-12-19 04:00:34 +00:00
|
|
|
MODULE_PARM_DESC(port_type_array, "Array of port types: HW_DEFAULT (0) is default "
|
|
|
|
|
"1 for IB, 2 for Ethernet");
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
|
|
|
|
|
struct mlx4_port_config {
|
|
|
|
|
struct list_head list;
|
|
|
|
|
enum mlx4_port_type port_type[MLX4_MAX_PORTS + 1];
|
|
|
|
|
struct pci_dev *pdev;
|
|
|
|
|
};
|
|
|
|
|
|
net/mlx4_core: mlx4_init_slave() shouldn't access comm channel before PF is ready
Currently, the PF call to pci_enable_sriov from the PF probe function
stalls for 10 seconds times the number of VFs probed on the host. This
happens because the way for such VFs to determine of the PF
initialization finished, is by attempting to issue reset on the
comm-channel and get timeout (after 10s).
The PF probe function is called from a kenernel workqueue, and therefore
during that time, rcu lock is being held and kernel's workqueue is
stalled. This blocks other processes that try to use the workqueue
or rcu lock. For example, interface renaming which is calling
rcu_synchronize is blocked, and timedout by systemd.
Changed mlx4_init_slave() to allow VF probed on the host to immediatly
detect that the PF is not ready, and return EPROBE_DEFER instantly.
Only when the PF finishes the initialization, allow such VFs to
access the comm channel.
This issue and fix are relevant only for probed VFs on the hypervisor,
there is no way to pass this information to a VM until comm channel is
ready, so in a VM, if PF is not ready, the first command will be timedout
after 10 seconds and return EPROBE_DEFER.
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-03-06 16:28:17 +00:00
|
|
|
static atomic_t pf_loading = ATOMIC_INIT(0);
|
|
|
|
|
|
net/mlx4_core: Set UAR page size to 4KB regardless of system page size
problem description:
The current code sets UAR page size equal to system page size.
The ConnectX-3 and ConnectX-3 Pro HWs require minimum 128 UAR pages.
The mlx4 kernel drivers are not loaded if there is less than 128 UAR pages.
solution:
Always set UAR page to 4KB. This allows more UAR pages if the OS
has PAGE_SIZE larger than 4KB. For example, PowerPC kernel use 64KB
system page size, with 4MB uar region, there are 4MB/2/64KB = 32
uars (half for uar, half for blueflame). This does not meet minimum 128
UAR pages requirement. With 4KB UAR page, there are 4MB/2/4KB = 512 uars
which meet the minimum requirement.
Note that only codes in mlx4_core that deal with firmware know that uar
page size is 4KB. Codes that deal with usr page in cq and qp context
(mlx4_ib, mlx4_en and part of mlx4_core) still have the same assumption
that uar page size equals to system page size.
Note that with this implementation, on 64KB system page size kernel, there
are 16 uars per system page but only one uars is used. The other 15
uars are ignored because of the above assumption.
Regarding SR-IOV, mlx4_core in hypervisor will set the uar page size
to 4KB and mlx4_core code in virtual OS will obtain the uar page size from
firmware.
Regarding backward compatibility in SR-IOV, if hypervisor has this new code,
the virtual OS must be updated. If hypervisor has old code, and the virtual
OS has this new code, the new code will be backward compatible with the
old code. If the uar size is big enough, this new code in VF continues to
work with 64 KB uar page size (on PowerPc kernel). If the uar size does not
meet 128 uars requirement, this new code not loaded in VF and print the same
error message as the old code in Hypervisor.
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-17 15:24:26 +00:00
|
|
|
static inline void mlx4_set_num_reserved_uars(struct mlx4_dev *dev,
|
|
|
|
|
struct mlx4_dev_cap *dev_cap)
|
|
|
|
|
{
|
|
|
|
|
/* The reserved_uars is calculated by system page size unit.
|
|
|
|
|
* Therefore, adjustment is added when the uar page size is less
|
|
|
|
|
* than the system page size
|
|
|
|
|
*/
|
|
|
|
|
dev->caps.reserved_uars =
|
|
|
|
|
max_t(int,
|
|
|
|
|
mlx4_get_num_reserved_uar(dev),
|
|
|
|
|
dev_cap->reserved_uars /
|
|
|
|
|
(1 << (PAGE_SHIFT - dev->uar_page_shift)));
|
|
|
|
|
}
|
|
|
|
|
|
2009-03-19 02:45:11 +00:00
|
|
|
int mlx4_check_port_params(struct mlx4_dev *dev,
|
|
|
|
|
enum mlx4_port_type *port_type)
|
2008-10-22 22:38:42 +00:00
|
|
|
{
|
|
|
|
|
int i;
|
|
|
|
|
|
2014-12-13 18:18:40 +00:00
|
|
|
if (!(dev->caps.flags & MLX4_DEV_CAP_FLAG_DPDP)) {
|
|
|
|
|
for (i = 0; i < dev->caps.num_ports - 1; i++) {
|
|
|
|
|
if (port_type[i] != port_type[i + 1]) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Only same port types supported on this HCA, aborting\n");
|
2009-03-19 02:45:11 +00:00
|
|
|
return -EINVAL;
|
|
|
|
|
}
|
2008-10-22 22:38:42 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < dev->caps.num_ports; i++) {
|
|
|
|
|
if (!(port_type[i] & dev->caps.supported_type[i+1])) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Requested port type for port %d is not supported on this HCA\n",
|
|
|
|
|
i + 1);
|
2008-10-22 22:38:42 +00:00
|
|
|
return -EINVAL;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void mlx4_set_port_mask(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
int i;
|
|
|
|
|
|
|
|
|
|
for (i = 1; i <= dev->caps.num_ports; ++i)
|
2011-12-13 04:10:41 +00:00
|
|
|
dev->caps.port_mask[i] = dev->caps.port_type[i];
|
2008-10-22 22:38:42 +00:00
|
|
|
}
|
2011-06-15 14:47:14 +00:00
|
|
|
|
2014-11-13 12:45:32 +00:00
|
|
|
enum {
|
|
|
|
|
MLX4_QUERY_FUNC_NUM_SYS_EQS = 1 << 0,
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
static int mlx4_query_func(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
|
|
|
|
|
{
|
|
|
|
|
int err = 0;
|
|
|
|
|
struct mlx4_func func;
|
|
|
|
|
|
|
|
|
|
if (dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_SYS_EQS) {
|
|
|
|
|
err = mlx4_QUERY_FUNC(dev, &func, 0);
|
|
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "QUERY_DEV_CAP command failed, aborting.\n");
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
dev_cap->max_eqs = func.max_eq;
|
|
|
|
|
dev_cap->reserved_eqs = func.rsvd_eqs;
|
|
|
|
|
dev_cap->reserved_uars = func.rsvd_uars;
|
|
|
|
|
err |= MLX4_QUERY_FUNC_NUM_SYS_EQS;
|
|
|
|
|
}
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2014-09-18 08:50:59 +00:00
|
|
|
static void mlx4_enable_cqe_eqe_stride(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_caps *dev_cap = &dev->caps;
|
|
|
|
|
|
|
|
|
|
/* FW not supporting or cancelled by user */
|
|
|
|
|
if (!(dev_cap->flags2 & MLX4_DEV_CAP_FLAG2_EQE_STRIDE) ||
|
|
|
|
|
!(dev_cap->flags2 & MLX4_DEV_CAP_FLAG2_CQE_STRIDE))
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
/* Must have 64B CQE_EQE enabled by FW to use bigger stride
|
|
|
|
|
* When FW has NCSI it may decide not to report 64B CQE/EQEs
|
|
|
|
|
*/
|
|
|
|
|
if (!(dev_cap->flags & MLX4_DEV_CAP_FLAG_64B_EQE) ||
|
|
|
|
|
!(dev_cap->flags & MLX4_DEV_CAP_FLAG_64B_CQE)) {
|
|
|
|
|
dev_cap->flags2 &= ~MLX4_DEV_CAP_FLAG2_CQE_STRIDE;
|
|
|
|
|
dev_cap->flags2 &= ~MLX4_DEV_CAP_FLAG2_EQE_STRIDE;
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (cache_line_size() == 128 || cache_line_size() == 256) {
|
|
|
|
|
mlx4_dbg(dev, "Enabling CQE stride cacheLine supported\n");
|
|
|
|
|
/* Changing the real data inside CQE size to 32B */
|
|
|
|
|
dev_cap->flags &= ~MLX4_DEV_CAP_FLAG_64B_CQE;
|
|
|
|
|
dev_cap->flags &= ~MLX4_DEV_CAP_FLAG_64B_EQE;
|
|
|
|
|
|
|
|
|
|
if (mlx4_is_master(dev))
|
|
|
|
|
dev_cap->function_caps |= MLX4_FUNC_CAP_EQE_CQE_STRIDE;
|
|
|
|
|
} else {
|
2015-02-03 15:57:17 +00:00
|
|
|
if (cache_line_size() != 32 && cache_line_size() != 64)
|
|
|
|
|
mlx4_dbg(dev, "Disabling CQE stride, cacheLine size unsupported\n");
|
2014-09-18 08:50:59 +00:00
|
|
|
dev_cap->flags2 &= ~MLX4_DEV_CAP_FLAG2_CQE_STRIDE;
|
|
|
|
|
dev_cap->flags2 &= ~MLX4_DEV_CAP_FLAG2_EQE_STRIDE;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2014-12-11 08:57:59 +00:00
|
|
|
static int _mlx4_dev_port(struct mlx4_dev *dev, int port,
|
|
|
|
|
struct mlx4_port_cap *port_cap)
|
|
|
|
|
{
|
|
|
|
|
dev->caps.vl_cap[port] = port_cap->max_vl;
|
|
|
|
|
dev->caps.ib_mtu_cap[port] = port_cap->ib_mtu;
|
|
|
|
|
dev->phys_caps.gid_phys_table_len[port] = port_cap->max_gids;
|
|
|
|
|
dev->phys_caps.pkey_phys_table_len[port] = port_cap->max_pkeys;
|
|
|
|
|
/* set gid and pkey table operating lengths by default
|
|
|
|
|
* to non-sriov values
|
|
|
|
|
*/
|
|
|
|
|
dev->caps.gid_table_len[port] = port_cap->max_gids;
|
|
|
|
|
dev->caps.pkey_table_len[port] = port_cap->max_pkeys;
|
|
|
|
|
dev->caps.port_width_cap[port] = port_cap->max_port_width;
|
|
|
|
|
dev->caps.eth_mtu_cap[port] = port_cap->eth_mtu;
|
2016-06-21 09:43:59 +00:00
|
|
|
dev->caps.max_tc_eth = port_cap->max_tc_eth;
|
2014-12-11 08:57:59 +00:00
|
|
|
dev->caps.def_mac[port] = port_cap->def_mac;
|
|
|
|
|
dev->caps.supported_type[port] = port_cap->supported_port_types;
|
|
|
|
|
dev->caps.suggested_type[port] = port_cap->suggested_type;
|
|
|
|
|
dev->caps.default_sense[port] = port_cap->default_sense;
|
|
|
|
|
dev->caps.trans_type[port] = port_cap->trans_type;
|
|
|
|
|
dev->caps.vendor_oui[port] = port_cap->vendor_oui;
|
|
|
|
|
dev->caps.wavelength[port] = port_cap->wavelength;
|
|
|
|
|
dev->caps.trans_code[port] = port_cap->trans_code;
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int mlx4_dev_port(struct mlx4_dev *dev, int port,
|
|
|
|
|
struct mlx4_port_cap *port_cap)
|
|
|
|
|
{
|
|
|
|
|
int err = 0;
|
|
|
|
|
|
|
|
|
|
err = mlx4_QUERY_PORT(dev, port, port_cap);
|
|
|
|
|
|
|
|
|
|
if (err)
|
|
|
|
|
mlx4_err(dev, "QUERY_PORT command failed.\n");
|
|
|
|
|
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2015-04-02 13:31:22 +00:00
|
|
|
static inline void mlx4_enable_ignore_fcs(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
if (!(dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_IGNORE_FCS))
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
if (mlx4_is_mfunc(dev)) {
|
|
|
|
|
mlx4_dbg(dev, "SRIOV mode - Disabling Ignore FCS");
|
|
|
|
|
dev->caps.flags2 &= ~MLX4_DEV_CAP_FLAG2_IGNORE_FCS;
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (!(dev->caps.flags & MLX4_DEV_CAP_FLAG_FCS_KEEP)) {
|
|
|
|
|
mlx4_dbg(dev,
|
|
|
|
|
"Keep FCS is not supported - Disabling Ignore FCS");
|
|
|
|
|
dev->caps.flags2 &= ~MLX4_DEV_CAP_FLAG2_IGNORE_FCS;
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2014-12-11 08:57:59 +00:00
|
|
|
#define MLX4_A0_STEERING_TABLE_SIZE 256
|
2007-10-10 22:43:54 +00:00
|
|
|
static int mlx4_dev_cap(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
|
2007-05-09 01:00:38 +00:00
|
|
|
{
|
|
|
|
|
int err;
|
2007-06-18 15:15:02 +00:00
|
|
|
int i;
|
2007-05-09 01:00:38 +00:00
|
|
|
|
|
|
|
|
err = mlx4_QUERY_DEV_CAP(dev, dev_cap);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "QUERY_DEV_CAP command failed, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
return err;
|
|
|
|
|
}
|
2014-12-14 14:18:05 +00:00
|
|
|
mlx4_dev_cap_dump(dev, dev_cap);
|
2007-05-09 01:00:38 +00:00
|
|
|
|
|
|
|
|
if (dev_cap->min_page_sz > PAGE_SIZE) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "HCA minimum page size of %d bigger than kernel PAGE_SIZE of %ld, aborting\n",
|
2007-05-09 01:00:38 +00:00
|
|
|
dev_cap->min_page_sz, PAGE_SIZE);
|
|
|
|
|
return -ENODEV;
|
|
|
|
|
}
|
|
|
|
|
if (dev_cap->num_ports > MLX4_MAX_PORTS) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "HCA has %d ports, but we only support %d, aborting\n",
|
2007-05-09 01:00:38 +00:00
|
|
|
dev_cap->num_ports, MLX4_MAX_PORTS);
|
|
|
|
|
return -ENODEV;
|
|
|
|
|
}
|
|
|
|
|
|
2015-01-25 14:59:35 +00:00
|
|
|
if (dev_cap->uar_size > pci_resource_len(dev->persist->pdev, 2)) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "HCA reported UAR size of 0x%x bigger than PCI resource 2 size of 0x%llx, aborting\n",
|
2007-05-09 01:00:38 +00:00
|
|
|
dev_cap->uar_size,
|
2015-01-25 14:59:35 +00:00
|
|
|
(unsigned long long)
|
|
|
|
|
pci_resource_len(dev->persist->pdev, 2));
|
2007-05-09 01:00:38 +00:00
|
|
|
return -ENODEV;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
dev->caps.num_ports = dev_cap->num_ports;
|
2014-11-13 12:45:32 +00:00
|
|
|
dev->caps.num_sys_eqs = dev_cap->num_sys_eqs;
|
|
|
|
|
dev->phys_caps.num_phys_eqs = dev_cap->flags2 & MLX4_DEV_CAP_FLAG2_SYS_EQS ?
|
|
|
|
|
dev->caps.num_sys_eqs :
|
|
|
|
|
MLX4_MAX_EQ_NUM;
|
2007-06-18 15:15:02 +00:00
|
|
|
for (i = 1; i <= dev->caps.num_ports; ++i) {
|
2014-12-11 08:57:59 +00:00
|
|
|
err = _mlx4_dev_port(dev, i, dev_cap->port_cap + i);
|
|
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "QUERY_PORT command failed, aborting\n");
|
|
|
|
|
return err;
|
|
|
|
|
}
|
2007-06-18 15:15:02 +00:00
|
|
|
}
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
dev->caps.uar_page_size = PAGE_SIZE;
|
2007-05-09 01:00:38 +00:00
|
|
|
dev->caps.num_uars = dev_cap->uar_size / PAGE_SIZE;
|
|
|
|
|
dev->caps.local_ca_ack_delay = dev_cap->local_ca_ack_delay;
|
|
|
|
|
dev->caps.bf_reg_size = dev_cap->bf_reg_size;
|
|
|
|
|
dev->caps.bf_regs_per_page = dev_cap->bf_regs_per_page;
|
|
|
|
|
dev->caps.max_sq_sg = dev_cap->max_sq_sg;
|
|
|
|
|
dev->caps.max_rq_sg = dev_cap->max_rq_sg;
|
|
|
|
|
dev->caps.max_wqes = dev_cap->max_qp_sz;
|
|
|
|
|
dev->caps.max_qp_init_rdma = dev_cap->max_requester_per_qp;
|
|
|
|
|
dev->caps.max_srq_wqes = dev_cap->max_srq_sz;
|
|
|
|
|
dev->caps.max_srq_sge = dev_cap->max_rq_sg - 1;
|
|
|
|
|
dev->caps.reserved_srqs = dev_cap->reserved_srqs;
|
|
|
|
|
dev->caps.max_sq_desc_sz = dev_cap->max_sq_desc_sz;
|
|
|
|
|
dev->caps.max_rq_desc_sz = dev_cap->max_rq_desc_sz;
|
|
|
|
|
/*
|
|
|
|
|
* Subtract 1 from the limit because we need to allocate a
|
|
|
|
|
* spare CQE so the HCA HW can tell the difference between an
|
|
|
|
|
* empty CQ and a full CQ.
|
|
|
|
|
*/
|
|
|
|
|
dev->caps.max_cqes = dev_cap->max_cq_sz - 1;
|
|
|
|
|
dev->caps.reserved_cqs = dev_cap->reserved_cqs;
|
|
|
|
|
dev->caps.reserved_eqs = dev_cap->reserved_eqs;
|
2011-12-13 04:16:56 +00:00
|
|
|
dev->caps.reserved_mtts = dev_cap->reserved_mtts;
|
2007-05-09 01:00:38 +00:00
|
|
|
dev->caps.reserved_mrws = dev_cap->reserved_mrws;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
dev->caps.reserved_pds = dev_cap->reserved_pds;
|
2011-06-02 16:01:33 +00:00
|
|
|
dev->caps.reserved_xrcds = (dev->caps.flags & MLX4_DEV_CAP_FLAG_XRC) ?
|
|
|
|
|
dev_cap->reserved_xrcds : 0;
|
|
|
|
|
dev->caps.max_xrcds = (dev->caps.flags & MLX4_DEV_CAP_FLAG_XRC) ?
|
|
|
|
|
dev_cap->max_xrcds : 0;
|
2011-12-13 04:16:56 +00:00
|
|
|
dev->caps.mtt_entry_sz = dev_cap->mtt_entry_sz;
|
|
|
|
|
|
2007-06-26 12:55:28 +00:00
|
|
|
dev->caps.max_msg_sz = dev_cap->max_msg_sz;
|
2007-05-09 01:00:38 +00:00
|
|
|
dev->caps.page_size_cap = ~(u32) (dev_cap->min_page_sz - 1);
|
|
|
|
|
dev->caps.flags = dev_cap->flags;
|
2012-04-29 14:04:25 +00:00
|
|
|
dev->caps.flags2 = dev_cap->flags2;
|
2008-07-23 15:12:26 +00:00
|
|
|
dev->caps.bmme_flags = dev_cap->bmme_flags;
|
|
|
|
|
dev->caps.reserved_lkey = dev_cap->reserved_lkey;
|
2007-05-09 01:00:38 +00:00
|
|
|
dev->caps.stat_rate_support = dev_cap->stat_rate_support;
|
2008-04-17 04:09:27 +00:00
|
|
|
dev->caps.max_gso_sz = dev_cap->max_gso_sz;
|
2012-04-29 14:04:25 +00:00
|
|
|
dev->caps.max_rss_tbl_sz = dev_cap->max_rss_tbl_sz;
|
2007-05-09 01:00:38 +00:00
|
|
|
|
net/mlx4_core: Set UAR page size to 4KB regardless of system page size
problem description:
The current code sets UAR page size equal to system page size.
The ConnectX-3 and ConnectX-3 Pro HWs require minimum 128 UAR pages.
The mlx4 kernel drivers are not loaded if there is less than 128 UAR pages.
solution:
Always set UAR page to 4KB. This allows more UAR pages if the OS
has PAGE_SIZE larger than 4KB. For example, PowerPC kernel use 64KB
system page size, with 4MB uar region, there are 4MB/2/64KB = 32
uars (half for uar, half for blueflame). This does not meet minimum 128
UAR pages requirement. With 4KB UAR page, there are 4MB/2/4KB = 512 uars
which meet the minimum requirement.
Note that only codes in mlx4_core that deal with firmware know that uar
page size is 4KB. Codes that deal with usr page in cq and qp context
(mlx4_ib, mlx4_en and part of mlx4_core) still have the same assumption
that uar page size equals to system page size.
Note that with this implementation, on 64KB system page size kernel, there
are 16 uars per system page but only one uars is used. The other 15
uars are ignored because of the above assumption.
Regarding SR-IOV, mlx4_core in hypervisor will set the uar page size
to 4KB and mlx4_core code in virtual OS will obtain the uar page size from
firmware.
Regarding backward compatibility in SR-IOV, if hypervisor has this new code,
the virtual OS must be updated. If hypervisor has old code, and the virtual
OS has this new code, the new code will be backward compatible with the
old code. If the uar size is big enough, this new code in VF continues to
work with 64 KB uar page size (on PowerPc kernel). If the uar size does not
meet 128 uars requirement, this new code not loaded in VF and print the same
error message as the old code in Hypervisor.
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-17 15:24:26 +00:00
|
|
|
/* Save uar page shift */
|
|
|
|
|
if (!mlx4_is_slave(dev)) {
|
|
|
|
|
/* Virtual PCI function needs to determine UAR page size from
|
|
|
|
|
* firmware. Only master PCI function can set the uar page size
|
|
|
|
|
*/
|
2016-03-17 16:49:42 +00:00
|
|
|
if (enable_4k_uar)
|
|
|
|
|
dev->uar_page_shift = DEFAULT_UAR_PAGE_SHIFT;
|
|
|
|
|
else
|
|
|
|
|
dev->uar_page_shift = PAGE_SHIFT;
|
|
|
|
|
|
net/mlx4_core: Set UAR page size to 4KB regardless of system page size
problem description:
The current code sets UAR page size equal to system page size.
The ConnectX-3 and ConnectX-3 Pro HWs require minimum 128 UAR pages.
The mlx4 kernel drivers are not loaded if there is less than 128 UAR pages.
solution:
Always set UAR page to 4KB. This allows more UAR pages if the OS
has PAGE_SIZE larger than 4KB. For example, PowerPC kernel use 64KB
system page size, with 4MB uar region, there are 4MB/2/64KB = 32
uars (half for uar, half for blueflame). This does not meet minimum 128
UAR pages requirement. With 4KB UAR page, there are 4MB/2/4KB = 512 uars
which meet the minimum requirement.
Note that only codes in mlx4_core that deal with firmware know that uar
page size is 4KB. Codes that deal with usr page in cq and qp context
(mlx4_ib, mlx4_en and part of mlx4_core) still have the same assumption
that uar page size equals to system page size.
Note that with this implementation, on 64KB system page size kernel, there
are 16 uars per system page but only one uars is used. The other 15
uars are ignored because of the above assumption.
Regarding SR-IOV, mlx4_core in hypervisor will set the uar page size
to 4KB and mlx4_core code in virtual OS will obtain the uar page size from
firmware.
Regarding backward compatibility in SR-IOV, if hypervisor has this new code,
the virtual OS must be updated. If hypervisor has old code, and the virtual
OS has this new code, the new code will be backward compatible with the
old code. If the uar size is big enough, this new code in VF continues to
work with 64 KB uar page size (on PowerPc kernel). If the uar size does not
meet 128 uars requirement, this new code not loaded in VF and print the same
error message as the old code in Hypervisor.
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-17 15:24:26 +00:00
|
|
|
mlx4_set_num_reserved_uars(dev, dev_cap);
|
|
|
|
|
}
|
|
|
|
|
|
2015-07-27 11:46:31 +00:00
|
|
|
if (dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_PHV_EN) {
|
|
|
|
|
struct mlx4_init_hca_param hca_param;
|
|
|
|
|
|
|
|
|
|
memset(&hca_param, 0, sizeof(hca_param));
|
|
|
|
|
err = mlx4_QUERY_HCA(dev, &hca_param);
|
|
|
|
|
/* Turn off PHV_EN flag in case phv_check_en is set.
|
|
|
|
|
* phv_check_en is a HW check that parse the packet and verify
|
|
|
|
|
* phv bit was reported correctly in the wqe. To allow QinQ
|
|
|
|
|
* PHV_EN flag should be set and phv_check_en must be cleared
|
|
|
|
|
* otherwise QinQ packets will be drop by the HW.
|
|
|
|
|
*/
|
|
|
|
|
if (err || hca_param.phv_check_en)
|
|
|
|
|
dev->caps.flags2 &= ~MLX4_DEV_CAP_FLAG2_PHV_EN;
|
|
|
|
|
}
|
|
|
|
|
|
2012-09-27 16:53:05 +00:00
|
|
|
/* Sense port always allowed on supported devices for ConnectX-1 and -2 */
|
|
|
|
|
if (mlx4_priv(dev)->pci_dev_data & MLX4_PCI_DEV_FORCE_SENSE_PORT)
|
2011-12-19 04:00:26 +00:00
|
|
|
dev->caps.flags |= MLX4_DEV_CAP_FLAG_SENSE_SUPPORT;
|
2012-09-27 17:01:19 +00:00
|
|
|
/* Don't do sense port on multifunction devices (for now at least) */
|
|
|
|
|
if (mlx4_is_mfunc(dev))
|
|
|
|
|
dev->caps.flags &= ~MLX4_DEV_CAP_FLAG_SENSE_SUPPORT;
|
2011-12-19 04:00:26 +00:00
|
|
|
|
2014-07-22 12:44:11 +00:00
|
|
|
if (mlx4_low_memory_profile()) {
|
|
|
|
|
dev->caps.log_num_macs = MLX4_MIN_LOG_NUM_MAC;
|
|
|
|
|
dev->caps.log_num_vlans = MLX4_MIN_LOG_NUM_VLANS;
|
|
|
|
|
} else {
|
|
|
|
|
dev->caps.log_num_macs = log_num_mac;
|
|
|
|
|
dev->caps.log_num_vlans = MLX4_LOG_NUM_VLANS;
|
|
|
|
|
}
|
2008-10-22 17:25:29 +00:00
|
|
|
|
|
|
|
|
for (i = 1; i <= dev->caps.num_ports; ++i) {
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
dev->caps.port_type[i] = MLX4_PORT_TYPE_NONE;
|
|
|
|
|
if (dev->caps.supported_type[i]) {
|
|
|
|
|
/* if only ETH is supported - assign ETH */
|
|
|
|
|
if (dev->caps.supported_type[i] == MLX4_PORT_TYPE_ETH)
|
|
|
|
|
dev->caps.port_type[i] = MLX4_PORT_TYPE_ETH;
|
2012-06-19 08:21:43 +00:00
|
|
|
/* if only IB is supported, assign IB */
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
else if (dev->caps.supported_type[i] ==
|
2012-06-19 08:21:43 +00:00
|
|
|
MLX4_PORT_TYPE_IB)
|
|
|
|
|
dev->caps.port_type[i] = MLX4_PORT_TYPE_IB;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
else {
|
2012-06-19 08:21:43 +00:00
|
|
|
/* if IB and ETH are supported, we set the port
|
|
|
|
|
* type according to user selection of port type;
|
|
|
|
|
* if user selected none, take the FW hint */
|
|
|
|
|
if (port_type_array[i - 1] == MLX4_PORT_TYPE_NONE)
|
2011-12-19 04:00:34 +00:00
|
|
|
dev->caps.port_type[i] = dev->caps.suggested_type[i] ?
|
|
|
|
|
MLX4_PORT_TYPE_ETH : MLX4_PORT_TYPE_IB;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
else
|
2012-06-19 08:21:43 +00:00
|
|
|
dev->caps.port_type[i] = port_type_array[i - 1];
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
}
|
|
|
|
|
}
|
2011-12-19 04:00:34 +00:00
|
|
|
/*
|
|
|
|
|
* Link sensing is allowed on the port if 3 conditions are true:
|
|
|
|
|
* 1. Both protocols are supported on the port.
|
|
|
|
|
* 2. Different types are supported on the port
|
|
|
|
|
* 3. FW declared that it supports link sensing
|
|
|
|
|
*/
|
2009-03-19 02:45:11 +00:00
|
|
|
mlx4_priv(dev)->sense.sense_allowed[i] =
|
2011-12-19 04:00:26 +00:00
|
|
|
((dev->caps.supported_type[i] == MLX4_PORT_TYPE_AUTO) &&
|
2011-12-19 04:00:34 +00:00
|
|
|
(dev->caps.flags & MLX4_DEV_CAP_FLAG_DPDP) &&
|
2011-12-19 04:00:26 +00:00
|
|
|
(dev->caps.flags & MLX4_DEV_CAP_FLAG_SENSE_SUPPORT));
|
2008-10-22 22:38:42 +00:00
|
|
|
|
2011-12-19 04:00:34 +00:00
|
|
|
/*
|
|
|
|
|
* If "default_sense" bit is set, we move the port to "AUTO" mode
|
|
|
|
|
* and perform sense_port FW command to try and set the correct
|
|
|
|
|
* port type from beginning
|
|
|
|
|
*/
|
2011-12-29 07:42:34 +00:00
|
|
|
if (mlx4_priv(dev)->sense.sense_allowed[i] && dev->caps.default_sense[i]) {
|
2011-12-19 04:00:34 +00:00
|
|
|
enum mlx4_port_type sensed_port = MLX4_PORT_TYPE_NONE;
|
|
|
|
|
dev->caps.possible_type[i] = MLX4_PORT_TYPE_AUTO;
|
|
|
|
|
mlx4_SENSE_PORT(dev, i, &sensed_port);
|
|
|
|
|
if (sensed_port != MLX4_PORT_TYPE_NONE)
|
|
|
|
|
dev->caps.port_type[i] = sensed_port;
|
|
|
|
|
} else {
|
|
|
|
|
dev->caps.possible_type[i] = dev->caps.port_type[i];
|
|
|
|
|
}
|
|
|
|
|
|
2014-12-11 08:57:59 +00:00
|
|
|
if (dev->caps.log_num_macs > dev_cap->port_cap[i].log_max_macs) {
|
|
|
|
|
dev->caps.log_num_macs = dev_cap->port_cap[i].log_max_macs;
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_warn(dev, "Requested number of MACs is too much for port %d, reducing to %d\n",
|
2008-10-22 17:25:29 +00:00
|
|
|
i, 1 << dev->caps.log_num_macs);
|
|
|
|
|
}
|
2014-12-11 08:57:59 +00:00
|
|
|
if (dev->caps.log_num_vlans > dev_cap->port_cap[i].log_max_vlans) {
|
|
|
|
|
dev->caps.log_num_vlans = dev_cap->port_cap[i].log_max_vlans;
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_warn(dev, "Requested number of VLANs is too much for port %d, reducing to %d\n",
|
2008-10-22 17:25:29 +00:00
|
|
|
i, 1 << dev->caps.log_num_vlans);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2015-06-14 14:13:06 +00:00
|
|
|
if (mlx4_is_master(dev) && (dev->caps.num_ports == 2) &&
|
|
|
|
|
(port_type_array[0] == MLX4_PORT_TYPE_IB) &&
|
|
|
|
|
(port_type_array[1] == MLX4_PORT_TYPE_ETH)) {
|
|
|
|
|
mlx4_warn(dev,
|
|
|
|
|
"Granular QoS per VF not supported with IB/Eth configuration\n");
|
|
|
|
|
dev->caps.flags2 &= ~MLX4_DEV_CAP_FLAG2_QOS_VPP;
|
|
|
|
|
}
|
|
|
|
|
|
2015-06-15 14:58:58 +00:00
|
|
|
dev->caps.max_counters = dev_cap->max_counters;
|
2011-06-15 14:47:14 +00:00
|
|
|
|
2008-10-22 17:25:29 +00:00
|
|
|
dev->caps.reserved_qps_cnt[MLX4_QP_REGION_FW] = dev_cap->reserved_qps;
|
|
|
|
|
dev->caps.reserved_qps_cnt[MLX4_QP_REGION_ETH_ADDR] =
|
|
|
|
|
dev->caps.reserved_qps_cnt[MLX4_QP_REGION_FC_ADDR] =
|
|
|
|
|
(1 << dev->caps.log_num_macs) *
|
|
|
|
|
(1 << dev->caps.log_num_vlans) *
|
|
|
|
|
dev->caps.num_ports;
|
|
|
|
|
dev->caps.reserved_qps_cnt[MLX4_QP_REGION_FC_EXCH] = MLX4_NUM_FEXCH;
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 08:58:00 +00:00
|
|
|
|
|
|
|
|
if (dev_cap->dmfs_high_rate_qpn_base > 0 &&
|
|
|
|
|
dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_FS_EN)
|
|
|
|
|
dev->caps.dmfs_high_rate_qpn_base = dev_cap->dmfs_high_rate_qpn_base;
|
|
|
|
|
else
|
|
|
|
|
dev->caps.dmfs_high_rate_qpn_base =
|
|
|
|
|
dev->caps.reserved_qps_cnt[MLX4_QP_REGION_FW];
|
|
|
|
|
|
|
|
|
|
if (dev_cap->dmfs_high_rate_qpn_range > 0 &&
|
|
|
|
|
dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_FS_EN) {
|
|
|
|
|
dev->caps.dmfs_high_rate_qpn_range = dev_cap->dmfs_high_rate_qpn_range;
|
|
|
|
|
dev->caps.dmfs_high_steer_mode = MLX4_STEERING_DMFS_A0_DEFAULT;
|
|
|
|
|
dev->caps.flags2 |= MLX4_DEV_CAP_FLAG2_FS_A0;
|
|
|
|
|
} else {
|
|
|
|
|
dev->caps.dmfs_high_steer_mode = MLX4_STEERING_DMFS_A0_NOT_SUPPORTED;
|
|
|
|
|
dev->caps.dmfs_high_rate_qpn_base =
|
|
|
|
|
dev->caps.reserved_qps_cnt[MLX4_QP_REGION_FW];
|
|
|
|
|
dev->caps.dmfs_high_rate_qpn_range = MLX4_A0_STEERING_TABLE_SIZE;
|
|
|
|
|
}
|
|
|
|
|
|
2015-03-18 12:57:34 +00:00
|
|
|
dev->caps.rl_caps = dev_cap->rl_caps;
|
|
|
|
|
|
net/mlx4: Add A0 hybrid steering
A0 hybrid steering is a form of high performance flow steering.
By using this mode, mlx4 cards use a fast limited table based steering,
in order to enable fast steering of unicast packets to a QP.
In order to implement A0 hybrid steering we allocate resources
from different zones:
(1) General range
(2) Special MAC-assigned QPs [RSS, Raw-Ethernet] each has its own region.
When we create a rss QP or a raw ethernet (A0 steerable and BF ready) QP,
we try hard to allocate the QP from range (2). Otherwise, we try hard not
to allocate from this range. However, when the system is pushed to its
limits and one needs every resource, the allocator uses every region it can.
Meaning, when we run out of raw-eth qps, the allocator allocates from the
general range (and the special-A0 area is no longer active). If we run out
of RSS qps, the mechanism tries to allocate from the raw-eth QP zone. If that
is also exhausted, the allocator will allocate from the general range
(and the A0 region is no longer active).
Note that if a raw-eth qp is allocated from the general range, it attempts
to allocate the range such that bits 6 and 7 (blueflame bits) in the
QP number are not set.
When the feature is used in SRIOV, the VF has to notify the PF what
kind of QP attributes it needs. In order to do that, along with the
"Eth QP blueflame" bit, we reserve a new "A0 steerable QP". According
to the combination of these bits, the PF tries to allocate a suitable QP.
In order to maintain backward compatibility (with older PFs), the PF
notifies which QP attributes it supports via QUERY_FUNC_CAP command.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 08:57:57 +00:00
|
|
|
dev->caps.reserved_qps_cnt[MLX4_QP_REGION_RSS_RAW_ETH] =
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 08:58:00 +00:00
|
|
|
dev->caps.dmfs_high_rate_qpn_range;
|
2008-10-22 17:25:29 +00:00
|
|
|
|
|
|
|
|
dev->caps.reserved_qps = dev->caps.reserved_qps_cnt[MLX4_QP_REGION_FW] +
|
|
|
|
|
dev->caps.reserved_qps_cnt[MLX4_QP_REGION_ETH_ADDR] +
|
|
|
|
|
dev->caps.reserved_qps_cnt[MLX4_QP_REGION_FC_ADDR] +
|
|
|
|
|
dev->caps.reserved_qps_cnt[MLX4_QP_REGION_FC_EXCH];
|
|
|
|
|
|
2012-08-03 08:40:41 +00:00
|
|
|
dev->caps.sqp_demux = (mlx4_is_master(dev)) ? MLX4_MAX_NUM_SLAVES : 0;
|
2012-10-21 14:59:24 +00:00
|
|
|
|
2013-08-01 16:55:01 +00:00
|
|
|
if (!enable_64b_cqe_eqe && !mlx4_is_slave(dev)) {
|
2012-10-21 14:59:24 +00:00
|
|
|
if (dev_cap->flags &
|
|
|
|
|
(MLX4_DEV_CAP_FLAG_64B_CQE | MLX4_DEV_CAP_FLAG_64B_EQE)) {
|
|
|
|
|
mlx4_warn(dev, "64B EQEs/CQEs supported by the device but not enabled\n");
|
|
|
|
|
dev->caps.flags &= ~MLX4_DEV_CAP_FLAG_64B_CQE;
|
|
|
|
|
dev->caps.flags &= ~MLX4_DEV_CAP_FLAG_64B_EQE;
|
|
|
|
|
}
|
2014-09-18 08:50:59 +00:00
|
|
|
|
|
|
|
|
if (dev_cap->flags2 &
|
|
|
|
|
(MLX4_DEV_CAP_FLAG2_CQE_STRIDE |
|
|
|
|
|
MLX4_DEV_CAP_FLAG2_EQE_STRIDE)) {
|
|
|
|
|
mlx4_warn(dev, "Disabling EQE/CQE stride per user request\n");
|
|
|
|
|
dev_cap->flags2 &= ~MLX4_DEV_CAP_FLAG2_CQE_STRIDE;
|
|
|
|
|
dev_cap->flags2 &= ~MLX4_DEV_CAP_FLAG2_EQE_STRIDE;
|
|
|
|
|
}
|
2012-10-21 14:59:24 +00:00
|
|
|
}
|
|
|
|
|
|
2013-01-10 15:18:35 +00:00
|
|
|
if ((dev->caps.flags &
|
2012-10-21 14:59:24 +00:00
|
|
|
(MLX4_DEV_CAP_FLAG_64B_CQE | MLX4_DEV_CAP_FLAG_64B_EQE)) &&
|
|
|
|
|
mlx4_is_master(dev))
|
|
|
|
|
dev->caps.function_caps |= MLX4_FUNC_CAP_64B_EQE_CQE;
|
|
|
|
|
|
net/mlx4: Change QP allocation scheme
When using BF (Blue-Flame), the QPN overrides the VLAN, CV, and SV fields
in the WQE. Thus, BF may only be used for QPNs with bits 6,7 unset.
The current Ethernet driver code reserves a Tx QP range with 256b alignment.
This is wrong because if there are more than 64 Tx QPs in use,
QPNs >= base + 65 will have bits 6/7 set.
This problem is not specific for the Ethernet driver, any entity that
tries to reserve more than 64 BF-enabled QPs should fail. Also, using
ranges is not necessary here and is wasteful.
The new mechanism introduced here will support reservation for
"Eth QPs eligible for BF" for all drivers: bare-metal, multi-PF, and VFs
(when hypervisors support WC in VMs). The flow we use is:
1. In mlx4_en, allocate Tx QPs one by one instead of a range allocation,
and request "BF enabled QPs" if BF is supported for the function
2. In the ALLOC_RES FW command, change param1 to:
a. param1[23:0] - number of QPs
b. param1[31-24] - flags controlling QPs reservation
Bit 31 refers to Eth blueflame supported QPs. Those QPs must have
bits 6 and 7 unset in order to be used in Ethernet.
Bits 24-30 of the flags are currently reserved.
When a function tries to allocate a QP, it states the required attributes
for this QP. Those attributes are considered "best-effort". If an attribute,
such as Ethernet BF enabled QP, is a must-have attribute, the function has
to check that attribute is supported before trying to do the allocation.
In a lower layer of the code, mlx4_qp_reserve_range masks out the bits
which are unsupported. If SRIOV is used, the PF validates those attributes
and masks out unsupported attributes as well. In order to notify VFs which
attributes are supported, the VF uses QUERY_FUNC_CAP command. This command's
mailbox is filled by the PF, which notifies which QP allocation attributes
it supports.
Signed-off-by: Eugenia Emantayev <eugenia@mellanox.co.il>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 08:57:54 +00:00
|
|
|
if (!mlx4_is_slave(dev)) {
|
2014-09-18 08:50:59 +00:00
|
|
|
mlx4_enable_cqe_eqe_stride(dev);
|
net/mlx4: Change QP allocation scheme
When using BF (Blue-Flame), the QPN overrides the VLAN, CV, and SV fields
in the WQE. Thus, BF may only be used for QPNs with bits 6,7 unset.
The current Ethernet driver code reserves a Tx QP range with 256b alignment.
This is wrong because if there are more than 64 Tx QPs in use,
QPNs >= base + 65 will have bits 6/7 set.
This problem is not specific for the Ethernet driver, any entity that
tries to reserve more than 64 BF-enabled QPs should fail. Also, using
ranges is not necessary here and is wasteful.
The new mechanism introduced here will support reservation for
"Eth QPs eligible for BF" for all drivers: bare-metal, multi-PF, and VFs
(when hypervisors support WC in VMs). The flow we use is:
1. In mlx4_en, allocate Tx QPs one by one instead of a range allocation,
and request "BF enabled QPs" if BF is supported for the function
2. In the ALLOC_RES FW command, change param1 to:
a. param1[23:0] - number of QPs
b. param1[31-24] - flags controlling QPs reservation
Bit 31 refers to Eth blueflame supported QPs. Those QPs must have
bits 6 and 7 unset in order to be used in Ethernet.
Bits 24-30 of the flags are currently reserved.
When a function tries to allocate a QP, it states the required attributes
for this QP. Those attributes are considered "best-effort". If an attribute,
such as Ethernet BF enabled QP, is a must-have attribute, the function has
to check that attribute is supported before trying to do the allocation.
In a lower layer of the code, mlx4_qp_reserve_range masks out the bits
which are unsupported. If SRIOV is used, the PF validates those attributes
and masks out unsupported attributes as well. In order to notify VFs which
attributes are supported, the VF uses QUERY_FUNC_CAP command. This command's
mailbox is filled by the PF, which notifies which QP allocation attributes
it supports.
Signed-off-by: Eugenia Emantayev <eugenia@mellanox.co.il>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 08:57:54 +00:00
|
|
|
dev->caps.alloc_res_qp_mask =
|
net/mlx4: Add A0 hybrid steering
A0 hybrid steering is a form of high performance flow steering.
By using this mode, mlx4 cards use a fast limited table based steering,
in order to enable fast steering of unicast packets to a QP.
In order to implement A0 hybrid steering we allocate resources
from different zones:
(1) General range
(2) Special MAC-assigned QPs [RSS, Raw-Ethernet] each has its own region.
When we create a rss QP or a raw ethernet (A0 steerable and BF ready) QP,
we try hard to allocate the QP from range (2). Otherwise, we try hard not
to allocate from this range. However, when the system is pushed to its
limits and one needs every resource, the allocator uses every region it can.
Meaning, when we run out of raw-eth qps, the allocator allocates from the
general range (and the special-A0 area is no longer active). If we run out
of RSS qps, the mechanism tries to allocate from the raw-eth QP zone. If that
is also exhausted, the allocator will allocate from the general range
(and the A0 region is no longer active).
Note that if a raw-eth qp is allocated from the general range, it attempts
to allocate the range such that bits 6 and 7 (blueflame bits) in the
QP number are not set.
When the feature is used in SRIOV, the VF has to notify the PF what
kind of QP attributes it needs. In order to do that, along with the
"Eth QP blueflame" bit, we reserve a new "A0 steerable QP". According
to the combination of these bits, the PF tries to allocate a suitable QP.
In order to maintain backward compatibility (with older PFs), the PF
notifies which QP attributes it supports via QUERY_FUNC_CAP command.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 08:57:57 +00:00
|
|
|
(dev->caps.bf_reg_size ? MLX4_RESERVE_ETH_BF_QP : 0) |
|
|
|
|
|
MLX4_RESERVE_A0_QP;
|
2015-04-02 13:31:17 +00:00
|
|
|
|
|
|
|
|
if (!(dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_ETS_CFG) &&
|
|
|
|
|
dev->caps.flags & MLX4_DEV_CAP_FLAG_SET_ETH_SCHED) {
|
|
|
|
|
mlx4_warn(dev, "Old device ETS support detected\n");
|
|
|
|
|
mlx4_warn(dev, "Consider upgrading device FW.\n");
|
|
|
|
|
dev->caps.flags2 |= MLX4_DEV_CAP_FLAG2_ETS_CFG;
|
|
|
|
|
}
|
|
|
|
|
|
net/mlx4: Change QP allocation scheme
When using BF (Blue-Flame), the QPN overrides the VLAN, CV, and SV fields
in the WQE. Thus, BF may only be used for QPNs with bits 6,7 unset.
The current Ethernet driver code reserves a Tx QP range with 256b alignment.
This is wrong because if there are more than 64 Tx QPs in use,
QPNs >= base + 65 will have bits 6/7 set.
This problem is not specific for the Ethernet driver, any entity that
tries to reserve more than 64 BF-enabled QPs should fail. Also, using
ranges is not necessary here and is wasteful.
The new mechanism introduced here will support reservation for
"Eth QPs eligible for BF" for all drivers: bare-metal, multi-PF, and VFs
(when hypervisors support WC in VMs). The flow we use is:
1. In mlx4_en, allocate Tx QPs one by one instead of a range allocation,
and request "BF enabled QPs" if BF is supported for the function
2. In the ALLOC_RES FW command, change param1 to:
a. param1[23:0] - number of QPs
b. param1[31-24] - flags controlling QPs reservation
Bit 31 refers to Eth blueflame supported QPs. Those QPs must have
bits 6 and 7 unset in order to be used in Ethernet.
Bits 24-30 of the flags are currently reserved.
When a function tries to allocate a QP, it states the required attributes
for this QP. Those attributes are considered "best-effort". If an attribute,
such as Ethernet BF enabled QP, is a must-have attribute, the function has
to check that attribute is supported before trying to do the allocation.
In a lower layer of the code, mlx4_qp_reserve_range masks out the bits
which are unsupported. If SRIOV is used, the PF validates those attributes
and masks out unsupported attributes as well. In order to notify VFs which
attributes are supported, the VF uses QUERY_FUNC_CAP command. This command's
mailbox is filled by the PF, which notifies which QP allocation attributes
it supports.
Signed-off-by: Eugenia Emantayev <eugenia@mellanox.co.il>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 08:57:54 +00:00
|
|
|
} else {
|
|
|
|
|
dev->caps.alloc_res_qp_mask = 0;
|
|
|
|
|
}
|
2014-09-18 08:50:59 +00:00
|
|
|
|
2015-04-02 13:31:22 +00:00
|
|
|
mlx4_enable_ignore_fcs(dev);
|
|
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
return 0;
|
|
|
|
|
}
|
2014-01-05 15:41:08 +00:00
|
|
|
|
|
|
|
|
static int mlx4_get_pcie_dev_link_caps(struct mlx4_dev *dev,
|
|
|
|
|
enum pci_bus_speed *speed,
|
|
|
|
|
enum pcie_link_width *width)
|
|
|
|
|
{
|
|
|
|
|
u32 lnkcap1, lnkcap2;
|
|
|
|
|
int err1, err2;
|
|
|
|
|
|
|
|
|
|
#define PCIE_MLW_CAP_SHIFT 4 /* start of MLW mask in link capabilities */
|
|
|
|
|
|
|
|
|
|
*speed = PCI_SPEED_UNKNOWN;
|
|
|
|
|
*width = PCIE_LNK_WIDTH_UNKNOWN;
|
|
|
|
|
|
2015-01-25 14:59:35 +00:00
|
|
|
err1 = pcie_capability_read_dword(dev->persist->pdev, PCI_EXP_LNKCAP,
|
|
|
|
|
&lnkcap1);
|
|
|
|
|
err2 = pcie_capability_read_dword(dev->persist->pdev, PCI_EXP_LNKCAP2,
|
|
|
|
|
&lnkcap2);
|
2014-01-05 15:41:08 +00:00
|
|
|
if (!err2 && lnkcap2) { /* PCIe r3.0-compliant */
|
|
|
|
|
if (lnkcap2 & PCI_EXP_LNKCAP2_SLS_8_0GB)
|
|
|
|
|
*speed = PCIE_SPEED_8_0GT;
|
|
|
|
|
else if (lnkcap2 & PCI_EXP_LNKCAP2_SLS_5_0GB)
|
|
|
|
|
*speed = PCIE_SPEED_5_0GT;
|
|
|
|
|
else if (lnkcap2 & PCI_EXP_LNKCAP2_SLS_2_5GB)
|
|
|
|
|
*speed = PCIE_SPEED_2_5GT;
|
|
|
|
|
}
|
|
|
|
|
if (!err1) {
|
|
|
|
|
*width = (lnkcap1 & PCI_EXP_LNKCAP_MLW) >> PCIE_MLW_CAP_SHIFT;
|
|
|
|
|
if (!lnkcap2) { /* pre-r3.0 */
|
|
|
|
|
if (lnkcap1 & PCI_EXP_LNKCAP_SLS_5_0GB)
|
|
|
|
|
*speed = PCIE_SPEED_5_0GT;
|
|
|
|
|
else if (lnkcap1 & PCI_EXP_LNKCAP_SLS_2_5GB)
|
|
|
|
|
*speed = PCIE_SPEED_2_5GT;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (*speed == PCI_SPEED_UNKNOWN || *width == PCIE_LNK_WIDTH_UNKNOWN) {
|
|
|
|
|
return err1 ? err1 :
|
|
|
|
|
err2 ? err2 : -EINVAL;
|
|
|
|
|
}
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void mlx4_check_pcie_caps(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
enum pcie_link_width width, width_cap;
|
|
|
|
|
enum pci_bus_speed speed, speed_cap;
|
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
#define PCIE_SPEED_STR(speed) \
|
|
|
|
|
(speed == PCIE_SPEED_8_0GT ? "8.0GT/s" : \
|
|
|
|
|
speed == PCIE_SPEED_5_0GT ? "5.0GT/s" : \
|
|
|
|
|
speed == PCIE_SPEED_2_5GT ? "2.5GT/s" : \
|
|
|
|
|
"Unknown")
|
|
|
|
|
|
|
|
|
|
err = mlx4_get_pcie_dev_link_caps(dev, &speed_cap, &width_cap);
|
|
|
|
|
if (err) {
|
|
|
|
|
mlx4_warn(dev,
|
|
|
|
|
"Unable to determine PCIe device BW capabilities\n");
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
2015-01-25 14:59:35 +00:00
|
|
|
err = pcie_get_minimum_link(dev->persist->pdev, &speed, &width);
|
2014-01-05 15:41:08 +00:00
|
|
|
if (err || speed == PCI_SPEED_UNKNOWN ||
|
|
|
|
|
width == PCIE_LNK_WIDTH_UNKNOWN) {
|
|
|
|
|
mlx4_warn(dev,
|
|
|
|
|
"Unable to determine PCI device chain minimum BW\n");
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (width != width_cap || speed != speed_cap)
|
|
|
|
|
mlx4_warn(dev,
|
|
|
|
|
"PCIe BW is different than device's capability\n");
|
|
|
|
|
|
|
|
|
|
mlx4_info(dev, "PCIe link speed is %s, device supports %s\n",
|
|
|
|
|
PCIE_SPEED_STR(speed), PCIE_SPEED_STR(speed_cap));
|
|
|
|
|
mlx4_info(dev, "PCIe link width is x%d, device supports x%d\n",
|
|
|
|
|
width, width_cap);
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/*The function checks if there are live vf, return the num of them*/
|
|
|
|
|
static int mlx4_how_many_lives_vf(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
struct mlx4_slave_state *s_state;
|
|
|
|
|
int i;
|
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
|
|
for (i = 1/*the ppf is 0*/; i < dev->num_slaves; ++i) {
|
|
|
|
|
s_state = &priv->mfunc.master.slave_state[i];
|
|
|
|
|
if (s_state->active && s_state->last_cmd !=
|
|
|
|
|
MLX4_COMM_CMD_RESET) {
|
|
|
|
|
mlx4_warn(dev, "%s: slave: %d is still active\n",
|
|
|
|
|
__func__, i);
|
|
|
|
|
ret++;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
return ret;
|
|
|
|
|
}
|
|
|
|
|
|
mlx4_core: Implement mechanism for reserved Q_Keys
The SR-IOV special QP tunneling mechanism uses proxy special QPs
(instead of the real special QPs) for MADs on guests. These proxy QPs
send their packets to a "tunnel" QP owned by the master. The master
then forwards the MAD (after any required paravirtualization) to the
real special QP, which sends out the MAD.
For security reasons (i.e., to prevent guests from sending MADs to
tunnel QPs belonging to other guests), each proxy-tunnel QP pair is
assigned a unique, reserved, Q_Key. These Q_Keys are available only
for proxy and tunnel QPs -- if the guest tries to use these Q_Keys
with other QPs, it will fail.
This patch introduces a mechanism for reserving a block of 64K Q_Keys
for proxy/tunneling use.
The patch introduces also two new fields into mlx4_dev: base_sqpn and
base_tunnel_sqpn.
In SR-IOV mode, the QP numbers for the "real," proxy, and tunnel sqps
are added to the reserved QPN area (so that they will not change).
There are 8 special QPs per port in the HCA, and each of them is
assigned both a proxy and a tunnel QP, for each VF and for the PF as
well in SR-IOV mode.
The QPNs for these QPs are arranged as follows:
1. The real SQP numbers (8)
2. The proxy SQPs (8 * (max number of VFs + max number of PFs)
3. The tunnel SQPs (8 * (max number of VFs + max number of PFs)
To support these QPs, two new fields are added to struct mlx4_dev:
base_sqp: this is the QP number of the first of the real SQPs
base_tunnel_sqp: this is the qp number of the first qp in the tunnel
sqp region. (On guests, this is the first tunnel
sqp of the 8 which are assigned to that guest).
In addition, in SR-IOV mode, sqp_start is the number of the first
proxy SQP in the proxy SQP region. (In guests, this is the first
proxy SQP of the 8 which are assigned to that guest)
Note that in non-SR-IOV mode, there are no proxies and no tunnels.
In this case, sqp_start is set to sqp_base -- which minimizes code
changes.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 08:21:42 +00:00
|
|
|
int mlx4_get_parav_qkey(struct mlx4_dev *dev, u32 qpn, u32 *qkey)
|
|
|
|
|
{
|
|
|
|
|
u32 qk = MLX4_RESERVED_QKEY_BASE;
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 08:40:57 +00:00
|
|
|
|
|
|
|
|
if (qpn >= dev->phys_caps.base_tunnel_sqpn + 8 * MLX4_MFUNC_MAX ||
|
|
|
|
|
qpn < dev->phys_caps.base_proxy_sqpn)
|
mlx4_core: Implement mechanism for reserved Q_Keys
The SR-IOV special QP tunneling mechanism uses proxy special QPs
(instead of the real special QPs) for MADs on guests. These proxy QPs
send their packets to a "tunnel" QP owned by the master. The master
then forwards the MAD (after any required paravirtualization) to the
real special QP, which sends out the MAD.
For security reasons (i.e., to prevent guests from sending MADs to
tunnel QPs belonging to other guests), each proxy-tunnel QP pair is
assigned a unique, reserved, Q_Key. These Q_Keys are available only
for proxy and tunnel QPs -- if the guest tries to use these Q_Keys
with other QPs, it will fail.
This patch introduces a mechanism for reserving a block of 64K Q_Keys
for proxy/tunneling use.
The patch introduces also two new fields into mlx4_dev: base_sqpn and
base_tunnel_sqpn.
In SR-IOV mode, the QP numbers for the "real," proxy, and tunnel sqps
are added to the reserved QPN area (so that they will not change).
There are 8 special QPs per port in the HCA, and each of them is
assigned both a proxy and a tunnel QP, for each VF and for the PF as
well in SR-IOV mode.
The QPNs for these QPs are arranged as follows:
1. The real SQP numbers (8)
2. The proxy SQPs (8 * (max number of VFs + max number of PFs)
3. The tunnel SQPs (8 * (max number of VFs + max number of PFs)
To support these QPs, two new fields are added to struct mlx4_dev:
base_sqp: this is the QP number of the first of the real SQPs
base_tunnel_sqp: this is the qp number of the first qp in the tunnel
sqp region. (On guests, this is the first tunnel
sqp of the 8 which are assigned to that guest).
In addition, in SR-IOV mode, sqp_start is the number of the first
proxy SQP in the proxy SQP region. (In guests, this is the first
proxy SQP of the 8 which are assigned to that guest)
Note that in non-SR-IOV mode, there are no proxies and no tunnels.
In this case, sqp_start is set to sqp_base -- which minimizes code
changes.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 08:21:42 +00:00
|
|
|
return -EINVAL;
|
|
|
|
|
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 08:40:57 +00:00
|
|
|
if (qpn >= dev->phys_caps.base_tunnel_sqpn)
|
mlx4_core: Implement mechanism for reserved Q_Keys
The SR-IOV special QP tunneling mechanism uses proxy special QPs
(instead of the real special QPs) for MADs on guests. These proxy QPs
send their packets to a "tunnel" QP owned by the master. The master
then forwards the MAD (after any required paravirtualization) to the
real special QP, which sends out the MAD.
For security reasons (i.e., to prevent guests from sending MADs to
tunnel QPs belonging to other guests), each proxy-tunnel QP pair is
assigned a unique, reserved, Q_Key. These Q_Keys are available only
for proxy and tunnel QPs -- if the guest tries to use these Q_Keys
with other QPs, it will fail.
This patch introduces a mechanism for reserving a block of 64K Q_Keys
for proxy/tunneling use.
The patch introduces also two new fields into mlx4_dev: base_sqpn and
base_tunnel_sqpn.
In SR-IOV mode, the QP numbers for the "real," proxy, and tunnel sqps
are added to the reserved QPN area (so that they will not change).
There are 8 special QPs per port in the HCA, and each of them is
assigned both a proxy and a tunnel QP, for each VF and for the PF as
well in SR-IOV mode.
The QPNs for these QPs are arranged as follows:
1. The real SQP numbers (8)
2. The proxy SQPs (8 * (max number of VFs + max number of PFs)
3. The tunnel SQPs (8 * (max number of VFs + max number of PFs)
To support these QPs, two new fields are added to struct mlx4_dev:
base_sqp: this is the QP number of the first of the real SQPs
base_tunnel_sqp: this is the qp number of the first qp in the tunnel
sqp region. (On guests, this is the first tunnel
sqp of the 8 which are assigned to that guest).
In addition, in SR-IOV mode, sqp_start is the number of the first
proxy SQP in the proxy SQP region. (In guests, this is the first
proxy SQP of the 8 which are assigned to that guest)
Note that in non-SR-IOV mode, there are no proxies and no tunnels.
In this case, sqp_start is set to sqp_base -- which minimizes code
changes.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 08:21:42 +00:00
|
|
|
/* tunnel qp */
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 08:40:57 +00:00
|
|
|
qk += qpn - dev->phys_caps.base_tunnel_sqpn;
|
mlx4_core: Implement mechanism for reserved Q_Keys
The SR-IOV special QP tunneling mechanism uses proxy special QPs
(instead of the real special QPs) for MADs on guests. These proxy QPs
send their packets to a "tunnel" QP owned by the master. The master
then forwards the MAD (after any required paravirtualization) to the
real special QP, which sends out the MAD.
For security reasons (i.e., to prevent guests from sending MADs to
tunnel QPs belonging to other guests), each proxy-tunnel QP pair is
assigned a unique, reserved, Q_Key. These Q_Keys are available only
for proxy and tunnel QPs -- if the guest tries to use these Q_Keys
with other QPs, it will fail.
This patch introduces a mechanism for reserving a block of 64K Q_Keys
for proxy/tunneling use.
The patch introduces also two new fields into mlx4_dev: base_sqpn and
base_tunnel_sqpn.
In SR-IOV mode, the QP numbers for the "real," proxy, and tunnel sqps
are added to the reserved QPN area (so that they will not change).
There are 8 special QPs per port in the HCA, and each of them is
assigned both a proxy and a tunnel QP, for each VF and for the PF as
well in SR-IOV mode.
The QPNs for these QPs are arranged as follows:
1. The real SQP numbers (8)
2. The proxy SQPs (8 * (max number of VFs + max number of PFs)
3. The tunnel SQPs (8 * (max number of VFs + max number of PFs)
To support these QPs, two new fields are added to struct mlx4_dev:
base_sqp: this is the QP number of the first of the real SQPs
base_tunnel_sqp: this is the qp number of the first qp in the tunnel
sqp region. (On guests, this is the first tunnel
sqp of the 8 which are assigned to that guest).
In addition, in SR-IOV mode, sqp_start is the number of the first
proxy SQP in the proxy SQP region. (In guests, this is the first
proxy SQP of the 8 which are assigned to that guest)
Note that in non-SR-IOV mode, there are no proxies and no tunnels.
In this case, sqp_start is set to sqp_base -- which minimizes code
changes.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 08:21:42 +00:00
|
|
|
else
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 08:40:57 +00:00
|
|
|
qk += qpn - dev->phys_caps.base_proxy_sqpn;
|
mlx4_core: Implement mechanism for reserved Q_Keys
The SR-IOV special QP tunneling mechanism uses proxy special QPs
(instead of the real special QPs) for MADs on guests. These proxy QPs
send their packets to a "tunnel" QP owned by the master. The master
then forwards the MAD (after any required paravirtualization) to the
real special QP, which sends out the MAD.
For security reasons (i.e., to prevent guests from sending MADs to
tunnel QPs belonging to other guests), each proxy-tunnel QP pair is
assigned a unique, reserved, Q_Key. These Q_Keys are available only
for proxy and tunnel QPs -- if the guest tries to use these Q_Keys
with other QPs, it will fail.
This patch introduces a mechanism for reserving a block of 64K Q_Keys
for proxy/tunneling use.
The patch introduces also two new fields into mlx4_dev: base_sqpn and
base_tunnel_sqpn.
In SR-IOV mode, the QP numbers for the "real," proxy, and tunnel sqps
are added to the reserved QPN area (so that they will not change).
There are 8 special QPs per port in the HCA, and each of them is
assigned both a proxy and a tunnel QP, for each VF and for the PF as
well in SR-IOV mode.
The QPNs for these QPs are arranged as follows:
1. The real SQP numbers (8)
2. The proxy SQPs (8 * (max number of VFs + max number of PFs)
3. The tunnel SQPs (8 * (max number of VFs + max number of PFs)
To support these QPs, two new fields are added to struct mlx4_dev:
base_sqp: this is the QP number of the first of the real SQPs
base_tunnel_sqp: this is the qp number of the first qp in the tunnel
sqp region. (On guests, this is the first tunnel
sqp of the 8 which are assigned to that guest).
In addition, in SR-IOV mode, sqp_start is the number of the first
proxy SQP in the proxy SQP region. (In guests, this is the first
proxy SQP of the 8 which are assigned to that guest)
Note that in non-SR-IOV mode, there are no proxies and no tunnels.
In this case, sqp_start is set to sqp_base -- which minimizes code
changes.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 08:21:42 +00:00
|
|
|
*qkey = qk;
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL(mlx4_get_parav_qkey);
|
|
|
|
|
|
2012-08-03 08:40:43 +00:00
|
|
|
void mlx4_sync_pkey_table(struct mlx4_dev *dev, int slave, int port, int i, int val)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = container_of(dev, struct mlx4_priv, dev);
|
|
|
|
|
|
|
|
|
|
if (!mlx4_is_master(dev))
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
priv->virt2phys_pkey[slave][port - 1][i] = val;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL(mlx4_sync_pkey_table);
|
|
|
|
|
|
2012-08-03 08:40:56 +00:00
|
|
|
void mlx4_put_slave_node_guid(struct mlx4_dev *dev, int slave, __be64 guid)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = container_of(dev, struct mlx4_priv, dev);
|
|
|
|
|
|
|
|
|
|
if (!mlx4_is_master(dev))
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
priv->slave_node_guids[slave] = guid;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL(mlx4_put_slave_node_guid);
|
|
|
|
|
|
|
|
|
|
__be64 mlx4_get_slave_node_guid(struct mlx4_dev *dev, int slave)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = container_of(dev, struct mlx4_priv, dev);
|
|
|
|
|
|
|
|
|
|
if (!mlx4_is_master(dev))
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
|
|
return priv->slave_node_guids[slave];
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL(mlx4_get_slave_node_guid);
|
|
|
|
|
|
2012-02-26 09:48:12 +00:00
|
|
|
int mlx4_is_slave_active(struct mlx4_dev *dev, int slave)
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
struct mlx4_slave_state *s_slave;
|
|
|
|
|
|
|
|
|
|
if (!mlx4_is_master(dev))
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
|
|
s_slave = &priv->mfunc.master.slave_state[slave];
|
|
|
|
|
return !!s_slave->active;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL(mlx4_is_slave_active);
|
|
|
|
|
|
2012-12-06 17:11:59 +00:00
|
|
|
static void slave_adjust_steering_mode(struct mlx4_dev *dev,
|
|
|
|
|
struct mlx4_dev_cap *dev_cap,
|
|
|
|
|
struct mlx4_init_hca_param *hca_param)
|
|
|
|
|
{
|
|
|
|
|
dev->caps.steering_mode = hca_param->steering_mode;
|
|
|
|
|
if (dev->caps.steering_mode == MLX4_STEERING_MODE_DEVICE_MANAGED) {
|
|
|
|
|
dev->caps.num_qp_per_mgm = dev_cap->fs_max_num_qp_per_entry;
|
|
|
|
|
dev->caps.fs_log_max_ucast_qp_range_size =
|
|
|
|
|
dev_cap->fs_log_max_ucast_qp_range_size;
|
|
|
|
|
} else
|
|
|
|
|
dev->caps.num_qp_per_mgm =
|
|
|
|
|
4 * ((1 << hca_param->log_mc_entry_sz)/16 - 2);
|
|
|
|
|
|
|
|
|
|
mlx4_dbg(dev, "Steering mode is: %s\n",
|
|
|
|
|
mlx4_steering_mode_str(dev->caps.steering_mode));
|
|
|
|
|
}
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
static int mlx4_slave_cap(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
int err;
|
|
|
|
|
u32 page_size;
|
|
|
|
|
struct mlx4_dev_cap dev_cap;
|
|
|
|
|
struct mlx4_func_cap func_cap;
|
|
|
|
|
struct mlx4_init_hca_param hca_param;
|
2014-11-13 12:45:28 +00:00
|
|
|
u8 i;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
|
|
|
|
|
memset(&hca_param, 0, sizeof(hca_param));
|
|
|
|
|
err = mlx4_QUERY_HCA(dev, &hca_param);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "QUERY_HCA command failed, aborting\n");
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2014-05-14 09:15:14 +00:00
|
|
|
/* fail if the hca has an unknown global capability
|
|
|
|
|
* at this time global_caps should be always zeroed
|
|
|
|
|
*/
|
|
|
|
|
if (hca_param.global_caps) {
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
mlx4_err(dev, "Unknown hca global capabilities\n");
|
|
|
|
|
return -ENOSYS;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
mlx4_log_num_mgm_entry_size = hca_param.log_mc_entry_sz;
|
|
|
|
|
|
2013-04-23 06:06:48 +00:00
|
|
|
dev->caps.hca_core_clock = hca_param.hca_core_clock;
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
memset(&dev_cap, 0, sizeof(dev_cap));
|
2012-05-30 09:14:53 +00:00
|
|
|
dev->caps.max_qp_dest_rdma = 1 << hca_param.log_rd_per_qp;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
err = mlx4_dev_cap(dev, &dev_cap);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "QUERY_DEV_CAP command failed, aborting\n");
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2012-05-30 09:14:53 +00:00
|
|
|
err = mlx4_QUERY_FW(dev);
|
|
|
|
|
if (err)
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "QUERY_FW command failed: could not get FW version\n");
|
2012-05-30 09:14:53 +00:00
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
page_size = ~dev->caps.page_size_cap + 1;
|
|
|
|
|
mlx4_warn(dev, "HCA minimum page size:%d\n", page_size);
|
|
|
|
|
if (page_size > PAGE_SIZE) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "HCA minimum page size of %d bigger than kernel PAGE_SIZE of %ld, aborting\n",
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
page_size, PAGE_SIZE);
|
|
|
|
|
return -ENODEV;
|
|
|
|
|
}
|
|
|
|
|
|
net/mlx4_core: Set UAR page size to 4KB regardless of system page size
problem description:
The current code sets UAR page size equal to system page size.
The ConnectX-3 and ConnectX-3 Pro HWs require minimum 128 UAR pages.
The mlx4 kernel drivers are not loaded if there is less than 128 UAR pages.
solution:
Always set UAR page to 4KB. This allows more UAR pages if the OS
has PAGE_SIZE larger than 4KB. For example, PowerPC kernel use 64KB
system page size, with 4MB uar region, there are 4MB/2/64KB = 32
uars (half for uar, half for blueflame). This does not meet minimum 128
UAR pages requirement. With 4KB UAR page, there are 4MB/2/4KB = 512 uars
which meet the minimum requirement.
Note that only codes in mlx4_core that deal with firmware know that uar
page size is 4KB. Codes that deal with usr page in cq and qp context
(mlx4_ib, mlx4_en and part of mlx4_core) still have the same assumption
that uar page size equals to system page size.
Note that with this implementation, on 64KB system page size kernel, there
are 16 uars per system page but only one uars is used. The other 15
uars are ignored because of the above assumption.
Regarding SR-IOV, mlx4_core in hypervisor will set the uar page size
to 4KB and mlx4_core code in virtual OS will obtain the uar page size from
firmware.
Regarding backward compatibility in SR-IOV, if hypervisor has this new code,
the virtual OS must be updated. If hypervisor has old code, and the virtual
OS has this new code, the new code will be backward compatible with the
old code. If the uar size is big enough, this new code in VF continues to
work with 64 KB uar page size (on PowerPc kernel). If the uar size does not
meet 128 uars requirement, this new code not loaded in VF and print the same
error message as the old code in Hypervisor.
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-17 15:24:26 +00:00
|
|
|
/* Set uar_page_shift for VF */
|
|
|
|
|
dev->uar_page_shift = hca_param.uar_page_sz + 12;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
|
net/mlx4_core: Set UAR page size to 4KB regardless of system page size
problem description:
The current code sets UAR page size equal to system page size.
The ConnectX-3 and ConnectX-3 Pro HWs require minimum 128 UAR pages.
The mlx4 kernel drivers are not loaded if there is less than 128 UAR pages.
solution:
Always set UAR page to 4KB. This allows more UAR pages if the OS
has PAGE_SIZE larger than 4KB. For example, PowerPC kernel use 64KB
system page size, with 4MB uar region, there are 4MB/2/64KB = 32
uars (half for uar, half for blueflame). This does not meet minimum 128
UAR pages requirement. With 4KB UAR page, there are 4MB/2/4KB = 512 uars
which meet the minimum requirement.
Note that only codes in mlx4_core that deal with firmware know that uar
page size is 4KB. Codes that deal with usr page in cq and qp context
(mlx4_ib, mlx4_en and part of mlx4_core) still have the same assumption
that uar page size equals to system page size.
Note that with this implementation, on 64KB system page size kernel, there
are 16 uars per system page but only one uars is used. The other 15
uars are ignored because of the above assumption.
Regarding SR-IOV, mlx4_core in hypervisor will set the uar page size
to 4KB and mlx4_core code in virtual OS will obtain the uar page size from
firmware.
Regarding backward compatibility in SR-IOV, if hypervisor has this new code,
the virtual OS must be updated. If hypervisor has old code, and the virtual
OS has this new code, the new code will be backward compatible with the
old code. If the uar size is big enough, this new code in VF continues to
work with 64 KB uar page size (on PowerPc kernel). If the uar size does not
meet 128 uars requirement, this new code not loaded in VF and print the same
error message as the old code in Hypervisor.
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-17 15:24:26 +00:00
|
|
|
/* Make sure the master uar page size is valid */
|
|
|
|
|
if (dev->uar_page_shift > PAGE_SHIFT) {
|
|
|
|
|
mlx4_err(dev,
|
|
|
|
|
"Invalid configuration: uar page size is larger than system page size\n");
|
|
|
|
|
return -ENODEV;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
}
|
|
|
|
|
|
net/mlx4_core: Set UAR page size to 4KB regardless of system page size
problem description:
The current code sets UAR page size equal to system page size.
The ConnectX-3 and ConnectX-3 Pro HWs require minimum 128 UAR pages.
The mlx4 kernel drivers are not loaded if there is less than 128 UAR pages.
solution:
Always set UAR page to 4KB. This allows more UAR pages if the OS
has PAGE_SIZE larger than 4KB. For example, PowerPC kernel use 64KB
system page size, with 4MB uar region, there are 4MB/2/64KB = 32
uars (half for uar, half for blueflame). This does not meet minimum 128
UAR pages requirement. With 4KB UAR page, there are 4MB/2/4KB = 512 uars
which meet the minimum requirement.
Note that only codes in mlx4_core that deal with firmware know that uar
page size is 4KB. Codes that deal with usr page in cq and qp context
(mlx4_ib, mlx4_en and part of mlx4_core) still have the same assumption
that uar page size equals to system page size.
Note that with this implementation, on 64KB system page size kernel, there
are 16 uars per system page but only one uars is used. The other 15
uars are ignored because of the above assumption.
Regarding SR-IOV, mlx4_core in hypervisor will set the uar page size
to 4KB and mlx4_core code in virtual OS will obtain the uar page size from
firmware.
Regarding backward compatibility in SR-IOV, if hypervisor has this new code,
the virtual OS must be updated. If hypervisor has old code, and the virtual
OS has this new code, the new code will be backward compatible with the
old code. If the uar size is big enough, this new code in VF continues to
work with 64 KB uar page size (on PowerPc kernel). If the uar size does not
meet 128 uars requirement, this new code not loaded in VF and print the same
error message as the old code in Hypervisor.
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-17 15:24:26 +00:00
|
|
|
/* Set reserved_uars based on the uar_page_shift */
|
|
|
|
|
mlx4_set_num_reserved_uars(dev, &dev_cap);
|
|
|
|
|
|
|
|
|
|
/* Although uar page size in FW differs from system page size,
|
|
|
|
|
* upper software layers (mlx4_ib, mlx4_en and part of mlx4_core)
|
|
|
|
|
* still works with assumption that uar page size == system page size
|
|
|
|
|
*/
|
|
|
|
|
dev->caps.uar_page_size = PAGE_SIZE;
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
memset(&func_cap, 0, sizeof(func_cap));
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 08:40:57 +00:00
|
|
|
err = mlx4_QUERY_FUNC_CAP(dev, 0, &func_cap);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "QUERY_FUNC_CAP general command failed, aborting (%d)\n",
|
|
|
|
|
err);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if ((func_cap.pf_context_behaviour | PF_CONTEXT_BEHAVIOUR_MASK) !=
|
|
|
|
|
PF_CONTEXT_BEHAVIOUR_MASK) {
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 08:58:00 +00:00
|
|
|
mlx4_err(dev, "Unknown pf context behaviour %x known flags %x\n",
|
|
|
|
|
func_cap.pf_context_behaviour, PF_CONTEXT_BEHAVIOUR_MASK);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
return -ENOSYS;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
dev->caps.num_ports = func_cap.num_ports;
|
mlx4: Structures and init/teardown for VF resource quotas
This is step #1 for implementing SRIOV resource quotas for VFs.
Quotas are implemented per resource type for VFs and the PF, to prevent
any entity from simply grabbing all the resources for itself and leaving
the other entities unable to obtain such resources.
Resources which are allocated using quotas: QPs, CQs, SRQs, MPTs, MTTs, MAC,
VLAN, and Counters.
The quota system works as follows:
Each entity (VF or PF) is given a max number of a given resource (its quota),
and a guaranteed minimum number for each resource (starvation prevention).
For QPs, CQs, SRQs, MPTs and MTTs:
50% of the available quantity for the resource is divided equally among
the PF and all the active VFs (i.e., the number of VFs in the mlx4_core module
parameter "num_vfs"). This 50% represents the "guaranteed minimum" pool.
The other 50% is the "free pool", allocated on a first-come-first-serve basis.
For each VF/PF, resources are first allocated from its "guaranteed-minimum"
pool. When that pool is exhausted, the driver attempts to allocate from
the resource "free-pool".
The quota (i.e., max) for the VFs and the PF is:
The free-pool amount (50% of the real max) + the guaranteed minimum
For MACs:
Guarantee 2 MACs per VF/PF per port. As a result, since we have only
128 MACs per port, reduce the allowable number of VFs from 64 to 63.
Any remaining MACs are put into a free pool.
For VLANs:
For the PF, the per-port quota is 128 and guarantee is 64
(to allow the PF to register at least a VLAN per VF in VST mode).
For the VFs, the per-port quota is 64 and the guarantee is 0.
We assume that VGT VFs are trusted not to abuse the VLAN resource.
For Counters:
For all functions (PF and VFs), the quota is 128 and the guarantee is 0.
In this patch, we define the needed structures, which are added to the
resource-tracker struct. In addition, we do initialization
for the resource quota, and adjust the query_device response to use quotas
rather than resource maxima.
As part of the implementation, we introduce a new field in
mlx4_dev: quotas. This field holds the resource quotas used
to report maxima to the upper layers (ib_core, via query_device).
The HCA maxima of these values are passed to the VFs (via
QUERY_HCA) so that they may continue to use these in handling
QPs, CQs, SRQs and MPTs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 08:03:23 +00:00
|
|
|
dev->quotas.qp = func_cap.qp_quota;
|
|
|
|
|
dev->quotas.srq = func_cap.srq_quota;
|
|
|
|
|
dev->quotas.cq = func_cap.cq_quota;
|
|
|
|
|
dev->quotas.mpt = func_cap.mpt_quota;
|
|
|
|
|
dev->quotas.mtt = func_cap.mtt_quota;
|
|
|
|
|
dev->caps.num_qps = 1 << hca_param.log_num_qps;
|
|
|
|
|
dev->caps.num_srqs = 1 << hca_param.log_num_srqs;
|
|
|
|
|
dev->caps.num_cqs = 1 << hca_param.log_num_cqs;
|
|
|
|
|
dev->caps.num_mpts = 1 << hca_param.log_mpt_sz;
|
|
|
|
|
dev->caps.num_eqs = func_cap.max_eq;
|
|
|
|
|
dev->caps.reserved_eqs = func_cap.reserved_eq;
|
2015-01-27 13:58:00 +00:00
|
|
|
dev->caps.reserved_lkey = func_cap.reserved_lkey;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
dev->caps.num_pds = MLX4_NUM_PDS;
|
|
|
|
|
dev->caps.num_mgms = 0;
|
|
|
|
|
dev->caps.num_amgms = 0;
|
|
|
|
|
|
|
|
|
|
if (dev->caps.num_ports > MLX4_MAX_PORTS) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "HCA has %d ports, but we only support %d, aborting\n",
|
|
|
|
|
dev->caps.num_ports, MLX4_MAX_PORTS);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
return -ENODEV;
|
|
|
|
|
}
|
|
|
|
|
|
net/mlx4_core: Replace VF zero mac with random mac in mlx4_core
By design, when no default MAC addresses are set in the Hypervisor for VFs,
the VFs are passed zero-macs. When such a MAC is received by the VF, it
generates a random MAC address and registers that MAC address
with the Hypervisor.
This random mac generation is currently done in the mlx4_en module.
There is a problem, though, if the mlx4_ib module is loaded by a VF before
the mlx4_en module. In this case, for RoCE, mlx4_ib will see the un-replaced
zero-mac and register that zero-mac as part of QP1 initialization.
Having a zero-mac in the port's MAC table creates problems for a
Baseboard Management Console. The BMC occasionally sends packets with a
zero-mac destination MAC. If there is a zero-mac present in the port's
MAC table, the FW will send such BMC packets to the host driver rather than
to the wire, and BMC will stop working.
To address this problem, we move the replacement of zero-mac addresses
with random-mac addresses to procedure mlx4_slave_cap(), which is part of the
driver startup for VFs, and is before activation of mlx4_ib and mlx4_en.
As a result, zero-mac addresses will never be registered in the port MAC table
by the driver.
In addition, when mlx4_en does initialize the net device, it needs to set
the NET_ADDR_RANDOM flag in the netdev structure if the address was
randomly generated. This is done so that udev on the VM does not create
a new device name after each VF probe (VM boot and such). To accomplish this,
we add a per-port flag in mlx4_dev which gets set whenever mlx4_core replaces
a zero-mac with a randomly-generated mac. This flag is examined when mlx4_en
initializes the net-device.
Fix was suggested by Matan Barak <matanb@mellanox.com>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-10-14 14:43:48 +00:00
|
|
|
mlx4_replace_zero_macs(dev);
|
|
|
|
|
|
2014-05-29 13:31:03 +00:00
|
|
|
dev->caps.qp0_qkey = kcalloc(dev->caps.num_ports, sizeof(u32), GFP_KERNEL);
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 08:40:57 +00:00
|
|
|
dev->caps.qp0_tunnel = kcalloc(dev->caps.num_ports, sizeof (u32), GFP_KERNEL);
|
|
|
|
|
dev->caps.qp0_proxy = kcalloc(dev->caps.num_ports, sizeof (u32), GFP_KERNEL);
|
|
|
|
|
dev->caps.qp1_tunnel = kcalloc(dev->caps.num_ports, sizeof (u32), GFP_KERNEL);
|
|
|
|
|
dev->caps.qp1_proxy = kcalloc(dev->caps.num_ports, sizeof (u32), GFP_KERNEL);
|
|
|
|
|
|
|
|
|
|
if (!dev->caps.qp0_tunnel || !dev->caps.qp0_proxy ||
|
2014-05-29 13:31:03 +00:00
|
|
|
!dev->caps.qp1_tunnel || !dev->caps.qp1_proxy ||
|
|
|
|
|
!dev->caps.qp0_qkey) {
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 08:40:57 +00:00
|
|
|
err = -ENOMEM;
|
|
|
|
|
goto err_mem;
|
|
|
|
|
}
|
|
|
|
|
|
mlx4: Put physical GID and P_Key table sizes in mlx4_phys_caps struct and paravirtualize them
To allow easy paravirtualization of P_Key and GID table sizes, keep
paravirtualized sizes in mlx4_dev->caps, but save the actual physical
sizes from FW in struct: mlx4_dev->phys_cap.
In addition, in SR-IOV mode, do the following:
1. Reduce reported P_Key table size by 1.
This is done to reserve the highest P_Key index for internal use,
for declaring an invalid P_Key in P_Key paravirtualization.
We require a P_Key index which always contain an invalid P_Key
value for this purpose (i.e., one which cannot be modified by
the subnet manager). The way to do this is to reduce the
P_Key table size reported to the subnet manager by 1, so that
it will not attempt to access the P_Key at index #127.
2. Paravirtualize the GID table size to 1. Thus, each guest sees
only a single GID (at its paravirtualized index 0).
In addition, since we are paravirtualizing the GID table size to 1, we
add paravirtualization of the master GID event here (i.e., we do not
do ib_dispatch_event() for the GUID change event on the master, since
its (only) GUID never changes).
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 08:21:44 +00:00
|
|
|
for (i = 1; i <= dev->caps.num_ports; ++i) {
|
2014-11-13 12:45:28 +00:00
|
|
|
err = mlx4_QUERY_FUNC_CAP(dev, i, &func_cap);
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 08:40:57 +00:00
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "QUERY_FUNC_CAP port command failed for port %d, aborting (%d)\n",
|
|
|
|
|
i, err);
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 08:40:57 +00:00
|
|
|
goto err_mem;
|
|
|
|
|
}
|
2014-05-29 13:31:03 +00:00
|
|
|
dev->caps.qp0_qkey[i - 1] = func_cap.qp0_qkey;
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 08:40:57 +00:00
|
|
|
dev->caps.qp0_tunnel[i - 1] = func_cap.qp0_tunnel_qpn;
|
|
|
|
|
dev->caps.qp0_proxy[i - 1] = func_cap.qp0_proxy_qpn;
|
|
|
|
|
dev->caps.qp1_tunnel[i - 1] = func_cap.qp1_tunnel_qpn;
|
|
|
|
|
dev->caps.qp1_proxy[i - 1] = func_cap.qp1_proxy_qpn;
|
2012-05-30 09:14:54 +00:00
|
|
|
dev->caps.port_mask[i] = dev->caps.port_type[i];
|
2013-12-19 19:20:12 +00:00
|
|
|
dev->caps.phys_port_id[i] = func_cap.phys_port_id;
|
2015-11-12 17:35:30 +00:00
|
|
|
err = mlx4_get_slave_pkey_gid_tbl_len(dev, i,
|
|
|
|
|
&dev->caps.gid_table_len[i],
|
|
|
|
|
&dev->caps.pkey_table_len[i]);
|
|
|
|
|
if (err)
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 08:40:57 +00:00
|
|
|
goto err_mem;
|
mlx4: Put physical GID and P_Key table sizes in mlx4_phys_caps struct and paravirtualize them
To allow easy paravirtualization of P_Key and GID table sizes, keep
paravirtualized sizes in mlx4_dev->caps, but save the actual physical
sizes from FW in struct: mlx4_dev->phys_cap.
In addition, in SR-IOV mode, do the following:
1. Reduce reported P_Key table size by 1.
This is done to reserve the highest P_Key index for internal use,
for declaring an invalid P_Key in P_Key paravirtualization.
We require a P_Key index which always contain an invalid P_Key
value for this purpose (i.e., one which cannot be modified by
the subnet manager). The way to do this is to reduce the
P_Key table size reported to the subnet manager by 1, so that
it will not attempt to access the P_Key at index #127.
2. Paravirtualize the GID table size to 1. Thus, each guest sees
only a single GID (at its paravirtualized index 0).
In addition, since we are paravirtualizing the GID table size to 1, we
add paravirtualization of the master GID event here (i.e., we do not
do ib_dispatch_event() for the GUID change event on the master, since
its (only) GUID never changes).
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 08:21:44 +00:00
|
|
|
}
|
2012-05-30 09:14:54 +00:00
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
if (dev->caps.uar_page_size * (dev->caps.num_uars -
|
|
|
|
|
dev->caps.reserved_uars) >
|
2015-01-25 14:59:35 +00:00
|
|
|
pci_resource_len(dev->persist->pdev,
|
|
|
|
|
2)) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "HCA reported UAR region size of 0x%x bigger than PCI resource 2 size of 0x%llx, aborting\n",
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
dev->caps.uar_page_size * dev->caps.num_uars,
|
2015-01-25 14:59:35 +00:00
|
|
|
(unsigned long long)
|
|
|
|
|
pci_resource_len(dev->persist->pdev, 2));
|
2015-11-12 17:35:30 +00:00
|
|
|
err = -ENOMEM;
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 08:40:57 +00:00
|
|
|
goto err_mem;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
}
|
|
|
|
|
|
2012-10-21 14:59:24 +00:00
|
|
|
if (hca_param.dev_cap_enabled & MLX4_DEV_CAP_64B_EQE_ENABLED) {
|
|
|
|
|
dev->caps.eqe_size = 64;
|
|
|
|
|
dev->caps.eqe_factor = 1;
|
|
|
|
|
} else {
|
|
|
|
|
dev->caps.eqe_size = 32;
|
|
|
|
|
dev->caps.eqe_factor = 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (hca_param.dev_cap_enabled & MLX4_DEV_CAP_64B_CQE_ENABLED) {
|
|
|
|
|
dev->caps.cqe_size = 64;
|
2014-09-18 08:50:59 +00:00
|
|
|
dev->caps.userspace_caps |= MLX4_USER_DEV_CAP_LARGE_CQE;
|
2012-10-21 14:59:24 +00:00
|
|
|
} else {
|
|
|
|
|
dev->caps.cqe_size = 32;
|
|
|
|
|
}
|
|
|
|
|
|
2014-09-18 08:50:59 +00:00
|
|
|
if (hca_param.dev_cap_enabled & MLX4_DEV_CAP_EQE_STRIDE_ENABLED) {
|
|
|
|
|
dev->caps.eqe_size = hca_param.eqe_size;
|
|
|
|
|
dev->caps.eqe_factor = 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (hca_param.dev_cap_enabled & MLX4_DEV_CAP_CQE_STRIDE_ENABLED) {
|
|
|
|
|
dev->caps.cqe_size = hca_param.cqe_size;
|
|
|
|
|
/* User still need to know when CQE > 32B */
|
|
|
|
|
dev->caps.userspace_caps |= MLX4_USER_DEV_CAP_LARGE_CQE;
|
|
|
|
|
}
|
|
|
|
|
|
2013-06-20 11:58:10 +00:00
|
|
|
dev->caps.flags2 &= ~MLX4_DEV_CAP_FLAG2_TS;
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_warn(dev, "Timestamping is not supported in slave mode\n");
|
2013-06-20 11:58:10 +00:00
|
|
|
|
2012-12-06 17:11:59 +00:00
|
|
|
slave_adjust_steering_mode(dev, &dev_cap, &hca_param);
|
2015-04-02 13:31:06 +00:00
|
|
|
mlx4_dbg(dev, "RSS support for IP fragments is %s\n",
|
|
|
|
|
hca_param.rss_ip_frags ? "on" : "off");
|
2012-12-06 17:11:59 +00:00
|
|
|
|
net/mlx4: Change QP allocation scheme
When using BF (Blue-Flame), the QPN overrides the VLAN, CV, and SV fields
in the WQE. Thus, BF may only be used for QPNs with bits 6,7 unset.
The current Ethernet driver code reserves a Tx QP range with 256b alignment.
This is wrong because if there are more than 64 Tx QPs in use,
QPNs >= base + 65 will have bits 6/7 set.
This problem is not specific for the Ethernet driver, any entity that
tries to reserve more than 64 BF-enabled QPs should fail. Also, using
ranges is not necessary here and is wasteful.
The new mechanism introduced here will support reservation for
"Eth QPs eligible for BF" for all drivers: bare-metal, multi-PF, and VFs
(when hypervisors support WC in VMs). The flow we use is:
1. In mlx4_en, allocate Tx QPs one by one instead of a range allocation,
and request "BF enabled QPs" if BF is supported for the function
2. In the ALLOC_RES FW command, change param1 to:
a. param1[23:0] - number of QPs
b. param1[31-24] - flags controlling QPs reservation
Bit 31 refers to Eth blueflame supported QPs. Those QPs must have
bits 6 and 7 unset in order to be used in Ethernet.
Bits 24-30 of the flags are currently reserved.
When a function tries to allocate a QP, it states the required attributes
for this QP. Those attributes are considered "best-effort". If an attribute,
such as Ethernet BF enabled QP, is a must-have attribute, the function has
to check that attribute is supported before trying to do the allocation.
In a lower layer of the code, mlx4_qp_reserve_range masks out the bits
which are unsupported. If SRIOV is used, the PF validates those attributes
and masks out unsupported attributes as well. In order to notify VFs which
attributes are supported, the VF uses QUERY_FUNC_CAP command. This command's
mailbox is filled by the PF, which notifies which QP allocation attributes
it supports.
Signed-off-by: Eugenia Emantayev <eugenia@mellanox.co.il>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 08:57:54 +00:00
|
|
|
if (func_cap.extra_flags & MLX4_QUERY_FUNC_FLAGS_BF_RES_QP &&
|
|
|
|
|
dev->caps.bf_reg_size)
|
|
|
|
|
dev->caps.alloc_res_qp_mask |= MLX4_RESERVE_ETH_BF_QP;
|
|
|
|
|
|
net/mlx4: Add A0 hybrid steering
A0 hybrid steering is a form of high performance flow steering.
By using this mode, mlx4 cards use a fast limited table based steering,
in order to enable fast steering of unicast packets to a QP.
In order to implement A0 hybrid steering we allocate resources
from different zones:
(1) General range
(2) Special MAC-assigned QPs [RSS, Raw-Ethernet] each has its own region.
When we create a rss QP or a raw ethernet (A0 steerable and BF ready) QP,
we try hard to allocate the QP from range (2). Otherwise, we try hard not
to allocate from this range. However, when the system is pushed to its
limits and one needs every resource, the allocator uses every region it can.
Meaning, when we run out of raw-eth qps, the allocator allocates from the
general range (and the special-A0 area is no longer active). If we run out
of RSS qps, the mechanism tries to allocate from the raw-eth QP zone. If that
is also exhausted, the allocator will allocate from the general range
(and the A0 region is no longer active).
Note that if a raw-eth qp is allocated from the general range, it attempts
to allocate the range such that bits 6 and 7 (blueflame bits) in the
QP number are not set.
When the feature is used in SRIOV, the VF has to notify the PF what
kind of QP attributes it needs. In order to do that, along with the
"Eth QP blueflame" bit, we reserve a new "A0 steerable QP". According
to the combination of these bits, the PF tries to allocate a suitable QP.
In order to maintain backward compatibility (with older PFs), the PF
notifies which QP attributes it supports via QUERY_FUNC_CAP command.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 08:57:57 +00:00
|
|
|
if (func_cap.extra_flags & MLX4_QUERY_FUNC_FLAGS_A0_RES_QP)
|
|
|
|
|
dev->caps.alloc_res_qp_mask |= MLX4_RESERVE_A0_QP;
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
return 0;
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 08:40:57 +00:00
|
|
|
|
|
|
|
|
err_mem:
|
2014-05-29 13:31:03 +00:00
|
|
|
kfree(dev->caps.qp0_qkey);
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 08:40:57 +00:00
|
|
|
kfree(dev->caps.qp0_tunnel);
|
|
|
|
|
kfree(dev->caps.qp0_proxy);
|
|
|
|
|
kfree(dev->caps.qp1_tunnel);
|
|
|
|
|
kfree(dev->caps.qp1_proxy);
|
2014-05-29 13:31:03 +00:00
|
|
|
dev->caps.qp0_qkey = NULL;
|
|
|
|
|
dev->caps.qp0_tunnel = NULL;
|
|
|
|
|
dev->caps.qp0_proxy = NULL;
|
|
|
|
|
dev->caps.qp1_tunnel = NULL;
|
|
|
|
|
dev->caps.qp1_proxy = NULL;
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 08:40:57 +00:00
|
|
|
|
|
|
|
|
return err;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
}
|
2007-05-09 01:00:38 +00:00
|
|
|
|
2013-10-15 14:55:24 +00:00
|
|
|
static void mlx4_request_modules(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
int port;
|
|
|
|
|
int has_ib_port = false;
|
|
|
|
|
int has_eth_port = false;
|
|
|
|
|
#define EN_DRV_NAME "mlx4_en"
|
|
|
|
|
#define IB_DRV_NAME "mlx4_ib"
|
|
|
|
|
|
|
|
|
|
for (port = 1; port <= dev->caps.num_ports; port++) {
|
|
|
|
|
if (dev->caps.port_type[port] == MLX4_PORT_TYPE_IB)
|
|
|
|
|
has_ib_port = true;
|
|
|
|
|
else if (dev->caps.port_type[port] == MLX4_PORT_TYPE_ETH)
|
|
|
|
|
has_eth_port = true;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (has_eth_port)
|
|
|
|
|
request_module_nowait(EN_DRV_NAME);
|
2014-05-04 14:07:24 +00:00
|
|
|
if (has_ib_port || (dev->caps.flags & MLX4_DEV_CAP_FLAG_IBOE))
|
|
|
|
|
request_module_nowait(IB_DRV_NAME);
|
2013-10-15 14:55:24 +00:00
|
|
|
}
|
|
|
|
|
|
2008-10-22 22:38:42 +00:00
|
|
|
/*
|
|
|
|
|
* Change the port configuration of the device.
|
|
|
|
|
* Every user of this function must hold the port mutex.
|
|
|
|
|
*/
|
2009-03-19 02:45:11 +00:00
|
|
|
int mlx4_change_port_types(struct mlx4_dev *dev,
|
|
|
|
|
enum mlx4_port_type *port_types)
|
2008-10-22 22:38:42 +00:00
|
|
|
{
|
|
|
|
|
int err = 0;
|
|
|
|
|
int change = 0;
|
|
|
|
|
int port;
|
|
|
|
|
|
|
|
|
|
for (port = 0; port < dev->caps.num_ports; port++) {
|
2009-03-19 02:45:11 +00:00
|
|
|
/* Change the port type only if the new type is different
|
|
|
|
|
* from the current, and not set to Auto */
|
2012-02-21 03:41:07 +00:00
|
|
|
if (port_types[port] != dev->caps.port_type[port + 1])
|
2008-10-22 22:38:42 +00:00
|
|
|
change = 1;
|
|
|
|
|
}
|
|
|
|
|
if (change) {
|
|
|
|
|
mlx4_unregister_device(dev);
|
|
|
|
|
for (port = 1; port <= dev->caps.num_ports; port++) {
|
|
|
|
|
mlx4_CLOSE_PORT(dev, port);
|
2012-02-23 07:04:35 +00:00
|
|
|
dev->caps.port_type[port] = port_types[port - 1];
|
mlx4: Put physical GID and P_Key table sizes in mlx4_phys_caps struct and paravirtualize them
To allow easy paravirtualization of P_Key and GID table sizes, keep
paravirtualized sizes in mlx4_dev->caps, but save the actual physical
sizes from FW in struct: mlx4_dev->phys_cap.
In addition, in SR-IOV mode, do the following:
1. Reduce reported P_Key table size by 1.
This is done to reserve the highest P_Key index for internal use,
for declaring an invalid P_Key in P_Key paravirtualization.
We require a P_Key index which always contain an invalid P_Key
value for this purpose (i.e., one which cannot be modified by
the subnet manager). The way to do this is to reduce the
P_Key table size reported to the subnet manager by 1, so that
it will not attempt to access the P_Key at index #127.
2. Paravirtualize the GID table size to 1. Thus, each guest sees
only a single GID (at its paravirtualized index 0).
In addition, since we are paravirtualizing the GID table size to 1, we
add paravirtualization of the master GID event here (i.e., we do not
do ib_dispatch_event() for the GUID change event on the master, since
its (only) GUID never changes).
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 08:21:44 +00:00
|
|
|
err = mlx4_SET_PORT(dev, port, -1);
|
2008-10-22 22:38:42 +00:00
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to set port %d, aborting\n",
|
|
|
|
|
port);
|
2008-10-22 22:38:42 +00:00
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
mlx4_set_port_mask(dev);
|
|
|
|
|
err = mlx4_register_device(dev);
|
2013-10-15 14:55:24 +00:00
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "Failed to register device\n");
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
mlx4_request_modules(dev);
|
2008-10-22 22:38:42 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
out:
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static ssize_t show_port_type(struct device *dev,
|
|
|
|
|
struct device_attribute *attr,
|
|
|
|
|
char *buf)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_port_info *info = container_of(attr, struct mlx4_port_info,
|
|
|
|
|
port_attr);
|
|
|
|
|
struct mlx4_dev *mdev = info->dev;
|
2009-03-19 02:45:11 +00:00
|
|
|
char type[8];
|
|
|
|
|
|
|
|
|
|
sprintf(type, "%s",
|
|
|
|
|
(mdev->caps.port_type[info->port] == MLX4_PORT_TYPE_IB) ?
|
|
|
|
|
"ib" : "eth");
|
|
|
|
|
if (mdev->caps.possible_type[info->port] == MLX4_PORT_TYPE_AUTO)
|
|
|
|
|
sprintf(buf, "auto (%s)\n", type);
|
|
|
|
|
else
|
|
|
|
|
sprintf(buf, "%s\n", type);
|
2008-10-22 22:38:42 +00:00
|
|
|
|
2009-03-19 02:45:11 +00:00
|
|
|
return strlen(buf);
|
2008-10-22 22:38:42 +00:00
|
|
|
}
|
|
|
|
|
|
2016-02-26 16:32:25 +00:00
|
|
|
static int __set_port_type(struct mlx4_port_info *info,
|
|
|
|
|
enum mlx4_port_type port_type)
|
2008-10-22 22:38:42 +00:00
|
|
|
{
|
|
|
|
|
struct mlx4_dev *mdev = info->dev;
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(mdev);
|
|
|
|
|
enum mlx4_port_type types[MLX4_MAX_PORTS];
|
2009-03-19 02:45:11 +00:00
|
|
|
enum mlx4_port_type new_types[MLX4_MAX_PORTS];
|
2008-10-22 22:38:42 +00:00
|
|
|
int i;
|
|
|
|
|
int err = 0;
|
|
|
|
|
|
2016-10-27 13:27:14 +00:00
|
|
|
if ((port_type & mdev->caps.supported_type[info->port]) != port_type) {
|
|
|
|
|
mlx4_err(mdev,
|
|
|
|
|
"Requested port type for port %d is not supported on this HCA\n",
|
|
|
|
|
info->port);
|
|
|
|
|
err = -EINVAL;
|
|
|
|
|
goto err_sup;
|
|
|
|
|
}
|
|
|
|
|
|
2009-03-19 02:45:11 +00:00
|
|
|
mlx4_stop_sense(mdev);
|
2008-10-22 22:38:42 +00:00
|
|
|
mutex_lock(&priv->port_mutex);
|
2016-02-26 16:32:25 +00:00
|
|
|
info->tmp_type = port_type;
|
|
|
|
|
|
2009-03-19 02:45:11 +00:00
|
|
|
/* Possible type is always the one that was delivered */
|
|
|
|
|
mdev->caps.possible_type[info->port] = info->tmp_type;
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < mdev->caps.num_ports; i++) {
|
2008-10-22 22:38:42 +00:00
|
|
|
types[i] = priv->port[i+1].tmp_type ? priv->port[i+1].tmp_type :
|
2009-03-19 02:45:11 +00:00
|
|
|
mdev->caps.possible_type[i+1];
|
|
|
|
|
if (types[i] == MLX4_PORT_TYPE_AUTO)
|
|
|
|
|
types[i] = mdev->caps.port_type[i+1];
|
|
|
|
|
}
|
2008-10-22 22:38:42 +00:00
|
|
|
|
2011-12-19 04:00:26 +00:00
|
|
|
if (!(mdev->caps.flags & MLX4_DEV_CAP_FLAG_DPDP) &&
|
|
|
|
|
!(mdev->caps.flags & MLX4_DEV_CAP_FLAG_SENSE_SUPPORT)) {
|
2009-03-19 02:45:11 +00:00
|
|
|
for (i = 1; i <= mdev->caps.num_ports; i++) {
|
|
|
|
|
if (mdev->caps.possible_type[i] == MLX4_PORT_TYPE_AUTO) {
|
|
|
|
|
mdev->caps.possible_type[i] = mdev->caps.port_type[i];
|
|
|
|
|
err = -EINVAL;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(mdev, "Auto sensing is not supported on this HCA. Set only 'eth' or 'ib' for both ports (should be the same)\n");
|
2009-03-19 02:45:11 +00:00
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
mlx4_do_sense_ports(mdev, new_types, types);
|
|
|
|
|
|
|
|
|
|
err = mlx4_check_port_params(mdev, new_types);
|
2008-10-22 22:38:42 +00:00
|
|
|
if (err)
|
|
|
|
|
goto out;
|
|
|
|
|
|
2009-03-19 02:45:11 +00:00
|
|
|
/* We are about to apply the changes after the configuration
|
|
|
|
|
* was verified, no need to remember the temporary types
|
|
|
|
|
* any more */
|
|
|
|
|
for (i = 0; i < mdev->caps.num_ports; i++)
|
|
|
|
|
priv->port[i + 1].tmp_type = 0;
|
2008-10-22 22:38:42 +00:00
|
|
|
|
2009-03-19 02:45:11 +00:00
|
|
|
err = mlx4_change_port_types(mdev, new_types);
|
2008-10-22 22:38:42 +00:00
|
|
|
|
|
|
|
|
out:
|
2009-03-19 02:45:11 +00:00
|
|
|
mlx4_start_sense(mdev);
|
2008-10-22 22:38:42 +00:00
|
|
|
mutex_unlock(&priv->port_mutex);
|
2016-10-27 13:27:14 +00:00
|
|
|
err_sup:
|
2016-02-26 16:32:25 +00:00
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static ssize_t set_port_type(struct device *dev,
|
|
|
|
|
struct device_attribute *attr,
|
|
|
|
|
const char *buf, size_t count)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_port_info *info = container_of(attr, struct mlx4_port_info,
|
|
|
|
|
port_attr);
|
|
|
|
|
struct mlx4_dev *mdev = info->dev;
|
|
|
|
|
enum mlx4_port_type port_type;
|
|
|
|
|
static DEFINE_MUTEX(set_port_type_mutex);
|
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
mutex_lock(&set_port_type_mutex);
|
|
|
|
|
|
|
|
|
|
if (!strcmp(buf, "ib\n")) {
|
|
|
|
|
port_type = MLX4_PORT_TYPE_IB;
|
|
|
|
|
} else if (!strcmp(buf, "eth\n")) {
|
|
|
|
|
port_type = MLX4_PORT_TYPE_ETH;
|
|
|
|
|
} else if (!strcmp(buf, "auto\n")) {
|
|
|
|
|
port_type = MLX4_PORT_TYPE_AUTO;
|
|
|
|
|
} else {
|
|
|
|
|
mlx4_err(mdev, "%s is not supported port type\n", buf);
|
|
|
|
|
err = -EINVAL;
|
|
|
|
|
goto err_out;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = __set_port_type(info, port_type);
|
|
|
|
|
|
2014-11-02 14:26:14 +00:00
|
|
|
err_out:
|
|
|
|
|
mutex_unlock(&set_port_type_mutex);
|
|
|
|
|
|
2008-10-22 22:38:42 +00:00
|
|
|
return err ? err : count;
|
|
|
|
|
}
|
|
|
|
|
|
2012-01-11 17:02:17 +00:00
|
|
|
enum ibta_mtu {
|
|
|
|
|
IB_MTU_256 = 1,
|
|
|
|
|
IB_MTU_512 = 2,
|
|
|
|
|
IB_MTU_1024 = 3,
|
|
|
|
|
IB_MTU_2048 = 4,
|
|
|
|
|
IB_MTU_4096 = 5
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
static inline int int_to_ibta_mtu(int mtu)
|
|
|
|
|
{
|
|
|
|
|
switch (mtu) {
|
|
|
|
|
case 256: return IB_MTU_256;
|
|
|
|
|
case 512: return IB_MTU_512;
|
|
|
|
|
case 1024: return IB_MTU_1024;
|
|
|
|
|
case 2048: return IB_MTU_2048;
|
|
|
|
|
case 4096: return IB_MTU_4096;
|
|
|
|
|
default: return -1;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static inline int ibta_mtu_to_int(enum ibta_mtu mtu)
|
|
|
|
|
{
|
|
|
|
|
switch (mtu) {
|
|
|
|
|
case IB_MTU_256: return 256;
|
|
|
|
|
case IB_MTU_512: return 512;
|
|
|
|
|
case IB_MTU_1024: return 1024;
|
|
|
|
|
case IB_MTU_2048: return 2048;
|
|
|
|
|
case IB_MTU_4096: return 4096;
|
|
|
|
|
default: return -1;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static ssize_t show_port_ib_mtu(struct device *dev,
|
|
|
|
|
struct device_attribute *attr,
|
|
|
|
|
char *buf)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_port_info *info = container_of(attr, struct mlx4_port_info,
|
|
|
|
|
port_mtu_attr);
|
|
|
|
|
struct mlx4_dev *mdev = info->dev;
|
|
|
|
|
|
|
|
|
|
if (mdev->caps.port_type[info->port] == MLX4_PORT_TYPE_ETH)
|
|
|
|
|
mlx4_warn(mdev, "port level mtu is only used for IB ports\n");
|
|
|
|
|
|
|
|
|
|
sprintf(buf, "%d\n",
|
|
|
|
|
ibta_mtu_to_int(mdev->caps.port_ib_mtu[info->port]));
|
|
|
|
|
return strlen(buf);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static ssize_t set_port_ib_mtu(struct device *dev,
|
|
|
|
|
struct device_attribute *attr,
|
|
|
|
|
const char *buf, size_t count)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_port_info *info = container_of(attr, struct mlx4_port_info,
|
|
|
|
|
port_mtu_attr);
|
|
|
|
|
struct mlx4_dev *mdev = info->dev;
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(mdev);
|
|
|
|
|
int err, port, mtu, ibta_mtu = -1;
|
|
|
|
|
|
|
|
|
|
if (mdev->caps.port_type[info->port] == MLX4_PORT_TYPE_ETH) {
|
|
|
|
|
mlx4_warn(mdev, "port level mtu is only used for IB ports\n");
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
}
|
|
|
|
|
|
2013-06-25 09:09:36 +00:00
|
|
|
err = kstrtoint(buf, 0, &mtu);
|
|
|
|
|
if (!err)
|
2012-01-11 17:02:17 +00:00
|
|
|
ibta_mtu = int_to_ibta_mtu(mtu);
|
|
|
|
|
|
2013-06-25 09:09:36 +00:00
|
|
|
if (err || ibta_mtu < 0) {
|
2012-01-11 17:02:17 +00:00
|
|
|
mlx4_err(mdev, "%s is invalid IBTA mtu\n", buf);
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
mdev->caps.port_ib_mtu[info->port] = ibta_mtu;
|
|
|
|
|
|
|
|
|
|
mlx4_stop_sense(mdev);
|
|
|
|
|
mutex_lock(&priv->port_mutex);
|
|
|
|
|
mlx4_unregister_device(mdev);
|
|
|
|
|
for (port = 1; port <= mdev->caps.num_ports; port++) {
|
|
|
|
|
mlx4_CLOSE_PORT(mdev, port);
|
mlx4: Put physical GID and P_Key table sizes in mlx4_phys_caps struct and paravirtualize them
To allow easy paravirtualization of P_Key and GID table sizes, keep
paravirtualized sizes in mlx4_dev->caps, but save the actual physical
sizes from FW in struct: mlx4_dev->phys_cap.
In addition, in SR-IOV mode, do the following:
1. Reduce reported P_Key table size by 1.
This is done to reserve the highest P_Key index for internal use,
for declaring an invalid P_Key in P_Key paravirtualization.
We require a P_Key index which always contain an invalid P_Key
value for this purpose (i.e., one which cannot be modified by
the subnet manager). The way to do this is to reduce the
P_Key table size reported to the subnet manager by 1, so that
it will not attempt to access the P_Key at index #127.
2. Paravirtualize the GID table size to 1. Thus, each guest sees
only a single GID (at its paravirtualized index 0).
In addition, since we are paravirtualizing the GID table size to 1, we
add paravirtualization of the master GID event here (i.e., we do not
do ib_dispatch_event() for the GUID change event on the master, since
its (only) GUID never changes).
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 08:21:44 +00:00
|
|
|
err = mlx4_SET_PORT(mdev, port, -1);
|
2012-01-11 17:02:17 +00:00
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(mdev, "Failed to set port %d, aborting\n",
|
|
|
|
|
port);
|
2012-01-11 17:02:17 +00:00
|
|
|
goto err_set_port;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
err = mlx4_register_device(mdev);
|
|
|
|
|
err_set_port:
|
|
|
|
|
mutex_unlock(&priv->port_mutex);
|
|
|
|
|
mlx4_start_sense(mdev);
|
|
|
|
|
return err ? err : count;
|
|
|
|
|
}
|
|
|
|
|
|
2015-12-06 16:07:43 +00:00
|
|
|
/* bond for multi-function device */
|
|
|
|
|
#define MAX_MF_BOND_ALLOWED_SLAVES 63
|
|
|
|
|
static int mlx4_mf_bond(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
int err = 0;
|
2016-03-02 15:47:45 +00:00
|
|
|
int nvfs;
|
2015-12-06 16:07:43 +00:00
|
|
|
struct mlx4_slaves_pport slaves_port1;
|
|
|
|
|
struct mlx4_slaves_pport slaves_port2;
|
|
|
|
|
DECLARE_BITMAP(slaves_port_1_2, MLX4_MFUNC_MAX);
|
|
|
|
|
|
|
|
|
|
slaves_port1 = mlx4_phys_to_slaves_pport(dev, 1);
|
|
|
|
|
slaves_port2 = mlx4_phys_to_slaves_pport(dev, 2);
|
|
|
|
|
bitmap_and(slaves_port_1_2,
|
|
|
|
|
slaves_port1.slaves, slaves_port2.slaves,
|
|
|
|
|
dev->persist->num_vfs + 1);
|
|
|
|
|
|
|
|
|
|
/* only single port vfs are allowed */
|
|
|
|
|
if (bitmap_weight(slaves_port_1_2, dev->persist->num_vfs + 1) > 1) {
|
|
|
|
|
mlx4_warn(dev, "HA mode unsupported for dual ported VFs\n");
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
}
|
|
|
|
|
|
2016-03-02 15:47:45 +00:00
|
|
|
/* number of virtual functions is number of total functions minus one
|
|
|
|
|
* physical function for each port.
|
|
|
|
|
*/
|
|
|
|
|
nvfs = bitmap_weight(slaves_port1.slaves, dev->persist->num_vfs + 1) +
|
|
|
|
|
bitmap_weight(slaves_port2.slaves, dev->persist->num_vfs + 1) - 2;
|
|
|
|
|
|
2015-12-06 16:07:43 +00:00
|
|
|
/* limit on maximum allowed VFs */
|
2016-03-02 15:47:45 +00:00
|
|
|
if (nvfs > MAX_MF_BOND_ALLOWED_SLAVES) {
|
|
|
|
|
mlx4_warn(dev, "HA mode is not supported for %d VFs (max %d are allowed)\n",
|
|
|
|
|
nvfs, MAX_MF_BOND_ALLOWED_SLAVES);
|
2015-12-06 16:07:43 +00:00
|
|
|
return -EINVAL;
|
2016-03-02 15:47:45 +00:00
|
|
|
}
|
2015-12-06 16:07:43 +00:00
|
|
|
|
|
|
|
|
if (dev->caps.steering_mode != MLX4_STEERING_MODE_DEVICE_MANAGED) {
|
|
|
|
|
mlx4_warn(dev, "HA mode unsupported for NON DMFS steering\n");
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_bond_mac_table(dev);
|
|
|
|
|
if (err)
|
|
|
|
|
return err;
|
|
|
|
|
err = mlx4_bond_vlan_table(dev);
|
|
|
|
|
if (err)
|
|
|
|
|
goto err1;
|
|
|
|
|
err = mlx4_bond_fs_rules(dev);
|
|
|
|
|
if (err)
|
|
|
|
|
goto err2;
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
err2:
|
|
|
|
|
(void)mlx4_unbond_vlan_table(dev);
|
|
|
|
|
err1:
|
|
|
|
|
(void)mlx4_unbond_mac_table(dev);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int mlx4_mf_unbond(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
int ret, ret1;
|
|
|
|
|
|
|
|
|
|
ret = mlx4_unbond_fs_rules(dev);
|
|
|
|
|
if (ret)
|
|
|
|
|
mlx4_warn(dev, "multifunction unbond for flow rules failedi (%d)\n", ret);
|
|
|
|
|
ret1 = mlx4_unbond_mac_table(dev);
|
|
|
|
|
if (ret1) {
|
|
|
|
|
mlx4_warn(dev, "multifunction unbond for MAC table failed (%d)\n", ret1);
|
|
|
|
|
ret = ret1;
|
|
|
|
|
}
|
|
|
|
|
ret1 = mlx4_unbond_vlan_table(dev);
|
|
|
|
|
if (ret1) {
|
|
|
|
|
mlx4_warn(dev, "multifunction unbond for VLAN table failed (%d)\n", ret1);
|
|
|
|
|
ret = ret1;
|
|
|
|
|
}
|
|
|
|
|
return ret;
|
|
|
|
|
}
|
|
|
|
|
|
2015-02-03 14:48:33 +00:00
|
|
|
int mlx4_bond(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
int ret = 0;
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
|
|
|
|
|
mutex_lock(&priv->bond_mutex);
|
|
|
|
|
|
2015-12-06 16:07:43 +00:00
|
|
|
if (!mlx4_is_bonded(dev)) {
|
2015-02-03 14:48:33 +00:00
|
|
|
ret = mlx4_do_bond(dev, true);
|
2015-12-06 16:07:43 +00:00
|
|
|
if (ret)
|
|
|
|
|
mlx4_err(dev, "Failed to bond device: %d\n", ret);
|
|
|
|
|
if (!ret && mlx4_is_master(dev)) {
|
|
|
|
|
ret = mlx4_mf_bond(dev);
|
|
|
|
|
if (ret) {
|
|
|
|
|
mlx4_err(dev, "bond for multifunction failed\n");
|
|
|
|
|
mlx4_do_bond(dev, false);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
2015-02-03 14:48:33 +00:00
|
|
|
|
|
|
|
|
mutex_unlock(&priv->bond_mutex);
|
2015-12-06 16:07:43 +00:00
|
|
|
if (!ret)
|
2015-02-03 14:48:33 +00:00
|
|
|
mlx4_dbg(dev, "Device is bonded\n");
|
2015-12-06 16:07:43 +00:00
|
|
|
|
2015-02-03 14:48:33 +00:00
|
|
|
return ret;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL_GPL(mlx4_bond);
|
|
|
|
|
|
|
|
|
|
int mlx4_unbond(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
int ret = 0;
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
|
|
|
|
|
mutex_lock(&priv->bond_mutex);
|
|
|
|
|
|
2015-12-06 16:07:43 +00:00
|
|
|
if (mlx4_is_bonded(dev)) {
|
|
|
|
|
int ret2 = 0;
|
|
|
|
|
|
2015-02-03 14:48:33 +00:00
|
|
|
ret = mlx4_do_bond(dev, false);
|
2015-12-06 16:07:43 +00:00
|
|
|
if (ret)
|
|
|
|
|
mlx4_err(dev, "Failed to unbond device: %d\n", ret);
|
|
|
|
|
if (mlx4_is_master(dev))
|
|
|
|
|
ret2 = mlx4_mf_unbond(dev);
|
|
|
|
|
if (ret2) {
|
|
|
|
|
mlx4_warn(dev, "Failed to unbond device for multifunction (%d)\n", ret2);
|
|
|
|
|
ret = ret2;
|
|
|
|
|
}
|
|
|
|
|
}
|
2015-02-03 14:48:33 +00:00
|
|
|
|
|
|
|
|
mutex_unlock(&priv->bond_mutex);
|
2015-12-06 16:07:43 +00:00
|
|
|
if (!ret)
|
2015-02-03 14:48:33 +00:00
|
|
|
mlx4_dbg(dev, "Device is unbonded\n");
|
2015-12-06 16:07:43 +00:00
|
|
|
|
2015-02-03 14:48:33 +00:00
|
|
|
return ret;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL_GPL(mlx4_unbond);
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
int mlx4_port_map_set(struct mlx4_dev *dev, struct mlx4_port_map *v2p)
|
|
|
|
|
{
|
|
|
|
|
u8 port1 = v2p->port1;
|
|
|
|
|
u8 port2 = v2p->port2;
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
if (!(dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_PORT_REMAP))
|
|
|
|
|
return -ENOTSUPP;
|
|
|
|
|
|
|
|
|
|
mutex_lock(&priv->bond_mutex);
|
|
|
|
|
|
|
|
|
|
/* zero means keep current mapping for this port */
|
|
|
|
|
if (port1 == 0)
|
|
|
|
|
port1 = priv->v2p.port1;
|
|
|
|
|
if (port2 == 0)
|
|
|
|
|
port2 = priv->v2p.port2;
|
|
|
|
|
|
|
|
|
|
if ((port1 < 1) || (port1 > MLX4_MAX_PORTS) ||
|
|
|
|
|
(port2 < 1) || (port2 > MLX4_MAX_PORTS) ||
|
|
|
|
|
(port1 == 2 && port2 == 1)) {
|
|
|
|
|
/* besides boundary checks cross mapping makes
|
|
|
|
|
* no sense and therefore not allowed */
|
|
|
|
|
err = -EINVAL;
|
|
|
|
|
} else if ((port1 == priv->v2p.port1) &&
|
|
|
|
|
(port2 == priv->v2p.port2)) {
|
|
|
|
|
err = 0;
|
|
|
|
|
} else {
|
|
|
|
|
err = mlx4_virt2phy_port_map(dev, port1, port2);
|
|
|
|
|
if (!err) {
|
|
|
|
|
mlx4_dbg(dev, "port map changed: [%d][%d]\n",
|
|
|
|
|
port1, port2);
|
|
|
|
|
priv->v2p.port1 = port1;
|
|
|
|
|
priv->v2p.port2 = port2;
|
|
|
|
|
} else {
|
|
|
|
|
mlx4_err(dev, "Failed to change port mape: %d\n", err);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
mutex_unlock(&priv->bond_mutex);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL_GPL(mlx4_port_map_set);
|
|
|
|
|
|
2008-02-05 04:20:41 +00:00
|
|
|
static int mlx4_load_fw(struct mlx4_dev *dev)
|
2007-05-09 01:00:38 +00:00
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
priv->fw.fw_icm = mlx4_alloc_icm(dev, priv->fw.fw_pages,
|
2007-08-01 09:28:20 +00:00
|
|
|
GFP_HIGHUSER | __GFP_NOWARN, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
if (!priv->fw.fw_icm) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Couldn't allocate FW area, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
return -ENOMEM;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_MAP_FA(dev, priv->fw.fw_icm);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "MAP_FA command failed, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_free;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_RUN_FW(dev);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "RUN_FW command failed, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_unmap_fa;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
|
|
err_unmap_fa:
|
|
|
|
|
mlx4_UNMAP_FA(dev);
|
|
|
|
|
|
|
|
|
|
err_free:
|
2007-08-01 09:28:20 +00:00
|
|
|
mlx4_free_icm(dev, priv->fw.fw_icm, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2008-02-05 04:20:41 +00:00
|
|
|
static int mlx4_init_cmpt_table(struct mlx4_dev *dev, u64 cmpt_base,
|
|
|
|
|
int cmpt_entry_sz)
|
2007-05-09 01:00:38 +00:00
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
int err;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
int num_eqs;
|
2007-05-09 01:00:38 +00:00
|
|
|
|
|
|
|
|
err = mlx4_init_icm_table(dev, &priv->qp_table.cmpt_table,
|
|
|
|
|
cmpt_base +
|
|
|
|
|
((u64) (MLX4_CMPT_TYPE_QP *
|
|
|
|
|
cmpt_entry_sz) << MLX4_CMPT_SHIFT),
|
|
|
|
|
cmpt_entry_sz, dev->caps.num_qps,
|
2008-10-22 17:25:29 +00:00
|
|
|
dev->caps.reserved_qps_cnt[MLX4_QP_REGION_FW],
|
|
|
|
|
0, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
if (err)
|
|
|
|
|
goto err;
|
|
|
|
|
|
|
|
|
|
err = mlx4_init_icm_table(dev, &priv->srq_table.cmpt_table,
|
|
|
|
|
cmpt_base +
|
|
|
|
|
((u64) (MLX4_CMPT_TYPE_SRQ *
|
|
|
|
|
cmpt_entry_sz) << MLX4_CMPT_SHIFT),
|
|
|
|
|
cmpt_entry_sz, dev->caps.num_srqs,
|
2007-08-01 09:28:20 +00:00
|
|
|
dev->caps.reserved_srqs, 0, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
if (err)
|
|
|
|
|
goto err_qp;
|
|
|
|
|
|
|
|
|
|
err = mlx4_init_icm_table(dev, &priv->cq_table.cmpt_table,
|
|
|
|
|
cmpt_base +
|
|
|
|
|
((u64) (MLX4_CMPT_TYPE_CQ *
|
|
|
|
|
cmpt_entry_sz) << MLX4_CMPT_SHIFT),
|
|
|
|
|
cmpt_entry_sz, dev->caps.num_cqs,
|
2007-08-01 09:28:20 +00:00
|
|
|
dev->caps.reserved_cqs, 0, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
if (err)
|
|
|
|
|
goto err_srq;
|
|
|
|
|
|
2014-11-13 12:45:32 +00:00
|
|
|
num_eqs = dev->phys_caps.num_phys_eqs;
|
2007-05-09 01:00:38 +00:00
|
|
|
err = mlx4_init_icm_table(dev, &priv->eq_table.cmpt_table,
|
|
|
|
|
cmpt_base +
|
|
|
|
|
((u64) (MLX4_CMPT_TYPE_EQ *
|
|
|
|
|
cmpt_entry_sz) << MLX4_CMPT_SHIFT),
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
cmpt_entry_sz, num_eqs, num_eqs, 0, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
if (err)
|
|
|
|
|
goto err_cq;
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
|
|
err_cq:
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->cq_table.cmpt_table);
|
|
|
|
|
|
|
|
|
|
err_srq:
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->srq_table.cmpt_table);
|
|
|
|
|
|
|
|
|
|
err_qp:
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->qp_table.cmpt_table);
|
|
|
|
|
|
|
|
|
|
err:
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2007-10-10 22:43:54 +00:00
|
|
|
static int mlx4_init_icm(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap,
|
|
|
|
|
struct mlx4_init_hca_param *init_hca, u64 icm_size)
|
2007-05-09 01:00:38 +00:00
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
u64 aux_pages;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
int num_eqs;
|
2007-05-09 01:00:38 +00:00
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
err = mlx4_SET_ICM_SIZE(dev, icm_size, &aux_pages);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "SET_ICM_SIZE command failed, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_dbg(dev, "%lld KB of HCA context requires %lld KB aux memory\n",
|
2007-05-09 01:00:38 +00:00
|
|
|
(unsigned long long) icm_size >> 10,
|
|
|
|
|
(unsigned long long) aux_pages << 2);
|
|
|
|
|
|
|
|
|
|
priv->fw.aux_icm = mlx4_alloc_icm(dev, aux_pages,
|
2007-08-01 09:28:20 +00:00
|
|
|
GFP_HIGHUSER | __GFP_NOWARN, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
if (!priv->fw.aux_icm) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Couldn't allocate aux memory, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
return -ENOMEM;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_MAP_ICM_AUX(dev, priv->fw.aux_icm);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "MAP_ICM_AUX command failed, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_free_aux;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_init_cmpt_table(dev, init_hca->cmpt_base, dev_cap->cmpt_entry_sz);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to map cMPT context memory, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_unmap_aux;
|
|
|
|
|
}
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
|
2014-11-13 12:45:32 +00:00
|
|
|
num_eqs = dev->phys_caps.num_phys_eqs;
|
2009-09-06 03:24:49 +00:00
|
|
|
err = mlx4_init_icm_table(dev, &priv->eq_table.table,
|
|
|
|
|
init_hca->eqc_base, dev_cap->eqc_entry_sz,
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
num_eqs, num_eqs, 0, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to map EQ context memory, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_unmap_cmpt;
|
|
|
|
|
}
|
|
|
|
|
|
2007-08-01 09:28:53 +00:00
|
|
|
/*
|
|
|
|
|
* Reserved MTT entries must be aligned up to a cacheline
|
|
|
|
|
* boundary, since the FW will write to them, while the driver
|
|
|
|
|
* writes to all other MTT entries. (The variable
|
|
|
|
|
* dev->caps.mtt_entry_sz below is really the MTT segment
|
|
|
|
|
* size, not the raw entry size)
|
|
|
|
|
*/
|
|
|
|
|
dev->caps.reserved_mtts =
|
|
|
|
|
ALIGN(dev->caps.reserved_mtts * dev->caps.mtt_entry_sz,
|
|
|
|
|
dma_get_cache_alignment()) / dev->caps.mtt_entry_sz;
|
|
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
err = mlx4_init_icm_table(dev, &priv->mr_table.mtt_table,
|
|
|
|
|
init_hca->mtt_base,
|
|
|
|
|
dev->caps.mtt_entry_sz,
|
2011-12-13 04:16:56 +00:00
|
|
|
dev->caps.num_mtts,
|
2007-08-01 09:28:20 +00:00
|
|
|
dev->caps.reserved_mtts, 1, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to map MTT context memory, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_unmap_eq;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_init_icm_table(dev, &priv->mr_table.dmpt_table,
|
|
|
|
|
init_hca->dmpt_base,
|
|
|
|
|
dev_cap->dmpt_entry_sz,
|
|
|
|
|
dev->caps.num_mpts,
|
2007-08-01 09:28:20 +00:00
|
|
|
dev->caps.reserved_mrws, 1, 1);
|
2007-05-09 01:00:38 +00:00
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to map dMPT context memory, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_unmap_mtt;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_init_icm_table(dev, &priv->qp_table.qp_table,
|
|
|
|
|
init_hca->qpc_base,
|
|
|
|
|
dev_cap->qpc_entry_sz,
|
|
|
|
|
dev->caps.num_qps,
|
2008-10-22 17:25:29 +00:00
|
|
|
dev->caps.reserved_qps_cnt[MLX4_QP_REGION_FW],
|
|
|
|
|
0, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to map QP context memory, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_unmap_dmpt;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_init_icm_table(dev, &priv->qp_table.auxc_table,
|
|
|
|
|
init_hca->auxc_base,
|
|
|
|
|
dev_cap->aux_entry_sz,
|
|
|
|
|
dev->caps.num_qps,
|
2008-10-22 17:25:29 +00:00
|
|
|
dev->caps.reserved_qps_cnt[MLX4_QP_REGION_FW],
|
|
|
|
|
0, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to map AUXC context memory, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_unmap_qp;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_init_icm_table(dev, &priv->qp_table.altc_table,
|
|
|
|
|
init_hca->altc_base,
|
|
|
|
|
dev_cap->altc_entry_sz,
|
|
|
|
|
dev->caps.num_qps,
|
2008-10-22 17:25:29 +00:00
|
|
|
dev->caps.reserved_qps_cnt[MLX4_QP_REGION_FW],
|
|
|
|
|
0, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to map ALTC context memory, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_unmap_auxc;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_init_icm_table(dev, &priv->qp_table.rdmarc_table,
|
|
|
|
|
init_hca->rdmarc_base,
|
|
|
|
|
dev_cap->rdmarc_entry_sz << priv->qp_table.rdmarc_shift,
|
|
|
|
|
dev->caps.num_qps,
|
2008-10-22 17:25:29 +00:00
|
|
|
dev->caps.reserved_qps_cnt[MLX4_QP_REGION_FW],
|
|
|
|
|
0, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "Failed to map RDMARC context memory, aborting\n");
|
|
|
|
|
goto err_unmap_altc;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_init_icm_table(dev, &priv->cq_table.table,
|
|
|
|
|
init_hca->cqc_base,
|
|
|
|
|
dev_cap->cqc_entry_sz,
|
|
|
|
|
dev->caps.num_cqs,
|
2007-08-01 09:28:20 +00:00
|
|
|
dev->caps.reserved_cqs, 0, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to map CQ context memory, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_unmap_rdmarc;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_init_icm_table(dev, &priv->srq_table.table,
|
|
|
|
|
init_hca->srqc_base,
|
|
|
|
|
dev_cap->srq_entry_sz,
|
|
|
|
|
dev->caps.num_srqs,
|
2007-08-01 09:28:20 +00:00
|
|
|
dev->caps.reserved_srqs, 0, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to map SRQ context memory, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_unmap_cq;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
{NET, IB}/mlx4: Add device managed flow steering firmware API
The driver is modified to support three operation modes.
If supported by firmware use the device managed flow steering
API, that which we call device managed steering mode. Else, if
the firmware supports the B0 steering mode use it, and finally,
if none of the above, use the A0 steering mode.
When the steering mode is device managed, the code is modified
such that L2 based rules set by the mlx4_en driver for Ethernet
unicast and multicast, and the IB stack multicast attach calls
done through the mlx4_ib driver are all routed to use the device
managed API.
When attaching rule using device managed flow steering API,
the firmware returns a 64 bit registration id, which is to be
provided during detach.
Currently the firmware is always programmed during HCA initialization
to use standard L2 hashing. Future work should be done to allow
configuring the flow-steering hash function with common, non
proprietary means.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-07-05 04:03:46 +00:00
|
|
|
* For flow steering device managed mode it is required to use
|
|
|
|
|
* mlx4_init_icm_table. For B0 steering mode it's not strictly
|
|
|
|
|
* required, but for simplicity just map the whole multicast
|
|
|
|
|
* group table now. The table isn't very big and it's a lot
|
|
|
|
|
* easier than trying to track ref counts.
|
2007-05-09 01:00:38 +00:00
|
|
|
*/
|
|
|
|
|
err = mlx4_init_icm_table(dev, &priv->mcg_table.table,
|
2011-12-13 04:16:02 +00:00
|
|
|
init_hca->mc_base,
|
|
|
|
|
mlx4_get_mgm_entry_size(dev),
|
2007-05-09 01:00:38 +00:00
|
|
|
dev->caps.num_mgms + dev->caps.num_amgms,
|
|
|
|
|
dev->caps.num_mgms + dev->caps.num_amgms,
|
2007-08-01 09:28:20 +00:00
|
|
|
0, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to map MCG context memory, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_unmap_srq;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
|
|
err_unmap_srq:
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->srq_table.table);
|
|
|
|
|
|
|
|
|
|
err_unmap_cq:
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->cq_table.table);
|
|
|
|
|
|
|
|
|
|
err_unmap_rdmarc:
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->qp_table.rdmarc_table);
|
|
|
|
|
|
|
|
|
|
err_unmap_altc:
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->qp_table.altc_table);
|
|
|
|
|
|
|
|
|
|
err_unmap_auxc:
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->qp_table.auxc_table);
|
|
|
|
|
|
|
|
|
|
err_unmap_qp:
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->qp_table.qp_table);
|
|
|
|
|
|
|
|
|
|
err_unmap_dmpt:
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->mr_table.dmpt_table);
|
|
|
|
|
|
|
|
|
|
err_unmap_mtt:
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->mr_table.mtt_table);
|
|
|
|
|
|
|
|
|
|
err_unmap_eq:
|
2009-09-06 03:24:49 +00:00
|
|
|
mlx4_cleanup_icm_table(dev, &priv->eq_table.table);
|
2007-05-09 01:00:38 +00:00
|
|
|
|
|
|
|
|
err_unmap_cmpt:
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->eq_table.cmpt_table);
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->cq_table.cmpt_table);
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->srq_table.cmpt_table);
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->qp_table.cmpt_table);
|
|
|
|
|
|
|
|
|
|
err_unmap_aux:
|
|
|
|
|
mlx4_UNMAP_ICM_AUX(dev);
|
|
|
|
|
|
|
|
|
|
err_free_aux:
|
2007-08-01 09:28:20 +00:00
|
|
|
mlx4_free_icm(dev, priv->fw.aux_icm, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void mlx4_free_icms(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->mcg_table.table);
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->srq_table.table);
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->cq_table.table);
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->qp_table.rdmarc_table);
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->qp_table.altc_table);
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->qp_table.auxc_table);
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->qp_table.qp_table);
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->mr_table.dmpt_table);
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->mr_table.mtt_table);
|
2009-09-06 03:24:49 +00:00
|
|
|
mlx4_cleanup_icm_table(dev, &priv->eq_table.table);
|
2007-05-09 01:00:38 +00:00
|
|
|
mlx4_cleanup_icm_table(dev, &priv->eq_table.cmpt_table);
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->cq_table.cmpt_table);
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->srq_table.cmpt_table);
|
|
|
|
|
mlx4_cleanup_icm_table(dev, &priv->qp_table.cmpt_table);
|
|
|
|
|
|
|
|
|
|
mlx4_UNMAP_ICM_AUX(dev);
|
2007-08-01 09:28:20 +00:00
|
|
|
mlx4_free_icm(dev, priv->fw.aux_icm, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
}
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
static void mlx4_slave_exit(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
|
mlx4_core: Fix crash on uninitialized priv->cmd.slave_sem
On an SR-IOV master device, __mlx4_init_one() calls mlx4_init_hca()
before mlx4_multi_func_init(). However, for unlucky configurations,
mlx4_init_hca() might call mlx4_SENSE_PORT() (via mlx4_dev_cap()), and
that calls mlx4_cmd_imm() with MLX4_CMD_WRAPPED set.
However, on a multifunction device with MLX4_CMD_WRAPPED, __mlx4_cmd()
calls into mlx4_slave_cmd(), and that immediately tries to do
down(&priv->cmd.slave_sem);
but priv->cmd.slave_sem isn't initialized until mlx4_multi_func_init()
(which we haven't called yet). The next thing it tries to do is access
priv->mfunc.vhcr, but that hasn't been allocated yet.
Fix this by moving the initialization of slave_sem and vhcr up into
mlx4_cmd_init(). Also, since slave_sem is really just being used as a
mutex, convert it into a slave_cmd_mutex.
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-09-26 04:24:07 +00:00
|
|
|
mutex_lock(&priv->cmd.slave_cmd_mutex);
|
2015-01-25 14:59:43 +00:00
|
|
|
if (mlx4_comm_cmd(dev, MLX4_COMM_CMD_RESET, 0, MLX4_COMM_CMD_NA_OP,
|
|
|
|
|
MLX4_COMM_TIME))
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_warn(dev, "Failed to close slave function\n");
|
mlx4_core: Fix crash on uninitialized priv->cmd.slave_sem
On an SR-IOV master device, __mlx4_init_one() calls mlx4_init_hca()
before mlx4_multi_func_init(). However, for unlucky configurations,
mlx4_init_hca() might call mlx4_SENSE_PORT() (via mlx4_dev_cap()), and
that calls mlx4_cmd_imm() with MLX4_CMD_WRAPPED set.
However, on a multifunction device with MLX4_CMD_WRAPPED, __mlx4_cmd()
calls into mlx4_slave_cmd(), and that immediately tries to do
down(&priv->cmd.slave_sem);
but priv->cmd.slave_sem isn't initialized until mlx4_multi_func_init()
(which we haven't called yet). The next thing it tries to do is access
priv->mfunc.vhcr, but that hasn't been allocated yet.
Fix this by moving the initialization of slave_sem and vhcr up into
mlx4_cmd_init(). Also, since slave_sem is really just being used as a
mutex, convert it into a slave_cmd_mutex.
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-09-26 04:24:07 +00:00
|
|
|
mutex_unlock(&priv->cmd.slave_cmd_mutex);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
}
|
|
|
|
|
|
2011-03-22 22:38:41 +00:00
|
|
|
static int map_bf_area(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
resource_size_t bf_start;
|
|
|
|
|
resource_size_t bf_len;
|
|
|
|
|
int err = 0;
|
|
|
|
|
|
2012-02-19 21:38:52 +00:00
|
|
|
if (!dev->caps.bf_reg_size)
|
|
|
|
|
return -ENXIO;
|
|
|
|
|
|
2015-01-25 14:59:35 +00:00
|
|
|
bf_start = pci_resource_start(dev->persist->pdev, 2) +
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
(dev->caps.num_uars << PAGE_SHIFT);
|
2015-01-25 14:59:35 +00:00
|
|
|
bf_len = pci_resource_len(dev->persist->pdev, 2) -
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
(dev->caps.num_uars << PAGE_SHIFT);
|
2011-03-22 22:38:41 +00:00
|
|
|
priv->bf_mapping = io_mapping_create_wc(bf_start, bf_len);
|
|
|
|
|
if (!priv->bf_mapping)
|
|
|
|
|
err = -ENOMEM;
|
|
|
|
|
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void unmap_bf_area(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
if (mlx4_priv(dev)->bf_mapping)
|
|
|
|
|
io_mapping_free(mlx4_priv(dev)->bf_mapping);
|
|
|
|
|
}
|
|
|
|
|
|
2013-04-23 06:06:49 +00:00
|
|
|
cycle_t mlx4_read_clock(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
u32 clockhi, clocklo, clockhi1;
|
|
|
|
|
cycle_t cycles;
|
|
|
|
|
int i;
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < 10; i++) {
|
|
|
|
|
clockhi = swab32(readl(priv->clock_mapping));
|
|
|
|
|
clocklo = swab32(readl(priv->clock_mapping + 4));
|
|
|
|
|
clockhi1 = swab32(readl(priv->clock_mapping));
|
|
|
|
|
if (clockhi == clockhi1)
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
cycles = (u64) clockhi << 32 | (u64) clocklo;
|
|
|
|
|
|
|
|
|
|
return cycles;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL_GPL(mlx4_read_clock);
|
|
|
|
|
|
|
|
|
|
|
2013-04-23 06:06:48 +00:00
|
|
|
static int map_internal_clock(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
|
|
|
|
|
priv->clock_mapping =
|
2015-01-25 14:59:35 +00:00
|
|
|
ioremap(pci_resource_start(dev->persist->pdev,
|
|
|
|
|
priv->fw.clock_bar) +
|
2013-04-23 06:06:48 +00:00
|
|
|
priv->fw.clock_offset, MLX4_CLOCK_SIZE);
|
|
|
|
|
|
|
|
|
|
if (!priv->clock_mapping)
|
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
2015-06-11 13:35:26 +00:00
|
|
|
int mlx4_get_internal_clock_params(struct mlx4_dev *dev,
|
|
|
|
|
struct mlx4_clock_params *params)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
|
|
|
|
|
if (mlx4_is_slave(dev))
|
|
|
|
|
return -ENOTSUPP;
|
|
|
|
|
|
|
|
|
|
if (!params)
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
|
|
params->bar = priv->fw.clock_bar;
|
|
|
|
|
params->offset = priv->fw.clock_offset;
|
|
|
|
|
params->size = MLX4_CLOCK_SIZE;
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL_GPL(mlx4_get_internal_clock_params);
|
|
|
|
|
|
2013-04-23 06:06:48 +00:00
|
|
|
static void unmap_internal_clock(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
|
|
|
|
|
if (priv->clock_mapping)
|
|
|
|
|
iounmap(priv->clock_mapping);
|
|
|
|
|
}
|
|
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
static void mlx4_close_hca(struct mlx4_dev *dev)
|
|
|
|
|
{
|
2013-04-23 06:06:48 +00:00
|
|
|
unmap_internal_clock(dev);
|
2011-03-22 22:38:41 +00:00
|
|
|
unmap_bf_area(dev);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
if (mlx4_is_slave(dev))
|
|
|
|
|
mlx4_slave_exit(dev);
|
|
|
|
|
else {
|
|
|
|
|
mlx4_CLOSE_HCA(dev, 0);
|
|
|
|
|
mlx4_free_icms(dev);
|
2014-11-13 12:45:30 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void mlx4_close_fw(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
if (!mlx4_is_slave(dev)) {
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
mlx4_UNMAP_FA(dev);
|
|
|
|
|
mlx4_free_icm(dev, mlx4_priv(dev)->fw.fw_icm, 0);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2015-01-25 14:59:42 +00:00
|
|
|
static int mlx4_comm_check_offline(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
#define COMM_CHAN_OFFLINE_OFFSET 0x09
|
|
|
|
|
|
|
|
|
|
u32 comm_flags;
|
|
|
|
|
u32 offline_bit;
|
|
|
|
|
unsigned long end;
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
|
|
|
|
|
end = msecs_to_jiffies(MLX4_COMM_OFFLINE_TIME_OUT) + jiffies;
|
|
|
|
|
while (time_before(jiffies, end)) {
|
|
|
|
|
comm_flags = swab32(readl((__iomem char *)priv->mfunc.comm +
|
|
|
|
|
MLX4_COMM_CHAN_FLAGS));
|
|
|
|
|
offline_bit = (comm_flags &
|
|
|
|
|
(u32)(1 << COMM_CHAN_OFFLINE_OFFSET));
|
|
|
|
|
if (!offline_bit)
|
|
|
|
|
return 0;
|
|
|
|
|
/* There are cases as part of AER/Reset flow that PF needs
|
|
|
|
|
* around 100 msec to load. We therefore sleep for 100 msec
|
|
|
|
|
* to allow other tasks to make use of that CPU during this
|
|
|
|
|
* time interval.
|
|
|
|
|
*/
|
|
|
|
|
msleep(100);
|
|
|
|
|
}
|
|
|
|
|
mlx4_err(dev, "Communication channel is offline.\n");
|
|
|
|
|
return -EIO;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void mlx4_reset_vf_support(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
#define COMM_CHAN_RST_OFFSET 0x1e
|
|
|
|
|
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
u32 comm_rst;
|
|
|
|
|
u32 comm_caps;
|
|
|
|
|
|
|
|
|
|
comm_caps = swab32(readl((__iomem char *)priv->mfunc.comm +
|
|
|
|
|
MLX4_COMM_CHAN_CAPS));
|
|
|
|
|
comm_rst = (comm_caps & (u32)(1 << COMM_CHAN_RST_OFFSET));
|
|
|
|
|
|
|
|
|
|
if (comm_rst)
|
|
|
|
|
dev->caps.vf_caps |= MLX4_VF_CAP_FLAG_RESET;
|
|
|
|
|
}
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
static int mlx4_init_slave(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
u64 dma = (u64) priv->mfunc.vhcr_dma;
|
|
|
|
|
int ret_from_reset = 0;
|
|
|
|
|
u32 slave_read;
|
|
|
|
|
u32 cmd_channel_ver;
|
|
|
|
|
|
net/mlx4_core: mlx4_init_slave() shouldn't access comm channel before PF is ready
Currently, the PF call to pci_enable_sriov from the PF probe function
stalls for 10 seconds times the number of VFs probed on the host. This
happens because the way for such VFs to determine of the PF
initialization finished, is by attempting to issue reset on the
comm-channel and get timeout (after 10s).
The PF probe function is called from a kenernel workqueue, and therefore
during that time, rcu lock is being held and kernel's workqueue is
stalled. This blocks other processes that try to use the workqueue
or rcu lock. For example, interface renaming which is calling
rcu_synchronize is blocked, and timedout by systemd.
Changed mlx4_init_slave() to allow VF probed on the host to immediatly
detect that the PF is not ready, and return EPROBE_DEFER instantly.
Only when the PF finishes the initialization, allow such VFs to
access the comm channel.
This issue and fix are relevant only for probed VFs on the hypervisor,
there is no way to pass this information to a VM until comm channel is
ready, so in a VM, if PF is not ready, the first command will be timedout
after 10 seconds and return EPROBE_DEFER.
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-03-06 16:28:17 +00:00
|
|
|
if (atomic_read(&pf_loading)) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_warn(dev, "PF is not ready - Deferring probe\n");
|
net/mlx4_core: mlx4_init_slave() shouldn't access comm channel before PF is ready
Currently, the PF call to pci_enable_sriov from the PF probe function
stalls for 10 seconds times the number of VFs probed on the host. This
happens because the way for such VFs to determine of the PF
initialization finished, is by attempting to issue reset on the
comm-channel and get timeout (after 10s).
The PF probe function is called from a kenernel workqueue, and therefore
during that time, rcu lock is being held and kernel's workqueue is
stalled. This blocks other processes that try to use the workqueue
or rcu lock. For example, interface renaming which is calling
rcu_synchronize is blocked, and timedout by systemd.
Changed mlx4_init_slave() to allow VF probed on the host to immediatly
detect that the PF is not ready, and return EPROBE_DEFER instantly.
Only when the PF finishes the initialization, allow such VFs to
access the comm channel.
This issue and fix are relevant only for probed VFs on the hypervisor,
there is no way to pass this information to a VM until comm channel is
ready, so in a VM, if PF is not ready, the first command will be timedout
after 10 seconds and return EPROBE_DEFER.
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-03-06 16:28:17 +00:00
|
|
|
return -EPROBE_DEFER;
|
|
|
|
|
}
|
|
|
|
|
|
mlx4_core: Fix crash on uninitialized priv->cmd.slave_sem
On an SR-IOV master device, __mlx4_init_one() calls mlx4_init_hca()
before mlx4_multi_func_init(). However, for unlucky configurations,
mlx4_init_hca() might call mlx4_SENSE_PORT() (via mlx4_dev_cap()), and
that calls mlx4_cmd_imm() with MLX4_CMD_WRAPPED set.
However, on a multifunction device with MLX4_CMD_WRAPPED, __mlx4_cmd()
calls into mlx4_slave_cmd(), and that immediately tries to do
down(&priv->cmd.slave_sem);
but priv->cmd.slave_sem isn't initialized until mlx4_multi_func_init()
(which we haven't called yet). The next thing it tries to do is access
priv->mfunc.vhcr, but that hasn't been allocated yet.
Fix this by moving the initialization of slave_sem and vhcr up into
mlx4_cmd_init(). Also, since slave_sem is really just being used as a
mutex, convert it into a slave_cmd_mutex.
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-09-26 04:24:07 +00:00
|
|
|
mutex_lock(&priv->cmd.slave_cmd_mutex);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
priv->cmd.max_cmds = 1;
|
2015-01-25 14:59:42 +00:00
|
|
|
if (mlx4_comm_check_offline(dev)) {
|
|
|
|
|
mlx4_err(dev, "PF is not responsive, skipping initialization\n");
|
|
|
|
|
goto err_offline;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
mlx4_reset_vf_support(dev);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
mlx4_warn(dev, "Sending reset\n");
|
|
|
|
|
ret_from_reset = mlx4_comm_cmd(dev, MLX4_COMM_CMD_RESET, 0,
|
2015-01-25 14:59:43 +00:00
|
|
|
MLX4_COMM_CMD_NA_OP, MLX4_COMM_TIME);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/* if we are in the middle of flr the slave will try
|
|
|
|
|
* NUM_OF_RESET_RETRIES times before leaving.*/
|
|
|
|
|
if (ret_from_reset) {
|
|
|
|
|
if (MLX4_DELAY_RESET_SLAVE == ret_from_reset) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_warn(dev, "slave is currently in the middle of FLR - Deferring probe\n");
|
2013-06-04 05:13:27 +00:00
|
|
|
mutex_unlock(&priv->cmd.slave_cmd_mutex);
|
|
|
|
|
return -EPROBE_DEFER;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
} else
|
|
|
|
|
goto err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* check the driver version - the slave I/F revision
|
|
|
|
|
* must match the master's */
|
|
|
|
|
slave_read = swab32(readl(&priv->mfunc.comm->slave_read));
|
|
|
|
|
cmd_channel_ver = mlx4_comm_get_version();
|
|
|
|
|
|
|
|
|
|
if (MLX4_COMM_GET_IF_REV(cmd_channel_ver) !=
|
|
|
|
|
MLX4_COMM_GET_IF_REV(slave_read)) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "slave driver version is not supported by the master\n");
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
goto err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
mlx4_warn(dev, "Sending vhcr0\n");
|
|
|
|
|
if (mlx4_comm_cmd(dev, MLX4_COMM_CMD_VHCR0, dma >> 48,
|
2015-01-25 14:59:43 +00:00
|
|
|
MLX4_COMM_CMD_NA_OP, MLX4_COMM_TIME))
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
goto err;
|
|
|
|
|
if (mlx4_comm_cmd(dev, MLX4_COMM_CMD_VHCR1, dma >> 32,
|
2015-01-25 14:59:43 +00:00
|
|
|
MLX4_COMM_CMD_NA_OP, MLX4_COMM_TIME))
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
goto err;
|
|
|
|
|
if (mlx4_comm_cmd(dev, MLX4_COMM_CMD_VHCR2, dma >> 16,
|
2015-01-25 14:59:43 +00:00
|
|
|
MLX4_COMM_CMD_NA_OP, MLX4_COMM_TIME))
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
goto err;
|
2015-01-25 14:59:43 +00:00
|
|
|
if (mlx4_comm_cmd(dev, MLX4_COMM_CMD_VHCR_EN, dma,
|
|
|
|
|
MLX4_COMM_CMD_NA_OP, MLX4_COMM_TIME))
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
goto err;
|
mlx4_core: Fix crash on uninitialized priv->cmd.slave_sem
On an SR-IOV master device, __mlx4_init_one() calls mlx4_init_hca()
before mlx4_multi_func_init(). However, for unlucky configurations,
mlx4_init_hca() might call mlx4_SENSE_PORT() (via mlx4_dev_cap()), and
that calls mlx4_cmd_imm() with MLX4_CMD_WRAPPED set.
However, on a multifunction device with MLX4_CMD_WRAPPED, __mlx4_cmd()
calls into mlx4_slave_cmd(), and that immediately tries to do
down(&priv->cmd.slave_sem);
but priv->cmd.slave_sem isn't initialized until mlx4_multi_func_init()
(which we haven't called yet). The next thing it tries to do is access
priv->mfunc.vhcr, but that hasn't been allocated yet.
Fix this by moving the initialization of slave_sem and vhcr up into
mlx4_cmd_init(). Also, since slave_sem is really just being used as a
mutex, convert it into a slave_cmd_mutex.
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-09-26 04:24:07 +00:00
|
|
|
|
|
|
|
|
mutex_unlock(&priv->cmd.slave_cmd_mutex);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
|
|
err:
|
2015-01-25 14:59:43 +00:00
|
|
|
mlx4_comm_cmd(dev, MLX4_COMM_CMD_RESET, 0, MLX4_COMM_CMD_NA_OP, 0);
|
2015-01-25 14:59:42 +00:00
|
|
|
err_offline:
|
mlx4_core: Fix crash on uninitialized priv->cmd.slave_sem
On an SR-IOV master device, __mlx4_init_one() calls mlx4_init_hca()
before mlx4_multi_func_init(). However, for unlucky configurations,
mlx4_init_hca() might call mlx4_SENSE_PORT() (via mlx4_dev_cap()), and
that calls mlx4_cmd_imm() with MLX4_CMD_WRAPPED set.
However, on a multifunction device with MLX4_CMD_WRAPPED, __mlx4_cmd()
calls into mlx4_slave_cmd(), and that immediately tries to do
down(&priv->cmd.slave_sem);
but priv->cmd.slave_sem isn't initialized until mlx4_multi_func_init()
(which we haven't called yet). The next thing it tries to do is access
priv->mfunc.vhcr, but that hasn't been allocated yet.
Fix this by moving the initialization of slave_sem and vhcr up into
mlx4_cmd_init(). Also, since slave_sem is really just being used as a
mutex, convert it into a slave_cmd_mutex.
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-09-26 04:24:07 +00:00
|
|
|
mutex_unlock(&priv->cmd.slave_cmd_mutex);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
return -EIO;
|
2007-05-09 01:00:38 +00:00
|
|
|
}
|
|
|
|
|
|
mlx4: Put physical GID and P_Key table sizes in mlx4_phys_caps struct and paravirtualize them
To allow easy paravirtualization of P_Key and GID table sizes, keep
paravirtualized sizes in mlx4_dev->caps, but save the actual physical
sizes from FW in struct: mlx4_dev->phys_cap.
In addition, in SR-IOV mode, do the following:
1. Reduce reported P_Key table size by 1.
This is done to reserve the highest P_Key index for internal use,
for declaring an invalid P_Key in P_Key paravirtualization.
We require a P_Key index which always contain an invalid P_Key
value for this purpose (i.e., one which cannot be modified by
the subnet manager). The way to do this is to reduce the
P_Key table size reported to the subnet manager by 1, so that
it will not attempt to access the P_Key at index #127.
2. Paravirtualize the GID table size to 1. Thus, each guest sees
only a single GID (at its paravirtualized index 0).
In addition, since we are paravirtualizing the GID table size to 1, we
add paravirtualization of the master GID event here (i.e., we do not
do ib_dispatch_event() for the GUID change event on the master, since
its (only) GUID never changes).
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 08:21:44 +00:00
|
|
|
static void mlx4_parav_master_pf_caps(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
int i;
|
|
|
|
|
|
|
|
|
|
for (i = 1; i <= dev->caps.num_ports; i++) {
|
2014-03-12 10:00:39 +00:00
|
|
|
if (dev->caps.port_type[i] == MLX4_PORT_TYPE_ETH)
|
|
|
|
|
dev->caps.gid_table_len[i] =
|
2014-03-19 16:11:52 +00:00
|
|
|
mlx4_get_slave_num_gids(dev, 0, i);
|
2014-03-12 10:00:39 +00:00
|
|
|
else
|
|
|
|
|
dev->caps.gid_table_len[i] = 1;
|
mlx4: Put physical GID and P_Key table sizes in mlx4_phys_caps struct and paravirtualize them
To allow easy paravirtualization of P_Key and GID table sizes, keep
paravirtualized sizes in mlx4_dev->caps, but save the actual physical
sizes from FW in struct: mlx4_dev->phys_cap.
In addition, in SR-IOV mode, do the following:
1. Reduce reported P_Key table size by 1.
This is done to reserve the highest P_Key index for internal use,
for declaring an invalid P_Key in P_Key paravirtualization.
We require a P_Key index which always contain an invalid P_Key
value for this purpose (i.e., one which cannot be modified by
the subnet manager). The way to do this is to reduce the
P_Key table size reported to the subnet manager by 1, so that
it will not attempt to access the P_Key at index #127.
2. Paravirtualize the GID table size to 1. Thus, each guest sees
only a single GID (at its paravirtualized index 0).
In addition, since we are paravirtualizing the GID table size to 1, we
add paravirtualization of the master GID event here (i.e., we do not
do ib_dispatch_event() for the GUID change event on the master, since
its (only) GUID never changes).
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 08:21:44 +00:00
|
|
|
dev->caps.pkey_table_len[i] =
|
|
|
|
|
dev->phys_caps.pkey_phys_table_len[i] - 1;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2012-12-06 17:12:00 +00:00
|
|
|
static int choose_log_fs_mgm_entry_size(int qp_per_entry)
|
|
|
|
|
{
|
|
|
|
|
int i = MLX4_MIN_MGM_LOG_ENTRY_SIZE;
|
|
|
|
|
|
|
|
|
|
for (i = MLX4_MIN_MGM_LOG_ENTRY_SIZE; i <= MLX4_MAX_MGM_LOG_ENTRY_SIZE;
|
|
|
|
|
i++) {
|
|
|
|
|
if (qp_per_entry <= 4 * ((1 << i) / 16 - 2))
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return (i <= MLX4_MAX_MGM_LOG_ENTRY_SIZE) ? i : -1;
|
|
|
|
|
}
|
|
|
|
|
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 08:58:00 +00:00
|
|
|
static const char *dmfs_high_rate_steering_mode_str(int dmfs_high_steer_mode)
|
|
|
|
|
{
|
|
|
|
|
switch (dmfs_high_steer_mode) {
|
|
|
|
|
case MLX4_STEERING_DMFS_A0_DEFAULT:
|
|
|
|
|
return "default performance";
|
|
|
|
|
|
|
|
|
|
case MLX4_STEERING_DMFS_A0_DYNAMIC:
|
|
|
|
|
return "dynamic hybrid mode";
|
|
|
|
|
|
|
|
|
|
case MLX4_STEERING_DMFS_A0_STATIC:
|
|
|
|
|
return "performance optimized for limited rule configuration (static)";
|
|
|
|
|
|
|
|
|
|
case MLX4_STEERING_DMFS_A0_DISABLE:
|
|
|
|
|
return "disabled performance optimized steering";
|
|
|
|
|
|
|
|
|
|
case MLX4_STEERING_DMFS_A0_NOT_SUPPORTED:
|
|
|
|
|
return "performance optimized steering not supported";
|
|
|
|
|
|
|
|
|
|
default:
|
|
|
|
|
return "Unrecognized mode";
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#define MLX4_DMFS_A0_STEERING (1UL << 2)
|
|
|
|
|
|
2012-12-06 17:11:59 +00:00
|
|
|
static void choose_steering_mode(struct mlx4_dev *dev,
|
|
|
|
|
struct mlx4_dev_cap *dev_cap)
|
|
|
|
|
{
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 08:58:00 +00:00
|
|
|
if (mlx4_log_num_mgm_entry_size <= 0) {
|
|
|
|
|
if ((-mlx4_log_num_mgm_entry_size) & MLX4_DMFS_A0_STEERING) {
|
|
|
|
|
if (dev->caps.dmfs_high_steer_mode ==
|
|
|
|
|
MLX4_STEERING_DMFS_A0_NOT_SUPPORTED)
|
|
|
|
|
mlx4_err(dev, "DMFS high rate mode not supported\n");
|
|
|
|
|
else
|
|
|
|
|
dev->caps.dmfs_high_steer_mode =
|
|
|
|
|
MLX4_STEERING_DMFS_A0_STATIC;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (mlx4_log_num_mgm_entry_size <= 0 &&
|
2012-12-06 17:12:00 +00:00
|
|
|
dev_cap->flags2 & MLX4_DEV_CAP_FLAG2_FS_EN &&
|
2012-12-06 17:11:59 +00:00
|
|
|
(!mlx4_is_mfunc(dev) ||
|
2015-01-25 14:59:35 +00:00
|
|
|
(dev_cap->fs_max_num_qp_per_entry >=
|
|
|
|
|
(dev->persist->num_vfs + 1))) &&
|
2012-12-06 17:12:00 +00:00
|
|
|
choose_log_fs_mgm_entry_size(dev_cap->fs_max_num_qp_per_entry) >=
|
|
|
|
|
MLX4_MIN_MGM_LOG_ENTRY_SIZE) {
|
|
|
|
|
dev->oper_log_mgm_entry_size =
|
|
|
|
|
choose_log_fs_mgm_entry_size(dev_cap->fs_max_num_qp_per_entry);
|
2012-12-06 17:11:59 +00:00
|
|
|
dev->caps.steering_mode = MLX4_STEERING_MODE_DEVICE_MANAGED;
|
|
|
|
|
dev->caps.num_qp_per_mgm = dev_cap->fs_max_num_qp_per_entry;
|
|
|
|
|
dev->caps.fs_log_max_ucast_qp_range_size =
|
|
|
|
|
dev_cap->fs_log_max_ucast_qp_range_size;
|
|
|
|
|
} else {
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 08:58:00 +00:00
|
|
|
if (dev->caps.dmfs_high_steer_mode !=
|
|
|
|
|
MLX4_STEERING_DMFS_A0_NOT_SUPPORTED)
|
|
|
|
|
dev->caps.dmfs_high_steer_mode = MLX4_STEERING_DMFS_A0_DISABLE;
|
2012-12-06 17:11:59 +00:00
|
|
|
if (dev->caps.flags & MLX4_DEV_CAP_FLAG_VEP_UC_STEER &&
|
|
|
|
|
dev->caps.flags & MLX4_DEV_CAP_FLAG_VEP_MC_STEER)
|
|
|
|
|
dev->caps.steering_mode = MLX4_STEERING_MODE_B0;
|
|
|
|
|
else {
|
|
|
|
|
dev->caps.steering_mode = MLX4_STEERING_MODE_A0;
|
|
|
|
|
|
|
|
|
|
if (dev->caps.flags & MLX4_DEV_CAP_FLAG_VEP_UC_STEER ||
|
|
|
|
|
dev->caps.flags & MLX4_DEV_CAP_FLAG_VEP_MC_STEER)
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_warn(dev, "Must have both UC_STEER and MC_STEER flags set to use B0 steering - falling back to A0 steering mode\n");
|
2012-12-06 17:11:59 +00:00
|
|
|
}
|
2012-12-06 17:12:00 +00:00
|
|
|
dev->oper_log_mgm_entry_size =
|
|
|
|
|
mlx4_log_num_mgm_entry_size > 0 ?
|
|
|
|
|
mlx4_log_num_mgm_entry_size :
|
|
|
|
|
MLX4_DEFAULT_MGM_LOG_ENTRY_SIZE;
|
2012-12-06 17:11:59 +00:00
|
|
|
dev->caps.num_qp_per_mgm = mlx4_get_qp_per_mgm(dev);
|
|
|
|
|
}
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_dbg(dev, "Steering mode is: %s, oper_log_mgm_entry_size = %d, modparam log_num_mgm_entry_size = %d\n",
|
2012-12-06 17:12:00 +00:00
|
|
|
mlx4_steering_mode_str(dev->caps.steering_mode),
|
|
|
|
|
dev->oper_log_mgm_entry_size,
|
|
|
|
|
mlx4_log_num_mgm_entry_size);
|
2012-12-06 17:11:59 +00:00
|
|
|
}
|
|
|
|
|
|
2013-12-23 14:09:43 +00:00
|
|
|
static void choose_tunnel_offload_mode(struct mlx4_dev *dev,
|
|
|
|
|
struct mlx4_dev_cap *dev_cap)
|
|
|
|
|
{
|
|
|
|
|
if (dev->caps.steering_mode == MLX4_STEERING_MODE_DEVICE_MANAGED &&
|
2015-01-15 13:28:54 +00:00
|
|
|
dev_cap->flags2 & MLX4_DEV_CAP_FLAG2_VXLAN_OFFLOADS)
|
2013-12-23 14:09:43 +00:00
|
|
|
dev->caps.tunnel_offload_mode = MLX4_TUNNEL_OFFLOAD_MODE_VXLAN;
|
|
|
|
|
else
|
|
|
|
|
dev->caps.tunnel_offload_mode = MLX4_TUNNEL_OFFLOAD_MODE_NONE;
|
|
|
|
|
|
|
|
|
|
mlx4_dbg(dev, "Tunneling offload mode is: %s\n", (dev->caps.tunnel_offload_mode
|
|
|
|
|
== MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) ? "vxlan" : "none");
|
|
|
|
|
}
|
|
|
|
|
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 08:58:00 +00:00
|
|
|
static int mlx4_validate_optimized_steering(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
int i;
|
|
|
|
|
struct mlx4_port_cap port_cap;
|
|
|
|
|
|
|
|
|
|
if (dev->caps.dmfs_high_steer_mode == MLX4_STEERING_DMFS_A0_NOT_SUPPORTED)
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
|
|
for (i = 1; i <= dev->caps.num_ports; i++) {
|
|
|
|
|
if (mlx4_dev_port(dev, i, &port_cap)) {
|
|
|
|
|
mlx4_err(dev,
|
|
|
|
|
"QUERY_DEV_CAP command failed, can't veify DMFS high rate steering.\n");
|
|
|
|
|
} else if ((dev->caps.dmfs_high_steer_mode !=
|
|
|
|
|
MLX4_STEERING_DMFS_A0_DEFAULT) &&
|
|
|
|
|
(port_cap.dmfs_optimized_state ==
|
|
|
|
|
!!(dev->caps.dmfs_high_steer_mode ==
|
|
|
|
|
MLX4_STEERING_DMFS_A0_DISABLE))) {
|
|
|
|
|
mlx4_err(dev,
|
|
|
|
|
"DMFS high rate steer mode differ, driver requested %s but %s in FW.\n",
|
|
|
|
|
dmfs_high_rate_steering_mode_str(
|
|
|
|
|
dev->caps.dmfs_high_steer_mode),
|
|
|
|
|
(port_cap.dmfs_optimized_state ?
|
|
|
|
|
"enabled" : "disabled"));
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
2014-11-13 12:45:30 +00:00
|
|
|
static int mlx4_init_fw(struct mlx4_dev *dev)
|
2007-05-09 01:00:38 +00:00
|
|
|
{
|
2008-07-15 06:48:53 +00:00
|
|
|
struct mlx4_mod_stat_cfg mlx4_cfg;
|
2014-11-13 12:45:30 +00:00
|
|
|
int err = 0;
|
2007-05-09 01:00:38 +00:00
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
if (!mlx4_is_slave(dev)) {
|
|
|
|
|
err = mlx4_QUERY_FW(dev);
|
|
|
|
|
if (err) {
|
|
|
|
|
if (err == -EACCES)
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_info(dev, "non-primary physical function, skipping\n");
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
else
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "QUERY_FW command failed, aborting\n");
|
2012-09-05 22:50:51 +00:00
|
|
|
return err;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
}
|
2007-05-09 01:00:38 +00:00
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
err = mlx4_load_fw(dev);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to start FW, aborting\n");
|
2012-09-05 22:50:51 +00:00
|
|
|
return err;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
}
|
2007-05-09 01:00:38 +00:00
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
mlx4_cfg.log_pg_sz_m = 1;
|
|
|
|
|
mlx4_cfg.log_pg_sz = 0;
|
|
|
|
|
err = mlx4_MOD_STAT_CFG(dev, &mlx4_cfg);
|
|
|
|
|
if (err)
|
|
|
|
|
mlx4_warn(dev, "Failed to override log_pg_sz parameter\n");
|
2014-11-13 12:45:30 +00:00
|
|
|
}
|
2008-07-15 06:48:53 +00:00
|
|
|
|
2014-11-13 12:45:30 +00:00
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int mlx4_init_hca(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
struct mlx4_adapter adapter;
|
|
|
|
|
struct mlx4_dev_cap dev_cap;
|
|
|
|
|
struct mlx4_profile profile;
|
|
|
|
|
struct mlx4_init_hca_param init_hca;
|
|
|
|
|
u64 icm_size;
|
|
|
|
|
struct mlx4_config_dev_params params;
|
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
if (!mlx4_is_slave(dev)) {
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
err = mlx4_dev_cap(dev, &dev_cap);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "QUERY_DEV_CAP command failed, aborting\n");
|
2014-12-30 09:59:50 +00:00
|
|
|
return err;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
}
|
2007-05-09 01:00:38 +00:00
|
|
|
|
2012-12-06 17:11:59 +00:00
|
|
|
choose_steering_mode(dev, &dev_cap);
|
2013-12-23 14:09:43 +00:00
|
|
|
choose_tunnel_offload_mode(dev, &dev_cap);
|
2012-12-06 17:11:59 +00:00
|
|
|
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 08:58:00 +00:00
|
|
|
if (dev->caps.dmfs_high_steer_mode == MLX4_STEERING_DMFS_A0_STATIC &&
|
|
|
|
|
mlx4_is_master(dev))
|
|
|
|
|
dev->caps.function_caps |= MLX4_FUNC_CAP_DMFS_A0_STATIC;
|
|
|
|
|
|
2013-12-19 19:20:12 +00:00
|
|
|
err = mlx4_get_phys_port_id(dev);
|
|
|
|
|
if (err)
|
|
|
|
|
mlx4_err(dev, "Fail to get physical port id\n");
|
|
|
|
|
|
mlx4: Put physical GID and P_Key table sizes in mlx4_phys_caps struct and paravirtualize them
To allow easy paravirtualization of P_Key and GID table sizes, keep
paravirtualized sizes in mlx4_dev->caps, but save the actual physical
sizes from FW in struct: mlx4_dev->phys_cap.
In addition, in SR-IOV mode, do the following:
1. Reduce reported P_Key table size by 1.
This is done to reserve the highest P_Key index for internal use,
for declaring an invalid P_Key in P_Key paravirtualization.
We require a P_Key index which always contain an invalid P_Key
value for this purpose (i.e., one which cannot be modified by
the subnet manager). The way to do this is to reduce the
P_Key table size reported to the subnet manager by 1, so that
it will not attempt to access the P_Key at index #127.
2. Paravirtualize the GID table size to 1. Thus, each guest sees
only a single GID (at its paravirtualized index 0).
In addition, since we are paravirtualizing the GID table size to 1, we
add paravirtualization of the master GID event here (i.e., we do not
do ib_dispatch_event() for the GUID change event on the master, since
its (only) GUID never changes).
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 08:21:44 +00:00
|
|
|
if (mlx4_is_master(dev))
|
|
|
|
|
mlx4_parav_master_pf_caps(dev);
|
|
|
|
|
|
2014-07-22 12:44:11 +00:00
|
|
|
if (mlx4_low_memory_profile()) {
|
|
|
|
|
mlx4_info(dev, "Running from within kdump kernel. Using low memory profile\n");
|
|
|
|
|
profile = low_mem_profile;
|
|
|
|
|
} else {
|
|
|
|
|
profile = default_profile;
|
|
|
|
|
}
|
{NET, IB}/mlx4: Add device managed flow steering firmware API
The driver is modified to support three operation modes.
If supported by firmware use the device managed flow steering
API, that which we call device managed steering mode. Else, if
the firmware supports the B0 steering mode use it, and finally,
if none of the above, use the A0 steering mode.
When the steering mode is device managed, the code is modified
such that L2 based rules set by the mlx4_en driver for Ethernet
unicast and multicast, and the IB stack multicast attach calls
done through the mlx4_ib driver are all routed to use the device
managed API.
When attaching rule using device managed flow steering API,
the firmware returns a 64 bit registration id, which is to be
provided during detach.
Currently the firmware is always programmed during HCA initialization
to use standard L2 hashing. Future work should be done to allow
configuring the flow-steering hash function with common, non
proprietary means.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-07-05 04:03:46 +00:00
|
|
|
if (dev->caps.steering_mode ==
|
|
|
|
|
MLX4_STEERING_MODE_DEVICE_MANAGED)
|
|
|
|
|
profile.num_mcg = MLX4_FS_NUM_MCG;
|
2007-05-09 01:00:38 +00:00
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
icm_size = mlx4_make_profile(dev, &profile, &dev_cap,
|
|
|
|
|
&init_hca);
|
|
|
|
|
if ((long long) icm_size < 0) {
|
|
|
|
|
err = icm_size;
|
2014-12-30 09:59:50 +00:00
|
|
|
return err;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
}
|
2007-05-09 01:00:38 +00:00
|
|
|
|
2012-02-09 16:10:06 +00:00
|
|
|
dev->caps.max_fmr_maps = (1 << (32 - ilog2(dev->caps.num_mpts))) - 1;
|
|
|
|
|
|
2016-03-17 16:49:42 +00:00
|
|
|
if (enable_4k_uar) {
|
|
|
|
|
init_hca.log_uar_sz = ilog2(dev->caps.num_uars) +
|
|
|
|
|
PAGE_SHIFT - DEFAULT_UAR_PAGE_SHIFT;
|
|
|
|
|
init_hca.uar_page_sz = DEFAULT_UAR_PAGE_SHIFT - 12;
|
|
|
|
|
} else {
|
|
|
|
|
init_hca.log_uar_sz = ilog2(dev->caps.num_uars);
|
|
|
|
|
init_hca.uar_page_sz = PAGE_SHIFT - 12;
|
|
|
|
|
}
|
net/mlx4_core: Set UAR page size to 4KB regardless of system page size
problem description:
The current code sets UAR page size equal to system page size.
The ConnectX-3 and ConnectX-3 Pro HWs require minimum 128 UAR pages.
The mlx4 kernel drivers are not loaded if there is less than 128 UAR pages.
solution:
Always set UAR page to 4KB. This allows more UAR pages if the OS
has PAGE_SIZE larger than 4KB. For example, PowerPC kernel use 64KB
system page size, with 4MB uar region, there are 4MB/2/64KB = 32
uars (half for uar, half for blueflame). This does not meet minimum 128
UAR pages requirement. With 4KB UAR page, there are 4MB/2/4KB = 512 uars
which meet the minimum requirement.
Note that only codes in mlx4_core that deal with firmware know that uar
page size is 4KB. Codes that deal with usr page in cq and qp context
(mlx4_ib, mlx4_en and part of mlx4_core) still have the same assumption
that uar page size equals to system page size.
Note that with this implementation, on 64KB system page size kernel, there
are 16 uars per system page but only one uars is used. The other 15
uars are ignored because of the above assumption.
Regarding SR-IOV, mlx4_core in hypervisor will set the uar page size
to 4KB and mlx4_core code in virtual OS will obtain the uar page size from
firmware.
Regarding backward compatibility in SR-IOV, if hypervisor has this new code,
the virtual OS must be updated. If hypervisor has old code, and the virtual
OS has this new code, the new code will be backward compatible with the
old code. If the uar size is big enough, this new code in VF continues to
work with 64 KB uar page size (on PowerPc kernel). If the uar size does not
meet 128 uars requirement, this new code not loaded in VF and print the same
error message as the old code in Hypervisor.
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-17 15:24:26 +00:00
|
|
|
|
2013-02-06 16:19:11 +00:00
|
|
|
init_hca.mw_enabled = 0;
|
|
|
|
|
if (dev->caps.flags & MLX4_DEV_CAP_FLAG_MEM_WINDOW ||
|
|
|
|
|
dev->caps.bmme_flags & MLX4_BMME_FLAG_TYPE_2_WIN)
|
|
|
|
|
init_hca.mw_enabled = INIT_HCA_TPT_MW_ENABLE;
|
2011-03-22 22:38:41 +00:00
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
err = mlx4_init_icm(dev, &dev_cap, &init_hca, icm_size);
|
|
|
|
|
if (err)
|
2014-12-30 09:59:50 +00:00
|
|
|
return err;
|
2007-05-09 01:00:38 +00:00
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
err = mlx4_INIT_HCA(dev, &init_hca);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "INIT_HCA command failed, aborting\n");
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
goto err_free_icm;
|
|
|
|
|
}
|
2014-11-13 12:45:32 +00:00
|
|
|
|
|
|
|
|
if (dev_cap.flags2 & MLX4_DEV_CAP_FLAG2_SYS_EQS) {
|
|
|
|
|
err = mlx4_query_func(dev, &dev_cap);
|
|
|
|
|
if (err < 0) {
|
|
|
|
|
mlx4_err(dev, "QUERY_FUNC command failed, aborting.\n");
|
2014-12-30 09:59:50 +00:00
|
|
|
goto err_close;
|
2014-11-13 12:45:32 +00:00
|
|
|
} else if (err & MLX4_QUERY_FUNC_NUM_SYS_EQS) {
|
|
|
|
|
dev->caps.num_eqs = dev_cap.max_eqs;
|
|
|
|
|
dev->caps.reserved_eqs = dev_cap.reserved_eqs;
|
|
|
|
|
dev->caps.reserved_uars = dev_cap.reserved_uars;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2013-04-23 06:06:48 +00:00
|
|
|
/*
|
|
|
|
|
* If TS is supported by FW
|
|
|
|
|
* read HCA frequency by QUERY_HCA command
|
|
|
|
|
*/
|
|
|
|
|
if (dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS) {
|
|
|
|
|
memset(&init_hca, 0, sizeof(init_hca));
|
|
|
|
|
err = mlx4_QUERY_HCA(dev, &init_hca);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "QUERY_HCA command failed, disable timestamp\n");
|
2013-04-23 06:06:48 +00:00
|
|
|
dev->caps.flags2 &= ~MLX4_DEV_CAP_FLAG2_TS;
|
|
|
|
|
} else {
|
|
|
|
|
dev->caps.hca_core_clock =
|
|
|
|
|
init_hca.hca_core_clock;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* In case we got HCA frequency 0 - disable timestamping
|
|
|
|
|
* to avoid dividing by zero
|
|
|
|
|
*/
|
|
|
|
|
if (!dev->caps.hca_core_clock) {
|
|
|
|
|
dev->caps.flags2 &= ~MLX4_DEV_CAP_FLAG2_TS;
|
|
|
|
|
mlx4_err(dev,
|
2014-05-07 19:52:57 +00:00
|
|
|
"HCA frequency is 0 - timestamping is not supported\n");
|
2013-04-23 06:06:48 +00:00
|
|
|
} else if (map_internal_clock(dev)) {
|
|
|
|
|
/*
|
|
|
|
|
* Map internal clock,
|
|
|
|
|
* in case of failure disable timestamping
|
|
|
|
|
*/
|
|
|
|
|
dev->caps.flags2 &= ~MLX4_DEV_CAP_FLAG2_TS;
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to map internal clock. Timestamping is not supported\n");
|
2013-04-23 06:06:48 +00:00
|
|
|
}
|
|
|
|
|
}
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 08:58:00 +00:00
|
|
|
|
|
|
|
|
if (dev->caps.dmfs_high_steer_mode !=
|
|
|
|
|
MLX4_STEERING_DMFS_A0_NOT_SUPPORTED) {
|
|
|
|
|
if (mlx4_validate_optimized_steering(dev))
|
|
|
|
|
mlx4_warn(dev, "Optimized steering validation failed\n");
|
|
|
|
|
|
|
|
|
|
if (dev->caps.dmfs_high_steer_mode ==
|
|
|
|
|
MLX4_STEERING_DMFS_A0_DISABLE) {
|
|
|
|
|
dev->caps.dmfs_high_rate_qpn_base =
|
|
|
|
|
dev->caps.reserved_qps_cnt[MLX4_QP_REGION_FW];
|
|
|
|
|
dev->caps.dmfs_high_rate_qpn_range =
|
|
|
|
|
MLX4_A0_STEERING_TABLE_SIZE;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
mlx4_dbg(dev, "DMFS high rate steer mode is: %s\n",
|
|
|
|
|
dmfs_high_rate_steering_mode_str(
|
|
|
|
|
dev->caps.dmfs_high_steer_mode));
|
|
|
|
|
}
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
} else {
|
|
|
|
|
err = mlx4_init_slave(dev);
|
|
|
|
|
if (err) {
|
2013-06-04 05:13:27 +00:00
|
|
|
if (err != -EPROBE_DEFER)
|
|
|
|
|
mlx4_err(dev, "Failed to initialize slave\n");
|
2012-09-05 22:50:51 +00:00
|
|
|
return err;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
}
|
2007-05-09 01:00:38 +00:00
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
err = mlx4_slave_cap(dev);
|
|
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "Failed to obtain slave caps\n");
|
|
|
|
|
goto err_close;
|
|
|
|
|
}
|
2007-05-09 01:00:38 +00:00
|
|
|
}
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
if (map_bf_area(dev))
|
|
|
|
|
mlx4_dbg(dev, "Failed to map blue flame area\n");
|
|
|
|
|
|
|
|
|
|
/*Only the master set the ports, all the rest got it from it.*/
|
|
|
|
|
if (!mlx4_is_slave(dev))
|
|
|
|
|
mlx4_set_port_mask(dev);
|
|
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
err = mlx4_QUERY_ADAPTER(dev, &adapter);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "QUERY_ADAPTER command failed, aborting\n");
|
2012-09-05 22:50:51 +00:00
|
|
|
goto unmap_bf;
|
2007-05-09 01:00:38 +00:00
|
|
|
}
|
|
|
|
|
|
2014-11-09 11:51:53 +00:00
|
|
|
/* Query CONFIG_DEV parameters */
|
|
|
|
|
err = mlx4_config_dev_retrieval(dev, ¶ms);
|
|
|
|
|
if (err && err != -ENOTSUPP) {
|
|
|
|
|
mlx4_err(dev, "Failed to query CONFIG_DEV parameters\n");
|
|
|
|
|
} else if (!err) {
|
|
|
|
|
dev->caps.rx_checksum_flags_port[1] = params.rx_csum_flags_port_1;
|
|
|
|
|
dev->caps.rx_checksum_flags_port[2] = params.rx_csum_flags_port_2;
|
|
|
|
|
}
|
2007-05-09 01:00:38 +00:00
|
|
|
priv->eq_table.inta_pin = adapter.inta_pin;
|
2007-09-18 07:14:18 +00:00
|
|
|
memcpy(dev->board_id, adapter.board_id, sizeof dev->board_id);
|
2007-05-09 01:00:38 +00:00
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
2012-09-05 22:50:51 +00:00
|
|
|
unmap_bf:
|
2013-04-23 06:06:48 +00:00
|
|
|
unmap_internal_clock(dev);
|
2012-09-05 22:50:51 +00:00
|
|
|
unmap_bf_area(dev);
|
|
|
|
|
|
2014-05-29 13:30:59 +00:00
|
|
|
if (mlx4_is_slave(dev)) {
|
2014-05-29 13:31:03 +00:00
|
|
|
kfree(dev->caps.qp0_qkey);
|
2014-05-29 13:30:59 +00:00
|
|
|
kfree(dev->caps.qp0_tunnel);
|
|
|
|
|
kfree(dev->caps.qp0_proxy);
|
|
|
|
|
kfree(dev->caps.qp1_tunnel);
|
|
|
|
|
kfree(dev->caps.qp1_proxy);
|
|
|
|
|
}
|
|
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
err_close:
|
2012-10-21 14:59:23 +00:00
|
|
|
if (mlx4_is_slave(dev))
|
|
|
|
|
mlx4_slave_exit(dev);
|
|
|
|
|
else
|
|
|
|
|
mlx4_CLOSE_HCA(dev, 0);
|
2007-05-09 01:00:38 +00:00
|
|
|
|
|
|
|
|
err_free_icm:
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
if (!mlx4_is_slave(dev))
|
|
|
|
|
mlx4_free_icms(dev);
|
2007-05-09 01:00:38 +00:00
|
|
|
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2011-06-15 14:47:14 +00:00
|
|
|
static int mlx4_init_counters_table(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
2015-06-15 14:58:58 +00:00
|
|
|
int nent_pow2;
|
2011-06-15 14:47:14 +00:00
|
|
|
|
|
|
|
|
if (!(dev->caps.flags & MLX4_DEV_CAP_FLAG_COUNTERS))
|
|
|
|
|
return -ENOENT;
|
|
|
|
|
|
2015-06-15 14:58:59 +00:00
|
|
|
if (!dev->caps.max_counters)
|
|
|
|
|
return -ENOSPC;
|
|
|
|
|
|
2015-06-15 14:58:58 +00:00
|
|
|
nent_pow2 = roundup_pow_of_two(dev->caps.max_counters);
|
|
|
|
|
/* reserve last counter index for sink counter */
|
|
|
|
|
return mlx4_bitmap_init(&priv->counters_bitmap, nent_pow2,
|
|
|
|
|
nent_pow2 - 1, 0,
|
|
|
|
|
nent_pow2 - dev->caps.max_counters + 1);
|
2011-06-15 14:47:14 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void mlx4_cleanup_counters_table(struct mlx4_dev *dev)
|
|
|
|
|
{
|
2015-06-15 14:58:56 +00:00
|
|
|
if (!(dev->caps.flags & MLX4_DEV_CAP_FLAG_COUNTERS))
|
|
|
|
|
return;
|
|
|
|
|
|
2015-06-15 14:58:59 +00:00
|
|
|
if (!dev->caps.max_counters)
|
|
|
|
|
return;
|
|
|
|
|
|
2011-06-15 14:47:14 +00:00
|
|
|
mlx4_bitmap_cleanup(&mlx4_priv(dev)->counters_bitmap);
|
|
|
|
|
}
|
|
|
|
|
|
2015-06-15 14:59:02 +00:00
|
|
|
static void mlx4_cleanup_default_counters(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
int port;
|
|
|
|
|
|
|
|
|
|
for (port = 0; port < dev->caps.num_ports; port++)
|
|
|
|
|
if (priv->def_counter[port] != -1)
|
|
|
|
|
mlx4_counter_free(dev, priv->def_counter[port]);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int mlx4_allocate_default_counters(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
int port, err = 0;
|
|
|
|
|
u32 idx;
|
|
|
|
|
|
|
|
|
|
for (port = 0; port < dev->caps.num_ports; port++)
|
|
|
|
|
priv->def_counter[port] = -1;
|
|
|
|
|
|
|
|
|
|
for (port = 0; port < dev->caps.num_ports; port++) {
|
|
|
|
|
err = mlx4_counter_alloc(dev, &idx);
|
|
|
|
|
|
|
|
|
|
if (!err || err == -ENOSPC) {
|
|
|
|
|
priv->def_counter[port] = idx;
|
|
|
|
|
} else if (err == -ENOENT) {
|
|
|
|
|
err = 0;
|
|
|
|
|
continue;
|
2015-07-22 13:53:46 +00:00
|
|
|
} else if (mlx4_is_slave(dev) && err == -EINVAL) {
|
|
|
|
|
priv->def_counter[port] = MLX4_SINK_COUNTER_INDEX(dev);
|
|
|
|
|
mlx4_warn(dev, "can't allocate counter from old PF driver, using index %d\n",
|
|
|
|
|
MLX4_SINK_COUNTER_INDEX(dev));
|
|
|
|
|
err = 0;
|
2015-06-15 14:59:02 +00:00
|
|
|
} else {
|
|
|
|
|
mlx4_err(dev, "%s: failed to allocate default counter port %d err %d\n",
|
|
|
|
|
__func__, port + 1, err);
|
|
|
|
|
mlx4_cleanup_default_counters(dev);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
mlx4_dbg(dev, "%s: default counter index %d for port %d\n",
|
|
|
|
|
__func__, priv->def_counter[port], port + 1);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2012-05-15 10:35:03 +00:00
|
|
|
int __mlx4_counter_alloc(struct mlx4_dev *dev, u32 *idx)
|
2011-06-15 14:47:14 +00:00
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
|
|
|
|
|
if (!(dev->caps.flags & MLX4_DEV_CAP_FLAG_COUNTERS))
|
|
|
|
|
return -ENOENT;
|
|
|
|
|
|
|
|
|
|
*idx = mlx4_bitmap_alloc(&priv->counters_bitmap);
|
2015-06-15 14:59:02 +00:00
|
|
|
if (*idx == -1) {
|
|
|
|
|
*idx = MLX4_SINK_COUNTER_INDEX(dev);
|
|
|
|
|
return -ENOSPC;
|
|
|
|
|
}
|
2011-06-15 14:47:14 +00:00
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
2012-05-15 10:35:03 +00:00
|
|
|
|
|
|
|
|
int mlx4_counter_alloc(struct mlx4_dev *dev, u32 *idx)
|
|
|
|
|
{
|
|
|
|
|
u64 out_param;
|
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
if (mlx4_is_mfunc(dev)) {
|
|
|
|
|
err = mlx4_cmd_imm(dev, 0, &out_param, RES_COUNTER,
|
|
|
|
|
RES_OP_RESERVE, MLX4_CMD_ALLOC_RES,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_WRAPPED);
|
|
|
|
|
if (!err)
|
|
|
|
|
*idx = get_param_l(&out_param);
|
|
|
|
|
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
return __mlx4_counter_alloc(dev, idx);
|
|
|
|
|
}
|
2011-06-15 14:47:14 +00:00
|
|
|
EXPORT_SYMBOL_GPL(mlx4_counter_alloc);
|
|
|
|
|
|
2015-06-15 14:58:57 +00:00
|
|
|
static int __mlx4_clear_if_stat(struct mlx4_dev *dev,
|
|
|
|
|
u8 counter_index)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_cmd_mailbox *if_stat_mailbox;
|
|
|
|
|
int err;
|
|
|
|
|
u32 if_stat_in_mod = (counter_index & 0xff) | MLX4_QUERY_IF_STAT_RESET;
|
|
|
|
|
|
|
|
|
|
if_stat_mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(if_stat_mailbox))
|
|
|
|
|
return PTR_ERR(if_stat_mailbox);
|
|
|
|
|
|
|
|
|
|
err = mlx4_cmd_box(dev, 0, if_stat_mailbox->dma, if_stat_in_mod, 0,
|
|
|
|
|
MLX4_CMD_QUERY_IF_STAT, MLX4_CMD_TIME_CLASS_C,
|
|
|
|
|
MLX4_CMD_NATIVE);
|
|
|
|
|
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, if_stat_mailbox);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2012-05-15 10:35:03 +00:00
|
|
|
void __mlx4_counter_free(struct mlx4_dev *dev, u32 idx)
|
2011-06-15 14:47:14 +00:00
|
|
|
{
|
2015-06-15 14:58:56 +00:00
|
|
|
if (!(dev->caps.flags & MLX4_DEV_CAP_FLAG_COUNTERS))
|
|
|
|
|
return;
|
|
|
|
|
|
2015-06-15 14:59:02 +00:00
|
|
|
if (idx == MLX4_SINK_COUNTER_INDEX(dev))
|
|
|
|
|
return;
|
|
|
|
|
|
2015-06-15 14:58:57 +00:00
|
|
|
__mlx4_clear_if_stat(dev, idx);
|
|
|
|
|
|
mlx4_core: Roll back round robin bitmap allocation commit for CQs, SRQs, and MPTs
Commit f4ec9e9 "mlx4_core: Change bitmap allocator to work in round-robin fashion"
introduced round-robin allocation (via bitmap) for all resources which allocate
via a bitmap.
Round robin allocation is desirable for mcgs, counters, pd's, UARs, and xrcds.
These are simply numbers, with no involvement of ICM memory mapping.
Round robin is required for QPs, since we had a problem with immediate
reuse of a 24-bit QP number (commit f4ec9e9).
However, for other resources which use the bitmap allocator and involve
mapping ICM memory -- MPTs, CQs, SRQs -- round-robin is not desirable.
What happens in these cases is the following:
ICM memory is allocated and mapped in chunks of 256K.
Since the resource allocation index goes up monotonically, the allocator
will eventually require mapping a new chunk. Now, chunks are also unmapped
when their reference count goes back to zero. Thus, if a single app is
running and starts/exits frequently we will have the following situation:
When the app starts, a new chunk must be allocated and mapped.
When the app exits, the chunk reference count goes back to zero, and the
chunk is unmapped and freed. Therefore, the app must pay the cost of allocation
and mapping of ICM memory each time it runs (although the price is paid only when
allocating the initial entry in the new chunk).
For apps which allocate MPTs/SRQs/CQs and which operate as described above,
this presented a performance problem.
We therefore roll back the round-robin allocator modification for MPTs, CQs, SRQs.
Reported-by: Matthew Finlay <matt@mellanox.com>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-12-08 14:50:17 +00:00
|
|
|
mlx4_bitmap_free(&mlx4_priv(dev)->counters_bitmap, idx, MLX4_USE_RR);
|
2011-06-15 14:47:14 +00:00
|
|
|
return;
|
|
|
|
|
}
|
2012-05-15 10:35:03 +00:00
|
|
|
|
|
|
|
|
void mlx4_counter_free(struct mlx4_dev *dev, u32 idx)
|
|
|
|
|
{
|
2013-03-07 03:46:54 +00:00
|
|
|
u64 in_param = 0;
|
2012-05-15 10:35:03 +00:00
|
|
|
|
|
|
|
|
if (mlx4_is_mfunc(dev)) {
|
|
|
|
|
set_param_l(&in_param, idx);
|
|
|
|
|
mlx4_cmd(dev, in_param, RES_COUNTER, RES_OP_RESERVE,
|
|
|
|
|
MLX4_CMD_FREE_RES, MLX4_CMD_TIME_CLASS_A,
|
|
|
|
|
MLX4_CMD_WRAPPED);
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
__mlx4_counter_free(dev, idx);
|
|
|
|
|
}
|
2011-06-15 14:47:14 +00:00
|
|
|
EXPORT_SYMBOL_GPL(mlx4_counter_free);
|
|
|
|
|
|
2015-06-15 14:59:02 +00:00
|
|
|
int mlx4_get_default_counter_index(struct mlx4_dev *dev, int port)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
|
|
|
|
|
return priv->def_counter[port - 1];
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL_GPL(mlx4_get_default_counter_index);
|
|
|
|
|
|
2015-03-03 08:54:48 +00:00
|
|
|
void mlx4_set_admin_guid(struct mlx4_dev *dev, __be64 guid, int entry, int port)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
|
|
|
|
|
priv->mfunc.master.vf_admin[entry].vport[port].guid = guid;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL_GPL(mlx4_set_admin_guid);
|
|
|
|
|
|
|
|
|
|
__be64 mlx4_get_admin_guid(struct mlx4_dev *dev, int entry, int port)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
|
|
|
|
|
return priv->mfunc.master.vf_admin[entry].vport[port].guid;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL_GPL(mlx4_get_admin_guid);
|
|
|
|
|
|
2015-03-03 09:23:32 +00:00
|
|
|
void mlx4_set_random_admin_guid(struct mlx4_dev *dev, int entry, int port)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
__be64 guid;
|
|
|
|
|
|
|
|
|
|
/* hw GUID */
|
|
|
|
|
if (entry == 0)
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
get_random_bytes((char *)&guid, sizeof(guid));
|
|
|
|
|
guid &= ~(cpu_to_be64(1ULL << 56));
|
|
|
|
|
guid |= cpu_to_be64(1ULL << 57);
|
|
|
|
|
priv->mfunc.master.vf_admin[entry].vport[port].guid = guid;
|
|
|
|
|
}
|
|
|
|
|
|
2007-10-10 22:43:54 +00:00
|
|
|
static int mlx4_setup_hca(struct mlx4_dev *dev)
|
2007-05-09 01:00:38 +00:00
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
int err;
|
2008-10-22 22:38:42 +00:00
|
|
|
int port;
|
2008-11-29 05:29:46 +00:00
|
|
|
__be32 ib_port_default_caps;
|
2007-05-09 01:00:38 +00:00
|
|
|
|
|
|
|
|
err = mlx4_init_uar_table(dev);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to initialize user access region table, aborting\n");
|
2016-07-02 12:31:05 +00:00
|
|
|
return err;
|
2007-05-09 01:00:38 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_uar_alloc(dev, &priv->driver_uar);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to allocate driver access region, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_uar_table_free;
|
|
|
|
|
}
|
|
|
|
|
|
2011-01-12 17:50:36 +00:00
|
|
|
priv->kar = ioremap((phys_addr_t) priv->driver_uar.pfn << PAGE_SHIFT, PAGE_SIZE);
|
2007-05-09 01:00:38 +00:00
|
|
|
if (!priv->kar) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Couldn't map kernel access region, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
err = -ENOMEM;
|
|
|
|
|
goto err_uar_free;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_init_pd_table(dev);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to initialize protection domain table, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_kar_unmap;
|
|
|
|
|
}
|
|
|
|
|
|
2011-06-02 16:01:33 +00:00
|
|
|
err = mlx4_init_xrcd_table(dev);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to initialize reliable connection domain table, aborting\n");
|
2011-06-02 16:01:33 +00:00
|
|
|
goto err_pd_table_free;
|
|
|
|
|
}
|
|
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
err = mlx4_init_mr_table(dev);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to initialize memory region table, aborting\n");
|
2011-06-02 16:01:33 +00:00
|
|
|
goto err_xrcd_table_free;
|
2007-05-09 01:00:38 +00:00
|
|
|
}
|
|
|
|
|
|
2013-07-28 15:54:21 +00:00
|
|
|
if (!mlx4_is_slave(dev)) {
|
|
|
|
|
err = mlx4_init_mcg_table(dev);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to initialize multicast group table, aborting\n");
|
2013-07-28 15:54:21 +00:00
|
|
|
goto err_mr_table_free;
|
|
|
|
|
}
|
2014-06-01 08:53:50 +00:00
|
|
|
err = mlx4_config_mad_demux(dev);
|
|
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "Failed in config_mad_demux, aborting\n");
|
|
|
|
|
goto err_mcg_table_free;
|
|
|
|
|
}
|
2013-07-28 15:54:21 +00:00
|
|
|
}
|
|
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
err = mlx4_init_eq_table(dev);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to initialize event queue table, aborting\n");
|
2013-07-28 15:54:21 +00:00
|
|
|
goto err_mcg_table_free;
|
2007-05-09 01:00:38 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_cmd_use_events(dev);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to switch to event-driven firmware commands, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_eq_table_free;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_NOP(dev);
|
|
|
|
|
if (err) {
|
2007-08-07 13:08:28 +00:00
|
|
|
if (dev->flags & MLX4_FLAG_MSI_X) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_warn(dev, "NOP command failed to generate MSI-X interrupt IRQ %d)\n",
|
2015-05-31 06:30:16 +00:00
|
|
|
priv->eq_table.eq[MLX4_EQ_ASYNC].irq);
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_warn(dev, "Trying again without MSI-X\n");
|
2007-08-07 13:08:28 +00:00
|
|
|
} else {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "NOP command failed to generate interrupt (IRQ %d), aborting\n",
|
2015-05-31 06:30:16 +00:00
|
|
|
priv->eq_table.eq[MLX4_EQ_ASYNC].irq);
|
2007-05-09 01:00:38 +00:00
|
|
|
mlx4_err(dev, "BIOS or ACPI interrupt routing problem?\n");
|
2007-08-07 13:08:28 +00:00
|
|
|
}
|
2007-05-09 01:00:38 +00:00
|
|
|
|
|
|
|
|
goto err_cmd_poll;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
mlx4_dbg(dev, "NOP command IRQ test passed\n");
|
|
|
|
|
|
|
|
|
|
err = mlx4_init_cq_table(dev);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to initialize completion queue table, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_cmd_poll;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_init_srq_table(dev);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to initialize shared receive queue table, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_cq_table_free;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_init_qp_table(dev);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to initialize queue pair table, aborting\n");
|
2007-05-09 01:00:38 +00:00
|
|
|
goto err_srq_table_free;
|
|
|
|
|
}
|
|
|
|
|
|
2015-06-15 14:58:59 +00:00
|
|
|
if (!mlx4_is_slave(dev)) {
|
|
|
|
|
err = mlx4_init_counters_table(dev);
|
|
|
|
|
if (err && err != -ENOENT) {
|
|
|
|
|
mlx4_err(dev, "Failed to initialize counters table, aborting\n");
|
|
|
|
|
goto err_qp_table_free;
|
|
|
|
|
}
|
2011-06-15 14:47:14 +00:00
|
|
|
}
|
|
|
|
|
|
2015-06-15 14:59:02 +00:00
|
|
|
err = mlx4_allocate_default_counters(dev);
|
|
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "Failed to allocate default counters, aborting\n");
|
|
|
|
|
goto err_counters_table_free;
|
2011-06-15 14:47:14 +00:00
|
|
|
}
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
if (!mlx4_is_slave(dev)) {
|
|
|
|
|
for (port = 1; port <= dev->caps.num_ports; port++) {
|
|
|
|
|
ib_port_default_caps = 0;
|
|
|
|
|
err = mlx4_get_port_ib_caps(dev, port,
|
|
|
|
|
&ib_port_default_caps);
|
|
|
|
|
if (err)
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_warn(dev, "failed to get port %d default ib capabilities (%d). Continuing with caps = 0\n",
|
|
|
|
|
port, err);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
dev->caps.ib_port_def_cap[port] = ib_port_default_caps;
|
|
|
|
|
|
2012-06-19 08:21:41 +00:00
|
|
|
/* initialize per-slave default ib port capabilities */
|
|
|
|
|
if (mlx4_is_master(dev)) {
|
|
|
|
|
int i;
|
|
|
|
|
for (i = 0; i < dev->num_slaves; i++) {
|
|
|
|
|
if (i == mlx4_master_func_num(dev))
|
|
|
|
|
continue;
|
|
|
|
|
priv->mfunc.master.slave_state[i].ib_cap_mask[port] =
|
2014-05-07 19:52:57 +00:00
|
|
|
ib_port_default_caps;
|
2012-06-19 08:21:41 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2012-01-11 17:02:17 +00:00
|
|
|
if (mlx4_is_mfunc(dev))
|
|
|
|
|
dev->caps.port_ib_mtu[port] = IB_MTU_2048;
|
|
|
|
|
else
|
|
|
|
|
dev->caps.port_ib_mtu[port] = IB_MTU_4096;
|
2011-10-24 09:02:34 +00:00
|
|
|
|
mlx4: Put physical GID and P_Key table sizes in mlx4_phys_caps struct and paravirtualize them
To allow easy paravirtualization of P_Key and GID table sizes, keep
paravirtualized sizes in mlx4_dev->caps, but save the actual physical
sizes from FW in struct: mlx4_dev->phys_cap.
In addition, in SR-IOV mode, do the following:
1. Reduce reported P_Key table size by 1.
This is done to reserve the highest P_Key index for internal use,
for declaring an invalid P_Key in P_Key paravirtualization.
We require a P_Key index which always contain an invalid P_Key
value for this purpose (i.e., one which cannot be modified by
the subnet manager). The way to do this is to reduce the
P_Key table size reported to the subnet manager by 1, so that
it will not attempt to access the P_Key at index #127.
2. Paravirtualize the GID table size to 1. Thus, each guest sees
only a single GID (at its paravirtualized index 0).
In addition, since we are paravirtualizing the GID table size to 1, we
add paravirtualization of the master GID event here (i.e., we do not
do ib_dispatch_event() for the GUID change event on the master, since
its (only) GUID never changes).
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 08:21:44 +00:00
|
|
|
err = mlx4_SET_PORT(dev, port, mlx4_is_master(dev) ?
|
|
|
|
|
dev->caps.pkey_table_len[port] : -1);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "Failed to set port %d, aborting\n",
|
2014-05-07 19:52:57 +00:00
|
|
|
port);
|
2015-06-15 14:59:02 +00:00
|
|
|
goto err_default_countes_free;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
}
|
2008-10-22 22:38:42 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
return 0;
|
|
|
|
|
|
2015-06-15 14:59:02 +00:00
|
|
|
err_default_countes_free:
|
|
|
|
|
mlx4_cleanup_default_counters(dev);
|
|
|
|
|
|
2011-06-15 14:47:14 +00:00
|
|
|
err_counters_table_free:
|
2015-06-15 14:58:59 +00:00
|
|
|
if (!mlx4_is_slave(dev))
|
|
|
|
|
mlx4_cleanup_counters_table(dev);
|
2011-06-15 14:47:14 +00:00
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
err_qp_table_free:
|
|
|
|
|
mlx4_cleanup_qp_table(dev);
|
|
|
|
|
|
|
|
|
|
err_srq_table_free:
|
|
|
|
|
mlx4_cleanup_srq_table(dev);
|
|
|
|
|
|
|
|
|
|
err_cq_table_free:
|
|
|
|
|
mlx4_cleanup_cq_table(dev);
|
|
|
|
|
|
|
|
|
|
err_cmd_poll:
|
|
|
|
|
mlx4_cmd_use_polling(dev);
|
|
|
|
|
|
|
|
|
|
err_eq_table_free:
|
|
|
|
|
mlx4_cleanup_eq_table(dev);
|
|
|
|
|
|
2013-07-28 15:54:21 +00:00
|
|
|
err_mcg_table_free:
|
|
|
|
|
if (!mlx4_is_slave(dev))
|
|
|
|
|
mlx4_cleanup_mcg_table(dev);
|
|
|
|
|
|
2007-07-12 14:50:45 +00:00
|
|
|
err_mr_table_free:
|
2007-05-09 01:00:38 +00:00
|
|
|
mlx4_cleanup_mr_table(dev);
|
|
|
|
|
|
2011-06-02 16:01:33 +00:00
|
|
|
err_xrcd_table_free:
|
|
|
|
|
mlx4_cleanup_xrcd_table(dev);
|
|
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
err_pd_table_free:
|
|
|
|
|
mlx4_cleanup_pd_table(dev);
|
|
|
|
|
|
|
|
|
|
err_kar_unmap:
|
|
|
|
|
iounmap(priv->kar);
|
|
|
|
|
|
|
|
|
|
err_uar_free:
|
|
|
|
|
mlx4_uar_free(dev, &priv->driver_uar);
|
|
|
|
|
|
|
|
|
|
err_uar_table_free:
|
|
|
|
|
mlx4_cleanup_uar_table(dev);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2015-05-31 06:30:17 +00:00
|
|
|
static int mlx4_init_affinity_hint(struct mlx4_dev *dev, int port, int eqn)
|
|
|
|
|
{
|
|
|
|
|
int requested_cpu = 0;
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
struct mlx4_eq *eq;
|
|
|
|
|
int off = 0;
|
|
|
|
|
int i;
|
|
|
|
|
|
|
|
|
|
if (eqn > dev->caps.num_comp_vectors)
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
|
|
for (i = 1; i < port; i++)
|
|
|
|
|
off += mlx4_get_eqs_per_port(dev, i);
|
|
|
|
|
|
|
|
|
|
requested_cpu = eqn - off - !!(eqn > MLX4_EQ_ASYNC);
|
|
|
|
|
|
|
|
|
|
/* Meaning EQs are shared, and this call comes from the second port */
|
|
|
|
|
if (requested_cpu < 0)
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
|
|
eq = &priv->eq_table.eq[eqn];
|
|
|
|
|
|
|
|
|
|
if (!zalloc_cpumask_var(&eq->affinity_mask, GFP_KERNEL))
|
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
|
|
cpumask_set_cpu(requested_cpu, eq->affinity_mask);
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
2008-02-05 04:20:41 +00:00
|
|
|
static void mlx4_enable_msi_x(struct mlx4_dev *dev)
|
2007-05-09 01:00:38 +00:00
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
2008-12-22 15:15:03 +00:00
|
|
|
struct msix_entry *entries;
|
2007-05-09 01:00:38 +00:00
|
|
|
int i;
|
2015-05-31 06:30:16 +00:00
|
|
|
int port = 0;
|
2007-05-09 01:00:38 +00:00
|
|
|
|
|
|
|
|
if (msi_x) {
|
2015-05-31 06:30:16 +00:00
|
|
|
int nreq = dev->caps.num_ports * num_online_cpus() + 1;
|
2014-11-13 12:45:32 +00:00
|
|
|
|
2013-01-17 05:30:43 +00:00
|
|
|
nreq = min_t(int, dev->caps.num_eqs - dev->caps.reserved_eqs,
|
|
|
|
|
nreq);
|
2015-10-07 16:31:46 +00:00
|
|
|
if (nreq > MAX_MSIX)
|
2015-08-27 19:43:25 +00:00
|
|
|
nreq = MAX_MSIX;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
|
2008-12-22 15:15:03 +00:00
|
|
|
entries = kcalloc(nreq, sizeof *entries, GFP_KERNEL);
|
|
|
|
|
if (!entries)
|
|
|
|
|
goto no_msi;
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < nreq; ++i)
|
2007-05-09 01:00:38 +00:00
|
|
|
entries[i].entry = i;
|
|
|
|
|
|
2015-01-25 14:59:35 +00:00
|
|
|
nreq = pci_enable_msix_range(dev->persist->pdev, entries, 2,
|
|
|
|
|
nreq);
|
2014-02-18 10:11:47 +00:00
|
|
|
|
2015-05-31 06:30:16 +00:00
|
|
|
if (nreq < 0 || nreq < MLX4_EQ_ASYNC) {
|
2009-04-21 17:11:06 +00:00
|
|
|
kfree(entries);
|
2007-05-09 01:00:38 +00:00
|
|
|
goto no_msi;
|
2011-03-22 22:37:47 +00:00
|
|
|
}
|
2015-05-31 06:30:16 +00:00
|
|
|
/* 1 is reserved for events (asyncrounous EQ) */
|
|
|
|
|
dev->caps.num_comp_vectors = nreq - 1;
|
|
|
|
|
|
|
|
|
|
priv->eq_table.eq[MLX4_EQ_ASYNC].irq = entries[0].vector;
|
|
|
|
|
bitmap_zero(priv->eq_table.eq[MLX4_EQ_ASYNC].actv_ports.ports,
|
|
|
|
|
dev->caps.num_ports);
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < dev->caps.num_comp_vectors + 1; i++) {
|
|
|
|
|
if (i == MLX4_EQ_ASYNC)
|
|
|
|
|
continue;
|
|
|
|
|
|
|
|
|
|
priv->eq_table.eq[i].irq =
|
|
|
|
|
entries[i + 1 - !!(i > MLX4_EQ_ASYNC)].vector;
|
|
|
|
|
|
2015-10-07 16:31:46 +00:00
|
|
|
if (MLX4_IS_LEGACY_EQ_MODE(dev->caps)) {
|
2015-05-31 06:30:16 +00:00
|
|
|
bitmap_fill(priv->eq_table.eq[i].actv_ports.ports,
|
|
|
|
|
dev->caps.num_ports);
|
2015-05-31 06:30:17 +00:00
|
|
|
/* We don't set affinity hint when there
|
|
|
|
|
* aren't enough EQs
|
|
|
|
|
*/
|
2015-05-31 06:30:16 +00:00
|
|
|
} else {
|
|
|
|
|
set_bit(port,
|
|
|
|
|
priv->eq_table.eq[i].actv_ports.ports);
|
2015-05-31 06:30:17 +00:00
|
|
|
if (mlx4_init_affinity_hint(dev, port + 1, i))
|
|
|
|
|
mlx4_warn(dev, "Couldn't init hint cpumask for EQ %d\n",
|
|
|
|
|
i);
|
2015-05-31 06:30:16 +00:00
|
|
|
}
|
|
|
|
|
/* We divide the Eqs evenly between the two ports.
|
|
|
|
|
* (dev->caps.num_comp_vectors / dev->caps.num_ports)
|
|
|
|
|
* refers to the number of Eqs per port
|
|
|
|
|
* (i.e eqs_per_port). Theoretically, we would like to
|
|
|
|
|
* write something like (i + 1) % eqs_per_port == 0.
|
|
|
|
|
* However, since there's an asynchronous Eq, we have
|
|
|
|
|
* to skip over it by comparing this condition to
|
|
|
|
|
* !!((i + 1) > MLX4_EQ_ASYNC).
|
|
|
|
|
*/
|
|
|
|
|
if ((dev->caps.num_comp_vectors > dev->caps.num_ports) &&
|
|
|
|
|
((i + 1) %
|
|
|
|
|
(dev->caps.num_comp_vectors / dev->caps.num_ports)) ==
|
|
|
|
|
!!((i + 1) > MLX4_EQ_ASYNC))
|
|
|
|
|
/* If dev->caps.num_comp_vectors < dev->caps.num_ports,
|
|
|
|
|
* everything is shared anyway.
|
|
|
|
|
*/
|
|
|
|
|
port++;
|
|
|
|
|
}
|
2007-05-09 01:00:38 +00:00
|
|
|
|
|
|
|
|
dev->flags |= MLX4_FLAG_MSI_X;
|
2008-12-22 15:15:03 +00:00
|
|
|
|
|
|
|
|
kfree(entries);
|
2007-05-09 01:00:38 +00:00
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
no_msi:
|
2008-12-22 15:15:03 +00:00
|
|
|
dev->caps.num_comp_vectors = 1;
|
|
|
|
|
|
2015-05-31 06:30:16 +00:00
|
|
|
BUG_ON(MLX4_EQ_ASYNC >= 2);
|
|
|
|
|
for (i = 0; i < 2; ++i) {
|
2015-01-25 14:59:35 +00:00
|
|
|
priv->eq_table.eq[i].irq = dev->persist->pdev->irq;
|
2015-05-31 06:30:16 +00:00
|
|
|
if (i != MLX4_EQ_ASYNC) {
|
|
|
|
|
bitmap_fill(priv->eq_table.eq[i].actv_ports.ports,
|
|
|
|
|
dev->caps.num_ports);
|
|
|
|
|
}
|
|
|
|
|
}
|
2007-05-09 01:00:38 +00:00
|
|
|
}
|
|
|
|
|
|
2008-10-22 22:38:42 +00:00
|
|
|
static int mlx4_init_port_info(struct mlx4_dev *dev, int port)
|
2008-10-22 18:44:46 +00:00
|
|
|
{
|
2016-02-26 16:32:24 +00:00
|
|
|
struct devlink *devlink = priv_to_devlink(mlx4_priv(dev));
|
2008-10-22 18:44:46 +00:00
|
|
|
struct mlx4_port_info *info = &mlx4_priv(dev)->port[port];
|
2016-02-26 16:32:24 +00:00
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
err = devlink_port_register(devlink, &info->devlink_port, port);
|
|
|
|
|
if (err)
|
|
|
|
|
return err;
|
2008-10-22 18:44:46 +00:00
|
|
|
|
|
|
|
|
info->dev = dev;
|
|
|
|
|
info->port = port;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
if (!mlx4_is_slave(dev)) {
|
|
|
|
|
mlx4_init_mac_table(dev, &info->mac_table);
|
|
|
|
|
mlx4_init_vlan_table(dev, &info->vlan_table);
|
2014-05-27 06:26:38 +00:00
|
|
|
mlx4_init_roce_gid_table(dev, &info->gid_table);
|
2013-02-07 02:25:22 +00:00
|
|
|
info->base_qpn = mlx4_get_base_qpn(dev, port);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
}
|
2008-10-22 22:38:42 +00:00
|
|
|
|
|
|
|
|
sprintf(info->dev_name, "mlx4_port%d", port);
|
|
|
|
|
info->port_attr.attr.name = info->dev_name;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
if (mlx4_is_mfunc(dev))
|
|
|
|
|
info->port_attr.attr.mode = S_IRUGO;
|
|
|
|
|
else {
|
|
|
|
|
info->port_attr.attr.mode = S_IRUGO | S_IWUSR;
|
|
|
|
|
info->port_attr.store = set_port_type;
|
|
|
|
|
}
|
2008-10-22 22:38:42 +00:00
|
|
|
info->port_attr.show = show_port_type;
|
2010-03-15 21:01:55 +00:00
|
|
|
sysfs_attr_init(&info->port_attr.attr);
|
2008-10-22 22:38:42 +00:00
|
|
|
|
2015-01-25 14:59:35 +00:00
|
|
|
err = device_create_file(&dev->persist->pdev->dev, &info->port_attr);
|
2008-10-22 22:38:42 +00:00
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "Failed to create file for port %d\n", port);
|
2016-02-26 16:32:24 +00:00
|
|
|
devlink_port_unregister(&info->devlink_port);
|
2008-10-22 22:38:42 +00:00
|
|
|
info->port = -1;
|
|
|
|
|
}
|
|
|
|
|
|
2012-01-11 17:02:17 +00:00
|
|
|
sprintf(info->dev_mtu_name, "mlx4_port%d_mtu", port);
|
|
|
|
|
info->port_mtu_attr.attr.name = info->dev_mtu_name;
|
|
|
|
|
if (mlx4_is_mfunc(dev))
|
|
|
|
|
info->port_mtu_attr.attr.mode = S_IRUGO;
|
|
|
|
|
else {
|
|
|
|
|
info->port_mtu_attr.attr.mode = S_IRUGO | S_IWUSR;
|
|
|
|
|
info->port_mtu_attr.store = set_port_ib_mtu;
|
|
|
|
|
}
|
|
|
|
|
info->port_mtu_attr.show = show_port_ib_mtu;
|
|
|
|
|
sysfs_attr_init(&info->port_mtu_attr.attr);
|
|
|
|
|
|
2015-01-25 14:59:35 +00:00
|
|
|
err = device_create_file(&dev->persist->pdev->dev,
|
|
|
|
|
&info->port_mtu_attr);
|
2012-01-11 17:02:17 +00:00
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "Failed to create mtu file for port %d\n", port);
|
2015-01-25 14:59:35 +00:00
|
|
|
device_remove_file(&info->dev->persist->pdev->dev,
|
|
|
|
|
&info->port_attr);
|
2016-09-20 11:55:31 +00:00
|
|
|
devlink_port_unregister(&info->devlink_port);
|
2012-01-11 17:02:17 +00:00
|
|
|
info->port = -1;
|
|
|
|
|
}
|
|
|
|
|
|
2008-10-22 22:38:42 +00:00
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void mlx4_cleanup_port_info(struct mlx4_port_info *info)
|
|
|
|
|
{
|
|
|
|
|
if (info->port < 0)
|
|
|
|
|
return;
|
|
|
|
|
|
2015-01-25 14:59:35 +00:00
|
|
|
device_remove_file(&info->dev->persist->pdev->dev, &info->port_attr);
|
|
|
|
|
device_remove_file(&info->dev->persist->pdev->dev,
|
|
|
|
|
&info->port_mtu_attr);
|
2016-09-20 11:55:31 +00:00
|
|
|
devlink_port_unregister(&info->devlink_port);
|
|
|
|
|
|
2015-05-31 06:30:16 +00:00
|
|
|
#ifdef CONFIG_RFS_ACCEL
|
|
|
|
|
free_irq_cpu_rmap(info->rmap);
|
|
|
|
|
info->rmap = NULL;
|
|
|
|
|
#endif
|
2008-10-22 18:44:46 +00:00
|
|
|
}
|
|
|
|
|
|
2011-03-22 22:38:24 +00:00
|
|
|
static int mlx4_init_steering(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
int num_entries = dev->caps.num_ports;
|
|
|
|
|
int i, j;
|
|
|
|
|
|
|
|
|
|
priv->steer = kzalloc(sizeof(struct mlx4_steer) * num_entries, GFP_KERNEL);
|
|
|
|
|
if (!priv->steer)
|
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
2012-02-14 06:37:41 +00:00
|
|
|
for (i = 0; i < num_entries; i++)
|
2011-03-22 22:38:24 +00:00
|
|
|
for (j = 0; j < MLX4_NUM_STEERS; j++) {
|
|
|
|
|
INIT_LIST_HEAD(&priv->steer[i].promisc_qps[j]);
|
|
|
|
|
INIT_LIST_HEAD(&priv->steer[i].steer_entries[j]);
|
|
|
|
|
}
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void mlx4_clear_steering(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
struct mlx4_steer_index *entry, *tmp_entry;
|
|
|
|
|
struct mlx4_promisc_qp *pqp, *tmp_pqp;
|
|
|
|
|
int num_entries = dev->caps.num_ports;
|
|
|
|
|
int i, j;
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < num_entries; i++) {
|
|
|
|
|
for (j = 0; j < MLX4_NUM_STEERS; j++) {
|
|
|
|
|
list_for_each_entry_safe(pqp, tmp_pqp,
|
|
|
|
|
&priv->steer[i].promisc_qps[j],
|
|
|
|
|
list) {
|
|
|
|
|
list_del(&pqp->list);
|
|
|
|
|
kfree(pqp);
|
|
|
|
|
}
|
|
|
|
|
list_for_each_entry_safe(entry, tmp_entry,
|
|
|
|
|
&priv->steer[i].steer_entries[j],
|
|
|
|
|
list) {
|
|
|
|
|
list_del(&entry->list);
|
|
|
|
|
list_for_each_entry_safe(pqp, tmp_pqp,
|
|
|
|
|
&entry->duplicates,
|
|
|
|
|
list) {
|
|
|
|
|
list_del(&pqp->list);
|
|
|
|
|
kfree(pqp);
|
|
|
|
|
}
|
|
|
|
|
kfree(entry);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
kfree(priv->steer);
|
|
|
|
|
}
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
static int extended_func_num(struct pci_dev *pdev)
|
|
|
|
|
{
|
|
|
|
|
return PCI_SLOT(pdev->devfn) * 8 + PCI_FUNC(pdev->devfn);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#define MLX4_OWNER_BASE 0x8069c
|
|
|
|
|
#define MLX4_OWNER_SIZE 4
|
|
|
|
|
|
|
|
|
|
static int mlx4_get_ownership(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
void __iomem *owner;
|
|
|
|
|
u32 ret;
|
|
|
|
|
|
2015-01-25 14:59:35 +00:00
|
|
|
if (pci_channel_offline(dev->persist->pdev))
|
2012-07-20 09:55:43 +00:00
|
|
|
return -EIO;
|
|
|
|
|
|
2015-01-25 14:59:35 +00:00
|
|
|
owner = ioremap(pci_resource_start(dev->persist->pdev, 0) +
|
|
|
|
|
MLX4_OWNER_BASE,
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
MLX4_OWNER_SIZE);
|
|
|
|
|
if (!owner) {
|
|
|
|
|
mlx4_err(dev, "Failed to obtain ownership bit\n");
|
|
|
|
|
return -ENOMEM;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
ret = readl(owner);
|
|
|
|
|
iounmap(owner);
|
|
|
|
|
return (int) !!ret;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void mlx4_free_ownership(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
void __iomem *owner;
|
|
|
|
|
|
2015-01-25 14:59:35 +00:00
|
|
|
if (pci_channel_offline(dev->persist->pdev))
|
2012-07-20 09:55:43 +00:00
|
|
|
return;
|
|
|
|
|
|
2015-01-25 14:59:35 +00:00
|
|
|
owner = ioremap(pci_resource_start(dev->persist->pdev, 0) +
|
|
|
|
|
MLX4_OWNER_BASE,
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
MLX4_OWNER_SIZE);
|
|
|
|
|
if (!owner) {
|
|
|
|
|
mlx4_err(dev, "Failed to obtain ownership bit\n");
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
writel(0, owner);
|
|
|
|
|
msleep(1000);
|
|
|
|
|
iounmap(owner);
|
|
|
|
|
}
|
|
|
|
|
|
2014-11-13 12:45:30 +00:00
|
|
|
#define SRIOV_VALID_STATE(flags) (!!((flags) & MLX4_FLAG_SRIOV) ==\
|
|
|
|
|
!!((flags) & MLX4_FLAG_MASTER))
|
|
|
|
|
|
|
|
|
|
static u64 mlx4_enable_sriov(struct mlx4_dev *dev, struct pci_dev *pdev,
|
2015-01-25 14:59:42 +00:00
|
|
|
u8 total_vfs, int existing_vfs, int reset_flow)
|
2014-11-13 12:45:30 +00:00
|
|
|
{
|
|
|
|
|
u64 dev_flags = dev->flags;
|
2014-12-14 14:18:04 +00:00
|
|
|
int err = 0;
|
2015-07-06 14:20:19 +00:00
|
|
|
int fw_enabled_sriov_vfs = min(pci_sriov_get_totalvfs(pdev),
|
|
|
|
|
MLX4_MAX_NUM_VF);
|
2014-11-13 12:45:30 +00:00
|
|
|
|
2015-01-25 14:59:42 +00:00
|
|
|
if (reset_flow) {
|
|
|
|
|
dev->dev_vfs = kcalloc(total_vfs, sizeof(*dev->dev_vfs),
|
|
|
|
|
GFP_KERNEL);
|
|
|
|
|
if (!dev->dev_vfs)
|
|
|
|
|
goto free_mem;
|
|
|
|
|
return dev_flags;
|
|
|
|
|
}
|
|
|
|
|
|
2014-12-14 14:18:04 +00:00
|
|
|
atomic_inc(&pf_loading);
|
|
|
|
|
if (dev->flags & MLX4_FLAG_SRIOV) {
|
|
|
|
|
if (existing_vfs != total_vfs) {
|
|
|
|
|
mlx4_err(dev, "SR-IOV was already enabled, but with num_vfs (%d) different than requested (%d)\n",
|
|
|
|
|
existing_vfs, total_vfs);
|
|
|
|
|
total_vfs = existing_vfs;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
dev->dev_vfs = kzalloc(total_vfs * sizeof(*dev->dev_vfs), GFP_KERNEL);
|
2014-11-13 12:45:30 +00:00
|
|
|
if (NULL == dev->dev_vfs) {
|
|
|
|
|
mlx4_err(dev, "Failed to allocate memory for VFs\n");
|
|
|
|
|
goto disable_sriov;
|
2014-12-14 14:18:04 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (!(dev->flags & MLX4_FLAG_SRIOV)) {
|
2015-07-06 14:20:19 +00:00
|
|
|
if (total_vfs > fw_enabled_sriov_vfs) {
|
|
|
|
|
mlx4_err(dev, "requested vfs (%d) > available vfs (%d). Continuing without SR_IOV\n",
|
|
|
|
|
total_vfs, fw_enabled_sriov_vfs);
|
|
|
|
|
err = -ENOMEM;
|
|
|
|
|
goto disable_sriov;
|
|
|
|
|
}
|
2014-12-14 14:18:04 +00:00
|
|
|
mlx4_warn(dev, "Enabling SR-IOV with %d VFs\n", total_vfs);
|
|
|
|
|
err = pci_enable_sriov(pdev, total_vfs);
|
|
|
|
|
}
|
|
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "Failed to enable SR-IOV, continuing without SR-IOV (err = %d)\n",
|
|
|
|
|
err);
|
|
|
|
|
goto disable_sriov;
|
|
|
|
|
} else {
|
|
|
|
|
mlx4_warn(dev, "Running in master mode\n");
|
|
|
|
|
dev_flags |= MLX4_FLAG_SRIOV |
|
|
|
|
|
MLX4_FLAG_MASTER;
|
|
|
|
|
dev_flags &= ~MLX4_FLAG_SLAVE;
|
2015-01-25 14:59:35 +00:00
|
|
|
dev->persist->num_vfs = total_vfs;
|
2014-11-13 12:45:30 +00:00
|
|
|
}
|
|
|
|
|
return dev_flags;
|
|
|
|
|
|
|
|
|
|
disable_sriov:
|
2014-12-14 14:18:04 +00:00
|
|
|
atomic_dec(&pf_loading);
|
2015-01-25 14:59:42 +00:00
|
|
|
free_mem:
|
2015-01-25 14:59:35 +00:00
|
|
|
dev->persist->num_vfs = 0;
|
2014-11-13 12:45:30 +00:00
|
|
|
kfree(dev->dev_vfs);
|
2015-06-02 21:07:23 +00:00
|
|
|
dev->dev_vfs = NULL;
|
2014-11-13 12:45:30 +00:00
|
|
|
return dev_flags & ~MLX4_FLAG_MASTER;
|
|
|
|
|
}
|
|
|
|
|
|
2014-11-13 12:45:33 +00:00
|
|
|
enum {
|
|
|
|
|
MLX4_DEV_CAP_CHECK_NUM_VFS_ABOVE_64 = -1,
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
static int mlx4_check_dev_cap(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap,
|
|
|
|
|
int *nvfs)
|
|
|
|
|
{
|
|
|
|
|
int requested_vfs = nvfs[0] + nvfs[1] + nvfs[2];
|
|
|
|
|
/* Checking for 64 VFs as a limitation of CX2 */
|
|
|
|
|
if (!(dev_cap->flags2 & MLX4_DEV_CAP_FLAG2_80_VFS) &&
|
|
|
|
|
requested_vfs >= 64) {
|
|
|
|
|
mlx4_err(dev, "Requested %d VFs, but FW does not support more than 64\n",
|
|
|
|
|
requested_vfs);
|
|
|
|
|
return MLX4_DEV_CAP_CHECK_NUM_VFS_ABOVE_64;
|
|
|
|
|
}
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
2016-04-20 13:01:16 +00:00
|
|
|
static int mlx4_pci_enable_device(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct pci_dev *pdev = dev->persist->pdev;
|
|
|
|
|
int err = 0;
|
|
|
|
|
|
|
|
|
|
mutex_lock(&dev->persist->pci_status_mutex);
|
|
|
|
|
if (dev->persist->pci_status == MLX4_PCI_STATUS_DISABLED) {
|
|
|
|
|
err = pci_enable_device(pdev);
|
|
|
|
|
if (!err)
|
|
|
|
|
dev->persist->pci_status = MLX4_PCI_STATUS_ENABLED;
|
|
|
|
|
}
|
|
|
|
|
mutex_unlock(&dev->persist->pci_status_mutex);
|
|
|
|
|
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void mlx4_pci_disable_device(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct pci_dev *pdev = dev->persist->pdev;
|
|
|
|
|
|
|
|
|
|
mutex_lock(&dev->persist->pci_status_mutex);
|
|
|
|
|
if (dev->persist->pci_status == MLX4_PCI_STATUS_ENABLED) {
|
|
|
|
|
pci_disable_device(pdev);
|
|
|
|
|
dev->persist->pci_status = MLX4_PCI_STATUS_DISABLED;
|
|
|
|
|
}
|
|
|
|
|
mutex_unlock(&dev->persist->pci_status_mutex);
|
|
|
|
|
}
|
|
|
|
|
|
2014-09-30 09:03:48 +00:00
|
|
|
static int mlx4_load_one(struct pci_dev *pdev, int pci_dev_data,
|
2015-01-25 14:59:42 +00:00
|
|
|
int total_vfs, int *nvfs, struct mlx4_priv *priv,
|
|
|
|
|
int reset_flow)
|
2007-05-09 01:00:38 +00:00
|
|
|
{
|
|
|
|
|
struct mlx4_dev *dev;
|
2014-09-30 09:03:48 +00:00
|
|
|
unsigned sum = 0;
|
2007-05-09 01:00:38 +00:00
|
|
|
int err;
|
2008-10-22 18:44:46 +00:00
|
|
|
int port;
|
2014-09-30 09:03:48 +00:00
|
|
|
int i;
|
2014-11-13 12:45:32 +00:00
|
|
|
struct mlx4_dev_cap *dev_cap = NULL;
|
2014-09-30 09:03:47 +00:00
|
|
|
int existing_vfs = 0;
|
2007-05-09 01:00:38 +00:00
|
|
|
|
2014-09-30 09:03:48 +00:00
|
|
|
dev = &priv->dev;
|
2007-05-09 01:00:38 +00:00
|
|
|
|
2007-06-07 18:51:58 +00:00
|
|
|
INIT_LIST_HEAD(&priv->ctx_list);
|
|
|
|
|
spin_lock_init(&priv->ctx_lock);
|
2007-05-09 01:00:38 +00:00
|
|
|
|
2008-10-22 22:38:42 +00:00
|
|
|
mutex_init(&priv->port_mutex);
|
2015-02-03 14:48:33 +00:00
|
|
|
mutex_init(&priv->bond_mutex);
|
2008-10-22 22:38:42 +00:00
|
|
|
|
2008-04-23 18:55:45 +00:00
|
|
|
INIT_LIST_HEAD(&priv->pgdir_list);
|
|
|
|
|
mutex_init(&priv->pgdir_mutex);
|
2016-06-13 14:50:25 +00:00
|
|
|
spin_lock_init(&priv->cmd.context_lock);
|
2008-04-23 18:55:45 +00:00
|
|
|
|
2011-03-22 22:38:41 +00:00
|
|
|
INIT_LIST_HEAD(&priv->bf_list);
|
|
|
|
|
mutex_init(&priv->bf_mutex);
|
|
|
|
|
|
2011-06-23 04:44:30 +00:00
|
|
|
dev->rev_id = pdev->revision;
|
2013-11-07 10:19:53 +00:00
|
|
|
dev->numa_node = dev_to_node(&pdev->dev);
|
2014-09-30 09:03:48 +00:00
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/* Detect if this device is a virtual function */
|
2012-09-27 16:23:41 +00:00
|
|
|
if (pci_dev_data & MLX4_PCI_DEV_IS_VF) {
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
mlx4_warn(dev, "Detected virtual function - running in slave mode\n");
|
|
|
|
|
dev->flags |= MLX4_FLAG_SLAVE;
|
|
|
|
|
} else {
|
|
|
|
|
/* We reset the device and enable SRIOV only for physical
|
|
|
|
|
* devices. Try to claim ownership on the device;
|
|
|
|
|
* if already taken, skip -- do not allow multiple PFs */
|
|
|
|
|
err = mlx4_get_ownership(dev);
|
|
|
|
|
if (err) {
|
|
|
|
|
if (err < 0)
|
2014-09-30 09:03:48 +00:00
|
|
|
return err;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
else {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_warn(dev, "Multiple PFs not yet supported - Skipping PF\n");
|
2014-09-30 09:03:48 +00:00
|
|
|
return -EINVAL;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
}
|
|
|
|
|
}
|
2011-06-23 04:44:30 +00:00
|
|
|
|
2013-07-28 15:54:21 +00:00
|
|
|
atomic_set(&priv->opreq_count, 0);
|
|
|
|
|
INIT_WORK(&priv->opreq_task, mlx4_opreq_action);
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/*
|
|
|
|
|
* Now reset the HCA before we touch the PCI capabilities or
|
|
|
|
|
* attempt a firmware command, since a boot ROM may have left
|
|
|
|
|
* the HCA in an undefined state.
|
|
|
|
|
*/
|
|
|
|
|
err = mlx4_reset(dev);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to reset HCA, aborting\n");
|
2014-09-30 09:03:48 +00:00
|
|
|
goto err_sriov;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
}
|
2014-11-13 12:45:32 +00:00
|
|
|
|
|
|
|
|
if (total_vfs) {
|
|
|
|
|
dev->flags = MLX4_FLAG_MASTER;
|
2014-12-14 14:18:04 +00:00
|
|
|
existing_vfs = pci_num_vf(pdev);
|
|
|
|
|
if (existing_vfs)
|
|
|
|
|
dev->flags |= MLX4_FLAG_SRIOV;
|
2015-01-25 14:59:35 +00:00
|
|
|
dev->persist->num_vfs = total_vfs;
|
2014-11-13 12:45:32 +00:00
|
|
|
}
|
2007-05-09 01:00:38 +00:00
|
|
|
}
|
|
|
|
|
|
2015-01-25 14:59:38 +00:00
|
|
|
/* on load remove any previous indication of internal error,
|
|
|
|
|
* device is up.
|
|
|
|
|
*/
|
|
|
|
|
dev->persist->state = MLX4_DEVICE_STATE_UP;
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
slave_start:
|
2012-09-05 22:50:52 +00:00
|
|
|
err = mlx4_cmd_init(dev);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to init command interface, aborting\n");
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
goto err_sriov;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* In slave functions, the communication channel must be initialized
|
|
|
|
|
* before posting commands. Also, init num_slaves before calling
|
|
|
|
|
* mlx4_init_hca */
|
|
|
|
|
if (mlx4_is_mfunc(dev)) {
|
2014-11-13 12:45:32 +00:00
|
|
|
if (mlx4_is_master(dev)) {
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
dev->num_slaves = MLX4_MAX_NUM_SLAVES;
|
2014-11-13 12:45:32 +00:00
|
|
|
|
|
|
|
|
} else {
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
dev->num_slaves = 0;
|
2013-01-24 01:54:17 +00:00
|
|
|
err = mlx4_multi_func_init(dev);
|
|
|
|
|
if (err) {
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "Failed to init slave mfunc interface, aborting\n");
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
goto err_cmd;
|
|
|
|
|
}
|
|
|
|
|
}
|
2007-05-09 01:00:38 +00:00
|
|
|
}
|
|
|
|
|
|
2014-11-13 12:45:30 +00:00
|
|
|
err = mlx4_init_fw(dev);
|
|
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "Failed to init fw, aborting.\n");
|
|
|
|
|
goto err_mfunc;
|
|
|
|
|
}
|
|
|
|
|
|
2014-11-13 12:45:32 +00:00
|
|
|
if (mlx4_is_master(dev)) {
|
2014-12-14 14:18:04 +00:00
|
|
|
/* when we hit the goto slave_start below, dev_cap already initialized */
|
2014-11-13 12:45:32 +00:00
|
|
|
if (!dev_cap) {
|
|
|
|
|
dev_cap = kzalloc(sizeof(*dev_cap), GFP_KERNEL);
|
|
|
|
|
|
|
|
|
|
if (!dev_cap) {
|
|
|
|
|
err = -ENOMEM;
|
|
|
|
|
goto err_fw;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = mlx4_QUERY_DEV_CAP(dev, dev_cap);
|
|
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "QUERY_DEV_CAP command failed, aborting.\n");
|
|
|
|
|
goto err_fw;
|
|
|
|
|
}
|
|
|
|
|
|
2014-11-13 12:45:33 +00:00
|
|
|
if (mlx4_check_dev_cap(dev, dev_cap, nvfs))
|
|
|
|
|
goto err_fw;
|
|
|
|
|
|
2014-11-13 12:45:32 +00:00
|
|
|
if (!(dev_cap->flags2 & MLX4_DEV_CAP_FLAG2_SYS_EQS)) {
|
2015-01-25 14:59:42 +00:00
|
|
|
u64 dev_flags = mlx4_enable_sriov(dev, pdev,
|
|
|
|
|
total_vfs,
|
|
|
|
|
existing_vfs,
|
|
|
|
|
reset_flow);
|
2014-11-13 12:45:32 +00:00
|
|
|
|
2015-06-02 21:07:24 +00:00
|
|
|
mlx4_close_fw(dev);
|
2014-11-13 12:45:32 +00:00
|
|
|
mlx4_cmd_cleanup(dev, MLX4_CMD_CLEANUP_ALL);
|
|
|
|
|
dev->flags = dev_flags;
|
|
|
|
|
if (!SRIOV_VALID_STATE(dev->flags)) {
|
|
|
|
|
mlx4_err(dev, "Invalid SRIOV state\n");
|
|
|
|
|
goto err_sriov;
|
|
|
|
|
}
|
|
|
|
|
err = mlx4_reset(dev);
|
|
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "Failed to reset HCA, aborting.\n");
|
|
|
|
|
goto err_sriov;
|
|
|
|
|
}
|
|
|
|
|
goto slave_start;
|
|
|
|
|
}
|
|
|
|
|
} else {
|
|
|
|
|
/* Legacy mode FW requires SRIOV to be enabled before
|
|
|
|
|
* doing QUERY_DEV_CAP, since max_eq's value is different if
|
|
|
|
|
* SRIOV is enabled.
|
|
|
|
|
*/
|
|
|
|
|
memset(dev_cap, 0, sizeof(*dev_cap));
|
|
|
|
|
err = mlx4_QUERY_DEV_CAP(dev, dev_cap);
|
|
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "QUERY_DEV_CAP command failed, aborting.\n");
|
|
|
|
|
goto err_fw;
|
|
|
|
|
}
|
2014-11-13 12:45:33 +00:00
|
|
|
|
|
|
|
|
if (mlx4_check_dev_cap(dev, dev_cap, nvfs))
|
|
|
|
|
goto err_fw;
|
2014-11-13 12:45:32 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
err = mlx4_init_hca(dev);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
if (err) {
|
|
|
|
|
if (err == -EACCES) {
|
|
|
|
|
/* Not primary Physical function
|
|
|
|
|
* Running in slave mode */
|
2014-11-13 12:45:29 +00:00
|
|
|
mlx4_cmd_cleanup(dev, MLX4_CMD_CLEANUP_ALL);
|
2014-11-13 12:45:30 +00:00
|
|
|
/* We're not a PF */
|
|
|
|
|
if (dev->flags & MLX4_FLAG_SRIOV) {
|
|
|
|
|
if (!existing_vfs)
|
|
|
|
|
pci_disable_sriov(pdev);
|
2015-01-25 14:59:42 +00:00
|
|
|
if (mlx4_is_master(dev) && !reset_flow)
|
2014-11-13 12:45:30 +00:00
|
|
|
atomic_dec(&pf_loading);
|
|
|
|
|
dev->flags &= ~MLX4_FLAG_SRIOV;
|
|
|
|
|
}
|
|
|
|
|
if (!mlx4_is_slave(dev))
|
|
|
|
|
mlx4_free_ownership(dev);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
dev->flags |= MLX4_FLAG_SLAVE;
|
|
|
|
|
dev->flags &= ~MLX4_FLAG_MASTER;
|
|
|
|
|
goto slave_start;
|
|
|
|
|
} else
|
2014-11-13 12:45:30 +00:00
|
|
|
goto err_fw;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
}
|
|
|
|
|
|
2014-11-13 12:45:32 +00:00
|
|
|
if (mlx4_is_master(dev) && (dev_cap->flags2 & MLX4_DEV_CAP_FLAG2_SYS_EQS)) {
|
2015-01-25 14:59:42 +00:00
|
|
|
u64 dev_flags = mlx4_enable_sriov(dev, pdev, total_vfs,
|
|
|
|
|
existing_vfs, reset_flow);
|
2014-11-13 12:45:32 +00:00
|
|
|
|
|
|
|
|
if ((dev->flags ^ dev_flags) & (MLX4_FLAG_MASTER | MLX4_FLAG_SLAVE)) {
|
|
|
|
|
mlx4_cmd_cleanup(dev, MLX4_CMD_CLEANUP_VHCR);
|
|
|
|
|
dev->flags = dev_flags;
|
|
|
|
|
err = mlx4_cmd_init(dev);
|
|
|
|
|
if (err) {
|
|
|
|
|
/* Only VHCR is cleaned up, so could still
|
|
|
|
|
* send FW commands
|
|
|
|
|
*/
|
|
|
|
|
mlx4_err(dev, "Failed to init VHCR command interface, aborting\n");
|
|
|
|
|
goto err_close;
|
|
|
|
|
}
|
|
|
|
|
} else {
|
|
|
|
|
dev->flags = dev_flags;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (!SRIOV_VALID_STATE(dev->flags)) {
|
|
|
|
|
mlx4_err(dev, "Invalid SRIOV state\n");
|
|
|
|
|
goto err_close;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2014-01-05 15:41:08 +00:00
|
|
|
/* check if the device is functioning at its maximum possible speed.
|
|
|
|
|
* No return code for this call, just warn the user in case of PCI
|
|
|
|
|
* express device capabilities are under-satisfied by the bus.
|
|
|
|
|
*/
|
2014-05-04 14:07:25 +00:00
|
|
|
if (!mlx4_is_slave(dev))
|
|
|
|
|
mlx4_check_pcie_caps(dev);
|
2014-01-05 15:41:08 +00:00
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/* In master functions, the communication channel must be initialized
|
|
|
|
|
* after obtaining its address from fw */
|
|
|
|
|
if (mlx4_is_master(dev)) {
|
2014-09-30 09:03:48 +00:00
|
|
|
if (dev->caps.num_ports < 2 &&
|
|
|
|
|
num_vfs_argc > 1) {
|
|
|
|
|
err = -EINVAL;
|
|
|
|
|
mlx4_err(dev,
|
|
|
|
|
"Error: Trying to configure VFs on port 2, but HCA has only %d physical ports\n",
|
|
|
|
|
dev->caps.num_ports);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
goto err_close;
|
|
|
|
|
}
|
2015-01-25 14:59:35 +00:00
|
|
|
memcpy(dev->persist->nvfs, nvfs, sizeof(dev->persist->nvfs));
|
2014-03-19 16:11:53 +00:00
|
|
|
|
2015-01-25 14:59:35 +00:00
|
|
|
for (i = 0;
|
|
|
|
|
i < sizeof(dev->persist->nvfs)/
|
|
|
|
|
sizeof(dev->persist->nvfs[0]); i++) {
|
2014-09-30 09:03:48 +00:00
|
|
|
unsigned j;
|
|
|
|
|
|
2015-01-25 14:59:35 +00:00
|
|
|
for (j = 0; j < dev->persist->nvfs[i]; ++sum, ++j) {
|
2014-09-30 09:03:48 +00:00
|
|
|
dev->dev_vfs[sum].min_port = i < 2 ? i + 1 : 1;
|
|
|
|
|
dev->dev_vfs[sum].n_ports = i < 2 ? 1 :
|
|
|
|
|
dev->caps.num_ports;
|
2014-03-19 16:11:50 +00:00
|
|
|
}
|
|
|
|
|
}
|
2014-09-30 09:03:48 +00:00
|
|
|
|
|
|
|
|
/* In master functions, the communication channel
|
|
|
|
|
* must be initialized after obtaining its address from fw
|
|
|
|
|
*/
|
|
|
|
|
err = mlx4_multi_func_init(dev);
|
|
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "Failed to init master mfunc interface, aborting.\n");
|
|
|
|
|
goto err_close;
|
|
|
|
|
}
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
}
|
2007-05-09 01:00:38 +00:00
|
|
|
|
2008-12-22 15:15:03 +00:00
|
|
|
err = mlx4_alloc_eq_table(dev);
|
|
|
|
|
if (err)
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
goto err_master_mfunc;
|
2008-12-22 15:15:03 +00:00
|
|
|
|
2015-05-31 06:30:16 +00:00
|
|
|
bitmap_zero(priv->msix_ctl.pool_bm, MAX_MSIX);
|
2012-02-21 03:39:32 +00:00
|
|
|
mutex_init(&priv->msix_ctl.pool_lock);
|
2011-03-22 22:37:47 +00:00
|
|
|
|
2007-08-07 13:08:28 +00:00
|
|
|
mlx4_enable_msi_x(dev);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
if ((mlx4_is_mfunc(dev)) &&
|
|
|
|
|
!(dev->flags & MLX4_FLAG_MSI_X)) {
|
2013-01-24 01:54:17 +00:00
|
|
|
err = -ENOSYS;
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_err(dev, "INTx is not supported in multi-function mode, aborting\n");
|
2011-03-22 22:38:24 +00:00
|
|
|
goto err_free_eq;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (!mlx4_is_slave(dev)) {
|
|
|
|
|
err = mlx4_init_steering(dev);
|
|
|
|
|
if (err)
|
2014-09-30 09:03:48 +00:00
|
|
|
goto err_disable_msix;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
}
|
2011-03-22 22:38:24 +00:00
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
err = mlx4_setup_hca(dev);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
if (err == -EBUSY && (dev->flags & MLX4_FLAG_MSI_X) &&
|
|
|
|
|
!mlx4_is_mfunc(dev)) {
|
2007-08-07 13:08:28 +00:00
|
|
|
dev->flags &= ~MLX4_FLAG_MSI_X;
|
2012-06-25 00:24:12 +00:00
|
|
|
dev->caps.num_comp_vectors = 1;
|
2007-08-07 13:08:28 +00:00
|
|
|
pci_disable_msix(pdev);
|
|
|
|
|
err = mlx4_setup_hca(dev);
|
|
|
|
|
}
|
|
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
if (err)
|
2011-03-22 22:38:24 +00:00
|
|
|
goto err_steer;
|
2007-05-09 01:00:38 +00:00
|
|
|
|
mlx4: Structures and init/teardown for VF resource quotas
This is step #1 for implementing SRIOV resource quotas for VFs.
Quotas are implemented per resource type for VFs and the PF, to prevent
any entity from simply grabbing all the resources for itself and leaving
the other entities unable to obtain such resources.
Resources which are allocated using quotas: QPs, CQs, SRQs, MPTs, MTTs, MAC,
VLAN, and Counters.
The quota system works as follows:
Each entity (VF or PF) is given a max number of a given resource (its quota),
and a guaranteed minimum number for each resource (starvation prevention).
For QPs, CQs, SRQs, MPTs and MTTs:
50% of the available quantity for the resource is divided equally among
the PF and all the active VFs (i.e., the number of VFs in the mlx4_core module
parameter "num_vfs"). This 50% represents the "guaranteed minimum" pool.
The other 50% is the "free pool", allocated on a first-come-first-serve basis.
For each VF/PF, resources are first allocated from its "guaranteed-minimum"
pool. When that pool is exhausted, the driver attempts to allocate from
the resource "free-pool".
The quota (i.e., max) for the VFs and the PF is:
The free-pool amount (50% of the real max) + the guaranteed minimum
For MACs:
Guarantee 2 MACs per VF/PF per port. As a result, since we have only
128 MACs per port, reduce the allowable number of VFs from 64 to 63.
Any remaining MACs are put into a free pool.
For VLANs:
For the PF, the per-port quota is 128 and guarantee is 64
(to allow the PF to register at least a VLAN per VF in VST mode).
For the VFs, the per-port quota is 64 and the guarantee is 0.
We assume that VGT VFs are trusted not to abuse the VLAN resource.
For Counters:
For all functions (PF and VFs), the quota is 128 and the guarantee is 0.
In this patch, we define the needed structures, which are added to the
resource-tracker struct. In addition, we do initialization
for the resource quota, and adjust the query_device response to use quotas
rather than resource maxima.
As part of the implementation, we introduce a new field in
mlx4_dev: quotas. This field holds the resource quotas used
to report maxima to the upper layers (ib_core, via query_device).
The HCA maxima of these values are passed to the VFs (via
QUERY_HCA) so that they may continue to use these in handling
QPs, CQs, SRQs and MPTs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 08:03:23 +00:00
|
|
|
mlx4_init_quotas(dev);
|
2015-01-25 14:59:42 +00:00
|
|
|
/* When PF resources are ready arm its comm channel to enable
|
|
|
|
|
* getting commands
|
|
|
|
|
*/
|
|
|
|
|
if (mlx4_is_master(dev)) {
|
|
|
|
|
err = mlx4_ARM_COMM_CHANNEL(dev);
|
|
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, " Failed to arm comm channel eq: %x\n",
|
|
|
|
|
err);
|
|
|
|
|
goto err_steer;
|
|
|
|
|
}
|
|
|
|
|
}
|
mlx4: Structures and init/teardown for VF resource quotas
This is step #1 for implementing SRIOV resource quotas for VFs.
Quotas are implemented per resource type for VFs and the PF, to prevent
any entity from simply grabbing all the resources for itself and leaving
the other entities unable to obtain such resources.
Resources which are allocated using quotas: QPs, CQs, SRQs, MPTs, MTTs, MAC,
VLAN, and Counters.
The quota system works as follows:
Each entity (VF or PF) is given a max number of a given resource (its quota),
and a guaranteed minimum number for each resource (starvation prevention).
For QPs, CQs, SRQs, MPTs and MTTs:
50% of the available quantity for the resource is divided equally among
the PF and all the active VFs (i.e., the number of VFs in the mlx4_core module
parameter "num_vfs"). This 50% represents the "guaranteed minimum" pool.
The other 50% is the "free pool", allocated on a first-come-first-serve basis.
For each VF/PF, resources are first allocated from its "guaranteed-minimum"
pool. When that pool is exhausted, the driver attempts to allocate from
the resource "free-pool".
The quota (i.e., max) for the VFs and the PF is:
The free-pool amount (50% of the real max) + the guaranteed minimum
For MACs:
Guarantee 2 MACs per VF/PF per port. As a result, since we have only
128 MACs per port, reduce the allowable number of VFs from 64 to 63.
Any remaining MACs are put into a free pool.
For VLANs:
For the PF, the per-port quota is 128 and guarantee is 64
(to allow the PF to register at least a VLAN per VF in VST mode).
For the VFs, the per-port quota is 64 and the guarantee is 0.
We assume that VGT VFs are trusted not to abuse the VLAN resource.
For Counters:
For all functions (PF and VFs), the quota is 128 and the guarantee is 0.
In this patch, we define the needed structures, which are added to the
resource-tracker struct. In addition, we do initialization
for the resource quota, and adjust the query_device response to use quotas
rather than resource maxima.
As part of the implementation, we introduce a new field in
mlx4_dev: quotas. This field holds the resource quotas used
to report maxima to the upper layers (ib_core, via query_device).
The HCA maxima of these values are passed to the VFs (via
QUERY_HCA) so that they may continue to use these in handling
QPs, CQs, SRQs and MPTs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 08:03:23 +00:00
|
|
|
|
2008-10-22 22:38:42 +00:00
|
|
|
for (port = 1; port <= dev->caps.num_ports; port++) {
|
|
|
|
|
err = mlx4_init_port_info(dev, port);
|
|
|
|
|
if (err)
|
|
|
|
|
goto err_port;
|
|
|
|
|
}
|
2008-10-22 18:44:46 +00:00
|
|
|
|
2015-02-03 14:48:33 +00:00
|
|
|
priv->v2p.port1 = 1;
|
|
|
|
|
priv->v2p.port2 = 2;
|
|
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
err = mlx4_register_device(dev);
|
|
|
|
|
if (err)
|
2008-10-22 22:38:42 +00:00
|
|
|
goto err_port;
|
2007-05-09 01:00:38 +00:00
|
|
|
|
2013-10-15 14:55:24 +00:00
|
|
|
mlx4_request_modules(dev);
|
|
|
|
|
|
2009-03-19 02:45:11 +00:00
|
|
|
mlx4_sense_init(dev);
|
|
|
|
|
mlx4_start_sense(dev);
|
|
|
|
|
|
2014-04-14 01:51:19 +00:00
|
|
|
priv->removed = 0;
|
2007-05-09 01:00:38 +00:00
|
|
|
|
2015-01-25 14:59:42 +00:00
|
|
|
if (mlx4_is_master(dev) && dev->persist->num_vfs && !reset_flow)
|
2014-04-14 08:17:22 +00:00
|
|
|
atomic_dec(&pf_loading);
|
|
|
|
|
|
2014-12-14 14:18:04 +00:00
|
|
|
kfree(dev_cap);
|
2007-05-09 01:00:38 +00:00
|
|
|
return 0;
|
|
|
|
|
|
2008-10-22 22:38:42 +00:00
|
|
|
err_port:
|
2010-01-06 20:54:39 +00:00
|
|
|
for (--port; port >= 1; --port)
|
2008-10-22 22:38:42 +00:00
|
|
|
mlx4_cleanup_port_info(&priv->port[port]);
|
|
|
|
|
|
2015-06-15 14:59:02 +00:00
|
|
|
mlx4_cleanup_default_counters(dev);
|
2015-06-15 14:58:59 +00:00
|
|
|
if (!mlx4_is_slave(dev))
|
|
|
|
|
mlx4_cleanup_counters_table(dev);
|
2007-05-09 01:00:38 +00:00
|
|
|
mlx4_cleanup_qp_table(dev);
|
|
|
|
|
mlx4_cleanup_srq_table(dev);
|
|
|
|
|
mlx4_cleanup_cq_table(dev);
|
|
|
|
|
mlx4_cmd_use_polling(dev);
|
|
|
|
|
mlx4_cleanup_eq_table(dev);
|
2013-07-28 15:54:21 +00:00
|
|
|
mlx4_cleanup_mcg_table(dev);
|
2007-05-09 01:00:38 +00:00
|
|
|
mlx4_cleanup_mr_table(dev);
|
2011-06-02 16:01:33 +00:00
|
|
|
mlx4_cleanup_xrcd_table(dev);
|
2007-05-09 01:00:38 +00:00
|
|
|
mlx4_cleanup_pd_table(dev);
|
|
|
|
|
mlx4_cleanup_uar_table(dev);
|
|
|
|
|
|
2011-03-22 22:38:24 +00:00
|
|
|
err_steer:
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
if (!mlx4_is_slave(dev))
|
|
|
|
|
mlx4_clear_steering(dev);
|
2011-03-22 22:38:24 +00:00
|
|
|
|
2014-09-30 09:03:48 +00:00
|
|
|
err_disable_msix:
|
|
|
|
|
if (dev->flags & MLX4_FLAG_MSI_X)
|
|
|
|
|
pci_disable_msix(pdev);
|
|
|
|
|
|
2008-12-22 15:15:03 +00:00
|
|
|
err_free_eq:
|
|
|
|
|
mlx4_free_eq_table(dev);
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
err_master_mfunc:
|
2015-01-27 13:58:01 +00:00
|
|
|
if (mlx4_is_master(dev)) {
|
|
|
|
|
mlx4_free_resource_tracker(dev, RES_TR_FREE_STRUCTS_ONLY);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
mlx4_multi_func_cleanup(dev);
|
2015-01-27 13:58:01 +00:00
|
|
|
}
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
|
2014-05-29 13:30:59 +00:00
|
|
|
if (mlx4_is_slave(dev)) {
|
2014-05-29 13:31:03 +00:00
|
|
|
kfree(dev->caps.qp0_qkey);
|
2014-05-29 13:30:59 +00:00
|
|
|
kfree(dev->caps.qp0_tunnel);
|
|
|
|
|
kfree(dev->caps.qp0_proxy);
|
|
|
|
|
kfree(dev->caps.qp1_tunnel);
|
|
|
|
|
kfree(dev->caps.qp1_proxy);
|
|
|
|
|
}
|
|
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
err_close:
|
|
|
|
|
mlx4_close_hca(dev);
|
|
|
|
|
|
2014-11-13 12:45:30 +00:00
|
|
|
err_fw:
|
|
|
|
|
mlx4_close_fw(dev);
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
err_mfunc:
|
|
|
|
|
if (mlx4_is_slave(dev))
|
|
|
|
|
mlx4_multi_func_cleanup(dev);
|
|
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
err_cmd:
|
2014-11-13 12:45:29 +00:00
|
|
|
mlx4_cmd_cleanup(dev, MLX4_CMD_CLEANUP_ALL);
|
2007-05-09 01:00:38 +00:00
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
err_sriov:
|
2015-01-25 14:59:42 +00:00
|
|
|
if (dev->flags & MLX4_FLAG_SRIOV && !existing_vfs) {
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
pci_disable_sriov(pdev);
|
2015-01-25 14:59:42 +00:00
|
|
|
dev->flags &= ~MLX4_FLAG_SRIOV;
|
|
|
|
|
}
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
|
2015-01-25 14:59:42 +00:00
|
|
|
if (mlx4_is_master(dev) && dev->persist->num_vfs && !reset_flow)
|
2014-04-14 08:17:22 +00:00
|
|
|
atomic_dec(&pf_loading);
|
|
|
|
|
|
2014-03-19 16:11:50 +00:00
|
|
|
kfree(priv->dev.dev_vfs);
|
|
|
|
|
|
2014-09-30 09:03:48 +00:00
|
|
|
if (!mlx4_is_slave(dev))
|
|
|
|
|
mlx4_free_ownership(dev);
|
|
|
|
|
|
2014-11-13 12:45:32 +00:00
|
|
|
kfree(dev_cap);
|
2014-09-30 09:03:48 +00:00
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int __mlx4_init_one(struct pci_dev *pdev, int pci_dev_data,
|
|
|
|
|
struct mlx4_priv *priv)
|
|
|
|
|
{
|
|
|
|
|
int err;
|
|
|
|
|
int nvfs[MLX4_MAX_PORTS + 1] = {0, 0, 0};
|
|
|
|
|
int prb_vf[MLX4_MAX_PORTS + 1] = {0, 0, 0};
|
|
|
|
|
const int param_map[MLX4_MAX_PORTS + 1][MLX4_MAX_PORTS + 1] = {
|
|
|
|
|
{2, 0, 0}, {0, 1, 2}, {0, 1, 2} };
|
|
|
|
|
unsigned total_vfs = 0;
|
|
|
|
|
unsigned int i;
|
|
|
|
|
|
|
|
|
|
pr_info(DRV_NAME ": Initializing %s\n", pci_name(pdev));
|
|
|
|
|
|
2016-04-20 13:01:16 +00:00
|
|
|
err = mlx4_pci_enable_device(&priv->dev);
|
2014-09-30 09:03:48 +00:00
|
|
|
if (err) {
|
|
|
|
|
dev_err(&pdev->dev, "Cannot enable PCI device, aborting\n");
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Due to requirement that all VFs and the PF are *guaranteed* 2 MACS
|
|
|
|
|
* per port, we must limit the number of VFs to 63 (since their are
|
|
|
|
|
* 128 MACs)
|
|
|
|
|
*/
|
|
|
|
|
for (i = 0; i < sizeof(nvfs)/sizeof(nvfs[0]) && i < num_vfs_argc;
|
|
|
|
|
total_vfs += nvfs[param_map[num_vfs_argc - 1][i]], i++) {
|
|
|
|
|
nvfs[param_map[num_vfs_argc - 1][i]] = num_vfs[i];
|
|
|
|
|
if (nvfs[i] < 0) {
|
|
|
|
|
dev_err(&pdev->dev, "num_vfs module parameter cannot be negative\n");
|
|
|
|
|
err = -EINVAL;
|
|
|
|
|
goto err_disable_pdev;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
for (i = 0; i < sizeof(prb_vf)/sizeof(prb_vf[0]) && i < probe_vfs_argc;
|
|
|
|
|
i++) {
|
|
|
|
|
prb_vf[param_map[probe_vfs_argc - 1][i]] = probe_vf[i];
|
|
|
|
|
if (prb_vf[i] < 0 || prb_vf[i] > nvfs[i]) {
|
|
|
|
|
dev_err(&pdev->dev, "probe_vf module parameter cannot be negative or greater than num_vfs\n");
|
|
|
|
|
err = -EINVAL;
|
|
|
|
|
goto err_disable_pdev;
|
|
|
|
|
}
|
|
|
|
|
}
|
2015-07-06 14:20:19 +00:00
|
|
|
if (total_vfs > MLX4_MAX_NUM_VF) {
|
2014-09-30 09:03:48 +00:00
|
|
|
dev_err(&pdev->dev,
|
2015-07-06 14:20:19 +00:00
|
|
|
"Requested more VF's (%d) than allowed by hw (%d)\n",
|
|
|
|
|
total_vfs, MLX4_MAX_NUM_VF);
|
2014-09-30 09:03:48 +00:00
|
|
|
err = -EINVAL;
|
|
|
|
|
goto err_disable_pdev;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < MLX4_MAX_PORTS; i++) {
|
2015-07-06 14:20:19 +00:00
|
|
|
if (nvfs[i] + nvfs[2] > MLX4_MAX_NUM_VF_P_PORT) {
|
2014-09-30 09:03:48 +00:00
|
|
|
dev_err(&pdev->dev,
|
2015-07-06 14:20:19 +00:00
|
|
|
"Requested more VF's (%d) for port (%d) than allowed by driver (%d)\n",
|
2014-09-30 09:03:48 +00:00
|
|
|
nvfs[i] + nvfs[2], i + 1,
|
2015-07-06 14:20:19 +00:00
|
|
|
MLX4_MAX_NUM_VF_P_PORT);
|
2014-09-30 09:03:48 +00:00
|
|
|
err = -EINVAL;
|
|
|
|
|
goto err_disable_pdev;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Check for BARs. */
|
|
|
|
|
if (!(pci_dev_data & MLX4_PCI_DEV_IS_VF) &&
|
|
|
|
|
!(pci_resource_flags(pdev, 0) & IORESOURCE_MEM)) {
|
|
|
|
|
dev_err(&pdev->dev, "Missing DCS, aborting (driver_data: 0x%x, pci_resource_flags(pdev, 0):0x%lx)\n",
|
|
|
|
|
pci_dev_data, pci_resource_flags(pdev, 0));
|
|
|
|
|
err = -ENODEV;
|
|
|
|
|
goto err_disable_pdev;
|
|
|
|
|
}
|
|
|
|
|
if (!(pci_resource_flags(pdev, 2) & IORESOURCE_MEM)) {
|
|
|
|
|
dev_err(&pdev->dev, "Missing UAR, aborting\n");
|
|
|
|
|
err = -ENODEV;
|
|
|
|
|
goto err_disable_pdev;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
err = pci_request_regions(pdev, DRV_NAME);
|
|
|
|
|
if (err) {
|
|
|
|
|
dev_err(&pdev->dev, "Couldn't get PCI resources, aborting\n");
|
|
|
|
|
goto err_disable_pdev;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
pci_set_master(pdev);
|
|
|
|
|
|
|
|
|
|
err = pci_set_dma_mask(pdev, DMA_BIT_MASK(64));
|
|
|
|
|
if (err) {
|
|
|
|
|
dev_warn(&pdev->dev, "Warning: couldn't set 64-bit PCI DMA mask\n");
|
|
|
|
|
err = pci_set_dma_mask(pdev, DMA_BIT_MASK(32));
|
|
|
|
|
if (err) {
|
|
|
|
|
dev_err(&pdev->dev, "Can't set PCI DMA mask, aborting\n");
|
|
|
|
|
goto err_release_regions;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
err = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(64));
|
|
|
|
|
if (err) {
|
|
|
|
|
dev_warn(&pdev->dev, "Warning: couldn't set 64-bit consistent PCI DMA mask\n");
|
|
|
|
|
err = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32));
|
|
|
|
|
if (err) {
|
|
|
|
|
dev_err(&pdev->dev, "Can't set consistent PCI DMA mask, aborting\n");
|
|
|
|
|
goto err_release_regions;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Allow large DMA segments, up to the firmware limit of 1 GB */
|
|
|
|
|
dma_set_max_seg_size(&pdev->dev, 1024 * 1024 * 1024);
|
|
|
|
|
/* Detect if this device is a virtual function */
|
|
|
|
|
if (pci_dev_data & MLX4_PCI_DEV_IS_VF) {
|
|
|
|
|
/* When acting as pf, we normally skip vfs unless explicitly
|
|
|
|
|
* requested to probe them.
|
|
|
|
|
*/
|
|
|
|
|
if (total_vfs) {
|
|
|
|
|
unsigned vfs_offset = 0;
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < sizeof(nvfs)/sizeof(nvfs[0]) &&
|
|
|
|
|
vfs_offset + nvfs[i] < extended_func_num(pdev);
|
|
|
|
|
vfs_offset += nvfs[i], i++)
|
|
|
|
|
;
|
|
|
|
|
if (i == sizeof(nvfs)/sizeof(nvfs[0])) {
|
|
|
|
|
err = -ENODEV;
|
|
|
|
|
goto err_release_regions;
|
|
|
|
|
}
|
|
|
|
|
if ((extended_func_num(pdev) - vfs_offset)
|
|
|
|
|
> prb_vf[i]) {
|
|
|
|
|
dev_warn(&pdev->dev, "Skipping virtual function:%d\n",
|
|
|
|
|
extended_func_num(pdev));
|
|
|
|
|
err = -ENODEV;
|
|
|
|
|
goto err_release_regions;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2015-01-25 14:59:37 +00:00
|
|
|
err = mlx4_catas_init(&priv->dev);
|
2014-09-30 09:03:48 +00:00
|
|
|
if (err)
|
|
|
|
|
goto err_release_regions;
|
2015-01-25 14:59:37 +00:00
|
|
|
|
2015-01-25 14:59:42 +00:00
|
|
|
err = mlx4_load_one(pdev, pci_dev_data, total_vfs, nvfs, priv, 0);
|
2015-01-25 14:59:37 +00:00
|
|
|
if (err)
|
|
|
|
|
goto err_catas;
|
|
|
|
|
|
2014-09-30 09:03:48 +00:00
|
|
|
return 0;
|
2007-05-09 01:00:38 +00:00
|
|
|
|
2015-01-25 14:59:37 +00:00
|
|
|
err_catas:
|
|
|
|
|
mlx4_catas_end(&priv->dev);
|
|
|
|
|
|
2009-09-06 03:24:48 +00:00
|
|
|
err_release_regions:
|
|
|
|
|
pci_release_regions(pdev);
|
2007-05-09 01:00:38 +00:00
|
|
|
|
|
|
|
|
err_disable_pdev:
|
2016-04-20 13:01:16 +00:00
|
|
|
mlx4_pci_disable_device(&priv->dev);
|
2007-05-09 01:00:38 +00:00
|
|
|
pci_set_drvdata(pdev, NULL);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2016-02-26 16:32:25 +00:00
|
|
|
static int mlx4_devlink_port_type_set(struct devlink_port *devlink_port,
|
|
|
|
|
enum devlink_port_type port_type)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_port_info *info = container_of(devlink_port,
|
|
|
|
|
struct mlx4_port_info,
|
|
|
|
|
devlink_port);
|
|
|
|
|
enum mlx4_port_type mlx4_port_type;
|
|
|
|
|
|
|
|
|
|
switch (port_type) {
|
|
|
|
|
case DEVLINK_PORT_TYPE_AUTO:
|
|
|
|
|
mlx4_port_type = MLX4_PORT_TYPE_AUTO;
|
|
|
|
|
break;
|
|
|
|
|
case DEVLINK_PORT_TYPE_ETH:
|
|
|
|
|
mlx4_port_type = MLX4_PORT_TYPE_ETH;
|
|
|
|
|
break;
|
|
|
|
|
case DEVLINK_PORT_TYPE_IB:
|
|
|
|
|
mlx4_port_type = MLX4_PORT_TYPE_IB;
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
return -EOPNOTSUPP;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return __set_port_type(info, mlx4_port_type);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static const struct devlink_ops mlx4_devlink_ops = {
|
|
|
|
|
.port_type_set = mlx4_devlink_port_type_set,
|
|
|
|
|
};
|
|
|
|
|
|
2012-12-06 14:30:56 +00:00
|
|
|
static int mlx4_init_one(struct pci_dev *pdev, const struct pci_device_id *id)
|
2007-10-10 22:43:54 +00:00
|
|
|
{
|
2016-02-26 16:32:24 +00:00
|
|
|
struct devlink *devlink;
|
2014-04-14 01:51:19 +00:00
|
|
|
struct mlx4_priv *priv;
|
|
|
|
|
struct mlx4_dev *dev;
|
2014-09-30 09:03:48 +00:00
|
|
|
int ret;
|
2014-04-14 01:51:19 +00:00
|
|
|
|
2010-07-10 07:22:46 +00:00
|
|
|
printk_once(KERN_INFO "%s", mlx4_version);
|
2007-10-10 22:43:54 +00:00
|
|
|
|
2016-02-26 16:32:25 +00:00
|
|
|
devlink = devlink_alloc(&mlx4_devlink_ops, sizeof(*priv));
|
2016-02-26 16:32:24 +00:00
|
|
|
if (!devlink)
|
2014-04-14 01:51:19 +00:00
|
|
|
return -ENOMEM;
|
2016-02-26 16:32:24 +00:00
|
|
|
priv = devlink_priv(devlink);
|
2014-04-14 01:51:19 +00:00
|
|
|
|
|
|
|
|
dev = &priv->dev;
|
2015-01-25 14:59:35 +00:00
|
|
|
dev->persist = kzalloc(sizeof(*dev->persist), GFP_KERNEL);
|
|
|
|
|
if (!dev->persist) {
|
2016-02-26 16:32:24 +00:00
|
|
|
ret = -ENOMEM;
|
|
|
|
|
goto err_devlink_free;
|
2015-01-25 14:59:35 +00:00
|
|
|
}
|
|
|
|
|
dev->persist->pdev = pdev;
|
|
|
|
|
dev->persist->dev = dev;
|
|
|
|
|
pci_set_drvdata(pdev, dev->persist);
|
2014-04-14 01:51:19 +00:00
|
|
|
priv->pci_dev_data = id->driver_data;
|
2015-01-25 14:59:38 +00:00
|
|
|
mutex_init(&dev->persist->device_state_mutex);
|
2015-01-25 14:59:40 +00:00
|
|
|
mutex_init(&dev->persist->interface_state_mutex);
|
2016-04-20 13:01:16 +00:00
|
|
|
mutex_init(&dev->persist->pci_status_mutex);
|
2014-04-14 01:51:19 +00:00
|
|
|
|
2016-02-26 16:32:24 +00:00
|
|
|
ret = devlink_register(devlink, &pdev->dev);
|
|
|
|
|
if (ret)
|
|
|
|
|
goto err_persist_free;
|
|
|
|
|
|
2014-09-30 09:03:48 +00:00
|
|
|
ret = __mlx4_init_one(pdev, id->driver_data, priv);
|
2016-02-26 16:32:24 +00:00
|
|
|
if (ret)
|
|
|
|
|
goto err_devlink_unregister;
|
2015-01-25 14:59:41 +00:00
|
|
|
|
2016-02-26 16:32:24 +00:00
|
|
|
pci_save_state(pdev);
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
|
|
err_devlink_unregister:
|
|
|
|
|
devlink_unregister(devlink);
|
|
|
|
|
err_persist_free:
|
|
|
|
|
kfree(dev->persist);
|
|
|
|
|
err_devlink_free:
|
|
|
|
|
devlink_free(devlink);
|
2014-09-30 09:03:48 +00:00
|
|
|
return ret;
|
2007-10-10 22:43:54 +00:00
|
|
|
}
|
|
|
|
|
|
2015-01-25 14:59:36 +00:00
|
|
|
static void mlx4_clean_dev(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_dev_persistent *persist = dev->persist;
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
2015-01-25 14:59:42 +00:00
|
|
|
unsigned long flags = (dev->flags & RESET_PERSIST_MASK_FLAGS);
|
2015-01-25 14:59:36 +00:00
|
|
|
|
|
|
|
|
memset(priv, 0, sizeof(*priv));
|
|
|
|
|
priv->dev.persist = persist;
|
2015-01-25 14:59:42 +00:00
|
|
|
priv->dev.flags = flags;
|
2015-01-25 14:59:36 +00:00
|
|
|
}
|
|
|
|
|
|
2014-09-30 09:03:48 +00:00
|
|
|
static void mlx4_unload_one(struct pci_dev *pdev)
|
2007-05-09 01:00:38 +00:00
|
|
|
{
|
2015-01-25 14:59:35 +00:00
|
|
|
struct mlx4_dev_persistent *persist = pci_get_drvdata(pdev);
|
|
|
|
|
struct mlx4_dev *dev = persist->dev;
|
2007-05-09 01:00:38 +00:00
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
2014-04-14 01:51:19 +00:00
|
|
|
int pci_dev_data;
|
2015-01-25 14:59:36 +00:00
|
|
|
int p, i;
|
2007-05-09 01:00:38 +00:00
|
|
|
|
2014-04-14 01:51:19 +00:00
|
|
|
if (priv->removed)
|
|
|
|
|
return;
|
2007-05-09 01:00:38 +00:00
|
|
|
|
2015-01-25 14:59:36 +00:00
|
|
|
/* saving current ports type for further use */
|
|
|
|
|
for (i = 0; i < dev->caps.num_ports; i++) {
|
|
|
|
|
dev->persist->curr_port_type[i] = dev->caps.port_type[i + 1];
|
|
|
|
|
dev->persist->curr_port_poss_type[i] = dev->caps.
|
|
|
|
|
possible_type[i + 1];
|
|
|
|
|
}
|
|
|
|
|
|
2014-04-14 01:51:19 +00:00
|
|
|
pci_dev_data = priv->pci_dev_data;
|
2007-05-09 01:00:38 +00:00
|
|
|
|
2014-04-14 01:51:19 +00:00
|
|
|
mlx4_stop_sense(dev);
|
|
|
|
|
mlx4_unregister_device(dev);
|
2007-05-09 01:00:38 +00:00
|
|
|
|
2014-04-14 01:51:19 +00:00
|
|
|
for (p = 1; p <= dev->caps.num_ports; p++) {
|
|
|
|
|
mlx4_cleanup_port_info(&priv->port[p]);
|
|
|
|
|
mlx4_CLOSE_PORT(dev, p);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (mlx4_is_master(dev))
|
|
|
|
|
mlx4_free_resource_tracker(dev,
|
|
|
|
|
RES_TR_FREE_SLAVES_ONLY);
|
|
|
|
|
|
2015-06-15 14:59:02 +00:00
|
|
|
mlx4_cleanup_default_counters(dev);
|
2015-06-15 14:58:59 +00:00
|
|
|
if (!mlx4_is_slave(dev))
|
|
|
|
|
mlx4_cleanup_counters_table(dev);
|
2014-04-14 01:51:19 +00:00
|
|
|
mlx4_cleanup_qp_table(dev);
|
|
|
|
|
mlx4_cleanup_srq_table(dev);
|
|
|
|
|
mlx4_cleanup_cq_table(dev);
|
|
|
|
|
mlx4_cmd_use_polling(dev);
|
|
|
|
|
mlx4_cleanup_eq_table(dev);
|
|
|
|
|
mlx4_cleanup_mcg_table(dev);
|
|
|
|
|
mlx4_cleanup_mr_table(dev);
|
|
|
|
|
mlx4_cleanup_xrcd_table(dev);
|
|
|
|
|
mlx4_cleanup_pd_table(dev);
|
2007-05-09 01:00:38 +00:00
|
|
|
|
2014-04-14 01:51:19 +00:00
|
|
|
if (mlx4_is_master(dev))
|
|
|
|
|
mlx4_free_resource_tracker(dev,
|
|
|
|
|
RES_TR_FREE_STRUCTS_ONLY);
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 08:40:57 +00:00
|
|
|
|
2014-04-14 01:51:19 +00:00
|
|
|
iounmap(priv->kar);
|
|
|
|
|
mlx4_uar_free(dev, &priv->driver_uar);
|
|
|
|
|
mlx4_cleanup_uar_table(dev);
|
|
|
|
|
if (!mlx4_is_slave(dev))
|
|
|
|
|
mlx4_clear_steering(dev);
|
|
|
|
|
mlx4_free_eq_table(dev);
|
|
|
|
|
if (mlx4_is_master(dev))
|
|
|
|
|
mlx4_multi_func_cleanup(dev);
|
|
|
|
|
mlx4_close_hca(dev);
|
2014-11-13 12:45:30 +00:00
|
|
|
mlx4_close_fw(dev);
|
2014-04-14 01:51:19 +00:00
|
|
|
if (mlx4_is_slave(dev))
|
|
|
|
|
mlx4_multi_func_cleanup(dev);
|
2014-11-13 12:45:29 +00:00
|
|
|
mlx4_cmd_cleanup(dev, MLX4_CMD_CLEANUP_ALL);
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 08:40:57 +00:00
|
|
|
|
2014-04-14 01:51:19 +00:00
|
|
|
if (dev->flags & MLX4_FLAG_MSI_X)
|
|
|
|
|
pci_disable_msix(pdev);
|
|
|
|
|
|
|
|
|
|
if (!mlx4_is_slave(dev))
|
|
|
|
|
mlx4_free_ownership(dev);
|
|
|
|
|
|
2014-05-29 13:31:03 +00:00
|
|
|
kfree(dev->caps.qp0_qkey);
|
2014-04-14 01:51:19 +00:00
|
|
|
kfree(dev->caps.qp0_tunnel);
|
|
|
|
|
kfree(dev->caps.qp0_proxy);
|
|
|
|
|
kfree(dev->caps.qp1_tunnel);
|
|
|
|
|
kfree(dev->caps.qp1_proxy);
|
|
|
|
|
kfree(dev->dev_vfs);
|
|
|
|
|
|
2015-01-25 14:59:36 +00:00
|
|
|
mlx4_clean_dev(dev);
|
2014-04-14 01:51:19 +00:00
|
|
|
priv->pci_dev_data = pci_dev_data;
|
|
|
|
|
priv->removed = 1;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void mlx4_remove_one(struct pci_dev *pdev)
|
|
|
|
|
{
|
2015-01-25 14:59:35 +00:00
|
|
|
struct mlx4_dev_persistent *persist = pci_get_drvdata(pdev);
|
|
|
|
|
struct mlx4_dev *dev = persist->dev;
|
2014-04-14 01:51:19 +00:00
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
2016-02-26 16:32:24 +00:00
|
|
|
struct devlink *devlink = priv_to_devlink(priv);
|
2015-01-25 14:59:42 +00:00
|
|
|
int active_vfs = 0;
|
2014-04-14 01:51:19 +00:00
|
|
|
|
2015-01-25 14:59:40 +00:00
|
|
|
mutex_lock(&persist->interface_state_mutex);
|
|
|
|
|
persist->interface_state |= MLX4_INTERFACE_STATE_DELETION;
|
|
|
|
|
mutex_unlock(&persist->interface_state_mutex);
|
|
|
|
|
|
2015-01-25 14:59:42 +00:00
|
|
|
/* Disabling SR-IOV is not allowed while there are active vf's */
|
|
|
|
|
if (mlx4_is_master(dev) && dev->flags & MLX4_FLAG_SRIOV) {
|
|
|
|
|
active_vfs = mlx4_how_many_lives_vf(dev);
|
|
|
|
|
if (active_vfs) {
|
|
|
|
|
pr_warn("Removing PF when there are active VF's !!\n");
|
|
|
|
|
pr_warn("Will not disable SR-IOV.\n");
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2015-01-25 14:59:40 +00:00
|
|
|
/* device marked to be under deletion running now without the lock
|
|
|
|
|
* letting other tasks to be terminated
|
|
|
|
|
*/
|
|
|
|
|
if (persist->interface_state & MLX4_INTERFACE_STATE_UP)
|
|
|
|
|
mlx4_unload_one(pdev);
|
|
|
|
|
else
|
|
|
|
|
mlx4_info(dev, "%s: interface is down\n", __func__);
|
2015-01-25 14:59:37 +00:00
|
|
|
mlx4_catas_end(dev);
|
2015-01-25 14:59:42 +00:00
|
|
|
if (dev->flags & MLX4_FLAG_SRIOV && !active_vfs) {
|
|
|
|
|
mlx4_warn(dev, "Disabling SR-IOV\n");
|
|
|
|
|
pci_disable_sriov(pdev);
|
|
|
|
|
}
|
|
|
|
|
|
2014-09-30 09:03:48 +00:00
|
|
|
pci_release_regions(pdev);
|
2016-04-20 13:01:16 +00:00
|
|
|
mlx4_pci_disable_device(dev);
|
2016-02-26 16:32:24 +00:00
|
|
|
devlink_unregister(devlink);
|
2015-01-25 14:59:35 +00:00
|
|
|
kfree(dev->persist);
|
2016-02-26 16:32:24 +00:00
|
|
|
devlink_free(devlink);
|
2014-04-14 01:51:19 +00:00
|
|
|
pci_set_drvdata(pdev, NULL);
|
2007-05-09 01:00:38 +00:00
|
|
|
}
|
|
|
|
|
|
2015-01-25 14:59:36 +00:00
|
|
|
static int restore_current_port_types(struct mlx4_dev *dev,
|
|
|
|
|
enum mlx4_port_type *types,
|
|
|
|
|
enum mlx4_port_type *poss_types)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
int err, i;
|
|
|
|
|
|
|
|
|
|
mlx4_stop_sense(dev);
|
|
|
|
|
|
|
|
|
|
mutex_lock(&priv->port_mutex);
|
|
|
|
|
for (i = 0; i < dev->caps.num_ports; i++)
|
|
|
|
|
dev->caps.possible_type[i + 1] = poss_types[i];
|
|
|
|
|
err = mlx4_change_port_types(dev, types);
|
|
|
|
|
mlx4_start_sense(dev);
|
|
|
|
|
mutex_unlock(&priv->port_mutex);
|
|
|
|
|
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2007-07-12 14:50:45 +00:00
|
|
|
int mlx4_restart_one(struct pci_dev *pdev)
|
|
|
|
|
{
|
2015-01-25 14:59:35 +00:00
|
|
|
struct mlx4_dev_persistent *persist = pci_get_drvdata(pdev);
|
|
|
|
|
struct mlx4_dev *dev = persist->dev;
|
2012-09-27 16:23:41 +00:00
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
2014-09-30 09:03:48 +00:00
|
|
|
int nvfs[MLX4_MAX_PORTS + 1] = {0, 0, 0};
|
|
|
|
|
int pci_dev_data, err, total_vfs;
|
2012-09-27 16:23:41 +00:00
|
|
|
|
|
|
|
|
pci_dev_data = priv->pci_dev_data;
|
2015-01-25 14:59:35 +00:00
|
|
|
total_vfs = dev->persist->num_vfs;
|
|
|
|
|
memcpy(nvfs, dev->persist->nvfs, sizeof(dev->persist->nvfs));
|
2014-09-30 09:03:48 +00:00
|
|
|
|
|
|
|
|
mlx4_unload_one(pdev);
|
2015-01-25 14:59:42 +00:00
|
|
|
err = mlx4_load_one(pdev, pci_dev_data, total_vfs, nvfs, priv, 1);
|
2014-09-30 09:03:48 +00:00
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "%s: ERROR: mlx4_load_one failed, pci_name=%s, err=%d\n",
|
|
|
|
|
__func__, pci_name(pdev), err);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2015-01-25 14:59:36 +00:00
|
|
|
err = restore_current_port_types(dev, dev->persist->curr_port_type,
|
|
|
|
|
dev->persist->curr_port_poss_type);
|
|
|
|
|
if (err)
|
|
|
|
|
mlx4_err(dev, "could not restore original port types (%d)\n",
|
|
|
|
|
err);
|
|
|
|
|
|
2014-09-30 09:03:48 +00:00
|
|
|
return err;
|
2007-07-12 14:50:45 +00:00
|
|
|
}
|
|
|
|
|
|
2014-08-08 13:56:03 +00:00
|
|
|
static const struct pci_device_id mlx4_pci_table[] = {
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/* MT25408 "Hermon" SDR */
|
2012-09-27 16:53:05 +00:00
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x6340), MLX4_PCI_DEV_FORCE_SENSE_PORT },
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/* MT25408 "Hermon" DDR */
|
2012-09-27 16:53:05 +00:00
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x634a), MLX4_PCI_DEV_FORCE_SENSE_PORT },
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/* MT25408 "Hermon" QDR */
|
2012-09-27 16:53:05 +00:00
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x6354), MLX4_PCI_DEV_FORCE_SENSE_PORT },
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/* MT25408 "Hermon" DDR PCIe gen2 */
|
2012-09-27 16:53:05 +00:00
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x6732), MLX4_PCI_DEV_FORCE_SENSE_PORT },
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/* MT25408 "Hermon" QDR PCIe gen2 */
|
2012-09-27 16:53:05 +00:00
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x673c), MLX4_PCI_DEV_FORCE_SENSE_PORT },
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/* MT25408 "Hermon" EN 10GigE */
|
2012-09-27 16:53:05 +00:00
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x6368), MLX4_PCI_DEV_FORCE_SENSE_PORT },
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/* MT25408 "Hermon" EN 10GigE PCIe gen2 */
|
2012-09-27 16:53:05 +00:00
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x6750), MLX4_PCI_DEV_FORCE_SENSE_PORT },
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/* MT25458 ConnectX EN 10GBASE-T 10GigE */
|
2012-09-27 16:53:05 +00:00
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x6372), MLX4_PCI_DEV_FORCE_SENSE_PORT },
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/* MT25458 ConnectX EN 10GBASE-T+Gen2 10GigE */
|
2012-09-27 16:53:05 +00:00
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x675a), MLX4_PCI_DEV_FORCE_SENSE_PORT },
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/* MT26468 ConnectX EN 10GigE PCIe gen2*/
|
2012-09-27 16:53:05 +00:00
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x6764), MLX4_PCI_DEV_FORCE_SENSE_PORT },
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/* MT26438 ConnectX EN 40GigE PCIe gen2 5GT/s */
|
2012-09-27 16:53:05 +00:00
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x6746), MLX4_PCI_DEV_FORCE_SENSE_PORT },
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/* MT26478 ConnectX2 40GigE PCIe gen2 */
|
2012-09-27 16:53:05 +00:00
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x676e), MLX4_PCI_DEV_FORCE_SENSE_PORT },
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/* MT25400 Family [ConnectX-2 Virtual Function] */
|
2012-09-27 16:23:41 +00:00
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x1002), MLX4_PCI_DEV_IS_VF },
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/* MT27500 Family [ConnectX-3] */
|
|
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x1003), 0 },
|
|
|
|
|
/* MT27500 Family [ConnectX-3 Virtual Function] */
|
2012-09-27 16:23:41 +00:00
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x1004), MLX4_PCI_DEV_IS_VF },
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x1005), 0 }, /* MT27510 Family */
|
|
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x1006), 0 }, /* MT27511 Family */
|
|
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x1007), 0 }, /* MT27520 Family */
|
|
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x1008), 0 }, /* MT27521 Family */
|
|
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x1009), 0 }, /* MT27530 Family */
|
|
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x100a), 0 }, /* MT27531 Family */
|
|
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x100b), 0 }, /* MT27540 Family */
|
|
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x100c), 0 }, /* MT27541 Family */
|
|
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x100d), 0 }, /* MT27550 Family */
|
|
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x100e), 0 }, /* MT27551 Family */
|
|
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x100f), 0 }, /* MT27560 Family */
|
|
|
|
|
{ PCI_VDEVICE(MELLANOX, 0x1010), 0 }, /* MT27561 Family */
|
2007-05-09 01:00:38 +00:00
|
|
|
{ 0, }
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
MODULE_DEVICE_TABLE(pci, mlx4_pci_table);
|
|
|
|
|
|
2012-07-20 09:55:43 +00:00
|
|
|
static pci_ers_result_t mlx4_pci_err_detected(struct pci_dev *pdev,
|
|
|
|
|
pci_channel_state_t state)
|
|
|
|
|
{
|
2015-01-25 14:59:41 +00:00
|
|
|
struct mlx4_dev_persistent *persist = pci_get_drvdata(pdev);
|
|
|
|
|
|
|
|
|
|
mlx4_err(persist->dev, "mlx4_pci_err_detected was called\n");
|
|
|
|
|
mlx4_enter_error_state(persist);
|
2012-07-20 09:55:43 +00:00
|
|
|
|
2015-01-25 14:59:41 +00:00
|
|
|
mutex_lock(&persist->interface_state_mutex);
|
|
|
|
|
if (persist->interface_state & MLX4_INTERFACE_STATE_UP)
|
|
|
|
|
mlx4_unload_one(pdev);
|
|
|
|
|
|
|
|
|
|
mutex_unlock(&persist->interface_state_mutex);
|
|
|
|
|
if (state == pci_channel_io_perm_failure)
|
|
|
|
|
return PCI_ERS_RESULT_DISCONNECT;
|
|
|
|
|
|
2016-04-20 13:01:16 +00:00
|
|
|
mlx4_pci_disable_device(persist->dev);
|
2015-01-25 14:59:41 +00:00
|
|
|
return PCI_ERS_RESULT_NEED_RESET;
|
2012-07-20 09:55:43 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static pci_ers_result_t mlx4_pci_slot_reset(struct pci_dev *pdev)
|
|
|
|
|
{
|
2015-01-25 14:59:41 +00:00
|
|
|
struct mlx4_dev_persistent *persist = pci_get_drvdata(pdev);
|
|
|
|
|
struct mlx4_dev *dev = persist->dev;
|
2016-04-20 13:01:15 +00:00
|
|
|
int err;
|
2014-03-27 01:28:31 +00:00
|
|
|
|
2015-01-25 14:59:41 +00:00
|
|
|
mlx4_err(dev, "mlx4_pci_slot_reset was called\n");
|
2016-04-20 13:01:16 +00:00
|
|
|
err = mlx4_pci_enable_device(dev);
|
2016-04-20 13:01:15 +00:00
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "Can not re-enable device, err=%d\n", err);
|
2015-01-25 14:59:41 +00:00
|
|
|
return PCI_ERS_RESULT_DISCONNECT;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
pci_set_master(pdev);
|
|
|
|
|
pci_restore_state(pdev);
|
|
|
|
|
pci_save_state(pdev);
|
2016-04-20 13:01:15 +00:00
|
|
|
return PCI_ERS_RESULT_RECOVERED;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void mlx4_pci_resume(struct pci_dev *pdev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_dev_persistent *persist = pci_get_drvdata(pdev);
|
|
|
|
|
struct mlx4_dev *dev = persist->dev;
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
int nvfs[MLX4_MAX_PORTS + 1] = {0, 0, 0};
|
|
|
|
|
int total_vfs;
|
|
|
|
|
int err;
|
2015-01-25 14:59:41 +00:00
|
|
|
|
2016-04-20 13:01:15 +00:00
|
|
|
mlx4_err(dev, "%s was called\n", __func__);
|
2015-01-25 14:59:41 +00:00
|
|
|
total_vfs = dev->persist->num_vfs;
|
|
|
|
|
memcpy(nvfs, dev->persist->nvfs, sizeof(dev->persist->nvfs));
|
|
|
|
|
|
|
|
|
|
mutex_lock(&persist->interface_state_mutex);
|
|
|
|
|
if (!(persist->interface_state & MLX4_INTERFACE_STATE_UP)) {
|
2016-04-20 13:01:15 +00:00
|
|
|
err = mlx4_load_one(pdev, priv->pci_dev_data, total_vfs, nvfs,
|
2015-01-25 14:59:42 +00:00
|
|
|
priv, 1);
|
2016-04-20 13:01:15 +00:00
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "%s: mlx4_load_one failed, err=%d\n",
|
|
|
|
|
__func__, err);
|
2015-01-25 14:59:41 +00:00
|
|
|
goto end;
|
|
|
|
|
}
|
|
|
|
|
|
2016-04-20 13:01:15 +00:00
|
|
|
err = restore_current_port_types(dev, dev->persist->
|
2015-01-25 14:59:41 +00:00
|
|
|
curr_port_type, dev->persist->
|
|
|
|
|
curr_port_poss_type);
|
2016-04-20 13:01:15 +00:00
|
|
|
if (err)
|
|
|
|
|
mlx4_err(dev, "could not restore original port types (%d)\n", err);
|
2015-01-25 14:59:41 +00:00
|
|
|
}
|
|
|
|
|
end:
|
|
|
|
|
mutex_unlock(&persist->interface_state_mutex);
|
2012-07-20 09:55:43 +00:00
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
2015-01-25 14:59:41 +00:00
|
|
|
static void mlx4_shutdown(struct pci_dev *pdev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_dev_persistent *persist = pci_get_drvdata(pdev);
|
|
|
|
|
|
|
|
|
|
mlx4_info(persist->dev, "mlx4_shutdown was called\n");
|
|
|
|
|
mutex_lock(&persist->interface_state_mutex);
|
2016-06-21 11:20:03 +00:00
|
|
|
if (persist->interface_state & MLX4_INTERFACE_STATE_UP) {
|
|
|
|
|
/* Notify mlx4 clients that the kernel is being shut down */
|
|
|
|
|
persist->interface_state |= MLX4_INTERFACE_STATE_SHUTDOWN;
|
2015-01-25 14:59:41 +00:00
|
|
|
mlx4_unload_one(pdev);
|
2016-06-21 11:20:03 +00:00
|
|
|
}
|
2015-01-25 14:59:41 +00:00
|
|
|
mutex_unlock(&persist->interface_state_mutex);
|
|
|
|
|
}
|
|
|
|
|
|
2012-09-07 16:33:15 +00:00
|
|
|
static const struct pci_error_handlers mlx4_err_handler = {
|
2012-07-20 09:55:43 +00:00
|
|
|
.error_detected = mlx4_pci_err_detected,
|
|
|
|
|
.slot_reset = mlx4_pci_slot_reset,
|
2016-04-20 13:01:15 +00:00
|
|
|
.resume = mlx4_pci_resume,
|
2012-07-20 09:55:43 +00:00
|
|
|
};
|
|
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
static struct pci_driver mlx4_driver = {
|
|
|
|
|
.name = DRV_NAME,
|
|
|
|
|
.id_table = mlx4_pci_table,
|
|
|
|
|
.probe = mlx4_init_one,
|
2015-01-25 14:59:41 +00:00
|
|
|
.shutdown = mlx4_shutdown,
|
2012-12-03 14:23:15 +00:00
|
|
|
.remove = mlx4_remove_one,
|
2012-07-20 09:55:43 +00:00
|
|
|
.err_handler = &mlx4_err_handler,
|
2007-05-09 01:00:38 +00:00
|
|
|
};
|
|
|
|
|
|
2008-10-22 22:38:42 +00:00
|
|
|
static int __init mlx4_verify_params(void)
|
|
|
|
|
{
|
|
|
|
|
if ((log_num_mac < 0) || (log_num_mac > 7)) {
|
2014-05-22 12:55:40 +00:00
|
|
|
pr_warn("mlx4_core: bad num_mac: %d\n", log_num_mac);
|
2008-10-22 22:38:42 +00:00
|
|
|
return -1;
|
|
|
|
|
}
|
|
|
|
|
|
2011-10-16 08:26:21 +00:00
|
|
|
if (log_num_vlan != 0)
|
2014-05-22 12:55:40 +00:00
|
|
|
pr_warn("mlx4_core: log_num_vlan - obsolete module param, using %d\n",
|
|
|
|
|
MLX4_LOG_NUM_VLANS);
|
2008-10-22 22:38:42 +00:00
|
|
|
|
2014-05-22 12:55:39 +00:00
|
|
|
if (use_prio != 0)
|
|
|
|
|
pr_warn("mlx4_core: use_prio - obsolete module param, ignored\n");
|
2008-10-22 22:38:42 +00:00
|
|
|
|
2010-09-20 06:42:38 +00:00
|
|
|
if ((log_mtts_per_seg < 1) || (log_mtts_per_seg > 7)) {
|
2014-05-22 12:55:40 +00:00
|
|
|
pr_warn("mlx4_core: bad log_mtts_per_seg: %d\n",
|
|
|
|
|
log_mtts_per_seg);
|
2009-05-27 21:38:34 +00:00
|
|
|
return -1;
|
|
|
|
|
}
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
/* Check if module param for ports type has legal combination */
|
|
|
|
|
if (port_type_array[0] == false && port_type_array[1] == true) {
|
2014-05-22 12:55:40 +00:00
|
|
|
pr_warn("Module parameter configuration ETH/IB is not supported. Switching to default configuration IB/IB\n");
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 04:18:30 +00:00
|
|
|
port_type_array[0] = true;
|
|
|
|
|
}
|
|
|
|
|
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 08:58:00 +00:00
|
|
|
if (mlx4_log_num_mgm_entry_size < -7 ||
|
|
|
|
|
(mlx4_log_num_mgm_entry_size > 0 &&
|
|
|
|
|
(mlx4_log_num_mgm_entry_size < MLX4_MIN_MGM_LOG_ENTRY_SIZE ||
|
|
|
|
|
mlx4_log_num_mgm_entry_size > MLX4_MAX_MGM_LOG_ENTRY_SIZE))) {
|
|
|
|
|
pr_warn("mlx4_core: mlx4_log_num_mgm_entry_size (%d) not in legal range (-7..0 or %d..%d)\n",
|
2014-05-07 19:52:57 +00:00
|
|
|
mlx4_log_num_mgm_entry_size,
|
|
|
|
|
MLX4_MIN_MGM_LOG_ENTRY_SIZE,
|
|
|
|
|
MLX4_MAX_MGM_LOG_ENTRY_SIZE);
|
2012-12-06 17:12:00 +00:00
|
|
|
return -1;
|
|
|
|
|
}
|
|
|
|
|
|
2008-10-22 22:38:42 +00:00
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
static int __init mlx4_init(void)
|
|
|
|
|
{
|
|
|
|
|
int ret;
|
|
|
|
|
|
2008-10-22 22:38:42 +00:00
|
|
|
if (mlx4_verify_params())
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
2009-03-19 02:45:11 +00:00
|
|
|
|
|
|
|
|
mlx4_wq = create_singlethread_workqueue("mlx4");
|
|
|
|
|
if (!mlx4_wq)
|
|
|
|
|
return -ENOMEM;
|
2007-07-12 14:50:45 +00:00
|
|
|
|
2007-05-09 01:00:38 +00:00
|
|
|
ret = pci_register_driver(&mlx4_driver);
|
2013-12-03 02:04:10 +00:00
|
|
|
if (ret < 0)
|
|
|
|
|
destroy_workqueue(mlx4_wq);
|
2007-05-09 01:00:38 +00:00
|
|
|
return ret < 0 ? ret : 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void __exit mlx4_cleanup(void)
|
|
|
|
|
{
|
|
|
|
|
pci_unregister_driver(&mlx4_driver);
|
2009-03-19 02:45:11 +00:00
|
|
|
destroy_workqueue(mlx4_wq);
|
2007-05-09 01:00:38 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
module_init(mlx4_init);
|
|
|
|
|
module_exit(mlx4_cleanup);
|
