2019-05-19 13:51:43 +00:00
// SPDX-License-Identifier: GPL-2.0-or-later
2005-09-20 21:46:58 +00:00
/*
2010-12-15 02:51:48 +00:00
* Linux MegaRAID driver for SAS based RAID controllers
2005-09-20 21:46:58 +00:00
*
2014-11-17 09:54:03 +00:00
* Copyright ( c ) 2003 - 2013 LSI Corporation
2018-10-17 06:37:43 +00:00
* Copyright ( c ) 2013 - 2016 Avago Technologies
* Copyright ( c ) 2016 - 2018 Broadcom Inc .
2005-09-20 21:46:58 +00:00
*
2018-10-17 06:37:43 +00:00
* Authors : Broadcom Inc .
2010-12-15 02:51:48 +00:00
* Sreenivas Bagalkote
* Sumant Patro
* Bo Yang
2014-11-17 09:54:03 +00:00
* Adam Radford
2018-10-17 06:37:43 +00:00
* Kashyap Desai < kashyap . desai @ broadcom . com >
* Sumit Saxena < sumit . saxena @ broadcom . com >
2005-09-20 21:46:58 +00:00
*
2018-10-17 06:37:43 +00:00
* Send feedback to : megaraidlinux . pdl @ broadcom . com
2005-09-20 21:46:58 +00:00
*/
# include <linux/kernel.h>
# include <linux/types.h>
# include <linux/pci.h>
# include <linux/list.h>
# include <linux/moduleparam.h>
# include <linux/module.h>
# include <linux/spinlock.h>
# include <linux/interrupt.h>
# include <linux/delay.h>
# include <linux/uio.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>
2016-12-24 19:46:01 +00:00
# include <linux/uaccess.h>
2017-02-10 08:59:22 +00:00
# include <asm/unaligned.h>
2005-10-04 16:36:04 +00:00
# include <linux/fs.h>
2005-09-20 21:46:58 +00:00
# include <linux/compat.h>
2007-02-14 20:41:55 +00:00
# include <linux/blkdev.h>
2006-01-11 12:16:10 +00:00
# include <linux/mutex.h>
2009-10-06 20:18:02 +00:00
# include <linux/poll.h>
2017-08-23 11:47:03 +00:00
# include <linux/vmalloc.h>
scsi: megaraid_sas: IRQ poll to avoid CPU hard lockups
Issue Description:
We have seen cpu lock up issues from field if system has a large (more than
96) logical cpu count. SAS3.0 controller (Invader series) supports max 96
MSI-X vector and SAS3.5 product (Ventura) supports max 128 MSI-X vectors.
This may be a generic issue (if PCI device support completion on multiple
reply queues).
Let me explain it w.r.t megaraid_sas supported h/w just to simplify the
problem and possible changes to handle such issues. MegaRAID controller
supports multiple reply queues in completion path. Driver creates MSI-X
vectors for controller as "minimum of (FW supported Reply queues, Logical
CPUs)". If submitter is not interrupted via completion on same CPU, there
is a loop in the IO path. This behavior can cause hard/soft CPU lockups, IO
timeout, system sluggish etc.
Example - one CPU (e.g. CPU A) is busy submitting the IOs and another CPU
(e.g. CPU B) is busy with processing the corresponding IO's reply
descriptors from reply descriptor queue upon receiving the interrupts from
HBA. If CPU A is continuously pumping the IOs then always CPU B (which is
executing the ISR) will see the valid reply descriptors in the reply
descriptor queue and it will be continuously processing those reply
descriptor in a loop without quitting the ISR handler.
megaraid_sas driver will exit ISR handler if it finds unused reply
descriptor in the reply descriptor queue. Since CPU A will be continuously
sending the IOs, CPU B may always see a valid reply descriptor (posted by
HBA Firmware after processing the IO) in the reply descriptor queue. In
worst case, driver will not quit from this loop in the ISR handler.
Eventually, CPU lockup will be detected by watchdog.
Above mentioned behavior is not common if "rq_affinity" set to 2 or
affinity_hint is honored by irqbalancer as "exact". If rq_affinity is set
to 2, submitter will be always interrupted via completion on same CPU. If
irqbalancer is using "exact" policy, interrupt will be delivered to
submitter CPU.
Problem statement:
If CPU count to MSI-X vectors (reply descriptor Queues) count ratio is not
1:1, we still have exposure of issue explained above and for that we don't
have any solution.
Exposure of soft/hard lockup is seen if CPU count is more than MSI-X
supported by device.
If CPUs count to MSI-X vectors count ratio is not 1:1, (Other way, if
CPU counts to MSI-X vector count ratio is something like X:1, where X > 1)
then 'exact' irqbalance policy OR rq_affinity = 2 won't help to avoid CPU
hard/soft lockups. There won't be any one to one mapping between
CPU to MSI-X vector instead one MSI-X interrupt (or reply descriptor queue)
is shared with group/set of CPUs and there is a possibility of having a
loop in the IO path within that CPU group and may observe lockups.
For example: Consider a system having two NUMA nodes and each node having
four logical CPUs and also consider that number of MSI-X vectors enabled on
the HBA is two, then CPUs count to MSI-X vector count ratio as 4:1.
e.g.
MSI-X vector 0 is affinity to CPU 0, CPU 1, CPU 2 & CPU 3 of NUMA node 0 and
MSI-X vector 1 is affinity to CPU 4, CPU 5, CPU 6 & CPU 7 of NUMA node 1.
numactl --hardware
available: 2 nodes (0-1)
node 0 cpus: 0 1 2 3 --> MSI-X 0
node 0 size: 65536 MB
node 0 free: 63176 MB
node 1 cpus: 4 5 6 7 --> MSI-X 1
node 1 size: 65536 MB
node 1 free: 63176 MB
Assume that user started an application which uses all the CPUs of NUMA
node 0 for issuing the IOs. Only one CPU from affinity list (it can be any
cpu since this behavior depends upon irqbalance) CPU0 will receive the
interrupts from MSI-X 0 for all the IOs. Eventually, CPU 0 IO submission
percentage will be decreasing and ISR processing percentage will be
increasing as it is more busy with processing the interrupts. Gradually IO
submission percentage on CPU 0 will be zero and it's ISR processing
percentage will be 100% as IO loop has already formed within the
NUMA node 0, i.e. CPU 1, CPU 2 & CPU 3 will be continuously busy with
submitting the heavy IOs and only CPU 0 is busy in the ISR path as it
always find the valid reply descriptor in the reply descriptor queue.
Eventually, we will observe the hard lockup here.
Chances of occurring of hard/soft lockups are directly proportional to
value of X. If value of X is high, then chances of observing CPU lockups is
high.
Solution:
Use IRQ poll interface defined in "irq_poll.c".
megaraid_sas driver will execute ISR routine in softirq context and it will
always quit the loop based on budget provided in IRQ poll interface.
Driver will switch to IRQ poll only when more than a threshold number of
reply descriptors are handled in one ISR. Currently threshold is set as
1/4th of HBA queue depth.
In these scenarios (i.e. where CPUs count to MSI-X vectors count ratio is
X:1 (where X > 1)), IRQ poll interface will avoid CPU hard lockups due to
voluntary exit from the reply queue processing based on budget.
Note - Only one MSI-X vector is busy doing processing.
Select CONFIG_IRQ_POLL from driver Kconfig for driver compilation.
Signed-off-by: Kashyap Desai <kashyap.desai@broadcom.com>
Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-05-07 17:05:35 +00:00
# include <linux/irq_poll.h>
2005-09-20 21:46:58 +00:00
# include <scsi/scsi.h>
# include <scsi/scsi_cmnd.h>
# include <scsi/scsi_device.h>
# include <scsi/scsi_host.h>
2011-07-26 22:42:52 +00:00
# include <scsi/scsi_tcq.h>
2019-05-07 17:05:37 +00:00
# include <scsi/scsi_dbg.h>
2010-12-21 21:34:31 +00:00
# include "megaraid_sas_fusion.h"
2005-09-20 21:46:58 +00:00
# include "megaraid_sas.h"
2010-10-12 13:18:50 +00:00
/*
* Number of sectors per IO command
* Will be set in megasas_init_mfi if user does not provide
*/
static unsigned int max_sectors ;
2019-05-29 16:00:40 +00:00
module_param_named ( max_sectors , max_sectors , int , 0444 ) ;
2010-10-12 13:18:50 +00:00
MODULE_PARM_DESC ( max_sectors ,
" Maximum number of sectors per IO command " ) ;
2010-12-21 18:17:40 +00:00
static int msix_disable ;
2019-05-29 16:00:40 +00:00
module_param ( msix_disable , int , 0444 ) ;
2010-12-21 18:17:40 +00:00
MODULE_PARM_DESC ( msix_disable , " Disable MSI-X interrupt handling. Default: 0 " ) ;
2012-10-02 02:26:59 +00:00
static unsigned int msix_vectors ;
2019-05-29 16:00:40 +00:00
module_param ( msix_vectors , int , 0444 ) ;
2012-10-02 02:26:59 +00:00
MODULE_PARM_DESC ( msix_vectors , " MSI-X max vector count. Default: Set by FW " ) ;
2014-03-10 09:51:56 +00:00
static int allow_vf_ioctls ;
2019-05-29 16:00:40 +00:00
module_param ( allow_vf_ioctls , int , 0444 ) ;
2014-03-10 09:51:56 +00:00
MODULE_PARM_DESC ( allow_vf_ioctls , " Allow ioctls in SR-IOV VF mode. Default: 0 " ) ;
2015-01-05 14:36:23 +00:00
static unsigned int throttlequeuedepth = MEGASAS_THROTTLE_QUEUE_DEPTH ;
2019-05-29 16:00:40 +00:00
module_param ( throttlequeuedepth , int , 0444 ) ;
2012-07-18 01:20:03 +00:00
MODULE_PARM_DESC ( throttlequeuedepth ,
" Adapter queue depth when throttled due to I/O timeout. Default: 16 " ) ;
2016-01-28 15:34:34 +00:00
unsigned int resetwaittime = MEGASAS_RESET_WAIT_TIME ;
2019-05-29 16:00:40 +00:00
module_param ( resetwaittime , int , 0444 ) ;
2018-10-17 06:37:45 +00:00
MODULE_PARM_DESC ( resetwaittime , " Wait time in (1-180s) after I/O timeout before resetting adapter. Default: 180s " ) ;
2012-07-18 01:20:24 +00:00
2014-09-12 13:27:48 +00:00
int smp_affinity_enable = 1 ;
2019-05-29 16:00:40 +00:00
module_param ( smp_affinity_enable , int , 0444 ) ;
2018-04-29 12:25:32 +00:00
MODULE_PARM_DESC ( smp_affinity_enable , " SMP affinity feature enable/disable Default: enable(1) " ) ;
2014-09-12 13:27:48 +00:00
2016-01-28 15:34:28 +00:00
int rdpq_enable = 1 ;
2019-05-29 16:00:40 +00:00
module_param ( rdpq_enable , int , 0444 ) ;
2018-10-17 06:37:45 +00:00
MODULE_PARM_DESC ( rdpq_enable , " Allocate reply queue in chunks for large queue depth enable/disable Default: enable(1) " ) ;
2016-01-28 15:34:28 +00:00
2016-01-28 15:34:30 +00:00
unsigned int dual_qdepth_disable ;
2019-05-29 16:00:40 +00:00
module_param ( dual_qdepth_disable , int , 0444 ) ;
2016-01-28 15:34:30 +00:00
MODULE_PARM_DESC ( dual_qdepth_disable , " Disable dual queue depth feature. Default: 0 " ) ;
2016-01-28 15:34:34 +00:00
unsigned int scmd_timeout = MEGASAS_DEFAULT_CMD_TIMEOUT ;
2019-05-29 16:00:40 +00:00
module_param ( scmd_timeout , int , 0444 ) ;
2016-01-28 15:34:34 +00:00
MODULE_PARM_DESC ( scmd_timeout , " scsi command timeout (10-90s), default 90s. See megasas_reset_timer. " ) ;
2019-06-25 11:04:35 +00:00
int perf_mode = - 1 ;
module_param ( perf_mode , int , 0444 ) ;
MODULE_PARM_DESC ( perf_mode , " Performance mode (only for Aero adapters), options: \n \t \t "
" 0 - balanced: High iops and low latency queues are allocated & \n \t \t "
" interrupt coalescing is enabled only on high iops queues \n \t \t "
" 1 - iops: High iops queues are not allocated & \n \t \t "
" interrupt coalescing is enabled on all queues \n \t \t "
" 2 - latency: High iops queues are not allocated & \n \t \t "
" interrupt coalescing is disabled on all queues \n \t \t "
" default mode is 'balanced' "
) ;
2019-06-28 09:50:40 +00:00
int event_log_level = MFI_EVT_CLASS_CRITICAL ;
module_param ( event_log_level , int , 0644 ) ;
MODULE_PARM_DESC ( event_log_level , " Asynchronous event logging level- range is: -2(CLASS_DEBUG) to 4(CLASS_DEAD), Default: 2(CLASS_CRITICAL) " ) ;
2019-07-26 07:32:14 +00:00
unsigned int enable_sdev_max_qd ;
module_param ( enable_sdev_max_qd , int , 0444 ) ;
MODULE_PARM_DESC ( enable_sdev_max_qd , " Enable sdev max qd as can_queue. Default: 0 " ) ;
2005-09-20 21:46:58 +00:00
MODULE_LICENSE ( " GPL " ) ;
MODULE_VERSION ( MEGASAS_VERSION ) ;
2018-10-17 06:37:43 +00:00
MODULE_AUTHOR ( " megaraidlinux.pdl@broadcom.com " ) ;
MODULE_DESCRIPTION ( " Broadcom MegaRAID SAS Driver " ) ;
2005-09-20 21:46:58 +00:00
2011-10-09 01:14:27 +00:00
int megasas_transition_to_ready ( struct megasas_instance * instance , int ocr ) ;
2010-09-23 02:36:29 +00:00
static int megasas_get_pd_list ( struct megasas_instance * instance ) ;
2013-09-06 22:27:14 +00:00
static int megasas_ld_list_query ( struct megasas_instance * instance ,
u8 query_type ) ;
2010-09-23 02:36:29 +00:00
static int megasas_issue_init_mfi ( struct megasas_instance * instance ) ;
static int megasas_register_aen ( struct megasas_instance * instance ,
u32 seq_num , u32 class_locale_word ) ;
2017-02-10 08:59:10 +00:00
static void megasas_get_pd_info ( struct megasas_instance * instance ,
struct scsi_device * sdev ) ;
2018-06-04 10:45:12 +00:00
2005-09-20 21:46:58 +00:00
/*
* PCI ID table for all supported controllers
*/
static struct pci_device_id megasas_pci_table [ ] = {
2006-08-15 09:17:21 +00:00
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_SAS1064R ) } ,
/* xscale IOP */
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_SAS1078R ) } ,
/* ppc IOP */
2008-03-17 08:13:07 +00:00
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_SAS1078DE ) } ,
/* ppc IOP */
2008-08-10 19:42:38 +00:00
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_SAS1078GEN2 ) } ,
/* gen2*/
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_SAS0079GEN2 ) } ,
/* gen2*/
2009-10-06 20:31:54 +00:00
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_SAS0073SKINNY ) } ,
/* skinny*/
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_SAS0071SKINNY ) } ,
/* skinny*/
2006-08-15 09:17:21 +00:00
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_VERDE_ZCR ) } ,
/* xscale IOP, vega */
{ PCI_DEVICE ( PCI_VENDOR_ID_DELL , PCI_DEVICE_ID_DELL_PERC5 ) } ,
/* xscale IOP */
2010-12-21 21:34:31 +00:00
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_FUSION ) } ,
/* Fusion */
2014-03-10 09:51:56 +00:00
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_PLASMA ) } ,
/* Plasma */
2011-10-09 01:15:06 +00:00
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_INVADER ) } ,
/* Invader */
2013-05-22 07:01:43 +00:00
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_FURY ) } ,
/* Fury */
2015-10-15 08:09:44 +00:00
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_INTRUDER ) } ,
/* Intruder */
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_INTRUDER_24 ) } ,
/* Intruder 24 port*/
2015-10-15 08:09:54 +00:00
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_CUTLASS_52 ) } ,
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_CUTLASS_53 ) } ,
2017-01-10 23:20:43 +00:00
/* VENTURA */
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_VENTURA ) } ,
2017-10-19 09:48:49 +00:00
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_CRUSADER ) } ,
2017-01-10 23:20:43 +00:00
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_HARPOON ) } ,
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_TOMCAT ) } ,
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_VENTURA_4PORT ) } ,
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_CRUSADER_4PORT ) } ,
2018-11-09 17:47:20 +00:00
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_AERO_10E1 ) } ,
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_AERO_10E2 ) } ,
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_AERO_10E5 ) } ,
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_AERO_10E6 ) } ,
2019-06-25 11:04:20 +00:00
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_AERO_10E0 ) } ,
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_AERO_10E3 ) } ,
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_AERO_10E4 ) } ,
{ PCI_DEVICE ( PCI_VENDOR_ID_LSI_LOGIC , PCI_DEVICE_ID_LSI_AERO_10E7 ) } ,
2006-08-15 09:17:21 +00:00
{ }
2005-09-20 21:46:58 +00:00
} ;
MODULE_DEVICE_TABLE ( pci , megasas_pci_table ) ;
static int megasas_mgmt_majorno ;
2014-03-10 09:51:56 +00:00
struct megasas_mgmt_info megasas_mgmt_info ;
2005-09-20 21:46:58 +00:00
static struct fasync_struct * megasas_async_queue ;
2006-01-11 12:16:10 +00:00
static DEFINE_MUTEX ( megasas_async_queue_mutex ) ;
2005-09-20 21:46:58 +00:00
2009-10-06 20:18:02 +00:00
static int megasas_poll_wait_aen ;
static DECLARE_WAIT_QUEUE_HEAD ( megasas_poll_wait ) ;
2009-10-06 20:20:59 +00:00
static u32 support_poll_for_event ;
2010-12-21 21:34:31 +00:00
u32 megasas_dbg_lvl ;
2010-10-11 12:59:20 +00:00
static u32 support_device_change ;
2018-01-05 13:33:04 +00:00
static bool support_nvme_encapsulation ;
2019-06-25 11:04:30 +00:00
static bool support_pci_lane_margining ;
2006-10-03 20:09:14 +00:00
2009-10-06 20:18:02 +00:00
/* define lock for aen poll */
spinlock_t poll_aen_lock ;
2019-05-07 17:05:49 +00:00
extern struct dentry * megasas_debugfs_root ;
extern void megasas_init_debugfs ( void ) ;
extern void megasas_exit_debugfs ( void ) ;
extern void megasas_setup_debugfs ( struct megasas_instance * instance ) ;
extern void megasas_destroy_debugfs ( struct megasas_instance * instance ) ;
2010-12-21 21:34:31 +00:00
void
2007-11-09 09:35:44 +00:00
megasas_complete_cmd ( struct megasas_instance * instance , struct megasas_cmd * cmd ,
u8 alt_status ) ;
2011-02-25 04:57:15 +00:00
static u32
2018-12-17 08:47:39 +00:00
megasas_read_fw_status_reg_gen2 ( struct megasas_instance * instance ) ;
2011-02-25 04:57:15 +00:00
static int
megasas_adp_reset_gen2 ( struct megasas_instance * instance ,
struct megasas_register_set __iomem * reg_set ) ;
2010-12-21 18:23:23 +00:00
static irqreturn_t megasas_isr ( int irq , void * devp ) ;
static u32
megasas_init_adapter_mfi ( struct megasas_instance * instance ) ;
u32
megasas_build_and_issue_cmd ( struct megasas_instance * instance ,
struct scsi_cmnd * scmd ) ;
static void megasas_complete_cmd_dpc ( unsigned long instance_addr ) ;
2010-12-21 21:34:31 +00:00
int
2014-03-10 09:51:56 +00:00
wait_and_poll ( struct megasas_instance * instance , struct megasas_cmd * cmd ,
int seconds ) ;
2010-12-21 21:34:31 +00:00
void megasas_fusion_ocr_wq ( struct work_struct * work ) ;
2014-03-10 09:51:56 +00:00
static int megasas_get_ld_vf_affiliation ( struct megasas_instance * instance ,
int initial ) ;
2017-10-19 09:48:59 +00:00
static int
2017-10-19 09:49:05 +00:00
megasas_set_dma_mask ( struct megasas_instance * instance ) ;
2017-10-19 09:48:59 +00:00
static int
megasas_alloc_ctrl_mem ( struct megasas_instance * instance ) ;
static inline void
megasas_free_ctrl_mem ( struct megasas_instance * instance ) ;
static inline int
megasas_alloc_ctrl_dma_buffers ( struct megasas_instance * instance ) ;
static inline void
megasas_free_ctrl_dma_buffers ( struct megasas_instance * instance ) ;
static inline void
megasas_init_ctrl_params ( struct megasas_instance * instance ) ;
2010-12-21 18:23:23 +00:00
2018-12-17 08:47:40 +00:00
u32 megasas_readl ( struct megasas_instance * instance ,
const volatile void __iomem * addr )
{
u32 i = 0 , ret_val ;
/*
* Due to a HW errata in Aero controllers , reads to certain
* Fusion registers could intermittently return all zeroes .
* This behavior is transient in nature and subsequent reads will
* return valid value . As a workaround in driver , retry readl for
* upto three times until a non - zero value is read .
*/
if ( instance - > adapter_type = = AERO_SERIES ) {
do {
ret_val = readl ( addr ) ;
i + + ;
} while ( ret_val = = 0 & & i < 3 ) ;
return ret_val ;
} else {
return readl ( addr ) ;
}
}
2017-10-19 09:49:05 +00:00
/**
* megasas_set_dma_settings - Populate DMA address , length and flags for DCMDs
* @ instance : Adapter soft state
* @ dcmd : DCMD frame inside MFI command
* @ dma_addr : DMA address of buffer to be passed to FW
* @ dma_len : Length of DMA buffer to be passed to FW
* @ return : void
*/
void megasas_set_dma_settings ( struct megasas_instance * instance ,
struct megasas_dcmd_frame * dcmd ,
dma_addr_t dma_addr , u32 dma_len )
{
if ( instance - > consistent_mask_64bit ) {
dcmd - > sgl . sge64 [ 0 ] . phys_addr = cpu_to_le64 ( dma_addr ) ;
dcmd - > sgl . sge64 [ 0 ] . length = cpu_to_le32 ( dma_len ) ;
dcmd - > flags = cpu_to_le16 ( dcmd - > flags | MFI_FRAME_SGL64 ) ;
} else {
dcmd - > sgl . sge32 [ 0 ] . phys_addr =
cpu_to_le32 ( lower_32_bits ( dma_addr ) ) ;
dcmd - > sgl . sge32 [ 0 ] . length = cpu_to_le32 ( dma_len ) ;
dcmd - > flags = cpu_to_le16 ( dcmd - > flags ) ;
}
}
2010-12-21 18:23:23 +00:00
2019-07-02 13:01:14 +00:00
static void
2010-12-21 18:23:23 +00:00
megasas_issue_dcmd ( struct megasas_instance * instance , struct megasas_cmd * cmd )
{
instance - > instancet - > fire_cmd ( instance ,
cmd - > frame_phys_addr , 0 , instance - > reg_set ) ;
2017-02-10 08:59:09 +00:00
return ;
2010-12-21 18:23:23 +00:00
}
2007-11-09 09:35:44 +00:00
2005-09-20 21:46:58 +00:00
/**
* megasas_get_cmd - Get a command from the free pool
* @ instance : Adapter soft state
*
* Returns a free command from the pool
*/
2010-12-21 21:34:31 +00:00
struct megasas_cmd * megasas_get_cmd ( struct megasas_instance
2005-09-20 21:46:58 +00:00
* instance )
{
unsigned long flags ;
struct megasas_cmd * cmd = NULL ;
2014-09-12 13:27:58 +00:00
spin_lock_irqsave ( & instance - > mfi_pool_lock , flags ) ;
2005-09-20 21:46:58 +00:00
if ( ! list_empty ( & instance - > cmd_pool ) ) {
cmd = list_entry ( ( & instance - > cmd_pool ) - > next ,
struct megasas_cmd , list ) ;
list_del_init ( & cmd - > list ) ;
} else {
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " Command pool empty! \n " ) ;
2005-09-20 21:46:58 +00:00
}
2014-09-12 13:27:58 +00:00
spin_unlock_irqrestore ( & instance - > mfi_pool_lock , flags ) ;
2005-09-20 21:46:58 +00:00
return cmd ;
}
/**
2015-04-23 11:01:24 +00:00
* megasas_return_cmd - Return a cmd to free command pool
2005-09-20 21:46:58 +00:00
* @ instance : Adapter soft state
* @ cmd : Command packet to be returned to free command pool
*/
2017-05-08 22:58:53 +00:00
void
2015-04-23 11:01:24 +00:00
megasas_return_cmd ( struct megasas_instance * instance , struct megasas_cmd * cmd )
2005-09-20 21:46:58 +00:00
{
2015-04-23 11:01:24 +00:00
unsigned long flags ;
u32 blk_tags ;
struct megasas_cmd_fusion * cmd_fusion ;
struct fusion_context * fusion = instance - > ctrl_context ;
/* This flag is used only for fusion adapter.
* Wait for Interrupt for Polled mode DCMD
2014-09-12 13:27:58 +00:00
*/
2015-04-23 11:01:24 +00:00
if ( cmd - > flags & DRV_DCMD_POLLED_MODE )
2014-09-12 13:27:58 +00:00
return ;
2005-09-20 21:46:58 +00:00
2015-04-23 11:01:24 +00:00
spin_lock_irqsave ( & instance - > mfi_pool_lock , flags ) ;
if ( fusion ) {
blk_tags = instance - > max_scsi_cmds + cmd - > index ;
cmd_fusion = fusion - > cmd_list [ blk_tags ] ;
megasas_return_cmd_fusion ( instance , cmd_fusion ) ;
}
2005-09-20 21:46:58 +00:00
cmd - > scmd = NULL ;
2010-12-21 21:34:31 +00:00
cmd - > frame_count = 0 ;
2015-04-23 11:01:24 +00:00
cmd - > flags = 0 ;
2017-02-10 08:59:28 +00:00
memset ( cmd - > frame , 0 , instance - > mfi_frame_size ) ;
cmd - > frame - > io . context = cpu_to_le32 ( cmd - > index ) ;
2015-04-23 11:01:24 +00:00
if ( ! fusion & & reset_devices )
2011-10-09 01:15:19 +00:00
cmd - > frame - > hdr . cmd = MFI_CMD_INVALID ;
2014-09-12 13:27:58 +00:00
list_add ( & cmd - > list , ( & instance - > cmd_pool ) - > next ) ;
spin_unlock_irqrestore ( & instance - > mfi_pool_lock , flags ) ;
2005-09-20 21:46:58 +00:00
2015-04-23 11:01:24 +00:00
}
2006-01-25 20:02:40 +00:00
2015-08-31 11:53:51 +00:00
static const char *
format_timestamp ( uint32_t timestamp )
{
static char buffer [ 32 ] ;
if ( ( timestamp & 0xff000000 ) = = 0xff000000 )
snprintf ( buffer , sizeof ( buffer ) , " boot + %us " , timestamp &
0x00ffffff ) ;
else
snprintf ( buffer , sizeof ( buffer ) , " %us " , timestamp ) ;
return buffer ;
}
static const char *
format_class ( int8_t class )
{
static char buffer [ 6 ] ;
switch ( class ) {
case MFI_EVT_CLASS_DEBUG :
return " debug " ;
case MFI_EVT_CLASS_PROGRESS :
return " progress " ;
case MFI_EVT_CLASS_INFO :
return " info " ;
case MFI_EVT_CLASS_WARNING :
return " WARN " ;
case MFI_EVT_CLASS_CRITICAL :
return " CRIT " ;
case MFI_EVT_CLASS_FATAL :
return " FATAL " ;
case MFI_EVT_CLASS_DEAD :
return " DEAD " ;
default :
snprintf ( buffer , sizeof ( buffer ) , " %d " , class ) ;
return buffer ;
}
}
/**
* megasas_decode_evt : Decode FW AEN event and print critical event
* for information .
* @ instance : Adapter soft state
*/
static void
megasas_decode_evt ( struct megasas_instance * instance )
{
struct megasas_evt_detail * evt_detail = instance - > evt_detail ;
union megasas_evt_class_locale class_locale ;
class_locale . word = le32_to_cpu ( evt_detail - > cl . word ) ;
2019-06-28 09:50:40 +00:00
if ( ( event_log_level < MFI_EVT_CLASS_DEBUG ) | |
( event_log_level > MFI_EVT_CLASS_DEAD ) ) {
printk ( KERN_WARNING " megaraid_sas: provided event log level is out of range, setting it to default 2(CLASS_CRITICAL), permissible range is: -2 to 4 \n " ) ;
event_log_level = MFI_EVT_CLASS_CRITICAL ;
}
if ( class_locale . members . class > = event_log_level )
2015-08-31 11:53:51 +00:00
dev_info ( & instance - > pdev - > dev , " %d (%s/0x%04x/%s) - %s \n " ,
le32_to_cpu ( evt_detail - > seq_num ) ,
format_timestamp ( le32_to_cpu ( evt_detail - > time_stamp ) ) ,
( class_locale . members . locale ) ,
format_class ( class_locale . members . class ) ,
evt_detail - > description ) ;
}
2006-01-25 20:02:40 +00:00
/**
2010-12-15 03:17:17 +00:00
* The following functions are defined for xscale
2006-01-25 20:02:40 +00:00
* ( deviceid : 1064 R , PERC5 ) controllers
*/
2005-09-20 21:46:58 +00:00
/**
2006-01-25 20:02:40 +00:00
* megasas_enable_intr_xscale - Enables interrupts
2005-09-20 21:46:58 +00:00
* @ regs : MFI register set
*/
static inline void
2013-05-22 07:04:14 +00:00
megasas_enable_intr_xscale ( struct megasas_instance * instance )
2005-09-20 21:46:58 +00:00
{
2013-05-22 07:04:14 +00:00
struct megasas_register_set __iomem * regs ;
2015-07-07 20:52:45 +00:00
2013-05-22 07:04:14 +00:00
regs = instance - > reg_set ;
2010-09-23 02:36:29 +00:00
writel ( 0 , & ( regs ) - > outbound_intr_mask ) ;
2005-09-20 21:46:58 +00:00
/* Dummy readl to force pci flush */
readl ( & regs - > outbound_intr_mask ) ;
}
2006-10-03 19:52:12 +00:00
/**
* megasas_disable_intr_xscale - Disables interrupt
* @ regs : MFI register set
*/
static inline void
2013-05-22 07:04:14 +00:00
megasas_disable_intr_xscale ( struct megasas_instance * instance )
2006-10-03 19:52:12 +00:00
{
2013-05-22 07:04:14 +00:00
struct megasas_register_set __iomem * regs ;
2006-10-03 19:52:12 +00:00
u32 mask = 0x1f ;
2015-07-07 20:52:45 +00:00
2013-05-22 07:04:14 +00:00
regs = instance - > reg_set ;
2006-10-03 19:52:12 +00:00
writel ( mask , & regs - > outbound_intr_mask ) ;
/* Dummy readl to force pci flush */
readl ( & regs - > outbound_intr_mask ) ;
}
2006-01-25 20:02:40 +00:00
/**
* megasas_read_fw_status_reg_xscale - returns the current FW status value
* @ regs : MFI register set
*/
static u32
2018-12-17 08:47:39 +00:00
megasas_read_fw_status_reg_xscale ( struct megasas_instance * instance )
2006-01-25 20:02:40 +00:00
{
2018-12-17 08:47:39 +00:00
return readl ( & instance - > reg_set - > outbound_msg_0 ) ;
2006-01-25 20:02:40 +00:00
}
/**
* megasas_clear_interrupt_xscale - Check & clear interrupt
* @ regs : MFI register set
*/
2010-12-15 03:17:17 +00:00
static int
2018-12-17 08:47:39 +00:00
megasas_clear_intr_xscale ( struct megasas_instance * instance )
2006-01-25 20:02:40 +00:00
{
u32 status ;
2010-09-23 02:36:29 +00:00
u32 mfiStatus = 0 ;
2018-12-17 08:47:39 +00:00
struct megasas_register_set __iomem * regs ;
regs = instance - > reg_set ;
2015-07-07 20:52:45 +00:00
2006-01-25 20:02:40 +00:00
/*
* Check if it is our interrupt
*/
status = readl ( & regs - > outbound_intr_status ) ;
2010-09-23 02:36:29 +00:00
if ( status & MFI_OB_INTR_STATUS_MASK )
mfiStatus = MFI_INTR_FLAG_REPLY_MESSAGE ;
if ( status & MFI_XSCALE_OMR0_CHANGE_INTERRUPT )
mfiStatus | = MFI_INTR_FLAG_FIRMWARE_STATE_CHANGE ;
2006-01-25 20:02:40 +00:00
/*
* Clear the interrupt by writing back the same value
*/
2010-09-23 02:36:29 +00:00
if ( mfiStatus )
writel ( status , & regs - > outbound_intr_status ) ;
2006-01-25 20:02:40 +00:00
2008-08-10 19:42:37 +00:00
/* Dummy readl to force pci flush */
readl ( & regs - > outbound_intr_status ) ;
2010-09-23 02:36:29 +00:00
return mfiStatus ;
2006-01-25 20:02:40 +00:00
}
/**
* megasas_fire_cmd_xscale - Sends command to the FW
* @ frame_phys_addr : Physical address of cmd
* @ frame_count : Number of frames for the command
* @ regs : MFI register set
*/
2010-12-15 03:17:17 +00:00
static inline void
2009-10-06 20:47:35 +00:00
megasas_fire_cmd_xscale ( struct megasas_instance * instance ,
dma_addr_t frame_phys_addr ,
u32 frame_count ,
struct megasas_register_set __iomem * regs )
2006-01-25 20:02:40 +00:00
{
2010-09-23 02:36:29 +00:00
unsigned long flags ;
2015-07-07 20:52:45 +00:00
2010-09-23 02:36:29 +00:00
spin_lock_irqsave ( & instance - > hba_lock , flags ) ;
2006-01-25 20:02:40 +00:00
writel ( ( frame_phys_addr > > 3 ) | ( frame_count ) ,
& ( regs ) - > inbound_queue_port ) ;
2010-09-23 02:36:29 +00:00
spin_unlock_irqrestore ( & instance - > hba_lock , flags ) ;
}
/**
* megasas_adp_reset_xscale - For controller reset
* @ regs : MFI register set
*/
static int
megasas_adp_reset_xscale ( struct megasas_instance * instance ,
struct megasas_register_set __iomem * regs )
{
u32 i ;
u32 pcidata ;
2015-07-07 20:52:45 +00:00
2010-09-23 02:36:29 +00:00
writel ( MFI_ADP_RESET , & regs - > inbound_doorbell ) ;
for ( i = 0 ; i < 3 ; i + + )
msleep ( 1000 ) ; /* sleep for 3 secs */
pcidata = 0 ;
pci_read_config_dword ( instance - > pdev , MFI_1068_PCSR_OFFSET , & pcidata ) ;
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " pcidata = %x \n " , pcidata ) ;
2010-09-23 02:36:29 +00:00
if ( pcidata & 0x2 ) {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " mfi 1068 offset read=%x \n " , pcidata ) ;
2010-09-23 02:36:29 +00:00
pcidata & = ~ 0x2 ;
pci_write_config_dword ( instance - > pdev ,
MFI_1068_PCSR_OFFSET , pcidata ) ;
for ( i = 0 ; i < 2 ; i + + )
msleep ( 1000 ) ; /* need to wait 2 secs again */
pcidata = 0 ;
pci_read_config_dword ( instance - > pdev ,
MFI_1068_FW_HANDSHAKE_OFFSET , & pcidata ) ;
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " 1068 offset handshake read=%x \n " , pcidata ) ;
2010-09-23 02:36:29 +00:00
if ( ( pcidata & 0xffff0000 ) = = MFI_1068_FW_READY ) {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " 1068 offset pcidt=%x \n " , pcidata ) ;
2010-09-23 02:36:29 +00:00
pcidata = 0 ;
pci_write_config_dword ( instance - > pdev ,
MFI_1068_FW_HANDSHAKE_OFFSET , pcidata ) ;
}
}
return 0 ;
}
/**
* megasas_check_reset_xscale - For controller reset check
* @ regs : MFI register set
*/
static int
megasas_check_reset_xscale ( struct megasas_instance * instance ,
struct megasas_register_set __iomem * regs )
{
2016-01-28 15:34:32 +00:00
if ( ( atomic_read ( & instance - > adprecovery ) ! = MEGASAS_HBA_OPERATIONAL ) & &
2013-09-06 10:20:52 +00:00
( le32_to_cpu ( * instance - > consumer ) = =
MEGASAS_ADPRESET_INPROG_SIGN ) )
2010-09-23 02:36:29 +00:00
return 1 ;
return 0 ;
2006-01-25 20:02:40 +00:00
}
static struct megasas_instance_template megasas_instance_template_xscale = {
. fire_cmd = megasas_fire_cmd_xscale ,
. enable_intr = megasas_enable_intr_xscale ,
2006-10-03 19:52:12 +00:00
. disable_intr = megasas_disable_intr_xscale ,
2006-01-25 20:02:40 +00:00
. clear_intr = megasas_clear_intr_xscale ,
. read_fw_status_reg = megasas_read_fw_status_reg_xscale ,
2010-09-23 02:36:29 +00:00
. adp_reset = megasas_adp_reset_xscale ,
. check_reset = megasas_check_reset_xscale ,
2010-12-21 18:23:23 +00:00
. service_isr = megasas_isr ,
. tasklet = megasas_complete_cmd_dpc ,
. init_adapter = megasas_init_adapter_mfi ,
. build_and_issue_cmd = megasas_build_and_issue_cmd ,
. issue_dcmd = megasas_issue_dcmd ,
2006-01-25 20:02:40 +00:00
} ;
/**
2010-12-15 03:17:17 +00:00
* This is the end of set of functions & definitions specific
2006-01-25 20:02:40 +00:00
* to xscale ( deviceid : 1064 R , PERC5 ) controllers
*/
2006-02-03 23:34:35 +00:00
/**
2010-12-15 03:17:17 +00:00
* The following functions are defined for ppc ( deviceid : 0x60 )
2015-07-07 20:52:45 +00:00
* controllers
2006-02-03 23:34:35 +00:00
*/
/**
* megasas_enable_intr_ppc - Enables interrupts
* @ regs : MFI register set
*/
static inline void
2013-05-22 07:04:14 +00:00
megasas_enable_intr_ppc ( struct megasas_instance * instance )
2006-02-03 23:34:35 +00:00
{
2013-05-22 07:04:14 +00:00
struct megasas_register_set __iomem * regs ;
2015-07-07 20:52:45 +00:00
2013-05-22 07:04:14 +00:00
regs = instance - > reg_set ;
2006-02-03 23:34:35 +00:00
writel ( 0xFFFFFFFF , & ( regs ) - > outbound_doorbell_clear ) ;
2010-12-15 03:17:17 +00:00
2010-09-23 02:36:29 +00:00
writel ( ~ 0x80000000 , & ( regs ) - > outbound_intr_mask ) ;
2006-02-03 23:34:35 +00:00
/* Dummy readl to force pci flush */
readl ( & regs - > outbound_intr_mask ) ;
}
2006-10-03 19:52:12 +00:00
/**
* megasas_disable_intr_ppc - Disable interrupt
* @ regs : MFI register set
*/
static inline void
2013-05-22 07:04:14 +00:00
megasas_disable_intr_ppc ( struct megasas_instance * instance )
2006-10-03 19:52:12 +00:00
{
2013-05-22 07:04:14 +00:00
struct megasas_register_set __iomem * regs ;
2006-10-03 19:52:12 +00:00
u32 mask = 0xFFFFFFFF ;
2015-07-07 20:52:45 +00:00
2013-05-22 07:04:14 +00:00
regs = instance - > reg_set ;
2006-10-03 19:52:12 +00:00
writel ( mask , & regs - > outbound_intr_mask ) ;
/* Dummy readl to force pci flush */
readl ( & regs - > outbound_intr_mask ) ;
}
2006-02-03 23:34:35 +00:00
/**
* megasas_read_fw_status_reg_ppc - returns the current FW status value
* @ regs : MFI register set
*/
static u32
2018-12-17 08:47:39 +00:00
megasas_read_fw_status_reg_ppc ( struct megasas_instance * instance )
2006-02-03 23:34:35 +00:00
{
2018-12-17 08:47:39 +00:00
return readl ( & instance - > reg_set - > outbound_scratch_pad_0 ) ;
2006-02-03 23:34:35 +00:00
}
/**
* megasas_clear_interrupt_ppc - Check & clear interrupt
* @ regs : MFI register set
*/
2010-12-15 03:17:17 +00:00
static int
2018-12-17 08:47:39 +00:00
megasas_clear_intr_ppc ( struct megasas_instance * instance )
2006-02-03 23:34:35 +00:00
{
2011-05-12 01:34:52 +00:00
u32 status , mfiStatus = 0 ;
2018-12-17 08:47:39 +00:00
struct megasas_register_set __iomem * regs ;
regs = instance - > reg_set ;
2011-05-12 01:34:52 +00:00
2006-02-03 23:34:35 +00:00
/*
* Check if it is our interrupt
*/
status = readl ( & regs - > outbound_intr_status ) ;
2011-05-12 01:34:52 +00:00
if ( status & MFI_REPLY_1078_MESSAGE_INTERRUPT )
mfiStatus = MFI_INTR_FLAG_REPLY_MESSAGE ;
if ( status & MFI_G2_OUTBOUND_DOORBELL_CHANGE_INTERRUPT )
mfiStatus | = MFI_INTR_FLAG_FIRMWARE_STATE_CHANGE ;
2006-02-03 23:34:35 +00:00
/*
* Clear the interrupt by writing back the same value
*/
writel ( status , & regs - > outbound_doorbell_clear ) ;
2008-08-10 19:42:37 +00:00
/* Dummy readl to force pci flush */
readl ( & regs - > outbound_doorbell_clear ) ;
2011-05-12 01:34:52 +00:00
return mfiStatus ;
2006-02-03 23:34:35 +00:00
}
2011-05-12 01:34:52 +00:00
2006-02-03 23:34:35 +00:00
/**
* megasas_fire_cmd_ppc - Sends command to the FW
* @ frame_phys_addr : Physical address of cmd
* @ frame_count : Number of frames for the command
* @ regs : MFI register set
*/
2010-12-15 03:17:17 +00:00
static inline void
2009-10-06 20:47:35 +00:00
megasas_fire_cmd_ppc ( struct megasas_instance * instance ,
dma_addr_t frame_phys_addr ,
u32 frame_count ,
struct megasas_register_set __iomem * regs )
2006-02-03 23:34:35 +00:00
{
2010-09-23 02:36:29 +00:00
unsigned long flags ;
2015-07-07 20:52:45 +00:00
2010-09-23 02:36:29 +00:00
spin_lock_irqsave ( & instance - > hba_lock , flags ) ;
2010-12-15 03:17:17 +00:00
writel ( ( frame_phys_addr | ( frame_count < < 1 ) ) | 1 ,
2006-02-03 23:34:35 +00:00
& ( regs ) - > inbound_queue_port ) ;
2010-09-23 02:36:29 +00:00
spin_unlock_irqrestore ( & instance - > hba_lock , flags ) ;
2006-02-03 23:34:35 +00:00
}
2010-09-23 02:36:29 +00:00
/**
* megasas_check_reset_ppc - For controller reset check
* @ regs : MFI register set
*/
static int
megasas_check_reset_ppc ( struct megasas_instance * instance ,
struct megasas_register_set __iomem * regs )
{
2016-01-28 15:34:32 +00:00
if ( atomic_read ( & instance - > adprecovery ) ! = MEGASAS_HBA_OPERATIONAL )
2011-05-12 01:34:52 +00:00
return 1 ;
2010-09-23 02:36:29 +00:00
return 0 ;
}
2011-05-12 01:34:52 +00:00
2006-02-03 23:34:35 +00:00
static struct megasas_instance_template megasas_instance_template_ppc = {
2010-12-15 03:17:17 +00:00
2006-02-03 23:34:35 +00:00
. fire_cmd = megasas_fire_cmd_ppc ,
. enable_intr = megasas_enable_intr_ppc ,
2006-10-03 19:52:12 +00:00
. disable_intr = megasas_disable_intr_ppc ,
2006-02-03 23:34:35 +00:00
. clear_intr = megasas_clear_intr_ppc ,
. read_fw_status_reg = megasas_read_fw_status_reg_ppc ,
2011-05-12 01:34:52 +00:00
. adp_reset = megasas_adp_reset_xscale ,
2010-09-23 02:36:29 +00:00
. check_reset = megasas_check_reset_ppc ,
2010-12-21 18:23:23 +00:00
. service_isr = megasas_isr ,
. tasklet = megasas_complete_cmd_dpc ,
. init_adapter = megasas_init_adapter_mfi ,
. build_and_issue_cmd = megasas_build_and_issue_cmd ,
. issue_dcmd = megasas_issue_dcmd ,
2006-02-03 23:34:35 +00:00
} ;
2009-10-06 20:31:54 +00:00
/**
* megasas_enable_intr_skinny - Enables interrupts
* @ regs : MFI register set
*/
static inline void
2013-05-22 07:04:14 +00:00
megasas_enable_intr_skinny ( struct megasas_instance * instance )
2009-10-06 20:31:54 +00:00
{
2013-05-22 07:04:14 +00:00
struct megasas_register_set __iomem * regs ;
2015-07-07 20:52:45 +00:00
2013-05-22 07:04:14 +00:00
regs = instance - > reg_set ;
2009-10-06 20:31:54 +00:00
writel ( 0xFFFFFFFF , & ( regs ) - > outbound_intr_mask ) ;
writel ( ~ MFI_SKINNY_ENABLE_INTERRUPT_MASK , & ( regs ) - > outbound_intr_mask ) ;
/* Dummy readl to force pci flush */
readl ( & regs - > outbound_intr_mask ) ;
}
/**
* megasas_disable_intr_skinny - Disables interrupt
* @ regs : MFI register set
*/
static inline void
2013-05-22 07:04:14 +00:00
megasas_disable_intr_skinny ( struct megasas_instance * instance )
2009-10-06 20:31:54 +00:00
{
2013-05-22 07:04:14 +00:00
struct megasas_register_set __iomem * regs ;
2009-10-06 20:31:54 +00:00
u32 mask = 0xFFFFFFFF ;
2015-07-07 20:52:45 +00:00
2013-05-22 07:04:14 +00:00
regs = instance - > reg_set ;
2009-10-06 20:31:54 +00:00
writel ( mask , & regs - > outbound_intr_mask ) ;
/* Dummy readl to force pci flush */
readl ( & regs - > outbound_intr_mask ) ;
}
/**
* megasas_read_fw_status_reg_skinny - returns the current FW status value
* @ regs : MFI register set
*/
static u32
2018-12-17 08:47:39 +00:00
megasas_read_fw_status_reg_skinny ( struct megasas_instance * instance )
2009-10-06 20:31:54 +00:00
{
2018-12-17 08:47:39 +00:00
return readl ( & instance - > reg_set - > outbound_scratch_pad_0 ) ;
2009-10-06 20:31:54 +00:00
}
/**
* megasas_clear_interrupt_skinny - Check & clear interrupt
* @ regs : MFI register set
*/
static int
2018-12-17 08:47:39 +00:00
megasas_clear_intr_skinny ( struct megasas_instance * instance )
2009-10-06 20:31:54 +00:00
{
u32 status ;
2011-02-25 04:57:15 +00:00
u32 mfiStatus = 0 ;
2018-12-17 08:47:39 +00:00
struct megasas_register_set __iomem * regs ;
regs = instance - > reg_set ;
2011-02-25 04:57:15 +00:00
2009-10-06 20:31:54 +00:00
/*
* Check if it is our interrupt
*/
status = readl ( & regs - > outbound_intr_status ) ;
if ( ! ( status & MFI_SKINNY_ENABLE_INTERRUPT_MASK ) ) {
2010-09-23 02:36:29 +00:00
return 0 ;
2009-10-06 20:31:54 +00:00
}
2011-02-25 04:57:15 +00:00
/*
* Check if it is our interrupt
*/
2018-12-17 08:47:39 +00:00
if ( ( megasas_read_fw_status_reg_skinny ( instance ) & MFI_STATE_MASK ) = =
2011-02-25 04:57:15 +00:00
MFI_STATE_FAULT ) {
mfiStatus = MFI_INTR_FLAG_FIRMWARE_STATE_CHANGE ;
} else
mfiStatus = MFI_INTR_FLAG_REPLY_MESSAGE ;
2009-10-06 20:31:54 +00:00
/*
* Clear the interrupt by writing back the same value
*/
writel ( status , & regs - > outbound_intr_status ) ;
/*
2015-07-07 20:52:45 +00:00
* dummy read to flush PCI
*/
2009-10-06 20:31:54 +00:00
readl ( & regs - > outbound_intr_status ) ;
2011-02-25 04:57:15 +00:00
return mfiStatus ;
2009-10-06 20:31:54 +00:00
}
/**
* megasas_fire_cmd_skinny - Sends command to the FW
* @ frame_phys_addr : Physical address of cmd
* @ frame_count : Number of frames for the command
* @ regs : MFI register set
*/
static inline void
2009-10-06 20:47:35 +00:00
megasas_fire_cmd_skinny ( struct megasas_instance * instance ,
dma_addr_t frame_phys_addr ,
u32 frame_count ,
2009-10-06 20:31:54 +00:00
struct megasas_register_set __iomem * regs )
{
2009-10-06 20:47:35 +00:00
unsigned long flags ;
2015-07-07 20:52:45 +00:00
2010-09-23 02:36:29 +00:00
spin_lock_irqsave ( & instance - > hba_lock , flags ) ;
2013-09-06 10:20:52 +00:00
writel ( upper_32_bits ( frame_phys_addr ) ,
& ( regs ) - > inbound_high_queue_port ) ;
writel ( ( lower_32_bits ( frame_phys_addr ) | ( frame_count < < 1 ) ) | 1 ,
& ( regs ) - > inbound_low_queue_port ) ;
2010-09-23 02:36:29 +00:00
spin_unlock_irqrestore ( & instance - > hba_lock , flags ) ;
}
/**
* megasas_check_reset_skinny - For controller reset check
* @ regs : MFI register set
*/
static int
megasas_check_reset_skinny ( struct megasas_instance * instance ,
struct megasas_register_set __iomem * regs )
{
2016-01-28 15:34:32 +00:00
if ( atomic_read ( & instance - > adprecovery ) ! = MEGASAS_HBA_OPERATIONAL )
2011-05-12 01:34:52 +00:00
return 1 ;
2010-09-23 02:36:29 +00:00
return 0 ;
2009-10-06 20:31:54 +00:00
}
static struct megasas_instance_template megasas_instance_template_skinny = {
. fire_cmd = megasas_fire_cmd_skinny ,
. enable_intr = megasas_enable_intr_skinny ,
. disable_intr = megasas_disable_intr_skinny ,
. clear_intr = megasas_clear_intr_skinny ,
. read_fw_status_reg = megasas_read_fw_status_reg_skinny ,
2011-02-25 04:57:15 +00:00
. adp_reset = megasas_adp_reset_gen2 ,
2010-09-23 02:36:29 +00:00
. check_reset = megasas_check_reset_skinny ,
2010-12-21 18:23:23 +00:00
. service_isr = megasas_isr ,
. tasklet = megasas_complete_cmd_dpc ,
. init_adapter = megasas_init_adapter_mfi ,
. build_and_issue_cmd = megasas_build_and_issue_cmd ,
. issue_dcmd = megasas_issue_dcmd ,
2009-10-06 20:31:54 +00:00
} ;
2008-08-10 19:42:38 +00:00
/**
* The following functions are defined for gen2 ( deviceid : 0x78 0x79 )
* controllers
*/
/**
* megasas_enable_intr_gen2 - Enables interrupts
* @ regs : MFI register set
*/
static inline void
2013-05-22 07:04:14 +00:00
megasas_enable_intr_gen2 ( struct megasas_instance * instance )
2008-08-10 19:42:38 +00:00
{
2013-05-22 07:04:14 +00:00
struct megasas_register_set __iomem * regs ;
2015-07-07 20:52:45 +00:00
2013-05-22 07:04:14 +00:00
regs = instance - > reg_set ;
2008-08-10 19:42:38 +00:00
writel ( 0xFFFFFFFF , & ( regs ) - > outbound_doorbell_clear ) ;
/* write ~0x00000005 (4 & 1) to the intr mask*/
writel ( ~ MFI_GEN2_ENABLE_INTERRUPT_MASK , & ( regs ) - > outbound_intr_mask ) ;
/* Dummy readl to force pci flush */
readl ( & regs - > outbound_intr_mask ) ;
}
/**
* megasas_disable_intr_gen2 - Disables interrupt
* @ regs : MFI register set
*/
static inline void
2013-05-22 07:04:14 +00:00
megasas_disable_intr_gen2 ( struct megasas_instance * instance )
2008-08-10 19:42:38 +00:00
{
2013-05-22 07:04:14 +00:00
struct megasas_register_set __iomem * regs ;
2008-08-10 19:42:38 +00:00
u32 mask = 0xFFFFFFFF ;
2015-07-07 20:52:45 +00:00
2013-05-22 07:04:14 +00:00
regs = instance - > reg_set ;
2008-08-10 19:42:38 +00:00
writel ( mask , & regs - > outbound_intr_mask ) ;
/* Dummy readl to force pci flush */
readl ( & regs - > outbound_intr_mask ) ;
}
/**
* megasas_read_fw_status_reg_gen2 - returns the current FW status value
* @ regs : MFI register set
*/
static u32
2018-12-17 08:47:39 +00:00
megasas_read_fw_status_reg_gen2 ( struct megasas_instance * instance )
2008-08-10 19:42:38 +00:00
{
2018-12-17 08:47:39 +00:00
return readl ( & instance - > reg_set - > outbound_scratch_pad_0 ) ;
2008-08-10 19:42:38 +00:00
}
/**
* megasas_clear_interrupt_gen2 - Check & clear interrupt
* @ regs : MFI register set
*/
static int
2018-12-17 08:47:39 +00:00
megasas_clear_intr_gen2 ( struct megasas_instance * instance )
2008-08-10 19:42:38 +00:00
{
u32 status ;
2010-09-23 02:36:29 +00:00
u32 mfiStatus = 0 ;
2018-12-17 08:47:39 +00:00
struct megasas_register_set __iomem * regs ;
regs = instance - > reg_set ;
2015-07-07 20:52:45 +00:00
2008-08-10 19:42:38 +00:00
/*
* Check if it is our interrupt
*/
status = readl ( & regs - > outbound_intr_status ) ;
2013-05-22 06:59:54 +00:00
if ( status & MFI_INTR_FLAG_REPLY_MESSAGE ) {
2010-09-23 02:36:29 +00:00
mfiStatus = MFI_INTR_FLAG_REPLY_MESSAGE ;
}
if ( status & MFI_G2_OUTBOUND_DOORBELL_CHANGE_INTERRUPT ) {
mfiStatus | = MFI_INTR_FLAG_FIRMWARE_STATE_CHANGE ;
}
2008-08-10 19:42:38 +00:00
/*
* Clear the interrupt by writing back the same value
*/
2010-09-23 02:36:29 +00:00
if ( mfiStatus )
writel ( status , & regs - > outbound_doorbell_clear ) ;
2008-08-10 19:42:38 +00:00
/* Dummy readl to force pci flush */
readl ( & regs - > outbound_intr_status ) ;
2010-09-23 02:36:29 +00:00
return mfiStatus ;
2008-08-10 19:42:38 +00:00
}
/**
* megasas_fire_cmd_gen2 - Sends command to the FW
* @ frame_phys_addr : Physical address of cmd
* @ frame_count : Number of frames for the command
* @ regs : MFI register set
*/
static inline void
2009-10-06 20:47:35 +00:00
megasas_fire_cmd_gen2 ( struct megasas_instance * instance ,
dma_addr_t frame_phys_addr ,
u32 frame_count ,
2008-08-10 19:42:38 +00:00
struct megasas_register_set __iomem * regs )
{
2010-09-23 02:36:29 +00:00
unsigned long flags ;
2015-07-07 20:52:45 +00:00
2010-09-23 02:36:29 +00:00
spin_lock_irqsave ( & instance - > hba_lock , flags ) ;
2008-08-10 19:42:38 +00:00
writel ( ( frame_phys_addr | ( frame_count < < 1 ) ) | 1 ,
& ( regs ) - > inbound_queue_port ) ;
2010-09-23 02:36:29 +00:00
spin_unlock_irqrestore ( & instance - > hba_lock , flags ) ;
}
/**
* megasas_adp_reset_gen2 - For controller reset
* @ regs : MFI register set
*/
static int
megasas_adp_reset_gen2 ( struct megasas_instance * instance ,
struct megasas_register_set __iomem * reg_set )
{
2015-07-07 20:52:45 +00:00
u32 retry = 0 ;
u32 HostDiag ;
u32 __iomem * seq_offset = & reg_set - > seq_offset ;
u32 __iomem * hostdiag_offset = & reg_set - > host_diag ;
2011-02-25 04:57:15 +00:00
if ( instance - > instancet = = & megasas_instance_template_skinny ) {
seq_offset = & reg_set - > fusion_seq_offset ;
hostdiag_offset = & reg_set - > fusion_host_diag ;
}
writel ( 0 , seq_offset ) ;
writel ( 4 , seq_offset ) ;
writel ( 0xb , seq_offset ) ;
writel ( 2 , seq_offset ) ;
writel ( 7 , seq_offset ) ;
writel ( 0xd , seq_offset ) ;
2010-09-23 02:36:29 +00:00
msleep ( 1000 ) ;
2011-02-25 04:57:15 +00:00
HostDiag = ( u32 ) readl ( hostdiag_offset ) ;
2010-09-23 02:36:29 +00:00
2015-07-07 20:52:45 +00:00
while ( ! ( HostDiag & DIAG_WRITE_ENABLE ) ) {
2010-09-23 02:36:29 +00:00
msleep ( 100 ) ;
2011-02-25 04:57:15 +00:00
HostDiag = ( u32 ) readl ( hostdiag_offset ) ;
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " RESETGEN2: retry=%x, hostdiag=%x \n " ,
2010-09-23 02:36:29 +00:00
retry , HostDiag ) ;
if ( retry + + > = 100 )
return 1 ;
}
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " ADP_RESET_GEN2: HostDiag=%x \n " , HostDiag ) ;
2010-09-23 02:36:29 +00:00
2011-02-25 04:57:15 +00:00
writel ( ( HostDiag | DIAG_RESET_ADAPTER ) , hostdiag_offset ) ;
2010-09-23 02:36:29 +00:00
ssleep ( 10 ) ;
2011-02-25 04:57:15 +00:00
HostDiag = ( u32 ) readl ( hostdiag_offset ) ;
2015-07-07 20:52:45 +00:00
while ( HostDiag & DIAG_RESET_ADAPTER ) {
2010-09-23 02:36:29 +00:00
msleep ( 100 ) ;
2011-02-25 04:57:15 +00:00
HostDiag = ( u32 ) readl ( hostdiag_offset ) ;
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " RESET_GEN2: retry=%x, hostdiag=%x \n " ,
2010-09-23 02:36:29 +00:00
retry , HostDiag ) ;
if ( retry + + > = 1000 )
return 1 ;
}
return 0 ;
}
/**
* megasas_check_reset_gen2 - For controller reset check
* @ regs : MFI register set
*/
static int
megasas_check_reset_gen2 ( struct megasas_instance * instance ,
struct megasas_register_set __iomem * regs )
{
2016-01-28 15:34:32 +00:00
if ( atomic_read ( & instance - > adprecovery ) ! = MEGASAS_HBA_OPERATIONAL )
2010-10-12 13:20:27 +00:00
return 1 ;
2010-09-23 02:36:29 +00:00
return 0 ;
2008-08-10 19:42:38 +00:00
}
static struct megasas_instance_template megasas_instance_template_gen2 = {
. fire_cmd = megasas_fire_cmd_gen2 ,
. enable_intr = megasas_enable_intr_gen2 ,
. disable_intr = megasas_disable_intr_gen2 ,
. clear_intr = megasas_clear_intr_gen2 ,
. read_fw_status_reg = megasas_read_fw_status_reg_gen2 ,
2010-09-23 02:36:29 +00:00
. adp_reset = megasas_adp_reset_gen2 ,
. check_reset = megasas_check_reset_gen2 ,
2010-12-21 18:23:23 +00:00
. service_isr = megasas_isr ,
. tasklet = megasas_complete_cmd_dpc ,
. init_adapter = megasas_init_adapter_mfi ,
. build_and_issue_cmd = megasas_build_and_issue_cmd ,
. issue_dcmd = megasas_issue_dcmd ,
2008-08-10 19:42:38 +00:00
} ;
2006-02-03 23:34:35 +00:00
/**
* This is the end of set of functions & definitions
2010-09-23 02:36:29 +00:00
* specific to gen2 ( deviceid : 0x78 , 0x79 ) controllers
2006-02-03 23:34:35 +00:00
*/
2010-12-21 21:34:31 +00:00
/*
* Template added for TB ( Fusion )
*/
extern struct megasas_instance_template megasas_instance_template_fusion ;
2005-09-20 21:46:58 +00:00
/**
* megasas_issue_polled - Issues a polling command
* @ instance : Adapter soft state
2010-12-15 03:17:17 +00:00
* @ cmd : Command packet to be issued
2005-09-20 21:46:58 +00:00
*
2015-05-06 13:31:02 +00:00
* For polling , MFI requires the cmd_status to be set to MFI_STAT_INVALID_STATUS before posting .
2005-09-20 21:46:58 +00:00
*/
2010-12-21 21:34:31 +00:00
int
2005-09-20 21:46:58 +00:00
megasas_issue_polled ( struct megasas_instance * instance , struct megasas_cmd * cmd )
{
struct megasas_header * frame_hdr = & cmd - > frame - > hdr ;
2016-01-28 15:34:23 +00:00
frame_hdr - > cmd_status = MFI_STAT_INVALID_STATUS ;
2013-09-06 10:20:52 +00:00
frame_hdr - > flags | = cpu_to_le16 ( MFI_FRAME_DONT_POST_IN_REPLY_QUEUE ) ;
2005-09-20 21:46:58 +00:00
2017-02-10 08:59:09 +00:00
if ( atomic_read ( & instance - > adprecovery ) = = MEGASAS_HW_CRITICAL_ERROR ) {
2016-01-28 15:34:23 +00:00
dev_err ( & instance - > pdev - > dev , " Failed from %s %d \n " ,
__func__ , __LINE__ ) ;
return DCMD_NOT_FIRED ;
}
2005-09-20 21:46:58 +00:00
2017-02-10 08:59:09 +00:00
instance - > instancet - > issue_dcmd ( instance , cmd ) ;
2016-01-28 15:34:23 +00:00
return wait_and_poll ( instance , cmd , instance - > requestorId ?
MEGASAS_ROUTINE_WAIT_TIME_VF : MFI_IO_TIMEOUT_SECS ) ;
2005-09-20 21:46:58 +00:00
}
/**
* megasas_issue_blocked_cmd - Synchronous wrapper around regular FW cmds
* @ instance : Adapter soft state
* @ cmd : Command to be issued
2014-02-12 18:06:15 +00:00
* @ timeout : Timeout in seconds
2005-09-20 21:46:58 +00:00
*
* This function waits on an event for the command to be returned from ISR .
2006-10-03 20:19:21 +00:00
* Max wait time is MEGASAS_INTERNAL_CMD_WAIT_TIME secs
2005-09-20 21:46:58 +00:00
* Used to issue ioctl commands .
*/
2014-09-12 13:27:58 +00:00
int
2005-09-20 21:46:58 +00:00
megasas_issue_blocked_cmd ( struct megasas_instance * instance ,
2014-02-12 18:06:15 +00:00
struct megasas_cmd * cmd , int timeout )
2005-09-20 21:46:58 +00:00
{
2014-02-12 18:06:15 +00:00
int ret = 0 ;
2015-05-06 13:31:02 +00:00
cmd - > cmd_status_drv = MFI_STAT_INVALID_STATUS ;
2005-09-20 21:46:58 +00:00
2017-02-10 08:59:09 +00:00
if ( atomic_read ( & instance - > adprecovery ) = = MEGASAS_HW_CRITICAL_ERROR ) {
2016-01-28 15:34:23 +00:00
dev_err ( & instance - > pdev - > dev , " Failed from %s %d \n " ,
__func__ , __LINE__ ) ;
return DCMD_NOT_FIRED ;
}
2017-02-10 08:59:09 +00:00
instance - > instancet - > issue_dcmd ( instance , cmd ) ;
2014-02-12 18:06:15 +00:00
if ( timeout ) {
ret = wait_event_timeout ( instance - > int_cmd_wait_q ,
2015-05-06 13:31:02 +00:00
cmd - > cmd_status_drv ! = MFI_STAT_INVALID_STATUS , timeout * HZ ) ;
2016-01-28 15:34:23 +00:00
if ( ! ret ) {
2019-05-07 17:05:38 +00:00
dev_err ( & instance - > pdev - > dev ,
" DCMD(opcode: 0x%x) is timed out, func:%s \n " ,
cmd - > frame - > dcmd . opcode , __func__ ) ;
2016-01-28 15:34:23 +00:00
return DCMD_TIMEOUT ;
}
2014-02-12 18:06:15 +00:00
} else
wait_event ( instance - > int_cmd_wait_q ,
2015-05-06 13:31:02 +00:00
cmd - > cmd_status_drv ! = MFI_STAT_INVALID_STATUS ) ;
2005-09-20 21:46:58 +00:00
2015-05-06 13:31:02 +00:00
return ( cmd - > cmd_status_drv = = MFI_STAT_OK ) ?
2016-01-28 15:34:23 +00:00
DCMD_SUCCESS : DCMD_FAILED ;
2005-09-20 21:46:58 +00:00
}
/**
* megasas_issue_blocked_abort_cmd - Aborts previously issued cmd
* @ instance : Adapter soft state
* @ cmd_to_abort : Previously issued cmd to be aborted
2014-02-12 18:06:15 +00:00
* @ timeout : Timeout in seconds
2005-09-20 21:46:58 +00:00
*
2014-02-12 18:06:15 +00:00
* MFI firmware can abort previously issued AEN comamnd ( automatic event
2005-09-20 21:46:58 +00:00
* notification ) . The megasas_issue_blocked_abort_cmd ( ) issues such abort
2006-10-03 20:19:21 +00:00
* cmd and waits for return status .
* Max wait time is MEGASAS_INTERNAL_CMD_WAIT_TIME secs
2005-09-20 21:46:58 +00:00
*/
static int
megasas_issue_blocked_abort_cmd ( struct megasas_instance * instance ,
2014-02-12 18:06:15 +00:00
struct megasas_cmd * cmd_to_abort , int timeout )
2005-09-20 21:46:58 +00:00
{
struct megasas_cmd * cmd ;
struct megasas_abort_frame * abort_fr ;
2014-02-12 18:06:15 +00:00
int ret = 0 ;
2019-05-07 17:05:38 +00:00
u32 opcode ;
2005-09-20 21:46:58 +00:00
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd )
return - 1 ;
abort_fr = & cmd - > frame - > abort ;
/*
* Prepare and issue the abort frame
*/
abort_fr - > cmd = MFI_CMD_ABORT ;
2015-05-06 13:31:02 +00:00
abort_fr - > cmd_status = MFI_STAT_INVALID_STATUS ;
2013-09-06 10:20:52 +00:00
abort_fr - > flags = cpu_to_le16 ( 0 ) ;
abort_fr - > abort_context = cpu_to_le32 ( cmd_to_abort - > index ) ;
abort_fr - > abort_mfi_phys_addr_lo =
cpu_to_le32 ( lower_32_bits ( cmd_to_abort - > frame_phys_addr ) ) ;
abort_fr - > abort_mfi_phys_addr_hi =
cpu_to_le32 ( upper_32_bits ( cmd_to_abort - > frame_phys_addr ) ) ;
2005-09-20 21:46:58 +00:00
cmd - > sync_cmd = 1 ;
2015-05-06 13:31:02 +00:00
cmd - > cmd_status_drv = MFI_STAT_INVALID_STATUS ;
2005-09-20 21:46:58 +00:00
2017-02-10 08:59:09 +00:00
if ( atomic_read ( & instance - > adprecovery ) = = MEGASAS_HW_CRITICAL_ERROR ) {
2016-01-28 15:34:23 +00:00
dev_err ( & instance - > pdev - > dev , " Failed from %s %d \n " ,
__func__ , __LINE__ ) ;
return DCMD_NOT_FIRED ;
}
2005-09-20 21:46:58 +00:00
2017-02-10 08:59:09 +00:00
instance - > instancet - > issue_dcmd ( instance , cmd ) ;
2014-02-12 18:06:15 +00:00
if ( timeout ) {
ret = wait_event_timeout ( instance - > abort_cmd_wait_q ,
2015-05-06 13:31:02 +00:00
cmd - > cmd_status_drv ! = MFI_STAT_INVALID_STATUS , timeout * HZ ) ;
2014-02-12 18:06:15 +00:00
if ( ! ret ) {
2019-05-07 17:05:38 +00:00
opcode = cmd_to_abort - > frame - > dcmd . opcode ;
dev_err ( & instance - > pdev - > dev ,
" Abort(to be aborted DCMD opcode: 0x%x) is timed out func:%s \n " ,
opcode , __func__ ) ;
2016-01-28 15:34:23 +00:00
return DCMD_TIMEOUT ;
2014-02-12 18:06:15 +00:00
}
} else
wait_event ( instance - > abort_cmd_wait_q ,
2015-05-06 13:31:02 +00:00
cmd - > cmd_status_drv ! = MFI_STAT_INVALID_STATUS ) ;
2014-02-12 18:06:15 +00:00
2010-09-23 02:36:29 +00:00
cmd - > sync_cmd = 0 ;
2005-09-20 21:46:58 +00:00
megasas_return_cmd ( instance , cmd ) ;
2016-01-28 15:34:23 +00:00
return ( cmd - > cmd_status_drv = = MFI_STAT_OK ) ?
DCMD_SUCCESS : DCMD_FAILED ;
2005-09-20 21:46:58 +00:00
}
/**
* megasas_make_sgl32 - Prepares 32 - bit SGL
* @ instance : Adapter soft state
* @ scp : SCSI command from the mid - layer
* @ mfi_sgl : SGL to be filled in
*
* If successful , this function returns the number of SG elements . Otherwise ,
* it returnes - 1.
*/
2006-01-14 21:20:43 +00:00
static int
2005-09-20 21:46:58 +00:00
megasas_make_sgl32 ( struct megasas_instance * instance , struct scsi_cmnd * scp ,
union megasas_sgl * mfi_sgl )
{
int i ;
int sge_count ;
struct scatterlist * os_sgl ;
2007-05-25 20:04:08 +00:00
sge_count = scsi_dma_map ( scp ) ;
BUG_ON ( sge_count < 0 ) ;
2005-09-20 21:46:58 +00:00
2007-05-25 20:04:08 +00:00
if ( sge_count ) {
scsi_for_each_sg ( scp , os_sgl , sge_count , i ) {
2013-09-06 10:20:52 +00:00
mfi_sgl - > sge32 [ i ] . length = cpu_to_le32 ( sg_dma_len ( os_sgl ) ) ;
mfi_sgl - > sge32 [ i ] . phys_addr = cpu_to_le32 ( sg_dma_address ( os_sgl ) ) ;
2007-05-25 20:04:08 +00:00
}
2005-09-20 21:46:58 +00:00
}
return sge_count ;
}
/**
* megasas_make_sgl64 - Prepares 64 - bit SGL
* @ instance : Adapter soft state
* @ scp : SCSI command from the mid - layer
* @ mfi_sgl : SGL to be filled in
*
* If successful , this function returns the number of SG elements . Otherwise ,
* it returnes - 1.
*/
2006-01-14 21:20:43 +00:00
static int
2005-09-20 21:46:58 +00:00
megasas_make_sgl64 ( struct megasas_instance * instance , struct scsi_cmnd * scp ,
union megasas_sgl * mfi_sgl )
{
int i ;
int sge_count ;
struct scatterlist * os_sgl ;
2007-05-25 20:04:08 +00:00
sge_count = scsi_dma_map ( scp ) ;
BUG_ON ( sge_count < 0 ) ;
2005-09-20 21:46:58 +00:00
2007-05-25 20:04:08 +00:00
if ( sge_count ) {
scsi_for_each_sg ( scp , os_sgl , sge_count , i ) {
2013-09-06 10:20:52 +00:00
mfi_sgl - > sge64 [ i ] . length = cpu_to_le32 ( sg_dma_len ( os_sgl ) ) ;
mfi_sgl - > sge64 [ i ] . phys_addr = cpu_to_le64 ( sg_dma_address ( os_sgl ) ) ;
2007-05-25 20:04:08 +00:00
}
2005-09-20 21:46:58 +00:00
}
return sge_count ;
}
2009-10-06 20:43:28 +00:00
/**
* megasas_make_sgl_skinny - Prepares IEEE SGL
* @ instance : Adapter soft state
* @ scp : SCSI command from the mid - layer
* @ mfi_sgl : SGL to be filled in
*
* If successful , this function returns the number of SG elements . Otherwise ,
* it returnes - 1.
*/
static int
megasas_make_sgl_skinny ( struct megasas_instance * instance ,
struct scsi_cmnd * scp , union megasas_sgl * mfi_sgl )
{
int i ;
int sge_count ;
struct scatterlist * os_sgl ;
sge_count = scsi_dma_map ( scp ) ;
if ( sge_count ) {
scsi_for_each_sg ( scp , os_sgl , sge_count , i ) {
2013-09-06 10:20:52 +00:00
mfi_sgl - > sge_skinny [ i ] . length =
cpu_to_le32 ( sg_dma_len ( os_sgl ) ) ;
2009-10-06 20:43:28 +00:00
mfi_sgl - > sge_skinny [ i ] . phys_addr =
2013-09-06 10:20:52 +00:00
cpu_to_le64 ( sg_dma_address ( os_sgl ) ) ;
mfi_sgl - > sge_skinny [ i ] . flag = cpu_to_le32 ( 0 ) ;
2009-10-06 20:43:28 +00:00
}
}
return sge_count ;
}
2006-10-03 19:40:47 +00:00
/**
* megasas_get_frame_count - Computes the number of frames
2008-03-17 07:36:43 +00:00
* @ frame_type : type of frame - io or pthru frame
2006-10-03 19:40:47 +00:00
* @ sge_count : number of sg elements
*
* Returns the number of frames required for numnber of sge ' s ( sge_count )
*/
2009-10-06 20:43:28 +00:00
static u32 megasas_get_frame_count ( struct megasas_instance * instance ,
u8 sge_count , u8 frame_type )
2006-10-03 19:40:47 +00:00
{
int num_cnt ;
int sge_bytes ;
u32 sge_sz ;
2015-07-07 20:52:45 +00:00
u32 frame_count = 0 ;
2006-10-03 19:40:47 +00:00
sge_sz = ( IS_DMA64 ) ? sizeof ( struct megasas_sge64 ) :
sizeof ( struct megasas_sge32 ) ;
2009-10-06 20:43:28 +00:00
if ( instance - > flag_ieee ) {
sge_sz = sizeof ( struct megasas_sge_skinny ) ;
}
2006-10-03 19:40:47 +00:00
/*
2008-03-17 07:36:43 +00:00
* Main frame can contain 2 SGEs for 64 - bit SGLs and
* 3 SGEs for 32 - bit SGLs for ldio &
* 1 SGEs for 64 - bit SGLs and
* 2 SGEs for 32 - bit SGLs for pthru frame
*/
if ( unlikely ( frame_type = = PTHRU_FRAME ) ) {
2009-10-06 20:43:28 +00:00
if ( instance - > flag_ieee = = 1 ) {
num_cnt = sge_count - 1 ;
} else if ( IS_DMA64 )
2008-03-17 07:36:43 +00:00
num_cnt = sge_count - 1 ;
else
num_cnt = sge_count - 2 ;
} else {
2009-10-06 20:43:28 +00:00
if ( instance - > flag_ieee = = 1 ) {
num_cnt = sge_count - 1 ;
} else if ( IS_DMA64 )
2008-03-17 07:36:43 +00:00
num_cnt = sge_count - 2 ;
else
num_cnt = sge_count - 3 ;
}
2006-10-03 19:40:47 +00:00
2015-07-07 20:52:45 +00:00
if ( num_cnt > 0 ) {
2006-10-03 19:40:47 +00:00
sge_bytes = sge_sz * num_cnt ;
frame_count = ( sge_bytes / MEGAMFI_FRAME_SIZE ) +
( ( sge_bytes % MEGAMFI_FRAME_SIZE ) ? 1 : 0 ) ;
}
/* Main frame */
2015-07-07 20:52:45 +00:00
frame_count + = 1 ;
2006-10-03 19:40:47 +00:00
if ( frame_count > 7 )
frame_count = 8 ;
return frame_count ;
}
2005-09-20 21:46:58 +00:00
/**
* megasas_build_dcdb - Prepares a direct cdb ( DCDB ) command
* @ instance : Adapter soft state
* @ scp : SCSI command
* @ cmd : Command to be prepared in
*
* This function prepares CDB commands . These are typcially pass - through
* commands to the devices .
*/
2006-01-14 21:20:43 +00:00
static int
2005-09-20 21:46:58 +00:00
megasas_build_dcdb ( struct megasas_instance * instance , struct scsi_cmnd * scp ,
struct megasas_cmd * cmd )
{
u32 is_logical ;
u32 device_id ;
u16 flags = 0 ;
struct megasas_pthru_frame * pthru ;
2017-02-10 08:59:05 +00:00
is_logical = MEGASAS_IS_LOGICAL ( scp - > device ) ;
2015-04-23 11:00:39 +00:00
device_id = MEGASAS_DEV_INDEX ( scp ) ;
2005-09-20 21:46:58 +00:00
pthru = ( struct megasas_pthru_frame * ) cmd - > frame ;
2018-10-10 17:31:25 +00:00
if ( scp - > sc_data_direction = = DMA_TO_DEVICE )
2005-09-20 21:46:58 +00:00
flags = MFI_FRAME_DIR_WRITE ;
2018-10-10 17:31:25 +00:00
else if ( scp - > sc_data_direction = = DMA_FROM_DEVICE )
2005-09-20 21:46:58 +00:00
flags = MFI_FRAME_DIR_READ ;
2018-10-10 17:31:25 +00:00
else if ( scp - > sc_data_direction = = DMA_NONE )
2005-09-20 21:46:58 +00:00
flags = MFI_FRAME_DIR_NONE ;
2009-10-06 20:43:28 +00:00
if ( instance - > flag_ieee = = 1 ) {
flags | = MFI_FRAME_IEEE ;
}
2005-09-20 21:46:58 +00:00
/*
* Prepare the DCDB frame
*/
pthru - > cmd = ( is_logical ) ? MFI_CMD_LD_SCSI_IO : MFI_CMD_PD_SCSI_IO ;
pthru - > cmd_status = 0x0 ;
pthru - > scsi_status = 0x0 ;
pthru - > target_id = device_id ;
pthru - > lun = scp - > device - > lun ;
pthru - > cdb_len = scp - > cmd_len ;
pthru - > timeout = 0 ;
2009-12-06 15:24:21 +00:00
pthru - > pad_0 = 0 ;
2013-09-06 10:20:52 +00:00
pthru - > flags = cpu_to_le16 ( flags ) ;
pthru - > data_xfer_len = cpu_to_le32 ( scsi_bufflen ( scp ) ) ;
2005-09-20 21:46:58 +00:00
memcpy ( pthru - > cdb , scp - > cmnd , scp - > cmd_len ) ;
2009-10-06 20:12:21 +00:00
/*
2015-07-07 20:52:45 +00:00
* If the command is for the tape device , set the
* pthru timeout to the os layer timeout value .
*/
2009-10-06 20:12:21 +00:00
if ( scp - > device - > type = = TYPE_TAPE ) {
if ( ( scp - > request - > timeout / HZ ) > 0xFFFF )
2015-04-23 11:03:09 +00:00
pthru - > timeout = cpu_to_le16 ( 0xFFFF ) ;
2009-10-06 20:12:21 +00:00
else
2013-09-06 10:20:52 +00:00
pthru - > timeout = cpu_to_le16 ( scp - > request - > timeout / HZ ) ;
2009-10-06 20:12:21 +00:00
}
2005-09-20 21:46:58 +00:00
/*
* Construct SGL
*/
2009-10-06 20:43:28 +00:00
if ( instance - > flag_ieee = = 1 ) {
2013-09-06 10:20:52 +00:00
pthru - > flags | = cpu_to_le16 ( MFI_FRAME_SGL64 ) ;
2009-10-06 20:43:28 +00:00
pthru - > sge_count = megasas_make_sgl_skinny ( instance , scp ,
& pthru - > sgl ) ;
} else if ( IS_DMA64 ) {
2013-09-06 10:20:52 +00:00
pthru - > flags | = cpu_to_le16 ( MFI_FRAME_SGL64 ) ;
2005-09-20 21:46:58 +00:00
pthru - > sge_count = megasas_make_sgl64 ( instance , scp ,
& pthru - > sgl ) ;
} else
pthru - > sge_count = megasas_make_sgl32 ( instance , scp ,
& pthru - > sgl ) ;
2009-12-06 15:30:19 +00:00
if ( pthru - > sge_count > instance - > max_num_sge ) {
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " DCDB too many SGE NUM=%x \n " ,
2009-12-06 15:30:19 +00:00
pthru - > sge_count ) ;
return 0 ;
}
2005-09-20 21:46:58 +00:00
/*
* Sense info specific
*/
pthru - > sense_len = SCSI_SENSE_BUFFERSIZE ;
2013-09-06 10:20:52 +00:00
pthru - > sense_buf_phys_addr_hi =
cpu_to_le32 ( upper_32_bits ( cmd - > sense_phys_addr ) ) ;
pthru - > sense_buf_phys_addr_lo =
cpu_to_le32 ( lower_32_bits ( cmd - > sense_phys_addr ) ) ;
2005-09-20 21:46:58 +00:00
/*
* Compute the total number of frames this command consumes . FW uses
* this number to pull sufficient number of frames from host memory .
*/
2009-10-06 20:43:28 +00:00
cmd - > frame_count = megasas_get_frame_count ( instance , pthru - > sge_count ,
2008-03-17 07:36:43 +00:00
PTHRU_FRAME ) ;
2005-09-20 21:46:58 +00:00
return cmd - > frame_count ;
}
/**
* megasas_build_ldio - Prepares IOs to logical devices
* @ instance : Adapter soft state
* @ scp : SCSI command
2009-07-16 15:13:03 +00:00
* @ cmd : Command to be prepared
2005-09-20 21:46:58 +00:00
*
* Frames ( and accompanying SGLs ) for regular SCSI IOs use this function .
*/
2006-01-14 21:20:43 +00:00
static int
2005-09-20 21:46:58 +00:00
megasas_build_ldio ( struct megasas_instance * instance , struct scsi_cmnd * scp ,
struct megasas_cmd * cmd )
{
u32 device_id ;
u8 sc = scp - > cmnd [ 0 ] ;
u16 flags = 0 ;
struct megasas_io_frame * ldio ;
2015-04-23 11:00:39 +00:00
device_id = MEGASAS_DEV_INDEX ( scp ) ;
2005-09-20 21:46:58 +00:00
ldio = ( struct megasas_io_frame * ) cmd - > frame ;
2018-10-10 17:31:25 +00:00
if ( scp - > sc_data_direction = = DMA_TO_DEVICE )
2005-09-20 21:46:58 +00:00
flags = MFI_FRAME_DIR_WRITE ;
2018-10-10 17:31:25 +00:00
else if ( scp - > sc_data_direction = = DMA_FROM_DEVICE )
2005-09-20 21:46:58 +00:00
flags = MFI_FRAME_DIR_READ ;
2009-10-06 20:43:28 +00:00
if ( instance - > flag_ieee = = 1 ) {
flags | = MFI_FRAME_IEEE ;
}
2005-09-20 21:46:58 +00:00
/*
2006-10-03 19:40:47 +00:00
* Prepare the Logical IO frame : 2 nd bit is zero for all read cmds
2005-09-20 21:46:58 +00:00
*/
ldio - > cmd = ( sc & 0x02 ) ? MFI_CMD_LD_WRITE : MFI_CMD_LD_READ ;
ldio - > cmd_status = 0x0 ;
ldio - > scsi_status = 0x0 ;
ldio - > target_id = device_id ;
ldio - > timeout = 0 ;
ldio - > reserved_0 = 0 ;
ldio - > pad_0 = 0 ;
2013-09-06 10:20:52 +00:00
ldio - > flags = cpu_to_le16 ( flags ) ;
2005-09-20 21:46:58 +00:00
ldio - > start_lba_hi = 0 ;
ldio - > access_byte = ( scp - > cmd_len ! = 6 ) ? scp - > cmnd [ 1 ] : 0 ;
/*
* 6 - byte READ ( 0x08 ) or WRITE ( 0x0A ) cdb
*/
if ( scp - > cmd_len = = 6 ) {
2013-09-06 10:20:52 +00:00
ldio - > lba_count = cpu_to_le32 ( ( u32 ) scp - > cmnd [ 4 ] ) ;
ldio - > start_lba_lo = cpu_to_le32 ( ( ( u32 ) scp - > cmnd [ 1 ] < < 16 ) |
( ( u32 ) scp - > cmnd [ 2 ] < < 8 ) |
( u32 ) scp - > cmnd [ 3 ] ) ;
2005-09-20 21:46:58 +00:00
2013-09-06 10:20:52 +00:00
ldio - > start_lba_lo & = cpu_to_le32 ( 0x1FFFFF ) ;
2005-09-20 21:46:58 +00:00
}
/*
* 10 - byte READ ( 0x28 ) or WRITE ( 0x2A ) cdb
*/
else if ( scp - > cmd_len = = 10 ) {
2013-09-06 10:20:52 +00:00
ldio - > lba_count = cpu_to_le32 ( ( u32 ) scp - > cmnd [ 8 ] |
( ( u32 ) scp - > cmnd [ 7 ] < < 8 ) ) ;
ldio - > start_lba_lo = cpu_to_le32 ( ( ( u32 ) scp - > cmnd [ 2 ] < < 24 ) |
( ( u32 ) scp - > cmnd [ 3 ] < < 16 ) |
( ( u32 ) scp - > cmnd [ 4 ] < < 8 ) |
( u32 ) scp - > cmnd [ 5 ] ) ;
2005-09-20 21:46:58 +00:00
}
/*
* 12 - byte READ ( 0xA8 ) or WRITE ( 0xAA ) cdb
*/
else if ( scp - > cmd_len = = 12 ) {
2013-09-06 10:20:52 +00:00
ldio - > lba_count = cpu_to_le32 ( ( ( u32 ) scp - > cmnd [ 6 ] < < 24 ) |
( ( u32 ) scp - > cmnd [ 7 ] < < 16 ) |
( ( u32 ) scp - > cmnd [ 8 ] < < 8 ) |
( u32 ) scp - > cmnd [ 9 ] ) ;
2005-09-20 21:46:58 +00:00
2013-09-06 10:20:52 +00:00
ldio - > start_lba_lo = cpu_to_le32 ( ( ( u32 ) scp - > cmnd [ 2 ] < < 24 ) |
( ( u32 ) scp - > cmnd [ 3 ] < < 16 ) |
( ( u32 ) scp - > cmnd [ 4 ] < < 8 ) |
( u32 ) scp - > cmnd [ 5 ] ) ;
2005-09-20 21:46:58 +00:00
}
/*
* 16 - byte READ ( 0x88 ) or WRITE ( 0x8A ) cdb
*/
else if ( scp - > cmd_len = = 16 ) {
2013-09-06 10:20:52 +00:00
ldio - > lba_count = cpu_to_le32 ( ( ( u32 ) scp - > cmnd [ 10 ] < < 24 ) |
( ( u32 ) scp - > cmnd [ 11 ] < < 16 ) |
( ( u32 ) scp - > cmnd [ 12 ] < < 8 ) |
( u32 ) scp - > cmnd [ 13 ] ) ;
2005-09-20 21:46:58 +00:00
2013-09-06 10:20:52 +00:00
ldio - > start_lba_lo = cpu_to_le32 ( ( ( u32 ) scp - > cmnd [ 6 ] < < 24 ) |
( ( u32 ) scp - > cmnd [ 7 ] < < 16 ) |
( ( u32 ) scp - > cmnd [ 8 ] < < 8 ) |
( u32 ) scp - > cmnd [ 9 ] ) ;
2005-09-20 21:46:58 +00:00
2013-09-06 10:20:52 +00:00
ldio - > start_lba_hi = cpu_to_le32 ( ( ( u32 ) scp - > cmnd [ 2 ] < < 24 ) |
( ( u32 ) scp - > cmnd [ 3 ] < < 16 ) |
( ( u32 ) scp - > cmnd [ 4 ] < < 8 ) |
( u32 ) scp - > cmnd [ 5 ] ) ;
2005-09-20 21:46:58 +00:00
}
/*
* Construct SGL
*/
2009-10-06 20:43:28 +00:00
if ( instance - > flag_ieee ) {
2013-09-06 10:20:52 +00:00
ldio - > flags | = cpu_to_le16 ( MFI_FRAME_SGL64 ) ;
2009-10-06 20:43:28 +00:00
ldio - > sge_count = megasas_make_sgl_skinny ( instance , scp ,
& ldio - > sgl ) ;
} else if ( IS_DMA64 ) {
2013-09-06 10:20:52 +00:00
ldio - > flags | = cpu_to_le16 ( MFI_FRAME_SGL64 ) ;
2005-09-20 21:46:58 +00:00
ldio - > sge_count = megasas_make_sgl64 ( instance , scp , & ldio - > sgl ) ;
} else
ldio - > sge_count = megasas_make_sgl32 ( instance , scp , & ldio - > sgl ) ;
2009-12-06 15:30:19 +00:00
if ( ldio - > sge_count > instance - > max_num_sge ) {
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " build_ld_io: sge_count = %x \n " ,
2009-12-06 15:30:19 +00:00
ldio - > sge_count ) ;
return 0 ;
}
2005-09-20 21:46:58 +00:00
/*
* Sense info specific
*/
ldio - > sense_len = SCSI_SENSE_BUFFERSIZE ;
ldio - > sense_buf_phys_addr_hi = 0 ;
2013-09-06 10:20:52 +00:00
ldio - > sense_buf_phys_addr_lo = cpu_to_le32 ( cmd - > sense_phys_addr ) ;
2005-09-20 21:46:58 +00:00
2006-10-03 19:40:47 +00:00
/*
* Compute the total number of frames this command consumes . FW uses
* this number to pull sufficient number of frames from host memory .
*/
2009-10-06 20:43:28 +00:00
cmd - > frame_count = megasas_get_frame_count ( instance ,
ldio - > sge_count , IO_FRAME ) ;
2005-09-20 21:46:58 +00:00
return cmd - > frame_count ;
}
/**
2015-01-05 14:36:03 +00:00
* megasas_cmd_type - Checks if the cmd is for logical drive / sysPD
* and whether it ' s RW or non RW
2006-01-25 19:53:25 +00:00
* @ scmd : SCSI command
2010-12-15 03:17:17 +00:00
*
2005-09-20 21:46:58 +00:00
*/
2015-01-05 14:36:03 +00:00
inline int megasas_cmd_type ( struct scsi_cmnd * cmd )
2005-09-20 21:46:58 +00:00
{
2015-01-05 14:36:03 +00:00
int ret ;
2006-01-25 19:53:25 +00:00
switch ( cmd - > cmnd [ 0 ] ) {
case READ_10 :
case WRITE_10 :
case READ_12 :
case WRITE_12 :
case READ_6 :
case WRITE_6 :
case READ_16 :
case WRITE_16 :
2017-02-10 08:59:05 +00:00
ret = ( MEGASAS_IS_LOGICAL ( cmd - > device ) ) ?
2015-01-05 14:36:03 +00:00
READ_WRITE_LDIO : READ_WRITE_SYSPDIO ;
break ;
2006-01-25 19:53:25 +00:00
default :
2017-02-10 08:59:05 +00:00
ret = ( MEGASAS_IS_LOGICAL ( cmd - > device ) ) ?
2015-01-05 14:36:03 +00:00
NON_READ_WRITE_LDIO : NON_READ_WRITE_SYSPDIO ;
2005-09-20 21:46:58 +00:00
}
2015-01-05 14:36:03 +00:00
return ret ;
2005-09-20 21:46:58 +00:00
}
2006-10-03 20:09:14 +00:00
/**
* megasas_dump_pending_frames - Dumps the frame address of all pending cmds
2015-07-07 20:52:45 +00:00
* in FW
2006-10-03 20:09:14 +00:00
* @ instance : Adapter soft state
*/
static inline void
megasas_dump_pending_frames ( struct megasas_instance * instance )
{
struct megasas_cmd * cmd ;
int i , n ;
union megasas_sgl * mfi_sgl ;
struct megasas_io_frame * ldio ;
struct megasas_pthru_frame * pthru ;
u32 sgcount ;
2017-02-10 08:59:25 +00:00
u16 max_cmd = instance - > max_fw_cmds ;
2006-10-03 20:09:14 +00:00
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " [%d]: Dumping Frame Phys Address of all pending cmds in FW \n " , instance - > host - > host_no ) ;
dev_err ( & instance - > pdev - > dev , " [%d]: Total OS Pending cmds : %d \n " , instance - > host - > host_no , atomic_read ( & instance - > fw_outstanding ) ) ;
2006-10-03 20:09:14 +00:00
if ( IS_DMA64 )
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " [%d]: 64 bit SGLs were sent to FW \n " , instance - > host - > host_no ) ;
2006-10-03 20:09:14 +00:00
else
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " [%d]: 32 bit SGLs were sent to FW \n " , instance - > host - > host_no ) ;
2006-10-03 20:09:14 +00:00
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " [%d]: Pending OS cmds in FW : \n " , instance - > host - > host_no ) ;
2006-10-03 20:09:14 +00:00
for ( i = 0 ; i < max_cmd ; i + + ) {
cmd = instance - > cmd_list [ i ] ;
2015-07-07 20:52:45 +00:00
if ( ! cmd - > scmd )
2006-10-03 20:09:14 +00:00
continue ;
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " [%d]: Frame addr :0x%08lx : " , instance - > host - > host_no , ( unsigned long ) cmd - > frame_phys_addr ) ;
2015-01-05 14:36:03 +00:00
if ( megasas_cmd_type ( cmd - > scmd ) = = READ_WRITE_LDIO ) {
2006-10-03 20:09:14 +00:00
ldio = ( struct megasas_io_frame * ) cmd - > frame ;
mfi_sgl = & ldio - > sgl ;
sgcount = ldio - > sge_count ;
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " [%d]: frame count : 0x%x, Cmd : 0x%x, Tgt id : 0x%x, "
2013-09-06 10:20:52 +00:00
" lba lo : 0x%x, lba_hi : 0x%x, sense_buf addr : 0x%x,sge count : 0x%x \n " ,
instance - > host - > host_no , cmd - > frame_count , ldio - > cmd , ldio - > target_id ,
le32_to_cpu ( ldio - > start_lba_lo ) , le32_to_cpu ( ldio - > start_lba_hi ) ,
le32_to_cpu ( ldio - > sense_buf_phys_addr_lo ) , sgcount ) ;
2015-07-07 20:52:45 +00:00
} else {
2006-10-03 20:09:14 +00:00
pthru = ( struct megasas_pthru_frame * ) cmd - > frame ;
mfi_sgl = & pthru - > sgl ;
sgcount = pthru - > sge_count ;
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " [%d]: frame count : 0x%x, Cmd : 0x%x, Tgt id : 0x%x, "
2013-09-06 10:20:52 +00:00
" lun : 0x%x, cdb_len : 0x%x, data xfer len : 0x%x, sense_buf addr : 0x%x,sge count : 0x%x \n " ,
instance - > host - > host_no , cmd - > frame_count , pthru - > cmd , pthru - > target_id ,
pthru - > lun , pthru - > cdb_len , le32_to_cpu ( pthru - > data_xfer_len ) ,
le32_to_cpu ( pthru - > sense_buf_phys_addr_lo ) , sgcount ) ;
2006-10-03 20:09:14 +00:00
}
2015-07-07 20:52:45 +00:00
if ( megasas_dbg_lvl & MEGASAS_DBG_LVL ) {
for ( n = 0 ; n < sgcount ; n + + ) {
if ( IS_DMA64 )
dev_err ( & instance - > pdev - > dev , " sgl len : 0x%x, sgl addr : 0x%llx \n " ,
le32_to_cpu ( mfi_sgl - > sge64 [ n ] . length ) ,
le64_to_cpu ( mfi_sgl - > sge64 [ n ] . phys_addr ) ) ;
else
dev_err ( & instance - > pdev - > dev , " sgl len : 0x%x, sgl addr : 0x%x \n " ,
le32_to_cpu ( mfi_sgl - > sge32 [ n ] . length ) ,
le32_to_cpu ( mfi_sgl - > sge32 [ n ] . phys_addr ) ) ;
2006-10-03 20:09:14 +00:00
}
}
} /*for max_cmd*/
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " [%d]: Pending Internal cmds in FW : \n " , instance - > host - > host_no ) ;
2006-10-03 20:09:14 +00:00
for ( i = 0 ; i < max_cmd ; i + + ) {
cmd = instance - > cmd_list [ i ] ;
2015-07-07 20:52:45 +00:00
if ( cmd - > sync_cmd = = 1 )
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " 0x%08lx : " , ( unsigned long ) cmd - > frame_phys_addr ) ;
2006-10-03 20:09:14 +00:00
}
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " [%d]: Dumping Done \n \n " , instance - > host - > host_no ) ;
2006-10-03 20:09:14 +00:00
}
2010-12-21 18:23:23 +00:00
u32
megasas_build_and_issue_cmd ( struct megasas_instance * instance ,
struct scsi_cmnd * scmd )
{
struct megasas_cmd * cmd ;
u32 frame_count ;
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd )
return SCSI_MLQUEUE_HOST_BUSY ;
/*
* Logical drive command
*/
2015-01-05 14:36:03 +00:00
if ( megasas_cmd_type ( scmd ) = = READ_WRITE_LDIO )
2010-12-21 18:23:23 +00:00
frame_count = megasas_build_ldio ( instance , scmd , cmd ) ;
else
frame_count = megasas_build_dcdb ( instance , scmd , cmd ) ;
if ( ! frame_count )
goto out_return_cmd ;
cmd - > scmd = scmd ;
scmd - > SCp . ptr = ( char * ) cmd ;
/*
* Issue the command to the FW
*/
atomic_inc ( & instance - > fw_outstanding ) ;
instance - > instancet - > fire_cmd ( instance , cmd - > frame_phys_addr ,
cmd - > frame_count - 1 , instance - > reg_set ) ;
return 0 ;
out_return_cmd :
megasas_return_cmd ( instance , cmd ) ;
2016-01-28 15:34:29 +00:00
return SCSI_MLQUEUE_HOST_BUSY ;
2010-12-21 18:23:23 +00:00
}
2005-09-20 21:46:58 +00:00
/**
* megasas_queue_command - Queue entry point
* @ scmd : SCSI command to be queued
* @ done : Callback entry point
*/
static int
2014-09-12 13:27:38 +00:00
megasas_queue_command ( struct Scsi_Host * shost , struct scsi_cmnd * scmd )
2005-09-20 21:46:58 +00:00
{
struct megasas_instance * instance ;
2016-01-28 15:34:25 +00:00
struct MR_PRIV_DEVICE * mr_device_priv_data ;
2005-09-20 21:46:58 +00:00
instance = ( struct megasas_instance * )
scmd - > device - > host - > hostdata ;
2007-02-14 20:34:46 +00:00
2014-11-17 09:54:08 +00:00
if ( instance - > unload = = 1 ) {
scmd - > result = DID_NO_CONNECT < < 16 ;
scmd - > scsi_done ( scmd ) ;
return 0 ;
}
2010-09-23 02:36:29 +00:00
if ( instance - > issuepend_done = = 0 )
2007-02-14 20:34:46 +00:00
return SCSI_MLQUEUE_HOST_BUSY ;
2013-05-22 06:59:28 +00:00
2014-03-10 09:51:56 +00:00
/* Check for an mpio path and adjust behavior */
2016-01-28 15:34:32 +00:00
if ( atomic_read ( & instance - > adprecovery ) = = MEGASAS_ADPRESET_SM_INFAULT ) {
2014-03-10 09:51:56 +00:00
if ( megasas_check_mpio_paths ( instance , scmd ) = =
2017-02-10 08:59:07 +00:00
( DID_REQUEUE < < 16 ) ) {
2014-03-10 09:51:56 +00:00
return SCSI_MLQUEUE_HOST_BUSY ;
} else {
scmd - > result = DID_NO_CONNECT < < 16 ;
2014-09-12 13:27:38 +00:00
scmd - > scsi_done ( scmd ) ;
2014-03-10 09:51:56 +00:00
return 0 ;
}
}
2016-01-28 15:34:32 +00:00
if ( atomic_read ( & instance - > adprecovery ) = = MEGASAS_HW_CRITICAL_ERROR ) {
2014-03-10 09:51:56 +00:00
scmd - > result = DID_NO_CONNECT < < 16 ;
2014-09-12 13:27:38 +00:00
scmd - > scsi_done ( scmd ) ;
2013-05-22 06:59:28 +00:00
return 0 ;
}
2016-01-28 15:34:25 +00:00
mr_device_priv_data = scmd - > device - > hostdata ;
if ( ! mr_device_priv_data ) {
scmd - > result = DID_NO_CONNECT < < 16 ;
scmd - > scsi_done ( scmd ) ;
return 0 ;
}
2016-01-28 15:34:32 +00:00
if ( atomic_read ( & instance - > adprecovery ) ! = MEGASAS_HBA_OPERATIONAL )
2010-09-23 02:36:29 +00:00
return SCSI_MLQUEUE_HOST_BUSY ;
2016-01-28 15:34:32 +00:00
if ( mr_device_priv_data - > tm_busy )
2016-01-28 15:34:25 +00:00
return SCSI_MLQUEUE_DEVICE_BUSY ;
2010-09-23 02:36:29 +00:00
2005-09-20 21:46:58 +00:00
scmd - > result = 0 ;
2017-02-10 08:59:05 +00:00
if ( MEGASAS_IS_LOGICAL ( scmd - > device ) & &
2014-09-12 13:27:33 +00:00
( scmd - > device - > id > = instance - > fw_supported_vd_count | |
scmd - > device - > lun ) ) {
2006-01-25 19:53:25 +00:00
scmd - > result = DID_BAD_TARGET < < 16 ;
goto out_done ;
2005-09-20 21:46:58 +00:00
}
2017-02-10 08:59:05 +00:00
if ( ( scmd - > cmnd [ 0 ] = = SYNCHRONIZE_CACHE ) & &
MEGASAS_IS_LOGICAL ( scmd - > device ) & &
( ! instance - > fw_sync_cache_support ) ) {
2007-02-14 21:00:55 +00:00
scmd - > result = DID_OK < < 16 ;
goto out_done ;
}
2016-01-28 15:34:29 +00:00
return instance - > instancet - > build_and_issue_cmd ( instance , scmd ) ;
2006-01-25 19:53:25 +00:00
out_done :
2014-09-12 13:27:38 +00:00
scmd - > scsi_done ( scmd ) ;
2006-01-25 19:53:25 +00:00
return 0 ;
2005-09-20 21:46:58 +00:00
}
2009-10-06 20:33:06 +00:00
static struct megasas_instance * megasas_lookup_instance ( u16 host_no )
{
int i ;
for ( i = 0 ; i < megasas_mgmt_info . max_index ; i + + ) {
if ( ( megasas_mgmt_info . instance [ i ] ) & &
( megasas_mgmt_info . instance [ i ] - > host - > host_no = = host_no ) )
return megasas_mgmt_info . instance [ i ] ;
}
return NULL ;
}
2015-10-15 08:10:44 +00:00
/*
2017-02-10 08:59:10 +00:00
* megasas_set_dynamic_target_properties -
* Device property set by driver may not be static and it is required to be
* updated after OCR
*
* set tm_capable .
* set dma alignment ( only for eedp protection enable vd ) .
2015-10-15 08:10:44 +00:00
*
* @ sdev : OS provided scsi device
*
* Returns void
*/
2018-06-04 10:45:12 +00:00
void megasas_set_dynamic_target_properties ( struct scsi_device * sdev ,
bool is_target_prop )
2015-10-15 08:10:44 +00:00
{
2017-02-10 08:59:10 +00:00
u16 pd_index = 0 , ld ;
u32 device_id ;
2015-10-15 08:10:44 +00:00
struct megasas_instance * instance ;
struct fusion_context * fusion ;
2016-01-28 15:34:25 +00:00
struct MR_PRIV_DEVICE * mr_device_priv_data ;
struct MR_PD_CFG_SEQ_NUM_SYNC * pd_sync ;
2015-10-15 08:10:44 +00:00
struct MR_LD_RAID * raid ;
struct MR_DRV_RAID_MAP_ALL * local_map_ptr ;
instance = megasas_lookup_instance ( sdev - > host - > host_no ) ;
fusion = instance - > ctrl_context ;
2016-01-28 15:34:25 +00:00
mr_device_priv_data = sdev - > hostdata ;
2015-10-15 08:10:44 +00:00
2017-02-10 08:59:06 +00:00
if ( ! fusion | | ! mr_device_priv_data )
2015-10-15 08:10:44 +00:00
return ;
2017-02-10 08:59:06 +00:00
if ( MEGASAS_IS_LOGICAL ( sdev ) ) {
2015-10-15 08:10:44 +00:00
device_id = ( ( sdev - > channel % 2 ) * MEGASAS_MAX_DEV_PER_CHANNEL )
+ sdev - > id ;
local_map_ptr = fusion - > ld_drv_map [ ( instance - > map_id & 1 ) ] ;
ld = MR_TargetIdToLdGet ( device_id , local_map_ptr ) ;
2017-02-10 08:59:06 +00:00
if ( ld > = instance - > fw_supported_vd_count )
return ;
2015-10-15 08:10:44 +00:00
raid = MR_LdRaidGet ( ld , local_map_ptr ) ;
if ( raid - > capability . ldPiMode = = MR_PROT_INFO_TYPE_CONTROLLER )
2017-02-10 08:59:10 +00:00
blk_queue_update_dma_alignment ( sdev - > request_queue , 0x7 ) ;
2017-02-10 08:59:06 +00:00
2016-01-28 15:34:25 +00:00
mr_device_priv_data - > is_tm_capable =
raid - > capability . tmCapable ;
2017-02-10 08:59:06 +00:00
} else if ( instance - > use_seqnum_jbod_fp ) {
pd_index = ( sdev - > channel * MEGASAS_MAX_DEV_PER_CHANNEL ) +
2017-02-10 08:59:10 +00:00
sdev - > id ;
2017-02-10 08:59:06 +00:00
pd_sync = ( void * ) fusion - > pd_seq_sync
[ ( instance - > pd_seq_map_id - 1 ) & 1 ] ;
mr_device_priv_data - > is_tm_capable =
2017-02-10 08:59:10 +00:00
pd_sync - > seq [ pd_index ] . capability . tmCapable ;
2015-10-15 08:10:44 +00:00
}
2018-06-04 10:45:12 +00:00
if ( is_target_prop & & instance - > tgt_prop - > reset_tmo ) {
/*
* If FW provides a target reset timeout value , driver will use
* it . If not set , fallback to default values .
*/
mr_device_priv_data - > target_reset_tmo =
min_t ( u8 , instance - > max_reset_tmo ,
instance - > tgt_prop - > reset_tmo ) ;
mr_device_priv_data - > task_abort_tmo = instance - > task_abort_tmo ;
} else {
mr_device_priv_data - > target_reset_tmo =
MEGASAS_DEFAULT_TM_TIMEOUT ;
mr_device_priv_data - > task_abort_tmo =
MEGASAS_DEFAULT_TM_TIMEOUT ;
}
2015-10-15 08:10:44 +00:00
}
2017-02-10 08:59:10 +00:00
/*
* megasas_set_nvme_device_properties -
* set nomerges = 2
* set virtual page boundary = 4 K ( current mr_nvme_pg_size is 4 K ) .
* set maximum io transfer = MDTS of NVME device provided by MR firmware .
*
* MR firmware provides value in KB . Caller of this function converts
* kb into bytes .
*
* e . a MDTS = 5 means 2 ^ 5 * nvme page size . ( In case of 4 K page size ,
* MR firmware provides value 128 as ( 32 * 4 K ) = 128 K .
*
* @ sdev : scsi device
* @ max_io_size : maximum io transfer size
*
*/
static inline void
megasas_set_nvme_device_properties ( struct scsi_device * sdev , u32 max_io_size )
2016-01-28 15:34:26 +00:00
{
struct megasas_instance * instance ;
2017-02-10 08:59:10 +00:00
u32 mr_nvme_pg_size ;
2016-01-28 15:34:26 +00:00
2017-02-10 08:59:10 +00:00
instance = ( struct megasas_instance * ) sdev - > host - > hostdata ;
mr_nvme_pg_size = max_t ( u32 , instance - > nvme_page_size ,
MR_DEFAULT_NVME_PAGE_SIZE ) ;
2016-01-28 15:34:26 +00:00
2017-02-10 08:59:10 +00:00
blk_queue_max_hw_sectors ( sdev - > request_queue , ( max_io_size / 512 ) ) ;
2016-01-28 15:34:26 +00:00
2018-03-08 01:10:10 +00:00
blk_queue_flag_set ( QUEUE_FLAG_NOMERGES , sdev - > request_queue ) ;
2017-02-10 08:59:10 +00:00
blk_queue_virt_boundary ( sdev - > request_queue , mr_nvme_pg_size - 1 ) ;
}
2016-01-28 15:34:26 +00:00
2017-02-10 08:59:10 +00:00
/*
2019-07-26 07:32:14 +00:00
* megasas_set_fw_assisted_qd -
* set device queue depth to can_queue
* set device queue depth to fw assisted qd
2017-02-10 08:59:10 +00:00
*
* @ sdev : scsi device
2017-02-10 08:59:11 +00:00
* @ is_target_prop true , if fw provided target properties .
2017-02-10 08:59:10 +00:00
*/
2019-07-26 07:32:14 +00:00
static void megasas_set_fw_assisted_qd ( struct scsi_device * sdev ,
2017-02-10 08:59:11 +00:00
bool is_target_prop )
2017-02-10 08:59:10 +00:00
{
u8 interface_type ;
u32 device_qd = MEGASAS_DEFAULT_CMD_PER_LUN ;
2017-02-10 08:59:11 +00:00
u32 tgt_device_qd ;
2017-02-10 08:59:10 +00:00
struct megasas_instance * instance ;
struct MR_PRIV_DEVICE * mr_device_priv_data ;
2016-01-28 15:34:26 +00:00
2017-02-10 08:59:10 +00:00
instance = megasas_lookup_instance ( sdev - > host - > host_no ) ;
mr_device_priv_data = sdev - > hostdata ;
interface_type = mr_device_priv_data - > interface_type ;
2016-01-28 15:34:26 +00:00
2017-02-10 08:59:10 +00:00
switch ( interface_type ) {
case SAS_PD :
device_qd = MEGASAS_SAS_QD ;
break ;
case SATA_PD :
device_qd = MEGASAS_SATA_QD ;
break ;
case NVME_PD :
device_qd = MEGASAS_NVME_QD ;
break ;
2016-01-28 15:34:26 +00:00
}
2017-02-10 08:59:10 +00:00
2017-02-10 08:59:11 +00:00
if ( is_target_prop ) {
tgt_device_qd = le32_to_cpu ( instance - > tgt_prop - > device_qdepth ) ;
if ( tgt_device_qd & &
( tgt_device_qd < = instance - > host - > can_queue ) )
device_qd = tgt_device_qd ;
2019-07-26 07:32:14 +00:00
}
2017-02-10 08:59:11 +00:00
2019-07-26 07:32:14 +00:00
if ( instance - > enable_sdev_max_qd & & interface_type ! = UNKNOWN_DRIVE )
device_qd = instance - > host - > can_queue ;
scsi_change_queue_depth ( sdev , device_qd ) ;
}
/*
* megasas_set_static_target_properties -
* Device property set by driver are static and it is not required to be
* updated after OCR .
*
* set io timeout
* set device queue depth
* set nvme device properties . see - megasas_set_nvme_device_properties
*
* @ sdev : scsi device
* @ is_target_prop true , if fw provided target properties .
*/
static void megasas_set_static_target_properties ( struct scsi_device * sdev ,
bool is_target_prop )
{
u32 max_io_size_kb = MR_DEFAULT_NVME_MDTS_KB ;
struct megasas_instance * instance ;
instance = megasas_lookup_instance ( sdev - > host - > host_no ) ;
/*
* The RAID firmware may require extended timeouts .
*/
blk_queue_rq_timeout ( sdev - > request_queue , scmd_timeout * HZ ) ;
/* max_io_size_kb will be set to non zero for
* nvme based vd and syspd .
*/
if ( is_target_prop )
2017-02-10 08:59:11 +00:00
max_io_size_kb = le32_to_cpu ( instance - > tgt_prop - > max_io_size_kb ) ;
2017-02-10 08:59:10 +00:00
if ( instance - > nvme_page_size & & max_io_size_kb )
megasas_set_nvme_device_properties ( sdev , ( max_io_size_kb < < 10 ) ) ;
2019-07-26 07:32:14 +00:00
megasas_set_fw_assisted_qd ( sdev , is_target_prop ) ;
2016-01-28 15:34:26 +00:00
}
2016-01-28 15:34:25 +00:00
2006-02-17 11:13:48 +00:00
static int megasas_slave_configure ( struct scsi_device * sdev )
{
2015-11-05 15:47:37 +00:00
u16 pd_index = 0 ;
struct megasas_instance * instance ;
2017-02-10 08:59:11 +00:00
int ret_target_prop = DCMD_FAILED ;
bool is_target_prop = false ;
2015-11-05 15:47:37 +00:00
instance = megasas_lookup_instance ( sdev - > host - > host_no ) ;
2016-03-10 10:14:37 +00:00
if ( instance - > pd_list_not_supported ) {
2017-02-10 08:59:05 +00:00
if ( ! MEGASAS_IS_LOGICAL ( sdev ) & & sdev - > type = = TYPE_DISK ) {
2015-11-05 15:47:37 +00:00
pd_index = ( sdev - > channel * MEGASAS_MAX_DEV_PER_CHANNEL ) +
sdev - > id ;
if ( instance - > pd_list [ pd_index ] . driveState ! =
MR_PD_STATE_SYSTEM )
return - ENXIO ;
}
}
2016-01-28 15:34:25 +00:00
2018-01-05 13:27:42 +00:00
mutex_lock ( & instance - > reset_mutex ) ;
2017-02-10 08:59:10 +00:00
/* Send DCMD to Firmware and cache the information */
if ( ( instance - > pd_info ) & & ! MEGASAS_IS_LOGICAL ( sdev ) )
megasas_get_pd_info ( instance , sdev ) ;
2017-02-10 08:59:11 +00:00
/* Some ventura firmware may not have instance->nvme_page_size set.
* Do not send MR_DCMD_DRV_GET_TARGET_PROP
*/
if ( ( instance - > tgt_prop ) & & ( instance - > nvme_page_size ) )
ret_target_prop = megasas_get_target_prop ( instance , sdev ) ;
is_target_prop = ( ret_target_prop = = DCMD_SUCCESS ) ? true : false ;
megasas_set_static_target_properties ( sdev , is_target_prop ) ;
2017-02-10 08:59:10 +00:00
/* This sdev property may change post OCR */
2018-06-04 10:45:12 +00:00
megasas_set_dynamic_target_properties ( sdev , is_target_prop ) ;
mutex_unlock ( & instance - > reset_mutex ) ;
2014-09-12 13:27:13 +00:00
2009-10-06 20:33:06 +00:00
return 0 ;
}
static int megasas_slave_alloc ( struct scsi_device * sdev )
{
2015-07-07 20:52:45 +00:00
u16 pd_index = 0 ;
2009-10-06 20:33:06 +00:00
struct megasas_instance * instance ;
2016-01-28 15:34:25 +00:00
struct MR_PRIV_DEVICE * mr_device_priv_data ;
2015-07-07 20:52:45 +00:00
2009-10-06 20:33:06 +00:00
instance = megasas_lookup_instance ( sdev - > host - > host_no ) ;
2017-02-10 08:59:05 +00:00
if ( ! MEGASAS_IS_LOGICAL ( sdev ) ) {
2009-10-06 20:33:06 +00:00
/*
* Open the OS scan to the SYSTEM PD
*/
pd_index =
( sdev - > channel * MEGASAS_MAX_DEV_PER_CHANNEL ) +
sdev - > id ;
2016-03-10 10:14:37 +00:00
if ( ( instance - > pd_list_not_supported | |
instance - > pd_list [ pd_index ] . driveState = =
2015-11-05 15:47:37 +00:00
MR_PD_STATE_SYSTEM ) ) {
2016-01-28 15:34:25 +00:00
goto scan_target ;
2009-10-06 20:33:06 +00:00
}
return - ENXIO ;
}
2016-01-28 15:34:25 +00:00
scan_target :
mr_device_priv_data = kzalloc ( sizeof ( * mr_device_priv_data ) ,
GFP_KERNEL ) ;
if ( ! mr_device_priv_data )
return - ENOMEM ;
sdev - > hostdata = mr_device_priv_data ;
2017-03-10 11:22:13 +00:00
atomic_set ( & mr_device_priv_data - > r1_ldio_hint ,
instance - > r1_ldio_hint_default ) ;
2006-02-17 11:13:48 +00:00
return 0 ;
}
2016-01-28 15:34:25 +00:00
static void megasas_slave_destroy ( struct scsi_device * sdev )
{
kfree ( sdev - > hostdata ) ;
sdev - > hostdata = NULL ;
}
2015-01-05 14:36:18 +00:00
/*
* megasas_complete_outstanding_ioctls - Complete outstanding ioctls after a
* kill adapter
* @ instance : Adapter soft state
*
*/
2015-04-23 11:03:23 +00:00
static void megasas_complete_outstanding_ioctls ( struct megasas_instance * instance )
2015-01-05 14:36:18 +00:00
{
int i ;
struct megasas_cmd * cmd_mfi ;
struct megasas_cmd_fusion * cmd_fusion ;
struct fusion_context * fusion = instance - > ctrl_context ;
/* Find all outstanding ioctls */
if ( fusion ) {
for ( i = 0 ; i < instance - > max_fw_cmds ; i + + ) {
cmd_fusion = fusion - > cmd_list [ i ] ;
if ( cmd_fusion - > sync_cmd_idx ! = ( u32 ) ULONG_MAX ) {
cmd_mfi = instance - > cmd_list [ cmd_fusion - > sync_cmd_idx ] ;
if ( cmd_mfi - > sync_cmd & &
2017-08-23 11:47:04 +00:00
( cmd_mfi - > frame - > hdr . cmd ! = MFI_CMD_ABORT ) ) {
cmd_mfi - > frame - > hdr . cmd_status =
MFI_STAT_WRONG_STATE ;
2015-01-05 14:36:18 +00:00
megasas_complete_cmd ( instance ,
cmd_mfi , DID_OK ) ;
2017-08-23 11:47:04 +00:00
}
2015-01-05 14:36:18 +00:00
}
}
} else {
for ( i = 0 ; i < instance - > max_fw_cmds ; i + + ) {
cmd_mfi = instance - > cmd_list [ i ] ;
if ( cmd_mfi - > sync_cmd & & cmd_mfi - > frame - > hdr . cmd ! =
MFI_CMD_ABORT )
megasas_complete_cmd ( instance , cmd_mfi , DID_OK ) ;
}
}
}
2010-12-21 21:34:31 +00:00
void megaraid_sas_kill_hba ( struct megasas_instance * instance )
2010-09-23 02:36:29 +00:00
{
2015-01-05 14:36:18 +00:00
/* Set critical error to block I/O & ioctls in case caller didn't */
2016-01-28 15:34:32 +00:00
atomic_set ( & instance - > adprecovery , MEGASAS_HW_CRITICAL_ERROR ) ;
2015-01-05 14:36:18 +00:00
/* Wait 1 second to ensure IO or ioctls in build have posted */
msleep ( 1000 ) ;
2010-09-23 02:36:29 +00:00
if ( ( instance - > pdev - > device = = PCI_DEVICE_ID_LSI_SAS0073SKINNY ) | |
2015-01-05 14:36:18 +00:00
( instance - > pdev - > device = = PCI_DEVICE_ID_LSI_SAS0071SKINNY ) | |
2017-10-19 09:48:50 +00:00
( instance - > adapter_type ! = MFI_SERIES ) ) {
2018-10-17 06:37:47 +00:00
if ( ! instance - > requestorId ) {
writel ( MFI_STOP_ADP , & instance - > reg_set - > doorbell ) ;
/* Flush */
readl ( & instance - > reg_set - > doorbell ) ;
}
2016-01-28 15:44:25 +00:00
if ( instance - > requestorId & & instance - > peerIsPresent )
2014-03-10 09:51:56 +00:00
memset ( instance - > ld_ids , 0xff , MEGASAS_MAX_LD_IDS ) ;
2010-09-23 02:36:29 +00:00
} else {
2015-01-05 14:36:18 +00:00
writel ( MFI_STOP_ADP ,
& instance - > reg_set - > inbound_doorbell ) ;
2010-12-21 21:34:31 +00:00
}
2015-01-05 14:36:18 +00:00
/* Complete outstanding ioctls when adapter is killed */
megasas_complete_outstanding_ioctls ( instance ) ;
2010-12-21 21:34:31 +00:00
}
/**
* megasas_check_and_restore_queue_depth - Check if queue depth needs to be
* restored to max value
* @ instance : Adapter soft state
*
*/
void
megasas_check_and_restore_queue_depth ( struct megasas_instance * instance )
{
unsigned long flags ;
2015-01-05 14:36:23 +00:00
2010-12-21 21:34:31 +00:00
if ( instance - > flag & MEGASAS_FW_BUSY
2012-07-18 01:20:03 +00:00
& & time_after ( jiffies , instance - > last_time + 5 * HZ )
& & atomic_read ( & instance - > fw_outstanding ) <
instance - > throttlequeuedepth + 1 ) {
2010-12-21 21:34:31 +00:00
spin_lock_irqsave ( instance - > host - > host_lock , flags ) ;
instance - > flag & = ~ MEGASAS_FW_BUSY ;
2016-01-28 15:34:30 +00:00
instance - > host - > can_queue = instance - > cur_can_queue ;
2010-12-21 21:34:31 +00:00
spin_unlock_irqrestore ( instance - > host - > host_lock , flags ) ;
2010-09-23 02:36:29 +00:00
}
}
2007-11-09 09:35:44 +00:00
/**
* megasas_complete_cmd_dpc - Returns FW ' s controller structure
* @ instance_addr : Address of adapter soft state
*
* Tasklet to complete cmds
*/
static void megasas_complete_cmd_dpc ( unsigned long instance_addr )
{
u32 producer ;
u32 consumer ;
u32 context ;
struct megasas_cmd * cmd ;
struct megasas_instance * instance =
( struct megasas_instance * ) instance_addr ;
unsigned long flags ;
/* If we have already declared adapter dead, donot complete cmds */
2016-01-28 15:34:32 +00:00
if ( atomic_read ( & instance - > adprecovery ) = = MEGASAS_HW_CRITICAL_ERROR )
2007-11-09 09:35:44 +00:00
return ;
spin_lock_irqsave ( & instance - > completion_lock , flags ) ;
2013-09-06 10:20:52 +00:00
producer = le32_to_cpu ( * instance - > producer ) ;
consumer = le32_to_cpu ( * instance - > consumer ) ;
2007-11-09 09:35:44 +00:00
while ( consumer ! = producer ) {
2013-09-06 10:20:52 +00:00
context = le32_to_cpu ( instance - > reply_queue [ consumer ] ) ;
2010-09-23 02:36:29 +00:00
if ( context > = instance - > max_fw_cmds ) {
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " Unexpected context value %x \n " ,
2010-09-23 02:36:29 +00:00
context ) ;
BUG ( ) ;
}
2007-11-09 09:35:44 +00:00
cmd = instance - > cmd_list [ context ] ;
megasas_complete_cmd ( instance , cmd , DID_OK ) ;
consumer + + ;
if ( consumer = = ( instance - > max_fw_cmds + 1 ) ) {
consumer = 0 ;
}
}
2013-09-06 10:20:52 +00:00
* instance - > consumer = cpu_to_le32 ( producer ) ;
2007-11-09 09:35:44 +00:00
spin_unlock_irqrestore ( & instance - > completion_lock , flags ) ;
/*
* Check if we can restore can_queue
*/
2010-12-21 21:34:31 +00:00
megasas_check_and_restore_queue_depth ( instance ) ;
2007-11-09 09:35:44 +00:00
}
2017-10-22 22:30:04 +00:00
static void megasas_sriov_heartbeat_handler ( struct timer_list * t ) ;
2014-03-10 09:51:56 +00:00
/**
2017-10-22 22:30:04 +00:00
* megasas_start_timer - Initializes sriov heartbeat timer object
2014-03-10 09:51:56 +00:00
* @ instance : Adapter soft state
*
*/
2017-10-22 22:30:04 +00:00
void megasas_start_timer ( struct megasas_instance * instance )
2014-03-10 09:51:56 +00:00
{
2017-10-22 22:30:04 +00:00
struct timer_list * timer = & instance - > sriov_heartbeat_timer ;
timer_setup ( timer , megasas_sriov_heartbeat_handler , 0 ) ;
timer - > expires = jiffies + MEGASAS_SRIOV_HEARTBEAT_INTERVAL_VF ;
2014-03-10 09:51:56 +00:00
add_timer ( timer ) ;
}
2010-10-12 13:20:27 +00:00
static void
megasas_internal_reset_defer_cmds ( struct megasas_instance * instance ) ;
static void
process_fw_state_change_wq ( struct work_struct * work ) ;
2019-07-02 13:01:14 +00:00
static void megasas_do_ocr ( struct megasas_instance * instance )
2010-10-12 13:20:27 +00:00
{
if ( ( instance - > pdev - > device = = PCI_DEVICE_ID_LSI_SAS1064R ) | |
( instance - > pdev - > device = = PCI_DEVICE_ID_DELL_PERC5 ) | |
( instance - > pdev - > device = = PCI_DEVICE_ID_LSI_VERDE_ZCR ) ) {
2013-09-06 10:20:52 +00:00
* instance - > consumer = cpu_to_le32 ( MEGASAS_ADPRESET_INPROG_SIGN ) ;
2010-10-12 13:20:27 +00:00
}
2013-05-22 07:04:14 +00:00
instance - > instancet - > disable_intr ( instance ) ;
2016-01-28 15:34:32 +00:00
atomic_set ( & instance - > adprecovery , MEGASAS_ADPRESET_SM_INFAULT ) ;
2010-10-12 13:20:27 +00:00
instance - > issuepend_done = 0 ;
atomic_set ( & instance - > fw_outstanding , 0 ) ;
megasas_internal_reset_defer_cmds ( instance ) ;
process_fw_state_change_wq ( & instance - > work_init ) ;
}
2014-07-09 22:17:56 +00:00
static int megasas_get_ld_vf_affiliation_111 ( struct megasas_instance * instance ,
int initial )
2014-03-10 09:51:56 +00:00
{
struct megasas_cmd * cmd ;
struct megasas_dcmd_frame * dcmd ;
struct MR_LD_VF_AFFILIATION_111 * new_affiliation_111 = NULL ;
dma_addr_t new_affiliation_111_h ;
int ld , retval = 0 ;
u8 thisVf ;
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " megasas_get_ld_vf_affiliation_111: "
" Failed to get cmd for scsi%d \n " ,
2014-03-10 09:51:56 +00:00
instance - > host - > host_no ) ;
return - ENOMEM ;
}
dcmd = & cmd - > frame - > dcmd ;
2014-07-09 22:17:56 +00:00
if ( ! instance - > vf_affiliation_111 ) {
2015-07-07 20:52:34 +00:00
dev_warn ( & instance - > pdev - > dev , " SR-IOV: Couldn't get LD/VF "
" affiliation for scsi%d \n " , instance - > host - > host_no ) ;
2014-03-10 09:51:56 +00:00
megasas_return_cmd ( instance , cmd ) ;
return - ENOMEM ;
}
if ( initial )
memset ( instance - > vf_affiliation_111 , 0 ,
sizeof ( struct MR_LD_VF_AFFILIATION_111 ) ) ;
else {
2014-07-09 22:17:56 +00:00
new_affiliation_111 =
cross-tree: phase out dma_zalloc_coherent()
We already need to zero out memory for dma_alloc_coherent(), as such
using dma_zalloc_coherent() is superflous. Phase it out.
This change was generated with the following Coccinelle SmPL patch:
@ replace_dma_zalloc_coherent @
expression dev, size, data, handle, flags;
@@
-dma_zalloc_coherent(dev, size, handle, flags)
+dma_alloc_coherent(dev, size, handle, flags)
Suggested-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
[hch: re-ran the script on the latest tree]
Signed-off-by: Christoph Hellwig <hch@lst.de>
2019-01-04 08:23:09 +00:00
dma_alloc_coherent ( & instance - > pdev - > dev ,
sizeof ( struct MR_LD_VF_AFFILIATION_111 ) ,
& new_affiliation_111_h , GFP_KERNEL ) ;
2014-07-09 22:17:56 +00:00
if ( ! new_affiliation_111 ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " SR-IOV: Couldn't allocate "
" memory for new affiliation for scsi%d \n " ,
2014-07-09 22:17:56 +00:00
instance - > host - > host_no ) ;
2014-03-10 09:51:56 +00:00
megasas_return_cmd ( instance , cmd ) ;
return - ENOMEM ;
}
}
memset ( dcmd - > mbox . b , 0 , MFI_MBOX_SIZE ) ;
dcmd - > cmd = MFI_CMD_DCMD ;
2015-05-06 13:31:02 +00:00
dcmd - > cmd_status = MFI_STAT_INVALID_STATUS ;
2014-03-10 09:51:56 +00:00
dcmd - > sge_count = 1 ;
2015-04-23 11:03:54 +00:00
dcmd - > flags = cpu_to_le16 ( MFI_FRAME_DIR_BOTH ) ;
2014-03-10 09:51:56 +00:00
dcmd - > timeout = 0 ;
dcmd - > pad_0 = 0 ;
2015-04-23 11:03:54 +00:00
dcmd - > data_xfer_len =
cpu_to_le32 ( sizeof ( struct MR_LD_VF_AFFILIATION_111 ) ) ;
dcmd - > opcode = cpu_to_le32 ( MR_DCMD_LD_VF_MAP_GET_ALL_LDS_111 ) ;
2014-03-10 09:51:56 +00:00
2014-07-09 22:17:56 +00:00
if ( initial )
dcmd - > sgl . sge32 [ 0 ] . phys_addr =
2015-04-23 11:03:54 +00:00
cpu_to_le32 ( instance - > vf_affiliation_111_h ) ;
2014-03-10 09:51:56 +00:00
else
2015-04-23 11:03:54 +00:00
dcmd - > sgl . sge32 [ 0 ] . phys_addr =
cpu_to_le32 ( new_affiliation_111_h ) ;
2014-07-09 22:17:56 +00:00
2015-04-23 11:03:54 +00:00
dcmd - > sgl . sge32 [ 0 ] . length = cpu_to_le32 (
sizeof ( struct MR_LD_VF_AFFILIATION_111 ) ) ;
2014-03-10 09:51:56 +00:00
2015-07-07 20:52:34 +00:00
dev_warn ( & instance - > pdev - > dev , " SR-IOV: Getting LD/VF affiliation for "
2014-03-10 09:51:56 +00:00
" scsi%d \n " , instance - > host - > host_no ) ;
2016-01-28 15:34:23 +00:00
if ( megasas_issue_blocked_cmd ( instance , cmd , 0 ) ! = DCMD_SUCCESS ) {
2015-07-07 20:52:34 +00:00
dev_warn ( & instance - > pdev - > dev , " SR-IOV: LD/VF affiliation DCMD "
" failed with status 0x%x for scsi%d \n " ,
2014-03-10 09:51:56 +00:00
dcmd - > cmd_status , instance - > host - > host_no ) ;
retval = 1 ; /* Do a scan if we couldn't get affiliation */
goto out ;
}
if ( ! initial ) {
2014-07-09 22:17:56 +00:00
thisVf = new_affiliation_111 - > thisVf ;
for ( ld = 0 ; ld < new_affiliation_111 - > vdCount ; ld + + )
if ( instance - > vf_affiliation_111 - > map [ ld ] . policy [ thisVf ] ! =
new_affiliation_111 - > map [ ld ] . policy [ thisVf ] ) {
2015-07-07 20:52:34 +00:00
dev_warn ( & instance - > pdev - > dev , " SR-IOV: "
" Got new LD/VF affiliation for scsi%d \n " ,
2014-03-10 09:51:56 +00:00
instance - > host - > host_no ) ;
2014-07-09 22:17:56 +00:00
memcpy ( instance - > vf_affiliation_111 ,
new_affiliation_111 ,
sizeof ( struct MR_LD_VF_AFFILIATION_111 ) ) ;
2014-03-10 09:51:56 +00:00
retval = 1 ;
goto out ;
}
2014-07-09 22:17:56 +00:00
}
out :
if ( new_affiliation_111 ) {
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & instance - > pdev - > dev ,
2014-07-09 22:17:56 +00:00
sizeof ( struct MR_LD_VF_AFFILIATION_111 ) ,
new_affiliation_111 ,
new_affiliation_111_h ) ;
}
2014-09-12 13:27:58 +00:00
2015-04-23 11:01:24 +00:00
megasas_return_cmd ( instance , cmd ) ;
2014-07-09 22:17:56 +00:00
return retval ;
}
static int megasas_get_ld_vf_affiliation_12 ( struct megasas_instance * instance ,
int initial )
{
struct megasas_cmd * cmd ;
struct megasas_dcmd_frame * dcmd ;
struct MR_LD_VF_AFFILIATION * new_affiliation = NULL ;
struct MR_LD_VF_MAP * newmap = NULL , * savedmap = NULL ;
dma_addr_t new_affiliation_h ;
int i , j , retval = 0 , found = 0 , doscan = 0 ;
u8 thisVf ;
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " megasas_get_ld_vf_affiliation12: "
" Failed to get cmd for scsi%d \n " ,
2014-07-09 22:17:56 +00:00
instance - > host - > host_no ) ;
return - ENOMEM ;
}
dcmd = & cmd - > frame - > dcmd ;
if ( ! instance - > vf_affiliation ) {
2015-07-07 20:52:34 +00:00
dev_warn ( & instance - > pdev - > dev , " SR-IOV: Couldn't get LD/VF "
" affiliation for scsi%d \n " , instance - > host - > host_no ) ;
2014-07-09 22:17:56 +00:00
megasas_return_cmd ( instance , cmd ) ;
return - ENOMEM ;
}
if ( initial )
memset ( instance - > vf_affiliation , 0 , ( MAX_LOGICAL_DRIVES + 1 ) *
sizeof ( struct MR_LD_VF_AFFILIATION ) ) ;
else {
new_affiliation =
cross-tree: phase out dma_zalloc_coherent()
We already need to zero out memory for dma_alloc_coherent(), as such
using dma_zalloc_coherent() is superflous. Phase it out.
This change was generated with the following Coccinelle SmPL patch:
@ replace_dma_zalloc_coherent @
expression dev, size, data, handle, flags;
@@
-dma_zalloc_coherent(dev, size, handle, flags)
+dma_alloc_coherent(dev, size, handle, flags)
Suggested-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
[hch: re-ran the script on the latest tree]
Signed-off-by: Christoph Hellwig <hch@lst.de>
2019-01-04 08:23:09 +00:00
dma_alloc_coherent ( & instance - > pdev - > dev ,
( MAX_LOGICAL_DRIVES + 1 ) * sizeof ( struct MR_LD_VF_AFFILIATION ) ,
& new_affiliation_h , GFP_KERNEL ) ;
2014-07-09 22:17:56 +00:00
if ( ! new_affiliation ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " SR-IOV: Couldn't allocate "
" memory for new affiliation for scsi%d \n " ,
2014-07-09 22:17:56 +00:00
instance - > host - > host_no ) ;
megasas_return_cmd ( instance , cmd ) ;
return - ENOMEM ;
}
}
memset ( dcmd - > mbox . b , 0 , MFI_MBOX_SIZE ) ;
dcmd - > cmd = MFI_CMD_DCMD ;
2015-05-06 13:31:02 +00:00
dcmd - > cmd_status = MFI_STAT_INVALID_STATUS ;
2014-07-09 22:17:56 +00:00
dcmd - > sge_count = 1 ;
2015-04-23 11:03:54 +00:00
dcmd - > flags = cpu_to_le16 ( MFI_FRAME_DIR_BOTH ) ;
2014-07-09 22:17:56 +00:00
dcmd - > timeout = 0 ;
dcmd - > pad_0 = 0 ;
2015-04-23 11:03:54 +00:00
dcmd - > data_xfer_len = cpu_to_le32 ( ( MAX_LOGICAL_DRIVES + 1 ) *
sizeof ( struct MR_LD_VF_AFFILIATION ) ) ;
dcmd - > opcode = cpu_to_le32 ( MR_DCMD_LD_VF_MAP_GET_ALL_LDS ) ;
2014-07-09 22:17:56 +00:00
if ( initial )
2015-04-23 11:03:54 +00:00
dcmd - > sgl . sge32 [ 0 ] . phys_addr =
cpu_to_le32 ( instance - > vf_affiliation_h ) ;
2014-07-09 22:17:56 +00:00
else
2015-04-23 11:03:54 +00:00
dcmd - > sgl . sge32 [ 0 ] . phys_addr =
cpu_to_le32 ( new_affiliation_h ) ;
2014-07-09 22:17:56 +00:00
2015-04-23 11:03:54 +00:00
dcmd - > sgl . sge32 [ 0 ] . length = cpu_to_le32 ( ( MAX_LOGICAL_DRIVES + 1 ) *
sizeof ( struct MR_LD_VF_AFFILIATION ) ) ;
2014-07-09 22:17:56 +00:00
2015-07-07 20:52:34 +00:00
dev_warn ( & instance - > pdev - > dev , " SR-IOV: Getting LD/VF affiliation for "
2014-07-09 22:17:56 +00:00
" scsi%d \n " , instance - > host - > host_no ) ;
2016-01-28 15:34:23 +00:00
if ( megasas_issue_blocked_cmd ( instance , cmd , 0 ) ! = DCMD_SUCCESS ) {
2015-07-07 20:52:34 +00:00
dev_warn ( & instance - > pdev - > dev , " SR-IOV: LD/VF affiliation DCMD "
" failed with status 0x%x for scsi%d \n " ,
2014-07-09 22:17:56 +00:00
dcmd - > cmd_status , instance - > host - > host_no ) ;
retval = 1 ; /* Do a scan if we couldn't get affiliation */
goto out ;
}
if ( ! initial ) {
if ( ! new_affiliation - > ldCount ) {
2015-07-07 20:52:34 +00:00
dev_warn ( & instance - > pdev - > dev , " SR-IOV: Got new LD/VF "
" affiliation for passive path for scsi%d \n " ,
2014-07-09 22:17:56 +00:00
instance - > host - > host_no ) ;
retval = 1 ;
goto out ;
}
newmap = new_affiliation - > map ;
savedmap = instance - > vf_affiliation - > map ;
thisVf = new_affiliation - > thisVf ;
for ( i = 0 ; i < new_affiliation - > ldCount ; i + + ) {
found = 0 ;
for ( j = 0 ; j < instance - > vf_affiliation - > ldCount ;
j + + ) {
if ( newmap - > ref . targetId = =
savedmap - > ref . targetId ) {
found = 1 ;
if ( newmap - > policy [ thisVf ] ! =
savedmap - > policy [ thisVf ] ) {
doscan = 1 ;
goto out ;
}
2014-03-10 09:51:56 +00:00
}
savedmap = ( struct MR_LD_VF_MAP * )
( ( unsigned char * ) savedmap +
savedmap - > size ) ;
2014-07-09 22:17:56 +00:00
}
if ( ! found & & newmap - > policy [ thisVf ] ! =
MR_LD_ACCESS_HIDDEN ) {
doscan = 1 ;
goto out ;
}
newmap = ( struct MR_LD_VF_MAP * )
( ( unsigned char * ) newmap + newmap - > size ) ;
}
newmap = new_affiliation - > map ;
savedmap = instance - > vf_affiliation - > map ;
for ( i = 0 ; i < instance - > vf_affiliation - > ldCount ; i + + ) {
found = 0 ;
for ( j = 0 ; j < new_affiliation - > ldCount ; j + + ) {
if ( savedmap - > ref . targetId = =
newmap - > ref . targetId ) {
found = 1 ;
if ( savedmap - > policy [ thisVf ] ! =
newmap - > policy [ thisVf ] ) {
doscan = 1 ;
goto out ;
}
}
2014-03-10 09:51:56 +00:00
newmap = ( struct MR_LD_VF_MAP * )
( ( unsigned char * ) newmap +
newmap - > size ) ;
}
2014-07-09 22:17:56 +00:00
if ( ! found & & savedmap - > policy [ thisVf ] ! =
MR_LD_ACCESS_HIDDEN ) {
doscan = 1 ;
goto out ;
}
savedmap = ( struct MR_LD_VF_MAP * )
( ( unsigned char * ) savedmap +
savedmap - > size ) ;
2014-03-10 09:51:56 +00:00
}
}
out :
2014-07-09 22:17:56 +00:00
if ( doscan ) {
2015-07-07 20:52:34 +00:00
dev_warn ( & instance - > pdev - > dev , " SR-IOV: Got new LD/VF "
" affiliation for scsi%d \n " , instance - > host - > host_no ) ;
2014-07-09 22:17:56 +00:00
memcpy ( instance - > vf_affiliation , new_affiliation ,
new_affiliation - > size ) ;
retval = 1 ;
2014-03-10 09:51:56 +00:00
}
2014-07-09 22:17:56 +00:00
if ( new_affiliation )
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & instance - > pdev - > dev ,
2014-07-09 22:17:56 +00:00
( MAX_LOGICAL_DRIVES + 1 ) *
sizeof ( struct MR_LD_VF_AFFILIATION ) ,
new_affiliation , new_affiliation_h ) ;
2015-04-23 11:01:24 +00:00
megasas_return_cmd ( instance , cmd ) ;
2014-03-10 09:51:56 +00:00
return retval ;
}
2014-07-09 22:17:56 +00:00
/* This function will get the current SR-IOV LD/VF affiliation */
static int megasas_get_ld_vf_affiliation ( struct megasas_instance * instance ,
int initial )
{
int retval ;
if ( instance - > PlasmaFW111 )
retval = megasas_get_ld_vf_affiliation_111 ( instance , initial ) ;
else
retval = megasas_get_ld_vf_affiliation_12 ( instance , initial ) ;
return retval ;
}
2014-03-10 09:51:56 +00:00
/* This function will tell FW to start the SR-IOV heartbeat */
int megasas_sriov_start_heartbeat ( struct megasas_instance * instance ,
int initial )
{
struct megasas_cmd * cmd ;
struct megasas_dcmd_frame * dcmd ;
int retval = 0 ;
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " megasas_sriov_start_heartbeat: "
" Failed to get cmd for scsi%d \n " ,
2014-03-10 09:51:56 +00:00
instance - > host - > host_no ) ;
return - ENOMEM ;
}
dcmd = & cmd - > frame - > dcmd ;
if ( initial ) {
instance - > hb_host_mem =
cross-tree: phase out dma_zalloc_coherent()
We already need to zero out memory for dma_alloc_coherent(), as such
using dma_zalloc_coherent() is superflous. Phase it out.
This change was generated with the following Coccinelle SmPL patch:
@ replace_dma_zalloc_coherent @
expression dev, size, data, handle, flags;
@@
-dma_zalloc_coherent(dev, size, handle, flags)
+dma_alloc_coherent(dev, size, handle, flags)
Suggested-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
[hch: re-ran the script on the latest tree]
Signed-off-by: Christoph Hellwig <hch@lst.de>
2019-01-04 08:23:09 +00:00
dma_alloc_coherent ( & instance - > pdev - > dev ,
sizeof ( struct MR_CTRL_HB_HOST_MEM ) ,
& instance - > hb_host_mem_h ,
GFP_KERNEL ) ;
2014-03-10 09:51:56 +00:00
if ( ! instance - > hb_host_mem ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " SR-IOV: Couldn't allocate "
" memory for heartbeat host memory for scsi%d \n " ,
instance - > host - > host_no ) ;
2014-03-10 09:51:56 +00:00
retval = - ENOMEM ;
goto out ;
}
}
memset ( dcmd - > mbox . b , 0 , MFI_MBOX_SIZE ) ;
2015-04-23 11:03:54 +00:00
dcmd - > mbox . s [ 0 ] = cpu_to_le16 ( sizeof ( struct MR_CTRL_HB_HOST_MEM ) ) ;
2014-03-10 09:51:56 +00:00
dcmd - > cmd = MFI_CMD_DCMD ;
2015-05-06 13:31:02 +00:00
dcmd - > cmd_status = MFI_STAT_INVALID_STATUS ;
2014-03-10 09:51:56 +00:00
dcmd - > sge_count = 1 ;
2015-04-23 11:03:54 +00:00
dcmd - > flags = cpu_to_le16 ( MFI_FRAME_DIR_BOTH ) ;
2014-03-10 09:51:56 +00:00
dcmd - > timeout = 0 ;
dcmd - > pad_0 = 0 ;
2015-04-23 11:03:54 +00:00
dcmd - > data_xfer_len = cpu_to_le32 ( sizeof ( struct MR_CTRL_HB_HOST_MEM ) ) ;
dcmd - > opcode = cpu_to_le32 ( MR_DCMD_CTRL_SHARED_HOST_MEM_ALLOC ) ;
2017-10-19 09:49:05 +00:00
megasas_set_dma_settings ( instance , dcmd , instance - > hb_host_mem_h ,
sizeof ( struct MR_CTRL_HB_HOST_MEM ) ) ;
2014-03-10 09:51:56 +00:00
2015-07-07 20:52:34 +00:00
dev_warn ( & instance - > pdev - > dev , " SR-IOV: Starting heartbeat for scsi%d \n " ,
2014-03-10 09:51:56 +00:00
instance - > host - > host_no ) ;
2017-10-19 09:48:50 +00:00
if ( ( instance - > adapter_type ! = MFI_SERIES ) & &
! instance - > mask_interrupts )
2015-04-23 11:01:24 +00:00
retval = megasas_issue_blocked_cmd ( instance , cmd ,
MEGASAS_ROUTINE_WAIT_TIME_VF ) ;
else
retval = megasas_issue_polled ( instance , cmd ) ;
2014-03-10 09:51:56 +00:00
2015-04-23 11:01:24 +00:00
if ( retval ) {
2015-05-06 13:31:02 +00:00
dev_warn ( & instance - > pdev - > dev , " SR-IOV: MR_DCMD_CTRL_SHARED_HOST "
" _MEM_ALLOC DCMD %s for scsi%d \n " ,
( dcmd - > cmd_status = = MFI_STAT_INVALID_STATUS ) ?
" timed out " : " failed " , instance - > host - > host_no ) ;
2014-03-10 09:51:56 +00:00
retval = 1 ;
}
out :
megasas_return_cmd ( instance , cmd ) ;
return retval ;
}
/* Handler for SR-IOV heartbeat */
2017-10-22 22:30:04 +00:00
static void megasas_sriov_heartbeat_handler ( struct timer_list * t )
2014-03-10 09:51:56 +00:00
{
struct megasas_instance * instance =
2017-10-22 22:30:04 +00:00
from_timer ( instance , t , sriov_heartbeat_timer ) ;
2014-03-10 09:51:56 +00:00
if ( instance - > hb_host_mem - > HB . fwCounter ! =
instance - > hb_host_mem - > HB . driverCounter ) {
instance - > hb_host_mem - > HB . driverCounter =
instance - > hb_host_mem - > HB . fwCounter ;
mod_timer ( & instance - > sriov_heartbeat_timer ,
jiffies + MEGASAS_SRIOV_HEARTBEAT_INTERVAL_VF ) ;
} else {
2015-07-07 20:52:34 +00:00
dev_warn ( & instance - > pdev - > dev , " SR-IOV: Heartbeat never "
2014-03-10 09:51:56 +00:00
" completed for scsi%d \n " , instance - > host - > host_no ) ;
schedule_work ( & instance - > work_init ) ;
}
}
2005-09-20 21:46:58 +00:00
/**
* megasas_wait_for_outstanding - Wait for all outstanding cmds
* @ instance : Adapter soft state
*
2011-03-31 01:57:33 +00:00
* This function waits for up to MEGASAS_RESET_WAIT_TIME seconds for FW to
2005-09-20 21:46:58 +00:00
* complete all its outstanding commands . Returns error if one or more IOs
* are pending after this time period . It also marks the controller dead .
*/
static int megasas_wait_for_outstanding ( struct megasas_instance * instance )
{
2016-01-28 15:34:33 +00:00
int i , sl , outstanding ;
2010-09-23 02:36:29 +00:00
u32 reset_index ;
2005-09-20 21:46:58 +00:00
u32 wait_time = MEGASAS_RESET_WAIT_TIME ;
2010-09-23 02:36:29 +00:00
unsigned long flags ;
struct list_head clist_local ;
struct megasas_cmd * reset_cmd ;
2010-10-12 13:20:27 +00:00
u32 fw_state ;
2010-09-23 02:36:29 +00:00
2016-01-28 15:34:33 +00:00
if ( atomic_read ( & instance - > adprecovery ) = = MEGASAS_HW_CRITICAL_ERROR ) {
dev_info ( & instance - > pdev - > dev , " %s:%d HBA is killed. \n " ,
__func__ , __LINE__ ) ;
return FAILED ;
}
2010-09-23 02:36:29 +00:00
2016-01-28 15:34:32 +00:00
if ( atomic_read ( & instance - > adprecovery ) ! = MEGASAS_HBA_OPERATIONAL ) {
2010-09-23 02:36:29 +00:00
INIT_LIST_HEAD ( & clist_local ) ;
spin_lock_irqsave ( & instance - > hba_lock , flags ) ;
list_splice_init ( & instance - > internal_reset_pending_q ,
& clist_local ) ;
spin_unlock_irqrestore ( & instance - > hba_lock , flags ) ;
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " HBA reset wait ... \n " ) ;
2010-09-23 02:36:29 +00:00
for ( i = 0 ; i < wait_time ; i + + ) {
msleep ( 1000 ) ;
2016-01-28 15:34:32 +00:00
if ( atomic_read ( & instance - > adprecovery ) = = MEGASAS_HBA_OPERATIONAL )
2010-09-23 02:36:29 +00:00
break ;
}
2016-01-28 15:34:32 +00:00
if ( atomic_read ( & instance - > adprecovery ) ! = MEGASAS_HBA_OPERATIONAL ) {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " reset: Stopping HBA. \n " ) ;
2016-01-28 15:34:32 +00:00
atomic_set ( & instance - > adprecovery , MEGASAS_HW_CRITICAL_ERROR ) ;
2010-09-23 02:36:29 +00:00
return FAILED ;
}
2015-07-07 20:52:45 +00:00
reset_index = 0 ;
2010-09-23 02:36:29 +00:00
while ( ! list_empty ( & clist_local ) ) {
2015-07-07 20:52:45 +00:00
reset_cmd = list_entry ( ( & clist_local ) - > next ,
2010-09-23 02:36:29 +00:00
struct megasas_cmd , list ) ;
list_del_init ( & reset_cmd - > list ) ;
if ( reset_cmd - > scmd ) {
2017-02-10 08:59:07 +00:00
reset_cmd - > scmd - > result = DID_REQUEUE < < 16 ;
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " %d:%p reset [%02x] \n " ,
2010-09-23 02:36:29 +00:00
reset_index , reset_cmd ,
2011-04-04 13:42:14 +00:00
reset_cmd - > scmd - > cmnd [ 0 ] ) ;
2010-09-23 02:36:29 +00:00
reset_cmd - > scmd - > scsi_done ( reset_cmd - > scmd ) ;
megasas_return_cmd ( instance , reset_cmd ) ;
} else if ( reset_cmd - > sync_cmd ) {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " %p synch cmds "
2010-09-23 02:36:29 +00:00
" reset queue \n " ,
reset_cmd ) ;
2015-05-06 13:31:02 +00:00
reset_cmd - > cmd_status_drv = MFI_STAT_INVALID_STATUS ;
2010-09-23 02:36:29 +00:00
instance - > instancet - > fire_cmd ( instance ,
reset_cmd - > frame_phys_addr ,
0 , instance - > reg_set ) ;
} else {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " %p unexpected "
2010-09-23 02:36:29 +00:00
" cmds lst \n " ,
reset_cmd ) ;
}
reset_index + + ;
}
return SUCCESS ;
}
2005-09-20 21:46:58 +00:00
2012-07-18 01:20:24 +00:00
for ( i = 0 ; i < resetwaittime ; i + + ) {
2016-01-28 15:34:33 +00:00
outstanding = atomic_read ( & instance - > fw_outstanding ) ;
2006-05-30 19:03:37 +00:00
if ( ! outstanding )
2005-09-20 21:46:58 +00:00
break ;
if ( ! ( i % MEGASAS_RESET_NOTICE_INTERVAL ) ) {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " [%2d]waiting for %d "
2006-05-30 19:03:37 +00:00
" commands to complete \n " , i , outstanding ) ;
2007-11-09 09:35:44 +00:00
/*
* Call cmd completion routine . Cmd to be
* be completed directly without depending on isr .
*/
megasas_complete_cmd_dpc ( ( unsigned long ) instance ) ;
2005-09-20 21:46:58 +00:00
}
msleep ( 1000 ) ;
}
2010-10-12 13:20:27 +00:00
i = 0 ;
2016-01-28 15:34:33 +00:00
outstanding = atomic_read ( & instance - > fw_outstanding ) ;
2018-12-17 08:47:39 +00:00
fw_state = instance - > instancet - > read_fw_status_reg ( instance ) & MFI_STATE_MASK ;
2016-01-28 15:34:33 +00:00
if ( ( ! outstanding & & ( fw_state = = MFI_STATE_OPERATIONAL ) ) )
goto no_outstanding ;
if ( instance - > disableOnlineCtrlReset )
goto kill_hba_and_failed ;
2010-10-12 13:20:27 +00:00
do {
2016-01-28 15:34:33 +00:00
if ( ( fw_state = = MFI_STATE_FAULT ) | | atomic_read ( & instance - > fw_outstanding ) ) {
dev_info ( & instance - > pdev - > dev ,
2019-04-17 13:51:09 +00:00
" %s:%d waiting_for_outstanding: before issue OCR. FW state = 0x%x, outstanding 0x%x \n " ,
2016-01-28 15:34:33 +00:00
__func__ , __LINE__ , fw_state , atomic_read ( & instance - > fw_outstanding ) ) ;
if ( i = = 3 )
goto kill_hba_and_failed ;
2010-10-12 13:20:27 +00:00
megasas_do_ocr ( instance ) ;
2016-01-28 15:34:33 +00:00
if ( atomic_read ( & instance - > adprecovery ) = = MEGASAS_HW_CRITICAL_ERROR ) {
dev_info ( & instance - > pdev - > dev , " %s:%d OCR failed and HBA is killed. \n " ,
__func__ , __LINE__ ) ;
return FAILED ;
}
dev_info ( & instance - > pdev - > dev , " %s:%d waiting_for_outstanding: after issue OCR. \n " ,
__func__ , __LINE__ ) ;
for ( sl = 0 ; sl < 10 ; sl + + )
msleep ( 500 ) ;
outstanding = atomic_read ( & instance - > fw_outstanding ) ;
2018-12-17 08:47:39 +00:00
fw_state = instance - > instancet - > read_fw_status_reg ( instance ) & MFI_STATE_MASK ;
2016-01-28 15:34:33 +00:00
if ( ( ! outstanding & & ( fw_state = = MFI_STATE_OPERATIONAL ) ) )
goto no_outstanding ;
2010-10-12 13:20:27 +00:00
}
i + + ;
} while ( i < = 3 ) ;
2016-01-28 15:34:33 +00:00
no_outstanding :
2010-10-12 13:20:27 +00:00
2016-01-28 15:34:33 +00:00
dev_info ( & instance - > pdev - > dev , " %s:%d no more pending commands remain after reset handling. \n " ,
__func__ , __LINE__ ) ;
return SUCCESS ;
2010-10-12 13:20:27 +00:00
2016-01-28 15:34:33 +00:00
kill_hba_and_failed :
2005-09-20 21:46:58 +00:00
2016-01-28 15:34:33 +00:00
/* Reset not supported, kill adapter */
dev_info ( & instance - > pdev - > dev , " %s:%d killing adapter scsi%d "
" disableOnlineCtrlReset %d fw_outstanding %d \n " ,
__func__ , __LINE__ , instance - > host - > host_no , instance - > disableOnlineCtrlReset ,
atomic_read ( & instance - > fw_outstanding ) ) ;
megasas_dump_pending_frames ( instance ) ;
megaraid_sas_kill_hba ( instance ) ;
2010-09-23 02:36:29 +00:00
2016-01-28 15:34:33 +00:00
return FAILED ;
2005-09-20 21:46:58 +00:00
}
/**
* megasas_generic_reset - Generic reset routine
* @ scmd : Mid - layer SCSI command
*
* This routine implements a generic reset handler for device , bus and host
* reset requests . Device , bus and host specific reset handlers can use this
* function after they do their specific tasks .
*/
static int megasas_generic_reset ( struct scsi_cmnd * scmd )
{
int ret_val ;
struct megasas_instance * instance ;
instance = ( struct megasas_instance * ) scmd - > device - > host - > hostdata ;
2011-04-04 13:42:14 +00:00
scmd_printk ( KERN_NOTICE , scmd , " megasas: RESET cmd=%x retries=%x \n " ,
scmd - > cmnd [ 0 ] , scmd - > retries ) ;
2005-09-20 21:46:58 +00:00
2016-01-28 15:34:32 +00:00
if ( atomic_read ( & instance - > adprecovery ) = = MEGASAS_HW_CRITICAL_ERROR ) {
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " cannot recover from previous reset failures \n " ) ;
2005-09-20 21:46:58 +00:00
return FAILED ;
}
ret_val = megasas_wait_for_outstanding ( instance ) ;
if ( ret_val = = SUCCESS )
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " reset successful \n " ) ;
2005-09-20 21:46:58 +00:00
else
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " failed to do reset \n " ) ;
2005-09-20 21:46:58 +00:00
return ret_val ;
}
2007-05-17 12:47:51 +00:00
/**
* megasas_reset_timer - quiesce the adapter if required
* @ scmd : scsi cmnd
*
* Sets the FW busy flag and reduces the host - > can_queue if the
* cmd has not been completed within the timeout period .
*/
static enum
2008-09-14 12:55:09 +00:00
blk_eh_timer_return megasas_reset_timer ( struct scsi_cmnd * scmd )
2007-05-17 12:47:51 +00:00
{
struct megasas_instance * instance ;
unsigned long flags ;
if ( time_after ( jiffies , scmd - > jiffies_at_alloc +
2016-01-28 15:34:34 +00:00
( scmd_timeout * 2 ) * HZ ) ) {
2018-05-29 13:52:29 +00:00
return BLK_EH_DONE ;
2007-05-17 12:47:51 +00:00
}
2011-10-13 23:01:12 +00:00
instance = ( struct megasas_instance * ) scmd - > device - > host - > hostdata ;
2007-05-17 12:47:51 +00:00
if ( ! ( instance - > flag & MEGASAS_FW_BUSY ) ) {
/* FW is busy, throttle IO */
spin_lock_irqsave ( instance - > host - > host_lock , flags ) ;
2012-07-18 01:20:03 +00:00
instance - > host - > can_queue = instance - > throttlequeuedepth ;
2007-05-17 12:47:51 +00:00
instance - > last_time = jiffies ;
instance - > flag | = MEGASAS_FW_BUSY ;
spin_unlock_irqrestore ( instance - > host - > host_lock , flags ) ;
}
2008-09-14 12:55:09 +00:00
return BLK_EH_RESET_TIMER ;
2007-05-17 12:47:51 +00:00
}
2017-02-10 08:59:15 +00:00
/**
2019-05-07 17:05:39 +00:00
* megasas_dump - This function will print hexdump of provided buffer .
* @ buf : Buffer to be dumped
* @ sz : Size in bytes
* @ format : Different formats of dumping e . g . format = n will
* cause only ' n ' 32 bit words to be dumped in a single
* line .
2017-02-10 08:59:15 +00:00
*/
2019-05-07 17:05:37 +00:00
inline void
2019-05-07 17:05:39 +00:00
megasas_dump ( void * buf , int sz , int format )
2017-02-10 08:59:15 +00:00
{
int i ;
2019-05-07 17:05:39 +00:00
__le32 * buf_loc = ( __le32 * ) buf ;
2017-02-10 08:59:15 +00:00
2019-05-07 17:05:39 +00:00
for ( i = 0 ; i < ( sz / sizeof ( __le32 ) ) ; i + + ) {
if ( ( i % format ) = = 0 ) {
if ( i ! = 0 )
printk ( KERN_CONT " \n " ) ;
printk ( KERN_CONT " %08x: " , ( i * 4 ) ) ;
}
printk ( KERN_CONT " %08x " , le32_to_cpu ( buf_loc [ i ] ) ) ;
2017-02-10 08:59:15 +00:00
}
2019-05-07 17:05:39 +00:00
printk ( KERN_CONT " \n " ) ;
2017-02-10 08:59:15 +00:00
}
2019-05-07 17:05:41 +00:00
/**
* megasas_dump_reg_set - This function will print hexdump of register set
* @ buf : Buffer to be dumped
* @ sz : Size in bytes
* @ format : Different formats of dumping e . g . format = n will
* cause only ' n ' 32 bit words to be dumped in a
* single line .
*/
inline void
megasas_dump_reg_set ( void __iomem * reg_set )
{
unsigned int i , sz = 256 ;
u32 __iomem * reg = ( u32 __iomem * ) reg_set ;
for ( i = 0 ; i < ( sz / sizeof ( u32 ) ) ; i + + )
printk ( " %08x: %08x \n " , ( i * 4 ) , readl ( & reg [ i ] ) ) ;
}
2019-05-07 17:05:37 +00:00
/**
* megasas_dump_fusion_io - This function will print key details
* of SCSI IO
* @ scmd : SCSI command pointer of SCSI IO
*/
void
megasas_dump_fusion_io ( struct scsi_cmnd * scmd )
{
struct megasas_cmd_fusion * cmd ;
union MEGASAS_REQUEST_DESCRIPTOR_UNION * req_desc ;
struct megasas_instance * instance ;
cmd = ( struct megasas_cmd_fusion * ) scmd - > SCp . ptr ;
instance = ( struct megasas_instance * ) scmd - > device - > host - > hostdata ;
scmd_printk ( KERN_INFO , scmd ,
" scmd: (0x%p) retries: 0x%x allowed: 0x%x \n " ,
scmd , scmd - > retries , scmd - > allowed ) ;
scsi_print_command ( scmd ) ;
if ( cmd ) {
req_desc = ( union MEGASAS_REQUEST_DESCRIPTOR_UNION * ) cmd - > request_desc ;
scmd_printk ( KERN_INFO , scmd , " Request descriptor details: \n " ) ;
scmd_printk ( KERN_INFO , scmd ,
" RequestFlags:0x%x MSIxIndex:0x%x SMID:0x%x LMID:0x%x DevHandle:0x%x \n " ,
req_desc - > SCSIIO . RequestFlags ,
req_desc - > SCSIIO . MSIxIndex , req_desc - > SCSIIO . SMID ,
req_desc - > SCSIIO . LMID , req_desc - > SCSIIO . DevHandle ) ;
printk ( KERN_INFO " IO request frame: \n " ) ;
megasas_dump ( cmd - > io_request ,
2019-05-07 17:05:39 +00:00
MEGA_MPI2_RAID_DEFAULT_IO_FRAME_SIZE , 8 ) ;
2019-05-07 17:05:37 +00:00
printk ( KERN_INFO " Chain frame: \n " ) ;
megasas_dump ( cmd - > sg_frame ,
2019-05-07 17:05:39 +00:00
instance - > max_chain_frame_sz , 8 ) ;
2019-05-07 17:05:37 +00:00
}
}
2019-05-07 17:05:40 +00:00
/*
* megasas_dump_sys_regs - This function will dump system registers through
* sysfs .
* @ reg_set : Pointer to System register set .
* @ buf : Buffer to which output is to be written .
* @ return : Number of bytes written to buffer .
*/
static inline ssize_t
megasas_dump_sys_regs ( void __iomem * reg_set , char * buf )
{
unsigned int i , sz = 256 ;
int bytes_wrote = 0 ;
char * loc = ( char * ) buf ;
u32 __iomem * reg = ( u32 __iomem * ) reg_set ;
2017-02-10 08:59:15 +00:00
2019-05-07 17:05:40 +00:00
for ( i = 0 ; i < sz / sizeof ( u32 ) ; i + + ) {
bytes_wrote + = snprintf ( loc + bytes_wrote , PAGE_SIZE ,
" %08x: %08x \n " , ( i * 4 ) ,
readl ( & reg [ i ] ) ) ;
2017-02-10 08:59:15 +00:00
}
2019-05-07 17:05:40 +00:00
return bytes_wrote ;
2017-02-10 08:59:15 +00:00
}
2005-09-20 21:46:58 +00:00
/**
* megasas_reset_bus_host - Bus & host reset handler entry point
*/
static int megasas_reset_bus_host ( struct scsi_cmnd * scmd )
{
int ret ;
2010-12-21 21:34:31 +00:00
struct megasas_instance * instance ;
2015-07-07 20:52:45 +00:00
2010-12-21 21:34:31 +00:00
instance = ( struct megasas_instance * ) scmd - > device - > host - > hostdata ;
2005-09-20 21:46:58 +00:00
2017-02-10 08:59:15 +00:00
scmd_printk ( KERN_INFO , scmd ,
2019-05-07 17:05:37 +00:00
" OCR is requested due to IO timeout!! \n " ) ;
scmd_printk ( KERN_INFO , scmd ,
" SCSI host state: %d SCSI host busy: %d FW outstanding: %d \n " ,
scmd - > device - > host - > shost_state ,
2018-06-24 14:03:26 +00:00
scsi_host_busy ( scmd - > device - > host ) ,
2017-02-10 08:59:15 +00:00
atomic_read ( & instance - > fw_outstanding ) ) ;
2005-09-20 21:46:58 +00:00
/*
2006-03-21 23:21:33 +00:00
* First wait for all commands to complete
2005-09-20 21:46:58 +00:00
*/
2017-10-19 09:48:50 +00:00
if ( instance - > adapter_type = = MFI_SERIES ) {
ret = megasas_generic_reset ( scmd ) ;
} else {
2019-05-07 17:05:37 +00:00
megasas_dump_fusion_io ( scmd ) ;
2017-02-10 08:59:15 +00:00
ret = megasas_reset_fusion ( scmd - > device - > host ,
SCSIIO_TIMEOUT_OCR ) ;
2017-10-19 09:48:50 +00:00
}
2005-09-20 21:46:58 +00:00
return ret ;
}
2016-04-15 07:23:32 +00:00
/**
* megasas_task_abort - Issues task abort request to firmware
* ( supported only for fusion adapters )
* @ scmd : SCSI command pointer
*/
static int megasas_task_abort ( struct scsi_cmnd * scmd )
{
int ret ;
struct megasas_instance * instance ;
instance = ( struct megasas_instance * ) scmd - > device - > host - > hostdata ;
2017-10-19 09:48:50 +00:00
if ( instance - > adapter_type ! = MFI_SERIES )
2016-04-15 07:23:32 +00:00
ret = megasas_task_abort_fusion ( scmd ) ;
else {
sdev_printk ( KERN_NOTICE , scmd - > device , " TASK ABORT not supported \n " ) ;
ret = FAILED ;
}
return ret ;
}
/**
* megasas_reset_target : Issues target reset request to firmware
* ( supported only for fusion adapters )
* @ scmd : SCSI command pointer
*/
static int megasas_reset_target ( struct scsi_cmnd * scmd )
{
int ret ;
struct megasas_instance * instance ;
instance = ( struct megasas_instance * ) scmd - > device - > host - > hostdata ;
2017-10-19 09:48:50 +00:00
if ( instance - > adapter_type ! = MFI_SERIES )
2016-04-15 07:23:32 +00:00
ret = megasas_reset_target_fusion ( scmd ) ;
else {
sdev_printk ( KERN_NOTICE , scmd - > device , " TARGET RESET not supported \n " ) ;
ret = FAILED ;
}
return ret ;
}
2007-02-14 20:41:55 +00:00
/**
* megasas_bios_param - Returns disk geometry for a disk
2015-07-07 20:52:45 +00:00
* @ sdev : device handle
2007-02-14 20:41:55 +00:00
* @ bdev : block device
* @ capacity : drive capacity
* @ geom : geometry parameters
*/
static int
megasas_bios_param ( struct scsi_device * sdev , struct block_device * bdev ,
sector_t capacity , int geom [ ] )
{
int heads ;
int sectors ;
sector_t cylinders ;
unsigned long tmp ;
2015-07-07 20:52:45 +00:00
2007-02-14 20:41:55 +00:00
/* Default heads (64) & sectors (32) */
heads = 64 ;
sectors = 32 ;
tmp = heads * sectors ;
cylinders = capacity ;
sector_div ( cylinders , tmp ) ;
/*
* Handle extended translation size for logical drives > 1 Gb
*/
if ( capacity > = 0x200000 ) {
heads = 255 ;
sectors = 63 ;
tmp = heads * sectors ;
cylinders = capacity ;
sector_div ( cylinders , tmp ) ;
}
geom [ 0 ] = heads ;
geom [ 1 ] = sectors ;
geom [ 2 ] = cylinders ;
return 0 ;
}
2009-10-06 20:50:17 +00:00
static void megasas_aen_polling ( struct work_struct * work ) ;
2005-09-20 21:46:58 +00:00
/**
* megasas_service_aen - Processes an event notification
* @ instance : Adapter soft state
* @ cmd : AEN command completed by the ISR
*
* For AEN , driver sends a command down to FW that is held by the FW till an
* event occurs . When an event of interest occurs , FW completes the command
* that it was previously holding .
*
* This routines sends SIGIO signal to processes that have registered with the
* driver for AEN .
*/
static void
megasas_service_aen ( struct megasas_instance * instance , struct megasas_cmd * cmd )
{
2009-10-06 20:18:02 +00:00
unsigned long flags ;
2015-07-07 20:52:45 +00:00
2005-09-20 21:46:58 +00:00
/*
* Don ' t signal app if it is just an aborted previously registered aen
*/
2009-10-06 20:18:02 +00:00
if ( ( ! cmd - > abort_aen ) & & ( instance - > unload = = 0 ) ) {
spin_lock_irqsave ( & poll_aen_lock , flags ) ;
megasas_poll_wait_aen = 1 ;
spin_unlock_irqrestore ( & poll_aen_lock , flags ) ;
wake_up ( & megasas_poll_wait ) ;
2005-09-20 21:46:58 +00:00
kill_fasync ( & megasas_async_queue , SIGIO , POLL_IN ) ;
2009-10-06 20:18:02 +00:00
}
2005-09-20 21:46:58 +00:00
else
cmd - > abort_aen = 0 ;
instance - > aen_cmd = NULL ;
2014-09-12 13:27:58 +00:00
2015-04-23 11:01:24 +00:00
megasas_return_cmd ( instance , cmd ) ;
2009-10-06 20:50:17 +00:00
2010-09-23 02:36:29 +00:00
if ( ( instance - > unload = = 0 ) & &
( ( instance - > issuepend_done = = 1 ) ) ) {
2009-10-06 20:50:17 +00:00
struct megasas_aen_event * ev ;
2015-07-07 20:52:45 +00:00
2009-10-06 20:50:17 +00:00
ev = kzalloc ( sizeof ( * ev ) , GFP_ATOMIC ) ;
if ( ! ev ) {
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " megasas_service_aen: out of memory \n " ) ;
2009-10-06 20:50:17 +00:00
} else {
ev - > instance = instance ;
instance - > ev = ev ;
2012-12-04 11:33:54 +00:00
INIT_DELAYED_WORK ( & ev - > hotplug_work ,
megasas_aen_polling ) ;
schedule_delayed_work ( & ev - > hotplug_work , 0 ) ;
2009-10-06 20:50:17 +00:00
}
}
2005-09-20 21:46:58 +00:00
}
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
static ssize_t
2019-05-29 16:00:41 +00:00
fw_crash_buffer_store ( struct device * cdev ,
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
struct device_attribute * attr , const char * buf , size_t count )
{
struct Scsi_Host * shost = class_to_shost ( cdev ) ;
struct megasas_instance * instance =
( struct megasas_instance * ) shost - > hostdata ;
int val = 0 ;
unsigned long flags ;
if ( kstrtoint ( buf , 0 , & val ) ! = 0 )
return - EINVAL ;
spin_lock_irqsave ( & instance - > crashdump_lock , flags ) ;
instance - > fw_crash_buffer_offset = val ;
spin_unlock_irqrestore ( & instance - > crashdump_lock , flags ) ;
return strlen ( buf ) ;
}
static ssize_t
2019-05-29 16:00:41 +00:00
fw_crash_buffer_show ( struct device * cdev ,
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
struct device_attribute * attr , char * buf )
{
struct Scsi_Host * shost = class_to_shost ( cdev ) ;
struct megasas_instance * instance =
( struct megasas_instance * ) shost - > hostdata ;
u32 size ;
unsigned long dmachunk = CRASH_DMA_BUF_SIZE ;
2019-07-22 16:15:24 +00:00
unsigned long chunk_left_bytes ;
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
unsigned long src_addr ;
unsigned long flags ;
u32 buff_offset ;
spin_lock_irqsave ( & instance - > crashdump_lock , flags ) ;
buff_offset = instance - > fw_crash_buffer_offset ;
if ( ! instance - > crash_dump_buf & &
! ( ( instance - > fw_crash_state = = AVAILABLE ) | |
( instance - > fw_crash_state = = COPYING ) ) ) {
dev_err ( & instance - > pdev - > dev ,
" Firmware crash dump is not available \n " ) ;
spin_unlock_irqrestore ( & instance - > crashdump_lock , flags ) ;
return - EINVAL ;
}
2015-07-07 20:52:45 +00:00
if ( buff_offset > ( instance - > fw_crash_buffer_size * dmachunk ) ) {
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
dev_err ( & instance - > pdev - > dev ,
" Firmware crash dump offset is out of range \n " ) ;
spin_unlock_irqrestore ( & instance - > crashdump_lock , flags ) ;
return 0 ;
}
size = ( instance - > fw_crash_buffer_size * dmachunk ) - buff_offset ;
2019-07-22 16:15:24 +00:00
chunk_left_bytes = dmachunk - ( buff_offset % dmachunk ) ;
size = ( size > chunk_left_bytes ) ? chunk_left_bytes : size ;
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
size = ( size > = PAGE_SIZE ) ? ( PAGE_SIZE - 1 ) : size ;
src_addr = ( unsigned long ) instance - > crash_buf [ buff_offset / dmachunk ] +
( buff_offset % dmachunk ) ;
2015-07-07 20:52:45 +00:00
memcpy ( buf , ( void * ) src_addr , size ) ;
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
spin_unlock_irqrestore ( & instance - > crashdump_lock , flags ) ;
return size ;
}
static ssize_t
2019-05-29 16:00:41 +00:00
fw_crash_buffer_size_show ( struct device * cdev ,
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
struct device_attribute * attr , char * buf )
{
struct Scsi_Host * shost = class_to_shost ( cdev ) ;
struct megasas_instance * instance =
( struct megasas_instance * ) shost - > hostdata ;
return snprintf ( buf , PAGE_SIZE , " %ld \n " , ( unsigned long )
( ( instance - > fw_crash_buffer_size ) * 1024 * 1024 ) / PAGE_SIZE ) ;
}
static ssize_t
2019-05-29 16:00:41 +00:00
fw_crash_state_store ( struct device * cdev ,
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
struct device_attribute * attr , const char * buf , size_t count )
{
struct Scsi_Host * shost = class_to_shost ( cdev ) ;
struct megasas_instance * instance =
( struct megasas_instance * ) shost - > hostdata ;
int val = 0 ;
unsigned long flags ;
if ( kstrtoint ( buf , 0 , & val ) ! = 0 )
return - EINVAL ;
if ( ( val < = AVAILABLE | | val > COPY_ERROR ) ) {
dev_err ( & instance - > pdev - > dev , " application updates invalid "
" firmware crash state \n " ) ;
return - EINVAL ;
}
instance - > fw_crash_state = val ;
if ( ( val = = COPIED ) | | ( val = = COPY_ERROR ) ) {
spin_lock_irqsave ( & instance - > crashdump_lock , flags ) ;
megasas_free_host_crash_buffer ( instance ) ;
spin_unlock_irqrestore ( & instance - > crashdump_lock , flags ) ;
if ( val = = COPY_ERROR )
dev_info ( & instance - > pdev - > dev , " application failed to "
" copy Firmware crash dump \n " ) ;
else
dev_info ( & instance - > pdev - > dev , " Firmware crash dump "
" copied successfully \n " ) ;
}
return strlen ( buf ) ;
}
static ssize_t
2019-05-29 16:00:41 +00:00
fw_crash_state_show ( struct device * cdev ,
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
struct device_attribute * attr , char * buf )
{
struct Scsi_Host * shost = class_to_shost ( cdev ) ;
struct megasas_instance * instance =
( struct megasas_instance * ) shost - > hostdata ;
2015-07-07 20:52:45 +00:00
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
return snprintf ( buf , PAGE_SIZE , " %d \n " , instance - > fw_crash_state ) ;
}
static ssize_t
2019-05-29 16:00:41 +00:00
page_size_show ( struct device * cdev ,
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
struct device_attribute * attr , char * buf )
{
return snprintf ( buf , PAGE_SIZE , " %ld \n " , ( unsigned long ) PAGE_SIZE - 1 ) ;
}
2016-01-28 15:34:30 +00:00
static ssize_t
2019-05-29 16:00:41 +00:00
ldio_outstanding_show ( struct device * cdev , struct device_attribute * attr ,
2016-01-28 15:34:30 +00:00
char * buf )
{
struct Scsi_Host * shost = class_to_shost ( cdev ) ;
struct megasas_instance * instance = ( struct megasas_instance * ) shost - > hostdata ;
return snprintf ( buf , PAGE_SIZE , " %d \n " , atomic_read ( & instance - > ldio_outstanding ) ) ;
}
2018-01-05 13:27:46 +00:00
static ssize_t
2019-05-29 16:00:41 +00:00
fw_cmds_outstanding_show ( struct device * cdev ,
2018-01-05 13:27:46 +00:00
struct device_attribute * attr , char * buf )
{
struct Scsi_Host * shost = class_to_shost ( cdev ) ;
struct megasas_instance * instance = ( struct megasas_instance * ) shost - > hostdata ;
return snprintf ( buf , PAGE_SIZE , " %d \n " , atomic_read ( & instance - > fw_outstanding ) ) ;
}
2019-07-26 07:32:14 +00:00
static ssize_t
enable_sdev_max_qd_show ( struct device * cdev ,
struct device_attribute * attr , char * buf )
{
struct Scsi_Host * shost = class_to_shost ( cdev ) ;
struct megasas_instance * instance = ( struct megasas_instance * ) shost - > hostdata ;
return snprintf ( buf , PAGE_SIZE , " %d \n " , instance - > enable_sdev_max_qd ) ;
}
static ssize_t
enable_sdev_max_qd_store ( struct device * cdev ,
struct device_attribute * attr , const char * buf , size_t count )
{
struct Scsi_Host * shost = class_to_shost ( cdev ) ;
struct megasas_instance * instance = ( struct megasas_instance * ) shost - > hostdata ;
u32 val = 0 ;
bool is_target_prop ;
int ret_target_prop = DCMD_FAILED ;
struct scsi_device * sdev ;
if ( kstrtou32 ( buf , 0 , & val ) ! = 0 ) {
pr_err ( " megasas: could not set enable_sdev_max_qd \n " ) ;
return - EINVAL ;
}
mutex_lock ( & instance - > reset_mutex ) ;
if ( val )
instance - > enable_sdev_max_qd = true ;
else
instance - > enable_sdev_max_qd = false ;
shost_for_each_device ( sdev , shost ) {
ret_target_prop = megasas_get_target_prop ( instance , sdev ) ;
is_target_prop = ( ret_target_prop = = DCMD_SUCCESS ) ? true : false ;
megasas_set_fw_assisted_qd ( sdev , is_target_prop ) ;
}
mutex_unlock ( & instance - > reset_mutex ) ;
return strlen ( buf ) ;
}
2019-05-07 17:05:40 +00:00
static ssize_t
2019-05-29 16:00:41 +00:00
dump_system_regs_show ( struct device * cdev ,
2019-05-07 17:05:40 +00:00
struct device_attribute * attr , char * buf )
{
struct Scsi_Host * shost = class_to_shost ( cdev ) ;
struct megasas_instance * instance =
( struct megasas_instance * ) shost - > hostdata ;
return megasas_dump_sys_regs ( instance - > reg_set , buf ) ;
}
2019-05-07 17:05:43 +00:00
static ssize_t
2019-05-29 16:00:41 +00:00
raid_map_id_show ( struct device * cdev , struct device_attribute * attr ,
2019-05-07 17:05:43 +00:00
char * buf )
{
struct Scsi_Host * shost = class_to_shost ( cdev ) ;
struct megasas_instance * instance =
( struct megasas_instance * ) shost - > hostdata ;
return snprintf ( buf , PAGE_SIZE , " %ld \n " ,
( unsigned long ) instance - > map_id ) ;
}
2019-05-29 16:00:41 +00:00
static DEVICE_ATTR_RW ( fw_crash_buffer ) ;
static DEVICE_ATTR_RO ( fw_crash_buffer_size ) ;
static DEVICE_ATTR_RW ( fw_crash_state ) ;
static DEVICE_ATTR_RO ( page_size ) ;
static DEVICE_ATTR_RO ( ldio_outstanding ) ;
static DEVICE_ATTR_RO ( fw_cmds_outstanding ) ;
2019-07-26 07:32:14 +00:00
static DEVICE_ATTR_RW ( enable_sdev_max_qd ) ;
2019-05-29 16:00:41 +00:00
static DEVICE_ATTR_RO ( dump_system_regs ) ;
static DEVICE_ATTR_RO ( raid_map_id ) ;
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
2019-07-02 13:01:14 +00:00
static struct device_attribute * megaraid_host_attrs [ ] = {
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
& dev_attr_fw_crash_buffer_size ,
& dev_attr_fw_crash_buffer ,
& dev_attr_fw_crash_state ,
& dev_attr_page_size ,
2016-01-28 15:34:30 +00:00
& dev_attr_ldio_outstanding ,
2018-01-05 13:27:46 +00:00
& dev_attr_fw_cmds_outstanding ,
2019-07-26 07:32:14 +00:00
& dev_attr_enable_sdev_max_qd ,
2019-05-07 17:05:40 +00:00
& dev_attr_dump_system_regs ,
2019-05-07 17:05:43 +00:00
& dev_attr_raid_map_id ,
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
NULL ,
} ;
2005-09-20 21:46:58 +00:00
/*
* Scsi host template for megaraid_sas driver
*/
static struct scsi_host_template megasas_template = {
. module = THIS_MODULE ,
2015-04-23 11:01:39 +00:00
. name = " Avago SAS based MegaRAID driver " ,
2005-09-20 21:46:58 +00:00
. proc_name = " megaraid_sas " ,
2006-02-17 11:13:48 +00:00
. slave_configure = megasas_slave_configure ,
2009-10-06 20:33:06 +00:00
. slave_alloc = megasas_slave_alloc ,
2016-01-28 15:34:25 +00:00
. slave_destroy = megasas_slave_destroy ,
2005-09-20 21:46:58 +00:00
. queuecommand = megasas_queue_command ,
2016-04-15 07:23:32 +00:00
. eh_target_reset_handler = megasas_reset_target ,
. eh_abort_handler = megasas_task_abort ,
2005-09-20 21:46:58 +00:00
. eh_host_reset_handler = megasas_reset_bus_host ,
2007-05-17 12:47:51 +00:00
. eh_timed_out = megasas_reset_timer ,
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
. shost_attrs = megaraid_host_attrs ,
2007-02-14 20:41:55 +00:00
. bios_param = megasas_bios_param ,
2014-11-13 14:08:42 +00:00
. change_queue_depth = scsi_change_queue_depth ,
2019-06-17 12:20:00 +00:00
. max_segment_size = 0xffffffff ,
2013-10-23 10:25:40 +00:00
. no_write_same = 1 ,
2005-09-20 21:46:58 +00:00
} ;
/**
* megasas_complete_int_cmd - Completes an internal command
* @ instance : Adapter soft state
* @ cmd : Command to be completed
*
* The megasas_issue_blocked_cmd ( ) function waits for a command to complete
* after it issues a command . This function wakes up that waiting routine by
* calling wake_up ( ) on the wait queue .
*/
static void
megasas_complete_int_cmd ( struct megasas_instance * instance ,
struct megasas_cmd * cmd )
{
2015-05-06 13:31:02 +00:00
cmd - > cmd_status_drv = cmd - > frame - > io . cmd_status ;
2005-09-20 21:46:58 +00:00
wake_up ( & instance - > int_cmd_wait_q ) ;
}
/**
* megasas_complete_abort - Completes aborting a command
* @ instance : Adapter soft state
* @ cmd : Cmd that was issued to abort another cmd
*
2010-12-15 03:17:17 +00:00
* The megasas_issue_blocked_abort_cmd ( ) function waits on abort_cmd_wait_q
* after it issues an abort on a previously issued command . This function
2005-09-20 21:46:58 +00:00
* wakes up all functions waiting on the same wait queue .
*/
static void
megasas_complete_abort ( struct megasas_instance * instance ,
struct megasas_cmd * cmd )
{
if ( cmd - > sync_cmd ) {
cmd - > sync_cmd = 0 ;
2015-05-06 13:31:02 +00:00
cmd - > cmd_status_drv = 0 ;
2005-09-20 21:46:58 +00:00
wake_up ( & instance - > abort_cmd_wait_q ) ;
}
}
/**
* megasas_complete_cmd - Completes a command
* @ instance : Adapter soft state
* @ cmd : Command to be completed
2010-12-15 03:17:17 +00:00
* @ alt_status : If non - zero , use this value as status to
2015-07-07 20:52:45 +00:00
* SCSI mid - layer instead of the value returned
* by the FW . This should be used if caller wants
* an alternate status ( as in the case of aborted
* commands )
2005-09-20 21:46:58 +00:00
*/
2010-12-21 21:34:31 +00:00
void
2005-09-20 21:46:58 +00:00
megasas_complete_cmd ( struct megasas_instance * instance , struct megasas_cmd * cmd ,
u8 alt_status )
{
int exception = 0 ;
struct megasas_header * hdr = & cmd - > frame - > hdr ;
2009-10-06 20:18:02 +00:00
unsigned long flags ;
2010-12-21 21:34:31 +00:00
struct fusion_context * fusion = instance - > ctrl_context ;
2015-08-31 11:53:11 +00:00
u32 opcode , status ;
2005-09-20 21:46:58 +00:00
2010-09-23 02:36:29 +00:00
/* flag for the retry reset */
cmd - > retry_for_fw_reset = 0 ;
2007-05-17 12:47:51 +00:00
if ( cmd - > scmd )
cmd - > scmd - > SCp . ptr = NULL ;
2005-09-20 21:46:58 +00:00
switch ( hdr - > cmd ) {
2011-10-09 01:15:19 +00:00
case MFI_CMD_INVALID :
/* Some older 1068 controller FW may keep a pended
MR_DCMD_CTRL_EVENT_GET_INFO left over from the main kernel
when booting the kdump kernel . Ignore this command to
prevent a kernel panic on shutdown of the kdump kernel . */
2015-07-07 20:52:34 +00:00
dev_warn ( & instance - > pdev - > dev , " MFI_CMD_INVALID command "
" completed \n " ) ;
dev_warn ( & instance - > pdev - > dev , " If you have a controller "
" other than PERC5, please upgrade your firmware \n " ) ;
2011-10-09 01:15:19 +00:00
break ;
2005-09-20 21:46:58 +00:00
case MFI_CMD_PD_SCSI_IO :
case MFI_CMD_LD_SCSI_IO :
/*
* MFI_CMD_PD_SCSI_IO and MFI_CMD_LD_SCSI_IO could have been
* issued either through an IO path or an IOCTL path . If it
* was via IOCTL , we will send it to internal completion .
*/
if ( cmd - > sync_cmd ) {
cmd - > sync_cmd = 0 ;
megasas_complete_int_cmd ( instance , cmd ) ;
break ;
}
2018-11-28 04:32:27 +00:00
/* fall through */
2005-09-20 21:46:58 +00:00
case MFI_CMD_LD_READ :
case MFI_CMD_LD_WRITE :
if ( alt_status ) {
cmd - > scmd - > result = alt_status < < 16 ;
exception = 1 ;
}
if ( exception ) {
2006-05-30 19:03:37 +00:00
atomic_dec ( & instance - > fw_outstanding ) ;
2005-09-20 21:46:58 +00:00
2007-05-25 20:04:08 +00:00
scsi_dma_unmap ( cmd - > scmd ) ;
2005-09-20 21:46:58 +00:00
cmd - > scmd - > scsi_done ( cmd - > scmd ) ;
megasas_return_cmd ( instance , cmd ) ;
break ;
}
switch ( hdr - > cmd_status ) {
case MFI_STAT_OK :
cmd - > scmd - > result = DID_OK < < 16 ;
break ;
case MFI_STAT_SCSI_IO_FAILED :
case MFI_STAT_LD_INIT_IN_PROGRESS :
cmd - > scmd - > result =
( DID_ERROR < < 16 ) | hdr - > scsi_status ;
break ;
case MFI_STAT_SCSI_DONE_WITH_ERROR :
cmd - > scmd - > result = ( DID_OK < < 16 ) | hdr - > scsi_status ;
if ( hdr - > scsi_status = = SAM_STAT_CHECK_CONDITION ) {
memset ( cmd - > scmd - > sense_buffer , 0 ,
SCSI_SENSE_BUFFERSIZE ) ;
memcpy ( cmd - > scmd - > sense_buffer , cmd - > sense ,
hdr - > sense_len ) ;
cmd - > scmd - > result | = DRIVER_SENSE < < 24 ;
}
break ;
case MFI_STAT_LD_OFFLINE :
case MFI_STAT_DEVICE_NOT_FOUND :
cmd - > scmd - > result = DID_BAD_TARGET < < 16 ;
break ;
default :
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " MFI FW status %#x \n " ,
2005-09-20 21:46:58 +00:00
hdr - > cmd_status ) ;
cmd - > scmd - > result = DID_ERROR < < 16 ;
break ;
}
2006-05-30 19:03:37 +00:00
atomic_dec ( & instance - > fw_outstanding ) ;
2005-09-20 21:46:58 +00:00
2007-05-25 20:04:08 +00:00
scsi_dma_unmap ( cmd - > scmd ) ;
2005-09-20 21:46:58 +00:00
cmd - > scmd - > scsi_done ( cmd - > scmd ) ;
megasas_return_cmd ( instance , cmd ) ;
break ;
case MFI_CMD_SMP :
case MFI_CMD_STP :
2018-01-05 13:33:04 +00:00
case MFI_CMD_NVME :
2019-06-25 11:04:30 +00:00
case MFI_CMD_TOOLBOX :
2017-10-19 09:49:02 +00:00
megasas_complete_int_cmd ( instance , cmd ) ;
break ;
2005-09-20 21:46:58 +00:00
case MFI_CMD_DCMD :
2013-09-06 10:20:52 +00:00
opcode = le32_to_cpu ( cmd - > frame - > dcmd . opcode ) ;
2010-12-21 21:34:31 +00:00
/* Check for LD map update */
2013-09-06 10:20:52 +00:00
if ( ( opcode = = MR_DCMD_LD_MAP_GET_INFO )
& & ( cmd - > frame - > dcmd . mbox . b [ 1 ] = = 1 ) ) {
2013-05-22 07:05:04 +00:00
fusion - > fast_path_io = 0 ;
2010-12-21 21:34:31 +00:00
spin_lock_irqsave ( instance - > host - > host_lock , flags ) ;
2018-01-05 13:27:44 +00:00
status = cmd - > frame - > hdr . cmd_status ;
2015-08-31 11:53:11 +00:00
instance - > map_update_cmd = NULL ;
2018-01-05 13:27:44 +00:00
if ( status ! = MFI_STAT_OK ) {
if ( status ! = MFI_STAT_NOT_FOUND )
2015-07-07 20:52:34 +00:00
dev_warn ( & instance - > pdev - > dev , " map syncfailed, status = 0x%x \n " ,
2010-12-21 21:34:31 +00:00
cmd - > frame - > hdr . cmd_status ) ;
else {
2015-04-23 11:01:24 +00:00
megasas_return_cmd ( instance , cmd ) ;
2010-12-21 21:34:31 +00:00
spin_unlock_irqrestore (
instance - > host - > host_lock ,
flags ) ;
break ;
}
2018-01-05 13:27:44 +00:00
}
2015-04-23 11:01:24 +00:00
megasas_return_cmd ( instance , cmd ) ;
2013-05-22 07:05:04 +00:00
/*
* Set fast path IO to ZERO .
* Validate Map will set proper value .
* Meanwhile all IOs will go as LD IO .
*/
2018-01-05 13:27:44 +00:00
if ( status = = MFI_STAT_OK & &
( MR_ValidateMapInfo ( instance , ( instance - > map_id + 1 ) ) ) ) {
instance - > map_id + + ;
2010-12-21 21:34:31 +00:00
fusion - > fast_path_io = 1 ;
2018-01-05 13:27:44 +00:00
} else {
2010-12-21 21:34:31 +00:00
fusion - > fast_path_io = 0 ;
2018-01-05 13:27:44 +00:00
}
2010-12-21 21:34:31 +00:00
megasas_sync_map_info ( instance ) ;
spin_unlock_irqrestore ( instance - > host - > host_lock ,
flags ) ;
break ;
}
2013-09-06 10:20:52 +00:00
if ( opcode = = MR_DCMD_CTRL_EVENT_GET_INFO | |
opcode = = MR_DCMD_CTRL_EVENT_GET ) {
2009-10-06 20:18:02 +00:00
spin_lock_irqsave ( & poll_aen_lock , flags ) ;
megasas_poll_wait_aen = 0 ;
spin_unlock_irqrestore ( & poll_aen_lock , flags ) ;
}
2005-09-20 21:46:58 +00:00
2015-08-31 11:53:11 +00:00
/* FW has an updated PD sequence */
if ( ( opcode = = MR_DCMD_SYSTEM_PD_MAP_GET_INFO ) & &
( cmd - > frame - > dcmd . mbox . b [ 0 ] = = 1 ) ) {
spin_lock_irqsave ( instance - > host - > host_lock , flags ) ;
status = cmd - > frame - > hdr . cmd_status ;
instance - > jbod_seq_cmd = NULL ;
megasas_return_cmd ( instance , cmd ) ;
if ( status = = MFI_STAT_OK ) {
instance - > pd_seq_map_id + + ;
/* Re-register a pd sync seq num cmd */
if ( megasas_sync_pd_seq_num ( instance , true ) )
instance - > use_seqnum_jbod_fp = false ;
} else
instance - > use_seqnum_jbod_fp = false ;
spin_unlock_irqrestore ( instance - > host - > host_lock , flags ) ;
break ;
}
2005-09-20 21:46:58 +00:00
/*
* See if got an event notification
*/
2013-09-06 10:20:52 +00:00
if ( opcode = = MR_DCMD_CTRL_EVENT_WAIT )
2005-09-20 21:46:58 +00:00
megasas_service_aen ( instance , cmd ) ;
else
megasas_complete_int_cmd ( instance , cmd ) ;
break ;
case MFI_CMD_ABORT :
/*
* Cmd issued to abort another cmd returned
*/
megasas_complete_abort ( instance , cmd ) ;
break ;
default :
2015-07-07 20:52:34 +00:00
dev_info ( & instance - > pdev - > dev , " Unknown command completed! [0x%X] \n " ,
2005-09-20 21:46:58 +00:00
hdr - > cmd ) ;
2017-10-19 09:49:02 +00:00
megasas_complete_int_cmd ( instance , cmd ) ;
2005-09-20 21:46:58 +00:00
break ;
}
}
2010-09-23 02:36:29 +00:00
/**
* megasas_issue_pending_cmds_again - issue all pending cmds
2015-07-07 20:52:45 +00:00
* in FW again because of the fw reset
2010-09-23 02:36:29 +00:00
* @ instance : Adapter soft state
*/
static inline void
megasas_issue_pending_cmds_again ( struct megasas_instance * instance )
{
struct megasas_cmd * cmd ;
struct list_head clist_local ;
union megasas_evt_class_locale class_locale ;
unsigned long flags ;
u32 seq_num ;
INIT_LIST_HEAD ( & clist_local ) ;
spin_lock_irqsave ( & instance - > hba_lock , flags ) ;
list_splice_init ( & instance - > internal_reset_pending_q , & clist_local ) ;
spin_unlock_irqrestore ( & instance - > hba_lock , flags ) ;
while ( ! list_empty ( & clist_local ) ) {
2015-07-07 20:52:45 +00:00
cmd = list_entry ( ( & clist_local ) - > next ,
2010-09-23 02:36:29 +00:00
struct megasas_cmd , list ) ;
list_del_init ( & cmd - > list ) ;
if ( cmd - > sync_cmd | | cmd - > scmd ) {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " command %p, %p:%d "
" detected to be pending while HBA reset \n " ,
2010-09-23 02:36:29 +00:00
cmd , cmd - > scmd , cmd - > sync_cmd ) ;
cmd - > retry_for_fw_reset + + ;
if ( cmd - > retry_for_fw_reset = = 3 ) {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " cmd %p, %p:%d "
2010-09-23 02:36:29 +00:00
" was tried multiple times during reset. "
" Shutting down the HBA \n " ,
cmd , cmd - > scmd , cmd - > sync_cmd ) ;
2015-01-05 14:36:18 +00:00
instance - > instancet - > disable_intr ( instance ) ;
atomic_set ( & instance - > fw_reset_no_pci_access , 1 ) ;
2010-09-23 02:36:29 +00:00
megaraid_sas_kill_hba ( instance ) ;
return ;
}
}
if ( cmd - > sync_cmd = = 1 ) {
if ( cmd - > scmd ) {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " unexpected "
2010-09-23 02:36:29 +00:00
" cmd attached to internal command! \n " ) ;
}
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " %p synchronous cmd "
2010-09-23 02:36:29 +00:00
" on the internal reset queue, "
" issue it again. \n " , cmd ) ;
2015-05-06 13:31:02 +00:00
cmd - > cmd_status_drv = MFI_STAT_INVALID_STATUS ;
2010-09-23 02:36:29 +00:00
instance - > instancet - > fire_cmd ( instance ,
2015-07-07 20:52:45 +00:00
cmd - > frame_phys_addr ,
2010-09-23 02:36:29 +00:00
0 , instance - > reg_set ) ;
} else if ( cmd - > scmd ) {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " %p scsi cmd [%02x] "
2010-09-23 02:36:29 +00:00
" detected on the internal queue, issue again. \n " ,
2011-04-04 13:42:14 +00:00
cmd , cmd - > scmd - > cmnd [ 0 ] ) ;
2010-09-23 02:36:29 +00:00
atomic_inc ( & instance - > fw_outstanding ) ;
instance - > instancet - > fire_cmd ( instance ,
cmd - > frame_phys_addr ,
cmd - > frame_count - 1 , instance - > reg_set ) ;
} else {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " %p unexpected cmd on the "
2010-09-23 02:36:29 +00:00
" internal reset defer list while re-issue!! \n " ,
cmd ) ;
}
}
if ( instance - > aen_cmd ) {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " aen_cmd in def process \n " ) ;
2010-09-23 02:36:29 +00:00
megasas_return_cmd ( instance , instance - > aen_cmd ) ;
2015-07-07 20:52:45 +00:00
instance - > aen_cmd = NULL ;
2010-09-23 02:36:29 +00:00
}
/*
2015-07-07 20:52:45 +00:00
* Initiate AEN ( Asynchronous Event Notification )
*/
2010-09-23 02:36:29 +00:00
seq_num = instance - > last_seq_num ;
class_locale . members . reserved = 0 ;
class_locale . members . locale = MR_EVT_LOCALE_ALL ;
class_locale . members . class = MR_EVT_CLASS_DEBUG ;
megasas_register_aen ( instance , seq_num , class_locale . word ) ;
}
/**
* Move the internal reset pending commands to a deferred queue .
*
* We move the commands pending at internal reset time to a
* pending queue . This queue would be flushed after successful
* completion of the internal reset sequence . if the internal reset
* did not complete in time , the kernel reset handler would flush
* these commands .
* */
static void
megasas_internal_reset_defer_cmds ( struct megasas_instance * instance )
{
struct megasas_cmd * cmd ;
int i ;
2017-02-10 08:59:25 +00:00
u16 max_cmd = instance - > max_fw_cmds ;
2010-09-23 02:36:29 +00:00
u32 defer_index ;
unsigned long flags ;
2015-07-07 20:52:45 +00:00
defer_index = 0 ;
2014-09-12 13:27:58 +00:00
spin_lock_irqsave ( & instance - > mfi_pool_lock , flags ) ;
2010-09-23 02:36:29 +00:00
for ( i = 0 ; i < max_cmd ; i + + ) {
cmd = instance - > cmd_list [ i ] ;
if ( cmd - > sync_cmd = = 1 | | cmd - > scmd ) {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " moving cmd[%d]:%p:%d:%p "
2010-09-23 02:36:29 +00:00
" on the defer queue as internal \n " ,
defer_index , cmd , cmd - > sync_cmd , cmd - > scmd ) ;
if ( ! list_empty ( & cmd - > list ) ) {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " ERROR while "
2010-09-23 02:36:29 +00:00
" moving this cmd:%p, %d %p, it was "
" discovered on some list? \n " ,
cmd , cmd - > sync_cmd , cmd - > scmd ) ;
list_del_init ( & cmd - > list ) ;
}
defer_index + + ;
list_add_tail ( & cmd - > list ,
& instance - > internal_reset_pending_q ) ;
}
}
2014-09-12 13:27:58 +00:00
spin_unlock_irqrestore ( & instance - > mfi_pool_lock , flags ) ;
2010-09-23 02:36:29 +00:00
}
static void
process_fw_state_change_wq ( struct work_struct * work )
{
struct megasas_instance * instance =
container_of ( work , struct megasas_instance , work_init ) ;
u32 wait ;
unsigned long flags ;
2016-01-28 15:34:32 +00:00
if ( atomic_read ( & instance - > adprecovery ) ! = MEGASAS_ADPRESET_SM_INFAULT ) {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " error, recovery st %x \n " ,
2016-01-28 15:34:32 +00:00
atomic_read ( & instance - > adprecovery ) ) ;
2010-09-23 02:36:29 +00:00
return ;
}
2016-01-28 15:34:32 +00:00
if ( atomic_read ( & instance - > adprecovery ) = = MEGASAS_ADPRESET_SM_INFAULT ) {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " FW detected to be in fault "
2010-09-23 02:36:29 +00:00
" state, restarting it... \n " ) ;
2013-05-22 07:04:14 +00:00
instance - > instancet - > disable_intr ( instance ) ;
2010-09-23 02:36:29 +00:00
atomic_set ( & instance - > fw_outstanding , 0 ) ;
atomic_set ( & instance - > fw_reset_no_pci_access , 1 ) ;
instance - > instancet - > adp_reset ( instance , instance - > reg_set ) ;
2015-07-07 20:52:45 +00:00
atomic_set ( & instance - > fw_reset_no_pci_access , 0 ) ;
2010-09-23 02:36:29 +00:00
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " FW restarted successfully, "
2010-09-23 02:36:29 +00:00
" initiating next stage... \n " ) ;
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " HBA recovery state machine, "
2010-09-23 02:36:29 +00:00
" state 2 starting... \n " ) ;
2015-07-07 20:52:45 +00:00
/* waiting for about 20 second before start the second init */
2010-09-23 02:36:29 +00:00
for ( wait = 0 ; wait < 30 ; wait + + ) {
msleep ( 1000 ) ;
}
2011-10-09 01:14:27 +00:00
if ( megasas_transition_to_ready ( instance , 1 ) ) {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " adapter not ready \n " ) ;
2010-09-23 02:36:29 +00:00
2015-01-05 14:36:18 +00:00
atomic_set ( & instance - > fw_reset_no_pci_access , 1 ) ;
2010-09-23 02:36:29 +00:00
megaraid_sas_kill_hba ( instance ) ;
return ;
}
if ( ( instance - > pdev - > device = = PCI_DEVICE_ID_LSI_SAS1064R ) | |
( instance - > pdev - > device = = PCI_DEVICE_ID_DELL_PERC5 ) | |
( instance - > pdev - > device = = PCI_DEVICE_ID_LSI_VERDE_ZCR )
) {
* instance - > consumer = * instance - > producer ;
} else {
* instance - > consumer = 0 ;
* instance - > producer = 0 ;
}
megasas_issue_init_mfi ( instance ) ;
spin_lock_irqsave ( & instance - > hba_lock , flags ) ;
2016-01-28 15:34:32 +00:00
atomic_set ( & instance - > adprecovery , MEGASAS_HBA_OPERATIONAL ) ;
2010-09-23 02:36:29 +00:00
spin_unlock_irqrestore ( & instance - > hba_lock , flags ) ;
2013-05-22 07:04:14 +00:00
instance - > instancet - > enable_intr ( instance ) ;
2010-09-23 02:36:29 +00:00
megasas_issue_pending_cmds_again ( instance ) ;
instance - > issuepend_done = 1 ;
}
}
2005-09-20 21:46:58 +00:00
/**
* megasas_deplete_reply_queue - Processes all completed commands
* @ instance : Adapter soft state
* @ alt_status : Alternate status to be returned to
2015-07-07 20:52:45 +00:00
* SCSI mid - layer instead of the status
* returned by the FW
2010-09-23 02:36:29 +00:00
* Note : this must be called with hba lock held
2005-09-20 21:46:58 +00:00
*/
2006-01-14 21:20:43 +00:00
static int
2010-09-23 02:36:29 +00:00
megasas_deplete_reply_queue ( struct megasas_instance * instance ,
u8 alt_status )
2005-09-20 21:46:58 +00:00
{
2010-09-23 02:36:29 +00:00
u32 mfiStatus ;
u32 fw_state ;
if ( ( mfiStatus = instance - > instancet - > check_reset ( instance ,
instance - > reg_set ) ) = = 1 ) {
return IRQ_HANDLED ;
}
2018-12-17 08:47:39 +00:00
mfiStatus = instance - > instancet - > clear_intr ( instance ) ;
if ( mfiStatus = = 0 ) {
2011-02-25 04:56:21 +00:00
/* Hardware may not set outbound_intr_status in MSI-X mode */
2011-10-09 01:15:13 +00:00
if ( ! instance - > msix_vectors )
2011-02-25 04:56:21 +00:00
return IRQ_NONE ;
2010-09-23 02:36:29 +00:00
}
instance - > mfiStatus = mfiStatus ;
if ( ( mfiStatus & MFI_INTR_FLAG_FIRMWARE_STATE_CHANGE ) ) {
fw_state = instance - > instancet - > read_fw_status_reg (
2018-12-17 08:47:39 +00:00
instance ) & MFI_STATE_MASK ;
2010-09-23 02:36:29 +00:00
if ( fw_state ! = MFI_STATE_FAULT ) {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " fw state:%x \n " ,
2010-09-23 02:36:29 +00:00
fw_state ) ;
}
if ( ( fw_state = = MFI_STATE_FAULT ) & &
( instance - > disableOnlineCtrlReset = = 0 ) ) {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " wait adp restart \n " ) ;
2010-09-23 02:36:29 +00:00
if ( ( instance - > pdev - > device = =
PCI_DEVICE_ID_LSI_SAS1064R ) | |
( instance - > pdev - > device = =
PCI_DEVICE_ID_DELL_PERC5 ) | |
( instance - > pdev - > device = =
PCI_DEVICE_ID_LSI_VERDE_ZCR ) ) {
* instance - > consumer =
2013-09-06 10:20:52 +00:00
cpu_to_le32 ( MEGASAS_ADPRESET_INPROG_SIGN ) ;
2010-09-23 02:36:29 +00:00
}
2013-05-22 07:04:14 +00:00
instance - > instancet - > disable_intr ( instance ) ;
2016-01-28 15:34:32 +00:00
atomic_set ( & instance - > adprecovery , MEGASAS_ADPRESET_SM_INFAULT ) ;
2010-09-23 02:36:29 +00:00
instance - > issuepend_done = 0 ;
atomic_set ( & instance - > fw_outstanding , 0 ) ;
megasas_internal_reset_defer_cmds ( instance ) ;
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " fwState=%x, stage:%d \n " ,
2016-01-28 15:34:32 +00:00
fw_state , atomic_read ( & instance - > adprecovery ) ) ;
2010-09-23 02:36:29 +00:00
schedule_work ( & instance - > work_init ) ;
return IRQ_HANDLED ;
} else {
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " fwstate:%x, dis_OCR=%x \n " ,
2010-09-23 02:36:29 +00:00
fw_state , instance - > disableOnlineCtrlReset ) ;
}
}
2005-09-20 21:46:58 +00:00
2006-10-03 20:13:18 +00:00
tasklet_schedule ( & instance - > isr_tasklet ) ;
2005-09-20 21:46:58 +00:00
return IRQ_HANDLED ;
}
/**
* megasas_isr - isr entry point
*/
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers
Maintain a per-CPU global "struct pt_regs *" variable which can be used instead
of passing regs around manually through all ~1800 interrupt handlers in the
Linux kernel.
The regs pointer is used in few places, but it potentially costs both stack
space and code to pass it around. On the FRV arch, removing the regs parameter
from all the genirq function results in a 20% speed up of the IRQ exit path
(ie: from leaving timer_interrupt() to leaving do_IRQ()).
Where appropriate, an arch may override the generic storage facility and do
something different with the variable. On FRV, for instance, the address is
maintained in GR28 at all times inside the kernel as part of general exception
handling.
Having looked over the code, it appears that the parameter may be handed down
through up to twenty or so layers of functions. Consider a USB character
device attached to a USB hub, attached to a USB controller that posts its
interrupts through a cascaded auxiliary interrupt controller. A character
device driver may want to pass regs to the sysrq handler through the input
layer which adds another few layers of parameter passing.
I've build this code with allyesconfig for x86_64 and i386. I've runtested the
main part of the code on FRV and i386, though I can't test most of the drivers.
I've also done partial conversion for powerpc and MIPS - these at least compile
with minimal configurations.
This will affect all archs. Mostly the changes should be relatively easy.
Take do_IRQ(), store the regs pointer at the beginning, saving the old one:
struct pt_regs *old_regs = set_irq_regs(regs);
And put the old one back at the end:
set_irq_regs(old_regs);
Don't pass regs through to generic_handle_irq() or __do_IRQ().
In timer_interrupt(), this sort of change will be necessary:
- update_process_times(user_mode(regs));
- profile_tick(CPU_PROFILING, regs);
+ update_process_times(user_mode(get_irq_regs()));
+ profile_tick(CPU_PROFILING);
I'd like to move update_process_times()'s use of get_irq_regs() into itself,
except that i386, alone of the archs, uses something other than user_mode().
Some notes on the interrupt handling in the drivers:
(*) input_dev() is now gone entirely. The regs pointer is no longer stored in
the input_dev struct.
(*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does
something different depending on whether it's been supplied with a regs
pointer or not.
(*) Various IRQ handler function pointers have been moved to type
irq_handler_t.
Signed-Off-By: David Howells <dhowells@redhat.com>
(cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
static irqreturn_t megasas_isr ( int irq , void * devp )
2005-09-20 21:46:58 +00:00
{
2011-10-09 01:15:13 +00:00
struct megasas_irq_context * irq_context = devp ;
struct megasas_instance * instance = irq_context - > instance ;
2010-09-23 02:36:29 +00:00
unsigned long flags ;
2015-07-07 20:52:45 +00:00
irqreturn_t rc ;
2010-09-23 02:36:29 +00:00
2011-10-09 01:15:13 +00:00
if ( atomic_read ( & instance - > fw_reset_no_pci_access ) )
2010-09-23 02:36:29 +00:00
return IRQ_HANDLED ;
spin_lock_irqsave ( & instance - > hba_lock , flags ) ;
2015-07-07 20:52:45 +00:00
rc = megasas_deplete_reply_queue ( instance , DID_OK ) ;
2010-09-23 02:36:29 +00:00
spin_unlock_irqrestore ( & instance - > hba_lock , flags ) ;
return rc ;
2005-09-20 21:46:58 +00:00
}
/**
* megasas_transition_to_ready - Move the FW to READY state
2006-01-25 20:02:40 +00:00
* @ instance : Adapter soft state
2005-09-20 21:46:58 +00:00
*
* During the initialization , FW passes can potentially be in any one of
* several possible states . If the FW in operational , waiting - for - handshake
* states , driver must take steps to bring it to ready state . Otherwise , it
* has to wait for the ready state .
*/
2010-12-21 21:34:31 +00:00
int
2011-10-09 01:14:27 +00:00
megasas_transition_to_ready ( struct megasas_instance * instance , int ocr )
2005-09-20 21:46:58 +00:00
{
int i ;
u8 max_wait ;
u32 fw_state ;
2009-10-06 20:52:20 +00:00
u32 abs_state , curr_abs_state ;
2005-09-20 21:46:58 +00:00
2018-12-17 08:47:39 +00:00
abs_state = instance - > instancet - > read_fw_status_reg ( instance ) ;
2014-04-01 11:59:50 +00:00
fw_state = abs_state & MFI_STATE_MASK ;
2005-09-20 21:46:58 +00:00
2006-10-03 19:28:49 +00:00
if ( fw_state ! = MFI_STATE_READY )
2015-07-07 20:52:34 +00:00
dev_info ( & instance - > pdev - > dev , " Waiting for FW to come to ready "
2010-12-15 03:17:17 +00:00
" state \n " ) ;
2006-10-03 19:28:49 +00:00
2005-09-20 21:46:58 +00:00
while ( fw_state ! = MFI_STATE_READY ) {
switch ( fw_state ) {
case MFI_STATE_FAULT :
2019-05-07 17:05:44 +00:00
dev_printk ( KERN_ERR , & instance - > pdev - > dev ,
" FW in FAULT state, Fault code:0x%x subcode:0x%x func:%s \n " ,
abs_state & MFI_STATE_FAULT_CODE ,
abs_state & MFI_STATE_FAULT_SUBCODE , __func__ ) ;
2011-10-09 01:14:27 +00:00
if ( ocr ) {
max_wait = MEGASAS_RESET_WAIT_TIME ;
break ;
2019-05-07 17:05:41 +00:00
} else {
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " System Register set: \n " ) ;
megasas_dump_reg_set ( instance - > reg_set ) ;
2011-10-09 01:14:27 +00:00
return - ENODEV ;
2019-05-07 17:05:41 +00:00
}
2005-09-20 21:46:58 +00:00
case MFI_STATE_WAIT_HANDSHAKE :
/*
* Set the CLR bit in inbound doorbell
*/
2009-10-06 20:47:35 +00:00
if ( ( instance - > pdev - > device = =
2009-10-06 20:31:54 +00:00
PCI_DEVICE_ID_LSI_SAS0073SKINNY ) | |
( instance - > pdev - > device = =
2010-12-21 21:34:31 +00:00
PCI_DEVICE_ID_LSI_SAS0071SKINNY ) | |
2017-10-19 09:48:50 +00:00
( instance - > adapter_type ! = MFI_SERIES ) )
2009-10-06 20:31:54 +00:00
writel (
MFI_INIT_CLEAR_HANDSHAKE | MFI_INIT_HOTPLUG ,
2010-12-21 21:34:31 +00:00
& instance - > reg_set - > doorbell ) ;
2015-10-15 08:09:34 +00:00
else
2009-10-06 20:31:54 +00:00
writel (
MFI_INIT_CLEAR_HANDSHAKE | MFI_INIT_HOTPLUG ,
& instance - > reg_set - > inbound_doorbell ) ;
2005-09-20 21:46:58 +00:00
2009-10-06 20:52:20 +00:00
max_wait = MEGASAS_RESET_WAIT_TIME ;
2005-09-20 21:46:58 +00:00
break ;
2006-10-03 19:28:49 +00:00
case MFI_STATE_BOOT_MESSAGE_PENDING :
2009-10-06 20:31:54 +00:00
if ( ( instance - > pdev - > device = =
2010-12-21 21:34:31 +00:00
PCI_DEVICE_ID_LSI_SAS0073SKINNY ) | |
( instance - > pdev - > device = =
PCI_DEVICE_ID_LSI_SAS0071SKINNY ) | |
2017-10-19 09:48:50 +00:00
( instance - > adapter_type ! = MFI_SERIES ) )
2009-10-06 20:31:54 +00:00
writel ( MFI_INIT_HOTPLUG ,
2010-12-21 21:34:31 +00:00
& instance - > reg_set - > doorbell ) ;
2015-10-15 08:09:34 +00:00
else
2009-10-06 20:31:54 +00:00
writel ( MFI_INIT_HOTPLUG ,
& instance - > reg_set - > inbound_doorbell ) ;
2006-10-03 19:28:49 +00:00
2009-10-06 20:52:20 +00:00
max_wait = MEGASAS_RESET_WAIT_TIME ;
2006-10-03 19:28:49 +00:00
break ;
2005-09-20 21:46:58 +00:00
case MFI_STATE_OPERATIONAL :
/*
2006-10-03 19:28:49 +00:00
* Bring it to READY state ; assuming max wait 10 secs
2005-09-20 21:46:58 +00:00
*/
2013-05-22 07:04:14 +00:00
instance - > instancet - > disable_intr ( instance ) ;
2009-10-06 20:31:54 +00:00
if ( ( instance - > pdev - > device = =
PCI_DEVICE_ID_LSI_SAS0073SKINNY ) | |
( instance - > pdev - > device = =
2010-12-21 21:34:31 +00:00
PCI_DEVICE_ID_LSI_SAS0071SKINNY ) | |
2017-10-19 09:48:50 +00:00
( instance - > adapter_type ! = MFI_SERIES ) ) {
2009-10-06 20:31:54 +00:00
writel ( MFI_RESET_FLAGS ,
2010-12-21 21:34:31 +00:00
& instance - > reg_set - > doorbell ) ;
2015-10-15 08:09:34 +00:00
2017-10-19 09:48:50 +00:00
if ( instance - > adapter_type ! = MFI_SERIES ) {
2010-12-21 21:34:31 +00:00
for ( i = 0 ; i < ( 10 * 1000 ) ; i + = 20 ) {
2018-12-17 08:47:40 +00:00
if ( megasas_readl (
instance ,
2010-12-21 21:34:31 +00:00
& instance - >
reg_set - >
doorbell ) & 1 )
msleep ( 20 ) ;
else
break ;
}
}
2009-10-06 20:31:54 +00:00
} else
writel ( MFI_RESET_FLAGS ,
& instance - > reg_set - > inbound_doorbell ) ;
2005-09-20 21:46:58 +00:00
2009-10-06 20:52:20 +00:00
max_wait = MEGASAS_RESET_WAIT_TIME ;
2005-09-20 21:46:58 +00:00
break ;
case MFI_STATE_UNDEFINED :
/*
* This state should not last for more than 2 seconds
*/
2009-10-06 20:52:20 +00:00
max_wait = MEGASAS_RESET_WAIT_TIME ;
2005-09-20 21:46:58 +00:00
break ;
case MFI_STATE_BB_INIT :
2009-10-06 20:52:20 +00:00
max_wait = MEGASAS_RESET_WAIT_TIME ;
2005-09-20 21:46:58 +00:00
break ;
case MFI_STATE_FW_INIT :
2009-10-06 20:52:20 +00:00
max_wait = MEGASAS_RESET_WAIT_TIME ;
2005-09-20 21:46:58 +00:00
break ;
case MFI_STATE_FW_INIT_2 :
2009-10-06 20:52:20 +00:00
max_wait = MEGASAS_RESET_WAIT_TIME ;
2005-09-20 21:46:58 +00:00
break ;
case MFI_STATE_DEVICE_SCAN :
2009-10-06 20:52:20 +00:00
max_wait = MEGASAS_RESET_WAIT_TIME ;
2005-09-20 21:46:58 +00:00
break ;
case MFI_STATE_FLUSH_CACHE :
2009-10-06 20:52:20 +00:00
max_wait = MEGASAS_RESET_WAIT_TIME ;
2005-09-20 21:46:58 +00:00
break ;
default :
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " Unknown state 0x%x \n " ,
2005-09-20 21:46:58 +00:00
fw_state ) ;
2019-05-07 17:05:41 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " System Register set: \n " ) ;
megasas_dump_reg_set ( instance - > reg_set ) ;
2005-09-20 21:46:58 +00:00
return - ENODEV ;
}
/*
* The cur_state should not last for more than max_wait secs
*/
2019-03-01 14:46:28 +00:00
for ( i = 0 ; i < max_wait * 50 ; i + + ) {
2014-04-01 11:59:50 +00:00
curr_abs_state = instance - > instancet - >
2018-12-17 08:47:39 +00:00
read_fw_status_reg ( instance ) ;
2005-09-20 21:46:58 +00:00
2009-10-06 20:52:20 +00:00
if ( abs_state = = curr_abs_state ) {
2019-03-01 14:46:28 +00:00
msleep ( 20 ) ;
2005-09-20 21:46:58 +00:00
} else
break ;
}
/*
* Return error if fw_state hasn ' t changed after max_wait
*/
2009-10-06 20:52:20 +00:00
if ( curr_abs_state = = abs_state ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " FW state [%d] hasn't changed "
2005-09-20 21:46:58 +00:00
" in %d secs \n " , fw_state , max_wait ) ;
2019-05-07 17:05:41 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " System Register set: \n " ) ;
megasas_dump_reg_set ( instance - > reg_set ) ;
2005-09-20 21:46:58 +00:00
return - ENODEV ;
}
2014-04-01 11:59:50 +00:00
abs_state = curr_abs_state ;
fw_state = curr_abs_state & MFI_STATE_MASK ;
2010-09-23 02:36:29 +00:00
}
2015-07-07 20:52:34 +00:00
dev_info ( & instance - > pdev - > dev , " FW now in Ready state \n " ) ;
2005-09-20 21:46:58 +00:00
return 0 ;
}
/**
* megasas_teardown_frame_pool - Destroy the cmd frame DMA pool
* @ instance : Adapter soft state
*/
static void megasas_teardown_frame_pool ( struct megasas_instance * instance )
{
int i ;
2017-02-10 08:59:25 +00:00
u16 max_cmd = instance - > max_mfi_cmds ;
2005-09-20 21:46:58 +00:00
struct megasas_cmd * cmd ;
if ( ! instance - > frame_dma_pool )
return ;
/*
* Return all frames to pool
*/
for ( i = 0 ; i < max_cmd ; i + + ) {
cmd = instance - > cmd_list [ i ] ;
if ( cmd - > frame )
2017-07-06 08:13:06 +00:00
dma_pool_free ( instance - > frame_dma_pool , cmd - > frame ,
2005-09-20 21:46:58 +00:00
cmd - > frame_phys_addr ) ;
if ( cmd - > sense )
2017-07-06 08:13:06 +00:00
dma_pool_free ( instance - > sense_dma_pool , cmd - > sense ,
2005-09-20 21:46:58 +00:00
cmd - > sense_phys_addr ) ;
}
/*
* Now destroy the pool itself
*/
2017-07-06 08:13:06 +00:00
dma_pool_destroy ( instance - > frame_dma_pool ) ;
dma_pool_destroy ( instance - > sense_dma_pool ) ;
2005-09-20 21:46:58 +00:00
instance - > frame_dma_pool = NULL ;
instance - > sense_dma_pool = NULL ;
}
/**
* megasas_create_frame_pool - Creates DMA pool for cmd frames
* @ instance : Adapter soft state
*
* Each command packet has an embedded DMA memory buffer that is used for
* filling MFI frame and the SG list that immediately follows the frame . This
* function creates those DMA memory buffers for each command packet by using
* PCI pool facility .
*/
static int megasas_create_frame_pool ( struct megasas_instance * instance )
{
int i ;
2017-02-10 08:59:25 +00:00
u16 max_cmd ;
2005-09-20 21:46:58 +00:00
u32 frame_count ;
struct megasas_cmd * cmd ;
2010-12-21 21:34:31 +00:00
max_cmd = instance - > max_mfi_cmds ;
2005-09-20 21:46:58 +00:00
/*
2015-01-05 14:35:58 +00:00
* For MFI controllers .
* max_num_sge = 60
* max_sge_sz = 16 byte ( sizeof megasas_sge_skinny )
* Total 960 byte ( 15 MFI frame of 64 byte )
*
* Fusion adapter require only 3 extra frame .
* max_num_sge = 16 ( defined as MAX_IOCTL_SGE )
* max_sge_sz = 12 byte ( sizeof megasas_sge64 )
* Total 192 byte ( 3 MFI frame of 64 byte )
2005-09-20 21:46:58 +00:00
*/
2017-10-19 09:48:50 +00:00
frame_count = ( instance - > adapter_type = = MFI_SERIES ) ?
( 15 + 1 ) : ( 3 + 1 ) ;
2017-02-10 08:59:28 +00:00
instance - > mfi_frame_size = MEGAMFI_FRAME_SIZE * frame_count ;
2005-09-20 21:46:58 +00:00
/*
* Use DMA pool facility provided by PCI layer
*/
2017-07-06 08:13:06 +00:00
instance - > frame_dma_pool = dma_pool_create ( " megasas frame pool " ,
& instance - > pdev - > dev ,
instance - > mfi_frame_size , 256 , 0 ) ;
2005-09-20 21:46:58 +00:00
if ( ! instance - > frame_dma_pool ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " failed to setup frame pool \n " ) ;
2005-09-20 21:46:58 +00:00
return - ENOMEM ;
}
2017-07-06 08:13:06 +00:00
instance - > sense_dma_pool = dma_pool_create ( " megasas sense pool " ,
& instance - > pdev - > dev , 128 ,
4 , 0 ) ;
2005-09-20 21:46:58 +00:00
if ( ! instance - > sense_dma_pool ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " failed to setup sense pool \n " ) ;
2005-09-20 21:46:58 +00:00
2017-07-06 08:13:06 +00:00
dma_pool_destroy ( instance - > frame_dma_pool ) ;
2005-09-20 21:46:58 +00:00
instance - > frame_dma_pool = NULL ;
return - ENOMEM ;
}
/*
* Allocate and attach a frame to each of the commands in cmd_list .
* By making cmd - > index as the context instead of the & cmd , we can
* always use 32 bit context regardless of the architecture
*/
for ( i = 0 ; i < max_cmd ; i + + ) {
cmd = instance - > cmd_list [ i ] ;
2018-02-15 16:25:06 +00:00
cmd - > frame = dma_pool_zalloc ( instance - > frame_dma_pool ,
2005-09-20 21:46:58 +00:00
GFP_KERNEL , & cmd - > frame_phys_addr ) ;
2017-07-06 08:13:06 +00:00
cmd - > sense = dma_pool_alloc ( instance - > sense_dma_pool ,
2005-09-20 21:46:58 +00:00
GFP_KERNEL , & cmd - > sense_phys_addr ) ;
/*
* megasas_teardown_frame_pool ( ) takes care of freeing
* whatever has been allocated
*/
if ( ! cmd - > frame | | ! cmd - > sense ) {
2017-07-06 08:13:06 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " dma_pool_alloc failed \n " ) ;
2005-09-20 21:46:58 +00:00
megasas_teardown_frame_pool ( instance ) ;
return - ENOMEM ;
}
2013-09-06 10:20:52 +00:00
cmd - > frame - > io . context = cpu_to_le32 ( cmd - > index ) ;
2009-10-06 20:50:17 +00:00
cmd - > frame - > io . pad_0 = 0 ;
2017-10-19 09:48:50 +00:00
if ( ( instance - > adapter_type = = MFI_SERIES ) & & reset_devices )
2011-10-09 01:15:19 +00:00
cmd - > frame - > hdr . cmd = MFI_CMD_INVALID ;
2005-09-20 21:46:58 +00:00
}
return 0 ;
}
/**
* megasas_free_cmds - Free all the cmds in the free cmd pool
* @ instance : Adapter soft state
*/
2010-12-21 21:34:31 +00:00
void megasas_free_cmds ( struct megasas_instance * instance )
2005-09-20 21:46:58 +00:00
{
int i ;
2015-07-07 20:52:45 +00:00
2005-09-20 21:46:58 +00:00
/* First free the MFI frame pool */
megasas_teardown_frame_pool ( instance ) ;
/* Free all the commands in the cmd_list */
2010-12-21 21:34:31 +00:00
for ( i = 0 ; i < instance - > max_mfi_cmds ; i + + )
2005-09-20 21:46:58 +00:00
kfree ( instance - > cmd_list [ i ] ) ;
/* Free the cmd_list buffer itself */
kfree ( instance - > cmd_list ) ;
instance - > cmd_list = NULL ;
INIT_LIST_HEAD ( & instance - > cmd_pool ) ;
}
/**
* megasas_alloc_cmds - Allocates the command packets
* @ instance : Adapter soft state
*
* Each command that is issued to the FW , whether IO commands from the OS or
* internal commands like IOCTLs , are wrapped in local data structure called
* megasas_cmd . The frame embedded in this megasas_cmd is actually issued to
* the FW .
*
* Each frame has a 32 - bit field called context ( tag ) . This context is used
* to get back the megasas_cmd from the frame when a frame gets completed in
* the ISR . Typically the address of the megasas_cmd itself would be used as
* the context . But we wanted to keep the differences between 32 and 64 bit
* systems to the mininum . We always use 32 bit integers for the context . In
* this driver , the 32 bit values are the indices into an array cmd_list .
* This array is used only to look up the megasas_cmd given the context . The
* free commands themselves are maintained in a linked list called cmd_pool .
*/
2010-12-21 21:34:31 +00:00
int megasas_alloc_cmds ( struct megasas_instance * instance )
2005-09-20 21:46:58 +00:00
{
int i ;
int j ;
2017-02-10 08:59:25 +00:00
u16 max_cmd ;
2005-09-20 21:46:58 +00:00
struct megasas_cmd * cmd ;
2010-12-21 21:34:31 +00:00
max_cmd = instance - > max_mfi_cmds ;
2005-09-20 21:46:58 +00:00
/*
* instance - > cmd_list is an array of struct megasas_cmd pointers .
* Allocate the dynamic array first and then allocate individual
* commands .
*/
some kmalloc/memset ->kzalloc (tree wide)
Transform some calls to kmalloc/memset to a single kzalloc (or kcalloc).
Here is a short excerpt of the semantic patch performing
this transformation:
@@
type T2;
expression x;
identifier f,fld;
expression E;
expression E1,E2;
expression e1,e2,e3,y;
statement S;
@@
x =
- kmalloc
+ kzalloc
(E1,E2)
... when != \(x->fld=E;\|y=f(...,x,...);\|f(...,x,...);\|x=E;\|while(...) S\|for(e1;e2;e3) S\)
- memset((T2)x,0,E1);
@@
expression E1,E2,E3;
@@
- kzalloc(E1 * E2,E3)
+ kcalloc(E1,E2,E3)
[akpm@linux-foundation.org: get kcalloc args the right way around]
Signed-off-by: Yoann Padioleau <padator@wanadoo.fr>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Acked-by: Russell King <rmk@arm.linux.org.uk>
Cc: Bryan Wu <bryan.wu@analog.com>
Acked-by: Jiri Slaby <jirislaby@gmail.com>
Cc: Dave Airlie <airlied@linux.ie>
Acked-by: Roland Dreier <rolandd@cisco.com>
Cc: Jiri Kosina <jkosina@suse.cz>
Acked-by: Dmitry Torokhov <dtor@mail.ru>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Mauro Carvalho Chehab <mchehab@infradead.org>
Acked-by: Pierre Ossman <drzeus-list@drzeus.cx>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: "David S. Miller" <davem@davemloft.net>
Acked-by: Greg KH <greg@kroah.com>
Cc: James Bottomley <James.Bottomley@steeleye.com>
Cc: "Antonino A. Daplas" <adaplas@pol.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 08:49:03 +00:00
instance - > cmd_list = kcalloc ( max_cmd , sizeof ( struct megasas_cmd * ) , GFP_KERNEL ) ;
2005-09-20 21:46:58 +00:00
if ( ! instance - > cmd_list ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " out of memory \n " ) ;
2005-09-20 21:46:58 +00:00
return - ENOMEM ;
}
2010-12-21 21:34:31 +00:00
memset ( instance - > cmd_list , 0 , sizeof ( struct megasas_cmd * ) * max_cmd ) ;
2005-09-20 21:46:58 +00:00
for ( i = 0 ; i < max_cmd ; i + + ) {
instance - > cmd_list [ i ] = kmalloc ( sizeof ( struct megasas_cmd ) ,
GFP_KERNEL ) ;
if ( ! instance - > cmd_list [ i ] ) {
for ( j = 0 ; j < i ; j + + )
kfree ( instance - > cmd_list [ j ] ) ;
kfree ( instance - > cmd_list ) ;
instance - > cmd_list = NULL ;
return - ENOMEM ;
}
}
for ( i = 0 ; i < max_cmd ; i + + ) {
cmd = instance - > cmd_list [ i ] ;
memset ( cmd , 0 , sizeof ( struct megasas_cmd ) ) ;
cmd - > index = i ;
2010-09-23 02:36:29 +00:00
cmd - > scmd = NULL ;
2005-09-20 21:46:58 +00:00
cmd - > instance = instance ;
list_add_tail ( & cmd - > list , & instance - > cmd_pool ) ;
}
/*
* Create a frame pool and assign one frame to each cmd
*/
if ( megasas_create_frame_pool ( instance ) ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " Error creating frame DMA pool \n " ) ;
2005-09-20 21:46:58 +00:00
megasas_free_cmds ( instance ) ;
2019-02-15 11:50:27 +00:00
return - ENOMEM ;
2005-09-20 21:46:58 +00:00
}
return 0 ;
}
2016-01-28 15:34:23 +00:00
/*
* dcmd_timeout_ocr_possible - Check if OCR is possible based on Driver / FW state .
* @ instance : Adapter soft state
*
* Return 0 for only Fusion adapter , if driver load / unload is not in progress
* or FW is not under OCR .
*/
inline int
dcmd_timeout_ocr_possible ( struct megasas_instance * instance ) {
2017-10-19 09:48:50 +00:00
if ( instance - > adapter_type = = MFI_SERIES )
2016-01-28 15:34:23 +00:00
return KILL_ADAPTER ;
else if ( instance - > unload | |
test_bit ( MEGASAS_FUSION_IN_RESET , & instance - > reset_flags ) )
return IGNORE_TIMEOUT ;
else
return INITIATE_OCR ;
}
2017-02-10 08:59:10 +00:00
static void
megasas_get_pd_info ( struct megasas_instance * instance , struct scsi_device * sdev )
2016-01-28 15:34:26 +00:00
{
int ret ;
struct megasas_cmd * cmd ;
struct megasas_dcmd_frame * dcmd ;
2017-02-10 08:59:10 +00:00
struct MR_PRIV_DEVICE * mr_device_priv_data ;
u16 device_id = 0 ;
device_id = ( sdev - > channel * MEGASAS_MAX_DEV_PER_CHANNEL ) + sdev - > id ;
2016-01-28 15:34:26 +00:00
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd ) {
dev_err ( & instance - > pdev - > dev , " Failed to get cmd %s \n " , __func__ ) ;
2017-02-10 08:59:10 +00:00
return ;
2016-01-28 15:34:26 +00:00
}
dcmd = & cmd - > frame - > dcmd ;
memset ( instance - > pd_info , 0 , sizeof ( * instance - > pd_info ) ) ;
memset ( dcmd - > mbox . b , 0 , MFI_MBOX_SIZE ) ;
dcmd - > mbox . s [ 0 ] = cpu_to_le16 ( device_id ) ;
dcmd - > cmd = MFI_CMD_DCMD ;
dcmd - > cmd_status = 0xFF ;
dcmd - > sge_count = 1 ;
2017-10-19 09:49:05 +00:00
dcmd - > flags = MFI_FRAME_DIR_READ ;
2016-01-28 15:34:26 +00:00
dcmd - > timeout = 0 ;
dcmd - > pad_0 = 0 ;
dcmd - > data_xfer_len = cpu_to_le32 ( sizeof ( struct MR_PD_INFO ) ) ;
dcmd - > opcode = cpu_to_le32 ( MR_DCMD_PD_GET_INFO ) ;
2017-10-19 09:49:05 +00:00
megasas_set_dma_settings ( instance , dcmd , instance - > pd_info_h ,
sizeof ( struct MR_PD_INFO ) ) ;
2016-01-28 15:34:26 +00:00
2017-10-19 09:48:50 +00:00
if ( ( instance - > adapter_type ! = MFI_SERIES ) & &
! instance - > mask_interrupts )
2016-01-28 15:34:26 +00:00
ret = megasas_issue_blocked_cmd ( instance , cmd , MFI_IO_TIMEOUT_SECS ) ;
else
ret = megasas_issue_polled ( instance , cmd ) ;
switch ( ret ) {
case DCMD_SUCCESS :
2017-02-10 08:59:10 +00:00
mr_device_priv_data = sdev - > hostdata ;
le16_to_cpus ( ( u16 * ) & instance - > pd_info - > state . ddf . pdType ) ;
mr_device_priv_data - > interface_type =
2016-01-28 15:34:26 +00:00
instance - > pd_info - > state . ddf . pdType . intf ;
break ;
case DCMD_TIMEOUT :
switch ( dcmd_timeout_ocr_possible ( instance ) ) {
case INITIATE_OCR :
cmd - > flags | = DRV_DCMD_SKIP_REFIRE ;
2019-06-25 11:04:23 +00:00
mutex_unlock ( & instance - > reset_mutex ) ;
2016-01-28 15:34:26 +00:00
megasas_reset_fusion ( instance - > host ,
MFI_IO_TIMEOUT_OCR ) ;
2019-06-25 11:04:23 +00:00
mutex_lock ( & instance - > reset_mutex ) ;
2016-01-28 15:34:26 +00:00
break ;
case KILL_ADAPTER :
megaraid_sas_kill_hba ( instance ) ;
break ;
case IGNORE_TIMEOUT :
dev_info ( & instance - > pdev - > dev , " Ignore DCMD timeout: %s %d \n " ,
__func__ , __LINE__ ) ;
break ;
}
break ;
}
if ( ret ! = DCMD_TIMEOUT )
megasas_return_cmd ( instance , cmd ) ;
2017-02-10 08:59:10 +00:00
return ;
2016-01-28 15:34:26 +00:00
}
2009-10-06 20:27:54 +00:00
/*
* megasas_get_pd_list_info - Returns FW ' s pd_list structure
* @ instance : Adapter soft state
* @ pd_list : pd_list structure
*
* Issues an internal command ( DCMD ) to get the FW ' s controller PD
* list structure . This information is mainly used to find out SYSTEM
* supported by the FW .
*/
static int
megasas_get_pd_list ( struct megasas_instance * instance )
{
int ret = 0 , pd_index = 0 ;
struct megasas_cmd * cmd ;
struct megasas_dcmd_frame * dcmd ;
struct MR_PD_LIST * ci ;
struct MR_PD_ADDRESS * pd_addr ;
2016-07-08 10:30:16 +00:00
if ( instance - > pd_list_not_supported ) {
dev_info ( & instance - > pdev - > dev , " MR_DCMD_PD_LIST_QUERY "
" not supported by firmware \n " ) ;
return ret ;
}
2017-10-19 09:48:56 +00:00
ci = instance - > pd_list_buf ;
2009-10-06 20:27:54 +00:00
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " (get_pd_list): Failed to get cmd \n " ) ;
2009-10-06 20:27:54 +00:00
return - ENOMEM ;
}
dcmd = & cmd - > frame - > dcmd ;
memset ( ci , 0 , sizeof ( * ci ) ) ;
memset ( dcmd - > mbox . b , 0 , MFI_MBOX_SIZE ) ;
dcmd - > mbox . b [ 0 ] = MR_PD_QUERY_TYPE_EXPOSED_TO_HOST ;
dcmd - > mbox . b [ 1 ] = 0 ;
dcmd - > cmd = MFI_CMD_DCMD ;
2015-05-06 13:31:02 +00:00
dcmd - > cmd_status = MFI_STAT_INVALID_STATUS ;
2009-10-06 20:27:54 +00:00
dcmd - > sge_count = 1 ;
2017-10-19 09:49:05 +00:00
dcmd - > flags = MFI_FRAME_DIR_READ ;
2009-10-06 20:27:54 +00:00
dcmd - > timeout = 0 ;
2009-12-06 15:24:21 +00:00
dcmd - > pad_0 = 0 ;
2013-09-06 10:20:52 +00:00
dcmd - > data_xfer_len = cpu_to_le32 ( MEGASAS_MAX_PD * sizeof ( struct MR_PD_LIST ) ) ;
dcmd - > opcode = cpu_to_le32 ( MR_DCMD_PD_LIST_QUERY ) ;
2009-10-06 20:27:54 +00:00
2017-10-19 09:49:05 +00:00
megasas_set_dma_settings ( instance , dcmd , instance - > pd_list_buf_h ,
( MEGASAS_MAX_PD * sizeof ( struct MR_PD_LIST ) ) ) ;
2009-10-06 20:27:54 +00:00
2017-10-19 09:48:50 +00:00
if ( ( instance - > adapter_type ! = MFI_SERIES ) & &
! instance - > mask_interrupts )
2014-09-12 13:27:58 +00:00
ret = megasas_issue_blocked_cmd ( instance , cmd ,
2016-01-28 15:34:23 +00:00
MFI_IO_TIMEOUT_SECS ) ;
2014-09-12 13:27:58 +00:00
else
ret = megasas_issue_polled ( instance , cmd ) ;
2009-10-06 20:27:54 +00:00
2016-01-28 15:34:23 +00:00
switch ( ret ) {
case DCMD_FAILED :
2016-03-10 10:14:37 +00:00
dev_info ( & instance - > pdev - > dev , " MR_DCMD_PD_LIST_QUERY "
" failed/not supported by firmware \n " ) ;
2017-10-19 09:48:50 +00:00
if ( instance - > adapter_type ! = MFI_SERIES )
2016-03-10 10:14:37 +00:00
megaraid_sas_kill_hba ( instance ) ;
else
instance - > pd_list_not_supported = 1 ;
2016-01-28 15:34:23 +00:00
break ;
case DCMD_TIMEOUT :
2009-10-06 20:27:54 +00:00
2016-01-28 15:34:23 +00:00
switch ( dcmd_timeout_ocr_possible ( instance ) ) {
case INITIATE_OCR :
cmd - > flags | = DRV_DCMD_SKIP_REFIRE ;
/*
* DCMD failed from AEN path .
* AEN path already hold reset_mutex to avoid PCI access
* while OCR is in progress .
*/
mutex_unlock ( & instance - > reset_mutex ) ;
megasas_reset_fusion ( instance - > host ,
MFI_IO_TIMEOUT_OCR ) ;
mutex_lock ( & instance - > reset_mutex ) ;
break ;
case KILL_ADAPTER :
megaraid_sas_kill_hba ( instance ) ;
break ;
case IGNORE_TIMEOUT :
dev_info ( & instance - > pdev - > dev , " Ignore DCMD timeout: %s %d \n " ,
__func__ , __LINE__ ) ;
break ;
}
2009-10-06 20:27:54 +00:00
2016-01-28 15:34:23 +00:00
break ;
case DCMD_SUCCESS :
pd_addr = ci - > addr ;
2019-05-07 17:05:47 +00:00
if ( megasas_dbg_lvl & LD_PD_DEBUG )
dev_info ( & instance - > pdev - > dev , " %s, sysPD count: 0x%x \n " ,
__func__ , le32_to_cpu ( ci - > count ) ) ;
2016-01-28 15:34:23 +00:00
if ( ( le32_to_cpu ( ci - > count ) >
( MEGASAS_MAX_PD_CHANNELS * MEGASAS_MAX_DEV_PER_CHANNEL ) ) )
break ;
2009-10-06 20:27:54 +00:00
2013-10-18 07:20:37 +00:00
memset ( instance - > local_pd_list , 0 ,
2016-01-28 15:34:23 +00:00
MEGASAS_MAX_PD * sizeof ( struct megasas_pd_list ) ) ;
2009-10-06 20:27:54 +00:00
2013-09-06 10:20:52 +00:00
for ( pd_index = 0 ; pd_index < le32_to_cpu ( ci - > count ) ; pd_index + + ) {
2013-10-18 07:20:37 +00:00
instance - > local_pd_list [ le16_to_cpu ( pd_addr - > deviceId ) ] . tid =
2016-01-28 15:34:23 +00:00
le16_to_cpu ( pd_addr - > deviceId ) ;
2013-10-18 07:20:37 +00:00
instance - > local_pd_list [ le16_to_cpu ( pd_addr - > deviceId ) ] . driveType =
2016-01-28 15:34:23 +00:00
pd_addr - > scsiDevType ;
2013-10-18 07:20:37 +00:00
instance - > local_pd_list [ le16_to_cpu ( pd_addr - > deviceId ) ] . driveState =
2016-01-28 15:34:23 +00:00
MR_PD_STATE_SYSTEM ;
2019-05-07 17:05:47 +00:00
if ( megasas_dbg_lvl & LD_PD_DEBUG )
dev_info ( & instance - > pdev - > dev ,
" PD%d: targetID: 0x%03x deviceType:0x%x \n " ,
pd_index , le16_to_cpu ( pd_addr - > deviceId ) ,
pd_addr - > scsiDevType ) ;
2009-10-06 20:27:54 +00:00
pd_addr + + ;
}
2016-01-28 15:34:23 +00:00
2013-10-18 07:20:37 +00:00
memcpy ( instance - > pd_list , instance - > local_pd_list ,
sizeof ( instance - > pd_list ) ) ;
2016-01-28 15:34:23 +00:00
break ;
2009-10-06 20:27:54 +00:00
}
2016-01-28 15:34:23 +00:00
if ( ret ! = DCMD_TIMEOUT )
megasas_return_cmd ( instance , cmd ) ;
2009-10-06 20:27:54 +00:00
return ret ;
}
2009-12-06 15:30:19 +00:00
/*
* megasas_get_ld_list_info - Returns FW ' s ld_list structure
* @ instance : Adapter soft state
* @ ld_list : ld_list structure
*
* Issues an internal command ( DCMD ) to get the FW ' s controller PD
* list structure . This information is mainly used to find out SYSTEM
* supported by the FW .
*/
static int
megasas_get_ld_list ( struct megasas_instance * instance )
{
int ret = 0 , ld_index = 0 , ids = 0 ;
struct megasas_cmd * cmd ;
struct megasas_dcmd_frame * dcmd ;
struct MR_LD_LIST * ci ;
dma_addr_t ci_h = 0 ;
2013-09-06 10:20:52 +00:00
u32 ld_count ;
2009-12-06 15:30:19 +00:00
2017-10-19 09:48:56 +00:00
ci = instance - > ld_list_buf ;
ci_h = instance - > ld_list_buf_h ;
2009-12-06 15:30:19 +00:00
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " megasas_get_ld_list: Failed to get cmd \n " ) ;
2009-12-06 15:30:19 +00:00
return - ENOMEM ;
}
dcmd = & cmd - > frame - > dcmd ;
memset ( ci , 0 , sizeof ( * ci ) ) ;
memset ( dcmd - > mbox . b , 0 , MFI_MBOX_SIZE ) ;
2014-09-12 13:27:33 +00:00
if ( instance - > supportmax256vd )
dcmd - > mbox . b [ 0 ] = 1 ;
2009-12-06 15:30:19 +00:00
dcmd - > cmd = MFI_CMD_DCMD ;
2015-05-06 13:31:02 +00:00
dcmd - > cmd_status = MFI_STAT_INVALID_STATUS ;
2009-12-06 15:30:19 +00:00
dcmd - > sge_count = 1 ;
2017-10-19 09:49:05 +00:00
dcmd - > flags = MFI_FRAME_DIR_READ ;
2009-12-06 15:30:19 +00:00
dcmd - > timeout = 0 ;
2013-09-06 10:20:52 +00:00
dcmd - > data_xfer_len = cpu_to_le32 ( sizeof ( struct MR_LD_LIST ) ) ;
dcmd - > opcode = cpu_to_le32 ( MR_DCMD_LD_GET_LIST ) ;
2009-12-06 15:30:19 +00:00
dcmd - > pad_0 = 0 ;
2017-10-19 09:49:05 +00:00
megasas_set_dma_settings ( instance , dcmd , ci_h ,
sizeof ( struct MR_LD_LIST ) ) ;
2017-10-19 09:48:50 +00:00
if ( ( instance - > adapter_type ! = MFI_SERIES ) & &
! instance - > mask_interrupts )
2014-09-12 13:27:58 +00:00
ret = megasas_issue_blocked_cmd ( instance , cmd ,
2016-01-28 15:34:23 +00:00
MFI_IO_TIMEOUT_SECS ) ;
2014-09-12 13:27:58 +00:00
else
ret = megasas_issue_polled ( instance , cmd ) ;
2013-09-06 10:20:52 +00:00
ld_count = le32_to_cpu ( ci - > ldCount ) ;
2016-01-28 15:34:23 +00:00
switch ( ret ) {
case DCMD_FAILED :
megaraid_sas_kill_hba ( instance ) ;
break ;
case DCMD_TIMEOUT :
switch ( dcmd_timeout_ocr_possible ( instance ) ) {
case INITIATE_OCR :
cmd - > flags | = DRV_DCMD_SKIP_REFIRE ;
/*
* DCMD failed from AEN path .
* AEN path already hold reset_mutex to avoid PCI access
* while OCR is in progress .
*/
mutex_unlock ( & instance - > reset_mutex ) ;
megasas_reset_fusion ( instance - > host ,
MFI_IO_TIMEOUT_OCR ) ;
mutex_lock ( & instance - > reset_mutex ) ;
break ;
case KILL_ADAPTER :
megaraid_sas_kill_hba ( instance ) ;
break ;
case IGNORE_TIMEOUT :
dev_info ( & instance - > pdev - > dev , " Ignore DCMD timeout: %s %d \n " ,
__func__ , __LINE__ ) ;
break ;
}
break ;
case DCMD_SUCCESS :
2019-05-07 17:05:47 +00:00
if ( megasas_dbg_lvl & LD_PD_DEBUG )
dev_info ( & instance - > pdev - > dev , " %s, LD count: 0x%x \n " ,
__func__ , ld_count ) ;
2016-01-28 15:34:23 +00:00
if ( ld_count > instance - > fw_supported_vd_count )
break ;
2009-12-06 15:30:19 +00:00
2014-09-12 13:27:33 +00:00
memset ( instance - > ld_ids , 0xff , MAX_LOGICAL_DRIVES_EXT ) ;
2009-12-06 15:30:19 +00:00
2013-09-06 10:20:52 +00:00
for ( ld_index = 0 ; ld_index < ld_count ; ld_index + + ) {
2009-12-06 15:30:19 +00:00
if ( ci - > ldList [ ld_index ] . state ! = 0 ) {
ids = ci - > ldList [ ld_index ] . ref . targetId ;
2016-01-28 15:34:23 +00:00
instance - > ld_ids [ ids ] = ci - > ldList [ ld_index ] . ref . targetId ;
2019-05-07 17:05:47 +00:00
if ( megasas_dbg_lvl & LD_PD_DEBUG )
dev_info ( & instance - > pdev - > dev ,
" LD%d: targetID: 0x%03x \n " ,
ld_index , ids ) ;
2009-12-06 15:30:19 +00:00
}
}
2016-01-28 15:34:23 +00:00
break ;
2009-12-06 15:30:19 +00:00
}
2016-01-28 15:34:23 +00:00
if ( ret ! = DCMD_TIMEOUT )
megasas_return_cmd ( instance , cmd ) ;
2009-12-06 15:30:19 +00:00
return ret ;
}
2013-09-06 22:27:14 +00:00
/**
* megasas_ld_list_query - Returns FW ' s ld_list structure
* @ instance : Adapter soft state
* @ ld_list : ld_list structure
*
* Issues an internal command ( DCMD ) to get the FW ' s controller PD
* list structure . This information is mainly used to find out SYSTEM
* supported by the FW .
*/
static int
megasas_ld_list_query ( struct megasas_instance * instance , u8 query_type )
{
int ret = 0 , ld_index = 0 , ids = 0 ;
struct megasas_cmd * cmd ;
struct megasas_dcmd_frame * dcmd ;
struct MR_LD_TARGETID_LIST * ci ;
dma_addr_t ci_h = 0 ;
2013-09-06 10:20:52 +00:00
u32 tgtid_count ;
2013-09-06 22:27:14 +00:00
2017-10-19 09:48:56 +00:00
ci = instance - > ld_targetid_list_buf ;
ci_h = instance - > ld_targetid_list_buf_h ;
2013-09-06 22:27:14 +00:00
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd ) {
2015-07-07 20:52:34 +00:00
dev_warn ( & instance - > pdev - > dev ,
" megasas_ld_list_query: Failed to get cmd \n " ) ;
2013-09-06 22:27:14 +00:00
return - ENOMEM ;
}
dcmd = & cmd - > frame - > dcmd ;
memset ( ci , 0 , sizeof ( * ci ) ) ;
memset ( dcmd - > mbox . b , 0 , MFI_MBOX_SIZE ) ;
dcmd - > mbox . b [ 0 ] = query_type ;
2014-09-12 13:27:33 +00:00
if ( instance - > supportmax256vd )
dcmd - > mbox . b [ 2 ] = 1 ;
2013-09-06 22:27:14 +00:00
dcmd - > cmd = MFI_CMD_DCMD ;
2015-05-06 13:31:02 +00:00
dcmd - > cmd_status = MFI_STAT_INVALID_STATUS ;
2013-09-06 22:27:14 +00:00
dcmd - > sge_count = 1 ;
2017-10-19 09:49:05 +00:00
dcmd - > flags = MFI_FRAME_DIR_READ ;
2013-09-06 22:27:14 +00:00
dcmd - > timeout = 0 ;
2013-09-06 10:20:52 +00:00
dcmd - > data_xfer_len = cpu_to_le32 ( sizeof ( struct MR_LD_TARGETID_LIST ) ) ;
dcmd - > opcode = cpu_to_le32 ( MR_DCMD_LD_LIST_QUERY ) ;
2013-09-06 22:27:14 +00:00
dcmd - > pad_0 = 0 ;
2017-10-19 09:49:05 +00:00
megasas_set_dma_settings ( instance , dcmd , ci_h ,
sizeof ( struct MR_LD_TARGETID_LIST ) ) ;
2017-10-19 09:48:50 +00:00
if ( ( instance - > adapter_type ! = MFI_SERIES ) & &
! instance - > mask_interrupts )
2016-01-28 15:34:23 +00:00
ret = megasas_issue_blocked_cmd ( instance , cmd , MFI_IO_TIMEOUT_SECS ) ;
2014-09-12 13:27:58 +00:00
else
ret = megasas_issue_polled ( instance , cmd ) ;
2013-09-06 22:27:14 +00:00
2016-01-28 15:34:23 +00:00
switch ( ret ) {
case DCMD_FAILED :
dev_info ( & instance - > pdev - > dev ,
" DCMD not supported by firmware - %s %d \n " ,
__func__ , __LINE__ ) ;
ret = megasas_get_ld_list ( instance ) ;
break ;
case DCMD_TIMEOUT :
switch ( dcmd_timeout_ocr_possible ( instance ) ) {
case INITIATE_OCR :
cmd - > flags | = DRV_DCMD_SKIP_REFIRE ;
/*
* DCMD failed from AEN path .
* AEN path already hold reset_mutex to avoid PCI access
* while OCR is in progress .
*/
mutex_unlock ( & instance - > reset_mutex ) ;
megasas_reset_fusion ( instance - > host ,
MFI_IO_TIMEOUT_OCR ) ;
mutex_lock ( & instance - > reset_mutex ) ;
break ;
case KILL_ADAPTER :
megaraid_sas_kill_hba ( instance ) ;
break ;
case IGNORE_TIMEOUT :
dev_info ( & instance - > pdev - > dev , " Ignore DCMD timeout: %s %d \n " ,
__func__ , __LINE__ ) ;
break ;
}
break ;
case DCMD_SUCCESS :
tgtid_count = le32_to_cpu ( ci - > count ) ;
2019-05-07 17:05:47 +00:00
if ( megasas_dbg_lvl & LD_PD_DEBUG )
dev_info ( & instance - > pdev - > dev , " %s, LD count: 0x%x \n " ,
__func__ , tgtid_count ) ;
2016-01-28 15:34:23 +00:00
if ( ( tgtid_count > ( instance - > fw_supported_vd_count ) ) )
break ;
2013-09-06 10:20:52 +00:00
2013-09-06 22:27:14 +00:00
memset ( instance - > ld_ids , 0xff , MEGASAS_MAX_LD_IDS ) ;
2013-09-06 10:20:52 +00:00
for ( ld_index = 0 ; ld_index < tgtid_count ; ld_index + + ) {
2013-09-06 22:27:14 +00:00
ids = ci - > targetId [ ld_index ] ;
instance - > ld_ids [ ids ] = ci - > targetId [ ld_index ] ;
2019-05-07 17:05:47 +00:00
if ( megasas_dbg_lvl & LD_PD_DEBUG )
dev_info ( & instance - > pdev - > dev , " LD%d: targetID: 0x%03x \n " ,
ld_index , ci - > targetId [ ld_index ] ) ;
2013-09-06 22:27:14 +00:00
}
2016-01-28 15:34:23 +00:00
break ;
2013-09-06 22:27:14 +00:00
}
2016-01-28 15:34:23 +00:00
if ( ret ! = DCMD_TIMEOUT )
megasas_return_cmd ( instance , cmd ) ;
2013-09-06 22:27:14 +00:00
return ret ;
}
2019-01-29 09:38:14 +00:00
/**
* dcmd . opcode - MR_DCMD_CTRL_DEVICE_LIST_GET
* dcmd . mbox - reserved
* dcmd . sge IN - ptr to return MR_HOST_DEVICE_LIST structure
* Desc : This DCMD will return the combined device list
* Status : MFI_STAT_OK - List returned successfully
* MFI_STAT_INVALID_CMD - Firmware support for the feature has been
* disabled
* @ instance : Adapter soft state
* @ is_probe : Driver probe check
* Return : 0 if DCMD succeeded
* non - zero if failed
*/
2019-04-13 14:21:38 +00:00
static int
2019-01-29 09:38:14 +00:00
megasas_host_device_list_query ( struct megasas_instance * instance ,
bool is_probe )
{
int ret , i , target_id ;
struct megasas_cmd * cmd ;
struct megasas_dcmd_frame * dcmd ;
struct MR_HOST_DEVICE_LIST * ci ;
u32 count ;
dma_addr_t ci_h ;
ci = instance - > host_device_list_buf ;
ci_h = instance - > host_device_list_buf_h ;
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd ) {
dev_warn ( & instance - > pdev - > dev ,
" %s: failed to get cmd \n " ,
__func__ ) ;
return - ENOMEM ;
}
dcmd = & cmd - > frame - > dcmd ;
memset ( ci , 0 , sizeof ( * ci ) ) ;
memset ( dcmd - > mbox . b , 0 , MFI_MBOX_SIZE ) ;
dcmd - > mbox . b [ 0 ] = is_probe ? 0 : 1 ;
dcmd - > cmd = MFI_CMD_DCMD ;
dcmd - > cmd_status = MFI_STAT_INVALID_STATUS ;
dcmd - > sge_count = 1 ;
dcmd - > flags = MFI_FRAME_DIR_READ ;
dcmd - > timeout = 0 ;
dcmd - > pad_0 = 0 ;
dcmd - > data_xfer_len = cpu_to_le32 ( HOST_DEVICE_LIST_SZ ) ;
dcmd - > opcode = cpu_to_le32 ( MR_DCMD_CTRL_DEVICE_LIST_GET ) ;
megasas_set_dma_settings ( instance , dcmd , ci_h , HOST_DEVICE_LIST_SZ ) ;
if ( ! instance - > mask_interrupts ) {
ret = megasas_issue_blocked_cmd ( instance , cmd ,
MFI_IO_TIMEOUT_SECS ) ;
} else {
ret = megasas_issue_polled ( instance , cmd ) ;
cmd - > flags | = DRV_DCMD_SKIP_REFIRE ;
}
switch ( ret ) {
case DCMD_SUCCESS :
/* Fill the internal pd_list and ld_ids array based on
* targetIds returned by FW
*/
count = le32_to_cpu ( ci - > count ) ;
2019-06-25 11:04:27 +00:00
if ( count > ( MEGASAS_MAX_PD + MAX_LOGICAL_DRIVES_EXT ) )
break ;
2019-05-07 17:05:47 +00:00
if ( megasas_dbg_lvl & LD_PD_DEBUG )
dev_info ( & instance - > pdev - > dev , " %s, Device count: 0x%x \n " ,
__func__ , count ) ;
2019-01-29 09:38:14 +00:00
memset ( instance - > local_pd_list , 0 ,
MEGASAS_MAX_PD * sizeof ( struct megasas_pd_list ) ) ;
memset ( instance - > ld_ids , 0xff , MAX_LOGICAL_DRIVES_EXT ) ;
for ( i = 0 ; i < count ; i + + ) {
target_id = le16_to_cpu ( ci - > host_device_list [ i ] . target_id ) ;
if ( ci - > host_device_list [ i ] . flags . u . bits . is_sys_pd ) {
instance - > local_pd_list [ target_id ] . tid = target_id ;
instance - > local_pd_list [ target_id ] . driveType =
ci - > host_device_list [ i ] . scsi_type ;
instance - > local_pd_list [ target_id ] . driveState =
MR_PD_STATE_SYSTEM ;
2019-05-07 17:05:47 +00:00
if ( megasas_dbg_lvl & LD_PD_DEBUG )
dev_info ( & instance - > pdev - > dev ,
" Device %d: PD targetID: 0x%03x deviceType:0x%x \n " ,
i , target_id , ci - > host_device_list [ i ] . scsi_type ) ;
2019-01-29 09:38:14 +00:00
} else {
instance - > ld_ids [ target_id ] = target_id ;
2019-05-07 17:05:47 +00:00
if ( megasas_dbg_lvl & LD_PD_DEBUG )
dev_info ( & instance - > pdev - > dev ,
" Device %d: LD targetID: 0x%03x \n " ,
i , target_id ) ;
2019-01-29 09:38:14 +00:00
}
}
memcpy ( instance - > pd_list , instance - > local_pd_list ,
sizeof ( instance - > pd_list ) ) ;
break ;
case DCMD_TIMEOUT :
switch ( dcmd_timeout_ocr_possible ( instance ) ) {
case INITIATE_OCR :
cmd - > flags | = DRV_DCMD_SKIP_REFIRE ;
2019-06-25 11:04:23 +00:00
mutex_unlock ( & instance - > reset_mutex ) ;
2019-01-29 09:38:14 +00:00
megasas_reset_fusion ( instance - > host ,
MFI_IO_TIMEOUT_OCR ) ;
2019-06-25 11:04:23 +00:00
mutex_lock ( & instance - > reset_mutex ) ;
2019-01-29 09:38:14 +00:00
break ;
case KILL_ADAPTER :
megaraid_sas_kill_hba ( instance ) ;
break ;
case IGNORE_TIMEOUT :
dev_info ( & instance - > pdev - > dev , " Ignore DCMD timeout: %s %d \n " ,
__func__ , __LINE__ ) ;
break ;
}
break ;
case DCMD_FAILED :
dev_err ( & instance - > pdev - > dev ,
" %s: MR_DCMD_CTRL_DEVICE_LIST_GET failed \n " ,
__func__ ) ;
break ;
}
if ( ret ! = DCMD_TIMEOUT )
megasas_return_cmd ( instance , cmd ) ;
return ret ;
}
2014-11-17 09:54:13 +00:00
/*
* megasas_update_ext_vd_details : Update details w . r . t Extended VD
* instance : Controller ' s instance
*/
static void megasas_update_ext_vd_details ( struct megasas_instance * instance )
{
struct fusion_context * fusion ;
2017-01-10 23:20:48 +00:00
u32 ventura_map_sz = 0 ;
2014-11-17 09:54:13 +00:00
fusion = instance - > ctrl_context ;
/* For MFI based controllers return dummy success */
if ( ! fusion )
return ;
instance - > supportmax256vd =
2017-10-19 09:48:57 +00:00
instance - > ctrl_info_buf - > adapterOperations3 . supportMaxExtLDs ;
2014-11-17 09:54:13 +00:00
/* Below is additional check to address future FW enhancement */
2017-10-19 09:48:57 +00:00
if ( instance - > ctrl_info_buf - > max_lds > 64 )
2014-11-17 09:54:13 +00:00
instance - > supportmax256vd = 1 ;
instance - > drv_supported_vd_count = MEGASAS_MAX_LD_CHANNELS
* MEGASAS_MAX_DEV_PER_CHANNEL ;
instance - > drv_supported_pd_count = MEGASAS_MAX_PD_CHANNELS
* MEGASAS_MAX_DEV_PER_CHANNEL ;
if ( instance - > supportmax256vd ) {
instance - > fw_supported_vd_count = MAX_LOGICAL_DRIVES_EXT ;
instance - > fw_supported_pd_count = MAX_PHYSICAL_DEVICES ;
} else {
instance - > fw_supported_vd_count = MAX_LOGICAL_DRIVES ;
instance - > fw_supported_pd_count = MAX_PHYSICAL_DEVICES ;
}
2015-04-23 11:01:09 +00:00
dev_info ( & instance - > pdev - > dev ,
2018-10-17 06:37:56 +00:00
" FW provided supportMaxExtLDs: %d \t max_lds: %d \n " ,
instance - > ctrl_info_buf - > adapterOperations3 . supportMaxExtLDs ? 1 : 0 ,
instance - > ctrl_info_buf - > max_lds ) ;
2014-11-17 09:54:13 +00:00
2017-01-10 23:20:48 +00:00
if ( instance - > max_raid_mapsize ) {
ventura_map_sz = instance - > max_raid_mapsize *
MR_MIN_MAP_SIZE ; /* 64k */
fusion - > current_map_sz = ventura_map_sz ;
fusion - > max_map_sz = ventura_map_sz ;
} else {
fusion - > old_map_sz = sizeof ( struct MR_FW_RAID_MAP ) +
( sizeof ( struct MR_LD_SPAN_MAP ) *
( instance - > fw_supported_vd_count - 1 ) ) ;
fusion - > new_map_sz = sizeof ( struct MR_FW_RAID_MAP_EXT ) ;
2014-11-17 09:54:13 +00:00
2017-01-10 23:20:48 +00:00
fusion - > max_map_sz =
max ( fusion - > old_map_sz , fusion - > new_map_sz ) ;
2014-11-17 09:54:13 +00:00
2017-01-10 23:20:48 +00:00
if ( instance - > supportmax256vd )
fusion - > current_map_sz = fusion - > new_map_sz ;
else
fusion - > current_map_sz = fusion - > old_map_sz ;
}
/* irrespective of FW raid maps, driver raid map is constant */
fusion - > drv_map_sz = sizeof ( struct MR_DRV_RAID_MAP_ALL ) ;
2014-11-17 09:54:13 +00:00
}
2018-10-17 06:37:40 +00:00
/*
* dcmd . opcode - MR_DCMD_CTRL_SNAPDUMP_GET_PROPERTIES
* dcmd . hdr . length - number of bytes to read
* dcmd . sge - Ptr to MR_SNAPDUMP_PROPERTIES
* Desc : Fill in snapdump properties
* Status : MFI_STAT_OK - Command successful
*/
void megasas_get_snapdump_properties ( struct megasas_instance * instance )
{
int ret = 0 ;
struct megasas_cmd * cmd ;
struct megasas_dcmd_frame * dcmd ;
struct MR_SNAPDUMP_PROPERTIES * ci ;
dma_addr_t ci_h = 0 ;
ci = instance - > snapdump_prop ;
ci_h = instance - > snapdump_prop_h ;
if ( ! ci )
return ;
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd ) {
dev_dbg ( & instance - > pdev - > dev , " Failed to get a free cmd \n " ) ;
return ;
}
dcmd = & cmd - > frame - > dcmd ;
memset ( ci , 0 , sizeof ( * ci ) ) ;
memset ( dcmd - > mbox . b , 0 , MFI_MBOX_SIZE ) ;
dcmd - > cmd = MFI_CMD_DCMD ;
dcmd - > cmd_status = MFI_STAT_INVALID_STATUS ;
dcmd - > sge_count = 1 ;
dcmd - > flags = MFI_FRAME_DIR_READ ;
dcmd - > timeout = 0 ;
dcmd - > pad_0 = 0 ;
dcmd - > data_xfer_len = cpu_to_le32 ( sizeof ( struct MR_SNAPDUMP_PROPERTIES ) ) ;
dcmd - > opcode = cpu_to_le32 ( MR_DCMD_CTRL_SNAPDUMP_GET_PROPERTIES ) ;
megasas_set_dma_settings ( instance , dcmd , ci_h ,
sizeof ( struct MR_SNAPDUMP_PROPERTIES ) ) ;
if ( ! instance - > mask_interrupts ) {
ret = megasas_issue_blocked_cmd ( instance , cmd ,
MFI_IO_TIMEOUT_SECS ) ;
} else {
ret = megasas_issue_polled ( instance , cmd ) ;
cmd - > flags | = DRV_DCMD_SKIP_REFIRE ;
}
switch ( ret ) {
case DCMD_SUCCESS :
instance - > snapdump_wait_time =
min_t ( u8 , ci - > trigger_min_num_sec_before_ocr ,
MEGASAS_MAX_SNAP_DUMP_WAIT_TIME ) ;
break ;
case DCMD_TIMEOUT :
switch ( dcmd_timeout_ocr_possible ( instance ) ) {
case INITIATE_OCR :
cmd - > flags | = DRV_DCMD_SKIP_REFIRE ;
2019-06-25 11:04:23 +00:00
mutex_unlock ( & instance - > reset_mutex ) ;
2018-10-17 06:37:40 +00:00
megasas_reset_fusion ( instance - > host ,
MFI_IO_TIMEOUT_OCR ) ;
2019-06-25 11:04:23 +00:00
mutex_lock ( & instance - > reset_mutex ) ;
2018-10-17 06:37:40 +00:00
break ;
case KILL_ADAPTER :
megaraid_sas_kill_hba ( instance ) ;
break ;
case IGNORE_TIMEOUT :
dev_info ( & instance - > pdev - > dev , " Ignore DCMD timeout: %s %d \n " ,
__func__ , __LINE__ ) ;
break ;
}
}
if ( ret ! = DCMD_TIMEOUT )
megasas_return_cmd ( instance , cmd ) ;
}
2005-09-20 21:46:58 +00:00
/**
* megasas_get_controller_info - Returns FW ' s controller structure
* @ instance : Adapter soft state
*
* Issues an internal command ( DCMD ) to get the FW ' s controller structure .
* This information is mainly used to find out the maximum IO transfer per
* command supported by the FW .
*/
2014-09-12 13:27:33 +00:00
int
2014-11-17 09:54:13 +00:00
megasas_get_ctrl_info ( struct megasas_instance * instance )
2005-09-20 21:46:58 +00:00
{
int ret = 0 ;
struct megasas_cmd * cmd ;
struct megasas_dcmd_frame * dcmd ;
struct megasas_ctrl_info * ci ;
dma_addr_t ci_h = 0 ;
2017-10-19 09:48:56 +00:00
ci = instance - > ctrl_info_buf ;
ci_h = instance - > ctrl_info_buf_h ;
2014-11-17 09:54:13 +00:00
2005-09-20 21:46:58 +00:00
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " Failed to get a free cmd \n " ) ;
2005-09-20 21:46:58 +00:00
return - ENOMEM ;
}
dcmd = & cmd - > frame - > dcmd ;
memset ( ci , 0 , sizeof ( * ci ) ) ;
memset ( dcmd - > mbox . b , 0 , MFI_MBOX_SIZE ) ;
dcmd - > cmd = MFI_CMD_DCMD ;
2015-05-06 13:31:02 +00:00
dcmd - > cmd_status = MFI_STAT_INVALID_STATUS ;
2005-09-20 21:46:58 +00:00
dcmd - > sge_count = 1 ;
2017-10-19 09:49:05 +00:00
dcmd - > flags = MFI_FRAME_DIR_READ ;
2005-09-20 21:46:58 +00:00
dcmd - > timeout = 0 ;
2009-12-06 15:24:21 +00:00
dcmd - > pad_0 = 0 ;
2013-09-06 10:20:52 +00:00
dcmd - > data_xfer_len = cpu_to_le32 ( sizeof ( struct megasas_ctrl_info ) ) ;
dcmd - > opcode = cpu_to_le32 ( MR_DCMD_CTRL_GET_INFO ) ;
2014-09-12 13:27:33 +00:00
dcmd - > mbox . b [ 0 ] = 1 ;
2005-09-20 21:46:58 +00:00
2017-10-19 09:49:05 +00:00
megasas_set_dma_settings ( instance , dcmd , ci_h ,
sizeof ( struct megasas_ctrl_info ) ) ;
2017-10-19 09:48:50 +00:00
if ( ( instance - > adapter_type ! = MFI_SERIES ) & &
2018-01-05 13:27:47 +00:00
! instance - > mask_interrupts ) {
2016-01-28 15:34:23 +00:00
ret = megasas_issue_blocked_cmd ( instance , cmd , MFI_IO_TIMEOUT_SECS ) ;
2018-01-05 13:27:47 +00:00
} else {
2014-09-12 13:27:58 +00:00
ret = megasas_issue_polled ( instance , cmd ) ;
2018-01-05 13:27:47 +00:00
cmd - > flags | = DRV_DCMD_SKIP_REFIRE ;
}
2014-09-12 13:27:58 +00:00
2016-01-28 15:34:23 +00:00
switch ( ret ) {
case DCMD_SUCCESS :
/* Save required controller information in
* CPU endianness format .
*/
2017-10-19 09:48:57 +00:00
le32_to_cpus ( ( u32 * ) & ci - > properties . OnOffProperties ) ;
2018-10-17 06:37:40 +00:00
le16_to_cpus ( ( u16 * ) & ci - > properties . on_off_properties2 ) ;
2017-10-19 09:48:57 +00:00
le32_to_cpus ( ( u32 * ) & ci - > adapterOperations2 ) ;
le32_to_cpus ( ( u32 * ) & ci - > adapterOperations3 ) ;
le16_to_cpus ( ( u16 * ) & ci - > adapter_operations4 ) ;
2019-06-25 11:04:30 +00:00
le32_to_cpus ( ( u32 * ) & ci - > adapter_operations5 ) ;
2016-01-28 15:34:23 +00:00
/* Update the latest Ext VD info.
* From Init path , store current firmware details .
* From OCR path , detect any firmware properties changes .
* in case of Firmware upgrade without system reboot .
*/
2014-11-17 09:54:13 +00:00
megasas_update_ext_vd_details ( instance ) ;
2019-06-25 11:04:26 +00:00
instance - > support_seqnum_jbod_fp =
2017-10-19 09:48:57 +00:00
ci - > adapterOperations3 . useSeqNumJbodFP ;
2017-01-10 23:20:52 +00:00
instance - > support_morethan256jbod =
2017-10-19 09:48:57 +00:00
ci - > adapter_operations4 . support_pd_map_target_id ;
2018-01-05 13:33:04 +00:00
instance - > support_nvme_passthru =
ci - > adapter_operations4 . support_nvme_passthru ;
2019-06-25 11:04:30 +00:00
instance - > support_pci_lane_margining =
ci - > adapter_operations5 . support_pci_lane_margining ;
2018-06-04 10:45:12 +00:00
instance - > task_abort_tmo = ci - > TaskAbortTO ;
instance - > max_reset_tmo = ci - > MaxResetTO ;
2016-01-28 15:34:23 +00:00
/*Check whether controller is iMR or MR */
2017-10-19 09:48:57 +00:00
instance - > is_imr = ( ci - > memory_size ? 0 : 1 ) ;
2018-10-17 06:37:40 +00:00
instance - > snapdump_wait_time =
( ci - > properties . on_off_properties2 . enable_snap_dump ?
MEGASAS_DEFAULT_SNAP_DUMP_WAIT_TIME : 0 ) ;
2019-01-29 09:38:14 +00:00
instance - > enable_fw_dev_list =
ci - > properties . on_off_properties2 . enable_fw_dev_list ;
2015-04-23 11:01:24 +00:00
dev_info ( & instance - > pdev - > dev ,
2016-01-28 15:34:23 +00:00
" controller type \t : %s(%dMB) \n " ,
instance - > is_imr ? " iMR " : " MR " ,
2017-10-19 09:48:57 +00:00
le16_to_cpu ( ci - > memory_size ) ) ;
2016-01-28 15:34:23 +00:00
2015-10-15 08:10:14 +00:00
instance - > disableOnlineCtrlReset =
2017-10-19 09:48:57 +00:00
ci - > properties . OnOffProperties . disableOnlineCtrlReset ;
2015-10-15 08:10:24 +00:00
instance - > secure_jbod_support =
2017-10-19 09:48:57 +00:00
ci - > adapterOperations3 . supportSecurityonJBOD ;
2016-01-28 15:34:23 +00:00
dev_info ( & instance - > pdev - > dev , " Online Controller Reset(OCR) \t : %s \n " ,
instance - > disableOnlineCtrlReset ? " Disabled " : " Enabled " ) ;
2015-10-15 08:10:24 +00:00
dev_info ( & instance - > pdev - > dev , " Secure JBOD support \t : %s \n " ,
instance - > secure_jbod_support ? " Yes " : " No " ) ;
2018-01-05 13:33:04 +00:00
dev_info ( & instance - > pdev - > dev , " NVMe passthru support \t : %s \n " ,
instance - > support_nvme_passthru ? " Yes " : " No " ) ;
2018-06-04 10:45:12 +00:00
dev_info ( & instance - > pdev - > dev ,
" FW provided TM TaskAbort/Reset timeout \t : %d secs/%d secs \n " ,
instance - > task_abort_tmo , instance - > max_reset_tmo ) ;
2019-06-25 11:04:26 +00:00
dev_info ( & instance - > pdev - > dev , " JBOD sequence map support \t : %s \n " ,
instance - > support_seqnum_jbod_fp ? " Yes " : " No " ) ;
2019-06-25 11:04:30 +00:00
dev_info ( & instance - > pdev - > dev , " PCI Lane Margining support \t : %s \n " ,
instance - > support_pci_lane_margining ? " Yes " : " No " ) ;
2018-06-04 10:45:12 +00:00
2016-01-28 15:34:23 +00:00
break ;
case DCMD_TIMEOUT :
switch ( dcmd_timeout_ocr_possible ( instance ) ) {
case INITIATE_OCR :
cmd - > flags | = DRV_DCMD_SKIP_REFIRE ;
2019-06-25 11:04:23 +00:00
mutex_unlock ( & instance - > reset_mutex ) ;
2016-01-28 15:34:23 +00:00
megasas_reset_fusion ( instance - > host ,
MFI_IO_TIMEOUT_OCR ) ;
2019-06-25 11:04:23 +00:00
mutex_lock ( & instance - > reset_mutex ) ;
2016-01-28 15:34:23 +00:00
break ;
case KILL_ADAPTER :
megaraid_sas_kill_hba ( instance ) ;
break ;
case IGNORE_TIMEOUT :
dev_info ( & instance - > pdev - > dev , " Ignore DCMD timeout: %s %d \n " ,
__func__ , __LINE__ ) ;
break ;
}
2018-06-04 10:45:09 +00:00
break ;
2016-01-28 15:34:23 +00:00
case DCMD_FAILED :
megaraid_sas_kill_hba ( instance ) ;
break ;
2014-11-17 09:54:13 +00:00
}
2005-09-20 21:46:58 +00:00
2018-06-04 10:45:09 +00:00
if ( ret ! = DCMD_TIMEOUT )
megasas_return_cmd ( instance , cmd ) ;
2016-01-28 15:34:23 +00:00
2005-09-20 21:46:58 +00:00
return ret ;
}
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
/*
* megasas_set_crash_dump_params - Sends address of crash dump DMA buffer
* to firmware
*
* @ instance : Adapter soft state
* @ crash_buf_state - tell FW to turn ON / OFF crash dump feature
MR_CRASH_BUF_TURN_OFF = 0
MR_CRASH_BUF_TURN_ON = 1
* @ return 0 on success non - zero on failure .
* Issues an internal command ( DCMD ) to set parameters for crash dump feature .
* Driver will send address of crash dump DMA buffer and set mbox to tell FW
* that driver supports crash dump feature . This DCMD will be sent only if
* crash dump feature is supported by the FW .
*
*/
int megasas_set_crash_dump_params ( struct megasas_instance * instance ,
u8 crash_buf_state )
{
int ret = 0 ;
struct megasas_cmd * cmd ;
struct megasas_dcmd_frame * dcmd ;
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd ) {
dev_err ( & instance - > pdev - > dev , " Failed to get a free cmd \n " ) ;
return - ENOMEM ;
}
dcmd = & cmd - > frame - > dcmd ;
memset ( dcmd - > mbox . b , 0 , MFI_MBOX_SIZE ) ;
dcmd - > mbox . b [ 0 ] = crash_buf_state ;
dcmd - > cmd = MFI_CMD_DCMD ;
2015-05-06 13:31:02 +00:00
dcmd - > cmd_status = MFI_STAT_INVALID_STATUS ;
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
dcmd - > sge_count = 1 ;
2017-10-19 09:49:05 +00:00
dcmd - > flags = MFI_FRAME_DIR_NONE ;
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
dcmd - > timeout = 0 ;
dcmd - > pad_0 = 0 ;
dcmd - > data_xfer_len = cpu_to_le32 ( CRASH_DMA_BUF_SIZE ) ;
dcmd - > opcode = cpu_to_le32 ( MR_DCMD_CTRL_SET_CRASH_DUMP_PARAMS ) ;
2017-10-19 09:49:05 +00:00
megasas_set_dma_settings ( instance , dcmd , instance - > crash_dump_h ,
CRASH_DMA_BUF_SIZE ) ;
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
2017-10-19 09:48:50 +00:00
if ( ( instance - > adapter_type ! = MFI_SERIES ) & &
! instance - > mask_interrupts )
2016-01-28 15:34:23 +00:00
ret = megasas_issue_blocked_cmd ( instance , cmd , MFI_IO_TIMEOUT_SECS ) ;
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
else
2014-09-12 13:27:58 +00:00
ret = megasas_issue_polled ( instance , cmd ) ;
2016-01-28 15:34:23 +00:00
if ( ret = = DCMD_TIMEOUT ) {
switch ( dcmd_timeout_ocr_possible ( instance ) ) {
case INITIATE_OCR :
cmd - > flags | = DRV_DCMD_SKIP_REFIRE ;
megasas_reset_fusion ( instance - > host ,
MFI_IO_TIMEOUT_OCR ) ;
break ;
case KILL_ADAPTER :
megaraid_sas_kill_hba ( instance ) ;
break ;
case IGNORE_TIMEOUT :
dev_info ( & instance - > pdev - > dev , " Ignore DCMD timeout: %s %d \n " ,
__func__ , __LINE__ ) ;
break ;
}
} else
megasas_return_cmd ( instance , cmd ) ;
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
return ret ;
}
2007-11-07 17:09:50 +00:00
/**
* megasas_issue_init_mfi - Initializes the FW
* @ instance : Adapter soft state
*
* Issues the INIT MFI cmd
*/
static int
megasas_issue_init_mfi ( struct megasas_instance * instance )
{
2015-04-23 11:02:54 +00:00
__le32 context ;
2007-11-07 17:09:50 +00:00
struct megasas_cmd * cmd ;
struct megasas_init_frame * init_frame ;
struct megasas_init_queue_info * initq_info ;
dma_addr_t init_frame_h ;
dma_addr_t initq_info_h ;
/*
* Prepare a init frame . Note the init frame points to queue info
* structure . Each frame has SGL allocated after first 64 bytes . For
* this frame - since we don ' t need any SGL - we use SGL ' s space as
* queue info structure
*
* We will not get a NULL command below . We just created the pool .
*/
cmd = megasas_get_cmd ( instance ) ;
init_frame = ( struct megasas_init_frame * ) cmd - > frame ;
initq_info = ( struct megasas_init_queue_info * )
( ( unsigned long ) init_frame + 64 ) ;
init_frame_h = cmd - > frame_phys_addr ;
initq_info_h = init_frame_h + 64 ;
context = init_frame - > context ;
memset ( init_frame , 0 , MEGAMFI_FRAME_SIZE ) ;
memset ( initq_info , 0 , sizeof ( struct megasas_init_queue_info ) ) ;
init_frame - > context = context ;
2013-09-06 10:20:52 +00:00
initq_info - > reply_queue_entries = cpu_to_le32 ( instance - > max_fw_cmds + 1 ) ;
initq_info - > reply_queue_start_phys_addr_lo = cpu_to_le32 ( instance - > reply_queue_h ) ;
2007-11-07 17:09:50 +00:00
2013-09-06 10:20:52 +00:00
initq_info - > producer_index_phys_addr_lo = cpu_to_le32 ( instance - > producer_h ) ;
initq_info - > consumer_index_phys_addr_lo = cpu_to_le32 ( instance - > consumer_h ) ;
2007-11-07 17:09:50 +00:00
init_frame - > cmd = MFI_CMD_INIT ;
2015-05-06 13:31:02 +00:00
init_frame - > cmd_status = MFI_STAT_INVALID_STATUS ;
2013-09-06 10:20:52 +00:00
init_frame - > queue_info_new_phys_addr_lo =
cpu_to_le32 ( lower_32_bits ( initq_info_h ) ) ;
init_frame - > queue_info_new_phys_addr_hi =
cpu_to_le32 ( upper_32_bits ( initq_info_h ) ) ;
2007-11-07 17:09:50 +00:00
2013-09-06 10:20:52 +00:00
init_frame - > data_xfer_len = cpu_to_le32 ( sizeof ( struct megasas_init_queue_info ) ) ;
2007-11-07 17:09:50 +00:00
/*
* disable the intr before firing the init frame to FW
*/
2013-05-22 07:04:14 +00:00
instance - > instancet - > disable_intr ( instance ) ;
2007-11-07 17:09:50 +00:00
/*
* Issue the init frame in polled mode
*/
if ( megasas_issue_polled ( instance , cmd ) ) {
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " Failed to init firmware \n " ) ;
2007-11-07 17:09:50 +00:00
megasas_return_cmd ( instance , cmd ) ;
goto fail_fw_init ;
}
megasas_return_cmd ( instance , cmd ) ;
return 0 ;
fail_fw_init :
return - EINVAL ;
}
2010-12-21 18:23:23 +00:00
static u32
megasas_init_adapter_mfi ( struct megasas_instance * instance )
2005-09-20 21:46:58 +00:00
{
u32 context_sz ;
u32 reply_q_sz ;
/*
* Get various operational parameters from status register
*/
2018-12-17 08:47:39 +00:00
instance - > max_fw_cmds = instance - > instancet - > read_fw_status_reg ( instance ) & 0x00FFFF ;
2006-10-03 19:28:49 +00:00
/*
* Reduce the max supported cmds by 1. This is to ensure that the
* reply_q_sz ( 1 more than the max cmd that driver may send )
* does not exceed max cmds that the FW can support
*/
instance - > max_fw_cmds = instance - > max_fw_cmds - 1 ;
2010-12-21 21:34:31 +00:00
instance - > max_mfi_cmds = instance - > max_fw_cmds ;
2018-12-17 08:47:39 +00:00
instance - > max_num_sge = ( instance - > instancet - > read_fw_status_reg ( instance ) & 0xFF0000 ) > >
2006-01-25 20:02:40 +00:00
0x10 ;
2015-04-23 11:00:54 +00:00
/*
* For MFI skinny adapters , MEGASAS_SKINNY_INT_CMDS commands
* are reserved for IOCTL + driver ' s internal DCMDs .
*/
if ( ( instance - > pdev - > device = = PCI_DEVICE_ID_LSI_SAS0073SKINNY ) | |
( instance - > pdev - > device = = PCI_DEVICE_ID_LSI_SAS0071SKINNY ) ) {
instance - > max_scsi_cmds = ( instance - > max_fw_cmds -
MEGASAS_SKINNY_INT_CMDS ) ;
sema_init ( & instance - > ioctl_sem , MEGASAS_SKINNY_INT_CMDS ) ;
} else {
instance - > max_scsi_cmds = ( instance - > max_fw_cmds -
MEGASAS_INT_CMDS ) ;
sema_init ( & instance - > ioctl_sem , ( MEGASAS_MFI_IOCTL_CMDS ) ) ;
}
2016-01-28 15:34:30 +00:00
instance - > cur_can_queue = instance - > max_scsi_cmds ;
2005-09-20 21:46:58 +00:00
/*
* Create a pool of commands
*/
if ( megasas_alloc_cmds ( instance ) )
goto fail_alloc_cmds ;
/*
* Allocate memory for reply queue . Length of reply queue should
* be _one_ more than the maximum commands handled by the firmware .
*
* Note : When FW completes commands , it places corresponding contex
* values in this circular reply queue . This circular queue is a fairly
* typical producer - consumer queue . FW is the producer ( of completed
* commands ) and the driver is the consumer .
*/
context_sz = sizeof ( u32 ) ;
reply_q_sz = context_sz * ( instance - > max_fw_cmds + 1 ) ;
2018-10-10 17:31:25 +00:00
instance - > reply_queue = dma_alloc_coherent ( & instance - > pdev - > dev ,
reply_q_sz , & instance - > reply_queue_h , GFP_KERNEL ) ;
2005-09-20 21:46:58 +00:00
if ( ! instance - > reply_queue ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " Out of DMA mem for reply queue \n " ) ;
2005-09-20 21:46:58 +00:00
goto fail_reply_queue ;
}
2007-11-07 17:09:50 +00:00
if ( megasas_issue_init_mfi ( instance ) )
2005-09-20 21:46:58 +00:00
goto fail_fw_init ;
2014-11-17 09:54:13 +00:00
if ( megasas_get_ctrl_info ( instance ) ) {
2014-09-12 13:27:33 +00:00
dev_err ( & instance - > pdev - > dev , " (%d): Could get controller info "
" Fail from %s %d \n " , instance - > unique_id ,
__func__ , __LINE__ ) ;
goto fail_fw_init ;
}
2010-09-23 02:36:29 +00:00
instance - > fw_support_ieee = 0 ;
instance - > fw_support_ieee =
2018-12-17 08:47:39 +00:00
( instance - > instancet - > read_fw_status_reg ( instance ) &
2010-09-23 02:36:29 +00:00
0x04000000 ) ;
2015-07-07 20:52:34 +00:00
dev_notice ( & instance - > pdev - > dev , " megasas_init_mfi: fw_support_ieee=%d " ,
2010-09-23 02:36:29 +00:00
instance - > fw_support_ieee ) ;
if ( instance - > fw_support_ieee )
instance - > flag_ieee = 1 ;
2010-12-21 18:23:23 +00:00
return 0 ;
fail_fw_init :
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & instance - > pdev - > dev , reply_q_sz ,
2010-12-21 18:23:23 +00:00
instance - > reply_queue , instance - > reply_queue_h ) ;
fail_reply_queue :
megasas_free_cmds ( instance ) ;
fail_alloc_cmds :
return 1 ;
}
scsi: megaraid_sas: IRQ poll to avoid CPU hard lockups
Issue Description:
We have seen cpu lock up issues from field if system has a large (more than
96) logical cpu count. SAS3.0 controller (Invader series) supports max 96
MSI-X vector and SAS3.5 product (Ventura) supports max 128 MSI-X vectors.
This may be a generic issue (if PCI device support completion on multiple
reply queues).
Let me explain it w.r.t megaraid_sas supported h/w just to simplify the
problem and possible changes to handle such issues. MegaRAID controller
supports multiple reply queues in completion path. Driver creates MSI-X
vectors for controller as "minimum of (FW supported Reply queues, Logical
CPUs)". If submitter is not interrupted via completion on same CPU, there
is a loop in the IO path. This behavior can cause hard/soft CPU lockups, IO
timeout, system sluggish etc.
Example - one CPU (e.g. CPU A) is busy submitting the IOs and another CPU
(e.g. CPU B) is busy with processing the corresponding IO's reply
descriptors from reply descriptor queue upon receiving the interrupts from
HBA. If CPU A is continuously pumping the IOs then always CPU B (which is
executing the ISR) will see the valid reply descriptors in the reply
descriptor queue and it will be continuously processing those reply
descriptor in a loop without quitting the ISR handler.
megaraid_sas driver will exit ISR handler if it finds unused reply
descriptor in the reply descriptor queue. Since CPU A will be continuously
sending the IOs, CPU B may always see a valid reply descriptor (posted by
HBA Firmware after processing the IO) in the reply descriptor queue. In
worst case, driver will not quit from this loop in the ISR handler.
Eventually, CPU lockup will be detected by watchdog.
Above mentioned behavior is not common if "rq_affinity" set to 2 or
affinity_hint is honored by irqbalancer as "exact". If rq_affinity is set
to 2, submitter will be always interrupted via completion on same CPU. If
irqbalancer is using "exact" policy, interrupt will be delivered to
submitter CPU.
Problem statement:
If CPU count to MSI-X vectors (reply descriptor Queues) count ratio is not
1:1, we still have exposure of issue explained above and for that we don't
have any solution.
Exposure of soft/hard lockup is seen if CPU count is more than MSI-X
supported by device.
If CPUs count to MSI-X vectors count ratio is not 1:1, (Other way, if
CPU counts to MSI-X vector count ratio is something like X:1, where X > 1)
then 'exact' irqbalance policy OR rq_affinity = 2 won't help to avoid CPU
hard/soft lockups. There won't be any one to one mapping between
CPU to MSI-X vector instead one MSI-X interrupt (or reply descriptor queue)
is shared with group/set of CPUs and there is a possibility of having a
loop in the IO path within that CPU group and may observe lockups.
For example: Consider a system having two NUMA nodes and each node having
four logical CPUs and also consider that number of MSI-X vectors enabled on
the HBA is two, then CPUs count to MSI-X vector count ratio as 4:1.
e.g.
MSI-X vector 0 is affinity to CPU 0, CPU 1, CPU 2 & CPU 3 of NUMA node 0 and
MSI-X vector 1 is affinity to CPU 4, CPU 5, CPU 6 & CPU 7 of NUMA node 1.
numactl --hardware
available: 2 nodes (0-1)
node 0 cpus: 0 1 2 3 --> MSI-X 0
node 0 size: 65536 MB
node 0 free: 63176 MB
node 1 cpus: 4 5 6 7 --> MSI-X 1
node 1 size: 65536 MB
node 1 free: 63176 MB
Assume that user started an application which uses all the CPUs of NUMA
node 0 for issuing the IOs. Only one CPU from affinity list (it can be any
cpu since this behavior depends upon irqbalance) CPU0 will receive the
interrupts from MSI-X 0 for all the IOs. Eventually, CPU 0 IO submission
percentage will be decreasing and ISR processing percentage will be
increasing as it is more busy with processing the interrupts. Gradually IO
submission percentage on CPU 0 will be zero and it's ISR processing
percentage will be 100% as IO loop has already formed within the
NUMA node 0, i.e. CPU 1, CPU 2 & CPU 3 will be continuously busy with
submitting the heavy IOs and only CPU 0 is busy in the ISR path as it
always find the valid reply descriptor in the reply descriptor queue.
Eventually, we will observe the hard lockup here.
Chances of occurring of hard/soft lockups are directly proportional to
value of X. If value of X is high, then chances of observing CPU lockups is
high.
Solution:
Use IRQ poll interface defined in "irq_poll.c".
megaraid_sas driver will execute ISR routine in softirq context and it will
always quit the loop based on budget provided in IRQ poll interface.
Driver will switch to IRQ poll only when more than a threshold number of
reply descriptors are handled in one ISR. Currently threshold is set as
1/4th of HBA queue depth.
In these scenarios (i.e. where CPUs count to MSI-X vectors count ratio is
X:1 (where X > 1)), IRQ poll interface will avoid CPU hard lockups due to
voluntary exit from the reply queue processing based on budget.
Note - Only one MSI-X vector is busy doing processing.
Select CONFIG_IRQ_POLL from driver Kconfig for driver compilation.
Signed-off-by: Kashyap Desai <kashyap.desai@broadcom.com>
Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-05-07 17:05:35 +00:00
static
void megasas_setup_irq_poll ( struct megasas_instance * instance )
{
struct megasas_irq_context * irq_ctx ;
u32 count , i ;
count = instance - > msix_vectors > 0 ? instance - > msix_vectors : 1 ;
/* Initialize IRQ poll */
for ( i = 0 ; i < count ; i + + ) {
irq_ctx = & instance - > irq_context [ i ] ;
irq_ctx - > os_irq = pci_irq_vector ( instance - > pdev , i ) ;
irq_ctx - > irq_poll_scheduled = false ;
irq_poll_init ( & irq_ctx - > irqpoll ,
instance - > threshold_reply_count ,
megasas_irqpoll ) ;
}
}
2015-04-23 11:00:09 +00:00
/*
2016-12-02 11:52:23 +00:00
* megasas_setup_irqs_ioapic - register legacy interrupts .
2015-04-23 11:00:09 +00:00
* @ instance : Adapter soft state
*
* Do not enable interrupt , only setup ISRs .
*
* Return 0 on success .
*/
static int
megasas_setup_irqs_ioapic ( struct megasas_instance * instance )
{
struct pci_dev * pdev ;
pdev = instance - > pdev ;
instance - > irq_context [ 0 ] . instance = instance ;
instance - > irq_context [ 0 ] . MSIxIndex = 0 ;
2019-10-07 05:18:28 +00:00
snprintf ( instance - > irq_context - > name , MEGASAS_MSIX_NAME_LEN , " %s%u " ,
" megasas " , instance - > host - > host_no ) ;
2016-12-02 11:52:23 +00:00
if ( request_irq ( pci_irq_vector ( pdev , 0 ) ,
instance - > instancet - > service_isr , IRQF_SHARED ,
2019-10-07 05:18:28 +00:00
instance - > irq_context - > name , & instance - > irq_context [ 0 ] ) ) {
2015-04-23 11:00:09 +00:00
dev_err ( & instance - > pdev - > dev ,
" Failed to register IRQ from %s %d \n " ,
__func__ , __LINE__ ) ;
return - 1 ;
}
2019-06-25 11:04:35 +00:00
instance - > perf_mode = MR_LATENCY_PERF_MODE ;
2019-06-25 11:04:31 +00:00
instance - > low_latency_index_start = 0 ;
2015-04-23 11:00:09 +00:00
return 0 ;
}
/**
* megasas_setup_irqs_msix - register MSI - x interrupts .
* @ instance : Adapter soft state
* @ is_probe : Driver probe check
*
* Do not enable interrupt , only setup ISRs .
*
* Return 0 on success .
*/
static int
megasas_setup_irqs_msix ( struct megasas_instance * instance , u8 is_probe )
{
2016-12-02 11:52:23 +00:00
int i , j ;
2015-04-23 11:00:09 +00:00
struct pci_dev * pdev ;
pdev = instance - > pdev ;
/* Try MSI-x */
for ( i = 0 ; i < instance - > msix_vectors ; i + + ) {
instance - > irq_context [ i ] . instance = instance ;
instance - > irq_context [ i ] . MSIxIndex = i ;
2019-10-07 05:18:28 +00:00
snprintf ( instance - > irq_context [ i ] . name , MEGASAS_MSIX_NAME_LEN , " %s%u-msix%u " ,
" megasas " , instance - > host - > host_no , i ) ;
2016-12-02 11:52:23 +00:00
if ( request_irq ( pci_irq_vector ( pdev , i ) ,
2019-10-07 05:18:28 +00:00
instance - > instancet - > service_isr , 0 , instance - > irq_context [ i ] . name ,
2015-04-23 11:00:09 +00:00
& instance - > irq_context [ i ] ) ) {
dev_err ( & instance - > pdev - > dev ,
" Failed to register IRQ for vector %d. \n " , i ) ;
2016-12-02 11:52:23 +00:00
for ( j = 0 ; j < i ; j + + )
free_irq ( pci_irq_vector ( pdev , j ) ,
& instance - > irq_context [ j ] ) ;
2015-04-23 11:00:09 +00:00
/* Retry irq register for IO_APIC*/
instance - > msix_vectors = 0 ;
2019-05-07 17:05:36 +00:00
instance - > msix_load_balance = false ;
2017-03-10 11:22:12 +00:00
if ( is_probe ) {
pci_free_irq_vectors ( instance - > pdev ) ;
2015-04-23 11:00:09 +00:00
return megasas_setup_irqs_ioapic ( instance ) ;
2017-03-10 11:22:12 +00:00
} else {
2015-04-23 11:00:09 +00:00
return - 1 ;
2017-03-10 11:22:12 +00:00
}
2015-04-23 11:00:09 +00:00
}
}
2019-05-07 17:05:36 +00:00
2015-04-23 11:00:09 +00:00
return 0 ;
}
/*
* megasas_destroy_irqs - unregister interrupts .
* @ instance : Adapter soft state
* return : void
*/
static void
megasas_destroy_irqs ( struct megasas_instance * instance ) {
int i ;
scsi: megaraid_sas: IRQ poll to avoid CPU hard lockups
Issue Description:
We have seen cpu lock up issues from field if system has a large (more than
96) logical cpu count. SAS3.0 controller (Invader series) supports max 96
MSI-X vector and SAS3.5 product (Ventura) supports max 128 MSI-X vectors.
This may be a generic issue (if PCI device support completion on multiple
reply queues).
Let me explain it w.r.t megaraid_sas supported h/w just to simplify the
problem and possible changes to handle such issues. MegaRAID controller
supports multiple reply queues in completion path. Driver creates MSI-X
vectors for controller as "minimum of (FW supported Reply queues, Logical
CPUs)". If submitter is not interrupted via completion on same CPU, there
is a loop in the IO path. This behavior can cause hard/soft CPU lockups, IO
timeout, system sluggish etc.
Example - one CPU (e.g. CPU A) is busy submitting the IOs and another CPU
(e.g. CPU B) is busy with processing the corresponding IO's reply
descriptors from reply descriptor queue upon receiving the interrupts from
HBA. If CPU A is continuously pumping the IOs then always CPU B (which is
executing the ISR) will see the valid reply descriptors in the reply
descriptor queue and it will be continuously processing those reply
descriptor in a loop without quitting the ISR handler.
megaraid_sas driver will exit ISR handler if it finds unused reply
descriptor in the reply descriptor queue. Since CPU A will be continuously
sending the IOs, CPU B may always see a valid reply descriptor (posted by
HBA Firmware after processing the IO) in the reply descriptor queue. In
worst case, driver will not quit from this loop in the ISR handler.
Eventually, CPU lockup will be detected by watchdog.
Above mentioned behavior is not common if "rq_affinity" set to 2 or
affinity_hint is honored by irqbalancer as "exact". If rq_affinity is set
to 2, submitter will be always interrupted via completion on same CPU. If
irqbalancer is using "exact" policy, interrupt will be delivered to
submitter CPU.
Problem statement:
If CPU count to MSI-X vectors (reply descriptor Queues) count ratio is not
1:1, we still have exposure of issue explained above and for that we don't
have any solution.
Exposure of soft/hard lockup is seen if CPU count is more than MSI-X
supported by device.
If CPUs count to MSI-X vectors count ratio is not 1:1, (Other way, if
CPU counts to MSI-X vector count ratio is something like X:1, where X > 1)
then 'exact' irqbalance policy OR rq_affinity = 2 won't help to avoid CPU
hard/soft lockups. There won't be any one to one mapping between
CPU to MSI-X vector instead one MSI-X interrupt (or reply descriptor queue)
is shared with group/set of CPUs and there is a possibility of having a
loop in the IO path within that CPU group and may observe lockups.
For example: Consider a system having two NUMA nodes and each node having
four logical CPUs and also consider that number of MSI-X vectors enabled on
the HBA is two, then CPUs count to MSI-X vector count ratio as 4:1.
e.g.
MSI-X vector 0 is affinity to CPU 0, CPU 1, CPU 2 & CPU 3 of NUMA node 0 and
MSI-X vector 1 is affinity to CPU 4, CPU 5, CPU 6 & CPU 7 of NUMA node 1.
numactl --hardware
available: 2 nodes (0-1)
node 0 cpus: 0 1 2 3 --> MSI-X 0
node 0 size: 65536 MB
node 0 free: 63176 MB
node 1 cpus: 4 5 6 7 --> MSI-X 1
node 1 size: 65536 MB
node 1 free: 63176 MB
Assume that user started an application which uses all the CPUs of NUMA
node 0 for issuing the IOs. Only one CPU from affinity list (it can be any
cpu since this behavior depends upon irqbalance) CPU0 will receive the
interrupts from MSI-X 0 for all the IOs. Eventually, CPU 0 IO submission
percentage will be decreasing and ISR processing percentage will be
increasing as it is more busy with processing the interrupts. Gradually IO
submission percentage on CPU 0 will be zero and it's ISR processing
percentage will be 100% as IO loop has already formed within the
NUMA node 0, i.e. CPU 1, CPU 2 & CPU 3 will be continuously busy with
submitting the heavy IOs and only CPU 0 is busy in the ISR path as it
always find the valid reply descriptor in the reply descriptor queue.
Eventually, we will observe the hard lockup here.
Chances of occurring of hard/soft lockups are directly proportional to
value of X. If value of X is high, then chances of observing CPU lockups is
high.
Solution:
Use IRQ poll interface defined in "irq_poll.c".
megaraid_sas driver will execute ISR routine in softirq context and it will
always quit the loop based on budget provided in IRQ poll interface.
Driver will switch to IRQ poll only when more than a threshold number of
reply descriptors are handled in one ISR. Currently threshold is set as
1/4th of HBA queue depth.
In these scenarios (i.e. where CPUs count to MSI-X vectors count ratio is
X:1 (where X > 1)), IRQ poll interface will avoid CPU hard lockups due to
voluntary exit from the reply queue processing based on budget.
Note - Only one MSI-X vector is busy doing processing.
Select CONFIG_IRQ_POLL from driver Kconfig for driver compilation.
Signed-off-by: Kashyap Desai <kashyap.desai@broadcom.com>
Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-05-07 17:05:35 +00:00
int count ;
struct megasas_irq_context * irq_ctx ;
count = instance - > msix_vectors > 0 ? instance - > msix_vectors : 1 ;
if ( instance - > adapter_type ! = MFI_SERIES ) {
for ( i = 0 ; i < count ; i + + ) {
irq_ctx = & instance - > irq_context [ i ] ;
irq_poll_disable ( & irq_ctx - > irqpoll ) ;
}
}
2015-04-23 11:00:09 +00:00
if ( instance - > msix_vectors )
for ( i = 0 ; i < instance - > msix_vectors ; i + + ) {
2016-12-02 11:52:23 +00:00
free_irq ( pci_irq_vector ( instance - > pdev , i ) ,
2015-04-23 11:00:09 +00:00
& instance - > irq_context [ i ] ) ;
}
else
2016-12-02 11:52:23 +00:00
free_irq ( pci_irq_vector ( instance - > pdev , 0 ) ,
& instance - > irq_context [ 0 ] ) ;
2015-04-23 11:00:09 +00:00
}
2015-08-31 11:53:11 +00:00
/**
* megasas_setup_jbod_map - setup jbod map for FP seq_number .
* @ instance : Adapter soft state
* @ is_probe : Driver probe check
*
* Return 0 on success .
*/
void
megasas_setup_jbod_map ( struct megasas_instance * instance )
{
int i ;
struct fusion_context * fusion = instance - > ctrl_context ;
u32 pd_seq_map_sz ;
pd_seq_map_sz = sizeof ( struct MR_PD_CFG_SEQ_NUM_SYNC ) +
( sizeof ( struct MR_PD_CFG_SEQ ) * ( MAX_PHYSICAL_DEVICES - 1 ) ) ;
2019-06-25 11:04:26 +00:00
instance - > use_seqnum_jbod_fp =
instance - > support_seqnum_jbod_fp ;
2015-08-31 11:53:11 +00:00
if ( reset_devices | | ! fusion | |
2019-06-25 11:04:26 +00:00
! instance - > support_seqnum_jbod_fp ) {
2015-08-31 11:53:11 +00:00
dev_info ( & instance - > pdev - > dev ,
2019-06-25 11:04:26 +00:00
" JBOD sequence map is disabled %s %d \n " ,
2015-08-31 11:53:11 +00:00
__func__ , __LINE__ ) ;
instance - > use_seqnum_jbod_fp = false ;
return ;
}
if ( fusion - > pd_seq_sync [ 0 ] )
goto skip_alloc ;
for ( i = 0 ; i < JBOD_MAPS_COUNT ; i + + ) {
fusion - > pd_seq_sync [ i ] = dma_alloc_coherent
( & instance - > pdev - > dev , pd_seq_map_sz ,
& fusion - > pd_seq_phys [ i ] , GFP_KERNEL ) ;
if ( ! fusion - > pd_seq_sync [ i ] ) {
dev_err ( & instance - > pdev - > dev ,
" Failed to allocate memory from %s %d \n " ,
__func__ , __LINE__ ) ;
if ( i = = 1 ) {
dma_free_coherent ( & instance - > pdev - > dev ,
pd_seq_map_sz , fusion - > pd_seq_sync [ 0 ] ,
fusion - > pd_seq_phys [ 0 ] ) ;
fusion - > pd_seq_sync [ 0 ] = NULL ;
}
instance - > use_seqnum_jbod_fp = false ;
return ;
}
}
skip_alloc :
if ( ! megasas_sync_pd_seq_num ( instance , false ) & &
! megasas_sync_pd_seq_num ( instance , true ) )
instance - > use_seqnum_jbod_fp = true ;
else
instance - > use_seqnum_jbod_fp = false ;
}
2018-03-13 09:42:40 +00:00
static void megasas_setup_reply_map ( struct megasas_instance * instance )
{
const struct cpumask * mask ;
2019-06-25 11:04:31 +00:00
unsigned int queue , cpu , low_latency_index_start ;
2018-03-13 09:42:40 +00:00
2019-06-25 11:04:31 +00:00
low_latency_index_start = instance - > low_latency_index_start ;
for ( queue = low_latency_index_start ; queue < instance - > msix_vectors ; queue + + ) {
2018-03-13 09:42:40 +00:00
mask = pci_irq_get_affinity ( instance - > pdev , queue ) ;
if ( ! mask )
goto fallback ;
for_each_cpu ( cpu , mask )
instance - > reply_map [ cpu ] = queue ;
}
return ;
fallback :
2019-06-25 11:04:31 +00:00
queue = low_latency_index_start ;
for_each_possible_cpu ( cpu ) {
instance - > reply_map [ cpu ] = queue ;
if ( queue = = ( instance - > msix_vectors - 1 ) )
queue = low_latency_index_start ;
else
queue + + ;
}
2018-03-13 09:42:40 +00:00
}
2019-01-29 09:38:12 +00:00
/**
* megasas_get_device_list - Get the PD and LD device list from FW .
* @ instance : Adapter soft state
* @ return : Success or failure
*
* Issue DCMDs to Firmware to get the PD and LD list .
2019-01-29 09:38:14 +00:00
* Based on the FW support , driver sends the HOST_DEVICE_LIST or combination
* of PD_LIST / LD_LIST_QUERY DCMDs to get the device list .
2019-01-29 09:38:12 +00:00
*/
static
int megasas_get_device_list ( struct megasas_instance * instance )
{
memset ( instance - > pd_list , 0 ,
( MEGASAS_MAX_PD * sizeof ( struct megasas_pd_list ) ) ) ;
memset ( instance - > ld_ids , 0xff , MEGASAS_MAX_LD_IDS ) ;
2019-01-29 09:38:14 +00:00
if ( instance - > enable_fw_dev_list ) {
if ( megasas_host_device_list_query ( instance , true ) )
return FAILED ;
} else {
if ( megasas_get_pd_list ( instance ) < 0 ) {
dev_err ( & instance - > pdev - > dev , " failed to get PD list \n " ) ;
return FAILED ;
}
2019-01-29 09:38:12 +00:00
2019-01-29 09:38:14 +00:00
if ( megasas_ld_list_query ( instance ,
MR_LD_QUERY_TYPE_EXPOSED_TO_HOST ) ) {
dev_err ( & instance - > pdev - > dev , " failed to get LD list \n " ) ;
return FAILED ;
}
2019-01-29 09:38:12 +00:00
}
return SUCCESS ;
}
2019-06-25 11:04:31 +00:00
2019-06-25 11:04:33 +00:00
/**
* megasas_set_high_iops_queue_affinity_hint - Set affinity hint for high IOPS queues
* @ instance : Adapter soft state
* return : void
*/
static inline void
megasas_set_high_iops_queue_affinity_hint ( struct megasas_instance * instance )
{
int i ;
int local_numa_node ;
2019-06-25 11:04:35 +00:00
if ( instance - > perf_mode = = MR_BALANCED_PERF_MODE ) {
2019-06-25 11:04:33 +00:00
local_numa_node = dev_to_node ( & instance - > pdev - > dev ) ;
for ( i = 0 ; i < instance - > low_latency_index_start ; i + + )
irq_set_affinity_hint ( pci_irq_vector ( instance - > pdev , i ) ,
cpumask_of_node ( local_numa_node ) ) ;
}
}
2019-06-25 11:04:31 +00:00
static int
__megasas_alloc_irq_vectors ( struct megasas_instance * instance )
{
int i , irq_flags ;
struct irq_affinity desc = { . pre_vectors = instance - > low_latency_index_start } ;
struct irq_affinity * descp = & desc ;
irq_flags = PCI_IRQ_MSIX ;
if ( instance - > smp_affinity_enable )
irq_flags | = PCI_IRQ_AFFINITY ;
else
descp = NULL ;
i = pci_alloc_irq_vectors_affinity ( instance - > pdev ,
instance - > low_latency_index_start ,
instance - > msix_vectors , irq_flags , descp ) ;
return i ;
}
/**
* megasas_alloc_irq_vectors - Allocate IRQ vectors / enable MSI - x vectors
* @ instance : Adapter soft state
* return : void
*/
static void
megasas_alloc_irq_vectors ( struct megasas_instance * instance )
{
int i ;
unsigned int num_msix_req ;
i = __megasas_alloc_irq_vectors ( instance ) ;
2019-06-25 11:04:35 +00:00
if ( ( instance - > perf_mode = = MR_BALANCED_PERF_MODE ) & &
( i ! = instance - > msix_vectors ) ) {
2019-06-25 11:04:31 +00:00
if ( instance - > msix_vectors )
pci_free_irq_vectors ( instance - > pdev ) ;
/* Disable Balanced IOPS mode and try realloc vectors */
2019-06-25 11:04:35 +00:00
instance - > perf_mode = MR_LATENCY_PERF_MODE ;
2019-06-25 11:04:31 +00:00
instance - > low_latency_index_start = 1 ;
num_msix_req = num_online_cpus ( ) + instance - > low_latency_index_start ;
instance - > msix_vectors = min ( num_msix_req ,
instance - > msix_vectors ) ;
i = __megasas_alloc_irq_vectors ( instance ) ;
}
dev_info ( & instance - > pdev - > dev ,
" requested/available msix %d/%d \n " , instance - > msix_vectors , i ) ;
if ( i > 0 )
instance - > msix_vectors = i ;
else
instance - > msix_vectors = 0 ;
2019-06-25 11:04:33 +00:00
if ( instance - > smp_affinity_enable )
megasas_set_high_iops_queue_affinity_hint ( instance ) ;
2019-06-25 11:04:31 +00:00
}
2010-12-21 18:23:23 +00:00
/**
* megasas_init_fw - Initializes the FW
* @ instance : Adapter soft state
*
* This is the main function for initializing firmware
*/
static int megasas_init_fw ( struct megasas_instance * instance )
{
u32 max_sectors_1 ;
2017-02-10 08:59:10 +00:00
u32 max_sectors_2 , tmp_sectors , msix_enable ;
2018-10-17 06:37:51 +00:00
u32 scratch_pad_1 , scratch_pad_2 , scratch_pad_3 , status_reg ;
2013-09-13 16:46:44 +00:00
resource_size_t base_addr ;
2019-05-07 17:05:42 +00:00
void * base_addr_phys ;
2014-09-12 13:27:33 +00:00
struct megasas_ctrl_info * ctrl_info = NULL ;
2010-12-21 18:23:23 +00:00
unsigned long bar_list ;
2019-05-07 17:05:48 +00:00
int i , j , loop ;
2014-03-10 09:51:56 +00:00
struct IOV_111 * iovPtr ;
2015-10-15 08:09:34 +00:00
struct fusion_context * fusion ;
2019-06-25 11:04:31 +00:00
bool intr_coalescing ;
unsigned int num_msix_req ;
2019-06-25 11:04:35 +00:00
u16 lnksta , speed ;
2015-10-15 08:09:34 +00:00
fusion = instance - > ctrl_context ;
2010-12-21 18:23:23 +00:00
/* Find first memory bar */
bar_list = pci_select_bars ( instance - > pdev , IORESOURCE_MEM ) ;
2016-08-21 08:28:25 +00:00
instance - > bar = find_first_bit ( & bar_list , BITS_PER_LONG ) ;
2016-08-06 06:37:34 +00:00
if ( pci_request_selected_regions ( instance - > pdev , 1 < < instance - > bar ,
2010-12-21 18:23:23 +00:00
" megasas: LSI " ) ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " IO memory region busy! \n " ) ;
2010-12-21 18:23:23 +00:00
return - EBUSY ;
}
2013-09-13 16:46:44 +00:00
base_addr = pci_resource_start ( instance - > pdev , instance - > bar ) ;
instance - > reg_set = ioremap_nocache ( base_addr , 8192 ) ;
2010-12-21 18:23:23 +00:00
if ( ! instance - > reg_set ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " Failed to map IO mem \n " ) ;
2010-12-21 18:23:23 +00:00
goto fail_ioremap ;
}
2019-05-07 17:05:42 +00:00
base_addr_phys = & base_addr ;
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev ,
" BAR:0x%lx BAR's base_addr(phys):%pa mapped virt_addr:0x%p \n " ,
instance - > bar , base_addr_phys , instance - > reg_set ) ;
2017-10-19 09:48:50 +00:00
if ( instance - > adapter_type ! = MFI_SERIES )
2010-12-21 21:34:31 +00:00
instance - > instancet = & megasas_instance_template_fusion ;
2017-01-10 23:20:49 +00:00
else {
switch ( instance - > pdev - > device ) {
case PCI_DEVICE_ID_LSI_SAS1078R :
case PCI_DEVICE_ID_LSI_SAS1078DE :
instance - > instancet = & megasas_instance_template_ppc ;
break ;
case PCI_DEVICE_ID_LSI_SAS1078GEN2 :
case PCI_DEVICE_ID_LSI_SAS0079GEN2 :
instance - > instancet = & megasas_instance_template_gen2 ;
break ;
case PCI_DEVICE_ID_LSI_SAS0073SKINNY :
case PCI_DEVICE_ID_LSI_SAS0071SKINNY :
instance - > instancet = & megasas_instance_template_skinny ;
break ;
case PCI_DEVICE_ID_LSI_SAS1064R :
case PCI_DEVICE_ID_DELL_PERC5 :
default :
instance - > instancet = & megasas_instance_template_xscale ;
instance - > pd_list_not_supported = 1 ;
break ;
}
2010-12-21 18:23:23 +00:00
}
2013-07-15 20:56:05 +00:00
if ( megasas_transition_to_ready ( instance , 0 ) ) {
2019-05-07 17:05:34 +00:00
dev_info ( & instance - > pdev - > dev ,
" Failed to transition controller to ready from %s! \n " ,
__func__ ) ;
2019-05-07 17:05:32 +00:00
if ( instance - > adapter_type ! = MFI_SERIES ) {
2018-10-17 06:37:42 +00:00
status_reg = instance - > instancet - > read_fw_status_reg (
2018-12-17 08:47:39 +00:00
instance ) ;
2019-05-07 17:05:32 +00:00
if ( status_reg & MFI_RESET_ADAPTER ) {
2019-05-07 17:05:34 +00:00
if ( megasas_adp_reset_wait_for_ready
( instance , true , 0 ) = = FAILED )
2019-05-07 17:05:32 +00:00
goto fail_ready_state ;
} else {
goto fail_ready_state ;
}
} else {
2018-10-17 06:37:42 +00:00
atomic_set ( & instance - > fw_reset_no_pci_access , 1 ) ;
instance - > instancet - > adp_reset
( instance , instance - > reg_set ) ;
atomic_set ( & instance - > fw_reset_no_pci_access , 0 ) ;
/*waiting for about 30 second before retry*/
ssleep ( 30 ) ;
2013-07-15 20:56:05 +00:00
2018-10-17 06:37:42 +00:00
if ( megasas_transition_to_ready ( instance , 0 ) )
goto fail_ready_state ;
}
2019-05-07 17:05:34 +00:00
dev_info ( & instance - > pdev - > dev ,
" FW restarted successfully from %s! \n " ,
__func__ ) ;
2013-07-15 20:56:05 +00:00
}
2010-12-21 18:23:23 +00:00
2017-10-19 09:48:59 +00:00
megasas_init_ctrl_params ( instance ) ;
2017-10-19 09:49:05 +00:00
if ( megasas_set_dma_mask ( instance ) )
2017-10-19 09:48:59 +00:00
goto fail_ready_state ;
if ( megasas_alloc_ctrl_mem ( instance ) )
goto fail_alloc_dma_buf ;
if ( megasas_alloc_ctrl_dma_buffers ( instance ) )
goto fail_alloc_dma_buf ;
fusion = instance - > ctrl_context ;
2018-12-17 08:47:37 +00:00
if ( instance - > adapter_type > = VENTURA_SERIES ) {
2018-10-17 06:37:51 +00:00
scratch_pad_2 =
2018-12-17 08:47:40 +00:00
megasas_readl ( instance ,
& instance - > reg_set - > outbound_scratch_pad_2 ) ;
2018-10-17 06:37:51 +00:00
instance - > max_raid_mapsize = ( ( scratch_pad_2 > >
2017-01-10 23:20:48 +00:00
MR_MAX_RAID_MAP_SIZE_OFFSET_SHIFT ) &
MR_MAX_RAID_MAP_SIZE_MASK ) ;
}
2013-05-22 07:04:14 +00:00
2019-07-26 07:32:14 +00:00
instance - > enable_sdev_max_qd = enable_sdev_max_qd ;
2019-06-25 11:04:29 +00:00
switch ( instance - > adapter_type ) {
case VENTURA_SERIES :
2019-06-25 11:04:28 +00:00
fusion - > pcie_bw_limitation = true ;
2019-06-25 11:04:29 +00:00
break ;
case AERO_SERIES :
fusion - > r56_div_offload = true ;
break ;
default :
break ;
}
2019-06-25 11:04:28 +00:00
2011-05-12 01:33:47 +00:00
/* Check if MSI-X is supported while in ready state */
2018-12-17 08:47:39 +00:00
msix_enable = ( instance - > instancet - > read_fw_status_reg ( instance ) &
2011-05-12 01:33:47 +00:00
0x4000000 ) > > 0x1a ;
2011-10-09 01:15:13 +00:00
if ( msix_enable & & ! msix_disable ) {
2016-12-02 11:52:23 +00:00
2018-12-17 08:47:40 +00:00
scratch_pad_1 = megasas_readl
( instance , & instance - > reg_set - > outbound_scratch_pad_1 ) ;
2011-10-09 01:15:13 +00:00
/* Check max MSI-X vectors */
2015-10-15 08:09:34 +00:00
if ( fusion ) {
2017-10-19 09:48:48 +00:00
if ( instance - > adapter_type = = THUNDERBOLT_SERIES ) {
/* Thunderbolt Series*/
2018-10-17 06:37:51 +00:00
instance - > msix_vectors = ( scratch_pad_1
2015-10-15 08:09:34 +00:00
& MR_MAX_REPLY_QUEUES_OFFSET ) + 1 ;
2018-10-17 06:37:46 +00:00
} else {
2018-10-17 06:37:51 +00:00
instance - > msix_vectors = ( ( scratch_pad_1
2015-10-15 08:09:34 +00:00
& MR_MAX_REPLY_QUEUES_EXT_OFFSET )
> > MR_MAX_REPLY_QUEUES_EXT_OFFSET_SHIFT ) + 1 ;
2018-10-17 06:37:46 +00:00
/*
* For Invader series , > 8 MSI - x vectors
* supported by FW / HW implies combined
* reply queue mode is enabled .
* For Ventura series , > 16 MSI - x vectors
* supported by FW / HW implies combined
* reply queue mode is enabled .
*/
switch ( instance - > adapter_type ) {
case INVADER_SERIES :
if ( instance - > msix_vectors > 8 )
instance - > msix_combined = true ;
break ;
2018-12-17 08:47:38 +00:00
case AERO_SERIES :
2018-10-17 06:37:46 +00:00
case VENTURA_SERIES :
if ( instance - > msix_vectors > 16 )
instance - > msix_combined = true ;
break ;
}
2017-01-10 23:20:44 +00:00
2016-01-28 15:34:28 +00:00
if ( rdpq_enable )
2018-10-17 06:37:51 +00:00
instance - > is_rdpq = ( scratch_pad_1 & MR_RDPQ_MODE_OFFSET ) ?
2016-01-28 15:34:28 +00:00
1 : 0 ;
2019-05-07 17:05:36 +00:00
2019-06-28 09:50:39 +00:00
if ( instance - > adapter_type > = INVADER_SERIES & &
! instance - > msix_combined ) {
2019-05-07 17:05:36 +00:00
instance - > msix_load_balance = true ;
instance - > smp_affinity_enable = false ;
}
2015-10-15 08:09:34 +00:00
/* Save 1-15 reply post index address to local memory
* Index 0 is already saved from reg offset
* MPI2_REPLY_POST_HOST_INDEX_OFFSET
*/
for ( loop = 1 ; loop < MR_MAX_MSIX_REG_ARRAY ; loop + + ) {
instance - > reply_post_host_index_addr [ loop ] =
( u32 __iomem * )
( ( u8 __iomem * ) instance - > reg_set +
MPI2_SUP_REPLY_POST_HOST_INDEX_OFFSET
+ ( loop * 0x10 ) ) ;
}
2013-05-22 07:04:14 +00:00
}
2019-05-07 17:05:48 +00:00
dev_info ( & instance - > pdev - > dev ,
" firmware supports msix \t : (%d) " ,
instance - > msix_vectors ) ;
2013-05-22 07:04:14 +00:00
if ( msix_vectors )
instance - > msix_vectors = min ( msix_vectors ,
instance - > msix_vectors ) ;
2015-10-15 08:09:34 +00:00
} else /* MFI adapters */
2011-10-09 01:15:13 +00:00
instance - > msix_vectors = 1 ;
2019-05-07 17:05:36 +00:00
2019-06-25 11:04:31 +00:00
/*
* For Aero ( if some conditions are met ) , driver will configure a
* few additional reply queues with interrupt coalescing enabled .
* These queues with interrupt coalescing enabled are called
* High IOPS queues and rest of reply queues ( based on number of
* logical CPUs ) are termed as Low latency queues .
*
* Total Number of reply queues = High IOPS queues + low latency queues
*
* For rest of fusion adapters , 1 additional reply queue will be
* reserved for management commands , rest of reply queues
* ( based on number of logical CPUs ) will be used for IOs and
* referenced as IO queues .
* Total Number of reply queues = 1 + IO queues
*
* MFI adapters supports single MSI - x so single reply queue
* will be used for IO and management commands .
*/
intr_coalescing = ( scratch_pad_1 & MR_INTR_COALESCING_SUPPORT_OFFSET ) ?
true : false ;
if ( intr_coalescing & &
( num_online_cpus ( ) > = MR_HIGH_IOPS_QUEUE_COUNT ) & &
( instance - > msix_vectors = = MEGASAS_MAX_MSIX_QUEUES ) )
2019-06-25 11:04:35 +00:00
instance - > perf_mode = MR_BALANCED_PERF_MODE ;
2014-08-18 06:01:44 +00:00
else
2019-06-25 11:04:35 +00:00
instance - > perf_mode = MR_LATENCY_PERF_MODE ;
if ( instance - > adapter_type = = AERO_SERIES ) {
pcie_capability_read_word ( instance - > pdev , PCI_EXP_LNKSTA , & lnksta ) ;
speed = lnksta & PCI_EXP_LNKSTA_CLS ;
/*
* For Aero , if PCIe link speed is < 16 GT / s , then driver should operate
* in latency perf mode and enable R1 PCI bandwidth algorithm
*/
if ( speed < 0x4 ) {
instance - > perf_mode = MR_LATENCY_PERF_MODE ;
fusion - > pcie_bw_limitation = true ;
}
/*
* Performance mode settings provided through module parameter - perf_mode will
* take affect only for :
* 1. Aero family of adapters .
* 2. When user sets module parameter - perf_mode in range of 0 - 2.
*/
if ( ( perf_mode > = MR_BALANCED_PERF_MODE ) & &
( perf_mode < = MR_LATENCY_PERF_MODE ) )
instance - > perf_mode = perf_mode ;
/*
* If intr coalescing is not supported by controller FW , then IOPS
* and Balanced modes are not feasible .
*/
if ( ! intr_coalescing )
instance - > perf_mode = MR_LATENCY_PERF_MODE ;
}
2019-06-25 11:04:31 +00:00
2019-06-25 11:04:35 +00:00
if ( instance - > perf_mode = = MR_BALANCED_PERF_MODE )
2019-06-25 11:04:31 +00:00
instance - > low_latency_index_start =
MR_HIGH_IOPS_QUEUE_COUNT ;
else
instance - > low_latency_index_start = 1 ;
num_msix_req = num_online_cpus ( ) + instance - > low_latency_index_start ;
instance - > msix_vectors = min ( num_msix_req ,
instance - > msix_vectors ) ;
megasas_alloc_irq_vectors ( instance ) ;
if ( ! instance - > msix_vectors )
2019-05-07 17:05:36 +00:00
instance - > msix_load_balance = false ;
2011-10-09 01:15:13 +00:00
}
2017-01-10 23:20:44 +00:00
/*
* MSI - X host index 0 is common for all adapter .
* It is used for all MPT based Adapters .
*/
if ( instance - > msix_combined ) {
instance - > reply_post_host_index_addr [ 0 ] =
( u32 * ) ( ( u8 * ) instance - > reg_set +
MPI2_SUP_REPLY_POST_HOST_INDEX_OFFSET ) ;
} else {
instance - > reply_post_host_index_addr [ 0 ] =
( u32 * ) ( ( u8 * ) instance - > reg_set +
MPI2_REPLY_POST_HOST_INDEX_OFFSET ) ;
}
2017-03-10 11:22:12 +00:00
if ( ! instance - > msix_vectors ) {
i = pci_alloc_irq_vectors ( instance - > pdev , 1 , 1 , PCI_IRQ_LEGACY ) ;
if ( i < 0 )
2018-10-17 06:37:44 +00:00
goto fail_init_adapter ;
2017-03-10 11:22:12 +00:00
}
2011-05-12 01:33:47 +00:00
2018-03-13 09:42:40 +00:00
megasas_setup_reply_map ( instance ) ;
2015-06-02 10:39:46 +00:00
dev_info ( & instance - > pdev - > dev ,
" current msix/online cpus \t : (%d/%d) \n " ,
instance - > msix_vectors , ( unsigned int ) num_online_cpus ( ) ) ;
2016-01-28 15:34:28 +00:00
dev_info ( & instance - > pdev - > dev ,
" RDPQ mode \t : (%s) \n " , instance - > is_rdpq ? " enabled " : " disabled " ) ;
2015-04-23 11:00:09 +00:00
2015-10-15 08:10:34 +00:00
tasklet_init ( & instance - > isr_tasklet , instance - > instancet - > tasklet ,
( unsigned long ) instance ) ;
2014-09-12 13:27:33 +00:00
/*
* Below are default value for legacy Firmware .
* non - fusion based controllers
*/
instance - > fw_supported_vd_count = MAX_LOGICAL_DRIVES ;
instance - > fw_supported_pd_count = MAX_PHYSICAL_DEVICES ;
2010-12-21 18:23:23 +00:00
/* Get operational params, sge flags, send init cmd to controller */
if ( instance - > instancet - > init_adapter ( instance ) )
2011-02-25 04:55:56 +00:00
goto fail_init_adapter ;
2010-12-21 18:23:23 +00:00
2018-12-17 08:47:37 +00:00
if ( instance - > adapter_type > = VENTURA_SERIES ) {
2018-10-17 06:37:51 +00:00
scratch_pad_3 =
2018-12-17 08:47:40 +00:00
megasas_readl ( instance ,
& instance - > reg_set - > outbound_scratch_pad_3 ) ;
2018-10-17 06:37:51 +00:00
if ( ( scratch_pad_3 & MR_NVME_PAGE_SIZE_MASK ) > =
2017-02-10 08:59:10 +00:00
MR_DEFAULT_NVME_PAGE_SHIFT )
instance - > nvme_page_size =
2018-10-17 06:37:51 +00:00
( 1 < < ( scratch_pad_3 & MR_NVME_PAGE_SIZE_MASK ) ) ;
2017-02-10 08:59:10 +00:00
dev_info ( & instance - > pdev - > dev ,
" NVME page size \t : (%d) \n " , instance - > nvme_page_size ) ;
}
2016-11-01 16:32:02 +00:00
if ( instance - > msix_vectors ?
megasas_setup_irqs_msix ( instance , 1 ) :
megasas_setup_irqs_ioapic ( instance ) )
goto fail_init_adapter ;
2015-06-02 10:39:46 +00:00
scsi: megaraid_sas: IRQ poll to avoid CPU hard lockups
Issue Description:
We have seen cpu lock up issues from field if system has a large (more than
96) logical cpu count. SAS3.0 controller (Invader series) supports max 96
MSI-X vector and SAS3.5 product (Ventura) supports max 128 MSI-X vectors.
This may be a generic issue (if PCI device support completion on multiple
reply queues).
Let me explain it w.r.t megaraid_sas supported h/w just to simplify the
problem and possible changes to handle such issues. MegaRAID controller
supports multiple reply queues in completion path. Driver creates MSI-X
vectors for controller as "minimum of (FW supported Reply queues, Logical
CPUs)". If submitter is not interrupted via completion on same CPU, there
is a loop in the IO path. This behavior can cause hard/soft CPU lockups, IO
timeout, system sluggish etc.
Example - one CPU (e.g. CPU A) is busy submitting the IOs and another CPU
(e.g. CPU B) is busy with processing the corresponding IO's reply
descriptors from reply descriptor queue upon receiving the interrupts from
HBA. If CPU A is continuously pumping the IOs then always CPU B (which is
executing the ISR) will see the valid reply descriptors in the reply
descriptor queue and it will be continuously processing those reply
descriptor in a loop without quitting the ISR handler.
megaraid_sas driver will exit ISR handler if it finds unused reply
descriptor in the reply descriptor queue. Since CPU A will be continuously
sending the IOs, CPU B may always see a valid reply descriptor (posted by
HBA Firmware after processing the IO) in the reply descriptor queue. In
worst case, driver will not quit from this loop in the ISR handler.
Eventually, CPU lockup will be detected by watchdog.
Above mentioned behavior is not common if "rq_affinity" set to 2 or
affinity_hint is honored by irqbalancer as "exact". If rq_affinity is set
to 2, submitter will be always interrupted via completion on same CPU. If
irqbalancer is using "exact" policy, interrupt will be delivered to
submitter CPU.
Problem statement:
If CPU count to MSI-X vectors (reply descriptor Queues) count ratio is not
1:1, we still have exposure of issue explained above and for that we don't
have any solution.
Exposure of soft/hard lockup is seen if CPU count is more than MSI-X
supported by device.
If CPUs count to MSI-X vectors count ratio is not 1:1, (Other way, if
CPU counts to MSI-X vector count ratio is something like X:1, where X > 1)
then 'exact' irqbalance policy OR rq_affinity = 2 won't help to avoid CPU
hard/soft lockups. There won't be any one to one mapping between
CPU to MSI-X vector instead one MSI-X interrupt (or reply descriptor queue)
is shared with group/set of CPUs and there is a possibility of having a
loop in the IO path within that CPU group and may observe lockups.
For example: Consider a system having two NUMA nodes and each node having
four logical CPUs and also consider that number of MSI-X vectors enabled on
the HBA is two, then CPUs count to MSI-X vector count ratio as 4:1.
e.g.
MSI-X vector 0 is affinity to CPU 0, CPU 1, CPU 2 & CPU 3 of NUMA node 0 and
MSI-X vector 1 is affinity to CPU 4, CPU 5, CPU 6 & CPU 7 of NUMA node 1.
numactl --hardware
available: 2 nodes (0-1)
node 0 cpus: 0 1 2 3 --> MSI-X 0
node 0 size: 65536 MB
node 0 free: 63176 MB
node 1 cpus: 4 5 6 7 --> MSI-X 1
node 1 size: 65536 MB
node 1 free: 63176 MB
Assume that user started an application which uses all the CPUs of NUMA
node 0 for issuing the IOs. Only one CPU from affinity list (it can be any
cpu since this behavior depends upon irqbalance) CPU0 will receive the
interrupts from MSI-X 0 for all the IOs. Eventually, CPU 0 IO submission
percentage will be decreasing and ISR processing percentage will be
increasing as it is more busy with processing the interrupts. Gradually IO
submission percentage on CPU 0 will be zero and it's ISR processing
percentage will be 100% as IO loop has already formed within the
NUMA node 0, i.e. CPU 1, CPU 2 & CPU 3 will be continuously busy with
submitting the heavy IOs and only CPU 0 is busy in the ISR path as it
always find the valid reply descriptor in the reply descriptor queue.
Eventually, we will observe the hard lockup here.
Chances of occurring of hard/soft lockups are directly proportional to
value of X. If value of X is high, then chances of observing CPU lockups is
high.
Solution:
Use IRQ poll interface defined in "irq_poll.c".
megaraid_sas driver will execute ISR routine in softirq context and it will
always quit the loop based on budget provided in IRQ poll interface.
Driver will switch to IRQ poll only when more than a threshold number of
reply descriptors are handled in one ISR. Currently threshold is set as
1/4th of HBA queue depth.
In these scenarios (i.e. where CPUs count to MSI-X vectors count ratio is
X:1 (where X > 1)), IRQ poll interface will avoid CPU hard lockups due to
voluntary exit from the reply queue processing based on budget.
Note - Only one MSI-X vector is busy doing processing.
Select CONFIG_IRQ_POLL from driver Kconfig for driver compilation.
Signed-off-by: Kashyap Desai <kashyap.desai@broadcom.com>
Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-05-07 17:05:35 +00:00
if ( instance - > adapter_type ! = MFI_SERIES )
megasas_setup_irq_poll ( instance ) ;
2015-04-23 11:00:09 +00:00
instance - > instancet - > enable_intr ( instance ) ;
2010-12-21 18:23:23 +00:00
2016-05-03 17:24:31 +00:00
dev_info ( & instance - > pdev - > dev , " INIT adapter done \n " ) ;
2010-12-21 18:23:23 +00:00
2015-08-31 11:53:11 +00:00
megasas_setup_jbod_map ( instance ) ;
2019-01-29 09:38:12 +00:00
if ( megasas_get_device_list ( instance ) ! = SUCCESS ) {
dev_err ( & instance - > pdev - > dev ,
" %s: megasas_get_device_list failed \n " ,
__func__ ) ;
2017-02-10 08:59:33 +00:00
goto fail_get_ld_pd_list ;
2014-01-16 10:25:36 +00:00
}
2009-10-06 20:27:54 +00:00
2017-01-10 23:20:46 +00:00
/* stream detection initialization */
2018-12-17 08:47:37 +00:00
if ( instance - > adapter_type > = VENTURA_SERIES ) {
2017-01-10 23:20:46 +00:00
fusion - > stream_detect_by_ld =
treewide: kzalloc() -> kcalloc()
The kzalloc() function has a 2-factor argument form, kcalloc(). This
patch replaces cases of:
kzalloc(a * b, gfp)
with:
kcalloc(a * b, gfp)
as well as handling cases of:
kzalloc(a * b * c, gfp)
with:
kzalloc(array3_size(a, b, c), gfp)
as it's slightly less ugly than:
kzalloc_array(array_size(a, b), c, gfp)
This does, however, attempt to ignore constant size factors like:
kzalloc(4 * 1024, gfp)
though any constants defined via macros get caught up in the conversion.
Any factors with a sizeof() of "unsigned char", "char", and "u8" were
dropped, since they're redundant.
The Coccinelle script used for this was:
// Fix redundant parens around sizeof().
@@
type TYPE;
expression THING, E;
@@
(
kzalloc(
- (sizeof(TYPE)) * E
+ sizeof(TYPE) * E
, ...)
|
kzalloc(
- (sizeof(THING)) * E
+ sizeof(THING) * E
, ...)
)
// Drop single-byte sizes and redundant parens.
@@
expression COUNT;
typedef u8;
typedef __u8;
@@
(
kzalloc(
- sizeof(u8) * (COUNT)
+ COUNT
, ...)
|
kzalloc(
- sizeof(__u8) * (COUNT)
+ COUNT
, ...)
|
kzalloc(
- sizeof(char) * (COUNT)
+ COUNT
, ...)
|
kzalloc(
- sizeof(unsigned char) * (COUNT)
+ COUNT
, ...)
|
kzalloc(
- sizeof(u8) * COUNT
+ COUNT
, ...)
|
kzalloc(
- sizeof(__u8) * COUNT
+ COUNT
, ...)
|
kzalloc(
- sizeof(char) * COUNT
+ COUNT
, ...)
|
kzalloc(
- sizeof(unsigned char) * COUNT
+ COUNT
, ...)
)
// 2-factor product with sizeof(type/expression) and identifier or constant.
@@
type TYPE;
expression THING;
identifier COUNT_ID;
constant COUNT_CONST;
@@
(
- kzalloc
+ kcalloc
(
- sizeof(TYPE) * (COUNT_ID)
+ COUNT_ID, sizeof(TYPE)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(TYPE) * COUNT_ID
+ COUNT_ID, sizeof(TYPE)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(TYPE) * (COUNT_CONST)
+ COUNT_CONST, sizeof(TYPE)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(TYPE) * COUNT_CONST
+ COUNT_CONST, sizeof(TYPE)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(THING) * (COUNT_ID)
+ COUNT_ID, sizeof(THING)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(THING) * COUNT_ID
+ COUNT_ID, sizeof(THING)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(THING) * (COUNT_CONST)
+ COUNT_CONST, sizeof(THING)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(THING) * COUNT_CONST
+ COUNT_CONST, sizeof(THING)
, ...)
)
// 2-factor product, only identifiers.
@@
identifier SIZE, COUNT;
@@
- kzalloc
+ kcalloc
(
- SIZE * COUNT
+ COUNT, SIZE
, ...)
// 3-factor product with 1 sizeof(type) or sizeof(expression), with
// redundant parens removed.
@@
expression THING;
identifier STRIDE, COUNT;
type TYPE;
@@
(
kzalloc(
- sizeof(TYPE) * (COUNT) * (STRIDE)
+ array3_size(COUNT, STRIDE, sizeof(TYPE))
, ...)
|
kzalloc(
- sizeof(TYPE) * (COUNT) * STRIDE
+ array3_size(COUNT, STRIDE, sizeof(TYPE))
, ...)
|
kzalloc(
- sizeof(TYPE) * COUNT * (STRIDE)
+ array3_size(COUNT, STRIDE, sizeof(TYPE))
, ...)
|
kzalloc(
- sizeof(TYPE) * COUNT * STRIDE
+ array3_size(COUNT, STRIDE, sizeof(TYPE))
, ...)
|
kzalloc(
- sizeof(THING) * (COUNT) * (STRIDE)
+ array3_size(COUNT, STRIDE, sizeof(THING))
, ...)
|
kzalloc(
- sizeof(THING) * (COUNT) * STRIDE
+ array3_size(COUNT, STRIDE, sizeof(THING))
, ...)
|
kzalloc(
- sizeof(THING) * COUNT * (STRIDE)
+ array3_size(COUNT, STRIDE, sizeof(THING))
, ...)
|
kzalloc(
- sizeof(THING) * COUNT * STRIDE
+ array3_size(COUNT, STRIDE, sizeof(THING))
, ...)
)
// 3-factor product with 2 sizeof(variable), with redundant parens removed.
@@
expression THING1, THING2;
identifier COUNT;
type TYPE1, TYPE2;
@@
(
kzalloc(
- sizeof(TYPE1) * sizeof(TYPE2) * COUNT
+ array3_size(COUNT, sizeof(TYPE1), sizeof(TYPE2))
, ...)
|
kzalloc(
- sizeof(TYPE1) * sizeof(THING2) * (COUNT)
+ array3_size(COUNT, sizeof(TYPE1), sizeof(TYPE2))
, ...)
|
kzalloc(
- sizeof(THING1) * sizeof(THING2) * COUNT
+ array3_size(COUNT, sizeof(THING1), sizeof(THING2))
, ...)
|
kzalloc(
- sizeof(THING1) * sizeof(THING2) * (COUNT)
+ array3_size(COUNT, sizeof(THING1), sizeof(THING2))
, ...)
|
kzalloc(
- sizeof(TYPE1) * sizeof(THING2) * COUNT
+ array3_size(COUNT, sizeof(TYPE1), sizeof(THING2))
, ...)
|
kzalloc(
- sizeof(TYPE1) * sizeof(THING2) * (COUNT)
+ array3_size(COUNT, sizeof(TYPE1), sizeof(THING2))
, ...)
)
// 3-factor product, only identifiers, with redundant parens removed.
@@
identifier STRIDE, SIZE, COUNT;
@@
(
kzalloc(
- (COUNT) * STRIDE * SIZE
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kzalloc(
- COUNT * (STRIDE) * SIZE
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kzalloc(
- COUNT * STRIDE * (SIZE)
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kzalloc(
- (COUNT) * (STRIDE) * SIZE
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kzalloc(
- COUNT * (STRIDE) * (SIZE)
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kzalloc(
- (COUNT) * STRIDE * (SIZE)
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kzalloc(
- (COUNT) * (STRIDE) * (SIZE)
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kzalloc(
- COUNT * STRIDE * SIZE
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
)
// Any remaining multi-factor products, first at least 3-factor products,
// when they're not all constants...
@@
expression E1, E2, E3;
constant C1, C2, C3;
@@
(
kzalloc(C1 * C2 * C3, ...)
|
kzalloc(
- (E1) * E2 * E3
+ array3_size(E1, E2, E3)
, ...)
|
kzalloc(
- (E1) * (E2) * E3
+ array3_size(E1, E2, E3)
, ...)
|
kzalloc(
- (E1) * (E2) * (E3)
+ array3_size(E1, E2, E3)
, ...)
|
kzalloc(
- E1 * E2 * E3
+ array3_size(E1, E2, E3)
, ...)
)
// And then all remaining 2 factors products when they're not all constants,
// keeping sizeof() as the second factor argument.
@@
expression THING, E1, E2;
type TYPE;
constant C1, C2, C3;
@@
(
kzalloc(sizeof(THING) * C2, ...)
|
kzalloc(sizeof(TYPE) * C2, ...)
|
kzalloc(C1 * C2 * C3, ...)
|
kzalloc(C1 * C2, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(TYPE) * (E2)
+ E2, sizeof(TYPE)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(TYPE) * E2
+ E2, sizeof(TYPE)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(THING) * (E2)
+ E2, sizeof(THING)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(THING) * E2
+ E2, sizeof(THING)
, ...)
|
- kzalloc
+ kcalloc
(
- (E1) * E2
+ E1, E2
, ...)
|
- kzalloc
+ kcalloc
(
- (E1) * (E2)
+ E1, E2
, ...)
|
- kzalloc
+ kcalloc
(
- E1 * E2
+ E1, E2
, ...)
)
Signed-off-by: Kees Cook <keescook@chromium.org>
2018-06-12 21:03:40 +00:00
kcalloc ( MAX_LOGICAL_DRIVES_EXT ,
sizeof ( struct LD_STREAM_DETECT * ) ,
GFP_KERNEL ) ;
2017-01-10 23:20:46 +00:00
if ( ! fusion - > stream_detect_by_ld ) {
dev_err ( & instance - > pdev - > dev ,
2017-02-10 08:59:37 +00:00
" unable to allocate stream detection for pool of LDs \n " ) ;
2017-01-10 23:20:46 +00:00
goto fail_get_ld_pd_list ;
}
for ( i = 0 ; i < MAX_LOGICAL_DRIVES_EXT ; + + i ) {
fusion - > stream_detect_by_ld [ i ] =
2018-01-05 13:27:36 +00:00
kzalloc ( sizeof ( struct LD_STREAM_DETECT ) ,
2017-01-10 23:20:46 +00:00
GFP_KERNEL ) ;
if ( ! fusion - > stream_detect_by_ld [ i ] ) {
dev_err ( & instance - > pdev - > dev ,
" unable to allocate stream detect by LD \n " ) ;
for ( j = 0 ; j < i ; + + j )
kfree ( fusion - > stream_detect_by_ld [ j ] ) ;
kfree ( fusion - > stream_detect_by_ld ) ;
fusion - > stream_detect_by_ld = NULL ;
goto fail_get_ld_pd_list ;
}
fusion - > stream_detect_by_ld [ i ] - > mru_bit_map
= MR_STREAM_BITMAP ;
}
}
2005-09-20 21:46:58 +00:00
/*
* Compute the max allowed sectors per IO : The controller info has two
* limits on max sectors . Driver should use the minimum of these two .
*
* 1 < < stripe_sz_ops . min = max sectors per strip
*
* Note that older firmwares ( < FW ver 30 ) didn ' t report information
* to calculate max_sectors_1 . So the number ended up as zero always .
*/
2007-11-09 09:14:00 +00:00
tmp_sectors = 0 ;
2017-10-19 09:48:57 +00:00
ctrl_info = instance - > ctrl_info_buf ;
2005-09-20 21:46:58 +00:00
2014-09-12 13:27:33 +00:00
max_sectors_1 = ( 1 < < ctrl_info - > stripe_sz_ops . min ) *
le16_to_cpu ( ctrl_info - > max_strips_per_io ) ;
max_sectors_2 = le32_to_cpu ( ctrl_info - > max_request_size ) ;
2013-05-22 07:05:33 +00:00
2015-07-07 20:52:45 +00:00
tmp_sectors = min_t ( u32 , max_sectors_1 , max_sectors_2 ) ;
2013-05-22 07:05:04 +00:00
2016-01-28 15:44:25 +00:00
instance - > peerIsPresent = ctrl_info - > cluster . peerIsPresent ;
instance - > passive = ctrl_info - > cluster . passive ;
memcpy ( instance - > clusterId , ctrl_info - > clusterId , sizeof ( instance - > clusterId ) ) ;
2014-09-12 13:27:33 +00:00
instance - > UnevenSpanSupport =
ctrl_info - > adapterOperations2 . supportUnevenSpans ;
if ( instance - > UnevenSpanSupport ) {
struct fusion_context * fusion = instance - > ctrl_context ;
2018-01-05 13:27:44 +00:00
if ( MR_ValidateMapInfo ( instance , instance - > map_id ) )
2014-09-12 13:27:33 +00:00
fusion - > fast_path_io = 1 ;
else
fusion - > fast_path_io = 0 ;
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
2014-09-12 13:27:33 +00:00
}
if ( ctrl_info - > host_interface . SRIOV ) {
2015-08-31 11:54:11 +00:00
instance - > requestorId = ctrl_info - > iov . requestorId ;
if ( instance - > pdev - > device = = PCI_DEVICE_ID_LSI_PLASMA ) {
if ( ! ctrl_info - > adapterOperations2 . activePassive )
instance - > PlasmaFW111 = 1 ;
dev_info ( & instance - > pdev - > dev , " SR-IOV: firmware type: %s \n " ,
instance - > PlasmaFW111 ? " 1.11 " : " new " ) ;
if ( instance - > PlasmaFW111 ) {
iovPtr = ( struct IOV_111 * )
( ( unsigned char * ) ctrl_info + IOV_111_OFFSET ) ;
instance - > requestorId = iovPtr - > requestorId ;
}
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
}
2015-08-31 11:54:11 +00:00
dev_info ( & instance - > pdev - > dev , " SRIOV: VF requestorId %d \n " ,
instance - > requestorId ) ;
2014-09-12 13:27:33 +00:00
}
instance - > crash_dump_fw_support =
ctrl_info - > adapterOperations3 . supportCrashDump ;
instance - > crash_dump_drv_support =
( instance - > crash_dump_fw_support & &
instance - > crash_dump_buf ) ;
2015-04-23 11:01:09 +00:00
if ( instance - > crash_dump_drv_support )
2014-09-12 13:27:33 +00:00
megasas_set_crash_dump_params ( instance ,
MR_CRASH_BUF_TURN_OFF ) ;
2015-04-23 11:01:09 +00:00
else {
2014-09-12 13:27:33 +00:00
if ( instance - > crash_dump_buf )
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & instance - > pdev - > dev ,
2014-09-12 13:27:33 +00:00
CRASH_DMA_BUF_SIZE ,
instance - > crash_dump_buf ,
instance - > crash_dump_h ) ;
instance - > crash_dump_buf = NULL ;
2007-11-09 09:14:00 +00:00
}
2015-01-05 14:36:03 +00:00
2018-10-17 06:37:40 +00:00
if ( instance - > snapdump_wait_time ) {
megasas_get_snapdump_properties ( instance ) ;
dev_info ( & instance - > pdev - > dev , " Snap dump wait time \t : %d \n " ,
instance - > snapdump_wait_time ) ;
}
2015-04-23 11:01:09 +00:00
dev_info ( & instance - > pdev - > dev ,
" pci id \t \t : (0x%04x)/(0x%04x)/(0x%04x)/(0x%04x) \n " ,
le16_to_cpu ( ctrl_info - > pci . vendor_id ) ,
le16_to_cpu ( ctrl_info - > pci . device_id ) ,
le16_to_cpu ( ctrl_info - > pci . sub_vendor_id ) ,
le16_to_cpu ( ctrl_info - > pci . sub_device_id ) ) ;
dev_info ( & instance - > pdev - > dev , " unevenspan support : %s \n " ,
instance - > UnevenSpanSupport ? " yes " : " no " ) ;
dev_info ( & instance - > pdev - > dev , " firmware crash dump : %s \n " ,
instance - > crash_dump_drv_support ? " yes " : " no " ) ;
2019-06-25 11:04:26 +00:00
dev_info ( & instance - > pdev - > dev , " JBOD sequence map : %s \n " ,
instance - > use_seqnum_jbod_fp ? " enabled " : " disabled " ) ;
2015-04-23 11:01:09 +00:00
2007-11-09 09:14:00 +00:00
instance - > max_sectors_per_req = instance - > max_num_sge *
2015-10-15 08:10:04 +00:00
SGE_BUFFER_SIZE / 512 ;
2007-11-09 09:14:00 +00:00
if ( tmp_sectors & & ( instance - > max_sectors_per_req > tmp_sectors ) )
instance - > max_sectors_per_req = tmp_sectors ;
2005-09-20 21:46:58 +00:00
2015-01-05 14:36:23 +00:00
/* Check for valid throttlequeuedepth module parameter */
if ( throttlequeuedepth & &
throttlequeuedepth < = instance - > max_scsi_cmds )
instance - > throttlequeuedepth = throttlequeuedepth ;
else
instance - > throttlequeuedepth =
MEGASAS_THROTTLE_QUEUE_DEPTH ;
2017-08-23 11:46:56 +00:00
if ( ( resetwaittime < 1 ) | |
( resetwaittime > MEGASAS_RESET_WAIT_TIME ) )
2016-01-28 15:34:34 +00:00
resetwaittime = MEGASAS_RESET_WAIT_TIME ;
if ( ( scmd_timeout < 10 ) | | ( scmd_timeout > MEGASAS_DEFAULT_CMD_TIMEOUT ) )
scmd_timeout = MEGASAS_DEFAULT_CMD_TIMEOUT ;
2007-11-09 09:40:16 +00:00
2014-03-10 09:51:56 +00:00
/* Launch SR-IOV heartbeat timer */
if ( instance - > requestorId ) {
2018-10-17 06:37:48 +00:00
if ( ! megasas_sriov_start_heartbeat ( instance , 1 ) ) {
2017-10-22 22:30:04 +00:00
megasas_start_timer ( instance ) ;
2018-10-17 06:37:48 +00:00
} else {
2014-03-10 09:51:56 +00:00
instance - > skip_heartbeat_timer_del = 1 ;
2018-10-17 06:37:48 +00:00
goto fail_get_ld_pd_list ;
}
2014-03-10 09:51:56 +00:00
}
2018-10-17 06:37:39 +00:00
/*
* Create and start watchdog thread which will monitor
* controller state every 1 sec and trigger OCR when
* it enters fault state
*/
if ( instance - > adapter_type ! = MFI_SERIES )
if ( megasas_fusion_start_watchdog ( instance ) ! = SUCCESS )
goto fail_start_watchdog ;
2005-09-20 21:46:58 +00:00
return 0 ;
2018-10-17 06:37:39 +00:00
fail_start_watchdog :
if ( instance - > requestorId & & ! instance - > skip_heartbeat_timer_del )
del_timer_sync ( & instance - > sriov_heartbeat_timer ) ;
2017-01-10 23:20:46 +00:00
fail_get_ld_pd_list :
instance - > instancet - > disable_intr ( instance ) ;
2016-12-02 11:52:23 +00:00
megasas_destroy_irqs ( instance ) ;
2018-10-17 06:37:44 +00:00
fail_init_adapter :
2015-04-23 11:00:09 +00:00
if ( instance - > msix_vectors )
2016-12-02 11:52:23 +00:00
pci_free_irq_vectors ( instance - > pdev ) ;
2015-04-23 11:00:09 +00:00
instance - > msix_vectors = 0 ;
2017-10-19 09:48:59 +00:00
fail_alloc_dma_buf :
megasas_free_ctrl_dma_buffers ( instance ) ;
megasas_free_ctrl_mem ( instance ) ;
2010-12-21 18:23:23 +00:00
fail_ready_state :
2005-09-20 21:46:58 +00:00
iounmap ( instance - > reg_set ) ;
2017-02-10 08:59:33 +00:00
fail_ioremap :
2016-08-06 06:37:34 +00:00
pci_release_selected_regions ( instance - > pdev , 1 < < instance - > bar ) ;
2005-09-20 21:46:58 +00:00
2017-02-10 08:59:33 +00:00
dev_err ( & instance - > pdev - > dev , " Failed from %s %d \n " ,
__func__ , __LINE__ ) ;
2005-09-20 21:46:58 +00:00
return - EINVAL ;
}
/**
* megasas_release_mfi - Reverses the FW initialization
2015-03-03 10:58:07 +00:00
* @ instance : Adapter soft state
2005-09-20 21:46:58 +00:00
*/
static void megasas_release_mfi ( struct megasas_instance * instance )
{
2010-12-21 21:34:31 +00:00
u32 reply_q_sz = sizeof ( u32 ) * ( instance - > max_mfi_cmds + 1 ) ;
2005-09-20 21:46:58 +00:00
2010-12-21 21:34:31 +00:00
if ( instance - > reply_queue )
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & instance - > pdev - > dev , reply_q_sz ,
2005-09-20 21:46:58 +00:00
instance - > reply_queue , instance - > reply_queue_h ) ;
megasas_free_cmds ( instance ) ;
iounmap ( instance - > reg_set ) ;
2016-08-06 06:37:34 +00:00
pci_release_selected_regions ( instance - > pdev , 1 < < instance - > bar ) ;
2005-09-20 21:46:58 +00:00
}
/**
* megasas_get_seq_num - Gets latest event sequence numbers
* @ instance : Adapter soft state
* @ eli : FW event log sequence numbers information
*
* FW maintains a log of all events in a non - volatile area . Upper layers would
* usually find out the latest sequence number of the events , the seq number at
* the boot etc . They would " read " all the events below the latest seq number
* by issuing a direct fw cmd ( DCMD ) . For the future events ( beyond latest seq
* number ) , they would subsribe to AEN ( asynchronous event notification ) and
* wait for the events to happen .
*/
static int
megasas_get_seq_num ( struct megasas_instance * instance ,
struct megasas_evt_log_info * eli )
{
struct megasas_cmd * cmd ;
struct megasas_dcmd_frame * dcmd ;
struct megasas_evt_log_info * el_info ;
dma_addr_t el_info_h = 0 ;
2018-01-05 13:27:38 +00:00
int ret ;
2005-09-20 21:46:58 +00:00
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd ) {
return - ENOMEM ;
}
dcmd = & cmd - > frame - > dcmd ;
cross-tree: phase out dma_zalloc_coherent()
We already need to zero out memory for dma_alloc_coherent(), as such
using dma_zalloc_coherent() is superflous. Phase it out.
This change was generated with the following Coccinelle SmPL patch:
@ replace_dma_zalloc_coherent @
expression dev, size, data, handle, flags;
@@
-dma_zalloc_coherent(dev, size, handle, flags)
+dma_alloc_coherent(dev, size, handle, flags)
Suggested-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
[hch: re-ran the script on the latest tree]
Signed-off-by: Christoph Hellwig <hch@lst.de>
2019-01-04 08:23:09 +00:00
el_info = dma_alloc_coherent ( & instance - > pdev - > dev ,
sizeof ( struct megasas_evt_log_info ) ,
& el_info_h , GFP_KERNEL ) ;
2005-09-20 21:46:58 +00:00
if ( ! el_info ) {
megasas_return_cmd ( instance , cmd ) ;
return - ENOMEM ;
}
memset ( dcmd - > mbox . b , 0 , MFI_MBOX_SIZE ) ;
dcmd - > cmd = MFI_CMD_DCMD ;
dcmd - > cmd_status = 0x0 ;
dcmd - > sge_count = 1 ;
2017-10-19 09:49:05 +00:00
dcmd - > flags = MFI_FRAME_DIR_READ ;
2005-09-20 21:46:58 +00:00
dcmd - > timeout = 0 ;
2009-12-06 15:24:21 +00:00
dcmd - > pad_0 = 0 ;
2013-09-06 10:20:52 +00:00
dcmd - > data_xfer_len = cpu_to_le32 ( sizeof ( struct megasas_evt_log_info ) ) ;
dcmd - > opcode = cpu_to_le32 ( MR_DCMD_CTRL_EVENT_GET_INFO ) ;
2017-10-19 09:49:05 +00:00
megasas_set_dma_settings ( instance , dcmd , el_info_h ,
sizeof ( struct megasas_evt_log_info ) ) ;
2005-09-20 21:46:58 +00:00
2018-01-05 13:27:38 +00:00
ret = megasas_issue_blocked_cmd ( instance , cmd , MFI_IO_TIMEOUT_SECS ) ;
if ( ret ! = DCMD_SUCCESS ) {
dev_err ( & instance - > pdev - > dev , " Failed from %s %d \n " ,
__func__ , __LINE__ ) ;
goto dcmd_failed ;
}
2005-09-20 21:46:58 +00:00
2018-01-05 13:27:38 +00:00
/*
* Copy the data back into callers buffer
*/
eli - > newest_seq_num = el_info - > newest_seq_num ;
eli - > oldest_seq_num = el_info - > oldest_seq_num ;
eli - > clear_seq_num = el_info - > clear_seq_num ;
eli - > shutdown_seq_num = el_info - > shutdown_seq_num ;
eli - > boot_seq_num = el_info - > boot_seq_num ;
2005-09-20 21:46:58 +00:00
2018-01-05 13:27:38 +00:00
dcmd_failed :
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & instance - > pdev - > dev ,
sizeof ( struct megasas_evt_log_info ) ,
el_info , el_info_h ) ;
2005-09-20 21:46:58 +00:00
2015-04-23 11:01:24 +00:00
megasas_return_cmd ( instance , cmd ) ;
2005-09-20 21:46:58 +00:00
2018-01-05 13:27:38 +00:00
return ret ;
2005-09-20 21:46:58 +00:00
}
/**
* megasas_register_aen - Registers for asynchronous event notification
* @ instance : Adapter soft state
* @ seq_num : The starting sequence number
* @ class_locale : Class of the event
*
* This function subscribes for AEN for events beyond the @ seq_num . It requests
* to be notified if and only if the event is of type @ class_locale
*/
static int
megasas_register_aen ( struct megasas_instance * instance , u32 seq_num ,
u32 class_locale_word )
{
int ret_val ;
struct megasas_cmd * cmd ;
struct megasas_dcmd_frame * dcmd ;
union megasas_evt_class_locale curr_aen ;
union megasas_evt_class_locale prev_aen ;
/*
* If there an AEN pending already ( aen_cmd ) , check if the
* class_locale of that pending AEN is inclusive of the new
* AEN request we currently have . If it is , then we don ' t have
* to do anything . In other words , whichever events the current
* AEN request is subscribing to , have already been subscribed
* to .
*
* If the old_cmd is _not_ inclusive , then we have to abort
* that command , form a class_locale that is superset of both
* old and current and re - issue to the FW
*/
curr_aen . word = class_locale_word ;
if ( instance - > aen_cmd ) {
2015-04-23 11:04:24 +00:00
prev_aen . word =
le32_to_cpu ( instance - > aen_cmd - > frame - > dcmd . mbox . w [ 1 ] ) ;
2005-09-20 21:46:58 +00:00
2017-08-23 11:47:01 +00:00
if ( ( curr_aen . members . class < MFI_EVT_CLASS_DEBUG ) | |
( curr_aen . members . class > MFI_EVT_CLASS_DEAD ) ) {
dev_info ( & instance - > pdev - > dev ,
" %s %d out of range class %d send by application \n " ,
__func__ , __LINE__ , curr_aen . members . class ) ;
return 0 ;
}
2005-09-20 21:46:58 +00:00
/*
* A class whose enum value is smaller is inclusive of all
* higher values . If a PROGRESS ( = - 1 ) was previously
* registered , then a new registration requests for higher
* classes need not be sent to FW . They are automatically
* included .
*
* Locale numbers don ' t have such hierarchy . They are bitmap
* values
*/
if ( ( prev_aen . members . class < = curr_aen . members . class ) & &
2013-09-16 09:48:06 +00:00
! ( ( prev_aen . members . locale & curr_aen . members . locale ) ^
2005-09-20 21:46:58 +00:00
curr_aen . members . locale ) ) {
/*
* Previously issued event registration includes
* current request . Nothing to do .
*/
return 0 ;
} else {
2013-09-16 09:48:06 +00:00
curr_aen . members . locale | = prev_aen . members . locale ;
2005-09-20 21:46:58 +00:00
if ( prev_aen . members . class < curr_aen . members . class )
curr_aen . members . class = prev_aen . members . class ;
instance - > aen_cmd - > abort_aen = 1 ;
ret_val = megasas_issue_blocked_abort_cmd ( instance ,
instance - >
2014-02-12 18:06:15 +00:00
aen_cmd , 30 ) ;
2005-09-20 21:46:58 +00:00
if ( ret_val ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " Failed to abort "
2005-09-20 21:46:58 +00:00
" previous AEN command \n " ) ;
return ret_val ;
}
}
}
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd )
return - ENOMEM ;
dcmd = & cmd - > frame - > dcmd ;
memset ( instance - > evt_detail , 0 , sizeof ( struct megasas_evt_detail ) ) ;
/*
* Prepare DCMD for aen registration
*/
memset ( dcmd - > mbox . b , 0 , MFI_MBOX_SIZE ) ;
dcmd - > cmd = MFI_CMD_DCMD ;
dcmd - > cmd_status = 0x0 ;
dcmd - > sge_count = 1 ;
2017-10-19 09:49:05 +00:00
dcmd - > flags = MFI_FRAME_DIR_READ ;
2005-09-20 21:46:58 +00:00
dcmd - > timeout = 0 ;
2009-12-06 15:24:21 +00:00
dcmd - > pad_0 = 0 ;
2013-09-06 10:20:52 +00:00
dcmd - > data_xfer_len = cpu_to_le32 ( sizeof ( struct megasas_evt_detail ) ) ;
dcmd - > opcode = cpu_to_le32 ( MR_DCMD_CTRL_EVENT_WAIT ) ;
dcmd - > mbox . w [ 0 ] = cpu_to_le32 ( seq_num ) ;
2010-09-23 02:36:29 +00:00
instance - > last_seq_num = seq_num ;
2013-09-06 10:20:52 +00:00
dcmd - > mbox . w [ 1 ] = cpu_to_le32 ( curr_aen . word ) ;
2017-10-19 09:49:05 +00:00
megasas_set_dma_settings ( instance , dcmd , instance - > evt_detail_h ,
sizeof ( struct megasas_evt_detail ) ) ;
2005-09-20 21:46:58 +00:00
2009-10-06 20:43:28 +00:00
if ( instance - > aen_cmd ! = NULL ) {
megasas_return_cmd ( instance , cmd ) ;
return 0 ;
}
2005-09-20 21:46:58 +00:00
/*
* Store reference to the cmd used to register for AEN . When an
* application wants us to register for AEN , we have to abort this
* cmd and re - register with a new EVENT LOCALE supplied by that app
*/
instance - > aen_cmd = cmd ;
/*
* Issue the aen registration frame
*/
2010-12-21 21:34:31 +00:00
instance - > instancet - > issue_dcmd ( instance , cmd ) ;
2005-09-20 21:46:58 +00:00
return 0 ;
}
2017-02-10 08:59:11 +00:00
/* megasas_get_target_prop - Send DCMD with below details to firmware.
*
* This DCMD will fetch few properties of LD / system PD defined
* in MR_TARGET_DEV_PROPERTIES . eg . Queue Depth , MDTS value .
*
* DCMD send by drivers whenever new target is added to the OS .
*
* dcmd . opcode - MR_DCMD_DEV_GET_TARGET_PROP
* dcmd . mbox . b [ 0 ] - DCMD is to be fired for LD or system PD .
* 0 = system PD , 1 = LD .
* dcmd . mbox . s [ 1 ] - TargetID for LD / system PD .
* dcmd . sge IN - Pointer to return MR_TARGET_DEV_PROPERTIES .
*
* @ instance : Adapter soft state
* @ sdev : OS provided scsi device
*
* Returns 0 on success non - zero on failure .
*/
2018-06-04 10:45:12 +00:00
int
2017-02-10 08:59:11 +00:00
megasas_get_target_prop ( struct megasas_instance * instance ,
struct scsi_device * sdev )
{
int ret ;
struct megasas_cmd * cmd ;
struct megasas_dcmd_frame * dcmd ;
2019-06-29 01:02:12 +00:00
u16 targetId = ( ( sdev - > channel % 2 ) * MEGASAS_MAX_DEV_PER_CHANNEL ) +
sdev - > id ;
2017-02-10 08:59:11 +00:00
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd ) {
dev_err ( & instance - > pdev - > dev ,
" Failed to get cmd %s \n " , __func__ ) ;
return - ENOMEM ;
}
dcmd = & cmd - > frame - > dcmd ;
memset ( instance - > tgt_prop , 0 , sizeof ( * instance - > tgt_prop ) ) ;
memset ( dcmd - > mbox . b , 0 , MFI_MBOX_SIZE ) ;
dcmd - > mbox . b [ 0 ] = MEGASAS_IS_LOGICAL ( sdev ) ;
dcmd - > mbox . s [ 1 ] = cpu_to_le16 ( targetId ) ;
dcmd - > cmd = MFI_CMD_DCMD ;
dcmd - > cmd_status = 0xFF ;
dcmd - > sge_count = 1 ;
2017-10-19 09:49:05 +00:00
dcmd - > flags = MFI_FRAME_DIR_READ ;
2017-02-10 08:59:11 +00:00
dcmd - > timeout = 0 ;
dcmd - > pad_0 = 0 ;
dcmd - > data_xfer_len =
cpu_to_le32 ( sizeof ( struct MR_TARGET_PROPERTIES ) ) ;
dcmd - > opcode = cpu_to_le32 ( MR_DCMD_DRV_GET_TARGET_PROP ) ;
2017-10-19 09:49:05 +00:00
megasas_set_dma_settings ( instance , dcmd , instance - > tgt_prop_h ,
sizeof ( struct MR_TARGET_PROPERTIES ) ) ;
2017-02-10 08:59:11 +00:00
2017-10-19 09:48:50 +00:00
if ( ( instance - > adapter_type ! = MFI_SERIES ) & &
! instance - > mask_interrupts )
2017-02-10 08:59:11 +00:00
ret = megasas_issue_blocked_cmd ( instance ,
cmd , MFI_IO_TIMEOUT_SECS ) ;
else
ret = megasas_issue_polled ( instance , cmd ) ;
switch ( ret ) {
case DCMD_TIMEOUT :
switch ( dcmd_timeout_ocr_possible ( instance ) ) {
case INITIATE_OCR :
cmd - > flags | = DRV_DCMD_SKIP_REFIRE ;
2019-06-25 11:04:23 +00:00
mutex_unlock ( & instance - > reset_mutex ) ;
2017-02-10 08:59:11 +00:00
megasas_reset_fusion ( instance - > host ,
MFI_IO_TIMEOUT_OCR ) ;
2019-06-25 11:04:23 +00:00
mutex_lock ( & instance - > reset_mutex ) ;
2017-02-10 08:59:11 +00:00
break ;
case KILL_ADAPTER :
megaraid_sas_kill_hba ( instance ) ;
break ;
case IGNORE_TIMEOUT :
dev_info ( & instance - > pdev - > dev ,
" Ignore DCMD timeout: %s %d \n " ,
__func__ , __LINE__ ) ;
break ;
}
break ;
default :
megasas_return_cmd ( instance , cmd ) ;
}
if ( ret ! = DCMD_SUCCESS )
dev_err ( & instance - > pdev - > dev ,
" return from %s %d return value %d \n " ,
__func__ , __LINE__ , ret ) ;
return ret ;
}
2005-09-20 21:46:58 +00:00
/**
* megasas_start_aen - Subscribes to AEN during driver load time
* @ instance : Adapter soft state
*/
static int megasas_start_aen ( struct megasas_instance * instance )
{
struct megasas_evt_log_info eli ;
union megasas_evt_class_locale class_locale ;
/*
* Get the latest sequence number from FW
*/
memset ( & eli , 0 , sizeof ( eli ) ) ;
if ( megasas_get_seq_num ( instance , & eli ) )
return - 1 ;
/*
* Register AEN with FW for latest sequence number plus 1
*/
class_locale . members . reserved = 0 ;
class_locale . members . locale = MR_EVT_LOCALE_ALL ;
class_locale . members . class = MR_EVT_CLASS_DEBUG ;
2013-09-06 10:20:52 +00:00
return megasas_register_aen ( instance ,
2015-04-23 11:03:24 +00:00
le32_to_cpu ( eli . newest_seq_num ) + 1 ,
2013-09-06 10:20:52 +00:00
class_locale . word ) ;
2005-09-20 21:46:58 +00:00
}
/**
* megasas_io_attach - Attaches this driver to SCSI mid - layer
* @ instance : Adapter soft state
*/
static int megasas_io_attach ( struct megasas_instance * instance )
{
struct Scsi_Host * host = instance - > host ;
/*
* Export parameters required by SCSI mid - layer
*/
host - > unique_id = instance - > unique_id ;
2015-01-05 14:36:23 +00:00
host - > can_queue = instance - > max_scsi_cmds ;
2005-09-20 21:46:58 +00:00
host - > this_id = instance - > init_id ;
host - > sg_tablesize = instance - > max_num_sge ;
2011-02-25 04:57:09 +00:00
if ( instance - > fw_support_ieee )
instance - > max_sectors_per_req = MEGASAS_MAX_SECTORS_IEEE ;
2010-10-12 13:18:50 +00:00
/*
* Check if the module parameter value for max_sectors can be used
*/
if ( max_sectors & & max_sectors < instance - > max_sectors_per_req )
instance - > max_sectors_per_req = max_sectors ;
else {
if ( max_sectors ) {
if ( ( ( instance - > pdev - > device = =
PCI_DEVICE_ID_LSI_SAS1078GEN2 ) | |
( instance - > pdev - > device = =
PCI_DEVICE_ID_LSI_SAS0079GEN2 ) ) & &
( max_sectors < = MEGASAS_MAX_SECTORS ) ) {
instance - > max_sectors_per_req = max_sectors ;
} else {
2015-07-07 20:52:34 +00:00
dev_info ( & instance - > pdev - > dev , " max_sectors should be > 0 "
2010-10-12 13:18:50 +00:00
" and <= %d (or < 1MB for GEN2 controller) \n " ,
instance - > max_sectors_per_req ) ;
}
}
}
2005-09-20 21:46:58 +00:00
host - > max_sectors = instance - > max_sectors_per_req ;
2010-12-21 21:34:31 +00:00
host - > cmd_per_lun = MEGASAS_DEFAULT_CMD_PER_LUN ;
2005-09-20 21:46:58 +00:00
host - > max_channel = MEGASAS_MAX_CHANNELS - 1 ;
host - > max_id = MEGASAS_MAX_DEV_PER_CHANNEL ;
host - > max_lun = MEGASAS_MAX_LUN ;
2006-02-03 23:34:17 +00:00
host - > max_cmd_len = 16 ;
2005-09-20 21:46:58 +00:00
/*
* Notify the mid - layer about the new controller
*/
if ( scsi_add_host ( host , & instance - > pdev - > dev ) ) {
2015-04-23 11:01:24 +00:00
dev_err ( & instance - > pdev - > dev ,
" Failed to add host from %s %d \n " ,
__func__ , __LINE__ ) ;
2005-09-20 21:46:58 +00:00
return - ENODEV ;
}
return 0 ;
}
2017-10-19 09:49:05 +00:00
/**
* megasas_set_dma_mask - Set DMA mask for supported controllers
*
* @ instance : Adapter soft state
* Description :
*
2018-12-18 13:59:54 +00:00
* For Ventura , driver / FW will operate in 63 bit DMA addresses .
2017-10-19 09:49:05 +00:00
*
* For invader -
* By default , driver / FW will operate in 32 bit DMA addresses
* for consistent DMA mapping but if 32 bit consistent
2018-12-18 13:59:54 +00:00
* DMA mask fails , driver will try with 63 bit consistent
* mask provided FW is true 63 bit DMA capable
2017-10-19 09:49:05 +00:00
*
* For older controllers ( Thunderbolt and MFI based adapters ) -
* driver / FW will operate in 32 bit consistent DMA addresses .
*/
2007-11-07 17:09:50 +00:00
static int
2017-10-19 09:49:05 +00:00
megasas_set_dma_mask ( struct megasas_instance * instance )
2007-11-07 17:09:50 +00:00
{
2017-10-19 09:49:05 +00:00
u64 consistent_mask ;
struct pci_dev * pdev ;
2018-10-17 06:37:51 +00:00
u32 scratch_pad_1 ;
2017-10-19 09:49:05 +00:00
pdev = instance - > pdev ;
2018-12-17 08:47:37 +00:00
consistent_mask = ( instance - > adapter_type > = VENTURA_SERIES ) ?
2018-12-18 13:59:54 +00:00
DMA_BIT_MASK ( 63 ) : DMA_BIT_MASK ( 32 ) ;
2017-10-19 09:49:05 +00:00
2007-11-07 17:09:50 +00:00
if ( IS_DMA64 ) {
2018-12-18 13:59:54 +00:00
if ( dma_set_mask ( & pdev - > dev , DMA_BIT_MASK ( 63 ) ) & &
2017-10-19 09:49:05 +00:00
dma_set_mask_and_coherent ( & pdev - > dev , DMA_BIT_MASK ( 32 ) ) )
goto fail_set_dma_mask ;
2018-12-18 13:59:54 +00:00
if ( ( * pdev - > dev . dma_mask = = DMA_BIT_MASK ( 63 ) ) & &
2017-10-19 09:49:05 +00:00
( dma_set_coherent_mask ( & pdev - > dev , consistent_mask ) & &
dma_set_mask_and_coherent ( & pdev - > dev , DMA_BIT_MASK ( 32 ) ) ) ) {
/*
* If 32 bit DMA mask fails , then try for 64 bit mask
* for FW capable of handling 64 bit DMA .
*/
2018-12-17 08:47:40 +00:00
scratch_pad_1 = megasas_readl
( instance , & instance - > reg_set - > outbound_scratch_pad_1 ) ;
2007-11-07 17:09:50 +00:00
2018-10-17 06:37:51 +00:00
if ( ! ( scratch_pad_1 & MR_CAN_HANDLE_64_BIT_DMA_OFFSET ) )
2017-10-19 09:49:05 +00:00
goto fail_set_dma_mask ;
else if ( dma_set_mask_and_coherent ( & pdev - > dev ,
2018-12-18 13:59:54 +00:00
DMA_BIT_MASK ( 63 ) ) )
2007-11-07 17:09:50 +00:00
goto fail_set_dma_mask ;
}
2017-10-19 09:49:05 +00:00
} else if ( dma_set_mask_and_coherent ( & pdev - > dev , DMA_BIT_MASK ( 32 ) ) )
goto fail_set_dma_mask ;
if ( pdev - > dev . coherent_dma_mask = = DMA_BIT_MASK ( 32 ) )
instance - > consistent_mask_64bit = false ;
else
instance - > consistent_mask_64bit = true ;
dev_info ( & pdev - > dev , " %s bit DMA mask and %s bit consistent mask \n " ,
2019-01-02 15:07:25 +00:00
( ( * pdev - > dev . dma_mask = = DMA_BIT_MASK ( 63 ) ) ? " 63 " : " 32 " ) ,
2018-12-18 13:59:54 +00:00
( instance - > consistent_mask_64bit ? " 63 " : " 32 " ) ) ;
2013-09-06 10:20:52 +00:00
2007-11-07 17:09:50 +00:00
return 0 ;
fail_set_dma_mask :
2017-10-19 09:49:05 +00:00
dev_err ( & pdev - > dev , " Failed to set DMA mask \n " ) ;
return - 1 ;
2007-11-07 17:09:50 +00:00
}
2017-10-19 09:48:48 +00:00
/*
* megasas_set_adapter_type - Set adapter type .
* Supported controllers can be divided in
2018-12-17 08:47:38 +00:00
* different categories -
* enum MR_ADAPTER_TYPE {
* MFI_SERIES = 1 ,
* THUNDERBOLT_SERIES = 2 ,
* INVADER_SERIES = 3 ,
* VENTURA_SERIES = 4 ,
* AERO_SERIES = 5 ,
* } ;
2017-10-19 09:48:48 +00:00
* @ instance : Adapter soft state
* return : void
2005-09-20 21:46:58 +00:00
*/
2017-10-19 09:48:48 +00:00
static inline void megasas_set_adapter_type ( struct megasas_instance * instance )
2005-09-20 21:46:58 +00:00
{
2017-10-19 09:48:49 +00:00
if ( ( instance - > pdev - > vendor = = PCI_VENDOR_ID_DELL ) & &
( instance - > pdev - > device = = PCI_DEVICE_ID_DELL_PERC5 ) ) {
2017-10-19 09:48:48 +00:00
instance - > adapter_type = MFI_SERIES ;
2017-10-19 09:48:49 +00:00
} else {
switch ( instance - > pdev - > device ) {
2018-11-09 17:47:20 +00:00
case PCI_DEVICE_ID_LSI_AERO_10E1 :
case PCI_DEVICE_ID_LSI_AERO_10E2 :
case PCI_DEVICE_ID_LSI_AERO_10E5 :
case PCI_DEVICE_ID_LSI_AERO_10E6 :
2018-12-17 08:47:38 +00:00
instance - > adapter_type = AERO_SERIES ;
break ;
2017-10-19 09:48:49 +00:00
case PCI_DEVICE_ID_LSI_VENTURA :
case PCI_DEVICE_ID_LSI_CRUSADER :
case PCI_DEVICE_ID_LSI_HARPOON :
case PCI_DEVICE_ID_LSI_TOMCAT :
case PCI_DEVICE_ID_LSI_VENTURA_4PORT :
case PCI_DEVICE_ID_LSI_CRUSADER_4PORT :
instance - > adapter_type = VENTURA_SERIES ;
break ;
case PCI_DEVICE_ID_LSI_FUSION :
case PCI_DEVICE_ID_LSI_PLASMA :
instance - > adapter_type = THUNDERBOLT_SERIES ;
break ;
case PCI_DEVICE_ID_LSI_INVADER :
case PCI_DEVICE_ID_LSI_INTRUDER :
case PCI_DEVICE_ID_LSI_INTRUDER_24 :
case PCI_DEVICE_ID_LSI_CUTLASS_52 :
case PCI_DEVICE_ID_LSI_CUTLASS_53 :
case PCI_DEVICE_ID_LSI_FURY :
instance - > adapter_type = INVADER_SERIES ;
break ;
default : /* For all other supported controllers */
instance - > adapter_type = MFI_SERIES ;
break ;
2011-02-25 04:56:28 +00:00
}
}
2017-10-19 09:48:48 +00:00
}
2005-09-20 21:46:58 +00:00
2017-10-19 09:48:54 +00:00
static inline int megasas_alloc_mfi_ctrl_mem ( struct megasas_instance * instance )
{
2018-10-10 17:31:25 +00:00
instance - > producer = dma_alloc_coherent ( & instance - > pdev - > dev ,
sizeof ( u32 ) , & instance - > producer_h , GFP_KERNEL ) ;
instance - > consumer = dma_alloc_coherent ( & instance - > pdev - > dev ,
sizeof ( u32 ) , & instance - > consumer_h , GFP_KERNEL ) ;
2005-09-20 21:46:58 +00:00
2017-10-19 09:48:54 +00:00
if ( ! instance - > producer | | ! instance - > consumer ) {
dev_err ( & instance - > pdev - > dev ,
" Failed to allocate memory for producer, consumer \n " ) ;
return - 1 ;
2005-09-20 21:46:58 +00:00
}
2017-10-19 09:48:54 +00:00
* instance - > producer = 0 ;
* instance - > consumer = 0 ;
return 0 ;
}
2005-09-20 21:46:58 +00:00
2017-10-19 09:48:54 +00:00
/**
* megasas_alloc_ctrl_mem - Allocate per controller memory for core data
* structures which are not common across MFI
* adapters and fusion adapters .
* For MFI based adapters , allocate producer and
* consumer buffers . For fusion adapters , allocate
* memory for fusion context .
* @ instance : Adapter soft state
* return : 0 for SUCCESS
*/
static int megasas_alloc_ctrl_mem ( struct megasas_instance * instance )
{
treewide: kzalloc() -> kcalloc()
The kzalloc() function has a 2-factor argument form, kcalloc(). This
patch replaces cases of:
kzalloc(a * b, gfp)
with:
kcalloc(a * b, gfp)
as well as handling cases of:
kzalloc(a * b * c, gfp)
with:
kzalloc(array3_size(a, b, c), gfp)
as it's slightly less ugly than:
kzalloc_array(array_size(a, b), c, gfp)
This does, however, attempt to ignore constant size factors like:
kzalloc(4 * 1024, gfp)
though any constants defined via macros get caught up in the conversion.
Any factors with a sizeof() of "unsigned char", "char", and "u8" were
dropped, since they're redundant.
The Coccinelle script used for this was:
// Fix redundant parens around sizeof().
@@
type TYPE;
expression THING, E;
@@
(
kzalloc(
- (sizeof(TYPE)) * E
+ sizeof(TYPE) * E
, ...)
|
kzalloc(
- (sizeof(THING)) * E
+ sizeof(THING) * E
, ...)
)
// Drop single-byte sizes and redundant parens.
@@
expression COUNT;
typedef u8;
typedef __u8;
@@
(
kzalloc(
- sizeof(u8) * (COUNT)
+ COUNT
, ...)
|
kzalloc(
- sizeof(__u8) * (COUNT)
+ COUNT
, ...)
|
kzalloc(
- sizeof(char) * (COUNT)
+ COUNT
, ...)
|
kzalloc(
- sizeof(unsigned char) * (COUNT)
+ COUNT
, ...)
|
kzalloc(
- sizeof(u8) * COUNT
+ COUNT
, ...)
|
kzalloc(
- sizeof(__u8) * COUNT
+ COUNT
, ...)
|
kzalloc(
- sizeof(char) * COUNT
+ COUNT
, ...)
|
kzalloc(
- sizeof(unsigned char) * COUNT
+ COUNT
, ...)
)
// 2-factor product with sizeof(type/expression) and identifier or constant.
@@
type TYPE;
expression THING;
identifier COUNT_ID;
constant COUNT_CONST;
@@
(
- kzalloc
+ kcalloc
(
- sizeof(TYPE) * (COUNT_ID)
+ COUNT_ID, sizeof(TYPE)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(TYPE) * COUNT_ID
+ COUNT_ID, sizeof(TYPE)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(TYPE) * (COUNT_CONST)
+ COUNT_CONST, sizeof(TYPE)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(TYPE) * COUNT_CONST
+ COUNT_CONST, sizeof(TYPE)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(THING) * (COUNT_ID)
+ COUNT_ID, sizeof(THING)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(THING) * COUNT_ID
+ COUNT_ID, sizeof(THING)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(THING) * (COUNT_CONST)
+ COUNT_CONST, sizeof(THING)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(THING) * COUNT_CONST
+ COUNT_CONST, sizeof(THING)
, ...)
)
// 2-factor product, only identifiers.
@@
identifier SIZE, COUNT;
@@
- kzalloc
+ kcalloc
(
- SIZE * COUNT
+ COUNT, SIZE
, ...)
// 3-factor product with 1 sizeof(type) or sizeof(expression), with
// redundant parens removed.
@@
expression THING;
identifier STRIDE, COUNT;
type TYPE;
@@
(
kzalloc(
- sizeof(TYPE) * (COUNT) * (STRIDE)
+ array3_size(COUNT, STRIDE, sizeof(TYPE))
, ...)
|
kzalloc(
- sizeof(TYPE) * (COUNT) * STRIDE
+ array3_size(COUNT, STRIDE, sizeof(TYPE))
, ...)
|
kzalloc(
- sizeof(TYPE) * COUNT * (STRIDE)
+ array3_size(COUNT, STRIDE, sizeof(TYPE))
, ...)
|
kzalloc(
- sizeof(TYPE) * COUNT * STRIDE
+ array3_size(COUNT, STRIDE, sizeof(TYPE))
, ...)
|
kzalloc(
- sizeof(THING) * (COUNT) * (STRIDE)
+ array3_size(COUNT, STRIDE, sizeof(THING))
, ...)
|
kzalloc(
- sizeof(THING) * (COUNT) * STRIDE
+ array3_size(COUNT, STRIDE, sizeof(THING))
, ...)
|
kzalloc(
- sizeof(THING) * COUNT * (STRIDE)
+ array3_size(COUNT, STRIDE, sizeof(THING))
, ...)
|
kzalloc(
- sizeof(THING) * COUNT * STRIDE
+ array3_size(COUNT, STRIDE, sizeof(THING))
, ...)
)
// 3-factor product with 2 sizeof(variable), with redundant parens removed.
@@
expression THING1, THING2;
identifier COUNT;
type TYPE1, TYPE2;
@@
(
kzalloc(
- sizeof(TYPE1) * sizeof(TYPE2) * COUNT
+ array3_size(COUNT, sizeof(TYPE1), sizeof(TYPE2))
, ...)
|
kzalloc(
- sizeof(TYPE1) * sizeof(THING2) * (COUNT)
+ array3_size(COUNT, sizeof(TYPE1), sizeof(TYPE2))
, ...)
|
kzalloc(
- sizeof(THING1) * sizeof(THING2) * COUNT
+ array3_size(COUNT, sizeof(THING1), sizeof(THING2))
, ...)
|
kzalloc(
- sizeof(THING1) * sizeof(THING2) * (COUNT)
+ array3_size(COUNT, sizeof(THING1), sizeof(THING2))
, ...)
|
kzalloc(
- sizeof(TYPE1) * sizeof(THING2) * COUNT
+ array3_size(COUNT, sizeof(TYPE1), sizeof(THING2))
, ...)
|
kzalloc(
- sizeof(TYPE1) * sizeof(THING2) * (COUNT)
+ array3_size(COUNT, sizeof(TYPE1), sizeof(THING2))
, ...)
)
// 3-factor product, only identifiers, with redundant parens removed.
@@
identifier STRIDE, SIZE, COUNT;
@@
(
kzalloc(
- (COUNT) * STRIDE * SIZE
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kzalloc(
- COUNT * (STRIDE) * SIZE
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kzalloc(
- COUNT * STRIDE * (SIZE)
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kzalloc(
- (COUNT) * (STRIDE) * SIZE
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kzalloc(
- COUNT * (STRIDE) * (SIZE)
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kzalloc(
- (COUNT) * STRIDE * (SIZE)
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kzalloc(
- (COUNT) * (STRIDE) * (SIZE)
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kzalloc(
- COUNT * STRIDE * SIZE
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
)
// Any remaining multi-factor products, first at least 3-factor products,
// when they're not all constants...
@@
expression E1, E2, E3;
constant C1, C2, C3;
@@
(
kzalloc(C1 * C2 * C3, ...)
|
kzalloc(
- (E1) * E2 * E3
+ array3_size(E1, E2, E3)
, ...)
|
kzalloc(
- (E1) * (E2) * E3
+ array3_size(E1, E2, E3)
, ...)
|
kzalloc(
- (E1) * (E2) * (E3)
+ array3_size(E1, E2, E3)
, ...)
|
kzalloc(
- E1 * E2 * E3
+ array3_size(E1, E2, E3)
, ...)
)
// And then all remaining 2 factors products when they're not all constants,
// keeping sizeof() as the second factor argument.
@@
expression THING, E1, E2;
type TYPE;
constant C1, C2, C3;
@@
(
kzalloc(sizeof(THING) * C2, ...)
|
kzalloc(sizeof(TYPE) * C2, ...)
|
kzalloc(C1 * C2 * C3, ...)
|
kzalloc(C1 * C2, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(TYPE) * (E2)
+ E2, sizeof(TYPE)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(TYPE) * E2
+ E2, sizeof(TYPE)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(THING) * (E2)
+ E2, sizeof(THING)
, ...)
|
- kzalloc
+ kcalloc
(
- sizeof(THING) * E2
+ E2, sizeof(THING)
, ...)
|
- kzalloc
+ kcalloc
(
- (E1) * E2
+ E1, E2
, ...)
|
- kzalloc
+ kcalloc
(
- (E1) * (E2)
+ E1, E2
, ...)
|
- kzalloc
+ kcalloc
(
- E1 * E2
+ E1, E2
, ...)
)
Signed-off-by: Kees Cook <keescook@chromium.org>
2018-06-12 21:03:40 +00:00
instance - > reply_map = kcalloc ( nr_cpu_ids , sizeof ( unsigned int ) ,
2018-03-13 09:42:40 +00:00
GFP_KERNEL ) ;
if ( ! instance - > reply_map )
return - ENOMEM ;
2017-10-19 09:48:54 +00:00
switch ( instance - > adapter_type ) {
case MFI_SERIES :
if ( megasas_alloc_mfi_ctrl_mem ( instance ) )
2018-03-13 09:42:40 +00:00
goto fail ;
2017-10-19 09:48:54 +00:00
break ;
2018-12-17 08:47:38 +00:00
case AERO_SERIES :
2017-10-19 09:48:54 +00:00
case VENTURA_SERIES :
case THUNDERBOLT_SERIES :
case INVADER_SERIES :
if ( megasas_alloc_fusion_context ( instance ) )
2018-03-13 09:42:40 +00:00
goto fail ;
2017-10-19 09:48:54 +00:00
break ;
}
2005-09-20 21:46:58 +00:00
2017-10-19 09:48:54 +00:00
return 0 ;
2018-03-13 09:42:40 +00:00
fail :
kfree ( instance - > reply_map ) ;
instance - > reply_map = NULL ;
return - ENOMEM ;
2017-10-19 09:48:54 +00:00
}
2005-09-20 21:46:58 +00:00
2017-10-19 09:48:54 +00:00
/*
* megasas_free_ctrl_mem - Free fusion context for fusion adapters and
* producer , consumer buffers for MFI adapters
*
* @ instance - Adapter soft instance
*
*/
static inline void megasas_free_ctrl_mem ( struct megasas_instance * instance )
{
2018-03-13 09:42:40 +00:00
kfree ( instance - > reply_map ) ;
2017-10-19 09:48:54 +00:00
if ( instance - > adapter_type = = MFI_SERIES ) {
if ( instance - > producer )
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & instance - > pdev - > dev , sizeof ( u32 ) ,
2017-10-19 09:48:54 +00:00
instance - > producer ,
instance - > producer_h ) ;
if ( instance - > consumer )
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & instance - > pdev - > dev , sizeof ( u32 ) ,
2017-10-19 09:48:54 +00:00
instance - > consumer ,
instance - > consumer_h ) ;
} else {
megasas_free_fusion_context ( instance ) ;
2005-09-20 21:46:58 +00:00
}
2017-10-19 09:48:54 +00:00
}
2005-09-20 21:46:58 +00:00
2017-10-19 09:48:55 +00:00
/**
* megasas_alloc_ctrl_dma_buffers - Allocate consistent DMA buffers during
* driver load time
*
* @ instance - Adapter soft instance
* @ return - O for SUCCESS
*/
static inline
int megasas_alloc_ctrl_dma_buffers ( struct megasas_instance * instance )
{
struct pci_dev * pdev = instance - > pdev ;
2017-10-19 09:48:56 +00:00
struct fusion_context * fusion = instance - > ctrl_context ;
2005-09-20 21:46:58 +00:00
2018-10-10 17:31:25 +00:00
instance - > evt_detail = dma_alloc_coherent ( & pdev - > dev ,
sizeof ( struct megasas_evt_detail ) ,
& instance - > evt_detail_h , GFP_KERNEL ) ;
2017-02-10 08:59:17 +00:00
2017-10-19 09:48:55 +00:00
if ( ! instance - > evt_detail ) {
dev_err ( & instance - > pdev - > dev ,
" Failed to allocate event detail buffer \n " ) ;
return - ENOMEM ;
}
2017-10-19 09:48:56 +00:00
if ( fusion ) {
fusion - > ioc_init_request =
dma_alloc_coherent ( & pdev - > dev ,
sizeof ( struct MPI2_IOC_INIT_REQUEST ) ,
& fusion - > ioc_init_request_phys ,
GFP_KERNEL ) ;
if ( ! fusion - > ioc_init_request ) {
dev_err ( & pdev - > dev ,
" Failed to allocate PD list buffer \n " ) ;
return - ENOMEM ;
2010-12-21 21:34:31 +00:00
}
2018-10-17 06:37:40 +00:00
instance - > snapdump_prop = dma_alloc_coherent ( & pdev - > dev ,
sizeof ( struct MR_SNAPDUMP_PROPERTIES ) ,
& instance - > snapdump_prop_h , GFP_KERNEL ) ;
if ( ! instance - > snapdump_prop )
dev_err ( & pdev - > dev ,
" Failed to allocate snapdump properties buffer \n " ) ;
2019-01-29 09:38:14 +00:00
instance - > host_device_list_buf = dma_alloc_coherent ( & pdev - > dev ,
HOST_DEVICE_LIST_SZ ,
& instance - > host_device_list_buf_h ,
GFP_KERNEL ) ;
if ( ! instance - > host_device_list_buf ) {
dev_err ( & pdev - > dev ,
" Failed to allocate targetid list buffer \n " ) ;
return - ENOMEM ;
}
2017-10-19 09:48:56 +00:00
}
2005-09-20 21:46:58 +00:00
2017-10-19 09:48:56 +00:00
instance - > pd_list_buf =
2018-10-10 17:31:25 +00:00
dma_alloc_coherent ( & pdev - > dev ,
2017-10-19 09:48:56 +00:00
MEGASAS_MAX_PD * sizeof ( struct MR_PD_LIST ) ,
2018-10-10 17:31:25 +00:00
& instance - > pd_list_buf_h , GFP_KERNEL ) ;
2017-10-19 09:48:56 +00:00
if ( ! instance - > pd_list_buf ) {
dev_err ( & pdev - > dev , " Failed to allocate PD list buffer \n " ) ;
return - ENOMEM ;
2005-09-20 21:46:58 +00:00
}
2017-10-19 09:48:56 +00:00
instance - > ctrl_info_buf =
2018-10-10 17:31:25 +00:00
dma_alloc_coherent ( & pdev - > dev ,
2017-10-19 09:48:56 +00:00
sizeof ( struct megasas_ctrl_info ) ,
2018-10-10 17:31:25 +00:00
& instance - > ctrl_info_buf_h , GFP_KERNEL ) ;
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
2017-10-19 09:48:56 +00:00
if ( ! instance - > ctrl_info_buf ) {
dev_err ( & pdev - > dev ,
" Failed to allocate controller info buffer \n " ) ;
return - ENOMEM ;
}
2005-09-20 21:46:58 +00:00
2017-10-19 09:48:56 +00:00
instance - > ld_list_buf =
2018-10-10 17:31:25 +00:00
dma_alloc_coherent ( & pdev - > dev ,
2017-10-19 09:48:56 +00:00
sizeof ( struct MR_LD_LIST ) ,
2018-10-10 17:31:25 +00:00
& instance - > ld_list_buf_h , GFP_KERNEL ) ;
2005-09-20 21:46:58 +00:00
2017-10-19 09:48:56 +00:00
if ( ! instance - > ld_list_buf ) {
dev_err ( & pdev - > dev , " Failed to allocate LD list buffer \n " ) ;
return - ENOMEM ;
}
instance - > ld_targetid_list_buf =
2018-10-10 17:31:25 +00:00
dma_alloc_coherent ( & pdev - > dev ,
sizeof ( struct MR_LD_TARGETID_LIST ) ,
& instance - > ld_targetid_list_buf_h , GFP_KERNEL ) ;
2017-10-19 09:48:56 +00:00
if ( ! instance - > ld_targetid_list_buf ) {
dev_err ( & pdev - > dev ,
" Failed to allocate LD targetid list buffer \n " ) ;
return - ENOMEM ;
2005-09-20 21:46:58 +00:00
}
2016-04-15 07:23:30 +00:00
if ( ! reset_devices ) {
2017-10-19 09:48:55 +00:00
instance - > system_info_buf =
2018-10-10 17:31:25 +00:00
dma_alloc_coherent ( & pdev - > dev ,
sizeof ( struct MR_DRV_SYSTEM_INFO ) ,
& instance - > system_info_h , GFP_KERNEL ) ;
2017-10-19 09:48:55 +00:00
instance - > pd_info =
2018-10-10 17:31:25 +00:00
dma_alloc_coherent ( & pdev - > dev ,
sizeof ( struct MR_PD_INFO ) ,
& instance - > pd_info_h , GFP_KERNEL ) ;
2017-10-19 09:48:55 +00:00
instance - > tgt_prop =
2018-10-10 17:31:25 +00:00
dma_alloc_coherent ( & pdev - > dev ,
sizeof ( struct MR_TARGET_PROPERTIES ) ,
& instance - > tgt_prop_h , GFP_KERNEL ) ;
2017-10-19 09:48:55 +00:00
instance - > crash_dump_buf =
2018-10-10 17:31:25 +00:00
dma_alloc_coherent ( & pdev - > dev , CRASH_DMA_BUF_SIZE ,
& instance - > crash_dump_h , GFP_KERNEL ) ;
2016-04-15 07:23:30 +00:00
2017-10-19 09:48:55 +00:00
if ( ! instance - > system_info_buf )
dev_err ( & instance - > pdev - > dev ,
" Failed to allocate system info buffer \n " ) ;
2016-01-28 15:34:26 +00:00
2016-04-15 07:23:30 +00:00
if ( ! instance - > pd_info )
2017-10-19 09:48:55 +00:00
dev_err ( & instance - > pdev - > dev ,
" Failed to allocate pd_info buffer \n " ) ;
2017-08-08 05:15:55 +00:00
2017-02-10 08:59:11 +00:00
if ( ! instance - > tgt_prop )
2017-10-19 09:48:55 +00:00
dev_err ( & instance - > pdev - > dev ,
" Failed to allocate tgt_prop buffer \n " ) ;
2017-02-10 08:59:11 +00:00
2016-04-15 07:23:30 +00:00
if ( ! instance - > crash_dump_buf )
2017-10-19 09:48:55 +00:00
dev_err ( & instance - > pdev - > dev ,
" Failed to allocate crash dump buffer \n " ) ;
2016-04-15 07:23:30 +00:00
}
2016-01-28 15:34:26 +00:00
2017-10-19 09:48:55 +00:00
return 0 ;
}
/*
* megasas_free_ctrl_dma_buffers - Free consistent DMA buffers allocated
* during driver load time
*
* @ instance - Adapter soft instance
*
*/
static inline
void megasas_free_ctrl_dma_buffers ( struct megasas_instance * instance )
{
struct pci_dev * pdev = instance - > pdev ;
2017-10-19 09:48:56 +00:00
struct fusion_context * fusion = instance - > ctrl_context ;
2017-10-19 09:48:55 +00:00
if ( instance - > evt_detail )
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & pdev - > dev , sizeof ( struct megasas_evt_detail ) ,
2017-10-19 09:48:55 +00:00
instance - > evt_detail ,
instance - > evt_detail_h ) ;
2017-10-19 09:48:56 +00:00
if ( fusion & & fusion - > ioc_init_request )
dma_free_coherent ( & pdev - > dev ,
sizeof ( struct MPI2_IOC_INIT_REQUEST ) ,
fusion - > ioc_init_request ,
fusion - > ioc_init_request_phys ) ;
if ( instance - > pd_list_buf )
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & pdev - > dev ,
2017-10-19 09:48:56 +00:00
MEGASAS_MAX_PD * sizeof ( struct MR_PD_LIST ) ,
instance - > pd_list_buf ,
instance - > pd_list_buf_h ) ;
if ( instance - > ld_list_buf )
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & pdev - > dev , sizeof ( struct MR_LD_LIST ) ,
2017-10-19 09:48:56 +00:00
instance - > ld_list_buf ,
instance - > ld_list_buf_h ) ;
if ( instance - > ld_targetid_list_buf )
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & pdev - > dev , sizeof ( struct MR_LD_TARGETID_LIST ) ,
2017-10-19 09:48:56 +00:00
instance - > ld_targetid_list_buf ,
instance - > ld_targetid_list_buf_h ) ;
if ( instance - > ctrl_info_buf )
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & pdev - > dev , sizeof ( struct megasas_ctrl_info ) ,
2017-10-19 09:48:56 +00:00
instance - > ctrl_info_buf ,
instance - > ctrl_info_buf_h ) ;
2017-10-19 09:48:55 +00:00
if ( instance - > system_info_buf )
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & pdev - > dev , sizeof ( struct MR_DRV_SYSTEM_INFO ) ,
2017-10-19 09:48:55 +00:00
instance - > system_info_buf ,
instance - > system_info_h ) ;
if ( instance - > pd_info )
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & pdev - > dev , sizeof ( struct MR_PD_INFO ) ,
2017-10-19 09:48:55 +00:00
instance - > pd_info , instance - > pd_info_h ) ;
if ( instance - > tgt_prop )
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & pdev - > dev , sizeof ( struct MR_TARGET_PROPERTIES ) ,
2017-10-19 09:48:55 +00:00
instance - > tgt_prop , instance - > tgt_prop_h ) ;
if ( instance - > crash_dump_buf )
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & pdev - > dev , CRASH_DMA_BUF_SIZE ,
2017-10-19 09:48:55 +00:00
instance - > crash_dump_buf ,
instance - > crash_dump_h ) ;
2018-10-17 06:37:40 +00:00
if ( instance - > snapdump_prop )
dma_free_coherent ( & pdev - > dev ,
sizeof ( struct MR_SNAPDUMP_PROPERTIES ) ,
instance - > snapdump_prop ,
instance - > snapdump_prop_h ) ;
2019-01-29 09:38:14 +00:00
if ( instance - > host_device_list_buf )
dma_free_coherent ( & pdev - > dev ,
HOST_DEVICE_LIST_SZ ,
instance - > host_device_list_buf ,
instance - > host_device_list_buf_h ) ;
2017-10-19 09:48:55 +00:00
}
2017-10-19 09:48:58 +00:00
/*
* megasas_init_ctrl_params - Initialize controller ' s instance
* parameters before FW init
* @ instance - Adapter soft instance
* @ return - void
*/
static inline void megasas_init_ctrl_params ( struct megasas_instance * instance )
{
instance - > fw_crash_state = UNAVAILABLE ;
megasas_poll_wait_aen = 0 ;
instance - > issuepend_done = 1 ;
atomic_set ( & instance - > adprecovery , MEGASAS_HBA_OPERATIONAL ) ;
2005-09-20 21:46:58 +00:00
/*
* Initialize locks and queues
*/
INIT_LIST_HEAD ( & instance - > cmd_pool ) ;
2010-09-23 02:36:29 +00:00
INIT_LIST_HEAD ( & instance - > internal_reset_pending_q ) ;
2005-09-20 21:46:58 +00:00
2017-10-19 09:48:58 +00:00
atomic_set ( & instance - > fw_outstanding , 0 ) ;
2019-05-07 17:05:36 +00:00
atomic64_set ( & instance - > total_io_count , 0 ) ;
2006-05-30 19:03:37 +00:00
2005-09-20 21:46:58 +00:00
init_waitqueue_head ( & instance - > int_cmd_wait_q ) ;
init_waitqueue_head ( & instance - > abort_cmd_wait_q ) ;
2017-10-19 09:48:58 +00:00
spin_lock_init ( & instance - > crashdump_lock ) ;
2014-09-12 13:27:58 +00:00
spin_lock_init ( & instance - > mfi_pool_lock ) ;
2010-09-23 02:36:29 +00:00
spin_lock_init ( & instance - > hba_lock ) ;
2017-01-10 23:20:46 +00:00
spin_lock_init ( & instance - > stream_lock ) ;
2007-11-09 09:35:44 +00:00
spin_lock_init ( & instance - > completion_lock ) ;
2005-09-20 21:46:58 +00:00
2017-10-19 09:48:58 +00:00
mutex_init ( & instance - > reset_mutex ) ;
2015-01-05 14:36:23 +00:00
2009-10-06 20:40:58 +00:00
if ( ( instance - > pdev - > device = = PCI_DEVICE_ID_LSI_SAS0073SKINNY ) | |
2017-10-19 09:48:58 +00:00
( instance - > pdev - > device = = PCI_DEVICE_ID_LSI_SAS0071SKINNY ) )
2009-10-06 20:43:28 +00:00
instance - > flag_ieee = 1 ;
2009-10-06 20:40:58 +00:00
2006-10-03 20:09:14 +00:00
megasas_dbg_lvl = 0 ;
2007-05-17 12:47:51 +00:00
instance - > flag = 0 ;
2009-10-06 20:47:35 +00:00
instance - > unload = 1 ;
2007-05-17 12:47:51 +00:00
instance - > last_time = 0 ;
2010-09-23 02:36:29 +00:00
instance - > disableOnlineCtrlReset = 1 ;
2013-05-22 07:05:04 +00:00
instance - > UnevenSpanSupport = 0 ;
2019-05-07 17:05:36 +00:00
instance - > smp_affinity_enable = smp_affinity_enable ? true : false ;
instance - > msix_load_balance = false ;
2010-09-23 02:36:29 +00:00
2018-10-17 06:37:39 +00:00
if ( instance - > adapter_type ! = MFI_SERIES )
2010-12-21 21:34:31 +00:00
INIT_WORK ( & instance - > work_init , megasas_fusion_ocr_wq ) ;
2018-10-17 06:37:39 +00:00
else
2010-12-21 21:34:31 +00:00
INIT_WORK ( & instance - > work_init , process_fw_state_change_wq ) ;
2017-10-19 09:48:58 +00:00
}
2005-09-20 21:46:58 +00:00
/**
* megasas_probe_one - PCI hotplug entry point
* @ pdev : PCI device structure
2010-12-15 03:17:17 +00:00
* @ id : PCI ids of supported hotplugged adapter
2005-09-20 21:46:58 +00:00
*/
2012-12-21 21:08:55 +00:00
static int megasas_probe_one ( struct pci_dev * pdev ,
const struct pci_device_id * id )
2005-09-20 21:46:58 +00:00
{
2015-04-23 11:00:09 +00:00
int rval , pos ;
2005-09-20 21:46:58 +00:00
struct Scsi_Host * host ;
struct megasas_instance * instance ;
2011-02-25 04:56:28 +00:00
u16 control = 0 ;
2018-11-09 17:47:20 +00:00
switch ( pdev - > device ) {
2019-06-25 11:04:20 +00:00
case PCI_DEVICE_ID_LSI_AERO_10E0 :
case PCI_DEVICE_ID_LSI_AERO_10E3 :
case PCI_DEVICE_ID_LSI_AERO_10E4 :
case PCI_DEVICE_ID_LSI_AERO_10E7 :
dev_err ( & pdev - > dev , " Adapter is in non secure mode \n " ) ;
return 1 ;
2018-11-09 17:47:20 +00:00
case PCI_DEVICE_ID_LSI_AERO_10E1 :
case PCI_DEVICE_ID_LSI_AERO_10E5 :
dev_info ( & pdev - > dev , " Adapter is in configurable secure mode \n " ) ;
break ;
}
2011-02-25 04:56:28 +00:00
/* Reset MSI-X in the kdump kernel */
if ( reset_devices ) {
pos = pci_find_capability ( pdev , PCI_CAP_ID_MSIX ) ;
if ( pos ) {
2013-04-18 00:08:44 +00:00
pci_read_config_word ( pdev , pos + PCI_MSIX_FLAGS ,
2011-02-25 04:56:28 +00:00
& control ) ;
if ( control & PCI_MSIX_FLAGS_ENABLE ) {
dev_info ( & pdev - > dev , " resetting MSI-X \n " ) ;
pci_write_config_word ( pdev ,
2013-04-18 00:08:44 +00:00
pos + PCI_MSIX_FLAGS ,
2011-02-25 04:56:28 +00:00
control &
~ PCI_MSIX_FLAGS_ENABLE ) ;
}
}
}
2005-09-20 21:46:58 +00:00
/*
* PCI prepping : enable device set bus mastering and dma mask
*/
2009-11-20 07:27:20 +00:00
rval = pci_enable_device_mem ( pdev ) ;
2005-09-20 21:46:58 +00:00
if ( rval ) {
return rval ;
}
pci_set_master ( pdev ) ;
host = scsi_host_alloc ( & megasas_template ,
sizeof ( struct megasas_instance ) ) ;
if ( ! host ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & pdev - > dev , " scsi_host_alloc failed \n " ) ;
2005-09-20 21:46:58 +00:00
goto fail_alloc_instance ;
}
instance = ( struct megasas_instance * ) host - > hostdata ;
memset ( instance , 0 , sizeof ( * instance ) ) ;
2015-07-07 20:52:45 +00:00
atomic_set ( & instance - > fw_reset_no_pci_access , 0 ) ;
2005-09-20 21:46:58 +00:00
/*
* Initialize PCI related and misc parameters
*/
2017-10-19 09:48:58 +00:00
instance - > pdev = pdev ;
2005-09-20 21:46:58 +00:00
instance - > host = host ;
instance - > unique_id = pdev - > bus - > number < < 8 | pdev - > devfn ;
instance - > init_id = MEGASAS_DEFAULT_INIT_ID ;
2015-01-05 14:36:23 +00:00
2017-10-19 09:48:58 +00:00
megasas_set_adapter_type ( instance ) ;
2006-10-03 20:09:14 +00:00
2011-02-25 04:56:12 +00:00
/*
* Initialize MFI Firmware
*/
if ( megasas_init_fw ( instance ) )
goto fail_init_mfi ;
2014-03-10 09:51:56 +00:00
if ( instance - > requestorId ) {
if ( instance - > PlasmaFW111 ) {
instance - > vf_affiliation_111 =
2018-10-10 17:31:25 +00:00
dma_alloc_coherent ( & pdev - > dev ,
sizeof ( struct MR_LD_VF_AFFILIATION_111 ) ,
& instance - > vf_affiliation_111_h ,
GFP_KERNEL ) ;
2014-03-10 09:51:56 +00:00
if ( ! instance - > vf_affiliation_111 )
2015-07-07 20:52:34 +00:00
dev_warn ( & pdev - > dev , " Can't allocate "
2014-03-10 09:51:56 +00:00
" memory for VF affiliation buffer \n " ) ;
} else {
instance - > vf_affiliation =
2018-10-10 17:31:25 +00:00
dma_alloc_coherent ( & pdev - > dev ,
( MAX_LOGICAL_DRIVES + 1 ) *
sizeof ( struct MR_LD_VF_AFFILIATION ) ,
& instance - > vf_affiliation_h ,
GFP_KERNEL ) ;
2014-03-10 09:51:56 +00:00
if ( ! instance - > vf_affiliation )
2015-07-07 20:52:34 +00:00
dev_warn ( & pdev - > dev , " Can't allocate "
2014-03-10 09:51:56 +00:00
" memory for VF affiliation buffer \n " ) ;
}
}
2005-09-20 21:46:58 +00:00
/*
* Store instance in PCI softstate
*/
pci_set_drvdata ( pdev , instance ) ;
/*
* Add this controller to megasas_mgmt_info structure so that it
* can be exported to management applications
*/
megasas_mgmt_info . count + + ;
megasas_mgmt_info . instance [ megasas_mgmt_info . max_index ] = instance ;
megasas_mgmt_info . max_index + + ;
2011-05-12 01:34:29 +00:00
/*
* Register with SCSI mid - layer
*/
if ( megasas_io_attach ( instance ) )
goto fail_io_attach ;
instance - > unload = 0 ;
2014-11-17 09:54:08 +00:00
/*
* Trigger SCSI to scan our drives
*/
2019-01-29 09:38:14 +00:00
if ( ! instance - > enable_fw_dev_list | |
( instance - > host_device_list_buf - > count > 0 ) )
scsi_scan_host ( host ) ;
2011-05-12 01:34:29 +00:00
2005-09-20 21:46:58 +00:00
/*
* Initiate AEN ( Asynchronous Event Notification )
*/
if ( megasas_start_aen ( instance ) ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & pdev - > dev , " start aen failed \n " ) ;
2005-09-20 21:46:58 +00:00
goto fail_start_aen ;
}
2019-05-07 17:05:49 +00:00
megasas_setup_debugfs ( instance ) ;
2014-07-09 22:17:57 +00:00
/* Get current SR-IOV LD/VF affiliation */
if ( instance - > requestorId )
megasas_get_ld_vf_affiliation ( instance , 1 ) ;
2005-09-20 21:46:58 +00:00
return 0 ;
2015-07-07 20:52:45 +00:00
fail_start_aen :
fail_io_attach :
2005-09-20 21:46:58 +00:00
megasas_mgmt_info . count - - ;
megasas_mgmt_info . max_index - - ;
2017-08-07 17:26:57 +00:00
megasas_mgmt_info . instance [ megasas_mgmt_info . max_index ] = NULL ;
2005-09-20 21:46:58 +00:00
2013-05-22 07:04:14 +00:00
instance - > instancet - > disable_intr ( instance ) ;
2015-04-23 11:00:09 +00:00
megasas_destroy_irqs ( instance ) ;
2017-10-19 09:48:50 +00:00
if ( instance - > adapter_type ! = MFI_SERIES )
2011-02-25 04:55:56 +00:00
megasas_release_fusion ( instance ) ;
else
megasas_release_mfi ( instance ) ;
2011-10-09 01:15:13 +00:00
if ( instance - > msix_vectors )
2016-12-02 11:52:23 +00:00
pci_free_irq_vectors ( instance - > pdev ) ;
2015-04-23 11:00:09 +00:00
fail_init_mfi :
2005-09-20 21:46:58 +00:00
scsi_host_put ( host ) ;
2015-07-07 20:52:45 +00:00
fail_alloc_instance :
2005-09-20 21:46:58 +00:00
pci_disable_device ( pdev ) ;
return - ENODEV ;
}
/**
* megasas_flush_cache - Requests FW to flush all its caches
* @ instance : Adapter soft state
*/
static void megasas_flush_cache ( struct megasas_instance * instance )
{
struct megasas_cmd * cmd ;
struct megasas_dcmd_frame * dcmd ;
2016-01-28 15:34:32 +00:00
if ( atomic_read ( & instance - > adprecovery ) = = MEGASAS_HW_CRITICAL_ERROR )
2010-09-23 02:36:29 +00:00
return ;
2005-09-20 21:46:58 +00:00
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd )
return ;
dcmd = & cmd - > frame - > dcmd ;
memset ( dcmd - > mbox . b , 0 , MFI_MBOX_SIZE ) ;
dcmd - > cmd = MFI_CMD_DCMD ;
dcmd - > cmd_status = 0x0 ;
dcmd - > sge_count = 0 ;
2013-09-06 10:20:52 +00:00
dcmd - > flags = cpu_to_le16 ( MFI_FRAME_DIR_NONE ) ;
2005-09-20 21:46:58 +00:00
dcmd - > timeout = 0 ;
2009-12-06 15:24:21 +00:00
dcmd - > pad_0 = 0 ;
2005-09-20 21:46:58 +00:00
dcmd - > data_xfer_len = 0 ;
2013-09-06 10:20:52 +00:00
dcmd - > opcode = cpu_to_le32 ( MR_DCMD_CTRL_CACHE_FLUSH ) ;
2005-09-20 21:46:58 +00:00
dcmd - > mbox . b [ 0 ] = MR_FLUSH_CTRL_CACHE | MR_FLUSH_DISK_CACHE ;
2016-01-28 15:34:23 +00:00
if ( megasas_issue_blocked_cmd ( instance , cmd , MFI_IO_TIMEOUT_SECS )
! = DCMD_SUCCESS ) {
dev_err ( & instance - > pdev - > dev ,
" return from %s %d \n " , __func__ , __LINE__ ) ;
return ;
}
2005-09-20 21:46:58 +00:00
2015-04-23 11:01:24 +00:00
megasas_return_cmd ( instance , cmd ) ;
2005-09-20 21:46:58 +00:00
}
/**
* megasas_shutdown_controller - Instructs FW to shutdown the controller
* @ instance : Adapter soft state
2007-11-07 17:09:50 +00:00
* @ opcode : Shutdown / Hibernate
2005-09-20 21:46:58 +00:00
*/
2007-11-07 17:09:50 +00:00
static void megasas_shutdown_controller ( struct megasas_instance * instance ,
u32 opcode )
2005-09-20 21:46:58 +00:00
{
struct megasas_cmd * cmd ;
struct megasas_dcmd_frame * dcmd ;
2016-01-28 15:34:32 +00:00
if ( atomic_read ( & instance - > adprecovery ) = = MEGASAS_HW_CRITICAL_ERROR )
2010-09-23 02:36:29 +00:00
return ;
2005-09-20 21:46:58 +00:00
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd )
return ;
if ( instance - > aen_cmd )
2014-02-12 18:06:15 +00:00
megasas_issue_blocked_abort_cmd ( instance ,
2016-01-28 15:34:23 +00:00
instance - > aen_cmd , MFI_IO_TIMEOUT_SECS ) ;
2010-12-21 21:34:31 +00:00
if ( instance - > map_update_cmd )
megasas_issue_blocked_abort_cmd ( instance ,
2016-01-28 15:34:23 +00:00
instance - > map_update_cmd , MFI_IO_TIMEOUT_SECS ) ;
2015-08-31 11:53:11 +00:00
if ( instance - > jbod_seq_cmd )
megasas_issue_blocked_abort_cmd ( instance ,
2016-01-28 15:34:23 +00:00
instance - > jbod_seq_cmd , MFI_IO_TIMEOUT_SECS ) ;
2015-08-31 11:53:11 +00:00
2005-09-20 21:46:58 +00:00
dcmd = & cmd - > frame - > dcmd ;
memset ( dcmd - > mbox . b , 0 , MFI_MBOX_SIZE ) ;
dcmd - > cmd = MFI_CMD_DCMD ;
dcmd - > cmd_status = 0x0 ;
dcmd - > sge_count = 0 ;
2013-09-06 10:20:52 +00:00
dcmd - > flags = cpu_to_le16 ( MFI_FRAME_DIR_NONE ) ;
2005-09-20 21:46:58 +00:00
dcmd - > timeout = 0 ;
2009-12-06 15:24:21 +00:00
dcmd - > pad_0 = 0 ;
2005-09-20 21:46:58 +00:00
dcmd - > data_xfer_len = 0 ;
2013-09-06 10:20:52 +00:00
dcmd - > opcode = cpu_to_le32 ( opcode ) ;
2005-09-20 21:46:58 +00:00
2016-01-28 15:34:23 +00:00
if ( megasas_issue_blocked_cmd ( instance , cmd , MFI_IO_TIMEOUT_SECS )
! = DCMD_SUCCESS ) {
dev_err ( & instance - > pdev - > dev ,
" return from %s %d \n " , __func__ , __LINE__ ) ;
return ;
}
2005-09-20 21:46:58 +00:00
2015-04-23 11:01:24 +00:00
megasas_return_cmd ( instance , cmd ) ;
2005-09-20 21:46:58 +00:00
}
2008-05-01 15:56:02 +00:00
# ifdef CONFIG_PM
2007-11-07 17:09:50 +00:00
/**
2007-11-09 09:40:16 +00:00
* megasas_suspend - driver suspend entry point
* @ pdev : PCI device structure
2007-11-07 17:09:50 +00:00
* @ state : PCI power state to suspend routine
*/
2008-05-01 15:56:02 +00:00
static int
2007-11-07 17:09:50 +00:00
megasas_suspend ( struct pci_dev * pdev , pm_message_t state )
{
struct megasas_instance * instance ;
instance = pci_get_drvdata ( pdev ) ;
2019-06-25 11:04:20 +00:00
if ( ! instance )
return 0 ;
2009-10-06 20:47:35 +00:00
instance - > unload = 1 ;
2007-11-07 17:09:50 +00:00
2019-05-07 17:05:46 +00:00
dev_info ( & pdev - > dev , " %s is called \n " , __func__ ) ;
2014-03-10 09:51:56 +00:00
/* Shutdown SR-IOV heartbeat timer */
if ( instance - > requestorId & & ! instance - > skip_heartbeat_timer_del )
del_timer_sync ( & instance - > sriov_heartbeat_timer ) ;
2018-10-17 06:37:39 +00:00
/* Stop the FW fault detection watchdog */
if ( instance - > adapter_type ! = MFI_SERIES )
megasas_fusion_stop_watchdog ( instance ) ;
2007-11-07 17:09:50 +00:00
megasas_flush_cache ( instance ) ;
megasas_shutdown_controller ( instance , MR_DCMD_HIBERNATE_SHUTDOWN ) ;
2009-10-06 20:50:17 +00:00
/* cancel the delayed work if this work still in queue */
if ( instance - > ev ! = NULL ) {
struct megasas_aen_event * ev = instance - > ev ;
2012-12-04 11:33:54 +00:00
cancel_delayed_work_sync ( & ev - > hotplug_work ) ;
2009-10-06 20:50:17 +00:00
instance - > ev = NULL ;
}
2007-11-07 17:09:50 +00:00
tasklet_kill ( & instance - > isr_tasklet ) ;
pci_set_drvdata ( instance - > pdev , instance ) ;
2013-05-22 07:04:14 +00:00
instance - > instancet - > disable_intr ( instance ) ;
2011-10-09 01:15:13 +00:00
2015-04-23 11:00:09 +00:00
megasas_destroy_irqs ( instance ) ;
2011-10-09 01:15:13 +00:00
if ( instance - > msix_vectors )
2016-12-02 11:52:23 +00:00
pci_free_irq_vectors ( instance - > pdev ) ;
2007-11-07 17:09:50 +00:00
pci_save_state ( pdev ) ;
pci_disable_device ( pdev ) ;
pci_set_power_state ( pdev , pci_choose_state ( pdev , state ) ) ;
return 0 ;
}
/**
* megasas_resume - driver resume entry point
* @ pdev : PCI device structure
*/
2008-05-01 15:56:02 +00:00
static int
2007-11-07 17:09:50 +00:00
megasas_resume ( struct pci_dev * pdev )
{
2015-04-23 11:00:09 +00:00
int rval ;
2007-11-07 17:09:50 +00:00
struct Scsi_Host * host ;
struct megasas_instance * instance ;
2016-12-02 11:52:23 +00:00
int irq_flags = PCI_IRQ_LEGACY ;
2007-11-07 17:09:50 +00:00
instance = pci_get_drvdata ( pdev ) ;
2019-06-25 11:04:20 +00:00
if ( ! instance )
return 0 ;
2007-11-07 17:09:50 +00:00
host = instance - > host ;
pci_set_power_state ( pdev , PCI_D0 ) ;
pci_enable_wake ( pdev , PCI_D0 , 0 ) ;
pci_restore_state ( pdev ) ;
2019-05-07 17:05:46 +00:00
dev_info ( & pdev - > dev , " %s is called \n " , __func__ ) ;
2007-11-07 17:09:50 +00:00
/*
* PCI prepping : enable device set bus mastering and dma mask
*/
2009-11-20 07:27:20 +00:00
rval = pci_enable_device_mem ( pdev ) ;
2007-11-07 17:09:50 +00:00
if ( rval ) {
2015-07-07 20:52:34 +00:00
dev_err ( & pdev - > dev , " Enable device failed \n " ) ;
2007-11-07 17:09:50 +00:00
return rval ;
}
pci_set_master ( pdev ) ;
2017-10-19 09:49:05 +00:00
/*
* We expect the FW state to be READY
*/
if ( megasas_transition_to_ready ( instance , 0 ) )
goto fail_ready_state ;
if ( megasas_set_dma_mask ( instance ) )
2007-11-07 17:09:50 +00:00
goto fail_set_dma_mask ;
/*
* Initialize MFI Firmware
*/
atomic_set ( & instance - > fw_outstanding , 0 ) ;
2018-01-05 13:27:39 +00:00
atomic_set ( & instance - > ldio_outstanding , 0 ) ;
2007-11-07 17:09:50 +00:00
2011-05-12 01:33:47 +00:00
/* Now re-enable MSI-X */
2016-12-02 11:52:23 +00:00
if ( instance - > msix_vectors ) {
irq_flags = PCI_IRQ_MSIX ;
2019-05-07 17:05:36 +00:00
if ( instance - > smp_affinity_enable )
2016-12-02 11:52:23 +00:00
irq_flags | = PCI_IRQ_AFFINITY ;
}
rval = pci_alloc_irq_vectors ( instance - > pdev , 1 ,
instance - > msix_vectors ?
instance - > msix_vectors : 1 , irq_flags ) ;
if ( rval < 0 )
2014-08-18 06:01:43 +00:00
goto fail_reenable_msix ;
2011-05-12 01:33:47 +00:00
2018-03-13 09:42:40 +00:00
megasas_setup_reply_map ( instance ) ;
2017-10-19 09:48:50 +00:00
if ( instance - > adapter_type ! = MFI_SERIES ) {
2010-12-21 21:34:31 +00:00
megasas_reset_reply_desc ( instance ) ;
if ( megasas_ioc_init_fusion ( instance ) ) {
megasas_free_cmds ( instance ) ;
megasas_free_cmds_fusion ( instance ) ;
goto fail_init_mfi ;
}
if ( ! megasas_get_map_info ( instance ) )
megasas_sync_map_info ( instance ) ;
2015-10-15 08:09:34 +00:00
} else {
2010-12-21 21:34:31 +00:00
* instance - > producer = 0 ;
* instance - > consumer = 0 ;
if ( megasas_issue_init_mfi ( instance ) )
goto fail_init_mfi ;
}
2007-11-07 17:09:50 +00:00
2018-06-04 10:45:10 +00:00
if ( megasas_get_ctrl_info ( instance ) ! = DCMD_SUCCESS )
goto fail_init_mfi ;
2010-12-21 21:34:31 +00:00
tasklet_init ( & instance - > isr_tasklet , instance - > instancet - > tasklet ,
( unsigned long ) instance ) ;
2007-11-07 17:09:50 +00:00
2015-04-23 11:00:09 +00:00
if ( instance - > msix_vectors ?
megasas_setup_irqs_msix ( instance , 0 ) :
megasas_setup_irqs_ioapic ( instance ) )
goto fail_init_mfi ;
2007-11-07 17:09:50 +00:00
scsi: megaraid_sas: IRQ poll to avoid CPU hard lockups
Issue Description:
We have seen cpu lock up issues from field if system has a large (more than
96) logical cpu count. SAS3.0 controller (Invader series) supports max 96
MSI-X vector and SAS3.5 product (Ventura) supports max 128 MSI-X vectors.
This may be a generic issue (if PCI device support completion on multiple
reply queues).
Let me explain it w.r.t megaraid_sas supported h/w just to simplify the
problem and possible changes to handle such issues. MegaRAID controller
supports multiple reply queues in completion path. Driver creates MSI-X
vectors for controller as "minimum of (FW supported Reply queues, Logical
CPUs)". If submitter is not interrupted via completion on same CPU, there
is a loop in the IO path. This behavior can cause hard/soft CPU lockups, IO
timeout, system sluggish etc.
Example - one CPU (e.g. CPU A) is busy submitting the IOs and another CPU
(e.g. CPU B) is busy with processing the corresponding IO's reply
descriptors from reply descriptor queue upon receiving the interrupts from
HBA. If CPU A is continuously pumping the IOs then always CPU B (which is
executing the ISR) will see the valid reply descriptors in the reply
descriptor queue and it will be continuously processing those reply
descriptor in a loop without quitting the ISR handler.
megaraid_sas driver will exit ISR handler if it finds unused reply
descriptor in the reply descriptor queue. Since CPU A will be continuously
sending the IOs, CPU B may always see a valid reply descriptor (posted by
HBA Firmware after processing the IO) in the reply descriptor queue. In
worst case, driver will not quit from this loop in the ISR handler.
Eventually, CPU lockup will be detected by watchdog.
Above mentioned behavior is not common if "rq_affinity" set to 2 or
affinity_hint is honored by irqbalancer as "exact". If rq_affinity is set
to 2, submitter will be always interrupted via completion on same CPU. If
irqbalancer is using "exact" policy, interrupt will be delivered to
submitter CPU.
Problem statement:
If CPU count to MSI-X vectors (reply descriptor Queues) count ratio is not
1:1, we still have exposure of issue explained above and for that we don't
have any solution.
Exposure of soft/hard lockup is seen if CPU count is more than MSI-X
supported by device.
If CPUs count to MSI-X vectors count ratio is not 1:1, (Other way, if
CPU counts to MSI-X vector count ratio is something like X:1, where X > 1)
then 'exact' irqbalance policy OR rq_affinity = 2 won't help to avoid CPU
hard/soft lockups. There won't be any one to one mapping between
CPU to MSI-X vector instead one MSI-X interrupt (or reply descriptor queue)
is shared with group/set of CPUs and there is a possibility of having a
loop in the IO path within that CPU group and may observe lockups.
For example: Consider a system having two NUMA nodes and each node having
four logical CPUs and also consider that number of MSI-X vectors enabled on
the HBA is two, then CPUs count to MSI-X vector count ratio as 4:1.
e.g.
MSI-X vector 0 is affinity to CPU 0, CPU 1, CPU 2 & CPU 3 of NUMA node 0 and
MSI-X vector 1 is affinity to CPU 4, CPU 5, CPU 6 & CPU 7 of NUMA node 1.
numactl --hardware
available: 2 nodes (0-1)
node 0 cpus: 0 1 2 3 --> MSI-X 0
node 0 size: 65536 MB
node 0 free: 63176 MB
node 1 cpus: 4 5 6 7 --> MSI-X 1
node 1 size: 65536 MB
node 1 free: 63176 MB
Assume that user started an application which uses all the CPUs of NUMA
node 0 for issuing the IOs. Only one CPU from affinity list (it can be any
cpu since this behavior depends upon irqbalance) CPU0 will receive the
interrupts from MSI-X 0 for all the IOs. Eventually, CPU 0 IO submission
percentage will be decreasing and ISR processing percentage will be
increasing as it is more busy with processing the interrupts. Gradually IO
submission percentage on CPU 0 will be zero and it's ISR processing
percentage will be 100% as IO loop has already formed within the
NUMA node 0, i.e. CPU 1, CPU 2 & CPU 3 will be continuously busy with
submitting the heavy IOs and only CPU 0 is busy in the ISR path as it
always find the valid reply descriptor in the reply descriptor queue.
Eventually, we will observe the hard lockup here.
Chances of occurring of hard/soft lockups are directly proportional to
value of X. If value of X is high, then chances of observing CPU lockups is
high.
Solution:
Use IRQ poll interface defined in "irq_poll.c".
megaraid_sas driver will execute ISR routine in softirq context and it will
always quit the loop based on budget provided in IRQ poll interface.
Driver will switch to IRQ poll only when more than a threshold number of
reply descriptors are handled in one ISR. Currently threshold is set as
1/4th of HBA queue depth.
In these scenarios (i.e. where CPUs count to MSI-X vectors count ratio is
X:1 (where X > 1)), IRQ poll interface will avoid CPU hard lockups due to
voluntary exit from the reply queue processing based on budget.
Note - Only one MSI-X vector is busy doing processing.
Select CONFIG_IRQ_POLL from driver Kconfig for driver compilation.
Signed-off-by: Kashyap Desai <kashyap.desai@broadcom.com>
Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-05-07 17:05:35 +00:00
if ( instance - > adapter_type ! = MFI_SERIES )
megasas_setup_irq_poll ( instance ) ;
2014-03-10 09:51:56 +00:00
/* Re-launch SR-IOV heartbeat timer */
if ( instance - > requestorId ) {
if ( ! megasas_sriov_start_heartbeat ( instance , 0 ) )
2017-10-22 22:30:04 +00:00
megasas_start_timer ( instance ) ;
2015-04-23 11:02:09 +00:00
else {
2014-03-10 09:51:56 +00:00
instance - > skip_heartbeat_timer_del = 1 ;
2015-04-23 11:02:09 +00:00
goto fail_init_mfi ;
}
2014-03-10 09:51:56 +00:00
}
2013-05-22 07:04:14 +00:00
instance - > instancet - > enable_intr ( instance ) ;
2015-08-31 11:53:11 +00:00
megasas_setup_jbod_map ( instance ) ;
2009-10-06 20:47:35 +00:00
instance - > unload = 0 ;
2011-05-12 01:34:29 +00:00
/*
* Initiate AEN ( Asynchronous Event Notification )
*/
if ( megasas_start_aen ( instance ) )
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " Start AEN failed \n " ) ;
2011-05-12 01:34:29 +00:00
2018-10-17 06:37:39 +00:00
/* Re-launch FW fault watchdog */
if ( instance - > adapter_type ! = MFI_SERIES )
if ( megasas_fusion_start_watchdog ( instance ) ! = SUCCESS )
goto fail_start_watchdog ;
2007-11-07 17:09:50 +00:00
return 0 ;
2018-10-17 06:37:39 +00:00
fail_start_watchdog :
if ( instance - > requestorId & & ! instance - > skip_heartbeat_timer_del )
del_timer_sync ( & instance - > sriov_heartbeat_timer ) ;
2007-11-07 17:09:50 +00:00
fail_init_mfi :
2017-10-19 09:48:55 +00:00
megasas_free_ctrl_dma_buffers ( instance ) ;
2017-10-19 09:48:54 +00:00
megasas_free_ctrl_mem ( instance ) ;
2007-11-07 17:09:50 +00:00
scsi_host_put ( host ) ;
2017-10-19 09:49:05 +00:00
fail_reenable_msix :
2007-11-07 17:09:50 +00:00
fail_set_dma_mask :
fail_ready_state :
pci_disable_device ( pdev ) ;
return - ENODEV ;
}
2008-05-01 15:56:02 +00:00
# else
# define megasas_suspend NULL
# define megasas_resume NULL
# endif
2007-11-07 17:09:50 +00:00
2016-10-21 13:33:31 +00:00
static inline int
megasas_wait_for_adapter_operational ( struct megasas_instance * instance )
{
int wait_time = MEGASAS_RESET_WAIT_TIME * 2 ;
int i ;
2018-06-04 10:45:11 +00:00
u8 adp_state ;
2016-10-21 13:33:31 +00:00
for ( i = 0 ; i < wait_time ; i + + ) {
2018-06-04 10:45:11 +00:00
adp_state = atomic_read ( & instance - > adprecovery ) ;
if ( ( adp_state = = MEGASAS_HBA_OPERATIONAL ) | |
( adp_state = = MEGASAS_HW_CRITICAL_ERROR ) )
2016-10-21 13:33:31 +00:00
break ;
if ( ! ( i % MEGASAS_RESET_NOTICE_INTERVAL ) )
dev_notice ( & instance - > pdev - > dev , " waiting for controller reset to finish \n " ) ;
msleep ( 1000 ) ;
}
2018-06-04 10:45:11 +00:00
if ( adp_state ! = MEGASAS_HBA_OPERATIONAL ) {
dev_info ( & instance - > pdev - > dev ,
" %s HBA failed to become operational, adp_state %d \n " ,
__func__ , adp_state ) ;
2016-10-21 13:33:31 +00:00
return 1 ;
}
return 0 ;
}
2005-09-20 21:46:58 +00:00
/**
* megasas_detach_one - PCI hot " un " plug entry point
* @ pdev : PCI device structure
*/
2012-12-21 21:08:55 +00:00
static void megasas_detach_one ( struct pci_dev * pdev )
2005-09-20 21:46:58 +00:00
{
int i ;
struct Scsi_Host * host ;
struct megasas_instance * instance ;
2010-12-21 21:34:31 +00:00
struct fusion_context * fusion ;
2015-08-31 11:53:11 +00:00
u32 pd_seq_map_sz ;
2005-09-20 21:46:58 +00:00
instance = pci_get_drvdata ( pdev ) ;
2019-06-25 11:04:20 +00:00
if ( ! instance )
return ;
2005-09-20 21:46:58 +00:00
host = instance - > host ;
2010-12-21 21:34:31 +00:00
fusion = instance - > ctrl_context ;
2005-09-20 21:46:58 +00:00
2014-03-10 09:51:56 +00:00
/* Shutdown SR-IOV heartbeat timer */
if ( instance - > requestorId & & ! instance - > skip_heartbeat_timer_del )
del_timer_sync ( & instance - > sriov_heartbeat_timer ) ;
2018-10-17 06:37:39 +00:00
/* Stop the FW fault detection watchdog */
if ( instance - > adapter_type ! = MFI_SERIES )
megasas_fusion_stop_watchdog ( instance ) ;
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
if ( instance - > fw_crash_state ! = UNAVAILABLE )
megasas_free_host_crash_buffer ( instance ) ;
2005-09-20 21:46:58 +00:00
scsi_remove_host ( instance - > host ) ;
2018-01-05 13:27:41 +00:00
instance - > unload = 1 ;
2016-10-21 13:33:31 +00:00
if ( megasas_wait_for_adapter_operational ( instance ) )
goto skip_firing_dcmds ;
2005-09-20 21:46:58 +00:00
megasas_flush_cache ( instance ) ;
2007-11-07 17:09:50 +00:00
megasas_shutdown_controller ( instance , MR_DCMD_CTRL_SHUTDOWN ) ;
2009-10-06 20:50:17 +00:00
2016-10-21 13:33:31 +00:00
skip_firing_dcmds :
2009-10-06 20:50:17 +00:00
/* cancel the delayed work if this work still in queue*/
if ( instance - > ev ! = NULL ) {
struct megasas_aen_event * ev = instance - > ev ;
2012-12-04 11:33:54 +00:00
cancel_delayed_work_sync ( & ev - > hotplug_work ) ;
2009-10-06 20:50:17 +00:00
instance - > ev = NULL ;
}
2014-02-12 18:06:15 +00:00
/* cancel all wait events */
wake_up_all ( & instance - > int_cmd_wait_q ) ;
2006-10-03 20:13:18 +00:00
tasklet_kill ( & instance - > isr_tasklet ) ;
2005-09-20 21:46:58 +00:00
/*
* Take the instance off the instance array . Note that we will not
* decrement the max_index . We let this array be sparse array
*/
for ( i = 0 ; i < megasas_mgmt_info . max_index ; i + + ) {
if ( megasas_mgmt_info . instance [ i ] = = instance ) {
megasas_mgmt_info . count - - ;
megasas_mgmt_info . instance [ i ] = NULL ;
break ;
}
}
2013-05-22 07:04:14 +00:00
instance - > instancet - > disable_intr ( instance ) ;
2005-09-20 21:46:58 +00:00
2015-04-23 11:00:09 +00:00
megasas_destroy_irqs ( instance ) ;
2011-10-09 01:15:13 +00:00
if ( instance - > msix_vectors )
2016-12-02 11:52:23 +00:00
pci_free_irq_vectors ( instance - > pdev ) ;
2005-09-20 21:46:58 +00:00
2018-12-17 08:47:37 +00:00
if ( instance - > adapter_type > = VENTURA_SERIES ) {
2017-01-10 23:20:46 +00:00
for ( i = 0 ; i < MAX_LOGICAL_DRIVES_EXT ; + + i )
kfree ( fusion - > stream_detect_by_ld [ i ] ) ;
kfree ( fusion - > stream_detect_by_ld ) ;
fusion - > stream_detect_by_ld = NULL ;
}
2017-10-19 09:48:50 +00:00
if ( instance - > adapter_type ! = MFI_SERIES ) {
2010-12-21 21:34:31 +00:00
megasas_release_fusion ( instance ) ;
2015-08-31 11:53:11 +00:00
pd_seq_map_sz = sizeof ( struct MR_PD_CFG_SEQ_NUM_SYNC ) +
( sizeof ( struct MR_PD_CFG_SEQ ) *
( MAX_PHYSICAL_DEVICES - 1 ) ) ;
2014-09-12 13:27:33 +00:00
for ( i = 0 ; i < 2 ; i + + ) {
2010-12-21 21:34:31 +00:00
if ( fusion - > ld_map [ i ] )
dma_free_coherent ( & instance - > pdev - > dev ,
2014-09-12 13:27:33 +00:00
fusion - > max_map_sz ,
2010-12-21 21:34:31 +00:00
fusion - > ld_map [ i ] ,
2014-09-12 13:27:33 +00:00
fusion - > ld_map_phys [ i ] ) ;
2017-08-23 11:47:03 +00:00
if ( fusion - > ld_drv_map [ i ] ) {
if ( is_vmalloc_addr ( fusion - > ld_drv_map [ i ] ) )
vfree ( fusion - > ld_drv_map [ i ] ) ;
else
free_pages ( ( ulong ) fusion - > ld_drv_map [ i ] ,
fusion - > drv_map_pages ) ;
}
2016-01-22 12:41:42 +00:00
if ( fusion - > pd_seq_sync [ i ] )
dma_free_coherent ( & instance - > pdev - > dev ,
pd_seq_map_sz ,
fusion - > pd_seq_sync [ i ] ,
fusion - > pd_seq_phys [ i ] ) ;
2014-09-12 13:27:33 +00:00
}
2015-10-15 08:09:34 +00:00
} else {
2010-12-21 21:34:31 +00:00
megasas_release_mfi ( instance ) ;
}
2005-09-20 21:46:58 +00:00
2014-03-10 09:51:56 +00:00
if ( instance - > vf_affiliation )
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & pdev - > dev , ( MAX_LOGICAL_DRIVES + 1 ) *
2014-03-10 09:51:56 +00:00
sizeof ( struct MR_LD_VF_AFFILIATION ) ,
instance - > vf_affiliation ,
instance - > vf_affiliation_h ) ;
if ( instance - > vf_affiliation_111 )
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & pdev - > dev ,
2014-03-10 09:51:56 +00:00
sizeof ( struct MR_LD_VF_AFFILIATION_111 ) ,
instance - > vf_affiliation_111 ,
instance - > vf_affiliation_111_h ) ;
if ( instance - > hb_host_mem )
2018-10-10 17:31:25 +00:00
dma_free_coherent ( & pdev - > dev , sizeof ( struct MR_CTRL_HB_HOST_MEM ) ,
2014-03-10 09:51:56 +00:00
instance - > hb_host_mem ,
instance - > hb_host_mem_h ) ;
2017-10-19 09:48:55 +00:00
megasas_free_ctrl_dma_buffers ( instance ) ;
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
2017-10-19 09:48:54 +00:00
megasas_free_ctrl_mem ( instance ) ;
2015-04-23 11:02:09 +00:00
2019-05-07 17:05:49 +00:00
megasas_destroy_debugfs ( instance ) ;
2005-09-20 21:46:58 +00:00
scsi_host_put ( host ) ;
pci_disable_device ( pdev ) ;
}
/**
* megasas_shutdown - Shutdown entry point
* @ device : Generic device structure
*/
static void megasas_shutdown ( struct pci_dev * pdev )
{
struct megasas_instance * instance = pci_get_drvdata ( pdev ) ;
2011-10-09 01:15:13 +00:00
2019-06-25 11:04:20 +00:00
if ( ! instance )
return ;
2009-10-06 20:47:35 +00:00
instance - > unload = 1 ;
2016-10-21 13:33:31 +00:00
if ( megasas_wait_for_adapter_operational ( instance ) )
goto skip_firing_dcmds ;
2005-09-20 21:46:58 +00:00
megasas_flush_cache ( instance ) ;
2008-08-10 19:42:37 +00:00
megasas_shutdown_controller ( instance , MR_DCMD_CTRL_SHUTDOWN ) ;
2016-10-21 13:33:31 +00:00
skip_firing_dcmds :
2013-05-22 07:04:14 +00:00
instance - > instancet - > disable_intr ( instance ) ;
2015-04-23 11:00:09 +00:00
megasas_destroy_irqs ( instance ) ;
2011-10-09 01:15:13 +00:00
if ( instance - > msix_vectors )
2016-12-02 11:52:23 +00:00
pci_free_irq_vectors ( instance - > pdev ) ;
2005-09-20 21:46:58 +00:00
}
/**
* megasas_mgmt_open - char node " open " entry point
*/
static int megasas_mgmt_open ( struct inode * inode , struct file * filep )
{
/*
* Allow only those users with admin rights
*/
if ( ! capable ( CAP_SYS_ADMIN ) )
return - EACCES ;
return 0 ;
}
/**
* megasas_mgmt_fasync - Async notifier registration from applications
*
* This function adds the calling process to a driver global queue . When an
* event occurs , SIGIO will be sent to all processes in this queue .
*/
static int megasas_mgmt_fasync ( int fd , struct file * filep , int mode )
{
int rc ;
2006-01-11 12:16:10 +00:00
mutex_lock ( & megasas_async_queue_mutex ) ;
2005-09-20 21:46:58 +00:00
rc = fasync_helper ( fd , filep , mode , & megasas_async_queue ) ;
2006-01-11 12:16:10 +00:00
mutex_unlock ( & megasas_async_queue_mutex ) ;
2005-09-20 21:46:58 +00:00
if ( rc > = 0 ) {
/* For sanity check when we get ioctl */
filep - > private_data = filep ;
return 0 ;
}
printk ( KERN_DEBUG " megasas: fasync_helper failed [%d] \n " , rc ) ;
return rc ;
}
2009-10-06 20:18:02 +00:00
/**
* megasas_mgmt_poll - char node " poll " entry point
* */
2017-07-03 10:39:46 +00:00
static __poll_t megasas_mgmt_poll ( struct file * file , poll_table * wait )
2009-10-06 20:18:02 +00:00
{
2017-07-03 10:39:46 +00:00
__poll_t mask ;
2009-10-06 20:18:02 +00:00
unsigned long flags ;
2015-07-07 20:52:45 +00:00
2009-10-06 20:18:02 +00:00
poll_wait ( file , & megasas_poll_wait , wait ) ;
spin_lock_irqsave ( & poll_aen_lock , flags ) ;
if ( megasas_poll_wait_aen )
2018-02-11 22:34:03 +00:00
mask = ( EPOLLIN | EPOLLRDNORM ) ;
2009-10-06 20:18:02 +00:00
else
mask = 0 ;
2014-09-12 13:27:33 +00:00
megasas_poll_wait_aen = 0 ;
2009-10-06 20:18:02 +00:00
spin_unlock_irqrestore ( & poll_aen_lock , flags ) ;
return mask ;
}
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
/*
* megasas_set_crash_dump_params_ioctl :
* Send CRASH_DUMP_MODE DCMD to all controllers
* @ cmd : MFI command frame
*/
2015-07-07 20:52:45 +00:00
static int megasas_set_crash_dump_params_ioctl ( struct megasas_cmd * cmd )
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
{
struct megasas_instance * local_instance ;
int i , error = 0 ;
int crash_support ;
crash_support = cmd - > frame - > dcmd . mbox . w [ 0 ] ;
for ( i = 0 ; i < megasas_mgmt_info . max_index ; i + + ) {
local_instance = megasas_mgmt_info . instance [ i ] ;
if ( local_instance & & local_instance - > crash_dump_drv_support ) {
2016-01-28 15:34:32 +00:00
if ( ( atomic_read ( & local_instance - > adprecovery ) = =
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
MEGASAS_HBA_OPERATIONAL ) & &
! megasas_set_crash_dump_params ( local_instance ,
crash_support ) ) {
local_instance - > crash_dump_app_support =
crash_support ;
dev_info ( & local_instance - > pdev - > dev ,
" Application firmware crash "
" dump mode set success \n " ) ;
error = 0 ;
} else {
dev_info ( & local_instance - > pdev - > dev ,
" Application firmware crash "
" dump mode set failed \n " ) ;
error = - 1 ;
}
}
}
return error ;
}
2005-09-20 21:46:58 +00:00
/**
* megasas_mgmt_fw_ioctl - Issues management ioctls to FW
* @ instance : Adapter soft state
* @ argp : User ' s ioctl packet
*/
static int
megasas_mgmt_fw_ioctl ( struct megasas_instance * instance ,
struct megasas_iocpacket __user * user_ioc ,
struct megasas_iocpacket * ioc )
{
2017-10-19 09:49:05 +00:00
struct megasas_sge64 * kern_sge64 = NULL ;
struct megasas_sge32 * kern_sge32 = NULL ;
2005-09-20 21:46:58 +00:00
struct megasas_cmd * cmd ;
void * kbuff_arr [ MAX_IOCTL_SGE ] ;
dma_addr_t buf_handle = 0 ;
int error = 0 , i ;
void * sense = NULL ;
dma_addr_t sense_handle ;
2009-10-06 20:52:20 +00:00
unsigned long * sense_ptr ;
2017-10-19 09:49:02 +00:00
u32 opcode = 0 ;
2005-09-20 21:46:58 +00:00
memset ( kbuff_arr , 0 , sizeof ( kbuff_arr ) ) ;
if ( ioc - > sge_count > MAX_IOCTL_SGE ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " SGE count [%d] > max limit [%d] \n " ,
2005-09-20 21:46:58 +00:00
ioc - > sge_count , MAX_IOCTL_SGE ) ;
return - EINVAL ;
}
2018-01-05 13:33:04 +00:00
if ( ( ioc - > frame . hdr . cmd > = MFI_CMD_OP_COUNT ) | |
( ( ioc - > frame . hdr . cmd = = MFI_CMD_NVME ) & &
2019-06-25 11:04:30 +00:00
! instance - > support_nvme_passthru ) | |
( ( ioc - > frame . hdr . cmd = = MFI_CMD_TOOLBOX ) & &
! instance - > support_pci_lane_margining ) ) {
2017-10-19 09:49:02 +00:00
dev_err ( & instance - > pdev - > dev ,
" Received invalid ioctl command 0x%x \n " ,
ioc - > frame . hdr . cmd ) ;
return - ENOTSUPP ;
}
2005-09-20 21:46:58 +00:00
cmd = megasas_get_cmd ( instance ) ;
if ( ! cmd ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " Failed to get a cmd packet \n " ) ;
2005-09-20 21:46:58 +00:00
return - ENOMEM ;
}
/*
* User ' s IOCTL packet has 2 frames ( maximum ) . Copy those two
* frames into our cmd ' s frames . cmd - > frame ' s context will get
* overwritten when we copy from user ' s frames . So set that value
* alone separately
*/
memcpy ( cmd - > frame , ioc - > frame . raw , 2 * MEGAMFI_FRAME_SIZE ) ;
2013-09-06 10:20:52 +00:00
cmd - > frame - > hdr . context = cpu_to_le32 ( cmd - > index ) ;
2009-10-06 20:18:02 +00:00
cmd - > frame - > hdr . pad_0 = 0 ;
2017-10-19 09:49:05 +00:00
cmd - > frame - > hdr . flags & = ( ~ MFI_FRAME_IEEE ) ;
if ( instance - > consistent_mask_64bit )
cmd - > frame - > hdr . flags | = cpu_to_le16 ( ( MFI_FRAME_SGL64 |
MFI_FRAME_SENSE64 ) ) ;
else
cmd - > frame - > hdr . flags & = cpu_to_le16 ( ~ ( MFI_FRAME_SGL64 |
2013-09-06 10:20:52 +00:00
MFI_FRAME_SENSE64 ) ) ;
2017-10-19 09:49:02 +00:00
if ( cmd - > frame - > hdr . cmd = = MFI_CMD_DCMD )
opcode = le32_to_cpu ( cmd - > frame - > dcmd . opcode ) ;
2005-09-20 21:46:58 +00:00
2017-08-23 11:47:00 +00:00
if ( opcode = = MR_DCMD_CTRL_SHUTDOWN ) {
2019-06-25 11:04:23 +00:00
mutex_lock ( & instance - > reset_mutex ) ;
2017-02-10 08:59:23 +00:00
if ( megasas_get_ctrl_info ( instance ) ! = DCMD_SUCCESS ) {
megasas_return_cmd ( instance , cmd ) ;
2019-06-25 11:04:23 +00:00
mutex_unlock ( & instance - > reset_mutex ) ;
2017-02-10 08:59:23 +00:00
return - 1 ;
}
2019-06-25 11:04:23 +00:00
mutex_unlock ( & instance - > reset_mutex ) ;
2017-02-10 08:59:23 +00:00
}
2017-08-23 11:47:00 +00:00
if ( opcode = = MR_DRIVER_SET_APP_CRASHDUMP_MODE ) {
megaraid_sas : Firmware crash dump feature support
Resending the patch. Addressed the review comments from Tomas Henzl.
Move buff_offset inside spinlock, corrected loop at crash dump buffer free,
reset_devices check is added to disable fw crash dump feature in kdump kernel.
This feature will provide similar interface as kernel crash dump feature.
When megaraid firmware encounter any crash, driver will collect the firmware raw image and
dump it into pre-configured location.
Driver will allocate two different segment of memory.
#1 Non-DMA able large buffer (will be allocated on demand) to capture actual FW crash dump.
#2 DMA buffer (persistence allocation) just to do a arbitrator job.
Firmware will keep writing Crash dump data in chucks of DMA buffer size into #2,
which will be copy back by driver to the host memory as described in #1.
Driver-Firmware interface:
==================
A.) Host driver can allocate maximum 512MB Host memory to store crash dump data.
This memory will be internal to the host and will not be exposed to the Firmware.
Driver may not be able to allocate 512 MB. In that case, driver will do possible memory
(available at run time) allocation to store crash dump data.
Let’s call this buffer as Host Crash Buffer.
Host Crash buffer will not be contigious as a whole, but it will have multiple chunk of contigious memory.
This will be internal to driver and firmware/application are unaware of it.
Partial allocation of Host Crash buffer may have valid information to debug depending upon
what was collected in that buffer and depending on nature of failure.
Complete Crash dump is the best case, but we do want to capture partial buffer just to grab something rather than nothing.
Host Crash buffer will be allocated only when FW Crash dump data is available,
and will be deallocated once application copy Host Crash buffer to the file.
Host Crash buffer size can be anything between 1MB to 512MB. (It will be multiple of 1MBs)
B.) Irrespective of underlying Firmware capability of crash dump support,
driver will allocate DMA buffer at start of the day for each MR controllers.
Let’s call this buffer as “DMA Crash Buffer”.
For this feature, size of DMA crash buffer will be 1MB.
(We will not gain much even if DMA buffer size is increased.)
C.) Driver will now read Controller Info sending existing dcmd “MR_DCMD_CTRL_GET_INFO”.
Driver should extract the information from ctrl info provided by firmware and
figure out if firmware support crash dump feature or not.
Driver will enable crash dump feature only if
“Firmware support Crash dump” +
“Driver was able to create DMA Crash Buffer”.
If either one from above is not set, Crash dump feature should be disable in driver.
Firmware will enable crash dump feature only if “Driver Send DCMD- MR_DCMD_SET_CRASH_BUF_PARA with MR_CRASH_BUF_TURN_ON”
Helper application/script should use sysfs parameter fw_crash_xxx to actually copy data from
host memory to the filesystem.
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-09-12 13:27:28 +00:00
error = megasas_set_crash_dump_params_ioctl ( cmd ) ;
megasas_return_cmd ( instance , cmd ) ;
return error ;
}
2005-09-20 21:46:58 +00:00
/*
* The management interface between applications and the fw uses
* MFI frames . E . g , RAID configuration changes , LD property changes
* etc are accomplishes through different kinds of MFI frames . The
* driver needs to care only about substituting user buffers with
* kernel buffers in SGLs . The location of SGL is embedded in the
* struct iocpacket itself .
*/
2017-10-19 09:49:05 +00:00
if ( instance - > consistent_mask_64bit )
kern_sge64 = ( struct megasas_sge64 * )
( ( unsigned long ) cmd - > frame + ioc - > sgl_off ) ;
else
kern_sge32 = ( struct megasas_sge32 * )
( ( unsigned long ) cmd - > frame + ioc - > sgl_off ) ;
2005-09-20 21:46:58 +00:00
/*
* For each user buffer , create a mirror buffer and copy in
*/
for ( i = 0 ; i < ioc - > sge_count ; i + + ) {
2011-01-19 09:01:14 +00:00
if ( ! ioc - > sgl [ i ] . iov_len )
continue ;
2007-02-14 20:55:45 +00:00
kbuff_arr [ i ] = dma_alloc_coherent ( & instance - > pdev - > dev ,
2005-09-20 21:46:58 +00:00
ioc - > sgl [ i ] . iov_len ,
2007-02-14 20:55:45 +00:00
& buf_handle , GFP_KERNEL ) ;
2005-09-20 21:46:58 +00:00
if ( ! kbuff_arr [ i ] ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " Failed to alloc "
" kernel SGL buffer for IOCTL \n " ) ;
2005-09-20 21:46:58 +00:00
error = - ENOMEM ;
goto out ;
}
/*
* We don ' t change the dma_coherent_mask , so
2018-10-10 17:31:25 +00:00
* dma_alloc_coherent only returns 32 bit addresses
2005-09-20 21:46:58 +00:00
*/
2017-10-19 09:49:05 +00:00
if ( instance - > consistent_mask_64bit ) {
kern_sge64 [ i ] . phys_addr = cpu_to_le64 ( buf_handle ) ;
kern_sge64 [ i ] . length = cpu_to_le32 ( ioc - > sgl [ i ] . iov_len ) ;
} else {
kern_sge32 [ i ] . phys_addr = cpu_to_le32 ( buf_handle ) ;
kern_sge32 [ i ] . length = cpu_to_le32 ( ioc - > sgl [ i ] . iov_len ) ;
}
2005-09-20 21:46:58 +00:00
/*
* We created a kernel buffer corresponding to the
* user buffer . Now copy in from the user buffer
*/
if ( copy_from_user ( kbuff_arr [ i ] , ioc - > sgl [ i ] . iov_base ,
( u32 ) ( ioc - > sgl [ i ] . iov_len ) ) ) {
error = - EFAULT ;
goto out ;
}
}
if ( ioc - > sense_len ) {
2007-02-14 20:55:45 +00:00
sense = dma_alloc_coherent ( & instance - > pdev - > dev , ioc - > sense_len ,
& sense_handle , GFP_KERNEL ) ;
2005-09-20 21:46:58 +00:00
if ( ! sense ) {
error = - ENOMEM ;
goto out ;
}
sense_ptr =
2009-10-06 20:52:20 +00:00
( unsigned long * ) ( ( unsigned long ) cmd - > frame + ioc - > sense_off ) ;
2017-10-19 09:49:05 +00:00
if ( instance - > consistent_mask_64bit )
* sense_ptr = cpu_to_le64 ( sense_handle ) ;
else
* sense_ptr = cpu_to_le32 ( sense_handle ) ;
2005-09-20 21:46:58 +00:00
}
/*
* Set the sync_cmd flag so that the ISR knows not to complete this
* cmd to the SCSI mid - layer
*/
cmd - > sync_cmd = 1 ;
2016-01-28 15:34:23 +00:00
if ( megasas_issue_blocked_cmd ( instance , cmd , 0 ) = = DCMD_NOT_FIRED ) {
cmd - > sync_cmd = 0 ;
dev_err ( & instance - > pdev - > dev ,
2017-10-19 09:49:02 +00:00
" return -EBUSY from %s %d cmd 0x%x opcode 0x%x cmd->cmd_status_drv 0x%x \n " ,
__func__ , __LINE__ , cmd - > frame - > hdr . cmd , opcode ,
cmd - > cmd_status_drv ) ;
2016-01-28 15:34:23 +00:00
return - EBUSY ;
}
2005-09-20 21:46:58 +00:00
cmd - > sync_cmd = 0 ;
2014-11-17 09:54:08 +00:00
if ( instance - > unload = = 1 ) {
dev_info ( & instance - > pdev - > dev , " Driver unload is in progress "
" don't submit data to application \n " ) ;
goto out ;
}
2005-09-20 21:46:58 +00:00
/*
* copy out the kernel buffers to user buffers
*/
for ( i = 0 ; i < ioc - > sge_count ; i + + ) {
if ( copy_to_user ( ioc - > sgl [ i ] . iov_base , kbuff_arr [ i ] ,
ioc - > sgl [ i ] . iov_len ) ) {
error = - EFAULT ;
goto out ;
}
}
/*
* copy out the sense
*/
if ( ioc - > sense_len ) {
/*
2008-03-18 07:13:06 +00:00
* sense_ptr points to the location that has the user
2005-09-20 21:46:58 +00:00
* sense buffer address
*/
2009-10-06 20:52:20 +00:00
sense_ptr = ( unsigned long * ) ( ( unsigned long ) ioc - > frame . raw +
ioc - > sense_off ) ;
2005-09-20 21:46:58 +00:00
2017-02-10 08:59:22 +00:00
if ( copy_to_user ( ( void __user * ) ( ( unsigned long )
get_unaligned ( ( unsigned long * ) sense_ptr ) ) ,
2008-03-18 07:13:06 +00:00
sense , ioc - > sense_len ) ) {
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " Failed to copy out to user "
2007-11-09 09:28:47 +00:00
" sense data \n " ) ;
2005-09-20 21:46:58 +00:00
error = - EFAULT ;
goto out ;
}
}
/*
* copy the status codes returned by the fw
*/
if ( copy_to_user ( & user_ioc - > frame . hdr . cmd_status ,
& cmd - > frame - > hdr . cmd_status , sizeof ( u8 ) ) ) {
2015-07-07 20:52:34 +00:00
dev_printk ( KERN_DEBUG , & instance - > pdev - > dev , " Error copying out cmd_status \n " ) ;
2005-09-20 21:46:58 +00:00
error = - EFAULT ;
}
2015-07-07 20:52:45 +00:00
out :
2005-09-20 21:46:58 +00:00
if ( sense ) {
2007-02-14 20:55:45 +00:00
dma_free_coherent ( & instance - > pdev - > dev , ioc - > sense_len ,
2005-09-20 21:46:58 +00:00
sense , sense_handle ) ;
}
2012-11-21 08:54:48 +00:00
for ( i = 0 ; i < ioc - > sge_count ; i + + ) {
2016-03-14 14:29:45 +00:00
if ( kbuff_arr [ i ] ) {
2017-10-19 09:49:05 +00:00
if ( instance - > consistent_mask_64bit )
dma_free_coherent ( & instance - > pdev - > dev ,
le32_to_cpu ( kern_sge64 [ i ] . length ) ,
kbuff_arr [ i ] ,
le64_to_cpu ( kern_sge64 [ i ] . phys_addr ) ) ;
else
dma_free_coherent ( & instance - > pdev - > dev ,
le32_to_cpu ( kern_sge32 [ i ] . length ) ,
kbuff_arr [ i ] ,
le32_to_cpu ( kern_sge32 [ i ] . phys_addr ) ) ;
2014-09-12 13:27:58 +00:00
kbuff_arr [ i ] = NULL ;
2016-03-14 14:29:45 +00:00
}
2005-09-20 21:46:58 +00:00
}
2015-04-23 11:01:24 +00:00
megasas_return_cmd ( instance , cmd ) ;
2005-09-20 21:46:58 +00:00
return error ;
}
static int megasas_mgmt_ioctl_fw ( struct file * file , unsigned long arg )
{
struct megasas_iocpacket __user * user_ioc =
( struct megasas_iocpacket __user * ) arg ;
struct megasas_iocpacket * ioc ;
struct megasas_instance * instance ;
int error ;
2016-08-21 08:39:04 +00:00
ioc = memdup_user ( user_ioc , sizeof ( * ioc ) ) ;
if ( IS_ERR ( ioc ) )
return PTR_ERR ( ioc ) ;
2005-09-20 21:46:58 +00:00
instance = megasas_lookup_instance ( ioc - > host_no ) ;
if ( ! instance ) {
error = - ENODEV ;
goto out_kfree_ioc ;
}
2014-03-10 09:51:56 +00:00
/* Block ioctls in VF mode */
if ( instance - > requestorId & & ! allow_vf_ioctls ) {
error = - ENODEV ;
goto out_kfree_ioc ;
}
2016-01-28 15:34:32 +00:00
if ( atomic_read ( & instance - > adprecovery ) = = MEGASAS_HW_CRITICAL_ERROR ) {
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " Controller in crit error \n " ) ;
2009-10-06 20:47:35 +00:00
error = - ENODEV ;
goto out_kfree_ioc ;
}
if ( instance - > unload = = 1 ) {
error = - ENODEV ;
goto out_kfree_ioc ;
}
2005-09-20 21:46:58 +00:00
if ( down_interruptible ( & instance - > ioctl_sem ) ) {
error = - ERESTARTSYS ;
goto out_kfree_ioc ;
}
2010-09-23 02:36:29 +00:00
2018-01-05 13:27:43 +00:00
if ( megasas_wait_for_adapter_operational ( instance ) ) {
2010-09-23 02:36:29 +00:00
error = - ENODEV ;
2013-04-16 07:44:19 +00:00
goto out_up ;
2010-09-23 02:36:29 +00:00
}
2005-09-20 21:46:58 +00:00
error = megasas_mgmt_fw_ioctl ( instance , user_ioc , ioc ) ;
2015-07-07 20:52:45 +00:00
out_up :
2005-09-20 21:46:58 +00:00
up ( & instance - > ioctl_sem ) ;
2015-07-07 20:52:45 +00:00
out_kfree_ioc :
2005-09-20 21:46:58 +00:00
kfree ( ioc ) ;
return error ;
}
static int megasas_mgmt_ioctl_aen ( struct file * file , unsigned long arg )
{
struct megasas_instance * instance ;
struct megasas_aen aen ;
int error ;
if ( file - > private_data ! = file ) {
printk ( KERN_DEBUG " megasas: fasync_helper was not "
" called first \n " ) ;
return - EINVAL ;
}
if ( copy_from_user ( & aen , ( void __user * ) arg , sizeof ( aen ) ) )
return - EFAULT ;
instance = megasas_lookup_instance ( aen . host_no ) ;
if ( ! instance )
return - ENODEV ;
2016-01-28 15:34:32 +00:00
if ( atomic_read ( & instance - > adprecovery ) = = MEGASAS_HW_CRITICAL_ERROR ) {
2010-09-23 02:36:29 +00:00
return - ENODEV ;
2009-10-06 20:47:35 +00:00
}
if ( instance - > unload = = 1 ) {
return - ENODEV ;
}
2018-01-05 13:27:43 +00:00
if ( megasas_wait_for_adapter_operational ( instance ) )
2010-09-23 02:36:29 +00:00
return - ENODEV ;
2016-01-28 15:34:22 +00:00
mutex_lock ( & instance - > reset_mutex ) ;
2005-09-20 21:46:58 +00:00
error = megasas_register_aen ( instance , aen . seq_num ,
aen . class_locale_word ) ;
2016-01-28 15:34:22 +00:00
mutex_unlock ( & instance - > reset_mutex ) ;
2005-09-20 21:46:58 +00:00
return error ;
}
/**
* megasas_mgmt_ioctl - char node ioctl entry point
*/
static long
megasas_mgmt_ioctl ( struct file * file , unsigned int cmd , unsigned long arg )
{
switch ( cmd ) {
case MEGASAS_IOC_FIRMWARE :
return megasas_mgmt_ioctl_fw ( file , arg ) ;
case MEGASAS_IOC_GET_AEN :
return megasas_mgmt_ioctl_aen ( file , arg ) ;
}
return - ENOTTY ;
}
# ifdef CONFIG_COMPAT
static int megasas_mgmt_compat_ioctl_fw ( struct file * file , unsigned long arg )
{
struct compat_megasas_iocpacket __user * cioc =
( struct compat_megasas_iocpacket __user * ) arg ;
struct megasas_iocpacket __user * ioc =
compat_alloc_user_space ( sizeof ( struct megasas_iocpacket ) ) ;
int i ;
int error = 0 ;
2010-02-11 17:01:50 +00:00
compat_uptr_t ptr ;
2015-10-15 08:10:54 +00:00
u32 local_sense_off ;
u32 local_sense_len ;
2016-01-28 15:44:26 +00:00
u32 user_sense_off ;
2005-09-20 21:46:58 +00:00
2006-10-04 10:34:03 +00:00
if ( clear_user ( ioc , sizeof ( * ioc ) ) )
return - EFAULT ;
2005-09-20 21:46:58 +00:00
if ( copy_in_user ( & ioc - > host_no , & cioc - > host_no , sizeof ( u16 ) ) | |
copy_in_user ( & ioc - > sgl_off , & cioc - > sgl_off , sizeof ( u32 ) ) | |
copy_in_user ( & ioc - > sense_off , & cioc - > sense_off , sizeof ( u32 ) ) | |
copy_in_user ( & ioc - > sense_len , & cioc - > sense_len , sizeof ( u32 ) ) | |
copy_in_user ( ioc - > frame . raw , cioc - > frame . raw , 128 ) | |
copy_in_user ( & ioc - > sge_count , & cioc - > sge_count , sizeof ( u32 ) ) )
return - EFAULT ;
2010-02-11 17:01:50 +00:00
/*
* The sense_ptr is used in megasas_mgmt_fw_ioctl only when
* sense_len is not null , so prepare the 64 bit value under
* the same condition .
*/
2016-01-28 15:44:26 +00:00
if ( get_user ( local_sense_off , & ioc - > sense_off ) | |
get_user ( local_sense_len , & ioc - > sense_len ) | |
get_user ( user_sense_off , & cioc - > sense_off ) )
2015-10-15 08:10:54 +00:00
return - EFAULT ;
scsi: megaraid_sas: fix a missing-check bug
In megasas_mgmt_compat_ioctl_fw(), to handle the structure
compat_megasas_iocpacket 'cioc', a user-space structure megasas_iocpacket
'ioc' is allocated before megasas_mgmt_ioctl_fw() is invoked to handle
the packet. Since the two data structures have different fields, the data
is copied from 'cioc' to 'ioc' field by field. In the copy process,
'sense_ptr' is prepared if the field 'sense_len' is not null, because it
will be used in megasas_mgmt_ioctl_fw(). To prepare 'sense_ptr', the
user-space data 'ioc->sense_off' and 'cioc->sense_off' are copied and
saved to kernel-space variables 'local_sense_off' and 'user_sense_off'
respectively. Given that 'ioc->sense_off' is also copied from
'cioc->sense_off', 'local_sense_off' and 'user_sense_off' should have the
same value. However, 'cioc' is in the user space and a malicious user can
race to change the value of 'cioc->sense_off' after it is copied to
'ioc->sense_off' but before it is copied to 'user_sense_off'. By doing
so, the attacker can inject different values into 'local_sense_off' and
'user_sense_off'. This can cause undefined behavior in the following
execution, because the two variables are supposed to be same.
This patch enforces a check on the two kernel variables 'local_sense_off'
and 'user_sense_off' to make sure they are the same after the copy. In
case they are not, an error code EINVAL will be returned.
Signed-off-by: Wenwen Wang <wang6495@umn.edu>
Acked-by: Sumit Saxena <sumit.saxena@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2018-10-06 18:34:21 +00:00
if ( local_sense_off ! = user_sense_off )
return - EINVAL ;
2015-10-15 08:10:54 +00:00
if ( local_sense_len ) {
2010-02-11 17:01:50 +00:00
void __user * * sense_ioc_ptr =
2016-01-28 15:44:26 +00:00
( void __user * * ) ( ( u8 * ) ( ( unsigned long ) & ioc - > frame . raw ) + local_sense_off ) ;
2010-02-11 17:01:50 +00:00
compat_uptr_t * sense_cioc_ptr =
2016-01-28 15:44:26 +00:00
( compat_uptr_t * ) ( ( ( unsigned long ) & cioc - > frame . raw ) + user_sense_off ) ;
2010-02-11 17:01:50 +00:00
if ( get_user ( ptr , sense_cioc_ptr ) | |
put_user ( compat_ptr ( ptr ) , sense_ioc_ptr ) )
return - EFAULT ;
}
2005-09-20 21:46:58 +00:00
2010-02-11 17:01:50 +00:00
for ( i = 0 ; i < MAX_IOCTL_SGE ; i + + ) {
2005-09-20 21:46:58 +00:00
if ( get_user ( ptr , & cioc - > sgl [ i ] . iov_base ) | |
put_user ( compat_ptr ( ptr ) , & ioc - > sgl [ i ] . iov_base ) | |
copy_in_user ( & ioc - > sgl [ i ] . iov_len ,
& cioc - > sgl [ i ] . iov_len , sizeof ( compat_size_t ) ) )
return - EFAULT ;
}
error = megasas_mgmt_ioctl_fw ( file , ( unsigned long ) ioc ) ;
if ( copy_in_user ( & cioc - > frame . hdr . cmd_status ,
& ioc - > frame . hdr . cmd_status , sizeof ( u8 ) ) ) {
printk ( KERN_DEBUG " megasas: error copy_in_user cmd_status \n " ) ;
return - EFAULT ;
}
return error ;
}
static long
megasas_mgmt_compat_ioctl ( struct file * file , unsigned int cmd ,
unsigned long arg )
{
switch ( cmd ) {
2006-01-25 19:53:25 +00:00
case MEGASAS_IOC_FIRMWARE32 :
return megasas_mgmt_compat_ioctl_fw ( file , arg ) ;
2005-09-20 21:46:58 +00:00
case MEGASAS_IOC_GET_AEN :
return megasas_mgmt_ioctl_aen ( file , arg ) ;
}
return - ENOTTY ;
}
# endif
/*
* File operations structure for management interface
*/
2007-02-12 08:55:34 +00:00
static const struct file_operations megasas_mgmt_fops = {
2005-09-20 21:46:58 +00:00
. owner = THIS_MODULE ,
. open = megasas_mgmt_open ,
. fasync = megasas_mgmt_fasync ,
. unlocked_ioctl = megasas_mgmt_ioctl ,
2009-10-06 20:18:02 +00:00
. poll = megasas_mgmt_poll ,
2005-09-20 21:46:58 +00:00
# ifdef CONFIG_COMPAT
. compat_ioctl = megasas_mgmt_compat_ioctl ,
# endif
llseek: automatically add .llseek fop
All file_operations should get a .llseek operation so we can make
nonseekable_open the default for future file operations without a
.llseek pointer.
The three cases that we can automatically detect are no_llseek, seq_lseek
and default_llseek. For cases where we can we can automatically prove that
the file offset is always ignored, we use noop_llseek, which maintains
the current behavior of not returning an error from a seek.
New drivers should normally not use noop_llseek but instead use no_llseek
and call nonseekable_open at open time. Existing drivers can be converted
to do the same when the maintainer knows for certain that no user code
relies on calling seek on the device file.
The generated code is often incorrectly indented and right now contains
comments that clarify for each added line why a specific variant was
chosen. In the version that gets submitted upstream, the comments will
be gone and I will manually fix the indentation, because there does not
seem to be a way to do that using coccinelle.
Some amount of new code is currently sitting in linux-next that should get
the same modifications, which I will do at the end of the merge window.
Many thanks to Julia Lawall for helping me learn to write a semantic
patch that does all this.
===== begin semantic patch =====
// This adds an llseek= method to all file operations,
// as a preparation for making no_llseek the default.
//
// The rules are
// - use no_llseek explicitly if we do nonseekable_open
// - use seq_lseek for sequential files
// - use default_llseek if we know we access f_pos
// - use noop_llseek if we know we don't access f_pos,
// but we still want to allow users to call lseek
//
@ open1 exists @
identifier nested_open;
@@
nested_open(...)
{
<+...
nonseekable_open(...)
...+>
}
@ open exists@
identifier open_f;
identifier i, f;
identifier open1.nested_open;
@@
int open_f(struct inode *i, struct file *f)
{
<+...
(
nonseekable_open(...)
|
nested_open(...)
)
...+>
}
@ read disable optional_qualifier exists @
identifier read_f;
identifier f, p, s, off;
type ssize_t, size_t, loff_t;
expression E;
identifier func;
@@
ssize_t read_f(struct file *f, char *p, size_t s, loff_t *off)
{
<+...
(
*off = E
|
*off += E
|
func(..., off, ...)
|
E = *off
)
...+>
}
@ read_no_fpos disable optional_qualifier exists @
identifier read_f;
identifier f, p, s, off;
type ssize_t, size_t, loff_t;
@@
ssize_t read_f(struct file *f, char *p, size_t s, loff_t *off)
{
... when != off
}
@ write @
identifier write_f;
identifier f, p, s, off;
type ssize_t, size_t, loff_t;
expression E;
identifier func;
@@
ssize_t write_f(struct file *f, const char *p, size_t s, loff_t *off)
{
<+...
(
*off = E
|
*off += E
|
func(..., off, ...)
|
E = *off
)
...+>
}
@ write_no_fpos @
identifier write_f;
identifier f, p, s, off;
type ssize_t, size_t, loff_t;
@@
ssize_t write_f(struct file *f, const char *p, size_t s, loff_t *off)
{
... when != off
}
@ fops0 @
identifier fops;
@@
struct file_operations fops = {
...
};
@ has_llseek depends on fops0 @
identifier fops0.fops;
identifier llseek_f;
@@
struct file_operations fops = {
...
.llseek = llseek_f,
...
};
@ has_read depends on fops0 @
identifier fops0.fops;
identifier read_f;
@@
struct file_operations fops = {
...
.read = read_f,
...
};
@ has_write depends on fops0 @
identifier fops0.fops;
identifier write_f;
@@
struct file_operations fops = {
...
.write = write_f,
...
};
@ has_open depends on fops0 @
identifier fops0.fops;
identifier open_f;
@@
struct file_operations fops = {
...
.open = open_f,
...
};
// use no_llseek if we call nonseekable_open
////////////////////////////////////////////
@ nonseekable1 depends on !has_llseek && has_open @
identifier fops0.fops;
identifier nso ~= "nonseekable_open";
@@
struct file_operations fops = {
... .open = nso, ...
+.llseek = no_llseek, /* nonseekable */
};
@ nonseekable2 depends on !has_llseek @
identifier fops0.fops;
identifier open.open_f;
@@
struct file_operations fops = {
... .open = open_f, ...
+.llseek = no_llseek, /* open uses nonseekable */
};
// use seq_lseek for sequential files
/////////////////////////////////////
@ seq depends on !has_llseek @
identifier fops0.fops;
identifier sr ~= "seq_read";
@@
struct file_operations fops = {
... .read = sr, ...
+.llseek = seq_lseek, /* we have seq_read */
};
// use default_llseek if there is a readdir
///////////////////////////////////////////
@ fops1 depends on !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
identifier readdir_e;
@@
// any other fop is used that changes pos
struct file_operations fops = {
... .readdir = readdir_e, ...
+.llseek = default_llseek, /* readdir is present */
};
// use default_llseek if at least one of read/write touches f_pos
/////////////////////////////////////////////////////////////////
@ fops2 depends on !fops1 && !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
identifier read.read_f;
@@
// read fops use offset
struct file_operations fops = {
... .read = read_f, ...
+.llseek = default_llseek, /* read accesses f_pos */
};
@ fops3 depends on !fops1 && !fops2 && !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
identifier write.write_f;
@@
// write fops use offset
struct file_operations fops = {
... .write = write_f, ...
+ .llseek = default_llseek, /* write accesses f_pos */
};
// Use noop_llseek if neither read nor write accesses f_pos
///////////////////////////////////////////////////////////
@ fops4 depends on !fops1 && !fops2 && !fops3 && !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
identifier read_no_fpos.read_f;
identifier write_no_fpos.write_f;
@@
// write fops use offset
struct file_operations fops = {
...
.write = write_f,
.read = read_f,
...
+.llseek = noop_llseek, /* read and write both use no f_pos */
};
@ depends on has_write && !has_read && !fops1 && !fops2 && !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
identifier write_no_fpos.write_f;
@@
struct file_operations fops = {
... .write = write_f, ...
+.llseek = noop_llseek, /* write uses no f_pos */
};
@ depends on has_read && !has_write && !fops1 && !fops2 && !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
identifier read_no_fpos.read_f;
@@
struct file_operations fops = {
... .read = read_f, ...
+.llseek = noop_llseek, /* read uses no f_pos */
};
@ depends on !has_read && !has_write && !fops1 && !fops2 && !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
@@
struct file_operations fops = {
...
+.llseek = noop_llseek, /* no read or write fn */
};
===== End semantic patch =====
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Julia Lawall <julia@diku.dk>
Cc: Christoph Hellwig <hch@infradead.org>
2010-08-15 16:52:59 +00:00
. llseek = noop_llseek ,
2005-09-20 21:46:58 +00:00
} ;
/*
* PCI hotplug support registration structure
*/
static struct pci_driver megasas_pci_driver = {
. name = " megaraid_sas " ,
. id_table = megasas_pci_table ,
. probe = megasas_probe_one ,
2012-12-21 21:08:55 +00:00
. remove = megasas_detach_one ,
2007-11-07 17:09:50 +00:00
. suspend = megasas_suspend ,
. resume = megasas_resume ,
2005-09-20 21:46:58 +00:00
. shutdown = megasas_shutdown ,
} ;
/*
* Sysfs driver attributes
*/
2017-07-19 12:50:06 +00:00
static ssize_t version_show ( struct device_driver * dd , char * buf )
2005-09-20 21:46:58 +00:00
{
return snprintf ( buf , strlen ( MEGASAS_VERSION ) + 2 , " %s \n " ,
MEGASAS_VERSION ) ;
}
2017-07-19 12:50:06 +00:00
static DRIVER_ATTR_RO ( version ) ;
2005-09-20 21:46:58 +00:00
2017-07-19 12:50:06 +00:00
static ssize_t release_date_show ( struct device_driver * dd , char * buf )
2015-04-23 11:01:54 +00:00
{
return snprintf ( buf , strlen ( MEGASAS_RELDATE ) + 2 , " %s \n " ,
MEGASAS_RELDATE ) ;
}
2017-07-19 12:50:06 +00:00
static DRIVER_ATTR_RO ( release_date ) ;
2015-04-23 11:01:54 +00:00
2017-07-19 12:50:06 +00:00
static ssize_t support_poll_for_event_show ( struct device_driver * dd , char * buf )
2009-10-06 20:20:59 +00:00
{
return sprintf ( buf , " %u \n " , support_poll_for_event ) ;
}
2017-07-19 12:50:06 +00:00
static DRIVER_ATTR_RO ( support_poll_for_event ) ;
2009-10-06 20:20:59 +00:00
2017-07-19 12:50:06 +00:00
static ssize_t support_device_change_show ( struct device_driver * dd , char * buf )
2010-10-11 12:59:20 +00:00
{
return sprintf ( buf , " %u \n " , support_device_change ) ;
}
2017-07-19 12:50:06 +00:00
static DRIVER_ATTR_RO ( support_device_change ) ;
2010-10-11 12:59:20 +00:00
2017-07-19 12:50:06 +00:00
static ssize_t dbg_lvl_show ( struct device_driver * dd , char * buf )
2006-10-03 20:09:14 +00:00
{
2007-11-09 09:40:16 +00:00
return sprintf ( buf , " %u \n " , megasas_dbg_lvl ) ;
2006-10-03 20:09:14 +00:00
}
2017-07-19 12:50:06 +00:00
static ssize_t dbg_lvl_store ( struct device_driver * dd , const char * buf ,
size_t count )
2006-10-03 20:09:14 +00:00
{
int retval = count ;
2015-07-07 20:52:45 +00:00
if ( sscanf ( buf , " %u " , & megasas_dbg_lvl ) < 1 ) {
2006-10-03 20:09:14 +00:00
printk ( KERN_ERR " megasas: could not set dbg_lvl \n " ) ;
retval = - EINVAL ;
}
return retval ;
}
2017-07-19 12:50:06 +00:00
static DRIVER_ATTR_RW ( dbg_lvl ) ;
2007-11-09 09:40:16 +00:00
2018-01-05 13:33:04 +00:00
static ssize_t
support_nvme_encapsulation_show ( struct device_driver * dd , char * buf )
{
return sprintf ( buf , " %u \n " , support_nvme_encapsulation ) ;
}
static DRIVER_ATTR_RO ( support_nvme_encapsulation ) ;
2019-06-25 11:04:30 +00:00
static ssize_t
support_pci_lane_margining_show ( struct device_driver * dd , char * buf )
{
return sprintf ( buf , " %u \n " , support_pci_lane_margining ) ;
}
static DRIVER_ATTR_RO ( support_pci_lane_margining ) ;
2017-02-10 08:59:16 +00:00
static inline void megasas_remove_scsi_device ( struct scsi_device * sdev )
{
sdev_printk ( KERN_INFO , sdev , " SCSI device is removed \n " ) ;
scsi_remove_device ( sdev ) ;
scsi_device_put ( sdev ) ;
}
2019-01-29 09:38:13 +00:00
/**
* megasas_update_device_list - Update the PD and LD device list from FW
* after an AEN event notification
* @ instance : Adapter soft state
* @ event_type : Indicates type of event ( PD or LD event )
*
* @ return : Success or failure
*
* Issue DCMDs to Firmware to update the internal device list in driver .
2019-01-29 09:38:14 +00:00
* Based on the FW support , driver sends the HOST_DEVICE_LIST or combination
* of PD_LIST / LD_LIST_QUERY DCMDs to get the device list .
2019-01-29 09:38:13 +00:00
*/
static
int megasas_update_device_list ( struct megasas_instance * instance ,
int event_type )
{
int dcmd_ret = DCMD_SUCCESS ;
2019-01-29 09:38:14 +00:00
if ( instance - > enable_fw_dev_list ) {
dcmd_ret = megasas_host_device_list_query ( instance , false ) ;
2019-01-29 09:38:13 +00:00
if ( dcmd_ret ! = DCMD_SUCCESS )
goto out ;
2019-01-29 09:38:14 +00:00
} else {
if ( event_type & SCAN_PD_CHANNEL ) {
dcmd_ret = megasas_get_pd_list ( instance ) ;
2019-01-29 09:38:13 +00:00
if ( dcmd_ret ! = DCMD_SUCCESS )
goto out ;
}
2019-01-29 09:38:14 +00:00
if ( event_type & SCAN_VD_CHANNEL ) {
if ( ! instance - > requestorId | |
( instance - > requestorId & &
megasas_get_ld_vf_affiliation ( instance , 0 ) ) ) {
dcmd_ret = megasas_ld_list_query ( instance ,
MR_LD_QUERY_TYPE_EXPOSED_TO_HOST ) ;
if ( dcmd_ret ! = DCMD_SUCCESS )
goto out ;
}
}
2019-01-29 09:38:13 +00:00
}
out :
return dcmd_ret ;
}
/**
* megasas_add_remove_devices - Add / remove devices to SCSI mid - layer
* after an AEN event notification
* @ instance : Adapter soft state
* @ scan_type : Indicates type of devices ( PD / LD ) to add
* @ return void
*/
static
void megasas_add_remove_devices ( struct megasas_instance * instance ,
int scan_type )
{
int i , j ;
u16 pd_index = 0 ;
u16 ld_index = 0 ;
2019-01-29 09:38:14 +00:00
u16 channel = 0 , id = 0 ;
2019-01-29 09:38:13 +00:00
struct Scsi_Host * host ;
struct scsi_device * sdev1 ;
2019-01-29 09:38:14 +00:00
struct MR_HOST_DEVICE_LIST * targetid_list = NULL ;
struct MR_HOST_DEVICE_LIST_ENTRY * targetid_entry = NULL ;
2019-01-29 09:38:13 +00:00
host = instance - > host ;
2019-01-29 09:38:14 +00:00
if ( instance - > enable_fw_dev_list ) {
targetid_list = instance - > host_device_list_buf ;
for ( i = 0 ; i < targetid_list - > count ; i + + ) {
targetid_entry = & targetid_list - > host_device_list [ i ] ;
if ( targetid_entry - > flags . u . bits . is_sys_pd ) {
channel = le16_to_cpu ( targetid_entry - > target_id ) /
MEGASAS_MAX_DEV_PER_CHANNEL ;
id = le16_to_cpu ( targetid_entry - > target_id ) %
MEGASAS_MAX_DEV_PER_CHANNEL ;
} else {
channel = MEGASAS_MAX_PD_CHANNELS +
( le16_to_cpu ( targetid_entry - > target_id ) /
MEGASAS_MAX_DEV_PER_CHANNEL ) ;
id = le16_to_cpu ( targetid_entry - > target_id ) %
MEGASAS_MAX_DEV_PER_CHANNEL ;
}
sdev1 = scsi_device_lookup ( host , channel , id , 0 ) ;
if ( ! sdev1 ) {
scsi_add_device ( host , channel , id , 0 ) ;
} else {
scsi_device_put ( sdev1 ) ;
}
}
}
2019-01-29 09:38:13 +00:00
if ( scan_type & SCAN_PD_CHANNEL ) {
for ( i = 0 ; i < MEGASAS_MAX_PD_CHANNELS ; i + + ) {
for ( j = 0 ; j < MEGASAS_MAX_DEV_PER_CHANNEL ; j + + ) {
pd_index = i * MEGASAS_MAX_DEV_PER_CHANNEL + j ;
sdev1 = scsi_device_lookup ( host , i , j , 0 ) ;
if ( instance - > pd_list [ pd_index ] . driveState = =
MR_PD_STATE_SYSTEM ) {
if ( ! sdev1 )
scsi_add_device ( host , i , j , 0 ) ;
else
scsi_device_put ( sdev1 ) ;
} else {
if ( sdev1 )
megasas_remove_scsi_device ( sdev1 ) ;
}
}
}
}
if ( scan_type & SCAN_VD_CHANNEL ) {
for ( i = 0 ; i < MEGASAS_MAX_LD_CHANNELS ; i + + ) {
for ( j = 0 ; j < MEGASAS_MAX_DEV_PER_CHANNEL ; j + + ) {
ld_index = ( i * MEGASAS_MAX_DEV_PER_CHANNEL ) + j ;
sdev1 = scsi_device_lookup ( host ,
MEGASAS_MAX_PD_CHANNELS + i , j , 0 ) ;
if ( instance - > ld_ids [ ld_index ] ! = 0xff ) {
if ( ! sdev1 )
scsi_add_device ( host , MEGASAS_MAX_PD_CHANNELS + i , j , 0 ) ;
else
scsi_device_put ( sdev1 ) ;
} else {
if ( sdev1 )
megasas_remove_scsi_device ( sdev1 ) ;
}
}
}
}
}
2009-10-06 20:50:17 +00:00
static void
megasas_aen_polling ( struct work_struct * work )
{
struct megasas_aen_event * ev =
2012-12-04 11:33:54 +00:00
container_of ( work , struct megasas_aen_event , hotplug_work . work ) ;
2009-10-06 20:50:17 +00:00
struct megasas_instance * instance = ev - > instance ;
union megasas_evt_class_locale class_locale ;
2019-01-29 09:38:13 +00:00
int event_type = 0 ;
2019-05-25 12:42:02 +00:00
u32 seq_num ;
2009-10-06 20:50:17 +00:00
int error ;
2016-01-28 15:34:23 +00:00
u8 dcmd_ret = DCMD_SUCCESS ;
2009-10-06 20:50:17 +00:00
if ( ! instance ) {
printk ( KERN_ERR " invalid instance! \n " ) ;
kfree ( ev ) ;
return ;
}
2014-03-10 09:51:56 +00:00
/* Don't run the event workqueue thread if OCR is running */
2016-01-28 15:34:22 +00:00
mutex_lock ( & instance - > reset_mutex ) ;
2014-03-10 09:51:56 +00:00
2009-10-06 20:50:17 +00:00
instance - > ev = NULL ;
if ( instance - > evt_detail ) {
2015-08-31 11:53:51 +00:00
megasas_decode_evt ( instance ) ;
2009-10-06 20:50:17 +00:00
2013-09-06 10:20:52 +00:00
switch ( le32_to_cpu ( instance - > evt_detail - > code ) ) {
2009-12-06 15:39:25 +00:00
2016-01-28 15:34:22 +00:00
case MR_EVT_PD_INSERTED :
2009-10-06 20:50:17 +00:00
case MR_EVT_PD_REMOVED :
2019-01-29 09:38:13 +00:00
event_type = SCAN_PD_CHANNEL ;
2009-12-06 15:39:25 +00:00
break ;
case MR_EVT_LD_OFFLINE :
2011-02-25 04:56:53 +00:00
case MR_EVT_CFG_CLEARED :
2009-12-06 15:39:25 +00:00
case MR_EVT_LD_DELETED :
case MR_EVT_LD_CREATED :
2019-01-29 09:38:13 +00:00
event_type = SCAN_VD_CHANNEL ;
2009-12-06 15:39:25 +00:00
break ;
2016-01-28 15:34:22 +00:00
2009-10-06 20:50:17 +00:00
case MR_EVT_CTRL_HOST_BUS_SCAN_REQUESTED :
2009-12-06 15:39:25 +00:00
case MR_EVT_FOREIGN_CFG_IMPORTED :
2010-12-21 21:34:31 +00:00
case MR_EVT_LD_STATE_CHANGE :
2019-01-29 09:38:13 +00:00
event_type = SCAN_PD_CHANNEL | SCAN_VD_CHANNEL ;
2016-01-28 15:34:22 +00:00
dev_info ( & instance - > pdev - > dev , " scanning for scsi%d... \n " ,
instance - > host - > host_no ) ;
2009-10-06 20:50:17 +00:00
break ;
2016-01-28 15:34:22 +00:00
2015-10-15 08:10:14 +00:00
case MR_EVT_CTRL_PROP_CHANGED :
2018-10-17 06:37:40 +00:00
dcmd_ret = megasas_get_ctrl_info ( instance ) ;
if ( dcmd_ret = = DCMD_SUCCESS & &
instance - > snapdump_wait_time ) {
megasas_get_snapdump_properties ( instance ) ;
dev_info ( & instance - > pdev - > dev ,
" Snap dump wait time \t : %d \n " ,
instance - > snapdump_wait_time ) ;
}
break ;
2009-10-06 20:50:17 +00:00
default :
2019-01-29 09:38:13 +00:00
event_type = 0 ;
2009-10-06 20:50:17 +00:00
break ;
}
} else {
2015-07-07 20:52:34 +00:00
dev_err ( & instance - > pdev - > dev , " invalid evt_detail! \n " ) ;
2016-01-28 15:34:22 +00:00
mutex_unlock ( & instance - > reset_mutex ) ;
2009-10-06 20:50:17 +00:00
kfree ( ev ) ;
return ;
}
2019-01-29 09:38:13 +00:00
if ( event_type )
dcmd_ret = megasas_update_device_list ( instance , event_type ) ;
2016-01-28 15:34:22 +00:00
2019-01-29 09:38:13 +00:00
mutex_unlock ( & instance - > reset_mutex ) ;
2009-12-06 15:39:25 +00:00
2019-01-29 09:38:13 +00:00
if ( event_type & & dcmd_ret = = DCMD_SUCCESS )
megasas_add_remove_devices ( instance , event_type ) ;
2009-10-06 20:50:17 +00:00
2016-01-28 15:34:23 +00:00
if ( dcmd_ret = = DCMD_SUCCESS )
2016-01-28 15:34:22 +00:00
seq_num = le32_to_cpu ( instance - > evt_detail - > seq_num ) + 1 ;
else
seq_num = instance - > last_seq_num ;
2009-10-06 20:50:17 +00:00
/* Register AEN with FW for latest sequence number plus 1 */
class_locale . members . reserved = 0 ;
class_locale . members . locale = MR_EVT_LOCALE_ALL ;
class_locale . members . class = MR_EVT_CLASS_DEBUG ;
2016-01-28 15:34:22 +00:00
if ( instance - > aen_cmd ! = NULL ) {
kfree ( ev ) ;
return ;
}
mutex_lock ( & instance - > reset_mutex ) ;
2009-10-06 20:50:17 +00:00
error = megasas_register_aen ( instance , seq_num ,
class_locale . word ) ;
if ( error )
2016-01-28 15:34:22 +00:00
dev_err ( & instance - > pdev - > dev ,
" register aen failed error %x \n " , error ) ;
2009-10-06 20:50:17 +00:00
2016-01-28 15:34:22 +00:00
mutex_unlock ( & instance - > reset_mutex ) ;
2009-10-06 20:50:17 +00:00
kfree ( ev ) ;
}
2005-09-20 21:46:58 +00:00
/**
* megasas_init - Driver load entry point
*/
static int __init megasas_init ( void )
{
int rval ;
2016-04-15 07:23:30 +00:00
/*
* Booted in kdump kernel , minimize memory footprints by
* disabling few features
*/
if ( reset_devices ) {
msix_vectors = 1 ;
rdpq_enable = 0 ;
dual_qdepth_disable = 1 ;
}
2005-09-20 21:46:58 +00:00
/*
* Announce driver version and other information
*/
2014-11-17 09:53:58 +00:00
pr_info ( " megasas: %s \n " , MEGASAS_VERSION ) ;
2005-09-20 21:46:58 +00:00
2012-07-18 01:20:44 +00:00
spin_lock_init ( & poll_aen_lock ) ;
2009-10-06 20:20:59 +00:00
support_poll_for_event = 2 ;
2010-10-11 12:59:20 +00:00
support_device_change = 1 ;
2018-01-05 13:33:04 +00:00
support_nvme_encapsulation = true ;
2019-06-25 11:04:30 +00:00
support_pci_lane_margining = true ;
2009-10-06 20:20:59 +00:00
2005-09-20 21:46:58 +00:00
memset ( & megasas_mgmt_info , 0 , sizeof ( megasas_mgmt_info ) ) ;
/*
* Register character device node
*/
rval = register_chrdev ( 0 , " megaraid_sas_ioctl " , & megasas_mgmt_fops ) ;
if ( rval < 0 ) {
printk ( KERN_DEBUG " megasas: failed to open device node \n " ) ;
return rval ;
}
megasas_mgmt_majorno = rval ;
2019-05-07 17:05:49 +00:00
megasas_init_debugfs ( ) ;
2005-09-20 21:46:58 +00:00
/*
* Register ourselves as PCI hotplug module
*/
2006-08-17 13:28:22 +00:00
rval = pci_register_driver ( & megasas_pci_driver ) ;
2005-09-20 21:46:58 +00:00
if ( rval ) {
2014-11-05 13:26:48 +00:00
printk ( KERN_DEBUG " megasas: PCI hotplug registration failed \n " ) ;
2006-10-04 10:34:03 +00:00
goto err_pcidrv ;
}
2019-06-28 09:50:40 +00:00
if ( ( event_log_level < MFI_EVT_CLASS_DEBUG ) | |
( event_log_level > MFI_EVT_CLASS_DEAD ) ) {
2019-07-20 21:58:40 +00:00
pr_warn ( " megaraid_sas: provided event log level is out of range, setting it to default 2(CLASS_CRITICAL), permissible range is: -2 to 4 \n " ) ;
2019-06-28 09:50:40 +00:00
event_log_level = MFI_EVT_CLASS_CRITICAL ;
}
2006-10-04 10:34:03 +00:00
rval = driver_create_file ( & megasas_pci_driver . driver ,
& driver_attr_version ) ;
if ( rval )
goto err_dcf_attr_ver ;
2009-10-06 20:20:59 +00:00
2015-04-23 11:01:54 +00:00
rval = driver_create_file ( & megasas_pci_driver . driver ,
& driver_attr_release_date ) ;
if ( rval )
goto err_dcf_rel_date ;
2009-10-06 20:20:59 +00:00
rval = driver_create_file ( & megasas_pci_driver . driver ,
& driver_attr_support_poll_for_event ) ;
if ( rval )
goto err_dcf_support_poll_for_event ;
2006-10-04 10:34:03 +00:00
rval = driver_create_file ( & megasas_pci_driver . driver ,
& driver_attr_dbg_lvl ) ;
if ( rval )
goto err_dcf_dbg_lvl ;
2010-10-11 12:59:20 +00:00
rval = driver_create_file ( & megasas_pci_driver . driver ,
& driver_attr_support_device_change ) ;
if ( rval )
goto err_dcf_support_device_change ;
2018-01-05 13:33:04 +00:00
rval = driver_create_file ( & megasas_pci_driver . driver ,
& driver_attr_support_nvme_encapsulation ) ;
if ( rval )
goto err_dcf_support_nvme_encapsulation ;
2019-06-25 11:04:30 +00:00
rval = driver_create_file ( & megasas_pci_driver . driver ,
& driver_attr_support_pci_lane_margining ) ;
if ( rval )
goto err_dcf_support_pci_lane_margining ;
2005-09-20 21:46:58 +00:00
return rval ;
2007-11-09 09:40:16 +00:00
2019-06-25 11:04:30 +00:00
err_dcf_support_pci_lane_margining :
driver_remove_file ( & megasas_pci_driver . driver ,
& driver_attr_support_nvme_encapsulation ) ;
2018-01-05 13:33:04 +00:00
err_dcf_support_nvme_encapsulation :
driver_remove_file ( & megasas_pci_driver . driver ,
& driver_attr_support_device_change ) ;
2010-10-11 12:59:20 +00:00
err_dcf_support_device_change :
2007-11-09 09:40:16 +00:00
driver_remove_file ( & megasas_pci_driver . driver ,
& driver_attr_dbg_lvl ) ;
2006-10-04 10:34:03 +00:00
err_dcf_dbg_lvl :
2009-10-06 20:20:59 +00:00
driver_remove_file ( & megasas_pci_driver . driver ,
& driver_attr_support_poll_for_event ) ;
err_dcf_support_poll_for_event :
2015-04-23 11:01:54 +00:00
driver_remove_file ( & megasas_pci_driver . driver ,
& driver_attr_release_date ) ;
err_dcf_rel_date :
2006-10-04 10:34:03 +00:00
driver_remove_file ( & megasas_pci_driver . driver , & driver_attr_version ) ;
err_dcf_attr_ver :
pci_unregister_driver ( & megasas_pci_driver ) ;
err_pcidrv :
2019-05-07 17:05:49 +00:00
megasas_exit_debugfs ( ) ;
2006-10-04 10:34:03 +00:00
unregister_chrdev ( megasas_mgmt_majorno , " megaraid_sas_ioctl " ) ;
2010-12-15 03:17:17 +00:00
return rval ;
2005-09-20 21:46:58 +00:00
}
/**
* megasas_exit - Driver unload entry point
*/
static void __exit megasas_exit ( void )
{
2006-10-03 20:09:14 +00:00
driver_remove_file ( & megasas_pci_driver . driver ,
& driver_attr_dbg_lvl ) ;
2010-10-11 12:59:20 +00:00
driver_remove_file ( & megasas_pci_driver . driver ,
& driver_attr_support_poll_for_event ) ;
driver_remove_file ( & megasas_pci_driver . driver ,
& driver_attr_support_device_change ) ;
2015-04-23 11:01:54 +00:00
driver_remove_file ( & megasas_pci_driver . driver ,
& driver_attr_release_date ) ;
2006-10-04 10:34:03 +00:00
driver_remove_file ( & megasas_pci_driver . driver , & driver_attr_version ) ;
2018-01-05 13:33:04 +00:00
driver_remove_file ( & megasas_pci_driver . driver ,
& driver_attr_support_nvme_encapsulation ) ;
2019-06-25 11:04:30 +00:00
driver_remove_file ( & megasas_pci_driver . driver ,
& driver_attr_support_pci_lane_margining ) ;
2005-09-20 21:46:58 +00:00
pci_unregister_driver ( & megasas_pci_driver ) ;
2019-05-07 17:05:49 +00:00
megasas_exit_debugfs ( ) ;
2005-09-20 21:46:58 +00:00
unregister_chrdev ( megasas_mgmt_majorno , " megaraid_sas_ioctl " ) ;
}
module_init ( megasas_init ) ;
module_exit ( megasas_exit ) ;