This callback further invokes the mlxfw module to flash the new
firmware file to the device.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Yotam Gigi <yotamg@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
This callback further invokes the mlxfw module to flash the new
firmware file to the device.
As the firmware flash process takes about 20 seconds and ethtool
takes the rtnl lock during the flash_device callback, we release
the rtnl lock at the beginning of the flash process and take it
again before leaving the callback.
This way, rtnl is not held during the process. To make sure the
device does not get deleted while being flashed, we take a
reference to it before releasing rtnl lock.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Yotam Gigi <yotamg@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Add mlx5 implementation for the ones defined by the mlxfw
shared module to be used while flashing the device firmware.
The callbacks do their job through the MCQI, MCC and MCDA registers.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Yotam Gigi <yotamg@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
To be used by the mlx5 callbacks exposed to the mlxfw module.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Yotam Gigi <yotamg@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
There are upcoming NIC (mlx5) use-cases where people want to avoid
building the mlxfw module, allow for that. The mlxsw module is
untouched and keeps selecting mlxfw.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Acked-by: Yotam Gigi <yotamg@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
For environments where flow-based ipv6 router is offloaded.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Use a macro for the static mapping between the enumeration of field
supported by the firmware for header re-write to the corresponding
network header field. This improves the readability of the code and
doesn't change any functionality.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Enable offloading of TC matching on ip ttl / hop-limit
As matching on ttl is supported only by newer HW brands (ConnectX-5),
we should do capability check before attempting to offload that.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
The code section for offloading matches on ip tos (L3) should come
before and not after the one that deals with tcp/udp (L4) matches.
Otherwise, we might come up with wrong min-inline requirement, when
one attempts to match on both L3 and L4.
Fixes: fd7da28b28 ('net/mlx5e: Offload TC matching on ip tos / traffic-class')
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Introduce a Page-Reuse mechanism in non-Striding RQ RX datapath.
A WQE (RX descriptor) buffer is a page, that in most cases was fully
wasted on a packet that is much smaller, requiring a new page for
the next round.
In this patch, we implement a page-reuse mechanism, that resembles a
`SW Striding RQ`.
We allow the WQE to reuse its allocated page as much as it could,
until the page is fully consumed. In each round, the WQE is capable
of receiving packet of maximal size (MTU). Yet, upon the reception of
a packet, the WQE knows the actual packet size, and consumes the exact
amount of memory needed to build a linear SKB. Then, it updates the
buffer pointer within the page accordingly, for the next round.
Feature is mutually exclusive with XDP (packet-per-page)
and LRO (session size is a power of two, needs unused page).
Performance tests:
iperf tcp tests show huge gain:
--------------------------------------------
num streams | BW before | BW after | ratio |
1 | 22.2 | 30.9 | 1.39x |
8 | 64.2 | 93.6 | 1.46x |
64 | 56.7 | 91.4 | 1.61x |
--------------------------------------------
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
In the RX memory scheme of non Striding RQ, we use linear SKBs.
Keeping NET_IP_ALIGN in headroom can improve performance on some archs.
In addition, take this headroom into account when calculating the
LRO WQE size.
These are not needed in Striding RQ as they're done implicitly
within the non-linear SKB allocation.
Fixes: 1bfecfca56 ("net/mlx5e: Build RX SKB on demand")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Build the SKB over the receive packet instead of the
whole page. Getting the SKB's linear data and shared_info
closer improves locality.
In addition, this opens up the possibility to make use of
other parts of the page in the downstream page-reuse patch.
Fixes: 1bfecfca56 ("net/mlx5e: Build RX SKB on demand")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Two entries being added at the same time to the IFLA
policy table, whilst parallel bug fixes to decnet
routing dst handling overlapping with the dst gc removal
in net-next.
Signed-off-by: David S. Miller <davem@davemloft.net>
Rename the qed_roce_if file to qed_rdma_if as it
represents a common interface for RoCE and iWARP.
this commit affects RDMA/qedr as well.
Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch places common iWARP / RoCE code in qed_rdma
and roce specific code in qed_roce
There is one new function ( qed_roce_setup ) added, the rest
of the patch removes content from the files and removes some
static definitions.
Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds files that will contain common code for RoCE/iWARP.
The files are currently identical to qed_roce.c / qed_roce.h and
intentionally not added to the makefile. The next patch in the series
will modify the files so that roce specific code is left in qed_roce
and common roce/iwarp code will be placed in qed_rdma
This patch is the result of a simple
cp qed_rdma.c qed_roce.c
cp qed_rdma.h qed_roce.h
Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The next patch in the series will duplicate qed_roce as part
of code preprations for iWARP support. Do some cleanup before
duplicating
Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We recently refactored i40e_do_reset() and its friends to be able to
hold the RTNL lock only for the portions that actually need to be
protected. However, a separate refactoring added several new callers of
these functions during the PCIe error recovery and suspend/resume
cycles.
When merging the changes together, it was not noticed that we could
reduce the RTNL scope by letting the reset function handle the lock
itself, as previously it was not possible.
Fix this by replacing these call sites to indicate that the reset
function should handle its own lock. This enables multiple PFs to reset
or resume simultaneously without serializing the resets via the RTNL
lock. The end result is that on systems with lots of PFs and VFs the
resets don't stall waiting for each other to finish.
It is probable that we can also do the same for i40e_do_reset_safe, but
this author did not research that change carefully enough to be
confident.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
When IWARP is enabled, we weren't clearing the PE_CRITERR, just logging
it and removing it from the mask. We need to do a corer to reset the
PE_CRITERR register, so set the bit for that as we handle the
interrupt.
We should also be checking for the error against the PFINT_ICR0 register,
and only need to clear it in the value getting written to
PFINT_ICR0_ENA.
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
When disabling interrupts, we should only be clearing the CAUSE_ENA bit,
not clearing the whole register. Clearing the whole register sets the
NEXTQ_IDX field to 0 instead of 0x7ff which can confuse the Firmware in
some reset sequences.
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
There exists a bug in which the driver does not correctly exit overflow
promiscuous mode. This can occur if "too many" mac filters are added,
putting the driver into overflow promiscuous mode, and the filters are
then removed. When the failed filters are removed, the driver reports
exiting overflow promiscuous mode which is correct, however traffic
continues to be received as if in promiscuous mode still.
The bug occurs because the conditional for toggling promiscuous mode was
set to only execute when promiscuous mode was enabled and not when it
was disabled as well. This patch fixes the conditional to correctly
execute when promiscuous mode is toggled and not just enabled. Without
this patch, the driver is unable to correctly exit overflow promiscuous
mode.
Signed-off-by: Alan Brady <alan.brady@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch adds support for OEM firmware version. If OEM specific
adapter is detected ethtool reports OEM product version in firmware
version string instead of etrack id.
Signed-off-by: Filip Sadowski <filip.sadowski@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Partition bandwidth control is not in just one form of MFP (multi-function
partitioning), so make the code more generic and be sure to nudge the Tx
scheduler for all MFP.
Copyright updated to 2017.
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch adds a check and message if the device is in
MFP mode as changing RSS input set is not supported in
MFP mode.
Signed-off-by: Carolyn Wyborny <carolyn.wyborny@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Changes parsing of FW 4.33 AQ command Get CEE DCBX OPER CFG (0x0A07).
Change is required because FW now creates the oper_prio_tc
nibbles reversed from those in the CEE Priority Group sub-TLV.
This change will only apply to FW 4.33 as future FW versions will use a
different function to parse the CEE data.
Signed-off-by: Greg Bowers <gregory.j.bowers@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This is a fix for the static code analysis issue where dcbcfg->numapps
could be greater than size of array (i.e dcbcfg->app[I40E_DCBX_MAX_APPS]).
The fix makes sure that the array is not accessed past the size of
of the array (i.e. I40E_DCBX_MAX_APPS).
Copyright updated to 2017.
Signed-off-by: Sudheer Mogilappagari <sudheer.mogilappagari@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The firmware expects the port number passed when setting up
the UDP tunnel configuration to be in Little Endian format.
The i40e_aq_add_udp_tunnel command byte swaps the value from
host order to Little Endian.
Since commit fe0b0cd97b ("i40e: send correct port number to
AdminQ when enabling UDP tunnels") we've correctly
sent the value in host order.
Let's also add a comment to the function explaining that it must
be in host order, as the port numbers are commonly stored as Big
Endian values.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
When searching for the vf_capability client routine, dev_info() was
used, instead of the normal dev_dbg(). This causes the message to be
displayed at standard log levels which can cause administrators to
worry. Avoid this by using dev_dbg instead.
Copyright updated to 2017.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Update a few flags related to FW interactions.
Copyright updated to 2017.
Signed-off-by: Alice Michael <alice.michael@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The variable num_active_queues represents the number of active queues we
have for the device. We assign this pretty early in i40evf_init_subtask.
Several code locations are written with loops over the tx_rings and
rx_rings structures, which don't get allocated until
i40evf_alloc_queues, and which get freed by i40evf_free_queues.
These call sites were written under the assumption that tx_rings and
rx_rings would always be allocated at least when num_active_queues is
non-zero.
Lets fix this by moving the assignment into the function where we
allocate queues. We'll use a temporary variable for storage so that we
don't assign the value in the adapter structure until after the rings
have been set up.
Finally, when we free the queues, we'll clear the value to ensure that
we do not loop over the rings memory that no longer exists.
This resolves a possible NULL pointer dereference in
i40evf_get_ethtool_stats which could occur if the VF fails to recover
from a reset, and then a user requests statistics.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch adds proper XDP_TX action support. For each Tx ring, an
additional XDP Tx ring is allocated and setup. This version does the
DMA mapping in the fast-path, which will penalize performance for
IOMMU enabled systems. Further, debugfs support is not wired up for
the XDP Tx rings.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This commit adds basic XDP support for i40e derived NICs. All XDP
actions will end up in XDP_DROP.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This series includes updates to mlx5 IPoIB netdevice driver (mlx5i),
1. We move ipoib files into separate directory, to allow it to grow
separately in its own space
2. Remove HW update carrier logic from IPoIB and VF representors profiles.
3. Add basic ethtool support. (Rings options/statistics and driver info).
4. Change MTU support.
5. Xmit path statistics reporting.
6. add PTP support.
For the new ethtool ops, PTP (ioctl) and change_mtu ndos in IPoIB, we didn't add new
implementation or new logic, we only reused those callbacks from the already existing
mlx5e (ethernet netdevice profile) and exposed them in IPoIB netdevice/ethtool ops.
Thanks,
Saeed.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQEcBAABAgAGBQJZSR3rAAoJEEg/ir3gV/o+gjQIANM2xA8GN9Aji4DSK3iZyWAK
fN0T1soixQ2oYN/gLUNsRFpU9dPvgir4X/2RjNQQaV5UR/S83LcnDcQzzuuKrZ1l
ij1Dd//+U3e2EZTenV5tzvcEOKqnhFObyg3vQR5gC/IYhtzXBFKsCkjzymmP+9Gj
9k2zCvfkGN6aP9WJe/iNCvbbEgv+roVsdCP66wNzsfWFN9BwfSpW/5mXqmShN/58
odIDOWlKoumvpmU0K2v8xm4IAsTK9gW0PUfXCRhSTORvywzlc+WcZAUd1vrNlzk7
qKhfo6OpBaPRpmAb1zBAgTPuKDU9n58tF1nxI+X1Sxm99yj9Tah1R9+Q/NnEcuI=
=qPAM
-----END PGP SIGNATURE-----
Merge tag 'mlx5-updates-2017-06-20' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
mlx5-updates-2017-06-20 (mlx5 IPoIB updates)
This series includes updates to mlx5 IPoIB netdevice driver (mlx5i),
1. We move ipoib files into separate directory, to allow it to grow
separately in its own space
2. Remove HW update carrier logic from IPoIB and VF representors profiles.
3. Add basic ethtool support. (Rings options/statistics and driver info).
4. Change MTU support.
5. Xmit path statistics reporting.
6. add PTP support.
For the new ethtool ops, PTP (ioctl) and change_mtu ndos in IPoIB, we didn't add new
implementation or new logic, we only reused those callbacks from the already existing
mlx5e (ethernet netdevice profile) and exposed them in IPoIB netdevice/ethtool ops.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
When having the skb pointer in the first descriptor, stmmac_tx_clean
can get called at a moment where the IP has only cleared the own bit
of the first descriptor, thus freeing the skb, even though there can
be several descriptors whose buffers point into the same skb.
By simply moving the skb pointer from the first descriptor to the last
descriptor, a skb will get freed only when the IP has cleared the
own bit of all the descriptors that are using that skb.
Signed-off-by: Niklas Cassel <niklas.cassel@axis.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Somehow two copies of the line 'up_write(&vf->efx->filter_sem);' got into
efx_ef10_sriov_set_vf_vlan(). This would put the mutex in a bad state and
cause all subsequent down attempts to hang.
Fixes: 671b53eec2 ("sfc: Ensure down_write(&filter_sem) and up_write() are matched before calling efx_net_open()")
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With -Wformat-truncation, gcc throws the following warning.
Fix this by increasing the size of devname to accommodate 15 character
netdev interface name and description.
Remove length format precision for %s. We can fit entire name.
Also increment the version.
drivers/net/ethernet/cisco/enic/enic_main.c: In function ‘enic_open’:
drivers/net/ethernet/cisco/enic/enic_main.c:1740:15: warning: ‘%u’ directive output may be truncated writing between 1 and 2 bytes into a region of size between 1 and 12 [-Wformat-truncation=]
"%.11s-rx-%u", netdev->name, i);
^~
drivers/net/ethernet/cisco/enic/enic_main.c:1740:5: note: directive argument in the range [0, 16]
"%.11s-rx-%u", netdev->name, i);
^~~~~~~~~~~~~
drivers/net/ethernet/cisco/enic/enic_main.c:1738:4: note: ‘snprintf’ output between 6 and 18 bytes into a destination of size 16
snprintf(enic->msix[intr].devname,
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sizeof(enic->msix[intr].devname),
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
"%.11s-rx-%u", netdev->name, i);
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Signed-off-by: Govindarajulu Varadarajan <gvaradar@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is nothing in the IP that prevents us from enabling TSO for IPv6.
Before patch:
ftp fe80::2aa:bbff:fecc:1336%eth0
ftp> get /dev/zero
882512708 bytes received in 00:14 (56.11 MiB/s)
After patch:
ftp fe80::2aa:bbff:fecc:1336%eth0
ftp> get /dev/zero
1203326784 bytes received in 00:12 (94.52 MiB/s)
Signed-off-by: Niklas Cassel <niklas.cassel@axis.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the ibmvnic driver is not in the VNIC_OPEN state, return from
ibmvnic_resume callback. If we are not in the VNIC_OPEN state, interrupts
may not be initialized and directly calling the interrupt handler will
cause a crash.
Signed-off-by: John Allen <jallen@linux.vnet.ibm.com>
Reviewed-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove the use of arch_setup_dma_ops() that was not exported
and was breaking loadable module compilation.
Signed-off-by: Madalin Bucur <madalin.bucur@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make sure dma_ops are set, to be later used by the Ethernet driver.
Signed-off-by: Madalin Bucur <madalin.bucur@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The get_rps_cpu() function will not do any RPS on the data flow when no
queue is setup and always use the current cpu where the IRQ was handled
to also handle the backlog. As we only have one physical queue we always
set this to 0 unconditionally.
Signed-off-by: John Crispin <john@phrozen.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Originally the driver only utilised the new QDMA engine. The current code
still assumes this is the case when locking the IRQ mask register. Since
RX now runs on the old style PDMA engine we can add a second lock. This
patch reduces the IRQ latency as the TX and RX path no longer need to wait
on each other under heavy load.
Signed-off-by: John Crispin <john@phrozen.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The PDMA engine used for RX allows IRQ aggregation. The patch sets up the
corresponding registers to aggregate 4 IRQs into one. Using aggregation
reduces the load on the core handling to a quarter thus reducing IRQ
latency and increasing RX performance by around 10%.
Signed-off-by: John Crispin <john@phrozen.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently PHY status changes are only printed for DSA ports. This patch
adds code to also print status changes for non-fixed links.
Signed-off-by: John Crispin <john@phrozen.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Allwinner V3s SoC has an Ethernet MAC and an internal PHY like the ones
in H3 SoC, however the MAC has no external *MII interfaces available at
GPIOs, thus only MII connection to internal PHY is supported.
Add this variant of EMAC to dwmac-sun8i driver.
The default value of the syscon EMAC-related register seems to have
changed from H3, but it seems to be a harmless change.
Signed-off-by: Icenowy Zheng <icenowy@aosc.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
The EPHY control part of the EMAC syscon register has a bit called
CLK_SEL. On the datasheet it says that if it's 0 the EPHY clock is 25MHz
and if it's 1 the clock is 24MHz.
However, according to the datasheets, no Allwinner SoC with EPHY has any
extra xtal input pins for the EPHY, and the system xtal is 24MHz.
That means the EPHY is not possible to get a 25MHz xtal input, and thus
the frequency can only be 24MHz.
It doesn't matter on H3 as the default value of H3 is 24MHz, however on
V3s the default value is wrongly set to 25MHz, which prevented the EPHY
from working properly.
Force the EPHY clock frequency to 24MHz.
Signed-off-by: Icenowy Zheng <icenowy@aosc.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Whenever firmware indicates that there's an async indication it needs
to handle, there's a switch-case where the right functionality is called
based on function's personality and information.
Before iWARP is added [as yet another client], switch over the SPQ into
a callback-registered mechanism, allowing registration of the relevant
event-processing logic based on the function's personality. This allows
us to tidy the code by removing protocol-specifics from a common file.
Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Driver needs to wait for all resources to return from FW before it can send
the FUNC_CLOSE ramrod.
Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rename the functions common to both iWARP and RoCE to have a prefix of
_rdma_ instead of _roce_.
Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Once we have iWARP support, the qede portion of the qedr<->qede would
serve all the RDMA protocols - so rename the file to be appropriate
to its function.
While we're at it, we're also moving a couple of inclusions to it into
.h files and adding includes to make sure it contains all type
definitions it requires.
Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If DCBx update occurs while QPs are open, stop sending edpms until all
QPs are closed.
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Configure device according to DCBx results so that EDPMs
made by RoCE would honor flow-control.
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
iWARP would require the chains to allocate/free their PBL memory
independently, so add the infrastructure to provide it externally.
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes the phy loopback self_test failed issue. when
Marvell Phy Module is loaded, it will powerdown fiber when doing
phy loopback self test, which cause phy loopback self_test fail.
Signed-off-by: Lin Yun Sheng <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add ioctl support to IPoIB device driver. For now, this
ioctl will support timestamp get and set.
Signed-off-by: Feras Daoud <ferasda@mellanox.com>
Signed-off-by: Eitan Rabin <rabin@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Enable PTP for IPoIB rdma_netdev and add the ability
to get the time stamping parameters using ethtool.
Signed-off-by: Feras Daoud <ferasda@mellanox.com>
Signed-off-by: Eitan Rabin <rabin@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Add the ndo that supports change mtu for IPoIB.
The callback called from the ipoib ULP driver, that gives the ability to
change the SW and HW resources accordingly in the lower driver.
Signed-off-by: Erez Shitrit <erezsh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
The mtu extra space that kept for the HW is specific for each link type,
and it is different in mlx5e and mlx5i modules.
Now it is kept in the priv structures, set by the mlx5e/mlx5i driver
accordingly.
Signed-off-by: Erez Shitrit <erezsh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Add function that sets the default values for ipoib, setting/clearing
abilities that IPoIB doesn't support, like RQ size in this case.
Signed-off-by: Erez Shitrit <erezsh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Updating the carrier involves specific HW setting, each profile should
use its own function for that.
Both IPoIB and VF representor don't need carrier update function, since
VF representor has only a logical link to VF and IPoIB manages its own
link via ib_core upper layer.
Signed-off-by: Erez Shitrit <erezsh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Port flow control supported only for ethernet ports,
therefore, prevent any call if the port type differs from
MLX5_CAP_PORT_TYPE_ETH.
Signed-off-by: Feras Daoud <ferasda@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
IPoIB netdevice driver was only introduced in previous kernel release
and it is growing in terms of features and LOC, move it to a separate
directory.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
During the module initialisation there is a possible race
(basically race between uld and lld) where neither the uld
nor lld notifies the uP about where to route the ctrl queue
completions. LLD skips notifying uP as the rdma queues were
not created by then (will leave it to ULD to notify the uP).
As the ULD comes up, it also skips notifying the uP as the
flag FULL_INIT_DONE is not set yet (ULD assumes that the
interface is not up yet).
Consequently, this race between uld and lld leaves uP
unnotified about where to send the ctrl queue completions
to, leading to iwarp RI_RES WR failure.
Here is the race:
CPU 0 CPU1
- allocates nic rx queus
- t4_sge_alloc_ctrl_txq()
(if rdma rsp queues exists,
tell uP to route ctrl queue
compl to rdma rspq)
- acquires the mutex_lock
- allocates rdma response queues
- if FULL_INIT_DONE set,
tell uP to route ctrl queue compl
to rdma rspq
- relinquishes mutex_lock
- acquires the mutex_lock
- enable_rx()
- set FULL_INIT_DONE
- relinquishes mutex_lock
This patch fixes the above issue.
Fixes: e7519f9926f1('cxgb4: avoid enabling napi twice to the same queue')
Signed-off-by: Raju Rangoju <rajur@chelsio.com>
Acked-by: Steve Wise <swise@opengridcomputing.com>
CC: Stable <stable@vger.kernel.org> # 4.9+
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
During the module initialisation there is a possible race
(basically race between uld and lld) where neither the uld
nor lld notifies the uP about where to route the ctrl queue
completions. LLD skips notifying uP as the rdma queues were
not created by then (will leave it to ULD to notify the uP).
As the ULD comes up, it also skips notifying the uP as the
flag FULL_INIT_DONE is not set yet (ULD assumes that the
interface is not up yet).
Consequently, this race between uld and lld leaves uP
unnotified about where to send the ctrl queue completions
to, leading to iwarp RI_RES WR failure.
Here is the race:
CPU 0 CPU1
- allocates nic rx queus
- t4_sge_alloc_ctrl_txq()
(if rdma rsp queues exists,
tell uP to route ctrl queue
compl to rdma rspq)
- acquires the mutex_lock
- allocates rdma response queues
- if FULL_INIT_DONE set,
tell uP to route ctrl queue compl
to rdma rspq
- relinquishes mutex_lock
- acquires the mutex_lock
- enable_rx()
- set FULL_INIT_DONE
- relinquishes mutex_lock
This patch fixes the above issue.
Fixes: e7519f9926f1('cxgb4: avoid enabling napi twice to the same queue')
Signed-off-by: Raju Rangoju <rajur@chelsio.com>
Acked-by: Steve Wise <swise@opengridcomputing.com>
CC: Stable <stable@vger.kernel.org> # 4.9+
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add general use per-vNIC mailbox area and use it for VLAN filtering
support. Initially proto is hardcoded to 802.1q.
Signed-off-by: Pablo Cascón <pablo.cascon@netronome.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Avoid NULL dereference in setup_sge_queues() when the adapter is
in non offload mode.
Fixes: 0fbc81b3ad ('chcr/cxgb4i/cxgbit/RDMA/cxgb4: Allocate resources dynamically for all cxgb4 ULD's')
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Each Octeon output ring can DMA packets to host memory in two modes: info-
pointer mode and buffer-pointer-only mode. In info-pointer mode, Octeon
takes two buffer pointers for each packet and places the length of the
packet along with specified number of bytes from the beginning of the
packet into one buffer and the rest of the packet in a separate buffer. In
buffer-pointer-only mode, Octeon takes single buffer pointer and places the
length of the packet at the beginning of the buffer followed by the packet
data.
This patch switches all Octeon output rings from info-pointer mode to
buffer-pointer-only mode. This results in fewer DMA setups and cache line
snoops.
Signed-off-by: Prasad Kanneganti <pkanneganti@cavium.com>
Signed-off-by: Derek Chickles <derek.chickles@cavium.com>
Signed-off-by: Satanand Burla <satananda.burla@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add implementation to support ethtool -K ethX rx-vlan-filter on/off.
Rename OCTNET_CMD_ENABLE_VLAN_FILTER command to OCTNET_CMD_VLAN_FILTER_CTL
and add OCTNET_CMD_VLAN_FILTER_ENABLE and OCTNET_CMD_VLAN_FILTER_DISABLE
parameters so that it can be used to enable or disable the filter.
Signed-off-by: Prasad Kanneganti <prasad.kanneganti@cavium.com>
Signed-off-by: Derek Chickles <derek.chickles@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This series provide some updates and cleanups for mlx5 core and netdevice
driver.
From Eli Cohen, add a missing event string.
From Or Gerlitz, some checkpatch cleanups.
From Moni, Disalbe HW level LAG when SRIOV is enabled.
From Tariq, A code reuse cleanup in aRFS flow.
From Itay Aveksis, Typo fix.
From Gal Pressman, ethtool statistics updates and "update stats" deferred work optimizations.
From Majd Dibbiny, Fast unload support on kernel shutdown.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQEcBAABAgAGBQJZQv0+AAoJEEg/ir3gV/o+1iQH/15I5Pr9KoCWSTN9aUglRupU
8HmJhkf7Novaro6WtIybgMGkdoNTrmHgyTEngAkRq5a5Ws/LrC/1wLH+lVMDh+Fx
/2a5cfPsK483gHWBtAbasBD8SHnsyTIeVnEhuDsevHQNkz3HGuKOgx5ZHF1sdkHU
bj/QU06LNPKAlMDI/wKod13MB4+AdTFemaJRCCgXFvu/p/EfVvB+TStdOsrxj1kx
lDIwkCykJSJsg38HoLXt7Z12nWwgHGf2De04RukKeJ6C6KTdKcUu5EYbaL9BSZZT
jiIayYjRgeXzNhY4R5yLPc0FkecNIgC90YJShUN3nR3PWa+ytaHpfJQPOS4/AW8=
=Tjmk
-----END PGP SIGNATURE-----
Merge tag 'mlx5-updates-2017-06-16' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
Mellanox mlx5 updates and cleanups 2017-06-16
mlx5-updates-2017-06-16
This series provide some updates and cleanups for mlx5 core and netdevice
driver.
From Eli Cohen, add a missing event string.
From Or Gerlitz, some checkpatch cleanups.
From Moni, Disalbe HW level LAG when SRIOV is enabled.
From Tariq, A code reuse cleanup in aRFS flow.
From Itay Aveksis, Typo fix.
From Gal Pressman, ethtool statistics updates and "update stats" deferred work optimizations.
From Majd Dibbiny, Fast unload support on kernel shutdown.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
When booting into the kdump/kexec kernel, pHyp and vios
are not prepared for the initialization crq request and
a failover transport event is generated. This is not
handled correctly.
At this point in initialization the driver is still in
the 'probing' state and cannot handle a full reset of the
driver as is normally done for a failover transport event.
To correct this we catch driver resets while still in the
'probing' state and return EAGAIN. This results in the
driver tearing down the main crq and calling ibmvnic_init()
again.
Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Cosmetic patch simplifying the smi read and write error paths. It also
align their error paths with the ones of the xsmi functions.
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds the xmdio xsmi interface support in the mvmdio driver.
This interface is used in Ethernet controllers on Marvell 370, 7k and 8k
(as of now). The xsmi interface supported by this driver complies with
the IEEE 802.3 clause 45. The xSMI interface is used by 10GbE devices.
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a check for the read and write smi operations, to ensure the
MII_ADDR_C45 bit isn't set. This will be needed as soon as the xSMI
support is added to the mvmdio driver.
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Put the two poll intervals (min and max) in the driver's ops
structure. This is needed to add the xmdio support later.
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce an ops structure to add an indirection on the is_done
function, as this is needed to add the xMDIO support later.
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The MDIO layer already provides per-bus locking, so there's no need for
MDIO bus drivers to do their own internal locking. Remove this.
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Cosmetic patch to use the GENMASK helper for masks.
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Cosmetic patch replacing spaces by tabs for defined values.
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Cosmetic fix reordering headers alphabetically.
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support to qede to report bpf_prog ID during XDP_QUERY_PROG.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Cc: Mintz Yuval <Yuval.Mintz@cavium.com>
Acked-by: Alexei Starovoitov <ast@fb.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support to nfp to report bpf_prog ID during XDP_QUERY_PROG.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Cc: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Alexei Starovoitov <ast@fb.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support to ixgbe to report bpf_prog ID during XDP_QUERY_PROG.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Cc: Alexander Duyck <alexander.h.duyck@intel.com>
Cc: John Fastabend <john.fastabend@gmail.com>
Acked-by: Alexei Starovoitov <ast@fb.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support to thunderx to report bpf_prog ID during XDP_QUERY_PROG.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Cc: Sunil Goutham <sgoutham@cavium.com>
Acked-by: Alexei Starovoitov <ast@fb.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support to bnxt to report bpf_prog ID during XDP_QUERY_PROG.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Cc: Michael Chan <michael.chan@broadcom.com>
Acked-by: Alexei Starovoitov <ast@fb.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support to mlx5e to report bpf_prog ID during XDP_QUERY_PROG.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Cc: Tariq Toukan <tariqt@mellanox.com>
Cc: Saeed Mahameed <saeedm@mellanox.com>
Acked-by: Alexei Starovoitov <ast@fb.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support to mlx4 to report bpf_prog ID during XDP_QUERY_PROG.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Cc: Tariq Toukan <tariqt@mellanox.com>
Cc: Saeed Mahameed <saeedm@mellanox.com>
Acked-by: Alexei Starovoitov <ast@fb.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
It seems like a historic accident that these return unsigned char *,
and in many places that means casts are required, more often than not.
Make these functions return void * and remove all the casts across
the tree, adding a (u8 *) cast only where the unsigned char pointer
was used directly, all done with the following spatch:
@@
expression SKB, LEN;
typedef u8;
identifier fn = { skb_push, __skb_push, skb_push_rcsum };
@@
- *(fn(SKB, LEN))
+ *(u8 *)fn(SKB, LEN)
@@
expression E, SKB, LEN;
identifier fn = { skb_push, __skb_push, skb_push_rcsum };
type T;
@@
- E = ((T *)(fn(SKB, LEN)))
+ E = fn(SKB, LEN)
@@
expression SKB, LEN;
identifier fn = { skb_push, __skb_push, skb_push_rcsum };
@@
- fn(SKB, LEN)[0]
+ *(u8 *)fn(SKB, LEN)
Note that the last part there converts from push(...)[0] to the
more idiomatic *(u8 *)push(...).
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It seems like a historic accident that these return unsigned char *,
and in many places that means casts are required, more often than not.
Make these functions (skb_put, __skb_put and pskb_put) return void *
and remove all the casts across the tree, adding a (u8 *) cast only
where the unsigned char pointer was used directly, all done with the
following spatch:
@@
expression SKB, LEN;
typedef u8;
identifier fn = { skb_put, __skb_put };
@@
- *(fn(SKB, LEN))
+ *(u8 *)fn(SKB, LEN)
@@
expression E, SKB, LEN;
identifier fn = { skb_put, __skb_put };
type T;
@@
- E = ((T *)(fn(SKB, LEN)))
+ E = fn(SKB, LEN)
which actually doesn't cover pskb_put since there are only three
users overall.
A handful of stragglers were converted manually, notably a macro in
drivers/isdn/i4l/isdn_bsdcomp.c and, oddly enough, one of the many
instances in net/bluetooth/hci_sock.c. In the former file, I also
had to fix one whitespace problem spatch introduced.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A common pattern with skb_put() is to just want to memcpy()
some data into the new space, introduce skb_put_data() for
this.
An spatch similar to the one for skb_put_zero() converts many
of the places using it:
@@
identifier p, p2;
expression len, skb, data;
type t, t2;
@@
(
-p = skb_put(skb, len);
+p = skb_put_data(skb, data, len);
|
-p = (t)skb_put(skb, len);
+p = skb_put_data(skb, data, len);
)
(
p2 = (t2)p;
-memcpy(p2, data, len);
|
-memcpy(p, data, len);
)
@@
type t, t2;
identifier p, p2;
expression skb, data;
@@
t *p;
...
(
-p = skb_put(skb, sizeof(t));
+p = skb_put_data(skb, data, sizeof(t));
|
-p = (t *)skb_put(skb, sizeof(t));
+p = skb_put_data(skb, data, sizeof(t));
)
(
p2 = (t2)p;
-memcpy(p2, data, sizeof(*p));
|
-memcpy(p, data, sizeof(*p));
)
@@
expression skb, len, data;
@@
-memcpy(skb_put(skb, len), data, len);
+skb_put_data(skb, data, len);
(again, manually post-processed to retain some comments)
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There were many places that my previous spatch didn't find,
as pointed out by yuan linyu in various patches.
The following spatch found many more and also removes the
now unnecessary casts:
@@
identifier p, p2;
expression len;
expression skb;
type t, t2;
@@
(
-p = skb_put(skb, len);
+p = skb_put_zero(skb, len);
|
-p = (t)skb_put(skb, len);
+p = skb_put_zero(skb, len);
)
... when != p
(
p2 = (t2)p;
-memset(p2, 0, len);
|
-memset(p, 0, len);
)
@@
type t, t2;
identifier p, p2;
expression skb;
@@
t *p;
...
(
-p = skb_put(skb, sizeof(t));
+p = skb_put_zero(skb, sizeof(t));
|
-p = (t *)skb_put(skb, sizeof(t));
+p = skb_put_zero(skb, sizeof(t));
)
... when != p
(
p2 = (t2)p;
-memset(p2, 0, sizeof(*p));
|
-memset(p, 0, sizeof(*p));
)
@@
expression skb, len;
@@
-memset(skb_put(skb, len), 0, len);
+skb_put_zero(skb, len);
Apply it to the tree (with one manual fixup to keep the
comment in vxlan.c, which spatch removed.)
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some code re-ordering, functionally equivalent.
- The !tx_info->inl check is evaluated anyway in both flows
(common case/end case). Run it first, this might finish
the flows earlier.
- dma_unmap calls are identical in both flows, get it out
of the if block into the common area.
Performance tests:
Tested on ConnectX3Pro, Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz
Gain is too small to be measurable, no degradation sensed.
Results are similar for IPv4 and IPv6.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Cc: kernel-team@fb.com
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Define LOG_TXBB_SIZE, log of TXBB_SIZE, and use it with a shift
operation instead of a multiplication with TXBB_SIZE.
Operations are equivalent as TXBB_SIZE is a power of two.
Performance tests:
Tested on ConnectX3Pro, Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz
Gain is too small to be measurable, no degradation sensed.
Results are similar for IPv4 and IPv6.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Cc: kernel-team@fb.com
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Increase the default TX ring size (from 512 to 1024) to match
the RX ring size.
This gives the XDP TX ring a better chance to keep up with the
rate of its RX ring in case of a high load of XDP_TX actions.
Tested:
Ethtool counter rx_xdp_tx_full used to increase, after applying this
patch it stopped.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Cc: kernel-team@fb.com
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of having their own NAPIs, XDP TX completion queues get
polled within the corresponding RX NAPI.
This prevents any possible race on TX ring prod/cons indices,
between the context that issues the transmits (RX NAPI) and the
context that handles the completions (was previously done in
a separate NAPI).
This also improves performance, as it decreases the number
of NAPIs running on a CPU, saving the overhead of syncing
and switching between the contexts.
Performance tests:
Tested on ConnectX3Pro, Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz
Single queue no-RSS optimization ON.
XDP_TX packet rate:
-------------------------------------
| Before | After | Gain |
IPv4 | 12.0 Mpps | 13.8 Mpps | 15% |
IPv6 | 12.0 Mpps | 13.8 Mpps | 15% |
-------------------------------------
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Cc: kernel-team@fb.com
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Several performance improvements in XDP TX datapath,
including:
- Ring a single doorbell for XDP TX ring per NAPI budget,
instead of doing it per a lower threshold (was 8).
This includes removing the flow of immediate doorbell ringing
in case of a full TX ring.
- Compiler branch predictor hints.
- Calculate values in compile time rather than in runtime.
Performance tests:
Tested on ConnectX3Pro, Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz
Single queue no-RSS optimization ON.
XDP_TX packet rate:
-------------------------------------
| Before | After | Gain |
IPv4 | 10.3 Mpps | 12.0 Mpps | 17% |
IPv6 | 10.3 Mpps | 12.0 Mpps | 17% |
-------------------------------------
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Cc: kernel-team@fb.com
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
