| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Before adding yet another possibly contended atomic_long_t,
it is time to add per-cpu storage for existing ones:
dev->tx_dropped, dev->rx_dropped, and dev->rx_nohandler
Because many devices do not have to increment such counters,
allocate the per-cpu storage on demand, so that dev_get_stats()
does not have to spend considerable time folding zero counters.
Note that some drivers have abused these counters which
were supposed to be only used by core networking stack.
v4: should use per_cpu_ptr() in dev_get_stats() (Jakub)
v3: added a READ_ONCE() in netdev_core_stats_alloc() (Paolo)
v2: add a missing include (reported by kernel test robot <lkp@intel.com>)
Change in netdev_core_stats_alloc() (Jakub)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: jeffreyji <jeffreyji@google.com>
Reviewed-by: Brian Vazquez <brianvv@google.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Link: https://lore.kernel.org/r/20220311051420.2608812-1-eric.dumazet@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Enable binding the nfp driver to NFP3800 and NFP3803 devices.
The PCIE_SRAM offset is different for the NFP3800 device, which also
only supports a single explicit group.
Changes to Dirk's work:
* 48-bit dma addressing is not ready yet. Keep 40-bit dma addressing
for NFP3800.
Signed-off-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
NFP3800 has slightly different queue controller range bounds.
Use the static chip data instead of defines. This commit
still assumes unchanged descriptor format. Later datapath
changes will allow adjusting for descriptor accounting.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The queue controller (QCP) is accessed based on a device specific
offset. The NFP3800 device also supports more queues.
Furthermore, the NFP3800 VFs also access the QCP differently to how the
NFP6000 VFs accesses it, though still indirectly. Fortunately, we can
remove the offset all together for both VF types. This is safe for
NFP6000 VFs since the offset was effectively a wrap around and only used
for convenience to have it set the same as the NFP6000 PF.
Use nfp_dev_info to store queue controller parameters.
Signed-off-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
In preparation for new chips instead of defines use dev_info constants
to store DMA mask length.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
NFP3800 uses a different PCIe configuration to CPP expansion BAR offsets.
We don't need to differentiate between the NFP4000, NFP5000 and NFP6000
since they all use the same offsets.
Signed-off-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
In preparation for supporting new chip add a driver data structure
which will hold per-chip-version information such as register
offsets.
Plumb it through to the relevant functions (nfpcore and nfp_net).
For now only a very simple member holding chip names is added,
following commits will add more.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Make sure the device ID tables are in ascending order.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The model number for NFP3800 and newer devices can be completely
derived from PluDevice register without subtracting 0x10.
Signed-off-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The PCI_DEVICE_ID_NETRONOME_NFP6000_VF is available for use and
should be used instead of the PCI_DEVICE_NFP6000VF. Meanwhile,
PCI_DEVICE_NFP6000VF PCI ID is removed for not being used.
Signed-off-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Multiple writes cause intermediate pointer values that do not
end on complete TX descriptors.
The QCP peripheral on the NFP provides a number of access
modes. In some access modes, the maximum amount to add must
be restricted to a 6bit value. The particular access mode
used by _nfp_qcp_ptr_add() has no such restrictions, so the
"< NFP_QCP_MAX_ADD" test is unnecessary.
Note that trying to add more that the configured ring size
in a single add will cause a QCP overflow, caught and handled
by the QCP peripheral.
Signed-off-by: Christo du Toit <christo.du.toit@netronome.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
NFP driver ABI contains a bit for ring prioritization which
was never implemented in the initially envisioned form.
Remove it, and open up the possibility of reclaiming for other uses.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: Fei Qin <fei.qin@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The print function dev_err() is redundant because platform_get_irq()
already prints an error.
Eliminate the follow coccicheck warning:
./drivers/net/ethernet/8390/mcf8390.c:414:2-9: line 414 is redundant
because platform_get_irq() already prints an error
Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Signed-off-by: Yang Li <yang.lee@linux.alibaba.com>
Link: https://lore.kernel.org/r/20220311001756.12234-1-yang.lee@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
netdev_name_node_alt_create() and netdev_name_node_alt_destroy()
are only called by rtnetlink, so no need for exports.
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20220310223952.558779-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
In function netvsc_process_raw_pkt for VM_PKT_DATA_USING_XFER_PAGES
case there is already a 'return' statement which results 'break'
as dead code
Signed-off-by: Saurabh Sengar <ssengar@linux.microsoft.com>
Reviewed-by: Haiyang Zhang <haiyangz@microsoft.com>
Link: https://lore.kernel.org/r/1646933534-29493-1-git-send-email-ssengar@linux.microsoft.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Both functions are only used by core networking code.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220310143229.895319-1-hch@lst.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
In review for commit 8ee7ec4890e2b ("net: ipa: embed interconnect
array in the power structure"), Jakub Kicinski suggested that a
follow-up patch use struct_size() when computing the size of the
IPA power structure, which ends with a flexible array member.
Do that.
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Alex Elder <elder@linaro.org>
Link: https://lore.kernel.org/r/20220311162423.872645-1-elder@linaro.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Add sysfs nodes for the frequency generator and signal counters.
Update SMA selector lists for these, and also add the new
'None', 'VCC' 'GND' selectors.
Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The timecard now has 4 general purpose timestampers.
Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Input signals can be steered to any of the frequency counters.
The counter measures the frequency over a number of seconds:
echo 0 > freq1/seconds = turns off measurement
echo 1 > freq1/seconds = sets period & turns on measurment.
Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The signal generators can be programmed either via the sysfs
file or through a PTP_CLK_REQ_PEROUT ioctl request.
Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Newer firmware provides 4 programmable signal generators, add
support for those here. The signal generators provide the
ability to set the period, duty cycle, phase offset, and polarity,
with new values defaulting to prior values.
The period and phase offset are specified in nanoseconds.
E.g: period [duty [phase [polarity]]]
echo 500000000 > signal # 1/2 second period
echo 1000000 40 100 > signal # 1ms period, 40% on, offset 100ns
echo 0 > signal # turn off generator
Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add the ability to group sysfs nodes behind a firmware feature
check. This way non-present sysfs attributes are omitted on
older firmware, which does not have newer features.
This will be used in the upcoming patches which adds more
features to the timecard.
Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
These will provide constant outputs.
Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
As there are may be 2 GNSS outputs, rename the first one for clarity.
This also works around a parsing issue when specifying selectors.
Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This adds support for the "IN: None" selector, which disables
the input on a sma pin. This should be compatible with old firmware
(the firmware will ignore it if not supported).
Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Assuming the firmware allows it, the direction of each SMA connector
is no longer fixed. Handle remapping directions for each pin.
Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When doing manual injection of the frame, it is required to check if the
TX FIFO is ready to accept the next word of the frame. For this we are
using 'readx_poll_timeout_atomic', the only problem is that before it
actually checks the status, is determining the time when to finish polling
the status. Which seems to be an expensive operation.
Therefore check the status of the TX FIFO before calling
'readx_poll_timeout_atomic'.
Doing this will improve the TX bitrate by ~70%. Because 99% the FIFO is
ready by that time. The measurements were done using iperf3.
Before:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.03 sec 55.2 MBytes 46.2 Mbits/sec 0 sender
[ 5] 0.00-10.04 sec 53.8 MBytes 45.0 Mbits/sec receiver
After:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.10 sec 95.0 MBytes 78.9 Mbits/sec 0 sender
[ 5] 0.00-10.11 sec 95.0 MBytes 78.8 Mbits/sec receiver
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Remove dev_err() messages after platform_get_irq*() failures.
platform_get_irq() already prints an error.
Generated by: scripts/coccinelle/api/platform_get_irq.cocci
Signed-off-by: Yihao Han <hanyihao@vivo.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Commit bf08824a0f47 ("flow_dissector: Add support for HSR") added support for
HSR within the flow dissector. However, it only works for HSR in version
1. Version 0 uses a different Ether Type. Add support for it.
Reported-by: Anthony Harivel <anthony.harivel@linutronix.de>
Signed-off-by: Kurt Kanzenbach <kurt@linutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
It is not recommened to use platform_get_resource(pdev, IORESOURCE_IRQ)
for requesting IRQ's resources any more, as they can be not ready yet in
case of DT-booting.
platform_get_irq() instead is a recommended way for getting IRQ even if
it was not retrieved earlier.
It also makes code simpler because we're getting "int" value right away
and no conversion from resource to int is required.
Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Minghao Chi <chi.minghao@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
platform_get_resource(pdev, IORESOURCE_IRQ, ..) relies on static
allocation of IRQ resources in DT core code, this causes an issue
when using hierarchical interrupt domains using "interrupts" property
in the node as this bypasses the hierarchical setup and messes up the
irq chaining.
In preparation for removal of static setup of IRQ resource from DT core
code use platform_get_irq() for DT users only.
While at it propagate error code in emac_dev_stop() in case
platform_get_irq_optional() fails.
Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Convert am65-cpsw driver and am65-cpsw ethtool to use Phylink APIs
as described at Documentation/networking/sfp-phylink.rst. All calls
to Phy APIs are replaced with their equivalent Phylink APIs.
No functional change intended. Use Phylink instead of conventional
Phylib, in preparation to add support for SGMII/QSGMII modes.
Signed-off-by: Siddharth Vadapalli <s-vadapalli@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Currently in case of target hardware restart, we just reconfig and
re-enable the security keys and enable the network queues to start
data traffic back from where it was interrupted.
Many ath10k wifi chipsets have sequence numbers for the data
packets assigned by firmware and the mac sequence number will
restart from zero after target hardware restart leading to mismatch
in the sequence number expected by the remote peer vs the sequence
number of the frame sent by the target firmware.
This mismatch in sequence number will cause out-of-order packets
on the remote peer and all the frames sent by the device are dropped
until we reach the sequence number which was sent before we restarted
the target hardware
In order to fix this, we trigger a sta disconnect, in case of target
hw restart. After this there will be a fresh connection and thereby
avoiding the dropping of frames by remote peer.
The right fix would be to pull the entire data path into the host
which is not feasible or would need lots of complex changes and
will still be inefficient.
Tested on ath10k using WCN3990, QCA6174
Signed-off-by: Youghandhar Chintala <youghand@codeaurora.org>
Link: https://lore.kernel.org/r/20220308115325.5246-2-youghand@codeaurora.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Today's implementation of csum_shift() leads to branching based on
parity of 'offset'
000002f8 <csum_block_add>:
2f8: 70 a5 00 01 andi. r5,r5,1
2fc: 41 a2 00 08 beq 304 <csum_block_add+0xc>
300: 54 84 c0 3e rotlwi r4,r4,24
304: 7c 63 20 14 addc r3,r3,r4
308: 7c 63 01 94 addze r3,r3
30c: 4e 80 00 20 blr
Use first bit of 'offset' directly as input of the rotation instead of
branching.
000002f8 <csum_block_add>:
2f8: 54 a5 1f 38 rlwinm r5,r5,3,28,28
2fc: 20 a5 00 20 subfic r5,r5,32
300: 5c 84 28 3e rotlw r4,r4,r5
304: 7c 63 20 14 addc r3,r3,r4
308: 7c 63 01 94 addze r3,r3
30c: 4e 80 00 20 blr
And change to left shift instead of right shift to skip one more
instruction. This has no impact on the final sum.
000002f8 <csum_block_add>:
2f8: 54 a5 1f 38 rlwinm r5,r5,3,28,28
2fc: 5c 84 28 3e rotlw r4,r4,r5
300: 7c 63 20 14 addc r3,r3,r4
304: 7c 63 01 94 addze r3,r3
308: 4e 80 00 20 blr
Seems like only powerpc benefits from a branchless implementation.
Other main architectures like ARM or X86 get better code with
the generic implementation and its branch.
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
While commit 6a01afcf8468 ("mac80211: mesh: Free ie data when leaving
mesh") fixed a memory leak on mesh leave / teardown it introduced a
potential memory corruption caused by a double free when rejoining the
mesh:
ieee80211_leave_mesh()
-> kfree(sdata->u.mesh.ie);
...
ieee80211_join_mesh()
-> copy_mesh_setup()
-> old_ie = ifmsh->ie;
-> kfree(old_ie);
This double free / kernel panics can be reproduced by using wpa_supplicant
with an encrypted mesh (if set up without encryption via "iw" then
ifmsh->ie is always NULL, which avoids this issue). And then calling:
$ iw dev mesh0 mesh leave
$ iw dev mesh0 mesh join my-mesh
Note that typically these commands are not used / working when using
wpa_supplicant. And it seems that wpa_supplicant or wpa_cli are going
through a NETDEV_DOWN/NETDEV_UP cycle between a mesh leave and mesh join
where the NETDEV_UP resets the mesh.ie to NULL via a memcpy of
default_mesh_setup in cfg80211_netdev_notifier_call, which then avoids
the memory corruption, too.
The issue was first observed in an application which was not using
wpa_supplicant but "Senf" instead, which implements its own calls to
nl80211.
Fixing the issue by removing the kfree()'ing of the mesh IE in the mesh
join function and leaving it solely up to the mesh leave to free the
mesh IE.
Cc: stable@vger.kernel.org
Fixes: 6a01afcf8468 ("mac80211: mesh: Free ie data when leaving mesh")
Reported-by: Matthias Kretschmer <mathias.kretschmer@fit.fraunhofer.de>
Signed-off-by: Linus Lüssing <ll@simonwunderlich.de>
Tested-by: Mathias Kretschmer <mathias.kretschmer@fit.fraunhofer.de>
Link: https://lore.kernel.org/r/20220310183513.28589-1-linus.luessing@c0d3.blue
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
There are no legacy rates on 60GHz or sub-1GHz band, so modify the check.
Signed-off-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: MeiChia Chiu <MeiChia.Chiu@mediatek.com>
Link: https://lore.kernel.org/r/20220308021645.16272-1-MeiChia.Chiu@mediatek.com
[Ghz -> GHz]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
It should be NL80211_IFTYPE_OCB instead.
Signed-off-by: Veerendranath Jakkam <quic_vjakkam@quicinc.com>
Link: https://lore.kernel.org/r/1645542399-4680-1-git-send-email-quic_vjakkam@quicinc.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Previous memory allocations in this function already use GFP_KERNEL, so
use __dev_alloc_skb() and an explicit GFP_KERNEL instead of an implicit
GFP_ATOMIC.
This gives more opportunities of successful allocation.
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Link: https://lore.kernel.org/r/194a0e2ff00c3fae88cc9fba47431747360c8242.1645345378.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Fix the following coccicheck warning:
./drivers/net/wireless/mac80211_hwsim.c:1040:0-23: WARNING:
hwsim_fops_rx_rssi should be defined with
DEFINE_DEBUGFS_ATTRIBUTE
Signed-off-by: Yihao Han <hanyihao@vivo.com>
Link: https://lore.kernel.org/r/20220218070228.6210-1-hanyihao@vivo.com
[fix indentation]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Unlike the legacy EEPROM callbacks, when using the netlink EEPROM query
(get_module_eeprom_by_page) the driver should not try to validate the
query parameters, but just perform the read requested by the userspace.
Recent discussion in the mailing list:
https://lore.kernel.org/netdev/20220120093051.70845141@kicinski-fedora-PC1C0HJN.hsd1.ca.comcast.net/
Signed-off-by: Gal Pressman <gal@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
The assumption that the first byte in the module mapping dword is the
module number shouldn't be hard-coded in the driver, but come from
mlx5_ifc structs.
While at it, fix the incorrect width for the 'rx_lane' and 'tx_lane'
fields.
Signed-off-by: Gal Pressman <gal@nvidia.com>
Reviewed-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
The MCIA register supports either 12 or 32 dwords, use the correct value
by querying the capability from the MCAM register.
Signed-off-by: Gal Pressman <gal@nvidia.com>
Reviewed-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
SMFS dr matchers are processed sequentially in hardware according to
their priorities, and not skipped if empty.
Currently, smfs ct fs creates four predefined dr matchers per ct
table (ct/ct nat) with hardcoded priority. Compared to dmfs ct fs
using autogroups, this might cause additional hops in fastpath for
traffic patterns that match later priorties, even if previous
priorites are empty, e.g user only using ipv6 UDP traffic will
have additional 3 hops.
Create the matchers dynamically, using the highest priority available,
on first rule usage, and remove them on last usage.
Signed-off-by: Paul Blakey <paulb@nvidia.com>
Reviewed-by: Oz Shlomo <ozsh@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
fs_core layer adds extra book keeping that is either unneeded for CT, or
unused by the underlying software steering, such as allocating FTEs and
FTE ids, saving the match key and mask, and autogroups management.
On top of that, direct steering has a translation layer (fs_dr) from PRM
commands to direct steering objects, for example, creating temporary
dr_action objects. This has a performance impact when dealing
with CT high insertion rate.
To use direct steering (smfs) directly for ct, add a tc ct fs smfs
implementation. Instead of dmfs autogroups, smfs ct fs uses one of 4
predefined dr matchers in CT and CT-NAT tables, for each combination
of tuple ethertype (ipv4/ipv6), and tuple ip_proto (udp/tcp) that
is currently used by nf flow table flow offload.
At rule insertions, validate the flow rule fits one of the predfined
matcher, and insert to it.
To fill the dr_actions of the rule efficiently, create the fwd to post_ct
tbl dr_action at fs init, the count dr_action at counter creation,
and re-use the already pre-allocated modify header dr_action.
Signed-off-by: Paul Blakey <paulb@nvidia.com>
Reviewed-by: Oz Shlomo <ozsh@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Add a thin layer that exports selected direct steering (dr) API
which will be used by a ct fs implementation in a following
patch.
Signed-off-by: Paul Blakey <paulb@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
If sw steering was used to create the table, dr steeering fs creates
a backing dr table for the mlx5 flow table.
Add helper to return this table so it can be used to create matchers and
add rules on it directly instead of passing via eswitch_offloads/fs_core
insertion.
Signed-off-by: Paul Blakey <paulb@nvidia.com>
Reviewed-by: Oz Shlomo <ozsh@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Currently, fs_core layer provides flow steering services to the driver
including: autogroups, allocating FTEs (flow table entries) and FTE ids,
and support of fte action modification. If then software steering is
configured, rule insertion will go through a translation layer from
firmware buffers to software steering objects (see fs_dr.c).
The connection tracking table is a system table that is not directly
controlled by the user and is a very high scale table. These fs_core
services introduces an overhead that may be optimized by using software
steering API directly.
Introduce ct flow steering interface to allow multiple flow steering
providers. Use the new interface to implement the current dmfs (device
managed flow steering) provider which uses fs_core insertion.
Signed-off-by: Paul Blakey <paulb@nvidia.com>
Reviewed-by: Oz Shlomo <ozsh@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
The function is node-aware and gets the node as an argument.
Use a node-aware allocation for the doorbell pgdir structure.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Prefer the aware allocation, use the device NUMA node.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
Eric Dumazet
Dirk van der Merwe
Jakub Kicinski
Christo du Toit
Yang Li
Jakub Kicinski
Saurabh Sengar
Christoph Hellwig
Alex Elder
Jonathan Lemon
Horatiu Vultur
Yihao Han
Kurt Kanzenbach
Minghao Chi
Lad Prabhakar
Siddharth Vadapalli
Youghandhar Chintala
Christophe Leroy
Linus Lüssing
MeiChia Chiu
Veerendranath Jakkam
Christophe JAILLET
Gal Pressman
Paul Blakey
Tariq Toukan