The driver implemented SIOCSHWTSTAMP ioctl command only, but it stores
configuration in the private data, so it is possible to report it back
to users. Implement both ndo_hwtstamp_set and ndo_hwtstamp_get
callbacks. To properly report RX filter type, store it in hwts_rx_en
instead of using this field as a simple flag. The logic didn't change
because receive path used this field as boolean flag.
Signed-off-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>
Reviewed-by: Kory Maincent <kory.maincent@bootlin.com>
Link: https://patch.msgid.link/20251103150952.3538205-7-vadim.fedorenko@linux.dev
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The driver implemented SIOCSHWTSTAMP ioctl command only, but it also
stores configuration in private data, so it's possible to report it back
to users. Implement both ndo_hwtstamp_set and ndo_hwtstamp_get
callbacks.
Reviewed-by: Kory Maincent <kory.maincent@bootlin.com>
Signed-off-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>
Link: https://patch.msgid.link/20251103150952.3538205-6-vadim.fedorenko@linux.dev
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The driver implemented SIOCSHWTSTAMP ioctl command only. But it stores
timestamping configuration, so it is possible to report it to users.
Implement both ndo_hwtstamp_set and ndo_hwtstamp_get callbacks. After
this the ndo_eth_ioctl effectively becomes phy_do_ioctl - adjust
callback accordingly.
Reviewed-by: Kory Maincent <kory.maincent@bootlin.com>
Signed-off-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>
Link: https://patch.msgid.link/20251103150952.3538205-5-vadim.fedorenko@linux.dev
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The driver implemented SIOCSHWTSTAMP ioctl command only, but there is a
way to get configuration back. Implement both ndo_hwtstamp_set and
ndo_hwtstamp_set callbacks.
Reviewed-by: Kory Maincent <kory.maincent@bootlin.com>
Signed-off-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>
Link: https://patch.msgid.link/20251103150952.3538205-4-vadim.fedorenko@linux.dev
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The driver implemented SIOCSHWTSTAMP ioctl command only, but there is a
way to get configured status. Implement both ndo_hwtstamp_set and
ndo_hwtstamp_get callbacks.
Reviewed-by: Kory Maincent <kory.maincent@bootlin.com>
Signed-off-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>
Link: https://patch.msgid.link/20251103150952.3538205-3-vadim.fedorenko@linux.dev
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Commit bd61848900 ("net: devmem: Implement TX path") declared this
but never implemented it.
Signed-off-by: Yue Haibing <yuehaibing@huawei.com>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Reviewed-by: Mina Almasry <almasrymina@google.com>
Link: https://patch.msgid.link/20251103072046.1670574-1-yuehaibing@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Matthieu Baerts says:
====================
mptcp: pm: in-kernel: fullmesh endp nb + bind cases
Here is a small optimisation for the in-kernel PM, joined by a small
behavioural change to avoid confusions, and followed by a few more
tests.
- Patch 1: record fullmesh endpoints numbers, not to iterate over all
endpoints to check if one is marked as fullmesh.
- Patch 2: when at least one endpoint is marked as fullmesh, only use
these endpoints when reacting to an ADD_ADDR, even if there are no
endpoints for this IP family: this is less confusing.
- Patch 3: reduce duplicated code to prepare the next patch.
- Patch 4: extra "bind" cases: the listen socket restrict the bind to
one IP address, not allowing MP_JOIN to extra IP addresses, except if
another listening socket accepts them.
====================
Link: https://patch.msgid.link/20251101-net-next-mptcp-fm-endp-nb-bind-v1-0-b4166772d6bb@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
By design, an MPTCP connection will not accept extra subflows where no
MPTCP listening sockets can accept such requests.
In other words, it means that if the 'server' listens on a specific
address / device, it cannot accept MP_JOIN sent to a different address /
device. Except if there is another MPTCP listening socket accepting
them.
This is what the new tests are validating:
- Forcing a bind on the main v4/v6 address, and checking that MP_JOIN
to announced addresses are not accepted.
- Also forcing a bind on the main v4/v6 address, but before, another
listening socket is created to accept additional subflows. Note that
'mptcpize run nc -l' -- or something else only doing: socket(MPTCP),
bind(<IP>), listen(0) -- would be enough, but here mptcp_connect is
reused not to depend on another tool just for that.
- Same as the previous one, but using v6 link-local addresses: this is
a bit particular because it is required to specify the outgoing
network interface when connecting to a link-local address announced
by the other peer. When using the routing rules, this doesn't work
(the outgoing interface is not known) ; but it does work with a
'laminar' endpoint having a specified interface.
Note that extra small modifications are needed for these tests to work:
- mptcp_connect's check_getpeername_connect() check should strip the
specified interface when comparing addresses.
- With IPv6 link-local addresses, it is required to wait for them to
be ready (no longer in 'tentative' mode) before using them, otherwise
the bind() will not be allowed.
Link: https://github.com/multipath-tcp/mptcp_net-next/issues/591
Reviewed-by: Geliang Tang <geliang@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20251101-net-next-mptcp-fm-endp-nb-bind-v1-4-b4166772d6bb@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The same extra long commands are present twice, with small differences:
the variable for the stdin file is different.
Use new dedicated variables in one command to avoid this code
duplication.
Reviewed-by: Geliang Tang <geliang@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20251101-net-next-mptcp-fm-endp-nb-bind-v1-3-b4166772d6bb@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Our documentation is saying that the in-kernel PM is only using fullmesh
endpoints to establish subflows to announced addresses when at least one
endpoint has a fullmesh flag. But this was not totally correct: only
fullmesh endpoints were used if at least one endpoint *from the same
address family as the received ADD_ADDR* has the fullmesh flag.
This is confusing, and it seems clearer not to have differences
depending on the address family.
So, now, when at least one MPTCP endpoint has a fullmesh flag, the local
addresses are picked from all fullmesh endpoints, which might be 0 if
there are no endpoints for the correct address family.
One selftest needs to be adapted for this behaviour change.
Reviewed-by: Geliang Tang <geliang@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20251101-net-next-mptcp-fm-endp-nb-bind-v1-2-b4166772d6bb@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Instead of iterating over all endpoints, under RCU read lock, just to
check if one of them as the fullmesh flag, we can keep a counter of
fullmesh endpoint, similar to what is done with the other flags.
This counter is now checked, before iterating over all endpoints.
Similar to the other counters, this new one is also exposed. A userspace
app can then know when it is being used in a fullmesh mode, with
potentially (too) many subflows.
Reviewed-by: Geliang Tang <geliang@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20251101-net-next-mptcp-fm-endp-nb-bind-v1-1-b4166772d6bb@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Tariq Toukan says:
====================
net/mlx5e: Reduce interface downtime on configuration change
This series significantly reduces the interface downtime while swapping
channels during a configuration change, on capable devices.
Here we remove an old requirement on operations ordering that became
obsolete on recent capable devices. This helps cutting the downtime by a
factor of magnitude, ~80% in our example.
Perf numbers:
Measured the number of dropped packets in a simple ping flood test,
during a configuration change operation, that switches the number of
channels from 247 to 248.
Before: 71 packets lost
After: 15 packets lost, ~80% saving.
====================
Link: https://patch.msgid.link/1761831159-1013140-1-git-send-email-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Cap bit tis_tir_td_order=1 indicates that an old firmware requirement /
limitation no longer exists. When unset, the latency of several firmware
commands significantly increases with the presence of high number of
co-existing channels (both old and new sets). Hence, we used to close
unneeded old channels before invoking those firmware commands.
Today, on capable devices, this is no longer the case. Minimize the
interface down time by deferring the old channels closure, after the
activation of the new ones.
Perf numbers:
Measured the number of dropped packets in a simple ping flood test,
during a configuration change operation, that switches the number of
channels from 247 to 248.
Before: 71 packets lost
After: 15 packets lost, ~80% saving.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Carolina Jubran <cjubran@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/1761831159-1013140-8-git-send-email-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Let the caller function mlx5e_safe_switch_params() maintain a copy
of the old channels, and pass it to mlx5e_switch_priv_channels().
This is in preparation for the next patch.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Carolina Jubran <cjubran@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/1761831159-1013140-7-git-send-email-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
On old firmware, (tis_tir_td_order=0), TIR of a transport domain should
either be created after all SQs of the same domain, or TIR.self_lb_en
should be reapplied using MODIFY_TIR, for self loopback filtering to
function correctly.
This is not necessary anymnore on new FW (tis_tir_td_order=1), thus
there's no need for calling modify_tir operations after creating a new
set of SQs to maintain the self loopback prevention functional.
Skip these operations.
This saves O(max_num_channels) MODIFY_TIR firmware commands in
operations like interface up or channels configuration change.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Carolina Jubran <cjubran@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/1761831159-1013140-6-git-send-email-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In IPoIB, the self loopback prevention configuration apply in activation
stage has two roles: fulfill a firmware requirement for old firmware
(tis_tir_td_order=0), and update the proper configuration as it was not
set in init.
Here we set the proper configuration in init, to allow skipping the
modify_tirs commands on new firmware in a downstream patch.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Carolina Jubran <cjubran@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/1761831159-1013140-5-git-send-email-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Until now, IPoIB was creating TIRs without setting self loopback
prevention, then modifying them in activation stage.
This is a preparation patch, that will be used by IPoIB to init TIRs
properly without the need for following calls of modify_tir.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Carolina Jubran <cjubran@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/1761831159-1013140-4-git-send-email-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Extend the TIR API and use it in mlx5e_modify_tirs_lb() instead of the
explicit modify_tir code.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Carolina Jubran <cjubran@nvidia.com>
Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/1761831159-1013140-3-git-send-email-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The re-application of self loopback prevention attributes in TIRs is
necessary in old firmwares (where tis_tir_td_order cap is cleared) after
recreation of SQs.
However, this is not needed in new firmware with tis_tir_td_order=1.
As a preparation patch, enhance the function structures to differentiate
between an explicit loopback prevention configuration apply, and the
re-apply operation required by old firmware.
Loopback selftests should now call mlx5e_modify_tirs_lb() directly, as
their use case is not related to the firmware limitation.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Carolina Jubran <cjubran@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/1761831159-1013140-2-git-send-email-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The original comments contained spelling errors and incomplete logical
descriptions, which could easily lead to misunderstandings of the code
logic. The specific modifications are as follows:
Correct the spelling error by changing "inut max" to "but not exceed the
maximum limit";
Add the note "If the user has not specified a value, the default maximum
limit is 8" to clarify the default value logic;
Improve the coherence of the statement to make the queue quantity rules
clearer.
After the modification, the comments can accurately reflect the code
behavior of "taking the smaller value between the number of CPUs and the
default maximum limit of 8 for the number of queues", enhancing code
maintainability.
Signed-off-by: Chu Guangqing <chuguangqing@inspur.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://patch.msgid.link/20251103032212.2462-1-chuguangqing@inspur.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Fix the spelling error of "separate".
Signed-off-by: Chu Guangqing <chuguangqing@inspur.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Link: https://patch.msgid.link/20251103074305.4727-1-chuguangqing@inspur.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Russell King says:
====================
net: stmmac: multi-interface stmmac
This series adds a callback for platform glue to configure the stmmac
core interface mode depending on the PHY interface mode that is being
used. This is currently only called just before the dwmac core is reset
since these signals are latched on reset.
Included in this series are changes to s32 to move its PHY_INTF_SEL_x
definitions out of the way of the dwmac core's signals which has more
entitlement to use this name. We convert dwmac-imx as an example.
Including other platform glue would make this series excessively large,
but once this core code is merged, the individual platform glue updates
can be posted one after another as they will be independent of each
other.
It is hoped that this callback can be used in future to reconfigure the
dwmac core when the interface mode changes to support PHYs that change
their interface mode, but we're nowhere near being able to do that yet.
====================
Link: https://patch.msgid.link/aQiWzyrXU_2hGJ4j@shell.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Rather than placing the phy_intf_sel() setup in the ->init() method,
move it to the new ->set_phy_intf_sel() method.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1vFt5C-0000000ChpR-2kAB@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pass the imx_priv_data instead of the plat_stmmacenet_data into the
set_intf_mode() SoC specific methods.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1vFt57-0000000ChpL-25kS@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Simplify the set_intf_mode() implementations, testing the phy_intf_sel
value rather than the PHY interface mode.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1vFt52-0000000ChpG-1bsd@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
i.MX implementations other than IMX8DXL involve setting the dwmac core
phy_intf_sel input. Use stmmac_get_phy_intf_sel() to decode the PHY
interface mode to the phy_intf_sel value, validating the result, and
passing it into the implementation specific .set_intf_mode() method
rather than each .set_intf_mode() method doing this.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1vFt4x-0000000ChpA-1Edr@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Use FIELD_PREP()/FIELD_GET() in the functions to construct the PHY
interface selection bitfield or to extract its value.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1vFt4s-0000000Chp4-0kwf@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Convert dwmac-imx to use the PHY_INTF_SEL_xxx definitions rather than
constants via:
- ensuring that the prefix for the MASK and value definitions is the
same.
- using FIELD_PREP() to shift the PHY_INTF_SEL_xxx definition to the
appropriate bitfield.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1vFt4n-0000000Choy-0IeG@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When dwmac is synthesised with support for multiple PHY interfaces, the
core provides phy_intf_sel inputs, sampled on reset, to configure the
PHY facing interface. Use stmmac_get_phy_intf_sel() in core code to
determine the dwmac phy_intf_sel input value, and provide a new
platform method called with this value just before we issue a soft
reset to the dwmac core.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1vFt4h-0000000Chos-3wxX@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Provide a function to translate the PHY interface mode to the
phy_intf_sel pin configuration for dwmac1000 and dwmac4 cores that
support multiple interfaces. We currently handle MII, GMII, RGMII,
SGMII, RMII and REVMII, but not TBI, RTBI nor SMII as drivers do not
appear to use these three and the driver doesn't currently support
these.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1vFt4c-0000000Choe-3SII@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add definitions for the active PHY interface found in DMA hardware
feature register 0, and also used to configure the core in multi-
interface designs via phy_intf_sel.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Link: https://patch.msgid.link/E1vFt4X-0000000ChoY-30p9@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
S32's PHY_INTF_SEL_x definitions conflict with those for the dwmac
cores as they use a different bitmapping. Add a S32 prefix so that
they are unique.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Reviewed-by: Jan Petrous (OSS) <jan.petrous@oss.nxp.com>
Link: https://patch.msgid.link/E1vFt4S-0000000ChoS-2Ahi@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
imx_dwmac_set_clk_tx_rate() is passed the interface mode from phylink
which will be the same as plat_dat->phy_interface. Use the passed-in
interface mode rather than plat_dat->phy_interface.
Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Tested-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1vFt4N-0000000ChoM-1llp@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
deliver_skb() should not be inlined as is it not called
in the fast path.
Add unlikely() clauses giving hints to the compiler about this fact.
Before this patch:
size net/core/dev.o
text data bss dec hex filename
121794 13330 176 135300 21084 net/core/dev.o
__netif_receive_skb_core() size on x86_64 : 4080 bytes.
After:
size net/core/dev.o
text data bss dec hex filenamee
120330 13338 176 133844 20ad4 net/core/dev.o
__netif_receive_skb_core() size on x86_64 : 2781 bytes.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20251103165256.1712169-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Gathering interface statistics can be a relatively expensive operation
on certain systems as it requires iterating over all the cpus.
RTEXT_FILTER_SKIP_STATS was first introduced [1] to skip AF_INET6
statistics from interface dumps and it was then extended [2] to
also exclude IFLA_VF_INFO.
The semantics of the flag does not seem to be limited to AF_INET
or VF statistics and having a way to query the interface status
(e.g: carrier, address) without retrieving its statistics seems
reasonable. So this patch extends the use RTEXT_FILTER_SKIP_STATS
to also affect IFLA_STATS.
[1] https://lore.kernel.org/all/20150911204848.GC9687@oracle.com/
[2] https://lore.kernel.org/all/20230611105108.122586-1-gal@nvidia.com/
Signed-off-by: Adrian Moreno <amorenoz@redhat.com>
Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Link: https://patch.msgid.link/20251103154006.1189707-1-amorenoz@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Jason Xing says:
====================
xsk: minor optimizations around locks
Two optimizations regarding xsk_tx_list_lock and cq_lock can yield a
performance increase because of avoiding disabling and enabling
interrupts frequently.
====================
Link: https://patch.msgid.link/20251030000646.18859-1-kerneljasonxing@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
- Split cq_lock into two smaller locks: cq_prod_lock and
cq_cached_prod_lock
- Avoid disabling/enabling interrupts in the hot xmit path
In either xsk_cq_cancel_locked() or xsk_cq_reserve_locked() function,
the race condition is only between multiple xsks sharing the same
pool. They are all in the process context rather than interrupt context,
so now the small lock named cq_cached_prod_lock can be used without
handling interrupts.
While cq_cached_prod_lock ensures the exclusive modification of
@cached_prod, cq_prod_lock in xsk_cq_submit_addr_locked() only cares
about @producer and corresponding @desc. Both of them don't necessarily
be consistent with @cached_prod protected by cq_cached_prod_lock.
That's the reason why the previous big lock can be split into two
smaller ones. Please note that SPSC rule is all about the global state
of producer and consumer that can affect both layers instead of local
or cached ones.
Frequently disabling and enabling interrupt are very time consuming
in some cases, especially in a per-descriptor granularity, which now
can be avoided after this optimization, even when the pool is shared by
multiple xsks.
With this patch, the performance number[1] could go from 1,872,565 pps
to 1,961,009 pps. It's a minor rise of around 5%.
[1]: taskset -c 1 ./xdpsock -i enp2s0f1 -q 0 -t -S -s 64
Signed-off-by: Jason Xing <kernelxing@tencent.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://patch.msgid.link/20251030000646.18859-3-kerneljasonxing@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The commit ac98d8aab6 ("xsk: wire upp Tx zero-copy functions")
originally introducing this lock put the deletion process in the
sk_destruct which can run in irq context obviously, so the
xxx_irqsave()/xxx_irqrestore() pair was used. But later another
commit 541d7fdd76 ("xsk: proper AF_XDP socket teardown ordering")
moved the deletion into xsk_release() that only happens in process
context. It means that since this commit, it doesn't necessarily
need that pair.
Now, there are two places that use this xsk_tx_list_lock and only
run in the process context. So avoid manipulating the irq then.
Signed-off-by: Jason Xing <kernelxing@tencent.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://patch.msgid.link/20251030000646.18859-2-kerneljasonxing@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
In a multi-network card or container environment, this is needed in order
to differentiate between trace events relating to net devices that exist
in different network namespaces and share the same name.
for xmit_timeout trace events:
[002] ..s1. 1838.311662: net_dev_xmit_timeout: dev=eth0 driver=virtio_net queue=10 net_cookie=3
[007] ..s1. 1839.335650: net_dev_xmit_timeout: dev=eth0 driver=virtio_net queue=10 net_cookie=4100
[007] ..s1. 1844.455659: net_dev_xmit_timeout: dev=eth0 driver=virtio_net queue=10 net_cookie=3
[002] ..s1. 1850.087647: net_dev_xmit_timeout: dev=eth0 driver=virtio_net queue=10 net_cookie=3
Cc: Eran Ben Elisha <eranbe@mellanox.com>
Cc: Jiri Pirko <jiri@mellanox.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Simon Horman <horms@kernel.org>
Cc: Paolo Abeni <pabeni@redhat.com>
Suggested-by: Ido Schimmel <idosch@idosch.org>
Signed-off-by: Tonghao Zhang <tonghao@bamaicloud.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://patch.msgid.link/20251028043244.82288-1-tonghao@bamaicloud.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Oleksij Rempel says:
====================
ethtool: introduce PHY MSE diagnostics UAPI and drivers
This series introduces a generic kernel-userspace API for retrieving PHY
Mean Square Error (MSE) diagnostics, together with netlink integration,
a fast-path reporting hook in LINKSTATE_GET, and initial driver
implementations for the KSZ9477 and DP83TD510E PHYs.
MSE is defined by the OPEN Alliance "Advanced diagnostic features for
100BASE-T1 automotive Ethernet PHYs" specification [1] as a measure of
slicer error rate, typically used internally to derive the Signal
Quality Indicator (SQI). While SQI is useful as a normalized quality
index, it hides raw measurement data, varies in scaling and thresholds
between vendors, and may not indicate certain failure modes - for
example, cases where autonegotiation would fail even though SQI reports
a good link. In practice, such scenarios can only be investigated in
fixed-link mode; here, MSE can provide an empirically estimated value
indicating conditions under which autonegotiation would not succeed.
Example output with current implementation:
root@DistroKit:~ ethtool lan1
Settings for lan1:
...
Speed: 1000Mb/s
Duplex: Full
...
Link detected: yes
SQI: 5/7
MSE: 3/127 (channel: worst)
root@DistroKit:~ ethtool --show-mse lan1
MSE diagnostics for lan1:
MSE Configuration:
Max Average MSE: 127
Refresh Rate: 2000000 ps
Symbols per Sample: 250
Supported capabilities: average channel-a channel-b channel-c
channel-d worst
MSE Snapshot (Channel: a):
Average MSE: 4
MSE Snapshot (Channel: b):
Average MSE: 3
MSE Snapshot (Channel: c):
Average MSE: 2
MSE Snapshot (Channel: d):
Average MSE: 3
[1] https://opensig.org/wp-content/uploads/2024/01/Advanced_PHY_features_for_automotive_Ethernet_V1.0.pdf
====================
Link: https://patch.msgid.link/20251027122801.982364-1-o.rempel@pengutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Implement get_mse_capability() and get_mse_snapshot() for the DP83TD510E
to expose its Mean Square Error (MSE) register via the new PHY MSE
UAPI.
The DP83TD510E does not document any peak MSE values; it only exposes
a single average MSE register used internally to derive SQI. This
implementation therefore advertises only PHY_MSE_CAP_AVG, along with
LINK and channel-A selectors. Scaling is fixed to 0xFFFF, and the
refresh interval/number of symbols are estimated from 10BASE-T1L
symbol rate (7.5 MBd) and typical diagnostic intervals (~1 ms).
For 10BASE-T1L deployments, SQI is a reliable indicator of link
modulation quality once the link is established, but it does not
indicate whether autonegotiation pulses will be correctly received
in marginal conditions. MSE provides a direct measurement of slicer
error rate that can be used to evaluate if autonegotiation is likely
to succeed under a given cable length and condition. In practice,
testing such scenarios often requires forcing a fixed-link setup to
isolate MSE behaviour from the autonegotiation process.
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Link: https://patch.msgid.link/20251027122801.982364-5-o.rempel@pengutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Implement the get_mse_capability() and get_mse_snapshot() PHY driver ops
for KSZ9477-series integrated PHYs to demonstrate the new PHY MSE
UAPI.
These PHYs do not expose a documented direct MSE register, but the
Signal Quality Indicator (SQI) registers are derived from the
internal MSE computation. This hook maps SQI readings into the MSE
interface so that tooling can retrieve the raw value together with
metadata for correct interpretation in userspace.
Behaviour:
- For 1000BASE-T, report per-channel (A–D) values and support a
WORST channel selector.
- For 100BASE-TX, only LINK-wide measurements are available.
- Report average MSE only, with a max scale based on
KSZ9477_MMD_SQI_MASK and a fixed refresh rate of 2 µs.
This mapping differs from the OPEN Alliance SQI definition, which
assigns thresholds such as pre-fail indices; the MSE interface
instead provides the raw measurement, leaving interpretation to
userspace.
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Link: https://patch.msgid.link/20251027122801.982364-4-o.rempel@pengutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Introduce the userspace entry point for PHY MSE diagnostics via
ethtool netlink. This exposes the core API added previously and
returns both capability information and one or more snapshots.
Userspace sends ETHTOOL_MSG_MSE_GET. The reply carries:
- ETHTOOL_A_MSE_CAPABILITIES: scale limits and timing information
- ETHTOOL_A_MSE_CHANNEL_* nests: one or more snapshots (per-channel
if available, otherwise WORST, otherwise LINK)
Link down returns -ENETDOWN.
Changes:
- YAML: add attribute sets (mse, mse-capabilities, mse-snapshot)
and the mse-get operation
- UAPI (generated): add ETHTOOL_A_MSE_* enums and message IDs,
ETHTOOL_MSG_MSE_GET/REPLY
- ethtool core: add net/ethtool/mse.c implementing the request,
register genl op, and hook into ethnl dispatch
- docs: document MSE_GET in ethtool-netlink.rst
The include/uapi/linux/ethtool_netlink_generated.h is generated
from Documentation/netlink/specs/ethtool.yaml.
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Link: https://patch.msgid.link/20251027122801.982364-3-o.rempel@pengutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add the base infrastructure for Mean Square Error (MSE) diagnostics,
as proposed by the OPEN Alliance "Advanced diagnostic features for
100BASE-T1 automotive Ethernet PHYs" [1] specification.
The OPEN Alliance spec defines only average MSE and average peak MSE
over a fixed number of symbols. However, other PHYs, such as the
KSZ9131, additionally expose a worst-peak MSE value latched since the
last channel capture. This API accounts for such vendor extensions by
adding a distinct capability bit and snapshot field.
Channel-to-pair mapping is normally straightforward, but in some cases
(e.g. 100BASE-TX with MDI-X resolution unknown) the mapping is ambiguous.
If hardware does not expose MDI-X status, the exact pair cannot be
determined. To avoid returning misleading per-channel data in this case,
a LINK selector is defined for aggregate MSE measurements.
All investigated devices differ in MSE capabilities, such
as sample rate, number of analyzed symbols, and scaling factors.
For example, the KSZ9131 uses different scaling for MSE and pMSE.
To make this visible to callers, scale limits and timing information
are returned via get_mse_capability().
Some PHYs sample very few symbols at high frequency (e.g. 2 us update
rate). To cover such cases and allow for future high-speed PHYs with
even shorter intervals, the refresh rate is reported as u64 in
picoseconds.
This patch introduces the internal PHY API for Mean Square Error
diagnostics. It defines new kernel-side data types and driver hooks:
- struct phy_mse_capability: describes supported metrics, scale
limits, update interval, and sampling length.
- struct phy_mse_snapshot: holds one correlated measurement set.
- New phy_driver ops: `get_mse_capability()` and `get_mse_snapshot()`.
These definitions form the core kernel API. No user-visible interfaces
are added in this commit.
Standardization notes:
OPEN Alliance defines presence and interpretation of some metrics but does
not fix numeric scales or sampling internals:
- SQI (3-bit, 0..7) is mandatory; correlation to SNR/BER is informative
(OA 100BASE-T1 TC1 v1.0 6.1.2; OA 1000BASE-T1 TC12 v2.2 6.1.2).
- MSE is optional; OA recommends 2^16 symbols and scaling to 0..511,
with a worst-case latch since last read (OA 100BASE-T1 TC1 v1.0 6.1.1; OA
1000BASE-T1 TC12 v2.2 6.1.1). Refresh is recommended (~0.8-2.0 ms for
100BASE-T1; ~80-200 us for 1000BASE-T1). Exact scaling/time windows
are vendor-specific.
- Peak MSE (pMSE) is defined only for 100BASE-T1 as optional, e.g.
128-symbol sliding window with 8-bit range and worst-case latch (OA
100BASE-T1 TC1 v1.0 6.1.3).
Therefore this API exposes which measures and selectors a PHY supports,
and documents where behavior is standard-referenced vs vendor-specific.
[1] <https://opensig.org/wp-content/uploads/2024/01/
Advanced_PHY_features_for_automotive_Ethernet_V1.0.pdf>
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Link: https://patch.msgid.link/20251027122801.982364-2-o.rempel@pengutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Samiullah Khawaja says:
====================
Add support to do threaded napi busy poll
Extend the already existing support of threaded napi poll to do continuous
busy polling.
This is used for doing continuous polling of napi to fetch descriptors
from backing RX/TX queues for low latency applications. Allow enabling
of threaded busypoll using netlink so this can be enabled on a set of
dedicated napis for low latency applications.
Once enabled user can fetch the PID of the kthread doing NAPI polling
and set affinity, priority and scheduler for it depending on the
low-latency requirements.
Extend the netlink interface to allow enabling/disabling threaded
busypolling at individual napi level.
We use this for our AF_XDP based hard low-latency usecase with usecs
level latency requirement. For our usecase we want low jitter and stable
latency at P99.
Following is an analysis and comparison of available (and compatible)
busy poll interfaces for a low latency usecase with stable P99. This can
be suitable for applications that want very low latency at the expense
of cpu usage and efficiency.
Already existing APIs (SO_BUSYPOLL and epoll) allow busy polling a NAPI
backing a socket, but the missing piece is a mechanism to busy poll a
NAPI instance in a dedicated thread while ignoring available events or
packets, regardless of the userspace API. Most existing mechanisms are
designed to work in a pattern where you poll until new packets or events
are received, after which userspace is expected to handle them.
As a result, one has to hack together a solution using a mechanism
intended to receive packets or events, not to simply NAPI poll. NAPI
threaded busy polling, on the other hand, provides this capability
natively, independent of any userspace API. This makes it really easy to
setup and manage.
For analysis we use an AF_XDP based benchmarking tool `xsk_rr`. The
description of the tool and how it tries to simulate the real workload
is following,
- It sends UDP packets between 2 machines.
- The client machine sends packets at a fixed frequency. To maintain the
frequency of the packet being sent, we use open-loop sampling. That is
the packets are sent in a separate thread.
- The server replies to the packet inline by reading the pkt from the
recv ring and replies using the tx ring.
- To simulate the application processing time, we use a configurable
delay in usecs on the client side after a reply is received from the
server.
The xsk_rr tool is posted separately as an RFC for tools/testing/selftest.
We use this tool with following napi polling configurations,
- Interrupts only
- SO_BUSYPOLL (inline in the same thread where the client receives the
packet).
- SO_BUSYPOLL (separate thread and separate core)
- Threaded NAPI busypoll
System is configured using following script in all 4 cases,
```
echo 0 | sudo tee /sys/class/net/eth0/threaded
echo 0 | sudo tee /proc/sys/kernel/timer_migration
echo off | sudo tee /sys/devices/system/cpu/smt/control
sudo ethtool -L eth0 rx 1 tx 1
sudo ethtool -G eth0 rx 1024
echo 0 | sudo tee /proc/sys/net/core/rps_sock_flow_entries
echo 0 | sudo tee /sys/class/net/eth0/queues/rx-0/rps_cpus
# pin IRQs on CPU 2
IRQS="$(gawk '/eth0-(TxRx-)?1/ {match($1, /([0-9]+)/, arr); \
print arr[0]}' < /proc/interrupts)"
for irq in "${IRQS}"; \
do echo 2 | sudo tee /proc/irq/$irq/smp_affinity_list; done
echo -1 | sudo tee /proc/sys/kernel/sched_rt_runtime_us
for i in /sys/devices/virtual/workqueue/*/cpumask; \
do echo $i; echo 1,2,3,4,5,6 > $i; done
if [[ -z "$1" ]]; then
echo 400 | sudo tee /proc/sys/net/core/busy_read
echo 100 | sudo tee /sys/class/net/eth0/napi_defer_hard_irqs
echo 15000 | sudo tee /sys/class/net/eth0/gro_flush_timeout
fi
sudo ethtool -C eth0 adaptive-rx off adaptive-tx off rx-usecs 0 tx-usecs 0
if [[ "$1" == "enable_threaded" ]]; then
echo 0 | sudo tee /proc/sys/net/core/busy_poll
echo 0 | sudo tee /proc/sys/net/core/busy_read
echo 100 | sudo tee /sys/class/net/eth0/napi_defer_hard_irqs
echo 15000 | sudo tee /sys/class/net/eth0/gro_flush_timeout
NAPI_ID=$(ynl --family netdev --output-json --do queue-get \
--json '{"ifindex": '${IFINDEX}', "id": '0', "type": "rx"}' | jq '."napi-id"')
ynl --family netdev --json '{"id": "'${NAPI_ID}'", "threaded": "busy-poll"}'
NAPI_T=$(ynl --family netdev --output-json --do napi-get \
--json '{"id": "'$NAPI_ID'"}' | jq '."pid"')
sudo chrt -f -p 50 $NAPI_T
# pin threaded poll thread to CPU 2
sudo taskset -pc 2 $NAPI_T
fi
if [[ "$1" == "enable_interrupt" ]]; then
echo 0 | sudo tee /proc/sys/net/core/busy_read
echo 0 | sudo tee /sys/class/net/eth0/napi_defer_hard_irqs
echo 15000 | sudo tee /sys/class/net/eth0/gro_flush_timeout
fi
```
To enable various configurations, script can be run as following,
- Interrupt Only
```
<script> enable_interrupt
```
- SO_BUSYPOLL (no arguments to script)
```
<script>
```
- NAPI threaded busypoll
```
<script> enable_threaded
```
Once configured, the workload is run with various configurations using
following commands. Set period (1/frequency) and delay in usecs to
produce results for packet frequency and application processing delay.
## Interrupt Only and SO_BUSYPOLL (inline)
- Server
```
sudo chrt -f 50 taskset -c 3-5 ./xsk_rr -o 0 -B 400 -i eth0 -4 \
-D <IP-dest> -S <IP-src> -M <MAC-dst> -m <MAC-src> -p 54321 -h -v
```
- Client
```
sudo chrt -f 50 taskset -c 3-5 ./xsk_rr -o 0 -B 400 -i eth0 -4 \
-S <IP-src> -D <IP-dest> -m <MAC-src> -M <MAC-dst> -p 54321 \
-P <Period-usecs> -d <Delay-usecs> -T -l 1 -v
```
## SO_BUSYPOLL(done in separate core using recvfrom)
Argument -t spawns a separate thread and continuously calls recvfrom.
- Server
```
sudo chrt -f 50 taskset -c 3-5 ./xsk_rr -o 0 -B 400 -i eth0 -4 \
-D <IP-dest> -S <IP-src> -M <MAC-dst> -m <MAC-src> -p 54321 \
-h -v -t
```
- Client
```
sudo chrt -f 50 taskset -c 3-5 ./xsk_rr -o 0 -B 400 -i eth0 -4 \
-S <IP-src> -D <IP-dest> -m <MAC-src> -M <MAC-dst> -p 54321 \
-P <Period-usecs> -d <Delay-usecs> -T -l 1 -v -t
```
## NAPI Threaded Busy Poll
Argument -n skips the recvfrom call as there is no recv kick needed.
- Server
```
sudo chrt -f 50 taskset -c 3-5 ./xsk_rr -o 0 -B 400 -i eth0 -4 \
-D <IP-dest> -S <IP-src> -M <MAC-dst> -m <MAC-src> -p 54321 \
-h -v -n
```
- Client
```
sudo chrt -f 50 taskset -c 3-5 ./xsk_rr -o 0 -B 400 -i eth0 -4 \
-S <IP-src> -D <IP-dest> -m <MAC-src> -M <MAC-dst> -p 54321 \
-P <Period-usecs> -d <Delay-usecs> -T -l 1 -v -n
```
| Experiment | interrupts | SO_BUSYPOLL | SO_BUSYPOLL(separate) | NAPI threaded |
|---|---|---|---|---|
| 12 Kpkt/s + 0us delay | | | | |
| | p5: 12700 | p5: 12900 | p5: 13300 | p5: 12800 |
| | p50: 13100 | p50: 13600 | p50: 14100 | p50: 13000 |
| | p95: 13200 | p95: 13800 | p95: 14400 | p95: 13000 |
| | p99: 13200 | p99: 13800 | p99: 14400 | p99: 13000 |
| 32 Kpkt/s + 30us delay | | | | |
| | p5: 19900 | p5: 16600 | p5: 13100 | p5: 12800 |
| | p50: 21100 | p50: 17000 | p50: 13700 | p50: 13000 |
| | p95: 21200 | p95: 17100 | p95: 14000 | p95: 13000 |
| | p99: 21200 | p99: 17100 | p99: 14000 | p99: 13000 |
| 125 Kpkt/s + 6us delay | | | | |
| | p5: 14600 | p5: 17100 | p5: 13300 | p5: 12900 |
| | p50: 15400 | p50: 17400 | p50: 13800 | p50: 13100 |
| | p95: 15600 | p95: 17600 | p95: 14000 | p95: 13100 |
| | p99: 15600 | p99: 17600 | p99: 14000 | p99: 13100 |
| 12 Kpkt/s + 78us delay | | | | |
| | p5: 14100 | p5: 16700 | p5: 13200 | p5: 12600 |
| | p50: 14300 | p50: 17100 | p50: 13900 | p50: 12800 |
| | p95: 14300 | p95: 17200 | p95: 14200 | p95: 12800 |
| | p99: 14300 | p99: 17200 | p99: 14200 | p99: 12800 |
| 25 Kpkt/s + 38us delay | | | | |
| | p5: 19900 | p5: 16600 | p5: 13000 | p5: 12700 |
| | p50: 21000 | p50: 17100 | p50: 13800 | p50: 12900 |
| | p95: 21100 | p95: 17100 | p95: 14100 | p95: 12900 |
| | p99: 21100 | p99: 17100 | p99: 14100 | p99: 12900 |
## Observations
- Here without application processing all the approaches give the same
latency within 1usecs range and NAPI threaded gives minimum latency.
- With application processing the latency increases by 3-4usecs when
doing inline polling.
- Using a dedicated core to drive napi polling keeps the latency same
even with application processing. This is observed both in userspace
and threaded napi (in kernel).
- Using napi threaded polling in kernel gives lower latency by
1-2usecs as compared to userspace driven polling in separate core.
- Even on a dedicated core, SO_BUSYPOLL adds around 1-2usecs of latency.
This is because it doesn't continuously busy poll until events are
ready. Instead, it returns after polling only once, requiring the
process to re-invoke the syscall for each poll, which requires a new
enter/leave kernel cycle and the setup/teardown of the busy poll for
every single poll attempt.
- With application processing userspace will get the packet from recv
ring and spend some time doing application processing and then do napi
polling. While application processing is happening a dedicated core
doing napi polling can pull the packet of the NAPI RX queue and
populate the AF_XDP recv ring. This means that when the application
thread is done with application processing it has new packets ready to
recv and process in recv ring.
- Napi threaded busy polling in the kernel with a dedicated core gives
the consistent P5-P99 latency.
Note well that threaded napi busy-polling has not been shown to yield
efficiency or throughput benefits. In contrast, dedicating an entire
core to busy-polling one NAPI (NIC queue) is rather inefficient.
However, in certain specific use cases, this mechanism results in lower
packet processing latency. The experimental testing reported here only
covers those use cases and does not present a comprehensive evaluation
of threaded napi busy-polling.
Following histogram is generated to measure the time spent in recvfrom
while using inline thread with SO_BUSYPOLL. The histogram is generated
using the following bpftrace command. In this experiment there are 32K
packets per second and the application processing delay is 30usecs. This
is to measure whether there is significant time spent pulling packets
from the descriptor queue that it will affect the overall latency if
done inline.
```
bpftrace -e '
kprobe:xsk_recvmsg {
@start[tid] = nsecs;
}
kretprobe:xsk_recvmsg {
if (@start[tid]) {
$sample = (nsecs - @start[tid]);
@xsk_recvfrom_hist = hist($sample);
delete(@start[tid]);
}
}
END { clear(@start);}'
```
Here in case of inline busypolling around 35 percent of calls are taking
1-2usecs and around 50 percent are taking 0.5-2usecs.
@xsk_recvfrom_hist:
[128, 256) 24073 |@@@@@@@@@@@@@@@@@@@@@@ |
[256, 512) 55633 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|
[512, 1K) 20974 |@@@@@@@@@@@@@@@@@@@ |
[1K, 2K) 34234 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ |
[2K, 4K) 3266 |@@@ |
[4K, 8K) 19 | |
====================
Link: https://patch.msgid.link/20251028203007.575686-1-skhawaja@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add testcase to run busy poll test with threaded napi busy poll enabled.
Signed-off-by: Samiullah Khawaja <skhawaja@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Acked-by: Martin Karsten <mkarsten@uwaterloo.ca>
Tested-by: Martin Karsten <mkarsten@uwaterloo.ca>
Link: https://patch.msgid.link/20251028203007.575686-3-skhawaja@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
