APPLIED[Q/P/N/J]: [SRU][R/Q/P/N/J][PATCH 0/1] i40e: avoid redundant VF link state updates
Stefan Bader
stefan.bader at canonical.com
Tue Nov 11 13:37:44 UTC 2025
On 11/11/2025 14:29, Stefan Bader wrote:
> On 03/11/2025 16:40, Robert Malz wrote:
>> BugLink: https://bugs.launchpad.net/bugs/2130552
>>
>> [ Impact ]
>>
>> * In the i40e driver, every VF link state change request is causing
>> a VF reset,
>> even if the requested change is already set. This leads to
>> unnecessary VF
>> resets and can cause performance degradation or instability in the VF
>> driver, particularly in DPDK environments.
>>
>> * With the proposed patch, i40e will skip VF link state change
>> requests when the
>> desired link state matches the current configuration. This prevents
>> unnecessary VF resets and reduces PF-VF communication overhead.
>>
>> * The issue was originally detected by analyzing DPDK VF driver
>> issues during
>> OpenStack Neutron restarts. In such cases, i40e will issue an
>> unnecessary VF
>> reset request which causes instability in the DPDK driver.
>>
>> [ Test Plan ]
>>
>> * The issue can be reproduced by triggering VF link state changes
>> multiple times
>> with the same requested link state.
>>
>> * To reproduce the problem, run the following command multiple times
>> on the same interface: 'ip link set <ifname> vf 0 state auto'.
>> Every time the command is executed, the PF driver will trigger a
>> VF reset.
>>
>> * Monitor for VF reset events using dmesg or appropriate logging
>> mechanisms
>> to confirm the issue and verify the fix.
>>
>> * Regression testing should include:
>> - Verify normal VF link state changes still work correctly when
>> state actually changes
>> - Validate that legitimate state changes still trigger resets when
>> needed
>>
>> [ Where problems could occur ]
>>
>> * Changes only affect the i40e .ndo_set_vf_link_state routine. In case
>> of any regressions, only the ability to set VF link state changes
>> could be
>> impacted.
>>
>> * The patch was verified locally and passed Intel internal test
>> execution.
>>
>> * Potential failure modes:
>> - Logic error in state comparison could prevent legitimate state
>> changes
>> - Incorrect state tracking could lead to inconsistent VF behavior
>>
>> * Risk mitigation:
>> - The change is isolated to a specific code path in the i40e driver
>> - Existing VF functionality remains unchanged when state changes
>> are needed
>> - The patch adds a simple comparison check before existing link
>> change routine
>>
>> [ Other Info ]
>>
>> * At the time of writing this document, the target patch is still in
>> the net-next
>> tree and has not yet been merged into linux-next.
>>
>> * Patch: a7ae783da0b919550e260aebfca1c6ef030b99a4 from netdev/net-next
>>
>> * Patch applies on most versions without any conflicts
>>
>> * No dependencies on other patches
>>
>> * Backport requirements: Patch should apply cleanly to Ubuntu kernel
>> versions
>>
>> * VF reset during link state change was added as a part of
>> 31deb12e85c35ddd2c037f0107d05d8674cab2c0 (upstream) commit, which is
>> included in the Kernels starting from 5.15
>>
>> Jay Vosburgh (1):
>> i40e: avoid redundant VF link state updates
>>
>> drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c | 12 ++++++++++++
>> 1 file changed, 12 insertions(+)
>>
>
>
> Applied to questing,plucky,noble,jammy:linux/master-next (to be included
> in s2025.10.13). Thanks.
Wanted to add that I changed net-next to linux-next as it landed there
by now (same SHA1).
>
> -Stefan
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_0xE8675DEECBEECEA3.asc
Type: application/pgp-keys
Size: 48643 bytes
Desc: OpenPGP public key
URL: <https://lists.ubuntu.com/archives/kernel-team/attachments/20251111/c00068bb/attachment-0001.key>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <https://lists.ubuntu.com/archives/kernel-team/attachments/20251111/c00068bb/attachment-0001.sig>
More information about the kernel-team
mailing list