[ovs-dev] [PATCH V3 0/4] dpif-netdev offload transitions

[Eli Britstein]

On 8/2/2021 10:58 PM, Ilya Maximets wrote:
> External email: Use caution opening links or attachments
>
>
> On 7/26/21 10:14 AM, Eli Britstein wrote:
>> This patch-set improves/fixes offloads transitions behavior.
>>
>> Patch #1 avoids flushing PMD offloads unnecessarily.
>> Patch #2 fixes a flow modifications bug.
>> Patch #3 fixes a race condition with flow modifications.
>> Patch #3 improves debuggability of flow modifications.
>>
>> v2-v1:
>> - Rebase.
>> v3-v2:
>> - Added patch #2 for fixes another flow modification scenario.
>> - Changed log to a separated function.
>>
>> GitHub Actions:
>> v1: https://github.com/elibritstein/OVS/actions/runs/769805954
>>      - This run has encountered some internal GitHub problems.
>>      - A previous good run, with the same code, only changed commit
>>        messages since:
>>      https://github.com/elibritstein/OVS/actions/runs/766660787
>> v2: https://github.com/elibritstein/OVS/actions/runs/1023045302
>> v3: https://github.com/elibritstein/OVS/actions/runs/1066584199
>>
>> Eli Britstein (4):
>>    dpif-netdev: Do not flush PMD offloads on reload
>>    dpif-netdev: Fix flow modification after failure
>>    dpif-netdev: Fix offloads of modified flows
>>    dpif-netdev: Log flow modification in debug level
>>
>>   lib/dpif-netdev.c | 144 ++++++++++++++++++++++++----------------------
>>   1 file changed, 74 insertions(+), 70 deletions(-)
>>
> Thanks!
> I renamed the function in the patch #4 from 'dump_*' to 'log_*'
> to avoid confusion with flow dump functions and applied the
> patch set.
>
> I added Ack from David to patch #1 since it's the same as in v2.
> Patches #2 and #3 backported down to 2.13.
Thanks. Indeed I forgot to add it.
>
> Technically, I think, we can consider patch #1 as a bug fix too
> and backport it down to 2.15.  Let me know if that's needed.

This patch depends on the flush series that was merged in 2.15, but 
didn't have Fixes tags, though it fixed a memory leak.

I think here it is not severe as a memory leak, only failure in 
offloads, but similar to the severity of patches #2,#3.

Though it's not a clear cut, if #2,#3 are backported, #1 can be too.

>
> Best regards, Ilya Maximets.