Skip to contentSkip to conversation

netdev call - Jan 16th

ANN Jakub Kicinski 5 participants 8 messages Jan 16, 2024 – Jan 17, 2024

Subsystems (2)
lore
Share & tools
lore
#1 lore raw
Quoted 7 lines · Jakub Kicinski

Hi,

The bi-weekly netdev call at https://bbb.lwn.net/b/jak-wkr-seg-hjn
is scheduled tomorrow at 8:30 am (PT) / 5:30 pm (~EU).

There's a minor CI update. Please suggest other topics.

I would like to discuss a process question for posting a fix to a stable kernel that isn't needed in the latest upstream as it was fixed another way.

This is related to this thread:

https://lore.kernel.org/linux-patches/[email protected]/

Thanks.

#2 lore raw
Quoted 9 lines · Andy Gospodarek
Quoted 7 lines · Jakub Kicinski

Hi,

The bi-weekly netdev call at https://bbb.lwn.net/b/jak-wkr-seg-hjn
is scheduled tomorrow at 8:30 am (PT) / 5:30 pm (~EU).

There's a minor CI update. Please suggest other topics.

Quoted 9 lines

I would like to discuss a process question for posting a fix to a stable kernel that isn't needed in the latest upstream as it was fixed another way.

This is related to this thread:

https://lore.kernel.org/linux-patches/[email protected]/

Thanks.

If you send it to stable with a tag like,

CC <[email protected]> # 5.15.x

or whatever kernel you need this has worked from me. This has worked for me if I understood the above question correctly. The relevant docs are in Documentation/process/stable-kernel-rules.rst. The following bit seems to explain it.

* For patches that may have kernel version prerequisites specify them using the following format in the sign-off area:

.. code-block:: none

     Cc: <[email protected]> # 3.3.x

The tag has the meaning of:

.. code-block:: none

git cherry-pick <this commit>

For each "-stable" tree starting with the specified version.

Note, such tagging is unnecessary if the stable team can derive the appropriate versions from Fixes: tags.

* To delay pick up of patches, use the following format:

.. code-block:: none

Cc <[email protected]> # after 4 weeks in mainline

* For any other requests, just add a note to the stable tag. This for example can be used to point out known problems:

.. code-block:: none

Cc <[email protected]> # see patch description, needs adjustments for <= 6.3
#3 lore raw
Quoted 51 lines · John Fastabend
Quoted 9 lines · Andy Gospodarek
Quoted 7 lines · Jakub Kicinski

Hi,

The bi-weekly netdev call at https://bbb.lwn.net/b/jak-wkr-seg-hjn
is scheduled tomorrow at 8:30 am (PT) / 5:30 pm (~EU).

There's a minor CI update. Please suggest other topics.

Quoted 9 lines

I would like to discuss a process question for posting a fix to a stable kernel that isn't needed in the latest upstream as it was fixed another way.

This is related to this thread:

https://lore.kernel.org/linux-patches/[email protected]/

Thanks.

If you send it to stable with a tag like,

CC <[email protected]> # 5.15.x

or whatever kernel you need this has worked from me. This has worked for me if I understood the above question correctly. The relevant docs are in Documentation/process/stable-kernel-rules.rst. The following bit seems to explain it.

* For patches that may have kernel version prerequisites specify them using the following format in the sign-off area:

.. code-block:: none

     Cc: <[email protected]> # 3.3.x

The tag has the meaning of:

.. code-block:: none

git cherry-pick <this commit>

For each "-stable" tree starting with the specified version.

Note, such tagging is unnecessary if the stable team can derive the appropriate versions from Fixes: tags.

* To delay pick up of patches, use the following format:

.. code-block:: none

Quoted 8 lines
Cc <[email protected]> # after 4 weeks in mainline

* For any other requests, just add a note to the stable tag. This for example can be used to point out known problems:

.. code-block:: none

Cc <[email protected]> # see patch description, needs adjustments for <= 6.3

Thanks, John.

This one is a bit odd what happened is that by the time this problem was reported (on an older kernel), the code changed out from underneath. The new code was bug-compatible with the old code (whic is an indicator that a good job was done porting bnxt_en to use DMA mapping from the page pool infra.

So at the time when the patch was posted, the tip of tree code did need a fixes tag of:

Fixes 578fcfd26e2a ("bnxt_en: Let the page pool manage the DMA mapping")

In any kernel pre-6.6 (or if we found the patch a few months back) the patch would have been different (since the dma mapping was handled differently) and the fixes tag would have been:

Fixes a7559bc8c17c ("bnxt: support transmit and free of aggregation buffers")

Greg's concern (at least as I read it) is that it looks like I'm asking him to take a patch that is not upstream. That totally makes sense, but now what?

Based on what I see in Documentation, I think I need to just resubmit to stable as that would follow "Option 2" and explain why it 'deviates from the original' patch.

#4 lore raw
Quoted 4 lines

Thanks, John.

This one is a bit odd what happened is that by the time this problem was reported (on an older kernel), the code changed out from underneath.

Hi Andy

Talk to Florian

He has dealt with Fixes like this in the past. It generally works out fine so long as you are explicit about what is going on, in comments under the ---. That, plus correctly marking what kernel version the patch is for.

Andrew

#5 lore raw
Quoted 7 lines · Andrew Lunn
Quoted 4 lines

Thanks, John.

This one is a bit odd what happened is that by the time this problem was reported (on an older kernel), the code changed out from underneath.

Hi Andy

Quoted 8 lines

Talk to Florian

He has dealt with Fixes like this in the past. It generally works out fine so long as you are explicit about what is going on, in comments under the ---. That, plus correctly marking what kernel version the patch is for.

Andrew

Will do! I know where to find him!

#6 lore raw
Quoted 10 lines · Andy Gospodarek
Quoted 9 lines · Andrew Lunn
Quoted 4 lines

Thanks, John.

This one is a bit odd what happened is that by the time this problem was reported (on an older kernel), the code changed out from underneath.

Hi Andy

Talk to Florian

Quoted 8 lines
Quoted 6 lines

He has dealt with Fixes like this in the past. It generally works out fine so long as you are explicit about what is going on, in comments under the ---. That, plus correctly marking what kernel version the patch is for.

Andrew

Will do! I know where to find him!

Hi there! It looks like doing a "targeted" set of fixes for pre-6.6 kernels is appropriate so you would just want to submit them with a [PATCH stable X.Y] subject to clearly denote which stable branch they apply against, and have a cover letter that explains why.

Signature · 1 line
Florian
author maintainer
#7 lore raw
In reply to Jakub Kicinski
Quoted 4 lines · On Mon, 15 Jan 2024 17:54:40 -0800 Jakub Kicinski wrote:

The bi-weekly netdev call at https://bbb.lwn.net/b/jak-wkr-seg-hjn
is scheduled tomorrow at 8:30 am (PT) / 5:30 pm (~EU).

There's a minor CI update. Please suggest other topics.

Sorry about the delay in sending minutes, here they are:

* BQL (Dave Taht)
* Embedded drivers missing BQL
* Complain to the driver maintainers
* Driver review time is best to request basic features like BQL
* Some of the problems are in the vendor downstream drivers
* Are vendors running flent and other latency tests?
* Jesse: yes, although they struggle at DC speeds
* Multi-queue BQL
* Not much interest among attendees
* Andrew: try to keep it in the core, fewer driver changes the better
* CI
* Most of the networking selftests are now run on pending patches
* https://netdev.bots.linux.dev/status.html
* Not reporting back to patchwork, yet, because there’s a bunch of pre-existing failures
* Paolo: we should extend the tests with mptcp tests
* DSA tests may also be useful - Andrew to investigate if they can run in SW
* We’re looking to integrate with other people running tests themselves

Unrelated to the call, but we have also posted a "2023 netdev retrospective": https://people.kernel.org/kuba/netdev-in-2023