[ANNOUNCE] v6.1-rc2-rt1

[ANNOUNCE] v6.1-rc2-rt1

[Sebastian Andrzej Siewior]

Dear RT folks!

I'm pleased to announce the v6.1-rc2-rt1 patch set. 

Changes since v6.0.3-rt12:

  - Update to v6.1-rc2.

Known issues
     - Valentin Schneider reported a few splats on ARM64, see
          https://lkml.kernel.org/r/20210810134127.1394269-1-valentin.schneider@arm.com

You can get this release via the git tree at:

    git://git.kernel.org/pub/scm/linux/kernel/git/rt/linux-rt-devel.git v6.1-rc2-rt1

The RT patch against v6.1-rc2 can be found here:

    https://cdn.kernel.org/pub/linux/kernel/projects/rt/6.1/older/patch-6.1-rc2-rt1.patch.xz

The split quilt queue is available at:

    https://cdn.kernel.org/pub/linux/kernel/projects/rt/6.1/older/patches-6.1-rc2-rt1.tar.xz

Sebastian

v6.0.3-rt12 does not compile (WAS: [ANNOUNCE] v6.1-rc2-rt1)

[Sebastian Andrzej Siewior]

Please keep the list in Cc:

On 2022-10-27 00:26:45 [+0200], David Runge wrote:
> Hi Sebastian,

Hi David,

> I'm following the linux-rt development for the Arch Linux package [1].
> With 6.0.3-rt12 I have run into a bit of a pickle though.

I released it the other as the last one for the v6.0 series. Didn't
announce it officially since I made no RT relevant change.

> I'm seeing:
> 
> ```
> drivers/net/ethernet/xilinx/xilinx_axienet_main.c: In function ‘axienet_get_stats64’:
> drivers/net/ethernet/xilinx/xilinx_axienet_main.c:1308:25: error: implicit declaration of function ‘u64_stats_fetch_begin_irq’; did you mean ‘u64_stats_fetch_begin’? [-Werror=implicit-function-declaration]
>  1308 |                 start = u64_stats_fetch_begin_irq(&lp->rx_stat_sync);
>       |                         ^~~~~~~~~~~~~~~~~~~~~~~~~
>       |                         u64_stats_fetch_begin
> drivers/net/ethernet/xilinx/xilinx_axienet_main.c:1311:18: error: implicit declaration of function ‘u64_stats_fetch_retry_irq’; did you mean ‘u64_stats_fetch_retry’? [-Werror=implicit-function-declaration]
>  1311 |         } while (u64_stats_fetch_retry_irq(&lp->rx_stat_sync, start));
>       |                  ^~~~~~~~~~~~~~~~~~~~~~~~~
>       |                  u64_stats_fetch_retry
> ```

I addressed this for the v6.1 release, wasn't aware that it also managed
its way into the v6.0 series.

> Grepping for either ‘u64_stats_fetch_begin_irq’, or
> ‘u64_stats_fetch_retry_irq’ in the codebase, they are indeed the only
> occurences of it in that file... which is strange.

Well, we are working on getting rid of this function. Upstream should
follow in ~v6.2.

> ```
> rg u64_stats_fetch_begin_irq
> drivers/net/ethernet/xilinx/xilinx_axienet_main.c
> 1308:           start = u64_stats_fetch_begin_irq(&lp->rx_stat_sync);
> 1314:           start = u64_stats_fetch_begin_irq(&lp->tx_stat_sync);
> ```
> 
> If I check our own kernel or Greg's v6.0.3 tag there are plenty of
> occurences of the function all across the codebase, but not as soon as
> the patchset is merged.

There is
   0002-net-Remove-the-obsolte-u64_stats_fetch_-_irq-users.patch

which gets rid of drivers/net users and then there is
   u64_stat-Remove-the-obsolete-fetch_irq-variants.patch

which removes the API once all users are gone.

> When doing the release for 6.0.0-rt11 I already noticed a few missing
> options but was under the impression that they might be missing due to
> the realtime options disabling them. However, now it looks as if there
> are entire subsets missing oO
> Does the patchset drop a lot of network driver code and/or do the last
> occurences there also have to be removed?
> As is, I am unable to build v6.0.3-rt12.

I'm going to release -rt13 in a bit with this fixed.
There are few options like SLAB or TRANSPARENT_HUGEPAGE which are not
available on PREEMPT_RT for one reason or another. I try to avoid
dropping options if possible.

> Best,
> David

Sebastian

Re: v6.0.3-rt12 does not compile (WAS: [ANNOUNCE] v6.1-rc2-rt1)

[David Runge]

[-- Attachment #1: Type: text/plain, Size: 903 bytes --]

On 2022-10-27 09:01:27 (+0200), Sebastian Andrzej Siewior wrote:
> Please keep the list in Cc:

Thanks, will keep that in mind for the future!

> I'm going to release -rt13 in a bit with this fixed.
> There are few options like SLAB or TRANSPARENT_HUGEPAGE which are not
> available on PREEMPT_RT for one reason or another. I try to avoid
> dropping options if possible.

Seems I forgot to reply...
Thanks for releasing a new version so fast and being on top of these
changes!

For reference: Will there be more rebases on stable versions of the most
recent kernel versions going forward?
The linux-stable-rt repository seems to only cover LTS kernels AFAIKT
and a few more releases in the current stable kernel version range would
allow downstreams to more easily backport patches (e.g. for security
issues) also for the realtime kernel.

Best,
David

-- 
https://sleepmap.de

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

Re: v6.0.3-rt12 does not compile (WAS: [ANNOUNCE] v6.1-rc2-rt1)

[Sebastian Andrzej Siewior]

On 2022-11-15 18:07:09 [+0100], David Runge wrote:
> 
> For reference: Will there be more rebases on stable versions of the most
> recent kernel versions going forward?

I don't plan to.

> The linux-stable-rt repository seems to only cover LTS kernels AFAIKT
> and a few more releases in the current stable kernel version range would
> allow downstreams to more easily backport patches (e.g. for security
> issues) also for the realtime kernel.

The non-LTS kernels are dropped quickly so I don't expect v6.0 to stay
for long. I plan to stay behind upstream as close as possible so I don't
have to backport anything for testing and development.

> Best,
> David

Sebastian