[ANNOUNCE] v7.3-rc4-rt1

[ANNOUNCE] v7.3-rc4-rt1

[Sebastian Andrzej Siewior]

Dear RT folks!

I'm pleased to announce the v7.3-rc4-rt1 patch set. 

Changes since v7.2-rt5:

  - Rebase to v7.3-rc4

The printk related changes are all gone. What is left in the queue is
PowerPC, i915 and a sysfs file:
- PowerPC
  Some progress has been made. What is left is a pseries related MMU
  change, a patch dropping KVM_MPIC (which is e500 only) and a stack
  protector related workaround.
  The stack protector workaround might work with a different fix since
  only arch/powerpc is using boot_init_stack_canary().

- i915
  The patches in the queue are different compared to what is worked on.
  Both sides seem to be stuck. The XE driver seemed to be an alternative
  on platforms which are supported. Now that I test it has the same
  vblank issues as the i915.

- sysfs file, /sys/kernel/realtime
  This file has been around since forever. It was used to "easily"
  detect if the PREEMPT_RT patch has been applied and the required
  Kconfig symbol has been enabled.
  There has been a discussion if this file should be upstreamed, if it
  should remain at the same location or if it should become a folder
  with features.
  Any change here implies that users need to update their scripts. Even
  if the file is present in !PREEMPT_RT kernels and returns 0 (as the
  file was created with the patch queue applied but CONFIG_PREEMPT_RT nt
  selected).
  Given that this never part of an upstream kernel, it is probably best to
  no include it. The uname(2) interface provides the same information.

I am considering not updating the queue anymore given how much has been
accomplished.

Known issues
    - Yoann Congal reported a bit spinlock in dm_exception_table_lock().
        https://lore.kernel.org/all/Z8GTjqgDe_5EkE3t@P-ASN-ECS-830T8C3.local

You can get this release via the git tree at:

    https://git.kernel.org/pub/scm/linux/kernel/git/rt/linux-rt-devel.git v7.3-rc4-rt1

The RT patch against v7.3-rc4 can be found here:

    https://cdn.kernel.org/pub/linux/kernel/projects/rt/7.3/older/patch-7.3-rc4-rt1.patch.xz

The split quilt queue is available at:

    https://cdn.kernel.org/pub/linux/kernel/projects/rt/7.3/older/patches-7.3-rc4-rt1.tar.xz

Sebastian

Re: [ANNOUNCE] v7.3-rc4-rt1

[Shrikanth Hegde]

On 9/22/26 8:11 PM, Sebastian Andrzej Siewior wrote:
> Dear RT folks!
> 
> I'm pleased to announce the v7.3-rc4-rt1 patch set.
> 
> Changes since v7.2-rt5:
> 
>    - Rebase to v7.3-rc4
> 
> The printk related changes are all gone. What is left in the queue is
> PowerPC, i915 and a sysfs file:
> - PowerPC
>    Some progress has been made. What is left is a pseries related MMU
>    change, a patch dropping KVM_MPIC (which is e500 only) and a stack
>    protector related workaround.
>    The stack protector workaround might work with a different fix since
>    only arch/powerpc is using boot_init_stack_canary().
> 

Thanks Sebastian.

It is nice to see a little progress has been made atleast.
But it was mostly my mistake that I couldn't get time to work on it.

Let me see if we can atleast get these ppc patches upstream.
Most seems sensible to me.

But then would come heavy lifting on function and performance verification.
That's a different beast altogether.

Re: [ANNOUNCE] v7.3-rc4-rt1

[Sebastian Andrzej Siewior]

On 2026-10-02 10:14:51 [+0530], Shrikanth Hegde wrote:
> It is nice to see a little progress has been made atleast.
> But it was mostly my mistake that I couldn't get time to work on it.
> 
> Let me see if we can atleast get these ppc patches upstream.
> Most seems sensible to me.
> 
> But then would come heavy lifting on function and performance verification.
> That's a different beast altogether.

Right. I remember the 32bit ports (e300, e500) were okay and from 64bit
I remember just e6500. I know John has a G4 and he didn't complain :)
So sadly it is almost just Book-E that I had.
The recent testing I did was on qemu (and there I hit the one bug I
reported) so it is just works or !works.

Sebastian