[0/1] arm64: enable PREEMPT_LAZY

Message

Valentin Schneider March 5, 2025, 10:49 a.m. UTC

Hey folks,

This is a resubmission of Mark's patch [1]. I didn't end up changing anything
other than fluffing up the changelog.

I ran this under an arm64 VM, added some trace_printk's and checked that
sched_switch's between runnable hackbench threads would occurr in the right
conditions:
o with TIF_NEED_RESCHED_LAZY set
  o at the tick
  o at do_notify_resume()
o with TIF_NEED_RESCHED set  

[1]: https://lore.kernel.org/linux-rt-users/[e-mail]/

Cheers,
Valentin

Mark Rutland (1):
  arm64: enable PREEMPT_LAZY

 arch/arm64/Kconfig                   |  1 +
 arch/arm64/include/asm/thread_info.h | 16 +++++++++-------
 arch/arm64/kernel/entry-common.c     |  2 +-
 3 files changed, 11 insertions(+), 8 deletions(-)

--
2.43.0

Comments

Sebastian Andrzej Siewior March 5, 2025, 11:06 a.m. UTC | #1

On 2025-03-05 11:49:25 [+0100], Valentin Schneider wrote:
> From: Mark Rutland <[e-mail]>
> 
> For an architecture to enable CONFIG_ARCH_HAS_RESCHED_LAZY, two things are
> required:
> 1) Adding a TIF_NEED_RESCHED_LAZY flag definition
> 2) Checking for TIF_NEED_RESCHED_LAZY in the appropriate locations
> 
> 2) is handled in a generic manner by CONFIG_GENERIC_ENTRY, which isn't
> (yet) implemented for arm64. However, outside of core scheduler code,
> TIF_NEED_RESCHED_LAZY only needs to be checked on a kernel exit, meaning:
> o return/entry to userspace.
> o return/entry to guest.
> 
> The return/entry to a guest is all handled by xfer_to_guest_mode_handle_work()
> which already does the right thing, so it can be left as-is.
> 
> arm64 doesn't use common entry's exit_to_user_mode_prepare(), so update its
> return to user path to check for TIF_NEED_RESCHED_LAZY and call into
> schedule() accordingly.
> 
> Link: https://lore.kernel.org/linux-rt-users/[e-mail]/
> Link: https://lore.kernel.org/all/[e-mail]/
> Signed-off-by: Mark Rutland <[e-mail]>
> [testdrive, _TIF_WORK_MASK fixlet and changelog.]
> Signed-off-by: Mike Galbraith <[e-mail]>
> [Another round of testing; changelog faff]
> Signed-off-by: Valentin Schneider <[e-mail]>

I somehow assumed it was already merged. For what it is worth, this
looks like patch Mike posted in
	https://lore.kernel.org/[e-mail]

and tested 

Reviewed-by: Sebastian Andrzej Siewior <[e-mail]>

Sebastian

Phil Auld March 5, 2025, 2:03 p.m. UTC | #2

On Wed, Mar 05, 2025 at 11:49:24AM +0100 Valentin Schneider wrote:
> Hey folks,
> 
> This is a resubmission of Mark's patch [1]. I didn't end up changing anything
> other than fluffing up the changelog.
> 
> I ran this under an arm64 VM, added some trace_printk's and checked that
> sched_switch's between runnable hackbench threads would occurr in the right
> conditions:
> o with TIF_NEED_RESCHED_LAZY set
>   o at the tick
>   o at do_notify_resume()
> o with TIF_NEED_RESCHED set  
> 
> [1]: https://lore.kernel.org/linux-rt-users/[e-mail]/
>

Thanks!  I was looking for this the other day :) 


Cheers,
Phil

> Cheers,
> Valentin
> 
> Mark Rutland (1):
>   arm64: enable PREEMPT_LAZY
> 
>  arch/arm64/Kconfig                   |  1 +
>  arch/arm64/include/asm/thread_info.h | 16 +++++++++-------
>  arch/arm64/kernel/entry-common.c     |  2 +-
>  3 files changed, 11 insertions(+), 8 deletions(-)
> 
> --
> 2.43.0
> 

--