Re: Another preempt folding issue?

From: Peter Zijlstra <peterz@infradead.org>
To: Stefan Bader <stefan.bader@canonical.com>
Cc: Borislav Petkov <bp@alien8.de>,
	Paolo Bonzini <pbonzini@redhat.com>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	kvm@vger.kernel.org, Marcelo Tosatti <mtosatti@redhat.com>,
	MASAO TAKAHASHI <masao-takahashi@kanno.co.jp>,
	Joerg Roedel <joro@8bytes.org>
Subject: Re: Another preempt folding issue?
Date: Fri, 14 Feb 2014 12:49:26 +0100	[thread overview]
Message-ID: <20140214114926.GB15586@twins.programming.kicks-ass.net> (raw)
In-Reply-To: <52FDFCFA.906@canonical.com>

On Fri, Feb 14, 2014 at 12:24:42PM +0100, Stefan Bader wrote:
> Oh and one thing I was wondering. Not sure I do understand it right... When
> initially converting to percpu counts, you changed the 32bit assembly like that:
> 
> --- a/arch/x86/kernel/entry_32.S
> +++ b/arch/x86/kernel/entry_32.S
> @@ -362,12 +362,9 @@ END(ret_from_exception)
>  #ifdef CONFIG_PREEMPT
>  ENTRY(resume_kernel)
>         DISABLE_INTERRUPTS(CLBR_ANY)
> -       cmpl $0,TI_preempt_count(%ebp)  # non-zero preempt_count ?
> -       jnz restore_all
>  need_resched:
> -       movl TI_flags(%ebp), %ecx       # need_resched set ?
> -       testb $_TIF_NEED_RESCHED, %cl
> -       jz restore_all
> +       cmpl $0,PER_CPU_VAR(__preempt_count)
> +       jnz restore_all
>         testl $X86_EFLAGS_IF,PT_EFLAGS(%esp)    # interrupts off (exception path
>         jz restore_all
>         call preempt_schedule_irq
> 
> This seems to say if preempt_count was 0 then then if the thread flag was set
> and interrupts were not off(?) it would do a preempt ipi and then come back to
> re-check the thread flag.

No not an IPI; it would reschedule.

So the old code:

  if preempt_count != 0; continue out
  if !TIF_NEED_RESCHED; continue out
  if IRQs-off in calling context; continue out
  preempt_schedule_irq

The new code:

  if preempt_count != 0; continue out
  if IRQs-off in calling context; continue out
  preempt_schedule_irq

> This would now be if preempt_count is 0 only... and I wonder whether that would
> change from doing that loop...

We can do away with the TIF_NEED_RESCHED test because that state is
folded into the preempt_count by means of PREEMPT_NEED_RESCHED.