Re: [PATCH v2 0/5] x86 switch_mm uninlining and IRQ improvements

[Borislav Petkov <bp@alien8.de>]

On Tue, Apr 26, 2016 at 09:39:04AM -0700, Andy Lutomirski wrote:
> Hi all-
> 
> I've been playing with context switching lately, and I'm going to start
> sending out some of the patches that should be mostly self-contained and
> ready for -tip.
> 
> Here's a little batch to start improving switch_mm.  It uninlines it
> and makes it run with IRQs off.  (AFAICT everyone who's modified it
> thought it ran with IRQs off, but that's not always the case.  I
> don't know of any bugs that this fixes, but it'll be needed for PCID
> to avoid introducing really nasty races.)
> 
> This may also help a bit with FSGSBASE -- not sure yet.  It certainly
> won't hurt.
> 
> It contains a trivial off-topic ARM patch to avoid breaking the build.
> 
> Andy Lutomirski (5):
>   arm: Include linux/preempt.h from asm/mmu_context.h
>   sched: Add switch_mm_irqs_off and use it in the scheduler
>   x86/mm: Build arch/x86/mm/tlb.c even on !SMP
>   x86/mm: Uninline switch_mm
>   x86/mm: Turn off IRQs in switch_mm
> 
>  arch/arm/include/asm/mmu_context.h |   1 +
>  arch/x86/include/asm/mmu_context.h | 101 ++------------------------------
>  arch/x86/mm/Makefile               |   3 +-
>  arch/x86/mm/tlb.c                  | 116 +++++++++++++++++++++++++++++++++++++
>  include/linux/mmu_context.h        |   7 +++
>  kernel/sched/core.c                |   6 +-
>  6 files changed, 133 insertions(+), 101 deletions(-)

Reviewed-by: Borislav Petkov <bp@suse.de>

There was a concern that maybe disabling IRQs in
exec_mmap->activate_mm()->switch_mm() would be a little bit of a
slowdown but that's not a hot path anyway.

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.