Linux Virtualization - Dec 2015 - [PATCH 08/34] asm-generic: smp_store_mb should use smp_mb

On Wednesday 30 December 2015 15:24:47 Michael S. Tsirkin
wrote:
> asm-generic variant of smp_store_mb() calls mb() which is stronger
> than implied by both the name and the documentation.
> 
> smp_store_mb is only used by core kernel code at the moment, so
> we know no one mis-uses it for an MMIO barrier.
> Make it call smp_mb consistently before some arch-specific
> code uses it as such by mistake.
> 
> Signed-off-by: Michael S. Tsirkin <mst at redhat.com>
> ---
>  include/asm-generic/barrier.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/include/asm-generic/barrier.h b/include/asm-generic/barrier.h
> index 538f8d1..987b2e0 100644
> --- a/include/asm-generic/barrier.h
> +++ b/include/asm-generic/barrier.h
> @@ -93,7 +93,7 @@
>  #endif	/* CONFIG_SMP */
>  
>  #ifndef smp_store_mb
> -#define smp_store_mb(var, value)  do { WRITE_ONCE(var, value); mb(); }
while (0)
> +#define smp_store_mb(var, value)  do { WRITE_ONCE(var, value); smp_mb(); }
while (0)
>  #endif
>  
>  #ifndef smp_mb__before_atomic
> 
The same patch is already in the tip tree scheduled for 4.5 as d5a73cadf3fd
("lcoking/barriers, arch: Use smp barriers in smp_store_release()").

I think you can drop your version.

	arnd

On Wed, Dec 30, 2015 at 02:44:21PM +0100, Arnd Bergmann
wrote:
> On Wednesday 30 December 2015 15:24:47 Michael S. Tsirkin wrote:
> > asm-generic variant of smp_store_mb() calls mb() which is stronger
> > than implied by both the name and the documentation.
> > 
> > smp_store_mb is only used by core kernel code at the moment, so
> > we know no one mis-uses it for an MMIO barrier.
> > Make it call smp_mb consistently before some arch-specific
> > code uses it as such by mistake.
> > 
> > Signed-off-by: Michael S. Tsirkin <mst at redhat.com>
> > ---
> >  include/asm-generic/barrier.h | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/include/asm-generic/barrier.h
b/include/asm-generic/barrier.h
> > index 538f8d1..987b2e0 100644
> > --- a/include/asm-generic/barrier.h
> > +++ b/include/asm-generic/barrier.h
> > @@ -93,7 +93,7 @@
> >  #endif	/* CONFIG_SMP */
> >  
> >  #ifndef smp_store_mb
> > -#define smp_store_mb(var, value)  do { WRITE_ONCE(var, value); mb();
} while (0)
> > +#define smp_store_mb(var, value)  do { WRITE_ONCE(var, value);
smp_mb(); } while (0)
> >  #endif
> >  
> >  #ifndef smp_mb__before_atomic
> > 
> 
> The same patch is already in the tip tree scheduled for 4.5 as d5a73cadf3fd
> ("lcoking/barriers, arch: Use smp barriers in
smp_store_release()").
Sorry which tree do you mean exactly?
> I think you can drop your version.
> 
> 	arnd
Will drop mine, thanks.
I kind of dislike that if I just drop it, some arches will temporarily
regress to a slower implementation.
I think I can just cherry-pick d5a73cadf3fd into my tree: git
normally figures such duplicates out nicely.
Does this sound good?

-- 
MST

On Wednesday 30 December 2015 22:30:38 Michael S. Tsirkin
wrote:
> On Wed, Dec 30, 2015 at 02:44:21PM +0100, Arnd Bergmann wrote:
> > On Wednesday 30 December 2015 15:24:47 Michael S. Tsirkin wrote:
> > >  #ifndef smp_store_mb
> > > -#define smp_store_mb(var, value)  do { WRITE_ONCE(var, value);
mb(); } while (0)
> > > +#define smp_store_mb(var, value)  do { WRITE_ONCE(var, value);
smp_mb(); } while (0)
> > >  #endif
> > >  
> > >  #ifndef smp_mb__before_atomic
> > > 
> > 
> > The same patch is already in the tip tree scheduled for 4.5 as
d5a73cadf3fd
> > ("lcoking/barriers, arch: Use smp barriers in
smp_store_release()").
> 
> Sorry which tree do you mean exactly?
$ git log --ancestry-path --oneline --merges d5a73cadf3fd..next/master | tail -n
17
cb17a685bed6 Merge remote-tracking branch 'tip/auto-latest'
f29c2e03f0b3 Merge branch 'x86/urgent'
8cd6990bf71d Merge branch 'x86/platform'
0541d92a5eb4 Merge branch 'x86/mm'
aa7c8013c8c0 Merge branch 'x86/fpu'
fcc9a1bd013c Merge branch 'x86/efi'
e74ef3f60886 Merge branch 'x86/cpu'
44a4f0063508 Merge branch 'x86/cleanups'
28c814578fcf Merge branch 'x86/cache'
d74ff99dada8 Merge branch 'x86/boot'
db3c55380b10 Merge branch 'x86/asm'
7cd91b91da20 Merge branch 'x86/apic'
7bfc343947e6 Merge branch 'timers/core'
1720bbcb66d1 Merge branch 'sched/core'
af9a59f26764 Merge branch 'ras/core'
984b85eca78d Merge branch 'perf/core'
d2b22d438aab Merge branch 'locking/core'

$ grep auto-latest Next/Trees 
tip             git    
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git#auto-latest

So locking/core of tip.git has the patch and gets merged into linux-next
through auto-latest in tip.git.
> > I think you can drop your version.
> > 
> >       arnd
> 
> Will drop mine, thanks.
> I kind of dislike that if I just drop it, some arches will temporarily
> regress to a slower implementation.
> I think I can just cherry-pick d5a73cadf3fd into my tree: git
> normally figures such duplicates out nicely.
> Does this sound good?
I don't think there is a perfect solution, you can either cherry-pick
it and get a duplicate commit in the git history, or you merge in the
whole locking/core branch from tip.

I'd say ask Ingo/PeterZ/Davidlohr which way they prefer.

	Arnd