Re: [PATCH v2 1/3] kasan, arm64: use ARCH_SLAB_MINALIGN instead of manual aligning

Re: [PATCH v2 1/3] kasan, arm64: use ARCH_SLAB_MINALIGN instead of manual aligning

[Andrew Morton]

On Wed,  2 Jan 2019 18:36:06 +0100 Andrey Konovalov <andreyknvl@google.com> wrote:

> Instead of changing cache->align to be aligned to KASAN_SHADOW_SCALE_SIZE
> in kasan_cache_create() we can reuse the ARCH_SLAB_MINALIGN macro.
> 
> ...
>
> --- a/arch/arm64/include/asm/kasan.h
> +++ b/arch/arm64/include/asm/kasan.h
> @@ -36,6 +36,10 @@
>  #define KASAN_SHADOW_OFFSET     (KASAN_SHADOW_END - (1ULL << \
>  					(64 - KASAN_SHADOW_SCALE_SHIFT)))
>  
> +#ifdef CONFIG_KASAN_SW_TAGS
> +#define ARCH_SLAB_MINALIGN	(1ULL << KASAN_SHADOW_SCALE_SHIFT)
> +#endif
> +
>  void kasan_init(void);
>  void kasan_copy_shadow(pgd_t *pgdir);
>  asmlinkage void kasan_early_init(void);
> diff --git a/include/linux/slab.h b/include/linux/slab.h
> index 11b45f7ae405..d87f913ab4e8 100644
> --- a/include/linux/slab.h
> +++ b/include/linux/slab.h
> @@ -16,6 +16,7 @@
>  #include <linux/overflow.h>
>  #include <linux/types.h>
>  #include <linux/workqueue.h>
> +#include <linux/kasan.h>
>  

This still seems unadvisable.  Like other architectures, arm defines
ARCH_SLAB_MINALIGN in arch/arm/include/asm/cache.h. 
arch/arm/include/asm64/cache.h doesn't define ARCH_SLAB_MINALIGN
afaict.

If arch/arm/include/asm64/cache.h later gets a definition of
ARCH_SLAB_MINALIGN then we again face the risk that different .c files
will see different values of ARCH_SLAB_MINALIGN depending on which
headers they include.

So what to say about this?  The architecture's ARCH_SLAB_MINALIGN
should be defined in the architecture's cache.h, end of story.  Not in
slab.h, not in kasan.h.

Re: [PATCH v2 1/3] kasan, arm64: use ARCH_SLAB_MINALIGN instead of manual aligning

[Will Deacon]

On Wed, Jan 02, 2019 at 12:14:36PM -0800, Andrew Morton wrote:
> On Wed,  2 Jan 2019 18:36:06 +0100 Andrey Konovalov <andreyknvl@google.com> wrote:
> 
> > Instead of changing cache->align to be aligned to KASAN_SHADOW_SCALE_SIZE
> > in kasan_cache_create() we can reuse the ARCH_SLAB_MINALIGN macro.
> > 
> > ...
> >
> > --- a/arch/arm64/include/asm/kasan.h
> > +++ b/arch/arm64/include/asm/kasan.h
> > @@ -36,6 +36,10 @@
> >  #define KASAN_SHADOW_OFFSET     (KASAN_SHADOW_END - (1ULL << \
> >  					(64 - KASAN_SHADOW_SCALE_SHIFT)))
> >  
> > +#ifdef CONFIG_KASAN_SW_TAGS
> > +#define ARCH_SLAB_MINALIGN	(1ULL << KASAN_SHADOW_SCALE_SHIFT)
> > +#endif
> > +
> >  void kasan_init(void);
> >  void kasan_copy_shadow(pgd_t *pgdir);
> >  asmlinkage void kasan_early_init(void);
> > diff --git a/include/linux/slab.h b/include/linux/slab.h
> > index 11b45f7ae405..d87f913ab4e8 100644
> > --- a/include/linux/slab.h
> > +++ b/include/linux/slab.h
> > @@ -16,6 +16,7 @@
> >  #include <linux/overflow.h>
> >  #include <linux/types.h>
> >  #include <linux/workqueue.h>
> > +#include <linux/kasan.h>
> >  
> 
> This still seems unadvisable.  Like other architectures, arm defines
> ARCH_SLAB_MINALIGN in arch/arm/include/asm/cache.h. 
> arch/arm/include/asm64/cache.h doesn't define ARCH_SLAB_MINALIGN
> afaict.
> 
> If arch/arm/include/asm64/cache.h later gets a definition of
> ARCH_SLAB_MINALIGN then we again face the risk that different .c files
> will see different values of ARCH_SLAB_MINALIGN depending on which
> headers they include.
> 
> So what to say about this?  The architecture's ARCH_SLAB_MINALIGN
> should be defined in the architecture's cache.h, end of story.  Not in
> slab.h, not in kasan.h.

Agreed. Also, as far as I can tell, this patch isn't actually a fix (unlike
the other two in this series) so it should be harmless to drop it for now.

Will

Re: [PATCH v2 1/3] kasan, arm64: use ARCH_SLAB_MINALIGN instead of manual aligning

[Andrey Konovalov]

On Wed, Jan 2, 2019 at 9:14 PM Andrew Morton <akpm@linux-foundation.org> wrote:
>
> On Wed,  2 Jan 2019 18:36:06 +0100 Andrey Konovalov <andreyknvl@google.com> wrote:
>
> > Instead of changing cache->align to be aligned to KASAN_SHADOW_SCALE_SIZE
> > in kasan_cache_create() we can reuse the ARCH_SLAB_MINALIGN macro.
> >
> > ...
> >
> > --- a/arch/arm64/include/asm/kasan.h
> > +++ b/arch/arm64/include/asm/kasan.h
> > @@ -36,6 +36,10 @@
> >  #define KASAN_SHADOW_OFFSET     (KASAN_SHADOW_END - (1ULL << \
> >                                       (64 - KASAN_SHADOW_SCALE_SHIFT)))
> >
> > +#ifdef CONFIG_KASAN_SW_TAGS
> > +#define ARCH_SLAB_MINALIGN   (1ULL << KASAN_SHADOW_SCALE_SHIFT)
> > +#endif
> > +
> >  void kasan_init(void);
> >  void kasan_copy_shadow(pgd_t *pgdir);
> >  asmlinkage void kasan_early_init(void);
> > diff --git a/include/linux/slab.h b/include/linux/slab.h
> > index 11b45f7ae405..d87f913ab4e8 100644
> > --- a/include/linux/slab.h
> > +++ b/include/linux/slab.h
> > @@ -16,6 +16,7 @@
> >  #include <linux/overflow.h>
> >  #include <linux/types.h>
> >  #include <linux/workqueue.h>
> > +#include <linux/kasan.h>
> >
>
> This still seems unadvisable.  Like other architectures, arm defines
> ARCH_SLAB_MINALIGN in arch/arm/include/asm/cache.h.
> arch/arm/include/asm64/cache.h doesn't define ARCH_SLAB_MINALIGN
> afaict.
>
> If arch/arm/include/asm64/cache.h later gets a definition of
> ARCH_SLAB_MINALIGN then we again face the risk that different .c files
> will see different values of ARCH_SLAB_MINALIGN depending on which
> headers they include.
>
> So what to say about this?  The architecture's ARCH_SLAB_MINALIGN
> should be defined in the architecture's cache.h, end of story.  Not in
> slab.h, not in kasan.h.

Done in v3, thanks!

>
>
> --
> You received this message because you are subscribed to the Google Groups "kasan-dev" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to kasan-dev+unsubscribe@googlegroups.com.
> To post to this group, send email to kasan-dev@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/kasan-dev/20190102121436.5c2b72d1b0ec49affadc9692%40linux-foundation.org.
> For more options, visit https://groups.google.com/d/optout.