Re: [RFC][PATCH 2/4] slab: Detect negative size values and saturate

[Kees Cook]

On Tue, Jul 09, 2024 at 08:57:55AM +0200, Przemek Kitszel wrote:
> On 7/8/24 21:18, Kees Cook wrote:
> > The allocator will already reject giant sizes seen from negative size
> > arguments, so this commit mainly services as an example for initial
> > type-based filtering. The size argument is checked for negative values
> > in signed arguments, saturating any if found instead of passing them on.
> > 
> > For example, now the size is checked:
> > 
> > Before:
> > 				/* %rdi unchecked */
> >   1eb:   be c0 0c 00 00          mov    $0xcc0,%esi
> >   1f0:   e8 00 00 00 00          call   1f5 <do_SLAB_NEGATIVE+0x15>
> >                          1f1: R_X86_64_PLT32 __kmalloc_noprof-0x4
> > 
> > After:
> >   6d0:   48 63 c7                movslq %edi,%rax
> >   6d3:   85 ff                   test   %edi,%edi
> >   6d5:   be c0 0c 00 00          mov    $0xcc0,%esi
> >   6da:   48 c7 c2 ff ff ff ff    mov    $0xffffffffffffffff,%rdx
> >   6e1:   48 0f 49 d0             cmovns %rax,%rdx
> >   6e5:   48 89 d7                mov    %rdx,%rdi
> >   6e8:   e8 00 00 00 00          call   6ed <do_SLAB_NEGATIVE+0x1d>
> >                          6e9: R_X86_64_PLT32     __kmalloc_noprof-0x4
> > 
> > Signed-off-by: Kees Cook <kees@xxxxxxxxxx>
> > ---
> > Cc: Christoph Lameter <cl@xxxxxxxxx>
> > Cc: Pekka Enberg <penberg@xxxxxxxxxx>
> > Cc: David Rientjes <rientjes@xxxxxxxxxx>
> > Cc: Joonsoo Kim <iamjoonsoo.kim@xxxxxxx>
> > Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> > Cc: Vlastimil Babka <vbabka@xxxxxxx>
> > Cc: Roman Gushchin <roman.gushchin@xxxxxxxxx>
> > Cc: Hyeonggon Yoo <42.hyeyoo@xxxxxxxxx>
> > Cc: linux-mm@xxxxxxxxx
> > ---
> >   include/linux/slab.h | 19 ++++++++++++++++++-
> >   1 file changed, 18 insertions(+), 1 deletion(-)
> > 
> > diff --git a/include/linux/slab.h b/include/linux/slab.h
> > index d99afce36098..7353756cbec6 100644
> > --- a/include/linux/slab.h
> > +++ b/include/linux/slab.h
> > @@ -684,7 +684,24 @@ static __always_inline __alloc_size(1) void *kmalloc_noprof(size_t size, gfp_t f
> >   	}
> >   	return __kmalloc_noprof(size, flags);
> >   }
> > -#define kmalloc(...)				alloc_hooks(kmalloc_noprof(__VA_ARGS__))
> > +#define kmalloc_sized(...)			alloc_hooks(kmalloc_noprof(__VA_ARGS__))
> > +
> > +#define __size_force_positive(x)				\
> > +	({							\
> > +		typeof(__force_integral_expr(x)) __forced_val =	\
> > +			__force_integral_expr(x);		\
> > +		__forced_val < 0 ? SIZE_MAX : __forced_val;	\
> > +	})
> > +
> > +#define kmalloc(p, gfp)		_Generic((p),    \
> > +	unsigned char:  kmalloc_sized(__force_integral_expr(p), gfp), \
> > +	unsigned short: kmalloc_sized(__force_integral_expr(p), gfp), \
> > +	unsigned int:   kmalloc_sized(__force_integral_expr(p), gfp), \
> > +	unsigned long:  kmalloc_sized(__force_integral_expr(p), gfp), \
> > +	signed char:    kmalloc_sized(__size_force_positive(p), gfp), \
> > +	signed short:   kmalloc_sized(__size_force_positive(p), gfp), \
> > +	signed int:     kmalloc_sized(__size_force_positive(p), gfp), \
> > +	signed long:    kmalloc_sized(__size_force_positive(p), gfp))
> 
> I like this idea and series very much, thank you!

Thanks!

> What about bool?
> What about long long?

Ah yes, I will add these. LKP also found a weird one (a bitfield!) that
I'm fixing at the source:
https://lore.kernel.org/lkml/20240709154953.work.953-kees@xxxxxxxxxx/

-- 
Kees Cook