[dm-devel] [PATCH] slab: implement kmalloc guard

[Mikulas Patocka]

On Mon, 15 Sep 2014, Joonsoo Kim wrote:

> On Thu, Sep 11, 2014 at 10:32:52PM -0400, Mikulas Patocka wrote:
> > 
> > 
> > On Mon, 8 Sep 2014, Christoph Lameter wrote:
> > 
> > > On Mon, 8 Sep 2014, Mikulas Patocka wrote:
> > > 
> > > > I don't know what you mean. If someone allocates 10000 objects with sizes
> > > > from 1 to 10000, you can't have 10000 slab caches - you can't have a slab
> > > > cache for each used size. Also - you can't create a slab cache in
> > > > interrupt context.
> > > 
> > > Oh you can create them up front on bootup. And I think only the small
> > > sizes matter. Allocations >=8K are pushed to the page allocator anyways.
> > 
> > Only for SLUB. For SLAB, large allocations are still use SLAB caches up to 
> > 4M. But anyway - having 8K preallocated slab caches is too much.
> > 
> > If you want to integrate this patch into the slab/slub subsystem, a better 
> > solution would be to store the exact size requested with kmalloc along the 
> > slab/slub object itself (before the preceding redzone). But it would 
> > result in duplicating the work - you'd have to repeat the logic in this 
> > patch three times - once for slab, once for slub and once for 
> > kmalloc_large/kmalloc_large_node.
> > 
> > I don't know if it would be better than this patch.
> 
> Hello,
> 
> Out of bound write could be detected by kernel address asanitizer(KASan).
> See following link.
> 
> https://lkml.org/lkml/2014/9/10/441
> 
> Although this patch also looks good to me, I think that KASan is
> better than this, because it could detect out of bound write and
> has more features for debugging.
> 
> Thanks.

Surely, KAsan detects more bugs. But it has also high overhead. The 
overhead of kmalloc guard is very low.

The kmalloc guard already helped to find one previously unknown bug: 
http://lkml.iu.edu/hypermail/linux/kernel/1409.1/02325.html

Mikulas