Re: [kernel-hardening] [PATCH 09/38] usercopy: Mark kmalloc caches as usercopy caches

[Kees Cook]

On Tue, Nov 12, 2019 at 08:17:57AM +0100, Jiri Slaby wrote:
> On 11. 01. 18, 3:02, Kees Cook wrote:
> > From: David Windsor <dave@xxxxxxxxxxxx>
> > 
> > Mark the kmalloc slab caches as entirely whitelisted. These caches
> > are frequently used to fulfill kernel allocations that contain data
> > to be copied to/from userspace. Internal-only uses are also common,
> > but are scattered in the kernel. For now, mark all the kmalloc caches
> > as whitelisted.
> > 
> > This patch is modified from Brad Spengler/PaX Team's PAX_USERCOPY
> > whitelisting code in the last public patch of grsecurity/PaX based on my
> > understanding of the code. Changes or omissions from the original code are
> > mine and don't reflect the original grsecurity/PaX code.
> > 
> > Signed-off-by: David Windsor <dave@xxxxxxxxxxxx>
> > [kees: merged in moved kmalloc hunks, adjust commit log]
> > Cc: Pekka Enberg <penberg@xxxxxxxxxx>
> > Cc: David Rientjes <rientjes@xxxxxxxxxx>
> > Cc: Joonsoo Kim <iamjoonsoo.kim@xxxxxxx>
> > Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> > Cc: linux-mm@xxxxxxxxx
> > Cc: linux-xfs@xxxxxxxxxxxxxxx
> > Signed-off-by: Kees Cook <keescook@xxxxxxxxxxxx>
> > Acked-by: Christoph Lameter <cl@xxxxxxxxx>
> > ---
> >  mm/slab.c        |  3 ++-
> >  mm/slab.h        |  3 ++-
> >  mm/slab_common.c | 10 ++++++----
> >  3 files changed, 10 insertions(+), 6 deletions(-)
> > 
> > diff --git a/mm/slab.c b/mm/slab.c
> > index b9b0df620bb9..dd367fe17a4e 100644
> > --- a/mm/slab.c
> > +++ b/mm/slab.c
> ...
> > @@ -1098,7 +1099,8 @@ void __init setup_kmalloc_cache_index_table(void)
> >  static void __init new_kmalloc_cache(int idx, slab_flags_t flags)
> >  {
> >  	kmalloc_caches[idx] = create_kmalloc_cache(kmalloc_info[idx].name,
> > -					kmalloc_info[idx].size, flags);
> > +					kmalloc_info[idx].size, flags, 0,
> > +					kmalloc_info[idx].size);
> >  }
> >  
> >  /*
> > @@ -1139,7 +1141,7 @@ void __init create_kmalloc_caches(slab_flags_t flags)
> >  
> >  			BUG_ON(!n);
> >  			kmalloc_dma_caches[i] = create_kmalloc_cache(n,
> > -				size, SLAB_CACHE_DMA | flags);
> > +				size, SLAB_CACHE_DMA | flags, 0, 0);
> 
> Hi,
> 
> was there any (undocumented) reason NOT to mark DMA caches as usercopy?
> 
> We are seeing this on s390x:
> 
> > usercopy: Kernel memory overwrite attempt detected to SLUB object
> 'dma-kmalloc-1k' (offset 0, size 11)!
> > ------------[ cut here ]------------
> > kernel BUG at mm/usercopy.c:99!

Interesting! I believe the rationale was that if the region is used for
DMA, allowing direct access to it from userspace could be prone to
races.

> See:
> https://bugzilla.suse.com/show_bug.cgi?id=1156053

For context from the bug, the trace is:

(<0000000000386c5a> usercopy_abort+0xa2/0xa8) 
 <000000000036097a> __check_heap_object+0x11a/0x120  
 <0000000000386b3a> __check_object_size+0x18a/0x208  
 <000000000079b4ba> skb_copy_datagram_from_iter+0x62/0x240  
 <000003ff804edd5c> iucv_sock_sendmsg+0x1fc/0x858 Ýaf_iucv¨  
 <0000000000785894> sock_sendmsg+0x54/0x90  
 <0000000000785944> sock_write_iter+0x74/0xa0  
 <000000000038a3f0> new_sync_write+0x110/0x180  
 <000000000038d42e> vfs_write+0xa6/0x1d0  
 <000000000038d748> ksys_write+0x60/0xe8  
 <000000000096a660> system_call+0xdc/0x2e0  

I know Al worked on fixing up usercopy checking for iters. I wonder if
there is redundant checking happening here? i.e. haven't iters already
done object size verifications, so they're not needed during iter copy
helpers?

> This indeed fixes it:
> --- a/mm/slab_common.c
> +++ b/mm/slab_common.c
> @@ -1290,7 +1290,8 @@ void __init create_kmalloc_caches(slab_flags_t flags)
>                         kmalloc_caches[KMALLOC_DMA][i] =
> create_kmalloc_cache(
>                                 kmalloc_info[i].name[KMALLOC_DMA],
>                                 kmalloc_info[i].size,
> -                               SLAB_CACHE_DMA | flags, 0, 0);
> +                               SLAB_CACHE_DMA | flags, 0,
> +                               kmalloc_info[i].size);
>                 }
>         }
>  #endif

How is iucv the only network protocol that has run into this? Do others
use a bounce buffer?

-- 
Kees Cook