Re: [PATCH] uaccess: Add minimum bounds check on kernel buffer size

[Kees Cook]

On Fri, Feb 03, 2023 at 10:23:13PM +0100, Arnd Bergmann wrote:
> On Fri, Feb 3, 2023, at 20:35, Kees Cook wrote:
> > While there is logic about the difference between ksize and usize,
> > copy_struct_from_user() didn't check the size of the destination buffer
> > (when it was known) against ksize. Add this check so there is an upper
> > bounds check on the possible memset() call, otherwise lower bounds
> > checks made by callers will trigger bounds warnings under -Warray-bounds.
> > Seen under GCC 13:
> >
> > In function 'copy_struct_from_user',
> >     inlined from 'iommufd_fops_ioctl' at
> > ../drivers/iommu/iommufd/main.c:333:8:
> > ../include/linux/fortify-string.h:59:33: warning: '__builtin_memset' 
> > offset [57, 4294967294] is out of the bounds [0, 56] of object 'buf' 
> > with type 'union ucmd_buffer' [-Warray-bounds=]
> >    59 | #define __underlying_memset     __builtin_memset
> >       |                                 ^
> > ../include/linux/fortify-string.h:453:9: note: in expansion of macro 
> > '__underlying_memset'
> >   453 |         __underlying_memset(p, c, __fortify_size); \
> >       |         ^~~~~~~~~~~~~~~~~~~
> > ../include/linux/fortify-string.h:461:25: note: in expansion of macro 
> > '__fortify_memset_chk'
> >   461 | #define memset(p, c, s) __fortify_memset_chk(p, c, s, \
> >       |                         ^~~~~~~~~~~~~~~~~~~~
> > ../include/linux/uaccess.h:334:17: note: in expansion of macro 'memset'
> >   334 |                 memset(dst + size, 0, rest);
> >       |                 ^~~~~~
> > ../drivers/iommu/iommufd/main.c: In function 'iommufd_fops_ioctl':
> > ../drivers/iommu/iommufd/main.c:311:27: note: 'buf' declared here
> >   311 |         union ucmd_buffer buf;
> >       |                           ^~~
> >
> 
> Hi Kees,
> 
> I started building with gcc-13.0.1 myself but ran into a lot of
> other -Warray-bounds warnings in randconfig builds, so I ended up
> turning it off once more with CONFIG_CC_NO_ARRAY_BOUNDS in order
> to keep building without warnings.

Understood. AFAIK, all the open bugs I (and you) filed with GCC 13 have
been fixed related to -Warray-bounds. The most recent was the misbehavior
between CONFIG_UBSAN_SHIFT and -Warray-bounds. (Though the shift checking
still exposes some warnings since it introduces an implicit bounds check
on the shift variable, but they're not _wrong_ any more.)

> Is there anything else I need to do to get to the point of
> just addressing actual issues instead of false positives?
> Do you already have a patch series for fixing the others?

I've been working through the list that I see when building with
-Warray-bounds and -fstrict-flex-arrays=3. Some are real bugs, as usual.

> > diff --git a/include/linux/uaccess.h b/include/linux/uaccess.h
> > index afb18f198843..ab9728138ad6 100644
> > --- a/include/linux/uaccess.h
> > +++ b/include/linux/uaccess.h
> > @@ -329,6 +329,10 @@ copy_struct_from_user(void *dst, size_t ksize, 
> > const void __user *src,
> >  	size_t size = min(ksize, usize);
> >  	size_t rest = max(ksize, usize) - size;
> > 
> > +	/* Double check if ksize is larger than a known object size. */
> > +	if (WARN_ON_ONCE(ksize > __builtin_object_size(dst, 1)))
> > +		return -E2BIG;
> > +
> 
> WARN_ON_ONCE() may be a little expensive since that adds two
> comparisons and a static variable to each copy, but it's probably
> fine. 

Yeah. IMO, copy_struct_from_user() is not fast path and having better
bounds checking when coming from userspace is well worth it.

-- 
Kees Cook