Re: [PATCH] fs/befs/linuxvfs.c: need signed cast for variable 'block'

[Kees Cook]

On Thu, Oct 31, 2013 at 12:06 PM, Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote:
> On Thu, Oct 31, 2013 at 09:53:59AM -0700, Kees Cook wrote:
>
>> If block (type sector_t) is unsigned, we shouldn't cast it signed.
>> This entire code path should be removed. What is BEFS's expected
>> maximum block size? (Looks like even befs_blocknr_t is u64, so nothing
>> seems trivially in danger of wrapping.) I would also note that all the
>> format strings are wrong too (%ld instead of %lu).
>
> FWIW, this
>         res = befs_fblock2brun(sb, ds, block, &run);
>         if (res != BEFS_OK) {
>                 befs_error(sb,
>                            "<--- befs_get_block() for inode %lu, block "
>                            "%ld ERROR", inode->i_ino, block);
>                 return -EFBIG;
>         }
> also looks wrong - ioctl(..., FIBMAP, ...) shouldn't be able to spew
> printks on a valid fs and hitting it with block number greater than
> file length will, AFAICS, trigger that.
>
> I agree that this code needs fixing, but just making gcc STFU about the
> comparison would only serve to hide the problem.  Anybody familiar with
> befs or willing to learn it?

Agreed. MAINTAINERS shows it as orphaned. Perhaps it should be moved
into staging?

-Kees

-- 
Kees Cook
Chrome OS Security
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/