Re: [Qemu-devel] [PATCH 09/12] crypto: fix stack-buffer-overflow error

[Daniel P. Berrange]

On Fri, Dec 08, 2017 at 01:58:22AM +0100, Marc-André Lureau wrote:
> ASAN complains about:
> 
> ==8856==ERROR: AddressSanitizer: stack-buffer-overflow on address 
> 0x7ffd8a1fe168 at pc 0x561136cb4451 bp 0x7ffd8a1fe130 sp 0x7ffd8a1fd8e0
> READ of size 16 at 0x7ffd8a1fe168 thread T0
>     #0 0x561136cb4450 in __asan_memcpy 
> (/home/elmarco/src/qq/build/tests/test-crypto-ivgen+0x110450)
>     #1 0x561136d2a6a7 in qcrypto_ivgen_essiv_calculate 
> /home/elmarco/src/qq/crypto/ivgen-essiv.c:83:5
>     #2 0x561136d29af8 in qcrypto_ivgen_calculate 
> /home/elmarco/src/qq/crypto/ivgen.c:72:12
>     #3 0x561136d07c8e in test_ivgen 
> /home/elmarco/src/qq/tests/test-crypto-ivgen.c:148:5
>     #4 0x7f77772c3b04 in test_case_run 
> /home/elmarco/src/gnome/glib/builddir/../glib/gtestutils.c:2237
>     #5 0x7f77772c3ec4 in g_test_run_suite_internal 
> /home/elmarco/src/gnome/glib/builddir/../glib/gtestutils.c:2321
>     #6 0x7f77772c3f6d in g_test_run_suite_internal 
> /home/elmarco/src/gnome/glib/builddir/../glib/gtestutils.c:2333
>     #7 0x7f77772c3f6d in g_test_run_suite_internal 
> /home/elmarco/src/gnome/glib/builddir/../glib/gtestutils.c:2333
>     #8 0x7f77772c3f6d in g_test_run_suite_internal 
> /home/elmarco/src/gnome/glib/builddir/../glib/gtestutils.c:2333
>     #9 0x7f77772c4184 in g_test_run_suite 
> /home/elmarco/src/gnome/glib/builddir/../glib/gtestutils.c:2408
>     #10 0x7f77772c2e0d in g_test_run 
> /home/elmarco/src/gnome/glib/builddir/../glib/gtestutils.c:1674
>     #11 0x561136d0799b in main 
> /home/elmarco/src/qq/tests/test-crypto-ivgen.c:173:12
>     #12 0x7f77756e6039 in __libc_start_main (/lib64/libc.so.6+0x21039)
>     #13 0x561136c13d89 in _start 
> (/home/elmarco/src/qq/build/tests/test-crypto-ivgen+0x6fd89)
> 
> Address 0x7ffd8a1fe168 is located in stack of thread T0 at offset 40 in frame
>     #0 0x561136d2a40f in qcrypto_ivgen_essiv_calculate 
> /home/elmarco/src/qq/crypto/ivgen-essiv.c:76
> 
>   This frame has 1 object(s):
>     [32, 40) 'sector.addr' <== Memory access at offset 40 overflows this 
> variable
> HINT: this may be a false positive if your program uses some custom stack 
> unwind mechanism or swapcontext
>       (longjmp and C++ exceptions *are* supported)
> SUMMARY: AddressSanitizer: stack-buffer-overflow 
> (/home/elmarco/src/qq/build/tests/test-crypto-ivgen+0x110450) in __asan_memcpy
> Shadow bytes around the buggy address:
>   0x100031437bd0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>   0x100031437be0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>   0x100031437bf0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>   0x100031437c00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>   0x100031437c10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
> =>0x100031437c20: 00 00 00 00 00 00 00 00 f1 f1 f1 f1 00[f3]f3 f3
>   0x100031437c30: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>   0x100031437c40: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>   0x100031437c50: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>   0x100031437c60: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>   0x100031437c70: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
> Shadow byte legend (one shadow byte represents 8 application bytes):
>   Addressable:           00
>   Partially addressable: 01 02 03 04 05 06 07
>   Heap left redzone:       fa
>   Freed heap region:       fd
>   Stack left redzone:      f1
>   Stack mid redzone:       f2
>   Stack right redzone:     f3
>   Stack after return:      f5
>   Stack use after scope:   f8
>   Global redzone:          f9
>   Global init order:       f6
>   Poisoned by user:        f7
>   Container overflow:      fc
>   Array cookie:            ac
>   Intra object redzone:    bb
>   ASan internal:           fe
>   Left alloca redzone:     ca
>   Right alloca redzone:    cb
> 
> It looks like the rest of the code copes with ndata being larger than
> sizeof(sector), so limit the memcpy() range.
> 
> Signed-off-by: Marc-André Lureau <address@hidden>
> ---
>  crypto/ivgen-essiv.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/crypto/ivgen-essiv.c b/crypto/ivgen-essiv.c
> index cba20bde6c..ad4d926c19 100644
> --- a/crypto/ivgen-essiv.c
> +++ b/crypto/ivgen-essiv.c
> @@ -79,7 +79,7 @@ static int qcrypto_ivgen_essiv_calculate(QCryptoIVGen 
> *ivgen,
>      uint8_t *data = g_new(uint8_t, ndata);
>  
>      sector = cpu_to_le64(sector);
> -    memcpy(data, (uint8_t *)&sector, ndata);
> +    memcpy(data, (uint8_t *)&sector, MIN(sizeof(sector), ndata));

Ok, so typically sizeof(sector) will be 8, and ndata will be 16 for
AES. So without this fix we're reading 8 extra bytes off the stack
that were just after 'sector'. This is harmless, because these extra
8 bytes will just be the '*iv' pointer, and the very next lines
overwrite that bogus extra data. Also explains why valgrind didn't
report it - the extra stack bytes are still valid memory region.
So I don't think this would crash or be a security problem, which is
good, as avoids any backport need. 

>      if (sizeof(sector) < ndata) {
>          memset(data + sizeof(sector), 0, ndata - sizeof(sector));
>      }

Reviewed-by: Daniel P. Berrange <address@hidden>

I'll add this to my qcrypto queue, but if feel free to include it in a pull
of your own if you submit this entire series at once.

Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|