github_mirrors/vectorscan
3
0
Fork 1
mirror of https://github.com/VectorCamp/vectorscan.git synced 2025-06-28 16:41:01 +03:00
Code Issues Packages Projects Releases Wiki Activity

Fix noodle SVE2 off by one bug (#309)

Browse Source 
By using svmatch on 16 bit lanes with a 8 bit predicate, we end up
including an undefined character in the pattern checks. The inactive
lane after load contains an undefined value, usually \0. Patterns
using \0 as the last character would then match this spurious
character, returning a match beyond the buffer's end. The fix checks
for such matches and rejects them.

Signed-off-by: Yoan Picchi <yoan.picchi@arm.com>
This commit is contained in:
ypicchi-arm 2024-08-05 07:42:56 +01:00 committed by GitHub
parent 6c8e33e597
commit aa4bc24439
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 9 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
src/hwlm/noodle_engine_sve.hpp
View File

@ -38,8 +38,12 @@ hwlm_error_t checkMatched(const struct noodTable *n, const u8 *buf, size_t len,
size_t matchPos = basePos + svcntp_b8(svptrue_b8(), brk);
DEBUG_PRINTF("match pos %zu\n", matchPos);
assert(matchPos < len);
hwlmcb_rv_t rv = final(n, buf, len, needsConfirm, cbi, matchPos);
RETURN_IF_TERMINATED(rv);
size_t end_of_match_pos = matchPos - cbi->offsetAdj + n->key_offset - 1;
// doubleMatched can add a fake \0 at the end of the buffer. This check get rid of any match that might include it
if(end_of_match_pos < len) {
hwlmcb_rv_t rv = final(n, buf, len, needsConfirm, cbi, matchPos);
RETURN_IF_TERMINATED(rv);
}
next_match = svpnext_b8(matched, next_match);
} while (unlikely(svptest_any(svptrue_b8(), next_match)));
return HWLM_SUCCESS;
@ -155,6 +159,9 @@ svbool_t doubleMatched(svuint16_t chars, const u8 *d,
// d - 1 won't underflow as the first position in buf has been dealt
// with meaning that d > buf
svuint16_t vec_rot = svreinterpret_u16(svld1_u8(pg_rot, d - 1));
// we reuse u8 predicates for u16 lanes. This means that we may actually check against one
// undefined extra character at the end of the buffer (usually \0). We check it later to
// reject this spurious match
*matched = svmatch(pg, vec, chars);
*matched_rot = svmatch(pg_rot, vec_rot, chars);
return svorr_z(svptrue_b8(), *matched, *matched_rot);