skel: validate yytables_fload header buffer and bound table dimensions#731
Open
MarkLee131 wants to merge 1 commit intowestes:masterfrom
Open
skel: validate yytables_fload header buffer and bound table dimensions#731MarkLee131 wants to merge 1 commit intowestes:masterfrom
MarkLee131 wants to merge 1 commit intowestes:masterfrom
Conversation
The binary tables-file loader trusts attacker-controlled bytes from the
file in three places. Fix all three:
1. yytbl_hdr_read at cpp-flex.skl:4010 did
th->th_name = th->th_version + strlen(th->th_version) + 1;
immediately after fread'ing th_hsize - 14 bytes into th_version.
If those bytes contain no NUL, strlen walks past the heap region.
Fix: use strnlen() and require both the version and the name string
to terminate with NULs inside the buffer; return -1 otherwise.
2. yytbl_fload at cpp-flex.skl:4278 then did
if (strcmp(th.th_name, key) != 0)
even when th_name == th_version + bytes, where the buffer has no
room for the name section. Fixed by the same validation in (1):
th_name is only set when a name NUL is found inside the buffer.
3. yytbl_data_load at cpp-flex.skl:4090-4092 computes
bytes = sizeof(yy_trans_info) * td_lolen * td_hilen
(or the equivalent non-STRUCT form) with td_lolen and td_hilen read
directly from the file as flex_uint32_t. A 50-byte tables file with
td_lolen near UINT32_MAX makes the loader call yyalloc(8 GB). Fix:
reject td_lolen or td_hilen above YYTBL_MAX_DIM (1<<24) before the
multiplication, mirroring 9c54eb6 which added the analogous guard
to the build-time tool's allocators.
Verified with three minimum-size PoCs (16, 16, 50 bytes) that the
unpatched loader trips ASAN on each, and the patched loader returns -1
cleanly. A valid tables file generated by `flex --tables-file=` still
loads as before (regression check passed).
Found via a libFuzzer harness against yytables_fload, expanding what
the upstream OSS-Fuzz projects/flex integration covers (their
flex-patch.diff neuters the m4 fork, which keeps the loader off the
fuzzed code path).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fix #730: The binary tables-file loader trusts attacker-controlled bytes from the file in three places.
yytbl_hdr_read at cpp-flex.skl:4010 did th->th_name = th->th_version + strlen(th->th_version) + 1; immediately after fread'ing th_hsize - 14 bytes into th_version. If those bytes contain no NUL, strlen walks past the heap region. Fix: use strnlen() and require both the version and the name string to terminate with NULs inside the buffer; return -1 otherwise.
yytbl_fload at cpp-flex.skl:4278 then did if (strcmp(th.th_name, key) != 0) even when th_name == th_version + bytes, where the buffer has no room for the name section. Fixed by the same validation in (1): th_name is only set when a name NUL is found inside the buffer.
yytbl_data_load at cpp-flex.skl:4090-4092 computes bytes = sizeof(yy_trans_info) * td_lolen * td_hilen (or the equivalent non-STRUCT form) with td_lolen and td_hilen read directly from the file as flex_uint32_t. A 50-byte tables file with td_lolen near UINT32_MAX makes the loader call yyalloc(8 GB). Fix: reject td_lolen or td_hilen above YYTBL_MAX_DIM (1<<24) before the multiplication, mirroring 9c54eb6 which added the analogous guard to the build-time tool's allocators.
Verified with three minimum-size PoCs (16, 16, 50 bytes) that the unpatched loader trips ASAN on each, and the patched loader returns -1 cleanly. A valid tables file generated by
flex --tables-file=still loads as before (regression check passed).