Fix sequence validation and seqStore bounds check #3439
Merged
Conversation
This file contains 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
Cyan4973
approved these changes
Jan 20, 2023
|
I presume there is no change in performance when sequence validation is not enabled ? |
embg
requested changes
Jan 20, 2023
I ran some benchmarks and there doesn't appear to be any change in performance. |
embg
approved these changes
Jan 23, 2023
lib/common/error_private.c
Outdated
| @@ -53,6 +53,7 @@ const char* ERR_getErrorString(ERR_enum code) | |||
| case PREFIX(dstBuffer_wrong): return "Destination buffer is wrong"; | |||
| case PREFIX(srcBuffer_wrong): return "Source buffer is wrong"; | |||
| case PREFIX(externalMatchFinder_failed): return "External matchfinder returned an error code"; | |||
| case PREFIX(invalid_external_sequences): return "External matchfinder returned an error code"; | |||
There was a problem hiding this comment.
Can you please change this error message to something like "External sequences are not valid"?
daniellerozenblit
force-pushed
the
sequence-validation-bug-fix
branch
from
79697ca to
1b65727
Compare
Cyan4973
reviewed
Jan 23, 2023
lib/common/error_private.c
Outdated
| @@ -53,6 +53,7 @@ const char* ERR_getErrorString(ERR_enum code) | |||
| case PREFIX(dstBuffer_wrong): return "Destination buffer is wrong"; | |||
| case PREFIX(srcBuffer_wrong): return "Source buffer is wrong"; | |||
| case PREFIX(externalMatchFinder_failed): return "External matchfinder returned an error code"; | |||
| case PREFIX(invalid_external_sequences): return "External sequences are not valid"; | |||
There was a problem hiding this comment.
minor: even error codes have a naming scheme!
Looking at existing ones, the pattern is typically topic_qualifier.
For this case, it would be something like :
externalSequences_invalid.
Cyan4973
approved these changes
Jan 23, 2023
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
While working on the Sequence Compression API fuzzer, I found bugs in in
ZSTD_validateSequence()as well as inZSTD_copySequencesToSeqStoreExplicitBlockDelim()that allow one additionalseqDefto be written to theseqStore, past the designated memory allocation.Bug 1
The validation method currently only checks that the
matchLengthof each sequence is greater than the global minimum match length of 3. This is not accurate in cases where theminMatchin the cctx does not properly correspond with the true minimum match length of the sequences.In most cases, this is fine. However, we currently allocate less memory for
minMatch > 3than forminMatch = 3. This means thatZSTD_validateSequence()does not always properly validate sequences given the memory constraints determined byminMatchin the cctx.Fix
Fix: See changes in
ZSTD_validateSequence()We now loosely check that match lengths respect the cctx's
minMatch. Meaning, we check that all match lengths are greater than 3 whenminMatchis greater than 3. This is enough to validate thatmaxNbSeqis large enough to fit all sequences.Bug 2
ZSTD_validateSequence()is optional, so we need to protect agains these cases even when it is not enabled. We have asserts that catch overwrites to theseqStore, but these are not run in the production build. There is also a check beforeZSTD_storeSeq()to ensure that we have not already passed the memory limit. However, this allows us to write one additional sequence pastMaxNbSeqbefore throwing an error.We are able to roundtrip without error in cases where we only write one sequence past the given memory limit.
Fix
Fix: See changes in
ZSTD_copySequencesToSeqStoreExplicitBlockDelim()andZSTD_copySequencesToSeqStoreNoBlockDelim(). I have changed these checks to return an error if we have already reachedmaxNbSeq, so that we do not attempt to write an additional sequence.