Fix misaligned reference and logic error in crc32 #1906
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
Previously, this code tried to turn a &[u8] into a &[u32] without checking alignment. This means it could and did create misaligned references, which is UB. This can be detected by running the tests with -Zbuild-std --target=x86_64-unknown-linux-gnu (or whatever your host is). This change adopts the approach from the murmurhash implementation. The previous implementation also ignored the tail bytes. The loop at the end treats num_bytes as if it is the full length of the slice, but it isn't, num_bytes number of bytes after the last 4-byte group. This can be observed for example by changing "hello" to just "hell" in the tests. Under the old implementation, the test will still pass. Now, the value that comes out changes, and "hello" and "hell" hash to different values.
Codecov Report
@@ Coverage Diff @@
## master #1906 +/- ##
==========================================
- Coverage 83.42% 83.42% -0.01%
==========================================
Files 214 214
Lines 57025 57018 -7
==========================================
- Hits 47574 47567 -7
Misses 9451 9451
Continue to review full report at Codecov.
|
tustvold
approved these changes
Jun 19, 2022
| while offset < num_bytes { | ||
| hash = _mm_crc32_u8(hash, bytes[offset]); | ||
| offset += 1; | ||
| let remainder = bytes.len() % 4; |
There was a problem hiding this comment.
mcheshkov
pushed a commit
to cube-js/arrow-rs
that referenced
this pull request
Jul 16, 2024
Previously, this code tried to turn a &[u8] into a &[u32] without checking alignment. This means it could and did create misaligned references, which is UB. This can be detected by running the tests with -Zbuild-std --target=x86_64-unknown-linux-gnu (or whatever your host is). This change adopts the approach from the murmurhash implementation. The previous implementation also ignored the tail bytes. The loop at the end treats num_bytes as if it is the full length of the slice, but it isn't, num_bytes number of bytes after the last 4-byte group. This can be observed for example by changing "hello" to just "hell" in the tests. Under the old implementation, the test will still pass. Now, the value that comes out changes, and "hello" and "hell" hash to different values.
mcheshkov
pushed a commit
to cube-js/arrow-rs
that referenced
this pull request
Jul 17, 2024
Previously, this code tried to turn a &[u8] into a &[u32] without checking alignment. This means it could and did create misaligned references, which is UB. This can be detected by running the tests with -Zbuild-std --target=x86_64-unknown-linux-gnu (or whatever your host is). This change adopts the approach from the murmurhash implementation. The previous implementation also ignored the tail bytes. The loop at the end treats num_bytes as if it is the full length of the slice, but it isn't, num_bytes number of bytes after the last 4-byte group. This can be observed for example by changing "hello" to just "hell" in the tests. Under the old implementation, the test will still pass. Now, the value that comes out changes, and "hello" and "hell" hash to different values.
mcheshkov
pushed a commit
to cube-js/arrow-rs
that referenced
this pull request
Aug 21, 2024
Previously, this code tried to turn a &[u8] into a &[u32] without checking alignment. This means it could and did create misaligned references, which is UB. This can be detected by running the tests with -Zbuild-std --target=x86_64-unknown-linux-gnu (or whatever your host is). This change adopts the approach from the murmurhash implementation. The previous implementation also ignored the tail bytes. The loop at the end treats num_bytes as if it is the full length of the slice, but it isn't, num_bytes number of bytes after the last 4-byte group. This can be observed for example by changing "hello" to just "hell" in the tests. Under the old implementation, the test will still pass. Now, the value that comes out changes, and "hello" and "hell" hash to different values. Can drop this after rebase on commit ded6316 "Fix misaligned reference and logic error in crc32 (apache#1906)", first released in 17.0.0
mcheshkov
pushed a commit
to cube-js/arrow-rs
that referenced
this pull request
Aug 21, 2024
Previously, this code tried to turn a &[u8] into a &[u32] without checking alignment. This means it could and did create misaligned references, which is UB. This can be detected by running the tests with -Zbuild-std --target=x86_64-unknown-linux-gnu (or whatever your host is). This change adopts the approach from the murmurhash implementation. The previous implementation also ignored the tail bytes. The loop at the end treats num_bytes as if it is the full length of the slice, but it isn't, num_bytes number of bytes after the last 4-byte group. This can be observed for example by changing "hello" to just "hell" in the tests. Under the old implementation, the test will still pass. Now, the value that comes out changes, and "hello" and "hell" hash to different values. Can drop this after rebase on commit ded6316 "Fix misaligned reference and logic error in crc32 (apache#1906)", first released in 17.0.0
mcheshkov
pushed a commit
to cube-js/arrow-rs
that referenced
this pull request
Aug 21, 2024
Previously, this code tried to turn a &[u8] into a &[u32] without checking alignment. This means it could and did create misaligned references, which is UB. This can be detected by running the tests with -Zbuild-std --target=x86_64-unknown-linux-gnu (or whatever your host is). This change adopts the approach from the murmurhash implementation. The previous implementation also ignored the tail bytes. The loop at the end treats num_bytes as if it is the full length of the slice, but it isn't, num_bytes number of bytes after the last 4-byte group. This can be observed for example by changing "hello" to just "hell" in the tests. Under the old implementation, the test will still pass. Now, the value that comes out changes, and "hello" and "hell" hash to different values. Can drop this after rebase on commit ded6316 "Fix misaligned reference and logic error in crc32 (apache#1906)", first released in 17.0.0
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.
Which issue does this PR close?
Closes #.
Rationale for this change
Previously, this code tried to turn a &[u8] into a &[u32] without
checking alignment. This means it could and did create misaligned
references, which is UB. This can be detected by running the tests with
-Zbuild-std --target=x86_64-unknown-linux-gnu (or whatever your host
is). This change adopts the approach from the murmurhash implementation.
The previous implementation also ignored the tail bytes. The loop at the
end treats num_bytes as if it is the full length of the slice, but it
isn't, num_bytes number of bytes after the last 4-byte group. This can
be observed for example by changing "hello" to just "hell" in the tests.
Under the old implementation, the test will still pass. Now, the value
that comes out changes, and "hello" and "hell" hash to different values.
What changes are included in this PR?
A soundness and correctness fix for crc32_hash
Are there any user-facing changes?
I don't know. Can users observe the values that come out of crc32_hash?