Cherry-pick: Harden bound checking tests of AES-XTS and replace SSE instructions that degraded performance for certain input lengths #2319
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
…ts. (aws#2286) This change hardens the tests introduced in aws#2227 Fix out-of-bound (OOB) input read in AES-XTS Decrypt in AVX-512 implementation. It adds a memory page preceding the input and output buffer that is protected against read and write in order to detect any under-read, in which case a segfault occurs. The suspected code that can potentially cause a "pre-bound" OOB is the cipher-stealing section in Encrypt [crypto/fipsmodule/aes/asm/aesni-xts-avx512.pl#L1809-L1810](https://github.com/aws/aws-lc/blob/v1.48.5/crypto/fipsmodule/aes/asm/aesni-xts-avx512.pl#L1809-L1810) and decrypt [crypto/fipsmodule/aes/asm/aesni-xts-avx512.pl#2572-L2573](https://github.com/aws/aws-lc/blob/v1.48.5/crypto/fipsmodule/aes/asm/aesni-xts-avx512.pl#L2572-L2573). The efficacy of the added test was shown by changing the decrypt cipher-stealing code for example to: ```diff --- a/crypto/fipsmodule/aes/asm/aesni-xts-avx512.pl +++ b/crypto/fipsmodule/aes/asm/aesni-xts-avx512.pl @@ -2569,7 +2569,7 @@ ___ vpshufb %xmm10,%xmm8,%xmm8 - vmovdqu -0x10($input,$length,1),%xmm3 + vmovdqu -0x12($input,$length,1),%xmm3 vmovdqu %xmm8,-0x10($output,$length,1) ``` With this change, a segmentation fault occurs in the test vector of input length 17 bytes (1 AES block + 1 byte); which is the smallest test vector that requires cipher stealing. At the changed line: - `$input` points at byte 16, i.e. past the first block - `$length` = 1, after [l.2429](https://github.com/aws/aws-lc/blob/v1.48.5/crypto/fipsmodule/aes/asm/aesni-xts-avx512.pl#L2429) - the read index with the diff change is `$input + $length - 18` = `$input -17`, which points at byte "-1", i.e. the byte right before byte 0 of the input, i.e. an underread, this causes a segfault at this vector. - Other larger changes, e.g. -0x20, will have the same result. Another test changes the location of the written output ```@@ -2607,7 +2607,7 @@ ___ .L_done_${rndsuffix}: # store last ciphertext value - vmovdqu %xmm8,-0x10($output) + vmovdqu %xmm8,-0x11($output) ___ } ``` - This test caused a segfault with the smallest input of 1 block = 16 bytes Similar tests in the encrypt path gave the same result of segfaulting when trying to read before the input beginning. By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license and the ISC license. (cherry picked from commit a39439b)
A bimodal performance occurred in the XTS encrypt AVX512 implementation. We have observed more than 80% drop in performance. This is caused by mixing SSE and AVX instructions in the AVX512 implementation. For a subset of input lengths, the code path contained a single move movdqa, an SSE instruction. Use vmovdqa instead. (cherry picked from commit 37c2b5e)
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## fips-2024-09-27 #2319 +/- ##
===================================================
- Coverage 78.60% 78.60% -0.01%
===================================================
Files 585 585
Lines 100601 100608 +7
Branches 14259 14262 +3
===================================================
+ Hits 79076 79079 +3
- Misses 20890 20892 +2
- Partials 635 637 +2 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
### Description of changes: BoringSSL now requires C++17 after google/boringssl@9ff8491. When we build speed.cc with BoringSSL's headers we need to specify our our build to use C++17. This is blocking all of our CI with failures like [this](https://us-west-2.codebuild.aws.amazon.com/project/eyJlbmNyeXB0ZWREYXRhIjoiWk5IUWJGRGxBcTJ2Mkp4WGF3dnBwYjc5V0ZZYSt5SVVGbkwvODkydTNTaVQ2V2FMN3hwa0tjSWNFemw2QWtCWW5welFWV3lpRFpKVitwejgvelhpRWh3NDNqcWhKalpPYW9hL2tLMDlJSDFPT1NkNyIsIml2UGFyYW1ldGVyU3BlYyI6Ilp4VXRNQXFGM1BZYVlaRkIiLCJtYXRlcmlhbFNldFNlcmlhbCI6MX0%3D/build/73b9a032-23e0-493c-b4f2-08fce7df7798). ### Call-outs: This does not change AWS-LC's normal C++11 requirement. ### Testing: The CI will build this. By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license and the ISC license. (cherry picked from commit 04a0f10)
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.
Original PRs: #2286 and #2140
Original commits: a39439b and 37c2b5e
Description of changes:
This is a follow-up to #2228 where an out-of-bound (OOB) read was fixed in the AVX512 implementation of AES-XTS and more tests were added.
This cherry-picks:
By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license and the ISC license.