SpecFuzz pass: Don't instrument MFENCE instructions #19
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
Before this change, the SpecFuzz pass would fail to compile code that used MFENCEs because it is treated as a write instruction with no memory operands by LLVM. This patch skips instrumenting MFENCEs. In the future, it might make sense to have SpecFuzz respond in some way to MFENCEs, but for now, skipping them is a decent workaround.
|
Related issue: #18 |
* The main file in the examples folder is named demo.c rather than main.c. * The second edit is a typo (compare the address from the example output to the address in the explanation of that output).
| @@ -37,7 +37,7 @@ Build a sample vulnerable program: | |||
| ```bash | |||
| $ cd example | |||
| $ make sf | |||
| clang-sf -fsanitize=address -O1 main.c -c -o main.sf.o | |||
| clang-sf -fsanitize=address -O1 demo.c -c -o main.sf.o | |||
There was a problem hiding this comment.
This change is already present in another pull request you'd made. No need to add it here as well.
There was a problem hiding this comment.
Whoops this and the other doc change is due to my own bad skill with Github PRs. I'll clean this up.
| @@ -48,7 +48,7 @@ $ ./demo-sf 11 | |||
| [SF], 1, 0x123, 0x456, 0, 0x789 | |||
| r = 0 | |||
| ``` | |||
| Here, the line `[SF], 1, 0x123, 0x456, 0, 0x52b519` means that SpecFuzz detected that the instruction | |||
| Here, the line `[SF], 1, 0x123, 0x456, 0, 0x789` means that SpecFuzz detected that the instruction | |||
| @@ -588,6 +588,8 @@ auto X86SpecFuzzPass::visitWrite(MachineInstr &MI, MachineBasicBlock &Parent) -> | |||
| // Pushes are a special case as the address is always in RSP | |||
| if (isPush(MI.getOpcode())) | |||
| return visitPush(MI, Parent); | |||
| if(MI.getOpcode() == X86::MFENCE) | |||
There was a problem hiding this comment.
There is a cleaner way to do it. Just add MFENCE to X86SpecFuzzPass::isExplicitlySerializing. In this case, MFENCE will be skipped completely.
There was a problem hiding this comment.
Makes sense. I'll send an update later.
There was a problem hiding this comment.
Could you, please, also add a corresponding test to the test suite? See the tests/ directory for examples.
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.
Before this change, the SpecFuzz pass would fail to compile code that used
MFENCEs because it is treated as a write instruction with no memory operands by
LLVM.
This patch skips instrumenting MFENCEs. In the future, it might make sense to
have SpecFuzz respond in some way to MFENCEs, but for now, skipping them is a
decent workaround.
Tested by using the test case here: #18, before the change it crashes, after the change it compiles