Details
Details
- Reviewers
eugenis
Diff Detail
Diff Detail
Event Timeline
Comment Actions
Two things:
- I had to skip the ioctls defined in emu10k1.h and asequencer.h (mostly because the struct definitions there depend on DECLARE_BITMAP, which isn't defined in any of the headers I have on my machine. But we could redefine that macro - it's rather trivial).
- I had to flip READ and WRITE to make it work. Looks like we interpret "WRITE" as "writes to memory", whereas in Linux "_IOW" means "writes to the device". So in most cases our WRITE maps to _IOR and READ maps to _IOR, but I suspect this may not be an exact mapping.
Comment Actions
I think you are correct about _IOW and _IOR.
Please add a test for at least one of these, to see that there is no systematic error here (I realize that testing all of them is hardly possible).
LGTM
lib/sanitizer_common/sanitizer_common_interceptors_ioctl.inc | ||
---|---|---|
663 | Why not add READWRITE here in the same commit? |
Comment Actions
to see that there is no systematic error here
Would you consider the fact that this fixed my content_unittests crashes to be sufficient verification? (Obviously for regression testing one test would do nothing).
lib/sanitizer_common/sanitizer_common_interceptors_ioctl.inc | ||
---|---|---|
663 | fixed |
Why not add READWRITE here in the same commit?