DynamoRIO / DynamoRIO/dynamorio
drcachesim offline traces do not handle large cache flush sizes properly
- Dominant language
- C
- Stars
- 3.2k
- Forks
- 629
- Avg merge
- 2d 15h
- Merged PRs (30d)
- 31
Description
The comment in trace_entry.h explains it pretty well:
```
// On ARM, a flush is requested via a SYS_cacheflush system call,
// and the flush size could be larger than USHRT_MAX.
// If the size is smaller than USHRT_MAX, we use one entry with non-zero size.
// Otherwise, we use two entries, one entry has type TRACE_TYPE_*_FLUSH for
// the start address of flush, and one entry has type TRACE_TYPE_*_FLUSH_END
// for the end address (exclusive) of flush.
// The size field of both entries should be 0.
```
This is done for online traces, but offline traces fail to check for size truncation and fail to use two entries in that case, instead blindly truncating the size.
Contributor guide
Assessment
This issue has not been assessed yet.