Bug 13227
Summary: | Potential BUG_ON assertion fails in drm_gem_object_free | ||
---|---|---|---|
Product: | Drivers | Reporter: | Alexey Khoroshilov (khoroshilov) |
Component: | Video(DRI - Intel) | Assignee: | Eric Anholt (eric) |
Status: | RESOLVED CODE_FIX | ||
Severity: | normal | CC: | chris, eric, gordon.jin, jbarnes |
Priority: | P1 | ||
Hardware: | All | ||
OS: | Linux | ||
Kernel Version: | 2.6.30-rc4 | Subsystem: | |
Regression: | No | Bisected commit-id: |
Description
Alexey Khoroshilov
2009-05-02 22:39:22 UTC
wondering if this issue still exist in the latest kernel? (In reply to comment #1) > wondering if this issue still exist in the latest kernel? I have checked Linus tree as of today. The first two issues have been fixed, the third and the fourth ones are still present. Is a more up-to-date tree available anywhere? Yep, these look like real locking bugs. Unfortunately just taking the struct mutex in the pread/pwrite ioctls won't work since the callees take those and have ordering requirements vs. the mmap sem. I guess we'll have to push the bounds checking and final unref into the callees as well. Eric, I think you messed with this code last to fix the mmap vs. struct mutex inversion? Eric? This was cleaned up with the introduction of drm_gem_object_unreference_unlocked() - though we may have missed a few, or even added some more since. |