diff options
author | Rahat Mahmood <rahat@google.com> | 2021-05-14 11:06:07 -0700 |
---|---|---|
committer | gVisor bot <gvisor-bot@google.com> | 2021-05-14 11:07:57 -0700 |
commit | 78ae3db1a39c0cd925c6b75807fa1dc76ba99986 (patch) | |
tree | f4efaef41e425289ed253596c8e632431e21fdc2 /test/util/cgroup_util.h | |
parent | 2b457d9ee9ba50da4a9208d957053fac2c77932d (diff) |
Fix cgroup hierarchy registration.
Previously, registration was racy because we were publishing
hierarchies in the registry without fully initializing the underlying
filesystem. This led to concurrent mount(2)s discovering the partially
intialized filesystems and dropping the final refs on them which cause
them to be freed prematurely.
Reported-by: syzbot+13f54e77bdf59f0171f0@syzkaller.appspotmail.com
Reported-by: syzbot+2c7f0a9127ac6a84f17e@syzkaller.appspotmail.com
PiperOrigin-RevId: 373824552
Diffstat (limited to 'test/util/cgroup_util.h')
-rw-r--r-- | test/util/cgroup_util.h | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/test/util/cgroup_util.h b/test/util/cgroup_util.h index b797a8b24..e3f696a89 100644 --- a/test/util/cgroup_util.h +++ b/test/util/cgroup_util.h @@ -83,6 +83,8 @@ class Mounter { PosixError Unmount(const Cgroup& c); + void release(const Cgroup& c); + private: // The destruction order of these members avoids errors during cleanup. We // first unmount (by executing the mounts_ cleanups), then delete the |