behlendorf approved this pull request. Looks good. While reviewing I took the opportunity to [port this to Linux](https://github.com/zfsonlinux/zfs/pull/7980).
My only concern is the use of the `cv_broadcast(); cv_destroy();` construct. This was always dodgy because once `cv_destroy()` was called no process could safely access the cv except to wake up and exit `cv_wait()`. `cv_destroy()` does block for all current waiters to exit so there's no use-after-free issue, but no new waiters are allowed, and according to the commit logs we ran in to some deadlocks with the VFS due to the waiting. I believe @lundman ran in to the same issue, I don't think this ever a problem for illumos of FreeBSD simply due to VFS differences. This is all to say there's a good chance that at least on Linux we're going to need to add back in our fix for this issue which was to defer the `cv_destroy()` until the end of `zfs_range_unlock()`. -- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/openzfs/openzfs/pull/680#pullrequestreview-160924430 ------------------------------------------ openzfs: openzfs-developer Permalink: https://openzfs.topicbox.com/groups/developer/T0ada5cf1c6037d04-M50c5b4189022d5338b133155 Delivery options: https://openzfs.topicbox.com/groups/developer/subscription
