[PULL] uuid_mutex fixes and cleanups part2

[PULL] uuid_mutex fixes and cleanups part2

[Anand Jain]

These patches we sent independently before and are in the mailing list.
They have been tested successfully using the xfstests. The cyclical
lockdep warning aren't due to these set of patches and they (2) can be
ignored as they are harmless because the threads involved are
ioctl/commit and mount thread which lockdep thinks is warn-able can't
co-exist in the same time space. So this set is ready to be pulled in.
Thanks.

  git@github.com:asj/btrfs-devel.git misc-next-jul18

1/7 has been discussed and reviewed at length, which is drop the
unnecessary uuid_mutex in the btrfs_free_extra_devids().

2/7 fixes the race which syzbot reported.

3/7 as when we sprout we hijack the seed fs_devices, we are bringing
back the seed fs_devices using the clone_fs_devices() instead we could
use our btrfs_scan_one_device() which makes the sprout-ing operation
much cleaner.

4-7/7 are a simple cleanup patches.

Anand Jain (7):
1. btrfs: drop uuid_mutex in btrfs_free_extra_devids()
2. btrfs: fix race between free_stale_devices and close_fs_devices
3. btrfs: do device clone using the btrfs_scan_one_device
4. btrfs: use the assigned fs_devices instead of the dereference
5. btrfs: warn for num_devices below 0
6. btrfs: add helper btrfs_num_devices() to deduce num_devices
7. btrfs: add helper function check device delete able

 fs/btrfs/volumes.c | 106 +++++++++++++++++++++++++++++++----------------------
 1 file changed, 62 insertions(+), 44 deletions(-)

-- 
2.7.0

Re: [PULL] uuid_mutex fixes and cleanups part2

[David Sterba]

On Wed, Jul 18, 2018 at 12:10:24PM +0800, Anand Jain wrote:
> These patches we sent independently before and are in the mailing list.
> They have been tested successfully using the xfstests. The cyclical
> lockdep warning aren't due to these set of patches and they (2) can be
> ignored as they are harmless because the threads involved are
> ioctl/commit and mount thread which lockdep thinks is warn-able can't
> co-exist in the same time space. So this set is ready to be pulled in.
> Thanks.
> 
>   git@github.com:asj/btrfs-devel.git misc-next-jul18
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Please note that this not a URL that I could pull, this scheme is used
for the write acces based on SSH.

The branch name should describe the contents of the branch and be
versioned, not date based.

> 1/7 has been discussed and reviewed at length, which is drop the
> unnecessary uuid_mutex in the btrfs_free_extra_devids().
> 
> 2/7 fixes the race which syzbot reported.
> 
> 3/7 as when we sprout we hijack the seed fs_devices, we are bringing
> back the seed fs_devices using the clone_fs_devices() instead we could
> use our btrfs_scan_one_device() which makes the sprout-ing operation
> much cleaner.
> 
> 4-7/7 are a simple cleanup patches.
> 
> Anand Jain (7):
> 1. btrfs: drop uuid_mutex in btrfs_free_extra_devids()
> 2. btrfs: fix race between free_stale_devices and close_fs_devices
> 3. btrfs: do device clone using the btrfs_scan_one_device
> 4. btrfs: use the assigned fs_devices instead of the dereference
> 5. btrfs: warn for num_devices below 0
> 6. btrfs: add helper btrfs_num_devices() to deduce num_devices
> 7. btrfs: add helper function check device delete able

I've commented in to the patches posted to mailinglist. I'll consider 1
and 2 for merge, need to look up the previous discussions again to
refresh the details. 4 is in misc-next and the rest has comments.