[0/5] Various fixes

Message

Josef Bacik Dec. 6, 2019, 2:37 p.m. UTC

These were discovered while reworking how we handle root refs.  They are all
relatively straightforward and mostly deal with error cases, with the exception
of

[PATCH 1/5] btrfs: drop log root for dropped roots
[PATCH 4/5] btrfs: skip log replay on orphaned roots

These two are pretty important and were uncovered with my fsstress patch.  The
first fixes a space leak in the case that we delete a subvol that has a tree log
attached to it.  The leak does not persist across mounts so it's not too bad,
but still pretty important.  The second patch I've only seen in production once
in the last 90 days, but could keep us from mounting if we have a subvol that
was deleted with a tree log that we didn't finish deleting.

The rest of these are just for various error conditions and are less important,
but should be safe enough to send along now if desired.  Thanks,

Josef

Comments

David Sterba Dec. 9, 2019, 6:16 p.m. UTC | #1

On Fri, Dec 06, 2019 at 09:37:13AM -0500, Josef Bacik wrote:
> These were discovered while reworking how we handle root refs.  They are all
> relatively straightforward and mostly deal with error cases, with the exception
> of
> 
> [PATCH 1/5] btrfs: drop log root for dropped roots
> [PATCH 4/5] btrfs: skip log replay on orphaned roots
> 
> These two are pretty important and were uncovered with my fsstress patch.  The
> first fixes a space leak in the case that we delete a subvol that has a tree log
> attached to it.  The leak does not persist across mounts so it's not too bad,
> but still pretty important.  The second patch I've only seen in production once
> in the last 90 days, but could keep us from mounting if we have a subvol that
> was deleted with a tree log that we didn't finish deleting.
> 
> The rest of these are just for various error conditions and are less important,
> but should be safe enough to send along now if desired.  Thanks,

2-5 added to misc-next, 1 has some comments. Thanks.