[PATCH v2 0/2] btrfs: fix an on-stack path leak and migrate to auto-release for on-stack paths

[Qu Wenruo <wqu@suse.com>]

[CHANGELOG]
v2:
- Add back an early btrfs_release_path() call in
  print_data_reloc_error()
  This is to avoid holding the path (and its extent buffers locked)
  during time consuming iterate_extent_inodes() calls.

  And add a comment on that early release, with updated commit message.

I thought patch "btrfs: make sure extent and csum paths are always released in
scrub_raid56_parity_stripe()" has already taught us that tag based
manual cleanup is never reliable, now there is another similar bug in
print_data_reloc_error().

This time it is harder to expose, as we always imply if the function
returned an error, they should do the proper cleanup.
But extent_to_logical() does not follow that assumption.

The first patch is the minimal fix for backport, the second patch is
going to solve the problem by using auto-release for all on-stack btrfs
paths.

Qu Wenruo (2):
  btrfs: fix a potential path leak in print_data_reloc_error()
  btrfs: introduce BTRFS_PATH_AUTO_RELEASE() helper

 fs/btrfs/ctree.h  |  9 +++++++++
 fs/btrfs/defrag.c |  5 +----
 fs/btrfs/inode.c  |  7 +++++--
 fs/btrfs/scrub.c  | 18 ++++++------------
 4 files changed, 21 insertions(+), 18 deletions(-)

-- 
2.52.0