[PATCH v2 0/2] btrfs: raid56: concurrency fix and a very tiny optimization

[Qu Wenruo <wqu@suse.com>]

We have a unprotected concurrency updateing rbio::error_bitmap.
The first patch is going to fix it.

While we are at rbio_update_error_bitmap(), there is also a tiny
optimization we can do for calculating the bio size.
Since we only care about the size of the bio, bio_for_each_bvec_all() is
much better than bio_for_each_segment_all(), as the former one can
handle multi-page bvec directly to reduce the loop.

Changelog:
v2:
- Use set_bit() in a loop instead of an asymmetrical spinlock.
  Since only endio can have concurrency accessing the bitmap, while
  the main thread only access them in a single thread, we will have
  asymmetrical spinlock schema, which is not ideal.
  Instead go set_bit() in a loop.

- Add a tiny optimization to calculate bio length in
  rbio_update_error_bitmap()
  

Qu Wenruo (2):
  btrfs: raid56: make error_bitmap update atomic
  btrfs: raid56: reduce the overhead to calculate the bio length

 fs/btrfs/raid56.c | 15 +++++++++++----
 1 file changed, 11 insertions(+), 4 deletions(-)

-- 
2.39.1