Two fixes for the writeback congestion accounting in
ceph_process_folio_batch(), both found while reworking the writeback
path for folios.

Patch 1 fixes a writeback_count leak when an fscrypt bounce page
allocation fails, which can eventually leave the mount permanently
congested. Patch 2 restores the set-only semantics for write_congested
that the folio batch refactor lost.

---
Tal Zussman (2):
      ceph: fix writeback_count leak when a bounce page allocation fails
      ceph: don't clear write_congested when collecting folios

 fs/ceph/addr.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
---
base-commit: ef063f72b23420b2889a8f631887fe0adae4af12
change-id: 20260904-ceph-writeback-count-leak-2a577e8c15f9

Best regards,
--  
Tal Zussman <tz2294@columbia.edu>