OSD: Remove shard lock that mClockScheduler use when updating conf

When mClockScheduler uses bench to determine conf. values a deadlock
can occur if the update scheduler config run during consume_map already started.

Fixes: https://tracker.ceph.com/issues/57699
Signed-off-by: Nitzan Mordechai <nmordec@redhat.com>
(cherry picked from commit 6450dae40120b0e3d9e46b4561b8212ef330470a)

diff --git a/src/osd/OSD.cc b/src/osd/OSD.cc
index 8d950135427696de20329755710d263acf0e3340..48d656f300bdc35c19d01f7a6db8cfab881b74eb 100644 (file)
--- a/src/osd/OSD.cc
+++ b/src/osd/OSD.cc
@@ -11028,7 +11028,6 @@ void OSDShard::unprime_split_children(spg_t parent, unsigned old_pg_num)
 
 void OSDShard::update_scheduler_config()
 {
-  std::lock_guard l(shard_lock);
   scheduler->update_configuration();
 }