From: Adam Kupczyk <akupczyk@redhat.com>
Date: Thu, 9 Jan 2020 18:02:04 +0000 (-0500)
Subject: bluestore/bluefs: make accounting resiliant to unlock()
X-Git-Tag: v15.1.0~216^2
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=82fa73a411338f4bd27eb9b5863bf5b800617a67;p=ceph.git

bluestore/bluefs: make accounting resiliant to unlock()

Moved vselector.sub_usage, so sub_usage and add_usage will not cross
_flush_bdev_safely(), which contains unlock().

Fixes: https://tracker.ceph.com/issues/43538
Signed-off-by: Adam Kupczyk <akupczyk@redhat.com>
---

diff --git a/src/os/bluestore/BlueFS.cc b/src/os/bluestore/BlueFS.cc
index 26572052dfa4..1ace095ae3f7 100644
--- a/src/os/bluestore/BlueFS.cc
+++ b/src/os/bluestore/BlueFS.cc
@@ -2240,7 +2240,6 @@ void BlueFS::_compact_log_async(std::unique_lock<ceph::mutex>& l)
   flush_bdev();  // FIXME?
 
   _flush_and_sync_log(l, 0, old_log_jump_to);
-  vselector->sub_usage(log_file->vselector_hint, log_file->fnode);
 
   // 2. prepare compacted log
   bluefs_transaction_t t;
@@ -2316,6 +2315,8 @@ void BlueFS::_compact_log_async(std::unique_lock<ceph::mutex>& l)
     ++from;
   }
 
+  vselector->sub_usage(log_file->vselector_hint, log_file->fnode);
+
   // clear the extents from old log file, they are added to new log
   log_file->fnode.clear_extents();
   // swap the log files. New log file is the log file now.