RepoMirrors/ceph
1
0
Fork 0
mirror of https://github.com/ceph/ceph synced 2025-04-01 14:51:13 +00:00
Code Issues Packages Projects Releases Wiki Activity

os/bluestore/BlueFS: prevent _compact_log_async reentry

Browse Source 
_should_compact_log uses new_log != nullptr to tell whether compaction is
already in progress, but we don't set it until we are midway through the
process.  Set it at the top of the method to prevent reentry.

See 455cc6cea2, which failed to implement
this properly.

Fixes: http://tracker.ceph.com/issues/21250
Signed-off-by: Sage Weil <sage@redhat.com>
This commit is contained in:
Sage Weil 2017-09-05 15:01:02 -04:00
parent 683212ae1f
commit 6bd9db304f
1 changed files with 6 additions and 3 deletions

9
src/os/bluestore/BlueFS.cc
View File

@ -1187,6 +1187,11 @@ void BlueFS::_compact_log_async(std::unique_lock<std::mutex>& l)
assert(!new_log);
assert(!new_log_writer);
// create a new log [writer] so that we know compaction is in progress
// (see _should_compact_log)
new_log = new File;
new_log->fnode.ino = 0; // so that _flush_range won't try to log the fnode
// 1. allocate new log space and jump to it.
old_log_jump_to = log_file->fnode.get_allocated();
uint64_t need = old_log_jump_to + cct->_conf->bluefs_max_log_runway;
@ -1228,9 +1233,7 @@ void BlueFS::_compact_log_async(std::unique_lock<std::mutex>& l)
dout(10) << __func__ << " new_log_jump_to 0x" << std::hex << new_log_jump_to
<< std::dec << dendl;
// create a new log [writer]
new_log = new File;
new_log->fnode.ino = 0; // so that _flush_range won't try to log the fnode
// allocate
int r = _allocate(BlueFS::BDEV_DB, new_log_jump_to,
&new_log->fnode.extents);
assert(r == 0);