From: Yan, Zheng <zyan@redhat.com>
Date: Mon, 10 Oct 2016 10:39:30 +0000 (+0800)
Subject: os/ObjectStore: properly clear object map when replaying OP_REMOVE
X-Git-Tag: v11.1.0~527^2~1
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=3223378194ad59850177cb87754cb1ada2e5e5f1;p=ceph.git

os/ObjectStore: properly clear object map when replaying OP_REMOVE

To remove an object, filestore needs to unlink corresponding object
file from filesystem and removes corresponding object keys from
DBObjectMap. When replaying OP_REMOVE operation, it's possible the
operation has completed partially, object file has been deleted, but
object keys in DBObjectMap hasn't.

The fix is force clear object keys if object file does not exists

Fixes: http://tracker.ceph.com/issues/17177
Signed-off-by: Yan, Zheng <zyan@redhat.com>
---

diff --git a/src/os/filestore/FileStore.cc b/src/os/filestore/FileStore.cc
index ff9283148619..7029571212d4 100644
--- a/src/os/filestore/FileStore.cc
+++ b/src/os/filestore/FileStore.cc
@@ -472,13 +472,7 @@ int FileStore::lfn_unlink(const coll_t& cid, const ghobject_t& o,
     }
 
     if (!force_clear_omap) {
-      if (hardlink == 0) {
-          if (!m_disable_wbthrottle) {
-	    wbthrottle.clear_object(o); // should be only non-cache ref
-          }
-	  fdcache.clear(o);
-	  return 0;
-      } else if (hardlink == 1) {
+      if (hardlink == 0 || hardlink == 1) {
 	  force_clear_omap = true;
       }
     }
@@ -505,6 +499,12 @@ int FileStore::lfn_unlink(const coll_t& cid, const ghobject_t& o,
       if (!backend->can_checkpoint())
 	object_map->sync(&o, &spos);
     }
+    if (hardlink == 0) {
+      if (!m_disable_wbthrottle) {
+	wbthrottle.clear_object(o); // should be only non-cache ref
+      }
+      return 0;
+    }
   }
   r = index->unlink(o);
   if (r < 0) {