From: Tianshan Qu <tianshan@xsky.com>
Date: Thu, 24 May 2018 13:02:39 +0000 (+0800)
Subject: rgw: fix index update in dir_suggest_changes
X-Git-Tag: v10.2.11~9^2
X-Git-Url: http://git.apps.os.sepia.ceph.com/?a=commitdiff_plain;h=2e3f9de81959d4dc8becb17e3e11d5a69e927d04;p=ceph.git

rgw: fix index update in dir_suggest_changes

1. do nothing if the index is no longer exists in cls
   rgw_dir_suggest_changes, whatever the pending state we got in
   check_disk_state, to avoid recreate index when list race with
   delete. the situations:

   * if dir_suggest_changes is done before complete op in Delete, the
     index will be deleted finally.

   * if dir_suggest_changes is done after complete op in Delete, the
     index will no exists and do nothing.

2. stale pending write op's index is existent but exists field is
   false, so remove cur_disk.exists check in op CEPH_RGW_UPDATE, make
   it recover the situation again.

fixes: http://tracker.ceph.com/issues/24627

jewel backport of: https://github.com/ceph/ceph/pull/22217

Signed-off-by: Tianshan Qu <tianshan@xsky.com>
(cherry picked from commit 8304c3e934d667fb89dbe0a79b252a5610b5eb62)
---

diff --git a/src/cls/rgw/cls_rgw.cc b/src/cls/rgw/cls_rgw.cc
index ff994dc5c66ed..5f467b224e34d 100644
--- a/src/cls/rgw/cls_rgw.cc
+++ b/src/cls/rgw/cls_rgw.cc
@@ -1927,7 +1927,7 @@ int rgw_dir_suggest_changes(cls_method_context_t hctx, bufferlist *in, bufferlis
     string cur_change_key;
     encode_obj_index_key(cur_change.key, &cur_change_key);
     int ret = cls_cxx_map_get_val(hctx, cur_change_key, &cur_disk_bl);
-    if (ret < 0 && ret != -ENOENT)
+    if (ret < 0)
       return -EINVAL;
 
     if (cur_disk_bl.length()) {
@@ -1983,18 +1983,6 @@ int rgw_dir_suggest_changes(cls_method_context_t hctx, bufferlist *in, bufferlis
         }
         break;
       case CEPH_RGW_UPDATE:
-	if (!cur_disk.exists) {
-	  // this update would only have been sent by the rgw client
-	  // if the rgw_bucket_dir_entry existed, however between that
-	  // check and now the entry has diappeared, so we were likely
-	  // in the midst of a delete op, and we will not recreate the
-	  // entry
-	  CLS_LOG(10,
-		  "CEPH_RGW_UPDATE not applied because rgw_bucket_dir_entry"
-		  " no longer exists\n");
-	  break;
-	}
-
         CLS_LOG(10, "CEPH_RGW_UPDATE name=%s instance=%s total_entries: %" PRId64 " -> %" PRId64 "\n",
                 cur_change.key.name.c_str(), cur_change.key.instance.c_str(), stats.num_entries, stats.num_entries + 1);
         stats.num_entries++;