From: Casey Bodley <cbodley@redhat.com>
Date: Fri, 28 Feb 2020 16:06:32 +0000 (-0500)
Subject: rgw: bucket_list_ordered advances past duplicate common prefixes
X-Git-Tag: v15.1.1~172^2~2
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=81cfd5da3f7e6d889e6440c23227d8e78c8aa06f;p=ceph.git

rgw: bucket_list_ordered advances past duplicate common prefixes

we may see the same common prefix from more than one shard. when we
detect a duplicate, we need to advance past it. otherwise, we may make
the wrong decision about is_truncated because the shards with
duplicates won't be at_end()

Fixes: https://tracker.ceph.com/issues/44353

Signed-off-by: Casey Bodley <cbodley@redhat.com>
---

diff --git a/src/rgw/rgw_rados.cc b/src/rgw/rgw_rados.cc
index 6142c9bd457a..2e0a9f77c3c2 100644
--- a/src/rgw/rgw_rados.cc
+++ b/src/rgw/rgw_rados.cc
@@ -8309,7 +8309,9 @@ int RGWRados::cls_bucket_list_ordered(RGWBucketInfo& bucket_info,
       // it's important that the values in the map refer to the index
       // into the results_trackers vector, which may not be the same
       // as the shard number (i.e., when not all shards are requested)
-      candidates[t.entry_name()] = tracker_idx;
+      if (!candidates.emplace(t.entry_name(), tracker_idx).second) {
+        t.advance(); // skip duplicate common prefixes
+      }
     }
     ++tracker_idx;
   }
@@ -8367,7 +8369,9 @@ int RGWRados::cls_bucket_list_ordered(RGWBucketInfo& bucket_info,
     // refresh the candidates map
     candidates.erase(candidates.begin());
     if (! tracker.advance().at_end()) {
-      candidates[tracker.entry_name()] = tracker_idx;
+      if (!candidates.emplace(tracker.entry_name(), tracker_idx).second) {
+        tracker.advance(); // skip duplicate common prefixes
+      }
     } else if (tracker.is_truncated()) {
       // once we exhaust one shard that is truncated, we need to stop,
       // as we cannot be certain that one of the next entries needs to