client: force sending cap revoke ack always

If just before the revoke request, which will increase the 'seq', is
sent out the clients released the corresponding caps and sent out
the cap update request with old 'seq', the mds will miss checking
the seqs and calculating the caps.

We should always send an ack for revoke requests.

Fixes: https://tracker.ceph.com/issues/61782
Signed-off-by: Xiubo Li <xiubli@redhat.com>
(cherry picked from commit 09a2e33c4aca6e72d75e2527ab5f467576d55dc8)

diff --git a/src/client/Client.cc b/src/client/Client.cc
index 8613f19ed94a2ec27eca87b72e3b3e62960c9e93..4bd7bd9dd8eeb3af04287a03e2c10a215377e880 100644 (file)
--- a/src/client/Client.cc
+++ b/src/client/Client.cc
@@ -5757,6 +5757,13 @@ void Client::handle_cap_grant(MetaSession *session, Inode *in, Cap *cap, const M
     }
   }
 
+  // just in case the caps was released just before we get the revoke msg
+  if (!check && m->get_op() == CEPH_CAP_OP_REVOKE) {
+    cap->wanted = 0; // don't let check_caps skip sending a response to MDS
+    check = true;
+    flags = CHECK_CAPS_NODELAY;
+  }
+
   if (check)
     check_caps(in, flags);