From: Patrick Donnelly <pdonnell@redhat.com>
Date: Fri, 1 Feb 2019 04:43:42 +0000 (-0800)
Subject: mds: resolve memory leak of recall gather
X-Git-Tag: v14.1.0~105^2
X-Git-Url: http://git.apps.os.sepia.ceph.com/?a=commitdiff_plain;h=d8000a556a3390317fb4eca0621cf36ff2e5eb82;p=ceph-ci.git

mds: resolve memory leak of recall gather

Small change in behavior: after all clients respond to session flush, immediate
begin next recall instead of waiting a whole second.

Fixes: http://tracker.ceph.com/issues/38137
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
---

diff --git a/src/mds/MDSRank.cc b/src/mds/MDSRank.cc
index 810eb0b8fa1..a42121bff5e 100644
--- a/src/mds/MDSRank.cc
+++ b/src/mds/MDSRank.cc
@@ -326,10 +326,13 @@ private:
 
     caps_recalled += count;
     if ((throttled || count > 0) && (recall_timeout == 0 || duration < recall_timeout)) {
-      auto timer = new FunctionContext([this](int _) {
-        recall_client_state();
-      });
-      mds->timer.add_event_after(1.0, timer);
+      C_ContextTimeout *ctx = new C_ContextTimeout(
+        mds, 1, new FunctionContext([this](int r) {
+          recall_client_state();
+      }));
+      ctx->start_timer();
+      gather->set_finisher(new MDSInternalContextWrapper(mds, ctx));
+      gather->activate();
     } else {
       if (!gather->has_subs()) {
         delete gather;