From: Greg Farnum <gfarnum@redhat.com>
Date: Tue, 24 Mar 2015 22:15:42 +0000 (-0700)
Subject: client: Hold on to exclusive caps on directories we "own"
X-Git-Tag: v9.0.0~74^2
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=refs%2Fpull%2F4168%2Fhead;p=ceph.git

client: Hold on to exclusive caps on directories we "own"

If a directory is complete, we *really* want to keep the exclusive cap
so that we don't end up needing to do MDS lookup requests on every cache
miss.

Fixes: #11226

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
---

diff --git a/src/client/Client.cc b/src/client/Client.cc
index a649f2276eb2..b272bbec9ded 100644
--- a/src/client/Client.cc
+++ b/src/client/Client.cc
@@ -2850,6 +2850,13 @@ void Client::check_caps(Inode *in, bool is_delayed)
   unsigned used = get_caps_used(in);
   unsigned cap_used;
 
+  if (in->is_dir() && (in->flags & I_COMPLETE)) {
+    // we do this here because we don't want to drop to Fs (and then
+    // drop the Fs if we do a create!) if that alone makes us send lookups
+    // to the MDS. Doing it in in->caps_wanted() has knock-on effects elsewhere
+    wanted |= CEPH_CAP_FILE_EXCL;
+  }
+
   int retain = wanted | used | CEPH_CAP_PIN;
   if (!unmounting) {
     if (wanted)