From: Sage Weil <sage@redhat.com>
Date: Thu, 3 Sep 2015 17:41:43 +0000 (-0400)
Subject: client: clarify setattr forced sync behavior
X-Git-Tag: v10.0.0~123^2~1
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=ac0314436a5bc6903a90bbd43f8ef5d7aa25b113;p=ceph.git

client: clarify setattr forced sync behavior

Signed-off-by: Sage Weil <sage@redhat.com>
---

diff --git a/src/client/Client.cc b/src/client/Client.cc
index b1da7a896bb..669a835a1f3 100644
--- a/src/client/Client.cc
+++ b/src/client/Client.cc
@@ -5807,6 +5807,17 @@ int Client::_setattr(Inode *in, struct stat *attr, int mask, int uid, int gid,
 		   << " != cap dirtier " << in->cap_dirtier_uid << ":"
 		   << in->cap_dirtier_gid << ", forcing sync setattr"
 		   << dendl;
+    /*
+     * This works because we implicitly flush the caps as part of the
+     * request, so the cap update check will happen with the writeback
+     * cap context, and then the setattr check will happen with the
+     * caller's context.
+     *
+     * In reality this pattern is likely pretty rare (different users
+     * setattr'ing the same file).  If that turns out not to be the
+     * case later, we can build a more complex pipelined cap writeback
+     * infrastructure...
+     */
     if (!mask)
       mask |= CEPH_SETATTR_CTIME;
     goto force_request;