From 6f293e47850a873d0ccc39882be7b3ef6e1043b6 Mon Sep 17 00:00:00 2001
From: Stefan Hajnoczi <stefanha@redhat.com>
Date: Thu, 5 Nov 2015 15:20:57 +0100
Subject: [PATCH 1/2] rbd: make qemu's cache setting override any ceph setting

Message-id: <1446736858-29005-2-git-send-email-stefanha@redhat.com>
Patchwork-id: 68293
O-Subject: [RHEL-7.2.z qemu-kvm PATCH 1/2] rbd: make qemu's cache setting override any ceph setting
Bugzilla: 1279389
RH-Acked-by: Max Reitz <mreitz@redhat.com>
RH-Acked-by: Jeffrey Cody <jcody@redhat.com>
RH-Acked-by: Kevin Wolf <kwolf@redhat.com>

From: Josh Durgin <jdurgin@redhat.com>

To be safe, when cache=none is used ceph settings should not be able
to override it to turn on caching. This was previously possible with
rbd_cache=true in the rbd device configuration or a ceph configuration
file. Similarly, rbd settings could have turned off caching when qemu
requested it, although this would just be a performance problem.

Fix this by changing rbd's cache setting to match qemu after all other
ceph settings have been applied.

Signed-off-by: Josh Durgin <jdurgin@redhat.com>
Reviewed-by: Jeff Cody <jcody@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit 99a3c89d5d538dc6c360e35dffb797cfe06e9cda)
Signed-off-by: Miroslav Rezanina <mrezanin@redhat.com>

Conflicts:
	block/rbd.c - downstream doesn't use Error **errp

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 block/rbd.c | 26 +++++++++++++-------------
 1 file changed, 13 insertions(+), 13 deletions(-)

diff --git a/block/rbd.c b/block/rbd.c
index 4eea455..11a39ac 100644
--- a/block/rbd.c
+++ b/block/rbd.c
@@ -501,19 +501,6 @@ static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags,
         s->snap = g_strdup(snap_buf);
     }
 
-    /*
-     * Fallback to more conservative semantics if setting cache
-     * options fails. Ignore errors from setting rbd_cache because the
-     * only possible error is that the option does not exist, and
-     * librbd defaults to no caching. If write through caching cannot
-     * be set up, fall back to no caching.
-     */
-    if (flags & BDRV_O_NOCACHE) {
-        rados_conf_set(s->cluster, "rbd_cache", "false");
-    } else {
-        rados_conf_set(s->cluster, "rbd_cache", "true");
-    }
-
     if (strstr(conf, "conf=") == NULL) {
         /* try default location, but ignore failure */
         rados_conf_read_file(s->cluster, NULL);
@@ -527,6 +514,19 @@ static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags,
         }
     }
 
+    /*
+     * Fallback to more conservative semantics if setting cache
+     * options fails. Ignore errors from setting rbd_cache because the
+     * only possible error is that the option does not exist, and
+     * librbd defaults to no caching. If write through caching cannot
+     * be set up, fall back to no caching.
+     */
+    if (flags & BDRV_O_NOCACHE) {
+        rados_conf_set(s->cluster, "rbd_cache", "false");
+    } else {
+        rados_conf_set(s->cluster, "rbd_cache", "true");
+    }
+
     r = rados_connect(s->cluster);
     if (r < 0) {
         error_report("error connecting");
-- 
1.8.3.1