OSDN Git Service

throttle: Check current timers before updating any_timer_armed[]
authorAlberto Garcia <berto@igalia.com>
Mon, 15 Jun 2015 15:41:15 +0000 (18:41 +0300)
committerStefan Hajnoczi <stefanha@redhat.com>
Tue, 23 Jun 2015 14:06:16 +0000 (15:06 +0100)
Calling throttle_group_config() cancels all timers from a particular
BlockDriverState, so any_timer_armed[] should be updated accordingly.

However, with the current code it may happen that a timer is armed in
a different BlockDriverState from the same group, so any_timer_armed[]
would be set to false in a situation where there is still a timer
armed.

The consequence is that we might end up with two timers armed. This
should not have any noticeable impact however, since all accesses to
the ThrottleGroup are protected by a lock, and the situation would
become normal again shortly thereafter as soon as all timers have been
fired.

The correct way to solve this is to check that we're actually
cancelling a timer before updating any_timer_armed[].

Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-id: 1434382875-3998-1-git-send-email-berto@igalia.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
block/throttle-groups.c

index efc462f..1abc6fc 100644 (file)
@@ -324,9 +324,14 @@ void throttle_group_config(BlockDriverState *bs, ThrottleConfig *cfg)
     ThrottleState *ts = bs->throttle_state;
     ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);
     qemu_mutex_lock(&tg->lock);
-    throttle_config(ts, tt, cfg);
     /* throttle_config() cancels the timers */
-    tg->any_timer_armed[0] = tg->any_timer_armed[1] = false;
+    if (timer_pending(tt->timers[0])) {
+        tg->any_timer_armed[0] = false;
+    }
+    if (timer_pending(tt->timers[1])) {
+        tg->any_timer_armed[1] = false;
+    }
+    throttle_config(ts, tt, cfg);
     qemu_mutex_unlock(&tg->lock);
 }