[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Qemu-devel] [PATCH v2 07/13] mirror: Protect source between bdrv_drain
From: |
Fam Zheng |
Subject: |
[Qemu-devel] [PATCH v2 07/13] mirror: Protect source between bdrv_drain and bdrv_swap |
Date: |
Tue, 2 Jun 2015 11:21:56 +0800 |
Source and target are in sync when we leave the mirror_run loop, they
should remain so until bdrv_swap. Before block_job_defer_to_main_loop
was introduced, it has been easy to prove that. Now that tricky things
can happen after mirror_run returns and before mirror_exit runs, for
example, ioeventfd handlers being called, or a device model timer
callback submitting more I/O.
So, skip the block_job_defer_to_main_loop if we're already in the main
context. This is a necessary special casing until BlockBackend really
honors bdrv_lock.
If we're not in the main context, we rely on the notifying mechanism to
do the right thing.
Signed-off-by: Fam Zheng <address@hidden>
---
block/mirror.c | 18 ++++++++++++++++--
1 file changed, 16 insertions(+), 2 deletions(-)
diff --git a/block/mirror.c b/block/mirror.c
index 58f391a..24cf687 100644
--- a/block/mirror.c
+++ b/block/mirror.c
@@ -57,6 +57,8 @@ typedef struct MirrorBlockJob {
int in_flight;
int sectors_in_flight;
int ret;
+ /* True if the source is locked by us */
+ bool need_unlock;
} MirrorBlockJob;
typedef struct MirrorOp {
@@ -358,6 +360,9 @@ static void mirror_exit(BlockJob *job, void *opaque)
if (replace_aio_context) {
aio_context_release(replace_aio_context);
}
+ if (s->need_unlock) {
+ bdrv_unlock(s->common.bs);
+ }
g_free(s->replaces);
bdrv_unref(s->target);
block_job_completed(&s->common, data->ret);
@@ -521,7 +526,8 @@ static void coroutine_fn mirror_run(void *opaque)
* mirror_populate runs.
*/
trace_mirror_before_drain(s, cnt);
- bdrv_drain(bs);
+ bdrv_lock(bs);
+ s->need_unlock = true;
cnt = bdrv_get_dirty_count(s->dirty_bitmap);
}
@@ -543,6 +549,10 @@ static void coroutine_fn mirror_run(void *opaque)
s->common.cancelled = false;
break;
}
+ if (s->need_unlock) {
+ bdrv_unlock(bs);
+ s->need_unlock = false;
+ }
last_pause_ns = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
}
@@ -565,7 +575,11 @@ immediate_exit:
data = g_malloc(sizeof(*data));
data->ret = ret;
- block_job_defer_to_main_loop(&s->common, mirror_exit, data);
+ if (bs->aio_context == qemu_get_aio_context()) {
+ mirror_exit(&s->common, data);
+ } else {
+ block_job_defer_to_main_loop(&s->common, mirror_exit, data);
+ }
}
static void mirror_set_speed(BlockJob *job, int64_t speed, Error **errp)
--
2.4.1
- [Qemu-devel] [PATCH v2 02/13] block: Introduce bdrv_lock and bdrv_unlock API, (continued)
- [Qemu-devel] [PATCH v2 03/13] blockdev: Lock BDS during internal snapshot transaction, Fam Zheng, 2015/06/01
- [Qemu-devel] [PATCH v2 04/13] blockdev: Lock BDS during external snapshot transaction, Fam Zheng, 2015/06/01
- [Qemu-devel] [PATCH v2 05/13] blockdev: Lock BDS during drive-backup transaction, Fam Zheng, 2015/06/01
- [Qemu-devel] [PATCH v2 06/13] blockdev: Lock BDS during blockdev-backup transaction, Fam Zheng, 2015/06/01
- [Qemu-devel] [PATCH v2 07/13] mirror: Protect source between bdrv_drain and bdrv_swap,
Fam Zheng <=
- [Qemu-devel] [PATCH v2 08/13] block: Add bdrv_add_lock_unlock_notifier, Fam Zheng, 2015/06/01
- [Qemu-devel] [PATCH v2 09/13] block-backend: Add blk_add_lock_unlock_notifier, Fam Zheng, 2015/06/01
- [Qemu-devel] [PATCH v2 10/13] virtio-blk: Move complete_request to 'ops' structure, Fam Zheng, 2015/06/01
- [Qemu-devel] [PATCH v2 11/13] virtio-blk: Don't handle output when backend is locked, Fam Zheng, 2015/06/01
- [Qemu-devel] [PATCH v2 12/13] virtio-scsi-dataplane: Add backend lock listener, Fam Zheng, 2015/06/01
- [Qemu-devel] [PATCH v2 13/13] nbd-server: Clear "can_read" when backend is locked, Fam Zheng, 2015/06/01