On 07/02/2015 10:51, w00214312 wrote:
> From: Bin Wu <wu.wu...@huawei.com>
>
> When we test the drive_mirror between different hosts by ndb devices,
> we find that, during the cancel phase the qemu process crashes sometimes.
> By checking the crash core file, we find the stack as follows, which means
> a coroutine re-enter error occurs:
This bug probably can be fixed simply by delaying the setting of
recv_coroutine.
What are the symptoms if you only apply your "qemu-coroutine-lock: fix
co_queue multi-adding bug" patch but not "qemu-coroutine: fix
qemu_co_queue_run_restart error"?
Can you try the patch below? (Compile-tested only).
diff --git a/block/nbd-client.c b/block/nbd-client.c
index 6e1c97c..23d6a71 100644
--- a/block/nbd-client.c
+++ b/block/nbd-client.c
@@ -104,10 +104,21 @@ static int nbd_co_send_request(NbdClientSession *s,
QEMUIOVector *qiov, int offset)
{
AioContext *aio_context;
- int rc, ret;
+ int rc, ret, i;
qemu_co_mutex_lock(&s->send_mutex);
+
+ for (i = 0; i < MAX_NBD_REQUESTS; i++) {
+ if (s->recv_coroutine[i] == NULL) {
+ s->recv_coroutine[i] = qemu_coroutine_self();
+ break;
+ }
+ }
+
+ assert(i < MAX_NBD_REQUESTS);
+ request->handle = INDEX_TO_HANDLE(s, i);
s->send_coroutine = qemu_coroutine_self();
+
aio_context = bdrv_get_aio_context(s->bs);
aio_set_fd_handler(aio_context, s->sock,
nbd_reply_ready, nbd_restart_write, s);
@@ -164,8 +175,6 @@ static void nbd_co_receive_reply(NbdClientSession *s,
static void nbd_coroutine_start(NbdClientSession *s,
struct nbd_request *request)
{
- int i;
-
/* Poor man semaphore. The free_sema is locked when no other request
* can be accepted, and unlocked after receiving one reply. */
if (s->in_flight >= MAX_NBD_REQUESTS - 1) {
@@ -174,15 +183,7 @@ static void nbd_coroutine_start(NbdClientSession *s,
}
s->in_flight++;
- for (i = 0; i < MAX_NBD_REQUESTS; i++) {
- if (s->recv_coroutine[i] == NULL) {
- s->recv_coroutine[i] = qemu_coroutine_self();
- break;
- }
- }
-
- assert(i < MAX_NBD_REQUESTS);
- request->handle = INDEX_TO_HANDLE(s, i);
+ /* s->recv_coroutine[i] is set as soon as we get the send_lock. */
}
static void nbd_coroutine_end(NbdClientSession *s,
