libceph: fix potential hang in ceph_osdc_notify()

If the cluster becomes unavailable, ceph_osdc_notify() may hang even
with osd_request_timeout option set because linger_notify_finish_wait()
waits for MWatchNotify NOTIFY_COMPLETE message with no associated OSD
request in flight -- it's completely asynchronous.

Introduce an additional timeout, derived from the specified notify
timeout.  While at it, switch both waits to killable which is more
correct.

Cc: [email protected]
Signed-off-by: Ilya Dryomov <[email protected]>
Reviewed-by: Dongsheng Yang <[email protected]>
Reviewed-by: Xiubo Li <[email protected]>

Author: idryomov

net/ceph/osd_client.c

@@ -3334,17 +3334,24 @@ static int linger_reg_commit_wait(struct ceph_osd_linger_request *lreq)
 	int ret;
 
 	dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
-	ret = wait_for_completion_interruptible(&lreq->reg_commit_wait);
+	ret = wait_for_completion_killable(&lreq->reg_commit_wait);
 	return ret ?: lreq->reg_commit_error;
 }
 
-static int linger_notify_finish_wait(struct ceph_osd_linger_request *lreq)
+static int linger_notify_finish_wait(struct ceph_osd_linger_request *lreq,
+				     unsigned long timeout)
 {
-	int ret;
+	long left;
 
 	dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
-	ret = wait_for_completion_interruptible(&lreq->notify_finish_wait);
-	return ret ?: lreq->notify_finish_error;
+	left = wait_for_completion_killable_timeout(&lreq->notify_finish_wait,
+						ceph_timeout_jiffies(timeout));
+	if (left <= 0)
+		left = left ?: -ETIMEDOUT;
+	else
+		left = lreq->notify_finish_error; /* completed */
+
+	return left;
 }
 
 /*
@@ -4896,7 +4903,8 @@ int ceph_osdc_notify(struct ceph_osd_client *osdc,
 	linger_submit(lreq);
 	ret = linger_reg_commit_wait(lreq);
 	if (!ret)
-		ret = linger_notify_finish_wait(lreq);
+		ret = linger_notify_finish_wait(lreq,
+				 msecs_to_jiffies(2 * timeout * MSEC_PER_SEC));
 	else
 		dout("lreq %p failed to initiate notify %d\n", lreq, ret);