rbd: retry watch re-registration periodically

Revamp watch code to support retrying watch re-registration:

- add rbd_dev->watch_state for more robust errcb handling
- store watch cookie separately to avoid dereferencing watch_handle
  which is set to NULL on unwatch
- move re-register code into a delayed work and retry re-registration
  every second, unless the client is blacklisted

Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Mike Christie <mchristi@redhat.com>
Tested-by: Mike Christie <mchristi@redhat.com>

1 files changed, 109 insertions, 29 deletions

diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c
index 1c805eea6767..cb96fb19e8a7 100644
--- a/drivers/block/rbd.c
+++ b/drivers/block/rbd.c
@@ -114,6 +114,8 @@ static int atomic_dec_return_safe(atomic_t *v)
 
 #define RBD_OBJ_PREFIX_LEN_MAX	64
 
+#define RBD_RETRY_DELAY		msecs_to_jiffies(1000)
+
 /* Feature bits */
 
 #define RBD_FEATURE_LAYERING	(1<<0)
@@ -319,6 +321,12 @@ struct rbd_img_request {
 #define for_each_obj_request_safe(ireq, oreq, n) \
 	list_for_each_entry_safe_reverse(oreq, n, &(ireq)->obj_requests, links)
 
+enum rbd_watch_state {
+	RBD_WATCH_STATE_UNREGISTERED,
+	RBD_WATCH_STATE_REGISTERED,
+	RBD_WATCH_STATE_ERROR,
+};
+
 struct rbd_mapping {
 	u64                     size;
 	u64                     features;
@@ -352,7 +360,11 @@ struct rbd_device {
 
 	struct ceph_file_layout	layout;		/* used for all rbd requests */
 
+	struct mutex		watch_mutex;
+	enum rbd_watch_state	watch_state;
 	struct ceph_osd_linger_request *watch_handle;
+	u64			watch_cookie;
+	struct delayed_work	watch_dwork;
 
 	struct workqueue_struct	*task_wq;
 
@@ -3083,9 +3095,6 @@ out_err:
 	obj_request_done_set(obj_request);
 }
 
-static int rbd_dev_header_watch_sync(struct rbd_device *rbd_dev);
-static void __rbd_dev_header_unwatch_sync(struct rbd_device *rbd_dev);
-
 static void rbd_watch_cb(void *arg, u64 notify_id, u64 cookie,
 			 u64 notifier_id, void *data, size_t data_len)
 {
@@ -3113,35 +3122,34 @@ static void rbd_watch_cb(void *arg, u64 notify_id, u64 cookie,
 		rbd_warn(rbd_dev, "notify_ack ret %d", ret);
 }
 
+static void __rbd_unregister_watch(struct rbd_device *rbd_dev);
+
 static void rbd_watch_errcb(void *arg, u64 cookie, int err)
 {
 	struct rbd_device *rbd_dev = arg;
-	int ret;
 
 	rbd_warn(rbd_dev, "encountered watch error: %d", err);
 
-	__rbd_dev_header_unwatch_sync(rbd_dev);
+	mutex_lock(&rbd_dev->watch_mutex);
+	if (rbd_dev->watch_state == RBD_WATCH_STATE_REGISTERED) {
+		__rbd_unregister_watch(rbd_dev);
+		rbd_dev->watch_state = RBD_WATCH_STATE_ERROR;
 
-	ret = rbd_dev_header_watch_sync(rbd_dev);
-	if (ret) {
-		rbd_warn(rbd_dev, "failed to reregister watch: %d", ret);
-		return;
+		queue_delayed_work(rbd_dev->task_wq, &rbd_dev->watch_dwork, 0);
 	}
-
-	ret = rbd_dev_refresh(rbd_dev);
-	if (ret)
-		rbd_warn(rbd_dev, "reregisteration refresh failed: %d", ret);
+	mutex_unlock(&rbd_dev->watch_mutex);
 }
 
 /*
- * Initiate a watch request, synchronously.
+ * watch_mutex must be locked
  */
-static int rbd_dev_header_watch_sync(struct rbd_device *rbd_dev)
+static int __rbd_register_watch(struct rbd_device *rbd_dev)
 {
 	struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc;
 	struct ceph_osd_linger_request *handle;
 
 	rbd_assert(!rbd_dev->watch_handle);
+	dout("%s rbd_dev %p\n", __func__, rbd_dev);
 
 	handle = ceph_osdc_watch(osdc, &rbd_dev->header_oid,
 				 &rbd_dev->header_oloc, rbd_watch_cb,
@@ -3153,13 +3161,16 @@ static int rbd_dev_header_watch_sync(struct rbd_device *rbd_dev)
 	return 0;
 }
 
-static void __rbd_dev_header_unwatch_sync(struct rbd_device *rbd_dev)
+/*
+ * watch_mutex must be locked
+ */
+static void __rbd_unregister_watch(struct rbd_device *rbd_dev)
 {
 	struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc;
 	int ret;
 
-	if (!rbd_dev->watch_handle)
-		return;
+	rbd_assert(rbd_dev->watch_handle);
+	dout("%s rbd_dev %p\n", __func__, rbd_dev);
 
 	ret = ceph_osdc_unwatch(osdc, rbd_dev->watch_handle);
 	if (ret)
@@ -3168,17 +3179,80 @@ static void __rbd_dev_header_unwatch_sync(struct rbd_device *rbd_dev)
 	rbd_dev->watch_handle = NULL;
 }
 
-/*
- * Tear down a watch request, synchronously.
- */
-static void rbd_dev_header_unwatch_sync(struct rbd_device *rbd_dev)
+static int rbd_register_watch(struct rbd_device *rbd_dev)
+{
+	int ret;
+
+	mutex_lock(&rbd_dev->watch_mutex);
+	rbd_assert(rbd_dev->watch_state == RBD_WATCH_STATE_UNREGISTERED);
+	ret = __rbd_register_watch(rbd_dev);
+	if (ret)
+		goto out;
+
+	rbd_dev->watch_state = RBD_WATCH_STATE_REGISTERED;
+	rbd_dev->watch_cookie = rbd_dev->watch_handle->linger_id;
+
+out:
+	mutex_unlock(&rbd_dev->watch_mutex);
+	return ret;
+}
+
+static void cancel_tasks_sync(struct rbd_device *rbd_dev)
 {
-	__rbd_dev_header_unwatch_sync(rbd_dev);
+	dout("%s rbd_dev %p\n", __func__, rbd_dev);
+
+	cancel_delayed_work_sync(&rbd_dev->watch_dwork);
+}
+
+static void rbd_unregister_watch(struct rbd_device *rbd_dev)
+{
+	cancel_tasks_sync(rbd_dev);
+
+	mutex_lock(&rbd_dev->watch_mutex);
+	if (rbd_dev->watch_state == RBD_WATCH_STATE_REGISTERED)
+		__rbd_unregister_watch(rbd_dev);
+	rbd_dev->watch_state = RBD_WATCH_STATE_UNREGISTERED;
+	mutex_unlock(&rbd_dev->watch_mutex);
 
-	dout("%s flushing notifies\n", __func__);
 	ceph_osdc_flush_notifies(&rbd_dev->rbd_client->client->osdc);
 }
 
+static void rbd_reregister_watch(struct work_struct *work)
+{
+	struct rbd_device *rbd_dev = container_of(to_delayed_work(work),
+					    struct rbd_device, watch_dwork);
+	int ret;
+
+	dout("%s rbd_dev %p\n", __func__, rbd_dev);
+
+	mutex_lock(&rbd_dev->watch_mutex);
+	if (rbd_dev->watch_state != RBD_WATCH_STATE_ERROR)
+		goto fail_unlock;
+
+	ret = __rbd_register_watch(rbd_dev);
+	if (ret) {
+		rbd_warn(rbd_dev, "failed to reregister watch: %d", ret);
+		if (ret != -EBLACKLISTED)
+			queue_delayed_work(rbd_dev->task_wq,
+					   &rbd_dev->watch_dwork,
+					   RBD_RETRY_DELAY);
+		goto fail_unlock;
+	}
+
+	rbd_dev->watch_state = RBD_WATCH_STATE_REGISTERED;
+	rbd_dev->watch_cookie = rbd_dev->watch_handle->linger_id;
+	mutex_unlock(&rbd_dev->watch_mutex);
+
+	ret = rbd_dev_refresh(rbd_dev);
+	if (ret)
+		rbd_warn(rbd_dev, "reregisteration refresh failed: %d", ret);
+
+	return;
+
+fail_unlock:
+	mutex_unlock(&rbd_dev->watch_mutex);
+}
+
 /*
  * Synchronous osd object method call.  Returns the number of bytes
  * returned in the outbound buffer, or a negative error code.
@@ -3945,6 +4019,8 @@ static void rbd_spec_free(struct kref *kref)
 
 static void rbd_dev_free(struct rbd_device *rbd_dev)
 {
+	WARN_ON(rbd_dev->watch_state != RBD_WATCH_STATE_UNREGISTERED);
+
 	ceph_oid_destroy(&rbd_dev->header_oid);
 	ceph_oloc_destroy(&rbd_dev->header_oloc);
 
@@ -3991,6 +4067,10 @@ static struct rbd_device *__rbd_dev_create(struct rbd_client *rbdc,
 	ceph_oid_init(&rbd_dev->header_oid);
 	ceph_oloc_init(&rbd_dev->header_oloc);
 
+	mutex_init(&rbd_dev->watch_mutex);
+	rbd_dev->watch_state = RBD_WATCH_STATE_UNREGISTERED;
+	INIT_DELAYED_WORK(&rbd_dev->watch_dwork, rbd_reregister_watch);
+
 	rbd_dev->dev.bus = &rbd_bus_type;
 	rbd_dev->dev.type = &rbd_device_type;
 	rbd_dev->dev.parent = &rbd_root_dev;
@@ -5222,7 +5302,7 @@ static int rbd_dev_image_probe(struct rbd_device *rbd_dev, int depth)
 		goto err_out_format;
 
 	if (!depth) {
-		ret = rbd_dev_header_watch_sync(rbd_dev);
+		ret = rbd_register_watch(rbd_dev);
 		if (ret) {
 			if (ret == -ENOENT)
 				pr_info("image %s/%s does not exist\n",
@@ -5281,7 +5361,7 @@ err_out_probe:
 	rbd_dev_unprobe(rbd_dev);
 err_out_watch:
 	if (!depth)
-		rbd_dev_header_unwatch_sync(rbd_dev);
+		rbd_unregister_watch(rbd_dev);
 err_out_format:
 	rbd_dev->image_format = 0;
 	kfree(rbd_dev->spec->image_id);
@@ -5348,11 +5428,11 @@ static ssize_t do_rbd_add(struct bus_type *bus,
 	rc = rbd_dev_device_setup(rbd_dev);
 	if (rc) {
 		/*
-		 * rbd_dev_header_unwatch_sync() can't be moved into
+		 * rbd_unregister_watch() can't be moved into
 		 * rbd_dev_image_release() without refactoring, see
 		 * commit 1f3ef78861ac.
 		 */
-		rbd_dev_header_unwatch_sync(rbd_dev);
+		rbd_unregister_watch(rbd_dev);
 		rbd_dev_image_release(rbd_dev);
 		goto out;
 	}
@@ -5473,7 +5553,7 @@ static ssize_t do_rbd_remove(struct bus_type *bus,
 	if (ret < 0 || already)
 		return ret;
 
-	rbd_dev_header_unwatch_sync(rbd_dev);
+	rbd_unregister_watch(rbd_dev);
 
 	/*
 	 * Don't free anything from rbd_dev->disk until after all