librbd: retry lock requests periodically until acquired

If the exclusive lock owner acks the lock release request but crashes
before it actually releases the lock, the requestor will wait forever.
Therefore, after a certain timeout, retry the request again until it
succeeds.

Fixes: #11537
Backport: hammer
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
(cherry picked from commit 37c74e6e5274208d1b6efaf315afec03ea7eaa82)

diff --git a/src/librbd/ImageWatcher.cc b/src/librbd/ImageWatcher.cc
index 798b4320343eadd7c837cf12bfe8402eadef8226..477e9cdb32bbf5059f767f7f7986bb4a860b57fd 100644 (file)
--- a/src/librbd/ImageWatcher.cc
+++ b/src/librbd/ImageWatcher.cc
@@ -602,6 +602,13 @@ void ImageWatcher::notify_request_lock() {
     lderr(m_image_ctx.cct) << "error requesting lock: " << cpp_strerror(r)
                           << dendl;
     schedule_retry_aio_requests(true);
+  } else {
+    // lock owner acked -- but resend if we don't see them release the lock
+    int retry_timeout = m_image_ctx.cct->_conf->client_notify_timeout;
+    FunctionContext *ctx = new FunctionContext(
+      boost::bind(&ImageWatcher::notify_request_lock, this));
+    m_task_finisher->add_event_after(TASK_CODE_REQUEST_LOCK,
+                                     retry_timeout, ctx);
   }
 }
 
@@ -728,6 +735,7 @@ void ImageWatcher::handle_payload(const HeaderUpdatePayload &payload,
 void ImageWatcher::handle_payload(const AcquiredLockPayload &payload,
                                   bufferlist *out) {
   ldout(m_image_ctx.cct, 10) << "image exclusively locked announcement" << dendl;
+  m_task_finisher->cancel(TASK_CODE_REQUEST_LOCK);
   if (payload.client_id.is_valid()) {
     Mutex::Locker l(m_owner_client_id_lock);
     if (payload.client_id == m_owner_client_id) {
@@ -747,6 +755,7 @@ void ImageWatcher::handle_payload(const AcquiredLockPayload &payload,
 void ImageWatcher::handle_payload(const ReleasedLockPayload &payload,
                                   bufferlist *out) {
   ldout(m_image_ctx.cct, 10) << "exclusive lock released" << dendl;
+  m_task_finisher->cancel(TASK_CODE_REQUEST_LOCK);
   if (payload.client_id.is_valid()) {
     Mutex::Locker l(m_owner_client_id_lock);
     if (payload.client_id != m_owner_client_id) {