ceph: drop cap releases in requests composed before cap reconnect

These cap releases are stale because MDS will re-establish client
caps according to the cap reconnect messages.

Note: MDS can detect stale cap messages, so these stale cap
releases are harmless even we don't drop them.

Signed-off-by: Yan, Zheng <zyan@redhat.com>

1 files changed, 13 insertions, 6 deletions

diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index 71c073f38e54..0a1a9e452148 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -1853,7 +1853,7 @@ static int set_request_path_attr(struct inode *rinode, struct dentry *rdentry,
  */
 static struct ceph_msg *create_request_message(struct ceph_mds_client *mdsc,
                                                struct ceph_mds_request *req,
-                                               int mds)
+                                               int mds, bool drop_cap_releases)
 {
         struct ceph_msg *msg;
         struct ceph_mds_request_head *head;
@@ -1937,6 +1937,12 @@ static struct ceph_msg *create_request_message(struct ceph_mds_client *mdsc,
                 releases += ceph_encode_inode_release(&p,
                       req->r_old_dentry->d_inode,
                       mds, req->r_old_inode_drop, req->r_old_inode_unless, 0);
+
+        if (drop_cap_releases) {
+                releases = 0;
+                p = msg->front.iov_base + req->r_request_release_offset;
+        }
+
         head->num_releases = cpu_to_le16(releases);
 
         /* time stamp */
@@ -1989,7 +1995,7 @@ static void complete_request(struct ceph_mds_client *mdsc,
  */
 static int __prepare_send_request(struct ceph_mds_client *mdsc,
                                   struct ceph_mds_request *req,
-                                  int mds)
+                                  int mds, bool drop_cap_releases)
 {
         struct ceph_mds_request_head *rhead;
         struct ceph_msg *msg;
@@ -2048,7 +2054,7 @@ static int __prepare_send_request(struct ceph_mds_client *mdsc,
                 ceph_msg_put(req->r_request);
                 req->r_request = NULL;
         }
-        msg = create_request_message(mdsc, req, mds);
+        msg = create_request_message(mdsc, req, mds, drop_cap_releases);
         if (IS_ERR(msg)) {
                 req->r_err = PTR_ERR(msg);
                 complete_request(mdsc, req);
@@ -2132,7 +2138,7 @@ static int __do_request(struct ceph_mds_client *mdsc,
         if (req->r_request_started == 0)   /* note request start time */
                 req->r_request_started = jiffies;
 
-        err = __prepare_send_request(mdsc, req, mds);
+        err = __prepare_send_request(mdsc, req, mds, false);
         if (!err) {
                 ceph_msg_get(req->r_request);
                 ceph_con_send(&session->s_con, req->r_request);
@@ -2658,7 +2664,7 @@ static void replay_unsafe_requests(struct ceph_mds_client *mdsc,
 
         mutex_lock(&mdsc->mutex);
         list_for_each_entry_safe(req, nreq, &session->s_unsafe, r_unsafe_item) {
-                err = __prepare_send_request(mdsc, req, session->s_mds);
+                err = __prepare_send_request(mdsc, req, session->s_mds, true);
                 if (!err) {
                         ceph_msg_get(req->r_request);
                         ceph_con_send(&session->s_con, req->r_request);
@@ -2679,7 +2685,8 @@ static void replay_unsafe_requests(struct ceph_mds_client *mdsc,
                         continue; /* only old requests */
                 if (req->r_session &&
                     req->r_session->s_mds == session->s_mds) {
-                        err = __prepare_send_request(mdsc, req, session->s_mds);
+                        err = __prepare_send_request(mdsc, req,
+                                                     session->s_mds, true);
                         if (!err) {
                                 ceph_msg_get(req->r_request);
                                 ceph_con_send(&session->s_con, req->r_request);