Ceph » Linux kernel client

should remove code in ceph_lock().

Yan, thanks for your review and comments. It is not right to remove the ref increase code in ceph_lock_message. Pls review the latest changes.

From 76ddd7ca3e0a20888d85ebaad129e5a95c39fd99 Mon Sep 17 00:00:00 2001
From: Zhi Zhang <willzzhang@tencent.com>
Date: Fri, 22 Mar 2019 14:16:33 +0800
Subject: [PATCH] ceph: remove duplicated filelock ref increase

Inode i_filelock_ref is increased in ceph_lock or ceph_flock, but it is
increased again in ceph_lock_message. This results in this ref won't
become zero. If CEPH_I_ERROR_FILELOCK flag is set in
remove_session_caps once, this flag can't be cleared even if client is
back to normal. So further file lock will return EIO.

Signed-off-by: Zhi Zhang <zhang.david2011@gmail.com>
---
 fs/ceph/locks.c | 13 -------------
 1 file changed, 13 deletions(-)

diff --git a/fs/ceph/locks.c b/fs/ceph/locks.c
index 9dae2ec..ac9b53b 100644
--- a/fs/ceph/locks.c
+++ b/fs/ceph/locks.c
@@ -237,15 +237,6 @@ int ceph_lock(struct file *file, int cmd, struct file_lock *fl)
        spin_lock(&ci->i_ceph_lock);
        if (ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) {
                err = -EIO;
-       } else if (op == CEPH_MDS_OP_SETFILELOCK) {
-               /*
-                * increasing i_filelock_ref closes race window between
-                * handling request reply and adding file_lock struct to
-                * inode. Otherwise, i_auth_cap may get trimmed in the
-                * window. Caller function will decrease the counter.
-                */
-               fl->fl_ops = &ceph_fl_lock_ops;
-               atomic_inc(&ci->i_filelock_ref);
        }
        spin_unlock(&ci->i_ceph_lock);
        if (err < 0) {
@@ -299,10 +290,6 @@ int ceph_flock(struct file *file, int cmd, struct file_lock *fl)
        spin_lock(&ci->i_ceph_lock);
        if (ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) {
                err = -EIO;
-       } else {
-               /* see comment in ceph_lock */
-               fl->fl_ops = &ceph_fl_lock_ops;
-               atomic_inc(&ci->i_filelock_ref);
        }
        spin_unlock(&ci->i_ceph_lock);
        if (err < 0) {
-- 
1.8.3.1