ceph: use i_ceph_lock instead of i_lock
We have been using i_lock to protect all kinds of data structures in the
ceph_inode_info struct, including lists of inodes that we need to iterate
over while avoiding races with inode destruction. That requires grabbing
a reference to the inode with the list lock protected, but igrab() now
takes i_lock to check the inode flags.
Changing the list lock ordering would be a painful process.
However, using a ceph-specific i_ceph_lock in the ceph inode instead of
i_lock is a simple mechanical change and avoids the ordering constraints
imposed by igrab().
Reported-by: Amon Ott <a.ott@m-privacy.de>
Signed-off-by: Sage Weil <sage@newdream.net>
diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index 264ab70..34bc350 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -732,21 +732,21 @@
}
}
- spin_lock(&inode->i_lock);
+ spin_lock(&ci->i_ceph_lock);
cap = NULL;
if (mode == USE_AUTH_MDS)
cap = ci->i_auth_cap;
if (!cap && !RB_EMPTY_ROOT(&ci->i_caps))
cap = rb_entry(rb_first(&ci->i_caps), struct ceph_cap, ci_node);
if (!cap) {
- spin_unlock(&inode->i_lock);
+ spin_unlock(&ci->i_ceph_lock);
goto random;
}
mds = cap->session->s_mds;
dout("choose_mds %p %llx.%llx mds%d (%scap %p)\n",
inode, ceph_vinop(inode), mds,
cap == ci->i_auth_cap ? "auth " : "", cap);
- spin_unlock(&inode->i_lock);
+ spin_unlock(&ci->i_ceph_lock);
return mds;
random:
@@ -951,7 +951,7 @@
dout("removing cap %p, ci is %p, inode is %p\n",
cap, ci, &ci->vfs_inode);
- spin_lock(&inode->i_lock);
+ spin_lock(&ci->i_ceph_lock);
__ceph_remove_cap(cap);
if (!__ceph_is_any_real_caps(ci)) {
struct ceph_mds_client *mdsc =
@@ -984,7 +984,7 @@
}
spin_unlock(&mdsc->cap_dirty_lock);
}
- spin_unlock(&inode->i_lock);
+ spin_unlock(&ci->i_ceph_lock);
while (drop--)
iput(inode);
return 0;
@@ -1015,10 +1015,10 @@
wake_up_all(&ci->i_cap_wq);
if (arg) {
- spin_lock(&inode->i_lock);
+ spin_lock(&ci->i_ceph_lock);
ci->i_wanted_max_size = 0;
ci->i_requested_max_size = 0;
- spin_unlock(&inode->i_lock);
+ spin_unlock(&ci->i_ceph_lock);
}
return 0;
}
@@ -1151,7 +1151,7 @@
if (session->s_trim_caps <= 0)
return -1;
- spin_lock(&inode->i_lock);
+ spin_lock(&ci->i_ceph_lock);
mine = cap->issued | cap->implemented;
used = __ceph_caps_used(ci);
oissued = __ceph_caps_issued_other(ci, cap);
@@ -1170,7 +1170,7 @@
__ceph_remove_cap(cap);
} else {
/* try to drop referring dentries */
- spin_unlock(&inode->i_lock);
+ spin_unlock(&ci->i_ceph_lock);
d_prune_aliases(inode);
dout("trim_caps_cb %p cap %p pruned, count now %d\n",
inode, cap, atomic_read(&inode->i_count));
@@ -1178,7 +1178,7 @@
}
out:
- spin_unlock(&inode->i_lock);
+ spin_unlock(&ci->i_ceph_lock);
return 0;
}
@@ -1296,7 +1296,7 @@
i_flushing_item);
struct inode *inode = &ci->vfs_inode;
- spin_lock(&inode->i_lock);
+ spin_lock(&ci->i_ceph_lock);
if (ci->i_cap_flush_seq <= want_flush_seq) {
dout("check_cap_flush still flushing %p "
"seq %lld <= %lld to mds%d\n", inode,
@@ -1304,7 +1304,7 @@
session->s_mds);
ret = 0;
}
- spin_unlock(&inode->i_lock);
+ spin_unlock(&ci->i_ceph_lock);
}
mutex_unlock(&session->s_mutex);
ceph_put_mds_session(session);
@@ -2011,10 +2011,10 @@
struct ceph_inode_info *ci = ceph_inode(inode);
dout("invalidate_dir_request %p (D_COMPLETE, lease(s))\n", inode);
- spin_lock(&inode->i_lock);
+ spin_lock(&ci->i_ceph_lock);
ceph_dir_clear_complete(inode);
ci->i_release_count++;
- spin_unlock(&inode->i_lock);
+ spin_unlock(&ci->i_ceph_lock);
if (req->r_dentry)
ceph_invalidate_dentry_lease(req->r_dentry);
@@ -2422,7 +2422,7 @@
if (err)
goto out_free;
- spin_lock(&inode->i_lock);
+ spin_lock(&ci->i_ceph_lock);
cap->seq = 0; /* reset cap seq */
cap->issue_seq = 0; /* and issue_seq */
@@ -2445,7 +2445,7 @@
rec.v1.pathbase = cpu_to_le64(pathbase);
reclen = sizeof(rec.v1);
}
- spin_unlock(&inode->i_lock);
+ spin_unlock(&ci->i_ceph_lock);
if (recon_state->flock) {
int num_fcntl_locks, num_flock_locks;
