summaryrefslogtreecommitdiffstats
path: root/fs/ceph
diff options
context:
space:
mode:
authorYan, Zheng <zyan@redhat.com>2015-01-08 21:30:12 +0800
committerIlya Dryomov <idryomov@gmail.com>2015-02-19 13:31:38 +0300
commitd3383a8e37f802818cde4cb489bb0735db637cf0 (patch)
treed4c662fa1a495440c7e6f5e9c851e2562287f930 /fs/ceph
parent73e39e4dba828fe1affefe6290456623319707bd (diff)
downloadlinux-d3383a8e37f802818cde4cb489bb0735db637cf0.tar.bz2
ceph: avoid block operation when !TASK_RUNNING (ceph_mdsc_sync)
check_cap_flush() calls mutex_lock(), which may block. So we can't use it as condition check function for wait_event(); Signed-off-by: Yan, Zheng <zyan@redhat.com>
Diffstat (limited to 'fs/ceph')
-rw-r--r--fs/ceph/caps.c2
-rw-r--r--fs/ceph/mds_client.c51
2 files changed, 35 insertions, 18 deletions
diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c
index 8ed1192606d9..844b57cb52bd 100644
--- a/fs/ceph/caps.c
+++ b/fs/ceph/caps.c
@@ -1450,8 +1450,8 @@ static int __mark_caps_flushing(struct inode *inode,
spin_lock(&mdsc->cap_dirty_lock);
list_del_init(&ci->i_dirty_item);
- ci->i_cap_flush_seq = ++mdsc->cap_flush_seq;
if (list_empty(&ci->i_flushing_item)) {
+ ci->i_cap_flush_seq = ++mdsc->cap_flush_seq;
list_add_tail(&ci->i_flushing_item, &session->s_cap_flushing);
mdsc->num_cap_flushing++;
dout(" inode %p now flushing seq %lld\n", inode,
diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index 85c67ae03e46..fdf5cc8737ee 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -1464,19 +1464,33 @@ out_unlocked:
return err;
}
+static int check_cap_flush(struct inode *inode, u64 want_flush_seq)
+{
+ struct ceph_inode_info *ci = ceph_inode(inode);
+ int ret;
+ spin_lock(&ci->i_ceph_lock);
+ if (ci->i_flushing_caps)
+ ret = ci->i_cap_flush_seq >= want_flush_seq;
+ else
+ ret = 1;
+ spin_unlock(&ci->i_ceph_lock);
+ return ret;
+}
+
/*
* flush all dirty inode data to disk.
*
* returns true if we've flushed through want_flush_seq
*/
-static int check_cap_flush(struct ceph_mds_client *mdsc, u64 want_flush_seq)
+static void wait_caps_flush(struct ceph_mds_client *mdsc, u64 want_flush_seq)
{
- int mds, ret = 1;
+ int mds;
dout("check_cap_flush want %lld\n", want_flush_seq);
mutex_lock(&mdsc->mutex);
- for (mds = 0; ret && mds < mdsc->max_sessions; mds++) {
+ for (mds = 0; mds < mdsc->max_sessions; mds++) {
struct ceph_mds_session *session = mdsc->sessions[mds];
+ struct inode *inode = NULL;
if (!session)
continue;
@@ -1489,29 +1503,29 @@ static int check_cap_flush(struct ceph_mds_client *mdsc, u64 want_flush_seq)
list_entry(session->s_cap_flushing.next,
struct ceph_inode_info,
i_flushing_item);
- struct inode *inode = &ci->vfs_inode;
- spin_lock(&ci->i_ceph_lock);
- if (ci->i_cap_flush_seq <= want_flush_seq) {
+ if (!check_cap_flush(&ci->vfs_inode, want_flush_seq)) {
dout("check_cap_flush still flushing %p "
- "seq %lld <= %lld to mds%d\n", inode,
- ci->i_cap_flush_seq, want_flush_seq,
- session->s_mds);
- ret = 0;
+ "seq %lld <= %lld to mds%d\n",
+ &ci->vfs_inode, ci->i_cap_flush_seq,
+ want_flush_seq, session->s_mds);
+ inode = igrab(&ci->vfs_inode);
}
- spin_unlock(&ci->i_ceph_lock);
}
mutex_unlock(&session->s_mutex);
ceph_put_mds_session(session);
- if (!ret)
- return ret;
+ if (inode) {
+ wait_event(mdsc->cap_flushing_wq,
+ check_cap_flush(inode, want_flush_seq));
+ iput(inode);
+ }
+
mutex_lock(&mdsc->mutex);
}
mutex_unlock(&mdsc->mutex);
dout("check_cap_flush ok, flushed thru %lld\n", want_flush_seq);
- return ret;
}
/*
@@ -3447,14 +3461,17 @@ void ceph_mdsc_sync(struct ceph_mds_client *mdsc)
dout("sync\n");
mutex_lock(&mdsc->mutex);
want_tid = mdsc->last_tid;
- want_flush = mdsc->cap_flush_seq;
mutex_unlock(&mdsc->mutex);
- dout("sync want tid %lld flush_seq %lld\n", want_tid, want_flush);
ceph_flush_dirty_caps(mdsc);
+ spin_lock(&mdsc->cap_dirty_lock);
+ want_flush = mdsc->cap_flush_seq;
+ spin_unlock(&mdsc->cap_dirty_lock);
+
+ dout("sync want tid %lld flush_seq %lld\n", want_tid, want_flush);
wait_unsafe_requests(mdsc, want_tid);
- wait_event(mdsc->cap_flushing_wq, check_cap_flush(mdsc, want_flush));
+ wait_caps_flush(mdsc, want_flush);
}
/*