1 files changed, 86 insertions, 65 deletions

diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index 959b1bf7c327..920e9f048bd8 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -150,14 +150,13 @@ static int parse_reply_info_in(void **p, void *end,
 			info->pool_ns_data = *p;
 			*p += info->pool_ns_len;
 		}
-		/* btime, change_attr */
-		{
-			struct ceph_timespec btime;
-			u64 change_attr;
-			ceph_decode_need(p, end, sizeof(btime), bad);
-			ceph_decode_copy(p, &btime, sizeof(btime));
-			ceph_decode_64_safe(p, end, change_attr, bad);
-		}
+
+		/* btime */
+		ceph_decode_need(p, end, sizeof(info->btime), bad);
+		ceph_decode_copy(p, &info->btime, sizeof(info->btime));
+
+		/* change attribute */
+		ceph_decode_64_safe(p, end, info->change_attr, bad);
 
 		/* dir pin */
 		if (struct_v >= 2) {
@@ -166,6 +165,15 @@ static int parse_reply_info_in(void **p, void *end,
 			info->dir_pin = -ENODATA;
 		}
 
+		/* snapshot birth time, remains zero for v<=2 */
+		if (struct_v >= 3) {
+			ceph_decode_need(p, end, sizeof(info->snap_btime), bad);
+			ceph_decode_copy(p, &info->snap_btime,
+					 sizeof(info->snap_btime));
+		} else {
+			memset(&info->snap_btime, 0, sizeof(info->snap_btime));
+		}
+
 		*p = end;
 	} else {
 		if (features & CEPH_FEATURE_MDS_INLINE_DATA) {
@@ -197,7 +205,14 @@ static int parse_reply_info_in(void **p, void *end,
 			}
 		}
 
+		if (features & CEPH_FEATURE_FS_BTIME) {
+			ceph_decode_need(p, end, sizeof(info->btime), bad);
+			ceph_decode_copy(p, &info->btime, sizeof(info->btime));
+			ceph_decode_64_safe(p, end, info->change_attr, bad);
+		}
+
 		info->dir_pin = -ENODATA;
+		/* info->snap_btime remains zero */
 	}
 	return 0;
 bad:
@@ -690,11 +705,12 @@ void ceph_mdsc_release_request(struct kref *kref)
 		ceph_msg_put(req->r_reply);
 	if (req->r_inode) {
 		ceph_put_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
-		iput(req->r_inode);
+		/* avoid calling iput_final() in mds dispatch threads */
+		ceph_async_iput(req->r_inode);
 	}
 	if (req->r_parent)
 		ceph_put_cap_refs(ceph_inode(req->r_parent), CEPH_CAP_PIN);
-	iput(req->r_target_inode);
+	ceph_async_iput(req->r_target_inode);
 	if (req->r_dentry)
 		dput(req->r_dentry);
 	if (req->r_old_dentry)
@@ -708,7 +724,7 @@ void ceph_mdsc_release_request(struct kref *kref)
 		 */
 		ceph_put_cap_refs(ceph_inode(req->r_old_dentry_dir),
 				  CEPH_CAP_PIN);
-		iput(req->r_old_dentry_dir);
+		ceph_async_iput(req->r_old_dentry_dir);
 	}
 	kfree(req->r_path1);
 	kfree(req->r_path2);
@@ -716,6 +732,7 @@ void ceph_mdsc_release_request(struct kref *kref)
 		ceph_pagelist_release(req->r_pagelist);
 	put_request_session(req);
 	ceph_unreserve_caps(req->r_mdsc, &req->r_caps_reservation);
+	WARN_ON_ONCE(!list_empty(&req->r_wait));
 	kfree(req);
 }
 
@@ -818,7 +835,8 @@ static void __unregister_request(struct ceph_mds_client *mdsc,
 	}
 
 	if (req->r_unsafe_dir) {
-		iput(req->r_unsafe_dir);
+		/* avoid calling iput_final() in mds dispatch threads */
+		ceph_async_iput(req->r_unsafe_dir);
 		req->r_unsafe_dir = NULL;
 	}
 
@@ -901,7 +919,7 @@ static int __choose_mds(struct ceph_mds_client *mdsc,
 		struct inode *dir;
 
 		rcu_read_lock();
-		parent = req->r_dentry->d_parent;
+		parent = READ_ONCE(req->r_dentry->d_parent);
 		dir = req->r_parent ? : d_inode_rcu(parent);
 
 		if (!dir || dir->i_sb != mdsc->fsc->sb) {
@@ -983,7 +1001,7 @@ static int __choose_mds(struct ceph_mds_client *mdsc,
 		cap = rb_entry(rb_first(&ci->i_caps), struct ceph_cap, ci_node);
 	if (!cap) {
 		spin_unlock(&ci->i_ceph_lock);
-		iput(inode);
+		ceph_async_iput(inode);
 		goto random;
 	}
 	mds = cap->session->s_mds;
@@ -992,7 +1010,9 @@ static int __choose_mds(struct ceph_mds_client *mdsc,
 	     cap == ci->i_auth_cap ? "auth " : "", cap);
 	spin_unlock(&ci->i_ceph_lock);
 out:
-	iput(inode);
+	/* avoid calling iput_final() while holding mdsc->mutex or
+	 * in mds dispatch threads */
+	ceph_async_iput(inode);
 	return mds;
 
 random:
@@ -1302,7 +1322,9 @@ int ceph_iterate_session_caps(struct ceph_mds_session *session,
 		spin_unlock(&session->s_cap_lock);
 
 		if (last_inode) {
-			iput(last_inode);
+			/* avoid calling iput_final() while holding
+			 * s_mutex or in mds dispatch threads */
+			ceph_async_iput(last_inode);
 			last_inode = NULL;
 		}
 		if (old_cap) {
@@ -1335,7 +1357,7 @@ out:
 	session->s_cap_iterator = NULL;
 	spin_unlock(&session->s_cap_lock);
 
-	iput(last_inode);
+	ceph_async_iput(last_inode);
 	if (old_cap)
 		ceph_put_cap(session->s_mdsc, old_cap);
 
@@ -1471,7 +1493,8 @@ static void remove_session_caps(struct ceph_mds_session *session)
 			spin_unlock(&session->s_cap_lock);
 
 			inode = ceph_find_inode(sb, vino);
-			iput(inode);
+			 /* avoid calling iput_final() while holding s_mutex */
+			ceph_async_iput(inode);
 
 			spin_lock(&session->s_cap_lock);
 		}
@@ -2114,9 +2137,10 @@ retry:
 		if (inode && ceph_snap(inode) == CEPH_SNAPDIR) {
 			dout("build_path path+%d: %p SNAPDIR\n",
 			     pos, temp);
-		} else if (stop_on_nosnap && inode &&
+		} else if (stop_on_nosnap && inode && dentry != temp &&
 			   ceph_snap(inode) == CEPH_NOSNAP) {
 			spin_unlock(&temp->d_lock);
+			pos++; /* get rid of any prepended '/' */
 			break;
 		} else {
 			pos -= temp->d_name.len;
@@ -2127,7 +2151,7 @@ retry:
 			memcpy(path + pos, temp->d_name.name, temp->d_name.len);
 		}
 		spin_unlock(&temp->d_lock);
-		temp = temp->d_parent;
+		temp = READ_ONCE(temp->d_parent);
 
 		/* Are we at the root? */
 		if (IS_ROOT(temp))
@@ -3719,42 +3743,35 @@ static void check_new_map(struct ceph_mds_client *mdsc,
 		     ceph_mdsmap_is_laggy(newmap, i) ? " (laggy)" : "",
 		     ceph_session_state_name(s->s_state));
 
-		if (i >= newmap->m_num_mds ||
-		    memcmp(ceph_mdsmap_get_addr(oldmap, i),
-			   ceph_mdsmap_get_addr(newmap, i),
-			   sizeof(struct ceph_entity_addr))) {
-			if (s->s_state == CEPH_MDS_SESSION_OPENING) {
-				/* the session never opened, just close it
-				 * out now */
-				get_session(s);
-				__unregister_session(mdsc, s);
-				__wake_requests(mdsc, &s->s_waiting);
-				ceph_put_mds_session(s);
-			} else if (i >= newmap->m_num_mds) {
-				/* force close session for stopped mds */
-				get_session(s);
-				__unregister_session(mdsc, s);
-				__wake_requests(mdsc, &s->s_waiting);
-				kick_requests(mdsc, i);
-				mutex_unlock(&mdsc->mutex);
+		if (i >= newmap->m_num_mds) {
+			/* force close session for stopped mds */
+			get_session(s);
+			__unregister_session(mdsc, s);
+			__wake_requests(mdsc, &s->s_waiting);
+			mutex_unlock(&mdsc->mutex);
 
-				mutex_lock(&s->s_mutex);
-				cleanup_session_requests(mdsc, s);
-				remove_session_caps(s);
-				mutex_unlock(&s->s_mutex);
+			mutex_lock(&s->s_mutex);
+			cleanup_session_requests(mdsc, s);
+			remove_session_caps(s);
+			mutex_unlock(&s->s_mutex);
 
-				ceph_put_mds_session(s);
+			ceph_put_mds_session(s);
 
-				mutex_lock(&mdsc->mutex);
-			} else {
-				/* just close it */
-				mutex_unlock(&mdsc->mutex);
-				mutex_lock(&s->s_mutex);
-				mutex_lock(&mdsc->mutex);
-				ceph_con_close(&s->s_con);
-				mutex_unlock(&s->s_mutex);
-				s->s_state = CEPH_MDS_SESSION_RESTARTING;
-			}
+			mutex_lock(&mdsc->mutex);
+			kick_requests(mdsc, i);
+			continue;
+		}
+
+		if (memcmp(ceph_mdsmap_get_addr(oldmap, i),
+			   ceph_mdsmap_get_addr(newmap, i),
+			   sizeof(struct ceph_entity_addr))) {
+			/* just close it */
+			mutex_unlock(&mdsc->mutex);
+			mutex_lock(&s->s_mutex);
+			mutex_lock(&mdsc->mutex);
+			ceph_con_close(&s->s_con);
+			mutex_unlock(&s->s_mutex);
+			s->s_state = CEPH_MDS_SESSION_RESTARTING;
 		} else if (oldstate == newstate) {
 			continue;  /* nothing new with this mds */
 		}
@@ -3912,8 +3929,9 @@ release:
 	ceph_con_send(&session->s_con, msg);
 
 out:
-	iput(inode);
 	mutex_unlock(&session->s_mutex);
+	/* avoid calling iput_final() in mds dispatch threads */
+	ceph_async_iput(inode);
 	return;
 
 bad:
@@ -3922,31 +3940,33 @@ bad:
 }
 
 void ceph_mdsc_lease_send_msg(struct ceph_mds_session *session,
-			      struct inode *inode,
 			      struct dentry *dentry, char action,
 			      u32 seq)
 {
 	struct ceph_msg *msg;
 	struct ceph_mds_lease *lease;
-	int len = sizeof(*lease) + sizeof(u32);
-	int dnamelen = 0;
+	struct inode *dir;
+	int len = sizeof(*lease) + sizeof(u32) + NAME_MAX;
 
-	dout("lease_send_msg inode %p dentry %p %s to mds%d\n",
-	     inode, dentry, ceph_lease_op_name(action), session->s_mds);
-	dnamelen = dentry->d_name.len;
-	len += dnamelen;
+	dout("lease_send_msg identry %p %s to mds%d\n",
+	     dentry, ceph_lease_op_name(action), session->s_mds);
 
 	msg = ceph_msg_new(CEPH_MSG_CLIENT_LEASE, len, GFP_NOFS, false);
 	if (!msg)
 		return;
 	lease = msg->front.iov_base;
 	lease->action = action;
-	lease->ino = cpu_to_le64(ceph_vino(inode).ino);
-	lease->first = lease->last = cpu_to_le64(ceph_vino(inode).snap);
 	lease->seq = cpu_to_le32(seq);
-	put_unaligned_le32(dnamelen, lease + 1);
-	memcpy((void *)(lease + 1) + 4, dentry->d_name.name, dnamelen);
 
+	spin_lock(&dentry->d_lock);
+	dir = d_inode(dentry->d_parent);
+	lease->ino = cpu_to_le64(ceph_ino(dir));
+	lease->first = lease->last = cpu_to_le64(ceph_snap(dir));
+
+	put_unaligned_le32(dentry->d_name.len, lease + 1);
+	memcpy((void *)(lease + 1) + 4,
+	       dentry->d_name.name, dentry->d_name.len);
+	spin_unlock(&dentry->d_lock);
 	/*
 	 * if this is a preemptive lease RELEASE, no need to
 	 * flush request stream, since the actual request will
@@ -4148,6 +4168,7 @@ static void wait_requests(struct ceph_mds_client *mdsc)
 		while ((req = __get_oldest_req(mdsc))) {
 			dout("wait_requests timed out on tid %llu\n",
 			     req->r_tid);
+			list_del_init(&req->r_wait);
 			__unregister_request(mdsc, req);
 		}
 	}