1 files changed, 52 insertions, 26 deletions

diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index a2600101ec22..24561a557e01 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -1,6 +1,7 @@
 #include "ceph_debug.h"
 
 #include <linux/wait.h>
+#include <linux/slab.h>
 #include <linux/sched.h>
 
 #include "mds_client.h"
@@ -328,6 +329,8 @@ static struct ceph_mds_session *register_session(struct ceph_mds_client *mdsc,
 	struct ceph_mds_session *s;
 
 	s = kzalloc(sizeof(*s), GFP_NOFS);
+	if (!s)
+		return ERR_PTR(-ENOMEM);
 	s->s_mdsc = mdsc;
 	s->s_mds = mds;
 	s->s_state = CEPH_MDS_SESSION_NEW;
@@ -529,7 +532,7 @@ static void __unregister_request(struct ceph_mds_client *mdsc,
 {
 	dout("__unregister_request %p tid %lld\n", req, req->r_tid);
 	rb_erase(&req->r_node, &mdsc->request_tree);
-	ceph_mdsc_put_request(req);
+	RB_CLEAR_NODE(&req->r_node);
 
 	if (req->r_unsafe_dir) {
 		struct ceph_inode_info *ci = ceph_inode(req->r_unsafe_dir);
@@ -538,6 +541,8 @@ static void __unregister_request(struct ceph_mds_client *mdsc,
 		list_del_init(&req->r_unsafe_dir_item);
 		spin_unlock(&ci->i_unsafe_lock);
 	}
+
+	ceph_mdsc_put_request(req);
 }
 
 /*
@@ -731,9 +736,10 @@ static void cleanup_cap_releases(struct ceph_mds_session *session)
 }
 
 /*
- * Helper to safely iterate over all caps associated with a session.
+ * Helper to safely iterate over all caps associated with a session, with
+ * special care taken to handle a racing __ceph_remove_cap().
  *
- * caller must hold session s_mutex
+ * Caller must hold session s_mutex.
  */
 static int iterate_session_caps(struct ceph_mds_session *session,
 				 int (*cb)(struct inode *, struct ceph_cap *,
@@ -862,6 +868,7 @@ static int send_renew_caps(struct ceph_mds_client *mdsc,
 	if (time_after_eq(jiffies, session->s_cap_ttl) &&
 	    time_after_eq(session->s_cap_ttl, session->s_renew_requested))
 		pr_info("mds%d caps stale\n", session->s_mds);
+	session->s_renew_requested = jiffies;
 
 	/* do not try to renew caps until a recovering mds has reconnected
 	 * with its clients. */
@@ -874,7 +881,6 @@ static int send_renew_caps(struct ceph_mds_client *mdsc,
 
 	dout("send_renew_caps to mds%d (%s)\n", session->s_mds,
 		ceph_mds_state_name(state));
-	session->s_renew_requested = jiffies;
 	msg = create_session_msg(CEPH_SESSION_REQUEST_RENEWCAPS,
 				 ++session->s_renew_seq);
 	if (IS_ERR(msg))
@@ -1566,8 +1572,13 @@ static int __do_request(struct ceph_mds_client *mdsc,
 
 	/* get, open session */
 	session = __ceph_lookup_mds_session(mdsc, mds);
-	if (!session)
+	if (!session) {
 		session = register_session(mdsc, mds);
+		if (IS_ERR(session)) {
+			err = PTR_ERR(session);
+			goto finish;
+		}
+	}
 	dout("do_request mds%d session %p state %s\n", mds, session,
 	     session_state_name(session->s_state));
 	if (session->s_state != CEPH_MDS_SESSION_OPEN &&
@@ -1770,7 +1781,7 @@ static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
 	dout("handle_reply %p\n", req);
 
 	/* correct session? */
-	if (!req->r_session && req->r_session != session) {
+	if (req->r_session != session) {
 		pr_err("mdsc_handle_reply got %llu on session mds%d"
 		       " not mds%d\n", tid, session->s_mds,
 		       req->r_session ? req->r_session->s_mds : -1);
@@ -2126,7 +2137,7 @@ static void send_mds_reconnect(struct ceph_mds_client *mdsc, int mds)
 	struct ceph_mds_session *session = NULL;
 	struct ceph_msg *reply;
 	struct rb_node *p;
-	int err;
+	int err = -ENOMEM;
 	struct ceph_pagelist *pagelist;
 
 	pr_info("reconnect to recovering mds%d\n", mds);
@@ -2175,7 +2186,7 @@ static void send_mds_reconnect(struct ceph_mds_client *mdsc, int mds)
 		goto fail;
 	err = iterate_session_caps(session, encode_caps_cb, pagelist);
 	if (err < 0)
-		goto out;
+		goto fail;
 
 	/*
 	 * snaprealms.  we provide mds with the ino, seq (version), and
@@ -2203,28 +2214,31 @@ send:
 	reply->nr_pages = calc_pages_for(0, pagelist->length);
 	ceph_con_send(&session->s_con, reply);
 
-	if (session) {
-		session->s_state = CEPH_MDS_SESSION_OPEN;
-		__wake_requests(mdsc, &session->s_waiting);
-	}
+	session->s_state = CEPH_MDS_SESSION_OPEN;
+	mutex_unlock(&session->s_mutex);
+
+	mutex_lock(&mdsc->mutex);
+	__wake_requests(mdsc, &session->s_waiting);
+	mutex_unlock(&mdsc->mutex);
+
+	ceph_put_mds_session(session);
 
-out:
 	up_read(&mdsc->snap_rwsem);
-	if (session) {
-		mutex_unlock(&session->s_mutex);
-		ceph_put_mds_session(session);
-	}
 	mutex_lock(&mdsc->mutex);
 	return;
 
 fail:
 	ceph_msg_put(reply);
+	up_read(&mdsc->snap_rwsem);
+	mutex_unlock(&session->s_mutex);
+	ceph_put_mds_session(session);
 fail_nomsg:
 	ceph_pagelist_release(pagelist);
 	kfree(pagelist);
 fail_nopagelist:
-	pr_err("ENOMEM preparing reconnect for mds%d\n", mds);
-	goto out;
+	pr_err("error %d preparing reconnect for mds%d\n", err, mds);
+	mutex_lock(&mdsc->mutex);
+	return;
 }
 
 
@@ -2682,29 +2696,41 @@ void ceph_mdsc_pre_umount(struct ceph_mds_client *mdsc)
  */
 static void wait_unsafe_requests(struct ceph_mds_client *mdsc, u64 want_tid)
 {
-	struct ceph_mds_request *req = NULL;
+	struct ceph_mds_request *req = NULL, *nextreq;
 	struct rb_node *n;
 
 	mutex_lock(&mdsc->mutex);
 	dout("wait_unsafe_requests want %lld\n", want_tid);
+restart:
 	req = __get_oldest_req(mdsc);
 	while (req && req->r_tid <= want_tid) {
+		/* find next request */
+		n = rb_next(&req->r_node);
+		if (n)
+			nextreq = rb_entry(n, struct ceph_mds_request, r_node);
+		else
+			nextreq = NULL;
 		if ((req->r_op & CEPH_MDS_OP_WRITE)) {
 			/* write op */
 			ceph_mdsc_get_request(req);
+			if (nextreq)
+				ceph_mdsc_get_request(nextreq);
 			mutex_unlock(&mdsc->mutex);
 			dout("wait_unsafe_requests  wait on %llu (want %llu)\n",
 			     req->r_tid, want_tid);
 			wait_for_completion(&req->r_safe_completion);
 			mutex_lock(&mdsc->mutex);
-			n = rb_next(&req->r_node);
 			ceph_mdsc_put_request(req);
-		} else {
-			n = rb_next(&req->r_node);
+			if (!nextreq)
+				break;  /* next dne before, so we're done! */
+			if (RB_EMPTY_NODE(&nextreq->r_node)) {
+				/* next request was removed from tree */
+				ceph_mdsc_put_request(nextreq);
+				goto restart;
+			}
+			ceph_mdsc_put_request(nextreq);  /* won't go away */
 		}
-		if (!n)
-			break;
-		req = rb_entry(n, struct ceph_mds_request, r_node);
+		req = nextreq;
 	}
 	mutex_unlock(&mdsc->mutex);
 	dout("wait_unsafe_requests done\n");