Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client: (23 commits)
  ceph: fix cap flush race reentrancy
  libceph: subscribe to osdmap when cluster is full
  libceph: handle new osdmap down/state change encoding
  rbd: handle online resize of underlying rbd image
  ceph: avoid inode lookup on nfs fh reconnect
  ceph: use LOOKUPINO to make unconnected nfs fh more reliable
  rbd: use snprintf for disk->disk_name
  rbd: cleanup: make kfree match kmalloc
  rbd: warn on update_snaps failure on notify
  ceph: check return value for start_request in writepages
  ceph: remove useless check
  libceph: add missing breaks in addr_set_port
  libceph: fix TAG_WAIT case
  ceph: fix broken comparison in readdir loop
  libceph: fix osdmap timestamp assignment
  ceph: fix rare potential cap leak
  libceph: use snprintf for unknown addrs
  libceph: use snprintf for formatting object name
  ceph: use snprintf for dirstat content
  libceph: fix uninitialized value when no get_authorizer method is set
  ...

drivers/block/rbd.c

@@ -1191,14 +1191,19 @@ static int rbd_req_sync_notify_ack(struct rbd_device *dev,
 static void rbd_watch_cb(u64 ver, u64 notify_id, u8 opcode, void *data)
 {
 	struct rbd_device *dev = (struct rbd_device *)data;
+	int rc;
+
 	if (!dev)
 		return;
 
 	dout("rbd_watch_cb %s notify_id=%lld opcode=%d\n", dev->obj_md_name,
 		notify_id, (int)opcode);
 	mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING);
-	__rbd_update_snaps(dev);
+	rc = __rbd_update_snaps(dev);
 	mutex_unlock(&ctl_mutex);
+	if (rc)
+		pr_warning(DRV_NAME "%d got notification but failed to update"
+			   " snaps: %d\n", dev->major, rc);
 
 	rbd_req_sync_notify_ack(dev, ver, notify_id, dev->obj_md_name);
 }
@@ -1597,7 +1602,7 @@ static int rbd_header_add_snap(struct rbd_device *dev,
 	int name_len = strlen(snap_name);
 	u64 new_snapid;
 	int ret;
-	void *data, *data_start, *data_end;
+	void *data, *p, *e;
 	u64 ver;
 
 	/* we should create a snapshot only if we're pointing at the head */
@@ -1614,16 +1619,16 @@ static int rbd_header_add_snap(struct rbd_device *dev,
 	if (!data)
 		return -ENOMEM;
 
-	data_start = data;
-	data_end = data + name_len + 16;
+	p = data;
+	e = data + name_len + 16;
 
-	ceph_encode_string_safe(&data, data_end, snap_name, name_len, bad);
-	ceph_encode_64_safe(&data, data_end, new_snapid, bad);
+	ceph_encode_string_safe(&p, e, snap_name, name_len, bad);
+	ceph_encode_64_safe(&p, e, new_snapid, bad);
 
 	ret = rbd_req_sync_exec(dev, dev->obj_md_name, "rbd", "snap_add",
-				data_start, data - data_start, &ver);
+				data, p - data, &ver);
 
-	kfree(data_start);
+	kfree(data);
 
 	if (ret < 0)
 		return ret;
@@ -1659,6 +1664,9 @@ static int __rbd_update_snaps(struct rbd_device *rbd_dev)
 	if (ret < 0)
 		return ret;
 
+	/* resized? */
+	set_capacity(rbd_dev->disk, h.image_size / 512ULL);
+
 	down_write(&rbd_dev->header.snap_rwsem);
 
 	snap_seq = rbd_dev->header.snapc->seq;
@@ -1716,7 +1724,8 @@ static int rbd_init_disk(struct rbd_device *rbd_dev)
 	if (!disk)
 		goto out;
 
-	sprintf(disk->disk_name, DRV_NAME "%d", rbd_dev->id);
+	snprintf(disk->disk_name, sizeof(disk->disk_name), DRV_NAME "%d",
+		 rbd_dev->id);
 	disk->major = rbd_dev->major;
 	disk->first_minor = 0;
 	disk->fops = &rbd_bd_ops;

fs/ceph/addr.c

@@ -848,7 +848,8 @@ static int ceph_writepages_start(struct address_space *mapping,
 		op->payload_len = cpu_to_le32(len);
 		req->r_request->hdr.data_len = cpu_to_le32(len);
 
-		ceph_osdc_start_request(&fsc->client->osdc, req, true);
+		rc = ceph_osdc_start_request(&fsc->client->osdc, req, true);
+		BUG_ON(rc);
 		req = NULL;
 
 		/* continue? */
@@ -880,8 +881,6 @@ static int ceph_writepages_start(struct address_space *mapping,
 out:
 	if (req)
 		ceph_osdc_put_request(req);
-	if (rc > 0)
-		rc = 0;  /* vfs expects us to return 0 */
 	ceph_put_snap_context(snapc);
 	dout("writepages done, rc = %d\n", rc);
 	return rc;

fs/ceph/caps.c

@@ -569,7 +569,8 @@ int ceph_add_cap(struct inode *inode,
 		list_add_tail(&cap->session_caps, &session->s_caps);
 		session->s_nr_caps++;
 		spin_unlock(&session->s_cap_lock);
-	}
+	} else if (new_cap)
+		ceph_put_cap(mdsc, new_cap);
 
 	if (!ci->i_snap_realm) {
 		/*
@@ -2634,6 +2635,7 @@ static void handle_cap_export(struct inode *inode, struct ceph_mds_caps *ex,
 			      struct ceph_mds_session *session,
 			      int *open_target_sessions)
 {
+	struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
 	struct ceph_inode_info *ci = ceph_inode(inode);
 	int mds = session->s_mds;
 	unsigned mseq = le32_to_cpu(ex->migrate_seq);
@@ -2670,6 +2672,19 @@ static void handle_cap_export(struct inode *inode, struct ceph_mds_caps *ex,
 			 * export targets, so that we get the matching IMPORT
 			 */
 			*open_target_sessions = 1;
+
+			/*
+			 * we can't flush dirty caps that we've seen the
+			 * EXPORT but no IMPORT for
+			 */
+			spin_lock(&mdsc->cap_dirty_lock);
+			if (!list_empty(&ci->i_dirty_item)) {
+				dout(" moving %p to cap_dirty_migrating\n",
+				     inode);
+				list_move(&ci->i_dirty_item,
+					  &mdsc->cap_dirty_migrating);
+			}
+			spin_unlock(&mdsc->cap_dirty_lock);
 		}
 		__ceph_remove_cap(cap);
 	}
@@ -2707,6 +2722,13 @@ static void handle_cap_import(struct ceph_mds_client *mdsc,
 		ci->i_cap_exporting_issued = 0;
 		ci->i_cap_exporting_mseq = 0;
 		ci->i_cap_exporting_mds = -1;
+
+		spin_lock(&mdsc->cap_dirty_lock);
+		if (!list_empty(&ci->i_dirty_item)) {
+			dout(" moving %p back to cap_dirty\n", inode);
+			list_move(&ci->i_dirty_item, &mdsc->cap_dirty);
+		}
+		spin_unlock(&mdsc->cap_dirty_lock);
 	} else {
 		dout("handle_cap_import inode %p ci %p mds%d mseq %d\n",
 		     inode, ci, mds, mseq);
@@ -2910,38 +2932,16 @@ void ceph_check_delayed_caps(struct ceph_mds_client *mdsc)
  */
 void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc)
 {
-	struct ceph_inode_info *ci, *nci = NULL;
-	struct inode *inode, *ninode = NULL;
-	struct list_head *p, *n;
+	struct ceph_inode_info *ci;
+	struct inode *inode;
 
 	dout("flush_dirty_caps\n");
 	spin_lock(&mdsc->cap_dirty_lock);
-	list_for_each_safe(p, n, &mdsc->cap_dirty) {
-		if (nci) {
-			ci = nci;
-			inode = ninode;
-			ci->i_ceph_flags &= ~CEPH_I_NOFLUSH;
-			dout("flush_dirty_caps inode %p (was next inode)\n",
-			     inode);
-		} else {
-			ci = list_entry(p, struct ceph_inode_info,
-					i_dirty_item);
-			inode = igrab(&ci->vfs_inode);
-			BUG_ON(!inode);
-			dout("flush_dirty_caps inode %p\n", inode);
-		}
-		if (n != &mdsc->cap_dirty) {
-			nci = list_entry(n, struct ceph_inode_info,
-					 i_dirty_item);
-			ninode = igrab(&nci->vfs_inode);
-			BUG_ON(!ninode);
-			nci->i_ceph_flags |= CEPH_I_NOFLUSH;
-			dout("flush_dirty_caps next inode %p, noflush\n",
-			     ninode);
-		} else {
-			nci = NULL;
-			ninode = NULL;
-		}
+	while (!list_empty(&mdsc->cap_dirty)) {
+		ci = list_first_entry(&mdsc->cap_dirty, struct ceph_inode_info,
+				      i_dirty_item);
+		inode = igrab(&ci->vfs_inode);
+		dout("flush_dirty_caps %p\n", inode);
 		spin_unlock(&mdsc->cap_dirty_lock);
 		if (inode) {
 			ceph_check_caps(ci, CHECK_CAPS_NODELAY|CHECK_CAPS_FLUSH,
@@ -2951,6 +2951,7 @@ void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc)
 		spin_lock(&mdsc->cap_dirty_lock);
 	}
 	spin_unlock(&mdsc->cap_dirty_lock);
+	dout("flush_dirty_caps done\n");
 }
 
 /*

fs/ceph/dir.c

@@ -360,7 +360,7 @@ static int ceph_readdir(struct file *filp, void *dirent, filldir_t filldir)
 	rinfo = &fi->last_readdir->r_reply_info;
 	dout("readdir frag %x num %d off %d chunkoff %d\n", frag,
 	     rinfo->dir_nr, off, fi->offset);
-	while (off - fi->offset >= 0 && off - fi->offset < rinfo->dir_nr) {
+	while (off >= fi->offset && off - fi->offset < rinfo->dir_nr) {
 		u64 pos = ceph_make_fpos(frag, off);
 		struct ceph_mds_reply_inode *in =
 			rinfo->dir_in[off - fi->offset].in;
@@ -1066,16 +1066,17 @@ static ssize_t ceph_read_dir(struct file *file, char __user *buf, size_t size,
 	struct inode *inode = file->f_dentry->d_inode;
 	struct ceph_inode_info *ci = ceph_inode(inode);
 	int left;
+	const int bufsize = 1024;
 
 	if (!ceph_test_mount_opt(ceph_sb_to_client(inode->i_sb), DIRSTAT))
 		return -EISDIR;
 
 	if (!cf->dir_info) {
-		cf->dir_info = kmalloc(1024, GFP_NOFS);
+		cf->dir_info = kmalloc(bufsize, GFP_NOFS);
 		if (!cf->dir_info)
 			return -ENOMEM;
 		cf->dir_info_len =
-			sprintf(cf->dir_info,
+			snprintf(cf->dir_info, bufsize,
 				"entries:   %20lld\n"
 				" files:    %20lld\n"
 				" subdirs:  %20lld\n"

fs/ceph/export.c

@@ -86,6 +86,7 @@ static int ceph_encode_fh(struct dentry *dentry, u32 *rawfh, int *max_len,
 static struct dentry *__fh_to_dentry(struct super_block *sb,
 				     struct ceph_nfs_fh *fh)
 {
+	struct ceph_mds_client *mdsc = ceph_sb_to_client(sb)->mdsc;
 	struct inode *inode;
 	struct dentry *dentry;
 	struct ceph_vino vino;
@@ -95,8 +96,24 @@ static struct dentry *__fh_to_dentry(struct super_block *sb,
 	vino.ino = fh->ino;
 	vino.snap = CEPH_NOSNAP;
 	inode = ceph_find_inode(sb, vino);
-	if (!inode)
-		return ERR_PTR(-ESTALE);
+	if (!inode) {
+		struct ceph_mds_request *req;
+
+		req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_LOOKUPINO,
+					       USE_ANY_MDS);
+		if (IS_ERR(req))
+			return ERR_CAST(req);
+
+		req->r_ino1 = vino;
+		req->r_num_caps = 1;
+		err = ceph_mdsc_do_request(mdsc, NULL, req);
+		inode = req->r_target_inode;
+		if (inode)
+			igrab(inode);
+		ceph_mdsc_put_request(req);
+		if (!inode)
+			return ERR_PTR(-ESTALE);
+	}
 
 	dentry = d_obtain_alias(inode);
 	if (IS_ERR(dentry)) {
@@ -148,8 +165,10 @@ static struct dentry *__cfh_to_dentry(struct super_block *sb,
 		snprintf(req->r_path2, 16, "%d", cfh->parent_name_hash);
 		req->r_num_caps = 1;
 		err = ceph_mdsc_do_request(mdsc, NULL, req);
+		inode = req->r_target_inode;
+		if (inode)
+			igrab(inode);
 		ceph_mdsc_put_request(req);
-		inode = ceph_find_inode(sb, vino);
 		if (!inode)
 			return ERR_PTR(err ? err : -ESTALE);
 	}

fs/ceph/mds_client.c

@@ -578,6 +578,7 @@ static void __register_request(struct ceph_mds_client *mdsc,
 	if (dir) {
 		struct ceph_inode_info *ci = ceph_inode(dir);
 
+		ihold(dir);
 		spin_lock(&ci->i_unsafe_lock);
 		req->r_unsafe_dir = dir;
 		list_add_tail(&req->r_unsafe_dir_item, &ci->i_unsafe_dirops);
@@ -598,6 +599,9 @@ static void __unregister_request(struct ceph_mds_client *mdsc,
 		spin_lock(&ci->i_unsafe_lock);
 		list_del_init(&req->r_unsafe_dir_item);
 		spin_unlock(&ci->i_unsafe_lock);
+
+		iput(req->r_unsafe_dir);
+		req->r_unsafe_dir = NULL;
 	}
 
 	ceph_mdsc_put_request(req);
@@ -2691,7 +2695,6 @@ static void handle_lease(struct ceph_mds_client *mdsc,
 {
 	struct super_block *sb = mdsc->fsc->sb;
 	struct inode *inode;
-	struct ceph_inode_info *ci;
 	struct dentry *parent, *dentry;
 	struct ceph_dentry_info *di;
 	int mds = session->s_mds;
@@ -2728,7 +2731,6 @@ static void handle_lease(struct ceph_mds_client *mdsc,
 		dout("handle_lease no inode %llx\n", vino.ino);
 		goto release;
 	}
-	ci = ceph_inode(inode);
 
 	/* dentry */
 	parent = d_find_alias(inode);
@@ -3002,6 +3004,7 @@ int ceph_mdsc_init(struct ceph_fs_client *fsc)
 	spin_lock_init(&mdsc->snap_flush_lock);
 	mdsc->cap_flush_seq = 0;
 	INIT_LIST_HEAD(&mdsc->cap_dirty);
+	INIT_LIST_HEAD(&mdsc->cap_dirty_migrating);
 	mdsc->num_cap_flushing = 0;
 	spin_lock_init(&mdsc->cap_dirty_lock);
 	init_waitqueue_head(&mdsc->cap_flushing_wq);

fs/ceph/mds_client.h

@@ -278,6 +278,7 @@ struct ceph_mds_client {
 
 	u64               cap_flush_seq;
 	struct list_head  cap_dirty;        /* inodes with dirty caps */
+	struct list_head  cap_dirty_migrating; /* ...that are migration... */
 	int               num_cap_flushing; /* # caps we are flushing */
 	spinlock_t        cap_dirty_lock;   /* protects above items */
 	wait_queue_head_t cap_flushing_wq;

include/linux/ceph/ceph_fs.h

@@ -313,6 +313,7 @@ enum {
 	CEPH_MDS_OP_GETATTR    = 0x00101,
 	CEPH_MDS_OP_LOOKUPHASH = 0x00102,
 	CEPH_MDS_OP_LOOKUPPARENT = 0x00103,
+	CEPH_MDS_OP_LOOKUPINO  = 0x00104,
 
 	CEPH_MDS_OP_SETXATTR   = 0x01105,
 	CEPH_MDS_OP_RMXATTR    = 0x01106,

net/ceph/messenger.c

@@ -76,7 +76,8 @@ const char *ceph_pr_addr(const struct sockaddr_storage *ss)
 		break;
 
 	default:
-		sprintf(s, "(unknown sockaddr family %d)", (int)ss->ss_family);
+		snprintf(s, MAX_ADDR_STR_LEN, "(unknown sockaddr family %d)",
+			 (int)ss->ss_family);
 	}
 
 	return s;
@@ -598,7 +599,7 @@ static void prepare_write_keepalive(struct ceph_connection *con)
  * Connection negotiation.
  */
 
-static void prepare_connect_authorizer(struct ceph_connection *con)
+static int prepare_connect_authorizer(struct ceph_connection *con)
 {
 	void *auth_buf;
 	int auth_len = 0;
@@ -612,13 +613,20 @@ static void prepare_connect_authorizer(struct ceph_connection *con)
 					 con->auth_retry);
 	mutex_lock(&con->mutex);
 
+	if (test_bit(CLOSED, &con->state) ||
+	    test_bit(OPENING, &con->state))
+		return -EAGAIN;
+
 	con->out_connect.authorizer_protocol = cpu_to_le32(auth_protocol);
 	con->out_connect.authorizer_len = cpu_to_le32(auth_len);
 
-	con->out_kvec[con->out_kvec_left].iov_base = auth_buf;
-	con->out_kvec[con->out_kvec_left].iov_len = auth_len;
-	con->out_kvec_left++;
-	con->out_kvec_bytes += auth_len;
+	if (auth_len) {
+		con->out_kvec[con->out_kvec_left].iov_base = auth_buf;
+		con->out_kvec[con->out_kvec_left].iov_len = auth_len;
+		con->out_kvec_left++;
+		con->out_kvec_bytes += auth_len;
+	}
+	return 0;
 }
 
 /*
@@ -640,9 +648,9 @@ static void prepare_write_banner(struct ceph_messenger *msgr,
 	set_bit(WRITE_PENDING, &con->state);
 }
 
-static void prepare_write_connect(struct ceph_messenger *msgr,
-				  struct ceph_connection *con,
-				  int after_banner)
+static int prepare_write_connect(struct ceph_messenger *msgr,
+				 struct ceph_connection *con,
+				 int after_banner)
 {
 	unsigned global_seq = get_global_seq(con->msgr, 0);
 	int proto;
@@ -683,7 +691,7 @@ static void prepare_write_connect(struct ceph_messenger *msgr,
 	con->out_more = 0;
 	set_bit(WRITE_PENDING, &con->state);
 
-	prepare_connect_authorizer(con);
+	return prepare_connect_authorizer(con);
 }
 
 
@@ -1065,8 +1073,10 @@ static void addr_set_port(struct sockaddr_storage *ss, int p)
 	switch (ss->ss_family) {
 	case AF_INET:
 		((struct sockaddr_in *)ss)->sin_port = htons(p);
+		break;
 	case AF_INET6:
 		((struct sockaddr_in6 *)ss)->sin6_port = htons(p);
+		break;
 	}
 }
 
@@ -1216,6 +1226,7 @@ static int process_connect(struct ceph_connection *con)
 	u64 sup_feat = con->msgr->supported_features;
 	u64 req_feat = con->msgr->required_features;
 	u64 server_feat = le64_to_cpu(con->in_reply.features);
+	int ret;
 
 	dout("process_connect on %p tag %d\n", con, (int)con->in_tag);
 
@@ -1250,7 +1261,9 @@ static int process_connect(struct ceph_connection *con)
 			return -1;
 		}
 		con->auth_retry = 1;
-		prepare_write_connect(con->msgr, con, 0);
+		ret = prepare_write_connect(con->msgr, con, 0);
+		if (ret < 0)
+			return ret;
 		prepare_read_connect(con);
 		break;
 
@@ -1277,6 +1290,9 @@ static int process_connect(struct ceph_connection *con)
 		if (con->ops->peer_reset)
 			con->ops->peer_reset(con);
 		mutex_lock(&con->mutex);
+		if (test_bit(CLOSED, &con->state) ||
+		    test_bit(OPENING, &con->state))
+			return -EAGAIN;
 		break;
 
 	case CEPH_MSGR_TAG_RETRY_SESSION:
@@ -1341,7 +1357,9 @@ static int process_connect(struct ceph_connection *con)
 		 * to WAIT.  This shouldn't happen if we are the
 		 * client.
 		 */
-		pr_err("process_connect peer connecting WAIT\n");
+		pr_err("process_connect got WAIT as client\n");
+		con->error_msg = "protocol error, got WAIT as client";
+		return -1;
 
 	default:
 		pr_err("connect protocol error, will retry\n");
@@ -1810,6 +1828,17 @@ static int try_read(struct ceph_connection *con)
 more:
 	dout("try_read tag %d in_base_pos %d\n", (int)con->in_tag,
 	     con->in_base_pos);
+
+	/*
+	 * process_connect and process_message drop and re-take
+	 * con->mutex.  make sure we handle a racing close or reopen.
+	 */
+	if (test_bit(CLOSED, &con->state) ||
+	    test_bit(OPENING, &con->state)) {
+		ret = -EAGAIN;
+		goto out;
+	}
+
 	if (test_bit(CONNECTING, &con->state)) {
 		if (!test_bit(NEGOTIATING, &con->state)) {
 			dout("try_read connecting\n");
@@ -1938,8 +1967,10 @@ static void con_work(struct work_struct *work)
 {
 	struct ceph_connection *con = container_of(work, struct ceph_connection,
 						   work.work);
+	int ret;
 
 	mutex_lock(&con->mutex);
+restart:
 	if (test_and_clear_bit(BACKOFF, &con->state)) {
 		dout("con_work %p backing off\n", con);
 		if (queue_delayed_work(ceph_msgr_wq, &con->work,
@@ -1969,18 +2000,31 @@ static void con_work(struct work_struct *work)
 		con_close_socket(con);
 	}
 
-	if (test_and_clear_bit(SOCK_CLOSED, &con->state) ||
-	    try_read(con) < 0 ||
-	    try_write(con) < 0) {
-		mutex_unlock(&con->mutex);
-		ceph_fault(con);     /* error/fault path */
-		goto done_unlocked;
-	}
+	if (test_and_clear_bit(SOCK_CLOSED, &con->state))
+		goto fault;
+
+	ret = try_read(con);
+	if (ret == -EAGAIN)
+		goto restart;
+	if (ret < 0)
+		goto fault;
+
+	ret = try_write(con);
+	if (ret == -EAGAIN)
+		goto restart;
+	if (ret < 0)
+		goto fault;
 
 done:
 	mutex_unlock(&con->mutex);
 done_unlocked:
 	con->ops->put(con);
+	return;
+
+fault:
+	mutex_unlock(&con->mutex);
+	ceph_fault(con);     /* error/fault path */
+	goto done_unlocked;
 }
 
 

net/ceph/osd_client.c

@@ -124,7 +124,7 @@ static void calc_layout(struct ceph_osd_client *osdc,
 	ceph_calc_raw_layout(osdc, layout, vino.snap, off,
 			     plen, &bno, req, op);
 
-	sprintf(req->r_oid, "%llx.%08llx", vino.ino, bno);
+	snprintf(req->r_oid, sizeof(req->r_oid), "%llx.%08llx", vino.ino, bno);
 	req->r_oid_len = strlen(req->r_oid);
 }
 
@@ -1421,6 +1421,15 @@ void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg)
 done:
 	downgrade_write(&osdc->map_sem);
 	ceph_monc_got_osdmap(&osdc->client->monc, osdc->osdmap->epoch);
+
+	/*
+	 * subscribe to subsequent osdmap updates if full to ensure
+	 * we find out when we are no longer full and stop returning
+	 * ENOSPC.
+	 */
+	if (ceph_osdmap_flag(osdc->osdmap, CEPH_OSDMAP_FULL))
+		ceph_monc_request_next_osdmap(&osdc->client->monc);
+
 	send_queued(osdc);
 	up_read(&osdc->map_sem);
 	wake_up_all(&osdc->client->auth_wq);
@@ -1677,8 +1686,14 @@ int ceph_osdc_start_request(struct ceph_osd_client *osdc,
 	 */
 	if (req->r_sent == 0) {
 		rc = __map_request(osdc, req);
-		if (rc < 0)
+		if (rc < 0) {
+			if (nofail) {
+				dout("osdc_start_request failed map, "
+				     " will retry %lld\n", req->r_tid);
+				rc = 0;
+			}
 			goto out_unlock;
+		}
 		if (req->r_osd == NULL) {
 			dout("send_request %p no up osds in pg\n", req);
 			ceph_monc_request_next_osdmap(&osdc->client->monc);

net/ceph/osdmap.c

@@ -765,7 +765,7 @@ struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
 	}
 
 	map->epoch++;
-	map->modified = map->modified;
+	map->modified = modified;
 	if (newcrush) {
 		if (map->crush)
 			crush_destroy(map->crush);
@@ -830,15 +830,20 @@ struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
 		map->osd_addr[osd] = addr;
 	}
 
-	/* new_down */
+	/* new_state */
 	ceph_decode_32_safe(p, end, len, bad);
 	while (len--) {
 		u32 osd;
+		u8 xorstate;
 		ceph_decode_32_safe(p, end, osd, bad);
+		xorstate = **(u8 **)p;
 		(*p)++;  /* clean flag */
-		pr_info("osd%d down\n", osd);
+		if (xorstate == 0)
+			xorstate = CEPH_OSD_UP;
+		if (xorstate & CEPH_OSD_UP)
+			pr_info("osd%d down\n", osd);
 		if (osd < map->max_osd)
-			map->osd_state[osd] &= ~CEPH_OSD_UP;
+			map->osd_state[osd] ^= xorstate;
 	}
 
 	/* new_weight */