NFS client updates for Linux 4.15

Stable bugfixes:
 - Revalidate "." and ".." correctly on open
 - Avoid RCU usage in tracepoints
 - Fix ugly referral attributes
 - Fix a typo in nomigration mount option
 - Revert "NFS: Move the flock open mode check into nfs_flock()"
 
 Features:
 - Implement a stronger send queue accounting system for NFS over RDMA
 - Switch some atomics to the new refcount_t type
 
 Other bugfixes and cleanups:
 - Clean up access mode bits
 - Remove special-case revalidations in nfs_opendir()
 - Improve invalidating NFS over RDMA memory for async operations that time out
 - Handle NFS over RDMA replies with a worqueue
 - Handle NFS over RDMA sends with a workqueue
 - Fix up replaying interrupted requests
 - Remove dead NFS over RDMA definitions
 - Update NFS over RDMA copyright information
 - Be more consistent with bool initialization and comparisons
 - Mark expected switch fall throughs
 - Various sunrpc tracepoint cleanups
 - Fix various OPEN races
 - Fix a typo in nfs_rename()
 - Use common error handling code in nfs_lock_and_join_request()
 - Check that some structures are properly cleaned up during net_exit()
 - Remove net pointer from dprintk()s
 -----BEGIN PGP SIGNATURE-----
 
 iQIzBAABCAAdFiEEnZ5MQTpR7cLU7KEp18tUv7ClQOsFAloPWGwACgkQ18tUv7Cl
 QOtMVhAAufCkDxqO2lmDH+0JyYUKMcoOMYtI8s2J1HrbEzTW/dVtI28fPAKEEd4m
 2JjNqnO516Jiv+g3E6eO4uunZRb4IB3AYT6YaTwmBFE+l7tpMdPb1xybOBP02Hji
 Y29kzLXwxxvnoxEqFalzCzV2BeRb2kAw6mayY9FxH6AfiEEQZfmxLCYgVuYa2jTC
 Z/B5E0GxAf28Aj0bIP8lLKbOkFijo851DB88UffEOZQGKUDlAd3GNUSSHb81Rj0N
 4ef7bKoGylkIpZ1PdTChdG1+RKqud02zrmQfmEwXui3eUwhOWy8hrKloNykqR5sj
 pgoDz79euAq4TDVyQKtutnbvVxfCcBeMYAXZhXkZLVcl+39in0kuLj4SxU5AmDhf
 ErnthG4W7jsLMM96kMvSTaoh4uwioviG1KmZfvuvUoMBSwtiX18hFTWtFKRD6x9e
 PNOqBdh8nkKYEFbEO4ksfYaWZJ5AuyFIQiIpj1gm+7sf039oN/zEuPV+jaEJG0oa
 Ef9IqHrQbbCUFYFjpBENr3HjU3igTTaxQ5iq+VYl4zg1pw6m6JTojqZ6qtQzqOYS
 O3N1ygeShsW934z8QcWjtEyeUXIB3JF9vUS3gEBgWPDyCltGXyq4Cq6Lod4s4JCb
 pWGI6wJLX1Fg6nq7cj0S4Or3QBgz2q8ZyBxssamhdvON/Ef5ccI=
 =2Zc1
 -----END PGP SIGNATURE-----

Merge tag 'nfs-for-4.15-1' of git://git.linux-nfs.org/projects/anna/linux-nfs

Pull NFS client updates from Anna Schumaker:
 "Stable bugfixes:
   - Revalidate "." and ".." correctly on open
   - Avoid RCU usage in tracepoints
   - Fix ugly referral attributes
   - Fix a typo in nomigration mount option
   - Revert "NFS: Move the flock open mode check into nfs_flock()"

  Features:
   - Implement a stronger send queue accounting system for NFS over RDMA
   - Switch some atomics to the new refcount_t type

  Other bugfixes and cleanups:
   - Clean up access mode bits
   - Remove special-case revalidations in nfs_opendir()
   - Improve invalidating NFS over RDMA memory for async operations that
     time out
   - Handle NFS over RDMA replies with a worqueue
   - Handle NFS over RDMA sends with a workqueue
   - Fix up replaying interrupted requests
   - Remove dead NFS over RDMA definitions
   - Update NFS over RDMA copyright information
   - Be more consistent with bool initialization and comparisons
   - Mark expected switch fall throughs
   - Various sunrpc tracepoint cleanups
   - Fix various OPEN races
   - Fix a typo in nfs_rename()
   - Use common error handling code in nfs_lock_and_join_request()
   - Check that some structures are properly cleaned up during
     net_exit()
   - Remove net pointer from dprintk()s"

* tag 'nfs-for-4.15-1' of git://git.linux-nfs.org/projects/anna/linux-nfs: (62 commits)
  NFS: Revert "NFS: Move the flock open mode check into nfs_flock()"
  NFS: Fix typo in nomigration mount option
  nfs: Fix ugly referral attributes
  NFS: super: mark expected switch fall-throughs
  sunrpc: remove net pointer from messages
  nfs: remove net pointer from messages
  sunrpc: exit_net cleanup check added
  nfs client: exit_net cleanup check added
  nfs/write: Use common error handling code in nfs_lock_and_join_requests()
  NFSv4: Replace closed stateids with the "invalid special stateid"
  NFSv4: nfs_set_open_stateid must not trigger state recovery for closed state
  NFSv4: Check the open stateid when searching for expired state
  NFSv4: Clean up nfs4_delegreturn_done
  NFSv4: cleanup nfs4_close_done
  NFSv4: Retry NFS4ERR_OLD_STATEID errors in layoutreturn
  pNFS: Retry NFS4ERR_OLD_STATEID errors in layoutreturn-on-close
  NFSv4: Don't try to CLOSE if the stateid 'other' field has changed
  NFSv4: Retry CLOSE and DELEGRETURN on NFS4ERR_OLD_STATEID.
  NFS: Fix a typo in nfs_rename()
  NFSv4: Fix open create exclusive when the server reboots
  ...

fs/nfs/cache_lib.c

@@ -67,7 +67,7 @@ out:
  */
 void nfs_cache_defer_req_put(struct nfs_cache_defer_req *dreq)
 {
-	if (atomic_dec_and_test(&dreq->count))
+	if (refcount_dec_and_test(&dreq->count))
 		kfree(dreq);
 }
 
@@ -87,7 +87,7 @@ static struct cache_deferred_req *nfs_dns_cache_defer(struct cache_req *req)
 
 	dreq = container_of(req, struct nfs_cache_defer_req, req);
 	dreq->deferred_req.revisit = nfs_dns_cache_revisit;
-	atomic_inc(&dreq->count);
+	refcount_inc(&dreq->count);
 
 	return &dreq->deferred_req;
 }
@@ -99,7 +99,7 @@ struct nfs_cache_defer_req *nfs_cache_defer_req_alloc(void)
 	dreq = kzalloc(sizeof(*dreq), GFP_KERNEL);
 	if (dreq) {
 		init_completion(&dreq->completion);
-		atomic_set(&dreq->count, 1);
+		refcount_set(&dreq->count, 1);
 		dreq->req.defer = nfs_dns_cache_defer;
 	}
 	return dreq;

fs/nfs/cache_lib.h

@@ -16,7 +16,7 @@ struct nfs_cache_defer_req {
 	struct cache_req req;
 	struct cache_deferred_req deferred_req;
 	struct completion completion;
-	atomic_t count;
+	refcount_t count;
 };
 
 extern int nfs_cache_upcall(struct cache_detail *cd, char *entry_name);

fs/nfs/callback.c

@@ -49,15 +49,15 @@ static int nfs4_callback_up_net(struct svc_serv *serv, struct net *net)
 	if (ret <= 0)
 		goto out_err;
 	nn->nfs_callback_tcpport = ret;
-	dprintk("NFS: Callback listener port = %u (af %u, net %p)\n",
+	dprintk("NFS: Callback listener port = %u (af %u, net %x)\n",
-			nn->nfs_callback_tcpport, PF_INET, net);
+		nn->nfs_callback_tcpport, PF_INET, net->ns.inum);
 
 	ret = svc_create_xprt(serv, "tcp", net, PF_INET6,
 				nfs_callback_set_tcpport, SVC_SOCK_ANONYMOUS);
 	if (ret > 0) {
 		nn->nfs_callback_tcpport6 = ret;
-		dprintk("NFS: Callback listener port = %u (af %u, net %p)\n",
+		dprintk("NFS: Callback listener port = %u (af %u, net %x\n",
-				nn->nfs_callback_tcpport6, PF_INET6, net);
+			nn->nfs_callback_tcpport6, PF_INET6, net->ns.inum);
 	} else if (ret != -EAFNOSUPPORT)
 		goto out_err;
 	return 0;
@@ -185,7 +185,7 @@ static void nfs_callback_down_net(u32 minorversion, struct svc_serv *serv, struc
 	if (--nn->cb_users[minorversion])
 		return;
 
-	dprintk("NFS: destroy per-net callback data; net=%p\n", net);
+	dprintk("NFS: destroy per-net callback data; net=%x\n", net->ns.inum);
 	svc_shutdown_net(serv, net);
 }
 
@@ -198,7 +198,7 @@ static int nfs_callback_up_net(int minorversion, struct svc_serv *serv,
 	if (nn->cb_users[minorversion]++)
 		return 0;
 
-	dprintk("NFS: create per-net callback data; net=%p\n", net);
+	dprintk("NFS: create per-net callback data; net=%x\n", net->ns.inum);
 
 	ret = svc_bind(serv, net);
 	if (ret < 0) {
@@ -223,7 +223,7 @@ err_socks:
 err_bind:
 	nn->cb_users[minorversion]--;
 	dprintk("NFS: Couldn't create callback socket: err = %d; "
-			"net = %p\n", ret, net);
+			"net = %x\n", ret, net->ns.inum);
 	return ret;
 }
 

fs/nfs/callback_proc.c

@@ -440,7 +440,7 @@ static bool referring_call_exists(struct nfs_client *clp,
 				  uint32_t nrclists,
 				  struct referring_call_list *rclists)
 {
-	bool status = 0;
+	bool status = false;
 	int i, j;
 	struct nfs4_session *session;
 	struct nfs4_slot_table *tbl;

fs/nfs/client.c

@@ -163,7 +163,7 @@ struct nfs_client *nfs_alloc_client(const struct nfs_client_initdata *cl_init)
 
 	clp->rpc_ops = clp->cl_nfs_mod->rpc_ops;
 
-	atomic_set(&clp->cl_count, 1);
+	refcount_set(&clp->cl_count, 1);
 	clp->cl_cons_state = NFS_CS_INITING;
 
 	memcpy(&clp->cl_addr, cl_init->addr, cl_init->addrlen);
@@ -269,7 +269,7 @@ void nfs_put_client(struct nfs_client *clp)
 
 	nn = net_generic(clp->cl_net, nfs_net_id);
 
-	if (atomic_dec_and_lock(&clp->cl_count, &nn->nfs_client_lock)) {
+	if (refcount_dec_and_lock(&clp->cl_count, &nn->nfs_client_lock)) {
 		list_del(&clp->cl_share_link);
 		nfs_cb_idr_remove_locked(clp);
 		spin_unlock(&nn->nfs_client_lock);
@@ -314,7 +314,7 @@ static struct nfs_client *nfs_match_client(const struct nfs_client_initdata *dat
 							   sap))
 				continue;
 
-		atomic_inc(&clp->cl_count);
+		refcount_inc(&clp->cl_count);
 		return clp;
 	}
 	return NULL;
@@ -1006,7 +1006,7 @@ struct nfs_server *nfs_clone_server(struct nfs_server *source,
 	/* Copy data from the source */
 	server->nfs_client = source->nfs_client;
 	server->destroy = source->destroy;
-	atomic_inc(&server->nfs_client->cl_count);
+	refcount_inc(&server->nfs_client->cl_count);
 	nfs_server_copy_userdata(server, source);
 
 	server->fsid = fattr->fsid;
@@ -1166,7 +1166,7 @@ static int nfs_server_list_show(struct seq_file *m, void *v)
 		   clp->rpc_ops->version,
 		   rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_HEX_ADDR),
 		   rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_HEX_PORT),
-		   atomic_read(&clp->cl_count),
+		   refcount_read(&clp->cl_count),
 		   clp->cl_hostname);
 	rcu_read_unlock();
 

fs/nfs/delegation.c

@@ -1040,6 +1040,33 @@ int nfs_delegations_present(struct nfs_client *clp)
 	return ret;
 }
 
+/**
+ * nfs4_refresh_delegation_stateid - Update delegation stateid seqid
+ * @dst: stateid to refresh
+ * @inode: inode to check
+ *
+ * Returns "true" and updates "dst->seqid" * if inode had a delegation
+ * that matches our delegation stateid. Otherwise "false" is returned.
+ */
+bool nfs4_refresh_delegation_stateid(nfs4_stateid *dst, struct inode *inode)
+{
+	struct nfs_delegation *delegation;
+	bool ret = false;
+	if (!inode)
+		goto out;
+
+	rcu_read_lock();
+	delegation = rcu_dereference(NFS_I(inode)->delegation);
+	if (delegation != NULL &&
+	    nfs4_stateid_match_other(dst, &delegation->stateid)) {
+		dst->seqid = delegation->stateid.seqid;
+		return ret;
+	}
+	rcu_read_unlock();
+out:
+	return ret;
+}
+
 /**
  * nfs4_copy_delegation_stateid - Copy inode's state ID information
  * @inode: inode to check

fs/nfs/delegation.h

@@ -62,6 +62,7 @@ int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4
 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid, fmode_t type);
 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid);
 bool nfs4_copy_delegation_stateid(struct inode *inode, fmode_t flags, nfs4_stateid *dst, struct rpc_cred **cred);
+bool nfs4_refresh_delegation_stateid(nfs4_stateid *dst, struct inode *inode);
 
 void nfs_mark_delegation_referenced(struct nfs_delegation *delegation);
 int nfs4_have_delegation(struct inode *inode, fmode_t flags);

fs/nfs/dir.c

@@ -118,13 +118,6 @@ nfs_opendir(struct inode *inode, struct file *filp)
 		goto out;
 	}
 	filp->private_data = ctx;
-	if (filp->f_path.dentry == filp->f_path.mnt->mnt_root) {
-		/* This is a mountpoint, so d_revalidate will never
-		 * have been called, so we need to refresh the
-		 * inode (for close-open consistency) ourselves.
-		 */
-		__nfs_revalidate_inode(NFS_SERVER(inode), inode);
-	}
 out:
 	put_rpccred(cred);
 	return res;
@@ -253,7 +246,7 @@ int nfs_readdir_search_for_pos(struct nfs_cache_array *array, nfs_readdir_descri
 	desc->cache_entry_index = index;
 	return 0;
 out_eof:
-	desc->eof = 1;
+	desc->eof = true;
 	return -EBADCOOKIE;
 }
 
@@ -307,7 +300,7 @@ int nfs_readdir_search_for_cookie(struct nfs_cache_array *array, nfs_readdir_des
 	if (array->eof_index >= 0) {
 		status = -EBADCOOKIE;
 		if (*desc->dir_cookie == array->last_cookie)
-			desc->eof = 1;
+			desc->eof = true;
 	}
 out:
 	return status;
@@ -761,7 +754,7 @@ int nfs_do_filldir(nfs_readdir_descriptor_t *desc)
 		ent = &array->array[i];
 		if (!dir_emit(desc->ctx, ent->string.name, ent->string.len,
 		    nfs_compat_user_ino64(ent->ino), ent->d_type)) {
-			desc->eof = 1;
+			desc->eof = true;
 			break;
 		}
 		desc->ctx->pos++;
@@ -773,7 +766,7 @@ int nfs_do_filldir(nfs_readdir_descriptor_t *desc)
 			ctx->duped = 1;
 	}
 	if (array->eof_index >= 0)
-		desc->eof = 1;
+		desc->eof = true;
 
 	kunmap(desc->page);
 	cache_page_release(desc);
@@ -873,7 +866,7 @@ static int nfs_readdir(struct file *file, struct dir_context *ctx)
 		if (res == -EBADCOOKIE) {
 			res = 0;
 			/* This means either end of directory */
-			if (*desc->dir_cookie && desc->eof == 0) {
+			if (*desc->dir_cookie && !desc->eof) {
 				/* Or that the server has 'lost' a cookie */
 				res = uncached_readdir(desc);
 				if (res == 0)
@@ -1241,8 +1234,7 @@ static int nfs_weak_revalidate(struct dentry *dentry, unsigned int flags)
 		return 0;
 	}
 
-	if (nfs_mapping_need_revalidate_inode(inode))
+	error = nfs_lookup_verify_inode(inode, flags);
-		error = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
 	dfprintk(LOOKUPCACHE, "NFS: %s: inode %lu is %s\n",
 			__func__, inode->i_ino, error ? "invalid" : "valid");
 	return !error;
@@ -1393,6 +1385,7 @@ static int nfs4_lookup_revalidate(struct dentry *, unsigned int);
 
 const struct dentry_operations nfs4_dentry_operations = {
 	.d_revalidate	= nfs4_lookup_revalidate,
+	.d_weak_revalidate	= nfs_weak_revalidate,
 	.d_delete	= nfs_dentry_delete,
 	.d_iput		= nfs_dentry_iput,
 	.d_automount	= nfs_d_automount,
@@ -2064,7 +2057,7 @@ out:
 		 * should mark the directories for revalidation.
 		 */
 		d_move(old_dentry, new_dentry);
-		nfs_set_verifier(new_dentry,
+		nfs_set_verifier(old_dentry,
 					nfs_save_change_attribute(new_dir));
 	} else if (error == -ENOENT)
 		nfs_dentry_handle_enoent(old_dentry);
@@ -2369,15 +2362,15 @@ void nfs_access_add_cache(struct inode *inode, struct nfs_access_entry *set)
 }
 EXPORT_SYMBOL_GPL(nfs_access_add_cache);
 
-#define NFS_MAY_READ (NFS4_ACCESS_READ)
+#define NFS_MAY_READ (NFS_ACCESS_READ)
-#define NFS_MAY_WRITE (NFS4_ACCESS_MODIFY | \
+#define NFS_MAY_WRITE (NFS_ACCESS_MODIFY | \
-		NFS4_ACCESS_EXTEND | \
+		NFS_ACCESS_EXTEND | \
-		NFS4_ACCESS_DELETE)
+		NFS_ACCESS_DELETE)
-#define NFS_FILE_MAY_WRITE (NFS4_ACCESS_MODIFY | \
+#define NFS_FILE_MAY_WRITE (NFS_ACCESS_MODIFY | \
-		NFS4_ACCESS_EXTEND)
+		NFS_ACCESS_EXTEND)
 #define NFS_DIR_MAY_WRITE NFS_MAY_WRITE
-#define NFS_MAY_LOOKUP (NFS4_ACCESS_LOOKUP)
+#define NFS_MAY_LOOKUP (NFS_ACCESS_LOOKUP)
-#define NFS_MAY_EXECUTE (NFS4_ACCESS_EXECUTE)
+#define NFS_MAY_EXECUTE (NFS_ACCESS_EXECUTE)
 static int
 nfs_access_calc_mask(u32 access_result, umode_t umode)
 {
@@ -2425,9 +2418,14 @@ static int nfs_do_access(struct inode *inode, struct rpc_cred *cred, int mask)
 	if (!may_block)
 		goto out;
 
-	/* Be clever: ask server to check for all possible rights */
+	/*
-	cache.mask = NFS_MAY_LOOKUP | NFS_MAY_EXECUTE
+	 * Determine which access bits we want to ask for...
-		     | NFS_MAY_WRITE | NFS_MAY_READ;
+	 */
+	cache.mask = NFS_ACCESS_READ | NFS_ACCESS_MODIFY | NFS_ACCESS_EXTEND;
+	if (S_ISDIR(inode->i_mode))
+		cache.mask |= NFS_ACCESS_DELETE | NFS_ACCESS_LOOKUP;
+	else
+		cache.mask |= NFS_ACCESS_EXECUTE;
 	cache.cred = cred;
 	status = NFS_PROTO(inode)->access(inode, &cache);
 	if (status != 0) {

fs/nfs/file.c

@@ -829,23 +829,9 @@ int nfs_flock(struct file *filp, int cmd, struct file_lock *fl)
 	if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FLOCK)
 		is_local = 1;
 
-	/*
+	/* We're simulating flock() locks using posix locks on the server */
-	 * VFS doesn't require the open mode to match a flock() lock's type.
+	if (fl->fl_type == F_UNLCK)
-	 * NFS, however, may simulate flock() locking with posix locking which
-	 * requires the open mode to match the lock type.
-	 */
-	switch (fl->fl_type) {
-	case F_UNLCK:
 		return do_unlk(filp, cmd, fl, is_local);
-	case F_RDLCK:
-		if (!(filp->f_mode & FMODE_READ))
-			return -EBADF;
-		break;
-	case F_WRLCK:
-		if (!(filp->f_mode & FMODE_WRITE))
-			return -EBADF;
-	}
-
 	return do_setlk(filp, cmd, fl, is_local);
 }
 EXPORT_SYMBOL_GPL(nfs_flock);

fs/nfs/filelayout/filelayout.c

@@ -471,10 +471,10 @@ filelayout_read_pagelist(struct nfs_pgio_header *hdr)
 		return PNFS_NOT_ATTEMPTED;
 
 	dprintk("%s USE DS: %s cl_count %d\n", __func__,
-		ds->ds_remotestr, atomic_read(&ds->ds_clp->cl_count));
+		ds->ds_remotestr, refcount_read(&ds->ds_clp->cl_count));
 
 	/* No multipath support. Use first DS */
-	atomic_inc(&ds->ds_clp->cl_count);
+	refcount_inc(&ds->ds_clp->cl_count);
 	hdr->ds_clp = ds->ds_clp;
 	hdr->ds_commit_idx = idx;
 	fh = nfs4_fl_select_ds_fh(lseg, j);
@@ -515,10 +515,10 @@ filelayout_write_pagelist(struct nfs_pgio_header *hdr, int sync)
 
 	dprintk("%s ino %lu sync %d req %zu@%llu DS: %s cl_count %d\n",
 		__func__, hdr->inode->i_ino, sync, (size_t) hdr->args.count,
-		offset, ds->ds_remotestr, atomic_read(&ds->ds_clp->cl_count));
+		offset, ds->ds_remotestr, refcount_read(&ds->ds_clp->cl_count));
 
 	hdr->pgio_done_cb = filelayout_write_done_cb;
-	atomic_inc(&ds->ds_clp->cl_count);
+	refcount_inc(&ds->ds_clp->cl_count);
 	hdr->ds_clp = ds->ds_clp;
 	hdr->ds_commit_idx = idx;
 	fh = nfs4_fl_select_ds_fh(lseg, j);
@@ -1064,9 +1064,9 @@ static int filelayout_initiate_commit(struct nfs_commit_data *data, int how)
 		goto out_err;
 
 	dprintk("%s ino %lu, how %d cl_count %d\n", __func__,
-		data->inode->i_ino, how, atomic_read(&ds->ds_clp->cl_count));
+		data->inode->i_ino, how, refcount_read(&ds->ds_clp->cl_count));
 	data->commit_done_cb = filelayout_commit_done_cb;
-	atomic_inc(&ds->ds_clp->cl_count);
+	refcount_inc(&ds->ds_clp->cl_count);
 	data->ds_clp = ds->ds_clp;
 	fh = select_ds_fh_from_commit(lseg, data->ds_commit_index);
 	if (fh)

fs/nfs/flexfilelayout/flexfilelayout.c

@@ -187,7 +187,7 @@ ff_layout_add_mirror(struct pnfs_layout_hdr *lo,
 			continue;
 		if (!ff_mirror_match_fh(mirror, pos))
 			continue;
-		if (atomic_inc_not_zero(&pos->ref)) {
+		if (refcount_inc_not_zero(&pos->ref)) {
 			spin_unlock(&inode->i_lock);
 			return pos;
 		}
@@ -218,7 +218,7 @@ static struct nfs4_ff_layout_mirror *ff_layout_alloc_mirror(gfp_t gfp_flags)
 	mirror = kzalloc(sizeof(*mirror), gfp_flags);
 	if (mirror != NULL) {
 		spin_lock_init(&mirror->lock);
-		atomic_set(&mirror->ref, 1);
+		refcount_set(&mirror->ref, 1);
 		INIT_LIST_HEAD(&mirror->mirrors);
 	}
 	return mirror;
@@ -242,7 +242,7 @@ static void ff_layout_free_mirror(struct nfs4_ff_layout_mirror *mirror)
 
 static void ff_layout_put_mirror(struct nfs4_ff_layout_mirror *mirror)
 {
-	if (mirror != NULL && atomic_dec_and_test(&mirror->ref))
+	if (mirror != NULL && refcount_dec_and_test(&mirror->ref))
 		ff_layout_free_mirror(mirror);
 }
 
@@ -1726,10 +1726,10 @@ ff_layout_read_pagelist(struct nfs_pgio_header *hdr)
 	vers = nfs4_ff_layout_ds_version(lseg, idx);
 
 	dprintk("%s USE DS: %s cl_count %d vers %d\n", __func__,
-		ds->ds_remotestr, atomic_read(&ds->ds_clp->cl_count), vers);
+		ds->ds_remotestr, refcount_read(&ds->ds_clp->cl_count), vers);
 
 	hdr->pgio_done_cb = ff_layout_read_done_cb;
-	atomic_inc(&ds->ds_clp->cl_count);
+	refcount_inc(&ds->ds_clp->cl_count);
 	hdr->ds_clp = ds->ds_clp;
 	fh = nfs4_ff_layout_select_ds_fh(lseg, idx);
 	if (fh)
@@ -1785,11 +1785,11 @@ ff_layout_write_pagelist(struct nfs_pgio_header *hdr, int sync)
 
 	dprintk("%s ino %lu sync %d req %zu@%llu DS: %s cl_count %d vers %d\n",
 		__func__, hdr->inode->i_ino, sync, (size_t) hdr->args.count,
-		offset, ds->ds_remotestr, atomic_read(&ds->ds_clp->cl_count),
+		offset, ds->ds_remotestr, refcount_read(&ds->ds_clp->cl_count),
 		vers);
 
 	hdr->pgio_done_cb = ff_layout_write_done_cb;
-	atomic_inc(&ds->ds_clp->cl_count);
+	refcount_inc(&ds->ds_clp->cl_count);
 	hdr->ds_clp = ds->ds_clp;
 	hdr->ds_commit_idx = idx;
 	fh = nfs4_ff_layout_select_ds_fh(lseg, idx);
@@ -1863,11 +1863,11 @@ static int ff_layout_initiate_commit(struct nfs_commit_data *data, int how)
 	vers = nfs4_ff_layout_ds_version(lseg, idx);
 
 	dprintk("%s ino %lu, how %d cl_count %d vers %d\n", __func__,
-		data->inode->i_ino, how, atomic_read(&ds->ds_clp->cl_count),
+		data->inode->i_ino, how, refcount_read(&ds->ds_clp->cl_count),
 		vers);
 	data->commit_done_cb = ff_layout_commit_done_cb;
 	data->cred = ds_cred;
-	atomic_inc(&ds->ds_clp->cl_count);
+	refcount_inc(&ds->ds_clp->cl_count);
 	data->ds_clp = ds->ds_clp;
 	fh = select_ds_fh_from_commit(lseg, data->ds_commit_index);
 	if (fh)
@@ -2286,7 +2286,7 @@ ff_layout_mirror_prepare_stats(struct pnfs_layout_hdr *lo,
 		if (!test_and_clear_bit(NFS4_FF_MIRROR_STAT_AVAIL, &mirror->flags))
 			continue;
 		/* mirror refcount put in cleanup_layoutstats */
-		if (!atomic_inc_not_zero(&mirror->ref))
+		if (!refcount_inc_not_zero(&mirror->ref))
 			continue;
 		dev = &mirror->mirror_ds->id_node; 
 		memcpy(&devinfo->dev_id, &dev->deviceid, NFS4_DEVICEID4_SIZE);

fs/nfs/flexfilelayout/flexfilelayout.h

@@ -14,6 +14,7 @@
 #define FF_FLAGS_NO_IO_THRU_MDS  2
 #define FF_FLAGS_NO_READ_IO      4
 
+#include <linux/refcount.h>
 #include "../pnfs.h"
 
 /* XXX: Let's filter out insanely large mirror count for now to avoid oom
@@ -82,7 +83,7 @@ struct nfs4_ff_layout_mirror {
 	nfs4_stateid			stateid;
 	struct rpc_cred	__rcu		*ro_cred;
 	struct rpc_cred	__rcu		*rw_cred;
-	atomic_t			ref;
+	refcount_t			ref;
 	spinlock_t			lock;
 	unsigned long			flags;
 	struct nfs4_ff_layoutstat	read_stat;

fs/nfs/inode.c

@@ -783,7 +783,7 @@ EXPORT_SYMBOL_GPL(nfs_getattr);
 
 static void nfs_init_lock_context(struct nfs_lock_context *l_ctx)
 {
-	atomic_set(&l_ctx->count, 1);
+	refcount_set(&l_ctx->count, 1);
 	l_ctx->lockowner = current->files;
 	INIT_LIST_HEAD(&l_ctx->list);
 	atomic_set(&l_ctx->io_count, 0);
@@ -797,7 +797,7 @@ static struct nfs_lock_context *__nfs_find_lock_context(struct nfs_open_context
 	do {
 		if (pos->lockowner != current->files)
 			continue;
-		atomic_inc(&pos->count);
+		refcount_inc(&pos->count);
 		return pos;
 	} while ((pos = list_entry(pos->list.next, typeof(*pos), list)) != head);
 	return NULL;
@@ -836,7 +836,7 @@ void nfs_put_lock_context(struct nfs_lock_context *l_ctx)
 	struct nfs_open_context *ctx = l_ctx->open_context;
 	struct inode *inode = d_inode(ctx->dentry);
 
-	if (!atomic_dec_and_lock(&l_ctx->count, &inode->i_lock))
+	if (!refcount_dec_and_lock(&l_ctx->count, &inode->i_lock))
 		return;
 	list_del(&l_ctx->list);
 	spin_unlock(&inode->i_lock);
@@ -913,7 +913,7 @@ EXPORT_SYMBOL_GPL(alloc_nfs_open_context);
 struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
 {
 	if (ctx != NULL)
-		atomic_inc(&ctx->lock_context.count);
+		refcount_inc(&ctx->lock_context.count);
 	return ctx;
 }
 EXPORT_SYMBOL_GPL(get_nfs_open_context);
@@ -924,11 +924,11 @@ static void __put_nfs_open_context(struct nfs_open_context *ctx, int is_sync)
 	struct super_block *sb = ctx->dentry->d_sb;
 
 	if (!list_empty(&ctx->list)) {
-		if (!atomic_dec_and_lock(&ctx->lock_context.count, &inode->i_lock))
+		if (!refcount_dec_and_lock(&ctx->lock_context.count, &inode->i_lock))
 			return;
 		list_del(&ctx->list);
 		spin_unlock(&inode->i_lock);
-	} else if (!atomic_dec_and_test(&ctx->lock_context.count))
+	} else if (!refcount_dec_and_test(&ctx->lock_context.count))
 		return;
 	if (inode != NULL)
 		NFS_PROTO(inode)->close_context(ctx, is_sync);
@@ -2084,8 +2084,12 @@ static int nfs_net_init(struct net *net)
 
 static void nfs_net_exit(struct net *net)
 {
+	struct nfs_net *nn = net_generic(net, nfs_net_id);
+
 	nfs_fs_proc_net_exit(net);
 	nfs_cleanup_cb_ident_idr(net);
+	WARN_ON_ONCE(!list_empty(&nn->nfs_client_list));
+	WARN_ON_ONCE(!list_empty(&nn->nfs_volume_list));
 }
 
 static struct pernet_operations nfs_net_ops = {

fs/nfs/nfs3proc.c

@@ -188,6 +188,7 @@ static int nfs3_proc_access(struct inode *inode, struct nfs_access_entry *entry)
 {
 	struct nfs3_accessargs	arg = {
 		.fh		= NFS_FH(inode),
+		.access		= entry->mask,
 	};
 	struct nfs3_accessres	res;
 	struct rpc_message msg = {
@@ -196,25 +197,9 @@ static int nfs3_proc_access(struct inode *inode, struct nfs_access_entry *entry)
 		.rpc_resp	= &res,
 		.rpc_cred	= entry->cred,
 	};
-	int mode = entry->mask;
 	int status = -ENOMEM;
 
 	dprintk("NFS call  access\n");
-
-	if (mode & MAY_READ)
-		arg.access |= NFS3_ACCESS_READ;
-	if (S_ISDIR(inode->i_mode)) {
-		if (mode & MAY_WRITE)
-			arg.access |= NFS3_ACCESS_MODIFY | NFS3_ACCESS_EXTEND | NFS3_ACCESS_DELETE;
-		if (mode & MAY_EXEC)
-			arg.access |= NFS3_ACCESS_LOOKUP;
-	} else {
-		if (mode & MAY_WRITE)
-			arg.access |= NFS3_ACCESS_MODIFY | NFS3_ACCESS_EXTEND;
-		if (mode & MAY_EXEC)
-			arg.access |= NFS3_ACCESS_EXECUTE;
-	}
-
 	res.fattr = nfs_alloc_fattr();
 	if (res.fattr == NULL)
 		goto out;

fs/nfs/nfs4_fs.h

@@ -145,7 +145,7 @@ struct nfs4_lock_state {
 	unsigned long		ls_flags;
 	struct nfs_seqid_counter	ls_seqid;
 	nfs4_stateid		ls_stateid;
-	atomic_t		ls_count;
+	refcount_t		ls_count;
 	fl_owner_t		ls_owner;
 };
 
@@ -162,6 +162,7 @@ enum {
 	NFS_STATE_POSIX_LOCKS,		/* Posix locks are supported */
 	NFS_STATE_RECOVERY_FAILED,	/* OPEN stateid state recovery failed */
 	NFS_STATE_MAY_NOTIFY_LOCK,	/* server may CB_NOTIFY_LOCK */
+	NFS_STATE_CHANGE_WAIT,		/* A state changing operation is outstanding */
 };
 
 struct nfs4_state {
@@ -185,6 +186,8 @@ struct nfs4_state {
 	unsigned int n_rdwr;		/* Number of read/write references */
 	fmode_t state;			/* State on the server (R,W, or RW) */
 	atomic_t count;
+
+	wait_queue_head_t waitq;
 };
 
 
@@ -458,6 +461,10 @@ extern int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl);
 extern int nfs4_select_rw_stateid(struct nfs4_state *, fmode_t,
 		const struct nfs_lock_context *, nfs4_stateid *,
 		struct rpc_cred **);
+extern bool nfs4_refresh_open_stateid(nfs4_stateid *dst,
+		struct nfs4_state *state);
+extern bool nfs4_copy_open_stateid(nfs4_stateid *dst,
+		struct nfs4_state *state);
 
 extern struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask);
 extern int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task);
@@ -465,7 +472,7 @@ extern void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid);
 extern void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid);
 extern void nfs_release_seqid(struct nfs_seqid *seqid);
 extern void nfs_free_seqid(struct nfs_seqid *seqid);
-extern int nfs4_setup_sequence(const struct nfs_client *client,
+extern int nfs4_setup_sequence(struct nfs_client *client,
 				struct nfs4_sequence_args *args,
 				struct nfs4_sequence_res *res,
 				struct rpc_task *task);
@@ -475,6 +482,7 @@ extern int nfs4_sequence_done(struct rpc_task *task,
 extern void nfs4_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp);
 
 extern const nfs4_stateid zero_stateid;
+extern const nfs4_stateid invalid_stateid;
 
 /* nfs4super.c */
 struct nfs_mount_info;

fs/nfs/nfs4client.c

@@ -483,7 +483,7 @@ static int nfs4_match_client(struct nfs_client  *pos,  struct nfs_client *new,
 	 * ID and serverowner fields.  Wait for CREATE_SESSION
 	 * to finish. */
 	if (pos->cl_cons_state > NFS_CS_READY) {
-		atomic_inc(&pos->cl_count);
+		refcount_inc(&pos->cl_count);
 		spin_unlock(&nn->nfs_client_lock);
 
 		nfs_put_client(*prev);
@@ -559,7 +559,7 @@ int nfs40_walk_client_list(struct nfs_client *new,
 		 * way that a SETCLIENTID_CONFIRM to pos can succeed is
 		 * if new and pos point to the same server:
 		 */
-		atomic_inc(&pos->cl_count);
+		refcount_inc(&pos->cl_count);
 		spin_unlock(&nn->nfs_client_lock);
 
 		nfs_put_client(prev);
@@ -715,7 +715,7 @@ int nfs41_walk_client_list(struct nfs_client *new,
 			continue;
 
 found:
-		atomic_inc(&pos->cl_count);
+		refcount_inc(&pos->cl_count);
 		*result = pos;
 		status = 0;
 		break;
@@ -749,7 +749,7 @@ nfs4_find_client_ident(struct net *net, int cb_ident)
 	spin_lock(&nn->nfs_client_lock);
 	clp = idr_find(&nn->cb_ident_idr, cb_ident);
 	if (clp)
-		atomic_inc(&clp->cl_count);
+		refcount_inc(&clp->cl_count);
 	spin_unlock(&nn->nfs_client_lock);
 	return clp;
 }
@@ -793,7 +793,7 @@ nfs4_find_client_sessionid(struct net *net, const struct sockaddr *addr,
 
 	spin_lock(&nn->nfs_client_lock);
 	list_for_each_entry(clp, &nn->nfs_client_list, cl_share_link) {
-		if (nfs4_cb_match_client(addr, clp, minorversion) == false)
+		if (!nfs4_cb_match_client(addr, clp, minorversion))
 			continue;
 
 		if (!nfs4_has_session(clp))
@@ -804,7 +804,7 @@ nfs4_find_client_sessionid(struct net *net, const struct sockaddr *addr,
 		    sid->data, NFS4_MAX_SESSIONID_LEN) != 0)
 			continue;
 
-		atomic_inc(&clp->cl_count);
+		refcount_inc(&clp->cl_count);
 		spin_unlock(&nn->nfs_client_lock);
 		return clp;
 	}

fs/nfs/nfs4proc.c

@@ -96,6 +96,10 @@ static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
 			    struct nfs_open_context *ctx, struct nfs4_label *ilabel,
 			    struct nfs4_label *olabel);
 #ifdef CONFIG_NFS_V4_1
+static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
+		struct rpc_cred *cred,
+		struct nfs4_slot *slot,
+		bool is_privileged);
 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 		struct rpc_cred *);
 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
@@ -254,15 +258,12 @@ const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 };
 
 const u32 nfs4_fs_locations_bitmap[3] = {
-	FATTR4_WORD0_TYPE
+	FATTR4_WORD0_CHANGE
-	| FATTR4_WORD0_CHANGE
 	| FATTR4_WORD0_SIZE
 	| FATTR4_WORD0_FSID
 	| FATTR4_WORD0_FILEID
 	| FATTR4_WORD0_FS_LOCATIONS,
-	FATTR4_WORD1_MODE
+	FATTR4_WORD1_OWNER
-	| FATTR4_WORD1_NUMLINKS
-	| FATTR4_WORD1_OWNER
 	| FATTR4_WORD1_OWNER_GROUP
 	| FATTR4_WORD1_RAWDEV
 	| FATTR4_WORD1_SPACE_USED
@@ -644,13 +645,14 @@ static int nfs40_sequence_done(struct rpc_task *task,
 
 #if defined(CONFIG_NFS_V4_1)
 
-static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
+static void nfs41_release_slot(struct nfs4_slot *slot)
 {
 	struct nfs4_session *session;
 	struct nfs4_slot_table *tbl;
-	struct nfs4_slot *slot = res->sr_slot;
 	bool send_new_highest_used_slotid = false;
 
+	if (!slot)
+		return;
 	tbl = slot->table;
 	session = tbl->session;
 
@@ -676,13 +678,18 @@ static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 		send_new_highest_used_slotid = false;
 out_unlock:
 	spin_unlock(&tbl->slot_tbl_lock);
-	res->sr_slot = NULL;
 	if (send_new_highest_used_slotid)
 		nfs41_notify_server(session->clp);
 	if (waitqueue_active(&tbl->slot_waitq))
 		wake_up_all(&tbl->slot_waitq);
 }
 
+static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
+{
+	nfs41_release_slot(res->sr_slot);
+	res->sr_slot = NULL;
+}
+
 static int nfs41_sequence_process(struct rpc_task *task,
 		struct nfs4_sequence_res *res)
 {
@@ -710,13 +717,6 @@ static int nfs41_sequence_process(struct rpc_task *task,
 	/* Check the SEQUENCE operation status */
 	switch (res->sr_status) {
 	case 0:
-		/* If previous op on slot was interrupted and we reused
-		 * the seq# and got a reply from the cache, then retry
-		 */
-		if (task->tk_status == -EREMOTEIO && interrupted) {
-			++slot->seq_nr;
-			goto retry_nowait;
-		}
 		/* Update the slot's sequence and clientid lease timer */
 		slot->seq_done = 1;
 		clp = session->clp;
@@ -750,16 +750,16 @@ static int nfs41_sequence_process(struct rpc_task *task,
 		 * The slot id we used was probably retired. Try again
 		 * using a different slot id.
 		 */
+		if (slot->seq_nr < slot->table->target_highest_slotid)
+			goto session_recover;
 		goto retry_nowait;
 	case -NFS4ERR_SEQ_MISORDERED:
 		/*
 		 * Was the last operation on this sequence interrupted?
 		 * If so, retry after bumping the sequence number.
 		 */
-		if (interrupted) {
+		if (interrupted)
-			++slot->seq_nr;
+			goto retry_new_seq;
-			goto retry_nowait;
-		}
 		/*
 		 * Could this slot have been previously retired?
 		 * If so, then the server may be expecting seq_nr = 1!
@@ -768,10 +768,11 @@ static int nfs41_sequence_process(struct rpc_task *task,
 			slot->seq_nr = 1;
 			goto retry_nowait;
 		}
-		break;
+		goto session_recover;
 	case -NFS4ERR_SEQ_FALSE_RETRY:
-		++slot->seq_nr;
+		if (interrupted)
-		goto retry_nowait;
+			goto retry_new_seq;
+		goto session_recover;
 	default:
 		/* Just update the slot sequence no. */
 		slot->seq_done = 1;
@@ -781,6 +782,11 @@ out:
 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 out_noaction:
 	return ret;
+session_recover:
+	nfs4_schedule_session_recovery(session, res->sr_status);
+	goto retry_nowait;
+retry_new_seq:
+	++slot->seq_nr;
 retry_nowait:
 	if (rpc_restart_call_prepare(task)) {
 		nfs41_sequence_free_slot(res);
@@ -857,6 +863,17 @@ static const struct rpc_call_ops nfs41_call_sync_ops = {
 	.rpc_call_done = nfs41_call_sync_done,
 };
 
+static void
+nfs4_sequence_process_interrupted(struct nfs_client *client,
+		struct nfs4_slot *slot, struct rpc_cred *cred)
+{
+	struct rpc_task *task;
+
+	task = _nfs41_proc_sequence(client, cred, slot, true);
+	if (!IS_ERR(task))
+		rpc_put_task_async(task);
+}
+
 #else	/* !CONFIG_NFS_V4_1 */
 
 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
@@ -877,9 +894,34 @@ int nfs4_sequence_done(struct rpc_task *task,
 }
 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 
+static void
+nfs4_sequence_process_interrupted(struct nfs_client *client,
+		struct nfs4_slot *slot, struct rpc_cred *cred)
+{
+	WARN_ON_ONCE(1);
+	slot->interrupted = 0;
+}
+
 #endif	/* !CONFIG_NFS_V4_1 */
 
-int nfs4_setup_sequence(const struct nfs_client *client,
+static
+void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
+		struct nfs4_sequence_res *res,
+		struct nfs4_slot *slot)
+{
+	if (!slot)
+		return;
+	slot->privileged = args->sa_privileged ? 1 : 0;
+	args->sa_slot = slot;
+
+	res->sr_slot = slot;
+	res->sr_timestamp = jiffies;
+	res->sr_status_flags = 0;
+	res->sr_status = 1;
+
+}
+
+int nfs4_setup_sequence(struct nfs_client *client,
 			struct nfs4_sequence_args *args,
 			struct nfs4_sequence_res *res,
 			struct rpc_task *task)
@@ -897,29 +939,28 @@ int nfs4_setup_sequence(const struct nfs_client *client,
 		task->tk_timeout = 0;
 	}
 
-	spin_lock(&tbl->slot_tbl_lock);
+	for (;;) {
-	/* The state manager will wait until the slot table is empty */
+		spin_lock(&tbl->slot_tbl_lock);
-	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
+		/* The state manager will wait until the slot table is empty */
-		goto out_sleep;
+		if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
+			goto out_sleep;
 
-	slot = nfs4_alloc_slot(tbl);
+		slot = nfs4_alloc_slot(tbl);
-	if (IS_ERR(slot)) {
+		if (IS_ERR(slot)) {
-		/* Try again in 1/4 second */
+			/* Try again in 1/4 second */
-		if (slot == ERR_PTR(-ENOMEM))
+			if (slot == ERR_PTR(-ENOMEM))
-			task->tk_timeout = HZ >> 2;
+				task->tk_timeout = HZ >> 2;
-		goto out_sleep;
+			goto out_sleep;
+		}
+		spin_unlock(&tbl->slot_tbl_lock);
+
+		if (likely(!slot->interrupted))
+			break;
+		nfs4_sequence_process_interrupted(client,
+				slot, task->tk_msg.rpc_cred);
 	}
-	spin_unlock(&tbl->slot_tbl_lock);
 
-	slot->privileged = args->sa_privileged ? 1 : 0;
+	nfs4_sequence_attach_slot(args, res, slot);
-	args->sa_slot = slot;
-
-	res->sr_slot = slot;
-	if (session) {
-		res->sr_timestamp = jiffies;
-		res->sr_status_flags = 0;
-		res->sr_status = 1;
-	}
 
 	trace_nfs4_setup_sequence(session, args);
 out_start:
@@ -1044,6 +1085,12 @@ struct nfs4_opendata {
 	int rpc_status;
 };
 
+struct nfs4_open_createattrs {
+	struct nfs4_label *label;
+	struct iattr *sattr;
+	const __u32 verf[2];
+};
+
 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
 		int err, struct nfs4_exception *exception)
 {
@@ -1113,8 +1160,7 @@ static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 
 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
-		const struct iattr *attrs,
+		const struct nfs4_open_createattrs *c,
-		struct nfs4_label *label,
 		enum open_claim_type4 claim,
 		gfp_t gfp_mask)
 {
@@ -1122,6 +1168,7 @@ static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 	struct inode *dir = d_inode(parent);
 	struct nfs_server *server = NFS_SERVER(dir);
 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
+	struct nfs4_label *label = (c != NULL) ? c->label : NULL;
 	struct nfs4_opendata *p;
 
 	p = kzalloc(sizeof(*p), gfp_mask);
@@ -1187,15 +1234,11 @@ static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 		p->o_arg.fh = NFS_FH(d_inode(dentry));
 	}
-	if (attrs != NULL && attrs->ia_valid != 0) {
+	if (c != NULL && c->sattr != NULL && c->sattr->ia_valid != 0) {
-		__u32 verf[2];
-
 		p->o_arg.u.attrs = &p->attrs;
-		memcpy(&p->attrs, attrs, sizeof(p->attrs));
+		memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
 
-		verf[0] = jiffies;
+		memcpy(p->o_arg.u.verifier.data, c->verf,
-		verf[1] = current->pid;
-		memcpy(p->o_arg.u.verifier.data, verf,
 				sizeof(p->o_arg.u.verifier.data));
 	}
 	p->c_arg.fh = &p->o_res.fh;
@@ -1334,6 +1377,25 @@ static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
 }
 #endif /* CONFIG_NFS_V4_1 */
 
+static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
+{
+	if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
+		wake_up_all(&state->waitq);
+}
+
+static void nfs_state_log_out_of_order_open_stateid(struct nfs4_state *state,
+		const nfs4_stateid *stateid)
+{
+	u32 state_seqid = be32_to_cpu(state->open_stateid.seqid);
+	u32 stateid_seqid = be32_to_cpu(stateid->seqid);
+
+	if (stateid_seqid == state_seqid + 1U ||
+	    (stateid_seqid == 1U && state_seqid == 0xffffffffU))
+		nfs_state_log_update_open_stateid(state);
+	else
+		set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
+}
+
 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
 {
 	struct nfs_client *clp = state->owner->so_server->nfs_client;
@@ -1349,18 +1411,32 @@ static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
 		nfs4_state_mark_reclaim_nograce(clp, state);
 }
 
+/*
+ * Check for whether or not the caller may update the open stateid
+ * to the value passed in by stateid.
+ *
+ * Note: This function relies heavily on the server implementing
+ * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
+ * correctly.
+ * i.e. The stateid seqids have to be initialised to 1, and
+ * are then incremented on every state transition.
+ */
 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
-		const nfs4_stateid *stateid, nfs4_stateid *freeme)
+		const nfs4_stateid *stateid)
 {
-	if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
+	if (test_bit(NFS_OPEN_STATE, &state->flags) == 0 ||
-		return true;
+	    !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
-	if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
+		if (stateid->seqid == cpu_to_be32(1))
-		nfs4_stateid_copy(freeme, &state->open_stateid);
+			nfs_state_log_update_open_stateid(state);
-		nfs_test_and_clear_all_open_stateid(state);
+		else
+			set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
 		return true;
 	}
-	if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
+
+	if (nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
+		nfs_state_log_out_of_order_open_stateid(state, stateid);
 		return true;
+	}
 	return false;
 }
 
@@ -1399,11 +1475,14 @@ static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
 	if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
 	    !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
 		nfs_resync_open_stateid_locked(state);
-		return;
+		goto out;
 	}
 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 		nfs4_stateid_copy(&state->stateid, stateid);
 	nfs4_stateid_copy(&state->open_stateid, stateid);
+	trace_nfs4_open_stateid_update(state->inode, stateid, 0);
+out:
+	nfs_state_log_update_open_stateid(state);
 }
 
 static void nfs_clear_open_stateid(struct nfs4_state *state,
@@ -1420,29 +1499,60 @@ static void nfs_clear_open_stateid(struct nfs4_state *state,
 }
 
 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
-		const nfs4_stateid *stateid, fmode_t fmode,
+		const nfs4_stateid *stateid, nfs4_stateid *freeme)
-		nfs4_stateid *freeme)
 {
-	switch (fmode) {
+	DEFINE_WAIT(wait);
-		case FMODE_READ:
+	int status = 0;
-			set_bit(NFS_O_RDONLY_STATE, &state->flags);
+	for (;;) {
+
+		if (!nfs_need_update_open_stateid(state, stateid))
+			return;
+		if (!test_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
 			break;
-		case FMODE_WRITE:
+		if (status)
-			set_bit(NFS_O_WRONLY_STATE, &state->flags);
 			break;
-		case FMODE_READ|FMODE_WRITE:
+		/* Rely on seqids for serialisation with NFSv4.0 */
-			set_bit(NFS_O_RDWR_STATE, &state->flags);
+		if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
+			break;
+
+		prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
+		/*
+		 * Ensure we process the state changes in the same order
+		 * in which the server processed them by delaying the
+		 * update of the stateid until we are in sequence.
+		 */
+		write_sequnlock(&state->seqlock);
+		spin_unlock(&state->owner->so_lock);
+		rcu_read_unlock();
+		trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
+		if (!signal_pending(current)) {
+			if (schedule_timeout(5*HZ) == 0)
+				status = -EAGAIN;
+			else
+				status = 0;
+		} else
+			status = -EINTR;
+		finish_wait(&state->waitq, &wait);
+		rcu_read_lock();
+		spin_lock(&state->owner->so_lock);
+		write_seqlock(&state->seqlock);
 	}
-	if (!nfs_need_update_open_stateid(state, stateid, freeme))
+
-		return;
+	if (test_bit(NFS_OPEN_STATE, &state->flags) &&
+	    !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
+		nfs4_stateid_copy(freeme, &state->open_stateid);
+		nfs_test_and_clear_all_open_stateid(state);
+	}
+
 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 		nfs4_stateid_copy(&state->stateid, stateid);
 	nfs4_stateid_copy(&state->open_stateid, stateid);
+	trace_nfs4_open_stateid_update(state->inode, stateid, status);
+	nfs_state_log_update_open_stateid(state);
 }
 
-static void __update_open_stateid(struct nfs4_state *state,
+static void nfs_state_set_open_stateid(struct nfs4_state *state,
 		const nfs4_stateid *open_stateid,
-		const nfs4_stateid *deleg_stateid,
 		fmode_t fmode,
 		nfs4_stateid *freeme)
 {
@@ -1450,17 +1560,34 @@ static void __update_open_stateid(struct nfs4_state *state,
 	 * Protect the call to nfs4_state_set_mode_locked and
 	 * serialise the stateid update
 	 */
-	spin_lock(&state->owner->so_lock);
 	write_seqlock(&state->seqlock);
-	if (deleg_stateid != NULL) {
+	nfs_set_open_stateid_locked(state, open_stateid, freeme);
-		nfs4_stateid_copy(&state->stateid, deleg_stateid);
+	switch (fmode) {
-		set_bit(NFS_DELEGATED_STATE, &state->flags);
+	case FMODE_READ:
+		set_bit(NFS_O_RDONLY_STATE, &state->flags);
+		break;
+	case FMODE_WRITE:
+		set_bit(NFS_O_WRONLY_STATE, &state->flags);
+		break;
+	case FMODE_READ|FMODE_WRITE:
+		set_bit(NFS_O_RDWR_STATE, &state->flags);
 	}
-	if (open_stateid != NULL)
+	set_bit(NFS_OPEN_STATE, &state->flags);
-		nfs_set_open_stateid_locked(state, open_stateid, fmode, freeme);
+	write_sequnlock(&state->seqlock);
+}
+
+static void nfs_state_set_delegation(struct nfs4_state *state,
+		const nfs4_stateid *deleg_stateid,
+		fmode_t fmode)
+{
+	/*
+	 * Protect the call to nfs4_state_set_mode_locked and
+	 * serialise the stateid update
+	 */
+	write_seqlock(&state->seqlock);
+	nfs4_stateid_copy(&state->stateid, deleg_stateid);
+	set_bit(NFS_DELEGATED_STATE, &state->flags);
 	write_sequnlock(&state->seqlock);
-	update_open_stateflags(state, fmode);
-	spin_unlock(&state->owner->so_lock);
 }
 
 static int update_open_stateid(struct nfs4_state *state,
@@ -1478,6 +1605,12 @@ static int update_open_stateid(struct nfs4_state *state,
 	fmode &= (FMODE_READ|FMODE_WRITE);
 
 	rcu_read_lock();
+	spin_lock(&state->owner->so_lock);
+	if (open_stateid != NULL) {
+		nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
+		ret = 1;
+	}
+
 	deleg_cur = rcu_dereference(nfsi->delegation);
 	if (deleg_cur == NULL)
 		goto no_delegation;
@@ -1494,18 +1627,16 @@ static int update_open_stateid(struct nfs4_state *state,
 		goto no_delegation_unlock;
 
 	nfs_mark_delegation_referenced(deleg_cur);
-	__update_open_stateid(state, open_stateid, &deleg_cur->stateid,
+	nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
-			fmode, &freeme);
 	ret = 1;
 no_delegation_unlock:
 	spin_unlock(&deleg_cur->lock);
 no_delegation:
+	if (ret)
+		update_open_stateflags(state, fmode);
+	spin_unlock(&state->owner->so_lock);
 	rcu_read_unlock();
 
-	if (!ret && open_stateid != NULL) {
-		__update_open_stateid(state, open_stateid, NULL, fmode, &freeme);
-		ret = 1;
-	}
 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
 		nfs4_schedule_state_manager(clp);
 	if (freeme.type != 0)
@@ -1761,7 +1892,7 @@ static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context
 	struct nfs4_opendata *opendata;
 
 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
-			NULL, NULL, claim, GFP_NOFS);
+			NULL, claim, GFP_NOFS);
 	if (opendata == NULL)
 		return ERR_PTR(-ENOMEM);
 	opendata->state = state;
@@ -2518,7 +2649,7 @@ static int nfs41_check_expired_locks(struct nfs4_state *state)
 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
 			struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
 
-			atomic_inc(&lsp->ls_count);
+			refcount_inc(&lsp->ls_count);
 			spin_unlock(&state->state_lock);
 
 			nfs4_put_lock_state(prev);
@@ -2692,8 +2823,7 @@ out:
 static int _nfs4_do_open(struct inode *dir,
 			struct nfs_open_context *ctx,
 			int flags,
-			struct iattr *sattr,
+			const struct nfs4_open_createattrs *c,
-			struct nfs4_label *label,
 			int *opened)
 {
 	struct nfs4_state_owner  *sp;
@@ -2705,6 +2835,8 @@ static int _nfs4_do_open(struct inode *dir,
 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
 	fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
+	struct iattr *sattr = c->sattr;
+	struct nfs4_label *label = c->label;
 	struct nfs4_label *olabel = NULL;
 	int status;
 
@@ -2723,8 +2855,8 @@ static int _nfs4_do_open(struct inode *dir,
 	status = -ENOMEM;
 	if (d_really_is_positive(dentry))
 		claim = NFS4_OPEN_CLAIM_FH;
-	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
+	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
-			label, claim, GFP_KERNEL);
+			c, claim, GFP_KERNEL);
 	if (opendata == NULL)
 		goto err_put_state_owner;
 
@@ -2805,10 +2937,18 @@ static struct nfs4_state *nfs4_do_open(struct inode *dir,
 	struct nfs_server *server = NFS_SERVER(dir);
 	struct nfs4_exception exception = { };
 	struct nfs4_state *res;
+	struct nfs4_open_createattrs c = {
+		.label = label,
+		.sattr = sattr,
+		.verf = {
+			[0] = (__u32)jiffies,
+			[1] = (__u32)current->pid,
+		},
+	};
 	int status;
 
 	do {
-		status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
+		status = _nfs4_do_open(dir, ctx, flags, &c, opened);
 		res = ctx->state;
 		trace_nfs4_open_file(ctx, flags, status);
 		if (status == 0)
@@ -3024,18 +3164,20 @@ static void nfs4_close_done(struct rpc_task *task, void *data)
 			calldata->arg.lr_args = NULL;
 			calldata->res.lr_res = NULL;
 			break;
+		case -NFS4ERR_OLD_STATEID:
+			if (nfs4_refresh_layout_stateid(&calldata->arg.lr_args->stateid,
+						calldata->inode))
+				goto lr_restart;
+			/* Fallthrough */
 		case -NFS4ERR_ADMIN_REVOKED:
 		case -NFS4ERR_DELEG_REVOKED:
 		case -NFS4ERR_EXPIRED:
 		case -NFS4ERR_BAD_STATEID:
-		case -NFS4ERR_OLD_STATEID:
 		case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
 		case -NFS4ERR_WRONG_CRED:
 			calldata->arg.lr_args = NULL;
 			calldata->res.lr_res = NULL;
-			calldata->res.lr_ret = 0;
+			goto lr_restart;
-			rpc_restart_call_prepare(task);
-			return;
 		}
 	}
 
@@ -3051,39 +3193,43 @@ static void nfs4_close_done(struct rpc_task *task, void *data)
 			if (calldata->arg.bitmask != NULL) {
 				calldata->arg.bitmask = NULL;
 				calldata->res.fattr = NULL;
-				task->tk_status = 0;
+				goto out_restart;
-				rpc_restart_call_prepare(task);
-				goto out_release;
 
 			}
 			break;
+		case -NFS4ERR_OLD_STATEID:
+			/* Did we race with OPEN? */
+			if (nfs4_refresh_open_stateid(&calldata->arg.stateid,
+						state))
+				goto out_restart;
+			goto out_release;
 		case -NFS4ERR_ADMIN_REVOKED:
 		case -NFS4ERR_STALE_STATEID:
 		case -NFS4ERR_EXPIRED:
 			nfs4_free_revoked_stateid(server,
 					&calldata->arg.stateid,
 					task->tk_msg.rpc_cred);
-		case -NFS4ERR_OLD_STATEID:
+			/* Fallthrough */
 		case -NFS4ERR_BAD_STATEID:
-			if (!nfs4_stateid_match(&calldata->arg.stateid,
+			break;
-						&state->open_stateid)) {
-				rpc_restart_call_prepare(task);
-				goto out_release;
-			}
-			if (calldata->arg.fmode == 0)
-				break;
 		default:
-			if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
+			if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
-				rpc_restart_call_prepare(task);
+				goto out_restart;
-				goto out_release;
-			}
 	}
 	nfs_clear_open_stateid(state, &calldata->arg.stateid,
 			res_stateid, calldata->arg.fmode);
 out_release:
+	task->tk_status = 0;
 	nfs_release_seqid(calldata->arg.seqid);
 	nfs_refresh_inode(calldata->inode, &calldata->fattr);
 	dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
+	return;
+lr_restart:
+	calldata->res.lr_ret = 0;
+out_restart:
+	task->tk_status = 0;
+	rpc_restart_call_prepare(task);
+	goto out_release;
 }
 
 static void nfs4_close_prepare(struct rpc_task *task, void *data)
@@ -3103,7 +3249,6 @@ static void nfs4_close_prepare(struct rpc_task *task, void *data)
 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
-	nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
 	/* Calculate the change in open mode */
 	calldata->arg.fmode = 0;
 	if (state->n_rdwr == 0) {
@@ -3121,7 +3266,7 @@ static void nfs4_close_prepare(struct rpc_task *task, void *data)
 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
 
 	if (!nfs4_valid_open_stateid(state) ||
-	    test_bit(NFS_OPEN_STATE, &state->flags) == 0)
+	    !nfs4_refresh_open_stateid(&calldata->arg.stateid, state))
 		call_close = 0;
 	spin_unlock(&state->owner->so_lock);
 
@@ -3215,6 +3360,8 @@ int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
 	calldata->inode = state->inode;
 	calldata->state = state;
 	calldata->arg.fh = NFS_FH(state->inode);
+	if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
+		goto out_free_calldata;
 	/* Serialization for the sequence id */
 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
 	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
@@ -3889,6 +4036,7 @@ static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry
 	struct nfs4_accessargs args = {
 		.fh = NFS_FH(inode),
 		.bitmask = server->cache_consistency_bitmask,
+		.access = entry->mask,
 	};
 	struct nfs4_accessres res = {
 		.server = server,
@@ -3899,26 +4047,8 @@ static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry
 		.rpc_resp = &res,
 		.rpc_cred = entry->cred,
 	};
-	int mode = entry->mask;
 	int status = 0;
 
-	/*
-	 * Determine which access bits we want to ask for...
-	 */
-	if (mode & MAY_READ)
-		args.access |= NFS4_ACCESS_READ;
-	if (S_ISDIR(inode->i_mode)) {
-		if (mode & MAY_WRITE)
-			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
-		if (mode & MAY_EXEC)
-			args.access |= NFS4_ACCESS_LOOKUP;
-	} else {
-		if (mode & MAY_WRITE)
-			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
-		if (mode & MAY_EXEC)
-			args.access |= NFS4_ACCESS_EXECUTE;
-	}
-
 	res.fattr = nfs_alloc_fattr();
 	if (res.fattr == NULL)
 		return -ENOMEM;
@@ -4843,7 +4973,7 @@ static void nfs4_renew_release(void *calldata)
 	struct nfs4_renewdata *data = calldata;
 	struct nfs_client *clp = data->client;
 
-	if (atomic_read(&clp->cl_count) > 1)
+	if (refcount_read(&clp->cl_count) > 1)
 		nfs4_schedule_state_renewal(clp);
 	nfs_put_client(clp);
 	kfree(data);
@@ -4891,7 +5021,7 @@ static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred,
 
 	if (renew_flags == 0)
 		return 0;
-	if (!atomic_inc_not_zero(&clp->cl_count))
+	if (!refcount_inc_not_zero(&clp->cl_count))
 		return -EIO;
 	data = kmalloc(sizeof(*data), GFP_NOFS);
 	if (data == NULL) {
@@ -5643,18 +5773,20 @@ static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
 			data->args.lr_args = NULL;
 			data->res.lr_res = NULL;
 			break;
+		case -NFS4ERR_OLD_STATEID:
+			if (nfs4_refresh_layout_stateid(&data->args.lr_args->stateid,
+						data->inode))
+				goto lr_restart;
+			/* Fallthrough */
 		case -NFS4ERR_ADMIN_REVOKED:
 		case -NFS4ERR_DELEG_REVOKED:
 		case -NFS4ERR_EXPIRED:
 		case -NFS4ERR_BAD_STATEID:
-		case -NFS4ERR_OLD_STATEID:
 		case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
 		case -NFS4ERR_WRONG_CRED:
 			data->args.lr_args = NULL;
 			data->res.lr_res = NULL;
-			data->res.lr_ret = 0;
+			goto lr_restart;
-			rpc_restart_call_prepare(task);
-			return;
 		}
 	}
 
@@ -5668,27 +5800,36 @@ static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
 		nfs4_free_revoked_stateid(data->res.server,
 				data->args.stateid,
 				task->tk_msg.rpc_cred);
+		/* Fallthrough */
 	case -NFS4ERR_BAD_STATEID:
-	case -NFS4ERR_OLD_STATEID:
 	case -NFS4ERR_STALE_STATEID:
 		task->tk_status = 0;
 		break;
+	case -NFS4ERR_OLD_STATEID:
+		if (nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
+			goto out_restart;
+		task->tk_status = 0;
+		break;
 	case -NFS4ERR_ACCESS:
 		if (data->args.bitmask) {
 			data->args.bitmask = NULL;
 			data->res.fattr = NULL;
-			task->tk_status = 0;
+			goto out_restart;
-			rpc_restart_call_prepare(task);
-			return;
 		}
+		/* Fallthrough */
 	default:
 		if (nfs4_async_handle_error(task, data->res.server,
 					    NULL, NULL) == -EAGAIN) {
-			rpc_restart_call_prepare(task);
+			goto out_restart;
-			return;
 		}
 	}
 	data->rpc_status = task->tk_status;
+	return;
+lr_restart:
+	data->res.lr_ret = 0;
+out_restart:
+	task->tk_status = 0;
+	rpc_restart_call_prepare(task);
 }
 
 static void nfs4_delegreturn_release(void *calldata)
@@ -5896,7 +6037,7 @@ static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
 	p->arg.seqid = seqid;
 	p->res.seqid = seqid;
 	p->lsp = lsp;
-	atomic_inc(&lsp->ls_count);
+	refcount_inc(&lsp->ls_count);
 	/* Ensure we don't close file until we're done freeing locks! */
 	p->ctx = get_nfs_open_context(ctx);
 	p->l_ctx = nfs_get_lock_context(ctx);
@@ -6112,7 +6253,7 @@ static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
 	p->res.lock_seqid = p->arg.lock_seqid;
 	p->lsp = lsp;
 	p->server = server;
-	atomic_inc(&lsp->ls_count);
+	refcount_inc(&lsp->ls_count);
 	p->ctx = get_nfs_open_context(ctx);
 	memcpy(&p->fl, fl, sizeof(p->fl));
 	return p;
@@ -6568,6 +6709,20 @@ nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
 	    !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
 		return -ENOLCK;
 
+	/*
+	 * Don't rely on the VFS having checked the file open mode,
+	 * since it won't do this for flock() locks.
+	 */
+	switch (request->fl_type) {
+	case F_RDLCK:
+		if (!(filp->f_mode & FMODE_READ))
+			return -EBADF;
+		break;
+	case F_WRLCK:
+		if (!(filp->f_mode & FMODE_WRITE))
+			return -EBADF;
+	}
+
 	status = nfs4_set_lock_state(state, request);
 	if (status != 0)
 		return status;
@@ -6763,9 +6918,7 @@ static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
 				   struct page *page)
 {
 	struct nfs_server *server = NFS_SERVER(dir);
-	u32 bitmask[3] = {
+	u32 bitmask[3];
-		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
-	};
 	struct nfs4_fs_locations_arg args = {
 		.dir_fh = NFS_FH(dir),
 		.name = name,
@@ -6784,12 +6937,15 @@ static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
 
 	dprintk("%s: start\n", __func__);
 
+	bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
+	bitmask[1] = nfs4_fattr_bitmap[1];
+
 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
 	 * is not supported */
 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
-		bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
+		bitmask[0] &= ~FATTR4_WORD0_FILEID;
 	else
-		bitmask[0] |= FATTR4_WORD0_FILEID;
+		bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
 
 	nfs_fattr_init(&fs_locations->fattr);
 	fs_locations->server = server;
@@ -7472,7 +7628,7 @@ nfs4_run_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
 	struct nfs41_exchange_id_data *calldata;
 	int status;
 
-	if (!atomic_inc_not_zero(&clp->cl_count))
+	if (!refcount_inc_not_zero(&clp->cl_count))
 		return ERR_PTR(-EIO);
 
 	status = -ENOMEM;
@@ -8072,7 +8228,7 @@ static void nfs41_sequence_release(void *data)
 	struct nfs4_sequence_data *calldata = data;
 	struct nfs_client *clp = calldata->clp;
 
-	if (atomic_read(&clp->cl_count) > 1)
+	if (refcount_read(&clp->cl_count) > 1)
 		nfs4_schedule_state_renewal(clp);
 	nfs_put_client(clp);
 	kfree(calldata);
@@ -8101,7 +8257,7 @@ static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
 	trace_nfs4_sequence(clp, task->tk_status);
 	if (task->tk_status < 0) {
 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
-		if (atomic_read(&clp->cl_count) == 1)
+		if (refcount_read(&clp->cl_count) == 1)
 			goto out;
 
 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
@@ -8135,6 +8291,7 @@ static const struct rpc_call_ops nfs41_sequence_ops = {
 
 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
 		struct rpc_cred *cred,
+		struct nfs4_slot *slot,
 		bool is_privileged)
 {
 	struct nfs4_sequence_data *calldata;
@@ -8148,15 +8305,18 @@ static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
 		.callback_ops = &nfs41_sequence_ops,
 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
 	};
+	struct rpc_task *ret;
 
-	if (!atomic_inc_not_zero(&clp->cl_count))
+	ret = ERR_PTR(-EIO);
-		return ERR_PTR(-EIO);
+	if (!refcount_inc_not_zero(&clp->cl_count))
+		goto out_err;
+
+	ret = ERR_PTR(-ENOMEM);
 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
-	if (calldata == NULL) {
+	if (calldata == NULL)
-		nfs_put_client(clp);
+		goto out_put_clp;
-		return ERR_PTR(-ENOMEM);
-	}
 	nfs4_init_sequence(&calldata->args, &calldata->res, 0);
+	nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
 	if (is_privileged)
 		nfs4_set_sequence_privileged(&calldata->args);
 	msg.rpc_argp = &calldata->args;
@@ -8164,7 +8324,15 @@ static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
 	calldata->clp = clp;
 	task_setup_data.callback_data = calldata;
 
-	return rpc_run_task(&task_setup_data);
+	ret = rpc_run_task(&task_setup_data);
+	if (IS_ERR(ret))
+		goto out_err;
+	return ret;
+out_put_clp:
+	nfs_put_client(clp);
+out_err:
+	nfs41_release_slot(slot);
+	return ret;
 }
 
 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
@@ -8174,7 +8342,7 @@ static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cr
 
 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
 		return -EAGAIN;
-	task = _nfs41_proc_sequence(clp, cred, false);
+	task = _nfs41_proc_sequence(clp, cred, NULL, false);
 	if (IS_ERR(task))
 		ret = PTR_ERR(task);
 	else
@@ -8188,7 +8356,7 @@ static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
 	struct rpc_task *task;
 	int ret;
 
-	task = _nfs41_proc_sequence(clp, cred, true);
+	task = _nfs41_proc_sequence(clp, cred, NULL, true);
 	if (IS_ERR(task)) {
 		ret = PTR_ERR(task);
 		goto out;
@@ -8588,18 +8756,27 @@ static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
 
 	server = NFS_SERVER(lrp->args.inode);
 	switch (task->tk_status) {
+	case -NFS4ERR_OLD_STATEID:
+		if (nfs4_refresh_layout_stateid(&lrp->args.stateid,
+					lrp->args.inode))
+			goto out_restart;
+		/* Fallthrough */
 	default:
 		task->tk_status = 0;
+		/* Fallthrough */
 	case 0:
 		break;
 	case -NFS4ERR_DELAY:
 		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
 			break;
-		nfs4_sequence_free_slot(&lrp->res.seq_res);
+		goto out_restart;
-		rpc_restart_call_prepare(task);
-		return;
 	}
 	dprintk("<-- %s\n", __func__);
+	return;
+out_restart:
+	task->tk_status = 0;
+	nfs4_sequence_free_slot(&lrp->res.seq_res);
+	rpc_restart_call_prepare(task);
 }
 
 static void nfs4_layoutreturn_release(void *calldata)

fs/nfs/nfs4state.c

@@ -69,6 +69,14 @@ const nfs4_stateid zero_stateid = {
 	{ .data = { 0 } },
 	.type = NFS4_SPECIAL_STATEID_TYPE,
 };
+const nfs4_stateid invalid_stateid = {
+	{
+		.seqid = cpu_to_be32(0xffffffffU),
+		.other = { 0 },
+	},
+	.type = NFS4_INVALID_STATEID_TYPE,
+};
+
 static DEFINE_MUTEX(nfs_clid_init_mutex);
 
 int nfs4_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
@@ -645,6 +653,7 @@ nfs4_alloc_open_state(void)
 	INIT_LIST_HEAD(&state->lock_states);
 	spin_lock_init(&state->state_lock);
 	seqlock_init(&state->seqlock);
+	init_waitqueue_head(&state->waitq);
 	return state;
 }
 
@@ -825,7 +834,7 @@ __nfs4_find_lock_state(struct nfs4_state *state,
 			ret = pos;
 	}
 	if (ret)
-		atomic_inc(&ret->ls_count);
+		refcount_inc(&ret->ls_count);
 	return ret;
 }
 
@@ -843,7 +852,7 @@ static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, f
 	if (lsp == NULL)
 		return NULL;
 	nfs4_init_seqid_counter(&lsp->ls_seqid);
-	atomic_set(&lsp->ls_count, 1);
+	refcount_set(&lsp->ls_count, 1);
 	lsp->ls_state = state;
 	lsp->ls_owner = fl_owner;
 	lsp->ls_seqid.owner_id = ida_simple_get(&server->lockowner_id, 0, 0, GFP_NOFS);
@@ -907,7 +916,7 @@ void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
 	if (lsp == NULL)
 		return;
 	state = lsp->ls_state;
-	if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock))
+	if (!refcount_dec_and_lock(&lsp->ls_count, &state->state_lock))
 		return;
 	list_del(&lsp->ls_locks);
 	if (list_empty(&state->lock_states))
@@ -927,7 +936,7 @@ static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
 	struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
 
 	dst->fl_u.nfs4_fl.owner = lsp;
-	atomic_inc(&lsp->ls_count);
+	refcount_inc(&lsp->ls_count);
 }
 
 static void nfs4_fl_release_lock(struct file_lock *fl)
@@ -985,18 +994,39 @@ out:
 	return ret;
 }
 
-static void nfs4_copy_open_stateid(nfs4_stateid *dst, struct nfs4_state *state)
+bool nfs4_refresh_open_stateid(nfs4_stateid *dst, struct nfs4_state *state)
 {
+	bool ret;
+	int seq;
+
+	do {
+		ret = false;
+		seq = read_seqbegin(&state->seqlock);
+		if (nfs4_state_match_open_stateid_other(state, dst)) {
+			dst->seqid = state->open_stateid.seqid;
+			ret = true;
+		}
+	} while (read_seqretry(&state->seqlock, seq));
+	return ret;
+}
+
+bool nfs4_copy_open_stateid(nfs4_stateid *dst, struct nfs4_state *state)
+{
+	bool ret;
 	const nfs4_stateid *src;
 	int seq;
 
 	do {
+		ret = false;
 		src = &zero_stateid;
 		seq = read_seqbegin(&state->seqlock);
-		if (test_bit(NFS_OPEN_STATE, &state->flags))
+		if (test_bit(NFS_OPEN_STATE, &state->flags)) {
 			src = &state->open_stateid;
+			ret = true;
+		}
 		nfs4_stateid_copy(dst, src);
 	} while (read_seqretry(&state->seqlock, seq));
+	return ret;
 }
 
 /*
@@ -1177,7 +1207,7 @@ void nfs4_schedule_state_manager(struct nfs_client *clp)
 	if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
 		return;
 	__module_get(THIS_MODULE);
-	atomic_inc(&clp->cl_count);
+	refcount_inc(&clp->cl_count);
 
 	/* The rcu_read_lock() is not strictly necessary, as the state
 	 * manager is the only thread that ever changes the rpc_xprt
@@ -1269,7 +1299,7 @@ int nfs4_wait_clnt_recover(struct nfs_client *clp)
 
 	might_sleep();
 
-	atomic_inc(&clp->cl_count);
+	refcount_inc(&clp->cl_count);
 	res = wait_on_bit_action(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 				 nfs_wait_bit_killable, TASK_KILLABLE);
 	if (res)
@@ -1409,6 +1439,11 @@ void nfs_inode_find_state_and_recover(struct inode *inode,
 			found = true;
 			continue;
 		}
+		if (nfs4_stateid_match_other(&state->open_stateid, stateid) &&
+		    nfs4_state_mark_reclaim_nograce(clp, state)) {
+			found = true;
+			continue;
+		}
 		if (nfs_state_lock_state_matches_stateid(state, stateid) &&
 		    nfs4_state_mark_reclaim_nograce(clp, state))
 			found = true;
@@ -2510,7 +2545,7 @@ static void nfs4_state_manager(struct nfs_client *clp)
 			break;
 		if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
 			break;
-	} while (atomic_read(&clp->cl_count) > 1);
+	} while (refcount_read(&clp->cl_count) > 1);
 	return;
 out_error:
 	if (strlen(section))

fs/nfs/nfs4trace.h

@@ -202,17 +202,13 @@ DECLARE_EVENT_CLASS(nfs4_clientid_event,
 		TP_ARGS(clp, error),
 
 		TP_STRUCT__entry(
-			__string(dstaddr,
+			__string(dstaddr, clp->cl_hostname)
-				rpc_peeraddr2str(clp->cl_rpcclient,
-					RPC_DISPLAY_ADDR))
 			__field(int, error)
 		),
 
 		TP_fast_assign(
 			__entry->error = error;
-			__assign_str(dstaddr,
+			__assign_str(dstaddr, clp->cl_hostname);
-				rpc_peeraddr2str(clp->cl_rpcclient,
-						RPC_DISPLAY_ADDR));
 		),
 
 		TP_printk(
@@ -1066,6 +1062,8 @@ DECLARE_EVENT_CLASS(nfs4_inode_stateid_event,
 
 DEFINE_NFS4_INODE_STATEID_EVENT(nfs4_setattr);
 DEFINE_NFS4_INODE_STATEID_EVENT(nfs4_delegreturn);
+DEFINE_NFS4_INODE_STATEID_EVENT(nfs4_open_stateid_update);
+DEFINE_NFS4_INODE_STATEID_EVENT(nfs4_open_stateid_update_wait);
 
 DECLARE_EVENT_CLASS(nfs4_getattr_event,
 		TP_PROTO(
@@ -1133,9 +1131,7 @@ DECLARE_EVENT_CLASS(nfs4_inode_callback_event,
 			__field(dev_t, dev)
 			__field(u32, fhandle)
 			__field(u64, fileid)
-			__string(dstaddr, clp ?
+			__string(dstaddr, clp ? clp->cl_hostname : "unknown")
-				rpc_peeraddr2str(clp->cl_rpcclient,
-					RPC_DISPLAY_ADDR) : "unknown")
 		),
 
 		TP_fast_assign(
@@ -1148,9 +1144,7 @@ DECLARE_EVENT_CLASS(nfs4_inode_callback_event,
 				__entry->fileid = 0;
 				__entry->dev = 0;
 			}
-			__assign_str(dstaddr, clp ?
+			__assign_str(dstaddr, clp ? clp->cl_hostname : "unknown")
-				rpc_peeraddr2str(clp->cl_rpcclient,
-					RPC_DISPLAY_ADDR) : "unknown")
 		),
 
 		TP_printk(
@@ -1192,9 +1186,7 @@ DECLARE_EVENT_CLASS(nfs4_inode_stateid_callback_event,
 			__field(dev_t, dev)
 			__field(u32, fhandle)
 			__field(u64, fileid)
-			__string(dstaddr, clp ?
+			__string(dstaddr, clp ? clp->cl_hostname : "unknown")
-				rpc_peeraddr2str(clp->cl_rpcclient,
-					RPC_DISPLAY_ADDR) : "unknown")
 			__field(int, stateid_seq)
 			__field(u32, stateid_hash)
 		),
@@ -1209,9 +1201,7 @@ DECLARE_EVENT_CLASS(nfs4_inode_stateid_callback_event,
 				__entry->fileid = 0;
 				__entry->dev = 0;
 			}
-			__assign_str(dstaddr, clp ?
+			__assign_str(dstaddr, clp ? clp->cl_hostname : "unknown")
-				rpc_peeraddr2str(clp->cl_rpcclient,
-					RPC_DISPLAY_ADDR) : "unknown")
 			__entry->stateid_seq =
 				be32_to_cpu(stateid->seqid);
 			__entry->stateid_hash =

fs/nfs/nfs4xdr.c

@@ -4385,6 +4385,14 @@ static int decode_delegation_stateid(struct xdr_stream *xdr, nfs4_stateid *state
 	return decode_stateid(xdr, stateid);
 }
 
+static int decode_invalid_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid)
+{
+	nfs4_stateid dummy;
+
+	nfs4_stateid_copy(stateid, &invalid_stateid);
+	return decode_stateid(xdr, &dummy);
+}
+
 static int decode_close(struct xdr_stream *xdr, struct nfs_closeres *res)
 {
 	int status;
@@ -4393,7 +4401,7 @@ static int decode_close(struct xdr_stream *xdr, struct nfs_closeres *res)
 	if (status != -EIO)
 		nfs_increment_open_seqid(status, res->seqid);
 	if (!status)
-		status = decode_open_stateid(xdr, &res->stateid);
+		status = decode_invalid_stateid(xdr, &res->stateid);
 	return status;
 }
 
@@ -6108,6 +6116,8 @@ static int decode_layoutreturn(struct xdr_stream *xdr,
 	res->lrs_present = be32_to_cpup(p);
 	if (res->lrs_present)
 		status = decode_layout_stateid(xdr, &res->stateid);
+	else
+		nfs4_stateid_copy(&res->stateid, &invalid_stateid);
 	return status;
 out_overflow:
 	print_overflow_msg(__func__, xdr);

fs/nfs/pnfs.c

@@ -251,7 +251,7 @@ EXPORT_SYMBOL_GPL(pnfs_unregister_layoutdriver);
 void
 pnfs_get_layout_hdr(struct pnfs_layout_hdr *lo)
 {
-	atomic_inc(&lo->plh_refcount);
+	refcount_inc(&lo->plh_refcount);
 }
 
 static struct pnfs_layout_hdr *
@@ -296,7 +296,7 @@ pnfs_put_layout_hdr(struct pnfs_layout_hdr *lo)
 
 	pnfs_layoutreturn_before_put_layout_hdr(lo);
 
-	if (atomic_dec_and_lock(&lo->plh_refcount, &inode->i_lock)) {
+	if (refcount_dec_and_lock(&lo->plh_refcount, &inode->i_lock)) {
 		if (!list_empty(&lo->plh_segs))
 			WARN_ONCE(1, "NFS: BUG unfreed layout segments.\n");
 		pnfs_detach_layout_hdr(lo);
@@ -354,6 +354,24 @@ pnfs_clear_lseg_state(struct pnfs_layout_segment *lseg,
 		pnfs_lseg_dec_and_remove_zero(lseg, free_me);
 }
 
+/*
+ * Update the seqid of a layout stateid
+ */
+bool nfs4_refresh_layout_stateid(nfs4_stateid *dst, struct inode *inode)
+{
+	struct pnfs_layout_hdr *lo;
+	bool ret = false;
+
+	spin_lock(&inode->i_lock);
+	lo = NFS_I(inode)->layout;
+	if (lo && nfs4_stateid_match_other(dst, &lo->plh_stateid)) {
+		dst->seqid = lo->plh_stateid.seqid;
+		ret = true;
+	}
+	spin_unlock(&inode->i_lock);
+	return ret;
+}
+
 /*
  * Mark a pnfs_layout_hdr and all associated layout segments as invalid
  *
@@ -395,14 +413,14 @@ pnfs_layout_set_fail_bit(struct pnfs_layout_hdr *lo, int fail_bit)
 {
 	lo->plh_retry_timestamp = jiffies;
 	if (!test_and_set_bit(fail_bit, &lo->plh_flags))
-		atomic_inc(&lo->plh_refcount);
+		refcount_inc(&lo->plh_refcount);
 }
 
 static void
 pnfs_layout_clear_fail_bit(struct pnfs_layout_hdr *lo, int fail_bit)
 {
 	if (test_and_clear_bit(fail_bit, &lo->plh_flags))
-		atomic_dec(&lo->plh_refcount);
+		refcount_dec(&lo->plh_refcount);
 }
 
 static void
@@ -450,7 +468,7 @@ pnfs_init_lseg(struct pnfs_layout_hdr *lo, struct pnfs_layout_segment *lseg,
 {
 	INIT_LIST_HEAD(&lseg->pls_list);
 	INIT_LIST_HEAD(&lseg->pls_lc_list);
-	atomic_set(&lseg->pls_refcount, 1);
+	refcount_set(&lseg->pls_refcount, 1);
 	set_bit(NFS_LSEG_VALID, &lseg->pls_flags);
 	lseg->pls_layout = lo;
 	lseg->pls_range = *range;
@@ -472,7 +490,7 @@ pnfs_layout_remove_lseg(struct pnfs_layout_hdr *lo,
 	WARN_ON(test_bit(NFS_LSEG_VALID, &lseg->pls_flags));
 	list_del_init(&lseg->pls_list);
 	/* Matched by pnfs_get_layout_hdr in pnfs_layout_insert_lseg */
-	atomic_dec(&lo->plh_refcount);
+	refcount_dec(&lo->plh_refcount);
 	if (test_bit(NFS_LSEG_LAYOUTRETURN, &lseg->pls_flags))
 		return;
 	if (list_empty(&lo->plh_segs) &&
@@ -507,13 +525,13 @@ pnfs_put_lseg(struct pnfs_layout_segment *lseg)
 		return;
 
 	dprintk("%s: lseg %p ref %d valid %d\n", __func__, lseg,
-		atomic_read(&lseg->pls_refcount),
+		refcount_read(&lseg->pls_refcount),
 		test_bit(NFS_LSEG_VALID, &lseg->pls_flags));
 
 	lo = lseg->pls_layout;
 	inode = lo->plh_inode;
 
-	if (atomic_dec_and_lock(&lseg->pls_refcount, &inode->i_lock)) {
+	if (refcount_dec_and_lock(&lseg->pls_refcount, &inode->i_lock)) {
 		if (test_bit(NFS_LSEG_VALID, &lseg->pls_flags)) {
 			spin_unlock(&inode->i_lock);
 			return;
@@ -551,7 +569,7 @@ pnfs_lseg_range_contained(const struct pnfs_layout_range *l1,
 static bool pnfs_lseg_dec_and_remove_zero(struct pnfs_layout_segment *lseg,
 		struct list_head *tmp_list)
 {
-	if (!atomic_dec_and_test(&lseg->pls_refcount))
+	if (!refcount_dec_and_test(&lseg->pls_refcount))
 		return false;
 	pnfs_layout_remove_lseg(lseg->pls_layout, lseg);
 	list_add(&lseg->pls_list, tmp_list);
@@ -570,7 +588,7 @@ static int mark_lseg_invalid(struct pnfs_layout_segment *lseg,
 		 * outstanding io is finished.
 		 */
 		dprintk("%s: lseg %p ref %d\n", __func__, lseg,
-			atomic_read(&lseg->pls_refcount));
+			refcount_read(&lseg->pls_refcount));
 		if (pnfs_lseg_dec_and_remove_zero(lseg, tmp_list))
 			rv = 1;
 	}
@@ -1451,7 +1469,7 @@ alloc_init_layout_hdr(struct inode *ino,
 	lo = pnfs_alloc_layout_hdr(ino, gfp_flags);
 	if (!lo)
 		return NULL;
-	atomic_set(&lo->plh_refcount, 1);
+	refcount_set(&lo->plh_refcount, 1);
 	INIT_LIST_HEAD(&lo->plh_layouts);
 	INIT_LIST_HEAD(&lo->plh_segs);
 	INIT_LIST_HEAD(&lo->plh_return_segs);
@@ -1513,7 +1531,7 @@ pnfs_lseg_range_match(const struct pnfs_layout_range *ls_range,
 	if ((range->iomode == IOMODE_RW &&
 	     ls_range->iomode != IOMODE_RW) ||
 	    (range->iomode != ls_range->iomode &&
-	     strict_iomode == true) ||
+	     strict_iomode) ||
 	    !pnfs_lseg_range_intersecting(ls_range, range))
 		return 0;
 
@@ -1546,7 +1564,7 @@ pnfs_find_lseg(struct pnfs_layout_hdr *lo,
 	}
 
 	dprintk("%s:Return lseg %p ref %d\n",
-		__func__, ret, ret ? atomic_read(&ret->pls_refcount) : 0);
+		__func__, ret, ret ? refcount_read(&ret->pls_refcount) : 0);
 	return ret;
 }
 

fs/nfs/pnfs.h

@@ -30,6 +30,7 @@
 #ifndef FS_NFS_PNFS_H
 #define FS_NFS_PNFS_H
 
+#include <linux/refcount.h>
 #include <linux/nfs_fs.h>
 #include <linux/nfs_page.h>
 #include <linux/workqueue.h>
@@ -54,7 +55,7 @@ struct nfs4_pnfs_ds {
 	char			*ds_remotestr;	/* comma sep list of addrs */
 	struct list_head	ds_addrs;
 	struct nfs_client	*ds_clp;
-	atomic_t		ds_count;
+	refcount_t		ds_count;
 	unsigned long		ds_state;
 #define NFS4DS_CONNECTING	0	/* ds is establishing connection */
 };
@@ -63,7 +64,7 @@ struct pnfs_layout_segment {
 	struct list_head pls_list;
 	struct list_head pls_lc_list;
 	struct pnfs_layout_range pls_range;
-	atomic_t pls_refcount;
+	refcount_t pls_refcount;
 	u32 pls_seq;
 	unsigned long pls_flags;
 	struct pnfs_layout_hdr *pls_layout;
@@ -179,7 +180,7 @@ struct pnfs_layoutdriver_type {
 };
 
 struct pnfs_layout_hdr {
-	atomic_t		plh_refcount;
+	refcount_t		plh_refcount;
 	atomic_t		plh_outstanding; /* number of RPCs out */
 	struct list_head	plh_layouts;   /* other client layouts */
 	struct list_head	plh_bulk_destroy;
@@ -251,6 +252,7 @@ int pnfs_destroy_layouts_byfsid(struct nfs_client *clp,
 		bool is_recall);
 int pnfs_destroy_layouts_byclid(struct nfs_client *clp,
 		bool is_recall);
+bool nfs4_refresh_layout_stateid(nfs4_stateid *dst, struct inode *inode);
 void pnfs_put_layout_hdr(struct pnfs_layout_hdr *lo);
 void pnfs_set_layout_stateid(struct pnfs_layout_hdr *lo,
 			     const nfs4_stateid *new,
@@ -393,7 +395,7 @@ static inline struct pnfs_layout_segment *
 pnfs_get_lseg(struct pnfs_layout_segment *lseg)
 {
 	if (lseg) {
-		atomic_inc(&lseg->pls_refcount);
+		refcount_inc(&lseg->pls_refcount);
 		smp_mb__after_atomic();
 	}
 	return lseg;
@@ -764,6 +766,11 @@ static inline void nfs4_pnfs_v3_ds_connect_unload(void)
 {
 }
 
+static inline bool nfs4_refresh_layout_stateid(nfs4_stateid *dst,
+		struct inode *inode)
+{
+	return false;
+}
 #endif /* CONFIG_NFS_V4_1 */
 
 #if IS_ENABLED(CONFIG_NFS_V4_2)

fs/nfs/pnfs_nfs.c

@@ -338,7 +338,7 @@ print_ds(struct nfs4_pnfs_ds *ds)
 		"        client %p\n"
 		"        cl_exchange_flags %x\n",
 		ds->ds_remotestr,
-		atomic_read(&ds->ds_count), ds->ds_clp,
+		refcount_read(&ds->ds_count), ds->ds_clp,
 		ds->ds_clp ? ds->ds_clp->cl_exchange_flags : 0);
 }
 
@@ -451,7 +451,7 @@ static void destroy_ds(struct nfs4_pnfs_ds *ds)
 
 void nfs4_pnfs_ds_put(struct nfs4_pnfs_ds *ds)
 {
-	if (atomic_dec_and_lock(&ds->ds_count,
+	if (refcount_dec_and_lock(&ds->ds_count,
 				&nfs4_ds_cache_lock)) {
 		list_del_init(&ds->ds_node);
 		spin_unlock(&nfs4_ds_cache_lock);
@@ -537,7 +537,7 @@ nfs4_pnfs_ds_add(struct list_head *dsaddrs, gfp_t gfp_flags)
 		INIT_LIST_HEAD(&ds->ds_addrs);
 		list_splice_init(dsaddrs, &ds->ds_addrs);
 		ds->ds_remotestr = remotestr;
-		atomic_set(&ds->ds_count, 1);
+		refcount_set(&ds->ds_count, 1);
 		INIT_LIST_HEAD(&ds->ds_node);
 		ds->ds_clp = NULL;
 		list_add(&ds->ds_node, &nfs4_data_server_cache);
@@ -546,10 +546,10 @@ nfs4_pnfs_ds_add(struct list_head *dsaddrs, gfp_t gfp_flags)
 	} else {
 		kfree(remotestr);
 		kfree(ds);
-		atomic_inc(&tmp_ds->ds_count);
+		refcount_inc(&tmp_ds->ds_count);
 		dprintk("%s data server %s found, inc'ed ds_count to %d\n",
 			__func__, tmp_ds->ds_remotestr,
-			atomic_read(&tmp_ds->ds_count));
+			refcount_read(&tmp_ds->ds_count));
 		ds = tmp_ds;
 	}
 	spin_unlock(&nfs4_ds_cache_lock);

fs/nfs/super.c

@@ -1332,7 +1332,7 @@ static int nfs_parse_mount_options(char *raw,
 			mnt->options |= NFS_OPTION_MIGRATION;
 			break;
 		case Opt_nomigration:
-			mnt->options &= NFS_OPTION_MIGRATION;
+			mnt->options &= ~NFS_OPTION_MIGRATION;
 			break;
 
 		/*
@@ -1456,18 +1456,21 @@ static int nfs_parse_mount_options(char *raw,
 			switch (token) {
 			case Opt_xprt_udp6:
 				protofamily = AF_INET6;
+				/* fall through */
 			case Opt_xprt_udp:
 				mnt->flags &= ~NFS_MOUNT_TCP;
 				mnt->nfs_server.protocol = XPRT_TRANSPORT_UDP;
 				break;
 			case Opt_xprt_tcp6:
 				protofamily = AF_INET6;
+				/* fall through */
 			case Opt_xprt_tcp:
 				mnt->flags |= NFS_MOUNT_TCP;
 				mnt->nfs_server.protocol = XPRT_TRANSPORT_TCP;
 				break;
 			case Opt_xprt_rdma6:
 				protofamily = AF_INET6;
+				/* fall through */
 			case Opt_xprt_rdma:
 				/* vector side protocols to TCP */
 				mnt->flags |= NFS_MOUNT_TCP;
@@ -1494,11 +1497,13 @@ static int nfs_parse_mount_options(char *raw,
 			switch (token) {
 			case Opt_xprt_udp6:
 				mountfamily = AF_INET6;
+				/* fall through */
 			case Opt_xprt_udp:
 				mnt->mount_server.protocol = XPRT_TRANSPORT_UDP;
 				break;
 			case Opt_xprt_tcp6:
 				mountfamily = AF_INET6;
+				/* fall through */
 			case Opt_xprt_tcp:
 				mnt->mount_server.protocol = XPRT_TRANSPORT_TCP;
 				break;
@@ -1988,9 +1993,9 @@ static int nfs23_validate_mount_data(void *options,
 	args->version = NFS_DEFAULT_VERSION;
 	switch (data->version) {
 	case 1:
-		data->namlen = 0;
+		data->namlen = 0; /* fall through */
 	case 2:
-		data->bsize = 0;
+		data->bsize = 0; /* fall through */
 	case 3:
 		if (data->flags & NFS_MOUNT_VER3)
 			goto out_no_v3;
@@ -1998,11 +2003,14 @@ static int nfs23_validate_mount_data(void *options,
 		memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
 		/* Turn off security negotiation */
 		extra_flags |= NFS_MOUNT_SECFLAVOUR;
+		/* fall through */
 	case 4:
 		if (data->flags & NFS_MOUNT_SECFLAVOUR)
 			goto out_no_sec;
+		/* fall through */
 	case 5:
 		memset(data->context, 0, sizeof(data->context));
+		/* fall through */
 	case 6:
 		if (data->flags & NFS_MOUNT_VER3) {
 			if (data->root.size > NFS3_FHSIZE || data->root.size == 0)

fs/nfs/write.c

@@ -487,10 +487,8 @@ try_again:
 	}
 
 	ret = nfs_page_group_lock(head);
-	if (ret < 0) {
+	if (ret < 0)
-		nfs_unlock_and_release_request(head);
+		goto release_request;
-		return ERR_PTR(ret);
-	}
 
 	/* lock each request in the page group */
 	total_bytes = head->wb_bytes;
@@ -515,8 +513,7 @@ try_again:
 			if (ret < 0) {
 				nfs_unroll_locks(inode, head, subreq);
 				nfs_release_request(subreq);
-				nfs_unlock_and_release_request(head);
+				goto release_request;
-				return ERR_PTR(ret);
 			}
 		}
 		/*
@@ -532,8 +529,8 @@ try_again:
 			nfs_page_group_unlock(head);
 			nfs_unroll_locks(inode, head, subreq);
 			nfs_unlock_and_release_request(subreq);
-			nfs_unlock_and_release_request(head);
+			ret = -EIO;
-			return ERR_PTR(-EIO);
+			goto release_request;
 		}
 	}
 
@@ -576,6 +573,10 @@ try_again:
 	/* still holds ref on head from nfs_page_find_head_request
 	 * and still has lock on head from lock loop */
 	return head;
+
+release_request:
+	nfs_unlock_and_release_request(head);
+	return ERR_PTR(ret);
 }
 
 static void nfs_write_error_remove_page(struct nfs_page *req)

include/linux/nfs_fs.h

@@ -23,6 +23,7 @@
 #include <linux/mm.h>
 #include <linux/pagemap.h>
 #include <linux/rbtree.h>
+#include <linux/refcount.h>
 #include <linux/rwsem.h>
 #include <linux/wait.h>
 
@@ -56,7 +57,7 @@ struct nfs_access_entry {
 };
 
 struct nfs_lock_context {
-	atomic_t count;
+	refcount_t count;
 	struct list_head list;
 	struct nfs_open_context *open_context;
 	fl_owner_t lockowner;
@@ -184,6 +185,16 @@ struct nfs_inode {
 	struct inode		vfs_inode;
 };
 
+/*
+ * Access bit flags
+ */
+#define NFS_ACCESS_READ        0x0001
+#define NFS_ACCESS_LOOKUP      0x0002
+#define NFS_ACCESS_MODIFY      0x0004
+#define NFS_ACCESS_EXTEND      0x0008
+#define NFS_ACCESS_DELETE      0x0010
+#define NFS_ACCESS_EXECUTE     0x0020
+
 /*
  * Cache validity bit flags
  */

include/linux/nfs_fs_sb.h

@@ -10,6 +10,7 @@
 #include <linux/sunrpc/xprt.h>
 
 #include <linux/atomic.h>
+#include <linux/refcount.h>
 
 struct nfs4_session;
 struct nfs_iostats;
@@ -25,7 +26,7 @@ struct nfs41_impl_id;
  * The nfs_client identifies our client state to the server.
  */
 struct nfs_client {
-	atomic_t		cl_count;
+	refcount_t		cl_count;
 	atomic_t		cl_mds_count;
 	int			cl_cons_state;	/* current construction state (-ve: init error) */
 #define NFS_CS_READY		0		/* ready to be used */

include/linux/sunrpc/rpc_rdma.h

@@ -1,4 +1,5 @@
 /*
+ * Copyright (c) 2015-2017 Oracle. All rights reserved.
  * Copyright (c) 2003-2007 Network Appliance, Inc. All rights reserved.
  *
  * This software is available to you under a choice of one of two
@@ -50,65 +51,6 @@ enum {
 	RPCRDMA_V1_DEF_INLINE_SIZE	= 1024,
 };
 
-struct rpcrdma_segment {
-	__be32 rs_handle;	/* Registered memory handle */
-	__be32 rs_length;	/* Length of the chunk in bytes */
-	__be64 rs_offset;	/* Chunk virtual address or offset */
-};
-
-/*
- * read chunk(s), encoded as a linked list.
- */
-struct rpcrdma_read_chunk {
-	__be32 rc_discrim;	/* 1 indicates presence */
-	__be32 rc_position;	/* Position in XDR stream */
-	struct rpcrdma_segment rc_target;
-};
-
-/*
- * write chunk, and reply chunk.
- */
-struct rpcrdma_write_chunk {
-	struct rpcrdma_segment wc_target;
-};
-
-/*
- * write chunk(s), encoded as a counted array.
- */
-struct rpcrdma_write_array {
-	__be32 wc_discrim;	/* 1 indicates presence */
-	__be32 wc_nchunks;	/* Array count */
-	struct rpcrdma_write_chunk wc_array[0];
-};
-
-struct rpcrdma_msg {
-	__be32 rm_xid;	/* Mirrors the RPC header xid */
-	__be32 rm_vers;	/* Version of this protocol */
-	__be32 rm_credit;	/* Buffers requested/granted */
-	__be32 rm_type;	/* Type of message (enum rpcrdma_proc) */
-	union {
-
-		struct {			/* no chunks */
-			__be32 rm_empty[3];	/* 3 empty chunk lists */
-		} rm_nochunks;
-
-		struct {			/* no chunks and padded */
-			__be32 rm_align;	/* Padding alignment */
-			__be32 rm_thresh;	/* Padding threshold */
-			__be32 rm_pempty[3];	/* 3 empty chunk lists */
-		} rm_padded;
-
-		struct {
-			__be32 rm_err;
-			__be32 rm_vers_low;
-			__be32 rm_vers_high;
-		} rm_error;
-
-		__be32 rm_chunks[0];	/* read, write and reply chunks */
-
-	} rm_body;
-};
-
 /*
  * XDR sizes, in quads
  */

include/trace/events/sunrpc.h

@@ -71,6 +71,36 @@ TRACE_EVENT(rpc_connect_status,
 		__entry->status)
 );
 
+TRACE_EVENT(rpc_request,
+	TP_PROTO(const struct rpc_task *task),
+
+	TP_ARGS(task),
+
+	TP_STRUCT__entry(
+		__field(unsigned int, task_id)
+		__field(unsigned int, client_id)
+		__field(int, version)
+		__field(bool, async)
+		__string(progname, task->tk_client->cl_program->name)
+		__string(procname, rpc_proc_name(task))
+	),
+
+	TP_fast_assign(
+		__entry->task_id = task->tk_pid;
+		__entry->client_id = task->tk_client->cl_clid;
+		__entry->version = task->tk_client->cl_vers;
+		__entry->async = RPC_IS_ASYNC(task);
+		__assign_str(progname, task->tk_client->cl_program->name)
+		__assign_str(procname, rpc_proc_name(task))
+	),
+
+	TP_printk("task:%u@%u %sv%d %s (%ssync)",
+		__entry->task_id, __entry->client_id,
+		__get_str(progname), __entry->version,
+		__get_str(procname), __entry->async ? "a": ""
+		)
+);
+
 DECLARE_EVENT_CLASS(rpc_task_running,
 
 	TP_PROTO(const struct rpc_clnt *clnt, const struct rpc_task *task, const void *action),
@@ -342,21 +372,21 @@ DECLARE_EVENT_CLASS(rpc_xprt_event,
 	TP_ARGS(xprt, xid, status),
 
 	TP_STRUCT__entry(
-		__field(__be32, xid)
+		__field(u32, xid)
 		__field(int, status)
 		__string(addr, xprt->address_strings[RPC_DISPLAY_ADDR])
 		__string(port, xprt->address_strings[RPC_DISPLAY_PORT])
 	),
 
 	TP_fast_assign(
-		__entry->xid = xid;
+		__entry->xid = be32_to_cpu(xid);
 		__entry->status = status;
 		__assign_str(addr, xprt->address_strings[RPC_DISPLAY_ADDR]);
 		__assign_str(port, xprt->address_strings[RPC_DISPLAY_PORT]);
 	),
 
-	TP_printk("peer=[%s]:%s xid=0x%x status=%d", __get_str(addr),
+	TP_printk("peer=[%s]:%s xid=0x%08x status=%d", __get_str(addr),
-			__get_str(port), be32_to_cpu(__entry->xid),
+			__get_str(port), __entry->xid,
 			__entry->status)
 );
 
@@ -417,7 +447,7 @@ TRACE_EVENT(xs_tcp_data_recv,
 	TP_STRUCT__entry(
 		__string(addr, xs->xprt.address_strings[RPC_DISPLAY_ADDR])
 		__string(port, xs->xprt.address_strings[RPC_DISPLAY_PORT])
-		__field(__be32, xid)
+		__field(u32, xid)
 		__field(unsigned long, flags)
 		__field(unsigned long, copied)
 		__field(unsigned int, reclen)
@@ -427,15 +457,15 @@ TRACE_EVENT(xs_tcp_data_recv,
 	TP_fast_assign(
 		__assign_str(addr, xs->xprt.address_strings[RPC_DISPLAY_ADDR]);
 		__assign_str(port, xs->xprt.address_strings[RPC_DISPLAY_PORT]);
-		__entry->xid = xs->tcp_xid;
+		__entry->xid = be32_to_cpu(xs->tcp_xid);
 		__entry->flags = xs->tcp_flags;
 		__entry->copied = xs->tcp_copied;
 		__entry->reclen = xs->tcp_reclen;
 		__entry->offset = xs->tcp_offset;
 	),
 
-	TP_printk("peer=[%s]:%s xid=0x%x flags=%s copied=%lu reclen=%u offset=%lu",
+	TP_printk("peer=[%s]:%s xid=0x%08x flags=%s copied=%lu reclen=%u offset=%lu",
-			__get_str(addr), __get_str(port), be32_to_cpu(__entry->xid),
+			__get_str(addr), __get_str(port), __entry->xid,
 			rpc_show_sock_xprt_flags(__entry->flags),
 			__entry->copied, __entry->reclen, __entry->offset)
 );
@@ -457,20 +487,20 @@ TRACE_EVENT(svc_recv,
 
 	TP_STRUCT__entry(
 		__field(struct sockaddr *, addr)
-		__field(__be32, xid)
+		__field(u32, xid)
 		__field(int, status)
 		__field(unsigned long, flags)
 	),
 
 	TP_fast_assign(
 		__entry->addr = (struct sockaddr *)&rqst->rq_addr;
-		__entry->xid = status > 0 ? rqst->rq_xid : 0;
+		__entry->xid = status > 0 ? be32_to_cpu(rqst->rq_xid) : 0;
 		__entry->status = status;
 		__entry->flags = rqst->rq_flags;
 	),
 
-	TP_printk("addr=%pIScp xid=0x%x status=%d flags=%s", __entry->addr,
+	TP_printk("addr=%pIScp xid=0x%08x status=%d flags=%s", __entry->addr,
-			be32_to_cpu(__entry->xid), __entry->status,
+			__entry->xid, __entry->status,
 			show_rqstp_flags(__entry->flags))
 );
 
@@ -481,21 +511,21 @@ DECLARE_EVENT_CLASS(svc_rqst_event,
 	TP_ARGS(rqst),
 
 	TP_STRUCT__entry(
-		__field(__be32, xid)
+		__field(u32, xid)
 		__field(unsigned long, flags)
 		__dynamic_array(unsigned char, addr, rqst->rq_addrlen)
 	),
 
 	TP_fast_assign(
-		__entry->xid = rqst->rq_xid;
+		__entry->xid = be32_to_cpu(rqst->rq_xid);
 		__entry->flags = rqst->rq_flags;
 		memcpy(__get_dynamic_array(addr),
 			&rqst->rq_addr, rqst->rq_addrlen);
 	),
 
-	TP_printk("addr=%pIScp rq_xid=0x%x flags=%s",
+	TP_printk("addr=%pIScp rq_xid=0x%08x flags=%s",
 		(struct sockaddr *)__get_dynamic_array(addr),
-		be32_to_cpu(__entry->xid),
+		__entry->xid,
 		show_rqstp_flags(__entry->flags))
 );
 
@@ -515,7 +545,7 @@ DECLARE_EVENT_CLASS(svc_rqst_status,
 
 	TP_STRUCT__entry(
 		__field(struct sockaddr *, addr)
-		__field(__be32, xid)
+		__field(u32, xid)
 		__field(int, dropme)
 		__field(int, status)
 		__field(unsigned long, flags)
@@ -523,13 +553,13 @@ DECLARE_EVENT_CLASS(svc_rqst_status,
 
 	TP_fast_assign(
 		__entry->addr = (struct sockaddr *)&rqst->rq_addr;
-		__entry->xid = rqst->rq_xid;
+		__entry->xid = be32_to_cpu(rqst->rq_xid);
 		__entry->status = status;
 		__entry->flags = rqst->rq_flags;
 	),
 
-	TP_printk("addr=%pIScp rq_xid=0x%x status=%d flags=%s",
+	TP_printk("addr=%pIScp rq_xid=0x%08x status=%d flags=%s",
-		__entry->addr, be32_to_cpu(__entry->xid),
+		__entry->addr, __entry->xid,
 		__entry->status, show_rqstp_flags(__entry->flags))
 );
 
@@ -678,18 +708,19 @@ DECLARE_EVENT_CLASS(svc_deferred_event,
 	TP_ARGS(dr),
 
 	TP_STRUCT__entry(
-		__field(__be32, xid)
+		__field(u32, xid)
 		__dynamic_array(unsigned char, addr, dr->addrlen)
 	),
 
 	TP_fast_assign(
-		__entry->xid = *(__be32 *)(dr->args + (dr->xprt_hlen>>2));
+		__entry->xid = be32_to_cpu(*(__be32 *)(dr->args +
+						       (dr->xprt_hlen>>2)));
 		memcpy(__get_dynamic_array(addr), &dr->addr, dr->addrlen);
 	),
 
-	TP_printk("addr=%pIScp xid=0x%x",
+	TP_printk("addr=%pIScp xid=0x%08x",
 		(struct sockaddr *)__get_dynamic_array(addr),
-		be32_to_cpu(__entry->xid))
+		__entry->xid)
 );
 
 DEFINE_EVENT(svc_deferred_event, svc_drop_deferred,

net/sunrpc/clnt.c

@@ -1491,7 +1491,6 @@ rpc_restart_call(struct rpc_task *task)
 }
 EXPORT_SYMBOL_GPL(rpc_restart_call);
 
-#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
 const char
 *rpc_proc_name(const struct rpc_task *task)
 {
@@ -1505,7 +1504,6 @@ const char
 	} else
 		return "no proc";
 }
-#endif
 
 /*
  * 0.  Initial state
@@ -1519,6 +1517,7 @@ call_start(struct rpc_task *task)
 	struct rpc_clnt	*clnt = task->tk_client;
 	int idx = task->tk_msg.rpc_proc->p_statidx;
 
+	trace_rpc_request(task);
 	dprintk("RPC: %5u call_start %s%d proc %s (%s)\n", task->tk_pid,
 			clnt->cl_program->name, clnt->cl_vers,
 			rpc_proc_name(task),
@@ -1586,6 +1585,7 @@ call_reserveresult(struct rpc_task *task)
 	switch (status) {
 	case -ENOMEM:
 		rpc_delay(task, HZ >> 2);
+		/* fall through */
 	case -EAGAIN:	/* woken up; retry */
 		task->tk_action = call_retry_reserve;
 		return;
@@ -1647,10 +1647,13 @@ call_refreshresult(struct rpc_task *task)
 		/* Use rate-limiting and a max number of retries if refresh
 		 * had status 0 but failed to update the cred.
 		 */
+		/* fall through */
 	case -ETIMEDOUT:
 		rpc_delay(task, 3*HZ);
+		/* fall through */
 	case -EAGAIN:
 		status = -EACCES;
+		/* fall through */
 	case -EKEYEXPIRED:
 		if (!task->tk_cred_retry)
 			break;
@@ -1911,6 +1914,7 @@ call_connect_status(struct rpc_task *task)
 			task->tk_action = call_bind;
 			return;
 		}
+		/* fall through */
 	case -ECONNRESET:
 	case -ECONNABORTED:
 	case -ENETUNREACH:
@@ -1924,6 +1928,7 @@ call_connect_status(struct rpc_task *task)
 			break;
 		/* retry with existing socket, after a delay */
 		rpc_delay(task, 3*HZ);
+		/* fall through */
 	case -EAGAIN:
 		/* Check for timeouts before looping back to call_bind */
 	case -ETIMEDOUT:
@@ -2025,6 +2030,7 @@ call_transmit_status(struct rpc_task *task)
 			rpc_exit(task, task->tk_status);
 			break;
 		}
+		/* fall through */
 	case -ECONNRESET:
 	case -ECONNABORTED:
 	case -EADDRINUSE:
@@ -2145,6 +2151,7 @@ call_status(struct rpc_task *task)
 		 * were a timeout.
 		 */
 		rpc_delay(task, 3*HZ);
+		/* fall through */
 	case -ETIMEDOUT:
 		task->tk_action = call_timeout;
 		break;
@@ -2152,14 +2159,17 @@ call_status(struct rpc_task *task)
 	case -ECONNRESET:
 	case -ECONNABORTED:
 		rpc_force_rebind(clnt);
+		/* fall through */
 	case -EADDRINUSE:
 		rpc_delay(task, 3*HZ);
+		/* fall through */
 	case -EPIPE:
 	case -ENOTCONN:
 		task->tk_action = call_bind;
 		break;
 	case -ENOBUFS:
 		rpc_delay(task, HZ>>2);
+		/* fall through */
 	case -EAGAIN:
 		task->tk_action = call_transmit;
 		break;

net/sunrpc/rpc_pipe.c

@@ -1410,8 +1410,8 @@ rpc_fill_super(struct super_block *sb, void *data, int silent)
 		return PTR_ERR(gssd_dentry);
 	}
 
-	dprintk("RPC:       sending pipefs MOUNT notification for net %p%s\n",
+	dprintk("RPC:       sending pipefs MOUNT notification for net %x%s\n",
-		net, NET_NAME(net));
+		net->ns.inum, NET_NAME(net));
 	mutex_lock(&sn->pipefs_sb_lock);
 	sn->pipefs_sb = sb;
 	err = blocking_notifier_call_chain(&rpc_pipefs_notifier_list,
@@ -1462,8 +1462,8 @@ static void rpc_kill_sb(struct super_block *sb)
 		goto out;
 	}
 	sn->pipefs_sb = NULL;
-	dprintk("RPC:       sending pipefs UMOUNT notification for net %p%s\n",
+	dprintk("RPC:       sending pipefs UMOUNT notification for net %x%s\n",
-		net, NET_NAME(net));
+		net->ns.inum, NET_NAME(net));
 	blocking_notifier_call_chain(&rpc_pipefs_notifier_list,
 					   RPC_PIPEFS_UMOUNT,
 					   sb);

net/sunrpc/rpcb_clnt.c

@@ -216,9 +216,9 @@ static void rpcb_set_local(struct net *net, struct rpc_clnt *clnt,
 	smp_wmb(); 
 	sn->rpcb_users = 1;
 	dprintk("RPC:       created new rpcb local clients (rpcb_local_clnt: "
-			"%p, rpcb_local_clnt4: %p) for net %p%s\n",
+		"%p, rpcb_local_clnt4: %p) for net %x%s\n",
-			sn->rpcb_local_clnt, sn->rpcb_local_clnt4,
+		sn->rpcb_local_clnt, sn->rpcb_local_clnt4,
-			net, (net == &init_net) ? " (init_net)" : "");
+		net->ns.inum, (net == &init_net) ? " (init_net)" : "");
 }
 
 /*

net/sunrpc/sched.c

@@ -274,10 +274,9 @@ static inline void rpc_task_set_debuginfo(struct rpc_task *task)
 
 static void rpc_set_active(struct rpc_task *task)
 {
-	trace_rpc_task_begin(task->tk_client, task, NULL);
-
 	rpc_task_set_debuginfo(task);
 	set_bit(RPC_TASK_ACTIVE, &task->tk_runstate);
+	trace_rpc_task_begin(task->tk_client, task, NULL);
 }
 
 /*

net/sunrpc/sunrpc_syms.c

@@ -65,10 +65,13 @@ err_proc:
 
 static __net_exit void sunrpc_exit_net(struct net *net)
 {
+	struct sunrpc_net *sn = net_generic(net, sunrpc_net_id);
+
 	rpc_pipefs_exit_net(net);
 	unix_gid_cache_destroy(net);
 	ip_map_cache_destroy(net);
 	rpc_proc_exit(net);
+	WARN_ON_ONCE(!list_empty(&sn->all_clients));
 }
 
 static struct pernet_operations sunrpc_net_ops = {

net/sunrpc/xprt.c

@@ -1139,6 +1139,7 @@ void xprt_alloc_slot(struct rpc_xprt *xprt, struct rpc_task *task)
 	case -EAGAIN:
 		xprt_add_backlog(xprt, task);
 		dprintk("RPC:       waiting for request slot\n");
+		/* fall through */
 	default:
 		task->tk_status = -EAGAIN;
 	}

net/sunrpc/xprtrdma/backchannel.c

@@ -43,7 +43,7 @@ static int rpcrdma_bc_setup_rqst(struct rpcrdma_xprt *r_xprt,
 	req = rpcrdma_create_req(r_xprt);
 	if (IS_ERR(req))
 		return PTR_ERR(req);
-	req->rl_backchannel = true;
+	__set_bit(RPCRDMA_REQ_F_BACKCHANNEL, &req->rl_flags);
 
 	rb = rpcrdma_alloc_regbuf(RPCRDMA_HDRBUF_SIZE,
 				  DMA_TO_DEVICE, GFP_KERNEL);
@@ -223,8 +223,8 @@ int rpcrdma_bc_marshal_reply(struct rpc_rqst *rqst)
 	*p++ = xdr_zero;
 	*p = xdr_zero;
 
-	if (!rpcrdma_prepare_send_sges(&r_xprt->rx_ia, req, RPCRDMA_HDRLEN_MIN,
+	if (rpcrdma_prepare_send_sges(r_xprt, req, RPCRDMA_HDRLEN_MIN,
-				       &rqst->rq_snd_buf, rpcrdma_noch))
+				      &rqst->rq_snd_buf, rpcrdma_noch))
 		return -EIO;
 	return 0;
 }

net/sunrpc/xprtrdma/fmr_ops.c

@@ -306,28 +306,9 @@ out_reset:
 	}
 }
 
-/* Use a slow, safe mechanism to invalidate all memory regions
- * that were registered for "req".
- */
-static void
-fmr_op_unmap_safe(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req,
-		  bool sync)
-{
-	struct rpcrdma_mw *mw;
-
-	while (!list_empty(&req->rl_registered)) {
-		mw = rpcrdma_pop_mw(&req->rl_registered);
-		if (sync)
-			fmr_op_recover_mr(mw);
-		else
-			rpcrdma_defer_mr_recovery(mw);
-	}
-}
-
 const struct rpcrdma_memreg_ops rpcrdma_fmr_memreg_ops = {
 	.ro_map				= fmr_op_map,
 	.ro_unmap_sync			= fmr_op_unmap_sync,
-	.ro_unmap_safe			= fmr_op_unmap_safe,
 	.ro_recover_mr			= fmr_op_recover_mr,
 	.ro_open			= fmr_op_open,
 	.ro_maxpages			= fmr_op_maxpages,

net/sunrpc/xprtrdma/frwr_ops.c

@@ -420,7 +420,6 @@ frwr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 			 IB_ACCESS_REMOTE_WRITE | IB_ACCESS_LOCAL_WRITE :
 			 IB_ACCESS_REMOTE_READ;
 
-	rpcrdma_set_signaled(&r_xprt->rx_ep, &reg_wr->wr);
 	rc = ib_post_send(ia->ri_id->qp, &reg_wr->wr, &bad_wr);
 	if (rc)
 		goto out_senderr;
@@ -508,12 +507,6 @@ frwr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct list_head *mws)
 	f->fr_cqe.done = frwr_wc_localinv_wake;
 	reinit_completion(&f->fr_linv_done);
 
-	/* Initialize CQ count, since there is always a signaled
-	 * WR being posted here.  The new cqcount depends on how
-	 * many SQEs are about to be consumed.
-	 */
-	rpcrdma_init_cqcount(&r_xprt->rx_ep, count);
-
 	/* Transport disconnect drains the receive CQ before it
 	 * replaces the QP. The RPC reply handler won't call us
 	 * unless ri_id->qp is a valid pointer.
@@ -546,7 +539,6 @@ reset_mrs:
 	/* Find and reset the MRs in the LOCAL_INV WRs that did not
 	 * get posted.
 	 */
-	rpcrdma_init_cqcount(&r_xprt->rx_ep, -count);
 	while (bad_wr) {
 		f = container_of(bad_wr, struct rpcrdma_frmr,
 				 fr_invwr);
@@ -559,28 +551,9 @@ reset_mrs:
 	goto unmap;
 }
 
-/* Use a slow, safe mechanism to invalidate all memory regions
- * that were registered for "req".
- */
-static void
-frwr_op_unmap_safe(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req,
-		   bool sync)
-{
-	struct rpcrdma_mw *mw;
-
-	while (!list_empty(&req->rl_registered)) {
-		mw = rpcrdma_pop_mw(&req->rl_registered);
-		if (sync)
-			frwr_op_recover_mr(mw);
-		else
-			rpcrdma_defer_mr_recovery(mw);
-	}
-}
-
 const struct rpcrdma_memreg_ops rpcrdma_frwr_memreg_ops = {
 	.ro_map				= frwr_op_map,
 	.ro_unmap_sync			= frwr_op_unmap_sync,
-	.ro_unmap_safe			= frwr_op_unmap_safe,
 	.ro_recover_mr			= frwr_op_recover_mr,
 	.ro_open			= frwr_op_open,
 	.ro_maxpages			= frwr_op_maxpages,

net/sunrpc/xprtrdma/rpc_rdma.c

@@ -1,4 +1,5 @@
 /*
+ * Copyright (c) 2014-2017 Oracle.  All rights reserved.
  * Copyright (c) 2003-2007 Network Appliance, Inc. All rights reserved.
  *
  * This software is available to you under a choice of one of two
@@ -75,11 +76,11 @@ static unsigned int rpcrdma_max_call_header_size(unsigned int maxsegs)
 
 	/* Maximum Read list size */
 	maxsegs += 2;	/* segment for head and tail buffers */
-	size = maxsegs * sizeof(struct rpcrdma_read_chunk);
+	size = maxsegs * rpcrdma_readchunk_maxsz * sizeof(__be32);
 
 	/* Minimal Read chunk size */
 	size += sizeof(__be32);	/* segment count */
-	size += sizeof(struct rpcrdma_segment);
+	size += rpcrdma_segment_maxsz * sizeof(__be32);
 	size += sizeof(__be32);	/* list discriminator */
 
 	dprintk("RPC:       %s: max call header size = %u\n",
@@ -102,7 +103,7 @@ static unsigned int rpcrdma_max_reply_header_size(unsigned int maxsegs)
 	/* Maximum Write list size */
 	maxsegs += 2;	/* segment for head and tail buffers */
 	size = sizeof(__be32);		/* segment count */
-	size += maxsegs * sizeof(struct rpcrdma_segment);
+	size += maxsegs * rpcrdma_segment_maxsz * sizeof(__be32);
 	size += sizeof(__be32);	/* list discriminator */
 
 	dprintk("RPC:       %s: max reply header size = %u\n",
@@ -511,27 +512,60 @@ rpcrdma_encode_reply_chunk(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req,
 	return 0;
 }
 
-/* Prepare the RPC-over-RDMA header SGE.
+/**
+ * rpcrdma_unmap_sendctx - DMA-unmap Send buffers
+ * @sc: sendctx containing SGEs to unmap
+ *
+ */
+void
+rpcrdma_unmap_sendctx(struct rpcrdma_sendctx *sc)
+{
+	struct rpcrdma_ia *ia = &sc->sc_xprt->rx_ia;
+	struct ib_sge *sge;
+	unsigned int count;
+
+	dprintk("RPC:       %s: unmapping %u sges for sc=%p\n",
+		__func__, sc->sc_unmap_count, sc);
+
+	/* The first two SGEs contain the transport header and
+	 * the inline buffer. These are always left mapped so
+	 * they can be cheaply re-used.
+	 */
+	sge = &sc->sc_sges[2];
+	for (count = sc->sc_unmap_count; count; ++sge, --count)
+		ib_dma_unmap_page(ia->ri_device,
+				  sge->addr, sge->length, DMA_TO_DEVICE);
+
+	if (test_and_clear_bit(RPCRDMA_REQ_F_TX_RESOURCES, &sc->sc_req->rl_flags)) {
+		smp_mb__after_atomic();
+		wake_up_bit(&sc->sc_req->rl_flags, RPCRDMA_REQ_F_TX_RESOURCES);
+	}
+}
+
+/* Prepare an SGE for the RPC-over-RDMA transport header.
  */
 static bool
 rpcrdma_prepare_hdr_sge(struct rpcrdma_ia *ia, struct rpcrdma_req *req,
 			u32 len)
 {
+	struct rpcrdma_sendctx *sc = req->rl_sendctx;
 	struct rpcrdma_regbuf *rb = req->rl_rdmabuf;
-	struct ib_sge *sge = &req->rl_send_sge[0];
+	struct ib_sge *sge = sc->sc_sges;
 
-	if (unlikely(!rpcrdma_regbuf_is_mapped(rb))) {
+	if (!rpcrdma_dma_map_regbuf(ia, rb))
-		if (!__rpcrdma_dma_map_regbuf(ia, rb))
+		goto out_regbuf;
-			return false;
+	sge->addr = rdmab_addr(rb);
-		sge->addr = rdmab_addr(rb);
-		sge->lkey = rdmab_lkey(rb);
-	}
 	sge->length = len;
+	sge->lkey = rdmab_lkey(rb);
 
 	ib_dma_sync_single_for_device(rdmab_device(rb), sge->addr,
 				      sge->length, DMA_TO_DEVICE);
-	req->rl_send_wr.num_sge++;
+	sc->sc_wr.num_sge++;
 	return true;
+
+out_regbuf:
+	pr_err("rpcrdma: failed to DMA map a Send buffer\n");
+	return false;
 }
 
 /* Prepare the Send SGEs. The head and tail iovec, and each entry
@@ -541,10 +575,11 @@ static bool
 rpcrdma_prepare_msg_sges(struct rpcrdma_ia *ia, struct rpcrdma_req *req,
 			 struct xdr_buf *xdr, enum rpcrdma_chunktype rtype)
 {
+	struct rpcrdma_sendctx *sc = req->rl_sendctx;
 	unsigned int sge_no, page_base, len, remaining;
 	struct rpcrdma_regbuf *rb = req->rl_sendbuf;
 	struct ib_device *device = ia->ri_device;
-	struct ib_sge *sge = req->rl_send_sge;
+	struct ib_sge *sge = sc->sc_sges;
 	u32 lkey = ia->ri_pd->local_dma_lkey;
 	struct page *page, **ppages;
 
@@ -552,7 +587,7 @@ rpcrdma_prepare_msg_sges(struct rpcrdma_ia *ia, struct rpcrdma_req *req,
 	 * DMA-mapped. Sync the content that has changed.
 	 */
 	if (!rpcrdma_dma_map_regbuf(ia, rb))
-		return false;
+		goto out_regbuf;
 	sge_no = 1;
 	sge[sge_no].addr = rdmab_addr(rb);
 	sge[sge_no].length = xdr->head[0].iov_len;
@@ -607,7 +642,7 @@ rpcrdma_prepare_msg_sges(struct rpcrdma_ia *ia, struct rpcrdma_req *req,
 			sge[sge_no].length = len;
 			sge[sge_no].lkey = lkey;
 
-			req->rl_mapped_sges++;
+			sc->sc_unmap_count++;
 			ppages++;
 			remaining -= len;
 			page_base = 0;
@@ -633,56 +668,61 @@ map_tail:
 			goto out_mapping_err;
 		sge[sge_no].length = len;
 		sge[sge_no].lkey = lkey;
-		req->rl_mapped_sges++;
+		sc->sc_unmap_count++;
 	}
 
 out:
-	req->rl_send_wr.num_sge = sge_no + 1;
+	sc->sc_wr.num_sge += sge_no;
+	if (sc->sc_unmap_count)
+		__set_bit(RPCRDMA_REQ_F_TX_RESOURCES, &req->rl_flags);
 	return true;
 
+out_regbuf:
+	pr_err("rpcrdma: failed to DMA map a Send buffer\n");
+	return false;
+
 out_mapping_overflow:
+	rpcrdma_unmap_sendctx(sc);
 	pr_err("rpcrdma: too many Send SGEs (%u)\n", sge_no);
 	return false;
 
 out_mapping_err:
+	rpcrdma_unmap_sendctx(sc);
 	pr_err("rpcrdma: Send mapping error\n");
 	return false;
 }
 
-bool
+/**
-rpcrdma_prepare_send_sges(struct rpcrdma_ia *ia, struct rpcrdma_req *req,
+ * rpcrdma_prepare_send_sges - Construct SGEs for a Send WR
-			  u32 hdrlen, struct xdr_buf *xdr,
+ * @r_xprt: controlling transport
-			  enum rpcrdma_chunktype rtype)
+ * @req: context of RPC Call being marshalled
+ * @hdrlen: size of transport header, in bytes
+ * @xdr: xdr_buf containing RPC Call
+ * @rtype: chunk type being encoded
+ *
+ * Returns 0 on success; otherwise a negative errno is returned.
+ */
+int
+rpcrdma_prepare_send_sges(struct rpcrdma_xprt *r_xprt,
+			  struct rpcrdma_req *req, u32 hdrlen,
+			  struct xdr_buf *xdr, enum rpcrdma_chunktype rtype)
 {
-	req->rl_send_wr.num_sge = 0;
+	req->rl_sendctx = rpcrdma_sendctx_get_locked(&r_xprt->rx_buf);
-	req->rl_mapped_sges = 0;
+	if (!req->rl_sendctx)
+		return -ENOBUFS;
+	req->rl_sendctx->sc_wr.num_sge = 0;
+	req->rl_sendctx->sc_unmap_count = 0;
+	req->rl_sendctx->sc_req = req;
+	__clear_bit(RPCRDMA_REQ_F_TX_RESOURCES, &req->rl_flags);
 
-	if (!rpcrdma_prepare_hdr_sge(ia, req, hdrlen))
+	if (!rpcrdma_prepare_hdr_sge(&r_xprt->rx_ia, req, hdrlen))
-		goto out_map;
+		return -EIO;
 
 	if (rtype != rpcrdma_areadch)
-		if (!rpcrdma_prepare_msg_sges(ia, req, xdr, rtype))
+		if (!rpcrdma_prepare_msg_sges(&r_xprt->rx_ia, req, xdr, rtype))
-			goto out_map;
+			return -EIO;
 
-	return true;
+	return 0;
-
-out_map:
-	pr_err("rpcrdma: failed to DMA map a Send buffer\n");
-	return false;
-}
-
-void
-rpcrdma_unmap_sges(struct rpcrdma_ia *ia, struct rpcrdma_req *req)
-{
-	struct ib_device *device = ia->ri_device;
-	struct ib_sge *sge;
-	int count;
-
-	sge = &req->rl_send_sge[2];
-	for (count = req->rl_mapped_sges; count--; sge++)
-		ib_dma_unmap_page(device, sge->addr, sge->length,
-				  DMA_TO_DEVICE);
-	req->rl_mapped_sges = 0;
 }
 
 /**
@@ -833,12 +873,10 @@ rpcrdma_marshal_req(struct rpcrdma_xprt *r_xprt, struct rpc_rqst *rqst)
 		transfertypes[rtype], transfertypes[wtype],
 		xdr_stream_pos(xdr));
 
-	if (!rpcrdma_prepare_send_sges(&r_xprt->rx_ia, req,
+	ret = rpcrdma_prepare_send_sges(r_xprt, req, xdr_stream_pos(xdr),
-				       xdr_stream_pos(xdr),
+					&rqst->rq_snd_buf, rtype);
-				       &rqst->rq_snd_buf, rtype)) {
+	if (ret)
-		ret = -EIO;
 		goto out_err;
-	}
 	return 0;
 
 out_err:
@@ -970,14 +1008,13 @@ rpcrdma_mark_remote_invalidation(struct list_head *mws,
  * straightforward to check the RPC header's direction field.
  */
 static bool
-rpcrdma_is_bcall(struct rpcrdma_xprt *r_xprt, struct rpcrdma_rep *rep,
+rpcrdma_is_bcall(struct rpcrdma_xprt *r_xprt, struct rpcrdma_rep *rep)
-		 __be32 xid, __be32 proc)
 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
 {
 	struct xdr_stream *xdr = &rep->rr_stream;
 	__be32 *p;
 
-	if (proc != rdma_msg)
+	if (rep->rr_proc != rdma_msg)
 		return false;
 
 	/* Peek at stream contents without advancing. */
@@ -992,7 +1029,7 @@ rpcrdma_is_bcall(struct rpcrdma_xprt *r_xprt, struct rpcrdma_rep *rep,
 		return false;
 
 	/* RPC header */
-	if (*p++ != xid)
+	if (*p++ != rep->rr_xid)
 		return false;
 	if (*p != cpu_to_be32(RPC_CALL))
 		return false;
@@ -1212,78 +1249,21 @@ rpcrdma_decode_error(struct rpcrdma_xprt *r_xprt, struct rpcrdma_rep *rep,
 	return -EREMOTEIO;
 }
 
-/* Process received RPC/RDMA messages.
+/* Perform XID lookup, reconstruction of the RPC reply, and
- *
+ * RPC completion while holding the transport lock to ensure
- * Errors must result in the RPC task either being awakened, or
+ * the rep, rqst, and rq_task pointers remain stable.
- * allowed to timeout, to discover the errors at that time.
  */
-void
+void rpcrdma_complete_rqst(struct rpcrdma_rep *rep)
-rpcrdma_reply_handler(struct work_struct *work)
 {
-	struct rpcrdma_rep *rep =
-			container_of(work, struct rpcrdma_rep, rr_work);
 	struct rpcrdma_xprt *r_xprt = rep->rr_rxprt;
 	struct rpc_xprt *xprt = &r_xprt->rx_xprt;
-	struct xdr_stream *xdr = &rep->rr_stream;
+	struct rpc_rqst *rqst = rep->rr_rqst;
-	struct rpcrdma_req *req;
-	struct rpc_rqst *rqst;
-	__be32 *p, xid, vers, proc;
 	unsigned long cwnd;
 	int status;
 
-	dprintk("RPC:       %s: incoming rep %p\n", __func__, rep);
-
-	if (rep->rr_hdrbuf.head[0].iov_len == 0)
-		goto out_badstatus;
-
-	xdr_init_decode(xdr, &rep->rr_hdrbuf,
-			rep->rr_hdrbuf.head[0].iov_base);
-
-	/* Fixed transport header fields */
-	p = xdr_inline_decode(xdr, 4 * sizeof(*p));
-	if (unlikely(!p))
-		goto out_shortreply;
-	xid = *p++;
-	vers = *p++;
-	p++;	/* credits */
-	proc = *p++;
-
-	if (rpcrdma_is_bcall(r_xprt, rep, xid, proc))
-		return;
-
-	/* Match incoming rpcrdma_rep to an rpcrdma_req to
-	 * get context for handling any incoming chunks.
-	 */
-	spin_lock(&xprt->recv_lock);
-	rqst = xprt_lookup_rqst(xprt, xid);
-	if (!rqst)
-		goto out_norqst;
-	xprt_pin_rqst(rqst);
-	spin_unlock(&xprt->recv_lock);
-	req = rpcr_to_rdmar(rqst);
-	req->rl_reply = rep;
-
-	dprintk("RPC:       %s: reply %p completes request %p (xid 0x%08x)\n",
-		__func__, rep, req, be32_to_cpu(xid));
-
-	/* Invalidate and unmap the data payloads before waking the
-	 * waiting application. This guarantees the memory regions
-	 * are properly fenced from the server before the application
-	 * accesses the data. It also ensures proper send flow control:
-	 * waking the next RPC waits until this RPC has relinquished
-	 * all its Send Queue entries.
-	 */
-	if (!list_empty(&req->rl_registered)) {
-		rpcrdma_mark_remote_invalidation(&req->rl_registered, rep);
-		r_xprt->rx_ia.ri_ops->ro_unmap_sync(r_xprt,
-						    &req->rl_registered);
-	}
-
 	xprt->reestablish_timeout = 0;
-	if (vers != rpcrdma_version)
-		goto out_badversion;
 
-	switch (proc) {
+	switch (rep->rr_proc) {
 	case rdma_msg:
 		status = rpcrdma_decode_msg(r_xprt, rep, rqst);
 		break;
@@ -1302,15 +1282,137 @@ rpcrdma_reply_handler(struct work_struct *work)
 out:
 	spin_lock(&xprt->recv_lock);
 	cwnd = xprt->cwnd;
-	xprt->cwnd = atomic_read(&r_xprt->rx_buf.rb_credits) << RPC_CWNDSHIFT;
+	xprt->cwnd = r_xprt->rx_buf.rb_credits << RPC_CWNDSHIFT;
 	if (xprt->cwnd > cwnd)
 		xprt_release_rqst_cong(rqst->rq_task);
 
 	xprt_complete_rqst(rqst->rq_task, status);
 	xprt_unpin_rqst(rqst);
 	spin_unlock(&xprt->recv_lock);
-	dprintk("RPC:       %s: xprt_complete_rqst(0x%p, 0x%p, %d)\n",
+	return;
-		__func__, xprt, rqst, status);
+
+/* If the incoming reply terminated a pending RPC, the next
+ * RPC call will post a replacement receive buffer as it is
+ * being marshaled.
+ */
+out_badheader:
+	dprintk("RPC: %5u %s: invalid rpcrdma reply (type %u)\n",
+		rqst->rq_task->tk_pid, __func__, be32_to_cpu(rep->rr_proc));
+	r_xprt->rx_stats.bad_reply_count++;
+	status = -EIO;
+	goto out;
+}
+
+void rpcrdma_release_rqst(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
+{
+	/* Invalidate and unmap the data payloads before waking
+	 * the waiting application. This guarantees the memory
+	 * regions are properly fenced from the server before the
+	 * application accesses the data. It also ensures proper
+	 * send flow control: waking the next RPC waits until this
+	 * RPC has relinquished all its Send Queue entries.
+	 */
+	if (!list_empty(&req->rl_registered))
+		r_xprt->rx_ia.ri_ops->ro_unmap_sync(r_xprt,
+						    &req->rl_registered);
+
+	/* Ensure that any DMA mapped pages associated with
+	 * the Send of the RPC Call have been unmapped before
+	 * allowing the RPC to complete. This protects argument
+	 * memory not controlled by the RPC client from being
+	 * re-used before we're done with it.
+	 */
+	if (test_bit(RPCRDMA_REQ_F_TX_RESOURCES, &req->rl_flags)) {
+		r_xprt->rx_stats.reply_waits_for_send++;
+		out_of_line_wait_on_bit(&req->rl_flags,
+					RPCRDMA_REQ_F_TX_RESOURCES,
+					bit_wait,
+					TASK_UNINTERRUPTIBLE);
+	}
+}
+
+/* Reply handling runs in the poll worker thread. Anything that
+ * might wait is deferred to a separate workqueue.
+ */
+void rpcrdma_deferred_completion(struct work_struct *work)
+{
+	struct rpcrdma_rep *rep =
+			container_of(work, struct rpcrdma_rep, rr_work);
+	struct rpcrdma_req *req = rpcr_to_rdmar(rep->rr_rqst);
+
+	rpcrdma_mark_remote_invalidation(&req->rl_registered, rep);
+	rpcrdma_release_rqst(rep->rr_rxprt, req);
+	rpcrdma_complete_rqst(rep);
+}
+
+/* Process received RPC/RDMA messages.
+ *
+ * Errors must result in the RPC task either being awakened, or
+ * allowed to timeout, to discover the errors at that time.
+ */
+void rpcrdma_reply_handler(struct rpcrdma_rep *rep)
+{
+	struct rpcrdma_xprt *r_xprt = rep->rr_rxprt;
+	struct rpc_xprt *xprt = &r_xprt->rx_xprt;
+	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
+	struct rpcrdma_req *req;
+	struct rpc_rqst *rqst;
+	u32 credits;
+	__be32 *p;
+
+	dprintk("RPC:       %s: incoming rep %p\n", __func__, rep);
+
+	if (rep->rr_hdrbuf.head[0].iov_len == 0)
+		goto out_badstatus;
+
+	xdr_init_decode(&rep->rr_stream, &rep->rr_hdrbuf,
+			rep->rr_hdrbuf.head[0].iov_base);
+
+	/* Fixed transport header fields */
+	p = xdr_inline_decode(&rep->rr_stream, 4 * sizeof(*p));
+	if (unlikely(!p))
+		goto out_shortreply;
+	rep->rr_xid = *p++;
+	rep->rr_vers = *p++;
+	credits = be32_to_cpu(*p++);
+	rep->rr_proc = *p++;
+
+	if (rep->rr_vers != rpcrdma_version)
+		goto out_badversion;
+
+	if (rpcrdma_is_bcall(r_xprt, rep))
+		return;
+
+	/* Match incoming rpcrdma_rep to an rpcrdma_req to
+	 * get context for handling any incoming chunks.
+	 */
+	spin_lock(&xprt->recv_lock);
+	rqst = xprt_lookup_rqst(xprt, rep->rr_xid);
+	if (!rqst)
+		goto out_norqst;
+	xprt_pin_rqst(rqst);
+
+	if (credits == 0)
+		credits = 1;	/* don't deadlock */
+	else if (credits > buf->rb_max_requests)
+		credits = buf->rb_max_requests;
+	buf->rb_credits = credits;
+
+	spin_unlock(&xprt->recv_lock);
+
+	req = rpcr_to_rdmar(rqst);
+	req->rl_reply = rep;
+	rep->rr_rqst = rqst;
+	clear_bit(RPCRDMA_REQ_F_PENDING, &req->rl_flags);
+
+	dprintk("RPC:       %s: reply %p completes request %p (xid 0x%08x)\n",
+		__func__, rep, req, be32_to_cpu(rep->rr_xid));
+
+	if (list_empty(&req->rl_registered) &&
+	    !test_bit(RPCRDMA_REQ_F_TX_RESOURCES, &req->rl_flags))
+		rpcrdma_complete_rqst(rep);
+	else
+		queue_work(rpcrdma_receive_wq, &rep->rr_work);
 	return;
 
 out_badstatus:
@@ -1321,37 +1423,22 @@ out_badstatus:
 	}
 	return;
 
-/* If the incoming reply terminated a pending RPC, the next
- * RPC call will post a replacement receive buffer as it is
- * being marshaled.
- */
 out_badversion:
 	dprintk("RPC:       %s: invalid version %d\n",
-		__func__, be32_to_cpu(vers));
+		__func__, be32_to_cpu(rep->rr_vers));
-	status = -EIO;
+	goto repost;
-	r_xprt->rx_stats.bad_reply_count++;
-	goto out;
 
-out_badheader:
+/* The RPC transaction has already been terminated, or the header
-	dprintk("RPC: %5u %s: invalid rpcrdma reply (type %u)\n",
+ * is corrupt.
-		rqst->rq_task->tk_pid, __func__, be32_to_cpu(proc));
-	r_xprt->rx_stats.bad_reply_count++;
-	status = -EIO;
-	goto out;
-
-/* The req was still available, but by the time the recv_lock
- * was acquired, the rqst and task had been released. Thus the RPC
- * has already been terminated.
  */
 out_norqst:
 	spin_unlock(&xprt->recv_lock);
 	dprintk("RPC:       %s: no match for incoming xid 0x%08x\n",
-		__func__, be32_to_cpu(xid));
+		__func__, be32_to_cpu(rep->rr_xid));
 	goto repost;
 
 out_shortreply:
 	dprintk("RPC:       %s: short/invalid reply\n", __func__);
-	goto repost;
 
 /* If no pending RPC transaction was matched, post a replacement
  * receive buffer before returning.

net/sunrpc/xprtrdma/transport.c

@@ -1,4 +1,5 @@
 /*
+ * Copyright (c) 2014-2017 Oracle.  All rights reserved.
  * Copyright (c) 2003-2007 Network Appliance, Inc. All rights reserved.
  *
  * This software is available to you under a choice of one of two
@@ -678,16 +679,14 @@ xprt_rdma_free(struct rpc_task *task)
 	struct rpc_rqst *rqst = task->tk_rqstp;
 	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(rqst->rq_xprt);
 	struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
-	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
 
-	if (req->rl_backchannel)
+	if (test_bit(RPCRDMA_REQ_F_BACKCHANNEL, &req->rl_flags))
 		return;
 
 	dprintk("RPC:       %s: called on 0x%p\n", __func__, req->rl_reply);
 
-	if (!list_empty(&req->rl_registered))
+	if (test_bit(RPCRDMA_REQ_F_PENDING, &req->rl_flags))
-		ia->ri_ops->ro_unmap_safe(r_xprt, req, !RPC_IS_ASYNC(task));
+		rpcrdma_release_rqst(r_xprt, req);
-	rpcrdma_unmap_sges(ia, req);
 	rpcrdma_buffer_put(req);
 }
 
@@ -728,7 +727,8 @@ xprt_rdma_send_request(struct rpc_task *task)
 
 	/* On retransmit, remove any previously registered chunks */
 	if (unlikely(!list_empty(&req->rl_registered)))
-		r_xprt->rx_ia.ri_ops->ro_unmap_safe(r_xprt, req, false);
+		r_xprt->rx_ia.ri_ops->ro_unmap_sync(r_xprt,
+						    &req->rl_registered);
 
 	rc = rpcrdma_marshal_req(r_xprt, rqst);
 	if (rc < 0)
@@ -742,6 +742,7 @@ xprt_rdma_send_request(struct rpc_task *task)
 		goto drop_connection;
 	req->rl_connect_cookie = xprt->connect_cookie;
 
+	set_bit(RPCRDMA_REQ_F_PENDING, &req->rl_flags);
 	if (rpcrdma_ep_post(&r_xprt->rx_ia, &r_xprt->rx_ep, req))
 		goto drop_connection;
 
@@ -789,11 +790,13 @@ void xprt_rdma_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
 		   r_xprt->rx_stats.failed_marshal_count,
 		   r_xprt->rx_stats.bad_reply_count,
 		   r_xprt->rx_stats.nomsg_call_count);
-	seq_printf(seq, "%lu %lu %lu %lu\n",
+	seq_printf(seq, "%lu %lu %lu %lu %lu %lu\n",
 		   r_xprt->rx_stats.mrs_recovered,
 		   r_xprt->rx_stats.mrs_orphaned,
 		   r_xprt->rx_stats.mrs_allocated,
-		   r_xprt->rx_stats.local_inv_needed);
+		   r_xprt->rx_stats.local_inv_needed,
+		   r_xprt->rx_stats.empty_sendctx_q,
+		   r_xprt->rx_stats.reply_waits_for_send);
 }
 
 static int

net/sunrpc/xprtrdma/verbs.c

@@ -1,4 +1,5 @@
 /*
+ * Copyright (c) 2014-2017 Oracle.  All rights reserved.
  * Copyright (c) 2003-2007 Network Appliance, Inc. All rights reserved.
  *
  * This software is available to you under a choice of one of two
@@ -49,9 +50,10 @@
 
 #include <linux/interrupt.h>
 #include <linux/slab.h>
-#include <linux/prefetch.h>
 #include <linux/sunrpc/addr.h>
 #include <linux/sunrpc/svc_rdma.h>
+
+#include <asm-generic/barrier.h>
 #include <asm/bitops.h>
 
 #include <rdma/ib_cm.h>
@@ -73,7 +75,7 @@ static void rpcrdma_create_mrs(struct rpcrdma_xprt *r_xprt);
 static void rpcrdma_destroy_mrs(struct rpcrdma_buffer *buf);
 static void rpcrdma_dma_unmap_regbuf(struct rpcrdma_regbuf *rb);
 
-static struct workqueue_struct *rpcrdma_receive_wq __read_mostly;
+struct workqueue_struct *rpcrdma_receive_wq __read_mostly;
 
 int
 rpcrdma_alloc_wq(void)
@@ -126,30 +128,17 @@ rpcrdma_qp_async_error_upcall(struct ib_event *event, void *context)
 static void
 rpcrdma_wc_send(struct ib_cq *cq, struct ib_wc *wc)
 {
+	struct ib_cqe *cqe = wc->wr_cqe;
+	struct rpcrdma_sendctx *sc =
+		container_of(cqe, struct rpcrdma_sendctx, sc_cqe);
+
 	/* WARNING: Only wr_cqe and status are reliable at this point */
 	if (wc->status != IB_WC_SUCCESS && wc->status != IB_WC_WR_FLUSH_ERR)
 		pr_err("rpcrdma: Send: %s (%u/0x%x)\n",
 		       ib_wc_status_msg(wc->status),
 		       wc->status, wc->vendor_err);
-}
 
-/* Perform basic sanity checking to avoid using garbage
+	rpcrdma_sendctx_put_locked(sc);
- * to update the credit grant value.
- */
-static void
-rpcrdma_update_granted_credits(struct rpcrdma_rep *rep)
-{
-	struct rpcrdma_buffer *buffer = &rep->rr_rxprt->rx_buf;
-	__be32 *p = rep->rr_rdmabuf->rg_base;
-	u32 credits;
-
-	credits = be32_to_cpup(p + 2);
-	if (credits == 0)
-		credits = 1;	/* don't deadlock */
-	else if (credits > buffer->rb_max_requests)
-		credits = buffer->rb_max_requests;
-
-	atomic_set(&buffer->rb_credits, credits);
 }
 
 /**
@@ -181,11 +170,8 @@ rpcrdma_wc_receive(struct ib_cq *cq, struct ib_wc *wc)
 				   rdmab_addr(rep->rr_rdmabuf),
 				   wc->byte_len, DMA_FROM_DEVICE);
 
-	if (wc->byte_len >= RPCRDMA_HDRLEN_ERR)
-		rpcrdma_update_granted_credits(rep);
-
 out_schedule:
-	queue_work(rpcrdma_receive_wq, &rep->rr_work);
+	rpcrdma_reply_handler(rep);
 	return;
 
 out_fail:
@@ -295,7 +281,7 @@ rpcrdma_conn_upcall(struct rdma_cm_id *id, struct rdma_cm_event *event)
 	case RDMA_CM_EVENT_DISCONNECTED:
 		connstate = -ECONNABORTED;
 connected:
-		atomic_set(&xprt->rx_buf.rb_credits, 1);
+		xprt->rx_buf.rb_credits = 1;
 		ep->rep_connected = connstate;
 		rpcrdma_conn_func(ep);
 		wake_up_all(&ep->rep_connect_wait);
@@ -564,16 +550,15 @@ rpcrdma_ep_create(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia,
 		ep->rep_attr.cap.max_recv_sge);
 
 	/* set trigger for requesting send completion */
-	ep->rep_cqinit = ep->rep_attr.cap.max_send_wr/2 - 1;
+	ep->rep_send_batch = min_t(unsigned int, RPCRDMA_MAX_SEND_BATCH,
-	if (ep->rep_cqinit <= 2)
+				   cdata->max_requests >> 2);
-		ep->rep_cqinit = 0;	/* always signal? */
+	ep->rep_send_count = ep->rep_send_batch;
-	rpcrdma_init_cqcount(ep, 0);
 	init_waitqueue_head(&ep->rep_connect_wait);
 	INIT_DELAYED_WORK(&ep->rep_connect_worker, rpcrdma_connect_worker);
 
 	sendcq = ib_alloc_cq(ia->ri_device, NULL,
 			     ep->rep_attr.cap.max_send_wr + 1,
-			     0, IB_POLL_SOFTIRQ);
+			     1, IB_POLL_WORKQUEUE);
 	if (IS_ERR(sendcq)) {
 		rc = PTR_ERR(sendcq);
 		dprintk("RPC:       %s: failed to create send CQ: %i\n",
@@ -583,7 +568,7 @@ rpcrdma_ep_create(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia,
 
 	recvcq = ib_alloc_cq(ia->ri_device, NULL,
 			     ep->rep_attr.cap.max_recv_wr + 1,
-			     0, IB_POLL_SOFTIRQ);
+			     0, IB_POLL_WORKQUEUE);
 	if (IS_ERR(recvcq)) {
 		rc = PTR_ERR(recvcq);
 		dprintk("RPC:       %s: failed to create recv CQ: %i\n",
@@ -846,6 +831,168 @@ rpcrdma_ep_disconnect(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia)
 	ib_drain_qp(ia->ri_id->qp);
 }
 
+/* Fixed-size circular FIFO queue. This implementation is wait-free and
+ * lock-free.
+ *
+ * Consumer is the code path that posts Sends. This path dequeues a
+ * sendctx for use by a Send operation. Multiple consumer threads
+ * are serialized by the RPC transport lock, which allows only one
+ * ->send_request call at a time.
+ *
+ * Producer is the code path that handles Send completions. This path
+ * enqueues a sendctx that has been completed. Multiple producer
+ * threads are serialized by the ib_poll_cq() function.
+ */
+
+/* rpcrdma_sendctxs_destroy() assumes caller has already quiesced
+ * queue activity, and ib_drain_qp has flushed all remaining Send
+ * requests.
+ */
+static void rpcrdma_sendctxs_destroy(struct rpcrdma_buffer *buf)
+{
+	unsigned long i;
+
+	for (i = 0; i <= buf->rb_sc_last; i++)
+		kfree(buf->rb_sc_ctxs[i]);
+	kfree(buf->rb_sc_ctxs);
+}
+
+static struct rpcrdma_sendctx *rpcrdma_sendctx_create(struct rpcrdma_ia *ia)
+{
+	struct rpcrdma_sendctx *sc;
+
+	sc = kzalloc(sizeof(*sc) +
+		     ia->ri_max_send_sges * sizeof(struct ib_sge),
+		     GFP_KERNEL);
+	if (!sc)
+		return NULL;
+
+	sc->sc_wr.wr_cqe = &sc->sc_cqe;
+	sc->sc_wr.sg_list = sc->sc_sges;
+	sc->sc_wr.opcode = IB_WR_SEND;
+	sc->sc_cqe.done = rpcrdma_wc_send;
+	return sc;
+}
+
+static int rpcrdma_sendctxs_create(struct rpcrdma_xprt *r_xprt)
+{
+	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
+	struct rpcrdma_sendctx *sc;
+	unsigned long i;
+
+	/* Maximum number of concurrent outstanding Send WRs. Capping
+	 * the circular queue size stops Send Queue overflow by causing
+	 * the ->send_request call to fail temporarily before too many
+	 * Sends are posted.
+	 */
+	i = buf->rb_max_requests + RPCRDMA_MAX_BC_REQUESTS;
+	dprintk("RPC:       %s: allocating %lu send_ctxs\n", __func__, i);
+	buf->rb_sc_ctxs = kcalloc(i, sizeof(sc), GFP_KERNEL);
+	if (!buf->rb_sc_ctxs)
+		return -ENOMEM;
+
+	buf->rb_sc_last = i - 1;
+	for (i = 0; i <= buf->rb_sc_last; i++) {
+		sc = rpcrdma_sendctx_create(&r_xprt->rx_ia);
+		if (!sc)
+			goto out_destroy;
+
+		sc->sc_xprt = r_xprt;
+		buf->rb_sc_ctxs[i] = sc;
+	}
+
+	return 0;
+
+out_destroy:
+	rpcrdma_sendctxs_destroy(buf);
+	return -ENOMEM;
+}
+
+/* The sendctx queue is not guaranteed to have a size that is a
+ * power of two, thus the helpers in circ_buf.h cannot be used.
+ * The other option is to use modulus (%), which can be expensive.
+ */
+static unsigned long rpcrdma_sendctx_next(struct rpcrdma_buffer *buf,
+					  unsigned long item)
+{
+	return likely(item < buf->rb_sc_last) ? item + 1 : 0;
+}
+
+/**
+ * rpcrdma_sendctx_get_locked - Acquire a send context
+ * @buf: transport buffers from which to acquire an unused context
+ *
+ * Returns pointer to a free send completion context; or NULL if
+ * the queue is empty.
+ *
+ * Usage: Called to acquire an SGE array before preparing a Send WR.
+ *
+ * The caller serializes calls to this function (per rpcrdma_buffer),
+ * and provides an effective memory barrier that flushes the new value
+ * of rb_sc_head.
+ */
+struct rpcrdma_sendctx *rpcrdma_sendctx_get_locked(struct rpcrdma_buffer *buf)
+{
+	struct rpcrdma_xprt *r_xprt;
+	struct rpcrdma_sendctx *sc;
+	unsigned long next_head;
+
+	next_head = rpcrdma_sendctx_next(buf, buf->rb_sc_head);
+
+	if (next_head == READ_ONCE(buf->rb_sc_tail))
+		goto out_emptyq;
+
+	/* ORDER: item must be accessed _before_ head is updated */
+	sc = buf->rb_sc_ctxs[next_head];
+
+	/* Releasing the lock in the caller acts as a memory
+	 * barrier that flushes rb_sc_head.
+	 */
+	buf->rb_sc_head = next_head;
+
+	return sc;
+
+out_emptyq:
+	/* The queue is "empty" if there have not been enough Send
+	 * completions recently. This is a sign the Send Queue is
+	 * backing up. Cause the caller to pause and try again.
+	 */
+	dprintk("RPC:       %s: empty sendctx queue\n", __func__);
+	r_xprt = container_of(buf, struct rpcrdma_xprt, rx_buf);
+	r_xprt->rx_stats.empty_sendctx_q++;
+	return NULL;
+}
+
+/**
+ * rpcrdma_sendctx_put_locked - Release a send context
+ * @sc: send context to release
+ *
+ * Usage: Called from Send completion to return a sendctxt
+ * to the queue.
+ *
+ * The caller serializes calls to this function (per rpcrdma_buffer).
+ */
+void rpcrdma_sendctx_put_locked(struct rpcrdma_sendctx *sc)
+{
+	struct rpcrdma_buffer *buf = &sc->sc_xprt->rx_buf;
+	unsigned long next_tail;
+
+	/* Unmap SGEs of previously completed by unsignaled
+	 * Sends by walking up the queue until @sc is found.
+	 */
+	next_tail = buf->rb_sc_tail;
+	do {
+		next_tail = rpcrdma_sendctx_next(buf, next_tail);
+
+		/* ORDER: item must be accessed _before_ tail is updated */
+		rpcrdma_unmap_sendctx(buf->rb_sc_ctxs[next_tail]);
+
+	} while (buf->rb_sc_ctxs[next_tail] != sc);
+
+	/* Paired with READ_ONCE */
+	smp_store_release(&buf->rb_sc_tail, next_tail);
+}
+
 static void
 rpcrdma_mr_recovery_worker(struct work_struct *work)
 {
@@ -941,13 +1088,8 @@ rpcrdma_create_req(struct rpcrdma_xprt *r_xprt)
 	spin_lock(&buffer->rb_reqslock);
 	list_add(&req->rl_all, &buffer->rb_allreqs);
 	spin_unlock(&buffer->rb_reqslock);
-	req->rl_cqe.done = rpcrdma_wc_send;
 	req->rl_buffer = &r_xprt->rx_buf;
 	INIT_LIST_HEAD(&req->rl_registered);
-	req->rl_send_wr.next = NULL;
-	req->rl_send_wr.wr_cqe = &req->rl_cqe;
-	req->rl_send_wr.sg_list = req->rl_send_sge;
-	req->rl_send_wr.opcode = IB_WR_SEND;
 	return req;
 }
 
@@ -974,7 +1116,7 @@ rpcrdma_create_rep(struct rpcrdma_xprt *r_xprt)
 
 	rep->rr_cqe.done = rpcrdma_wc_receive;
 	rep->rr_rxprt = r_xprt;
-	INIT_WORK(&rep->rr_work, rpcrdma_reply_handler);
+	INIT_WORK(&rep->rr_work, rpcrdma_deferred_completion);
 	rep->rr_recv_wr.next = NULL;
 	rep->rr_recv_wr.wr_cqe = &rep->rr_cqe;
 	rep->rr_recv_wr.sg_list = &rep->rr_rdmabuf->rg_iov;
@@ -995,7 +1137,6 @@ rpcrdma_buffer_create(struct rpcrdma_xprt *r_xprt)
 
 	buf->rb_max_requests = r_xprt->rx_data.max_requests;
 	buf->rb_bc_srv_max_requests = 0;
-	atomic_set(&buf->rb_credits, 1);
 	spin_lock_init(&buf->rb_mwlock);
 	spin_lock_init(&buf->rb_lock);
 	spin_lock_init(&buf->rb_recovery_lock);
@@ -1022,7 +1163,6 @@ rpcrdma_buffer_create(struct rpcrdma_xprt *r_xprt)
 			rc = PTR_ERR(req);
 			goto out;
 		}
-		req->rl_backchannel = false;
 		list_add(&req->rl_list, &buf->rb_send_bufs);
 	}
 
@@ -1040,6 +1180,10 @@ rpcrdma_buffer_create(struct rpcrdma_xprt *r_xprt)
 		list_add(&rep->rr_list, &buf->rb_recv_bufs);
 	}
 
+	rc = rpcrdma_sendctxs_create(r_xprt);
+	if (rc)
+		goto out;
+
 	return 0;
 out:
 	rpcrdma_buffer_destroy(buf);
@@ -1116,6 +1260,8 @@ rpcrdma_buffer_destroy(struct rpcrdma_buffer *buf)
 	cancel_delayed_work_sync(&buf->rb_recovery_worker);
 	cancel_delayed_work_sync(&buf->rb_refresh_worker);
 
+	rpcrdma_sendctxs_destroy(buf);
+
 	while (!list_empty(&buf->rb_recv_bufs)) {
 		struct rpcrdma_rep *rep;
 
@@ -1231,7 +1377,6 @@ rpcrdma_buffer_put(struct rpcrdma_req *req)
 	struct rpcrdma_buffer *buffers = req->rl_buffer;
 	struct rpcrdma_rep *rep = req->rl_reply;
 
-	req->rl_send_wr.num_sge = 0;
 	req->rl_reply = NULL;
 
 	spin_lock(&buffers->rb_lock);
@@ -1363,7 +1508,7 @@ rpcrdma_ep_post(struct rpcrdma_ia *ia,
 		struct rpcrdma_ep *ep,
 		struct rpcrdma_req *req)
 {
-	struct ib_send_wr *send_wr = &req->rl_send_wr;
+	struct ib_send_wr *send_wr = &req->rl_sendctx->sc_wr;
 	struct ib_send_wr *send_wr_fail;
 	int rc;
 
@@ -1377,7 +1522,14 @@ rpcrdma_ep_post(struct rpcrdma_ia *ia,
 	dprintk("RPC:       %s: posting %d s/g entries\n",
 		__func__, send_wr->num_sge);
 
-	rpcrdma_set_signaled(ep, send_wr);
+	if (!ep->rep_send_count ||
+	    test_bit(RPCRDMA_REQ_F_TX_RESOURCES, &req->rl_flags)) {
+		send_wr->send_flags |= IB_SEND_SIGNALED;
+		ep->rep_send_count = ep->rep_send_batch;
+	} else {
+		send_wr->send_flags &= ~IB_SEND_SIGNALED;
+		--ep->rep_send_count;
+	}
 	rc = ib_post_send(ia->ri_id->qp, send_wr, &send_wr_fail);
 	if (rc)
 		goto out_postsend_err;

net/sunrpc/xprtrdma/xprt_rdma.h

@@ -1,4 +1,5 @@
 /*
+ * Copyright (c) 2014-2017 Oracle.  All rights reserved.
  * Copyright (c) 2003-2007 Network Appliance, Inc. All rights reserved.
  *
  * This software is available to you under a choice of one of two
@@ -93,8 +94,8 @@ enum {
  */
 
 struct rpcrdma_ep {
-	atomic_t		rep_cqcount;
+	unsigned int		rep_send_count;
-	int			rep_cqinit;
+	unsigned int		rep_send_batch;
 	int			rep_connected;
 	struct ib_qp_init_attr	rep_attr;
 	wait_queue_head_t 	rep_connect_wait;
@@ -104,25 +105,6 @@ struct rpcrdma_ep {
 	struct delayed_work	rep_connect_worker;
 };
 
-static inline void
-rpcrdma_init_cqcount(struct rpcrdma_ep *ep, int count)
-{
-	atomic_set(&ep->rep_cqcount, ep->rep_cqinit - count);
-}
-
-/* To update send queue accounting, provider must take a
- * send completion every now and then.
- */
-static inline void
-rpcrdma_set_signaled(struct rpcrdma_ep *ep, struct ib_send_wr *send_wr)
-{
-	send_wr->send_flags = 0;
-	if (unlikely(atomic_sub_return(1, &ep->rep_cqcount) <= 0)) {
-		rpcrdma_init_cqcount(ep, 0);
-		send_wr->send_flags = IB_SEND_SIGNALED;
-	}
-}
-
 /* Pre-allocate extra Work Requests for handling backward receives
  * and sends. This is a fixed value because the Work Queues are
  * allocated when the forward channel is set up.
@@ -164,12 +146,6 @@ rdmab_lkey(struct rpcrdma_regbuf *rb)
 	return rb->rg_iov.lkey;
 }
 
-static inline struct rpcrdma_msg *
-rdmab_to_msg(struct rpcrdma_regbuf *rb)
-{
-	return (struct rpcrdma_msg *)rb->rg_base;
-}
-
 static inline struct ib_device *
 rdmab_device(struct rpcrdma_regbuf *rb)
 {
@@ -202,22 +178,24 @@ enum {
 };
 
 /*
- * struct rpcrdma_rep -- this structure encapsulates state required to recv
+ * struct rpcrdma_rep -- this structure encapsulates state required
- * and complete a reply, asychronously. It needs several pieces of
+ * to receive and complete an RPC Reply, asychronously. It needs
- * state:
+ * several pieces of state:
- *   o recv buffer (posted to provider)
- *   o ib_sge (also donated to provider)
- *   o status of reply (length, success or not)
- *   o bookkeeping state to get run by reply handler (list, etc)
  *
- * These are allocated during initialization, per-transport instance.
+ *   o receive buffer and ib_sge (donated to provider)
+ *   o status of receive (success or not, length, inv rkey)
+ *   o bookkeeping state to get run by reply handler (XDR stream)
  *
- * N of these are associated with a transport instance, and stored in
+ * These structures are allocated during transport initialization.
- * struct rpcrdma_buffer. N is the max number of outstanding requests.
+ * N of these are associated with a transport instance, managed by
+ * struct rpcrdma_buffer. N is the max number of outstanding RPCs.
  */
 
 struct rpcrdma_rep {
 	struct ib_cqe		rr_cqe;
+	__be32			rr_xid;
+	__be32			rr_vers;
+	__be32			rr_proc;
 	int			rr_wc_flags;
 	u32			rr_inv_rkey;
 	struct rpcrdma_regbuf	*rr_rdmabuf;
@@ -225,10 +203,34 @@ struct rpcrdma_rep {
 	struct work_struct	rr_work;
 	struct xdr_buf		rr_hdrbuf;
 	struct xdr_stream	rr_stream;
+	struct rpc_rqst		*rr_rqst;
 	struct list_head	rr_list;
 	struct ib_recv_wr	rr_recv_wr;
 };
 
+/* struct rpcrdma_sendctx - DMA mapped SGEs to unmap after Send completes
+ */
+struct rpcrdma_req;
+struct rpcrdma_xprt;
+struct rpcrdma_sendctx {
+	struct ib_send_wr	sc_wr;
+	struct ib_cqe		sc_cqe;
+	struct rpcrdma_xprt	*sc_xprt;
+	struct rpcrdma_req	*sc_req;
+	unsigned int		sc_unmap_count;
+	struct ib_sge		sc_sges[];
+};
+
+/* Limit the number of SGEs that can be unmapped during one
+ * Send completion. This caps the amount of work a single
+ * completion can do before returning to the provider.
+ *
+ * Setting this to zero disables Send completion batching.
+ */
+enum {
+	RPCRDMA_MAX_SEND_BATCH = 7,
+};
+
 /*
  * struct rpcrdma_mw - external memory region metadata
  *
@@ -340,26 +342,30 @@ enum {
 struct rpcrdma_buffer;
 struct rpcrdma_req {
 	struct list_head	rl_list;
-	unsigned int		rl_mapped_sges;
 	unsigned int		rl_connect_cookie;
 	struct rpcrdma_buffer	*rl_buffer;
 	struct rpcrdma_rep	*rl_reply;
 	struct xdr_stream	rl_stream;
 	struct xdr_buf		rl_hdrbuf;
-	struct ib_send_wr	rl_send_wr;
+	struct rpcrdma_sendctx	*rl_sendctx;
-	struct ib_sge		rl_send_sge[RPCRDMA_MAX_SEND_SGES];
 	struct rpcrdma_regbuf	*rl_rdmabuf;	/* xprt header */
 	struct rpcrdma_regbuf	*rl_sendbuf;	/* rq_snd_buf */
 	struct rpcrdma_regbuf	*rl_recvbuf;	/* rq_rcv_buf */
 
-	struct ib_cqe		rl_cqe;
 	struct list_head	rl_all;
-	bool			rl_backchannel;
+	unsigned long		rl_flags;
 
 	struct list_head	rl_registered;	/* registered segments */
 	struct rpcrdma_mr_seg	rl_segments[RPCRDMA_MAX_SEGS];
 };
 
+/* rl_flags */
+enum {
+	RPCRDMA_REQ_F_BACKCHANNEL = 0,
+	RPCRDMA_REQ_F_PENDING,
+	RPCRDMA_REQ_F_TX_RESOURCES,
+};
+
 static inline void
 rpcrdma_set_xprtdata(struct rpc_rqst *rqst, struct rpcrdma_req *req)
 {
@@ -399,12 +405,17 @@ struct rpcrdma_buffer {
 	struct list_head	rb_mws;
 	struct list_head	rb_all;
 
+	unsigned long		rb_sc_head;
+	unsigned long		rb_sc_tail;
+	unsigned long		rb_sc_last;
+	struct rpcrdma_sendctx	**rb_sc_ctxs;
+
 	spinlock_t		rb_lock;	/* protect buf lists */
 	int			rb_send_count, rb_recv_count;
 	struct list_head	rb_send_bufs;
 	struct list_head	rb_recv_bufs;
 	u32			rb_max_requests;
-	atomic_t		rb_credits;	/* most recent credit grant */
+	u32			rb_credits;	/* most recent credit grant */
 
 	u32			rb_bc_srv_max_requests;
 	spinlock_t		rb_reqslock;	/* protect rb_allreqs */
@@ -453,10 +464,12 @@ struct rpcrdma_stats {
 	unsigned long		mrs_recovered;
 	unsigned long		mrs_orphaned;
 	unsigned long		mrs_allocated;
+	unsigned long		empty_sendctx_q;
 
 	/* accessed when receiving a reply */
 	unsigned long long	total_rdma_reply;
 	unsigned long long	fixup_copy_count;
+	unsigned long		reply_waits_for_send;
 	unsigned long		local_inv_needed;
 	unsigned long		nomsg_call_count;
 	unsigned long		bcall_count;
@@ -473,8 +486,6 @@ struct rpcrdma_memreg_ops {
 				  struct rpcrdma_mw **);
 	void		(*ro_unmap_sync)(struct rpcrdma_xprt *,
 					 struct list_head *);
-	void		(*ro_unmap_safe)(struct rpcrdma_xprt *,
-					 struct rpcrdma_req *, bool);
 	void		(*ro_recover_mr)(struct rpcrdma_mw *);
 	int		(*ro_open)(struct rpcrdma_ia *,
 				   struct rpcrdma_ep *,
@@ -532,6 +543,8 @@ void rpcrdma_ia_close(struct rpcrdma_ia *);
 bool frwr_is_supported(struct rpcrdma_ia *);
 bool fmr_is_supported(struct rpcrdma_ia *);
 
+extern struct workqueue_struct *rpcrdma_receive_wq;
+
 /*
  * Endpoint calls - xprtrdma/verbs.c
  */
@@ -554,6 +567,8 @@ struct rpcrdma_rep *rpcrdma_create_rep(struct rpcrdma_xprt *);
 void rpcrdma_destroy_req(struct rpcrdma_req *);
 int rpcrdma_buffer_create(struct rpcrdma_xprt *);
 void rpcrdma_buffer_destroy(struct rpcrdma_buffer *);
+struct rpcrdma_sendctx *rpcrdma_sendctx_get_locked(struct rpcrdma_buffer *buf);
+void rpcrdma_sendctx_put_locked(struct rpcrdma_sendctx *sc);
 
 struct rpcrdma_mw *rpcrdma_get_mw(struct rpcrdma_xprt *);
 void rpcrdma_put_mw(struct rpcrdma_xprt *, struct rpcrdma_mw *);
@@ -610,12 +625,18 @@ enum rpcrdma_chunktype {
 	rpcrdma_replych
 };
 
-bool rpcrdma_prepare_send_sges(struct rpcrdma_ia *, struct rpcrdma_req *,
+int rpcrdma_prepare_send_sges(struct rpcrdma_xprt *r_xprt,
-			       u32, struct xdr_buf *, enum rpcrdma_chunktype);
+			      struct rpcrdma_req *req, u32 hdrlen,
-void rpcrdma_unmap_sges(struct rpcrdma_ia *, struct rpcrdma_req *);
+			      struct xdr_buf *xdr,
+			      enum rpcrdma_chunktype rtype);
+void rpcrdma_unmap_sendctx(struct rpcrdma_sendctx *sc);
 int rpcrdma_marshal_req(struct rpcrdma_xprt *r_xprt, struct rpc_rqst *rqst);
 void rpcrdma_set_max_header_sizes(struct rpcrdma_xprt *);
-void rpcrdma_reply_handler(struct work_struct *work);
+void rpcrdma_complete_rqst(struct rpcrdma_rep *rep);
+void rpcrdma_reply_handler(struct rpcrdma_rep *rep);
+void rpcrdma_release_rqst(struct rpcrdma_xprt *r_xprt,
+			  struct rpcrdma_req *req);
+void rpcrdma_deferred_completion(struct work_struct *work);
 
 static inline void rpcrdma_set_xdrlen(struct xdr_buf *xdr, size_t len)
 {

net/sunrpc/xprtsock.c

@@ -552,6 +552,7 @@ static int xs_local_send_request(struct rpc_task *task)
 	default:
 		dprintk("RPC:       sendmsg returned unrecognized error %d\n",
 			-status);
+		/* fall through */
 	case -EPIPE:
 		xs_close(xprt);
 		status = -ENOTCONN;
@@ -1611,6 +1612,7 @@ static void xs_tcp_state_change(struct sock *sk)
 		xprt->connect_cookie++;
 		clear_bit(XPRT_CONNECTED, &xprt->state);
 		xs_tcp_force_close(xprt);
+		/* fall through */
 	case TCP_CLOSING:
 		/*
 		 * If the server closed down the connection, make sure that
@@ -2368,6 +2370,7 @@ static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
 	switch (ret) {
 	case 0:
 		xs_set_srcport(transport, sock);
+		/* fall through */
 	case -EINPROGRESS:
 		/* SYN_SENT! */
 		if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
@@ -2419,6 +2422,7 @@ static void xs_tcp_setup_socket(struct work_struct *work)
 	default:
 		printk("%s: connect returned unhandled error %d\n",
 			__func__, status);
+		/* fall through */
 	case -EADDRNOTAVAIL:
 		/* We're probably in TIME_WAIT. Get rid of existing socket,
 		 * and retry