Merge tag 'nfs-for-5.14-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

Pull NFS client updates from Trond Myklebust:
 "Highlights include:

  Features:

   - Multiple patches to add support for fcntl() leases over NFSv4.

   - A sysfs interface to display more information about the various
     transport connections used by the RPC client

   - A sysfs interface to allow a suitably privileged user to offline a
     transport that may no longer point to a valid server

   - A sysfs interface to allow a suitably privileged user to change the
     server IP address used by the RPC client

  Stable fixes:

   - Two sunrpc fixes for deadlocks involving privileged rpc_wait_queues

  Bugfixes:

   - SUNRPC: Avoid a KASAN slab-out-of-bounds bug in xdr_set_page_base()

   - SUNRPC: prevent port reuse on transports which don't request it.

   - NFSv3: Fix memory leak in posix_acl_create()

   - NFS: Various fixes to attribute revalidation timeouts

   - NFSv4: Fix handling of non-atomic change attribute updates

   - NFSv4: If a server is down, don't cause mounts to other servers to
     hang as well

   - pNFS: Fix an Oops in pnfs_mark_request_commit() when doing O_DIRECT

   - NFS: Fix mount failures due to incorrect setting of the
     has_sec_mnt_opts filesystem flag

   - NFS: Ensure nfs_readpage returns promptly when an internal error
     occurs

   - NFS: Fix fscache read from NFS after cache error

   - pNFS: Various bugfixes around the LAYOUTGET operation"

* tag 'nfs-for-5.14-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs: (46 commits)
  NFSv4/pNFS: Return an error if _nfs4_pnfs_v3_ds_connect can't load NFSv3
  NFSv4/pNFS: Don't call _nfs4_pnfs_v3_ds_connect multiple times
  NFSv4/pnfs: Clean up layout get on open
  NFSv4/pnfs: Fix layoutget behaviour after invalidation
  NFSv4/pnfs: Fix the layout barrier update
  NFS: Fix fscache read from NFS after cache error
  NFS: Ensure nfs_readpage returns promptly when internal error occurs
  sunrpc: remove an offlined xprt using sysfs
  sunrpc: provide showing transport's state info in the sysfs directory
  sunrpc: display xprt's queuelen of assigned tasks via sysfs
  sunrpc: provide multipath info in the sysfs directory
  NFSv4.1 identify and mark RPC tasks that can move between transports
  sunrpc: provide transport info in the sysfs directory
  SUNRPC: take a xprt offline using sysfs
  sunrpc: add dst_attr attributes to the sysfs xprt directory
  SUNRPC for TCP display xprt's source port in sysfs xprt_info
  SUNRPC query transport's source port
  SUNRPC display xprt's main value in sysfs's xprt_info
  SUNRPC mark the first transport
  sunrpc: add add sysfs directory per xprt under each xprt_switch
  ...

11 files changed, 770 insertions, 10 deletions

diff --git a/net/sunrpc/Makefile b/net/sunrpc/Makefile
index 9488600451e8..1c8de397d6ad 100644
--- a/net/sunrpc/Makefile
+++ b/net/sunrpc/Makefile
@@ -12,7 +12,7 @@ sunrpc-y := clnt.o xprt.o socklib.o xprtsock.o sched.o \
 	    auth.o auth_null.o auth_unix.o \
 	    svc.o svcsock.o svcauth.o svcauth_unix.o \
 	    addr.o rpcb_clnt.o timer.o xdr.o \
-	    sunrpc_syms.o cache.o rpc_pipe.o \
+	    sunrpc_syms.o cache.o rpc_pipe.o sysfs.o \
 	    svc_xprt.o \
 	    xprtmultipath.o
 sunrpc-$(CONFIG_SUNRPC_DEBUG) += debugfs.o
diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index 42623d6b8f0e..8b4de70e8ead 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -41,6 +41,7 @@
 #include <trace/events/sunrpc.h>
 
 #include "sunrpc.h"
+#include "sysfs.h"
 #include "netns.h"
 
 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
@@ -327,6 +328,7 @@ err_auth:
 out:
 	if (pipefs_sb)
 		rpc_put_sb_net(net);
+	rpc_sysfs_client_destroy(clnt);
 	rpc_clnt_debugfs_unregister(clnt);
 	return err;
 }
@@ -410,6 +412,7 @@ static struct rpc_clnt * rpc_new_client(const struct rpc_create_args *args,
 	}
 
 	rpc_clnt_set_transport(clnt, xprt, timeout);
+	xprt->main = true;
 	xprt_iter_init(&clnt->cl_xpi, xps);
 	xprt_switch_put(xps);
 
@@ -423,6 +426,7 @@ static struct rpc_clnt * rpc_new_client(const struct rpc_create_args *args,
 	/* save the nodename */
 	rpc_clnt_set_nodename(clnt, nodename);
 
+	rpc_sysfs_client_setup(clnt, xps, rpc_net_ns(clnt));
 	err = rpc_client_register(clnt, args->authflavor, args->client_name);
 	if (err)
 		goto out_no_path;
@@ -733,6 +737,7 @@ int rpc_switch_client_transport(struct rpc_clnt *clnt,
 
 	rpc_unregister_client(clnt);
 	__rpc_clnt_remove_pipedir(clnt);
+	rpc_sysfs_client_destroy(clnt);
 	rpc_clnt_debugfs_unregister(clnt);
 
 	/*
@@ -879,6 +884,7 @@ static void rpc_free_client_work(struct work_struct *work)
 	 * so they cannot be called in rpciod, so they are handled separately
 	 * here.
 	 */
+	rpc_sysfs_client_destroy(clnt);
 	rpc_clnt_debugfs_unregister(clnt);
 	rpc_free_clid(clnt);
 	rpc_clnt_remove_pipedir(clnt);
@@ -2100,6 +2106,30 @@ call_connect_status(struct rpc_task *task)
 	case -ENOTCONN:
 	case -EAGAIN:
 	case -ETIMEDOUT:
+		if (!(task->tk_flags & RPC_TASK_NO_ROUND_ROBIN) &&
+		    (task->tk_flags & RPC_TASK_MOVEABLE) &&
+		    test_bit(XPRT_REMOVE, &xprt->state)) {
+			struct rpc_xprt *saved = task->tk_xprt;
+			struct rpc_xprt_switch *xps;
+
+			rcu_read_lock();
+			xps = xprt_switch_get(rcu_dereference(clnt->cl_xpi.xpi_xpswitch));
+			rcu_read_unlock();
+			if (xps->xps_nxprts > 1) {
+				long value;
+
+				xprt_release(task);
+				value = atomic_long_dec_return(&xprt->queuelen);
+				if (value == 0)
+					rpc_xprt_switch_remove_xprt(xps, saved);
+				xprt_put(saved);
+				task->tk_xprt = NULL;
+				task->tk_action = call_start;
+			}
+			xprt_switch_put(xps);
+			if (!task->tk_xprt)
+				return;
+		}
 		goto out_retry;
 	case -ENOBUFS:
 		rpc_delay(task, HZ >> 2);
diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c
index 39ed0e0afe6d..c045f63d11fa 100644
--- a/net/sunrpc/sched.c
+++ b/net/sunrpc/sched.c
@@ -592,10 +592,20 @@ static struct rpc_task *__rpc_find_next_queued_priority(struct rpc_wait_queue *q
 	struct rpc_task *task;
 
 	/*
+	 * Service the privileged queue.
+	 */
+	q = &queue->tasks[RPC_NR_PRIORITY - 1];
+	if (queue->maxpriority > RPC_PRIORITY_PRIVILEGED && !list_empty(q)) {
+		task = list_first_entry(q, struct rpc_task, u.tk_wait.list);
+		goto out;
+	}
+
+	/*
 	 * Service a batch of tasks from a single owner.
 	 */
 	q = &queue->tasks[queue->priority];
-	if (!list_empty(q) && --queue->nr) {
+	if (!list_empty(q) && queue->nr) {
+		queue->nr--;
 		task = list_first_entry(q, struct rpc_task, u.tk_wait.list);
 		goto out;
 	}
diff --git a/net/sunrpc/sunrpc_syms.c b/net/sunrpc/sunrpc_syms.c
index 236fadc4a439..691c0000e9ea 100644
--- a/net/sunrpc/sunrpc_syms.c
+++ b/net/sunrpc/sunrpc_syms.c
@@ -24,6 +24,7 @@
 #include <linux/sunrpc/xprtsock.h>
 
 #include "sunrpc.h"
+#include "sysfs.h"
 #include "netns.h"
 
 unsigned int sunrpc_net_id;
@@ -103,6 +104,10 @@ init_sunrpc(void)
 	if (err)
 		goto out4;
 
+	err = rpc_sysfs_init();
+	if (err)
+		goto out5;
+
 	sunrpc_debugfs_init();
 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
 	rpc_register_sysctl();
@@ -111,6 +116,8 @@ init_sunrpc(void)
 	init_socket_xprt();	/* clnt sock transport */
 	return 0;
 
+out5:
+	unregister_rpc_pipefs();
 out4:
 	unregister_pernet_subsys(&sunrpc_net_ops);
 out3:
@@ -124,7 +131,10 @@ out:
 static void __exit
 cleanup_sunrpc(void)
 {
+	rpc_sysfs_exit();
 	rpc_cleanup_clids();
+	xprt_cleanup_ids();
+	xprt_multipath_cleanup_ids();
 	rpcauth_remove_module();
 	cleanup_socket_xprt();
 	svc_cleanup_xprt_sock();
diff --git a/net/sunrpc/sysfs.c b/net/sunrpc/sysfs.c
new file mode 100644
index 000000000000..64da3bfd28e6
--- /dev/null
+++ b/net/sunrpc/sysfs.c
@@ -0,0 +1,588 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (c) 2020 Anna Schumaker <Anna.Schumaker@Netapp.com>
+ */
+#include <linux/sunrpc/clnt.h>
+#include <linux/kobject.h>
+#include <linux/sunrpc/addr.h>
+#include <linux/sunrpc/xprtsock.h>
+
+#include "sysfs.h"
+
+struct xprt_addr {
+	const char *addr;
+	struct rcu_head rcu;
+};
+
+static void free_xprt_addr(struct rcu_head *head)
+{
+	struct xprt_addr *addr = container_of(head, struct xprt_addr, rcu);
+
+	kfree(addr->addr);
+	kfree(addr);
+}
+
+static struct kset *rpc_sunrpc_kset;
+static struct kobject *rpc_sunrpc_client_kobj, *rpc_sunrpc_xprt_switch_kobj;
+
+static void rpc_sysfs_object_release(struct kobject *kobj)
+{
+	kfree(kobj);
+}
+
+static const struct kobj_ns_type_operations *
+rpc_sysfs_object_child_ns_type(struct kobject *kobj)
+{
+	return &net_ns_type_operations;
+}
+
+static struct kobj_type rpc_sysfs_object_type = {
+	.release = rpc_sysfs_object_release,
+	.sysfs_ops = &kobj_sysfs_ops,
+	.child_ns_type = rpc_sysfs_object_child_ns_type,
+};
+
+static struct kobject *rpc_sysfs_object_alloc(const char *name,
+					      struct kset *kset,
+					      struct kobject *parent)
+{
+	struct kobject *kobj;
+
+	kobj = kzalloc(sizeof(*kobj), GFP_KERNEL);
+	if (kobj) {
+		kobj->kset = kset;
+		if (kobject_init_and_add(kobj, &rpc_sysfs_object_type,
+					 parent, "%s", name) == 0)
+			return kobj;
+		kobject_put(kobj);
+	}
+	return NULL;
+}
+
+static inline struct rpc_xprt *
+rpc_sysfs_xprt_kobj_get_xprt(struct kobject *kobj)
+{
+	struct rpc_sysfs_xprt *x = container_of(kobj,
+		struct rpc_sysfs_xprt, kobject);
+
+	return xprt_get(x->xprt);
+}
+
+static inline struct rpc_xprt_switch *
+rpc_sysfs_xprt_kobj_get_xprt_switch(struct kobject *kobj)
+{
+	struct rpc_sysfs_xprt *x = container_of(kobj,
+		struct rpc_sysfs_xprt, kobject);
+
+	return xprt_switch_get(x->xprt_switch);
+}
+
+static inline struct rpc_xprt_switch *
+rpc_sysfs_xprt_switch_kobj_get_xprt(struct kobject *kobj)
+{
+	struct rpc_sysfs_xprt_switch *x = container_of(kobj,
+		struct rpc_sysfs_xprt_switch, kobject);
+
+	return xprt_switch_get(x->xprt_switch);
+}
+
+static ssize_t rpc_sysfs_xprt_dstaddr_show(struct kobject *kobj,
+					   struct kobj_attribute *attr,
+					   char *buf)
+{
+	struct rpc_xprt *xprt = rpc_sysfs_xprt_kobj_get_xprt(kobj);
+	ssize_t ret;
+
+	if (!xprt)
+		return 0;
+	ret = sprintf(buf, "%s\n", xprt->address_strings[RPC_DISPLAY_ADDR]);
+	xprt_put(xprt);
+	return ret + 1;
+}
+
+static ssize_t rpc_sysfs_xprt_info_show(struct kobject *kobj,
+					struct kobj_attribute *attr,
+					char *buf)
+{
+	struct rpc_xprt *xprt = rpc_sysfs_xprt_kobj_get_xprt(kobj);
+	ssize_t ret;
+
+	if (!xprt)
+		return 0;
+
+	ret = sprintf(buf, "last_used=%lu\ncur_cong=%lu\ncong_win=%lu\n"
+		       "max_num_slots=%u\nmin_num_slots=%u\nnum_reqs=%u\n"
+		       "binding_q_len=%u\nsending_q_len=%u\npending_q_len=%u\n"
+		       "backlog_q_len=%u\nmain_xprt=%d\nsrc_port=%u\n"
+		       "tasks_queuelen=%ld\n",
+		       xprt->last_used, xprt->cong, xprt->cwnd, xprt->max_reqs,
+		       xprt->min_reqs, xprt->num_reqs, xprt->binding.qlen,
+		       xprt->sending.qlen, xprt->pending.qlen,
+		       xprt->backlog.qlen, xprt->main,
+		       (xprt->xprt_class->ident == XPRT_TRANSPORT_TCP) ?
+		       get_srcport(xprt) : 0,
+		       atomic_long_read(&xprt->queuelen));
+	xprt_put(xprt);
+	return ret + 1;
+}
+
+static ssize_t rpc_sysfs_xprt_state_show(struct kobject *kobj,
+					 struct kobj_attribute *attr,
+					 char *buf)
+{
+	struct rpc_xprt *xprt = rpc_sysfs_xprt_kobj_get_xprt(kobj);
+	ssize_t ret;
+	int locked, connected, connecting, close_wait, bound, binding,
+	    closing, congested, cwnd_wait, write_space, offline, remove;
+
+	if (!xprt)
+		return 0;
+
+	if (!xprt->state) {
+		ret = sprintf(buf, "state=CLOSED\n");
+	} else {
+		locked = test_bit(XPRT_LOCKED, &xprt->state);
+		connected = test_bit(XPRT_CONNECTED, &xprt->state);
+		connecting = test_bit(XPRT_CONNECTING, &xprt->state);
+		close_wait = test_bit(XPRT_CLOSE_WAIT, &xprt->state);
+		bound = test_bit(XPRT_BOUND, &xprt->state);
+		binding = test_bit(XPRT_BINDING, &xprt->state);
+		closing = test_bit(XPRT_CLOSING, &xprt->state);
+		congested = test_bit(XPRT_CONGESTED, &xprt->state);
+		cwnd_wait = test_bit(XPRT_CWND_WAIT, &xprt->state);
+		write_space = test_bit(XPRT_WRITE_SPACE, &xprt->state);
+		offline = test_bit(XPRT_OFFLINE, &xprt->state);
+		remove = test_bit(XPRT_REMOVE, &xprt->state);
+
+		ret = sprintf(buf, "state=%s %s %s %s %s %s %s %s %s %s %s %s\n",
+			      locked ? "LOCKED" : "",
+			      connected ? "CONNECTED" : "",
+			      connecting ? "CONNECTING" : "",
+			      close_wait ? "CLOSE_WAIT" : "",
+			      bound ? "BOUND" : "",
+			      binding ? "BOUNDING" : "",
+			      closing ? "CLOSING" : "",
+			      congested ? "CONGESTED" : "",
+			      cwnd_wait ? "CWND_WAIT" : "",
+			      write_space ? "WRITE_SPACE" : "",
+			      offline ? "OFFLINE" : "",
+			      remove ? "REMOVE" : "");
+	}
+
+	xprt_put(xprt);
+	return ret + 1;
+}
+
+static ssize_t rpc_sysfs_xprt_switch_info_show(struct kobject *kobj,
+					       struct kobj_attribute *attr,
+					       char *buf)
+{
+	struct rpc_xprt_switch *xprt_switch =
+		rpc_sysfs_xprt_switch_kobj_get_xprt(kobj);
+	ssize_t ret;
+
+	if (!xprt_switch)
+		return 0;
+	ret = sprintf(buf, "num_xprts=%u\nnum_active=%u\nqueue_len=%ld\n",
+		      xprt_switch->xps_nxprts, xprt_switch->xps_nactive,
+		      atomic_long_read(&xprt_switch->xps_queuelen));
+	xprt_switch_put(xprt_switch);
+	return ret + 1;
+}
+
+static ssize_t rpc_sysfs_xprt_dstaddr_store(struct kobject *kobj,
+					    struct kobj_attribute *attr,
+					    const char *buf, size_t count)
+{
+	struct rpc_xprt *xprt = rpc_sysfs_xprt_kobj_get_xprt(kobj);
+	struct sockaddr *saddr;
+	char *dst_addr;
+	int port;
+	struct xprt_addr *saved_addr;
+	size_t buf_len;
+
+	if (!xprt)
+		return 0;
+	if (!(xprt->xprt_class->ident == XPRT_TRANSPORT_TCP ||
+	      xprt->xprt_class->ident == XPRT_TRANSPORT_RDMA)) {
+		xprt_put(xprt);
+		return -EOPNOTSUPP;
+	}
+
+	if (wait_on_bit_lock(&xprt->state, XPRT_LOCKED, TASK_KILLABLE)) {
+		count = -EINTR;
+		goto out_put;
+	}
+	saddr = (struct sockaddr *)&xprt->addr;
+	port = rpc_get_port(saddr);
+
+	/* buf_len is the len until the first occurence of either
+	 * '\n' or '\0'
+	 */
+	buf_len = strcspn(buf, "\n");
+
+	dst_addr = kstrndup(buf, buf_len, GFP_KERNEL);
+	if (!dst_addr)
+		goto out_err;
+	saved_addr = kzalloc(sizeof(*saved_addr), GFP_KERNEL);
+	if (!saved_addr)
+		goto out_err_free;
+	saved_addr->addr =
+		rcu_dereference_raw(xprt->address_strings[RPC_DISPLAY_ADDR]);
+	rcu_assign_pointer(xprt->address_strings[RPC_DISPLAY_ADDR], dst_addr);
+	call_rcu(&saved_addr->rcu, free_xprt_addr);
+	xprt->addrlen = rpc_pton(xprt->xprt_net, buf, buf_len, saddr,
+				 sizeof(*saddr));
+	rpc_set_port(saddr, port);
+
+	xprt_force_disconnect(xprt);
+out:
+	xprt_release_write(xprt, NULL);
+out_put:
+	xprt_put(xprt);
+	return count;
+out_err_free:
+	kfree(dst_addr);
+out_err:
+	count = -ENOMEM;
+	goto out;
+}
+
+static ssize_t rpc_sysfs_xprt_state_change(struct kobject *kobj,
+					   struct kobj_attribute *attr,
+					   const char *buf, size_t count)
+{
+	struct rpc_xprt *xprt = rpc_sysfs_xprt_kobj_get_xprt(kobj);
+	int offline = 0, online = 0, remove = 0;
+	struct rpc_xprt_switch *xps = rpc_sysfs_xprt_kobj_get_xprt_switch(kobj);
+
+	if (!xprt)
+		return 0;
+
+	if (!strncmp(buf, "offline", 7))
+		offline = 1;
+	else if (!strncmp(buf, "online", 6))
+		online = 1;
+	else if (!strncmp(buf, "remove", 6))
+		remove = 1;
+	else
+		return -EINVAL;
+
+	if (wait_on_bit_lock(&xprt->state, XPRT_LOCKED, TASK_KILLABLE)) {
+		count = -EINTR;
+		goto out_put;
+	}
+	if (xprt->main) {
+		count = -EINVAL;
+		goto release_tasks;
+	}
+	if (offline) {
+		set_bit(XPRT_OFFLINE, &xprt->state);
+		spin_lock(&xps->xps_lock);
+		xps->xps_nactive--;
+		spin_unlock(&xps->xps_lock);
+	} else if (online) {
+		clear_bit(XPRT_OFFLINE, &xprt->state);
+		spin_lock(&xps->xps_lock);
+		xps->xps_nactive++;
+		spin_unlock(&xps->xps_lock);
+	} else if (remove) {
+		if (test_bit(XPRT_OFFLINE, &xprt->state)) {
+			set_bit(XPRT_REMOVE, &xprt->state);
+			xprt_force_disconnect(xprt);
+			if (test_bit(XPRT_CONNECTED, &xprt->state)) {
+				if (!xprt->sending.qlen &&
+				    !xprt->pending.qlen &&
+				    !xprt->backlog.qlen &&
+				    !atomic_long_read(&xprt->queuelen))
+					rpc_xprt_switch_remove_xprt(xps, xprt);
+			}
+		} else {
+			count = -EINVAL;
+		}
+	}
+
+release_tasks:
+	xprt_release_write(xprt, NULL);
+out_put:
+	xprt_put(xprt);
+	xprt_switch_put(xps);
+	return count;
+}
+
+int rpc_sysfs_init(void)
+{
+	rpc_sunrpc_kset = kset_create_and_add("sunrpc", NULL, kernel_kobj);
+	if (!rpc_sunrpc_kset)
+		return -ENOMEM;
+	rpc_sunrpc_client_kobj =
+		rpc_sysfs_object_alloc("rpc-clients", rpc_sunrpc_kset, NULL);
+	if (!rpc_sunrpc_client_kobj)
+		goto err_client;
+	rpc_sunrpc_xprt_switch_kobj =
+		rpc_sysfs_object_alloc("xprt-switches", rpc_sunrpc_kset, NULL);
+	if (!rpc_sunrpc_xprt_switch_kobj)
+		goto err_switch;
+	return 0;
+err_switch:
+	kobject_put(rpc_sunrpc_client_kobj);
+	rpc_sunrpc_client_kobj = NULL;
+err_client:
+	kset_unregister(rpc_sunrpc_kset);
+	rpc_sunrpc_kset = NULL;
+	return -ENOMEM;
+}
+
+static void rpc_sysfs_client_release(struct kobject *kobj)
+{
+	struct rpc_sysfs_client *c;
+
+	c = container_of(kobj, struct rpc_sysfs_client, kobject);
+	kfree(c);
+}
+
+static void rpc_sysfs_xprt_switch_release(struct kobject *kobj)
+{
+	struct rpc_sysfs_xprt_switch *xprt_switch;
+
+	xprt_switch = container_of(kobj, struct rpc_sysfs_xprt_switch, kobject);
+	kfree(xprt_switch);
+}
+
+static void rpc_sysfs_xprt_release(struct kobject *kobj)
+{
+	struct rpc_sysfs_xprt *xprt;
+
+	xprt = container_of(kobj, struct rpc_sysfs_xprt, kobject);
+	kfree(xprt);
+}
+
+static const void *rpc_sysfs_client_namespace(struct kobject *kobj)
+{
+	return container_of(kobj, struct rpc_sysfs_client, kobject)->net;
+}
+
+static const void *rpc_sysfs_xprt_switch_namespace(struct kobject *kobj)
+{
+	return container_of(kobj, struct rpc_sysfs_xprt_switch, kobject)->net;
+}
+
+static const void *rpc_sysfs_xprt_namespace(struct kobject *kobj)
+{
+	return container_of(kobj, struct rpc_sysfs_xprt,
+			    kobject)->xprt->xprt_net;
+}
+
+static struct kobj_attribute rpc_sysfs_xprt_dstaddr = __ATTR(dstaddr,
+	0644, rpc_sysfs_xprt_dstaddr_show, rpc_sysfs_xprt_dstaddr_store);
+
+static struct kobj_attribute rpc_sysfs_xprt_info = __ATTR(xprt_info,
+	0444, rpc_sysfs_xprt_info_show, NULL);
+
+static struct kobj_attribute rpc_sysfs_xprt_change_state = __ATTR(xprt_state,
+	0644, rpc_sysfs_xprt_state_show, rpc_sysfs_xprt_state_change);
+
+static struct attribute *rpc_sysfs_xprt_attrs[] = {
+	&rpc_sysfs_xprt_dstaddr.attr,
+	&rpc_sysfs_xprt_info.attr,
+	&rpc_sysfs_xprt_change_state.attr,
+	NULL,
+};
+
+static struct kobj_attribute rpc_sysfs_xprt_switch_info =
+	__ATTR(xprt_switch_info, 0444, rpc_sysfs_xprt_switch_info_show, NULL);
+
+static struct attribute *rpc_sysfs_xprt_switch_attrs[] = {
+	&rpc_sysfs_xprt_switch_info.attr,
+	NULL,
+};
+
+static struct kobj_type rpc_sysfs_client_type = {
+	.release = rpc_sysfs_client_release,
+	.sysfs_ops = &kobj_sysfs_ops,
+	.namespace = rpc_sysfs_client_namespace,
+};
+
+static struct kobj_type rpc_sysfs_xprt_switch_type = {
+	.release = rpc_sysfs_xprt_switch_release,
+	.default_attrs = rpc_sysfs_xprt_switch_attrs,
+	.sysfs_ops = &kobj_sysfs_ops,
+	.namespace = rpc_sysfs_xprt_switch_namespace,
+};
+
+static struct kobj_type rpc_sysfs_xprt_type = {
+	.release = rpc_sysfs_xprt_release,
+	.default_attrs = rpc_sysfs_xprt_attrs,
+	.sysfs_ops = &kobj_sysfs_ops,
+	.namespace = rpc_sysfs_xprt_namespace,
+};
+
+void rpc_sysfs_exit(void)
+{
+	kobject_put(rpc_sunrpc_client_kobj);
+	kobject_put(rpc_sunrpc_xprt_switch_kobj);
+	kset_unregister(rpc_sunrpc_kset);
+}
+
+static struct rpc_sysfs_client *rpc_sysfs_client_alloc(struct kobject *parent,
+						       struct net *net,
+						       int clid)
+{
+	struct rpc_sysfs_client *p;
+
+	p = kzalloc(sizeof(*p), GFP_KERNEL);
+	if (p) {
+		p->net = net;
+		p->kobject.kset = rpc_sunrpc_kset;
+		if (kobject_init_and_add(&p->kobject, &rpc_sysfs_client_type,
+					 parent, "clnt-%d", clid) == 0)
+			return p;
+		kobject_put(&p->kobject);
+	}
+	return NULL;
+}
+
+static struct rpc_sysfs_xprt_switch *
+rpc_sysfs_xprt_switch_alloc(struct kobject *parent,
+			    struct rpc_xprt_switch *xprt_switch,
+			    struct net *net,
+			    gfp_t gfp_flags)
+{
+	struct rpc_sysfs_xprt_switch *p;
+
+	p = kzalloc(sizeof(*p), gfp_flags);
+	if (p) {
+		p->net = net;
+		p->kobject.kset = rpc_sunrpc_kset;
+		if (kobject_init_and_add(&p->kobject,
+					 &rpc_sysfs_xprt_switch_type,
+					 parent, "switch-%d",
+					 xprt_switch->xps_id) == 0)
+			return p;
+		kobject_put(&p->kobject);
+	}
+	return NULL;
+}
+
+static struct rpc_sysfs_xprt *rpc_sysfs_xprt_alloc(struct kobject *parent,
+						   struct rpc_xprt *xprt,
+						   gfp_t gfp_flags)
+{
+	struct rpc_sysfs_xprt *p;
+
+	p = kzalloc(sizeof(*p), gfp_flags);
+	if (!p)
+		goto out;
+	p->kobject.kset = rpc_sunrpc_kset;
+	if (kobject_init_and_add(&p->kobject, &rpc_sysfs_xprt_type,
+				 parent, "xprt-%d-%s", xprt->id,
+				 xprt->address_strings[RPC_DISPLAY_PROTO]) == 0)
+		return p;
+	kobject_put(&p->kobject);
+out:
+	return NULL;
+}
+
+void rpc_sysfs_client_setup(struct rpc_clnt *clnt,
+			    struct rpc_xprt_switch *xprt_switch,
+			    struct net *net)
+{
+	struct rpc_sysfs_client *rpc_client;
+
+	rpc_client = rpc_sysfs_client_alloc(rpc_sunrpc_client_kobj,
+					    net, clnt->cl_clid);
+	if (rpc_client) {
+		char name[] = "switch";
+		struct rpc_sysfs_xprt_switch *xswitch =
+			(struct rpc_sysfs_xprt_switch *)xprt_switch->xps_sysfs;
+		int ret;
+
+		clnt->cl_sysfs = rpc_client;
+		rpc_client->clnt = clnt;
+		rpc_client->xprt_switch = xprt_switch;
+		kobject_uevent(&rpc_client->kobject, KOBJ_ADD);
+		ret = sysfs_create_link_nowarn(&rpc_client->kobject,
+					       &xswitch->kobject, name);
+		if (ret)
+			pr_warn("can't create link to %s in sysfs (%d)\n",
+				name, ret);
+	}
+}
+
+void rpc_sysfs_xprt_switch_setup(struct rpc_xprt_switch *xprt_switch,
+				 struct rpc_xprt *xprt,
+				 gfp_t gfp_flags)
+{
+	struct rpc_sysfs_xprt_switch *rpc_xprt_switch;
+	struct net *net;
+
+	if (xprt_switch->xps_net)
+		net = xprt_switch->xps_net;
+	else
+		net = xprt->xprt_net;
+	rpc_xprt_switch =
+		rpc_sysfs_xprt_switch_alloc(rpc_sunrpc_xprt_switch_kobj,
+					    xprt_switch, net, gfp_flags);
+	if (rpc_xprt_switch) {
+		xprt_switch->xps_sysfs = rpc_xprt_switch;
+		rpc_xprt_switch->xprt_switch = xprt_switch;
+		rpc_xprt_switch->xprt = xprt;
+		kobject_uevent(&rpc_xprt_switch->kobject, KOBJ_ADD);
+	}
+}
+
+void rpc_sysfs_xprt_setup(struct rpc_xprt_switch *xprt_switch,
+			  struct rpc_xprt *xprt,
+			  gfp_t gfp_flags)
+{
+	struct rpc_sysfs_xprt *rpc_xprt;
+	struct rpc_sysfs_xprt_switch *switch_obj =
+		(struct rpc_sysfs_xprt_switch *)xprt_switch->xps_sysfs;
+
+	rpc_xprt = rpc_sysfs_xprt_alloc(&switch_obj->kobject, xprt, gfp_flags);
+	if (rpc_xprt) {
+		xprt->xprt_sysfs = rpc_xprt;
+		rpc_xprt->xprt = xprt;
+		rpc_xprt->xprt_switch = xprt_switch;
+		kobject_uevent(&rpc_xprt->kobject, KOBJ_ADD);
+	}
+}
+
+void rpc_sysfs_client_destroy(struct rpc_clnt *clnt)
+{
+	struct rpc_sysfs_client *rpc_client = clnt->cl_sysfs;
+
+	if (rpc_client) {
+		char name[] = "switch";
+
+		sysfs_remove_link(&rpc_client->kobject, name);
+		kobject_uevent(&rpc_client->kobject, KOBJ_REMOVE);
+		kobject_del(&rpc_client->kobject);
+		kobject_put(&rpc_client->kobject);
+		clnt->cl_sysfs = NULL;
+	}
+}
+
+void rpc_sysfs_xprt_switch_destroy(struct rpc_xprt_switch *xprt_switch)
+{
+	struct rpc_sysfs_xprt_switch *rpc_xprt_switch = xprt_switch->xps_sysfs;
+
+	if (rpc_xprt_switch) {
+		kobject_uevent(&rpc_xprt_switch->kobject, KOBJ_REMOVE);
+		kobject_del(&rpc_xprt_switch->kobject);
+		kobject_put(&rpc_xprt_switch->kobject);
+		xprt_switch->xps_sysfs = NULL;
+	}
+}
+
+void rpc_sysfs_xprt_destroy(struct rpc_xprt *xprt)
+{
+	struct rpc_sysfs_xprt *rpc_xprt = xprt->xprt_sysfs;
+
+	if (rpc_xprt) {
+		kobject_uevent(&rpc_xprt->kobject, KOBJ_REMOVE);
+		kobject_del(&rpc_xprt->kobject);
+		kobject_put(&rpc_xprt->kobject);
+		xprt->xprt_sysfs = NULL;
+	}
+}
diff --git a/net/sunrpc/sysfs.h b/net/sunrpc/sysfs.h
new file mode 100644
index 000000000000..6620cebd1037
--- /dev/null
+++ b/net/sunrpc/sysfs.h
@@ -0,0 +1,42 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (c) 2020 Anna Schumaker <Anna.Schumaker@Netapp.com>
+ */
+#ifndef __SUNRPC_SYSFS_H
+#define __SUNRPC_SYSFS_H
+
+struct rpc_sysfs_client {
+	struct kobject kobject;
+	struct net *net;
+	struct rpc_clnt *clnt;
+	struct rpc_xprt_switch *xprt_switch;
+};
+
+struct rpc_sysfs_xprt_switch {
+	struct kobject kobject;
+	struct net *net;
+	struct rpc_xprt_switch *xprt_switch;
+	struct rpc_xprt *xprt;
+};
+
+struct rpc_sysfs_xprt {
+	struct kobject kobject;
+	struct rpc_xprt *xprt;
+	struct rpc_xprt_switch *xprt_switch;
+};
+
+int rpc_sysfs_init(void);
+void rpc_sysfs_exit(void);
+
+void rpc_sysfs_client_setup(struct rpc_clnt *clnt,
+			    struct rpc_xprt_switch *xprt_switch,
+			    struct net *net);
+void rpc_sysfs_client_destroy(struct rpc_clnt *clnt);
+void rpc_sysfs_xprt_switch_setup(struct rpc_xprt_switch *xprt_switch,
+				 struct rpc_xprt *xprt, gfp_t gfp_flags);
+void rpc_sysfs_xprt_switch_destroy(struct rpc_xprt_switch *xprt);
+void rpc_sysfs_xprt_setup(struct rpc_xprt_switch *xprt_switch,
+			  struct rpc_xprt *xprt, gfp_t gfp_flags);
+void rpc_sysfs_xprt_destroy(struct rpc_xprt *xprt);
+
+#endif
diff --git a/net/sunrpc/xdr.c b/net/sunrpc/xdr.c
index 3964ff74ee51..ca10ba2626f2 100644
--- a/net/sunrpc/xdr.c
+++ b/net/sunrpc/xdr.c
@@ -1230,10 +1230,9 @@ static unsigned int xdr_set_page_base(struct xdr_stream *xdr,
 	void *kaddr;
 
 	maxlen = xdr->buf->page_len;
-	if (base >= maxlen) {
-		base = maxlen;
-		maxlen = 0;
-	} else
+	if (base >= maxlen)
+		return 0;
+	else
 		maxlen -= base;
 	if (len > maxlen)
 		len = maxlen;
diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
index 3509a7f139b9..fb6db09725c7 100644
--- a/net/sunrpc/xprt.c
+++ b/net/sunrpc/xprt.c
@@ -55,6 +55,7 @@
 #include <trace/events/sunrpc.h>
 
 #include "sunrpc.h"
+#include "sysfs.h"
 
 /*
  * Local variables
@@ -443,7 +444,7 @@ void xprt_release_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task)
 }
 EXPORT_SYMBOL_GPL(xprt_release_xprt_cong);
 
-static inline void xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
+void xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
 {
 	if (xprt->snd_task != task)
 		return;
@@ -1746,6 +1747,30 @@ static void xprt_free_all_slots(struct rpc_xprt *xprt)
 	}
 }
 
+static DEFINE_IDA(rpc_xprt_ids);
+
+void xprt_cleanup_ids(void)
+{
+	ida_destroy(&rpc_xprt_ids);
+}
+
+static int xprt_alloc_id(struct rpc_xprt *xprt)
+{
+	int id;
+
+	id = ida_simple_get(&rpc_xprt_ids, 0, 0, GFP_KERNEL);
+	if (id < 0)
+		return id;
+
+	xprt->id = id;
+	return 0;
+}
+
+static void xprt_free_id(struct rpc_xprt *xprt)
+{
+	ida_simple_remove(&rpc_xprt_ids, xprt->id);
+}
+
 struct rpc_xprt *xprt_alloc(struct net *net, size_t size,
 		unsigned int num_prealloc,
 		unsigned int max_alloc)
@@ -1758,6 +1783,7 @@ struct rpc_xprt *xprt_alloc(struct net *net, size_t size,
 	if (xprt == NULL)
 		goto out;
 
+	xprt_alloc_id(xprt);
 	xprt_init(xprt, net);
 
 	for (i = 0; i < num_prealloc; i++) {
@@ -1786,6 +1812,8 @@ void xprt_free(struct rpc_xprt *xprt)
 {
 	put_net(xprt->xprt_net);
 	xprt_free_all_slots(xprt);
+	xprt_free_id(xprt);
+	rpc_sysfs_xprt_destroy(xprt);
 	kfree_rcu(xprt, rcu);
 }
 EXPORT_SYMBOL_GPL(xprt_free);
diff --git a/net/sunrpc/xprtmultipath.c b/net/sunrpc/xprtmultipath.c
index 1b4073131c6f..c60820e45082 100644
--- a/net/sunrpc/xprtmultipath.c
+++ b/net/sunrpc/xprtmultipath.c
@@ -19,6 +19,8 @@
 #include <linux/sunrpc/addr.h>
 #include <linux/sunrpc/xprtmultipath.h>
 
+#include "sysfs.h"
+
 typedef struct rpc_xprt *(*xprt_switch_find_xprt_t)(struct rpc_xprt_switch *xps,
 		const struct rpc_xprt *cur);
 
@@ -55,6 +57,7 @@ void rpc_xprt_switch_add_xprt(struct rpc_xprt_switch *xps,
 	if (xps->xps_net == xprt->xprt_net || xps->xps_net == NULL)
 		xprt_switch_add_xprt_locked(xps, xprt);
 	spin_unlock(&xps->xps_lock);
+	rpc_sysfs_xprt_setup(xps, xprt, GFP_KERNEL);
 }
 
 static void xprt_switch_remove_xprt_locked(struct rpc_xprt_switch *xps,
@@ -62,7 +65,8 @@ static void xprt_switch_remove_xprt_locked(struct rpc_xprt_switch *xps,
 {
 	if (unlikely(xprt == NULL))
 		return;
-	xps->xps_nactive--;
+	if (!test_bit(XPRT_OFFLINE, &xprt->state))
+		xps->xps_nactive--;
 	xps->xps_nxprts--;
 	if (xps->xps_nxprts == 0)
 		xps->xps_net = NULL;
@@ -86,6 +90,30 @@ void rpc_xprt_switch_remove_xprt(struct rpc_xprt_switch *xps,
 	xprt_put(xprt);
 }
 
+static DEFINE_IDA(rpc_xprtswitch_ids);
+
+void xprt_multipath_cleanup_ids(void)
+{
+	ida_destroy(&rpc_xprtswitch_ids);
+}
+
+static int xprt_switch_alloc_id(struct rpc_xprt_switch *xps, gfp_t gfp_flags)
+{
+	int id;
+
+	id = ida_simple_get(&rpc_xprtswitch_ids, 0, 0, gfp_flags);
+	if (id < 0)
+		return id;
+
+	xps->xps_id = id;
+	return 0;
+}
+
+static void xprt_switch_free_id(struct rpc_xprt_switch *xps)
+{
+	ida_simple_remove(&rpc_xprtswitch_ids, xps->xps_id);
+}
+
 /**
  * xprt_switch_alloc - Allocate a new struct rpc_xprt_switch
  * @xprt: pointer to struct rpc_xprt
@@ -103,12 +131,15 @@ struct rpc_xprt_switch *xprt_switch_alloc(struct rpc_xprt *xprt,
 	if (xps != NULL) {
 		spin_lock_init(&xps->xps_lock);
 		kref_init(&xps->xps_kref);
+		xprt_switch_alloc_id(xps, gfp_flags);
 		xps->xps_nxprts = xps->xps_nactive = 0;
 		atomic_long_set(&xps->xps_queuelen, 0);
 		xps->xps_net = NULL;
 		INIT_LIST_HEAD(&xps->xps_xprt_list);
 		xps->xps_iter_ops = &rpc_xprt_iter_singular;
+		rpc_sysfs_xprt_switch_setup(xps, xprt, gfp_flags);
 		xprt_switch_add_xprt_locked(xps, xprt);
+		rpc_sysfs_xprt_setup(xps, xprt, gfp_flags);
 	}
 
 	return xps;
@@ -136,6 +167,8 @@ static void xprt_switch_free(struct kref *kref)
 			struct rpc_xprt_switch, xps_kref);
 
 	xprt_switch_free_entries(xps);
+	rpc_sysfs_xprt_switch_destroy(xps);
+	xprt_switch_free_id(xps);
 	kfree_rcu(xps, xps_rcu);
 }
 
@@ -198,7 +231,8 @@ void xprt_iter_default_rewind(struct rpc_xprt_iter *xpi)
 static
 bool xprt_is_active(const struct rpc_xprt *xprt)
 {
-	return kref_read(&xprt->kref) != 0;
+	return (kref_read(&xprt->kref) != 0 &&
+		!test_bit(XPRT_OFFLINE, &xprt->state));
 }
 
 static
diff --git a/net/sunrpc/xprtrdma/transport.c b/net/sunrpc/xprtrdma/transport.c
index 19a49d26b1e4..9c2ffc67c0fd 100644
--- a/net/sunrpc/xprtrdma/transport.c
+++ b/net/sunrpc/xprtrdma/transport.c
@@ -73,6 +73,7 @@ unsigned int xprt_rdma_max_inline_read = RPCRDMA_DEF_INLINE;
 unsigned int xprt_rdma_max_inline_write = RPCRDMA_DEF_INLINE;
 unsigned int xprt_rdma_memreg_strategy		= RPCRDMA_FRWR;
 int xprt_rdma_pad_optimize;
+static struct xprt_class xprt_rdma;
 
 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
 
@@ -349,6 +350,7 @@ xprt_setup_rdma(struct xprt_create *args)
 	/* Ensure xprt->addr holds valid server TCP (not RDMA)
 	 * address, for any side protocols which peek at it */
 	xprt->prot = IPPROTO_TCP;
+	xprt->xprt_class = &xprt_rdma;
 	xprt->addrlen = args->addrlen;
 	memcpy(&xprt->addr, sap, xprt->addrlen);
 
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 316d04945587..e573dcecdd66 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -91,6 +91,11 @@ static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT;
 
 static struct ctl_table_header *sunrpc_table_header;
 
+static struct xprt_class xs_local_transport;
+static struct xprt_class xs_udp_transport;
+static struct xprt_class xs_tcp_transport;
+static struct xprt_class xs_bc_tcp_transport;
+
 /*
  * FIXME: changing the UDP slot table size should also resize the UDP
  *        socket buffers for existing UDP transports
@@ -1648,6 +1653,13 @@ static int xs_get_srcport(struct sock_xprt *transport)
 	return port;
 }
 
+unsigned short get_srcport(struct rpc_xprt *xprt)
+{
+	struct sock_xprt *sock = container_of(xprt, struct sock_xprt, xprt);
+	return sock->srcport;
+}
+EXPORT_SYMBOL(get_srcport);
+
 static unsigned short xs_next_srcport(struct sock_xprt *transport, unsigned short port)
 {
 	if (transport->srcport != 0)
@@ -1689,7 +1701,8 @@ static int xs_bind(struct sock_xprt *transport, struct socket *sock)
 		err = kernel_bind(sock, (struct sockaddr *)&myaddr,
 				transport->xprt.addrlen);
 		if (err == 0) {
-			transport->srcport = port;
+			if (transport->xprt.reuseport)
+				transport->srcport = port;
 			break;
 		}
 		last = port;
@@ -2779,6 +2792,7 @@ static struct rpc_xprt *xs_setup_local(struct xprt_create *args)
 	transport = container_of(xprt, struct sock_xprt, xprt);
 
 	xprt->prot = 0;
+	xprt->xprt_class = &xs_local_transport;
 	xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
 
 	xprt->bind_timeout = XS_BIND_TO;
@@ -2848,6 +2862,7 @@ static struct rpc_xprt *xs_setup_udp(struct xprt_create *args)
 	transport = container_of(xprt, struct sock_xprt, xprt);
 
 	xprt->prot = IPPROTO_UDP;
+	xprt->xprt_class = &xs_udp_transport;
 	/* XXX: header size can vary due to auth type, IPv6, etc. */
 	xprt->max_payload = (1U << 16) - (MAX_HEADER << 3);
 
@@ -2928,6 +2943,7 @@ static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args)
 	transport = container_of(xprt, struct sock_xprt, xprt);
 
 	xprt->prot = IPPROTO_TCP;
+	xprt->xprt_class = &xs_tcp_transport;
 	xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
 
 	xprt->bind_timeout = XS_BIND_TO;
@@ -3001,6 +3017,7 @@ static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args)
 	transport = container_of(xprt, struct sock_xprt, xprt);
 
 	xprt->prot = IPPROTO_TCP;
+	xprt->xprt_class = &xs_bc_tcp_transport;
 	xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
 	xprt->timeout = &xs_tcp_default_timeout;