Merge tag 'nfsd-6.7' of git://git.kernel.org/pub/scm/linux/kernel/git/cel/linux

Pull nfsd updates from Chuck Lever:
 "This release completes the SunRPC thread scheduler work that was begun
  in v6.6. The scheduler can now find an svc thread to wake in constant
  time and without a list walk. Thanks again to Neil Brown for this
  overhaul.

  Lorenzo Bianconi contributed infrastructure for a netlink-based NFSD
  control plane. The long-term plan is to provide the same functionality
  as found in /proc/fs/nfsd, plus some interesting additions, and then
  migrate the NFSD user space utilities to netlink.

  A long series to overhaul NFSD's NFSv4 operation encoding was applied
  in this release. The goals are to bring this family of encoding
  functions in line with the matching NFSv4 decoding functions and with
  the NFSv2 and NFSv3 XDR functions, preparing the way for better memory
  safety and maintainability.

  A further improvement to NFSD's write delegation support was
  contributed by Dai Ngo. This adds a CB_GETATTR callback, enabling the
  server to retrieve cached size and mtime data from clients holding
  write delegations. If the server can retrieve this information, it
  does not have to recall the delegation in some cases.

  The usual panoply of bug fixes and minor improvements round out this
  release. As always I am grateful to all contributors, reviewers, and
  testers"

* tag 'nfsd-6.7' of git://git.kernel.org/pub/scm/linux/kernel/git/cel/linux: (127 commits)
  svcrdma: Fix tracepoint printk format
  svcrdma: Drop connection after an RDMA Read error
  NFSD: clean up alloc_init_deleg()
  NFSD: Fix frame size warning in svc_export_parse()
  NFSD: Rewrite synopsis of nfsd_percpu_counters_init()
  nfsd: Clean up errors in nfs3proc.c
  nfsd: Clean up errors in nfs4state.c
  NFSD: Clean up errors in stats.c
  NFSD: simplify error paths in nfsd_svc()
  NFSD: Clean up nfsd4_encode_seek()
  NFSD: Clean up nfsd4_encode_offset_status()
  NFSD: Clean up nfsd4_encode_copy_notify()
  NFSD: Clean up nfsd4_encode_copy()
  NFSD: Clean up nfsd4_encode_test_stateid()
  NFSD: Clean up nfsd4_encode_exchange_id()
  NFSD: Clean up nfsd4_do_encode_secinfo()
  NFSD: Clean up nfsd4_encode_access()
  NFSD: Clean up nfsd4_encode_readdir()
  NFSD: Clean up nfsd4_encode_entry4()
  NFSD: Add an nfsd4_encode_nfs_cookie4() helper
  ...

5 files changed, 193 insertions, 220 deletions

diff --git a/net/sunrpc/backchannel_rqst.c b/net/sunrpc/backchannel_rqst.c
index 65a6c6429a53..caa94cf57123 100644
--- a/net/sunrpc/backchannel_rqst.c
+++ b/net/sunrpc/backchannel_rqst.c
@@ -83,7 +83,6 @@ static struct rpc_rqst *xprt_alloc_bc_req(struct rpc_xprt *xprt)
 		return NULL;
 
 	req->rq_xprt = xprt;
-	INIT_LIST_HEAD(&req->rq_bc_list);
 
 	/* Preallocate one XDR receive buffer */
 	if (xprt_alloc_xdr_buf(&req->rq_rcv_buf, gfp_flags) < 0) {
@@ -349,10 +348,8 @@ found:
 }
 
 /*
- * Add callback request to callback list.  The callback
- * service sleeps on the sv_cb_waitq waiting for new
- * requests.  Wake it up after adding enqueing the
- * request.
+ * Add callback request to callback list.  Wake a thread
+ * on the first pool (usually the only pool) to handle it.
  */
 void xprt_complete_bc_request(struct rpc_rqst *req, uint32_t copied)
 {
@@ -369,8 +366,6 @@ void xprt_complete_bc_request(struct rpc_rqst *req, uint32_t copied)
 
 	dprintk("RPC:       add callback request to list\n");
 	xprt_get(xprt);
-	spin_lock(&bc_serv->sv_cb_lock);
-	list_add(&req->rq_bc_list, &bc_serv->sv_cb_list);
-	wake_up(&bc_serv->sv_cb_waitq);
-	spin_unlock(&bc_serv->sv_cb_lock);
+	lwq_enqueue(&req->rq_bc_list, &bc_serv->sv_cb_list);
+	svc_pool_wake_idle_thread(&bc_serv->sv_pools[0]);
 }
diff --git a/net/sunrpc/svc.c b/net/sunrpc/svc.c
index 812fda9d45dd..3f2ea7a0496f 100644
--- a/net/sunrpc/svc.c
+++ b/net/sunrpc/svc.c
@@ -438,9 +438,7 @@ EXPORT_SYMBOL_GPL(svc_bind);
 static void
 __svc_init_bc(struct svc_serv *serv)
 {
-	INIT_LIST_HEAD(&serv->sv_cb_list);
-	spin_lock_init(&serv->sv_cb_lock);
-	init_waitqueue_head(&serv->sv_cb_waitq);
+	lwq_init(&serv->sv_cb_list);
 }
 #else
 static void
@@ -509,9 +507,9 @@ __svc_create(struct svc_program *prog, unsigned int bufsize, int npools,
 				i, serv->sv_name);
 
 		pool->sp_id = i;
-		INIT_LIST_HEAD(&pool->sp_sockets);
+		lwq_init(&pool->sp_xprts);
 		INIT_LIST_HEAD(&pool->sp_all_threads);
-		spin_lock_init(&pool->sp_lock);
+		init_llist_head(&pool->sp_idle_threads);
 
 		percpu_counter_init(&pool->sp_messages_arrived, 0, GFP_KERNEL);
 		percpu_counter_init(&pool->sp_sockets_queued, 0, GFP_KERNEL);
@@ -575,11 +573,12 @@ svc_destroy(struct kref *ref)
 	timer_shutdown_sync(&serv->sv_temptimer);
 
 	/*
-	 * The last user is gone and thus all sockets have to be destroyed to
-	 * the point. Check this.
+	 * Remaining transports at this point are not expected.
 	 */
-	BUG_ON(!list_empty(&serv->sv_permsocks));
-	BUG_ON(!list_empty(&serv->sv_tempsocks));
+	WARN_ONCE(!list_empty(&serv->sv_permsocks),
+		  "SVC: permsocks remain for %s\n", serv->sv_program->pg_name);
+	WARN_ONCE(!list_empty(&serv->sv_tempsocks),
+		  "SVC: tempsocks remain for %s\n", serv->sv_program->pg_name);
 
 	cache_clean_deferred(serv);
 
@@ -642,7 +641,6 @@ svc_rqst_alloc(struct svc_serv *serv, struct svc_pool *pool, int node)
 
 	folio_batch_init(&rqstp->rq_fbatch);
 
-	__set_bit(RQ_BUSY, &rqstp->rq_flags);
 	rqstp->rq_server = serv;
 	rqstp->rq_pool = pool;
 
@@ -682,10 +680,13 @@ svc_prepare_thread(struct svc_serv *serv, struct svc_pool *pool, int node)
 	serv->sv_nrthreads += 1;
 	spin_unlock_bh(&serv->sv_lock);
 
-	spin_lock_bh(&pool->sp_lock);
-	pool->sp_nrthreads++;
+	atomic_inc(&pool->sp_nrthreads);
+
+	/* Protected by whatever lock the service uses when calling
+	 * svc_set_num_threads()
+	 */
 	list_add_rcu(&rqstp->rq_all, &pool->sp_all_threads);
-	spin_unlock_bh(&pool->sp_lock);
+
 	return rqstp;
 }
 
@@ -701,23 +702,25 @@ svc_prepare_thread(struct svc_serv *serv, struct svc_pool *pool, int node)
 void svc_pool_wake_idle_thread(struct svc_pool *pool)
 {
 	struct svc_rqst	*rqstp;
+	struct llist_node *ln;
 
 	rcu_read_lock();
-	list_for_each_entry_rcu(rqstp, &pool->sp_all_threads, rq_all) {
-		if (test_and_set_bit(RQ_BUSY, &rqstp->rq_flags))
-			continue;
-
+	ln = READ_ONCE(pool->sp_idle_threads.first);
+	if (ln) {
+		rqstp = llist_entry(ln, struct svc_rqst, rq_idle);
 		WRITE_ONCE(rqstp->rq_qtime, ktime_get());
-		wake_up_process(rqstp->rq_task);
+		if (!task_is_running(rqstp->rq_task)) {
+			wake_up_process(rqstp->rq_task);
+			trace_svc_wake_up(rqstp->rq_task->pid);
+			percpu_counter_inc(&pool->sp_threads_woken);
+		}
 		rcu_read_unlock();
-		percpu_counter_inc(&pool->sp_threads_woken);
-		trace_svc_wake_up(rqstp->rq_task->pid);
 		return;
 	}
 	rcu_read_unlock();
 
-	set_bit(SP_CONGESTED, &pool->sp_flags);
 }
+EXPORT_SYMBOL_GPL(svc_pool_wake_idle_thread);
 
 static struct svc_pool *
 svc_pool_next(struct svc_serv *serv, struct svc_pool *pool, unsigned int *state)
@@ -725,36 +728,38 @@ svc_pool_next(struct svc_serv *serv, struct svc_pool *pool, unsigned int *state)
 	return pool ? pool : &serv->sv_pools[(*state)++ % serv->sv_nrpools];
 }
 
-static struct task_struct *
-svc_pool_victim(struct svc_serv *serv, struct svc_pool *pool, unsigned int *state)
+static struct svc_pool *
+svc_pool_victim(struct svc_serv *serv, struct svc_pool *target_pool,
+		unsigned int *state)
 {
+	struct svc_pool *pool;
 	unsigned int i;
-	struct task_struct *task = NULL;
+
+retry:
+	pool = target_pool;
 
 	if (pool != NULL) {
-		spin_lock_bh(&pool->sp_lock);
+		if (atomic_inc_not_zero(&pool->sp_nrthreads))
+			goto found_pool;
+		return NULL;
 	} else {
 		for (i = 0; i < serv->sv_nrpools; i++) {
 			pool = &serv->sv_pools[--(*state) % serv->sv_nrpools];
-			spin_lock_bh(&pool->sp_lock);
-			if (!list_empty(&pool->sp_all_threads))
+			if (atomic_inc_not_zero(&pool->sp_nrthreads))
 				goto found_pool;
-			spin_unlock_bh(&pool->sp_lock);
 		}
 		return NULL;
 	}
 
 found_pool:
-	if (!list_empty(&pool->sp_all_threads)) {
-		struct svc_rqst *rqstp;
-
-		rqstp = list_entry(pool->sp_all_threads.next, struct svc_rqst, rq_all);
-		set_bit(RQ_VICTIM, &rqstp->rq_flags);
-		list_del_rcu(&rqstp->rq_all);
-		task = rqstp->rq_task;
-	}
-	spin_unlock_bh(&pool->sp_lock);
-	return task;
+	set_bit(SP_VICTIM_REMAINS, &pool->sp_flags);
+	set_bit(SP_NEED_VICTIM, &pool->sp_flags);
+	if (!atomic_dec_and_test(&pool->sp_nrthreads))
+		return pool;
+	/* Nothing left in this pool any more */
+	clear_bit(SP_NEED_VICTIM, &pool->sp_flags);
+	clear_bit(SP_VICTIM_REMAINS, &pool->sp_flags);
+	goto retry;
 }
 
 static int
@@ -795,18 +800,16 @@ svc_start_kthreads(struct svc_serv *serv, struct svc_pool *pool, int nrservs)
 static int
 svc_stop_kthreads(struct svc_serv *serv, struct svc_pool *pool, int nrservs)
 {
-	struct svc_rqst	*rqstp;
-	struct task_struct *task;
 	unsigned int state = serv->sv_nrthreads-1;
+	struct svc_pool *victim;
 
 	do {
-		task = svc_pool_victim(serv, pool, &state);
-		if (task == NULL)
+		victim = svc_pool_victim(serv, pool, &state);
+		if (!victim)
 			break;
-		rqstp = kthread_data(task);
-		/* Did we lose a race to svo_function threadfn? */
-		if (kthread_stop(task) == -EINTR)
-			svc_exit_thread(rqstp);
+		svc_pool_wake_idle_thread(victim);
+		wait_on_bit(&victim->sp_flags, SP_VICTIM_REMAINS,
+			    TASK_IDLE);
 		nrservs++;
 	} while (nrservs < 0);
 	return 0;
@@ -832,13 +835,10 @@ svc_stop_kthreads(struct svc_serv *serv, struct svc_pool *pool, int nrservs)
 int
 svc_set_num_threads(struct svc_serv *serv, struct svc_pool *pool, int nrservs)
 {
-	if (pool == NULL) {
+	if (!pool)
 		nrservs -= serv->sv_nrthreads;
-	} else {
-		spin_lock_bh(&pool->sp_lock);
-		nrservs -= pool->sp_nrthreads;
-		spin_unlock_bh(&pool->sp_lock);
-	}
+	else
+		nrservs -= atomic_read(&pool->sp_nrthreads);
 
 	if (nrservs > 0)
 		return svc_start_kthreads(serv, pool, nrservs);
@@ -924,11 +924,9 @@ svc_exit_thread(struct svc_rqst *rqstp)
 	struct svc_serv	*serv = rqstp->rq_server;
 	struct svc_pool	*pool = rqstp->rq_pool;
 
-	spin_lock_bh(&pool->sp_lock);
-	pool->sp_nrthreads--;
-	if (!test_and_set_bit(RQ_VICTIM, &rqstp->rq_flags))
-		list_del_rcu(&rqstp->rq_all);
-	spin_unlock_bh(&pool->sp_lock);
+	list_del_rcu(&rqstp->rq_all);
+
+	atomic_dec(&pool->sp_nrthreads);
 
 	spin_lock_bh(&serv->sv_lock);
 	serv->sv_nrthreads -= 1;
@@ -938,6 +936,11 @@ svc_exit_thread(struct svc_rqst *rqstp)
 	svc_rqst_free(rqstp);
 
 	svc_put(serv);
+	/* That svc_put() cannot be the last, because the thread
+	 * waiting for SP_VICTIM_REMAINS to clear must hold
+	 * a reference. So it is still safe to access pool.
+	 */
+	clear_and_wake_up_bit(SP_VICTIM_REMAINS, &pool->sp_flags);
 }
 EXPORT_SYMBOL_GPL(svc_exit_thread);
 
@@ -1544,24 +1547,20 @@ out_drop:
 }
 
 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
-/*
- * Process a backchannel RPC request that arrived over an existing
- * outbound connection
+/**
+ * svc_process_bc - process a reverse-direction RPC request
+ * @req: RPC request to be used for client-side processing
+ * @rqstp: server-side execution context
+ *
  */
-int
-bc_svc_process(struct svc_serv *serv, struct rpc_rqst *req,
-	       struct svc_rqst *rqstp)
+void svc_process_bc(struct rpc_rqst *req, struct svc_rqst *rqstp)
 {
 	struct rpc_task *task;
 	int proc_error;
-	int error;
-
-	dprintk("svc: %s(%p)\n", __func__, req);
 
 	/* Build the svc_rqst used by the common processing routine */
 	rqstp->rq_xid = req->rq_xid;
 	rqstp->rq_prot = req->rq_xprt->prot;
-	rqstp->rq_server = serv;
 	rqstp->rq_bc_net = req->rq_xprt->xprt_net;
 
 	rqstp->rq_addrlen = sizeof(req->rq_xprt->addr);
@@ -1590,10 +1589,8 @@ bc_svc_process(struct svc_serv *serv, struct rpc_rqst *req,
 	 * been processed by the caller.
 	 */
 	svcxdr_init_decode(rqstp);
-	if (!xdr_inline_decode(&rqstp->rq_arg_stream, XDR_UNIT * 2)) {
-		error = -EINVAL;
-		goto out;
-	}
+	if (!xdr_inline_decode(&rqstp->rq_arg_stream, XDR_UNIT * 2))
+		return;
 
 	/* Parse and execute the bc call */
 	proc_error = svc_process_common(rqstp);
@@ -1602,26 +1599,18 @@ bc_svc_process(struct svc_serv *serv, struct rpc_rqst *req,
 	if (!proc_error) {
 		/* Processing error: drop the request */
 		xprt_free_bc_request(req);
-		error = -EINVAL;
-		goto out;
+		return;
 	}
 	/* Finally, send the reply synchronously */
 	memcpy(&req->rq_snd_buf, &rqstp->rq_res, sizeof(req->rq_snd_buf));
 	task = rpc_run_bc_task(req);
-	if (IS_ERR(task)) {
-		error = PTR_ERR(task);
-		goto out;
-	}
+	if (IS_ERR(task))
+		return;
 
 	WARN_ON_ONCE(atomic_read(&task->tk_count) != 1);
-	error = task->tk_status;
 	rpc_put_task(task);
-
-out:
-	dprintk("svc: %s(), error=%d\n", __func__, error);
-	return error;
 }
-EXPORT_SYMBOL_GPL(bc_svc_process);
+EXPORT_SYMBOL_GPL(svc_process_bc);
 #endif /* CONFIG_SUNRPC_BACKCHANNEL */
 
 /**
diff --git a/net/sunrpc/svc_xprt.c b/net/sunrpc/svc_xprt.c
index 4cfe9640df48..fee83d1024bc 100644
--- a/net/sunrpc/svc_xprt.c
+++ b/net/sunrpc/svc_xprt.c
@@ -9,7 +9,6 @@
 #include <linux/sched/mm.h>
 #include <linux/errno.h>
 #include <linux/freezer.h>
-#include <linux/kthread.h>
 #include <linux/slab.h>
 #include <net/sock.h>
 #include <linux/sunrpc/addr.h>
@@ -17,6 +16,7 @@
 #include <linux/sunrpc/svc_xprt.h>
 #include <linux/sunrpc/svcsock.h>
 #include <linux/sunrpc/xprt.h>
+#include <linux/sunrpc/bc_xprt.h>
 #include <linux/module.h>
 #include <linux/netdevice.h>
 #include <trace/events/sunrpc.h>
@@ -201,7 +201,6 @@ void svc_xprt_init(struct net *net, struct svc_xprt_class *xcl,
 	kref_init(&xprt->xpt_ref);
 	xprt->xpt_server = serv;
 	INIT_LIST_HEAD(&xprt->xpt_list);
-	INIT_LIST_HEAD(&xprt->xpt_ready);
 	INIT_LIST_HEAD(&xprt->xpt_deferred);
 	INIT_LIST_HEAD(&xprt->xpt_users);
 	mutex_init(&xprt->xpt_mutex);
@@ -472,9 +471,7 @@ void svc_xprt_enqueue(struct svc_xprt *xprt)
 	pool = svc_pool_for_cpu(xprt->xpt_server);
 
 	percpu_counter_inc(&pool->sp_sockets_queued);
-	spin_lock_bh(&pool->sp_lock);
-	list_add_tail(&xprt->xpt_ready, &pool->sp_sockets);
-	spin_unlock_bh(&pool->sp_lock);
+	lwq_enqueue(&xprt->xpt_ready, &pool->sp_xprts);
 
 	svc_pool_wake_idle_thread(pool);
 }
@@ -487,18 +484,9 @@ static struct svc_xprt *svc_xprt_dequeue(struct svc_pool *pool)
 {
 	struct svc_xprt	*xprt = NULL;
 
-	if (list_empty(&pool->sp_sockets))
-		goto out;
-
-	spin_lock_bh(&pool->sp_lock);
-	if (likely(!list_empty(&pool->sp_sockets))) {
-		xprt = list_first_entry(&pool->sp_sockets,
-					struct svc_xprt, xpt_ready);
-		list_del_init(&xprt->xpt_ready);
+	xprt = lwq_dequeue(&pool->sp_xprts, struct svc_xprt, xpt_ready);
+	if (xprt)
 		svc_xprt_get(xprt);
-	}
-	spin_unlock_bh(&pool->sp_lock);
-out:
 	return xprt;
 }
 
@@ -674,7 +662,7 @@ static bool svc_alloc_arg(struct svc_rqst *rqstp)
 			continue;
 
 		set_current_state(TASK_IDLE);
-		if (kthread_should_stop()) {
+		if (svc_thread_should_stop(rqstp)) {
 			set_current_state(TASK_RUNNING);
 			return false;
 		}
@@ -699,7 +687,7 @@ static bool svc_alloc_arg(struct svc_rqst *rqstp)
 }
 
 static bool
-rqst_should_sleep(struct svc_rqst *rqstp)
+svc_thread_should_sleep(struct svc_rqst *rqstp)
 {
 	struct svc_pool		*pool = rqstp->rq_pool;
 
@@ -708,65 +696,51 @@ rqst_should_sleep(struct svc_rqst *rqstp)
 		return false;
 
 	/* was a socket queued? */
-	if (!list_empty(&pool->sp_sockets))
+	if (!lwq_empty(&pool->sp_xprts))
 		return false;
 
 	/* are we shutting down? */
-	if (kthread_should_stop())
+	if (svc_thread_should_stop(rqstp))
 		return false;
 
-	/* are we freezing? */
-	if (freezing(current))
-		return false;
+#if defined(CONFIG_SUNRPC_BACKCHANNEL)
+	if (svc_is_backchannel(rqstp)) {
+		if (!lwq_empty(&rqstp->rq_server->sv_cb_list))
+			return false;
+	}
+#endif
 
 	return true;
 }
 
-static struct svc_xprt *svc_get_next_xprt(struct svc_rqst *rqstp)
+static void svc_thread_wait_for_work(struct svc_rqst *rqstp)
 {
-	struct svc_pool		*pool = rqstp->rq_pool;
-
-	/* rq_xprt should be clear on entry */
-	WARN_ON_ONCE(rqstp->rq_xprt);
-
-	rqstp->rq_xprt = svc_xprt_dequeue(pool);
-	if (rqstp->rq_xprt)
-		goto out_found;
-
-	set_current_state(TASK_IDLE);
-	smp_mb__before_atomic();
-	clear_bit(SP_CONGESTED, &pool->sp_flags);
-	clear_bit(RQ_BUSY, &rqstp->rq_flags);
-	smp_mb__after_atomic();
-
-	if (likely(rqst_should_sleep(rqstp)))
-		schedule();
-	else
+	struct svc_pool *pool = rqstp->rq_pool;
+
+	if (svc_thread_should_sleep(rqstp)) {
+		set_current_state(TASK_IDLE | TASK_FREEZABLE);
+		llist_add(&rqstp->rq_idle, &pool->sp_idle_threads);
+		if (likely(svc_thread_should_sleep(rqstp)))
+			schedule();
+
+		while (!llist_del_first_this(&pool->sp_idle_threads,
+					     &rqstp->rq_idle)) {
+			/* Work just became available.  This thread can only
+			 * handle it after removing rqstp from the idle
+			 * list. If that attempt failed, some other thread
+			 * must have queued itself after finding no
+			 * work to do, so that thread has taken responsibly
+			 * for this new work.  This thread can safely sleep
+			 * until woken again.
+			 */
+			schedule();
+			set_current_state(TASK_IDLE | TASK_FREEZABLE);
+		}
 		__set_current_state(TASK_RUNNING);
-
+	} else {
+		cond_resched();
+	}
 	try_to_freeze();
-
-	set_bit(RQ_BUSY, &rqstp->rq_flags);
-	smp_mb__after_atomic();
-	clear_bit(SP_TASK_PENDING, &pool->sp_flags);
-	rqstp->rq_xprt = svc_xprt_dequeue(pool);
-	if (rqstp->rq_xprt)
-		goto out_found;
-
-	if (kthread_should_stop())
-		return NULL;
-	return NULL;
-out_found:
-	clear_bit(SP_TASK_PENDING, &pool->sp_flags);
-	/* Normally we will wait up to 5 seconds for any required
-	 * cache information to be provided.
-	 */
-	if (!test_bit(SP_CONGESTED, &pool->sp_flags))
-		rqstp->rq_chandle.thread_wait = 5*HZ;
-	else
-		rqstp->rq_chandle.thread_wait = 1*HZ;
-	trace_svc_xprt_dequeue(rqstp);
-	return rqstp->rq_xprt;
 }
 
 static void svc_add_new_temp_xprt(struct svc_serv *serv, struct svc_xprt *newxpt)
@@ -785,7 +759,7 @@ static void svc_add_new_temp_xprt(struct svc_serv *serv, struct svc_xprt *newxpt
 	svc_xprt_received(newxpt);
 }
 
-static int svc_handle_xprt(struct svc_rqst *rqstp, struct svc_xprt *xprt)
+static void svc_handle_xprt(struct svc_rqst *rqstp, struct svc_xprt *xprt)
 {
 	struct svc_serv *serv = rqstp->rq_server;
 	int len = 0;
@@ -826,11 +800,35 @@ static int svc_handle_xprt(struct svc_rqst *rqstp, struct svc_xprt *xprt)
 			len = xprt->xpt_ops->xpo_recvfrom(rqstp);
 		rqstp->rq_reserved = serv->sv_max_mesg;
 		atomic_add(rqstp->rq_reserved, &xprt->xpt_reserved);
+		if (len <= 0)
+			goto out;
+
+		trace_svc_xdr_recvfrom(&rqstp->rq_arg);
+
+		clear_bit(XPT_OLD, &xprt->xpt_flags);
+
+		rqstp->rq_chandle.defer = svc_defer;
+
+		if (serv->sv_stats)
+			serv->sv_stats->netcnt++;
+		percpu_counter_inc(&rqstp->rq_pool->sp_messages_arrived);
+		rqstp->rq_stime = ktime_get();
+		svc_process(rqstp);
 	} else
 		svc_xprt_received(xprt);
 
 out:
-	return len;
+	rqstp->rq_res.len = 0;
+	svc_xprt_release(rqstp);
+}
+
+static void svc_thread_wake_next(struct svc_rqst *rqstp)
+{
+	if (!svc_thread_should_sleep(rqstp))
+		/* More work pending after I dequeued some,
+		 * wake another worker
+		 */
+		svc_pool_wake_idle_thread(rqstp->rq_pool);
 }
 
 /**
@@ -843,44 +841,51 @@ out:
  */
 void svc_recv(struct svc_rqst *rqstp)
 {
-	struct svc_xprt		*xprt = NULL;
-	struct svc_serv		*serv = rqstp->rq_server;
-	int			len;
+	struct svc_pool *pool = rqstp->rq_pool;
 
 	if (!svc_alloc_arg(rqstp))
-		goto out;
+		return;
 
-	try_to_freeze();
-	cond_resched();
-	if (kthread_should_stop())
-		goto out;
+	svc_thread_wait_for_work(rqstp);
 
-	xprt = svc_get_next_xprt(rqstp);
-	if (!xprt)
-		goto out;
+	clear_bit(SP_TASK_PENDING, &pool->sp_flags);
 
-	len = svc_handle_xprt(rqstp, xprt);
+	if (svc_thread_should_stop(rqstp)) {
+		svc_thread_wake_next(rqstp);
+		return;
+	}
 
-	/* No data, incomplete (TCP) read, or accept() */
-	if (len <= 0)
-		goto out_release;
+	rqstp->rq_xprt = svc_xprt_dequeue(pool);
+	if (rqstp->rq_xprt) {
+		struct svc_xprt *xprt = rqstp->rq_xprt;
 
-	trace_svc_xdr_recvfrom(&rqstp->rq_arg);
+		svc_thread_wake_next(rqstp);
+		/* Normally we will wait up to 5 seconds for any required
+		 * cache information to be provided.  When there are no
+		 * idle threads, we reduce the wait time.
+		 */
+		if (pool->sp_idle_threads.first)
+			rqstp->rq_chandle.thread_wait = 5 * HZ;
+		else
+			rqstp->rq_chandle.thread_wait = 1 * HZ;
 
-	clear_bit(XPT_OLD, &xprt->xpt_flags);
+		trace_svc_xprt_dequeue(rqstp);
+		svc_handle_xprt(rqstp, xprt);
+	}
 
-	rqstp->rq_chandle.defer = svc_defer;
+#if defined(CONFIG_SUNRPC_BACKCHANNEL)
+	if (svc_is_backchannel(rqstp)) {
+		struct svc_serv *serv = rqstp->rq_server;
+		struct rpc_rqst *req;
 
-	if (serv->sv_stats)
-		serv->sv_stats->netcnt++;
-	percpu_counter_inc(&rqstp->rq_pool->sp_messages_arrived);
-	rqstp->rq_stime = ktime_get();
-	svc_process(rqstp);
-out:
-	return;
-out_release:
-	rqstp->rq_res.len = 0;
-	svc_xprt_release(rqstp);
+		req = lwq_dequeue(&serv->sv_cb_list,
+				  struct rpc_rqst, rq_bc_list);
+		if (req) {
+			svc_thread_wake_next(rqstp);
+			svc_process_bc(req, rqstp);
+		}
+	}
+#endif
 }
 EXPORT_SYMBOL_GPL(svc_recv);
 
@@ -890,7 +895,6 @@ EXPORT_SYMBOL_GPL(svc_recv);
 void svc_drop(struct svc_rqst *rqstp)
 {
 	trace_svc_drop(rqstp);
-	svc_xprt_release(rqstp);
 }
 EXPORT_SYMBOL_GPL(svc_drop);
 
@@ -906,8 +910,6 @@ void svc_send(struct svc_rqst *rqstp)
 	int status;
 
 	xprt = rqstp->rq_xprt;
-	if (!xprt)
-		return;
 
 	/* calculate over-all length */
 	xb = &rqstp->rq_res;
@@ -920,7 +922,6 @@ void svc_send(struct svc_rqst *rqstp)
 	status = xprt->xpt_ops->xpo_sendto(rqstp);
 
 	trace_svc_send(rqstp, status);
-	svc_xprt_release(rqstp);
 }
 
 /*
@@ -1031,7 +1032,6 @@ static void svc_delete_xprt(struct svc_xprt *xprt)
 
 	spin_lock_bh(&serv->sv_lock);
 	list_del_init(&xprt->xpt_list);
-	WARN_ON_ONCE(!list_empty(&xprt->xpt_ready));
 	if (test_bit(XPT_TEMP, &xprt->xpt_flags))
 		serv->sv_tmpcnt--;
 	spin_unlock_bh(&serv->sv_lock);
@@ -1082,36 +1082,26 @@ static int svc_close_list(struct svc_serv *serv, struct list_head *xprt_list, st
 	return ret;
 }
 
-static struct svc_xprt *svc_dequeue_net(struct svc_serv *serv, struct net *net)
+static void svc_clean_up_xprts(struct svc_serv *serv, struct net *net)
 {
-	struct svc_pool *pool;
 	struct svc_xprt *xprt;
-	struct svc_xprt *tmp;
 	int i;
 
 	for (i = 0; i < serv->sv_nrpools; i++) {
-		pool = &serv->sv_pools[i];
-
-		spin_lock_bh(&pool->sp_lock);
-		list_for_each_entry_safe(xprt, tmp, &pool->sp_sockets, xpt_ready) {
-			if (xprt->xpt_net != net)
-				continue;
-			list_del_init(&xprt->xpt_ready);
-			spin_unlock_bh(&pool->sp_lock);
-			return xprt;
+		struct svc_pool *pool = &serv->sv_pools[i];
+		struct llist_node *q, **t1, *t2;
+
+		q = lwq_dequeue_all(&pool->sp_xprts);
+		lwq_for_each_safe(xprt, t1, t2, &q, xpt_ready) {
+			if (xprt->xpt_net == net) {
+				set_bit(XPT_CLOSE, &xprt->xpt_flags);
+				svc_delete_xprt(xprt);
+				xprt = NULL;
+			}
 		}
-		spin_unlock_bh(&pool->sp_lock);
-	}
-	return NULL;
-}
-
-static void svc_clean_up_xprts(struct svc_serv *serv, struct net *net)
-{
-	struct svc_xprt *xprt;
 
-	while ((xprt = svc_dequeue_net(serv, net))) {
-		set_bit(XPT_CLOSE, &xprt->xpt_flags);
-		svc_delete_xprt(xprt);
+		if (q)
+			lwq_enqueue_batch(q, &pool->sp_xprts);
 	}
 }
 
diff --git a/net/sunrpc/xprtrdma/backchannel.c b/net/sunrpc/xprtrdma/backchannel.c
index e4d84a13c566..8c817e755262 100644
--- a/net/sunrpc/xprtrdma/backchannel.c
+++ b/net/sunrpc/xprtrdma/backchannel.c
@@ -263,11 +263,9 @@ void rpcrdma_bc_receive_call(struct rpcrdma_xprt *r_xprt,
 	/* Queue rqst for ULP's callback service */
 	bc_serv = xprt->bc_serv;
 	xprt_get(xprt);
-	spin_lock(&bc_serv->sv_cb_lock);
-	list_add(&rqst->rq_bc_list, &bc_serv->sv_cb_list);
-	spin_unlock(&bc_serv->sv_cb_lock);
+	lwq_enqueue(&rqst->rq_bc_list, &bc_serv->sv_cb_list);
 
-	wake_up(&bc_serv->sv_cb_waitq);
+	svc_pool_wake_idle_thread(&bc_serv->sv_pools[0]);
 
 	r_xprt->rx_stats.bcall_count++;
 	return;
diff --git a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
index 85c8bcaebb80..3b05f90a3e50 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
@@ -852,7 +852,8 @@ out_readfail:
 	if (ret == -EINVAL)
 		svc_rdma_send_error(rdma_xprt, ctxt, ret);
 	svc_rdma_recv_ctxt_put(rdma_xprt, ctxt);
-	return ret;
+	svc_xprt_deferred_close(xprt);
+	return -ENOTCONN;
 
 out_backchannel:
 	svc_rdma_handle_bc_reply(rqstp, ctxt);