1 files changed, 52 insertions, 90 deletions

diff --git a/net/sunrpc/xprtrdma/transport.c b/net/sunrpc/xprtrdma/transport.c
index 3cfeba68ee9a..10bb2b929c6d 100644
--- a/net/sunrpc/xprtrdma/transport.c
+++ b/net/sunrpc/xprtrdma/transport.c
@@ -60,19 +60,16 @@
 #include "xprt_rdma.h"
 #include <trace/events/rpcrdma.h>
 
-#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
-# define RPCDBG_FACILITY	RPCDBG_TRANS
-#endif
-
 /*
  * tunables
  */
 
-unsigned int xprt_rdma_slot_table_entries = RPCRDMA_DEF_SLOT_TABLE;
+static unsigned int xprt_rdma_slot_table_entries = RPCRDMA_DEF_SLOT_TABLE;
 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)
 
@@ -238,26 +235,35 @@ xprt_rdma_connect_worker(struct work_struct *work)
 	struct rpcrdma_xprt *r_xprt = container_of(work, struct rpcrdma_xprt,
 						   rx_connect_worker.work);
 	struct rpc_xprt *xprt = &r_xprt->rx_xprt;
+	unsigned int pflags = current->flags;
 	int rc;
 
-	rc = rpcrdma_ep_connect(&r_xprt->rx_ep, &r_xprt->rx_ia);
+	if (atomic_read(&xprt->swapper))
+		current->flags |= PF_MEMALLOC;
+	rc = rpcrdma_xprt_connect(r_xprt);
 	xprt_clear_connecting(xprt);
-	if (r_xprt->rx_ep.rep_connected > 0) {
+	if (!rc) {
+		xprt->connect_cookie++;
 		xprt->stat.connect_count++;
 		xprt->stat.connect_time += (long)jiffies -
 					   xprt->stat.connect_start;
 		xprt_set_connected(xprt);
 		rc = -EAGAIN;
-	}
+	} else
+		rpcrdma_xprt_disconnect(r_xprt);
+	xprt_unlock_connect(xprt, r_xprt);
 	xprt_wake_pending_tasks(xprt, rc);
+	current_restore_flags(pflags, PF_MEMALLOC);
 }
 
 /**
  * xprt_rdma_inject_disconnect - inject a connection fault
  * @xprt: transport context
  *
- * If @xprt is connected, disconnect it to simulate spurious connection
- * loss.
+ * If @xprt is connected, disconnect it to simulate spurious
+ * connection loss. Caller must hold @xprt's send lock to
+ * ensure that data structures and hardware resources are
+ * stable during the rdma_disconnect() call.
  */
 static void
 xprt_rdma_inject_disconnect(struct rpc_xprt *xprt)
@@ -265,7 +271,7 @@ xprt_rdma_inject_disconnect(struct rpc_xprt *xprt)
 	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
 
 	trace_xprtrdma_op_inject_dsc(r_xprt);
-	rdma_disconnect(r_xprt->rx_ia.ri_id);
+	rdma_disconnect(r_xprt->rx_ep->re_id);
 }
 
 /**
@@ -280,13 +286,10 @@ xprt_rdma_destroy(struct rpc_xprt *xprt)
 {
 	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
 
-	trace_xprtrdma_op_destroy(r_xprt);
-
 	cancel_delayed_work_sync(&r_xprt->rx_connect_worker);
 
-	rpcrdma_ep_destroy(r_xprt);
+	rpcrdma_xprt_disconnect(r_xprt);
 	rpcrdma_buffer_destroy(&r_xprt->rx_buf);
-	rpcrdma_ia_close(&r_xprt->rx_ia);
 
 	xprt_rdma_free_addresses(xprt);
 	xprt_free(xprt);
@@ -316,10 +319,15 @@ xprt_setup_rdma(struct xprt_create *args)
 	if (args->addrlen > sizeof(xprt->addr))
 		return ERR_PTR(-EBADF);
 
+	if (!try_module_get(THIS_MODULE))
+		return ERR_PTR(-EIO);
+
 	xprt = xprt_alloc(args->net, sizeof(struct rpcrdma_xprt), 0,
 			  xprt_rdma_slot_table_entries);
-	if (!xprt)
+	if (!xprt) {
+		module_put(THIS_MODULE);
 		return ERR_PTR(-ENOMEM);
+	}
 
 	xprt->timeout = &xprt_rdma_default_timeout;
 	xprt->connect_timeout = xprt->timeout->to_initval;
@@ -339,6 +347,7 @@ 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);
 
@@ -347,43 +356,20 @@ xprt_setup_rdma(struct xprt_create *args)
 	xprt_rdma_format_addresses(xprt, sap);
 
 	new_xprt = rpcx_to_rdmax(xprt);
-	rc = rpcrdma_ia_open(new_xprt);
-	if (rc)
-		goto out1;
-
-	rc = rpcrdma_ep_create(new_xprt);
-	if (rc)
-		goto out2;
-
 	rc = rpcrdma_buffer_create(new_xprt);
-	if (rc)
-		goto out3;
-
-	if (!try_module_get(THIS_MODULE))
-		goto out4;
+	if (rc) {
+		xprt_rdma_free_addresses(xprt);
+		xprt_free(xprt);
+		module_put(THIS_MODULE);
+		return ERR_PTR(rc);
+	}
 
 	INIT_DELAYED_WORK(&new_xprt->rx_connect_worker,
 			  xprt_rdma_connect_worker);
+
 	xprt->max_payload = RPCRDMA_MAX_DATA_SEGS << PAGE_SHIFT;
 
-	dprintk("RPC:       %s: %s:%s\n", __func__,
-		xprt->address_strings[RPC_DISPLAY_ADDR],
-		xprt->address_strings[RPC_DISPLAY_PORT]);
-	trace_xprtrdma_create(new_xprt);
 	return xprt;
-
-out4:
-	rpcrdma_buffer_destroy(&new_xprt->rx_buf);
-	rc = -ENODEV;
-out3:
-	rpcrdma_ep_destroy(new_xprt);
-out2:
-	rpcrdma_ia_close(&new_xprt->rx_ia);
-out1:
-	trace_xprtrdma_op_destroy(new_xprt);
-	xprt_rdma_free_addresses(xprt);
-	xprt_free(xprt);
-	return ERR_PTR(rc);
 }
 
 /**
@@ -398,26 +384,9 @@ out1:
 void xprt_rdma_close(struct rpc_xprt *xprt)
 {
 	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
-	struct rpcrdma_ep *ep = &r_xprt->rx_ep;
-	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
-
-	might_sleep();
 
-	trace_xprtrdma_op_close(r_xprt);
+	rpcrdma_xprt_disconnect(r_xprt);
 
-	/* Prevent marshaling and sending of new requests */
-	xprt_clear_connected(xprt);
-
-	if (test_and_clear_bit(RPCRDMA_IAF_REMOVING, &ia->ri_flags)) {
-		rpcrdma_ia_remove(ia);
-		goto out;
-	}
-
-	if (ep->rep_connected == -ENODEV)
-		return;
-	rpcrdma_ep_disconnect(ep, ia);
-
-out:
 	xprt->reestablish_timeout = 0;
 	++xprt->connect_cookie;
 	xprt_disconnect_done(xprt);
@@ -445,9 +414,6 @@ xprt_rdma_set_port(struct rpc_xprt *xprt, u16 port)
 	kfree(xprt->address_strings[RPC_DISPLAY_HEX_PORT]);
 	snprintf(buf, sizeof(buf), "%4hx", port);
 	xprt->address_strings[RPC_DISPLAY_HEX_PORT] = kstrdup(buf, GFP_KERNEL);
-
-	trace_xprtrdma_op_setport(container_of(xprt, struct rpcrdma_xprt,
-					       rx_xprt));
 }
 
 /**
@@ -517,16 +483,18 @@ static void
 xprt_rdma_connect(struct rpc_xprt *xprt, struct rpc_task *task)
 {
 	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
+	struct rpcrdma_ep *ep = r_xprt->rx_ep;
 	unsigned long delay;
 
+	WARN_ON_ONCE(!xprt_lock_connect(xprt, task, r_xprt));
+
 	delay = 0;
-	if (r_xprt->rx_ep.rep_connected != 0) {
+	if (ep && ep->re_connect_status != 0) {
 		delay = xprt_reconnect_delay(xprt);
 		xprt_reconnect_backoff(xprt, RPCRDMA_INIT_REEST_TO);
 	}
 	trace_xprtrdma_op_connect(r_xprt, delay);
-	queue_delayed_work(xprtiod_workqueue, &r_xprt->rx_connect_worker,
-			   delay);
+	queue_delayed_work(system_long_wq, &r_xprt->rx_connect_worker, delay);
 }
 
 /**
@@ -552,9 +520,8 @@ xprt_rdma_alloc_slot(struct rpc_xprt *xprt, struct rpc_task *task)
 	return;
 
 out_sleep:
-	set_bit(XPRT_CONGESTED, &xprt->state);
-	rpc_sleep_on(&xprt->backlog, task, NULL);
-	task->tk_status = -EAGAIN;
+	task->tk_status = -ENOMEM;
+	xprt_add_backlog(xprt, task);
 }
 
 /**
@@ -569,10 +536,11 @@ xprt_rdma_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *rqst)
 	struct rpcrdma_xprt *r_xprt =
 		container_of(xprt, struct rpcrdma_xprt, rx_xprt);
 
-	memset(rqst, 0, sizeof(*rqst));
-	rpcrdma_buffer_put(&r_xprt->rx_buf, rpcr_to_rdmar(rqst));
-	if (unlikely(!rpc_wake_up_next(&xprt->backlog)))
-		clear_bit(XPRT_CONGESTED, &xprt->state);
+	rpcrdma_reply_put(&r_xprt->rx_buf, rpcr_to_rdmar(rqst));
+	if (!xprt_wake_up_backlog(xprt, rqst)) {
+		memset(rqst, 0, sizeof(*rqst));
+		rpcrdma_buffer_put(&r_xprt->rx_buf, rpcr_to_rdmar(rqst));
+	}
 }
 
 static bool rpcrdma_check_regbuf(struct rpcrdma_xprt *r_xprt,
@@ -602,11 +570,7 @@ xprt_rdma_allocate(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);
-	gfp_t flags;
-
-	flags = RPCRDMA_DEF_GFP;
-	if (RPC_IS_SWAPPER(task))
-		flags = __GFP_MEMALLOC | GFP_NOWAIT | __GFP_NOWARN;
+	gfp_t flags = rpc_task_gfp_mask();
 
 	if (!rpcrdma_check_regbuf(r_xprt, req->rl_sendbuf, rqst->rq_callsize,
 				  flags))
@@ -617,11 +581,9 @@ xprt_rdma_allocate(struct rpc_task *task)
 
 	rqst->rq_buffer = rdmab_data(req->rl_sendbuf);
 	rqst->rq_rbuffer = rdmab_data(req->rl_recvbuf);
-	trace_xprtrdma_op_allocate(task, req);
 	return 0;
 
 out_fail:
-	trace_xprtrdma_op_allocate(task, NULL);
 	return -ENOMEM;
 }
 
@@ -635,13 +597,12 @@ static void
 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);
 
-	trace_xprtrdma_op_free(task, req);
-
-	if (!list_empty(&req->rl_registered))
-		frwr_unmap_sync(r_xprt, req);
+	if (unlikely(!list_empty(&req->rl_registered))) {
+		trace_xprtrdma_mrs_zap(task);
+		frwr_unmap_sync(rpcx_to_rdmax(rqst->rq_xprt), req);
+	}
 
 	/* XXX: If the RPC is completing because of a signal and
 	 * not because a reply was received, we ought to ensure
@@ -694,7 +655,7 @@ xprt_rdma_send_request(struct rpc_rqst *rqst)
 		goto drop_connection;
 	rqst->rq_xtime = ktime_get();
 
-	if (rpcrdma_ep_post(&r_xprt->rx_ia, &r_xprt->rx_ep, req))
+	if (frwr_send(r_xprt, req))
 		goto drop_connection;
 
 	rqst->rq_xmit_bytes_sent += rqst->rq_snd_buf.len;
@@ -806,6 +767,7 @@ static struct xprt_class xprt_rdma = {
 	.owner			= THIS_MODULE,
 	.ident			= XPRT_TRANSPORT_RDMA,
 	.setup			= xprt_setup_rdma,
+	.netid			= { "rdma", "rdma6", "" },
 };
 
 void xprt_rdma_cleanup(void)