xref: /openbmc/linux/net/sunrpc/xprtrdma/verbs.c (revision 877013bc)
1 // SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause
2 /*
3  * Copyright (c) 2014-2017 Oracle.  All rights reserved.
4  * Copyright (c) 2003-2007 Network Appliance, Inc. All rights reserved.
5  *
6  * This software is available to you under a choice of one of two
7  * licenses.  You may choose to be licensed under the terms of the GNU
8  * General Public License (GPL) Version 2, available from the file
9  * COPYING in the main directory of this source tree, or the BSD-type
10  * license below:
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  *
16  *      Redistributions of source code must retain the above copyright
17  *      notice, this list of conditions and the following disclaimer.
18  *
19  *      Redistributions in binary form must reproduce the above
20  *      copyright notice, this list of conditions and the following
21  *      disclaimer in the documentation and/or other materials provided
22  *      with the distribution.
23  *
24  *      Neither the name of the Network Appliance, Inc. nor the names of
25  *      its contributors may be used to endorse or promote products
26  *      derived from this software without specific prior written
27  *      permission.
28  *
29  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
30  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
31  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
32  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
33  * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
34  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
35  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
36  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
37  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
38  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
39  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
40  */
41 
42 /*
43  * verbs.c
44  *
45  * Encapsulates the major functions managing:
46  *  o adapters
47  *  o endpoints
48  *  o connections
49  *  o buffer memory
50  */
51 
52 #include <linux/interrupt.h>
53 #include <linux/slab.h>
54 #include <linux/sunrpc/addr.h>
55 #include <linux/sunrpc/svc_rdma.h>
56 #include <linux/log2.h>
57 
58 #include <asm-generic/barrier.h>
59 #include <asm/bitops.h>
60 
61 #include <rdma/ib_cm.h>
62 
63 #include "xprt_rdma.h"
64 #include <trace/events/rpcrdma.h>
65 
66 /*
67  * Globals/Macros
68  */
69 
70 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
71 # define RPCDBG_FACILITY	RPCDBG_TRANS
72 #endif
73 
74 /*
75  * internal functions
76  */
77 static int rpcrdma_sendctxs_create(struct rpcrdma_xprt *r_xprt);
78 static void rpcrdma_sendctxs_destroy(struct rpcrdma_xprt *r_xprt);
79 static void rpcrdma_sendctx_put_locked(struct rpcrdma_xprt *r_xprt,
80 				       struct rpcrdma_sendctx *sc);
81 static int rpcrdma_reqs_setup(struct rpcrdma_xprt *r_xprt);
82 static void rpcrdma_reqs_reset(struct rpcrdma_xprt *r_xprt);
83 static void rpcrdma_rep_destroy(struct rpcrdma_rep *rep);
84 static void rpcrdma_reps_unmap(struct rpcrdma_xprt *r_xprt);
85 static void rpcrdma_mrs_create(struct rpcrdma_xprt *r_xprt);
86 static void rpcrdma_mrs_destroy(struct rpcrdma_xprt *r_xprt);
87 static void rpcrdma_ep_get(struct rpcrdma_ep *ep);
88 static int rpcrdma_ep_put(struct rpcrdma_ep *ep);
89 static struct rpcrdma_regbuf *
90 rpcrdma_regbuf_alloc(size_t size, enum dma_data_direction direction,
91 		     gfp_t flags);
92 static void rpcrdma_regbuf_dma_unmap(struct rpcrdma_regbuf *rb);
93 static void rpcrdma_regbuf_free(struct rpcrdma_regbuf *rb);
94 
95 /* Wait for outstanding transport work to finish. ib_drain_qp
96  * handles the drains in the wrong order for us, so open code
97  * them here.
98  */
99 static void rpcrdma_xprt_drain(struct rpcrdma_xprt *r_xprt)
100 {
101 	struct rpcrdma_ep *ep = r_xprt->rx_ep;
102 	struct rdma_cm_id *id = ep->re_id;
103 
104 	/* Wait for rpcrdma_post_recvs() to leave its critical
105 	 * section.
106 	 */
107 	if (atomic_inc_return(&ep->re_receiving) > 1)
108 		wait_for_completion(&ep->re_done);
109 
110 	/* Flush Receives, then wait for deferred Reply work
111 	 * to complete.
112 	 */
113 	ib_drain_rq(id->qp);
114 
115 	/* Deferred Reply processing might have scheduled
116 	 * local invalidations.
117 	 */
118 	ib_drain_sq(id->qp);
119 
120 	rpcrdma_ep_put(ep);
121 }
122 
123 /* Ensure xprt_force_disconnect() is invoked exactly once when a
124  * connection is closed or lost. (The important thing is it needs
125  * to be invoked "at least" once).
126  */
127 static void rpcrdma_force_disconnect(struct rpcrdma_ep *ep)
128 {
129 	if (atomic_add_unless(&ep->re_force_disconnect, 1, 1))
130 		xprt_force_disconnect(ep->re_xprt);
131 }
132 
133 /**
134  * rpcrdma_flush_disconnect - Disconnect on flushed completion
135  * @r_xprt: transport to disconnect
136  * @wc: work completion entry
137  *
138  * Must be called in process context.
139  */
140 void rpcrdma_flush_disconnect(struct rpcrdma_xprt *r_xprt, struct ib_wc *wc)
141 {
142 	if (wc->status != IB_WC_SUCCESS)
143 		rpcrdma_force_disconnect(r_xprt->rx_ep);
144 }
145 
146 /**
147  * rpcrdma_wc_send - Invoked by RDMA provider for each polled Send WC
148  * @cq:	completion queue
149  * @wc:	WCE for a completed Send WR
150  *
151  */
152 static void rpcrdma_wc_send(struct ib_cq *cq, struct ib_wc *wc)
153 {
154 	struct ib_cqe *cqe = wc->wr_cqe;
155 	struct rpcrdma_sendctx *sc =
156 		container_of(cqe, struct rpcrdma_sendctx, sc_cqe);
157 	struct rpcrdma_xprt *r_xprt = cq->cq_context;
158 
159 	/* WARNING: Only wr_cqe and status are reliable at this point */
160 	trace_xprtrdma_wc_send(wc, &sc->sc_cid);
161 	rpcrdma_sendctx_put_locked(r_xprt, sc);
162 	rpcrdma_flush_disconnect(r_xprt, wc);
163 }
164 
165 /**
166  * rpcrdma_wc_receive - Invoked by RDMA provider for each polled Receive WC
167  * @cq:	completion queue
168  * @wc:	WCE for a completed Receive WR
169  *
170  */
171 static void rpcrdma_wc_receive(struct ib_cq *cq, struct ib_wc *wc)
172 {
173 	struct ib_cqe *cqe = wc->wr_cqe;
174 	struct rpcrdma_rep *rep = container_of(cqe, struct rpcrdma_rep,
175 					       rr_cqe);
176 	struct rpcrdma_xprt *r_xprt = cq->cq_context;
177 
178 	/* WARNING: Only wr_cqe and status are reliable at this point */
179 	trace_xprtrdma_wc_receive(wc, &rep->rr_cid);
180 	--r_xprt->rx_ep->re_receive_count;
181 	if (wc->status != IB_WC_SUCCESS)
182 		goto out_flushed;
183 
184 	/* status == SUCCESS means all fields in wc are trustworthy */
185 	rpcrdma_set_xdrlen(&rep->rr_hdrbuf, wc->byte_len);
186 	rep->rr_wc_flags = wc->wc_flags;
187 	rep->rr_inv_rkey = wc->ex.invalidate_rkey;
188 
189 	ib_dma_sync_single_for_cpu(rdmab_device(rep->rr_rdmabuf),
190 				   rdmab_addr(rep->rr_rdmabuf),
191 				   wc->byte_len, DMA_FROM_DEVICE);
192 
193 	rpcrdma_reply_handler(rep);
194 	return;
195 
196 out_flushed:
197 	rpcrdma_flush_disconnect(r_xprt, wc);
198 	rpcrdma_rep_put(&r_xprt->rx_buf, rep);
199 }
200 
201 static void rpcrdma_update_cm_private(struct rpcrdma_ep *ep,
202 				      struct rdma_conn_param *param)
203 {
204 	const struct rpcrdma_connect_private *pmsg = param->private_data;
205 	unsigned int rsize, wsize;
206 
207 	/* Default settings for RPC-over-RDMA Version One */
208 	ep->re_implicit_roundup = xprt_rdma_pad_optimize;
209 	rsize = RPCRDMA_V1_DEF_INLINE_SIZE;
210 	wsize = RPCRDMA_V1_DEF_INLINE_SIZE;
211 
212 	if (pmsg &&
213 	    pmsg->cp_magic == rpcrdma_cmp_magic &&
214 	    pmsg->cp_version == RPCRDMA_CMP_VERSION) {
215 		ep->re_implicit_roundup = true;
216 		rsize = rpcrdma_decode_buffer_size(pmsg->cp_send_size);
217 		wsize = rpcrdma_decode_buffer_size(pmsg->cp_recv_size);
218 	}
219 
220 	if (rsize < ep->re_inline_recv)
221 		ep->re_inline_recv = rsize;
222 	if (wsize < ep->re_inline_send)
223 		ep->re_inline_send = wsize;
224 
225 	rpcrdma_set_max_header_sizes(ep);
226 }
227 
228 /**
229  * rpcrdma_cm_event_handler - Handle RDMA CM events
230  * @id: rdma_cm_id on which an event has occurred
231  * @event: details of the event
232  *
233  * Called with @id's mutex held. Returns 1 if caller should
234  * destroy @id, otherwise 0.
235  */
236 static int
237 rpcrdma_cm_event_handler(struct rdma_cm_id *id, struct rdma_cm_event *event)
238 {
239 	struct sockaddr *sap = (struct sockaddr *)&id->route.addr.dst_addr;
240 	struct rpcrdma_ep *ep = id->context;
241 
242 	might_sleep();
243 
244 	switch (event->event) {
245 	case RDMA_CM_EVENT_ADDR_RESOLVED:
246 	case RDMA_CM_EVENT_ROUTE_RESOLVED:
247 		ep->re_async_rc = 0;
248 		complete(&ep->re_done);
249 		return 0;
250 	case RDMA_CM_EVENT_ADDR_ERROR:
251 		ep->re_async_rc = -EPROTO;
252 		complete(&ep->re_done);
253 		return 0;
254 	case RDMA_CM_EVENT_ROUTE_ERROR:
255 		ep->re_async_rc = -ENETUNREACH;
256 		complete(&ep->re_done);
257 		return 0;
258 	case RDMA_CM_EVENT_DEVICE_REMOVAL:
259 		pr_info("rpcrdma: removing device %s for %pISpc\n",
260 			ep->re_id->device->name, sap);
261 		fallthrough;
262 	case RDMA_CM_EVENT_ADDR_CHANGE:
263 		ep->re_connect_status = -ENODEV;
264 		goto disconnected;
265 	case RDMA_CM_EVENT_ESTABLISHED:
266 		rpcrdma_ep_get(ep);
267 		ep->re_connect_status = 1;
268 		rpcrdma_update_cm_private(ep, &event->param.conn);
269 		trace_xprtrdma_inline_thresh(ep);
270 		wake_up_all(&ep->re_connect_wait);
271 		break;
272 	case RDMA_CM_EVENT_CONNECT_ERROR:
273 		ep->re_connect_status = -ENOTCONN;
274 		goto wake_connect_worker;
275 	case RDMA_CM_EVENT_UNREACHABLE:
276 		ep->re_connect_status = -ENETUNREACH;
277 		goto wake_connect_worker;
278 	case RDMA_CM_EVENT_REJECTED:
279 		dprintk("rpcrdma: connection to %pISpc rejected: %s\n",
280 			sap, rdma_reject_msg(id, event->status));
281 		ep->re_connect_status = -ECONNREFUSED;
282 		if (event->status == IB_CM_REJ_STALE_CONN)
283 			ep->re_connect_status = -ENOTCONN;
284 wake_connect_worker:
285 		wake_up_all(&ep->re_connect_wait);
286 		return 0;
287 	case RDMA_CM_EVENT_DISCONNECTED:
288 		ep->re_connect_status = -ECONNABORTED;
289 disconnected:
290 		rpcrdma_force_disconnect(ep);
291 		return rpcrdma_ep_put(ep);
292 	default:
293 		break;
294 	}
295 
296 	dprintk("RPC:       %s: %pISpc on %s/frwr: %s\n", __func__, sap,
297 		ep->re_id->device->name, rdma_event_msg(event->event));
298 	return 0;
299 }
300 
301 static struct rdma_cm_id *rpcrdma_create_id(struct rpcrdma_xprt *r_xprt,
302 					    struct rpcrdma_ep *ep)
303 {
304 	unsigned long wtimeout = msecs_to_jiffies(RDMA_RESOLVE_TIMEOUT) + 1;
305 	struct rpc_xprt *xprt = &r_xprt->rx_xprt;
306 	struct rdma_cm_id *id;
307 	int rc;
308 
309 	init_completion(&ep->re_done);
310 
311 	id = rdma_create_id(xprt->xprt_net, rpcrdma_cm_event_handler, ep,
312 			    RDMA_PS_TCP, IB_QPT_RC);
313 	if (IS_ERR(id))
314 		return id;
315 
316 	ep->re_async_rc = -ETIMEDOUT;
317 	rc = rdma_resolve_addr(id, NULL, (struct sockaddr *)&xprt->addr,
318 			       RDMA_RESOLVE_TIMEOUT);
319 	if (rc)
320 		goto out;
321 	rc = wait_for_completion_interruptible_timeout(&ep->re_done, wtimeout);
322 	if (rc < 0)
323 		goto out;
324 
325 	rc = ep->re_async_rc;
326 	if (rc)
327 		goto out;
328 
329 	ep->re_async_rc = -ETIMEDOUT;
330 	rc = rdma_resolve_route(id, RDMA_RESOLVE_TIMEOUT);
331 	if (rc)
332 		goto out;
333 	rc = wait_for_completion_interruptible_timeout(&ep->re_done, wtimeout);
334 	if (rc < 0)
335 		goto out;
336 	rc = ep->re_async_rc;
337 	if (rc)
338 		goto out;
339 
340 	return id;
341 
342 out:
343 	rdma_destroy_id(id);
344 	return ERR_PTR(rc);
345 }
346 
347 static void rpcrdma_ep_destroy(struct kref *kref)
348 {
349 	struct rpcrdma_ep *ep = container_of(kref, struct rpcrdma_ep, re_kref);
350 
351 	if (ep->re_id->qp) {
352 		rdma_destroy_qp(ep->re_id);
353 		ep->re_id->qp = NULL;
354 	}
355 
356 	if (ep->re_attr.recv_cq)
357 		ib_free_cq(ep->re_attr.recv_cq);
358 	ep->re_attr.recv_cq = NULL;
359 	if (ep->re_attr.send_cq)
360 		ib_free_cq(ep->re_attr.send_cq);
361 	ep->re_attr.send_cq = NULL;
362 
363 	if (ep->re_pd)
364 		ib_dealloc_pd(ep->re_pd);
365 	ep->re_pd = NULL;
366 
367 	kfree(ep);
368 	module_put(THIS_MODULE);
369 }
370 
371 static noinline void rpcrdma_ep_get(struct rpcrdma_ep *ep)
372 {
373 	kref_get(&ep->re_kref);
374 }
375 
376 /* Returns:
377  *     %0 if @ep still has a positive kref count, or
378  *     %1 if @ep was destroyed successfully.
379  */
380 static noinline int rpcrdma_ep_put(struct rpcrdma_ep *ep)
381 {
382 	return kref_put(&ep->re_kref, rpcrdma_ep_destroy);
383 }
384 
385 static int rpcrdma_ep_create(struct rpcrdma_xprt *r_xprt)
386 {
387 	struct rpcrdma_connect_private *pmsg;
388 	struct ib_device *device;
389 	struct rdma_cm_id *id;
390 	struct rpcrdma_ep *ep;
391 	int rc;
392 
393 	ep = kzalloc(sizeof(*ep), GFP_NOFS);
394 	if (!ep)
395 		return -ENOTCONN;
396 	ep->re_xprt = &r_xprt->rx_xprt;
397 	kref_init(&ep->re_kref);
398 
399 	id = rpcrdma_create_id(r_xprt, ep);
400 	if (IS_ERR(id)) {
401 		kfree(ep);
402 		return PTR_ERR(id);
403 	}
404 	__module_get(THIS_MODULE);
405 	device = id->device;
406 	ep->re_id = id;
407 	reinit_completion(&ep->re_done);
408 
409 	ep->re_max_requests = r_xprt->rx_xprt.max_reqs;
410 	ep->re_inline_send = xprt_rdma_max_inline_write;
411 	ep->re_inline_recv = xprt_rdma_max_inline_read;
412 	rc = frwr_query_device(ep, device);
413 	if (rc)
414 		goto out_destroy;
415 
416 	r_xprt->rx_buf.rb_max_requests = cpu_to_be32(ep->re_max_requests);
417 
418 	ep->re_attr.srq = NULL;
419 	ep->re_attr.cap.max_inline_data = 0;
420 	ep->re_attr.sq_sig_type = IB_SIGNAL_REQ_WR;
421 	ep->re_attr.qp_type = IB_QPT_RC;
422 	ep->re_attr.port_num = ~0;
423 
424 	dprintk("RPC:       %s: requested max: dtos: send %d recv %d; "
425 		"iovs: send %d recv %d\n",
426 		__func__,
427 		ep->re_attr.cap.max_send_wr,
428 		ep->re_attr.cap.max_recv_wr,
429 		ep->re_attr.cap.max_send_sge,
430 		ep->re_attr.cap.max_recv_sge);
431 
432 	ep->re_send_batch = ep->re_max_requests >> 3;
433 	ep->re_send_count = ep->re_send_batch;
434 	init_waitqueue_head(&ep->re_connect_wait);
435 
436 	ep->re_attr.send_cq = ib_alloc_cq_any(device, r_xprt,
437 					      ep->re_attr.cap.max_send_wr,
438 					      IB_POLL_WORKQUEUE);
439 	if (IS_ERR(ep->re_attr.send_cq)) {
440 		rc = PTR_ERR(ep->re_attr.send_cq);
441 		goto out_destroy;
442 	}
443 
444 	ep->re_attr.recv_cq = ib_alloc_cq_any(device, r_xprt,
445 					      ep->re_attr.cap.max_recv_wr,
446 					      IB_POLL_WORKQUEUE);
447 	if (IS_ERR(ep->re_attr.recv_cq)) {
448 		rc = PTR_ERR(ep->re_attr.recv_cq);
449 		goto out_destroy;
450 	}
451 	ep->re_receive_count = 0;
452 
453 	/* Initialize cma parameters */
454 	memset(&ep->re_remote_cma, 0, sizeof(ep->re_remote_cma));
455 
456 	/* Prepare RDMA-CM private message */
457 	pmsg = &ep->re_cm_private;
458 	pmsg->cp_magic = rpcrdma_cmp_magic;
459 	pmsg->cp_version = RPCRDMA_CMP_VERSION;
460 	pmsg->cp_flags |= RPCRDMA_CMP_F_SND_W_INV_OK;
461 	pmsg->cp_send_size = rpcrdma_encode_buffer_size(ep->re_inline_send);
462 	pmsg->cp_recv_size = rpcrdma_encode_buffer_size(ep->re_inline_recv);
463 	ep->re_remote_cma.private_data = pmsg;
464 	ep->re_remote_cma.private_data_len = sizeof(*pmsg);
465 
466 	/* Client offers RDMA Read but does not initiate */
467 	ep->re_remote_cma.initiator_depth = 0;
468 	ep->re_remote_cma.responder_resources =
469 		min_t(int, U8_MAX, device->attrs.max_qp_rd_atom);
470 
471 	/* Limit transport retries so client can detect server
472 	 * GID changes quickly. RPC layer handles re-establishing
473 	 * transport connection and retransmission.
474 	 */
475 	ep->re_remote_cma.retry_count = 6;
476 
477 	/* RPC-over-RDMA handles its own flow control. In addition,
478 	 * make all RNR NAKs visible so we know that RPC-over-RDMA
479 	 * flow control is working correctly (no NAKs should be seen).
480 	 */
481 	ep->re_remote_cma.flow_control = 0;
482 	ep->re_remote_cma.rnr_retry_count = 0;
483 
484 	ep->re_pd = ib_alloc_pd(device, 0);
485 	if (IS_ERR(ep->re_pd)) {
486 		rc = PTR_ERR(ep->re_pd);
487 		goto out_destroy;
488 	}
489 
490 	rc = rdma_create_qp(id, ep->re_pd, &ep->re_attr);
491 	if (rc)
492 		goto out_destroy;
493 
494 	r_xprt->rx_ep = ep;
495 	return 0;
496 
497 out_destroy:
498 	rpcrdma_ep_put(ep);
499 	rdma_destroy_id(id);
500 	return rc;
501 }
502 
503 /**
504  * rpcrdma_xprt_connect - Connect an unconnected transport
505  * @r_xprt: controlling transport instance
506  *
507  * Returns 0 on success or a negative errno.
508  */
509 int rpcrdma_xprt_connect(struct rpcrdma_xprt *r_xprt)
510 {
511 	struct rpc_xprt *xprt = &r_xprt->rx_xprt;
512 	struct rpcrdma_ep *ep;
513 	int rc;
514 
515 	rc = rpcrdma_ep_create(r_xprt);
516 	if (rc)
517 		return rc;
518 	ep = r_xprt->rx_ep;
519 
520 	xprt_clear_connected(xprt);
521 	rpcrdma_reset_cwnd(r_xprt);
522 
523 	/* Bump the ep's reference count while there are
524 	 * outstanding Receives.
525 	 */
526 	rpcrdma_ep_get(ep);
527 	rpcrdma_post_recvs(r_xprt, 1, true);
528 
529 	rc = rdma_connect(ep->re_id, &ep->re_remote_cma);
530 	if (rc)
531 		goto out;
532 
533 	if (xprt->reestablish_timeout < RPCRDMA_INIT_REEST_TO)
534 		xprt->reestablish_timeout = RPCRDMA_INIT_REEST_TO;
535 	wait_event_interruptible(ep->re_connect_wait,
536 				 ep->re_connect_status != 0);
537 	if (ep->re_connect_status <= 0) {
538 		rc = ep->re_connect_status;
539 		goto out;
540 	}
541 
542 	rc = rpcrdma_sendctxs_create(r_xprt);
543 	if (rc) {
544 		rc = -ENOTCONN;
545 		goto out;
546 	}
547 
548 	rc = rpcrdma_reqs_setup(r_xprt);
549 	if (rc) {
550 		rc = -ENOTCONN;
551 		goto out;
552 	}
553 	rpcrdma_mrs_create(r_xprt);
554 
555 out:
556 	trace_xprtrdma_connect(r_xprt, rc);
557 	return rc;
558 }
559 
560 /**
561  * rpcrdma_xprt_disconnect - Disconnect underlying transport
562  * @r_xprt: controlling transport instance
563  *
564  * Caller serializes. Either the transport send lock is held,
565  * or we're being called to destroy the transport.
566  *
567  * On return, @r_xprt is completely divested of all hardware
568  * resources and prepared for the next ->connect operation.
569  */
570 void rpcrdma_xprt_disconnect(struct rpcrdma_xprt *r_xprt)
571 {
572 	struct rpcrdma_ep *ep = r_xprt->rx_ep;
573 	struct rdma_cm_id *id;
574 	int rc;
575 
576 	if (!ep)
577 		return;
578 
579 	id = ep->re_id;
580 	rc = rdma_disconnect(id);
581 	trace_xprtrdma_disconnect(r_xprt, rc);
582 
583 	rpcrdma_xprt_drain(r_xprt);
584 	rpcrdma_reps_unmap(r_xprt);
585 	rpcrdma_reqs_reset(r_xprt);
586 	rpcrdma_mrs_destroy(r_xprt);
587 	rpcrdma_sendctxs_destroy(r_xprt);
588 
589 	if (rpcrdma_ep_put(ep))
590 		rdma_destroy_id(id);
591 
592 	r_xprt->rx_ep = NULL;
593 }
594 
595 /* Fixed-size circular FIFO queue. This implementation is wait-free and
596  * lock-free.
597  *
598  * Consumer is the code path that posts Sends. This path dequeues a
599  * sendctx for use by a Send operation. Multiple consumer threads
600  * are serialized by the RPC transport lock, which allows only one
601  * ->send_request call at a time.
602  *
603  * Producer is the code path that handles Send completions. This path
604  * enqueues a sendctx that has been completed. Multiple producer
605  * threads are serialized by the ib_poll_cq() function.
606  */
607 
608 /* rpcrdma_sendctxs_destroy() assumes caller has already quiesced
609  * queue activity, and rpcrdma_xprt_drain has flushed all remaining
610  * Send requests.
611  */
612 static void rpcrdma_sendctxs_destroy(struct rpcrdma_xprt *r_xprt)
613 {
614 	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
615 	unsigned long i;
616 
617 	if (!buf->rb_sc_ctxs)
618 		return;
619 	for (i = 0; i <= buf->rb_sc_last; i++)
620 		kfree(buf->rb_sc_ctxs[i]);
621 	kfree(buf->rb_sc_ctxs);
622 	buf->rb_sc_ctxs = NULL;
623 }
624 
625 static struct rpcrdma_sendctx *rpcrdma_sendctx_create(struct rpcrdma_ep *ep)
626 {
627 	struct rpcrdma_sendctx *sc;
628 
629 	sc = kzalloc(struct_size(sc, sc_sges, ep->re_attr.cap.max_send_sge),
630 		     GFP_KERNEL);
631 	if (!sc)
632 		return NULL;
633 
634 	sc->sc_cqe.done = rpcrdma_wc_send;
635 	sc->sc_cid.ci_queue_id = ep->re_attr.send_cq->res.id;
636 	sc->sc_cid.ci_completion_id =
637 		atomic_inc_return(&ep->re_completion_ids);
638 	return sc;
639 }
640 
641 static int rpcrdma_sendctxs_create(struct rpcrdma_xprt *r_xprt)
642 {
643 	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
644 	struct rpcrdma_sendctx *sc;
645 	unsigned long i;
646 
647 	/* Maximum number of concurrent outstanding Send WRs. Capping
648 	 * the circular queue size stops Send Queue overflow by causing
649 	 * the ->send_request call to fail temporarily before too many
650 	 * Sends are posted.
651 	 */
652 	i = r_xprt->rx_ep->re_max_requests + RPCRDMA_MAX_BC_REQUESTS;
653 	buf->rb_sc_ctxs = kcalloc(i, sizeof(sc), GFP_KERNEL);
654 	if (!buf->rb_sc_ctxs)
655 		return -ENOMEM;
656 
657 	buf->rb_sc_last = i - 1;
658 	for (i = 0; i <= buf->rb_sc_last; i++) {
659 		sc = rpcrdma_sendctx_create(r_xprt->rx_ep);
660 		if (!sc)
661 			return -ENOMEM;
662 
663 		buf->rb_sc_ctxs[i] = sc;
664 	}
665 
666 	buf->rb_sc_head = 0;
667 	buf->rb_sc_tail = 0;
668 	return 0;
669 }
670 
671 /* The sendctx queue is not guaranteed to have a size that is a
672  * power of two, thus the helpers in circ_buf.h cannot be used.
673  * The other option is to use modulus (%), which can be expensive.
674  */
675 static unsigned long rpcrdma_sendctx_next(struct rpcrdma_buffer *buf,
676 					  unsigned long item)
677 {
678 	return likely(item < buf->rb_sc_last) ? item + 1 : 0;
679 }
680 
681 /**
682  * rpcrdma_sendctx_get_locked - Acquire a send context
683  * @r_xprt: controlling transport instance
684  *
685  * Returns pointer to a free send completion context; or NULL if
686  * the queue is empty.
687  *
688  * Usage: Called to acquire an SGE array before preparing a Send WR.
689  *
690  * The caller serializes calls to this function (per transport), and
691  * provides an effective memory barrier that flushes the new value
692  * of rb_sc_head.
693  */
694 struct rpcrdma_sendctx *rpcrdma_sendctx_get_locked(struct rpcrdma_xprt *r_xprt)
695 {
696 	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
697 	struct rpcrdma_sendctx *sc;
698 	unsigned long next_head;
699 
700 	next_head = rpcrdma_sendctx_next(buf, buf->rb_sc_head);
701 
702 	if (next_head == READ_ONCE(buf->rb_sc_tail))
703 		goto out_emptyq;
704 
705 	/* ORDER: item must be accessed _before_ head is updated */
706 	sc = buf->rb_sc_ctxs[next_head];
707 
708 	/* Releasing the lock in the caller acts as a memory
709 	 * barrier that flushes rb_sc_head.
710 	 */
711 	buf->rb_sc_head = next_head;
712 
713 	return sc;
714 
715 out_emptyq:
716 	/* The queue is "empty" if there have not been enough Send
717 	 * completions recently. This is a sign the Send Queue is
718 	 * backing up. Cause the caller to pause and try again.
719 	 */
720 	xprt_wait_for_buffer_space(&r_xprt->rx_xprt);
721 	r_xprt->rx_stats.empty_sendctx_q++;
722 	return NULL;
723 }
724 
725 /**
726  * rpcrdma_sendctx_put_locked - Release a send context
727  * @r_xprt: controlling transport instance
728  * @sc: send context to release
729  *
730  * Usage: Called from Send completion to return a sendctxt
731  * to the queue.
732  *
733  * The caller serializes calls to this function (per transport).
734  */
735 static void rpcrdma_sendctx_put_locked(struct rpcrdma_xprt *r_xprt,
736 				       struct rpcrdma_sendctx *sc)
737 {
738 	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
739 	unsigned long next_tail;
740 
741 	/* Unmap SGEs of previously completed but unsignaled
742 	 * Sends by walking up the queue until @sc is found.
743 	 */
744 	next_tail = buf->rb_sc_tail;
745 	do {
746 		next_tail = rpcrdma_sendctx_next(buf, next_tail);
747 
748 		/* ORDER: item must be accessed _before_ tail is updated */
749 		rpcrdma_sendctx_unmap(buf->rb_sc_ctxs[next_tail]);
750 
751 	} while (buf->rb_sc_ctxs[next_tail] != sc);
752 
753 	/* Paired with READ_ONCE */
754 	smp_store_release(&buf->rb_sc_tail, next_tail);
755 
756 	xprt_write_space(&r_xprt->rx_xprt);
757 }
758 
759 static void
760 rpcrdma_mrs_create(struct rpcrdma_xprt *r_xprt)
761 {
762 	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
763 	struct rpcrdma_ep *ep = r_xprt->rx_ep;
764 	unsigned int count;
765 
766 	for (count = 0; count < ep->re_max_rdma_segs; count++) {
767 		struct rpcrdma_mr *mr;
768 		int rc;
769 
770 		mr = kzalloc(sizeof(*mr), GFP_NOFS);
771 		if (!mr)
772 			break;
773 
774 		rc = frwr_mr_init(r_xprt, mr);
775 		if (rc) {
776 			kfree(mr);
777 			break;
778 		}
779 
780 		spin_lock(&buf->rb_lock);
781 		rpcrdma_mr_push(mr, &buf->rb_mrs);
782 		list_add(&mr->mr_all, &buf->rb_all_mrs);
783 		spin_unlock(&buf->rb_lock);
784 	}
785 
786 	r_xprt->rx_stats.mrs_allocated += count;
787 	trace_xprtrdma_createmrs(r_xprt, count);
788 }
789 
790 static void
791 rpcrdma_mr_refresh_worker(struct work_struct *work)
792 {
793 	struct rpcrdma_buffer *buf = container_of(work, struct rpcrdma_buffer,
794 						  rb_refresh_worker);
795 	struct rpcrdma_xprt *r_xprt = container_of(buf, struct rpcrdma_xprt,
796 						   rx_buf);
797 
798 	rpcrdma_mrs_create(r_xprt);
799 	xprt_write_space(&r_xprt->rx_xprt);
800 }
801 
802 /**
803  * rpcrdma_mrs_refresh - Wake the MR refresh worker
804  * @r_xprt: controlling transport instance
805  *
806  */
807 void rpcrdma_mrs_refresh(struct rpcrdma_xprt *r_xprt)
808 {
809 	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
810 	struct rpcrdma_ep *ep = r_xprt->rx_ep;
811 
812 	/* If there is no underlying connection, it's no use
813 	 * to wake the refresh worker.
814 	 */
815 	if (ep->re_connect_status == 1) {
816 		/* The work is scheduled on a WQ_MEM_RECLAIM
817 		 * workqueue in order to prevent MR allocation
818 		 * from recursing into NFS during direct reclaim.
819 		 */
820 		queue_work(xprtiod_workqueue, &buf->rb_refresh_worker);
821 	}
822 }
823 
824 /**
825  * rpcrdma_req_create - Allocate an rpcrdma_req object
826  * @r_xprt: controlling r_xprt
827  * @size: initial size, in bytes, of send and receive buffers
828  * @flags: GFP flags passed to memory allocators
829  *
830  * Returns an allocated and fully initialized rpcrdma_req or NULL.
831  */
832 struct rpcrdma_req *rpcrdma_req_create(struct rpcrdma_xprt *r_xprt, size_t size,
833 				       gfp_t flags)
834 {
835 	struct rpcrdma_buffer *buffer = &r_xprt->rx_buf;
836 	struct rpcrdma_req *req;
837 
838 	req = kzalloc(sizeof(*req), flags);
839 	if (req == NULL)
840 		goto out1;
841 
842 	req->rl_sendbuf = rpcrdma_regbuf_alloc(size, DMA_TO_DEVICE, flags);
843 	if (!req->rl_sendbuf)
844 		goto out2;
845 
846 	req->rl_recvbuf = rpcrdma_regbuf_alloc(size, DMA_NONE, flags);
847 	if (!req->rl_recvbuf)
848 		goto out3;
849 
850 	INIT_LIST_HEAD(&req->rl_free_mrs);
851 	INIT_LIST_HEAD(&req->rl_registered);
852 	spin_lock(&buffer->rb_lock);
853 	list_add(&req->rl_all, &buffer->rb_allreqs);
854 	spin_unlock(&buffer->rb_lock);
855 	return req;
856 
857 out3:
858 	kfree(req->rl_sendbuf);
859 out2:
860 	kfree(req);
861 out1:
862 	return NULL;
863 }
864 
865 /**
866  * rpcrdma_req_setup - Per-connection instance setup of an rpcrdma_req object
867  * @r_xprt: controlling transport instance
868  * @req: rpcrdma_req object to set up
869  *
870  * Returns zero on success, and a negative errno on failure.
871  */
872 int rpcrdma_req_setup(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
873 {
874 	struct rpcrdma_regbuf *rb;
875 	size_t maxhdrsize;
876 
877 	/* Compute maximum header buffer size in bytes */
878 	maxhdrsize = rpcrdma_fixed_maxsz + 3 +
879 		     r_xprt->rx_ep->re_max_rdma_segs * rpcrdma_readchunk_maxsz;
880 	maxhdrsize *= sizeof(__be32);
881 	rb = rpcrdma_regbuf_alloc(__roundup_pow_of_two(maxhdrsize),
882 				  DMA_TO_DEVICE, GFP_KERNEL);
883 	if (!rb)
884 		goto out;
885 
886 	if (!__rpcrdma_regbuf_dma_map(r_xprt, rb))
887 		goto out_free;
888 
889 	req->rl_rdmabuf = rb;
890 	xdr_buf_init(&req->rl_hdrbuf, rdmab_data(rb), rdmab_length(rb));
891 	return 0;
892 
893 out_free:
894 	rpcrdma_regbuf_free(rb);
895 out:
896 	return -ENOMEM;
897 }
898 
899 /* ASSUMPTION: the rb_allreqs list is stable for the duration,
900  * and thus can be walked without holding rb_lock. Eg. the
901  * caller is holding the transport send lock to exclude
902  * device removal or disconnection.
903  */
904 static int rpcrdma_reqs_setup(struct rpcrdma_xprt *r_xprt)
905 {
906 	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
907 	struct rpcrdma_req *req;
908 	int rc;
909 
910 	list_for_each_entry(req, &buf->rb_allreqs, rl_all) {
911 		rc = rpcrdma_req_setup(r_xprt, req);
912 		if (rc)
913 			return rc;
914 	}
915 	return 0;
916 }
917 
918 static void rpcrdma_req_reset(struct rpcrdma_req *req)
919 {
920 	/* Credits are valid for only one connection */
921 	req->rl_slot.rq_cong = 0;
922 
923 	rpcrdma_regbuf_free(req->rl_rdmabuf);
924 	req->rl_rdmabuf = NULL;
925 
926 	rpcrdma_regbuf_dma_unmap(req->rl_sendbuf);
927 	rpcrdma_regbuf_dma_unmap(req->rl_recvbuf);
928 
929 	frwr_reset(req);
930 }
931 
932 /* ASSUMPTION: the rb_allreqs list is stable for the duration,
933  * and thus can be walked without holding rb_lock. Eg. the
934  * caller is holding the transport send lock to exclude
935  * device removal or disconnection.
936  */
937 static void rpcrdma_reqs_reset(struct rpcrdma_xprt *r_xprt)
938 {
939 	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
940 	struct rpcrdma_req *req;
941 
942 	list_for_each_entry(req, &buf->rb_allreqs, rl_all)
943 		rpcrdma_req_reset(req);
944 }
945 
946 static noinline
947 struct rpcrdma_rep *rpcrdma_rep_create(struct rpcrdma_xprt *r_xprt,
948 				       bool temp)
949 {
950 	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
951 	struct rpcrdma_rep *rep;
952 
953 	rep = kzalloc(sizeof(*rep), GFP_KERNEL);
954 	if (rep == NULL)
955 		goto out;
956 
957 	rep->rr_rdmabuf = rpcrdma_regbuf_alloc(r_xprt->rx_ep->re_inline_recv,
958 					       DMA_FROM_DEVICE, GFP_KERNEL);
959 	if (!rep->rr_rdmabuf)
960 		goto out_free;
961 
962 	if (!rpcrdma_regbuf_dma_map(r_xprt, rep->rr_rdmabuf))
963 		goto out_free_regbuf;
964 
965 	rep->rr_cid.ci_completion_id =
966 		atomic_inc_return(&r_xprt->rx_ep->re_completion_ids);
967 
968 	xdr_buf_init(&rep->rr_hdrbuf, rdmab_data(rep->rr_rdmabuf),
969 		     rdmab_length(rep->rr_rdmabuf));
970 	rep->rr_cqe.done = rpcrdma_wc_receive;
971 	rep->rr_rxprt = r_xprt;
972 	rep->rr_recv_wr.next = NULL;
973 	rep->rr_recv_wr.wr_cqe = &rep->rr_cqe;
974 	rep->rr_recv_wr.sg_list = &rep->rr_rdmabuf->rg_iov;
975 	rep->rr_recv_wr.num_sge = 1;
976 	rep->rr_temp = temp;
977 
978 	spin_lock(&buf->rb_lock);
979 	list_add(&rep->rr_all, &buf->rb_all_reps);
980 	spin_unlock(&buf->rb_lock);
981 	return rep;
982 
983 out_free_regbuf:
984 	rpcrdma_regbuf_free(rep->rr_rdmabuf);
985 out_free:
986 	kfree(rep);
987 out:
988 	return NULL;
989 }
990 
991 static void rpcrdma_rep_free(struct rpcrdma_rep *rep)
992 {
993 	rpcrdma_regbuf_free(rep->rr_rdmabuf);
994 	kfree(rep);
995 }
996 
997 static void rpcrdma_rep_destroy(struct rpcrdma_rep *rep)
998 {
999 	struct rpcrdma_buffer *buf = &rep->rr_rxprt->rx_buf;
1000 
1001 	spin_lock(&buf->rb_lock);
1002 	list_del(&rep->rr_all);
1003 	spin_unlock(&buf->rb_lock);
1004 
1005 	rpcrdma_rep_free(rep);
1006 }
1007 
1008 static struct rpcrdma_rep *rpcrdma_rep_get_locked(struct rpcrdma_buffer *buf)
1009 {
1010 	struct llist_node *node;
1011 
1012 	/* Calls to llist_del_first are required to be serialized */
1013 	node = llist_del_first(&buf->rb_free_reps);
1014 	if (!node)
1015 		return NULL;
1016 	return llist_entry(node, struct rpcrdma_rep, rr_node);
1017 }
1018 
1019 /**
1020  * rpcrdma_rep_put - Release rpcrdma_rep back to free list
1021  * @buf: buffer pool
1022  * @rep: rep to release
1023  *
1024  */
1025 void rpcrdma_rep_put(struct rpcrdma_buffer *buf, struct rpcrdma_rep *rep)
1026 {
1027 	llist_add(&rep->rr_node, &buf->rb_free_reps);
1028 }
1029 
1030 /* Caller must ensure the QP is quiescent (RQ is drained) before
1031  * invoking this function, to guarantee rb_all_reps is not
1032  * changing.
1033  */
1034 static void rpcrdma_reps_unmap(struct rpcrdma_xprt *r_xprt)
1035 {
1036 	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
1037 	struct rpcrdma_rep *rep;
1038 
1039 	list_for_each_entry(rep, &buf->rb_all_reps, rr_all) {
1040 		rpcrdma_regbuf_dma_unmap(rep->rr_rdmabuf);
1041 		rep->rr_temp = true;	/* Mark this rep for destruction */
1042 	}
1043 }
1044 
1045 static void rpcrdma_reps_destroy(struct rpcrdma_buffer *buf)
1046 {
1047 	struct rpcrdma_rep *rep;
1048 
1049 	spin_lock(&buf->rb_lock);
1050 	while ((rep = list_first_entry_or_null(&buf->rb_all_reps,
1051 					       struct rpcrdma_rep,
1052 					       rr_all)) != NULL) {
1053 		list_del(&rep->rr_all);
1054 		spin_unlock(&buf->rb_lock);
1055 
1056 		rpcrdma_rep_free(rep);
1057 
1058 		spin_lock(&buf->rb_lock);
1059 	}
1060 	spin_unlock(&buf->rb_lock);
1061 }
1062 
1063 /**
1064  * rpcrdma_buffer_create - Create initial set of req/rep objects
1065  * @r_xprt: transport instance to (re)initialize
1066  *
1067  * Returns zero on success, otherwise a negative errno.
1068  */
1069 int rpcrdma_buffer_create(struct rpcrdma_xprt *r_xprt)
1070 {
1071 	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
1072 	int i, rc;
1073 
1074 	buf->rb_bc_srv_max_requests = 0;
1075 	spin_lock_init(&buf->rb_lock);
1076 	INIT_LIST_HEAD(&buf->rb_mrs);
1077 	INIT_LIST_HEAD(&buf->rb_all_mrs);
1078 	INIT_WORK(&buf->rb_refresh_worker, rpcrdma_mr_refresh_worker);
1079 
1080 	INIT_LIST_HEAD(&buf->rb_send_bufs);
1081 	INIT_LIST_HEAD(&buf->rb_allreqs);
1082 	INIT_LIST_HEAD(&buf->rb_all_reps);
1083 
1084 	rc = -ENOMEM;
1085 	for (i = 0; i < r_xprt->rx_xprt.max_reqs; i++) {
1086 		struct rpcrdma_req *req;
1087 
1088 		req = rpcrdma_req_create(r_xprt, RPCRDMA_V1_DEF_INLINE_SIZE * 2,
1089 					 GFP_KERNEL);
1090 		if (!req)
1091 			goto out;
1092 		list_add(&req->rl_list, &buf->rb_send_bufs);
1093 	}
1094 
1095 	init_llist_head(&buf->rb_free_reps);
1096 
1097 	return 0;
1098 out:
1099 	rpcrdma_buffer_destroy(buf);
1100 	return rc;
1101 }
1102 
1103 /**
1104  * rpcrdma_req_destroy - Destroy an rpcrdma_req object
1105  * @req: unused object to be destroyed
1106  *
1107  * Relies on caller holding the transport send lock to protect
1108  * removing req->rl_all from buf->rb_all_reqs safely.
1109  */
1110 void rpcrdma_req_destroy(struct rpcrdma_req *req)
1111 {
1112 	struct rpcrdma_mr *mr;
1113 
1114 	list_del(&req->rl_all);
1115 
1116 	while ((mr = rpcrdma_mr_pop(&req->rl_free_mrs))) {
1117 		struct rpcrdma_buffer *buf = &mr->mr_xprt->rx_buf;
1118 
1119 		spin_lock(&buf->rb_lock);
1120 		list_del(&mr->mr_all);
1121 		spin_unlock(&buf->rb_lock);
1122 
1123 		frwr_mr_release(mr);
1124 	}
1125 
1126 	rpcrdma_regbuf_free(req->rl_recvbuf);
1127 	rpcrdma_regbuf_free(req->rl_sendbuf);
1128 	rpcrdma_regbuf_free(req->rl_rdmabuf);
1129 	kfree(req);
1130 }
1131 
1132 /**
1133  * rpcrdma_mrs_destroy - Release all of a transport's MRs
1134  * @r_xprt: controlling transport instance
1135  *
1136  * Relies on caller holding the transport send lock to protect
1137  * removing mr->mr_list from req->rl_free_mrs safely.
1138  */
1139 static void rpcrdma_mrs_destroy(struct rpcrdma_xprt *r_xprt)
1140 {
1141 	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
1142 	struct rpcrdma_mr *mr;
1143 
1144 	cancel_work_sync(&buf->rb_refresh_worker);
1145 
1146 	spin_lock(&buf->rb_lock);
1147 	while ((mr = list_first_entry_or_null(&buf->rb_all_mrs,
1148 					      struct rpcrdma_mr,
1149 					      mr_all)) != NULL) {
1150 		list_del(&mr->mr_list);
1151 		list_del(&mr->mr_all);
1152 		spin_unlock(&buf->rb_lock);
1153 
1154 		frwr_mr_release(mr);
1155 
1156 		spin_lock(&buf->rb_lock);
1157 	}
1158 	spin_unlock(&buf->rb_lock);
1159 }
1160 
1161 /**
1162  * rpcrdma_buffer_destroy - Release all hw resources
1163  * @buf: root control block for resources
1164  *
1165  * ORDERING: relies on a prior rpcrdma_xprt_drain :
1166  * - No more Send or Receive completions can occur
1167  * - All MRs, reps, and reqs are returned to their free lists
1168  */
1169 void
1170 rpcrdma_buffer_destroy(struct rpcrdma_buffer *buf)
1171 {
1172 	rpcrdma_reps_destroy(buf);
1173 
1174 	while (!list_empty(&buf->rb_send_bufs)) {
1175 		struct rpcrdma_req *req;
1176 
1177 		req = list_first_entry(&buf->rb_send_bufs,
1178 				       struct rpcrdma_req, rl_list);
1179 		list_del(&req->rl_list);
1180 		rpcrdma_req_destroy(req);
1181 	}
1182 }
1183 
1184 /**
1185  * rpcrdma_mr_get - Allocate an rpcrdma_mr object
1186  * @r_xprt: controlling transport
1187  *
1188  * Returns an initialized rpcrdma_mr or NULL if no free
1189  * rpcrdma_mr objects are available.
1190  */
1191 struct rpcrdma_mr *
1192 rpcrdma_mr_get(struct rpcrdma_xprt *r_xprt)
1193 {
1194 	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
1195 	struct rpcrdma_mr *mr;
1196 
1197 	spin_lock(&buf->rb_lock);
1198 	mr = rpcrdma_mr_pop(&buf->rb_mrs);
1199 	spin_unlock(&buf->rb_lock);
1200 	return mr;
1201 }
1202 
1203 /**
1204  * rpcrdma_buffer_get - Get a request buffer
1205  * @buffers: Buffer pool from which to obtain a buffer
1206  *
1207  * Returns a fresh rpcrdma_req, or NULL if none are available.
1208  */
1209 struct rpcrdma_req *
1210 rpcrdma_buffer_get(struct rpcrdma_buffer *buffers)
1211 {
1212 	struct rpcrdma_req *req;
1213 
1214 	spin_lock(&buffers->rb_lock);
1215 	req = list_first_entry_or_null(&buffers->rb_send_bufs,
1216 				       struct rpcrdma_req, rl_list);
1217 	if (req)
1218 		list_del_init(&req->rl_list);
1219 	spin_unlock(&buffers->rb_lock);
1220 	return req;
1221 }
1222 
1223 /**
1224  * rpcrdma_buffer_put - Put request/reply buffers back into pool
1225  * @buffers: buffer pool
1226  * @req: object to return
1227  *
1228  */
1229 void rpcrdma_buffer_put(struct rpcrdma_buffer *buffers, struct rpcrdma_req *req)
1230 {
1231 	if (req->rl_reply)
1232 		rpcrdma_rep_put(buffers, req->rl_reply);
1233 	req->rl_reply = NULL;
1234 
1235 	spin_lock(&buffers->rb_lock);
1236 	list_add(&req->rl_list, &buffers->rb_send_bufs);
1237 	spin_unlock(&buffers->rb_lock);
1238 }
1239 
1240 /* Returns a pointer to a rpcrdma_regbuf object, or NULL.
1241  *
1242  * xprtrdma uses a regbuf for posting an outgoing RDMA SEND, or for
1243  * receiving the payload of RDMA RECV operations. During Long Calls
1244  * or Replies they may be registered externally via frwr_map.
1245  */
1246 static struct rpcrdma_regbuf *
1247 rpcrdma_regbuf_alloc(size_t size, enum dma_data_direction direction,
1248 		     gfp_t flags)
1249 {
1250 	struct rpcrdma_regbuf *rb;
1251 
1252 	rb = kmalloc(sizeof(*rb), flags);
1253 	if (!rb)
1254 		return NULL;
1255 	rb->rg_data = kmalloc(size, flags);
1256 	if (!rb->rg_data) {
1257 		kfree(rb);
1258 		return NULL;
1259 	}
1260 
1261 	rb->rg_device = NULL;
1262 	rb->rg_direction = direction;
1263 	rb->rg_iov.length = size;
1264 	return rb;
1265 }
1266 
1267 /**
1268  * rpcrdma_regbuf_realloc - re-allocate a SEND/RECV buffer
1269  * @rb: regbuf to reallocate
1270  * @size: size of buffer to be allocated, in bytes
1271  * @flags: GFP flags
1272  *
1273  * Returns true if reallocation was successful. If false is
1274  * returned, @rb is left untouched.
1275  */
1276 bool rpcrdma_regbuf_realloc(struct rpcrdma_regbuf *rb, size_t size, gfp_t flags)
1277 {
1278 	void *buf;
1279 
1280 	buf = kmalloc(size, flags);
1281 	if (!buf)
1282 		return false;
1283 
1284 	rpcrdma_regbuf_dma_unmap(rb);
1285 	kfree(rb->rg_data);
1286 
1287 	rb->rg_data = buf;
1288 	rb->rg_iov.length = size;
1289 	return true;
1290 }
1291 
1292 /**
1293  * __rpcrdma_regbuf_dma_map - DMA-map a regbuf
1294  * @r_xprt: controlling transport instance
1295  * @rb: regbuf to be mapped
1296  *
1297  * Returns true if the buffer is now DMA mapped to @r_xprt's device
1298  */
1299 bool __rpcrdma_regbuf_dma_map(struct rpcrdma_xprt *r_xprt,
1300 			      struct rpcrdma_regbuf *rb)
1301 {
1302 	struct ib_device *device = r_xprt->rx_ep->re_id->device;
1303 
1304 	if (rb->rg_direction == DMA_NONE)
1305 		return false;
1306 
1307 	rb->rg_iov.addr = ib_dma_map_single(device, rdmab_data(rb),
1308 					    rdmab_length(rb), rb->rg_direction);
1309 	if (ib_dma_mapping_error(device, rdmab_addr(rb))) {
1310 		trace_xprtrdma_dma_maperr(rdmab_addr(rb));
1311 		return false;
1312 	}
1313 
1314 	rb->rg_device = device;
1315 	rb->rg_iov.lkey = r_xprt->rx_ep->re_pd->local_dma_lkey;
1316 	return true;
1317 }
1318 
1319 static void rpcrdma_regbuf_dma_unmap(struct rpcrdma_regbuf *rb)
1320 {
1321 	if (!rb)
1322 		return;
1323 
1324 	if (!rpcrdma_regbuf_is_mapped(rb))
1325 		return;
1326 
1327 	ib_dma_unmap_single(rb->rg_device, rdmab_addr(rb), rdmab_length(rb),
1328 			    rb->rg_direction);
1329 	rb->rg_device = NULL;
1330 }
1331 
1332 static void rpcrdma_regbuf_free(struct rpcrdma_regbuf *rb)
1333 {
1334 	rpcrdma_regbuf_dma_unmap(rb);
1335 	if (rb)
1336 		kfree(rb->rg_data);
1337 	kfree(rb);
1338 }
1339 
1340 /**
1341  * rpcrdma_post_sends - Post WRs to a transport's Send Queue
1342  * @r_xprt: controlling transport instance
1343  * @req: rpcrdma_req containing the Send WR to post
1344  *
1345  * Returns 0 if the post was successful, otherwise -ENOTCONN
1346  * is returned.
1347  */
1348 int rpcrdma_post_sends(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
1349 {
1350 	if (frwr_send(r_xprt, req))
1351 		return -ENOTCONN;
1352 	return 0;
1353 }
1354 
1355 /**
1356  * rpcrdma_post_recvs - Refill the Receive Queue
1357  * @r_xprt: controlling transport instance
1358  * @needed: current credit grant
1359  * @temp: mark Receive buffers to be deleted after one use
1360  *
1361  */
1362 void rpcrdma_post_recvs(struct rpcrdma_xprt *r_xprt, int needed, bool temp)
1363 {
1364 	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
1365 	struct rpcrdma_ep *ep = r_xprt->rx_ep;
1366 	struct ib_recv_wr *wr, *bad_wr;
1367 	struct rpcrdma_rep *rep;
1368 	int count, rc;
1369 
1370 	rc = 0;
1371 	count = 0;
1372 
1373 	if (likely(ep->re_receive_count > needed))
1374 		goto out;
1375 	needed -= ep->re_receive_count;
1376 	if (!temp)
1377 		needed += RPCRDMA_MAX_RECV_BATCH;
1378 
1379 	if (atomic_inc_return(&ep->re_receiving) > 1)
1380 		goto out;
1381 
1382 	/* fast path: all needed reps can be found on the free list */
1383 	wr = NULL;
1384 	while (needed) {
1385 		rep = rpcrdma_rep_get_locked(buf);
1386 		if (rep && rep->rr_temp) {
1387 			rpcrdma_rep_destroy(rep);
1388 			continue;
1389 		}
1390 		if (!rep)
1391 			rep = rpcrdma_rep_create(r_xprt, temp);
1392 		if (!rep)
1393 			break;
1394 
1395 		rep->rr_cid.ci_queue_id = ep->re_attr.recv_cq->res.id;
1396 		trace_xprtrdma_post_recv(rep);
1397 		rep->rr_recv_wr.next = wr;
1398 		wr = &rep->rr_recv_wr;
1399 		--needed;
1400 		++count;
1401 	}
1402 	if (!wr)
1403 		goto out;
1404 
1405 	rc = ib_post_recv(ep->re_id->qp, wr,
1406 			  (const struct ib_recv_wr **)&bad_wr);
1407 	if (atomic_dec_return(&ep->re_receiving) > 0)
1408 		complete(&ep->re_done);
1409 
1410 out:
1411 	trace_xprtrdma_post_recvs(r_xprt, count, rc);
1412 	if (rc) {
1413 		for (wr = bad_wr; wr;) {
1414 			struct rpcrdma_rep *rep;
1415 
1416 			rep = container_of(wr, struct rpcrdma_rep, rr_recv_wr);
1417 			wr = wr->next;
1418 			rpcrdma_rep_put(buf, rep);
1419 			--count;
1420 		}
1421 	}
1422 	ep->re_receive_count += count;
1423 	return;
1424 }
1425