verbs.c 52.2 KB
Newer Older
1
/*
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37
 * Copyright (c) 2003-2007 Network Appliance, Inc. All rights reserved.
 *
 * This software is available to you under a choice of one of two
 * licenses.  You may choose to be licensed under the terms of the GNU
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the BSD-type
 * license below:
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 *
 *      Redistributions of source code must retain the above copyright
 *      notice, this list of conditions and the following disclaimer.
 *
 *      Redistributions in binary form must reproduce the above
 *      copyright notice, this list of conditions and the following
 *      disclaimer in the documentation and/or other materials provided
 *      with the distribution.
 *
 *      Neither the name of the Network Appliance, Inc. nor the names of
 *      its contributors may be used to endorse or promote products
 *      derived from this software without specific prior written
 *      permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
38 39
 */

40 41 42 43 44 45 46 47 48 49
/*
 * verbs.c
 *
 * Encapsulates the major functions managing:
 *  o adapters
 *  o endpoints
 *  o connections
 *  o buffer memory
 */

50
#include <linux/interrupt.h>
51
#include <linux/slab.h>
52
#include <asm/bitops.h>
53

54 55
#include "xprt_rdma.h"

56 57 58 59 60 61 62 63
/*
 * Globals/Macros
 */

#ifdef RPC_DEBUG
# define RPCDBG_FACILITY	RPCDBG_TRANS
#endif

64 65
static void rpcrdma_reset_frmrs(struct rpcrdma_ia *);

66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146
/*
 * internal functions
 */

/*
 * handle replies in tasklet context, using a single, global list
 * rdma tasklet function -- just turn around and call the func
 * for all replies on the list
 */

static DEFINE_SPINLOCK(rpcrdma_tk_lock_g);
static LIST_HEAD(rpcrdma_tasklets_g);

static void
rpcrdma_run_tasklet(unsigned long data)
{
	struct rpcrdma_rep *rep;
	void (*func)(struct rpcrdma_rep *);
	unsigned long flags;

	data = data;
	spin_lock_irqsave(&rpcrdma_tk_lock_g, flags);
	while (!list_empty(&rpcrdma_tasklets_g)) {
		rep = list_entry(rpcrdma_tasklets_g.next,
				 struct rpcrdma_rep, rr_list);
		list_del(&rep->rr_list);
		func = rep->rr_func;
		rep->rr_func = NULL;
		spin_unlock_irqrestore(&rpcrdma_tk_lock_g, flags);

		if (func)
			func(rep);
		else
			rpcrdma_recv_buffer_put(rep);

		spin_lock_irqsave(&rpcrdma_tk_lock_g, flags);
	}
	spin_unlock_irqrestore(&rpcrdma_tk_lock_g, flags);
}

static DECLARE_TASKLET(rpcrdma_tasklet_g, rpcrdma_run_tasklet, 0UL);

static inline void
rpcrdma_schedule_tasklet(struct rpcrdma_rep *rep)
{
	unsigned long flags;

	spin_lock_irqsave(&rpcrdma_tk_lock_g, flags);
	list_add_tail(&rep->rr_list, &rpcrdma_tasklets_g);
	spin_unlock_irqrestore(&rpcrdma_tk_lock_g, flags);
	tasklet_schedule(&rpcrdma_tasklet_g);
}

static void
rpcrdma_qp_async_error_upcall(struct ib_event *event, void *context)
{
	struct rpcrdma_ep *ep = context;

	dprintk("RPC:       %s: QP error %X on device %s ep %p\n",
		__func__, event->event, event->device->name, context);
	if (ep->rep_connected == 1) {
		ep->rep_connected = -EIO;
		ep->rep_func(ep);
		wake_up_all(&ep->rep_connect_wait);
	}
}

static void
rpcrdma_cq_async_error_upcall(struct ib_event *event, void *context)
{
	struct rpcrdma_ep *ep = context;

	dprintk("RPC:       %s: CQ error %X on device %s ep %p\n",
		__func__, event->event, event->device->name, context);
	if (ep->rep_connected == 1) {
		ep->rep_connected = -EIO;
		ep->rep_func(ep);
		wake_up_all(&ep->rep_connect_wait);
	}
}

147 148
static void
rpcrdma_sendcq_process_wc(struct ib_wc *wc)
149
{
150
	struct rpcrdma_mw *frmr = (struct rpcrdma_mw *)(unsigned long)wc->wr_id;
151

152 153
	dprintk("RPC:       %s: frmr %p status %X opcode %d\n",
		__func__, frmr, wc->status, wc->opcode);
154

155
	if (wc->wr_id == 0ULL)
156
		return;
157 158
	if (wc->status != IB_WC_SUCCESS) {
		frmr->r.frmr.fr_state = FRMR_IS_STALE;
159
		return;
160
	}
161

162
	frmr->r.frmr.fr_state = FRMR_IS_INVALID;
163 164
}

165
static int
166
rpcrdma_sendcq_poll(struct ib_cq *cq, struct rpcrdma_ep *ep)
167
{
168
	struct ib_wc *wcs;
169
	int budget, count, rc;
170

171
	budget = RPCRDMA_WC_BUDGET / RPCRDMA_POLLSIZE;
172 173 174 175 176 177 178 179 180 181
	do {
		wcs = ep->rep_send_wcs;

		rc = ib_poll_cq(cq, RPCRDMA_POLLSIZE, wcs);
		if (rc <= 0)
			return rc;

		count = rc;
		while (count-- > 0)
			rpcrdma_sendcq_process_wc(wcs++);
182
	} while (rc == RPCRDMA_POLLSIZE && --budget);
183
	return 0;
184
}
185

186 187 188 189 190 191 192 193 194 195 196
/*
 * Handle send, fast_reg_mr, and local_inv completions.
 *
 * Send events are typically suppressed and thus do not result
 * in an upcall. Occasionally one is signaled, however. This
 * prevents the provider's completion queue from wrapping and
 * losing a completion.
 */
static void
rpcrdma_sendcq_upcall(struct ib_cq *cq, void *cq_context)
{
197
	struct rpcrdma_ep *ep = (struct rpcrdma_ep *)cq_context;
198 199
	int rc;

200
	rc = rpcrdma_sendcq_poll(cq, ep);
201 202 203 204
	if (rc) {
		dprintk("RPC:       %s: ib_poll_cq failed: %i\n",
			__func__, rc);
		return;
205 206
	}

207 208 209 210 211
	rc = ib_req_notify_cq(cq,
			IB_CQ_NEXT_COMP | IB_CQ_REPORT_MISSED_EVENTS);
	if (rc == 0)
		return;
	if (rc < 0) {
212 213 214 215 216
		dprintk("RPC:       %s: ib_req_notify_cq failed: %i\n",
			__func__, rc);
		return;
	}

217
	rpcrdma_sendcq_poll(cq, ep);
218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255
}

static void
rpcrdma_recvcq_process_wc(struct ib_wc *wc)
{
	struct rpcrdma_rep *rep =
			(struct rpcrdma_rep *)(unsigned long)wc->wr_id;

	dprintk("RPC:       %s: rep %p status %X opcode %X length %u\n",
		__func__, rep, wc->status, wc->opcode, wc->byte_len);

	if (wc->status != IB_WC_SUCCESS) {
		rep->rr_len = ~0U;
		goto out_schedule;
	}
	if (wc->opcode != IB_WC_RECV)
		return;

	rep->rr_len = wc->byte_len;
	ib_dma_sync_single_for_cpu(rdmab_to_ia(rep->rr_buffer)->ri_id->device,
			rep->rr_iov.addr, rep->rr_len, DMA_FROM_DEVICE);

	if (rep->rr_len >= 16) {
		struct rpcrdma_msg *p = (struct rpcrdma_msg *)rep->rr_base;
		unsigned int credits = ntohl(p->rm_credit);

		if (credits == 0)
			credits = 1;	/* don't deadlock */
		else if (credits > rep->rr_buffer->rb_max_requests)
			credits = rep->rr_buffer->rb_max_requests;
		atomic_set(&rep->rr_buffer->rb_credits, credits);
	}

out_schedule:
	rpcrdma_schedule_tasklet(rep);
}

static int
256
rpcrdma_recvcq_poll(struct ib_cq *cq, struct rpcrdma_ep *ep)
257
{
258
	struct ib_wc *wcs;
259
	int budget, count, rc;
260

261
	budget = RPCRDMA_WC_BUDGET / RPCRDMA_POLLSIZE;
262 263 264 265 266 267 268 269 270 271
	do {
		wcs = ep->rep_recv_wcs;

		rc = ib_poll_cq(cq, RPCRDMA_POLLSIZE, wcs);
		if (rc <= 0)
			return rc;

		count = rc;
		while (count-- > 0)
			rpcrdma_recvcq_process_wc(wcs++);
272
	} while (rc == RPCRDMA_POLLSIZE && --budget);
273
	return 0;
274 275 276
}

/*
277
 * Handle receive completions.
278 279 280 281 282 283 284 285 286 287 288
 *
 * It is reentrant but processes single events in order to maintain
 * ordering of receives to keep server credits.
 *
 * It is the responsibility of the scheduled tasklet to return
 * recv buffers to the pool. NOTE: this affects synchronization of
 * connection shutdown. That is, the structures required for
 * the completion of the reply handler must remain intact until
 * all memory has been reclaimed.
 */
static void
289
rpcrdma_recvcq_upcall(struct ib_cq *cq, void *cq_context)
290
{
291
	struct rpcrdma_ep *ep = (struct rpcrdma_ep *)cq_context;
292 293
	int rc;

294
	rc = rpcrdma_recvcq_poll(cq, ep);
295 296 297
	if (rc) {
		dprintk("RPC:       %s: ib_poll_cq failed: %i\n",
			__func__, rc);
298
		return;
299
	}
300

301 302 303 304 305
	rc = ib_req_notify_cq(cq,
			IB_CQ_NEXT_COMP | IB_CQ_REPORT_MISSED_EVENTS);
	if (rc == 0)
		return;
	if (rc < 0) {
306
		dprintk("RPC:       %s: ib_req_notify_cq failed: %i\n",
307 308 309 310
			__func__, rc);
		return;
	}

311
	rpcrdma_recvcq_poll(cq, ep);
312 313
}

314 315 316 317 318 319 320
static void
rpcrdma_flush_cqs(struct rpcrdma_ep *ep)
{
	rpcrdma_recvcq_upcall(ep->rep_attr.recv_cq, ep);
	rpcrdma_sendcq_upcall(ep->rep_attr.send_cq, ep);
}

321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343
#ifdef RPC_DEBUG
static const char * const conn[] = {
	"address resolved",
	"address error",
	"route resolved",
	"route error",
	"connect request",
	"connect response",
	"connect error",
	"unreachable",
	"rejected",
	"established",
	"disconnected",
	"device removal"
};
#endif

static int
rpcrdma_conn_upcall(struct rdma_cm_id *id, struct rdma_cm_event *event)
{
	struct rpcrdma_xprt *xprt = id->context;
	struct rpcrdma_ia *ia = &xprt->rx_ia;
	struct rpcrdma_ep *ep = &xprt->rx_ep;
344
#ifdef RPC_DEBUG
345
	struct sockaddr_in *addr = (struct sockaddr_in *) &ep->rep_remote_addr;
346
#endif
347 348 349 350 351 352 353
	struct ib_qp_attr attr;
	struct ib_qp_init_attr iattr;
	int connstate = 0;

	switch (event->event) {
	case RDMA_CM_EVENT_ADDR_RESOLVED:
	case RDMA_CM_EVENT_ROUTE_RESOLVED:
354
		ia->ri_async_rc = 0;
355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392
		complete(&ia->ri_done);
		break;
	case RDMA_CM_EVENT_ADDR_ERROR:
		ia->ri_async_rc = -EHOSTUNREACH;
		dprintk("RPC:       %s: CM address resolution error, ep 0x%p\n",
			__func__, ep);
		complete(&ia->ri_done);
		break;
	case RDMA_CM_EVENT_ROUTE_ERROR:
		ia->ri_async_rc = -ENETUNREACH;
		dprintk("RPC:       %s: CM route resolution error, ep 0x%p\n",
			__func__, ep);
		complete(&ia->ri_done);
		break;
	case RDMA_CM_EVENT_ESTABLISHED:
		connstate = 1;
		ib_query_qp(ia->ri_id->qp, &attr,
			IB_QP_MAX_QP_RD_ATOMIC | IB_QP_MAX_DEST_RD_ATOMIC,
			&iattr);
		dprintk("RPC:       %s: %d responder resources"
			" (%d initiator)\n",
			__func__, attr.max_dest_rd_atomic, attr.max_rd_atomic);
		goto connected;
	case RDMA_CM_EVENT_CONNECT_ERROR:
		connstate = -ENOTCONN;
		goto connected;
	case RDMA_CM_EVENT_UNREACHABLE:
		connstate = -ENETDOWN;
		goto connected;
	case RDMA_CM_EVENT_REJECTED:
		connstate = -ECONNREFUSED;
		goto connected;
	case RDMA_CM_EVENT_DISCONNECTED:
		connstate = -ECONNABORTED;
		goto connected;
	case RDMA_CM_EVENT_DEVICE_REMOVAL:
		connstate = -ENODEV;
connected:
H
Harvey Harrison 已提交
393
		dprintk("RPC:       %s: %s: %pI4:%u (ep 0x%p event 0x%x)\n",
394 395 396
			__func__,
			(event->event <= 11) ? conn[event->event] :
						"unknown connection error",
H
Harvey Harrison 已提交
397
			&addr->sin_addr.s_addr,
398 399 400 401 402 403 404 405 406 407
			ntohs(addr->sin_port),
			ep, event->event);
		atomic_set(&rpcx_to_rdmax(ep->rep_xprt)->rx_buf.rb_credits, 1);
		dprintk("RPC:       %s: %sconnected\n",
					__func__, connstate > 0 ? "" : "dis");
		ep->rep_connected = connstate;
		ep->rep_func(ep);
		wake_up_all(&ep->rep_connect_wait);
		break;
	default:
408
		dprintk("RPC:       %s: unexpected CM event %d\n",
409 410 411 412
			__func__, event->event);
		break;
	}

413 414 415 416
#ifdef RPC_DEBUG
	if (connstate == 1) {
		int ird = attr.max_dest_rd_atomic;
		int tird = ep->rep_remote_cma.responder_resources;
H
Harvey Harrison 已提交
417
		printk(KERN_INFO "rpcrdma: connection to %pI4:%u "
418
			"on %s, memreg %d slots %d ird %d%s\n",
H
Harvey Harrison 已提交
419
			&addr->sin_addr.s_addr,
420 421 422 423 424 425
			ntohs(addr->sin_port),
			ia->ri_id->device->name,
			ia->ri_memreg_strategy,
			xprt->rx_buf.rb_max_requests,
			ird, ird < 4 && ird < tird / 2 ? " (low!)" : "");
	} else if (connstate < 0) {
H
Harvey Harrison 已提交
426 427
		printk(KERN_INFO "rpcrdma: connection to %pI4:%u closed (%d)\n",
			&addr->sin_addr.s_addr,
428 429 430 431 432
			ntohs(addr->sin_port),
			connstate);
	}
#endif

433 434 435 436 437 438 439 440 441 442
	return 0;
}

static struct rdma_cm_id *
rpcrdma_create_id(struct rpcrdma_xprt *xprt,
			struct rpcrdma_ia *ia, struct sockaddr *addr)
{
	struct rdma_cm_id *id;
	int rc;

443 444
	init_completion(&ia->ri_done);

445
	id = rdma_create_id(rpcrdma_conn_upcall, xprt, RDMA_PS_TCP, IB_QPT_RC);
446 447 448 449 450 451 452
	if (IS_ERR(id)) {
		rc = PTR_ERR(id);
		dprintk("RPC:       %s: rdma_create_id() failed %i\n",
			__func__, rc);
		return id;
	}

453
	ia->ri_async_rc = -ETIMEDOUT;
454 455 456 457 458 459
	rc = rdma_resolve_addr(id, NULL, addr, RDMA_RESOLVE_TIMEOUT);
	if (rc) {
		dprintk("RPC:       %s: rdma_resolve_addr() failed %i\n",
			__func__, rc);
		goto out;
	}
460 461
	wait_for_completion_interruptible_timeout(&ia->ri_done,
				msecs_to_jiffies(RDMA_RESOLVE_TIMEOUT) + 1);
462 463 464 465
	rc = ia->ri_async_rc;
	if (rc)
		goto out;

466
	ia->ri_async_rc = -ETIMEDOUT;
467 468 469 470 471 472
	rc = rdma_resolve_route(id, RDMA_RESOLVE_TIMEOUT);
	if (rc) {
		dprintk("RPC:       %s: rdma_resolve_route() failed %i\n",
			__func__, rc);
		goto out;
	}
473 474
	wait_for_completion_interruptible_timeout(&ia->ri_done,
				msecs_to_jiffies(RDMA_RESOLVE_TIMEOUT) + 1);
475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514
	rc = ia->ri_async_rc;
	if (rc)
		goto out;

	return id;

out:
	rdma_destroy_id(id);
	return ERR_PTR(rc);
}

/*
 * Drain any cq, prior to teardown.
 */
static void
rpcrdma_clean_cq(struct ib_cq *cq)
{
	struct ib_wc wc;
	int count = 0;

	while (1 == ib_poll_cq(cq, 1, &wc))
		++count;

	if (count)
		dprintk("RPC:       %s: flushed %d events (last 0x%x)\n",
			__func__, count, wc.opcode);
}

/*
 * Exported functions.
 */

/*
 * Open and initialize an Interface Adapter.
 *  o initializes fields of struct rpcrdma_ia, including
 *    interface and provider attributes and protection zone.
 */
int
rpcrdma_ia_open(struct rpcrdma_xprt *xprt, struct sockaddr *addr, int memreg)
{
515 516
	int rc, mem_priv;
	struct ib_device_attr devattr;
517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532
	struct rpcrdma_ia *ia = &xprt->rx_ia;

	ia->ri_id = rpcrdma_create_id(xprt, ia, addr);
	if (IS_ERR(ia->ri_id)) {
		rc = PTR_ERR(ia->ri_id);
		goto out1;
	}

	ia->ri_pd = ib_alloc_pd(ia->ri_id->device);
	if (IS_ERR(ia->ri_pd)) {
		rc = PTR_ERR(ia->ri_pd);
		dprintk("RPC:       %s: ib_alloc_pd() failed %i\n",
			__func__, rc);
		goto out2;
	}

533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549
	/*
	 * Query the device to determine if the requested memory
	 * registration strategy is supported. If it isn't, set the
	 * strategy to a globally supported model.
	 */
	rc = ib_query_device(ia->ri_id->device, &devattr);
	if (rc) {
		dprintk("RPC:       %s: ib_query_device failed %d\n",
			__func__, rc);
		goto out2;
	}

	if (devattr.device_cap_flags & IB_DEVICE_LOCAL_DMA_LKEY) {
		ia->ri_have_dma_lkey = 1;
		ia->ri_dma_lkey = ia->ri_id->device->local_dma_lkey;
	}

550
	if (memreg == RPCRDMA_FRMR) {
551 552 553 554 555
		/* Requires both frmr reg and local dma lkey */
		if ((devattr.device_cap_flags &
		     (IB_DEVICE_MEM_MGT_EXTENSIONS|IB_DEVICE_LOCAL_DMA_LKEY)) !=
		    (IB_DEVICE_MEM_MGT_EXTENSIONS|IB_DEVICE_LOCAL_DMA_LKEY)) {
			dprintk("RPC:       %s: FRMR registration "
556 557
				"not supported by HCA\n", __func__);
			memreg = RPCRDMA_MTHCAFMR;
558 559 560 561 562
		} else {
			/* Mind the ia limit on FRMR page list depth */
			ia->ri_max_frmr_depth = min_t(unsigned int,
				RPCRDMA_MAX_DATA_SEGS,
				devattr.max_fast_reg_page_list_len);
563
		}
564 565 566 567 568 569 570 571
	}
	if (memreg == RPCRDMA_MTHCAFMR) {
		if (!ia->ri_id->device->alloc_fmr) {
			dprintk("RPC:       %s: MTHCAFMR registration "
				"not supported by HCA\n", __func__);
#if RPCRDMA_PERSISTENT_REGISTRATION
			memreg = RPCRDMA_ALLPHYSICAL;
#else
572
			rc = -ENOMEM;
573 574 575
			goto out2;
#endif
		}
576 577
	}

578 579 580 581 582 583 584 585
	/*
	 * Optionally obtain an underlying physical identity mapping in
	 * order to do a memory window-based bind. This base registration
	 * is protected from remote access - that is enabled only by binding
	 * for the specific bytes targeted during each RPC operation, and
	 * revoked after the corresponding completion similar to a storage
	 * adapter.
	 */
586
	switch (memreg) {
587
	case RPCRDMA_FRMR:
588
		break;
589
#if RPCRDMA_PERSISTENT_REGISTRATION
590 591 592 593 594
	case RPCRDMA_ALLPHYSICAL:
		mem_priv = IB_ACCESS_LOCAL_WRITE |
				IB_ACCESS_REMOTE_WRITE |
				IB_ACCESS_REMOTE_READ;
		goto register_setup;
595
#endif
596 597
	case RPCRDMA_MTHCAFMR:
		if (ia->ri_have_dma_lkey)
598
			break;
599
		mem_priv = IB_ACCESS_LOCAL_WRITE;
600
#if RPCRDMA_PERSISTENT_REGISTRATION
601
	register_setup:
602
#endif
603 604 605
		ia->ri_bind_mem = ib_get_dma_mr(ia->ri_pd, mem_priv);
		if (IS_ERR(ia->ri_bind_mem)) {
			printk(KERN_ALERT "%s: ib_get_dma_mr for "
606
				"phys register failed with %lX\n",
607
				__func__, PTR_ERR(ia->ri_bind_mem));
608 609
			rc = -ENOMEM;
			goto out2;
610
		}
611 612
		break;
	default:
613 614 615
		printk(KERN_ERR "RPC: Unsupported memory "
				"registration mode: %d\n", memreg);
		rc = -ENOMEM;
616
		goto out2;
617
	}
618 619
	dprintk("RPC:       %s: memory registration strategy is %d\n",
		__func__, memreg);
620 621 622 623

	/* Else will do memory reg/dereg for each chunk */
	ia->ri_memreg_strategy = memreg;

624
	rwlock_init(&ia->ri_qplock);
625 626 627
	return 0;
out2:
	rdma_destroy_id(ia->ri_id);
628
	ia->ri_id = NULL;
629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648
out1:
	return rc;
}

/*
 * Clean up/close an IA.
 *   o if event handles and PD have been initialized, free them.
 *   o close the IA
 */
void
rpcrdma_ia_close(struct rpcrdma_ia *ia)
{
	int rc;

	dprintk("RPC:       %s: entering\n", __func__);
	if (ia->ri_bind_mem != NULL) {
		rc = ib_dereg_mr(ia->ri_bind_mem);
		dprintk("RPC:       %s: ib_dereg_mr returned %i\n",
			__func__, rc);
	}
649 650 651 652 653 654
	if (ia->ri_id != NULL && !IS_ERR(ia->ri_id)) {
		if (ia->ri_id->qp)
			rdma_destroy_qp(ia->ri_id);
		rdma_destroy_id(ia->ri_id);
		ia->ri_id = NULL;
	}
655 656 657 658 659 660 661 662 663 664 665 666 667 668 669
	if (ia->ri_pd != NULL && !IS_ERR(ia->ri_pd)) {
		rc = ib_dealloc_pd(ia->ri_pd);
		dprintk("RPC:       %s: ib_dealloc_pd returned %i\n",
			__func__, rc);
	}
}

/*
 * Create unconnected endpoint.
 */
int
rpcrdma_ep_create(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia,
				struct rpcrdma_create_data_internal *cdata)
{
	struct ib_device_attr devattr;
670
	struct ib_cq *sendcq, *recvcq;
C
Chuck Lever 已提交
671
	int rc, err;
672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689

	rc = ib_query_device(ia->ri_id->device, &devattr);
	if (rc) {
		dprintk("RPC:       %s: ib_query_device failed %d\n",
			__func__, rc);
		return rc;
	}

	/* check provider's send/recv wr limits */
	if (cdata->max_requests > devattr.max_qp_wr)
		cdata->max_requests = devattr.max_qp_wr;

	ep->rep_attr.event_handler = rpcrdma_qp_async_error_upcall;
	ep->rep_attr.qp_context = ep;
	/* send_cq and recv_cq initialized below */
	ep->rep_attr.srq = NULL;
	ep->rep_attr.cap.max_send_wr = cdata->max_requests;
	switch (ia->ri_memreg_strategy) {
690 691 692
	case RPCRDMA_FRMR: {
		int depth = 7;

693 694 695
		/* Add room for frmr register and invalidate WRs.
		 * 1. FRMR reg WR for head
		 * 2. FRMR invalidate WR for head
696 697
		 * 3. N FRMR reg WRs for pagelist
		 * 4. N FRMR invalidate WRs for pagelist
698 699 700 701
		 * 5. FRMR reg WR for tail
		 * 6. FRMR invalidate WR for tail
		 * 7. The RDMA_SEND WR
		 */
702 703 704 705 706 707 708 709 710 711 712 713 714 715 716

		/* Calculate N if the device max FRMR depth is smaller than
		 * RPCRDMA_MAX_DATA_SEGS.
		 */
		if (ia->ri_max_frmr_depth < RPCRDMA_MAX_DATA_SEGS) {
			int delta = RPCRDMA_MAX_DATA_SEGS -
				    ia->ri_max_frmr_depth;

			do {
				depth += 2; /* FRMR reg + invalidate */
				delta -= ia->ri_max_frmr_depth;
			} while (delta > 0);

		}
		ep->rep_attr.cap.max_send_wr *= depth;
717
		if (ep->rep_attr.cap.max_send_wr > devattr.max_qp_wr) {
718
			cdata->max_requests = devattr.max_qp_wr / depth;
719 720
			if (!cdata->max_requests)
				return -EINVAL;
721 722
			ep->rep_attr.cap.max_send_wr = cdata->max_requests *
						       depth;
723
		}
724
		break;
725
	}
726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745
	default:
		break;
	}
	ep->rep_attr.cap.max_recv_wr = cdata->max_requests;
	ep->rep_attr.cap.max_send_sge = (cdata->padding ? 4 : 2);
	ep->rep_attr.cap.max_recv_sge = 1;
	ep->rep_attr.cap.max_inline_data = 0;
	ep->rep_attr.sq_sig_type = IB_SIGNAL_REQ_WR;
	ep->rep_attr.qp_type = IB_QPT_RC;
	ep->rep_attr.port_num = ~0;

	dprintk("RPC:       %s: requested max: dtos: send %d recv %d; "
		"iovs: send %d recv %d\n",
		__func__,
		ep->rep_attr.cap.max_send_wr,
		ep->rep_attr.cap.max_recv_wr,
		ep->rep_attr.cap.max_send_sge,
		ep->rep_attr.cap.max_recv_sge);

	/* set trigger for requesting send completion */
746
	ep->rep_cqinit = ep->rep_attr.cap.max_send_wr/2 - 1;
747 748 749 750 751
	if (ep->rep_cqinit <= 2)
		ep->rep_cqinit = 0;
	INIT_CQCOUNT(ep);
	ep->rep_ia = ia;
	init_waitqueue_head(&ep->rep_connect_wait);
752
	INIT_DELAYED_WORK(&ep->rep_connect_worker, rpcrdma_connect_worker);
753

754
	sendcq = ib_create_cq(ia->ri_id->device, rpcrdma_sendcq_upcall,
755
				  rpcrdma_cq_async_error_upcall, ep,
756
				  ep->rep_attr.cap.max_send_wr + 1, 0);
757 758 759
	if (IS_ERR(sendcq)) {
		rc = PTR_ERR(sendcq);
		dprintk("RPC:       %s: failed to create send CQ: %i\n",
760 761 762 763
			__func__, rc);
		goto out1;
	}

764
	rc = ib_req_notify_cq(sendcq, IB_CQ_NEXT_COMP);
765 766 767 768 769 770
	if (rc) {
		dprintk("RPC:       %s: ib_req_notify_cq failed: %i\n",
			__func__, rc);
		goto out2;
	}

771
	recvcq = ib_create_cq(ia->ri_id->device, rpcrdma_recvcq_upcall,
772
				  rpcrdma_cq_async_error_upcall, ep,
773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790
				  ep->rep_attr.cap.max_recv_wr + 1, 0);
	if (IS_ERR(recvcq)) {
		rc = PTR_ERR(recvcq);
		dprintk("RPC:       %s: failed to create recv CQ: %i\n",
			__func__, rc);
		goto out2;
	}

	rc = ib_req_notify_cq(recvcq, IB_CQ_NEXT_COMP);
	if (rc) {
		dprintk("RPC:       %s: ib_req_notify_cq failed: %i\n",
			__func__, rc);
		ib_destroy_cq(recvcq);
		goto out2;
	}

	ep->rep_attr.send_cq = sendcq;
	ep->rep_attr.recv_cq = recvcq;
791 792 793 794 795 796 797 798

	/* Initialize cma parameters */

	/* RPC/RDMA does not use private data */
	ep->rep_remote_cma.private_data = NULL;
	ep->rep_remote_cma.private_data_len = 0;

	/* Client offers RDMA Read but does not initiate */
799
	ep->rep_remote_cma.initiator_depth = 0;
800
	if (devattr.max_qp_rd_atom > 32)	/* arbitrary but <= 255 */
801 802
		ep->rep_remote_cma.responder_resources = 32;
	else
803 804 805 806 807 808 809 810 811
		ep->rep_remote_cma.responder_resources = devattr.max_qp_rd_atom;

	ep->rep_remote_cma.retry_count = 7;
	ep->rep_remote_cma.flow_control = 0;
	ep->rep_remote_cma.rnr_retry_count = 0;

	return 0;

out2:
812
	err = ib_destroy_cq(sendcq);
C
Chuck Lever 已提交
813 814 815
	if (err)
		dprintk("RPC:       %s: ib_destroy_cq returned %i\n",
			__func__, err);
816 817 818 819 820 821 822 823 824 825 826
out1:
	return rc;
}

/*
 * rpcrdma_ep_destroy
 *
 * Disconnect and destroy endpoint. After this, the only
 * valid operations on the ep are to free it (if dynamically
 * allocated) or re-create it.
 */
827
void
828 829 830 831 832 833 834
rpcrdma_ep_destroy(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia)
{
	int rc;

	dprintk("RPC:       %s: entering, connected is %d\n",
		__func__, ep->rep_connected);

835 836
	cancel_delayed_work_sync(&ep->rep_connect_worker);

837 838 839 840 841
	if (ia->ri_id->qp) {
		rc = rpcrdma_ep_disconnect(ep, ia);
		if (rc)
			dprintk("RPC:       %s: rpcrdma_ep_disconnect"
				" returned %i\n", __func__, rc);
842 843
		rdma_destroy_qp(ia->ri_id);
		ia->ri_id->qp = NULL;
844 845 846 847 848 849 850 851
	}

	/* padding - could be done in rpcrdma_buffer_destroy... */
	if (ep->rep_pad_mr) {
		rpcrdma_deregister_internal(ia, ep->rep_pad_mr, &ep->rep_pad);
		ep->rep_pad_mr = NULL;
	}

852 853 854 855 856 857 858 859
	rpcrdma_clean_cq(ep->rep_attr.recv_cq);
	rc = ib_destroy_cq(ep->rep_attr.recv_cq);
	if (rc)
		dprintk("RPC:       %s: ib_destroy_cq returned %i\n",
			__func__, rc);

	rpcrdma_clean_cq(ep->rep_attr.send_cq);
	rc = ib_destroy_cq(ep->rep_attr.send_cq);
860 861 862 863 864 865 866 867 868 869 870
	if (rc)
		dprintk("RPC:       %s: ib_destroy_cq returned %i\n",
			__func__, rc);
}

/*
 * Connect unconnected endpoint.
 */
int
rpcrdma_ep_connect(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia)
{
871
	struct rdma_cm_id *id, *old;
872 873 874
	int rc = 0;
	int retry_count = 0;

875
	if (ep->rep_connected != 0) {
876 877
		struct rpcrdma_xprt *xprt;
retry:
878
		dprintk("RPC:       %s: reconnecting...\n", __func__);
879 880 881 882
		rc = rpcrdma_ep_disconnect(ep, ia);
		if (rc && rc != -ENOTCONN)
			dprintk("RPC:       %s: rpcrdma_ep_disconnect"
				" status %i\n", __func__, rc);
883
		rpcrdma_flush_cqs(ep);
884

885 886 887
		if (ia->ri_memreg_strategy == RPCRDMA_FRMR)
			rpcrdma_reset_frmrs(ia);

888 889 890 891
		xprt = container_of(ia, struct rpcrdma_xprt, rx_ia);
		id = rpcrdma_create_id(xprt, ia,
				(struct sockaddr *)&xprt->rx_data.addr);
		if (IS_ERR(id)) {
892
			rc = -EHOSTUNREACH;
893 894 895 896 897 898 899 900 901 902 903 904 905
			goto out;
		}
		/* TEMP TEMP TEMP - fail if new device:
		 * Deregister/remarshal *all* requests!
		 * Close and recreate adapter, pd, etc!
		 * Re-determine all attributes still sane!
		 * More stuff I haven't thought of!
		 * Rrrgh!
		 */
		if (ia->ri_id->device != id->device) {
			printk("RPC:       %s: can't reconnect on "
				"different device!\n", __func__);
			rdma_destroy_id(id);
906
			rc = -ENETUNREACH;
907 908 909
			goto out;
		}
		/* END TEMP */
910 911 912 913 914 915 916 917
		rc = rdma_create_qp(id, ia->ri_pd, &ep->rep_attr);
		if (rc) {
			dprintk("RPC:       %s: rdma_create_qp failed %i\n",
				__func__, rc);
			rdma_destroy_id(id);
			rc = -ENETUNREACH;
			goto out;
		}
918 919 920

		write_lock(&ia->ri_qplock);
		old = ia->ri_id;
921
		ia->ri_id = id;
922 923 924 925
		write_unlock(&ia->ri_qplock);

		rdma_destroy_qp(old);
		rdma_destroy_id(old);
926 927 928 929 930 931 932 933 934
	} else {
		dprintk("RPC:       %s: connecting...\n", __func__);
		rc = rdma_create_qp(ia->ri_id, ia->ri_pd, &ep->rep_attr);
		if (rc) {
			dprintk("RPC:       %s: rdma_create_qp failed %i\n",
				__func__, rc);
			/* do not update ep->rep_connected */
			return -ENETUNREACH;
		}
935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953
	}

	ep->rep_connected = 0;

	rc = rdma_connect(ia->ri_id, &ep->rep_remote_cma);
	if (rc) {
		dprintk("RPC:       %s: rdma_connect() failed with %i\n",
				__func__, rc);
		goto out;
	}

	wait_event_interruptible(ep->rep_connect_wait, ep->rep_connected != 0);

	/*
	 * Check state. A non-peer reject indicates no listener
	 * (ECONNREFUSED), which may be a transient state. All
	 * others indicate a transport condition which has already
	 * undergone a best-effort.
	 */
954 955
	if (ep->rep_connected == -ECONNREFUSED &&
	    ++retry_count <= RDMA_CONNECT_RETRY_MAX) {
956 957 958 959 960 961
		dprintk("RPC:       %s: non-peer_reject, retry\n", __func__);
		goto retry;
	}
	if (ep->rep_connected <= 0) {
		/* Sometimes, the only way to reliably connect to remote
		 * CMs is to use same nonzero values for ORD and IRD. */
962 963 964 965 966 967 968 969
		if (retry_count++ <= RDMA_CONNECT_RETRY_MAX + 1 &&
		    (ep->rep_remote_cma.responder_resources == 0 ||
		     ep->rep_remote_cma.initiator_depth !=
				ep->rep_remote_cma.responder_resources)) {
			if (ep->rep_remote_cma.responder_resources == 0)
				ep->rep_remote_cma.responder_resources = 1;
			ep->rep_remote_cma.initiator_depth =
				ep->rep_remote_cma.responder_resources;
970
			goto retry;
971
		}
972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996
		rc = ep->rep_connected;
	} else {
		dprintk("RPC:       %s: connected\n", __func__);
	}

out:
	if (rc)
		ep->rep_connected = rc;
	return rc;
}

/*
 * rpcrdma_ep_disconnect
 *
 * This is separate from destroy to facilitate the ability
 * to reconnect without recreating the endpoint.
 *
 * This call is not reentrant, and must not be made in parallel
 * on the same endpoint.
 */
int
rpcrdma_ep_disconnect(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia)
{
	int rc;

997
	rpcrdma_flush_cqs(ep);
998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019
	rc = rdma_disconnect(ia->ri_id);
	if (!rc) {
		/* returns without wait if not connected */
		wait_event_interruptible(ep->rep_connect_wait,
							ep->rep_connected != 1);
		dprintk("RPC:       %s: after wait, %sconnected\n", __func__,
			(ep->rep_connected == 1) ? "still " : "dis");
	} else {
		dprintk("RPC:       %s: rdma_disconnect %i\n", __func__, rc);
		ep->rep_connected = rc;
	}
	return rc;
}

/*
 * Initialize buffer memory
 */
int
rpcrdma_buffer_create(struct rpcrdma_buffer *buf, struct rpcrdma_ep *ep,
	struct rpcrdma_ia *ia, struct rpcrdma_create_data_internal *cdata)
{
	char *p;
1020
	size_t len, rlen, wlen;
1021
	int i, rc;
1022
	struct rpcrdma_mw *r;
1023 1024 1025 1026 1027 1028 1029 1030 1031 1032

	buf->rb_max_requests = cdata->max_requests;
	spin_lock_init(&buf->rb_lock);
	atomic_set(&buf->rb_credits, 1);

	/* Need to allocate:
	 *   1.  arrays for send and recv pointers
	 *   2.  arrays of struct rpcrdma_req to fill in pointers
	 *   3.  array of struct rpcrdma_rep for replies
	 *   4.  padding, if any
1033
	 *   5.  mw's, fmr's or frmr's, if any
1034 1035 1036 1037 1038 1039 1040
	 * Send/recv buffers in req/rep need to be registered
	 */

	len = buf->rb_max_requests *
		(sizeof(struct rpcrdma_req *) + sizeof(struct rpcrdma_rep *));
	len += cdata->padding;
	switch (ia->ri_memreg_strategy) {
1041 1042 1043 1044
	case RPCRDMA_FRMR:
		len += buf->rb_max_requests * RPCRDMA_MAX_SEGS *
				sizeof(struct rpcrdma_mw);
		break;
1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080
	case RPCRDMA_MTHCAFMR:
		/* TBD we are perhaps overallocating here */
		len += (buf->rb_max_requests + 1) * RPCRDMA_MAX_SEGS *
				sizeof(struct rpcrdma_mw);
		break;
	default:
		break;
	}

	/* allocate 1, 4 and 5 in one shot */
	p = kzalloc(len, GFP_KERNEL);
	if (p == NULL) {
		dprintk("RPC:       %s: req_t/rep_t/pad kzalloc(%zd) failed\n",
			__func__, len);
		rc = -ENOMEM;
		goto out;
	}
	buf->rb_pool = p;	/* for freeing it later */

	buf->rb_send_bufs = (struct rpcrdma_req **) p;
	p = (char *) &buf->rb_send_bufs[buf->rb_max_requests];
	buf->rb_recv_bufs = (struct rpcrdma_rep **) p;
	p = (char *) &buf->rb_recv_bufs[buf->rb_max_requests];

	/*
	 * Register the zeroed pad buffer, if any.
	 */
	if (cdata->padding) {
		rc = rpcrdma_register_internal(ia, p, cdata->padding,
					    &ep->rep_pad_mr, &ep->rep_pad);
		if (rc)
			goto out;
	}
	p += cdata->padding;

	INIT_LIST_HEAD(&buf->rb_mws);
1081
	INIT_LIST_HEAD(&buf->rb_all);
1082
	r = (struct rpcrdma_mw *)p;
1083
	switch (ia->ri_memreg_strategy) {
1084 1085 1086
	case RPCRDMA_FRMR:
		for (i = buf->rb_max_requests * RPCRDMA_MAX_SEGS; i; i--) {
			r->r.frmr.fr_mr = ib_alloc_fast_reg_mr(ia->ri_pd,
1087
						ia->ri_max_frmr_depth);
1088 1089 1090 1091 1092 1093
			if (IS_ERR(r->r.frmr.fr_mr)) {
				rc = PTR_ERR(r->r.frmr.fr_mr);
				dprintk("RPC:       %s: ib_alloc_fast_reg_mr"
					" failed %i\n", __func__, rc);
				goto out;
			}
1094 1095 1096
			r->r.frmr.fr_pgl = ib_alloc_fast_reg_page_list(
						ia->ri_id->device,
						ia->ri_max_frmr_depth);
1097 1098 1099 1100 1101
			if (IS_ERR(r->r.frmr.fr_pgl)) {
				rc = PTR_ERR(r->r.frmr.fr_pgl);
				dprintk("RPC:       %s: "
					"ib_alloc_fast_reg_page_list "
					"failed %i\n", __func__, rc);
A
Allen Andrews 已提交
1102 1103

				ib_dereg_mr(r->r.frmr.fr_mr);
1104 1105
				goto out;
			}
1106
			list_add(&r->mw_all, &buf->rb_all);
1107 1108 1109 1110
			list_add(&r->mw_list, &buf->rb_mws);
			++r;
		}
		break;
1111 1112 1113
	case RPCRDMA_MTHCAFMR:
		/* TBD we are perhaps overallocating here */
		for (i = (buf->rb_max_requests+1) * RPCRDMA_MAX_SEGS; i; i--) {
1114 1115
			static struct ib_fmr_attr fa =
				{ RPCRDMA_MAX_DATA_SEGS, 1, PAGE_SHIFT };
1116 1117 1118 1119 1120 1121 1122 1123 1124
			r->r.fmr = ib_alloc_fmr(ia->ri_pd,
				IB_ACCESS_REMOTE_WRITE | IB_ACCESS_REMOTE_READ,
				&fa);
			if (IS_ERR(r->r.fmr)) {
				rc = PTR_ERR(r->r.fmr);
				dprintk("RPC:       %s: ib_alloc_fmr"
					" failed %i\n", __func__, rc);
				goto out;
			}
1125
			list_add(&r->mw_all, &buf->rb_all);
1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137
			list_add(&r->mw_list, &buf->rb_mws);
			++r;
		}
		break;
	default:
		break;
	}

	/*
	 * Allocate/init the request/reply buffers. Doing this
	 * using kmalloc for now -- one for each buf.
	 */
1138 1139 1140 1141 1142
	wlen = 1 << fls(cdata->inline_wsize + sizeof(struct rpcrdma_req));
	rlen = 1 << fls(cdata->inline_rsize + sizeof(struct rpcrdma_rep));
	dprintk("RPC:       %s: wlen = %zu, rlen = %zu\n",
		__func__, wlen, rlen);

1143 1144 1145 1146
	for (i = 0; i < buf->rb_max_requests; i++) {
		struct rpcrdma_req *req;
		struct rpcrdma_rep *rep;

1147
		req = kmalloc(wlen, GFP_KERNEL);
1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158
		if (req == NULL) {
			dprintk("RPC:       %s: request buffer %d alloc"
				" failed\n", __func__, i);
			rc = -ENOMEM;
			goto out;
		}
		memset(req, 0, sizeof(struct rpcrdma_req));
		buf->rb_send_bufs[i] = req;
		buf->rb_send_bufs[i]->rl_buffer = buf;

		rc = rpcrdma_register_internal(ia, req->rl_base,
1159
				wlen - offsetof(struct rpcrdma_req, rl_base),
1160 1161 1162 1163 1164
				&buf->rb_send_bufs[i]->rl_handle,
				&buf->rb_send_bufs[i]->rl_iov);
		if (rc)
			goto out;

1165 1166
		buf->rb_send_bufs[i]->rl_size = wlen -
						sizeof(struct rpcrdma_req);
1167

1168
		rep = kmalloc(rlen, GFP_KERNEL);
1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179
		if (rep == NULL) {
			dprintk("RPC:       %s: reply buffer %d alloc failed\n",
				__func__, i);
			rc = -ENOMEM;
			goto out;
		}
		memset(rep, 0, sizeof(struct rpcrdma_rep));
		buf->rb_recv_bufs[i] = rep;
		buf->rb_recv_bufs[i]->rr_buffer = buf;

		rc = rpcrdma_register_internal(ia, rep->rr_base,
1180
				rlen - offsetof(struct rpcrdma_rep, rr_base),
1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206
				&buf->rb_recv_bufs[i]->rr_handle,
				&buf->rb_recv_bufs[i]->rr_iov);
		if (rc)
			goto out;

	}
	dprintk("RPC:       %s: max_requests %d\n",
		__func__, buf->rb_max_requests);
	/* done */
	return 0;
out:
	rpcrdma_buffer_destroy(buf);
	return rc;
}

/*
 * Unregister and destroy buffer memory. Need to deal with
 * partial initialization, so it's callable from failed create.
 * Must be called before destroying endpoint, as registrations
 * reference it.
 */
void
rpcrdma_buffer_destroy(struct rpcrdma_buffer *buf)
{
	int rc, i;
	struct rpcrdma_ia *ia = rdmab_to_ia(buf);
1207
	struct rpcrdma_mw *r;
1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231

	/* clean up in reverse order from create
	 *   1.  recv mr memory (mr free, then kfree)
	 *   2.  send mr memory (mr free, then kfree)
	 *   3.  padding (if any) [moved to rpcrdma_ep_destroy]
	 *   4.  arrays
	 */
	dprintk("RPC:       %s: entering\n", __func__);

	for (i = 0; i < buf->rb_max_requests; i++) {
		if (buf->rb_recv_bufs && buf->rb_recv_bufs[i]) {
			rpcrdma_deregister_internal(ia,
					buf->rb_recv_bufs[i]->rr_handle,
					&buf->rb_recv_bufs[i]->rr_iov);
			kfree(buf->rb_recv_bufs[i]);
		}
		if (buf->rb_send_bufs && buf->rb_send_bufs[i]) {
			rpcrdma_deregister_internal(ia,
					buf->rb_send_bufs[i]->rl_handle,
					&buf->rb_send_bufs[i]->rl_iov);
			kfree(buf->rb_send_bufs[i]);
		}
	}

A
Allen Andrews 已提交
1232 1233 1234
	while (!list_empty(&buf->rb_mws)) {
		r = list_entry(buf->rb_mws.next,
			struct rpcrdma_mw, mw_list);
1235
		list_del(&r->mw_all);
A
Allen Andrews 已提交
1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259
		list_del(&r->mw_list);
		switch (ia->ri_memreg_strategy) {
		case RPCRDMA_FRMR:
			rc = ib_dereg_mr(r->r.frmr.fr_mr);
			if (rc)
				dprintk("RPC:       %s:"
					" ib_dereg_mr"
					" failed %i\n",
					__func__, rc);
			ib_free_fast_reg_page_list(r->r.frmr.fr_pgl);
			break;
		case RPCRDMA_MTHCAFMR:
			rc = ib_dealloc_fmr(r->r.fmr);
			if (rc)
				dprintk("RPC:       %s:"
					" ib_dealloc_fmr"
					" failed %i\n",
					__func__, rc);
			break;
		default:
			break;
		}
	}

1260 1261 1262
	kfree(buf->rb_pool);
}

1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318
/* After a disconnect, a flushed FAST_REG_MR can leave an FRMR in
 * an unusable state. Find FRMRs in this state and dereg / reg
 * each.  FRMRs that are VALID and attached to an rpcrdma_req are
 * also torn down.
 *
 * This gives all in-use FRMRs a fresh rkey and leaves them INVALID.
 *
 * This is invoked only in the transport connect worker in order
 * to serialize with rpcrdma_register_frmr_external().
 */
static void
rpcrdma_reset_frmrs(struct rpcrdma_ia *ia)
{
	struct rpcrdma_xprt *r_xprt =
				container_of(ia, struct rpcrdma_xprt, rx_ia);
	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
	struct list_head *pos;
	struct rpcrdma_mw *r;
	int rc;

	list_for_each(pos, &buf->rb_all) {
		r = list_entry(pos, struct rpcrdma_mw, mw_all);

		if (r->r.frmr.fr_state == FRMR_IS_INVALID)
			continue;

		rc = ib_dereg_mr(r->r.frmr.fr_mr);
		if (rc)
			dprintk("RPC:       %s: ib_dereg_mr failed %i\n",
				__func__, rc);
		ib_free_fast_reg_page_list(r->r.frmr.fr_pgl);

		r->r.frmr.fr_mr = ib_alloc_fast_reg_mr(ia->ri_pd,
					ia->ri_max_frmr_depth);
		if (IS_ERR(r->r.frmr.fr_mr)) {
			rc = PTR_ERR(r->r.frmr.fr_mr);
			dprintk("RPC:       %s: ib_alloc_fast_reg_mr"
				" failed %i\n", __func__, rc);
			continue;
		}
		r->r.frmr.fr_pgl = ib_alloc_fast_reg_page_list(
					ia->ri_id->device,
					ia->ri_max_frmr_depth);
		if (IS_ERR(r->r.frmr.fr_pgl)) {
			rc = PTR_ERR(r->r.frmr.fr_pgl);
			dprintk("RPC:       %s: "
				"ib_alloc_fast_reg_page_list "
				"failed %i\n", __func__, rc);

			ib_dereg_mr(r->r.frmr.fr_mr);
			continue;
		}
		r->r.frmr.fr_state = FRMR_IS_INVALID;
	}
}

1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357
/* "*mw" can be NULL when rpcrdma_buffer_get_mrs() fails, leaving
 * some req segments uninitialized.
 */
static void
rpcrdma_buffer_put_mr(struct rpcrdma_mw **mw, struct rpcrdma_buffer *buf)
{
	if (*mw) {
		list_add_tail(&(*mw)->mw_list, &buf->rb_mws);
		*mw = NULL;
	}
}

/* Cycle mw's back in reverse order, and "spin" them.
 * This delays and scrambles reuse as much as possible.
 */
static void
rpcrdma_buffer_put_mrs(struct rpcrdma_req *req, struct rpcrdma_buffer *buf)
{
	struct rpcrdma_mr_seg *seg = req->rl_segments;
	struct rpcrdma_mr_seg *seg1 = seg;
	int i;

	for (i = 1, seg++; i < RPCRDMA_MAX_SEGS; seg++, i++)
		rpcrdma_buffer_put_mr(&seg->mr_chunk.rl_mw, buf);
	rpcrdma_buffer_put_mr(&seg1->mr_chunk.rl_mw, buf);
}

static void
rpcrdma_buffer_put_sendbuf(struct rpcrdma_req *req, struct rpcrdma_buffer *buf)
{
	buf->rb_send_bufs[--buf->rb_send_index] = req;
	req->rl_niovs = 0;
	if (req->rl_reply) {
		buf->rb_recv_bufs[--buf->rb_recv_index] = req->rl_reply;
		req->rl_reply->rr_func = NULL;
		req->rl_reply = NULL;
	}
}

1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440
/* rpcrdma_unmap_one() was already done by rpcrdma_deregister_frmr_external().
 * Redo only the ib_post_send().
 */
static void
rpcrdma_retry_local_inv(struct rpcrdma_mw *r, struct rpcrdma_ia *ia)
{
	struct rpcrdma_xprt *r_xprt =
				container_of(ia, struct rpcrdma_xprt, rx_ia);
	struct ib_send_wr invalidate_wr, *bad_wr;
	int rc;

	dprintk("RPC:       %s: FRMR %p is stale\n", __func__, r);

	/* When this FRMR is re-inserted into rb_mws, it is no longer stale */
	r->r.frmr.fr_state = FRMR_IS_VALID;

	memset(&invalidate_wr, 0, sizeof(invalidate_wr));
	invalidate_wr.wr_id = (unsigned long)(void *)r;
	invalidate_wr.opcode = IB_WR_LOCAL_INV;
	invalidate_wr.send_flags = IB_SEND_SIGNALED;
	invalidate_wr.ex.invalidate_rkey = r->r.frmr.fr_mr->rkey;
	DECR_CQCOUNT(&r_xprt->rx_ep);

	dprintk("RPC:       %s: frmr %p invalidating rkey %08x\n",
		__func__, r, r->r.frmr.fr_mr->rkey);

	read_lock(&ia->ri_qplock);
	rc = ib_post_send(ia->ri_id->qp, &invalidate_wr, &bad_wr);
	read_unlock(&ia->ri_qplock);
	if (rc) {
		/* Force rpcrdma_buffer_get() to retry */
		r->r.frmr.fr_state = FRMR_IS_STALE;
		dprintk("RPC:       %s: ib_post_send failed, %i\n",
			__func__, rc);
	}
}

static void
rpcrdma_retry_flushed_linv(struct list_head *stale,
			   struct rpcrdma_buffer *buf)
{
	struct rpcrdma_ia *ia = rdmab_to_ia(buf);
	struct list_head *pos;
	struct rpcrdma_mw *r;
	unsigned long flags;

	list_for_each(pos, stale) {
		r = list_entry(pos, struct rpcrdma_mw, mw_list);
		rpcrdma_retry_local_inv(r, ia);
	}

	spin_lock_irqsave(&buf->rb_lock, flags);
	list_splice_tail(stale, &buf->rb_mws);
	spin_unlock_irqrestore(&buf->rb_lock, flags);
}

static struct rpcrdma_req *
rpcrdma_buffer_get_frmrs(struct rpcrdma_req *req, struct rpcrdma_buffer *buf,
			 struct list_head *stale)
{
	struct rpcrdma_mw *r;
	int i;

	i = RPCRDMA_MAX_SEGS - 1;
	while (!list_empty(&buf->rb_mws)) {
		r = list_entry(buf->rb_mws.next,
			       struct rpcrdma_mw, mw_list);
		list_del(&r->mw_list);
		if (r->r.frmr.fr_state == FRMR_IS_STALE) {
			list_add(&r->mw_list, stale);
			continue;
		}
		req->rl_segments[i].mr_chunk.rl_mw = r;
		if (unlikely(i-- == 0))
			return req;	/* Success */
	}

	/* Not enough entries on rb_mws for this req */
	rpcrdma_buffer_put_sendbuf(req, buf);
	rpcrdma_buffer_put_mrs(req, buf);
	return NULL;
}

1441
static struct rpcrdma_req *
1442
rpcrdma_buffer_get_fmrs(struct rpcrdma_req *req, struct rpcrdma_buffer *buf)
1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462
{
	struct rpcrdma_mw *r;
	int i;

	i = RPCRDMA_MAX_SEGS - 1;
	while (!list_empty(&buf->rb_mws)) {
		r = list_entry(buf->rb_mws.next,
			       struct rpcrdma_mw, mw_list);
		list_del(&r->mw_list);
		req->rl_segments[i].mr_chunk.rl_mw = r;
		if (unlikely(i-- == 0))
			return req;	/* Success */
	}

	/* Not enough entries on rb_mws for this req */
	rpcrdma_buffer_put_sendbuf(req, buf);
	rpcrdma_buffer_put_mrs(req, buf);
	return NULL;
}

1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474
/*
 * Get a set of request/reply buffers.
 *
 * Reply buffer (if needed) is attached to send buffer upon return.
 * Rule:
 *    rb_send_index and rb_recv_index MUST always be pointing to the
 *    *next* available buffer (non-NULL). They are incremented after
 *    removing buffers, and decremented *before* returning them.
 */
struct rpcrdma_req *
rpcrdma_buffer_get(struct rpcrdma_buffer *buffers)
{
1475
	struct rpcrdma_ia *ia = rdmab_to_ia(buffers);
1476
	struct list_head stale;
1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497
	struct rpcrdma_req *req;
	unsigned long flags;

	spin_lock_irqsave(&buffers->rb_lock, flags);
	if (buffers->rb_send_index == buffers->rb_max_requests) {
		spin_unlock_irqrestore(&buffers->rb_lock, flags);
		dprintk("RPC:       %s: out of request buffers\n", __func__);
		return ((struct rpcrdma_req *)NULL);
	}

	req = buffers->rb_send_bufs[buffers->rb_send_index];
	if (buffers->rb_send_index < buffers->rb_recv_index) {
		dprintk("RPC:       %s: %d extra receives outstanding (ok)\n",
			__func__,
			buffers->rb_recv_index - buffers->rb_send_index);
		req->rl_reply = NULL;
	} else {
		req->rl_reply = buffers->rb_recv_bufs[buffers->rb_recv_index];
		buffers->rb_recv_bufs[buffers->rb_recv_index++] = NULL;
	}
	buffers->rb_send_bufs[buffers->rb_send_index++] = NULL;
1498 1499

	INIT_LIST_HEAD(&stale);
1500 1501
	switch (ia->ri_memreg_strategy) {
	case RPCRDMA_FRMR:
1502 1503
		req = rpcrdma_buffer_get_frmrs(req, buffers, &stale);
		break;
1504
	case RPCRDMA_MTHCAFMR:
1505
		req = rpcrdma_buffer_get_fmrs(req, buffers);
1506 1507 1508
		break;
	default:
		break;
1509 1510
	}
	spin_unlock_irqrestore(&buffers->rb_lock, flags);
1511 1512
	if (!list_empty(&stale))
		rpcrdma_retry_flushed_linv(&stale, buffers);
1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527
	return req;
}

/*
 * Put request/reply buffers back into pool.
 * Pre-decrement counter/array index.
 */
void
rpcrdma_buffer_put(struct rpcrdma_req *req)
{
	struct rpcrdma_buffer *buffers = req->rl_buffer;
	struct rpcrdma_ia *ia = rdmab_to_ia(buffers);
	unsigned long flags;

	spin_lock_irqsave(&buffers->rb_lock, flags);
1528
	rpcrdma_buffer_put_sendbuf(req, buffers);
1529
	switch (ia->ri_memreg_strategy) {
1530
	case RPCRDMA_FRMR:
1531
	case RPCRDMA_MTHCAFMR:
1532
		rpcrdma_buffer_put_mrs(req, buffers);
1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562
		break;
	default:
		break;
	}
	spin_unlock_irqrestore(&buffers->rb_lock, flags);
}

/*
 * Recover reply buffers from pool.
 * This happens when recovering from error conditions.
 * Post-increment counter/array index.
 */
void
rpcrdma_recv_buffer_get(struct rpcrdma_req *req)
{
	struct rpcrdma_buffer *buffers = req->rl_buffer;
	unsigned long flags;

	if (req->rl_iov.length == 0)	/* special case xprt_rdma_allocate() */
		buffers = ((struct rpcrdma_req *) buffers)->rl_buffer;
	spin_lock_irqsave(&buffers->rb_lock, flags);
	if (buffers->rb_recv_index < buffers->rb_max_requests) {
		req->rl_reply = buffers->rb_recv_bufs[buffers->rb_recv_index];
		buffers->rb_recv_bufs[buffers->rb_recv_index++] = NULL;
	}
	spin_unlock_irqrestore(&buffers->rb_lock, flags);
}

/*
 * Put reply buffers back into pool when not attached to
1563
 * request. This happens in error conditions.
1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593
 */
void
rpcrdma_recv_buffer_put(struct rpcrdma_rep *rep)
{
	struct rpcrdma_buffer *buffers = rep->rr_buffer;
	unsigned long flags;

	rep->rr_func = NULL;
	spin_lock_irqsave(&buffers->rb_lock, flags);
	buffers->rb_recv_bufs[--buffers->rb_recv_index] = rep;
	spin_unlock_irqrestore(&buffers->rb_lock, flags);
}

/*
 * Wrappers for internal-use kmalloc memory registration, used by buffer code.
 */

int
rpcrdma_register_internal(struct rpcrdma_ia *ia, void *va, int len,
				struct ib_mr **mrp, struct ib_sge *iov)
{
	struct ib_phys_buf ipb;
	struct ib_mr *mr;
	int rc;

	/*
	 * All memory passed here was kmalloc'ed, therefore phys-contiguous.
	 */
	iov->addr = ib_dma_map_single(ia->ri_id->device,
			va, len, DMA_BIDIRECTIONAL);
1594 1595 1596
	if (ib_dma_mapping_error(ia->ri_id->device, iov->addr))
		return -ENOMEM;

1597 1598
	iov->length = len;

1599 1600 1601 1602 1603
	if (ia->ri_have_dma_lkey) {
		*mrp = NULL;
		iov->lkey = ia->ri_dma_lkey;
		return 0;
	} else if (ia->ri_bind_mem != NULL) {
1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615
		*mrp = NULL;
		iov->lkey = ia->ri_bind_mem->lkey;
		return 0;
	}

	ipb.addr = iov->addr;
	ipb.size = iov->length;
	mr = ib_reg_phys_mr(ia->ri_pd, &ipb, 1,
			IB_ACCESS_LOCAL_WRITE, &iov->addr);

	dprintk("RPC:       %s: phys convert: 0x%llx "
			"registered 0x%llx length %d\n",
1616 1617
			__func__, (unsigned long long)ipb.addr,
			(unsigned long long)iov->addr, len);
1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666

	if (IS_ERR(mr)) {
		*mrp = NULL;
		rc = PTR_ERR(mr);
		dprintk("RPC:       %s: failed with %i\n", __func__, rc);
	} else {
		*mrp = mr;
		iov->lkey = mr->lkey;
		rc = 0;
	}

	return rc;
}

int
rpcrdma_deregister_internal(struct rpcrdma_ia *ia,
				struct ib_mr *mr, struct ib_sge *iov)
{
	int rc;

	ib_dma_unmap_single(ia->ri_id->device,
			iov->addr, iov->length, DMA_BIDIRECTIONAL);

	if (NULL == mr)
		return 0;

	rc = ib_dereg_mr(mr);
	if (rc)
		dprintk("RPC:       %s: ib_dereg_mr failed %i\n", __func__, rc);
	return rc;
}

/*
 * Wrappers for chunk registration, shared by read/write chunk code.
 */

static void
rpcrdma_map_one(struct rpcrdma_ia *ia, struct rpcrdma_mr_seg *seg, int writing)
{
	seg->mr_dir = writing ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
	seg->mr_dmalen = seg->mr_len;
	if (seg->mr_page)
		seg->mr_dma = ib_dma_map_page(ia->ri_id->device,
				seg->mr_page, offset_in_page(seg->mr_offset),
				seg->mr_dmalen, seg->mr_dir);
	else
		seg->mr_dma = ib_dma_map_single(ia->ri_id->device,
				seg->mr_offset,
				seg->mr_dmalen, seg->mr_dir);
1667 1668 1669
	if (ib_dma_mapping_error(ia->ri_id->device, seg->mr_dma)) {
		dprintk("RPC:       %s: mr_dma %llx mr_offset %p mr_dma_len %zu\n",
			__func__,
R
Randy Dunlap 已提交
1670 1671
			(unsigned long long)seg->mr_dma,
			seg->mr_offset, seg->mr_dmalen);
1672
	}
1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685
}

static void
rpcrdma_unmap_one(struct rpcrdma_ia *ia, struct rpcrdma_mr_seg *seg)
{
	if (seg->mr_page)
		ib_dma_unmap_page(ia->ri_id->device,
				seg->mr_dma, seg->mr_dmalen, seg->mr_dir);
	else
		ib_dma_unmap_single(ia->ri_id->device,
				seg->mr_dma, seg->mr_dmalen, seg->mr_dir);
}

1686 1687 1688 1689 1690 1691
static int
rpcrdma_register_frmr_external(struct rpcrdma_mr_seg *seg,
			int *nsegs, int writing, struct rpcrdma_ia *ia,
			struct rpcrdma_xprt *r_xprt)
{
	struct rpcrdma_mr_seg *seg1 = seg;
1692 1693 1694
	struct rpcrdma_mw *mw = seg1->mr_chunk.rl_mw;
	struct rpcrdma_frmr *frmr = &mw->r.frmr;
	struct ib_mr *mr = frmr->fr_mr;
1695
	struct ib_send_wr frmr_wr, *bad_wr;
1696 1697 1698
	u8 key;
	int len, pageoff;
	int i, rc;
1699 1700 1701
	int seg_len;
	u64 pa;
	int page_no;
1702 1703 1704 1705 1706

	pageoff = offset_in_page(seg1->mr_offset);
	seg1->mr_offset -= pageoff;	/* start of page */
	seg1->mr_len += pageoff;
	len = -pageoff;
1707 1708
	if (*nsegs > ia->ri_max_frmr_depth)
		*nsegs = ia->ri_max_frmr_depth;
1709
	for (page_no = i = 0; i < *nsegs;) {
1710
		rpcrdma_map_one(ia, seg, writing);
1711 1712
		pa = seg->mr_dma;
		for (seg_len = seg->mr_len; seg_len > 0; seg_len -= PAGE_SIZE) {
1713
			frmr->fr_pgl->page_list[page_no++] = pa;
1714 1715
			pa += PAGE_SIZE;
		}
1716 1717 1718 1719 1720 1721 1722 1723 1724
		len += seg->mr_len;
		++seg;
		++i;
		/* Check for holes */
		if ((i < *nsegs && offset_in_page(seg->mr_offset)) ||
		    offset_in_page((seg-1)->mr_offset + (seg-1)->mr_len))
			break;
	}
	dprintk("RPC:       %s: Using frmr %p to map %d segments\n",
1725
		__func__, mw, i);
1726

1727 1728
	frmr->fr_state = FRMR_IS_VALID;

1729
	memset(&frmr_wr, 0, sizeof frmr_wr);
1730
	frmr_wr.wr_id = (unsigned long)(void *)mw;
1731
	frmr_wr.opcode = IB_WR_FAST_REG_MR;
1732
	frmr_wr.wr.fast_reg.iova_start = seg1->mr_dma;
1733
	frmr_wr.wr.fast_reg.page_list = frmr->fr_pgl;
1734
	frmr_wr.wr.fast_reg.page_list_len = page_no;
1735
	frmr_wr.wr.fast_reg.page_shift = PAGE_SHIFT;
1736
	frmr_wr.wr.fast_reg.length = page_no << PAGE_SHIFT;
1737
	if (frmr_wr.wr.fast_reg.length < len) {
1738 1739
		rc = -EIO;
		goto out_err;
1740 1741 1742
	}

	/* Bump the key */
1743 1744
	key = (u8)(mr->rkey & 0x000000FF);
	ib_update_fast_reg_key(mr, ++key);
1745

1746
	frmr_wr.wr.fast_reg.access_flags = (writing ?
1747 1748
				IB_ACCESS_REMOTE_WRITE | IB_ACCESS_LOCAL_WRITE :
				IB_ACCESS_REMOTE_READ);
1749
	frmr_wr.wr.fast_reg.rkey = mr->rkey;
1750 1751
	DECR_CQCOUNT(&r_xprt->rx_ep);

1752
	rc = ib_post_send(ia->ri_id->qp, &frmr_wr, &bad_wr);
1753 1754 1755
	if (rc) {
		dprintk("RPC:       %s: failed ib_post_send for register,"
			" status %i\n", __func__, rc);
1756
		ib_update_fast_reg_key(mr, --key);
1757
		goto out_err;
1758
	} else {
1759
		seg1->mr_rkey = mr->rkey;
1760 1761 1762 1763 1764
		seg1->mr_base = seg1->mr_dma + pageoff;
		seg1->mr_nsegs = i;
		seg1->mr_len = len;
	}
	*nsegs = i;
1765 1766
	return 0;
out_err:
1767
	frmr->fr_state = FRMR_IS_INVALID;
1768 1769
	while (i--)
		rpcrdma_unmap_one(ia, --seg);
1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781
	return rc;
}

static int
rpcrdma_deregister_frmr_external(struct rpcrdma_mr_seg *seg,
			struct rpcrdma_ia *ia, struct rpcrdma_xprt *r_xprt)
{
	struct rpcrdma_mr_seg *seg1 = seg;
	struct ib_send_wr invalidate_wr, *bad_wr;
	int rc;

	memset(&invalidate_wr, 0, sizeof invalidate_wr);
1782
	invalidate_wr.wr_id = (unsigned long)(void *)seg1->mr_chunk.rl_mw;
1783
	invalidate_wr.opcode = IB_WR_LOCAL_INV;
1784
	invalidate_wr.send_flags = IB_SEND_SIGNALED;
1785 1786 1787
	invalidate_wr.ex.invalidate_rkey = seg1->mr_chunk.rl_mw->r.frmr.fr_mr->rkey;
	DECR_CQCOUNT(&r_xprt->rx_ep);

1788 1789 1790
	read_lock(&ia->ri_qplock);
	while (seg1->mr_nsegs--)
		rpcrdma_unmap_one(ia, seg++);
1791
	rc = ib_post_send(ia->ri_id->qp, &invalidate_wr, &bad_wr);
1792
	read_unlock(&ia->ri_qplock);
1793 1794 1795 1796 1797 1798
	if (rc)
		dprintk("RPC:       %s: failed ib_post_send for invalidate,"
			" status %i\n", __func__, rc);
	return rc;
}

1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852
static int
rpcrdma_register_fmr_external(struct rpcrdma_mr_seg *seg,
			int *nsegs, int writing, struct rpcrdma_ia *ia)
{
	struct rpcrdma_mr_seg *seg1 = seg;
	u64 physaddrs[RPCRDMA_MAX_DATA_SEGS];
	int len, pageoff, i, rc;

	pageoff = offset_in_page(seg1->mr_offset);
	seg1->mr_offset -= pageoff;	/* start of page */
	seg1->mr_len += pageoff;
	len = -pageoff;
	if (*nsegs > RPCRDMA_MAX_DATA_SEGS)
		*nsegs = RPCRDMA_MAX_DATA_SEGS;
	for (i = 0; i < *nsegs;) {
		rpcrdma_map_one(ia, seg, writing);
		physaddrs[i] = seg->mr_dma;
		len += seg->mr_len;
		++seg;
		++i;
		/* Check for holes */
		if ((i < *nsegs && offset_in_page(seg->mr_offset)) ||
		    offset_in_page((seg-1)->mr_offset + (seg-1)->mr_len))
			break;
	}
	rc = ib_map_phys_fmr(seg1->mr_chunk.rl_mw->r.fmr,
				physaddrs, i, seg1->mr_dma);
	if (rc) {
		dprintk("RPC:       %s: failed ib_map_phys_fmr "
			"%u@0x%llx+%i (%d)... status %i\n", __func__,
			len, (unsigned long long)seg1->mr_dma,
			pageoff, i, rc);
		while (i--)
			rpcrdma_unmap_one(ia, --seg);
	} else {
		seg1->mr_rkey = seg1->mr_chunk.rl_mw->r.fmr->rkey;
		seg1->mr_base = seg1->mr_dma + pageoff;
		seg1->mr_nsegs = i;
		seg1->mr_len = len;
	}
	*nsegs = i;
	return rc;
}

static int
rpcrdma_deregister_fmr_external(struct rpcrdma_mr_seg *seg,
			struct rpcrdma_ia *ia)
{
	struct rpcrdma_mr_seg *seg1 = seg;
	LIST_HEAD(l);
	int rc;

	list_add(&seg1->mr_chunk.rl_mw->r.fmr->list, &l);
	rc = ib_unmap_fmr(&l);
1853
	read_lock(&ia->ri_qplock);
1854 1855
	while (seg1->mr_nsegs--)
		rpcrdma_unmap_one(ia, seg++);
1856
	read_unlock(&ia->ri_qplock);
1857 1858 1859 1860 1861 1862
	if (rc)
		dprintk("RPC:       %s: failed ib_unmap_fmr,"
			" status %i\n", __func__, rc);
	return rc;
}

1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881
int
rpcrdma_register_external(struct rpcrdma_mr_seg *seg,
			int nsegs, int writing, struct rpcrdma_xprt *r_xprt)
{
	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
	int rc = 0;

	switch (ia->ri_memreg_strategy) {

#if RPCRDMA_PERSISTENT_REGISTRATION
	case RPCRDMA_ALLPHYSICAL:
		rpcrdma_map_one(ia, seg, writing);
		seg->mr_rkey = ia->ri_bind_mem->rkey;
		seg->mr_base = seg->mr_dma;
		seg->mr_nsegs = 1;
		nsegs = 1;
		break;
#endif

1882 1883 1884 1885 1886
	/* Registration using frmr registration */
	case RPCRDMA_FRMR:
		rc = rpcrdma_register_frmr_external(seg, &nsegs, writing, ia, r_xprt);
		break;

1887
	/* Registration using fmr memory registration */
1888
	case RPCRDMA_MTHCAFMR:
1889
		rc = rpcrdma_register_fmr_external(seg, &nsegs, writing, ia);
1890 1891 1892
		break;

	default:
1893
		return -1;
1894 1895 1896 1897 1898 1899 1900 1901 1902
	}
	if (rc)
		return -1;

	return nsegs;
}

int
rpcrdma_deregister_external(struct rpcrdma_mr_seg *seg,
1903
		struct rpcrdma_xprt *r_xprt)
1904 1905 1906 1907 1908 1909 1910 1911
{
	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
	int nsegs = seg->mr_nsegs, rc;

	switch (ia->ri_memreg_strategy) {

#if RPCRDMA_PERSISTENT_REGISTRATION
	case RPCRDMA_ALLPHYSICAL:
1912
		read_lock(&ia->ri_qplock);
1913
		rpcrdma_unmap_one(ia, seg);
1914
		read_unlock(&ia->ri_qplock);
1915 1916 1917
		break;
#endif

1918 1919 1920 1921
	case RPCRDMA_FRMR:
		rc = rpcrdma_deregister_frmr_external(seg, ia, r_xprt);
		break;

1922
	case RPCRDMA_MTHCAFMR:
1923
		rc = rpcrdma_deregister_fmr_external(seg, ia);
1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
		break;

	default:
		break;
	}
	return nsegs;
}

/*
 * Prepost any receive buffer, then post send.
 *
 * Receive buffer is donated to hardware, reclaimed upon recv completion.
 */
int
rpcrdma_ep_post(struct rpcrdma_ia *ia,
		struct rpcrdma_ep *ep,
		struct rpcrdma_req *req)
{
	struct ib_send_wr send_wr, *send_wr_fail;
	struct rpcrdma_rep *rep = req->rl_reply;
	int rc;

	if (rep) {
		rc = rpcrdma_ep_post_recv(ia, ep, rep);
		if (rc)
			goto out;
		req->rl_reply = NULL;
	}

	send_wr.next = NULL;
	send_wr.wr_id = 0ULL;	/* no send cookie */
	send_wr.sg_list = req->rl_send_iov;
	send_wr.num_sge = req->rl_niovs;
	send_wr.opcode = IB_WR_SEND;
	if (send_wr.num_sge == 4)	/* no need to sync any pad (constant) */
		ib_dma_sync_single_for_device(ia->ri_id->device,
			req->rl_send_iov[3].addr, req->rl_send_iov[3].length,
			DMA_TO_DEVICE);
	ib_dma_sync_single_for_device(ia->ri_id->device,
		req->rl_send_iov[1].addr, req->rl_send_iov[1].length,
		DMA_TO_DEVICE);
	ib_dma_sync_single_for_device(ia->ri_id->device,
		req->rl_send_iov[0].addr, req->rl_send_iov[0].length,
		DMA_TO_DEVICE);

	if (DECR_CQCOUNT(ep) > 0)
		send_wr.send_flags = 0;
	else { /* Provider must take a send completion every now and then */
		INIT_CQCOUNT(ep);
		send_wr.send_flags = IB_SEND_SIGNALED;
	}

	rc = ib_post_send(ia->ri_id->qp, &send_wr, &send_wr_fail);
	if (rc)
		dprintk("RPC:       %s: ib_post_send returned %i\n", __func__,
			rc);
out:
	return rc;
}

/*
 * (Re)post a receive buffer.
 */
int
rpcrdma_ep_post_recv(struct rpcrdma_ia *ia,
		     struct rpcrdma_ep *ep,
		     struct rpcrdma_rep *rep)
{
	struct ib_recv_wr recv_wr, *recv_wr_fail;
	int rc;

	recv_wr.next = NULL;
	recv_wr.wr_id = (u64) (unsigned long) rep;
	recv_wr.sg_list = &rep->rr_iov;
	recv_wr.num_sge = 1;

	ib_dma_sync_single_for_cpu(ia->ri_id->device,
		rep->rr_iov.addr, rep->rr_iov.length, DMA_BIDIRECTIONAL);

	rc = ib_post_recv(ia->ri_id->qp, &recv_wr, &recv_wr_fail);

	if (rc)
		dprintk("RPC:       %s: ib_post_recv returned %i\n", __func__,
			rc);
	return rc;
}
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050

/* Physical mapping means one Read/Write list entry per-page.
 * All list entries must fit within an inline buffer
 *
 * NB: The server must return a Write list for NFS READ,
 *     which has the same constraint. Factor in the inline
 *     rsize as well.
 */
static size_t
rpcrdma_physical_max_payload(struct rpcrdma_xprt *r_xprt)
{
	struct rpcrdma_create_data_internal *cdata = &r_xprt->rx_data;
	unsigned int inline_size, pages;

	inline_size = min_t(unsigned int,
			    cdata->inline_wsize, cdata->inline_rsize);
	inline_size -= RPCRDMA_HDRLEN_MIN;
	pages = inline_size / sizeof(struct rpcrdma_segment);
	return pages << PAGE_SHIFT;
}

static size_t
rpcrdma_mr_max_payload(struct rpcrdma_xprt *r_xprt)
{
	return RPCRDMA_MAX_DATA_SEGS << PAGE_SHIFT;
}

size_t
rpcrdma_max_payload(struct rpcrdma_xprt *r_xprt)
{
	size_t result;

	switch (r_xprt->rx_ia.ri_memreg_strategy) {
	case RPCRDMA_ALLPHYSICAL:
		result = rpcrdma_physical_max_payload(r_xprt);
		break;
	default:
		result = rpcrdma_mr_max_payload(r_xprt);
	}
	return result;
}