ib_srp.c 66.9 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
/*
 * Copyright (c) 2005 Cisco Systems.  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
 * OpenIB.org BSD 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.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 */

33 34
#define pr_fmt(fmt) PFX fmt

35 36 37 38 39 40 41
#include <linux/module.h>
#include <linux/init.h>
#include <linux/slab.h>
#include <linux/err.h>
#include <linux/string.h>
#include <linux/parser.h>
#include <linux/random.h>
42
#include <linux/jiffies.h>
43

A
Arun Sharma 已提交
44
#include <linux/atomic.h>
45 46 47 48 49

#include <scsi/scsi.h>
#include <scsi/scsi_device.h>
#include <scsi/scsi_dbg.h>
#include <scsi/srp.h>
50
#include <scsi/scsi_transport_srp.h>
51 52 53 54 55 56 57 58 59 60 61 62 63

#include "ib_srp.h"

#define DRV_NAME	"ib_srp"
#define PFX		DRV_NAME ": "
#define DRV_VERSION	"0.2"
#define DRV_RELDATE	"November 1, 2005"

MODULE_AUTHOR("Roland Dreier");
MODULE_DESCRIPTION("InfiniBand SCSI RDMA Protocol initiator "
		   "v" DRV_VERSION " (" DRV_RELDATE ")");
MODULE_LICENSE("Dual BSD/GPL");

64 65
static unsigned int srp_sg_tablesize;
static unsigned int cmd_sg_entries;
66 67
static unsigned int indirect_sg_entries;
static bool allow_ext_sg;
68
static int topspin_workarounds = 1;
69

70 71
module_param(srp_sg_tablesize, uint, 0444);
MODULE_PARM_DESC(srp_sg_tablesize, "Deprecated name for cmd_sg_entries");
72

73 74 75
module_param(cmd_sg_entries, uint, 0444);
MODULE_PARM_DESC(cmd_sg_entries,
		 "Default number of gather/scatter entries in the SRP command (default is 12, max 255)");
76

77 78 79 80 81 82 83 84
module_param(indirect_sg_entries, uint, 0444);
MODULE_PARM_DESC(indirect_sg_entries,
		 "Default max number of gather/scatter entries (default is 12, max is " __stringify(SCSI_MAX_SG_CHAIN_SEGMENTS) ")");

module_param(allow_ext_sg, bool, 0444);
MODULE_PARM_DESC(allow_ext_sg,
		  "Default behavior when there are more than cmd_sg_entries S/G entries after mapping; fails the request when false (default false)");

85 86 87 88 89 90
module_param(topspin_workarounds, int, 0444);
MODULE_PARM_DESC(topspin_workarounds,
		 "Enable workarounds for Topspin/Cisco SRP target bugs if != 0");

static void srp_add_one(struct ib_device *device);
static void srp_remove_one(struct ib_device *device);
91 92
static void srp_recv_completion(struct ib_cq *cq, void *target_ptr);
static void srp_send_completion(struct ib_cq *cq, void *target_ptr);
93 94
static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event);

95 96
static struct scsi_transport_template *ib_srp_transport_template;

97 98 99 100 101 102
static struct ib_client srp_client = {
	.name   = "srp",
	.add    = srp_add_one,
	.remove = srp_remove_one
};

103 104
static struct ib_sa_client srp_sa_client;

105 106 107 108 109 110 111 112 113 114
static inline struct srp_target_port *host_to_target(struct Scsi_Host *host)
{
	return (struct srp_target_port *) host->hostdata;
}

static const char *srp_target_info(struct Scsi_Host *host)
{
	return host_to_target(host)->target_name;
}

115 116 117
static int srp_target_is_topspin(struct srp_target_port *target)
{
	static const u8 topspin_oui[3] = { 0x00, 0x05, 0xad };
118
	static const u8 cisco_oui[3]   = { 0x00, 0x1b, 0x0d };
119 120

	return topspin_workarounds &&
121 122
		(!memcmp(&target->ioc_guid, topspin_oui, sizeof topspin_oui) ||
		 !memcmp(&target->ioc_guid, cisco_oui, sizeof cisco_oui));
123 124
}

125 126 127 128 129 130 131 132 133 134 135 136 137 138
static struct srp_iu *srp_alloc_iu(struct srp_host *host, size_t size,
				   gfp_t gfp_mask,
				   enum dma_data_direction direction)
{
	struct srp_iu *iu;

	iu = kmalloc(sizeof *iu, gfp_mask);
	if (!iu)
		goto out;

	iu->buf = kzalloc(size, gfp_mask);
	if (!iu->buf)
		goto out_free_iu;

139 140 141
	iu->dma = ib_dma_map_single(host->srp_dev->dev, iu->buf, size,
				    direction);
	if (ib_dma_mapping_error(host->srp_dev->dev, iu->dma))
142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161
		goto out_free_buf;

	iu->size      = size;
	iu->direction = direction;

	return iu;

out_free_buf:
	kfree(iu->buf);
out_free_iu:
	kfree(iu);
out:
	return NULL;
}

static void srp_free_iu(struct srp_host *host, struct srp_iu *iu)
{
	if (!iu)
		return;

162 163
	ib_dma_unmap_single(host->srp_dev->dev, iu->dma, iu->size,
			    iu->direction);
164 165 166 167 168 169
	kfree(iu->buf);
	kfree(iu);
}

static void srp_qp_event(struct ib_event *event, void *context)
{
170
	pr_debug("QP event %d\n", event->event);
171 172 173 174 175 176 177 178 179 180 181 182
}

static int srp_init_qp(struct srp_target_port *target,
		       struct ib_qp *qp)
{
	struct ib_qp_attr *attr;
	int ret;

	attr = kmalloc(sizeof *attr, GFP_KERNEL);
	if (!attr)
		return -ENOMEM;

183 184 185 186
	ret = ib_find_pkey(target->srp_host->srp_dev->dev,
			   target->srp_host->port,
			   be16_to_cpu(target->path.pkey),
			   &attr->pkey_index);
187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205
	if (ret)
		goto out;

	attr->qp_state        = IB_QPS_INIT;
	attr->qp_access_flags = (IB_ACCESS_REMOTE_READ |
				    IB_ACCESS_REMOTE_WRITE);
	attr->port_num        = target->srp_host->port;

	ret = ib_modify_qp(qp, attr,
			   IB_QP_STATE		|
			   IB_QP_PKEY_INDEX	|
			   IB_QP_ACCESS_FLAGS	|
			   IB_QP_PORT);

out:
	kfree(attr);
	return ret;
}

D
David Dillow 已提交
206 207 208 209
static int srp_new_cm_id(struct srp_target_port *target)
{
	struct ib_cm_id *new_cm_id;

210
	new_cm_id = ib_create_cm_id(target->srp_host->srp_dev->dev,
D
David Dillow 已提交
211 212 213 214 215 216 217 218 219 220 221
				    srp_cm_handler, target);
	if (IS_ERR(new_cm_id))
		return PTR_ERR(new_cm_id);

	if (target->cm_id)
		ib_destroy_cm_id(target->cm_id);
	target->cm_id = new_cm_id;

	return 0;
}

222 223 224
static int srp_create_target_ib(struct srp_target_port *target)
{
	struct ib_qp_init_attr *init_attr;
225 226
	struct ib_cq *recv_cq, *send_cq;
	struct ib_qp *qp;
227 228 229 230 231 232
	int ret;

	init_attr = kzalloc(sizeof *init_attr, GFP_KERNEL);
	if (!init_attr)
		return -ENOMEM;

233 234 235 236
	recv_cq = ib_create_cq(target->srp_host->srp_dev->dev,
			       srp_recv_completion, NULL, target, SRP_RQ_SIZE, 0);
	if (IS_ERR(recv_cq)) {
		ret = PTR_ERR(recv_cq);
237
		goto err;
238 239
	}

240 241 242 243
	send_cq = ib_create_cq(target->srp_host->srp_dev->dev,
			       srp_send_completion, NULL, target, SRP_SQ_SIZE, 0);
	if (IS_ERR(send_cq)) {
		ret = PTR_ERR(send_cq);
244
		goto err_recv_cq;
245 246
	}

247
	ib_req_notify_cq(recv_cq, IB_CQ_NEXT_COMP);
248 249 250 251 252 253 254 255

	init_attr->event_handler       = srp_qp_event;
	init_attr->cap.max_send_wr     = SRP_SQ_SIZE;
	init_attr->cap.max_recv_wr     = SRP_RQ_SIZE;
	init_attr->cap.max_recv_sge    = 1;
	init_attr->cap.max_send_sge    = 1;
	init_attr->sq_sig_type         = IB_SIGNAL_ALL_WR;
	init_attr->qp_type             = IB_QPT_RC;
256 257
	init_attr->send_cq             = send_cq;
	init_attr->recv_cq             = recv_cq;
258

259 260 261
	qp = ib_create_qp(target->srp_host->srp_dev->pd, init_attr);
	if (IS_ERR(qp)) {
		ret = PTR_ERR(qp);
262
		goto err_send_cq;
263 264
	}

265
	ret = srp_init_qp(target, qp);
266 267
	if (ret)
		goto err_qp;
268

269 270 271 272 273 274 275 276 277 278 279
	if (target->qp)
		ib_destroy_qp(target->qp);
	if (target->recv_cq)
		ib_destroy_cq(target->recv_cq);
	if (target->send_cq)
		ib_destroy_cq(target->send_cq);

	target->qp = qp;
	target->recv_cq = recv_cq;
	target->send_cq = send_cq;

280 281 282 283
	kfree(init_attr);
	return 0;

err_qp:
284
	ib_destroy_qp(qp);
285 286

err_send_cq:
287
	ib_destroy_cq(send_cq);
288 289

err_recv_cq:
290
	ib_destroy_cq(recv_cq);
291 292

err:
293 294 295 296 297 298 299 300 301
	kfree(init_attr);
	return ret;
}

static void srp_free_target_ib(struct srp_target_port *target)
{
	int i;

	ib_destroy_qp(target->qp);
302 303
	ib_destroy_cq(target->send_cq);
	ib_destroy_cq(target->recv_cq);
304

305 306 307
	target->qp = NULL;
	target->send_cq = target->recv_cq = NULL;

308 309
	for (i = 0; i < SRP_RQ_SIZE; ++i)
		srp_free_iu(target->srp_host, target->rx_ring[i]);
310
	for (i = 0; i < SRP_SQ_SIZE; ++i)
311 312 313 314 315 316 317 318 319 320 321
		srp_free_iu(target->srp_host, target->tx_ring[i]);
}

static void srp_path_rec_completion(int status,
				    struct ib_sa_path_rec *pathrec,
				    void *target_ptr)
{
	struct srp_target_port *target = target_ptr;

	target->status = status;
	if (status)
322 323
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Got failed path rec status %d\n", status);
324 325 326 327 328 329 330 331 332 333 334
	else
		target->path = *pathrec;
	complete(&target->done);
}

static int srp_lookup_path(struct srp_target_port *target)
{
	target->path.numb_path = 1;

	init_completion(&target->done);

335
	target->path_query_id = ib_sa_path_rec_get(&srp_sa_client,
336
						   target->srp_host->srp_dev->dev,
337 338
						   target->srp_host->port,
						   &target->path,
339
						   IB_SA_PATH_REC_SERVICE_ID	|
340 341 342 343 344 345 346 347 348 349 350 351 352 353
						   IB_SA_PATH_REC_DGID		|
						   IB_SA_PATH_REC_SGID		|
						   IB_SA_PATH_REC_NUMB_PATH	|
						   IB_SA_PATH_REC_PKEY,
						   SRP_PATH_REC_TIMEOUT_MS,
						   GFP_KERNEL,
						   srp_path_rec_completion,
						   target, &target->path_query);
	if (target->path_query_id < 0)
		return target->path_query_id;

	wait_for_completion(&target->done);

	if (target->status < 0)
354 355
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Path record query failed\n");
356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396

	return target->status;
}

static int srp_send_req(struct srp_target_port *target)
{
	struct {
		struct ib_cm_req_param param;
		struct srp_login_req   priv;
	} *req = NULL;
	int status;

	req = kzalloc(sizeof *req, GFP_KERNEL);
	if (!req)
		return -ENOMEM;

	req->param.primary_path 	      = &target->path;
	req->param.alternate_path 	      = NULL;
	req->param.service_id 		      = target->service_id;
	req->param.qp_num 		      = target->qp->qp_num;
	req->param.qp_type 		      = target->qp->qp_type;
	req->param.private_data 	      = &req->priv;
	req->param.private_data_len 	      = sizeof req->priv;
	req->param.flow_control 	      = 1;

	get_random_bytes(&req->param.starting_psn, 4);
	req->param.starting_psn 	     &= 0xffffff;

	/*
	 * Pick some arbitrary defaults here; we could make these
	 * module parameters if anyone cared about setting them.
	 */
	req->param.responder_resources	      = 4;
	req->param.remote_cm_response_timeout = 20;
	req->param.local_cm_response_timeout  = 20;
	req->param.retry_count 		      = 7;
	req->param.rnr_retry_count 	      = 7;
	req->param.max_cm_retries 	      = 15;

	req->priv.opcode     	= SRP_LOGIN_REQ;
	req->priv.tag        	= 0;
397
	req->priv.req_it_iu_len = cpu_to_be32(target->max_iu_len);
398 399
	req->priv.req_buf_fmt 	= cpu_to_be16(SRP_BUF_FORMAT_DIRECT |
					      SRP_BUF_FORMAT_INDIRECT);
400
	/*
R
Roland Dreier 已提交
401
	 * In the published SRP specification (draft rev. 16a), the
402 403 404 405 406 407 408 409 410
	 * port identifier format is 8 bytes of ID extension followed
	 * by 8 bytes of GUID.  Older drafts put the two halves in the
	 * opposite order, so that the GUID comes first.
	 *
	 * Targets conforming to these obsolete drafts can be
	 * recognized by the I/O Class they report.
	 */
	if (target->io_class == SRP_REV10_IB_IO_CLASS) {
		memcpy(req->priv.initiator_port_id,
411
		       &target->path.sgid.global.interface_id, 8);
412
		memcpy(req->priv.initiator_port_id + 8,
413
		       &target->initiator_ext, 8);
414 415 416 417
		memcpy(req->priv.target_port_id,     &target->ioc_guid, 8);
		memcpy(req->priv.target_port_id + 8, &target->id_ext, 8);
	} else {
		memcpy(req->priv.initiator_port_id,
418 419 420
		       &target->initiator_ext, 8);
		memcpy(req->priv.initiator_port_id + 8,
		       &target->path.sgid.global.interface_id, 8);
421 422 423 424
		memcpy(req->priv.target_port_id,     &target->id_ext, 8);
		memcpy(req->priv.target_port_id + 8, &target->ioc_guid, 8);
	}

425 426
	/*
	 * Topspin/Cisco SRP targets will reject our login unless we
427 428
	 * zero out the first 8 bytes of our initiator port ID and set
	 * the second 8 bytes to the local node GUID.
429
	 */
430
	if (srp_target_is_topspin(target)) {
431 432 433 434
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Topspin/Cisco initiator port ID workaround "
			     "activated for target GUID %016llx\n",
			     (unsigned long long) be64_to_cpu(target->ioc_guid));
435
		memset(req->priv.initiator_port_id, 0, 8);
436
		memcpy(req->priv.initiator_port_id + 8,
437
		       &target->srp_host->srp_dev->dev->node_guid, 8);
438 439 440 441 442 443 444 445 446
	}

	status = ib_send_cm_req(target->cm_id, &req->param);

	kfree(req);

	return status;
}

447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463
static bool srp_queue_remove_work(struct srp_target_port *target)
{
	bool changed = false;

	spin_lock_irq(&target->lock);
	if (target->state != SRP_TARGET_REMOVED) {
		target->state = SRP_TARGET_REMOVED;
		changed = true;
	}
	spin_unlock_irq(&target->lock);

	if (changed)
		queue_work(system_long_wq, &target->remove_work);

	return changed;
}

464 465 466 467 468 469 470 471 472 473 474 475 476 477 478
static bool srp_change_conn_state(struct srp_target_port *target,
				  bool connected)
{
	bool changed = false;

	spin_lock_irq(&target->lock);
	if (target->connected != connected) {
		target->connected = connected;
		changed = true;
	}
	spin_unlock_irq(&target->lock);

	return changed;
}

479 480
static void srp_disconnect_target(struct srp_target_port *target)
{
481 482
	if (srp_change_conn_state(target, false)) {
		/* XXX should send SRP_I_LOGOUT request */
483

484 485 486 487
		if (ib_send_cm_dreq(target->cm_id, NULL, 0)) {
			shost_printk(KERN_DEBUG, target->scsi_host,
				     PFX "Sending CM DREQ failed\n");
		}
488
	}
489 490
}

491 492
static void srp_free_req_data(struct srp_target_port *target)
{
493
	struct ib_device *ibdev = target->srp_host->srp_dev->dev;
494 495 496 497 498 499
	struct srp_request *req;
	int i;

	for (i = 0, req = target->req_ring; i < SRP_CMD_SQ_SIZE; ++i, ++req) {
		kfree(req->fmr_list);
		kfree(req->map_page);
500 501 502 503 504 505
		if (req->indirect_dma_addr) {
			ib_dma_unmap_single(ibdev, req->indirect_dma_addr,
					    target->indirect_size,
					    DMA_TO_DEVICE);
		}
		kfree(req->indirect_desc);
506 507 508
	}
}

509 510 511 512 513 514 515 516 517 518 519 520 521 522 523
/**
 * srp_del_scsi_host_attr() - Remove attributes defined in the host template.
 * @shost: SCSI host whose attributes to remove from sysfs.
 *
 * Note: Any attributes defined in the host template and that did not exist
 * before invocation of this function will be ignored.
 */
static void srp_del_scsi_host_attr(struct Scsi_Host *shost)
{
	struct device_attribute **attr;

	for (attr = shost->hostt->shost_attrs; attr && *attr; ++attr)
		device_remove_file(&shost->shost_dev, *attr);
}

524 525
static void srp_remove_target(struct srp_target_port *target)
{
526 527
	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);

528 529 530
	srp_del_scsi_host_attr(target->scsi_host);
	srp_remove_host(target->scsi_host);
	scsi_remove_host(target->scsi_host);
531
	srp_disconnect_target(target);
532 533 534 535 536 537
	ib_destroy_cm_id(target->cm_id);
	srp_free_target_ib(target);
	srp_free_req_data(target);
	scsi_host_put(target->scsi_host);
}

D
David Howells 已提交
538
static void srp_remove_work(struct work_struct *work)
539
{
D
David Howells 已提交
540
	struct srp_target_port *target =
541
		container_of(work, struct srp_target_port, remove_work);
542

543
	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);
544

545
	spin_lock(&target->srp_host->target_lock);
546
	list_del(&target->list);
547
	spin_unlock(&target->srp_host->target_lock);
548

549
	srp_remove_target(target);
550 551 552 553
}

static int srp_connect_target(struct srp_target_port *target)
{
D
David Dillow 已提交
554
	int retries = 3;
555 556
	int ret;

557 558
	WARN_ON_ONCE(target->connected);

559 560
	target->qp_in_error = false;

561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579
	ret = srp_lookup_path(target);
	if (ret)
		return ret;

	while (1) {
		init_completion(&target->done);
		ret = srp_send_req(target);
		if (ret)
			return ret;
		wait_for_completion(&target->done);

		/*
		 * The CM event handling code will set status to
		 * SRP_PORT_REDIRECT if we get a port redirect REJ
		 * back, or SRP_DLID_REDIRECT if we get a lid/qp
		 * redirect REJ back.
		 */
		switch (target->status) {
		case 0:
580
			srp_change_conn_state(target, true);
581 582 583 584 585 586 587 588 589 590 591
			return 0;

		case SRP_PORT_REDIRECT:
			ret = srp_lookup_path(target);
			if (ret)
				return ret;
			break;

		case SRP_DLID_REDIRECT:
			break;

D
David Dillow 已提交
592 593 594 595 596 597 598 599 600 601 602 603 604 605 606
		case SRP_STALE_CONN:
			/* Our current CM id was stale, and is now in timewait.
			 * Try to reconnect with a new one.
			 */
			if (!retries-- || srp_new_cm_id(target)) {
				shost_printk(KERN_ERR, target->scsi_host, PFX
					     "giving up on stale connection\n");
				target->status = -ECONNRESET;
				return target->status;
			}

			shost_printk(KERN_ERR, target->scsi_host, PFX
				     "retrying stale connection\n");
			break;

607 608 609 610 611 612
		default:
			return target->status;
		}
	}
}

613 614 615 616
static void srp_unmap_data(struct scsi_cmnd *scmnd,
			   struct srp_target_port *target,
			   struct srp_request *req)
{
617 618 619
	struct ib_device *ibdev = target->srp_host->srp_dev->dev;
	struct ib_pool_fmr **pfmr;

620
	if (!scsi_sglist(scmnd) ||
621 622 623 624
	    (scmnd->sc_data_direction != DMA_TO_DEVICE &&
	     scmnd->sc_data_direction != DMA_FROM_DEVICE))
		return;

625 626 627
	pfmr = req->fmr_list;
	while (req->nfmr--)
		ib_fmr_pool_unmap(*pfmr++);
628

629 630
	ib_dma_unmap_sg(ibdev, scsi_sglist(scmnd), scsi_sg_count(scmnd),
			scmnd->sc_data_direction);
631 632
}

B
Bart Van Assche 已提交
633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668
/**
 * srp_claim_req - Take ownership of the scmnd associated with a request.
 * @target: SRP target port.
 * @req: SRP request.
 * @scmnd: If NULL, take ownership of @req->scmnd. If not NULL, only take
 *         ownership of @req->scmnd if it equals @scmnd.
 *
 * Return value:
 * Either NULL or a pointer to the SCSI command the caller became owner of.
 */
static struct scsi_cmnd *srp_claim_req(struct srp_target_port *target,
				       struct srp_request *req,
				       struct scsi_cmnd *scmnd)
{
	unsigned long flags;

	spin_lock_irqsave(&target->lock, flags);
	if (!scmnd) {
		scmnd = req->scmnd;
		req->scmnd = NULL;
	} else if (req->scmnd == scmnd) {
		req->scmnd = NULL;
	} else {
		scmnd = NULL;
	}
	spin_unlock_irqrestore(&target->lock, flags);

	return scmnd;
}

/**
 * srp_free_req() - Unmap data and add request to the free request list.
 */
static void srp_free_req(struct srp_target_port *target,
			 struct srp_request *req, struct scsi_cmnd *scmnd,
			 s32 req_lim_delta)
669
{
670 671
	unsigned long flags;

B
Bart Van Assche 已提交
672 673
	srp_unmap_data(scmnd, target, req);

674
	spin_lock_irqsave(&target->lock, flags);
675
	target->req_lim += req_lim_delta;
676
	list_add_tail(&req->list, &target->free_reqs);
677
	spin_unlock_irqrestore(&target->lock, flags);
678 679 680 681
}

static void srp_reset_req(struct srp_target_port *target, struct srp_request *req)
{
B
Bart Van Assche 已提交
682 683 684
	struct scsi_cmnd *scmnd = srp_claim_req(target, req, NULL);

	if (scmnd) {
685
		srp_free_req(target, req, scmnd, 0);
B
Bart Van Assche 已提交
686 687 688
		scmnd->result = DID_RESET << 16;
		scmnd->scsi_done(scmnd);
	}
689 690
}

691 692
static int srp_reconnect_target(struct srp_target_port *target)
{
693
	struct Scsi_Host *shost = target->scsi_host;
694
	int i, ret;
695

696
	if (target->state != SRP_TARGET_LIVE)
697 698
		return -EAGAIN;

699 700
	scsi_target_block(&shost->shost_gendev);

701 702 703 704 705
	srp_disconnect_target(target);
	/*
	 * Now get a new local CM ID so that we avoid confusing the
	 * target in case things are really fouled up.
	 */
D
David Dillow 已提交
706 707
	ret = srp_new_cm_id(target);
	if (ret)
708
		goto unblock;
709

710
	ret = srp_create_target_ib(target);
711
	if (ret)
712
		goto unblock;
713

714 715 716 717 718
	for (i = 0; i < SRP_CMD_SQ_SIZE; ++i) {
		struct srp_request *req = &target->req_ring[i];
		if (req->scmnd)
			srp_reset_req(target, req);
	}
719

720
	INIT_LIST_HEAD(&target->free_tx);
721
	for (i = 0; i < SRP_SQ_SIZE; ++i)
722
		list_add(&target->tx_ring[i]->list, &target->free_tx);
723 724

	ret = srp_connect_target(target);
725 726 727 728 729

unblock:
	scsi_target_unblock(&shost->shost_gendev, ret == 0 ? SDEV_RUNNING :
			    SDEV_TRANSPORT_OFFLINE);

730 731 732
	if (ret)
		goto err;

733
	shost_printk(KERN_INFO, target->scsi_host, PFX "reconnect succeeded\n");
734 735 736 737

	return ret;

err:
738 739
	shost_printk(KERN_ERR, target->scsi_host,
		     PFX "reconnect failed (%d), removing target port.\n", ret);
740 741 742

	/*
	 * We couldn't reconnect, so kill our target port off.
743 744 745
	 * However, we have to defer the real removal because we
	 * are in the context of the SCSI error handler now, which
	 * will deadlock if we call scsi_remove_host().
746
	 */
747
	srp_queue_remove_work(target);
748 749 750 751

	return ret;
}

752 753
static void srp_map_desc(struct srp_map_state *state, dma_addr_t dma_addr,
			 unsigned int dma_len, u32 rkey)
754
{
755
	struct srp_direct_buf *desc = state->desc;
756

757 758 759
	desc->va = cpu_to_be64(dma_addr);
	desc->key = cpu_to_be32(rkey);
	desc->len = cpu_to_be32(dma_len);
760

761 762 763 764
	state->total_len += dma_len;
	state->desc++;
	state->ndesc++;
}
765

766 767 768 769 770 771
static int srp_map_finish_fmr(struct srp_map_state *state,
			      struct srp_target_port *target)
{
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_pool_fmr *fmr;
	u64 io_addr = 0;
772

773 774
	if (!state->npages)
		return 0;
775

776 777 778 779 780
	if (state->npages == 1) {
		srp_map_desc(state, state->base_dma_addr, state->fmr_len,
			     target->rkey);
		state->npages = state->fmr_len = 0;
		return 0;
781 782
	}

783 784 785 786
	fmr = ib_fmr_pool_map_phys(dev->fmr_pool, state->pages,
				   state->npages, io_addr);
	if (IS_ERR(fmr))
		return PTR_ERR(fmr);
787

788 789
	*state->next_fmr++ = fmr;
	state->nfmr++;
790

791 792 793 794 795 796 797 798 799 800 801 802 803
	srp_map_desc(state, 0, state->fmr_len, fmr->fmr->rkey);
	state->npages = state->fmr_len = 0;
	return 0;
}

static void srp_map_update_start(struct srp_map_state *state,
				 struct scatterlist *sg, int sg_index,
				 dma_addr_t dma_addr)
{
	state->unmapped_sg = sg;
	state->unmapped_index = sg_index;
	state->unmapped_addr = dma_addr;
}
804

805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826
static int srp_map_sg_entry(struct srp_map_state *state,
			    struct srp_target_port *target,
			    struct scatterlist *sg, int sg_index,
			    int use_fmr)
{
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = dev->dev;
	dma_addr_t dma_addr = ib_sg_dma_address(ibdev, sg);
	unsigned int dma_len = ib_sg_dma_len(ibdev, sg);
	unsigned int len;
	int ret;

	if (!dma_len)
		return 0;

	if (use_fmr == SRP_MAP_NO_FMR) {
		/* Once we're in direct map mode for a request, we don't
		 * go back to FMR mode, so no need to update anything
		 * other than the descriptor.
		 */
		srp_map_desc(state, dma_addr, dma_len, target->rkey);
		return 0;
827
	}
828

829 830 831 832 833 834 835 836 837 838 839 840 841 842
	/* If we start at an offset into the FMR page, don't merge into
	 * the current FMR. Finish it out, and use the kernel's MR for this
	 * sg entry. This is to avoid potential bugs on some SRP targets
	 * that were never quite defined, but went away when the initiator
	 * avoided using FMR on such page fragments.
	 */
	if (dma_addr & ~dev->fmr_page_mask || dma_len > dev->fmr_max_size) {
		ret = srp_map_finish_fmr(state, target);
		if (ret)
			return ret;

		srp_map_desc(state, dma_addr, dma_len, target->rkey);
		srp_map_update_start(state, NULL, 0, 0);
		return 0;
843 844
	}

845 846 847 848 849 850 851
	/* If this is the first sg to go into the FMR, save our position.
	 * We need to know the first unmapped entry, its index, and the
	 * first unmapped address within that entry to be able to restart
	 * mapping after an error.
	 */
	if (!state->unmapped_sg)
		srp_map_update_start(state, sg, sg_index, dma_addr);
852

853 854 855 856 857
	while (dma_len) {
		if (state->npages == SRP_FMR_SIZE) {
			ret = srp_map_finish_fmr(state, target);
			if (ret)
				return ret;
858

859 860 861 862
			srp_map_update_start(state, sg, sg_index, dma_addr);
		}

		len = min_t(unsigned int, dma_len, dev->fmr_page_size);
863

864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881
		if (!state->npages)
			state->base_dma_addr = dma_addr;
		state->pages[state->npages++] = dma_addr;
		state->fmr_len += len;
		dma_addr += len;
		dma_len -= len;
	}

	/* If the last entry of the FMR wasn't a full page, then we need to
	 * close it out and start a new one -- we can only merge at page
	 * boundries.
	 */
	ret = 0;
	if (len != dev->fmr_page_size) {
		ret = srp_map_finish_fmr(state, target);
		if (!ret)
			srp_map_update_start(state, NULL, 0, 0);
	}
882 883 884
	return ret;
}

885 886 887
static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_target_port *target,
			struct srp_request *req)
{
888
	struct scatterlist *scat, *sg;
889
	struct srp_cmd *cmd = req->cmd->buf;
890
	int i, len, nents, count, use_fmr;
891 892
	struct srp_device *dev;
	struct ib_device *ibdev;
893 894 895 896
	struct srp_map_state state;
	struct srp_indirect_buf *indirect_hdr;
	u32 table_len;
	u8 fmt;
897

898
	if (!scsi_sglist(scmnd) || scmnd->sc_data_direction == DMA_NONE)
899 900 901 902
		return sizeof (struct srp_cmd);

	if (scmnd->sc_data_direction != DMA_FROM_DEVICE &&
	    scmnd->sc_data_direction != DMA_TO_DEVICE) {
903 904 905
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled data direction %d\n",
			     scmnd->sc_data_direction);
906 907 908
		return -EINVAL;
	}

909 910
	nents = scsi_sg_count(scmnd);
	scat  = scsi_sglist(scmnd);
911

912
	dev = target->srp_host->srp_dev;
913 914 915
	ibdev = dev->dev;

	count = ib_dma_map_sg(ibdev, scat, nents, scmnd->sc_data_direction);
916 917
	if (unlikely(count == 0))
		return -EIO;
918 919 920

	fmt = SRP_DATA_DESC_DIRECT;
	len = sizeof (struct srp_cmd) +	sizeof (struct srp_direct_buf);
921

922
	if (count == 1) {
923 924 925 926 927 928
		/*
		 * The midlayer only generated a single gather/scatter
		 * entry, or DMA mapping coalesced everything to a
		 * single entry.  So a direct descriptor along with
		 * the DMA MR suffices.
		 */
929
		struct srp_direct_buf *buf = (void *) cmd->add_data;
930

931
		buf->va  = cpu_to_be64(ib_sg_dma_address(ibdev, scat));
932
		buf->key = cpu_to_be32(target->rkey);
933
		buf->len = cpu_to_be32(ib_sg_dma_len(ibdev, scat));
934 935 936 937 938 939 940 941 942 943 944

		req->nfmr = 0;
		goto map_complete;
	}

	/* We have more than one scatter/gather entry, so build our indirect
	 * descriptor table, trying to merge as many entries with FMR as we
	 * can.
	 */
	indirect_hdr = (void *) cmd->add_data;

945 946 947
	ib_dma_sync_single_for_cpu(ibdev, req->indirect_dma_addr,
				   target->indirect_size, DMA_TO_DEVICE);

948
	memset(&state, 0, sizeof(state));
949
	state.desc	= req->indirect_desc;
950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972
	state.pages	= req->map_page;
	state.next_fmr	= req->fmr_list;

	use_fmr = dev->fmr_pool ? SRP_MAP_ALLOW_FMR : SRP_MAP_NO_FMR;

	for_each_sg(scat, sg, count, i) {
		if (srp_map_sg_entry(&state, target, sg, i, use_fmr)) {
			/* FMR mapping failed, so backtrack to the first
			 * unmapped entry and continue on without using FMR.
			 */
			dma_addr_t dma_addr;
			unsigned int dma_len;

backtrack:
			sg = state.unmapped_sg;
			i = state.unmapped_index;

			dma_addr = ib_sg_dma_address(ibdev, sg);
			dma_len = ib_sg_dma_len(ibdev, sg);
			dma_len -= (state.unmapped_addr - dma_addr);
			dma_addr = state.unmapped_addr;
			use_fmr = SRP_MAP_NO_FMR;
			srp_map_desc(&state, dma_addr, dma_len, target->rkey);
973
		}
974
	}
975

976 977
	if (use_fmr == SRP_MAP_ALLOW_FMR && srp_map_finish_fmr(&state, target))
		goto backtrack;
978

979 980 981 982 983
	/* We've mapped the request, now pull as much of the indirect
	 * descriptor table as we can into the command buffer. If this
	 * target is not using an external indirect table, we are
	 * guaranteed to fit into the command, as the SCSI layer won't
	 * give us more S/G entries than we allow.
984 985 986 987 988 989 990
	 */
	req->nfmr = state.nfmr;
	if (state.ndesc == 1) {
		/* FMR mapping was able to collapse this to one entry,
		 * so use a direct descriptor.
		 */
		struct srp_direct_buf *buf = (void *) cmd->add_data;
991

992
		*buf = req->indirect_desc[0];
993
		goto map_complete;
994 995
	}

996 997 998 999 1000 1001 1002 1003
	if (unlikely(target->cmd_sg_cnt < state.ndesc &&
						!target->allow_ext_sg)) {
		shost_printk(KERN_ERR, target->scsi_host,
			     "Could not fit S/G list into SRP_CMD\n");
		return -EIO;
	}

	count = min(state.ndesc, target->cmd_sg_cnt);
1004 1005 1006 1007
	table_len = state.ndesc * sizeof (struct srp_direct_buf);

	fmt = SRP_DATA_DESC_INDIRECT;
	len = sizeof(struct srp_cmd) + sizeof (struct srp_indirect_buf);
1008
	len += count * sizeof (struct srp_direct_buf);
1009

1010 1011
	memcpy(indirect_hdr->desc_list, req->indirect_desc,
	       count * sizeof (struct srp_direct_buf));
1012

1013
	indirect_hdr->table_desc.va = cpu_to_be64(req->indirect_dma_addr);
1014 1015 1016 1017 1018
	indirect_hdr->table_desc.key = cpu_to_be32(target->rkey);
	indirect_hdr->table_desc.len = cpu_to_be32(table_len);
	indirect_hdr->len = cpu_to_be32(state.total_len);

	if (scmnd->sc_data_direction == DMA_TO_DEVICE)
1019
		cmd->data_out_desc_cnt = count;
1020
	else
1021 1022 1023 1024
		cmd->data_in_desc_cnt = count;

	ib_dma_sync_single_for_device(ibdev, req->indirect_dma_addr, table_len,
				      DMA_TO_DEVICE);
1025 1026

map_complete:
1027 1028 1029 1030 1031 1032 1033 1034
	if (scmnd->sc_data_direction == DMA_TO_DEVICE)
		cmd->buf_fmt = fmt << 4;
	else
		cmd->buf_fmt = fmt;

	return len;
}

1035 1036 1037 1038 1039 1040 1041 1042
/*
 * Return an IU and possible credit to the free pool
 */
static void srp_put_tx_iu(struct srp_target_port *target, struct srp_iu *iu,
			  enum srp_iu_type iu_type)
{
	unsigned long flags;

1043
	spin_lock_irqsave(&target->lock, flags);
1044 1045 1046
	list_add(&iu->list, &target->free_tx);
	if (iu_type != SRP_IU_RSP)
		++target->req_lim;
1047
	spin_unlock_irqrestore(&target->lock, flags);
1048 1049
}

1050
/*
1051 1052
 * Must be called with target->lock held to protect req_lim and free_tx.
 * If IU is not sent, it must be returned using srp_put_tx_iu().
1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070
 *
 * Note:
 * An upper limit for the number of allocated information units for each
 * request type is:
 * - SRP_IU_CMD: SRP_CMD_SQ_SIZE, since the SCSI mid-layer never queues
 *   more than Scsi_Host.can_queue requests.
 * - SRP_IU_TSK_MGMT: SRP_TSK_MGMT_SQ_SIZE.
 * - SRP_IU_RSP: 1, since a conforming SRP target never sends more than
 *   one unanswered SRP request to an initiator.
 */
static struct srp_iu *__srp_get_tx_iu(struct srp_target_port *target,
				      enum srp_iu_type iu_type)
{
	s32 rsv = (iu_type == SRP_IU_TSK_MGMT) ? 0 : SRP_TSK_MGMT_SQ_SIZE;
	struct srp_iu *iu;

	srp_send_completion(target->send_cq, target);

1071
	if (list_empty(&target->free_tx))
1072 1073 1074
		return NULL;

	/* Initiator responses to target requests do not consume credits */
1075 1076 1077 1078 1079 1080 1081
	if (iu_type != SRP_IU_RSP) {
		if (target->req_lim <= rsv) {
			++target->zero_req_lim;
			return NULL;
		}

		--target->req_lim;
1082 1083
	}

1084
	iu = list_first_entry(&target->free_tx, struct srp_iu, list);
1085
	list_del(&iu->list);
1086 1087 1088
	return iu;
}

1089 1090
static int srp_post_send(struct srp_target_port *target,
			 struct srp_iu *iu, int len)
1091 1092 1093 1094 1095 1096
{
	struct ib_sge list;
	struct ib_send_wr wr, *bad_wr;

	list.addr   = iu->dma;
	list.length = len;
1097
	list.lkey   = target->lkey;
1098 1099

	wr.next       = NULL;
1100
	wr.wr_id      = (uintptr_t) iu;
1101 1102 1103 1104 1105
	wr.sg_list    = &list;
	wr.num_sge    = 1;
	wr.opcode     = IB_WR_SEND;
	wr.send_flags = IB_SEND_SIGNALED;

1106
	return ib_post_send(target->qp, &wr, &bad_wr);
1107 1108
}

1109
static int srp_post_recv(struct srp_target_port *target, struct srp_iu *iu)
1110 1111
{
	struct ib_recv_wr wr, *bad_wr;
1112
	struct ib_sge list;
1113 1114 1115

	list.addr   = iu->dma;
	list.length = iu->size;
1116
	list.lkey   = target->lkey;
1117 1118

	wr.next     = NULL;
1119
	wr.wr_id    = (uintptr_t) iu;
1120 1121 1122
	wr.sg_list  = &list;
	wr.num_sge  = 1;

1123
	return ib_post_recv(target->qp, &wr, &bad_wr);
1124 1125
}

1126 1127 1128 1129 1130 1131 1132
static void srp_process_rsp(struct srp_target_port *target, struct srp_rsp *rsp)
{
	struct srp_request *req;
	struct scsi_cmnd *scmnd;
	unsigned long flags;

	if (unlikely(rsp->tag & SRP_TAG_TSK_MGMT)) {
1133
		spin_lock_irqsave(&target->lock, flags);
1134
		target->req_lim += be32_to_cpu(rsp->req_lim_delta);
1135
		spin_unlock_irqrestore(&target->lock, flags);
1136

1137 1138 1139 1140
		target->tsk_mgmt_status = -1;
		if (be32_to_cpu(rsp->resp_data_len) >= 4)
			target->tsk_mgmt_status = rsp->data[3];
		complete(&target->tsk_mgmt_done);
1141
	} else {
1142
		req = &target->req_ring[rsp->tag];
B
Bart Van Assche 已提交
1143 1144
		scmnd = srp_claim_req(target, req, NULL);
		if (!scmnd) {
1145 1146 1147
			shost_printk(KERN_ERR, target->scsi_host,
				     "Null scmnd for RSP w/tag %016llx\n",
				     (unsigned long long) rsp->tag);
B
Bart Van Assche 已提交
1148 1149 1150 1151 1152 1153 1154

			spin_lock_irqsave(&target->lock, flags);
			target->req_lim += be32_to_cpu(rsp->req_lim_delta);
			spin_unlock_irqrestore(&target->lock, flags);

			return;
		}
1155 1156 1157 1158 1159 1160 1161 1162 1163 1164
		scmnd->result = rsp->status;

		if (rsp->flags & SRP_RSP_FLAG_SNSVALID) {
			memcpy(scmnd->sense_buffer, rsp->data +
			       be32_to_cpu(rsp->resp_data_len),
			       min_t(int, be32_to_cpu(rsp->sense_data_len),
				     SCSI_SENSE_BUFFERSIZE));
		}

		if (rsp->flags & (SRP_RSP_FLAG_DOOVER | SRP_RSP_FLAG_DOUNDER))
1165
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_out_res_cnt));
1166
		else if (rsp->flags & (SRP_RSP_FLAG_DIOVER | SRP_RSP_FLAG_DIUNDER))
1167
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_in_res_cnt));
1168

B
Bart Van Assche 已提交
1169 1170 1171
		srp_free_req(target, req, scmnd,
			     be32_to_cpu(rsp->req_lim_delta));

1172 1173
		scmnd->host_scribble = NULL;
		scmnd->scsi_done(scmnd);
1174 1175 1176
	}
}

1177 1178 1179
static int srp_response_common(struct srp_target_port *target, s32 req_delta,
			       void *rsp, int len)
{
1180
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1181 1182
	unsigned long flags;
	struct srp_iu *iu;
1183
	int err;
1184

1185
	spin_lock_irqsave(&target->lock, flags);
1186 1187
	target->req_lim += req_delta;
	iu = __srp_get_tx_iu(target, SRP_IU_RSP);
1188
	spin_unlock_irqrestore(&target->lock, flags);
1189

1190 1191 1192
	if (!iu) {
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "no IU available to send response\n");
1193
		return 1;
1194 1195 1196 1197 1198 1199
	}

	ib_dma_sync_single_for_cpu(dev, iu->dma, len, DMA_TO_DEVICE);
	memcpy(iu->buf, rsp, len);
	ib_dma_sync_single_for_device(dev, iu->dma, len, DMA_TO_DEVICE);

1200 1201
	err = srp_post_send(target, iu, len);
	if (err) {
1202 1203
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "unable to post response: %d\n", err);
1204 1205
		srp_put_tx_iu(target, iu, SRP_IU_RSP);
	}
1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240

	return err;
}

static void srp_process_cred_req(struct srp_target_port *target,
				 struct srp_cred_req *req)
{
	struct srp_cred_rsp rsp = {
		.opcode = SRP_CRED_RSP,
		.tag = req->tag,
	};
	s32 delta = be32_to_cpu(req->req_lim_delta);

	if (srp_response_common(target, delta, &rsp, sizeof rsp))
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "problems processing SRP_CRED_REQ\n");
}

static void srp_process_aer_req(struct srp_target_port *target,
				struct srp_aer_req *req)
{
	struct srp_aer_rsp rsp = {
		.opcode = SRP_AER_RSP,
		.tag = req->tag,
	};
	s32 delta = be32_to_cpu(req->req_lim_delta);

	shost_printk(KERN_ERR, target->scsi_host, PFX
		     "ignoring AER for LUN %llu\n", be64_to_cpu(req->lun));

	if (srp_response_common(target, delta, &rsp, sizeof rsp))
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "problems processing SRP_AER_REQ\n");
}

1241 1242
static void srp_handle_recv(struct srp_target_port *target, struct ib_wc *wc)
{
1243
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1244
	struct srp_iu *iu = (struct srp_iu *) (uintptr_t) wc->wr_id;
1245
	int res;
1246 1247
	u8 opcode;

1248 1249
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_ti_iu_len,
				   DMA_FROM_DEVICE);
1250 1251 1252 1253

	opcode = *(u8 *) iu->buf;

	if (0) {
1254 1255
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "recv completion, opcode 0x%02x\n", opcode);
B
Bart Van Assche 已提交
1256 1257
		print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET, 8, 1,
			       iu->buf, wc->byte_len, true);
1258 1259 1260 1261 1262 1263 1264
	}

	switch (opcode) {
	case SRP_RSP:
		srp_process_rsp(target, iu->buf);
		break;

1265 1266 1267 1268 1269 1270 1271 1272
	case SRP_CRED_REQ:
		srp_process_cred_req(target, iu->buf);
		break;

	case SRP_AER_REQ:
		srp_process_aer_req(target, iu->buf);
		break;

1273 1274
	case SRP_T_LOGOUT:
		/* XXX Handle target logout */
1275 1276
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Got target logout request\n");
1277 1278 1279
		break;

	default:
1280 1281
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled SRP opcode 0x%02x\n", opcode);
1282 1283 1284
		break;
	}

1285 1286
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_ti_iu_len,
				      DMA_FROM_DEVICE);
1287

1288
	res = srp_post_recv(target, iu);
1289 1290 1291
	if (res != 0)
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Recv failed with error code %d\n", res);
1292 1293
}

1294 1295 1296 1297
static void srp_handle_qp_err(enum ib_wc_status wc_status,
			      enum ib_wc_opcode wc_opcode,
			      struct srp_target_port *target)
{
1298
	if (target->connected && !target->qp_in_error) {
1299 1300 1301 1302 1303
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "failed %s status %d\n",
			     wc_opcode & IB_WC_RECV ? "receive" : "send",
			     wc_status);
	}
1304 1305 1306
	target->qp_in_error = true;
}

1307
static void srp_recv_completion(struct ib_cq *cq, void *target_ptr)
1308 1309 1310 1311 1312 1313
{
	struct srp_target_port *target = target_ptr;
	struct ib_wc wc;

	ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
	while (ib_poll_cq(cq, 1, &wc) > 0) {
1314 1315 1316 1317
		if (likely(wc.status == IB_WC_SUCCESS)) {
			srp_handle_recv(target, &wc);
		} else {
			srp_handle_qp_err(wc.status, wc.opcode, target);
1318
		}
1319 1320 1321 1322 1323 1324 1325
	}
}

static void srp_send_completion(struct ib_cq *cq, void *target_ptr)
{
	struct srp_target_port *target = target_ptr;
	struct ib_wc wc;
1326
	struct srp_iu *iu;
1327 1328

	while (ib_poll_cq(cq, 1, &wc) > 0) {
1329 1330 1331 1332 1333
		if (likely(wc.status == IB_WC_SUCCESS)) {
			iu = (struct srp_iu *) (uintptr_t) wc.wr_id;
			list_add(&iu->list, &target->free_tx);
		} else {
			srp_handle_qp_err(wc.status, wc.opcode, target);
1334
		}
1335 1336 1337
	}
}

1338
static int srp_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *scmnd)
1339
{
1340
	struct srp_target_port *target = host_to_target(shost);
1341 1342 1343
	struct srp_request *req;
	struct srp_iu *iu;
	struct srp_cmd *cmd;
1344
	struct ib_device *dev;
1345
	unsigned long flags;
1346 1347
	int len;

1348
	spin_lock_irqsave(&target->lock, flags);
1349
	iu = __srp_get_tx_iu(target, SRP_IU_CMD);
1350
	if (!iu)
1351 1352 1353 1354 1355
		goto err_unlock;

	req = list_first_entry(&target->free_reqs, struct srp_request, list);
	list_del(&req->list);
	spin_unlock_irqrestore(&target->lock, flags);
1356

1357
	dev = target->srp_host->srp_dev->dev;
1358
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_iu_len,
1359
				   DMA_TO_DEVICE);
1360 1361

	scmnd->result        = 0;
1362
	scmnd->host_scribble = (void *) req;
1363 1364 1365 1366 1367 1368

	cmd = iu->buf;
	memset(cmd, 0, sizeof *cmd);

	cmd->opcode = SRP_CMD;
	cmd->lun    = cpu_to_be64((u64) scmnd->device->lun << 48);
1369
	cmd->tag    = req->index;
1370 1371 1372 1373 1374 1375 1376
	memcpy(cmd->cdb, scmnd->cmnd, scmnd->cmd_len);

	req->scmnd    = scmnd;
	req->cmd      = iu;

	len = srp_map_data(scmnd, target, req);
	if (len < 0) {
1377 1378
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Failed to map data\n");
1379
		goto err_iu;
1380 1381
	}

1382
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_iu_len,
1383
				      DMA_TO_DEVICE);
1384

1385
	if (srp_post_send(target, iu, len)) {
1386
		shost_printk(KERN_ERR, target->scsi_host, PFX "Send failed\n");
1387 1388 1389 1390 1391 1392 1393 1394
		goto err_unmap;
	}

	return 0;

err_unmap:
	srp_unmap_data(scmnd, target, req);

1395 1396 1397
err_iu:
	srp_put_tx_iu(target, iu, SRP_IU_CMD);

1398
	spin_lock_irqsave(&target->lock, flags);
1399
	list_add(&req->list, &target->free_reqs);
1400 1401

err_unlock:
1402
	spin_unlock_irqrestore(&target->lock, flags);
1403

1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418
	return SCSI_MLQUEUE_HOST_BUSY;
}

static int srp_alloc_iu_bufs(struct srp_target_port *target)
{
	int i;

	for (i = 0; i < SRP_RQ_SIZE; ++i) {
		target->rx_ring[i] = srp_alloc_iu(target->srp_host,
						  target->max_ti_iu_len,
						  GFP_KERNEL, DMA_FROM_DEVICE);
		if (!target->rx_ring[i])
			goto err;
	}

1419
	for (i = 0; i < SRP_SQ_SIZE; ++i) {
1420
		target->tx_ring[i] = srp_alloc_iu(target->srp_host,
1421
						  target->max_iu_len,
1422 1423 1424
						  GFP_KERNEL, DMA_TO_DEVICE);
		if (!target->tx_ring[i])
			goto err;
1425 1426

		list_add(&target->tx_ring[i]->list, &target->free_tx);
1427 1428 1429 1430 1431 1432 1433 1434 1435 1436
	}

	return 0;

err:
	for (i = 0; i < SRP_RQ_SIZE; ++i) {
		srp_free_iu(target->srp_host, target->rx_ring[i]);
		target->rx_ring[i] = NULL;
	}

1437
	for (i = 0; i < SRP_SQ_SIZE; ++i) {
1438 1439 1440 1441 1442 1443 1444
		srp_free_iu(target->srp_host, target->tx_ring[i]);
		target->tx_ring[i] = NULL;
	}

	return -ENOMEM;
}

1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471
static uint32_t srp_compute_rq_tmo(struct ib_qp_attr *qp_attr, int attr_mask)
{
	uint64_t T_tr_ns, max_compl_time_ms;
	uint32_t rq_tmo_jiffies;

	/*
	 * According to section 11.2.4.2 in the IBTA spec (Modify Queue Pair,
	 * table 91), both the QP timeout and the retry count have to be set
	 * for RC QP's during the RTR to RTS transition.
	 */
	WARN_ON_ONCE((attr_mask & (IB_QP_TIMEOUT | IB_QP_RETRY_CNT)) !=
		     (IB_QP_TIMEOUT | IB_QP_RETRY_CNT));

	/*
	 * Set target->rq_tmo_jiffies to one second more than the largest time
	 * it can take before an error completion is generated. See also
	 * C9-140..142 in the IBTA spec for more information about how to
	 * convert the QP Local ACK Timeout value to nanoseconds.
	 */
	T_tr_ns = 4096 * (1ULL << qp_attr->timeout);
	max_compl_time_ms = qp_attr->retry_cnt * 4 * T_tr_ns;
	do_div(max_compl_time_ms, NSEC_PER_MSEC);
	rq_tmo_jiffies = msecs_to_jiffies(max_compl_time_ms + 1000);

	return rq_tmo_jiffies;
}

1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530
static void srp_cm_rep_handler(struct ib_cm_id *cm_id,
			       struct srp_login_rsp *lrsp,
			       struct srp_target_port *target)
{
	struct ib_qp_attr *qp_attr = NULL;
	int attr_mask = 0;
	int ret;
	int i;

	if (lrsp->opcode == SRP_LOGIN_RSP) {
		target->max_ti_iu_len = be32_to_cpu(lrsp->max_ti_iu_len);
		target->req_lim       = be32_to_cpu(lrsp->req_lim_delta);

		/*
		 * Reserve credits for task management so we don't
		 * bounce requests back to the SCSI mid-layer.
		 */
		target->scsi_host->can_queue
			= min(target->req_lim - SRP_TSK_MGMT_SQ_SIZE,
			      target->scsi_host->can_queue);
	} else {
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled RSP opcode %#x\n", lrsp->opcode);
		ret = -ECONNRESET;
		goto error;
	}

	if (!target->rx_ring[0]) {
		ret = srp_alloc_iu_bufs(target);
		if (ret)
			goto error;
	}

	ret = -ENOMEM;
	qp_attr = kmalloc(sizeof *qp_attr, GFP_KERNEL);
	if (!qp_attr)
		goto error;

	qp_attr->qp_state = IB_QPS_RTR;
	ret = ib_cm_init_qp_attr(cm_id, qp_attr, &attr_mask);
	if (ret)
		goto error_free;

	ret = ib_modify_qp(target->qp, qp_attr, attr_mask);
	if (ret)
		goto error_free;

	for (i = 0; i < SRP_RQ_SIZE; i++) {
		struct srp_iu *iu = target->rx_ring[i];
		ret = srp_post_recv(target, iu);
		if (ret)
			goto error_free;
	}

	qp_attr->qp_state = IB_QPS_RTS;
	ret = ib_cm_init_qp_attr(cm_id, qp_attr, &attr_mask);
	if (ret)
		goto error_free;

1531 1532
	target->rq_tmo_jiffies = srp_compute_rq_tmo(qp_attr, attr_mask);

1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545
	ret = ib_modify_qp(target->qp, qp_attr, attr_mask);
	if (ret)
		goto error_free;

	ret = ib_send_cm_rtu(cm_id, NULL, 0);

error_free:
	kfree(qp_attr);

error:
	target->status = ret;
}

1546 1547 1548 1549
static void srp_cm_rej_handler(struct ib_cm_id *cm_id,
			       struct ib_cm_event *event,
			       struct srp_target_port *target)
{
1550
	struct Scsi_Host *shost = target->scsi_host;
1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566
	struct ib_class_port_info *cpi;
	int opcode;

	switch (event->param.rej_rcvd.reason) {
	case IB_CM_REJ_PORT_CM_REDIRECT:
		cpi = event->param.rej_rcvd.ari;
		target->path.dlid = cpi->redirect_lid;
		target->path.pkey = cpi->redirect_pkey;
		cm_id->remote_cm_qpn = be32_to_cpu(cpi->redirect_qp) & 0x00ffffff;
		memcpy(target->path.dgid.raw, cpi->redirect_gid, 16);

		target->status = target->path.dlid ?
			SRP_DLID_REDIRECT : SRP_PORT_REDIRECT;
		break;

	case IB_CM_REJ_PORT_REDIRECT:
1567
		if (srp_target_is_topspin(target)) {
1568 1569 1570 1571 1572 1573 1574 1575
			/*
			 * Topspin/Cisco SRP gateways incorrectly send
			 * reject reason code 25 when they mean 24
			 * (port redirect).
			 */
			memcpy(target->path.dgid.raw,
			       event->param.rej_rcvd.ari, 16);

1576 1577 1578 1579
			shost_printk(KERN_DEBUG, shost,
				     PFX "Topspin/Cisco redirect to target port GID %016llx%016llx\n",
				     (unsigned long long) be64_to_cpu(target->path.dgid.global.subnet_prefix),
				     (unsigned long long) be64_to_cpu(target->path.dgid.global.interface_id));
1580 1581 1582

			target->status = SRP_PORT_REDIRECT;
		} else {
1583 1584
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_PORT_REDIRECT\n");
1585 1586 1587 1588 1589
			target->status = -ECONNRESET;
		}
		break;

	case IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID:
1590 1591
		shost_printk(KERN_WARNING, shost,
			    "  REJ reason: IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID\n");
1592 1593 1594 1595 1596 1597 1598 1599 1600 1601
		target->status = -ECONNRESET;
		break;

	case IB_CM_REJ_CONSUMER_DEFINED:
		opcode = *(u8 *) event->private_data;
		if (opcode == SRP_LOGIN_REJ) {
			struct srp_login_rej *rej = event->private_data;
			u32 reason = be32_to_cpu(rej->reason);

			if (reason == SRP_LOGIN_REJ_REQ_IT_IU_LENGTH_TOO_LARGE)
1602 1603
				shost_printk(KERN_WARNING, shost,
					     PFX "SRP_LOGIN_REJ: requested max_it_iu_len too large\n");
1604
			else
1605 1606
				shost_printk(KERN_WARNING, shost,
					    PFX "SRP LOGIN REJECTED, reason 0x%08x\n", reason);
1607
		} else
1608 1609 1610
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_CONSUMER_DEFINED,"
				     " opcode 0x%02x\n", opcode);
1611 1612 1613
		target->status = -ECONNRESET;
		break;

D
David Dillow 已提交
1614 1615 1616 1617 1618
	case IB_CM_REJ_STALE_CONN:
		shost_printk(KERN_WARNING, shost, "  REJ reason: stale connection\n");
		target->status = SRP_STALE_CONN;
		break;

1619
	default:
1620 1621
		shost_printk(KERN_WARNING, shost, "  REJ reason 0x%x\n",
			     event->param.rej_rcvd.reason);
1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632
		target->status = -ECONNRESET;
	}
}

static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event)
{
	struct srp_target_port *target = cm_id->context;
	int comp = 0;

	switch (event->event) {
	case IB_CM_REQ_ERROR:
1633 1634
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Sending CM REQ failed\n");
1635 1636 1637 1638 1639 1640
		comp = 1;
		target->status = -ECONNRESET;
		break;

	case IB_CM_REP_RECEIVED:
		comp = 1;
1641
		srp_cm_rep_handler(cm_id, event->private_data, target);
1642 1643 1644
		break;

	case IB_CM_REJ_RECEIVED:
1645
		shost_printk(KERN_DEBUG, target->scsi_host, PFX "REJ received\n");
1646 1647 1648 1649 1650
		comp = 1;

		srp_cm_rej_handler(cm_id, event, target);
		break;

1651
	case IB_CM_DREQ_RECEIVED:
1652 1653
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "DREQ received - connection closed\n");
1654
		srp_change_conn_state(target, false);
1655
		if (ib_send_cm_drep(cm_id, NULL, 0))
1656 1657
			shost_printk(KERN_ERR, target->scsi_host,
				     PFX "Sending CM DREP failed\n");
1658 1659 1660
		break;

	case IB_CM_TIMEWAIT_EXIT:
1661 1662
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "connection closed\n");
1663 1664 1665 1666

		target->status = 0;
		break;

1667 1668 1669 1670 1671
	case IB_CM_MRA_RECEIVED:
	case IB_CM_DREQ_ERROR:
	case IB_CM_DREP_RECEIVED:
		break;

1672
	default:
1673 1674
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled CM event %d\n", event->event);
1675 1676 1677 1678 1679 1680 1681 1682 1683
		break;
	}

	if (comp)
		complete(&target->done);

	return 0;
}

1684
static int srp_send_tsk_mgmt(struct srp_target_port *target,
1685
			     u64 req_tag, unsigned int lun, u8 func)
1686
{
1687
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1688 1689 1690
	struct srp_iu *iu;
	struct srp_tsk_mgmt *tsk_mgmt;

1691
	init_completion(&target->tsk_mgmt_done);
1692

1693
	spin_lock_irq(&target->lock);
1694
	iu = __srp_get_tx_iu(target, SRP_IU_TSK_MGMT);
1695
	spin_unlock_irq(&target->lock);
1696

1697
	if (!iu)
1698
		return -1;
1699

1700 1701
	ib_dma_sync_single_for_cpu(dev, iu->dma, sizeof *tsk_mgmt,
				   DMA_TO_DEVICE);
1702 1703 1704 1705
	tsk_mgmt = iu->buf;
	memset(tsk_mgmt, 0, sizeof *tsk_mgmt);

	tsk_mgmt->opcode 	= SRP_TSK_MGMT;
1706 1707
	tsk_mgmt->lun		= cpu_to_be64((u64) lun << 48);
	tsk_mgmt->tag		= req_tag | SRP_TAG_TSK_MGMT;
1708
	tsk_mgmt->tsk_mgmt_func = func;
1709
	tsk_mgmt->task_tag	= req_tag;
1710

1711 1712
	ib_dma_sync_single_for_device(dev, iu->dma, sizeof *tsk_mgmt,
				      DMA_TO_DEVICE);
1713 1714 1715 1716
	if (srp_post_send(target, iu, sizeof *tsk_mgmt)) {
		srp_put_tx_iu(target, iu, SRP_IU_TSK_MGMT);
		return -1;
	}
1717

1718
	if (!wait_for_completion_timeout(&target->tsk_mgmt_done,
1719
					 msecs_to_jiffies(SRP_ABORT_TIMEOUT_MS)))
1720
		return -1;
1721

1722 1723 1724
	return 0;
}

1725 1726
static int srp_abort(struct scsi_cmnd *scmnd)
{
1727
	struct srp_target_port *target = host_to_target(scmnd->device->host);
1728
	struct srp_request *req = (struct srp_request *) scmnd->host_scribble;
1729

1730
	shost_printk(KERN_ERR, target->scsi_host, "SRP abort called\n");
1731

B
Bart Van Assche 已提交
1732
	if (!req || target->qp_in_error || !srp_claim_req(target, req, scmnd))
1733
		return FAILED;
B
Bart Van Assche 已提交
1734 1735 1736 1737
	srp_send_tsk_mgmt(target, req->index, scmnd->device->lun,
			  SRP_TSK_ABORT_TASK);
	srp_free_req(target, req, scmnd, 0);
	scmnd->result = DID_ABORT << 16;
1738
	scmnd->scsi_done(scmnd);
1739

B
Bart Van Assche 已提交
1740
	return SUCCESS;
1741 1742 1743 1744
}

static int srp_reset_device(struct scsi_cmnd *scmnd)
{
1745
	struct srp_target_port *target = host_to_target(scmnd->device->host);
1746
	int i;
1747

1748
	shost_printk(KERN_ERR, target->scsi_host, "SRP reset_device called\n");
1749

1750 1751
	if (target->qp_in_error)
		return FAILED;
1752 1753
	if (srp_send_tsk_mgmt(target, SRP_TAG_NO_REQ, scmnd->device->lun,
			      SRP_TSK_LUN_RESET))
1754
		return FAILED;
1755
	if (target->tsk_mgmt_status)
1756 1757
		return FAILED;

1758 1759
	for (i = 0; i < SRP_CMD_SQ_SIZE; ++i) {
		struct srp_request *req = &target->req_ring[i];
1760
		if (req->scmnd && req->scmnd->device == scmnd->device)
1761
			srp_reset_req(target, req);
1762
	}
1763 1764

	return SUCCESS;
1765 1766 1767 1768 1769 1770 1771
}

static int srp_reset_host(struct scsi_cmnd *scmnd)
{
	struct srp_target_port *target = host_to_target(scmnd->device->host);
	int ret = FAILED;

1772
	shost_printk(KERN_ERR, target->scsi_host, PFX "SRP reset_host called\n");
1773 1774 1775 1776 1777 1778 1779

	if (!srp_reconnect_target(target))
		ret = SUCCESS;

	return ret;
}

1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794
static int srp_slave_configure(struct scsi_device *sdev)
{
	struct Scsi_Host *shost = sdev->host;
	struct srp_target_port *target = host_to_target(shost);
	struct request_queue *q = sdev->request_queue;
	unsigned long timeout;

	if (sdev->type == TYPE_DISK) {
		timeout = max_t(unsigned, 30 * HZ, target->rq_tmo_jiffies);
		blk_queue_rq_timeout(q, timeout);
	}

	return 0;
}

1795 1796
static ssize_t show_id_ext(struct device *dev, struct device_attribute *attr,
			   char *buf)
1797
{
1798
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1799 1800 1801 1802 1803

	return sprintf(buf, "0x%016llx\n",
		       (unsigned long long) be64_to_cpu(target->id_ext));
}

1804 1805
static ssize_t show_ioc_guid(struct device *dev, struct device_attribute *attr,
			     char *buf)
1806
{
1807
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1808 1809 1810 1811 1812

	return sprintf(buf, "0x%016llx\n",
		       (unsigned long long) be64_to_cpu(target->ioc_guid));
}

1813 1814
static ssize_t show_service_id(struct device *dev,
			       struct device_attribute *attr, char *buf)
1815
{
1816
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1817 1818 1819 1820 1821

	return sprintf(buf, "0x%016llx\n",
		       (unsigned long long) be64_to_cpu(target->service_id));
}

1822 1823
static ssize_t show_pkey(struct device *dev, struct device_attribute *attr,
			 char *buf)
1824
{
1825
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1826 1827 1828 1829

	return sprintf(buf, "0x%04x\n", be16_to_cpu(target->path.pkey));
}

1830 1831
static ssize_t show_dgid(struct device *dev, struct device_attribute *attr,
			 char *buf)
1832
{
1833
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1834

H
Harvey Harrison 已提交
1835
	return sprintf(buf, "%pI6\n", target->path.dgid.raw);
1836 1837
}

1838 1839
static ssize_t show_orig_dgid(struct device *dev,
			      struct device_attribute *attr, char *buf)
1840
{
1841
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1842

H
Harvey Harrison 已提交
1843
	return sprintf(buf, "%pI6\n", target->orig_dgid);
1844 1845
}

1846 1847 1848 1849 1850 1851 1852 1853
static ssize_t show_req_lim(struct device *dev,
			    struct device_attribute *attr, char *buf)
{
	struct srp_target_port *target = host_to_target(class_to_shost(dev));

	return sprintf(buf, "%d\n", target->req_lim);
}

1854 1855
static ssize_t show_zero_req_lim(struct device *dev,
				 struct device_attribute *attr, char *buf)
1856
{
1857
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1858 1859 1860 1861

	return sprintf(buf, "%d\n", target->zero_req_lim);
}

1862 1863
static ssize_t show_local_ib_port(struct device *dev,
				  struct device_attribute *attr, char *buf)
1864
{
1865
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1866 1867 1868 1869

	return sprintf(buf, "%d\n", target->srp_host->port);
}

1870 1871
static ssize_t show_local_ib_device(struct device *dev,
				    struct device_attribute *attr, char *buf)
1872
{
1873
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1874

1875
	return sprintf(buf, "%s\n", target->srp_host->srp_dev->dev->name);
1876 1877
}

1878 1879 1880 1881 1882 1883 1884 1885
static ssize_t show_cmd_sg_entries(struct device *dev,
				   struct device_attribute *attr, char *buf)
{
	struct srp_target_port *target = host_to_target(class_to_shost(dev));

	return sprintf(buf, "%u\n", target->cmd_sg_cnt);
}

1886 1887 1888 1889 1890 1891 1892 1893
static ssize_t show_allow_ext_sg(struct device *dev,
				 struct device_attribute *attr, char *buf)
{
	struct srp_target_port *target = host_to_target(class_to_shost(dev));

	return sprintf(buf, "%s\n", target->allow_ext_sg ? "true" : "false");
}

1894 1895 1896 1897 1898 1899
static DEVICE_ATTR(id_ext,	    S_IRUGO, show_id_ext,	   NULL);
static DEVICE_ATTR(ioc_guid,	    S_IRUGO, show_ioc_guid,	   NULL);
static DEVICE_ATTR(service_id,	    S_IRUGO, show_service_id,	   NULL);
static DEVICE_ATTR(pkey,	    S_IRUGO, show_pkey,		   NULL);
static DEVICE_ATTR(dgid,	    S_IRUGO, show_dgid,		   NULL);
static DEVICE_ATTR(orig_dgid,	    S_IRUGO, show_orig_dgid,	   NULL);
1900
static DEVICE_ATTR(req_lim,         S_IRUGO, show_req_lim,         NULL);
1901 1902 1903
static DEVICE_ATTR(zero_req_lim,    S_IRUGO, show_zero_req_lim,	   NULL);
static DEVICE_ATTR(local_ib_port,   S_IRUGO, show_local_ib_port,   NULL);
static DEVICE_ATTR(local_ib_device, S_IRUGO, show_local_ib_device, NULL);
1904
static DEVICE_ATTR(cmd_sg_entries,  S_IRUGO, show_cmd_sg_entries,  NULL);
1905
static DEVICE_ATTR(allow_ext_sg,    S_IRUGO, show_allow_ext_sg,    NULL);
1906 1907 1908 1909 1910 1911 1912 1913

static struct device_attribute *srp_host_attrs[] = {
	&dev_attr_id_ext,
	&dev_attr_ioc_guid,
	&dev_attr_service_id,
	&dev_attr_pkey,
	&dev_attr_dgid,
	&dev_attr_orig_dgid,
1914
	&dev_attr_req_lim,
1915 1916 1917
	&dev_attr_zero_req_lim,
	&dev_attr_local_ib_port,
	&dev_attr_local_ib_device,
1918
	&dev_attr_cmd_sg_entries,
1919
	&dev_attr_allow_ext_sg,
1920 1921 1922
	NULL
};

1923 1924
static struct scsi_host_template srp_template = {
	.module				= THIS_MODULE,
R
Roland Dreier 已提交
1925 1926
	.name				= "InfiniBand SRP initiator",
	.proc_name			= DRV_NAME,
1927
	.slave_configure		= srp_slave_configure,
1928 1929 1930 1931 1932
	.info				= srp_target_info,
	.queuecommand			= srp_queuecommand,
	.eh_abort_handler		= srp_abort,
	.eh_device_reset_handler	= srp_reset_device,
	.eh_host_reset_handler		= srp_reset_host,
1933
	.sg_tablesize			= SRP_DEF_SG_TABLESIZE,
1934
	.can_queue			= SRP_CMD_SQ_SIZE,
1935
	.this_id			= -1,
1936
	.cmd_per_lun			= SRP_CMD_SQ_SIZE,
1937 1938
	.use_clustering			= ENABLE_CLUSTERING,
	.shost_attrs			= srp_host_attrs
1939 1940 1941 1942
};

static int srp_add_target(struct srp_host *host, struct srp_target_port *target)
{
1943 1944 1945
	struct srp_rport_identifiers ids;
	struct srp_rport *rport;

1946 1947 1948
	sprintf(target->target_name, "SRP.T10:%016llX",
		 (unsigned long long) be64_to_cpu(target->id_ext));

1949
	if (scsi_add_host(target->scsi_host, host->srp_dev->dev->dma_device))
1950 1951
		return -ENODEV;

1952 1953
	memcpy(ids.port_id, &target->id_ext, 8);
	memcpy(ids.port_id + 8, &target->ioc_guid, 8);
1954
	ids.roles = SRP_RPORT_ROLE_TARGET;
1955 1956 1957 1958 1959 1960
	rport = srp_rport_add(target->scsi_host, &ids);
	if (IS_ERR(rport)) {
		scsi_remove_host(target->scsi_host);
		return PTR_ERR(rport);
	}

1961
	spin_lock(&host->target_lock);
1962
	list_add_tail(&target->list, &host->target_list);
1963
	spin_unlock(&host->target_lock);
1964 1965

	target->state = SRP_TARGET_LIVE;
1966
	target->connected = false;
1967 1968

	scsi_scan_target(&target->scsi_host->shost_gendev,
1969
			 0, target->scsi_id, SCAN_WILD_CARD, 0);
1970 1971 1972 1973

	return 0;
}

1974
static void srp_release_dev(struct device *dev)
1975 1976
{
	struct srp_host *host =
1977
		container_of(dev, struct srp_host, dev);
1978 1979 1980 1981 1982 1983

	complete(&host->released);
}

static struct class srp_class = {
	.name    = "infiniband_srp",
1984
	.dev_release = srp_release_dev
1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002
};

/*
 * Target ports are added by writing
 *
 *     id_ext=<SRP ID ext>,ioc_guid=<SRP IOC GUID>,dgid=<dest GID>,
 *     pkey=<P_Key>,service_id=<service ID>
 *
 * to the add_target sysfs attribute.
 */
enum {
	SRP_OPT_ERR		= 0,
	SRP_OPT_ID_EXT		= 1 << 0,
	SRP_OPT_IOC_GUID	= 1 << 1,
	SRP_OPT_DGID		= 1 << 2,
	SRP_OPT_PKEY		= 1 << 3,
	SRP_OPT_SERVICE_ID	= 1 << 4,
	SRP_OPT_MAX_SECT	= 1 << 5,
2003
	SRP_OPT_MAX_CMD_PER_LUN	= 1 << 6,
2004
	SRP_OPT_IO_CLASS	= 1 << 7,
2005
	SRP_OPT_INITIATOR_EXT	= 1 << 8,
2006
	SRP_OPT_CMD_SG_ENTRIES	= 1 << 9,
2007 2008
	SRP_OPT_ALLOW_EXT_SG	= 1 << 10,
	SRP_OPT_SG_TABLESIZE	= 1 << 11,
2009 2010 2011 2012 2013 2014 2015
	SRP_OPT_ALL		= (SRP_OPT_ID_EXT	|
				   SRP_OPT_IOC_GUID	|
				   SRP_OPT_DGID		|
				   SRP_OPT_PKEY		|
				   SRP_OPT_SERVICE_ID),
};

2016
static const match_table_t srp_opt_tokens = {
2017 2018 2019 2020 2021 2022 2023
	{ SRP_OPT_ID_EXT,		"id_ext=%s" 		},
	{ SRP_OPT_IOC_GUID,		"ioc_guid=%s" 		},
	{ SRP_OPT_DGID,			"dgid=%s" 		},
	{ SRP_OPT_PKEY,			"pkey=%x" 		},
	{ SRP_OPT_SERVICE_ID,		"service_id=%s"		},
	{ SRP_OPT_MAX_SECT,		"max_sect=%d" 		},
	{ SRP_OPT_MAX_CMD_PER_LUN,	"max_cmd_per_lun=%d" 	},
2024
	{ SRP_OPT_IO_CLASS,		"io_class=%x"		},
2025
	{ SRP_OPT_INITIATOR_EXT,	"initiator_ext=%s"	},
2026
	{ SRP_OPT_CMD_SG_ENTRIES,	"cmd_sg_entries=%u"	},
2027 2028
	{ SRP_OPT_ALLOW_EXT_SG,		"allow_ext_sg=%u"	},
	{ SRP_OPT_SG_TABLESIZE,		"sg_tablesize=%u"	},
2029
	{ SRP_OPT_ERR,			NULL 			}
2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057
};

static int srp_parse_options(const char *buf, struct srp_target_port *target)
{
	char *options, *sep_opt;
	char *p;
	char dgid[3];
	substring_t args[MAX_OPT_ARGS];
	int opt_mask = 0;
	int token;
	int ret = -EINVAL;
	int i;

	options = kstrdup(buf, GFP_KERNEL);
	if (!options)
		return -ENOMEM;

	sep_opt = options;
	while ((p = strsep(&sep_opt, ",")) != NULL) {
		if (!*p)
			continue;

		token = match_token(p, srp_opt_tokens, args);
		opt_mask |= token;

		switch (token) {
		case SRP_OPT_ID_EXT:
			p = match_strdup(args);
2058 2059 2060 2061
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2062 2063 2064 2065 2066 2067
			target->id_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_IOC_GUID:
			p = match_strdup(args);
2068 2069 2070 2071
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2072 2073 2074 2075 2076 2077
			target->ioc_guid = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_DGID:
			p = match_strdup(args);
2078 2079 2080 2081
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2082
			if (strlen(p) != 32) {
2083
				pr_warn("bad dest GID parameter '%s'\n", p);
2084
				kfree(p);
2085 2086 2087 2088 2089 2090 2091
				goto out;
			}

			for (i = 0; i < 16; ++i) {
				strlcpy(dgid, p + i * 2, 3);
				target->path.dgid.raw[i] = simple_strtoul(dgid, NULL, 16);
			}
2092
			kfree(p);
2093
			memcpy(target->orig_dgid, target->path.dgid.raw, 16);
2094 2095 2096 2097
			break;

		case SRP_OPT_PKEY:
			if (match_hex(args, &token)) {
2098
				pr_warn("bad P_Key parameter '%s'\n", p);
2099 2100 2101 2102 2103 2104 2105
				goto out;
			}
			target->path.pkey = cpu_to_be16(token);
			break;

		case SRP_OPT_SERVICE_ID:
			p = match_strdup(args);
2106 2107 2108 2109
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2110
			target->service_id = cpu_to_be64(simple_strtoull(p, NULL, 16));
2111
			target->path.service_id = target->service_id;
2112 2113 2114 2115 2116
			kfree(p);
			break;

		case SRP_OPT_MAX_SECT:
			if (match_int(args, &token)) {
2117
				pr_warn("bad max sect parameter '%s'\n", p);
2118 2119 2120 2121 2122
				goto out;
			}
			target->scsi_host->max_sectors = token;
			break;

2123 2124
		case SRP_OPT_MAX_CMD_PER_LUN:
			if (match_int(args, &token)) {
2125 2126
				pr_warn("bad max cmd_per_lun parameter '%s'\n",
					p);
2127 2128
				goto out;
			}
2129
			target->scsi_host->cmd_per_lun = min(token, SRP_CMD_SQ_SIZE);
2130 2131
			break;

2132 2133
		case SRP_OPT_IO_CLASS:
			if (match_hex(args, &token)) {
2134
				pr_warn("bad IO class parameter '%s'\n", p);
2135 2136 2137 2138
				goto out;
			}
			if (token != SRP_REV10_IB_IO_CLASS &&
			    token != SRP_REV16A_IB_IO_CLASS) {
2139 2140 2141
				pr_warn("unknown IO class parameter value %x specified (use %x or %x).\n",
					token, SRP_REV10_IB_IO_CLASS,
					SRP_REV16A_IB_IO_CLASS);
2142 2143 2144 2145 2146
				goto out;
			}
			target->io_class = token;
			break;

2147 2148
		case SRP_OPT_INITIATOR_EXT:
			p = match_strdup(args);
2149 2150 2151 2152
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2153 2154 2155 2156
			target->initiator_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

2157 2158
		case SRP_OPT_CMD_SG_ENTRIES:
			if (match_int(args, &token) || token < 1 || token > 255) {
2159 2160
				pr_warn("bad max cmd_sg_entries parameter '%s'\n",
					p);
2161 2162 2163 2164 2165
				goto out;
			}
			target->cmd_sg_cnt = token;
			break;

2166 2167
		case SRP_OPT_ALLOW_EXT_SG:
			if (match_int(args, &token)) {
2168
				pr_warn("bad allow_ext_sg parameter '%s'\n", p);
2169 2170 2171 2172 2173 2174 2175 2176
				goto out;
			}
			target->allow_ext_sg = !!token;
			break;

		case SRP_OPT_SG_TABLESIZE:
			if (match_int(args, &token) || token < 1 ||
					token > SCSI_MAX_SG_CHAIN_SEGMENTS) {
2177 2178
				pr_warn("bad max sg_tablesize parameter '%s'\n",
					p);
2179 2180 2181 2182 2183
				goto out;
			}
			target->sg_tablesize = token;
			break;

2184
		default:
2185 2186
			pr_warn("unknown parameter or missing value '%s' in target creation request\n",
				p);
2187 2188 2189 2190 2191 2192 2193 2194 2195 2196
			goto out;
		}
	}

	if ((opt_mask & SRP_OPT_ALL) == SRP_OPT_ALL)
		ret = 0;
	else
		for (i = 0; i < ARRAY_SIZE(srp_opt_tokens); ++i)
			if ((srp_opt_tokens[i].token & SRP_OPT_ALL) &&
			    !(srp_opt_tokens[i].token & opt_mask))
2197 2198
				pr_warn("target creation request is missing parameter '%s'\n",
					srp_opt_tokens[i].pattern);
2199 2200 2201 2202 2203 2204

out:
	kfree(options);
	return ret;
}

2205 2206
static ssize_t srp_create_target(struct device *dev,
				 struct device_attribute *attr,
2207 2208 2209
				 const char *buf, size_t count)
{
	struct srp_host *host =
2210
		container_of(dev, struct srp_host, dev);
2211 2212
	struct Scsi_Host *target_host;
	struct srp_target_port *target;
2213 2214
	struct ib_device *ibdev = host->srp_dev->dev;
	dma_addr_t dma_addr;
2215
	int i, ret;
2216 2217 2218 2219 2220 2221

	target_host = scsi_host_alloc(&srp_template,
				      sizeof (struct srp_target_port));
	if (!target_host)
		return -ENOMEM;

2222
	target_host->transportt  = ib_srp_transport_template;
2223 2224
	target_host->max_channel = 0;
	target_host->max_id      = 1;
A
Arne Redlich 已提交
2225 2226
	target_host->max_lun     = SRP_MAX_LUN;
	target_host->max_cmd_len = sizeof ((struct srp_cmd *) (void *) 0L)->cdb;
R
Roland Dreier 已提交
2227

2228 2229
	target = host_to_target(target_host);

2230 2231 2232 2233 2234 2235
	target->io_class	= SRP_REV16A_IB_IO_CLASS;
	target->scsi_host	= target_host;
	target->srp_host	= host;
	target->lkey		= host->srp_dev->mr->lkey;
	target->rkey		= host->srp_dev->mr->rkey;
	target->cmd_sg_cnt	= cmd_sg_entries;
2236 2237
	target->sg_tablesize	= indirect_sg_entries ? : cmd_sg_entries;
	target->allow_ext_sg	= allow_ext_sg;
2238 2239 2240 2241 2242

	ret = srp_parse_options(buf, target);
	if (ret)
		goto err;

2243 2244
	if (!host->srp_dev->fmr_pool && !target->allow_ext_sg &&
				target->cmd_sg_cnt < target->sg_tablesize) {
2245
		pr_warn("No FMR pool and no external indirect descriptors, limiting sg_tablesize to cmd_sg_cnt\n");
2246 2247 2248 2249 2250 2251
		target->sg_tablesize = target->cmd_sg_cnt;
	}

	target_host->sg_tablesize = target->sg_tablesize;
	target->indirect_size = target->sg_tablesize *
				sizeof (struct srp_direct_buf);
2252 2253 2254 2255
	target->max_iu_len = sizeof (struct srp_cmd) +
			     sizeof (struct srp_indirect_buf) +
			     target->cmd_sg_cnt * sizeof (struct srp_direct_buf);

2256
	INIT_WORK(&target->remove_work, srp_remove_work);
2257 2258 2259 2260 2261 2262 2263 2264 2265 2266
	spin_lock_init(&target->lock);
	INIT_LIST_HEAD(&target->free_tx);
	INIT_LIST_HEAD(&target->free_reqs);
	for (i = 0; i < SRP_CMD_SQ_SIZE; ++i) {
		struct srp_request *req = &target->req_ring[i];

		req->fmr_list = kmalloc(target->cmd_sg_cnt * sizeof (void *),
					GFP_KERNEL);
		req->map_page = kmalloc(SRP_FMR_SIZE * sizeof (void *),
					GFP_KERNEL);
2267 2268
		req->indirect_desc = kmalloc(target->indirect_size, GFP_KERNEL);
		if (!req->fmr_list || !req->map_page || !req->indirect_desc)
2269 2270
			goto err_free_mem;

2271 2272 2273 2274 2275 2276 2277
		dma_addr = ib_dma_map_single(ibdev, req->indirect_desc,
					     target->indirect_size,
					     DMA_TO_DEVICE);
		if (ib_dma_mapping_error(ibdev, dma_addr))
			goto err_free_mem;

		req->indirect_dma_addr = dma_addr;
2278 2279 2280 2281
		req->index = i;
		list_add_tail(&req->list, &target->free_reqs);
	}

2282
	ib_query_gid(ibdev, host->port, 0, &target->path.sgid);
2283

2284 2285
	shost_printk(KERN_DEBUG, target->scsi_host, PFX
		     "new target: id_ext %016llx ioc_guid %016llx pkey %04x "
H
Harvey Harrison 已提交
2286
		     "service_id %016llx dgid %pI6\n",
2287 2288 2289 2290
	       (unsigned long long) be64_to_cpu(target->id_ext),
	       (unsigned long long) be64_to_cpu(target->ioc_guid),
	       be16_to_cpu(target->path.pkey),
	       (unsigned long long) be64_to_cpu(target->service_id),
2291
	       target->path.dgid.raw);
2292 2293 2294

	ret = srp_create_target_ib(target);
	if (ret)
2295
		goto err_free_mem;
2296

D
David Dillow 已提交
2297 2298
	ret = srp_new_cm_id(target);
	if (ret)
2299
		goto err_free_ib;
2300 2301 2302

	ret = srp_connect_target(target);
	if (ret) {
2303 2304
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Connection failed\n");
2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319
		goto err_cm_id;
	}

	ret = srp_add_target(host, target);
	if (ret)
		goto err_disconnect;

	return count;

err_disconnect:
	srp_disconnect_target(target);

err_cm_id:
	ib_destroy_cm_id(target->cm_id);

2320
err_free_ib:
2321 2322
	srp_free_target_ib(target);

2323 2324 2325
err_free_mem:
	srp_free_req_data(target);

2326 2327 2328 2329 2330 2331
err:
	scsi_host_put(target_host);

	return ret;
}

2332
static DEVICE_ATTR(add_target, S_IWUSR, NULL, srp_create_target);
2333

2334 2335
static ssize_t show_ibdev(struct device *dev, struct device_attribute *attr,
			  char *buf)
2336
{
2337
	struct srp_host *host = container_of(dev, struct srp_host, dev);
2338

2339
	return sprintf(buf, "%s\n", host->srp_dev->dev->name);
2340 2341
}

2342
static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
2343

2344 2345
static ssize_t show_port(struct device *dev, struct device_attribute *attr,
			 char *buf)
2346
{
2347
	struct srp_host *host = container_of(dev, struct srp_host, dev);
2348 2349 2350 2351

	return sprintf(buf, "%d\n", host->port);
}

2352
static DEVICE_ATTR(port, S_IRUGO, show_port, NULL);
2353

2354
static struct srp_host *srp_add_port(struct srp_device *device, u8 port)
2355 2356 2357 2358 2359 2360 2361 2362
{
	struct srp_host *host;

	host = kzalloc(sizeof *host, GFP_KERNEL);
	if (!host)
		return NULL;

	INIT_LIST_HEAD(&host->target_list);
2363
	spin_lock_init(&host->target_lock);
2364
	init_completion(&host->released);
2365
	host->srp_dev = device;
2366 2367
	host->port = port;

2368 2369
	host->dev.class = &srp_class;
	host->dev.parent = device->dev->dma_device;
2370
	dev_set_name(&host->dev, "srp-%s-%d", device->dev->name, port);
2371

2372
	if (device_register(&host->dev))
2373
		goto free_host;
2374
	if (device_create_file(&host->dev, &dev_attr_add_target))
2375
		goto err_class;
2376
	if (device_create_file(&host->dev, &dev_attr_ibdev))
2377
		goto err_class;
2378
	if (device_create_file(&host->dev, &dev_attr_port))
2379 2380 2381 2382 2383
		goto err_class;

	return host;

err_class:
2384
	device_unregister(&host->dev);
2385

2386
free_host:
2387 2388 2389 2390 2391 2392 2393
	kfree(host);

	return NULL;
}

static void srp_add_one(struct ib_device *device)
{
2394 2395 2396
	struct srp_device *srp_dev;
	struct ib_device_attr *dev_attr;
	struct ib_fmr_pool_param fmr_param;
2397
	struct srp_host *host;
2398
	int max_pages_per_fmr, fmr_page_shift, s, e, p;
2399

2400 2401
	dev_attr = kmalloc(sizeof *dev_attr, GFP_KERNEL);
	if (!dev_attr)
2402
		return;
2403

2404
	if (ib_query_device(device, dev_attr)) {
2405
		pr_warn("Query device failed for %s\n", device->name);
2406 2407 2408 2409 2410 2411 2412 2413 2414
		goto free_attr;
	}

	srp_dev = kmalloc(sizeof *srp_dev, GFP_KERNEL);
	if (!srp_dev)
		goto free_attr;

	/*
	 * Use the smallest page size supported by the HCA, down to a
2415 2416
	 * minimum of 4096 bytes. We're unlikely to build large sglists
	 * out of smaller entries.
2417
	 */
2418 2419 2420 2421
	fmr_page_shift		= max(12, ffs(dev_attr->page_size_cap) - 1);
	srp_dev->fmr_page_size	= 1 << fmr_page_shift;
	srp_dev->fmr_page_mask	= ~((u64) srp_dev->fmr_page_size - 1);
	srp_dev->fmr_max_size	= srp_dev->fmr_page_size * SRP_FMR_SIZE;
2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436

	INIT_LIST_HEAD(&srp_dev->dev_list);

	srp_dev->dev = device;
	srp_dev->pd  = ib_alloc_pd(device);
	if (IS_ERR(srp_dev->pd))
		goto free_dev;

	srp_dev->mr = ib_get_dma_mr(srp_dev->pd,
				    IB_ACCESS_LOCAL_WRITE |
				    IB_ACCESS_REMOTE_READ |
				    IB_ACCESS_REMOTE_WRITE);
	if (IS_ERR(srp_dev->mr))
		goto err_pd;

2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454
	for (max_pages_per_fmr = SRP_FMR_SIZE;
			max_pages_per_fmr >= SRP_FMR_MIN_SIZE;
			max_pages_per_fmr /= 2, srp_dev->fmr_max_size /= 2) {
		memset(&fmr_param, 0, sizeof fmr_param);
		fmr_param.pool_size	    = SRP_FMR_POOL_SIZE;
		fmr_param.dirty_watermark   = SRP_FMR_DIRTY_SIZE;
		fmr_param.cache		    = 1;
		fmr_param.max_pages_per_fmr = max_pages_per_fmr;
		fmr_param.page_shift	    = fmr_page_shift;
		fmr_param.access	    = (IB_ACCESS_LOCAL_WRITE |
					       IB_ACCESS_REMOTE_WRITE |
					       IB_ACCESS_REMOTE_READ);

		srp_dev->fmr_pool = ib_create_fmr_pool(srp_dev->pd, &fmr_param);
		if (!IS_ERR(srp_dev->fmr_pool))
			break;
	}

2455 2456
	if (IS_ERR(srp_dev->fmr_pool))
		srp_dev->fmr_pool = NULL;
2457

T
Tom Tucker 已提交
2458
	if (device->node_type == RDMA_NODE_IB_SWITCH) {
2459 2460 2461 2462 2463 2464 2465 2466
		s = 0;
		e = 0;
	} else {
		s = 1;
		e = device->phys_port_cnt;
	}

	for (p = s; p <= e; ++p) {
2467
		host = srp_add_port(srp_dev, p);
2468
		if (host)
2469
			list_add_tail(&host->list, &srp_dev->dev_list);
2470 2471
	}

2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483
	ib_set_client_data(device, &srp_client, srp_dev);

	goto free_attr;

err_pd:
	ib_dealloc_pd(srp_dev->pd);

free_dev:
	kfree(srp_dev);

free_attr:
	kfree(dev_attr);
2484 2485 2486 2487
}

static void srp_remove_one(struct ib_device *device)
{
2488
	struct srp_device *srp_dev;
2489
	struct srp_host *host, *tmp_host;
2490
	struct srp_target_port *target;
2491

2492
	srp_dev = ib_get_client_data(device, &srp_client);
2493

2494
	list_for_each_entry_safe(host, tmp_host, &srp_dev->dev_list, list) {
2495
		device_unregister(&host->dev);
2496 2497 2498 2499 2500 2501 2502
		/*
		 * Wait for the sysfs entry to go away, so that no new
		 * target ports can be created.
		 */
		wait_for_completion(&host->released);

		/*
2503
		 * Remove all target ports.
2504
		 */
2505
		spin_lock(&host->target_lock);
2506 2507
		list_for_each_entry(target, &host->target_list, list)
			srp_queue_remove_work(target);
2508
		spin_unlock(&host->target_lock);
2509 2510

		/*
2511
		 * Wait for target port removal tasks.
2512
		 */
2513
		flush_workqueue(system_long_wq);
2514 2515 2516 2517

		kfree(host);
	}

2518 2519 2520 2521 2522 2523
	if (srp_dev->fmr_pool)
		ib_destroy_fmr_pool(srp_dev->fmr_pool);
	ib_dereg_mr(srp_dev->mr);
	ib_dealloc_pd(srp_dev->pd);

	kfree(srp_dev);
2524 2525
}

2526 2527 2528
static struct srp_function_template ib_srp_transport_functions = {
};

2529 2530 2531 2532
static int __init srp_init_module(void)
{
	int ret;

2533
	BUILD_BUG_ON(FIELD_SIZEOF(struct ib_wc, wr_id) < sizeof(void *));
2534

2535
	if (srp_sg_tablesize) {
2536
		pr_warn("srp_sg_tablesize is deprecated, please use cmd_sg_entries\n");
2537 2538 2539 2540 2541 2542 2543 2544
		if (!cmd_sg_entries)
			cmd_sg_entries = srp_sg_tablesize;
	}

	if (!cmd_sg_entries)
		cmd_sg_entries = SRP_DEF_SG_TABLESIZE;

	if (cmd_sg_entries > 255) {
2545
		pr_warn("Clamping cmd_sg_entries to 255\n");
2546
		cmd_sg_entries = 255;
2547 2548
	}

2549 2550 2551
	if (!indirect_sg_entries)
		indirect_sg_entries = cmd_sg_entries;
	else if (indirect_sg_entries < cmd_sg_entries) {
2552 2553
		pr_warn("Bumping up indirect_sg_entries to match cmd_sg_entries (%u)\n",
			cmd_sg_entries);
2554 2555 2556
		indirect_sg_entries = cmd_sg_entries;
	}

2557 2558 2559 2560 2561
	ib_srp_transport_template =
		srp_attach_transport(&ib_srp_transport_functions);
	if (!ib_srp_transport_template)
		return -ENOMEM;

2562 2563
	ret = class_register(&srp_class);
	if (ret) {
2564
		pr_err("couldn't register class infiniband_srp\n");
2565
		srp_release_transport(ib_srp_transport_template);
2566 2567 2568
		return ret;
	}

2569 2570
	ib_sa_register_client(&srp_sa_client);

2571 2572
	ret = ib_register_client(&srp_client);
	if (ret) {
2573
		pr_err("couldn't register IB client\n");
2574
		srp_release_transport(ib_srp_transport_template);
2575
		ib_sa_unregister_client(&srp_sa_client);
2576 2577 2578 2579 2580 2581 2582 2583 2584 2585
		class_unregister(&srp_class);
		return ret;
	}

	return 0;
}

static void __exit srp_cleanup_module(void)
{
	ib_unregister_client(&srp_client);
2586
	ib_sa_unregister_client(&srp_sa_client);
2587
	class_unregister(&srp_class);
2588
	srp_release_transport(ib_srp_transport_template);
2589 2590 2591 2592
}

module_init(srp_init_module);
module_exit(srp_cleanup_module);