ib_srp.c 69.9 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

#include "ib_srp.h"

#define DRV_NAME	"ib_srp"
#define PFX		DRV_NAME ": "
56 57
#define DRV_VERSION	"1.0"
#define DRV_RELDATE	"July 1, 2013"
58 59 60 61 62 63

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
	recv_cq = ib_create_cq(target->srp_host->srp_dev->dev,
234 235
			       srp_recv_completion, NULL, target, SRP_RQ_SIZE,
			       target->comp_vector);
236 237
	if (IS_ERR(recv_cq)) {
		ret = PTR_ERR(recv_cq);
238
		goto err;
239 240
	}

241
	send_cq = ib_create_cq(target->srp_host->srp_dev->dev,
242 243
			       srp_send_completion, NULL, target, SRP_SQ_SIZE,
			       target->comp_vector);
244 245
	if (IS_ERR(send_cq)) {
		ret = PTR_ERR(send_cq);
246
		goto err_recv_cq;
247 248
	}

249
	ib_req_notify_cq(recv_cq, IB_CQ_NEXT_COMP);
250 251 252 253 254 255 256 257

	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;
258 259
	init_attr->send_cq             = send_cq;
	init_attr->recv_cq             = recv_cq;
260

261 262 263
	qp = ib_create_qp(target->srp_host->srp_dev->pd, init_attr);
	if (IS_ERR(qp)) {
		ret = PTR_ERR(qp);
264
		goto err_send_cq;
265 266
	}

267
	ret = srp_init_qp(target, qp);
268 269
	if (ret)
		goto err_qp;
270

271 272 273 274 275 276 277 278 279 280 281
	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;

282 283 284 285
	kfree(init_attr);
	return 0;

err_qp:
286
	ib_destroy_qp(qp);
287 288

err_send_cq:
289
	ib_destroy_cq(send_cq);
290 291

err_recv_cq:
292
	ib_destroy_cq(recv_cq);
293 294

err:
295 296 297 298 299 300 301 302 303
	kfree(init_attr);
	return ret;
}

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

	ib_destroy_qp(target->qp);
304 305
	ib_destroy_cq(target->send_cq);
	ib_destroy_cq(target->recv_cq);
306

307 308 309
	target->qp = NULL;
	target->send_cq = target->recv_cq = NULL;

310 311
	for (i = 0; i < SRP_RQ_SIZE; ++i)
		srp_free_iu(target->srp_host, target->rx_ring[i]);
312
	for (i = 0; i < SRP_SQ_SIZE; ++i)
313 314 315 316 317 318 319 320 321 322 323
		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)
324 325
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Got failed path rec status %d\n", status);
326 327 328 329 330 331 332 333 334 335 336
	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);

337
	target->path_query_id = ib_sa_path_rec_get(&srp_sa_client,
338
						   target->srp_host->srp_dev->dev,
339 340
						   target->srp_host->port,
						   &target->path,
341
						   IB_SA_PATH_REC_SERVICE_ID	|
342 343 344 345 346 347 348 349 350 351 352 353 354 355
						   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)
356 357
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Path record query failed\n");
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

	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;
393
	req->param.retry_count                = target->tl_retry_count;
394 395 396 397 398
	req->param.rnr_retry_count 	      = 7;
	req->param.max_cm_retries 	      = 15;

	req->priv.opcode     	= SRP_LOGIN_REQ;
	req->priv.tag        	= 0;
399
	req->priv.req_it_iu_len = cpu_to_be32(target->max_iu_len);
400 401
	req->priv.req_buf_fmt 	= cpu_to_be16(SRP_BUF_FORMAT_DIRECT |
					      SRP_BUF_FORMAT_INDIRECT);
402
	/*
R
Roland Dreier 已提交
403
	 * In the published SRP specification (draft rev. 16a), the
404 405 406 407 408 409 410 411 412
	 * 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,
413
		       &target->path.sgid.global.interface_id, 8);
414
		memcpy(req->priv.initiator_port_id + 8,
415
		       &target->initiator_ext, 8);
416 417 418 419
		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,
420 421 422
		       &target->initiator_ext, 8);
		memcpy(req->priv.initiator_port_id + 8,
		       &target->path.sgid.global.interface_id, 8);
423 424 425 426
		memcpy(req->priv.target_port_id,     &target->id_ext, 8);
		memcpy(req->priv.target_port_id + 8, &target->ioc_guid, 8);
	}

427 428
	/*
	 * Topspin/Cisco SRP targets will reject our login unless we
429 430
	 * zero out the first 8 bytes of our initiator port ID and set
	 * the second 8 bytes to the local node GUID.
431
	 */
432
	if (srp_target_is_topspin(target)) {
433 434 435 436
		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));
437
		memset(req->priv.initiator_port_id, 0, 8);
438
		memcpy(req->priv.initiator_port_id + 8,
439
		       &target->srp_host->srp_dev->dev->node_guid, 8);
440 441 442 443 444 445 446 447 448
	}

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

	kfree(req);

	return status;
}

449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465
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;
}

466 467 468 469 470 471 472 473 474 475 476 477 478 479 480
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;
}

481 482
static void srp_disconnect_target(struct srp_target_port *target)
{
483 484
	if (srp_change_conn_state(target, false)) {
		/* XXX should send SRP_I_LOGOUT request */
485

486 487 488 489
		if (ib_send_cm_dreq(target->cm_id, NULL, 0)) {
			shost_printk(KERN_DEBUG, target->scsi_host,
				     PFX "Sending CM DREQ failed\n");
		}
490
	}
491 492
}

493 494
static void srp_free_req_data(struct srp_target_port *target)
{
495
	struct ib_device *ibdev = target->srp_host->srp_dev->dev;
496 497 498 499 500 501
	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);
502 503 504 505 506 507
		if (req->indirect_dma_addr) {
			ib_dma_unmap_single(ibdev, req->indirect_dma_addr,
					    target->indirect_size,
					    DMA_TO_DEVICE);
		}
		kfree(req->indirect_desc);
508 509 510
	}
}

511 512 513 514 515 516 517 518 519 520 521 522 523 524 525
/**
 * 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);
}

526 527
static void srp_remove_target(struct srp_target_port *target)
{
528 529
	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);

530 531 532
	srp_del_scsi_host_attr(target->scsi_host);
	srp_remove_host(target->scsi_host);
	scsi_remove_host(target->scsi_host);
533
	srp_disconnect_target(target);
534 535 536 537 538 539
	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
540
static void srp_remove_work(struct work_struct *work)
541
{
D
David Howells 已提交
542
	struct srp_target_port *target =
543
		container_of(work, struct srp_target_port, remove_work);
544

545
	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);
546

547 548
	srp_remove_target(target);

549
	spin_lock(&target->srp_host->target_lock);
550
	list_del(&target->list);
551
	spin_unlock(&target->srp_host->target_lock);
552 553
}

554 555 556 557 558 559 560
static void srp_rport_delete(struct srp_rport *rport)
{
	struct srp_target_port *target = rport->lld_data;

	srp_queue_remove_work(target);
}

561 562
static int srp_connect_target(struct srp_target_port *target)
{
D
David Dillow 已提交
563
	int retries = 3;
564 565
	int ret;

566 567
	WARN_ON_ONCE(target->connected);

568 569
	target->qp_in_error = false;

570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588
	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:
589
			srp_change_conn_state(target, true);
590 591 592 593 594 595 596 597 598 599 600
			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
601 602 603 604 605 606 607 608 609 610 611 612 613 614 615
		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;

616 617 618 619 620 621
		default:
			return target->status;
		}
	}
}

622 623 624 625
static void srp_unmap_data(struct scsi_cmnd *scmnd,
			   struct srp_target_port *target,
			   struct srp_request *req)
{
626 627 628
	struct ib_device *ibdev = target->srp_host->srp_dev->dev;
	struct ib_pool_fmr **pfmr;

629
	if (!scsi_sglist(scmnd) ||
630 631 632 633
	    (scmnd->sc_data_direction != DMA_TO_DEVICE &&
	     scmnd->sc_data_direction != DMA_FROM_DEVICE))
		return;

634 635 636
	pfmr = req->fmr_list;
	while (req->nfmr--)
		ib_fmr_pool_unmap(*pfmr++);
637

638 639
	ib_dma_unmap_sg(ibdev, scsi_sglist(scmnd), scsi_sg_count(scmnd),
			scmnd->sc_data_direction);
640 641
}

B
Bart Van Assche 已提交
642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677
/**
 * 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)
678
{
679 680
	unsigned long flags;

B
Bart Van Assche 已提交
681 682
	srp_unmap_data(scmnd, target, req);

683
	spin_lock_irqsave(&target->lock, flags);
684
	target->req_lim += req_lim_delta;
685
	list_add_tail(&req->list, &target->free_reqs);
686
	spin_unlock_irqrestore(&target->lock, flags);
687 688 689 690
}

static void srp_reset_req(struct srp_target_port *target, struct srp_request *req)
{
B
Bart Van Assche 已提交
691 692 693
	struct scsi_cmnd *scmnd = srp_claim_req(target, req, NULL);

	if (scmnd) {
694
		srp_free_req(target, req, scmnd, 0);
B
Bart Van Assche 已提交
695 696 697
		scmnd->result = DID_RESET << 16;
		scmnd->scsi_done(scmnd);
	}
698 699
}

700 701
static int srp_reconnect_target(struct srp_target_port *target)
{
702
	struct Scsi_Host *shost = target->scsi_host;
703
	int i, ret;
704

705 706
	scsi_target_block(&shost->shost_gendev);

707 708
	srp_disconnect_target(target);
	/*
709 710 711
	 * Now get a new local CM ID so that we avoid confusing the target in
	 * case things are really fouled up. Doing so also ensures that all CM
	 * callbacks will have finished before a new QP is allocated.
712
	 */
D
David Dillow 已提交
713
	ret = srp_new_cm_id(target);
714 715 716 717 718 719 720 721 722
	/*
	 * Whether or not creating a new CM ID succeeded, create a new
	 * QP. This guarantees that all completion callback function
	 * invocations have finished before request resetting starts.
	 */
	if (ret == 0)
		ret = srp_create_target_ib(target);
	else
		srp_create_target_ib(target);
723

724 725 726 727 728
	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);
	}
729

730
	INIT_LIST_HEAD(&target->free_tx);
731
	for (i = 0; i < SRP_SQ_SIZE; ++i)
732
		list_add(&target->tx_ring[i]->list, &target->free_tx);
733

734 735
	if (ret == 0)
		ret = srp_connect_target(target);
736 737 738

	scsi_target_unblock(&shost->shost_gendev, ret == 0 ? SDEV_RUNNING :
			    SDEV_TRANSPORT_OFFLINE);
739
	target->transport_offline = !!ret;
740

741 742 743
	if (ret)
		goto err;

744
	shost_printk(KERN_INFO, target->scsi_host, PFX "reconnect succeeded\n");
745 746 747 748

	return ret;

err:
749 750
	shost_printk(KERN_ERR, target->scsi_host,
		     PFX "reconnect failed (%d), removing target port.\n", ret);
751 752 753

	/*
	 * We couldn't reconnect, so kill our target port off.
754 755 756
	 * 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().
757
	 */
758
	srp_queue_remove_work(target);
759 760 761 762

	return ret;
}

763 764
static void srp_map_desc(struct srp_map_state *state, dma_addr_t dma_addr,
			 unsigned int dma_len, u32 rkey)
765
{
766
	struct srp_direct_buf *desc = state->desc;
767

768 769 770
	desc->va = cpu_to_be64(dma_addr);
	desc->key = cpu_to_be32(rkey);
	desc->len = cpu_to_be32(dma_len);
771

772 773 774 775
	state->total_len += dma_len;
	state->desc++;
	state->ndesc++;
}
776

777 778 779 780 781 782
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;
783

784 785
	if (!state->npages)
		return 0;
786

787 788 789 790 791
	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;
792 793
	}

794 795 796 797
	fmr = ib_fmr_pool_map_phys(dev->fmr_pool, state->pages,
				   state->npages, io_addr);
	if (IS_ERR(fmr))
		return PTR_ERR(fmr);
798

799 800
	*state->next_fmr++ = fmr;
	state->nfmr++;
801

802 803 804 805 806 807 808 809 810 811 812 813 814
	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;
}
815

816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837
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;
838
	}
839

840 841 842 843 844 845 846 847 848 849 850 851 852 853
	/* 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;
854 855
	}

856 857 858 859 860 861 862
	/* 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);
863

864 865 866 867 868
	while (dma_len) {
		if (state->npages == SRP_FMR_SIZE) {
			ret = srp_map_finish_fmr(state, target);
			if (ret)
				return ret;
869

870 871 872 873
			srp_map_update_start(state, sg, sg_index, dma_addr);
		}

		len = min_t(unsigned int, dma_len, dev->fmr_page_size);
874

875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892
		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);
	}
893 894 895
	return ret;
}

896 897 898
static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_target_port *target,
			struct srp_request *req)
{
899
	struct scatterlist *scat, *sg;
900
	struct srp_cmd *cmd = req->cmd->buf;
901
	int i, len, nents, count, use_fmr;
902 903
	struct srp_device *dev;
	struct ib_device *ibdev;
904 905 906 907
	struct srp_map_state state;
	struct srp_indirect_buf *indirect_hdr;
	u32 table_len;
	u8 fmt;
908

909
	if (!scsi_sglist(scmnd) || scmnd->sc_data_direction == DMA_NONE)
910 911 912 913
		return sizeof (struct srp_cmd);

	if (scmnd->sc_data_direction != DMA_FROM_DEVICE &&
	    scmnd->sc_data_direction != DMA_TO_DEVICE) {
914 915 916
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled data direction %d\n",
			     scmnd->sc_data_direction);
917 918 919
		return -EINVAL;
	}

920 921
	nents = scsi_sg_count(scmnd);
	scat  = scsi_sglist(scmnd);
922

923
	dev = target->srp_host->srp_dev;
924 925 926
	ibdev = dev->dev;

	count = ib_dma_map_sg(ibdev, scat, nents, scmnd->sc_data_direction);
927 928
	if (unlikely(count == 0))
		return -EIO;
929 930 931

	fmt = SRP_DATA_DESC_DIRECT;
	len = sizeof (struct srp_cmd) +	sizeof (struct srp_direct_buf);
932

933
	if (count == 1) {
934 935 936 937 938 939
		/*
		 * 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.
		 */
940
		struct srp_direct_buf *buf = (void *) cmd->add_data;
941

942
		buf->va  = cpu_to_be64(ib_sg_dma_address(ibdev, scat));
943
		buf->key = cpu_to_be32(target->rkey);
944
		buf->len = cpu_to_be32(ib_sg_dma_len(ibdev, scat));
945 946 947 948 949 950 951 952 953 954 955

		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;

956 957 958
	ib_dma_sync_single_for_cpu(ibdev, req->indirect_dma_addr,
				   target->indirect_size, DMA_TO_DEVICE);

959
	memset(&state, 0, sizeof(state));
960
	state.desc	= req->indirect_desc;
961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983
	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);
984
		}
985
	}
986

987 988
	if (use_fmr == SRP_MAP_ALLOW_FMR && srp_map_finish_fmr(&state, target))
		goto backtrack;
989

990 991 992 993 994
	/* 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.
995 996 997 998 999 1000 1001
	 */
	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;
1002

1003
		*buf = req->indirect_desc[0];
1004
		goto map_complete;
1005 1006
	}

1007 1008 1009 1010 1011 1012 1013 1014
	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);
1015 1016 1017 1018
	table_len = state.ndesc * sizeof (struct srp_direct_buf);

	fmt = SRP_DATA_DESC_INDIRECT;
	len = sizeof(struct srp_cmd) + sizeof (struct srp_indirect_buf);
1019
	len += count * sizeof (struct srp_direct_buf);
1020

1021 1022
	memcpy(indirect_hdr->desc_list, req->indirect_desc,
	       count * sizeof (struct srp_direct_buf));
1023

1024
	indirect_hdr->table_desc.va = cpu_to_be64(req->indirect_dma_addr);
1025 1026 1027 1028 1029
	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)
1030
		cmd->data_out_desc_cnt = count;
1031
	else
1032 1033 1034 1035
		cmd->data_in_desc_cnt = count;

	ib_dma_sync_single_for_device(ibdev, req->indirect_dma_addr, table_len,
				      DMA_TO_DEVICE);
1036 1037

map_complete:
1038 1039 1040 1041 1042 1043 1044 1045
	if (scmnd->sc_data_direction == DMA_TO_DEVICE)
		cmd->buf_fmt = fmt << 4;
	else
		cmd->buf_fmt = fmt;

	return len;
}

1046 1047 1048 1049 1050 1051 1052 1053
/*
 * 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;

1054
	spin_lock_irqsave(&target->lock, flags);
1055 1056 1057
	list_add(&iu->list, &target->free_tx);
	if (iu_type != SRP_IU_RSP)
		++target->req_lim;
1058
	spin_unlock_irqrestore(&target->lock, flags);
1059 1060
}

1061
/*
1062 1063
 * 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().
1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081
 *
 * 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);

1082
	if (list_empty(&target->free_tx))
1083 1084 1085
		return NULL;

	/* Initiator responses to target requests do not consume credits */
1086 1087 1088 1089 1090 1091 1092
	if (iu_type != SRP_IU_RSP) {
		if (target->req_lim <= rsv) {
			++target->zero_req_lim;
			return NULL;
		}

		--target->req_lim;
1093 1094
	}

1095
	iu = list_first_entry(&target->free_tx, struct srp_iu, list);
1096
	list_del(&iu->list);
1097 1098 1099
	return iu;
}

1100 1101
static int srp_post_send(struct srp_target_port *target,
			 struct srp_iu *iu, int len)
1102 1103 1104 1105 1106 1107
{
	struct ib_sge list;
	struct ib_send_wr wr, *bad_wr;

	list.addr   = iu->dma;
	list.length = len;
1108
	list.lkey   = target->lkey;
1109 1110

	wr.next       = NULL;
1111
	wr.wr_id      = (uintptr_t) iu;
1112 1113 1114 1115 1116
	wr.sg_list    = &list;
	wr.num_sge    = 1;
	wr.opcode     = IB_WR_SEND;
	wr.send_flags = IB_SEND_SIGNALED;

1117
	return ib_post_send(target->qp, &wr, &bad_wr);
1118 1119
}

1120
static int srp_post_recv(struct srp_target_port *target, struct srp_iu *iu)
1121 1122
{
	struct ib_recv_wr wr, *bad_wr;
1123
	struct ib_sge list;
1124 1125 1126

	list.addr   = iu->dma;
	list.length = iu->size;
1127
	list.lkey   = target->lkey;
1128 1129

	wr.next     = NULL;
1130
	wr.wr_id    = (uintptr_t) iu;
1131 1132 1133
	wr.sg_list  = &list;
	wr.num_sge  = 1;

1134
	return ib_post_recv(target->qp, &wr, &bad_wr);
1135 1136
}

1137 1138 1139 1140 1141 1142 1143
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)) {
1144
		spin_lock_irqsave(&target->lock, flags);
1145
		target->req_lim += be32_to_cpu(rsp->req_lim_delta);
1146
		spin_unlock_irqrestore(&target->lock, flags);
1147

1148 1149 1150 1151
		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);
1152
	} else {
1153
		req = &target->req_ring[rsp->tag];
B
Bart Van Assche 已提交
1154 1155
		scmnd = srp_claim_req(target, req, NULL);
		if (!scmnd) {
1156 1157 1158
			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
1159 1160 1161 1162 1163 1164 1165

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

			return;
		}
1166 1167 1168 1169 1170 1171 1172 1173 1174 1175
		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))
1176
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_out_res_cnt));
1177
		else if (rsp->flags & (SRP_RSP_FLAG_DIOVER | SRP_RSP_FLAG_DIUNDER))
1178
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_in_res_cnt));
1179

B
Bart Van Assche 已提交
1180 1181 1182
		srp_free_req(target, req, scmnd,
			     be32_to_cpu(rsp->req_lim_delta));

1183 1184
		scmnd->host_scribble = NULL;
		scmnd->scsi_done(scmnd);
1185 1186 1187
	}
}

1188 1189 1190
static int srp_response_common(struct srp_target_port *target, s32 req_delta,
			       void *rsp, int len)
{
1191
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1192 1193
	unsigned long flags;
	struct srp_iu *iu;
1194
	int err;
1195

1196
	spin_lock_irqsave(&target->lock, flags);
1197 1198
	target->req_lim += req_delta;
	iu = __srp_get_tx_iu(target, SRP_IU_RSP);
1199
	spin_unlock_irqrestore(&target->lock, flags);
1200

1201 1202 1203
	if (!iu) {
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "no IU available to send response\n");
1204
		return 1;
1205 1206 1207 1208 1209 1210
	}

	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);

1211 1212
	err = srp_post_send(target, iu, len);
	if (err) {
1213 1214
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "unable to post response: %d\n", err);
1215 1216
		srp_put_tx_iu(target, iu, SRP_IU_RSP);
	}
1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251

	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");
}

1252 1253
static void srp_handle_recv(struct srp_target_port *target, struct ib_wc *wc)
{
1254
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1255
	struct srp_iu *iu = (struct srp_iu *) (uintptr_t) wc->wr_id;
1256
	int res;
1257 1258
	u8 opcode;

1259 1260
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_ti_iu_len,
				   DMA_FROM_DEVICE);
1261 1262 1263 1264

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

	if (0) {
1265 1266
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "recv completion, opcode 0x%02x\n", opcode);
B
Bart Van Assche 已提交
1267 1268
		print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET, 8, 1,
			       iu->buf, wc->byte_len, true);
1269 1270 1271 1272 1273 1274 1275
	}

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

1276 1277 1278 1279 1280 1281 1282 1283
	case SRP_CRED_REQ:
		srp_process_cred_req(target, iu->buf);
		break;

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

1284 1285
	case SRP_T_LOGOUT:
		/* XXX Handle target logout */
1286 1287
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Got target logout request\n");
1288 1289 1290
		break;

	default:
1291 1292
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled SRP opcode 0x%02x\n", opcode);
1293 1294 1295
		break;
	}

1296 1297
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_ti_iu_len,
				      DMA_FROM_DEVICE);
1298

1299
	res = srp_post_recv(target, iu);
1300 1301 1302
	if (res != 0)
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Recv failed with error code %d\n", res);
1303 1304
}

1305 1306 1307 1308
static void srp_handle_qp_err(enum ib_wc_status wc_status,
			      enum ib_wc_opcode wc_opcode,
			      struct srp_target_port *target)
{
1309
	if (target->connected && !target->qp_in_error) {
1310 1311 1312 1313 1314
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "failed %s status %d\n",
			     wc_opcode & IB_WC_RECV ? "receive" : "send",
			     wc_status);
	}
1315 1316 1317
	target->qp_in_error = true;
}

1318
static void srp_recv_completion(struct ib_cq *cq, void *target_ptr)
1319 1320 1321 1322 1323 1324
{
	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) {
1325 1326 1327 1328
		if (likely(wc.status == IB_WC_SUCCESS)) {
			srp_handle_recv(target, &wc);
		} else {
			srp_handle_qp_err(wc.status, wc.opcode, target);
1329
		}
1330 1331 1332 1333 1334 1335 1336
	}
}

static void srp_send_completion(struct ib_cq *cq, void *target_ptr)
{
	struct srp_target_port *target = target_ptr;
	struct ib_wc wc;
1337
	struct srp_iu *iu;
1338 1339

	while (ib_poll_cq(cq, 1, &wc) > 0) {
1340 1341 1342 1343 1344
		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);
1345
		}
1346 1347 1348
	}
}

1349
static int srp_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *scmnd)
1350
{
1351
	struct srp_target_port *target = host_to_target(shost);
1352 1353 1354
	struct srp_request *req;
	struct srp_iu *iu;
	struct srp_cmd *cmd;
1355
	struct ib_device *dev;
1356
	unsigned long flags;
1357 1358
	int len;

1359 1360 1361 1362 1363 1364
	if (unlikely(target->transport_offline)) {
		scmnd->result = DID_NO_CONNECT << 16;
		scmnd->scsi_done(scmnd);
		return 0;
	}

1365
	spin_lock_irqsave(&target->lock, flags);
1366
	iu = __srp_get_tx_iu(target, SRP_IU_CMD);
1367
	if (!iu)
1368 1369 1370 1371 1372
		goto err_unlock;

	req = list_first_entry(&target->free_reqs, struct srp_request, list);
	list_del(&req->list);
	spin_unlock_irqrestore(&target->lock, flags);
1373

1374
	dev = target->srp_host->srp_dev->dev;
1375
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_iu_len,
1376
				   DMA_TO_DEVICE);
1377 1378

	scmnd->result        = 0;
1379
	scmnd->host_scribble = (void *) req;
1380 1381 1382 1383 1384 1385

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

	cmd->opcode = SRP_CMD;
	cmd->lun    = cpu_to_be64((u64) scmnd->device->lun << 48);
1386
	cmd->tag    = req->index;
1387 1388 1389 1390 1391 1392 1393
	memcpy(cmd->cdb, scmnd->cmnd, scmnd->cmd_len);

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

	len = srp_map_data(scmnd, target, req);
	if (len < 0) {
1394 1395
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Failed to map data\n");
1396
		goto err_iu;
1397 1398
	}

1399
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_iu_len,
1400
				      DMA_TO_DEVICE);
1401

1402
	if (srp_post_send(target, iu, len)) {
1403
		shost_printk(KERN_ERR, target->scsi_host, PFX "Send failed\n");
1404 1405 1406 1407 1408 1409 1410 1411
		goto err_unmap;
	}

	return 0;

err_unmap:
	srp_unmap_data(scmnd, target, req);

1412 1413 1414
err_iu:
	srp_put_tx_iu(target, iu, SRP_IU_CMD);

1415
	spin_lock_irqsave(&target->lock, flags);
1416
	list_add(&req->list, &target->free_reqs);
1417 1418

err_unlock:
1419
	spin_unlock_irqrestore(&target->lock, flags);
1420

1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435
	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;
	}

1436
	for (i = 0; i < SRP_SQ_SIZE; ++i) {
1437
		target->tx_ring[i] = srp_alloc_iu(target->srp_host,
1438
						  target->max_iu_len,
1439 1440 1441
						  GFP_KERNEL, DMA_TO_DEVICE);
		if (!target->tx_ring[i])
			goto err;
1442 1443

		list_add(&target->tx_ring[i]->list, &target->free_tx);
1444 1445 1446 1447 1448 1449 1450 1451 1452 1453
	}

	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;
	}

1454
	for (i = 0; i < SRP_SQ_SIZE; ++i) {
1455 1456 1457 1458 1459 1460 1461
		srp_free_iu(target->srp_host, target->tx_ring[i]);
		target->tx_ring[i] = NULL;
	}

	return -ENOMEM;
}

1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488
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;
}

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 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547
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;

1548 1549
	target->rq_tmo_jiffies = srp_compute_rq_tmo(qp_attr, attr_mask);

1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562
	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;
}

1563 1564 1565 1566
static void srp_cm_rej_handler(struct ib_cm_id *cm_id,
			       struct ib_cm_event *event,
			       struct srp_target_port *target)
{
1567
	struct Scsi_Host *shost = target->scsi_host;
1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583
	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:
1584
		if (srp_target_is_topspin(target)) {
1585 1586 1587 1588 1589 1590 1591 1592
			/*
			 * 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);

1593 1594 1595 1596
			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));
1597 1598 1599

			target->status = SRP_PORT_REDIRECT;
		} else {
1600 1601
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_PORT_REDIRECT\n");
1602 1603 1604 1605 1606
			target->status = -ECONNRESET;
		}
		break;

	case IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID:
1607 1608
		shost_printk(KERN_WARNING, shost,
			    "  REJ reason: IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID\n");
1609 1610 1611 1612 1613 1614 1615 1616 1617 1618
		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)
1619 1620
				shost_printk(KERN_WARNING, shost,
					     PFX "SRP_LOGIN_REJ: requested max_it_iu_len too large\n");
1621
			else
1622 1623
				shost_printk(KERN_WARNING, shost,
					    PFX "SRP LOGIN REJECTED, reason 0x%08x\n", reason);
1624
		} else
1625 1626 1627
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_CONSUMER_DEFINED,"
				     " opcode 0x%02x\n", opcode);
1628 1629 1630
		target->status = -ECONNRESET;
		break;

D
David Dillow 已提交
1631 1632 1633 1634 1635
	case IB_CM_REJ_STALE_CONN:
		shost_printk(KERN_WARNING, shost, "  REJ reason: stale connection\n");
		target->status = SRP_STALE_CONN;
		break;

1636
	default:
1637 1638
		shost_printk(KERN_WARNING, shost, "  REJ reason 0x%x\n",
			     event->param.rej_rcvd.reason);
1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649
		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:
1650 1651
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Sending CM REQ failed\n");
1652 1653 1654 1655 1656 1657
		comp = 1;
		target->status = -ECONNRESET;
		break;

	case IB_CM_REP_RECEIVED:
		comp = 1;
1658
		srp_cm_rep_handler(cm_id, event->private_data, target);
1659 1660 1661
		break;

	case IB_CM_REJ_RECEIVED:
1662
		shost_printk(KERN_DEBUG, target->scsi_host, PFX "REJ received\n");
1663 1664 1665 1666 1667
		comp = 1;

		srp_cm_rej_handler(cm_id, event, target);
		break;

1668
	case IB_CM_DREQ_RECEIVED:
1669 1670
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "DREQ received - connection closed\n");
1671
		srp_change_conn_state(target, false);
1672
		if (ib_send_cm_drep(cm_id, NULL, 0))
1673 1674
			shost_printk(KERN_ERR, target->scsi_host,
				     PFX "Sending CM DREP failed\n");
1675 1676 1677
		break;

	case IB_CM_TIMEWAIT_EXIT:
1678 1679
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "connection closed\n");
1680 1681 1682 1683

		target->status = 0;
		break;

1684 1685 1686 1687 1688
	case IB_CM_MRA_RECEIVED:
	case IB_CM_DREQ_ERROR:
	case IB_CM_DREP_RECEIVED:
		break;

1689
	default:
1690 1691
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled CM event %d\n", event->event);
1692 1693 1694 1695 1696 1697 1698 1699 1700
		break;
	}

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

	return 0;
}

1701
static int srp_send_tsk_mgmt(struct srp_target_port *target,
1702
			     u64 req_tag, unsigned int lun, u8 func)
1703
{
1704
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1705 1706 1707
	struct srp_iu *iu;
	struct srp_tsk_mgmt *tsk_mgmt;

1708 1709 1710
	if (!target->connected || target->qp_in_error)
		return -1;

1711
	init_completion(&target->tsk_mgmt_done);
1712

1713
	spin_lock_irq(&target->lock);
1714
	iu = __srp_get_tx_iu(target, SRP_IU_TSK_MGMT);
1715
	spin_unlock_irq(&target->lock);
1716

1717
	if (!iu)
1718
		return -1;
1719

1720 1721
	ib_dma_sync_single_for_cpu(dev, iu->dma, sizeof *tsk_mgmt,
				   DMA_TO_DEVICE);
1722 1723 1724 1725
	tsk_mgmt = iu->buf;
	memset(tsk_mgmt, 0, sizeof *tsk_mgmt);

	tsk_mgmt->opcode 	= SRP_TSK_MGMT;
1726 1727
	tsk_mgmt->lun		= cpu_to_be64((u64) lun << 48);
	tsk_mgmt->tag		= req_tag | SRP_TAG_TSK_MGMT;
1728
	tsk_mgmt->tsk_mgmt_func = func;
1729
	tsk_mgmt->task_tag	= req_tag;
1730

1731 1732
	ib_dma_sync_single_for_device(dev, iu->dma, sizeof *tsk_mgmt,
				      DMA_TO_DEVICE);
1733 1734 1735 1736
	if (srp_post_send(target, iu, sizeof *tsk_mgmt)) {
		srp_put_tx_iu(target, iu, SRP_IU_TSK_MGMT);
		return -1;
	}
1737

1738
	if (!wait_for_completion_timeout(&target->tsk_mgmt_done,
1739
					 msecs_to_jiffies(SRP_ABORT_TIMEOUT_MS)))
1740
		return -1;
1741

1742 1743 1744
	return 0;
}

1745 1746
static int srp_abort(struct scsi_cmnd *scmnd)
{
1747
	struct srp_target_port *target = host_to_target(scmnd->device->host);
1748
	struct srp_request *req = (struct srp_request *) scmnd->host_scribble;
1749
	int ret;
1750

1751
	shost_printk(KERN_ERR, target->scsi_host, "SRP abort called\n");
1752

1753
	if (!req || !srp_claim_req(target, req, scmnd))
1754
		return FAILED;
1755
	if (srp_send_tsk_mgmt(target, req->index, scmnd->device->lun,
1756
			      SRP_TSK_ABORT_TASK) == 0)
1757
		ret = SUCCESS;
1758 1759
	else if (target->transport_offline)
		ret = FAST_IO_FAIL;
1760 1761
	else
		ret = FAILED;
B
Bart Van Assche 已提交
1762 1763
	srp_free_req(target, req, scmnd, 0);
	scmnd->result = DID_ABORT << 16;
1764
	scmnd->scsi_done(scmnd);
1765

1766
	return ret;
1767 1768 1769 1770
}

static int srp_reset_device(struct scsi_cmnd *scmnd)
{
1771
	struct srp_target_port *target = host_to_target(scmnd->device->host);
1772
	int i;
1773

1774
	shost_printk(KERN_ERR, target->scsi_host, "SRP reset_device called\n");
1775

1776 1777
	if (srp_send_tsk_mgmt(target, SRP_TAG_NO_REQ, scmnd->device->lun,
			      SRP_TSK_LUN_RESET))
1778
		return FAILED;
1779
	if (target->tsk_mgmt_status)
1780 1781
		return FAILED;

1782 1783
	for (i = 0; i < SRP_CMD_SQ_SIZE; ++i) {
		struct srp_request *req = &target->req_ring[i];
1784
		if (req->scmnd && req->scmnd->device == scmnd->device)
1785
			srp_reset_req(target, req);
1786
	}
1787 1788

	return SUCCESS;
1789 1790 1791 1792 1793 1794 1795
}

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

1796
	shost_printk(KERN_ERR, target->scsi_host, PFX "SRP reset_host called\n");
1797 1798 1799 1800 1801 1802 1803

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

	return ret;
}

1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818
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;
}

1819 1820
static ssize_t show_id_ext(struct device *dev, struct device_attribute *attr,
			   char *buf)
1821
{
1822
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1823 1824 1825 1826 1827

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

1828 1829
static ssize_t show_ioc_guid(struct device *dev, struct device_attribute *attr,
			     char *buf)
1830
{
1831
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1832 1833 1834 1835 1836

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

1837 1838
static ssize_t show_service_id(struct device *dev,
			       struct device_attribute *attr, char *buf)
1839
{
1840
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1841 1842 1843 1844 1845

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

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

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

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

H
Harvey Harrison 已提交
1859
	return sprintf(buf, "%pI6\n", target->path.dgid.raw);
1860 1861
}

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

H
Harvey Harrison 已提交
1867
	return sprintf(buf, "%pI6\n", target->orig_dgid);
1868 1869
}

1870 1871 1872 1873 1874 1875 1876 1877
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);
}

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

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

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

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

1894 1895
static ssize_t show_local_ib_device(struct device *dev,
				    struct device_attribute *attr, char *buf)
1896
{
1897
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1898

1899
	return sprintf(buf, "%s\n", target->srp_host->srp_dev->dev->name);
1900 1901
}

1902 1903 1904 1905 1906 1907 1908 1909
static ssize_t show_comp_vector(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->comp_vector);
}

1910 1911 1912 1913 1914 1915 1916 1917
static ssize_t show_tl_retry_count(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->tl_retry_count);
}

1918 1919 1920 1921 1922 1923 1924 1925
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);
}

1926 1927 1928 1929 1930 1931 1932 1933
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");
}

1934 1935 1936 1937 1938 1939
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);
1940
static DEVICE_ATTR(req_lim,         S_IRUGO, show_req_lim,         NULL);
1941 1942 1943
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);
1944
static DEVICE_ATTR(comp_vector,     S_IRUGO, show_comp_vector,     NULL);
1945
static DEVICE_ATTR(tl_retry_count,  S_IRUGO, show_tl_retry_count,  NULL);
1946
static DEVICE_ATTR(cmd_sg_entries,  S_IRUGO, show_cmd_sg_entries,  NULL);
1947
static DEVICE_ATTR(allow_ext_sg,    S_IRUGO, show_allow_ext_sg,    NULL);
1948 1949 1950 1951 1952 1953 1954 1955

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,
1956
	&dev_attr_req_lim,
1957 1958 1959
	&dev_attr_zero_req_lim,
	&dev_attr_local_ib_port,
	&dev_attr_local_ib_device,
1960
	&dev_attr_comp_vector,
1961
	&dev_attr_tl_retry_count,
1962
	&dev_attr_cmd_sg_entries,
1963
	&dev_attr_allow_ext_sg,
1964 1965 1966
	NULL
};

1967 1968
static struct scsi_host_template srp_template = {
	.module				= THIS_MODULE,
R
Roland Dreier 已提交
1969 1970
	.name				= "InfiniBand SRP initiator",
	.proc_name			= DRV_NAME,
1971
	.slave_configure		= srp_slave_configure,
1972 1973 1974 1975 1976
	.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,
B
Bart Van Assche 已提交
1977
	.skip_settle_delay		= true,
1978
	.sg_tablesize			= SRP_DEF_SG_TABLESIZE,
1979
	.can_queue			= SRP_CMD_SQ_SIZE,
1980
	.this_id			= -1,
1981
	.cmd_per_lun			= SRP_CMD_SQ_SIZE,
1982 1983
	.use_clustering			= ENABLE_CLUSTERING,
	.shost_attrs			= srp_host_attrs
1984 1985 1986 1987
};

static int srp_add_target(struct srp_host *host, struct srp_target_port *target)
{
1988 1989 1990
	struct srp_rport_identifiers ids;
	struct srp_rport *rport;

1991 1992 1993
	sprintf(target->target_name, "SRP.T10:%016llX",
		 (unsigned long long) be64_to_cpu(target->id_ext));

1994
	if (scsi_add_host(target->scsi_host, host->srp_dev->dev->dma_device))
1995 1996
		return -ENODEV;

1997 1998
	memcpy(ids.port_id, &target->id_ext, 8);
	memcpy(ids.port_id + 8, &target->ioc_guid, 8);
1999
	ids.roles = SRP_RPORT_ROLE_TARGET;
2000 2001 2002 2003 2004 2005
	rport = srp_rport_add(target->scsi_host, &ids);
	if (IS_ERR(rport)) {
		scsi_remove_host(target->scsi_host);
		return PTR_ERR(rport);
	}

2006 2007
	rport->lld_data = target;

2008
	spin_lock(&host->target_lock);
2009
	list_add_tail(&target->list, &host->target_list);
2010
	spin_unlock(&host->target_lock);
2011 2012 2013 2014

	target->state = SRP_TARGET_LIVE;

	scsi_scan_target(&target->scsi_host->shost_gendev,
2015
			 0, target->scsi_id, SCAN_WILD_CARD, 0);
2016 2017 2018 2019

	return 0;
}

2020
static void srp_release_dev(struct device *dev)
2021 2022
{
	struct srp_host *host =
2023
		container_of(dev, struct srp_host, dev);
2024 2025 2026 2027 2028 2029

	complete(&host->released);
}

static struct class srp_class = {
	.name    = "infiniband_srp",
2030
	.dev_release = srp_release_dev
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 2058 2059 2060 2061 2062
/**
 * srp_conn_unique() - check whether the connection to a target is unique
 */
static bool srp_conn_unique(struct srp_host *host,
			    struct srp_target_port *target)
{
	struct srp_target_port *t;
	bool ret = false;

	if (target->state == SRP_TARGET_REMOVED)
		goto out;

	ret = true;

	spin_lock(&host->target_lock);
	list_for_each_entry(t, &host->target_list, list) {
		if (t != target &&
		    target->id_ext == t->id_ext &&
		    target->ioc_guid == t->ioc_guid &&
		    target->initiator_ext == t->initiator_ext) {
			ret = false;
			break;
		}
	}
	spin_unlock(&host->target_lock);

out:
	return ret;
}

2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078
/*
 * 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,
2079
	SRP_OPT_MAX_CMD_PER_LUN	= 1 << 6,
2080
	SRP_OPT_IO_CLASS	= 1 << 7,
2081
	SRP_OPT_INITIATOR_EXT	= 1 << 8,
2082
	SRP_OPT_CMD_SG_ENTRIES	= 1 << 9,
2083 2084
	SRP_OPT_ALLOW_EXT_SG	= 1 << 10,
	SRP_OPT_SG_TABLESIZE	= 1 << 11,
2085
	SRP_OPT_COMP_VECTOR	= 1 << 12,
2086
	SRP_OPT_TL_RETRY_COUNT	= 1 << 13,
2087 2088 2089 2090 2091 2092 2093
	SRP_OPT_ALL		= (SRP_OPT_ID_EXT	|
				   SRP_OPT_IOC_GUID	|
				   SRP_OPT_DGID		|
				   SRP_OPT_PKEY		|
				   SRP_OPT_SERVICE_ID),
};

2094
static const match_table_t srp_opt_tokens = {
2095 2096 2097 2098 2099 2100 2101
	{ 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" 	},
2102
	{ SRP_OPT_IO_CLASS,		"io_class=%x"		},
2103
	{ SRP_OPT_INITIATOR_EXT,	"initiator_ext=%s"	},
2104
	{ SRP_OPT_CMD_SG_ENTRIES,	"cmd_sg_entries=%u"	},
2105 2106
	{ SRP_OPT_ALLOW_EXT_SG,		"allow_ext_sg=%u"	},
	{ SRP_OPT_SG_TABLESIZE,		"sg_tablesize=%u"	},
2107
	{ SRP_OPT_COMP_VECTOR,		"comp_vector=%u"	},
2108
	{ SRP_OPT_TL_RETRY_COUNT,	"tl_retry_count=%u"	},
2109
	{ SRP_OPT_ERR,			NULL 			}
2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137
};

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);
2138 2139 2140 2141
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2142 2143 2144 2145 2146 2147
			target->id_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_IOC_GUID:
			p = match_strdup(args);
2148 2149 2150 2151
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2152 2153 2154 2155 2156 2157
			target->ioc_guid = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_DGID:
			p = match_strdup(args);
2158 2159 2160 2161
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2162
			if (strlen(p) != 32) {
2163
				pr_warn("bad dest GID parameter '%s'\n", p);
2164
				kfree(p);
2165 2166 2167 2168 2169 2170 2171
				goto out;
			}

			for (i = 0; i < 16; ++i) {
				strlcpy(dgid, p + i * 2, 3);
				target->path.dgid.raw[i] = simple_strtoul(dgid, NULL, 16);
			}
2172
			kfree(p);
2173
			memcpy(target->orig_dgid, target->path.dgid.raw, 16);
2174 2175 2176 2177
			break;

		case SRP_OPT_PKEY:
			if (match_hex(args, &token)) {
2178
				pr_warn("bad P_Key parameter '%s'\n", p);
2179 2180 2181 2182 2183 2184 2185
				goto out;
			}
			target->path.pkey = cpu_to_be16(token);
			break;

		case SRP_OPT_SERVICE_ID:
			p = match_strdup(args);
2186 2187 2188 2189
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2190
			target->service_id = cpu_to_be64(simple_strtoull(p, NULL, 16));
2191
			target->path.service_id = target->service_id;
2192 2193 2194 2195 2196
			kfree(p);
			break;

		case SRP_OPT_MAX_SECT:
			if (match_int(args, &token)) {
2197
				pr_warn("bad max sect parameter '%s'\n", p);
2198 2199 2200 2201 2202
				goto out;
			}
			target->scsi_host->max_sectors = token;
			break;

2203 2204
		case SRP_OPT_MAX_CMD_PER_LUN:
			if (match_int(args, &token)) {
2205 2206
				pr_warn("bad max cmd_per_lun parameter '%s'\n",
					p);
2207 2208
				goto out;
			}
2209
			target->scsi_host->cmd_per_lun = min(token, SRP_CMD_SQ_SIZE);
2210 2211
			break;

2212 2213
		case SRP_OPT_IO_CLASS:
			if (match_hex(args, &token)) {
2214
				pr_warn("bad IO class parameter '%s'\n", p);
2215 2216 2217 2218
				goto out;
			}
			if (token != SRP_REV10_IB_IO_CLASS &&
			    token != SRP_REV16A_IB_IO_CLASS) {
2219 2220 2221
				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);
2222 2223 2224 2225 2226
				goto out;
			}
			target->io_class = token;
			break;

2227 2228
		case SRP_OPT_INITIATOR_EXT:
			p = match_strdup(args);
2229 2230 2231 2232
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2233 2234 2235 2236
			target->initiator_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

2237 2238
		case SRP_OPT_CMD_SG_ENTRIES:
			if (match_int(args, &token) || token < 1 || token > 255) {
2239 2240
				pr_warn("bad max cmd_sg_entries parameter '%s'\n",
					p);
2241 2242 2243 2244 2245
				goto out;
			}
			target->cmd_sg_cnt = token;
			break;

2246 2247
		case SRP_OPT_ALLOW_EXT_SG:
			if (match_int(args, &token)) {
2248
				pr_warn("bad allow_ext_sg parameter '%s'\n", p);
2249 2250 2251 2252 2253 2254 2255 2256
				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) {
2257 2258
				pr_warn("bad max sg_tablesize parameter '%s'\n",
					p);
2259 2260 2261 2262 2263
				goto out;
			}
			target->sg_tablesize = token;
			break;

2264 2265 2266 2267 2268 2269 2270 2271
		case SRP_OPT_COMP_VECTOR:
			if (match_int(args, &token) || token < 0) {
				pr_warn("bad comp_vector parameter '%s'\n", p);
				goto out;
			}
			target->comp_vector = token;
			break;

2272 2273 2274 2275 2276 2277 2278 2279 2280
		case SRP_OPT_TL_RETRY_COUNT:
			if (match_int(args, &token) || token < 2 || token > 7) {
				pr_warn("bad tl_retry_count parameter '%s' (must be a number between 2 and 7)\n",
					p);
				goto out;
			}
			target->tl_retry_count = token;
			break;

2281
		default:
2282 2283
			pr_warn("unknown parameter or missing value '%s' in target creation request\n",
				p);
2284 2285 2286 2287 2288 2289 2290 2291 2292 2293
			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))
2294 2295
				pr_warn("target creation request is missing parameter '%s'\n",
					srp_opt_tokens[i].pattern);
2296 2297 2298 2299 2300 2301

out:
	kfree(options);
	return ret;
}

2302 2303
static ssize_t srp_create_target(struct device *dev,
				 struct device_attribute *attr,
2304 2305 2306
				 const char *buf, size_t count)
{
	struct srp_host *host =
2307
		container_of(dev, struct srp_host, dev);
2308 2309
	struct Scsi_Host *target_host;
	struct srp_target_port *target;
2310 2311
	struct ib_device *ibdev = host->srp_dev->dev;
	dma_addr_t dma_addr;
2312
	int i, ret;
2313 2314 2315 2316 2317 2318

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

2319
	target_host->transportt  = ib_srp_transport_template;
2320 2321
	target_host->max_channel = 0;
	target_host->max_id      = 1;
A
Arne Redlich 已提交
2322 2323
	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
2324

2325 2326
	target = host_to_target(target_host);

2327 2328 2329 2330 2331 2332
	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;
2333 2334
	target->sg_tablesize	= indirect_sg_entries ? : cmd_sg_entries;
	target->allow_ext_sg	= allow_ext_sg;
2335
	target->tl_retry_count	= 7;
2336 2337 2338 2339 2340

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

2341 2342 2343 2344 2345 2346 2347 2348 2349 2350
	if (!srp_conn_unique(target->srp_host, target)) {
		shost_printk(KERN_INFO, target->scsi_host,
			     PFX "Already connected to target port with id_ext=%016llx;ioc_guid=%016llx;initiator_ext=%016llx\n",
			     be64_to_cpu(target->id_ext),
			     be64_to_cpu(target->ioc_guid),
			     be64_to_cpu(target->initiator_ext));
		ret = -EEXIST;
		goto err;
	}

2351 2352
	if (!host->srp_dev->fmr_pool && !target->allow_ext_sg &&
				target->cmd_sg_cnt < target->sg_tablesize) {
2353
		pr_warn("No FMR pool and no external indirect descriptors, limiting sg_tablesize to cmd_sg_cnt\n");
2354 2355 2356 2357 2358 2359
		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);
2360 2361 2362 2363
	target->max_iu_len = sizeof (struct srp_cmd) +
			     sizeof (struct srp_indirect_buf) +
			     target->cmd_sg_cnt * sizeof (struct srp_direct_buf);

2364
	INIT_WORK(&target->remove_work, srp_remove_work);
2365 2366 2367 2368 2369 2370 2371 2372 2373 2374
	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);
2375 2376
		req->indirect_desc = kmalloc(target->indirect_size, GFP_KERNEL);
		if (!req->fmr_list || !req->map_page || !req->indirect_desc)
2377 2378
			goto err_free_mem;

2379 2380 2381 2382 2383 2384 2385
		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;
2386 2387 2388 2389
		req->index = i;
		list_add_tail(&req->list, &target->free_reqs);
	}

2390
	ib_query_gid(ibdev, host->port, 0, &target->path.sgid);
2391

2392 2393
	shost_printk(KERN_DEBUG, target->scsi_host, PFX
		     "new target: id_ext %016llx ioc_guid %016llx pkey %04x "
H
Harvey Harrison 已提交
2394
		     "service_id %016llx dgid %pI6\n",
2395 2396 2397 2398
	       (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),
2399
	       target->path.dgid.raw);
2400 2401 2402

	ret = srp_create_target_ib(target);
	if (ret)
2403
		goto err_free_mem;
2404

D
David Dillow 已提交
2405 2406
	ret = srp_new_cm_id(target);
	if (ret)
2407
		goto err_free_ib;
2408 2409 2410

	ret = srp_connect_target(target);
	if (ret) {
2411 2412
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Connection failed\n");
2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427
		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);

2428
err_free_ib:
2429 2430
	srp_free_target_ib(target);

2431 2432 2433
err_free_mem:
	srp_free_req_data(target);

2434 2435 2436 2437 2438 2439
err:
	scsi_host_put(target_host);

	return ret;
}

2440
static DEVICE_ATTR(add_target, S_IWUSR, NULL, srp_create_target);
2441

2442 2443
static ssize_t show_ibdev(struct device *dev, struct device_attribute *attr,
			  char *buf)
2444
{
2445
	struct srp_host *host = container_of(dev, struct srp_host, dev);
2446

2447
	return sprintf(buf, "%s\n", host->srp_dev->dev->name);
2448 2449
}

2450
static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
2451

2452 2453
static ssize_t show_port(struct device *dev, struct device_attribute *attr,
			 char *buf)
2454
{
2455
	struct srp_host *host = container_of(dev, struct srp_host, dev);
2456 2457 2458 2459

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

2460
static DEVICE_ATTR(port, S_IRUGO, show_port, NULL);
2461

2462
static struct srp_host *srp_add_port(struct srp_device *device, u8 port)
2463 2464 2465 2466 2467 2468 2469 2470
{
	struct srp_host *host;

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

	INIT_LIST_HEAD(&host->target_list);
2471
	spin_lock_init(&host->target_lock);
2472
	init_completion(&host->released);
2473
	host->srp_dev = device;
2474 2475
	host->port = port;

2476 2477
	host->dev.class = &srp_class;
	host->dev.parent = device->dev->dma_device;
2478
	dev_set_name(&host->dev, "srp-%s-%d", device->dev->name, port);
2479

2480
	if (device_register(&host->dev))
2481
		goto free_host;
2482
	if (device_create_file(&host->dev, &dev_attr_add_target))
2483
		goto err_class;
2484
	if (device_create_file(&host->dev, &dev_attr_ibdev))
2485
		goto err_class;
2486
	if (device_create_file(&host->dev, &dev_attr_port))
2487 2488 2489 2490 2491
		goto err_class;

	return host;

err_class:
2492
	device_unregister(&host->dev);
2493

2494
free_host:
2495 2496 2497 2498 2499 2500 2501
	kfree(host);

	return NULL;
}

static void srp_add_one(struct ib_device *device)
{
2502 2503 2504
	struct srp_device *srp_dev;
	struct ib_device_attr *dev_attr;
	struct ib_fmr_pool_param fmr_param;
2505
	struct srp_host *host;
2506
	int max_pages_per_fmr, fmr_page_shift, s, e, p;
2507

2508 2509
	dev_attr = kmalloc(sizeof *dev_attr, GFP_KERNEL);
	if (!dev_attr)
2510
		return;
2511

2512
	if (ib_query_device(device, dev_attr)) {
2513
		pr_warn("Query device failed for %s\n", device->name);
2514 2515 2516 2517 2518 2519 2520 2521 2522
		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
2523 2524
	 * minimum of 4096 bytes. We're unlikely to build large sglists
	 * out of smaller entries.
2525
	 */
2526 2527 2528 2529
	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;
2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544

	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;

2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562
	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;
	}

2563 2564
	if (IS_ERR(srp_dev->fmr_pool))
		srp_dev->fmr_pool = NULL;
2565

T
Tom Tucker 已提交
2566
	if (device->node_type == RDMA_NODE_IB_SWITCH) {
2567 2568 2569 2570 2571 2572 2573 2574
		s = 0;
		e = 0;
	} else {
		s = 1;
		e = device->phys_port_cnt;
	}

	for (p = s; p <= e; ++p) {
2575
		host = srp_add_port(srp_dev, p);
2576
		if (host)
2577
			list_add_tail(&host->list, &srp_dev->dev_list);
2578 2579
	}

2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591
	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);
2592 2593 2594 2595
}

static void srp_remove_one(struct ib_device *device)
{
2596
	struct srp_device *srp_dev;
2597
	struct srp_host *host, *tmp_host;
2598
	struct srp_target_port *target;
2599

2600
	srp_dev = ib_get_client_data(device, &srp_client);
2601 2602
	if (!srp_dev)
		return;
2603

2604
	list_for_each_entry_safe(host, tmp_host, &srp_dev->dev_list, list) {
2605
		device_unregister(&host->dev);
2606 2607 2608 2609 2610 2611 2612
		/*
		 * Wait for the sysfs entry to go away, so that no new
		 * target ports can be created.
		 */
		wait_for_completion(&host->released);

		/*
2613
		 * Remove all target ports.
2614
		 */
2615
		spin_lock(&host->target_lock);
2616 2617
		list_for_each_entry(target, &host->target_list, list)
			srp_queue_remove_work(target);
2618
		spin_unlock(&host->target_lock);
2619 2620

		/*
2621
		 * Wait for target port removal tasks.
2622
		 */
2623
		flush_workqueue(system_long_wq);
2624 2625 2626 2627

		kfree(host);
	}

2628 2629 2630 2631 2632 2633
	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);
2634 2635
}

2636
static struct srp_function_template ib_srp_transport_functions = {
2637
	.rport_delete		 = srp_rport_delete,
2638 2639
};

2640 2641 2642 2643
static int __init srp_init_module(void)
{
	int ret;

2644
	BUILD_BUG_ON(FIELD_SIZEOF(struct ib_wc, wr_id) < sizeof(void *));
2645

2646
	if (srp_sg_tablesize) {
2647
		pr_warn("srp_sg_tablesize is deprecated, please use cmd_sg_entries\n");
2648 2649 2650 2651 2652 2653 2654 2655
		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) {
2656
		pr_warn("Clamping cmd_sg_entries to 255\n");
2657
		cmd_sg_entries = 255;
2658 2659
	}

2660 2661 2662
	if (!indirect_sg_entries)
		indirect_sg_entries = cmd_sg_entries;
	else if (indirect_sg_entries < cmd_sg_entries) {
2663 2664
		pr_warn("Bumping up indirect_sg_entries to match cmd_sg_entries (%u)\n",
			cmd_sg_entries);
2665 2666 2667
		indirect_sg_entries = cmd_sg_entries;
	}

2668 2669 2670 2671 2672
	ib_srp_transport_template =
		srp_attach_transport(&ib_srp_transport_functions);
	if (!ib_srp_transport_template)
		return -ENOMEM;

2673 2674
	ret = class_register(&srp_class);
	if (ret) {
2675
		pr_err("couldn't register class infiniband_srp\n");
2676
		srp_release_transport(ib_srp_transport_template);
2677 2678 2679
		return ret;
	}

2680 2681
	ib_sa_register_client(&srp_sa_client);

2682 2683
	ret = ib_register_client(&srp_client);
	if (ret) {
2684
		pr_err("couldn't register IB client\n");
2685
		srp_release_transport(ib_srp_transport_template);
2686
		ib_sa_unregister_client(&srp_sa_client);
2687 2688 2689 2690 2691 2692 2693 2694 2695 2696
		class_unregister(&srp_class);
		return ret;
	}

	return 0;
}

static void __exit srp_cleanup_module(void)
{
	ib_unregister_client(&srp_client);
2697
	ib_sa_unregister_client(&srp_sa_client);
2698
	class_unregister(&srp_class);
2699
	srp_release_transport(ib_srp_transport_template);
2700 2701 2702 2703
}

module_init(srp_init_module);
module_exit(srp_cleanup_module);