ib_srp.c 78.5 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

#include <scsi/scsi.h>
#include <scsi/scsi_device.h>
#include <scsi/scsi_dbg.h>
49
#include <scsi/scsi_tcq.h>
50
#include <scsi/srp.h>
51
#include <scsi/scsi_transport_srp.h>
52 53 54 55 56

#include "ib_srp.h"

#define DRV_NAME	"ib_srp"
#define PFX		DRV_NAME ": "
57 58
#define DRV_VERSION	"1.0"
#define DRV_RELDATE	"July 1, 2013"
59 60 61 62 63 64

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

65 66
static unsigned int srp_sg_tablesize;
static unsigned int cmd_sg_entries;
67 68
static unsigned int indirect_sg_entries;
static bool allow_ext_sg;
69
static int topspin_workarounds = 1;
70

71 72
module_param(srp_sg_tablesize, uint, 0444);
MODULE_PARM_DESC(srp_sg_tablesize, "Deprecated name for cmd_sg_entries");
73

74 75 76
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)");
77

78 79 80 81 82 83 84 85
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)");

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

90 91
static struct kernel_param_ops srp_tmo_ops;

92 93 94 95 96
static int srp_reconnect_delay = 10;
module_param_cb(reconnect_delay, &srp_tmo_ops, &srp_reconnect_delay,
		S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(reconnect_delay, "Time between successive reconnect attempts");

97 98 99 100 101 102 103 104
static int srp_fast_io_fail_tmo = 15;
module_param_cb(fast_io_fail_tmo, &srp_tmo_ops, &srp_fast_io_fail_tmo,
		S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(fast_io_fail_tmo,
		 "Number of seconds between the observation of a transport"
		 " layer error and failing all I/O. \"off\" means that this"
		 " functionality is disabled.");

105
static int srp_dev_loss_tmo = 600;
106 107 108 109 110 111 112 113 114 115
module_param_cb(dev_loss_tmo, &srp_tmo_ops, &srp_dev_loss_tmo,
		S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(dev_loss_tmo,
		 "Maximum number of seconds that the SRP transport should"
		 " insulate transport layer errors. After this time has been"
		 " exceeded the SCSI host is removed. Should be"
		 " between 1 and " __stringify(SCSI_DEVICE_BLOCK_MAX_TIMEOUT)
		 " if fast_io_fail_tmo has not been set. \"off\" means that"
		 " this functionality is disabled.");

116 117
static void srp_add_one(struct ib_device *device);
static void srp_remove_one(struct ib_device *device);
118 119
static void srp_recv_completion(struct ib_cq *cq, void *target_ptr);
static void srp_send_completion(struct ib_cq *cq, void *target_ptr);
120 121
static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event);

122 123
static struct scsi_transport_template *ib_srp_transport_template;

124 125 126 127 128 129
static struct ib_client srp_client = {
	.name   = "srp",
	.add    = srp_add_one,
	.remove = srp_remove_one
};

130 131
static struct ib_sa_client srp_sa_client;

132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152
static int srp_tmo_get(char *buffer, const struct kernel_param *kp)
{
	int tmo = *(int *)kp->arg;

	if (tmo >= 0)
		return sprintf(buffer, "%d", tmo);
	else
		return sprintf(buffer, "off");
}

static int srp_tmo_set(const char *val, const struct kernel_param *kp)
{
	int tmo, res;

	if (strncmp(val, "off", 3) != 0) {
		res = kstrtoint(val, 0, &tmo);
		if (res)
			goto out;
	} else {
		tmo = -1;
	}
153 154 155 156 157
	if (kp->arg == &srp_reconnect_delay)
		res = srp_tmo_valid(tmo, srp_fast_io_fail_tmo,
				    srp_dev_loss_tmo);
	else if (kp->arg == &srp_fast_io_fail_tmo)
		res = srp_tmo_valid(srp_reconnect_delay, tmo, srp_dev_loss_tmo);
158
	else
159 160
		res = srp_tmo_valid(srp_reconnect_delay, srp_fast_io_fail_tmo,
				    tmo);
161 162 163 164 165 166 167 168 169 170 171 172 173
	if (res)
		goto out;
	*(int *)kp->arg = tmo;

out:
	return res;
}

static struct kernel_param_ops srp_tmo_ops = {
	.get = srp_tmo_get,
	.set = srp_tmo_set,
};

174 175 176 177 178 179 180 181 182 183
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;
}

184 185 186
static int srp_target_is_topspin(struct srp_target_port *target)
{
	static const u8 topspin_oui[3] = { 0x00, 0x05, 0xad };
187
	static const u8 cisco_oui[3]   = { 0x00, 0x1b, 0x0d };
188 189

	return topspin_workarounds &&
190 191
		(!memcmp(&target->ioc_guid, topspin_oui, sizeof topspin_oui) ||
		 !memcmp(&target->ioc_guid, cisco_oui, sizeof cisco_oui));
192 193
}

194 195 196 197 198 199 200 201 202 203 204 205 206 207
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;

208 209 210
	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))
211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230
		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;

231 232
	ib_dma_unmap_single(host->srp_dev->dev, iu->dma, iu->size,
			    iu->direction);
233 234 235 236 237 238
	kfree(iu->buf);
	kfree(iu);
}

static void srp_qp_event(struct ib_event *event, void *context)
{
239
	pr_debug("QP event %d\n", event->event);
240 241 242 243 244 245 246 247 248 249 250 251
}

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;

252 253 254 255
	ret = ib_find_pkey(target->srp_host->srp_dev->dev,
			   target->srp_host->port,
			   be16_to_cpu(target->path.pkey),
			   &attr->pkey_index);
256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274
	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
275 276 277 278
static int srp_new_cm_id(struct srp_target_port *target)
{
	struct ib_cm_id *new_cm_id;

279
	new_cm_id = ib_create_cm_id(target->srp_host->srp_dev->dev,
D
David Dillow 已提交
280 281 282 283 284 285 286 287 288 289 290
				    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;
}

291 292 293
static int srp_create_target_ib(struct srp_target_port *target)
{
	struct ib_qp_init_attr *init_attr;
294 295
	struct ib_cq *recv_cq, *send_cq;
	struct ib_qp *qp;
296 297 298 299 300 301
	int ret;

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

302
	recv_cq = ib_create_cq(target->srp_host->srp_dev->dev,
303 304
			       srp_recv_completion, NULL, target,
			       target->queue_size, target->comp_vector);
305 306
	if (IS_ERR(recv_cq)) {
		ret = PTR_ERR(recv_cq);
307
		goto err;
308 309
	}

310
	send_cq = ib_create_cq(target->srp_host->srp_dev->dev,
311 312
			       srp_send_completion, NULL, target,
			       target->queue_size, target->comp_vector);
313 314
	if (IS_ERR(send_cq)) {
		ret = PTR_ERR(send_cq);
315
		goto err_recv_cq;
316 317
	}

318
	ib_req_notify_cq(recv_cq, IB_CQ_NEXT_COMP);
319 320

	init_attr->event_handler       = srp_qp_event;
321 322
	init_attr->cap.max_send_wr     = target->queue_size;
	init_attr->cap.max_recv_wr     = target->queue_size;
323 324 325 326
	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;
327 328
	init_attr->send_cq             = send_cq;
	init_attr->recv_cq             = recv_cq;
329

330 331 332
	qp = ib_create_qp(target->srp_host->srp_dev->pd, init_attr);
	if (IS_ERR(qp)) {
		ret = PTR_ERR(qp);
333
		goto err_send_cq;
334 335
	}

336
	ret = srp_init_qp(target, qp);
337 338
	if (ret)
		goto err_qp;
339

340 341 342 343 344 345 346 347 348 349 350
	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;

351 352 353 354
	kfree(init_attr);
	return 0;

err_qp:
355
	ib_destroy_qp(qp);
356 357

err_send_cq:
358
	ib_destroy_cq(send_cq);
359 360

err_recv_cq:
361
	ib_destroy_cq(recv_cq);
362 363

err:
364 365 366 367
	kfree(init_attr);
	return ret;
}

368 369 370 371
/*
 * Note: this function may be called without srp_alloc_iu_bufs() having been
 * invoked. Hence the target->[rt]x_ring checks.
 */
372 373 374 375 376
static void srp_free_target_ib(struct srp_target_port *target)
{
	int i;

	ib_destroy_qp(target->qp);
377 378
	ib_destroy_cq(target->send_cq);
	ib_destroy_cq(target->recv_cq);
379

380 381 382
	target->qp = NULL;
	target->send_cq = target->recv_cq = NULL;

383 384 385 386 387 388 389 390 391 392 393 394
	if (target->rx_ring) {
		for (i = 0; i < target->queue_size; ++i)
			srp_free_iu(target->srp_host, target->rx_ring[i]);
		kfree(target->rx_ring);
		target->rx_ring = NULL;
	}
	if (target->tx_ring) {
		for (i = 0; i < target->queue_size; ++i)
			srp_free_iu(target->srp_host, target->tx_ring[i]);
		kfree(target->tx_ring);
		target->tx_ring = NULL;
	}
395 396 397 398 399 400 401 402 403 404
}

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)
405 406
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Got failed path rec status %d\n", status);
407 408 409 410 411 412 413
	else
		target->path = *pathrec;
	complete(&target->done);
}

static int srp_lookup_path(struct srp_target_port *target)
{
414 415
	int ret;

416 417 418 419
	target->path.numb_path = 1;

	init_completion(&target->done);

420
	target->path_query_id = ib_sa_path_rec_get(&srp_sa_client,
421
						   target->srp_host->srp_dev->dev,
422 423
						   target->srp_host->port,
						   &target->path,
424
						   IB_SA_PATH_REC_SERVICE_ID	|
425 426 427 428 429 430 431 432 433 434 435
						   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;

436 437 438
	ret = wait_for_completion_interruptible(&target->done);
	if (ret < 0)
		return ret;
439 440

	if (target->status < 0)
441 442
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Path record query failed\n");
443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477

	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;
478
	req->param.retry_count                = target->tl_retry_count;
479 480 481 482 483
	req->param.rnr_retry_count 	      = 7;
	req->param.max_cm_retries 	      = 15;

	req->priv.opcode     	= SRP_LOGIN_REQ;
	req->priv.tag        	= 0;
484
	req->priv.req_it_iu_len = cpu_to_be32(target->max_iu_len);
485 486
	req->priv.req_buf_fmt 	= cpu_to_be16(SRP_BUF_FORMAT_DIRECT |
					      SRP_BUF_FORMAT_INDIRECT);
487
	/*
R
Roland Dreier 已提交
488
	 * In the published SRP specification (draft rev. 16a), the
489 490 491 492 493 494 495 496 497
	 * 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,
498
		       &target->path.sgid.global.interface_id, 8);
499
		memcpy(req->priv.initiator_port_id + 8,
500
		       &target->initiator_ext, 8);
501 502 503 504
		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,
505 506 507
		       &target->initiator_ext, 8);
		memcpy(req->priv.initiator_port_id + 8,
		       &target->path.sgid.global.interface_id, 8);
508 509 510 511
		memcpy(req->priv.target_port_id,     &target->id_ext, 8);
		memcpy(req->priv.target_port_id + 8, &target->ioc_guid, 8);
	}

512 513
	/*
	 * Topspin/Cisco SRP targets will reject our login unless we
514 515
	 * zero out the first 8 bytes of our initiator port ID and set
	 * the second 8 bytes to the local node GUID.
516
	 */
517
	if (srp_target_is_topspin(target)) {
518 519 520 521
		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));
522
		memset(req->priv.initiator_port_id, 0, 8);
523
		memcpy(req->priv.initiator_port_id + 8,
524
		       &target->srp_host->srp_dev->dev->node_guid, 8);
525 526 527 528 529 530 531 532 533
	}

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

	kfree(req);

	return status;
}

534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550
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;
}

551 552 553 554 555 556 557 558 559 560 561 562 563 564 565
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;
}

566 567
static void srp_disconnect_target(struct srp_target_port *target)
{
568 569
	if (srp_change_conn_state(target, false)) {
		/* XXX should send SRP_I_LOGOUT request */
570

571 572 573 574
		if (ib_send_cm_dreq(target->cm_id, NULL, 0)) {
			shost_printk(KERN_DEBUG, target->scsi_host,
				     PFX "Sending CM DREQ failed\n");
		}
575
	}
576 577
}

578 579
static void srp_free_req_data(struct srp_target_port *target)
{
580
	struct ib_device *ibdev = target->srp_host->srp_dev->dev;
581 582 583
	struct srp_request *req;
	int i;

584 585 586 587 588
	if (!target->req_ring)
		return;

	for (i = 0; i < target->req_ring_size; ++i) {
		req = &target->req_ring[i];
589 590
		kfree(req->fmr_list);
		kfree(req->map_page);
591 592 593 594 595 596
		if (req->indirect_dma_addr) {
			ib_dma_unmap_single(ibdev, req->indirect_dma_addr,
					    target->indirect_size,
					    DMA_TO_DEVICE);
		}
		kfree(req->indirect_desc);
597
	}
598 599 600

	kfree(target->req_ring);
	target->req_ring = NULL;
601 602
}

603 604 605 606 607 608 609 610 611 612
static int srp_alloc_req_data(struct srp_target_port *target)
{
	struct srp_device *srp_dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = srp_dev->dev;
	struct srp_request *req;
	dma_addr_t dma_addr;
	int i, ret = -ENOMEM;

	INIT_LIST_HEAD(&target->free_reqs);

613 614 615 616 617 618
	target->req_ring = kzalloc(target->req_ring_size *
				   sizeof(*target->req_ring), GFP_KERNEL);
	if (!target->req_ring)
		goto out;

	for (i = 0; i < target->req_ring_size; ++i) {
619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643
		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);
		req->indirect_desc = kmalloc(target->indirect_size, GFP_KERNEL);
		if (!req->fmr_list || !req->map_page || !req->indirect_desc)
			goto out;

		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 out;

		req->indirect_dma_addr = dma_addr;
		req->index = i;
		list_add_tail(&req->list, &target->free_reqs);
	}
	ret = 0;

out:
	return ret;
}

644 645 646 647 648 649 650 651 652 653 654 655 656 657 658
/**
 * 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);
}

659 660
static void srp_remove_target(struct srp_target_port *target)
{
661 662
	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);

663
	srp_del_scsi_host_attr(target->scsi_host);
664
	srp_rport_get(target->rport);
665 666
	srp_remove_host(target->scsi_host);
	scsi_remove_host(target->scsi_host);
667
	srp_stop_rport_timers(target->rport);
668
	srp_disconnect_target(target);
669 670
	ib_destroy_cm_id(target->cm_id);
	srp_free_target_ib(target);
671
	cancel_work_sync(&target->tl_err_work);
672
	srp_rport_put(target->rport);
673
	srp_free_req_data(target);
674 675 676 677 678

	spin_lock(&target->srp_host->target_lock);
	list_del(&target->list);
	spin_unlock(&target->srp_host->target_lock);

679 680 681
	scsi_host_put(target->scsi_host);
}

D
David Howells 已提交
682
static void srp_remove_work(struct work_struct *work)
683
{
D
David Howells 已提交
684
	struct srp_target_port *target =
685
		container_of(work, struct srp_target_port, remove_work);
686

687
	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);
688

689
	srp_remove_target(target);
690 691
}

692 693 694 695 696 697 698
static void srp_rport_delete(struct srp_rport *rport)
{
	struct srp_target_port *target = rport->lld_data;

	srp_queue_remove_work(target);
}

699 700
static int srp_connect_target(struct srp_target_port *target)
{
D
David Dillow 已提交
701
	int retries = 3;
702 703
	int ret;

704 705
	WARN_ON_ONCE(target->connected);

706 707
	target->qp_in_error = false;

708 709 710 711 712 713 714 715 716
	ret = srp_lookup_path(target);
	if (ret)
		return ret;

	while (1) {
		init_completion(&target->done);
		ret = srp_send_req(target);
		if (ret)
			return ret;
717 718 719
		ret = wait_for_completion_interruptible(&target->done);
		if (ret < 0)
			return ret;
720 721 722 723 724 725 726 727 728

		/*
		 * 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:
729
			srp_change_conn_state(target, true);
730 731 732 733 734 735 736 737 738 739 740
			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
741 742 743 744 745 746 747 748 749 750 751 752 753 754 755
		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;

756 757 758 759 760 761
		default:
			return target->status;
		}
	}
}

762 763 764 765
static void srp_unmap_data(struct scsi_cmnd *scmnd,
			   struct srp_target_port *target,
			   struct srp_request *req)
{
766 767 768
	struct ib_device *ibdev = target->srp_host->srp_dev->dev;
	struct ib_pool_fmr **pfmr;

769
	if (!scsi_sglist(scmnd) ||
770 771 772 773
	    (scmnd->sc_data_direction != DMA_TO_DEVICE &&
	     scmnd->sc_data_direction != DMA_FROM_DEVICE))
		return;

774 775 776
	pfmr = req->fmr_list;
	while (req->nfmr--)
		ib_fmr_pool_unmap(*pfmr++);
777

778 779
	ib_dma_unmap_sg(ibdev, scsi_sglist(scmnd), scsi_sg_count(scmnd),
			scmnd->sc_data_direction);
780 781
}

B
Bart Van Assche 已提交
782 783 784 785
/**
 * srp_claim_req - Take ownership of the scmnd associated with a request.
 * @target: SRP target port.
 * @req: SRP request.
786
 * @sdev: If not NULL, only take ownership for this SCSI device.
B
Bart Van Assche 已提交
787 788 789 790 791 792 793 794
 * @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,
795
				       struct scsi_device *sdev,
B
Bart Van Assche 已提交
796 797 798 799 800
				       struct scsi_cmnd *scmnd)
{
	unsigned long flags;

	spin_lock_irqsave(&target->lock, flags);
801 802 803
	if (req->scmnd &&
	    (!sdev || req->scmnd->device == sdev) &&
	    (!scmnd || req->scmnd == scmnd)) {
B
Bart Van Assche 已提交
804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819
		scmnd = req->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)
820
{
821 822
	unsigned long flags;

B
Bart Van Assche 已提交
823 824
	srp_unmap_data(scmnd, target, req);

825
	spin_lock_irqsave(&target->lock, flags);
826
	target->req_lim += req_lim_delta;
827
	list_add_tail(&req->list, &target->free_reqs);
828
	spin_unlock_irqrestore(&target->lock, flags);
829 830
}

831
static void srp_finish_req(struct srp_target_port *target,
832 833
			   struct srp_request *req, struct scsi_device *sdev,
			   int result)
834
{
835
	struct scsi_cmnd *scmnd = srp_claim_req(target, req, sdev, NULL);
B
Bart Van Assche 已提交
836 837

	if (scmnd) {
838
		srp_free_req(target, req, scmnd, 0);
839
		scmnd->result = result;
B
Bart Van Assche 已提交
840 841
		scmnd->scsi_done(scmnd);
	}
842 843
}

844
static void srp_terminate_io(struct srp_rport *rport)
845
{
846
	struct srp_target_port *target = rport->lld_data;
847 848
	struct Scsi_Host *shost = target->scsi_host;
	struct scsi_device *sdev;
849 850
	int i;

851 852 853 854 855 856 857
	/*
	 * Invoking srp_terminate_io() while srp_queuecommand() is running
	 * is not safe. Hence the warning statement below.
	 */
	shost_for_each_device(sdev, shost)
		WARN_ON_ONCE(sdev->request_queue->request_fn_active);

858
	for (i = 0; i < target->req_ring_size; ++i) {
859
		struct srp_request *req = &target->req_ring[i];
860
		srp_finish_req(target, req, NULL, DID_TRANSPORT_FAILFAST << 16);
861 862
	}
}
863

864 865 866 867 868 869 870 871 872 873 874 875 876
/*
 * It is up to the caller to ensure that srp_rport_reconnect() calls are
 * serialized and that no concurrent srp_queuecommand(), srp_abort(),
 * srp_reset_device() or srp_reset_host() calls will occur while this function
 * is in progress. One way to realize that is not to call this function
 * directly but to call srp_reconnect_rport() instead since that last function
 * serializes calls of this function via rport->mutex and also blocks
 * srp_queuecommand() calls before invoking this function.
 */
static int srp_rport_reconnect(struct srp_rport *rport)
{
	struct srp_target_port *target = rport->lld_data;
	int i, ret;
877

878 879
	srp_disconnect_target(target);
	/*
880 881 882
	 * 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.
883
	 */
D
David Dillow 已提交
884
	ret = srp_new_cm_id(target);
885 886 887 888 889 890 891 892 893
	/*
	 * 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);
894

895
	for (i = 0; i < target->req_ring_size; ++i) {
896
		struct srp_request *req = &target->req_ring[i];
897
		srp_finish_req(target, req, NULL, DID_RESET << 16);
898
	}
899

900
	INIT_LIST_HEAD(&target->free_tx);
901
	for (i = 0; i < target->queue_size; ++i)
902
		list_add(&target->tx_ring[i]->list, &target->free_tx);
903

904 905
	if (ret == 0)
		ret = srp_connect_target(target);
906

907 908 909
	if (ret == 0)
		shost_printk(KERN_INFO, target->scsi_host,
			     PFX "reconnect succeeded\n");
910 911 912 913

	return ret;
}

914 915
static void srp_map_desc(struct srp_map_state *state, dma_addr_t dma_addr,
			 unsigned int dma_len, u32 rkey)
916
{
917
	struct srp_direct_buf *desc = state->desc;
918

919 920 921
	desc->va = cpu_to_be64(dma_addr);
	desc->key = cpu_to_be32(rkey);
	desc->len = cpu_to_be32(dma_len);
922

923 924 925 926
	state->total_len += dma_len;
	state->desc++;
	state->ndesc++;
}
927

928 929 930 931 932 933
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;
934

935 936
	if (!state->npages)
		return 0;
937

938 939 940 941 942
	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;
943 944
	}

945 946 947 948
	fmr = ib_fmr_pool_map_phys(dev->fmr_pool, state->pages,
				   state->npages, io_addr);
	if (IS_ERR(fmr))
		return PTR_ERR(fmr);
949

950 951
	*state->next_fmr++ = fmr;
	state->nfmr++;
952

953 954 955 956 957 958 959 960 961 962 963 964 965
	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;
}
966

967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988
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;
989
	}
990

991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004
	/* 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;
1005 1006
	}

1007 1008 1009 1010 1011 1012 1013
	/* 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);
1014

1015 1016 1017 1018 1019
	while (dma_len) {
		if (state->npages == SRP_FMR_SIZE) {
			ret = srp_map_finish_fmr(state, target);
			if (ret)
				return ret;
1020

1021 1022 1023 1024
			srp_map_update_start(state, sg, sg_index, dma_addr);
		}

		len = min_t(unsigned int, dma_len, dev->fmr_page_size);
1025

1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043
		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);
	}
1044 1045 1046
	return ret;
}

1047 1048 1049
static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_target_port *target,
			struct srp_request *req)
{
1050
	struct scatterlist *scat, *sg;
1051
	struct srp_cmd *cmd = req->cmd->buf;
1052
	int i, len, nents, count, use_fmr;
1053 1054
	struct srp_device *dev;
	struct ib_device *ibdev;
1055 1056 1057 1058
	struct srp_map_state state;
	struct srp_indirect_buf *indirect_hdr;
	u32 table_len;
	u8 fmt;
1059

1060
	if (!scsi_sglist(scmnd) || scmnd->sc_data_direction == DMA_NONE)
1061 1062 1063 1064
		return sizeof (struct srp_cmd);

	if (scmnd->sc_data_direction != DMA_FROM_DEVICE &&
	    scmnd->sc_data_direction != DMA_TO_DEVICE) {
1065 1066 1067
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled data direction %d\n",
			     scmnd->sc_data_direction);
1068 1069 1070
		return -EINVAL;
	}

1071 1072
	nents = scsi_sg_count(scmnd);
	scat  = scsi_sglist(scmnd);
1073

1074
	dev = target->srp_host->srp_dev;
1075 1076 1077
	ibdev = dev->dev;

	count = ib_dma_map_sg(ibdev, scat, nents, scmnd->sc_data_direction);
1078 1079
	if (unlikely(count == 0))
		return -EIO;
1080 1081 1082

	fmt = SRP_DATA_DESC_DIRECT;
	len = sizeof (struct srp_cmd) +	sizeof (struct srp_direct_buf);
1083

1084
	if (count == 1) {
1085 1086 1087 1088 1089 1090
		/*
		 * 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.
		 */
1091
		struct srp_direct_buf *buf = (void *) cmd->add_data;
1092

1093
		buf->va  = cpu_to_be64(ib_sg_dma_address(ibdev, scat));
1094
		buf->key = cpu_to_be32(target->rkey);
1095
		buf->len = cpu_to_be32(ib_sg_dma_len(ibdev, scat));
1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106

		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;

1107 1108 1109
	ib_dma_sync_single_for_cpu(ibdev, req->indirect_dma_addr,
				   target->indirect_size, DMA_TO_DEVICE);

1110
	memset(&state, 0, sizeof(state));
1111
	state.desc	= req->indirect_desc;
1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134
	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);
1135
		}
1136
	}
1137

1138 1139
	if (use_fmr == SRP_MAP_ALLOW_FMR && srp_map_finish_fmr(&state, target))
		goto backtrack;
1140

1141 1142 1143 1144 1145
	/* 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.
1146 1147 1148 1149 1150 1151 1152
	 */
	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;
1153

1154
		*buf = req->indirect_desc[0];
1155
		goto map_complete;
1156 1157
	}

1158 1159 1160 1161 1162 1163 1164 1165
	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);
1166 1167 1168 1169
	table_len = state.ndesc * sizeof (struct srp_direct_buf);

	fmt = SRP_DATA_DESC_INDIRECT;
	len = sizeof(struct srp_cmd) + sizeof (struct srp_indirect_buf);
1170
	len += count * sizeof (struct srp_direct_buf);
1171

1172 1173
	memcpy(indirect_hdr->desc_list, req->indirect_desc,
	       count * sizeof (struct srp_direct_buf));
1174

1175
	indirect_hdr->table_desc.va = cpu_to_be64(req->indirect_dma_addr);
1176 1177 1178 1179 1180
	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)
1181
		cmd->data_out_desc_cnt = count;
1182
	else
1183 1184 1185 1186
		cmd->data_in_desc_cnt = count;

	ib_dma_sync_single_for_device(ibdev, req->indirect_dma_addr, table_len,
				      DMA_TO_DEVICE);
1187 1188

map_complete:
1189 1190 1191 1192 1193 1194 1195 1196
	if (scmnd->sc_data_direction == DMA_TO_DEVICE)
		cmd->buf_fmt = fmt << 4;
	else
		cmd->buf_fmt = fmt;

	return len;
}

1197 1198 1199 1200 1201 1202 1203 1204
/*
 * 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;

1205
	spin_lock_irqsave(&target->lock, flags);
1206 1207 1208
	list_add(&iu->list, &target->free_tx);
	if (iu_type != SRP_IU_RSP)
		++target->req_lim;
1209
	spin_unlock_irqrestore(&target->lock, flags);
1210 1211
}

1212
/*
1213 1214
 * 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().
1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232
 *
 * 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);

1233
	if (list_empty(&target->free_tx))
1234 1235 1236
		return NULL;

	/* Initiator responses to target requests do not consume credits */
1237 1238 1239 1240 1241 1242 1243
	if (iu_type != SRP_IU_RSP) {
		if (target->req_lim <= rsv) {
			++target->zero_req_lim;
			return NULL;
		}

		--target->req_lim;
1244 1245
	}

1246
	iu = list_first_entry(&target->free_tx, struct srp_iu, list);
1247
	list_del(&iu->list);
1248 1249 1250
	return iu;
}

1251 1252
static int srp_post_send(struct srp_target_port *target,
			 struct srp_iu *iu, int len)
1253 1254 1255 1256 1257 1258
{
	struct ib_sge list;
	struct ib_send_wr wr, *bad_wr;

	list.addr   = iu->dma;
	list.length = len;
1259
	list.lkey   = target->lkey;
1260 1261

	wr.next       = NULL;
1262
	wr.wr_id      = (uintptr_t) iu;
1263 1264 1265 1266 1267
	wr.sg_list    = &list;
	wr.num_sge    = 1;
	wr.opcode     = IB_WR_SEND;
	wr.send_flags = IB_SEND_SIGNALED;

1268
	return ib_post_send(target->qp, &wr, &bad_wr);
1269 1270
}

1271
static int srp_post_recv(struct srp_target_port *target, struct srp_iu *iu)
1272 1273
{
	struct ib_recv_wr wr, *bad_wr;
1274
	struct ib_sge list;
1275 1276 1277

	list.addr   = iu->dma;
	list.length = iu->size;
1278
	list.lkey   = target->lkey;
1279 1280

	wr.next     = NULL;
1281
	wr.wr_id    = (uintptr_t) iu;
1282 1283 1284
	wr.sg_list  = &list;
	wr.num_sge  = 1;

1285
	return ib_post_recv(target->qp, &wr, &bad_wr);
1286 1287
}

1288 1289 1290 1291 1292 1293 1294
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)) {
1295
		spin_lock_irqsave(&target->lock, flags);
1296
		target->req_lim += be32_to_cpu(rsp->req_lim_delta);
1297
		spin_unlock_irqrestore(&target->lock, flags);
1298

1299 1300 1301 1302
		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);
1303
	} else {
1304
		req = &target->req_ring[rsp->tag];
1305
		scmnd = srp_claim_req(target, req, NULL, NULL);
B
Bart Van Assche 已提交
1306
		if (!scmnd) {
1307 1308 1309
			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
1310 1311 1312 1313 1314 1315 1316

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

			return;
		}
1317 1318 1319 1320 1321 1322 1323 1324 1325 1326
		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))
1327
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_out_res_cnt));
1328
		else if (rsp->flags & (SRP_RSP_FLAG_DIOVER | SRP_RSP_FLAG_DIUNDER))
1329
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_in_res_cnt));
1330

B
Bart Van Assche 已提交
1331 1332 1333
		srp_free_req(target, req, scmnd,
			     be32_to_cpu(rsp->req_lim_delta));

1334 1335
		scmnd->host_scribble = NULL;
		scmnd->scsi_done(scmnd);
1336 1337 1338
	}
}

1339 1340 1341
static int srp_response_common(struct srp_target_port *target, s32 req_delta,
			       void *rsp, int len)
{
1342
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1343 1344
	unsigned long flags;
	struct srp_iu *iu;
1345
	int err;
1346

1347
	spin_lock_irqsave(&target->lock, flags);
1348 1349
	target->req_lim += req_delta;
	iu = __srp_get_tx_iu(target, SRP_IU_RSP);
1350
	spin_unlock_irqrestore(&target->lock, flags);
1351

1352 1353 1354
	if (!iu) {
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "no IU available to send response\n");
1355
		return 1;
1356 1357 1358 1359 1360 1361
	}

	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);

1362 1363
	err = srp_post_send(target, iu, len);
	if (err) {
1364 1365
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "unable to post response: %d\n", err);
1366 1367
		srp_put_tx_iu(target, iu, SRP_IU_RSP);
	}
1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402

	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");
}

1403 1404
static void srp_handle_recv(struct srp_target_port *target, struct ib_wc *wc)
{
1405
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1406
	struct srp_iu *iu = (struct srp_iu *) (uintptr_t) wc->wr_id;
1407
	int res;
1408 1409
	u8 opcode;

1410 1411
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_ti_iu_len,
				   DMA_FROM_DEVICE);
1412 1413 1414 1415

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

	if (0) {
1416 1417
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "recv completion, opcode 0x%02x\n", opcode);
B
Bart Van Assche 已提交
1418 1419
		print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET, 8, 1,
			       iu->buf, wc->byte_len, true);
1420 1421 1422 1423 1424 1425 1426
	}

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

1427 1428 1429 1430 1431 1432 1433 1434
	case SRP_CRED_REQ:
		srp_process_cred_req(target, iu->buf);
		break;

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

1435 1436
	case SRP_T_LOGOUT:
		/* XXX Handle target logout */
1437 1438
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Got target logout request\n");
1439 1440 1441
		break;

	default:
1442 1443
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled SRP opcode 0x%02x\n", opcode);
1444 1445 1446
		break;
	}

1447 1448
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_ti_iu_len,
				      DMA_FROM_DEVICE);
1449

1450
	res = srp_post_recv(target, iu);
1451 1452 1453
	if (res != 0)
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Recv failed with error code %d\n", res);
1454 1455
}

1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470
/**
 * srp_tl_err_work() - handle a transport layer error
 *
 * Note: This function may get invoked before the rport has been created,
 * hence the target->rport test.
 */
static void srp_tl_err_work(struct work_struct *work)
{
	struct srp_target_port *target;

	target = container_of(work, struct srp_target_port, tl_err_work);
	if (target->rport)
		srp_start_tl_fail_timers(target->rport);
}

1471
static void srp_handle_qp_err(enum ib_wc_status wc_status, bool send_err,
1472 1473
			      struct srp_target_port *target)
{
1474
	if (target->connected && !target->qp_in_error) {
1475 1476
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "failed %s status %d\n",
1477
			     send_err ? "send" : "receive",
1478
			     wc_status);
1479
		queue_work(system_long_wq, &target->tl_err_work);
1480
	}
1481 1482 1483
	target->qp_in_error = true;
}

1484
static void srp_recv_completion(struct ib_cq *cq, void *target_ptr)
1485 1486 1487 1488 1489 1490
{
	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) {
1491 1492 1493
		if (likely(wc.status == IB_WC_SUCCESS)) {
			srp_handle_recv(target, &wc);
		} else {
1494
			srp_handle_qp_err(wc.status, false, target);
1495
		}
1496 1497 1498 1499 1500 1501 1502
	}
}

static void srp_send_completion(struct ib_cq *cq, void *target_ptr)
{
	struct srp_target_port *target = target_ptr;
	struct ib_wc wc;
1503
	struct srp_iu *iu;
1504 1505

	while (ib_poll_cq(cq, 1, &wc) > 0) {
1506 1507 1508 1509
		if (likely(wc.status == IB_WC_SUCCESS)) {
			iu = (struct srp_iu *) (uintptr_t) wc.wr_id;
			list_add(&iu->list, &target->free_tx);
		} else {
1510
			srp_handle_qp_err(wc.status, true, target);
1511
		}
1512 1513 1514
	}
}

1515
static int srp_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *scmnd)
1516
{
1517
	struct srp_target_port *target = host_to_target(shost);
1518
	struct srp_rport *rport = target->rport;
1519 1520 1521
	struct srp_request *req;
	struct srp_iu *iu;
	struct srp_cmd *cmd;
1522
	struct ib_device *dev;
1523
	unsigned long flags;
1524
	int len, result;
1525 1526 1527 1528 1529 1530 1531 1532 1533 1534
	const bool in_scsi_eh = !in_interrupt() && current == shost->ehandler;

	/*
	 * The SCSI EH thread is the only context from which srp_queuecommand()
	 * can get invoked for blocked devices (SDEV_BLOCK /
	 * SDEV_CREATED_BLOCK). Avoid racing with srp_reconnect_rport() by
	 * locking the rport mutex if invoked from inside the SCSI EH.
	 */
	if (in_scsi_eh)
		mutex_lock(&rport->mutex);
1535

1536 1537 1538
	result = srp_chkready(target->rport);
	if (unlikely(result)) {
		scmnd->result = result;
1539
		scmnd->scsi_done(scmnd);
1540
		goto unlock_rport;
1541 1542
	}

1543
	spin_lock_irqsave(&target->lock, flags);
1544
	iu = __srp_get_tx_iu(target, SRP_IU_CMD);
1545
	if (!iu)
1546 1547 1548 1549 1550
		goto err_unlock;

	req = list_first_entry(&target->free_reqs, struct srp_request, list);
	list_del(&req->list);
	spin_unlock_irqrestore(&target->lock, flags);
1551

1552
	dev = target->srp_host->srp_dev->dev;
1553
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_iu_len,
1554
				   DMA_TO_DEVICE);
1555 1556

	scmnd->result        = 0;
1557
	scmnd->host_scribble = (void *) req;
1558 1559 1560 1561 1562 1563

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

	cmd->opcode = SRP_CMD;
	cmd->lun    = cpu_to_be64((u64) scmnd->device->lun << 48);
1564
	cmd->tag    = req->index;
1565 1566 1567 1568 1569 1570 1571
	memcpy(cmd->cdb, scmnd->cmnd, scmnd->cmd_len);

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

	len = srp_map_data(scmnd, target, req);
	if (len < 0) {
1572 1573
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Failed to map data\n");
1574
		goto err_iu;
1575 1576
	}

1577
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_iu_len,
1578
				      DMA_TO_DEVICE);
1579

1580
	if (srp_post_send(target, iu, len)) {
1581
		shost_printk(KERN_ERR, target->scsi_host, PFX "Send failed\n");
1582 1583 1584
		goto err_unmap;
	}

1585 1586 1587 1588
unlock_rport:
	if (in_scsi_eh)
		mutex_unlock(&rport->mutex);

1589 1590 1591 1592 1593
	return 0;

err_unmap:
	srp_unmap_data(scmnd, target, req);

1594 1595 1596
err_iu:
	srp_put_tx_iu(target, iu, SRP_IU_CMD);

1597 1598 1599 1600 1601 1602
	/*
	 * Avoid that the loops that iterate over the request ring can
	 * encounter a dangling SCSI command pointer.
	 */
	req->scmnd = NULL;

1603
	spin_lock_irqsave(&target->lock, flags);
1604
	list_add(&req->list, &target->free_reqs);
1605 1606

err_unlock:
1607
	spin_unlock_irqrestore(&target->lock, flags);
1608

1609 1610 1611
	if (in_scsi_eh)
		mutex_unlock(&rport->mutex);

1612 1613 1614
	return SCSI_MLQUEUE_HOST_BUSY;
}

1615 1616 1617 1618
/*
 * Note: the resources allocated in this function are freed in
 * srp_free_target_ib().
 */
1619 1620 1621 1622
static int srp_alloc_iu_bufs(struct srp_target_port *target)
{
	int i;

1623 1624 1625 1626 1627 1628 1629 1630 1631 1632
	target->rx_ring = kzalloc(target->queue_size * sizeof(*target->rx_ring),
				  GFP_KERNEL);
	if (!target->rx_ring)
		goto err_no_ring;
	target->tx_ring = kzalloc(target->queue_size * sizeof(*target->tx_ring),
				  GFP_KERNEL);
	if (!target->tx_ring)
		goto err_no_ring;

	for (i = 0; i < target->queue_size; ++i) {
1633 1634 1635 1636 1637 1638 1639
		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;
	}

1640
	for (i = 0; i < target->queue_size; ++i) {
1641
		target->tx_ring[i] = srp_alloc_iu(target->srp_host,
1642
						  target->max_iu_len,
1643 1644 1645
						  GFP_KERNEL, DMA_TO_DEVICE);
		if (!target->tx_ring[i])
			goto err;
1646 1647

		list_add(&target->tx_ring[i]->list, &target->free_tx);
1648 1649 1650 1651 1652
	}

	return 0;

err:
1653
	for (i = 0; i < target->queue_size; ++i) {
1654 1655 1656 1657
		srp_free_iu(target->srp_host, target->rx_ring[i]);
		srp_free_iu(target->srp_host, target->tx_ring[i]);
	}

1658 1659 1660 1661 1662 1663 1664

err_no_ring:
	kfree(target->tx_ring);
	target->tx_ring = NULL;
	kfree(target->rx_ring);
	target->rx_ring = NULL;

1665 1666 1667
	return -ENOMEM;
}

1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694
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;
}

1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714
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);
1715 1716 1717
		target->scsi_host->cmd_per_lun
			= min_t(int, target->scsi_host->can_queue,
				target->scsi_host->cmd_per_lun);
1718 1719 1720 1721 1722 1723 1724
	} else {
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled RSP opcode %#x\n", lrsp->opcode);
		ret = -ECONNRESET;
		goto error;
	}

1725
	if (!target->rx_ring) {
1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744
		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;

1745
	for (i = 0; i < target->queue_size; i++) {
1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756
		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;

1757 1758
	target->rq_tmo_jiffies = srp_compute_rq_tmo(qp_attr, attr_mask);

1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771
	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;
}

1772 1773 1774 1775
static void srp_cm_rej_handler(struct ib_cm_id *cm_id,
			       struct ib_cm_event *event,
			       struct srp_target_port *target)
{
1776
	struct Scsi_Host *shost = target->scsi_host;
1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792
	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:
1793
		if (srp_target_is_topspin(target)) {
1794 1795 1796 1797 1798 1799 1800 1801
			/*
			 * 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);

1802 1803 1804 1805
			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));
1806 1807 1808

			target->status = SRP_PORT_REDIRECT;
		} else {
1809 1810
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_PORT_REDIRECT\n");
1811 1812 1813 1814 1815
			target->status = -ECONNRESET;
		}
		break;

	case IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID:
1816 1817
		shost_printk(KERN_WARNING, shost,
			    "  REJ reason: IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID\n");
1818 1819 1820 1821 1822 1823 1824 1825 1826 1827
		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)
1828 1829
				shost_printk(KERN_WARNING, shost,
					     PFX "SRP_LOGIN_REJ: requested max_it_iu_len too large\n");
1830
			else
B
Bart Van Assche 已提交
1831 1832 1833 1834
				shost_printk(KERN_WARNING, shost, PFX
					     "SRP LOGIN from %pI6 to %pI6 REJECTED, reason 0x%08x\n",
					     target->path.sgid.raw,
					     target->orig_dgid, reason);
1835
		} else
1836 1837 1838
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_CONSUMER_DEFINED,"
				     " opcode 0x%02x\n", opcode);
1839 1840 1841
		target->status = -ECONNRESET;
		break;

D
David Dillow 已提交
1842 1843 1844 1845 1846
	case IB_CM_REJ_STALE_CONN:
		shost_printk(KERN_WARNING, shost, "  REJ reason: stale connection\n");
		target->status = SRP_STALE_CONN;
		break;

1847
	default:
1848 1849
		shost_printk(KERN_WARNING, shost, "  REJ reason 0x%x\n",
			     event->param.rej_rcvd.reason);
1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860
		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:
1861 1862
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Sending CM REQ failed\n");
1863 1864 1865 1866 1867 1868
		comp = 1;
		target->status = -ECONNRESET;
		break;

	case IB_CM_REP_RECEIVED:
		comp = 1;
1869
		srp_cm_rep_handler(cm_id, event->private_data, target);
1870 1871 1872
		break;

	case IB_CM_REJ_RECEIVED:
1873
		shost_printk(KERN_DEBUG, target->scsi_host, PFX "REJ received\n");
1874 1875 1876 1877 1878
		comp = 1;

		srp_cm_rej_handler(cm_id, event, target);
		break;

1879
	case IB_CM_DREQ_RECEIVED:
1880 1881
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "DREQ received - connection closed\n");
1882
		srp_change_conn_state(target, false);
1883
		if (ib_send_cm_drep(cm_id, NULL, 0))
1884 1885
			shost_printk(KERN_ERR, target->scsi_host,
				     PFX "Sending CM DREP failed\n");
1886
		queue_work(system_long_wq, &target->tl_err_work);
1887 1888 1889
		break;

	case IB_CM_TIMEWAIT_EXIT:
1890 1891
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "connection closed\n");
1892
		comp = 1;
1893 1894 1895 1896

		target->status = 0;
		break;

1897 1898 1899 1900 1901
	case IB_CM_MRA_RECEIVED:
	case IB_CM_DREQ_ERROR:
	case IB_CM_DREP_RECEIVED:
		break;

1902
	default:
1903 1904
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled CM event %d\n", event->event);
1905 1906 1907 1908 1909 1910 1911 1912 1913
		break;
	}

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

	return 0;
}

1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964
/**
 * srp_change_queue_type - changing device queue tag type
 * @sdev: scsi device struct
 * @tag_type: requested tag type
 *
 * Returns queue tag type.
 */
static int
srp_change_queue_type(struct scsi_device *sdev, int tag_type)
{
	if (sdev->tagged_supported) {
		scsi_set_tag_type(sdev, tag_type);
		if (tag_type)
			scsi_activate_tcq(sdev, sdev->queue_depth);
		else
			scsi_deactivate_tcq(sdev, sdev->queue_depth);
	} else
		tag_type = 0;

	return tag_type;
}

/**
 * srp_change_queue_depth - setting device queue depth
 * @sdev: scsi device struct
 * @qdepth: requested queue depth
 * @reason: SCSI_QDEPTH_DEFAULT/SCSI_QDEPTH_QFULL/SCSI_QDEPTH_RAMP_UP
 * (see include/scsi/scsi_host.h for definition)
 *
 * Returns queue depth.
 */
static int
srp_change_queue_depth(struct scsi_device *sdev, int qdepth, int reason)
{
	struct Scsi_Host *shost = sdev->host;
	int max_depth;
	if (reason == SCSI_QDEPTH_DEFAULT || reason == SCSI_QDEPTH_RAMP_UP) {
		max_depth = shost->can_queue;
		if (!sdev->tagged_supported)
			max_depth = 1;
		if (qdepth > max_depth)
			qdepth = max_depth;
		scsi_adjust_queue_depth(sdev, scsi_get_tag_type(sdev), qdepth);
	} else if (reason == SCSI_QDEPTH_QFULL)
		scsi_track_queue_full(sdev, qdepth);
	else
		return -EOPNOTSUPP;

	return sdev->queue_depth;
}

1965
static int srp_send_tsk_mgmt(struct srp_target_port *target,
1966
			     u64 req_tag, unsigned int lun, u8 func)
1967
{
1968
	struct srp_rport *rport = target->rport;
1969
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1970 1971 1972
	struct srp_iu *iu;
	struct srp_tsk_mgmt *tsk_mgmt;

1973 1974 1975
	if (!target->connected || target->qp_in_error)
		return -1;

1976
	init_completion(&target->tsk_mgmt_done);
1977

1978 1979 1980 1981 1982
	/*
	 * Lock the rport mutex to avoid that srp_create_target_ib() is
	 * invoked while a task management function is being sent.
	 */
	mutex_lock(&rport->mutex);
1983
	spin_lock_irq(&target->lock);
1984
	iu = __srp_get_tx_iu(target, SRP_IU_TSK_MGMT);
1985
	spin_unlock_irq(&target->lock);
1986

1987 1988 1989
	if (!iu) {
		mutex_unlock(&rport->mutex);

1990
		return -1;
1991
	}
1992

1993 1994
	ib_dma_sync_single_for_cpu(dev, iu->dma, sizeof *tsk_mgmt,
				   DMA_TO_DEVICE);
1995 1996 1997 1998
	tsk_mgmt = iu->buf;
	memset(tsk_mgmt, 0, sizeof *tsk_mgmt);

	tsk_mgmt->opcode 	= SRP_TSK_MGMT;
1999 2000
	tsk_mgmt->lun		= cpu_to_be64((u64) lun << 48);
	tsk_mgmt->tag		= req_tag | SRP_TAG_TSK_MGMT;
2001
	tsk_mgmt->tsk_mgmt_func = func;
2002
	tsk_mgmt->task_tag	= req_tag;
2003

2004 2005
	ib_dma_sync_single_for_device(dev, iu->dma, sizeof *tsk_mgmt,
				      DMA_TO_DEVICE);
2006 2007
	if (srp_post_send(target, iu, sizeof *tsk_mgmt)) {
		srp_put_tx_iu(target, iu, SRP_IU_TSK_MGMT);
2008 2009
		mutex_unlock(&rport->mutex);

2010 2011
		return -1;
	}
2012
	mutex_unlock(&rport->mutex);
2013

2014
	if (!wait_for_completion_timeout(&target->tsk_mgmt_done,
2015
					 msecs_to_jiffies(SRP_ABORT_TIMEOUT_MS)))
2016
		return -1;
2017

2018 2019 2020
	return 0;
}

2021 2022
static int srp_abort(struct scsi_cmnd *scmnd)
{
2023
	struct srp_target_port *target = host_to_target(scmnd->device->host);
2024
	struct srp_request *req = (struct srp_request *) scmnd->host_scribble;
2025
	int ret;
2026

2027
	shost_printk(KERN_ERR, target->scsi_host, "SRP abort called\n");
2028

2029
	if (!req || !srp_claim_req(target, req, NULL, scmnd))
2030
		return SUCCESS;
2031
	if (srp_send_tsk_mgmt(target, req->index, scmnd->device->lun,
2032
			      SRP_TSK_ABORT_TASK) == 0)
2033
		ret = SUCCESS;
2034
	else if (target->rport->state == SRP_RPORT_LOST)
2035
		ret = FAST_IO_FAIL;
2036 2037
	else
		ret = FAILED;
B
Bart Van Assche 已提交
2038 2039
	srp_free_req(target, req, scmnd, 0);
	scmnd->result = DID_ABORT << 16;
2040
	scmnd->scsi_done(scmnd);
2041

2042
	return ret;
2043 2044 2045 2046
}

static int srp_reset_device(struct scsi_cmnd *scmnd)
{
2047
	struct srp_target_port *target = host_to_target(scmnd->device->host);
2048
	int i;
2049

2050
	shost_printk(KERN_ERR, target->scsi_host, "SRP reset_device called\n");
2051

2052 2053
	if (srp_send_tsk_mgmt(target, SRP_TAG_NO_REQ, scmnd->device->lun,
			      SRP_TSK_LUN_RESET))
2054
		return FAILED;
2055
	if (target->tsk_mgmt_status)
2056 2057
		return FAILED;

2058
	for (i = 0; i < target->req_ring_size; ++i) {
2059
		struct srp_request *req = &target->req_ring[i];
2060
		srp_finish_req(target, req, scmnd->device, DID_RESET << 16);
2061
	}
2062 2063

	return SUCCESS;
2064 2065 2066 2067 2068 2069
}

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

2070
	shost_printk(KERN_ERR, target->scsi_host, PFX "SRP reset_host called\n");
2071

2072
	return srp_reconnect_rport(target->rport) == 0 ? SUCCESS : FAILED;
2073 2074
}

2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089
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;
}

2090 2091
static ssize_t show_id_ext(struct device *dev, struct device_attribute *attr,
			   char *buf)
2092
{
2093
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2094 2095 2096 2097 2098

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

2099 2100
static ssize_t show_ioc_guid(struct device *dev, struct device_attribute *attr,
			     char *buf)
2101
{
2102
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2103 2104 2105 2106 2107

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

2108 2109
static ssize_t show_service_id(struct device *dev,
			       struct device_attribute *attr, char *buf)
2110
{
2111
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2112 2113 2114 2115 2116

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

2117 2118
static ssize_t show_pkey(struct device *dev, struct device_attribute *attr,
			 char *buf)
2119
{
2120
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2121 2122 2123 2124

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

B
Bart Van Assche 已提交
2125 2126 2127 2128 2129 2130 2131 2132
static ssize_t show_sgid(struct device *dev, struct device_attribute *attr,
			 char *buf)
{
	struct srp_target_port *target = host_to_target(class_to_shost(dev));

	return sprintf(buf, "%pI6\n", target->path.sgid.raw);
}

2133 2134
static ssize_t show_dgid(struct device *dev, struct device_attribute *attr,
			 char *buf)
2135
{
2136
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2137

H
Harvey Harrison 已提交
2138
	return sprintf(buf, "%pI6\n", target->path.dgid.raw);
2139 2140
}

2141 2142
static ssize_t show_orig_dgid(struct device *dev,
			      struct device_attribute *attr, char *buf)
2143
{
2144
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2145

H
Harvey Harrison 已提交
2146
	return sprintf(buf, "%pI6\n", target->orig_dgid);
2147 2148
}

2149 2150 2151 2152 2153 2154 2155 2156
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);
}

2157 2158
static ssize_t show_zero_req_lim(struct device *dev,
				 struct device_attribute *attr, char *buf)
2159
{
2160
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2161 2162 2163 2164

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

2165 2166
static ssize_t show_local_ib_port(struct device *dev,
				  struct device_attribute *attr, char *buf)
2167
{
2168
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2169 2170 2171 2172

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

2173 2174
static ssize_t show_local_ib_device(struct device *dev,
				    struct device_attribute *attr, char *buf)
2175
{
2176
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2177

2178
	return sprintf(buf, "%s\n", target->srp_host->srp_dev->dev->name);
2179 2180
}

2181 2182 2183 2184 2185 2186 2187 2188
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);
}

2189 2190 2191 2192 2193 2194 2195 2196
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);
}

2197 2198 2199 2200 2201 2202 2203 2204
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);
}

2205 2206 2207 2208 2209 2210 2211 2212
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");
}

2213 2214 2215 2216
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);
B
Bart Van Assche 已提交
2217
static DEVICE_ATTR(sgid,	    S_IRUGO, show_sgid,		   NULL);
2218 2219
static DEVICE_ATTR(dgid,	    S_IRUGO, show_dgid,		   NULL);
static DEVICE_ATTR(orig_dgid,	    S_IRUGO, show_orig_dgid,	   NULL);
2220
static DEVICE_ATTR(req_lim,         S_IRUGO, show_req_lim,         NULL);
2221 2222 2223
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);
2224
static DEVICE_ATTR(comp_vector,     S_IRUGO, show_comp_vector,     NULL);
2225
static DEVICE_ATTR(tl_retry_count,  S_IRUGO, show_tl_retry_count,  NULL);
2226
static DEVICE_ATTR(cmd_sg_entries,  S_IRUGO, show_cmd_sg_entries,  NULL);
2227
static DEVICE_ATTR(allow_ext_sg,    S_IRUGO, show_allow_ext_sg,    NULL);
2228 2229 2230 2231 2232 2233

static struct device_attribute *srp_host_attrs[] = {
	&dev_attr_id_ext,
	&dev_attr_ioc_guid,
	&dev_attr_service_id,
	&dev_attr_pkey,
B
Bart Van Assche 已提交
2234
	&dev_attr_sgid,
2235 2236
	&dev_attr_dgid,
	&dev_attr_orig_dgid,
2237
	&dev_attr_req_lim,
2238 2239 2240
	&dev_attr_zero_req_lim,
	&dev_attr_local_ib_port,
	&dev_attr_local_ib_device,
2241
	&dev_attr_comp_vector,
2242
	&dev_attr_tl_retry_count,
2243
	&dev_attr_cmd_sg_entries,
2244
	&dev_attr_allow_ext_sg,
2245 2246 2247
	NULL
};

2248 2249
static struct scsi_host_template srp_template = {
	.module				= THIS_MODULE,
R
Roland Dreier 已提交
2250 2251
	.name				= "InfiniBand SRP initiator",
	.proc_name			= DRV_NAME,
2252
	.slave_configure		= srp_slave_configure,
2253 2254
	.info				= srp_target_info,
	.queuecommand			= srp_queuecommand,
2255 2256
	.change_queue_depth             = srp_change_queue_depth,
	.change_queue_type              = srp_change_queue_type,
2257 2258 2259
	.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
2260
	.skip_settle_delay		= true,
2261
	.sg_tablesize			= SRP_DEF_SG_TABLESIZE,
2262
	.can_queue			= SRP_DEFAULT_CMD_SQ_SIZE,
2263
	.this_id			= -1,
2264
	.cmd_per_lun			= SRP_DEFAULT_CMD_SQ_SIZE,
2265 2266
	.use_clustering			= ENABLE_CLUSTERING,
	.shost_attrs			= srp_host_attrs
2267 2268 2269 2270
};

static int srp_add_target(struct srp_host *host, struct srp_target_port *target)
{
2271 2272 2273
	struct srp_rport_identifiers ids;
	struct srp_rport *rport;

2274 2275 2276
	sprintf(target->target_name, "SRP.T10:%016llX",
		 (unsigned long long) be64_to_cpu(target->id_ext));

2277
	if (scsi_add_host(target->scsi_host, host->srp_dev->dev->dma_device))
2278 2279
		return -ENODEV;

2280 2281
	memcpy(ids.port_id, &target->id_ext, 8);
	memcpy(ids.port_id + 8, &target->ioc_guid, 8);
2282
	ids.roles = SRP_RPORT_ROLE_TARGET;
2283 2284 2285 2286 2287 2288
	rport = srp_rport_add(target->scsi_host, &ids);
	if (IS_ERR(rport)) {
		scsi_remove_host(target->scsi_host);
		return PTR_ERR(rport);
	}

2289
	rport->lld_data = target;
2290
	target->rport = rport;
2291

2292
	spin_lock(&host->target_lock);
2293
	list_add_tail(&target->list, &host->target_list);
2294
	spin_unlock(&host->target_lock);
2295 2296 2297 2298

	target->state = SRP_TARGET_LIVE;

	scsi_scan_target(&target->scsi_host->shost_gendev,
2299
			 0, target->scsi_id, SCAN_WILD_CARD, 0);
2300 2301 2302 2303

	return 0;
}

2304
static void srp_release_dev(struct device *dev)
2305 2306
{
	struct srp_host *host =
2307
		container_of(dev, struct srp_host, dev);
2308 2309 2310 2311 2312 2313

	complete(&host->released);
}

static struct class srp_class = {
	.name    = "infiniband_srp",
2314
	.dev_release = srp_release_dev
2315 2316
};

2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346
/**
 * 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;
}

2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362
/*
 * 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,
2363
	SRP_OPT_MAX_CMD_PER_LUN	= 1 << 6,
2364
	SRP_OPT_IO_CLASS	= 1 << 7,
2365
	SRP_OPT_INITIATOR_EXT	= 1 << 8,
2366
	SRP_OPT_CMD_SG_ENTRIES	= 1 << 9,
2367 2368
	SRP_OPT_ALLOW_EXT_SG	= 1 << 10,
	SRP_OPT_SG_TABLESIZE	= 1 << 11,
2369
	SRP_OPT_COMP_VECTOR	= 1 << 12,
2370
	SRP_OPT_TL_RETRY_COUNT	= 1 << 13,
2371
	SRP_OPT_QUEUE_SIZE	= 1 << 14,
2372 2373 2374 2375 2376 2377 2378
	SRP_OPT_ALL		= (SRP_OPT_ID_EXT	|
				   SRP_OPT_IOC_GUID	|
				   SRP_OPT_DGID		|
				   SRP_OPT_PKEY		|
				   SRP_OPT_SERVICE_ID),
};

2379
static const match_table_t srp_opt_tokens = {
2380 2381 2382 2383 2384 2385 2386
	{ 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" 	},
2387
	{ SRP_OPT_IO_CLASS,		"io_class=%x"		},
2388
	{ SRP_OPT_INITIATOR_EXT,	"initiator_ext=%s"	},
2389
	{ SRP_OPT_CMD_SG_ENTRIES,	"cmd_sg_entries=%u"	},
2390 2391
	{ SRP_OPT_ALLOW_EXT_SG,		"allow_ext_sg=%u"	},
	{ SRP_OPT_SG_TABLESIZE,		"sg_tablesize=%u"	},
2392
	{ SRP_OPT_COMP_VECTOR,		"comp_vector=%u"	},
2393
	{ SRP_OPT_TL_RETRY_COUNT,	"tl_retry_count=%u"	},
2394
	{ SRP_OPT_QUEUE_SIZE,		"queue_size=%d"		},
2395
	{ SRP_OPT_ERR,			NULL 			}
2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423
};

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);
2424 2425 2426 2427
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2428 2429 2430 2431 2432 2433
			target->id_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_IOC_GUID:
			p = match_strdup(args);
2434 2435 2436 2437
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2438 2439 2440 2441 2442 2443
			target->ioc_guid = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_DGID:
			p = match_strdup(args);
2444 2445 2446 2447
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2448
			if (strlen(p) != 32) {
2449
				pr_warn("bad dest GID parameter '%s'\n", p);
2450
				kfree(p);
2451 2452 2453 2454 2455 2456 2457
				goto out;
			}

			for (i = 0; i < 16; ++i) {
				strlcpy(dgid, p + i * 2, 3);
				target->path.dgid.raw[i] = simple_strtoul(dgid, NULL, 16);
			}
2458
			kfree(p);
2459
			memcpy(target->orig_dgid, target->path.dgid.raw, 16);
2460 2461 2462 2463
			break;

		case SRP_OPT_PKEY:
			if (match_hex(args, &token)) {
2464
				pr_warn("bad P_Key parameter '%s'\n", p);
2465 2466 2467 2468 2469 2470 2471
				goto out;
			}
			target->path.pkey = cpu_to_be16(token);
			break;

		case SRP_OPT_SERVICE_ID:
			p = match_strdup(args);
2472 2473 2474 2475
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2476
			target->service_id = cpu_to_be64(simple_strtoull(p, NULL, 16));
2477
			target->path.service_id = target->service_id;
2478 2479 2480 2481 2482
			kfree(p);
			break;

		case SRP_OPT_MAX_SECT:
			if (match_int(args, &token)) {
2483
				pr_warn("bad max sect parameter '%s'\n", p);
2484 2485 2486 2487 2488
				goto out;
			}
			target->scsi_host->max_sectors = token;
			break;

2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500
		case SRP_OPT_QUEUE_SIZE:
			if (match_int(args, &token) || token < 1) {
				pr_warn("bad queue_size parameter '%s'\n", p);
				goto out;
			}
			target->scsi_host->can_queue = token;
			target->queue_size = token + SRP_RSP_SQ_SIZE +
					     SRP_TSK_MGMT_SQ_SIZE;
			if (!(opt_mask & SRP_OPT_MAX_CMD_PER_LUN))
				target->scsi_host->cmd_per_lun = token;
			break;

2501
		case SRP_OPT_MAX_CMD_PER_LUN:
2502
			if (match_int(args, &token) || token < 1) {
2503 2504
				pr_warn("bad max cmd_per_lun parameter '%s'\n",
					p);
2505 2506
				goto out;
			}
2507
			target->scsi_host->cmd_per_lun = token;
2508 2509
			break;

2510 2511
		case SRP_OPT_IO_CLASS:
			if (match_hex(args, &token)) {
2512
				pr_warn("bad IO class parameter '%s'\n", p);
2513 2514 2515 2516
				goto out;
			}
			if (token != SRP_REV10_IB_IO_CLASS &&
			    token != SRP_REV16A_IB_IO_CLASS) {
2517 2518 2519
				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);
2520 2521 2522 2523 2524
				goto out;
			}
			target->io_class = token;
			break;

2525 2526
		case SRP_OPT_INITIATOR_EXT:
			p = match_strdup(args);
2527 2528 2529 2530
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2531 2532 2533 2534
			target->initiator_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

2535 2536
		case SRP_OPT_CMD_SG_ENTRIES:
			if (match_int(args, &token) || token < 1 || token > 255) {
2537 2538
				pr_warn("bad max cmd_sg_entries parameter '%s'\n",
					p);
2539 2540 2541 2542 2543
				goto out;
			}
			target->cmd_sg_cnt = token;
			break;

2544 2545
		case SRP_OPT_ALLOW_EXT_SG:
			if (match_int(args, &token)) {
2546
				pr_warn("bad allow_ext_sg parameter '%s'\n", p);
2547 2548 2549 2550 2551 2552 2553 2554
				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) {
2555 2556
				pr_warn("bad max sg_tablesize parameter '%s'\n",
					p);
2557 2558 2559 2560 2561
				goto out;
			}
			target->sg_tablesize = token;
			break;

2562 2563 2564 2565 2566 2567 2568 2569
		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;

2570 2571 2572 2573 2574 2575 2576 2577 2578
		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;

2579
		default:
2580 2581
			pr_warn("unknown parameter or missing value '%s' in target creation request\n",
				p);
2582 2583 2584 2585 2586 2587 2588 2589 2590 2591
			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))
2592 2593
				pr_warn("target creation request is missing parameter '%s'\n",
					srp_opt_tokens[i].pattern);
2594

2595 2596 2597 2598 2599 2600
	if (target->scsi_host->cmd_per_lun > target->scsi_host->can_queue
	    && (opt_mask & SRP_OPT_MAX_CMD_PER_LUN))
		pr_warn("cmd_per_lun = %d > queue_size = %d\n",
			target->scsi_host->cmd_per_lun,
			target->scsi_host->can_queue);

2601 2602 2603 2604 2605
out:
	kfree(options);
	return ret;
}

2606 2607
static ssize_t srp_create_target(struct device *dev,
				 struct device_attribute *attr,
2608 2609 2610
				 const char *buf, size_t count)
{
	struct srp_host *host =
2611
		container_of(dev, struct srp_host, dev);
2612 2613
	struct Scsi_Host *target_host;
	struct srp_target_port *target;
2614
	struct ib_device *ibdev = host->srp_dev->dev;
2615
	int ret;
2616 2617 2618 2619 2620 2621

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

2622
	target_host->transportt  = ib_srp_transport_template;
2623 2624
	target_host->max_channel = 0;
	target_host->max_id      = 1;
A
Arne Redlich 已提交
2625 2626
	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
2627

2628 2629
	target = host_to_target(target_host);

2630 2631 2632 2633 2634 2635
	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;
2636 2637
	target->sg_tablesize	= indirect_sg_entries ? : cmd_sg_entries;
	target->allow_ext_sg	= allow_ext_sg;
2638
	target->tl_retry_count	= 7;
2639
	target->queue_size	= SRP_DEFAULT_QUEUE_SIZE;
2640

2641 2642
	mutex_lock(&host->add_target_mutex);

2643 2644 2645 2646
	ret = srp_parse_options(buf, target);
	if (ret)
		goto err;

2647 2648
	target->req_ring_size = target->queue_size - SRP_TSK_MGMT_SQ_SIZE;

2649 2650 2651 2652 2653 2654 2655 2656 2657 2658
	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;
	}

2659 2660
	if (!host->srp_dev->fmr_pool && !target->allow_ext_sg &&
				target->cmd_sg_cnt < target->sg_tablesize) {
2661
		pr_warn("No FMR pool and no external indirect descriptors, limiting sg_tablesize to cmd_sg_cnt\n");
2662 2663 2664 2665 2666 2667
		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);
2668 2669 2670 2671
	target->max_iu_len = sizeof (struct srp_cmd) +
			     sizeof (struct srp_indirect_buf) +
			     target->cmd_sg_cnt * sizeof (struct srp_direct_buf);

2672
	INIT_WORK(&target->tl_err_work, srp_tl_err_work);
2673
	INIT_WORK(&target->remove_work, srp_remove_work);
2674 2675
	spin_lock_init(&target->lock);
	INIT_LIST_HEAD(&target->free_tx);
2676 2677 2678
	ret = srp_alloc_req_data(target);
	if (ret)
		goto err_free_mem;
2679

2680 2681 2682
	ret = ib_query_gid(ibdev, host->port, 0, &target->path.sgid);
	if (ret)
		goto err_free_mem;
2683 2684 2685

	ret = srp_create_target_ib(target);
	if (ret)
2686
		goto err_free_mem;
2687

D
David Dillow 已提交
2688 2689
	ret = srp_new_cm_id(target);
	if (ret)
2690
		goto err_free_ib;
2691 2692 2693

	ret = srp_connect_target(target);
	if (ret) {
2694 2695
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Connection failed\n");
2696 2697 2698 2699 2700 2701 2702
		goto err_cm_id;
	}

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

B
Bart Van Assche 已提交
2703 2704 2705 2706 2707 2708 2709 2710
	shost_printk(KERN_DEBUG, target->scsi_host, PFX
		     "new target: id_ext %016llx ioc_guid %016llx pkey %04x service_id %016llx sgid %pI6 dgid %pI6\n",
		     be64_to_cpu(target->id_ext),
		     be64_to_cpu(target->ioc_guid),
		     be16_to_cpu(target->path.pkey),
		     be64_to_cpu(target->service_id),
		     target->path.sgid.raw, target->path.dgid.raw);

2711 2712 2713 2714 2715
	ret = count;

out:
	mutex_unlock(&host->add_target_mutex);
	return ret;
2716 2717 2718 2719 2720 2721 2722

err_disconnect:
	srp_disconnect_target(target);

err_cm_id:
	ib_destroy_cm_id(target->cm_id);

2723
err_free_ib:
2724 2725
	srp_free_target_ib(target);

2726 2727 2728
err_free_mem:
	srp_free_req_data(target);

2729 2730
err:
	scsi_host_put(target_host);
2731
	goto out;
2732 2733
}

2734
static DEVICE_ATTR(add_target, S_IWUSR, NULL, srp_create_target);
2735

2736 2737
static ssize_t show_ibdev(struct device *dev, struct device_attribute *attr,
			  char *buf)
2738
{
2739
	struct srp_host *host = container_of(dev, struct srp_host, dev);
2740

2741
	return sprintf(buf, "%s\n", host->srp_dev->dev->name);
2742 2743
}

2744
static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
2745

2746 2747
static ssize_t show_port(struct device *dev, struct device_attribute *attr,
			 char *buf)
2748
{
2749
	struct srp_host *host = container_of(dev, struct srp_host, dev);
2750 2751 2752 2753

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

2754
static DEVICE_ATTR(port, S_IRUGO, show_port, NULL);
2755

2756
static struct srp_host *srp_add_port(struct srp_device *device, u8 port)
2757 2758 2759 2760 2761 2762 2763 2764
{
	struct srp_host *host;

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

	INIT_LIST_HEAD(&host->target_list);
2765
	spin_lock_init(&host->target_lock);
2766
	init_completion(&host->released);
2767
	mutex_init(&host->add_target_mutex);
2768
	host->srp_dev = device;
2769 2770
	host->port = port;

2771 2772
	host->dev.class = &srp_class;
	host->dev.parent = device->dev->dma_device;
2773
	dev_set_name(&host->dev, "srp-%s-%d", device->dev->name, port);
2774

2775
	if (device_register(&host->dev))
2776
		goto free_host;
2777
	if (device_create_file(&host->dev, &dev_attr_add_target))
2778
		goto err_class;
2779
	if (device_create_file(&host->dev, &dev_attr_ibdev))
2780
		goto err_class;
2781
	if (device_create_file(&host->dev, &dev_attr_port))
2782 2783 2784 2785 2786
		goto err_class;

	return host;

err_class:
2787
	device_unregister(&host->dev);
2788

2789
free_host:
2790 2791 2792 2793 2794 2795 2796
	kfree(host);

	return NULL;
}

static void srp_add_one(struct ib_device *device)
{
2797 2798 2799
	struct srp_device *srp_dev;
	struct ib_device_attr *dev_attr;
	struct ib_fmr_pool_param fmr_param;
2800
	struct srp_host *host;
2801
	int max_pages_per_fmr, fmr_page_shift, s, e, p;
2802

2803 2804
	dev_attr = kmalloc(sizeof *dev_attr, GFP_KERNEL);
	if (!dev_attr)
2805
		return;
2806

2807
	if (ib_query_device(device, dev_attr)) {
2808
		pr_warn("Query device failed for %s\n", device->name);
2809 2810 2811 2812 2813 2814 2815 2816 2817
		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
2818 2819
	 * minimum of 4096 bytes. We're unlikely to build large sglists
	 * out of smaller entries.
2820
	 */
2821 2822 2823 2824
	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;
2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839

	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;

2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857
	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;
	}

2858 2859
	if (IS_ERR(srp_dev->fmr_pool))
		srp_dev->fmr_pool = NULL;
2860

T
Tom Tucker 已提交
2861
	if (device->node_type == RDMA_NODE_IB_SWITCH) {
2862 2863 2864 2865 2866 2867 2868 2869
		s = 0;
		e = 0;
	} else {
		s = 1;
		e = device->phys_port_cnt;
	}

	for (p = s; p <= e; ++p) {
2870
		host = srp_add_port(srp_dev, p);
2871
		if (host)
2872
			list_add_tail(&host->list, &srp_dev->dev_list);
2873 2874
	}

2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886
	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);
2887 2888 2889 2890
}

static void srp_remove_one(struct ib_device *device)
{
2891
	struct srp_device *srp_dev;
2892
	struct srp_host *host, *tmp_host;
2893
	struct srp_target_port *target;
2894

2895
	srp_dev = ib_get_client_data(device, &srp_client);
2896 2897
	if (!srp_dev)
		return;
2898

2899
	list_for_each_entry_safe(host, tmp_host, &srp_dev->dev_list, list) {
2900
		device_unregister(&host->dev);
2901 2902 2903 2904 2905 2906 2907
		/*
		 * Wait for the sysfs entry to go away, so that no new
		 * target ports can be created.
		 */
		wait_for_completion(&host->released);

		/*
2908
		 * Remove all target ports.
2909
		 */
2910
		spin_lock(&host->target_lock);
2911 2912
		list_for_each_entry(target, &host->target_list, list)
			srp_queue_remove_work(target);
2913
		spin_unlock(&host->target_lock);
2914 2915

		/*
2916
		 * Wait for target port removal tasks.
2917
		 */
2918
		flush_workqueue(system_long_wq);
2919 2920 2921 2922

		kfree(host);
	}

2923 2924 2925 2926 2927 2928
	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);
2929 2930
}

2931
static struct srp_function_template ib_srp_transport_functions = {
2932 2933
	.has_rport_state	 = true,
	.reset_timer_if_blocked	 = true,
2934
	.reconnect_delay	 = &srp_reconnect_delay,
2935 2936 2937
	.fast_io_fail_tmo	 = &srp_fast_io_fail_tmo,
	.dev_loss_tmo		 = &srp_dev_loss_tmo,
	.reconnect		 = srp_rport_reconnect,
2938
	.rport_delete		 = srp_rport_delete,
2939
	.terminate_rport_io	 = srp_terminate_io,
2940 2941
};

2942 2943 2944 2945
static int __init srp_init_module(void)
{
	int ret;

2946
	BUILD_BUG_ON(FIELD_SIZEOF(struct ib_wc, wr_id) < sizeof(void *));
2947

2948
	if (srp_sg_tablesize) {
2949
		pr_warn("srp_sg_tablesize is deprecated, please use cmd_sg_entries\n");
2950 2951 2952 2953 2954 2955 2956 2957
		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) {
2958
		pr_warn("Clamping cmd_sg_entries to 255\n");
2959
		cmd_sg_entries = 255;
2960 2961
	}

2962 2963 2964
	if (!indirect_sg_entries)
		indirect_sg_entries = cmd_sg_entries;
	else if (indirect_sg_entries < cmd_sg_entries) {
2965 2966
		pr_warn("Bumping up indirect_sg_entries to match cmd_sg_entries (%u)\n",
			cmd_sg_entries);
2967 2968 2969
		indirect_sg_entries = cmd_sg_entries;
	}

2970 2971 2972 2973 2974
	ib_srp_transport_template =
		srp_attach_transport(&ib_srp_transport_functions);
	if (!ib_srp_transport_template)
		return -ENOMEM;

2975 2976
	ret = class_register(&srp_class);
	if (ret) {
2977
		pr_err("couldn't register class infiniband_srp\n");
2978
		srp_release_transport(ib_srp_transport_template);
2979 2980 2981
		return ret;
	}

2982 2983
	ib_sa_register_client(&srp_sa_client);

2984 2985
	ret = ib_register_client(&srp_client);
	if (ret) {
2986
		pr_err("couldn't register IB client\n");
2987
		srp_release_transport(ib_srp_transport_template);
2988
		ib_sa_unregister_client(&srp_sa_client);
2989 2990 2991 2992 2993 2994 2995 2996 2997 2998
		class_unregister(&srp_class);
		return ret;
	}

	return 0;
}

static void __exit srp_cleanup_module(void)
{
	ib_unregister_client(&srp_client);
2999
	ib_sa_unregister_client(&srp_sa_client);
3000
	class_unregister(&srp_class);
3001
	srp_release_transport(ib_srp_transport_template);
3002 3003 3004 3005
}

module_init(srp_init_module);
module_exit(srp_cleanup_module);