ib_srp.c 93.7 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
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
34

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
#include <rdma/ib_cache.h>
44

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

#include <scsi/scsi.h>
#include <scsi/scsi_device.h>
#include <scsi/scsi_dbg.h>
50
#include <scsi/scsi_tcq.h>
51
#include <scsi/srp.h>
52
#include <scsi/scsi_transport_srp.h>
53 54 55 56 57

#include "ib_srp.h"

#define DRV_NAME	"ib_srp"
#define PFX		DRV_NAME ": "
58 59
#define DRV_VERSION	"1.0"
#define DRV_RELDATE	"July 1, 2013"
60 61 62 63 64 65

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

66 67
static unsigned int srp_sg_tablesize;
static unsigned int cmd_sg_entries;
68 69
static unsigned int indirect_sg_entries;
static bool allow_ext_sg;
70
static bool prefer_fr;
71
static bool register_always;
72
static int topspin_workarounds = 1;
73

74 75
module_param(srp_sg_tablesize, uint, 0444);
MODULE_PARM_DESC(srp_sg_tablesize, "Deprecated name for cmd_sg_entries");
76

77 78 79
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)");
80

81 82 83 84 85 86 87 88
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)");

89 90 91 92
module_param(topspin_workarounds, int, 0444);
MODULE_PARM_DESC(topspin_workarounds,
		 "Enable workarounds for Topspin/Cisco SRP target bugs if != 0");

93 94 95 96
module_param(prefer_fr, bool, 0444);
MODULE_PARM_DESC(prefer_fr,
"Whether to use fast registration if both FMR and fast registration are supported");

97 98 99 100
module_param(register_always, bool, 0444);
MODULE_PARM_DESC(register_always,
		 "Use memory registration even for contiguous memory regions");

101 102
static struct kernel_param_ops srp_tmo_ops;

103 104 105 106 107
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");

108 109 110 111 112 113 114 115
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.");

116
static int srp_dev_loss_tmo = 600;
117 118 119 120 121 122 123 124 125 126
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.");

B
Bart Van Assche 已提交
127 128 129 130 131
static unsigned ch_count;
module_param(ch_count, uint, 0444);
MODULE_PARM_DESC(ch_count,
		 "Number of RDMA channels to use for communication with an SRP target. Using more than one channel improves performance if the HCA supports multiple completion vectors. The default value is the minimum of four times the number of online CPU sockets and the number of completion vectors supported by the HCA.");

132 133
static void srp_add_one(struct ib_device *device);
static void srp_remove_one(struct ib_device *device);
134 135
static void srp_recv_completion(struct ib_cq *cq, void *ch_ptr);
static void srp_send_completion(struct ib_cq *cq, void *ch_ptr);
136 137
static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event);

138
static struct scsi_transport_template *ib_srp_transport_template;
139
static struct workqueue_struct *srp_remove_wq;
140

141 142 143 144 145 146
static struct ib_client srp_client = {
	.name   = "srp",
	.add    = srp_add_one,
	.remove = srp_remove_one
};

147 148
static struct ib_sa_client srp_sa_client;

149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169
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;
	}
170 171 172 173 174
	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);
175
	else
176 177
		res = srp_tmo_valid(srp_reconnect_delay, srp_fast_io_fail_tmo,
				    tmo);
178 179 180 181 182 183 184 185 186 187 188 189 190
	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,
};

191 192 193 194 195 196 197 198 199 200
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;
}

201 202 203
static int srp_target_is_topspin(struct srp_target_port *target)
{
	static const u8 topspin_oui[3] = { 0x00, 0x05, 0xad };
204
	static const u8 cisco_oui[3]   = { 0x00, 0x1b, 0x0d };
205 206

	return topspin_workarounds &&
207 208
		(!memcmp(&target->ioc_guid, topspin_oui, sizeof topspin_oui) ||
		 !memcmp(&target->ioc_guid, cisco_oui, sizeof cisco_oui));
209 210
}

211 212 213 214 215 216 217 218 219 220 221 222 223 224
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;

225 226 227
	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))
228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247
		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;

248 249
	ib_dma_unmap_single(host->srp_dev->dev, iu->dma, iu->size,
			    iu->direction);
250 251 252 253 254 255
	kfree(iu->buf);
	kfree(iu);
}

static void srp_qp_event(struct ib_event *event, void *context)
{
256
	pr_debug("QP event %d\n", event->event);
257 258 259 260 261 262 263 264 265 266 267 268
}

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;

269 270 271 272
	ret = ib_find_cached_pkey(target->srp_host->srp_dev->dev,
				  target->srp_host->port,
				  be16_to_cpu(target->pkey),
				  &attr->pkey_index);
273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291
	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;
}

292
static int srp_new_cm_id(struct srp_rdma_ch *ch)
D
David Dillow 已提交
293
{
294
	struct srp_target_port *target = ch->target;
D
David Dillow 已提交
295 296
	struct ib_cm_id *new_cm_id;

297
	new_cm_id = ib_create_cm_id(target->srp_host->srp_dev->dev,
298
				    srp_cm_handler, ch);
D
David Dillow 已提交
299 300 301
	if (IS_ERR(new_cm_id))
		return PTR_ERR(new_cm_id);

302 303 304 305 306 307 308
	if (ch->cm_id)
		ib_destroy_cm_id(ch->cm_id);
	ch->cm_id = new_cm_id;
	ch->path.sgid = target->sgid;
	ch->path.dgid = target->orig_dgid;
	ch->path.pkey = target->pkey;
	ch->path.service_id = target->service_id;
D
David Dillow 已提交
309 310 311 312

	return 0;
}

313 314 315 316 317 318 319 320 321
static struct ib_fmr_pool *srp_alloc_fmr_pool(struct srp_target_port *target)
{
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_fmr_pool_param fmr_param;

	memset(&fmr_param, 0, sizeof(fmr_param));
	fmr_param.pool_size	    = target->scsi_host->can_queue;
	fmr_param.dirty_watermark   = fmr_param.pool_size / 4;
	fmr_param.cache		    = 1;
322 323
	fmr_param.max_pages_per_fmr = dev->max_pages_per_mr;
	fmr_param.page_shift	    = ilog2(dev->mr_page_size);
324 325 326 327 328 329 330
	fmr_param.access	    = (IB_ACCESS_LOCAL_WRITE |
				       IB_ACCESS_REMOTE_WRITE |
				       IB_ACCESS_REMOTE_READ);

	return ib_create_fmr_pool(dev->pd, &fmr_param);
}

331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456
/**
 * srp_destroy_fr_pool() - free the resources owned by a pool
 * @pool: Fast registration pool to be destroyed.
 */
static void srp_destroy_fr_pool(struct srp_fr_pool *pool)
{
	int i;
	struct srp_fr_desc *d;

	if (!pool)
		return;

	for (i = 0, d = &pool->desc[0]; i < pool->size; i++, d++) {
		if (d->frpl)
			ib_free_fast_reg_page_list(d->frpl);
		if (d->mr)
			ib_dereg_mr(d->mr);
	}
	kfree(pool);
}

/**
 * srp_create_fr_pool() - allocate and initialize a pool for fast registration
 * @device:            IB device to allocate fast registration descriptors for.
 * @pd:                Protection domain associated with the FR descriptors.
 * @pool_size:         Number of descriptors to allocate.
 * @max_page_list_len: Maximum fast registration work request page list length.
 */
static struct srp_fr_pool *srp_create_fr_pool(struct ib_device *device,
					      struct ib_pd *pd, int pool_size,
					      int max_page_list_len)
{
	struct srp_fr_pool *pool;
	struct srp_fr_desc *d;
	struct ib_mr *mr;
	struct ib_fast_reg_page_list *frpl;
	int i, ret = -EINVAL;

	if (pool_size <= 0)
		goto err;
	ret = -ENOMEM;
	pool = kzalloc(sizeof(struct srp_fr_pool) +
		       pool_size * sizeof(struct srp_fr_desc), GFP_KERNEL);
	if (!pool)
		goto err;
	pool->size = pool_size;
	pool->max_page_list_len = max_page_list_len;
	spin_lock_init(&pool->lock);
	INIT_LIST_HEAD(&pool->free_list);

	for (i = 0, d = &pool->desc[0]; i < pool->size; i++, d++) {
		mr = ib_alloc_fast_reg_mr(pd, max_page_list_len);
		if (IS_ERR(mr)) {
			ret = PTR_ERR(mr);
			goto destroy_pool;
		}
		d->mr = mr;
		frpl = ib_alloc_fast_reg_page_list(device, max_page_list_len);
		if (IS_ERR(frpl)) {
			ret = PTR_ERR(frpl);
			goto destroy_pool;
		}
		d->frpl = frpl;
		list_add_tail(&d->entry, &pool->free_list);
	}

out:
	return pool;

destroy_pool:
	srp_destroy_fr_pool(pool);

err:
	pool = ERR_PTR(ret);
	goto out;
}

/**
 * srp_fr_pool_get() - obtain a descriptor suitable for fast registration
 * @pool: Pool to obtain descriptor from.
 */
static struct srp_fr_desc *srp_fr_pool_get(struct srp_fr_pool *pool)
{
	struct srp_fr_desc *d = NULL;
	unsigned long flags;

	spin_lock_irqsave(&pool->lock, flags);
	if (!list_empty(&pool->free_list)) {
		d = list_first_entry(&pool->free_list, typeof(*d), entry);
		list_del(&d->entry);
	}
	spin_unlock_irqrestore(&pool->lock, flags);

	return d;
}

/**
 * srp_fr_pool_put() - put an FR descriptor back in the free list
 * @pool: Pool the descriptor was allocated from.
 * @desc: Pointer to an array of fast registration descriptor pointers.
 * @n:    Number of descriptors to put back.
 *
 * Note: The caller must already have queued an invalidation request for
 * desc->mr->rkey before calling this function.
 */
static void srp_fr_pool_put(struct srp_fr_pool *pool, struct srp_fr_desc **desc,
			    int n)
{
	unsigned long flags;
	int i;

	spin_lock_irqsave(&pool->lock, flags);
	for (i = 0; i < n; i++)
		list_add(&desc[i]->entry, &pool->free_list);
	spin_unlock_irqrestore(&pool->lock, flags);
}

static struct srp_fr_pool *srp_alloc_fr_pool(struct srp_target_port *target)
{
	struct srp_device *dev = target->srp_host->srp_dev;

	return srp_create_fr_pool(dev->dev, dev->pd,
				  target->scsi_host->can_queue,
				  dev->max_pages_per_mr);
}

457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473
/**
 * srp_destroy_qp() - destroy an RDMA queue pair
 * @ch: SRP RDMA channel.
 *
 * Change a queue pair into the error state and wait until all receive
 * completions have been processed before destroying it. This avoids that
 * the receive completion handler can access the queue pair while it is
 * being destroyed.
 */
static void srp_destroy_qp(struct srp_rdma_ch *ch)
{
	static struct ib_qp_attr attr = { .qp_state = IB_QPS_ERR };
	static struct ib_recv_wr wr = { .wr_id = SRP_LAST_WR_ID };
	struct ib_recv_wr *bad_wr;
	int ret;

	/* Destroying a QP and reusing ch->done is only safe if not connected */
474
	WARN_ON_ONCE(ch->connected);
475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490

	ret = ib_modify_qp(ch->qp, &attr, IB_QP_STATE);
	WARN_ONCE(ret, "ib_cm_init_qp_attr() returned %d\n", ret);
	if (ret)
		goto out;

	init_completion(&ch->done);
	ret = ib_post_recv(ch->qp, &wr, &bad_wr);
	WARN_ONCE(ret, "ib_post_recv() returned %d\n", ret);
	if (ret == 0)
		wait_for_completion(&ch->done);

out:
	ib_destroy_qp(ch->qp);
}

491
static int srp_create_ch_ib(struct srp_rdma_ch *ch)
492
{
493
	struct srp_target_port *target = ch->target;
494
	struct srp_device *dev = target->srp_host->srp_dev;
495
	struct ib_qp_init_attr *init_attr;
496 497
	struct ib_cq *recv_cq, *send_cq;
	struct ib_qp *qp;
498
	struct ib_fmr_pool *fmr_pool = NULL;
499 500
	struct srp_fr_pool *fr_pool = NULL;
	const int m = 1 + dev->use_fast_reg;
501 502 503 504 505 506
	int ret;

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

507
	/* + 1 for SRP_LAST_WR_ID */
508
	recv_cq = ib_create_cq(dev->dev, srp_recv_completion, NULL, ch,
509
			       target->queue_size + 1, ch->comp_vector);
510 511
	if (IS_ERR(recv_cq)) {
		ret = PTR_ERR(recv_cq);
512
		goto err;
513 514
	}

515 516
	send_cq = ib_create_cq(dev->dev, srp_send_completion, NULL, ch,
			       m * target->queue_size, ch->comp_vector);
517 518
	if (IS_ERR(send_cq)) {
		ret = PTR_ERR(send_cq);
519
		goto err_recv_cq;
520 521
	}

522
	ib_req_notify_cq(recv_cq, IB_CQ_NEXT_COMP);
523 524

	init_attr->event_handler       = srp_qp_event;
525
	init_attr->cap.max_send_wr     = m * target->queue_size;
526
	init_attr->cap.max_recv_wr     = target->queue_size + 1;
527 528
	init_attr->cap.max_recv_sge    = 1;
	init_attr->cap.max_send_sge    = 1;
529
	init_attr->sq_sig_type         = IB_SIGNAL_REQ_WR;
530
	init_attr->qp_type             = IB_QPT_RC;
531 532
	init_attr->send_cq             = send_cq;
	init_attr->recv_cq             = recv_cq;
533

534
	qp = ib_create_qp(dev->pd, init_attr);
535 536
	if (IS_ERR(qp)) {
		ret = PTR_ERR(qp);
537
		goto err_send_cq;
538 539
	}

540
	ret = srp_init_qp(target, qp);
541 542
	if (ret)
		goto err_qp;
543

544 545 546 547 548 549 550 551
	if (dev->use_fast_reg && dev->has_fr) {
		fr_pool = srp_alloc_fr_pool(target);
		if (IS_ERR(fr_pool)) {
			ret = PTR_ERR(fr_pool);
			shost_printk(KERN_WARNING, target->scsi_host, PFX
				     "FR pool allocation failed (%d)\n", ret);
			goto err_qp;
		}
552 553 554
		if (ch->fr_pool)
			srp_destroy_fr_pool(ch->fr_pool);
		ch->fr_pool = fr_pool;
555
	} else if (!dev->use_fast_reg && dev->has_fmr) {
556 557 558 559 560 561 562
		fmr_pool = srp_alloc_fmr_pool(target);
		if (IS_ERR(fmr_pool)) {
			ret = PTR_ERR(fmr_pool);
			shost_printk(KERN_WARNING, target->scsi_host, PFX
				     "FMR pool allocation failed (%d)\n", ret);
			goto err_qp;
		}
563 564 565
		if (ch->fmr_pool)
			ib_destroy_fmr_pool(ch->fmr_pool);
		ch->fmr_pool = fmr_pool;
566 567
	}

568
	if (ch->qp)
569
		srp_destroy_qp(ch);
570 571 572 573
	if (ch->recv_cq)
		ib_destroy_cq(ch->recv_cq);
	if (ch->send_cq)
		ib_destroy_cq(ch->send_cq);
574

575 576 577
	ch->qp = qp;
	ch->recv_cq = recv_cq;
	ch->send_cq = send_cq;
578

579 580 581 582
	kfree(init_attr);
	return 0;

err_qp:
583
	ib_destroy_qp(qp);
584 585

err_send_cq:
586
	ib_destroy_cq(send_cq);
587 588

err_recv_cq:
589
	ib_destroy_cq(recv_cq);
590 591

err:
592 593 594 595
	kfree(init_attr);
	return ret;
}

596 597
/*
 * Note: this function may be called without srp_alloc_iu_bufs() having been
598
 * invoked. Hence the ch->[rt]x_ring checks.
599
 */
600 601
static void srp_free_ch_ib(struct srp_target_port *target,
			   struct srp_rdma_ch *ch)
602
{
603
	struct srp_device *dev = target->srp_host->srp_dev;
604 605
	int i;

B
Bart Van Assche 已提交
606 607 608
	if (!ch->target)
		return;

609 610 611
	if (ch->cm_id) {
		ib_destroy_cm_id(ch->cm_id);
		ch->cm_id = NULL;
612 613
	}

B
Bart Van Assche 已提交
614 615 616 617
	/* If srp_new_cm_id() succeeded but srp_create_ch_ib() not, return. */
	if (!ch->qp)
		return;

618
	if (dev->use_fast_reg) {
619 620
		if (ch->fr_pool)
			srp_destroy_fr_pool(ch->fr_pool);
621
	} else {
622 623
		if (ch->fmr_pool)
			ib_destroy_fmr_pool(ch->fmr_pool);
624
	}
625
	srp_destroy_qp(ch);
626 627
	ib_destroy_cq(ch->send_cq);
	ib_destroy_cq(ch->recv_cq);
628

B
Bart Van Assche 已提交
629 630 631 632 633 634 635 636
	/*
	 * Avoid that the SCSI error handler tries to use this channel after
	 * it has been freed. The SCSI error handler can namely continue
	 * trying to perform recovery actions after scsi_remove_host()
	 * returned.
	 */
	ch->target = NULL;

637 638
	ch->qp = NULL;
	ch->send_cq = ch->recv_cq = NULL;
639

640
	if (ch->rx_ring) {
641
		for (i = 0; i < target->queue_size; ++i)
642 643 644
			srp_free_iu(target->srp_host, ch->rx_ring[i]);
		kfree(ch->rx_ring);
		ch->rx_ring = NULL;
645
	}
646
	if (ch->tx_ring) {
647
		for (i = 0; i < target->queue_size; ++i)
648 649 650
			srp_free_iu(target->srp_host, ch->tx_ring[i]);
		kfree(ch->tx_ring);
		ch->tx_ring = NULL;
651
	}
652 653 654 655
}

static void srp_path_rec_completion(int status,
				    struct ib_sa_path_rec *pathrec,
656
				    void *ch_ptr)
657
{
658 659
	struct srp_rdma_ch *ch = ch_ptr;
	struct srp_target_port *target = ch->target;
660

661
	ch->status = status;
662
	if (status)
663 664
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Got failed path rec status %d\n", status);
665
	else
666 667
		ch->path = *pathrec;
	complete(&ch->done);
668 669
}

670
static int srp_lookup_path(struct srp_rdma_ch *ch)
671
{
672
	struct srp_target_port *target = ch->target;
673 674
	int ret;

675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695
	ch->path.numb_path = 1;

	init_completion(&ch->done);

	ch->path_query_id = ib_sa_path_rec_get(&srp_sa_client,
					       target->srp_host->srp_dev->dev,
					       target->srp_host->port,
					       &ch->path,
					       IB_SA_PATH_REC_SERVICE_ID |
					       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,
					       ch, &ch->path_query);
	if (ch->path_query_id < 0)
		return ch->path_query_id;

	ret = wait_for_completion_interruptible(&ch->done);
696 697
	if (ret < 0)
		return ret;
698

699
	if (ch->status < 0)
700 701
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Path record query failed\n");
702

703
	return ch->status;
704 705
}

B
Bart Van Assche 已提交
706
static int srp_send_req(struct srp_rdma_ch *ch, bool multich)
707
{
708
	struct srp_target_port *target = ch->target;
709 710 711 712 713 714 715 716 717 718
	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;

719
	req->param.primary_path		      = &ch->path;
720 721
	req->param.alternate_path 	      = NULL;
	req->param.service_id 		      = target->service_id;
722 723
	req->param.qp_num		      = ch->qp->qp_num;
	req->param.qp_type		      = ch->qp->qp_type;
724 725 726 727 728 729 730 731 732 733 734 735 736 737
	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;
738
	req->param.retry_count                = target->tl_retry_count;
739 740 741 742 743
	req->param.rnr_retry_count 	      = 7;
	req->param.max_cm_retries 	      = 15;

	req->priv.opcode     	= SRP_LOGIN_REQ;
	req->priv.tag        	= 0;
744
	req->priv.req_it_iu_len = cpu_to_be32(target->max_iu_len);
745 746
	req->priv.req_buf_fmt 	= cpu_to_be16(SRP_BUF_FORMAT_DIRECT |
					      SRP_BUF_FORMAT_INDIRECT);
B
Bart Van Assche 已提交
747 748
	req->priv.req_flags	= (multich ? SRP_MULTICHAN_MULTI :
				   SRP_MULTICHAN_SINGLE);
749
	/*
R
Roland Dreier 已提交
750
	 * In the published SRP specification (draft rev. 16a), the
751 752 753 754 755 756 757 758 759
	 * 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,
760
		       &target->sgid.global.interface_id, 8);
761
		memcpy(req->priv.initiator_port_id + 8,
762
		       &target->initiator_ext, 8);
763 764 765 766
		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,
767 768
		       &target->initiator_ext, 8);
		memcpy(req->priv.initiator_port_id + 8,
769
		       &target->sgid.global.interface_id, 8);
770 771 772 773
		memcpy(req->priv.target_port_id,     &target->id_ext, 8);
		memcpy(req->priv.target_port_id + 8, &target->ioc_guid, 8);
	}

774 775
	/*
	 * Topspin/Cisco SRP targets will reject our login unless we
776 777
	 * zero out the first 8 bytes of our initiator port ID and set
	 * the second 8 bytes to the local node GUID.
778
	 */
779
	if (srp_target_is_topspin(target)) {
780 781 782 783
		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));
784
		memset(req->priv.initiator_port_id, 0, 8);
785
		memcpy(req->priv.initiator_port_id + 8,
786
		       &target->srp_host->srp_dev->dev->node_guid, 8);
787 788
	}

789
	status = ib_send_cm_req(ch->cm_id, &req->param);
790 791 792 793 794 795

	kfree(req);

	return status;
}

796 797 798 799 800 801 802 803 804 805 806 807
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)
808
		queue_work(srp_remove_wq, &target->remove_work);
809 810 811 812

	return changed;
}

813 814
static void srp_disconnect_target(struct srp_target_port *target)
{
B
Bart Van Assche 已提交
815 816
	struct srp_rdma_ch *ch;
	int i;
817

818
	/* XXX should send SRP_I_LOGOUT request */
819

820 821 822 823 824 825
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		ch->connected = false;
		if (ch->cm_id && ib_send_cm_dreq(ch->cm_id, NULL, 0)) {
			shost_printk(KERN_DEBUG, target->scsi_host,
				     PFX "Sending CM DREQ failed\n");
826
		}
827
	}
828 829
}

830 831
static void srp_free_req_data(struct srp_target_port *target,
			      struct srp_rdma_ch *ch)
832
{
833 834
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = dev->dev;
835 836 837
	struct srp_request *req;
	int i;

B
Bart Van Assche 已提交
838
	if (!ch->target || !ch->req_ring)
839 840 841
		return;

	for (i = 0; i < target->req_ring_size; ++i) {
842
		req = &ch->req_ring[i];
843 844 845 846
		if (dev->use_fast_reg)
			kfree(req->fr_list);
		else
			kfree(req->fmr_list);
847
		kfree(req->map_page);
848 849 850 851 852 853
		if (req->indirect_dma_addr) {
			ib_dma_unmap_single(ibdev, req->indirect_dma_addr,
					    target->indirect_size,
					    DMA_TO_DEVICE);
		}
		kfree(req->indirect_desc);
854
	}
855

856 857
	kfree(ch->req_ring);
	ch->req_ring = NULL;
858 859
}

860
static int srp_alloc_req_data(struct srp_rdma_ch *ch)
861
{
862
	struct srp_target_port *target = ch->target;
863 864 865
	struct srp_device *srp_dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = srp_dev->dev;
	struct srp_request *req;
866
	void *mr_list;
867 868 869
	dma_addr_t dma_addr;
	int i, ret = -ENOMEM;

870 871 872
	ch->req_ring = kcalloc(target->req_ring_size, sizeof(*ch->req_ring),
			       GFP_KERNEL);
	if (!ch->req_ring)
873 874 875
		goto out;

	for (i = 0; i < target->req_ring_size; ++i) {
876
		req = &ch->req_ring[i];
877 878 879 880 881 882 883 884
		mr_list = kmalloc(target->cmd_sg_cnt * sizeof(void *),
				  GFP_KERNEL);
		if (!mr_list)
			goto out;
		if (srp_dev->use_fast_reg)
			req->fr_list = mr_list;
		else
			req->fmr_list = mr_list;
885
		req->map_page = kmalloc(srp_dev->max_pages_per_mr *
886
					sizeof(void *), GFP_KERNEL);
887 888
		if (!req->map_page)
			goto out;
889
		req->indirect_desc = kmalloc(target->indirect_size, GFP_KERNEL);
890
		if (!req->indirect_desc)
891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906
			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;
	}
	ret = 0;

out:
	return ret;
}

907 908 909 910 911 912 913 914 915 916 917 918 919 920 921
/**
 * 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);
}

922 923
static void srp_remove_target(struct srp_target_port *target)
{
B
Bart Van Assche 已提交
924 925
	struct srp_rdma_ch *ch;
	int i;
926

927 928
	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);

929
	srp_del_scsi_host_attr(target->scsi_host);
930
	srp_rport_get(target->rport);
931 932
	srp_remove_host(target->scsi_host);
	scsi_remove_host(target->scsi_host);
933
	srp_stop_rport_timers(target->rport);
934
	srp_disconnect_target(target);
B
Bart Van Assche 已提交
935 936 937 938
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		srp_free_ch_ib(target, ch);
	}
939
	cancel_work_sync(&target->tl_err_work);
940
	srp_rport_put(target->rport);
B
Bart Van Assche 已提交
941 942 943 944 945 946
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		srp_free_req_data(target, ch);
	}
	kfree(target->ch);
	target->ch = NULL;
947 948 949 950 951

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

952 953 954
	scsi_host_put(target->scsi_host);
}

D
David Howells 已提交
955
static void srp_remove_work(struct work_struct *work)
956
{
D
David Howells 已提交
957
	struct srp_target_port *target =
958
		container_of(work, struct srp_target_port, remove_work);
959

960
	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);
961

962
	srp_remove_target(target);
963 964
}

965 966 967 968 969 970 971
static void srp_rport_delete(struct srp_rport *rport)
{
	struct srp_target_port *target = rport->lld_data;

	srp_queue_remove_work(target);
}

972 973 974 975 976 977 978 979 980 981 982 983 984 985
/**
 * srp_connected_ch() - number of connected channels
 * @target: SRP target port.
 */
static int srp_connected_ch(struct srp_target_port *target)
{
	int i, c = 0;

	for (i = 0; i < target->ch_count; i++)
		c += target->ch[i].connected;

	return c;
}

B
Bart Van Assche 已提交
986
static int srp_connect_ch(struct srp_rdma_ch *ch, bool multich)
987
{
988
	struct srp_target_port *target = ch->target;
989 990
	int ret;

991
	WARN_ON_ONCE(!multich && srp_connected_ch(target) > 0);
992

993
	ret = srp_lookup_path(ch);
994 995 996 997
	if (ret)
		return ret;

	while (1) {
998
		init_completion(&ch->done);
B
Bart Van Assche 已提交
999
		ret = srp_send_req(ch, multich);
1000 1001
		if (ret)
			return ret;
1002
		ret = wait_for_completion_interruptible(&ch->done);
1003 1004
		if (ret < 0)
			return ret;
1005 1006 1007 1008 1009 1010 1011

		/*
		 * 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.
		 */
1012
		switch (ch->status) {
1013
		case 0:
1014
			ch->connected = true;
1015 1016 1017
			return 0;

		case SRP_PORT_REDIRECT:
1018
			ret = srp_lookup_path(ch);
1019 1020 1021 1022 1023 1024 1025
			if (ret)
				return ret;
			break;

		case SRP_DLID_REDIRECT:
			break;

D
David Dillow 已提交
1026 1027
		case SRP_STALE_CONN:
			shost_printk(KERN_ERR, target->scsi_host, PFX
1028
				     "giving up on stale connection\n");
1029 1030
			ch->status = -ECONNRESET;
			return ch->status;
D
David Dillow 已提交
1031

1032
		default:
1033
			return ch->status;
1034 1035 1036 1037
		}
	}
}

1038
static int srp_inv_rkey(struct srp_rdma_ch *ch, u32 rkey)
1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049
{
	struct ib_send_wr *bad_wr;
	struct ib_send_wr wr = {
		.opcode		    = IB_WR_LOCAL_INV,
		.wr_id		    = LOCAL_INV_WR_ID_MASK,
		.next		    = NULL,
		.num_sge	    = 0,
		.send_flags	    = 0,
		.ex.invalidate_rkey = rkey,
	};

1050
	return ib_post_send(ch->qp, &wr, &bad_wr);
1051 1052
}

1053
static void srp_unmap_data(struct scsi_cmnd *scmnd,
1054
			   struct srp_rdma_ch *ch,
1055 1056
			   struct srp_request *req)
{
1057
	struct srp_target_port *target = ch->target;
1058 1059 1060
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = dev->dev;
	int i, res;
1061

1062
	if (!scsi_sglist(scmnd) ||
1063 1064 1065 1066
	    (scmnd->sc_data_direction != DMA_TO_DEVICE &&
	     scmnd->sc_data_direction != DMA_FROM_DEVICE))
		return;

1067 1068 1069 1070
	if (dev->use_fast_reg) {
		struct srp_fr_desc **pfr;

		for (i = req->nmdesc, pfr = req->fr_list; i > 0; i--, pfr++) {
1071
			res = srp_inv_rkey(ch, (*pfr)->mr->rkey);
1072 1073 1074 1075 1076 1077 1078 1079 1080
			if (res < 0) {
				shost_printk(KERN_ERR, target->scsi_host, PFX
				  "Queueing INV WR for rkey %#x failed (%d)\n",
				  (*pfr)->mr->rkey, res);
				queue_work(system_long_wq,
					   &target->tl_err_work);
			}
		}
		if (req->nmdesc)
1081
			srp_fr_pool_put(ch->fr_pool, req->fr_list,
1082 1083 1084 1085 1086 1087 1088
					req->nmdesc);
	} else {
		struct ib_pool_fmr **pfmr;

		for (i = req->nmdesc, pfmr = req->fmr_list; i > 0; i--, pfmr++)
			ib_fmr_pool_unmap(*pfmr);
	}
1089

1090 1091
	ib_dma_unmap_sg(ibdev, scsi_sglist(scmnd), scsi_sg_count(scmnd),
			scmnd->sc_data_direction);
1092 1093
}

B
Bart Van Assche 已提交
1094 1095
/**
 * srp_claim_req - Take ownership of the scmnd associated with a request.
1096
 * @ch: SRP RDMA channel.
B
Bart Van Assche 已提交
1097
 * @req: SRP request.
1098
 * @sdev: If not NULL, only take ownership for this SCSI device.
B
Bart Van Assche 已提交
1099 1100 1101 1102 1103 1104
 * @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.
 */
1105
static struct scsi_cmnd *srp_claim_req(struct srp_rdma_ch *ch,
B
Bart Van Assche 已提交
1106
				       struct srp_request *req,
1107
				       struct scsi_device *sdev,
B
Bart Van Assche 已提交
1108 1109 1110 1111
				       struct scsi_cmnd *scmnd)
{
	unsigned long flags;

1112
	spin_lock_irqsave(&ch->lock, flags);
1113 1114 1115
	if (req->scmnd &&
	    (!sdev || req->scmnd->device == sdev) &&
	    (!scmnd || req->scmnd == scmnd)) {
B
Bart Van Assche 已提交
1116 1117 1118 1119 1120
		scmnd = req->scmnd;
		req->scmnd = NULL;
	} else {
		scmnd = NULL;
	}
1121
	spin_unlock_irqrestore(&ch->lock, flags);
B
Bart Van Assche 已提交
1122 1123 1124 1125 1126 1127

	return scmnd;
}

/**
 * srp_free_req() - Unmap data and add request to the free request list.
1128
 * @ch:     SRP RDMA channel.
1129 1130 1131
 * @req:    Request to be freed.
 * @scmnd:  SCSI command associated with @req.
 * @req_lim_delta: Amount to be added to @target->req_lim.
B
Bart Van Assche 已提交
1132
 */
1133 1134
static void srp_free_req(struct srp_rdma_ch *ch, struct srp_request *req,
			 struct scsi_cmnd *scmnd, s32 req_lim_delta)
1135
{
1136 1137
	unsigned long flags;

1138
	srp_unmap_data(scmnd, ch, req);
B
Bart Van Assche 已提交
1139

1140 1141 1142
	spin_lock_irqsave(&ch->lock, flags);
	ch->req_lim += req_lim_delta;
	spin_unlock_irqrestore(&ch->lock, flags);
1143 1144
}

1145 1146
static void srp_finish_req(struct srp_rdma_ch *ch, struct srp_request *req,
			   struct scsi_device *sdev, int result)
1147
{
1148
	struct scsi_cmnd *scmnd = srp_claim_req(ch, req, sdev, NULL);
B
Bart Van Assche 已提交
1149 1150

	if (scmnd) {
1151
		srp_free_req(ch, req, scmnd, 0);
1152
		scmnd->result = result;
B
Bart Van Assche 已提交
1153 1154
		scmnd->scsi_done(scmnd);
	}
1155 1156
}

1157
static void srp_terminate_io(struct srp_rport *rport)
1158
{
1159
	struct srp_target_port *target = rport->lld_data;
B
Bart Van Assche 已提交
1160
	struct srp_rdma_ch *ch;
1161 1162
	struct Scsi_Host *shost = target->scsi_host;
	struct scsi_device *sdev;
B
Bart Van Assche 已提交
1163
	int i, j;
1164

1165 1166 1167 1168 1169 1170 1171
	/*
	 * 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);

B
Bart Van Assche 已提交
1172 1173
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
1174

B
Bart Van Assche 已提交
1175 1176 1177 1178 1179 1180
		for (j = 0; j < target->req_ring_size; ++j) {
			struct srp_request *req = &ch->req_ring[j];

			srp_finish_req(ch, req, NULL,
				       DID_TRANSPORT_FAILFAST << 16);
		}
1181 1182
	}
}
1183

1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195
/*
 * 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;
B
Bart Van Assche 已提交
1196 1197 1198
	struct srp_rdma_ch *ch;
	int i, j, ret = 0;
	bool multich = false;
1199

1200
	srp_disconnect_target(target);
1201 1202 1203 1204

	if (target->state == SRP_TARGET_SCANNING)
		return -ENODEV;

1205
	/*
1206 1207 1208
	 * 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.
1209
	 */
B
Bart Van Assche 已提交
1210 1211 1212 1213 1214
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		if (!ch->target)
			break;
		ret += srp_new_cm_id(ch);
1215
	}
B
Bart Van Assche 已提交
1216 1217 1218 1219 1220 1221
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		if (!ch->target)
			break;
		for (j = 0; j < target->req_ring_size; ++j) {
			struct srp_request *req = &ch->req_ring[j];
1222

B
Bart Van Assche 已提交
1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235
			srp_finish_req(ch, req, NULL, DID_RESET << 16);
		}
	}
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		if (!ch->target)
			break;
		/*
		 * 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.
		 */
		ret += srp_create_ch_ib(ch);
1236

B
Bart Van Assche 已提交
1237 1238 1239 1240
		INIT_LIST_HEAD(&ch->free_tx);
		for (j = 0; j < target->queue_size; ++j)
			list_add(&ch->tx_ring[j]->list, &ch->free_tx);
	}
1241 1242 1243

	target->qp_in_error = false;

B
Bart Van Assche 已提交
1244 1245 1246 1247 1248 1249 1250 1251 1252 1253
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		if (ret || !ch->target) {
			if (i > 1)
				ret = 0;
			break;
		}
		ret = srp_connect_ch(ch, multich);
		multich = true;
	}
1254

1255 1256 1257
	if (ret == 0)
		shost_printk(KERN_INFO, target->scsi_host,
			     PFX "reconnect succeeded\n");
1258 1259 1260 1261

	return ret;
}

1262 1263
static void srp_map_desc(struct srp_map_state *state, dma_addr_t dma_addr,
			 unsigned int dma_len, u32 rkey)
1264
{
1265
	struct srp_direct_buf *desc = state->desc;
1266

1267 1268 1269
	desc->va = cpu_to_be64(dma_addr);
	desc->key = cpu_to_be32(rkey);
	desc->len = cpu_to_be32(dma_len);
1270

1271 1272 1273 1274
	state->total_len += dma_len;
	state->desc++;
	state->ndesc++;
}
1275

1276
static int srp_map_finish_fmr(struct srp_map_state *state,
1277
			      struct srp_rdma_ch *ch)
1278 1279 1280
{
	struct ib_pool_fmr *fmr;
	u64 io_addr = 0;
1281

1282
	fmr = ib_fmr_pool_map_phys(ch->fmr_pool, state->pages,
1283 1284 1285
				   state->npages, io_addr);
	if (IS_ERR(fmr))
		return PTR_ERR(fmr);
1286

1287
	*state->next_fmr++ = fmr;
1288
	state->nmdesc++;
1289

1290
	srp_map_desc(state, 0, state->dma_len, fmr->fmr->rkey);
1291

1292 1293 1294
	return 0;
}

1295
static int srp_map_finish_fr(struct srp_map_state *state,
1296
			     struct srp_rdma_ch *ch)
1297
{
1298
	struct srp_target_port *target = ch->target;
1299 1300 1301 1302 1303 1304
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_send_wr *bad_wr;
	struct ib_send_wr wr;
	struct srp_fr_desc *desc;
	u32 rkey;

1305
	desc = srp_fr_pool_get(ch->fr_pool);
1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333
	if (!desc)
		return -ENOMEM;

	rkey = ib_inc_rkey(desc->mr->rkey);
	ib_update_fast_reg_key(desc->mr, rkey);

	memcpy(desc->frpl->page_list, state->pages,
	       sizeof(state->pages[0]) * state->npages);

	memset(&wr, 0, sizeof(wr));
	wr.opcode = IB_WR_FAST_REG_MR;
	wr.wr_id = FAST_REG_WR_ID_MASK;
	wr.wr.fast_reg.iova_start = state->base_dma_addr;
	wr.wr.fast_reg.page_list = desc->frpl;
	wr.wr.fast_reg.page_list_len = state->npages;
	wr.wr.fast_reg.page_shift = ilog2(dev->mr_page_size);
	wr.wr.fast_reg.length = state->dma_len;
	wr.wr.fast_reg.access_flags = (IB_ACCESS_LOCAL_WRITE |
				       IB_ACCESS_REMOTE_READ |
				       IB_ACCESS_REMOTE_WRITE);
	wr.wr.fast_reg.rkey = desc->mr->lkey;

	*state->next_fr++ = desc;
	state->nmdesc++;

	srp_map_desc(state, state->base_dma_addr, state->dma_len,
		     desc->mr->rkey);

1334
	return ib_post_send(ch->qp, &wr, &bad_wr);
1335 1336
}

1337
static int srp_finish_mapping(struct srp_map_state *state,
1338
			      struct srp_rdma_ch *ch)
1339
{
1340
	struct srp_target_port *target = ch->target;
1341 1342 1343 1344 1345
	int ret = 0;

	if (state->npages == 0)
		return 0;

1346
	if (state->npages == 1 && !register_always)
1347
		srp_map_desc(state, state->base_dma_addr, state->dma_len,
1348 1349
			     target->rkey);
	else
1350
		ret = target->srp_host->srp_dev->use_fast_reg ?
1351 1352
			srp_map_finish_fr(state, ch) :
			srp_map_finish_fmr(state, ch);
1353 1354 1355

	if (ret == 0) {
		state->npages = 0;
1356
		state->dma_len = 0;
1357 1358 1359 1360 1361
	}

	return ret;
}

1362 1363 1364 1365 1366 1367 1368 1369
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;
}
1370

1371
static int srp_map_sg_entry(struct srp_map_state *state,
1372
			    struct srp_rdma_ch *ch,
1373
			    struct scatterlist *sg, int sg_index,
1374
			    bool use_mr)
1375
{
1376
	struct srp_target_port *target = ch->target;
1377 1378 1379 1380 1381 1382 1383 1384 1385 1386
	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;

1387 1388 1389 1390
	if (!use_mr) {
		/*
		 * Once we're in direct map mode for a request, we don't
		 * go back to FMR or FR mode, so no need to update anything
1391 1392 1393 1394
		 * other than the descriptor.
		 */
		srp_map_desc(state, dma_addr, dma_len, target->rkey);
		return 0;
1395
	}
1396

1397 1398 1399 1400 1401
	/*
	 * Since not all RDMA HW drivers support non-zero page offsets for
	 * FMR, if we start at an offset into a page, don't merge into the
	 * current FMR mapping. Finish it out, and use the kernel's MR for
	 * this sg entry.
1402
	 */
1403 1404
	if ((!dev->use_fast_reg && dma_addr & ~dev->mr_page_mask) ||
	    dma_len > dev->mr_max_size) {
1405
		ret = srp_finish_mapping(state, ch);
1406 1407 1408 1409 1410 1411
		if (ret)
			return ret;

		srp_map_desc(state, dma_addr, dma_len, target->rkey);
		srp_map_update_start(state, NULL, 0, 0);
		return 0;
1412 1413
	}

1414 1415 1416 1417 1418
	/*
	 * If this is the first sg that will be mapped via FMR or via FR, 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.
1419 1420 1421
	 */
	if (!state->unmapped_sg)
		srp_map_update_start(state, sg, sg_index, dma_addr);
1422

1423
	while (dma_len) {
1424 1425
		unsigned offset = dma_addr & ~dev->mr_page_mask;
		if (state->npages == dev->max_pages_per_mr || offset != 0) {
1426
			ret = srp_finish_mapping(state, ch);
1427 1428
			if (ret)
				return ret;
1429

1430 1431 1432
			srp_map_update_start(state, sg, sg_index, dma_addr);
		}

1433
		len = min_t(unsigned int, dma_len, dev->mr_page_size - offset);
1434

1435 1436
		if (!state->npages)
			state->base_dma_addr = dma_addr;
1437
		state->pages[state->npages++] = dma_addr & dev->mr_page_mask;
1438
		state->dma_len += len;
1439 1440 1441 1442
		dma_addr += len;
		dma_len -= len;
	}

1443 1444
	/*
	 * If the last entry of the MR wasn't a full page, then we need to
1445 1446 1447 1448
	 * close it out and start a new one -- we can only merge at page
	 * boundries.
	 */
	ret = 0;
1449
	if (len != dev->mr_page_size) {
1450
		ret = srp_finish_mapping(state, ch);
1451 1452 1453
		if (!ret)
			srp_map_update_start(state, NULL, 0, 0);
	}
1454 1455 1456
	return ret;
}

1457 1458 1459
static int srp_map_sg(struct srp_map_state *state, struct srp_rdma_ch *ch,
		      struct srp_request *req, struct scatterlist *scat,
		      int count)
1460
{
1461
	struct srp_target_port *target = ch->target;
1462 1463 1464
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = dev->dev;
	struct scatterlist *sg;
1465 1466
	int i;
	bool use_mr;
1467 1468 1469

	state->desc	= req->indirect_desc;
	state->pages	= req->map_page;
1470 1471
	if (dev->use_fast_reg) {
		state->next_fr = req->fr_list;
1472
		use_mr = !!ch->fr_pool;
1473 1474
	} else {
		state->next_fmr = req->fmr_list;
1475
		use_mr = !!ch->fmr_pool;
1476
	}
1477 1478

	for_each_sg(scat, sg, count, i) {
1479
		if (srp_map_sg_entry(state, ch, sg, i, use_mr)) {
1480 1481 1482 1483
			/*
			 * Memory registration failed, so backtrack to the
			 * first unmapped entry and continue on without using
			 * memory registration.
1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495
			 */
			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;
1496
			use_mr = false;
1497 1498 1499 1500
			srp_map_desc(state, dma_addr, dma_len, target->rkey);
		}
	}

1501
	if (use_mr && srp_finish_mapping(state, ch))
1502 1503
		goto backtrack;

1504
	req->nmdesc = state->nmdesc;
1505 1506

	return 0;
1507 1508
}

1509
static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_rdma_ch *ch,
1510 1511
			struct srp_request *req)
{
1512
	struct srp_target_port *target = ch->target;
1513
	struct scatterlist *scat;
1514
	struct srp_cmd *cmd = req->cmd->buf;
1515
	int len, nents, count;
1516 1517
	struct srp_device *dev;
	struct ib_device *ibdev;
1518 1519 1520 1521
	struct srp_map_state state;
	struct srp_indirect_buf *indirect_hdr;
	u32 table_len;
	u8 fmt;
1522

1523
	if (!scsi_sglist(scmnd) || scmnd->sc_data_direction == DMA_NONE)
1524 1525 1526 1527
		return sizeof (struct srp_cmd);

	if (scmnd->sc_data_direction != DMA_FROM_DEVICE &&
	    scmnd->sc_data_direction != DMA_TO_DEVICE) {
1528 1529 1530
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled data direction %d\n",
			     scmnd->sc_data_direction);
1531 1532 1533
		return -EINVAL;
	}

1534 1535
	nents = scsi_sg_count(scmnd);
	scat  = scsi_sglist(scmnd);
1536

1537
	dev = target->srp_host->srp_dev;
1538 1539 1540
	ibdev = dev->dev;

	count = ib_dma_map_sg(ibdev, scat, nents, scmnd->sc_data_direction);
1541 1542
	if (unlikely(count == 0))
		return -EIO;
1543 1544 1545

	fmt = SRP_DATA_DESC_DIRECT;
	len = sizeof (struct srp_cmd) +	sizeof (struct srp_direct_buf);
1546

1547
	if (count == 1 && !register_always) {
1548 1549 1550 1551 1552 1553
		/*
		 * 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.
		 */
1554
		struct srp_direct_buf *buf = (void *) cmd->add_data;
1555

1556
		buf->va  = cpu_to_be64(ib_sg_dma_address(ibdev, scat));
1557
		buf->key = cpu_to_be32(target->rkey);
1558
		buf->len = cpu_to_be32(ib_sg_dma_len(ibdev, scat));
1559

1560
		req->nmdesc = 0;
1561 1562 1563
		goto map_complete;
	}

1564 1565 1566
	/*
	 * We have more than one scatter/gather entry, so build our indirect
	 * descriptor table, trying to merge as many entries as we can.
1567 1568 1569
	 */
	indirect_hdr = (void *) cmd->add_data;

1570 1571 1572
	ib_dma_sync_single_for_cpu(ibdev, req->indirect_dma_addr,
				   target->indirect_size, DMA_TO_DEVICE);

1573
	memset(&state, 0, sizeof(state));
1574
	srp_map_sg(&state, ch, req, scat, count);
1575

1576 1577 1578 1579 1580
	/* 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.
1581 1582
	 */
	if (state.ndesc == 1) {
1583 1584
		/*
		 * Memory registration collapsed the sg-list into one entry,
1585 1586 1587
		 * so use a direct descriptor.
		 */
		struct srp_direct_buf *buf = (void *) cmd->add_data;
1588

1589
		*buf = req->indirect_desc[0];
1590
		goto map_complete;
1591 1592
	}

1593 1594 1595 1596 1597 1598 1599 1600
	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);
1601 1602 1603 1604
	table_len = state.ndesc * sizeof (struct srp_direct_buf);

	fmt = SRP_DATA_DESC_INDIRECT;
	len = sizeof(struct srp_cmd) + sizeof (struct srp_indirect_buf);
1605
	len += count * sizeof (struct srp_direct_buf);
1606

1607 1608
	memcpy(indirect_hdr->desc_list, req->indirect_desc,
	       count * sizeof (struct srp_direct_buf));
1609

1610
	indirect_hdr->table_desc.va = cpu_to_be64(req->indirect_dma_addr);
1611 1612 1613 1614 1615
	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)
1616
		cmd->data_out_desc_cnt = count;
1617
	else
1618 1619 1620 1621
		cmd->data_in_desc_cnt = count;

	ib_dma_sync_single_for_device(ibdev, req->indirect_dma_addr, table_len,
				      DMA_TO_DEVICE);
1622 1623

map_complete:
1624 1625 1626 1627 1628 1629 1630 1631
	if (scmnd->sc_data_direction == DMA_TO_DEVICE)
		cmd->buf_fmt = fmt << 4;
	else
		cmd->buf_fmt = fmt;

	return len;
}

1632 1633 1634
/*
 * Return an IU and possible credit to the free pool
 */
1635
static void srp_put_tx_iu(struct srp_rdma_ch *ch, struct srp_iu *iu,
1636 1637 1638 1639
			  enum srp_iu_type iu_type)
{
	unsigned long flags;

1640 1641
	spin_lock_irqsave(&ch->lock, flags);
	list_add(&iu->list, &ch->free_tx);
1642
	if (iu_type != SRP_IU_RSP)
1643 1644
		++ch->req_lim;
	spin_unlock_irqrestore(&ch->lock, flags);
1645 1646
}

1647
/*
1648
 * Must be called with ch->lock held to protect req_lim and free_tx.
1649
 * If IU is not sent, it must be returned using srp_put_tx_iu().
1650 1651 1652 1653 1654 1655 1656 1657 1658 1659
 *
 * 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.
 */
1660
static struct srp_iu *__srp_get_tx_iu(struct srp_rdma_ch *ch,
1661 1662
				      enum srp_iu_type iu_type)
{
1663
	struct srp_target_port *target = ch->target;
1664 1665 1666
	s32 rsv = (iu_type == SRP_IU_TSK_MGMT) ? 0 : SRP_TSK_MGMT_SQ_SIZE;
	struct srp_iu *iu;

1667
	srp_send_completion(ch->send_cq, ch);
1668

1669
	if (list_empty(&ch->free_tx))
1670 1671 1672
		return NULL;

	/* Initiator responses to target requests do not consume credits */
1673
	if (iu_type != SRP_IU_RSP) {
1674
		if (ch->req_lim <= rsv) {
1675 1676 1677 1678
			++target->zero_req_lim;
			return NULL;
		}

1679
		--ch->req_lim;
1680 1681
	}

1682
	iu = list_first_entry(&ch->free_tx, struct srp_iu, list);
1683
	list_del(&iu->list);
1684 1685 1686
	return iu;
}

1687
static int srp_post_send(struct srp_rdma_ch *ch, struct srp_iu *iu, int len)
1688
{
1689
	struct srp_target_port *target = ch->target;
1690 1691 1692 1693 1694
	struct ib_sge list;
	struct ib_send_wr wr, *bad_wr;

	list.addr   = iu->dma;
	list.length = len;
1695
	list.lkey   = target->lkey;
1696 1697

	wr.next       = NULL;
1698
	wr.wr_id      = (uintptr_t) iu;
1699 1700 1701 1702 1703
	wr.sg_list    = &list;
	wr.num_sge    = 1;
	wr.opcode     = IB_WR_SEND;
	wr.send_flags = IB_SEND_SIGNALED;

1704
	return ib_post_send(ch->qp, &wr, &bad_wr);
1705 1706
}

1707
static int srp_post_recv(struct srp_rdma_ch *ch, struct srp_iu *iu)
1708
{
1709
	struct srp_target_port *target = ch->target;
1710
	struct ib_recv_wr wr, *bad_wr;
1711
	struct ib_sge list;
1712 1713 1714

	list.addr   = iu->dma;
	list.length = iu->size;
1715
	list.lkey   = target->lkey;
1716 1717

	wr.next     = NULL;
1718
	wr.wr_id    = (uintptr_t) iu;
1719 1720 1721
	wr.sg_list  = &list;
	wr.num_sge  = 1;

1722
	return ib_post_recv(ch->qp, &wr, &bad_wr);
1723 1724
}

1725
static void srp_process_rsp(struct srp_rdma_ch *ch, struct srp_rsp *rsp)
1726
{
1727
	struct srp_target_port *target = ch->target;
1728 1729 1730 1731 1732
	struct srp_request *req;
	struct scsi_cmnd *scmnd;
	unsigned long flags;

	if (unlikely(rsp->tag & SRP_TAG_TSK_MGMT)) {
1733 1734 1735
		spin_lock_irqsave(&ch->lock, flags);
		ch->req_lim += be32_to_cpu(rsp->req_lim_delta);
		spin_unlock_irqrestore(&ch->lock, flags);
1736

1737
		ch->tsk_mgmt_status = -1;
1738
		if (be32_to_cpu(rsp->resp_data_len) >= 4)
1739 1740
			ch->tsk_mgmt_status = rsp->data[3];
		complete(&ch->tsk_mgmt_done);
1741
	} else {
B
Bart Van Assche 已提交
1742 1743 1744 1745 1746
		scmnd = scsi_host_find_tag(target->scsi_host, rsp->tag);
		if (scmnd) {
			req = (void *)scmnd->host_scribble;
			scmnd = srp_claim_req(ch, req, NULL, scmnd);
		}
B
Bart Van Assche 已提交
1747
		if (!scmnd) {
1748
			shost_printk(KERN_ERR, target->scsi_host,
B
Bart Van Assche 已提交
1749 1750
				     "Null scmnd for RSP w/tag %#016llx received on ch %td / QP %#x\n",
				     rsp->tag, ch - target->ch, ch->qp->qp_num);
B
Bart Van Assche 已提交
1751

1752 1753 1754
			spin_lock_irqsave(&ch->lock, flags);
			ch->req_lim += be32_to_cpu(rsp->req_lim_delta);
			spin_unlock_irqrestore(&ch->lock, flags);
B
Bart Van Assche 已提交
1755 1756 1757

			return;
		}
1758 1759 1760 1761 1762 1763 1764 1765 1766
		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));
		}

B
Bart Van Assche 已提交
1767
		if (unlikely(rsp->flags & SRP_RSP_FLAG_DIUNDER))
1768
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_in_res_cnt));
B
Bart Van Assche 已提交
1769 1770 1771 1772 1773 1774
		else if (unlikely(rsp->flags & SRP_RSP_FLAG_DIOVER))
			scsi_set_resid(scmnd, -be32_to_cpu(rsp->data_in_res_cnt));
		else if (unlikely(rsp->flags & SRP_RSP_FLAG_DOUNDER))
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_out_res_cnt));
		else if (unlikely(rsp->flags & SRP_RSP_FLAG_DOOVER))
			scsi_set_resid(scmnd, -be32_to_cpu(rsp->data_out_res_cnt));
1775

1776
		srp_free_req(ch, req, scmnd,
B
Bart Van Assche 已提交
1777 1778
			     be32_to_cpu(rsp->req_lim_delta));

1779 1780
		scmnd->host_scribble = NULL;
		scmnd->scsi_done(scmnd);
1781 1782 1783
	}
}

1784
static int srp_response_common(struct srp_rdma_ch *ch, s32 req_delta,
1785 1786
			       void *rsp, int len)
{
1787
	struct srp_target_port *target = ch->target;
1788
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1789 1790
	unsigned long flags;
	struct srp_iu *iu;
1791
	int err;
1792

1793 1794 1795 1796
	spin_lock_irqsave(&ch->lock, flags);
	ch->req_lim += req_delta;
	iu = __srp_get_tx_iu(ch, SRP_IU_RSP);
	spin_unlock_irqrestore(&ch->lock, flags);
1797

1798 1799 1800
	if (!iu) {
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "no IU available to send response\n");
1801
		return 1;
1802 1803 1804 1805 1806 1807
	}

	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);

1808
	err = srp_post_send(ch, iu, len);
1809
	if (err) {
1810 1811
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "unable to post response: %d\n", err);
1812
		srp_put_tx_iu(ch, iu, SRP_IU_RSP);
1813
	}
1814 1815 1816 1817

	return err;
}

1818
static void srp_process_cred_req(struct srp_rdma_ch *ch,
1819 1820 1821 1822 1823 1824 1825 1826
				 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);

1827 1828
	if (srp_response_common(ch, delta, &rsp, sizeof(rsp)))
		shost_printk(KERN_ERR, ch->target->scsi_host, PFX
1829 1830 1831
			     "problems processing SRP_CRED_REQ\n");
}

1832
static void srp_process_aer_req(struct srp_rdma_ch *ch,
1833 1834
				struct srp_aer_req *req)
{
1835
	struct srp_target_port *target = ch->target;
1836 1837 1838 1839 1840 1841 1842 1843 1844
	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));

1845
	if (srp_response_common(ch, delta, &rsp, sizeof(rsp)))
1846 1847 1848 1849
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "problems processing SRP_AER_REQ\n");
}

1850
static void srp_handle_recv(struct srp_rdma_ch *ch, struct ib_wc *wc)
1851
{
1852
	struct srp_target_port *target = ch->target;
1853
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1854
	struct srp_iu *iu = (struct srp_iu *) (uintptr_t) wc->wr_id;
1855
	int res;
1856 1857
	u8 opcode;

1858
	ib_dma_sync_single_for_cpu(dev, iu->dma, ch->max_ti_iu_len,
1859
				   DMA_FROM_DEVICE);
1860 1861 1862 1863

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

	if (0) {
1864 1865
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "recv completion, opcode 0x%02x\n", opcode);
B
Bart Van Assche 已提交
1866 1867
		print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET, 8, 1,
			       iu->buf, wc->byte_len, true);
1868 1869 1870 1871
	}

	switch (opcode) {
	case SRP_RSP:
1872
		srp_process_rsp(ch, iu->buf);
1873 1874
		break;

1875
	case SRP_CRED_REQ:
1876
		srp_process_cred_req(ch, iu->buf);
1877 1878 1879
		break;

	case SRP_AER_REQ:
1880
		srp_process_aer_req(ch, iu->buf);
1881 1882
		break;

1883 1884
	case SRP_T_LOGOUT:
		/* XXX Handle target logout */
1885 1886
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Got target logout request\n");
1887 1888 1889
		break;

	default:
1890 1891
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled SRP opcode 0x%02x\n", opcode);
1892 1893 1894
		break;
	}

1895
	ib_dma_sync_single_for_device(dev, iu->dma, ch->max_ti_iu_len,
1896
				      DMA_FROM_DEVICE);
1897

1898
	res = srp_post_recv(ch, iu);
1899 1900 1901
	if (res != 0)
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Recv failed with error code %d\n", res);
1902 1903
}

1904 1905
/**
 * srp_tl_err_work() - handle a transport layer error
1906
 * @work: Work structure embedded in an SRP target port.
1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919
 *
 * 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);
}

1920
static void srp_handle_qp_err(u64 wr_id, enum ib_wc_status wc_status,
1921
			      bool send_err, struct srp_rdma_ch *ch)
1922
{
1923 1924 1925 1926 1927 1928 1929
	struct srp_target_port *target = ch->target;

	if (wr_id == SRP_LAST_WR_ID) {
		complete(&ch->done);
		return;
	}

1930
	if (ch->connected && !target->qp_in_error) {
1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944
		if (wr_id & LOCAL_INV_WR_ID_MASK) {
			shost_printk(KERN_ERR, target->scsi_host, PFX
				     "LOCAL_INV failed with status %d\n",
				     wc_status);
		} else if (wr_id & FAST_REG_WR_ID_MASK) {
			shost_printk(KERN_ERR, target->scsi_host, PFX
				     "FAST_REG_MR failed status %d\n",
				     wc_status);
		} else {
			shost_printk(KERN_ERR, target->scsi_host,
				     PFX "failed %s status %d for iu %p\n",
				     send_err ? "send" : "receive",
				     wc_status, (void *)(uintptr_t)wr_id);
		}
1945
		queue_work(system_long_wq, &target->tl_err_work);
1946
	}
1947 1948 1949
	target->qp_in_error = true;
}

1950
static void srp_recv_completion(struct ib_cq *cq, void *ch_ptr)
1951
{
1952
	struct srp_rdma_ch *ch = ch_ptr;
1953 1954 1955 1956
	struct ib_wc wc;

	ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
	while (ib_poll_cq(cq, 1, &wc) > 0) {
1957
		if (likely(wc.status == IB_WC_SUCCESS)) {
1958
			srp_handle_recv(ch, &wc);
1959
		} else {
1960
			srp_handle_qp_err(wc.wr_id, wc.status, false, ch);
1961
		}
1962 1963 1964
	}
}

1965
static void srp_send_completion(struct ib_cq *cq, void *ch_ptr)
1966
{
1967
	struct srp_rdma_ch *ch = ch_ptr;
1968
	struct ib_wc wc;
1969
	struct srp_iu *iu;
1970 1971

	while (ib_poll_cq(cq, 1, &wc) > 0) {
1972 1973
		if (likely(wc.status == IB_WC_SUCCESS)) {
			iu = (struct srp_iu *) (uintptr_t) wc.wr_id;
1974
			list_add(&iu->list, &ch->free_tx);
1975
		} else {
1976
			srp_handle_qp_err(wc.wr_id, wc.status, true, ch);
1977
		}
1978 1979 1980
	}
}

1981
static int srp_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *scmnd)
1982
{
1983
	struct srp_target_port *target = host_to_target(shost);
1984
	struct srp_rport *rport = target->rport;
1985
	struct srp_rdma_ch *ch;
1986 1987 1988
	struct srp_request *req;
	struct srp_iu *iu;
	struct srp_cmd *cmd;
1989
	struct ib_device *dev;
1990
	unsigned long flags;
B
Bart Van Assche 已提交
1991 1992
	u32 tag;
	u16 idx;
1993
	int len, ret;
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
	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);
2004

2005 2006 2007
	scmnd->result = srp_chkready(target->rport);
	if (unlikely(scmnd->result))
		goto err;
2008

B
Bart Van Assche 已提交
2009 2010
	WARN_ON_ONCE(scmnd->request->tag < 0);
	tag = blk_mq_unique_tag(scmnd->request);
B
Bart Van Assche 已提交
2011
	ch = &target->ch[blk_mq_unique_tag_to_hwq(tag)];
B
Bart Van Assche 已提交
2012 2013 2014 2015
	idx = blk_mq_unique_tag_to_tag(tag);
	WARN_ONCE(idx >= target->req_ring_size, "%s: tag %#x: idx %d >= %d\n",
		  dev_name(&shost->shost_gendev), tag, idx,
		  target->req_ring_size);
2016 2017 2018 2019

	spin_lock_irqsave(&ch->lock, flags);
	iu = __srp_get_tx_iu(ch, SRP_IU_CMD);
	spin_unlock_irqrestore(&ch->lock, flags);
2020

B
Bart Van Assche 已提交
2021 2022 2023 2024
	if (!iu)
		goto err;

	req = &ch->req_ring[idx];
2025
	dev = target->srp_host->srp_dev->dev;
2026
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_iu_len,
2027
				   DMA_TO_DEVICE);
2028

2029
	scmnd->host_scribble = (void *) req;
2030 2031 2032 2033 2034 2035

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

	cmd->opcode = SRP_CMD;
	cmd->lun    = cpu_to_be64((u64) scmnd->device->lun << 48);
B
Bart Van Assche 已提交
2036
	cmd->tag    = tag;
2037 2038 2039 2040 2041
	memcpy(cmd->cdb, scmnd->cmnd, scmnd->cmd_len);

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

2042
	len = srp_map_data(scmnd, ch, req);
2043
	if (len < 0) {
2044
		shost_printk(KERN_ERR, target->scsi_host,
2045 2046 2047 2048
			     PFX "Failed to map data (%d)\n", len);
		/*
		 * If we ran out of memory descriptors (-ENOMEM) because an
		 * application is queuing many requests with more than
2049
		 * max_pages_per_mr sg-list elements, tell the SCSI mid-layer
2050 2051 2052 2053
		 * to reduce queue depth temporarily.
		 */
		scmnd->result = len == -ENOMEM ?
			DID_OK << 16 | QUEUE_FULL << 1 : DID_ERROR << 16;
2054
		goto err_iu;
2055 2056
	}

2057
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_iu_len,
2058
				      DMA_TO_DEVICE);
2059

2060
	if (srp_post_send(ch, iu, len)) {
2061
		shost_printk(KERN_ERR, target->scsi_host, PFX "Send failed\n");
2062 2063 2064
		goto err_unmap;
	}

2065 2066
	ret = 0;

2067 2068 2069 2070
unlock_rport:
	if (in_scsi_eh)
		mutex_unlock(&rport->mutex);

2071
	return ret;
2072 2073

err_unmap:
2074
	srp_unmap_data(scmnd, ch, req);
2075

2076
err_iu:
2077
	srp_put_tx_iu(ch, iu, SRP_IU_CMD);
2078

2079 2080 2081 2082 2083 2084
	/*
	 * Avoid that the loops that iterate over the request ring can
	 * encounter a dangling SCSI command pointer.
	 */
	req->scmnd = NULL;

2085 2086 2087 2088 2089 2090 2091
err:
	if (scmnd->result) {
		scmnd->scsi_done(scmnd);
		ret = 0;
	} else {
		ret = SCSI_MLQUEUE_HOST_BUSY;
	}
2092

2093
	goto unlock_rport;
2094 2095
}

2096 2097
/*
 * Note: the resources allocated in this function are freed in
2098
 * srp_free_ch_ib().
2099
 */
2100
static int srp_alloc_iu_bufs(struct srp_rdma_ch *ch)
2101
{
2102
	struct srp_target_port *target = ch->target;
2103 2104
	int i;

2105 2106 2107
	ch->rx_ring = kcalloc(target->queue_size, sizeof(*ch->rx_ring),
			      GFP_KERNEL);
	if (!ch->rx_ring)
2108
		goto err_no_ring;
2109 2110 2111
	ch->tx_ring = kcalloc(target->queue_size, sizeof(*ch->tx_ring),
			      GFP_KERNEL);
	if (!ch->tx_ring)
2112 2113 2114
		goto err_no_ring;

	for (i = 0; i < target->queue_size; ++i) {
2115 2116 2117 2118
		ch->rx_ring[i] = srp_alloc_iu(target->srp_host,
					      ch->max_ti_iu_len,
					      GFP_KERNEL, DMA_FROM_DEVICE);
		if (!ch->rx_ring[i])
2119 2120 2121
			goto err;
	}

2122
	for (i = 0; i < target->queue_size; ++i) {
2123 2124 2125 2126
		ch->tx_ring[i] = srp_alloc_iu(target->srp_host,
					      target->max_iu_len,
					      GFP_KERNEL, DMA_TO_DEVICE);
		if (!ch->tx_ring[i])
2127
			goto err;
2128

2129
		list_add(&ch->tx_ring[i]->list, &ch->free_tx);
2130 2131 2132 2133 2134
	}

	return 0;

err:
2135
	for (i = 0; i < target->queue_size; ++i) {
2136 2137
		srp_free_iu(target->srp_host, ch->rx_ring[i]);
		srp_free_iu(target->srp_host, ch->tx_ring[i]);
2138 2139
	}

2140 2141

err_no_ring:
2142 2143 2144 2145
	kfree(ch->tx_ring);
	ch->tx_ring = NULL;
	kfree(ch->rx_ring);
	ch->rx_ring = NULL;
2146

2147 2148 2149
	return -ENOMEM;
}

2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176
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;
}

2177 2178
static void srp_cm_rep_handler(struct ib_cm_id *cm_id,
			       struct srp_login_rsp *lrsp,
2179
			       struct srp_rdma_ch *ch)
2180
{
2181
	struct srp_target_port *target = ch->target;
2182 2183 2184 2185 2186 2187
	struct ib_qp_attr *qp_attr = NULL;
	int attr_mask = 0;
	int ret;
	int i;

	if (lrsp->opcode == SRP_LOGIN_RSP) {
2188 2189
		ch->max_ti_iu_len = be32_to_cpu(lrsp->max_ti_iu_len);
		ch->req_lim       = be32_to_cpu(lrsp->req_lim_delta);
2190 2191 2192 2193 2194 2195

		/*
		 * Reserve credits for task management so we don't
		 * bounce requests back to the SCSI mid-layer.
		 */
		target->scsi_host->can_queue
2196
			= min(ch->req_lim - SRP_TSK_MGMT_SQ_SIZE,
2197
			      target->scsi_host->can_queue);
2198 2199 2200
		target->scsi_host->cmd_per_lun
			= min_t(int, target->scsi_host->can_queue,
				target->scsi_host->cmd_per_lun);
2201 2202 2203 2204 2205 2206 2207
	} else {
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled RSP opcode %#x\n", lrsp->opcode);
		ret = -ECONNRESET;
		goto error;
	}

2208 2209
	if (!ch->rx_ring) {
		ret = srp_alloc_iu_bufs(ch);
2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223
		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;

2224
	ret = ib_modify_qp(ch->qp, qp_attr, attr_mask);
2225 2226 2227
	if (ret)
		goto error_free;

2228
	for (i = 0; i < target->queue_size; i++) {
2229 2230 2231
		struct srp_iu *iu = ch->rx_ring[i];

		ret = srp_post_recv(ch, iu);
2232 2233 2234 2235 2236 2237 2238 2239 2240
		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;

2241 2242
	target->rq_tmo_jiffies = srp_compute_rq_tmo(qp_attr, attr_mask);

2243
	ret = ib_modify_qp(ch->qp, qp_attr, attr_mask);
2244 2245 2246 2247 2248 2249 2250 2251 2252
	if (ret)
		goto error_free;

	ret = ib_send_cm_rtu(cm_id, NULL, 0);

error_free:
	kfree(qp_attr);

error:
2253
	ch->status = ret;
2254 2255
}

2256 2257
static void srp_cm_rej_handler(struct ib_cm_id *cm_id,
			       struct ib_cm_event *event,
2258
			       struct srp_rdma_ch *ch)
2259
{
2260
	struct srp_target_port *target = ch->target;
2261
	struct Scsi_Host *shost = target->scsi_host;
2262 2263 2264 2265 2266 2267
	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;
2268 2269
		ch->path.dlid = cpi->redirect_lid;
		ch->path.pkey = cpi->redirect_pkey;
2270
		cm_id->remote_cm_qpn = be32_to_cpu(cpi->redirect_qp) & 0x00ffffff;
2271
		memcpy(ch->path.dgid.raw, cpi->redirect_gid, 16);
2272

2273
		ch->status = ch->path.dlid ?
2274 2275 2276 2277
			SRP_DLID_REDIRECT : SRP_PORT_REDIRECT;
		break;

	case IB_CM_REJ_PORT_REDIRECT:
2278
		if (srp_target_is_topspin(target)) {
2279 2280 2281 2282 2283
			/*
			 * Topspin/Cisco SRP gateways incorrectly send
			 * reject reason code 25 when they mean 24
			 * (port redirect).
			 */
2284
			memcpy(ch->path.dgid.raw,
2285 2286
			       event->param.rej_rcvd.ari, 16);

2287 2288
			shost_printk(KERN_DEBUG, shost,
				     PFX "Topspin/Cisco redirect to target port GID %016llx%016llx\n",
2289 2290
				     be64_to_cpu(ch->path.dgid.global.subnet_prefix),
				     be64_to_cpu(ch->path.dgid.global.interface_id));
2291

2292
			ch->status = SRP_PORT_REDIRECT;
2293
		} else {
2294 2295
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_PORT_REDIRECT\n");
2296
			ch->status = -ECONNRESET;
2297 2298 2299 2300
		}
		break;

	case IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID:
2301 2302
		shost_printk(KERN_WARNING, shost,
			    "  REJ reason: IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID\n");
2303
		ch->status = -ECONNRESET;
2304 2305 2306 2307 2308 2309 2310 2311 2312
		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)
2313 2314
				shost_printk(KERN_WARNING, shost,
					     PFX "SRP_LOGIN_REJ: requested max_it_iu_len too large\n");
2315
			else
B
Bart Van Assche 已提交
2316 2317
				shost_printk(KERN_WARNING, shost, PFX
					     "SRP LOGIN from %pI6 to %pI6 REJECTED, reason 0x%08x\n",
2318 2319
					     target->sgid.raw,
					     target->orig_dgid.raw, reason);
2320
		} else
2321 2322 2323
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_CONSUMER_DEFINED,"
				     " opcode 0x%02x\n", opcode);
2324
		ch->status = -ECONNRESET;
2325 2326
		break;

D
David Dillow 已提交
2327 2328
	case IB_CM_REJ_STALE_CONN:
		shost_printk(KERN_WARNING, shost, "  REJ reason: stale connection\n");
2329
		ch->status = SRP_STALE_CONN;
D
David Dillow 已提交
2330 2331
		break;

2332
	default:
2333 2334
		shost_printk(KERN_WARNING, shost, "  REJ reason 0x%x\n",
			     event->param.rej_rcvd.reason);
2335
		ch->status = -ECONNRESET;
2336 2337 2338 2339 2340
	}
}

static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event)
{
2341 2342
	struct srp_rdma_ch *ch = cm_id->context;
	struct srp_target_port *target = ch->target;
2343 2344 2345 2346
	int comp = 0;

	switch (event->event) {
	case IB_CM_REQ_ERROR:
2347 2348
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Sending CM REQ failed\n");
2349
		comp = 1;
2350
		ch->status = -ECONNRESET;
2351 2352 2353 2354
		break;

	case IB_CM_REP_RECEIVED:
		comp = 1;
2355
		srp_cm_rep_handler(cm_id, event->private_data, ch);
2356 2357 2358
		break;

	case IB_CM_REJ_RECEIVED:
2359
		shost_printk(KERN_DEBUG, target->scsi_host, PFX "REJ received\n");
2360 2361
		comp = 1;

2362
		srp_cm_rej_handler(cm_id, event, ch);
2363 2364
		break;

2365
	case IB_CM_DREQ_RECEIVED:
2366 2367
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "DREQ received - connection closed\n");
2368
		ch->connected = false;
2369
		if (ib_send_cm_drep(cm_id, NULL, 0))
2370 2371
			shost_printk(KERN_ERR, target->scsi_host,
				     PFX "Sending CM DREP failed\n");
2372
		queue_work(system_long_wq, &target->tl_err_work);
2373 2374 2375
		break;

	case IB_CM_TIMEWAIT_EXIT:
2376 2377
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "connection closed\n");
2378
		comp = 1;
2379

2380
		ch->status = 0;
2381 2382
		break;

2383 2384 2385 2386 2387
	case IB_CM_MRA_RECEIVED:
	case IB_CM_DREQ_ERROR:
	case IB_CM_DREP_RECEIVED:
		break;

2388
	default:
2389 2390
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled CM event %d\n", event->event);
2391 2392 2393 2394
		break;
	}

	if (comp)
2395
		complete(&ch->done);
2396 2397 2398 2399

	return 0;
}

2400 2401 2402 2403 2404 2405 2406 2407
/**
 * srp_change_queue_depth - setting device queue depth
 * @sdev: scsi device struct
 * @qdepth: requested queue depth
 *
 * Returns queue depth.
 */
static int
2408
srp_change_queue_depth(struct scsi_device *sdev, int qdepth)
2409
{
2410
	if (!sdev->tagged_supported)
2411
		qdepth = 1;
2412
	return scsi_change_queue_depth(sdev, qdepth);
2413 2414
}

2415 2416
static int srp_send_tsk_mgmt(struct srp_rdma_ch *ch, u64 req_tag,
			     unsigned int lun, u8 func)
2417
{
2418
	struct srp_target_port *target = ch->target;
2419
	struct srp_rport *rport = target->rport;
2420
	struct ib_device *dev = target->srp_host->srp_dev->dev;
2421 2422 2423
	struct srp_iu *iu;
	struct srp_tsk_mgmt *tsk_mgmt;

2424
	if (!ch->connected || target->qp_in_error)
2425 2426
		return -1;

2427
	init_completion(&ch->tsk_mgmt_done);
2428

2429
	/*
2430
	 * Lock the rport mutex to avoid that srp_create_ch_ib() is
2431 2432 2433
	 * invoked while a task management function is being sent.
	 */
	mutex_lock(&rport->mutex);
2434 2435 2436
	spin_lock_irq(&ch->lock);
	iu = __srp_get_tx_iu(ch, SRP_IU_TSK_MGMT);
	spin_unlock_irq(&ch->lock);
2437

2438 2439 2440
	if (!iu) {
		mutex_unlock(&rport->mutex);

2441
		return -1;
2442
	}
2443

2444 2445
	ib_dma_sync_single_for_cpu(dev, iu->dma, sizeof *tsk_mgmt,
				   DMA_TO_DEVICE);
2446 2447 2448 2449
	tsk_mgmt = iu->buf;
	memset(tsk_mgmt, 0, sizeof *tsk_mgmt);

	tsk_mgmt->opcode 	= SRP_TSK_MGMT;
2450 2451
	tsk_mgmt->lun		= cpu_to_be64((u64) lun << 48);
	tsk_mgmt->tag		= req_tag | SRP_TAG_TSK_MGMT;
2452
	tsk_mgmt->tsk_mgmt_func = func;
2453
	tsk_mgmt->task_tag	= req_tag;
2454

2455 2456
	ib_dma_sync_single_for_device(dev, iu->dma, sizeof *tsk_mgmt,
				      DMA_TO_DEVICE);
2457 2458
	if (srp_post_send(ch, iu, sizeof(*tsk_mgmt))) {
		srp_put_tx_iu(ch, iu, SRP_IU_TSK_MGMT);
2459 2460
		mutex_unlock(&rport->mutex);

2461 2462
		return -1;
	}
2463
	mutex_unlock(&rport->mutex);
2464

2465
	if (!wait_for_completion_timeout(&ch->tsk_mgmt_done,
2466
					 msecs_to_jiffies(SRP_ABORT_TIMEOUT_MS)))
2467
		return -1;
2468

2469 2470 2471
	return 0;
}

2472 2473
static int srp_abort(struct scsi_cmnd *scmnd)
{
2474
	struct srp_target_port *target = host_to_target(scmnd->device->host);
2475
	struct srp_request *req = (struct srp_request *) scmnd->host_scribble;
B
Bart Van Assche 已提交
2476
	u32 tag;
B
Bart Van Assche 已提交
2477
	u16 ch_idx;
2478
	struct srp_rdma_ch *ch;
2479
	int ret;
2480

2481
	shost_printk(KERN_ERR, target->scsi_host, "SRP abort called\n");
2482

B
Bart Van Assche 已提交
2483
	if (!req)
2484
		return SUCCESS;
B
Bart Van Assche 已提交
2485
	tag = blk_mq_unique_tag(scmnd->request);
B
Bart Van Assche 已提交
2486 2487 2488 2489 2490 2491 2492 2493
	ch_idx = blk_mq_unique_tag_to_hwq(tag);
	if (WARN_ON_ONCE(ch_idx >= target->ch_count))
		return SUCCESS;
	ch = &target->ch[ch_idx];
	if (!srp_claim_req(ch, req, NULL, scmnd))
		return SUCCESS;
	shost_printk(KERN_ERR, target->scsi_host,
		     "Sending SRP abort for tag %#x\n", tag);
B
Bart Van Assche 已提交
2494
	if (srp_send_tsk_mgmt(ch, tag, scmnd->device->lun,
2495
			      SRP_TSK_ABORT_TASK) == 0)
2496
		ret = SUCCESS;
2497
	else if (target->rport->state == SRP_RPORT_LOST)
2498
		ret = FAST_IO_FAIL;
2499 2500
	else
		ret = FAILED;
2501
	srp_free_req(ch, req, scmnd, 0);
B
Bart Van Assche 已提交
2502
	scmnd->result = DID_ABORT << 16;
2503
	scmnd->scsi_done(scmnd);
2504

2505
	return ret;
2506 2507 2508 2509
}

static int srp_reset_device(struct scsi_cmnd *scmnd)
{
2510
	struct srp_target_port *target = host_to_target(scmnd->device->host);
B
Bart Van Assche 已提交
2511
	struct srp_rdma_ch *ch;
2512
	int i;
2513

2514
	shost_printk(KERN_ERR, target->scsi_host, "SRP reset_device called\n");
2515

B
Bart Van Assche 已提交
2516
	ch = &target->ch[0];
2517
	if (srp_send_tsk_mgmt(ch, SRP_TAG_NO_REQ, scmnd->device->lun,
2518
			      SRP_TSK_LUN_RESET))
2519
		return FAILED;
2520
	if (ch->tsk_mgmt_status)
2521 2522
		return FAILED;

B
Bart Van Assche 已提交
2523 2524 2525 2526
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		for (i = 0; i < target->req_ring_size; ++i) {
			struct srp_request *req = &ch->req_ring[i];
2527

B
Bart Van Assche 已提交
2528 2529
			srp_finish_req(ch, req, scmnd->device, DID_RESET << 16);
		}
2530
	}
2531 2532

	return SUCCESS;
2533 2534 2535 2536 2537 2538
}

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

2539
	shost_printk(KERN_ERR, target->scsi_host, PFX "SRP reset_host called\n");
2540

2541
	return srp_reconnect_rport(target->rport) == 0 ? SUCCESS : FAILED;
2542 2543
}

2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558
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;
}

2559 2560
static ssize_t show_id_ext(struct device *dev, struct device_attribute *attr,
			   char *buf)
2561
{
2562
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2563 2564 2565 2566 2567

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

2568 2569
static ssize_t show_ioc_guid(struct device *dev, struct device_attribute *attr,
			     char *buf)
2570
{
2571
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2572 2573 2574 2575 2576

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

2577 2578
static ssize_t show_service_id(struct device *dev,
			       struct device_attribute *attr, char *buf)
2579
{
2580
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2581 2582 2583 2584 2585

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

2586 2587
static ssize_t show_pkey(struct device *dev, struct device_attribute *attr,
			 char *buf)
2588
{
2589
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2590

2591
	return sprintf(buf, "0x%04x\n", be16_to_cpu(target->pkey));
2592 2593
}

B
Bart Van Assche 已提交
2594 2595 2596 2597 2598
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));

2599
	return sprintf(buf, "%pI6\n", target->sgid.raw);
B
Bart Van Assche 已提交
2600 2601
}

2602 2603
static ssize_t show_dgid(struct device *dev, struct device_attribute *attr,
			 char *buf)
2604
{
2605
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
B
Bart Van Assche 已提交
2606
	struct srp_rdma_ch *ch = &target->ch[0];
2607

2608
	return sprintf(buf, "%pI6\n", ch->path.dgid.raw);
2609 2610
}

2611 2612
static ssize_t show_orig_dgid(struct device *dev,
			      struct device_attribute *attr, char *buf)
2613
{
2614
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2615

2616
	return sprintf(buf, "%pI6\n", target->orig_dgid.raw);
2617 2618
}

2619 2620 2621 2622
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));
B
Bart Van Assche 已提交
2623 2624
	struct srp_rdma_ch *ch;
	int i, req_lim = INT_MAX;
2625

B
Bart Van Assche 已提交
2626 2627 2628 2629 2630
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		req_lim = min(req_lim, ch->req_lim);
	}
	return sprintf(buf, "%d\n", req_lim);
2631 2632
}

2633 2634
static ssize_t show_zero_req_lim(struct device *dev,
				 struct device_attribute *attr, char *buf)
2635
{
2636
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2637 2638 2639 2640

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

2641 2642
static ssize_t show_local_ib_port(struct device *dev,
				  struct device_attribute *attr, char *buf)
2643
{
2644
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2645 2646 2647 2648

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

2649 2650
static ssize_t show_local_ib_device(struct device *dev,
				    struct device_attribute *attr, char *buf)
2651
{
2652
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2653

2654
	return sprintf(buf, "%s\n", target->srp_host->srp_dev->dev->name);
2655 2656
}

B
Bart Van Assche 已提交
2657 2658 2659 2660 2661 2662 2663 2664
static ssize_t show_ch_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->ch_count);
}

2665 2666 2667 2668 2669 2670 2671 2672
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);
}

2673 2674 2675 2676 2677 2678 2679 2680
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);
}

2681 2682 2683 2684 2685 2686 2687 2688
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);
}

2689 2690 2691 2692 2693 2694 2695 2696
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");
}

2697 2698 2699 2700
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
2701
static DEVICE_ATTR(sgid,	    S_IRUGO, show_sgid,		   NULL);
2702 2703
static DEVICE_ATTR(dgid,	    S_IRUGO, show_dgid,		   NULL);
static DEVICE_ATTR(orig_dgid,	    S_IRUGO, show_orig_dgid,	   NULL);
2704
static DEVICE_ATTR(req_lim,         S_IRUGO, show_req_lim,         NULL);
2705 2706 2707
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);
B
Bart Van Assche 已提交
2708
static DEVICE_ATTR(ch_count,        S_IRUGO, show_ch_count,        NULL);
2709
static DEVICE_ATTR(comp_vector,     S_IRUGO, show_comp_vector,     NULL);
2710
static DEVICE_ATTR(tl_retry_count,  S_IRUGO, show_tl_retry_count,  NULL);
2711
static DEVICE_ATTR(cmd_sg_entries,  S_IRUGO, show_cmd_sg_entries,  NULL);
2712
static DEVICE_ATTR(allow_ext_sg,    S_IRUGO, show_allow_ext_sg,    NULL);
2713 2714 2715 2716 2717 2718

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
2719
	&dev_attr_sgid,
2720 2721
	&dev_attr_dgid,
	&dev_attr_orig_dgid,
2722
	&dev_attr_req_lim,
2723 2724 2725
	&dev_attr_zero_req_lim,
	&dev_attr_local_ib_port,
	&dev_attr_local_ib_device,
B
Bart Van Assche 已提交
2726
	&dev_attr_ch_count,
2727
	&dev_attr_comp_vector,
2728
	&dev_attr_tl_retry_count,
2729
	&dev_attr_cmd_sg_entries,
2730
	&dev_attr_allow_ext_sg,
2731 2732 2733
	NULL
};

2734 2735
static struct scsi_host_template srp_template = {
	.module				= THIS_MODULE,
R
Roland Dreier 已提交
2736 2737
	.name				= "InfiniBand SRP initiator",
	.proc_name			= DRV_NAME,
2738
	.slave_configure		= srp_slave_configure,
2739 2740
	.info				= srp_target_info,
	.queuecommand			= srp_queuecommand,
2741
	.change_queue_depth             = srp_change_queue_depth,
2742 2743 2744
	.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
2745
	.skip_settle_delay		= true,
2746
	.sg_tablesize			= SRP_DEF_SG_TABLESIZE,
2747
	.can_queue			= SRP_DEFAULT_CMD_SQ_SIZE,
2748
	.this_id			= -1,
2749
	.cmd_per_lun			= SRP_DEFAULT_CMD_SQ_SIZE,
2750
	.use_clustering			= ENABLE_CLUSTERING,
B
Bart Van Assche 已提交
2751 2752
	.shost_attrs			= srp_host_attrs,
	.use_blk_tags			= 1,
2753
	.track_queue_depth		= 1,
2754 2755
};

2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766
static int srp_sdev_count(struct Scsi_Host *host)
{
	struct scsi_device *sdev;
	int c = 0;

	shost_for_each_device(sdev, host)
		c++;

	return c;
}

2767 2768
static int srp_add_target(struct srp_host *host, struct srp_target_port *target)
{
2769 2770 2771
	struct srp_rport_identifiers ids;
	struct srp_rport *rport;

2772
	target->state = SRP_TARGET_SCANNING;
2773 2774 2775
	sprintf(target->target_name, "SRP.T10:%016llX",
		 (unsigned long long) be64_to_cpu(target->id_ext));

2776
	if (scsi_add_host(target->scsi_host, host->srp_dev->dev->dma_device))
2777 2778
		return -ENODEV;

2779 2780
	memcpy(ids.port_id, &target->id_ext, 8);
	memcpy(ids.port_id + 8, &target->ioc_guid, 8);
2781
	ids.roles = SRP_RPORT_ROLE_TARGET;
2782 2783 2784 2785 2786 2787
	rport = srp_rport_add(target->scsi_host, &ids);
	if (IS_ERR(rport)) {
		scsi_remove_host(target->scsi_host);
		return PTR_ERR(rport);
	}

2788
	rport->lld_data = target;
2789
	target->rport = rport;
2790

2791
	spin_lock(&host->target_lock);
2792
	list_add_tail(&target->list, &host->target_list);
2793
	spin_unlock(&host->target_lock);
2794 2795

	scsi_scan_target(&target->scsi_host->shost_gendev,
2796
			 0, target->scsi_id, SCAN_WILD_CARD, 0);
2797

2798 2799
	if (srp_connected_ch(target) < target->ch_count ||
	    target->qp_in_error) {
2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815
		shost_printk(KERN_INFO, target->scsi_host,
			     PFX "SCSI scan failed - removing SCSI host\n");
		srp_queue_remove_work(target);
		goto out;
	}

	pr_debug(PFX "%s: SCSI scan succeeded - detected %d LUNs\n",
		 dev_name(&target->scsi_host->shost_gendev),
		 srp_sdev_count(target->scsi_host));

	spin_lock_irq(&target->lock);
	if (target->state == SRP_TARGET_SCANNING)
		target->state = SRP_TARGET_LIVE;
	spin_unlock_irq(&target->lock);

out:
2816 2817 2818
	return 0;
}

2819
static void srp_release_dev(struct device *dev)
2820 2821
{
	struct srp_host *host =
2822
		container_of(dev, struct srp_host, dev);
2823 2824 2825 2826 2827 2828

	complete(&host->released);
}

static struct class srp_class = {
	.name    = "infiniband_srp",
2829
	.dev_release = srp_release_dev
2830 2831
};

2832 2833
/**
 * srp_conn_unique() - check whether the connection to a target is unique
2834 2835
 * @host:   SRP host.
 * @target: SRP target port.
2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863
 */
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;
}

2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879
/*
 * 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,
2880
	SRP_OPT_MAX_CMD_PER_LUN	= 1 << 6,
2881
	SRP_OPT_IO_CLASS	= 1 << 7,
2882
	SRP_OPT_INITIATOR_EXT	= 1 << 8,
2883
	SRP_OPT_CMD_SG_ENTRIES	= 1 << 9,
2884 2885
	SRP_OPT_ALLOW_EXT_SG	= 1 << 10,
	SRP_OPT_SG_TABLESIZE	= 1 << 11,
2886
	SRP_OPT_COMP_VECTOR	= 1 << 12,
2887
	SRP_OPT_TL_RETRY_COUNT	= 1 << 13,
2888
	SRP_OPT_QUEUE_SIZE	= 1 << 14,
2889 2890 2891 2892 2893 2894 2895
	SRP_OPT_ALL		= (SRP_OPT_ID_EXT	|
				   SRP_OPT_IOC_GUID	|
				   SRP_OPT_DGID		|
				   SRP_OPT_PKEY		|
				   SRP_OPT_SERVICE_ID),
};

2896
static const match_table_t srp_opt_tokens = {
2897 2898 2899 2900 2901 2902 2903
	{ 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" 	},
2904
	{ SRP_OPT_IO_CLASS,		"io_class=%x"		},
2905
	{ SRP_OPT_INITIATOR_EXT,	"initiator_ext=%s"	},
2906
	{ SRP_OPT_CMD_SG_ENTRIES,	"cmd_sg_entries=%u"	},
2907 2908
	{ SRP_OPT_ALLOW_EXT_SG,		"allow_ext_sg=%u"	},
	{ SRP_OPT_SG_TABLESIZE,		"sg_tablesize=%u"	},
2909
	{ SRP_OPT_COMP_VECTOR,		"comp_vector=%u"	},
2910
	{ SRP_OPT_TL_RETRY_COUNT,	"tl_retry_count=%u"	},
2911
	{ SRP_OPT_QUEUE_SIZE,		"queue_size=%d"		},
2912
	{ SRP_OPT_ERR,			NULL 			}
2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930
};

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;
2931
	while ((p = strsep(&sep_opt, ",\n")) != NULL) {
2932 2933 2934 2935 2936 2937 2938 2939 2940
		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);
2941 2942 2943 2944
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2945 2946 2947 2948 2949 2950
			target->id_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_IOC_GUID:
			p = match_strdup(args);
2951 2952 2953 2954
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2955 2956 2957 2958 2959 2960
			target->ioc_guid = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_DGID:
			p = match_strdup(args);
2961 2962 2963 2964
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2965
			if (strlen(p) != 32) {
2966
				pr_warn("bad dest GID parameter '%s'\n", p);
2967
				kfree(p);
2968 2969 2970 2971
				goto out;
			}

			for (i = 0; i < 16; ++i) {
2972 2973 2974 2975 2976 2977 2978
				strlcpy(dgid, p + i * 2, sizeof(dgid));
				if (sscanf(dgid, "%hhx",
					   &target->orig_dgid.raw[i]) < 1) {
					ret = -EINVAL;
					kfree(p);
					goto out;
				}
2979
			}
2980
			kfree(p);
2981 2982 2983 2984
			break;

		case SRP_OPT_PKEY:
			if (match_hex(args, &token)) {
2985
				pr_warn("bad P_Key parameter '%s'\n", p);
2986 2987
				goto out;
			}
2988
			target->pkey = cpu_to_be16(token);
2989 2990 2991 2992
			break;

		case SRP_OPT_SERVICE_ID:
			p = match_strdup(args);
2993 2994 2995 2996
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2997 2998 2999 3000 3001 3002
			target->service_id = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_MAX_SECT:
			if (match_int(args, &token)) {
3003
				pr_warn("bad max sect parameter '%s'\n", p);
3004 3005 3006 3007 3008
				goto out;
			}
			target->scsi_host->max_sectors = token;
			break;

3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020
		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;

3021
		case SRP_OPT_MAX_CMD_PER_LUN:
3022
			if (match_int(args, &token) || token < 1) {
3023 3024
				pr_warn("bad max cmd_per_lun parameter '%s'\n",
					p);
3025 3026
				goto out;
			}
3027
			target->scsi_host->cmd_per_lun = token;
3028 3029
			break;

3030 3031
		case SRP_OPT_IO_CLASS:
			if (match_hex(args, &token)) {
3032
				pr_warn("bad IO class parameter '%s'\n", p);
3033 3034 3035 3036
				goto out;
			}
			if (token != SRP_REV10_IB_IO_CLASS &&
			    token != SRP_REV16A_IB_IO_CLASS) {
3037 3038 3039
				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);
3040 3041 3042 3043 3044
				goto out;
			}
			target->io_class = token;
			break;

3045 3046
		case SRP_OPT_INITIATOR_EXT:
			p = match_strdup(args);
3047 3048 3049 3050
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
3051 3052 3053 3054
			target->initiator_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

3055 3056
		case SRP_OPT_CMD_SG_ENTRIES:
			if (match_int(args, &token) || token < 1 || token > 255) {
3057 3058
				pr_warn("bad max cmd_sg_entries parameter '%s'\n",
					p);
3059 3060 3061 3062 3063
				goto out;
			}
			target->cmd_sg_cnt = token;
			break;

3064 3065
		case SRP_OPT_ALLOW_EXT_SG:
			if (match_int(args, &token)) {
3066
				pr_warn("bad allow_ext_sg parameter '%s'\n", p);
3067 3068 3069 3070 3071 3072 3073 3074
				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) {
3075 3076
				pr_warn("bad max sg_tablesize parameter '%s'\n",
					p);
3077 3078 3079 3080 3081
				goto out;
			}
			target->sg_tablesize = token;
			break;

3082 3083 3084 3085 3086 3087 3088 3089
		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;

3090 3091 3092 3093 3094 3095 3096 3097 3098
		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;

3099
		default:
3100 3101
			pr_warn("unknown parameter or missing value '%s' in target creation request\n",
				p);
3102 3103 3104 3105 3106 3107 3108 3109 3110 3111
			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))
3112 3113
				pr_warn("target creation request is missing parameter '%s'\n",
					srp_opt_tokens[i].pattern);
3114

3115 3116 3117 3118 3119 3120
	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);

3121 3122 3123 3124 3125
out:
	kfree(options);
	return ret;
}

3126 3127
static ssize_t srp_create_target(struct device *dev,
				 struct device_attribute *attr,
3128 3129 3130
				 const char *buf, size_t count)
{
	struct srp_host *host =
3131
		container_of(dev, struct srp_host, dev);
3132 3133
	struct Scsi_Host *target_host;
	struct srp_target_port *target;
3134
	struct srp_rdma_ch *ch;
3135 3136
	struct srp_device *srp_dev = host->srp_dev;
	struct ib_device *ibdev = srp_dev->dev;
B
Bart Van Assche 已提交
3137 3138
	int ret, node_idx, node, cpu, i;
	bool multich = false;
3139 3140 3141 3142 3143 3144

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

3145
	target_host->transportt  = ib_srp_transport_template;
3146 3147
	target_host->max_channel = 0;
	target_host->max_id      = 1;
A
Arne Redlich 已提交
3148 3149
	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
3150

3151 3152
	target = host_to_target(target_host);

3153 3154 3155 3156 3157 3158
	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;
3159 3160
	target->sg_tablesize	= indirect_sg_entries ? : cmd_sg_entries;
	target->allow_ext_sg	= allow_ext_sg;
3161
	target->tl_retry_count	= 7;
3162
	target->queue_size	= SRP_DEFAULT_QUEUE_SIZE;
3163

3164 3165 3166 3167 3168 3169
	/*
	 * Avoid that the SCSI host can be removed by srp_remove_target()
	 * before this function returns.
	 */
	scsi_host_get(target->scsi_host);

3170 3171
	mutex_lock(&host->add_target_mutex);

3172 3173
	ret = srp_parse_options(buf, target);
	if (ret)
3174
		goto out;
3175

B
Bart Van Assche 已提交
3176 3177
	ret = scsi_init_shared_tag_map(target_host, target_host->can_queue);
	if (ret)
3178
		goto out;
B
Bart Van Assche 已提交
3179

3180 3181
	target->req_ring_size = target->queue_size - SRP_TSK_MGMT_SQ_SIZE;

3182 3183 3184 3185 3186 3187 3188
	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;
3189
		goto out;
3190 3191
	}

3192
	if (!srp_dev->has_fmr && !srp_dev->has_fr && !target->allow_ext_sg &&
3193
	    target->cmd_sg_cnt < target->sg_tablesize) {
3194
		pr_warn("No MR pool and no external indirect descriptors, limiting sg_tablesize to cmd_sg_cnt\n");
3195 3196 3197 3198 3199 3200
		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);
3201 3202 3203 3204
	target->max_iu_len = sizeof (struct srp_cmd) +
			     sizeof (struct srp_indirect_buf) +
			     target->cmd_sg_cnt * sizeof (struct srp_direct_buf);

3205
	INIT_WORK(&target->tl_err_work, srp_tl_err_work);
3206
	INIT_WORK(&target->remove_work, srp_remove_work);
3207
	spin_lock_init(&target->lock);
3208
	ret = ib_query_gid(ibdev, host->port, 0, &target->sgid);
3209
	if (ret)
3210
		goto out;
3211

B
Bart Van Assche 已提交
3212 3213 3214 3215 3216 3217 3218 3219 3220
	ret = -ENOMEM;
	target->ch_count = max_t(unsigned, num_online_nodes(),
				 min(ch_count ? :
				     min(4 * num_online_nodes(),
					 ibdev->num_comp_vectors),
				     num_online_cpus()));
	target->ch = kcalloc(target->ch_count, sizeof(*target->ch),
			     GFP_KERNEL);
	if (!target->ch)
3221
		goto out;
3222

B
Bart Van Assche 已提交
3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250
	node_idx = 0;
	for_each_online_node(node) {
		const int ch_start = (node_idx * target->ch_count /
				      num_online_nodes());
		const int ch_end = ((node_idx + 1) * target->ch_count /
				    num_online_nodes());
		const int cv_start = (node_idx * ibdev->num_comp_vectors /
				      num_online_nodes() + target->comp_vector)
				     % ibdev->num_comp_vectors;
		const int cv_end = ((node_idx + 1) * ibdev->num_comp_vectors /
				    num_online_nodes() + target->comp_vector)
				   % ibdev->num_comp_vectors;
		int cpu_idx = 0;

		for_each_online_cpu(cpu) {
			if (cpu_to_node(cpu) != node)
				continue;
			if (ch_start + cpu_idx >= ch_end)
				continue;
			ch = &target->ch[ch_start + cpu_idx];
			ch->target = target;
			ch->comp_vector = cv_start == cv_end ? cv_start :
				cv_start + cpu_idx % (cv_end - cv_start);
			spin_lock_init(&ch->lock);
			INIT_LIST_HEAD(&ch->free_tx);
			ret = srp_new_cm_id(ch);
			if (ret)
				goto err_disconnect;
3251

B
Bart Van Assche 已提交
3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279
			ret = srp_create_ch_ib(ch);
			if (ret)
				goto err_disconnect;

			ret = srp_alloc_req_data(ch);
			if (ret)
				goto err_disconnect;

			ret = srp_connect_ch(ch, multich);
			if (ret) {
				shost_printk(KERN_ERR, target->scsi_host,
					     PFX "Connection %d/%d failed\n",
					     ch_start + cpu_idx,
					     target->ch_count);
				if (node_idx == 0 && cpu_idx == 0) {
					goto err_disconnect;
				} else {
					srp_free_ch_ib(target, ch);
					srp_free_req_data(target, ch);
					target->ch_count = ch - target->ch;
					break;
				}
			}

			multich = true;
			cpu_idx++;
		}
		node_idx++;
3280 3281
	}

B
Bart Van Assche 已提交
3282 3283
	target->scsi_host->nr_hw_queues = target->ch_count;

3284 3285 3286 3287
	ret = srp_add_target(host, target);
	if (ret)
		goto err_disconnect;

3288 3289 3290 3291 3292
	if (target->state != SRP_TARGET_REMOVED) {
		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),
3293
			     be16_to_cpu(target->pkey),
3294
			     be64_to_cpu(target->service_id),
3295
			     target->sgid.raw, target->orig_dgid.raw);
3296
	}
B
Bart Van Assche 已提交
3297

3298 3299 3300 3301
	ret = count;

out:
	mutex_unlock(&host->add_target_mutex);
3302 3303 3304

	scsi_host_put(target->scsi_host);

3305
	return ret;
3306 3307 3308 3309

err_disconnect:
	srp_disconnect_target(target);

B
Bart Van Assche 已提交
3310 3311 3312 3313 3314
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		srp_free_ch_ib(target, ch);
		srp_free_req_data(target, ch);
	}
3315

B
Bart Van Assche 已提交
3316
	kfree(target->ch);
3317
	goto out;
3318 3319
}

3320
static DEVICE_ATTR(add_target, S_IWUSR, NULL, srp_create_target);
3321

3322 3323
static ssize_t show_ibdev(struct device *dev, struct device_attribute *attr,
			  char *buf)
3324
{
3325
	struct srp_host *host = container_of(dev, struct srp_host, dev);
3326

3327
	return sprintf(buf, "%s\n", host->srp_dev->dev->name);
3328 3329
}

3330
static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
3331

3332 3333
static ssize_t show_port(struct device *dev, struct device_attribute *attr,
			 char *buf)
3334
{
3335
	struct srp_host *host = container_of(dev, struct srp_host, dev);
3336 3337 3338 3339

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

3340
static DEVICE_ATTR(port, S_IRUGO, show_port, NULL);
3341

3342
static struct srp_host *srp_add_port(struct srp_device *device, u8 port)
3343 3344 3345 3346 3347 3348 3349 3350
{
	struct srp_host *host;

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

	INIT_LIST_HEAD(&host->target_list);
3351
	spin_lock_init(&host->target_lock);
3352
	init_completion(&host->released);
3353
	mutex_init(&host->add_target_mutex);
3354
	host->srp_dev = device;
3355 3356
	host->port = port;

3357 3358
	host->dev.class = &srp_class;
	host->dev.parent = device->dev->dma_device;
3359
	dev_set_name(&host->dev, "srp-%s-%d", device->dev->name, port);
3360

3361
	if (device_register(&host->dev))
3362
		goto free_host;
3363
	if (device_create_file(&host->dev, &dev_attr_add_target))
3364
		goto err_class;
3365
	if (device_create_file(&host->dev, &dev_attr_ibdev))
3366
		goto err_class;
3367
	if (device_create_file(&host->dev, &dev_attr_port))
3368 3369 3370 3371 3372
		goto err_class;

	return host;

err_class:
3373
	device_unregister(&host->dev);
3374

3375
free_host:
3376 3377 3378 3379 3380 3381 3382
	kfree(host);

	return NULL;
}

static void srp_add_one(struct ib_device *device)
{
3383 3384
	struct srp_device *srp_dev;
	struct ib_device_attr *dev_attr;
3385
	struct srp_host *host;
3386 3387
	int mr_page_shift, s, e, p;
	u64 max_pages_per_mr;
3388

3389 3390
	dev_attr = kmalloc(sizeof *dev_attr, GFP_KERNEL);
	if (!dev_attr)
3391
		return;
3392

3393
	if (ib_query_device(device, dev_attr)) {
3394
		pr_warn("Query device failed for %s\n", device->name);
3395 3396 3397 3398 3399 3400 3401
		goto free_attr;
	}

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

3402 3403
	srp_dev->has_fmr = (device->alloc_fmr && device->dealloc_fmr &&
			    device->map_phys_fmr && device->unmap_fmr);
3404 3405 3406 3407 3408 3409 3410
	srp_dev->has_fr = (dev_attr->device_cap_flags &
			   IB_DEVICE_MEM_MGT_EXTENSIONS);
	if (!srp_dev->has_fmr && !srp_dev->has_fr)
		dev_warn(&device->dev, "neither FMR nor FR is supported\n");

	srp_dev->use_fast_reg = (srp_dev->has_fr &&
				 (!srp_dev->has_fmr || prefer_fr));
3411

3412 3413
	/*
	 * Use the smallest page size supported by the HCA, down to a
3414 3415
	 * minimum of 4096 bytes. We're unlikely to build large sglists
	 * out of smaller entries.
3416
	 */
3417 3418 3419 3420 3421 3422 3423
	mr_page_shift		= max(12, ffs(dev_attr->page_size_cap) - 1);
	srp_dev->mr_page_size	= 1 << mr_page_shift;
	srp_dev->mr_page_mask	= ~((u64) srp_dev->mr_page_size - 1);
	max_pages_per_mr	= dev_attr->max_mr_size;
	do_div(max_pages_per_mr, srp_dev->mr_page_size);
	srp_dev->max_pages_per_mr = min_t(u64, SRP_MAX_PAGES_PER_MR,
					  max_pages_per_mr);
3424 3425 3426 3427 3428
	if (srp_dev->use_fast_reg) {
		srp_dev->max_pages_per_mr =
			min_t(u32, srp_dev->max_pages_per_mr,
			      dev_attr->max_fast_reg_page_list_len);
	}
3429 3430
	srp_dev->mr_max_size	= srp_dev->mr_page_size *
				   srp_dev->max_pages_per_mr;
3431
	pr_debug("%s: mr_page_shift = %d, dev_attr->max_mr_size = %#llx, dev_attr->max_fast_reg_page_list_len = %u, max_pages_per_mr = %d, mr_max_size = %#x\n",
3432
		 device->name, mr_page_shift, dev_attr->max_mr_size,
3433
		 dev_attr->max_fast_reg_page_list_len,
3434
		 srp_dev->max_pages_per_mr, srp_dev->mr_max_size);
3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449

	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;

T
Tom Tucker 已提交
3450
	if (device->node_type == RDMA_NODE_IB_SWITCH) {
3451 3452 3453 3454 3455 3456 3457 3458
		s = 0;
		e = 0;
	} else {
		s = 1;
		e = device->phys_port_cnt;
	}

	for (p = s; p <= e; ++p) {
3459
		host = srp_add_port(srp_dev, p);
3460
		if (host)
3461
			list_add_tail(&host->list, &srp_dev->dev_list);
3462 3463
	}

3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475
	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);
3476 3477 3478 3479
}

static void srp_remove_one(struct ib_device *device)
{
3480
	struct srp_device *srp_dev;
3481
	struct srp_host *host, *tmp_host;
3482
	struct srp_target_port *target;
3483

3484
	srp_dev = ib_get_client_data(device, &srp_client);
3485 3486
	if (!srp_dev)
		return;
3487

3488
	list_for_each_entry_safe(host, tmp_host, &srp_dev->dev_list, list) {
3489
		device_unregister(&host->dev);
3490 3491 3492 3493 3494 3495 3496
		/*
		 * Wait for the sysfs entry to go away, so that no new
		 * target ports can be created.
		 */
		wait_for_completion(&host->released);

		/*
3497
		 * Remove all target ports.
3498
		 */
3499
		spin_lock(&host->target_lock);
3500 3501
		list_for_each_entry(target, &host->target_list, list)
			srp_queue_remove_work(target);
3502
		spin_unlock(&host->target_lock);
3503 3504

		/*
3505
		 * Wait for tl_err and target port removal tasks.
3506
		 */
3507
		flush_workqueue(system_long_wq);
3508
		flush_workqueue(srp_remove_wq);
3509 3510 3511 3512

		kfree(host);
	}

3513 3514 3515 3516
	ib_dereg_mr(srp_dev->mr);
	ib_dealloc_pd(srp_dev->pd);

	kfree(srp_dev);
3517 3518
}

3519
static struct srp_function_template ib_srp_transport_functions = {
3520 3521
	.has_rport_state	 = true,
	.reset_timer_if_blocked	 = true,
3522
	.reconnect_delay	 = &srp_reconnect_delay,
3523 3524 3525
	.fast_io_fail_tmo	 = &srp_fast_io_fail_tmo,
	.dev_loss_tmo		 = &srp_dev_loss_tmo,
	.reconnect		 = srp_rport_reconnect,
3526
	.rport_delete		 = srp_rport_delete,
3527
	.terminate_rport_io	 = srp_terminate_io,
3528 3529
};

3530 3531 3532 3533
static int __init srp_init_module(void)
{
	int ret;

3534
	BUILD_BUG_ON(FIELD_SIZEOF(struct ib_wc, wr_id) < sizeof(void *));
3535

3536
	if (srp_sg_tablesize) {
3537
		pr_warn("srp_sg_tablesize is deprecated, please use cmd_sg_entries\n");
3538 3539 3540 3541 3542 3543 3544 3545
		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) {
3546
		pr_warn("Clamping cmd_sg_entries to 255\n");
3547
		cmd_sg_entries = 255;
3548 3549
	}

3550 3551 3552
	if (!indirect_sg_entries)
		indirect_sg_entries = cmd_sg_entries;
	else if (indirect_sg_entries < cmd_sg_entries) {
3553 3554
		pr_warn("Bumping up indirect_sg_entries to match cmd_sg_entries (%u)\n",
			cmd_sg_entries);
3555 3556 3557
		indirect_sg_entries = cmd_sg_entries;
	}

3558
	srp_remove_wq = create_workqueue("srp_remove");
3559 3560
	if (!srp_remove_wq) {
		ret = -ENOMEM;
3561 3562 3563 3564
		goto out;
	}

	ret = -ENOMEM;
3565 3566 3567
	ib_srp_transport_template =
		srp_attach_transport(&ib_srp_transport_functions);
	if (!ib_srp_transport_template)
3568
		goto destroy_wq;
3569

3570 3571
	ret = class_register(&srp_class);
	if (ret) {
3572
		pr_err("couldn't register class infiniband_srp\n");
3573
		goto release_tr;
3574 3575
	}

3576 3577
	ib_sa_register_client(&srp_sa_client);

3578 3579
	ret = ib_register_client(&srp_client);
	if (ret) {
3580
		pr_err("couldn't register IB client\n");
3581
		goto unreg_sa;
3582 3583
	}

3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596
out:
	return ret;

unreg_sa:
	ib_sa_unregister_client(&srp_sa_client);
	class_unregister(&srp_class);

release_tr:
	srp_release_transport(ib_srp_transport_template);

destroy_wq:
	destroy_workqueue(srp_remove_wq);
	goto out;
3597 3598 3599 3600 3601
}

static void __exit srp_cleanup_module(void)
{
	ib_unregister_client(&srp_client);
3602
	ib_sa_unregister_client(&srp_sa_client);
3603
	class_unregister(&srp_class);
3604
	srp_release_transport(ib_srp_transport_template);
3605
	destroy_workqueue(srp_remove_wq);
3606 3607 3608 3609
}

module_init(srp_init_module);
module_exit(srp_cleanup_module);