ib_srp.c 94.0 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

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

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

#include "ib_srp.h"

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

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

65 66
static unsigned int srp_sg_tablesize;
static unsigned int cmd_sg_entries;
67 68
static unsigned int indirect_sg_entries;
static bool allow_ext_sg;
69
static bool prefer_fr;
70
static bool register_always;
71
static int topspin_workarounds = 1;
72

73 74
module_param(srp_sg_tablesize, uint, 0444);
MODULE_PARM_DESC(srp_sg_tablesize, "Deprecated name for cmd_sg_entries");
75

76 77 78
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)");
79

80 81 82 83 84 85 86 87
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)");

88 89 90 91
module_param(topspin_workarounds, int, 0444);
MODULE_PARM_DESC(topspin_workarounds,
		 "Enable workarounds for Topspin/Cisco SRP target bugs if != 0");

92 93 94 95
module_param(prefer_fr, bool, 0444);
MODULE_PARM_DESC(prefer_fr,
"Whether to use fast registration if both FMR and fast registration are supported");

96 97 98 99
module_param(register_always, bool, 0444);
MODULE_PARM_DESC(register_always,
		 "Use memory registration even for contiguous memory regions");

100 101
static struct kernel_param_ops srp_tmo_ops;

102 103 104 105 106
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");

107 108 109 110 111 112 113 114
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.");

115
static int srp_dev_loss_tmo = 600;
116 117 118 119 120 121 122 123 124 125
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
126 127 128 129 130
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.");

131 132
static void srp_add_one(struct ib_device *device);
static void srp_remove_one(struct ib_device *device);
133 134
static void srp_recv_completion(struct ib_cq *cq, void *ch_ptr);
static void srp_send_completion(struct ib_cq *cq, void *ch_ptr);
135 136
static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event);

137
static struct scsi_transport_template *ib_srp_transport_template;
138
static struct workqueue_struct *srp_remove_wq;
139

140 141 142 143 144 145
static struct ib_client srp_client = {
	.name   = "srp",
	.add    = srp_add_one,
	.remove = srp_remove_one
};

146 147
static struct ib_sa_client srp_sa_client;

148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168
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;
	}
169 170 171 172 173
	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);
174
	else
175 176
		res = srp_tmo_valid(srp_reconnect_delay, srp_fast_io_fail_tmo,
				    tmo);
177 178 179 180 181 182 183 184 185 186 187 188 189
	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,
};

190 191 192 193 194 195 196 197 198 199
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;
}

200 201 202
static int srp_target_is_topspin(struct srp_target_port *target)
{
	static const u8 topspin_oui[3] = { 0x00, 0x05, 0xad };
203
	static const u8 cisco_oui[3]   = { 0x00, 0x1b, 0x0d };
204 205

	return topspin_workarounds &&
206 207
		(!memcmp(&target->ioc_guid, topspin_oui, sizeof topspin_oui) ||
		 !memcmp(&target->ioc_guid, cisco_oui, sizeof cisco_oui));
208 209
}

210 211 212 213 214 215 216 217 218 219 220 221 222 223
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;

224 225 226
	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))
227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246
		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;

247 248
	ib_dma_unmap_single(host->srp_dev->dev, iu->dma, iu->size,
			    iu->direction);
249 250 251 252 253 254
	kfree(iu->buf);
	kfree(iu);
}

static void srp_qp_event(struct ib_event *event, void *context)
{
255
	pr_debug("QP event %d\n", event->event);
256 257 258 259 260 261 262 263 264 265 266 267
}

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;

268 269
	ret = ib_find_pkey(target->srp_host->srp_dev->dev,
			   target->srp_host->port,
270
			   be16_to_cpu(target->pkey),
271
			   &attr->pkey_index);
272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290
	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;
}

291
static int srp_new_cm_id(struct srp_rdma_ch *ch)
D
David Dillow 已提交
292
{
293
	struct srp_target_port *target = ch->target;
D
David Dillow 已提交
294 295
	struct ib_cm_id *new_cm_id;

296
	new_cm_id = ib_create_cm_id(target->srp_host->srp_dev->dev,
297
				    srp_cm_handler, ch);
D
David Dillow 已提交
298 299 300
	if (IS_ERR(new_cm_id))
		return PTR_ERR(new_cm_id);

301 302 303 304 305 306 307
	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
308 309 310 311

	return 0;
}

312 313 314 315 316 317 318 319 320
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;
321 322
	fmr_param.max_pages_per_fmr = dev->max_pages_per_mr;
	fmr_param.page_shift	    = ilog2(dev->mr_page_size);
323 324 325 326 327 328 329
	fmr_param.access	    = (IB_ACCESS_LOCAL_WRITE |
				       IB_ACCESS_REMOTE_WRITE |
				       IB_ACCESS_REMOTE_READ);

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

330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455
/**
 * 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);
}

456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490
/**
 * 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)
{
	struct srp_target_port *target = ch->target;
	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 */
	WARN_ON_ONCE(target->connected);

	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;
}

813 814 815 816 817 818 819 820 821 822 823 824 825 826 827
static bool srp_change_conn_state(struct srp_target_port *target,
				  bool connected)
{
	bool changed = false;

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

	return changed;
}

828 829
static void srp_disconnect_target(struct srp_target_port *target)
{
B
Bart Van Assche 已提交
830 831
	struct srp_rdma_ch *ch;
	int i;
832

833 834
	if (srp_change_conn_state(target, false)) {
		/* XXX should send SRP_I_LOGOUT request */
835

B
Bart Van Assche 已提交
836 837 838 839 840 841
		for (i = 0; i < target->ch_count; i++) {
			ch = &target->ch[i];
			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");
			}
842
		}
843
	}
844 845
}

846 847
static void srp_free_req_data(struct srp_target_port *target,
			      struct srp_rdma_ch *ch)
848
{
849 850
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = dev->dev;
851 852 853
	struct srp_request *req;
	int i;

B
Bart Van Assche 已提交
854
	if (!ch->target || !ch->req_ring)
855 856 857
		return;

	for (i = 0; i < target->req_ring_size; ++i) {
858
		req = &ch->req_ring[i];
859 860 861 862
		if (dev->use_fast_reg)
			kfree(req->fr_list);
		else
			kfree(req->fmr_list);
863
		kfree(req->map_page);
864 865 866 867 868 869
		if (req->indirect_dma_addr) {
			ib_dma_unmap_single(ibdev, req->indirect_dma_addr,
					    target->indirect_size,
					    DMA_TO_DEVICE);
		}
		kfree(req->indirect_desc);
870
	}
871

872 873
	kfree(ch->req_ring);
	ch->req_ring = NULL;
874 875
}

876
static int srp_alloc_req_data(struct srp_rdma_ch *ch)
877
{
878
	struct srp_target_port *target = ch->target;
879 880 881
	struct srp_device *srp_dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = srp_dev->dev;
	struct srp_request *req;
882
	void *mr_list;
883 884 885
	dma_addr_t dma_addr;
	int i, ret = -ENOMEM;

886 887 888
	ch->req_ring = kcalloc(target->req_ring_size, sizeof(*ch->req_ring),
			       GFP_KERNEL);
	if (!ch->req_ring)
889 890 891
		goto out;

	for (i = 0; i < target->req_ring_size; ++i) {
892
		req = &ch->req_ring[i];
893 894 895 896 897 898 899 900
		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;
901
		req->map_page = kmalloc(srp_dev->max_pages_per_mr *
902
					sizeof(void *), GFP_KERNEL);
903 904
		if (!req->map_page)
			goto out;
905
		req->indirect_desc = kmalloc(target->indirect_size, GFP_KERNEL);
906
		if (!req->indirect_desc)
907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922
			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;
}

923 924 925 926 927 928 929 930 931 932 933 934 935 936 937
/**
 * 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);
}

938 939
static void srp_remove_target(struct srp_target_port *target)
{
B
Bart Van Assche 已提交
940 941
	struct srp_rdma_ch *ch;
	int i;
942

943 944
	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);

945
	srp_del_scsi_host_attr(target->scsi_host);
946
	srp_rport_get(target->rport);
947 948
	srp_remove_host(target->scsi_host);
	scsi_remove_host(target->scsi_host);
949
	srp_stop_rport_timers(target->rport);
950
	srp_disconnect_target(target);
B
Bart Van Assche 已提交
951 952 953 954
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		srp_free_ch_ib(target, ch);
	}
955
	cancel_work_sync(&target->tl_err_work);
956
	srp_rport_put(target->rport);
B
Bart Van Assche 已提交
957 958 959 960 961 962
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		srp_free_req_data(target, ch);
	}
	kfree(target->ch);
	target->ch = NULL;
963 964 965 966 967

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

968 969 970
	scsi_host_put(target->scsi_host);
}

D
David Howells 已提交
971
static void srp_remove_work(struct work_struct *work)
972
{
D
David Howells 已提交
973
	struct srp_target_port *target =
974
		container_of(work, struct srp_target_port, remove_work);
975

976
	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);
977

978
	srp_remove_target(target);
979 980
}

981 982 983 984 985 986 987
static void srp_rport_delete(struct srp_rport *rport)
{
	struct srp_target_port *target = rport->lld_data;

	srp_queue_remove_work(target);
}

B
Bart Van Assche 已提交
988
static int srp_connect_ch(struct srp_rdma_ch *ch, bool multich)
989
{
990
	struct srp_target_port *target = ch->target;
991 992
	int ret;

B
Bart Van Assche 已提交
993
	WARN_ON_ONCE(!multich && target->connected);
994

995 996
	target->qp_in_error = false;

997
	ret = srp_lookup_path(ch);
998 999 1000 1001
	if (ret)
		return ret;

	while (1) {
1002
		init_completion(&ch->done);
B
Bart Van Assche 已提交
1003
		ret = srp_send_req(ch, multich);
1004 1005
		if (ret)
			return ret;
1006
		ret = wait_for_completion_interruptible(&ch->done);
1007 1008
		if (ret < 0)
			return ret;
1009 1010 1011 1012 1013 1014 1015

		/*
		 * 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.
		 */
1016
		switch (ch->status) {
1017
		case 0:
1018
			srp_change_conn_state(target, true);
1019 1020 1021
			return 0;

		case SRP_PORT_REDIRECT:
1022
			ret = srp_lookup_path(ch);
1023 1024 1025 1026 1027 1028 1029
			if (ret)
				return ret;
			break;

		case SRP_DLID_REDIRECT:
			break;

D
David Dillow 已提交
1030 1031
		case SRP_STALE_CONN:
			shost_printk(KERN_ERR, target->scsi_host, PFX
1032
				     "giving up on stale connection\n");
1033 1034
			ch->status = -ECONNRESET;
			return ch->status;
D
David Dillow 已提交
1035

1036
		default:
1037
			return ch->status;
1038 1039 1040 1041
		}
	}
}

1042
static int srp_inv_rkey(struct srp_rdma_ch *ch, u32 rkey)
1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053
{
	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,
	};

1054
	return ib_post_send(ch->qp, &wr, &bad_wr);
1055 1056
}

1057
static void srp_unmap_data(struct scsi_cmnd *scmnd,
1058
			   struct srp_rdma_ch *ch,
1059 1060
			   struct srp_request *req)
{
1061
	struct srp_target_port *target = ch->target;
1062 1063 1064
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = dev->dev;
	int i, res;
1065

1066
	if (!scsi_sglist(scmnd) ||
1067 1068 1069 1070
	    (scmnd->sc_data_direction != DMA_TO_DEVICE &&
	     scmnd->sc_data_direction != DMA_FROM_DEVICE))
		return;

1071 1072 1073 1074
	if (dev->use_fast_reg) {
		struct srp_fr_desc **pfr;

		for (i = req->nmdesc, pfr = req->fr_list; i > 0; i--, pfr++) {
1075
			res = srp_inv_rkey(ch, (*pfr)->mr->rkey);
1076 1077 1078 1079 1080 1081 1082 1083 1084
			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)
1085
			srp_fr_pool_put(ch->fr_pool, req->fr_list,
1086 1087 1088 1089 1090 1091 1092
					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);
	}
1093

1094 1095
	ib_dma_unmap_sg(ibdev, scsi_sglist(scmnd), scsi_sg_count(scmnd),
			scmnd->sc_data_direction);
1096 1097
}

B
Bart Van Assche 已提交
1098 1099
/**
 * srp_claim_req - Take ownership of the scmnd associated with a request.
1100
 * @ch: SRP RDMA channel.
B
Bart Van Assche 已提交
1101
 * @req: SRP request.
1102
 * @sdev: If not NULL, only take ownership for this SCSI device.
B
Bart Van Assche 已提交
1103 1104 1105 1106 1107 1108
 * @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.
 */
1109
static struct scsi_cmnd *srp_claim_req(struct srp_rdma_ch *ch,
B
Bart Van Assche 已提交
1110
				       struct srp_request *req,
1111
				       struct scsi_device *sdev,
B
Bart Van Assche 已提交
1112 1113 1114 1115
				       struct scsi_cmnd *scmnd)
{
	unsigned long flags;

1116
	spin_lock_irqsave(&ch->lock, flags);
1117 1118 1119
	if (req->scmnd &&
	    (!sdev || req->scmnd->device == sdev) &&
	    (!scmnd || req->scmnd == scmnd)) {
B
Bart Van Assche 已提交
1120 1121 1122 1123 1124
		scmnd = req->scmnd;
		req->scmnd = NULL;
	} else {
		scmnd = NULL;
	}
1125
	spin_unlock_irqrestore(&ch->lock, flags);
B
Bart Van Assche 已提交
1126 1127 1128 1129 1130 1131

	return scmnd;
}

/**
 * srp_free_req() - Unmap data and add request to the free request list.
1132
 * @ch:     SRP RDMA channel.
1133 1134 1135
 * @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
1136
 */
1137 1138
static void srp_free_req(struct srp_rdma_ch *ch, struct srp_request *req,
			 struct scsi_cmnd *scmnd, s32 req_lim_delta)
1139
{
1140 1141
	unsigned long flags;

1142
	srp_unmap_data(scmnd, ch, req);
B
Bart Van Assche 已提交
1143

1144 1145 1146
	spin_lock_irqsave(&ch->lock, flags);
	ch->req_lim += req_lim_delta;
	spin_unlock_irqrestore(&ch->lock, flags);
1147 1148
}

1149 1150
static void srp_finish_req(struct srp_rdma_ch *ch, struct srp_request *req,
			   struct scsi_device *sdev, int result)
1151
{
1152
	struct scsi_cmnd *scmnd = srp_claim_req(ch, req, sdev, NULL);
B
Bart Van Assche 已提交
1153 1154

	if (scmnd) {
1155
		srp_free_req(ch, req, scmnd, 0);
1156
		scmnd->result = result;
B
Bart Van Assche 已提交
1157 1158
		scmnd->scsi_done(scmnd);
	}
1159 1160
}

1161
static void srp_terminate_io(struct srp_rport *rport)
1162
{
1163
	struct srp_target_port *target = rport->lld_data;
B
Bart Van Assche 已提交
1164
	struct srp_rdma_ch *ch;
1165 1166
	struct Scsi_Host *shost = target->scsi_host;
	struct scsi_device *sdev;
B
Bart Van Assche 已提交
1167
	int i, j;
1168

1169 1170 1171 1172 1173 1174 1175
	/*
	 * 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
1176 1177
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
1178

B
Bart Van Assche 已提交
1179 1180 1181 1182 1183 1184
		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);
		}
1185 1186
	}
}
1187

1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199
/*
 * 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
1200 1201 1202
	struct srp_rdma_ch *ch;
	int i, j, ret = 0;
	bool multich = false;
1203

1204
	srp_disconnect_target(target);
1205 1206 1207 1208

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

1209
	/*
1210 1211 1212
	 * 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.
1213
	 */
B
Bart Van Assche 已提交
1214 1215 1216 1217 1218
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		if (!ch->target)
			break;
		ret += srp_new_cm_id(ch);
1219
	}
B
Bart Van Assche 已提交
1220 1221 1222 1223 1224 1225
	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];
1226

B
Bart Van Assche 已提交
1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239
			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);
1240

B
Bart Van Assche 已提交
1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254
		INIT_LIST_HEAD(&ch->free_tx);
		for (j = 0; j < target->queue_size; ++j)
			list_add(&ch->tx_ring[j]->list, &ch->free_tx);
	}
	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;
	}
1255

1256 1257 1258
	if (ret == 0)
		shost_printk(KERN_INFO, target->scsi_host,
			     PFX "reconnect succeeded\n");
1259 1260 1261 1262

	return ret;
}

1263 1264
static void srp_map_desc(struct srp_map_state *state, dma_addr_t dma_addr,
			 unsigned int dma_len, u32 rkey)
1265
{
1266
	struct srp_direct_buf *desc = state->desc;
1267

1268 1269 1270
	desc->va = cpu_to_be64(dma_addr);
	desc->key = cpu_to_be32(rkey);
	desc->len = cpu_to_be32(dma_len);
1271

1272 1273 1274 1275
	state->total_len += dma_len;
	state->desc++;
	state->ndesc++;
}
1276

1277
static int srp_map_finish_fmr(struct srp_map_state *state,
1278
			      struct srp_rdma_ch *ch)
1279 1280 1281
{
	struct ib_pool_fmr *fmr;
	u64 io_addr = 0;
1282

1283
	fmr = ib_fmr_pool_map_phys(ch->fmr_pool, state->pages,
1284 1285 1286
				   state->npages, io_addr);
	if (IS_ERR(fmr))
		return PTR_ERR(fmr);
1287

1288
	*state->next_fmr++ = fmr;
1289
	state->nmdesc++;
1290

1291
	srp_map_desc(state, 0, state->dma_len, fmr->fmr->rkey);
1292

1293 1294 1295
	return 0;
}

1296
static int srp_map_finish_fr(struct srp_map_state *state,
1297
			     struct srp_rdma_ch *ch)
1298
{
1299
	struct srp_target_port *target = ch->target;
1300 1301 1302 1303 1304 1305
	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;

1306
	desc = srp_fr_pool_get(ch->fr_pool);
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 1334
	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);

1335
	return ib_post_send(ch->qp, &wr, &bad_wr);
1336 1337
}

1338
static int srp_finish_mapping(struct srp_map_state *state,
1339
			      struct srp_rdma_ch *ch)
1340
{
1341
	struct srp_target_port *target = ch->target;
1342 1343 1344 1345 1346
	int ret = 0;

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

1347
	if (state->npages == 1 && !register_always)
1348
		srp_map_desc(state, state->base_dma_addr, state->dma_len,
1349 1350
			     target->rkey);
	else
1351
		ret = target->srp_host->srp_dev->use_fast_reg ?
1352 1353
			srp_map_finish_fr(state, ch) :
			srp_map_finish_fmr(state, ch);
1354 1355 1356

	if (ret == 0) {
		state->npages = 0;
1357
		state->dma_len = 0;
1358 1359 1360 1361 1362
	}

	return ret;
}

1363 1364 1365 1366 1367 1368 1369 1370
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;
}
1371

1372
static int srp_map_sg_entry(struct srp_map_state *state,
1373
			    struct srp_rdma_ch *ch,
1374
			    struct scatterlist *sg, int sg_index,
1375
			    bool use_mr)
1376
{
1377
	struct srp_target_port *target = ch->target;
1378 1379 1380 1381 1382 1383 1384 1385 1386 1387
	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;

1388 1389 1390 1391
	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
1392 1393 1394 1395
		 * other than the descriptor.
		 */
		srp_map_desc(state, dma_addr, dma_len, target->rkey);
		return 0;
1396
	}
1397

1398 1399 1400 1401 1402
	/*
	 * 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.
1403
	 */
1404 1405
	if ((!dev->use_fast_reg && dma_addr & ~dev->mr_page_mask) ||
	    dma_len > dev->mr_max_size) {
1406
		ret = srp_finish_mapping(state, ch);
1407 1408 1409 1410 1411 1412
		if (ret)
			return ret;

		srp_map_desc(state, dma_addr, dma_len, target->rkey);
		srp_map_update_start(state, NULL, 0, 0);
		return 0;
1413 1414
	}

1415 1416 1417 1418 1419
	/*
	 * 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.
1420 1421 1422
	 */
	if (!state->unmapped_sg)
		srp_map_update_start(state, sg, sg_index, dma_addr);
1423

1424
	while (dma_len) {
1425 1426
		unsigned offset = dma_addr & ~dev->mr_page_mask;
		if (state->npages == dev->max_pages_per_mr || offset != 0) {
1427
			ret = srp_finish_mapping(state, ch);
1428 1429
			if (ret)
				return ret;
1430

1431 1432 1433
			srp_map_update_start(state, sg, sg_index, dma_addr);
		}

1434
		len = min_t(unsigned int, dma_len, dev->mr_page_size - offset);
1435

1436 1437
		if (!state->npages)
			state->base_dma_addr = dma_addr;
1438
		state->pages[state->npages++] = dma_addr & dev->mr_page_mask;
1439
		state->dma_len += len;
1440 1441 1442 1443
		dma_addr += len;
		dma_len -= len;
	}

1444 1445
	/*
	 * If the last entry of the MR wasn't a full page, then we need to
1446 1447 1448 1449
	 * close it out and start a new one -- we can only merge at page
	 * boundries.
	 */
	ret = 0;
1450
	if (len != dev->mr_page_size) {
1451
		ret = srp_finish_mapping(state, ch);
1452 1453 1454
		if (!ret)
			srp_map_update_start(state, NULL, 0, 0);
	}
1455 1456 1457
	return ret;
}

1458 1459 1460
static int srp_map_sg(struct srp_map_state *state, struct srp_rdma_ch *ch,
		      struct srp_request *req, struct scatterlist *scat,
		      int count)
1461
{
1462
	struct srp_target_port *target = ch->target;
1463 1464 1465
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = dev->dev;
	struct scatterlist *sg;
1466 1467
	int i;
	bool use_mr;
1468 1469 1470

	state->desc	= req->indirect_desc;
	state->pages	= req->map_page;
1471 1472
	if (dev->use_fast_reg) {
		state->next_fr = req->fr_list;
1473
		use_mr = !!ch->fr_pool;
1474 1475
	} else {
		state->next_fmr = req->fmr_list;
1476
		use_mr = !!ch->fmr_pool;
1477
	}
1478 1479

	for_each_sg(scat, sg, count, i) {
1480
		if (srp_map_sg_entry(state, ch, sg, i, use_mr)) {
1481 1482 1483 1484
			/*
			 * Memory registration failed, so backtrack to the
			 * first unmapped entry and continue on without using
			 * memory registration.
1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496
			 */
			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;
1497
			use_mr = false;
1498 1499 1500 1501
			srp_map_desc(state, dma_addr, dma_len, target->rkey);
		}
	}

1502
	if (use_mr && srp_finish_mapping(state, ch))
1503 1504
		goto backtrack;

1505
	req->nmdesc = state->nmdesc;
1506 1507

	return 0;
1508 1509
}

1510
static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_rdma_ch *ch,
1511 1512
			struct srp_request *req)
{
1513
	struct srp_target_port *target = ch->target;
1514
	struct scatterlist *scat;
1515
	struct srp_cmd *cmd = req->cmd->buf;
1516
	int len, nents, count;
1517 1518
	struct srp_device *dev;
	struct ib_device *ibdev;
1519 1520 1521 1522
	struct srp_map_state state;
	struct srp_indirect_buf *indirect_hdr;
	u32 table_len;
	u8 fmt;
1523

1524
	if (!scsi_sglist(scmnd) || scmnd->sc_data_direction == DMA_NONE)
1525 1526 1527 1528
		return sizeof (struct srp_cmd);

	if (scmnd->sc_data_direction != DMA_FROM_DEVICE &&
	    scmnd->sc_data_direction != DMA_TO_DEVICE) {
1529 1530 1531
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled data direction %d\n",
			     scmnd->sc_data_direction);
1532 1533 1534
		return -EINVAL;
	}

1535 1536
	nents = scsi_sg_count(scmnd);
	scat  = scsi_sglist(scmnd);
1537

1538
	dev = target->srp_host->srp_dev;
1539 1540 1541
	ibdev = dev->dev;

	count = ib_dma_map_sg(ibdev, scat, nents, scmnd->sc_data_direction);
1542 1543
	if (unlikely(count == 0))
		return -EIO;
1544 1545 1546

	fmt = SRP_DATA_DESC_DIRECT;
	len = sizeof (struct srp_cmd) +	sizeof (struct srp_direct_buf);
1547

1548
	if (count == 1 && !register_always) {
1549 1550 1551 1552 1553 1554
		/*
		 * 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.
		 */
1555
		struct srp_direct_buf *buf = (void *) cmd->add_data;
1556

1557
		buf->va  = cpu_to_be64(ib_sg_dma_address(ibdev, scat));
1558
		buf->key = cpu_to_be32(target->rkey);
1559
		buf->len = cpu_to_be32(ib_sg_dma_len(ibdev, scat));
1560

1561
		req->nmdesc = 0;
1562 1563 1564
		goto map_complete;
	}

1565 1566 1567
	/*
	 * We have more than one scatter/gather entry, so build our indirect
	 * descriptor table, trying to merge as many entries as we can.
1568 1569 1570
	 */
	indirect_hdr = (void *) cmd->add_data;

1571 1572 1573
	ib_dma_sync_single_for_cpu(ibdev, req->indirect_dma_addr,
				   target->indirect_size, DMA_TO_DEVICE);

1574
	memset(&state, 0, sizeof(state));
1575
	srp_map_sg(&state, ch, req, scat, count);
1576

1577 1578 1579 1580 1581
	/* 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.
1582 1583
	 */
	if (state.ndesc == 1) {
1584 1585
		/*
		 * Memory registration collapsed the sg-list into one entry,
1586 1587 1588
		 * so use a direct descriptor.
		 */
		struct srp_direct_buf *buf = (void *) cmd->add_data;
1589

1590
		*buf = req->indirect_desc[0];
1591
		goto map_complete;
1592 1593
	}

1594 1595 1596 1597 1598 1599 1600 1601
	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);
1602 1603 1604 1605
	table_len = state.ndesc * sizeof (struct srp_direct_buf);

	fmt = SRP_DATA_DESC_INDIRECT;
	len = sizeof(struct srp_cmd) + sizeof (struct srp_indirect_buf);
1606
	len += count * sizeof (struct srp_direct_buf);
1607

1608 1609
	memcpy(indirect_hdr->desc_list, req->indirect_desc,
	       count * sizeof (struct srp_direct_buf));
1610

1611
	indirect_hdr->table_desc.va = cpu_to_be64(req->indirect_dma_addr);
1612 1613 1614 1615 1616
	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)
1617
		cmd->data_out_desc_cnt = count;
1618
	else
1619 1620 1621 1622
		cmd->data_in_desc_cnt = count;

	ib_dma_sync_single_for_device(ibdev, req->indirect_dma_addr, table_len,
				      DMA_TO_DEVICE);
1623 1624

map_complete:
1625 1626 1627 1628 1629 1630 1631 1632
	if (scmnd->sc_data_direction == DMA_TO_DEVICE)
		cmd->buf_fmt = fmt << 4;
	else
		cmd->buf_fmt = fmt;

	return len;
}

1633 1634 1635
/*
 * Return an IU and possible credit to the free pool
 */
1636
static void srp_put_tx_iu(struct srp_rdma_ch *ch, struct srp_iu *iu,
1637 1638 1639 1640
			  enum srp_iu_type iu_type)
{
	unsigned long flags;

1641 1642
	spin_lock_irqsave(&ch->lock, flags);
	list_add(&iu->list, &ch->free_tx);
1643
	if (iu_type != SRP_IU_RSP)
1644 1645
		++ch->req_lim;
	spin_unlock_irqrestore(&ch->lock, flags);
1646 1647
}

1648
/*
1649
 * Must be called with ch->lock held to protect req_lim and free_tx.
1650
 * If IU is not sent, it must be returned using srp_put_tx_iu().
1651 1652 1653 1654 1655 1656 1657 1658 1659 1660
 *
 * 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.
 */
1661
static struct srp_iu *__srp_get_tx_iu(struct srp_rdma_ch *ch,
1662 1663
				      enum srp_iu_type iu_type)
{
1664
	struct srp_target_port *target = ch->target;
1665 1666 1667
	s32 rsv = (iu_type == SRP_IU_TSK_MGMT) ? 0 : SRP_TSK_MGMT_SQ_SIZE;
	struct srp_iu *iu;

1668
	srp_send_completion(ch->send_cq, ch);
1669

1670
	if (list_empty(&ch->free_tx))
1671 1672 1673
		return NULL;

	/* Initiator responses to target requests do not consume credits */
1674
	if (iu_type != SRP_IU_RSP) {
1675
		if (ch->req_lim <= rsv) {
1676 1677 1678 1679
			++target->zero_req_lim;
			return NULL;
		}

1680
		--ch->req_lim;
1681 1682
	}

1683
	iu = list_first_entry(&ch->free_tx, struct srp_iu, list);
1684
	list_del(&iu->list);
1685 1686 1687
	return iu;
}

1688
static int srp_post_send(struct srp_rdma_ch *ch, struct srp_iu *iu, int len)
1689
{
1690
	struct srp_target_port *target = ch->target;
1691 1692 1693 1694 1695
	struct ib_sge list;
	struct ib_send_wr wr, *bad_wr;

	list.addr   = iu->dma;
	list.length = len;
1696
	list.lkey   = target->lkey;
1697 1698

	wr.next       = NULL;
1699
	wr.wr_id      = (uintptr_t) iu;
1700 1701 1702 1703 1704
	wr.sg_list    = &list;
	wr.num_sge    = 1;
	wr.opcode     = IB_WR_SEND;
	wr.send_flags = IB_SEND_SIGNALED;

1705
	return ib_post_send(ch->qp, &wr, &bad_wr);
1706 1707
}

1708
static int srp_post_recv(struct srp_rdma_ch *ch, struct srp_iu *iu)
1709
{
1710
	struct srp_target_port *target = ch->target;
1711
	struct ib_recv_wr wr, *bad_wr;
1712
	struct ib_sge list;
1713 1714 1715

	list.addr   = iu->dma;
	list.length = iu->size;
1716
	list.lkey   = target->lkey;
1717 1718

	wr.next     = NULL;
1719
	wr.wr_id    = (uintptr_t) iu;
1720 1721 1722
	wr.sg_list  = &list;
	wr.num_sge  = 1;

1723
	return ib_post_recv(ch->qp, &wr, &bad_wr);
1724 1725
}

1726
static void srp_process_rsp(struct srp_rdma_ch *ch, struct srp_rsp *rsp)
1727
{
1728
	struct srp_target_port *target = ch->target;
1729 1730 1731 1732 1733
	struct srp_request *req;
	struct scsi_cmnd *scmnd;
	unsigned long flags;

	if (unlikely(rsp->tag & SRP_TAG_TSK_MGMT)) {
1734 1735 1736
		spin_lock_irqsave(&ch->lock, flags);
		ch->req_lim += be32_to_cpu(rsp->req_lim_delta);
		spin_unlock_irqrestore(&ch->lock, flags);
1737

1738
		ch->tsk_mgmt_status = -1;
1739
		if (be32_to_cpu(rsp->resp_data_len) >= 4)
1740 1741
			ch->tsk_mgmt_status = rsp->data[3];
		complete(&ch->tsk_mgmt_done);
1742
	} else {
B
Bart Van Assche 已提交
1743 1744 1745 1746 1747
		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
1748
		if (!scmnd) {
1749
			shost_printk(KERN_ERR, target->scsi_host,
B
Bart Van Assche 已提交
1750 1751
				     "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
1752

1753 1754 1755
			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
1756 1757 1758

			return;
		}
1759 1760 1761 1762 1763 1764 1765 1766 1767
		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
1768
		if (unlikely(rsp->flags & SRP_RSP_FLAG_DIUNDER))
1769
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_in_res_cnt));
B
Bart Van Assche 已提交
1770 1771 1772 1773 1774 1775
		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));
1776

1777
		srp_free_req(ch, req, scmnd,
B
Bart Van Assche 已提交
1778 1779
			     be32_to_cpu(rsp->req_lim_delta));

1780 1781
		scmnd->host_scribble = NULL;
		scmnd->scsi_done(scmnd);
1782 1783 1784
	}
}

1785
static int srp_response_common(struct srp_rdma_ch *ch, s32 req_delta,
1786 1787
			       void *rsp, int len)
{
1788
	struct srp_target_port *target = ch->target;
1789
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1790 1791
	unsigned long flags;
	struct srp_iu *iu;
1792
	int err;
1793

1794 1795 1796 1797
	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);
1798

1799 1800 1801
	if (!iu) {
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "no IU available to send response\n");
1802
		return 1;
1803 1804 1805 1806 1807 1808
	}

	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);

1809
	err = srp_post_send(ch, iu, len);
1810
	if (err) {
1811 1812
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "unable to post response: %d\n", err);
1813
		srp_put_tx_iu(ch, iu, SRP_IU_RSP);
1814
	}
1815 1816 1817 1818

	return err;
}

1819
static void srp_process_cred_req(struct srp_rdma_ch *ch,
1820 1821 1822 1823 1824 1825 1826 1827
				 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);

1828 1829
	if (srp_response_common(ch, delta, &rsp, sizeof(rsp)))
		shost_printk(KERN_ERR, ch->target->scsi_host, PFX
1830 1831 1832
			     "problems processing SRP_CRED_REQ\n");
}

1833
static void srp_process_aer_req(struct srp_rdma_ch *ch,
1834 1835
				struct srp_aer_req *req)
{
1836
	struct srp_target_port *target = ch->target;
1837 1838 1839 1840 1841 1842 1843 1844 1845
	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));

1846
	if (srp_response_common(ch, delta, &rsp, sizeof(rsp)))
1847 1848 1849 1850
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "problems processing SRP_AER_REQ\n");
}

1851
static void srp_handle_recv(struct srp_rdma_ch *ch, struct ib_wc *wc)
1852
{
1853
	struct srp_target_port *target = ch->target;
1854
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1855
	struct srp_iu *iu = (struct srp_iu *) (uintptr_t) wc->wr_id;
1856
	int res;
1857 1858
	u8 opcode;

1859
	ib_dma_sync_single_for_cpu(dev, iu->dma, ch->max_ti_iu_len,
1860
				   DMA_FROM_DEVICE);
1861 1862 1863 1864

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

	if (0) {
1865 1866
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "recv completion, opcode 0x%02x\n", opcode);
B
Bart Van Assche 已提交
1867 1868
		print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET, 8, 1,
			       iu->buf, wc->byte_len, true);
1869 1870 1871 1872
	}

	switch (opcode) {
	case SRP_RSP:
1873
		srp_process_rsp(ch, iu->buf);
1874 1875
		break;

1876
	case SRP_CRED_REQ:
1877
		srp_process_cred_req(ch, iu->buf);
1878 1879 1880
		break;

	case SRP_AER_REQ:
1881
		srp_process_aer_req(ch, iu->buf);
1882 1883
		break;

1884 1885
	case SRP_T_LOGOUT:
		/* XXX Handle target logout */
1886 1887
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Got target logout request\n");
1888 1889 1890
		break;

	default:
1891 1892
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled SRP opcode 0x%02x\n", opcode);
1893 1894 1895
		break;
	}

1896
	ib_dma_sync_single_for_device(dev, iu->dma, ch->max_ti_iu_len,
1897
				      DMA_FROM_DEVICE);
1898

1899
	res = srp_post_recv(ch, iu);
1900 1901 1902
	if (res != 0)
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Recv failed with error code %d\n", res);
1903 1904
}

1905 1906
/**
 * srp_tl_err_work() - handle a transport layer error
1907
 * @work: Work structure embedded in an SRP target port.
1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920
 *
 * 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);
}

1921
static void srp_handle_qp_err(u64 wr_id, enum ib_wc_status wc_status,
1922
			      bool send_err, struct srp_rdma_ch *ch)
1923
{
1924 1925 1926 1927 1928 1929 1930
	struct srp_target_port *target = ch->target;

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

1931
	if (target->connected && !target->qp_in_error) {
1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945
		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);
		}
1946
		queue_work(system_long_wq, &target->tl_err_work);
1947
	}
1948 1949 1950
	target->qp_in_error = true;
}

1951
static void srp_recv_completion(struct ib_cq *cq, void *ch_ptr)
1952
{
1953
	struct srp_rdma_ch *ch = ch_ptr;
1954 1955 1956 1957
	struct ib_wc wc;

	ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
	while (ib_poll_cq(cq, 1, &wc) > 0) {
1958
		if (likely(wc.status == IB_WC_SUCCESS)) {
1959
			srp_handle_recv(ch, &wc);
1960
		} else {
1961
			srp_handle_qp_err(wc.wr_id, wc.status, false, ch);
1962
		}
1963 1964 1965
	}
}

1966
static void srp_send_completion(struct ib_cq *cq, void *ch_ptr)
1967
{
1968
	struct srp_rdma_ch *ch = ch_ptr;
1969
	struct ib_wc wc;
1970
	struct srp_iu *iu;
1971 1972

	while (ib_poll_cq(cq, 1, &wc) > 0) {
1973 1974
		if (likely(wc.status == IB_WC_SUCCESS)) {
			iu = (struct srp_iu *) (uintptr_t) wc.wr_id;
1975
			list_add(&iu->list, &ch->free_tx);
1976
		} else {
1977
			srp_handle_qp_err(wc.wr_id, wc.status, true, ch);
1978
		}
1979 1980 1981
	}
}

1982
static int srp_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *scmnd)
1983
{
1984
	struct srp_target_port *target = host_to_target(shost);
1985
	struct srp_rport *rport = target->rport;
1986
	struct srp_rdma_ch *ch;
1987 1988 1989
	struct srp_request *req;
	struct srp_iu *iu;
	struct srp_cmd *cmd;
1990
	struct ib_device *dev;
1991
	unsigned long flags;
B
Bart Van Assche 已提交
1992 1993
	u32 tag;
	u16 idx;
1994
	int len, ret;
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
	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);
2005

2006 2007 2008
	scmnd->result = srp_chkready(target->rport);
	if (unlikely(scmnd->result))
		goto err;
2009

B
Bart Van Assche 已提交
2010 2011
	WARN_ON_ONCE(scmnd->request->tag < 0);
	tag = blk_mq_unique_tag(scmnd->request);
B
Bart Van Assche 已提交
2012
	ch = &target->ch[blk_mq_unique_tag_to_hwq(tag)];
B
Bart Van Assche 已提交
2013 2014 2015 2016
	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);
2017 2018 2019 2020

	spin_lock_irqsave(&ch->lock, flags);
	iu = __srp_get_tx_iu(ch, SRP_IU_CMD);
	spin_unlock_irqrestore(&ch->lock, flags);
2021

B
Bart Van Assche 已提交
2022 2023 2024 2025
	if (!iu)
		goto err;

	req = &ch->req_ring[idx];
2026
	dev = target->srp_host->srp_dev->dev;
2027
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_iu_len,
2028
				   DMA_TO_DEVICE);
2029

2030
	scmnd->host_scribble = (void *) req;
2031 2032 2033 2034 2035 2036

	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
2037
	cmd->tag    = tag;
2038 2039 2040 2041 2042
	memcpy(cmd->cdb, scmnd->cmnd, scmnd->cmd_len);

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

2043
	len = srp_map_data(scmnd, ch, req);
2044
	if (len < 0) {
2045
		shost_printk(KERN_ERR, target->scsi_host,
2046 2047 2048 2049
			     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
2050
		 * max_pages_per_mr sg-list elements, tell the SCSI mid-layer
2051 2052 2053 2054
		 * to reduce queue depth temporarily.
		 */
		scmnd->result = len == -ENOMEM ?
			DID_OK << 16 | QUEUE_FULL << 1 : DID_ERROR << 16;
2055
		goto err_iu;
2056 2057
	}

2058
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_iu_len,
2059
				      DMA_TO_DEVICE);
2060

2061
	if (srp_post_send(ch, iu, len)) {
2062
		shost_printk(KERN_ERR, target->scsi_host, PFX "Send failed\n");
2063 2064 2065
		goto err_unmap;
	}

2066 2067
	ret = 0;

2068 2069 2070 2071
unlock_rport:
	if (in_scsi_eh)
		mutex_unlock(&rport->mutex);

2072
	return ret;
2073 2074

err_unmap:
2075
	srp_unmap_data(scmnd, ch, req);
2076

2077
err_iu:
2078
	srp_put_tx_iu(ch, iu, SRP_IU_CMD);
2079

2080 2081 2082 2083 2084 2085
	/*
	 * Avoid that the loops that iterate over the request ring can
	 * encounter a dangling SCSI command pointer.
	 */
	req->scmnd = NULL;

2086 2087 2088 2089 2090 2091 2092
err:
	if (scmnd->result) {
		scmnd->scsi_done(scmnd);
		ret = 0;
	} else {
		ret = SCSI_MLQUEUE_HOST_BUSY;
	}
2093

2094
	goto unlock_rport;
2095 2096
}

2097 2098
/*
 * Note: the resources allocated in this function are freed in
2099
 * srp_free_ch_ib().
2100
 */
2101
static int srp_alloc_iu_bufs(struct srp_rdma_ch *ch)
2102
{
2103
	struct srp_target_port *target = ch->target;
2104 2105
	int i;

2106 2107 2108
	ch->rx_ring = kcalloc(target->queue_size, sizeof(*ch->rx_ring),
			      GFP_KERNEL);
	if (!ch->rx_ring)
2109
		goto err_no_ring;
2110 2111 2112
	ch->tx_ring = kcalloc(target->queue_size, sizeof(*ch->tx_ring),
			      GFP_KERNEL);
	if (!ch->tx_ring)
2113 2114 2115
		goto err_no_ring;

	for (i = 0; i < target->queue_size; ++i) {
2116 2117 2118 2119
		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])
2120 2121 2122
			goto err;
	}

2123
	for (i = 0; i < target->queue_size; ++i) {
2124 2125 2126 2127
		ch->tx_ring[i] = srp_alloc_iu(target->srp_host,
					      target->max_iu_len,
					      GFP_KERNEL, DMA_TO_DEVICE);
		if (!ch->tx_ring[i])
2128
			goto err;
2129

2130
		list_add(&ch->tx_ring[i]->list, &ch->free_tx);
2131 2132 2133 2134 2135
	}

	return 0;

err:
2136
	for (i = 0; i < target->queue_size; ++i) {
2137 2138
		srp_free_iu(target->srp_host, ch->rx_ring[i]);
		srp_free_iu(target->srp_host, ch->tx_ring[i]);
2139 2140
	}

2141 2142

err_no_ring:
2143 2144 2145 2146
	kfree(ch->tx_ring);
	ch->tx_ring = NULL;
	kfree(ch->rx_ring);
	ch->rx_ring = NULL;
2147

2148 2149 2150
	return -ENOMEM;
}

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 2177
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;
}

2178 2179
static void srp_cm_rep_handler(struct ib_cm_id *cm_id,
			       struct srp_login_rsp *lrsp,
2180
			       struct srp_rdma_ch *ch)
2181
{
2182
	struct srp_target_port *target = ch->target;
2183 2184 2185 2186 2187 2188
	struct ib_qp_attr *qp_attr = NULL;
	int attr_mask = 0;
	int ret;
	int i;

	if (lrsp->opcode == SRP_LOGIN_RSP) {
2189 2190
		ch->max_ti_iu_len = be32_to_cpu(lrsp->max_ti_iu_len);
		ch->req_lim       = be32_to_cpu(lrsp->req_lim_delta);
2191 2192 2193 2194 2195 2196

		/*
		 * Reserve credits for task management so we don't
		 * bounce requests back to the SCSI mid-layer.
		 */
		target->scsi_host->can_queue
2197
			= min(ch->req_lim - SRP_TSK_MGMT_SQ_SIZE,
2198
			      target->scsi_host->can_queue);
2199 2200 2201
		target->scsi_host->cmd_per_lun
			= min_t(int, target->scsi_host->can_queue,
				target->scsi_host->cmd_per_lun);
2202 2203 2204 2205 2206 2207 2208
	} else {
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled RSP opcode %#x\n", lrsp->opcode);
		ret = -ECONNRESET;
		goto error;
	}

2209 2210
	if (!ch->rx_ring) {
		ret = srp_alloc_iu_bufs(ch);
2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224
		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;

2225
	ret = ib_modify_qp(ch->qp, qp_attr, attr_mask);
2226 2227 2228
	if (ret)
		goto error_free;

2229
	for (i = 0; i < target->queue_size; i++) {
2230 2231 2232
		struct srp_iu *iu = ch->rx_ring[i];

		ret = srp_post_recv(ch, iu);
2233 2234 2235 2236 2237 2238 2239 2240 2241
		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;

2242 2243
	target->rq_tmo_jiffies = srp_compute_rq_tmo(qp_attr, attr_mask);

2244
	ret = ib_modify_qp(ch->qp, qp_attr, attr_mask);
2245 2246 2247 2248 2249 2250 2251 2252 2253
	if (ret)
		goto error_free;

	ret = ib_send_cm_rtu(cm_id, NULL, 0);

error_free:
	kfree(qp_attr);

error:
2254
	ch->status = ret;
2255 2256
}

2257 2258
static void srp_cm_rej_handler(struct ib_cm_id *cm_id,
			       struct ib_cm_event *event,
2259
			       struct srp_rdma_ch *ch)
2260
{
2261
	struct srp_target_port *target = ch->target;
2262
	struct Scsi_Host *shost = target->scsi_host;
2263 2264 2265 2266 2267 2268
	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;
2269 2270
		ch->path.dlid = cpi->redirect_lid;
		ch->path.pkey = cpi->redirect_pkey;
2271
		cm_id->remote_cm_qpn = be32_to_cpu(cpi->redirect_qp) & 0x00ffffff;
2272
		memcpy(ch->path.dgid.raw, cpi->redirect_gid, 16);
2273

2274
		ch->status = ch->path.dlid ?
2275 2276 2277 2278
			SRP_DLID_REDIRECT : SRP_PORT_REDIRECT;
		break;

	case IB_CM_REJ_PORT_REDIRECT:
2279
		if (srp_target_is_topspin(target)) {
2280 2281 2282 2283 2284
			/*
			 * Topspin/Cisco SRP gateways incorrectly send
			 * reject reason code 25 when they mean 24
			 * (port redirect).
			 */
2285
			memcpy(ch->path.dgid.raw,
2286 2287
			       event->param.rej_rcvd.ari, 16);

2288 2289
			shost_printk(KERN_DEBUG, shost,
				     PFX "Topspin/Cisco redirect to target port GID %016llx%016llx\n",
2290 2291
				     be64_to_cpu(ch->path.dgid.global.subnet_prefix),
				     be64_to_cpu(ch->path.dgid.global.interface_id));
2292

2293
			ch->status = SRP_PORT_REDIRECT;
2294
		} else {
2295 2296
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_PORT_REDIRECT\n");
2297
			ch->status = -ECONNRESET;
2298 2299 2300 2301
		}
		break;

	case IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID:
2302 2303
		shost_printk(KERN_WARNING, shost,
			    "  REJ reason: IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID\n");
2304
		ch->status = -ECONNRESET;
2305 2306 2307 2308 2309 2310 2311 2312 2313
		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)
2314 2315
				shost_printk(KERN_WARNING, shost,
					     PFX "SRP_LOGIN_REJ: requested max_it_iu_len too large\n");
2316
			else
B
Bart Van Assche 已提交
2317 2318
				shost_printk(KERN_WARNING, shost, PFX
					     "SRP LOGIN from %pI6 to %pI6 REJECTED, reason 0x%08x\n",
2319 2320
					     target->sgid.raw,
					     target->orig_dgid.raw, reason);
2321
		} else
2322 2323 2324
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_CONSUMER_DEFINED,"
				     " opcode 0x%02x\n", opcode);
2325
		ch->status = -ECONNRESET;
2326 2327
		break;

D
David Dillow 已提交
2328 2329
	case IB_CM_REJ_STALE_CONN:
		shost_printk(KERN_WARNING, shost, "  REJ reason: stale connection\n");
2330
		ch->status = SRP_STALE_CONN;
D
David Dillow 已提交
2331 2332
		break;

2333
	default:
2334 2335
		shost_printk(KERN_WARNING, shost, "  REJ reason 0x%x\n",
			     event->param.rej_rcvd.reason);
2336
		ch->status = -ECONNRESET;
2337 2338 2339 2340 2341
	}
}

static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event)
{
2342 2343
	struct srp_rdma_ch *ch = cm_id->context;
	struct srp_target_port *target = ch->target;
2344 2345 2346 2347
	int comp = 0;

	switch (event->event) {
	case IB_CM_REQ_ERROR:
2348 2349
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Sending CM REQ failed\n");
2350
		comp = 1;
2351
		ch->status = -ECONNRESET;
2352 2353 2354 2355
		break;

	case IB_CM_REP_RECEIVED:
		comp = 1;
2356
		srp_cm_rep_handler(cm_id, event->private_data, ch);
2357 2358 2359
		break;

	case IB_CM_REJ_RECEIVED:
2360
		shost_printk(KERN_DEBUG, target->scsi_host, PFX "REJ received\n");
2361 2362
		comp = 1;

2363
		srp_cm_rej_handler(cm_id, event, ch);
2364 2365
		break;

2366
	case IB_CM_DREQ_RECEIVED:
2367 2368
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "DREQ received - connection closed\n");
2369
		srp_change_conn_state(target, false);
2370
		if (ib_send_cm_drep(cm_id, NULL, 0))
2371 2372
			shost_printk(KERN_ERR, target->scsi_host,
				     PFX "Sending CM DREP failed\n");
2373
		queue_work(system_long_wq, &target->tl_err_work);
2374 2375 2376
		break;

	case IB_CM_TIMEWAIT_EXIT:
2377 2378
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "connection closed\n");
2379
		comp = 1;
2380

2381
		ch->status = 0;
2382 2383
		break;

2384 2385 2386 2387 2388
	case IB_CM_MRA_RECEIVED:
	case IB_CM_DREQ_ERROR:
	case IB_CM_DREP_RECEIVED:
		break;

2389
	default:
2390 2391
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled CM event %d\n", event->event);
2392 2393 2394 2395
		break;
	}

	if (comp)
2396
		complete(&ch->done);
2397 2398 2399 2400

	return 0;
}

2401 2402 2403 2404
/**
 * srp_change_queue_depth - setting device queue depth
 * @sdev: scsi device struct
 * @qdepth: requested queue depth
2405
 * @reason: SCSI_QDEPTH_DEFAULT
2406 2407 2408 2409 2410 2411 2412
 * (see include/scsi/scsi_host.h for definition)
 *
 * Returns queue depth.
 */
static int
srp_change_queue_depth(struct scsi_device *sdev, int qdepth, int reason)
{
2413
	if (!sdev->tagged_supported)
2414
		qdepth = 1;
2415
	scsi_adjust_queue_depth(sdev, qdepth);
2416 2417 2418
	return sdev->queue_depth;
}

2419 2420
static int srp_send_tsk_mgmt(struct srp_rdma_ch *ch, u64 req_tag,
			     unsigned int lun, u8 func)
2421
{
2422
	struct srp_target_port *target = ch->target;
2423
	struct srp_rport *rport = target->rport;
2424
	struct ib_device *dev = target->srp_host->srp_dev->dev;
2425 2426 2427
	struct srp_iu *iu;
	struct srp_tsk_mgmt *tsk_mgmt;

2428 2429 2430
	if (!target->connected || target->qp_in_error)
		return -1;

2431
	init_completion(&ch->tsk_mgmt_done);
2432

2433
	/*
2434
	 * Lock the rport mutex to avoid that srp_create_ch_ib() is
2435 2436 2437
	 * invoked while a task management function is being sent.
	 */
	mutex_lock(&rport->mutex);
2438 2439 2440
	spin_lock_irq(&ch->lock);
	iu = __srp_get_tx_iu(ch, SRP_IU_TSK_MGMT);
	spin_unlock_irq(&ch->lock);
2441

2442 2443 2444
	if (!iu) {
		mutex_unlock(&rport->mutex);

2445
		return -1;
2446
	}
2447

2448 2449
	ib_dma_sync_single_for_cpu(dev, iu->dma, sizeof *tsk_mgmt,
				   DMA_TO_DEVICE);
2450 2451 2452 2453
	tsk_mgmt = iu->buf;
	memset(tsk_mgmt, 0, sizeof *tsk_mgmt);

	tsk_mgmt->opcode 	= SRP_TSK_MGMT;
2454 2455
	tsk_mgmt->lun		= cpu_to_be64((u64) lun << 48);
	tsk_mgmt->tag		= req_tag | SRP_TAG_TSK_MGMT;
2456
	tsk_mgmt->tsk_mgmt_func = func;
2457
	tsk_mgmt->task_tag	= req_tag;
2458

2459 2460
	ib_dma_sync_single_for_device(dev, iu->dma, sizeof *tsk_mgmt,
				      DMA_TO_DEVICE);
2461 2462
	if (srp_post_send(ch, iu, sizeof(*tsk_mgmt))) {
		srp_put_tx_iu(ch, iu, SRP_IU_TSK_MGMT);
2463 2464
		mutex_unlock(&rport->mutex);

2465 2466
		return -1;
	}
2467
	mutex_unlock(&rport->mutex);
2468

2469
	if (!wait_for_completion_timeout(&ch->tsk_mgmt_done,
2470
					 msecs_to_jiffies(SRP_ABORT_TIMEOUT_MS)))
2471
		return -1;
2472

2473 2474 2475
	return 0;
}

2476 2477
static int srp_abort(struct scsi_cmnd *scmnd)
{
2478
	struct srp_target_port *target = host_to_target(scmnd->device->host);
2479
	struct srp_request *req = (struct srp_request *) scmnd->host_scribble;
B
Bart Van Assche 已提交
2480
	u32 tag;
B
Bart Van Assche 已提交
2481
	u16 ch_idx;
2482
	struct srp_rdma_ch *ch;
2483
	int ret;
2484

2485
	shost_printk(KERN_ERR, target->scsi_host, "SRP abort called\n");
2486

B
Bart Van Assche 已提交
2487
	if (!req)
2488
		return SUCCESS;
B
Bart Van Assche 已提交
2489
	tag = blk_mq_unique_tag(scmnd->request);
B
Bart Van Assche 已提交
2490 2491 2492 2493 2494 2495 2496 2497
	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
2498
	if (srp_send_tsk_mgmt(ch, tag, scmnd->device->lun,
2499
			      SRP_TSK_ABORT_TASK) == 0)
2500
		ret = SUCCESS;
2501
	else if (target->rport->state == SRP_RPORT_LOST)
2502
		ret = FAST_IO_FAIL;
2503 2504
	else
		ret = FAILED;
2505
	srp_free_req(ch, req, scmnd, 0);
B
Bart Van Assche 已提交
2506
	scmnd->result = DID_ABORT << 16;
2507
	scmnd->scsi_done(scmnd);
2508

2509
	return ret;
2510 2511 2512 2513
}

static int srp_reset_device(struct scsi_cmnd *scmnd)
{
2514
	struct srp_target_port *target = host_to_target(scmnd->device->host);
B
Bart Van Assche 已提交
2515
	struct srp_rdma_ch *ch;
2516
	int i;
2517

2518
	shost_printk(KERN_ERR, target->scsi_host, "SRP reset_device called\n");
2519

B
Bart Van Assche 已提交
2520
	ch = &target->ch[0];
2521
	if (srp_send_tsk_mgmt(ch, SRP_TAG_NO_REQ, scmnd->device->lun,
2522
			      SRP_TSK_LUN_RESET))
2523
		return FAILED;
2524
	if (ch->tsk_mgmt_status)
2525 2526
		return FAILED;

B
Bart Van Assche 已提交
2527 2528 2529 2530
	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];
2531

B
Bart Van Assche 已提交
2532 2533
			srp_finish_req(ch, req, scmnd->device, DID_RESET << 16);
		}
2534
	}
2535 2536

	return SUCCESS;
2537 2538 2539 2540 2541 2542
}

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

2543
	shost_printk(KERN_ERR, target->scsi_host, PFX "SRP reset_host called\n");
2544

2545
	return srp_reconnect_rport(target->rport) == 0 ? SUCCESS : FAILED;
2546 2547
}

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

2563 2564
static ssize_t show_id_ext(struct device *dev, struct device_attribute *attr,
			   char *buf)
2565
{
2566
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2567 2568 2569 2570 2571

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

2572 2573
static ssize_t show_ioc_guid(struct device *dev, struct device_attribute *attr,
			     char *buf)
2574
{
2575
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2576 2577 2578 2579 2580

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

2581 2582
static ssize_t show_service_id(struct device *dev,
			       struct device_attribute *attr, char *buf)
2583
{
2584
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2585 2586 2587 2588 2589

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

2590 2591
static ssize_t show_pkey(struct device *dev, struct device_attribute *attr,
			 char *buf)
2592
{
2593
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2594

2595
	return sprintf(buf, "0x%04x\n", be16_to_cpu(target->pkey));
2596 2597
}

B
Bart Van Assche 已提交
2598 2599 2600 2601 2602
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));

2603
	return sprintf(buf, "%pI6\n", target->sgid.raw);
B
Bart Van Assche 已提交
2604 2605
}

2606 2607
static ssize_t show_dgid(struct device *dev, struct device_attribute *attr,
			 char *buf)
2608
{
2609
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
B
Bart Van Assche 已提交
2610
	struct srp_rdma_ch *ch = &target->ch[0];
2611

2612
	return sprintf(buf, "%pI6\n", ch->path.dgid.raw);
2613 2614
}

2615 2616
static ssize_t show_orig_dgid(struct device *dev,
			      struct device_attribute *attr, char *buf)
2617
{
2618
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2619

2620
	return sprintf(buf, "%pI6\n", target->orig_dgid.raw);
2621 2622
}

2623 2624 2625 2626
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
2627 2628
	struct srp_rdma_ch *ch;
	int i, req_lim = INT_MAX;
2629

B
Bart Van Assche 已提交
2630 2631 2632 2633 2634
	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);
2635 2636
}

2637 2638
static ssize_t show_zero_req_lim(struct device *dev,
				 struct device_attribute *attr, char *buf)
2639
{
2640
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2641 2642 2643 2644

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

2645 2646
static ssize_t show_local_ib_port(struct device *dev,
				  struct device_attribute *attr, char *buf)
2647
{
2648
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2649 2650 2651 2652

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

2653 2654
static ssize_t show_local_ib_device(struct device *dev,
				    struct device_attribute *attr, char *buf)
2655
{
2656
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2657

2658
	return sprintf(buf, "%s\n", target->srp_host->srp_dev->dev->name);
2659 2660
}

B
Bart Van Assche 已提交
2661 2662 2663 2664 2665 2666 2667 2668
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);
}

2669 2670 2671 2672 2673 2674 2675 2676
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);
}

2677 2678 2679 2680 2681 2682 2683 2684
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);
}

2685 2686 2687 2688 2689 2690 2691 2692
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);
}

2693 2694 2695 2696 2697 2698 2699 2700
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");
}

2701 2702 2703 2704
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
2705
static DEVICE_ATTR(sgid,	    S_IRUGO, show_sgid,		   NULL);
2706 2707
static DEVICE_ATTR(dgid,	    S_IRUGO, show_dgid,		   NULL);
static DEVICE_ATTR(orig_dgid,	    S_IRUGO, show_orig_dgid,	   NULL);
2708
static DEVICE_ATTR(req_lim,         S_IRUGO, show_req_lim,         NULL);
2709 2710 2711
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
2712
static DEVICE_ATTR(ch_count,        S_IRUGO, show_ch_count,        NULL);
2713
static DEVICE_ATTR(comp_vector,     S_IRUGO, show_comp_vector,     NULL);
2714
static DEVICE_ATTR(tl_retry_count,  S_IRUGO, show_tl_retry_count,  NULL);
2715
static DEVICE_ATTR(cmd_sg_entries,  S_IRUGO, show_cmd_sg_entries,  NULL);
2716
static DEVICE_ATTR(allow_ext_sg,    S_IRUGO, show_allow_ext_sg,    NULL);
2717 2718 2719 2720 2721 2722

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
2723
	&dev_attr_sgid,
2724 2725
	&dev_attr_dgid,
	&dev_attr_orig_dgid,
2726
	&dev_attr_req_lim,
2727 2728 2729
	&dev_attr_zero_req_lim,
	&dev_attr_local_ib_port,
	&dev_attr_local_ib_device,
B
Bart Van Assche 已提交
2730
	&dev_attr_ch_count,
2731
	&dev_attr_comp_vector,
2732
	&dev_attr_tl_retry_count,
2733
	&dev_attr_cmd_sg_entries,
2734
	&dev_attr_allow_ext_sg,
2735 2736 2737
	NULL
};

2738 2739
static struct scsi_host_template srp_template = {
	.module				= THIS_MODULE,
R
Roland Dreier 已提交
2740 2741
	.name				= "InfiniBand SRP initiator",
	.proc_name			= DRV_NAME,
2742
	.slave_configure		= srp_slave_configure,
2743 2744
	.info				= srp_target_info,
	.queuecommand			= srp_queuecommand,
2745
	.change_queue_depth             = srp_change_queue_depth,
2746
	.change_queue_type              = scsi_change_queue_type,
2747 2748 2749
	.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
2750
	.skip_settle_delay		= true,
2751
	.sg_tablesize			= SRP_DEF_SG_TABLESIZE,
2752
	.can_queue			= SRP_DEFAULT_CMD_SQ_SIZE,
2753
	.this_id			= -1,
2754
	.cmd_per_lun			= SRP_DEFAULT_CMD_SQ_SIZE,
2755
	.use_clustering			= ENABLE_CLUSTERING,
B
Bart Van Assche 已提交
2756 2757
	.shost_attrs			= srp_host_attrs,
	.use_blk_tags			= 1,
2758
	.track_queue_depth		= 1,
2759 2760
};

2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771
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;
}

2772 2773
static int srp_add_target(struct srp_host *host, struct srp_target_port *target)
{
2774 2775 2776
	struct srp_rport_identifiers ids;
	struct srp_rport *rport;

2777
	target->state = SRP_TARGET_SCANNING;
2778 2779 2780
	sprintf(target->target_name, "SRP.T10:%016llX",
		 (unsigned long long) be64_to_cpu(target->id_ext));

2781
	if (scsi_add_host(target->scsi_host, host->srp_dev->dev->dma_device))
2782 2783
		return -ENODEV;

2784 2785
	memcpy(ids.port_id, &target->id_ext, 8);
	memcpy(ids.port_id + 8, &target->ioc_guid, 8);
2786
	ids.roles = SRP_RPORT_ROLE_TARGET;
2787 2788 2789 2790 2791 2792
	rport = srp_rport_add(target->scsi_host, &ids);
	if (IS_ERR(rport)) {
		scsi_remove_host(target->scsi_host);
		return PTR_ERR(rport);
	}

2793
	rport->lld_data = target;
2794
	target->rport = rport;
2795

2796
	spin_lock(&host->target_lock);
2797
	list_add_tail(&target->list, &host->target_list);
2798
	spin_unlock(&host->target_lock);
2799 2800

	scsi_scan_target(&target->scsi_host->shost_gendev,
2801
			 0, target->scsi_id, SCAN_WILD_CARD, 0);
2802

2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819
	if (!target->connected || target->qp_in_error) {
		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:
2820 2821 2822
	return 0;
}

2823
static void srp_release_dev(struct device *dev)
2824 2825
{
	struct srp_host *host =
2826
		container_of(dev, struct srp_host, dev);
2827 2828 2829 2830 2831 2832

	complete(&host->released);
}

static struct class srp_class = {
	.name    = "infiniband_srp",
2833
	.dev_release = srp_release_dev
2834 2835
};

2836 2837
/**
 * srp_conn_unique() - check whether the connection to a target is unique
2838 2839
 * @host:   SRP host.
 * @target: SRP target port.
2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867
 */
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;
}

2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883
/*
 * 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,
2884
	SRP_OPT_MAX_CMD_PER_LUN	= 1 << 6,
2885
	SRP_OPT_IO_CLASS	= 1 << 7,
2886
	SRP_OPT_INITIATOR_EXT	= 1 << 8,
2887
	SRP_OPT_CMD_SG_ENTRIES	= 1 << 9,
2888 2889
	SRP_OPT_ALLOW_EXT_SG	= 1 << 10,
	SRP_OPT_SG_TABLESIZE	= 1 << 11,
2890
	SRP_OPT_COMP_VECTOR	= 1 << 12,
2891
	SRP_OPT_TL_RETRY_COUNT	= 1 << 13,
2892
	SRP_OPT_QUEUE_SIZE	= 1 << 14,
2893 2894 2895 2896 2897 2898 2899
	SRP_OPT_ALL		= (SRP_OPT_ID_EXT	|
				   SRP_OPT_IOC_GUID	|
				   SRP_OPT_DGID		|
				   SRP_OPT_PKEY		|
				   SRP_OPT_SERVICE_ID),
};

2900
static const match_table_t srp_opt_tokens = {
2901 2902 2903 2904 2905 2906 2907
	{ 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" 	},
2908
	{ SRP_OPT_IO_CLASS,		"io_class=%x"		},
2909
	{ SRP_OPT_INITIATOR_EXT,	"initiator_ext=%s"	},
2910
	{ SRP_OPT_CMD_SG_ENTRIES,	"cmd_sg_entries=%u"	},
2911 2912
	{ SRP_OPT_ALLOW_EXT_SG,		"allow_ext_sg=%u"	},
	{ SRP_OPT_SG_TABLESIZE,		"sg_tablesize=%u"	},
2913
	{ SRP_OPT_COMP_VECTOR,		"comp_vector=%u"	},
2914
	{ SRP_OPT_TL_RETRY_COUNT,	"tl_retry_count=%u"	},
2915
	{ SRP_OPT_QUEUE_SIZE,		"queue_size=%d"		},
2916
	{ SRP_OPT_ERR,			NULL 			}
2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944
};

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

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

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

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

		switch (token) {
		case SRP_OPT_ID_EXT:
			p = match_strdup(args);
2945 2946 2947 2948
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2949 2950 2951 2952 2953 2954
			target->id_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_IOC_GUID:
			p = match_strdup(args);
2955 2956 2957 2958
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2959 2960 2961 2962 2963 2964
			target->ioc_guid = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_DGID:
			p = match_strdup(args);
2965 2966 2967 2968
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2969
			if (strlen(p) != 32) {
2970
				pr_warn("bad dest GID parameter '%s'\n", p);
2971
				kfree(p);
2972 2973 2974 2975
				goto out;
			}

			for (i = 0; i < 16; ++i) {
2976 2977 2978 2979 2980 2981 2982
				strlcpy(dgid, p + i * 2, sizeof(dgid));
				if (sscanf(dgid, "%hhx",
					   &target->orig_dgid.raw[i]) < 1) {
					ret = -EINVAL;
					kfree(p);
					goto out;
				}
2983
			}
2984
			kfree(p);
2985 2986 2987 2988
			break;

		case SRP_OPT_PKEY:
			if (match_hex(args, &token)) {
2989
				pr_warn("bad P_Key parameter '%s'\n", p);
2990 2991
				goto out;
			}
2992
			target->pkey = cpu_to_be16(token);
2993 2994 2995 2996
			break;

		case SRP_OPT_SERVICE_ID:
			p = match_strdup(args);
2997 2998 2999 3000
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
3001 3002 3003 3004 3005 3006
			target->service_id = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_MAX_SECT:
			if (match_int(args, &token)) {
3007
				pr_warn("bad max sect parameter '%s'\n", p);
3008 3009 3010 3011 3012
				goto out;
			}
			target->scsi_host->max_sectors = token;
			break;

3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024
		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;

3025
		case SRP_OPT_MAX_CMD_PER_LUN:
3026
			if (match_int(args, &token) || token < 1) {
3027 3028
				pr_warn("bad max cmd_per_lun parameter '%s'\n",
					p);
3029 3030
				goto out;
			}
3031
			target->scsi_host->cmd_per_lun = token;
3032 3033
			break;

3034 3035
		case SRP_OPT_IO_CLASS:
			if (match_hex(args, &token)) {
3036
				pr_warn("bad IO class parameter '%s'\n", p);
3037 3038 3039 3040
				goto out;
			}
			if (token != SRP_REV10_IB_IO_CLASS &&
			    token != SRP_REV16A_IB_IO_CLASS) {
3041 3042 3043
				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);
3044 3045 3046 3047 3048
				goto out;
			}
			target->io_class = token;
			break;

3049 3050
		case SRP_OPT_INITIATOR_EXT:
			p = match_strdup(args);
3051 3052 3053 3054
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
3055 3056 3057 3058
			target->initiator_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

3059 3060
		case SRP_OPT_CMD_SG_ENTRIES:
			if (match_int(args, &token) || token < 1 || token > 255) {
3061 3062
				pr_warn("bad max cmd_sg_entries parameter '%s'\n",
					p);
3063 3064 3065 3066 3067
				goto out;
			}
			target->cmd_sg_cnt = token;
			break;

3068 3069
		case SRP_OPT_ALLOW_EXT_SG:
			if (match_int(args, &token)) {
3070
				pr_warn("bad allow_ext_sg parameter '%s'\n", p);
3071 3072 3073 3074 3075 3076 3077 3078
				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) {
3079 3080
				pr_warn("bad max sg_tablesize parameter '%s'\n",
					p);
3081 3082 3083 3084 3085
				goto out;
			}
			target->sg_tablesize = token;
			break;

3086 3087 3088 3089 3090 3091 3092 3093
		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;

3094 3095 3096 3097 3098 3099 3100 3101 3102
		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;

3103
		default:
3104 3105
			pr_warn("unknown parameter or missing value '%s' in target creation request\n",
				p);
3106 3107 3108 3109 3110 3111 3112 3113 3114 3115
			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))
3116 3117
				pr_warn("target creation request is missing parameter '%s'\n",
					srp_opt_tokens[i].pattern);
3118

3119 3120 3121 3122 3123 3124
	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);

3125 3126 3127 3128 3129
out:
	kfree(options);
	return ret;
}

3130 3131
static ssize_t srp_create_target(struct device *dev,
				 struct device_attribute *attr,
3132 3133 3134
				 const char *buf, size_t count)
{
	struct srp_host *host =
3135
		container_of(dev, struct srp_host, dev);
3136 3137
	struct Scsi_Host *target_host;
	struct srp_target_port *target;
3138
	struct srp_rdma_ch *ch;
3139 3140
	struct srp_device *srp_dev = host->srp_dev;
	struct ib_device *ibdev = srp_dev->dev;
B
Bart Van Assche 已提交
3141 3142
	int ret, node_idx, node, cpu, i;
	bool multich = false;
3143 3144 3145 3146 3147 3148

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

3149
	target_host->transportt  = ib_srp_transport_template;
3150 3151
	target_host->max_channel = 0;
	target_host->max_id      = 1;
A
Arne Redlich 已提交
3152 3153
	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
3154

3155 3156
	target = host_to_target(target_host);

3157 3158 3159 3160 3161 3162
	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;
3163 3164
	target->sg_tablesize	= indirect_sg_entries ? : cmd_sg_entries;
	target->allow_ext_sg	= allow_ext_sg;
3165
	target->tl_retry_count	= 7;
3166
	target->queue_size	= SRP_DEFAULT_QUEUE_SIZE;
3167

3168 3169 3170 3171 3172 3173
	/*
	 * Avoid that the SCSI host can be removed by srp_remove_target()
	 * before this function returns.
	 */
	scsi_host_get(target->scsi_host);

3174 3175
	mutex_lock(&host->add_target_mutex);

3176 3177 3178 3179
	ret = srp_parse_options(buf, target);
	if (ret)
		goto err;

B
Bart Van Assche 已提交
3180 3181 3182 3183
	ret = scsi_init_shared_tag_map(target_host, target_host->can_queue);
	if (ret)
		goto err;

3184 3185
	target->req_ring_size = target->queue_size - SRP_TSK_MGMT_SQ_SIZE;

3186 3187 3188 3189 3190 3191 3192 3193 3194 3195
	if (!srp_conn_unique(target->srp_host, target)) {
		shost_printk(KERN_INFO, target->scsi_host,
			     PFX "Already connected to target port with id_ext=%016llx;ioc_guid=%016llx;initiator_ext=%016llx\n",
			     be64_to_cpu(target->id_ext),
			     be64_to_cpu(target->ioc_guid),
			     be64_to_cpu(target->initiator_ext));
		ret = -EEXIST;
		goto err;
	}

3196
	if (!srp_dev->has_fmr && !srp_dev->has_fr && !target->allow_ext_sg &&
3197
	    target->cmd_sg_cnt < target->sg_tablesize) {
3198
		pr_warn("No MR pool and no external indirect descriptors, limiting sg_tablesize to cmd_sg_cnt\n");
3199 3200 3201 3202 3203 3204
		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);
3205 3206 3207 3208
	target->max_iu_len = sizeof (struct srp_cmd) +
			     sizeof (struct srp_indirect_buf) +
			     target->cmd_sg_cnt * sizeof (struct srp_direct_buf);

3209
	INIT_WORK(&target->tl_err_work, srp_tl_err_work);
3210
	INIT_WORK(&target->remove_work, srp_remove_work);
3211
	spin_lock_init(&target->lock);
3212
	ret = ib_query_gid(ibdev, host->port, 0, &target->sgid);
3213
	if (ret)
B
Bart Van Assche 已提交
3214
		goto err;
3215

B
Bart Van Assche 已提交
3216 3217 3218 3219 3220 3221 3222 3223 3224 3225
	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)
		goto err;
3226

B
Bart Van Assche 已提交
3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254
	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;
3255

B
Bart Van Assche 已提交
3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283
			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++;
3284 3285
	}

B
Bart Van Assche 已提交
3286 3287
	target->scsi_host->nr_hw_queues = target->ch_count;

3288 3289 3290 3291
	ret = srp_add_target(host, target);
	if (ret)
		goto err_disconnect;

3292 3293 3294 3295 3296
	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),
3297
			     be16_to_cpu(target->pkey),
3298
			     be64_to_cpu(target->service_id),
3299
			     target->sgid.raw, target->orig_dgid.raw);
3300
	}
B
Bart Van Assche 已提交
3301

3302 3303 3304 3305
	ret = count;

out:
	mutex_unlock(&host->add_target_mutex);
3306 3307 3308

	scsi_host_put(target->scsi_host);

3309
	return ret;
3310 3311 3312 3313

err_disconnect:
	srp_disconnect_target(target);

B
Bart Van Assche 已提交
3314 3315 3316 3317 3318
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		srp_free_ch_ib(target, ch);
		srp_free_req_data(target, ch);
	}
3319

B
Bart Van Assche 已提交
3320
	kfree(target->ch);
3321

3322 3323
err:
	scsi_host_put(target_host);
3324
	goto out;
3325 3326
}

3327
static DEVICE_ATTR(add_target, S_IWUSR, NULL, srp_create_target);
3328

3329 3330
static ssize_t show_ibdev(struct device *dev, struct device_attribute *attr,
			  char *buf)
3331
{
3332
	struct srp_host *host = container_of(dev, struct srp_host, dev);
3333

3334
	return sprintf(buf, "%s\n", host->srp_dev->dev->name);
3335 3336
}

3337
static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
3338

3339 3340
static ssize_t show_port(struct device *dev, struct device_attribute *attr,
			 char *buf)
3341
{
3342
	struct srp_host *host = container_of(dev, struct srp_host, dev);
3343 3344 3345 3346

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

3347
static DEVICE_ATTR(port, S_IRUGO, show_port, NULL);
3348

3349
static struct srp_host *srp_add_port(struct srp_device *device, u8 port)
3350 3351 3352 3353 3354 3355 3356 3357
{
	struct srp_host *host;

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

	INIT_LIST_HEAD(&host->target_list);
3358
	spin_lock_init(&host->target_lock);
3359
	init_completion(&host->released);
3360
	mutex_init(&host->add_target_mutex);
3361
	host->srp_dev = device;
3362 3363
	host->port = port;

3364 3365
	host->dev.class = &srp_class;
	host->dev.parent = device->dev->dma_device;
3366
	dev_set_name(&host->dev, "srp-%s-%d", device->dev->name, port);
3367

3368
	if (device_register(&host->dev))
3369
		goto free_host;
3370
	if (device_create_file(&host->dev, &dev_attr_add_target))
3371
		goto err_class;
3372
	if (device_create_file(&host->dev, &dev_attr_ibdev))
3373
		goto err_class;
3374
	if (device_create_file(&host->dev, &dev_attr_port))
3375 3376 3377 3378 3379
		goto err_class;

	return host;

err_class:
3380
	device_unregister(&host->dev);
3381

3382
free_host:
3383 3384 3385 3386 3387 3388 3389
	kfree(host);

	return NULL;
}

static void srp_add_one(struct ib_device *device)
{
3390 3391
	struct srp_device *srp_dev;
	struct ib_device_attr *dev_attr;
3392
	struct srp_host *host;
3393 3394
	int mr_page_shift, s, e, p;
	u64 max_pages_per_mr;
3395

3396 3397
	dev_attr = kmalloc(sizeof *dev_attr, GFP_KERNEL);
	if (!dev_attr)
3398
		return;
3399

3400
	if (ib_query_device(device, dev_attr)) {
3401
		pr_warn("Query device failed for %s\n", device->name);
3402 3403 3404 3405 3406 3407 3408
		goto free_attr;
	}

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

3409 3410
	srp_dev->has_fmr = (device->alloc_fmr && device->dealloc_fmr &&
			    device->map_phys_fmr && device->unmap_fmr);
3411 3412 3413 3414 3415 3416 3417
	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));
3418

3419 3420
	/*
	 * Use the smallest page size supported by the HCA, down to a
3421 3422
	 * minimum of 4096 bytes. We're unlikely to build large sglists
	 * out of smaller entries.
3423
	 */
3424 3425 3426 3427 3428 3429 3430
	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);
3431 3432 3433 3434 3435
	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);
	}
3436 3437
	srp_dev->mr_max_size	= srp_dev->mr_page_size *
				   srp_dev->max_pages_per_mr;
3438
	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",
3439
		 device->name, mr_page_shift, dev_attr->max_mr_size,
3440
		 dev_attr->max_fast_reg_page_list_len,
3441
		 srp_dev->max_pages_per_mr, srp_dev->mr_max_size);
3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456

	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
3457
	if (device->node_type == RDMA_NODE_IB_SWITCH) {
3458 3459 3460 3461 3462 3463 3464 3465
		s = 0;
		e = 0;
	} else {
		s = 1;
		e = device->phys_port_cnt;
	}

	for (p = s; p <= e; ++p) {
3466
		host = srp_add_port(srp_dev, p);
3467
		if (host)
3468
			list_add_tail(&host->list, &srp_dev->dev_list);
3469 3470
	}

3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482
	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);
3483 3484 3485 3486
}

static void srp_remove_one(struct ib_device *device)
{
3487
	struct srp_device *srp_dev;
3488
	struct srp_host *host, *tmp_host;
3489
	struct srp_target_port *target;
3490

3491
	srp_dev = ib_get_client_data(device, &srp_client);
3492 3493
	if (!srp_dev)
		return;
3494

3495
	list_for_each_entry_safe(host, tmp_host, &srp_dev->dev_list, list) {
3496
		device_unregister(&host->dev);
3497 3498 3499 3500 3501 3502 3503
		/*
		 * Wait for the sysfs entry to go away, so that no new
		 * target ports can be created.
		 */
		wait_for_completion(&host->released);

		/*
3504
		 * Remove all target ports.
3505
		 */
3506
		spin_lock(&host->target_lock);
3507 3508
		list_for_each_entry(target, &host->target_list, list)
			srp_queue_remove_work(target);
3509
		spin_unlock(&host->target_lock);
3510 3511

		/*
3512
		 * Wait for tl_err and target port removal tasks.
3513
		 */
3514
		flush_workqueue(system_long_wq);
3515
		flush_workqueue(srp_remove_wq);
3516 3517 3518 3519

		kfree(host);
	}

3520 3521 3522 3523
	ib_dereg_mr(srp_dev->mr);
	ib_dealloc_pd(srp_dev->pd);

	kfree(srp_dev);
3524 3525
}

3526
static struct srp_function_template ib_srp_transport_functions = {
3527 3528
	.has_rport_state	 = true,
	.reset_timer_if_blocked	 = true,
3529
	.reconnect_delay	 = &srp_reconnect_delay,
3530 3531 3532
	.fast_io_fail_tmo	 = &srp_fast_io_fail_tmo,
	.dev_loss_tmo		 = &srp_dev_loss_tmo,
	.reconnect		 = srp_rport_reconnect,
3533
	.rport_delete		 = srp_rport_delete,
3534
	.terminate_rport_io	 = srp_terminate_io,
3535 3536
};

3537 3538 3539 3540
static int __init srp_init_module(void)
{
	int ret;

3541
	BUILD_BUG_ON(FIELD_SIZEOF(struct ib_wc, wr_id) < sizeof(void *));
3542

3543
	if (srp_sg_tablesize) {
3544
		pr_warn("srp_sg_tablesize is deprecated, please use cmd_sg_entries\n");
3545 3546 3547 3548 3549 3550 3551 3552
		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) {
3553
		pr_warn("Clamping cmd_sg_entries to 255\n");
3554
		cmd_sg_entries = 255;
3555 3556
	}

3557 3558 3559
	if (!indirect_sg_entries)
		indirect_sg_entries = cmd_sg_entries;
	else if (indirect_sg_entries < cmd_sg_entries) {
3560 3561
		pr_warn("Bumping up indirect_sg_entries to match cmd_sg_entries (%u)\n",
			cmd_sg_entries);
3562 3563 3564
		indirect_sg_entries = cmd_sg_entries;
	}

3565
	srp_remove_wq = create_workqueue("srp_remove");
3566 3567
	if (!srp_remove_wq) {
		ret = -ENOMEM;
3568 3569 3570 3571
		goto out;
	}

	ret = -ENOMEM;
3572 3573 3574
	ib_srp_transport_template =
		srp_attach_transport(&ib_srp_transport_functions);
	if (!ib_srp_transport_template)
3575
		goto destroy_wq;
3576

3577 3578
	ret = class_register(&srp_class);
	if (ret) {
3579
		pr_err("couldn't register class infiniband_srp\n");
3580
		goto release_tr;
3581 3582
	}

3583 3584
	ib_sa_register_client(&srp_sa_client);

3585 3586
	ret = ib_register_client(&srp_client);
	if (ret) {
3587
		pr_err("couldn't register IB client\n");
3588
		goto unreg_sa;
3589 3590
	}

3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603
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;
3604 3605 3606 3607 3608
}

static void __exit srp_cleanup_module(void)
{
	ib_unregister_client(&srp_client);
3609
	ib_sa_unregister_client(&srp_sa_client);
3610
	class_unregister(&srp_class);
3611
	srp_release_transport(ib_srp_transport_template);
3612
	destroy_workqueue(srp_remove_wq);
3613 3614 3615 3616
}

module_init(srp_init_module);
module_exit(srp_cleanup_module);