ib_srp.c 94.3 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
/*
 * Copyright (c) 2005 Cisco Systems.  All rights reserved.
 *
 * This software is available to you under a choice of one of two
 * licenses.  You may choose to be licensed under the terms of the GNU
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - Redistributions in binary form must reproduce the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer in the documentation and/or other materials
 *        provided with the distribution.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 */

33
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
34

35 36 37 38 39 40 41
#include <linux/module.h>
#include <linux/init.h>
#include <linux/slab.h>
#include <linux/err.h>
#include <linux/string.h>
#include <linux/parser.h>
#include <linux/random.h>
42
#include <linux/jiffies.h>
43
#include <rdma/ib_cache.h>
44

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

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

#include "ib_srp.h"

#define DRV_NAME	"ib_srp"
#define PFX		DRV_NAME ": "
58 59
#define DRV_VERSION	"2.0"
#define DRV_RELDATE	"July 26, 2015"
60 61

MODULE_AUTHOR("Roland Dreier");
62
MODULE_DESCRIPTION("InfiniBand SCSI RDMA Protocol initiator");
63
MODULE_LICENSE("Dual BSD/GPL");
64 65
MODULE_VERSION(DRV_VERSION);
MODULE_INFO(release_date, DRV_RELDATE);
66

67 68
static unsigned int srp_sg_tablesize;
static unsigned int cmd_sg_entries;
69 70
static unsigned int indirect_sg_entries;
static bool allow_ext_sg;
71 72
static bool prefer_fr = true;
static bool register_always = true;
73
static int topspin_workarounds = 1;
74

75 76
module_param(srp_sg_tablesize, uint, 0444);
MODULE_PARM_DESC(srp_sg_tablesize, "Deprecated name for cmd_sg_entries");
77

78 79 80
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)");
81

82 83 84 85 86 87 88 89
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)");

90 91 92 93
module_param(topspin_workarounds, int, 0444);
MODULE_PARM_DESC(topspin_workarounds,
		 "Enable workarounds for Topspin/Cisco SRP target bugs if != 0");

94 95 96 97
module_param(prefer_fr, bool, 0444);
MODULE_PARM_DESC(prefer_fr,
"Whether to use fast registration if both FMR and fast registration are supported");

98 99 100 101
module_param(register_always, bool, 0444);
MODULE_PARM_DESC(register_always,
		 "Use memory registration even for contiguous memory regions");

102
static const struct kernel_param_ops srp_tmo_ops;
103

104 105 106 107 108
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");

109 110 111 112 113 114 115 116
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.");

117
static int srp_dev_loss_tmo = 600;
118 119 120 121 122 123 124 125 126 127
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
128 129 130 131 132
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.");

133
static void srp_add_one(struct ib_device *device);
134
static void srp_remove_one(struct ib_device *device, void *client_data);
135 136
static void srp_recv_completion(struct ib_cq *cq, void *ch_ptr);
static void srp_send_completion(struct ib_cq *cq, void *ch_ptr);
137 138
static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event);

139
static struct scsi_transport_template *ib_srp_transport_template;
140
static struct workqueue_struct *srp_remove_wq;
141

142 143 144 145 146 147
static struct ib_client srp_client = {
	.name   = "srp",
	.add    = srp_add_one,
	.remove = srp_remove_one
};

148 149
static struct ib_sa_client srp_sa_client;

150 151 152 153 154 155 156 157 158 159 160 161 162 163
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;

164 165 166 167
	res = srp_parse_tmo(&tmo, val);
	if (res)
		goto out;

168 169 170 171 172
	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);
173
	else
174 175
		res = srp_tmo_valid(srp_reconnect_delay, srp_fast_io_fail_tmo,
				    tmo);
176 177 178 179 180 181 182 183
	if (res)
		goto out;
	*(int *)kp->arg = tmo;

out:
	return res;
}

184
static const struct kernel_param_ops srp_tmo_ops = {
185 186 187 188
	.get = srp_tmo_get,
	.set = srp_tmo_set,
};

189 190 191 192 193 194 195 196 197 198
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;
}

199 200 201
static int srp_target_is_topspin(struct srp_target_port *target)
{
	static const u8 topspin_oui[3] = { 0x00, 0x05, 0xad };
202
	static const u8 cisco_oui[3]   = { 0x00, 0x1b, 0x0d };
203 204

	return topspin_workarounds &&
205 206
		(!memcmp(&target->ioc_guid, topspin_oui, sizeof topspin_oui) ||
		 !memcmp(&target->ioc_guid, cisco_oui, sizeof cisco_oui));
207 208
}

209 210 211 212 213 214 215 216 217 218 219 220 221 222
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;

223 224 225
	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))
226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245
		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;

246 247
	ib_dma_unmap_single(host->srp_dev->dev, iu->dma, iu->size,
			    iu->direction);
248 249 250 251 252 253
	kfree(iu->buf);
	kfree(iu);
}

static void srp_qp_event(struct ib_event *event, void *context)
{
254 255
	pr_debug("QP event %s (%d)\n",
		 ib_event_msg(event->event), 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;

268 269 270 271
	ret = ib_find_cached_pkey(target->srp_host->srp_dev->dev,
				  target->srp_host->port,
				  be16_to_cpu(target->pkey),
				  &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
/**
 * 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++) {
S
Sagi Grimberg 已提交
381 382
		mr = ib_alloc_mr(pd, IB_MR_TYPE_MEM_REG,
				 max_page_list_len);
383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456
		if (IS_ERR(mr)) {
			ret = PTR_ERR(mr);
			goto destroy_pool;
		}
		d->mr = mr;
		frpl = ib_alloc_fast_reg_page_list(device, max_page_list_len);
		if (IS_ERR(frpl)) {
			ret = PTR_ERR(frpl);
			goto destroy_pool;
		}
		d->frpl = frpl;
		list_add_tail(&d->entry, &pool->free_list);
	}

out:
	return pool;

destroy_pool:
	srp_destroy_fr_pool(pool);

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

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

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

	return d;
}

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

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

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

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

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

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

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

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

out:
	ib_destroy_qp(ch->qp);
}

491
static int srp_create_ch_ib(struct srp_rdma_ch *ch)
492
{
493
	struct srp_target_port *target = ch->target;
494
	struct srp_device *dev = target->srp_host->srp_dev;
495
	struct ib_qp_init_attr *init_attr;
496 497
	struct ib_cq *recv_cq, *send_cq;
	struct ib_qp *qp;
498
	struct ib_fmr_pool *fmr_pool = NULL;
499 500
	struct srp_fr_pool *fr_pool = NULL;
	const int m = 1 + dev->use_fast_reg;
501
	struct ib_cq_init_attr cq_attr = {};
502 503 504 505 506 507
	int ret;

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

508
	/* + 1 for SRP_LAST_WR_ID */
509 510
	cq_attr.cqe = target->queue_size + 1;
	cq_attr.comp_vector = ch->comp_vector;
511
	recv_cq = ib_create_cq(dev->dev, srp_recv_completion, NULL, ch,
512
			       &cq_attr);
513 514
	if (IS_ERR(recv_cq)) {
		ret = PTR_ERR(recv_cq);
515
		goto err;
516 517
	}

518 519
	cq_attr.cqe = m * target->queue_size;
	cq_attr.comp_vector = ch->comp_vector;
520
	send_cq = ib_create_cq(dev->dev, srp_send_completion, NULL, ch,
521
			       &cq_attr);
522 523
	if (IS_ERR(send_cq)) {
		ret = PTR_ERR(send_cq);
524
		goto err_recv_cq;
525 526
	}

527
	ib_req_notify_cq(recv_cq, IB_CQ_NEXT_COMP);
528 529

	init_attr->event_handler       = srp_qp_event;
530
	init_attr->cap.max_send_wr     = m * target->queue_size;
531
	init_attr->cap.max_recv_wr     = target->queue_size + 1;
532 533
	init_attr->cap.max_recv_sge    = 1;
	init_attr->cap.max_send_sge    = 1;
534
	init_attr->sq_sig_type         = IB_SIGNAL_REQ_WR;
535
	init_attr->qp_type             = IB_QPT_RC;
536 537
	init_attr->send_cq             = send_cq;
	init_attr->recv_cq             = recv_cq;
538

539
	qp = ib_create_qp(dev->pd, init_attr);
540 541
	if (IS_ERR(qp)) {
		ret = PTR_ERR(qp);
542
		goto err_send_cq;
543 544
	}

545
	ret = srp_init_qp(target, qp);
546 547
	if (ret)
		goto err_qp;
548

549
	if (dev->use_fast_reg) {
550 551 552 553 554 555 556
		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;
		}
557
	} else if (dev->use_fmr) {
558 559 560 561 562 563 564 565 566
		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;
		}
	}

567
	if (ch->qp)
568
		srp_destroy_qp(ch);
569 570 571 572
	if (ch->recv_cq)
		ib_destroy_cq(ch->recv_cq);
	if (ch->send_cq)
		ib_destroy_cq(ch->send_cq);
573

574 575 576
	ch->qp = qp;
	ch->recv_cq = recv_cq;
	ch->send_cq = send_cq;
577

578 579 580 581 582 583 584 585 586 587
	if (dev->use_fast_reg) {
		if (ch->fr_pool)
			srp_destroy_fr_pool(ch->fr_pool);
		ch->fr_pool = fr_pool;
	} else if (dev->use_fmr) {
		if (ch->fmr_pool)
			ib_destroy_fmr_pool(ch->fmr_pool);
		ch->fmr_pool = fmr_pool;
	}

588 589 590 591
	kfree(init_attr);
	return 0;

err_qp:
592
	ib_destroy_qp(qp);
593 594

err_send_cq:
595
	ib_destroy_cq(send_cq);
596 597

err_recv_cq:
598
	ib_destroy_cq(recv_cq);
599 600

err:
601 602 603 604
	kfree(init_attr);
	return ret;
}

605 606
/*
 * Note: this function may be called without srp_alloc_iu_bufs() having been
607
 * invoked. Hence the ch->[rt]x_ring checks.
608
 */
609 610
static void srp_free_ch_ib(struct srp_target_port *target,
			   struct srp_rdma_ch *ch)
611
{
612
	struct srp_device *dev = target->srp_host->srp_dev;
613 614
	int i;

B
Bart Van Assche 已提交
615 616 617
	if (!ch->target)
		return;

618 619 620
	if (ch->cm_id) {
		ib_destroy_cm_id(ch->cm_id);
		ch->cm_id = NULL;
621 622
	}

B
Bart Van Assche 已提交
623 624 625 626
	/* If srp_new_cm_id() succeeded but srp_create_ch_ib() not, return. */
	if (!ch->qp)
		return;

627
	if (dev->use_fast_reg) {
628 629
		if (ch->fr_pool)
			srp_destroy_fr_pool(ch->fr_pool);
630
	} else if (dev->use_fmr) {
631 632
		if (ch->fmr_pool)
			ib_destroy_fmr_pool(ch->fmr_pool);
633
	}
634
	srp_destroy_qp(ch);
635 636
	ib_destroy_cq(ch->send_cq);
	ib_destroy_cq(ch->recv_cq);
637

B
Bart Van Assche 已提交
638 639 640 641 642 643 644 645
	/*
	 * 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;

646 647
	ch->qp = NULL;
	ch->send_cq = ch->recv_cq = NULL;
648

649
	if (ch->rx_ring) {
650
		for (i = 0; i < target->queue_size; ++i)
651 652 653
			srp_free_iu(target->srp_host, ch->rx_ring[i]);
		kfree(ch->rx_ring);
		ch->rx_ring = NULL;
654
	}
655
	if (ch->tx_ring) {
656
		for (i = 0; i < target->queue_size; ++i)
657 658 659
			srp_free_iu(target->srp_host, ch->tx_ring[i]);
		kfree(ch->tx_ring);
		ch->tx_ring = NULL;
660
	}
661 662 663 664
}

static void srp_path_rec_completion(int status,
				    struct ib_sa_path_rec *pathrec,
665
				    void *ch_ptr)
666
{
667 668
	struct srp_rdma_ch *ch = ch_ptr;
	struct srp_target_port *target = ch->target;
669

670
	ch->status = status;
671
	if (status)
672 673
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Got failed path rec status %d\n", status);
674
	else
675 676
		ch->path = *pathrec;
	complete(&ch->done);
677 678
}

679
static int srp_lookup_path(struct srp_rdma_ch *ch)
680
{
681
	struct srp_target_port *target = ch->target;
682 683
	int ret;

684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704
	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);
705 706
	if (ret < 0)
		return ret;
707

708
	if (ch->status < 0)
709 710
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Path record query failed\n");
711

712
	return ch->status;
713 714
}

B
Bart Van Assche 已提交
715
static int srp_send_req(struct srp_rdma_ch *ch, bool multich)
716
{
717
	struct srp_target_port *target = ch->target;
718 719 720 721 722 723 724 725 726 727
	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;

728
	req->param.primary_path		      = &ch->path;
729 730
	req->param.alternate_path 	      = NULL;
	req->param.service_id 		      = target->service_id;
731 732
	req->param.qp_num		      = ch->qp->qp_num;
	req->param.qp_type		      = ch->qp->qp_type;
733 734 735 736 737 738 739 740 741 742 743 744 745 746
	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;
747
	req->param.retry_count                = target->tl_retry_count;
748 749 750 751 752
	req->param.rnr_retry_count 	      = 7;
	req->param.max_cm_retries 	      = 15;

	req->priv.opcode     	= SRP_LOGIN_REQ;
	req->priv.tag        	= 0;
753
	req->priv.req_it_iu_len = cpu_to_be32(target->max_iu_len);
754 755
	req->priv.req_buf_fmt 	= cpu_to_be16(SRP_BUF_FORMAT_DIRECT |
					      SRP_BUF_FORMAT_INDIRECT);
B
Bart Van Assche 已提交
756 757
	req->priv.req_flags	= (multich ? SRP_MULTICHAN_MULTI :
				   SRP_MULTICHAN_SINGLE);
758
	/*
R
Roland Dreier 已提交
759
	 * In the published SRP specification (draft rev. 16a), the
760 761 762 763 764 765 766 767 768
	 * 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,
769
		       &target->sgid.global.interface_id, 8);
770
		memcpy(req->priv.initiator_port_id + 8,
771
		       &target->initiator_ext, 8);
772 773 774 775
		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,
776 777
		       &target->initiator_ext, 8);
		memcpy(req->priv.initiator_port_id + 8,
778
		       &target->sgid.global.interface_id, 8);
779 780 781 782
		memcpy(req->priv.target_port_id,     &target->id_ext, 8);
		memcpy(req->priv.target_port_id + 8, &target->ioc_guid, 8);
	}

783 784
	/*
	 * Topspin/Cisco SRP targets will reject our login unless we
785 786
	 * zero out the first 8 bytes of our initiator port ID and set
	 * the second 8 bytes to the local node GUID.
787
	 */
788
	if (srp_target_is_topspin(target)) {
789 790 791
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Topspin/Cisco initiator port ID workaround "
			     "activated for target GUID %016llx\n",
792
			     be64_to_cpu(target->ioc_guid));
793
		memset(req->priv.initiator_port_id, 0, 8);
794
		memcpy(req->priv.initiator_port_id + 8,
795
		       &target->srp_host->srp_dev->dev->node_guid, 8);
796 797
	}

798
	status = ib_send_cm_req(ch->cm_id, &req->param);
799 800 801 802 803 804

	kfree(req);

	return status;
}

805 806 807 808 809 810 811 812 813 814 815 816
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)
817
		queue_work(srp_remove_wq, &target->remove_work);
818 819 820 821

	return changed;
}

822 823
static void srp_disconnect_target(struct srp_target_port *target)
{
B
Bart Van Assche 已提交
824 825
	struct srp_rdma_ch *ch;
	int i;
826

827
	/* XXX should send SRP_I_LOGOUT request */
828

829 830 831 832 833 834
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		ch->connected = false;
		if (ch->cm_id && ib_send_cm_dreq(ch->cm_id, NULL, 0)) {
			shost_printk(KERN_DEBUG, target->scsi_host,
				     PFX "Sending CM DREQ failed\n");
835
		}
836
	}
837 838
}

839 840
static void srp_free_req_data(struct srp_target_port *target,
			      struct srp_rdma_ch *ch)
841
{
842 843
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = dev->dev;
844 845 846
	struct srp_request *req;
	int i;

847
	if (!ch->req_ring)
848 849 850
		return;

	for (i = 0; i < target->req_ring_size; ++i) {
851
		req = &ch->req_ring[i];
852 853 854 855
		if (dev->use_fast_reg)
			kfree(req->fr_list);
		else
			kfree(req->fmr_list);
856
		kfree(req->map_page);
857 858 859 860 861 862
		if (req->indirect_dma_addr) {
			ib_dma_unmap_single(ibdev, req->indirect_dma_addr,
					    target->indirect_size,
					    DMA_TO_DEVICE);
		}
		kfree(req->indirect_desc);
863
	}
864

865 866
	kfree(ch->req_ring);
	ch->req_ring = NULL;
867 868
}

869
static int srp_alloc_req_data(struct srp_rdma_ch *ch)
870
{
871
	struct srp_target_port *target = ch->target;
872 873 874
	struct srp_device *srp_dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = srp_dev->dev;
	struct srp_request *req;
875
	void *mr_list;
876 877 878
	dma_addr_t dma_addr;
	int i, ret = -ENOMEM;

879 880 881
	ch->req_ring = kcalloc(target->req_ring_size, sizeof(*ch->req_ring),
			       GFP_KERNEL);
	if (!ch->req_ring)
882 883 884
		goto out;

	for (i = 0; i < target->req_ring_size; ++i) {
885
		req = &ch->req_ring[i];
886 887 888 889 890 891 892 893
		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;
894
		req->map_page = kmalloc(srp_dev->max_pages_per_mr *
895
					sizeof(void *), GFP_KERNEL);
896 897
		if (!req->map_page)
			goto out;
898
		req->indirect_desc = kmalloc(target->indirect_size, GFP_KERNEL);
899
		if (!req->indirect_desc)
900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915
			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;
}

916 917 918 919 920 921 922 923 924 925 926 927 928 929 930
/**
 * 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);
}

931 932
static void srp_remove_target(struct srp_target_port *target)
{
B
Bart Van Assche 已提交
933 934
	struct srp_rdma_ch *ch;
	int i;
935

936 937
	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);

938
	srp_del_scsi_host_attr(target->scsi_host);
939
	srp_rport_get(target->rport);
940 941
	srp_remove_host(target->scsi_host);
	scsi_remove_host(target->scsi_host);
942
	srp_stop_rport_timers(target->rport);
943
	srp_disconnect_target(target);
B
Bart Van Assche 已提交
944 945 946 947
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		srp_free_ch_ib(target, ch);
	}
948
	cancel_work_sync(&target->tl_err_work);
949
	srp_rport_put(target->rport);
B
Bart Van Assche 已提交
950 951 952 953 954 955
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		srp_free_req_data(target, ch);
	}
	kfree(target->ch);
	target->ch = NULL;
956 957 958 959 960

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

961 962 963
	scsi_host_put(target->scsi_host);
}

D
David Howells 已提交
964
static void srp_remove_work(struct work_struct *work)
965
{
D
David Howells 已提交
966
	struct srp_target_port *target =
967
		container_of(work, struct srp_target_port, remove_work);
968

969
	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);
970

971
	srp_remove_target(target);
972 973
}

974 975 976 977 978 979 980
static void srp_rport_delete(struct srp_rport *rport)
{
	struct srp_target_port *target = rport->lld_data;

	srp_queue_remove_work(target);
}

981 982 983 984 985 986 987 988 989 990 991 992 993 994
/**
 * srp_connected_ch() - number of connected channels
 * @target: SRP target port.
 */
static int srp_connected_ch(struct srp_target_port *target)
{
	int i, c = 0;

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

	return c;
}

B
Bart Van Assche 已提交
995
static int srp_connect_ch(struct srp_rdma_ch *ch, bool multich)
996
{
997
	struct srp_target_port *target = ch->target;
998 999
	int ret;

1000
	WARN_ON_ONCE(!multich && srp_connected_ch(target) > 0);
1001

1002
	ret = srp_lookup_path(ch);
1003 1004 1005 1006
	if (ret)
		return ret;

	while (1) {
1007
		init_completion(&ch->done);
B
Bart Van Assche 已提交
1008
		ret = srp_send_req(ch, multich);
1009 1010
		if (ret)
			return ret;
1011
		ret = wait_for_completion_interruptible(&ch->done);
1012 1013
		if (ret < 0)
			return ret;
1014 1015 1016 1017 1018 1019 1020

		/*
		 * 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.
		 */
1021
		switch (ch->status) {
1022
		case 0:
1023
			ch->connected = true;
1024 1025 1026
			return 0;

		case SRP_PORT_REDIRECT:
1027
			ret = srp_lookup_path(ch);
1028 1029 1030 1031 1032 1033 1034
			if (ret)
				return ret;
			break;

		case SRP_DLID_REDIRECT:
			break;

D
David Dillow 已提交
1035 1036
		case SRP_STALE_CONN:
			shost_printk(KERN_ERR, target->scsi_host, PFX
1037
				     "giving up on stale connection\n");
1038 1039
			ch->status = -ECONNRESET;
			return ch->status;
D
David Dillow 已提交
1040

1041
		default:
1042
			return ch->status;
1043 1044 1045 1046
		}
	}
}

1047
static int srp_inv_rkey(struct srp_rdma_ch *ch, u32 rkey)
1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058
{
	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,
	};

1059
	return ib_post_send(ch->qp, &wr, &bad_wr);
1060 1061
}

1062
static void srp_unmap_data(struct scsi_cmnd *scmnd,
1063
			   struct srp_rdma_ch *ch,
1064 1065
			   struct srp_request *req)
{
1066
	struct srp_target_port *target = ch->target;
1067 1068 1069
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = dev->dev;
	int i, res;
1070

1071
	if (!scsi_sglist(scmnd) ||
1072 1073 1074 1075
	    (scmnd->sc_data_direction != DMA_TO_DEVICE &&
	     scmnd->sc_data_direction != DMA_FROM_DEVICE))
		return;

1076 1077 1078 1079
	if (dev->use_fast_reg) {
		struct srp_fr_desc **pfr;

		for (i = req->nmdesc, pfr = req->fr_list; i > 0; i--, pfr++) {
1080
			res = srp_inv_rkey(ch, (*pfr)->mr->rkey);
1081 1082 1083 1084 1085 1086 1087 1088 1089
			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)
1090
			srp_fr_pool_put(ch->fr_pool, req->fr_list,
1091
					req->nmdesc);
1092
	} else if (dev->use_fmr) {
1093 1094 1095 1096 1097
		struct ib_pool_fmr **pfmr;

		for (i = req->nmdesc, pfmr = req->fmr_list; i > 0; i--, pfmr++)
			ib_fmr_pool_unmap(*pfmr);
	}
1098

1099 1100
	ib_dma_unmap_sg(ibdev, scsi_sglist(scmnd), scsi_sg_count(scmnd),
			scmnd->sc_data_direction);
1101 1102
}

B
Bart Van Assche 已提交
1103 1104
/**
 * srp_claim_req - Take ownership of the scmnd associated with a request.
1105
 * @ch: SRP RDMA channel.
B
Bart Van Assche 已提交
1106
 * @req: SRP request.
1107
 * @sdev: If not NULL, only take ownership for this SCSI device.
B
Bart Van Assche 已提交
1108 1109 1110 1111 1112 1113
 * @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.
 */
1114
static struct scsi_cmnd *srp_claim_req(struct srp_rdma_ch *ch,
B
Bart Van Assche 已提交
1115
				       struct srp_request *req,
1116
				       struct scsi_device *sdev,
B
Bart Van Assche 已提交
1117 1118 1119 1120
				       struct scsi_cmnd *scmnd)
{
	unsigned long flags;

1121
	spin_lock_irqsave(&ch->lock, flags);
1122 1123 1124
	if (req->scmnd &&
	    (!sdev || req->scmnd->device == sdev) &&
	    (!scmnd || req->scmnd == scmnd)) {
B
Bart Van Assche 已提交
1125 1126 1127 1128 1129
		scmnd = req->scmnd;
		req->scmnd = NULL;
	} else {
		scmnd = NULL;
	}
1130
	spin_unlock_irqrestore(&ch->lock, flags);
B
Bart Van Assche 已提交
1131 1132 1133 1134 1135 1136

	return scmnd;
}

/**
 * srp_free_req() - Unmap data and add request to the free request list.
1137
 * @ch:     SRP RDMA channel.
1138 1139 1140
 * @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
1141
 */
1142 1143
static void srp_free_req(struct srp_rdma_ch *ch, struct srp_request *req,
			 struct scsi_cmnd *scmnd, s32 req_lim_delta)
1144
{
1145 1146
	unsigned long flags;

1147
	srp_unmap_data(scmnd, ch, req);
B
Bart Van Assche 已提交
1148

1149 1150 1151
	spin_lock_irqsave(&ch->lock, flags);
	ch->req_lim += req_lim_delta;
	spin_unlock_irqrestore(&ch->lock, flags);
1152 1153
}

1154 1155
static void srp_finish_req(struct srp_rdma_ch *ch, struct srp_request *req,
			   struct scsi_device *sdev, int result)
1156
{
1157
	struct scsi_cmnd *scmnd = srp_claim_req(ch, req, sdev, NULL);
B
Bart Van Assche 已提交
1158 1159

	if (scmnd) {
1160
		srp_free_req(ch, req, scmnd, 0);
1161
		scmnd->result = result;
B
Bart Van Assche 已提交
1162 1163
		scmnd->scsi_done(scmnd);
	}
1164 1165
}

1166
static void srp_terminate_io(struct srp_rport *rport)
1167
{
1168
	struct srp_target_port *target = rport->lld_data;
B
Bart Van Assche 已提交
1169
	struct srp_rdma_ch *ch;
1170 1171
	struct Scsi_Host *shost = target->scsi_host;
	struct scsi_device *sdev;
B
Bart Van Assche 已提交
1172
	int i, j;
1173

1174 1175 1176 1177 1178 1179 1180
	/*
	 * 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
1181 1182
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
1183

B
Bart Van Assche 已提交
1184 1185 1186 1187 1188 1189
		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);
		}
1190 1191
	}
}
1192

1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204
/*
 * 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
1205 1206 1207
	struct srp_rdma_ch *ch;
	int i, j, ret = 0;
	bool multich = false;
1208

1209
	srp_disconnect_target(target);
1210 1211 1212 1213

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

1214
	/*
1215 1216 1217
	 * 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.
1218
	 */
B
Bart Van Assche 已提交
1219 1220 1221
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		ret += srp_new_cm_id(ch);
1222
	}
B
Bart Van Assche 已提交
1223 1224 1225 1226
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		for (j = 0; j < target->req_ring_size; ++j) {
			struct srp_request *req = &ch->req_ring[j];
1227

B
Bart Van Assche 已提交
1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238
			srp_finish_req(ch, req, NULL, DID_RESET << 16);
		}
	}
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		/*
		 * 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);
1239

B
Bart Van Assche 已提交
1240 1241 1242 1243
		INIT_LIST_HEAD(&ch->free_tx);
		for (j = 0; j < target->queue_size; ++j)
			list_add(&ch->tx_ring[j]->list, &ch->free_tx);
	}
1244 1245 1246

	target->qp_in_error = false;

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

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

	return ret;
}

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

1267 1268
	WARN_ON_ONCE(!dma_len);

1269 1270 1271
	desc->va = cpu_to_be64(dma_addr);
	desc->key = cpu_to_be32(rkey);
	desc->len = cpu_to_be32(dma_len);
1272

1273 1274 1275 1276
	state->total_len += dma_len;
	state->desc++;
	state->ndesc++;
}
1277

1278
static int srp_map_finish_fmr(struct srp_map_state *state,
1279
			      struct srp_rdma_ch *ch)
1280
{
1281 1282
	struct srp_target_port *target = ch->target;
	struct srp_device *dev = target->srp_host->srp_dev;
1283 1284
	struct ib_pool_fmr *fmr;
	u64 io_addr = 0;
1285

1286 1287 1288
	if (state->fmr.next >= state->fmr.end)
		return -ENOMEM;

1289
	fmr = ib_fmr_pool_map_phys(ch->fmr_pool, state->pages,
1290 1291 1292
				   state->npages, io_addr);
	if (IS_ERR(fmr))
		return PTR_ERR(fmr);
1293

1294
	*state->fmr.next++ = fmr;
1295
	state->nmdesc++;
1296

1297 1298
	srp_map_desc(state, state->base_dma_addr & ~dev->mr_page_mask,
		     state->dma_len, fmr->fmr->rkey);
1299

1300 1301 1302
	return 0;
}

1303
static int srp_map_finish_fr(struct srp_map_state *state,
1304
			     struct srp_rdma_ch *ch)
1305
{
1306
	struct srp_target_port *target = ch->target;
1307 1308 1309 1310 1311 1312
	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;

1313 1314 1315
	if (state->fr.next >= state->fr.end)
		return -ENOMEM;

1316
	desc = srp_fr_pool_get(ch->fr_pool);
1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338
	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;

1339
	*state->fr.next++ = desc;
1340 1341 1342 1343 1344
	state->nmdesc++;

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

1345
	return ib_post_send(ch->qp, &wr, &bad_wr);
1346 1347
}

1348
static int srp_finish_mapping(struct srp_map_state *state,
1349
			      struct srp_rdma_ch *ch)
1350
{
1351
	struct srp_target_port *target = ch->target;
1352
	struct srp_device *dev = target->srp_host->srp_dev;
1353 1354
	int ret = 0;

1355 1356
	WARN_ON_ONCE(!dev->use_fast_reg && !dev->use_fmr);

1357 1358 1359
	if (state->npages == 0)
		return 0;

1360
	if (state->npages == 1 && target->global_mr)
1361
		srp_map_desc(state, state->base_dma_addr, state->dma_len,
1362
			     target->global_mr->rkey);
1363
	else
1364
		ret = dev->use_fast_reg ? srp_map_finish_fr(state, ch) :
1365
			srp_map_finish_fmr(state, ch);
1366 1367 1368

	if (ret == 0) {
		state->npages = 0;
1369
		state->dma_len = 0;
1370 1371 1372 1373 1374
	}

	return ret;
}

1375
static int srp_map_sg_entry(struct srp_map_state *state,
1376
			    struct srp_rdma_ch *ch,
1377
			    struct scatterlist *sg, int sg_index)
1378
{
1379
	struct srp_target_port *target = ch->target;
1380 1381 1382 1383
	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);
1384
	unsigned int len = 0;
1385 1386
	int ret;

1387
	WARN_ON_ONCE(!dma_len);
1388

1389
	while (dma_len) {
1390 1391
		unsigned offset = dma_addr & ~dev->mr_page_mask;
		if (state->npages == dev->max_pages_per_mr || offset != 0) {
1392
			ret = srp_finish_mapping(state, ch);
1393 1394 1395 1396
			if (ret)
				return ret;
		}

1397
		len = min_t(unsigned int, dma_len, dev->mr_page_size - offset);
1398

1399 1400
		if (!state->npages)
			state->base_dma_addr = dma_addr;
1401
		state->pages[state->npages++] = dma_addr & dev->mr_page_mask;
1402
		state->dma_len += len;
1403 1404 1405 1406
		dma_addr += len;
		dma_len -= len;
	}

1407 1408
	/*
	 * If the last entry of the MR wasn't a full page, then we need to
1409 1410 1411 1412
	 * close it out and start a new one -- we can only merge at page
	 * boundries.
	 */
	ret = 0;
1413
	if (len != dev->mr_page_size)
1414
		ret = srp_finish_mapping(state, ch);
1415 1416 1417
	return ret;
}

1418 1419 1420
static int srp_map_sg(struct srp_map_state *state, struct srp_rdma_ch *ch,
		      struct srp_request *req, struct scatterlist *scat,
		      int count)
1421
{
1422
	struct srp_target_port *target = ch->target;
1423 1424
	struct srp_device *dev = target->srp_host->srp_dev;
	struct scatterlist *sg;
1425
	int i, ret;
1426 1427 1428

	state->desc	= req->indirect_desc;
	state->pages	= req->map_page;
1429
	if (dev->use_fast_reg) {
1430 1431
		state->fr.next = req->fr_list;
		state->fr.end = req->fr_list + target->cmd_sg_cnt;
1432
	} else if (dev->use_fmr) {
1433 1434
		state->fmr.next = req->fmr_list;
		state->fmr.end = req->fmr_list + target->cmd_sg_cnt;
1435
	}
1436

1437
	if (dev->use_fast_reg || dev->use_fmr) {
1438 1439 1440 1441 1442
		for_each_sg(scat, sg, count, i) {
			ret = srp_map_sg_entry(state, ch, sg, i);
			if (ret)
				goto out;
		}
1443 1444 1445
		ret = srp_finish_mapping(state, ch);
		if (ret)
			goto out;
1446 1447 1448
	} else {
		for_each_sg(scat, sg, count, i) {
			srp_map_desc(state, ib_sg_dma_address(dev->dev, sg),
1449 1450
				     ib_sg_dma_len(dev->dev, sg),
				     target->global_mr->rkey);
1451
		}
1452
	}
1453

1454
	req->nmdesc = state->nmdesc;
1455
	ret = 0;
1456

1457 1458
out:
	return ret;
1459 1460
}

1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499
/*
 * Register the indirect data buffer descriptor with the HCA.
 *
 * Note: since the indirect data buffer descriptor has been allocated with
 * kmalloc() it is guaranteed that this buffer is a physically contiguous
 * memory buffer.
 */
static int srp_map_idb(struct srp_rdma_ch *ch, struct srp_request *req,
		       void **next_mr, void **end_mr, u32 idb_len,
		       __be32 *idb_rkey)
{
	struct srp_target_port *target = ch->target;
	struct srp_device *dev = target->srp_host->srp_dev;
	struct srp_map_state state;
	struct srp_direct_buf idb_desc;
	u64 idb_pages[1];
	int ret;

	memset(&state, 0, sizeof(state));
	memset(&idb_desc, 0, sizeof(idb_desc));
	state.gen.next = next_mr;
	state.gen.end = end_mr;
	state.desc = &idb_desc;
	state.pages = idb_pages;
	state.pages[0] = (req->indirect_dma_addr &
			  dev->mr_page_mask);
	state.npages = 1;
	state.base_dma_addr = req->indirect_dma_addr;
	state.dma_len = idb_len;
	ret = srp_finish_mapping(&state, ch);
	if (ret < 0)
		goto out;

	*idb_rkey = idb_desc.key;

out:
	return ret;
}

1500
static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_rdma_ch *ch,
1501 1502
			struct srp_request *req)
{
1503
	struct srp_target_port *target = ch->target;
1504
	struct scatterlist *scat;
1505
	struct srp_cmd *cmd = req->cmd->buf;
1506
	int len, nents, count, ret;
1507 1508
	struct srp_device *dev;
	struct ib_device *ibdev;
1509 1510
	struct srp_map_state state;
	struct srp_indirect_buf *indirect_hdr;
1511 1512
	u32 idb_len, table_len;
	__be32 idb_rkey;
1513
	u8 fmt;
1514

1515
	if (!scsi_sglist(scmnd) || scmnd->sc_data_direction == DMA_NONE)
1516 1517 1518 1519
		return sizeof (struct srp_cmd);

	if (scmnd->sc_data_direction != DMA_FROM_DEVICE &&
	    scmnd->sc_data_direction != DMA_TO_DEVICE) {
1520 1521 1522
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled data direction %d\n",
			     scmnd->sc_data_direction);
1523 1524 1525
		return -EINVAL;
	}

1526 1527
	nents = scsi_sg_count(scmnd);
	scat  = scsi_sglist(scmnd);
1528

1529
	dev = target->srp_host->srp_dev;
1530 1531 1532
	ibdev = dev->dev;

	count = ib_dma_map_sg(ibdev, scat, nents, scmnd->sc_data_direction);
1533 1534
	if (unlikely(count == 0))
		return -EIO;
1535 1536 1537

	fmt = SRP_DATA_DESC_DIRECT;
	len = sizeof (struct srp_cmd) +	sizeof (struct srp_direct_buf);
1538

1539
	if (count == 1 && target->global_mr) {
1540 1541 1542 1543 1544 1545
		/*
		 * 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.
		 */
1546
		struct srp_direct_buf *buf = (void *) cmd->add_data;
1547

1548
		buf->va  = cpu_to_be64(ib_sg_dma_address(ibdev, scat));
1549
		buf->key = cpu_to_be32(target->global_mr->rkey);
1550
		buf->len = cpu_to_be32(ib_sg_dma_len(ibdev, scat));
1551

1552
		req->nmdesc = 0;
1553 1554 1555
		goto map_complete;
	}

1556 1557 1558
	/*
	 * We have more than one scatter/gather entry, so build our indirect
	 * descriptor table, trying to merge as many entries as we can.
1559 1560 1561
	 */
	indirect_hdr = (void *) cmd->add_data;

1562 1563 1564
	ib_dma_sync_single_for_cpu(ibdev, req->indirect_dma_addr,
				   target->indirect_size, DMA_TO_DEVICE);

1565
	memset(&state, 0, sizeof(state));
1566
	srp_map_sg(&state, ch, req, scat, count);
1567

1568 1569 1570 1571 1572
	/* 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.
1573 1574
	 */
	if (state.ndesc == 1) {
1575 1576
		/*
		 * Memory registration collapsed the sg-list into one entry,
1577 1578 1579
		 * so use a direct descriptor.
		 */
		struct srp_direct_buf *buf = (void *) cmd->add_data;
1580

1581
		*buf = req->indirect_desc[0];
1582
		goto map_complete;
1583 1584
	}

1585 1586 1587 1588 1589 1590 1591 1592
	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);
1593
	table_len = state.ndesc * sizeof (struct srp_direct_buf);
1594
	idb_len = sizeof(struct srp_indirect_buf) + table_len;
1595 1596 1597

	fmt = SRP_DATA_DESC_INDIRECT;
	len = sizeof(struct srp_cmd) + sizeof (struct srp_indirect_buf);
1598
	len += count * sizeof (struct srp_direct_buf);
1599

1600 1601
	memcpy(indirect_hdr->desc_list, req->indirect_desc,
	       count * sizeof (struct srp_direct_buf));
1602

1603
	if (!target->global_mr) {
1604 1605 1606 1607 1608 1609
		ret = srp_map_idb(ch, req, state.gen.next, state.gen.end,
				  idb_len, &idb_rkey);
		if (ret < 0)
			return ret;
		req->nmdesc++;
	} else {
1610
		idb_rkey = target->global_mr->rkey;
1611 1612
	}

1613
	indirect_hdr->table_desc.va = cpu_to_be64(req->indirect_dma_addr);
1614
	indirect_hdr->table_desc.key = idb_rkey;
1615 1616 1617 1618
	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)
1619
		cmd->data_out_desc_cnt = count;
1620
	else
1621 1622 1623 1624
		cmd->data_in_desc_cnt = count;

	ib_dma_sync_single_for_device(ibdev, req->indirect_dma_addr, table_len,
				      DMA_TO_DEVICE);
1625 1626

map_complete:
1627 1628 1629 1630 1631 1632 1633 1634
	if (scmnd->sc_data_direction == DMA_TO_DEVICE)
		cmd->buf_fmt = fmt << 4;
	else
		cmd->buf_fmt = fmt;

	return len;
}

1635 1636 1637
/*
 * Return an IU and possible credit to the free pool
 */
1638
static void srp_put_tx_iu(struct srp_rdma_ch *ch, struct srp_iu *iu,
1639 1640 1641 1642
			  enum srp_iu_type iu_type)
{
	unsigned long flags;

1643 1644
	spin_lock_irqsave(&ch->lock, flags);
	list_add(&iu->list, &ch->free_tx);
1645
	if (iu_type != SRP_IU_RSP)
1646 1647
		++ch->req_lim;
	spin_unlock_irqrestore(&ch->lock, flags);
1648 1649
}

1650
/*
1651
 * Must be called with ch->lock held to protect req_lim and free_tx.
1652
 * If IU is not sent, it must be returned using srp_put_tx_iu().
1653 1654 1655 1656 1657 1658 1659 1660 1661 1662
 *
 * 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.
 */
1663
static struct srp_iu *__srp_get_tx_iu(struct srp_rdma_ch *ch,
1664 1665
				      enum srp_iu_type iu_type)
{
1666
	struct srp_target_port *target = ch->target;
1667 1668 1669
	s32 rsv = (iu_type == SRP_IU_TSK_MGMT) ? 0 : SRP_TSK_MGMT_SQ_SIZE;
	struct srp_iu *iu;

1670
	srp_send_completion(ch->send_cq, ch);
1671

1672
	if (list_empty(&ch->free_tx))
1673 1674 1675
		return NULL;

	/* Initiator responses to target requests do not consume credits */
1676
	if (iu_type != SRP_IU_RSP) {
1677
		if (ch->req_lim <= rsv) {
1678 1679 1680 1681
			++target->zero_req_lim;
			return NULL;
		}

1682
		--ch->req_lim;
1683 1684
	}

1685
	iu = list_first_entry(&ch->free_tx, struct srp_iu, list);
1686
	list_del(&iu->list);
1687 1688 1689
	return iu;
}

1690
static int srp_post_send(struct srp_rdma_ch *ch, struct srp_iu *iu, int len)
1691
{
1692
	struct srp_target_port *target = ch->target;
1693 1694 1695 1696 1697
	struct ib_sge list;
	struct ib_send_wr wr, *bad_wr;

	list.addr   = iu->dma;
	list.length = len;
1698
	list.lkey   = target->lkey;
1699 1700

	wr.next       = NULL;
1701
	wr.wr_id      = (uintptr_t) iu;
1702 1703 1704 1705 1706
	wr.sg_list    = &list;
	wr.num_sge    = 1;
	wr.opcode     = IB_WR_SEND;
	wr.send_flags = IB_SEND_SIGNALED;

1707
	return ib_post_send(ch->qp, &wr, &bad_wr);
1708 1709
}

1710
static int srp_post_recv(struct srp_rdma_ch *ch, struct srp_iu *iu)
1711
{
1712
	struct srp_target_port *target = ch->target;
1713
	struct ib_recv_wr wr, *bad_wr;
1714
	struct ib_sge list;
1715 1716 1717

	list.addr   = iu->dma;
	list.length = iu->size;
1718
	list.lkey   = target->lkey;
1719 1720

	wr.next     = NULL;
1721
	wr.wr_id    = (uintptr_t) iu;
1722 1723 1724
	wr.sg_list  = &list;
	wr.num_sge  = 1;

1725
	return ib_post_recv(ch->qp, &wr, &bad_wr);
1726 1727
}

1728
static void srp_process_rsp(struct srp_rdma_ch *ch, struct srp_rsp *rsp)
1729
{
1730
	struct srp_target_port *target = ch->target;
1731 1732 1733 1734 1735
	struct srp_request *req;
	struct scsi_cmnd *scmnd;
	unsigned long flags;

	if (unlikely(rsp->tag & SRP_TAG_TSK_MGMT)) {
1736 1737 1738
		spin_lock_irqsave(&ch->lock, flags);
		ch->req_lim += be32_to_cpu(rsp->req_lim_delta);
		spin_unlock_irqrestore(&ch->lock, flags);
1739

1740
		ch->tsk_mgmt_status = -1;
1741
		if (be32_to_cpu(rsp->resp_data_len) >= 4)
1742 1743
			ch->tsk_mgmt_status = rsp->data[3];
		complete(&ch->tsk_mgmt_done);
1744
	} else {
B
Bart Van Assche 已提交
1745 1746 1747 1748 1749
		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
1750
		if (!scmnd) {
1751
			shost_printk(KERN_ERR, target->scsi_host,
B
Bart Van Assche 已提交
1752 1753
				     "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
1754

1755 1756 1757
			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
1758 1759 1760

			return;
		}
1761 1762 1763 1764 1765 1766 1767 1768 1769
		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
1770
		if (unlikely(rsp->flags & SRP_RSP_FLAG_DIUNDER))
1771
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_in_res_cnt));
B
Bart Van Assche 已提交
1772 1773 1774 1775 1776 1777
		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));
1778

1779
		srp_free_req(ch, req, scmnd,
B
Bart Van Assche 已提交
1780 1781
			     be32_to_cpu(rsp->req_lim_delta));

1782 1783
		scmnd->host_scribble = NULL;
		scmnd->scsi_done(scmnd);
1784 1785 1786
	}
}

1787
static int srp_response_common(struct srp_rdma_ch *ch, s32 req_delta,
1788 1789
			       void *rsp, int len)
{
1790
	struct srp_target_port *target = ch->target;
1791
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1792 1793
	unsigned long flags;
	struct srp_iu *iu;
1794
	int err;
1795

1796 1797 1798 1799
	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);
1800

1801 1802 1803
	if (!iu) {
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "no IU available to send response\n");
1804
		return 1;
1805 1806 1807 1808 1809 1810
	}

	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);

1811
	err = srp_post_send(ch, iu, len);
1812
	if (err) {
1813 1814
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "unable to post response: %d\n", err);
1815
		srp_put_tx_iu(ch, iu, SRP_IU_RSP);
1816
	}
1817 1818 1819 1820

	return err;
}

1821
static void srp_process_cred_req(struct srp_rdma_ch *ch,
1822 1823 1824 1825 1826 1827 1828 1829
				 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);

1830 1831
	if (srp_response_common(ch, delta, &rsp, sizeof(rsp)))
		shost_printk(KERN_ERR, ch->target->scsi_host, PFX
1832 1833 1834
			     "problems processing SRP_CRED_REQ\n");
}

1835
static void srp_process_aer_req(struct srp_rdma_ch *ch,
1836 1837
				struct srp_aer_req *req)
{
1838
	struct srp_target_port *target = ch->target;
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
B
Bart Van Assche 已提交
1846
		     "ignoring AER for LUN %llu\n", scsilun_to_int(&req->lun));
1847

1848
	if (srp_response_common(ch, delta, &rsp, sizeof(rsp)))
1849 1850 1851 1852
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "problems processing SRP_AER_REQ\n");
}

1853
static void srp_handle_recv(struct srp_rdma_ch *ch, struct ib_wc *wc)
1854
{
1855
	struct srp_target_port *target = ch->target;
1856
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1857
	struct srp_iu *iu = (struct srp_iu *) (uintptr_t) wc->wr_id;
1858
	int res;
1859 1860
	u8 opcode;

1861
	ib_dma_sync_single_for_cpu(dev, iu->dma, ch->max_ti_iu_len,
1862
				   DMA_FROM_DEVICE);
1863 1864 1865 1866

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

	if (0) {
1867 1868
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "recv completion, opcode 0x%02x\n", opcode);
B
Bart Van Assche 已提交
1869 1870
		print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET, 8, 1,
			       iu->buf, wc->byte_len, true);
1871 1872 1873 1874
	}

	switch (opcode) {
	case SRP_RSP:
1875
		srp_process_rsp(ch, iu->buf);
1876 1877
		break;

1878
	case SRP_CRED_REQ:
1879
		srp_process_cred_req(ch, iu->buf);
1880 1881 1882
		break;

	case SRP_AER_REQ:
1883
		srp_process_aer_req(ch, iu->buf);
1884 1885
		break;

1886 1887
	case SRP_T_LOGOUT:
		/* XXX Handle target logout */
1888 1889
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Got target logout request\n");
1890 1891 1892
		break;

	default:
1893 1894
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled SRP opcode 0x%02x\n", opcode);
1895 1896 1897
		break;
	}

1898
	ib_dma_sync_single_for_device(dev, iu->dma, ch->max_ti_iu_len,
1899
				      DMA_FROM_DEVICE);
1900

1901
	res = srp_post_recv(ch, iu);
1902 1903 1904
	if (res != 0)
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Recv failed with error code %d\n", res);
1905 1906
}

1907 1908
/**
 * srp_tl_err_work() - handle a transport layer error
1909
 * @work: Work structure embedded in an SRP target port.
1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922
 *
 * 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);
}

1923
static void srp_handle_qp_err(u64 wr_id, enum ib_wc_status wc_status,
1924
			      bool send_err, struct srp_rdma_ch *ch)
1925
{
1926 1927 1928 1929 1930 1931 1932
	struct srp_target_port *target = ch->target;

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

1933
	if (ch->connected && !target->qp_in_error) {
1934 1935
		if (wr_id & LOCAL_INV_WR_ID_MASK) {
			shost_printk(KERN_ERR, target->scsi_host, PFX
1936 1937
				     "LOCAL_INV failed with status %s (%d)\n",
				     ib_wc_status_msg(wc_status), wc_status);
1938 1939
		} else if (wr_id & FAST_REG_WR_ID_MASK) {
			shost_printk(KERN_ERR, target->scsi_host, PFX
1940 1941
				     "FAST_REG_MR failed status %s (%d)\n",
				     ib_wc_status_msg(wc_status), wc_status);
1942 1943
		} else {
			shost_printk(KERN_ERR, target->scsi_host,
1944
				     PFX "failed %s status %s (%d) for iu %p\n",
1945
				     send_err ? "send" : "receive",
1946 1947
				     ib_wc_status_msg(wc_status), wc_status,
				     (void *)(uintptr_t)wr_id);
1948
		}
1949
		queue_work(system_long_wq, &target->tl_err_work);
1950
	}
1951 1952 1953
	target->qp_in_error = true;
}

1954
static void srp_recv_completion(struct ib_cq *cq, void *ch_ptr)
1955
{
1956
	struct srp_rdma_ch *ch = ch_ptr;
1957 1958 1959 1960
	struct ib_wc wc;

	ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
	while (ib_poll_cq(cq, 1, &wc) > 0) {
1961
		if (likely(wc.status == IB_WC_SUCCESS)) {
1962
			srp_handle_recv(ch, &wc);
1963
		} else {
1964
			srp_handle_qp_err(wc.wr_id, wc.status, false, ch);
1965
		}
1966 1967 1968
	}
}

1969
static void srp_send_completion(struct ib_cq *cq, void *ch_ptr)
1970
{
1971
	struct srp_rdma_ch *ch = ch_ptr;
1972
	struct ib_wc wc;
1973
	struct srp_iu *iu;
1974 1975

	while (ib_poll_cq(cq, 1, &wc) > 0) {
1976 1977
		if (likely(wc.status == IB_WC_SUCCESS)) {
			iu = (struct srp_iu *) (uintptr_t) wc.wr_id;
1978
			list_add(&iu->list, &ch->free_tx);
1979
		} else {
1980
			srp_handle_qp_err(wc.wr_id, wc.status, true, ch);
1981
		}
1982 1983 1984
	}
}

1985
static int srp_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *scmnd)
1986
{
1987
	struct srp_target_port *target = host_to_target(shost);
1988
	struct srp_rport *rport = target->rport;
1989
	struct srp_rdma_ch *ch;
1990 1991 1992
	struct srp_request *req;
	struct srp_iu *iu;
	struct srp_cmd *cmd;
1993
	struct ib_device *dev;
1994
	unsigned long flags;
B
Bart Van Assche 已提交
1995 1996
	u32 tag;
	u16 idx;
1997
	int len, ret;
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
	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);
2008

2009 2010 2011
	scmnd->result = srp_chkready(target->rport);
	if (unlikely(scmnd->result))
		goto err;
2012

B
Bart Van Assche 已提交
2013 2014
	WARN_ON_ONCE(scmnd->request->tag < 0);
	tag = blk_mq_unique_tag(scmnd->request);
B
Bart Van Assche 已提交
2015
	ch = &target->ch[blk_mq_unique_tag_to_hwq(tag)];
B
Bart Van Assche 已提交
2016 2017 2018 2019
	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);
2020 2021 2022 2023

	spin_lock_irqsave(&ch->lock, flags);
	iu = __srp_get_tx_iu(ch, SRP_IU_CMD);
	spin_unlock_irqrestore(&ch->lock, flags);
2024

B
Bart Van Assche 已提交
2025 2026 2027 2028
	if (!iu)
		goto err;

	req = &ch->req_ring[idx];
2029
	dev = target->srp_host->srp_dev->dev;
2030
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_iu_len,
2031
				   DMA_TO_DEVICE);
2032

2033
	scmnd->host_scribble = (void *) req;
2034 2035 2036 2037 2038

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

	cmd->opcode = SRP_CMD;
B
Bart Van Assche 已提交
2039
	int_to_scsilun(scmnd->device->lun, &cmd->lun);
B
Bart Van Assche 已提交
2040
	cmd->tag    = tag;
2041 2042 2043 2044 2045
	memcpy(cmd->cdb, scmnd->cmnd, scmnd->cmd_len);

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

2046
	len = srp_map_data(scmnd, ch, req);
2047
	if (len < 0) {
2048
		shost_printk(KERN_ERR, target->scsi_host,
2049 2050 2051 2052
			     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
2053
		 * max_pages_per_mr sg-list elements, tell the SCSI mid-layer
2054 2055 2056 2057
		 * to reduce queue depth temporarily.
		 */
		scmnd->result = len == -ENOMEM ?
			DID_OK << 16 | QUEUE_FULL << 1 : DID_ERROR << 16;
2058
		goto err_iu;
2059 2060
	}

2061
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_iu_len,
2062
				      DMA_TO_DEVICE);
2063

2064
	if (srp_post_send(ch, iu, len)) {
2065
		shost_printk(KERN_ERR, target->scsi_host, PFX "Send failed\n");
2066 2067 2068
		goto err_unmap;
	}

2069 2070
	ret = 0;

2071 2072 2073 2074
unlock_rport:
	if (in_scsi_eh)
		mutex_unlock(&rport->mutex);

2075
	return ret;
2076 2077

err_unmap:
2078
	srp_unmap_data(scmnd, ch, req);
2079

2080
err_iu:
2081
	srp_put_tx_iu(ch, iu, SRP_IU_CMD);
2082

2083 2084 2085 2086 2087 2088
	/*
	 * Avoid that the loops that iterate over the request ring can
	 * encounter a dangling SCSI command pointer.
	 */
	req->scmnd = NULL;

2089 2090 2091 2092 2093 2094 2095
err:
	if (scmnd->result) {
		scmnd->scsi_done(scmnd);
		ret = 0;
	} else {
		ret = SCSI_MLQUEUE_HOST_BUSY;
	}
2096

2097
	goto unlock_rport;
2098 2099
}

2100 2101
/*
 * Note: the resources allocated in this function are freed in
2102
 * srp_free_ch_ib().
2103
 */
2104
static int srp_alloc_iu_bufs(struct srp_rdma_ch *ch)
2105
{
2106
	struct srp_target_port *target = ch->target;
2107 2108
	int i;

2109 2110 2111
	ch->rx_ring = kcalloc(target->queue_size, sizeof(*ch->rx_ring),
			      GFP_KERNEL);
	if (!ch->rx_ring)
2112
		goto err_no_ring;
2113 2114 2115
	ch->tx_ring = kcalloc(target->queue_size, sizeof(*ch->tx_ring),
			      GFP_KERNEL);
	if (!ch->tx_ring)
2116 2117 2118
		goto err_no_ring;

	for (i = 0; i < target->queue_size; ++i) {
2119 2120 2121 2122
		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])
2123 2124 2125
			goto err;
	}

2126
	for (i = 0; i < target->queue_size; ++i) {
2127 2128 2129 2130
		ch->tx_ring[i] = srp_alloc_iu(target->srp_host,
					      target->max_iu_len,
					      GFP_KERNEL, DMA_TO_DEVICE);
		if (!ch->tx_ring[i])
2131
			goto err;
2132

2133
		list_add(&ch->tx_ring[i]->list, &ch->free_tx);
2134 2135 2136 2137 2138
	}

	return 0;

err:
2139
	for (i = 0; i < target->queue_size; ++i) {
2140 2141
		srp_free_iu(target->srp_host, ch->rx_ring[i]);
		srp_free_iu(target->srp_host, ch->tx_ring[i]);
2142 2143
	}

2144 2145

err_no_ring:
2146 2147 2148 2149
	kfree(ch->tx_ring);
	ch->tx_ring = NULL;
	kfree(ch->rx_ring);
	ch->rx_ring = NULL;
2150

2151 2152 2153
	return -ENOMEM;
}

2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180
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;
}

2181
static void srp_cm_rep_handler(struct ib_cm_id *cm_id,
2182
			       const struct srp_login_rsp *lrsp,
2183
			       struct srp_rdma_ch *ch)
2184
{
2185
	struct srp_target_port *target = ch->target;
2186 2187 2188 2189 2190 2191
	struct ib_qp_attr *qp_attr = NULL;
	int attr_mask = 0;
	int ret;
	int i;

	if (lrsp->opcode == SRP_LOGIN_RSP) {
2192 2193
		ch->max_ti_iu_len = be32_to_cpu(lrsp->max_ti_iu_len);
		ch->req_lim       = be32_to_cpu(lrsp->req_lim_delta);
2194 2195 2196 2197 2198 2199

		/*
		 * Reserve credits for task management so we don't
		 * bounce requests back to the SCSI mid-layer.
		 */
		target->scsi_host->can_queue
2200
			= min(ch->req_lim - SRP_TSK_MGMT_SQ_SIZE,
2201
			      target->scsi_host->can_queue);
2202 2203 2204
		target->scsi_host->cmd_per_lun
			= min_t(int, target->scsi_host->can_queue,
				target->scsi_host->cmd_per_lun);
2205 2206 2207 2208 2209 2210 2211
	} else {
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled RSP opcode %#x\n", lrsp->opcode);
		ret = -ECONNRESET;
		goto error;
	}

2212 2213
	if (!ch->rx_ring) {
		ret = srp_alloc_iu_bufs(ch);
2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227
		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;

2228
	ret = ib_modify_qp(ch->qp, qp_attr, attr_mask);
2229 2230 2231
	if (ret)
		goto error_free;

2232
	for (i = 0; i < target->queue_size; i++) {
2233 2234 2235
		struct srp_iu *iu = ch->rx_ring[i];

		ret = srp_post_recv(ch, iu);
2236 2237 2238 2239 2240 2241 2242 2243 2244
		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;

2245 2246
	target->rq_tmo_jiffies = srp_compute_rq_tmo(qp_attr, attr_mask);

2247
	ret = ib_modify_qp(ch->qp, qp_attr, attr_mask);
2248 2249 2250 2251 2252 2253 2254 2255 2256
	if (ret)
		goto error_free;

	ret = ib_send_cm_rtu(cm_id, NULL, 0);

error_free:
	kfree(qp_attr);

error:
2257
	ch->status = ret;
2258 2259
}

2260 2261
static void srp_cm_rej_handler(struct ib_cm_id *cm_id,
			       struct ib_cm_event *event,
2262
			       struct srp_rdma_ch *ch)
2263
{
2264
	struct srp_target_port *target = ch->target;
2265
	struct Scsi_Host *shost = target->scsi_host;
2266 2267 2268 2269 2270 2271
	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;
2272 2273
		ch->path.dlid = cpi->redirect_lid;
		ch->path.pkey = cpi->redirect_pkey;
2274
		cm_id->remote_cm_qpn = be32_to_cpu(cpi->redirect_qp) & 0x00ffffff;
2275
		memcpy(ch->path.dgid.raw, cpi->redirect_gid, 16);
2276

2277
		ch->status = ch->path.dlid ?
2278 2279 2280 2281
			SRP_DLID_REDIRECT : SRP_PORT_REDIRECT;
		break;

	case IB_CM_REJ_PORT_REDIRECT:
2282
		if (srp_target_is_topspin(target)) {
2283 2284 2285 2286 2287
			/*
			 * Topspin/Cisco SRP gateways incorrectly send
			 * reject reason code 25 when they mean 24
			 * (port redirect).
			 */
2288
			memcpy(ch->path.dgid.raw,
2289 2290
			       event->param.rej_rcvd.ari, 16);

2291 2292
			shost_printk(KERN_DEBUG, shost,
				     PFX "Topspin/Cisco redirect to target port GID %016llx%016llx\n",
2293 2294
				     be64_to_cpu(ch->path.dgid.global.subnet_prefix),
				     be64_to_cpu(ch->path.dgid.global.interface_id));
2295

2296
			ch->status = SRP_PORT_REDIRECT;
2297
		} else {
2298 2299
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_PORT_REDIRECT\n");
2300
			ch->status = -ECONNRESET;
2301 2302 2303 2304
		}
		break;

	case IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID:
2305 2306
		shost_printk(KERN_WARNING, shost,
			    "  REJ reason: IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID\n");
2307
		ch->status = -ECONNRESET;
2308 2309 2310 2311 2312 2313 2314 2315 2316
		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)
2317 2318
				shost_printk(KERN_WARNING, shost,
					     PFX "SRP_LOGIN_REJ: requested max_it_iu_len too large\n");
2319
			else
B
Bart Van Assche 已提交
2320 2321
				shost_printk(KERN_WARNING, shost, PFX
					     "SRP LOGIN from %pI6 to %pI6 REJECTED, reason 0x%08x\n",
2322 2323
					     target->sgid.raw,
					     target->orig_dgid.raw, reason);
2324
		} else
2325 2326 2327
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_CONSUMER_DEFINED,"
				     " opcode 0x%02x\n", opcode);
2328
		ch->status = -ECONNRESET;
2329 2330
		break;

D
David Dillow 已提交
2331 2332
	case IB_CM_REJ_STALE_CONN:
		shost_printk(KERN_WARNING, shost, "  REJ reason: stale connection\n");
2333
		ch->status = SRP_STALE_CONN;
D
David Dillow 已提交
2334 2335
		break;

2336
	default:
2337 2338
		shost_printk(KERN_WARNING, shost, "  REJ reason 0x%x\n",
			     event->param.rej_rcvd.reason);
2339
		ch->status = -ECONNRESET;
2340 2341 2342 2343 2344
	}
}

static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event)
{
2345 2346
	struct srp_rdma_ch *ch = cm_id->context;
	struct srp_target_port *target = ch->target;
2347 2348 2349 2350
	int comp = 0;

	switch (event->event) {
	case IB_CM_REQ_ERROR:
2351 2352
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Sending CM REQ failed\n");
2353
		comp = 1;
2354
		ch->status = -ECONNRESET;
2355 2356 2357 2358
		break;

	case IB_CM_REP_RECEIVED:
		comp = 1;
2359
		srp_cm_rep_handler(cm_id, event->private_data, ch);
2360 2361 2362
		break;

	case IB_CM_REJ_RECEIVED:
2363
		shost_printk(KERN_DEBUG, target->scsi_host, PFX "REJ received\n");
2364 2365
		comp = 1;

2366
		srp_cm_rej_handler(cm_id, event, ch);
2367 2368
		break;

2369
	case IB_CM_DREQ_RECEIVED:
2370 2371
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "DREQ received - connection closed\n");
2372
		ch->connected = false;
2373
		if (ib_send_cm_drep(cm_id, NULL, 0))
2374 2375
			shost_printk(KERN_ERR, target->scsi_host,
				     PFX "Sending CM DREP failed\n");
2376
		queue_work(system_long_wq, &target->tl_err_work);
2377 2378 2379
		break;

	case IB_CM_TIMEWAIT_EXIT:
2380 2381
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "connection closed\n");
2382
		comp = 1;
2383

2384
		ch->status = 0;
2385 2386
		break;

2387 2388 2389 2390 2391
	case IB_CM_MRA_RECEIVED:
	case IB_CM_DREQ_ERROR:
	case IB_CM_DREP_RECEIVED:
		break;

2392
	default:
2393 2394
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled CM event %d\n", event->event);
2395 2396 2397 2398
		break;
	}

	if (comp)
2399
		complete(&ch->done);
2400 2401 2402 2403

	return 0;
}

2404 2405 2406 2407 2408 2409 2410 2411
/**
 * srp_change_queue_depth - setting device queue depth
 * @sdev: scsi device struct
 * @qdepth: requested queue depth
 *
 * Returns queue depth.
 */
static int
2412
srp_change_queue_depth(struct scsi_device *sdev, int qdepth)
2413
{
2414
	if (!sdev->tagged_supported)
2415
		qdepth = 1;
2416
	return scsi_change_queue_depth(sdev, qdepth);
2417 2418
}

B
Bart Van Assche 已提交
2419 2420
static int srp_send_tsk_mgmt(struct srp_rdma_ch *ch, u64 req_tag, u64 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
	if (!ch->connected || target->qp_in_error)
2429 2430
		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;
B
Bart Van Assche 已提交
2454
	int_to_scsilun(lun, &tsk_mgmt->lun);
2455
	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
	return sprintf(buf, "0x%016llx\n", be64_to_cpu(target->id_ext));
2569 2570
}

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

2576
	return sprintf(buf, "0x%016llx\n", be64_to_cpu(target->ioc_guid));
2577 2578
}

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

2584
	return sprintf(buf, "0x%016llx\n", be64_to_cpu(target->service_id));
2585 2586
}

2587 2588
static ssize_t show_pkey(struct device *dev, struct device_attribute *attr,
			 char *buf)
2589
{
2590
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2591

2592
	return sprintf(buf, "0x%04x\n", be16_to_cpu(target->pkey));
2593 2594
}

B
Bart Van Assche 已提交
2595 2596 2597 2598 2599
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));

2600
	return sprintf(buf, "%pI6\n", target->sgid.raw);
B
Bart Van Assche 已提交
2601 2602
}

2603 2604
static ssize_t show_dgid(struct device *dev, struct device_attribute *attr,
			 char *buf)
2605
{
2606
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
B
Bart Van Assche 已提交
2607
	struct srp_rdma_ch *ch = &target->ch[0];
2608

2609
	return sprintf(buf, "%pI6\n", ch->path.dgid.raw);
2610 2611
}

2612 2613
static ssize_t show_orig_dgid(struct device *dev,
			      struct device_attribute *attr, char *buf)
2614
{
2615
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2616

2617
	return sprintf(buf, "%pI6\n", target->orig_dgid.raw);
2618 2619
}

2620 2621 2622 2623
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
2624 2625
	struct srp_rdma_ch *ch;
	int i, req_lim = INT_MAX;
2626

B
Bart Van Assche 已提交
2627 2628 2629 2630 2631
	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);
2632 2633
}

2634 2635
static ssize_t show_zero_req_lim(struct device *dev,
				 struct device_attribute *attr, char *buf)
2636
{
2637
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2638 2639 2640 2641

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

2642 2643
static ssize_t show_local_ib_port(struct device *dev,
				  struct device_attribute *attr, char *buf)
2644
{
2645
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2646 2647 2648 2649

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

2650 2651
static ssize_t show_local_ib_device(struct device *dev,
				    struct device_attribute *attr, char *buf)
2652
{
2653
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2654

2655
	return sprintf(buf, "%s\n", target->srp_host->srp_dev->dev->name);
2656 2657
}

B
Bart Van Assche 已提交
2658 2659 2660 2661 2662 2663 2664 2665
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);
}

2666 2667 2668 2669 2670 2671 2672 2673
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);
}

2674 2675 2676 2677 2678 2679 2680 2681
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);
}

2682 2683 2684 2685 2686 2687 2688 2689
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);
}

2690 2691 2692 2693 2694 2695 2696 2697
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");
}

2698 2699 2700 2701
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
2702
static DEVICE_ATTR(sgid,	    S_IRUGO, show_sgid,		   NULL);
2703 2704
static DEVICE_ATTR(dgid,	    S_IRUGO, show_dgid,		   NULL);
static DEVICE_ATTR(orig_dgid,	    S_IRUGO, show_orig_dgid,	   NULL);
2705
static DEVICE_ATTR(req_lim,         S_IRUGO, show_req_lim,         NULL);
2706 2707 2708
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
2709
static DEVICE_ATTR(ch_count,        S_IRUGO, show_ch_count,        NULL);
2710
static DEVICE_ATTR(comp_vector,     S_IRUGO, show_comp_vector,     NULL);
2711
static DEVICE_ATTR(tl_retry_count,  S_IRUGO, show_tl_retry_count,  NULL);
2712
static DEVICE_ATTR(cmd_sg_entries,  S_IRUGO, show_cmd_sg_entries,  NULL);
2713
static DEVICE_ATTR(allow_ext_sg,    S_IRUGO, show_allow_ext_sg,    NULL);
2714 2715 2716 2717 2718 2719

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
2720
	&dev_attr_sgid,
2721 2722
	&dev_attr_dgid,
	&dev_attr_orig_dgid,
2723
	&dev_attr_req_lim,
2724 2725 2726
	&dev_attr_zero_req_lim,
	&dev_attr_local_ib_port,
	&dev_attr_local_ib_device,
B
Bart Van Assche 已提交
2727
	&dev_attr_ch_count,
2728
	&dev_attr_comp_vector,
2729
	&dev_attr_tl_retry_count,
2730
	&dev_attr_cmd_sg_entries,
2731
	&dev_attr_allow_ext_sg,
2732 2733 2734
	NULL
};

2735 2736
static struct scsi_host_template srp_template = {
	.module				= THIS_MODULE,
R
Roland Dreier 已提交
2737 2738
	.name				= "InfiniBand SRP initiator",
	.proc_name			= DRV_NAME,
2739
	.slave_configure		= srp_slave_configure,
2740 2741
	.info				= srp_target_info,
	.queuecommand			= srp_queuecommand,
2742
	.change_queue_depth             = srp_change_queue_depth,
2743 2744 2745
	.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
2746
	.skip_settle_delay		= true,
2747
	.sg_tablesize			= SRP_DEF_SG_TABLESIZE,
2748
	.can_queue			= SRP_DEFAULT_CMD_SQ_SIZE,
2749
	.this_id			= -1,
2750
	.cmd_per_lun			= SRP_DEFAULT_CMD_SQ_SIZE,
2751
	.use_clustering			= ENABLE_CLUSTERING,
B
Bart Van Assche 已提交
2752 2753
	.shost_attrs			= srp_host_attrs,
	.use_blk_tags			= 1,
2754
	.track_queue_depth		= 1,
2755 2756
};

2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767
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;
}

2768 2769 2770 2771 2772 2773 2774
/*
 * Return values:
 * < 0 upon failure. Caller is responsible for SRP target port cleanup.
 * 0 and target->state == SRP_TARGET_REMOVED if asynchronous target port
 *    removal has been scheduled.
 * 0 and target->state != SRP_TARGET_REMOVED upon success.
 */
2775 2776
static int srp_add_target(struct srp_host *host, struct srp_target_port *target)
{
2777 2778 2779
	struct srp_rport_identifiers ids;
	struct srp_rport *rport;

2780
	target->state = SRP_TARGET_SCANNING;
2781
	sprintf(target->target_name, "SRP.T10:%016llX",
2782
		be64_to_cpu(target->id_ext));
2783

2784
	if (scsi_add_host(target->scsi_host, host->srp_dev->dev->dma_device))
2785 2786
		return -ENODEV;

2787 2788
	memcpy(ids.port_id, &target->id_ext, 8);
	memcpy(ids.port_id + 8, &target->ioc_guid, 8);
2789
	ids.roles = SRP_RPORT_ROLE_TARGET;
2790 2791 2792 2793 2794 2795
	rport = srp_rport_add(target->scsi_host, &ids);
	if (IS_ERR(rport)) {
		scsi_remove_host(target->scsi_host);
		return PTR_ERR(rport);
	}

2796
	rport->lld_data = target;
2797
	target->rport = rport;
2798

2799
	spin_lock(&host->target_lock);
2800
	list_add_tail(&target->list, &host->target_list);
2801
	spin_unlock(&host->target_lock);
2802 2803

	scsi_scan_target(&target->scsi_host->shost_gendev,
2804
			 0, target->scsi_id, SCAN_WILD_CARD, 0);
2805

2806 2807
	if (srp_connected_ch(target) < target->ch_count ||
	    target->qp_in_error) {
2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823
		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:
2824 2825 2826
	return 0;
}

2827
static void srp_release_dev(struct device *dev)
2828 2829
{
	struct srp_host *host =
2830
		container_of(dev, struct srp_host, dev);
2831 2832 2833 2834 2835 2836

	complete(&host->released);
}

static struct class srp_class = {
	.name    = "infiniband_srp",
2837
	.dev_release = srp_release_dev
2838 2839
};

2840 2841
/**
 * srp_conn_unique() - check whether the connection to a target is unique
2842 2843
 * @host:   SRP host.
 * @target: SRP target port.
2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871
 */
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;
}

2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887
/*
 * 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,
2888
	SRP_OPT_MAX_CMD_PER_LUN	= 1 << 6,
2889
	SRP_OPT_IO_CLASS	= 1 << 7,
2890
	SRP_OPT_INITIATOR_EXT	= 1 << 8,
2891
	SRP_OPT_CMD_SG_ENTRIES	= 1 << 9,
2892 2893
	SRP_OPT_ALLOW_EXT_SG	= 1 << 10,
	SRP_OPT_SG_TABLESIZE	= 1 << 11,
2894
	SRP_OPT_COMP_VECTOR	= 1 << 12,
2895
	SRP_OPT_TL_RETRY_COUNT	= 1 << 13,
2896
	SRP_OPT_QUEUE_SIZE	= 1 << 14,
2897 2898 2899 2900 2901 2902 2903
	SRP_OPT_ALL		= (SRP_OPT_ID_EXT	|
				   SRP_OPT_IOC_GUID	|
				   SRP_OPT_DGID		|
				   SRP_OPT_PKEY		|
				   SRP_OPT_SERVICE_ID),
};

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

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;
2939
	while ((p = strsep(&sep_opt, ",\n")) != NULL) {
2940 2941 2942 2943 2944 2945 2946 2947 2948
		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);
2949 2950 2951 2952
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2953 2954 2955 2956 2957 2958
			target->id_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_IOC_GUID:
			p = match_strdup(args);
2959 2960 2961 2962
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2963 2964 2965 2966 2967 2968
			target->ioc_guid = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_DGID:
			p = match_strdup(args);
2969 2970 2971 2972
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2973
			if (strlen(p) != 32) {
2974
				pr_warn("bad dest GID parameter '%s'\n", p);
2975
				kfree(p);
2976 2977 2978 2979
				goto out;
			}

			for (i = 0; i < 16; ++i) {
2980 2981 2982 2983 2984 2985 2986
				strlcpy(dgid, p + i * 2, sizeof(dgid));
				if (sscanf(dgid, "%hhx",
					   &target->orig_dgid.raw[i]) < 1) {
					ret = -EINVAL;
					kfree(p);
					goto out;
				}
2987
			}
2988
			kfree(p);
2989 2990 2991 2992
			break;

		case SRP_OPT_PKEY:
			if (match_hex(args, &token)) {
2993
				pr_warn("bad P_Key parameter '%s'\n", p);
2994 2995
				goto out;
			}
2996
			target->pkey = cpu_to_be16(token);
2997 2998 2999 3000
			break;

		case SRP_OPT_SERVICE_ID:
			p = match_strdup(args);
3001 3002 3003 3004
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
3005 3006 3007 3008 3009 3010
			target->service_id = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_MAX_SECT:
			if (match_int(args, &token)) {
3011
				pr_warn("bad max sect parameter '%s'\n", p);
3012 3013 3014 3015 3016
				goto out;
			}
			target->scsi_host->max_sectors = token;
			break;

3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028
		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;

3029
		case SRP_OPT_MAX_CMD_PER_LUN:
3030
			if (match_int(args, &token) || token < 1) {
3031 3032
				pr_warn("bad max cmd_per_lun parameter '%s'\n",
					p);
3033 3034
				goto out;
			}
3035
			target->scsi_host->cmd_per_lun = token;
3036 3037
			break;

3038 3039
		case SRP_OPT_IO_CLASS:
			if (match_hex(args, &token)) {
3040
				pr_warn("bad IO class parameter '%s'\n", p);
3041 3042 3043 3044
				goto out;
			}
			if (token != SRP_REV10_IB_IO_CLASS &&
			    token != SRP_REV16A_IB_IO_CLASS) {
3045 3046 3047
				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);
3048 3049 3050 3051 3052
				goto out;
			}
			target->io_class = token;
			break;

3053 3054
		case SRP_OPT_INITIATOR_EXT:
			p = match_strdup(args);
3055 3056 3057 3058
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
3059 3060 3061 3062
			target->initiator_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

3063 3064
		case SRP_OPT_CMD_SG_ENTRIES:
			if (match_int(args, &token) || token < 1 || token > 255) {
3065 3066
				pr_warn("bad max cmd_sg_entries parameter '%s'\n",
					p);
3067 3068 3069 3070 3071
				goto out;
			}
			target->cmd_sg_cnt = token;
			break;

3072 3073
		case SRP_OPT_ALLOW_EXT_SG:
			if (match_int(args, &token)) {
3074
				pr_warn("bad allow_ext_sg parameter '%s'\n", p);
3075 3076 3077 3078 3079 3080 3081 3082
				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) {
3083 3084
				pr_warn("bad max sg_tablesize parameter '%s'\n",
					p);
3085 3086 3087 3088 3089
				goto out;
			}
			target->sg_tablesize = token;
			break;

3090 3091 3092 3093 3094 3095 3096 3097
		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;

3098 3099 3100 3101 3102 3103 3104 3105 3106
		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;

3107
		default:
3108 3109
			pr_warn("unknown parameter or missing value '%s' in target creation request\n",
				p);
3110 3111 3112 3113 3114 3115 3116 3117 3118 3119
			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))
3120 3121
				pr_warn("target creation request is missing parameter '%s'\n",
					srp_opt_tokens[i].pattern);
3122

3123 3124 3125 3126 3127 3128
	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);

3129 3130 3131 3132 3133
out:
	kfree(options);
	return ret;
}

3134 3135
static ssize_t srp_create_target(struct device *dev,
				 struct device_attribute *attr,
3136 3137 3138
				 const char *buf, size_t count)
{
	struct srp_host *host =
3139
		container_of(dev, struct srp_host, dev);
3140 3141
	struct Scsi_Host *target_host;
	struct srp_target_port *target;
3142
	struct srp_rdma_ch *ch;
3143 3144
	struct srp_device *srp_dev = host->srp_dev;
	struct ib_device *ibdev = srp_dev->dev;
B
Bart Van Assche 已提交
3145 3146
	int ret, node_idx, node, cpu, i;
	bool multich = false;
3147 3148 3149 3150 3151 3152

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

3153
	target_host->transportt  = ib_srp_transport_template;
3154 3155
	target_host->max_channel = 0;
	target_host->max_id      = 1;
B
Bart Van Assche 已提交
3156
	target_host->max_lun     = -1LL;
A
Arne Redlich 已提交
3157
	target_host->max_cmd_len = sizeof ((struct srp_cmd *) (void *) 0L)->cdb;
R
Roland Dreier 已提交
3158

3159 3160
	target = host_to_target(target_host);

3161 3162 3163
	target->io_class	= SRP_REV16A_IB_IO_CLASS;
	target->scsi_host	= target_host;
	target->srp_host	= host;
J
Jason Gunthorpe 已提交
3164
	target->lkey		= host->srp_dev->pd->local_dma_lkey;
3165
	target->global_mr	= host->srp_dev->global_mr;
3166
	target->cmd_sg_cnt	= cmd_sg_entries;
3167 3168
	target->sg_tablesize	= indirect_sg_entries ? : cmd_sg_entries;
	target->allow_ext_sg	= allow_ext_sg;
3169
	target->tl_retry_count	= 7;
3170
	target->queue_size	= SRP_DEFAULT_QUEUE_SIZE;
3171

3172 3173 3174 3175 3176 3177
	/*
	 * Avoid that the SCSI host can be removed by srp_remove_target()
	 * before this function returns.
	 */
	scsi_host_get(target->scsi_host);

3178 3179
	mutex_lock(&host->add_target_mutex);

3180 3181
	ret = srp_parse_options(buf, target);
	if (ret)
3182
		goto out;
3183

B
Bart Van Assche 已提交
3184 3185
	ret = scsi_init_shared_tag_map(target_host, target_host->can_queue);
	if (ret)
3186
		goto out;
B
Bart Van Assche 已提交
3187

3188 3189
	target->req_ring_size = target->queue_size - SRP_TSK_MGMT_SQ_SIZE;

3190 3191 3192 3193 3194 3195 3196
	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;
3197
		goto out;
3198 3199
	}

3200
	if (!srp_dev->has_fmr && !srp_dev->has_fr && !target->allow_ext_sg &&
3201
	    target->cmd_sg_cnt < target->sg_tablesize) {
3202
		pr_warn("No MR pool and no external indirect descriptors, limiting sg_tablesize to cmd_sg_cnt\n");
3203 3204 3205 3206 3207 3208
		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);
3209 3210 3211 3212
	target->max_iu_len = sizeof (struct srp_cmd) +
			     sizeof (struct srp_indirect_buf) +
			     target->cmd_sg_cnt * sizeof (struct srp_direct_buf);

3213
	INIT_WORK(&target->tl_err_work, srp_tl_err_work);
3214
	INIT_WORK(&target->remove_work, srp_remove_work);
3215
	spin_lock_init(&target->lock);
3216
	ret = ib_query_gid(ibdev, host->port, 0, &target->sgid);
3217
	if (ret)
3218
		goto out;
3219

B
Bart Van Assche 已提交
3220 3221 3222 3223 3224 3225 3226 3227 3228
	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)
3229
		goto out;
3230

B
Bart Van Assche 已提交
3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258
	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;
3259

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

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

			ret = srp_connect_ch(ch, multich);
			if (ret) {
				shost_printk(KERN_ERR, target->scsi_host,
					     PFX "Connection %d/%d failed\n",
					     ch_start + cpu_idx,
					     target->ch_count);
				if (node_idx == 0 && cpu_idx == 0) {
					goto err_disconnect;
				} else {
					srp_free_ch_ib(target, ch);
					srp_free_req_data(target, ch);
					target->ch_count = ch - target->ch;
3280
					goto connected;
B
Bart Van Assche 已提交
3281 3282 3283 3284 3285 3286 3287
				}
			}

			multich = true;
			cpu_idx++;
		}
		node_idx++;
3288 3289
	}

3290
connected:
B
Bart Van Assche 已提交
3291 3292
	target->scsi_host->nr_hw_queues = target->ch_count;

3293 3294 3295 3296
	ret = srp_add_target(host, target);
	if (ret)
		goto err_disconnect;

3297 3298 3299 3300 3301
	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),
3302
			     be16_to_cpu(target->pkey),
3303
			     be64_to_cpu(target->service_id),
3304
			     target->sgid.raw, target->orig_dgid.raw);
3305
	}
B
Bart Van Assche 已提交
3306

3307 3308 3309 3310
	ret = count;

out:
	mutex_unlock(&host->add_target_mutex);
3311 3312

	scsi_host_put(target->scsi_host);
3313 3314
	if (ret < 0)
		scsi_host_put(target->scsi_host);
3315

3316
	return ret;
3317 3318 3319 3320

err_disconnect:
	srp_disconnect_target(target);

B
Bart Van Assche 已提交
3321 3322 3323 3324 3325
	for (i = 0; i < target->ch_count; i++) {
		ch = &target->ch[i];
		srp_free_ch_ib(target, ch);
		srp_free_req_data(target, ch);
	}
3326

B
Bart Van Assche 已提交
3327
	kfree(target->ch);
3328
	goto out;
3329 3330
}

3331
static DEVICE_ATTR(add_target, S_IWUSR, NULL, srp_create_target);
3332

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

3338
	return sprintf(buf, "%s\n", host->srp_dev->dev->name);
3339 3340
}

3341
static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
3342

3343 3344
static ssize_t show_port(struct device *dev, struct device_attribute *attr,
			 char *buf)
3345
{
3346
	struct srp_host *host = container_of(dev, struct srp_host, dev);
3347 3348 3349 3350

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

3351
static DEVICE_ATTR(port, S_IRUGO, show_port, NULL);
3352

3353
static struct srp_host *srp_add_port(struct srp_device *device, u8 port)
3354 3355 3356 3357 3358 3359 3360 3361
{
	struct srp_host *host;

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

	INIT_LIST_HEAD(&host->target_list);
3362
	spin_lock_init(&host->target_lock);
3363
	init_completion(&host->released);
3364
	mutex_init(&host->add_target_mutex);
3365
	host->srp_dev = device;
3366 3367
	host->port = port;

3368 3369
	host->dev.class = &srp_class;
	host->dev.parent = device->dev->dma_device;
3370
	dev_set_name(&host->dev, "srp-%s-%d", device->dev->name, port);
3371

3372
	if (device_register(&host->dev))
3373
		goto free_host;
3374
	if (device_create_file(&host->dev, &dev_attr_add_target))
3375
		goto err_class;
3376
	if (device_create_file(&host->dev, &dev_attr_ibdev))
3377
		goto err_class;
3378
	if (device_create_file(&host->dev, &dev_attr_port))
3379 3380 3381 3382 3383
		goto err_class;

	return host;

err_class:
3384
	device_unregister(&host->dev);
3385

3386
free_host:
3387 3388 3389 3390 3391 3392 3393
	kfree(host);

	return NULL;
}

static void srp_add_one(struct ib_device *device)
{
3394 3395
	struct srp_device *srp_dev;
	struct ib_device_attr *dev_attr;
3396
	struct srp_host *host;
3397
	int mr_page_shift, p;
3398
	u64 max_pages_per_mr;
3399

3400 3401
	dev_attr = kmalloc(sizeof *dev_attr, GFP_KERNEL);
	if (!dev_attr)
3402
		return;
3403

3404
	if (ib_query_device(device, dev_attr)) {
3405
		pr_warn("Query device failed for %s\n", device->name);
3406 3407 3408 3409 3410 3411 3412
		goto free_attr;
	}

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

3413 3414
	srp_dev->has_fmr = (device->alloc_fmr && device->dealloc_fmr &&
			    device->map_phys_fmr && device->unmap_fmr);
3415 3416 3417 3418 3419 3420 3421
	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));
3422
	srp_dev->use_fmr = !srp_dev->use_fast_reg && srp_dev->has_fmr;
3423

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

	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;

3455 3456 3457 3458 3459 3460 3461 3462 3463 3464
	if (!register_always || (!srp_dev->has_fmr && !srp_dev->has_fr)) {
		srp_dev->global_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->global_mr))
			goto err_pd;
	} else {
		srp_dev->global_mr = NULL;
	}
3465

3466
	for (p = rdma_start_port(device); p <= rdma_end_port(device); ++p) {
3467
		host = srp_add_port(srp_dev, p);
3468
		if (host)
3469
			list_add_tail(&host->list, &srp_dev->dev_list);
3470 3471
	}

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

3486
static void srp_remove_one(struct ib_device *device, void *client_data)
3487
{
3488
	struct srp_device *srp_dev;
3489
	struct srp_host *host, *tmp_host;
3490
	struct srp_target_port *target;
3491

3492
	srp_dev = client_data;
3493 3494
	if (!srp_dev)
		return;
3495

3496
	list_for_each_entry_safe(host, tmp_host, &srp_dev->dev_list, list) {
3497
		device_unregister(&host->dev);
3498 3499 3500 3501 3502 3503 3504
		/*
		 * Wait for the sysfs entry to go away, so that no new
		 * target ports can be created.
		 */
		wait_for_completion(&host->released);

		/*
3505
		 * Remove all target ports.
3506
		 */
3507
		spin_lock(&host->target_lock);
3508 3509
		list_for_each_entry(target, &host->target_list, list)
			srp_queue_remove_work(target);
3510
		spin_unlock(&host->target_lock);
3511 3512

		/*
3513
		 * Wait for tl_err and target port removal tasks.
3514
		 */
3515
		flush_workqueue(system_long_wq);
3516
		flush_workqueue(srp_remove_wq);
3517 3518 3519 3520

		kfree(host);
	}

3521 3522
	if (srp_dev->global_mr)
		ib_dereg_mr(srp_dev->global_mr);
3523 3524 3525
	ib_dealloc_pd(srp_dev->pd);

	kfree(srp_dev);
3526 3527
}

3528
static struct srp_function_template ib_srp_transport_functions = {
3529 3530
	.has_rport_state	 = true,
	.reset_timer_if_blocked	 = true,
3531
	.reconnect_delay	 = &srp_reconnect_delay,
3532 3533 3534
	.fast_io_fail_tmo	 = &srp_fast_io_fail_tmo,
	.dev_loss_tmo		 = &srp_dev_loss_tmo,
	.reconnect		 = srp_rport_reconnect,
3535
	.rport_delete		 = srp_rport_delete,
3536
	.terminate_rport_io	 = srp_terminate_io,
3537 3538
};

3539 3540 3541 3542
static int __init srp_init_module(void)
{
	int ret;

3543
	BUILD_BUG_ON(FIELD_SIZEOF(struct ib_wc, wr_id) < sizeof(void *));
3544

3545
	if (srp_sg_tablesize) {
3546
		pr_warn("srp_sg_tablesize is deprecated, please use cmd_sg_entries\n");
3547 3548 3549 3550 3551 3552 3553 3554
		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) {
3555
		pr_warn("Clamping cmd_sg_entries to 255\n");
3556
		cmd_sg_entries = 255;
3557 3558
	}

3559 3560 3561
	if (!indirect_sg_entries)
		indirect_sg_entries = cmd_sg_entries;
	else if (indirect_sg_entries < cmd_sg_entries) {
3562 3563
		pr_warn("Bumping up indirect_sg_entries to match cmd_sg_entries (%u)\n",
			cmd_sg_entries);
3564 3565 3566
		indirect_sg_entries = cmd_sg_entries;
	}

3567
	srp_remove_wq = create_workqueue("srp_remove");
3568 3569
	if (!srp_remove_wq) {
		ret = -ENOMEM;
3570 3571 3572 3573
		goto out;
	}

	ret = -ENOMEM;
3574 3575 3576
	ib_srp_transport_template =
		srp_attach_transport(&ib_srp_transport_functions);
	if (!ib_srp_transport_template)
3577
		goto destroy_wq;
3578

3579 3580
	ret = class_register(&srp_class);
	if (ret) {
3581
		pr_err("couldn't register class infiniband_srp\n");
3582
		goto release_tr;
3583 3584
	}

3585 3586
	ib_sa_register_client(&srp_sa_client);

3587 3588
	ret = ib_register_client(&srp_client);
	if (ret) {
3589
		pr_err("couldn't register IB client\n");
3590
		goto unreg_sa;
3591 3592
	}

3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605
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;
3606 3607 3608 3609 3610
}

static void __exit srp_cleanup_module(void)
{
	ib_unregister_client(&srp_client);
3611
	ib_sa_unregister_client(&srp_sa_client);
3612
	class_unregister(&srp_class);
3613
	srp_release_transport(ib_srp_transport_template);
3614
	destroy_workqueue(srp_remove_wq);
3615 3616 3617 3618
}

module_init(srp_init_module);
module_exit(srp_cleanup_module);