ib_srp.c 89.2 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
/*
 * Copyright (c) 2005 Cisco Systems.  All rights reserved.
 *
 * This software is available to you under a choice of one of two
 * licenses.  You may choose to be licensed under the terms of the GNU
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - Redistributions in binary form must reproduce the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer in the documentation and/or other materials
 *        provided with the distribution.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 */

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

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

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

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

#include "ib_srp.h"

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

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

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

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

76 77 78
module_param(cmd_sg_entries, uint, 0444);
MODULE_PARM_DESC(cmd_sg_entries,
		 "Default number of gather/scatter entries in the SRP command (default is 12, max 255)");
79

80 81 82 83 84 85 86 87
module_param(indirect_sg_entries, uint, 0444);
MODULE_PARM_DESC(indirect_sg_entries,
		 "Default max number of gather/scatter entries (default is 12, max is " __stringify(SCSI_MAX_SG_CHAIN_SEGMENTS) ")");

module_param(allow_ext_sg, bool, 0444);
MODULE_PARM_DESC(allow_ext_sg,
		  "Default behavior when there are more than cmd_sg_entries S/G entries after mapping; fails the request when false (default false)");

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

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

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

100 101
static struct kernel_param_ops srp_tmo_ops;

102 103 104 105 106
static int srp_reconnect_delay = 10;
module_param_cb(reconnect_delay, &srp_tmo_ops, &srp_reconnect_delay,
		S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(reconnect_delay, "Time between successive reconnect attempts");

107 108 109 110 111 112 113 114
static int srp_fast_io_fail_tmo = 15;
module_param_cb(fast_io_fail_tmo, &srp_tmo_ops, &srp_fast_io_fail_tmo,
		S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(fast_io_fail_tmo,
		 "Number of seconds between the observation of a transport"
		 " layer error and failing all I/O. \"off\" means that this"
		 " functionality is disabled.");

115
static int srp_dev_loss_tmo = 600;
116 117 118 119 120 121 122 123 124 125
module_param_cb(dev_loss_tmo, &srp_tmo_ops, &srp_dev_loss_tmo,
		S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(dev_loss_tmo,
		 "Maximum number of seconds that the SRP transport should"
		 " insulate transport layer errors. After this time has been"
		 " exceeded the SCSI host is removed. Should be"
		 " between 1 and " __stringify(SCSI_DEVICE_BLOCK_MAX_TIMEOUT)
		 " if fast_io_fail_tmo has not been set. \"off\" means that"
		 " this functionality is disabled.");

126 127
static void srp_add_one(struct ib_device *device);
static void srp_remove_one(struct ib_device *device);
128 129
static void srp_recv_completion(struct ib_cq *cq, void *ch_ptr);
static void srp_send_completion(struct ib_cq *cq, void *ch_ptr);
130 131
static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event);

132
static struct scsi_transport_template *ib_srp_transport_template;
133
static struct workqueue_struct *srp_remove_wq;
134

135 136 137 138 139 140
static struct ib_client srp_client = {
	.name   = "srp",
	.add    = srp_add_one,
	.remove = srp_remove_one
};

141 142
static struct ib_sa_client srp_sa_client;

143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163
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;

	if (strncmp(val, "off", 3) != 0) {
		res = kstrtoint(val, 0, &tmo);
		if (res)
			goto out;
	} else {
		tmo = -1;
	}
164 165 166 167 168
	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);
169
	else
170 171
		res = srp_tmo_valid(srp_reconnect_delay, srp_fast_io_fail_tmo,
				    tmo);
172 173 174 175 176 177 178 179 180 181 182 183 184
	if (res)
		goto out;
	*(int *)kp->arg = tmo;

out:
	return res;
}

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

185 186 187 188 189 190 191 192 193 194
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;
}

195 196 197
static int srp_target_is_topspin(struct srp_target_port *target)
{
	static const u8 topspin_oui[3] = { 0x00, 0x05, 0xad };
198
	static const u8 cisco_oui[3]   = { 0x00, 0x1b, 0x0d };
199 200

	return topspin_workarounds &&
201 202
		(!memcmp(&target->ioc_guid, topspin_oui, sizeof topspin_oui) ||
		 !memcmp(&target->ioc_guid, cisco_oui, sizeof cisco_oui));
203 204
}

205 206 207 208 209 210 211 212 213 214 215 216 217 218
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;

219 220 221
	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))
222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241
		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;

242 243
	ib_dma_unmap_single(host->srp_dev->dev, iu->dma, iu->size,
			    iu->direction);
244 245 246 247 248 249
	kfree(iu->buf);
	kfree(iu);
}

static void srp_qp_event(struct ib_event *event, void *context)
{
250
	pr_debug("QP event %d\n", event->event);
251 252 253 254 255 256 257 258 259 260 261 262
}

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;

263 264
	ret = ib_find_pkey(target->srp_host->srp_dev->dev,
			   target->srp_host->port,
265
			   be16_to_cpu(target->pkey),
266
			   &attr->pkey_index);
267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285
	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;
}

286
static int srp_new_cm_id(struct srp_rdma_ch *ch)
D
David Dillow 已提交
287
{
288
	struct srp_target_port *target = ch->target;
D
David Dillow 已提交
289 290
	struct ib_cm_id *new_cm_id;

291
	new_cm_id = ib_create_cm_id(target->srp_host->srp_dev->dev,
292
				    srp_cm_handler, ch);
D
David Dillow 已提交
293 294 295
	if (IS_ERR(new_cm_id))
		return PTR_ERR(new_cm_id);

296 297 298 299 300 301 302
	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
303 304 305 306

	return 0;
}

307 308 309 310 311 312 313 314 315
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;
316 317
	fmr_param.max_pages_per_fmr = dev->max_pages_per_mr;
	fmr_param.page_shift	    = ilog2(dev->mr_page_size);
318 319 320 321 322 323 324
	fmr_param.access	    = (IB_ACCESS_LOCAL_WRITE |
				       IB_ACCESS_REMOTE_WRITE |
				       IB_ACCESS_REMOTE_READ);

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

325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450
/**
 * srp_destroy_fr_pool() - free the resources owned by a pool
 * @pool: Fast registration pool to be destroyed.
 */
static void srp_destroy_fr_pool(struct srp_fr_pool *pool)
{
	int i;
	struct srp_fr_desc *d;

	if (!pool)
		return;

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

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

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

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

out:
	return pool;

destroy_pool:
	srp_destroy_fr_pool(pool);

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

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

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

	return d;
}

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

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

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

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

451
static int srp_create_ch_ib(struct srp_rdma_ch *ch)
452
{
453
	struct srp_target_port *target = ch->target;
454
	struct srp_device *dev = target->srp_host->srp_dev;
455
	struct ib_qp_init_attr *init_attr;
456 457
	struct ib_cq *recv_cq, *send_cq;
	struct ib_qp *qp;
458
	struct ib_fmr_pool *fmr_pool = NULL;
459 460
	struct srp_fr_pool *fr_pool = NULL;
	const int m = 1 + dev->use_fast_reg;
461 462 463 464 465 466
	int ret;

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

467 468
	recv_cq = ib_create_cq(dev->dev, srp_recv_completion, NULL, ch,
			       target->queue_size, ch->comp_vector);
469 470
	if (IS_ERR(recv_cq)) {
		ret = PTR_ERR(recv_cq);
471
		goto err;
472 473
	}

474 475
	send_cq = ib_create_cq(dev->dev, srp_send_completion, NULL, ch,
			       m * target->queue_size, ch->comp_vector);
476 477
	if (IS_ERR(send_cq)) {
		ret = PTR_ERR(send_cq);
478
		goto err_recv_cq;
479 480
	}

481
	ib_req_notify_cq(recv_cq, IB_CQ_NEXT_COMP);
482 483

	init_attr->event_handler       = srp_qp_event;
484
	init_attr->cap.max_send_wr     = m * target->queue_size;
485
	init_attr->cap.max_recv_wr     = target->queue_size;
486 487
	init_attr->cap.max_recv_sge    = 1;
	init_attr->cap.max_send_sge    = 1;
488
	init_attr->sq_sig_type         = IB_SIGNAL_REQ_WR;
489
	init_attr->qp_type             = IB_QPT_RC;
490 491
	init_attr->send_cq             = send_cq;
	init_attr->recv_cq             = recv_cq;
492

493
	qp = ib_create_qp(dev->pd, init_attr);
494 495
	if (IS_ERR(qp)) {
		ret = PTR_ERR(qp);
496
		goto err_send_cq;
497 498
	}

499
	ret = srp_init_qp(target, qp);
500 501
	if (ret)
		goto err_qp;
502

503 504 505 506 507 508 509 510
	if (dev->use_fast_reg && dev->has_fr) {
		fr_pool = srp_alloc_fr_pool(target);
		if (IS_ERR(fr_pool)) {
			ret = PTR_ERR(fr_pool);
			shost_printk(KERN_WARNING, target->scsi_host, PFX
				     "FR pool allocation failed (%d)\n", ret);
			goto err_qp;
		}
511 512 513
		if (ch->fr_pool)
			srp_destroy_fr_pool(ch->fr_pool);
		ch->fr_pool = fr_pool;
514
	} else if (!dev->use_fast_reg && dev->has_fmr) {
515 516 517 518 519 520 521
		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;
		}
522 523 524
		if (ch->fmr_pool)
			ib_destroy_fmr_pool(ch->fmr_pool);
		ch->fmr_pool = fmr_pool;
525 526
	}

527 528 529 530 531 532
	if (ch->qp)
		ib_destroy_qp(ch->qp);
	if (ch->recv_cq)
		ib_destroy_cq(ch->recv_cq);
	if (ch->send_cq)
		ib_destroy_cq(ch->send_cq);
533

534 535 536
	ch->qp = qp;
	ch->recv_cq = recv_cq;
	ch->send_cq = send_cq;
537

538 539 540 541
	kfree(init_attr);
	return 0;

err_qp:
542
	ib_destroy_qp(qp);
543 544

err_send_cq:
545
	ib_destroy_cq(send_cq);
546 547

err_recv_cq:
548
	ib_destroy_cq(recv_cq);
549 550

err:
551 552 553 554
	kfree(init_attr);
	return ret;
}

555 556
/*
 * Note: this function may be called without srp_alloc_iu_bufs() having been
557
 * invoked. Hence the ch->[rt]x_ring checks.
558
 */
559 560
static void srp_free_ch_ib(struct srp_target_port *target,
			   struct srp_rdma_ch *ch)
561
{
562
	struct srp_device *dev = target->srp_host->srp_dev;
563 564
	int i;

565 566 567
	if (ch->cm_id) {
		ib_destroy_cm_id(ch->cm_id);
		ch->cm_id = NULL;
568 569
	}

570
	if (dev->use_fast_reg) {
571 572
		if (ch->fr_pool)
			srp_destroy_fr_pool(ch->fr_pool);
573
	} else {
574 575
		if (ch->fmr_pool)
			ib_destroy_fmr_pool(ch->fmr_pool);
576
	}
577 578 579
	ib_destroy_qp(ch->qp);
	ib_destroy_cq(ch->send_cq);
	ib_destroy_cq(ch->recv_cq);
580

581 582
	ch->qp = NULL;
	ch->send_cq = ch->recv_cq = NULL;
583

584
	if (ch->rx_ring) {
585
		for (i = 0; i < target->queue_size; ++i)
586 587 588
			srp_free_iu(target->srp_host, ch->rx_ring[i]);
		kfree(ch->rx_ring);
		ch->rx_ring = NULL;
589
	}
590
	if (ch->tx_ring) {
591
		for (i = 0; i < target->queue_size; ++i)
592 593 594
			srp_free_iu(target->srp_host, ch->tx_ring[i]);
		kfree(ch->tx_ring);
		ch->tx_ring = NULL;
595
	}
596 597 598 599
}

static void srp_path_rec_completion(int status,
				    struct ib_sa_path_rec *pathrec,
600
				    void *ch_ptr)
601
{
602 603
	struct srp_rdma_ch *ch = ch_ptr;
	struct srp_target_port *target = ch->target;
604

605
	ch->status = status;
606
	if (status)
607 608
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Got failed path rec status %d\n", status);
609
	else
610 611
		ch->path = *pathrec;
	complete(&ch->done);
612 613
}

614
static int srp_lookup_path(struct srp_rdma_ch *ch)
615
{
616
	struct srp_target_port *target = ch->target;
617 618
	int ret;

619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639
	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);
640 641
	if (ret < 0)
		return ret;
642

643
	if (ch->status < 0)
644 645
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Path record query failed\n");
646

647
	return ch->status;
648 649
}

650
static int srp_send_req(struct srp_rdma_ch *ch)
651
{
652
	struct srp_target_port *target = ch->target;
653 654 655 656 657 658 659 660 661 662
	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;

663
	req->param.primary_path		      = &ch->path;
664 665
	req->param.alternate_path 	      = NULL;
	req->param.service_id 		      = target->service_id;
666 667
	req->param.qp_num		      = ch->qp->qp_num;
	req->param.qp_type		      = ch->qp->qp_type;
668 669 670 671 672 673 674 675 676 677 678 679 680 681
	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;
682
	req->param.retry_count                = target->tl_retry_count;
683 684 685 686 687
	req->param.rnr_retry_count 	      = 7;
	req->param.max_cm_retries 	      = 15;

	req->priv.opcode     	= SRP_LOGIN_REQ;
	req->priv.tag        	= 0;
688
	req->priv.req_it_iu_len = cpu_to_be32(target->max_iu_len);
689 690
	req->priv.req_buf_fmt 	= cpu_to_be16(SRP_BUF_FORMAT_DIRECT |
					      SRP_BUF_FORMAT_INDIRECT);
691
	/*
R
Roland Dreier 已提交
692
	 * In the published SRP specification (draft rev. 16a), the
693 694 695 696 697 698 699 700 701
	 * 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,
702
		       &target->sgid.global.interface_id, 8);
703
		memcpy(req->priv.initiator_port_id + 8,
704
		       &target->initiator_ext, 8);
705 706 707 708
		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,
709 710
		       &target->initiator_ext, 8);
		memcpy(req->priv.initiator_port_id + 8,
711
		       &target->sgid.global.interface_id, 8);
712 713 714 715
		memcpy(req->priv.target_port_id,     &target->id_ext, 8);
		memcpy(req->priv.target_port_id + 8, &target->ioc_guid, 8);
	}

716 717
	/*
	 * Topspin/Cisco SRP targets will reject our login unless we
718 719
	 * zero out the first 8 bytes of our initiator port ID and set
	 * the second 8 bytes to the local node GUID.
720
	 */
721
	if (srp_target_is_topspin(target)) {
722 723 724 725
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Topspin/Cisco initiator port ID workaround "
			     "activated for target GUID %016llx\n",
			     (unsigned long long) be64_to_cpu(target->ioc_guid));
726
		memset(req->priv.initiator_port_id, 0, 8);
727
		memcpy(req->priv.initiator_port_id + 8,
728
		       &target->srp_host->srp_dev->dev->node_guid, 8);
729 730
	}

731
	status = ib_send_cm_req(ch->cm_id, &req->param);
732 733 734 735 736 737

	kfree(req);

	return status;
}

738 739 740 741 742 743 744 745 746 747 748 749
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)
750
		queue_work(srp_remove_wq, &target->remove_work);
751 752 753 754

	return changed;
}

755 756 757 758 759 760 761 762 763 764 765 766 767 768 769
static bool srp_change_conn_state(struct srp_target_port *target,
				  bool connected)
{
	bool changed = false;

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

	return changed;
}

770 771
static void srp_disconnect_target(struct srp_target_port *target)
{
772 773
	struct srp_rdma_ch *ch = &target->ch;

774 775
	if (srp_change_conn_state(target, false)) {
		/* XXX should send SRP_I_LOGOUT request */
776

777
		if (ib_send_cm_dreq(ch->cm_id, NULL, 0)) {
778 779 780
			shost_printk(KERN_DEBUG, target->scsi_host,
				     PFX "Sending CM DREQ failed\n");
		}
781
	}
782 783
}

784 785
static void srp_free_req_data(struct srp_target_port *target,
			      struct srp_rdma_ch *ch)
786
{
787 788
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = dev->dev;
789 790 791
	struct srp_request *req;
	int i;

792
	if (!ch->req_ring)
793 794 795
		return;

	for (i = 0; i < target->req_ring_size; ++i) {
796
		req = &ch->req_ring[i];
797 798 799 800
		if (dev->use_fast_reg)
			kfree(req->fr_list);
		else
			kfree(req->fmr_list);
801
		kfree(req->map_page);
802 803 804 805 806 807
		if (req->indirect_dma_addr) {
			ib_dma_unmap_single(ibdev, req->indirect_dma_addr,
					    target->indirect_size,
					    DMA_TO_DEVICE);
		}
		kfree(req->indirect_desc);
808
	}
809

810 811
	kfree(ch->req_ring);
	ch->req_ring = NULL;
812 813
}

814
static int srp_alloc_req_data(struct srp_rdma_ch *ch)
815
{
816
	struct srp_target_port *target = ch->target;
817 818 819
	struct srp_device *srp_dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = srp_dev->dev;
	struct srp_request *req;
820
	void *mr_list;
821 822 823
	dma_addr_t dma_addr;
	int i, ret = -ENOMEM;

824 825 826
	ch->req_ring = kcalloc(target->req_ring_size, sizeof(*ch->req_ring),
			       GFP_KERNEL);
	if (!ch->req_ring)
827 828 829
		goto out;

	for (i = 0; i < target->req_ring_size; ++i) {
830
		req = &ch->req_ring[i];
831 832 833 834 835 836 837 838
		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;
839
		req->map_page = kmalloc(srp_dev->max_pages_per_mr *
840
					sizeof(void *), GFP_KERNEL);
841 842
		if (!req->map_page)
			goto out;
843
		req->indirect_desc = kmalloc(target->indirect_size, GFP_KERNEL);
844
		if (!req->indirect_desc)
845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860
			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;
}

861 862 863 864 865 866 867 868 869 870 871 872 873 874 875
/**
 * 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);
}

876 877
static void srp_remove_target(struct srp_target_port *target)
{
878 879
	struct srp_rdma_ch *ch = &target->ch;

880 881
	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);

882
	srp_del_scsi_host_attr(target->scsi_host);
883
	srp_rport_get(target->rport);
884 885
	srp_remove_host(target->scsi_host);
	scsi_remove_host(target->scsi_host);
886
	srp_stop_rport_timers(target->rport);
887
	srp_disconnect_target(target);
888
	srp_free_ch_ib(target, ch);
889
	cancel_work_sync(&target->tl_err_work);
890
	srp_rport_put(target->rport);
891
	srp_free_req_data(target, ch);
892 893 894 895 896

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

897 898 899
	scsi_host_put(target->scsi_host);
}

D
David Howells 已提交
900
static void srp_remove_work(struct work_struct *work)
901
{
D
David Howells 已提交
902
	struct srp_target_port *target =
903
		container_of(work, struct srp_target_port, remove_work);
904

905
	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);
906

907
	srp_remove_target(target);
908 909
}

910 911 912 913 914 915 916
static void srp_rport_delete(struct srp_rport *rport)
{
	struct srp_target_port *target = rport->lld_data;

	srp_queue_remove_work(target);
}

917
static int srp_connect_ch(struct srp_rdma_ch *ch)
918
{
919
	struct srp_target_port *target = ch->target;
920 921
	int ret;

922 923
	WARN_ON_ONCE(target->connected);

924 925
	target->qp_in_error = false;

926
	ret = srp_lookup_path(ch);
927 928 929 930
	if (ret)
		return ret;

	while (1) {
931 932
		init_completion(&ch->done);
		ret = srp_send_req(ch);
933 934
		if (ret)
			return ret;
935
		ret = wait_for_completion_interruptible(&ch->done);
936 937
		if (ret < 0)
			return ret;
938 939 940 941 942 943 944

		/*
		 * 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.
		 */
945
		switch (ch->status) {
946
		case 0:
947
			srp_change_conn_state(target, true);
948 949 950
			return 0;

		case SRP_PORT_REDIRECT:
951
			ret = srp_lookup_path(ch);
952 953 954 955 956 957 958
			if (ret)
				return ret;
			break;

		case SRP_DLID_REDIRECT:
			break;

D
David Dillow 已提交
959 960
		case SRP_STALE_CONN:
			shost_printk(KERN_ERR, target->scsi_host, PFX
961
				     "giving up on stale connection\n");
962 963
			ch->status = -ECONNRESET;
			return ch->status;
D
David Dillow 已提交
964

965
		default:
966
			return ch->status;
967 968 969 970
		}
	}
}

971
static int srp_inv_rkey(struct srp_rdma_ch *ch, u32 rkey)
972 973 974 975 976 977 978 979 980 981 982
{
	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,
	};

983
	return ib_post_send(ch->qp, &wr, &bad_wr);
984 985
}

986
static void srp_unmap_data(struct scsi_cmnd *scmnd,
987
			   struct srp_rdma_ch *ch,
988 989
			   struct srp_request *req)
{
990
	struct srp_target_port *target = ch->target;
991 992 993
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = dev->dev;
	int i, res;
994

995
	if (!scsi_sglist(scmnd) ||
996 997 998 999
	    (scmnd->sc_data_direction != DMA_TO_DEVICE &&
	     scmnd->sc_data_direction != DMA_FROM_DEVICE))
		return;

1000 1001 1002 1003
	if (dev->use_fast_reg) {
		struct srp_fr_desc **pfr;

		for (i = req->nmdesc, pfr = req->fr_list; i > 0; i--, pfr++) {
1004
			res = srp_inv_rkey(ch, (*pfr)->mr->rkey);
1005 1006 1007 1008 1009 1010 1011 1012 1013
			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)
1014
			srp_fr_pool_put(ch->fr_pool, req->fr_list,
1015 1016 1017 1018 1019 1020 1021
					req->nmdesc);
	} else {
		struct ib_pool_fmr **pfmr;

		for (i = req->nmdesc, pfmr = req->fmr_list; i > 0; i--, pfmr++)
			ib_fmr_pool_unmap(*pfmr);
	}
1022

1023 1024
	ib_dma_unmap_sg(ibdev, scsi_sglist(scmnd), scsi_sg_count(scmnd),
			scmnd->sc_data_direction);
1025 1026
}

B
Bart Van Assche 已提交
1027 1028
/**
 * srp_claim_req - Take ownership of the scmnd associated with a request.
1029
 * @ch: SRP RDMA channel.
B
Bart Van Assche 已提交
1030
 * @req: SRP request.
1031
 * @sdev: If not NULL, only take ownership for this SCSI device.
B
Bart Van Assche 已提交
1032 1033 1034 1035 1036 1037
 * @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.
 */
1038
static struct scsi_cmnd *srp_claim_req(struct srp_rdma_ch *ch,
B
Bart Van Assche 已提交
1039
				       struct srp_request *req,
1040
				       struct scsi_device *sdev,
B
Bart Van Assche 已提交
1041 1042 1043 1044
				       struct scsi_cmnd *scmnd)
{
	unsigned long flags;

1045
	spin_lock_irqsave(&ch->lock, flags);
1046 1047 1048
	if (req->scmnd &&
	    (!sdev || req->scmnd->device == sdev) &&
	    (!scmnd || req->scmnd == scmnd)) {
B
Bart Van Assche 已提交
1049 1050 1051 1052 1053
		scmnd = req->scmnd;
		req->scmnd = NULL;
	} else {
		scmnd = NULL;
	}
1054
	spin_unlock_irqrestore(&ch->lock, flags);
B
Bart Van Assche 已提交
1055 1056 1057 1058 1059 1060

	return scmnd;
}

/**
 * srp_free_req() - Unmap data and add request to the free request list.
1061
 * @ch:     SRP RDMA channel.
1062 1063 1064
 * @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
1065
 */
1066 1067
static void srp_free_req(struct srp_rdma_ch *ch, struct srp_request *req,
			 struct scsi_cmnd *scmnd, s32 req_lim_delta)
1068
{
1069 1070
	unsigned long flags;

1071
	srp_unmap_data(scmnd, ch, req);
B
Bart Van Assche 已提交
1072

1073 1074 1075
	spin_lock_irqsave(&ch->lock, flags);
	ch->req_lim += req_lim_delta;
	spin_unlock_irqrestore(&ch->lock, flags);
1076 1077
}

1078 1079
static void srp_finish_req(struct srp_rdma_ch *ch, struct srp_request *req,
			   struct scsi_device *sdev, int result)
1080
{
1081
	struct scsi_cmnd *scmnd = srp_claim_req(ch, req, sdev, NULL);
B
Bart Van Assche 已提交
1082 1083

	if (scmnd) {
1084
		srp_free_req(ch, req, scmnd, 0);
1085
		scmnd->result = result;
B
Bart Van Assche 已提交
1086 1087
		scmnd->scsi_done(scmnd);
	}
1088 1089
}

1090
static void srp_terminate_io(struct srp_rport *rport)
1091
{
1092
	struct srp_target_port *target = rport->lld_data;
1093
	struct srp_rdma_ch *ch = &target->ch;
1094 1095
	struct Scsi_Host *shost = target->scsi_host;
	struct scsi_device *sdev;
1096 1097
	int i;

1098 1099 1100 1101 1102 1103 1104
	/*
	 * 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);

1105
	for (i = 0; i < target->req_ring_size; ++i) {
1106 1107 1108
		struct srp_request *req = &ch->req_ring[i];

		srp_finish_req(ch, req, NULL, DID_TRANSPORT_FAILFAST << 16);
1109 1110
	}
}
1111

1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123
/*
 * 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;
1124
	struct srp_rdma_ch *ch = &target->ch;
1125
	int i, ret;
1126

1127
	srp_disconnect_target(target);
1128 1129 1130 1131

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

1132
	/*
1133 1134 1135
	 * 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.
1136
	 */
1137
	ret = srp_new_cm_id(ch);
1138

1139
	for (i = 0; i < target->req_ring_size; ++i) {
1140 1141 1142
		struct srp_request *req = &ch->req_ring[i];

		srp_finish_req(ch, req, NULL, DID_RESET << 16);
1143
	}
1144

1145 1146 1147 1148 1149
	/*
	 * Whether or not creating a new CM ID succeeded, create a new
	 * QP. This guarantees that all callback functions for the old QP have
	 * finished before any send requests are posted on the new QP.
	 */
1150
	ret += srp_create_ch_ib(ch);
1151

1152
	INIT_LIST_HEAD(&ch->free_tx);
1153
	for (i = 0; i < target->queue_size; ++i)
1154
		list_add(&ch->tx_ring[i]->list, &ch->free_tx);
1155

1156
	if (ret == 0)
1157
		ret = srp_connect_ch(ch);
1158

1159 1160 1161
	if (ret == 0)
		shost_printk(KERN_INFO, target->scsi_host,
			     PFX "reconnect succeeded\n");
1162 1163 1164 1165

	return ret;
}

1166 1167
static void srp_map_desc(struct srp_map_state *state, dma_addr_t dma_addr,
			 unsigned int dma_len, u32 rkey)
1168
{
1169
	struct srp_direct_buf *desc = state->desc;
1170

1171 1172 1173
	desc->va = cpu_to_be64(dma_addr);
	desc->key = cpu_to_be32(rkey);
	desc->len = cpu_to_be32(dma_len);
1174

1175 1176 1177 1178
	state->total_len += dma_len;
	state->desc++;
	state->ndesc++;
}
1179

1180
static int srp_map_finish_fmr(struct srp_map_state *state,
1181
			      struct srp_rdma_ch *ch)
1182 1183 1184
{
	struct ib_pool_fmr *fmr;
	u64 io_addr = 0;
1185

1186
	fmr = ib_fmr_pool_map_phys(ch->fmr_pool, state->pages,
1187 1188 1189
				   state->npages, io_addr);
	if (IS_ERR(fmr))
		return PTR_ERR(fmr);
1190

1191
	*state->next_fmr++ = fmr;
1192
	state->nmdesc++;
1193

1194
	srp_map_desc(state, 0, state->dma_len, fmr->fmr->rkey);
1195

1196 1197 1198
	return 0;
}

1199
static int srp_map_finish_fr(struct srp_map_state *state,
1200
			     struct srp_rdma_ch *ch)
1201
{
1202
	struct srp_target_port *target = ch->target;
1203 1204 1205 1206 1207 1208
	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;

1209
	desc = srp_fr_pool_get(ch->fr_pool);
1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237
	if (!desc)
		return -ENOMEM;

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

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

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

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

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

1238
	return ib_post_send(ch->qp, &wr, &bad_wr);
1239 1240
}

1241
static int srp_finish_mapping(struct srp_map_state *state,
1242
			      struct srp_rdma_ch *ch)
1243
{
1244
	struct srp_target_port *target = ch->target;
1245 1246 1247 1248 1249
	int ret = 0;

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

1250
	if (state->npages == 1 && !register_always)
1251
		srp_map_desc(state, state->base_dma_addr, state->dma_len,
1252 1253
			     target->rkey);
	else
1254
		ret = target->srp_host->srp_dev->use_fast_reg ?
1255 1256
			srp_map_finish_fr(state, ch) :
			srp_map_finish_fmr(state, ch);
1257 1258 1259

	if (ret == 0) {
		state->npages = 0;
1260
		state->dma_len = 0;
1261 1262 1263 1264 1265
	}

	return ret;
}

1266 1267 1268 1269 1270 1271 1272 1273
static void srp_map_update_start(struct srp_map_state *state,
				 struct scatterlist *sg, int sg_index,
				 dma_addr_t dma_addr)
{
	state->unmapped_sg = sg;
	state->unmapped_index = sg_index;
	state->unmapped_addr = dma_addr;
}
1274

1275
static int srp_map_sg_entry(struct srp_map_state *state,
1276
			    struct srp_rdma_ch *ch,
1277
			    struct scatterlist *sg, int sg_index,
1278
			    bool use_mr)
1279
{
1280
	struct srp_target_port *target = ch->target;
1281 1282 1283 1284 1285 1286 1287 1288 1289 1290
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = dev->dev;
	dma_addr_t dma_addr = ib_sg_dma_address(ibdev, sg);
	unsigned int dma_len = ib_sg_dma_len(ibdev, sg);
	unsigned int len;
	int ret;

	if (!dma_len)
		return 0;

1291 1292 1293 1294
	if (!use_mr) {
		/*
		 * Once we're in direct map mode for a request, we don't
		 * go back to FMR or FR mode, so no need to update anything
1295 1296 1297 1298
		 * other than the descriptor.
		 */
		srp_map_desc(state, dma_addr, dma_len, target->rkey);
		return 0;
1299
	}
1300

1301 1302 1303 1304 1305
	/*
	 * Since not all RDMA HW drivers support non-zero page offsets for
	 * FMR, if we start at an offset into a page, don't merge into the
	 * current FMR mapping. Finish it out, and use the kernel's MR for
	 * this sg entry.
1306
	 */
1307 1308
	if ((!dev->use_fast_reg && dma_addr & ~dev->mr_page_mask) ||
	    dma_len > dev->mr_max_size) {
1309
		ret = srp_finish_mapping(state, ch);
1310 1311 1312 1313 1314 1315
		if (ret)
			return ret;

		srp_map_desc(state, dma_addr, dma_len, target->rkey);
		srp_map_update_start(state, NULL, 0, 0);
		return 0;
1316 1317
	}

1318 1319 1320 1321 1322
	/*
	 * If this is the first sg that will be mapped via FMR or via FR, save
	 * our position. We need to know the first unmapped entry, its index,
	 * and the first unmapped address within that entry to be able to
	 * restart mapping after an error.
1323 1324 1325
	 */
	if (!state->unmapped_sg)
		srp_map_update_start(state, sg, sg_index, dma_addr);
1326

1327
	while (dma_len) {
1328 1329
		unsigned offset = dma_addr & ~dev->mr_page_mask;
		if (state->npages == dev->max_pages_per_mr || offset != 0) {
1330
			ret = srp_finish_mapping(state, ch);
1331 1332
			if (ret)
				return ret;
1333

1334 1335 1336
			srp_map_update_start(state, sg, sg_index, dma_addr);
		}

1337
		len = min_t(unsigned int, dma_len, dev->mr_page_size - offset);
1338

1339 1340
		if (!state->npages)
			state->base_dma_addr = dma_addr;
1341
		state->pages[state->npages++] = dma_addr & dev->mr_page_mask;
1342
		state->dma_len += len;
1343 1344 1345 1346
		dma_addr += len;
		dma_len -= len;
	}

1347 1348
	/*
	 * If the last entry of the MR wasn't a full page, then we need to
1349 1350 1351 1352
	 * close it out and start a new one -- we can only merge at page
	 * boundries.
	 */
	ret = 0;
1353
	if (len != dev->mr_page_size) {
1354
		ret = srp_finish_mapping(state, ch);
1355 1356 1357
		if (!ret)
			srp_map_update_start(state, NULL, 0, 0);
	}
1358 1359 1360
	return ret;
}

1361 1362 1363
static int srp_map_sg(struct srp_map_state *state, struct srp_rdma_ch *ch,
		      struct srp_request *req, struct scatterlist *scat,
		      int count)
1364
{
1365
	struct srp_target_port *target = ch->target;
1366 1367 1368
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = dev->dev;
	struct scatterlist *sg;
1369 1370
	int i;
	bool use_mr;
1371 1372 1373

	state->desc	= req->indirect_desc;
	state->pages	= req->map_page;
1374 1375
	if (dev->use_fast_reg) {
		state->next_fr = req->fr_list;
1376
		use_mr = !!ch->fr_pool;
1377 1378
	} else {
		state->next_fmr = req->fmr_list;
1379
		use_mr = !!ch->fmr_pool;
1380
	}
1381 1382

	for_each_sg(scat, sg, count, i) {
1383
		if (srp_map_sg_entry(state, ch, sg, i, use_mr)) {
1384 1385 1386 1387
			/*
			 * Memory registration failed, so backtrack to the
			 * first unmapped entry and continue on without using
			 * memory registration.
1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399
			 */
			dma_addr_t dma_addr;
			unsigned int dma_len;

backtrack:
			sg = state->unmapped_sg;
			i = state->unmapped_index;

			dma_addr = ib_sg_dma_address(ibdev, sg);
			dma_len = ib_sg_dma_len(ibdev, sg);
			dma_len -= (state->unmapped_addr - dma_addr);
			dma_addr = state->unmapped_addr;
1400
			use_mr = false;
1401 1402 1403 1404
			srp_map_desc(state, dma_addr, dma_len, target->rkey);
		}
	}

1405
	if (use_mr && srp_finish_mapping(state, ch))
1406 1407
		goto backtrack;

1408
	req->nmdesc = state->nmdesc;
1409 1410

	return 0;
1411 1412
}

1413
static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_rdma_ch *ch,
1414 1415
			struct srp_request *req)
{
1416
	struct srp_target_port *target = ch->target;
1417
	struct scatterlist *scat;
1418
	struct srp_cmd *cmd = req->cmd->buf;
1419
	int len, nents, count;
1420 1421
	struct srp_device *dev;
	struct ib_device *ibdev;
1422 1423 1424 1425
	struct srp_map_state state;
	struct srp_indirect_buf *indirect_hdr;
	u32 table_len;
	u8 fmt;
1426

1427
	if (!scsi_sglist(scmnd) || scmnd->sc_data_direction == DMA_NONE)
1428 1429 1430 1431
		return sizeof (struct srp_cmd);

	if (scmnd->sc_data_direction != DMA_FROM_DEVICE &&
	    scmnd->sc_data_direction != DMA_TO_DEVICE) {
1432 1433 1434
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled data direction %d\n",
			     scmnd->sc_data_direction);
1435 1436 1437
		return -EINVAL;
	}

1438 1439
	nents = scsi_sg_count(scmnd);
	scat  = scsi_sglist(scmnd);
1440

1441
	dev = target->srp_host->srp_dev;
1442 1443 1444
	ibdev = dev->dev;

	count = ib_dma_map_sg(ibdev, scat, nents, scmnd->sc_data_direction);
1445 1446
	if (unlikely(count == 0))
		return -EIO;
1447 1448 1449

	fmt = SRP_DATA_DESC_DIRECT;
	len = sizeof (struct srp_cmd) +	sizeof (struct srp_direct_buf);
1450

1451
	if (count == 1 && !register_always) {
1452 1453 1454 1455 1456 1457
		/*
		 * 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.
		 */
1458
		struct srp_direct_buf *buf = (void *) cmd->add_data;
1459

1460
		buf->va  = cpu_to_be64(ib_sg_dma_address(ibdev, scat));
1461
		buf->key = cpu_to_be32(target->rkey);
1462
		buf->len = cpu_to_be32(ib_sg_dma_len(ibdev, scat));
1463

1464
		req->nmdesc = 0;
1465 1466 1467
		goto map_complete;
	}

1468 1469 1470
	/*
	 * We have more than one scatter/gather entry, so build our indirect
	 * descriptor table, trying to merge as many entries as we can.
1471 1472 1473
	 */
	indirect_hdr = (void *) cmd->add_data;

1474 1475 1476
	ib_dma_sync_single_for_cpu(ibdev, req->indirect_dma_addr,
				   target->indirect_size, DMA_TO_DEVICE);

1477
	memset(&state, 0, sizeof(state));
1478
	srp_map_sg(&state, ch, req, scat, count);
1479

1480 1481 1482 1483 1484
	/* 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.
1485 1486
	 */
	if (state.ndesc == 1) {
1487 1488
		/*
		 * Memory registration collapsed the sg-list into one entry,
1489 1490 1491
		 * so use a direct descriptor.
		 */
		struct srp_direct_buf *buf = (void *) cmd->add_data;
1492

1493
		*buf = req->indirect_desc[0];
1494
		goto map_complete;
1495 1496
	}

1497 1498 1499 1500 1501 1502 1503 1504
	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);
1505 1506 1507 1508
	table_len = state.ndesc * sizeof (struct srp_direct_buf);

	fmt = SRP_DATA_DESC_INDIRECT;
	len = sizeof(struct srp_cmd) + sizeof (struct srp_indirect_buf);
1509
	len += count * sizeof (struct srp_direct_buf);
1510

1511 1512
	memcpy(indirect_hdr->desc_list, req->indirect_desc,
	       count * sizeof (struct srp_direct_buf));
1513

1514
	indirect_hdr->table_desc.va = cpu_to_be64(req->indirect_dma_addr);
1515 1516 1517 1518 1519
	indirect_hdr->table_desc.key = cpu_to_be32(target->rkey);
	indirect_hdr->table_desc.len = cpu_to_be32(table_len);
	indirect_hdr->len = cpu_to_be32(state.total_len);

	if (scmnd->sc_data_direction == DMA_TO_DEVICE)
1520
		cmd->data_out_desc_cnt = count;
1521
	else
1522 1523 1524 1525
		cmd->data_in_desc_cnt = count;

	ib_dma_sync_single_for_device(ibdev, req->indirect_dma_addr, table_len,
				      DMA_TO_DEVICE);
1526 1527

map_complete:
1528 1529 1530 1531 1532 1533 1534 1535
	if (scmnd->sc_data_direction == DMA_TO_DEVICE)
		cmd->buf_fmt = fmt << 4;
	else
		cmd->buf_fmt = fmt;

	return len;
}

1536 1537 1538
/*
 * Return an IU and possible credit to the free pool
 */
1539
static void srp_put_tx_iu(struct srp_rdma_ch *ch, struct srp_iu *iu,
1540 1541 1542 1543
			  enum srp_iu_type iu_type)
{
	unsigned long flags;

1544 1545
	spin_lock_irqsave(&ch->lock, flags);
	list_add(&iu->list, &ch->free_tx);
1546
	if (iu_type != SRP_IU_RSP)
1547 1548
		++ch->req_lim;
	spin_unlock_irqrestore(&ch->lock, flags);
1549 1550
}

1551
/*
1552
 * Must be called with ch->lock held to protect req_lim and free_tx.
1553
 * If IU is not sent, it must be returned using srp_put_tx_iu().
1554 1555 1556 1557 1558 1559 1560 1561 1562 1563
 *
 * 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.
 */
1564
static struct srp_iu *__srp_get_tx_iu(struct srp_rdma_ch *ch,
1565 1566
				      enum srp_iu_type iu_type)
{
1567
	struct srp_target_port *target = ch->target;
1568 1569 1570
	s32 rsv = (iu_type == SRP_IU_TSK_MGMT) ? 0 : SRP_TSK_MGMT_SQ_SIZE;
	struct srp_iu *iu;

1571
	srp_send_completion(ch->send_cq, ch);
1572

1573
	if (list_empty(&ch->free_tx))
1574 1575 1576
		return NULL;

	/* Initiator responses to target requests do not consume credits */
1577
	if (iu_type != SRP_IU_RSP) {
1578
		if (ch->req_lim <= rsv) {
1579 1580 1581 1582
			++target->zero_req_lim;
			return NULL;
		}

1583
		--ch->req_lim;
1584 1585
	}

1586
	iu = list_first_entry(&ch->free_tx, struct srp_iu, list);
1587
	list_del(&iu->list);
1588 1589 1590
	return iu;
}

1591
static int srp_post_send(struct srp_rdma_ch *ch, struct srp_iu *iu, int len)
1592
{
1593
	struct srp_target_port *target = ch->target;
1594 1595 1596 1597 1598
	struct ib_sge list;
	struct ib_send_wr wr, *bad_wr;

	list.addr   = iu->dma;
	list.length = len;
1599
	list.lkey   = target->lkey;
1600 1601

	wr.next       = NULL;
1602
	wr.wr_id      = (uintptr_t) iu;
1603 1604 1605 1606 1607
	wr.sg_list    = &list;
	wr.num_sge    = 1;
	wr.opcode     = IB_WR_SEND;
	wr.send_flags = IB_SEND_SIGNALED;

1608
	return ib_post_send(ch->qp, &wr, &bad_wr);
1609 1610
}

1611
static int srp_post_recv(struct srp_rdma_ch *ch, struct srp_iu *iu)
1612
{
1613
	struct srp_target_port *target = ch->target;
1614
	struct ib_recv_wr wr, *bad_wr;
1615
	struct ib_sge list;
1616 1617 1618

	list.addr   = iu->dma;
	list.length = iu->size;
1619
	list.lkey   = target->lkey;
1620 1621

	wr.next     = NULL;
1622
	wr.wr_id    = (uintptr_t) iu;
1623 1624 1625
	wr.sg_list  = &list;
	wr.num_sge  = 1;

1626
	return ib_post_recv(ch->qp, &wr, &bad_wr);
1627 1628
}

1629
static void srp_process_rsp(struct srp_rdma_ch *ch, struct srp_rsp *rsp)
1630
{
1631
	struct srp_target_port *target = ch->target;
1632 1633 1634 1635 1636
	struct srp_request *req;
	struct scsi_cmnd *scmnd;
	unsigned long flags;

	if (unlikely(rsp->tag & SRP_TAG_TSK_MGMT)) {
1637 1638 1639
		spin_lock_irqsave(&ch->lock, flags);
		ch->req_lim += be32_to_cpu(rsp->req_lim_delta);
		spin_unlock_irqrestore(&ch->lock, flags);
1640

1641
		ch->tsk_mgmt_status = -1;
1642
		if (be32_to_cpu(rsp->resp_data_len) >= 4)
1643 1644
			ch->tsk_mgmt_status = rsp->data[3];
		complete(&ch->tsk_mgmt_done);
1645
	} else {
B
Bart Van Assche 已提交
1646 1647 1648 1649 1650
		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
1651
		if (!scmnd) {
1652 1653 1654
			shost_printk(KERN_ERR, target->scsi_host,
				     "Null scmnd for RSP w/tag %016llx\n",
				     (unsigned long long) rsp->tag);
B
Bart Van Assche 已提交
1655

1656 1657 1658
			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
1659 1660 1661

			return;
		}
1662 1663 1664 1665 1666 1667 1668 1669 1670
		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
1671
		if (unlikely(rsp->flags & SRP_RSP_FLAG_DIUNDER))
1672
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_in_res_cnt));
B
Bart Van Assche 已提交
1673 1674 1675 1676 1677 1678
		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));
1679

1680
		srp_free_req(ch, req, scmnd,
B
Bart Van Assche 已提交
1681 1682
			     be32_to_cpu(rsp->req_lim_delta));

1683 1684
		scmnd->host_scribble = NULL;
		scmnd->scsi_done(scmnd);
1685 1686 1687
	}
}

1688
static int srp_response_common(struct srp_rdma_ch *ch, s32 req_delta,
1689 1690
			       void *rsp, int len)
{
1691
	struct srp_target_port *target = ch->target;
1692
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1693 1694
	unsigned long flags;
	struct srp_iu *iu;
1695
	int err;
1696

1697 1698 1699 1700
	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);
1701

1702 1703 1704
	if (!iu) {
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "no IU available to send response\n");
1705
		return 1;
1706 1707 1708 1709 1710 1711
	}

	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);

1712
	err = srp_post_send(ch, iu, len);
1713
	if (err) {
1714 1715
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "unable to post response: %d\n", err);
1716
		srp_put_tx_iu(ch, iu, SRP_IU_RSP);
1717
	}
1718 1719 1720 1721

	return err;
}

1722
static void srp_process_cred_req(struct srp_rdma_ch *ch,
1723 1724 1725 1726 1727 1728 1729 1730
				 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);

1731 1732
	if (srp_response_common(ch, delta, &rsp, sizeof(rsp)))
		shost_printk(KERN_ERR, ch->target->scsi_host, PFX
1733 1734 1735
			     "problems processing SRP_CRED_REQ\n");
}

1736
static void srp_process_aer_req(struct srp_rdma_ch *ch,
1737 1738
				struct srp_aer_req *req)
{
1739
	struct srp_target_port *target = ch->target;
1740 1741 1742 1743 1744 1745 1746 1747 1748
	struct srp_aer_rsp rsp = {
		.opcode = SRP_AER_RSP,
		.tag = req->tag,
	};
	s32 delta = be32_to_cpu(req->req_lim_delta);

	shost_printk(KERN_ERR, target->scsi_host, PFX
		     "ignoring AER for LUN %llu\n", be64_to_cpu(req->lun));

1749
	if (srp_response_common(ch, delta, &rsp, sizeof(rsp)))
1750 1751 1752 1753
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "problems processing SRP_AER_REQ\n");
}

1754
static void srp_handle_recv(struct srp_rdma_ch *ch, struct ib_wc *wc)
1755
{
1756
	struct srp_target_port *target = ch->target;
1757
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1758
	struct srp_iu *iu = (struct srp_iu *) (uintptr_t) wc->wr_id;
1759
	int res;
1760 1761
	u8 opcode;

1762
	ib_dma_sync_single_for_cpu(dev, iu->dma, ch->max_ti_iu_len,
1763
				   DMA_FROM_DEVICE);
1764 1765 1766 1767

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

	if (0) {
1768 1769
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "recv completion, opcode 0x%02x\n", opcode);
B
Bart Van Assche 已提交
1770 1771
		print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET, 8, 1,
			       iu->buf, wc->byte_len, true);
1772 1773 1774 1775
	}

	switch (opcode) {
	case SRP_RSP:
1776
		srp_process_rsp(ch, iu->buf);
1777 1778
		break;

1779
	case SRP_CRED_REQ:
1780
		srp_process_cred_req(ch, iu->buf);
1781 1782 1783
		break;

	case SRP_AER_REQ:
1784
		srp_process_aer_req(ch, iu->buf);
1785 1786
		break;

1787 1788
	case SRP_T_LOGOUT:
		/* XXX Handle target logout */
1789 1790
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Got target logout request\n");
1791 1792 1793
		break;

	default:
1794 1795
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled SRP opcode 0x%02x\n", opcode);
1796 1797 1798
		break;
	}

1799
	ib_dma_sync_single_for_device(dev, iu->dma, ch->max_ti_iu_len,
1800
				      DMA_FROM_DEVICE);
1801

1802
	res = srp_post_recv(ch, iu);
1803 1804 1805
	if (res != 0)
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Recv failed with error code %d\n", res);
1806 1807
}

1808 1809
/**
 * srp_tl_err_work() - handle a transport layer error
1810
 * @work: Work structure embedded in an SRP target port.
1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823
 *
 * 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);
}

1824 1825
static void srp_handle_qp_err(u64 wr_id, enum ib_wc_status wc_status,
			      bool send_err, struct srp_target_port *target)
1826
{
1827
	if (target->connected && !target->qp_in_error) {
1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841
		if (wr_id & LOCAL_INV_WR_ID_MASK) {
			shost_printk(KERN_ERR, target->scsi_host, PFX
				     "LOCAL_INV failed with status %d\n",
				     wc_status);
		} else if (wr_id & FAST_REG_WR_ID_MASK) {
			shost_printk(KERN_ERR, target->scsi_host, PFX
				     "FAST_REG_MR failed status %d\n",
				     wc_status);
		} else {
			shost_printk(KERN_ERR, target->scsi_host,
				     PFX "failed %s status %d for iu %p\n",
				     send_err ? "send" : "receive",
				     wc_status, (void *)(uintptr_t)wr_id);
		}
1842
		queue_work(system_long_wq, &target->tl_err_work);
1843
	}
1844 1845 1846
	target->qp_in_error = true;
}

1847
static void srp_recv_completion(struct ib_cq *cq, void *ch_ptr)
1848
{
1849
	struct srp_rdma_ch *ch = ch_ptr;
1850 1851 1852 1853
	struct ib_wc wc;

	ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
	while (ib_poll_cq(cq, 1, &wc) > 0) {
1854
		if (likely(wc.status == IB_WC_SUCCESS)) {
1855
			srp_handle_recv(ch, &wc);
1856
		} else {
1857 1858
			srp_handle_qp_err(wc.wr_id, wc.status, false,
					  ch->target);
1859
		}
1860 1861 1862
	}
}

1863
static void srp_send_completion(struct ib_cq *cq, void *ch_ptr)
1864
{
1865
	struct srp_rdma_ch *ch = ch_ptr;
1866
	struct ib_wc wc;
1867
	struct srp_iu *iu;
1868 1869

	while (ib_poll_cq(cq, 1, &wc) > 0) {
1870 1871
		if (likely(wc.status == IB_WC_SUCCESS)) {
			iu = (struct srp_iu *) (uintptr_t) wc.wr_id;
1872
			list_add(&iu->list, &ch->free_tx);
1873
		} else {
1874 1875
			srp_handle_qp_err(wc.wr_id, wc.status, true,
					  ch->target);
1876
		}
1877 1878 1879
	}
}

1880
static int srp_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *scmnd)
1881
{
1882
	struct srp_target_port *target = host_to_target(shost);
1883
	struct srp_rport *rport = target->rport;
1884
	struct srp_rdma_ch *ch;
1885 1886 1887
	struct srp_request *req;
	struct srp_iu *iu;
	struct srp_cmd *cmd;
1888
	struct ib_device *dev;
1889
	unsigned long flags;
B
Bart Van Assche 已提交
1890 1891
	u32 tag;
	u16 idx;
1892
	int len, ret;
1893 1894 1895 1896 1897 1898 1899 1900 1901 1902
	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);
1903

1904 1905 1906
	scmnd->result = srp_chkready(target->rport);
	if (unlikely(scmnd->result))
		goto err;
1907

B
Bart Van Assche 已提交
1908 1909
	WARN_ON_ONCE(scmnd->request->tag < 0);
	tag = blk_mq_unique_tag(scmnd->request);
1910
	ch = &target->ch;
B
Bart Van Assche 已提交
1911 1912 1913 1914
	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);
1915 1916 1917 1918

	spin_lock_irqsave(&ch->lock, flags);
	iu = __srp_get_tx_iu(ch, SRP_IU_CMD);
	spin_unlock_irqrestore(&ch->lock, flags);
1919

B
Bart Van Assche 已提交
1920 1921 1922 1923
	if (!iu)
		goto err;

	req = &ch->req_ring[idx];
1924
	dev = target->srp_host->srp_dev->dev;
1925
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_iu_len,
1926
				   DMA_TO_DEVICE);
1927

1928
	scmnd->host_scribble = (void *) req;
1929 1930 1931 1932 1933 1934

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

	cmd->opcode = SRP_CMD;
	cmd->lun    = cpu_to_be64((u64) scmnd->device->lun << 48);
B
Bart Van Assche 已提交
1935
	cmd->tag    = tag;
1936 1937 1938 1939 1940
	memcpy(cmd->cdb, scmnd->cmnd, scmnd->cmd_len);

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

1941
	len = srp_map_data(scmnd, ch, req);
1942
	if (len < 0) {
1943
		shost_printk(KERN_ERR, target->scsi_host,
1944 1945 1946 1947
			     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
1948
		 * max_pages_per_mr sg-list elements, tell the SCSI mid-layer
1949 1950 1951 1952
		 * to reduce queue depth temporarily.
		 */
		scmnd->result = len == -ENOMEM ?
			DID_OK << 16 | QUEUE_FULL << 1 : DID_ERROR << 16;
1953
		goto err_iu;
1954 1955
	}

1956
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_iu_len,
1957
				      DMA_TO_DEVICE);
1958

1959
	if (srp_post_send(ch, iu, len)) {
1960
		shost_printk(KERN_ERR, target->scsi_host, PFX "Send failed\n");
1961 1962 1963
		goto err_unmap;
	}

1964 1965
	ret = 0;

1966 1967 1968 1969
unlock_rport:
	if (in_scsi_eh)
		mutex_unlock(&rport->mutex);

1970
	return ret;
1971 1972

err_unmap:
1973
	srp_unmap_data(scmnd, ch, req);
1974

1975
err_iu:
1976
	srp_put_tx_iu(ch, iu, SRP_IU_CMD);
1977

1978 1979 1980 1981 1982 1983
	/*
	 * Avoid that the loops that iterate over the request ring can
	 * encounter a dangling SCSI command pointer.
	 */
	req->scmnd = NULL;

1984 1985 1986 1987 1988 1989 1990
err:
	if (scmnd->result) {
		scmnd->scsi_done(scmnd);
		ret = 0;
	} else {
		ret = SCSI_MLQUEUE_HOST_BUSY;
	}
1991

1992
	goto unlock_rport;
1993 1994
}

1995 1996
/*
 * Note: the resources allocated in this function are freed in
1997
 * srp_free_ch_ib().
1998
 */
1999
static int srp_alloc_iu_bufs(struct srp_rdma_ch *ch)
2000
{
2001
	struct srp_target_port *target = ch->target;
2002 2003
	int i;

2004 2005 2006
	ch->rx_ring = kcalloc(target->queue_size, sizeof(*ch->rx_ring),
			      GFP_KERNEL);
	if (!ch->rx_ring)
2007
		goto err_no_ring;
2008 2009 2010
	ch->tx_ring = kcalloc(target->queue_size, sizeof(*ch->tx_ring),
			      GFP_KERNEL);
	if (!ch->tx_ring)
2011 2012 2013
		goto err_no_ring;

	for (i = 0; i < target->queue_size; ++i) {
2014 2015 2016 2017
		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])
2018 2019 2020
			goto err;
	}

2021
	for (i = 0; i < target->queue_size; ++i) {
2022 2023 2024 2025
		ch->tx_ring[i] = srp_alloc_iu(target->srp_host,
					      target->max_iu_len,
					      GFP_KERNEL, DMA_TO_DEVICE);
		if (!ch->tx_ring[i])
2026
			goto err;
2027

2028
		list_add(&ch->tx_ring[i]->list, &ch->free_tx);
2029 2030 2031 2032 2033
	}

	return 0;

err:
2034
	for (i = 0; i < target->queue_size; ++i) {
2035 2036
		srp_free_iu(target->srp_host, ch->rx_ring[i]);
		srp_free_iu(target->srp_host, ch->tx_ring[i]);
2037 2038
	}

2039 2040

err_no_ring:
2041 2042 2043 2044
	kfree(ch->tx_ring);
	ch->tx_ring = NULL;
	kfree(ch->rx_ring);
	ch->rx_ring = NULL;
2045

2046 2047 2048
	return -ENOMEM;
}

2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075
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;
}

2076 2077
static void srp_cm_rep_handler(struct ib_cm_id *cm_id,
			       struct srp_login_rsp *lrsp,
2078
			       struct srp_rdma_ch *ch)
2079
{
2080
	struct srp_target_port *target = ch->target;
2081 2082 2083 2084 2085 2086
	struct ib_qp_attr *qp_attr = NULL;
	int attr_mask = 0;
	int ret;
	int i;

	if (lrsp->opcode == SRP_LOGIN_RSP) {
2087 2088
		ch->max_ti_iu_len = be32_to_cpu(lrsp->max_ti_iu_len);
		ch->req_lim       = be32_to_cpu(lrsp->req_lim_delta);
2089 2090 2091 2092 2093 2094

		/*
		 * Reserve credits for task management so we don't
		 * bounce requests back to the SCSI mid-layer.
		 */
		target->scsi_host->can_queue
2095
			= min(ch->req_lim - SRP_TSK_MGMT_SQ_SIZE,
2096
			      target->scsi_host->can_queue);
2097 2098 2099
		target->scsi_host->cmd_per_lun
			= min_t(int, target->scsi_host->can_queue,
				target->scsi_host->cmd_per_lun);
2100 2101 2102 2103 2104 2105 2106
	} else {
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled RSP opcode %#x\n", lrsp->opcode);
		ret = -ECONNRESET;
		goto error;
	}

2107 2108
	if (!ch->rx_ring) {
		ret = srp_alloc_iu_bufs(ch);
2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122
		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;

2123
	ret = ib_modify_qp(ch->qp, qp_attr, attr_mask);
2124 2125 2126
	if (ret)
		goto error_free;

2127
	for (i = 0; i < target->queue_size; i++) {
2128 2129 2130
		struct srp_iu *iu = ch->rx_ring[i];

		ret = srp_post_recv(ch, iu);
2131 2132 2133 2134 2135 2136 2137 2138 2139
		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;

2140 2141
	target->rq_tmo_jiffies = srp_compute_rq_tmo(qp_attr, attr_mask);

2142
	ret = ib_modify_qp(ch->qp, qp_attr, attr_mask);
2143 2144 2145 2146 2147 2148 2149 2150 2151
	if (ret)
		goto error_free;

	ret = ib_send_cm_rtu(cm_id, NULL, 0);

error_free:
	kfree(qp_attr);

error:
2152
	ch->status = ret;
2153 2154
}

2155 2156
static void srp_cm_rej_handler(struct ib_cm_id *cm_id,
			       struct ib_cm_event *event,
2157
			       struct srp_rdma_ch *ch)
2158
{
2159
	struct srp_target_port *target = ch->target;
2160
	struct Scsi_Host *shost = target->scsi_host;
2161 2162 2163 2164 2165 2166
	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;
2167 2168
		ch->path.dlid = cpi->redirect_lid;
		ch->path.pkey = cpi->redirect_pkey;
2169
		cm_id->remote_cm_qpn = be32_to_cpu(cpi->redirect_qp) & 0x00ffffff;
2170
		memcpy(ch->path.dgid.raw, cpi->redirect_gid, 16);
2171

2172
		ch->status = ch->path.dlid ?
2173 2174 2175 2176
			SRP_DLID_REDIRECT : SRP_PORT_REDIRECT;
		break;

	case IB_CM_REJ_PORT_REDIRECT:
2177
		if (srp_target_is_topspin(target)) {
2178 2179 2180 2181 2182
			/*
			 * Topspin/Cisco SRP gateways incorrectly send
			 * reject reason code 25 when they mean 24
			 * (port redirect).
			 */
2183
			memcpy(ch->path.dgid.raw,
2184 2185
			       event->param.rej_rcvd.ari, 16);

2186 2187
			shost_printk(KERN_DEBUG, shost,
				     PFX "Topspin/Cisco redirect to target port GID %016llx%016llx\n",
2188 2189
				     be64_to_cpu(ch->path.dgid.global.subnet_prefix),
				     be64_to_cpu(ch->path.dgid.global.interface_id));
2190

2191
			ch->status = SRP_PORT_REDIRECT;
2192
		} else {
2193 2194
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_PORT_REDIRECT\n");
2195
			ch->status = -ECONNRESET;
2196 2197 2198 2199
		}
		break;

	case IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID:
2200 2201
		shost_printk(KERN_WARNING, shost,
			    "  REJ reason: IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID\n");
2202
		ch->status = -ECONNRESET;
2203 2204 2205 2206 2207 2208 2209 2210 2211
		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)
2212 2213
				shost_printk(KERN_WARNING, shost,
					     PFX "SRP_LOGIN_REJ: requested max_it_iu_len too large\n");
2214
			else
B
Bart Van Assche 已提交
2215 2216
				shost_printk(KERN_WARNING, shost, PFX
					     "SRP LOGIN from %pI6 to %pI6 REJECTED, reason 0x%08x\n",
2217 2218
					     target->sgid.raw,
					     target->orig_dgid.raw, reason);
2219
		} else
2220 2221 2222
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_CONSUMER_DEFINED,"
				     " opcode 0x%02x\n", opcode);
2223
		ch->status = -ECONNRESET;
2224 2225
		break;

D
David Dillow 已提交
2226 2227
	case IB_CM_REJ_STALE_CONN:
		shost_printk(KERN_WARNING, shost, "  REJ reason: stale connection\n");
2228
		ch->status = SRP_STALE_CONN;
D
David Dillow 已提交
2229 2230
		break;

2231
	default:
2232 2233
		shost_printk(KERN_WARNING, shost, "  REJ reason 0x%x\n",
			     event->param.rej_rcvd.reason);
2234
		ch->status = -ECONNRESET;
2235 2236 2237 2238 2239
	}
}

static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event)
{
2240 2241
	struct srp_rdma_ch *ch = cm_id->context;
	struct srp_target_port *target = ch->target;
2242 2243 2244 2245
	int comp = 0;

	switch (event->event) {
	case IB_CM_REQ_ERROR:
2246 2247
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Sending CM REQ failed\n");
2248
		comp = 1;
2249
		ch->status = -ECONNRESET;
2250 2251 2252 2253
		break;

	case IB_CM_REP_RECEIVED:
		comp = 1;
2254
		srp_cm_rep_handler(cm_id, event->private_data, ch);
2255 2256 2257
		break;

	case IB_CM_REJ_RECEIVED:
2258
		shost_printk(KERN_DEBUG, target->scsi_host, PFX "REJ received\n");
2259 2260
		comp = 1;

2261
		srp_cm_rej_handler(cm_id, event, ch);
2262 2263
		break;

2264
	case IB_CM_DREQ_RECEIVED:
2265 2266
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "DREQ received - connection closed\n");
2267
		srp_change_conn_state(target, false);
2268
		if (ib_send_cm_drep(cm_id, NULL, 0))
2269 2270
			shost_printk(KERN_ERR, target->scsi_host,
				     PFX "Sending CM DREP failed\n");
2271
		queue_work(system_long_wq, &target->tl_err_work);
2272 2273 2274
		break;

	case IB_CM_TIMEWAIT_EXIT:
2275 2276
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "connection closed\n");
2277
		comp = 1;
2278

2279
		ch->status = 0;
2280 2281
		break;

2282 2283 2284 2285 2286
	case IB_CM_MRA_RECEIVED:
	case IB_CM_DREQ_ERROR:
	case IB_CM_DREP_RECEIVED:
		break;

2287
	default:
2288 2289
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled CM event %d\n", event->event);
2290 2291 2292 2293
		break;
	}

	if (comp)
2294
		complete(&ch->done);
2295 2296 2297 2298

	return 0;
}

2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318
/**
 * srp_change_queue_depth - setting device queue depth
 * @sdev: scsi device struct
 * @qdepth: requested queue depth
 * @reason: SCSI_QDEPTH_DEFAULT/SCSI_QDEPTH_QFULL/SCSI_QDEPTH_RAMP_UP
 * (see include/scsi/scsi_host.h for definition)
 *
 * Returns queue depth.
 */
static int
srp_change_queue_depth(struct scsi_device *sdev, int qdepth, int reason)
{
	struct Scsi_Host *shost = sdev->host;
	int max_depth;
	if (reason == SCSI_QDEPTH_DEFAULT || reason == SCSI_QDEPTH_RAMP_UP) {
		max_depth = shost->can_queue;
		if (!sdev->tagged_supported)
			max_depth = 1;
		if (qdepth > max_depth)
			qdepth = max_depth;
2319
		scsi_adjust_queue_depth(sdev, qdepth);
2320 2321 2322 2323 2324 2325 2326 2327
	} else if (reason == SCSI_QDEPTH_QFULL)
		scsi_track_queue_full(sdev, qdepth);
	else
		return -EOPNOTSUPP;

	return sdev->queue_depth;
}

2328 2329
static int srp_send_tsk_mgmt(struct srp_rdma_ch *ch, u64 req_tag,
			     unsigned int lun, u8 func)
2330
{
2331
	struct srp_target_port *target = ch->target;
2332
	struct srp_rport *rport = target->rport;
2333
	struct ib_device *dev = target->srp_host->srp_dev->dev;
2334 2335 2336
	struct srp_iu *iu;
	struct srp_tsk_mgmt *tsk_mgmt;

2337 2338 2339
	if (!target->connected || target->qp_in_error)
		return -1;

2340
	init_completion(&ch->tsk_mgmt_done);
2341

2342
	/*
2343
	 * Lock the rport mutex to avoid that srp_create_ch_ib() is
2344 2345 2346
	 * invoked while a task management function is being sent.
	 */
	mutex_lock(&rport->mutex);
2347 2348 2349
	spin_lock_irq(&ch->lock);
	iu = __srp_get_tx_iu(ch, SRP_IU_TSK_MGMT);
	spin_unlock_irq(&ch->lock);
2350

2351 2352 2353
	if (!iu) {
		mutex_unlock(&rport->mutex);

2354
		return -1;
2355
	}
2356

2357 2358
	ib_dma_sync_single_for_cpu(dev, iu->dma, sizeof *tsk_mgmt,
				   DMA_TO_DEVICE);
2359 2360 2361 2362
	tsk_mgmt = iu->buf;
	memset(tsk_mgmt, 0, sizeof *tsk_mgmt);

	tsk_mgmt->opcode 	= SRP_TSK_MGMT;
2363 2364
	tsk_mgmt->lun		= cpu_to_be64((u64) lun << 48);
	tsk_mgmt->tag		= req_tag | SRP_TAG_TSK_MGMT;
2365
	tsk_mgmt->tsk_mgmt_func = func;
2366
	tsk_mgmt->task_tag	= req_tag;
2367

2368 2369
	ib_dma_sync_single_for_device(dev, iu->dma, sizeof *tsk_mgmt,
				      DMA_TO_DEVICE);
2370 2371
	if (srp_post_send(ch, iu, sizeof(*tsk_mgmt))) {
		srp_put_tx_iu(ch, iu, SRP_IU_TSK_MGMT);
2372 2373
		mutex_unlock(&rport->mutex);

2374 2375
		return -1;
	}
2376
	mutex_unlock(&rport->mutex);
2377

2378
	if (!wait_for_completion_timeout(&ch->tsk_mgmt_done,
2379
					 msecs_to_jiffies(SRP_ABORT_TIMEOUT_MS)))
2380
		return -1;
2381

2382 2383 2384
	return 0;
}

2385 2386
static int srp_abort(struct scsi_cmnd *scmnd)
{
2387
	struct srp_target_port *target = host_to_target(scmnd->device->host);
2388
	struct srp_request *req = (struct srp_request *) scmnd->host_scribble;
B
Bart Van Assche 已提交
2389
	u32 tag;
2390
	struct srp_rdma_ch *ch;
2391
	int ret;
2392

2393
	shost_printk(KERN_ERR, target->scsi_host, "SRP abort called\n");
2394

2395 2396
	ch = &target->ch;
	if (!req || !srp_claim_req(ch, req, NULL, scmnd))
2397
		return SUCCESS;
B
Bart Van Assche 已提交
2398 2399
	tag = blk_mq_unique_tag(scmnd->request);
	if (srp_send_tsk_mgmt(ch, tag, scmnd->device->lun,
2400
			      SRP_TSK_ABORT_TASK) == 0)
2401
		ret = SUCCESS;
2402
	else if (target->rport->state == SRP_RPORT_LOST)
2403
		ret = FAST_IO_FAIL;
2404 2405
	else
		ret = FAILED;
2406
	srp_free_req(ch, req, scmnd, 0);
B
Bart Van Assche 已提交
2407
	scmnd->result = DID_ABORT << 16;
2408
	scmnd->scsi_done(scmnd);
2409

2410
	return ret;
2411 2412 2413 2414
}

static int srp_reset_device(struct scsi_cmnd *scmnd)
{
2415
	struct srp_target_port *target = host_to_target(scmnd->device->host);
2416
	struct srp_rdma_ch *ch = &target->ch;
2417
	int i;
2418

2419
	shost_printk(KERN_ERR, target->scsi_host, "SRP reset_device called\n");
2420

2421
	if (srp_send_tsk_mgmt(ch, SRP_TAG_NO_REQ, scmnd->device->lun,
2422
			      SRP_TSK_LUN_RESET))
2423
		return FAILED;
2424
	if (ch->tsk_mgmt_status)
2425 2426
		return FAILED;

2427
	for (i = 0; i < target->req_ring_size; ++i) {
2428 2429 2430
		struct srp_request *req = &ch->req_ring[i];

		srp_finish_req(ch, req, scmnd->device, DID_RESET << 16);
2431
	}
2432 2433

	return SUCCESS;
2434 2435 2436 2437 2438 2439
}

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

2440
	shost_printk(KERN_ERR, target->scsi_host, PFX "SRP reset_host called\n");
2441

2442
	return srp_reconnect_rport(target->rport) == 0 ? SUCCESS : FAILED;
2443 2444
}

2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459
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;
}

2460 2461
static ssize_t show_id_ext(struct device *dev, struct device_attribute *attr,
			   char *buf)
2462
{
2463
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2464 2465 2466 2467 2468

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

2469 2470
static ssize_t show_ioc_guid(struct device *dev, struct device_attribute *attr,
			     char *buf)
2471
{
2472
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2473 2474 2475 2476 2477

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

2478 2479
static ssize_t show_service_id(struct device *dev,
			       struct device_attribute *attr, char *buf)
2480
{
2481
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2482 2483 2484 2485 2486

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

2487 2488
static ssize_t show_pkey(struct device *dev, struct device_attribute *attr,
			 char *buf)
2489
{
2490
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2491

2492
	return sprintf(buf, "0x%04x\n", be16_to_cpu(target->pkey));
2493 2494
}

B
Bart Van Assche 已提交
2495 2496 2497 2498 2499
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));

2500
	return sprintf(buf, "%pI6\n", target->sgid.raw);
B
Bart Van Assche 已提交
2501 2502
}

2503 2504
static ssize_t show_dgid(struct device *dev, struct device_attribute *attr,
			 char *buf)
2505
{
2506
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2507
	struct srp_rdma_ch *ch = &target->ch;
2508

2509
	return sprintf(buf, "%pI6\n", ch->path.dgid.raw);
2510 2511
}

2512 2513
static ssize_t show_orig_dgid(struct device *dev,
			      struct device_attribute *attr, char *buf)
2514
{
2515
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2516

2517
	return sprintf(buf, "%pI6\n", target->orig_dgid.raw);
2518 2519
}

2520 2521 2522 2523 2524
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));

2525
	return sprintf(buf, "%d\n", target->ch.req_lim);
2526 2527
}

2528 2529
static ssize_t show_zero_req_lim(struct device *dev,
				 struct device_attribute *attr, char *buf)
2530
{
2531
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2532 2533 2534 2535

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

2536 2537
static ssize_t show_local_ib_port(struct device *dev,
				  struct device_attribute *attr, char *buf)
2538
{
2539
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2540 2541 2542 2543

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

2544 2545
static ssize_t show_local_ib_device(struct device *dev,
				    struct device_attribute *attr, char *buf)
2546
{
2547
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2548

2549
	return sprintf(buf, "%s\n", target->srp_host->srp_dev->dev->name);
2550 2551
}

2552 2553 2554 2555 2556 2557 2558 2559
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);
}

2560 2561 2562 2563 2564 2565 2566 2567
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);
}

2568 2569 2570 2571 2572 2573 2574 2575
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);
}

2576 2577 2578 2579 2580 2581 2582 2583
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");
}

2584 2585 2586 2587
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
2588
static DEVICE_ATTR(sgid,	    S_IRUGO, show_sgid,		   NULL);
2589 2590
static DEVICE_ATTR(dgid,	    S_IRUGO, show_dgid,		   NULL);
static DEVICE_ATTR(orig_dgid,	    S_IRUGO, show_orig_dgid,	   NULL);
2591
static DEVICE_ATTR(req_lim,         S_IRUGO, show_req_lim,         NULL);
2592 2593 2594
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);
2595
static DEVICE_ATTR(comp_vector,     S_IRUGO, show_comp_vector,     NULL);
2596
static DEVICE_ATTR(tl_retry_count,  S_IRUGO, show_tl_retry_count,  NULL);
2597
static DEVICE_ATTR(cmd_sg_entries,  S_IRUGO, show_cmd_sg_entries,  NULL);
2598
static DEVICE_ATTR(allow_ext_sg,    S_IRUGO, show_allow_ext_sg,    NULL);
2599 2600 2601 2602 2603 2604

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
2605
	&dev_attr_sgid,
2606 2607
	&dev_attr_dgid,
	&dev_attr_orig_dgid,
2608
	&dev_attr_req_lim,
2609 2610 2611
	&dev_attr_zero_req_lim,
	&dev_attr_local_ib_port,
	&dev_attr_local_ib_device,
2612
	&dev_attr_comp_vector,
2613
	&dev_attr_tl_retry_count,
2614
	&dev_attr_cmd_sg_entries,
2615
	&dev_attr_allow_ext_sg,
2616 2617 2618
	NULL
};

2619 2620
static struct scsi_host_template srp_template = {
	.module				= THIS_MODULE,
R
Roland Dreier 已提交
2621 2622
	.name				= "InfiniBand SRP initiator",
	.proc_name			= DRV_NAME,
2623
	.slave_configure		= srp_slave_configure,
2624 2625
	.info				= srp_target_info,
	.queuecommand			= srp_queuecommand,
2626
	.change_queue_depth             = srp_change_queue_depth,
2627
	.change_queue_type              = scsi_change_queue_type,
2628 2629 2630
	.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
2631
	.skip_settle_delay		= true,
2632
	.sg_tablesize			= SRP_DEF_SG_TABLESIZE,
2633
	.can_queue			= SRP_DEFAULT_CMD_SQ_SIZE,
2634
	.this_id			= -1,
2635
	.cmd_per_lun			= SRP_DEFAULT_CMD_SQ_SIZE,
2636
	.use_clustering			= ENABLE_CLUSTERING,
B
Bart Van Assche 已提交
2637 2638
	.shost_attrs			= srp_host_attrs,
	.use_blk_tags			= 1,
2639 2640
};

2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651
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;
}

2652 2653
static int srp_add_target(struct srp_host *host, struct srp_target_port *target)
{
2654 2655 2656
	struct srp_rport_identifiers ids;
	struct srp_rport *rport;

2657
	target->state = SRP_TARGET_SCANNING;
2658 2659 2660
	sprintf(target->target_name, "SRP.T10:%016llX",
		 (unsigned long long) be64_to_cpu(target->id_ext));

2661
	if (scsi_add_host(target->scsi_host, host->srp_dev->dev->dma_device))
2662 2663
		return -ENODEV;

2664 2665
	memcpy(ids.port_id, &target->id_ext, 8);
	memcpy(ids.port_id + 8, &target->ioc_guid, 8);
2666
	ids.roles = SRP_RPORT_ROLE_TARGET;
2667 2668 2669 2670 2671 2672
	rport = srp_rport_add(target->scsi_host, &ids);
	if (IS_ERR(rport)) {
		scsi_remove_host(target->scsi_host);
		return PTR_ERR(rport);
	}

2673
	rport->lld_data = target;
2674
	target->rport = rport;
2675

2676
	spin_lock(&host->target_lock);
2677
	list_add_tail(&target->list, &host->target_list);
2678
	spin_unlock(&host->target_lock);
2679 2680

	scsi_scan_target(&target->scsi_host->shost_gendev,
2681
			 0, target->scsi_id, SCAN_WILD_CARD, 0);
2682

2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699
	if (!target->connected || target->qp_in_error) {
		shost_printk(KERN_INFO, target->scsi_host,
			     PFX "SCSI scan failed - removing SCSI host\n");
		srp_queue_remove_work(target);
		goto out;
	}

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

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

out:
2700 2701 2702
	return 0;
}

2703
static void srp_release_dev(struct device *dev)
2704 2705
{
	struct srp_host *host =
2706
		container_of(dev, struct srp_host, dev);
2707 2708 2709 2710 2711 2712

	complete(&host->released);
}

static struct class srp_class = {
	.name    = "infiniband_srp",
2713
	.dev_release = srp_release_dev
2714 2715
};

2716 2717
/**
 * srp_conn_unique() - check whether the connection to a target is unique
2718 2719
 * @host:   SRP host.
 * @target: SRP target port.
2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747
 */
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;
}

2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763
/*
 * 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,
2764
	SRP_OPT_MAX_CMD_PER_LUN	= 1 << 6,
2765
	SRP_OPT_IO_CLASS	= 1 << 7,
2766
	SRP_OPT_INITIATOR_EXT	= 1 << 8,
2767
	SRP_OPT_CMD_SG_ENTRIES	= 1 << 9,
2768 2769
	SRP_OPT_ALLOW_EXT_SG	= 1 << 10,
	SRP_OPT_SG_TABLESIZE	= 1 << 11,
2770
	SRP_OPT_COMP_VECTOR	= 1 << 12,
2771
	SRP_OPT_TL_RETRY_COUNT	= 1 << 13,
2772
	SRP_OPT_QUEUE_SIZE	= 1 << 14,
2773 2774 2775 2776 2777 2778 2779
	SRP_OPT_ALL		= (SRP_OPT_ID_EXT	|
				   SRP_OPT_IOC_GUID	|
				   SRP_OPT_DGID		|
				   SRP_OPT_PKEY		|
				   SRP_OPT_SERVICE_ID),
};

2780
static const match_table_t srp_opt_tokens = {
2781 2782 2783 2784 2785 2786 2787
	{ 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" 	},
2788
	{ SRP_OPT_IO_CLASS,		"io_class=%x"		},
2789
	{ SRP_OPT_INITIATOR_EXT,	"initiator_ext=%s"	},
2790
	{ SRP_OPT_CMD_SG_ENTRIES,	"cmd_sg_entries=%u"	},
2791 2792
	{ SRP_OPT_ALLOW_EXT_SG,		"allow_ext_sg=%u"	},
	{ SRP_OPT_SG_TABLESIZE,		"sg_tablesize=%u"	},
2793
	{ SRP_OPT_COMP_VECTOR,		"comp_vector=%u"	},
2794
	{ SRP_OPT_TL_RETRY_COUNT,	"tl_retry_count=%u"	},
2795
	{ SRP_OPT_QUEUE_SIZE,		"queue_size=%d"		},
2796
	{ SRP_OPT_ERR,			NULL 			}
2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824
};

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

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

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

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

		switch (token) {
		case SRP_OPT_ID_EXT:
			p = match_strdup(args);
2825 2826 2827 2828
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2829 2830 2831 2832 2833 2834
			target->id_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_IOC_GUID:
			p = match_strdup(args);
2835 2836 2837 2838
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2839 2840 2841 2842 2843 2844
			target->ioc_guid = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_DGID:
			p = match_strdup(args);
2845 2846 2847 2848
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2849
			if (strlen(p) != 32) {
2850
				pr_warn("bad dest GID parameter '%s'\n", p);
2851
				kfree(p);
2852 2853 2854 2855
				goto out;
			}

			for (i = 0; i < 16; ++i) {
2856 2857 2858 2859 2860 2861 2862
				strlcpy(dgid, p + i * 2, sizeof(dgid));
				if (sscanf(dgid, "%hhx",
					   &target->orig_dgid.raw[i]) < 1) {
					ret = -EINVAL;
					kfree(p);
					goto out;
				}
2863
			}
2864
			kfree(p);
2865 2866 2867 2868
			break;

		case SRP_OPT_PKEY:
			if (match_hex(args, &token)) {
2869
				pr_warn("bad P_Key parameter '%s'\n", p);
2870 2871
				goto out;
			}
2872
			target->pkey = cpu_to_be16(token);
2873 2874 2875 2876
			break;

		case SRP_OPT_SERVICE_ID:
			p = match_strdup(args);
2877 2878 2879 2880
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2881 2882 2883 2884 2885 2886
			target->service_id = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_MAX_SECT:
			if (match_int(args, &token)) {
2887
				pr_warn("bad max sect parameter '%s'\n", p);
2888 2889 2890 2891 2892
				goto out;
			}
			target->scsi_host->max_sectors = token;
			break;

2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904
		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;

2905
		case SRP_OPT_MAX_CMD_PER_LUN:
2906
			if (match_int(args, &token) || token < 1) {
2907 2908
				pr_warn("bad max cmd_per_lun parameter '%s'\n",
					p);
2909 2910
				goto out;
			}
2911
			target->scsi_host->cmd_per_lun = token;
2912 2913
			break;

2914 2915
		case SRP_OPT_IO_CLASS:
			if (match_hex(args, &token)) {
2916
				pr_warn("bad IO class parameter '%s'\n", p);
2917 2918 2919 2920
				goto out;
			}
			if (token != SRP_REV10_IB_IO_CLASS &&
			    token != SRP_REV16A_IB_IO_CLASS) {
2921 2922 2923
				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);
2924 2925 2926 2927 2928
				goto out;
			}
			target->io_class = token;
			break;

2929 2930
		case SRP_OPT_INITIATOR_EXT:
			p = match_strdup(args);
2931 2932 2933 2934
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2935 2936 2937 2938
			target->initiator_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

2939 2940
		case SRP_OPT_CMD_SG_ENTRIES:
			if (match_int(args, &token) || token < 1 || token > 255) {
2941 2942
				pr_warn("bad max cmd_sg_entries parameter '%s'\n",
					p);
2943 2944 2945 2946 2947
				goto out;
			}
			target->cmd_sg_cnt = token;
			break;

2948 2949
		case SRP_OPT_ALLOW_EXT_SG:
			if (match_int(args, &token)) {
2950
				pr_warn("bad allow_ext_sg parameter '%s'\n", p);
2951 2952 2953 2954 2955 2956 2957 2958
				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) {
2959 2960
				pr_warn("bad max sg_tablesize parameter '%s'\n",
					p);
2961 2962 2963 2964 2965
				goto out;
			}
			target->sg_tablesize = token;
			break;

2966 2967 2968 2969 2970 2971 2972 2973
		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;

2974 2975 2976 2977 2978 2979 2980 2981 2982
		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;

2983
		default:
2984 2985
			pr_warn("unknown parameter or missing value '%s' in target creation request\n",
				p);
2986 2987 2988 2989 2990 2991 2992 2993 2994 2995
			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))
2996 2997
				pr_warn("target creation request is missing parameter '%s'\n",
					srp_opt_tokens[i].pattern);
2998

2999 3000 3001 3002 3003 3004
	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);

3005 3006 3007 3008 3009
out:
	kfree(options);
	return ret;
}

3010 3011
static ssize_t srp_create_target(struct device *dev,
				 struct device_attribute *attr,
3012 3013 3014
				 const char *buf, size_t count)
{
	struct srp_host *host =
3015
		container_of(dev, struct srp_host, dev);
3016 3017
	struct Scsi_Host *target_host;
	struct srp_target_port *target;
3018
	struct srp_rdma_ch *ch;
3019 3020
	struct srp_device *srp_dev = host->srp_dev;
	struct ib_device *ibdev = srp_dev->dev;
3021
	int ret;
3022 3023 3024 3025 3026 3027

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

3028
	target_host->transportt  = ib_srp_transport_template;
3029 3030
	target_host->max_channel = 0;
	target_host->max_id      = 1;
A
Arne Redlich 已提交
3031 3032
	target_host->max_lun     = SRP_MAX_LUN;
	target_host->max_cmd_len = sizeof ((struct srp_cmd *) (void *) 0L)->cdb;
R
Roland Dreier 已提交
3033

3034 3035
	target = host_to_target(target_host);

3036 3037 3038 3039 3040 3041
	target->io_class	= SRP_REV16A_IB_IO_CLASS;
	target->scsi_host	= target_host;
	target->srp_host	= host;
	target->lkey		= host->srp_dev->mr->lkey;
	target->rkey		= host->srp_dev->mr->rkey;
	target->cmd_sg_cnt	= cmd_sg_entries;
3042 3043
	target->sg_tablesize	= indirect_sg_entries ? : cmd_sg_entries;
	target->allow_ext_sg	= allow_ext_sg;
3044
	target->tl_retry_count	= 7;
3045
	target->queue_size	= SRP_DEFAULT_QUEUE_SIZE;
3046

3047 3048 3049 3050 3051 3052
	/*
	 * Avoid that the SCSI host can be removed by srp_remove_target()
	 * before this function returns.
	 */
	scsi_host_get(target->scsi_host);

3053 3054
	mutex_lock(&host->add_target_mutex);

3055 3056 3057 3058
	ret = srp_parse_options(buf, target);
	if (ret)
		goto err;

B
Bart Van Assche 已提交
3059 3060 3061 3062
	ret = scsi_init_shared_tag_map(target_host, target_host->can_queue);
	if (ret)
		goto err;

3063 3064
	target->req_ring_size = target->queue_size - SRP_TSK_MGMT_SQ_SIZE;

3065 3066 3067 3068 3069 3070 3071 3072 3073 3074
	if (!srp_conn_unique(target->srp_host, target)) {
		shost_printk(KERN_INFO, target->scsi_host,
			     PFX "Already connected to target port with id_ext=%016llx;ioc_guid=%016llx;initiator_ext=%016llx\n",
			     be64_to_cpu(target->id_ext),
			     be64_to_cpu(target->ioc_guid),
			     be64_to_cpu(target->initiator_ext));
		ret = -EEXIST;
		goto err;
	}

3075
	if (!srp_dev->has_fmr && !srp_dev->has_fr && !target->allow_ext_sg &&
3076
	    target->cmd_sg_cnt < target->sg_tablesize) {
3077
		pr_warn("No MR pool and no external indirect descriptors, limiting sg_tablesize to cmd_sg_cnt\n");
3078 3079 3080 3081 3082 3083
		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);
3084 3085 3086 3087
	target->max_iu_len = sizeof (struct srp_cmd) +
			     sizeof (struct srp_indirect_buf) +
			     target->cmd_sg_cnt * sizeof (struct srp_direct_buf);

3088
	INIT_WORK(&target->tl_err_work, srp_tl_err_work);
3089
	INIT_WORK(&target->remove_work, srp_remove_work);
3090
	spin_lock_init(&target->lock);
3091 3092 3093 3094 3095 3096
	ch = &target->ch;
	ch->target = target;
	ch->comp_vector = target->comp_vector;
	spin_lock_init(&ch->lock);
	INIT_LIST_HEAD(&ch->free_tx);
	ret = srp_alloc_req_data(ch);
3097 3098
	if (ret)
		goto err_free_mem;
3099

3100
	ret = ib_query_gid(ibdev, host->port, 0, &target->sgid);
3101 3102
	if (ret)
		goto err_free_mem;
3103

3104
	ret = srp_create_ch_ib(ch);
3105
	if (ret)
3106
		goto err_free_mem;
3107

3108
	ret = srp_new_cm_id(ch);
D
David Dillow 已提交
3109
	if (ret)
3110
		goto err_free_ib;
3111

3112
	ret = srp_connect_ch(ch);
3113
	if (ret) {
3114 3115
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Connection failed\n");
3116
		goto err_free_ib;
3117 3118 3119 3120 3121 3122
	}

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

3123 3124 3125 3126 3127
	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),
3128
			     be16_to_cpu(target->pkey),
3129
			     be64_to_cpu(target->service_id),
3130
			     target->sgid.raw, target->orig_dgid.raw);
3131
	}
B
Bart Van Assche 已提交
3132

3133 3134 3135 3136
	ret = count;

out:
	mutex_unlock(&host->add_target_mutex);
3137 3138 3139

	scsi_host_put(target->scsi_host);

3140
	return ret;
3141 3142 3143 3144

err_disconnect:
	srp_disconnect_target(target);

3145
err_free_ib:
3146
	srp_free_ch_ib(target, ch);
3147

3148
err_free_mem:
3149
	srp_free_req_data(target, ch);
3150

3151 3152
err:
	scsi_host_put(target_host);
3153
	goto out;
3154 3155
}

3156
static DEVICE_ATTR(add_target, S_IWUSR, NULL, srp_create_target);
3157

3158 3159
static ssize_t show_ibdev(struct device *dev, struct device_attribute *attr,
			  char *buf)
3160
{
3161
	struct srp_host *host = container_of(dev, struct srp_host, dev);
3162

3163
	return sprintf(buf, "%s\n", host->srp_dev->dev->name);
3164 3165
}

3166
static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
3167

3168 3169
static ssize_t show_port(struct device *dev, struct device_attribute *attr,
			 char *buf)
3170
{
3171
	struct srp_host *host = container_of(dev, struct srp_host, dev);
3172 3173 3174 3175

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

3176
static DEVICE_ATTR(port, S_IRUGO, show_port, NULL);
3177

3178
static struct srp_host *srp_add_port(struct srp_device *device, u8 port)
3179 3180 3181 3182 3183 3184 3185 3186
{
	struct srp_host *host;

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

	INIT_LIST_HEAD(&host->target_list);
3187
	spin_lock_init(&host->target_lock);
3188
	init_completion(&host->released);
3189
	mutex_init(&host->add_target_mutex);
3190
	host->srp_dev = device;
3191 3192
	host->port = port;

3193 3194
	host->dev.class = &srp_class;
	host->dev.parent = device->dev->dma_device;
3195
	dev_set_name(&host->dev, "srp-%s-%d", device->dev->name, port);
3196

3197
	if (device_register(&host->dev))
3198
		goto free_host;
3199
	if (device_create_file(&host->dev, &dev_attr_add_target))
3200
		goto err_class;
3201
	if (device_create_file(&host->dev, &dev_attr_ibdev))
3202
		goto err_class;
3203
	if (device_create_file(&host->dev, &dev_attr_port))
3204 3205 3206 3207 3208
		goto err_class;

	return host;

err_class:
3209
	device_unregister(&host->dev);
3210

3211
free_host:
3212 3213 3214 3215 3216 3217 3218
	kfree(host);

	return NULL;
}

static void srp_add_one(struct ib_device *device)
{
3219 3220
	struct srp_device *srp_dev;
	struct ib_device_attr *dev_attr;
3221
	struct srp_host *host;
3222 3223
	int mr_page_shift, s, e, p;
	u64 max_pages_per_mr;
3224

3225 3226
	dev_attr = kmalloc(sizeof *dev_attr, GFP_KERNEL);
	if (!dev_attr)
3227
		return;
3228

3229
	if (ib_query_device(device, dev_attr)) {
3230
		pr_warn("Query device failed for %s\n", device->name);
3231 3232 3233 3234 3235 3236 3237
		goto free_attr;
	}

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

3238 3239
	srp_dev->has_fmr = (device->alloc_fmr && device->dealloc_fmr &&
			    device->map_phys_fmr && device->unmap_fmr);
3240 3241 3242 3243 3244 3245 3246
	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));
3247

3248 3249
	/*
	 * Use the smallest page size supported by the HCA, down to a
3250 3251
	 * minimum of 4096 bytes. We're unlikely to build large sglists
	 * out of smaller entries.
3252
	 */
3253 3254 3255 3256 3257 3258 3259
	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);
3260 3261 3262 3263 3264
	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);
	}
3265 3266
	srp_dev->mr_max_size	= srp_dev->mr_page_size *
				   srp_dev->max_pages_per_mr;
3267
	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",
3268
		 device->name, mr_page_shift, dev_attr->max_mr_size,
3269
		 dev_attr->max_fast_reg_page_list_len,
3270
		 srp_dev->max_pages_per_mr, srp_dev->mr_max_size);
3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285

	INIT_LIST_HEAD(&srp_dev->dev_list);

	srp_dev->dev = device;
	srp_dev->pd  = ib_alloc_pd(device);
	if (IS_ERR(srp_dev->pd))
		goto free_dev;

	srp_dev->mr = ib_get_dma_mr(srp_dev->pd,
				    IB_ACCESS_LOCAL_WRITE |
				    IB_ACCESS_REMOTE_READ |
				    IB_ACCESS_REMOTE_WRITE);
	if (IS_ERR(srp_dev->mr))
		goto err_pd;

T
Tom Tucker 已提交
3286
	if (device->node_type == RDMA_NODE_IB_SWITCH) {
3287 3288 3289 3290 3291 3292 3293 3294
		s = 0;
		e = 0;
	} else {
		s = 1;
		e = device->phys_port_cnt;
	}

	for (p = s; p <= e; ++p) {
3295
		host = srp_add_port(srp_dev, p);
3296
		if (host)
3297
			list_add_tail(&host->list, &srp_dev->dev_list);
3298 3299
	}

3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311
	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);
3312 3313 3314 3315
}

static void srp_remove_one(struct ib_device *device)
{
3316
	struct srp_device *srp_dev;
3317
	struct srp_host *host, *tmp_host;
3318
	struct srp_target_port *target;
3319

3320
	srp_dev = ib_get_client_data(device, &srp_client);
3321 3322
	if (!srp_dev)
		return;
3323

3324
	list_for_each_entry_safe(host, tmp_host, &srp_dev->dev_list, list) {
3325
		device_unregister(&host->dev);
3326 3327 3328 3329 3330 3331 3332
		/*
		 * Wait for the sysfs entry to go away, so that no new
		 * target ports can be created.
		 */
		wait_for_completion(&host->released);

		/*
3333
		 * Remove all target ports.
3334
		 */
3335
		spin_lock(&host->target_lock);
3336 3337
		list_for_each_entry(target, &host->target_list, list)
			srp_queue_remove_work(target);
3338
		spin_unlock(&host->target_lock);
3339 3340

		/*
3341
		 * Wait for tl_err and target port removal tasks.
3342
		 */
3343
		flush_workqueue(system_long_wq);
3344
		flush_workqueue(srp_remove_wq);
3345 3346 3347 3348

		kfree(host);
	}

3349 3350 3351 3352
	ib_dereg_mr(srp_dev->mr);
	ib_dealloc_pd(srp_dev->pd);

	kfree(srp_dev);
3353 3354
}

3355
static struct srp_function_template ib_srp_transport_functions = {
3356 3357
	.has_rport_state	 = true,
	.reset_timer_if_blocked	 = true,
3358
	.reconnect_delay	 = &srp_reconnect_delay,
3359 3360 3361
	.fast_io_fail_tmo	 = &srp_fast_io_fail_tmo,
	.dev_loss_tmo		 = &srp_dev_loss_tmo,
	.reconnect		 = srp_rport_reconnect,
3362
	.rport_delete		 = srp_rport_delete,
3363
	.terminate_rport_io	 = srp_terminate_io,
3364 3365
};

3366 3367 3368 3369
static int __init srp_init_module(void)
{
	int ret;

3370
	BUILD_BUG_ON(FIELD_SIZEOF(struct ib_wc, wr_id) < sizeof(void *));
3371

3372
	if (srp_sg_tablesize) {
3373
		pr_warn("srp_sg_tablesize is deprecated, please use cmd_sg_entries\n");
3374 3375 3376 3377 3378 3379 3380 3381
		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) {
3382
		pr_warn("Clamping cmd_sg_entries to 255\n");
3383
		cmd_sg_entries = 255;
3384 3385
	}

3386 3387 3388
	if (!indirect_sg_entries)
		indirect_sg_entries = cmd_sg_entries;
	else if (indirect_sg_entries < cmd_sg_entries) {
3389 3390
		pr_warn("Bumping up indirect_sg_entries to match cmd_sg_entries (%u)\n",
			cmd_sg_entries);
3391 3392 3393
		indirect_sg_entries = cmd_sg_entries;
	}

3394
	srp_remove_wq = create_workqueue("srp_remove");
3395 3396
	if (!srp_remove_wq) {
		ret = -ENOMEM;
3397 3398 3399 3400
		goto out;
	}

	ret = -ENOMEM;
3401 3402 3403
	ib_srp_transport_template =
		srp_attach_transport(&ib_srp_transport_functions);
	if (!ib_srp_transport_template)
3404
		goto destroy_wq;
3405

3406 3407
	ret = class_register(&srp_class);
	if (ret) {
3408
		pr_err("couldn't register class infiniband_srp\n");
3409
		goto release_tr;
3410 3411
	}

3412 3413
	ib_sa_register_client(&srp_sa_client);

3414 3415
	ret = ib_register_client(&srp_client);
	if (ret) {
3416
		pr_err("couldn't register IB client\n");
3417
		goto unreg_sa;
3418 3419
	}

3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432
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;
3433 3434 3435 3436 3437
}

static void __exit srp_cleanup_module(void)
{
	ib_unregister_client(&srp_client);
3438
	ib_sa_unregister_client(&srp_sa_client);
3439
	class_unregister(&srp_class);
3440
	srp_release_transport(ib_srp_transport_template);
3441
	destroy_workqueue(srp_remove_wq);
3442 3443 3444 3445
}

module_init(srp_init_module);
module_exit(srp_cleanup_module);