ib_srp.c 88.6 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 *target_ptr);
static void srp_send_completion(struct ib_cq *cq, void *target_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;
}

D
David Dillow 已提交
286 287 288 289
static int srp_new_cm_id(struct srp_target_port *target)
{
	struct ib_cm_id *new_cm_id;

290
	new_cm_id = ib_create_cm_id(target->srp_host->srp_dev->dev,
D
David Dillow 已提交
291 292 293 294 295 296 297
				    srp_cm_handler, target);
	if (IS_ERR(new_cm_id))
		return PTR_ERR(new_cm_id);

	if (target->cm_id)
		ib_destroy_cm_id(target->cm_id);
	target->cm_id = new_cm_id;
298 299 300 301
	target->path.sgid = target->sgid;
	target->path.dgid = target->orig_dgid;
	target->path.pkey = target->pkey;
	target->path.service_id = target->service_id;
D
David Dillow 已提交
302 303 304 305

	return 0;
}

306 307 308 309 310 311 312 313 314
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;
315 316
	fmr_param.max_pages_per_fmr = dev->max_pages_per_mr;
	fmr_param.page_shift	    = ilog2(dev->mr_page_size);
317 318 319 320 321 322 323
	fmr_param.access	    = (IB_ACCESS_LOCAL_WRITE |
				       IB_ACCESS_REMOTE_WRITE |
				       IB_ACCESS_REMOTE_READ);

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

324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449
/**
 * 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);
}

450 451
static int srp_create_target_ib(struct srp_target_port *target)
{
452
	struct srp_device *dev = target->srp_host->srp_dev;
453
	struct ib_qp_init_attr *init_attr;
454 455
	struct ib_cq *recv_cq, *send_cq;
	struct ib_qp *qp;
456
	struct ib_fmr_pool *fmr_pool = NULL;
457 458
	struct srp_fr_pool *fr_pool = NULL;
	const int m = 1 + dev->use_fast_reg;
459 460 461 462 463 464
	int ret;

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

465
	recv_cq = ib_create_cq(dev->dev, srp_recv_completion, NULL, target,
466
			       target->queue_size, target->comp_vector);
467 468
	if (IS_ERR(recv_cq)) {
		ret = PTR_ERR(recv_cq);
469
		goto err;
470 471
	}

472
	send_cq = ib_create_cq(dev->dev, srp_send_completion, NULL, target,
473
			       m * target->queue_size, target->comp_vector);
474 475
	if (IS_ERR(send_cq)) {
		ret = PTR_ERR(send_cq);
476
		goto err_recv_cq;
477 478
	}

479
	ib_req_notify_cq(recv_cq, IB_CQ_NEXT_COMP);
480 481

	init_attr->event_handler       = srp_qp_event;
482
	init_attr->cap.max_send_wr     = m * target->queue_size;
483
	init_attr->cap.max_recv_wr     = target->queue_size;
484 485
	init_attr->cap.max_recv_sge    = 1;
	init_attr->cap.max_send_sge    = 1;
486
	init_attr->sq_sig_type         = IB_SIGNAL_REQ_WR;
487
	init_attr->qp_type             = IB_QPT_RC;
488 489
	init_attr->send_cq             = send_cq;
	init_attr->recv_cq             = recv_cq;
490

491
	qp = ib_create_qp(dev->pd, init_attr);
492 493
	if (IS_ERR(qp)) {
		ret = PTR_ERR(qp);
494
		goto err_send_cq;
495 496
	}

497
	ret = srp_init_qp(target, qp);
498 499
	if (ret)
		goto err_qp;
500

501 502 503 504 505 506 507 508 509 510 511 512
	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;
		}
		if (target->fr_pool)
			srp_destroy_fr_pool(target->fr_pool);
		target->fr_pool = fr_pool;
	} else if (!dev->use_fast_reg && dev->has_fmr) {
513 514 515 516 517 518 519 520 521 522 523 524
		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;
		}
		if (target->fmr_pool)
			ib_destroy_fmr_pool(target->fmr_pool);
		target->fmr_pool = fmr_pool;
	}

525 526 527 528 529 530 531 532 533 534 535
	if (target->qp)
		ib_destroy_qp(target->qp);
	if (target->recv_cq)
		ib_destroy_cq(target->recv_cq);
	if (target->send_cq)
		ib_destroy_cq(target->send_cq);

	target->qp = qp;
	target->recv_cq = recv_cq;
	target->send_cq = send_cq;

536 537 538 539
	kfree(init_attr);
	return 0;

err_qp:
540
	ib_destroy_qp(qp);
541 542

err_send_cq:
543
	ib_destroy_cq(send_cq);
544 545

err_recv_cq:
546
	ib_destroy_cq(recv_cq);
547 548

err:
549 550 551 552
	kfree(init_attr);
	return ret;
}

553 554 555 556
/*
 * Note: this function may be called without srp_alloc_iu_bufs() having been
 * invoked. Hence the target->[rt]x_ring checks.
 */
557 558
static void srp_free_target_ib(struct srp_target_port *target)
{
559
	struct srp_device *dev = target->srp_host->srp_dev;
560 561
	int i;

562 563 564 565 566
	if (target->cm_id) {
		ib_destroy_cm_id(target->cm_id);
		target->cm_id = NULL;
	}

567 568 569 570 571 572 573
	if (dev->use_fast_reg) {
		if (target->fr_pool)
			srp_destroy_fr_pool(target->fr_pool);
	} else {
		if (target->fmr_pool)
			ib_destroy_fmr_pool(target->fmr_pool);
	}
574
	ib_destroy_qp(target->qp);
575 576
	ib_destroy_cq(target->send_cq);
	ib_destroy_cq(target->recv_cq);
577

578 579 580
	target->qp = NULL;
	target->send_cq = target->recv_cq = NULL;

581 582 583 584 585 586 587 588 589 590 591 592
	if (target->rx_ring) {
		for (i = 0; i < target->queue_size; ++i)
			srp_free_iu(target->srp_host, target->rx_ring[i]);
		kfree(target->rx_ring);
		target->rx_ring = NULL;
	}
	if (target->tx_ring) {
		for (i = 0; i < target->queue_size; ++i)
			srp_free_iu(target->srp_host, target->tx_ring[i]);
		kfree(target->tx_ring);
		target->tx_ring = NULL;
	}
593 594 595 596 597 598 599 600 601 602
}

static void srp_path_rec_completion(int status,
				    struct ib_sa_path_rec *pathrec,
				    void *target_ptr)
{
	struct srp_target_port *target = target_ptr;

	target->status = status;
	if (status)
603 604
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Got failed path rec status %d\n", status);
605 606 607 608 609 610 611
	else
		target->path = *pathrec;
	complete(&target->done);
}

static int srp_lookup_path(struct srp_target_port *target)
{
612 613
	int ret;

614 615 616 617
	target->path.numb_path = 1;

	init_completion(&target->done);

618
	target->path_query_id = ib_sa_path_rec_get(&srp_sa_client,
619
						   target->srp_host->srp_dev->dev,
620 621
						   target->srp_host->port,
						   &target->path,
622
						   IB_SA_PATH_REC_SERVICE_ID	|
623 624 625 626 627 628 629 630 631 632 633
						   IB_SA_PATH_REC_DGID		|
						   IB_SA_PATH_REC_SGID		|
						   IB_SA_PATH_REC_NUMB_PATH	|
						   IB_SA_PATH_REC_PKEY,
						   SRP_PATH_REC_TIMEOUT_MS,
						   GFP_KERNEL,
						   srp_path_rec_completion,
						   target, &target->path_query);
	if (target->path_query_id < 0)
		return target->path_query_id;

634 635 636
	ret = wait_for_completion_interruptible(&target->done);
	if (ret < 0)
		return ret;
637 638

	if (target->status < 0)
639 640
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Path record query failed\n");
641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675

	return target->status;
}

static int srp_send_req(struct srp_target_port *target)
{
	struct {
		struct ib_cm_req_param param;
		struct srp_login_req   priv;
	} *req = NULL;
	int status;

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

	req->param.primary_path 	      = &target->path;
	req->param.alternate_path 	      = NULL;
	req->param.service_id 		      = target->service_id;
	req->param.qp_num 		      = target->qp->qp_num;
	req->param.qp_type 		      = target->qp->qp_type;
	req->param.private_data 	      = &req->priv;
	req->param.private_data_len 	      = sizeof req->priv;
	req->param.flow_control 	      = 1;

	get_random_bytes(&req->param.starting_psn, 4);
	req->param.starting_psn 	     &= 0xffffff;

	/*
	 * Pick some arbitrary defaults here; we could make these
	 * module parameters if anyone cared about setting them.
	 */
	req->param.responder_resources	      = 4;
	req->param.remote_cm_response_timeout = 20;
	req->param.local_cm_response_timeout  = 20;
676
	req->param.retry_count                = target->tl_retry_count;
677 678 679 680 681
	req->param.rnr_retry_count 	      = 7;
	req->param.max_cm_retries 	      = 15;

	req->priv.opcode     	= SRP_LOGIN_REQ;
	req->priv.tag        	= 0;
682
	req->priv.req_it_iu_len = cpu_to_be32(target->max_iu_len);
683 684
	req->priv.req_buf_fmt 	= cpu_to_be16(SRP_BUF_FORMAT_DIRECT |
					      SRP_BUF_FORMAT_INDIRECT);
685
	/*
R
Roland Dreier 已提交
686
	 * In the published SRP specification (draft rev. 16a), the
687 688 689 690 691 692 693 694 695
	 * 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,
696
		       &target->sgid.global.interface_id, 8);
697
		memcpy(req->priv.initiator_port_id + 8,
698
		       &target->initiator_ext, 8);
699 700 701 702
		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,
703 704
		       &target->initiator_ext, 8);
		memcpy(req->priv.initiator_port_id + 8,
705
		       &target->sgid.global.interface_id, 8);
706 707 708 709
		memcpy(req->priv.target_port_id,     &target->id_ext, 8);
		memcpy(req->priv.target_port_id + 8, &target->ioc_guid, 8);
	}

710 711
	/*
	 * Topspin/Cisco SRP targets will reject our login unless we
712 713
	 * zero out the first 8 bytes of our initiator port ID and set
	 * the second 8 bytes to the local node GUID.
714
	 */
715
	if (srp_target_is_topspin(target)) {
716 717 718 719
		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));
720
		memset(req->priv.initiator_port_id, 0, 8);
721
		memcpy(req->priv.initiator_port_id + 8,
722
		       &target->srp_host->srp_dev->dev->node_guid, 8);
723 724 725 726 727 728 729 730 731
	}

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

	kfree(req);

	return status;
}

732 733 734 735 736 737 738 739 740 741 742 743
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)
744
		queue_work(srp_remove_wq, &target->remove_work);
745 746 747 748

	return changed;
}

749 750 751 752 753 754 755 756 757 758 759 760 761 762 763
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;
}

764 765
static void srp_disconnect_target(struct srp_target_port *target)
{
766 767
	if (srp_change_conn_state(target, false)) {
		/* XXX should send SRP_I_LOGOUT request */
768

769 770 771 772
		if (ib_send_cm_dreq(target->cm_id, NULL, 0)) {
			shost_printk(KERN_DEBUG, target->scsi_host,
				     PFX "Sending CM DREQ failed\n");
		}
773
	}
774 775
}

776 777
static void srp_free_req_data(struct srp_target_port *target)
{
778 779
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = dev->dev;
780 781 782
	struct srp_request *req;
	int i;

783 784 785 786 787
	if (!target->req_ring)
		return;

	for (i = 0; i < target->req_ring_size; ++i) {
		req = &target->req_ring[i];
788 789 790 791
		if (dev->use_fast_reg)
			kfree(req->fr_list);
		else
			kfree(req->fmr_list);
792
		kfree(req->map_page);
793 794 795 796 797 798
		if (req->indirect_dma_addr) {
			ib_dma_unmap_single(ibdev, req->indirect_dma_addr,
					    target->indirect_size,
					    DMA_TO_DEVICE);
		}
		kfree(req->indirect_desc);
799
	}
800 801 802

	kfree(target->req_ring);
	target->req_ring = NULL;
803 804
}

805 806 807 808 809
static int srp_alloc_req_data(struct srp_target_port *target)
{
	struct srp_device *srp_dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = srp_dev->dev;
	struct srp_request *req;
810
	void *mr_list;
811 812 813 814 815
	dma_addr_t dma_addr;
	int i, ret = -ENOMEM;

	INIT_LIST_HEAD(&target->free_reqs);

816 817 818 819 820 821
	target->req_ring = kzalloc(target->req_ring_size *
				   sizeof(*target->req_ring), GFP_KERNEL);
	if (!target->req_ring)
		goto out;

	for (i = 0; i < target->req_ring_size; ++i) {
822
		req = &target->req_ring[i];
823 824 825 826 827 828 829 830
		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;
831
		req->map_page = kmalloc(srp_dev->max_pages_per_mr *
832
					sizeof(void *), GFP_KERNEL);
833 834
		if (!req->map_page)
			goto out;
835
		req->indirect_desc = kmalloc(target->indirect_size, GFP_KERNEL);
836
		if (!req->indirect_desc)
837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854
			goto out;

		dma_addr = ib_dma_map_single(ibdev, req->indirect_desc,
					     target->indirect_size,
					     DMA_TO_DEVICE);
		if (ib_dma_mapping_error(ibdev, dma_addr))
			goto out;

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

out:
	return ret;
}

855 856 857 858 859 860 861 862 863 864 865 866 867 868 869
/**
 * 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);
}

870 871
static void srp_remove_target(struct srp_target_port *target)
{
872 873
	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);

874
	srp_del_scsi_host_attr(target->scsi_host);
875
	srp_rport_get(target->rport);
876 877
	srp_remove_host(target->scsi_host);
	scsi_remove_host(target->scsi_host);
878
	srp_stop_rport_timers(target->rport);
879
	srp_disconnect_target(target);
880
	srp_free_target_ib(target);
881
	cancel_work_sync(&target->tl_err_work);
882
	srp_rport_put(target->rport);
883
	srp_free_req_data(target);
884 885 886 887 888

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

889 890 891
	scsi_host_put(target->scsi_host);
}

D
David Howells 已提交
892
static void srp_remove_work(struct work_struct *work)
893
{
D
David Howells 已提交
894
	struct srp_target_port *target =
895
		container_of(work, struct srp_target_port, remove_work);
896

897
	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);
898

899
	srp_remove_target(target);
900 901
}

902 903 904 905 906 907 908
static void srp_rport_delete(struct srp_rport *rport)
{
	struct srp_target_port *target = rport->lld_data;

	srp_queue_remove_work(target);
}

909 910 911 912
static int srp_connect_target(struct srp_target_port *target)
{
	int ret;

913 914
	WARN_ON_ONCE(target->connected);

915 916
	target->qp_in_error = false;

917 918 919 920 921 922 923 924 925
	ret = srp_lookup_path(target);
	if (ret)
		return ret;

	while (1) {
		init_completion(&target->done);
		ret = srp_send_req(target);
		if (ret)
			return ret;
926 927 928
		ret = wait_for_completion_interruptible(&target->done);
		if (ret < 0)
			return ret;
929 930 931 932 933 934 935 936 937

		/*
		 * The CM event handling code will set status to
		 * SRP_PORT_REDIRECT if we get a port redirect REJ
		 * back, or SRP_DLID_REDIRECT if we get a lid/qp
		 * redirect REJ back.
		 */
		switch (target->status) {
		case 0:
938
			srp_change_conn_state(target, true);
939 940 941 942 943 944 945 946 947 948 949
			return 0;

		case SRP_PORT_REDIRECT:
			ret = srp_lookup_path(target);
			if (ret)
				return ret;
			break;

		case SRP_DLID_REDIRECT:
			break;

D
David Dillow 已提交
950 951
		case SRP_STALE_CONN:
			shost_printk(KERN_ERR, target->scsi_host, PFX
952 953 954
				     "giving up on stale connection\n");
			target->status = -ECONNRESET;
			return target->status;
D
David Dillow 已提交
955

956 957 958 959 960 961
		default:
			return target->status;
		}
	}
}

962 963 964 965 966 967 968 969 970 971 972 973 974 975 976
static int srp_inv_rkey(struct srp_target_port *target, u32 rkey)
{
	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,
	};

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

977 978 979 980
static void srp_unmap_data(struct scsi_cmnd *scmnd,
			   struct srp_target_port *target,
			   struct srp_request *req)
{
981 982 983
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = dev->dev;
	int i, res;
984

985
	if (!scsi_sglist(scmnd) ||
986 987 988 989
	    (scmnd->sc_data_direction != DMA_TO_DEVICE &&
	     scmnd->sc_data_direction != DMA_FROM_DEVICE))
		return;

990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011
	if (dev->use_fast_reg) {
		struct srp_fr_desc **pfr;

		for (i = req->nmdesc, pfr = req->fr_list; i > 0; i--, pfr++) {
			res = srp_inv_rkey(target, (*pfr)->mr->rkey);
			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)
			srp_fr_pool_put(target->fr_pool, req->fr_list,
					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);
	}
1012

1013 1014
	ib_dma_unmap_sg(ibdev, scsi_sglist(scmnd), scsi_sg_count(scmnd),
			scmnd->sc_data_direction);
1015 1016
}

B
Bart Van Assche 已提交
1017 1018 1019 1020
/**
 * srp_claim_req - Take ownership of the scmnd associated with a request.
 * @target: SRP target port.
 * @req: SRP request.
1021
 * @sdev: If not NULL, only take ownership for this SCSI device.
B
Bart Van Assche 已提交
1022 1023 1024 1025 1026 1027 1028 1029
 * @scmnd: If NULL, take ownership of @req->scmnd. If not NULL, only take
 *         ownership of @req->scmnd if it equals @scmnd.
 *
 * Return value:
 * Either NULL or a pointer to the SCSI command the caller became owner of.
 */
static struct scsi_cmnd *srp_claim_req(struct srp_target_port *target,
				       struct srp_request *req,
1030
				       struct scsi_device *sdev,
B
Bart Van Assche 已提交
1031 1032 1033 1034 1035
				       struct scsi_cmnd *scmnd)
{
	unsigned long flags;

	spin_lock_irqsave(&target->lock, flags);
1036 1037 1038
	if (req->scmnd &&
	    (!sdev || req->scmnd->device == sdev) &&
	    (!scmnd || req->scmnd == scmnd)) {
B
Bart Van Assche 已提交
1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050
		scmnd = req->scmnd;
		req->scmnd = NULL;
	} else {
		scmnd = NULL;
	}
	spin_unlock_irqrestore(&target->lock, flags);

	return scmnd;
}

/**
 * srp_free_req() - Unmap data and add request to the free request list.
1051 1052 1053 1054
 * @target: SRP target port.
 * @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
1055 1056 1057 1058
 */
static void srp_free_req(struct srp_target_port *target,
			 struct srp_request *req, struct scsi_cmnd *scmnd,
			 s32 req_lim_delta)
1059
{
1060 1061
	unsigned long flags;

B
Bart Van Assche 已提交
1062 1063
	srp_unmap_data(scmnd, target, req);

1064
	spin_lock_irqsave(&target->lock, flags);
1065
	target->req_lim += req_lim_delta;
1066
	list_add_tail(&req->list, &target->free_reqs);
1067
	spin_unlock_irqrestore(&target->lock, flags);
1068 1069
}

1070
static void srp_finish_req(struct srp_target_port *target,
1071 1072
			   struct srp_request *req, struct scsi_device *sdev,
			   int result)
1073
{
1074
	struct scsi_cmnd *scmnd = srp_claim_req(target, req, sdev, NULL);
B
Bart Van Assche 已提交
1075 1076

	if (scmnd) {
1077
		srp_free_req(target, req, scmnd, 0);
1078
		scmnd->result = result;
B
Bart Van Assche 已提交
1079 1080
		scmnd->scsi_done(scmnd);
	}
1081 1082
}

1083
static void srp_terminate_io(struct srp_rport *rport)
1084
{
1085
	struct srp_target_port *target = rport->lld_data;
1086 1087
	struct Scsi_Host *shost = target->scsi_host;
	struct scsi_device *sdev;
1088 1089
	int i;

1090 1091 1092 1093 1094 1095 1096
	/*
	 * 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);

1097
	for (i = 0; i < target->req_ring_size; ++i) {
1098
		struct srp_request *req = &target->req_ring[i];
1099
		srp_finish_req(target, req, NULL, DID_TRANSPORT_FAILFAST << 16);
1100 1101
	}
}
1102

1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115
/*
 * It is up to the caller to ensure that srp_rport_reconnect() calls are
 * serialized and that no concurrent srp_queuecommand(), srp_abort(),
 * srp_reset_device() or srp_reset_host() calls will occur while this function
 * is in progress. One way to realize that is not to call this function
 * directly but to call srp_reconnect_rport() instead since that last function
 * serializes calls of this function via rport->mutex and also blocks
 * srp_queuecommand() calls before invoking this function.
 */
static int srp_rport_reconnect(struct srp_rport *rport)
{
	struct srp_target_port *target = rport->lld_data;
	int i, ret;
1116

1117
	srp_disconnect_target(target);
1118 1119 1120 1121

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

1122
	/*
1123 1124 1125
	 * 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.
1126
	 */
D
David Dillow 已提交
1127
	ret = srp_new_cm_id(target);
1128

1129
	for (i = 0; i < target->req_ring_size; ++i) {
1130
		struct srp_request *req = &target->req_ring[i];
1131
		srp_finish_req(target, req, NULL, DID_RESET << 16);
1132
	}
1133

1134 1135 1136 1137 1138 1139 1140
	/*
	 * 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.
	 */
	ret += srp_create_target_ib(target);

1141
	INIT_LIST_HEAD(&target->free_tx);
1142
	for (i = 0; i < target->queue_size; ++i)
1143
		list_add(&target->tx_ring[i]->list, &target->free_tx);
1144

1145 1146
	if (ret == 0)
		ret = srp_connect_target(target);
1147

1148 1149 1150
	if (ret == 0)
		shost_printk(KERN_INFO, target->scsi_host,
			     PFX "reconnect succeeded\n");
1151 1152 1153 1154

	return ret;
}

1155 1156
static void srp_map_desc(struct srp_map_state *state, dma_addr_t dma_addr,
			 unsigned int dma_len, u32 rkey)
1157
{
1158
	struct srp_direct_buf *desc = state->desc;
1159

1160 1161 1162
	desc->va = cpu_to_be64(dma_addr);
	desc->key = cpu_to_be32(rkey);
	desc->len = cpu_to_be32(dma_len);
1163

1164 1165 1166 1167
	state->total_len += dma_len;
	state->desc++;
	state->ndesc++;
}
1168

1169 1170 1171 1172 1173
static int srp_map_finish_fmr(struct srp_map_state *state,
			      struct srp_target_port *target)
{
	struct ib_pool_fmr *fmr;
	u64 io_addr = 0;
1174

1175
	fmr = ib_fmr_pool_map_phys(target->fmr_pool, state->pages,
1176 1177 1178
				   state->npages, io_addr);
	if (IS_ERR(fmr))
		return PTR_ERR(fmr);
1179

1180
	*state->next_fmr++ = fmr;
1181
	state->nmdesc++;
1182

1183
	srp_map_desc(state, 0, state->dma_len, fmr->fmr->rkey);
1184

1185 1186 1187
	return 0;
}

1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228
static int srp_map_finish_fr(struct srp_map_state *state,
			     struct srp_target_port *target)
{
	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;

	desc = srp_fr_pool_get(target->fr_pool);
	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);

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

1229 1230 1231 1232 1233 1234 1235 1236
static int srp_finish_mapping(struct srp_map_state *state,
			      struct srp_target_port *target)
{
	int ret = 0;

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

1237
	if (state->npages == 1 && !register_always)
1238
		srp_map_desc(state, state->base_dma_addr, state->dma_len,
1239 1240
			     target->rkey);
	else
1241 1242 1243
		ret = target->srp_host->srp_dev->use_fast_reg ?
			srp_map_finish_fr(state, target) :
			srp_map_finish_fmr(state, target);
1244 1245 1246

	if (ret == 0) {
		state->npages = 0;
1247
		state->dma_len = 0;
1248 1249 1250 1251 1252
	}

	return ret;
}

1253 1254 1255 1256 1257 1258 1259 1260
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;
}
1261

1262 1263 1264
static int srp_map_sg_entry(struct srp_map_state *state,
			    struct srp_target_port *target,
			    struct scatterlist *sg, int sg_index,
1265
			    bool use_mr)
1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276
{
	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;

1277 1278 1279 1280
	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
1281 1282 1283 1284
		 * other than the descriptor.
		 */
		srp_map_desc(state, dma_addr, dma_len, target->rkey);
		return 0;
1285
	}
1286

1287 1288 1289 1290 1291
	/*
	 * 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.
1292
	 */
1293 1294
	if ((!dev->use_fast_reg && dma_addr & ~dev->mr_page_mask) ||
	    dma_len > dev->mr_max_size) {
1295
		ret = srp_finish_mapping(state, target);
1296 1297 1298 1299 1300 1301
		if (ret)
			return ret;

		srp_map_desc(state, dma_addr, dma_len, target->rkey);
		srp_map_update_start(state, NULL, 0, 0);
		return 0;
1302 1303
	}

1304 1305 1306 1307 1308
	/*
	 * 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.
1309 1310 1311
	 */
	if (!state->unmapped_sg)
		srp_map_update_start(state, sg, sg_index, dma_addr);
1312

1313
	while (dma_len) {
1314 1315
		unsigned offset = dma_addr & ~dev->mr_page_mask;
		if (state->npages == dev->max_pages_per_mr || offset != 0) {
1316
			ret = srp_finish_mapping(state, target);
1317 1318
			if (ret)
				return ret;
1319

1320 1321 1322
			srp_map_update_start(state, sg, sg_index, dma_addr);
		}

1323
		len = min_t(unsigned int, dma_len, dev->mr_page_size - offset);
1324

1325 1326
		if (!state->npages)
			state->base_dma_addr = dma_addr;
1327
		state->pages[state->npages++] = dma_addr & dev->mr_page_mask;
1328
		state->dma_len += len;
1329 1330 1331 1332
		dma_addr += len;
		dma_len -= len;
	}

1333 1334
	/*
	 * If the last entry of the MR wasn't a full page, then we need to
1335 1336 1337 1338
	 * close it out and start a new one -- we can only merge at page
	 * boundries.
	 */
	ret = 0;
1339
	if (len != dev->mr_page_size) {
1340
		ret = srp_finish_mapping(state, target);
1341 1342 1343
		if (!ret)
			srp_map_update_start(state, NULL, 0, 0);
	}
1344 1345 1346
	return ret;
}

1347 1348 1349
static int srp_map_sg(struct srp_map_state *state,
		      struct srp_target_port *target, struct srp_request *req,
		      struct scatterlist *scat, int count)
1350 1351 1352 1353
{
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = dev->dev;
	struct scatterlist *sg;
1354 1355
	int i;
	bool use_mr;
1356 1357 1358

	state->desc	= req->indirect_desc;
	state->pages	= req->map_page;
1359 1360 1361 1362 1363 1364 1365
	if (dev->use_fast_reg) {
		state->next_fr = req->fr_list;
		use_mr = !!target->fr_pool;
	} else {
		state->next_fmr = req->fmr_list;
		use_mr = !!target->fmr_pool;
	}
1366 1367

	for_each_sg(scat, sg, count, i) {
1368 1369 1370 1371 1372
		if (srp_map_sg_entry(state, target, sg, i, use_mr)) {
			/*
			 * Memory registration failed, so backtrack to the
			 * first unmapped entry and continue on without using
			 * memory registration.
1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384
			 */
			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;
1385
			use_mr = false;
1386 1387 1388 1389
			srp_map_desc(state, dma_addr, dma_len, target->rkey);
		}
	}

1390
	if (use_mr && srp_finish_mapping(state, target))
1391 1392
		goto backtrack;

1393
	req->nmdesc = state->nmdesc;
1394 1395

	return 0;
1396 1397
}

1398 1399 1400
static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_target_port *target,
			struct srp_request *req)
{
1401
	struct scatterlist *scat;
1402
	struct srp_cmd *cmd = req->cmd->buf;
1403
	int len, nents, count;
1404 1405
	struct srp_device *dev;
	struct ib_device *ibdev;
1406 1407 1408 1409
	struct srp_map_state state;
	struct srp_indirect_buf *indirect_hdr;
	u32 table_len;
	u8 fmt;
1410

1411
	if (!scsi_sglist(scmnd) || scmnd->sc_data_direction == DMA_NONE)
1412 1413 1414 1415
		return sizeof (struct srp_cmd);

	if (scmnd->sc_data_direction != DMA_FROM_DEVICE &&
	    scmnd->sc_data_direction != DMA_TO_DEVICE) {
1416 1417 1418
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled data direction %d\n",
			     scmnd->sc_data_direction);
1419 1420 1421
		return -EINVAL;
	}

1422 1423
	nents = scsi_sg_count(scmnd);
	scat  = scsi_sglist(scmnd);
1424

1425
	dev = target->srp_host->srp_dev;
1426 1427 1428
	ibdev = dev->dev;

	count = ib_dma_map_sg(ibdev, scat, nents, scmnd->sc_data_direction);
1429 1430
	if (unlikely(count == 0))
		return -EIO;
1431 1432 1433

	fmt = SRP_DATA_DESC_DIRECT;
	len = sizeof (struct srp_cmd) +	sizeof (struct srp_direct_buf);
1434

1435
	if (count == 1 && !register_always) {
1436 1437 1438 1439 1440 1441
		/*
		 * 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.
		 */
1442
		struct srp_direct_buf *buf = (void *) cmd->add_data;
1443

1444
		buf->va  = cpu_to_be64(ib_sg_dma_address(ibdev, scat));
1445
		buf->key = cpu_to_be32(target->rkey);
1446
		buf->len = cpu_to_be32(ib_sg_dma_len(ibdev, scat));
1447

1448
		req->nmdesc = 0;
1449 1450 1451
		goto map_complete;
	}

1452 1453 1454
	/*
	 * We have more than one scatter/gather entry, so build our indirect
	 * descriptor table, trying to merge as many entries as we can.
1455 1456 1457
	 */
	indirect_hdr = (void *) cmd->add_data;

1458 1459 1460
	ib_dma_sync_single_for_cpu(ibdev, req->indirect_dma_addr,
				   target->indirect_size, DMA_TO_DEVICE);

1461
	memset(&state, 0, sizeof(state));
1462
	srp_map_sg(&state, target, req, scat, count);
1463

1464 1465 1466 1467 1468
	/* 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.
1469 1470
	 */
	if (state.ndesc == 1) {
1471 1472
		/*
		 * Memory registration collapsed the sg-list into one entry,
1473 1474 1475
		 * so use a direct descriptor.
		 */
		struct srp_direct_buf *buf = (void *) cmd->add_data;
1476

1477
		*buf = req->indirect_desc[0];
1478
		goto map_complete;
1479 1480
	}

1481 1482 1483 1484 1485 1486 1487 1488
	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);
1489 1490 1491 1492
	table_len = state.ndesc * sizeof (struct srp_direct_buf);

	fmt = SRP_DATA_DESC_INDIRECT;
	len = sizeof(struct srp_cmd) + sizeof (struct srp_indirect_buf);
1493
	len += count * sizeof (struct srp_direct_buf);
1494

1495 1496
	memcpy(indirect_hdr->desc_list, req->indirect_desc,
	       count * sizeof (struct srp_direct_buf));
1497

1498
	indirect_hdr->table_desc.va = cpu_to_be64(req->indirect_dma_addr);
1499 1500 1501 1502 1503
	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)
1504
		cmd->data_out_desc_cnt = count;
1505
	else
1506 1507 1508 1509
		cmd->data_in_desc_cnt = count;

	ib_dma_sync_single_for_device(ibdev, req->indirect_dma_addr, table_len,
				      DMA_TO_DEVICE);
1510 1511

map_complete:
1512 1513 1514 1515 1516 1517 1518 1519
	if (scmnd->sc_data_direction == DMA_TO_DEVICE)
		cmd->buf_fmt = fmt << 4;
	else
		cmd->buf_fmt = fmt;

	return len;
}

1520 1521 1522 1523 1524 1525 1526 1527
/*
 * Return an IU and possible credit to the free pool
 */
static void srp_put_tx_iu(struct srp_target_port *target, struct srp_iu *iu,
			  enum srp_iu_type iu_type)
{
	unsigned long flags;

1528
	spin_lock_irqsave(&target->lock, flags);
1529 1530 1531
	list_add(&iu->list, &target->free_tx);
	if (iu_type != SRP_IU_RSP)
		++target->req_lim;
1532
	spin_unlock_irqrestore(&target->lock, flags);
1533 1534
}

1535
/*
1536 1537
 * Must be called with target->lock held to protect req_lim and free_tx.
 * If IU is not sent, it must be returned using srp_put_tx_iu().
1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555
 *
 * Note:
 * An upper limit for the number of allocated information units for each
 * request type is:
 * - SRP_IU_CMD: SRP_CMD_SQ_SIZE, since the SCSI mid-layer never queues
 *   more than Scsi_Host.can_queue requests.
 * - SRP_IU_TSK_MGMT: SRP_TSK_MGMT_SQ_SIZE.
 * - SRP_IU_RSP: 1, since a conforming SRP target never sends more than
 *   one unanswered SRP request to an initiator.
 */
static struct srp_iu *__srp_get_tx_iu(struct srp_target_port *target,
				      enum srp_iu_type iu_type)
{
	s32 rsv = (iu_type == SRP_IU_TSK_MGMT) ? 0 : SRP_TSK_MGMT_SQ_SIZE;
	struct srp_iu *iu;

	srp_send_completion(target->send_cq, target);

1556
	if (list_empty(&target->free_tx))
1557 1558 1559
		return NULL;

	/* Initiator responses to target requests do not consume credits */
1560 1561 1562 1563 1564 1565 1566
	if (iu_type != SRP_IU_RSP) {
		if (target->req_lim <= rsv) {
			++target->zero_req_lim;
			return NULL;
		}

		--target->req_lim;
1567 1568
	}

1569
	iu = list_first_entry(&target->free_tx, struct srp_iu, list);
1570
	list_del(&iu->list);
1571 1572 1573
	return iu;
}

1574 1575
static int srp_post_send(struct srp_target_port *target,
			 struct srp_iu *iu, int len)
1576 1577 1578 1579 1580 1581
{
	struct ib_sge list;
	struct ib_send_wr wr, *bad_wr;

	list.addr   = iu->dma;
	list.length = len;
1582
	list.lkey   = target->lkey;
1583 1584

	wr.next       = NULL;
1585
	wr.wr_id      = (uintptr_t) iu;
1586 1587 1588 1589 1590
	wr.sg_list    = &list;
	wr.num_sge    = 1;
	wr.opcode     = IB_WR_SEND;
	wr.send_flags = IB_SEND_SIGNALED;

1591
	return ib_post_send(target->qp, &wr, &bad_wr);
1592 1593
}

1594
static int srp_post_recv(struct srp_target_port *target, struct srp_iu *iu)
1595 1596
{
	struct ib_recv_wr wr, *bad_wr;
1597
	struct ib_sge list;
1598 1599 1600

	list.addr   = iu->dma;
	list.length = iu->size;
1601
	list.lkey   = target->lkey;
1602 1603

	wr.next     = NULL;
1604
	wr.wr_id    = (uintptr_t) iu;
1605 1606 1607
	wr.sg_list  = &list;
	wr.num_sge  = 1;

1608
	return ib_post_recv(target->qp, &wr, &bad_wr);
1609 1610
}

1611 1612 1613 1614 1615 1616 1617
static void srp_process_rsp(struct srp_target_port *target, struct srp_rsp *rsp)
{
	struct srp_request *req;
	struct scsi_cmnd *scmnd;
	unsigned long flags;

	if (unlikely(rsp->tag & SRP_TAG_TSK_MGMT)) {
1618
		spin_lock_irqsave(&target->lock, flags);
1619
		target->req_lim += be32_to_cpu(rsp->req_lim_delta);
1620
		spin_unlock_irqrestore(&target->lock, flags);
1621

1622 1623 1624 1625
		target->tsk_mgmt_status = -1;
		if (be32_to_cpu(rsp->resp_data_len) >= 4)
			target->tsk_mgmt_status = rsp->data[3];
		complete(&target->tsk_mgmt_done);
1626
	} else {
1627
		req = &target->req_ring[rsp->tag];
1628
		scmnd = srp_claim_req(target, req, NULL, NULL);
B
Bart Van Assche 已提交
1629
		if (!scmnd) {
1630 1631 1632
			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
1633 1634 1635 1636 1637 1638 1639

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

			return;
		}
1640 1641 1642 1643 1644 1645 1646 1647 1648
		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
1649
		if (unlikely(rsp->flags & SRP_RSP_FLAG_DIUNDER))
1650
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_in_res_cnt));
B
Bart Van Assche 已提交
1651 1652 1653 1654 1655 1656
		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));
1657

B
Bart Van Assche 已提交
1658 1659 1660
		srp_free_req(target, req, scmnd,
			     be32_to_cpu(rsp->req_lim_delta));

1661 1662
		scmnd->host_scribble = NULL;
		scmnd->scsi_done(scmnd);
1663 1664 1665
	}
}

1666 1667 1668
static int srp_response_common(struct srp_target_port *target, s32 req_delta,
			       void *rsp, int len)
{
1669
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1670 1671
	unsigned long flags;
	struct srp_iu *iu;
1672
	int err;
1673

1674
	spin_lock_irqsave(&target->lock, flags);
1675 1676
	target->req_lim += req_delta;
	iu = __srp_get_tx_iu(target, SRP_IU_RSP);
1677
	spin_unlock_irqrestore(&target->lock, flags);
1678

1679 1680 1681
	if (!iu) {
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "no IU available to send response\n");
1682
		return 1;
1683 1684 1685 1686 1687 1688
	}

	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);

1689 1690
	err = srp_post_send(target, iu, len);
	if (err) {
1691 1692
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "unable to post response: %d\n", err);
1693 1694
		srp_put_tx_iu(target, iu, SRP_IU_RSP);
	}
1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729

	return err;
}

static void srp_process_cred_req(struct srp_target_port *target,
				 struct srp_cred_req *req)
{
	struct srp_cred_rsp rsp = {
		.opcode = SRP_CRED_RSP,
		.tag = req->tag,
	};
	s32 delta = be32_to_cpu(req->req_lim_delta);

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

static void srp_process_aer_req(struct srp_target_port *target,
				struct srp_aer_req *req)
{
	struct srp_aer_rsp rsp = {
		.opcode = SRP_AER_RSP,
		.tag = req->tag,
	};
	s32 delta = be32_to_cpu(req->req_lim_delta);

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

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

1730 1731
static void srp_handle_recv(struct srp_target_port *target, struct ib_wc *wc)
{
1732
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1733
	struct srp_iu *iu = (struct srp_iu *) (uintptr_t) wc->wr_id;
1734
	int res;
1735 1736
	u8 opcode;

1737 1738
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_ti_iu_len,
				   DMA_FROM_DEVICE);
1739 1740 1741 1742

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

	if (0) {
1743 1744
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "recv completion, opcode 0x%02x\n", opcode);
B
Bart Van Assche 已提交
1745 1746
		print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET, 8, 1,
			       iu->buf, wc->byte_len, true);
1747 1748 1749 1750 1751 1752 1753
	}

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

1754 1755 1756 1757 1758 1759 1760 1761
	case SRP_CRED_REQ:
		srp_process_cred_req(target, iu->buf);
		break;

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

1762 1763
	case SRP_T_LOGOUT:
		/* XXX Handle target logout */
1764 1765
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Got target logout request\n");
1766 1767 1768
		break;

	default:
1769 1770
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled SRP opcode 0x%02x\n", opcode);
1771 1772 1773
		break;
	}

1774 1775
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_ti_iu_len,
				      DMA_FROM_DEVICE);
1776

1777
	res = srp_post_recv(target, iu);
1778 1779 1780
	if (res != 0)
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Recv failed with error code %d\n", res);
1781 1782
}

1783 1784
/**
 * srp_tl_err_work() - handle a transport layer error
1785
 * @work: Work structure embedded in an SRP target port.
1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798
 *
 * 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);
}

1799 1800
static void srp_handle_qp_err(u64 wr_id, enum ib_wc_status wc_status,
			      bool send_err, struct srp_target_port *target)
1801
{
1802
	if (target->connected && !target->qp_in_error) {
1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816
		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);
		}
1817
		queue_work(system_long_wq, &target->tl_err_work);
1818
	}
1819 1820 1821
	target->qp_in_error = true;
}

1822
static void srp_recv_completion(struct ib_cq *cq, void *target_ptr)
1823 1824 1825 1826 1827 1828
{
	struct srp_target_port *target = target_ptr;
	struct ib_wc wc;

	ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
	while (ib_poll_cq(cq, 1, &wc) > 0) {
1829 1830 1831
		if (likely(wc.status == IB_WC_SUCCESS)) {
			srp_handle_recv(target, &wc);
		} else {
1832
			srp_handle_qp_err(wc.wr_id, wc.status, false, target);
1833
		}
1834 1835 1836 1837 1838 1839 1840
	}
}

static void srp_send_completion(struct ib_cq *cq, void *target_ptr)
{
	struct srp_target_port *target = target_ptr;
	struct ib_wc wc;
1841
	struct srp_iu *iu;
1842 1843

	while (ib_poll_cq(cq, 1, &wc) > 0) {
1844 1845 1846 1847
		if (likely(wc.status == IB_WC_SUCCESS)) {
			iu = (struct srp_iu *) (uintptr_t) wc.wr_id;
			list_add(&iu->list, &target->free_tx);
		} else {
1848
			srp_handle_qp_err(wc.wr_id, wc.status, true, target);
1849
		}
1850 1851 1852
	}
}

1853
static int srp_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *scmnd)
1854
{
1855
	struct srp_target_port *target = host_to_target(shost);
1856
	struct srp_rport *rport = target->rport;
1857 1858 1859
	struct srp_request *req;
	struct srp_iu *iu;
	struct srp_cmd *cmd;
1860
	struct ib_device *dev;
1861
	unsigned long flags;
1862
	int len, ret;
1863 1864 1865 1866 1867 1868 1869 1870 1871 1872
	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);
1873

1874 1875 1876
	scmnd->result = srp_chkready(target->rport);
	if (unlikely(scmnd->result))
		goto err;
1877

1878
	spin_lock_irqsave(&target->lock, flags);
1879
	iu = __srp_get_tx_iu(target, SRP_IU_CMD);
1880
	if (!iu)
1881 1882 1883 1884 1885
		goto err_unlock;

	req = list_first_entry(&target->free_reqs, struct srp_request, list);
	list_del(&req->list);
	spin_unlock_irqrestore(&target->lock, flags);
1886

1887
	dev = target->srp_host->srp_dev->dev;
1888
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_iu_len,
1889
				   DMA_TO_DEVICE);
1890

1891
	scmnd->host_scribble = (void *) req;
1892 1893 1894 1895 1896 1897

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

	cmd->opcode = SRP_CMD;
	cmd->lun    = cpu_to_be64((u64) scmnd->device->lun << 48);
1898
	cmd->tag    = req->index;
1899 1900 1901 1902 1903 1904 1905
	memcpy(cmd->cdb, scmnd->cmnd, scmnd->cmd_len);

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

	len = srp_map_data(scmnd, target, req);
	if (len < 0) {
1906
		shost_printk(KERN_ERR, target->scsi_host,
1907 1908 1909 1910
			     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
1911
		 * max_pages_per_mr sg-list elements, tell the SCSI mid-layer
1912 1913 1914 1915
		 * to reduce queue depth temporarily.
		 */
		scmnd->result = len == -ENOMEM ?
			DID_OK << 16 | QUEUE_FULL << 1 : DID_ERROR << 16;
1916
		goto err_iu;
1917 1918
	}

1919
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_iu_len,
1920
				      DMA_TO_DEVICE);
1921

1922
	if (srp_post_send(target, iu, len)) {
1923
		shost_printk(KERN_ERR, target->scsi_host, PFX "Send failed\n");
1924 1925 1926
		goto err_unmap;
	}

1927 1928
	ret = 0;

1929 1930 1931 1932
unlock_rport:
	if (in_scsi_eh)
		mutex_unlock(&rport->mutex);

1933
	return ret;
1934 1935 1936 1937

err_unmap:
	srp_unmap_data(scmnd, target, req);

1938 1939 1940
err_iu:
	srp_put_tx_iu(target, iu, SRP_IU_CMD);

1941 1942 1943 1944 1945 1946
	/*
	 * Avoid that the loops that iterate over the request ring can
	 * encounter a dangling SCSI command pointer.
	 */
	req->scmnd = NULL;

1947
	spin_lock_irqsave(&target->lock, flags);
1948
	list_add(&req->list, &target->free_reqs);
1949 1950

err_unlock:
1951
	spin_unlock_irqrestore(&target->lock, flags);
1952

1953 1954 1955 1956 1957 1958 1959
err:
	if (scmnd->result) {
		scmnd->scsi_done(scmnd);
		ret = 0;
	} else {
		ret = SCSI_MLQUEUE_HOST_BUSY;
	}
1960

1961
	goto unlock_rport;
1962 1963
}

1964 1965 1966 1967
/*
 * Note: the resources allocated in this function are freed in
 * srp_free_target_ib().
 */
1968 1969 1970 1971
static int srp_alloc_iu_bufs(struct srp_target_port *target)
{
	int i;

1972 1973 1974 1975 1976 1977 1978 1979 1980 1981
	target->rx_ring = kzalloc(target->queue_size * sizeof(*target->rx_ring),
				  GFP_KERNEL);
	if (!target->rx_ring)
		goto err_no_ring;
	target->tx_ring = kzalloc(target->queue_size * sizeof(*target->tx_ring),
				  GFP_KERNEL);
	if (!target->tx_ring)
		goto err_no_ring;

	for (i = 0; i < target->queue_size; ++i) {
1982 1983 1984 1985 1986 1987 1988
		target->rx_ring[i] = srp_alloc_iu(target->srp_host,
						  target->max_ti_iu_len,
						  GFP_KERNEL, DMA_FROM_DEVICE);
		if (!target->rx_ring[i])
			goto err;
	}

1989
	for (i = 0; i < target->queue_size; ++i) {
1990
		target->tx_ring[i] = srp_alloc_iu(target->srp_host,
1991
						  target->max_iu_len,
1992 1993 1994
						  GFP_KERNEL, DMA_TO_DEVICE);
		if (!target->tx_ring[i])
			goto err;
1995 1996

		list_add(&target->tx_ring[i]->list, &target->free_tx);
1997 1998 1999 2000 2001
	}

	return 0;

err:
2002
	for (i = 0; i < target->queue_size; ++i) {
2003 2004 2005 2006
		srp_free_iu(target->srp_host, target->rx_ring[i]);
		srp_free_iu(target->srp_host, target->tx_ring[i]);
	}

2007 2008 2009 2010 2011 2012 2013

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

2014 2015 2016
	return -ENOMEM;
}

2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043
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;
}

2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063
static void srp_cm_rep_handler(struct ib_cm_id *cm_id,
			       struct srp_login_rsp *lrsp,
			       struct srp_target_port *target)
{
	struct ib_qp_attr *qp_attr = NULL;
	int attr_mask = 0;
	int ret;
	int i;

	if (lrsp->opcode == SRP_LOGIN_RSP) {
		target->max_ti_iu_len = be32_to_cpu(lrsp->max_ti_iu_len);
		target->req_lim       = be32_to_cpu(lrsp->req_lim_delta);

		/*
		 * Reserve credits for task management so we don't
		 * bounce requests back to the SCSI mid-layer.
		 */
		target->scsi_host->can_queue
			= min(target->req_lim - SRP_TSK_MGMT_SQ_SIZE,
			      target->scsi_host->can_queue);
2064 2065 2066
		target->scsi_host->cmd_per_lun
			= min_t(int, target->scsi_host->can_queue,
				target->scsi_host->cmd_per_lun);
2067 2068 2069 2070 2071 2072 2073
	} else {
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled RSP opcode %#x\n", lrsp->opcode);
		ret = -ECONNRESET;
		goto error;
	}

2074
	if (!target->rx_ring) {
2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093
		ret = srp_alloc_iu_bufs(target);
		if (ret)
			goto error;
	}

	ret = -ENOMEM;
	qp_attr = kmalloc(sizeof *qp_attr, GFP_KERNEL);
	if (!qp_attr)
		goto error;

	qp_attr->qp_state = IB_QPS_RTR;
	ret = ib_cm_init_qp_attr(cm_id, qp_attr, &attr_mask);
	if (ret)
		goto error_free;

	ret = ib_modify_qp(target->qp, qp_attr, attr_mask);
	if (ret)
		goto error_free;

2094
	for (i = 0; i < target->queue_size; i++) {
2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105
		struct srp_iu *iu = target->rx_ring[i];
		ret = srp_post_recv(target, iu);
		if (ret)
			goto error_free;
	}

	qp_attr->qp_state = IB_QPS_RTS;
	ret = ib_cm_init_qp_attr(cm_id, qp_attr, &attr_mask);
	if (ret)
		goto error_free;

2106 2107
	target->rq_tmo_jiffies = srp_compute_rq_tmo(qp_attr, attr_mask);

2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120
	ret = ib_modify_qp(target->qp, qp_attr, attr_mask);
	if (ret)
		goto error_free;

	ret = ib_send_cm_rtu(cm_id, NULL, 0);

error_free:
	kfree(qp_attr);

error:
	target->status = ret;
}

2121 2122 2123 2124
static void srp_cm_rej_handler(struct ib_cm_id *cm_id,
			       struct ib_cm_event *event,
			       struct srp_target_port *target)
{
2125
	struct Scsi_Host *shost = target->scsi_host;
2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141
	struct ib_class_port_info *cpi;
	int opcode;

	switch (event->param.rej_rcvd.reason) {
	case IB_CM_REJ_PORT_CM_REDIRECT:
		cpi = event->param.rej_rcvd.ari;
		target->path.dlid = cpi->redirect_lid;
		target->path.pkey = cpi->redirect_pkey;
		cm_id->remote_cm_qpn = be32_to_cpu(cpi->redirect_qp) & 0x00ffffff;
		memcpy(target->path.dgid.raw, cpi->redirect_gid, 16);

		target->status = target->path.dlid ?
			SRP_DLID_REDIRECT : SRP_PORT_REDIRECT;
		break;

	case IB_CM_REJ_PORT_REDIRECT:
2142
		if (srp_target_is_topspin(target)) {
2143 2144 2145 2146 2147 2148 2149 2150
			/*
			 * Topspin/Cisco SRP gateways incorrectly send
			 * reject reason code 25 when they mean 24
			 * (port redirect).
			 */
			memcpy(target->path.dgid.raw,
			       event->param.rej_rcvd.ari, 16);

2151 2152 2153 2154
			shost_printk(KERN_DEBUG, shost,
				     PFX "Topspin/Cisco redirect to target port GID %016llx%016llx\n",
				     (unsigned long long) be64_to_cpu(target->path.dgid.global.subnet_prefix),
				     (unsigned long long) be64_to_cpu(target->path.dgid.global.interface_id));
2155 2156 2157

			target->status = SRP_PORT_REDIRECT;
		} else {
2158 2159
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_PORT_REDIRECT\n");
2160 2161 2162 2163 2164
			target->status = -ECONNRESET;
		}
		break;

	case IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID:
2165 2166
		shost_printk(KERN_WARNING, shost,
			    "  REJ reason: IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID\n");
2167 2168 2169 2170 2171 2172 2173 2174 2175 2176
		target->status = -ECONNRESET;
		break;

	case IB_CM_REJ_CONSUMER_DEFINED:
		opcode = *(u8 *) event->private_data;
		if (opcode == SRP_LOGIN_REJ) {
			struct srp_login_rej *rej = event->private_data;
			u32 reason = be32_to_cpu(rej->reason);

			if (reason == SRP_LOGIN_REJ_REQ_IT_IU_LENGTH_TOO_LARGE)
2177 2178
				shost_printk(KERN_WARNING, shost,
					     PFX "SRP_LOGIN_REJ: requested max_it_iu_len too large\n");
2179
			else
B
Bart Van Assche 已提交
2180 2181
				shost_printk(KERN_WARNING, shost, PFX
					     "SRP LOGIN from %pI6 to %pI6 REJECTED, reason 0x%08x\n",
2182 2183
					     target->sgid.raw,
					     target->orig_dgid.raw, reason);
2184
		} else
2185 2186 2187
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_CONSUMER_DEFINED,"
				     " opcode 0x%02x\n", opcode);
2188 2189 2190
		target->status = -ECONNRESET;
		break;

D
David Dillow 已提交
2191 2192 2193 2194 2195
	case IB_CM_REJ_STALE_CONN:
		shost_printk(KERN_WARNING, shost, "  REJ reason: stale connection\n");
		target->status = SRP_STALE_CONN;
		break;

2196
	default:
2197 2198
		shost_printk(KERN_WARNING, shost, "  REJ reason 0x%x\n",
			     event->param.rej_rcvd.reason);
2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209
		target->status = -ECONNRESET;
	}
}

static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event)
{
	struct srp_target_port *target = cm_id->context;
	int comp = 0;

	switch (event->event) {
	case IB_CM_REQ_ERROR:
2210 2211
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Sending CM REQ failed\n");
2212 2213 2214 2215 2216 2217
		comp = 1;
		target->status = -ECONNRESET;
		break;

	case IB_CM_REP_RECEIVED:
		comp = 1;
2218
		srp_cm_rep_handler(cm_id, event->private_data, target);
2219 2220 2221
		break;

	case IB_CM_REJ_RECEIVED:
2222
		shost_printk(KERN_DEBUG, target->scsi_host, PFX "REJ received\n");
2223 2224 2225 2226 2227
		comp = 1;

		srp_cm_rej_handler(cm_id, event, target);
		break;

2228
	case IB_CM_DREQ_RECEIVED:
2229 2230
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "DREQ received - connection closed\n");
2231
		srp_change_conn_state(target, false);
2232
		if (ib_send_cm_drep(cm_id, NULL, 0))
2233 2234
			shost_printk(KERN_ERR, target->scsi_host,
				     PFX "Sending CM DREP failed\n");
2235
		queue_work(system_long_wq, &target->tl_err_work);
2236 2237 2238
		break;

	case IB_CM_TIMEWAIT_EXIT:
2239 2240
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "connection closed\n");
2241
		comp = 1;
2242 2243 2244 2245

		target->status = 0;
		break;

2246 2247 2248 2249 2250
	case IB_CM_MRA_RECEIVED:
	case IB_CM_DREQ_ERROR:
	case IB_CM_DREP_RECEIVED:
		break;

2251
	default:
2252 2253
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled CM event %d\n", event->event);
2254 2255 2256 2257 2258 2259 2260 2261 2262
		break;
	}

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

	return 0;
}

2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282
/**
 * 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;
2283
		scsi_adjust_queue_depth(sdev, qdepth);
2284 2285 2286 2287 2288 2289 2290 2291
	} else if (reason == SCSI_QDEPTH_QFULL)
		scsi_track_queue_full(sdev, qdepth);
	else
		return -EOPNOTSUPP;

	return sdev->queue_depth;
}

2292
static int srp_send_tsk_mgmt(struct srp_target_port *target,
2293
			     u64 req_tag, unsigned int lun, u8 func)
2294
{
2295
	struct srp_rport *rport = target->rport;
2296
	struct ib_device *dev = target->srp_host->srp_dev->dev;
2297 2298 2299
	struct srp_iu *iu;
	struct srp_tsk_mgmt *tsk_mgmt;

2300 2301 2302
	if (!target->connected || target->qp_in_error)
		return -1;

2303
	init_completion(&target->tsk_mgmt_done);
2304

2305 2306 2307 2308 2309
	/*
	 * Lock the rport mutex to avoid that srp_create_target_ib() is
	 * invoked while a task management function is being sent.
	 */
	mutex_lock(&rport->mutex);
2310
	spin_lock_irq(&target->lock);
2311
	iu = __srp_get_tx_iu(target, SRP_IU_TSK_MGMT);
2312
	spin_unlock_irq(&target->lock);
2313

2314 2315 2316
	if (!iu) {
		mutex_unlock(&rport->mutex);

2317
		return -1;
2318
	}
2319

2320 2321
	ib_dma_sync_single_for_cpu(dev, iu->dma, sizeof *tsk_mgmt,
				   DMA_TO_DEVICE);
2322 2323 2324 2325
	tsk_mgmt = iu->buf;
	memset(tsk_mgmt, 0, sizeof *tsk_mgmt);

	tsk_mgmt->opcode 	= SRP_TSK_MGMT;
2326 2327
	tsk_mgmt->lun		= cpu_to_be64((u64) lun << 48);
	tsk_mgmt->tag		= req_tag | SRP_TAG_TSK_MGMT;
2328
	tsk_mgmt->tsk_mgmt_func = func;
2329
	tsk_mgmt->task_tag	= req_tag;
2330

2331 2332
	ib_dma_sync_single_for_device(dev, iu->dma, sizeof *tsk_mgmt,
				      DMA_TO_DEVICE);
2333 2334
	if (srp_post_send(target, iu, sizeof *tsk_mgmt)) {
		srp_put_tx_iu(target, iu, SRP_IU_TSK_MGMT);
2335 2336
		mutex_unlock(&rport->mutex);

2337 2338
		return -1;
	}
2339
	mutex_unlock(&rport->mutex);
2340

2341
	if (!wait_for_completion_timeout(&target->tsk_mgmt_done,
2342
					 msecs_to_jiffies(SRP_ABORT_TIMEOUT_MS)))
2343
		return -1;
2344

2345 2346 2347
	return 0;
}

2348 2349
static int srp_abort(struct scsi_cmnd *scmnd)
{
2350
	struct srp_target_port *target = host_to_target(scmnd->device->host);
2351
	struct srp_request *req = (struct srp_request *) scmnd->host_scribble;
2352
	int ret;
2353

2354
	shost_printk(KERN_ERR, target->scsi_host, "SRP abort called\n");
2355

2356
	if (!req || !srp_claim_req(target, req, NULL, scmnd))
2357
		return SUCCESS;
2358
	if (srp_send_tsk_mgmt(target, req->index, scmnd->device->lun,
2359
			      SRP_TSK_ABORT_TASK) == 0)
2360
		ret = SUCCESS;
2361
	else if (target->rport->state == SRP_RPORT_LOST)
2362
		ret = FAST_IO_FAIL;
2363 2364
	else
		ret = FAILED;
B
Bart Van Assche 已提交
2365 2366
	srp_free_req(target, req, scmnd, 0);
	scmnd->result = DID_ABORT << 16;
2367
	scmnd->scsi_done(scmnd);
2368

2369
	return ret;
2370 2371 2372 2373
}

static int srp_reset_device(struct scsi_cmnd *scmnd)
{
2374
	struct srp_target_port *target = host_to_target(scmnd->device->host);
2375
	int i;
2376

2377
	shost_printk(KERN_ERR, target->scsi_host, "SRP reset_device called\n");
2378

2379 2380
	if (srp_send_tsk_mgmt(target, SRP_TAG_NO_REQ, scmnd->device->lun,
			      SRP_TSK_LUN_RESET))
2381
		return FAILED;
2382
	if (target->tsk_mgmt_status)
2383 2384
		return FAILED;

2385
	for (i = 0; i < target->req_ring_size; ++i) {
2386
		struct srp_request *req = &target->req_ring[i];
2387
		srp_finish_req(target, req, scmnd->device, DID_RESET << 16);
2388
	}
2389 2390

	return SUCCESS;
2391 2392 2393 2394 2395 2396
}

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

2397
	shost_printk(KERN_ERR, target->scsi_host, PFX "SRP reset_host called\n");
2398

2399
	return srp_reconnect_rport(target->rport) == 0 ? SUCCESS : FAILED;
2400 2401
}

2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416
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;
}

2417 2418
static ssize_t show_id_ext(struct device *dev, struct device_attribute *attr,
			   char *buf)
2419
{
2420
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2421 2422 2423 2424 2425

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

2426 2427
static ssize_t show_ioc_guid(struct device *dev, struct device_attribute *attr,
			     char *buf)
2428
{
2429
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2430 2431 2432 2433 2434

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

2435 2436
static ssize_t show_service_id(struct device *dev,
			       struct device_attribute *attr, char *buf)
2437
{
2438
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2439 2440 2441 2442 2443

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

2444 2445
static ssize_t show_pkey(struct device *dev, struct device_attribute *attr,
			 char *buf)
2446
{
2447
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2448

2449
	return sprintf(buf, "0x%04x\n", be16_to_cpu(target->pkey));
2450 2451
}

B
Bart Van Assche 已提交
2452 2453 2454 2455 2456
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));

2457
	return sprintf(buf, "%pI6\n", target->sgid.raw);
B
Bart Van Assche 已提交
2458 2459
}

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

H
Harvey Harrison 已提交
2465
	return sprintf(buf, "%pI6\n", target->path.dgid.raw);
2466 2467
}

2468 2469
static ssize_t show_orig_dgid(struct device *dev,
			      struct device_attribute *attr, char *buf)
2470
{
2471
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2472

2473
	return sprintf(buf, "%pI6\n", target->orig_dgid.raw);
2474 2475
}

2476 2477 2478 2479 2480 2481 2482 2483
static ssize_t show_req_lim(struct device *dev,
			    struct device_attribute *attr, char *buf)
{
	struct srp_target_port *target = host_to_target(class_to_shost(dev));

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

2484 2485
static ssize_t show_zero_req_lim(struct device *dev,
				 struct device_attribute *attr, char *buf)
2486
{
2487
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2488 2489 2490 2491

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

2492 2493
static ssize_t show_local_ib_port(struct device *dev,
				  struct device_attribute *attr, char *buf)
2494
{
2495
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2496 2497 2498 2499

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

2500 2501
static ssize_t show_local_ib_device(struct device *dev,
				    struct device_attribute *attr, char *buf)
2502
{
2503
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
2504

2505
	return sprintf(buf, "%s\n", target->srp_host->srp_dev->dev->name);
2506 2507
}

2508 2509 2510 2511 2512 2513 2514 2515
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);
}

2516 2517 2518 2519 2520 2521 2522 2523
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);
}

2524 2525 2526 2527 2528 2529 2530 2531
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);
}

2532 2533 2534 2535 2536 2537 2538 2539
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");
}

2540 2541 2542 2543
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
2544
static DEVICE_ATTR(sgid,	    S_IRUGO, show_sgid,		   NULL);
2545 2546
static DEVICE_ATTR(dgid,	    S_IRUGO, show_dgid,		   NULL);
static DEVICE_ATTR(orig_dgid,	    S_IRUGO, show_orig_dgid,	   NULL);
2547
static DEVICE_ATTR(req_lim,         S_IRUGO, show_req_lim,         NULL);
2548 2549 2550
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);
2551
static DEVICE_ATTR(comp_vector,     S_IRUGO, show_comp_vector,     NULL);
2552
static DEVICE_ATTR(tl_retry_count,  S_IRUGO, show_tl_retry_count,  NULL);
2553
static DEVICE_ATTR(cmd_sg_entries,  S_IRUGO, show_cmd_sg_entries,  NULL);
2554
static DEVICE_ATTR(allow_ext_sg,    S_IRUGO, show_allow_ext_sg,    NULL);
2555 2556 2557 2558 2559 2560

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
2561
	&dev_attr_sgid,
2562 2563
	&dev_attr_dgid,
	&dev_attr_orig_dgid,
2564
	&dev_attr_req_lim,
2565 2566 2567
	&dev_attr_zero_req_lim,
	&dev_attr_local_ib_port,
	&dev_attr_local_ib_device,
2568
	&dev_attr_comp_vector,
2569
	&dev_attr_tl_retry_count,
2570
	&dev_attr_cmd_sg_entries,
2571
	&dev_attr_allow_ext_sg,
2572 2573 2574
	NULL
};

2575 2576
static struct scsi_host_template srp_template = {
	.module				= THIS_MODULE,
R
Roland Dreier 已提交
2577 2578
	.name				= "InfiniBand SRP initiator",
	.proc_name			= DRV_NAME,
2579
	.slave_configure		= srp_slave_configure,
2580 2581
	.info				= srp_target_info,
	.queuecommand			= srp_queuecommand,
2582
	.change_queue_depth             = srp_change_queue_depth,
2583
	.change_queue_type              = scsi_change_queue_type,
2584 2585 2586
	.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
2587
	.skip_settle_delay		= true,
2588
	.sg_tablesize			= SRP_DEF_SG_TABLESIZE,
2589
	.can_queue			= SRP_DEFAULT_CMD_SQ_SIZE,
2590
	.this_id			= -1,
2591
	.cmd_per_lun			= SRP_DEFAULT_CMD_SQ_SIZE,
2592 2593
	.use_clustering			= ENABLE_CLUSTERING,
	.shost_attrs			= srp_host_attrs
2594 2595
};

2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606
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;
}

2607 2608
static int srp_add_target(struct srp_host *host, struct srp_target_port *target)
{
2609 2610 2611
	struct srp_rport_identifiers ids;
	struct srp_rport *rport;

2612
	target->state = SRP_TARGET_SCANNING;
2613 2614 2615
	sprintf(target->target_name, "SRP.T10:%016llX",
		 (unsigned long long) be64_to_cpu(target->id_ext));

2616
	if (scsi_add_host(target->scsi_host, host->srp_dev->dev->dma_device))
2617 2618
		return -ENODEV;

2619 2620
	memcpy(ids.port_id, &target->id_ext, 8);
	memcpy(ids.port_id + 8, &target->ioc_guid, 8);
2621
	ids.roles = SRP_RPORT_ROLE_TARGET;
2622 2623 2624 2625 2626 2627
	rport = srp_rport_add(target->scsi_host, &ids);
	if (IS_ERR(rport)) {
		scsi_remove_host(target->scsi_host);
		return PTR_ERR(rport);
	}

2628
	rport->lld_data = target;
2629
	target->rport = rport;
2630

2631
	spin_lock(&host->target_lock);
2632
	list_add_tail(&target->list, &host->target_list);
2633
	spin_unlock(&host->target_lock);
2634 2635

	scsi_scan_target(&target->scsi_host->shost_gendev,
2636
			 0, target->scsi_id, SCAN_WILD_CARD, 0);
2637

2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654
	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:
2655 2656 2657
	return 0;
}

2658
static void srp_release_dev(struct device *dev)
2659 2660
{
	struct srp_host *host =
2661
		container_of(dev, struct srp_host, dev);
2662 2663 2664 2665 2666 2667

	complete(&host->released);
}

static struct class srp_class = {
	.name    = "infiniband_srp",
2668
	.dev_release = srp_release_dev
2669 2670
};

2671 2672
/**
 * srp_conn_unique() - check whether the connection to a target is unique
2673 2674
 * @host:   SRP host.
 * @target: SRP target port.
2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702
 */
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;
}

2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718
/*
 * 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,
2719
	SRP_OPT_MAX_CMD_PER_LUN	= 1 << 6,
2720
	SRP_OPT_IO_CLASS	= 1 << 7,
2721
	SRP_OPT_INITIATOR_EXT	= 1 << 8,
2722
	SRP_OPT_CMD_SG_ENTRIES	= 1 << 9,
2723 2724
	SRP_OPT_ALLOW_EXT_SG	= 1 << 10,
	SRP_OPT_SG_TABLESIZE	= 1 << 11,
2725
	SRP_OPT_COMP_VECTOR	= 1 << 12,
2726
	SRP_OPT_TL_RETRY_COUNT	= 1 << 13,
2727
	SRP_OPT_QUEUE_SIZE	= 1 << 14,
2728 2729 2730 2731 2732 2733 2734
	SRP_OPT_ALL		= (SRP_OPT_ID_EXT	|
				   SRP_OPT_IOC_GUID	|
				   SRP_OPT_DGID		|
				   SRP_OPT_PKEY		|
				   SRP_OPT_SERVICE_ID),
};

2735
static const match_table_t srp_opt_tokens = {
2736 2737 2738 2739 2740 2741 2742
	{ 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" 	},
2743
	{ SRP_OPT_IO_CLASS,		"io_class=%x"		},
2744
	{ SRP_OPT_INITIATOR_EXT,	"initiator_ext=%s"	},
2745
	{ SRP_OPT_CMD_SG_ENTRIES,	"cmd_sg_entries=%u"	},
2746 2747
	{ SRP_OPT_ALLOW_EXT_SG,		"allow_ext_sg=%u"	},
	{ SRP_OPT_SG_TABLESIZE,		"sg_tablesize=%u"	},
2748
	{ SRP_OPT_COMP_VECTOR,		"comp_vector=%u"	},
2749
	{ SRP_OPT_TL_RETRY_COUNT,	"tl_retry_count=%u"	},
2750
	{ SRP_OPT_QUEUE_SIZE,		"queue_size=%d"		},
2751
	{ SRP_OPT_ERR,			NULL 			}
2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779
};

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);
2780 2781 2782 2783
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2784 2785 2786 2787 2788 2789
			target->id_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_IOC_GUID:
			p = match_strdup(args);
2790 2791 2792 2793
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2794 2795 2796 2797 2798 2799
			target->ioc_guid = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_DGID:
			p = match_strdup(args);
2800 2801 2802 2803
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2804
			if (strlen(p) != 32) {
2805
				pr_warn("bad dest GID parameter '%s'\n", p);
2806
				kfree(p);
2807 2808 2809 2810
				goto out;
			}

			for (i = 0; i < 16; ++i) {
2811 2812 2813 2814 2815 2816 2817
				strlcpy(dgid, p + i * 2, sizeof(dgid));
				if (sscanf(dgid, "%hhx",
					   &target->orig_dgid.raw[i]) < 1) {
					ret = -EINVAL;
					kfree(p);
					goto out;
				}
2818
			}
2819
			kfree(p);
2820 2821 2822 2823
			break;

		case SRP_OPT_PKEY:
			if (match_hex(args, &token)) {
2824
				pr_warn("bad P_Key parameter '%s'\n", p);
2825 2826
				goto out;
			}
2827
			target->pkey = cpu_to_be16(token);
2828 2829 2830 2831
			break;

		case SRP_OPT_SERVICE_ID:
			p = match_strdup(args);
2832 2833 2834 2835
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2836 2837 2838 2839 2840 2841
			target->service_id = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_MAX_SECT:
			if (match_int(args, &token)) {
2842
				pr_warn("bad max sect parameter '%s'\n", p);
2843 2844 2845 2846 2847
				goto out;
			}
			target->scsi_host->max_sectors = token;
			break;

2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859
		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;

2860
		case SRP_OPT_MAX_CMD_PER_LUN:
2861
			if (match_int(args, &token) || token < 1) {
2862 2863
				pr_warn("bad max cmd_per_lun parameter '%s'\n",
					p);
2864 2865
				goto out;
			}
2866
			target->scsi_host->cmd_per_lun = token;
2867 2868
			break;

2869 2870
		case SRP_OPT_IO_CLASS:
			if (match_hex(args, &token)) {
2871
				pr_warn("bad IO class parameter '%s'\n", p);
2872 2873 2874 2875
				goto out;
			}
			if (token != SRP_REV10_IB_IO_CLASS &&
			    token != SRP_REV16A_IB_IO_CLASS) {
2876 2877 2878
				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);
2879 2880 2881 2882 2883
				goto out;
			}
			target->io_class = token;
			break;

2884 2885
		case SRP_OPT_INITIATOR_EXT:
			p = match_strdup(args);
2886 2887 2888 2889
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2890 2891 2892 2893
			target->initiator_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

2894 2895
		case SRP_OPT_CMD_SG_ENTRIES:
			if (match_int(args, &token) || token < 1 || token > 255) {
2896 2897
				pr_warn("bad max cmd_sg_entries parameter '%s'\n",
					p);
2898 2899 2900 2901 2902
				goto out;
			}
			target->cmd_sg_cnt = token;
			break;

2903 2904
		case SRP_OPT_ALLOW_EXT_SG:
			if (match_int(args, &token)) {
2905
				pr_warn("bad allow_ext_sg parameter '%s'\n", p);
2906 2907 2908 2909 2910 2911 2912 2913
				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) {
2914 2915
				pr_warn("bad max sg_tablesize parameter '%s'\n",
					p);
2916 2917 2918 2919 2920
				goto out;
			}
			target->sg_tablesize = token;
			break;

2921 2922 2923 2924 2925 2926 2927 2928
		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;

2929 2930 2931 2932 2933 2934 2935 2936 2937
		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;

2938
		default:
2939 2940
			pr_warn("unknown parameter or missing value '%s' in target creation request\n",
				p);
2941 2942 2943 2944 2945 2946 2947 2948 2949 2950
			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))
2951 2952
				pr_warn("target creation request is missing parameter '%s'\n",
					srp_opt_tokens[i].pattern);
2953

2954 2955 2956 2957 2958 2959
	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);

2960 2961 2962 2963 2964
out:
	kfree(options);
	return ret;
}

2965 2966
static ssize_t srp_create_target(struct device *dev,
				 struct device_attribute *attr,
2967 2968 2969
				 const char *buf, size_t count)
{
	struct srp_host *host =
2970
		container_of(dev, struct srp_host, dev);
2971 2972
	struct Scsi_Host *target_host;
	struct srp_target_port *target;
2973 2974
	struct srp_device *srp_dev = host->srp_dev;
	struct ib_device *ibdev = srp_dev->dev;
2975
	int ret;
2976 2977 2978 2979 2980 2981

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

2982
	target_host->transportt  = ib_srp_transport_template;
2983 2984
	target_host->max_channel = 0;
	target_host->max_id      = 1;
A
Arne Redlich 已提交
2985 2986
	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
2987

2988 2989
	target = host_to_target(target_host);

2990 2991 2992 2993 2994 2995
	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;
2996 2997
	target->sg_tablesize	= indirect_sg_entries ? : cmd_sg_entries;
	target->allow_ext_sg	= allow_ext_sg;
2998
	target->tl_retry_count	= 7;
2999
	target->queue_size	= SRP_DEFAULT_QUEUE_SIZE;
3000

3001 3002 3003 3004 3005 3006
	/*
	 * Avoid that the SCSI host can be removed by srp_remove_target()
	 * before this function returns.
	 */
	scsi_host_get(target->scsi_host);

3007 3008
	mutex_lock(&host->add_target_mutex);

3009 3010 3011 3012
	ret = srp_parse_options(buf, target);
	if (ret)
		goto err;

3013 3014
	target->req_ring_size = target->queue_size - SRP_TSK_MGMT_SQ_SIZE;

3015 3016 3017 3018 3019 3020 3021 3022 3023 3024
	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;
	}

3025
	if (!srp_dev->has_fmr && !srp_dev->has_fr && !target->allow_ext_sg &&
3026
	    target->cmd_sg_cnt < target->sg_tablesize) {
3027
		pr_warn("No MR pool and no external indirect descriptors, limiting sg_tablesize to cmd_sg_cnt\n");
3028 3029 3030 3031 3032 3033
		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);
3034 3035 3036 3037
	target->max_iu_len = sizeof (struct srp_cmd) +
			     sizeof (struct srp_indirect_buf) +
			     target->cmd_sg_cnt * sizeof (struct srp_direct_buf);

3038
	INIT_WORK(&target->tl_err_work, srp_tl_err_work);
3039
	INIT_WORK(&target->remove_work, srp_remove_work);
3040 3041
	spin_lock_init(&target->lock);
	INIT_LIST_HEAD(&target->free_tx);
3042 3043 3044
	ret = srp_alloc_req_data(target);
	if (ret)
		goto err_free_mem;
3045

3046
	ret = ib_query_gid(ibdev, host->port, 0, &target->sgid);
3047 3048
	if (ret)
		goto err_free_mem;
3049 3050 3051

	ret = srp_create_target_ib(target);
	if (ret)
3052
		goto err_free_mem;
3053

D
David Dillow 已提交
3054 3055
	ret = srp_new_cm_id(target);
	if (ret)
3056
		goto err_free_ib;
3057 3058 3059

	ret = srp_connect_target(target);
	if (ret) {
3060 3061
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Connection failed\n");
3062
		goto err_free_ib;
3063 3064 3065 3066 3067 3068
	}

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

3069 3070 3071 3072 3073
	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),
3074
			     be16_to_cpu(target->pkey),
3075
			     be64_to_cpu(target->service_id),
3076
			     target->sgid.raw, target->orig_dgid.raw);
3077
	}
B
Bart Van Assche 已提交
3078

3079 3080 3081 3082
	ret = count;

out:
	mutex_unlock(&host->add_target_mutex);
3083 3084 3085

	scsi_host_put(target->scsi_host);

3086
	return ret;
3087 3088 3089 3090

err_disconnect:
	srp_disconnect_target(target);

3091
err_free_ib:
3092 3093
	srp_free_target_ib(target);

3094 3095 3096
err_free_mem:
	srp_free_req_data(target);

3097 3098
err:
	scsi_host_put(target_host);
3099
	goto out;
3100 3101
}

3102
static DEVICE_ATTR(add_target, S_IWUSR, NULL, srp_create_target);
3103

3104 3105
static ssize_t show_ibdev(struct device *dev, struct device_attribute *attr,
			  char *buf)
3106
{
3107
	struct srp_host *host = container_of(dev, struct srp_host, dev);
3108

3109
	return sprintf(buf, "%s\n", host->srp_dev->dev->name);
3110 3111
}

3112
static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
3113

3114 3115
static ssize_t show_port(struct device *dev, struct device_attribute *attr,
			 char *buf)
3116
{
3117
	struct srp_host *host = container_of(dev, struct srp_host, dev);
3118 3119 3120 3121

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

3122
static DEVICE_ATTR(port, S_IRUGO, show_port, NULL);
3123

3124
static struct srp_host *srp_add_port(struct srp_device *device, u8 port)
3125 3126 3127 3128 3129 3130 3131 3132
{
	struct srp_host *host;

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

	INIT_LIST_HEAD(&host->target_list);
3133
	spin_lock_init(&host->target_lock);
3134
	init_completion(&host->released);
3135
	mutex_init(&host->add_target_mutex);
3136
	host->srp_dev = device;
3137 3138
	host->port = port;

3139 3140
	host->dev.class = &srp_class;
	host->dev.parent = device->dev->dma_device;
3141
	dev_set_name(&host->dev, "srp-%s-%d", device->dev->name, port);
3142

3143
	if (device_register(&host->dev))
3144
		goto free_host;
3145
	if (device_create_file(&host->dev, &dev_attr_add_target))
3146
		goto err_class;
3147
	if (device_create_file(&host->dev, &dev_attr_ibdev))
3148
		goto err_class;
3149
	if (device_create_file(&host->dev, &dev_attr_port))
3150 3151 3152 3153 3154
		goto err_class;

	return host;

err_class:
3155
	device_unregister(&host->dev);
3156

3157
free_host:
3158 3159 3160 3161 3162 3163 3164
	kfree(host);

	return NULL;
}

static void srp_add_one(struct ib_device *device)
{
3165 3166
	struct srp_device *srp_dev;
	struct ib_device_attr *dev_attr;
3167
	struct srp_host *host;
3168 3169
	int mr_page_shift, s, e, p;
	u64 max_pages_per_mr;
3170

3171 3172
	dev_attr = kmalloc(sizeof *dev_attr, GFP_KERNEL);
	if (!dev_attr)
3173
		return;
3174

3175
	if (ib_query_device(device, dev_attr)) {
3176
		pr_warn("Query device failed for %s\n", device->name);
3177 3178 3179 3180 3181 3182 3183
		goto free_attr;
	}

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

3184 3185
	srp_dev->has_fmr = (device->alloc_fmr && device->dealloc_fmr &&
			    device->map_phys_fmr && device->unmap_fmr);
3186 3187 3188 3189 3190 3191 3192
	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));
3193

3194 3195
	/*
	 * Use the smallest page size supported by the HCA, down to a
3196 3197
	 * minimum of 4096 bytes. We're unlikely to build large sglists
	 * out of smaller entries.
3198
	 */
3199 3200 3201 3202 3203 3204 3205
	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);
3206 3207 3208 3209 3210
	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);
	}
3211 3212
	srp_dev->mr_max_size	= srp_dev->mr_page_size *
				   srp_dev->max_pages_per_mr;
3213
	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",
3214
		 device->name, mr_page_shift, dev_attr->max_mr_size,
3215
		 dev_attr->max_fast_reg_page_list_len,
3216
		 srp_dev->max_pages_per_mr, srp_dev->mr_max_size);
3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231

	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
3232
	if (device->node_type == RDMA_NODE_IB_SWITCH) {
3233 3234 3235 3236 3237 3238 3239 3240
		s = 0;
		e = 0;
	} else {
		s = 1;
		e = device->phys_port_cnt;
	}

	for (p = s; p <= e; ++p) {
3241
		host = srp_add_port(srp_dev, p);
3242
		if (host)
3243
			list_add_tail(&host->list, &srp_dev->dev_list);
3244 3245
	}

3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257
	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);
3258 3259 3260 3261
}

static void srp_remove_one(struct ib_device *device)
{
3262
	struct srp_device *srp_dev;
3263
	struct srp_host *host, *tmp_host;
3264
	struct srp_target_port *target;
3265

3266
	srp_dev = ib_get_client_data(device, &srp_client);
3267 3268
	if (!srp_dev)
		return;
3269

3270
	list_for_each_entry_safe(host, tmp_host, &srp_dev->dev_list, list) {
3271
		device_unregister(&host->dev);
3272 3273 3274 3275 3276 3277 3278
		/*
		 * Wait for the sysfs entry to go away, so that no new
		 * target ports can be created.
		 */
		wait_for_completion(&host->released);

		/*
3279
		 * Remove all target ports.
3280
		 */
3281
		spin_lock(&host->target_lock);
3282 3283
		list_for_each_entry(target, &host->target_list, list)
			srp_queue_remove_work(target);
3284
		spin_unlock(&host->target_lock);
3285 3286

		/*
3287
		 * Wait for tl_err and target port removal tasks.
3288
		 */
3289
		flush_workqueue(system_long_wq);
3290
		flush_workqueue(srp_remove_wq);
3291 3292 3293 3294

		kfree(host);
	}

3295 3296 3297 3298
	ib_dereg_mr(srp_dev->mr);
	ib_dealloc_pd(srp_dev->pd);

	kfree(srp_dev);
3299 3300
}

3301
static struct srp_function_template ib_srp_transport_functions = {
3302 3303
	.has_rport_state	 = true,
	.reset_timer_if_blocked	 = true,
3304
	.reconnect_delay	 = &srp_reconnect_delay,
3305 3306 3307
	.fast_io_fail_tmo	 = &srp_fast_io_fail_tmo,
	.dev_loss_tmo		 = &srp_dev_loss_tmo,
	.reconnect		 = srp_rport_reconnect,
3308
	.rport_delete		 = srp_rport_delete,
3309
	.terminate_rport_io	 = srp_terminate_io,
3310 3311
};

3312 3313 3314 3315
static int __init srp_init_module(void)
{
	int ret;

3316
	BUILD_BUG_ON(FIELD_SIZEOF(struct ib_wc, wr_id) < sizeof(void *));
3317

3318
	if (srp_sg_tablesize) {
3319
		pr_warn("srp_sg_tablesize is deprecated, please use cmd_sg_entries\n");
3320 3321 3322 3323 3324 3325 3326 3327
		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) {
3328
		pr_warn("Clamping cmd_sg_entries to 255\n");
3329
		cmd_sg_entries = 255;
3330 3331
	}

3332 3333 3334
	if (!indirect_sg_entries)
		indirect_sg_entries = cmd_sg_entries;
	else if (indirect_sg_entries < cmd_sg_entries) {
3335 3336
		pr_warn("Bumping up indirect_sg_entries to match cmd_sg_entries (%u)\n",
			cmd_sg_entries);
3337 3338 3339
		indirect_sg_entries = cmd_sg_entries;
	}

3340
	srp_remove_wq = create_workqueue("srp_remove");
3341 3342
	if (!srp_remove_wq) {
		ret = -ENOMEM;
3343 3344 3345 3346
		goto out;
	}

	ret = -ENOMEM;
3347 3348 3349
	ib_srp_transport_template =
		srp_attach_transport(&ib_srp_transport_functions);
	if (!ib_srp_transport_template)
3350
		goto destroy_wq;
3351

3352 3353
	ret = class_register(&srp_class);
	if (ret) {
3354
		pr_err("couldn't register class infiniband_srp\n");
3355
		goto release_tr;
3356 3357
	}

3358 3359
	ib_sa_register_client(&srp_sa_client);

3360 3361
	ret = ib_register_client(&srp_client);
	if (ret) {
3362
		pr_err("couldn't register IB client\n");
3363
		goto unreg_sa;
3364 3365
	}

3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378
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;
3379 3380 3381 3382 3383
}

static void __exit srp_cleanup_module(void)
{
	ib_unregister_client(&srp_client);
3384
	ib_sa_unregister_client(&srp_sa_client);
3385
	class_unregister(&srp_class);
3386
	srp_release_transport(ib_srp_transport_template);
3387
	destroy_workqueue(srp_remove_wq);
3388 3389 3390 3391
}

module_init(srp_init_module);
module_exit(srp_cleanup_module);