ib_srp.c 72.0 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
/*
 * Copyright (c) 2005 Cisco Systems.  All rights reserved.
 *
 * This software is available to you under a choice of one of two
 * licenses.  You may choose to be licensed under the terms of the GNU
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - Redistributions in binary form must reproduce the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer in the documentation and/or other materials
 *        provided with the distribution.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 */

33 34
#define pr_fmt(fmt) PFX fmt

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 49

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

#include "ib_srp.h"

#define DRV_NAME	"ib_srp"
#define PFX		DRV_NAME ": "
56 57
#define DRV_VERSION	"1.0"
#define DRV_RELDATE	"July 1, 2013"
58 59 60 61 62 63

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

64 65
static unsigned int srp_sg_tablesize;
static unsigned int cmd_sg_entries;
66 67
static unsigned int indirect_sg_entries;
static bool allow_ext_sg;
68
static int topspin_workarounds = 1;
69

70 71
module_param(srp_sg_tablesize, uint, 0444);
MODULE_PARM_DESC(srp_sg_tablesize, "Deprecated name for cmd_sg_entries");
72

73 74 75
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)");
76

77 78 79 80 81 82 83 84
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)");

85 86 87 88
module_param(topspin_workarounds, int, 0444);
MODULE_PARM_DESC(topspin_workarounds,
		 "Enable workarounds for Topspin/Cisco SRP target bugs if != 0");

89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109
static struct kernel_param_ops srp_tmo_ops;

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.");

static int srp_dev_loss_tmo = 60;
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.");

110 111
static void srp_add_one(struct ib_device *device);
static void srp_remove_one(struct ib_device *device);
112 113
static void srp_recv_completion(struct ib_cq *cq, void *target_ptr);
static void srp_send_completion(struct ib_cq *cq, void *target_ptr);
114 115
static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event);

116 117
static struct scsi_transport_template *ib_srp_transport_template;

118 119 120 121 122 123
static struct ib_client srp_client = {
	.name   = "srp",
	.add    = srp_add_one,
	.remove = srp_remove_one
};

124 125
static struct ib_sa_client srp_sa_client;

126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163
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;
	}
	if (kp->arg == &srp_fast_io_fail_tmo)
		res = srp_tmo_valid(tmo, srp_dev_loss_tmo);
	else
		res = srp_tmo_valid(srp_fast_io_fail_tmo, tmo);
	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,
};

164 165 166 167 168 169 170 171 172 173
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;
}

174 175 176
static int srp_target_is_topspin(struct srp_target_port *target)
{
	static const u8 topspin_oui[3] = { 0x00, 0x05, 0xad };
177
	static const u8 cisco_oui[3]   = { 0x00, 0x1b, 0x0d };
178 179

	return topspin_workarounds &&
180 181
		(!memcmp(&target->ioc_guid, topspin_oui, sizeof topspin_oui) ||
		 !memcmp(&target->ioc_guid, cisco_oui, sizeof cisco_oui));
182 183
}

184 185 186 187 188 189 190 191 192 193 194 195 196 197
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;

198 199 200
	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))
201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220
		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;

221 222
	ib_dma_unmap_single(host->srp_dev->dev, iu->dma, iu->size,
			    iu->direction);
223 224 225 226 227 228
	kfree(iu->buf);
	kfree(iu);
}

static void srp_qp_event(struct ib_event *event, void *context)
{
229
	pr_debug("QP event %d\n", event->event);
230 231 232 233 234 235 236 237 238 239 240 241
}

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;

242 243 244 245
	ret = ib_find_pkey(target->srp_host->srp_dev->dev,
			   target->srp_host->port,
			   be16_to_cpu(target->path.pkey),
			   &attr->pkey_index);
246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264
	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
265 266 267 268
static int srp_new_cm_id(struct srp_target_port *target)
{
	struct ib_cm_id *new_cm_id;

269
	new_cm_id = ib_create_cm_id(target->srp_host->srp_dev->dev,
D
David Dillow 已提交
270 271 272 273 274 275 276 277 278 279 280
				    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;

	return 0;
}

281 282 283
static int srp_create_target_ib(struct srp_target_port *target)
{
	struct ib_qp_init_attr *init_attr;
284 285
	struct ib_cq *recv_cq, *send_cq;
	struct ib_qp *qp;
286 287 288 289 290 291
	int ret;

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

292
	recv_cq = ib_create_cq(target->srp_host->srp_dev->dev,
293 294
			       srp_recv_completion, NULL, target, SRP_RQ_SIZE,
			       target->comp_vector);
295 296
	if (IS_ERR(recv_cq)) {
		ret = PTR_ERR(recv_cq);
297
		goto err;
298 299
	}

300
	send_cq = ib_create_cq(target->srp_host->srp_dev->dev,
301 302
			       srp_send_completion, NULL, target, SRP_SQ_SIZE,
			       target->comp_vector);
303 304
	if (IS_ERR(send_cq)) {
		ret = PTR_ERR(send_cq);
305
		goto err_recv_cq;
306 307
	}

308
	ib_req_notify_cq(recv_cq, IB_CQ_NEXT_COMP);
309 310 311 312 313 314 315 316

	init_attr->event_handler       = srp_qp_event;
	init_attr->cap.max_send_wr     = SRP_SQ_SIZE;
	init_attr->cap.max_recv_wr     = SRP_RQ_SIZE;
	init_attr->cap.max_recv_sge    = 1;
	init_attr->cap.max_send_sge    = 1;
	init_attr->sq_sig_type         = IB_SIGNAL_ALL_WR;
	init_attr->qp_type             = IB_QPT_RC;
317 318
	init_attr->send_cq             = send_cq;
	init_attr->recv_cq             = recv_cq;
319

320 321 322
	qp = ib_create_qp(target->srp_host->srp_dev->pd, init_attr);
	if (IS_ERR(qp)) {
		ret = PTR_ERR(qp);
323
		goto err_send_cq;
324 325
	}

326
	ret = srp_init_qp(target, qp);
327 328
	if (ret)
		goto err_qp;
329

330 331 332 333 334 335 336 337 338 339 340
	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;

341 342 343 344
	kfree(init_attr);
	return 0;

err_qp:
345
	ib_destroy_qp(qp);
346 347

err_send_cq:
348
	ib_destroy_cq(send_cq);
349 350

err_recv_cq:
351
	ib_destroy_cq(recv_cq);
352 353

err:
354 355 356 357 358 359 360 361 362
	kfree(init_attr);
	return ret;
}

static void srp_free_target_ib(struct srp_target_port *target)
{
	int i;

	ib_destroy_qp(target->qp);
363 364
	ib_destroy_cq(target->send_cq);
	ib_destroy_cq(target->recv_cq);
365

366 367 368
	target->qp = NULL;
	target->send_cq = target->recv_cq = NULL;

369 370
	for (i = 0; i < SRP_RQ_SIZE; ++i)
		srp_free_iu(target->srp_host, target->rx_ring[i]);
371
	for (i = 0; i < SRP_SQ_SIZE; ++i)
372 373 374 375 376 377 378 379 380 381 382
		srp_free_iu(target->srp_host, target->tx_ring[i]);
}

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)
383 384
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Got failed path rec status %d\n", status);
385 386 387 388 389 390 391 392 393 394 395
	else
		target->path = *pathrec;
	complete(&target->done);
}

static int srp_lookup_path(struct srp_target_port *target)
{
	target->path.numb_path = 1;

	init_completion(&target->done);

396
	target->path_query_id = ib_sa_path_rec_get(&srp_sa_client,
397
						   target->srp_host->srp_dev->dev,
398 399
						   target->srp_host->port,
						   &target->path,
400
						   IB_SA_PATH_REC_SERVICE_ID	|
401 402 403 404 405 406 407 408 409 410 411 412 413 414
						   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;

	wait_for_completion(&target->done);

	if (target->status < 0)
415 416
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Path record query failed\n");
417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451

	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;
452
	req->param.retry_count                = target->tl_retry_count;
453 454 455 456 457
	req->param.rnr_retry_count 	      = 7;
	req->param.max_cm_retries 	      = 15;

	req->priv.opcode     	= SRP_LOGIN_REQ;
	req->priv.tag        	= 0;
458
	req->priv.req_it_iu_len = cpu_to_be32(target->max_iu_len);
459 460
	req->priv.req_buf_fmt 	= cpu_to_be16(SRP_BUF_FORMAT_DIRECT |
					      SRP_BUF_FORMAT_INDIRECT);
461
	/*
R
Roland Dreier 已提交
462
	 * In the published SRP specification (draft rev. 16a), the
463 464 465 466 467 468 469 470 471
	 * 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,
472
		       &target->path.sgid.global.interface_id, 8);
473
		memcpy(req->priv.initiator_port_id + 8,
474
		       &target->initiator_ext, 8);
475 476 477 478
		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,
479 480 481
		       &target->initiator_ext, 8);
		memcpy(req->priv.initiator_port_id + 8,
		       &target->path.sgid.global.interface_id, 8);
482 483 484 485
		memcpy(req->priv.target_port_id,     &target->id_ext, 8);
		memcpy(req->priv.target_port_id + 8, &target->ioc_guid, 8);
	}

486 487
	/*
	 * Topspin/Cisco SRP targets will reject our login unless we
488 489
	 * zero out the first 8 bytes of our initiator port ID and set
	 * the second 8 bytes to the local node GUID.
490
	 */
491
	if (srp_target_is_topspin(target)) {
492 493 494 495
		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));
496
		memset(req->priv.initiator_port_id, 0, 8);
497
		memcpy(req->priv.initiator_port_id + 8,
498
		       &target->srp_host->srp_dev->dev->node_guid, 8);
499 500 501 502 503 504 505 506 507
	}

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

	kfree(req);

	return status;
}

508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524
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)
		queue_work(system_long_wq, &target->remove_work);

	return changed;
}

525 526 527 528 529 530 531 532 533 534 535 536 537 538 539
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;
}

540 541
static void srp_disconnect_target(struct srp_target_port *target)
{
542 543
	if (srp_change_conn_state(target, false)) {
		/* XXX should send SRP_I_LOGOUT request */
544

545 546 547 548
		if (ib_send_cm_dreq(target->cm_id, NULL, 0)) {
			shost_printk(KERN_DEBUG, target->scsi_host,
				     PFX "Sending CM DREQ failed\n");
		}
549
	}
550 551
}

552 553
static void srp_free_req_data(struct srp_target_port *target)
{
554
	struct ib_device *ibdev = target->srp_host->srp_dev->dev;
555 556 557 558 559 560
	struct srp_request *req;
	int i;

	for (i = 0, req = target->req_ring; i < SRP_CMD_SQ_SIZE; ++i, ++req) {
		kfree(req->fmr_list);
		kfree(req->map_page);
561 562 563 564 565 566
		if (req->indirect_dma_addr) {
			ib_dma_unmap_single(ibdev, req->indirect_dma_addr,
					    target->indirect_size,
					    DMA_TO_DEVICE);
		}
		kfree(req->indirect_desc);
567 568 569
	}
}

570 571 572 573 574 575 576 577 578 579 580 581 582 583 584
/**
 * 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);
}

585 586
static void srp_remove_target(struct srp_target_port *target)
{
587 588
	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);

589
	srp_del_scsi_host_attr(target->scsi_host);
590
	srp_rport_get(target->rport);
591 592
	srp_remove_host(target->scsi_host);
	scsi_remove_host(target->scsi_host);
593
	srp_disconnect_target(target);
594 595
	ib_destroy_cm_id(target->cm_id);
	srp_free_target_ib(target);
596
	srp_rport_put(target->rport);
597 598 599 600
	srp_free_req_data(target);
	scsi_host_put(target->scsi_host);
}

D
David Howells 已提交
601
static void srp_remove_work(struct work_struct *work)
602
{
D
David Howells 已提交
603
	struct srp_target_port *target =
604
		container_of(work, struct srp_target_port, remove_work);
605

606
	WARN_ON_ONCE(target->state != SRP_TARGET_REMOVED);
607

608 609
	srp_remove_target(target);

610
	spin_lock(&target->srp_host->target_lock);
611
	list_del(&target->list);
612
	spin_unlock(&target->srp_host->target_lock);
613 614
}

615 616 617 618 619 620 621
static void srp_rport_delete(struct srp_rport *rport)
{
	struct srp_target_port *target = rport->lld_data;

	srp_queue_remove_work(target);
}

622 623
static int srp_connect_target(struct srp_target_port *target)
{
D
David Dillow 已提交
624
	int retries = 3;
625 626
	int ret;

627 628
	WARN_ON_ONCE(target->connected);

629 630
	target->qp_in_error = false;

631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649
	ret = srp_lookup_path(target);
	if (ret)
		return ret;

	while (1) {
		init_completion(&target->done);
		ret = srp_send_req(target);
		if (ret)
			return ret;
		wait_for_completion(&target->done);

		/*
		 * 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:
650
			srp_change_conn_state(target, true);
651 652 653 654 655 656 657 658 659 660 661
			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
662 663 664 665 666 667 668 669 670 671 672 673 674 675 676
		case SRP_STALE_CONN:
			/* Our current CM id was stale, and is now in timewait.
			 * Try to reconnect with a new one.
			 */
			if (!retries-- || srp_new_cm_id(target)) {
				shost_printk(KERN_ERR, target->scsi_host, PFX
					     "giving up on stale connection\n");
				target->status = -ECONNRESET;
				return target->status;
			}

			shost_printk(KERN_ERR, target->scsi_host, PFX
				     "retrying stale connection\n");
			break;

677 678 679 680 681 682
		default:
			return target->status;
		}
	}
}

683 684 685 686
static void srp_unmap_data(struct scsi_cmnd *scmnd,
			   struct srp_target_port *target,
			   struct srp_request *req)
{
687 688 689
	struct ib_device *ibdev = target->srp_host->srp_dev->dev;
	struct ib_pool_fmr **pfmr;

690
	if (!scsi_sglist(scmnd) ||
691 692 693 694
	    (scmnd->sc_data_direction != DMA_TO_DEVICE &&
	     scmnd->sc_data_direction != DMA_FROM_DEVICE))
		return;

695 696 697
	pfmr = req->fmr_list;
	while (req->nfmr--)
		ib_fmr_pool_unmap(*pfmr++);
698

699 700
	ib_dma_unmap_sg(ibdev, scsi_sglist(scmnd), scsi_sg_count(scmnd),
			scmnd->sc_data_direction);
701 702
}

B
Bart Van Assche 已提交
703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738
/**
 * srp_claim_req - Take ownership of the scmnd associated with a request.
 * @target: SRP target port.
 * @req: SRP request.
 * @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,
				       struct scsi_cmnd *scmnd)
{
	unsigned long flags;

	spin_lock_irqsave(&target->lock, flags);
	if (!scmnd) {
		scmnd = req->scmnd;
		req->scmnd = NULL;
	} else if (req->scmnd == 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.
 */
static void srp_free_req(struct srp_target_port *target,
			 struct srp_request *req, struct scsi_cmnd *scmnd,
			 s32 req_lim_delta)
739
{
740 741
	unsigned long flags;

B
Bart Van Assche 已提交
742 743
	srp_unmap_data(scmnd, target, req);

744
	spin_lock_irqsave(&target->lock, flags);
745
	target->req_lim += req_lim_delta;
746
	list_add_tail(&req->list, &target->free_reqs);
747
	spin_unlock_irqrestore(&target->lock, flags);
748 749
}

750 751
static void srp_finish_req(struct srp_target_port *target,
			   struct srp_request *req, int result)
752
{
B
Bart Van Assche 已提交
753 754 755
	struct scsi_cmnd *scmnd = srp_claim_req(target, req, NULL);

	if (scmnd) {
756
		srp_free_req(target, req, scmnd, 0);
757
		scmnd->result = result;
B
Bart Van Assche 已提交
758 759
		scmnd->scsi_done(scmnd);
	}
760 761
}

762
static void srp_terminate_io(struct srp_rport *rport)
763
{
764 765 766 767 768 769 770 771
	struct srp_target_port *target = rport->lld_data;
	int i;

	for (i = 0; i < SRP_CMD_SQ_SIZE; ++i) {
		struct srp_request *req = &target->req_ring[i];
		srp_finish_req(target, req, DID_TRANSPORT_FAILFAST << 16);
	}
}
772

773 774 775 776 777 778 779 780 781 782 783 784 785
/*
 * 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;
786

787 788
	srp_disconnect_target(target);
	/*
789 790 791
	 * 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.
792
	 */
D
David Dillow 已提交
793
	ret = srp_new_cm_id(target);
794 795 796 797 798 799 800 801 802
	/*
	 * Whether or not creating a new CM ID succeeded, create a new
	 * QP. This guarantees that all completion callback function
	 * invocations have finished before request resetting starts.
	 */
	if (ret == 0)
		ret = srp_create_target_ib(target);
	else
		srp_create_target_ib(target);
803

804 805
	for (i = 0; i < SRP_CMD_SQ_SIZE; ++i) {
		struct srp_request *req = &target->req_ring[i];
806
		srp_finish_req(target, req, DID_RESET << 16);
807
	}
808

809
	INIT_LIST_HEAD(&target->free_tx);
810
	for (i = 0; i < SRP_SQ_SIZE; ++i)
811
		list_add(&target->tx_ring[i]->list, &target->free_tx);
812

813 814
	if (ret == 0)
		ret = srp_connect_target(target);
815

816 817 818
	if (ret == 0)
		shost_printk(KERN_INFO, target->scsi_host,
			     PFX "reconnect succeeded\n");
819 820 821 822

	return ret;
}

823 824
static void srp_map_desc(struct srp_map_state *state, dma_addr_t dma_addr,
			 unsigned int dma_len, u32 rkey)
825
{
826
	struct srp_direct_buf *desc = state->desc;
827

828 829 830
	desc->va = cpu_to_be64(dma_addr);
	desc->key = cpu_to_be32(rkey);
	desc->len = cpu_to_be32(dma_len);
831

832 833 834 835
	state->total_len += dma_len;
	state->desc++;
	state->ndesc++;
}
836

837 838 839 840 841 842
static int srp_map_finish_fmr(struct srp_map_state *state,
			      struct srp_target_port *target)
{
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_pool_fmr *fmr;
	u64 io_addr = 0;
843

844 845
	if (!state->npages)
		return 0;
846

847 848 849 850 851
	if (state->npages == 1) {
		srp_map_desc(state, state->base_dma_addr, state->fmr_len,
			     target->rkey);
		state->npages = state->fmr_len = 0;
		return 0;
852 853
	}

854 855 856 857
	fmr = ib_fmr_pool_map_phys(dev->fmr_pool, state->pages,
				   state->npages, io_addr);
	if (IS_ERR(fmr))
		return PTR_ERR(fmr);
858

859 860
	*state->next_fmr++ = fmr;
	state->nfmr++;
861

862 863 864 865 866 867 868 869 870 871 872 873 874
	srp_map_desc(state, 0, state->fmr_len, fmr->fmr->rkey);
	state->npages = state->fmr_len = 0;
	return 0;
}

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;
}
875

876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897
static int srp_map_sg_entry(struct srp_map_state *state,
			    struct srp_target_port *target,
			    struct scatterlist *sg, int sg_index,
			    int use_fmr)
{
	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;

	if (use_fmr == SRP_MAP_NO_FMR) {
		/* Once we're in direct map mode for a request, we don't
		 * go back to FMR mode, so no need to update anything
		 * other than the descriptor.
		 */
		srp_map_desc(state, dma_addr, dma_len, target->rkey);
		return 0;
898
	}
899

900 901 902 903 904 905 906 907 908 909 910 911 912 913
	/* If we start at an offset into the FMR page, don't merge into
	 * the current FMR. Finish it out, and use the kernel's MR for this
	 * sg entry. This is to avoid potential bugs on some SRP targets
	 * that were never quite defined, but went away when the initiator
	 * avoided using FMR on such page fragments.
	 */
	if (dma_addr & ~dev->fmr_page_mask || dma_len > dev->fmr_max_size) {
		ret = srp_map_finish_fmr(state, target);
		if (ret)
			return ret;

		srp_map_desc(state, dma_addr, dma_len, target->rkey);
		srp_map_update_start(state, NULL, 0, 0);
		return 0;
914 915
	}

916 917 918 919 920 921 922
	/* If this is the first sg to go into the FMR, 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.
	 */
	if (!state->unmapped_sg)
		srp_map_update_start(state, sg, sg_index, dma_addr);
923

924 925 926 927 928
	while (dma_len) {
		if (state->npages == SRP_FMR_SIZE) {
			ret = srp_map_finish_fmr(state, target);
			if (ret)
				return ret;
929

930 931 932 933
			srp_map_update_start(state, sg, sg_index, dma_addr);
		}

		len = min_t(unsigned int, dma_len, dev->fmr_page_size);
934

935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952
		if (!state->npages)
			state->base_dma_addr = dma_addr;
		state->pages[state->npages++] = dma_addr;
		state->fmr_len += len;
		dma_addr += len;
		dma_len -= len;
	}

	/* If the last entry of the FMR wasn't a full page, then we need to
	 * close it out and start a new one -- we can only merge at page
	 * boundries.
	 */
	ret = 0;
	if (len != dev->fmr_page_size) {
		ret = srp_map_finish_fmr(state, target);
		if (!ret)
			srp_map_update_start(state, NULL, 0, 0);
	}
953 954 955
	return ret;
}

956 957 958
static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_target_port *target,
			struct srp_request *req)
{
959
	struct scatterlist *scat, *sg;
960
	struct srp_cmd *cmd = req->cmd->buf;
961
	int i, len, nents, count, use_fmr;
962 963
	struct srp_device *dev;
	struct ib_device *ibdev;
964 965 966 967
	struct srp_map_state state;
	struct srp_indirect_buf *indirect_hdr;
	u32 table_len;
	u8 fmt;
968

969
	if (!scsi_sglist(scmnd) || scmnd->sc_data_direction == DMA_NONE)
970 971 972 973
		return sizeof (struct srp_cmd);

	if (scmnd->sc_data_direction != DMA_FROM_DEVICE &&
	    scmnd->sc_data_direction != DMA_TO_DEVICE) {
974 975 976
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled data direction %d\n",
			     scmnd->sc_data_direction);
977 978 979
		return -EINVAL;
	}

980 981
	nents = scsi_sg_count(scmnd);
	scat  = scsi_sglist(scmnd);
982

983
	dev = target->srp_host->srp_dev;
984 985 986
	ibdev = dev->dev;

	count = ib_dma_map_sg(ibdev, scat, nents, scmnd->sc_data_direction);
987 988
	if (unlikely(count == 0))
		return -EIO;
989 990 991

	fmt = SRP_DATA_DESC_DIRECT;
	len = sizeof (struct srp_cmd) +	sizeof (struct srp_direct_buf);
992

993
	if (count == 1) {
994 995 996 997 998 999
		/*
		 * 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.
		 */
1000
		struct srp_direct_buf *buf = (void *) cmd->add_data;
1001

1002
		buf->va  = cpu_to_be64(ib_sg_dma_address(ibdev, scat));
1003
		buf->key = cpu_to_be32(target->rkey);
1004
		buf->len = cpu_to_be32(ib_sg_dma_len(ibdev, scat));
1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015

		req->nfmr = 0;
		goto map_complete;
	}

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

1016 1017 1018
	ib_dma_sync_single_for_cpu(ibdev, req->indirect_dma_addr,
				   target->indirect_size, DMA_TO_DEVICE);

1019
	memset(&state, 0, sizeof(state));
1020
	state.desc	= req->indirect_desc;
1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043
	state.pages	= req->map_page;
	state.next_fmr	= req->fmr_list;

	use_fmr = dev->fmr_pool ? SRP_MAP_ALLOW_FMR : SRP_MAP_NO_FMR;

	for_each_sg(scat, sg, count, i) {
		if (srp_map_sg_entry(&state, target, sg, i, use_fmr)) {
			/* FMR mapping failed, so backtrack to the first
			 * unmapped entry and continue on without using FMR.
			 */
			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;
			use_fmr = SRP_MAP_NO_FMR;
			srp_map_desc(&state, dma_addr, dma_len, target->rkey);
1044
		}
1045
	}
1046

1047 1048
	if (use_fmr == SRP_MAP_ALLOW_FMR && srp_map_finish_fmr(&state, target))
		goto backtrack;
1049

1050 1051 1052 1053 1054
	/* 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.
1055 1056 1057 1058 1059 1060 1061
	 */
	req->nfmr = state.nfmr;
	if (state.ndesc == 1) {
		/* FMR mapping was able to collapse this to one entry,
		 * so use a direct descriptor.
		 */
		struct srp_direct_buf *buf = (void *) cmd->add_data;
1062

1063
		*buf = req->indirect_desc[0];
1064
		goto map_complete;
1065 1066
	}

1067 1068 1069 1070 1071 1072 1073 1074
	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);
1075 1076 1077 1078
	table_len = state.ndesc * sizeof (struct srp_direct_buf);

	fmt = SRP_DATA_DESC_INDIRECT;
	len = sizeof(struct srp_cmd) + sizeof (struct srp_indirect_buf);
1079
	len += count * sizeof (struct srp_direct_buf);
1080

1081 1082
	memcpy(indirect_hdr->desc_list, req->indirect_desc,
	       count * sizeof (struct srp_direct_buf));
1083

1084
	indirect_hdr->table_desc.va = cpu_to_be64(req->indirect_dma_addr);
1085 1086 1087 1088 1089
	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)
1090
		cmd->data_out_desc_cnt = count;
1091
	else
1092 1093 1094 1095
		cmd->data_in_desc_cnt = count;

	ib_dma_sync_single_for_device(ibdev, req->indirect_dma_addr, table_len,
				      DMA_TO_DEVICE);
1096 1097

map_complete:
1098 1099 1100 1101 1102 1103 1104 1105
	if (scmnd->sc_data_direction == DMA_TO_DEVICE)
		cmd->buf_fmt = fmt << 4;
	else
		cmd->buf_fmt = fmt;

	return len;
}

1106 1107 1108 1109 1110 1111 1112 1113
/*
 * 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;

1114
	spin_lock_irqsave(&target->lock, flags);
1115 1116 1117
	list_add(&iu->list, &target->free_tx);
	if (iu_type != SRP_IU_RSP)
		++target->req_lim;
1118
	spin_unlock_irqrestore(&target->lock, flags);
1119 1120
}

1121
/*
1122 1123
 * 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().
1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141
 *
 * 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);

1142
	if (list_empty(&target->free_tx))
1143 1144 1145
		return NULL;

	/* Initiator responses to target requests do not consume credits */
1146 1147 1148 1149 1150 1151 1152
	if (iu_type != SRP_IU_RSP) {
		if (target->req_lim <= rsv) {
			++target->zero_req_lim;
			return NULL;
		}

		--target->req_lim;
1153 1154
	}

1155
	iu = list_first_entry(&target->free_tx, struct srp_iu, list);
1156
	list_del(&iu->list);
1157 1158 1159
	return iu;
}

1160 1161
static int srp_post_send(struct srp_target_port *target,
			 struct srp_iu *iu, int len)
1162 1163 1164 1165 1166 1167
{
	struct ib_sge list;
	struct ib_send_wr wr, *bad_wr;

	list.addr   = iu->dma;
	list.length = len;
1168
	list.lkey   = target->lkey;
1169 1170

	wr.next       = NULL;
1171
	wr.wr_id      = (uintptr_t) iu;
1172 1173 1174 1175 1176
	wr.sg_list    = &list;
	wr.num_sge    = 1;
	wr.opcode     = IB_WR_SEND;
	wr.send_flags = IB_SEND_SIGNALED;

1177
	return ib_post_send(target->qp, &wr, &bad_wr);
1178 1179
}

1180
static int srp_post_recv(struct srp_target_port *target, struct srp_iu *iu)
1181 1182
{
	struct ib_recv_wr wr, *bad_wr;
1183
	struct ib_sge list;
1184 1185 1186

	list.addr   = iu->dma;
	list.length = iu->size;
1187
	list.lkey   = target->lkey;
1188 1189

	wr.next     = NULL;
1190
	wr.wr_id    = (uintptr_t) iu;
1191 1192 1193
	wr.sg_list  = &list;
	wr.num_sge  = 1;

1194
	return ib_post_recv(target->qp, &wr, &bad_wr);
1195 1196
}

1197 1198 1199 1200 1201 1202 1203
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)) {
1204
		spin_lock_irqsave(&target->lock, flags);
1205
		target->req_lim += be32_to_cpu(rsp->req_lim_delta);
1206
		spin_unlock_irqrestore(&target->lock, flags);
1207

1208 1209 1210 1211
		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);
1212
	} else {
1213
		req = &target->req_ring[rsp->tag];
B
Bart Van Assche 已提交
1214 1215
		scmnd = srp_claim_req(target, req, NULL);
		if (!scmnd) {
1216 1217 1218
			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
1219 1220 1221 1222 1223 1224 1225

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

			return;
		}
1226 1227 1228 1229 1230 1231 1232 1233 1234 1235
		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));
		}

		if (rsp->flags & (SRP_RSP_FLAG_DOOVER | SRP_RSP_FLAG_DOUNDER))
1236
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_out_res_cnt));
1237
		else if (rsp->flags & (SRP_RSP_FLAG_DIOVER | SRP_RSP_FLAG_DIUNDER))
1238
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_in_res_cnt));
1239

B
Bart Van Assche 已提交
1240 1241 1242
		srp_free_req(target, req, scmnd,
			     be32_to_cpu(rsp->req_lim_delta));

1243 1244
		scmnd->host_scribble = NULL;
		scmnd->scsi_done(scmnd);
1245 1246 1247
	}
}

1248 1249 1250
static int srp_response_common(struct srp_target_port *target, s32 req_delta,
			       void *rsp, int len)
{
1251
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1252 1253
	unsigned long flags;
	struct srp_iu *iu;
1254
	int err;
1255

1256
	spin_lock_irqsave(&target->lock, flags);
1257 1258
	target->req_lim += req_delta;
	iu = __srp_get_tx_iu(target, SRP_IU_RSP);
1259
	spin_unlock_irqrestore(&target->lock, flags);
1260

1261 1262 1263
	if (!iu) {
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "no IU available to send response\n");
1264
		return 1;
1265 1266 1267 1268 1269 1270
	}

	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);

1271 1272
	err = srp_post_send(target, iu, len);
	if (err) {
1273 1274
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "unable to post response: %d\n", err);
1275 1276
		srp_put_tx_iu(target, iu, SRP_IU_RSP);
	}
1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311

	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");
}

1312 1313
static void srp_handle_recv(struct srp_target_port *target, struct ib_wc *wc)
{
1314
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1315
	struct srp_iu *iu = (struct srp_iu *) (uintptr_t) wc->wr_id;
1316
	int res;
1317 1318
	u8 opcode;

1319 1320
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_ti_iu_len,
				   DMA_FROM_DEVICE);
1321 1322 1323 1324

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

	if (0) {
1325 1326
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "recv completion, opcode 0x%02x\n", opcode);
B
Bart Van Assche 已提交
1327 1328
		print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET, 8, 1,
			       iu->buf, wc->byte_len, true);
1329 1330 1331 1332 1333 1334 1335
	}

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

1336 1337 1338 1339 1340 1341 1342 1343
	case SRP_CRED_REQ:
		srp_process_cred_req(target, iu->buf);
		break;

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

1344 1345
	case SRP_T_LOGOUT:
		/* XXX Handle target logout */
1346 1347
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Got target logout request\n");
1348 1349 1350
		break;

	default:
1351 1352
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled SRP opcode 0x%02x\n", opcode);
1353 1354 1355
		break;
	}

1356 1357
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_ti_iu_len,
				      DMA_FROM_DEVICE);
1358

1359
	res = srp_post_recv(target, iu);
1360 1361 1362
	if (res != 0)
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Recv failed with error code %d\n", res);
1363 1364
}

1365 1366 1367 1368
static void srp_handle_qp_err(enum ib_wc_status wc_status,
			      enum ib_wc_opcode wc_opcode,
			      struct srp_target_port *target)
{
1369
	if (target->connected && !target->qp_in_error) {
1370 1371 1372 1373 1374
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "failed %s status %d\n",
			     wc_opcode & IB_WC_RECV ? "receive" : "send",
			     wc_status);
	}
1375 1376 1377
	target->qp_in_error = true;
}

1378
static void srp_recv_completion(struct ib_cq *cq, void *target_ptr)
1379 1380 1381 1382 1383 1384
{
	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) {
1385 1386 1387 1388
		if (likely(wc.status == IB_WC_SUCCESS)) {
			srp_handle_recv(target, &wc);
		} else {
			srp_handle_qp_err(wc.status, wc.opcode, target);
1389
		}
1390 1391 1392 1393 1394 1395 1396
	}
}

static void srp_send_completion(struct ib_cq *cq, void *target_ptr)
{
	struct srp_target_port *target = target_ptr;
	struct ib_wc wc;
1397
	struct srp_iu *iu;
1398 1399

	while (ib_poll_cq(cq, 1, &wc) > 0) {
1400 1401 1402 1403 1404
		if (likely(wc.status == IB_WC_SUCCESS)) {
			iu = (struct srp_iu *) (uintptr_t) wc.wr_id;
			list_add(&iu->list, &target->free_tx);
		} else {
			srp_handle_qp_err(wc.status, wc.opcode, target);
1405
		}
1406 1407 1408
	}
}

1409
static int srp_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *scmnd)
1410
{
1411
	struct srp_target_port *target = host_to_target(shost);
1412 1413 1414
	struct srp_request *req;
	struct srp_iu *iu;
	struct srp_cmd *cmd;
1415
	struct ib_device *dev;
1416
	unsigned long flags;
1417
	int len, result;
1418

1419 1420 1421
	result = srp_chkready(target->rport);
	if (unlikely(result)) {
		scmnd->result = result;
1422 1423 1424 1425
		scmnd->scsi_done(scmnd);
		return 0;
	}

1426
	spin_lock_irqsave(&target->lock, flags);
1427
	iu = __srp_get_tx_iu(target, SRP_IU_CMD);
1428
	if (!iu)
1429 1430 1431 1432 1433
		goto err_unlock;

	req = list_first_entry(&target->free_reqs, struct srp_request, list);
	list_del(&req->list);
	spin_unlock_irqrestore(&target->lock, flags);
1434

1435
	dev = target->srp_host->srp_dev->dev;
1436
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_iu_len,
1437
				   DMA_TO_DEVICE);
1438 1439

	scmnd->result        = 0;
1440
	scmnd->host_scribble = (void *) req;
1441 1442 1443 1444 1445 1446

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

	cmd->opcode = SRP_CMD;
	cmd->lun    = cpu_to_be64((u64) scmnd->device->lun << 48);
1447
	cmd->tag    = req->index;
1448 1449 1450 1451 1452 1453 1454
	memcpy(cmd->cdb, scmnd->cmnd, scmnd->cmd_len);

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

	len = srp_map_data(scmnd, target, req);
	if (len < 0) {
1455 1456
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Failed to map data\n");
1457
		goto err_iu;
1458 1459
	}

1460
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_iu_len,
1461
				      DMA_TO_DEVICE);
1462

1463
	if (srp_post_send(target, iu, len)) {
1464
		shost_printk(KERN_ERR, target->scsi_host, PFX "Send failed\n");
1465 1466 1467 1468 1469 1470 1471 1472
		goto err_unmap;
	}

	return 0;

err_unmap:
	srp_unmap_data(scmnd, target, req);

1473 1474 1475
err_iu:
	srp_put_tx_iu(target, iu, SRP_IU_CMD);

1476
	spin_lock_irqsave(&target->lock, flags);
1477
	list_add(&req->list, &target->free_reqs);
1478 1479

err_unlock:
1480
	spin_unlock_irqrestore(&target->lock, flags);
1481

1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496
	return SCSI_MLQUEUE_HOST_BUSY;
}

static int srp_alloc_iu_bufs(struct srp_target_port *target)
{
	int i;

	for (i = 0; i < SRP_RQ_SIZE; ++i) {
		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;
	}

1497
	for (i = 0; i < SRP_SQ_SIZE; ++i) {
1498
		target->tx_ring[i] = srp_alloc_iu(target->srp_host,
1499
						  target->max_iu_len,
1500 1501 1502
						  GFP_KERNEL, DMA_TO_DEVICE);
		if (!target->tx_ring[i])
			goto err;
1503 1504

		list_add(&target->tx_ring[i]->list, &target->free_tx);
1505 1506 1507 1508 1509 1510 1511 1512 1513 1514
	}

	return 0;

err:
	for (i = 0; i < SRP_RQ_SIZE; ++i) {
		srp_free_iu(target->srp_host, target->rx_ring[i]);
		target->rx_ring[i] = NULL;
	}

1515
	for (i = 0; i < SRP_SQ_SIZE; ++i) {
1516 1517 1518 1519 1520 1521 1522
		srp_free_iu(target->srp_host, target->tx_ring[i]);
		target->tx_ring[i] = NULL;
	}

	return -ENOMEM;
}

1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549
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;
}

1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608
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);
	} else {
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled RSP opcode %#x\n", lrsp->opcode);
		ret = -ECONNRESET;
		goto error;
	}

	if (!target->rx_ring[0]) {
		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;

	for (i = 0; i < SRP_RQ_SIZE; i++) {
		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;

1609 1610
	target->rq_tmo_jiffies = srp_compute_rq_tmo(qp_attr, attr_mask);

1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623
	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;
}

1624 1625 1626 1627
static void srp_cm_rej_handler(struct ib_cm_id *cm_id,
			       struct ib_cm_event *event,
			       struct srp_target_port *target)
{
1628
	struct Scsi_Host *shost = target->scsi_host;
1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644
	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:
1645
		if (srp_target_is_topspin(target)) {
1646 1647 1648 1649 1650 1651 1652 1653
			/*
			 * 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);

1654 1655 1656 1657
			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));
1658 1659 1660

			target->status = SRP_PORT_REDIRECT;
		} else {
1661 1662
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_PORT_REDIRECT\n");
1663 1664 1665 1666 1667
			target->status = -ECONNRESET;
		}
		break;

	case IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID:
1668 1669
		shost_printk(KERN_WARNING, shost,
			    "  REJ reason: IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID\n");
1670 1671 1672 1673 1674 1675 1676 1677 1678 1679
		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)
1680 1681
				shost_printk(KERN_WARNING, shost,
					     PFX "SRP_LOGIN_REJ: requested max_it_iu_len too large\n");
1682
			else
1683 1684
				shost_printk(KERN_WARNING, shost,
					    PFX "SRP LOGIN REJECTED, reason 0x%08x\n", reason);
1685
		} else
1686 1687 1688
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_CONSUMER_DEFINED,"
				     " opcode 0x%02x\n", opcode);
1689 1690 1691
		target->status = -ECONNRESET;
		break;

D
David Dillow 已提交
1692 1693 1694 1695 1696
	case IB_CM_REJ_STALE_CONN:
		shost_printk(KERN_WARNING, shost, "  REJ reason: stale connection\n");
		target->status = SRP_STALE_CONN;
		break;

1697
	default:
1698 1699
		shost_printk(KERN_WARNING, shost, "  REJ reason 0x%x\n",
			     event->param.rej_rcvd.reason);
1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710
		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:
1711 1712
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Sending CM REQ failed\n");
1713 1714 1715 1716 1717 1718
		comp = 1;
		target->status = -ECONNRESET;
		break;

	case IB_CM_REP_RECEIVED:
		comp = 1;
1719
		srp_cm_rep_handler(cm_id, event->private_data, target);
1720 1721 1722
		break;

	case IB_CM_REJ_RECEIVED:
1723
		shost_printk(KERN_DEBUG, target->scsi_host, PFX "REJ received\n");
1724 1725 1726 1727 1728
		comp = 1;

		srp_cm_rej_handler(cm_id, event, target);
		break;

1729
	case IB_CM_DREQ_RECEIVED:
1730 1731
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "DREQ received - connection closed\n");
1732
		srp_change_conn_state(target, false);
1733
		if (ib_send_cm_drep(cm_id, NULL, 0))
1734 1735
			shost_printk(KERN_ERR, target->scsi_host,
				     PFX "Sending CM DREP failed\n");
1736 1737 1738
		break;

	case IB_CM_TIMEWAIT_EXIT:
1739 1740
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "connection closed\n");
1741 1742 1743 1744

		target->status = 0;
		break;

1745 1746 1747 1748 1749
	case IB_CM_MRA_RECEIVED:
	case IB_CM_DREQ_ERROR:
	case IB_CM_DREP_RECEIVED:
		break;

1750
	default:
1751 1752
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled CM event %d\n", event->event);
1753 1754 1755 1756 1757 1758 1759 1760 1761
		break;
	}

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

	return 0;
}

1762
static int srp_send_tsk_mgmt(struct srp_target_port *target,
1763
			     u64 req_tag, unsigned int lun, u8 func)
1764
{
1765
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1766 1767 1768
	struct srp_iu *iu;
	struct srp_tsk_mgmt *tsk_mgmt;

1769 1770 1771
	if (!target->connected || target->qp_in_error)
		return -1;

1772
	init_completion(&target->tsk_mgmt_done);
1773

1774
	spin_lock_irq(&target->lock);
1775
	iu = __srp_get_tx_iu(target, SRP_IU_TSK_MGMT);
1776
	spin_unlock_irq(&target->lock);
1777

1778
	if (!iu)
1779
		return -1;
1780

1781 1782
	ib_dma_sync_single_for_cpu(dev, iu->dma, sizeof *tsk_mgmt,
				   DMA_TO_DEVICE);
1783 1784 1785 1786
	tsk_mgmt = iu->buf;
	memset(tsk_mgmt, 0, sizeof *tsk_mgmt);

	tsk_mgmt->opcode 	= SRP_TSK_MGMT;
1787 1788
	tsk_mgmt->lun		= cpu_to_be64((u64) lun << 48);
	tsk_mgmt->tag		= req_tag | SRP_TAG_TSK_MGMT;
1789
	tsk_mgmt->tsk_mgmt_func = func;
1790
	tsk_mgmt->task_tag	= req_tag;
1791

1792 1793
	ib_dma_sync_single_for_device(dev, iu->dma, sizeof *tsk_mgmt,
				      DMA_TO_DEVICE);
1794 1795 1796 1797
	if (srp_post_send(target, iu, sizeof *tsk_mgmt)) {
		srp_put_tx_iu(target, iu, SRP_IU_TSK_MGMT);
		return -1;
	}
1798

1799
	if (!wait_for_completion_timeout(&target->tsk_mgmt_done,
1800
					 msecs_to_jiffies(SRP_ABORT_TIMEOUT_MS)))
1801
		return -1;
1802

1803 1804 1805
	return 0;
}

1806 1807
static int srp_abort(struct scsi_cmnd *scmnd)
{
1808
	struct srp_target_port *target = host_to_target(scmnd->device->host);
1809
	struct srp_request *req = (struct srp_request *) scmnd->host_scribble;
1810
	int ret;
1811

1812
	shost_printk(KERN_ERR, target->scsi_host, "SRP abort called\n");
1813

1814
	if (!req || !srp_claim_req(target, req, scmnd))
1815
		return FAILED;
1816
	if (srp_send_tsk_mgmt(target, req->index, scmnd->device->lun,
1817
			      SRP_TSK_ABORT_TASK) == 0)
1818
		ret = SUCCESS;
1819
	else if (target->rport->state == SRP_RPORT_LOST)
1820
		ret = FAST_IO_FAIL;
1821 1822
	else
		ret = FAILED;
B
Bart Van Assche 已提交
1823 1824
	srp_free_req(target, req, scmnd, 0);
	scmnd->result = DID_ABORT << 16;
1825
	scmnd->scsi_done(scmnd);
1826

1827
	return ret;
1828 1829 1830 1831
}

static int srp_reset_device(struct scsi_cmnd *scmnd)
{
1832
	struct srp_target_port *target = host_to_target(scmnd->device->host);
1833
	int i;
1834

1835
	shost_printk(KERN_ERR, target->scsi_host, "SRP reset_device called\n");
1836

1837 1838
	if (srp_send_tsk_mgmt(target, SRP_TAG_NO_REQ, scmnd->device->lun,
			      SRP_TSK_LUN_RESET))
1839
		return FAILED;
1840
	if (target->tsk_mgmt_status)
1841 1842
		return FAILED;

1843 1844
	for (i = 0; i < SRP_CMD_SQ_SIZE; ++i) {
		struct srp_request *req = &target->req_ring[i];
1845
		if (req->scmnd && req->scmnd->device == scmnd->device)
1846
			srp_finish_req(target, req, DID_RESET << 16);
1847
	}
1848 1849

	return SUCCESS;
1850 1851 1852 1853 1854 1855
}

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

1856
	shost_printk(KERN_ERR, target->scsi_host, PFX "SRP reset_host called\n");
1857

1858
	return srp_reconnect_rport(target->rport) == 0 ? SUCCESS : FAILED;
1859 1860
}

1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875
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;
}

1876 1877
static ssize_t show_id_ext(struct device *dev, struct device_attribute *attr,
			   char *buf)
1878
{
1879
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1880 1881 1882 1883 1884

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

1885 1886
static ssize_t show_ioc_guid(struct device *dev, struct device_attribute *attr,
			     char *buf)
1887
{
1888
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1889 1890 1891 1892 1893

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

1894 1895
static ssize_t show_service_id(struct device *dev,
			       struct device_attribute *attr, char *buf)
1896
{
1897
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1898 1899 1900 1901 1902

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

1903 1904
static ssize_t show_pkey(struct device *dev, struct device_attribute *attr,
			 char *buf)
1905
{
1906
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1907 1908 1909 1910

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

1911 1912
static ssize_t show_dgid(struct device *dev, struct device_attribute *attr,
			 char *buf)
1913
{
1914
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1915

H
Harvey Harrison 已提交
1916
	return sprintf(buf, "%pI6\n", target->path.dgid.raw);
1917 1918
}

1919 1920
static ssize_t show_orig_dgid(struct device *dev,
			      struct device_attribute *attr, char *buf)
1921
{
1922
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1923

H
Harvey Harrison 已提交
1924
	return sprintf(buf, "%pI6\n", target->orig_dgid);
1925 1926
}

1927 1928 1929 1930 1931 1932 1933 1934
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);
}

1935 1936
static ssize_t show_zero_req_lim(struct device *dev,
				 struct device_attribute *attr, char *buf)
1937
{
1938
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1939 1940 1941 1942

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

1943 1944
static ssize_t show_local_ib_port(struct device *dev,
				  struct device_attribute *attr, char *buf)
1945
{
1946
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1947 1948 1949 1950

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

1951 1952
static ssize_t show_local_ib_device(struct device *dev,
				    struct device_attribute *attr, char *buf)
1953
{
1954
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1955

1956
	return sprintf(buf, "%s\n", target->srp_host->srp_dev->dev->name);
1957 1958
}

1959 1960 1961 1962 1963 1964 1965 1966
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);
}

1967 1968 1969 1970 1971 1972 1973 1974
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);
}

1975 1976 1977 1978 1979 1980 1981 1982
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);
}

1983 1984 1985 1986 1987 1988 1989 1990
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");
}

1991 1992 1993 1994 1995 1996
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);
static DEVICE_ATTR(dgid,	    S_IRUGO, show_dgid,		   NULL);
static DEVICE_ATTR(orig_dgid,	    S_IRUGO, show_orig_dgid,	   NULL);
1997
static DEVICE_ATTR(req_lim,         S_IRUGO, show_req_lim,         NULL);
1998 1999 2000
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);
2001
static DEVICE_ATTR(comp_vector,     S_IRUGO, show_comp_vector,     NULL);
2002
static DEVICE_ATTR(tl_retry_count,  S_IRUGO, show_tl_retry_count,  NULL);
2003
static DEVICE_ATTR(cmd_sg_entries,  S_IRUGO, show_cmd_sg_entries,  NULL);
2004
static DEVICE_ATTR(allow_ext_sg,    S_IRUGO, show_allow_ext_sg,    NULL);
2005 2006 2007 2008 2009 2010 2011 2012

static struct device_attribute *srp_host_attrs[] = {
	&dev_attr_id_ext,
	&dev_attr_ioc_guid,
	&dev_attr_service_id,
	&dev_attr_pkey,
	&dev_attr_dgid,
	&dev_attr_orig_dgid,
2013
	&dev_attr_req_lim,
2014 2015 2016
	&dev_attr_zero_req_lim,
	&dev_attr_local_ib_port,
	&dev_attr_local_ib_device,
2017
	&dev_attr_comp_vector,
2018
	&dev_attr_tl_retry_count,
2019
	&dev_attr_cmd_sg_entries,
2020
	&dev_attr_allow_ext_sg,
2021 2022 2023
	NULL
};

2024 2025
static struct scsi_host_template srp_template = {
	.module				= THIS_MODULE,
R
Roland Dreier 已提交
2026 2027
	.name				= "InfiniBand SRP initiator",
	.proc_name			= DRV_NAME,
2028
	.slave_configure		= srp_slave_configure,
2029 2030 2031 2032 2033
	.info				= srp_target_info,
	.queuecommand			= srp_queuecommand,
	.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
2034
	.skip_settle_delay		= true,
2035
	.sg_tablesize			= SRP_DEF_SG_TABLESIZE,
2036
	.can_queue			= SRP_CMD_SQ_SIZE,
2037
	.this_id			= -1,
2038
	.cmd_per_lun			= SRP_CMD_SQ_SIZE,
2039 2040
	.use_clustering			= ENABLE_CLUSTERING,
	.shost_attrs			= srp_host_attrs
2041 2042 2043 2044
};

static int srp_add_target(struct srp_host *host, struct srp_target_port *target)
{
2045 2046 2047
	struct srp_rport_identifiers ids;
	struct srp_rport *rport;

2048 2049 2050
	sprintf(target->target_name, "SRP.T10:%016llX",
		 (unsigned long long) be64_to_cpu(target->id_ext));

2051
	if (scsi_add_host(target->scsi_host, host->srp_dev->dev->dma_device))
2052 2053
		return -ENODEV;

2054 2055
	memcpy(ids.port_id, &target->id_ext, 8);
	memcpy(ids.port_id + 8, &target->ioc_guid, 8);
2056
	ids.roles = SRP_RPORT_ROLE_TARGET;
2057 2058 2059 2060 2061 2062
	rport = srp_rport_add(target->scsi_host, &ids);
	if (IS_ERR(rport)) {
		scsi_remove_host(target->scsi_host);
		return PTR_ERR(rport);
	}

2063
	rport->lld_data = target;
2064
	target->rport = rport;
2065

2066
	spin_lock(&host->target_lock);
2067
	list_add_tail(&target->list, &host->target_list);
2068
	spin_unlock(&host->target_lock);
2069 2070 2071 2072

	target->state = SRP_TARGET_LIVE;

	scsi_scan_target(&target->scsi_host->shost_gendev,
2073
			 0, target->scsi_id, SCAN_WILD_CARD, 0);
2074 2075 2076 2077

	return 0;
}

2078
static void srp_release_dev(struct device *dev)
2079 2080
{
	struct srp_host *host =
2081
		container_of(dev, struct srp_host, dev);
2082 2083 2084 2085 2086 2087

	complete(&host->released);
}

static struct class srp_class = {
	.name    = "infiniband_srp",
2088
	.dev_release = srp_release_dev
2089 2090
};

2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120
/**
 * srp_conn_unique() - check whether the connection to a target is unique
 */
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;
}

2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136
/*
 * 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,
2137
	SRP_OPT_MAX_CMD_PER_LUN	= 1 << 6,
2138
	SRP_OPT_IO_CLASS	= 1 << 7,
2139
	SRP_OPT_INITIATOR_EXT	= 1 << 8,
2140
	SRP_OPT_CMD_SG_ENTRIES	= 1 << 9,
2141 2142
	SRP_OPT_ALLOW_EXT_SG	= 1 << 10,
	SRP_OPT_SG_TABLESIZE	= 1 << 11,
2143
	SRP_OPT_COMP_VECTOR	= 1 << 12,
2144
	SRP_OPT_TL_RETRY_COUNT	= 1 << 13,
2145 2146 2147 2148 2149 2150 2151
	SRP_OPT_ALL		= (SRP_OPT_ID_EXT	|
				   SRP_OPT_IOC_GUID	|
				   SRP_OPT_DGID		|
				   SRP_OPT_PKEY		|
				   SRP_OPT_SERVICE_ID),
};

2152
static const match_table_t srp_opt_tokens = {
2153 2154 2155 2156 2157 2158 2159
	{ 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" 	},
2160
	{ SRP_OPT_IO_CLASS,		"io_class=%x"		},
2161
	{ SRP_OPT_INITIATOR_EXT,	"initiator_ext=%s"	},
2162
	{ SRP_OPT_CMD_SG_ENTRIES,	"cmd_sg_entries=%u"	},
2163 2164
	{ SRP_OPT_ALLOW_EXT_SG,		"allow_ext_sg=%u"	},
	{ SRP_OPT_SG_TABLESIZE,		"sg_tablesize=%u"	},
2165
	{ SRP_OPT_COMP_VECTOR,		"comp_vector=%u"	},
2166
	{ SRP_OPT_TL_RETRY_COUNT,	"tl_retry_count=%u"	},
2167
	{ SRP_OPT_ERR,			NULL 			}
2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195
};

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);
2196 2197 2198 2199
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2200 2201 2202 2203 2204 2205
			target->id_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_IOC_GUID:
			p = match_strdup(args);
2206 2207 2208 2209
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2210 2211 2212 2213 2214 2215
			target->ioc_guid = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_DGID:
			p = match_strdup(args);
2216 2217 2218 2219
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2220
			if (strlen(p) != 32) {
2221
				pr_warn("bad dest GID parameter '%s'\n", p);
2222
				kfree(p);
2223 2224 2225 2226 2227 2228 2229
				goto out;
			}

			for (i = 0; i < 16; ++i) {
				strlcpy(dgid, p + i * 2, 3);
				target->path.dgid.raw[i] = simple_strtoul(dgid, NULL, 16);
			}
2230
			kfree(p);
2231
			memcpy(target->orig_dgid, target->path.dgid.raw, 16);
2232 2233 2234 2235
			break;

		case SRP_OPT_PKEY:
			if (match_hex(args, &token)) {
2236
				pr_warn("bad P_Key parameter '%s'\n", p);
2237 2238 2239 2240 2241 2242 2243
				goto out;
			}
			target->path.pkey = cpu_to_be16(token);
			break;

		case SRP_OPT_SERVICE_ID:
			p = match_strdup(args);
2244 2245 2246 2247
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2248
			target->service_id = cpu_to_be64(simple_strtoull(p, NULL, 16));
2249
			target->path.service_id = target->service_id;
2250 2251 2252 2253 2254
			kfree(p);
			break;

		case SRP_OPT_MAX_SECT:
			if (match_int(args, &token)) {
2255
				pr_warn("bad max sect parameter '%s'\n", p);
2256 2257 2258 2259 2260
				goto out;
			}
			target->scsi_host->max_sectors = token;
			break;

2261 2262
		case SRP_OPT_MAX_CMD_PER_LUN:
			if (match_int(args, &token)) {
2263 2264
				pr_warn("bad max cmd_per_lun parameter '%s'\n",
					p);
2265 2266
				goto out;
			}
2267
			target->scsi_host->cmd_per_lun = min(token, SRP_CMD_SQ_SIZE);
2268 2269
			break;

2270 2271
		case SRP_OPT_IO_CLASS:
			if (match_hex(args, &token)) {
2272
				pr_warn("bad IO class parameter '%s'\n", p);
2273 2274 2275 2276
				goto out;
			}
			if (token != SRP_REV10_IB_IO_CLASS &&
			    token != SRP_REV16A_IB_IO_CLASS) {
2277 2278 2279
				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);
2280 2281 2282 2283 2284
				goto out;
			}
			target->io_class = token;
			break;

2285 2286
		case SRP_OPT_INITIATOR_EXT:
			p = match_strdup(args);
2287 2288 2289 2290
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2291 2292 2293 2294
			target->initiator_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

2295 2296
		case SRP_OPT_CMD_SG_ENTRIES:
			if (match_int(args, &token) || token < 1 || token > 255) {
2297 2298
				pr_warn("bad max cmd_sg_entries parameter '%s'\n",
					p);
2299 2300 2301 2302 2303
				goto out;
			}
			target->cmd_sg_cnt = token;
			break;

2304 2305
		case SRP_OPT_ALLOW_EXT_SG:
			if (match_int(args, &token)) {
2306
				pr_warn("bad allow_ext_sg parameter '%s'\n", p);
2307 2308 2309 2310 2311 2312 2313 2314
				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) {
2315 2316
				pr_warn("bad max sg_tablesize parameter '%s'\n",
					p);
2317 2318 2319 2320 2321
				goto out;
			}
			target->sg_tablesize = token;
			break;

2322 2323 2324 2325 2326 2327 2328 2329
		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;

2330 2331 2332 2333 2334 2335 2336 2337 2338
		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;

2339
		default:
2340 2341
			pr_warn("unknown parameter or missing value '%s' in target creation request\n",
				p);
2342 2343 2344 2345 2346 2347 2348 2349 2350 2351
			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))
2352 2353
				pr_warn("target creation request is missing parameter '%s'\n",
					srp_opt_tokens[i].pattern);
2354 2355 2356 2357 2358 2359

out:
	kfree(options);
	return ret;
}

2360 2361
static ssize_t srp_create_target(struct device *dev,
				 struct device_attribute *attr,
2362 2363 2364
				 const char *buf, size_t count)
{
	struct srp_host *host =
2365
		container_of(dev, struct srp_host, dev);
2366 2367
	struct Scsi_Host *target_host;
	struct srp_target_port *target;
2368 2369
	struct ib_device *ibdev = host->srp_dev->dev;
	dma_addr_t dma_addr;
2370
	int i, ret;
2371 2372 2373 2374 2375 2376

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

2377
	target_host->transportt  = ib_srp_transport_template;
2378 2379
	target_host->max_channel = 0;
	target_host->max_id      = 1;
A
Arne Redlich 已提交
2380 2381
	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
2382

2383 2384
	target = host_to_target(target_host);

2385 2386 2387 2388 2389 2390
	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;
2391 2392
	target->sg_tablesize	= indirect_sg_entries ? : cmd_sg_entries;
	target->allow_ext_sg	= allow_ext_sg;
2393
	target->tl_retry_count	= 7;
2394 2395 2396 2397 2398

	ret = srp_parse_options(buf, target);
	if (ret)
		goto err;

2399 2400 2401 2402 2403 2404 2405 2406 2407 2408
	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;
	}

2409 2410
	if (!host->srp_dev->fmr_pool && !target->allow_ext_sg &&
				target->cmd_sg_cnt < target->sg_tablesize) {
2411
		pr_warn("No FMR pool and no external indirect descriptors, limiting sg_tablesize to cmd_sg_cnt\n");
2412 2413 2414 2415 2416 2417
		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);
2418 2419 2420 2421
	target->max_iu_len = sizeof (struct srp_cmd) +
			     sizeof (struct srp_indirect_buf) +
			     target->cmd_sg_cnt * sizeof (struct srp_direct_buf);

2422
	INIT_WORK(&target->remove_work, srp_remove_work);
2423 2424 2425 2426 2427 2428 2429 2430 2431 2432
	spin_lock_init(&target->lock);
	INIT_LIST_HEAD(&target->free_tx);
	INIT_LIST_HEAD(&target->free_reqs);
	for (i = 0; i < SRP_CMD_SQ_SIZE; ++i) {
		struct srp_request *req = &target->req_ring[i];

		req->fmr_list = kmalloc(target->cmd_sg_cnt * sizeof (void *),
					GFP_KERNEL);
		req->map_page = kmalloc(SRP_FMR_SIZE * sizeof (void *),
					GFP_KERNEL);
2433 2434
		req->indirect_desc = kmalloc(target->indirect_size, GFP_KERNEL);
		if (!req->fmr_list || !req->map_page || !req->indirect_desc)
2435 2436
			goto err_free_mem;

2437 2438 2439 2440 2441 2442 2443
		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 err_free_mem;

		req->indirect_dma_addr = dma_addr;
2444 2445 2446 2447
		req->index = i;
		list_add_tail(&req->list, &target->free_reqs);
	}

2448
	ib_query_gid(ibdev, host->port, 0, &target->path.sgid);
2449

2450 2451
	shost_printk(KERN_DEBUG, target->scsi_host, PFX
		     "new target: id_ext %016llx ioc_guid %016llx pkey %04x "
H
Harvey Harrison 已提交
2452
		     "service_id %016llx dgid %pI6\n",
2453 2454 2455 2456
	       (unsigned long long) be64_to_cpu(target->id_ext),
	       (unsigned long long) be64_to_cpu(target->ioc_guid),
	       be16_to_cpu(target->path.pkey),
	       (unsigned long long) be64_to_cpu(target->service_id),
2457
	       target->path.dgid.raw);
2458 2459 2460

	ret = srp_create_target_ib(target);
	if (ret)
2461
		goto err_free_mem;
2462

D
David Dillow 已提交
2463 2464
	ret = srp_new_cm_id(target);
	if (ret)
2465
		goto err_free_ib;
2466 2467 2468

	ret = srp_connect_target(target);
	if (ret) {
2469 2470
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Connection failed\n");
2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485
		goto err_cm_id;
	}

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

	return count;

err_disconnect:
	srp_disconnect_target(target);

err_cm_id:
	ib_destroy_cm_id(target->cm_id);

2486
err_free_ib:
2487 2488
	srp_free_target_ib(target);

2489 2490 2491
err_free_mem:
	srp_free_req_data(target);

2492 2493 2494 2495 2496 2497
err:
	scsi_host_put(target_host);

	return ret;
}

2498
static DEVICE_ATTR(add_target, S_IWUSR, NULL, srp_create_target);
2499

2500 2501
static ssize_t show_ibdev(struct device *dev, struct device_attribute *attr,
			  char *buf)
2502
{
2503
	struct srp_host *host = container_of(dev, struct srp_host, dev);
2504

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

2508
static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
2509

2510 2511
static ssize_t show_port(struct device *dev, struct device_attribute *attr,
			 char *buf)
2512
{
2513
	struct srp_host *host = container_of(dev, struct srp_host, dev);
2514 2515 2516 2517

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

2518
static DEVICE_ATTR(port, S_IRUGO, show_port, NULL);
2519

2520
static struct srp_host *srp_add_port(struct srp_device *device, u8 port)
2521 2522 2523 2524 2525 2526 2527 2528
{
	struct srp_host *host;

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

	INIT_LIST_HEAD(&host->target_list);
2529
	spin_lock_init(&host->target_lock);
2530
	init_completion(&host->released);
2531
	host->srp_dev = device;
2532 2533
	host->port = port;

2534 2535
	host->dev.class = &srp_class;
	host->dev.parent = device->dev->dma_device;
2536
	dev_set_name(&host->dev, "srp-%s-%d", device->dev->name, port);
2537

2538
	if (device_register(&host->dev))
2539
		goto free_host;
2540
	if (device_create_file(&host->dev, &dev_attr_add_target))
2541
		goto err_class;
2542
	if (device_create_file(&host->dev, &dev_attr_ibdev))
2543
		goto err_class;
2544
	if (device_create_file(&host->dev, &dev_attr_port))
2545 2546 2547 2548 2549
		goto err_class;

	return host;

err_class:
2550
	device_unregister(&host->dev);
2551

2552
free_host:
2553 2554 2555 2556 2557 2558 2559
	kfree(host);

	return NULL;
}

static void srp_add_one(struct ib_device *device)
{
2560 2561 2562
	struct srp_device *srp_dev;
	struct ib_device_attr *dev_attr;
	struct ib_fmr_pool_param fmr_param;
2563
	struct srp_host *host;
2564
	int max_pages_per_fmr, fmr_page_shift, s, e, p;
2565

2566 2567
	dev_attr = kmalloc(sizeof *dev_attr, GFP_KERNEL);
	if (!dev_attr)
2568
		return;
2569

2570
	if (ib_query_device(device, dev_attr)) {
2571
		pr_warn("Query device failed for %s\n", device->name);
2572 2573 2574 2575 2576 2577 2578 2579 2580
		goto free_attr;
	}

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

	/*
	 * Use the smallest page size supported by the HCA, down to a
2581 2582
	 * minimum of 4096 bytes. We're unlikely to build large sglists
	 * out of smaller entries.
2583
	 */
2584 2585 2586 2587
	fmr_page_shift		= max(12, ffs(dev_attr->page_size_cap) - 1);
	srp_dev->fmr_page_size	= 1 << fmr_page_shift;
	srp_dev->fmr_page_mask	= ~((u64) srp_dev->fmr_page_size - 1);
	srp_dev->fmr_max_size	= srp_dev->fmr_page_size * SRP_FMR_SIZE;
2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602

	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;

2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620
	for (max_pages_per_fmr = SRP_FMR_SIZE;
			max_pages_per_fmr >= SRP_FMR_MIN_SIZE;
			max_pages_per_fmr /= 2, srp_dev->fmr_max_size /= 2) {
		memset(&fmr_param, 0, sizeof fmr_param);
		fmr_param.pool_size	    = SRP_FMR_POOL_SIZE;
		fmr_param.dirty_watermark   = SRP_FMR_DIRTY_SIZE;
		fmr_param.cache		    = 1;
		fmr_param.max_pages_per_fmr = max_pages_per_fmr;
		fmr_param.page_shift	    = fmr_page_shift;
		fmr_param.access	    = (IB_ACCESS_LOCAL_WRITE |
					       IB_ACCESS_REMOTE_WRITE |
					       IB_ACCESS_REMOTE_READ);

		srp_dev->fmr_pool = ib_create_fmr_pool(srp_dev->pd, &fmr_param);
		if (!IS_ERR(srp_dev->fmr_pool))
			break;
	}

2621 2622
	if (IS_ERR(srp_dev->fmr_pool))
		srp_dev->fmr_pool = NULL;
2623

T
Tom Tucker 已提交
2624
	if (device->node_type == RDMA_NODE_IB_SWITCH) {
2625 2626 2627 2628 2629 2630 2631 2632
		s = 0;
		e = 0;
	} else {
		s = 1;
		e = device->phys_port_cnt;
	}

	for (p = s; p <= e; ++p) {
2633
		host = srp_add_port(srp_dev, p);
2634
		if (host)
2635
			list_add_tail(&host->list, &srp_dev->dev_list);
2636 2637
	}

2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649
	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);
2650 2651 2652 2653
}

static void srp_remove_one(struct ib_device *device)
{
2654
	struct srp_device *srp_dev;
2655
	struct srp_host *host, *tmp_host;
2656
	struct srp_target_port *target;
2657

2658
	srp_dev = ib_get_client_data(device, &srp_client);
2659 2660
	if (!srp_dev)
		return;
2661

2662
	list_for_each_entry_safe(host, tmp_host, &srp_dev->dev_list, list) {
2663
		device_unregister(&host->dev);
2664 2665 2666 2667 2668 2669 2670
		/*
		 * Wait for the sysfs entry to go away, so that no new
		 * target ports can be created.
		 */
		wait_for_completion(&host->released);

		/*
2671
		 * Remove all target ports.
2672
		 */
2673
		spin_lock(&host->target_lock);
2674 2675
		list_for_each_entry(target, &host->target_list, list)
			srp_queue_remove_work(target);
2676
		spin_unlock(&host->target_lock);
2677 2678

		/*
2679
		 * Wait for target port removal tasks.
2680
		 */
2681
		flush_workqueue(system_long_wq);
2682 2683 2684 2685

		kfree(host);
	}

2686 2687 2688 2689 2690 2691
	if (srp_dev->fmr_pool)
		ib_destroy_fmr_pool(srp_dev->fmr_pool);
	ib_dereg_mr(srp_dev->mr);
	ib_dealloc_pd(srp_dev->pd);

	kfree(srp_dev);
2692 2693
}

2694
static struct srp_function_template ib_srp_transport_functions = {
2695 2696 2697 2698 2699
	.has_rport_state	 = true,
	.reset_timer_if_blocked	 = true,
	.fast_io_fail_tmo	 = &srp_fast_io_fail_tmo,
	.dev_loss_tmo		 = &srp_dev_loss_tmo,
	.reconnect		 = srp_rport_reconnect,
2700
	.rport_delete		 = srp_rport_delete,
2701
	.terminate_rport_io	 = srp_terminate_io,
2702 2703
};

2704 2705 2706 2707
static int __init srp_init_module(void)
{
	int ret;

2708
	BUILD_BUG_ON(FIELD_SIZEOF(struct ib_wc, wr_id) < sizeof(void *));
2709

2710
	if (srp_sg_tablesize) {
2711
		pr_warn("srp_sg_tablesize is deprecated, please use cmd_sg_entries\n");
2712 2713 2714 2715 2716 2717 2718 2719
		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) {
2720
		pr_warn("Clamping cmd_sg_entries to 255\n");
2721
		cmd_sg_entries = 255;
2722 2723
	}

2724 2725 2726
	if (!indirect_sg_entries)
		indirect_sg_entries = cmd_sg_entries;
	else if (indirect_sg_entries < cmd_sg_entries) {
2727 2728
		pr_warn("Bumping up indirect_sg_entries to match cmd_sg_entries (%u)\n",
			cmd_sg_entries);
2729 2730 2731
		indirect_sg_entries = cmd_sg_entries;
	}

2732 2733 2734 2735 2736
	ib_srp_transport_template =
		srp_attach_transport(&ib_srp_transport_functions);
	if (!ib_srp_transport_template)
		return -ENOMEM;

2737 2738
	ret = class_register(&srp_class);
	if (ret) {
2739
		pr_err("couldn't register class infiniband_srp\n");
2740
		srp_release_transport(ib_srp_transport_template);
2741 2742 2743
		return ret;
	}

2744 2745
	ib_sa_register_client(&srp_sa_client);

2746 2747
	ret = ib_register_client(&srp_client);
	if (ret) {
2748
		pr_err("couldn't register IB client\n");
2749
		srp_release_transport(ib_srp_transport_template);
2750
		ib_sa_unregister_client(&srp_sa_client);
2751 2752 2753 2754 2755 2756 2757 2758 2759 2760
		class_unregister(&srp_class);
		return ret;
	}

	return 0;
}

static void __exit srp_cleanup_module(void)
{
	ib_unregister_client(&srp_client);
2761
	ib_sa_unregister_client(&srp_sa_client);
2762
	class_unregister(&srp_class);
2763
	srp_release_transport(ib_srp_transport_template);
2764 2765 2766 2767
}

module_init(srp_init_module);
module_exit(srp_cleanup_module);