ib_srp.c 57.5 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 33 34 35 36 37 38 39
/*
 * 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.
 */

#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>
40
#include <linux/jiffies.h>
41 42 43 44 45 46 47

#include <asm/atomic.h>

#include <scsi/scsi.h>
#include <scsi/scsi_device.h>
#include <scsi/scsi_dbg.h>
#include <scsi/srp.h>
48
#include <scsi/scsi_transport_srp.h>
49 50 51 52 53 54 55 56 57 58 59 60 61

#include "ib_srp.h"

#define DRV_NAME	"ib_srp"
#define PFX		DRV_NAME ": "
#define DRV_VERSION	"0.2"
#define DRV_RELDATE	"November 1, 2005"

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

62 63 64 65 66
static int srp_sg_tablesize = SRP_DEF_SG_TABLESIZE;
static int srp_max_iu_len;

module_param(srp_sg_tablesize, int, 0444);
MODULE_PARM_DESC(srp_sg_tablesize,
67
		 "Max number of gather/scatter entries per I/O (default is 12, max 255)");
68

69 70 71 72 73 74
static int topspin_workarounds = 1;

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

75 76 77 78 79 80
static int mellanox_workarounds = 1;

module_param(mellanox_workarounds, int, 0444);
MODULE_PARM_DESC(mellanox_workarounds,
		 "Enable workarounds for Mellanox SRP target bugs if != 0");

81 82
static void srp_add_one(struct ib_device *device);
static void srp_remove_one(struct ib_device *device);
83 84
static void srp_recv_completion(struct ib_cq *cq, void *target_ptr);
static void srp_send_completion(struct ib_cq *cq, void *target_ptr);
85 86
static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event);

87 88
static struct scsi_transport_template *ib_srp_transport_template;

89 90 91 92 93 94
static struct ib_client srp_client = {
	.name   = "srp",
	.add    = srp_add_one,
	.remove = srp_remove_one
};

95 96
static struct ib_sa_client srp_sa_client;

97 98 99 100 101 102 103 104 105 106
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;
}

107 108 109
static int srp_target_is_topspin(struct srp_target_port *target)
{
	static const u8 topspin_oui[3] = { 0x00, 0x05, 0xad };
110
	static const u8 cisco_oui[3]   = { 0x00, 0x1b, 0x0d };
111 112

	return topspin_workarounds &&
113 114
		(!memcmp(&target->ioc_guid, topspin_oui, sizeof topspin_oui) ||
		 !memcmp(&target->ioc_guid, cisco_oui, sizeof cisco_oui));
115 116 117 118 119 120 121 122 123 124
}

static int srp_target_is_mellanox(struct srp_target_port *target)
{
	static const u8 mellanox_oui[3] = { 0x00, 0x02, 0xc9 };

	return mellanox_workarounds &&
		!memcmp(&target->ioc_guid, mellanox_oui, sizeof mellanox_oui);
}

125 126 127 128 129 130 131 132 133 134 135 136 137 138
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;

139 140 141
	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))
142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161
		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;

162 163
	ib_dma_unmap_single(host->srp_dev->dev, iu->dma, iu->size,
			    iu->direction);
164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182
	kfree(iu->buf);
	kfree(iu);
}

static void srp_qp_event(struct ib_event *event, void *context)
{
	printk(KERN_ERR PFX "QP event %d\n", event->event);
}

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;

183 184 185 186
	ret = ib_find_pkey(target->srp_host->srp_dev->dev,
			   target->srp_host->port,
			   be16_to_cpu(target->path.pkey),
			   &attr->pkey_index);
187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205
	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
206 207 208 209
static int srp_new_cm_id(struct srp_target_port *target)
{
	struct ib_cm_id *new_cm_id;

210
	new_cm_id = ib_create_cm_id(target->srp_host->srp_dev->dev,
D
David Dillow 已提交
211 212 213 214 215 216 217 218 219 220 221
				    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;
}

222 223 224 225 226 227 228 229 230
static int srp_create_target_ib(struct srp_target_port *target)
{
	struct ib_qp_init_attr *init_attr;
	int ret;

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

231 232 233 234
	target->recv_cq = ib_create_cq(target->srp_host->srp_dev->dev,
				       srp_recv_completion, NULL, target, SRP_RQ_SIZE, 0);
	if (IS_ERR(target->recv_cq)) {
		ret = PTR_ERR(target->recv_cq);
235
		goto err;
236 237
	}

238 239 240 241
	target->send_cq = ib_create_cq(target->srp_host->srp_dev->dev,
				       srp_send_completion, NULL, target, SRP_SQ_SIZE, 0);
	if (IS_ERR(target->send_cq)) {
		ret = PTR_ERR(target->send_cq);
242
		goto err_recv_cq;
243 244 245
	}

	ib_req_notify_cq(target->recv_cq, IB_CQ_NEXT_COMP);
246 247 248 249 250 251 252 253

	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;
254 255
	init_attr->send_cq             = target->send_cq;
	init_attr->recv_cq             = target->recv_cq;
256

257
	target->qp = ib_create_qp(target->srp_host->srp_dev->pd, init_attr);
258 259
	if (IS_ERR(target->qp)) {
		ret = PTR_ERR(target->qp);
260
		goto err_send_cq;
261 262 263
	}

	ret = srp_init_qp(target, target->qp);
264 265
	if (ret)
		goto err_qp;
266

267 268 269 270 271 272 273 274 275 276 277 278 279
	kfree(init_attr);
	return 0;

err_qp:
	ib_destroy_qp(target->qp);

err_send_cq:
	ib_destroy_cq(target->send_cq);

err_recv_cq:
	ib_destroy_cq(target->recv_cq);

err:
280 281 282 283 284 285 286 287 288
	kfree(init_attr);
	return ret;
}

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

	ib_destroy_qp(target->qp);
289 290
	ib_destroy_cq(target->send_cq);
	ib_destroy_cq(target->recv_cq);
291 292 293

	for (i = 0; i < SRP_RQ_SIZE; ++i)
		srp_free_iu(target->srp_host, target->rx_ring[i]);
294
	for (i = 0; i < SRP_SQ_SIZE; ++i)
295 296 297 298 299 300 301 302 303 304 305
		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)
306 307
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Got failed path rec status %d\n", status);
308 309 310 311 312 313 314 315 316 317 318
	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);

319
	target->path_query_id = ib_sa_path_rec_get(&srp_sa_client,
320
						   target->srp_host->srp_dev->dev,
321 322
						   target->srp_host->port,
						   &target->path,
323
						   IB_SA_PATH_REC_SERVICE_ID	|
324 325 326 327 328 329 330 331 332 333 334 335 336 337
						   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)
338 339
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Path record query failed\n");
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

	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;
	req->param.retry_count 		      = 7;
	req->param.rnr_retry_count 	      = 7;
	req->param.max_cm_retries 	      = 15;

	req->priv.opcode     	= SRP_LOGIN_REQ;
	req->priv.tag        	= 0;
381
	req->priv.req_it_iu_len = cpu_to_be32(srp_max_iu_len);
382 383
	req->priv.req_buf_fmt 	= cpu_to_be16(SRP_BUF_FORMAT_DIRECT |
					      SRP_BUF_FORMAT_INDIRECT);
384
	/*
R
Roland Dreier 已提交
385
	 * In the published SRP specification (draft rev. 16a), the
386 387 388 389 390 391 392 393 394
	 * 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,
395
		       &target->path.sgid.global.interface_id, 8);
396
		memcpy(req->priv.initiator_port_id + 8,
397
		       &target->initiator_ext, 8);
398 399 400 401
		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,
402 403 404
		       &target->initiator_ext, 8);
		memcpy(req->priv.initiator_port_id + 8,
		       &target->path.sgid.global.interface_id, 8);
405 406 407 408
		memcpy(req->priv.target_port_id,     &target->id_ext, 8);
		memcpy(req->priv.target_port_id + 8, &target->ioc_guid, 8);
	}

409 410
	/*
	 * Topspin/Cisco SRP targets will reject our login unless we
411 412
	 * zero out the first 8 bytes of our initiator port ID and set
	 * the second 8 bytes to the local node GUID.
413
	 */
414
	if (srp_target_is_topspin(target)) {
415 416 417 418
		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));
419
		memset(req->priv.initiator_port_id, 0, 8);
420
		memcpy(req->priv.initiator_port_id + 8,
421
		       &target->srp_host->srp_dev->dev->node_guid, 8);
422 423 424 425 426 427 428 429 430 431 432 433 434 435
	}

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

	kfree(req);

	return status;
}

static void srp_disconnect_target(struct srp_target_port *target)
{
	/* XXX should send SRP_I_LOGOUT request */

	init_completion(&target->done);
436
	if (ib_send_cm_dreq(target->cm_id, NULL, 0)) {
437 438
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Sending CM DREQ failed\n");
439 440
		return;
	}
441 442 443
	wait_for_completion(&target->done);
}

D
David Howells 已提交
444
static void srp_remove_work(struct work_struct *work)
445
{
D
David Howells 已提交
446 447
	struct srp_target_port *target =
		container_of(work, struct srp_target_port, work);
448 449 450 451 452 453 454 455 456

	spin_lock_irq(target->scsi_host->host_lock);
	if (target->state != SRP_TARGET_DEAD) {
		spin_unlock_irq(target->scsi_host->host_lock);
		return;
	}
	target->state = SRP_TARGET_REMOVED;
	spin_unlock_irq(target->scsi_host->host_lock);

457
	spin_lock(&target->srp_host->target_lock);
458
	list_del(&target->list);
459
	spin_unlock(&target->srp_host->target_lock);
460

461
	srp_remove_host(target->scsi_host);
462 463 464 465 466 467 468 469
	scsi_remove_host(target->scsi_host);
	ib_destroy_cm_id(target->cm_id);
	srp_free_target_ib(target);
	scsi_host_put(target->scsi_host);
}

static int srp_connect_target(struct srp_target_port *target)
{
D
David Dillow 已提交
470
	int retries = 3;
471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502
	int ret;

	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:
			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
503 504 505 506 507 508 509 510 511 512 513 514 515 516 517
		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;

518 519 520 521 522 523
		default:
			return target->status;
		}
	}
}

524 525 526 527
static void srp_unmap_data(struct scsi_cmnd *scmnd,
			   struct srp_target_port *target,
			   struct srp_request *req)
{
528
	if (!scsi_sglist(scmnd) ||
529 530 531 532
	    (scmnd->sc_data_direction != DMA_TO_DEVICE &&
	     scmnd->sc_data_direction != DMA_FROM_DEVICE))
		return;

533 534 535 536 537
	if (req->fmr) {
		ib_fmr_pool_unmap(req->fmr);
		req->fmr = NULL;
	}

538
	ib_dma_unmap_sg(target->srp_host->srp_dev->dev, scsi_sglist(scmnd),
539
			scsi_sg_count(scmnd), scmnd->sc_data_direction);
540 541
}

542 543 544
static void srp_remove_req(struct srp_target_port *target, struct srp_request *req)
{
	srp_unmap_data(req->scmnd, target, req);
545
	req->scmnd = NULL;
546 547 548 549 550 551 552 553 554 555
	list_move_tail(&req->list, &target->free_reqs);
}

static void srp_reset_req(struct srp_target_port *target, struct srp_request *req)
{
	req->scmnd->result = DID_RESET << 16;
	req->scmnd->scsi_done(req->scmnd);
	srp_remove_req(target, req);
}

556 557 558
static int srp_reconnect_target(struct srp_target_port *target)
{
	struct ib_qp_attr qp_attr;
559
	struct srp_request *req, *tmp;
560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575
	struct ib_wc wc;
	int ret;

	spin_lock_irq(target->scsi_host->host_lock);
	if (target->state != SRP_TARGET_LIVE) {
		spin_unlock_irq(target->scsi_host->host_lock);
		return -EAGAIN;
	}
	target->state = SRP_TARGET_CONNECTING;
	spin_unlock_irq(target->scsi_host->host_lock);

	srp_disconnect_target(target);
	/*
	 * Now get a new local CM ID so that we avoid confusing the
	 * target in case things are really fouled up.
	 */
D
David Dillow 已提交
576 577
	ret = srp_new_cm_id(target);
	if (ret)
578 579 580 581 582 583 584 585 586 587 588
		goto err;

	qp_attr.qp_state = IB_QPS_RESET;
	ret = ib_modify_qp(target->qp, &qp_attr, IB_QP_STATE);
	if (ret)
		goto err;

	ret = srp_init_qp(target, target->qp);
	if (ret)
		goto err;

589 590 591
	while (ib_poll_cq(target->recv_cq, 1, &wc) > 0)
		; /* nothing */
	while (ib_poll_cq(target->send_cq, 1, &wc) > 0)
592 593
		; /* nothing */

594
	spin_lock_irq(target->scsi_host->host_lock);
595 596
	list_for_each_entry_safe(req, tmp, &target->req_queue, list)
		srp_reset_req(target, req);
597
	spin_unlock_irq(target->scsi_host->host_lock);
598 599 600 601 602

	target->rx_head	 = 0;
	target->tx_head	 = 0;
	target->tx_tail  = 0;

603
	target->qp_in_error = 0;
604 605 606 607 608 609 610 611 612 613 614 615 616 617 618
	ret = srp_connect_target(target);
	if (ret)
		goto err;

	spin_lock_irq(target->scsi_host->host_lock);
	if (target->state == SRP_TARGET_CONNECTING) {
		ret = 0;
		target->state = SRP_TARGET_LIVE;
	} else
		ret = -EAGAIN;
	spin_unlock_irq(target->scsi_host->host_lock);

	return ret;

err:
619 620
	shost_printk(KERN_ERR, target->scsi_host,
		     PFX "reconnect failed (%d), removing target port.\n", ret);
621 622 623 624 625 626 627 628 629 630

	/*
	 * We couldn't reconnect, so kill our target port off.
	 * However, we have to defer the real removal because we might
	 * be in the context of the SCSI error handler now, which
	 * would deadlock if we call scsi_remove_host().
	 */
	spin_lock_irq(target->scsi_host->host_lock);
	if (target->state == SRP_TARGET_CONNECTING) {
		target->state = SRP_TARGET_DEAD;
D
David Howells 已提交
631
		INIT_WORK(&target->work, srp_remove_work);
632 633 634 635 636 637 638
		schedule_work(&target->work);
	}
	spin_unlock_irq(target->scsi_host->host_lock);

	return ret;
}

639
static int srp_map_fmr(struct srp_target_port *target, struct scatterlist *scat,
640 641 642 643 644 645 646 647 648
		       int sg_cnt, struct srp_request *req,
		       struct srp_direct_buf *buf)
{
	u64 io_addr = 0;
	u64 *dma_pages;
	u32 len;
	int page_cnt;
	int i, j;
	int ret;
649
	struct srp_device *dev = target->srp_host->srp_dev;
650
	struct ib_device *ibdev = dev->dev;
651
	struct scatterlist *sg;
652 653 654 655

	if (!dev->fmr_pool)
		return -ENODEV;

656 657
	if (srp_target_is_mellanox(target) &&
	    (ib_sg_dma_address(ibdev, &scat[0]) & ~dev->fmr_page_mask))
658 659
		return -EINVAL;

660
	len = page_cnt = 0;
661 662
	scsi_for_each_sg(req->scmnd, sg, sg_cnt, i) {
		unsigned int dma_len = ib_sg_dma_len(ibdev, sg);
663

664
		if (ib_sg_dma_address(ibdev, sg) & ~dev->fmr_page_mask) {
665 666 667 668 669
			if (i > 0)
				return -EINVAL;
			else
				++page_cnt;
		}
670
		if ((ib_sg_dma_address(ibdev, sg) + dma_len) &
671 672 673 674 675 676 677
		    ~dev->fmr_page_mask) {
			if (i < sg_cnt - 1)
				return -EINVAL;
			else
				++page_cnt;
		}

678
		len += dma_len;
679 680 681 682 683 684 685 686 687 688 689
	}

	page_cnt += len >> dev->fmr_page_shift;
	if (page_cnt > SRP_FMR_SIZE)
		return -ENOMEM;

	dma_pages = kmalloc(sizeof (u64) * page_cnt, GFP_ATOMIC);
	if (!dma_pages)
		return -ENOMEM;

	page_cnt = 0;
690 691
	scsi_for_each_sg(req->scmnd, sg, sg_cnt, i) {
		unsigned int dma_len = ib_sg_dma_len(ibdev, sg);
692 693

		for (j = 0; j < dma_len; j += dev->fmr_page_size)
694
			dma_pages[page_cnt++] =
695
				(ib_sg_dma_address(ibdev, sg) &
696 697
				 dev->fmr_page_mask) + j;
	}
698 699

	req->fmr = ib_fmr_pool_map_phys(dev->fmr_pool,
700
					dma_pages, page_cnt, io_addr);
701 702
	if (IS_ERR(req->fmr)) {
		ret = PTR_ERR(req->fmr);
V
Vu Pham 已提交
703
		req->fmr = NULL;
704 705 706
		goto out;
	}

707 708
	buf->va  = cpu_to_be64(ib_sg_dma_address(ibdev, &scat[0]) &
			       ~dev->fmr_page_mask);
709 710 711 712 713 714 715 716 717 718 719
	buf->key = cpu_to_be32(req->fmr->fmr->rkey);
	buf->len = cpu_to_be32(len);

	ret = 0;

out:
	kfree(dma_pages);

	return ret;
}

720 721 722
static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_target_port *target,
			struct srp_request *req)
{
723
	struct scatterlist *scat;
724
	struct srp_cmd *cmd = req->cmd->buf;
725
	int len, nents, count;
726
	u8 fmt = SRP_DATA_DESC_DIRECT;
727 728
	struct srp_device *dev;
	struct ib_device *ibdev;
729

730
	if (!scsi_sglist(scmnd) || scmnd->sc_data_direction == DMA_NONE)
731 732 733 734
		return sizeof (struct srp_cmd);

	if (scmnd->sc_data_direction != DMA_FROM_DEVICE &&
	    scmnd->sc_data_direction != DMA_TO_DEVICE) {
735 736 737
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled data direction %d\n",
			     scmnd->sc_data_direction);
738 739 740
		return -EINVAL;
	}

741 742
	nents = scsi_sg_count(scmnd);
	scat  = scsi_sglist(scmnd);
743

744
	dev = target->srp_host->srp_dev;
745 746 747
	ibdev = dev->dev;

	count = ib_dma_map_sg(ibdev, scat, nents, scmnd->sc_data_direction);
748 749 750

	fmt = SRP_DATA_DESC_DIRECT;
	len = sizeof (struct srp_cmd) +	sizeof (struct srp_direct_buf);
751

752
	if (count == 1) {
753 754 755 756 757 758
		/*
		 * 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.
		 */
759
		struct srp_direct_buf *buf = (void *) cmd->add_data;
760

761 762 763
		buf->va  = cpu_to_be64(ib_sg_dma_address(ibdev, scat));
		buf->key = cpu_to_be32(dev->mr->rkey);
		buf->len = cpu_to_be32(ib_sg_dma_len(ibdev, scat));
764
	} else if (srp_map_fmr(target, scat, count, req,
765 766 767 768 769 770
			       (void *) cmd->add_data)) {
		/*
		 * FMR mapping failed, and the scatterlist has more
		 * than one entry.  Generate an indirect memory
		 * descriptor.
		 */
771
		struct srp_indirect_buf *buf = (void *) cmd->add_data;
772
		struct scatterlist *sg;
773
		u32 datalen = 0;
774
		int i;
775

776
		fmt = SRP_DATA_DESC_INDIRECT;
777 778 779 780
		len = sizeof (struct srp_cmd) +
			sizeof (struct srp_indirect_buf) +
			count * sizeof (struct srp_direct_buf);

781 782
		scsi_for_each_sg(scmnd, sg, count, i) {
			unsigned int dma_len = ib_sg_dma_len(ibdev, sg);
783

784
			buf->desc_list[i].va  =
785
				cpu_to_be64(ib_sg_dma_address(ibdev, sg));
786
			buf->desc_list[i].key =
787 788 789
				cpu_to_be32(dev->mr->rkey);
			buf->desc_list[i].len = cpu_to_be32(dma_len);
			datalen += dma_len;
790
		}
791

792 793 794 795 796
		if (scmnd->sc_data_direction == DMA_TO_DEVICE)
			cmd->data_out_desc_cnt = count;
		else
			cmd->data_in_desc_cnt = count;

797 798
		buf->table_desc.va  =
			cpu_to_be64(req->cmd->dma + sizeof *cmd + sizeof *buf);
799
		buf->table_desc.key =
800
			cpu_to_be32(target->srp_host->srp_dev->mr->rkey);
801 802 803 804
		buf->table_desc.len =
			cpu_to_be32(count * sizeof (struct srp_direct_buf));

		buf->len = cpu_to_be32(datalen);
805 806 807 808 809 810 811 812 813 814
	}

	if (scmnd->sc_data_direction == DMA_TO_DEVICE)
		cmd->buf_fmt = fmt << 4;
	else
		cmd->buf_fmt = fmt;

	return len;
}

815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883
/*
 * Must be called with target->scsi_host->host_lock held to protect
 * req_lim and tx_head.  Lock cannot be dropped between call here and
 * call to __srp_post_send().
 *
 * 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);

	if (target->tx_head - target->tx_tail >= SRP_SQ_SIZE)
		return NULL;

	/* Initiator responses to target requests do not consume credits */
	if (target->req_lim <= rsv && iu_type != SRP_IU_RSP) {
		++target->zero_req_lim;
		return NULL;
	}

	iu = target->tx_ring[target->tx_head & SRP_SQ_MASK];
	iu->type = iu_type;
	return iu;
}

/*
 * Must be called with target->scsi_host->host_lock held to protect
 * req_lim and tx_head.
 */
static int __srp_post_send(struct srp_target_port *target,
			   struct srp_iu *iu, int len)
{
	struct ib_sge list;
	struct ib_send_wr wr, *bad_wr;
	int ret = 0;

	list.addr   = iu->dma;
	list.length = len;
	list.lkey   = target->srp_host->srp_dev->mr->lkey;

	wr.next       = NULL;
	wr.wr_id      = target->tx_head & SRP_SQ_MASK;
	wr.sg_list    = &list;
	wr.num_sge    = 1;
	wr.opcode     = IB_WR_SEND;
	wr.send_flags = IB_SEND_SIGNALED;

	ret = ib_post_send(target->qp, &wr, &bad_wr);

	if (!ret) {
		++target->tx_head;
		if (iu->type != SRP_IU_RSP)
			--target->req_lim;
	}

	return ret;
}

884 885 886 887 888 889 890 891 892 893 894
static int srp_post_recv(struct srp_target_port *target)
{
	unsigned long flags;
	struct srp_iu *iu;
	struct ib_sge list;
	struct ib_recv_wr wr, *bad_wr;
	unsigned int next;
	int ret;

	spin_lock_irqsave(target->scsi_host->host_lock, flags);

895
	next	 = target->rx_head & SRP_RQ_MASK;
896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915
	wr.wr_id = next;
	iu	 = target->rx_ring[next];

	list.addr   = iu->dma;
	list.length = iu->size;
	list.lkey   = target->srp_host->srp_dev->mr->lkey;

	wr.next     = NULL;
	wr.sg_list  = &list;
	wr.num_sge  = 1;

	ret = ib_post_recv(target->qp, &wr, &bad_wr);
	if (!ret)
		++target->rx_head;

	spin_unlock_irqrestore(target->scsi_host->host_lock, flags);

	return ret;
}

916 917 918 919 920 921 922 923 924 925 926 927 928 929
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;
	s32 delta;

	delta = (s32) be32_to_cpu(rsp->req_lim_delta);

	spin_lock_irqsave(target->scsi_host->host_lock, flags);

	target->req_lim += delta;

	if (unlikely(rsp->tag & SRP_TAG_TSK_MGMT)) {
930 931 932 933
		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);
934
	} else {
935
		req = &target->req_ring[rsp->tag];
936
		scmnd = req->scmnd;
937
		if (!scmnd)
938 939 940
			shost_printk(KERN_ERR, target->scsi_host,
				     "Null scmnd for RSP w/tag %016llx\n",
				     (unsigned long long) rsp->tag);
941 942 943 944 945 946 947 948 949 950
		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))
951
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_out_res_cnt));
952
		else if (rsp->flags & (SRP_RSP_FLAG_DIOVER | SRP_RSP_FLAG_DIUNDER))
953
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_in_res_cnt));
954

955 956 957
		scmnd->host_scribble = NULL;
		scmnd->scsi_done(scmnd);
		srp_remove_req(target, req);
958 959 960 961 962
	}

	spin_unlock_irqrestore(target->scsi_host->host_lock, flags);
}

963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027
static int srp_response_common(struct srp_target_port *target, s32 req_delta,
			       void *rsp, int len)
{
	struct ib_device *dev;
	unsigned long flags;
	struct srp_iu *iu;
	int err = 1;

	dev = target->srp_host->srp_dev->dev;

	spin_lock_irqsave(target->scsi_host->host_lock, flags);
	target->req_lim += req_delta;

	iu = __srp_get_tx_iu(target, SRP_IU_RSP);
	if (!iu) {
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "no IU available to send response\n");
		goto out;
	}

	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);

	err = __srp_post_send(target, iu, len);
	if (err)
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "unable to post response: %d\n", err);

out:
	spin_unlock_irqrestore(target->scsi_host->host_lock, flags);
	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");
}

1028 1029
static void srp_handle_recv(struct srp_target_port *target, struct ib_wc *wc)
{
1030
	struct ib_device *dev;
1031
	struct srp_iu *iu;
1032
	int res;
1033 1034
	u8 opcode;

1035
	iu = target->rx_ring[wc->wr_id];
1036

1037
	dev = target->srp_host->srp_dev->dev;
1038 1039
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_ti_iu_len,
				   DMA_FROM_DEVICE);
1040 1041 1042 1043

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

	if (0) {
1044 1045
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "recv completion, opcode 0x%02x\n", opcode);
B
Bart Van Assche 已提交
1046 1047
		print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET, 8, 1,
			       iu->buf, wc->byte_len, true);
1048 1049 1050 1051 1052 1053 1054
	}

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

1055 1056 1057 1058 1059 1060 1061 1062
	case SRP_CRED_REQ:
		srp_process_cred_req(target, iu->buf);
		break;

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

1063 1064
	case SRP_T_LOGOUT:
		/* XXX Handle target logout */
1065 1066
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Got target logout request\n");
1067 1068 1069
		break;

	default:
1070 1071
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled SRP opcode 0x%02x\n", opcode);
1072 1073 1074
		break;
	}

1075 1076
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_ti_iu_len,
				      DMA_FROM_DEVICE);
1077 1078 1079 1080 1081

	res = srp_post_recv(target);
	if (res != 0)
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Recv failed with error code %d\n", res);
1082 1083
}

1084
static void srp_recv_completion(struct ib_cq *cq, void *target_ptr)
1085 1086 1087 1088 1089 1090 1091
{
	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) {
		if (wc.status) {
1092
			shost_printk(KERN_ERR, target->scsi_host,
1093
				     PFX "failed receive status %d\n",
1094
				     wc.status);
1095
			target->qp_in_error = 1;
1096 1097 1098
			break;
		}

1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117
		srp_handle_recv(target, &wc);
	}
}

static void srp_send_completion(struct ib_cq *cq, void *target_ptr)
{
	struct srp_target_port *target = target_ptr;
	struct ib_wc wc;

	while (ib_poll_cq(cq, 1, &wc) > 0) {
		if (wc.status) {
			shost_printk(KERN_ERR, target->scsi_host,
				     PFX "failed send status %d\n",
				     wc.status);
			target->qp_in_error = 1;
			break;
		}

		++target->tx_tail;
1118 1119 1120
	}
}

J
Jeff Garzik 已提交
1121
static int srp_queuecommand_lck(struct scsi_cmnd *scmnd,
1122 1123 1124 1125 1126 1127
			    void (*done)(struct scsi_cmnd *))
{
	struct srp_target_port *target = host_to_target(scmnd->device->host);
	struct srp_request *req;
	struct srp_iu *iu;
	struct srp_cmd *cmd;
1128
	struct ib_device *dev;
1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140
	int len;

	if (target->state == SRP_TARGET_CONNECTING)
		goto err;

	if (target->state == SRP_TARGET_DEAD ||
	    target->state == SRP_TARGET_REMOVED) {
		scmnd->result = DID_BAD_TARGET << 16;
		done(scmnd);
		return 0;
	}

1141
	iu = __srp_get_tx_iu(target, SRP_IU_CMD);
1142 1143 1144
	if (!iu)
		goto err;

1145
	dev = target->srp_host->srp_dev->dev;
1146 1147
	ib_dma_sync_single_for_cpu(dev, iu->dma, srp_max_iu_len,
				   DMA_TO_DEVICE);
1148

B
Bart Van Assche 已提交
1149
	req = list_first_entry(&target->free_reqs, struct srp_request, list);
1150 1151 1152

	scmnd->scsi_done     = done;
	scmnd->result        = 0;
1153
	scmnd->host_scribble = (void *) req;
1154 1155 1156 1157 1158 1159

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

	cmd->opcode = SRP_CMD;
	cmd->lun    = cpu_to_be64((u64) scmnd->device->lun << 48);
1160
	cmd->tag    = req->index;
1161 1162 1163 1164 1165 1166 1167
	memcpy(cmd->cdb, scmnd->cmnd, scmnd->cmd_len);

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

	len = srp_map_data(scmnd, target, req);
	if (len < 0) {
1168 1169
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Failed to map data\n");
1170 1171 1172
		goto err;
	}

1173 1174
	ib_dma_sync_single_for_device(dev, iu->dma, srp_max_iu_len,
				      DMA_TO_DEVICE);
1175 1176

	if (__srp_post_send(target, iu, len)) {
1177
		shost_printk(KERN_ERR, target->scsi_host, PFX "Send failed\n");
1178 1179 1180
		goto err_unmap;
	}

1181
	list_move_tail(&req->list, &target->req_queue);
1182 1183 1184 1185 1186 1187 1188 1189 1190 1191

	return 0;

err_unmap:
	srp_unmap_data(scmnd, target, req);

err:
	return SCSI_MLQUEUE_HOST_BUSY;
}

J
Jeff Garzik 已提交
1192 1193
static DEF_SCSI_QCMD(srp_queuecommand)

1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205
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;
	}

1206
	for (i = 0; i < SRP_SQ_SIZE; ++i) {
1207
		target->tx_ring[i] = srp_alloc_iu(target->srp_host,
1208
						  srp_max_iu_len,
1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221
						  GFP_KERNEL, DMA_TO_DEVICE);
		if (!target->tx_ring[i])
			goto err;
	}

	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;
	}

1222
	for (i = 0; i < SRP_SQ_SIZE; ++i) {
1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233
		srp_free_iu(target->srp_host, target->tx_ring[i]);
		target->tx_ring[i] = NULL;
	}

	return -ENOMEM;
}

static void srp_cm_rej_handler(struct ib_cm_id *cm_id,
			       struct ib_cm_event *event,
			       struct srp_target_port *target)
{
1234
	struct Scsi_Host *shost = target->scsi_host;
1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250
	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:
1251
		if (srp_target_is_topspin(target)) {
1252 1253 1254 1255 1256 1257 1258 1259
			/*
			 * 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);

1260 1261 1262 1263
			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));
1264 1265 1266

			target->status = SRP_PORT_REDIRECT;
		} else {
1267 1268
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_PORT_REDIRECT\n");
1269 1270 1271 1272 1273
			target->status = -ECONNRESET;
		}
		break;

	case IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID:
1274 1275
		shost_printk(KERN_WARNING, shost,
			    "  REJ reason: IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID\n");
1276 1277 1278 1279 1280 1281 1282 1283 1284 1285
		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)
1286 1287
				shost_printk(KERN_WARNING, shost,
					     PFX "SRP_LOGIN_REJ: requested max_it_iu_len too large\n");
1288
			else
1289 1290
				shost_printk(KERN_WARNING, shost,
					    PFX "SRP LOGIN REJECTED, reason 0x%08x\n", reason);
1291
		} else
1292 1293 1294
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_CONSUMER_DEFINED,"
				     " opcode 0x%02x\n", opcode);
1295 1296 1297
		target->status = -ECONNRESET;
		break;

D
David Dillow 已提交
1298 1299 1300 1301 1302
	case IB_CM_REJ_STALE_CONN:
		shost_printk(KERN_WARNING, shost, "  REJ reason: stale connection\n");
		target->status = SRP_STALE_CONN;
		break;

1303
	default:
1304 1305
		shost_printk(KERN_WARNING, shost, "  REJ reason 0x%x\n",
			     event->param.rej_rcvd.reason);
1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316
		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;
	struct ib_qp_attr *qp_attr = NULL;
	int attr_mask = 0;
	int comp = 0;
	int opcode = 0;
1317
	int i;
1318 1319 1320

	switch (event->event) {
	case IB_CM_REQ_ERROR:
1321 1322
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Sending CM REQ failed\n");
1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336
		comp = 1;
		target->status = -ECONNRESET;
		break;

	case IB_CM_REP_RECEIVED:
		comp = 1;
		opcode = *(u8 *) event->private_data;

		if (opcode == SRP_LOGIN_RSP) {
			struct srp_login_rsp *rsp = event->private_data;

			target->max_ti_iu_len = be32_to_cpu(rsp->max_ti_iu_len);
			target->req_lim       = be32_to_cpu(rsp->req_lim_delta);

1337 1338 1339 1340 1341 1342 1343
			/*
			 * 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);
1344
		} else {
1345 1346
			shost_printk(KERN_WARNING, target->scsi_host,
				    PFX "Unhandled RSP opcode %#x\n", opcode);
1347 1348 1349 1350
			target->status = -ECONNRESET;
			break;
		}

V
Vu Pham 已提交
1351 1352 1353 1354 1355
		if (!target->rx_ring[0]) {
			target->status = srp_alloc_iu_bufs(target);
			if (target->status)
				break;
		}
1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371

		qp_attr = kmalloc(sizeof *qp_attr, GFP_KERNEL);
		if (!qp_attr) {
			target->status = -ENOMEM;
			break;
		}

		qp_attr->qp_state = IB_QPS_RTR;
		target->status = ib_cm_init_qp_attr(cm_id, qp_attr, &attr_mask);
		if (target->status)
			break;

		target->status = ib_modify_qp(target->qp, qp_attr, attr_mask);
		if (target->status)
			break;

1372 1373 1374 1375 1376
		for (i = 0; i < SRP_RQ_SIZE; i++) {
			target->status = srp_post_recv(target);
			if (target->status)
				break;
		}
1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395
		if (target->status)
			break;

		qp_attr->qp_state = IB_QPS_RTS;
		target->status = ib_cm_init_qp_attr(cm_id, qp_attr, &attr_mask);
		if (target->status)
			break;

		target->status = ib_modify_qp(target->qp, qp_attr, attr_mask);
		if (target->status)
			break;

		target->status = ib_send_cm_rtu(cm_id, NULL, 0);
		if (target->status)
			break;

		break;

	case IB_CM_REJ_RECEIVED:
1396
		shost_printk(KERN_DEBUG, target->scsi_host, PFX "REJ received\n");
1397 1398 1399 1400 1401
		comp = 1;

		srp_cm_rej_handler(cm_id, event, target);
		break;

1402
	case IB_CM_DREQ_RECEIVED:
1403 1404
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "DREQ received - connection closed\n");
1405
		if (ib_send_cm_drep(cm_id, NULL, 0))
1406 1407
			shost_printk(KERN_ERR, target->scsi_host,
				     PFX "Sending CM DREP failed\n");
1408 1409 1410
		break;

	case IB_CM_TIMEWAIT_EXIT:
1411 1412
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "connection closed\n");
1413 1414 1415 1416 1417

		comp = 1;
		target->status = 0;
		break;

1418 1419 1420 1421 1422
	case IB_CM_MRA_RECEIVED:
	case IB_CM_DREQ_ERROR:
	case IB_CM_DREP_RECEIVED:
		break;

1423
	default:
1424 1425
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled CM event %d\n", event->event);
1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436
		break;
	}

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

	kfree(qp_attr);

	return 0;
}

1437
static int srp_send_tsk_mgmt(struct srp_target_port *target,
1438
			     u64 req_tag, unsigned int lun, u8 func)
1439
{
1440
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1441 1442 1443 1444 1445
	struct srp_iu *iu;
	struct srp_tsk_mgmt *tsk_mgmt;

	spin_lock_irq(target->scsi_host->host_lock);

1446
	if (target->state == SRP_TARGET_DEAD ||
1447
	    target->state == SRP_TARGET_REMOVED)
1448 1449
		goto out;

1450
	init_completion(&target->tsk_mgmt_done);
1451

1452
	iu = __srp_get_tx_iu(target, SRP_IU_TSK_MGMT);
1453 1454 1455
	if (!iu)
		goto out;

1456 1457
	ib_dma_sync_single_for_cpu(dev, iu->dma, sizeof *tsk_mgmt,
				   DMA_TO_DEVICE);
1458 1459 1460 1461
	tsk_mgmt = iu->buf;
	memset(tsk_mgmt, 0, sizeof *tsk_mgmt);

	tsk_mgmt->opcode 	= SRP_TSK_MGMT;
1462 1463
	tsk_mgmt->lun		= cpu_to_be64((u64) lun << 48);
	tsk_mgmt->tag		= req_tag | SRP_TAG_TSK_MGMT;
1464
	tsk_mgmt->tsk_mgmt_func = func;
1465
	tsk_mgmt->task_tag	= req_tag;
1466

1467 1468
	ib_dma_sync_single_for_device(dev, iu->dma, sizeof *tsk_mgmt,
				      DMA_TO_DEVICE);
1469 1470 1471 1472
	if (__srp_post_send(target, iu, sizeof *tsk_mgmt))
		goto out;

	spin_unlock_irq(target->scsi_host->host_lock);
1473

1474
	if (!wait_for_completion_timeout(&target->tsk_mgmt_done,
1475
					 msecs_to_jiffies(SRP_ABORT_TIMEOUT_MS)))
1476
		return -1;
1477

1478
	return 0;
1479 1480 1481

out:
	spin_unlock_irq(target->scsi_host->host_lock);
1482 1483 1484
	return -1;
}

1485 1486
static int srp_abort(struct scsi_cmnd *scmnd)
{
1487
	struct srp_target_port *target = host_to_target(scmnd->device->host);
1488
	struct srp_request *req = (struct srp_request *) scmnd->host_scribble;
1489 1490
	int ret = SUCCESS;

1491
	shost_printk(KERN_ERR, target->scsi_host, "SRP abort called\n");
1492

1493
	if (!req || target->qp_in_error)
1494
		return FAILED;
1495 1496
	if (srp_send_tsk_mgmt(target, req->index, scmnd->device->lun,
			      SRP_TSK_ABORT_TASK))
1497 1498 1499 1500
		return FAILED;

	spin_lock_irq(target->scsi_host->host_lock);

1501 1502 1503 1504 1505 1506 1507
	if (req->scmnd) {
		if (!target->tsk_mgmt_status) {
			srp_remove_req(target, req);
			scmnd->result = DID_ABORT << 16;
		} else
			ret = FAILED;
	}
1508 1509 1510 1511

	spin_unlock_irq(target->scsi_host->host_lock);

	return ret;
1512 1513 1514 1515
}

static int srp_reset_device(struct scsi_cmnd *scmnd)
{
1516 1517 1518
	struct srp_target_port *target = host_to_target(scmnd->device->host);
	struct srp_request *req, *tmp;

1519
	shost_printk(KERN_ERR, target->scsi_host, "SRP reset_device called\n");
1520

1521 1522
	if (target->qp_in_error)
		return FAILED;
1523 1524
	if (srp_send_tsk_mgmt(target, SRP_TAG_NO_REQ, scmnd->device->lun,
			      SRP_TSK_LUN_RESET))
1525
		return FAILED;
1526
	if (target->tsk_mgmt_status)
1527 1528 1529 1530 1531
		return FAILED;

	spin_lock_irq(target->scsi_host->host_lock);

	list_for_each_entry_safe(req, tmp, &target->req_queue, list)
1532
		if (req->scmnd && req->scmnd->device == scmnd->device)
1533
			srp_reset_req(target, req);
1534 1535 1536 1537

	spin_unlock_irq(target->scsi_host->host_lock);

	return SUCCESS;
1538 1539 1540 1541 1542 1543 1544
}

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

1545
	shost_printk(KERN_ERR, target->scsi_host, PFX "SRP reset_host called\n");
1546 1547 1548 1549 1550 1551 1552

	if (!srp_reconnect_target(target))
		ret = SUCCESS;

	return ret;
}

1553 1554
static ssize_t show_id_ext(struct device *dev, struct device_attribute *attr,
			   char *buf)
1555
{
1556
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1557 1558 1559 1560 1561 1562 1563 1564 1565

	if (target->state == SRP_TARGET_DEAD ||
	    target->state == SRP_TARGET_REMOVED)
		return -ENODEV;

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

1566 1567
static ssize_t show_ioc_guid(struct device *dev, struct device_attribute *attr,
			     char *buf)
1568
{
1569
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1570 1571 1572 1573 1574 1575 1576 1577 1578

	if (target->state == SRP_TARGET_DEAD ||
	    target->state == SRP_TARGET_REMOVED)
		return -ENODEV;

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

1579 1580
static ssize_t show_service_id(struct device *dev,
			       struct device_attribute *attr, char *buf)
1581
{
1582
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1583 1584 1585 1586 1587 1588 1589 1590 1591

	if (target->state == SRP_TARGET_DEAD ||
	    target->state == SRP_TARGET_REMOVED)
		return -ENODEV;

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

1592 1593
static ssize_t show_pkey(struct device *dev, struct device_attribute *attr,
			 char *buf)
1594
{
1595
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1596 1597 1598 1599 1600 1601 1602 1603

	if (target->state == SRP_TARGET_DEAD ||
	    target->state == SRP_TARGET_REMOVED)
		return -ENODEV;

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

1604 1605
static ssize_t show_dgid(struct device *dev, struct device_attribute *attr,
			 char *buf)
1606
{
1607
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1608 1609 1610 1611 1612

	if (target->state == SRP_TARGET_DEAD ||
	    target->state == SRP_TARGET_REMOVED)
		return -ENODEV;

H
Harvey Harrison 已提交
1613
	return sprintf(buf, "%pI6\n", target->path.dgid.raw);
1614 1615
}

1616 1617
static ssize_t show_orig_dgid(struct device *dev,
			      struct device_attribute *attr, char *buf)
1618
{
1619
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1620 1621 1622 1623 1624

	if (target->state == SRP_TARGET_DEAD ||
	    target->state == SRP_TARGET_REMOVED)
		return -ENODEV;

H
Harvey Harrison 已提交
1625
	return sprintf(buf, "%pI6\n", target->orig_dgid);
1626 1627
}

1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639
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));

	if (target->state == SRP_TARGET_DEAD ||
	    target->state == SRP_TARGET_REMOVED)
		return -ENODEV;

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

1640 1641
static ssize_t show_zero_req_lim(struct device *dev,
				 struct device_attribute *attr, char *buf)
1642
{
1643
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1644 1645 1646 1647 1648 1649 1650 1651

	if (target->state == SRP_TARGET_DEAD ||
	    target->state == SRP_TARGET_REMOVED)
		return -ENODEV;

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

1652 1653
static ssize_t show_local_ib_port(struct device *dev,
				  struct device_attribute *attr, char *buf)
1654
{
1655
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1656 1657 1658 1659

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

1660 1661
static ssize_t show_local_ib_device(struct device *dev,
				    struct device_attribute *attr, char *buf)
1662
{
1663
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1664

1665
	return sprintf(buf, "%s\n", target->srp_host->srp_dev->dev->name);
1666 1667
}

1668 1669 1670 1671 1672 1673
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);
1674
static DEVICE_ATTR(req_lim,         S_IRUGO, show_req_lim,         NULL);
1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685
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);

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,
1686
	&dev_attr_req_lim,
1687 1688 1689
	&dev_attr_zero_req_lim,
	&dev_attr_local_ib_port,
	&dev_attr_local_ib_device,
1690 1691 1692
	NULL
};

1693 1694
static struct scsi_host_template srp_template = {
	.module				= THIS_MODULE,
R
Roland Dreier 已提交
1695 1696
	.name				= "InfiniBand SRP initiator",
	.proc_name			= DRV_NAME,
1697 1698 1699 1700 1701
	.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,
1702
	.can_queue			= SRP_CMD_SQ_SIZE,
1703
	.this_id			= -1,
1704
	.cmd_per_lun			= SRP_CMD_SQ_SIZE,
1705 1706
	.use_clustering			= ENABLE_CLUSTERING,
	.shost_attrs			= srp_host_attrs
1707 1708 1709 1710
};

static int srp_add_target(struct srp_host *host, struct srp_target_port *target)
{
1711 1712 1713
	struct srp_rport_identifiers ids;
	struct srp_rport *rport;

1714 1715 1716
	sprintf(target->target_name, "SRP.T10:%016llX",
		 (unsigned long long) be64_to_cpu(target->id_ext));

1717
	if (scsi_add_host(target->scsi_host, host->srp_dev->dev->dma_device))
1718 1719
		return -ENODEV;

1720 1721
	memcpy(ids.port_id, &target->id_ext, 8);
	memcpy(ids.port_id + 8, &target->ioc_guid, 8);
1722
	ids.roles = SRP_RPORT_ROLE_TARGET;
1723 1724 1725 1726 1727 1728
	rport = srp_rport_add(target->scsi_host, &ids);
	if (IS_ERR(rport)) {
		scsi_remove_host(target->scsi_host);
		return PTR_ERR(rport);
	}

1729
	spin_lock(&host->target_lock);
1730
	list_add_tail(&target->list, &host->target_list);
1731
	spin_unlock(&host->target_lock);
1732 1733 1734 1735

	target->state = SRP_TARGET_LIVE;

	scsi_scan_target(&target->scsi_host->shost_gendev,
1736
			 0, target->scsi_id, SCAN_WILD_CARD, 0);
1737 1738 1739 1740

	return 0;
}

1741
static void srp_release_dev(struct device *dev)
1742 1743
{
	struct srp_host *host =
1744
		container_of(dev, struct srp_host, dev);
1745 1746 1747 1748 1749 1750

	complete(&host->released);
}

static struct class srp_class = {
	.name    = "infiniband_srp",
1751
	.dev_release = srp_release_dev
1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769
};

/*
 * 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,
1770
	SRP_OPT_MAX_CMD_PER_LUN	= 1 << 6,
1771
	SRP_OPT_IO_CLASS	= 1 << 7,
1772
	SRP_OPT_INITIATOR_EXT	= 1 << 8,
1773 1774 1775 1776 1777 1778 1779
	SRP_OPT_ALL		= (SRP_OPT_ID_EXT	|
				   SRP_OPT_IOC_GUID	|
				   SRP_OPT_DGID		|
				   SRP_OPT_PKEY		|
				   SRP_OPT_SERVICE_ID),
};

1780
static const match_table_t srp_opt_tokens = {
1781 1782 1783 1784 1785 1786 1787
	{ 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" 	},
1788
	{ SRP_OPT_IO_CLASS,		"io_class=%x"		},
1789
	{ SRP_OPT_INITIATOR_EXT,	"initiator_ext=%s"	},
1790
	{ SRP_OPT_ERR,			NULL 			}
1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818
};

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);
1819 1820 1821 1822
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1823 1824 1825 1826 1827 1828
			target->id_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_IOC_GUID:
			p = match_strdup(args);
1829 1830 1831 1832
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1833 1834 1835 1836 1837 1838
			target->ioc_guid = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_DGID:
			p = match_strdup(args);
1839 1840 1841 1842
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1843 1844
			if (strlen(p) != 32) {
				printk(KERN_WARNING PFX "bad dest GID parameter '%s'\n", p);
1845
				kfree(p);
1846 1847 1848 1849 1850 1851 1852
				goto out;
			}

			for (i = 0; i < 16; ++i) {
				strlcpy(dgid, p + i * 2, 3);
				target->path.dgid.raw[i] = simple_strtoul(dgid, NULL, 16);
			}
1853
			kfree(p);
1854
			memcpy(target->orig_dgid, target->path.dgid.raw, 16);
1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866
			break;

		case SRP_OPT_PKEY:
			if (match_hex(args, &token)) {
				printk(KERN_WARNING PFX "bad P_Key parameter '%s'\n", p);
				goto out;
			}
			target->path.pkey = cpu_to_be16(token);
			break;

		case SRP_OPT_SERVICE_ID:
			p = match_strdup(args);
1867 1868 1869 1870
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1871
			target->service_id = cpu_to_be64(simple_strtoull(p, NULL, 16));
1872
			target->path.service_id = target->service_id;
1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883
			kfree(p);
			break;

		case SRP_OPT_MAX_SECT:
			if (match_int(args, &token)) {
				printk(KERN_WARNING PFX "bad max sect parameter '%s'\n", p);
				goto out;
			}
			target->scsi_host->max_sectors = token;
			break;

1884 1885 1886 1887 1888
		case SRP_OPT_MAX_CMD_PER_LUN:
			if (match_int(args, &token)) {
				printk(KERN_WARNING PFX "bad max cmd_per_lun parameter '%s'\n", p);
				goto out;
			}
1889
			target->scsi_host->cmd_per_lun = min(token, SRP_CMD_SQ_SIZE);
1890 1891
			break;

1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906
		case SRP_OPT_IO_CLASS:
			if (match_hex(args, &token)) {
				printk(KERN_WARNING PFX "bad  IO class parameter '%s' \n", p);
				goto out;
			}
			if (token != SRP_REV10_IB_IO_CLASS &&
			    token != SRP_REV16A_IB_IO_CLASS) {
				printk(KERN_WARNING PFX "unknown IO class parameter value"
				       " %x specified (use %x or %x).\n",
				       token, SRP_REV10_IB_IO_CLASS, SRP_REV16A_IB_IO_CLASS);
				goto out;
			}
			target->io_class = token;
			break;

1907 1908
		case SRP_OPT_INITIATOR_EXT:
			p = match_strdup(args);
1909 1910 1911 1912
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1913 1914 1915 1916
			target->initiator_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938
		default:
			printk(KERN_WARNING PFX "unknown parameter or missing value "
			       "'%s' in target creation request\n", p);
			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))
				printk(KERN_WARNING PFX "target creation request is "
				       "missing parameter '%s'\n",
				       srp_opt_tokens[i].pattern);

out:
	kfree(options);
	return ret;
}

1939 1940
static ssize_t srp_create_target(struct device *dev,
				 struct device_attribute *attr,
1941 1942 1943
				 const char *buf, size_t count)
{
	struct srp_host *host =
1944
		container_of(dev, struct srp_host, dev);
1945 1946 1947 1948 1949 1950 1951 1952 1953 1954
	struct Scsi_Host *target_host;
	struct srp_target_port *target;
	int ret;
	int i;

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

1955
	target_host->transportt = ib_srp_transport_template;
A
Arne Redlich 已提交
1956 1957
	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
1958

1959 1960
	target = host_to_target(target_host);

1961
	target->io_class   = SRP_REV16A_IB_IO_CLASS;
1962 1963 1964
	target->scsi_host  = target_host;
	target->srp_host   = host;

1965
	INIT_LIST_HEAD(&target->free_reqs);
1966
	INIT_LIST_HEAD(&target->req_queue);
1967
	for (i = 0; i < SRP_CMD_SQ_SIZE; ++i) {
1968 1969 1970
		target->req_ring[i].index = i;
		list_add_tail(&target->req_ring[i].list, &target->free_reqs);
	}
1971 1972 1973 1974 1975

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

1976
	ib_query_gid(host->srp_dev->dev, host->port, 0, &target->path.sgid);
1977

1978 1979
	shost_printk(KERN_DEBUG, target->scsi_host, PFX
		     "new target: id_ext %016llx ioc_guid %016llx pkey %04x "
H
Harvey Harrison 已提交
1980
		     "service_id %016llx dgid %pI6\n",
1981 1982 1983 1984
	       (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),
1985
	       target->path.dgid.raw);
1986 1987 1988 1989 1990

	ret = srp_create_target_ib(target);
	if (ret)
		goto err;

D
David Dillow 已提交
1991 1992
	ret = srp_new_cm_id(target);
	if (ret)
1993 1994
		goto err_free;

1995
	target->qp_in_error = 0;
1996 1997
	ret = srp_connect_target(target);
	if (ret) {
1998 1999
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Connection failed\n");
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023
		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);

err_free:
	srp_free_target_ib(target);

err:
	scsi_host_put(target_host);

	return ret;
}

2024
static DEVICE_ATTR(add_target, S_IWUSR, NULL, srp_create_target);
2025

2026 2027
static ssize_t show_ibdev(struct device *dev, struct device_attribute *attr,
			  char *buf)
2028
{
2029
	struct srp_host *host = container_of(dev, struct srp_host, dev);
2030

2031
	return sprintf(buf, "%s\n", host->srp_dev->dev->name);
2032 2033
}

2034
static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
2035

2036 2037
static ssize_t show_port(struct device *dev, struct device_attribute *attr,
			 char *buf)
2038
{
2039
	struct srp_host *host = container_of(dev, struct srp_host, dev);
2040 2041 2042 2043

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

2044
static DEVICE_ATTR(port, S_IRUGO, show_port, NULL);
2045

2046
static struct srp_host *srp_add_port(struct srp_device *device, u8 port)
2047 2048 2049 2050 2051 2052 2053 2054
{
	struct srp_host *host;

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

	INIT_LIST_HEAD(&host->target_list);
2055
	spin_lock_init(&host->target_lock);
2056
	init_completion(&host->released);
2057
	host->srp_dev = device;
2058 2059
	host->port = port;

2060 2061
	host->dev.class = &srp_class;
	host->dev.parent = device->dev->dma_device;
2062
	dev_set_name(&host->dev, "srp-%s-%d", device->dev->name, port);
2063

2064
	if (device_register(&host->dev))
2065
		goto free_host;
2066
	if (device_create_file(&host->dev, &dev_attr_add_target))
2067
		goto err_class;
2068
	if (device_create_file(&host->dev, &dev_attr_ibdev))
2069
		goto err_class;
2070
	if (device_create_file(&host->dev, &dev_attr_port))
2071 2072 2073 2074 2075
		goto err_class;

	return host;

err_class:
2076
	device_unregister(&host->dev);
2077

2078
free_host:
2079 2080 2081 2082 2083 2084 2085
	kfree(host);

	return NULL;
}

static void srp_add_one(struct ib_device *device)
{
2086 2087 2088
	struct srp_device *srp_dev;
	struct ib_device_attr *dev_attr;
	struct ib_fmr_pool_param fmr_param;
2089 2090 2091
	struct srp_host *host;
	int s, e, p;

2092 2093
	dev_attr = kmalloc(sizeof *dev_attr, GFP_KERNEL);
	if (!dev_attr)
2094
		return;
2095

2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112
	if (ib_query_device(device, dev_attr)) {
		printk(KERN_WARNING PFX "Query device failed for %s\n",
		       device->name);
		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
	 * minimum of 512 bytes (which is the smallest sector that a
	 * SCSI command will ever carry).
	 */
	srp_dev->fmr_page_shift = max(9, ffs(dev_attr->page_size_cap) - 1);
	srp_dev->fmr_page_size  = 1 << srp_dev->fmr_page_shift;
2113
	srp_dev->fmr_page_mask  = ~((u64) srp_dev->fmr_page_size - 1);
2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141

	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;

	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 = SRP_FMR_SIZE;
	fmr_param.page_shift	    = srp_dev->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))
		srp_dev->fmr_pool = NULL;
2142

T
Tom Tucker 已提交
2143
	if (device->node_type == RDMA_NODE_IB_SWITCH) {
2144 2145 2146 2147 2148 2149 2150 2151
		s = 0;
		e = 0;
	} else {
		s = 1;
		e = device->phys_port_cnt;
	}

	for (p = s; p <= e; ++p) {
2152
		host = srp_add_port(srp_dev, p);
2153
		if (host)
2154
			list_add_tail(&host->list, &srp_dev->dev_list);
2155 2156
	}

2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168
	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);
2169 2170 2171 2172
}

static void srp_remove_one(struct ib_device *device)
{
2173
	struct srp_device *srp_dev;
2174 2175 2176 2177
	struct srp_host *host, *tmp_host;
	LIST_HEAD(target_list);
	struct srp_target_port *target, *tmp_target;

2178
	srp_dev = ib_get_client_data(device, &srp_client);
2179

2180
	list_for_each_entry_safe(host, tmp_host, &srp_dev->dev_list, list) {
2181
		device_unregister(&host->dev);
2182 2183 2184 2185 2186 2187 2188 2189 2190 2191
		/*
		 * Wait for the sysfs entry to go away, so that no new
		 * target ports can be created.
		 */
		wait_for_completion(&host->released);

		/*
		 * Mark all target ports as removed, so we stop queueing
		 * commands and don't try to reconnect.
		 */
2192
		spin_lock(&host->target_lock);
2193
		list_for_each_entry(target, &host->target_list, list) {
2194 2195 2196
			spin_lock_irq(target->scsi_host->host_lock);
			target->state = SRP_TARGET_REMOVED;
			spin_unlock_irq(target->scsi_host->host_lock);
2197
		}
2198
		spin_unlock(&host->target_lock);
2199 2200 2201 2202 2203 2204 2205 2206 2207 2208

		/*
		 * Wait for any reconnection tasks that may have
		 * started before we marked our target ports as
		 * removed, and any target port removal tasks.
		 */
		flush_scheduled_work();

		list_for_each_entry_safe(target, tmp_target,
					 &host->target_list, list) {
2209
			srp_remove_host(target->scsi_host);
2210
			scsi_remove_host(target->scsi_host);
2211 2212 2213 2214 2215 2216 2217 2218 2219
			srp_disconnect_target(target);
			ib_destroy_cm_id(target->cm_id);
			srp_free_target_ib(target);
			scsi_host_put(target->scsi_host);
		}

		kfree(host);
	}

2220 2221 2222 2223 2224 2225
	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);
2226 2227
}

2228 2229 2230
static struct srp_function_template ib_srp_transport_functions = {
};

2231 2232 2233 2234
static int __init srp_init_module(void)
{
	int ret;

2235 2236 2237
	BUILD_BUG_ON_NOT_POWER_OF_2(SRP_SQ_SIZE);
	BUILD_BUG_ON_NOT_POWER_OF_2(SRP_RQ_SIZE);

2238 2239 2240 2241 2242
	if (srp_sg_tablesize > 255) {
		printk(KERN_WARNING PFX "Clamping srp_sg_tablesize to 255\n");
		srp_sg_tablesize = 255;
	}

2243 2244 2245 2246 2247
	ib_srp_transport_template =
		srp_attach_transport(&ib_srp_transport_functions);
	if (!ib_srp_transport_template)
		return -ENOMEM;

2248 2249 2250 2251 2252
	srp_template.sg_tablesize = srp_sg_tablesize;
	srp_max_iu_len = (sizeof (struct srp_cmd) +
			  sizeof (struct srp_indirect_buf) +
			  srp_sg_tablesize * 16);

2253 2254 2255
	ret = class_register(&srp_class);
	if (ret) {
		printk(KERN_ERR PFX "couldn't register class infiniband_srp\n");
2256
		srp_release_transport(ib_srp_transport_template);
2257 2258 2259
		return ret;
	}

2260 2261
	ib_sa_register_client(&srp_sa_client);

2262 2263 2264
	ret = ib_register_client(&srp_client);
	if (ret) {
		printk(KERN_ERR PFX "couldn't register IB client\n");
2265
		srp_release_transport(ib_srp_transport_template);
2266
		ib_sa_unregister_client(&srp_sa_client);
2267 2268 2269 2270 2271 2272 2273 2274 2275 2276
		class_unregister(&srp_class);
		return ret;
	}

	return 0;
}

static void __exit srp_cleanup_module(void)
{
	ib_unregister_client(&srp_client);
2277
	ib_sa_unregister_client(&srp_sa_client);
2278
	class_unregister(&srp_class);
2279
	srp_release_transport(ib_srp_transport_template);
2280 2281 2282 2283
}

module_init(srp_init_module);
module_exit(srp_cleanup_module);