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);
}

444 445 446 447 448 449
static bool srp_change_state(struct srp_target_port *target,
			    enum srp_target_state old,
			    enum srp_target_state new)
{
	bool changed = false;

450
	spin_lock_irq(&target->lock);
451 452 453 454
	if (target->state == old) {
		target->state = new;
		changed = true;
	}
455
	spin_unlock_irq(&target->lock);
456 457 458
	return changed;
}

D
David Howells 已提交
459
static void srp_remove_work(struct work_struct *work)
460
{
D
David Howells 已提交
461 462
	struct srp_target_port *target =
		container_of(work, struct srp_target_port, work);
463

464
	if (!srp_change_state(target, SRP_TARGET_DEAD, SRP_TARGET_REMOVED))
465 466
		return;

467
	spin_lock(&target->srp_host->target_lock);
468
	list_del(&target->list);
469
	spin_unlock(&target->srp_host->target_lock);
470

471
	srp_remove_host(target->scsi_host);
472 473 474 475 476 477 478 479
	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
480
	int retries = 3;
481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512
	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
513 514 515 516 517 518 519 520 521 522 523 524 525 526 527
		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;

528 529 530 531 532 533
		default:
			return target->status;
		}
	}
}

534 535 536 537
static void srp_unmap_data(struct scsi_cmnd *scmnd,
			   struct srp_target_port *target,
			   struct srp_request *req)
{
538
	if (!scsi_sglist(scmnd) ||
539 540 541 542
	    (scmnd->sc_data_direction != DMA_TO_DEVICE &&
	     scmnd->sc_data_direction != DMA_FROM_DEVICE))
		return;

543 544 545 546 547
	if (req->fmr) {
		ib_fmr_pool_unmap(req->fmr);
		req->fmr = NULL;
	}

548
	ib_dma_unmap_sg(target->srp_host->srp_dev->dev, scsi_sglist(scmnd),
549
			scsi_sg_count(scmnd), scmnd->sc_data_direction);
550 551
}

552 553
static void srp_remove_req(struct srp_target_port *target,
			   struct srp_request *req, s32 req_lim_delta)
554
{
555 556
	unsigned long flags;

557
	srp_unmap_data(req->scmnd, target, req);
558
	spin_lock_irqsave(&target->lock, flags);
559
	target->req_lim += req_lim_delta;
560
	req->scmnd = NULL;
561
	list_add_tail(&req->list, &target->free_reqs);
562
	spin_unlock_irqrestore(&target->lock, flags);
563 564 565 566 567 568
}

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);
569
	srp_remove_req(target, req, 0);
570 571
}

572 573 574 575
static int srp_reconnect_target(struct srp_target_port *target)
{
	struct ib_qp_attr qp_attr;
	struct ib_wc wc;
576
	int i, ret;
577

578
	if (!srp_change_state(target, SRP_TARGET_LIVE, SRP_TARGET_CONNECTING))
579 580 581 582 583 584 585
		return -EAGAIN;

	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
586 587
	ret = srp_new_cm_id(target);
	if (ret)
588 589 590 591 592 593 594 595 596 597 598
		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;

599 600 601
	while (ib_poll_cq(target->recv_cq, 1, &wc) > 0)
		; /* nothing */
	while (ib_poll_cq(target->send_cq, 1, &wc) > 0)
602 603
		; /* nothing */

604 605 606 607 608
	for (i = 0; i < SRP_CMD_SQ_SIZE; ++i) {
		struct srp_request *req = &target->req_ring[i];
		if (req->scmnd)
			srp_reset_req(target, req);
	}
609

610
	INIT_LIST_HEAD(&target->free_tx);
611
	for (i = 0; i < SRP_SQ_SIZE; ++i)
612
		list_add(&target->tx_ring[i]->list, &target->free_tx);
613

614
	target->qp_in_error = 0;
615 616 617 618
	ret = srp_connect_target(target);
	if (ret)
		goto err;

619
	if (!srp_change_state(target, SRP_TARGET_CONNECTING, SRP_TARGET_LIVE))
620 621 622 623 624
		ret = -EAGAIN;

	return ret;

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

	/*
	 * We couldn't reconnect, so kill our target port off.
630 631 632 633 634 635
	 * However, we have to defer the real removal because we
	 * are in the context of the SCSI error handler now, which
	 * will deadlock if we call scsi_remove_host().
	 *
	 * Schedule our work inside the lock to avoid a race with
	 * the flush_scheduled_work() in srp_remove_one().
636
	 */
637
	spin_lock_irq(&target->lock);
638 639
	if (target->state == SRP_TARGET_CONNECTING) {
		target->state = SRP_TARGET_DEAD;
D
David Howells 已提交
640
		INIT_WORK(&target->work, srp_remove_work);
T
Tejun Heo 已提交
641
		queue_work(ib_wq, &target->work);
642
	}
643
	spin_unlock_irq(&target->lock);
644 645 646 647

	return ret;
}

648
static int srp_map_fmr(struct srp_target_port *target, struct scatterlist *scat,
649 650 651 652 653 654 655 656 657
		       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;
658
	struct srp_device *dev = target->srp_host->srp_dev;
659
	struct ib_device *ibdev = dev->dev;
660
	struct scatterlist *sg;
661 662 663 664

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

665 666
	if (srp_target_is_mellanox(target) &&
	    (ib_sg_dma_address(ibdev, &scat[0]) & ~dev->fmr_page_mask))
667 668
		return -EINVAL;

669
	len = page_cnt = 0;
670 671
	scsi_for_each_sg(req->scmnd, sg, sg_cnt, i) {
		unsigned int dma_len = ib_sg_dma_len(ibdev, sg);
672

673
		if (ib_sg_dma_address(ibdev, sg) & ~dev->fmr_page_mask) {
674 675 676 677 678
			if (i > 0)
				return -EINVAL;
			else
				++page_cnt;
		}
679
		if ((ib_sg_dma_address(ibdev, sg) + dma_len) &
680 681 682 683 684 685 686
		    ~dev->fmr_page_mask) {
			if (i < sg_cnt - 1)
				return -EINVAL;
			else
				++page_cnt;
		}

687
		len += dma_len;
688 689 690 691 692 693 694 695 696 697 698
	}

	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;
699 700
	scsi_for_each_sg(req->scmnd, sg, sg_cnt, i) {
		unsigned int dma_len = ib_sg_dma_len(ibdev, sg);
701 702

		for (j = 0; j < dma_len; j += dev->fmr_page_size)
703
			dma_pages[page_cnt++] =
704
				(ib_sg_dma_address(ibdev, sg) &
705 706
				 dev->fmr_page_mask) + j;
	}
707 708

	req->fmr = ib_fmr_pool_map_phys(dev->fmr_pool,
709
					dma_pages, page_cnt, io_addr);
710 711
	if (IS_ERR(req->fmr)) {
		ret = PTR_ERR(req->fmr);
V
Vu Pham 已提交
712
		req->fmr = NULL;
713 714 715
		goto out;
	}

716 717
	buf->va  = cpu_to_be64(ib_sg_dma_address(ibdev, &scat[0]) &
			       ~dev->fmr_page_mask);
718 719 720 721 722 723 724 725 726 727 728
	buf->key = cpu_to_be32(req->fmr->fmr->rkey);
	buf->len = cpu_to_be32(len);

	ret = 0;

out:
	kfree(dma_pages);

	return ret;
}

729 730 731
static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_target_port *target,
			struct srp_request *req)
{
732
	struct scatterlist *scat;
733
	struct srp_cmd *cmd = req->cmd->buf;
734
	int len, nents, count;
735
	u8 fmt = SRP_DATA_DESC_DIRECT;
736 737
	struct srp_device *dev;
	struct ib_device *ibdev;
738

739
	if (!scsi_sglist(scmnd) || scmnd->sc_data_direction == DMA_NONE)
740 741 742 743
		return sizeof (struct srp_cmd);

	if (scmnd->sc_data_direction != DMA_FROM_DEVICE &&
	    scmnd->sc_data_direction != DMA_TO_DEVICE) {
744 745 746
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled data direction %d\n",
			     scmnd->sc_data_direction);
747 748 749
		return -EINVAL;
	}

750 751
	nents = scsi_sg_count(scmnd);
	scat  = scsi_sglist(scmnd);
752

753
	dev = target->srp_host->srp_dev;
754 755 756
	ibdev = dev->dev;

	count = ib_dma_map_sg(ibdev, scat, nents, scmnd->sc_data_direction);
757 758 759

	fmt = SRP_DATA_DESC_DIRECT;
	len = sizeof (struct srp_cmd) +	sizeof (struct srp_direct_buf);
760

761
	if (count == 1) {
762 763 764 765 766 767
		/*
		 * 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.
		 */
768
		struct srp_direct_buf *buf = (void *) cmd->add_data;
769

770
		buf->va  = cpu_to_be64(ib_sg_dma_address(ibdev, scat));
771
		buf->key = cpu_to_be32(target->rkey);
772
		buf->len = cpu_to_be32(ib_sg_dma_len(ibdev, scat));
773
	} else if (srp_map_fmr(target, scat, count, req,
774 775 776 777 778 779
			       (void *) cmd->add_data)) {
		/*
		 * FMR mapping failed, and the scatterlist has more
		 * than one entry.  Generate an indirect memory
		 * descriptor.
		 */
780
		struct srp_indirect_buf *buf = (void *) cmd->add_data;
781
		struct scatterlist *sg;
782
		u32 datalen = 0;
783
		int i;
784

785
		fmt = SRP_DATA_DESC_INDIRECT;
786 787 788 789
		len = sizeof (struct srp_cmd) +
			sizeof (struct srp_indirect_buf) +
			count * sizeof (struct srp_direct_buf);

790 791
		scsi_for_each_sg(scmnd, sg, count, i) {
			unsigned int dma_len = ib_sg_dma_len(ibdev, sg);
792

793
			buf->desc_list[i].va  =
794
				cpu_to_be64(ib_sg_dma_address(ibdev, sg));
795
			buf->desc_list[i].key =
796
				cpu_to_be32(target->rkey);
797 798
			buf->desc_list[i].len = cpu_to_be32(dma_len);
			datalen += dma_len;
799
		}
800

801 802 803 804 805
		if (scmnd->sc_data_direction == DMA_TO_DEVICE)
			cmd->data_out_desc_cnt = count;
		else
			cmd->data_in_desc_cnt = count;

806 807
		buf->table_desc.va  =
			cpu_to_be64(req->cmd->dma + sizeof *cmd + sizeof *buf);
808
		buf->table_desc.key =
809
			cpu_to_be32(target->rkey);
810 811 812 813
		buf->table_desc.len =
			cpu_to_be32(count * sizeof (struct srp_direct_buf));

		buf->len = cpu_to_be32(datalen);
814 815 816 817 818 819 820 821 822 823
	}

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

	return len;
}

824 825 826 827 828 829 830 831
/*
 * 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;

832
	spin_lock_irqsave(&target->lock, flags);
833 834 835
	list_add(&iu->list, &target->free_tx);
	if (iu_type != SRP_IU_RSP)
		++target->req_lim;
836
	spin_unlock_irqrestore(&target->lock, flags);
837 838
}

839
/*
840 841
 * 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().
842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859
 *
 * 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);

860
	if (list_empty(&target->free_tx))
861 862 863
		return NULL;

	/* Initiator responses to target requests do not consume credits */
864 865 866 867 868 869 870
	if (iu_type != SRP_IU_RSP) {
		if (target->req_lim <= rsv) {
			++target->zero_req_lim;
			return NULL;
		}

		--target->req_lim;
871 872
	}

873
	iu = list_first_entry(&target->free_tx, struct srp_iu, list);
874
	list_del(&iu->list);
875 876 877
	return iu;
}

878 879
static int srp_post_send(struct srp_target_port *target,
			 struct srp_iu *iu, int len)
880 881 882 883 884 885
{
	struct ib_sge list;
	struct ib_send_wr wr, *bad_wr;

	list.addr   = iu->dma;
	list.length = len;
886
	list.lkey   = target->lkey;
887 888

	wr.next       = NULL;
889
	wr.wr_id      = (uintptr_t) iu;
890 891 892 893 894
	wr.sg_list    = &list;
	wr.num_sge    = 1;
	wr.opcode     = IB_WR_SEND;
	wr.send_flags = IB_SEND_SIGNALED;

895
	return ib_post_send(target->qp, &wr, &bad_wr);
896 897
}

898
static int srp_post_recv(struct srp_target_port *target, struct srp_iu *iu)
899 900
{
	struct ib_recv_wr wr, *bad_wr;
901
	struct ib_sge list;
902 903 904

	list.addr   = iu->dma;
	list.length = iu->size;
905
	list.lkey   = target->lkey;
906 907

	wr.next     = NULL;
908
	wr.wr_id    = (uintptr_t) iu;
909 910 911
	wr.sg_list  = &list;
	wr.num_sge  = 1;

912
	return ib_post_recv(target->qp, &wr, &bad_wr);
913 914
}

915 916 917 918 919 920 921
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)) {
922
		spin_lock_irqsave(&target->lock, flags);
923
		target->req_lim += be32_to_cpu(rsp->req_lim_delta);
924
		spin_unlock_irqrestore(&target->lock, flags);
925

926 927 928 929
		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);
930
	} else {
931
		req = &target->req_ring[rsp->tag];
932
		scmnd = req->scmnd;
933
		if (!scmnd)
934 935 936
			shost_printk(KERN_ERR, target->scsi_host,
				     "Null scmnd for RSP w/tag %016llx\n",
				     (unsigned long long) rsp->tag);
937 938 939 940 941 942 943 944 945 946
		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))
947
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_out_res_cnt));
948
		else if (rsp->flags & (SRP_RSP_FLAG_DIOVER | SRP_RSP_FLAG_DIUNDER))
949
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_in_res_cnt));
950

951
		srp_remove_req(target, req, be32_to_cpu(rsp->req_lim_delta));
952 953
		scmnd->host_scribble = NULL;
		scmnd->scsi_done(scmnd);
954 955 956
	}
}

957 958 959
static int srp_response_common(struct srp_target_port *target, s32 req_delta,
			       void *rsp, int len)
{
960
	struct ib_device *dev = target->srp_host->srp_dev->dev;
961 962
	unsigned long flags;
	struct srp_iu *iu;
963
	int err;
964

965
	spin_lock_irqsave(&target->lock, flags);
966 967
	target->req_lim += req_delta;
	iu = __srp_get_tx_iu(target, SRP_IU_RSP);
968
	spin_unlock_irqrestore(&target->lock, flags);
969

970 971 972
	if (!iu) {
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "no IU available to send response\n");
973
		return 1;
974 975 976 977 978 979
	}

	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);

980 981
	err = srp_post_send(target, iu, len);
	if (err) {
982 983
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "unable to post response: %d\n", err);
984 985
		srp_put_tx_iu(target, iu, SRP_IU_RSP);
	}
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

	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");
}

1021 1022
static void srp_handle_recv(struct srp_target_port *target, struct ib_wc *wc)
{
1023 1024
	struct ib_device *dev = target->srp_host->srp_dev->dev;
	struct srp_iu *iu = (struct srp_iu *) wc->wr_id;
1025
	int res;
1026 1027
	u8 opcode;

1028 1029
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_ti_iu_len,
				   DMA_FROM_DEVICE);
1030 1031 1032 1033

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

	if (0) {
1034 1035
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "recv completion, opcode 0x%02x\n", opcode);
B
Bart Van Assche 已提交
1036 1037
		print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET, 8, 1,
			       iu->buf, wc->byte_len, true);
1038 1039 1040 1041 1042 1043 1044
	}

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

1045 1046 1047 1048 1049 1050 1051 1052
	case SRP_CRED_REQ:
		srp_process_cred_req(target, iu->buf);
		break;

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

1053 1054
	case SRP_T_LOGOUT:
		/* XXX Handle target logout */
1055 1056
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Got target logout request\n");
1057 1058 1059
		break;

	default:
1060 1061
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled SRP opcode 0x%02x\n", opcode);
1062 1063 1064
		break;
	}

1065 1066
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_ti_iu_len,
				      DMA_FROM_DEVICE);
1067

1068
	res = srp_post_recv(target, iu);
1069 1070 1071
	if (res != 0)
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Recv failed with error code %d\n", res);
1072 1073
}

1074
static void srp_recv_completion(struct ib_cq *cq, void *target_ptr)
1075 1076 1077 1078 1079 1080 1081
{
	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) {
1082
			shost_printk(KERN_ERR, target->scsi_host,
1083
				     PFX "failed receive status %d\n",
1084
				     wc.status);
1085
			target->qp_in_error = 1;
1086 1087 1088
			break;
		}

1089 1090 1091 1092 1093 1094 1095 1096
		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;
1097
	struct srp_iu *iu;
1098 1099 1100 1101 1102 1103 1104 1105 1106 1107

	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;
		}

1108 1109
		iu = (struct srp_iu *) wc.wr_id;
		list_add(&iu->list, &target->free_tx);
1110 1111 1112
	}
}

1113
static int srp_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *scmnd)
1114
{
1115
	struct srp_target_port *target = host_to_target(shost);
1116 1117 1118
	struct srp_request *req;
	struct srp_iu *iu;
	struct srp_cmd *cmd;
1119
	struct ib_device *dev;
1120
	unsigned long flags;
1121 1122 1123 1124 1125 1126 1127 1128
	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;
1129
		scmnd->scsi_done(scmnd);
1130 1131 1132
		return 0;
	}

1133
	spin_lock_irqsave(&target->lock, flags);
1134
	iu = __srp_get_tx_iu(target, SRP_IU_CMD);
1135
	if (!iu)
1136 1137 1138 1139 1140
		goto err_unlock;

	req = list_first_entry(&target->free_reqs, struct srp_request, list);
	list_del(&req->list);
	spin_unlock_irqrestore(&target->lock, flags);
1141

1142
	dev = target->srp_host->srp_dev->dev;
1143 1144
	ib_dma_sync_single_for_cpu(dev, iu->dma, srp_max_iu_len,
				   DMA_TO_DEVICE);
1145 1146

	scmnd->result        = 0;
1147
	scmnd->host_scribble = (void *) req;
1148 1149 1150 1151 1152 1153

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

	cmd->opcode = SRP_CMD;
	cmd->lun    = cpu_to_be64((u64) scmnd->device->lun << 48);
1154
	cmd->tag    = req->index;
1155 1156 1157 1158 1159 1160 1161
	memcpy(cmd->cdb, scmnd->cmnd, scmnd->cmd_len);

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

	len = srp_map_data(scmnd, target, req);
	if (len < 0) {
1162 1163
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Failed to map data\n");
1164
		goto err_iu;
1165 1166
	}

1167 1168
	ib_dma_sync_single_for_device(dev, iu->dma, srp_max_iu_len,
				      DMA_TO_DEVICE);
1169

1170
	if (srp_post_send(target, iu, len)) {
1171
		shost_printk(KERN_ERR, target->scsi_host, PFX "Send failed\n");
1172 1173 1174 1175 1176 1177 1178 1179
		goto err_unmap;
	}

	return 0;

err_unmap:
	srp_unmap_data(scmnd, target, req);

1180 1181 1182
err_iu:
	srp_put_tx_iu(target, iu, SRP_IU_CMD);

1183
	spin_lock_irqsave(&target->lock, flags);
1184
	list_add(&req->list, &target->free_reqs);
1185 1186

err_unlock:
1187
	spin_unlock_irqrestore(&target->lock, flags);
1188

1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204
err:
	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;
	}

1205
	for (i = 0; i < SRP_SQ_SIZE; ++i) {
1206
		target->tx_ring[i] = srp_alloc_iu(target->srp_host,
1207
						  srp_max_iu_len,
1208 1209 1210
						  GFP_KERNEL, DMA_TO_DEVICE);
		if (!target->tx_ring[i])
			goto err;
1211 1212

		list_add(&target->tx_ring[i]->list, &target->free_tx);
1213 1214 1215 1216 1217 1218 1219 1220 1221 1222
	}

	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;
	}

1223
	for (i = 0; i < SRP_SQ_SIZE; ++i) {
1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234
		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)
{
1235
	struct Scsi_Host *shost = target->scsi_host;
1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251
	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:
1252
		if (srp_target_is_topspin(target)) {
1253 1254 1255 1256 1257 1258 1259 1260
			/*
			 * 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);

1261 1262 1263 1264
			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));
1265 1266 1267

			target->status = SRP_PORT_REDIRECT;
		} else {
1268 1269
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_PORT_REDIRECT\n");
1270 1271 1272 1273 1274
			target->status = -ECONNRESET;
		}
		break;

	case IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID:
1275 1276
		shost_printk(KERN_WARNING, shost,
			    "  REJ reason: IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID\n");
1277 1278 1279 1280 1281 1282 1283 1284 1285 1286
		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)
1287 1288
				shost_printk(KERN_WARNING, shost,
					     PFX "SRP_LOGIN_REJ: requested max_it_iu_len too large\n");
1289
			else
1290 1291
				shost_printk(KERN_WARNING, shost,
					    PFX "SRP LOGIN REJECTED, reason 0x%08x\n", reason);
1292
		} else
1293 1294 1295
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_CONSUMER_DEFINED,"
				     " opcode 0x%02x\n", opcode);
1296 1297 1298
		target->status = -ECONNRESET;
		break;

D
David Dillow 已提交
1299 1300 1301 1302 1303
	case IB_CM_REJ_STALE_CONN:
		shost_printk(KERN_WARNING, shost, "  REJ reason: stale connection\n");
		target->status = SRP_STALE_CONN;
		break;

1304
	default:
1305 1306
		shost_printk(KERN_WARNING, shost, "  REJ reason 0x%x\n",
			     event->param.rej_rcvd.reason);
1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317
		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;
1318
	int i;
1319 1320 1321

	switch (event->event) {
	case IB_CM_REQ_ERROR:
1322 1323
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Sending CM REQ failed\n");
1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337
		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);

1338 1339 1340 1341 1342 1343 1344
			/*
			 * 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);
1345
		} else {
1346 1347
			shost_printk(KERN_WARNING, target->scsi_host,
				    PFX "Unhandled RSP opcode %#x\n", opcode);
1348 1349 1350 1351
			target->status = -ECONNRESET;
			break;
		}

V
Vu Pham 已提交
1352 1353 1354 1355 1356
		if (!target->rx_ring[0]) {
			target->status = srp_alloc_iu_bufs(target);
			if (target->status)
				break;
		}
1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372

		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;

1373
		for (i = 0; i < SRP_RQ_SIZE; i++) {
1374 1375
			struct srp_iu *iu = target->rx_ring[i];
			target->status = srp_post_recv(target, iu);
1376 1377 1378
			if (target->status)
				break;
		}
1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397
		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:
1398
		shost_printk(KERN_DEBUG, target->scsi_host, PFX "REJ received\n");
1399 1400 1401 1402 1403
		comp = 1;

		srp_cm_rej_handler(cm_id, event, target);
		break;

1404
	case IB_CM_DREQ_RECEIVED:
1405 1406
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "DREQ received - connection closed\n");
1407
		if (ib_send_cm_drep(cm_id, NULL, 0))
1408 1409
			shost_printk(KERN_ERR, target->scsi_host,
				     PFX "Sending CM DREP failed\n");
1410 1411 1412
		break;

	case IB_CM_TIMEWAIT_EXIT:
1413 1414
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "connection closed\n");
1415 1416 1417 1418 1419

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

1420 1421 1422 1423 1424
	case IB_CM_MRA_RECEIVED:
	case IB_CM_DREQ_ERROR:
	case IB_CM_DREP_RECEIVED:
		break;

1425
	default:
1426 1427
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled CM event %d\n", event->event);
1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438
		break;
	}

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

	kfree(qp_attr);

	return 0;
}

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

1446
	if (target->state == SRP_TARGET_DEAD ||
1447
	    target->state == SRP_TARGET_REMOVED)
1448
		return -1;
1449

1450
	init_completion(&target->tsk_mgmt_done);
1451

1452
	spin_lock_irq(&target->lock);
1453
	iu = __srp_get_tx_iu(target, SRP_IU_TSK_MGMT);
1454
	spin_unlock_irq(&target->lock);
1455

1456
	if (!iu)
1457
		return -1;
1458

1459 1460
	ib_dma_sync_single_for_cpu(dev, iu->dma, sizeof *tsk_mgmt,
				   DMA_TO_DEVICE);
1461 1462 1463 1464
	tsk_mgmt = iu->buf;
	memset(tsk_mgmt, 0, sizeof *tsk_mgmt);

	tsk_mgmt->opcode 	= SRP_TSK_MGMT;
1465 1466
	tsk_mgmt->lun		= cpu_to_be64((u64) lun << 48);
	tsk_mgmt->tag		= req_tag | SRP_TAG_TSK_MGMT;
1467
	tsk_mgmt->tsk_mgmt_func = func;
1468
	tsk_mgmt->task_tag	= req_tag;
1469

1470 1471
	ib_dma_sync_single_for_device(dev, iu->dma, sizeof *tsk_mgmt,
				      DMA_TO_DEVICE);
1472 1473 1474 1475
	if (srp_post_send(target, iu, sizeof *tsk_mgmt)) {
		srp_put_tx_iu(target, iu, SRP_IU_TSK_MGMT);
		return -1;
	}
1476

1477
	if (!wait_for_completion_timeout(&target->tsk_mgmt_done,
1478
					 msecs_to_jiffies(SRP_ABORT_TIMEOUT_MS)))
1479
		return -1;
1480

1481 1482 1483
	return 0;
}

1484 1485
static int srp_abort(struct scsi_cmnd *scmnd)
{
1486
	struct srp_target_port *target = host_to_target(scmnd->device->host);
1487
	struct srp_request *req = (struct srp_request *) scmnd->host_scribble;
1488 1489
	int ret = SUCCESS;

1490
	shost_printk(KERN_ERR, target->scsi_host, "SRP abort called\n");
1491

1492
	if (!req || target->qp_in_error)
1493
		return FAILED;
1494 1495
	if (srp_send_tsk_mgmt(target, req->index, scmnd->device->lun,
			      SRP_TSK_ABORT_TASK))
1496 1497
		return FAILED;

1498 1499
	if (req->scmnd) {
		if (!target->tsk_mgmt_status) {
1500
			srp_remove_req(target, req, 0);
1501 1502 1503 1504
			scmnd->result = DID_ABORT << 16;
		} else
			ret = FAILED;
	}
1505 1506

	return ret;
1507 1508 1509 1510
}

static int srp_reset_device(struct scsi_cmnd *scmnd)
{
1511
	struct srp_target_port *target = host_to_target(scmnd->device->host);
1512
	int i;
1513

1514
	shost_printk(KERN_ERR, target->scsi_host, "SRP reset_device called\n");
1515

1516 1517
	if (target->qp_in_error)
		return FAILED;
1518 1519
	if (srp_send_tsk_mgmt(target, SRP_TAG_NO_REQ, scmnd->device->lun,
			      SRP_TSK_LUN_RESET))
1520
		return FAILED;
1521
	if (target->tsk_mgmt_status)
1522 1523
		return FAILED;

1524 1525
	for (i = 0; i < SRP_CMD_SQ_SIZE; ++i) {
		struct srp_request *req = &target->req_ring[i];
1526
		if (req->scmnd && req->scmnd->device == scmnd->device)
1527
			srp_reset_req(target, req);
1528
	}
1529 1530

	return SUCCESS;
1531 1532 1533 1534 1535 1536 1537
}

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

1538
	shost_printk(KERN_ERR, target->scsi_host, PFX "SRP reset_host called\n");
1539 1540 1541 1542 1543 1544 1545

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

	return ret;
}

1546 1547
static ssize_t show_id_ext(struct device *dev, struct device_attribute *attr,
			   char *buf)
1548
{
1549
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1550 1551 1552 1553 1554 1555 1556 1557 1558

	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));
}

1559 1560
static ssize_t show_ioc_guid(struct device *dev, struct device_attribute *attr,
			     char *buf)
1561
{
1562
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1563 1564 1565 1566 1567 1568 1569 1570 1571

	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));
}

1572 1573
static ssize_t show_service_id(struct device *dev,
			       struct device_attribute *attr, char *buf)
1574
{
1575
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1576 1577 1578 1579 1580 1581 1582 1583 1584

	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));
}

1585 1586
static ssize_t show_pkey(struct device *dev, struct device_attribute *attr,
			 char *buf)
1587
{
1588
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1589 1590 1591 1592 1593 1594 1595 1596

	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));
}

1597 1598
static ssize_t show_dgid(struct device *dev, struct device_attribute *attr,
			 char *buf)
1599
{
1600
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1601 1602 1603 1604 1605

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

H
Harvey Harrison 已提交
1606
	return sprintf(buf, "%pI6\n", target->path.dgid.raw);
1607 1608
}

1609 1610
static ssize_t show_orig_dgid(struct device *dev,
			      struct device_attribute *attr, char *buf)
1611
{
1612
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1613 1614 1615 1616 1617

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

H
Harvey Harrison 已提交
1618
	return sprintf(buf, "%pI6\n", target->orig_dgid);
1619 1620
}

1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632
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);
}

1633 1634
static ssize_t show_zero_req_lim(struct device *dev,
				 struct device_attribute *attr, char *buf)
1635
{
1636
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1637 1638 1639 1640 1641 1642 1643 1644

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

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

1645 1646
static ssize_t show_local_ib_port(struct device *dev,
				  struct device_attribute *attr, char *buf)
1647
{
1648
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1649 1650 1651 1652

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

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

1658
	return sprintf(buf, "%s\n", target->srp_host->srp_dev->dev->name);
1659 1660
}

1661 1662 1663 1664 1665 1666
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);
1667
static DEVICE_ATTR(req_lim,         S_IRUGO, show_req_lim,         NULL);
1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678
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,
1679
	&dev_attr_req_lim,
1680 1681 1682
	&dev_attr_zero_req_lim,
	&dev_attr_local_ib_port,
	&dev_attr_local_ib_device,
1683 1684 1685
	NULL
};

1686 1687
static struct scsi_host_template srp_template = {
	.module				= THIS_MODULE,
R
Roland Dreier 已提交
1688 1689
	.name				= "InfiniBand SRP initiator",
	.proc_name			= DRV_NAME,
1690 1691 1692 1693 1694
	.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,
1695
	.can_queue			= SRP_CMD_SQ_SIZE,
1696
	.this_id			= -1,
1697
	.cmd_per_lun			= SRP_CMD_SQ_SIZE,
1698 1699
	.use_clustering			= ENABLE_CLUSTERING,
	.shost_attrs			= srp_host_attrs
1700 1701 1702 1703
};

static int srp_add_target(struct srp_host *host, struct srp_target_port *target)
{
1704 1705 1706
	struct srp_rport_identifiers ids;
	struct srp_rport *rport;

1707 1708 1709
	sprintf(target->target_name, "SRP.T10:%016llX",
		 (unsigned long long) be64_to_cpu(target->id_ext));

1710
	if (scsi_add_host(target->scsi_host, host->srp_dev->dev->dma_device))
1711 1712
		return -ENODEV;

1713 1714
	memcpy(ids.port_id, &target->id_ext, 8);
	memcpy(ids.port_id + 8, &target->ioc_guid, 8);
1715
	ids.roles = SRP_RPORT_ROLE_TARGET;
1716 1717 1718 1719 1720 1721
	rport = srp_rport_add(target->scsi_host, &ids);
	if (IS_ERR(rport)) {
		scsi_remove_host(target->scsi_host);
		return PTR_ERR(rport);
	}

1722
	spin_lock(&host->target_lock);
1723
	list_add_tail(&target->list, &host->target_list);
1724
	spin_unlock(&host->target_lock);
1725 1726 1727 1728

	target->state = SRP_TARGET_LIVE;

	scsi_scan_target(&target->scsi_host->shost_gendev,
1729
			 0, target->scsi_id, SCAN_WILD_CARD, 0);
1730 1731 1732 1733

	return 0;
}

1734
static void srp_release_dev(struct device *dev)
1735 1736
{
	struct srp_host *host =
1737
		container_of(dev, struct srp_host, dev);
1738 1739 1740 1741 1742 1743

	complete(&host->released);
}

static struct class srp_class = {
	.name    = "infiniband_srp",
1744
	.dev_release = srp_release_dev
1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762
};

/*
 * 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,
1763
	SRP_OPT_MAX_CMD_PER_LUN	= 1 << 6,
1764
	SRP_OPT_IO_CLASS	= 1 << 7,
1765
	SRP_OPT_INITIATOR_EXT	= 1 << 8,
1766 1767 1768 1769 1770 1771 1772
	SRP_OPT_ALL		= (SRP_OPT_ID_EXT	|
				   SRP_OPT_IOC_GUID	|
				   SRP_OPT_DGID		|
				   SRP_OPT_PKEY		|
				   SRP_OPT_SERVICE_ID),
};

1773
static const match_table_t srp_opt_tokens = {
1774 1775 1776 1777 1778 1779 1780
	{ 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" 	},
1781
	{ SRP_OPT_IO_CLASS,		"io_class=%x"		},
1782
	{ SRP_OPT_INITIATOR_EXT,	"initiator_ext=%s"	},
1783
	{ SRP_OPT_ERR,			NULL 			}
1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811
};

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);
1812 1813 1814 1815
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1816 1817 1818 1819 1820 1821
			target->id_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_IOC_GUID:
			p = match_strdup(args);
1822 1823 1824 1825
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1826 1827 1828 1829 1830 1831
			target->ioc_guid = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_DGID:
			p = match_strdup(args);
1832 1833 1834 1835
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1836 1837
			if (strlen(p) != 32) {
				printk(KERN_WARNING PFX "bad dest GID parameter '%s'\n", p);
1838
				kfree(p);
1839 1840 1841 1842 1843 1844 1845
				goto out;
			}

			for (i = 0; i < 16; ++i) {
				strlcpy(dgid, p + i * 2, 3);
				target->path.dgid.raw[i] = simple_strtoul(dgid, NULL, 16);
			}
1846
			kfree(p);
1847
			memcpy(target->orig_dgid, target->path.dgid.raw, 16);
1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859
			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);
1860 1861 1862 1863
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1864
			target->service_id = cpu_to_be64(simple_strtoull(p, NULL, 16));
1865
			target->path.service_id = target->service_id;
1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876
			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;

1877 1878 1879 1880 1881
		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;
			}
1882
			target->scsi_host->cmd_per_lun = min(token, SRP_CMD_SQ_SIZE);
1883 1884
			break;

1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899
		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;

1900 1901
		case SRP_OPT_INITIATOR_EXT:
			p = match_strdup(args);
1902 1903 1904 1905
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1906 1907 1908 1909
			target->initiator_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931
		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;
}

1932 1933
static ssize_t srp_create_target(struct device *dev,
				 struct device_attribute *attr,
1934 1935 1936
				 const char *buf, size_t count)
{
	struct srp_host *host =
1937
		container_of(dev, struct srp_host, dev);
1938 1939 1940 1941 1942 1943 1944 1945 1946 1947
	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;

1948
	target_host->transportt = ib_srp_transport_template;
A
Arne Redlich 已提交
1949 1950
	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
1951

1952 1953
	target = host_to_target(target_host);

1954
	target->io_class   = SRP_REV16A_IB_IO_CLASS;
1955 1956
	target->scsi_host  = target_host;
	target->srp_host   = host;
1957 1958
	target->lkey	   = host->srp_dev->mr->lkey;
	target->rkey	   = host->srp_dev->mr->rkey;
1959

1960
	spin_lock_init(&target->lock);
1961
	INIT_LIST_HEAD(&target->free_tx);
1962
	INIT_LIST_HEAD(&target->free_reqs);
1963
	for (i = 0; i < SRP_CMD_SQ_SIZE; ++i) {
1964 1965 1966
		target->req_ring[i].index = i;
		list_add_tail(&target->req_ring[i].list, &target->free_reqs);
	}
1967 1968 1969 1970 1971

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

1972
	ib_query_gid(host->srp_dev->dev, host->port, 0, &target->path.sgid);
1973

1974 1975
	shost_printk(KERN_DEBUG, target->scsi_host, PFX
		     "new target: id_ext %016llx ioc_guid %016llx pkey %04x "
H
Harvey Harrison 已提交
1976
		     "service_id %016llx dgid %pI6\n",
1977 1978 1979 1980
	       (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),
1981
	       target->path.dgid.raw);
1982 1983 1984 1985 1986

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

D
David Dillow 已提交
1987 1988
	ret = srp_new_cm_id(target);
	if (ret)
1989 1990
		goto err_free;

1991
	target->qp_in_error = 0;
1992 1993
	ret = srp_connect_target(target);
	if (ret) {
1994 1995
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Connection failed\n");
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
		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;
}

2020
static DEVICE_ATTR(add_target, S_IWUSR, NULL, srp_create_target);
2021

2022 2023
static ssize_t show_ibdev(struct device *dev, struct device_attribute *attr,
			  char *buf)
2024
{
2025
	struct srp_host *host = container_of(dev, struct srp_host, dev);
2026

2027
	return sprintf(buf, "%s\n", host->srp_dev->dev->name);
2028 2029
}

2030
static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
2031

2032 2033
static ssize_t show_port(struct device *dev, struct device_attribute *attr,
			 char *buf)
2034
{
2035
	struct srp_host *host = container_of(dev, struct srp_host, dev);
2036 2037 2038 2039

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

2040
static DEVICE_ATTR(port, S_IRUGO, show_port, NULL);
2041

2042
static struct srp_host *srp_add_port(struct srp_device *device, u8 port)
2043 2044 2045 2046 2047 2048 2049 2050
{
	struct srp_host *host;

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

	INIT_LIST_HEAD(&host->target_list);
2051
	spin_lock_init(&host->target_lock);
2052
	init_completion(&host->released);
2053
	host->srp_dev = device;
2054 2055
	host->port = port;

2056 2057
	host->dev.class = &srp_class;
	host->dev.parent = device->dev->dma_device;
2058
	dev_set_name(&host->dev, "srp-%s-%d", device->dev->name, port);
2059

2060
	if (device_register(&host->dev))
2061
		goto free_host;
2062
	if (device_create_file(&host->dev, &dev_attr_add_target))
2063
		goto err_class;
2064
	if (device_create_file(&host->dev, &dev_attr_ibdev))
2065
		goto err_class;
2066
	if (device_create_file(&host->dev, &dev_attr_port))
2067 2068 2069 2070 2071
		goto err_class;

	return host;

err_class:
2072
	device_unregister(&host->dev);
2073

2074
free_host:
2075 2076 2077 2078 2079 2080 2081
	kfree(host);

	return NULL;
}

static void srp_add_one(struct ib_device *device)
{
2082 2083 2084
	struct srp_device *srp_dev;
	struct ib_device_attr *dev_attr;
	struct ib_fmr_pool_param fmr_param;
2085 2086 2087
	struct srp_host *host;
	int s, e, p;

2088 2089
	dev_attr = kmalloc(sizeof *dev_attr, GFP_KERNEL);
	if (!dev_attr)
2090
		return;
2091

2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108
	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;
2109
	srp_dev->fmr_page_mask  = ~((u64) srp_dev->fmr_page_size - 1);
2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137

	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;
2138

T
Tom Tucker 已提交
2139
	if (device->node_type == RDMA_NODE_IB_SWITCH) {
2140 2141 2142 2143 2144 2145 2146 2147
		s = 0;
		e = 0;
	} else {
		s = 1;
		e = device->phys_port_cnt;
	}

	for (p = s; p <= e; ++p) {
2148
		host = srp_add_port(srp_dev, p);
2149
		if (host)
2150
			list_add_tail(&host->list, &srp_dev->dev_list);
2151 2152
	}

2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164
	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);
2165 2166 2167 2168
}

static void srp_remove_one(struct ib_device *device)
{
2169
	struct srp_device *srp_dev;
2170 2171 2172 2173
	struct srp_host *host, *tmp_host;
	LIST_HEAD(target_list);
	struct srp_target_port *target, *tmp_target;

2174
	srp_dev = ib_get_client_data(device, &srp_client);
2175

2176
	list_for_each_entry_safe(host, tmp_host, &srp_dev->dev_list, list) {
2177
		device_unregister(&host->dev);
2178 2179 2180 2181 2182 2183 2184 2185 2186 2187
		/*
		 * 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.
		 */
2188
		spin_lock(&host->target_lock);
2189
		list_for_each_entry(target, &host->target_list, list) {
2190
			spin_lock_irq(&target->lock);
2191
			target->state = SRP_TARGET_REMOVED;
2192
			spin_unlock_irq(&target->lock);
2193
		}
2194
		spin_unlock(&host->target_lock);
2195 2196 2197 2198 2199 2200

		/*
		 * Wait for any reconnection tasks that may have
		 * started before we marked our target ports as
		 * removed, and any target port removal tasks.
		 */
T
Tejun Heo 已提交
2201
		flush_workqueue(ib_wq);
2202 2203 2204

		list_for_each_entry_safe(target, tmp_target,
					 &host->target_list, list) {
2205
			srp_remove_host(target->scsi_host);
2206
			scsi_remove_host(target->scsi_host);
2207 2208 2209 2210 2211 2212 2213 2214 2215
			srp_disconnect_target(target);
			ib_destroy_cm_id(target->cm_id);
			srp_free_target_ib(target);
			scsi_host_put(target->scsi_host);
		}

		kfree(host);
	}

2216 2217 2218 2219 2220 2221
	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);
2222 2223
}

2224 2225 2226
static struct srp_function_template ib_srp_transport_functions = {
};

2227 2228 2229 2230
static int __init srp_init_module(void)
{
	int ret;

2231
	BUILD_BUG_ON(FIELD_SIZEOF(struct ib_wc, wr_id) < sizeof(void *));
2232

2233 2234 2235 2236 2237
	if (srp_sg_tablesize > 255) {
		printk(KERN_WARNING PFX "Clamping srp_sg_tablesize to 255\n");
		srp_sg_tablesize = 255;
	}

2238 2239 2240 2241 2242
	ib_srp_transport_template =
		srp_attach_transport(&ib_srp_transport_functions);
	if (!ib_srp_transport_template)
		return -ENOMEM;

2243 2244 2245 2246 2247
	srp_template.sg_tablesize = srp_sg_tablesize;
	srp_max_iu_len = (sizeof (struct srp_cmd) +
			  sizeof (struct srp_indirect_buf) +
			  srp_sg_tablesize * 16);

2248 2249 2250
	ret = class_register(&srp_class);
	if (ret) {
		printk(KERN_ERR PFX "couldn't register class infiniband_srp\n");
2251
		srp_release_transport(ib_srp_transport_template);
2252 2253 2254
		return ret;
	}

2255 2256
	ib_sa_register_client(&srp_sa_client);

2257 2258 2259
	ret = ib_register_client(&srp_client);
	if (ret) {
		printk(KERN_ERR PFX "couldn't register IB client\n");
2260
		srp_release_transport(ib_srp_transport_template);
2261
		ib_sa_unregister_client(&srp_sa_client);
2262 2263 2264 2265 2266 2267 2268 2269 2270 2271
		class_unregister(&srp_class);
		return ret;
	}

	return 0;
}

static void __exit srp_cleanup_module(void)
{
	ib_unregister_client(&srp_client);
2272
	ib_sa_unregister_client(&srp_sa_client);
2273
	class_unregister(&srp_class);
2274
	srp_release_transport(ib_srp_transport_template);
2275 2276 2277 2278
}

module_init(srp_init_module);
module_exit(srp_cleanup_module);