ib_srp.c 61.6 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 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
static unsigned int srp_sg_tablesize;
static unsigned int cmd_sg_entries;
static int topspin_workarounds = 1;
65

66 67
module_param(srp_sg_tablesize, uint, 0444);
MODULE_PARM_DESC(srp_sg_tablesize, "Deprecated name for cmd_sg_entries");
68

69 70 71
module_param(cmd_sg_entries, uint, 0444);
MODULE_PARM_DESC(cmd_sg_entries,
		 "Default number of gather/scatter entries in the SRP command (default is 12, max 255)");
72 73 74 75 76 77 78

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

static void srp_add_one(struct ib_device *device);
static void srp_remove_one(struct ib_device *device);
79 80
static void srp_recv_completion(struct ib_cq *cq, void *target_ptr);
static void srp_send_completion(struct ib_cq *cq, void *target_ptr);
81 82
static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event);

83 84
static struct scsi_transport_template *ib_srp_transport_template;

85 86 87 88 89 90
static struct ib_client srp_client = {
	.name   = "srp",
	.add    = srp_add_one,
	.remove = srp_remove_one
};

91 92
static struct ib_sa_client srp_sa_client;

93 94 95 96 97 98 99 100 101 102
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;
}

103 104 105
static int srp_target_is_topspin(struct srp_target_port *target)
{
	static const u8 topspin_oui[3] = { 0x00, 0x05, 0xad };
106
	static const u8 cisco_oui[3]   = { 0x00, 0x1b, 0x0d };
107 108

	return topspin_workarounds &&
109 110
		(!memcmp(&target->ioc_guid, topspin_oui, sizeof topspin_oui) ||
		 !memcmp(&target->ioc_guid, cisco_oui, sizeof cisco_oui));
111 112
}

113 114 115 116 117 118 119 120 121 122 123 124 125 126
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;

127 128 129
	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))
130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149
		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;

150 151
	ib_dma_unmap_single(host->srp_dev->dev, iu->dma, iu->size,
			    iu->direction);
152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170
	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;

171 172 173 174
	ret = ib_find_pkey(target->srp_host->srp_dev->dev,
			   target->srp_host->port,
			   be16_to_cpu(target->path.pkey),
			   &attr->pkey_index);
175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193
	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
194 195 196 197
static int srp_new_cm_id(struct srp_target_port *target)
{
	struct ib_cm_id *new_cm_id;

198
	new_cm_id = ib_create_cm_id(target->srp_host->srp_dev->dev,
D
David Dillow 已提交
199 200 201 202 203 204 205 206 207 208 209
				    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;
}

210 211 212 213 214 215 216 217 218
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;

219 220 221 222
	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);
223
		goto err;
224 225
	}

226 227 228 229
	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);
230
		goto err_recv_cq;
231 232 233
	}

	ib_req_notify_cq(target->recv_cq, IB_CQ_NEXT_COMP);
234 235 236 237 238 239 240 241

	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;
242 243
	init_attr->send_cq             = target->send_cq;
	init_attr->recv_cq             = target->recv_cq;
244

245
	target->qp = ib_create_qp(target->srp_host->srp_dev->pd, init_attr);
246 247
	if (IS_ERR(target->qp)) {
		ret = PTR_ERR(target->qp);
248
		goto err_send_cq;
249 250 251
	}

	ret = srp_init_qp(target, target->qp);
252 253
	if (ret)
		goto err_qp;
254

255 256 257 258 259 260 261 262 263 264 265 266 267
	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:
268 269 270 271 272 273 274 275 276
	kfree(init_attr);
	return ret;
}

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

	ib_destroy_qp(target->qp);
277 278
	ib_destroy_cq(target->send_cq);
	ib_destroy_cq(target->recv_cq);
279 280 281

	for (i = 0; i < SRP_RQ_SIZE; ++i)
		srp_free_iu(target->srp_host, target->rx_ring[i]);
282
	for (i = 0; i < SRP_SQ_SIZE; ++i)
283 284 285 286 287 288 289 290 291 292 293
		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)
294 295
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Got failed path rec status %d\n", status);
296 297 298 299 300 301 302 303 304 305 306
	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);

307
	target->path_query_id = ib_sa_path_rec_get(&srp_sa_client,
308
						   target->srp_host->srp_dev->dev,
309 310
						   target->srp_host->port,
						   &target->path,
311
						   IB_SA_PATH_REC_SERVICE_ID	|
312 313 314 315 316 317 318 319 320 321 322 323 324 325
						   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)
326 327
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Path record query failed\n");
328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368

	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;
369
	req->priv.req_it_iu_len = cpu_to_be32(target->max_iu_len);
370 371
	req->priv.req_buf_fmt 	= cpu_to_be16(SRP_BUF_FORMAT_DIRECT |
					      SRP_BUF_FORMAT_INDIRECT);
372
	/*
R
Roland Dreier 已提交
373
	 * In the published SRP specification (draft rev. 16a), the
374 375 376 377 378 379 380 381 382
	 * 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,
383
		       &target->path.sgid.global.interface_id, 8);
384
		memcpy(req->priv.initiator_port_id + 8,
385
		       &target->initiator_ext, 8);
386 387 388 389
		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,
390 391 392
		       &target->initiator_ext, 8);
		memcpy(req->priv.initiator_port_id + 8,
		       &target->path.sgid.global.interface_id, 8);
393 394 395 396
		memcpy(req->priv.target_port_id,     &target->id_ext, 8);
		memcpy(req->priv.target_port_id + 8, &target->ioc_guid, 8);
	}

397 398
	/*
	 * Topspin/Cisco SRP targets will reject our login unless we
399 400
	 * zero out the first 8 bytes of our initiator port ID and set
	 * the second 8 bytes to the local node GUID.
401
	 */
402
	if (srp_target_is_topspin(target)) {
403 404 405 406
		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));
407
		memset(req->priv.initiator_port_id, 0, 8);
408
		memcpy(req->priv.initiator_port_id + 8,
409
		       &target->srp_host->srp_dev->dev->node_guid, 8);
410 411 412 413 414 415 416 417 418 419 420 421 422 423
	}

	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);
424
	if (ib_send_cm_dreq(target->cm_id, NULL, 0)) {
425 426
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Sending CM DREQ failed\n");
427 428
		return;
	}
429 430 431
	wait_for_completion(&target->done);
}

432 433 434 435 436 437
static bool srp_change_state(struct srp_target_port *target,
			    enum srp_target_state old,
			    enum srp_target_state new)
{
	bool changed = false;

438
	spin_lock_irq(&target->lock);
439 440 441 442
	if (target->state == old) {
		target->state = new;
		changed = true;
	}
443
	spin_unlock_irq(&target->lock);
444 445 446
	return changed;
}

447 448 449 450 451 452 453 454 455 456 457
static void srp_free_req_data(struct srp_target_port *target)
{
	struct srp_request *req;
	int i;

	for (i = 0, req = target->req_ring; i < SRP_CMD_SQ_SIZE; ++i, ++req) {
		kfree(req->fmr_list);
		kfree(req->map_page);
	}
}

D
David Howells 已提交
458
static void srp_remove_work(struct work_struct *work)
459
{
D
David Howells 已提交
460 461
	struct srp_target_port *target =
		container_of(work, struct srp_target_port, work);
462

463
	if (!srp_change_state(target, SRP_TARGET_DEAD, SRP_TARGET_REMOVED))
464 465
		return;

466
	spin_lock(&target->srp_host->target_lock);
467
	list_del(&target->list);
468
	spin_unlock(&target->srp_host->target_lock);
469

470
	srp_remove_host(target->scsi_host);
471 472 473
	scsi_remove_host(target->scsi_host);
	ib_destroy_cm_id(target->cm_id);
	srp_free_target_ib(target);
474
	srp_free_req_data(target);
475 476 477 478 479
	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)
{
538 539 540
	struct ib_device *ibdev = target->srp_host->srp_dev->dev;
	struct ib_pool_fmr **pfmr;

541
	if (!scsi_sglist(scmnd) ||
542 543 544 545
	    (scmnd->sc_data_direction != DMA_TO_DEVICE &&
	     scmnd->sc_data_direction != DMA_FROM_DEVICE))
		return;

546 547 548
	pfmr = req->fmr_list;
	while (req->nfmr--)
		ib_fmr_pool_unmap(*pfmr++);
549

550 551
	ib_dma_unmap_sg(ibdev, scsi_sglist(scmnd), scsi_sg_count(scmnd),
			scmnd->sc_data_direction);
552 553
}

554 555
static void srp_remove_req(struct srp_target_port *target,
			   struct srp_request *req, s32 req_lim_delta)
556
{
557 558
	unsigned long flags;

559
	srp_unmap_data(req->scmnd, target, req);
560
	spin_lock_irqsave(&target->lock, flags);
561
	target->req_lim += req_lim_delta;
562
	req->scmnd = NULL;
563
	list_add_tail(&req->list, &target->free_reqs);
564
	spin_unlock_irqrestore(&target->lock, flags);
565 566 567 568 569 570
}

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);
571
	srp_remove_req(target, req, 0);
572 573
}

574 575 576 577
static int srp_reconnect_target(struct srp_target_port *target)
{
	struct ib_qp_attr qp_attr;
	struct ib_wc wc;
578
	int i, ret;
579

580
	if (!srp_change_state(target, SRP_TARGET_LIVE, SRP_TARGET_CONNECTING))
581 582 583 584 585 586 587
		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
588 589
	ret = srp_new_cm_id(target);
	if (ret)
590 591 592 593 594 595 596 597 598 599 600
		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;

601 602 603
	while (ib_poll_cq(target->recv_cq, 1, &wc) > 0)
		; /* nothing */
	while (ib_poll_cq(target->send_cq, 1, &wc) > 0)
604 605
		; /* nothing */

606 607 608 609 610
	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);
	}
611

612
	INIT_LIST_HEAD(&target->free_tx);
613
	for (i = 0; i < SRP_SQ_SIZE; ++i)
614
		list_add(&target->tx_ring[i]->list, &target->free_tx);
615

616
	target->qp_in_error = 0;
617 618 619 620
	ret = srp_connect_target(target);
	if (ret)
		goto err;

621
	if (!srp_change_state(target, SRP_TARGET_CONNECTING, SRP_TARGET_LIVE))
622 623 624 625 626
		ret = -EAGAIN;

	return ret;

err:
627 628
	shost_printk(KERN_ERR, target->scsi_host,
		     PFX "reconnect failed (%d), removing target port.\n", ret);
629 630 631

	/*
	 * We couldn't reconnect, so kill our target port off.
632 633 634 635 636 637
	 * 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().
638
	 */
639
	spin_lock_irq(&target->lock);
640 641
	if (target->state == SRP_TARGET_CONNECTING) {
		target->state = SRP_TARGET_DEAD;
D
David Howells 已提交
642
		INIT_WORK(&target->work, srp_remove_work);
T
Tejun Heo 已提交
643
		queue_work(ib_wq, &target->work);
644
	}
645
	spin_unlock_irq(&target->lock);
646 647 648 649

	return ret;
}

650 651
static void srp_map_desc(struct srp_map_state *state, dma_addr_t dma_addr,
			 unsigned int dma_len, u32 rkey)
652
{
653
	struct srp_direct_buf *desc = state->desc;
654

655 656 657
	desc->va = cpu_to_be64(dma_addr);
	desc->key = cpu_to_be32(rkey);
	desc->len = cpu_to_be32(dma_len);
658

659 660 661 662
	state->total_len += dma_len;
	state->desc++;
	state->ndesc++;
}
663

664 665 666 667 668 669
static int srp_map_finish_fmr(struct srp_map_state *state,
			      struct srp_target_port *target)
{
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_pool_fmr *fmr;
	u64 io_addr = 0;
670

671 672
	if (!state->npages)
		return 0;
673

674 675 676 677 678
	if (state->npages == 1) {
		srp_map_desc(state, state->base_dma_addr, state->fmr_len,
			     target->rkey);
		state->npages = state->fmr_len = 0;
		return 0;
679 680
	}

681 682 683 684
	fmr = ib_fmr_pool_map_phys(dev->fmr_pool, state->pages,
				   state->npages, io_addr);
	if (IS_ERR(fmr))
		return PTR_ERR(fmr);
685

686 687
	*state->next_fmr++ = fmr;
	state->nfmr++;
688

689 690 691 692 693 694 695 696 697 698 699 700 701
	srp_map_desc(state, 0, state->fmr_len, fmr->fmr->rkey);
	state->npages = state->fmr_len = 0;
	return 0;
}

static void srp_map_update_start(struct srp_map_state *state,
				 struct scatterlist *sg, int sg_index,
				 dma_addr_t dma_addr)
{
	state->unmapped_sg = sg;
	state->unmapped_index = sg_index;
	state->unmapped_addr = dma_addr;
}
702

703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724
static int srp_map_sg_entry(struct srp_map_state *state,
			    struct srp_target_port *target,
			    struct scatterlist *sg, int sg_index,
			    int use_fmr)
{
	struct srp_device *dev = target->srp_host->srp_dev;
	struct ib_device *ibdev = dev->dev;
	dma_addr_t dma_addr = ib_sg_dma_address(ibdev, sg);
	unsigned int dma_len = ib_sg_dma_len(ibdev, sg);
	unsigned int len;
	int ret;

	if (!dma_len)
		return 0;

	if (use_fmr == SRP_MAP_NO_FMR) {
		/* Once we're in direct map mode for a request, we don't
		 * go back to FMR mode, so no need to update anything
		 * other than the descriptor.
		 */
		srp_map_desc(state, dma_addr, dma_len, target->rkey);
		return 0;
725
	}
726

727 728 729 730 731 732 733 734 735 736 737 738 739 740
	/* If we start at an offset into the FMR page, don't merge into
	 * the current FMR. Finish it out, and use the kernel's MR for this
	 * sg entry. This is to avoid potential bugs on some SRP targets
	 * that were never quite defined, but went away when the initiator
	 * avoided using FMR on such page fragments.
	 */
	if (dma_addr & ~dev->fmr_page_mask || dma_len > dev->fmr_max_size) {
		ret = srp_map_finish_fmr(state, target);
		if (ret)
			return ret;

		srp_map_desc(state, dma_addr, dma_len, target->rkey);
		srp_map_update_start(state, NULL, 0, 0);
		return 0;
741 742
	}

743 744 745 746 747 748 749
	/* If this is the first sg to go into the FMR, save our position.
	 * We need to know the first unmapped entry, its index, and the
	 * first unmapped address within that entry to be able to restart
	 * mapping after an error.
	 */
	if (!state->unmapped_sg)
		srp_map_update_start(state, sg, sg_index, dma_addr);
750

751 752 753 754 755
	while (dma_len) {
		if (state->npages == SRP_FMR_SIZE) {
			ret = srp_map_finish_fmr(state, target);
			if (ret)
				return ret;
756

757 758 759 760
			srp_map_update_start(state, sg, sg_index, dma_addr);
		}

		len = min_t(unsigned int, dma_len, dev->fmr_page_size);
761

762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779
		if (!state->npages)
			state->base_dma_addr = dma_addr;
		state->pages[state->npages++] = dma_addr;
		state->fmr_len += len;
		dma_addr += len;
		dma_len -= len;
	}

	/* If the last entry of the FMR wasn't a full page, then we need to
	 * close it out and start a new one -- we can only merge at page
	 * boundries.
	 */
	ret = 0;
	if (len != dev->fmr_page_size) {
		ret = srp_map_finish_fmr(state, target);
		if (!ret)
			srp_map_update_start(state, NULL, 0, 0);
	}
780 781 782
	return ret;
}

783 784 785
static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_target_port *target,
			struct srp_request *req)
{
786
	struct scatterlist *scat, *sg;
787
	struct srp_cmd *cmd = req->cmd->buf;
788
	int i, len, nents, count, use_fmr;
789 790
	struct srp_device *dev;
	struct ib_device *ibdev;
791 792 793 794 795
	struct srp_map_state state;
	struct srp_indirect_buf *indirect_hdr;
	dma_addr_t indirect_addr;
	u32 table_len;
	u8 fmt;
796

797
	if (!scsi_sglist(scmnd) || scmnd->sc_data_direction == DMA_NONE)
798 799 800 801
		return sizeof (struct srp_cmd);

	if (scmnd->sc_data_direction != DMA_FROM_DEVICE &&
	    scmnd->sc_data_direction != DMA_TO_DEVICE) {
802 803 804
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled data direction %d\n",
			     scmnd->sc_data_direction);
805 806 807
		return -EINVAL;
	}

808 809
	nents = scsi_sg_count(scmnd);
	scat  = scsi_sglist(scmnd);
810

811
	dev = target->srp_host->srp_dev;
812 813 814
	ibdev = dev->dev;

	count = ib_dma_map_sg(ibdev, scat, nents, scmnd->sc_data_direction);
815 816
	if (unlikely(count == 0))
		return -EIO;
817 818 819

	fmt = SRP_DATA_DESC_DIRECT;
	len = sizeof (struct srp_cmd) +	sizeof (struct srp_direct_buf);
820

821
	if (count == 1) {
822 823 824 825 826 827
		/*
		 * 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.
		 */
828
		struct srp_direct_buf *buf = (void *) cmd->add_data;
829

830
		buf->va  = cpu_to_be64(ib_sg_dma_address(ibdev, scat));
831
		buf->key = cpu_to_be32(target->rkey);
832
		buf->len = cpu_to_be32(ib_sg_dma_len(ibdev, scat));
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

		req->nfmr = 0;
		goto map_complete;
	}

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

	memset(&state, 0, sizeof(state));
	state.desc	= indirect_hdr->desc_list;
	state.pages	= req->map_page;
	state.next_fmr	= req->fmr_list;

	use_fmr = dev->fmr_pool ? SRP_MAP_ALLOW_FMR : SRP_MAP_NO_FMR;

	for_each_sg(scat, sg, count, i) {
		if (srp_map_sg_entry(&state, target, sg, i, use_fmr)) {
			/* FMR mapping failed, so backtrack to the first
			 * unmapped entry and continue on without using FMR.
			 */
			dma_addr_t dma_addr;
			unsigned int dma_len;

backtrack:
			sg = state.unmapped_sg;
			i = state.unmapped_index;

			dma_addr = ib_sg_dma_address(ibdev, sg);
			dma_len = ib_sg_dma_len(ibdev, sg);
			dma_len -= (state.unmapped_addr - dma_addr);
			dma_addr = state.unmapped_addr;
			use_fmr = SRP_MAP_NO_FMR;
			srp_map_desc(&state, dma_addr, dma_len, target->rkey);
869
		}
870
	}
871

872 873
	if (use_fmr == SRP_MAP_ALLOW_FMR && srp_map_finish_fmr(&state, target))
		goto backtrack;
874

875 876 877 878 879 880 881 882
	/* We've mapped the request, fill in the command buffer.
	 */
	req->nfmr = state.nfmr;
	if (state.ndesc == 1) {
		/* FMR mapping was able to collapse this to one entry,
		 * so use a direct descriptor.
		 */
		struct srp_direct_buf *buf = (void *) cmd->add_data;
883

884 885
		*buf = indirect_hdr->desc_list[0];
		goto map_complete;
886 887
	}

888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906
	table_len = state.ndesc * sizeof (struct srp_direct_buf);

	fmt = SRP_DATA_DESC_INDIRECT;
	len = sizeof(struct srp_cmd) + sizeof (struct srp_indirect_buf);
	len += table_len;

	indirect_addr = req->cmd->dma + sizeof *cmd + sizeof *indirect_hdr;

	indirect_hdr->table_desc.va = cpu_to_be64(indirect_addr);
	indirect_hdr->table_desc.key = cpu_to_be32(target->rkey);
	indirect_hdr->table_desc.len = cpu_to_be32(table_len);
	indirect_hdr->len = cpu_to_be32(state.total_len);

	if (scmnd->sc_data_direction == DMA_TO_DEVICE)
		cmd->data_out_desc_cnt = state.ndesc;
	else
		cmd->data_in_desc_cnt = state.ndesc;

map_complete:
907 908 909 910 911 912 913 914
	if (scmnd->sc_data_direction == DMA_TO_DEVICE)
		cmd->buf_fmt = fmt << 4;
	else
		cmd->buf_fmt = fmt;

	return len;
}

915 916 917 918 919 920 921 922
/*
 * 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;

923
	spin_lock_irqsave(&target->lock, flags);
924 925 926
	list_add(&iu->list, &target->free_tx);
	if (iu_type != SRP_IU_RSP)
		++target->req_lim;
927
	spin_unlock_irqrestore(&target->lock, flags);
928 929
}

930
/*
931 932
 * 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().
933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950
 *
 * 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);

951
	if (list_empty(&target->free_tx))
952 953 954
		return NULL;

	/* Initiator responses to target requests do not consume credits */
955 956 957 958 959 960 961
	if (iu_type != SRP_IU_RSP) {
		if (target->req_lim <= rsv) {
			++target->zero_req_lim;
			return NULL;
		}

		--target->req_lim;
962 963
	}

964
	iu = list_first_entry(&target->free_tx, struct srp_iu, list);
965
	list_del(&iu->list);
966 967 968
	return iu;
}

969 970
static int srp_post_send(struct srp_target_port *target,
			 struct srp_iu *iu, int len)
971 972 973 974 975 976
{
	struct ib_sge list;
	struct ib_send_wr wr, *bad_wr;

	list.addr   = iu->dma;
	list.length = len;
977
	list.lkey   = target->lkey;
978 979

	wr.next       = NULL;
980
	wr.wr_id      = (uintptr_t) iu;
981 982 983 984 985
	wr.sg_list    = &list;
	wr.num_sge    = 1;
	wr.opcode     = IB_WR_SEND;
	wr.send_flags = IB_SEND_SIGNALED;

986
	return ib_post_send(target->qp, &wr, &bad_wr);
987 988
}

989
static int srp_post_recv(struct srp_target_port *target, struct srp_iu *iu)
990 991
{
	struct ib_recv_wr wr, *bad_wr;
992
	struct ib_sge list;
993 994 995

	list.addr   = iu->dma;
	list.length = iu->size;
996
	list.lkey   = target->lkey;
997 998

	wr.next     = NULL;
999
	wr.wr_id    = (uintptr_t) iu;
1000 1001 1002
	wr.sg_list  = &list;
	wr.num_sge  = 1;

1003
	return ib_post_recv(target->qp, &wr, &bad_wr);
1004 1005
}

1006 1007 1008 1009 1010 1011 1012
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)) {
1013
		spin_lock_irqsave(&target->lock, flags);
1014
		target->req_lim += be32_to_cpu(rsp->req_lim_delta);
1015
		spin_unlock_irqrestore(&target->lock, flags);
1016

1017 1018 1019 1020
		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);
1021
	} else {
1022
		req = &target->req_ring[rsp->tag];
1023
		scmnd = req->scmnd;
1024
		if (!scmnd)
1025 1026 1027
			shost_printk(KERN_ERR, target->scsi_host,
				     "Null scmnd for RSP w/tag %016llx\n",
				     (unsigned long long) rsp->tag);
1028 1029 1030 1031 1032 1033 1034 1035 1036 1037
		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))
1038
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_out_res_cnt));
1039
		else if (rsp->flags & (SRP_RSP_FLAG_DIOVER | SRP_RSP_FLAG_DIUNDER))
1040
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_in_res_cnt));
1041

1042
		srp_remove_req(target, req, be32_to_cpu(rsp->req_lim_delta));
1043 1044
		scmnd->host_scribble = NULL;
		scmnd->scsi_done(scmnd);
1045 1046 1047
	}
}

1048 1049 1050
static int srp_response_common(struct srp_target_port *target, s32 req_delta,
			       void *rsp, int len)
{
1051
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1052 1053
	unsigned long flags;
	struct srp_iu *iu;
1054
	int err;
1055

1056
	spin_lock_irqsave(&target->lock, flags);
1057 1058
	target->req_lim += req_delta;
	iu = __srp_get_tx_iu(target, SRP_IU_RSP);
1059
	spin_unlock_irqrestore(&target->lock, flags);
1060

1061 1062 1063
	if (!iu) {
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "no IU available to send response\n");
1064
		return 1;
1065 1066 1067 1068 1069 1070
	}

	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);

1071 1072
	err = srp_post_send(target, iu, len);
	if (err) {
1073 1074
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "unable to post response: %d\n", err);
1075 1076
		srp_put_tx_iu(target, iu, SRP_IU_RSP);
	}
1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111

	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");
}

1112 1113
static void srp_handle_recv(struct srp_target_port *target, struct ib_wc *wc)
{
1114 1115
	struct ib_device *dev = target->srp_host->srp_dev->dev;
	struct srp_iu *iu = (struct srp_iu *) wc->wr_id;
1116
	int res;
1117 1118
	u8 opcode;

1119 1120
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_ti_iu_len,
				   DMA_FROM_DEVICE);
1121 1122 1123 1124

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

	if (0) {
1125 1126
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "recv completion, opcode 0x%02x\n", opcode);
B
Bart Van Assche 已提交
1127 1128
		print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET, 8, 1,
			       iu->buf, wc->byte_len, true);
1129 1130 1131 1132 1133 1134 1135
	}

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

1136 1137 1138 1139 1140 1141 1142 1143
	case SRP_CRED_REQ:
		srp_process_cred_req(target, iu->buf);
		break;

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

1144 1145
	case SRP_T_LOGOUT:
		/* XXX Handle target logout */
1146 1147
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Got target logout request\n");
1148 1149 1150
		break;

	default:
1151 1152
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled SRP opcode 0x%02x\n", opcode);
1153 1154 1155
		break;
	}

1156 1157
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_ti_iu_len,
				      DMA_FROM_DEVICE);
1158

1159
	res = srp_post_recv(target, iu);
1160 1161 1162
	if (res != 0)
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Recv failed with error code %d\n", res);
1163 1164
}

1165
static void srp_recv_completion(struct ib_cq *cq, void *target_ptr)
1166 1167 1168 1169 1170 1171 1172
{
	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) {
1173
			shost_printk(KERN_ERR, target->scsi_host,
1174
				     PFX "failed receive status %d\n",
1175
				     wc.status);
1176
			target->qp_in_error = 1;
1177 1178 1179
			break;
		}

1180 1181 1182 1183 1184 1185 1186 1187
		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;
1188
	struct srp_iu *iu;
1189 1190 1191 1192 1193 1194 1195 1196 1197 1198

	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;
		}

1199 1200
		iu = (struct srp_iu *) wc.wr_id;
		list_add(&iu->list, &target->free_tx);
1201 1202 1203
	}
}

1204
static int srp_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *scmnd)
1205
{
1206
	struct srp_target_port *target = host_to_target(shost);
1207 1208 1209
	struct srp_request *req;
	struct srp_iu *iu;
	struct srp_cmd *cmd;
1210
	struct ib_device *dev;
1211
	unsigned long flags;
1212 1213 1214 1215 1216 1217 1218 1219
	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;
1220
		scmnd->scsi_done(scmnd);
1221 1222 1223
		return 0;
	}

1224
	spin_lock_irqsave(&target->lock, flags);
1225
	iu = __srp_get_tx_iu(target, SRP_IU_CMD);
1226
	if (!iu)
1227 1228 1229 1230 1231
		goto err_unlock;

	req = list_first_entry(&target->free_reqs, struct srp_request, list);
	list_del(&req->list);
	spin_unlock_irqrestore(&target->lock, flags);
1232

1233
	dev = target->srp_host->srp_dev->dev;
1234
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_iu_len,
1235
				   DMA_TO_DEVICE);
1236 1237

	scmnd->result        = 0;
1238
	scmnd->host_scribble = (void *) req;
1239 1240 1241 1242 1243 1244

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

	cmd->opcode = SRP_CMD;
	cmd->lun    = cpu_to_be64((u64) scmnd->device->lun << 48);
1245
	cmd->tag    = req->index;
1246 1247 1248 1249 1250 1251 1252
	memcpy(cmd->cdb, scmnd->cmnd, scmnd->cmd_len);

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

	len = srp_map_data(scmnd, target, req);
	if (len < 0) {
1253 1254
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Failed to map data\n");
1255
		goto err_iu;
1256 1257
	}

1258
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_iu_len,
1259
				      DMA_TO_DEVICE);
1260

1261
	if (srp_post_send(target, iu, len)) {
1262
		shost_printk(KERN_ERR, target->scsi_host, PFX "Send failed\n");
1263 1264 1265 1266 1267 1268 1269 1270
		goto err_unmap;
	}

	return 0;

err_unmap:
	srp_unmap_data(scmnd, target, req);

1271 1272 1273
err_iu:
	srp_put_tx_iu(target, iu, SRP_IU_CMD);

1274
	spin_lock_irqsave(&target->lock, flags);
1275
	list_add(&req->list, &target->free_reqs);
1276 1277

err_unlock:
1278
	spin_unlock_irqrestore(&target->lock, flags);
1279

1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295
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;
	}

1296
	for (i = 0; i < SRP_SQ_SIZE; ++i) {
1297
		target->tx_ring[i] = srp_alloc_iu(target->srp_host,
1298
						  target->max_iu_len,
1299 1300 1301
						  GFP_KERNEL, DMA_TO_DEVICE);
		if (!target->tx_ring[i])
			goto err;
1302 1303

		list_add(&target->tx_ring[i]->list, &target->free_tx);
1304 1305 1306 1307 1308 1309 1310 1311 1312 1313
	}

	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;
	}

1314
	for (i = 0; i < SRP_SQ_SIZE; ++i) {
1315 1316 1317 1318 1319 1320 1321
		srp_free_iu(target->srp_host, target->tx_ring[i]);
		target->tx_ring[i] = NULL;
	}

	return -ENOMEM;
}

1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393
static void srp_cm_rep_handler(struct ib_cm_id *cm_id,
			       struct srp_login_rsp *lrsp,
			       struct srp_target_port *target)
{
	struct ib_qp_attr *qp_attr = NULL;
	int attr_mask = 0;
	int ret;
	int i;

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

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

	if (!target->rx_ring[0]) {
		ret = srp_alloc_iu_bufs(target);
		if (ret)
			goto error;
	}

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

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

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

	for (i = 0; i < SRP_RQ_SIZE; i++) {
		struct srp_iu *iu = target->rx_ring[i];
		ret = srp_post_recv(target, iu);
		if (ret)
			goto error_free;
	}

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

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

	ret = ib_send_cm_rtu(cm_id, NULL, 0);

error_free:
	kfree(qp_attr);

error:
	target->status = ret;
}

1394 1395 1396 1397
static void srp_cm_rej_handler(struct ib_cm_id *cm_id,
			       struct ib_cm_event *event,
			       struct srp_target_port *target)
{
1398
	struct Scsi_Host *shost = target->scsi_host;
1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414
	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:
1415
		if (srp_target_is_topspin(target)) {
1416 1417 1418 1419 1420 1421 1422 1423
			/*
			 * 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);

1424 1425 1426 1427
			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));
1428 1429 1430

			target->status = SRP_PORT_REDIRECT;
		} else {
1431 1432
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_PORT_REDIRECT\n");
1433 1434 1435 1436 1437
			target->status = -ECONNRESET;
		}
		break;

	case IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID:
1438 1439
		shost_printk(KERN_WARNING, shost,
			    "  REJ reason: IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID\n");
1440 1441 1442 1443 1444 1445 1446 1447 1448 1449
		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)
1450 1451
				shost_printk(KERN_WARNING, shost,
					     PFX "SRP_LOGIN_REJ: requested max_it_iu_len too large\n");
1452
			else
1453 1454
				shost_printk(KERN_WARNING, shost,
					    PFX "SRP LOGIN REJECTED, reason 0x%08x\n", reason);
1455
		} else
1456 1457 1458
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_CONSUMER_DEFINED,"
				     " opcode 0x%02x\n", opcode);
1459 1460 1461
		target->status = -ECONNRESET;
		break;

D
David Dillow 已提交
1462 1463 1464 1465 1466
	case IB_CM_REJ_STALE_CONN:
		shost_printk(KERN_WARNING, shost, "  REJ reason: stale connection\n");
		target->status = SRP_STALE_CONN;
		break;

1467
	default:
1468 1469
		shost_printk(KERN_WARNING, shost, "  REJ reason 0x%x\n",
			     event->param.rej_rcvd.reason);
1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480
		target->status = -ECONNRESET;
	}
}

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

	switch (event->event) {
	case IB_CM_REQ_ERROR:
1481 1482
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Sending CM REQ failed\n");
1483 1484 1485 1486 1487 1488
		comp = 1;
		target->status = -ECONNRESET;
		break;

	case IB_CM_REP_RECEIVED:
		comp = 1;
1489
		srp_cm_rep_handler(cm_id, event->private_data, target);
1490 1491 1492
		break;

	case IB_CM_REJ_RECEIVED:
1493
		shost_printk(KERN_DEBUG, target->scsi_host, PFX "REJ received\n");
1494 1495 1496 1497 1498
		comp = 1;

		srp_cm_rej_handler(cm_id, event, target);
		break;

1499
	case IB_CM_DREQ_RECEIVED:
1500 1501
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "DREQ received - connection closed\n");
1502
		if (ib_send_cm_drep(cm_id, NULL, 0))
1503 1504
			shost_printk(KERN_ERR, target->scsi_host,
				     PFX "Sending CM DREP failed\n");
1505 1506 1507
		break;

	case IB_CM_TIMEWAIT_EXIT:
1508 1509
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "connection closed\n");
1510 1511 1512 1513 1514

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

1515 1516 1517 1518 1519
	case IB_CM_MRA_RECEIVED:
	case IB_CM_DREQ_ERROR:
	case IB_CM_DREP_RECEIVED:
		break;

1520
	default:
1521 1522
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled CM event %d\n", event->event);
1523 1524 1525 1526 1527 1528 1529 1530 1531
		break;
	}

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

	return 0;
}

1532
static int srp_send_tsk_mgmt(struct srp_target_port *target,
1533
			     u64 req_tag, unsigned int lun, u8 func)
1534
{
1535
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1536 1537 1538
	struct srp_iu *iu;
	struct srp_tsk_mgmt *tsk_mgmt;

1539
	if (target->state == SRP_TARGET_DEAD ||
1540
	    target->state == SRP_TARGET_REMOVED)
1541
		return -1;
1542

1543
	init_completion(&target->tsk_mgmt_done);
1544

1545
	spin_lock_irq(&target->lock);
1546
	iu = __srp_get_tx_iu(target, SRP_IU_TSK_MGMT);
1547
	spin_unlock_irq(&target->lock);
1548

1549
	if (!iu)
1550
		return -1;
1551

1552 1553
	ib_dma_sync_single_for_cpu(dev, iu->dma, sizeof *tsk_mgmt,
				   DMA_TO_DEVICE);
1554 1555 1556 1557
	tsk_mgmt = iu->buf;
	memset(tsk_mgmt, 0, sizeof *tsk_mgmt);

	tsk_mgmt->opcode 	= SRP_TSK_MGMT;
1558 1559
	tsk_mgmt->lun		= cpu_to_be64((u64) lun << 48);
	tsk_mgmt->tag		= req_tag | SRP_TAG_TSK_MGMT;
1560
	tsk_mgmt->tsk_mgmt_func = func;
1561
	tsk_mgmt->task_tag	= req_tag;
1562

1563 1564
	ib_dma_sync_single_for_device(dev, iu->dma, sizeof *tsk_mgmt,
				      DMA_TO_DEVICE);
1565 1566 1567 1568
	if (srp_post_send(target, iu, sizeof *tsk_mgmt)) {
		srp_put_tx_iu(target, iu, SRP_IU_TSK_MGMT);
		return -1;
	}
1569

1570
	if (!wait_for_completion_timeout(&target->tsk_mgmt_done,
1571
					 msecs_to_jiffies(SRP_ABORT_TIMEOUT_MS)))
1572
		return -1;
1573

1574 1575 1576
	return 0;
}

1577 1578
static int srp_abort(struct scsi_cmnd *scmnd)
{
1579
	struct srp_target_port *target = host_to_target(scmnd->device->host);
1580
	struct srp_request *req = (struct srp_request *) scmnd->host_scribble;
1581 1582
	int ret = SUCCESS;

1583
	shost_printk(KERN_ERR, target->scsi_host, "SRP abort called\n");
1584

1585
	if (!req || target->qp_in_error)
1586
		return FAILED;
1587 1588
	if (srp_send_tsk_mgmt(target, req->index, scmnd->device->lun,
			      SRP_TSK_ABORT_TASK))
1589 1590
		return FAILED;

1591 1592
	if (req->scmnd) {
		if (!target->tsk_mgmt_status) {
1593
			srp_remove_req(target, req, 0);
1594 1595 1596 1597
			scmnd->result = DID_ABORT << 16;
		} else
			ret = FAILED;
	}
1598 1599

	return ret;
1600 1601 1602 1603
}

static int srp_reset_device(struct scsi_cmnd *scmnd)
{
1604
	struct srp_target_port *target = host_to_target(scmnd->device->host);
1605
	int i;
1606

1607
	shost_printk(KERN_ERR, target->scsi_host, "SRP reset_device called\n");
1608

1609 1610
	if (target->qp_in_error)
		return FAILED;
1611 1612
	if (srp_send_tsk_mgmt(target, SRP_TAG_NO_REQ, scmnd->device->lun,
			      SRP_TSK_LUN_RESET))
1613
		return FAILED;
1614
	if (target->tsk_mgmt_status)
1615 1616
		return FAILED;

1617 1618
	for (i = 0; i < SRP_CMD_SQ_SIZE; ++i) {
		struct srp_request *req = &target->req_ring[i];
1619
		if (req->scmnd && req->scmnd->device == scmnd->device)
1620
			srp_reset_req(target, req);
1621
	}
1622 1623

	return SUCCESS;
1624 1625 1626 1627 1628 1629 1630
}

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

1631
	shost_printk(KERN_ERR, target->scsi_host, PFX "SRP reset_host called\n");
1632 1633 1634 1635 1636 1637 1638

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

	return ret;
}

1639 1640
static ssize_t show_id_ext(struct device *dev, struct device_attribute *attr,
			   char *buf)
1641
{
1642
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1643 1644 1645 1646 1647 1648 1649 1650 1651

	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));
}

1652 1653
static ssize_t show_ioc_guid(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 1660 1661 1662 1663 1664

	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));
}

1665 1666
static ssize_t show_service_id(struct device *dev,
			       struct device_attribute *attr, char *buf)
1667
{
1668
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1669 1670 1671 1672 1673 1674 1675 1676 1677

	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));
}

1678 1679
static ssize_t show_pkey(struct device *dev, struct device_attribute *attr,
			 char *buf)
1680
{
1681
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1682 1683 1684 1685 1686 1687 1688 1689

	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));
}

1690 1691
static ssize_t show_dgid(struct device *dev, struct device_attribute *attr,
			 char *buf)
1692
{
1693
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1694 1695 1696 1697 1698

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

H
Harvey Harrison 已提交
1699
	return sprintf(buf, "%pI6\n", target->path.dgid.raw);
1700 1701
}

1702 1703
static ssize_t show_orig_dgid(struct device *dev,
			      struct device_attribute *attr, char *buf)
1704
{
1705
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1706 1707 1708 1709 1710

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

H
Harvey Harrison 已提交
1711
	return sprintf(buf, "%pI6\n", target->orig_dgid);
1712 1713
}

1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725
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);
}

1726 1727
static ssize_t show_zero_req_lim(struct device *dev,
				 struct device_attribute *attr, char *buf)
1728
{
1729
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1730 1731 1732 1733 1734 1735 1736 1737

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

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

1738 1739
static ssize_t show_local_ib_port(struct device *dev,
				  struct device_attribute *attr, char *buf)
1740
{
1741
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1742 1743 1744 1745

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

1746 1747
static ssize_t show_local_ib_device(struct device *dev,
				    struct device_attribute *attr, char *buf)
1748
{
1749
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1750

1751
	return sprintf(buf, "%s\n", target->srp_host->srp_dev->dev->name);
1752 1753
}

1754 1755 1756 1757 1758 1759 1760 1761
static ssize_t show_cmd_sg_entries(struct device *dev,
				   struct device_attribute *attr, char *buf)
{
	struct srp_target_port *target = host_to_target(class_to_shost(dev));

	return sprintf(buf, "%u\n", target->cmd_sg_cnt);
}

1762 1763 1764 1765 1766 1767
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);
1768
static DEVICE_ATTR(req_lim,         S_IRUGO, show_req_lim,         NULL);
1769 1770 1771
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);
1772
static DEVICE_ATTR(cmd_sg_entries,  S_IRUGO, show_cmd_sg_entries,  NULL);
1773 1774 1775 1776 1777 1778 1779 1780

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,
1781
	&dev_attr_req_lim,
1782 1783 1784
	&dev_attr_zero_req_lim,
	&dev_attr_local_ib_port,
	&dev_attr_local_ib_device,
1785
	&dev_attr_cmd_sg_entries,
1786 1787 1788
	NULL
};

1789 1790
static struct scsi_host_template srp_template = {
	.module				= THIS_MODULE,
R
Roland Dreier 已提交
1791 1792
	.name				= "InfiniBand SRP initiator",
	.proc_name			= DRV_NAME,
1793 1794 1795 1796 1797
	.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,
1798
	.sg_tablesize			= SRP_DEF_SG_TABLESIZE,
1799
	.can_queue			= SRP_CMD_SQ_SIZE,
1800
	.this_id			= -1,
1801
	.cmd_per_lun			= SRP_CMD_SQ_SIZE,
1802 1803
	.use_clustering			= ENABLE_CLUSTERING,
	.shost_attrs			= srp_host_attrs
1804 1805 1806 1807
};

static int srp_add_target(struct srp_host *host, struct srp_target_port *target)
{
1808 1809 1810
	struct srp_rport_identifiers ids;
	struct srp_rport *rport;

1811 1812 1813
	sprintf(target->target_name, "SRP.T10:%016llX",
		 (unsigned long long) be64_to_cpu(target->id_ext));

1814
	if (scsi_add_host(target->scsi_host, host->srp_dev->dev->dma_device))
1815 1816
		return -ENODEV;

1817 1818
	memcpy(ids.port_id, &target->id_ext, 8);
	memcpy(ids.port_id + 8, &target->ioc_guid, 8);
1819
	ids.roles = SRP_RPORT_ROLE_TARGET;
1820 1821 1822 1823 1824 1825
	rport = srp_rport_add(target->scsi_host, &ids);
	if (IS_ERR(rport)) {
		scsi_remove_host(target->scsi_host);
		return PTR_ERR(rport);
	}

1826
	spin_lock(&host->target_lock);
1827
	list_add_tail(&target->list, &host->target_list);
1828
	spin_unlock(&host->target_lock);
1829 1830 1831 1832

	target->state = SRP_TARGET_LIVE;

	scsi_scan_target(&target->scsi_host->shost_gendev,
1833
			 0, target->scsi_id, SCAN_WILD_CARD, 0);
1834 1835 1836 1837

	return 0;
}

1838
static void srp_release_dev(struct device *dev)
1839 1840
{
	struct srp_host *host =
1841
		container_of(dev, struct srp_host, dev);
1842 1843 1844 1845 1846 1847

	complete(&host->released);
}

static struct class srp_class = {
	.name    = "infiniband_srp",
1848
	.dev_release = srp_release_dev
1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866
};

/*
 * 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,
1867
	SRP_OPT_MAX_CMD_PER_LUN	= 1 << 6,
1868
	SRP_OPT_IO_CLASS	= 1 << 7,
1869
	SRP_OPT_INITIATOR_EXT	= 1 << 8,
1870
	SRP_OPT_CMD_SG_ENTRIES	= 1 << 9,
1871 1872 1873 1874 1875 1876 1877
	SRP_OPT_ALL		= (SRP_OPT_ID_EXT	|
				   SRP_OPT_IOC_GUID	|
				   SRP_OPT_DGID		|
				   SRP_OPT_PKEY		|
				   SRP_OPT_SERVICE_ID),
};

1878
static const match_table_t srp_opt_tokens = {
1879 1880 1881 1882 1883 1884 1885
	{ 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" 	},
1886
	{ SRP_OPT_IO_CLASS,		"io_class=%x"		},
1887
	{ SRP_OPT_INITIATOR_EXT,	"initiator_ext=%s"	},
1888
	{ SRP_OPT_CMD_SG_ENTRIES,	"cmd_sg_entries=%u"	},
1889
	{ SRP_OPT_ERR,			NULL 			}
1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917
};

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);
1918 1919 1920 1921
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1922 1923 1924 1925 1926 1927
			target->id_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_IOC_GUID:
			p = match_strdup(args);
1928 1929 1930 1931
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1932 1933 1934 1935 1936 1937
			target->ioc_guid = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_DGID:
			p = match_strdup(args);
1938 1939 1940 1941
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1942 1943
			if (strlen(p) != 32) {
				printk(KERN_WARNING PFX "bad dest GID parameter '%s'\n", p);
1944
				kfree(p);
1945 1946 1947 1948 1949 1950 1951
				goto out;
			}

			for (i = 0; i < 16; ++i) {
				strlcpy(dgid, p + i * 2, 3);
				target->path.dgid.raw[i] = simple_strtoul(dgid, NULL, 16);
			}
1952
			kfree(p);
1953
			memcpy(target->orig_dgid, target->path.dgid.raw, 16);
1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965
			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);
1966 1967 1968 1969
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1970
			target->service_id = cpu_to_be64(simple_strtoull(p, NULL, 16));
1971
			target->path.service_id = target->service_id;
1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982
			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;

1983 1984 1985 1986 1987
		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;
			}
1988
			target->scsi_host->cmd_per_lun = min(token, SRP_CMD_SQ_SIZE);
1989 1990
			break;

1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
		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;

2006 2007
		case SRP_OPT_INITIATOR_EXT:
			p = match_strdup(args);
2008 2009 2010 2011
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
2012 2013 2014 2015
			target->initiator_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

2016 2017 2018 2019 2020 2021 2022 2023
		case SRP_OPT_CMD_SG_ENTRIES:
			if (match_int(args, &token) || token < 1 || token > 255) {
				printk(KERN_WARNING PFX "bad max cmd_sg_entries parameter '%s'\n", p);
				goto out;
			}
			target->cmd_sg_cnt = token;
			break;

2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045
		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;
}

2046 2047
static ssize_t srp_create_target(struct device *dev,
				 struct device_attribute *attr,
2048 2049 2050
				 const char *buf, size_t count)
{
	struct srp_host *host =
2051
		container_of(dev, struct srp_host, dev);
2052 2053
	struct Scsi_Host *target_host;
	struct srp_target_port *target;
2054
	int i, ret;
2055 2056 2057 2058 2059 2060

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

2061
	target_host->transportt  = ib_srp_transport_template;
A
Arne Redlich 已提交
2062 2063
	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
2064

2065 2066
	target = host_to_target(target_host);

2067 2068 2069 2070 2071 2072
	target->io_class	= SRP_REV16A_IB_IO_CLASS;
	target->scsi_host	= target_host;
	target->srp_host	= host;
	target->lkey		= host->srp_dev->mr->lkey;
	target->rkey		= host->srp_dev->mr->rkey;
	target->cmd_sg_cnt	= cmd_sg_entries;
2073 2074 2075 2076 2077

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

2078 2079 2080 2081 2082
	target_host->sg_tablesize = target->cmd_sg_cnt;
	target->max_iu_len = sizeof (struct srp_cmd) +
			     sizeof (struct srp_indirect_buf) +
			     target->cmd_sg_cnt * sizeof (struct srp_direct_buf);

2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099
	spin_lock_init(&target->lock);
	INIT_LIST_HEAD(&target->free_tx);
	INIT_LIST_HEAD(&target->free_reqs);
	for (i = 0; i < SRP_CMD_SQ_SIZE; ++i) {
		struct srp_request *req = &target->req_ring[i];

		req->fmr_list = kmalloc(target->cmd_sg_cnt * sizeof (void *),
					GFP_KERNEL);
		req->map_page = kmalloc(SRP_FMR_SIZE * sizeof (void *),
					GFP_KERNEL);
		if (!req->fmr_list || !req->map_page)
			goto err_free_mem;

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

2100
	ib_query_gid(host->srp_dev->dev, host->port, 0, &target->path.sgid);
2101

2102 2103
	shost_printk(KERN_DEBUG, target->scsi_host, PFX
		     "new target: id_ext %016llx ioc_guid %016llx pkey %04x "
H
Harvey Harrison 已提交
2104
		     "service_id %016llx dgid %pI6\n",
2105 2106 2107 2108
	       (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),
2109
	       target->path.dgid.raw);
2110 2111 2112

	ret = srp_create_target_ib(target);
	if (ret)
2113
		goto err_free_mem;
2114

D
David Dillow 已提交
2115 2116
	ret = srp_new_cm_id(target);
	if (ret)
2117
		goto err_free_ib;
2118

2119
	target->qp_in_error = 0;
2120 2121
	ret = srp_connect_target(target);
	if (ret) {
2122 2123
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Connection failed\n");
2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138
		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);

2139
err_free_ib:
2140 2141
	srp_free_target_ib(target);

2142 2143 2144
err_free_mem:
	srp_free_req_data(target);

2145 2146 2147 2148 2149 2150
err:
	scsi_host_put(target_host);

	return ret;
}

2151
static DEVICE_ATTR(add_target, S_IWUSR, NULL, srp_create_target);
2152

2153 2154
static ssize_t show_ibdev(struct device *dev, struct device_attribute *attr,
			  char *buf)
2155
{
2156
	struct srp_host *host = container_of(dev, struct srp_host, dev);
2157

2158
	return sprintf(buf, "%s\n", host->srp_dev->dev->name);
2159 2160
}

2161
static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
2162

2163 2164
static ssize_t show_port(struct device *dev, struct device_attribute *attr,
			 char *buf)
2165
{
2166
	struct srp_host *host = container_of(dev, struct srp_host, dev);
2167 2168 2169 2170

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

2171
static DEVICE_ATTR(port, S_IRUGO, show_port, NULL);
2172

2173
static struct srp_host *srp_add_port(struct srp_device *device, u8 port)
2174 2175 2176 2177 2178 2179 2180 2181
{
	struct srp_host *host;

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

	INIT_LIST_HEAD(&host->target_list);
2182
	spin_lock_init(&host->target_lock);
2183
	init_completion(&host->released);
2184
	host->srp_dev = device;
2185 2186
	host->port = port;

2187 2188
	host->dev.class = &srp_class;
	host->dev.parent = device->dev->dma_device;
2189
	dev_set_name(&host->dev, "srp-%s-%d", device->dev->name, port);
2190

2191
	if (device_register(&host->dev))
2192
		goto free_host;
2193
	if (device_create_file(&host->dev, &dev_attr_add_target))
2194
		goto err_class;
2195
	if (device_create_file(&host->dev, &dev_attr_ibdev))
2196
		goto err_class;
2197
	if (device_create_file(&host->dev, &dev_attr_port))
2198 2199 2200 2201 2202
		goto err_class;

	return host;

err_class:
2203
	device_unregister(&host->dev);
2204

2205
free_host:
2206 2207 2208 2209 2210 2211 2212
	kfree(host);

	return NULL;
}

static void srp_add_one(struct ib_device *device)
{
2213 2214 2215
	struct srp_device *srp_dev;
	struct ib_device_attr *dev_attr;
	struct ib_fmr_pool_param fmr_param;
2216
	struct srp_host *host;
2217
	int fmr_page_shift, s, e, p;
2218

2219 2220
	dev_attr = kmalloc(sizeof *dev_attr, GFP_KERNEL);
	if (!dev_attr)
2221
		return;
2222

2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234
	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
2235 2236
	 * minimum of 4096 bytes. We're unlikely to build large sglists
	 * out of smaller entries.
2237
	 */
2238 2239 2240 2241
	fmr_page_shift		= max(12, ffs(dev_attr->page_size_cap) - 1);
	srp_dev->fmr_page_size	= 1 << fmr_page_shift;
	srp_dev->fmr_page_mask	= ~((u64) srp_dev->fmr_page_size - 1);
	srp_dev->fmr_max_size	= srp_dev->fmr_page_size * SRP_FMR_SIZE;
2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261

	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;
2262
	fmr_param.page_shift	    = fmr_page_shift;
2263 2264 2265 2266 2267 2268 2269
	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;
2270

T
Tom Tucker 已提交
2271
	if (device->node_type == RDMA_NODE_IB_SWITCH) {
2272 2273 2274 2275 2276 2277 2278 2279
		s = 0;
		e = 0;
	} else {
		s = 1;
		e = device->phys_port_cnt;
	}

	for (p = s; p <= e; ++p) {
2280
		host = srp_add_port(srp_dev, p);
2281
		if (host)
2282
			list_add_tail(&host->list, &srp_dev->dev_list);
2283 2284
	}

2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296
	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);
2297 2298 2299 2300
}

static void srp_remove_one(struct ib_device *device)
{
2301
	struct srp_device *srp_dev;
2302 2303 2304 2305
	struct srp_host *host, *tmp_host;
	LIST_HEAD(target_list);
	struct srp_target_port *target, *tmp_target;

2306
	srp_dev = ib_get_client_data(device, &srp_client);
2307

2308
	list_for_each_entry_safe(host, tmp_host, &srp_dev->dev_list, list) {
2309
		device_unregister(&host->dev);
2310 2311 2312 2313 2314 2315 2316 2317 2318 2319
		/*
		 * 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.
		 */
2320
		spin_lock(&host->target_lock);
2321
		list_for_each_entry(target, &host->target_list, list) {
2322
			spin_lock_irq(&target->lock);
2323
			target->state = SRP_TARGET_REMOVED;
2324
			spin_unlock_irq(&target->lock);
2325
		}
2326
		spin_unlock(&host->target_lock);
2327 2328 2329 2330 2331 2332

		/*
		 * 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
2333
		flush_workqueue(ib_wq);
2334 2335 2336

		list_for_each_entry_safe(target, tmp_target,
					 &host->target_list, list) {
2337
			srp_remove_host(target->scsi_host);
2338
			scsi_remove_host(target->scsi_host);
2339 2340 2341
			srp_disconnect_target(target);
			ib_destroy_cm_id(target->cm_id);
			srp_free_target_ib(target);
2342
			srp_free_req_data(target);
2343 2344 2345 2346 2347 2348
			scsi_host_put(target->scsi_host);
		}

		kfree(host);
	}

2349 2350 2351 2352 2353 2354
	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);
2355 2356
}

2357 2358 2359
static struct srp_function_template ib_srp_transport_functions = {
};

2360 2361 2362 2363
static int __init srp_init_module(void)
{
	int ret;

2364
	BUILD_BUG_ON(FIELD_SIZEOF(struct ib_wc, wr_id) < sizeof(void *));
2365

2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377
	if (srp_sg_tablesize) {
		printk(KERN_WARNING PFX "srp_sg_tablesize is deprecated, please use cmd_sg_entries\n");
		if (!cmd_sg_entries)
			cmd_sg_entries = srp_sg_tablesize;
	}

	if (!cmd_sg_entries)
		cmd_sg_entries = SRP_DEF_SG_TABLESIZE;

	if (cmd_sg_entries > 255) {
		printk(KERN_WARNING PFX "Clamping cmd_sg_entries to 255\n");
		cmd_sg_entries = 255;
2378 2379
	}

2380 2381 2382 2383 2384
	ib_srp_transport_template =
		srp_attach_transport(&ib_srp_transport_functions);
	if (!ib_srp_transport_template)
		return -ENOMEM;

2385 2386 2387
	ret = class_register(&srp_class);
	if (ret) {
		printk(KERN_ERR PFX "couldn't register class infiniband_srp\n");
2388
		srp_release_transport(ib_srp_transport_template);
2389 2390 2391
		return ret;
	}

2392 2393
	ib_sa_register_client(&srp_sa_client);

2394 2395 2396
	ret = ib_register_client(&srp_client);
	if (ret) {
		printk(KERN_ERR PFX "couldn't register IB client\n");
2397
		srp_release_transport(ib_srp_transport_template);
2398
		ib_sa_unregister_client(&srp_sa_client);
2399 2400 2401 2402 2403 2404 2405 2406 2407 2408
		class_unregister(&srp_class);
		return ret;
	}

	return 0;
}

static void __exit srp_cleanup_module(void)
{
	ib_unregister_client(&srp_client);
2409
	ib_sa_unregister_client(&srp_sa_client);
2410
	class_unregister(&srp_class);
2411
	srp_release_transport(ib_srp_transport_template);
2412 2413 2414 2415
}

module_init(srp_init_module);
module_exit(srp_cleanup_module);