ib_isert.c 92.0 KB
Newer Older
1 2 3
/*******************************************************************************
 * This file contains iSCSI extentions for RDMA (iSER) Verbs
 *
4
 * (c) Copyright 2013 Datera, Inc.
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
 *
 * Nicholas A. Bellinger <nab@linux-iscsi.org>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 ****************************************************************************/

#include <linux/string.h>
#include <linux/module.h>
#include <linux/scatterlist.h>
#include <linux/socket.h>
#include <linux/in.h>
#include <linux/in6.h>
#include <rdma/ib_verbs.h>
#include <rdma/rdma_cm.h>
#include <target/target_core_base.h>
#include <target/target_core_fabric.h>
#include <target/iscsi/iscsi_transport.h>
30
#include <linux/semaphore.h>
31 32 33 34 35 36 37

#include "isert_proto.h"
#include "ib_isert.h"

#define	ISERT_MAX_CONN		8
#define ISER_MAX_RX_CQ_LEN	(ISERT_QP_MAX_RECV_DTOS * ISERT_MAX_CONN)
#define ISER_MAX_TX_CQ_LEN	(ISERT_QP_MAX_REQ_DTOS  * ISERT_MAX_CONN)
38 39
#define ISER_MAX_CQ_LEN		(ISER_MAX_RX_CQ_LEN + ISER_MAX_TX_CQ_LEN + \
				 ISERT_MAX_CONN)
40

S
Sagi Grimberg 已提交
41
static int isert_debug_level;
42 43 44
module_param_named(debug_level, isert_debug_level, int, 0644);
MODULE_PARM_DESC(debug_level, "Enable debug tracing if > 0 (default:0)");

45 46 47
static DEFINE_MUTEX(device_list_mutex);
static LIST_HEAD(device_list);
static struct workqueue_struct *isert_comp_wq;
48
static struct workqueue_struct *isert_release_wq;
49

50 51 52 53 54
static void
isert_unmap_cmd(struct isert_cmd *isert_cmd, struct isert_conn *isert_conn);
static int
isert_map_rdma(struct iscsi_conn *conn, struct iscsi_cmd *cmd,
	       struct isert_rdma_wr *wr);
55
static void
56
isert_unreg_rdma(struct isert_cmd *isert_cmd, struct isert_conn *isert_conn);
57
static int
58 59
isert_reg_rdma(struct iscsi_conn *conn, struct iscsi_cmd *cmd,
	       struct isert_rdma_wr *wr);
60 61
static int
isert_put_response(struct iscsi_conn *conn, struct iscsi_cmd *cmd);
62 63 64 65
static int
isert_rdma_post_recvl(struct isert_conn *isert_conn);
static int
isert_rdma_accept(struct isert_conn *isert_conn);
66
struct rdma_cm_id *isert_setup_id(struct isert_np *isert_np);
67

68 69
static void isert_release_work(struct work_struct *work);

70 71 72
static inline bool
isert_prot_cmd(struct isert_conn *conn, struct se_cmd *cmd)
{
73
	return (conn->pi_support &&
74 75 76 77
		cmd->prot_op != TARGET_PROT_NORMAL);
}


78 79 80
static void
isert_qp_event_callback(struct ib_event *e, void *context)
{
81
	struct isert_conn *isert_conn = context;
82

83 84 85
	isert_err("%s (%d): conn %p\n",
		  ib_event_msg(e->event), e->event, isert_conn);

86 87
	switch (e->event) {
	case IB_EVENT_COMM_EST:
88
		rdma_notify(isert_conn->cm_id, IB_EVENT_COMM_EST);
89 90
		break;
	case IB_EVENT_QP_LAST_WQE_REACHED:
91
		isert_warn("Reached TX IB_EVENT_QP_LAST_WQE_REACHED\n");
92 93 94 95 96 97 98 99 100 101 102 103 104
		break;
	default:
		break;
	}
}

static int
isert_query_device(struct ib_device *ib_dev, struct ib_device_attr *devattr)
{
	int ret;

	ret = ib_query_device(ib_dev, devattr);
	if (ret) {
105
		isert_err("ib_query_device() failed: %d\n", ret);
106 107
		return ret;
	}
108 109
	isert_dbg("devattr->max_sge: %d\n", devattr->max_sge);
	isert_dbg("devattr->max_sge_rd: %d\n", devattr->max_sge_rd);
110 111 112 113

	return 0;
}

114 115
static struct isert_comp *
isert_comp_get(struct isert_conn *isert_conn)
116
{
117
	struct isert_device *device = isert_conn->device;
118
	struct isert_comp *comp;
119
	int i, min = 0;
120 121

	mutex_lock(&device_list_mutex);
122 123 124 125 126 127
	for (i = 0; i < device->comps_used; i++)
		if (device->comps[i].active_qps <
		    device->comps[min].active_qps)
			min = i;
	comp = &device->comps[min];
	comp->active_qps++;
128 129
	mutex_unlock(&device_list_mutex);

130
	isert_info("conn %p, using comp %p min_index: %d\n",
131
		   isert_conn, comp, min);
132 133 134 135 136 137 138 139 140

	return comp;
}

static void
isert_comp_put(struct isert_comp *comp)
{
	mutex_lock(&device_list_mutex);
	comp->active_qps--;
141
	mutex_unlock(&device_list_mutex);
142 143 144 145 146 147 148
}

static struct ib_qp *
isert_create_qp(struct isert_conn *isert_conn,
		struct isert_comp *comp,
		struct rdma_cm_id *cma_id)
{
149
	struct isert_device *device = isert_conn->device;
150 151
	struct ib_qp_init_attr attr;
	int ret;
152 153 154 155

	memset(&attr, 0, sizeof(struct ib_qp_init_attr));
	attr.event_handler = isert_qp_event_callback;
	attr.qp_context = isert_conn;
156 157
	attr.send_cq = comp->cq;
	attr.recv_cq = comp->cq;
158
	attr.cap.max_send_wr = ISERT_QP_MAX_REQ_DTOS;
159
	attr.cap.max_recv_wr = ISERT_QP_MAX_RECV_DTOS + 1;
160 161
	/*
	 * FIXME: Use devattr.max_sge - 2 for max_send_sge as
162 163 164 165
	 * work-around for RDMA_READs with ConnectX-2.
	 *
	 * Also, still make sure to have at least two SGEs for
	 * outgoing control PDU responses.
166
	 */
167
	attr.cap.max_send_sge = max(2, device->dev_attr.max_sge - 2);
168 169 170 171 172
	isert_conn->max_sge = attr.cap.max_send_sge;

	attr.cap.max_recv_sge = 1;
	attr.sq_sig_type = IB_SIGNAL_REQ_WR;
	attr.qp_type = IB_QPT_RC;
173
	if (device->pi_capable)
174
		attr.create_flags |= IB_QP_CREATE_SIGNATURE_EN;
175

176
	ret = rdma_create_qp(cma_id, device->pd, &attr);
177
	if (ret) {
178
		isert_err("rdma_create_qp failed for cma_id %d\n", ret);
179 180 181 182 183 184 185 186 187 188 189 190 191
		return ERR_PTR(ret);
	}

	return cma_id->qp;
}

static int
isert_conn_setup_qp(struct isert_conn *isert_conn, struct rdma_cm_id *cma_id)
{
	struct isert_comp *comp;
	int ret;

	comp = isert_comp_get(isert_conn);
192 193 194
	isert_conn->qp = isert_create_qp(isert_conn, comp, cma_id);
	if (IS_ERR(isert_conn->qp)) {
		ret = PTR_ERR(isert_conn->qp);
195
		goto err;
196 197 198
	}

	return 0;
199
err:
200
	isert_comp_put(comp);
201
	return ret;
202 203 204 205 206
}

static void
isert_cq_event_callback(struct ib_event *e, void *context)
{
207
	isert_dbg("event: %d\n", e->event);
208 209 210 211 212
}

static int
isert_alloc_rx_descriptors(struct isert_conn *isert_conn)
{
213
	struct isert_device *device = isert_conn->device;
214
	struct ib_device *ib_dev = device->ib_device;
215 216 217 218 219
	struct iser_rx_desc *rx_desc;
	struct ib_sge *rx_sg;
	u64 dma_addr;
	int i, j;

220
	isert_conn->rx_descs = kzalloc(ISERT_QP_MAX_RECV_DTOS *
221
				sizeof(struct iser_rx_desc), GFP_KERNEL);
222
	if (!isert_conn->rx_descs)
223 224
		goto fail;

225
	rx_desc = isert_conn->rx_descs;
226 227 228 229 230 231 232 233 234 235 236 237

	for (i = 0; i < ISERT_QP_MAX_RECV_DTOS; i++, rx_desc++)  {
		dma_addr = ib_dma_map_single(ib_dev, (void *)rx_desc,
					ISER_RX_PAYLOAD_SIZE, DMA_FROM_DEVICE);
		if (ib_dma_mapping_error(ib_dev, dma_addr))
			goto dma_map_fail;

		rx_desc->dma_addr = dma_addr;

		rx_sg = &rx_desc->rx_sg;
		rx_sg->addr = rx_desc->dma_addr;
		rx_sg->length = ISER_RX_PAYLOAD_SIZE;
238
		rx_sg->lkey = device->pd->local_dma_lkey;
239 240 241 242 243
	}

	return 0;

dma_map_fail:
244
	rx_desc = isert_conn->rx_descs;
245 246 247 248
	for (j = 0; j < i; j++, rx_desc++) {
		ib_dma_unmap_single(ib_dev, rx_desc->dma_addr,
				    ISER_RX_PAYLOAD_SIZE, DMA_FROM_DEVICE);
	}
249 250
	kfree(isert_conn->rx_descs);
	isert_conn->rx_descs = NULL;
251
fail:
252 253
	isert_err("conn %p failed to allocate rx descriptors\n", isert_conn);

254 255 256 257 258 259
	return -ENOMEM;
}

static void
isert_free_rx_descriptors(struct isert_conn *isert_conn)
{
260
	struct ib_device *ib_dev = isert_conn->device->ib_device;
261 262 263
	struct iser_rx_desc *rx_desc;
	int i;

264
	if (!isert_conn->rx_descs)
265 266
		return;

267
	rx_desc = isert_conn->rx_descs;
268 269 270 271 272
	for (i = 0; i < ISERT_QP_MAX_RECV_DTOS; i++, rx_desc++)  {
		ib_dma_unmap_single(ib_dev, rx_desc->dma_addr,
				    ISER_RX_PAYLOAD_SIZE, DMA_FROM_DEVICE);
	}

273 274
	kfree(isert_conn->rx_descs);
	isert_conn->rx_descs = NULL;
275 276
}

277 278
static void isert_cq_work(struct work_struct *);
static void isert_cq_callback(struct ib_cq *, void *);
279

280 281
static void
isert_free_comps(struct isert_device *device)
282
{
283
	int i;
284

285 286
	for (i = 0; i < device->comps_used; i++) {
		struct isert_comp *comp = &device->comps[i];
287

288 289 290 291
		if (comp->cq) {
			cancel_work_sync(&comp->work);
			ib_destroy_cq(comp->cq);
		}
292
	}
293 294
	kfree(device->comps);
}
295

296 297 298 299 300
static int
isert_alloc_comps(struct isert_device *device,
		  struct ib_device_attr *attr)
{
	int i, max_cqe, ret = 0;
301

302
	device->comps_used = min(ISERT_MAX_CQ, min_t(int, num_online_cpus(),
303 304
				 device->ib_device->num_comp_vectors));

305
	isert_info("Using %d CQs, %s supports %d vectors support "
306 307 308 309 310 311 312 313
		   "Fast registration %d pi_capable %d\n",
		   device->comps_used, device->ib_device->name,
		   device->ib_device->num_comp_vectors, device->use_fastreg,
		   device->pi_capable);

	device->comps = kcalloc(device->comps_used, sizeof(struct isert_comp),
				GFP_KERNEL);
	if (!device->comps) {
314
		isert_err("Unable to allocate completion contexts\n");
315 316
		return -ENOMEM;
	}
317

318 319
	max_cqe = min(ISER_MAX_CQ_LEN, attr->max_cqe);

320
	for (i = 0; i < device->comps_used; i++) {
321
		struct ib_cq_init_attr cq_attr = {};
322 323 324
		struct isert_comp *comp = &device->comps[i];

		comp->device = device;
325
		INIT_WORK(&comp->work, isert_cq_work);
326 327
		cq_attr.cqe = max_cqe;
		cq_attr.comp_vector = i;
328 329 330 331
		comp->cq = ib_create_cq(device->ib_device,
					isert_cq_callback,
					isert_cq_event_callback,
					(void *)comp,
332
					&cq_attr);
333
		if (IS_ERR(comp->cq)) {
334
			isert_err("Unable to allocate cq\n");
335 336
			ret = PTR_ERR(comp->cq);
			comp->cq = NULL;
337
			goto out_cq;
338
		}
339

340
		ret = ib_req_notify_cq(comp->cq, IB_CQ_NEXT_COMP);
341
		if (ret)
342 343 344
			goto out_cq;
	}

345 346 347 348 349 350 351 352 353 354
	return 0;
out_cq:
	isert_free_comps(device);
	return ret;
}

static int
isert_create_device_ib_res(struct isert_device *device)
{
	struct ib_device_attr *dev_attr;
355
	int ret;
356 357

	dev_attr = &device->dev_attr;
358
	ret = isert_query_device(device->ib_device, dev_attr);
359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377
	if (ret)
		return ret;

	/* asign function handlers */
	if (dev_attr->device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS &&
	    dev_attr->device_cap_flags & IB_DEVICE_SIGNATURE_HANDOVER) {
		device->use_fastreg = 1;
		device->reg_rdma_mem = isert_reg_rdma;
		device->unreg_rdma_mem = isert_unreg_rdma;
	} else {
		device->use_fastreg = 0;
		device->reg_rdma_mem = isert_map_rdma;
		device->unreg_rdma_mem = isert_unmap_cmd;
	}

	ret = isert_alloc_comps(device, dev_attr);
	if (ret)
		return ret;

378 379 380 381 382 383 384 385
	device->pd = ib_alloc_pd(device->ib_device);
	if (IS_ERR(device->pd)) {
		ret = PTR_ERR(device->pd);
		isert_err("failed to allocate pd, device %p, ret=%d\n",
			  device, ret);
		goto out_cq;
	}

386 387 388
	/* Check signature cap */
	device->pi_capable = dev_attr->device_cap_flags &
			     IB_DEVICE_SIGNATURE_HANDOVER ? true : false;
389

390 391 392
	return 0;

out_cq:
393
	isert_free_comps(device);
394 395 396 397 398 399
	return ret;
}

static void
isert_free_device_ib_res(struct isert_device *device)
{
400
	isert_info("device %p\n", device);
401

402
	ib_dealloc_pd(device->pd);
403
	isert_free_comps(device);
404 405 406
}

static void
407
isert_device_put(struct isert_device *device)
408 409 410
{
	mutex_lock(&device_list_mutex);
	device->refcount--;
411
	isert_info("device %p refcount %d\n", device, device->refcount);
412 413 414 415 416 417 418 419 420
	if (!device->refcount) {
		isert_free_device_ib_res(device);
		list_del(&device->dev_node);
		kfree(device);
	}
	mutex_unlock(&device_list_mutex);
}

static struct isert_device *
421
isert_device_get(struct rdma_cm_id *cma_id)
422 423 424 425 426 427 428 429
{
	struct isert_device *device;
	int ret;

	mutex_lock(&device_list_mutex);
	list_for_each_entry(device, &device_list, dev_node) {
		if (device->ib_device->node_guid == cma_id->device->node_guid) {
			device->refcount++;
430 431
			isert_info("Found iser device %p refcount %d\n",
				   device, device->refcount);
432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454
			mutex_unlock(&device_list_mutex);
			return device;
		}
	}

	device = kzalloc(sizeof(struct isert_device), GFP_KERNEL);
	if (!device) {
		mutex_unlock(&device_list_mutex);
		return ERR_PTR(-ENOMEM);
	}

	INIT_LIST_HEAD(&device->dev_node);

	device->ib_device = cma_id->device;
	ret = isert_create_device_ib_res(device);
	if (ret) {
		kfree(device);
		mutex_unlock(&device_list_mutex);
		return ERR_PTR(ret);
	}

	device->refcount++;
	list_add_tail(&device->dev_node, &device_list);
455 456
	isert_info("Created a new iser device %p refcount %d\n",
		   device, device->refcount);
457 458 459 460 461
	mutex_unlock(&device_list_mutex);

	return device;
}

462
static void
463
isert_conn_free_fastreg_pool(struct isert_conn *isert_conn)
464 465 466 467
{
	struct fast_reg_descriptor *fr_desc, *tmp;
	int i = 0;

468
	if (list_empty(&isert_conn->fr_pool))
469 470
		return;

471
	isert_info("Freeing conn %p fastreg pool", isert_conn);
472 473

	list_for_each_entry_safe(fr_desc, tmp,
474
				 &isert_conn->fr_pool, list) {
475 476 477
		list_del(&fr_desc->list);
		ib_free_fast_reg_page_list(fr_desc->data_frpl);
		ib_dereg_mr(fr_desc->data_mr);
478 479 480
		if (fr_desc->pi_ctx) {
			ib_free_fast_reg_page_list(fr_desc->pi_ctx->prot_frpl);
			ib_dereg_mr(fr_desc->pi_ctx->prot_mr);
481
			ib_dereg_mr(fr_desc->pi_ctx->sig_mr);
482 483
			kfree(fr_desc->pi_ctx);
		}
484 485 486 487
		kfree(fr_desc);
		++i;
	}

488
	if (i < isert_conn->fr_pool_size)
489
		isert_warn("Pool still has %d regions registered\n",
490
			isert_conn->fr_pool_size - i);
491 492
}

493 494 495 496 497 498 499 500 501 502
static int
isert_create_pi_ctx(struct fast_reg_descriptor *desc,
		    struct ib_device *device,
		    struct ib_pd *pd)
{
	struct pi_context *pi_ctx;
	int ret;

	pi_ctx = kzalloc(sizeof(*desc->pi_ctx), GFP_KERNEL);
	if (!pi_ctx) {
503
		isert_err("Failed to allocate pi context\n");
504 505 506 507 508 509
		return -ENOMEM;
	}

	pi_ctx->prot_frpl = ib_alloc_fast_reg_page_list(device,
					    ISCSI_ISER_SG_TABLESIZE);
	if (IS_ERR(pi_ctx->prot_frpl)) {
510
		isert_err("Failed to allocate prot frpl err=%ld\n",
511 512 513 514 515
			  PTR_ERR(pi_ctx->prot_frpl));
		ret = PTR_ERR(pi_ctx->prot_frpl);
		goto err_pi_ctx;
	}

516 517
	pi_ctx->prot_mr = ib_alloc_mr(pd, IB_MR_TYPE_MEM_REG,
				      ISCSI_ISER_SG_TABLESIZE);
518
	if (IS_ERR(pi_ctx->prot_mr)) {
519
		isert_err("Failed to allocate prot frmr err=%ld\n",
520 521 522 523 524 525
			  PTR_ERR(pi_ctx->prot_mr));
		ret = PTR_ERR(pi_ctx->prot_mr);
		goto err_prot_frpl;
	}
	desc->ind |= ISERT_PROT_KEY_VALID;

S
Sagi Grimberg 已提交
526
	pi_ctx->sig_mr = ib_alloc_mr(pd, IB_MR_TYPE_SIGNATURE, 2);
527
	if (IS_ERR(pi_ctx->sig_mr)) {
528
		isert_err("Failed to allocate signature enabled mr err=%ld\n",
529 530 531 532 533 534 535 536 537 538 539 540
			  PTR_ERR(pi_ctx->sig_mr));
		ret = PTR_ERR(pi_ctx->sig_mr);
		goto err_prot_mr;
	}

	desc->pi_ctx = pi_ctx;
	desc->ind |= ISERT_SIG_KEY_VALID;
	desc->ind &= ~ISERT_PROTECTED;

	return 0;

err_prot_mr:
541
	ib_dereg_mr(pi_ctx->prot_mr);
542
err_prot_frpl:
543
	ib_free_fast_reg_page_list(pi_ctx->prot_frpl);
544
err_pi_ctx:
545
	kfree(pi_ctx);
546 547 548 549

	return ret;
}

550 551
static int
isert_create_fr_desc(struct ib_device *ib_device, struct ib_pd *pd,
552
		     struct fast_reg_descriptor *fr_desc)
553
{
554 555
	int ret;

556 557 558
	fr_desc->data_frpl = ib_alloc_fast_reg_page_list(ib_device,
							 ISCSI_ISER_SG_TABLESIZE);
	if (IS_ERR(fr_desc->data_frpl)) {
559
		isert_err("Failed to allocate data frpl err=%ld\n",
560
			  PTR_ERR(fr_desc->data_frpl));
561 562 563
		return PTR_ERR(fr_desc->data_frpl);
	}

564 565
	fr_desc->data_mr = ib_alloc_mr(pd, IB_MR_TYPE_MEM_REG,
				       ISCSI_ISER_SG_TABLESIZE);
566
	if (IS_ERR(fr_desc->data_mr)) {
567
		isert_err("Failed to allocate data frmr err=%ld\n",
568
			  PTR_ERR(fr_desc->data_mr));
569 570
		ret = PTR_ERR(fr_desc->data_mr);
		goto err_data_frpl;
571
	}
572 573
	fr_desc->ind |= ISERT_DATA_KEY_VALID;

574
	isert_dbg("Created fr_desc %p\n", fr_desc);
575 576

	return 0;
577

578 579 580 581
err_data_frpl:
	ib_free_fast_reg_page_list(fr_desc->data_frpl);

	return ret;
582 583 584
}

static int
585
isert_conn_create_fastreg_pool(struct isert_conn *isert_conn)
586 587
{
	struct fast_reg_descriptor *fr_desc;
588
	struct isert_device *device = isert_conn->device;
589 590 591 592 593 594 595 596 597
	struct se_session *se_sess = isert_conn->conn->sess->se_sess;
	struct se_node_acl *se_nacl = se_sess->se_node_acl;
	int i, ret, tag_num;
	/*
	 * Setup the number of FRMRs based upon the number of tags
	 * available to session in iscsi_target_locate_portal().
	 */
	tag_num = max_t(u32, ISCSIT_MIN_TAGS, se_nacl->queue_depth);
	tag_num = (tag_num * 2) + ISCSIT_EXTRA_TAGS;
598

599
	isert_conn->fr_pool_size = 0;
600
	for (i = 0; i < tag_num; i++) {
601 602
		fr_desc = kzalloc(sizeof(*fr_desc), GFP_KERNEL);
		if (!fr_desc) {
603
			isert_err("Failed to allocate fast_reg descriptor\n");
604 605 606 607
			ret = -ENOMEM;
			goto err;
		}

608
		ret = isert_create_fr_desc(device->ib_device,
609
					   device->pd, fr_desc);
610
		if (ret) {
611
			isert_err("Failed to create fastreg descriptor err=%d\n",
612
			       ret);
613
			kfree(fr_desc);
614 615 616
			goto err;
		}

617 618
		list_add_tail(&fr_desc->list, &isert_conn->fr_pool);
		isert_conn->fr_pool_size++;
619 620
	}

621
	isert_dbg("Creating conn %p fastreg pool size=%d",
622
		 isert_conn, isert_conn->fr_pool_size);
623 624 625 626

	return 0;

err:
627
	isert_conn_free_fastreg_pool(isert_conn);
628 629 630
	return ret;
}

631 632
static void
isert_init_conn(struct isert_conn *isert_conn)
633 634
{
	isert_conn->state = ISER_CONN_INIT;
635
	INIT_LIST_HEAD(&isert_conn->node);
636
	init_completion(&isert_conn->login_comp);
637
	init_completion(&isert_conn->login_req_comp);
638 639 640 641 642
	init_completion(&isert_conn->wait);
	kref_init(&isert_conn->kref);
	mutex_init(&isert_conn->mutex);
	spin_lock_init(&isert_conn->pool_lock);
	INIT_LIST_HEAD(&isert_conn->fr_pool);
643
	INIT_WORK(&isert_conn->release_work, isert_release_work);
644
}
645

646 647 648
static void
isert_free_login_buf(struct isert_conn *isert_conn)
{
649
	struct ib_device *ib_dev = isert_conn->device->ib_device;
650 651 652 653 654 655 656 657 658 659 660 661 662 663

	ib_dma_unmap_single(ib_dev, isert_conn->login_rsp_dma,
			    ISER_RX_LOGIN_SIZE, DMA_TO_DEVICE);
	ib_dma_unmap_single(ib_dev, isert_conn->login_req_dma,
			    ISCSI_DEF_MAX_RECV_SEG_LEN,
			    DMA_FROM_DEVICE);
	kfree(isert_conn->login_buf);
}

static int
isert_alloc_login_buf(struct isert_conn *isert_conn,
		      struct ib_device *ib_dev)
{
	int ret;
664 665 666 667

	isert_conn->login_buf = kzalloc(ISCSI_DEF_MAX_RECV_SEG_LEN +
					ISER_RX_LOGIN_SIZE, GFP_KERNEL);
	if (!isert_conn->login_buf) {
668
		isert_err("Unable to allocate isert_conn->login_buf\n");
669
		return -ENOMEM;
670 671 672 673 674
	}

	isert_conn->login_req_buf = isert_conn->login_buf;
	isert_conn->login_rsp_buf = isert_conn->login_buf +
				    ISCSI_DEF_MAX_RECV_SEG_LEN;
675

676
	isert_dbg("Set login_buf: %p login_req_buf: %p login_rsp_buf: %p\n",
677 678 679 680 681 682 683 684 685
		 isert_conn->login_buf, isert_conn->login_req_buf,
		 isert_conn->login_rsp_buf);

	isert_conn->login_req_dma = ib_dma_map_single(ib_dev,
				(void *)isert_conn->login_req_buf,
				ISCSI_DEF_MAX_RECV_SEG_LEN, DMA_FROM_DEVICE);

	ret = ib_dma_mapping_error(ib_dev, isert_conn->login_req_dma);
	if (ret) {
686
		isert_err("login_req_dma mapping error: %d\n", ret);
687 688 689 690 691 692 693 694 695 696
		isert_conn->login_req_dma = 0;
		goto out_login_buf;
	}

	isert_conn->login_rsp_dma = ib_dma_map_single(ib_dev,
					(void *)isert_conn->login_rsp_buf,
					ISER_RX_LOGIN_SIZE, DMA_TO_DEVICE);

	ret = ib_dma_mapping_error(ib_dev, isert_conn->login_rsp_dma);
	if (ret) {
697
		isert_err("login_rsp_dma mapping error: %d\n", ret);
698 699 700 701
		isert_conn->login_rsp_dma = 0;
		goto out_req_dma_map;
	}

702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736
	return 0;

out_req_dma_map:
	ib_dma_unmap_single(ib_dev, isert_conn->login_req_dma,
			    ISCSI_DEF_MAX_RECV_SEG_LEN, DMA_FROM_DEVICE);
out_login_buf:
	kfree(isert_conn->login_buf);
	return ret;
}

static int
isert_connect_request(struct rdma_cm_id *cma_id, struct rdma_cm_event *event)
{
	struct isert_np *isert_np = cma_id->context;
	struct iscsi_np *np = isert_np->np;
	struct isert_conn *isert_conn;
	struct isert_device *device;
	int ret = 0;

	spin_lock_bh(&np->np_thread_lock);
	if (!np->enabled) {
		spin_unlock_bh(&np->np_thread_lock);
		isert_dbg("iscsi_np is not enabled, reject connect request\n");
		return rdma_reject(cma_id, NULL, 0);
	}
	spin_unlock_bh(&np->np_thread_lock);

	isert_dbg("cma_id: %p, portal: %p\n",
		 cma_id, cma_id->context);

	isert_conn = kzalloc(sizeof(struct isert_conn), GFP_KERNEL);
	if (!isert_conn)
		return -ENOMEM;

	isert_init_conn(isert_conn);
737
	isert_conn->cm_id = cma_id;
738 739 740 741 742

	ret = isert_alloc_login_buf(isert_conn, cma_id->device);
	if (ret)
		goto out;

743
	device = isert_device_get(cma_id);
744 745 746 747
	if (IS_ERR(device)) {
		ret = PTR_ERR(device);
		goto out_rsp_dma_map;
	}
748
	isert_conn->device = device;
749

750 751 752 753
	/* Set max inflight RDMA READ requests */
	isert_conn->initiator_depth = min_t(u8,
				event->param.conn.initiator_depth,
				device->dev_attr.max_qp_init_rd_atom);
754
	isert_dbg("Using initiator_depth: %u\n", isert_conn->initiator_depth);
755

756
	ret = isert_conn_setup_qp(isert_conn, cma_id);
757 758 759
	if (ret)
		goto out_conn_dev;

760 761 762 763 764 765 766 767
	ret = isert_rdma_post_recvl(isert_conn);
	if (ret)
		goto out_conn_dev;

	ret = isert_rdma_accept(isert_conn);
	if (ret)
		goto out_conn_dev;

768
	mutex_lock(&isert_np->mutex);
769
	list_add_tail(&isert_conn->node, &isert_np->accepted);
770
	mutex_unlock(&isert_np->mutex);
771 772 773 774

	return 0;

out_conn_dev:
775
	isert_device_put(device);
776
out_rsp_dma_map:
777
	isert_free_login_buf(isert_conn);
778 779
out:
	kfree(isert_conn);
780
	rdma_reject(cma_id, NULL, 0);
781 782 783 784 785 786
	return ret;
}

static void
isert_connect_release(struct isert_conn *isert_conn)
{
787
	struct isert_device *device = isert_conn->device;
788

789
	isert_dbg("conn %p\n", isert_conn);
790

791 792 793
	BUG_ON(!device);

	if (device->use_fastreg)
794
		isert_conn_free_fastreg_pool(isert_conn);
795

796
	isert_free_rx_descriptors(isert_conn);
797 798
	if (isert_conn->cm_id)
		rdma_destroy_id(isert_conn->cm_id);
799

800 801
	if (isert_conn->qp) {
		struct isert_comp *comp = isert_conn->qp->recv_cq->cq_context;
802

803
		isert_comp_put(comp);
804
		ib_destroy_qp(isert_conn->qp);
805 806
	}

807 808 809
	if (isert_conn->login_buf)
		isert_free_login_buf(isert_conn);

810
	isert_device_put(device);
811

812
	kfree(isert_conn);
813 814 815 816 817
}

static void
isert_connected_handler(struct rdma_cm_id *cma_id)
{
818
	struct isert_conn *isert_conn = cma_id->qp->qp_context;
819
	struct isert_np *isert_np = cma_id->context;
820

821
	isert_info("conn %p\n", isert_conn);
822

823
	mutex_lock(&isert_conn->mutex);
824 825
	isert_conn->state = ISER_CONN_UP;
	kref_get(&isert_conn->kref);
826
	mutex_unlock(&isert_conn->mutex);
827 828 829 830 831 832 833

	mutex_lock(&isert_np->mutex);
	list_move_tail(&isert_conn->node, &isert_np->pending);
	mutex_unlock(&isert_np->mutex);

	isert_info("np %p: Allow accept_np to continue\n", isert_np);
	up(&isert_np->sem);
834 835 836
}

static void
837
isert_release_kref(struct kref *kref)
838 839
{
	struct isert_conn *isert_conn = container_of(kref,
840
				struct isert_conn, kref);
841

842 843
	isert_info("conn %p final kref %s/%d\n", isert_conn, current->comm,
		   current->pid);
844 845 846 847 848 849 850

	isert_connect_release(isert_conn);
}

static void
isert_put_conn(struct isert_conn *isert_conn)
{
851
	kref_put(&isert_conn->kref, isert_release_kref);
852 853
}

854 855 856 857 858
/**
 * isert_conn_terminate() - Initiate connection termination
 * @isert_conn: isert connection struct
 *
 * Notes:
859
 * In case the connection state is FULL_FEATURE, move state
860
 * to TEMINATING and start teardown sequence (rdma_disconnect).
861
 * In case the connection state is UP, complete flush as well.
862
 *
863
 * This routine must be called with mutex held. Thus it is
864 865 866 867 868 869 870
 * safe to call multiple times.
 */
static void
isert_conn_terminate(struct isert_conn *isert_conn)
{
	int err;

871 872 873 874 875
	switch (isert_conn->state) {
	case ISER_CONN_TERMINATING:
		break;
	case ISER_CONN_UP:
	case ISER_CONN_FULL_FEATURE: /* FALLTHRU */
876
		isert_info("Terminating conn %p state %d\n",
877
			   isert_conn, isert_conn->state);
878
		isert_conn->state = ISER_CONN_TERMINATING;
879
		err = rdma_disconnect(isert_conn->cm_id);
880
		if (err)
881
			isert_warn("Failed rdma_disconnect isert_conn %p\n",
882
				   isert_conn);
883 884
		break;
	default:
885
		isert_warn("conn %p teminating in state %d\n",
886
			   isert_conn, isert_conn->state);
887 888 889
	}
}

890
static int
891 892
isert_np_cma_handler(struct isert_np *isert_np,
		     enum rdma_cm_event_type event)
893
{
894 895
	isert_dbg("%s (%d): isert np %p\n",
		  rdma_event_msg(event), event, isert_np);
896

897 898
	switch (event) {
	case RDMA_CM_EVENT_DEVICE_REMOVAL:
899
		isert_np->cm_id = NULL;
900 901
		break;
	case RDMA_CM_EVENT_ADDR_CHANGE:
902 903
		isert_np->cm_id = isert_setup_id(isert_np);
		if (IS_ERR(isert_np->cm_id)) {
904
			isert_err("isert np %p setup id failed: %ld\n",
905 906
				  isert_np, PTR_ERR(isert_np->cm_id));
			isert_np->cm_id = NULL;
907 908 909
		}
		break;
	default:
910
		isert_err("isert np %p Unexpected event %d\n",
911
			  isert_np, event);
912 913
	}

914 915 916 917 918 919 920 921 922
	return -1;
}

static int
isert_disconnected_handler(struct rdma_cm_id *cma_id,
			   enum rdma_cm_event_type event)
{
	struct isert_np *isert_np = cma_id->context;
	struct isert_conn *isert_conn;
923
	bool terminating = false;
924

925
	if (isert_np->cm_id == cma_id)
926 927
		return isert_np_cma_handler(cma_id->context, event);

928
	isert_conn = cma_id->qp->qp_context;
929

930
	mutex_lock(&isert_conn->mutex);
931
	terminating = (isert_conn->state == ISER_CONN_TERMINATING);
932
	isert_conn_terminate(isert_conn);
933
	mutex_unlock(&isert_conn->mutex);
934

935 936
	isert_info("conn %p completing wait\n", isert_conn);
	complete(&isert_conn->wait);
937

938 939 940
	if (terminating)
		goto out;

941
	mutex_lock(&isert_np->mutex);
942 943
	if (!list_empty(&isert_conn->node)) {
		list_del_init(&isert_conn->node);
944 945 946
		isert_put_conn(isert_conn);
		queue_work(isert_release_wq, &isert_conn->release_work);
	}
947
	mutex_unlock(&isert_np->mutex);
948 949

out:
950
	return 0;
951 952
}

953
static int
954 955
isert_connect_error(struct rdma_cm_id *cma_id)
{
956
	struct isert_conn *isert_conn = cma_id->qp->qp_context;
957

958
	list_del_init(&isert_conn->node);
959
	isert_conn->cm_id = NULL;
960
	isert_put_conn(isert_conn);
961 962

	return -1;
963 964
}

965 966 967 968 969
static int
isert_cma_handler(struct rdma_cm_id *cma_id, struct rdma_cm_event *event)
{
	int ret = 0;

970 971
	isert_info("%s (%d): status %d id %p np %p\n",
		   rdma_event_msg(event->event), event->event,
972
		   event->status, cma_id, cma_id->context);
973 974 975 976

	switch (event->event) {
	case RDMA_CM_EVENT_CONNECT_REQUEST:
		ret = isert_connect_request(cma_id, event);
977
		if (ret)
978
			isert_err("failed handle connect request %d\n", ret);
979 980 981 982
		break;
	case RDMA_CM_EVENT_ESTABLISHED:
		isert_connected_handler(cma_id);
		break;
983 984 985 986
	case RDMA_CM_EVENT_ADDR_CHANGE:    /* FALLTHRU */
	case RDMA_CM_EVENT_DISCONNECTED:   /* FALLTHRU */
	case RDMA_CM_EVENT_DEVICE_REMOVAL: /* FALLTHRU */
	case RDMA_CM_EVENT_TIMEWAIT_EXIT:  /* FALLTHRU */
987
		ret = isert_disconnected_handler(cma_id, event->event);
988
		break;
989 990
	case RDMA_CM_EVENT_REJECTED:       /* FALLTHRU */
	case RDMA_CM_EVENT_UNREACHABLE:    /* FALLTHRU */
991
	case RDMA_CM_EVENT_CONNECT_ERROR:
992
		ret = isert_connect_error(cma_id);
993
		break;
994
	default:
995
		isert_err("Unhandled RDMA CMA event: %d\n", event->event);
996 997 998 999 1000 1001 1002
		break;
	}

	return ret;
}

static int
1003
isert_post_recvm(struct isert_conn *isert_conn, u32 count)
1004 1005 1006 1007 1008
{
	struct ib_recv_wr *rx_wr, *rx_wr_failed;
	int i, ret;
	struct iser_rx_desc *rx_desc;

1009
	for (rx_wr = isert_conn->rx_wr, i = 0; i < count; i++, rx_wr++) {
1010 1011 1012 1013 1014
		rx_desc = &isert_conn->rx_descs[i];
		rx_wr->wr_id = (uintptr_t)rx_desc;
		rx_wr->sg_list = &rx_desc->rx_sg;
		rx_wr->num_sge = 1;
		rx_wr->next = rx_wr + 1;
1015 1016 1017 1018 1019
	}
	rx_wr--;
	rx_wr->next = NULL; /* mark end of work requests list */

	isert_conn->post_recv_buf_count += count;
1020
	ret = ib_post_recv(isert_conn->qp, isert_conn->rx_wr,
1021
			   &rx_wr_failed);
1022
	if (ret) {
1023
		isert_err("ib_post_recv() failed with ret: %d\n", ret);
1024 1025
		isert_conn->post_recv_buf_count -= count;
	}
1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047

	return ret;
}

static int
isert_post_recv(struct isert_conn *isert_conn, struct iser_rx_desc *rx_desc)
{
	struct ib_recv_wr *rx_wr_failed, rx_wr;
	int ret;

	rx_wr.wr_id = (uintptr_t)rx_desc;
	rx_wr.sg_list = &rx_desc->rx_sg;
	rx_wr.num_sge = 1;
	rx_wr.next = NULL;

	isert_conn->post_recv_buf_count++;
	ret = ib_post_recv(isert_conn->qp, &rx_wr, &rx_wr_failed);
	if (ret) {
		isert_err("ib_post_recv() failed with ret: %d\n", ret);
		isert_conn->post_recv_buf_count--;
	}

1048 1049 1050 1051 1052 1053
	return ret;
}

static int
isert_post_send(struct isert_conn *isert_conn, struct iser_tx_desc *tx_desc)
{
1054
	struct ib_device *ib_dev = isert_conn->cm_id->device;
1055 1056 1057 1058 1059 1060 1061
	struct ib_send_wr send_wr, *send_wr_failed;
	int ret;

	ib_dma_sync_single_for_device(ib_dev, tx_desc->dma_addr,
				      ISER_HEADERS_LEN, DMA_TO_DEVICE);

	send_wr.next	= NULL;
1062
	send_wr.wr_id	= (uintptr_t)tx_desc;
1063 1064 1065 1066 1067
	send_wr.sg_list	= tx_desc->tx_sg;
	send_wr.num_sge	= tx_desc->num_sge;
	send_wr.opcode	= IB_WR_SEND;
	send_wr.send_flags = IB_SEND_SIGNALED;

1068
	ret = ib_post_send(isert_conn->qp, &send_wr, &send_wr_failed);
1069
	if (ret)
1070
		isert_err("ib_post_send() failed, ret: %d\n", ret);
1071 1072 1073 1074 1075 1076 1077 1078 1079

	return ret;
}

static void
isert_create_send_desc(struct isert_conn *isert_conn,
		       struct isert_cmd *isert_cmd,
		       struct iser_tx_desc *tx_desc)
{
1080
	struct isert_device *device = isert_conn->device;
1081
	struct ib_device *ib_dev = device->ib_device;
1082 1083 1084 1085 1086 1087 1088 1089 1090 1091

	ib_dma_sync_single_for_cpu(ib_dev, tx_desc->dma_addr,
				   ISER_HEADERS_LEN, DMA_TO_DEVICE);

	memset(&tx_desc->iser_header, 0, sizeof(struct iser_hdr));
	tx_desc->iser_header.flags = ISER_VER;

	tx_desc->num_sge = 1;
	tx_desc->isert_cmd = isert_cmd;

1092 1093
	if (tx_desc->tx_sg[0].lkey != device->pd->local_dma_lkey) {
		tx_desc->tx_sg[0].lkey = device->pd->local_dma_lkey;
1094
		isert_dbg("tx_desc %p lkey mismatch, fixing\n", tx_desc);
1095 1096 1097 1098 1099 1100 1101
	}
}

static int
isert_init_tx_hdrs(struct isert_conn *isert_conn,
		   struct iser_tx_desc *tx_desc)
{
1102
	struct isert_device *device = isert_conn->device;
1103
	struct ib_device *ib_dev = device->ib_device;
1104 1105 1106 1107 1108
	u64 dma_addr;

	dma_addr = ib_dma_map_single(ib_dev, (void *)tx_desc,
			ISER_HEADERS_LEN, DMA_TO_DEVICE);
	if (ib_dma_mapping_error(ib_dev, dma_addr)) {
1109
		isert_err("ib_dma_mapping_error() failed\n");
1110 1111 1112 1113 1114 1115
		return -ENOMEM;
	}

	tx_desc->dma_addr = dma_addr;
	tx_desc->tx_sg[0].addr	= tx_desc->dma_addr;
	tx_desc->tx_sg[0].length = ISER_HEADERS_LEN;
1116
	tx_desc->tx_sg[0].lkey = device->pd->local_dma_lkey;
1117

1118 1119 1120
	isert_dbg("Setup tx_sg[0].addr: 0x%llx length: %u lkey: 0x%x\n",
		  tx_desc->tx_sg[0].addr, tx_desc->tx_sg[0].length,
		  tx_desc->tx_sg[0].lkey);
1121 1122 1123 1124 1125

	return 0;
}

static void
1126
isert_init_send_wr(struct isert_conn *isert_conn, struct isert_cmd *isert_cmd,
1127
		   struct ib_send_wr *send_wr)
1128
{
1129 1130
	struct iser_tx_desc *tx_desc = &isert_cmd->tx_desc;

1131
	isert_cmd->rdma_wr.iser_ib_op = ISER_IB_SEND;
1132
	send_wr->wr_id = (uintptr_t)&isert_cmd->tx_desc;
1133
	send_wr->opcode = IB_WR_SEND;
1134
	send_wr->sg_list = &tx_desc->tx_sg[0];
1135
	send_wr->num_sge = isert_cmd->tx_desc.num_sge;
1136
	send_wr->send_flags = IB_SEND_SIGNALED;
1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148
}

static int
isert_rdma_post_recvl(struct isert_conn *isert_conn)
{
	struct ib_recv_wr rx_wr, *rx_wr_fail;
	struct ib_sge sge;
	int ret;

	memset(&sge, 0, sizeof(struct ib_sge));
	sge.addr = isert_conn->login_req_dma;
	sge.length = ISER_RX_LOGIN_SIZE;
1149
	sge.lkey = isert_conn->device->pd->local_dma_lkey;
1150

1151
	isert_dbg("Setup sge: addr: %llx length: %d 0x%08x\n",
1152 1153 1154
		sge.addr, sge.length, sge.lkey);

	memset(&rx_wr, 0, sizeof(struct ib_recv_wr));
1155
	rx_wr.wr_id = (uintptr_t)isert_conn->login_req_buf;
1156 1157 1158 1159
	rx_wr.sg_list = &sge;
	rx_wr.num_sge = 1;

	isert_conn->post_recv_buf_count++;
1160
	ret = ib_post_recv(isert_conn->qp, &rx_wr, &rx_wr_fail);
1161
	if (ret) {
1162
		isert_err("ib_post_recv() failed: %d\n", ret);
1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173
		isert_conn->post_recv_buf_count--;
	}

	return ret;
}

static int
isert_put_login_tx(struct iscsi_conn *conn, struct iscsi_login *login,
		   u32 length)
{
	struct isert_conn *isert_conn = conn->context;
1174
	struct isert_device *device = isert_conn->device;
1175
	struct ib_device *ib_dev = device->ib_device;
1176
	struct iser_tx_desc *tx_desc = &isert_conn->login_tx_desc;
1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198
	int ret;

	isert_create_send_desc(isert_conn, NULL, tx_desc);

	memcpy(&tx_desc->iscsi_header, &login->rsp[0],
	       sizeof(struct iscsi_hdr));

	isert_init_tx_hdrs(isert_conn, tx_desc);

	if (length > 0) {
		struct ib_sge *tx_dsg = &tx_desc->tx_sg[1];

		ib_dma_sync_single_for_cpu(ib_dev, isert_conn->login_rsp_dma,
					   length, DMA_TO_DEVICE);

		memcpy(isert_conn->login_rsp_buf, login->rsp_buf, length);

		ib_dma_sync_single_for_device(ib_dev, isert_conn->login_rsp_dma,
					      length, DMA_TO_DEVICE);

		tx_dsg->addr	= isert_conn->login_rsp_dma;
		tx_dsg->length	= length;
1199
		tx_dsg->lkey	= isert_conn->device->pd->local_dma_lkey;
1200 1201 1202 1203
		tx_desc->num_sge = 2;
	}
	if (!login->login_failed) {
		if (login->login_complete) {
1204
			if (!conn->sess->sess_ops->SessionType &&
1205
			    isert_conn->device->use_fastreg) {
1206
				ret = isert_conn_create_fastreg_pool(isert_conn);
1207
				if (ret) {
1208
					isert_err("Conn: %p failed to create"
1209 1210 1211 1212 1213
					       " fastreg pool\n", isert_conn);
					return ret;
				}
			}

1214 1215 1216 1217
			ret = isert_alloc_rx_descriptors(isert_conn);
			if (ret)
				return ret;

1218 1219
			ret = isert_post_recvm(isert_conn,
					       ISERT_QP_MAX_RECV_DTOS);
1220 1221 1222
			if (ret)
				return ret;

1223
			/* Now we are in FULL_FEATURE phase */
1224
			mutex_lock(&isert_conn->mutex);
1225
			isert_conn->state = ISER_CONN_FULL_FEATURE;
1226
			mutex_unlock(&isert_conn->mutex);
1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242
			goto post_send;
		}

		ret = isert_rdma_post_recvl(isert_conn);
		if (ret)
			return ret;
	}
post_send:
	ret = isert_post_send(isert_conn, tx_desc);
	if (ret)
		return ret;

	return 0;
}

static void
1243
isert_rx_login_req(struct isert_conn *isert_conn)
1244
{
1245 1246
	struct iser_rx_desc *rx_desc = (void *)isert_conn->login_req_buf;
	int rx_buflen = isert_conn->login_req_len;
1247 1248 1249 1250
	struct iscsi_conn *conn = isert_conn->conn;
	struct iscsi_login *login = conn->conn_login;
	int size;

1251
	isert_info("conn %p\n", isert_conn);
1252 1253

	WARN_ON_ONCE(!login);
1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278

	if (login->first_request) {
		struct iscsi_login_req *login_req =
			(struct iscsi_login_req *)&rx_desc->iscsi_header;
		/*
		 * Setup the initial iscsi_login values from the leading
		 * login request PDU.
		 */
		login->leading_connection = (!login_req->tsih) ? 1 : 0;
		login->current_stage =
			(login_req->flags & ISCSI_FLAG_LOGIN_CURRENT_STAGE_MASK)
			 >> 2;
		login->version_min	= login_req->min_version;
		login->version_max	= login_req->max_version;
		memcpy(login->isid, login_req->isid, 6);
		login->cmd_sn		= be32_to_cpu(login_req->cmdsn);
		login->init_task_tag	= login_req->itt;
		login->initial_exp_statsn = be32_to_cpu(login_req->exp_statsn);
		login->cid		= be16_to_cpu(login_req->cid);
		login->tsih		= be16_to_cpu(login_req->tsih);
	}

	memcpy(&login->req[0], (void *)&rx_desc->iscsi_header, ISCSI_HDR_LEN);

	size = min(rx_buflen, MAX_KEY_VALUE_PAIRS);
1279 1280 1281
	isert_dbg("Using login payload size: %d, rx_buflen: %d "
		  "MAX_KEY_VALUE_PAIRS: %d\n", size, rx_buflen,
		  MAX_KEY_VALUE_PAIRS);
1282 1283
	memcpy(login->req_buf, &rx_desc->data[0], size);

1284
	if (login->first_request) {
1285
		complete(&isert_conn->login_comp);
1286 1287 1288
		return;
	}
	schedule_delayed_work(&conn->login_work, 0);
1289 1290 1291
}

static struct iscsi_cmd
1292
*isert_allocate_cmd(struct iscsi_conn *conn, struct iser_rx_desc *rx_desc)
1293
{
1294
	struct isert_conn *isert_conn = conn->context;
1295
	struct isert_cmd *isert_cmd;
1296
	struct iscsi_cmd *cmd;
1297

1298
	cmd = iscsit_allocate_cmd(conn, TASK_INTERRUPTIBLE);
1299
	if (!cmd) {
1300
		isert_err("Unable to allocate iscsi_cmd + isert_cmd\n");
1301 1302
		return NULL;
	}
1303
	isert_cmd = iscsit_priv_cmd(cmd);
1304
	isert_cmd->conn = isert_conn;
1305
	isert_cmd->iscsi_cmd = cmd;
1306
	isert_cmd->rx_desc = rx_desc;
1307

1308
	return cmd;
1309 1310 1311 1312
}

static int
isert_handle_scsi_cmd(struct isert_conn *isert_conn,
1313 1314
		      struct isert_cmd *isert_cmd, struct iscsi_cmd *cmd,
		      struct iser_rx_desc *rx_desc, unsigned char *buf)
1315 1316 1317 1318 1319
{
	struct iscsi_conn *conn = isert_conn->conn;
	struct iscsi_scsi_req *hdr = (struct iscsi_scsi_req *)buf;
	int imm_data, imm_data_len, unsol_data, sg_nents, rc;
	bool dump_payload = false;
1320
	unsigned int data_len;
1321 1322 1323 1324 1325 1326 1327 1328

	rc = iscsit_setup_scsi_cmd(conn, cmd, buf);
	if (rc < 0)
		return rc;

	imm_data = cmd->immediate_data;
	imm_data_len = cmd->first_burst_len;
	unsol_data = cmd->unsolicited_data;
1329
	data_len = cmd->se_cmd.data_length;
1330

1331 1332
	if (imm_data && imm_data_len == data_len)
		cmd->se_cmd.se_cmd_flags |= SCF_PASSTHROUGH_SG_TO_MEM_NOALLOC;
1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343
	rc = iscsit_process_scsi_cmd(conn, cmd, hdr);
	if (rc < 0) {
		return 0;
	} else if (rc > 0) {
		dump_payload = true;
		goto sequence_cmd;
	}

	if (!imm_data)
		return 0;

1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357
	if (imm_data_len != data_len) {
		sg_nents = max(1UL, DIV_ROUND_UP(imm_data_len, PAGE_SIZE));
		sg_copy_from_buffer(cmd->se_cmd.t_data_sg, sg_nents,
				    &rx_desc->data[0], imm_data_len);
		isert_dbg("Copy Immediate sg_nents: %u imm_data_len: %d\n",
			  sg_nents, imm_data_len);
	} else {
		sg_init_table(&isert_cmd->sg, 1);
		cmd->se_cmd.t_data_sg = &isert_cmd->sg;
		cmd->se_cmd.t_data_nents = 1;
		sg_set_buf(&isert_cmd->sg, &rx_desc->data[0], imm_data_len);
		isert_dbg("Transfer Immediate imm_data_len: %d\n",
			  imm_data_len);
	}
1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368

	cmd->write_data_done += imm_data_len;

	if (cmd->write_data_done == cmd->se_cmd.data_length) {
		spin_lock_bh(&cmd->istate_lock);
		cmd->cmd_flags |= ICF_GOT_LAST_DATAOUT;
		cmd->i_state = ISTATE_RECEIVED_LAST_DATAOUT;
		spin_unlock_bh(&cmd->istate_lock);
	}

sequence_cmd:
1369
	rc = iscsit_sequence_cmd(conn, cmd, buf, hdr->cmdsn);
1370 1371 1372

	if (!rc && dump_payload == false && unsol_data)
		iscsit_set_unsoliticed_dataout(cmd);
1373
	else if (dump_payload && imm_data)
1374
		target_put_sess_cmd(&cmd->se_cmd);
1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398

	return 0;
}

static int
isert_handle_iscsi_dataout(struct isert_conn *isert_conn,
			   struct iser_rx_desc *rx_desc, unsigned char *buf)
{
	struct scatterlist *sg_start;
	struct iscsi_conn *conn = isert_conn->conn;
	struct iscsi_cmd *cmd = NULL;
	struct iscsi_data *hdr = (struct iscsi_data *)buf;
	u32 unsol_data_len = ntoh24(hdr->dlength);
	int rc, sg_nents, sg_off, page_off;

	rc = iscsit_check_dataout_hdr(conn, buf, &cmd);
	if (rc < 0)
		return rc;
	else if (!cmd)
		return 0;
	/*
	 * FIXME: Unexpected unsolicited_data out
	 */
	if (!cmd->unsolicited_data) {
1399
		isert_err("Received unexpected solicited data payload\n");
1400 1401 1402 1403
		dump_stack();
		return -1;
	}

1404 1405 1406 1407
	isert_dbg("Unsolicited DataOut unsol_data_len: %u, "
		  "write_data_done: %u, data_length: %u\n",
		  unsol_data_len,  cmd->write_data_done,
		  cmd->se_cmd.data_length);
1408 1409 1410 1411 1412 1413 1414 1415 1416

	sg_off = cmd->write_data_done / PAGE_SIZE;
	sg_start = &cmd->se_cmd.t_data_sg[sg_off];
	sg_nents = max(1UL, DIV_ROUND_UP(unsol_data_len, PAGE_SIZE));
	page_off = cmd->write_data_done % PAGE_SIZE;
	/*
	 * FIXME: Non page-aligned unsolicited_data out
	 */
	if (page_off) {
1417
		isert_err("unexpected non-page aligned data payload\n");
1418 1419 1420
		dump_stack();
		return -1;
	}
1421 1422 1423
	isert_dbg("Copying DataOut: sg_start: %p, sg_off: %u "
		  "sg_nents: %u from %p %u\n", sg_start, sg_off,
		  sg_nents, &rx_desc->data[0], unsol_data_len);
1424 1425 1426 1427 1428 1429 1430 1431

	sg_copy_from_buffer(sg_start, sg_nents, &rx_desc->data[0],
			    unsol_data_len);

	rc = iscsit_check_dataout_payload(cmd, hdr, false);
	if (rc < 0)
		return rc;

1432 1433 1434 1435 1436 1437 1438 1439 1440
	/*
	 * multiple data-outs on the same command can arrive -
	 * so post the buffer before hand
	 */
	rc = isert_post_recv(isert_conn, rx_desc);
	if (rc) {
		isert_err("ib_post_recv failed with %d\n", rc);
		return rc;
	}
1441 1442 1443
	return 0;
}

1444 1445
static int
isert_handle_nop_out(struct isert_conn *isert_conn, struct isert_cmd *isert_cmd,
1446 1447
		     struct iscsi_cmd *cmd, struct iser_rx_desc *rx_desc,
		     unsigned char *buf)
1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462
{
	struct iscsi_conn *conn = isert_conn->conn;
	struct iscsi_nopout *hdr = (struct iscsi_nopout *)buf;
	int rc;

	rc = iscsit_setup_nop_out(conn, cmd, hdr);
	if (rc < 0)
		return rc;
	/*
	 * FIXME: Add support for NOPOUT payload using unsolicited RDMA payload
	 */

	return iscsit_process_nop_out(conn, cmd, hdr);
}

1463 1464
static int
isert_handle_text_cmd(struct isert_conn *isert_conn, struct isert_cmd *isert_cmd,
1465 1466
		      struct iscsi_cmd *cmd, struct iser_rx_desc *rx_desc,
		      struct iscsi_text *hdr)
1467 1468 1469 1470
{
	struct iscsi_conn *conn = isert_conn->conn;
	u32 payload_length = ntoh24(hdr->dlength);
	int rc;
1471
	unsigned char *text_in = NULL;
1472 1473 1474 1475 1476

	rc = iscsit_setup_text_cmd(conn, cmd, hdr);
	if (rc < 0)
		return rc;

1477 1478 1479 1480 1481 1482 1483
	if (payload_length) {
		text_in = kzalloc(payload_length, GFP_KERNEL);
		if (!text_in) {
			isert_err("Unable to allocate text_in of payload_length: %u\n",
				  payload_length);
			return -ENOMEM;
		}
1484 1485 1486 1487 1488 1489 1490 1491
	}
	cmd->text_in_ptr = text_in;

	memcpy(cmd->text_in_ptr, &rx_desc->data[0], payload_length);

	return iscsit_process_text_cmd(conn, cmd, hdr);
}

1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503
static int
isert_rx_opcode(struct isert_conn *isert_conn, struct iser_rx_desc *rx_desc,
		uint32_t read_stag, uint64_t read_va,
		uint32_t write_stag, uint64_t write_va)
{
	struct iscsi_hdr *hdr = &rx_desc->iscsi_header;
	struct iscsi_conn *conn = isert_conn->conn;
	struct iscsi_cmd *cmd;
	struct isert_cmd *isert_cmd;
	int ret = -EINVAL;
	u8 opcode = (hdr->opcode & ISCSI_OPCODE_MASK);

1504
	if (conn->sess->sess_ops->SessionType &&
1505
	   (!(opcode & ISCSI_OP_TEXT) || !(opcode & ISCSI_OP_LOGOUT))) {
1506
		isert_err("Got illegal opcode: 0x%02x in SessionType=Discovery,"
1507
			  " ignoring\n", opcode);
1508 1509 1510
		return 0;
	}

1511 1512
	switch (opcode) {
	case ISCSI_OP_SCSI_CMD:
1513
		cmd = isert_allocate_cmd(conn, rx_desc);
1514 1515 1516
		if (!cmd)
			break;

1517
		isert_cmd = iscsit_priv_cmd(cmd);
1518 1519 1520 1521 1522
		isert_cmd->read_stag = read_stag;
		isert_cmd->read_va = read_va;
		isert_cmd->write_stag = write_stag;
		isert_cmd->write_va = write_va;

1523
		ret = isert_handle_scsi_cmd(isert_conn, isert_cmd, cmd,
1524 1525 1526
					rx_desc, (unsigned char *)hdr);
		break;
	case ISCSI_OP_NOOP_OUT:
1527
		cmd = isert_allocate_cmd(conn, rx_desc);
1528 1529 1530
		if (!cmd)
			break;

1531 1532
		isert_cmd = iscsit_priv_cmd(cmd);
		ret = isert_handle_nop_out(isert_conn, isert_cmd, cmd,
1533
					   rx_desc, (unsigned char *)hdr);
1534 1535 1536 1537 1538 1539
		break;
	case ISCSI_OP_SCSI_DATA_OUT:
		ret = isert_handle_iscsi_dataout(isert_conn, rx_desc,
						(unsigned char *)hdr);
		break;
	case ISCSI_OP_SCSI_TMFUNC:
1540
		cmd = isert_allocate_cmd(conn, rx_desc);
1541 1542 1543 1544 1545 1546 1547
		if (!cmd)
			break;

		ret = iscsit_handle_task_mgt_cmd(conn, cmd,
						(unsigned char *)hdr);
		break;
	case ISCSI_OP_LOGOUT:
1548
		cmd = isert_allocate_cmd(conn, rx_desc);
1549 1550 1551 1552 1553
		if (!cmd)
			break;

		ret = iscsit_handle_logout_cmd(conn, cmd, (unsigned char *)hdr);
		break;
1554
	case ISCSI_OP_TEXT:
1555
		if (be32_to_cpu(hdr->ttt) != 0xFFFFFFFF)
1556
			cmd = iscsit_find_cmd_from_itt(conn, hdr->itt);
1557 1558 1559 1560 1561
		else
			cmd = isert_allocate_cmd(conn, rx_desc);

		if (!cmd)
			break;
1562

1563 1564
		isert_cmd = iscsit_priv_cmd(cmd);
		ret = isert_handle_text_cmd(isert_conn, isert_cmd, cmd,
1565 1566
					    rx_desc, (struct iscsi_text *)hdr);
		break;
1567
	default:
1568
		isert_err("Got unknown iSCSI OpCode: 0x%02x\n", opcode);
1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587
		dump_stack();
		break;
	}

	return ret;
}

static void
isert_rx_do_work(struct iser_rx_desc *rx_desc, struct isert_conn *isert_conn)
{
	struct iser_hdr *iser_hdr = &rx_desc->iser_header;
	uint64_t read_va = 0, write_va = 0;
	uint32_t read_stag = 0, write_stag = 0;

	switch (iser_hdr->flags & 0xF0) {
	case ISCSI_CTRL:
		if (iser_hdr->flags & ISER_RSV) {
			read_stag = be32_to_cpu(iser_hdr->read_stag);
			read_va = be64_to_cpu(iser_hdr->read_va);
1588 1589
			isert_dbg("ISER_RSV: read_stag: 0x%x read_va: 0x%llx\n",
				  read_stag, (unsigned long long)read_va);
1590 1591 1592 1593
		}
		if (iser_hdr->flags & ISER_WSV) {
			write_stag = be32_to_cpu(iser_hdr->write_stag);
			write_va = be64_to_cpu(iser_hdr->write_va);
1594 1595
			isert_dbg("ISER_WSV: write_stag: 0x%x write_va: 0x%llx\n",
				  write_stag, (unsigned long long)write_va);
1596 1597
		}

1598
		isert_dbg("ISER ISCSI_CTRL PDU\n");
1599 1600
		break;
	case ISER_HELLO:
1601
		isert_err("iSER Hello message\n");
1602 1603
		break;
	default:
1604
		isert_warn("Unknown iSER hdr flags: 0x%02x\n", iser_hdr->flags);
1605 1606 1607
		break;
	}

1608 1609
	isert_rx_opcode(isert_conn, rx_desc,
			read_stag, read_va, write_stag, write_va);
1610 1611 1612
}

static void
1613 1614 1615
isert_rcv_completion(struct iser_rx_desc *desc,
		     struct isert_conn *isert_conn,
		     u32 xfer_len)
1616
{
1617
	struct ib_device *ib_dev = isert_conn->cm_id->device;
1618 1619
	struct iscsi_hdr *hdr;
	u64 rx_dma;
1620
	int rx_buflen;
1621 1622 1623 1624

	if ((char *)desc == isert_conn->login_req_buf) {
		rx_dma = isert_conn->login_req_dma;
		rx_buflen = ISER_RX_LOGIN_SIZE;
1625
		isert_dbg("login_buf: Using rx_dma: 0x%llx, rx_buflen: %d\n",
1626 1627 1628 1629
			 rx_dma, rx_buflen);
	} else {
		rx_dma = desc->dma_addr;
		rx_buflen = ISER_RX_PAYLOAD_SIZE;
1630
		isert_dbg("req_buf: Using rx_dma: 0x%llx, rx_buflen: %d\n",
1631 1632 1633 1634 1635 1636
			 rx_dma, rx_buflen);
	}

	ib_dma_sync_single_for_cpu(ib_dev, rx_dma, rx_buflen, DMA_FROM_DEVICE);

	hdr = &desc->iscsi_header;
1637
	isert_dbg("iSCSI opcode: 0x%02x, ITT: 0x%08x, flags: 0x%02x dlen: %d\n",
1638 1639 1640
		 hdr->opcode, hdr->itt, hdr->flags,
		 (int)(xfer_len - ISER_HEADERS_LEN));

1641 1642 1643 1644 1645 1646 1647 1648
	if ((char *)desc == isert_conn->login_req_buf) {
		isert_conn->login_req_len = xfer_len - ISER_HEADERS_LEN;
		if (isert_conn->conn) {
			struct iscsi_login *login = isert_conn->conn->conn_login;

			if (login && !login->first_request)
				isert_rx_login_req(isert_conn);
		}
1649
		mutex_lock(&isert_conn->mutex);
1650
		complete(&isert_conn->login_req_comp);
1651
		mutex_unlock(&isert_conn->mutex);
1652
	} else {
1653
		isert_rx_do_work(desc, isert_conn);
1654
	}
1655 1656 1657 1658 1659 1660 1661

	ib_dma_sync_single_for_device(ib_dev, rx_dma, rx_buflen,
				      DMA_FROM_DEVICE);

	isert_conn->post_recv_buf_count--;
}

1662 1663 1664 1665 1666
static int
isert_map_data_buf(struct isert_conn *isert_conn, struct isert_cmd *isert_cmd,
		   struct scatterlist *sg, u32 nents, u32 length, u32 offset,
		   enum iser_ib_op_code op, struct isert_data_buf *data)
{
1667
	struct ib_device *ib_dev = isert_conn->cm_id->device;
1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684

	data->dma_dir = op == ISER_IB_RDMA_WRITE ?
			      DMA_TO_DEVICE : DMA_FROM_DEVICE;

	data->len = length - offset;
	data->offset = offset;
	data->sg_off = data->offset / PAGE_SIZE;

	data->sg = &sg[data->sg_off];
	data->nents = min_t(unsigned int, nents - data->sg_off,
					  ISCSI_ISER_SG_TABLESIZE);
	data->len = min_t(unsigned int, data->len, ISCSI_ISER_SG_TABLESIZE *
					PAGE_SIZE);

	data->dma_nents = ib_dma_map_sg(ib_dev, data->sg, data->nents,
					data->dma_dir);
	if (unlikely(!data->dma_nents)) {
1685
		isert_err("Cmd: unable to dma map SGs %p\n", sg);
1686 1687 1688
		return -EINVAL;
	}

1689
	isert_dbg("Mapped cmd: %p count: %u sg: %p sg_nents: %u rdma_len %d\n",
1690
		  isert_cmd, data->dma_nents, data->sg, data->nents, data->len);
1691 1692 1693 1694 1695 1696 1697

	return 0;
}

static void
isert_unmap_data_buf(struct isert_conn *isert_conn, struct isert_data_buf *data)
{
1698
	struct ib_device *ib_dev = isert_conn->cm_id->device;
1699 1700 1701 1702 1703 1704 1705

	ib_dma_unmap_sg(ib_dev, data->sg, data->nents, data->dma_dir);
	memset(data, 0, sizeof(*data));
}



1706 1707 1708 1709 1710
static void
isert_unmap_cmd(struct isert_cmd *isert_cmd, struct isert_conn *isert_conn)
{
	struct isert_rdma_wr *wr = &isert_cmd->rdma_wr;

1711
	isert_dbg("Cmd %p\n", isert_cmd);
1712 1713

	if (wr->data.sg) {
1714
		isert_dbg("Cmd %p unmap_sg op\n", isert_cmd);
1715
		isert_unmap_data_buf(isert_conn, &wr->data);
1716 1717
	}

C
Christoph Hellwig 已提交
1718
	if (wr->rdma_wr) {
1719
		isert_dbg("Cmd %p free send_wr\n", isert_cmd);
C
Christoph Hellwig 已提交
1720 1721
		kfree(wr->rdma_wr);
		wr->rdma_wr = NULL;
1722
	}
1723

1724
	if (wr->ib_sge) {
1725
		isert_dbg("Cmd %p free ib_sge\n", isert_cmd);
1726 1727 1728
		kfree(wr->ib_sge);
		wr->ib_sge = NULL;
	}
1729 1730
}

1731
static void
1732
isert_unreg_rdma(struct isert_cmd *isert_cmd, struct isert_conn *isert_conn)
1733 1734 1735
{
	struct isert_rdma_wr *wr = &isert_cmd->rdma_wr;

1736
	isert_dbg("Cmd %p\n", isert_cmd);
1737 1738

	if (wr->fr_desc) {
1739
		isert_dbg("Cmd %p free fr_desc %p\n", isert_cmd, wr->fr_desc);
1740 1741 1742 1743
		if (wr->fr_desc->ind & ISERT_PROTECTED) {
			isert_unmap_data_buf(isert_conn, &wr->prot);
			wr->fr_desc->ind &= ~ISERT_PROTECTED;
		}
1744 1745 1746
		spin_lock_bh(&isert_conn->pool_lock);
		list_add_tail(&wr->fr_desc->list, &isert_conn->fr_pool);
		spin_unlock_bh(&isert_conn->pool_lock);
1747 1748 1749
		wr->fr_desc = NULL;
	}

1750
	if (wr->data.sg) {
1751
		isert_dbg("Cmd %p unmap_sg op\n", isert_cmd);
1752
		isert_unmap_data_buf(isert_conn, &wr->data);
1753 1754 1755
	}

	wr->ib_sge = NULL;
C
Christoph Hellwig 已提交
1756
	wr->rdma_wr = NULL;
1757 1758
}

1759
static void
1760
isert_put_cmd(struct isert_cmd *isert_cmd, bool comp_err)
1761
{
1762
	struct iscsi_cmd *cmd = isert_cmd->iscsi_cmd;
1763
	struct isert_conn *isert_conn = isert_cmd->conn;
1764
	struct iscsi_conn *conn = isert_conn->conn;
1765
	struct isert_device *device = isert_conn->device;
1766
	struct iscsi_text_rsp *hdr;
1767

1768
	isert_dbg("Cmd %p\n", isert_cmd);
1769 1770 1771 1772 1773

	switch (cmd->iscsi_opcode) {
	case ISCSI_OP_SCSI_CMD:
		spin_lock_bh(&conn->cmd_lock);
		if (!list_empty(&cmd->i_conn_node))
1774
			list_del_init(&cmd->i_conn_node);
1775 1776
		spin_unlock_bh(&conn->cmd_lock);

1777
		if (cmd->data_direction == DMA_TO_DEVICE) {
1778
			iscsit_stop_dataout_timer(cmd);
1779 1780 1781 1782 1783 1784 1785 1786 1787 1788
			/*
			 * Check for special case during comp_err where
			 * WRITE_PENDING has been handed off from core,
			 * but requires an extra target_put_sess_cmd()
			 * before transport_generic_free_cmd() below.
			 */
			if (comp_err &&
			    cmd->se_cmd.t_state == TRANSPORT_WRITE_PENDING) {
				struct se_cmd *se_cmd = &cmd->se_cmd;

1789
				target_put_sess_cmd(se_cmd);
1790 1791
			}
		}
1792

1793
		device->unreg_rdma_mem(isert_cmd, isert_conn);
1794 1795
		transport_generic_free_cmd(&cmd->se_cmd, 0);
		break;
1796
	case ISCSI_OP_SCSI_TMFUNC:
1797 1798
		spin_lock_bh(&conn->cmd_lock);
		if (!list_empty(&cmd->i_conn_node))
1799
			list_del_init(&cmd->i_conn_node);
1800 1801
		spin_unlock_bh(&conn->cmd_lock);

1802 1803 1804 1805
		transport_generic_free_cmd(&cmd->se_cmd, 0);
		break;
	case ISCSI_OP_REJECT:
	case ISCSI_OP_NOOP_OUT:
1806
	case ISCSI_OP_TEXT:
1807 1808 1809 1810 1811
		hdr = (struct iscsi_text_rsp *)&isert_cmd->tx_desc.iscsi_header;
		/* If the continue bit is on, keep the command alive */
		if (hdr->flags & ISCSI_FLAG_TEXT_CONTINUE)
			break;

1812 1813
		spin_lock_bh(&conn->cmd_lock);
		if (!list_empty(&cmd->i_conn_node))
1814
			list_del_init(&cmd->i_conn_node);
1815 1816 1817 1818 1819 1820 1821 1822
		spin_unlock_bh(&conn->cmd_lock);

		/*
		 * Handle special case for REJECT when iscsi_add_reject*() has
		 * overwritten the original iscsi_opcode assignment, and the
		 * associated cmd->se_cmd needs to be released.
		 */
		if (cmd->se_cmd.se_tfo != NULL) {
1823
			isert_dbg("Calling transport_generic_free_cmd for 0x%02x\n",
1824
				 cmd->iscsi_opcode);
1825 1826 1827 1828 1829 1830 1831
			transport_generic_free_cmd(&cmd->se_cmd, 0);
			break;
		}
		/*
		 * Fall-through
		 */
	default:
1832
		iscsit_release_cmd(cmd);
1833 1834 1835 1836 1837 1838 1839 1840
		break;
	}
}

static void
isert_unmap_tx_desc(struct iser_tx_desc *tx_desc, struct ib_device *ib_dev)
{
	if (tx_desc->dma_addr != 0) {
1841
		isert_dbg("unmap single for tx_desc->dma_addr\n");
1842 1843 1844 1845 1846 1847 1848 1849
		ib_dma_unmap_single(ib_dev, tx_desc->dma_addr,
				    ISER_HEADERS_LEN, DMA_TO_DEVICE);
		tx_desc->dma_addr = 0;
	}
}

static void
isert_completion_put(struct iser_tx_desc *tx_desc, struct isert_cmd *isert_cmd,
1850
		     struct ib_device *ib_dev, bool comp_err)
1851
{
1852
	if (isert_cmd->pdu_buf_dma != 0) {
1853
		isert_dbg("unmap single for isert_cmd->pdu_buf_dma\n");
1854 1855 1856
		ib_dma_unmap_single(ib_dev, isert_cmd->pdu_buf_dma,
				    isert_cmd->pdu_buf_len, DMA_TO_DEVICE);
		isert_cmd->pdu_buf_dma = 0;
1857 1858 1859
	}

	isert_unmap_tx_desc(tx_desc, ib_dev);
1860
	isert_put_cmd(isert_cmd, comp_err);
1861 1862
}

1863 1864 1865 1866 1867 1868 1869 1870
static int
isert_check_pi_status(struct se_cmd *se_cmd, struct ib_mr *sig_mr)
{
	struct ib_mr_status mr_status;
	int ret;

	ret = ib_check_mr_status(sig_mr, IB_MR_CHECK_SIG_STATUS, &mr_status);
	if (ret) {
1871
		isert_err("ib_check_mr_status failed, ret %d\n", ret);
1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893
		goto fail_mr_status;
	}

	if (mr_status.fail_status & IB_MR_CHECK_SIG_STATUS) {
		u64 sec_offset_err;
		u32 block_size = se_cmd->se_dev->dev_attrib.block_size + 8;

		switch (mr_status.sig_err.err_type) {
		case IB_SIG_BAD_GUARD:
			se_cmd->pi_err = TCM_LOGICAL_BLOCK_GUARD_CHECK_FAILED;
			break;
		case IB_SIG_BAD_REFTAG:
			se_cmd->pi_err = TCM_LOGICAL_BLOCK_REF_TAG_CHECK_FAILED;
			break;
		case IB_SIG_BAD_APPTAG:
			se_cmd->pi_err = TCM_LOGICAL_BLOCK_APP_TAG_CHECK_FAILED;
			break;
		}
		sec_offset_err = mr_status.sig_err.sig_err_offset;
		do_div(sec_offset_err, block_size);
		se_cmd->bad_sector = sec_offset_err + se_cmd->t_task_lba;

1894 1895 1896 1897 1898 1899
		isert_err("PI error found type %d at sector 0x%llx "
			  "expected 0x%x vs actual 0x%x\n",
			  mr_status.sig_err.err_type,
			  (unsigned long long)se_cmd->bad_sector,
			  mr_status.sig_err.expected,
			  mr_status.sig_err.actual);
1900 1901 1902 1903 1904 1905 1906
		ret = 1;
	}

fail_mr_status:
	return ret;
}

1907 1908 1909 1910
static void
isert_completion_rdma_write(struct iser_tx_desc *tx_desc,
			    struct isert_cmd *isert_cmd)
{
1911
	struct isert_rdma_wr *wr = &isert_cmd->rdma_wr;
1912
	struct iscsi_cmd *cmd = isert_cmd->iscsi_cmd;
1913
	struct se_cmd *se_cmd = &cmd->se_cmd;
1914
	struct isert_conn *isert_conn = isert_cmd->conn;
1915
	struct isert_device *device = isert_conn->device;
1916 1917 1918
	int ret = 0;

	if (wr->fr_desc && wr->fr_desc->ind & ISERT_PROTECTED) {
1919 1920 1921
		ret = isert_check_pi_status(se_cmd,
					    wr->fr_desc->pi_ctx->sig_mr);
		wr->fr_desc->ind &= ~ISERT_PROTECTED;
1922
	}
1923 1924

	device->unreg_rdma_mem(isert_cmd, isert_conn);
C
Christoph Hellwig 已提交
1925
	wr->rdma_wr_num = 0;
1926 1927 1928 1929 1930
	if (ret)
		transport_send_check_condition_and_sense(se_cmd,
							 se_cmd->pi_err, 0);
	else
		isert_put_response(isert_conn->conn, cmd);
1931 1932
}

1933 1934 1935 1936 1937
static void
isert_completion_rdma_read(struct iser_tx_desc *tx_desc,
			   struct isert_cmd *isert_cmd)
{
	struct isert_rdma_wr *wr = &isert_cmd->rdma_wr;
1938
	struct iscsi_cmd *cmd = isert_cmd->iscsi_cmd;
1939
	struct se_cmd *se_cmd = &cmd->se_cmd;
1940
	struct isert_conn *isert_conn = isert_cmd->conn;
1941
	struct isert_device *device = isert_conn->device;
1942
	int ret = 0;
1943

1944
	if (wr->fr_desc && wr->fr_desc->ind & ISERT_PROTECTED) {
1945 1946 1947
		ret = isert_check_pi_status(se_cmd,
					    wr->fr_desc->pi_ctx->sig_mr);
		wr->fr_desc->ind &= ~ISERT_PROTECTED;
1948 1949
	}

1950
	iscsit_stop_dataout_timer(cmd);
1951
	device->unreg_rdma_mem(isert_cmd, isert_conn);
1952
	cmd->write_data_done = wr->data.len;
C
Christoph Hellwig 已提交
1953
	wr->rdma_wr_num = 0;
1954

1955
	isert_dbg("Cmd: %p RDMA_READ comp calling execute_cmd\n", isert_cmd);
1956 1957 1958 1959 1960
	spin_lock_bh(&cmd->istate_lock);
	cmd->cmd_flags |= ICF_GOT_LAST_DATAOUT;
	cmd->i_state = ISTATE_RECEIVED_LAST_DATAOUT;
	spin_unlock_bh(&cmd->istate_lock);

1961
	if (ret) {
1962
		target_put_sess_cmd(se_cmd);
1963 1964
		transport_send_check_condition_and_sense(se_cmd,
							 se_cmd->pi_err, 0);
1965
	} else {
1966
		target_execute_cmd(se_cmd);
1967
	}
1968 1969 1970 1971 1972 1973 1974 1975
}

static void
isert_do_control_comp(struct work_struct *work)
{
	struct isert_cmd *isert_cmd = container_of(work,
			struct isert_cmd, comp_work);
	struct isert_conn *isert_conn = isert_cmd->conn;
1976
	struct ib_device *ib_dev = isert_conn->cm_id->device;
1977
	struct iscsi_cmd *cmd = isert_cmd->iscsi_cmd;
1978

1979 1980
	isert_dbg("Cmd %p i_state %d\n", isert_cmd, cmd->i_state);

1981 1982 1983
	switch (cmd->i_state) {
	case ISTATE_SEND_TASKMGTRSP:
		iscsit_tmr_post_handler(cmd, cmd->conn);
1984 1985
	case ISTATE_SEND_REJECT:   /* FALLTHRU */
	case ISTATE_SEND_TEXTRSP:  /* FALLTHRU */
1986
		cmd->i_state = ISTATE_SENT_STATUS;
1987 1988
		isert_completion_put(&isert_cmd->tx_desc, isert_cmd,
				     ib_dev, false);
1989
		break;
1990 1991 1992 1993
	case ISTATE_SEND_LOGOUTRSP:
		iscsit_logout_post_handler(cmd, cmd->conn);
		break;
	default:
1994
		isert_err("Unknown i_state %d\n", cmd->i_state);
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
		dump_stack();
		break;
	}
}

static void
isert_response_completion(struct iser_tx_desc *tx_desc,
			  struct isert_cmd *isert_cmd,
			  struct isert_conn *isert_conn,
			  struct ib_device *ib_dev)
{
2006
	struct iscsi_cmd *cmd = isert_cmd->iscsi_cmd;
2007 2008

	if (cmd->i_state == ISTATE_SEND_TASKMGTRSP ||
2009
	    cmd->i_state == ISTATE_SEND_LOGOUTRSP ||
2010 2011
	    cmd->i_state == ISTATE_SEND_REJECT ||
	    cmd->i_state == ISTATE_SEND_TEXTRSP) {
2012 2013 2014 2015 2016 2017
		isert_unmap_tx_desc(tx_desc, ib_dev);

		INIT_WORK(&isert_cmd->comp_work, isert_do_control_comp);
		queue_work(isert_comp_wq, &isert_cmd->comp_work);
		return;
	}
2018

2019
	cmd->i_state = ISTATE_SENT_STATUS;
2020
	isert_completion_put(tx_desc, isert_cmd, ib_dev, false);
2021 2022 2023
}

static void
2024
isert_snd_completion(struct iser_tx_desc *tx_desc,
2025
		      struct isert_conn *isert_conn)
2026
{
2027
	struct ib_device *ib_dev = isert_conn->cm_id->device;
2028 2029 2030 2031 2032 2033 2034 2035 2036
	struct isert_cmd *isert_cmd = tx_desc->isert_cmd;
	struct isert_rdma_wr *wr;

	if (!isert_cmd) {
		isert_unmap_tx_desc(tx_desc, ib_dev);
		return;
	}
	wr = &isert_cmd->rdma_wr;

2037 2038
	isert_dbg("Cmd %p iser_ib_op %d\n", isert_cmd, wr->iser_ib_op);

2039 2040 2041 2042 2043 2044
	switch (wr->iser_ib_op) {
	case ISER_IB_SEND:
		isert_response_completion(tx_desc, isert_cmd,
					  isert_conn, ib_dev);
		break;
	case ISER_IB_RDMA_WRITE:
2045
		isert_completion_rdma_write(tx_desc, isert_cmd);
2046 2047 2048 2049 2050
		break;
	case ISER_IB_RDMA_READ:
		isert_completion_rdma_read(tx_desc, isert_cmd);
		break;
	default:
2051
		isert_err("Unknown wr->iser_ib_op: 0x%x\n", wr->iser_ib_op);
2052 2053 2054 2055 2056
		dump_stack();
		break;
	}
}

2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070
/**
 * is_isert_tx_desc() - Indicate if the completion wr_id
 *     is a TX descriptor or not.
 * @isert_conn: iser connection
 * @wr_id: completion WR identifier
 *
 * Since we cannot rely on wc opcode in FLUSH errors
 * we must work around it by checking if the wr_id address
 * falls in the iser connection rx_descs buffer. If so
 * it is an RX descriptor, otherwize it is a TX.
 */
static inline bool
is_isert_tx_desc(struct isert_conn *isert_conn, void *wr_id)
{
2071 2072
	void *start = isert_conn->rx_descs;
	int len = ISERT_QP_MAX_RECV_DTOS * sizeof(*isert_conn->rx_descs);
2073 2074 2075 2076 2077 2078 2079

	if (wr_id >= start && wr_id < start + len)
		return false;

	return true;
}

2080
static void
2081
isert_cq_comp_err(struct isert_conn *isert_conn, struct ib_wc *wc)
2082
{
2083
	if (wc->wr_id == ISER_BEACON_WRID) {
2084
		isert_info("conn %p completing wait_comp_err\n",
2085
			   isert_conn);
2086
		complete(&isert_conn->wait_comp_err);
2087
	} else if (is_isert_tx_desc(isert_conn, (void *)(uintptr_t)wc->wr_id)) {
2088
		struct ib_device *ib_dev = isert_conn->cm_id->device;
2089
		struct isert_cmd *isert_cmd;
2090
		struct iser_tx_desc *desc;
2091

2092 2093
		desc = (struct iser_tx_desc *)(uintptr_t)wc->wr_id;
		isert_cmd = desc->isert_cmd;
2094 2095 2096 2097 2098 2099
		if (!isert_cmd)
			isert_unmap_tx_desc(desc, ib_dev);
		else
			isert_completion_put(desc, isert_cmd, ib_dev, true);
	} else {
		isert_conn->post_recv_buf_count--;
2100 2101
		if (!isert_conn->post_recv_buf_count)
			iscsit_cause_connection_reinstatement(isert_conn->conn, 0);
2102
	}
2103 2104 2105
}

static void
2106
isert_handle_wc(struct ib_wc *wc)
2107 2108 2109
{
	struct isert_conn *isert_conn;
	struct iser_tx_desc *tx_desc;
2110
	struct iser_rx_desc *rx_desc;
2111

2112 2113 2114 2115
	isert_conn = wc->qp->qp_context;
	if (likely(wc->status == IB_WC_SUCCESS)) {
		if (wc->opcode == IB_WC_RECV) {
			rx_desc = (struct iser_rx_desc *)(uintptr_t)wc->wr_id;
2116
			isert_rcv_completion(rx_desc, isert_conn, wc->byte_len);
2117
		} else {
2118
			tx_desc = (struct iser_tx_desc *)(uintptr_t)wc->wr_id;
2119
			isert_snd_completion(tx_desc, isert_conn);
2120
		}
2121 2122
	} else {
		if (wc->status != IB_WC_WR_FLUSH_ERR)
2123 2124 2125
			isert_err("%s (%d): wr id %llx vend_err %x\n",
				  ib_wc_status_msg(wc->status), wc->status,
				  wc->wr_id, wc->vendor_err);
2126
		else
2127 2128 2129
			isert_dbg("%s (%d): wr id %llx\n",
				  ib_wc_status_msg(wc->status), wc->status,
				  wc->wr_id);
2130

2131 2132 2133
		if (wc->wr_id != ISER_FASTREG_LI_WRID)
			isert_cq_comp_err(isert_conn, wc);
	}
2134 2135 2136
}

static void
2137
isert_cq_work(struct work_struct *work)
2138
{
2139
	enum { isert_poll_budget = 65536 };
2140
	struct isert_comp *comp = container_of(work, struct isert_comp,
2141
					       work);
2142 2143
	struct ib_wc *const wcs = comp->wcs;
	int i, n, completed = 0;
2144

2145 2146 2147
	while ((n = ib_poll_cq(comp->cq, ARRAY_SIZE(comp->wcs), wcs)) > 0) {
		for (i = 0; i < n; i++)
			isert_handle_wc(&wcs[i]);
2148

2149 2150
		completed += n;
		if (completed >= isert_poll_budget)
2151 2152 2153
			break;
	}

2154
	ib_req_notify_cq(comp->cq, IB_CQ_NEXT_COMP);
2155 2156 2157
}

static void
2158
isert_cq_callback(struct ib_cq *cq, void *context)
2159
{
2160
	struct isert_comp *comp = context;
2161

2162
	queue_work(isert_comp_wq, &comp->work);
2163 2164 2165 2166 2167 2168 2169 2170
}

static int
isert_post_response(struct isert_conn *isert_conn, struct isert_cmd *isert_cmd)
{
	struct ib_send_wr *wr_failed;
	int ret;

2171 2172 2173 2174 2175 2176
	ret = isert_post_recv(isert_conn, isert_cmd->rx_desc);
	if (ret) {
		isert_err("ib_post_recv failed with %d\n", ret);
		return ret;
	}

2177
	ret = ib_post_send(isert_conn->qp, &isert_cmd->tx_desc.send_wr,
2178 2179
			   &wr_failed);
	if (ret) {
2180
		isert_err("ib_post_send failed with %d\n", ret);
2181 2182 2183 2184 2185 2186 2187 2188
		return ret;
	}
	return ret;
}

static int
isert_put_response(struct iscsi_conn *conn, struct iscsi_cmd *cmd)
{
2189
	struct isert_cmd *isert_cmd = iscsit_priv_cmd(cmd);
2190
	struct isert_conn *isert_conn = conn->context;
2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203
	struct ib_send_wr *send_wr = &isert_cmd->tx_desc.send_wr;
	struct iscsi_scsi_rsp *hdr = (struct iscsi_scsi_rsp *)
				&isert_cmd->tx_desc.iscsi_header;

	isert_create_send_desc(isert_conn, isert_cmd, &isert_cmd->tx_desc);
	iscsit_build_rsp_pdu(cmd, conn, true, hdr);
	isert_init_tx_hdrs(isert_conn, &isert_cmd->tx_desc);
	/*
	 * Attach SENSE DATA payload to iSCSI Response PDU
	 */
	if (cmd->se_cmd.sense_buffer &&
	    ((cmd->se_cmd.se_cmd_flags & SCF_TRANSPORT_TASK_SENSE) ||
	    (cmd->se_cmd.se_cmd_flags & SCF_EMULATED_TASK_SENSE))) {
2204
		struct isert_device *device = isert_conn->device;
2205
		struct ib_device *ib_dev = device->ib_device;
2206
		struct ib_sge *tx_dsg = &isert_cmd->tx_desc.tx_sg[1];
2207
		u32 padding, pdu_len;
2208 2209 2210 2211 2212 2213 2214

		put_unaligned_be16(cmd->se_cmd.scsi_sense_length,
				   cmd->sense_buffer);
		cmd->se_cmd.scsi_sense_length += sizeof(__be16);

		padding = -(cmd->se_cmd.scsi_sense_length) & 3;
		hton24(hdr->dlength, (u32)cmd->se_cmd.scsi_sense_length);
2215
		pdu_len = cmd->se_cmd.scsi_sense_length + padding;
2216

2217 2218
		isert_cmd->pdu_buf_dma = ib_dma_map_single(ib_dev,
				(void *)cmd->sense_buffer, pdu_len,
2219 2220
				DMA_TO_DEVICE);

2221 2222 2223
		isert_cmd->pdu_buf_len = pdu_len;
		tx_dsg->addr	= isert_cmd->pdu_buf_dma;
		tx_dsg->length	= pdu_len;
2224
		tx_dsg->lkey	= device->pd->local_dma_lkey;
2225 2226 2227
		isert_cmd->tx_desc.num_sge = 2;
	}

2228
	isert_init_send_wr(isert_conn, isert_cmd, send_wr);
2229

2230
	isert_dbg("Posting SCSI Response\n");
2231 2232 2233 2234

	return isert_post_response(isert_conn, isert_cmd);
}

2235 2236 2237 2238
static void
isert_aborted_task(struct iscsi_conn *conn, struct iscsi_cmd *cmd)
{
	struct isert_cmd *isert_cmd = iscsit_priv_cmd(cmd);
2239
	struct isert_conn *isert_conn = conn->context;
2240
	struct isert_device *device = isert_conn->device;
2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252

	spin_lock_bh(&conn->cmd_lock);
	if (!list_empty(&cmd->i_conn_node))
		list_del_init(&cmd->i_conn_node);
	spin_unlock_bh(&conn->cmd_lock);

	if (cmd->data_direction == DMA_TO_DEVICE)
		iscsit_stop_dataout_timer(cmd);

	device->unreg_rdma_mem(isert_cmd, isert_conn);
}

2253 2254 2255
static enum target_prot_op
isert_get_sup_prot_ops(struct iscsi_conn *conn)
{
2256
	struct isert_conn *isert_conn = conn->context;
2257
	struct isert_device *device = isert_conn->device;
2258

2259 2260
	if (conn->tpg->tpg_attrib.t10_pi) {
		if (device->pi_capable) {
2261
			isert_info("conn %p PI offload enabled\n", isert_conn);
2262 2263 2264 2265 2266
			isert_conn->pi_support = true;
			return TARGET_PROT_ALL;
		}
	}

2267
	isert_info("conn %p PI offload disabled\n", isert_conn);
2268
	isert_conn->pi_support = false;
2269 2270 2271 2272

	return TARGET_PROT_NORMAL;
}

2273 2274 2275 2276
static int
isert_put_nopin(struct iscsi_cmd *cmd, struct iscsi_conn *conn,
		bool nopout_response)
{
2277
	struct isert_cmd *isert_cmd = iscsit_priv_cmd(cmd);
2278
	struct isert_conn *isert_conn = conn->context;
2279 2280 2281 2282 2283 2284 2285
	struct ib_send_wr *send_wr = &isert_cmd->tx_desc.send_wr;

	isert_create_send_desc(isert_conn, isert_cmd, &isert_cmd->tx_desc);
	iscsit_build_nopin_rsp(cmd, conn, (struct iscsi_nopin *)
			       &isert_cmd->tx_desc.iscsi_header,
			       nopout_response);
	isert_init_tx_hdrs(isert_conn, &isert_cmd->tx_desc);
2286
	isert_init_send_wr(isert_conn, isert_cmd, send_wr);
2287

2288
	isert_dbg("conn %p Posting NOPIN Response\n", isert_conn);
2289 2290 2291 2292 2293 2294 2295

	return isert_post_response(isert_conn, isert_cmd);
}

static int
isert_put_logout_rsp(struct iscsi_cmd *cmd, struct iscsi_conn *conn)
{
2296
	struct isert_cmd *isert_cmd = iscsit_priv_cmd(cmd);
2297
	struct isert_conn *isert_conn = conn->context;
2298 2299 2300 2301 2302 2303
	struct ib_send_wr *send_wr = &isert_cmd->tx_desc.send_wr;

	isert_create_send_desc(isert_conn, isert_cmd, &isert_cmd->tx_desc);
	iscsit_build_logout_rsp(cmd, conn, (struct iscsi_logout_rsp *)
				&isert_cmd->tx_desc.iscsi_header);
	isert_init_tx_hdrs(isert_conn, &isert_cmd->tx_desc);
2304
	isert_init_send_wr(isert_conn, isert_cmd, send_wr);
2305

2306
	isert_dbg("conn %p Posting Logout Response\n", isert_conn);
2307 2308 2309 2310 2311 2312 2313

	return isert_post_response(isert_conn, isert_cmd);
}

static int
isert_put_tm_rsp(struct iscsi_cmd *cmd, struct iscsi_conn *conn)
{
2314
	struct isert_cmd *isert_cmd = iscsit_priv_cmd(cmd);
2315
	struct isert_conn *isert_conn = conn->context;
2316 2317 2318 2319 2320 2321
	struct ib_send_wr *send_wr = &isert_cmd->tx_desc.send_wr;

	isert_create_send_desc(isert_conn, isert_cmd, &isert_cmd->tx_desc);
	iscsit_build_task_mgt_rsp(cmd, conn, (struct iscsi_tm_rsp *)
				  &isert_cmd->tx_desc.iscsi_header);
	isert_init_tx_hdrs(isert_conn, &isert_cmd->tx_desc);
2322
	isert_init_send_wr(isert_conn, isert_cmd, send_wr);
2323

2324
	isert_dbg("conn %p Posting Task Management Response\n", isert_conn);
2325 2326 2327 2328 2329 2330 2331

	return isert_post_response(isert_conn, isert_cmd);
}

static int
isert_put_reject(struct iscsi_cmd *cmd, struct iscsi_conn *conn)
{
2332
	struct isert_cmd *isert_cmd = iscsit_priv_cmd(cmd);
2333
	struct isert_conn *isert_conn = conn->context;
2334
	struct ib_send_wr *send_wr = &isert_cmd->tx_desc.send_wr;
2335
	struct isert_device *device = isert_conn->device;
2336
	struct ib_device *ib_dev = device->ib_device;
2337 2338 2339
	struct ib_sge *tx_dsg = &isert_cmd->tx_desc.tx_sg[1];
	struct iscsi_reject *hdr =
		(struct iscsi_reject *)&isert_cmd->tx_desc.iscsi_header;
2340 2341

	isert_create_send_desc(isert_conn, isert_cmd, &isert_cmd->tx_desc);
2342
	iscsit_build_reject(cmd, conn, hdr);
2343
	isert_init_tx_hdrs(isert_conn, &isert_cmd->tx_desc);
2344 2345

	hton24(hdr->dlength, ISCSI_HDR_LEN);
2346
	isert_cmd->pdu_buf_dma = ib_dma_map_single(ib_dev,
2347 2348
			(void *)cmd->buf_ptr, ISCSI_HDR_LEN,
			DMA_TO_DEVICE);
2349 2350
	isert_cmd->pdu_buf_len = ISCSI_HDR_LEN;
	tx_dsg->addr	= isert_cmd->pdu_buf_dma;
2351
	tx_dsg->length	= ISCSI_HDR_LEN;
2352
	tx_dsg->lkey	= device->pd->local_dma_lkey;
2353 2354
	isert_cmd->tx_desc.num_sge = 2;

2355
	isert_init_send_wr(isert_conn, isert_cmd, send_wr);
2356

2357
	isert_dbg("conn %p Posting Reject\n", isert_conn);
2358 2359 2360 2361

	return isert_post_response(isert_conn, isert_cmd);
}

2362 2363 2364
static int
isert_put_text_rsp(struct iscsi_cmd *cmd, struct iscsi_conn *conn)
{
2365
	struct isert_cmd *isert_cmd = iscsit_priv_cmd(cmd);
2366
	struct isert_conn *isert_conn = conn->context;
2367 2368 2369 2370 2371 2372 2373
	struct ib_send_wr *send_wr = &isert_cmd->tx_desc.send_wr;
	struct iscsi_text_rsp *hdr =
		(struct iscsi_text_rsp *)&isert_cmd->tx_desc.iscsi_header;
	u32 txt_rsp_len;
	int rc;

	isert_create_send_desc(isert_conn, isert_cmd, &isert_cmd->tx_desc);
2374
	rc = iscsit_build_text_rsp(cmd, conn, hdr, ISCSI_INFINIBAND);
2375 2376 2377 2378 2379 2380 2381
	if (rc < 0)
		return rc;

	txt_rsp_len = rc;
	isert_init_tx_hdrs(isert_conn, &isert_cmd->tx_desc);

	if (txt_rsp_len) {
2382
		struct isert_device *device = isert_conn->device;
2383
		struct ib_device *ib_dev = device->ib_device;
2384 2385 2386 2387 2388 2389 2390 2391 2392
		struct ib_sge *tx_dsg = &isert_cmd->tx_desc.tx_sg[1];
		void *txt_rsp_buf = cmd->buf_ptr;

		isert_cmd->pdu_buf_dma = ib_dma_map_single(ib_dev,
				txt_rsp_buf, txt_rsp_len, DMA_TO_DEVICE);

		isert_cmd->pdu_buf_len = txt_rsp_len;
		tx_dsg->addr	= isert_cmd->pdu_buf_dma;
		tx_dsg->length	= txt_rsp_len;
2393
		tx_dsg->lkey	= device->pd->local_dma_lkey;
2394 2395
		isert_cmd->tx_desc.num_sge = 2;
	}
2396
	isert_init_send_wr(isert_conn, isert_cmd, send_wr);
2397

2398
	isert_dbg("conn %p Text Response\n", isert_conn);
2399 2400 2401 2402

	return isert_post_response(isert_conn, isert_cmd);
}

2403 2404
static int
isert_build_rdma_wr(struct isert_conn *isert_conn, struct isert_cmd *isert_cmd,
C
Christoph Hellwig 已提交
2405
		    struct ib_sge *ib_sge, struct ib_rdma_wr *rdma_wr,
2406 2407
		    u32 data_left, u32 offset)
{
2408
	struct iscsi_cmd *cmd = isert_cmd->iscsi_cmd;
2409
	struct scatterlist *sg_start, *tmp_sg;
2410
	struct isert_device *device = isert_conn->device;
2411
	struct ib_device *ib_dev = device->ib_device;
2412 2413 2414 2415 2416 2417 2418 2419
	u32 sg_off, page_off;
	int i = 0, sg_nents;

	sg_off = offset / PAGE_SIZE;
	sg_start = &cmd->se_cmd.t_data_sg[sg_off];
	sg_nents = min(cmd->se_cmd.t_data_nents - sg_off, isert_conn->max_sge);
	page_off = offset % PAGE_SIZE;

C
Christoph Hellwig 已提交
2420 2421
	rdma_wr->wr.sg_list = ib_sge;
	rdma_wr->wr.wr_id = (uintptr_t)&isert_cmd->tx_desc;
2422 2423 2424 2425
	/*
	 * Perform mapping of TCM scatterlist memory ib_sge dma_addr.
	 */
	for_each_sg(sg_start, tmp_sg, sg_nents, i) {
2426 2427 2428 2429
		isert_dbg("RDMA from SGL dma_addr: 0x%llx dma_len: %u, "
			  "page_off: %u\n",
			  (unsigned long long)tmp_sg->dma_address,
			  tmp_sg->length, page_off);
2430 2431 2432 2433

		ib_sge->addr = ib_sg_dma_address(ib_dev, tmp_sg) + page_off;
		ib_sge->length = min_t(u32, data_left,
				ib_sg_dma_len(ib_dev, tmp_sg) - page_off);
2434
		ib_sge->lkey = device->pd->local_dma_lkey;
2435

2436 2437
		isert_dbg("RDMA ib_sge: addr: 0x%llx  length: %u lkey: %x\n",
			  ib_sge->addr, ib_sge->length, ib_sge->lkey);
2438 2439
		page_off = 0;
		data_left -= ib_sge->length;
2440 2441
		if (!data_left)
			break;
2442
		ib_sge++;
2443
		isert_dbg("Incrementing ib_sge pointer to %p\n", ib_sge);
2444 2445
	}

C
Christoph Hellwig 已提交
2446
	rdma_wr->wr.num_sge = ++i;
2447
	isert_dbg("Set outgoing sg_list: %p num_sg: %u from TCM SGLs\n",
C
Christoph Hellwig 已提交
2448
		  rdma_wr->wr.sg_list, rdma_wr->wr.num_sge);
2449

C
Christoph Hellwig 已提交
2450
	return rdma_wr->wr.num_sge;
2451 2452 2453
}

static int
2454 2455
isert_map_rdma(struct iscsi_conn *conn, struct iscsi_cmd *cmd,
	       struct isert_rdma_wr *wr)
2456 2457
{
	struct se_cmd *se_cmd = &cmd->se_cmd;
2458
	struct isert_cmd *isert_cmd = iscsit_priv_cmd(cmd);
2459
	struct isert_conn *isert_conn = conn->context;
2460
	struct isert_data_buf *data = &wr->data;
C
Christoph Hellwig 已提交
2461
	struct ib_rdma_wr *rdma_wr;
2462
	struct ib_sge *ib_sge;
2463 2464
	u32 offset, data_len, data_left, rdma_write_max, va_offset = 0;
	int ret = 0, i, ib_sge_cnt;
2465

2466
	isert_cmd->tx_desc.isert_cmd = isert_cmd;
2467

2468 2469 2470 2471 2472 2473
	offset = wr->iser_ib_op == ISER_IB_RDMA_READ ? cmd->write_data_done : 0;
	ret = isert_map_data_buf(isert_conn, isert_cmd, se_cmd->t_data_sg,
				 se_cmd->t_data_nents, se_cmd->data_length,
				 offset, wr->iser_ib_op, &wr->data);
	if (ret)
		return ret;
2474

2475 2476
	data_left = data->len;
	offset = data->offset;
2477

2478
	ib_sge = kzalloc(sizeof(struct ib_sge) * data->nents, GFP_KERNEL);
2479
	if (!ib_sge) {
2480
		isert_warn("Unable to allocate ib_sge\n");
2481
		ret = -ENOMEM;
2482
		goto unmap_cmd;
2483
	}
2484
	wr->ib_sge = ib_sge;
2485

C
Christoph Hellwig 已提交
2486 2487
	wr->rdma_wr_num = DIV_ROUND_UP(data->nents, isert_conn->max_sge);
	wr->rdma_wr = kzalloc(sizeof(struct ib_rdma_wr) * wr->rdma_wr_num,
2488
				GFP_KERNEL);
C
Christoph Hellwig 已提交
2489 2490
	if (!wr->rdma_wr) {
		isert_dbg("Unable to allocate wr->rdma_wr\n");
2491
		ret = -ENOMEM;
2492
		goto unmap_cmd;
2493 2494 2495 2496 2497
	}

	wr->isert_cmd = isert_cmd;
	rdma_write_max = isert_conn->max_sge * PAGE_SIZE;

C
Christoph Hellwig 已提交
2498 2499
	for (i = 0; i < wr->rdma_wr_num; i++) {
		rdma_wr = &isert_cmd->rdma_wr.rdma_wr[i];
2500 2501
		data_len = min(data_left, rdma_write_max);

C
Christoph Hellwig 已提交
2502
		rdma_wr->wr.send_flags = 0;
2503
		if (wr->iser_ib_op == ISER_IB_RDMA_WRITE) {
C
Christoph Hellwig 已提交
2504 2505 2506 2507 2508
			rdma_wr->wr.opcode = IB_WR_RDMA_WRITE;
			rdma_wr->remote_addr = isert_cmd->read_va + offset;
			rdma_wr->rkey = isert_cmd->read_stag;
			if (i + 1 == wr->rdma_wr_num)
				rdma_wr->wr.next = &isert_cmd->tx_desc.send_wr;
2509
			else
C
Christoph Hellwig 已提交
2510
				rdma_wr->wr.next = &wr->rdma_wr[i + 1].wr;
2511
		} else {
C
Christoph Hellwig 已提交
2512 2513 2514 2515 2516
			rdma_wr->wr.opcode = IB_WR_RDMA_READ;
			rdma_wr->remote_addr = isert_cmd->write_va + va_offset;
			rdma_wr->rkey = isert_cmd->write_stag;
			if (i + 1 == wr->rdma_wr_num)
				rdma_wr->wr.send_flags = IB_SEND_SIGNALED;
2517
			else
C
Christoph Hellwig 已提交
2518
				rdma_wr->wr.next = &wr->rdma_wr[i + 1].wr;
2519
		}
2520 2521

		ib_sge_cnt = isert_build_rdma_wr(isert_conn, isert_cmd, ib_sge,
C
Christoph Hellwig 已提交
2522
					rdma_wr, data_len, offset);
2523 2524 2525
		ib_sge += ib_sge_cnt;

		offset += data_len;
2526
		va_offset += data_len;
2527 2528
		data_left -= data_len;
	}
2529 2530

	return 0;
2531 2532 2533
unmap_cmd:
	isert_unmap_data_buf(isert_conn, data);

2534 2535 2536
	return ret;
}

2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553
static int
isert_map_fr_pagelist(struct ib_device *ib_dev,
		      struct scatterlist *sg_start, int sg_nents, u64 *fr_pl)
{
	u64 start_addr, end_addr, page, chunk_start = 0;
	struct scatterlist *tmp_sg;
	int i = 0, new_chunk, last_ent, n_pages;

	n_pages = 0;
	new_chunk = 1;
	last_ent = sg_nents - 1;
	for_each_sg(sg_start, tmp_sg, sg_nents, i) {
		start_addr = ib_sg_dma_address(ib_dev, tmp_sg);
		if (new_chunk)
			chunk_start = start_addr;
		end_addr = start_addr + ib_sg_dma_len(ib_dev, tmp_sg);

2554 2555 2556
		isert_dbg("SGL[%d] dma_addr: 0x%llx len: %u\n",
			  i, (unsigned long long)tmp_sg->dma_address,
			  tmp_sg->length);
2557 2558 2559 2560 2561 2562 2563 2564 2565 2566

		if ((end_addr & ~PAGE_MASK) && i < last_ent) {
			new_chunk = 0;
			continue;
		}
		new_chunk = 1;

		page = chunk_start & PAGE_MASK;
		do {
			fr_pl[n_pages++] = page;
2567 2568
			isert_dbg("Mapped page_list[%d] page_addr: 0x%llx\n",
				  n_pages - 1, page);
2569 2570 2571 2572 2573 2574 2575
			page += PAGE_SIZE;
		} while (page < end_addr);
	}

	return n_pages;
}

2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590
static inline void
isert_inv_rkey(struct ib_send_wr *inv_wr, struct ib_mr *mr)
{
	u32 rkey;

	memset(inv_wr, 0, sizeof(*inv_wr));
	inv_wr->wr_id = ISER_FASTREG_LI_WRID;
	inv_wr->opcode = IB_WR_LOCAL_INV;
	inv_wr->ex.invalidate_rkey = mr->rkey;

	/* Bump the key */
	rkey = ib_inc_rkey(mr->rkey);
	ib_update_fast_reg_key(mr, rkey);
}

2591
static int
2592 2593 2594
isert_fast_reg_mr(struct isert_conn *isert_conn,
		  struct fast_reg_descriptor *fr_desc,
		  struct isert_data_buf *mem,
2595
		  enum isert_indicator ind,
2596
		  struct ib_sge *sge)
2597
{
2598
	struct isert_device *device = isert_conn->device;
2599
	struct ib_device *ib_dev = device->ib_device;
2600 2601
	struct ib_mr *mr;
	struct ib_fast_reg_page_list *frpl;
C
Christoph Hellwig 已提交
2602 2603
	struct ib_fast_reg_wr fr_wr;
	struct ib_send_wr inv_wr, *bad_wr, *wr = NULL;
2604 2605
	int ret, pagelist_len;
	u32 page_off;
2606

2607
	if (mem->dma_nents == 1) {
2608
		sge->lkey = device->pd->local_dma_lkey;
2609 2610
		sge->addr = ib_sg_dma_address(ib_dev, &mem->sg[0]);
		sge->length = ib_sg_dma_len(ib_dev, &mem->sg[0]);
2611 2612
		isert_dbg("sge: addr: 0x%llx  length: %u lkey: %x\n",
			 sge->addr, sge->length, sge->lkey);
2613 2614 2615
		return 0;
	}

2616 2617 2618 2619 2620 2621 2622 2623 2624 2625
	if (ind == ISERT_DATA_KEY_VALID) {
		/* Registering data buffer */
		mr = fr_desc->data_mr;
		frpl = fr_desc->data_frpl;
	} else {
		/* Registering protection buffer */
		mr = fr_desc->pi_ctx->prot_mr;
		frpl = fr_desc->pi_ctx->prot_frpl;
	}

2626
	page_off = mem->offset % PAGE_SIZE;
2627

2628
	isert_dbg("Use fr_desc %p sg_nents %d offset %u\n",
2629
		  fr_desc, mem->nents, mem->offset);
2630

2631
	pagelist_len = isert_map_fr_pagelist(ib_dev, mem->sg, mem->nents,
2632
					     &frpl->page_list[0]);
2633

2634 2635
	if (!(fr_desc->ind & ind)) {
		isert_inv_rkey(&inv_wr, mr);
2636 2637 2638 2639 2640
		wr = &inv_wr;
	}

	/* Prepare FASTREG WR */
	memset(&fr_wr, 0, sizeof(fr_wr));
C
Christoph Hellwig 已提交
2641 2642 2643 2644 2645 2646 2647 2648 2649
	fr_wr.wr.wr_id = ISER_FASTREG_LI_WRID;
	fr_wr.wr.opcode = IB_WR_FAST_REG_MR;
	fr_wr.iova_start = frpl->page_list[0] + page_off;
	fr_wr.page_list = frpl;
	fr_wr.page_list_len = pagelist_len;
	fr_wr.page_shift = PAGE_SHIFT;
	fr_wr.length = mem->len;
	fr_wr.rkey = mr->rkey;
	fr_wr.access_flags = IB_ACCESS_LOCAL_WRITE;
2650 2651

	if (!wr)
C
Christoph Hellwig 已提交
2652
		wr = &fr_wr.wr;
2653
	else
C
Christoph Hellwig 已提交
2654
		wr->next = &fr_wr.wr;
2655

2656
	ret = ib_post_send(isert_conn->qp, wr, &bad_wr);
2657
	if (ret) {
2658
		isert_err("fast registration failed, ret:%d\n", ret);
2659 2660
		return ret;
	}
2661
	fr_desc->ind &= ~ind;
2662

2663 2664
	sge->lkey = mr->lkey;
	sge->addr = frpl->page_list[0] + page_off;
2665
	sge->length = mem->len;
2666

2667 2668
	isert_dbg("sge: addr: 0x%llx  length: %u lkey: %x\n",
		  sge->addr, sge->length, sge->lkey);
2669 2670 2671 2672

	return ret;
}

2673 2674 2675 2676
static inline void
isert_set_dif_domain(struct se_cmd *se_cmd, struct ib_sig_attrs *sig_attrs,
		     struct ib_sig_domain *domain)
{
2677
	domain->sig_type = IB_SIG_TYPE_T10_DIF;
2678 2679 2680
	domain->sig.dif.bg_type = IB_T10DIF_CRC;
	domain->sig.dif.pi_interval = se_cmd->se_dev->dev_attrib.block_size;
	domain->sig.dif.ref_tag = se_cmd->reftag_seed;
2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691
	/*
	 * At the moment we hard code those, but if in the future
	 * the target core would like to use it, we will take it
	 * from se_cmd.
	 */
	domain->sig.dif.apptag_check_mask = 0xffff;
	domain->sig.dif.app_escape = true;
	domain->sig.dif.ref_escape = true;
	if (se_cmd->prot_type == TARGET_DIF_TYPE1_PROT ||
	    se_cmd->prot_type == TARGET_DIF_TYPE2_PROT)
		domain->sig.dif.ref_remap = true;
2692 2693
};

2694 2695 2696 2697 2698 2699
static int
isert_set_sig_attrs(struct se_cmd *se_cmd, struct ib_sig_attrs *sig_attrs)
{
	switch (se_cmd->prot_op) {
	case TARGET_PROT_DIN_INSERT:
	case TARGET_PROT_DOUT_STRIP:
2700
		sig_attrs->mem.sig_type = IB_SIG_TYPE_NONE;
2701
		isert_set_dif_domain(se_cmd, sig_attrs, &sig_attrs->wire);
2702 2703 2704
		break;
	case TARGET_PROT_DOUT_INSERT:
	case TARGET_PROT_DIN_STRIP:
2705
		sig_attrs->wire.sig_type = IB_SIG_TYPE_NONE;
2706
		isert_set_dif_domain(se_cmd, sig_attrs, &sig_attrs->mem);
2707 2708 2709
		break;
	case TARGET_PROT_DIN_PASS:
	case TARGET_PROT_DOUT_PASS:
2710 2711
		isert_set_dif_domain(se_cmd, sig_attrs, &sig_attrs->wire);
		isert_set_dif_domain(se_cmd, sig_attrs, &sig_attrs->mem);
2712 2713
		break;
	default:
2714
		isert_err("Unsupported PI operation %d\n", se_cmd->prot_op);
2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729
		return -EINVAL;
	}

	return 0;
}

static inline u8
isert_set_prot_checks(u8 prot_checks)
{
	return (prot_checks & TARGET_DIF_CHECK_GUARD  ? 0xc0 : 0) |
	       (prot_checks & TARGET_DIF_CHECK_REFTAG ? 0x30 : 0) |
	       (prot_checks & TARGET_DIF_CHECK_REFTAG ? 0x0f : 0);
}

static int
2730 2731 2732 2733
isert_reg_sig_mr(struct isert_conn *isert_conn,
		 struct se_cmd *se_cmd,
		 struct isert_rdma_wr *rdma_wr,
		 struct fast_reg_descriptor *fr_desc)
2734
{
C
Christoph Hellwig 已提交
2735 2736
	struct ib_sig_handover_wr sig_wr;
	struct ib_send_wr inv_wr, *bad_wr, *wr = NULL;
2737 2738 2739 2740 2741 2742 2743 2744
	struct pi_context *pi_ctx = fr_desc->pi_ctx;
	struct ib_sig_attrs sig_attrs;
	int ret;

	memset(&sig_attrs, 0, sizeof(sig_attrs));
	ret = isert_set_sig_attrs(se_cmd, &sig_attrs);
	if (ret)
		goto err;
2745

2746 2747 2748
	sig_attrs.check_mask = isert_set_prot_checks(se_cmd->prot_checks);

	if (!(fr_desc->ind & ISERT_SIG_KEY_VALID)) {
2749
		isert_inv_rkey(&inv_wr, pi_ctx->sig_mr);
2750 2751 2752 2753
		wr = &inv_wr;
	}

	memset(&sig_wr, 0, sizeof(sig_wr));
C
Christoph Hellwig 已提交
2754 2755 2756 2757 2758 2759 2760
	sig_wr.wr.opcode = IB_WR_REG_SIG_MR;
	sig_wr.wr.wr_id = ISER_FASTREG_LI_WRID;
	sig_wr.wr.sg_list = &rdma_wr->ib_sg[DATA];
	sig_wr.wr.num_sge = 1;
	sig_wr.access_flags = IB_ACCESS_LOCAL_WRITE;
	sig_wr.sig_attrs = &sig_attrs;
	sig_wr.sig_mr = pi_ctx->sig_mr;
2761
	if (se_cmd->t_prot_sg)
C
Christoph Hellwig 已提交
2762
		sig_wr.prot = &rdma_wr->ib_sg[PROT];
2763 2764

	if (!wr)
C
Christoph Hellwig 已提交
2765
		wr = &sig_wr.wr;
2766
	else
C
Christoph Hellwig 已提交
2767
		wr->next = &sig_wr.wr;
2768

2769
	ret = ib_post_send(isert_conn->qp, wr, &bad_wr);
2770
	if (ret) {
2771
		isert_err("fast registration failed, ret:%d\n", ret);
2772 2773 2774 2775
		goto err;
	}
	fr_desc->ind &= ~ISERT_SIG_KEY_VALID;

2776 2777 2778
	rdma_wr->ib_sg[SIG].lkey = pi_ctx->sig_mr->lkey;
	rdma_wr->ib_sg[SIG].addr = 0;
	rdma_wr->ib_sg[SIG].length = se_cmd->data_length;
2779 2780 2781 2782 2783 2784
	if (se_cmd->prot_op != TARGET_PROT_DIN_STRIP &&
	    se_cmd->prot_op != TARGET_PROT_DOUT_INSERT)
		/*
		 * We have protection guards on the wire
		 * so we need to set a larget transfer
		 */
2785
		rdma_wr->ib_sg[SIG].length += se_cmd->prot_length;
2786

2787
	isert_dbg("sig_sge: addr: 0x%llx  length: %u lkey: %x\n",
2788 2789
		  rdma_wr->ib_sg[SIG].addr, rdma_wr->ib_sg[SIG].length,
		  rdma_wr->ib_sg[SIG].lkey);
2790
err:
2791 2792 2793
	return ret;
}

2794 2795 2796 2797 2798
static int
isert_handle_prot_cmd(struct isert_conn *isert_conn,
		      struct isert_cmd *isert_cmd,
		      struct isert_rdma_wr *wr)
{
2799
	struct isert_device *device = isert_conn->device;
2800 2801 2802 2803 2804 2805
	struct se_cmd *se_cmd = &isert_cmd->iscsi_cmd->se_cmd;
	int ret;

	if (!wr->fr_desc->pi_ctx) {
		ret = isert_create_pi_ctx(wr->fr_desc,
					  device->ib_device,
2806
					  device->pd);
2807
		if (ret) {
2808
			isert_err("conn %p failed to allocate pi_ctx\n",
2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820
				  isert_conn);
			return ret;
		}
	}

	if (se_cmd->t_prot_sg) {
		ret = isert_map_data_buf(isert_conn, isert_cmd,
					 se_cmd->t_prot_sg,
					 se_cmd->t_prot_nents,
					 se_cmd->prot_length,
					 0, wr->iser_ib_op, &wr->prot);
		if (ret) {
2821
			isert_err("conn %p failed to map protection buffer\n",
2822 2823 2824 2825 2826 2827 2828 2829
				  isert_conn);
			return ret;
		}

		memset(&wr->ib_sg[PROT], 0, sizeof(wr->ib_sg[PROT]));
		ret = isert_fast_reg_mr(isert_conn, wr->fr_desc, &wr->prot,
					ISERT_PROT_KEY_VALID, &wr->ib_sg[PROT]);
		if (ret) {
2830
			isert_err("conn %p failed to fast reg mr\n",
2831 2832 2833 2834 2835 2836 2837
				  isert_conn);
			goto unmap_prot_cmd;
		}
	}

	ret = isert_reg_sig_mr(isert_conn, se_cmd, wr, wr->fr_desc);
	if (ret) {
2838
		isert_err("conn %p failed to fast reg mr\n",
2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852
			  isert_conn);
		goto unmap_prot_cmd;
	}
	wr->fr_desc->ind |= ISERT_PROTECTED;

	return 0;

unmap_prot_cmd:
	if (se_cmd->t_prot_sg)
		isert_unmap_data_buf(isert_conn, &wr->prot);

	return ret;
}

2853
static int
2854 2855
isert_reg_rdma(struct iscsi_conn *conn, struct iscsi_cmd *cmd,
	       struct isert_rdma_wr *wr)
2856 2857 2858
{
	struct se_cmd *se_cmd = &cmd->se_cmd;
	struct isert_cmd *isert_cmd = iscsit_priv_cmd(cmd);
2859 2860
	struct isert_conn *isert_conn = conn->context;
	struct fast_reg_descriptor *fr_desc = NULL;
C
Christoph Hellwig 已提交
2861
	struct ib_rdma_wr *rdma_wr;
2862
	struct ib_sge *ib_sg;
2863 2864
	u32 offset;
	int ret = 0;
2865 2866
	unsigned long flags;

2867
	isert_cmd->tx_desc.isert_cmd = isert_cmd;
2868

2869 2870 2871 2872 2873 2874
	offset = wr->iser_ib_op == ISER_IB_RDMA_READ ? cmd->write_data_done : 0;
	ret = isert_map_data_buf(isert_conn, isert_cmd, se_cmd->t_data_sg,
				 se_cmd->t_data_nents, se_cmd->data_length,
				 offset, wr->iser_ib_op, &wr->data);
	if (ret)
		return ret;
2875

2876
	if (wr->data.dma_nents != 1 || isert_prot_cmd(isert_conn, se_cmd)) {
2877 2878
		spin_lock_irqsave(&isert_conn->pool_lock, flags);
		fr_desc = list_first_entry(&isert_conn->fr_pool,
2879 2880
					   struct fast_reg_descriptor, list);
		list_del(&fr_desc->list);
2881
		spin_unlock_irqrestore(&isert_conn->pool_lock, flags);
2882
		wr->fr_desc = fr_desc;
2883 2884
	}

2885
	ret = isert_fast_reg_mr(isert_conn, fr_desc, &wr->data,
2886
				ISERT_DATA_KEY_VALID, &wr->ib_sg[DATA]);
2887 2888
	if (ret)
		goto unmap_cmd;
2889

2890
	if (isert_prot_cmd(isert_conn, se_cmd)) {
2891
		ret = isert_handle_prot_cmd(isert_conn, isert_cmd, wr);
2892
		if (ret)
2893
			goto unmap_cmd;
2894

2895 2896 2897 2898
		ib_sg = &wr->ib_sg[SIG];
	} else {
		ib_sg = &wr->ib_sg[DATA];
	}
2899

2900
	memcpy(&wr->s_ib_sge, ib_sg, sizeof(*ib_sg));
2901
	wr->ib_sge = &wr->s_ib_sge;
C
Christoph Hellwig 已提交
2902 2903 2904
	wr->rdma_wr_num = 1;
	memset(&wr->s_rdma_wr, 0, sizeof(wr->s_rdma_wr));
	wr->rdma_wr = &wr->s_rdma_wr;
2905 2906
	wr->isert_cmd = isert_cmd;

C
Christoph Hellwig 已提交
2907 2908 2909 2910
	rdma_wr = &isert_cmd->rdma_wr.s_rdma_wr;
	rdma_wr->wr.sg_list = &wr->s_ib_sge;
	rdma_wr->wr.num_sge = 1;
	rdma_wr->wr.wr_id = (uintptr_t)&isert_cmd->tx_desc;
2911
	if (wr->iser_ib_op == ISER_IB_RDMA_WRITE) {
C
Christoph Hellwig 已提交
2912 2913 2914 2915
		rdma_wr->wr.opcode = IB_WR_RDMA_WRITE;
		rdma_wr->remote_addr = isert_cmd->read_va;
		rdma_wr->rkey = isert_cmd->read_stag;
		rdma_wr->wr.send_flags = !isert_prot_cmd(isert_conn, se_cmd) ?
2916
				      0 : IB_SEND_SIGNALED;
2917
	} else {
C
Christoph Hellwig 已提交
2918 2919 2920 2921
		rdma_wr->wr.opcode = IB_WR_RDMA_READ;
		rdma_wr->remote_addr = isert_cmd->write_va;
		rdma_wr->rkey = isert_cmd->write_stag;
		rdma_wr->wr.send_flags = IB_SEND_SIGNALED;
2922 2923
	}

2924
	return 0;
2925

2926 2927
unmap_cmd:
	if (fr_desc) {
2928 2929 2930
		spin_lock_irqsave(&isert_conn->pool_lock, flags);
		list_add_tail(&fr_desc->list, &isert_conn->fr_pool);
		spin_unlock_irqrestore(&isert_conn->pool_lock, flags);
2931
	}
2932
	isert_unmap_data_buf(isert_conn, &wr->data);
2933 2934 2935 2936

	return ret;
}

2937 2938 2939 2940
static int
isert_put_datain(struct iscsi_conn *conn, struct iscsi_cmd *cmd)
{
	struct se_cmd *se_cmd = &cmd->se_cmd;
2941
	struct isert_cmd *isert_cmd = iscsit_priv_cmd(cmd);
2942
	struct isert_rdma_wr *wr = &isert_cmd->rdma_wr;
2943
	struct isert_conn *isert_conn = conn->context;
2944
	struct isert_device *device = isert_conn->device;
2945 2946 2947
	struct ib_send_wr *wr_failed;
	int rc;

2948
	isert_dbg("Cmd: %p RDMA_WRITE data_length: %u\n",
2949
		 isert_cmd, se_cmd->data_length);
2950

2951
	wr->iser_ib_op = ISER_IB_RDMA_WRITE;
2952
	rc = device->reg_rdma_mem(conn, cmd, wr);
2953
	if (rc) {
2954
		isert_err("Cmd: %p failed to prepare RDMA res\n", isert_cmd);
2955 2956 2957
		return rc;
	}

2958
	if (!isert_prot_cmd(isert_conn, se_cmd)) {
2959 2960 2961 2962 2963 2964 2965 2966 2967
		/*
		 * Build isert_conn->tx_desc for iSCSI response PDU and attach
		 */
		isert_create_send_desc(isert_conn, isert_cmd,
				       &isert_cmd->tx_desc);
		iscsit_build_rsp_pdu(cmd, conn, true, (struct iscsi_scsi_rsp *)
				     &isert_cmd->tx_desc.iscsi_header);
		isert_init_tx_hdrs(isert_conn, &isert_cmd->tx_desc);
		isert_init_send_wr(isert_conn, isert_cmd,
2968
				   &isert_cmd->tx_desc.send_wr);
C
Christoph Hellwig 已提交
2969 2970
		isert_cmd->rdma_wr.s_rdma_wr.wr.next = &isert_cmd->tx_desc.send_wr;
		wr->rdma_wr_num += 1;
2971 2972 2973 2974 2975 2976

		rc = isert_post_recv(isert_conn, isert_cmd->rx_desc);
		if (rc) {
			isert_err("ib_post_recv failed with %d\n", rc);
			return rc;
		}
2977
	}
2978

C
Christoph Hellwig 已提交
2979
	rc = ib_post_send(isert_conn->qp, &wr->rdma_wr->wr, &wr_failed);
2980
	if (rc)
2981
		isert_warn("ib_post_send() failed for IB_WR_RDMA_WRITE\n");
2982

2983
	if (!isert_prot_cmd(isert_conn, se_cmd))
2984
		isert_dbg("Cmd: %p posted RDMA_WRITE + Response for iSER Data "
2985 2986
			 "READ\n", isert_cmd);
	else
2987
		isert_dbg("Cmd: %p posted RDMA_WRITE for iSER Data READ\n",
2988
			 isert_cmd);
2989

2990
	return 1;
2991 2992 2993 2994 2995 2996
}

static int
isert_get_dataout(struct iscsi_conn *conn, struct iscsi_cmd *cmd, bool recovery)
{
	struct se_cmd *se_cmd = &cmd->se_cmd;
2997
	struct isert_cmd *isert_cmd = iscsit_priv_cmd(cmd);
2998
	struct isert_rdma_wr *wr = &isert_cmd->rdma_wr;
2999
	struct isert_conn *isert_conn = conn->context;
3000
	struct isert_device *device = isert_conn->device;
3001 3002
	struct ib_send_wr *wr_failed;
	int rc;
3003

3004
	isert_dbg("Cmd: %p RDMA_READ data_length: %u write_data_done: %u\n",
3005
		 isert_cmd, se_cmd->data_length, cmd->write_data_done);
3006
	wr->iser_ib_op = ISER_IB_RDMA_READ;
3007
	rc = device->reg_rdma_mem(conn, cmd, wr);
3008
	if (rc) {
3009
		isert_err("Cmd: %p failed to prepare RDMA res\n", isert_cmd);
3010
		return rc;
3011 3012
	}

C
Christoph Hellwig 已提交
3013
	rc = ib_post_send(isert_conn->qp, &wr->rdma_wr->wr, &wr_failed);
3014
	if (rc)
3015
		isert_warn("ib_post_send() failed for IB_WR_RDMA_READ\n");
3016

3017
	isert_dbg("Cmd: %p posted RDMA_READ memory for ISER Data WRITE\n",
3018
		 isert_cmd);
3019

3020
	return 0;
3021 3022 3023 3024 3025
}

static int
isert_immediate_queue(struct iscsi_conn *conn, struct iscsi_cmd *cmd, int state)
{
3026 3027
	struct isert_cmd *isert_cmd = iscsit_priv_cmd(cmd);
	int ret = 0;
3028 3029

	switch (state) {
3030 3031 3032 3033 3034 3035
	case ISTATE_REMOVE:
		spin_lock_bh(&conn->cmd_lock);
		list_del_init(&cmd->i_conn_node);
		spin_unlock_bh(&conn->cmd_lock);
		isert_put_cmd(isert_cmd, true);
		break;
3036 3037 3038 3039
	case ISTATE_SEND_NOPIN_WANT_RESPONSE:
		ret = isert_put_nopin(cmd, conn, false);
		break;
	default:
3040
		isert_err("Unknown immediate state: 0x%02x\n", state);
3041 3042 3043 3044 3045 3046 3047 3048 3049 3050
		ret = -EINVAL;
		break;
	}

	return ret;
}

static int
isert_response_queue(struct iscsi_conn *conn, struct iscsi_cmd *cmd, int state)
{
S
Sagi Grimberg 已提交
3051
	struct isert_conn *isert_conn = conn->context;
3052 3053 3054 3055 3056
	int ret;

	switch (state) {
	case ISTATE_SEND_LOGOUTRSP:
		ret = isert_put_logout_rsp(cmd, conn);
S
Sagi Grimberg 已提交
3057 3058
		if (!ret)
			isert_conn->logout_posted = true;
3059 3060 3061 3062 3063 3064 3065 3066 3067 3068
		break;
	case ISTATE_SEND_NOPIN:
		ret = isert_put_nopin(cmd, conn, true);
		break;
	case ISTATE_SEND_TASKMGTRSP:
		ret = isert_put_tm_rsp(cmd, conn);
		break;
	case ISTATE_SEND_REJECT:
		ret = isert_put_reject(cmd, conn);
		break;
3069 3070 3071
	case ISTATE_SEND_TEXTRSP:
		ret = isert_put_text_rsp(cmd, conn);
		break;
3072 3073 3074 3075 3076 3077 3078 3079
	case ISTATE_SEND_STATUS:
		/*
		 * Special case for sending non GOOD SCSI status from TX thread
		 * context during pre se_cmd excecution failure.
		 */
		ret = isert_put_response(conn, cmd);
		break;
	default:
3080
		isert_err("Unknown response state: 0x%02x\n", state);
3081 3082 3083 3084 3085 3086 3087
		ret = -EINVAL;
		break;
	}

	return ret;
}

3088 3089 3090 3091 3092 3093 3094 3095 3096
struct rdma_cm_id *
isert_setup_id(struct isert_np *isert_np)
{
	struct iscsi_np *np = isert_np->np;
	struct rdma_cm_id *id;
	struct sockaddr *sa;
	int ret;

	sa = (struct sockaddr *)&np->np_sockaddr;
3097
	isert_dbg("ksockaddr: %p, sa: %p\n", &np->np_sockaddr, sa);
3098

3099
	id = rdma_create_id(&init_net, isert_cma_handler, isert_np,
3100 3101
			    RDMA_PS_TCP, IB_QPT_RC);
	if (IS_ERR(id)) {
3102
		isert_err("rdma_create_id() failed: %ld\n", PTR_ERR(id));
3103 3104 3105
		ret = PTR_ERR(id);
		goto out;
	}
3106
	isert_dbg("id %p context %p\n", id, id->context);
3107 3108 3109

	ret = rdma_bind_addr(id, sa);
	if (ret) {
3110
		isert_err("rdma_bind_addr() failed: %d\n", ret);
3111 3112 3113
		goto out_id;
	}

3114
	ret = rdma_listen(id, 0);
3115
	if (ret) {
3116
		isert_err("rdma_listen() failed: %d\n", ret);
3117 3118 3119 3120 3121 3122 3123 3124 3125 3126
		goto out_id;
	}

	return id;
out_id:
	rdma_destroy_id(id);
out:
	return ERR_PTR(ret);
}

3127 3128
static int
isert_setup_np(struct iscsi_np *np,
3129
	       struct sockaddr_storage *ksockaddr)
3130 3131 3132 3133 3134 3135 3136
{
	struct isert_np *isert_np;
	struct rdma_cm_id *isert_lid;
	int ret;

	isert_np = kzalloc(sizeof(struct isert_np), GFP_KERNEL);
	if (!isert_np) {
3137
		isert_err("Unable to allocate struct isert_np\n");
3138 3139
		return -ENOMEM;
	}
3140 3141
	sema_init(&isert_np->sem, 0);
	mutex_init(&isert_np->mutex);
3142 3143
	INIT_LIST_HEAD(&isert_np->accepted);
	INIT_LIST_HEAD(&isert_np->pending);
3144
	isert_np->np = np;
3145 3146 3147 3148 3149 3150

	/*
	 * Setup the np->np_sockaddr from the passed sockaddr setup
	 * in iscsi_target_configfs.c code..
	 */
	memcpy(&np->np_sockaddr, ksockaddr,
3151
	       sizeof(struct sockaddr_storage));
3152

3153
	isert_lid = isert_setup_id(isert_np);
3154 3155 3156 3157 3158
	if (IS_ERR(isert_lid)) {
		ret = PTR_ERR(isert_lid);
		goto out;
	}

3159
	isert_np->cm_id = isert_lid;
3160 3161 3162 3163 3164 3165
	np->np_context = isert_np;

	return 0;

out:
	kfree(isert_np);
3166

3167 3168 3169 3170 3171 3172
	return ret;
}

static int
isert_rdma_accept(struct isert_conn *isert_conn)
{
3173
	struct rdma_cm_id *cm_id = isert_conn->cm_id;
3174 3175 3176 3177 3178 3179 3180 3181 3182 3183
	struct rdma_conn_param cp;
	int ret;

	memset(&cp, 0, sizeof(struct rdma_conn_param));
	cp.initiator_depth = isert_conn->initiator_depth;
	cp.retry_count = 7;
	cp.rnr_retry_count = 7;

	ret = rdma_accept(cm_id, &cp);
	if (ret) {
3184
		isert_err("rdma_accept() failed with: %d\n", ret);
3185 3186 3187 3188 3189 3190 3191 3192 3193
		return ret;
	}

	return 0;
}

static int
isert_get_login_rx(struct iscsi_conn *conn, struct iscsi_login *login)
{
3194
	struct isert_conn *isert_conn = conn->context;
3195 3196
	int ret;

3197
	isert_info("before login_req comp conn: %p\n", isert_conn);
3198 3199
	ret = wait_for_completion_interruptible(&isert_conn->login_req_comp);
	if (ret) {
3200
		isert_err("isert_conn %p interrupted before got login req\n",
3201 3202 3203 3204 3205
			  isert_conn);
		return ret;
	}
	reinit_completion(&isert_conn->login_req_comp);

3206 3207 3208 3209 3210 3211 3212 3213
	/*
	 * For login requests after the first PDU, isert_rx_login_req() will
	 * kick schedule_delayed_work(&conn->login_work) as the packet is
	 * received, which turns this callback from iscsi_target_do_login_rx()
	 * into a NOP.
	 */
	if (!login->first_request)
		return 0;
3214

3215 3216
	isert_rx_login_req(isert_conn);

3217 3218
	isert_info("before login_comp conn: %p\n", conn);
	ret = wait_for_completion_interruptible(&isert_conn->login_comp);
3219 3220 3221
	if (ret)
		return ret;

3222
	isert_info("processing login->req: %p\n", login->req);
3223

3224 3225 3226 3227 3228 3229 3230
	return 0;
}

static void
isert_set_conn_info(struct iscsi_np *np, struct iscsi_conn *conn,
		    struct isert_conn *isert_conn)
{
3231
	struct rdma_cm_id *cm_id = isert_conn->cm_id;
3232 3233 3234 3235
	struct rdma_route *cm_route = &cm_id->route;

	conn->login_family = np->np_sockaddr.ss_family;

3236 3237
	conn->login_sockaddr = cm_route->addr.dst_addr;
	conn->local_sockaddr = cm_route->addr.src_addr;
3238 3239 3240 3241 3242
}

static int
isert_accept_np(struct iscsi_np *np, struct iscsi_conn *conn)
{
3243
	struct isert_np *isert_np = np->np_context;
3244
	struct isert_conn *isert_conn;
3245
	int ret;
3246 3247

accept_wait:
3248
	ret = down_interruptible(&isert_np->sem);
3249
	if (ret)
3250 3251 3252
		return -ENODEV;

	spin_lock_bh(&np->np_thread_lock);
3253
	if (np->np_thread_state >= ISCSI_NP_THREAD_RESET) {
3254
		spin_unlock_bh(&np->np_thread_lock);
3255
		isert_dbg("np_thread_state %d\n",
3256 3257 3258 3259 3260
			 np->np_thread_state);
		/**
		 * No point in stalling here when np_thread
		 * is in state RESET/SHUTDOWN/EXIT - bail
		 **/
3261 3262 3263 3264
		return -ENODEV;
	}
	spin_unlock_bh(&np->np_thread_lock);

3265
	mutex_lock(&isert_np->mutex);
3266
	if (list_empty(&isert_np->pending)) {
3267
		mutex_unlock(&isert_np->mutex);
3268 3269
		goto accept_wait;
	}
3270 3271 3272
	isert_conn = list_first_entry(&isert_np->pending,
			struct isert_conn, node);
	list_del_init(&isert_conn->node);
3273
	mutex_unlock(&isert_np->mutex);
3274 3275 3276 3277 3278 3279

	conn->context = isert_conn;
	isert_conn->conn = conn;

	isert_set_conn_info(np, conn, isert_conn);

3280
	isert_dbg("Processing isert_conn: %p\n", isert_conn);
3281

3282 3283 3284 3285 3286 3287
	return 0;
}

static void
isert_free_np(struct iscsi_np *np)
{
3288
	struct isert_np *isert_np = np->np_context;
3289
	struct isert_conn *isert_conn, *n;
3290

3291 3292
	if (isert_np->cm_id)
		rdma_destroy_id(isert_np->cm_id);
3293

3294 3295 3296 3297 3298
	/*
	 * FIXME: At this point we don't have a good way to insure
	 * that at this point we don't have hanging connections that
	 * completed RDMA establishment but didn't start iscsi login
	 * process. So work-around this by cleaning up what ever piled
3299
	 * up in accepted and pending lists.
3300
	 */
3301
	mutex_lock(&isert_np->mutex);
3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314
	if (!list_empty(&isert_np->pending)) {
		isert_info("Still have isert pending connections\n");
		list_for_each_entry_safe(isert_conn, n,
					 &isert_np->pending,
					 node) {
			isert_info("cleaning isert_conn %p state (%d)\n",
				   isert_conn, isert_conn->state);
			isert_connect_release(isert_conn);
		}
	}

	if (!list_empty(&isert_np->accepted)) {
		isert_info("Still have isert accepted connections\n");
3315
		list_for_each_entry_safe(isert_conn, n,
3316 3317
					 &isert_np->accepted,
					 node) {
3318
			isert_info("cleaning isert_conn %p state (%d)\n",
3319 3320 3321 3322
				   isert_conn, isert_conn->state);
			isert_connect_release(isert_conn);
		}
	}
3323
	mutex_unlock(&isert_np->mutex);
3324

3325 3326 3327 3328
	np->np_context = NULL;
	kfree(isert_np);
}

3329 3330 3331 3332 3333 3334
static void isert_release_work(struct work_struct *work)
{
	struct isert_conn *isert_conn = container_of(work,
						     struct isert_conn,
						     release_work);

3335
	isert_info("Starting release conn %p\n", isert_conn);
3336

3337
	wait_for_completion(&isert_conn->wait);
3338

3339
	mutex_lock(&isert_conn->mutex);
3340
	isert_conn->state = ISER_CONN_DOWN;
3341
	mutex_unlock(&isert_conn->mutex);
3342

3343
	isert_info("Destroying conn %p\n", isert_conn);
3344 3345 3346
	isert_put_conn(isert_conn);
}

S
Sagi Grimberg 已提交
3347 3348 3349 3350 3351
static void
isert_wait4logout(struct isert_conn *isert_conn)
{
	struct iscsi_conn *conn = isert_conn->conn;

3352 3353
	isert_info("conn %p\n", isert_conn);

S
Sagi Grimberg 已提交
3354
	if (isert_conn->logout_posted) {
3355
		isert_info("conn %p wait for conn_logout_comp\n", isert_conn);
S
Sagi Grimberg 已提交
3356 3357 3358 3359 3360
		wait_for_completion_timeout(&conn->conn_logout_comp,
					    SECONDS_FOR_LOGOUT_COMP * HZ);
	}
}

3361 3362 3363
static void
isert_wait4cmds(struct iscsi_conn *conn)
{
3364 3365
	isert_info("iscsi_conn %p\n", conn);

3366 3367 3368 3369 3370 3371
	if (conn->sess) {
		target_sess_cmd_list_set_waiting(conn->sess->se_sess);
		target_wait_for_sess_cmds(conn->sess->se_sess);
	}
}

3372 3373 3374 3375 3376
static void
isert_wait4flush(struct isert_conn *isert_conn)
{
	struct ib_recv_wr *bad_wr;

3377 3378
	isert_info("conn %p\n", isert_conn);

3379
	init_completion(&isert_conn->wait_comp_err);
3380 3381
	isert_conn->beacon.wr_id = ISER_BEACON_WRID;
	/* post an indication that all flush errors were consumed */
3382
	if (ib_post_recv(isert_conn->qp, &isert_conn->beacon, &bad_wr)) {
3383
		isert_err("conn %p failed to post beacon", isert_conn);
3384 3385 3386
		return;
	}

3387
	wait_for_completion(&isert_conn->wait_comp_err);
3388 3389
}

3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424
/**
 * isert_put_unsol_pending_cmds() - Drop commands waiting for
 *     unsolicitate dataout
 * @conn:    iscsi connection
 *
 * We might still have commands that are waiting for unsolicited
 * dataouts messages. We must put the extra reference on those
 * before blocking on the target_wait_for_session_cmds
 */
static void
isert_put_unsol_pending_cmds(struct iscsi_conn *conn)
{
	struct iscsi_cmd *cmd, *tmp;
	static LIST_HEAD(drop_cmd_list);

	spin_lock_bh(&conn->cmd_lock);
	list_for_each_entry_safe(cmd, tmp, &conn->conn_cmd_list, i_conn_node) {
		if ((cmd->cmd_flags & ICF_NON_IMMEDIATE_UNSOLICITED_DATA) &&
		    (cmd->write_data_done < conn->sess->sess_ops->FirstBurstLength) &&
		    (cmd->write_data_done < cmd->se_cmd.data_length))
			list_move_tail(&cmd->i_conn_node, &drop_cmd_list);
	}
	spin_unlock_bh(&conn->cmd_lock);

	list_for_each_entry_safe(cmd, tmp, &drop_cmd_list, i_conn_node) {
		list_del_init(&cmd->i_conn_node);
		if (cmd->i_state != ISTATE_REMOVE) {
			struct isert_cmd *isert_cmd = iscsit_priv_cmd(cmd);

			isert_info("conn %p dropping cmd %p\n", conn, cmd);
			isert_put_cmd(isert_cmd, true);
		}
	}
}

3425
static void isert_wait_conn(struct iscsi_conn *conn)
3426 3427 3428
{
	struct isert_conn *isert_conn = conn->context;

3429
	isert_info("Starting conn %p\n", isert_conn);
3430

3431
	mutex_lock(&isert_conn->mutex);
3432
	/*
3433
	 * Only wait for wait_comp_err if the isert_conn made it
3434 3435
	 * into full feature phase..
	 */
3436
	if (isert_conn->state == ISER_CONN_INIT) {
3437
		mutex_unlock(&isert_conn->mutex);
3438
		return;
3439
	}
3440
	isert_conn_terminate(isert_conn);
3441
	mutex_unlock(&isert_conn->mutex);
3442

3443
	isert_wait4flush(isert_conn);
3444 3445
	isert_put_unsol_pending_cmds(conn);
	isert_wait4cmds(conn);
S
Sagi Grimberg 已提交
3446
	isert_wait4logout(isert_conn);
3447

3448
	queue_work(isert_release_wq, &isert_conn->release_work);
3449 3450 3451 3452 3453
}

static void isert_free_conn(struct iscsi_conn *conn)
{
	struct isert_conn *isert_conn = conn->context;
3454

3455
	isert_wait4flush(isert_conn);
3456 3457 3458 3459 3460 3461
	isert_put_conn(isert_conn);
}

static struct iscsit_transport iser_target_transport = {
	.name			= "IB/iSER",
	.transport_type		= ISCSI_INFINIBAND,
3462
	.priv_size		= sizeof(struct isert_cmd),
3463 3464 3465 3466
	.owner			= THIS_MODULE,
	.iscsit_setup_np	= isert_setup_np,
	.iscsit_accept_np	= isert_accept_np,
	.iscsit_free_np		= isert_free_np,
3467
	.iscsit_wait_conn	= isert_wait_conn,
3468 3469 3470 3471 3472 3473 3474 3475
	.iscsit_free_conn	= isert_free_conn,
	.iscsit_get_login_rx	= isert_get_login_rx,
	.iscsit_put_login_tx	= isert_put_login_tx,
	.iscsit_immediate_queue	= isert_immediate_queue,
	.iscsit_response_queue	= isert_response_queue,
	.iscsit_get_dataout	= isert_get_dataout,
	.iscsit_queue_data_in	= isert_put_datain,
	.iscsit_queue_status	= isert_put_response,
3476
	.iscsit_aborted_task	= isert_aborted_task,
3477
	.iscsit_get_sup_prot_ops = isert_get_sup_prot_ops,
3478 3479 3480 3481 3482 3483
};

static int __init isert_init(void)
{
	int ret;

3484 3485
	isert_comp_wq = alloc_workqueue("isert_comp_wq",
					WQ_UNBOUND | WQ_HIGHPRI, 0);
3486
	if (!isert_comp_wq) {
3487
		isert_err("Unable to allocate isert_comp_wq\n");
3488
		ret = -ENOMEM;
3489
		return -ENOMEM;
3490 3491
	}

3492 3493 3494
	isert_release_wq = alloc_workqueue("isert_release_wq", WQ_UNBOUND,
					WQ_UNBOUND_MAX_ACTIVE);
	if (!isert_release_wq) {
3495
		isert_err("Unable to allocate isert_release_wq\n");
3496 3497 3498 3499
		ret = -ENOMEM;
		goto destroy_comp_wq;
	}

3500
	iscsit_register_transport(&iser_target_transport);
3501
	isert_info("iSER_TARGET[0] - Loaded iser_target_transport\n");
3502

3503 3504
	return 0;

3505 3506
destroy_comp_wq:
	destroy_workqueue(isert_comp_wq);
3507

3508 3509 3510 3511 3512
	return ret;
}

static void __exit isert_exit(void)
{
3513
	flush_scheduled_work();
3514
	destroy_workqueue(isert_release_wq);
3515 3516
	destroy_workqueue(isert_comp_wq);
	iscsit_unregister_transport(&iser_target_transport);
3517
	isert_info("iSER_TARGET[0] - Released iser_target_transport\n");
3518 3519 3520
}

MODULE_DESCRIPTION("iSER-Target for mainline target infrastructure");
S
Sagi Grimberg 已提交
3521
MODULE_VERSION("1.0");
3522 3523 3524 3525 3526
MODULE_AUTHOR("nab@Linux-iSCSI.org");
MODULE_LICENSE("GPL");

module_init(isert_init);
module_exit(isert_exit);