rdma.c 51.9 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
/*
 * NVMe over Fabrics RDMA host code.
 * Copyright (c) 2015-2016 HGST, a Western Digital Company.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
 * version 2, as published by the Free Software Foundation.
 *
 * This program is distributed in the hope 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.
 */
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
#include <linux/module.h>
#include <linux/init.h>
#include <linux/slab.h>
I
Israel Rukshin 已提交
18
#include <rdma/mr_pool.h>
19 20 21 22
#include <linux/err.h>
#include <linux/string.h>
#include <linux/atomic.h>
#include <linux/blk-mq.h>
23
#include <linux/blk-mq-rdma.h>
24 25 26 27 28 29 30 31 32 33 34 35 36 37 38
#include <linux/types.h>
#include <linux/list.h>
#include <linux/mutex.h>
#include <linux/scatterlist.h>
#include <linux/nvme.h>
#include <asm/unaligned.h>

#include <rdma/ib_verbs.h>
#include <rdma/rdma_cm.h>
#include <linux/nvme-rdma.h>

#include "nvme.h"
#include "fabrics.h"


39
#define NVME_RDMA_CONNECT_TIMEOUT_MS	3000		/* 3 second */
40 41 42 43 44 45

#define NVME_RDMA_MAX_SEGMENTS		256

#define NVME_RDMA_MAX_INLINE_SEGMENTS	1

struct nvme_rdma_device {
46 47
	struct ib_device	*dev;
	struct ib_pd		*pd;
48 49 50 51 52 53 54 55 56 57 58 59
	struct kref		ref;
	struct list_head	entry;
};

struct nvme_rdma_qe {
	struct ib_cqe		cqe;
	void			*data;
	u64			dma;
};

struct nvme_rdma_queue;
struct nvme_rdma_request {
60
	struct nvme_request	req;
61 62
	struct ib_mr		*mr;
	struct nvme_rdma_qe	sqe;
63 64 65
	union nvme_result	result;
	__le16			status;
	refcount_t		ref;
66 67 68 69 70 71 72 73 74 75 76 77
	struct ib_sge		sge[1 + NVME_RDMA_MAX_INLINE_SEGMENTS];
	u32			num_sge;
	int			nents;
	bool			inline_data;
	struct ib_reg_wr	reg_wr;
	struct ib_cqe		reg_cqe;
	struct nvme_rdma_queue  *queue;
	struct sg_table		sg_table;
	struct scatterlist	first_sgl[];
};

enum nvme_rdma_queue_flags {
78 79
	NVME_RDMA_Q_ALLOCATED		= 0,
	NVME_RDMA_Q_LIVE		= 1,
80
	NVME_RDMA_Q_TR_READY		= 2,
81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116
};

struct nvme_rdma_queue {
	struct nvme_rdma_qe	*rsp_ring;
	int			queue_size;
	size_t			cmnd_capsule_len;
	struct nvme_rdma_ctrl	*ctrl;
	struct nvme_rdma_device	*device;
	struct ib_cq		*ib_cq;
	struct ib_qp		*qp;

	unsigned long		flags;
	struct rdma_cm_id	*cm_id;
	int			cm_error;
	struct completion	cm_done;
};

struct nvme_rdma_ctrl {
	/* read only in the hot path */
	struct nvme_rdma_queue	*queues;

	/* other member variables */
	struct blk_mq_tag_set	tag_set;
	struct work_struct	err_work;

	struct nvme_rdma_qe	async_event_sqe;

	struct delayed_work	reconnect_work;

	struct list_head	list;

	struct blk_mq_tag_set	admin_tag_set;
	struct nvme_rdma_device	*device;

	u32			max_fr_pages;

117 118
	struct sockaddr_storage addr;
	struct sockaddr_storage src_addr;
119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147

	struct nvme_ctrl	ctrl;
};

static inline struct nvme_rdma_ctrl *to_rdma_ctrl(struct nvme_ctrl *ctrl)
{
	return container_of(ctrl, struct nvme_rdma_ctrl, ctrl);
}

static LIST_HEAD(device_list);
static DEFINE_MUTEX(device_list_mutex);

static LIST_HEAD(nvme_rdma_ctrl_list);
static DEFINE_MUTEX(nvme_rdma_ctrl_mutex);

/*
 * Disabling this option makes small I/O goes faster, but is fundamentally
 * unsafe.  With it turned off we will have to register a global rkey that
 * allows read and write access to all physical memory.
 */
static bool register_always = true;
module_param(register_always, bool, 0444);
MODULE_PARM_DESC(register_always,
	 "Use memory registration even for contiguous memory regions");

static int nvme_rdma_cm_handler(struct rdma_cm_id *cm_id,
		struct rdma_cm_event *event);
static void nvme_rdma_recv_done(struct ib_cq *cq, struct ib_wc *wc);

148 149 150
static const struct blk_mq_ops nvme_rdma_mq_ops;
static const struct blk_mq_ops nvme_rdma_admin_mq_ops;

151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227
/* XXX: really should move to a generic header sooner or later.. */
static inline void put_unaligned_le24(u32 val, u8 *p)
{
	*p++ = val;
	*p++ = val >> 8;
	*p++ = val >> 16;
}

static inline int nvme_rdma_queue_idx(struct nvme_rdma_queue *queue)
{
	return queue - queue->ctrl->queues;
}

static inline size_t nvme_rdma_inline_data_size(struct nvme_rdma_queue *queue)
{
	return queue->cmnd_capsule_len - sizeof(struct nvme_command);
}

static void nvme_rdma_free_qe(struct ib_device *ibdev, struct nvme_rdma_qe *qe,
		size_t capsule_size, enum dma_data_direction dir)
{
	ib_dma_unmap_single(ibdev, qe->dma, capsule_size, dir);
	kfree(qe->data);
}

static int nvme_rdma_alloc_qe(struct ib_device *ibdev, struct nvme_rdma_qe *qe,
		size_t capsule_size, enum dma_data_direction dir)
{
	qe->data = kzalloc(capsule_size, GFP_KERNEL);
	if (!qe->data)
		return -ENOMEM;

	qe->dma = ib_dma_map_single(ibdev, qe->data, capsule_size, dir);
	if (ib_dma_mapping_error(ibdev, qe->dma)) {
		kfree(qe->data);
		return -ENOMEM;
	}

	return 0;
}

static void nvme_rdma_free_ring(struct ib_device *ibdev,
		struct nvme_rdma_qe *ring, size_t ib_queue_size,
		size_t capsule_size, enum dma_data_direction dir)
{
	int i;

	for (i = 0; i < ib_queue_size; i++)
		nvme_rdma_free_qe(ibdev, &ring[i], capsule_size, dir);
	kfree(ring);
}

static struct nvme_rdma_qe *nvme_rdma_alloc_ring(struct ib_device *ibdev,
		size_t ib_queue_size, size_t capsule_size,
		enum dma_data_direction dir)
{
	struct nvme_rdma_qe *ring;
	int i;

	ring = kcalloc(ib_queue_size, sizeof(struct nvme_rdma_qe), GFP_KERNEL);
	if (!ring)
		return NULL;

	for (i = 0; i < ib_queue_size; i++) {
		if (nvme_rdma_alloc_qe(ibdev, &ring[i], capsule_size, dir))
			goto out_free_ring;
	}

	return ring;

out_free_ring:
	nvme_rdma_free_ring(ibdev, ring, i, capsule_size, dir);
	return NULL;
}

static void nvme_rdma_qp_event(struct ib_event *event, void *context)
{
228 229 230
	pr_debug("QP event %s (%d)\n",
		 ib_event_msg(event->event), event->event);

231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264
}

static int nvme_rdma_wait_for_cm(struct nvme_rdma_queue *queue)
{
	wait_for_completion_interruptible_timeout(&queue->cm_done,
			msecs_to_jiffies(NVME_RDMA_CONNECT_TIMEOUT_MS) + 1);
	return queue->cm_error;
}

static int nvme_rdma_create_qp(struct nvme_rdma_queue *queue, const int factor)
{
	struct nvme_rdma_device *dev = queue->device;
	struct ib_qp_init_attr init_attr;
	int ret;

	memset(&init_attr, 0, sizeof(init_attr));
	init_attr.event_handler = nvme_rdma_qp_event;
	/* +1 for drain */
	init_attr.cap.max_send_wr = factor * queue->queue_size + 1;
	/* +1 for drain */
	init_attr.cap.max_recv_wr = queue->queue_size + 1;
	init_attr.cap.max_recv_sge = 1;
	init_attr.cap.max_send_sge = 1 + NVME_RDMA_MAX_INLINE_SEGMENTS;
	init_attr.sq_sig_type = IB_SIGNAL_REQ_WR;
	init_attr.qp_type = IB_QPT_RC;
	init_attr.send_cq = queue->ib_cq;
	init_attr.recv_cq = queue->ib_cq;

	ret = rdma_create_qp(queue->cm_id, dev->pd, &init_attr);

	queue->qp = queue->cm_id->qp;
	return ret;
}

265 266
static void nvme_rdma_exit_request(struct blk_mq_tag_set *set,
		struct request *rq, unsigned int hctx_idx)
267
{
268
	struct nvme_rdma_ctrl *ctrl = set->driver_data;
269
	struct nvme_rdma_request *req = blk_mq_rq_to_pdu(rq);
270
	int queue_idx = (set == &ctrl->tag_set) ? hctx_idx + 1 : 0;
271 272 273 274 275 276 277
	struct nvme_rdma_queue *queue = &ctrl->queues[queue_idx];
	struct nvme_rdma_device *dev = queue->device;

	nvme_rdma_free_qe(dev->dev, &req->sqe, sizeof(struct nvme_command),
			DMA_TO_DEVICE);
}

278 279 280
static int nvme_rdma_init_request(struct blk_mq_tag_set *set,
		struct request *rq, unsigned int hctx_idx,
		unsigned int numa_node)
281
{
282
	struct nvme_rdma_ctrl *ctrl = set->driver_data;
283
	struct nvme_rdma_request *req = blk_mq_rq_to_pdu(rq);
284
	int queue_idx = (set == &ctrl->tag_set) ? hctx_idx + 1 : 0;
285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305
	struct nvme_rdma_queue *queue = &ctrl->queues[queue_idx];
	struct nvme_rdma_device *dev = queue->device;
	struct ib_device *ibdev = dev->dev;
	int ret;

	ret = nvme_rdma_alloc_qe(ibdev, &req->sqe, sizeof(struct nvme_command),
			DMA_TO_DEVICE);
	if (ret)
		return ret;

	req->queue = queue;

	return 0;
}

static int nvme_rdma_init_hctx(struct blk_mq_hw_ctx *hctx, void *data,
		unsigned int hctx_idx)
{
	struct nvme_rdma_ctrl *ctrl = data;
	struct nvme_rdma_queue *queue = &ctrl->queues[hctx_idx + 1];

306
	BUG_ON(hctx_idx >= ctrl->ctrl.queue_count);
307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365

	hctx->driver_data = queue;
	return 0;
}

static int nvme_rdma_init_admin_hctx(struct blk_mq_hw_ctx *hctx, void *data,
		unsigned int hctx_idx)
{
	struct nvme_rdma_ctrl *ctrl = data;
	struct nvme_rdma_queue *queue = &ctrl->queues[0];

	BUG_ON(hctx_idx != 0);

	hctx->driver_data = queue;
	return 0;
}

static void nvme_rdma_free_dev(struct kref *ref)
{
	struct nvme_rdma_device *ndev =
		container_of(ref, struct nvme_rdma_device, ref);

	mutex_lock(&device_list_mutex);
	list_del(&ndev->entry);
	mutex_unlock(&device_list_mutex);

	ib_dealloc_pd(ndev->pd);
	kfree(ndev);
}

static void nvme_rdma_dev_put(struct nvme_rdma_device *dev)
{
	kref_put(&dev->ref, nvme_rdma_free_dev);
}

static int nvme_rdma_dev_get(struct nvme_rdma_device *dev)
{
	return kref_get_unless_zero(&dev->ref);
}

static struct nvme_rdma_device *
nvme_rdma_find_get_device(struct rdma_cm_id *cm_id)
{
	struct nvme_rdma_device *ndev;

	mutex_lock(&device_list_mutex);
	list_for_each_entry(ndev, &device_list, entry) {
		if (ndev->dev->node_guid == cm_id->device->node_guid &&
		    nvme_rdma_dev_get(ndev))
			goto out_unlock;
	}

	ndev = kzalloc(sizeof(*ndev), GFP_KERNEL);
	if (!ndev)
		goto out_err;

	ndev->dev = cm_id->device;
	kref_init(&ndev->ref);

366 367
	ndev->pd = ib_alloc_pd(ndev->dev,
		register_always ? 0 : IB_PD_UNSAFE_GLOBAL_RKEY);
368 369 370 371 372 373 374
	if (IS_ERR(ndev->pd))
		goto out_free_dev;

	if (!(ndev->dev->attrs.device_cap_flags &
	      IB_DEVICE_MEM_MGT_EXTENSIONS)) {
		dev_err(&ndev->dev->dev,
			"Memory registrations not supported.\n");
375
		goto out_free_pd;
376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393
	}

	list_add(&ndev->entry, &device_list);
out_unlock:
	mutex_unlock(&device_list_mutex);
	return ndev;

out_free_pd:
	ib_dealloc_pd(ndev->pd);
out_free_dev:
	kfree(ndev);
out_err:
	mutex_unlock(&device_list_mutex);
	return NULL;
}

static void nvme_rdma_destroy_queue_ib(struct nvme_rdma_queue *queue)
{
394 395 396 397 398 399 400 401
	struct nvme_rdma_device *dev;
	struct ib_device *ibdev;

	if (!test_and_clear_bit(NVME_RDMA_Q_TR_READY, &queue->flags))
		return;

	dev = queue->device;
	ibdev = dev->dev;
402

I
Israel Rukshin 已提交
403 404
	ib_mr_pool_destroy(queue->qp, &queue->qp->rdma_mrs);

405 406 407 408 409 410
	/*
	 * The cm_id object might have been destroyed during RDMA connection
	 * establishment error flow to avoid getting other cma events, thus
	 * the destruction of the QP shouldn't use rdma_cm API.
	 */
	ib_destroy_qp(queue->qp);
411 412 413 414 415 416 417 418
	ib_free_cq(queue->ib_cq);

	nvme_rdma_free_ring(ibdev, queue->rsp_ring, queue->queue_size,
			sizeof(struct nvme_completion), DMA_FROM_DEVICE);

	nvme_rdma_dev_put(dev);
}

I
Israel Rukshin 已提交
419 420 421 422 423 424
static int nvme_rdma_get_max_fr_pages(struct ib_device *ibdev)
{
	return min_t(u32, NVME_RDMA_MAX_SEGMENTS,
		     ibdev->attrs.max_fast_reg_page_list_len);
}

425
static int nvme_rdma_create_queue_ib(struct nvme_rdma_queue *queue)
426
{
427
	struct ib_device *ibdev;
428 429 430 431 432
	const int send_wr_factor = 3;			/* MR, SEND, INV */
	const int cq_factor = send_wr_factor + 1;	/* + RECV */
	int comp_vector, idx = nvme_rdma_queue_idx(queue);
	int ret;

433 434 435 436 437 438 439
	queue->device = nvme_rdma_find_get_device(queue->cm_id);
	if (!queue->device) {
		dev_err(queue->cm_id->device->dev.parent,
			"no client data found!\n");
		return -ECONNREFUSED;
	}
	ibdev = queue->device->dev;
440 441

	/*
442 443
	 * Spread I/O queues completion vectors according their queue index.
	 * Admin queues can always go on completion vector 0.
444
	 */
445
	comp_vector = idx == 0 ? idx : idx - 1;
446 447

	/* +1 for ib_stop_cq */
448 449 450
	queue->ib_cq = ib_alloc_cq(ibdev, queue,
				cq_factor * queue->queue_size + 1,
				comp_vector, IB_POLL_SOFTIRQ);
451 452
	if (IS_ERR(queue->ib_cq)) {
		ret = PTR_ERR(queue->ib_cq);
453
		goto out_put_dev;
454 455 456 457 458 459 460 461 462 463 464 465 466
	}

	ret = nvme_rdma_create_qp(queue, send_wr_factor);
	if (ret)
		goto out_destroy_ib_cq;

	queue->rsp_ring = nvme_rdma_alloc_ring(ibdev, queue->queue_size,
			sizeof(struct nvme_completion), DMA_FROM_DEVICE);
	if (!queue->rsp_ring) {
		ret = -ENOMEM;
		goto out_destroy_qp;
	}

I
Israel Rukshin 已提交
467 468 469 470 471 472 473 474 475 476 477
	ret = ib_mr_pool_init(queue->qp, &queue->qp->rdma_mrs,
			      queue->queue_size,
			      IB_MR_TYPE_MEM_REG,
			      nvme_rdma_get_max_fr_pages(ibdev));
	if (ret) {
		dev_err(queue->ctrl->ctrl.device,
			"failed to initialize MR pool sized %d for QID %d\n",
			queue->queue_size, idx);
		goto out_destroy_ring;
	}

478 479
	set_bit(NVME_RDMA_Q_TR_READY, &queue->flags);

480 481
	return 0;

I
Israel Rukshin 已提交
482 483 484
out_destroy_ring:
	nvme_rdma_free_ring(ibdev, queue->rsp_ring, queue->queue_size,
			    sizeof(struct nvme_completion), DMA_FROM_DEVICE);
485
out_destroy_qp:
486
	rdma_destroy_qp(queue->cm_id);
487 488
out_destroy_ib_cq:
	ib_free_cq(queue->ib_cq);
489 490
out_put_dev:
	nvme_rdma_dev_put(queue->device);
491 492 493
	return ret;
}

494
static int nvme_rdma_alloc_queue(struct nvme_rdma_ctrl *ctrl,
495 496 497
		int idx, size_t queue_size)
{
	struct nvme_rdma_queue *queue;
498
	struct sockaddr *src_addr = NULL;
499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519
	int ret;

	queue = &ctrl->queues[idx];
	queue->ctrl = ctrl;
	init_completion(&queue->cm_done);

	if (idx > 0)
		queue->cmnd_capsule_len = ctrl->ctrl.ioccsz * 16;
	else
		queue->cmnd_capsule_len = sizeof(struct nvme_command);

	queue->queue_size = queue_size;

	queue->cm_id = rdma_create_id(&init_net, nvme_rdma_cm_handler, queue,
			RDMA_PS_TCP, IB_QPT_RC);
	if (IS_ERR(queue->cm_id)) {
		dev_info(ctrl->ctrl.device,
			"failed to create CM ID: %ld\n", PTR_ERR(queue->cm_id));
		return PTR_ERR(queue->cm_id);
	}

520
	if (ctrl->ctrl.opts->mask & NVMF_OPT_HOST_TRADDR)
521
		src_addr = (struct sockaddr *)&ctrl->src_addr;
522

523 524 525
	queue->cm_error = -ETIMEDOUT;
	ret = rdma_resolve_addr(queue->cm_id, src_addr,
			(struct sockaddr *)&ctrl->addr,
526 527 528 529 530 531 532 533 534 535
			NVME_RDMA_CONNECT_TIMEOUT_MS);
	if (ret) {
		dev_info(ctrl->ctrl.device,
			"rdma_resolve_addr failed (%d).\n", ret);
		goto out_destroy_cm_id;
	}

	ret = nvme_rdma_wait_for_cm(queue);
	if (ret) {
		dev_info(ctrl->ctrl.device,
536
			"rdma connection establishment failed (%d)\n", ret);
537 538 539
		goto out_destroy_cm_id;
	}

540
	set_bit(NVME_RDMA_Q_ALLOCATED, &queue->flags);
541 542 543 544 545

	return 0;

out_destroy_cm_id:
	rdma_destroy_id(queue->cm_id);
546
	nvme_rdma_destroy_queue_ib(queue);
547 548 549 550 551
	return ret;
}

static void nvme_rdma_stop_queue(struct nvme_rdma_queue *queue)
{
552 553 554
	if (!test_and_clear_bit(NVME_RDMA_Q_LIVE, &queue->flags))
		return;

555 556 557 558 559 560
	rdma_disconnect(queue->cm_id);
	ib_drain_qp(queue->qp);
}

static void nvme_rdma_free_queue(struct nvme_rdma_queue *queue)
{
561
	if (!test_and_clear_bit(NVME_RDMA_Q_ALLOCATED, &queue->flags))
562 563
		return;

564 565 566 567 568 569
	if (nvme_rdma_queue_idx(queue) == 0) {
		nvme_rdma_free_qe(queue->device->dev,
			&queue->ctrl->async_event_sqe,
			sizeof(struct nvme_command), DMA_TO_DEVICE);
	}

570 571 572 573
	nvme_rdma_destroy_queue_ib(queue);
	rdma_destroy_id(queue->cm_id);
}

574
static void nvme_rdma_free_io_queues(struct nvme_rdma_ctrl *ctrl)
575
{
576 577 578 579
	int i;

	for (i = 1; i < ctrl->ctrl.queue_count; i++)
		nvme_rdma_free_queue(&ctrl->queues[i]);
580 581
}

582
static void nvme_rdma_stop_io_queues(struct nvme_rdma_ctrl *ctrl)
583 584 585
{
	int i;

586
	for (i = 1; i < ctrl->ctrl.queue_count; i++)
587
		nvme_rdma_stop_queue(&ctrl->queues[i]);
588 589
}

590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607
static int nvme_rdma_start_queue(struct nvme_rdma_ctrl *ctrl, int idx)
{
	int ret;

	if (idx)
		ret = nvmf_connect_io_queue(&ctrl->ctrl, idx);
	else
		ret = nvmf_connect_admin_queue(&ctrl->ctrl);

	if (!ret)
		set_bit(NVME_RDMA_Q_LIVE, &ctrl->queues[idx].flags);
	else
		dev_info(ctrl->ctrl.device,
			"failed to connect queue: %d ret=%d\n", idx, ret);
	return ret;
}

static int nvme_rdma_start_io_queues(struct nvme_rdma_ctrl *ctrl)
608 609 610
{
	int i, ret = 0;

611
	for (i = 1; i < ctrl->ctrl.queue_count; i++) {
612 613
		ret = nvme_rdma_start_queue(ctrl, i);
		if (ret)
614
			goto out_stop_queues;
615 616
	}

617 618
	return 0;

619
out_stop_queues:
620 621
	for (i--; i >= 1; i--)
		nvme_rdma_stop_queue(&ctrl->queues[i]);
622 623 624
	return ret;
}

625
static int nvme_rdma_alloc_io_queues(struct nvme_rdma_ctrl *ctrl)
626
{
627
	struct nvmf_ctrl_options *opts = ctrl->ctrl.opts;
628
	struct ib_device *ibdev = ctrl->device->dev;
629
	unsigned int nr_io_queues;
630 631
	int i, ret;

632
	nr_io_queues = min(opts->nr_io_queues, num_online_cpus());
633 634 635 636 637 638 639 640 641

	/*
	 * we map queues according to the device irq vectors for
	 * optimal locality so we don't need more queues than
	 * completion vectors.
	 */
	nr_io_queues = min_t(unsigned int, nr_io_queues,
				ibdev->num_comp_vectors);

642 643 644 645
	ret = nvme_set_queue_count(&ctrl->ctrl, &nr_io_queues);
	if (ret)
		return ret;

646 647
	ctrl->ctrl.queue_count = nr_io_queues + 1;
	if (ctrl->ctrl.queue_count < 2)
648 649 650 651 652
		return 0;

	dev_info(ctrl->ctrl.device,
		"creating %d I/O queues.\n", nr_io_queues);

653
	for (i = 1; i < ctrl->ctrl.queue_count; i++) {
654 655 656
		ret = nvme_rdma_alloc_queue(ctrl, i,
				ctrl->ctrl.sqsize + 1);
		if (ret)
657 658 659 660 661 662
			goto out_free_queues;
	}

	return 0;

out_free_queues:
663
	for (i--; i >= 1; i--)
664
		nvme_rdma_free_queue(&ctrl->queues[i]);
665 666 667 668

	return ret;
}

669 670
static void nvme_rdma_free_tagset(struct nvme_ctrl *nctrl,
		struct blk_mq_tag_set *set)
671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688
{
	struct nvme_rdma_ctrl *ctrl = to_rdma_ctrl(nctrl);

	blk_mq_free_tag_set(set);
	nvme_rdma_dev_put(ctrl->device);
}

static struct blk_mq_tag_set *nvme_rdma_alloc_tagset(struct nvme_ctrl *nctrl,
		bool admin)
{
	struct nvme_rdma_ctrl *ctrl = to_rdma_ctrl(nctrl);
	struct blk_mq_tag_set *set;
	int ret;

	if (admin) {
		set = &ctrl->admin_tag_set;
		memset(set, 0, sizeof(*set));
		set->ops = &nvme_rdma_admin_mq_ops;
K
Keith Busch 已提交
689
		set->queue_depth = NVME_AQ_MQ_TAG_DEPTH;
690 691 692 693 694 695 696
		set->reserved_tags = 2; /* connect + keep-alive */
		set->numa_node = NUMA_NO_NODE;
		set->cmd_size = sizeof(struct nvme_rdma_request) +
			SG_CHUNK_SIZE * sizeof(struct scatterlist);
		set->driver_data = ctrl;
		set->nr_hw_queues = 1;
		set->timeout = ADMIN_TIMEOUT;
697
		set->flags = BLK_MQ_F_NO_SCHED;
698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734
	} else {
		set = &ctrl->tag_set;
		memset(set, 0, sizeof(*set));
		set->ops = &nvme_rdma_mq_ops;
		set->queue_depth = nctrl->opts->queue_size;
		set->reserved_tags = 1; /* fabric connect */
		set->numa_node = NUMA_NO_NODE;
		set->flags = BLK_MQ_F_SHOULD_MERGE;
		set->cmd_size = sizeof(struct nvme_rdma_request) +
			SG_CHUNK_SIZE * sizeof(struct scatterlist);
		set->driver_data = ctrl;
		set->nr_hw_queues = nctrl->queue_count - 1;
		set->timeout = NVME_IO_TIMEOUT;
	}

	ret = blk_mq_alloc_tag_set(set);
	if (ret)
		goto out;

	/*
	 * We need a reference on the device as long as the tag_set is alive,
	 * as the MRs in the request structures need a valid ib_device.
	 */
	ret = nvme_rdma_dev_get(ctrl->device);
	if (!ret) {
		ret = -EINVAL;
		goto out_free_tagset;
	}

	return set;

out_free_tagset:
	blk_mq_free_tag_set(set);
out:
	return ERR_PTR(ret);
}

735 736
static void nvme_rdma_destroy_admin_queue(struct nvme_rdma_ctrl *ctrl,
		bool remove)
737
{
738
	nvme_rdma_stop_queue(&ctrl->queues[0]);
739 740
	if (remove) {
		blk_cleanup_queue(ctrl->ctrl.admin_q);
741
		nvme_rdma_free_tagset(&ctrl->ctrl, ctrl->ctrl.admin_tagset);
742
	}
743
	nvme_rdma_free_queue(&ctrl->queues[0]);
744 745
}

746 747
static int nvme_rdma_configure_admin_queue(struct nvme_rdma_ctrl *ctrl,
		bool new)
748 749 750
{
	int error;

751
	error = nvme_rdma_alloc_queue(ctrl, 0, NVME_AQ_DEPTH);
752 753 754 755 756
	if (error)
		return error;

	ctrl->device = ctrl->queues[0].device;

I
Israel Rukshin 已提交
757
	ctrl->max_fr_pages = nvme_rdma_get_max_fr_pages(ctrl->device->dev);
758

759 760
	if (new) {
		ctrl->ctrl.admin_tagset = nvme_rdma_alloc_tagset(&ctrl->ctrl, true);
761 762
		if (IS_ERR(ctrl->ctrl.admin_tagset)) {
			error = PTR_ERR(ctrl->ctrl.admin_tagset);
763
			goto out_free_queue;
764
		}
765

766 767 768 769 770
		ctrl->ctrl.admin_q = blk_mq_init_queue(&ctrl->admin_tag_set);
		if (IS_ERR(ctrl->ctrl.admin_q)) {
			error = PTR_ERR(ctrl->ctrl.admin_q);
			goto out_free_tagset;
		}
771 772
	}

773
	error = nvme_rdma_start_queue(ctrl, 0);
774 775 776
	if (error)
		goto out_cleanup_queue;

777
	error = ctrl->ctrl.ops->reg_read64(&ctrl->ctrl, NVME_REG_CAP,
778 779 780 781 782 783 784 785 786 787 788 789 790 791 792
			&ctrl->ctrl.cap);
	if (error) {
		dev_err(ctrl->ctrl.device,
			"prop_get NVME_REG_CAP failed\n");
		goto out_cleanup_queue;
	}

	ctrl->ctrl.sqsize =
		min_t(int, NVME_CAP_MQES(ctrl->ctrl.cap), ctrl->ctrl.sqsize);

	error = nvme_enable_ctrl(&ctrl->ctrl, ctrl->ctrl.cap);
	if (error)
		goto out_cleanup_queue;

	ctrl->ctrl.max_hw_sectors =
793
		(ctrl->max_fr_pages - 1) << (ilog2(SZ_4K) - 9);
794 795 796 797 798 799 800 801 802 803 804 805 806 807

	error = nvme_init_identify(&ctrl->ctrl);
	if (error)
		goto out_cleanup_queue;

	error = nvme_rdma_alloc_qe(ctrl->queues[0].device->dev,
			&ctrl->async_event_sqe, sizeof(struct nvme_command),
			DMA_TO_DEVICE);
	if (error)
		goto out_cleanup_queue;

	return 0;

out_cleanup_queue:
808 809
	if (new)
		blk_cleanup_queue(ctrl->ctrl.admin_q);
810
out_free_tagset:
811
	if (new)
812
		nvme_rdma_free_tagset(&ctrl->ctrl, ctrl->ctrl.admin_tagset);
813 814 815 816 817
out_free_queue:
	nvme_rdma_free_queue(&ctrl->queues[0]);
	return error;
}

818 819 820 821 822 823
static void nvme_rdma_destroy_io_queues(struct nvme_rdma_ctrl *ctrl,
		bool remove)
{
	nvme_rdma_stop_io_queues(ctrl);
	if (remove) {
		blk_cleanup_queue(ctrl->ctrl.connect_q);
824
		nvme_rdma_free_tagset(&ctrl->ctrl, ctrl->ctrl.tagset);
825 826 827 828 829 830 831 832
	}
	nvme_rdma_free_io_queues(ctrl);
}

static int nvme_rdma_configure_io_queues(struct nvme_rdma_ctrl *ctrl, bool new)
{
	int ret;

833
	ret = nvme_rdma_alloc_io_queues(ctrl);
834 835 836 837 838
	if (ret)
		return ret;

	if (new) {
		ctrl->ctrl.tagset = nvme_rdma_alloc_tagset(&ctrl->ctrl, false);
839 840
		if (IS_ERR(ctrl->ctrl.tagset)) {
			ret = PTR_ERR(ctrl->ctrl.tagset);
841
			goto out_free_io_queues;
842
		}
843 844 845 846 847 848 849 850 851 852 853

		ctrl->ctrl.connect_q = blk_mq_init_queue(&ctrl->tag_set);
		if (IS_ERR(ctrl->ctrl.connect_q)) {
			ret = PTR_ERR(ctrl->ctrl.connect_q);
			goto out_free_tag_set;
		}
	} else {
		blk_mq_update_nr_hw_queues(&ctrl->tag_set,
			ctrl->ctrl.queue_count - 1);
	}

854
	ret = nvme_rdma_start_io_queues(ctrl);
855 856 857 858 859 860 861 862 863 864
	if (ret)
		goto out_cleanup_connect_q;

	return 0;

out_cleanup_connect_q:
	if (new)
		blk_cleanup_queue(ctrl->ctrl.connect_q);
out_free_tag_set:
	if (new)
865
		nvme_rdma_free_tagset(&ctrl->ctrl, ctrl->ctrl.tagset);
866 867 868
out_free_io_queues:
	nvme_rdma_free_io_queues(ctrl);
	return ret;
869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887
}

static void nvme_rdma_free_ctrl(struct nvme_ctrl *nctrl)
{
	struct nvme_rdma_ctrl *ctrl = to_rdma_ctrl(nctrl);

	if (list_empty(&ctrl->list))
		goto free_ctrl;

	mutex_lock(&nvme_rdma_ctrl_mutex);
	list_del(&ctrl->list);
	mutex_unlock(&nvme_rdma_ctrl_mutex);

	kfree(ctrl->queues);
	nvmf_free_options(nctrl->opts);
free_ctrl:
	kfree(ctrl);
}

S
Sagi Grimberg 已提交
888 889 890 891 892 893 894 895 896 897 898 899
static void nvme_rdma_reconnect_or_remove(struct nvme_rdma_ctrl *ctrl)
{
	/* If we are resetting/deleting then do nothing */
	if (ctrl->ctrl.state != NVME_CTRL_RECONNECTING) {
		WARN_ON_ONCE(ctrl->ctrl.state == NVME_CTRL_NEW ||
			ctrl->ctrl.state == NVME_CTRL_LIVE);
		return;
	}

	if (nvmf_should_reconnect(&ctrl->ctrl)) {
		dev_info(ctrl->ctrl.device, "Reconnecting in %d seconds...\n",
			ctrl->ctrl.opts->reconnect_delay);
900
		queue_delayed_work(nvme_wq, &ctrl->reconnect_work,
S
Sagi Grimberg 已提交
901 902 903
				ctrl->ctrl.opts->reconnect_delay * HZ);
	} else {
		dev_info(ctrl->ctrl.device, "Removing controller...\n");
904
		nvme_delete_ctrl(&ctrl->ctrl);
S
Sagi Grimberg 已提交
905 906 907
	}
}

908 909 910 911 912 913 914
static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work)
{
	struct nvme_rdma_ctrl *ctrl = container_of(to_delayed_work(work),
			struct nvme_rdma_ctrl, reconnect_work);
	bool changed;
	int ret;

915
	++ctrl->ctrl.nr_reconnects;
S
Sagi Grimberg 已提交
916

917
	ret = nvme_rdma_configure_admin_queue(ctrl, false);
918
	if (ret)
919
		goto requeue;
920

921
	if (ctrl->ctrl.queue_count > 1) {
922
		ret = nvme_rdma_configure_io_queues(ctrl, false);
923
		if (ret)
924
			goto destroy_admin;
925 926 927
	}

	changed = nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_LIVE);
928 929 930 931 932 933
	if (!changed) {
		/* state change failure is ok if we're in DELETING state */
		WARN_ON_ONCE(ctrl->ctrl.state != NVME_CTRL_DELETING);
		return;
	}

934
	nvme_start_ctrl(&ctrl->ctrl);
935

936 937 938 939
	dev_info(ctrl->ctrl.device, "Successfully reconnected (%d attempts)\n",
			ctrl->ctrl.nr_reconnects);

	ctrl->ctrl.nr_reconnects = 0;
940 941 942

	return;

943 944
destroy_admin:
	nvme_rdma_destroy_admin_queue(ctrl, false);
945
requeue:
S
Sagi Grimberg 已提交
946
	dev_info(ctrl->ctrl.device, "Failed reconnect attempt %d\n",
947
			ctrl->ctrl.nr_reconnects);
S
Sagi Grimberg 已提交
948
	nvme_rdma_reconnect_or_remove(ctrl);
949 950 951 952 953 954 955
}

static void nvme_rdma_error_recovery_work(struct work_struct *work)
{
	struct nvme_rdma_ctrl *ctrl = container_of(work,
			struct nvme_rdma_ctrl, err_work);

956
	nvme_stop_keep_alive(&ctrl->ctrl);
957

958
	if (ctrl->ctrl.queue_count > 1) {
959 960 961
		nvme_stop_queues(&ctrl->ctrl);
		blk_mq_tagset_busy_iter(&ctrl->tag_set,
					nvme_cancel_request, &ctrl->ctrl);
962 963 964 965
		nvme_rdma_destroy_io_queues(ctrl, false);
	}

	blk_mq_quiesce_queue(ctrl->ctrl.admin_q);
966 967
	blk_mq_tagset_busy_iter(&ctrl->admin_tag_set,
				nvme_cancel_request, &ctrl->ctrl);
968
	nvme_rdma_destroy_admin_queue(ctrl, false);
969

970 971 972 973
	/*
	 * queues are not a live anymore, so restart the queues to fail fast
	 * new IO
	 */
974
	blk_mq_unquiesce_queue(ctrl->ctrl.admin_q);
975 976
	nvme_start_queues(&ctrl->ctrl);

S
Sagi Grimberg 已提交
977
	nvme_rdma_reconnect_or_remove(ctrl);
978 979 980 981 982 983 984
}

static void nvme_rdma_error_recovery(struct nvme_rdma_ctrl *ctrl)
{
	if (!nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_RECONNECTING))
		return;

985
	queue_work(nvme_wq, &ctrl->err_work);
986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009
}

static void nvme_rdma_wr_error(struct ib_cq *cq, struct ib_wc *wc,
		const char *op)
{
	struct nvme_rdma_queue *queue = cq->cq_context;
	struct nvme_rdma_ctrl *ctrl = queue->ctrl;

	if (ctrl->ctrl.state == NVME_CTRL_LIVE)
		dev_info(ctrl->ctrl.device,
			     "%s for CQE 0x%p failed with status %s (%d)\n",
			     op, wc->wr_cqe,
			     ib_wc_status_msg(wc->status), wc->status);
	nvme_rdma_error_recovery(ctrl);
}

static void nvme_rdma_memreg_done(struct ib_cq *cq, struct ib_wc *wc)
{
	if (unlikely(wc->status != IB_WC_SUCCESS))
		nvme_rdma_wr_error(cq, wc, "MEMREG");
}

static void nvme_rdma_inv_rkey_done(struct ib_cq *cq, struct ib_wc *wc)
{
1010 1011 1012 1013 1014
	struct nvme_rdma_request *req =
		container_of(wc->wr_cqe, struct nvme_rdma_request, reg_cqe);
	struct request *rq = blk_mq_rq_from_pdu(req);

	if (unlikely(wc->status != IB_WC_SUCCESS)) {
1015
		nvme_rdma_wr_error(cq, wc, "LOCAL_INV");
1016 1017 1018 1019 1020 1021
		return;
	}

	if (refcount_dec_and_test(&req->ref))
		nvme_end_request(rq, req->status, req->result);

1022 1023 1024 1025 1026 1027 1028 1029 1030 1031
}

static int nvme_rdma_inv_rkey(struct nvme_rdma_queue *queue,
		struct nvme_rdma_request *req)
{
	struct ib_send_wr *bad_wr;
	struct ib_send_wr wr = {
		.opcode		    = IB_WR_LOCAL_INV,
		.next		    = NULL,
		.num_sge	    = 0,
1032
		.send_flags	    = IB_SEND_SIGNALED,
1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051
		.ex.invalidate_rkey = req->mr->rkey,
	};

	req->reg_cqe.done = nvme_rdma_inv_rkey_done;
	wr.wr_cqe = &req->reg_cqe;

	return ib_post_send(queue->qp, &wr, &bad_wr);
}

static void nvme_rdma_unmap_data(struct nvme_rdma_queue *queue,
		struct request *rq)
{
	struct nvme_rdma_request *req = blk_mq_rq_to_pdu(rq);
	struct nvme_rdma_device *dev = queue->device;
	struct ib_device *ibdev = dev->dev;

	if (!blk_rq_bytes(rq))
		return;

I
Israel Rukshin 已提交
1052 1053 1054 1055 1056
	if (req->mr) {
		ib_mr_pool_put(queue->qp, &queue->qp->rdma_mrs, req->mr);
		req->mr = NULL;
	}

1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100
	ib_dma_unmap_sg(ibdev, req->sg_table.sgl,
			req->nents, rq_data_dir(rq) ==
				    WRITE ? DMA_TO_DEVICE : DMA_FROM_DEVICE);

	nvme_cleanup_cmd(rq);
	sg_free_table_chained(&req->sg_table, true);
}

static int nvme_rdma_set_sg_null(struct nvme_command *c)
{
	struct nvme_keyed_sgl_desc *sg = &c->common.dptr.ksgl;

	sg->addr = 0;
	put_unaligned_le24(0, sg->length);
	put_unaligned_le32(0, sg->key);
	sg->type = NVME_KEY_SGL_FMT_DATA_DESC << 4;
	return 0;
}

static int nvme_rdma_map_sg_inline(struct nvme_rdma_queue *queue,
		struct nvme_rdma_request *req, struct nvme_command *c)
{
	struct nvme_sgl_desc *sg = &c->common.dptr.sgl;

	req->sge[1].addr = sg_dma_address(req->sg_table.sgl);
	req->sge[1].length = sg_dma_len(req->sg_table.sgl);
	req->sge[1].lkey = queue->device->pd->local_dma_lkey;

	sg->addr = cpu_to_le64(queue->ctrl->ctrl.icdoff);
	sg->length = cpu_to_le32(sg_dma_len(req->sg_table.sgl));
	sg->type = (NVME_SGL_FMT_DATA_DESC << 4) | NVME_SGL_FMT_OFFSET;

	req->inline_data = true;
	req->num_sge++;
	return 0;
}

static int nvme_rdma_map_sg_single(struct nvme_rdma_queue *queue,
		struct nvme_rdma_request *req, struct nvme_command *c)
{
	struct nvme_keyed_sgl_desc *sg = &c->common.dptr.ksgl;

	sg->addr = cpu_to_le64(sg_dma_address(req->sg_table.sgl));
	put_unaligned_le24(sg_dma_len(req->sg_table.sgl), sg->length);
1101
	put_unaligned_le32(queue->device->pd->unsafe_global_rkey, sg->key);
1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112
	sg->type = NVME_KEY_SGL_FMT_DATA_DESC << 4;
	return 0;
}

static int nvme_rdma_map_sg_fr(struct nvme_rdma_queue *queue,
		struct nvme_rdma_request *req, struct nvme_command *c,
		int count)
{
	struct nvme_keyed_sgl_desc *sg = &c->common.dptr.ksgl;
	int nr;

I
Israel Rukshin 已提交
1113 1114 1115 1116
	req->mr = ib_mr_pool_get(queue->qp, &queue->qp->rdma_mrs);
	if (WARN_ON_ONCE(!req->mr))
		return -EAGAIN;

1117 1118 1119 1120 1121
	/*
	 * Align the MR to a 4K page size to match the ctrl page size and
	 * the block virtual boundary.
	 */
	nr = ib_map_mr_sg(req->mr, req->sg_table.sgl, count, NULL, SZ_4K);
1122
	if (unlikely(nr < count)) {
I
Israel Rukshin 已提交
1123 1124
		ib_mr_pool_put(queue->qp, &queue->qp->rdma_mrs, req->mr);
		req->mr = NULL;
1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152
		if (nr < 0)
			return nr;
		return -EINVAL;
	}

	ib_update_fast_reg_key(req->mr, ib_inc_rkey(req->mr->rkey));

	req->reg_cqe.done = nvme_rdma_memreg_done;
	memset(&req->reg_wr, 0, sizeof(req->reg_wr));
	req->reg_wr.wr.opcode = IB_WR_REG_MR;
	req->reg_wr.wr.wr_cqe = &req->reg_cqe;
	req->reg_wr.wr.num_sge = 0;
	req->reg_wr.mr = req->mr;
	req->reg_wr.key = req->mr->rkey;
	req->reg_wr.access = IB_ACCESS_LOCAL_WRITE |
			     IB_ACCESS_REMOTE_READ |
			     IB_ACCESS_REMOTE_WRITE;

	sg->addr = cpu_to_le64(req->mr->iova);
	put_unaligned_le24(req->mr->length, sg->length);
	put_unaligned_le32(req->mr->rkey, sg->key);
	sg->type = (NVME_KEY_SGL_FMT_DATA_DESC << 4) |
			NVME_SGL_FMT_INVALIDATE;

	return 0;
}

static int nvme_rdma_map_data(struct nvme_rdma_queue *queue,
1153
		struct request *rq, struct nvme_command *c)
1154 1155 1156 1157
{
	struct nvme_rdma_request *req = blk_mq_rq_to_pdu(rq);
	struct nvme_rdma_device *dev = queue->device;
	struct ib_device *ibdev = dev->dev;
1158
	int count, ret;
1159 1160 1161

	req->num_sge = 1;
	req->inline_data = false;
1162
	refcount_set(&req->ref, 2); /* send and recv completions */
1163 1164 1165 1166 1167 1168 1169

	c->common.flags |= NVME_CMD_SGL_METABUF;

	if (!blk_rq_bytes(rq))
		return nvme_rdma_set_sg_null(c);

	req->sg_table.sgl = req->first_sgl;
1170 1171
	ret = sg_alloc_table_chained(&req->sg_table,
			blk_rq_nr_phys_segments(rq), req->sg_table.sgl);
1172 1173 1174
	if (ret)
		return -ENOMEM;

1175
	req->nents = blk_rq_map_sg(rq->q, rq, req->sg_table.sgl);
1176

1177
	count = ib_dma_map_sg(ibdev, req->sg_table.sgl, req->nents,
1178 1179 1180 1181 1182 1183 1184
		    rq_data_dir(rq) == WRITE ? DMA_TO_DEVICE : DMA_FROM_DEVICE);
	if (unlikely(count <= 0)) {
		sg_free_table_chained(&req->sg_table, true);
		return -EIO;
	}

	if (count == 1) {
1185 1186 1187
		if (rq_data_dir(rq) == WRITE && nvme_rdma_queue_idx(queue) &&
		    blk_rq_payload_bytes(rq) <=
				nvme_rdma_inline_data_size(queue))
1188 1189
			return nvme_rdma_map_sg_inline(queue, req, c);

1190
		if (dev->pd->flags & IB_PD_UNSAFE_GLOBAL_RKEY)
1191 1192 1193 1194 1195 1196 1197 1198
			return nvme_rdma_map_sg_single(queue, req, c);
	}

	return nvme_rdma_map_sg_fr(queue, req, c, count);
}

static void nvme_rdma_send_done(struct ib_cq *cq, struct ib_wc *wc)
{
1199 1200 1201 1202 1203 1204 1205
	struct nvme_rdma_qe *qe =
		container_of(wc->wr_cqe, struct nvme_rdma_qe, cqe);
	struct nvme_rdma_request *req =
		container_of(qe, struct nvme_rdma_request, sqe);
	struct request *rq = blk_mq_rq_from_pdu(req);

	if (unlikely(wc->status != IB_WC_SUCCESS)) {
1206
		nvme_rdma_wr_error(cq, wc, "SEND");
1207 1208 1209 1210 1211
		return;
	}

	if (refcount_dec_and_test(&req->ref))
		nvme_end_request(rq, req->status, req->result);
1212 1213 1214 1215
}

static int nvme_rdma_post_send(struct nvme_rdma_queue *queue,
		struct nvme_rdma_qe *qe, struct ib_sge *sge, u32 num_sge,
1216
		struct ib_send_wr *first)
1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229
{
	struct ib_send_wr wr, *bad_wr;
	int ret;

	sge->addr   = qe->dma;
	sge->length = sizeof(struct nvme_command),
	sge->lkey   = queue->device->pd->local_dma_lkey;

	wr.next       = NULL;
	wr.wr_cqe     = &qe->cqe;
	wr.sg_list    = sge;
	wr.num_sge    = num_sge;
	wr.opcode     = IB_WR_SEND;
1230
	wr.send_flags = IB_SEND_SIGNALED;
1231 1232 1233 1234 1235 1236 1237

	if (first)
		first->next = &wr;
	else
		first = &wr;

	ret = ib_post_send(queue->qp, first, &bad_wr);
1238
	if (unlikely(ret)) {
1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263
		dev_err(queue->ctrl->ctrl.device,
			     "%s failed with error code %d\n", __func__, ret);
	}
	return ret;
}

static int nvme_rdma_post_recv(struct nvme_rdma_queue *queue,
		struct nvme_rdma_qe *qe)
{
	struct ib_recv_wr wr, *bad_wr;
	struct ib_sge list;
	int ret;

	list.addr   = qe->dma;
	list.length = sizeof(struct nvme_completion);
	list.lkey   = queue->device->pd->local_dma_lkey;

	qe->cqe.done = nvme_rdma_recv_done;

	wr.next     = NULL;
	wr.wr_cqe   = &qe->cqe;
	wr.sg_list  = &list;
	wr.num_sge  = 1;

	ret = ib_post_recv(queue->qp, &wr, &bad_wr);
1264
	if (unlikely(ret)) {
1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279
		dev_err(queue->ctrl->ctrl.device,
			"%s failed with error code %d\n", __func__, ret);
	}
	return ret;
}

static struct blk_mq_tags *nvme_rdma_tagset(struct nvme_rdma_queue *queue)
{
	u32 queue_idx = nvme_rdma_queue_idx(queue);

	if (queue_idx == 0)
		return queue->ctrl->admin_tag_set.tags[queue_idx];
	return queue->ctrl->tag_set.tags[queue_idx - 1];
}

1280 1281 1282 1283 1284 1285
static void nvme_rdma_async_done(struct ib_cq *cq, struct ib_wc *wc)
{
	if (unlikely(wc->status != IB_WC_SUCCESS))
		nvme_rdma_wr_error(cq, wc, "ASYNC");
}

1286
static void nvme_rdma_submit_async_event(struct nvme_ctrl *arg)
1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299
{
	struct nvme_rdma_ctrl *ctrl = to_rdma_ctrl(arg);
	struct nvme_rdma_queue *queue = &ctrl->queues[0];
	struct ib_device *dev = queue->device->dev;
	struct nvme_rdma_qe *sqe = &ctrl->async_event_sqe;
	struct nvme_command *cmd = sqe->data;
	struct ib_sge sge;
	int ret;

	ib_dma_sync_single_for_cpu(dev, sqe->dma, sizeof(*cmd), DMA_TO_DEVICE);

	memset(cmd, 0, sizeof(*cmd));
	cmd->common.opcode = nvme_admin_async_event;
K
Keith Busch 已提交
1300
	cmd->common.command_id = NVME_AQ_BLK_MQ_DEPTH;
1301 1302 1303
	cmd->common.flags |= NVME_CMD_SGL_METABUF;
	nvme_rdma_set_sg_null(cmd);

1304 1305
	sqe->cqe.done = nvme_rdma_async_done;

1306 1307 1308
	ib_dma_sync_single_for_device(dev, sqe->dma, sizeof(*cmd),
			DMA_TO_DEVICE);

1309
	ret = nvme_rdma_post_send(queue, sqe, &sge, 1, NULL);
1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329
	WARN_ON_ONCE(ret);
}

static int nvme_rdma_process_nvme_rsp(struct nvme_rdma_queue *queue,
		struct nvme_completion *cqe, struct ib_wc *wc, int tag)
{
	struct request *rq;
	struct nvme_rdma_request *req;
	int ret = 0;

	rq = blk_mq_tag_to_rq(nvme_rdma_tagset(queue), cqe->command_id);
	if (!rq) {
		dev_err(queue->ctrl->ctrl.device,
			"tag 0x%x on QP %#x not found\n",
			cqe->command_id, queue->qp->qp_num);
		nvme_rdma_error_recovery(queue->ctrl);
		return ret;
	}
	req = blk_mq_rq_to_pdu(rq);

1330 1331
	req->status = cqe->status;
	req->result = cqe->result;
1332

1333 1334 1335 1336 1337 1338 1339
	if (wc->wc_flags & IB_WC_WITH_INVALIDATE) {
		if (unlikely(wc->ex.invalidate_rkey != req->mr->rkey)) {
			dev_err(queue->ctrl->ctrl.device,
				"Bogus remote invalidation for rkey %#x\n",
				req->mr->rkey);
			nvme_rdma_error_recovery(queue->ctrl);
		}
I
Israel Rukshin 已提交
1340
	} else if (req->mr) {
1341 1342 1343 1344 1345 1346 1347 1348 1349 1350
		ret = nvme_rdma_inv_rkey(queue, req);
		if (unlikely(ret < 0)) {
			dev_err(queue->ctrl->ctrl.device,
				"Queueing INV WR for rkey %#x failed (%d)\n",
				req->mr->rkey, ret);
			nvme_rdma_error_recovery(queue->ctrl);
		}
		/* the local invalidation completion will end the request */
		return 0;
	}
1351

1352 1353 1354 1355 1356 1357
	if (refcount_dec_and_test(&req->ref)) {
		if (rq->tag == tag)
			ret = 1;
		nvme_end_request(rq, req->status, req->result);
	}

1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383
	return ret;
}

static int __nvme_rdma_recv_done(struct ib_cq *cq, struct ib_wc *wc, int tag)
{
	struct nvme_rdma_qe *qe =
		container_of(wc->wr_cqe, struct nvme_rdma_qe, cqe);
	struct nvme_rdma_queue *queue = cq->cq_context;
	struct ib_device *ibdev = queue->device->dev;
	struct nvme_completion *cqe = qe->data;
	const size_t len = sizeof(struct nvme_completion);
	int ret = 0;

	if (unlikely(wc->status != IB_WC_SUCCESS)) {
		nvme_rdma_wr_error(cq, wc, "RECV");
		return 0;
	}

	ib_dma_sync_single_for_cpu(ibdev, qe->dma, len, DMA_FROM_DEVICE);
	/*
	 * AEN requests are special as they don't time out and can
	 * survive any kind of queue freeze and often don't respond to
	 * aborts.  We don't even bother to allocate a struct request
	 * for them but rather special case them here.
	 */
	if (unlikely(nvme_rdma_queue_idx(queue) == 0 &&
K
Keith Busch 已提交
1384
			cqe->command_id >= NVME_AQ_BLK_MQ_DEPTH))
1385 1386
		nvme_complete_async_event(&queue->ctrl->ctrl, cqe->status,
				&cqe->result);
1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419
	else
		ret = nvme_rdma_process_nvme_rsp(queue, cqe, wc, tag);
	ib_dma_sync_single_for_device(ibdev, qe->dma, len, DMA_FROM_DEVICE);

	nvme_rdma_post_recv(queue, qe);
	return ret;
}

static void nvme_rdma_recv_done(struct ib_cq *cq, struct ib_wc *wc)
{
	__nvme_rdma_recv_done(cq, wc, -1);
}

static int nvme_rdma_conn_established(struct nvme_rdma_queue *queue)
{
	int ret, i;

	for (i = 0; i < queue->queue_size; i++) {
		ret = nvme_rdma_post_recv(queue, &queue->rsp_ring[i]);
		if (ret)
			goto out_destroy_queue_ib;
	}

	return 0;

out_destroy_queue_ib:
	nvme_rdma_destroy_queue_ib(queue);
	return ret;
}

static int nvme_rdma_conn_rejected(struct nvme_rdma_queue *queue,
		struct rdma_cm_event *ev)
{
1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430
	struct rdma_cm_id *cm_id = queue->cm_id;
	int status = ev->status;
	const char *rej_msg;
	const struct nvme_rdma_cm_rej *rej_data;
	u8 rej_data_len;

	rej_msg = rdma_reject_msg(cm_id, status);
	rej_data = rdma_consumer_reject_data(cm_id, ev, &rej_data_len);

	if (rej_data && rej_data_len >= sizeof(u16)) {
		u16 sts = le16_to_cpu(rej_data->sts);
1431 1432

		dev_err(queue->ctrl->ctrl.device,
1433 1434
		      "Connect rejected: status %d (%s) nvme status %d (%s).\n",
		      status, rej_msg, sts, nvme_rdma_cm_msg(sts));
1435 1436
	} else {
		dev_err(queue->ctrl->ctrl.device,
1437
			"Connect rejected: status %d (%s).\n", status, rej_msg);
1438 1439 1440 1441 1442 1443 1444 1445 1446
	}

	return -ECONNRESET;
}

static int nvme_rdma_addr_resolved(struct nvme_rdma_queue *queue)
{
	int ret;

1447 1448 1449
	ret = nvme_rdma_create_queue_ib(queue);
	if (ret)
		return ret;
1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469

	ret = rdma_resolve_route(queue->cm_id, NVME_RDMA_CONNECT_TIMEOUT_MS);
	if (ret) {
		dev_err(queue->ctrl->ctrl.device,
			"rdma_resolve_route failed (%d).\n",
			queue->cm_error);
		goto out_destroy_queue;
	}

	return 0;

out_destroy_queue:
	nvme_rdma_destroy_queue_ib(queue);
	return ret;
}

static int nvme_rdma_route_resolved(struct nvme_rdma_queue *queue)
{
	struct nvme_rdma_ctrl *ctrl = queue->ctrl;
	struct rdma_conn_param param = { };
1470
	struct nvme_rdma_cm_req priv = { };
1471 1472 1473 1474 1475 1476
	int ret;

	param.qp_num = queue->qp->qp_num;
	param.flow_control = 1;

	param.responder_resources = queue->device->dev->attrs.max_qp_rd_atom;
1477 1478
	/* maximum retry count */
	param.retry_count = 7;
1479 1480 1481 1482 1483 1484
	param.rnr_retry_count = 7;
	param.private_data = &priv;
	param.private_data_len = sizeof(priv);

	priv.recfmt = cpu_to_le16(NVME_RDMA_CM_FMT_1_0);
	priv.qid = cpu_to_le16(nvme_rdma_queue_idx(queue));
1485 1486 1487 1488 1489
	/*
	 * set the admin queue depth to the minimum size
	 * specified by the Fabrics standard.
	 */
	if (priv.qid == 0) {
1490 1491
		priv.hrqsize = cpu_to_le16(NVME_AQ_DEPTH);
		priv.hsqsize = cpu_to_le16(NVME_AQ_DEPTH - 1);
1492
	} else {
1493 1494 1495 1496 1497
		/*
		 * current interpretation of the fabrics spec
		 * is at minimum you make hrqsize sqsize+1, or a
		 * 1's based representation of sqsize.
		 */
1498
		priv.hrqsize = cpu_to_le16(queue->queue_size);
1499
		priv.hsqsize = cpu_to_le16(queue->ctrl->ctrl.sqsize);
1500
	}
1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538

	ret = rdma_connect(queue->cm_id, &param);
	if (ret) {
		dev_err(ctrl->ctrl.device,
			"rdma_connect failed (%d).\n", ret);
		goto out_destroy_queue_ib;
	}

	return 0;

out_destroy_queue_ib:
	nvme_rdma_destroy_queue_ib(queue);
	return ret;
}

static int nvme_rdma_cm_handler(struct rdma_cm_id *cm_id,
		struct rdma_cm_event *ev)
{
	struct nvme_rdma_queue *queue = cm_id->context;
	int cm_error = 0;

	dev_dbg(queue->ctrl->ctrl.device, "%s (%d): status %d id %p\n",
		rdma_event_msg(ev->event), ev->event,
		ev->status, cm_id);

	switch (ev->event) {
	case RDMA_CM_EVENT_ADDR_RESOLVED:
		cm_error = nvme_rdma_addr_resolved(queue);
		break;
	case RDMA_CM_EVENT_ROUTE_RESOLVED:
		cm_error = nvme_rdma_route_resolved(queue);
		break;
	case RDMA_CM_EVENT_ESTABLISHED:
		queue->cm_error = nvme_rdma_conn_established(queue);
		/* complete cm_done regardless of success/failure */
		complete(&queue->cm_done);
		return 0;
	case RDMA_CM_EVENT_REJECTED:
1539
		nvme_rdma_destroy_queue_ib(queue);
1540 1541 1542 1543 1544
		cm_error = nvme_rdma_conn_rejected(queue, ev);
		break;
	case RDMA_CM_EVENT_ROUTE_ERROR:
	case RDMA_CM_EVENT_CONNECT_ERROR:
	case RDMA_CM_EVENT_UNREACHABLE:
1545 1546
		nvme_rdma_destroy_queue_ib(queue);
	case RDMA_CM_EVENT_ADDR_ERROR:
1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558
		dev_dbg(queue->ctrl->ctrl.device,
			"CM error event %d\n", ev->event);
		cm_error = -ECONNRESET;
		break;
	case RDMA_CM_EVENT_DISCONNECTED:
	case RDMA_CM_EVENT_ADDR_CHANGE:
	case RDMA_CM_EVENT_TIMEWAIT_EXIT:
		dev_dbg(queue->ctrl->ctrl.device,
			"disconnect received - connection closed\n");
		nvme_rdma_error_recovery(queue->ctrl);
		break;
	case RDMA_CM_EVENT_DEVICE_REMOVAL:
1559 1560
		/* device removal is handled via the ib_client API */
		break;
1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580
	default:
		dev_err(queue->ctrl->ctrl.device,
			"Unexpected RDMA CM event (%d)\n", ev->event);
		nvme_rdma_error_recovery(queue->ctrl);
		break;
	}

	if (cm_error) {
		queue->cm_error = cm_error;
		complete(&queue->cm_done);
	}

	return 0;
}

static enum blk_eh_timer_return
nvme_rdma_timeout(struct request *rq, bool reserved)
{
	struct nvme_rdma_request *req = blk_mq_rq_to_pdu(rq);

1581 1582 1583 1584
	dev_warn(req->queue->ctrl->ctrl.device,
		 "I/O %d QID %d timeout, reset controller\n",
		 rq->tag, nvme_rdma_queue_idx(req->queue));

1585 1586 1587 1588
	/* queue error recovery */
	nvme_rdma_error_recovery(req->queue->ctrl);

	/* fail with DNR on cmd timeout */
1589
	nvme_req(rq)->status = NVME_SC_ABORT_REQ | NVME_SC_DNR;
1590 1591 1592 1593

	return BLK_EH_HANDLED;
}

1594 1595 1596
/*
 * We cannot accept any other command until the Connect command has completed.
 */
C
Christoph Hellwig 已提交
1597
static inline blk_status_t
1598 1599 1600 1601 1602
nvme_rdma_is_ready(struct nvme_rdma_queue *queue, struct request *rq)
{
	if (unlikely(!test_bit(NVME_RDMA_Q_LIVE, &queue->flags)))
		return nvmf_check_init_req(&queue->ctrl->ctrl, rq);
	return BLK_STS_OK;
1603 1604
}

1605
static blk_status_t nvme_rdma_queue_rq(struct blk_mq_hw_ctx *hctx,
1606 1607 1608 1609 1610 1611 1612 1613 1614
		const struct blk_mq_queue_data *bd)
{
	struct nvme_ns *ns = hctx->queue->queuedata;
	struct nvme_rdma_queue *queue = hctx->driver_data;
	struct request *rq = bd->rq;
	struct nvme_rdma_request *req = blk_mq_rq_to_pdu(rq);
	struct nvme_rdma_qe *sqe = &req->sqe;
	struct nvme_command *c = sqe->data;
	struct ib_device *dev;
1615 1616
	blk_status_t ret;
	int err;
1617 1618 1619

	WARN_ON_ONCE(rq->tag < 0);

1620
	ret = nvme_rdma_is_ready(queue, rq);
1621
	if (unlikely(ret))
C
Christoph Hellwig 已提交
1622
		return ret;
1623

1624 1625 1626 1627 1628
	dev = queue->device->dev;
	ib_dma_sync_single_for_cpu(dev, sqe->dma,
			sizeof(struct nvme_command), DMA_TO_DEVICE);

	ret = nvme_setup_cmd(ns, rq, c);
1629
	if (ret)
1630 1631 1632 1633
		return ret;

	blk_mq_start_request(rq);

1634
	err = nvme_rdma_map_data(queue, rq, c);
1635
	if (unlikely(err < 0)) {
1636
		dev_err(queue->ctrl->ctrl.device,
1637
			     "Failed to map data (%d)\n", err);
1638 1639 1640 1641
		nvme_cleanup_cmd(rq);
		goto err;
	}

1642 1643
	sqe->cqe.done = nvme_rdma_send_done;

1644 1645 1646
	ib_dma_sync_single_for_device(dev, sqe->dma,
			sizeof(struct nvme_command), DMA_TO_DEVICE);

1647
	err = nvme_rdma_post_send(queue, sqe, req->sge, req->num_sge,
I
Israel Rukshin 已提交
1648
			req->mr ? &req->reg_wr.wr : NULL);
1649
	if (unlikely(err)) {
1650 1651 1652 1653
		nvme_rdma_unmap_data(queue, rq);
		goto err;
	}

1654
	return BLK_STS_OK;
1655
err:
1656 1657 1658
	if (err == -ENOMEM || err == -EAGAIN)
		return BLK_STS_RESOURCE;
	return BLK_STS_IOERR;
1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685
}

static int nvme_rdma_poll(struct blk_mq_hw_ctx *hctx, unsigned int tag)
{
	struct nvme_rdma_queue *queue = hctx->driver_data;
	struct ib_cq *cq = queue->ib_cq;
	struct ib_wc wc;
	int found = 0;

	while (ib_poll_cq(cq, 1, &wc) > 0) {
		struct ib_cqe *cqe = wc.wr_cqe;

		if (cqe) {
			if (cqe->done == nvme_rdma_recv_done)
				found |= __nvme_rdma_recv_done(cq, &wc, tag);
			else
				cqe->done(cq, &wc);
		}
	}

	return found;
}

static void nvme_rdma_complete_rq(struct request *rq)
{
	struct nvme_rdma_request *req = blk_mq_rq_to_pdu(rq);

1686 1687
	nvme_rdma_unmap_data(req->queue, rq);
	nvme_complete_rq(rq);
1688 1689
}

1690 1691 1692 1693 1694 1695 1696
static int nvme_rdma_map_queues(struct blk_mq_tag_set *set)
{
	struct nvme_rdma_ctrl *ctrl = set->driver_data;

	return blk_mq_rdma_map_queues(set, ctrl->device->dev, 0);
}

1697
static const struct blk_mq_ops nvme_rdma_mq_ops = {
1698 1699 1700 1701 1702 1703 1704
	.queue_rq	= nvme_rdma_queue_rq,
	.complete	= nvme_rdma_complete_rq,
	.init_request	= nvme_rdma_init_request,
	.exit_request	= nvme_rdma_exit_request,
	.init_hctx	= nvme_rdma_init_hctx,
	.poll		= nvme_rdma_poll,
	.timeout	= nvme_rdma_timeout,
1705
	.map_queues	= nvme_rdma_map_queues,
1706 1707
};

1708
static const struct blk_mq_ops nvme_rdma_admin_mq_ops = {
1709 1710
	.queue_rq	= nvme_rdma_queue_rq,
	.complete	= nvme_rdma_complete_rq,
1711 1712
	.init_request	= nvme_rdma_init_request,
	.exit_request	= nvme_rdma_exit_request,
1713 1714 1715 1716
	.init_hctx	= nvme_rdma_init_admin_hctx,
	.timeout	= nvme_rdma_timeout,
};

1717
static void nvme_rdma_shutdown_ctrl(struct nvme_rdma_ctrl *ctrl, bool shutdown)
1718 1719 1720 1721
{
	cancel_work_sync(&ctrl->err_work);
	cancel_delayed_work_sync(&ctrl->reconnect_work);

1722
	if (ctrl->ctrl.queue_count > 1) {
1723 1724 1725
		nvme_stop_queues(&ctrl->ctrl);
		blk_mq_tagset_busy_iter(&ctrl->tag_set,
					nvme_cancel_request, &ctrl->ctrl);
1726
		nvme_rdma_destroy_io_queues(ctrl, shutdown);
1727 1728
	}

1729
	if (shutdown)
1730
		nvme_shutdown_ctrl(&ctrl->ctrl);
1731 1732
	else
		nvme_disable_ctrl(&ctrl->ctrl, ctrl->ctrl.cap);
1733

1734
	blk_mq_quiesce_queue(ctrl->ctrl.admin_q);
1735 1736
	blk_mq_tagset_busy_iter(&ctrl->admin_tag_set,
				nvme_cancel_request, &ctrl->ctrl);
1737
	blk_mq_unquiesce_queue(ctrl->ctrl.admin_q);
1738
	nvme_rdma_destroy_admin_queue(ctrl, shutdown);
1739 1740
}

1741
static void nvme_rdma_delete_ctrl(struct nvme_ctrl *ctrl)
1742
{
1743
	nvme_rdma_shutdown_ctrl(to_rdma_ctrl(ctrl), true);
1744 1745 1746 1747
}

static void nvme_rdma_reset_ctrl_work(struct work_struct *work)
{
1748 1749
	struct nvme_rdma_ctrl *ctrl =
		container_of(work, struct nvme_rdma_ctrl, ctrl.reset_work);
1750 1751 1752
	int ret;
	bool changed;

1753
	nvme_stop_ctrl(&ctrl->ctrl);
1754
	nvme_rdma_shutdown_ctrl(ctrl, false);
1755

1756
	ret = nvme_rdma_configure_admin_queue(ctrl, false);
1757 1758
	if (ret)
		goto out_fail;
1759

1760
	if (ctrl->ctrl.queue_count > 1) {
1761
		ret = nvme_rdma_configure_io_queues(ctrl, false);
1762
		if (ret)
1763
			goto out_fail;
1764 1765 1766
	}

	changed = nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_LIVE);
1767 1768 1769 1770 1771
	if (!changed) {
		/* state change failure is ok if we're in DELETING state */
		WARN_ON_ONCE(ctrl->ctrl.state != NVME_CTRL_DELETING);
		return;
	}
1772

1773
	nvme_start_ctrl(&ctrl->ctrl);
1774 1775 1776

	return;

1777
out_fail:
1778
	dev_warn(ctrl->ctrl.device, "Removing after reset failure\n");
1779 1780 1781 1782
	nvme_remove_namespaces(&ctrl->ctrl);
	nvme_rdma_shutdown_ctrl(ctrl, true);
	nvme_uninit_ctrl(&ctrl->ctrl);
	nvme_put_ctrl(&ctrl->ctrl);
1783 1784 1785 1786 1787
}

static const struct nvme_ctrl_ops nvme_rdma_ctrl_ops = {
	.name			= "rdma",
	.module			= THIS_MODULE,
1788
	.flags			= NVME_F_FABRICS,
1789 1790 1791 1792 1793
	.reg_read32		= nvmf_reg_read32,
	.reg_read64		= nvmf_reg_read64,
	.reg_write32		= nvmf_reg_write32,
	.free_ctrl		= nvme_rdma_free_ctrl,
	.submit_async_event	= nvme_rdma_submit_async_event,
1794
	.delete_ctrl		= nvme_rdma_delete_ctrl,
1795 1796 1797
	.get_address		= nvmf_get_address,
};

1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874
static inline bool
__nvme_rdma_options_match(struct nvme_rdma_ctrl *ctrl,
	struct nvmf_ctrl_options *opts)
{
	char *stdport = __stringify(NVME_RDMA_IP_PORT);


	if (!nvmf_ctlr_matches_baseopts(&ctrl->ctrl, opts) ||
	    strcmp(opts->traddr, ctrl->ctrl.opts->traddr))
		return false;

	if (opts->mask & NVMF_OPT_TRSVCID &&
	    ctrl->ctrl.opts->mask & NVMF_OPT_TRSVCID) {
		if (strcmp(opts->trsvcid, ctrl->ctrl.opts->trsvcid))
			return false;
	} else if (opts->mask & NVMF_OPT_TRSVCID) {
		if (strcmp(opts->trsvcid, stdport))
			return false;
	} else if (ctrl->ctrl.opts->mask & NVMF_OPT_TRSVCID) {
		if (strcmp(stdport, ctrl->ctrl.opts->trsvcid))
			return false;
	}
	/* else, it's a match as both have stdport. Fall to next checks */

	/*
	 * checking the local address is rough. In most cases, one
	 * is not specified and the host port is selected by the stack.
	 *
	 * Assume no match if:
	 *  local address is specified and address is not the same
	 *  local address is not specified but remote is, or vice versa
	 *    (admin using specific host_traddr when it matters).
	 */
	if (opts->mask & NVMF_OPT_HOST_TRADDR &&
	    ctrl->ctrl.opts->mask & NVMF_OPT_HOST_TRADDR) {
		if (strcmp(opts->host_traddr, ctrl->ctrl.opts->host_traddr))
			return false;
	} else if (opts->mask & NVMF_OPT_HOST_TRADDR ||
		   ctrl->ctrl.opts->mask & NVMF_OPT_HOST_TRADDR)
		return false;
	/*
	 * if neither controller had an host port specified, assume it's
	 * a match as everything else matched.
	 */

	return true;
}

/*
 * Fails a connection request if it matches an existing controller
 * (association) with the same tuple:
 * <Host NQN, Host ID, local address, remote address, remote port, SUBSYS NQN>
 *
 * if local address is not specified in the request, it will match an
 * existing controller with all the other parameters the same and no
 * local port address specified as well.
 *
 * The ports don't need to be compared as they are intrinsically
 * already matched by the port pointers supplied.
 */
static bool
nvme_rdma_existing_controller(struct nvmf_ctrl_options *opts)
{
	struct nvme_rdma_ctrl *ctrl;
	bool found = false;

	mutex_lock(&nvme_rdma_ctrl_mutex);
	list_for_each_entry(ctrl, &nvme_rdma_ctrl_list, list) {
		found = __nvme_rdma_options_match(ctrl, opts);
		if (found)
			break;
	}
	mutex_unlock(&nvme_rdma_ctrl_mutex);

	return found;
}

1875 1876 1877 1878 1879 1880
static struct nvme_ctrl *nvme_rdma_create_ctrl(struct device *dev,
		struct nvmf_ctrl_options *opts)
{
	struct nvme_rdma_ctrl *ctrl;
	int ret;
	bool changed;
1881
	char *port;
1882 1883 1884 1885 1886 1887 1888

	ctrl = kzalloc(sizeof(*ctrl), GFP_KERNEL);
	if (!ctrl)
		return ERR_PTR(-ENOMEM);
	ctrl->ctrl.opts = opts;
	INIT_LIST_HEAD(&ctrl->list);

1889 1890 1891 1892 1893 1894 1895
	if (opts->mask & NVMF_OPT_TRSVCID)
		port = opts->trsvcid;
	else
		port = __stringify(NVME_RDMA_IP_PORT);

	ret = inet_pton_with_scope(&init_net, AF_UNSPEC,
			opts->traddr, port, &ctrl->addr);
1896
	if (ret) {
1897
		pr_err("malformed address passed: %s:%s\n", opts->traddr, port);
1898 1899 1900
		goto out_free_ctrl;
	}

1901
	if (opts->mask & NVMF_OPT_HOST_TRADDR) {
1902 1903
		ret = inet_pton_with_scope(&init_net, AF_UNSPEC,
			opts->host_traddr, NULL, &ctrl->src_addr);
1904
		if (ret) {
1905
			pr_err("malformed src address passed: %s\n",
1906 1907 1908 1909 1910
			       opts->host_traddr);
			goto out_free_ctrl;
		}
	}

1911 1912 1913 1914 1915
	if (!opts->duplicate_connect && nvme_rdma_existing_controller(opts)) {
		ret = -EALREADY;
		goto out_free_ctrl;
	}

1916 1917 1918 1919 1920 1921 1922 1923
	ret = nvme_init_ctrl(&ctrl->ctrl, dev, &nvme_rdma_ctrl_ops,
				0 /* no quirks, we're perfect! */);
	if (ret)
		goto out_free_ctrl;

	INIT_DELAYED_WORK(&ctrl->reconnect_work,
			nvme_rdma_reconnect_ctrl_work);
	INIT_WORK(&ctrl->err_work, nvme_rdma_error_recovery_work);
1924
	INIT_WORK(&ctrl->ctrl.reset_work, nvme_rdma_reset_ctrl_work);
1925

1926
	ctrl->ctrl.queue_count = opts->nr_io_queues + 1; /* +1 for admin queue */
1927
	ctrl->ctrl.sqsize = opts->queue_size - 1;
1928 1929 1930
	ctrl->ctrl.kato = opts->kato;

	ret = -ENOMEM;
1931
	ctrl->queues = kcalloc(ctrl->ctrl.queue_count, sizeof(*ctrl->queues),
1932 1933 1934 1935
				GFP_KERNEL);
	if (!ctrl->queues)
		goto out_uninit_ctrl;

1936
	ret = nvme_rdma_configure_admin_queue(ctrl, true);
1937 1938 1939 1940 1941 1942
	if (ret)
		goto out_kfree_queues;

	/* sanity check icdoff */
	if (ctrl->ctrl.icdoff) {
		dev_err(ctrl->ctrl.device, "icdoff is not supported!\n");
1943
		ret = -EINVAL;
1944 1945 1946 1947 1948 1949
		goto out_remove_admin_queue;
	}

	/* sanity check keyed sgls */
	if (!(ctrl->ctrl.sgls & (1 << 20))) {
		dev_err(ctrl->ctrl.device, "Mandatory keyed sgls are not support\n");
1950
		ret = -EINVAL;
1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961
		goto out_remove_admin_queue;
	}

	if (opts->queue_size > ctrl->ctrl.maxcmd) {
		/* warn if maxcmd is lower than queue_size */
		dev_warn(ctrl->ctrl.device,
			"queue_size %zu > ctrl maxcmd %u, clamping down\n",
			opts->queue_size, ctrl->ctrl.maxcmd);
		opts->queue_size = ctrl->ctrl.maxcmd;
	}

1962 1963 1964 1965 1966 1967 1968 1969
	if (opts->queue_size > ctrl->ctrl.sqsize + 1) {
		/* warn if sqsize is lower than queue_size */
		dev_warn(ctrl->ctrl.device,
			"queue_size %zu > ctrl sqsize %u, clamping down\n",
			opts->queue_size, ctrl->ctrl.sqsize + 1);
		opts->queue_size = ctrl->ctrl.sqsize + 1;
	}

1970
	if (opts->nr_io_queues) {
1971
		ret = nvme_rdma_configure_io_queues(ctrl, true);
1972 1973 1974 1975 1976 1977 1978
		if (ret)
			goto out_remove_admin_queue;
	}

	changed = nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_LIVE);
	WARN_ON_ONCE(!changed);

1979
	dev_info(ctrl->ctrl.device, "new ctrl: NQN \"%s\", addr %pISpcs\n",
1980 1981
		ctrl->ctrl.opts->subsysnqn, &ctrl->addr);

1982
	nvme_get_ctrl(&ctrl->ctrl);
1983 1984 1985 1986 1987

	mutex_lock(&nvme_rdma_ctrl_mutex);
	list_add_tail(&ctrl->list, &nvme_rdma_ctrl_list);
	mutex_unlock(&nvme_rdma_ctrl_mutex);

1988
	nvme_start_ctrl(&ctrl->ctrl);
1989 1990 1991 1992

	return &ctrl->ctrl;

out_remove_admin_queue:
1993
	nvme_rdma_destroy_admin_queue(ctrl, true);
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
out_kfree_queues:
	kfree(ctrl->queues);
out_uninit_ctrl:
	nvme_uninit_ctrl(&ctrl->ctrl);
	nvme_put_ctrl(&ctrl->ctrl);
	if (ret > 0)
		ret = -EIO;
	return ERR_PTR(ret);
out_free_ctrl:
	kfree(ctrl);
	return ERR_PTR(ret);
}

static struct nvmf_transport_ops nvme_rdma_transport = {
	.name		= "rdma",
2009
	.module		= THIS_MODULE,
2010
	.required_opts	= NVMF_OPT_TRADDR,
2011
	.allowed_opts	= NVMF_OPT_TRSVCID | NVMF_OPT_RECONNECT_DELAY |
S
Sagi Grimberg 已提交
2012
			  NVMF_OPT_HOST_TRADDR | NVMF_OPT_CTRL_LOSS_TMO,
2013 2014 2015
	.create_ctrl	= nvme_rdma_create_ctrl,
};

2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027
static void nvme_rdma_remove_one(struct ib_device *ib_device, void *client_data)
{
	struct nvme_rdma_ctrl *ctrl;

	/* Delete all controllers using this device */
	mutex_lock(&nvme_rdma_ctrl_mutex);
	list_for_each_entry(ctrl, &nvme_rdma_ctrl_list, list) {
		if (ctrl->device->dev != ib_device)
			continue;
		dev_info(ctrl->ctrl.device,
			"Removing ctrl: NQN \"%s\", addr %pISp\n",
			ctrl->ctrl.opts->subsysnqn, &ctrl->addr);
2028
		nvme_delete_ctrl(&ctrl->ctrl);
2029 2030 2031
	}
	mutex_unlock(&nvme_rdma_ctrl_mutex);

2032
	flush_workqueue(nvme_delete_wq);
2033 2034 2035 2036 2037 2038 2039
}

static struct ib_client nvme_rdma_ib_client = {
	.name   = "nvme_rdma",
	.remove = nvme_rdma_remove_one
};

2040 2041
static int __init nvme_rdma_init_module(void)
{
2042 2043 2044
	int ret;

	ret = ib_register_client(&nvme_rdma_ib_client);
2045
	if (ret)
2046
		return ret;
2047 2048 2049 2050

	ret = nvmf_register_transport(&nvme_rdma_transport);
	if (ret)
		goto err_unreg_client;
2051

2052
	return 0;
2053

2054 2055 2056
err_unreg_client:
	ib_unregister_client(&nvme_rdma_ib_client);
	return ret;
2057 2058 2059 2060 2061
}

static void __exit nvme_rdma_cleanup_module(void)
{
	nvmf_unregister_transport(&nvme_rdma_transport);
2062
	ib_unregister_client(&nvme_rdma_ib_client);
2063 2064 2065 2066 2067 2068
}

module_init(nvme_rdma_init_module);
module_exit(nvme_rdma_cleanup_module);

MODULE_LICENSE("GPL v2");