rdma.c 51.4 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 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
};

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;

116 117
	struct sockaddr_storage addr;
	struct sockaddr_storage src_addr;
118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146

	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);

147 148 149
static const struct blk_mq_ops nvme_rdma_mq_ops;
static const struct blk_mq_ops nvme_rdma_admin_mq_ops;

150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226
/* 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)
{
227 228 229
	pr_debug("QP event %s (%d)\n",
		 ib_event_msg(event->event), event->event);

230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263
}

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;
}

264 265
static void nvme_rdma_exit_request(struct blk_mq_tag_set *set,
		struct request *rq, unsigned int hctx_idx)
266
{
267
	struct nvme_rdma_ctrl *ctrl = set->driver_data;
268
	struct nvme_rdma_request *req = blk_mq_rq_to_pdu(rq);
269
	int queue_idx = (set == &ctrl->tag_set) ? hctx_idx + 1 : 0;
270 271 272 273 274 275 276
	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);
}

277 278 279
static int nvme_rdma_init_request(struct blk_mq_tag_set *set,
		struct request *rq, unsigned int hctx_idx,
		unsigned int numa_node)
280
{
281
	struct nvme_rdma_ctrl *ctrl = set->driver_data;
282
	struct nvme_rdma_request *req = blk_mq_rq_to_pdu(rq);
283
	int queue_idx = (set == &ctrl->tag_set) ? hctx_idx + 1 : 0;
284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304
	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];

305
	BUG_ON(hctx_idx >= ctrl->ctrl.queue_count);
306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364

	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);

365 366
	ndev->pd = ib_alloc_pd(ndev->dev,
		register_always ? 0 : IB_PD_UNSAFE_GLOBAL_RKEY);
367 368 369 370 371 372 373
	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");
374
		goto out_free_pd;
375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392
	}

	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)
{
393 394
	struct nvme_rdma_device *dev = queue->device;
	struct ib_device *ibdev = dev->dev;
395

I
Israel Rukshin 已提交
396 397
	ib_mr_pool_destroy(queue->qp, &queue->qp->rdma_mrs);

398 399 400 401 402 403 404 405 406
	rdma_destroy_qp(queue->cm_id);
	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
407 408 409 410 411 412
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);
}

413
static int nvme_rdma_create_queue_ib(struct nvme_rdma_queue *queue)
414
{
415
	struct ib_device *ibdev;
416 417 418 419 420
	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;

421 422 423 424 425 426 427
	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;
428 429

	/*
430 431
	 * Spread I/O queues completion vectors according their queue index.
	 * Admin queues can always go on completion vector 0.
432
	 */
433
	comp_vector = idx == 0 ? idx : idx - 1;
434 435

	/* +1 for ib_stop_cq */
436 437 438
	queue->ib_cq = ib_alloc_cq(ibdev, queue,
				cq_factor * queue->queue_size + 1,
				comp_vector, IB_POLL_SOFTIRQ);
439 440
	if (IS_ERR(queue->ib_cq)) {
		ret = PTR_ERR(queue->ib_cq);
441
		goto out_put_dev;
442 443 444 445 446 447 448 449 450 451 452 453 454
	}

	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
455 456 457 458 459 460 461 462 463 464 465
	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;
	}

466 467
	return 0;

I
Israel Rukshin 已提交
468 469 470
out_destroy_ring:
	nvme_rdma_free_ring(ibdev, queue->rsp_ring, queue->queue_size,
			    sizeof(struct nvme_completion), DMA_FROM_DEVICE);
471
out_destroy_qp:
472
	rdma_destroy_qp(queue->cm_id);
473 474
out_destroy_ib_cq:
	ib_free_cq(queue->ib_cq);
475 476
out_put_dev:
	nvme_rdma_dev_put(queue->device);
477 478 479
	return ret;
}

480
static int nvme_rdma_alloc_queue(struct nvme_rdma_ctrl *ctrl,
481 482 483
		int idx, size_t queue_size)
{
	struct nvme_rdma_queue *queue;
484
	struct sockaddr *src_addr = NULL;
485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505
	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);
	}

506
	if (ctrl->ctrl.opts->mask & NVMF_OPT_HOST_TRADDR)
507
		src_addr = (struct sockaddr *)&ctrl->src_addr;
508

509 510 511
	queue->cm_error = -ETIMEDOUT;
	ret = rdma_resolve_addr(queue->cm_id, src_addr,
			(struct sockaddr *)&ctrl->addr,
512 513 514 515 516 517 518 519 520 521
			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,
522
			"rdma connection establishment failed (%d)\n", ret);
523 524 525
		goto out_destroy_cm_id;
	}

526
	set_bit(NVME_RDMA_Q_ALLOCATED, &queue->flags);
527 528 529 530 531 532 533 534 535 536

	return 0;

out_destroy_cm_id:
	rdma_destroy_id(queue->cm_id);
	return ret;
}

static void nvme_rdma_stop_queue(struct nvme_rdma_queue *queue)
{
537 538 539
	if (!test_and_clear_bit(NVME_RDMA_Q_LIVE, &queue->flags))
		return;

540 541 542 543 544 545
	rdma_disconnect(queue->cm_id);
	ib_drain_qp(queue->qp);
}

static void nvme_rdma_free_queue(struct nvme_rdma_queue *queue)
{
546
	if (!test_and_clear_bit(NVME_RDMA_Q_ALLOCATED, &queue->flags))
547 548
		return;

549 550 551 552 553 554
	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);
	}

555 556 557 558
	nvme_rdma_destroy_queue_ib(queue);
	rdma_destroy_id(queue->cm_id);
}

559
static void nvme_rdma_free_io_queues(struct nvme_rdma_ctrl *ctrl)
560
{
561 562 563 564
	int i;

	for (i = 1; i < ctrl->ctrl.queue_count; i++)
		nvme_rdma_free_queue(&ctrl->queues[i]);
565 566
}

567
static void nvme_rdma_stop_io_queues(struct nvme_rdma_ctrl *ctrl)
568 569 570
{
	int i;

571
	for (i = 1; i < ctrl->ctrl.queue_count; i++)
572
		nvme_rdma_stop_queue(&ctrl->queues[i]);
573 574
}

575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592
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)
593 594 595
{
	int i, ret = 0;

596
	for (i = 1; i < ctrl->ctrl.queue_count; i++) {
597 598
		ret = nvme_rdma_start_queue(ctrl, i);
		if (ret)
599
			goto out_stop_queues;
600 601
	}

602 603
	return 0;

604
out_stop_queues:
605 606
	for (i--; i >= 1; i--)
		nvme_rdma_stop_queue(&ctrl->queues[i]);
607 608 609
	return ret;
}

610
static int nvme_rdma_alloc_io_queues(struct nvme_rdma_ctrl *ctrl)
611
{
612
	struct nvmf_ctrl_options *opts = ctrl->ctrl.opts;
613
	struct ib_device *ibdev = ctrl->device->dev;
614
	unsigned int nr_io_queues;
615 616
	int i, ret;

617
	nr_io_queues = min(opts->nr_io_queues, num_online_cpus());
618 619 620 621 622 623 624 625 626

	/*
	 * 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);

627 628 629 630
	ret = nvme_set_queue_count(&ctrl->ctrl, &nr_io_queues);
	if (ret)
		return ret;

631 632
	ctrl->ctrl.queue_count = nr_io_queues + 1;
	if (ctrl->ctrl.queue_count < 2)
633 634 635 636 637
		return 0;

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

638
	for (i = 1; i < ctrl->ctrl.queue_count; i++) {
639 640 641
		ret = nvme_rdma_alloc_queue(ctrl, i,
				ctrl->ctrl.sqsize + 1);
		if (ret)
642 643 644 645 646 647
			goto out_free_queues;
	}

	return 0;

out_free_queues:
648
	for (i--; i >= 1; i--)
649
		nvme_rdma_free_queue(&ctrl->queues[i]);
650 651 652 653

	return ret;
}

654 655
static void nvme_rdma_free_tagset(struct nvme_ctrl *nctrl,
		struct blk_mq_tag_set *set)
656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673
{
	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
674
		set->queue_depth = NVME_AQ_MQ_TAG_DEPTH;
675 676 677 678 679 680 681
		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;
682
		set->flags = BLK_MQ_F_NO_SCHED;
683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719
	} 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);
}

720 721
static void nvme_rdma_destroy_admin_queue(struct nvme_rdma_ctrl *ctrl,
		bool remove)
722
{
723
	nvme_rdma_stop_queue(&ctrl->queues[0]);
724 725
	if (remove) {
		blk_cleanup_queue(ctrl->ctrl.admin_q);
726
		nvme_rdma_free_tagset(&ctrl->ctrl, ctrl->ctrl.admin_tagset);
727
	}
728
	nvme_rdma_free_queue(&ctrl->queues[0]);
729 730
}

731 732
static int nvme_rdma_configure_admin_queue(struct nvme_rdma_ctrl *ctrl,
		bool new)
733 734 735
{
	int error;

736
	error = nvme_rdma_alloc_queue(ctrl, 0, NVME_AQ_DEPTH);
737 738 739 740 741
	if (error)
		return error;

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

I
Israel Rukshin 已提交
742
	ctrl->max_fr_pages = nvme_rdma_get_max_fr_pages(ctrl->device->dev);
743

744 745
	if (new) {
		ctrl->ctrl.admin_tagset = nvme_rdma_alloc_tagset(&ctrl->ctrl, true);
746 747
		if (IS_ERR(ctrl->ctrl.admin_tagset)) {
			error = PTR_ERR(ctrl->ctrl.admin_tagset);
748
			goto out_free_queue;
749
		}
750

751 752 753 754 755
		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;
		}
756 757
	}

758
	error = nvme_rdma_start_queue(ctrl, 0);
759 760 761
	if (error)
		goto out_cleanup_queue;

762
	error = ctrl->ctrl.ops->reg_read64(&ctrl->ctrl, NVME_REG_CAP,
763 764 765 766 767 768 769 770 771 772 773 774 775 776 777
			&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 =
778
		(ctrl->max_fr_pages - 1) << (ilog2(SZ_4K) - 9);
779 780 781 782 783 784 785 786 787 788 789 790 791 792

	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:
793 794
	if (new)
		blk_cleanup_queue(ctrl->ctrl.admin_q);
795
out_free_tagset:
796
	if (new)
797
		nvme_rdma_free_tagset(&ctrl->ctrl, ctrl->ctrl.admin_tagset);
798 799 800 801 802
out_free_queue:
	nvme_rdma_free_queue(&ctrl->queues[0]);
	return error;
}

803 804 805 806 807 808
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);
809
		nvme_rdma_free_tagset(&ctrl->ctrl, ctrl->ctrl.tagset);
810 811 812 813 814 815 816 817
	}
	nvme_rdma_free_io_queues(ctrl);
}

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

818
	ret = nvme_rdma_alloc_io_queues(ctrl);
819 820 821 822 823
	if (ret)
		return ret;

	if (new) {
		ctrl->ctrl.tagset = nvme_rdma_alloc_tagset(&ctrl->ctrl, false);
824 825
		if (IS_ERR(ctrl->ctrl.tagset)) {
			ret = PTR_ERR(ctrl->ctrl.tagset);
826
			goto out_free_io_queues;
827
		}
828 829 830 831 832 833 834 835 836 837 838

		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);
	}

839
	ret = nvme_rdma_start_io_queues(ctrl);
840 841 842 843 844 845 846 847 848 849
	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)
850
		nvme_rdma_free_tagset(&ctrl->ctrl, ctrl->ctrl.tagset);
851 852 853
out_free_io_queues:
	nvme_rdma_free_io_queues(ctrl);
	return ret;
854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872
}

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
873 874 875 876 877 878 879 880 881 882 883 884
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);
885
		queue_delayed_work(nvme_wq, &ctrl->reconnect_work,
S
Sagi Grimberg 已提交
886 887 888
				ctrl->ctrl.opts->reconnect_delay * HZ);
	} else {
		dev_info(ctrl->ctrl.device, "Removing controller...\n");
889
		nvme_delete_ctrl(&ctrl->ctrl);
S
Sagi Grimberg 已提交
890 891 892
	}
}

893 894 895 896 897 898 899
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;

900
	++ctrl->ctrl.nr_reconnects;
S
Sagi Grimberg 已提交
901

902
	ret = nvme_rdma_configure_admin_queue(ctrl, false);
903
	if (ret)
904
		goto requeue;
905

906
	if (ctrl->ctrl.queue_count > 1) {
907
		ret = nvme_rdma_configure_io_queues(ctrl, false);
908
		if (ret)
909
			goto destroy_admin;
910 911 912
	}

	changed = nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_LIVE);
913 914 915 916 917 918
	if (!changed) {
		/* state change failure is ok if we're in DELETING state */
		WARN_ON_ONCE(ctrl->ctrl.state != NVME_CTRL_DELETING);
		return;
	}

919
	nvme_start_ctrl(&ctrl->ctrl);
920

921 922 923 924
	dev_info(ctrl->ctrl.device, "Successfully reconnected (%d attempts)\n",
			ctrl->ctrl.nr_reconnects);

	ctrl->ctrl.nr_reconnects = 0;
925 926 927

	return;

928 929
destroy_admin:
	nvme_rdma_destroy_admin_queue(ctrl, false);
930
requeue:
S
Sagi Grimberg 已提交
931
	dev_info(ctrl->ctrl.device, "Failed reconnect attempt %d\n",
932
			ctrl->ctrl.nr_reconnects);
S
Sagi Grimberg 已提交
933
	nvme_rdma_reconnect_or_remove(ctrl);
934 935 936 937 938 939 940
}

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);

941
	nvme_stop_keep_alive(&ctrl->ctrl);
942

943
	if (ctrl->ctrl.queue_count > 1) {
944 945 946
		nvme_stop_queues(&ctrl->ctrl);
		blk_mq_tagset_busy_iter(&ctrl->tag_set,
					nvme_cancel_request, &ctrl->ctrl);
947 948 949 950
		nvme_rdma_destroy_io_queues(ctrl, false);
	}

	blk_mq_quiesce_queue(ctrl->ctrl.admin_q);
951 952
	blk_mq_tagset_busy_iter(&ctrl->admin_tag_set,
				nvme_cancel_request, &ctrl->ctrl);
953
	nvme_rdma_destroy_admin_queue(ctrl, false);
954

955 956 957 958
	/*
	 * queues are not a live anymore, so restart the queues to fail fast
	 * new IO
	 */
959
	blk_mq_unquiesce_queue(ctrl->ctrl.admin_q);
960 961
	nvme_start_queues(&ctrl->ctrl);

S
Sagi Grimberg 已提交
962
	nvme_rdma_reconnect_or_remove(ctrl);
963 964 965 966 967 968 969
}

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

970
	queue_work(nvme_wq, &ctrl->err_work);
971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994
}

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)
{
995 996 997 998 999
	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)) {
1000
		nvme_rdma_wr_error(cq, wc, "LOCAL_INV");
1001 1002 1003 1004 1005 1006
		return;
	}

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

1007 1008 1009 1010 1011 1012 1013 1014 1015 1016
}

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,
1017
		.send_flags	    = IB_SEND_SIGNALED,
1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036
		.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
1037 1038 1039 1040 1041
	if (req->mr) {
		ib_mr_pool_put(queue->qp, &queue->qp->rdma_mrs, req->mr);
		req->mr = NULL;
	}

1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 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
	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);
1086
	put_unaligned_le32(queue->device->pd->unsafe_global_rkey, sg->key);
1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097
	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
1098 1099 1100 1101
	req->mr = ib_mr_pool_get(queue->qp, &queue->qp->rdma_mrs);
	if (WARN_ON_ONCE(!req->mr))
		return -EAGAIN;

1102 1103 1104 1105 1106
	/*
	 * 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);
1107
	if (unlikely(nr < count)) {
I
Israel Rukshin 已提交
1108 1109
		ib_mr_pool_put(queue->qp, &queue->qp->rdma_mrs, req->mr);
		req->mr = NULL;
1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137
		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,
1138
		struct request *rq, struct nvme_command *c)
1139 1140 1141 1142
{
	struct nvme_rdma_request *req = blk_mq_rq_to_pdu(rq);
	struct nvme_rdma_device *dev = queue->device;
	struct ib_device *ibdev = dev->dev;
1143
	int count, ret;
1144 1145 1146

	req->num_sge = 1;
	req->inline_data = false;
1147
	refcount_set(&req->ref, 2); /* send and recv completions */
1148 1149 1150 1151 1152 1153 1154

	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;
1155 1156
	ret = sg_alloc_table_chained(&req->sg_table,
			blk_rq_nr_phys_segments(rq), req->sg_table.sgl);
1157 1158 1159
	if (ret)
		return -ENOMEM;

1160
	req->nents = blk_rq_map_sg(rq->q, rq, req->sg_table.sgl);
1161

1162
	count = ib_dma_map_sg(ibdev, req->sg_table.sgl, req->nents,
1163 1164 1165 1166 1167 1168 1169
		    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) {
1170 1171 1172
		if (rq_data_dir(rq) == WRITE && nvme_rdma_queue_idx(queue) &&
		    blk_rq_payload_bytes(rq) <=
				nvme_rdma_inline_data_size(queue))
1173 1174
			return nvme_rdma_map_sg_inline(queue, req, c);

1175
		if (dev->pd->flags & IB_PD_UNSAFE_GLOBAL_RKEY)
1176 1177 1178 1179 1180 1181 1182 1183
			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)
{
1184 1185 1186 1187 1188 1189 1190
	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)) {
1191
		nvme_rdma_wr_error(cq, wc, "SEND");
1192 1193 1194 1195 1196
		return;
	}

	if (refcount_dec_and_test(&req->ref))
		nvme_end_request(rq, req->status, req->result);
1197 1198 1199 1200
}

static int nvme_rdma_post_send(struct nvme_rdma_queue *queue,
		struct nvme_rdma_qe *qe, struct ib_sge *sge, u32 num_sge,
1201
		struct ib_send_wr *first)
1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214
{
	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;
1215
	wr.send_flags = IB_SEND_SIGNALED;
1216 1217 1218 1219 1220 1221 1222

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

	ret = ib_post_send(queue->qp, first, &bad_wr);
1223
	if (unlikely(ret)) {
1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248
		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);
1249
	if (unlikely(ret)) {
1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264
		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];
}

1265 1266 1267 1268 1269 1270
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");
}

1271
static void nvme_rdma_submit_async_event(struct nvme_ctrl *arg)
1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284
{
	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
1285
	cmd->common.command_id = NVME_AQ_BLK_MQ_DEPTH;
1286 1287 1288
	cmd->common.flags |= NVME_CMD_SGL_METABUF;
	nvme_rdma_set_sg_null(cmd);

1289 1290
	sqe->cqe.done = nvme_rdma_async_done;

1291 1292 1293
	ib_dma_sync_single_for_device(dev, sqe->dma, sizeof(*cmd),
			DMA_TO_DEVICE);

1294
	ret = nvme_rdma_post_send(queue, sqe, &sge, 1, NULL);
1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314
	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);

1315 1316
	req->status = cqe->status;
	req->result = cqe->result;
1317

1318 1319 1320 1321 1322 1323 1324
	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
1325
	} else if (req->mr) {
1326 1327 1328 1329 1330 1331 1332 1333 1334 1335
		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;
	}
1336

1337 1338 1339 1340 1341 1342
	if (refcount_dec_and_test(&req->ref)) {
		if (rq->tag == tag)
			ret = 1;
		nvme_end_request(rq, req->status, req->result);
	}

1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368
	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
1369
			cqe->command_id >= NVME_AQ_BLK_MQ_DEPTH))
1370 1371
		nvme_complete_async_event(&queue->ctrl->ctrl, cqe->status,
				&cqe->result);
1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404
	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)
{
1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415
	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);
1416 1417

		dev_err(queue->ctrl->ctrl.device,
1418 1419
		      "Connect rejected: status %d (%s) nvme status %d (%s).\n",
		      status, rej_msg, sts, nvme_rdma_cm_msg(sts));
1420 1421
	} else {
		dev_err(queue->ctrl->ctrl.device,
1422
			"Connect rejected: status %d (%s).\n", status, rej_msg);
1423 1424 1425 1426 1427 1428 1429 1430 1431
	}

	return -ECONNRESET;
}

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

1432 1433 1434
	ret = nvme_rdma_create_queue_ib(queue);
	if (ret)
		return ret;
1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454

	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 = { };
1455
	struct nvme_rdma_cm_req priv = { };
1456 1457 1458 1459 1460 1461
	int ret;

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

	param.responder_resources = queue->device->dev->attrs.max_qp_rd_atom;
1462 1463
	/* maximum retry count */
	param.retry_count = 7;
1464 1465 1466 1467 1468 1469
	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));
1470 1471 1472 1473 1474
	/*
	 * set the admin queue depth to the minimum size
	 * specified by the Fabrics standard.
	 */
	if (priv.qid == 0) {
1475 1476
		priv.hrqsize = cpu_to_le16(NVME_AQ_DEPTH);
		priv.hsqsize = cpu_to_le16(NVME_AQ_DEPTH - 1);
1477
	} else {
1478 1479 1480 1481 1482
		/*
		 * current interpretation of the fabrics spec
		 * is at minimum you make hrqsize sqsize+1, or a
		 * 1's based representation of sqsize.
		 */
1483
		priv.hrqsize = cpu_to_le16(queue->queue_size);
1484
		priv.hsqsize = cpu_to_le16(queue->ctrl->ctrl.sqsize);
1485
	}
1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523

	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:
1524
		nvme_rdma_destroy_queue_ib(queue);
1525 1526 1527 1528 1529
		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:
1530 1531
		nvme_rdma_destroy_queue_ib(queue);
	case RDMA_CM_EVENT_ADDR_ERROR:
1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543
		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:
1544 1545
		/* device removal is handled via the ib_client API */
		break;
1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565
	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);

1566 1567 1568 1569
	dev_warn(req->queue->ctrl->ctrl.device,
		 "I/O %d QID %d timeout, reset controller\n",
		 rq->tag, nvme_rdma_queue_idx(req->queue));

1570 1571 1572 1573
	/* queue error recovery */
	nvme_rdma_error_recovery(req->queue->ctrl);

	/* fail with DNR on cmd timeout */
1574
	nvme_req(rq)->status = NVME_SC_ABORT_REQ | NVME_SC_DNR;
1575 1576 1577 1578

	return BLK_EH_HANDLED;
}

1579 1580 1581
/*
 * We cannot accept any other command until the Connect command has completed.
 */
C
Christoph Hellwig 已提交
1582
static inline blk_status_t
1583 1584 1585 1586 1587
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;
1588 1589
}

1590
static blk_status_t nvme_rdma_queue_rq(struct blk_mq_hw_ctx *hctx,
1591 1592 1593 1594 1595 1596 1597 1598 1599
		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;
1600 1601
	blk_status_t ret;
	int err;
1602 1603 1604

	WARN_ON_ONCE(rq->tag < 0);

1605
	ret = nvme_rdma_is_ready(queue, rq);
1606
	if (unlikely(ret))
C
Christoph Hellwig 已提交
1607
		return ret;
1608

1609 1610 1611 1612 1613
	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);
1614
	if (ret)
1615 1616 1617 1618
		return ret;

	blk_mq_start_request(rq);

1619
	err = nvme_rdma_map_data(queue, rq, c);
1620
	if (unlikely(err < 0)) {
1621
		dev_err(queue->ctrl->ctrl.device,
1622
			     "Failed to map data (%d)\n", err);
1623 1624 1625 1626
		nvme_cleanup_cmd(rq);
		goto err;
	}

1627 1628
	sqe->cqe.done = nvme_rdma_send_done;

1629 1630 1631
	ib_dma_sync_single_for_device(dev, sqe->dma,
			sizeof(struct nvme_command), DMA_TO_DEVICE);

1632
	err = nvme_rdma_post_send(queue, sqe, req->sge, req->num_sge,
I
Israel Rukshin 已提交
1633
			req->mr ? &req->reg_wr.wr : NULL);
1634
	if (unlikely(err)) {
1635 1636 1637 1638
		nvme_rdma_unmap_data(queue, rq);
		goto err;
	}

1639
	return BLK_STS_OK;
1640
err:
1641 1642 1643
	if (err == -ENOMEM || err == -EAGAIN)
		return BLK_STS_RESOURCE;
	return BLK_STS_IOERR;
1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670
}

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);

1671 1672
	nvme_rdma_unmap_data(req->queue, rq);
	nvme_complete_rq(rq);
1673 1674
}

1675 1676 1677 1678 1679 1680 1681
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);
}

1682
static const struct blk_mq_ops nvme_rdma_mq_ops = {
1683 1684 1685 1686 1687 1688 1689
	.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,
1690
	.map_queues	= nvme_rdma_map_queues,
1691 1692
};

1693
static const struct blk_mq_ops nvme_rdma_admin_mq_ops = {
1694 1695
	.queue_rq	= nvme_rdma_queue_rq,
	.complete	= nvme_rdma_complete_rq,
1696 1697
	.init_request	= nvme_rdma_init_request,
	.exit_request	= nvme_rdma_exit_request,
1698 1699 1700 1701
	.init_hctx	= nvme_rdma_init_admin_hctx,
	.timeout	= nvme_rdma_timeout,
};

1702
static void nvme_rdma_shutdown_ctrl(struct nvme_rdma_ctrl *ctrl, bool shutdown)
1703 1704 1705 1706
{
	cancel_work_sync(&ctrl->err_work);
	cancel_delayed_work_sync(&ctrl->reconnect_work);

1707
	if (ctrl->ctrl.queue_count > 1) {
1708 1709 1710
		nvme_stop_queues(&ctrl->ctrl);
		blk_mq_tagset_busy_iter(&ctrl->tag_set,
					nvme_cancel_request, &ctrl->ctrl);
1711
		nvme_rdma_destroy_io_queues(ctrl, shutdown);
1712 1713
	}

1714
	if (shutdown)
1715
		nvme_shutdown_ctrl(&ctrl->ctrl);
1716 1717
	else
		nvme_disable_ctrl(&ctrl->ctrl, ctrl->ctrl.cap);
1718

1719
	blk_mq_quiesce_queue(ctrl->ctrl.admin_q);
1720 1721
	blk_mq_tagset_busy_iter(&ctrl->admin_tag_set,
				nvme_cancel_request, &ctrl->ctrl);
1722
	blk_mq_unquiesce_queue(ctrl->ctrl.admin_q);
1723
	nvme_rdma_destroy_admin_queue(ctrl, shutdown);
1724 1725
}

1726
static void nvme_rdma_delete_ctrl(struct nvme_ctrl *ctrl)
1727
{
1728
	nvme_rdma_shutdown_ctrl(to_rdma_ctrl(ctrl), true);
1729 1730 1731 1732
}

static void nvme_rdma_reset_ctrl_work(struct work_struct *work)
{
1733 1734
	struct nvme_rdma_ctrl *ctrl =
		container_of(work, struct nvme_rdma_ctrl, ctrl.reset_work);
1735 1736 1737
	int ret;
	bool changed;

1738
	nvme_stop_ctrl(&ctrl->ctrl);
1739
	nvme_rdma_shutdown_ctrl(ctrl, false);
1740

1741
	ret = nvme_rdma_configure_admin_queue(ctrl, false);
1742 1743
	if (ret)
		goto out_fail;
1744

1745
	if (ctrl->ctrl.queue_count > 1) {
1746
		ret = nvme_rdma_configure_io_queues(ctrl, false);
1747
		if (ret)
1748
			goto out_fail;
1749 1750 1751
	}

	changed = nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_LIVE);
1752 1753 1754 1755 1756
	if (!changed) {
		/* state change failure is ok if we're in DELETING state */
		WARN_ON_ONCE(ctrl->ctrl.state != NVME_CTRL_DELETING);
		return;
	}
1757

1758
	nvme_start_ctrl(&ctrl->ctrl);
1759 1760 1761

	return;

1762
out_fail:
1763
	dev_warn(ctrl->ctrl.device, "Removing after reset failure\n");
1764 1765 1766 1767
	nvme_remove_namespaces(&ctrl->ctrl);
	nvme_rdma_shutdown_ctrl(ctrl, true);
	nvme_uninit_ctrl(&ctrl->ctrl);
	nvme_put_ctrl(&ctrl->ctrl);
1768 1769 1770 1771 1772
}

static const struct nvme_ctrl_ops nvme_rdma_ctrl_ops = {
	.name			= "rdma",
	.module			= THIS_MODULE,
1773
	.flags			= NVME_F_FABRICS,
1774 1775 1776 1777 1778
	.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,
1779
	.delete_ctrl		= nvme_rdma_delete_ctrl,
1780 1781 1782
	.get_address		= nvmf_get_address,
};

1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 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
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;
}

1860 1861 1862 1863 1864 1865
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;
1866
	char *port;
1867 1868 1869 1870 1871 1872 1873

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

1874 1875 1876 1877 1878 1879 1880
	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);
1881
	if (ret) {
1882
		pr_err("malformed address passed: %s:%s\n", opts->traddr, port);
1883 1884 1885
		goto out_free_ctrl;
	}

1886
	if (opts->mask & NVMF_OPT_HOST_TRADDR) {
1887 1888
		ret = inet_pton_with_scope(&init_net, AF_UNSPEC,
			opts->host_traddr, NULL, &ctrl->src_addr);
1889
		if (ret) {
1890
			pr_err("malformed src address passed: %s\n",
1891 1892 1893 1894 1895
			       opts->host_traddr);
			goto out_free_ctrl;
		}
	}

1896 1897 1898 1899 1900
	if (!opts->duplicate_connect && nvme_rdma_existing_controller(opts)) {
		ret = -EALREADY;
		goto out_free_ctrl;
	}

1901 1902 1903 1904 1905 1906 1907 1908
	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);
1909
	INIT_WORK(&ctrl->ctrl.reset_work, nvme_rdma_reset_ctrl_work);
1910

1911
	ctrl->ctrl.queue_count = opts->nr_io_queues + 1; /* +1 for admin queue */
1912
	ctrl->ctrl.sqsize = opts->queue_size - 1;
1913 1914 1915
	ctrl->ctrl.kato = opts->kato;

	ret = -ENOMEM;
1916
	ctrl->queues = kcalloc(ctrl->ctrl.queue_count, sizeof(*ctrl->queues),
1917 1918 1919 1920
				GFP_KERNEL);
	if (!ctrl->queues)
		goto out_uninit_ctrl;

1921
	ret = nvme_rdma_configure_admin_queue(ctrl, true);
1922 1923 1924 1925 1926 1927
	if (ret)
		goto out_kfree_queues;

	/* sanity check icdoff */
	if (ctrl->ctrl.icdoff) {
		dev_err(ctrl->ctrl.device, "icdoff is not supported!\n");
1928
		ret = -EINVAL;
1929 1930 1931 1932 1933 1934
		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");
1935
		ret = -EINVAL;
1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946
		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;
	}

1947 1948 1949 1950 1951 1952 1953 1954
	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;
	}

1955
	if (opts->nr_io_queues) {
1956
		ret = nvme_rdma_configure_io_queues(ctrl, true);
1957 1958 1959 1960 1961 1962 1963
		if (ret)
			goto out_remove_admin_queue;
	}

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

1964
	dev_info(ctrl->ctrl.device, "new ctrl: NQN \"%s\", addr %pISpcs\n",
1965 1966
		ctrl->ctrl.opts->subsysnqn, &ctrl->addr);

1967
	nvme_get_ctrl(&ctrl->ctrl);
1968 1969 1970 1971 1972

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

1973
	nvme_start_ctrl(&ctrl->ctrl);
1974 1975 1976 1977

	return &ctrl->ctrl;

out_remove_admin_queue:
1978
	nvme_rdma_destroy_admin_queue(ctrl, true);
1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994
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",
	.required_opts	= NVMF_OPT_TRADDR,
1995
	.allowed_opts	= NVMF_OPT_TRSVCID | NVMF_OPT_RECONNECT_DELAY |
S
Sagi Grimberg 已提交
1996
			  NVMF_OPT_HOST_TRADDR | NVMF_OPT_CTRL_LOSS_TMO,
1997 1998 1999
	.create_ctrl	= nvme_rdma_create_ctrl,
};

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
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);
2012
		nvme_delete_ctrl(&ctrl->ctrl);
2013 2014 2015
	}
	mutex_unlock(&nvme_rdma_ctrl_mutex);

2016
	flush_workqueue(nvme_wq);
2017 2018 2019 2020 2021 2022 2023
}

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

2024 2025
static int __init nvme_rdma_init_module(void)
{
2026 2027 2028
	int ret;

	ret = ib_register_client(&nvme_rdma_ib_client);
2029
	if (ret)
2030
		return ret;
2031 2032 2033 2034

	ret = nvmf_register_transport(&nvme_rdma_transport);
	if (ret)
		goto err_unreg_client;
2035

2036
	return 0;
2037

2038 2039 2040
err_unreg_client:
	ib_unregister_client(&nvme_rdma_ib_client);
	return ret;
2041 2042 2043 2044 2045
}

static void __exit nvme_rdma_cleanup_module(void)
{
	nvmf_unregister_transport(&nvme_rdma_transport);
2046
	ib_unregister_client(&nvme_rdma_ib_client);
2047 2048 2049 2050 2051 2052
}

module_init(nvme_rdma_init_module);
module_exit(nvme_rdma_cleanup_module);

MODULE_LICENSE("GPL v2");