summaryrefslogtreecommitdiff
path: root/drivers/nvme/nvme.c
blob: f6465ea7f482a7b7ffe933152cc1374514c54711 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
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
116
117
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
147
148
149
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
227
228
229
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
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
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
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
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
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
// SPDX-License-Identifier: GPL-2.0+
/*
 * Copyright (C) 2017 NXP Semiconductors
 * Copyright (C) 2017 Bin Meng <bmeng.cn@gmail.com>
 */

#include <common.h>
#include <blk.h>
#include <cpu_func.h>
#include <dm.h>
#include <errno.h>
#include <log.h>
#include <malloc.h>
#include <memalign.h>
#include <pci.h>
#include <time.h>
#include <dm/device-internal.h>
#include <linux/compat.h>
#include "nvme.h"

#define NVME_Q_DEPTH		2
#define NVME_AQ_DEPTH		2
#define NVME_SQ_SIZE(depth)	(depth * sizeof(struct nvme_command))
#define NVME_CQ_SIZE(depth)	(depth * sizeof(struct nvme_completion))
#define NVME_CQ_ALLOCATION	ALIGN(NVME_CQ_SIZE(NVME_Q_DEPTH), \
				      ARCH_DMA_MINALIGN)
#define ADMIN_TIMEOUT		60
#define IO_TIMEOUT		30
#define MAX_PRP_POOL		512

enum nvme_queue_id {
	NVME_ADMIN_Q,
	NVME_IO_Q,
	NVME_Q_NUM,
};

/*
 * An NVM Express queue. Each device has at least two (one for admin
 * commands and one for I/O commands).
 */
struct nvme_queue {
	struct nvme_dev *dev;
	struct nvme_command *sq_cmds;
	struct nvme_completion *cqes;
	wait_queue_head_t sq_full;
	u32 __iomem *q_db;
	u16 q_depth;
	s16 cq_vector;
	u16 sq_head;
	u16 sq_tail;
	u16 cq_head;
	u16 qid;
	u8 cq_phase;
	u8 cqe_seen;
	unsigned long cmdid_data[];
};

static int nvme_wait_ready(struct nvme_dev *dev, bool enabled)
{
	u32 bit = enabled ? NVME_CSTS_RDY : 0;
	int timeout;
	ulong start;

	/* Timeout field in the CAP register is in 500 millisecond units */
	timeout = NVME_CAP_TIMEOUT(dev->cap) * 500;

	start = get_timer(0);
	while (get_timer(start) < timeout) {
		if ((readl(&dev->bar->csts) & NVME_CSTS_RDY) == bit)
			return 0;
	}

	return -ETIME;
}

static int nvme_setup_prps(struct nvme_dev *dev, u64 *prp2,
			   int total_len, u64 dma_addr)
{
	u32 page_size = dev->page_size;
	int offset = dma_addr & (page_size - 1);
	u64 *prp_pool;
	int length = total_len;
	int i, nprps;
	u32 prps_per_page = page_size >> 3;
	u32 num_pages;

	length -= (page_size - offset);

	if (length <= 0) {
		*prp2 = 0;
		return 0;
	}

	if (length)
		dma_addr += (page_size - offset);

	if (length <= page_size) {
		*prp2 = dma_addr;
		return 0;
	}

	nprps = DIV_ROUND_UP(length, page_size);
	num_pages = DIV_ROUND_UP(nprps, prps_per_page);

	if (nprps > dev->prp_entry_num) {
		free(dev->prp_pool);
		/*
		 * Always increase in increments of pages.  It doesn't waste
		 * much memory and reduces the number of allocations.
		 */
		dev->prp_pool = memalign(page_size, num_pages * page_size);
		if (!dev->prp_pool) {
			printf("Error: malloc prp_pool fail\n");
			return -ENOMEM;
		}
		dev->prp_entry_num = prps_per_page * num_pages;
	}

	prp_pool = dev->prp_pool;
	i = 0;
	while (nprps) {
		if (i == ((page_size >> 3) - 1)) {
			*(prp_pool + i) = cpu_to_le64((ulong)prp_pool +
					page_size);
			i = 0;
			prp_pool += page_size;
		}
		*(prp_pool + i++) = cpu_to_le64(dma_addr);
		dma_addr += page_size;
		nprps--;
	}
	*prp2 = (ulong)dev->prp_pool;

	flush_dcache_range((ulong)dev->prp_pool, (ulong)dev->prp_pool +
			   dev->prp_entry_num * sizeof(u64));

	return 0;
}

static __le16 nvme_get_cmd_id(void)
{
	static unsigned short cmdid;

	return cpu_to_le16((cmdid < USHRT_MAX) ? cmdid++ : 0);
}

static u16 nvme_read_completion_status(struct nvme_queue *nvmeq, u16 index)
{
	/*
	 * Single CQ entries are always smaller than a cache line, so we
	 * can't invalidate them individually. However CQ entries are
	 * read only by the CPU, so it's safe to always invalidate all of them,
	 * as the cache line should never become dirty.
	 */
	ulong start = (ulong)&nvmeq->cqes[0];
	ulong stop = start + NVME_CQ_ALLOCATION;

	invalidate_dcache_range(start, stop);

	return readw(&(nvmeq->cqes[index].status));
}

/**
 * nvme_submit_cmd() - copy a command into a queue and ring the doorbell
 *
 * @nvmeq:	The queue to use
 * @cmd:	The command to send
 */
static void nvme_submit_cmd(struct nvme_queue *nvmeq, struct nvme_command *cmd)
{
	u16 tail = nvmeq->sq_tail;

	memcpy(&nvmeq->sq_cmds[tail], cmd, sizeof(*cmd));
	flush_dcache_range((ulong)&nvmeq->sq_cmds[tail],
			   (ulong)&nvmeq->sq_cmds[tail] + sizeof(*cmd));

	if (++tail == nvmeq->q_depth)
		tail = 0;
	writel(tail, nvmeq->q_db);
	nvmeq->sq_tail = tail;
}

static int nvme_submit_sync_cmd(struct nvme_queue *nvmeq,
				struct nvme_command *cmd,
				u32 *result, unsigned timeout)
{
	u16 head = nvmeq->cq_head;
	u16 phase = nvmeq->cq_phase;
	u16 status;
	ulong start_time;
	ulong timeout_us = timeout * 100000;

	cmd->common.command_id = nvme_get_cmd_id();
	nvme_submit_cmd(nvmeq, cmd);

	start_time = timer_get_us();

	for (;;) {
		status = nvme_read_completion_status(nvmeq, head);
		if ((status & 0x01) == phase)
			break;
		if (timeout_us > 0 && (timer_get_us() - start_time)
		    >= timeout_us)
			return -ETIMEDOUT;
	}

	status >>= 1;
	if (status) {
		printf("ERROR: status = %x, phase = %d, head = %d\n",
		       status, phase, head);
		status = 0;
		if (++head == nvmeq->q_depth) {
			head = 0;
			phase = !phase;
		}
		writel(head, nvmeq->q_db + nvmeq->dev->db_stride);
		nvmeq->cq_head = head;
		nvmeq->cq_phase = phase;

		return -EIO;
	}

	if (result)
		*result = readl(&(nvmeq->cqes[head].result));

	if (++head == nvmeq->q_depth) {
		head = 0;
		phase = !phase;
	}
	writel(head, nvmeq->q_db + nvmeq->dev->db_stride);
	nvmeq->cq_head = head;
	nvmeq->cq_phase = phase;

	return status;
}

static int nvme_submit_admin_cmd(struct nvme_dev *dev, struct nvme_command *cmd,
				 u32 *result)
{
	return nvme_submit_sync_cmd(dev->queues[NVME_ADMIN_Q], cmd,
				    result, ADMIN_TIMEOUT);
}

static struct nvme_queue *nvme_alloc_queue(struct nvme_dev *dev,
					   int qid, int depth)
{
	struct nvme_queue *nvmeq = malloc(sizeof(*nvmeq));
	if (!nvmeq)
		return NULL;
	memset(nvmeq, 0, sizeof(*nvmeq));

	nvmeq->cqes = (void *)memalign(4096, NVME_CQ_ALLOCATION);
	if (!nvmeq->cqes)
		goto free_nvmeq;
	memset((void *)nvmeq->cqes, 0, NVME_CQ_SIZE(depth));

	nvmeq->sq_cmds = (void *)memalign(4096, NVME_SQ_SIZE(depth));
	if (!nvmeq->sq_cmds)
		goto free_queue;
	memset((void *)nvmeq->sq_cmds, 0, NVME_SQ_SIZE(depth));

	nvmeq->dev = dev;

	nvmeq->cq_head = 0;
	nvmeq->cq_phase = 1;
	nvmeq->q_db = &dev->dbs[qid * 2 * dev->db_stride];
	nvmeq->q_depth = depth;
	nvmeq->qid = qid;
	dev->queue_count++;
	dev->queues[qid] = nvmeq;

	return nvmeq;

 free_queue:
	free((void *)nvmeq->cqes);
 free_nvmeq:
	free(nvmeq);

	return NULL;
}

static int nvme_delete_queue(struct nvme_dev *dev, u8 opcode, u16 id)
{
	struct nvme_command c;

	memset(&c, 0, sizeof(c));
	c.delete_queue.opcode = opcode;
	c.delete_queue.qid = cpu_to_le16(id);

	return nvme_submit_admin_cmd(dev, &c, NULL);
}

static int nvme_delete_sq(struct nvme_dev *dev, u16 sqid)
{
	return nvme_delete_queue(dev, nvme_admin_delete_sq, sqid);
}

static int nvme_delete_cq(struct nvme_dev *dev, u16 cqid)
{
	return nvme_delete_queue(dev, nvme_admin_delete_cq, cqid);
}

static int nvme_enable_ctrl(struct nvme_dev *dev)
{
	dev->ctrl_config &= ~NVME_CC_SHN_MASK;
	dev->ctrl_config |= NVME_CC_ENABLE;
	writel(dev->ctrl_config, &dev->bar->cc);

	return nvme_wait_ready(dev, true);
}

static int nvme_disable_ctrl(struct nvme_dev *dev)
{
	dev->ctrl_config &= ~NVME_CC_SHN_MASK;
	dev->ctrl_config &= ~NVME_CC_ENABLE;
	writel(dev->ctrl_config, &dev->bar->cc);

	return nvme_wait_ready(dev, false);
}

static void nvme_free_queue(struct nvme_queue *nvmeq)
{
	free((void *)nvmeq->cqes);
	free(nvmeq->sq_cmds);
	free(nvmeq);
}

static void nvme_free_queues(struct nvme_dev *dev, int lowest)
{
	int i;

	for (i = dev->queue_count - 1; i >= lowest; i--) {
		struct nvme_queue *nvmeq = dev->queues[i];
		dev->queue_count--;
		dev->queues[i] = NULL;
		nvme_free_queue(nvmeq);
	}
}

static void nvme_init_queue(struct nvme_queue *nvmeq, u16 qid)
{
	struct nvme_dev *dev = nvmeq->dev;

	nvmeq->sq_tail = 0;
	nvmeq->cq_head = 0;
	nvmeq->cq_phase = 1;
	nvmeq->q_db = &dev->dbs[qid * 2 * dev->db_stride];
	memset((void *)nvmeq->cqes, 0, NVME_CQ_SIZE(nvmeq->q_depth));
	flush_dcache_range((ulong)nvmeq->cqes,
			   (ulong)nvmeq->cqes + NVME_CQ_ALLOCATION);
	dev->online_queues++;
}

static int nvme_configure_admin_queue(struct nvme_dev *dev)
{
	int result;
	u32 aqa;
	u64 cap = dev->cap;
	struct nvme_queue *nvmeq;
	/* most architectures use 4KB as the page size */
	unsigned page_shift = 12;
	unsigned dev_page_min = NVME_CAP_MPSMIN(cap) + 12;
	unsigned dev_page_max = NVME_CAP_MPSMAX(cap) + 12;

	if (page_shift < dev_page_min) {
		debug("Device minimum page size (%u) too large for host (%u)\n",
		      1 << dev_page_min, 1 << page_shift);
		return -ENODEV;
	}

	if (page_shift > dev_page_max) {
		debug("Device maximum page size (%u) smaller than host (%u)\n",
		      1 << dev_page_max, 1 << page_shift);
		page_shift = dev_page_max;
	}

	result = nvme_disable_ctrl(dev);
	if (result < 0)
		return result;

	nvmeq = dev->queues[NVME_ADMIN_Q];
	if (!nvmeq) {
		nvmeq = nvme_alloc_queue(dev, 0, NVME_AQ_DEPTH);
		if (!nvmeq)
			return -ENOMEM;
	}

	aqa = nvmeq->q_depth - 1;
	aqa |= aqa << 16;

	dev->page_size = 1 << page_shift;

	dev->ctrl_config = NVME_CC_CSS_NVM;
	dev->ctrl_config |= (page_shift - 12) << NVME_CC_MPS_SHIFT;
	dev->ctrl_config |= NVME_CC_ARB_RR | NVME_CC_SHN_NONE;
	dev->ctrl_config |= NVME_CC_IOSQES | NVME_CC_IOCQES;

	writel(aqa, &dev->bar->aqa);
	nvme_writeq((ulong)nvmeq->sq_cmds, &dev->bar->asq);
	nvme_writeq((ulong)nvmeq->cqes, &dev->bar->acq);

	result = nvme_enable_ctrl(dev);
	if (result)
		goto free_nvmeq;

	nvmeq->cq_vector = 0;

	nvme_init_queue(dev->queues[NVME_ADMIN_Q], 0);

	return result;

 free_nvmeq:
	nvme_free_queues(dev, 0);

	return result;
}

static int nvme_alloc_cq(struct nvme_dev *dev, u16 qid,
			    struct nvme_queue *nvmeq)
{
	struct nvme_command c;
	int flags = NVME_QUEUE_PHYS_CONTIG | NVME_CQ_IRQ_ENABLED;

	memset(&c, 0, sizeof(c));
	c.create_cq.opcode = nvme_admin_create_cq;
	c.create_cq.prp1 = cpu_to_le64((ulong)nvmeq->cqes);
	c.create_cq.cqid = cpu_to_le16(qid);
	c.create_cq.qsize = cpu_to_le16(nvmeq->q_depth - 1);
	c.create_cq.cq_flags = cpu_to_le16(flags);
	c.create_cq.irq_vector = cpu_to_le16(nvmeq->cq_vector);

	return nvme_submit_admin_cmd(dev, &c, NULL);
}

static int nvme_alloc_sq(struct nvme_dev *dev, u16 qid,
			    struct nvme_queue *nvmeq)
{
	struct nvme_command c;
	int flags = NVME_QUEUE_PHYS_CONTIG | NVME_SQ_PRIO_MEDIUM;

	memset(&c, 0, sizeof(c));
	c.create_sq.opcode = nvme_admin_create_sq;
	c.create_sq.prp1 = cpu_to_le64((ulong)nvmeq->sq_cmds);
	c.create_sq.sqid = cpu_to_le16(qid);
	c.create_sq.qsize = cpu_to_le16(nvmeq->q_depth - 1);
	c.create_sq.sq_flags = cpu_to_le16(flags);
	c.create_sq.cqid = cpu_to_le16(qid);

	return nvme_submit_admin_cmd(dev, &c, NULL);
}

int nvme_identify(struct nvme_dev *dev, unsigned nsid,
		  unsigned cns, dma_addr_t dma_addr)
{
	struct nvme_command c;
	u32 page_size = dev->page_size;
	int offset = dma_addr & (page_size - 1);
	int length = sizeof(struct nvme_id_ctrl);
	int ret;

	memset(&c, 0, sizeof(c));
	c.identify.opcode = nvme_admin_identify;
	c.identify.nsid = cpu_to_le32(nsid);
	c.identify.prp1 = cpu_to_le64(dma_addr);

	length -= (page_size - offset);
	if (length <= 0) {
		c.identify.prp2 = 0;
	} else {
		dma_addr += (page_size - offset);
		c.identify.prp2 = cpu_to_le64(dma_addr);
	}

	c.identify.cns = cpu_to_le32(cns);

	invalidate_dcache_range(dma_addr,
				dma_addr + sizeof(struct nvme_id_ctrl));

	ret = nvme_submit_admin_cmd(dev, &c, NULL);
	if (!ret)
		invalidate_dcache_range(dma_addr,
					dma_addr + sizeof(struct nvme_id_ctrl));

	return ret;
}

int nvme_get_features(struct nvme_dev *dev, unsigned fid, unsigned nsid,
		      dma_addr_t dma_addr, u32 *result)
{
	struct nvme_command c;
	int ret;

	memset(&c, 0, sizeof(c));
	c.features.opcode = nvme_admin_get_features;
	c.features.nsid = cpu_to_le32(nsid);
	c.features.prp1 = cpu_to_le64(dma_addr);
	c.features.fid = cpu_to_le32(fid);

	ret = nvme_submit_admin_cmd(dev, &c, result);

	/*
	 * TODO: Add some cache invalidation when a DMA buffer is involved
	 * in the request, here and before the command gets submitted. The
	 * buffer size varies by feature, also some features use a different
	 * field in the command packet to hold the buffer address.
	 * Section 5.21.1 (Set Features command) in the NVMe specification
	 * details the buffer requirements for each feature.
	 *
	 * At the moment there is no user of this function.
	 */

	return ret;
}

int nvme_set_features(struct nvme_dev *dev, unsigned fid, unsigned dword11,
		      dma_addr_t dma_addr, u32 *result)
{
	struct nvme_command c;

	memset(&c, 0, sizeof(c));
	c.features.opcode = nvme_admin_set_features;
	c.features.prp1 = cpu_to_le64(dma_addr);
	c.features.fid = cpu_to_le32(fid);
	c.features.dword11 = cpu_to_le32(dword11);

	/*
	 * TODO: Add a cache clean (aka flush) operation when a DMA buffer is
	 * involved in the request. The buffer size varies by feature, also
	 * some features use a different field in the command packet to hold
	 * the buffer address. Section 5.21.1 (Set Features command) in the
	 * NVMe specification details the buffer requirements for each
	 * feature.
	 * At the moment the only user of this function is not using
	 * any DMA buffer at all.
	 */

	return nvme_submit_admin_cmd(dev, &c, result);
}

static int nvme_create_queue(struct nvme_queue *nvmeq, int qid)
{
	struct nvme_dev *dev = nvmeq->dev;
	int result;

	nvmeq->cq_vector = qid - 1;
	result = nvme_alloc_cq(dev, qid, nvmeq);
	if (result < 0)
		goto release_cq;

	result = nvme_alloc_sq(dev, qid, nvmeq);
	if (result < 0)
		goto release_sq;

	nvme_init_queue(nvmeq, qid);

	return result;

 release_sq:
	nvme_delete_sq(dev, qid);
 release_cq:
	nvme_delete_cq(dev, qid);

	return result;
}

static int nvme_set_queue_count(struct nvme_dev *dev, int count)
{
	int status;
	u32 result;
	u32 q_count = (count - 1) | ((count - 1) << 16);

	status = nvme_set_features(dev, NVME_FEAT_NUM_QUEUES,
			q_count, 0, &result);

	if (status < 0)
		return status;
	if (status > 1)
		return 0;

	return min(result & 0xffff, result >> 16) + 1;
}

static void nvme_create_io_queues(struct nvme_dev *dev)
{
	unsigned int i;

	for (i = dev->queue_count; i <= dev->max_qid; i++)
		if (!nvme_alloc_queue(dev, i, dev->q_depth))
			break;

	for (i = dev->online_queues; i <= dev->queue_count - 1; i++)
		if (nvme_create_queue(dev->queues[i], i))
			break;
}

static int nvme_setup_io_queues(struct nvme_dev *dev)
{
	int nr_io_queues;
	int result;

	nr_io_queues = 1;
	result = nvme_set_queue_count(dev, nr_io_queues);
	if (result <= 0)
		return result;

	dev->max_qid = nr_io_queues;

	/* Free previously allocated queues */
	nvme_free_queues(dev, nr_io_queues + 1);
	nvme_create_io_queues(dev);

	return 0;
}

static int nvme_get_info_from_identify(struct nvme_dev *dev)
{
	struct nvme_id_ctrl *ctrl;
	int ret;
	int shift = NVME_CAP_MPSMIN(dev->cap) + 12;

	ctrl = memalign(dev->page_size, sizeof(struct nvme_id_ctrl));
	if (!ctrl)
		return -ENOMEM;

	ret = nvme_identify(dev, 0, 1, (dma_addr_t)(long)ctrl);
	if (ret) {
		free(ctrl);
		return -EIO;
	}

	dev->nn = le32_to_cpu(ctrl->nn);
	dev->vwc = ctrl->vwc;
	memcpy(dev->serial, ctrl->sn, sizeof(ctrl->sn));
	memcpy(dev->model, ctrl->mn, sizeof(ctrl->mn));
	memcpy(dev->firmware_rev, ctrl->fr, sizeof(ctrl->fr));
	if (ctrl->mdts)
		dev->max_transfer_shift = (ctrl->mdts + shift);
	else {
		/*
		 * Maximum Data Transfer Size (MDTS) field indicates the maximum
		 * data transfer size between the host and the controller. The
		 * host should not submit a command that exceeds this transfer
		 * size. The value is in units of the minimum memory page size
		 * and is reported as a power of two (2^n).
		 *
		 * The spec also says: a value of 0h indicates no restrictions
		 * on transfer size. But in nvme_blk_read/write() below we have
		 * the following algorithm for maximum number of logic blocks
		 * per transfer:
		 *
		 * u16 lbas = 1 << (dev->max_transfer_shift - ns->lba_shift);
		 *
		 * In order for lbas not to overflow, the maximum number is 15
		 * which means dev->max_transfer_shift = 15 + 9 (ns->lba_shift).
		 * Let's use 20 which provides 1MB size.
		 */
		dev->max_transfer_shift = 20;
	}

	free(ctrl);
	return 0;
}

int nvme_get_namespace_id(struct udevice *udev, u32 *ns_id, u8 *eui64)
{
	struct nvme_ns *ns = dev_get_priv(udev);

	if (ns_id)
		*ns_id = ns->ns_id;
	if (eui64)
		memcpy(eui64, ns->eui64, sizeof(ns->eui64));

	return 0;
}

int nvme_scan_namespace(void)
{
	struct uclass *uc;
	struct udevice *dev;
	int ret;

	ret = uclass_get(UCLASS_NVME, &uc);
	if (ret)
		return ret;

	uclass_foreach_dev(dev, uc) {
		ret = device_probe(dev);
		if (ret)
			return ret;
	}

	return 0;
}

static int nvme_blk_probe(struct udevice *udev)
{
	struct nvme_dev *ndev = dev_get_priv(udev->parent);
	struct blk_desc *desc = dev_get_uclass_plat(udev);
	struct nvme_ns *ns = dev_get_priv(udev);
	u8 flbas;
	struct pci_child_plat *pplat;
	struct nvme_id_ns *id;

	id = memalign(ndev->page_size, sizeof(struct nvme_id_ns));
	if (!id)
		return -ENOMEM;

	ns->dev = ndev;
	/* extract the namespace id from the block device name */
	ns->ns_id = trailing_strtol(udev->name);
	if (nvme_identify(ndev, ns->ns_id, 0, (dma_addr_t)(long)id)) {
		free(id);
		return -EIO;
	}

	memcpy(&ns->eui64, &id->eui64, sizeof(id->eui64));
	flbas = id->flbas & NVME_NS_FLBAS_LBA_MASK;
	ns->flbas = flbas;
	ns->lba_shift = id->lbaf[flbas].ds;
	list_add(&ns->list, &ndev->namespaces);

	desc->lba = le64_to_cpu(id->nsze);
	desc->log2blksz = ns->lba_shift;
	desc->blksz = 1 << ns->lba_shift;
	desc->bdev = udev;
	pplat = dev_get_parent_plat(udev->parent);
	sprintf(desc->vendor, "0x%.4x", pplat->vendor);
	memcpy(desc->product, ndev->serial, sizeof(ndev->serial));
	memcpy(desc->revision, ndev->firmware_rev, sizeof(ndev->firmware_rev));

	free(id);
	return 0;
}

static ulong nvme_blk_rw(struct udevice *udev, lbaint_t blknr,
			 lbaint_t blkcnt, void *buffer, bool read)
{
	struct nvme_ns *ns = dev_get_priv(udev);
	struct nvme_dev *dev = ns->dev;
	struct nvme_command c;
	struct blk_desc *desc = dev_get_uclass_plat(udev);
	int status;
	u64 prp2;
	u64 total_len = blkcnt << desc->log2blksz;
	u64 temp_len = total_len;

	u64 slba = blknr;
	u16 lbas = 1 << (dev->max_transfer_shift - ns->lba_shift);
	u64 total_lbas = blkcnt;

	flush_dcache_range((unsigned long)buffer,
			   (unsigned long)buffer + total_len);

	c.rw.opcode = read ? nvme_cmd_read : nvme_cmd_write;
	c.rw.flags = 0;
	c.rw.nsid = cpu_to_le32(ns->ns_id);
	c.rw.control = 0;
	c.rw.dsmgmt = 0;
	c.rw.reftag = 0;
	c.rw.apptag = 0;
	c.rw.appmask = 0;
	c.rw.metadata = 0;

	while (total_lbas) {
		if (total_lbas < lbas) {
			lbas = (u16)total_lbas;
			total_lbas = 0;
		} else {
			total_lbas -= lbas;
		}

		if (nvme_setup_prps(dev, &prp2,
				    lbas << ns->lba_shift, (ulong)buffer))
			return -EIO;
		c.rw.slba = cpu_to_le64(slba);
		slba += lbas;
		c.rw.length = cpu_to_le16(lbas - 1);
		c.rw.prp1 = cpu_to_le64((ulong)buffer);
		c.rw.prp2 = cpu_to_le64(prp2);
		status = nvme_submit_sync_cmd(dev->queues[NVME_IO_Q],
				&c, NULL, IO_TIMEOUT);
		if (status)
			break;
		temp_len -= (u32)lbas << ns->lba_shift;
		buffer += lbas << ns->lba_shift;
	}

	if (read)
		invalidate_dcache_range((unsigned long)buffer,
					(unsigned long)buffer + total_len);

	return (total_len - temp_len) >> desc->log2blksz;
}

static ulong nvme_blk_read(struct udevice *udev, lbaint_t blknr,
			   lbaint_t blkcnt, void *buffer)
{
	return nvme_blk_rw(udev, blknr, blkcnt, buffer, true);
}

static ulong nvme_blk_write(struct udevice *udev, lbaint_t blknr,
			    lbaint_t blkcnt, const void *buffer)
{
	return nvme_blk_rw(udev, blknr, blkcnt, (void *)buffer, false);
}

static const struct blk_ops nvme_blk_ops = {
	.read	= nvme_blk_read,
	.write	= nvme_blk_write,
};

U_BOOT_DRIVER(nvme_blk) = {
	.name	= "nvme-blk",
	.id	= UCLASS_BLK,
	.probe	= nvme_blk_probe,
	.ops	= &nvme_blk_ops,
	.priv_auto	= sizeof(struct nvme_ns),
};

static int nvme_bind(struct udevice *udev)
{
	static int ndev_num;
	char name[20];

	sprintf(name, "nvme#%d", ndev_num++);

	return device_set_name(udev, name);
}

static int nvme_probe(struct udevice *udev)
{
	int ret;
	struct nvme_dev *ndev = dev_get_priv(udev);
	struct nvme_id_ns *id;

	ndev->instance = trailing_strtol(udev->name);

	INIT_LIST_HEAD(&ndev->namespaces);
	ndev->bar = dm_pci_map_bar(udev, PCI_BASE_ADDRESS_0,
			PCI_REGION_MEM);
	if (readl(&ndev->bar->csts) == -1) {
		ret = -ENODEV;
		printf("Error: %s: Out of memory!\n", udev->name);
		goto free_nvme;
	}

	ndev->queues = malloc(NVME_Q_NUM * sizeof(struct nvme_queue *));
	if (!ndev->queues) {
		ret = -ENOMEM;
		printf("Error: %s: Out of memory!\n", udev->name);
		goto free_nvme;
	}
	memset(ndev->queues, 0, NVME_Q_NUM * sizeof(struct nvme_queue *));

	ndev->cap = nvme_readq(&ndev->bar->cap);
	ndev->q_depth = min_t(int, NVME_CAP_MQES(ndev->cap) + 1, NVME_Q_DEPTH);
	ndev->db_stride = 1 << NVME_CAP_STRIDE(ndev->cap);
	ndev->dbs = ((void __iomem *)ndev->bar) + 4096;

	ret = nvme_configure_admin_queue(ndev);
	if (ret)
		goto free_queue;

	/* Allocate after the page size is known */
	ndev->prp_pool = memalign(ndev->page_size, MAX_PRP_POOL);
	if (!ndev->prp_pool) {
		ret = -ENOMEM;
		printf("Error: %s: Out of memory!\n", udev->name);
		goto free_nvme;
	}
	ndev->prp_entry_num = MAX_PRP_POOL >> 3;

	ret = nvme_setup_io_queues(ndev);
	if (ret)
		goto free_queue;

	nvme_get_info_from_identify(ndev);

	/* Create a blk device for each namespace */

	id = memalign(ndev->page_size, sizeof(struct nvme_id_ns));
	if (!id) {
		ret = -ENOMEM;
		goto free_queue;
	}

	for (int i = 1; i <= ndev->nn; i++) {
		struct udevice *ns_udev;
		char name[20];

		memset(id, 0, sizeof(*id));
		if (nvme_identify(ndev, i, 0, (dma_addr_t)(long)id)) {
			ret = -EIO;
			goto free_id;
		}

		/* skip inactive namespace */
		if (!id->nsze)
			continue;

		/*
		 * Encode the namespace id to the device name so that
		 * we can extract it when doing the probe.
		 */
		sprintf(name, "blk#%d", i);

		/* The real blksz and size will be set by nvme_blk_probe() */
		ret = blk_create_devicef(udev, "nvme-blk", name, IF_TYPE_NVME,
					 -1, 512, 0, &ns_udev);
		if (ret)
			goto free_id;
	}

	free(id);
	return 0;

free_id:
	free(id);
free_queue:
	free((void *)ndev->queues);
free_nvme:
	return ret;
}

U_BOOT_DRIVER(nvme) = {
	.name	= "nvme",
	.id	= UCLASS_NVME,
	.bind	= nvme_bind,
	.probe	= nvme_probe,
	.priv_auto	= sizeof(struct nvme_dev),
};

struct pci_device_id nvme_supported[] = {
	{ PCI_DEVICE_CLASS(PCI_CLASS_STORAGE_EXPRESS, ~0) },
	{}
};

U_BOOT_PCI_DEVICE(nvme, nvme_supported);