summaryrefslogtreecommitdiffstats
path: root/module/zfs/vdev_disk.c
blob: 4e362226a88067d870f73fcdf4590363eeca699f (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
/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */
/*
 * Copyright (C) 2008-2010 Lawrence Livermore National Security, LLC.
 * Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
 * Rewritten for Linux by Brian Behlendorf <behlendorf1@llnl.gov>.
 * LLNL-CODE-403049.
 * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
 */

#include <sys/zfs_context.h>
#include <sys/spa.h>
#include <sys/vdev_disk.h>
#include <sys/vdev_impl.h>
#include <sys/fs/zfs.h>
#include <sys/zio.h>
#include <sys/sunldi.h>

char *zfs_vdev_scheduler = VDEV_SCHEDULER;
static void *zfs_vdev_holder = VDEV_HOLDER;

/*
 * Virtual device vector for disks.
 */
typedef struct dio_request {
	struct completion	dr_comp;	/* Completion for sync IO */
	zio_t			*dr_zio;	/* Parent ZIO */
	atomic_t		dr_ref;		/* References */
	int			dr_wait;	/* Wait for IO */
	int			dr_error;	/* Bio error */
	int			dr_bio_count;	/* Count of bio's */
	struct bio		*dr_bio[0];	/* Attached bio's */
} dio_request_t;


#ifdef HAVE_OPEN_BDEV_EXCLUSIVE
static fmode_t
vdev_bdev_mode(int smode)
{
	fmode_t mode = 0;

	ASSERT3S(smode & (FREAD | FWRITE), !=, 0);

	if (smode & FREAD)
		mode |= FMODE_READ;

	if (smode & FWRITE)
		mode |= FMODE_WRITE;

	return (mode);
}
#else
static int
vdev_bdev_mode(int smode)
{
	int mode = 0;

	ASSERT3S(smode & (FREAD | FWRITE), !=, 0);

	if ((smode & FREAD) && !(smode & FWRITE))
		mode = MS_RDONLY;

	return (mode);
}
#endif /* HAVE_OPEN_BDEV_EXCLUSIVE */

static uint64_t
bdev_capacity(struct block_device *bdev)
{
	struct hd_struct *part = bdev->bd_part;

	/* The partition capacity referenced by the block device */
	if (part)
		return (part->nr_sects << 9);

	/* Otherwise assume the full device capacity */
	return (get_capacity(bdev->bd_disk) << 9);
}

static void
vdev_disk_error(zio_t *zio)
{
#ifdef ZFS_DEBUG
	printk("ZFS: zio error=%d type=%d offset=%llu size=%llu "
	    "flags=%x\n", zio->io_error, zio->io_type,
	    (u_longlong_t)zio->io_offset, (u_longlong_t)zio->io_size,
	    zio->io_flags);
#endif
}

/*
 * Use the Linux 'noop' elevator for zfs managed block devices.  This
 * strikes the ideal balance by allowing the zfs elevator to do all
 * request ordering and prioritization.  While allowing the Linux
 * elevator to do the maximum front/back merging allowed by the
 * physical device.  This yields the largest possible requests for
 * the device with the lowest total overhead.
 */
static int
vdev_elevator_switch(vdev_t *v, char *elevator)
{
	vdev_disk_t *vd = v->vdev_tsd;
	struct block_device *bdev = vd->vd_bdev;
	struct request_queue *q = bdev_get_queue(bdev);
	char *device = bdev->bd_disk->disk_name;
	int error;

	/*
	 * Skip devices which are not whole disks (partitions).
	 * Device-mapper devices are excepted since they may be whole
	 * disks despite the vdev_wholedisk flag, in which case we can
	 * and should switch the elevator. If the device-mapper device
	 * does not have an elevator (i.e. dm-raid, dm-crypt, etc.) the
	 * "Skip devices without schedulers" check below will fail.
	 */
	if (!v->vdev_wholedisk && strncmp(device, "dm-", 3) != 0)
		return (0);

	/* Skip devices without schedulers (loop, ram, dm, etc) */
	if (!q->elevator || !blk_queue_stackable(q))
		return (0);

	/* Leave existing scheduler when set to "none" */
	if ((strncmp(elevator, "none", 4) == 0) && (strlen(elevator) == 4))
		return (0);

#ifdef HAVE_ELEVATOR_CHANGE
	error = elevator_change(q, elevator);
#else
	/*
	 * For pre-2.6.36 kernels elevator_change() is not available.
	 * Therefore we fall back to using a usermodehelper to echo the
	 * elevator into sysfs;  This requires /bin/echo and sysfs to be
	 * mounted which may not be true early in the boot process.
	 */
#define	SET_SCHEDULER_CMD \
	"exec 0</dev/null " \
	"     1>/sys/block/%s/queue/scheduler " \
	"     2>/dev/null; " \
	"echo %s"

	{
		char *argv[] = { "/bin/sh", "-c", NULL, NULL };
		char *envp[] = { NULL };

		argv[2] = kmem_asprintf(SET_SCHEDULER_CMD, device, elevator);
		error = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC);
		strfree(argv[2]);
	}
#endif /* HAVE_ELEVATOR_CHANGE */
	if (error)
		printk("ZFS: Unable to set \"%s\" scheduler for %s (%s): %d\n",
		    elevator, v->vdev_path, device, error);

	return (error);
}

/*
 * Expanding a whole disk vdev involves invoking BLKRRPART on the
 * whole disk device. This poses a problem, because BLKRRPART will
 * return EBUSY if one of the disk's partitions is open. That's why
 * we have to do it here, just before opening the data partition.
 * Unfortunately, BLKRRPART works by dropping all partitions and
 * recreating them, which means that for a short time window, all
 * /dev/sdxN device files disappear (until udev recreates them).
 * This means two things:
 *  - When we open the data partition just after a BLKRRPART, we
 *    can't do it using the normal device file path because of the
 *    obvious race condition with udev. Instead, we use reliable
 *    kernel APIs to get a handle to the new partition device from
 *    the whole disk device.
 *  - Because vdev_disk_open() initially needs to find the device
 *    using its path, multiple vdev_disk_open() invocations in
 *    short succession on the same disk with BLKRRPARTs in the
 *    middle have a high probability of failure (because of the
 *    race condition with udev). A typical situation where this
 *    might happen is when the zpool userspace tool does a
 *    TRYIMPORT immediately followed by an IMPORT. For this
 *    reason, we only invoke BLKRRPART in the module when strictly
 *    necessary (zpool online -e case), and rely on userspace to
 *    do it when possible.
 */
static struct block_device *
vdev_disk_rrpart(const char *path, int mode, vdev_disk_t *vd)
{
#if defined(HAVE_3ARG_BLKDEV_GET) && defined(HAVE_GET_GENDISK)
	struct block_device *bdev, *result = ERR_PTR(-ENXIO);
	struct gendisk *disk;
	int error, partno;

	bdev = vdev_bdev_open(path, vdev_bdev_mode(mode), zfs_vdev_holder);
	if (IS_ERR(bdev))
		return (bdev);

	disk = get_gendisk(bdev->bd_dev, &partno);
	vdev_bdev_close(bdev, vdev_bdev_mode(mode));

	if (disk) {
		bdev = bdget(disk_devt(disk));
		if (bdev) {
			error = blkdev_get(bdev, vdev_bdev_mode(mode), vd);
			if (error == 0)
				error = ioctl_by_bdev(bdev, BLKRRPART, 0);
			vdev_bdev_close(bdev, vdev_bdev_mode(mode));
		}

		bdev = bdget_disk(disk, partno);
		if (bdev) {
			error = blkdev_get(bdev,
			    vdev_bdev_mode(mode) | FMODE_EXCL, vd);
			if (error == 0)
				result = bdev;
		}
		put_disk(disk);
	}

	return (result);
#else
	return (ERR_PTR(-EOPNOTSUPP));
#endif /* defined(HAVE_3ARG_BLKDEV_GET) && defined(HAVE_GET_GENDISK) */
}

static int
vdev_disk_open(vdev_t *v, uint64_t *psize, uint64_t *max_psize,
    uint64_t *ashift)
{
	struct block_device *bdev = ERR_PTR(-ENXIO);
	vdev_disk_t *vd;
	int count = 0, mode, block_size;

	/* Must have a pathname and it must be absolute. */
	if (v->vdev_path == NULL || v->vdev_path[0] != '/') {
		v->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
		return (SET_ERROR(EINVAL));
	}

	/*
	 * Reopen the device if it's not currently open. Otherwise,
	 * just update the physical size of the device.
	 */
	if (v->vdev_tsd != NULL) {
		ASSERT(v->vdev_reopening);
		vd = v->vdev_tsd;
		goto skip_open;
	}

	vd = kmem_zalloc(sizeof (vdev_disk_t), KM_SLEEP);
	if (vd == NULL)
		return (SET_ERROR(ENOMEM));

	/*
	 * Devices are always opened by the path provided at configuration
	 * time.  This means that if the provided path is a udev by-id path
	 * then drives may be recabled without an issue.  If the provided
	 * path is a udev by-path path, then the physical location information
	 * will be preserved.  This can be critical for more complicated
	 * configurations where drives are located in specific physical
	 * locations to maximize the systems tolerence to component failure.
	 * Alternatively, you can provide your own udev rule to flexibly map
	 * the drives as you see fit.  It is not advised that you use the
	 * /dev/[hd]d devices which may be reordered due to probing order.
	 * Devices in the wrong locations will be detected by the higher
	 * level vdev validation.
	 *
	 * The specified paths may be briefly removed and recreated in
	 * response to udev events.  This should be exceptionally unlikely
	 * because the zpool command makes every effort to verify these paths
	 * have already settled prior to reaching this point.  Therefore,
	 * a ENOENT failure at this point is highly likely to be transient
	 * and it is reasonable to sleep and retry before giving up.  In
	 * practice delays have been observed to be on the order of 100ms.
	 */
	mode = spa_mode(v->vdev_spa);
	if (v->vdev_wholedisk && v->vdev_expanding)
		bdev = vdev_disk_rrpart(v->vdev_path, mode, vd);

	while (IS_ERR(bdev) && count < 50) {
		bdev = vdev_bdev_open(v->vdev_path,
		    vdev_bdev_mode(mode), zfs_vdev_holder);
		if (unlikely(PTR_ERR(bdev) == -ENOENT)) {
			msleep(10);
			count++;
		} else if (IS_ERR(bdev)) {
			break;
		}
	}

	if (IS_ERR(bdev)) {
		dprintf("failed open v->vdev_path=%s, error=%d count=%d\n",
		    v->vdev_path, -PTR_ERR(bdev), count);
		kmem_free(vd, sizeof (vdev_disk_t));
		return (SET_ERROR(-PTR_ERR(bdev)));
	}

	v->vdev_tsd = vd;
	vd->vd_bdev = bdev;

skip_open:
	/*  Determine the physical block size */
	block_size = vdev_bdev_block_size(vd->vd_bdev);

	/* Clear the nowritecache bit, causes vdev_reopen() to try again. */
	v->vdev_nowritecache = B_FALSE;

	/* Inform the ZIO pipeline that we are non-rotational */
	v->vdev_nonrot = blk_queue_nonrot(bdev_get_queue(vd->vd_bdev));

	/* Physical volume size in bytes */
	*psize = bdev_capacity(vd->vd_bdev);

	/* TODO: report possible expansion size */
	*max_psize = *psize;

	/* Based on the minimum sector size set the block size */
	*ashift = highbit64(MAX(block_size, SPA_MINBLOCKSIZE)) - 1;

	/* Try to set the io scheduler elevator algorithm */
	(void) vdev_elevator_switch(v, zfs_vdev_scheduler);

	return (0);
}

static void
vdev_disk_close(vdev_t *v)
{
	vdev_disk_t *vd = v->vdev_tsd;

	if (v->vdev_reopening || vd == NULL)
		return;

	if (vd->vd_bdev != NULL)
		vdev_bdev_close(vd->vd_bdev,
		    vdev_bdev_mode(spa_mode(v->vdev_spa)));

	kmem_free(vd, sizeof (vdev_disk_t));
	v->vdev_tsd = NULL;
}

static dio_request_t *
vdev_disk_dio_alloc(int bio_count)
{
	dio_request_t *dr;
	int i;

	dr = kmem_zalloc(sizeof (dio_request_t) +
	    sizeof (struct bio *) * bio_count, KM_SLEEP);
	if (dr) {
		init_completion(&dr->dr_comp);
		atomic_set(&dr->dr_ref, 0);
		dr->dr_bio_count = bio_count;
		dr->dr_error = 0;

		for (i = 0; i < dr->dr_bio_count; i++)
			dr->dr_bio[i] = NULL;
	}

	return (dr);
}

static void
vdev_disk_dio_free(dio_request_t *dr)
{
	int i;

	for (i = 0; i < dr->dr_bio_count; i++)
		if (dr->dr_bio[i])
			bio_put(dr->dr_bio[i]);

	kmem_free(dr, sizeof (dio_request_t) +
	    sizeof (struct bio *) * dr->dr_bio_count);
}

static void
vdev_disk_dio_get(dio_request_t *dr)
{
	atomic_inc(&dr->dr_ref);
}

static int
vdev_disk_dio_put(dio_request_t *dr)
{
	int rc = atomic_dec_return(&dr->dr_ref);

	/*
	 * Free the dio_request when the last reference is dropped and
	 * ensure zio_interpret is called only once with the correct zio
	 */
	if (rc == 0) {
		zio_t *zio = dr->dr_zio;
		int error = dr->dr_error;

		vdev_disk_dio_free(dr);

		if (zio) {
			zio->io_error = error;
			ASSERT3S(zio->io_error, >=, 0);
			if (zio->io_error)
				vdev_disk_error(zio);
			zio_interrupt(zio);
		}
	}

	return (rc);
}

BIO_END_IO_PROTO(vdev_disk_physio_completion, bio, error)
{
	dio_request_t *dr = bio->bi_private;
	int rc;
	int wait;

	if (dr->dr_error == 0) {
#ifdef HAVE_1ARG_BIO_END_IO_T
		dr->dr_error = -(bio->bi_error);
#else
		if (error)
			dr->dr_error = -(error);
		else if (!test_bit(BIO_UPTODATE, &bio->bi_flags))
			dr->dr_error = EIO;
#endif
	}

	wait = dr->dr_wait;
	/* Drop reference aquired by __vdev_disk_physio */
	rc = vdev_disk_dio_put(dr);

	/* Wake up synchronous waiter this is the last outstanding bio */
	if (wait && rc == 1)
		complete(&dr->dr_comp);
}

static inline unsigned long
bio_nr_pages(void *bio_ptr, unsigned int bio_size)
{
	return ((((unsigned long)bio_ptr + bio_size + PAGE_SIZE - 1) >>
	    PAGE_SHIFT) - ((unsigned long)bio_ptr >> PAGE_SHIFT));
}

static unsigned int
bio_map(struct bio *bio, void *bio_ptr, unsigned int bio_size)
{
	unsigned int offset, size, i;
	struct page *page;

	offset = offset_in_page(bio_ptr);
	for (i = 0; i < bio->bi_max_vecs; i++) {
		size = PAGE_SIZE - offset;

		if (bio_size <= 0)
			break;

		if (size > bio_size)
			size = bio_size;

		if (is_vmalloc_addr(bio_ptr))
			page = vmalloc_to_page(bio_ptr);
		else
			page = virt_to_page(bio_ptr);

		/*
		 * Some network related block device uses tcp_sendpage, which
		 * doesn't behave well when using 0-count page, this is a
		 * safety net to catch them.
		 */
		ASSERT3S(page_count(page), >, 0);

		if (bio_add_page(bio, page, size, offset) != size)
			break;

		bio_ptr  += size;
		bio_size -= size;
		offset = 0;
	}

	return (bio_size);
}

static inline void
vdev_submit_bio(int rw, struct bio *bio)
{
#ifdef HAVE_CURRENT_BIO_TAIL
	struct bio **bio_tail = current->bio_tail;
	current->bio_tail = NULL;
	submit_bio(rw, bio);
	current->bio_tail = bio_tail;
#else
	struct bio_list *bio_list = current->bio_list;
	current->bio_list = NULL;
	submit_bio(rw, bio);
	current->bio_list = bio_list;
#endif
}

static int
__vdev_disk_physio(struct block_device *bdev, zio_t *zio, caddr_t kbuf_ptr,
    size_t kbuf_size, uint64_t kbuf_offset, int flags, int wait)
{
	dio_request_t *dr;
	caddr_t bio_ptr;
	uint64_t bio_offset;
	int rw, bio_size, bio_count = 16;
	int i = 0, error = 0;

	ASSERT3U(kbuf_offset + kbuf_size, <=, bdev->bd_inode->i_size);

retry:
	dr = vdev_disk_dio_alloc(bio_count);
	if (dr == NULL)
		return (ENOMEM);

	if (zio && !(zio->io_flags & (ZIO_FLAG_IO_RETRY | ZIO_FLAG_TRYHARD)))
		bio_set_flags_failfast(bdev, &flags);

	rw = flags;
	dr->dr_zio = zio;
	dr->dr_wait = wait;

	/*
	 * When the IO size exceeds the maximum bio size for the request
	 * queue we are forced to break the IO in multiple bio's and wait
	 * for them all to complete.  Ideally, all pool users will set
	 * their volume block size to match the maximum request size and
	 * the common case will be one bio per vdev IO request.
	 */
	bio_ptr    = kbuf_ptr;
	bio_offset = kbuf_offset;
	bio_size   = kbuf_size;
	for (i = 0; i <= dr->dr_bio_count; i++) {

		/* Finished constructing bio's for given buffer */
		if (bio_size <= 0)
			break;

		/*
		 * By default only 'bio_count' bio's per dio are allowed.
		 * However, if we find ourselves in a situation where more
		 * are needed we allocate a larger dio and warn the user.
		 */
		if (dr->dr_bio_count == i) {
			vdev_disk_dio_free(dr);
			bio_count *= 2;
			goto retry;
		}

		/* bio_alloc() with __GFP_WAIT never returns NULL */
		dr->dr_bio[i] = bio_alloc(GFP_NOIO,
		    MIN(bio_nr_pages(bio_ptr, bio_size), BIO_MAX_PAGES));
		if (unlikely(dr->dr_bio[i] == NULL)) {
			vdev_disk_dio_free(dr);
			return (ENOMEM);
		}

		/* Matching put called by vdev_disk_physio_completion */
		vdev_disk_dio_get(dr);

		dr->dr_bio[i]->bi_bdev = bdev;
		BIO_BI_SECTOR(dr->dr_bio[i]) = bio_offset >> 9;
		dr->dr_bio[i]->bi_rw = rw;
		dr->dr_bio[i]->bi_end_io = vdev_disk_physio_completion;
		dr->dr_bio[i]->bi_private = dr;

		/* Remaining size is returned to become the new size */
		bio_size = bio_map(dr->dr_bio[i], bio_ptr, bio_size);

		/* Advance in buffer and construct another bio if needed */
		bio_ptr    += BIO_BI_SIZE(dr->dr_bio[i]);
		bio_offset += BIO_BI_SIZE(dr->dr_bio[i]);
	}

	/* Extra reference to protect dio_request during vdev_submit_bio */
	vdev_disk_dio_get(dr);

	/* Submit all bio's associated with this dio */
	for (i = 0; i < dr->dr_bio_count; i++)
		if (dr->dr_bio[i])
			vdev_submit_bio(rw, dr->dr_bio[i]);

	/*
	 * On synchronous blocking requests we wait for all bio the completion
	 * callbacks to run.  We will be woken when the last callback runs
	 * for this dio.  We are responsible for putting the last dio_request
	 * reference will in turn put back the last bio references.  The
	 * only synchronous consumer is vdev_disk_read_rootlabel() all other
	 * IO originating from vdev_disk_io_start() is asynchronous.
	 */
	if (wait) {
		wait_for_completion(&dr->dr_comp);
		error = dr->dr_error;
		ASSERT3S(atomic_read(&dr->dr_ref), ==, 1);
	}

	(void) vdev_disk_dio_put(dr);

	return (error);
}

int
vdev_disk_physio(struct block_device *bdev, caddr_t kbuf,
    size_t size, uint64_t offset, int flags)
{
	bio_set_flags_failfast(bdev, &flags);
	return (__vdev_disk_physio(bdev, NULL, kbuf, size, offset, flags, 1));
}

BIO_END_IO_PROTO(vdev_disk_io_flush_completion, bio, rc)
{
	zio_t *zio = bio->bi_private;
#ifdef HAVE_1ARG_BIO_END_IO_T
	int rc = bio->bi_error;
#endif

	zio->io_error = -rc;
	if (rc && (rc == -EOPNOTSUPP))
		zio->io_vd->vdev_nowritecache = B_TRUE;

	bio_put(bio);
	ASSERT3S(zio->io_error, >=, 0);
	if (zio->io_error)
		vdev_disk_error(zio);
	zio_interrupt(zio);
}

static int
vdev_disk_io_flush(struct block_device *bdev, zio_t *zio)
{
	struct request_queue *q;
	struct bio *bio;

	q = bdev_get_queue(bdev);
	if (!q)
		return (ENXIO);

	bio = bio_alloc(GFP_NOIO, 0);
	/* bio_alloc() with __GFP_WAIT never returns NULL */
	if (unlikely(bio == NULL))
		return (ENOMEM);

	bio->bi_end_io = vdev_disk_io_flush_completion;
	bio->bi_private = zio;
	bio->bi_bdev = bdev;
	vdev_submit_bio(VDEV_WRITE_FLUSH_FUA, bio);
	invalidate_bdev(bdev);

	return (0);
}

static void
vdev_disk_io_start(zio_t *zio)
{
	vdev_t *v = zio->io_vd;
	vdev_disk_t *vd = v->vdev_tsd;
	zio_priority_t pri = zio->io_priority;
	int flags, error;

	switch (zio->io_type) {
	case ZIO_TYPE_IOCTL:

		if (!vdev_readable(v)) {
			zio->io_error = SET_ERROR(ENXIO);
			zio_interrupt(zio);
			return;
		}

		switch (zio->io_cmd) {
		case DKIOCFLUSHWRITECACHE:

			if (zfs_nocacheflush)
				break;

			if (v->vdev_nowritecache) {
				zio->io_error = SET_ERROR(ENOTSUP);
				break;
			}

			error = vdev_disk_io_flush(vd->vd_bdev, zio);
			if (error == 0)
				return;

			zio->io_error = error;
			if (error == ENOTSUP)
				v->vdev_nowritecache = B_TRUE;

			break;

		default:
			zio->io_error = SET_ERROR(ENOTSUP);
		}

		zio_execute(zio);
		return;
	case ZIO_TYPE_WRITE:
		if ((pri == ZIO_PRIORITY_SYNC_WRITE) && (v->vdev_nonrot))
			flags = WRITE_SYNC;
		else
			flags = WRITE;
		break;

	case ZIO_TYPE_READ:
		if ((pri == ZIO_PRIORITY_SYNC_READ) && (v->vdev_nonrot))
			flags = READ_SYNC;
		else
			flags = READ;
		break;

	default:
		zio->io_error = SET_ERROR(ENOTSUP);
		zio_interrupt(zio);
		return;
	}

	error = __vdev_disk_physio(vd->vd_bdev, zio, zio->io_data,
	    zio->io_size, zio->io_offset, flags, 0);
	if (error) {
		zio->io_error = error;
		zio_interrupt(zio);
		return;
	}
}

static void
vdev_disk_io_done(zio_t *zio)
{
	/*
	 * If the device returned EIO, we revalidate the media.  If it is
	 * determined the media has changed this triggers the asynchronous
	 * removal of the device from the configuration.
	 */
	if (zio->io_error == EIO) {
		vdev_t *v = zio->io_vd;
		vdev_disk_t *vd = v->vdev_tsd;

		if (check_disk_change(vd->vd_bdev)) {
			vdev_bdev_invalidate(vd->vd_bdev);
			v->vdev_remove_wanted = B_TRUE;
			spa_async_request(zio->io_spa, SPA_ASYNC_REMOVE);
		}
	}
}

static void
vdev_disk_hold(vdev_t *vd)
{
	ASSERT(spa_config_held(vd->vdev_spa, SCL_STATE, RW_WRITER));

	/* We must have a pathname, and it must be absolute. */
	if (vd->vdev_path == NULL || vd->vdev_path[0] != '/')
		return;

	/*
	 * Only prefetch path and devid info if the device has
	 * never been opened.
	 */
	if (vd->vdev_tsd != NULL)
		return;

	/* XXX: Implement me as a vnode lookup for the device */
	vd->vdev_name_vp = NULL;
	vd->vdev_devid_vp = NULL;
}

static void
vdev_disk_rele(vdev_t *vd)
{
	ASSERT(spa_config_held(vd->vdev_spa, SCL_STATE, RW_WRITER));

	/* XXX: Implement me as a vnode rele for the device */
}

vdev_ops_t vdev_disk_ops = {
	vdev_disk_open,
	vdev_disk_close,
	vdev_default_asize,
	vdev_disk_io_start,
	vdev_disk_io_done,
	NULL,
	vdev_disk_hold,
	vdev_disk_rele,
	VDEV_TYPE_DISK,		/* name of this vdev type */
	B_TRUE			/* leaf vdev */
};

/*
 * Given the root disk device devid or pathname, read the label from
 * the device, and construct a configuration nvlist.
 */
int
vdev_disk_read_rootlabel(char *devpath, char *devid, nvlist_t **config)
{
	struct block_device *bdev;
	vdev_label_t *label;
	uint64_t s, size;
	int i;

	bdev = vdev_bdev_open(devpath, vdev_bdev_mode(FREAD), zfs_vdev_holder);
	if (IS_ERR(bdev))
		return (-PTR_ERR(bdev));

	s = bdev_capacity(bdev);
	if (s == 0) {
		vdev_bdev_close(bdev, vdev_bdev_mode(FREAD));
		return (EIO);
	}

	size = P2ALIGN_TYPED(s, sizeof (vdev_label_t), uint64_t);
	label = vmem_alloc(sizeof (vdev_label_t), KM_SLEEP);

	for (i = 0; i < VDEV_LABELS; i++) {
		uint64_t offset, state, txg = 0;

		/* read vdev label */
		offset = vdev_label_offset(size, i, 0);
		if (vdev_disk_physio(bdev, (caddr_t)label,
		    VDEV_SKIP_SIZE + VDEV_PHYS_SIZE, offset, READ_SYNC) != 0)
			continue;

		if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist,
		    sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0) {
			*config = NULL;
			continue;
		}

		if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE,
		    &state) != 0 || state >= POOL_STATE_DESTROYED) {
			nvlist_free(*config);
			*config = NULL;
			continue;
		}

		if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG,
		    &txg) != 0 || txg == 0) {
			nvlist_free(*config);
			*config = NULL;
			continue;
		}

		break;
	}

	vmem_free(label, sizeof (vdev_label_t));
	vdev_bdev_close(bdev, vdev_bdev_mode(FREAD));

	return (0);
}

module_param(zfs_vdev_scheduler, charp, 0644);
MODULE_PARM_DESC(zfs_vdev_scheduler, "I/O scheduler");