dm-raid.c 51.6 KB
Newer Older
NeilBrown's avatar
NeilBrown committed
1
2
/*
 * Copyright (C) 2010-2011 Neil Brown
3
 * Copyright (C) 2010-2016 Red Hat, Inc. All rights reserved.
NeilBrown's avatar
NeilBrown committed
4
5
6
7
8
 *
 * This file is released under the GPL.
 */

#include <linux/slab.h>
9
#include <linux/module.h>
NeilBrown's avatar
NeilBrown committed
10
11

#include "md.h"
12
#include "raid1.h"
NeilBrown's avatar
NeilBrown committed
13
#include "raid5.h"
14
#include "raid10.h"
NeilBrown's avatar
NeilBrown committed
15
16
#include "bitmap.h"

Alasdair G Kergon's avatar
Alasdair G Kergon committed
17
18
#include <linux/device-mapper.h>

NeilBrown's avatar
NeilBrown committed
19
#define DM_MSG_PREFIX "raid"
20
#define	MAX_RAID_DEVICES	253 /* md-raid kernel limit */
NeilBrown's avatar
NeilBrown committed
21

22
23
static bool devices_handle_discard_safely = false;

NeilBrown's avatar
NeilBrown committed
24
/*
25
26
 * The following flags are used by dm-raid.c to set up the array state.
 * They must be cleared before md_run is called.
NeilBrown's avatar
NeilBrown committed
27
 */
28
#define FirstUse 10             /* rdev flag */
NeilBrown's avatar
NeilBrown committed
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44

struct raid_dev {
	/*
	 * Two DM devices, one to hold metadata and one to hold the
	 * actual data/parity.  The reason for this is to not confuse
	 * ti->len and give more flexibility in altering size and
	 * characteristics.
	 *
	 * While it is possible for this device to be associated
	 * with a different physical device than the data_dev, it
	 * is intended for it to be the same.
	 *    |--------- Physical Device ---------|
	 *    |- meta_dev -|------ data_dev ------|
	 */
	struct dm_dev *meta_dev;
	struct dm_dev *data_dev;
45
	struct md_rdev rdev;
NeilBrown's avatar
NeilBrown committed
46
47
48
};

/*
Heinz Mauelshagen's avatar
Heinz Mauelshagen committed
49
 * Flags for rs->ctr_flags field.
50
51
52
 *
 * 1 = no flag value
 * 2 = flag with value
NeilBrown's avatar
NeilBrown committed
53
 */
54
55
56
57
58
59
60
61
62
63
64
65
#define CTR_FLAG_SYNC              0x1   /* 1 */ /* Not with raid0! */
#define CTR_FLAG_NOSYNC            0x2   /* 1 */ /* Not with raid0! */
#define CTR_FLAG_REBUILD           0x4   /* 2 */ /* Not with raid0! */
#define CTR_FLAG_DAEMON_SLEEP      0x8   /* 2 */ /* Not with raid0! */
#define CTR_FLAG_MIN_RECOVERY_RATE 0x10  /* 2 */ /* Not with raid0! */
#define CTR_FLAG_MAX_RECOVERY_RATE 0x20  /* 2 */ /* Not with raid0! */
#define CTR_FLAG_MAX_WRITE_BEHIND  0x40  /* 2 */ /* Only with raid1! */
#define CTR_FLAG_WRITE_MOSTLY      0x80  /* 2 */ /* Only with raid1! */
#define CTR_FLAG_STRIPE_CACHE      0x100 /* 2 */ /* Only with raid4/5/6! */
#define CTR_FLAG_REGION_SIZE       0x200 /* 2 */ /* Not with raid0! */
#define CTR_FLAG_RAID10_COPIES     0x400 /* 2 */ /* Only with raid10 */
#define CTR_FLAG_RAID10_FORMAT     0x800 /* 2 */ /* Only with raid10 */
66

NeilBrown's avatar
NeilBrown committed
67
68
69
struct raid_set {
	struct dm_target *ti;

70
	uint32_t bitmap_loaded;
Heinz Mauelshagen's avatar
Heinz Mauelshagen committed
71
	uint32_t ctr_flags;
NeilBrown's avatar
NeilBrown committed
72

73
	struct mddev md;
NeilBrown's avatar
NeilBrown committed
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
	struct raid_type *raid_type;
	struct dm_target_callbacks callbacks;

	struct raid_dev dev[0];
};

/* Supported raid types and properties. */
static struct raid_type {
	const char *name;		/* RAID algorithm. */
	const char *descr;		/* Descriptor text for logging. */
	const unsigned parity_devs;	/* # of parity devices. */
	const unsigned minimal_devs;	/* minimal # of devices in set. */
	const unsigned level;		/* RAID level. */
	const unsigned algorithm;	/* RAID algorithm. */
} raid_types[] = {
89
	{"raid0",    "RAID0 (striping)",                0, 2, 0, 0 /* NONE */},
90
	{"raid1",    "RAID1 (mirroring)",               0, 2, 1, 0 /* NONE */},
91
	{"raid10",   "RAID10 (striped mirrors)",        0, 2, 10, UINT_MAX /* Varies */},
NeilBrown's avatar
NeilBrown committed
92
93
94
95
96
97
98
99
100
101
	{"raid4",    "RAID4 (dedicated parity disk)",	1, 2, 5, ALGORITHM_PARITY_0},
	{"raid5_la", "RAID5 (left asymmetric)",		1, 2, 5, ALGORITHM_LEFT_ASYMMETRIC},
	{"raid5_ra", "RAID5 (right asymmetric)",	1, 2, 5, ALGORITHM_RIGHT_ASYMMETRIC},
	{"raid5_ls", "RAID5 (left symmetric)",		1, 2, 5, ALGORITHM_LEFT_SYMMETRIC},
	{"raid5_rs", "RAID5 (right symmetric)",		1, 2, 5, ALGORITHM_RIGHT_SYMMETRIC},
	{"raid6_zr", "RAID6 (zero restart)",		2, 4, 6, ALGORITHM_ROTATING_ZERO_RESTART},
	{"raid6_nr", "RAID6 (N restart)",		2, 4, 6, ALGORITHM_ROTATING_N_RESTART},
	{"raid6_nc", "RAID6 (N continue)",		2, 4, 6, ALGORITHM_ROTATING_N_CONTINUE}
};

102
103
104
105
106
107
/* True, if @v is in inclusive range [@min, @max] */
static bool _in_range(long v, long min, long max)
{
	return v >= min && v <= max;
}

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
/* ctr flag bit manipulation... */
/* Set single @flag in @flags */
static void _set_flag(uint32_t flag, uint32_t *flags)
{
	WARN_ON_ONCE(hweight32(flag) != 1);
	*flags |= flag;
}

/* Test single @flag in @flags */
static bool _test_flag(uint32_t flag, uint32_t flags)
{
	WARN_ON_ONCE(hweight32(flag) != 1);
	return (flag & flags) ? true : false;
}

/* Return true if single @flag is set in @*flags, else set it and return false */
static bool _test_and_set_flag(uint32_t flag, uint32_t *flags)
{
	if (_test_flag(flag, *flags))
		return true;

	_set_flag(flag, flags);
	return false;
}
/* ...ctr and runtime flag bit manipulation */

/* All table line arguments are defined here */
static struct arg_name_flag {
	const uint32_t flag;
	const char *name;
} _arg_name_flags[] = {
	{ CTR_FLAG_SYNC, "sync"},
	{ CTR_FLAG_NOSYNC, "nosync"},
	{ CTR_FLAG_REBUILD, "rebuild"},
	{ CTR_FLAG_DAEMON_SLEEP, "daemon_sleep"},
	{ CTR_FLAG_MIN_RECOVERY_RATE, "min_recovery_rate"},
	{ CTR_FLAG_MAX_RECOVERY_RATE, "max_recovery_rate"},
	{ CTR_FLAG_MAX_WRITE_BEHIND, "max_write_behind"},
	{ CTR_FLAG_WRITE_MOSTLY, "writemostly"},
	{ CTR_FLAG_STRIPE_CACHE, "stripe_cache"},
	{ CTR_FLAG_REGION_SIZE, "region_size"},
	{ CTR_FLAG_RAID10_COPIES, "raid10_copies"},
	{ CTR_FLAG_RAID10_FORMAT, "raid10_format"},
};

/* Return argument name string for given @flag */
static const char *_argname_by_flag(const uint32_t flag)
{
	if (hweight32(flag) == 1) {
		struct arg_name_flag *anf = _arg_name_flags + ARRAY_SIZE(_arg_name_flags);

		while (anf-- > _arg_name_flags)
			if (_test_flag(flag, anf->flag))
				return anf->name;

	} else
		DMERR("%s called with more than one flag!", __func__);

	return NULL;
}

/*
 * Convenience functions to set ti->error to @errmsg and
 * return @r in order to shorten code in a lot of places
 */
static int ti_error_ret(struct dm_target *ti, const char *errmsg, int r)
{
	ti->error = (char *) errmsg;
	return r;
}

static int ti_error_einval(struct dm_target *ti, const char *errmsg)
{
	return ti_error_ret(ti, errmsg, -EINVAL);
}
/* END: convenience functions to set ti->error to @errmsg... */

185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
static char *raid10_md_layout_to_format(int layout)
{
	/*
	 * Bit 16 and 17 stand for "offset" and "use_far_sets"
	 * Refer to MD's raid10.c for details
	 */
	if ((layout & 0x10000) && (layout & 0x20000))
		return "offset";

	if ((layout & 0xFF) > 1)
		return "near";

	return "far";
}

200
201
static unsigned raid10_md_layout_to_copies(int layout)
{
202
203
204
	if ((layout & 0xFF) > 1)
		return layout & 0xFF;
	return (layout >> 8) & 0xFF;
205
206
207
208
}

static int raid10_format_to_md_layout(char *format, unsigned copies)
{
209
210
	unsigned n = 1, f = 1;

Heinz Mauelshagen's avatar
Heinz Mauelshagen committed
211
	if (!strcasecmp("near", format))
212
213
214
215
		n = copies;
	else
		f = copies;

Heinz Mauelshagen's avatar
Heinz Mauelshagen committed
216
	if (!strcasecmp("offset", format))
217
218
		return 0x30000 | (f << 8) | n;

Heinz Mauelshagen's avatar
Heinz Mauelshagen committed
219
	if (!strcasecmp("far", format))
220
221
222
		return 0x20000 | (f << 8) | n;

	return (f << 8) | n;
223
224
}

225
static struct raid_type *get_raid_type(const char *name)
NeilBrown's avatar
NeilBrown committed
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
{
	int i;

	for (i = 0; i < ARRAY_SIZE(raid_types); i++)
		if (!strcmp(raid_types[i].name, name))
			return &raid_types[i];

	return NULL;
}

static struct raid_set *context_alloc(struct dm_target *ti, struct raid_type *raid_type, unsigned raid_devs)
{
	unsigned i;
	struct raid_set *rs;

241
242
	if (raid_devs <= raid_type->parity_devs)
		return ERR_PTR(ti_error_einval(ti, "Insufficient number of devices"));
NeilBrown's avatar
NeilBrown committed
243
244

	rs = kzalloc(sizeof(*rs) + raid_devs * sizeof(rs->dev[0]), GFP_KERNEL);
245
246
	if (!rs)
		return ERR_PTR(ti_error_ret(ti, "Cannot allocate raid context", -ENOMEM));
NeilBrown's avatar
NeilBrown committed
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268

	mddev_init(&rs->md);

	rs->ti = ti;
	rs->raid_type = raid_type;
	rs->md.raid_disks = raid_devs;
	rs->md.level = raid_type->level;
	rs->md.new_level = rs->md.level;
	rs->md.layout = raid_type->algorithm;
	rs->md.new_layout = rs->md.layout;
	rs->md.delta_disks = 0;
	rs->md.recovery_cp = 0;

	for (i = 0; i < raid_devs; i++)
		md_rdev_init(&rs->dev[i].rdev);

	/*
	 * Remaining items to be initialized by further RAID params:
	 *  rs->md.persistent
	 *  rs->md.external
	 *  rs->md.chunk_sectors
	 *  rs->md.new_chunk_sectors
269
	 *  rs->md.dev_sectors
NeilBrown's avatar
NeilBrown committed
270
271
272
273
274
275
276
277
278
	 */

	return rs;
}

static void context_free(struct raid_set *rs)
{
	int i;

279
280
281
	for (i = 0; i < rs->md.raid_disks; i++) {
		if (rs->dev[i].meta_dev)
			dm_put_device(rs->ti, rs->dev[i].meta_dev);
282
		md_rdev_clear(&rs->dev[i].rdev);
NeilBrown's avatar
NeilBrown committed
283
284
		if (rs->dev[i].data_dev)
			dm_put_device(rs->ti, rs->dev[i].data_dev);
285
	}
NeilBrown's avatar
NeilBrown committed
286
287
288
289
290
291
292
293
294

	kfree(rs);
}

/*
 * For every device we have two words
 *  <meta_dev>: meta device name or '-' if missing
 *  <data_dev>: data device name or '-' if missing
 *
295
296
297
298
299
300
301
302
303
304
 * The following are permitted:
 *    - -
 *    - <data_dev>
 *    <meta_dev> <data_dev>
 *
 * The following is not allowed:
 *    <meta_dev> -
 *
 * This code parses those words.  If there is a failure,
 * the caller must use context_free to unwind the operations.
NeilBrown's avatar
NeilBrown committed
305
 */
306
static int parse_dev_params(struct raid_set *rs, struct dm_arg_set *as)
NeilBrown's avatar
NeilBrown committed
307
308
309
310
{
	int i;
	int rebuild = 0;
	int metadata_available = 0;
311
	int r = 0;
312
	const char *arg;
NeilBrown's avatar
NeilBrown committed
313

314
315
316
317
318
319
	/* Put off the number of raid devices argument to get to dev pairs */
	arg = dm_shift_arg(as);
	if (!arg)
		return -EINVAL;

	for (i = 0; i < rs->md.raid_disks; i++) {
NeilBrown's avatar
NeilBrown committed
320
321
322
323
324
325
326
327
328
329
330
331
		rs->dev[i].rdev.raid_disk = i;

		rs->dev[i].meta_dev = NULL;
		rs->dev[i].data_dev = NULL;

		/*
		 * There are no offsets, since there is a separate device
		 * for data and metadata.
		 */
		rs->dev[i].rdev.data_offset = 0;
		rs->dev[i].rdev.mddev = &rs->md;

332
333
334
335
336
337
		arg = dm_shift_arg(as);
		if (!arg)
			return -EINVAL;

		if (strcmp(arg, "-")) {
			r = dm_get_device(rs->ti, arg,
338
339
					    dm_table_get_mode(rs->ti->table),
					    &rs->dev[i].meta_dev);
340
			if (r)
341
				return ti_error_ret(rs->ti, "RAID metadata device lookup failure", r);
342
343
344

			rs->dev[i].rdev.sb_page = alloc_page(GFP_KERNEL);
			if (!rs->dev[i].rdev.sb_page)
345
				return ti_error_ret(rs->ti, "Failed to allocate superblock page", -ENOMEM);
NeilBrown's avatar
NeilBrown committed
346
347
		}

348
349
350
351
352
		arg = dm_shift_arg(as);
		if (!arg)
			return -EINVAL;

		if (!strcmp(arg, "-")) {
NeilBrown's avatar
NeilBrown committed
353
			if (!test_bit(In_sync, &rs->dev[i].rdev.flags) &&
354
355
			    (!rs->dev[i].rdev.recovery_offset))
				return ti_error_einval(rs->ti, "Drive designated for rebuild not specified");
NeilBrown's avatar
NeilBrown committed
356

357
			if (rs->dev[i].meta_dev)
358
				return ti_error_einval(rs->ti, "No data device supplied with metadata device");
359

NeilBrown's avatar
NeilBrown committed
360
361
362
			continue;
		}

363
		r = dm_get_device(rs->ti, arg,
NeilBrown's avatar
NeilBrown committed
364
365
				    dm_table_get_mode(rs->ti->table),
				    &rs->dev[i].data_dev);
366
367
		if (r)
			return ti_error_ret(rs->ti, "RAID device lookup failure", r);
NeilBrown's avatar
NeilBrown committed
368

369
370
371
372
		if (rs->dev[i].meta_dev) {
			metadata_available = 1;
			rs->dev[i].rdev.meta_bdev = rs->dev[i].meta_dev->bdev;
		}
NeilBrown's avatar
NeilBrown committed
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
		rs->dev[i].rdev.bdev = rs->dev[i].data_dev->bdev;
		list_add(&rs->dev[i].rdev.same_set, &rs->md.disks);
		if (!test_bit(In_sync, &rs->dev[i].rdev.flags))
			rebuild++;
	}

	if (metadata_available) {
		rs->md.external = 0;
		rs->md.persistent = 1;
		rs->md.major_version = 2;
	} else if (rebuild && !rs->md.recovery_cp) {
		/*
		 * Without metadata, we will not be able to tell if the array
		 * is in-sync or not - we must assume it is not.  Therefore,
		 * it is impossible to rebuild a drive.
		 *
		 * Even if there is metadata, the on-disk information may
		 * indicate that the array is not in-sync and it will then
		 * fail at that time.
		 *
		 * User could specify 'nosync' option if desperate.
		 */
		DMERR("Unable to rebuild drive while array is not in-sync");
396
		return ti_error_einval(rs->ti, "Unable to rebuild drive while array is not in-sync");
NeilBrown's avatar
NeilBrown committed
397
398
399
400
401
	}

	return 0;
}

402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
/*
 * validate_region_size
 * @rs
 * @region_size:  region size in sectors.  If 0, pick a size (4MiB default).
 *
 * Set rs->md.bitmap_info.chunksize (which really refers to 'region size').
 * Ensure that (ti->len/region_size < 2^21) - required by MD bitmap.
 *
 * Returns: 0 on success, -EINVAL on failure.
 */
static int validate_region_size(struct raid_set *rs, unsigned long region_size)
{
	unsigned long min_region_size = rs->ti->len / (1 << 21);

	if (!region_size) {
		/*
		 * Choose a reasonable default.  All figures in sectors.
		 */
		if (min_region_size > (1 << 13)) {
421
			/* If not a power of 2, make it the next power of 2 */
422
			region_size = roundup_pow_of_two(min_region_size);
423
424
425
426
427
428
429
430
431
432
			DMINFO("Choosing default region size of %lu sectors",
			       region_size);
		} else {
			DMINFO("Choosing default region size of 4MiB");
			region_size = 1 << 13; /* sectors */
		}
	} else {
		/*
		 * Validate user-supplied value.
		 */
433
434
		if (region_size > rs->ti->len)
			return ti_error_einval(rs->ti, "Supplied region size is too large");
435
436
437
438

		if (region_size < min_region_size) {
			DMERR("Supplied region_size (%lu sectors) below minimum (%lu)",
			      region_size, min_region_size);
439
			return ti_error_einval(rs->ti, "Supplied region size is too small");
440
441
		}

442
443
		if (!is_power_of_2(region_size))
			return ti_error_einval(rs->ti, "Region size is not a power of 2");
444

445
446
		if (region_size < rs->md.chunk_sectors)
			return ti_error_einval(rs->ti, "Region size is smaller than the chunk size");
447
448
449
450
451
452
453
454
455
456
	}

	/*
	 * Convert sectors to bytes.
	 */
	rs->md.bitmap_info.chunksize = (region_size << 9);

	return 0;
}

457
/*
458
 * validate_raid_redundancy
459
460
 * @rs
 *
461
462
 * Determine if there are enough devices in the array that haven't
 * failed (or are being rebuilt) to form a usable array.
463
464
465
 *
 * Returns: 0 on success, -EINVAL on failure.
 */
466
static int validate_raid_redundancy(struct raid_set *rs)
467
468
{
	unsigned i, rebuild_cnt = 0;
469
	unsigned rebuilds_per_group = 0, copies, d;
470
	unsigned group_size, last_group_start;
471
472

	for (i = 0; i < rs->md.raid_disks; i++)
473
474
		if (!test_bit(In_sync, &rs->dev[i].rdev.flags) ||
		    !rs->dev[i].rdev.sb_page)
475
476
477
478
479
480
481
482
483
484
485
486
487
488
			rebuild_cnt++;

	switch (rs->raid_type->level) {
	case 1:
		if (rebuild_cnt >= rs->md.raid_disks)
			goto too_many;
		break;
	case 4:
	case 5:
	case 6:
		if (rebuild_cnt > rs->raid_type->parity_devs)
			goto too_many;
		break;
	case 10:
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
		copies = raid10_md_layout_to_copies(rs->md.layout);
		if (rebuild_cnt < copies)
			break;

		/*
		 * It is possible to have a higher rebuild count for RAID10,
		 * as long as the failed devices occur in different mirror
		 * groups (i.e. different stripes).
		 *
		 * When checking "near" format, make sure no adjacent devices
		 * have failed beyond what can be handled.  In addition to the
		 * simple case where the number of devices is a multiple of the
		 * number of copies, we must also handle cases where the number
		 * of devices is not a multiple of the number of copies.
		 * E.g.    dev1 dev2 dev3 dev4 dev5
		 *          A    A    B    B    C
		 *          C    D    D    E    E
		 */
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
		if (!strcmp("near", raid10_md_layout_to_format(rs->md.layout))) {
			for (i = 0; i < rs->md.raid_disks * copies; i++) {
				if (!(i % copies))
					rebuilds_per_group = 0;
				d = i % rs->md.raid_disks;
				if ((!rs->dev[d].rdev.sb_page ||
				     !test_bit(In_sync, &rs->dev[d].rdev.flags)) &&
				    (++rebuilds_per_group >= copies))
					goto too_many;
			}
			break;
		}

		/*
		 * When checking "far" and "offset" formats, we need to ensure
		 * that the device that holds its copy is not also dead or
		 * being rebuilt.  (Note that "far" and "offset" formats only
		 * support two copies right now.  These formats also only ever
		 * use the 'use_far_sets' variant.)
		 *
		 * This check is somewhat complicated by the need to account
		 * for arrays that are not a multiple of (far) copies.  This
		 * results in the need to treat the last (potentially larger)
		 * set differently.
		 */
		group_size = (rs->md.raid_disks / copies);
		last_group_start = (rs->md.raid_disks / group_size) - 1;
		last_group_start *= group_size;
		for (i = 0; i < rs->md.raid_disks; i++) {
			if (!(i % copies) && !(i > last_group_start))
537
				rebuilds_per_group = 0;
538
539
			if ((!rs->dev[i].rdev.sb_page ||
			     !test_bit(In_sync, &rs->dev[i].rdev.flags)) &&
540
			    (++rebuilds_per_group >= copies))
541
					goto too_many;
542
543
		}
		break;
544
	default:
545
546
		if (rebuild_cnt)
			return -EINVAL;
547
548
549
550
551
552
553
554
	}

	return 0;

too_many:
	return -EINVAL;
}

NeilBrown's avatar
NeilBrown committed
555
556
557
558
/*
 * Possible arguments are...
 *	<chunk_size> [optional_args]
 *
559
560
561
562
563
 * Argument definitions
 *    <chunk_size>			The number of sectors per disk that
 *                                      will form the "stripe"
 *    [[no]sync]			Force or prevent recovery of the
 *                                      entire array
NeilBrown's avatar
NeilBrown committed
564
 *    [rebuild <idx>]			Rebuild the drive indicated by the index
565
566
 *    [daemon_sleep <ms>]		Time between bitmap daemon work to
 *                                      clear bits
NeilBrown's avatar
NeilBrown committed
567
568
 *    [min_recovery_rate <kB/sec/disk>]	Throttle RAID initialization
 *    [max_recovery_rate <kB/sec/disk>]	Throttle RAID initialization
569
 *    [write_mostly <idx>]		Indicate a write mostly drive via index
NeilBrown's avatar
NeilBrown committed
570
571
 *    [max_write_behind <sectors>]	See '-write-behind=' (man mdadm)
 *    [stripe_cache <sectors>]		Stripe cache size for higher RAIDs
572
 *    [region_size <sectors>]           Defines granularity of bitmap
573
574
575
 *
 * RAID10-only options:
 *    [raid10_copies <# copies>]        Number of copies.  (Default: 2)
576
 *    [raid10_format <near|far|offset>] Layout algorithm.  (Default: near)
NeilBrown's avatar
NeilBrown committed
577
 */
578
static int parse_raid_params(struct raid_set *rs, struct dm_arg_set *as,
NeilBrown's avatar
NeilBrown committed
579
580
			     unsigned num_raid_params)
{
581
582
	char *raid10_format = "near";
	unsigned raid10_copies = 2;
583
	unsigned i;
584
	unsigned value, region_size = 0;
585
	sector_t sectors_per_dev = rs->ti->len;
586
	sector_t max_io_len;
587
	const char *arg, *key;
588
	struct raid_dev *rd;
589
590
591
592

	arg = dm_shift_arg(as);
	num_raid_params--; /* Account for chunk_size argument */

593
594
	if (kstrtouint(arg, 10, &value) < 0)
		return ti_error_einval(rs->ti, "Bad numerical argument given for chunk_size");
NeilBrown's avatar
NeilBrown committed
595
596
597

	/*
	 * First, parse the in-order required arguments
598
	 * "chunk_size" is the only argument of this type.
NeilBrown's avatar
NeilBrown committed
599
	 */
600
	if (rs->raid_type->level == 1) {
601
602
603
		if (value)
			DMERR("Ignoring chunk size parameter for RAID 1");
		value = 0;
604
605
606
607
	} else if (!is_power_of_2(value))
		return ti_error_einval(rs->ti, "Chunk size must be a power of 2");
	else if (value < 8)
		return ti_error_einval(rs->ti, "Chunk size value is too small");
NeilBrown's avatar
NeilBrown committed
608
609
610
611

	rs->md.new_chunk_sectors = rs->md.chunk_sectors = value;

	/*
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
	 * We set each individual device as In_sync with a completed
	 * 'recovery_offset'.  If there has been a device failure or
	 * replacement then one of the following cases applies:
	 *
	 *   1) User specifies 'rebuild'.
	 *      - Device is reset when param is read.
	 *   2) A new device is supplied.
	 *      - No matching superblock found, resets device.
	 *   3) Device failure was transient and returns on reload.
	 *      - Failure noticed, resets device for bitmap replay.
	 *   4) Device hadn't completed recovery after previous failure.
	 *      - Superblock is read and overrides recovery_offset.
	 *
	 * What is found in the superblocks of the devices is always
	 * authoritative, unless 'rebuild' or '[no]sync' was specified.
NeilBrown's avatar
NeilBrown committed
627
	 */
628
	for (i = 0; i < rs->md.raid_disks; i++) {
NeilBrown's avatar
NeilBrown committed
629
		set_bit(In_sync, &rs->dev[i].rdev.flags);
630
631
		rs->dev[i].rdev.recovery_offset = MaxSector;
	}
NeilBrown's avatar
NeilBrown committed
632

633
634
635
	/*
	 * Second, parse the unordered optional arguments
	 */
NeilBrown's avatar
NeilBrown committed
636
	for (i = 0; i < num_raid_params; i++) {
637
		arg = dm_shift_arg(as);
638
639
		if (!arg)
			return ti_error_einval(rs->ti, "Not enough raid parameters given");
640
641

		if (!strcasecmp(arg, "nosync")) {
NeilBrown's avatar
NeilBrown committed
642
			rs->md.recovery_cp = MaxSector;
643
			_set_flag(CTR_FLAG_NOSYNC, &rs->ctr_flags);
NeilBrown's avatar
NeilBrown committed
644
645
			continue;
		}
646
		if (!strcasecmp(arg, "sync")) {
NeilBrown's avatar
NeilBrown committed
647
			rs->md.recovery_cp = 0;
648
			_set_flag(CTR_FLAG_SYNC, &rs->ctr_flags);
NeilBrown's avatar
NeilBrown committed
649
650
651
			continue;
		}

652
653
654
		key = arg;
		arg = dm_shift_arg(as);
		i++; /* Account for the argument pairs */
655
656
		if (!arg)
			return ti_error_einval(rs->ti, "Wrong number of raid parameters given");
657

658
659
660
661
662
663
664
665
666
		/*
		 * Parameters that take a string value are checked here.
		 */

		if (!strcasecmp(key, _argname_by_flag(CTR_FLAG_RAID10_FORMAT))) {
			if (_test_and_set_flag(CTR_FLAG_RAID10_FORMAT, &rs->ctr_flags))
				return ti_error_einval(rs->ti, "Only one raid10_format argument pair allowed");
			if (rs->raid_type->level != 10)
				return ti_error_einval(rs->ti, "'raid10_format' is an invalid parameter for this RAID type");
667
668
			if (strcmp("near", arg) &&
			    strcmp("far", arg) &&
669
670
671
			    strcmp("offset", arg))
				return ti_error_einval(rs->ti, "Invalid 'raid10_format' value given");

672
			raid10_format = (char *) arg;
673
674
675
			continue;
		}

676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
		if (kstrtouint(arg, 10, &value) < 0)
			return ti_error_einval(rs->ti, "Bad numerical argument given in raid params");

		if (!strcasecmp(key, _argname_by_flag(CTR_FLAG_REBUILD))) {
			/*
			 * "rebuild" is being passed in by userspace to provide
			 * indexes of replaced devices and to set up additional
			 * devices on raid level takeover.
 			 */
			if (!_in_range(value, 0, rs->md.raid_disks - 1))
				return ti_error_einval(rs->ti, "Invalid rebuild index given");

			rd = rs->dev + value;
			clear_bit(In_sync, &rd->rdev.flags);
			clear_bit(Faulty, &rd->rdev.flags);
			rd->rdev.recovery_offset = 0;
			_set_flag(CTR_FLAG_REBUILD, &rs->ctr_flags);
		} else if (!strcasecmp(key, _argname_by_flag(CTR_FLAG_WRITE_MOSTLY))) {
			if (rs->raid_type->level != 1)
				return ti_error_einval(rs->ti, "write_mostly option is only valid for RAID1");

			if (!_in_range(value, 0, rs->md.raid_disks - 1))
				return ti_error_einval(rs->ti, "Invalid write_mostly index given");
NeilBrown's avatar
NeilBrown committed
699

700
			set_bit(WriteMostly, &rs->dev[value].rdev.flags);
701
702
703
704
705
706
707
			_set_flag(CTR_FLAG_WRITE_MOSTLY, &rs->ctr_flags);
		} else if (!strcasecmp(key, _argname_by_flag(CTR_FLAG_MAX_WRITE_BEHIND))) {
			if (rs->raid_type->level != 1)
				return ti_error_einval(rs->ti, "max_write_behind option is only valid for RAID1");

			if (_test_and_set_flag(CTR_FLAG_MAX_WRITE_BEHIND, &rs->ctr_flags))
				return ti_error_einval(rs->ti, "Only one max_write_behind argument pair allowed");
NeilBrown's avatar
NeilBrown committed
708
709
710
711
712
713

			/*
			 * In device-mapper, we specify things in sectors, but
			 * MD records this value in kB
			 */
			value /= 2;
714
715
716
			if (value > COUNTER_MAX)
				return ti_error_einval(rs->ti, "Max write-behind limit out of range");

NeilBrown's avatar
NeilBrown committed
717
			rs->md.bitmap_info.max_write_behind = value;
718
719
720
721
722
		} else if (!strcasecmp(key, _argname_by_flag(CTR_FLAG_DAEMON_SLEEP))) {
			if (_test_and_set_flag(CTR_FLAG_DAEMON_SLEEP, &rs->ctr_flags))
				return ti_error_einval(rs->ti, "Only one daemon_sleep argument pair allowed");
			if (!value || (value > MAX_SCHEDULE_TIMEOUT))
				return ti_error_einval(rs->ti, "daemon sleep period out of range");
NeilBrown's avatar
NeilBrown committed
723
			rs->md.bitmap_info.daemon_sleep = value;
724
725
726
		} else if (!strcasecmp(key, _argname_by_flag(CTR_FLAG_STRIPE_CACHE))) {
			if (_test_and_set_flag(CTR_FLAG_STRIPE_CACHE, &rs->ctr_flags))
				return ti_error_einval(rs->ti, "Only one stripe_cache argument pair allowed");
NeilBrown's avatar
NeilBrown committed
727
728
729
730
731
732
			/*
			 * In device-mapper, we specify things in sectors, but
			 * MD records this value in kB
			 */
			value /= 2;

733
734
735
736
737
738
739
740
741
742
			if (!_in_range(rs->raid_type->level, 4, 6))
				return ti_error_einval(rs->ti, "Inappropriate argument: stripe_cache");
			if (raid5_set_cache_size(&rs->md, (int)value))
				return ti_error_einval(rs->ti, "Bad stripe_cache size");

		} else if (!strcasecmp(key, _argname_by_flag(CTR_FLAG_MIN_RECOVERY_RATE))) {
			if (_test_and_set_flag(CTR_FLAG_MIN_RECOVERY_RATE, &rs->ctr_flags))
				return ti_error_einval(rs->ti, "Only one min_recovery_rate argument pair allowed");
			if (value > INT_MAX)
				return ti_error_einval(rs->ti, "min_recovery_rate out of range");
NeilBrown's avatar
NeilBrown committed
743
			rs->md.sync_speed_min = (int)value;
744
745
746
747
748
		} else if (!strcasecmp(key, _argname_by_flag(CTR_FLAG_MAX_RECOVERY_RATE))) {
			if (_test_and_set_flag(CTR_FLAG_MIN_RECOVERY_RATE, &rs->ctr_flags))
				return ti_error_einval(rs->ti, "Only one max_recovery_rate argument pair allowed");
			if (value > INT_MAX)
				return ti_error_einval(rs->ti, "max_recovery_rate out of range");
NeilBrown's avatar
NeilBrown committed
749
			rs->md.sync_speed_max = (int)value;
750
751
752
753
		} else if (!strcasecmp(key, _argname_by_flag(CTR_FLAG_REGION_SIZE))) {
			if (_test_and_set_flag(CTR_FLAG_REGION_SIZE, &rs->ctr_flags))
				return ti_error_einval(rs->ti, "Only one region_size argument pair allowed");

754
			region_size = value;
755
756
757
758
759
760
761
		} else if (!strcasecmp(key, _argname_by_flag(CTR_FLAG_RAID10_COPIES))) {
			if (_test_and_set_flag(CTR_FLAG_RAID10_COPIES, &rs->ctr_flags))
				return ti_error_einval(rs->ti, "Only one raid10_copies argument pair allowed");

			if (!_in_range(value, 2, rs->md.raid_disks))
				return ti_error_einval(rs->ti, "Bad value for 'raid10_copies'");

762
			raid10_copies = value;
NeilBrown's avatar
NeilBrown committed
763
764
		} else {
			DMERR("Unable to parse RAID parameter: %s", key);
765
			return ti_error_einval(rs->ti, "Unable to parse RAID parameters");
NeilBrown's avatar
NeilBrown committed
766
767
768
		}
	}

769
770
771
772
	if (validate_region_size(rs, region_size))
		return -EINVAL;

	if (rs->md.chunk_sectors)
773
		max_io_len = rs->md.chunk_sectors;
774
	else
775
		max_io_len = region_size;
776

777
778
	if (dm_set_target_max_io_len(rs->ti, max_io_len))
		return -EINVAL;
779

780
	if (rs->raid_type->level == 10) {
781
782
		if (raid10_copies > rs->md.raid_disks)
			return ti_error_einval(rs->ti, "Not enough devices to satisfy specification");
783

784
785
786
787
		/*
		 * If the format is not "near", we only support
		 * two copies at the moment.
		 */
788
789
		if (strcmp("near", raid10_format) && (raid10_copies > 2))
			return ti_error_einval(rs->ti, "Too many copies for given RAID10 format.");
790

791
792
793
794
795
796
797
		/* (Len * #mirrors) / #devices */
		sectors_per_dev = rs->ti->len * raid10_copies;
		sector_div(sectors_per_dev, rs->md.raid_disks);

		rs->md.layout = raid10_format_to_md_layout(raid10_format,
							   raid10_copies);
		rs->md.new_layout = rs->md.layout;
798
	} else if ((!rs->raid_type->level || rs->raid_type->level > 1) &&
799
		   sector_div(sectors_per_dev,
800
801
802
			      (rs->md.raid_disks - rs->raid_type->parity_devs)))
		return ti_error_einval(rs->ti, "Target length not divisible by number of data devices");

803
804
	rs->md.dev_sectors = sectors_per_dev;

NeilBrown's avatar
NeilBrown committed
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
	/* Assume there are no metadata devices until the drives are parsed */
	rs->md.persistent = 0;
	rs->md.external = 1;

	return 0;
}

static void do_table_event(struct work_struct *ws)
{
	struct raid_set *rs = container_of(ws, struct raid_set, md.event_work);

	dm_table_event(rs->ti->table);
}

static int raid_is_congested(struct dm_target_callbacks *cb, int bits)
{
	struct raid_set *rs = container_of(cb, struct raid_set, callbacks);

823
	return mddev_congested(&rs->md, bits);
NeilBrown's avatar
NeilBrown committed
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
/*
 * This structure is never routinely used by userspace, unlike md superblocks.
 * Devices with this superblock should only ever be accessed via device-mapper.
 */
#define DM_RAID_MAGIC 0x64526D44
struct dm_raid_superblock {
	__le32 magic;		/* "DmRd" */
	__le32 features;	/* Used to indicate possible future changes */

	__le32 num_devices;	/* Number of devices in this array. (Max 64) */
	__le32 array_position;	/* The position of this drive in the array */

	__le64 events;		/* Incremented by md when superblock updated */
	__le64 failed_devices;	/* Bit field of devices to indicate failures */

	/*
	 * This offset tracks the progress of the repair or replacement of
	 * an individual drive.
	 */
	__le64 disk_recovery_offset;

	/*
	 * This offset tracks the progress of the initial array
	 * synchronisation/parity calculation.
	 */
	__le64 array_resync_offset;

	/*
	 * RAID characteristics
	 */
	__le32 level;
	__le32 layout;
	__le32 stripe_sectors;

860
	/* Remainder of a logical block is zero-filled when writing (see super_sync()). */
861
862
} __packed;

863
static int read_disk_sb(struct md_rdev *rdev, int size)
864
865
866
867
868
869
{
	BUG_ON(!rdev->sb_page);

	if (rdev->sb_loaded)
		return 0;

Mike Christie's avatar
Mike Christie committed
870
	if (!sync_page_io(rdev, 0, size, rdev->sb_page, REQ_OP_READ, 0, 1)) {
871
872
		DMERR("Failed to read superblock of device at position %d",
		      rdev->raid_disk);
873
		md_error(rdev->mddev, rdev);
874
875
876
877
878
879
880
881
		return -EINVAL;
	}

	rdev->sb_loaded = 1;

	return 0;
}

882
static void super_sync(struct mddev *mddev, struct md_rdev *rdev)
883
{
884
	int i;
885
886
	uint64_t failed_devices;
	struct dm_raid_superblock *sb;
887
	struct raid_set *rs = container_of(mddev, struct raid_set, md);
888
889
890
891

	sb = page_address(rdev->sb_page);
	failed_devices = le64_to_cpu(sb->failed_devices);

892
893
894
895
	for (i = 0; i < mddev->raid_disks; i++)
		if (!rs->dev[i].data_dev ||
		    test_bit(Faulty, &(rs->dev[i].rdev.flags)))
			failed_devices |= (1ULL << i);
896

897
	memset(sb + 1, 0, rdev->sb_size - sizeof(*sb));
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

	sb->magic = cpu_to_le32(DM_RAID_MAGIC);
	sb->features = cpu_to_le32(0);	/* No features yet */

	sb->num_devices = cpu_to_le32(mddev->raid_disks);
	sb->array_position = cpu_to_le32(rdev->raid_disk);

	sb->events = cpu_to_le64(mddev->events);
	sb->failed_devices = cpu_to_le64(failed_devices);

	sb->disk_recovery_offset = cpu_to_le64(rdev->recovery_offset);
	sb->array_resync_offset = cpu_to_le64(mddev->recovery_cp);

	sb->level = cpu_to_le32(mddev->level);
	sb->layout = cpu_to_le32(mddev->layout);
	sb->stripe_sectors = cpu_to_le32(mddev->chunk_sectors);
}

/*
 * super_load
 *
 * This function creates a superblock if one is not found on the device
 * and will decide which superblock to use if there's a choice.
 *
 * Return: 1 if use rdev, 0 if use refdev, -Exxx otherwise
 */
924
static int super_load(struct md_rdev *rdev, struct md_rdev *refdev)
925
{
926
	int r;
927
928
929
930
931
	struct dm_raid_superblock *sb;
	struct dm_raid_superblock *refsb;
	uint64_t events_sb, events_refsb;

	rdev->sb_start = 0;
932
933
934
935
936
	rdev->sb_size = bdev_logical_block_size(rdev->meta_bdev);
	if (rdev->sb_size < sizeof(*sb) || rdev->sb_size > PAGE_SIZE) {
		DMERR("superblock size of a logical block is no longer valid");
		return -EINVAL;
	}
937

938
939
940
	r = read_disk_sb(rdev, rdev->sb_size);
	if (r)
		return r;
941
942

	sb = page_address(rdev->sb_page);
943
944
945
946
947
948
949
950

	/*
	 * Two cases that we want to write new superblocks and rebuild:
	 * 1) New device (no matching magic number)
	 * 2) Device specified for rebuild (!In_sync w/ offset == 0)
	 */
	if ((sb->magic != cpu_to_le32(DM_RAID_MAGIC)) ||
	    (!test_bit(In_sync, &rdev->flags) && !rdev->recovery_offset)) {
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
		super_sync(rdev->mddev, rdev);

		set_bit(FirstUse, &rdev->flags);

		/* Force writing of superblocks to disk */
		set_bit(MD_CHANGE_DEVS, &rdev->mddev->flags);

		/* Any superblock is better than none, choose that if given */
		return refdev ? 0 : 1;
	}

	if (!refdev)
		return 1;

	events_sb = le64_to_cpu(sb->events);

	refsb = page_address(refdev->sb_page);
	events_refsb = le64_to_cpu(refsb->events);

	return (events_sb > events_refsb) ? 1 : 0;
}

973
static int super_init_validation(struct mddev *mddev, struct md_rdev *rdev)
974
975
976
977
978
979
980
981
{
	int role;
	struct raid_set *rs = container_of(mddev, struct raid_set, md);
	uint64_t events_sb;
	uint64_t failed_devices;
	struct dm_raid_superblock *sb;
	uint32_t new_devs = 0;
	uint32_t rebuilds = 0;
NeilBrown's avatar
NeilBrown committed
982
	struct md_rdev *r;
983
984
985
986
987
988
989
990
991
992
993
994
995
996
	struct dm_raid_superblock *sb2;

	sb = page_address(rdev->sb_page);
	events_sb = le64_to_cpu(sb->events);
	failed_devices = le64_to_cpu(sb->failed_devices);

	/*
	 * Initialise to 1 if this is a new superblock.
	 */
	mddev->events = events_sb ? : 1;

	/*
	 * Reshaping is not currently allowed
	 */
997
998
999
1000
	if (le32_to_cpu(sb->level) != mddev->level) {
		DMERR("Reshaping arrays not yet supported. (RAID level change)");
		return -EINVAL;
	}