Line data Source code
1 : /*
2 : * Copyright (C) 2001 Sistina Software (UK) Limited.
3 : * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
4 : *
5 : * This file is released under the GPL.
6 : */
7 :
8 : #include "dm-core.h"
9 :
10 : #include <linux/module.h>
11 : #include <linux/vmalloc.h>
12 : #include <linux/blkdev.h>
13 : #include <linux/namei.h>
14 : #include <linux/ctype.h>
15 : #include <linux/string.h>
16 : #include <linux/slab.h>
17 : #include <linux/interrupt.h>
18 : #include <linux/mutex.h>
19 : #include <linux/delay.h>
20 : #include <linux/atomic.h>
21 : #include <linux/blk-mq.h>
22 : #include <linux/mount.h>
23 : #include <linux/dax.h>
24 :
25 : #define DM_MSG_PREFIX "table"
26 :
27 : #define NODE_SIZE L1_CACHE_BYTES
28 : #define KEYS_PER_NODE (NODE_SIZE / sizeof(sector_t))
29 : #define CHILDREN_PER_NODE (KEYS_PER_NODE + 1)
30 :
31 : /*
32 : * Similar to ceiling(log_size(n))
33 : */
34 0 : static unsigned int int_log(unsigned int n, unsigned int base)
35 : {
36 0 : int result = 0;
37 :
38 0 : while (n > 1) {
39 0 : n = dm_div_up(n, base);
40 0 : result++;
41 : }
42 :
43 0 : return result;
44 : }
45 :
46 : /*
47 : * Calculate the index of the child node of the n'th node k'th key.
48 : */
49 0 : static inline unsigned int get_child(unsigned int n, unsigned int k)
50 : {
51 0 : return (n * CHILDREN_PER_NODE) + k;
52 : }
53 :
54 : /*
55 : * Return the n'th node of level l from table t.
56 : */
57 0 : static inline sector_t *get_node(struct dm_table *t,
58 : unsigned int l, unsigned int n)
59 : {
60 0 : return t->index[l] + (n * KEYS_PER_NODE);
61 : }
62 :
63 : /*
64 : * Return the highest key that you could lookup from the n'th
65 : * node on level l of the btree.
66 : */
67 0 : static sector_t high(struct dm_table *t, unsigned int l, unsigned int n)
68 : {
69 0 : for (; l < t->depth - 1; l++)
70 0 : n = get_child(n, CHILDREN_PER_NODE - 1);
71 :
72 0 : if (n >= t->counts[l])
73 : return (sector_t) - 1;
74 :
75 0 : return get_node(t, l, n)[KEYS_PER_NODE - 1];
76 : }
77 :
78 : /*
79 : * Fills in a level of the btree based on the highs of the level
80 : * below it.
81 : */
82 0 : static int setup_btree_index(unsigned int l, struct dm_table *t)
83 : {
84 0 : unsigned int n, k;
85 0 : sector_t *node;
86 :
87 0 : for (n = 0U; n < t->counts[l]; n++) {
88 0 : node = get_node(t, l, n);
89 :
90 0 : for (k = 0U; k < KEYS_PER_NODE; k++)
91 0 : node[k] = high(t, l + 1, get_child(n, k));
92 : }
93 :
94 0 : return 0;
95 : }
96 :
97 0 : void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size)
98 : {
99 0 : unsigned long size;
100 0 : void *addr;
101 :
102 : /*
103 : * Check that we're not going to overflow.
104 : */
105 0 : if (nmemb > (ULONG_MAX / elem_size))
106 : return NULL;
107 :
108 0 : size = nmemb * elem_size;
109 0 : addr = vzalloc(size);
110 :
111 0 : return addr;
112 : }
113 : EXPORT_SYMBOL(dm_vcalloc);
114 :
115 : /*
116 : * highs, and targets are managed as dynamic arrays during a
117 : * table load.
118 : */
119 0 : static int alloc_targets(struct dm_table *t, unsigned int num)
120 : {
121 0 : sector_t *n_highs;
122 0 : struct dm_target *n_targets;
123 :
124 : /*
125 : * Allocate both the target array and offset array at once.
126 : */
127 0 : n_highs = (sector_t *) dm_vcalloc(num, sizeof(struct dm_target) +
128 : sizeof(sector_t));
129 0 : if (!n_highs)
130 : return -ENOMEM;
131 :
132 0 : n_targets = (struct dm_target *) (n_highs + num);
133 :
134 0 : memset(n_highs, -1, sizeof(*n_highs) * num);
135 0 : vfree(t->highs);
136 :
137 0 : t->num_allocated = num;
138 0 : t->highs = n_highs;
139 0 : t->targets = n_targets;
140 :
141 0 : return 0;
142 : }
143 :
144 0 : int dm_table_create(struct dm_table **result, fmode_t mode,
145 : unsigned num_targets, struct mapped_device *md)
146 : {
147 0 : struct dm_table *t = kzalloc(sizeof(*t), GFP_KERNEL);
148 :
149 0 : if (!t)
150 : return -ENOMEM;
151 :
152 0 : INIT_LIST_HEAD(&t->devices);
153 :
154 0 : if (!num_targets)
155 0 : num_targets = KEYS_PER_NODE;
156 :
157 0 : num_targets = dm_round_up(num_targets, KEYS_PER_NODE);
158 :
159 0 : if (!num_targets) {
160 0 : kfree(t);
161 0 : return -ENOMEM;
162 : }
163 :
164 0 : if (alloc_targets(t, num_targets)) {
165 0 : kfree(t);
166 0 : return -ENOMEM;
167 : }
168 :
169 0 : t->type = DM_TYPE_NONE;
170 0 : t->mode = mode;
171 0 : t->md = md;
172 0 : *result = t;
173 0 : return 0;
174 : }
175 :
176 0 : static void free_devices(struct list_head *devices, struct mapped_device *md)
177 : {
178 0 : struct list_head *tmp, *next;
179 :
180 0 : list_for_each_safe(tmp, next, devices) {
181 0 : struct dm_dev_internal *dd =
182 0 : list_entry(tmp, struct dm_dev_internal, list);
183 0 : DMWARN("%s: dm_table_destroy: dm_put_device call missing for %s",
184 : dm_device_name(md), dd->dm_dev->name);
185 0 : dm_put_table_device(md, dd->dm_dev);
186 0 : kfree(dd);
187 : }
188 0 : }
189 :
190 : static void dm_table_destroy_keyslot_manager(struct dm_table *t);
191 :
192 0 : void dm_table_destroy(struct dm_table *t)
193 : {
194 0 : unsigned int i;
195 :
196 0 : if (!t)
197 : return;
198 :
199 : /* free the indexes */
200 0 : if (t->depth >= 2)
201 0 : vfree(t->index[t->depth - 2]);
202 :
203 : /* free the targets */
204 0 : for (i = 0; i < t->num_targets; i++) {
205 0 : struct dm_target *tgt = t->targets + i;
206 :
207 0 : if (tgt->type->dtr)
208 0 : tgt->type->dtr(tgt);
209 :
210 0 : dm_put_target_type(tgt->type);
211 : }
212 :
213 0 : vfree(t->highs);
214 :
215 : /* free the device list */
216 0 : free_devices(&t->devices, t->md);
217 :
218 0 : dm_free_md_mempools(t->mempools);
219 :
220 0 : dm_table_destroy_keyslot_manager(t);
221 :
222 0 : kfree(t);
223 : }
224 :
225 : /*
226 : * See if we've already got a device in the list.
227 : */
228 0 : static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev)
229 : {
230 0 : struct dm_dev_internal *dd;
231 :
232 0 : list_for_each_entry (dd, l, list)
233 0 : if (dd->dm_dev->bdev->bd_dev == dev)
234 : return dd;
235 :
236 : return NULL;
237 : }
238 :
239 : /*
240 : * If possible, this checks an area of a destination device is invalid.
241 : */
242 0 : static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev,
243 : sector_t start, sector_t len, void *data)
244 : {
245 0 : struct queue_limits *limits = data;
246 0 : struct block_device *bdev = dev->bdev;
247 0 : sector_t dev_size =
248 0 : i_size_read(bdev->bd_inode) >> SECTOR_SHIFT;
249 0 : unsigned short logical_block_size_sectors =
250 0 : limits->logical_block_size >> SECTOR_SHIFT;
251 0 : char b[BDEVNAME_SIZE];
252 :
253 0 : if (!dev_size)
254 : return 0;
255 :
256 0 : if ((start >= dev_size) || (start + len > dev_size)) {
257 0 : DMWARN("%s: %s too small for target: "
258 : "start=%llu, len=%llu, dev_size=%llu",
259 : dm_device_name(ti->table->md), bdevname(bdev, b),
260 : (unsigned long long)start,
261 : (unsigned long long)len,
262 : (unsigned long long)dev_size);
263 0 : return 1;
264 : }
265 :
266 : /*
267 : * If the target is mapped to zoned block device(s), check
268 : * that the zones are not partially mapped.
269 : */
270 0 : if (bdev_zoned_model(bdev) != BLK_ZONED_NONE) {
271 : unsigned int zone_sectors = bdev_zone_sectors(bdev);
272 :
273 : if (start & (zone_sectors - 1)) {
274 : DMWARN("%s: start=%llu not aligned to h/w zone size %u of %s",
275 : dm_device_name(ti->table->md),
276 : (unsigned long long)start,
277 : zone_sectors, bdevname(bdev, b));
278 : return 1;
279 : }
280 :
281 : /*
282 : * Note: The last zone of a zoned block device may be smaller
283 : * than other zones. So for a target mapping the end of a
284 : * zoned block device with such a zone, len would not be zone
285 : * aligned. We do not allow such last smaller zone to be part
286 : * of the mapping here to ensure that mappings with multiple
287 : * devices do not end up with a smaller zone in the middle of
288 : * the sector range.
289 : */
290 : if (len & (zone_sectors - 1)) {
291 : DMWARN("%s: len=%llu not aligned to h/w zone size %u of %s",
292 : dm_device_name(ti->table->md),
293 : (unsigned long long)len,
294 : zone_sectors, bdevname(bdev, b));
295 : return 1;
296 : }
297 : }
298 :
299 0 : if (logical_block_size_sectors <= 1)
300 : return 0;
301 :
302 0 : if (start & (logical_block_size_sectors - 1)) {
303 0 : DMWARN("%s: start=%llu not aligned to h/w "
304 : "logical block size %u of %s",
305 : dm_device_name(ti->table->md),
306 : (unsigned long long)start,
307 : limits->logical_block_size, bdevname(bdev, b));
308 0 : return 1;
309 : }
310 :
311 0 : if (len & (logical_block_size_sectors - 1)) {
312 0 : DMWARN("%s: len=%llu not aligned to h/w "
313 : "logical block size %u of %s",
314 : dm_device_name(ti->table->md),
315 : (unsigned long long)len,
316 : limits->logical_block_size, bdevname(bdev, b));
317 0 : return 1;
318 : }
319 :
320 : return 0;
321 : }
322 :
323 : /*
324 : * This upgrades the mode on an already open dm_dev, being
325 : * careful to leave things as they were if we fail to reopen the
326 : * device and not to touch the existing bdev field in case
327 : * it is accessed concurrently.
328 : */
329 0 : static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode,
330 : struct mapped_device *md)
331 : {
332 0 : int r;
333 0 : struct dm_dev *old_dev, *new_dev;
334 :
335 0 : old_dev = dd->dm_dev;
336 :
337 0 : r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev,
338 0 : dd->dm_dev->mode | new_mode, &new_dev);
339 0 : if (r)
340 : return r;
341 :
342 0 : dd->dm_dev = new_dev;
343 0 : dm_put_table_device(md, old_dev);
344 :
345 0 : return 0;
346 : }
347 :
348 : /*
349 : * Convert the path to a device
350 : */
351 0 : dev_t dm_get_dev_t(const char *path)
352 : {
353 0 : dev_t dev;
354 :
355 0 : if (lookup_bdev(path, &dev))
356 0 : dev = name_to_dev_t(path);
357 0 : return dev;
358 : }
359 : EXPORT_SYMBOL_GPL(dm_get_dev_t);
360 :
361 : /*
362 : * Add a device to the list, or just increment the usage count if
363 : * it's already present.
364 : */
365 0 : int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode,
366 : struct dm_dev **result)
367 : {
368 0 : int r;
369 0 : dev_t dev;
370 0 : unsigned int major, minor;
371 0 : char dummy;
372 0 : struct dm_dev_internal *dd;
373 0 : struct dm_table *t = ti->table;
374 :
375 0 : BUG_ON(!t);
376 :
377 0 : if (sscanf(path, "%u:%u%c", &major, &minor, &dummy) == 2) {
378 : /* Extract the major/minor numbers */
379 0 : dev = MKDEV(major, minor);
380 0 : if (MAJOR(dev) != major || MINOR(dev) != minor)
381 : return -EOVERFLOW;
382 : } else {
383 0 : dev = dm_get_dev_t(path);
384 0 : if (!dev)
385 : return -ENODEV;
386 : }
387 :
388 0 : dd = find_device(&t->devices, dev);
389 0 : if (!dd) {
390 0 : dd = kmalloc(sizeof(*dd), GFP_KERNEL);
391 0 : if (!dd)
392 : return -ENOMEM;
393 :
394 0 : if ((r = dm_get_table_device(t->md, dev, mode, &dd->dm_dev))) {
395 0 : kfree(dd);
396 0 : return r;
397 : }
398 :
399 0 : refcount_set(&dd->count, 1);
400 0 : list_add(&dd->list, &t->devices);
401 0 : goto out;
402 :
403 0 : } else if (dd->dm_dev->mode != (mode | dd->dm_dev->mode)) {
404 0 : r = upgrade_mode(dd, mode, t->md);
405 0 : if (r)
406 : return r;
407 : }
408 0 : refcount_inc(&dd->count);
409 0 : out:
410 0 : *result = dd->dm_dev;
411 0 : return 0;
412 : }
413 : EXPORT_SYMBOL(dm_get_device);
414 :
415 0 : static int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev,
416 : sector_t start, sector_t len, void *data)
417 : {
418 0 : struct queue_limits *limits = data;
419 0 : struct block_device *bdev = dev->bdev;
420 0 : struct request_queue *q = bdev_get_queue(bdev);
421 0 : char b[BDEVNAME_SIZE];
422 :
423 0 : if (unlikely(!q)) {
424 0 : DMWARN("%s: Cannot set limits for nonexistent device %s",
425 : dm_device_name(ti->table->md), bdevname(bdev, b));
426 0 : return 0;
427 : }
428 :
429 0 : if (blk_stack_limits(limits, &q->limits,
430 0 : get_start_sect(bdev) + start) < 0)
431 0 : DMWARN("%s: adding target device %s caused an alignment inconsistency: "
432 : "physical_block_size=%u, logical_block_size=%u, "
433 : "alignment_offset=%u, start=%llu",
434 : dm_device_name(ti->table->md), bdevname(bdev, b),
435 : q->limits.physical_block_size,
436 : q->limits.logical_block_size,
437 : q->limits.alignment_offset,
438 : (unsigned long long) start << SECTOR_SHIFT);
439 : return 0;
440 : }
441 :
442 : /*
443 : * Decrement a device's use count and remove it if necessary.
444 : */
445 0 : void dm_put_device(struct dm_target *ti, struct dm_dev *d)
446 : {
447 0 : int found = 0;
448 0 : struct list_head *devices = &ti->table->devices;
449 0 : struct dm_dev_internal *dd;
450 :
451 0 : list_for_each_entry(dd, devices, list) {
452 0 : if (dd->dm_dev == d) {
453 : found = 1;
454 : break;
455 : }
456 : }
457 0 : if (!found) {
458 0 : DMWARN("%s: device %s not in table devices list",
459 : dm_device_name(ti->table->md), d->name);
460 0 : return;
461 : }
462 0 : if (refcount_dec_and_test(&dd->count)) {
463 0 : dm_put_table_device(ti->table->md, d);
464 0 : list_del(&dd->list);
465 0 : kfree(dd);
466 : }
467 : }
468 : EXPORT_SYMBOL(dm_put_device);
469 :
470 : /*
471 : * Checks to see if the target joins onto the end of the table.
472 : */
473 0 : static int adjoin(struct dm_table *table, struct dm_target *ti)
474 : {
475 0 : struct dm_target *prev;
476 :
477 0 : if (!table->num_targets)
478 0 : return !ti->begin;
479 :
480 0 : prev = &table->targets[table->num_targets - 1];
481 0 : return (ti->begin == (prev->begin + prev->len));
482 : }
483 :
484 : /*
485 : * Used to dynamically allocate the arg array.
486 : *
487 : * We do first allocation with GFP_NOIO because dm-mpath and dm-thin must
488 : * process messages even if some device is suspended. These messages have a
489 : * small fixed number of arguments.
490 : *
491 : * On the other hand, dm-switch needs to process bulk data using messages and
492 : * excessive use of GFP_NOIO could cause trouble.
493 : */
494 0 : static char **realloc_argv(unsigned *size, char **old_argv)
495 : {
496 0 : char **argv;
497 0 : unsigned new_size;
498 0 : gfp_t gfp;
499 :
500 0 : if (*size) {
501 0 : new_size = *size * 2;
502 0 : gfp = GFP_KERNEL;
503 : } else {
504 : new_size = 8;
505 : gfp = GFP_NOIO;
506 : }
507 0 : argv = kmalloc_array(new_size, sizeof(*argv), gfp);
508 0 : if (argv && old_argv) {
509 0 : memcpy(argv, old_argv, *size * sizeof(*argv));
510 0 : *size = new_size;
511 : }
512 :
513 0 : kfree(old_argv);
514 0 : return argv;
515 : }
516 :
517 : /*
518 : * Destructively splits up the argument list to pass to ctr.
519 : */
520 0 : int dm_split_args(int *argc, char ***argvp, char *input)
521 : {
522 0 : char *start, *end = input, *out, **argv = NULL;
523 0 : unsigned array_size = 0;
524 :
525 0 : *argc = 0;
526 :
527 0 : if (!input) {
528 0 : *argvp = NULL;
529 0 : return 0;
530 : }
531 :
532 0 : argv = realloc_argv(&array_size, argv);
533 0 : if (!argv)
534 : return -ENOMEM;
535 :
536 0 : while (1) {
537 : /* Skip whitespace */
538 0 : start = skip_spaces(end);
539 :
540 0 : if (!*start)
541 : break; /* success, we hit the end */
542 :
543 : /* 'out' is used to remove any back-quotes */
544 : end = out = start;
545 0 : while (*end) {
546 : /* Everything apart from '\0' can be quoted */
547 0 : if (*end == '\\' && *(end + 1)) {
548 0 : *out++ = *(end + 1);
549 0 : end += 2;
550 0 : continue;
551 : }
552 :
553 0 : if (isspace(*end))
554 : break; /* end of token */
555 :
556 0 : *out++ = *end++;
557 : }
558 :
559 : /* have we already filled the array ? */
560 0 : if ((*argc + 1) > array_size) {
561 0 : argv = realloc_argv(&array_size, argv);
562 0 : if (!argv)
563 : return -ENOMEM;
564 : }
565 :
566 : /* we know this is whitespace */
567 0 : if (*end)
568 0 : end++;
569 :
570 : /* terminate the string and put it in the array */
571 0 : *out = '\0';
572 0 : argv[*argc] = start;
573 0 : (*argc)++;
574 : }
575 :
576 0 : *argvp = argv;
577 0 : return 0;
578 : }
579 :
580 : /*
581 : * Impose necessary and sufficient conditions on a devices's table such
582 : * that any incoming bio which respects its logical_block_size can be
583 : * processed successfully. If it falls across the boundary between
584 : * two or more targets, the size of each piece it gets split into must
585 : * be compatible with the logical_block_size of the target processing it.
586 : */
587 0 : static int validate_hardware_logical_block_alignment(struct dm_table *table,
588 : struct queue_limits *limits)
589 : {
590 : /*
591 : * This function uses arithmetic modulo the logical_block_size
592 : * (in units of 512-byte sectors).
593 : */
594 0 : unsigned short device_logical_block_size_sects =
595 0 : limits->logical_block_size >> SECTOR_SHIFT;
596 :
597 : /*
598 : * Offset of the start of the next table entry, mod logical_block_size.
599 : */
600 0 : unsigned short next_target_start = 0;
601 :
602 : /*
603 : * Given an aligned bio that extends beyond the end of a
604 : * target, how many sectors must the next target handle?
605 : */
606 0 : unsigned short remaining = 0;
607 :
608 0 : struct dm_target *ti;
609 0 : struct queue_limits ti_limits;
610 0 : unsigned i;
611 :
612 : /*
613 : * Check each entry in the table in turn.
614 : */
615 0 : for (i = 0; i < dm_table_get_num_targets(table); i++) {
616 0 : ti = dm_table_get_target(table, i);
617 :
618 0 : blk_set_stacking_limits(&ti_limits);
619 :
620 : /* combine all target devices' limits */
621 0 : if (ti->type->iterate_devices)
622 0 : ti->type->iterate_devices(ti, dm_set_device_limits,
623 : &ti_limits);
624 :
625 : /*
626 : * If the remaining sectors fall entirely within this
627 : * table entry are they compatible with its logical_block_size?
628 : */
629 0 : if (remaining < ti->len &&
630 0 : remaining & ((ti_limits.logical_block_size >>
631 0 : SECTOR_SHIFT) - 1))
632 : break; /* Error */
633 :
634 0 : next_target_start =
635 0 : (unsigned short) ((next_target_start + ti->len) &
636 0 : (device_logical_block_size_sects - 1));
637 0 : remaining = next_target_start ?
638 : device_logical_block_size_sects - next_target_start : 0;
639 : }
640 :
641 0 : if (remaining) {
642 0 : DMWARN("%s: table line %u (start sect %llu len %llu) "
643 : "not aligned to h/w logical block size %u",
644 : dm_device_name(table->md), i,
645 : (unsigned long long) ti->begin,
646 : (unsigned long long) ti->len,
647 : limits->logical_block_size);
648 0 : return -EINVAL;
649 : }
650 :
651 : return 0;
652 : }
653 :
654 0 : int dm_table_add_target(struct dm_table *t, const char *type,
655 : sector_t start, sector_t len, char *params)
656 : {
657 0 : int r = -EINVAL, argc;
658 0 : char **argv;
659 0 : struct dm_target *tgt;
660 :
661 0 : if (t->singleton) {
662 0 : DMERR("%s: target type %s must appear alone in table",
663 : dm_device_name(t->md), t->targets->type->name);
664 0 : return -EINVAL;
665 : }
666 :
667 0 : BUG_ON(t->num_targets >= t->num_allocated);
668 :
669 0 : tgt = t->targets + t->num_targets;
670 0 : memset(tgt, 0, sizeof(*tgt));
671 :
672 0 : if (!len) {
673 0 : DMERR("%s: zero-length target", dm_device_name(t->md));
674 0 : return -EINVAL;
675 : }
676 :
677 0 : tgt->type = dm_get_target_type(type);
678 0 : if (!tgt->type) {
679 0 : DMERR("%s: %s: unknown target type", dm_device_name(t->md), type);
680 0 : return -EINVAL;
681 : }
682 :
683 0 : if (dm_target_needs_singleton(tgt->type)) {
684 0 : if (t->num_targets) {
685 0 : tgt->error = "singleton target type must appear alone in table";
686 0 : goto bad;
687 : }
688 0 : t->singleton = true;
689 : }
690 :
691 0 : if (dm_target_always_writeable(tgt->type) && !(t->mode & FMODE_WRITE)) {
692 0 : tgt->error = "target type may not be included in a read-only table";
693 0 : goto bad;
694 : }
695 :
696 0 : if (t->immutable_target_type) {
697 0 : if (t->immutable_target_type != tgt->type) {
698 0 : tgt->error = "immutable target type cannot be mixed with other target types";
699 0 : goto bad;
700 : }
701 0 : } else if (dm_target_is_immutable(tgt->type)) {
702 0 : if (t->num_targets) {
703 0 : tgt->error = "immutable target type cannot be mixed with other target types";
704 0 : goto bad;
705 : }
706 0 : t->immutable_target_type = tgt->type;
707 : }
708 :
709 0 : if (dm_target_has_integrity(tgt->type))
710 0 : t->integrity_added = 1;
711 :
712 0 : tgt->table = t;
713 0 : tgt->begin = start;
714 0 : tgt->len = len;
715 0 : tgt->error = "Unknown error";
716 :
717 : /*
718 : * Does this target adjoin the previous one ?
719 : */
720 0 : if (!adjoin(t, tgt)) {
721 0 : tgt->error = "Gap in table";
722 0 : goto bad;
723 : }
724 :
725 0 : r = dm_split_args(&argc, &argv, params);
726 0 : if (r) {
727 0 : tgt->error = "couldn't split parameters (insufficient memory)";
728 0 : goto bad;
729 : }
730 :
731 0 : r = tgt->type->ctr(tgt, argc, argv);
732 0 : kfree(argv);
733 0 : if (r)
734 0 : goto bad;
735 :
736 0 : t->highs[t->num_targets++] = tgt->begin + tgt->len - 1;
737 :
738 0 : if (!tgt->num_discard_bios && tgt->discards_supported)
739 0 : DMWARN("%s: %s: ignoring discards_supported because num_discard_bios is zero.",
740 : dm_device_name(t->md), type);
741 :
742 : return 0;
743 :
744 0 : bad:
745 0 : DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error);
746 0 : dm_put_target_type(tgt->type);
747 0 : return r;
748 : }
749 :
750 : /*
751 : * Target argument parsing helpers.
752 : */
753 0 : static int validate_next_arg(const struct dm_arg *arg,
754 : struct dm_arg_set *arg_set,
755 : unsigned *value, char **error, unsigned grouped)
756 : {
757 0 : const char *arg_str = dm_shift_arg(arg_set);
758 0 : char dummy;
759 :
760 0 : if (!arg_str ||
761 0 : (sscanf(arg_str, "%u%c", value, &dummy) != 1) ||
762 0 : (*value < arg->min) ||
763 0 : (*value > arg->max) ||
764 0 : (grouped && arg_set->argc < *value)) {
765 0 : *error = arg->error;
766 0 : return -EINVAL;
767 : }
768 :
769 : return 0;
770 : }
771 :
772 0 : int dm_read_arg(const struct dm_arg *arg, struct dm_arg_set *arg_set,
773 : unsigned *value, char **error)
774 : {
775 0 : return validate_next_arg(arg, arg_set, value, error, 0);
776 : }
777 : EXPORT_SYMBOL(dm_read_arg);
778 :
779 0 : int dm_read_arg_group(const struct dm_arg *arg, struct dm_arg_set *arg_set,
780 : unsigned *value, char **error)
781 : {
782 0 : return validate_next_arg(arg, arg_set, value, error, 1);
783 : }
784 : EXPORT_SYMBOL(dm_read_arg_group);
785 :
786 0 : const char *dm_shift_arg(struct dm_arg_set *as)
787 : {
788 0 : char *r;
789 :
790 0 : if (as->argc) {
791 0 : as->argc--;
792 0 : r = *as->argv;
793 0 : as->argv++;
794 0 : return r;
795 : }
796 :
797 : return NULL;
798 : }
799 : EXPORT_SYMBOL(dm_shift_arg);
800 :
801 0 : void dm_consume_args(struct dm_arg_set *as, unsigned num_args)
802 : {
803 0 : BUG_ON(as->argc < num_args);
804 0 : as->argc -= num_args;
805 0 : as->argv += num_args;
806 0 : }
807 : EXPORT_SYMBOL(dm_consume_args);
808 :
809 0 : static bool __table_type_bio_based(enum dm_queue_mode table_type)
810 : {
811 0 : return (table_type == DM_TYPE_BIO_BASED ||
812 0 : table_type == DM_TYPE_DAX_BIO_BASED);
813 : }
814 :
815 0 : static bool __table_type_request_based(enum dm_queue_mode table_type)
816 : {
817 0 : return table_type == DM_TYPE_REQUEST_BASED;
818 : }
819 :
820 0 : void dm_table_set_type(struct dm_table *t, enum dm_queue_mode type)
821 : {
822 0 : t->type = type;
823 0 : }
824 : EXPORT_SYMBOL_GPL(dm_table_set_type);
825 :
826 : /* validate the dax capability of the target device span */
827 0 : int device_not_dax_capable(struct dm_target *ti, struct dm_dev *dev,
828 : sector_t start, sector_t len, void *data)
829 : {
830 0 : int blocksize = *(int *) data, id;
831 0 : bool rc;
832 :
833 0 : id = dax_read_lock();
834 0 : rc = !dax_supported(dev->dax_dev, dev->bdev, blocksize, start, len);
835 0 : dax_read_unlock(id);
836 :
837 0 : return rc;
838 : }
839 :
840 : /* Check devices support synchronous DAX */
841 0 : static int device_not_dax_synchronous_capable(struct dm_target *ti, struct dm_dev *dev,
842 : sector_t start, sector_t len, void *data)
843 : {
844 0 : return !dev->dax_dev || !dax_synchronous(dev->dax_dev);
845 : }
846 :
847 0 : bool dm_table_supports_dax(struct dm_table *t,
848 : iterate_devices_callout_fn iterate_fn, int *blocksize)
849 : {
850 0 : struct dm_target *ti;
851 0 : unsigned i;
852 :
853 : /* Ensure that all targets support DAX. */
854 0 : for (i = 0; i < dm_table_get_num_targets(t); i++) {
855 0 : ti = dm_table_get_target(t, i);
856 :
857 0 : if (!ti->type->direct_access)
858 : return false;
859 :
860 0 : if (!ti->type->iterate_devices ||
861 0 : ti->type->iterate_devices(ti, iterate_fn, blocksize))
862 0 : return false;
863 : }
864 :
865 : return true;
866 : }
867 :
868 0 : static int device_is_rq_stackable(struct dm_target *ti, struct dm_dev *dev,
869 : sector_t start, sector_t len, void *data)
870 : {
871 0 : struct block_device *bdev = dev->bdev;
872 0 : struct request_queue *q = bdev_get_queue(bdev);
873 :
874 : /* request-based cannot stack on partitions! */
875 0 : if (bdev_is_partition(bdev))
876 : return false;
877 :
878 0 : return queue_is_mq(q);
879 : }
880 :
881 0 : static int dm_table_determine_type(struct dm_table *t)
882 : {
883 0 : unsigned i;
884 0 : unsigned bio_based = 0, request_based = 0, hybrid = 0;
885 0 : struct dm_target *tgt;
886 0 : struct list_head *devices = dm_table_get_devices(t);
887 0 : enum dm_queue_mode live_md_type = dm_get_md_type(t->md);
888 0 : int page_size = PAGE_SIZE;
889 :
890 0 : if (t->type != DM_TYPE_NONE) {
891 : /* target already set the table's type */
892 0 : if (t->type == DM_TYPE_BIO_BASED) {
893 : /* possibly upgrade to a variant of bio-based */
894 0 : goto verify_bio_based;
895 : }
896 0 : BUG_ON(t->type == DM_TYPE_DAX_BIO_BASED);
897 0 : goto verify_rq_based;
898 : }
899 :
900 0 : for (i = 0; i < t->num_targets; i++) {
901 0 : tgt = t->targets + i;
902 0 : if (dm_target_hybrid(tgt))
903 : hybrid = 1;
904 0 : else if (dm_target_request_based(tgt))
905 : request_based = 1;
906 : else
907 0 : bio_based = 1;
908 :
909 0 : if (bio_based && request_based) {
910 0 : DMERR("Inconsistent table: different target types"
911 : " can't be mixed up");
912 0 : return -EINVAL;
913 : }
914 : }
915 :
916 0 : if (hybrid && !bio_based && !request_based) {
917 : /*
918 : * The targets can work either way.
919 : * Determine the type from the live device.
920 : * Default to bio-based if device is new.
921 : */
922 0 : if (__table_type_request_based(live_md_type))
923 : request_based = 1;
924 : else
925 : bio_based = 1;
926 : }
927 :
928 0 : if (bio_based) {
929 0 : verify_bio_based:
930 : /* We must use this table as bio-based */
931 0 : t->type = DM_TYPE_BIO_BASED;
932 0 : if (dm_table_supports_dax(t, device_not_dax_capable, &page_size) ||
933 0 : (list_empty(devices) && live_md_type == DM_TYPE_DAX_BIO_BASED)) {
934 0 : t->type = DM_TYPE_DAX_BIO_BASED;
935 : }
936 0 : return 0;
937 : }
938 :
939 0 : BUG_ON(!request_based); /* No targets in this table */
940 :
941 0 : t->type = DM_TYPE_REQUEST_BASED;
942 :
943 0 : verify_rq_based:
944 : /*
945 : * Request-based dm supports only tables that have a single target now.
946 : * To support multiple targets, request splitting support is needed,
947 : * and that needs lots of changes in the block-layer.
948 : * (e.g. request completion process for partial completion.)
949 : */
950 0 : if (t->num_targets > 1) {
951 0 : DMERR("request-based DM doesn't support multiple targets");
952 0 : return -EINVAL;
953 : }
954 :
955 0 : if (list_empty(devices)) {
956 0 : int srcu_idx;
957 0 : struct dm_table *live_table = dm_get_live_table(t->md, &srcu_idx);
958 :
959 : /* inherit live table's type */
960 0 : if (live_table)
961 0 : t->type = live_table->type;
962 0 : dm_put_live_table(t->md, srcu_idx);
963 0 : return 0;
964 : }
965 :
966 0 : tgt = dm_table_get_immutable_target(t);
967 0 : if (!tgt) {
968 0 : DMERR("table load rejected: immutable target is required");
969 0 : return -EINVAL;
970 0 : } else if (tgt->max_io_len) {
971 0 : DMERR("table load rejected: immutable target that splits IO is not supported");
972 0 : return -EINVAL;
973 : }
974 :
975 : /* Non-request-stackable devices can't be used for request-based dm */
976 0 : if (!tgt->type->iterate_devices ||
977 0 : !tgt->type->iterate_devices(tgt, device_is_rq_stackable, NULL)) {
978 0 : DMERR("table load rejected: including non-request-stackable devices");
979 0 : return -EINVAL;
980 : }
981 :
982 : return 0;
983 : }
984 :
985 0 : enum dm_queue_mode dm_table_get_type(struct dm_table *t)
986 : {
987 0 : return t->type;
988 : }
989 :
990 0 : struct target_type *dm_table_get_immutable_target_type(struct dm_table *t)
991 : {
992 0 : return t->immutable_target_type;
993 : }
994 :
995 0 : struct dm_target *dm_table_get_immutable_target(struct dm_table *t)
996 : {
997 : /* Immutable target is implicitly a singleton */
998 0 : if (t->num_targets > 1 ||
999 0 : !dm_target_is_immutable(t->targets[0].type))
1000 0 : return NULL;
1001 :
1002 : return t->targets;
1003 : }
1004 :
1005 0 : struct dm_target *dm_table_get_wildcard_target(struct dm_table *t)
1006 : {
1007 0 : struct dm_target *ti;
1008 0 : unsigned i;
1009 :
1010 0 : for (i = 0; i < dm_table_get_num_targets(t); i++) {
1011 0 : ti = dm_table_get_target(t, i);
1012 0 : if (dm_target_is_wildcard(ti->type))
1013 0 : return ti;
1014 : }
1015 :
1016 : return NULL;
1017 : }
1018 :
1019 0 : bool dm_table_bio_based(struct dm_table *t)
1020 : {
1021 0 : return __table_type_bio_based(dm_table_get_type(t));
1022 : }
1023 :
1024 0 : bool dm_table_request_based(struct dm_table *t)
1025 : {
1026 0 : return __table_type_request_based(dm_table_get_type(t));
1027 : }
1028 :
1029 0 : static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device *md)
1030 : {
1031 0 : enum dm_queue_mode type = dm_table_get_type(t);
1032 0 : unsigned per_io_data_size = 0;
1033 0 : unsigned min_pool_size = 0;
1034 0 : struct dm_target *ti;
1035 0 : unsigned i;
1036 :
1037 0 : if (unlikely(type == DM_TYPE_NONE)) {
1038 0 : DMWARN("no table type is set, can't allocate mempools");
1039 0 : return -EINVAL;
1040 : }
1041 :
1042 0 : if (__table_type_bio_based(type))
1043 0 : for (i = 0; i < t->num_targets; i++) {
1044 0 : ti = t->targets + i;
1045 0 : per_io_data_size = max(per_io_data_size, ti->per_io_data_size);
1046 0 : min_pool_size = max(min_pool_size, ti->num_flush_bios);
1047 : }
1048 :
1049 0 : t->mempools = dm_alloc_md_mempools(md, type, t->integrity_supported,
1050 : per_io_data_size, min_pool_size);
1051 0 : if (!t->mempools)
1052 0 : return -ENOMEM;
1053 :
1054 : return 0;
1055 : }
1056 :
1057 0 : void dm_table_free_md_mempools(struct dm_table *t)
1058 : {
1059 0 : dm_free_md_mempools(t->mempools);
1060 0 : t->mempools = NULL;
1061 0 : }
1062 :
1063 0 : struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t)
1064 : {
1065 0 : return t->mempools;
1066 : }
1067 :
1068 0 : static int setup_indexes(struct dm_table *t)
1069 : {
1070 0 : int i;
1071 0 : unsigned int total = 0;
1072 0 : sector_t *indexes;
1073 :
1074 : /* allocate the space for *all* the indexes */
1075 0 : for (i = t->depth - 2; i >= 0; i--) {
1076 0 : t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE);
1077 0 : total += t->counts[i];
1078 : }
1079 :
1080 0 : indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE);
1081 0 : if (!indexes)
1082 : return -ENOMEM;
1083 :
1084 : /* set up internal nodes, bottom-up */
1085 0 : for (i = t->depth - 2; i >= 0; i--) {
1086 0 : t->index[i] = indexes;
1087 0 : indexes += (KEYS_PER_NODE * t->counts[i]);
1088 0 : setup_btree_index(i, t);
1089 : }
1090 :
1091 : return 0;
1092 : }
1093 :
1094 : /*
1095 : * Builds the btree to index the map.
1096 : */
1097 0 : static int dm_table_build_index(struct dm_table *t)
1098 : {
1099 0 : int r = 0;
1100 0 : unsigned int leaf_nodes;
1101 :
1102 : /* how many indexes will the btree have ? */
1103 0 : leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE);
1104 0 : t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE);
1105 :
1106 : /* leaf layer has already been set up */
1107 0 : t->counts[t->depth - 1] = leaf_nodes;
1108 0 : t->index[t->depth - 1] = t->highs;
1109 :
1110 0 : if (t->depth >= 2)
1111 0 : r = setup_indexes(t);
1112 :
1113 0 : return r;
1114 : }
1115 :
1116 0 : static bool integrity_profile_exists(struct gendisk *disk)
1117 : {
1118 0 : return !!blk_get_integrity(disk);
1119 : }
1120 :
1121 : /*
1122 : * Get a disk whose integrity profile reflects the table's profile.
1123 : * Returns NULL if integrity support was inconsistent or unavailable.
1124 : */
1125 0 : static struct gendisk * dm_table_get_integrity_disk(struct dm_table *t)
1126 : {
1127 0 : struct list_head *devices = dm_table_get_devices(t);
1128 0 : struct dm_dev_internal *dd = NULL;
1129 0 : struct gendisk *prev_disk = NULL, *template_disk = NULL;
1130 0 : unsigned i;
1131 :
1132 0 : for (i = 0; i < dm_table_get_num_targets(t); i++) {
1133 0 : struct dm_target *ti = dm_table_get_target(t, i);
1134 0 : if (!dm_target_passes_integrity(ti->type))
1135 0 : goto no_integrity;
1136 : }
1137 :
1138 0 : list_for_each_entry(dd, devices, list) {
1139 : template_disk = dd->dm_dev->bdev->bd_disk;
1140 : if (!integrity_profile_exists(template_disk))
1141 : goto no_integrity;
1142 : else if (prev_disk &&
1143 : blk_integrity_compare(prev_disk, template_disk) < 0)
1144 : goto no_integrity;
1145 : prev_disk = template_disk;
1146 : }
1147 :
1148 : return template_disk;
1149 :
1150 : no_integrity:
1151 : if (prev_disk)
1152 : DMWARN("%s: integrity not set: %s and %s profile mismatch",
1153 : dm_device_name(t->md),
1154 : prev_disk->disk_name,
1155 : template_disk->disk_name);
1156 : return NULL;
1157 : }
1158 :
1159 : /*
1160 : * Register the mapped device for blk_integrity support if the
1161 : * underlying devices have an integrity profile. But all devices may
1162 : * not have matching profiles (checking all devices isn't reliable
1163 : * during table load because this table may use other DM device(s) which
1164 : * must be resumed before they will have an initialized integity
1165 : * profile). Consequently, stacked DM devices force a 2 stage integrity
1166 : * profile validation: First pass during table load, final pass during
1167 : * resume.
1168 : */
1169 0 : static int dm_table_register_integrity(struct dm_table *t)
1170 : {
1171 0 : struct mapped_device *md = t->md;
1172 0 : struct gendisk *template_disk = NULL;
1173 :
1174 : /* If target handles integrity itself do not register it here. */
1175 0 : if (t->integrity_added)
1176 : return 0;
1177 :
1178 0 : template_disk = dm_table_get_integrity_disk(t);
1179 0 : if (!template_disk)
1180 : return 0;
1181 :
1182 0 : if (!integrity_profile_exists(dm_disk(md))) {
1183 0 : t->integrity_supported = true;
1184 : /*
1185 : * Register integrity profile during table load; we can do
1186 : * this because the final profile must match during resume.
1187 : */
1188 0 : blk_integrity_register(dm_disk(md),
1189 : blk_get_integrity(template_disk));
1190 0 : return 0;
1191 : }
1192 :
1193 : /*
1194 : * If DM device already has an initialized integrity
1195 : * profile the new profile should not conflict.
1196 : */
1197 : if (blk_integrity_compare(dm_disk(md), template_disk) < 0) {
1198 : DMWARN("%s: conflict with existing integrity profile: "
1199 : "%s profile mismatch",
1200 : dm_device_name(t->md),
1201 : template_disk->disk_name);
1202 : return 1;
1203 : }
1204 :
1205 : /* Preserve existing integrity profile */
1206 : t->integrity_supported = true;
1207 : return 0;
1208 : }
1209 :
1210 : #ifdef CONFIG_BLK_INLINE_ENCRYPTION
1211 :
1212 : struct dm_keyslot_manager {
1213 : struct blk_keyslot_manager ksm;
1214 : struct mapped_device *md;
1215 : };
1216 :
1217 : struct dm_keyslot_evict_args {
1218 : const struct blk_crypto_key *key;
1219 : int err;
1220 : };
1221 :
1222 : static int dm_keyslot_evict_callback(struct dm_target *ti, struct dm_dev *dev,
1223 : sector_t start, sector_t len, void *data)
1224 : {
1225 : struct dm_keyslot_evict_args *args = data;
1226 : int err;
1227 :
1228 : err = blk_crypto_evict_key(bdev_get_queue(dev->bdev), args->key);
1229 : if (!args->err)
1230 : args->err = err;
1231 : /* Always try to evict the key from all devices. */
1232 : return 0;
1233 : }
1234 :
1235 : /*
1236 : * When an inline encryption key is evicted from a device-mapper device, evict
1237 : * it from all the underlying devices.
1238 : */
1239 : static int dm_keyslot_evict(struct blk_keyslot_manager *ksm,
1240 : const struct blk_crypto_key *key, unsigned int slot)
1241 : {
1242 : struct dm_keyslot_manager *dksm = container_of(ksm,
1243 : struct dm_keyslot_manager,
1244 : ksm);
1245 : struct mapped_device *md = dksm->md;
1246 : struct dm_keyslot_evict_args args = { key };
1247 : struct dm_table *t;
1248 : int srcu_idx;
1249 : int i;
1250 : struct dm_target *ti;
1251 :
1252 : t = dm_get_live_table(md, &srcu_idx);
1253 : if (!t)
1254 : return 0;
1255 : for (i = 0; i < dm_table_get_num_targets(t); i++) {
1256 : ti = dm_table_get_target(t, i);
1257 : if (!ti->type->iterate_devices)
1258 : continue;
1259 : ti->type->iterate_devices(ti, dm_keyslot_evict_callback, &args);
1260 : }
1261 : dm_put_live_table(md, srcu_idx);
1262 : return args.err;
1263 : }
1264 :
1265 : static struct blk_ksm_ll_ops dm_ksm_ll_ops = {
1266 : .keyslot_evict = dm_keyslot_evict,
1267 : };
1268 :
1269 : static int device_intersect_crypto_modes(struct dm_target *ti,
1270 : struct dm_dev *dev, sector_t start,
1271 : sector_t len, void *data)
1272 : {
1273 : struct blk_keyslot_manager *parent = data;
1274 : struct blk_keyslot_manager *child = bdev_get_queue(dev->bdev)->ksm;
1275 :
1276 : blk_ksm_intersect_modes(parent, child);
1277 : return 0;
1278 : }
1279 :
1280 : void dm_destroy_keyslot_manager(struct blk_keyslot_manager *ksm)
1281 : {
1282 : struct dm_keyslot_manager *dksm = container_of(ksm,
1283 : struct dm_keyslot_manager,
1284 : ksm);
1285 :
1286 : if (!ksm)
1287 : return;
1288 :
1289 : blk_ksm_destroy(ksm);
1290 : kfree(dksm);
1291 : }
1292 :
1293 : static void dm_table_destroy_keyslot_manager(struct dm_table *t)
1294 : {
1295 : dm_destroy_keyslot_manager(t->ksm);
1296 : t->ksm = NULL;
1297 : }
1298 :
1299 : /*
1300 : * Constructs and initializes t->ksm with a keyslot manager that
1301 : * represents the common set of crypto capabilities of the devices
1302 : * described by the dm_table. However, if the constructed keyslot
1303 : * manager does not support a superset of the crypto capabilities
1304 : * supported by the current keyslot manager of the mapped_device,
1305 : * it returns an error instead, since we don't support restricting
1306 : * crypto capabilities on table changes. Finally, if the constructed
1307 : * keyslot manager doesn't actually support any crypto modes at all,
1308 : * it just returns NULL.
1309 : */
1310 : static int dm_table_construct_keyslot_manager(struct dm_table *t)
1311 : {
1312 : struct dm_keyslot_manager *dksm;
1313 : struct blk_keyslot_manager *ksm;
1314 : struct dm_target *ti;
1315 : unsigned int i;
1316 : bool ksm_is_empty = true;
1317 :
1318 : dksm = kmalloc(sizeof(*dksm), GFP_KERNEL);
1319 : if (!dksm)
1320 : return -ENOMEM;
1321 : dksm->md = t->md;
1322 :
1323 : ksm = &dksm->ksm;
1324 : blk_ksm_init_passthrough(ksm);
1325 : ksm->ksm_ll_ops = dm_ksm_ll_ops;
1326 : ksm->max_dun_bytes_supported = UINT_MAX;
1327 : memset(ksm->crypto_modes_supported, 0xFF,
1328 : sizeof(ksm->crypto_modes_supported));
1329 :
1330 : for (i = 0; i < dm_table_get_num_targets(t); i++) {
1331 : ti = dm_table_get_target(t, i);
1332 :
1333 : if (!dm_target_passes_crypto(ti->type)) {
1334 : blk_ksm_intersect_modes(ksm, NULL);
1335 : break;
1336 : }
1337 : if (!ti->type->iterate_devices)
1338 : continue;
1339 : ti->type->iterate_devices(ti, device_intersect_crypto_modes,
1340 : ksm);
1341 : }
1342 :
1343 : if (t->md->queue && !blk_ksm_is_superset(ksm, t->md->queue->ksm)) {
1344 : DMWARN("Inline encryption capabilities of new DM table were more restrictive than the old table's. This is not supported!");
1345 : dm_destroy_keyslot_manager(ksm);
1346 : return -EINVAL;
1347 : }
1348 :
1349 : /*
1350 : * If the new KSM doesn't actually support any crypto modes, we may as
1351 : * well represent it with a NULL ksm.
1352 : */
1353 : ksm_is_empty = true;
1354 : for (i = 0; i < ARRAY_SIZE(ksm->crypto_modes_supported); i++) {
1355 : if (ksm->crypto_modes_supported[i]) {
1356 : ksm_is_empty = false;
1357 : break;
1358 : }
1359 : }
1360 :
1361 : if (ksm_is_empty) {
1362 : dm_destroy_keyslot_manager(ksm);
1363 : ksm = NULL;
1364 : }
1365 :
1366 : /*
1367 : * t->ksm is only set temporarily while the table is being set
1368 : * up, and it gets set to NULL after the capabilities have
1369 : * been transferred to the request_queue.
1370 : */
1371 : t->ksm = ksm;
1372 :
1373 : return 0;
1374 : }
1375 :
1376 : static void dm_update_keyslot_manager(struct request_queue *q,
1377 : struct dm_table *t)
1378 : {
1379 : if (!t->ksm)
1380 : return;
1381 :
1382 : /* Make the ksm less restrictive */
1383 : if (!q->ksm) {
1384 : blk_ksm_register(t->ksm, q);
1385 : } else {
1386 : blk_ksm_update_capabilities(q->ksm, t->ksm);
1387 : dm_destroy_keyslot_manager(t->ksm);
1388 : }
1389 : t->ksm = NULL;
1390 : }
1391 :
1392 : #else /* CONFIG_BLK_INLINE_ENCRYPTION */
1393 :
1394 0 : static int dm_table_construct_keyslot_manager(struct dm_table *t)
1395 : {
1396 0 : return 0;
1397 : }
1398 :
1399 0 : void dm_destroy_keyslot_manager(struct blk_keyslot_manager *ksm)
1400 : {
1401 0 : }
1402 :
1403 0 : static void dm_table_destroy_keyslot_manager(struct dm_table *t)
1404 : {
1405 0 : }
1406 :
1407 0 : static void dm_update_keyslot_manager(struct request_queue *q,
1408 : struct dm_table *t)
1409 : {
1410 0 : }
1411 :
1412 : #endif /* !CONFIG_BLK_INLINE_ENCRYPTION */
1413 :
1414 : /*
1415 : * Prepares the table for use by building the indices,
1416 : * setting the type, and allocating mempools.
1417 : */
1418 0 : int dm_table_complete(struct dm_table *t)
1419 : {
1420 0 : int r;
1421 :
1422 0 : r = dm_table_determine_type(t);
1423 0 : if (r) {
1424 0 : DMERR("unable to determine table type");
1425 0 : return r;
1426 : }
1427 :
1428 0 : r = dm_table_build_index(t);
1429 0 : if (r) {
1430 0 : DMERR("unable to build btrees");
1431 0 : return r;
1432 : }
1433 :
1434 0 : r = dm_table_register_integrity(t);
1435 0 : if (r) {
1436 0 : DMERR("could not register integrity profile.");
1437 0 : return r;
1438 : }
1439 :
1440 0 : r = dm_table_construct_keyslot_manager(t);
1441 0 : if (r) {
1442 : DMERR("could not construct keyslot manager.");
1443 : return r;
1444 : }
1445 :
1446 0 : r = dm_table_alloc_md_mempools(t, t->md);
1447 0 : if (r)
1448 0 : DMERR("unable to allocate mempools");
1449 :
1450 : return r;
1451 : }
1452 :
1453 : static DEFINE_MUTEX(_event_lock);
1454 0 : void dm_table_event_callback(struct dm_table *t,
1455 : void (*fn)(void *), void *context)
1456 : {
1457 0 : mutex_lock(&_event_lock);
1458 0 : t->event_fn = fn;
1459 0 : t->event_context = context;
1460 0 : mutex_unlock(&_event_lock);
1461 0 : }
1462 :
1463 0 : void dm_table_event(struct dm_table *t)
1464 : {
1465 0 : mutex_lock(&_event_lock);
1466 0 : if (t->event_fn)
1467 0 : t->event_fn(t->event_context);
1468 0 : mutex_unlock(&_event_lock);
1469 0 : }
1470 : EXPORT_SYMBOL(dm_table_event);
1471 :
1472 0 : inline sector_t dm_table_get_size(struct dm_table *t)
1473 : {
1474 0 : return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0;
1475 : }
1476 : EXPORT_SYMBOL(dm_table_get_size);
1477 :
1478 0 : struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index)
1479 : {
1480 0 : if (index >= t->num_targets)
1481 : return NULL;
1482 :
1483 0 : return t->targets + index;
1484 : }
1485 :
1486 : /*
1487 : * Search the btree for the correct target.
1488 : *
1489 : * Caller should check returned pointer for NULL
1490 : * to trap I/O beyond end of device.
1491 : */
1492 0 : struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector)
1493 : {
1494 0 : unsigned int l, n = 0, k = 0;
1495 0 : sector_t *node;
1496 :
1497 0 : if (unlikely(sector >= dm_table_get_size(t)))
1498 : return NULL;
1499 :
1500 0 : for (l = 0; l < t->depth; l++) {
1501 0 : n = get_child(n, k);
1502 0 : node = get_node(t, l, n);
1503 :
1504 0 : for (k = 0; k < KEYS_PER_NODE; k++)
1505 0 : if (node[k] >= sector)
1506 : break;
1507 : }
1508 :
1509 0 : return &t->targets[(KEYS_PER_NODE * n) + k];
1510 : }
1511 :
1512 : /*
1513 : * type->iterate_devices() should be called when the sanity check needs to
1514 : * iterate and check all underlying data devices. iterate_devices() will
1515 : * iterate all underlying data devices until it encounters a non-zero return
1516 : * code, returned by whether the input iterate_devices_callout_fn, or
1517 : * iterate_devices() itself internally.
1518 : *
1519 : * For some target type (e.g. dm-stripe), one call of iterate_devices() may
1520 : * iterate multiple underlying devices internally, in which case a non-zero
1521 : * return code returned by iterate_devices_callout_fn will stop the iteration
1522 : * in advance.
1523 : *
1524 : * Cases requiring _any_ underlying device supporting some kind of attribute,
1525 : * should use the iteration structure like dm_table_any_dev_attr(), or call
1526 : * it directly. @func should handle semantics of positive examples, e.g.
1527 : * capable of something.
1528 : *
1529 : * Cases requiring _all_ underlying devices supporting some kind of attribute,
1530 : * should use the iteration structure like dm_table_supports_nowait() or
1531 : * dm_table_supports_discards(). Or introduce dm_table_all_devs_attr() that
1532 : * uses an @anti_func that handle semantics of counter examples, e.g. not
1533 : * capable of something. So: return !dm_table_any_dev_attr(t, anti_func, data);
1534 : */
1535 0 : static bool dm_table_any_dev_attr(struct dm_table *t,
1536 : iterate_devices_callout_fn func, void *data)
1537 : {
1538 0 : struct dm_target *ti;
1539 0 : unsigned int i;
1540 :
1541 0 : for (i = 0; i < dm_table_get_num_targets(t); i++) {
1542 0 : ti = dm_table_get_target(t, i);
1543 :
1544 0 : if (ti->type->iterate_devices &&
1545 0 : ti->type->iterate_devices(ti, func, data))
1546 : return true;
1547 : }
1548 :
1549 : return false;
1550 : }
1551 :
1552 0 : static int count_device(struct dm_target *ti, struct dm_dev *dev,
1553 : sector_t start, sector_t len, void *data)
1554 : {
1555 0 : unsigned *num_devices = data;
1556 :
1557 0 : (*num_devices)++;
1558 :
1559 0 : return 0;
1560 : }
1561 :
1562 : /*
1563 : * Check whether a table has no data devices attached using each
1564 : * target's iterate_devices method.
1565 : * Returns false if the result is unknown because a target doesn't
1566 : * support iterate_devices.
1567 : */
1568 0 : bool dm_table_has_no_data_devices(struct dm_table *table)
1569 : {
1570 0 : struct dm_target *ti;
1571 0 : unsigned i, num_devices;
1572 :
1573 0 : for (i = 0; i < dm_table_get_num_targets(table); i++) {
1574 0 : ti = dm_table_get_target(table, i);
1575 :
1576 0 : if (!ti->type->iterate_devices)
1577 : return false;
1578 :
1579 0 : num_devices = 0;
1580 0 : ti->type->iterate_devices(ti, count_device, &num_devices);
1581 0 : if (num_devices)
1582 : return false;
1583 : }
1584 :
1585 : return true;
1586 : }
1587 :
1588 0 : static int device_not_zoned_model(struct dm_target *ti, struct dm_dev *dev,
1589 : sector_t start, sector_t len, void *data)
1590 : {
1591 0 : struct request_queue *q = bdev_get_queue(dev->bdev);
1592 0 : enum blk_zoned_model *zoned_model = data;
1593 :
1594 0 : return blk_queue_zoned_model(q) != *zoned_model;
1595 : }
1596 :
1597 0 : static bool dm_table_supports_zoned_model(struct dm_table *t,
1598 : enum blk_zoned_model zoned_model)
1599 : {
1600 0 : struct dm_target *ti;
1601 0 : unsigned i;
1602 :
1603 0 : for (i = 0; i < dm_table_get_num_targets(t); i++) {
1604 0 : ti = dm_table_get_target(t, i);
1605 :
1606 0 : if (zoned_model == BLK_ZONED_HM &&
1607 : !dm_target_supports_zoned_hm(ti->type))
1608 : return false;
1609 :
1610 0 : if (!ti->type->iterate_devices ||
1611 0 : ti->type->iterate_devices(ti, device_not_zoned_model, &zoned_model))
1612 0 : return false;
1613 : }
1614 :
1615 : return true;
1616 : }
1617 :
1618 0 : static int device_not_matches_zone_sectors(struct dm_target *ti, struct dm_dev *dev,
1619 : sector_t start, sector_t len, void *data)
1620 : {
1621 0 : struct request_queue *q = bdev_get_queue(dev->bdev);
1622 0 : unsigned int *zone_sectors = data;
1623 :
1624 0 : return blk_queue_zone_sectors(q) != *zone_sectors;
1625 : }
1626 :
1627 0 : static int validate_hardware_zoned_model(struct dm_table *table,
1628 : enum blk_zoned_model zoned_model,
1629 : unsigned int zone_sectors)
1630 : {
1631 0 : if (zoned_model == BLK_ZONED_NONE)
1632 : return 0;
1633 :
1634 0 : if (!dm_table_supports_zoned_model(table, zoned_model)) {
1635 0 : DMERR("%s: zoned model is not consistent across all devices",
1636 : dm_device_name(table->md));
1637 0 : return -EINVAL;
1638 : }
1639 :
1640 : /* Check zone size validity and compatibility */
1641 0 : if (!zone_sectors || !is_power_of_2(zone_sectors))
1642 : return -EINVAL;
1643 :
1644 0 : if (dm_table_any_dev_attr(table, device_not_matches_zone_sectors, &zone_sectors)) {
1645 0 : DMERR("%s: zone sectors is not consistent across all devices",
1646 : dm_device_name(table->md));
1647 0 : return -EINVAL;
1648 : }
1649 :
1650 : return 0;
1651 : }
1652 :
1653 : /*
1654 : * Establish the new table's queue_limits and validate them.
1655 : */
1656 0 : int dm_calculate_queue_limits(struct dm_table *table,
1657 : struct queue_limits *limits)
1658 : {
1659 0 : struct dm_target *ti;
1660 0 : struct queue_limits ti_limits;
1661 0 : unsigned i;
1662 0 : enum blk_zoned_model zoned_model = BLK_ZONED_NONE;
1663 0 : unsigned int zone_sectors = 0;
1664 :
1665 0 : blk_set_stacking_limits(limits);
1666 :
1667 0 : for (i = 0; i < dm_table_get_num_targets(table); i++) {
1668 0 : blk_set_stacking_limits(&ti_limits);
1669 :
1670 0 : ti = dm_table_get_target(table, i);
1671 :
1672 0 : if (!ti->type->iterate_devices)
1673 0 : goto combine_limits;
1674 :
1675 : /*
1676 : * Combine queue limits of all the devices this target uses.
1677 : */
1678 0 : ti->type->iterate_devices(ti, dm_set_device_limits,
1679 : &ti_limits);
1680 :
1681 0 : if (zoned_model == BLK_ZONED_NONE && ti_limits.zoned != BLK_ZONED_NONE) {
1682 : /*
1683 : * After stacking all limits, validate all devices
1684 : * in table support this zoned model and zone sectors.
1685 : */
1686 0 : zoned_model = ti_limits.zoned;
1687 0 : zone_sectors = ti_limits.chunk_sectors;
1688 : }
1689 :
1690 : /* Set I/O hints portion of queue limits */
1691 0 : if (ti->type->io_hints)
1692 0 : ti->type->io_hints(ti, &ti_limits);
1693 :
1694 : /*
1695 : * Check each device area is consistent with the target's
1696 : * overall queue limits.
1697 : */
1698 0 : if (ti->type->iterate_devices(ti, device_area_is_invalid,
1699 : &ti_limits))
1700 : return -EINVAL;
1701 :
1702 0 : combine_limits:
1703 : /*
1704 : * Merge this target's queue limits into the overall limits
1705 : * for the table.
1706 : */
1707 0 : if (blk_stack_limits(limits, &ti_limits, 0) < 0)
1708 0 : DMWARN("%s: adding target device "
1709 : "(start sect %llu len %llu) "
1710 : "caused an alignment inconsistency",
1711 : dm_device_name(table->md),
1712 : (unsigned long long) ti->begin,
1713 : (unsigned long long) ti->len);
1714 : }
1715 :
1716 : /*
1717 : * Verify that the zoned model and zone sectors, as determined before
1718 : * any .io_hints override, are the same across all devices in the table.
1719 : * - this is especially relevant if .io_hints is emulating a disk-managed
1720 : * zoned model (aka BLK_ZONED_NONE) on host-managed zoned block devices.
1721 : * BUT...
1722 : */
1723 0 : if (limits->zoned != BLK_ZONED_NONE) {
1724 : /*
1725 : * ...IF the above limits stacking determined a zoned model
1726 : * validate that all of the table's devices conform to it.
1727 : */
1728 0 : zoned_model = limits->zoned;
1729 0 : zone_sectors = limits->chunk_sectors;
1730 : }
1731 0 : if (validate_hardware_zoned_model(table, zoned_model, zone_sectors))
1732 : return -EINVAL;
1733 :
1734 0 : return validate_hardware_logical_block_alignment(table, limits);
1735 : }
1736 :
1737 : /*
1738 : * Verify that all devices have an integrity profile that matches the
1739 : * DM device's registered integrity profile. If the profiles don't
1740 : * match then unregister the DM device's integrity profile.
1741 : */
1742 0 : static void dm_table_verify_integrity(struct dm_table *t)
1743 : {
1744 0 : struct gendisk *template_disk = NULL;
1745 :
1746 0 : if (t->integrity_added)
1747 : return;
1748 :
1749 0 : if (t->integrity_supported) {
1750 : /*
1751 : * Verify that the original integrity profile
1752 : * matches all the devices in this table.
1753 : */
1754 0 : template_disk = dm_table_get_integrity_disk(t);
1755 0 : if (template_disk &&
1756 0 : blk_integrity_compare(dm_disk(t->md), template_disk) >= 0)
1757 0 : return;
1758 : }
1759 :
1760 0 : if (integrity_profile_exists(dm_disk(t->md))) {
1761 : DMWARN("%s: unable to establish an integrity profile",
1762 : dm_device_name(t->md));
1763 0 : blk_integrity_unregister(dm_disk(t->md));
1764 : }
1765 : }
1766 :
1767 0 : static int device_flush_capable(struct dm_target *ti, struct dm_dev *dev,
1768 : sector_t start, sector_t len, void *data)
1769 : {
1770 0 : unsigned long flush = (unsigned long) data;
1771 0 : struct request_queue *q = bdev_get_queue(dev->bdev);
1772 :
1773 0 : return (q->queue_flags & flush);
1774 : }
1775 :
1776 0 : static bool dm_table_supports_flush(struct dm_table *t, unsigned long flush)
1777 : {
1778 0 : struct dm_target *ti;
1779 0 : unsigned i;
1780 :
1781 : /*
1782 : * Require at least one underlying device to support flushes.
1783 : * t->devices includes internal dm devices such as mirror logs
1784 : * so we need to use iterate_devices here, which targets
1785 : * supporting flushes must provide.
1786 : */
1787 0 : for (i = 0; i < dm_table_get_num_targets(t); i++) {
1788 0 : ti = dm_table_get_target(t, i);
1789 :
1790 0 : if (!ti->num_flush_bios)
1791 0 : continue;
1792 :
1793 0 : if (ti->flush_supported)
1794 : return true;
1795 :
1796 0 : if (ti->type->iterate_devices &&
1797 0 : ti->type->iterate_devices(ti, device_flush_capable, (void *) flush))
1798 : return true;
1799 : }
1800 :
1801 : return false;
1802 : }
1803 :
1804 0 : static int device_dax_write_cache_enabled(struct dm_target *ti,
1805 : struct dm_dev *dev, sector_t start,
1806 : sector_t len, void *data)
1807 : {
1808 0 : struct dax_device *dax_dev = dev->dax_dev;
1809 :
1810 0 : if (!dax_dev)
1811 : return false;
1812 :
1813 0 : if (dax_write_cache_enabled(dax_dev))
1814 : return true;
1815 : return false;
1816 : }
1817 :
1818 0 : static int device_is_rotational(struct dm_target *ti, struct dm_dev *dev,
1819 : sector_t start, sector_t len, void *data)
1820 : {
1821 0 : struct request_queue *q = bdev_get_queue(dev->bdev);
1822 :
1823 0 : return !blk_queue_nonrot(q);
1824 : }
1825 :
1826 0 : static int device_is_not_random(struct dm_target *ti, struct dm_dev *dev,
1827 : sector_t start, sector_t len, void *data)
1828 : {
1829 0 : struct request_queue *q = bdev_get_queue(dev->bdev);
1830 :
1831 0 : return !blk_queue_add_random(q);
1832 : }
1833 :
1834 0 : static int device_not_write_same_capable(struct dm_target *ti, struct dm_dev *dev,
1835 : sector_t start, sector_t len, void *data)
1836 : {
1837 0 : struct request_queue *q = bdev_get_queue(dev->bdev);
1838 :
1839 0 : return !q->limits.max_write_same_sectors;
1840 : }
1841 :
1842 0 : static bool dm_table_supports_write_same(struct dm_table *t)
1843 : {
1844 0 : struct dm_target *ti;
1845 0 : unsigned i;
1846 :
1847 0 : for (i = 0; i < dm_table_get_num_targets(t); i++) {
1848 0 : ti = dm_table_get_target(t, i);
1849 :
1850 0 : if (!ti->num_write_same_bios)
1851 : return false;
1852 :
1853 0 : if (!ti->type->iterate_devices ||
1854 0 : ti->type->iterate_devices(ti, device_not_write_same_capable, NULL))
1855 0 : return false;
1856 : }
1857 :
1858 : return true;
1859 : }
1860 :
1861 0 : static int device_not_write_zeroes_capable(struct dm_target *ti, struct dm_dev *dev,
1862 : sector_t start, sector_t len, void *data)
1863 : {
1864 0 : struct request_queue *q = bdev_get_queue(dev->bdev);
1865 :
1866 0 : return !q->limits.max_write_zeroes_sectors;
1867 : }
1868 :
1869 0 : static bool dm_table_supports_write_zeroes(struct dm_table *t)
1870 : {
1871 0 : struct dm_target *ti;
1872 0 : unsigned i = 0;
1873 :
1874 0 : while (i < dm_table_get_num_targets(t)) {
1875 0 : ti = dm_table_get_target(t, i++);
1876 :
1877 0 : if (!ti->num_write_zeroes_bios)
1878 : return false;
1879 :
1880 0 : if (!ti->type->iterate_devices ||
1881 0 : ti->type->iterate_devices(ti, device_not_write_zeroes_capable, NULL))
1882 0 : return false;
1883 : }
1884 :
1885 : return true;
1886 : }
1887 :
1888 0 : static int device_not_nowait_capable(struct dm_target *ti, struct dm_dev *dev,
1889 : sector_t start, sector_t len, void *data)
1890 : {
1891 0 : struct request_queue *q = bdev_get_queue(dev->bdev);
1892 :
1893 0 : return !blk_queue_nowait(q);
1894 : }
1895 :
1896 0 : static bool dm_table_supports_nowait(struct dm_table *t)
1897 : {
1898 0 : struct dm_target *ti;
1899 0 : unsigned i = 0;
1900 :
1901 0 : while (i < dm_table_get_num_targets(t)) {
1902 0 : ti = dm_table_get_target(t, i++);
1903 :
1904 0 : if (!dm_target_supports_nowait(ti->type))
1905 : return false;
1906 :
1907 0 : if (!ti->type->iterate_devices ||
1908 0 : ti->type->iterate_devices(ti, device_not_nowait_capable, NULL))
1909 0 : return false;
1910 : }
1911 :
1912 : return true;
1913 : }
1914 :
1915 0 : static int device_not_discard_capable(struct dm_target *ti, struct dm_dev *dev,
1916 : sector_t start, sector_t len, void *data)
1917 : {
1918 0 : struct request_queue *q = bdev_get_queue(dev->bdev);
1919 :
1920 0 : return !blk_queue_discard(q);
1921 : }
1922 :
1923 0 : static bool dm_table_supports_discards(struct dm_table *t)
1924 : {
1925 0 : struct dm_target *ti;
1926 0 : unsigned i;
1927 :
1928 0 : for (i = 0; i < dm_table_get_num_targets(t); i++) {
1929 0 : ti = dm_table_get_target(t, i);
1930 :
1931 0 : if (!ti->num_discard_bios)
1932 : return false;
1933 :
1934 : /*
1935 : * Either the target provides discard support (as implied by setting
1936 : * 'discards_supported') or it relies on _all_ data devices having
1937 : * discard support.
1938 : */
1939 0 : if (!ti->discards_supported &&
1940 0 : (!ti->type->iterate_devices ||
1941 0 : ti->type->iterate_devices(ti, device_not_discard_capable, NULL)))
1942 0 : return false;
1943 : }
1944 :
1945 : return true;
1946 : }
1947 :
1948 0 : static int device_not_secure_erase_capable(struct dm_target *ti,
1949 : struct dm_dev *dev, sector_t start,
1950 : sector_t len, void *data)
1951 : {
1952 0 : struct request_queue *q = bdev_get_queue(dev->bdev);
1953 :
1954 0 : return !blk_queue_secure_erase(q);
1955 : }
1956 :
1957 0 : static bool dm_table_supports_secure_erase(struct dm_table *t)
1958 : {
1959 0 : struct dm_target *ti;
1960 0 : unsigned int i;
1961 :
1962 0 : for (i = 0; i < dm_table_get_num_targets(t); i++) {
1963 0 : ti = dm_table_get_target(t, i);
1964 :
1965 0 : if (!ti->num_secure_erase_bios)
1966 : return false;
1967 :
1968 0 : if (!ti->type->iterate_devices ||
1969 0 : ti->type->iterate_devices(ti, device_not_secure_erase_capable, NULL))
1970 0 : return false;
1971 : }
1972 :
1973 : return true;
1974 : }
1975 :
1976 0 : static int device_requires_stable_pages(struct dm_target *ti,
1977 : struct dm_dev *dev, sector_t start,
1978 : sector_t len, void *data)
1979 : {
1980 0 : struct request_queue *q = bdev_get_queue(dev->bdev);
1981 :
1982 0 : return blk_queue_stable_writes(q);
1983 : }
1984 :
1985 0 : void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
1986 : struct queue_limits *limits)
1987 : {
1988 0 : bool wc = false, fua = false;
1989 0 : int page_size = PAGE_SIZE;
1990 :
1991 : /*
1992 : * Copy table's limits to the DM device's request_queue
1993 : */
1994 0 : q->limits = *limits;
1995 :
1996 0 : if (dm_table_supports_nowait(t))
1997 0 : blk_queue_flag_set(QUEUE_FLAG_NOWAIT, q);
1998 : else
1999 0 : blk_queue_flag_clear(QUEUE_FLAG_NOWAIT, q);
2000 :
2001 0 : if (!dm_table_supports_discards(t)) {
2002 0 : blk_queue_flag_clear(QUEUE_FLAG_DISCARD, q);
2003 : /* Must also clear discard limits... */
2004 0 : q->limits.max_discard_sectors = 0;
2005 0 : q->limits.max_hw_discard_sectors = 0;
2006 0 : q->limits.discard_granularity = 0;
2007 0 : q->limits.discard_alignment = 0;
2008 0 : q->limits.discard_misaligned = 0;
2009 : } else
2010 0 : blk_queue_flag_set(QUEUE_FLAG_DISCARD, q);
2011 :
2012 0 : if (dm_table_supports_secure_erase(t))
2013 0 : blk_queue_flag_set(QUEUE_FLAG_SECERASE, q);
2014 :
2015 0 : if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_WC))) {
2016 0 : wc = true;
2017 0 : if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_FUA)))
2018 0 : fua = true;
2019 : }
2020 0 : blk_queue_write_cache(q, wc, fua);
2021 :
2022 0 : if (dm_table_supports_dax(t, device_not_dax_capable, &page_size)) {
2023 0 : blk_queue_flag_set(QUEUE_FLAG_DAX, q);
2024 0 : if (dm_table_supports_dax(t, device_not_dax_synchronous_capable, NULL))
2025 0 : set_dax_synchronous(t->md->dax_dev);
2026 : }
2027 : else
2028 0 : blk_queue_flag_clear(QUEUE_FLAG_DAX, q);
2029 :
2030 0 : if (dm_table_any_dev_attr(t, device_dax_write_cache_enabled, NULL))
2031 0 : dax_write_cache(t->md->dax_dev, true);
2032 :
2033 : /* Ensure that all underlying devices are non-rotational. */
2034 0 : if (dm_table_any_dev_attr(t, device_is_rotational, NULL))
2035 0 : blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
2036 : else
2037 0 : blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
2038 :
2039 0 : if (!dm_table_supports_write_same(t))
2040 0 : q->limits.max_write_same_sectors = 0;
2041 0 : if (!dm_table_supports_write_zeroes(t))
2042 0 : q->limits.max_write_zeroes_sectors = 0;
2043 :
2044 0 : dm_table_verify_integrity(t);
2045 :
2046 : /*
2047 : * Some devices don't use blk_integrity but still want stable pages
2048 : * because they do their own checksumming.
2049 : * If any underlying device requires stable pages, a table must require
2050 : * them as well. Only targets that support iterate_devices are considered:
2051 : * don't want error, zero, etc to require stable pages.
2052 : */
2053 0 : if (dm_table_any_dev_attr(t, device_requires_stable_pages, NULL))
2054 0 : blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, q);
2055 : else
2056 0 : blk_queue_flag_clear(QUEUE_FLAG_STABLE_WRITES, q);
2057 :
2058 : /*
2059 : * Determine whether or not this queue's I/O timings contribute
2060 : * to the entropy pool, Only request-based targets use this.
2061 : * Clear QUEUE_FLAG_ADD_RANDOM if any underlying device does not
2062 : * have it set.
2063 : */
2064 0 : if (blk_queue_add_random(q) &&
2065 0 : dm_table_any_dev_attr(t, device_is_not_random, NULL))
2066 0 : blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
2067 :
2068 : /*
2069 : * For a zoned target, the number of zones should be updated for the
2070 : * correct value to be exposed in sysfs queue/nr_zones. For a BIO based
2071 : * target, this is all that is needed.
2072 : */
2073 : #ifdef CONFIG_BLK_DEV_ZONED
2074 : if (blk_queue_is_zoned(q)) {
2075 : WARN_ON_ONCE(queue_is_mq(q));
2076 : q->nr_zones = blkdev_nr_zones(t->md->disk);
2077 : }
2078 : #endif
2079 :
2080 0 : dm_update_keyslot_manager(q, t);
2081 0 : blk_queue_update_readahead(q);
2082 0 : }
2083 :
2084 0 : unsigned int dm_table_get_num_targets(struct dm_table *t)
2085 : {
2086 0 : return t->num_targets;
2087 : }
2088 :
2089 0 : struct list_head *dm_table_get_devices(struct dm_table *t)
2090 : {
2091 0 : return &t->devices;
2092 : }
2093 :
2094 0 : fmode_t dm_table_get_mode(struct dm_table *t)
2095 : {
2096 0 : return t->mode;
2097 : }
2098 : EXPORT_SYMBOL(dm_table_get_mode);
2099 :
2100 : enum suspend_mode {
2101 : PRESUSPEND,
2102 : PRESUSPEND_UNDO,
2103 : POSTSUSPEND,
2104 : };
2105 :
2106 0 : static void suspend_targets(struct dm_table *t, enum suspend_mode mode)
2107 : {
2108 0 : int i = t->num_targets;
2109 0 : struct dm_target *ti = t->targets;
2110 :
2111 0 : lockdep_assert_held(&t->md->suspend_lock);
2112 :
2113 0 : while (i--) {
2114 0 : switch (mode) {
2115 0 : case PRESUSPEND:
2116 0 : if (ti->type->presuspend)
2117 0 : ti->type->presuspend(ti);
2118 : break;
2119 0 : case PRESUSPEND_UNDO:
2120 0 : if (ti->type->presuspend_undo)
2121 0 : ti->type->presuspend_undo(ti);
2122 : break;
2123 0 : case POSTSUSPEND:
2124 0 : if (ti->type->postsuspend)
2125 0 : ti->type->postsuspend(ti);
2126 : break;
2127 : }
2128 0 : ti++;
2129 : }
2130 0 : }
2131 :
2132 0 : void dm_table_presuspend_targets(struct dm_table *t)
2133 : {
2134 0 : if (!t)
2135 : return;
2136 :
2137 0 : suspend_targets(t, PRESUSPEND);
2138 : }
2139 :
2140 0 : void dm_table_presuspend_undo_targets(struct dm_table *t)
2141 : {
2142 0 : if (!t)
2143 : return;
2144 :
2145 0 : suspend_targets(t, PRESUSPEND_UNDO);
2146 : }
2147 :
2148 0 : void dm_table_postsuspend_targets(struct dm_table *t)
2149 : {
2150 0 : if (!t)
2151 : return;
2152 :
2153 0 : suspend_targets(t, POSTSUSPEND);
2154 : }
2155 :
2156 0 : int dm_table_resume_targets(struct dm_table *t)
2157 : {
2158 0 : int i, r = 0;
2159 :
2160 0 : lockdep_assert_held(&t->md->suspend_lock);
2161 :
2162 0 : for (i = 0; i < t->num_targets; i++) {
2163 0 : struct dm_target *ti = t->targets + i;
2164 :
2165 0 : if (!ti->type->preresume)
2166 0 : continue;
2167 :
2168 0 : r = ti->type->preresume(ti);
2169 0 : if (r) {
2170 0 : DMERR("%s: %s: preresume failed, error = %d",
2171 : dm_device_name(t->md), ti->type->name, r);
2172 0 : return r;
2173 : }
2174 : }
2175 :
2176 0 : for (i = 0; i < t->num_targets; i++) {
2177 0 : struct dm_target *ti = t->targets + i;
2178 :
2179 0 : if (ti->type->resume)
2180 0 : ti->type->resume(ti);
2181 : }
2182 :
2183 : return 0;
2184 : }
2185 :
2186 0 : struct mapped_device *dm_table_get_md(struct dm_table *t)
2187 : {
2188 0 : return t->md;
2189 : }
2190 : EXPORT_SYMBOL(dm_table_get_md);
2191 :
2192 0 : const char *dm_table_device_name(struct dm_table *t)
2193 : {
2194 0 : return dm_device_name(t->md);
2195 : }
2196 : EXPORT_SYMBOL_GPL(dm_table_device_name);
2197 :
2198 0 : void dm_table_run_md_queue_async(struct dm_table *t)
2199 : {
2200 0 : if (!dm_table_request_based(t))
2201 : return;
2202 :
2203 0 : if (t->md->queue)
2204 0 : blk_mq_run_hw_queues(t->md->queue, true);
2205 : }
2206 : EXPORT_SYMBOL(dm_table_run_md_queue_async);
2207 :
|
