1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Code extracted from drivers/block/genhd.c
4 * Copyright (C) 1991-1998 Linus Torvalds
5 * Re-organised Feb 1998 Russell King
6 *
7 * We now have independent partition support from the
8 * block drivers, which allows all the partition code to
9 * be grouped in one location, and it to be mostly self
10 * contained.
11 */
12
13#include <linux/init.h>
14#include <linux/module.h>
15#include <linux/fs.h>
16#include <linux/slab.h>
17#include <linux/kmod.h>
18#include <linux/ctype.h>
19#include <linux/genhd.h>
20#include <linux/blktrace_api.h>
21
22#include "partitions/check.h"
23
24#ifdef CONFIG_BLK_DEV_MD
25extern void md_autodetect_dev(dev_t dev);
26#endif
27
28/*
29 * disk_name() is used by partition check code and the genhd driver.
30 * It formats the devicename of the indicated disk into
31 * the supplied buffer (of size at least 32), and returns
32 * a pointer to that same buffer (for convenience).
33 */
34
35char *disk_name(struct gendisk *hd, int partno, char *buf)
36{
37 if (!partno)
38 snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name);
39 else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1]))
40 snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno);
41 else
42 snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno);
43
44 return buf;
45}
46
47const char *bdevname(struct block_device *bdev, char *buf)
48{
49 return disk_name(bdev->bd_disk, bdev->bd_part->partno, buf);
50}
51
52EXPORT_SYMBOL(bdevname);
53
54const char *bio_devname(struct bio *bio, char *buf)
55{
56 return disk_name(bio->bi_disk, bio->bi_partno, buf);
57}
58EXPORT_SYMBOL(bio_devname);
59
60/*
61 * There's very little reason to use this, you should really
62 * have a struct block_device just about everywhere and use
63 * bdevname() instead.
64 */
65const char *__bdevname(dev_t dev, char *buffer)
66{
67 scnprintf(buffer, BDEVNAME_SIZE, "unknown-block(%u,%u)",
68 MAJOR(dev), MINOR(dev));
69 return buffer;
70}
71
72EXPORT_SYMBOL(__bdevname);
73
74static ssize_t part_partition_show(struct device *dev,
75 struct device_attribute *attr, char *buf)
76{
77 struct hd_struct *p = dev_to_part(dev);
78
79 return sprintf(buf, "%d\n", p->partno);
80}
81
82static ssize_t part_start_show(struct device *dev,
83 struct device_attribute *attr, char *buf)
84{
85 struct hd_struct *p = dev_to_part(dev);
86
87 return sprintf(buf, "%llu\n",(unsigned long long)p->start_sect);
88}
89
90ssize_t part_size_show(struct device *dev,
91 struct device_attribute *attr, char *buf)
92{
93 struct hd_struct *p = dev_to_part(dev);
94 return sprintf(buf, "%llu\n",(unsigned long long)part_nr_sects_read(p));
95}
96
97static ssize_t part_ro_show(struct device *dev,
98 struct device_attribute *attr, char *buf)
99{
100 struct hd_struct *p = dev_to_part(dev);
101 return sprintf(buf, "%d\n", p->policy ? 1 : 0);
102}
103
104static ssize_t part_alignment_offset_show(struct device *dev,
105 struct device_attribute *attr, char *buf)
106{
107 struct hd_struct *p = dev_to_part(dev);
108 return sprintf(buf, "%llu\n", (unsigned long long)p->alignment_offset);
109}
110
111static ssize_t part_discard_alignment_show(struct device *dev,
112 struct device_attribute *attr, char *buf)
113{
114 struct hd_struct *p = dev_to_part(dev);
115 return sprintf(buf, "%u\n", p->discard_alignment);
116}
117
118ssize_t part_stat_show(struct device *dev,
119 struct device_attribute *attr, char *buf)
120{
121 struct hd_struct *p = dev_to_part(dev);
122 struct request_queue *q = part_to_disk(p)->queue;
123 unsigned int inflight;
124
125 inflight = part_in_flight(q, p);
126 return sprintf(buf,
127 "%8lu %8lu %8llu %8u "
128 "%8lu %8lu %8llu %8u "
129 "%8u %8u %8u "
130 "%8lu %8lu %8llu %8u"
131 "\n",
132 part_stat_read(p, ios[STAT_READ]),
133 part_stat_read(p, merges[STAT_READ]),
134 (unsigned long long)part_stat_read(p, sectors[STAT_READ]),
135 (unsigned int)part_stat_read_msecs(p, STAT_READ),
136 part_stat_read(p, ios[STAT_WRITE]),
137 part_stat_read(p, merges[STAT_WRITE]),
138 (unsigned long long)part_stat_read(p, sectors[STAT_WRITE]),
139 (unsigned int)part_stat_read_msecs(p, STAT_WRITE),
140 inflight,
141 jiffies_to_msecs(part_stat_read(p, io_ticks)),
142 jiffies_to_msecs(part_stat_read(p, time_in_queue)),
143 part_stat_read(p, ios[STAT_DISCARD]),
144 part_stat_read(p, merges[STAT_DISCARD]),
145 (unsigned long long)part_stat_read(p, sectors[STAT_DISCARD]),
146 (unsigned int)part_stat_read_msecs(p, STAT_DISCARD));
147}
148
149ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
150 char *buf)
151{
152 struct hd_struct *p = dev_to_part(dev);
153 struct request_queue *q = part_to_disk(p)->queue;
154 unsigned int inflight[2];
155
156 part_in_flight_rw(q, p, inflight);
157 return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
158}
159
160#ifdef CONFIG_FAIL_MAKE_REQUEST
161ssize_t part_fail_show(struct device *dev,
162 struct device_attribute *attr, char *buf)
163{
164 struct hd_struct *p = dev_to_part(dev);
165
166 return sprintf(buf, "%d\n", p->make_it_fail);
167}
168
169ssize_t part_fail_store(struct device *dev,
170 struct device_attribute *attr,
171 const char *buf, size_t count)
172{
173 struct hd_struct *p = dev_to_part(dev);
174 int i;
175
176 if (count > 0 && sscanf(buf, "%d", &i) > 0)
177 p->make_it_fail = (i == 0) ? 0 : 1;
178
179 return count;
180}
181#endif
182
183static DEVICE_ATTR(partition, 0444, part_partition_show, NULL);
184static DEVICE_ATTR(start, 0444, part_start_show, NULL);
185static DEVICE_ATTR(size, 0444, part_size_show, NULL);
186static DEVICE_ATTR(ro, 0444, part_ro_show, NULL);
187static DEVICE_ATTR(alignment_offset, 0444, part_alignment_offset_show, NULL);
188static DEVICE_ATTR(discard_alignment, 0444, part_discard_alignment_show, NULL);
189static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
190static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
191#ifdef CONFIG_FAIL_MAKE_REQUEST
192static struct device_attribute dev_attr_fail =
193 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
194#endif
195
196static struct attribute *part_attrs[] = {
197 &dev_attr_partition.attr,
198 &dev_attr_start.attr,
199 &dev_attr_size.attr,
200 &dev_attr_ro.attr,
201 &dev_attr_alignment_offset.attr,
202 &dev_attr_discard_alignment.attr,
203 &dev_attr_stat.attr,
204 &dev_attr_inflight.attr,
205#ifdef CONFIG_FAIL_MAKE_REQUEST
206 &dev_attr_fail.attr,
207#endif
208 NULL
209};
210
211static struct attribute_group part_attr_group = {
212 .attrs = part_attrs,
213};
214
215static const struct attribute_group *part_attr_groups[] = {
216 &part_attr_group,
217#ifdef CONFIG_BLK_DEV_IO_TRACE
218 &blk_trace_attr_group,
219#endif
220 NULL
221};
222
223static void part_release(struct device *dev)
224{
225 struct hd_struct *p = dev_to_part(dev);
226 blk_free_devt(dev->devt);
227 hd_free_part(p);
228 kfree(p);
229}
230
231static int part_uevent(struct device *dev, struct kobj_uevent_env *env)
232{
233 struct hd_struct *part = dev_to_part(dev);
234
235 add_uevent_var(env, "PARTN=%u", part->partno);
236 if (part->info && part->info->volname[0])
237 add_uevent_var(env, "PARTNAME=%s", part->info->volname);
238 return 0;
239}
240
241struct device_type part_type = {
242 .name = "partition",
243 .groups = part_attr_groups,
244 .release = part_release,
245 .uevent = part_uevent,
246};
247
248static void delete_partition_work_fn(struct work_struct *work)
249{
250 struct hd_struct *part = container_of(to_rcu_work(work), struct hd_struct,
251 rcu_work);
252
253 part->start_sect = 0;
254 part->nr_sects = 0;
255 part_stat_set_all(part, 0);
256 put_device(part_to_dev(part));
257}
258
259void __delete_partition(struct percpu_ref *ref)
260{
261 struct hd_struct *part = container_of(ref, struct hd_struct, ref);
262 INIT_RCU_WORK(&part->rcu_work, delete_partition_work_fn);
263 queue_rcu_work(system_wq, &part->rcu_work);
264}
265
266/*
267 * Must be called either with bd_mutex held, before a disk can be opened or
268 * after all disk users are gone.
269 */
270void delete_partition(struct gendisk *disk, int partno)
271{
272 struct disk_part_tbl *ptbl =
273 rcu_dereference_protected(disk->part_tbl, 1);
274 struct hd_struct *part;
275
276 if (partno >= ptbl->len)
277 return;
278
279 part = rcu_dereference_protected(ptbl->part[partno], 1);
280 if (!part)
281 return;
282
283 rcu_assign_pointer(ptbl->part[partno], NULL);
284 rcu_assign_pointer(ptbl->last_lookup, NULL);
285 kobject_put(part->holder_dir);
286 device_del(part_to_dev(part));
287
288 hd_struct_kill(part);
289}
290
291static ssize_t whole_disk_show(struct device *dev,
292 struct device_attribute *attr, char *buf)
293{
294 return 0;
295}
296static DEVICE_ATTR(whole_disk, 0444, whole_disk_show, NULL);
297
298/*
299 * Must be called either with bd_mutex held, before a disk can be opened or
300 * after all disk users are gone.
301 */
302struct hd_struct *add_partition(struct gendisk *disk, int partno,
303 sector_t start, sector_t len, int flags,
304 struct partition_meta_info *info)
305{
306 struct hd_struct *p;
307 dev_t devt = MKDEV(0, 0);
308 struct device *ddev = disk_to_dev(disk);
309 struct device *pdev;
310 struct disk_part_tbl *ptbl;
311 const char *dname;
312 int err;
313
314 err = disk_expand_part_tbl(disk, partno);
315 if (err)
316 return ERR_PTR(err);
317 ptbl = rcu_dereference_protected(disk->part_tbl, 1);
318
319 if (ptbl->part[partno])
320 return ERR_PTR(-EBUSY);
321
322 p = kzalloc(sizeof(*p), GFP_KERNEL);
323 if (!p)
324 return ERR_PTR(-EBUSY);
325
326 if (!init_part_stats(p)) {
327 err = -ENOMEM;
328 goto out_free;
329 }
330
331 seqcount_init(&p->nr_sects_seq);
332 pdev = part_to_dev(p);
333
334 p->start_sect = start;
335 p->alignment_offset =
336 queue_limit_alignment_offset(&disk->queue->limits, start);
337 p->discard_alignment =
338 queue_limit_discard_alignment(&disk->queue->limits, start);
339 p->nr_sects = len;
340 p->partno = partno;
341 p->policy = get_disk_ro(disk);
342
343 if (info) {
344 struct partition_meta_info *pinfo = alloc_part_info(disk);
345 if (!pinfo) {
346 err = -ENOMEM;
347 goto out_free_stats;
348 }
349 memcpy(pinfo, info, sizeof(*info));
350 p->info = pinfo;
351 }
352
353 dname = dev_name(ddev);
354 if (isdigit(dname[strlen(dname) - 1]))
355 dev_set_name(pdev, "%sp%d", dname, partno);
356 else
357 dev_set_name(pdev, "%s%d", dname, partno);
358
359 device_initialize(pdev);
360 pdev->class = &block_class;
361 pdev->type = &part_type;
362 pdev->parent = ddev;
363
364 err = blk_alloc_devt(p, &devt);
365 if (err)
366 goto out_free_info;
367 pdev->devt = devt;
368
369 /* delay uevent until 'holders' subdir is created */
370 dev_set_uevent_suppress(pdev, 1);
371 err = device_add(pdev);
372 if (err)
373 goto out_put;
374
375 err = -ENOMEM;
376 p->holder_dir = kobject_create_and_add("holders", &pdev->kobj);
377 if (!p->holder_dir)
378 goto out_del;
379
380 dev_set_uevent_suppress(pdev, 0);
381 if (flags & ADDPART_FLAG_WHOLEDISK) {
382 err = device_create_file(pdev, &dev_attr_whole_disk);
383 if (err)
384 goto out_del;
385 }
386
387 err = hd_ref_init(p);
388 if (err) {
389 if (flags & ADDPART_FLAG_WHOLEDISK)
390 goto out_remove_file;
391 goto out_del;
392 }
393
394 /* everything is up and running, commence */
395 rcu_assign_pointer(ptbl->part[partno], p);
396
397 /* suppress uevent if the disk suppresses it */
398 if (!dev_get_uevent_suppress(ddev))
399 kobject_uevent(&pdev->kobj, KOBJ_ADD);
400 return p;
401
402out_free_info:
403 free_part_info(p);
404out_free_stats:
405 free_part_stats(p);
406out_free:
407 kfree(p);
408 return ERR_PTR(err);
409out_remove_file:
410 device_remove_file(pdev, &dev_attr_whole_disk);
411out_del:
412 kobject_put(p->holder_dir);
413 device_del(pdev);
414out_put:
415 put_device(pdev);
416 return ERR_PTR(err);
417}
418
419static bool disk_unlock_native_capacity(struct gendisk *disk)
420{
421 const struct block_device_operations *bdops = disk->fops;
422
423 if (bdops->unlock_native_capacity &&
424 !(disk->flags & GENHD_FL_NATIVE_CAPACITY)) {
425 printk(KERN_CONT "enabling native capacity\n");
426 bdops->unlock_native_capacity(disk);
427 disk->flags |= GENHD_FL_NATIVE_CAPACITY;
428 return true;
429 } else {
430 printk(KERN_CONT "truncated\n");
431 return false;
432 }
433}
434
435static int drop_partitions(struct gendisk *disk, struct block_device *bdev)
436{
437 struct disk_part_iter piter;
438 struct hd_struct *part;
439 int res;
440
441 if (bdev->bd_part_count || bdev->bd_super)
442 return -EBUSY;
443 res = invalidate_partition(disk, 0);
444 if (res)
445 return res;
446
447 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
448 while ((part = disk_part_iter_next(&piter)))
449 delete_partition(disk, part->partno);
450 disk_part_iter_exit(&piter);
451
452 return 0;
453}
454
455static bool part_zone_aligned(struct gendisk *disk,
456 struct block_device *bdev,
457 sector_t from, sector_t size)
458{
459 unsigned int zone_sectors = bdev_zone_sectors(bdev);
460
461 /*
462 * If this function is called, then the disk is a zoned block device
463 * (host-aware or host-managed). This can be detected even if the
464 * zoned block device support is disabled (CONFIG_BLK_DEV_ZONED not
465 * set). In this case, however, only host-aware devices will be seen
466 * as a block device is not created for host-managed devices. Without
467 * zoned block device support, host-aware drives can still be used as
468 * regular block devices (no zone operation) and their zone size will
469 * be reported as 0. Allow this case.
470 */
471 if (!zone_sectors)
472 return true;
473
474 /*
475 * Check partition start and size alignement. If the drive has a
476 * smaller last runt zone, ignore it and allow the partition to
477 * use it. Check the zone size too: it should be a power of 2 number
478 * of sectors.
479 */
480 if (WARN_ON_ONCE(!is_power_of_2(zone_sectors))) {
481 u32 rem;
482
483 div_u64_rem(from, zone_sectors, &rem);
484 if (rem)
485 return false;
486 if ((from + size) < get_capacity(disk)) {
487 div_u64_rem(size, zone_sectors, &rem);
488 if (rem)
489 return false;
490 }
491
492 } else {
493
494 if (from & (zone_sectors - 1))
495 return false;
496 if ((from + size) < get_capacity(disk) &&
497 (size & (zone_sectors - 1)))
498 return false;
499
500 }
501
502 return true;
503}
504
505int rescan_partitions(struct gendisk *disk, struct block_device *bdev)
506{
507 struct parsed_partitions *state = NULL;
508 struct hd_struct *part;
509 int p, highest, res;
510rescan:
511 if (state && !IS_ERR(state)) {
512 free_partitions(state);
513 state = NULL;
514 }
515
516 res = drop_partitions(disk, bdev);
517 if (res)
518 return res;
519
520 if (disk->fops->revalidate_disk)
521 disk->fops->revalidate_disk(disk);
522 check_disk_size_change(disk, bdev, true);
523 bdev->bd_invalidated = 0;
524 if (!get_capacity(disk) || !(state = check_partition(disk, bdev)))
525 return 0;
526 if (IS_ERR(state)) {
527 /*
528 * I/O error reading the partition table. If any
529 * partition code tried to read beyond EOD, retry
530 * after unlocking native capacity.
531 */
532 if (PTR_ERR(state) == -ENOSPC) {
533 printk(KERN_WARNING "%s: partition table beyond EOD, ",
534 disk->disk_name);
535 if (disk_unlock_native_capacity(disk))
536 goto rescan;
537 }
538 return -EIO;
539 }
540 /*
541 * If any partition code tried to read beyond EOD, try
542 * unlocking native capacity even if partition table is
543 * successfully read as we could be missing some partitions.
544 */
545 if (state->access_beyond_eod) {
546 printk(KERN_WARNING
547 "%s: partition table partially beyond EOD, ",
548 disk->disk_name);
549 if (disk_unlock_native_capacity(disk))
550 goto rescan;
551 }
552
553 /* tell userspace that the media / partition table may have changed */
554 kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
555
556 /* Detect the highest partition number and preallocate
557 * disk->part_tbl. This is an optimization and not strictly
558 * necessary.
559 */
560 for (p = 1, highest = 0; p < state->limit; p++)
561 if (state->parts[p].size)
562 highest = p;
563
564 disk_expand_part_tbl(disk, highest);
565
566 /* add partitions */
567 for (p = 1; p < state->limit; p++) {
568 sector_t size, from;
569
570 size = state->parts[p].size;
571 if (!size)
572 continue;
573
574 from = state->parts[p].from;
575 if (from >= get_capacity(disk)) {
576 printk(KERN_WARNING
577 "%s: p%d start %llu is beyond EOD, ",
578 disk->disk_name, p, (unsigned long long) from);
579 if (disk_unlock_native_capacity(disk))
580 goto rescan;
581 continue;
582 }
583
584 if (from + size > get_capacity(disk)) {
585 printk(KERN_WARNING
586 "%s: p%d size %llu extends beyond EOD, ",
587 disk->disk_name, p, (unsigned long long) size);
588
589 if (disk_unlock_native_capacity(disk)) {
590 /* free state and restart */
591 goto rescan;
592 } else {
593 /*
594 * we can not ignore partitions of broken tables
595 * created by for example camera firmware, but
596 * we limit them to the end of the disk to avoid
597 * creating invalid block devices
598 */
599 size = get_capacity(disk) - from;
600 }
601 }
602
603 /*
604 * On a zoned block device, partitions should be aligned on the
605 * device zone size (i.e. zone boundary crossing not allowed).
606 * Otherwise, resetting the write pointer of the last zone of
607 * one partition may impact the following partition.
608 */
609 if (bdev_is_zoned(bdev) &&
610 !part_zone_aligned(disk, bdev, from, size)) {
611 printk(KERN_WARNING
612 "%s: p%d start %llu+%llu is not zone aligned\n",
613 disk->disk_name, p, (unsigned long long) from,
614 (unsigned long long) size);
615 continue;
616 }
617
618 part = add_partition(disk, p, from, size,
619 state->parts[p].flags,
620 &state->parts[p].info);
621 if (IS_ERR(part)) {
622 printk(KERN_ERR " %s: p%d could not be added: %ld\n",
623 disk->disk_name, p, -PTR_ERR(part));
624 continue;
625 }
626#ifdef CONFIG_BLK_DEV_MD
627 if (state->parts[p].flags & ADDPART_FLAG_RAID)
628 md_autodetect_dev(part_to_dev(part)->devt);
629#endif
630 }
631 free_partitions(state);
632 return 0;
633}
634
635int invalidate_partitions(struct gendisk *disk, struct block_device *bdev)
636{
637 int res;
638
639 if (!bdev->bd_invalidated)
640 return 0;
641
642 res = drop_partitions(disk, bdev);
643 if (res)
644 return res;
645
646 set_capacity(disk, 0);
647 check_disk_size_change(disk, bdev, false);
648 bdev->bd_invalidated = 0;
649 /* tell userspace that the media / partition table may have changed */
650 kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
651
652 return 0;
653}
654
655unsigned char *read_dev_sector(struct block_device *bdev, sector_t n, Sector *p)
656{
657 struct address_space *mapping = bdev->bd_inode->i_mapping;
658 struct page *page;
659
660 page = read_mapping_page(mapping, (pgoff_t)(n >> (PAGE_SHIFT-9)), NULL);
661 if (!IS_ERR(page)) {
662 if (PageError(page))
663 goto fail;
664 p->v = page;
665 return (unsigned char *)page_address(page) + ((n & ((1 << (PAGE_SHIFT - 9)) - 1)) << 9);
666fail:
667 put_page(page);
668 }
669 p->v = NULL;
670 return NULL;
671}
672
673EXPORT_SYMBOL(read_dev_sector);
674