1 // SPDX-License-Identifier: GPL-2.0
3 * Code extracted from drivers/block/genhd.c
4 * Copyright (C) 1991-1998 Linus Torvalds
5 * Re-organised Feb 1998 Russell King
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
13 #include <linux/init.h>
14 #include <linux/module.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>
22 #include "partitions/check.h"
24 #ifdef CONFIG_BLK_DEV_MD
25 extern void md_autodetect_dev(dev_t dev);
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).
35 char *disk_name(struct gendisk *hd, int partno, char *buf)
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);
42 snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno);
47 const char *bdevname(struct block_device *bdev, char *buf)
49 return disk_name(bdev->bd_disk, bdev->bd_part->partno, buf);
52 EXPORT_SYMBOL(bdevname);
54 const char *bio_devname(struct bio *bio, char *buf)
56 return disk_name(bio->bi_disk, bio->bi_partno, buf);
58 EXPORT_SYMBOL(bio_devname);
61 * There's very little reason to use this, you should really
62 * have a struct block_device just about everywhere and use
65 const char *__bdevname(dev_t dev, char *buffer)
67 scnprintf(buffer, BDEVNAME_SIZE, "unknown-block(%u,%u)",
68 MAJOR(dev), MINOR(dev));
72 EXPORT_SYMBOL(__bdevname);
74 static ssize_t part_partition_show(struct device *dev,
75 struct device_attribute *attr, char *buf)
77 struct hd_struct *p = dev_to_part(dev);
79 return sprintf(buf, "%d\n", p->partno);
82 static ssize_t part_start_show(struct device *dev,
83 struct device_attribute *attr, char *buf)
85 struct hd_struct *p = dev_to_part(dev);
87 return sprintf(buf, "%llu\n",(unsigned long long)p->start_sect);
90 ssize_t part_size_show(struct device *dev,
91 struct device_attribute *attr, char *buf)
93 struct hd_struct *p = dev_to_part(dev);
94 return sprintf(buf, "%llu\n",(unsigned long long)part_nr_sects_read(p));
97 static ssize_t part_ro_show(struct device *dev,
98 struct device_attribute *attr, char *buf)
100 struct hd_struct *p = dev_to_part(dev);
101 return sprintf(buf, "%d\n", p->policy ? 1 : 0);
104 static ssize_t part_alignment_offset_show(struct device *dev,
105 struct device_attribute *attr, char *buf)
107 struct hd_struct *p = dev_to_part(dev);
108 return sprintf(buf, "%llu\n", (unsigned long long)p->alignment_offset);
111 static ssize_t part_discard_alignment_show(struct device *dev,
112 struct device_attribute *attr, char *buf)
114 struct hd_struct *p = dev_to_part(dev);
115 return sprintf(buf, "%u\n", p->discard_alignment);
118 ssize_t part_stat_show(struct device *dev,
119 struct device_attribute *attr, char *buf)
121 struct hd_struct *p = dev_to_part(dev);
122 struct request_queue *q = part_to_disk(p)->queue;
123 unsigned int inflight[2];
126 cpu = part_stat_lock();
127 part_round_stats(q, cpu, p);
129 part_in_flight(q, p, inflight);
131 "%8lu %8lu %8llu %8u "
132 "%8lu %8lu %8llu %8u "
135 part_stat_read(p, ios[READ]),
136 part_stat_read(p, merges[READ]),
137 (unsigned long long)part_stat_read(p, sectors[READ]),
138 jiffies_to_msecs(part_stat_read(p, ticks[READ])),
139 part_stat_read(p, ios[WRITE]),
140 part_stat_read(p, merges[WRITE]),
141 (unsigned long long)part_stat_read(p, sectors[WRITE]),
142 jiffies_to_msecs(part_stat_read(p, ticks[WRITE])),
144 jiffies_to_msecs(part_stat_read(p, io_ticks)),
145 jiffies_to_msecs(part_stat_read(p, time_in_queue)));
148 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
151 struct hd_struct *p = dev_to_part(dev);
152 struct request_queue *q = part_to_disk(p)->queue;
153 unsigned int inflight[2];
155 part_in_flight_rw(q, p, inflight);
156 return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
159 #ifdef CONFIG_FAIL_MAKE_REQUEST
160 ssize_t part_fail_show(struct device *dev,
161 struct device_attribute *attr, char *buf)
163 struct hd_struct *p = dev_to_part(dev);
165 return sprintf(buf, "%d\n", p->make_it_fail);
168 ssize_t part_fail_store(struct device *dev,
169 struct device_attribute *attr,
170 const char *buf, size_t count)
172 struct hd_struct *p = dev_to_part(dev);
175 if (count > 0 && sscanf(buf, "%d", &i) > 0)
176 p->make_it_fail = (i == 0) ? 0 : 1;
182 static DEVICE_ATTR(partition, S_IRUGO, part_partition_show, NULL);
183 static DEVICE_ATTR(start, S_IRUGO, part_start_show, NULL);
184 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
185 static DEVICE_ATTR(ro, S_IRUGO, part_ro_show, NULL);
186 static DEVICE_ATTR(alignment_offset, S_IRUGO, part_alignment_offset_show, NULL);
187 static DEVICE_ATTR(discard_alignment, S_IRUGO, part_discard_alignment_show,
189 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
190 static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
191 #ifdef CONFIG_FAIL_MAKE_REQUEST
192 static struct device_attribute dev_attr_fail =
193 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
196 static struct attribute *part_attrs[] = {
197 &dev_attr_partition.attr,
198 &dev_attr_start.attr,
201 &dev_attr_alignment_offset.attr,
202 &dev_attr_discard_alignment.attr,
204 &dev_attr_inflight.attr,
205 #ifdef CONFIG_FAIL_MAKE_REQUEST
211 static struct attribute_group part_attr_group = {
215 static const struct attribute_group *part_attr_groups[] = {
217 #ifdef CONFIG_BLK_DEV_IO_TRACE
218 &blk_trace_attr_group,
223 static void part_release(struct device *dev)
225 struct hd_struct *p = dev_to_part(dev);
226 blk_free_devt(dev->devt);
231 static int part_uevent(struct device *dev, struct kobj_uevent_env *env)
233 struct hd_struct *part = dev_to_part(dev);
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);
241 struct device_type part_type = {
243 .groups = part_attr_groups,
244 .release = part_release,
245 .uevent = part_uevent,
248 static void delete_partition_rcu_cb(struct rcu_head *head)
250 struct hd_struct *part = container_of(head, struct hd_struct, rcu_head);
252 part->start_sect = 0;
254 part_stat_set_all(part, 0);
255 put_device(part_to_dev(part));
258 void __delete_partition(struct percpu_ref *ref)
260 struct hd_struct *part = container_of(ref, struct hd_struct, ref);
261 call_rcu(&part->rcu_head, delete_partition_rcu_cb);
265 * Must be called either with bd_mutex held, before a disk can be opened or
266 * after all disk users are gone.
268 void delete_partition(struct gendisk *disk, int partno)
270 struct disk_part_tbl *ptbl =
271 rcu_dereference_protected(disk->part_tbl, 1);
272 struct hd_struct *part;
273 struct block_device *bdev;
275 if (partno >= ptbl->len)
278 part = rcu_dereference_protected(ptbl->part[partno], 1);
282 rcu_assign_pointer(ptbl->part[partno], NULL);
283 rcu_assign_pointer(ptbl->last_lookup, NULL);
284 kobject_put(part->holder_dir);
285 device_del(part_to_dev(part));
287 bdev = bdget(part_devt(part));
289 remove_inode_hash(bdev->bd_inode);
292 hd_struct_kill(part);
295 static ssize_t whole_disk_show(struct device *dev,
296 struct device_attribute *attr, char *buf)
300 static DEVICE_ATTR(whole_disk, S_IRUSR | S_IRGRP | S_IROTH,
301 whole_disk_show, NULL);
304 * Must be called either with bd_mutex held, before a disk can be opened or
305 * after all disk users are gone.
307 struct hd_struct *add_partition(struct gendisk *disk, int partno,
308 sector_t start, sector_t len, int flags,
309 struct partition_meta_info *info)
312 dev_t devt = MKDEV(0, 0);
313 struct device *ddev = disk_to_dev(disk);
315 struct disk_part_tbl *ptbl;
319 err = disk_expand_part_tbl(disk, partno);
322 ptbl = rcu_dereference_protected(disk->part_tbl, 1);
324 if (ptbl->part[partno])
325 return ERR_PTR(-EBUSY);
327 p = kzalloc(sizeof(*p), GFP_KERNEL);
329 return ERR_PTR(-EBUSY);
331 if (!init_part_stats(p)) {
336 seqcount_init(&p->nr_sects_seq);
337 pdev = part_to_dev(p);
339 p->start_sect = start;
340 p->alignment_offset =
341 queue_limit_alignment_offset(&disk->queue->limits, start);
342 p->discard_alignment =
343 queue_limit_discard_alignment(&disk->queue->limits, start);
346 p->policy = get_disk_ro(disk);
349 struct partition_meta_info *pinfo = alloc_part_info(disk);
354 memcpy(pinfo, info, sizeof(*info));
358 dname = dev_name(ddev);
359 if (isdigit(dname[strlen(dname) - 1]))
360 dev_set_name(pdev, "%sp%d", dname, partno);
362 dev_set_name(pdev, "%s%d", dname, partno);
364 device_initialize(pdev);
365 pdev->class = &block_class;
366 pdev->type = &part_type;
369 err = blk_alloc_devt(p, &devt);
374 /* delay uevent until 'holders' subdir is created */
375 dev_set_uevent_suppress(pdev, 1);
376 err = device_add(pdev);
381 p->holder_dir = kobject_create_and_add("holders", &pdev->kobj);
385 dev_set_uevent_suppress(pdev, 0);
386 if (flags & ADDPART_FLAG_WHOLEDISK) {
387 err = device_create_file(pdev, &dev_attr_whole_disk);
392 err = hd_ref_init(p);
394 if (flags & ADDPART_FLAG_WHOLEDISK)
395 goto out_remove_file;
399 /* everything is up and running, commence */
400 rcu_assign_pointer(ptbl->part[partno], p);
402 /* suppress uevent if the disk suppresses it */
403 if (!dev_get_uevent_suppress(ddev))
404 kobject_uevent(&pdev->kobj, KOBJ_ADD);
415 device_remove_file(pdev, &dev_attr_whole_disk);
417 kobject_put(p->holder_dir);
424 static bool disk_unlock_native_capacity(struct gendisk *disk)
426 const struct block_device_operations *bdops = disk->fops;
428 if (bdops->unlock_native_capacity &&
429 !(disk->flags & GENHD_FL_NATIVE_CAPACITY)) {
430 printk(KERN_CONT "enabling native capacity\n");
431 bdops->unlock_native_capacity(disk);
432 disk->flags |= GENHD_FL_NATIVE_CAPACITY;
435 printk(KERN_CONT "truncated\n");
440 static int drop_partitions(struct gendisk *disk, struct block_device *bdev)
442 struct disk_part_iter piter;
443 struct hd_struct *part;
446 if (bdev->bd_part_count || bdev->bd_super)
448 res = invalidate_partition(disk, 0);
452 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
453 while ((part = disk_part_iter_next(&piter)))
454 delete_partition(disk, part->partno);
455 disk_part_iter_exit(&piter);
460 static bool part_zone_aligned(struct gendisk *disk,
461 struct block_device *bdev,
462 sector_t from, sector_t size)
464 unsigned int zone_sectors = bdev_zone_sectors(bdev);
467 * If this function is called, then the disk is a zoned block device
468 * (host-aware or host-managed). This can be detected even if the
469 * zoned block device support is disabled (CONFIG_BLK_DEV_ZONED not
470 * set). In this case, however, only host-aware devices will be seen
471 * as a block device is not created for host-managed devices. Without
472 * zoned block device support, host-aware drives can still be used as
473 * regular block devices (no zone operation) and their zone size will
474 * be reported as 0. Allow this case.
480 * Check partition start and size alignement. If the drive has a
481 * smaller last runt zone, ignore it and allow the partition to
482 * use it. Check the zone size too: it should be a power of 2 number
485 if (WARN_ON_ONCE(!is_power_of_2(zone_sectors))) {
488 div_u64_rem(from, zone_sectors, &rem);
491 if ((from + size) < get_capacity(disk)) {
492 div_u64_rem(size, zone_sectors, &rem);
499 if (from & (zone_sectors - 1))
501 if ((from + size) < get_capacity(disk) &&
502 (size & (zone_sectors - 1)))
510 int rescan_partitions(struct gendisk *disk, struct block_device *bdev)
512 struct parsed_partitions *state = NULL;
513 struct hd_struct *part;
516 if (state && !IS_ERR(state)) {
517 free_partitions(state);
521 res = drop_partitions(disk, bdev);
525 if (disk->fops->revalidate_disk)
526 disk->fops->revalidate_disk(disk);
527 check_disk_size_change(disk, bdev);
528 bdev->bd_invalidated = 0;
529 if (!get_capacity(disk) || !(state = check_partition(disk, bdev)))
533 * I/O error reading the partition table. If any
534 * partition code tried to read beyond EOD, retry
535 * after unlocking native capacity.
537 if (PTR_ERR(state) == -ENOSPC) {
538 printk(KERN_WARNING "%s: partition table beyond EOD, ",
540 if (disk_unlock_native_capacity(disk))
546 * If any partition code tried to read beyond EOD, try
547 * unlocking native capacity even if partition table is
548 * successfully read as we could be missing some partitions.
550 if (state->access_beyond_eod) {
552 "%s: partition table partially beyond EOD, ",
554 if (disk_unlock_native_capacity(disk))
558 /* tell userspace that the media / partition table may have changed */
559 kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
561 /* Detect the highest partition number and preallocate
562 * disk->part_tbl. This is an optimization and not strictly
565 for (p = 1, highest = 0; p < state->limit; p++)
566 if (state->parts[p].size)
569 disk_expand_part_tbl(disk, highest);
572 for (p = 1; p < state->limit; p++) {
575 size = state->parts[p].size;
579 from = state->parts[p].from;
580 if (from >= get_capacity(disk)) {
582 "%s: p%d start %llu is beyond EOD, ",
583 disk->disk_name, p, (unsigned long long) from);
584 if (disk_unlock_native_capacity(disk))
589 if (from + size > get_capacity(disk)) {
591 "%s: p%d size %llu extends beyond EOD, ",
592 disk->disk_name, p, (unsigned long long) size);
594 if (disk_unlock_native_capacity(disk)) {
595 /* free state and restart */
599 * we can not ignore partitions of broken tables
600 * created by for example camera firmware, but
601 * we limit them to the end of the disk to avoid
602 * creating invalid block devices
604 size = get_capacity(disk) - from;
609 * On a zoned block device, partitions should be aligned on the
610 * device zone size (i.e. zone boundary crossing not allowed).
611 * Otherwise, resetting the write pointer of the last zone of
612 * one partition may impact the following partition.
614 if (bdev_is_zoned(bdev) &&
615 !part_zone_aligned(disk, bdev, from, size)) {
617 "%s: p%d start %llu+%llu is not zone aligned\n",
618 disk->disk_name, p, (unsigned long long) from,
619 (unsigned long long) size);
623 part = add_partition(disk, p, from, size,
624 state->parts[p].flags,
625 &state->parts[p].info);
627 printk(KERN_ERR " %s: p%d could not be added: %ld\n",
628 disk->disk_name, p, -PTR_ERR(part));
631 #ifdef CONFIG_BLK_DEV_MD
632 if (state->parts[p].flags & ADDPART_FLAG_RAID)
633 md_autodetect_dev(part_to_dev(part)->devt);
636 free_partitions(state);
640 int invalidate_partitions(struct gendisk *disk, struct block_device *bdev)
644 if (!bdev->bd_invalidated)
647 res = drop_partitions(disk, bdev);
651 set_capacity(disk, 0);
652 check_disk_size_change(disk, bdev);
653 bdev->bd_invalidated = 0;
654 /* tell userspace that the media / partition table may have changed */
655 kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
660 unsigned char *read_dev_sector(struct block_device *bdev, sector_t n, Sector *p)
662 struct address_space *mapping = bdev->bd_inode->i_mapping;
665 page = read_mapping_page(mapping, (pgoff_t)(n >> (PAGE_SHIFT-9)), NULL);
670 return (unsigned char *)page_address(page) + ((n & ((1 << (PAGE_SHIFT - 9)) - 1)) << 9);
678 EXPORT_SYMBOL(read_dev_sector);