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);
55 * There's very little reason to use this, you should really
56 * have a struct block_device just about everywhere and use
59 const char *__bdevname(dev_t dev, char *buffer)
61 scnprintf(buffer, BDEVNAME_SIZE, "unknown-block(%u,%u)",
62 MAJOR(dev), MINOR(dev));
66 EXPORT_SYMBOL(__bdevname);
68 static ssize_t part_partition_show(struct device *dev,
69 struct device_attribute *attr, char *buf)
71 struct hd_struct *p = dev_to_part(dev);
73 return sprintf(buf, "%d\n", p->partno);
76 static ssize_t part_start_show(struct device *dev,
77 struct device_attribute *attr, char *buf)
79 struct hd_struct *p = dev_to_part(dev);
81 return sprintf(buf, "%llu\n",(unsigned long long)p->start_sect);
84 ssize_t part_size_show(struct device *dev,
85 struct device_attribute *attr, char *buf)
87 struct hd_struct *p = dev_to_part(dev);
88 return sprintf(buf, "%llu\n",(unsigned long long)part_nr_sects_read(p));
91 static ssize_t part_ro_show(struct device *dev,
92 struct device_attribute *attr, char *buf)
94 struct hd_struct *p = dev_to_part(dev);
95 return sprintf(buf, "%d\n", p->policy ? 1 : 0);
98 static ssize_t part_alignment_offset_show(struct device *dev,
99 struct device_attribute *attr, char *buf)
101 struct hd_struct *p = dev_to_part(dev);
102 return sprintf(buf, "%llu\n", (unsigned long long)p->alignment_offset);
105 static ssize_t part_discard_alignment_show(struct device *dev,
106 struct device_attribute *attr, char *buf)
108 struct hd_struct *p = dev_to_part(dev);
109 return sprintf(buf, "%u\n", p->discard_alignment);
112 ssize_t part_stat_show(struct device *dev,
113 struct device_attribute *attr, char *buf)
115 struct hd_struct *p = dev_to_part(dev);
116 struct request_queue *q = part_to_disk(p)->queue;
117 unsigned int inflight[2];
120 cpu = part_stat_lock();
121 part_round_stats(q, cpu, p);
123 part_in_flight(q, p, inflight);
125 "%8lu %8lu %8llu %8u "
126 "%8lu %8lu %8llu %8u "
129 part_stat_read(p, ios[READ]),
130 part_stat_read(p, merges[READ]),
131 (unsigned long long)part_stat_read(p, sectors[READ]),
132 jiffies_to_msecs(part_stat_read(p, ticks[READ])),
133 part_stat_read(p, ios[WRITE]),
134 part_stat_read(p, merges[WRITE]),
135 (unsigned long long)part_stat_read(p, sectors[WRITE]),
136 jiffies_to_msecs(part_stat_read(p, ticks[WRITE])),
138 jiffies_to_msecs(part_stat_read(p, io_ticks)),
139 jiffies_to_msecs(part_stat_read(p, time_in_queue)));
142 ssize_t part_inflight_show(struct device *dev,
143 struct device_attribute *attr, char *buf)
145 struct hd_struct *p = dev_to_part(dev);
147 return sprintf(buf, "%8u %8u\n", atomic_read(&p->in_flight[0]),
148 atomic_read(&p->in_flight[1]));
151 #ifdef CONFIG_FAIL_MAKE_REQUEST
152 ssize_t part_fail_show(struct device *dev,
153 struct device_attribute *attr, char *buf)
155 struct hd_struct *p = dev_to_part(dev);
157 return sprintf(buf, "%d\n", p->make_it_fail);
160 ssize_t part_fail_store(struct device *dev,
161 struct device_attribute *attr,
162 const char *buf, size_t count)
164 struct hd_struct *p = dev_to_part(dev);
167 if (count > 0 && sscanf(buf, "%d", &i) > 0)
168 p->make_it_fail = (i == 0) ? 0 : 1;
174 static DEVICE_ATTR(partition, S_IRUGO, part_partition_show, NULL);
175 static DEVICE_ATTR(start, S_IRUGO, part_start_show, NULL);
176 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
177 static DEVICE_ATTR(ro, S_IRUGO, part_ro_show, NULL);
178 static DEVICE_ATTR(alignment_offset, S_IRUGO, part_alignment_offset_show, NULL);
179 static DEVICE_ATTR(discard_alignment, S_IRUGO, part_discard_alignment_show,
181 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
182 static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
183 #ifdef CONFIG_FAIL_MAKE_REQUEST
184 static struct device_attribute dev_attr_fail =
185 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
188 static struct attribute *part_attrs[] = {
189 &dev_attr_partition.attr,
190 &dev_attr_start.attr,
193 &dev_attr_alignment_offset.attr,
194 &dev_attr_discard_alignment.attr,
196 &dev_attr_inflight.attr,
197 #ifdef CONFIG_FAIL_MAKE_REQUEST
203 static struct attribute_group part_attr_group = {
207 static const struct attribute_group *part_attr_groups[] = {
209 #ifdef CONFIG_BLK_DEV_IO_TRACE
210 &blk_trace_attr_group,
215 static void part_release(struct device *dev)
217 struct hd_struct *p = dev_to_part(dev);
218 blk_free_devt(dev->devt);
223 static int part_uevent(struct device *dev, struct kobj_uevent_env *env)
225 struct hd_struct *part = dev_to_part(dev);
227 add_uevent_var(env, "PARTN=%u", part->partno);
228 if (part->info && part->info->volname[0])
229 add_uevent_var(env, "PARTNAME=%s", part->info->volname);
233 struct device_type part_type = {
235 .groups = part_attr_groups,
236 .release = part_release,
237 .uevent = part_uevent,
240 static void delete_partition_rcu_cb(struct rcu_head *head)
242 struct hd_struct *part = container_of(head, struct hd_struct, rcu_head);
244 part->start_sect = 0;
246 part_stat_set_all(part, 0);
247 put_device(part_to_dev(part));
250 void __delete_partition(struct percpu_ref *ref)
252 struct hd_struct *part = container_of(ref, struct hd_struct, ref);
253 call_rcu(&part->rcu_head, delete_partition_rcu_cb);
257 * Must be called either with bd_mutex held, before a disk can be opened or
258 * after all disk users are gone.
260 void delete_partition(struct gendisk *disk, int partno)
262 struct disk_part_tbl *ptbl =
263 rcu_dereference_protected(disk->part_tbl, 1);
264 struct hd_struct *part;
266 if (partno >= ptbl->len)
269 part = rcu_dereference_protected(ptbl->part[partno], 1);
273 rcu_assign_pointer(ptbl->part[partno], NULL);
274 rcu_assign_pointer(ptbl->last_lookup, NULL);
275 kobject_put(part->holder_dir);
276 device_del(part_to_dev(part));
278 hd_struct_kill(part);
281 static ssize_t whole_disk_show(struct device *dev,
282 struct device_attribute *attr, char *buf)
286 static DEVICE_ATTR(whole_disk, S_IRUSR | S_IRGRP | S_IROTH,
287 whole_disk_show, NULL);
290 * Must be called either with bd_mutex held, before a disk can be opened or
291 * after all disk users are gone.
293 struct hd_struct *add_partition(struct gendisk *disk, int partno,
294 sector_t start, sector_t len, int flags,
295 struct partition_meta_info *info)
298 dev_t devt = MKDEV(0, 0);
299 struct device *ddev = disk_to_dev(disk);
301 struct disk_part_tbl *ptbl;
305 err = disk_expand_part_tbl(disk, partno);
308 ptbl = rcu_dereference_protected(disk->part_tbl, 1);
310 if (ptbl->part[partno])
311 return ERR_PTR(-EBUSY);
313 p = kzalloc(sizeof(*p), GFP_KERNEL);
315 return ERR_PTR(-EBUSY);
317 if (!init_part_stats(p)) {
322 seqcount_init(&p->nr_sects_seq);
323 pdev = part_to_dev(p);
325 p->start_sect = start;
326 p->alignment_offset =
327 queue_limit_alignment_offset(&disk->queue->limits, start);
328 p->discard_alignment =
329 queue_limit_discard_alignment(&disk->queue->limits, start);
332 p->policy = get_disk_ro(disk);
335 struct partition_meta_info *pinfo = alloc_part_info(disk);
340 memcpy(pinfo, info, sizeof(*info));
344 dname = dev_name(ddev);
345 if (isdigit(dname[strlen(dname) - 1]))
346 dev_set_name(pdev, "%sp%d", dname, partno);
348 dev_set_name(pdev, "%s%d", dname, partno);
350 device_initialize(pdev);
351 pdev->class = &block_class;
352 pdev->type = &part_type;
355 err = blk_alloc_devt(p, &devt);
360 /* delay uevent until 'holders' subdir is created */
361 dev_set_uevent_suppress(pdev, 1);
362 err = device_add(pdev);
367 p->holder_dir = kobject_create_and_add("holders", &pdev->kobj);
371 dev_set_uevent_suppress(pdev, 0);
372 if (flags & ADDPART_FLAG_WHOLEDISK) {
373 err = device_create_file(pdev, &dev_attr_whole_disk);
378 err = hd_ref_init(p);
380 if (flags & ADDPART_FLAG_WHOLEDISK)
381 goto out_remove_file;
385 /* everything is up and running, commence */
386 rcu_assign_pointer(ptbl->part[partno], p);
388 /* suppress uevent if the disk suppresses it */
389 if (!dev_get_uevent_suppress(ddev))
390 kobject_uevent(&pdev->kobj, KOBJ_ADD);
401 device_remove_file(pdev, &dev_attr_whole_disk);
403 kobject_put(p->holder_dir);
410 static bool disk_unlock_native_capacity(struct gendisk *disk)
412 const struct block_device_operations *bdops = disk->fops;
414 if (bdops->unlock_native_capacity &&
415 !(disk->flags & GENHD_FL_NATIVE_CAPACITY)) {
416 printk(KERN_CONT "enabling native capacity\n");
417 bdops->unlock_native_capacity(disk);
418 disk->flags |= GENHD_FL_NATIVE_CAPACITY;
421 printk(KERN_CONT "truncated\n");
426 static int drop_partitions(struct gendisk *disk, struct block_device *bdev)
428 struct disk_part_iter piter;
429 struct hd_struct *part;
432 if (bdev->bd_part_count || bdev->bd_super)
434 res = invalidate_partition(disk, 0);
438 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
439 while ((part = disk_part_iter_next(&piter)))
440 delete_partition(disk, part->partno);
441 disk_part_iter_exit(&piter);
446 static bool part_zone_aligned(struct gendisk *disk,
447 struct block_device *bdev,
448 sector_t from, sector_t size)
450 unsigned int zone_sectors = bdev_zone_sectors(bdev);
453 * If this function is called, then the disk is a zoned block device
454 * (host-aware or host-managed). This can be detected even if the
455 * zoned block device support is disabled (CONFIG_BLK_DEV_ZONED not
456 * set). In this case, however, only host-aware devices will be seen
457 * as a block device is not created for host-managed devices. Without
458 * zoned block device support, host-aware drives can still be used as
459 * regular block devices (no zone operation) and their zone size will
460 * be reported as 0. Allow this case.
466 * Check partition start and size alignement. If the drive has a
467 * smaller last runt zone, ignore it and allow the partition to
468 * use it. Check the zone size too: it should be a power of 2 number
471 if (WARN_ON_ONCE(!is_power_of_2(zone_sectors))) {
474 div_u64_rem(from, zone_sectors, &rem);
477 if ((from + size) < get_capacity(disk)) {
478 div_u64_rem(size, zone_sectors, &rem);
485 if (from & (zone_sectors - 1))
487 if ((from + size) < get_capacity(disk) &&
488 (size & (zone_sectors - 1)))
496 int rescan_partitions(struct gendisk *disk, struct block_device *bdev)
498 struct parsed_partitions *state = NULL;
499 struct hd_struct *part;
502 if (state && !IS_ERR(state)) {
503 free_partitions(state);
507 res = drop_partitions(disk, bdev);
511 if (disk->fops->revalidate_disk)
512 disk->fops->revalidate_disk(disk);
513 check_disk_size_change(disk, bdev);
514 bdev->bd_invalidated = 0;
515 if (!get_capacity(disk) || !(state = check_partition(disk, bdev)))
519 * I/O error reading the partition table. If any
520 * partition code tried to read beyond EOD, retry
521 * after unlocking native capacity.
523 if (PTR_ERR(state) == -ENOSPC) {
524 printk(KERN_WARNING "%s: partition table beyond EOD, ",
526 if (disk_unlock_native_capacity(disk))
532 * If any partition code tried to read beyond EOD, try
533 * unlocking native capacity even if partition table is
534 * successfully read as we could be missing some partitions.
536 if (state->access_beyond_eod) {
538 "%s: partition table partially beyond EOD, ",
540 if (disk_unlock_native_capacity(disk))
544 /* tell userspace that the media / partition table may have changed */
545 kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
547 /* Detect the highest partition number and preallocate
548 * disk->part_tbl. This is an optimization and not strictly
551 for (p = 1, highest = 0; p < state->limit; p++)
552 if (state->parts[p].size)
555 disk_expand_part_tbl(disk, highest);
558 for (p = 1; p < state->limit; p++) {
561 size = state->parts[p].size;
565 from = state->parts[p].from;
566 if (from >= get_capacity(disk)) {
568 "%s: p%d start %llu is beyond EOD, ",
569 disk->disk_name, p, (unsigned long long) from);
570 if (disk_unlock_native_capacity(disk))
575 if (from + size > get_capacity(disk)) {
577 "%s: p%d size %llu extends beyond EOD, ",
578 disk->disk_name, p, (unsigned long long) size);
580 if (disk_unlock_native_capacity(disk)) {
581 /* free state and restart */
585 * we can not ignore partitions of broken tables
586 * created by for example camera firmware, but
587 * we limit them to the end of the disk to avoid
588 * creating invalid block devices
590 size = get_capacity(disk) - from;
595 * On a zoned block device, partitions should be aligned on the
596 * device zone size (i.e. zone boundary crossing not allowed).
597 * Otherwise, resetting the write pointer of the last zone of
598 * one partition may impact the following partition.
600 if (bdev_is_zoned(bdev) &&
601 !part_zone_aligned(disk, bdev, from, size)) {
603 "%s: p%d start %llu+%llu is not zone aligned\n",
604 disk->disk_name, p, (unsigned long long) from,
605 (unsigned long long) size);
609 part = add_partition(disk, p, from, size,
610 state->parts[p].flags,
611 &state->parts[p].info);
613 printk(KERN_ERR " %s: p%d could not be added: %ld\n",
614 disk->disk_name, p, -PTR_ERR(part));
617 #ifdef CONFIG_BLK_DEV_MD
618 if (state->parts[p].flags & ADDPART_FLAG_RAID)
619 md_autodetect_dev(part_to_dev(part)->devt);
622 free_partitions(state);
626 int invalidate_partitions(struct gendisk *disk, struct block_device *bdev)
630 if (!bdev->bd_invalidated)
633 res = drop_partitions(disk, bdev);
637 set_capacity(disk, 0);
638 check_disk_size_change(disk, bdev);
639 bdev->bd_invalidated = 0;
640 /* tell userspace that the media / partition table may have changed */
641 kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
646 unsigned char *read_dev_sector(struct block_device *bdev, sector_t n, Sector *p)
648 struct address_space *mapping = bdev->bd_inode->i_mapping;
651 page = read_mapping_page(mapping, (pgoff_t)(n >> (PAGE_SHIFT-9)), NULL);
656 return (unsigned char *)page_address(page) + ((n & ((1 << (PAGE_SHIFT - 9)) - 1)) << 9);
664 EXPORT_SYMBOL(read_dev_sector);