1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 1991-1998 Linus Torvalds
4 * Re-organised Feb 1998 Russell King
7 #include <linux/slab.h>
8 #include <linux/ctype.h>
9 #include <linux/genhd.h>
10 #include <linux/vmalloc.h>
11 #include <linux/blktrace_api.h>
12 #include <linux/raid/detect.h>
15 static int (*check_part[])(struct parsed_partitions *) = {
17 * Probe partition formats with tables at disk address 0
18 * that also have an ADFS boot block at 0xdc0.
20 #ifdef CONFIG_ACORN_PARTITION_ICS
23 #ifdef CONFIG_ACORN_PARTITION_POWERTEC
24 adfspart_check_POWERTEC,
26 #ifdef CONFIG_ACORN_PARTITION_EESOX
31 * Now move on to formats that only have partition info at
32 * disk address 0xdc0. Since these may also have stale
33 * PC/BIOS partition tables, they need to come before
36 #ifdef CONFIG_ACORN_PARTITION_CUMANA
37 adfspart_check_CUMANA,
39 #ifdef CONFIG_ACORN_PARTITION_ADFS
43 #ifdef CONFIG_CMDLINE_PARTITION
46 #ifdef CONFIG_EFI_PARTITION
47 efi_partition, /* this must come before msdos */
49 #ifdef CONFIG_SGI_PARTITION
52 #ifdef CONFIG_LDM_PARTITION
53 ldm_partition, /* this must come before msdos */
55 #ifdef CONFIG_MSDOS_PARTITION
58 #ifdef CONFIG_OSF_PARTITION
61 #ifdef CONFIG_SUN_PARTITION
64 #ifdef CONFIG_AMIGA_PARTITION
67 #ifdef CONFIG_ATARI_PARTITION
70 #ifdef CONFIG_MAC_PARTITION
73 #ifdef CONFIG_ULTRIX_PARTITION
76 #ifdef CONFIG_IBM_PARTITION
79 #ifdef CONFIG_KARMA_PARTITION
82 #ifdef CONFIG_SYSV68_PARTITION
88 static void bdev_set_nr_sectors(struct block_device *bdev, sector_t sectors)
90 spin_lock(&bdev->bd_size_lock);
91 i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
92 spin_unlock(&bdev->bd_size_lock);
95 static struct parsed_partitions *allocate_partitions(struct gendisk *hd)
97 struct parsed_partitions *state;
100 state = kzalloc(sizeof(*state), GFP_KERNEL);
104 nr = disk_max_parts(hd);
105 state->parts = vzalloc(array_size(nr, sizeof(state->parts[0])));
116 static void free_partitions(struct parsed_partitions *state)
122 static struct parsed_partitions *check_partition(struct gendisk *hd,
123 struct block_device *bdev)
125 struct parsed_partitions *state;
128 state = allocate_partitions(hd);
131 state->pp_buf = (char *)__get_free_page(GFP_KERNEL);
132 if (!state->pp_buf) {
133 free_partitions(state);
136 state->pp_buf[0] = '\0';
139 disk_name(hd, 0, state->name);
140 snprintf(state->pp_buf, PAGE_SIZE, " %s:", state->name);
141 if (isdigit(state->name[strlen(state->name)-1]))
142 sprintf(state->name, "p");
145 while (!res && check_part[i]) {
146 memset(state->parts, 0, state->limit * sizeof(state->parts[0]));
147 res = check_part[i++](state);
150 * We have hit an I/O error which we don't report now.
151 * But record it, and let the others do their job.
159 printk(KERN_INFO "%s", state->pp_buf);
161 free_page((unsigned long)state->pp_buf);
164 if (state->access_beyond_eod)
167 * The partition is unrecognized. So report I/O errors if there were any
172 strlcat(state->pp_buf,
173 " unable to read partition table\n", PAGE_SIZE);
174 printk(KERN_INFO "%s", state->pp_buf);
177 free_page((unsigned long)state->pp_buf);
178 free_partitions(state);
182 static ssize_t part_partition_show(struct device *dev,
183 struct device_attribute *attr, char *buf)
185 struct hd_struct *p = dev_to_part(dev);
187 return sprintf(buf, "%d\n", p->bdev->bd_partno);
190 static ssize_t part_start_show(struct device *dev,
191 struct device_attribute *attr, char *buf)
193 struct hd_struct *p = dev_to_part(dev);
195 return sprintf(buf, "%llu\n", p->bdev->bd_start_sect);
198 static ssize_t part_ro_show(struct device *dev,
199 struct device_attribute *attr, char *buf)
201 struct hd_struct *p = dev_to_part(dev);
202 return sprintf(buf, "%d\n", p->bdev->bd_read_only);
205 static ssize_t part_alignment_offset_show(struct device *dev,
206 struct device_attribute *attr, char *buf)
208 struct hd_struct *p = dev_to_part(dev);
210 return sprintf(buf, "%u\n",
211 queue_limit_alignment_offset(&part_to_disk(p)->queue->limits,
212 p->bdev->bd_start_sect));
215 static ssize_t part_discard_alignment_show(struct device *dev,
216 struct device_attribute *attr, char *buf)
218 struct hd_struct *p = dev_to_part(dev);
220 return sprintf(buf, "%u\n",
221 queue_limit_discard_alignment(&part_to_disk(p)->queue->limits,
222 p->bdev->bd_start_sect));
225 static DEVICE_ATTR(partition, 0444, part_partition_show, NULL);
226 static DEVICE_ATTR(start, 0444, part_start_show, NULL);
227 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
228 static DEVICE_ATTR(ro, 0444, part_ro_show, NULL);
229 static DEVICE_ATTR(alignment_offset, 0444, part_alignment_offset_show, NULL);
230 static DEVICE_ATTR(discard_alignment, 0444, part_discard_alignment_show, NULL);
231 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
232 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
233 #ifdef CONFIG_FAIL_MAKE_REQUEST
234 static struct device_attribute dev_attr_fail =
235 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
238 static struct attribute *part_attrs[] = {
239 &dev_attr_partition.attr,
240 &dev_attr_start.attr,
243 &dev_attr_alignment_offset.attr,
244 &dev_attr_discard_alignment.attr,
246 &dev_attr_inflight.attr,
247 #ifdef CONFIG_FAIL_MAKE_REQUEST
253 static struct attribute_group part_attr_group = {
257 static const struct attribute_group *part_attr_groups[] = {
259 #ifdef CONFIG_BLK_DEV_IO_TRACE
260 &blk_trace_attr_group,
265 static void part_release(struct device *dev)
267 struct hd_struct *p = dev_to_part(dev);
269 blk_free_devt(dev->devt);
273 static int part_uevent(struct device *dev, struct kobj_uevent_env *env)
275 struct hd_struct *part = dev_to_part(dev);
277 add_uevent_var(env, "PARTN=%u", part->bdev->bd_partno);
278 if (part->bdev->bd_meta_info && part->bdev->bd_meta_info->volname[0])
279 add_uevent_var(env, "PARTNAME=%s",
280 part->bdev->bd_meta_info->volname);
284 struct device_type part_type = {
286 .groups = part_attr_groups,
287 .release = part_release,
288 .uevent = part_uevent,
292 * Must be called either with bd_mutex held, before a disk can be opened or
293 * after all disk users are gone.
295 void delete_partition(struct hd_struct *part)
297 struct gendisk *disk = part_to_disk(part);
298 struct disk_part_tbl *ptbl =
299 rcu_dereference_protected(disk->part_tbl, 1);
301 rcu_assign_pointer(ptbl->part[part->bdev->bd_partno], NULL);
302 rcu_assign_pointer(ptbl->last_lookup, NULL);
304 kobject_put(part->bdev->bd_holder_dir);
305 device_del(part_to_dev(part));
308 * Remove the block device from the inode hash, so that it cannot be
309 * looked up any more even when openers still hold references.
311 remove_inode_hash(part->bdev->bd_inode);
313 put_device(part_to_dev(part));
316 static ssize_t whole_disk_show(struct device *dev,
317 struct device_attribute *attr, char *buf)
321 static DEVICE_ATTR(whole_disk, 0444, whole_disk_show, NULL);
324 * Must be called either with bd_mutex held, before a disk can be opened or
325 * after all disk users are gone.
327 static struct hd_struct *add_partition(struct gendisk *disk, int partno,
328 sector_t start, sector_t len, int flags,
329 struct partition_meta_info *info)
332 dev_t devt = MKDEV(0, 0);
333 struct device *ddev = disk_to_dev(disk);
335 struct block_device *bdev;
336 struct disk_part_tbl *ptbl;
341 * Partitions are not supported on zoned block devices that are used as
344 switch (disk->queue->limits.zoned) {
346 pr_warn("%s: partitions not supported on host managed zoned block device\n",
348 return ERR_PTR(-ENXIO);
350 pr_info("%s: disabling host aware zoned block device support due to partitions\n",
352 disk->queue->limits.zoned = BLK_ZONED_NONE;
358 err = disk_expand_part_tbl(disk, partno);
361 ptbl = rcu_dereference_protected(disk->part_tbl, 1);
363 if (ptbl->part[partno])
364 return ERR_PTR(-EBUSY);
366 bdev = bdev_alloc(disk, partno);
368 return ERR_PTR(-ENOMEM);
371 pdev = part_to_dev(p);
373 bdev->bd_start_sect = start;
374 bdev_set_nr_sectors(bdev, len);
375 bdev->bd_read_only = get_disk_ro(disk);
379 bdev->bd_meta_info = kmemdup(info, sizeof(*info), GFP_KERNEL);
380 if (!bdev->bd_meta_info)
384 dname = dev_name(ddev);
385 if (isdigit(dname[strlen(dname) - 1]))
386 dev_set_name(pdev, "%sp%d", dname, partno);
388 dev_set_name(pdev, "%s%d", dname, partno);
390 device_initialize(pdev);
391 pdev->class = &block_class;
392 pdev->type = &part_type;
395 err = blk_alloc_devt(bdev, &devt);
400 /* delay uevent until 'holders' subdir is created */
401 dev_set_uevent_suppress(pdev, 1);
402 err = device_add(pdev);
407 bdev->bd_holder_dir = kobject_create_and_add("holders", &pdev->kobj);
408 if (!bdev->bd_holder_dir)
411 dev_set_uevent_suppress(pdev, 0);
412 if (flags & ADDPART_FLAG_WHOLEDISK) {
413 err = device_create_file(pdev, &dev_attr_whole_disk);
418 /* everything is up and running, commence */
419 bdev_add(bdev, devt);
420 rcu_assign_pointer(ptbl->part[partno], bdev);
422 /* suppress uevent if the disk suppresses it */
423 if (!dev_get_uevent_suppress(ddev))
424 kobject_uevent(&pdev->kobj, KOBJ_ADD);
431 kobject_put(bdev->bd_holder_dir);
438 static bool partition_overlaps(struct gendisk *disk, sector_t start,
439 sector_t length, int skip_partno)
441 struct disk_part_iter piter;
442 struct block_device *part;
443 bool overlap = false;
445 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
446 while ((part = disk_part_iter_next(&piter))) {
447 if (part->bd_partno == skip_partno ||
448 start >= part->bd_start_sect + bdev_nr_sectors(part) ||
449 start + length <= part->bd_start_sect)
455 disk_part_iter_exit(&piter);
459 int bdev_add_partition(struct block_device *bdev, int partno,
460 sector_t start, sector_t length)
462 struct hd_struct *part;
464 mutex_lock(&bdev->bd_mutex);
465 if (partition_overlaps(bdev->bd_disk, start, length, -1)) {
466 mutex_unlock(&bdev->bd_mutex);
470 part = add_partition(bdev->bd_disk, partno, start, length,
471 ADDPART_FLAG_NONE, NULL);
472 mutex_unlock(&bdev->bd_mutex);
473 return PTR_ERR_OR_ZERO(part);
476 int bdev_del_partition(struct block_device *bdev, int partno)
478 struct block_device *bdevp;
479 struct hd_struct *part = NULL;
482 bdevp = bdget_disk(bdev->bd_disk, partno);
486 mutex_lock(&bdevp->bd_mutex);
487 mutex_lock_nested(&bdev->bd_mutex, 1);
490 part = disk_get_part(bdev->bd_disk, partno);
495 if (bdevp->bd_openers)
498 sync_blockdev(bdevp);
499 invalidate_bdev(bdevp);
501 delete_partition(part);
504 mutex_unlock(&bdev->bd_mutex);
505 mutex_unlock(&bdevp->bd_mutex);
512 int bdev_resize_partition(struct block_device *bdev, int partno,
513 sector_t start, sector_t length)
515 struct block_device *bdevp;
516 struct hd_struct *part;
519 part = disk_get_part(bdev->bd_disk, partno);
524 bdevp = bdget_part(part);
528 mutex_lock(&bdevp->bd_mutex);
529 mutex_lock_nested(&bdev->bd_mutex, 1);
532 if (start != part->bdev->bd_start_sect)
536 if (partition_overlaps(bdev->bd_disk, start, length, partno))
539 bdev_set_nr_sectors(bdevp, length);
543 mutex_unlock(&bdevp->bd_mutex);
544 mutex_unlock(&bdev->bd_mutex);
551 static bool disk_unlock_native_capacity(struct gendisk *disk)
553 const struct block_device_operations *bdops = disk->fops;
555 if (bdops->unlock_native_capacity &&
556 !(disk->flags & GENHD_FL_NATIVE_CAPACITY)) {
557 printk(KERN_CONT "enabling native capacity\n");
558 bdops->unlock_native_capacity(disk);
559 disk->flags |= GENHD_FL_NATIVE_CAPACITY;
562 printk(KERN_CONT "truncated\n");
567 int blk_drop_partitions(struct block_device *bdev)
569 struct disk_part_iter piter;
570 struct block_device *part;
572 if (bdev->bd_part_count)
576 invalidate_bdev(bdev);
578 disk_part_iter_init(&piter, bdev->bd_disk, DISK_PITER_INCL_EMPTY);
579 while ((part = disk_part_iter_next(&piter)))
580 delete_partition(part->bd_part);
581 disk_part_iter_exit(&piter);
586 /* for historic reasons in the DASD driver */
587 EXPORT_SYMBOL_GPL(blk_drop_partitions);
590 static bool blk_add_partition(struct gendisk *disk, struct block_device *bdev,
591 struct parsed_partitions *state, int p)
593 sector_t size = state->parts[p].size;
594 sector_t from = state->parts[p].from;
595 struct hd_struct *part;
600 if (from >= get_capacity(disk)) {
602 "%s: p%d start %llu is beyond EOD, ",
603 disk->disk_name, p, (unsigned long long) from);
604 if (disk_unlock_native_capacity(disk))
609 if (from + size > get_capacity(disk)) {
611 "%s: p%d size %llu extends beyond EOD, ",
612 disk->disk_name, p, (unsigned long long) size);
614 if (disk_unlock_native_capacity(disk))
618 * We can not ignore partitions of broken tables created by for
619 * example camera firmware, but we limit them to the end of the
620 * disk to avoid creating invalid block devices.
622 size = get_capacity(disk) - from;
625 part = add_partition(disk, p, from, size, state->parts[p].flags,
626 &state->parts[p].info);
627 if (IS_ERR(part) && PTR_ERR(part) != -ENXIO) {
628 printk(KERN_ERR " %s: p%d could not be added: %ld\n",
629 disk->disk_name, p, -PTR_ERR(part));
633 if (IS_BUILTIN(CONFIG_BLK_DEV_MD) &&
634 (state->parts[p].flags & ADDPART_FLAG_RAID))
635 md_autodetect_dev(part_to_dev(part)->devt);
640 int blk_add_partitions(struct gendisk *disk, struct block_device *bdev)
642 struct parsed_partitions *state;
643 int ret = -EAGAIN, p, highest;
645 if (!disk_part_scan_enabled(disk))
648 state = check_partition(disk, bdev);
653 * I/O error reading the partition table. If we tried to read
654 * beyond EOD, retry after unlocking the native capacity.
656 if (PTR_ERR(state) == -ENOSPC) {
657 printk(KERN_WARNING "%s: partition table beyond EOD, ",
659 if (disk_unlock_native_capacity(disk))
666 * Partitions are not supported on host managed zoned block devices.
668 if (disk->queue->limits.zoned == BLK_ZONED_HM) {
669 pr_warn("%s: ignoring partition table on host managed zoned block device\n",
676 * If we read beyond EOD, try unlocking native capacity even if the
677 * partition table was successfully read as we could be missing some
680 if (state->access_beyond_eod) {
682 "%s: partition table partially beyond EOD, ",
684 if (disk_unlock_native_capacity(disk))
688 /* tell userspace that the media / partition table may have changed */
689 kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
692 * Detect the highest partition number and preallocate disk->part_tbl.
693 * This is an optimization and not strictly necessary.
695 for (p = 1, highest = 0; p < state->limit; p++)
696 if (state->parts[p].size)
698 disk_expand_part_tbl(disk, highest);
700 for (p = 1; p < state->limit; p++)
701 if (!blk_add_partition(disk, bdev, state, p))
706 free_partitions(state);
710 void *read_part_sector(struct parsed_partitions *state, sector_t n, Sector *p)
712 struct address_space *mapping = state->bdev->bd_inode->i_mapping;
715 if (n >= get_capacity(state->bdev->bd_disk)) {
716 state->access_beyond_eod = true;
720 page = read_mapping_page(mapping,
721 (pgoff_t)(n >> (PAGE_SHIFT - 9)), NULL);
728 return (unsigned char *)page_address(page) +
729 ((n & ((1 << (PAGE_SHIFT - 9)) - 1)) << SECTOR_SHIFT);