1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 1991-1998 Linus Torvalds
4 * Re-organised Feb 1998 Russell King
5 * Copyright (C) 2020 Christoph Hellwig
8 #include <linux/slab.h>
9 #include <linux/ctype.h>
10 #include <linux/genhd.h>
11 #include <linux/vmalloc.h>
12 #include <linux/blktrace_api.h>
13 #include <linux/raid/detect.h>
16 static int (*check_part[])(struct parsed_partitions *) = {
18 * Probe partition formats with tables at disk address 0
19 * that also have an ADFS boot block at 0xdc0.
21 #ifdef CONFIG_ACORN_PARTITION_ICS
24 #ifdef CONFIG_ACORN_PARTITION_POWERTEC
25 adfspart_check_POWERTEC,
27 #ifdef CONFIG_ACORN_PARTITION_EESOX
32 * Now move on to formats that only have partition info at
33 * disk address 0xdc0. Since these may also have stale
34 * PC/BIOS partition tables, they need to come before
37 #ifdef CONFIG_ACORN_PARTITION_CUMANA
38 adfspart_check_CUMANA,
40 #ifdef CONFIG_ACORN_PARTITION_ADFS
44 #ifdef CONFIG_CMDLINE_PARTITION
47 #ifdef CONFIG_EFI_PARTITION
48 efi_partition, /* this must come before msdos */
50 #ifdef CONFIG_SGI_PARTITION
53 #ifdef CONFIG_LDM_PARTITION
54 ldm_partition, /* this must come before msdos */
56 #ifdef CONFIG_MSDOS_PARTITION
59 #ifdef CONFIG_OSF_PARTITION
62 #ifdef CONFIG_SUN_PARTITION
65 #ifdef CONFIG_AMIGA_PARTITION
68 #ifdef CONFIG_ATARI_PARTITION
71 #ifdef CONFIG_MAC_PARTITION
74 #ifdef CONFIG_ULTRIX_PARTITION
77 #ifdef CONFIG_IBM_PARTITION
80 #ifdef CONFIG_KARMA_PARTITION
83 #ifdef CONFIG_SYSV68_PARTITION
89 static void bdev_set_nr_sectors(struct block_device *bdev, sector_t sectors)
91 spin_lock(&bdev->bd_size_lock);
92 i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
93 spin_unlock(&bdev->bd_size_lock);
96 static struct parsed_partitions *allocate_partitions(struct gendisk *hd)
98 struct parsed_partitions *state;
101 state = kzalloc(sizeof(*state), GFP_KERNEL);
105 nr = disk_max_parts(hd);
106 state->parts = vzalloc(array_size(nr, sizeof(state->parts[0])));
117 static void free_partitions(struct parsed_partitions *state)
123 static struct parsed_partitions *check_partition(struct gendisk *hd)
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';
138 state->bdev = hd->part0;
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 return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_partno);
188 static ssize_t part_start_show(struct device *dev,
189 struct device_attribute *attr, char *buf)
191 return sprintf(buf, "%llu\n", dev_to_bdev(dev)->bd_start_sect);
194 static ssize_t part_ro_show(struct device *dev,
195 struct device_attribute *attr, char *buf)
197 return sprintf(buf, "%d\n", bdev_read_only(dev_to_bdev(dev)));
200 static ssize_t part_alignment_offset_show(struct device *dev,
201 struct device_attribute *attr, char *buf)
203 struct block_device *bdev = dev_to_bdev(dev);
205 return sprintf(buf, "%u\n",
206 queue_limit_alignment_offset(&bdev->bd_disk->queue->limits,
207 bdev->bd_start_sect));
210 static ssize_t part_discard_alignment_show(struct device *dev,
211 struct device_attribute *attr, char *buf)
213 struct block_device *bdev = dev_to_bdev(dev);
215 return sprintf(buf, "%u\n",
216 queue_limit_discard_alignment(&bdev->bd_disk->queue->limits,
217 bdev->bd_start_sect));
220 static DEVICE_ATTR(partition, 0444, part_partition_show, NULL);
221 static DEVICE_ATTR(start, 0444, part_start_show, NULL);
222 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
223 static DEVICE_ATTR(ro, 0444, part_ro_show, NULL);
224 static DEVICE_ATTR(alignment_offset, 0444, part_alignment_offset_show, NULL);
225 static DEVICE_ATTR(discard_alignment, 0444, part_discard_alignment_show, NULL);
226 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
227 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
228 #ifdef CONFIG_FAIL_MAKE_REQUEST
229 static struct device_attribute dev_attr_fail =
230 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
233 static struct attribute *part_attrs[] = {
234 &dev_attr_partition.attr,
235 &dev_attr_start.attr,
238 &dev_attr_alignment_offset.attr,
239 &dev_attr_discard_alignment.attr,
241 &dev_attr_inflight.attr,
242 #ifdef CONFIG_FAIL_MAKE_REQUEST
248 static struct attribute_group part_attr_group = {
252 static const struct attribute_group *part_attr_groups[] = {
254 #ifdef CONFIG_BLK_DEV_IO_TRACE
255 &blk_trace_attr_group,
260 static void part_release(struct device *dev)
262 if (MAJOR(dev->devt) == BLOCK_EXT_MAJOR)
263 blk_free_ext_minor(MINOR(dev->devt));
264 bdput(dev_to_bdev(dev));
267 static int part_uevent(struct device *dev, struct kobj_uevent_env *env)
269 struct block_device *part = dev_to_bdev(dev);
271 add_uevent_var(env, "PARTN=%u", part->bd_partno);
272 if (part->bd_meta_info && part->bd_meta_info->volname[0])
273 add_uevent_var(env, "PARTNAME=%s", part->bd_meta_info->volname);
277 struct device_type part_type = {
279 .groups = part_attr_groups,
280 .release = part_release,
281 .uevent = part_uevent,
285 * Must be called either with open_mutex held, before a disk can be opened or
286 * after all disk users are gone.
288 static void delete_partition(struct block_device *part)
291 __invalidate_device(part, true);
293 xa_erase(&part->bd_disk->part_tbl, part->bd_partno);
294 kobject_put(part->bd_holder_dir);
295 device_del(&part->bd_device);
298 * Remove the block device from the inode hash, so that it cannot be
299 * looked up any more even when openers still hold references.
301 remove_inode_hash(part->bd_inode);
303 put_device(&part->bd_device);
306 static ssize_t whole_disk_show(struct device *dev,
307 struct device_attribute *attr, char *buf)
311 static DEVICE_ATTR(whole_disk, 0444, whole_disk_show, NULL);
314 * Must be called either with open_mutex held, before a disk can be opened or
315 * after all disk users are gone.
317 static struct block_device *add_partition(struct gendisk *disk, int partno,
318 sector_t start, sector_t len, int flags,
319 struct partition_meta_info *info)
321 dev_t devt = MKDEV(0, 0);
322 struct device *ddev = disk_to_dev(disk);
324 struct block_device *bdev;
328 lockdep_assert_held(&disk->open_mutex);
330 if (partno >= disk_max_parts(disk))
331 return ERR_PTR(-EINVAL);
334 * Partitions are not supported on zoned block devices that are used as
337 switch (disk->queue->limits.zoned) {
339 pr_warn("%s: partitions not supported on host managed zoned block device\n",
341 return ERR_PTR(-ENXIO);
343 pr_info("%s: disabling host aware zoned block device support due to partitions\n",
345 blk_queue_set_zoned(disk, BLK_ZONED_NONE);
351 if (xa_load(&disk->part_tbl, partno))
352 return ERR_PTR(-EBUSY);
354 bdev = bdev_alloc(disk, partno);
356 return ERR_PTR(-ENOMEM);
358 bdev->bd_start_sect = start;
359 bdev_set_nr_sectors(bdev, len);
363 bdev->bd_meta_info = kmemdup(info, sizeof(*info), GFP_KERNEL);
364 if (!bdev->bd_meta_info)
368 pdev = &bdev->bd_device;
369 dname = dev_name(ddev);
370 if (isdigit(dname[strlen(dname) - 1]))
371 dev_set_name(pdev, "%sp%d", dname, partno);
373 dev_set_name(pdev, "%s%d", dname, partno);
375 device_initialize(pdev);
376 pdev->class = &block_class;
377 pdev->type = &part_type;
380 /* in consecutive minor range? */
381 if (bdev->bd_partno < disk->minors) {
382 devt = MKDEV(disk->major, disk->first_minor + bdev->bd_partno);
384 err = blk_alloc_ext_minor();
387 devt = MKDEV(BLOCK_EXT_MAJOR, err);
391 /* delay uevent until 'holders' subdir is created */
392 dev_set_uevent_suppress(pdev, 1);
393 err = device_add(pdev);
398 bdev->bd_holder_dir = kobject_create_and_add("holders", &pdev->kobj);
399 if (!bdev->bd_holder_dir)
402 dev_set_uevent_suppress(pdev, 0);
403 if (flags & ADDPART_FLAG_WHOLEDISK) {
404 err = device_create_file(pdev, &dev_attr_whole_disk);
409 /* everything is up and running, commence */
410 err = xa_insert(&disk->part_tbl, partno, bdev, GFP_KERNEL);
413 bdev_add(bdev, devt);
415 /* suppress uevent if the disk suppresses it */
416 if (!dev_get_uevent_suppress(ddev))
417 kobject_uevent(&pdev->kobj, KOBJ_ADD);
424 kobject_put(bdev->bd_holder_dir);
431 static bool partition_overlaps(struct gendisk *disk, sector_t start,
432 sector_t length, int skip_partno)
434 struct block_device *part;
435 bool overlap = false;
439 xa_for_each_start(&disk->part_tbl, idx, part, 1) {
440 if (part->bd_partno != skip_partno &&
441 start < part->bd_start_sect + bdev_nr_sectors(part) &&
442 start + length > part->bd_start_sect) {
452 int bdev_add_partition(struct block_device *bdev, int partno,
453 sector_t start, sector_t length)
455 struct block_device *part;
457 mutex_lock(&bdev->bd_disk->open_mutex);
458 if (partition_overlaps(bdev->bd_disk, start, length, -1)) {
459 mutex_unlock(&bdev->bd_disk->open_mutex);
463 part = add_partition(bdev->bd_disk, partno, start, length,
464 ADDPART_FLAG_NONE, NULL);
465 mutex_unlock(&bdev->bd_disk->open_mutex);
466 return PTR_ERR_OR_ZERO(part);
469 int bdev_del_partition(struct block_device *bdev, int partno)
471 struct block_device *part = NULL;
474 mutex_lock(&bdev->bd_disk->open_mutex);
475 part = xa_load(&bdev->bd_disk->part_tbl, partno);
480 if (part->bd_openers)
483 delete_partition(part);
486 mutex_unlock(&bdev->bd_disk->open_mutex);
490 int bdev_resize_partition(struct block_device *bdev, int partno,
491 sector_t start, sector_t length)
493 struct block_device *part = NULL;
496 mutex_lock(&bdev->bd_disk->open_mutex);
497 part = xa_load(&bdev->bd_disk->part_tbl, partno);
502 if (start != part->bd_start_sect)
506 if (partition_overlaps(bdev->bd_disk, start, length, partno))
509 bdev_set_nr_sectors(part, length);
513 mutex_unlock(&bdev->bd_disk->open_mutex);
517 static bool disk_unlock_native_capacity(struct gendisk *disk)
519 const struct block_device_operations *bdops = disk->fops;
521 if (bdops->unlock_native_capacity &&
522 !(disk->flags & GENHD_FL_NATIVE_CAPACITY)) {
523 printk(KERN_CONT "enabling native capacity\n");
524 bdops->unlock_native_capacity(disk);
525 disk->flags |= GENHD_FL_NATIVE_CAPACITY;
528 printk(KERN_CONT "truncated\n");
533 void blk_drop_partitions(struct gendisk *disk)
535 struct block_device *part;
538 lockdep_assert_held(&disk->open_mutex);
540 xa_for_each_start(&disk->part_tbl, idx, part, 1) {
543 delete_partition(part);
548 static bool blk_add_partition(struct gendisk *disk,
549 struct parsed_partitions *state, int p)
551 sector_t size = state->parts[p].size;
552 sector_t from = state->parts[p].from;
553 struct block_device *part;
558 if (from >= get_capacity(disk)) {
560 "%s: p%d start %llu is beyond EOD, ",
561 disk->disk_name, p, (unsigned long long) from);
562 if (disk_unlock_native_capacity(disk))
567 if (from + size > get_capacity(disk)) {
569 "%s: p%d size %llu extends beyond EOD, ",
570 disk->disk_name, p, (unsigned long long) size);
572 if (disk_unlock_native_capacity(disk))
576 * We can not ignore partitions of broken tables created by for
577 * example camera firmware, but we limit them to the end of the
578 * disk to avoid creating invalid block devices.
580 size = get_capacity(disk) - from;
583 part = add_partition(disk, p, from, size, state->parts[p].flags,
584 &state->parts[p].info);
585 if (IS_ERR(part) && PTR_ERR(part) != -ENXIO) {
586 printk(KERN_ERR " %s: p%d could not be added: %ld\n",
587 disk->disk_name, p, -PTR_ERR(part));
591 if (IS_BUILTIN(CONFIG_BLK_DEV_MD) &&
592 (state->parts[p].flags & ADDPART_FLAG_RAID))
593 md_autodetect_dev(part->bd_dev);
598 static int blk_add_partitions(struct gendisk *disk)
600 struct parsed_partitions *state;
601 int ret = -EAGAIN, p;
603 if (!disk_part_scan_enabled(disk))
606 state = check_partition(disk);
611 * I/O error reading the partition table. If we tried to read
612 * beyond EOD, retry after unlocking the native capacity.
614 if (PTR_ERR(state) == -ENOSPC) {
615 printk(KERN_WARNING "%s: partition table beyond EOD, ",
617 if (disk_unlock_native_capacity(disk))
624 * Partitions are not supported on host managed zoned block devices.
626 if (disk->queue->limits.zoned == BLK_ZONED_HM) {
627 pr_warn("%s: ignoring partition table on host managed zoned block device\n",
634 * If we read beyond EOD, try unlocking native capacity even if the
635 * partition table was successfully read as we could be missing some
638 if (state->access_beyond_eod) {
640 "%s: partition table partially beyond EOD, ",
642 if (disk_unlock_native_capacity(disk))
646 /* tell userspace that the media / partition table may have changed */
647 kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
649 for (p = 1; p < state->limit; p++)
650 if (!blk_add_partition(disk, state, p))
655 free_partitions(state);
659 int bdev_disk_changed(struct gendisk *disk, bool invalidate)
663 lockdep_assert_held(&disk->open_mutex);
665 if (!(disk->flags & GENHD_FL_UP))
669 if (disk->open_partitions)
671 sync_blockdev(disk->part0);
672 invalidate_bdev(disk->part0);
673 blk_drop_partitions(disk);
675 clear_bit(GD_NEED_PART_SCAN, &disk->state);
678 * Historically we only set the capacity to zero for devices that
679 * support partitions (independ of actually having partitions created).
680 * Doing that is rather inconsistent, but changing it broke legacy
681 * udisks polling for legacy ide-cdrom devices. Use the crude check
682 * below to get the sane behavior for most device while not breaking
683 * userspace for this particular setup.
686 if (disk_part_scan_enabled(disk) ||
687 !(disk->flags & GENHD_FL_REMOVABLE))
688 set_capacity(disk, 0);
691 if (get_capacity(disk)) {
692 ret = blk_add_partitions(disk);
695 } else if (invalidate) {
697 * Tell userspace that the media / partition table may have
700 kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
706 * Only exported for loop and dasd for historic reasons. Don't use in new
709 EXPORT_SYMBOL_GPL(bdev_disk_changed);
711 void *read_part_sector(struct parsed_partitions *state, sector_t n, Sector *p)
713 struct address_space *mapping = state->bdev->bd_inode->i_mapping;
716 if (n >= get_capacity(state->bdev->bd_disk)) {
717 state->access_beyond_eod = true;
721 page = read_mapping_page(mapping,
722 (pgoff_t)(n >> (PAGE_SHIFT - 9)), NULL);
729 return (unsigned char *)page_address(page) +
730 ((n & ((1 << (PAGE_SHIFT - 9)) - 1)) << SECTOR_SHIFT);