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 struct parsed_partitions *allocate_partitions(struct gendisk *hd)
90 struct parsed_partitions *state;
93 state = kzalloc(sizeof(*state), GFP_KERNEL);
97 nr = disk_max_parts(hd);
98 state->parts = vzalloc(array_size(nr, sizeof(state->parts[0])));
109 static void free_partitions(struct parsed_partitions *state)
115 static struct parsed_partitions *check_partition(struct gendisk *hd,
116 struct block_device *bdev)
118 struct parsed_partitions *state;
121 state = allocate_partitions(hd);
124 state->pp_buf = (char *)__get_free_page(GFP_KERNEL);
125 if (!state->pp_buf) {
126 free_partitions(state);
129 state->pp_buf[0] = '\0';
132 disk_name(hd, 0, state->name);
133 snprintf(state->pp_buf, PAGE_SIZE, " %s:", state->name);
134 if (isdigit(state->name[strlen(state->name)-1]))
135 sprintf(state->name, "p");
138 while (!res && check_part[i]) {
139 memset(state->parts, 0, state->limit * sizeof(state->parts[0]));
140 res = check_part[i++](state);
143 * We have hit an I/O error which we don't report now.
144 * But record it, and let the others do their job.
152 printk(KERN_INFO "%s", state->pp_buf);
154 free_page((unsigned long)state->pp_buf);
157 if (state->access_beyond_eod)
160 * The partition is unrecognized. So report I/O errors if there were any
165 strlcat(state->pp_buf,
166 " unable to read partition table\n", PAGE_SIZE);
167 printk(KERN_INFO "%s", state->pp_buf);
170 free_page((unsigned long)state->pp_buf);
171 free_partitions(state);
175 static ssize_t part_partition_show(struct device *dev,
176 struct device_attribute *attr, char *buf)
178 struct hd_struct *p = dev_to_part(dev);
180 return sprintf(buf, "%d\n", p->partno);
183 static ssize_t part_start_show(struct device *dev,
184 struct device_attribute *attr, char *buf)
186 struct hd_struct *p = dev_to_part(dev);
188 return sprintf(buf, "%llu\n",(unsigned long long)p->start_sect);
191 static ssize_t part_ro_show(struct device *dev,
192 struct device_attribute *attr, char *buf)
194 struct hd_struct *p = dev_to_part(dev);
195 return sprintf(buf, "%d\n", p->policy ? 1 : 0);
198 static ssize_t part_alignment_offset_show(struct device *dev,
199 struct device_attribute *attr, char *buf)
201 struct hd_struct *p = dev_to_part(dev);
202 return sprintf(buf, "%llu\n", (unsigned long long)p->alignment_offset);
205 static ssize_t part_discard_alignment_show(struct device *dev,
206 struct device_attribute *attr, char *buf)
208 struct hd_struct *p = dev_to_part(dev);
209 return sprintf(buf, "%u\n", p->discard_alignment);
212 static DEVICE_ATTR(partition, 0444, part_partition_show, NULL);
213 static DEVICE_ATTR(start, 0444, part_start_show, NULL);
214 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
215 static DEVICE_ATTR(ro, 0444, part_ro_show, NULL);
216 static DEVICE_ATTR(alignment_offset, 0444, part_alignment_offset_show, NULL);
217 static DEVICE_ATTR(discard_alignment, 0444, part_discard_alignment_show, NULL);
218 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
219 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
220 #ifdef CONFIG_FAIL_MAKE_REQUEST
221 static struct device_attribute dev_attr_fail =
222 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
225 static struct attribute *part_attrs[] = {
226 &dev_attr_partition.attr,
227 &dev_attr_start.attr,
230 &dev_attr_alignment_offset.attr,
231 &dev_attr_discard_alignment.attr,
233 &dev_attr_inflight.attr,
234 #ifdef CONFIG_FAIL_MAKE_REQUEST
240 static struct attribute_group part_attr_group = {
244 static const struct attribute_group *part_attr_groups[] = {
246 #ifdef CONFIG_BLK_DEV_IO_TRACE
247 &blk_trace_attr_group,
252 static void part_release(struct device *dev)
254 struct hd_struct *p = dev_to_part(dev);
255 blk_free_devt(dev->devt);
260 static int part_uevent(struct device *dev, struct kobj_uevent_env *env)
262 struct hd_struct *part = dev_to_part(dev);
264 add_uevent_var(env, "PARTN=%u", part->partno);
265 if (part->info && part->info->volname[0])
266 add_uevent_var(env, "PARTNAME=%s", part->info->volname);
270 struct device_type part_type = {
272 .groups = part_attr_groups,
273 .release = part_release,
274 .uevent = part_uevent,
277 static void hd_struct_free_work(struct work_struct *work)
279 struct hd_struct *part =
280 container_of(to_rcu_work(work), struct hd_struct, rcu_work);
282 part->start_sect = 0;
284 part_stat_set_all(part, 0);
285 put_device(part_to_dev(part));
288 static void hd_struct_free(struct percpu_ref *ref)
290 struct hd_struct *part = container_of(ref, struct hd_struct, ref);
291 struct gendisk *disk = part_to_disk(part);
292 struct disk_part_tbl *ptbl =
293 rcu_dereference_protected(disk->part_tbl, 1);
295 rcu_assign_pointer(ptbl->last_lookup, NULL);
296 put_device(disk_to_dev(disk));
298 INIT_RCU_WORK(&part->rcu_work, hd_struct_free_work);
299 queue_rcu_work(system_wq, &part->rcu_work);
302 int hd_ref_init(struct hd_struct *part)
304 if (percpu_ref_init(&part->ref, hd_struct_free, 0, GFP_KERNEL))
310 * Must be called either with bd_mutex held, before a disk can be opened or
311 * after all disk users are gone.
313 void delete_partition(struct gendisk *disk, struct hd_struct *part)
315 struct disk_part_tbl *ptbl =
316 rcu_dereference_protected(disk->part_tbl, 1);
319 * ->part_tbl is referenced in this part's release handler, so
320 * we have to hold the disk device
322 get_device(disk_to_dev(part_to_disk(part)));
323 rcu_assign_pointer(ptbl->part[part->partno], NULL);
324 kobject_put(part->holder_dir);
325 device_del(part_to_dev(part));
328 * Remove gendisk pointer from idr so that it cannot be looked up
329 * while RCU period before freeing gendisk is running to prevent
330 * use-after-free issues. Note that the device number stays
331 * "in-use" until we really free the gendisk.
333 blk_invalidate_devt(part_devt(part));
334 percpu_ref_kill(&part->ref);
337 static ssize_t whole_disk_show(struct device *dev,
338 struct device_attribute *attr, char *buf)
342 static DEVICE_ATTR(whole_disk, 0444, whole_disk_show, NULL);
345 * Must be called either with bd_mutex held, before a disk can be opened or
346 * after all disk users are gone.
348 static struct hd_struct *add_partition(struct gendisk *disk, int partno,
349 sector_t start, sector_t len, int flags,
350 struct partition_meta_info *info)
353 dev_t devt = MKDEV(0, 0);
354 struct device *ddev = disk_to_dev(disk);
356 struct disk_part_tbl *ptbl;
361 * Partitions are not supported on zoned block devices that are used as
364 switch (disk->queue->limits.zoned) {
366 pr_warn("%s: partitions not supported on host managed zoned block device\n",
368 return ERR_PTR(-ENXIO);
370 pr_info("%s: disabling host aware zoned block device support due to partitions\n",
372 disk->queue->limits.zoned = BLK_ZONED_NONE;
378 err = disk_expand_part_tbl(disk, partno);
381 ptbl = rcu_dereference_protected(disk->part_tbl, 1);
383 if (ptbl->part[partno])
384 return ERR_PTR(-EBUSY);
386 p = kzalloc(sizeof(*p), GFP_KERNEL);
388 return ERR_PTR(-EBUSY);
390 p->dkstats = alloc_percpu(struct disk_stats);
396 hd_sects_seq_init(p);
397 pdev = part_to_dev(p);
399 p->start_sect = start;
400 p->alignment_offset =
401 queue_limit_alignment_offset(&disk->queue->limits, start);
402 p->discard_alignment =
403 queue_limit_discard_alignment(&disk->queue->limits, start);
406 p->policy = get_disk_ro(disk);
409 struct partition_meta_info *pinfo;
411 pinfo = kzalloc_node(sizeof(*pinfo), GFP_KERNEL, disk->node_id);
416 memcpy(pinfo, info, sizeof(*info));
420 dname = dev_name(ddev);
421 if (isdigit(dname[strlen(dname) - 1]))
422 dev_set_name(pdev, "%sp%d", dname, partno);
424 dev_set_name(pdev, "%s%d", dname, partno);
426 device_initialize(pdev);
427 pdev->class = &block_class;
428 pdev->type = &part_type;
431 err = blk_alloc_devt(p, &devt);
436 /* delay uevent until 'holders' subdir is created */
437 dev_set_uevent_suppress(pdev, 1);
438 err = device_add(pdev);
443 p->holder_dir = kobject_create_and_add("holders", &pdev->kobj);
447 dev_set_uevent_suppress(pdev, 0);
448 if (flags & ADDPART_FLAG_WHOLEDISK) {
449 err = device_create_file(pdev, &dev_attr_whole_disk);
454 err = hd_ref_init(p);
456 if (flags & ADDPART_FLAG_WHOLEDISK)
457 goto out_remove_file;
461 /* everything is up and running, commence */
462 rcu_assign_pointer(ptbl->part[partno], p);
464 /* suppress uevent if the disk suppresses it */
465 if (!dev_get_uevent_suppress(ddev))
466 kobject_uevent(&pdev->kobj, KOBJ_ADD);
472 free_percpu(p->dkstats);
477 device_remove_file(pdev, &dev_attr_whole_disk);
479 kobject_put(p->holder_dir);
486 static bool partition_overlaps(struct gendisk *disk, sector_t start,
487 sector_t length, int skip_partno)
489 struct disk_part_iter piter;
490 struct hd_struct *part;
491 bool overlap = false;
493 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
494 while ((part = disk_part_iter_next(&piter))) {
495 if (part->partno == skip_partno ||
496 start >= part->start_sect + part->nr_sects ||
497 start + length <= part->start_sect)
503 disk_part_iter_exit(&piter);
507 int bdev_add_partition(struct block_device *bdev, int partno,
508 sector_t start, sector_t length)
510 struct hd_struct *part;
512 mutex_lock(&bdev->bd_mutex);
513 if (partition_overlaps(bdev->bd_disk, start, length, -1)) {
514 mutex_unlock(&bdev->bd_mutex);
518 part = add_partition(bdev->bd_disk, partno, start, length,
519 ADDPART_FLAG_NONE, NULL);
520 mutex_unlock(&bdev->bd_mutex);
521 return PTR_ERR_OR_ZERO(part);
524 int bdev_del_partition(struct block_device *bdev, int partno)
526 struct block_device *bdevp;
527 struct hd_struct *part;
530 part = disk_get_part(bdev->bd_disk, partno);
535 bdevp = bdget(part_devt(part));
539 mutex_lock(&bdevp->bd_mutex);
542 if (bdevp->bd_openers)
545 sync_blockdev(bdevp);
546 invalidate_bdev(bdevp);
548 mutex_lock_nested(&bdev->bd_mutex, 1);
549 delete_partition(bdev->bd_disk, part);
550 mutex_unlock(&bdev->bd_mutex);
554 mutex_unlock(&bdevp->bd_mutex);
561 int bdev_resize_partition(struct block_device *bdev, int partno,
562 sector_t start, sector_t length)
564 struct block_device *bdevp;
565 struct hd_struct *part;
568 part = disk_get_part(bdev->bd_disk, partno);
573 bdevp = bdget(part_devt(part));
577 mutex_lock(&bdevp->bd_mutex);
578 mutex_lock_nested(&bdev->bd_mutex, 1);
581 if (start != part->start_sect)
585 if (partition_overlaps(bdev->bd_disk, start, length, partno))
588 part_nr_sects_write(part, (sector_t)length);
589 i_size_write(bdevp->bd_inode, length << SECTOR_SHIFT);
593 mutex_unlock(&bdevp->bd_mutex);
594 mutex_unlock(&bdev->bd_mutex);
601 static bool disk_unlock_native_capacity(struct gendisk *disk)
603 const struct block_device_operations *bdops = disk->fops;
605 if (bdops->unlock_native_capacity &&
606 !(disk->flags & GENHD_FL_NATIVE_CAPACITY)) {
607 printk(KERN_CONT "enabling native capacity\n");
608 bdops->unlock_native_capacity(disk);
609 disk->flags |= GENHD_FL_NATIVE_CAPACITY;
612 printk(KERN_CONT "truncated\n");
617 int blk_drop_partitions(struct block_device *bdev)
619 struct disk_part_iter piter;
620 struct hd_struct *part;
622 if (bdev->bd_part_count)
626 invalidate_bdev(bdev);
628 disk_part_iter_init(&piter, bdev->bd_disk, DISK_PITER_INCL_EMPTY);
629 while ((part = disk_part_iter_next(&piter)))
630 delete_partition(bdev->bd_disk, part);
631 disk_part_iter_exit(&piter);
636 /* for historic reasons in the DASD driver */
637 EXPORT_SYMBOL_GPL(blk_drop_partitions);
640 static bool blk_add_partition(struct gendisk *disk, struct block_device *bdev,
641 struct parsed_partitions *state, int p)
643 sector_t size = state->parts[p].size;
644 sector_t from = state->parts[p].from;
645 struct hd_struct *part;
650 if (from >= get_capacity(disk)) {
652 "%s: p%d start %llu is beyond EOD, ",
653 disk->disk_name, p, (unsigned long long) from);
654 if (disk_unlock_native_capacity(disk))
659 if (from + size > get_capacity(disk)) {
661 "%s: p%d size %llu extends beyond EOD, ",
662 disk->disk_name, p, (unsigned long long) size);
664 if (disk_unlock_native_capacity(disk))
668 * We can not ignore partitions of broken tables created by for
669 * example camera firmware, but we limit them to the end of the
670 * disk to avoid creating invalid block devices.
672 size = get_capacity(disk) - from;
675 part = add_partition(disk, p, from, size, state->parts[p].flags,
676 &state->parts[p].info);
677 if (IS_ERR(part) && PTR_ERR(part) != -ENXIO) {
678 printk(KERN_ERR " %s: p%d could not be added: %ld\n",
679 disk->disk_name, p, -PTR_ERR(part));
683 if (IS_BUILTIN(CONFIG_BLK_DEV_MD) &&
684 (state->parts[p].flags & ADDPART_FLAG_RAID))
685 md_autodetect_dev(part_to_dev(part)->devt);
690 int blk_add_partitions(struct gendisk *disk, struct block_device *bdev)
692 struct parsed_partitions *state;
693 int ret = -EAGAIN, p, highest;
695 if (!disk_part_scan_enabled(disk))
698 state = check_partition(disk, bdev);
703 * I/O error reading the partition table. If we tried to read
704 * beyond EOD, retry after unlocking the native capacity.
706 if (PTR_ERR(state) == -ENOSPC) {
707 printk(KERN_WARNING "%s: partition table beyond EOD, ",
709 if (disk_unlock_native_capacity(disk))
716 * Partitions are not supported on host managed zoned block devices.
718 if (disk->queue->limits.zoned == BLK_ZONED_HM) {
719 pr_warn("%s: ignoring partition table on host managed zoned block device\n",
726 * If we read beyond EOD, try unlocking native capacity even if the
727 * partition table was successfully read as we could be missing some
730 if (state->access_beyond_eod) {
732 "%s: partition table partially beyond EOD, ",
734 if (disk_unlock_native_capacity(disk))
738 /* tell userspace that the media / partition table may have changed */
739 kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
742 * Detect the highest partition number and preallocate disk->part_tbl.
743 * This is an optimization and not strictly necessary.
745 for (p = 1, highest = 0; p < state->limit; p++)
746 if (state->parts[p].size)
748 disk_expand_part_tbl(disk, highest);
750 for (p = 1; p < state->limit; p++)
751 if (!blk_add_partition(disk, bdev, state, p))
756 free_partitions(state);
760 void *read_part_sector(struct parsed_partitions *state, sector_t n, Sector *p)
762 struct address_space *mapping = state->bdev->bd_inode->i_mapping;
765 if (n >= get_capacity(state->bdev->bd_disk)) {
766 state->access_beyond_eod = true;
770 page = read_mapping_page(mapping,
771 (pgoff_t)(n >> (PAGE_SHIFT - 9)), NULL);
778 return (unsigned char *)page_address(page) +
779 ((n & ((1 << (PAGE_SHIFT - 9)) - 1)) << SECTOR_SHIFT);