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 delete_partition_work_fn(struct work_struct *work)
279 struct hd_struct *part = container_of(to_rcu_work(work), struct hd_struct,
282 part->start_sect = 0;
284 part_stat_set_all(part, 0);
285 put_device(part_to_dev(part));
288 void __delete_partition(struct percpu_ref *ref)
290 struct hd_struct *part = container_of(ref, struct hd_struct, ref);
291 INIT_RCU_WORK(&part->rcu_work, delete_partition_work_fn);
292 queue_rcu_work(system_wq, &part->rcu_work);
296 * Must be called either with bd_mutex held, before a disk can be opened or
297 * after all disk users are gone.
299 void delete_partition(struct gendisk *disk, int partno)
301 struct disk_part_tbl *ptbl =
302 rcu_dereference_protected(disk->part_tbl, 1);
303 struct hd_struct *part;
305 if (partno >= ptbl->len)
308 part = rcu_dereference_protected(ptbl->part[partno], 1);
312 rcu_assign_pointer(ptbl->part[partno], NULL);
313 rcu_assign_pointer(ptbl->last_lookup, NULL);
314 kobject_put(part->holder_dir);
315 device_del(part_to_dev(part));
318 * Remove gendisk pointer from idr so that it cannot be looked up
319 * while RCU period before freeing gendisk is running to prevent
320 * use-after-free issues. Note that the device number stays
321 * "in-use" until we really free the gendisk.
323 blk_invalidate_devt(part_devt(part));
324 hd_struct_kill(part);
327 static ssize_t whole_disk_show(struct device *dev,
328 struct device_attribute *attr, char *buf)
332 static DEVICE_ATTR(whole_disk, 0444, whole_disk_show, NULL);
335 * Must be called either with bd_mutex held, before a disk can be opened or
336 * after all disk users are gone.
338 struct hd_struct *add_partition(struct gendisk *disk, int partno,
339 sector_t start, sector_t len, int flags,
340 struct partition_meta_info *info)
343 dev_t devt = MKDEV(0, 0);
344 struct device *ddev = disk_to_dev(disk);
346 struct disk_part_tbl *ptbl;
351 * Partitions are not supported on zoned block devices that are used as
354 switch (disk->queue->limits.zoned) {
356 pr_warn("%s: partitions not supported on host managed zoned block device\n",
358 return ERR_PTR(-ENXIO);
360 pr_info("%s: disabling host aware zoned block device support due to partitions\n",
362 disk->queue->limits.zoned = BLK_ZONED_NONE;
368 err = disk_expand_part_tbl(disk, partno);
371 ptbl = rcu_dereference_protected(disk->part_tbl, 1);
373 if (ptbl->part[partno])
374 return ERR_PTR(-EBUSY);
376 p = kzalloc(sizeof(*p), GFP_KERNEL);
378 return ERR_PTR(-EBUSY);
380 if (!init_part_stats(p)) {
385 seqcount_init(&p->nr_sects_seq);
386 pdev = part_to_dev(p);
388 p->start_sect = start;
389 p->alignment_offset =
390 queue_limit_alignment_offset(&disk->queue->limits, start);
391 p->discard_alignment =
392 queue_limit_discard_alignment(&disk->queue->limits, start);
395 p->policy = get_disk_ro(disk);
398 struct partition_meta_info *pinfo;
400 pinfo = kzalloc_node(sizeof(*pinfo), GFP_KERNEL, disk->node_id);
405 memcpy(pinfo, info, sizeof(*info));
409 dname = dev_name(ddev);
410 if (isdigit(dname[strlen(dname) - 1]))
411 dev_set_name(pdev, "%sp%d", dname, partno);
413 dev_set_name(pdev, "%s%d", dname, partno);
415 device_initialize(pdev);
416 pdev->class = &block_class;
417 pdev->type = &part_type;
420 err = blk_alloc_devt(p, &devt);
425 /* delay uevent until 'holders' subdir is created */
426 dev_set_uevent_suppress(pdev, 1);
427 err = device_add(pdev);
432 p->holder_dir = kobject_create_and_add("holders", &pdev->kobj);
436 dev_set_uevent_suppress(pdev, 0);
437 if (flags & ADDPART_FLAG_WHOLEDISK) {
438 err = device_create_file(pdev, &dev_attr_whole_disk);
443 err = hd_ref_init(p);
445 if (flags & ADDPART_FLAG_WHOLEDISK)
446 goto out_remove_file;
450 /* everything is up and running, commence */
451 rcu_assign_pointer(ptbl->part[partno], p);
453 /* suppress uevent if the disk suppresses it */
454 if (!dev_get_uevent_suppress(ddev))
455 kobject_uevent(&pdev->kobj, KOBJ_ADD);
466 device_remove_file(pdev, &dev_attr_whole_disk);
468 kobject_put(p->holder_dir);
475 static bool disk_unlock_native_capacity(struct gendisk *disk)
477 const struct block_device_operations *bdops = disk->fops;
479 if (bdops->unlock_native_capacity &&
480 !(disk->flags & GENHD_FL_NATIVE_CAPACITY)) {
481 printk(KERN_CONT "enabling native capacity\n");
482 bdops->unlock_native_capacity(disk);
483 disk->flags |= GENHD_FL_NATIVE_CAPACITY;
486 printk(KERN_CONT "truncated\n");
491 int blk_drop_partitions(struct gendisk *disk, struct block_device *bdev)
493 struct disk_part_iter piter;
494 struct hd_struct *part;
497 if (!disk_part_scan_enabled(disk))
499 if (bdev->bd_part_count || bdev->bd_openers > 1)
501 res = invalidate_partition(disk, 0);
505 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
506 while ((part = disk_part_iter_next(&piter)))
507 delete_partition(disk, part->partno);
508 disk_part_iter_exit(&piter);
513 static bool blk_add_partition(struct gendisk *disk, struct block_device *bdev,
514 struct parsed_partitions *state, int p)
516 sector_t size = state->parts[p].size;
517 sector_t from = state->parts[p].from;
518 struct hd_struct *part;
523 if (from >= get_capacity(disk)) {
525 "%s: p%d start %llu is beyond EOD, ",
526 disk->disk_name, p, (unsigned long long) from);
527 if (disk_unlock_native_capacity(disk))
532 if (from + size > get_capacity(disk)) {
534 "%s: p%d size %llu extends beyond EOD, ",
535 disk->disk_name, p, (unsigned long long) size);
537 if (disk_unlock_native_capacity(disk))
541 * We can not ignore partitions of broken tables created by for
542 * example camera firmware, but we limit them to the end of the
543 * disk to avoid creating invalid block devices.
545 size = get_capacity(disk) - from;
548 part = add_partition(disk, p, from, size, state->parts[p].flags,
549 &state->parts[p].info);
550 if (IS_ERR(part) && PTR_ERR(part) != -ENXIO) {
551 printk(KERN_ERR " %s: p%d could not be added: %ld\n",
552 disk->disk_name, p, -PTR_ERR(part));
556 if (IS_BUILTIN(CONFIG_BLK_DEV_MD) &&
557 (state->parts[p].flags & ADDPART_FLAG_RAID))
558 md_autodetect_dev(part_to_dev(part)->devt);
563 int blk_add_partitions(struct gendisk *disk, struct block_device *bdev)
565 struct parsed_partitions *state;
566 int ret = -EAGAIN, p, highest;
568 if (!disk_part_scan_enabled(disk))
571 state = check_partition(disk, bdev);
576 * I/O error reading the partition table. If we tried to read
577 * beyond EOD, retry after unlocking the native capacity.
579 if (PTR_ERR(state) == -ENOSPC) {
580 printk(KERN_WARNING "%s: partition table beyond EOD, ",
582 if (disk_unlock_native_capacity(disk))
589 * Partitions are not supported on host managed zoned block devices.
591 if (disk->queue->limits.zoned == BLK_ZONED_HM) {
592 pr_warn("%s: ignoring partition table on host managed zoned block device\n",
599 * If we read beyond EOD, try unlocking native capacity even if the
600 * partition table was successfully read as we could be missing some
603 if (state->access_beyond_eod) {
605 "%s: partition table partially beyond EOD, ",
607 if (disk_unlock_native_capacity(disk))
611 /* tell userspace that the media / partition table may have changed */
612 kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
615 * Detect the highest partition number and preallocate disk->part_tbl.
616 * This is an optimization and not strictly necessary.
618 for (p = 1, highest = 0; p < state->limit; p++)
619 if (state->parts[p].size)
621 disk_expand_part_tbl(disk, highest);
623 for (p = 1; p < state->limit; p++)
624 if (!blk_add_partition(disk, bdev, state, p))
629 free_partitions(state);
633 void *read_part_sector(struct parsed_partitions *state, sector_t n, Sector *p)
635 struct address_space *mapping = state->bdev->bd_inode->i_mapping;
638 if (n >= get_capacity(state->bdev->bd_disk)) {
639 state->access_beyond_eod = true;
643 page = read_mapping_page(mapping,
644 (pgoff_t)(n >> (PAGE_SHIFT - 9)), NULL);
651 return (unsigned char *)page_address(page) +
652 ((n & ((1 << (PAGE_SHIFT - 9)) - 1)) << SECTOR_SHIFT);