Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
94ea4158 | 2 | /* |
387048bf CH |
3 | * Copyright (C) 1991-1998 Linus Torvalds |
4 | * Re-organised Feb 1998 Russell King | |
7b51e703 | 5 | * Copyright (C) 2020 Christoph Hellwig |
94ea4158 | 6 | */ |
94ea4158 | 7 | #include <linux/fs.h> |
b81e0c23 | 8 | #include <linux/major.h> |
94ea4158 | 9 | #include <linux/slab.h> |
94ea4158 | 10 | #include <linux/ctype.h> |
387048bf | 11 | #include <linux/vmalloc.h> |
74cc979c | 12 | #include <linux/raid/detect.h> |
387048bf CH |
13 | #include "check.h" |
14 | ||
539050f9 | 15 | static int (*const check_part[])(struct parsed_partitions *) = { |
387048bf CH |
16 | /* |
17 | * Probe partition formats with tables at disk address 0 | |
18 | * that also have an ADFS boot block at 0xdc0. | |
19 | */ | |
20 | #ifdef CONFIG_ACORN_PARTITION_ICS | |
21 | adfspart_check_ICS, | |
22 | #endif | |
23 | #ifdef CONFIG_ACORN_PARTITION_POWERTEC | |
24 | adfspart_check_POWERTEC, | |
25 | #endif | |
26 | #ifdef CONFIG_ACORN_PARTITION_EESOX | |
27 | adfspart_check_EESOX, | |
28 | #endif | |
29 | ||
30 | /* | |
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 | |
34 | * the msdos entry. | |
35 | */ | |
36 | #ifdef CONFIG_ACORN_PARTITION_CUMANA | |
37 | adfspart_check_CUMANA, | |
38 | #endif | |
39 | #ifdef CONFIG_ACORN_PARTITION_ADFS | |
40 | adfspart_check_ADFS, | |
41 | #endif | |
42 | ||
43 | #ifdef CONFIG_CMDLINE_PARTITION | |
44 | cmdline_partition, | |
45 | #endif | |
46 | #ifdef CONFIG_EFI_PARTITION | |
47 | efi_partition, /* this must come before msdos */ | |
48 | #endif | |
49 | #ifdef CONFIG_SGI_PARTITION | |
50 | sgi_partition, | |
51 | #endif | |
52 | #ifdef CONFIG_LDM_PARTITION | |
53 | ldm_partition, /* this must come before msdos */ | |
54 | #endif | |
55 | #ifdef CONFIG_MSDOS_PARTITION | |
56 | msdos_partition, | |
57 | #endif | |
58 | #ifdef CONFIG_OSF_PARTITION | |
59 | osf_partition, | |
60 | #endif | |
61 | #ifdef CONFIG_SUN_PARTITION | |
62 | sun_partition, | |
63 | #endif | |
64 | #ifdef CONFIG_AMIGA_PARTITION | |
65 | amiga_partition, | |
66 | #endif | |
67 | #ifdef CONFIG_ATARI_PARTITION | |
68 | atari_partition, | |
69 | #endif | |
70 | #ifdef CONFIG_MAC_PARTITION | |
71 | mac_partition, | |
72 | #endif | |
73 | #ifdef CONFIG_ULTRIX_PARTITION | |
74 | ultrix_partition, | |
75 | #endif | |
76 | #ifdef CONFIG_IBM_PARTITION | |
77 | ibm_partition, | |
78 | #endif | |
79 | #ifdef CONFIG_KARMA_PARTITION | |
80 | karma_partition, | |
81 | #endif | |
82 | #ifdef CONFIG_SYSV68_PARTITION | |
83 | sysv68_partition, | |
84 | #endif | |
85 | NULL | |
86 | }; | |
87 | ||
88 | static struct parsed_partitions *allocate_partitions(struct gendisk *hd) | |
89 | { | |
90 | struct parsed_partitions *state; | |
1ebe2e5f | 91 | int nr = DISK_MAX_PARTS; |
387048bf CH |
92 | |
93 | state = kzalloc(sizeof(*state), GFP_KERNEL); | |
94 | if (!state) | |
95 | return NULL; | |
96 | ||
387048bf CH |
97 | state->parts = vzalloc(array_size(nr, sizeof(state->parts[0]))); |
98 | if (!state->parts) { | |
99 | kfree(state); | |
100 | return NULL; | |
101 | } | |
102 | ||
103 | state->limit = nr; | |
104 | ||
105 | return state; | |
106 | } | |
107 | ||
108 | static void free_partitions(struct parsed_partitions *state) | |
109 | { | |
110 | vfree(state->parts); | |
111 | kfree(state); | |
112 | } | |
113 | ||
0384264e | 114 | static struct parsed_partitions *check_partition(struct gendisk *hd) |
387048bf CH |
115 | { |
116 | struct parsed_partitions *state; | |
117 | int i, res, err; | |
118 | ||
119 | state = allocate_partitions(hd); | |
120 | if (!state) | |
121 | return NULL; | |
122 | state->pp_buf = (char *)__get_free_page(GFP_KERNEL); | |
123 | if (!state->pp_buf) { | |
124 | free_partitions(state); | |
125 | return NULL; | |
126 | } | |
127 | state->pp_buf[0] = '\0'; | |
128 | ||
a08aa9bc | 129 | state->disk = hd; |
1d703547 | 130 | snprintf(state->name, BDEVNAME_SIZE, "%s", hd->disk_name); |
387048bf CH |
131 | snprintf(state->pp_buf, PAGE_SIZE, " %s:", state->name); |
132 | if (isdigit(state->name[strlen(state->name)-1])) | |
133 | sprintf(state->name, "p"); | |
134 | ||
135 | i = res = err = 0; | |
136 | while (!res && check_part[i]) { | |
137 | memset(state->parts, 0, state->limit * sizeof(state->parts[0])); | |
138 | res = check_part[i++](state); | |
139 | if (res < 0) { | |
140 | /* | |
141 | * We have hit an I/O error which we don't report now. | |
142 | * But record it, and let the others do their job. | |
143 | */ | |
144 | err = res; | |
145 | res = 0; | |
146 | } | |
147 | ||
148 | } | |
149 | if (res > 0) { | |
150 | printk(KERN_INFO "%s", state->pp_buf); | |
151 | ||
152 | free_page((unsigned long)state->pp_buf); | |
153 | return state; | |
154 | } | |
155 | if (state->access_beyond_eod) | |
156 | err = -ENOSPC; | |
157 | /* | |
158 | * The partition is unrecognized. So report I/O errors if there were any | |
159 | */ | |
160 | if (err) | |
161 | res = err; | |
162 | if (res) { | |
163 | strlcat(state->pp_buf, | |
164 | " unable to read partition table\n", PAGE_SIZE); | |
165 | printk(KERN_INFO "%s", state->pp_buf); | |
166 | } | |
94ea4158 | 167 | |
387048bf CH |
168 | free_page((unsigned long)state->pp_buf); |
169 | free_partitions(state); | |
170 | return ERR_PTR(res); | |
171 | } | |
94ea4158 | 172 | |
94ea4158 AV |
173 | static ssize_t part_partition_show(struct device *dev, |
174 | struct device_attribute *attr, char *buf) | |
175 | { | |
0d02129e | 176 | return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_partno); |
94ea4158 AV |
177 | } |
178 | ||
179 | static ssize_t part_start_show(struct device *dev, | |
180 | struct device_attribute *attr, char *buf) | |
181 | { | |
0d02129e | 182 | return sprintf(buf, "%llu\n", dev_to_bdev(dev)->bd_start_sect); |
94ea4158 AV |
183 | } |
184 | ||
94ea4158 AV |
185 | static ssize_t part_ro_show(struct device *dev, |
186 | struct device_attribute *attr, char *buf) | |
187 | { | |
52f019d4 | 188 | return sprintf(buf, "%d\n", bdev_read_only(dev_to_bdev(dev))); |
94ea4158 AV |
189 | } |
190 | ||
191 | static ssize_t part_alignment_offset_show(struct device *dev, | |
192 | struct device_attribute *attr, char *buf) | |
193 | { | |
64dcc7c2 | 194 | return sprintf(buf, "%u\n", bdev_alignment_offset(dev_to_bdev(dev))); |
94ea4158 AV |
195 | } |
196 | ||
197 | static ssize_t part_discard_alignment_show(struct device *dev, | |
198 | struct device_attribute *attr, char *buf) | |
199 | { | |
f0f975a4 | 200 | return sprintf(buf, "%u\n", bdev_discard_alignment(dev_to_bdev(dev))); |
94ea4158 AV |
201 | } |
202 | ||
5657a819 JP |
203 | static DEVICE_ATTR(partition, 0444, part_partition_show, NULL); |
204 | static DEVICE_ATTR(start, 0444, part_start_show, NULL); | |
205 | static DEVICE_ATTR(size, 0444, part_size_show, NULL); | |
206 | static DEVICE_ATTR(ro, 0444, part_ro_show, NULL); | |
207 | static DEVICE_ATTR(alignment_offset, 0444, part_alignment_offset_show, NULL); | |
208 | static DEVICE_ATTR(discard_alignment, 0444, part_discard_alignment_show, NULL); | |
209 | static DEVICE_ATTR(stat, 0444, part_stat_show, NULL); | |
210 | static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL); | |
94ea4158 AV |
211 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
212 | static struct device_attribute dev_attr_fail = | |
5657a819 | 213 | __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store); |
94ea4158 AV |
214 | #endif |
215 | ||
216 | static struct attribute *part_attrs[] = { | |
217 | &dev_attr_partition.attr, | |
218 | &dev_attr_start.attr, | |
219 | &dev_attr_size.attr, | |
220 | &dev_attr_ro.attr, | |
221 | &dev_attr_alignment_offset.attr, | |
222 | &dev_attr_discard_alignment.attr, | |
223 | &dev_attr_stat.attr, | |
224 | &dev_attr_inflight.attr, | |
225 | #ifdef CONFIG_FAIL_MAKE_REQUEST | |
226 | &dev_attr_fail.attr, | |
227 | #endif | |
228 | NULL | |
229 | }; | |
230 | ||
0bd47800 | 231 | static const struct attribute_group part_attr_group = { |
94ea4158 AV |
232 | .attrs = part_attrs, |
233 | }; | |
234 | ||
235 | static const struct attribute_group *part_attr_groups[] = { | |
236 | &part_attr_group, | |
237 | #ifdef CONFIG_BLK_DEV_IO_TRACE | |
238 | &blk_trace_attr_group, | |
239 | #endif | |
240 | NULL | |
241 | }; | |
242 | ||
243 | static void part_release(struct device *dev) | |
244 | { | |
9d3b8813 | 245 | put_disk(dev_to_bdev(dev)->bd_disk); |
2f4731dc | 246 | iput(dev_to_bdev(dev)->bd_inode); |
94ea4158 AV |
247 | } |
248 | ||
162736b0 | 249 | static int part_uevent(const struct device *dev, struct kobj_uevent_env *env) |
0d9c51a6 | 250 | { |
162736b0 | 251 | const struct block_device *part = dev_to_bdev(dev); |
0d9c51a6 | 252 | |
0d02129e CH |
253 | add_uevent_var(env, "PARTN=%u", part->bd_partno); |
254 | if (part->bd_meta_info && part->bd_meta_info->volname[0]) | |
255 | add_uevent_var(env, "PARTNAME=%s", part->bd_meta_info->volname); | |
0d9c51a6 SM |
256 | return 0; |
257 | } | |
258 | ||
cdb37f73 | 259 | const struct device_type part_type = { |
94ea4158 AV |
260 | .name = "partition", |
261 | .groups = part_attr_groups, | |
262 | .release = part_release, | |
0d9c51a6 | 263 | .uevent = part_uevent, |
94ea4158 AV |
264 | }; |
265 | ||
eec1be4c | 266 | void drop_partition(struct block_device *part) |
94ea4158 | 267 | { |
a45e43ca CH |
268 | lockdep_assert_held(&part->bd_disk->open_mutex); |
269 | ||
eec1be4c CH |
270 | xa_erase(&part->bd_disk->part_tbl, part->bd_partno); |
271 | kobject_put(part->bd_holder_dir); | |
272 | ||
273 | device_del(&part->bd_device); | |
274 | put_device(&part->bd_device); | |
275 | } | |
276 | ||
94ea4158 AV |
277 | static ssize_t whole_disk_show(struct device *dev, |
278 | struct device_attribute *attr, char *buf) | |
279 | { | |
280 | return 0; | |
281 | } | |
a378f6a4 | 282 | static const DEVICE_ATTR(whole_disk, 0444, whole_disk_show, NULL); |
94ea4158 | 283 | |
6d2cf6f2 | 284 | /* |
a8698707 | 285 | * Must be called either with open_mutex held, before a disk can be opened or |
6d2cf6f2 BVA |
286 | * after all disk users are gone. |
287 | */ | |
0d02129e | 288 | static struct block_device *add_partition(struct gendisk *disk, int partno, |
94ea4158 AV |
289 | sector_t start, sector_t len, int flags, |
290 | struct partition_meta_info *info) | |
291 | { | |
94ea4158 AV |
292 | dev_t devt = MKDEV(0, 0); |
293 | struct device *ddev = disk_to_dev(disk); | |
294 | struct device *pdev; | |
22ae8ce8 | 295 | struct block_device *bdev; |
94ea4158 AV |
296 | const char *dname; |
297 | int err; | |
298 | ||
0e0ccdec CH |
299 | lockdep_assert_held(&disk->open_mutex); |
300 | ||
1ebe2e5f | 301 | if (partno >= DISK_MAX_PARTS) |
e82fc785 ML |
302 | return ERR_PTR(-EINVAL); |
303 | ||
b7205307 CH |
304 | /* |
305 | * Partitions are not supported on zoned block devices that are used as | |
306 | * such. | |
307 | */ | |
308 | switch (disk->queue->limits.zoned) { | |
309 | case BLK_ZONED_HM: | |
310 | pr_warn("%s: partitions not supported on host managed zoned block device\n", | |
311 | disk->disk_name); | |
312 | return ERR_PTR(-ENXIO); | |
313 | case BLK_ZONED_HA: | |
314 | pr_info("%s: disabling host aware zoned block device support due to partitions\n", | |
315 | disk->disk_name); | |
6b2bd274 | 316 | disk_set_zoned(disk, BLK_ZONED_NONE); |
b7205307 CH |
317 | break; |
318 | case BLK_ZONED_NONE: | |
319 | break; | |
320 | } | |
321 | ||
a33df75c | 322 | if (xa_load(&disk->part_tbl, partno)) |
94ea4158 AV |
323 | return ERR_PTR(-EBUSY); |
324 | ||
9d3b8813 CH |
325 | /* ensure we always have a reference to the whole disk */ |
326 | get_device(disk_to_dev(disk)); | |
327 | ||
328 | err = -ENOMEM; | |
22ae8ce8 CH |
329 | bdev = bdev_alloc(disk, partno); |
330 | if (!bdev) | |
9d3b8813 | 331 | goto out_put_disk; |
c83f6bf9 | 332 | |
29ff57c6 | 333 | bdev->bd_start_sect = start; |
a782483c | 334 | bdev_set_nr_sectors(bdev, len); |
94ea4158 | 335 | |
0d02129e | 336 | pdev = &bdev->bd_device; |
94ea4158 AV |
337 | dname = dev_name(ddev); |
338 | if (isdigit(dname[strlen(dname) - 1])) | |
339 | dev_set_name(pdev, "%sp%d", dname, partno); | |
340 | else | |
341 | dev_set_name(pdev, "%s%d", dname, partno); | |
342 | ||
343 | device_initialize(pdev); | |
344 | pdev->class = &block_class; | |
345 | pdev->type = &part_type; | |
346 | pdev->parent = ddev; | |
347 | ||
7c3f828b CH |
348 | /* in consecutive minor range? */ |
349 | if (bdev->bd_partno < disk->minors) { | |
350 | devt = MKDEV(disk->major, disk->first_minor + bdev->bd_partno); | |
351 | } else { | |
352 | err = blk_alloc_ext_minor(); | |
353 | if (err < 0) | |
354 | goto out_put; | |
355 | devt = MKDEV(BLOCK_EXT_MAJOR, err); | |
356 | } | |
94ea4158 AV |
357 | pdev->devt = devt; |
358 | ||
0468c532 CH |
359 | if (info) { |
360 | err = -ENOMEM; | |
361 | bdev->bd_meta_info = kmemdup(info, sizeof(*info), GFP_KERNEL); | |
362 | if (!bdev->bd_meta_info) | |
363 | goto out_put; | |
364 | } | |
365 | ||
94ea4158 AV |
366 | /* delay uevent until 'holders' subdir is created */ |
367 | dev_set_uevent_suppress(pdev, 1); | |
368 | err = device_add(pdev); | |
369 | if (err) | |
370 | goto out_put; | |
371 | ||
372 | err = -ENOMEM; | |
1bdd5ae0 CH |
373 | bdev->bd_holder_dir = kobject_create_and_add("holders", &pdev->kobj); |
374 | if (!bdev->bd_holder_dir) | |
94ea4158 AV |
375 | goto out_del; |
376 | ||
377 | dev_set_uevent_suppress(pdev, 0); | |
378 | if (flags & ADDPART_FLAG_WHOLEDISK) { | |
379 | err = device_create_file(pdev, &dev_attr_whole_disk); | |
380 | if (err) | |
381 | goto out_del; | |
382 | } | |
383 | ||
384 | /* everything is up and running, commence */ | |
a33df75c CH |
385 | err = xa_insert(&disk->part_tbl, partno, bdev, GFP_KERNEL); |
386 | if (err) | |
387 | goto out_del; | |
22ae8ce8 | 388 | bdev_add(bdev, devt); |
94ea4158 AV |
389 | |
390 | /* suppress uevent if the disk suppresses it */ | |
391 | if (!dev_get_uevent_suppress(ddev)) | |
392 | kobject_uevent(&pdev->kobj, KOBJ_ADD); | |
0d02129e | 393 | return bdev; |
94ea4158 | 394 | |
94ea4158 | 395 | out_del: |
1bdd5ae0 | 396 | kobject_put(bdev->bd_holder_dir); |
94ea4158 AV |
397 | device_del(pdev); |
398 | out_put: | |
399 | put_device(pdev); | |
9fbfabfd | 400 | return ERR_PTR(err); |
9d3b8813 CH |
401 | out_put_disk: |
402 | put_disk(disk); | |
94ea4158 AV |
403 | return ERR_PTR(err); |
404 | } | |
405 | ||
fa9156ae CH |
406 | static bool partition_overlaps(struct gendisk *disk, sector_t start, |
407 | sector_t length, int skip_partno) | |
408 | { | |
ad1eaa53 | 409 | struct block_device *part; |
fa9156ae | 410 | bool overlap = false; |
e3069123 CH |
411 | unsigned long idx; |
412 | ||
413 | rcu_read_lock(); | |
414 | xa_for_each_start(&disk->part_tbl, idx, part, 1) { | |
415 | if (part->bd_partno != skip_partno && | |
416 | start < part->bd_start_sect + bdev_nr_sectors(part) && | |
417 | start + length > part->bd_start_sect) { | |
418 | overlap = true; | |
419 | break; | |
420 | } | |
fa9156ae | 421 | } |
e3069123 | 422 | rcu_read_unlock(); |
fa9156ae | 423 | |
fa9156ae CH |
424 | return overlap; |
425 | } | |
426 | ||
7f6be376 CH |
427 | int bdev_add_partition(struct gendisk *disk, int partno, sector_t start, |
428 | sector_t length) | |
fa9156ae | 429 | { |
6d4e80db | 430 | sector_t capacity = get_capacity(disk), end; |
0d02129e | 431 | struct block_device *part; |
b5cfbd35 | 432 | int ret; |
fa9156ae | 433 | |
b5cfbd35 | 434 | mutex_lock(&disk->open_mutex); |
6d4e80db ML |
435 | if (check_add_overflow(start, length, &end)) { |
436 | ret = -EINVAL; | |
437 | goto out; | |
438 | } | |
439 | ||
440 | if (start >= capacity || end > capacity) { | |
441 | ret = -EINVAL; | |
442 | goto out; | |
443 | } | |
444 | ||
50b4aecf | 445 | if (!disk_live(disk)) { |
b5cfbd35 YY |
446 | ret = -ENXIO; |
447 | goto out; | |
448 | } | |
449 | ||
450 | if (partition_overlaps(disk, start, length, -1)) { | |
451 | ret = -EBUSY; | |
452 | goto out; | |
fa9156ae CH |
453 | } |
454 | ||
b5cfbd35 | 455 | part = add_partition(disk, partno, start, length, |
fa9156ae | 456 | ADDPART_FLAG_NONE, NULL); |
b5cfbd35 YY |
457 | ret = PTR_ERR_OR_ZERO(part); |
458 | out: | |
459 | mutex_unlock(&disk->open_mutex); | |
460 | return ret; | |
fa9156ae CH |
461 | } |
462 | ||
926fbb16 | 463 | int bdev_del_partition(struct gendisk *disk, int partno) |
fa9156ae | 464 | { |
0e0ccdec CH |
465 | struct block_device *part = NULL; |
466 | int ret = -ENXIO; | |
fa9156ae | 467 | |
926fbb16 CH |
468 | mutex_lock(&disk->open_mutex); |
469 | part = xa_load(&disk->part_tbl, partno); | |
0e0ccdec CH |
470 | if (!part) |
471 | goto out_unlock; | |
08fc1ab6 | 472 | |
fa9156ae | 473 | ret = -EBUSY; |
9acf381f | 474 | if (atomic_read(&part->bd_openers)) |
fa9156ae CH |
475 | goto out_unlock; |
476 | ||
c30b9787 CB |
477 | /* |
478 | * We verified that @part->bd_openers is zero above and so | |
479 | * @part->bd_holder{_ops} can't be set. And since we hold | |
480 | * @disk->open_mutex the device can't be claimed by anyone. | |
481 | * | |
482 | * So no need to call @part->bd_holder_ops->mark_dead() here. | |
483 | * Just delete the partition and invalidate it. | |
484 | */ | |
485 | ||
486 | remove_inode_hash(part->bd_inode); | |
487 | invalidate_bdev(part); | |
488 | drop_partition(part); | |
fa9156ae CH |
489 | ret = 0; |
490 | out_unlock: | |
926fbb16 | 491 | mutex_unlock(&disk->open_mutex); |
fa9156ae CH |
492 | return ret; |
493 | } | |
494 | ||
3d2e7989 CH |
495 | int bdev_resize_partition(struct gendisk *disk, int partno, sector_t start, |
496 | sector_t length) | |
fa9156ae | 497 | { |
0e0ccdec CH |
498 | struct block_device *part = NULL; |
499 | int ret = -ENXIO; | |
fa9156ae | 500 | |
3d2e7989 CH |
501 | mutex_lock(&disk->open_mutex); |
502 | part = xa_load(&disk->part_tbl, partno); | |
fa9156ae | 503 | if (!part) |
0e0ccdec | 504 | goto out_unlock; |
fa9156ae | 505 | |
fa9156ae | 506 | ret = -EINVAL; |
0d02129e | 507 | if (start != part->bd_start_sect) |
fa9156ae CH |
508 | goto out_unlock; |
509 | ||
510 | ret = -EBUSY; | |
3d2e7989 | 511 | if (partition_overlaps(disk, start, length, partno)) |
fa9156ae CH |
512 | goto out_unlock; |
513 | ||
0d02129e | 514 | bdev_set_nr_sectors(part, length); |
fa9156ae CH |
515 | |
516 | ret = 0; | |
517 | out_unlock: | |
3d2e7989 | 518 | mutex_unlock(&disk->open_mutex); |
fa9156ae CH |
519 | return ret; |
520 | } | |
521 | ||
94ea4158 AV |
522 | static bool disk_unlock_native_capacity(struct gendisk *disk) |
523 | { | |
86416916 CH |
524 | if (!disk->fops->unlock_native_capacity || |
525 | test_and_set_bit(GD_NATIVE_CAPACITY, &disk->state)) { | |
94ea4158 AV |
526 | printk(KERN_CONT "truncated\n"); |
527 | return false; | |
528 | } | |
86416916 CH |
529 | |
530 | printk(KERN_CONT "enabling native capacity\n"); | |
531 | disk->fops->unlock_native_capacity(disk); | |
532 | return true; | |
94ea4158 AV |
533 | } |
534 | ||
0384264e | 535 | static bool blk_add_partition(struct gendisk *disk, |
f902b026 | 536 | struct parsed_partitions *state, int p) |
fe316bf2 | 537 | { |
f902b026 CH |
538 | sector_t size = state->parts[p].size; |
539 | sector_t from = state->parts[p].from; | |
0d02129e | 540 | struct block_device *part; |
f902b026 CH |
541 | |
542 | if (!size) | |
543 | return true; | |
544 | ||
545 | if (from >= get_capacity(disk)) { | |
546 | printk(KERN_WARNING | |
547 | "%s: p%d start %llu is beyond EOD, ", | |
548 | disk->disk_name, p, (unsigned long long) from); | |
549 | if (disk_unlock_native_capacity(disk)) | |
550 | return false; | |
551 | return true; | |
fe316bf2 JN |
552 | } |
553 | ||
f902b026 CH |
554 | if (from + size > get_capacity(disk)) { |
555 | printk(KERN_WARNING | |
556 | "%s: p%d size %llu extends beyond EOD, ", | |
557 | disk->disk_name, p, (unsigned long long) size); | |
fe316bf2 | 558 | |
f902b026 CH |
559 | if (disk_unlock_native_capacity(disk)) |
560 | return false; | |
561 | ||
562 | /* | |
563 | * We can not ignore partitions of broken tables created by for | |
564 | * example camera firmware, but we limit them to the end of the | |
565 | * disk to avoid creating invalid block devices. | |
566 | */ | |
567 | size = get_capacity(disk) - from; | |
568 | } | |
569 | ||
570 | part = add_partition(disk, p, from, size, state->parts[p].flags, | |
571 | &state->parts[p].info); | |
b7205307 | 572 | if (IS_ERR(part) && PTR_ERR(part) != -ENXIO) { |
f902b026 CH |
573 | printk(KERN_ERR " %s: p%d could not be added: %ld\n", |
574 | disk->disk_name, p, -PTR_ERR(part)); | |
575 | return true; | |
576 | } | |
577 | ||
74cc979c CH |
578 | if (IS_BUILTIN(CONFIG_BLK_DEV_MD) && |
579 | (state->parts[p].flags & ADDPART_FLAG_RAID)) | |
0d02129e | 580 | md_autodetect_dev(part->bd_dev); |
74cc979c | 581 | |
f902b026 CH |
582 | return true; |
583 | } | |
584 | ||
0384264e | 585 | static int blk_add_partitions(struct gendisk *disk) |
f902b026 CH |
586 | { |
587 | struct parsed_partitions *state; | |
a33df75c | 588 | int ret = -EAGAIN, p; |
fe316bf2 | 589 | |
1ebe2e5f | 590 | if (disk->flags & GENHD_FL_NO_PART) |
142fe8f4 CH |
591 | return 0; |
592 | ||
748008e1 ML |
593 | if (test_bit(GD_SUPPRESS_PART_SCAN, &disk->state)) |
594 | return 0; | |
595 | ||
0384264e | 596 | state = check_partition(disk); |
f902b026 | 597 | if (!state) |
94ea4158 AV |
598 | return 0; |
599 | if (IS_ERR(state)) { | |
600 | /* | |
f902b026 CH |
601 | * I/O error reading the partition table. If we tried to read |
602 | * beyond EOD, retry after unlocking the native capacity. | |
94ea4158 AV |
603 | */ |
604 | if (PTR_ERR(state) == -ENOSPC) { | |
605 | printk(KERN_WARNING "%s: partition table beyond EOD, ", | |
606 | disk->disk_name); | |
607 | if (disk_unlock_native_capacity(disk)) | |
f902b026 | 608 | return -EAGAIN; |
94ea4158 AV |
609 | } |
610 | return -EIO; | |
611 | } | |
5eac3eb3 | 612 | |
f902b026 | 613 | /* |
b7205307 | 614 | * Partitions are not supported on host managed zoned block devices. |
f902b026 | 615 | */ |
b7205307 CH |
616 | if (disk->queue->limits.zoned == BLK_ZONED_HM) { |
617 | pr_warn("%s: ignoring partition table on host managed zoned block device\n", | |
5eac3eb3 | 618 | disk->disk_name); |
f902b026 CH |
619 | ret = 0; |
620 | goto out_free_state; | |
5eac3eb3 DLM |
621 | } |
622 | ||
94ea4158 | 623 | /* |
f902b026 CH |
624 | * If we read beyond EOD, try unlocking native capacity even if the |
625 | * partition table was successfully read as we could be missing some | |
626 | * partitions. | |
94ea4158 AV |
627 | */ |
628 | if (state->access_beyond_eod) { | |
629 | printk(KERN_WARNING | |
630 | "%s: partition table partially beyond EOD, ", | |
631 | disk->disk_name); | |
632 | if (disk_unlock_native_capacity(disk)) | |
f902b026 | 633 | goto out_free_state; |
94ea4158 AV |
634 | } |
635 | ||
636 | /* tell userspace that the media / partition table may have changed */ | |
637 | kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE); | |
638 | ||
f902b026 | 639 | for (p = 1; p < state->limit; p++) |
0384264e | 640 | if (!blk_add_partition(disk, state, p)) |
f902b026 | 641 | goto out_free_state; |
94ea4158 | 642 | |
f902b026 CH |
643 | ret = 0; |
644 | out_free_state: | |
ac2e5327 | 645 | free_partitions(state); |
f902b026 | 646 | return ret; |
fe316bf2 JN |
647 | } |
648 | ||
0384264e | 649 | int bdev_disk_changed(struct gendisk *disk, bool invalidate) |
630161cf | 650 | { |
00080f7f CH |
651 | struct block_device *part; |
652 | unsigned long idx; | |
630161cf CH |
653 | int ret = 0; |
654 | ||
655 | lockdep_assert_held(&disk->open_mutex); | |
656 | ||
50b4aecf | 657 | if (!disk_live(disk)) |
630161cf CH |
658 | return -ENXIO; |
659 | ||
660 | rescan: | |
661 | if (disk->open_partitions) | |
662 | return -EBUSY; | |
0384264e CH |
663 | sync_blockdev(disk->part0); |
664 | invalidate_bdev(disk->part0); | |
630161cf | 665 | |
51b4cb4f CB |
666 | xa_for_each_start(&disk->part_tbl, idx, part, 1) { |
667 | /* | |
668 | * Remove the block device from the inode hash, so that | |
669 | * it cannot be looked up any more even when openers | |
670 | * still hold references. | |
671 | */ | |
672 | remove_inode_hash(part->bd_inode); | |
673 | ||
674 | /* | |
675 | * If @disk->open_partitions isn't elevated but there's | |
676 | * still an active holder of that block device things | |
677 | * are broken. | |
678 | */ | |
679 | WARN_ON_ONCE(atomic_read(&part->bd_openers)); | |
680 | invalidate_bdev(part); | |
681 | drop_partition(part); | |
682 | } | |
630161cf CH |
683 | clear_bit(GD_NEED_PART_SCAN, &disk->state); |
684 | ||
685 | /* | |
686 | * Historically we only set the capacity to zero for devices that | |
687 | * support partitions (independ of actually having partitions created). | |
688 | * Doing that is rather inconsistent, but changing it broke legacy | |
689 | * udisks polling for legacy ide-cdrom devices. Use the crude check | |
690 | * below to get the sane behavior for most device while not breaking | |
691 | * userspace for this particular setup. | |
692 | */ | |
693 | if (invalidate) { | |
1ebe2e5f | 694 | if (!(disk->flags & GENHD_FL_NO_PART) || |
630161cf CH |
695 | !(disk->flags & GENHD_FL_REMOVABLE)) |
696 | set_capacity(disk, 0); | |
697 | } | |
698 | ||
699 | if (get_capacity(disk)) { | |
0384264e | 700 | ret = blk_add_partitions(disk); |
630161cf CH |
701 | if (ret == -EAGAIN) |
702 | goto rescan; | |
703 | } else if (invalidate) { | |
704 | /* | |
705 | * Tell userspace that the media / partition table may have | |
706 | * changed. | |
707 | */ | |
708 | kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE); | |
709 | } | |
710 | ||
711 | return ret; | |
712 | } | |
713 | /* | |
714 | * Only exported for loop and dasd for historic reasons. Don't use in new | |
715 | * code! | |
716 | */ | |
717 | EXPORT_SYMBOL_GPL(bdev_disk_changed); | |
718 | ||
1a9fba3a | 719 | void *read_part_sector(struct parsed_partitions *state, sector_t n, Sector *p) |
d1a5f2b4 | 720 | { |
a08aa9bc | 721 | struct address_space *mapping = state->disk->part0->bd_inode->i_mapping; |
4fdc08d4 | 722 | struct folio *folio; |
94ea4158 | 723 | |
a08aa9bc | 724 | if (n >= get_capacity(state->disk)) { |
1a9fba3a | 725 | state->access_beyond_eod = true; |
98d8ba69 | 726 | goto out; |
94ea4158 | 727 | } |
1a9fba3a | 728 | |
4fdc08d4 MWO |
729 | folio = read_mapping_folio(mapping, n >> PAGE_SECTORS_SHIFT, NULL); |
730 | if (IS_ERR(folio)) | |
1a9fba3a | 731 | goto out; |
1a9fba3a | 732 | |
4fdc08d4 MWO |
733 | p->v = folio; |
734 | return folio_address(folio) + offset_in_folio(folio, n * SECTOR_SIZE); | |
1a9fba3a | 735 | out: |
94ea4158 AV |
736 | p->v = NULL; |
737 | return NULL; | |
738 | } |