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 | |
94ea4158 | 5 | */ |
94ea4158 AV |
6 | #include <linux/fs.h> |
7 | #include <linux/slab.h> | |
94ea4158 AV |
8 | #include <linux/ctype.h> |
9 | #include <linux/genhd.h> | |
387048bf | 10 | #include <linux/vmalloc.h> |
94ea4158 | 11 | #include <linux/blktrace_api.h> |
74cc979c | 12 | #include <linux/raid/detect.h> |
387048bf CH |
13 | #include "check.h" |
14 | ||
15 | static int (*check_part[])(struct parsed_partitions *) = { | |
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; | |
91 | int nr; | |
92 | ||
93 | state = kzalloc(sizeof(*state), GFP_KERNEL); | |
94 | if (!state) | |
95 | return NULL; | |
96 | ||
97 | nr = disk_max_parts(hd); | |
98 | state->parts = vzalloc(array_size(nr, sizeof(state->parts[0]))); | |
99 | if (!state->parts) { | |
100 | kfree(state); | |
101 | return NULL; | |
102 | } | |
103 | ||
104 | state->limit = nr; | |
105 | ||
106 | return state; | |
107 | } | |
108 | ||
109 | static void free_partitions(struct parsed_partitions *state) | |
110 | { | |
111 | vfree(state->parts); | |
112 | kfree(state); | |
113 | } | |
114 | ||
115 | static struct parsed_partitions *check_partition(struct gendisk *hd, | |
116 | struct block_device *bdev) | |
117 | { | |
118 | struct parsed_partitions *state; | |
119 | int i, res, err; | |
120 | ||
121 | state = allocate_partitions(hd); | |
122 | if (!state) | |
123 | return NULL; | |
124 | state->pp_buf = (char *)__get_free_page(GFP_KERNEL); | |
125 | if (!state->pp_buf) { | |
126 | free_partitions(state); | |
127 | return NULL; | |
128 | } | |
129 | state->pp_buf[0] = '\0'; | |
130 | ||
131 | state->bdev = bdev; | |
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"); | |
136 | ||
137 | i = res = err = 0; | |
138 | while (!res && check_part[i]) { | |
139 | memset(state->parts, 0, state->limit * sizeof(state->parts[0])); | |
140 | res = check_part[i++](state); | |
141 | if (res < 0) { | |
142 | /* | |
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. | |
145 | */ | |
146 | err = res; | |
147 | res = 0; | |
148 | } | |
149 | ||
150 | } | |
151 | if (res > 0) { | |
152 | printk(KERN_INFO "%s", state->pp_buf); | |
153 | ||
154 | free_page((unsigned long)state->pp_buf); | |
155 | return state; | |
156 | } | |
157 | if (state->access_beyond_eod) | |
158 | err = -ENOSPC; | |
159 | /* | |
160 | * The partition is unrecognized. So report I/O errors if there were any | |
161 | */ | |
162 | if (err) | |
163 | res = err; | |
164 | if (res) { | |
165 | strlcat(state->pp_buf, | |
166 | " unable to read partition table\n", PAGE_SIZE); | |
167 | printk(KERN_INFO "%s", state->pp_buf); | |
168 | } | |
94ea4158 | 169 | |
387048bf CH |
170 | free_page((unsigned long)state->pp_buf); |
171 | free_partitions(state); | |
172 | return ERR_PTR(res); | |
173 | } | |
94ea4158 | 174 | |
94ea4158 AV |
175 | static ssize_t part_partition_show(struct device *dev, |
176 | struct device_attribute *attr, char *buf) | |
177 | { | |
178 | struct hd_struct *p = dev_to_part(dev); | |
179 | ||
180 | return sprintf(buf, "%d\n", p->partno); | |
181 | } | |
182 | ||
183 | static ssize_t part_start_show(struct device *dev, | |
184 | struct device_attribute *attr, char *buf) | |
185 | { | |
186 | struct hd_struct *p = dev_to_part(dev); | |
187 | ||
188 | return sprintf(buf, "%llu\n",(unsigned long long)p->start_sect); | |
189 | } | |
190 | ||
94ea4158 AV |
191 | static ssize_t part_ro_show(struct device *dev, |
192 | struct device_attribute *attr, char *buf) | |
193 | { | |
194 | struct hd_struct *p = dev_to_part(dev); | |
195 | return sprintf(buf, "%d\n", p->policy ? 1 : 0); | |
196 | } | |
197 | ||
198 | static ssize_t part_alignment_offset_show(struct device *dev, | |
199 | struct device_attribute *attr, char *buf) | |
200 | { | |
201 | struct hd_struct *p = dev_to_part(dev); | |
7b8917f5 CH |
202 | |
203 | return sprintf(buf, "%u\n", | |
204 | queue_limit_alignment_offset(&part_to_disk(p)->queue->limits, | |
205 | p->start_sect)); | |
94ea4158 AV |
206 | } |
207 | ||
208 | static ssize_t part_discard_alignment_show(struct device *dev, | |
209 | struct device_attribute *attr, char *buf) | |
210 | { | |
211 | struct hd_struct *p = dev_to_part(dev); | |
7cf34d97 CH |
212 | |
213 | return sprintf(buf, "%u\n", | |
214 | queue_limit_discard_alignment(&part_to_disk(p)->queue->limits, | |
215 | p->start_sect)); | |
94ea4158 AV |
216 | } |
217 | ||
5657a819 JP |
218 | static DEVICE_ATTR(partition, 0444, part_partition_show, NULL); |
219 | static DEVICE_ATTR(start, 0444, part_start_show, NULL); | |
220 | static DEVICE_ATTR(size, 0444, part_size_show, NULL); | |
221 | static DEVICE_ATTR(ro, 0444, part_ro_show, NULL); | |
222 | static DEVICE_ATTR(alignment_offset, 0444, part_alignment_offset_show, NULL); | |
223 | static DEVICE_ATTR(discard_alignment, 0444, part_discard_alignment_show, NULL); | |
224 | static DEVICE_ATTR(stat, 0444, part_stat_show, NULL); | |
225 | static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL); | |
94ea4158 AV |
226 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
227 | static struct device_attribute dev_attr_fail = | |
5657a819 | 228 | __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store); |
94ea4158 AV |
229 | #endif |
230 | ||
231 | static struct attribute *part_attrs[] = { | |
232 | &dev_attr_partition.attr, | |
233 | &dev_attr_start.attr, | |
234 | &dev_attr_size.attr, | |
235 | &dev_attr_ro.attr, | |
236 | &dev_attr_alignment_offset.attr, | |
237 | &dev_attr_discard_alignment.attr, | |
238 | &dev_attr_stat.attr, | |
239 | &dev_attr_inflight.attr, | |
240 | #ifdef CONFIG_FAIL_MAKE_REQUEST | |
241 | &dev_attr_fail.attr, | |
242 | #endif | |
243 | NULL | |
244 | }; | |
245 | ||
246 | static struct attribute_group part_attr_group = { | |
247 | .attrs = part_attrs, | |
248 | }; | |
249 | ||
250 | static const struct attribute_group *part_attr_groups[] = { | |
251 | &part_attr_group, | |
252 | #ifdef CONFIG_BLK_DEV_IO_TRACE | |
253 | &blk_trace_attr_group, | |
254 | #endif | |
255 | NULL | |
256 | }; | |
257 | ||
258 | static void part_release(struct device *dev) | |
259 | { | |
260 | struct hd_struct *p = dev_to_part(dev); | |
2da78092 | 261 | blk_free_devt(dev->devt); |
b54e5ed8 | 262 | hd_free_part(p); |
94ea4158 AV |
263 | kfree(p); |
264 | } | |
265 | ||
0d9c51a6 SM |
266 | static int part_uevent(struct device *dev, struct kobj_uevent_env *env) |
267 | { | |
268 | struct hd_struct *part = dev_to_part(dev); | |
269 | ||
270 | add_uevent_var(env, "PARTN=%u", part->partno); | |
271 | if (part->info && part->info->volname[0]) | |
272 | add_uevent_var(env, "PARTNAME=%s", part->info->volname); | |
273 | return 0; | |
274 | } | |
275 | ||
94ea4158 AV |
276 | struct device_type part_type = { |
277 | .name = "partition", | |
278 | .groups = part_attr_groups, | |
279 | .release = part_release, | |
0d9c51a6 | 280 | .uevent = part_uevent, |
94ea4158 AV |
281 | }; |
282 | ||
8da2892e | 283 | static void hd_struct_free_work(struct work_struct *work) |
94ea4158 | 284 | { |
8da2892e CH |
285 | struct hd_struct *part = |
286 | container_of(to_rcu_work(work), struct hd_struct, rcu_work); | |
cafe01ef ML |
287 | struct gendisk *disk = part_to_disk(part); |
288 | ||
289 | /* | |
290 | * Release the disk reference acquired in delete_partition here. | |
291 | * We can't release it in hd_struct_free because the final put_device | |
292 | * needs process context and thus can't be run directly from a | |
293 | * percpu_ref ->release handler. | |
294 | */ | |
295 | put_device(disk_to_dev(disk)); | |
94ea4158 AV |
296 | |
297 | part->start_sect = 0; | |
298 | part->nr_sects = 0; | |
299 | part_stat_set_all(part, 0); | |
300 | put_device(part_to_dev(part)); | |
301 | } | |
302 | ||
8da2892e | 303 | static void hd_struct_free(struct percpu_ref *ref) |
94ea4158 | 304 | { |
6c71013e | 305 | struct hd_struct *part = container_of(ref, struct hd_struct, ref); |
b7d6c303 ML |
306 | struct gendisk *disk = part_to_disk(part); |
307 | struct disk_part_tbl *ptbl = | |
308 | rcu_dereference_protected(disk->part_tbl, 1); | |
309 | ||
310 | rcu_assign_pointer(ptbl->last_lookup, NULL); | |
8da2892e CH |
311 | |
312 | INIT_RCU_WORK(&part->rcu_work, hd_struct_free_work); | |
94a2c3a3 | 313 | queue_rcu_work(system_wq, &part->rcu_work); |
94ea4158 AV |
314 | } |
315 | ||
8da2892e CH |
316 | int hd_ref_init(struct hd_struct *part) |
317 | { | |
318 | if (percpu_ref_init(&part->ref, hd_struct_free, 0, GFP_KERNEL)) | |
319 | return -ENOMEM; | |
320 | return 0; | |
321 | } | |
322 | ||
6d2cf6f2 BVA |
323 | /* |
324 | * Must be called either with bd_mutex held, before a disk can be opened or | |
325 | * after all disk users are gone. | |
326 | */ | |
8328eb28 | 327 | void delete_partition(struct hd_struct *part) |
94ea4158 | 328 | { |
8328eb28 | 329 | struct gendisk *disk = part_to_disk(part); |
6d2cf6f2 BVA |
330 | struct disk_part_tbl *ptbl = |
331 | rcu_dereference_protected(disk->part_tbl, 1); | |
94ea4158 | 332 | |
b7d6c303 ML |
333 | /* |
334 | * ->part_tbl is referenced in this part's release handler, so | |
335 | * we have to hold the disk device | |
336 | */ | |
8328eb28 | 337 | get_device(disk_to_dev(disk)); |
cddae808 | 338 | rcu_assign_pointer(ptbl->part[part->partno], NULL); |
94ea4158 AV |
339 | kobject_put(part->holder_dir); |
340 | device_del(part_to_dev(part)); | |
341 | ||
6fcc44d1 | 342 | /* |
22ae8ce8 CH |
343 | * Remove the block device from the inode hash, so that it cannot be |
344 | * looked up any more even when openers still hold references. | |
6fcc44d1 | 345 | */ |
22ae8ce8 CH |
346 | remove_inode_hash(part->bdev->bd_inode); |
347 | ||
4377b48d | 348 | percpu_ref_kill(&part->ref); |
94ea4158 AV |
349 | } |
350 | ||
351 | static ssize_t whole_disk_show(struct device *dev, | |
352 | struct device_attribute *attr, char *buf) | |
353 | { | |
354 | return 0; | |
355 | } | |
5657a819 | 356 | static DEVICE_ATTR(whole_disk, 0444, whole_disk_show, NULL); |
94ea4158 | 357 | |
6d2cf6f2 BVA |
358 | /* |
359 | * Must be called either with bd_mutex held, before a disk can be opened or | |
360 | * after all disk users are gone. | |
361 | */ | |
fa9156ae | 362 | static struct hd_struct *add_partition(struct gendisk *disk, int partno, |
94ea4158 AV |
363 | sector_t start, sector_t len, int flags, |
364 | struct partition_meta_info *info) | |
365 | { | |
366 | struct hd_struct *p; | |
367 | dev_t devt = MKDEV(0, 0); | |
368 | struct device *ddev = disk_to_dev(disk); | |
369 | struct device *pdev; | |
22ae8ce8 | 370 | struct block_device *bdev; |
94ea4158 AV |
371 | struct disk_part_tbl *ptbl; |
372 | const char *dname; | |
373 | int err; | |
374 | ||
b7205307 CH |
375 | /* |
376 | * Partitions are not supported on zoned block devices that are used as | |
377 | * such. | |
378 | */ | |
379 | switch (disk->queue->limits.zoned) { | |
380 | case BLK_ZONED_HM: | |
381 | pr_warn("%s: partitions not supported on host managed zoned block device\n", | |
382 | disk->disk_name); | |
383 | return ERR_PTR(-ENXIO); | |
384 | case BLK_ZONED_HA: | |
385 | pr_info("%s: disabling host aware zoned block device support due to partitions\n", | |
386 | disk->disk_name); | |
387 | disk->queue->limits.zoned = BLK_ZONED_NONE; | |
388 | break; | |
389 | case BLK_ZONED_NONE: | |
390 | break; | |
391 | } | |
392 | ||
94ea4158 AV |
393 | err = disk_expand_part_tbl(disk, partno); |
394 | if (err) | |
395 | return ERR_PTR(err); | |
6d2cf6f2 | 396 | ptbl = rcu_dereference_protected(disk->part_tbl, 1); |
94ea4158 AV |
397 | |
398 | if (ptbl->part[partno]) | |
399 | return ERR_PTR(-EBUSY); | |
400 | ||
401 | p = kzalloc(sizeof(*p), GFP_KERNEL); | |
402 | if (!p) | |
403 | return ERR_PTR(-EBUSY); | |
404 | ||
22ae8ce8 | 405 | err = -ENOMEM; |
58d4f14f | 406 | p->dkstats = alloc_percpu(struct disk_stats); |
22ae8ce8 | 407 | if (!p->dkstats) |
94ea4158 | 408 | goto out_free; |
22ae8ce8 CH |
409 | |
410 | bdev = bdev_alloc(disk, partno); | |
411 | if (!bdev) | |
412 | goto out_free_stats; | |
413 | p->bdev = bdev; | |
c83f6bf9 | 414 | |
07c4e1e8 | 415 | hd_sects_seq_init(p); |
94ea4158 AV |
416 | pdev = part_to_dev(p); |
417 | ||
418 | p->start_sect = start; | |
94ea4158 AV |
419 | p->nr_sects = len; |
420 | p->partno = partno; | |
421 | p->policy = get_disk_ro(disk); | |
422 | ||
423 | if (info) { | |
f17c21c1 CH |
424 | struct partition_meta_info *pinfo; |
425 | ||
426 | pinfo = kzalloc_node(sizeof(*pinfo), GFP_KERNEL, disk->node_id); | |
22ae8ce8 CH |
427 | if (!pinfo) |
428 | goto out_bdput; | |
94ea4158 AV |
429 | memcpy(pinfo, info, sizeof(*info)); |
430 | p->info = pinfo; | |
431 | } | |
432 | ||
433 | dname = dev_name(ddev); | |
434 | if (isdigit(dname[strlen(dname) - 1])) | |
435 | dev_set_name(pdev, "%sp%d", dname, partno); | |
436 | else | |
437 | dev_set_name(pdev, "%s%d", dname, partno); | |
438 | ||
439 | device_initialize(pdev); | |
440 | pdev->class = &block_class; | |
441 | pdev->type = &part_type; | |
442 | pdev->parent = ddev; | |
443 | ||
444 | err = blk_alloc_devt(p, &devt); | |
445 | if (err) | |
446 | goto out_free_info; | |
447 | pdev->devt = devt; | |
448 | ||
449 | /* delay uevent until 'holders' subdir is created */ | |
450 | dev_set_uevent_suppress(pdev, 1); | |
451 | err = device_add(pdev); | |
452 | if (err) | |
453 | goto out_put; | |
454 | ||
455 | err = -ENOMEM; | |
456 | p->holder_dir = kobject_create_and_add("holders", &pdev->kobj); | |
457 | if (!p->holder_dir) | |
458 | goto out_del; | |
459 | ||
460 | dev_set_uevent_suppress(pdev, 0); | |
461 | if (flags & ADDPART_FLAG_WHOLEDISK) { | |
462 | err = device_create_file(pdev, &dev_attr_whole_disk); | |
463 | if (err) | |
464 | goto out_del; | |
465 | } | |
466 | ||
b30a337c ML |
467 | err = hd_ref_init(p); |
468 | if (err) { | |
469 | if (flags & ADDPART_FLAG_WHOLEDISK) | |
470 | goto out_remove_file; | |
471 | goto out_del; | |
472 | } | |
473 | ||
94ea4158 | 474 | /* everything is up and running, commence */ |
22ae8ce8 | 475 | bdev_add(bdev, devt); |
94ea4158 AV |
476 | rcu_assign_pointer(ptbl->part[partno], p); |
477 | ||
478 | /* suppress uevent if the disk suppresses it */ | |
479 | if (!dev_get_uevent_suppress(ddev)) | |
480 | kobject_uevent(&pdev->kobj, KOBJ_ADD); | |
b30a337c | 481 | return p; |
94ea4158 AV |
482 | |
483 | out_free_info: | |
f17c21c1 | 484 | kfree(p->info); |
22ae8ce8 CH |
485 | out_bdput: |
486 | bdput(bdev); | |
94ea4158 | 487 | out_free_stats: |
58d4f14f | 488 | free_percpu(p->dkstats); |
94ea4158 AV |
489 | out_free: |
490 | kfree(p); | |
491 | return ERR_PTR(err); | |
b30a337c ML |
492 | out_remove_file: |
493 | device_remove_file(pdev, &dev_attr_whole_disk); | |
94ea4158 AV |
494 | out_del: |
495 | kobject_put(p->holder_dir); | |
496 | device_del(pdev); | |
497 | out_put: | |
498 | put_device(pdev); | |
94ea4158 AV |
499 | return ERR_PTR(err); |
500 | } | |
501 | ||
fa9156ae CH |
502 | static bool partition_overlaps(struct gendisk *disk, sector_t start, |
503 | sector_t length, int skip_partno) | |
504 | { | |
505 | struct disk_part_iter piter; | |
506 | struct hd_struct *part; | |
507 | bool overlap = false; | |
508 | ||
509 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY); | |
510 | while ((part = disk_part_iter_next(&piter))) { | |
511 | if (part->partno == skip_partno || | |
512 | start >= part->start_sect + part->nr_sects || | |
513 | start + length <= part->start_sect) | |
514 | continue; | |
515 | overlap = true; | |
516 | break; | |
517 | } | |
518 | ||
519 | disk_part_iter_exit(&piter); | |
520 | return overlap; | |
521 | } | |
522 | ||
523 | int bdev_add_partition(struct block_device *bdev, int partno, | |
524 | sector_t start, sector_t length) | |
525 | { | |
526 | struct hd_struct *part; | |
527 | ||
528 | mutex_lock(&bdev->bd_mutex); | |
529 | if (partition_overlaps(bdev->bd_disk, start, length, -1)) { | |
530 | mutex_unlock(&bdev->bd_mutex); | |
531 | return -EBUSY; | |
532 | } | |
533 | ||
534 | part = add_partition(bdev->bd_disk, partno, start, length, | |
535 | ADDPART_FLAG_NONE, NULL); | |
536 | mutex_unlock(&bdev->bd_mutex); | |
537 | return PTR_ERR_OR_ZERO(part); | |
538 | } | |
539 | ||
540 | int bdev_del_partition(struct block_device *bdev, int partno) | |
541 | { | |
542 | struct block_device *bdevp; | |
08fc1ab6 CH |
543 | struct hd_struct *part = NULL; |
544 | int ret; | |
fa9156ae | 545 | |
08fc1ab6 | 546 | bdevp = bdget_disk(bdev->bd_disk, partno); |
fa9156ae | 547 | if (!bdevp) |
88ce2a53 | 548 | return -ENXIO; |
fa9156ae CH |
549 | |
550 | mutex_lock(&bdevp->bd_mutex); | |
08fc1ab6 CH |
551 | mutex_lock_nested(&bdev->bd_mutex, 1); |
552 | ||
553 | ret = -ENXIO; | |
554 | part = disk_get_part(bdev->bd_disk, partno); | |
555 | if (!part) | |
556 | goto out_unlock; | |
fa9156ae CH |
557 | |
558 | ret = -EBUSY; | |
559 | if (bdevp->bd_openers) | |
560 | goto out_unlock; | |
561 | ||
d5f3178e | 562 | sync_blockdev(bdevp); |
fa9156ae CH |
563 | invalidate_bdev(bdevp); |
564 | ||
8328eb28 | 565 | delete_partition(part); |
fa9156ae CH |
566 | ret = 0; |
567 | out_unlock: | |
08fc1ab6 | 568 | mutex_unlock(&bdev->bd_mutex); |
fa9156ae CH |
569 | mutex_unlock(&bdevp->bd_mutex); |
570 | bdput(bdevp); | |
08fc1ab6 CH |
571 | if (part) |
572 | disk_put_part(part); | |
fa9156ae CH |
573 | return ret; |
574 | } | |
575 | ||
576 | int bdev_resize_partition(struct block_device *bdev, int partno, | |
577 | sector_t start, sector_t length) | |
578 | { | |
579 | struct block_device *bdevp; | |
580 | struct hd_struct *part; | |
581 | int ret = 0; | |
582 | ||
583 | part = disk_get_part(bdev->bd_disk, partno); | |
584 | if (!part) | |
585 | return -ENXIO; | |
586 | ||
587 | ret = -ENOMEM; | |
10ed1666 | 588 | bdevp = bdget_part(part); |
fa9156ae CH |
589 | if (!bdevp) |
590 | goto out_put_part; | |
591 | ||
592 | mutex_lock(&bdevp->bd_mutex); | |
593 | mutex_lock_nested(&bdev->bd_mutex, 1); | |
594 | ||
595 | ret = -EINVAL; | |
596 | if (start != part->start_sect) | |
597 | goto out_unlock; | |
598 | ||
599 | ret = -EBUSY; | |
600 | if (partition_overlaps(bdev->bd_disk, start, length, partno)) | |
601 | goto out_unlock; | |
602 | ||
c2b4bb8c CH |
603 | part_nr_sects_write(part, length); |
604 | bd_set_nr_sectors(bdevp, length); | |
fa9156ae CH |
605 | |
606 | ret = 0; | |
607 | out_unlock: | |
608 | mutex_unlock(&bdevp->bd_mutex); | |
609 | mutex_unlock(&bdev->bd_mutex); | |
610 | bdput(bdevp); | |
611 | out_put_part: | |
612 | disk_put_part(part); | |
613 | return ret; | |
614 | } | |
615 | ||
94ea4158 AV |
616 | static bool disk_unlock_native_capacity(struct gendisk *disk) |
617 | { | |
618 | const struct block_device_operations *bdops = disk->fops; | |
619 | ||
620 | if (bdops->unlock_native_capacity && | |
621 | !(disk->flags & GENHD_FL_NATIVE_CAPACITY)) { | |
622 | printk(KERN_CONT "enabling native capacity\n"); | |
623 | bdops->unlock_native_capacity(disk); | |
624 | disk->flags |= GENHD_FL_NATIVE_CAPACITY; | |
625 | return true; | |
626 | } else { | |
627 | printk(KERN_CONT "truncated\n"); | |
628 | return false; | |
629 | } | |
630 | } | |
631 | ||
d46430bf | 632 | int blk_drop_partitions(struct block_device *bdev) |
94ea4158 | 633 | { |
94ea4158 AV |
634 | struct disk_part_iter piter; |
635 | struct hd_struct *part; | |
94ea4158 | 636 | |
10c70d95 | 637 | if (bdev->bd_part_count) |
94ea4158 | 638 | return -EBUSY; |
e669c1da CH |
639 | |
640 | sync_blockdev(bdev); | |
641 | invalidate_bdev(bdev); | |
94ea4158 | 642 | |
d46430bf | 643 | disk_part_iter_init(&piter, bdev->bd_disk, DISK_PITER_INCL_EMPTY); |
94ea4158 | 644 | while ((part = disk_part_iter_next(&piter))) |
8328eb28 | 645 | delete_partition(part); |
94ea4158 AV |
646 | disk_part_iter_exit(&piter); |
647 | ||
fe316bf2 JN |
648 | return 0; |
649 | } | |
21be6cdc CH |
650 | #ifdef CONFIG_S390 |
651 | /* for historic reasons in the DASD driver */ | |
652 | EXPORT_SYMBOL_GPL(blk_drop_partitions); | |
653 | #endif | |
fe316bf2 | 654 | |
f902b026 CH |
655 | static bool blk_add_partition(struct gendisk *disk, struct block_device *bdev, |
656 | struct parsed_partitions *state, int p) | |
fe316bf2 | 657 | { |
f902b026 CH |
658 | sector_t size = state->parts[p].size; |
659 | sector_t from = state->parts[p].from; | |
fe316bf2 | 660 | struct hd_struct *part; |
f902b026 CH |
661 | |
662 | if (!size) | |
663 | return true; | |
664 | ||
665 | if (from >= get_capacity(disk)) { | |
666 | printk(KERN_WARNING | |
667 | "%s: p%d start %llu is beyond EOD, ", | |
668 | disk->disk_name, p, (unsigned long long) from); | |
669 | if (disk_unlock_native_capacity(disk)) | |
670 | return false; | |
671 | return true; | |
fe316bf2 JN |
672 | } |
673 | ||
f902b026 CH |
674 | if (from + size > get_capacity(disk)) { |
675 | printk(KERN_WARNING | |
676 | "%s: p%d size %llu extends beyond EOD, ", | |
677 | disk->disk_name, p, (unsigned long long) size); | |
fe316bf2 | 678 | |
f902b026 CH |
679 | if (disk_unlock_native_capacity(disk)) |
680 | return false; | |
681 | ||
682 | /* | |
683 | * We can not ignore partitions of broken tables created by for | |
684 | * example camera firmware, but we limit them to the end of the | |
685 | * disk to avoid creating invalid block devices. | |
686 | */ | |
687 | size = get_capacity(disk) - from; | |
688 | } | |
689 | ||
690 | part = add_partition(disk, p, from, size, state->parts[p].flags, | |
691 | &state->parts[p].info); | |
b7205307 | 692 | if (IS_ERR(part) && PTR_ERR(part) != -ENXIO) { |
f902b026 CH |
693 | printk(KERN_ERR " %s: p%d could not be added: %ld\n", |
694 | disk->disk_name, p, -PTR_ERR(part)); | |
695 | return true; | |
696 | } | |
697 | ||
74cc979c CH |
698 | if (IS_BUILTIN(CONFIG_BLK_DEV_MD) && |
699 | (state->parts[p].flags & ADDPART_FLAG_RAID)) | |
f902b026 | 700 | md_autodetect_dev(part_to_dev(part)->devt); |
74cc979c | 701 | |
f902b026 CH |
702 | return true; |
703 | } | |
704 | ||
a1548b67 | 705 | int blk_add_partitions(struct gendisk *disk, struct block_device *bdev) |
f902b026 CH |
706 | { |
707 | struct parsed_partitions *state; | |
708 | int ret = -EAGAIN, p, highest; | |
fe316bf2 | 709 | |
142fe8f4 CH |
710 | if (!disk_part_scan_enabled(disk)) |
711 | return 0; | |
712 | ||
f902b026 CH |
713 | state = check_partition(disk, bdev); |
714 | if (!state) | |
94ea4158 AV |
715 | return 0; |
716 | if (IS_ERR(state)) { | |
717 | /* | |
f902b026 CH |
718 | * I/O error reading the partition table. If we tried to read |
719 | * beyond EOD, retry after unlocking the native capacity. | |
94ea4158 AV |
720 | */ |
721 | if (PTR_ERR(state) == -ENOSPC) { | |
722 | printk(KERN_WARNING "%s: partition table beyond EOD, ", | |
723 | disk->disk_name); | |
724 | if (disk_unlock_native_capacity(disk)) | |
f902b026 | 725 | return -EAGAIN; |
94ea4158 AV |
726 | } |
727 | return -EIO; | |
728 | } | |
5eac3eb3 | 729 | |
f902b026 | 730 | /* |
b7205307 | 731 | * Partitions are not supported on host managed zoned block devices. |
f902b026 | 732 | */ |
b7205307 CH |
733 | if (disk->queue->limits.zoned == BLK_ZONED_HM) { |
734 | pr_warn("%s: ignoring partition table on host managed zoned block device\n", | |
5eac3eb3 | 735 | disk->disk_name); |
f902b026 CH |
736 | ret = 0; |
737 | goto out_free_state; | |
5eac3eb3 DLM |
738 | } |
739 | ||
94ea4158 | 740 | /* |
f902b026 CH |
741 | * If we read beyond EOD, try unlocking native capacity even if the |
742 | * partition table was successfully read as we could be missing some | |
743 | * partitions. | |
94ea4158 AV |
744 | */ |
745 | if (state->access_beyond_eod) { | |
746 | printk(KERN_WARNING | |
747 | "%s: partition table partially beyond EOD, ", | |
748 | disk->disk_name); | |
749 | if (disk_unlock_native_capacity(disk)) | |
f902b026 | 750 | goto out_free_state; |
94ea4158 AV |
751 | } |
752 | ||
753 | /* tell userspace that the media / partition table may have changed */ | |
754 | kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE); | |
755 | ||
f902b026 CH |
756 | /* |
757 | * Detect the highest partition number and preallocate disk->part_tbl. | |
758 | * This is an optimization and not strictly necessary. | |
94ea4158 AV |
759 | */ |
760 | for (p = 1, highest = 0; p < state->limit; p++) | |
761 | if (state->parts[p].size) | |
762 | highest = p; | |
94ea4158 AV |
763 | disk_expand_part_tbl(disk, highest); |
764 | ||
f902b026 CH |
765 | for (p = 1; p < state->limit; p++) |
766 | if (!blk_add_partition(disk, bdev, state, p)) | |
767 | goto out_free_state; | |
94ea4158 | 768 | |
f902b026 CH |
769 | ret = 0; |
770 | out_free_state: | |
ac2e5327 | 771 | free_partitions(state); |
f902b026 | 772 | return ret; |
fe316bf2 JN |
773 | } |
774 | ||
1a9fba3a | 775 | void *read_part_sector(struct parsed_partitions *state, sector_t n, Sector *p) |
d1a5f2b4 | 776 | { |
1a9fba3a | 777 | struct address_space *mapping = state->bdev->bd_inode->i_mapping; |
94ea4158 AV |
778 | struct page *page; |
779 | ||
1a9fba3a CH |
780 | if (n >= get_capacity(state->bdev->bd_disk)) { |
781 | state->access_beyond_eod = true; | |
782 | return NULL; | |
94ea4158 | 783 | } |
1a9fba3a CH |
784 | |
785 | page = read_mapping_page(mapping, | |
786 | (pgoff_t)(n >> (PAGE_SHIFT - 9)), NULL); | |
787 | if (IS_ERR(page)) | |
788 | goto out; | |
789 | if (PageError(page)) | |
790 | goto out_put_page; | |
791 | ||
792 | p->v = page; | |
793 | return (unsigned char *)page_address(page) + | |
794 | ((n & ((1 << (PAGE_SHIFT - 9)) - 1)) << SECTOR_SHIFT); | |
795 | out_put_page: | |
796 | put_page(page); | |
797 | out: | |
94ea4158 AV |
798 | p->v = NULL; |
799 | return NULL; | |
800 | } |