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 | ||
a782483c CH |
88 | static void bdev_set_nr_sectors(struct block_device *bdev, sector_t sectors) |
89 | { | |
90 | spin_lock(&bdev->bd_size_lock); | |
91 | i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT); | |
92 | spin_unlock(&bdev->bd_size_lock); | |
93 | } | |
94 | ||
387048bf CH |
95 | static struct parsed_partitions *allocate_partitions(struct gendisk *hd) |
96 | { | |
97 | struct parsed_partitions *state; | |
98 | int nr; | |
99 | ||
100 | state = kzalloc(sizeof(*state), GFP_KERNEL); | |
101 | if (!state) | |
102 | return NULL; | |
103 | ||
104 | nr = disk_max_parts(hd); | |
105 | state->parts = vzalloc(array_size(nr, sizeof(state->parts[0]))); | |
106 | if (!state->parts) { | |
107 | kfree(state); | |
108 | return NULL; | |
109 | } | |
110 | ||
111 | state->limit = nr; | |
112 | ||
113 | return state; | |
114 | } | |
115 | ||
116 | static void free_partitions(struct parsed_partitions *state) | |
117 | { | |
118 | vfree(state->parts); | |
119 | kfree(state); | |
120 | } | |
121 | ||
122 | static struct parsed_partitions *check_partition(struct gendisk *hd, | |
123 | struct block_device *bdev) | |
124 | { | |
125 | struct parsed_partitions *state; | |
126 | int i, res, err; | |
127 | ||
128 | state = allocate_partitions(hd); | |
129 | if (!state) | |
130 | return NULL; | |
131 | state->pp_buf = (char *)__get_free_page(GFP_KERNEL); | |
132 | if (!state->pp_buf) { | |
133 | free_partitions(state); | |
134 | return NULL; | |
135 | } | |
136 | state->pp_buf[0] = '\0'; | |
137 | ||
138 | state->bdev = bdev; | |
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"); | |
143 | ||
144 | i = res = err = 0; | |
145 | while (!res && check_part[i]) { | |
146 | memset(state->parts, 0, state->limit * sizeof(state->parts[0])); | |
147 | res = check_part[i++](state); | |
148 | if (res < 0) { | |
149 | /* | |
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. | |
152 | */ | |
153 | err = res; | |
154 | res = 0; | |
155 | } | |
156 | ||
157 | } | |
158 | if (res > 0) { | |
159 | printk(KERN_INFO "%s", state->pp_buf); | |
160 | ||
161 | free_page((unsigned long)state->pp_buf); | |
162 | return state; | |
163 | } | |
164 | if (state->access_beyond_eod) | |
165 | err = -ENOSPC; | |
166 | /* | |
167 | * The partition is unrecognized. So report I/O errors if there were any | |
168 | */ | |
169 | if (err) | |
170 | res = err; | |
171 | if (res) { | |
172 | strlcat(state->pp_buf, | |
173 | " unable to read partition table\n", PAGE_SIZE); | |
174 | printk(KERN_INFO "%s", state->pp_buf); | |
175 | } | |
94ea4158 | 176 | |
387048bf CH |
177 | free_page((unsigned long)state->pp_buf); |
178 | free_partitions(state); | |
179 | return ERR_PTR(res); | |
180 | } | |
94ea4158 | 181 | |
94ea4158 AV |
182 | static ssize_t part_partition_show(struct device *dev, |
183 | struct device_attribute *attr, char *buf) | |
184 | { | |
185 | struct hd_struct *p = dev_to_part(dev); | |
186 | ||
187 | return sprintf(buf, "%d\n", p->partno); | |
188 | } | |
189 | ||
190 | static ssize_t part_start_show(struct device *dev, | |
191 | struct device_attribute *attr, char *buf) | |
192 | { | |
193 | struct hd_struct *p = dev_to_part(dev); | |
194 | ||
195 | return sprintf(buf, "%llu\n",(unsigned long long)p->start_sect); | |
196 | } | |
197 | ||
94ea4158 AV |
198 | static ssize_t part_ro_show(struct device *dev, |
199 | struct device_attribute *attr, char *buf) | |
200 | { | |
201 | struct hd_struct *p = dev_to_part(dev); | |
202 | return sprintf(buf, "%d\n", p->policy ? 1 : 0); | |
203 | } | |
204 | ||
205 | static ssize_t part_alignment_offset_show(struct device *dev, | |
206 | struct device_attribute *attr, char *buf) | |
207 | { | |
208 | struct hd_struct *p = dev_to_part(dev); | |
7b8917f5 CH |
209 | |
210 | return sprintf(buf, "%u\n", | |
211 | queue_limit_alignment_offset(&part_to_disk(p)->queue->limits, | |
212 | p->start_sect)); | |
94ea4158 AV |
213 | } |
214 | ||
215 | static ssize_t part_discard_alignment_show(struct device *dev, | |
216 | struct device_attribute *attr, char *buf) | |
217 | { | |
218 | struct hd_struct *p = dev_to_part(dev); | |
7cf34d97 CH |
219 | |
220 | return sprintf(buf, "%u\n", | |
221 | queue_limit_discard_alignment(&part_to_disk(p)->queue->limits, | |
222 | p->start_sect)); | |
94ea4158 AV |
223 | } |
224 | ||
5657a819 JP |
225 | static DEVICE_ATTR(partition, 0444, part_partition_show, NULL); |
226 | static DEVICE_ATTR(start, 0444, part_start_show, NULL); | |
227 | static DEVICE_ATTR(size, 0444, part_size_show, NULL); | |
228 | static DEVICE_ATTR(ro, 0444, part_ro_show, NULL); | |
229 | static DEVICE_ATTR(alignment_offset, 0444, part_alignment_offset_show, NULL); | |
230 | static DEVICE_ATTR(discard_alignment, 0444, part_discard_alignment_show, NULL); | |
231 | static DEVICE_ATTR(stat, 0444, part_stat_show, NULL); | |
232 | static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL); | |
94ea4158 AV |
233 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
234 | static struct device_attribute dev_attr_fail = | |
5657a819 | 235 | __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store); |
94ea4158 AV |
236 | #endif |
237 | ||
238 | static struct attribute *part_attrs[] = { | |
239 | &dev_attr_partition.attr, | |
240 | &dev_attr_start.attr, | |
241 | &dev_attr_size.attr, | |
242 | &dev_attr_ro.attr, | |
243 | &dev_attr_alignment_offset.attr, | |
244 | &dev_attr_discard_alignment.attr, | |
245 | &dev_attr_stat.attr, | |
246 | &dev_attr_inflight.attr, | |
247 | #ifdef CONFIG_FAIL_MAKE_REQUEST | |
248 | &dev_attr_fail.attr, | |
249 | #endif | |
250 | NULL | |
251 | }; | |
252 | ||
253 | static struct attribute_group part_attr_group = { | |
254 | .attrs = part_attrs, | |
255 | }; | |
256 | ||
257 | static const struct attribute_group *part_attr_groups[] = { | |
258 | &part_attr_group, | |
259 | #ifdef CONFIG_BLK_DEV_IO_TRACE | |
260 | &blk_trace_attr_group, | |
261 | #endif | |
262 | NULL | |
263 | }; | |
264 | ||
265 | static void part_release(struct device *dev) | |
266 | { | |
267 | struct hd_struct *p = dev_to_part(dev); | |
2da78092 | 268 | blk_free_devt(dev->devt); |
b54e5ed8 | 269 | hd_free_part(p); |
94ea4158 AV |
270 | kfree(p); |
271 | } | |
272 | ||
0d9c51a6 SM |
273 | static int part_uevent(struct device *dev, struct kobj_uevent_env *env) |
274 | { | |
275 | struct hd_struct *part = dev_to_part(dev); | |
276 | ||
277 | add_uevent_var(env, "PARTN=%u", part->partno); | |
278 | if (part->info && part->info->volname[0]) | |
279 | add_uevent_var(env, "PARTNAME=%s", part->info->volname); | |
280 | return 0; | |
281 | } | |
282 | ||
94ea4158 AV |
283 | struct device_type part_type = { |
284 | .name = "partition", | |
285 | .groups = part_attr_groups, | |
286 | .release = part_release, | |
0d9c51a6 | 287 | .uevent = part_uevent, |
94ea4158 AV |
288 | }; |
289 | ||
8da2892e | 290 | static void hd_struct_free_work(struct work_struct *work) |
94ea4158 | 291 | { |
8da2892e CH |
292 | struct hd_struct *part = |
293 | container_of(to_rcu_work(work), struct hd_struct, rcu_work); | |
cafe01ef ML |
294 | struct gendisk *disk = part_to_disk(part); |
295 | ||
296 | /* | |
297 | * Release the disk reference acquired in delete_partition here. | |
298 | * We can't release it in hd_struct_free because the final put_device | |
299 | * needs process context and thus can't be run directly from a | |
300 | * percpu_ref ->release handler. | |
301 | */ | |
302 | put_device(disk_to_dev(disk)); | |
94ea4158 AV |
303 | |
304 | part->start_sect = 0; | |
a782483c | 305 | bdev_set_nr_sectors(part->bdev, 0); |
94ea4158 AV |
306 | part_stat_set_all(part, 0); |
307 | put_device(part_to_dev(part)); | |
308 | } | |
309 | ||
8da2892e | 310 | static void hd_struct_free(struct percpu_ref *ref) |
94ea4158 | 311 | { |
6c71013e | 312 | struct hd_struct *part = container_of(ref, struct hd_struct, ref); |
b7d6c303 ML |
313 | struct gendisk *disk = part_to_disk(part); |
314 | struct disk_part_tbl *ptbl = | |
315 | rcu_dereference_protected(disk->part_tbl, 1); | |
316 | ||
317 | rcu_assign_pointer(ptbl->last_lookup, NULL); | |
8da2892e CH |
318 | |
319 | INIT_RCU_WORK(&part->rcu_work, hd_struct_free_work); | |
94a2c3a3 | 320 | queue_rcu_work(system_wq, &part->rcu_work); |
94ea4158 AV |
321 | } |
322 | ||
8da2892e CH |
323 | int hd_ref_init(struct hd_struct *part) |
324 | { | |
325 | if (percpu_ref_init(&part->ref, hd_struct_free, 0, GFP_KERNEL)) | |
326 | return -ENOMEM; | |
327 | return 0; | |
328 | } | |
329 | ||
6d2cf6f2 BVA |
330 | /* |
331 | * Must be called either with bd_mutex held, before a disk can be opened or | |
332 | * after all disk users are gone. | |
333 | */ | |
8328eb28 | 334 | void delete_partition(struct hd_struct *part) |
94ea4158 | 335 | { |
8328eb28 | 336 | struct gendisk *disk = part_to_disk(part); |
6d2cf6f2 BVA |
337 | struct disk_part_tbl *ptbl = |
338 | rcu_dereference_protected(disk->part_tbl, 1); | |
94ea4158 | 339 | |
b7d6c303 ML |
340 | /* |
341 | * ->part_tbl is referenced in this part's release handler, so | |
342 | * we have to hold the disk device | |
343 | */ | |
8328eb28 | 344 | get_device(disk_to_dev(disk)); |
cddae808 | 345 | rcu_assign_pointer(ptbl->part[part->partno], NULL); |
94ea4158 AV |
346 | kobject_put(part->holder_dir); |
347 | device_del(part_to_dev(part)); | |
348 | ||
6fcc44d1 | 349 | /* |
22ae8ce8 CH |
350 | * Remove the block device from the inode hash, so that it cannot be |
351 | * looked up any more even when openers still hold references. | |
6fcc44d1 | 352 | */ |
22ae8ce8 CH |
353 | remove_inode_hash(part->bdev->bd_inode); |
354 | ||
4377b48d | 355 | percpu_ref_kill(&part->ref); |
94ea4158 AV |
356 | } |
357 | ||
358 | static ssize_t whole_disk_show(struct device *dev, | |
359 | struct device_attribute *attr, char *buf) | |
360 | { | |
361 | return 0; | |
362 | } | |
5657a819 | 363 | static DEVICE_ATTR(whole_disk, 0444, whole_disk_show, NULL); |
94ea4158 | 364 | |
6d2cf6f2 BVA |
365 | /* |
366 | * Must be called either with bd_mutex held, before a disk can be opened or | |
367 | * after all disk users are gone. | |
368 | */ | |
fa9156ae | 369 | static struct hd_struct *add_partition(struct gendisk *disk, int partno, |
94ea4158 AV |
370 | sector_t start, sector_t len, int flags, |
371 | struct partition_meta_info *info) | |
372 | { | |
373 | struct hd_struct *p; | |
374 | dev_t devt = MKDEV(0, 0); | |
375 | struct device *ddev = disk_to_dev(disk); | |
376 | struct device *pdev; | |
22ae8ce8 | 377 | struct block_device *bdev; |
94ea4158 AV |
378 | struct disk_part_tbl *ptbl; |
379 | const char *dname; | |
380 | int err; | |
381 | ||
b7205307 CH |
382 | /* |
383 | * Partitions are not supported on zoned block devices that are used as | |
384 | * such. | |
385 | */ | |
386 | switch (disk->queue->limits.zoned) { | |
387 | case BLK_ZONED_HM: | |
388 | pr_warn("%s: partitions not supported on host managed zoned block device\n", | |
389 | disk->disk_name); | |
390 | return ERR_PTR(-ENXIO); | |
391 | case BLK_ZONED_HA: | |
392 | pr_info("%s: disabling host aware zoned block device support due to partitions\n", | |
393 | disk->disk_name); | |
394 | disk->queue->limits.zoned = BLK_ZONED_NONE; | |
395 | break; | |
396 | case BLK_ZONED_NONE: | |
397 | break; | |
398 | } | |
399 | ||
94ea4158 AV |
400 | err = disk_expand_part_tbl(disk, partno); |
401 | if (err) | |
402 | return ERR_PTR(err); | |
6d2cf6f2 | 403 | ptbl = rcu_dereference_protected(disk->part_tbl, 1); |
94ea4158 AV |
404 | |
405 | if (ptbl->part[partno]) | |
406 | return ERR_PTR(-EBUSY); | |
407 | ||
408 | p = kzalloc(sizeof(*p), GFP_KERNEL); | |
409 | if (!p) | |
410 | return ERR_PTR(-EBUSY); | |
411 | ||
22ae8ce8 CH |
412 | bdev = bdev_alloc(disk, partno); |
413 | if (!bdev) | |
15e3d2c5 | 414 | goto out_free; |
22ae8ce8 | 415 | p->bdev = bdev; |
c83f6bf9 | 416 | |
94ea4158 AV |
417 | pdev = part_to_dev(p); |
418 | ||
419 | p->start_sect = start; | |
a782483c | 420 | bdev_set_nr_sectors(bdev, len); |
94ea4158 AV |
421 | p->partno = partno; |
422 | p->policy = get_disk_ro(disk); | |
423 | ||
424 | if (info) { | |
f17c21c1 CH |
425 | struct partition_meta_info *pinfo; |
426 | ||
427 | pinfo = kzalloc_node(sizeof(*pinfo), GFP_KERNEL, disk->node_id); | |
22ae8ce8 CH |
428 | if (!pinfo) |
429 | goto out_bdput; | |
94ea4158 AV |
430 | memcpy(pinfo, info, sizeof(*info)); |
431 | p->info = pinfo; | |
432 | } | |
433 | ||
434 | dname = dev_name(ddev); | |
435 | if (isdigit(dname[strlen(dname) - 1])) | |
436 | dev_set_name(pdev, "%sp%d", dname, partno); | |
437 | else | |
438 | dev_set_name(pdev, "%s%d", dname, partno); | |
439 | ||
440 | device_initialize(pdev); | |
441 | pdev->class = &block_class; | |
442 | pdev->type = &part_type; | |
443 | pdev->parent = ddev; | |
444 | ||
445 | err = blk_alloc_devt(p, &devt); | |
446 | if (err) | |
447 | goto out_free_info; | |
448 | pdev->devt = devt; | |
449 | ||
450 | /* delay uevent until 'holders' subdir is created */ | |
451 | dev_set_uevent_suppress(pdev, 1); | |
452 | err = device_add(pdev); | |
453 | if (err) | |
454 | goto out_put; | |
455 | ||
456 | err = -ENOMEM; | |
457 | p->holder_dir = kobject_create_and_add("holders", &pdev->kobj); | |
458 | if (!p->holder_dir) | |
459 | goto out_del; | |
460 | ||
461 | dev_set_uevent_suppress(pdev, 0); | |
462 | if (flags & ADDPART_FLAG_WHOLEDISK) { | |
463 | err = device_create_file(pdev, &dev_attr_whole_disk); | |
464 | if (err) | |
465 | goto out_del; | |
466 | } | |
467 | ||
b30a337c ML |
468 | err = hd_ref_init(p); |
469 | if (err) { | |
470 | if (flags & ADDPART_FLAG_WHOLEDISK) | |
471 | goto out_remove_file; | |
472 | goto out_del; | |
473 | } | |
474 | ||
94ea4158 | 475 | /* everything is up and running, commence */ |
22ae8ce8 | 476 | bdev_add(bdev, devt); |
94ea4158 AV |
477 | rcu_assign_pointer(ptbl->part[partno], p); |
478 | ||
479 | /* suppress uevent if the disk suppresses it */ | |
480 | if (!dev_get_uevent_suppress(ddev)) | |
481 | kobject_uevent(&pdev->kobj, KOBJ_ADD); | |
b30a337c | 482 | return p; |
94ea4158 AV |
483 | |
484 | out_free_info: | |
f17c21c1 | 485 | kfree(p->info); |
22ae8ce8 CH |
486 | out_bdput: |
487 | bdput(bdev); | |
94ea4158 AV |
488 | out_free: |
489 | kfree(p); | |
490 | return ERR_PTR(err); | |
b30a337c ML |
491 | out_remove_file: |
492 | device_remove_file(pdev, &dev_attr_whole_disk); | |
94ea4158 AV |
493 | out_del: |
494 | kobject_put(p->holder_dir); | |
495 | device_del(pdev); | |
496 | out_put: | |
497 | put_device(pdev); | |
94ea4158 AV |
498 | return ERR_PTR(err); |
499 | } | |
500 | ||
fa9156ae CH |
501 | static bool partition_overlaps(struct gendisk *disk, sector_t start, |
502 | sector_t length, int skip_partno) | |
503 | { | |
504 | struct disk_part_iter piter; | |
505 | struct hd_struct *part; | |
506 | bool overlap = false; | |
507 | ||
508 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY); | |
509 | while ((part = disk_part_iter_next(&piter))) { | |
510 | if (part->partno == skip_partno || | |
a782483c | 511 | start >= part->start_sect + bdev_nr_sectors(part->bdev) || |
fa9156ae CH |
512 | start + length <= part->start_sect) |
513 | continue; | |
514 | overlap = true; | |
515 | break; | |
516 | } | |
517 | ||
518 | disk_part_iter_exit(&piter); | |
519 | return overlap; | |
520 | } | |
521 | ||
522 | int bdev_add_partition(struct block_device *bdev, int partno, | |
523 | sector_t start, sector_t length) | |
524 | { | |
525 | struct hd_struct *part; | |
526 | ||
527 | mutex_lock(&bdev->bd_mutex); | |
528 | if (partition_overlaps(bdev->bd_disk, start, length, -1)) { | |
529 | mutex_unlock(&bdev->bd_mutex); | |
530 | return -EBUSY; | |
531 | } | |
532 | ||
533 | part = add_partition(bdev->bd_disk, partno, start, length, | |
534 | ADDPART_FLAG_NONE, NULL); | |
535 | mutex_unlock(&bdev->bd_mutex); | |
536 | return PTR_ERR_OR_ZERO(part); | |
537 | } | |
538 | ||
539 | int bdev_del_partition(struct block_device *bdev, int partno) | |
540 | { | |
541 | struct block_device *bdevp; | |
08fc1ab6 CH |
542 | struct hd_struct *part = NULL; |
543 | int ret; | |
fa9156ae | 544 | |
08fc1ab6 | 545 | bdevp = bdget_disk(bdev->bd_disk, partno); |
fa9156ae | 546 | if (!bdevp) |
88ce2a53 | 547 | return -ENXIO; |
fa9156ae CH |
548 | |
549 | mutex_lock(&bdevp->bd_mutex); | |
08fc1ab6 CH |
550 | mutex_lock_nested(&bdev->bd_mutex, 1); |
551 | ||
552 | ret = -ENXIO; | |
553 | part = disk_get_part(bdev->bd_disk, partno); | |
554 | if (!part) | |
555 | goto out_unlock; | |
fa9156ae CH |
556 | |
557 | ret = -EBUSY; | |
558 | if (bdevp->bd_openers) | |
559 | goto out_unlock; | |
560 | ||
d5f3178e | 561 | sync_blockdev(bdevp); |
fa9156ae CH |
562 | invalidate_bdev(bdevp); |
563 | ||
8328eb28 | 564 | delete_partition(part); |
fa9156ae CH |
565 | ret = 0; |
566 | out_unlock: | |
08fc1ab6 | 567 | mutex_unlock(&bdev->bd_mutex); |
fa9156ae CH |
568 | mutex_unlock(&bdevp->bd_mutex); |
569 | bdput(bdevp); | |
08fc1ab6 CH |
570 | if (part) |
571 | disk_put_part(part); | |
fa9156ae CH |
572 | return ret; |
573 | } | |
574 | ||
575 | int bdev_resize_partition(struct block_device *bdev, int partno, | |
576 | sector_t start, sector_t length) | |
577 | { | |
578 | struct block_device *bdevp; | |
579 | struct hd_struct *part; | |
580 | int ret = 0; | |
581 | ||
582 | part = disk_get_part(bdev->bd_disk, partno); | |
583 | if (!part) | |
584 | return -ENXIO; | |
585 | ||
586 | ret = -ENOMEM; | |
10ed1666 | 587 | bdevp = bdget_part(part); |
fa9156ae CH |
588 | if (!bdevp) |
589 | goto out_put_part; | |
590 | ||
591 | mutex_lock(&bdevp->bd_mutex); | |
592 | mutex_lock_nested(&bdev->bd_mutex, 1); | |
593 | ||
594 | ret = -EINVAL; | |
595 | if (start != part->start_sect) | |
596 | goto out_unlock; | |
597 | ||
598 | ret = -EBUSY; | |
599 | if (partition_overlaps(bdev->bd_disk, start, length, partno)) | |
600 | goto out_unlock; | |
601 | ||
a782483c | 602 | bdev_set_nr_sectors(bdevp, length); |
fa9156ae CH |
603 | |
604 | ret = 0; | |
605 | out_unlock: | |
606 | mutex_unlock(&bdevp->bd_mutex); | |
607 | mutex_unlock(&bdev->bd_mutex); | |
608 | bdput(bdevp); | |
609 | out_put_part: | |
610 | disk_put_part(part); | |
611 | return ret; | |
612 | } | |
613 | ||
94ea4158 AV |
614 | static bool disk_unlock_native_capacity(struct gendisk *disk) |
615 | { | |
616 | const struct block_device_operations *bdops = disk->fops; | |
617 | ||
618 | if (bdops->unlock_native_capacity && | |
619 | !(disk->flags & GENHD_FL_NATIVE_CAPACITY)) { | |
620 | printk(KERN_CONT "enabling native capacity\n"); | |
621 | bdops->unlock_native_capacity(disk); | |
622 | disk->flags |= GENHD_FL_NATIVE_CAPACITY; | |
623 | return true; | |
624 | } else { | |
625 | printk(KERN_CONT "truncated\n"); | |
626 | return false; | |
627 | } | |
628 | } | |
629 | ||
d46430bf | 630 | int blk_drop_partitions(struct block_device *bdev) |
94ea4158 | 631 | { |
94ea4158 AV |
632 | struct disk_part_iter piter; |
633 | struct hd_struct *part; | |
94ea4158 | 634 | |
10c70d95 | 635 | if (bdev->bd_part_count) |
94ea4158 | 636 | return -EBUSY; |
e669c1da CH |
637 | |
638 | sync_blockdev(bdev); | |
639 | invalidate_bdev(bdev); | |
94ea4158 | 640 | |
d46430bf | 641 | disk_part_iter_init(&piter, bdev->bd_disk, DISK_PITER_INCL_EMPTY); |
94ea4158 | 642 | while ((part = disk_part_iter_next(&piter))) |
8328eb28 | 643 | delete_partition(part); |
94ea4158 AV |
644 | disk_part_iter_exit(&piter); |
645 | ||
fe316bf2 JN |
646 | return 0; |
647 | } | |
21be6cdc CH |
648 | #ifdef CONFIG_S390 |
649 | /* for historic reasons in the DASD driver */ | |
650 | EXPORT_SYMBOL_GPL(blk_drop_partitions); | |
651 | #endif | |
fe316bf2 | 652 | |
f902b026 CH |
653 | static bool blk_add_partition(struct gendisk *disk, struct block_device *bdev, |
654 | struct parsed_partitions *state, int p) | |
fe316bf2 | 655 | { |
f902b026 CH |
656 | sector_t size = state->parts[p].size; |
657 | sector_t from = state->parts[p].from; | |
fe316bf2 | 658 | struct hd_struct *part; |
f902b026 CH |
659 | |
660 | if (!size) | |
661 | return true; | |
662 | ||
663 | if (from >= get_capacity(disk)) { | |
664 | printk(KERN_WARNING | |
665 | "%s: p%d start %llu is beyond EOD, ", | |
666 | disk->disk_name, p, (unsigned long long) from); | |
667 | if (disk_unlock_native_capacity(disk)) | |
668 | return false; | |
669 | return true; | |
fe316bf2 JN |
670 | } |
671 | ||
f902b026 CH |
672 | if (from + size > get_capacity(disk)) { |
673 | printk(KERN_WARNING | |
674 | "%s: p%d size %llu extends beyond EOD, ", | |
675 | disk->disk_name, p, (unsigned long long) size); | |
fe316bf2 | 676 | |
f902b026 CH |
677 | if (disk_unlock_native_capacity(disk)) |
678 | return false; | |
679 | ||
680 | /* | |
681 | * We can not ignore partitions of broken tables created by for | |
682 | * example camera firmware, but we limit them to the end of the | |
683 | * disk to avoid creating invalid block devices. | |
684 | */ | |
685 | size = get_capacity(disk) - from; | |
686 | } | |
687 | ||
688 | part = add_partition(disk, p, from, size, state->parts[p].flags, | |
689 | &state->parts[p].info); | |
b7205307 | 690 | if (IS_ERR(part) && PTR_ERR(part) != -ENXIO) { |
f902b026 CH |
691 | printk(KERN_ERR " %s: p%d could not be added: %ld\n", |
692 | disk->disk_name, p, -PTR_ERR(part)); | |
693 | return true; | |
694 | } | |
695 | ||
74cc979c CH |
696 | if (IS_BUILTIN(CONFIG_BLK_DEV_MD) && |
697 | (state->parts[p].flags & ADDPART_FLAG_RAID)) | |
f902b026 | 698 | md_autodetect_dev(part_to_dev(part)->devt); |
74cc979c | 699 | |
f902b026 CH |
700 | return true; |
701 | } | |
702 | ||
a1548b67 | 703 | int blk_add_partitions(struct gendisk *disk, struct block_device *bdev) |
f902b026 CH |
704 | { |
705 | struct parsed_partitions *state; | |
706 | int ret = -EAGAIN, p, highest; | |
fe316bf2 | 707 | |
142fe8f4 CH |
708 | if (!disk_part_scan_enabled(disk)) |
709 | return 0; | |
710 | ||
f902b026 CH |
711 | state = check_partition(disk, bdev); |
712 | if (!state) | |
94ea4158 AV |
713 | return 0; |
714 | if (IS_ERR(state)) { | |
715 | /* | |
f902b026 CH |
716 | * I/O error reading the partition table. If we tried to read |
717 | * beyond EOD, retry after unlocking the native capacity. | |
94ea4158 AV |
718 | */ |
719 | if (PTR_ERR(state) == -ENOSPC) { | |
720 | printk(KERN_WARNING "%s: partition table beyond EOD, ", | |
721 | disk->disk_name); | |
722 | if (disk_unlock_native_capacity(disk)) | |
f902b026 | 723 | return -EAGAIN; |
94ea4158 AV |
724 | } |
725 | return -EIO; | |
726 | } | |
5eac3eb3 | 727 | |
f902b026 | 728 | /* |
b7205307 | 729 | * Partitions are not supported on host managed zoned block devices. |
f902b026 | 730 | */ |
b7205307 CH |
731 | if (disk->queue->limits.zoned == BLK_ZONED_HM) { |
732 | pr_warn("%s: ignoring partition table on host managed zoned block device\n", | |
5eac3eb3 | 733 | disk->disk_name); |
f902b026 CH |
734 | ret = 0; |
735 | goto out_free_state; | |
5eac3eb3 DLM |
736 | } |
737 | ||
94ea4158 | 738 | /* |
f902b026 CH |
739 | * If we read beyond EOD, try unlocking native capacity even if the |
740 | * partition table was successfully read as we could be missing some | |
741 | * partitions. | |
94ea4158 AV |
742 | */ |
743 | if (state->access_beyond_eod) { | |
744 | printk(KERN_WARNING | |
745 | "%s: partition table partially beyond EOD, ", | |
746 | disk->disk_name); | |
747 | if (disk_unlock_native_capacity(disk)) | |
f902b026 | 748 | goto out_free_state; |
94ea4158 AV |
749 | } |
750 | ||
751 | /* tell userspace that the media / partition table may have changed */ | |
752 | kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE); | |
753 | ||
f902b026 CH |
754 | /* |
755 | * Detect the highest partition number and preallocate disk->part_tbl. | |
756 | * This is an optimization and not strictly necessary. | |
94ea4158 AV |
757 | */ |
758 | for (p = 1, highest = 0; p < state->limit; p++) | |
759 | if (state->parts[p].size) | |
760 | highest = p; | |
94ea4158 AV |
761 | disk_expand_part_tbl(disk, highest); |
762 | ||
f902b026 CH |
763 | for (p = 1; p < state->limit; p++) |
764 | if (!blk_add_partition(disk, bdev, state, p)) | |
765 | goto out_free_state; | |
94ea4158 | 766 | |
f902b026 CH |
767 | ret = 0; |
768 | out_free_state: | |
ac2e5327 | 769 | free_partitions(state); |
f902b026 | 770 | return ret; |
fe316bf2 JN |
771 | } |
772 | ||
1a9fba3a | 773 | void *read_part_sector(struct parsed_partitions *state, sector_t n, Sector *p) |
d1a5f2b4 | 774 | { |
1a9fba3a | 775 | struct address_space *mapping = state->bdev->bd_inode->i_mapping; |
94ea4158 AV |
776 | struct page *page; |
777 | ||
1a9fba3a CH |
778 | if (n >= get_capacity(state->bdev->bd_disk)) { |
779 | state->access_beyond_eod = true; | |
780 | return NULL; | |
94ea4158 | 781 | } |
1a9fba3a CH |
782 | |
783 | page = read_mapping_page(mapping, | |
784 | (pgoff_t)(n >> (PAGE_SHIFT - 9)), NULL); | |
785 | if (IS_ERR(page)) | |
786 | goto out; | |
787 | if (PageError(page)) | |
788 | goto out_put_page; | |
789 | ||
790 | p->v = page; | |
791 | return (unsigned char *)page_address(page) + | |
792 | ((n & ((1 << (PAGE_SHIFT - 9)) - 1)) << SECTOR_SHIFT); | |
793 | out_put_page: | |
794 | put_page(page); | |
795 | out: | |
94ea4158 AV |
796 | p->v = NULL; |
797 | return NULL; | |
798 | } |