blob: 736f5b979a2104640c3db8438d963fbc3ed21772
1 | /* |
2 | * Code extracted from drivers/block/genhd.c |
3 | * Copyright (C) 1991-1998 Linus Torvalds |
4 | * Re-organised Feb 1998 Russell King |
5 | * |
6 | * We now have independent partition support from the |
7 | * block drivers, which allows all the partition code to |
8 | * be grouped in one location, and it to be mostly self |
9 | * contained. |
10 | */ |
11 | |
12 | #include <linux/init.h> |
13 | #include <linux/module.h> |
14 | #include <linux/fs.h> |
15 | #include <linux/slab.h> |
16 | #include <linux/kmod.h> |
17 | #include <linux/ctype.h> |
18 | #include <linux/genhd.h> |
19 | #include <linux/dax.h> |
20 | #include <linux/blktrace_api.h> |
21 | |
22 | #include "partitions/check.h" |
23 | |
24 | #ifdef CONFIG_BLK_DEV_MD |
25 | extern void md_autodetect_dev(dev_t dev); |
26 | #endif |
27 | |
28 | /* |
29 | * disk_name() is used by partition check code and the genhd driver. |
30 | * It formats the devicename of the indicated disk into |
31 | * the supplied buffer (of size at least 32), and returns |
32 | * a pointer to that same buffer (for convenience). |
33 | */ |
34 | |
35 | char *disk_name(struct gendisk *hd, int partno, char *buf) |
36 | { |
37 | if (!partno) |
38 | snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name); |
39 | else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1])) |
40 | snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno); |
41 | else |
42 | snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno); |
43 | |
44 | return buf; |
45 | } |
46 | |
47 | const char *bdevname(struct block_device *bdev, char *buf) |
48 | { |
49 | return disk_name(bdev->bd_disk, bdev->bd_part->partno, buf); |
50 | } |
51 | |
52 | EXPORT_SYMBOL(bdevname); |
53 | |
54 | /* |
55 | * There's very little reason to use this, you should really |
56 | * have a struct block_device just about everywhere and use |
57 | * bdevname() instead. |
58 | */ |
59 | const char *__bdevname(dev_t dev, char *buffer) |
60 | { |
61 | scnprintf(buffer, BDEVNAME_SIZE, "unknown-block(%u,%u)", |
62 | MAJOR(dev), MINOR(dev)); |
63 | return buffer; |
64 | } |
65 | |
66 | EXPORT_SYMBOL(__bdevname); |
67 | |
68 | static ssize_t part_partition_show(struct device *dev, |
69 | struct device_attribute *attr, char *buf) |
70 | { |
71 | struct hd_struct *p = dev_to_part(dev); |
72 | |
73 | return sprintf(buf, "%d\n", p->partno); |
74 | } |
75 | |
76 | static ssize_t part_start_show(struct device *dev, |
77 | struct device_attribute *attr, char *buf) |
78 | { |
79 | struct hd_struct *p = dev_to_part(dev); |
80 | |
81 | return sprintf(buf, "%llu\n",(unsigned long long)p->start_sect); |
82 | } |
83 | |
84 | ssize_t part_size_show(struct device *dev, |
85 | struct device_attribute *attr, char *buf) |
86 | { |
87 | struct hd_struct *p = dev_to_part(dev); |
88 | return sprintf(buf, "%llu\n",(unsigned long long)part_nr_sects_read(p)); |
89 | } |
90 | |
91 | static ssize_t part_ro_show(struct device *dev, |
92 | struct device_attribute *attr, char *buf) |
93 | { |
94 | struct hd_struct *p = dev_to_part(dev); |
95 | return sprintf(buf, "%d\n", p->policy ? 1 : 0); |
96 | } |
97 | |
98 | static ssize_t part_alignment_offset_show(struct device *dev, |
99 | struct device_attribute *attr, char *buf) |
100 | { |
101 | struct hd_struct *p = dev_to_part(dev); |
102 | return sprintf(buf, "%llu\n", (unsigned long long)p->alignment_offset); |
103 | } |
104 | |
105 | static ssize_t part_discard_alignment_show(struct device *dev, |
106 | struct device_attribute *attr, char *buf) |
107 | { |
108 | struct hd_struct *p = dev_to_part(dev); |
109 | return sprintf(buf, "%u\n", p->discard_alignment); |
110 | } |
111 | |
112 | ssize_t part_stat_show(struct device *dev, |
113 | struct device_attribute *attr, char *buf) |
114 | { |
115 | struct hd_struct *p = dev_to_part(dev); |
116 | int cpu; |
117 | |
118 | cpu = part_stat_lock(); |
119 | part_round_stats(cpu, p); |
120 | part_stat_unlock(); |
121 | return sprintf(buf, |
122 | "%8lu %8lu %8llu %8u " |
123 | "%8lu %8lu %8llu %8u " |
124 | "%8u %8u %8u" |
125 | "\n", |
126 | part_stat_read(p, ios[READ]), |
127 | part_stat_read(p, merges[READ]), |
128 | (unsigned long long)part_stat_read(p, sectors[READ]), |
129 | jiffies_to_msecs(part_stat_read(p, ticks[READ])), |
130 | part_stat_read(p, ios[WRITE]), |
131 | part_stat_read(p, merges[WRITE]), |
132 | (unsigned long long)part_stat_read(p, sectors[WRITE]), |
133 | jiffies_to_msecs(part_stat_read(p, ticks[WRITE])), |
134 | part_in_flight(p), |
135 | jiffies_to_msecs(part_stat_read(p, io_ticks)), |
136 | jiffies_to_msecs(part_stat_read(p, time_in_queue))); |
137 | } |
138 | |
139 | ssize_t part_inflight_show(struct device *dev, |
140 | struct device_attribute *attr, char *buf) |
141 | { |
142 | struct hd_struct *p = dev_to_part(dev); |
143 | |
144 | return sprintf(buf, "%8u %8u\n", atomic_read(&p->in_flight[0]), |
145 | atomic_read(&p->in_flight[1])); |
146 | } |
147 | |
148 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
149 | ssize_t part_fail_show(struct device *dev, |
150 | struct device_attribute *attr, char *buf) |
151 | { |
152 | struct hd_struct *p = dev_to_part(dev); |
153 | |
154 | return sprintf(buf, "%d\n", p->make_it_fail); |
155 | } |
156 | |
157 | ssize_t part_fail_store(struct device *dev, |
158 | struct device_attribute *attr, |
159 | const char *buf, size_t count) |
160 | { |
161 | struct hd_struct *p = dev_to_part(dev); |
162 | int i; |
163 | |
164 | if (count > 0 && sscanf(buf, "%d", &i) > 0) |
165 | p->make_it_fail = (i == 0) ? 0 : 1; |
166 | |
167 | return count; |
168 | } |
169 | #endif |
170 | |
171 | static DEVICE_ATTR(partition, S_IRUGO, part_partition_show, NULL); |
172 | static DEVICE_ATTR(start, S_IRUGO, part_start_show, NULL); |
173 | static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL); |
174 | static DEVICE_ATTR(ro, S_IRUGO, part_ro_show, NULL); |
175 | static DEVICE_ATTR(alignment_offset, S_IRUGO, part_alignment_offset_show, NULL); |
176 | static DEVICE_ATTR(discard_alignment, S_IRUGO, part_discard_alignment_show, |
177 | NULL); |
178 | static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL); |
179 | static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL); |
180 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
181 | static struct device_attribute dev_attr_fail = |
182 | __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store); |
183 | #endif |
184 | |
185 | static struct attribute *part_attrs[] = { |
186 | &dev_attr_partition.attr, |
187 | &dev_attr_start.attr, |
188 | &dev_attr_size.attr, |
189 | &dev_attr_ro.attr, |
190 | &dev_attr_alignment_offset.attr, |
191 | &dev_attr_discard_alignment.attr, |
192 | &dev_attr_stat.attr, |
193 | &dev_attr_inflight.attr, |
194 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
195 | &dev_attr_fail.attr, |
196 | #endif |
197 | NULL |
198 | }; |
199 | |
200 | static struct attribute_group part_attr_group = { |
201 | .attrs = part_attrs, |
202 | }; |
203 | |
204 | static const struct attribute_group *part_attr_groups[] = { |
205 | &part_attr_group, |
206 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
207 | &blk_trace_attr_group, |
208 | #endif |
209 | NULL |
210 | }; |
211 | |
212 | static void part_release(struct device *dev) |
213 | { |
214 | struct hd_struct *p = dev_to_part(dev); |
215 | blk_free_devt(dev->devt); |
216 | hd_free_part(p); |
217 | kfree(p); |
218 | } |
219 | |
220 | static int part_uevent(struct device *dev, struct kobj_uevent_env *env) |
221 | { |
222 | struct hd_struct *part = dev_to_part(dev); |
223 | |
224 | add_uevent_var(env, "PARTN=%u", part->partno); |
225 | if (part->info && part->info->volname[0]) |
226 | add_uevent_var(env, "PARTNAME=%s", part->info->volname); |
227 | return 0; |
228 | } |
229 | |
230 | struct device_type part_type = { |
231 | .name = "partition", |
232 | .groups = part_attr_groups, |
233 | .release = part_release, |
234 | .uevent = part_uevent, |
235 | }; |
236 | |
237 | static void delete_partition_rcu_cb(struct rcu_head *head) |
238 | { |
239 | struct hd_struct *part = container_of(head, struct hd_struct, rcu_head); |
240 | |
241 | part->start_sect = 0; |
242 | part->nr_sects = 0; |
243 | part_stat_set_all(part, 0); |
244 | put_device(part_to_dev(part)); |
245 | } |
246 | |
247 | void __delete_partition(struct percpu_ref *ref) |
248 | { |
249 | struct hd_struct *part = container_of(ref, struct hd_struct, ref); |
250 | call_rcu(&part->rcu_head, delete_partition_rcu_cb); |
251 | } |
252 | |
253 | void delete_partition(struct gendisk *disk, int partno) |
254 | { |
255 | struct disk_part_tbl *ptbl = disk->part_tbl; |
256 | struct hd_struct *part; |
257 | |
258 | if (partno >= ptbl->len) |
259 | return; |
260 | |
261 | part = ptbl->part[partno]; |
262 | if (!part) |
263 | return; |
264 | |
265 | rcu_assign_pointer(ptbl->part[partno], NULL); |
266 | rcu_assign_pointer(ptbl->last_lookup, NULL); |
267 | kobject_put(part->holder_dir); |
268 | device_del(part_to_dev(part)); |
269 | |
270 | hd_struct_kill(part); |
271 | } |
272 | |
273 | static ssize_t whole_disk_show(struct device *dev, |
274 | struct device_attribute *attr, char *buf) |
275 | { |
276 | return 0; |
277 | } |
278 | static DEVICE_ATTR(whole_disk, S_IRUSR | S_IRGRP | S_IROTH, |
279 | whole_disk_show, NULL); |
280 | |
281 | struct hd_struct *add_partition(struct gendisk *disk, int partno, |
282 | sector_t start, sector_t len, int flags, |
283 | struct partition_meta_info *info) |
284 | { |
285 | struct hd_struct *p; |
286 | dev_t devt = MKDEV(0, 0); |
287 | struct device *ddev = disk_to_dev(disk); |
288 | struct device *pdev; |
289 | struct disk_part_tbl *ptbl; |
290 | const char *dname; |
291 | int err; |
292 | |
293 | err = disk_expand_part_tbl(disk, partno); |
294 | |
295 | if (err) |
296 | return ERR_PTR(err); |
297 | ptbl = disk->part_tbl; |
298 | |
299 | if (ptbl->part[partno]) |
300 | return ERR_PTR(-EBUSY); |
301 | |
302 | p = kzalloc(sizeof(*p), GFP_KERNEL); |
303 | if (!p) |
304 | return ERR_PTR(-EBUSY); |
305 | |
306 | if (!init_part_stats(p)) { |
307 | err = -ENOMEM; |
308 | goto out_free; |
309 | } |
310 | |
311 | seqcount_init(&p->nr_sects_seq); |
312 | pdev = part_to_dev(p); |
313 | |
314 | p->start_sect = start; |
315 | p->alignment_offset = |
316 | queue_limit_alignment_offset(&disk->queue->limits, start); |
317 | p->discard_alignment = |
318 | queue_limit_discard_alignment(&disk->queue->limits, start); |
319 | p->nr_sects = len; |
320 | p->partno = partno; |
321 | p->policy = get_disk_ro(disk); |
322 | |
323 | if (info) { |
324 | struct partition_meta_info *pinfo = alloc_part_info(disk); |
325 | if (!pinfo) { |
326 | err = -ENOMEM; |
327 | goto out_free_stats; |
328 | } |
329 | memcpy(pinfo, info, sizeof(*info)); |
330 | p->info = pinfo; |
331 | } |
332 | |
333 | dname = dev_name(ddev); |
334 | |
335 | if (isdigit(dname[strlen(dname) - 1])) |
336 | dev_set_name(pdev, "%sp%d", dname, partno); |
337 | else |
338 | dev_set_name(pdev, "%s%d", dname, partno); |
339 | |
340 | device_initialize(pdev); |
341 | pdev->class = &block_class; |
342 | pdev->type = &part_type; |
343 | pdev->parent = ddev; |
344 | |
345 | err = blk_alloc_devt(p, &devt); |
346 | if (err) |
347 | goto out_free_info; |
348 | pdev->devt = devt; |
349 | |
350 | /* delay uevent until 'holders' subdir is created */ |
351 | dev_set_uevent_suppress(pdev, 1); |
352 | err = device_add(pdev); |
353 | if (err) |
354 | goto out_put; |
355 | |
356 | err = -ENOMEM; |
357 | p->holder_dir = kobject_create_and_add("holders", &pdev->kobj); |
358 | if (!p->holder_dir) |
359 | goto out_del; |
360 | |
361 | dev_set_uevent_suppress(pdev, 0); |
362 | if (flags & ADDPART_FLAG_WHOLEDISK) { |
363 | err = device_create_file(pdev, &dev_attr_whole_disk); |
364 | if (err) |
365 | goto out_del; |
366 | } |
367 | |
368 | err = hd_ref_init(p); |
369 | if (err) { |
370 | if (flags & ADDPART_FLAG_WHOLEDISK) |
371 | goto out_remove_file; |
372 | goto out_del; |
373 | } |
374 | |
375 | /* everything is up and running, commence */ |
376 | rcu_assign_pointer(ptbl->part[partno], p); |
377 | |
378 | /* suppress uevent if the disk suppresses it */ |
379 | if (!dev_get_uevent_suppress(ddev)) |
380 | kobject_uevent(&pdev->kobj, KOBJ_ADD); |
381 | return p; |
382 | |
383 | out_free_info: |
384 | free_part_info(p); |
385 | out_free_stats: |
386 | free_part_stats(p); |
387 | out_free: |
388 | kfree(p); |
389 | return ERR_PTR(err); |
390 | out_remove_file: |
391 | device_remove_file(pdev, &dev_attr_whole_disk); |
392 | out_del: |
393 | kobject_put(p->holder_dir); |
394 | device_del(pdev); |
395 | out_put: |
396 | put_device(pdev); |
397 | blk_free_devt(devt); |
398 | return ERR_PTR(err); |
399 | } |
400 | |
401 | static bool disk_unlock_native_capacity(struct gendisk *disk) |
402 | { |
403 | const struct block_device_operations *bdops = disk->fops; |
404 | |
405 | if (bdops->unlock_native_capacity && |
406 | !(disk->flags & GENHD_FL_NATIVE_CAPACITY)) { |
407 | printk(KERN_CONT "enabling native capacity\n"); |
408 | bdops->unlock_native_capacity(disk); |
409 | disk->flags |= GENHD_FL_NATIVE_CAPACITY; |
410 | return true; |
411 | } else { |
412 | printk(KERN_CONT "truncated\n"); |
413 | return false; |
414 | } |
415 | } |
416 | |
417 | static int drop_partitions(struct gendisk *disk, struct block_device *bdev) |
418 | { |
419 | struct disk_part_iter piter; |
420 | struct hd_struct *part; |
421 | int res; |
422 | |
423 | if (bdev->bd_part_count || bdev->bd_super) |
424 | return -EBUSY; |
425 | res = invalidate_partition(disk, 0); |
426 | if (res) |
427 | return res; |
428 | |
429 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY); |
430 | while ((part = disk_part_iter_next(&piter))) |
431 | delete_partition(disk, part->partno); |
432 | disk_part_iter_exit(&piter); |
433 | |
434 | return 0; |
435 | } |
436 | |
437 | int rescan_partitions(struct gendisk *disk, struct block_device *bdev) |
438 | { |
439 | struct parsed_partitions *state = NULL; |
440 | struct hd_struct *part; |
441 | int p, highest, res; |
442 | rescan: |
443 | if (state && !IS_ERR(state)) { |
444 | free_partitions(state); |
445 | state = NULL; |
446 | } |
447 | |
448 | res = drop_partitions(disk, bdev); |
449 | if (res) |
450 | return res; |
451 | |
452 | if (disk->fops->revalidate_disk) |
453 | disk->fops->revalidate_disk(disk); |
454 | check_disk_size_change(disk, bdev); |
455 | bdev->bd_invalidated = 0; |
456 | if (!get_capacity(disk) || !(state = check_partition(disk, bdev))) |
457 | return 0; |
458 | if (IS_ERR(state)) { |
459 | /* |
460 | * I/O error reading the partition table. If any |
461 | * partition code tried to read beyond EOD, retry |
462 | * after unlocking native capacity. |
463 | */ |
464 | if (PTR_ERR(state) == -ENOSPC) { |
465 | printk(KERN_WARNING "%s: partition table beyond EOD, ", |
466 | disk->disk_name); |
467 | if (disk_unlock_native_capacity(disk)) |
468 | goto rescan; |
469 | } |
470 | return -EIO; |
471 | } |
472 | /* |
473 | * If any partition code tried to read beyond EOD, try |
474 | * unlocking native capacity even if partition table is |
475 | * successfully read as we could be missing some partitions. |
476 | */ |
477 | if (state->access_beyond_eod) { |
478 | printk(KERN_WARNING |
479 | "%s: partition table partially beyond EOD, ", |
480 | disk->disk_name); |
481 | if (disk_unlock_native_capacity(disk)) |
482 | goto rescan; |
483 | } |
484 | |
485 | /* tell userspace that the media / partition table may have changed */ |
486 | kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE); |
487 | |
488 | /* Detect the highest partition number and preallocate |
489 | * disk->part_tbl. This is an optimization and not strictly |
490 | * necessary. |
491 | */ |
492 | for (p = 1, highest = 0; p < state->limit; p++) |
493 | if (state->parts[p].size) |
494 | highest = p; |
495 | |
496 | disk_expand_part_tbl(disk, highest); |
497 | |
498 | /* add partitions */ |
499 | for (p = 1; p < state->limit; p++) { |
500 | sector_t size, from; |
501 | |
502 | size = state->parts[p].size; |
503 | if (!size) |
504 | continue; |
505 | |
506 | from = state->parts[p].from; |
507 | if (from >= get_capacity(disk)) { |
508 | printk(KERN_WARNING |
509 | "%s: p%d start %llu is beyond EOD, ", |
510 | disk->disk_name, p, (unsigned long long) from); |
511 | if (disk_unlock_native_capacity(disk)) |
512 | goto rescan; |
513 | continue; |
514 | } |
515 | |
516 | if (from + size > get_capacity(disk)) { |
517 | printk(KERN_WARNING |
518 | "%s: p%d size %llu extends beyond EOD, ", |
519 | disk->disk_name, p, (unsigned long long) size); |
520 | |
521 | if (disk_unlock_native_capacity(disk)) { |
522 | /* free state and restart */ |
523 | goto rescan; |
524 | } else { |
525 | /* |
526 | * we can not ignore partitions of broken tables |
527 | * created by for example camera firmware, but |
528 | * we limit them to the end of the disk to avoid |
529 | * creating invalid block devices |
530 | */ |
531 | size = get_capacity(disk) - from; |
532 | } |
533 | } |
534 | |
535 | part = add_partition(disk, p, from, size, |
536 | state->parts[p].flags, |
537 | &state->parts[p].info); |
538 | if (IS_ERR(part)) { |
539 | printk(KERN_ERR " %s: p%d could not be added: %ld\n", |
540 | disk->disk_name, p, -PTR_ERR(part)); |
541 | continue; |
542 | } |
543 | #ifdef CONFIG_BLK_DEV_MD |
544 | if (state->parts[p].flags & ADDPART_FLAG_RAID) |
545 | md_autodetect_dev(part_to_dev(part)->devt); |
546 | #endif |
547 | } |
548 | free_partitions(state); |
549 | return 0; |
550 | } |
551 | |
552 | int invalidate_partitions(struct gendisk *disk, struct block_device *bdev) |
553 | { |
554 | int res; |
555 | |
556 | if (!bdev->bd_invalidated) |
557 | return 0; |
558 | |
559 | res = drop_partitions(disk, bdev); |
560 | if (res) |
561 | return res; |
562 | |
563 | set_capacity(disk, 0); |
564 | check_disk_size_change(disk, bdev); |
565 | bdev->bd_invalidated = 0; |
566 | /* tell userspace that the media / partition table may have changed */ |
567 | kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE); |
568 | |
569 | return 0; |
570 | } |
571 | |
572 | static struct page *read_pagecache_sector(struct block_device *bdev, sector_t n) |
573 | { |
574 | struct address_space *mapping = bdev->bd_inode->i_mapping; |
575 | |
576 | return read_mapping_page(mapping, (pgoff_t)(n >> (PAGE_SHIFT-9)), |
577 | NULL); |
578 | } |
579 | |
580 | unsigned char *read_dev_sector(struct block_device *bdev, sector_t n, Sector *p) |
581 | { |
582 | struct page *page; |
583 | |
584 | /* don't populate page cache for dax capable devices */ |
585 | if (IS_DAX(bdev->bd_inode)) |
586 | page = read_dax_sector(bdev, n); |
587 | else |
588 | page = read_pagecache_sector(bdev, n); |
589 | |
590 | if (!IS_ERR(page)) { |
591 | if (PageError(page)) |
592 | goto fail; |
593 | p->v = page; |
594 | return (unsigned char *)page_address(page) + ((n & ((1 << (PAGE_SHIFT - 9)) - 1)) << 9); |
595 | fail: |
596 | put_page(page); |
597 | } |
598 | p->v = NULL; |
599 | return NULL; |
600 | } |
601 | |
602 | EXPORT_SYMBOL(read_dev_sector); |
603 |