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
disk_name(struct gendisk * hd,int partno,char * buf)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
bdevname(struct block_device * bdev,char * buf)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 */
__bdevname(dev_t dev,char * buffer)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
part_partition_show(struct device * dev,struct device_attribute * attr,char * buf)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
part_start_show(struct device * dev,struct device_attribute * attr,char * buf)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
part_size_show(struct device * dev,struct device_attribute * attr,char * buf)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
part_ro_show(struct device * dev,struct device_attribute * attr,char * buf)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
part_alignment_offset_show(struct device * dev,struct device_attribute * attr,char * buf)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
part_discard_alignment_show(struct device * dev,struct device_attribute * attr,char * buf)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
part_stat_show(struct device * dev,struct device_attribute * attr,char * buf)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
part_inflight_show(struct device * dev,struct device_attribute * attr,char * buf)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
part_fail_show(struct device * dev,struct device_attribute * attr,char * buf)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
part_fail_store(struct device * dev,struct device_attribute * attr,const char * buf,size_t count)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
part_release(struct device * dev)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
part_uevent(struct device * dev,struct kobj_uevent_env * env)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
delete_partition_rcu_cb(struct rcu_head * head)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
__delete_partition(struct percpu_ref * ref)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
delete_partition(struct gendisk * disk,int partno)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
whole_disk_show(struct device * dev,struct device_attribute * attr,char * buf)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
add_partition(struct gendisk * disk,int partno,sector_t start,sector_t len,int flags,struct partition_meta_info * info)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 if (err)
295 return ERR_PTR(err);
296 ptbl = disk->part_tbl;
297
298 if (ptbl->part[partno])
299 return ERR_PTR(-EBUSY);
300
301 p = kzalloc(sizeof(*p), GFP_KERNEL);
302 if (!p)
303 return ERR_PTR(-EBUSY);
304
305 if (!init_part_stats(p)) {
306 err = -ENOMEM;
307 goto out_free;
308 }
309
310 seqcount_init(&p->nr_sects_seq);
311 pdev = part_to_dev(p);
312
313 p->start_sect = start;
314 p->alignment_offset =
315 queue_limit_alignment_offset(&disk->queue->limits, start);
316 p->discard_alignment =
317 queue_limit_discard_alignment(&disk->queue->limits, start);
318 p->nr_sects = len;
319 p->partno = partno;
320 p->policy = get_disk_ro(disk);
321
322 if (info) {
323 struct partition_meta_info *pinfo = alloc_part_info(disk);
324 if (!pinfo) {
325 err = -ENOMEM;
326 goto out_free_stats;
327 }
328 memcpy(pinfo, info, sizeof(*info));
329 p->info = pinfo;
330 }
331
332 dname = dev_name(ddev);
333 if (isdigit(dname[strlen(dname) - 1]))
334 dev_set_name(pdev, "%sp%d", dname, partno);
335 else
336 dev_set_name(pdev, "%s%d", dname, partno);
337
338 device_initialize(pdev);
339 pdev->class = &block_class;
340 pdev->type = &part_type;
341 pdev->parent = ddev;
342
343 err = blk_alloc_devt(p, &devt);
344 if (err)
345 goto out_free_info;
346 pdev->devt = devt;
347
348 /* delay uevent until 'holders' subdir is created */
349 dev_set_uevent_suppress(pdev, 1);
350 err = device_add(pdev);
351 if (err)
352 goto out_put;
353
354 err = -ENOMEM;
355 p->holder_dir = kobject_create_and_add("holders", &pdev->kobj);
356 if (!p->holder_dir)
357 goto out_del;
358
359 dev_set_uevent_suppress(pdev, 0);
360 if (flags & ADDPART_FLAG_WHOLEDISK) {
361 err = device_create_file(pdev, &dev_attr_whole_disk);
362 if (err)
363 goto out_del;
364 }
365
366 err = hd_ref_init(p);
367 if (err) {
368 if (flags & ADDPART_FLAG_WHOLEDISK)
369 goto out_remove_file;
370 goto out_del;
371 }
372
373 /* everything is up and running, commence */
374 rcu_assign_pointer(ptbl->part[partno], p);
375
376 /* suppress uevent if the disk suppresses it */
377 if (!dev_get_uevent_suppress(ddev))
378 kobject_uevent(&pdev->kobj, KOBJ_ADD);
379 return p;
380
381 out_free_info:
382 free_part_info(p);
383 out_free_stats:
384 free_part_stats(p);
385 out_free:
386 kfree(p);
387 return ERR_PTR(err);
388 out_remove_file:
389 device_remove_file(pdev, &dev_attr_whole_disk);
390 out_del:
391 kobject_put(p->holder_dir);
392 device_del(pdev);
393 out_put:
394 put_device(pdev);
395 blk_free_devt(devt);
396 return ERR_PTR(err);
397 }
398
disk_unlock_native_capacity(struct gendisk * disk)399 static bool disk_unlock_native_capacity(struct gendisk *disk)
400 {
401 const struct block_device_operations *bdops = disk->fops;
402
403 if (bdops->unlock_native_capacity &&
404 !(disk->flags & GENHD_FL_NATIVE_CAPACITY)) {
405 printk(KERN_CONT "enabling native capacity\n");
406 bdops->unlock_native_capacity(disk);
407 disk->flags |= GENHD_FL_NATIVE_CAPACITY;
408 return true;
409 } else {
410 printk(KERN_CONT "truncated\n");
411 return false;
412 }
413 }
414
drop_partitions(struct gendisk * disk,struct block_device * bdev)415 static int drop_partitions(struct gendisk *disk, struct block_device *bdev)
416 {
417 struct disk_part_iter piter;
418 struct hd_struct *part;
419 int res;
420
421 if (bdev->bd_part_count || bdev->bd_super)
422 return -EBUSY;
423 res = invalidate_partition(disk, 0);
424 if (res)
425 return res;
426
427 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
428 while ((part = disk_part_iter_next(&piter)))
429 delete_partition(disk, part->partno);
430 disk_part_iter_exit(&piter);
431
432 return 0;
433 }
434
rescan_partitions(struct gendisk * disk,struct block_device * bdev)435 int rescan_partitions(struct gendisk *disk, struct block_device *bdev)
436 {
437 struct parsed_partitions *state = NULL;
438 struct hd_struct *part;
439 int p, highest, res;
440 rescan:
441 if (state && !IS_ERR(state)) {
442 free_partitions(state);
443 state = NULL;
444 }
445
446 res = drop_partitions(disk, bdev);
447 if (res)
448 return res;
449
450 if (disk->fops->revalidate_disk)
451 disk->fops->revalidate_disk(disk);
452 check_disk_size_change(disk, bdev);
453 bdev->bd_invalidated = 0;
454 if (!get_capacity(disk) || !(state = check_partition(disk, bdev)))
455 return 0;
456 if (IS_ERR(state)) {
457 /*
458 * I/O error reading the partition table. If any
459 * partition code tried to read beyond EOD, retry
460 * after unlocking native capacity.
461 */
462 if (PTR_ERR(state) == -ENOSPC) {
463 printk(KERN_WARNING "%s: partition table beyond EOD, ",
464 disk->disk_name);
465 if (disk_unlock_native_capacity(disk))
466 goto rescan;
467 }
468 return -EIO;
469 }
470 /*
471 * If any partition code tried to read beyond EOD, try
472 * unlocking native capacity even if partition table is
473 * successfully read as we could be missing some partitions.
474 */
475 if (state->access_beyond_eod) {
476 printk(KERN_WARNING
477 "%s: partition table partially beyond EOD, ",
478 disk->disk_name);
479 if (disk_unlock_native_capacity(disk))
480 goto rescan;
481 }
482
483 /* tell userspace that the media / partition table may have changed */
484 kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
485
486 /* Detect the highest partition number and preallocate
487 * disk->part_tbl. This is an optimization and not strictly
488 * necessary.
489 */
490 for (p = 1, highest = 0; p < state->limit; p++)
491 if (state->parts[p].size)
492 highest = p;
493
494 disk_expand_part_tbl(disk, highest);
495
496 /* add partitions */
497 for (p = 1; p < state->limit; p++) {
498 sector_t size, from;
499
500 size = state->parts[p].size;
501 if (!size)
502 continue;
503
504 from = state->parts[p].from;
505 if (from >= get_capacity(disk)) {
506 printk(KERN_WARNING
507 "%s: p%d start %llu is beyond EOD, ",
508 disk->disk_name, p, (unsigned long long) from);
509 if (disk_unlock_native_capacity(disk))
510 goto rescan;
511 continue;
512 }
513
514 if (from + size > get_capacity(disk)) {
515 printk(KERN_WARNING
516 "%s: p%d size %llu extends beyond EOD, ",
517 disk->disk_name, p, (unsigned long long) size);
518
519 if (disk_unlock_native_capacity(disk)) {
520 /* free state and restart */
521 goto rescan;
522 } else {
523 /*
524 * we can not ignore partitions of broken tables
525 * created by for example camera firmware, but
526 * we limit them to the end of the disk to avoid
527 * creating invalid block devices
528 */
529 size = get_capacity(disk) - from;
530 }
531 }
532
533 part = add_partition(disk, p, from, size,
534 state->parts[p].flags,
535 &state->parts[p].info);
536 if (IS_ERR(part)) {
537 printk(KERN_ERR " %s: p%d could not be added: %ld\n",
538 disk->disk_name, p, -PTR_ERR(part));
539 continue;
540 }
541 #ifdef CONFIG_BLK_DEV_MD
542 if (state->parts[p].flags & ADDPART_FLAG_RAID)
543 md_autodetect_dev(part_to_dev(part)->devt);
544 #endif
545 }
546 free_partitions(state);
547 return 0;
548 }
549
invalidate_partitions(struct gendisk * disk,struct block_device * bdev)550 int invalidate_partitions(struct gendisk *disk, struct block_device *bdev)
551 {
552 int res;
553
554 if (!bdev->bd_invalidated)
555 return 0;
556
557 res = drop_partitions(disk, bdev);
558 if (res)
559 return res;
560
561 set_capacity(disk, 0);
562 check_disk_size_change(disk, bdev);
563 bdev->bd_invalidated = 0;
564 /* tell userspace that the media / partition table may have changed */
565 kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
566
567 return 0;
568 }
569
read_pagecache_sector(struct block_device * bdev,sector_t n)570 static struct page *read_pagecache_sector(struct block_device *bdev, sector_t n)
571 {
572 struct address_space *mapping = bdev->bd_inode->i_mapping;
573
574 return read_mapping_page(mapping, (pgoff_t)(n >> (PAGE_SHIFT-9)),
575 NULL);
576 }
577
read_dev_sector(struct block_device * bdev,sector_t n,Sector * p)578 unsigned char *read_dev_sector(struct block_device *bdev, sector_t n, Sector *p)
579 {
580 struct page *page;
581
582 /* don't populate page cache for dax capable devices */
583 if (IS_DAX(bdev->bd_inode))
584 page = read_dax_sector(bdev, n);
585 else
586 page = read_pagecache_sector(bdev, n);
587
588 if (!IS_ERR(page)) {
589 if (PageError(page))
590 goto fail;
591 p->v = page;
592 return (unsigned char *)page_address(page) + ((n & ((1 << (PAGE_SHIFT - 9)) - 1)) << 9);
593 fail:
594 put_page(page);
595 }
596 p->v = NULL;
597 return NULL;
598 }
599
600 EXPORT_SYMBOL(read_dev_sector);
601