]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - fs/partitions/check.c
Linux 2.6.38-rc2
[mirror_ubuntu-zesty-kernel.git] / fs / partitions / check.c
1 /*
2 * fs/partitions/check.c
3 *
4 * Code extracted from drivers/block/genhd.c
5 * Copyright (C) 1991-1998 Linus Torvalds
6 * Re-organised Feb 1998 Russell King
7 *
8 * We now have independent partition support from the
9 * block drivers, which allows all the partition code to
10 * be grouped in one location, and it to be mostly self
11 * contained.
12 *
13 * Added needed MAJORS for new pairs, {hdi,hdj}, {hdk,hdl}
14 */
15
16 #include <linux/init.h>
17 #include <linux/module.h>
18 #include <linux/fs.h>
19 #include <linux/slab.h>
20 #include <linux/kmod.h>
21 #include <linux/ctype.h>
22 #include <linux/genhd.h>
23 #include <linux/blktrace_api.h>
24
25 #include "check.h"
26
27 #include "acorn.h"
28 #include "amiga.h"
29 #include "atari.h"
30 #include "ldm.h"
31 #include "mac.h"
32 #include "msdos.h"
33 #include "osf.h"
34 #include "sgi.h"
35 #include "sun.h"
36 #include "ibm.h"
37 #include "ultrix.h"
38 #include "efi.h"
39 #include "karma.h"
40 #include "sysv68.h"
41
42 #ifdef CONFIG_BLK_DEV_MD
43 extern void md_autodetect_dev(dev_t dev);
44 #endif
45
46 int warn_no_part = 1; /*This is ugly: should make genhd removable media aware*/
47
48 static int (*check_part[])(struct parsed_partitions *) = {
49 /*
50 * Probe partition formats with tables at disk address 0
51 * that also have an ADFS boot block at 0xdc0.
52 */
53 #ifdef CONFIG_ACORN_PARTITION_ICS
54 adfspart_check_ICS,
55 #endif
56 #ifdef CONFIG_ACORN_PARTITION_POWERTEC
57 adfspart_check_POWERTEC,
58 #endif
59 #ifdef CONFIG_ACORN_PARTITION_EESOX
60 adfspart_check_EESOX,
61 #endif
62
63 /*
64 * Now move on to formats that only have partition info at
65 * disk address 0xdc0. Since these may also have stale
66 * PC/BIOS partition tables, they need to come before
67 * the msdos entry.
68 */
69 #ifdef CONFIG_ACORN_PARTITION_CUMANA
70 adfspart_check_CUMANA,
71 #endif
72 #ifdef CONFIG_ACORN_PARTITION_ADFS
73 adfspart_check_ADFS,
74 #endif
75
76 #ifdef CONFIG_EFI_PARTITION
77 efi_partition, /* this must come before msdos */
78 #endif
79 #ifdef CONFIG_SGI_PARTITION
80 sgi_partition,
81 #endif
82 #ifdef CONFIG_LDM_PARTITION
83 ldm_partition, /* this must come before msdos */
84 #endif
85 #ifdef CONFIG_MSDOS_PARTITION
86 msdos_partition,
87 #endif
88 #ifdef CONFIG_OSF_PARTITION
89 osf_partition,
90 #endif
91 #ifdef CONFIG_SUN_PARTITION
92 sun_partition,
93 #endif
94 #ifdef CONFIG_AMIGA_PARTITION
95 amiga_partition,
96 #endif
97 #ifdef CONFIG_ATARI_PARTITION
98 atari_partition,
99 #endif
100 #ifdef CONFIG_MAC_PARTITION
101 mac_partition,
102 #endif
103 #ifdef CONFIG_ULTRIX_PARTITION
104 ultrix_partition,
105 #endif
106 #ifdef CONFIG_IBM_PARTITION
107 ibm_partition,
108 #endif
109 #ifdef CONFIG_KARMA_PARTITION
110 karma_partition,
111 #endif
112 #ifdef CONFIG_SYSV68_PARTITION
113 sysv68_partition,
114 #endif
115 NULL
116 };
117
118 /*
119 * disk_name() is used by partition check code and the genhd driver.
120 * It formats the devicename of the indicated disk into
121 * the supplied buffer (of size at least 32), and returns
122 * a pointer to that same buffer (for convenience).
123 */
124
125 char *disk_name(struct gendisk *hd, int partno, char *buf)
126 {
127 if (!partno)
128 snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name);
129 else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1]))
130 snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno);
131 else
132 snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno);
133
134 return buf;
135 }
136
137 const char *bdevname(struct block_device *bdev, char *buf)
138 {
139 return disk_name(bdev->bd_disk, bdev->bd_part->partno, buf);
140 }
141
142 EXPORT_SYMBOL(bdevname);
143
144 /*
145 * There's very little reason to use this, you should really
146 * have a struct block_device just about everywhere and use
147 * bdevname() instead.
148 */
149 const char *__bdevname(dev_t dev, char *buffer)
150 {
151 scnprintf(buffer, BDEVNAME_SIZE, "unknown-block(%u,%u)",
152 MAJOR(dev), MINOR(dev));
153 return buffer;
154 }
155
156 EXPORT_SYMBOL(__bdevname);
157
158 static struct parsed_partitions *
159 check_partition(struct gendisk *hd, struct block_device *bdev)
160 {
161 struct parsed_partitions *state;
162 int i, res, err;
163
164 state = kzalloc(sizeof(struct parsed_partitions), GFP_KERNEL);
165 if (!state)
166 return NULL;
167 state->pp_buf = (char *)__get_free_page(GFP_KERNEL);
168 if (!state->pp_buf) {
169 kfree(state);
170 return NULL;
171 }
172 state->pp_buf[0] = '\0';
173
174 state->bdev = bdev;
175 disk_name(hd, 0, state->name);
176 snprintf(state->pp_buf, PAGE_SIZE, " %s:", state->name);
177 if (isdigit(state->name[strlen(state->name)-1]))
178 sprintf(state->name, "p");
179
180 state->limit = disk_max_parts(hd);
181 i = res = err = 0;
182 while (!res && check_part[i]) {
183 memset(&state->parts, 0, sizeof(state->parts));
184 res = check_part[i++](state);
185 if (res < 0) {
186 /* We have hit an I/O error which we don't report now.
187 * But record it, and let the others do their job.
188 */
189 err = res;
190 res = 0;
191 }
192
193 }
194 if (res > 0) {
195 printk(KERN_INFO "%s", state->pp_buf);
196
197 free_page((unsigned long)state->pp_buf);
198 return state;
199 }
200 if (state->access_beyond_eod)
201 err = -ENOSPC;
202 if (err)
203 /* The partition is unrecognized. So report I/O errors if there were any */
204 res = err;
205 if (!res)
206 strlcat(state->pp_buf, " unknown partition table\n", PAGE_SIZE);
207 else if (warn_no_part)
208 strlcat(state->pp_buf, " unable to read partition table\n", PAGE_SIZE);
209
210 printk(KERN_INFO "%s", state->pp_buf);
211
212 free_page((unsigned long)state->pp_buf);
213 kfree(state);
214 return ERR_PTR(res);
215 }
216
217 static ssize_t part_partition_show(struct device *dev,
218 struct device_attribute *attr, char *buf)
219 {
220 struct hd_struct *p = dev_to_part(dev);
221
222 return sprintf(buf, "%d\n", p->partno);
223 }
224
225 static ssize_t part_start_show(struct device *dev,
226 struct device_attribute *attr, char *buf)
227 {
228 struct hd_struct *p = dev_to_part(dev);
229
230 return sprintf(buf, "%llu\n",(unsigned long long)p->start_sect);
231 }
232
233 ssize_t part_size_show(struct device *dev,
234 struct device_attribute *attr, char *buf)
235 {
236 struct hd_struct *p = dev_to_part(dev);
237 return sprintf(buf, "%llu\n",(unsigned long long)p->nr_sects);
238 }
239
240 ssize_t part_ro_show(struct device *dev,
241 struct device_attribute *attr, char *buf)
242 {
243 struct hd_struct *p = dev_to_part(dev);
244 return sprintf(buf, "%d\n", p->policy ? 1 : 0);
245 }
246
247 ssize_t part_alignment_offset_show(struct device *dev,
248 struct device_attribute *attr, char *buf)
249 {
250 struct hd_struct *p = dev_to_part(dev);
251 return sprintf(buf, "%llu\n", (unsigned long long)p->alignment_offset);
252 }
253
254 ssize_t part_discard_alignment_show(struct device *dev,
255 struct device_attribute *attr, char *buf)
256 {
257 struct hd_struct *p = dev_to_part(dev);
258 return sprintf(buf, "%u\n", p->discard_alignment);
259 }
260
261 ssize_t part_stat_show(struct device *dev,
262 struct device_attribute *attr, char *buf)
263 {
264 struct hd_struct *p = dev_to_part(dev);
265 int cpu;
266
267 cpu = part_stat_lock();
268 part_round_stats(cpu, p);
269 part_stat_unlock();
270 return sprintf(buf,
271 "%8lu %8lu %8llu %8u "
272 "%8lu %8lu %8llu %8u "
273 "%8u %8u %8u"
274 "\n",
275 part_stat_read(p, ios[READ]),
276 part_stat_read(p, merges[READ]),
277 (unsigned long long)part_stat_read(p, sectors[READ]),
278 jiffies_to_msecs(part_stat_read(p, ticks[READ])),
279 part_stat_read(p, ios[WRITE]),
280 part_stat_read(p, merges[WRITE]),
281 (unsigned long long)part_stat_read(p, sectors[WRITE]),
282 jiffies_to_msecs(part_stat_read(p, ticks[WRITE])),
283 part_in_flight(p),
284 jiffies_to_msecs(part_stat_read(p, io_ticks)),
285 jiffies_to_msecs(part_stat_read(p, time_in_queue)));
286 }
287
288 ssize_t part_inflight_show(struct device *dev,
289 struct device_attribute *attr, char *buf)
290 {
291 struct hd_struct *p = dev_to_part(dev);
292
293 return sprintf(buf, "%8u %8u\n", p->in_flight[0], p->in_flight[1]);
294 }
295
296 #ifdef CONFIG_FAIL_MAKE_REQUEST
297 ssize_t part_fail_show(struct device *dev,
298 struct device_attribute *attr, char *buf)
299 {
300 struct hd_struct *p = dev_to_part(dev);
301
302 return sprintf(buf, "%d\n", p->make_it_fail);
303 }
304
305 ssize_t part_fail_store(struct device *dev,
306 struct device_attribute *attr,
307 const char *buf, size_t count)
308 {
309 struct hd_struct *p = dev_to_part(dev);
310 int i;
311
312 if (count > 0 && sscanf(buf, "%d", &i) > 0)
313 p->make_it_fail = (i == 0) ? 0 : 1;
314
315 return count;
316 }
317 #endif
318
319 static DEVICE_ATTR(partition, S_IRUGO, part_partition_show, NULL);
320 static DEVICE_ATTR(start, S_IRUGO, part_start_show, NULL);
321 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
322 static DEVICE_ATTR(ro, S_IRUGO, part_ro_show, NULL);
323 static DEVICE_ATTR(alignment_offset, S_IRUGO, part_alignment_offset_show, NULL);
324 static DEVICE_ATTR(discard_alignment, S_IRUGO, part_discard_alignment_show,
325 NULL);
326 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
327 static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
328 #ifdef CONFIG_FAIL_MAKE_REQUEST
329 static struct device_attribute dev_attr_fail =
330 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
331 #endif
332
333 static struct attribute *part_attrs[] = {
334 &dev_attr_partition.attr,
335 &dev_attr_start.attr,
336 &dev_attr_size.attr,
337 &dev_attr_ro.attr,
338 &dev_attr_alignment_offset.attr,
339 &dev_attr_discard_alignment.attr,
340 &dev_attr_stat.attr,
341 &dev_attr_inflight.attr,
342 #ifdef CONFIG_FAIL_MAKE_REQUEST
343 &dev_attr_fail.attr,
344 #endif
345 NULL
346 };
347
348 static struct attribute_group part_attr_group = {
349 .attrs = part_attrs,
350 };
351
352 static const struct attribute_group *part_attr_groups[] = {
353 &part_attr_group,
354 #ifdef CONFIG_BLK_DEV_IO_TRACE
355 &blk_trace_attr_group,
356 #endif
357 NULL
358 };
359
360 static void part_release(struct device *dev)
361 {
362 struct hd_struct *p = dev_to_part(dev);
363 free_part_stats(p);
364 free_part_info(p);
365 kfree(p);
366 }
367
368 struct device_type part_type = {
369 .name = "partition",
370 .groups = part_attr_groups,
371 .release = part_release,
372 };
373
374 static void delete_partition_rcu_cb(struct rcu_head *head)
375 {
376 struct hd_struct *part = container_of(head, struct hd_struct, rcu_head);
377
378 part->start_sect = 0;
379 part->nr_sects = 0;
380 part_stat_set_all(part, 0);
381 put_device(part_to_dev(part));
382 }
383
384 void __delete_partition(struct hd_struct *part)
385 {
386 call_rcu(&part->rcu_head, delete_partition_rcu_cb);
387 }
388
389 void delete_partition(struct gendisk *disk, int partno)
390 {
391 struct disk_part_tbl *ptbl = disk->part_tbl;
392 struct hd_struct *part;
393
394 if (partno >= ptbl->len)
395 return;
396
397 part = ptbl->part[partno];
398 if (!part)
399 return;
400
401 blk_free_devt(part_devt(part));
402 rcu_assign_pointer(ptbl->part[partno], NULL);
403 rcu_assign_pointer(ptbl->last_lookup, NULL);
404 kobject_put(part->holder_dir);
405 device_del(part_to_dev(part));
406
407 hd_struct_put(part);
408 }
409
410 static ssize_t whole_disk_show(struct device *dev,
411 struct device_attribute *attr, char *buf)
412 {
413 return 0;
414 }
415 static DEVICE_ATTR(whole_disk, S_IRUSR | S_IRGRP | S_IROTH,
416 whole_disk_show, NULL);
417
418 struct hd_struct *add_partition(struct gendisk *disk, int partno,
419 sector_t start, sector_t len, int flags,
420 struct partition_meta_info *info)
421 {
422 struct hd_struct *p;
423 dev_t devt = MKDEV(0, 0);
424 struct device *ddev = disk_to_dev(disk);
425 struct device *pdev;
426 struct disk_part_tbl *ptbl;
427 const char *dname;
428 int err;
429
430 err = disk_expand_part_tbl(disk, partno);
431 if (err)
432 return ERR_PTR(err);
433 ptbl = disk->part_tbl;
434
435 if (ptbl->part[partno])
436 return ERR_PTR(-EBUSY);
437
438 p = kzalloc(sizeof(*p), GFP_KERNEL);
439 if (!p)
440 return ERR_PTR(-EBUSY);
441
442 if (!init_part_stats(p)) {
443 err = -ENOMEM;
444 goto out_free;
445 }
446 pdev = part_to_dev(p);
447
448 p->start_sect = start;
449 p->alignment_offset =
450 queue_limit_alignment_offset(&disk->queue->limits, start);
451 p->discard_alignment =
452 queue_limit_discard_alignment(&disk->queue->limits, start);
453 p->nr_sects = len;
454 p->partno = partno;
455 p->policy = get_disk_ro(disk);
456
457 if (info) {
458 struct partition_meta_info *pinfo = alloc_part_info(disk);
459 if (!pinfo)
460 goto out_free_stats;
461 memcpy(pinfo, info, sizeof(*info));
462 p->info = pinfo;
463 }
464
465 dname = dev_name(ddev);
466 if (isdigit(dname[strlen(dname) - 1]))
467 dev_set_name(pdev, "%sp%d", dname, partno);
468 else
469 dev_set_name(pdev, "%s%d", dname, partno);
470
471 device_initialize(pdev);
472 pdev->class = &block_class;
473 pdev->type = &part_type;
474 pdev->parent = ddev;
475
476 err = blk_alloc_devt(p, &devt);
477 if (err)
478 goto out_free_info;
479 pdev->devt = devt;
480
481 /* delay uevent until 'holders' subdir is created */
482 dev_set_uevent_suppress(pdev, 1);
483 err = device_add(pdev);
484 if (err)
485 goto out_put;
486
487 err = -ENOMEM;
488 p->holder_dir = kobject_create_and_add("holders", &pdev->kobj);
489 if (!p->holder_dir)
490 goto out_del;
491
492 dev_set_uevent_suppress(pdev, 0);
493 if (flags & ADDPART_FLAG_WHOLEDISK) {
494 err = device_create_file(pdev, &dev_attr_whole_disk);
495 if (err)
496 goto out_del;
497 }
498
499 /* everything is up and running, commence */
500 rcu_assign_pointer(ptbl->part[partno], p);
501
502 /* suppress uevent if the disk supresses it */
503 if (!dev_get_uevent_suppress(ddev))
504 kobject_uevent(&pdev->kobj, KOBJ_ADD);
505
506 hd_ref_init(p);
507 return p;
508
509 out_free_info:
510 free_part_info(p);
511 out_free_stats:
512 free_part_stats(p);
513 out_free:
514 kfree(p);
515 return ERR_PTR(err);
516 out_del:
517 kobject_put(p->holder_dir);
518 device_del(pdev);
519 out_put:
520 put_device(pdev);
521 blk_free_devt(devt);
522 return ERR_PTR(err);
523 }
524
525 static bool disk_unlock_native_capacity(struct gendisk *disk)
526 {
527 const struct block_device_operations *bdops = disk->fops;
528
529 if (bdops->unlock_native_capacity &&
530 !(disk->flags & GENHD_FL_NATIVE_CAPACITY)) {
531 printk(KERN_CONT "enabling native capacity\n");
532 bdops->unlock_native_capacity(disk);
533 disk->flags |= GENHD_FL_NATIVE_CAPACITY;
534 return true;
535 } else {
536 printk(KERN_CONT "truncated\n");
537 return false;
538 }
539 }
540
541 int rescan_partitions(struct gendisk *disk, struct block_device *bdev)
542 {
543 struct parsed_partitions *state = NULL;
544 struct disk_part_iter piter;
545 struct hd_struct *part;
546 int p, highest, res;
547 rescan:
548 if (state && !IS_ERR(state)) {
549 kfree(state);
550 state = NULL;
551 }
552
553 if (bdev->bd_part_count)
554 return -EBUSY;
555 res = invalidate_partition(disk, 0);
556 if (res)
557 return res;
558
559 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
560 while ((part = disk_part_iter_next(&piter)))
561 delete_partition(disk, part->partno);
562 disk_part_iter_exit(&piter);
563
564 if (disk->fops->revalidate_disk)
565 disk->fops->revalidate_disk(disk);
566 check_disk_size_change(disk, bdev);
567 bdev->bd_invalidated = 0;
568 if (!get_capacity(disk) || !(state = check_partition(disk, bdev)))
569 return 0;
570 if (IS_ERR(state)) {
571 /*
572 * I/O error reading the partition table. If any
573 * partition code tried to read beyond EOD, retry
574 * after unlocking native capacity.
575 */
576 if (PTR_ERR(state) == -ENOSPC) {
577 printk(KERN_WARNING "%s: partition table beyond EOD, ",
578 disk->disk_name);
579 if (disk_unlock_native_capacity(disk))
580 goto rescan;
581 }
582 return -EIO;
583 }
584 /*
585 * If any partition code tried to read beyond EOD, try
586 * unlocking native capacity even if partition table is
587 * sucessfully read as we could be missing some partitions.
588 */
589 if (state->access_beyond_eod) {
590 printk(KERN_WARNING
591 "%s: partition table partially beyond EOD, ",
592 disk->disk_name);
593 if (disk_unlock_native_capacity(disk))
594 goto rescan;
595 }
596
597 /* tell userspace that the media / partition table may have changed */
598 kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
599
600 /* Detect the highest partition number and preallocate
601 * disk->part_tbl. This is an optimization and not strictly
602 * necessary.
603 */
604 for (p = 1, highest = 0; p < state->limit; p++)
605 if (state->parts[p].size)
606 highest = p;
607
608 disk_expand_part_tbl(disk, highest);
609
610 /* add partitions */
611 for (p = 1; p < state->limit; p++) {
612 sector_t size, from;
613 struct partition_meta_info *info = NULL;
614
615 size = state->parts[p].size;
616 if (!size)
617 continue;
618
619 from = state->parts[p].from;
620 if (from >= get_capacity(disk)) {
621 printk(KERN_WARNING
622 "%s: p%d start %llu is beyond EOD, ",
623 disk->disk_name, p, (unsigned long long) from);
624 if (disk_unlock_native_capacity(disk))
625 goto rescan;
626 continue;
627 }
628
629 if (from + size > get_capacity(disk)) {
630 printk(KERN_WARNING
631 "%s: p%d size %llu extends beyond EOD, ",
632 disk->disk_name, p, (unsigned long long) size);
633
634 if (disk_unlock_native_capacity(disk)) {
635 /* free state and restart */
636 goto rescan;
637 } else {
638 /*
639 * we can not ignore partitions of broken tables
640 * created by for example camera firmware, but
641 * we limit them to the end of the disk to avoid
642 * creating invalid block devices
643 */
644 size = get_capacity(disk) - from;
645 }
646 }
647
648 if (state->parts[p].has_info)
649 info = &state->parts[p].info;
650 part = add_partition(disk, p, from, size,
651 state->parts[p].flags,
652 &state->parts[p].info);
653 if (IS_ERR(part)) {
654 printk(KERN_ERR " %s: p%d could not be added: %ld\n",
655 disk->disk_name, p, -PTR_ERR(part));
656 continue;
657 }
658 #ifdef CONFIG_BLK_DEV_MD
659 if (state->parts[p].flags & ADDPART_FLAG_RAID)
660 md_autodetect_dev(part_to_dev(part)->devt);
661 #endif
662 }
663 kfree(state);
664 return 0;
665 }
666
667 unsigned char *read_dev_sector(struct block_device *bdev, sector_t n, Sector *p)
668 {
669 struct address_space *mapping = bdev->bd_inode->i_mapping;
670 struct page *page;
671
672 page = read_mapping_page(mapping, (pgoff_t)(n >> (PAGE_CACHE_SHIFT-9)),
673 NULL);
674 if (!IS_ERR(page)) {
675 if (PageError(page))
676 goto fail;
677 p->v = page;
678 return (unsigned char *)page_address(page) + ((n & ((1 << (PAGE_CACHE_SHIFT - 9)) - 1)) << 9);
679 fail:
680 page_cache_release(page);
681 }
682 p->v = NULL;
683 return NULL;
684 }
685
686 EXPORT_SYMBOL(read_dev_sector);