Revert "block: fix accounting bug on cross partition merges"
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / partitions / check.c
blob0a8b0ad0c7e25c803ca57a9a814934620ad80039
1 /*
2 * fs/partitions/check.c
4 * Code extracted from drivers/block/genhd.c
5 * Copyright (C) 1991-1998 Linus Torvalds
6 * Re-organised Feb 1998 Russell King
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.
13 * Added needed MAJORS for new pairs, {hdi,hdj}, {hdk,hdl}
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>
25 #include "check.h"
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"
42 #ifdef CONFIG_BLK_DEV_MD
43 extern void md_autodetect_dev(dev_t dev);
44 #endif
46 int warn_no_part = 1; /*This is ugly: should make genhd removable media aware*/
48 static int (*check_part[])(struct parsed_partitions *) = {
50 * Probe partition formats with tables at disk address 0
51 * that also have an ADFS boot block at 0xdc0.
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
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.
69 #ifdef CONFIG_ACORN_PARTITION_CUMANA
70 adfspart_check_CUMANA,
71 #endif
72 #ifdef CONFIG_ACORN_PARTITION_ADFS
73 adfspart_check_ADFS,
74 #endif
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
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).
125 char *disk_name(struct gendisk *hd, int partno, char *buf)
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);
134 return buf;
137 const char *bdevname(struct block_device *bdev, char *buf)
139 return disk_name(bdev->bd_disk, bdev->bd_part->partno, buf);
142 EXPORT_SYMBOL(bdevname);
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.
149 const char *__bdevname(dev_t dev, char *buffer)
151 scnprintf(buffer, BDEVNAME_SIZE, "unknown-block(%u,%u)",
152 MAJOR(dev), MINOR(dev));
153 return buffer;
156 EXPORT_SYMBOL(__bdevname);
158 static struct parsed_partitions *
159 check_partition(struct gendisk *hd, struct block_device *bdev)
161 struct parsed_partitions *state;
162 int i, res, err;
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;
172 state->pp_buf[0] = '\0';
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");
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.
189 err = res;
190 res = 0;
194 if (res > 0) {
195 printk(KERN_INFO "%s", state->pp_buf);
197 free_page((unsigned long)state->pp_buf);
198 return state;
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);
210 printk(KERN_INFO "%s", state->pp_buf);
212 free_page((unsigned long)state->pp_buf);
213 kfree(state);
214 return ERR_PTR(res);
217 static ssize_t part_partition_show(struct device *dev,
218 struct device_attribute *attr, char *buf)
220 struct hd_struct *p = dev_to_part(dev);
222 return sprintf(buf, "%d\n", p->partno);
225 static ssize_t part_start_show(struct device *dev,
226 struct device_attribute *attr, char *buf)
228 struct hd_struct *p = dev_to_part(dev);
230 return sprintf(buf, "%llu\n",(unsigned long long)p->start_sect);
233 ssize_t part_size_show(struct device *dev,
234 struct device_attribute *attr, char *buf)
236 struct hd_struct *p = dev_to_part(dev);
237 return sprintf(buf, "%llu\n",(unsigned long long)p->nr_sects);
240 ssize_t part_alignment_offset_show(struct device *dev,
241 struct device_attribute *attr, char *buf)
243 struct hd_struct *p = dev_to_part(dev);
244 return sprintf(buf, "%llu\n", (unsigned long long)p->alignment_offset);
247 ssize_t part_discard_alignment_show(struct device *dev,
248 struct device_attribute *attr, char *buf)
250 struct hd_struct *p = dev_to_part(dev);
251 return sprintf(buf, "%u\n", p->discard_alignment);
254 ssize_t part_stat_show(struct device *dev,
255 struct device_attribute *attr, char *buf)
257 struct hd_struct *p = dev_to_part(dev);
258 int cpu;
260 cpu = part_stat_lock();
261 part_round_stats(cpu, p);
262 part_stat_unlock();
263 return sprintf(buf,
264 "%8lu %8lu %8llu %8u "
265 "%8lu %8lu %8llu %8u "
266 "%8u %8u %8u"
267 "\n",
268 part_stat_read(p, ios[READ]),
269 part_stat_read(p, merges[READ]),
270 (unsigned long long)part_stat_read(p, sectors[READ]),
271 jiffies_to_msecs(part_stat_read(p, ticks[READ])),
272 part_stat_read(p, ios[WRITE]),
273 part_stat_read(p, merges[WRITE]),
274 (unsigned long long)part_stat_read(p, sectors[WRITE]),
275 jiffies_to_msecs(part_stat_read(p, ticks[WRITE])),
276 part_in_flight(p),
277 jiffies_to_msecs(part_stat_read(p, io_ticks)),
278 jiffies_to_msecs(part_stat_read(p, time_in_queue)));
281 ssize_t part_inflight_show(struct device *dev,
282 struct device_attribute *attr, char *buf)
284 struct hd_struct *p = dev_to_part(dev);
286 return sprintf(buf, "%8u %8u\n", p->in_flight[0], p->in_flight[1]);
289 #ifdef CONFIG_FAIL_MAKE_REQUEST
290 ssize_t part_fail_show(struct device *dev,
291 struct device_attribute *attr, char *buf)
293 struct hd_struct *p = dev_to_part(dev);
295 return sprintf(buf, "%d\n", p->make_it_fail);
298 ssize_t part_fail_store(struct device *dev,
299 struct device_attribute *attr,
300 const char *buf, size_t count)
302 struct hd_struct *p = dev_to_part(dev);
303 int i;
305 if (count > 0 && sscanf(buf, "%d", &i) > 0)
306 p->make_it_fail = (i == 0) ? 0 : 1;
308 return count;
310 #endif
312 static DEVICE_ATTR(partition, S_IRUGO, part_partition_show, NULL);
313 static DEVICE_ATTR(start, S_IRUGO, part_start_show, NULL);
314 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
315 static DEVICE_ATTR(alignment_offset, S_IRUGO, part_alignment_offset_show, NULL);
316 static DEVICE_ATTR(discard_alignment, S_IRUGO, part_discard_alignment_show,
317 NULL);
318 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
319 static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
320 #ifdef CONFIG_FAIL_MAKE_REQUEST
321 static struct device_attribute dev_attr_fail =
322 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
323 #endif
325 static struct attribute *part_attrs[] = {
326 &dev_attr_partition.attr,
327 &dev_attr_start.attr,
328 &dev_attr_size.attr,
329 &dev_attr_alignment_offset.attr,
330 &dev_attr_discard_alignment.attr,
331 &dev_attr_stat.attr,
332 &dev_attr_inflight.attr,
333 #ifdef CONFIG_FAIL_MAKE_REQUEST
334 &dev_attr_fail.attr,
335 #endif
336 NULL
339 static struct attribute_group part_attr_group = {
340 .attrs = part_attrs,
343 static const struct attribute_group *part_attr_groups[] = {
344 &part_attr_group,
345 #ifdef CONFIG_BLK_DEV_IO_TRACE
346 &blk_trace_attr_group,
347 #endif
348 NULL
351 static void part_release(struct device *dev)
353 struct hd_struct *p = dev_to_part(dev);
354 free_part_stats(p);
355 free_part_info(p);
356 kfree(p);
359 struct device_type part_type = {
360 .name = "partition",
361 .groups = part_attr_groups,
362 .release = part_release,
365 static void delete_partition_rcu_cb(struct rcu_head *head)
367 struct hd_struct *part = container_of(head, struct hd_struct, rcu_head);
369 part->start_sect = 0;
370 part->nr_sects = 0;
371 part_stat_set_all(part, 0);
372 put_device(part_to_dev(part));
375 void delete_partition(struct gendisk *disk, int partno)
377 struct disk_part_tbl *ptbl = disk->part_tbl;
378 struct hd_struct *part;
380 if (partno >= ptbl->len)
381 return;
383 part = ptbl->part[partno];
384 if (!part)
385 return;
387 blk_free_devt(part_devt(part));
388 rcu_assign_pointer(ptbl->part[partno], NULL);
389 rcu_assign_pointer(ptbl->last_lookup, NULL);
390 kobject_put(part->holder_dir);
391 device_del(part_to_dev(part));
393 call_rcu(&part->rcu_head, delete_partition_rcu_cb);
396 static ssize_t whole_disk_show(struct device *dev,
397 struct device_attribute *attr, char *buf)
399 return 0;
401 static DEVICE_ATTR(whole_disk, S_IRUSR | S_IRGRP | S_IROTH,
402 whole_disk_show, NULL);
404 struct hd_struct *add_partition(struct gendisk *disk, int partno,
405 sector_t start, sector_t len, int flags,
406 struct partition_meta_info *info)
408 struct hd_struct *p;
409 dev_t devt = MKDEV(0, 0);
410 struct device *ddev = disk_to_dev(disk);
411 struct device *pdev;
412 struct disk_part_tbl *ptbl;
413 const char *dname;
414 int err;
416 err = disk_expand_part_tbl(disk, partno);
417 if (err)
418 return ERR_PTR(err);
419 ptbl = disk->part_tbl;
421 if (ptbl->part[partno])
422 return ERR_PTR(-EBUSY);
424 p = kzalloc(sizeof(*p), GFP_KERNEL);
425 if (!p)
426 return ERR_PTR(-EBUSY);
428 if (!init_part_stats(p)) {
429 err = -ENOMEM;
430 goto out_free;
432 pdev = part_to_dev(p);
434 p->start_sect = start;
435 p->alignment_offset =
436 queue_limit_alignment_offset(&disk->queue->limits, start);
437 p->discard_alignment =
438 queue_limit_discard_alignment(&disk->queue->limits, start);
439 p->nr_sects = len;
440 p->partno = partno;
441 p->policy = get_disk_ro(disk);
443 if (info) {
444 struct partition_meta_info *pinfo = alloc_part_info(disk);
445 if (!pinfo)
446 goto out_free_stats;
447 memcpy(pinfo, info, sizeof(*info));
448 p->info = pinfo;
451 dname = dev_name(ddev);
452 if (isdigit(dname[strlen(dname) - 1]))
453 dev_set_name(pdev, "%sp%d", dname, partno);
454 else
455 dev_set_name(pdev, "%s%d", dname, partno);
457 device_initialize(pdev);
458 pdev->class = &block_class;
459 pdev->type = &part_type;
460 pdev->parent = ddev;
462 err = blk_alloc_devt(p, &devt);
463 if (err)
464 goto out_free_info;
465 pdev->devt = devt;
467 /* delay uevent until 'holders' subdir is created */
468 dev_set_uevent_suppress(pdev, 1);
469 err = device_add(pdev);
470 if (err)
471 goto out_put;
473 err = -ENOMEM;
474 p->holder_dir = kobject_create_and_add("holders", &pdev->kobj);
475 if (!p->holder_dir)
476 goto out_del;
478 dev_set_uevent_suppress(pdev, 0);
479 if (flags & ADDPART_FLAG_WHOLEDISK) {
480 err = device_create_file(pdev, &dev_attr_whole_disk);
481 if (err)
482 goto out_del;
485 /* everything is up and running, commence */
486 rcu_assign_pointer(ptbl->part[partno], p);
488 /* suppress uevent if the disk supresses it */
489 if (!dev_get_uevent_suppress(ddev))
490 kobject_uevent(&pdev->kobj, KOBJ_ADD);
492 return p;
494 out_free_info:
495 free_part_info(p);
496 out_free_stats:
497 free_part_stats(p);
498 out_free:
499 kfree(p);
500 return ERR_PTR(err);
501 out_del:
502 kobject_put(p->holder_dir);
503 device_del(pdev);
504 out_put:
505 put_device(pdev);
506 blk_free_devt(devt);
507 return ERR_PTR(err);
510 /* Not exported, helper to add_disk(). */
511 void register_disk(struct gendisk *disk)
513 struct device *ddev = disk_to_dev(disk);
514 struct block_device *bdev;
515 struct disk_part_iter piter;
516 struct hd_struct *part;
517 int err;
519 ddev->parent = disk->driverfs_dev;
521 dev_set_name(ddev, disk->disk_name);
523 /* delay uevents, until we scanned partition table */
524 dev_set_uevent_suppress(ddev, 1);
526 if (device_add(ddev))
527 return;
528 if (!sysfs_deprecated) {
529 err = sysfs_create_link(block_depr, &ddev->kobj,
530 kobject_name(&ddev->kobj));
531 if (err) {
532 device_del(ddev);
533 return;
536 disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj);
537 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
539 /* No minors to use for partitions */
540 if (!disk_partitionable(disk))
541 goto exit;
543 /* No such device (e.g., media were just removed) */
544 if (!get_capacity(disk))
545 goto exit;
547 bdev = bdget_disk(disk, 0);
548 if (!bdev)
549 goto exit;
551 bdev->bd_invalidated = 1;
552 err = blkdev_get(bdev, FMODE_READ);
553 if (err < 0)
554 goto exit;
555 blkdev_put(bdev, FMODE_READ);
557 exit:
558 /* announce disk after possible partitions are created */
559 dev_set_uevent_suppress(ddev, 0);
560 kobject_uevent(&ddev->kobj, KOBJ_ADD);
562 /* announce possible partitions */
563 disk_part_iter_init(&piter, disk, 0);
564 while ((part = disk_part_iter_next(&piter)))
565 kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD);
566 disk_part_iter_exit(&piter);
569 static bool disk_unlock_native_capacity(struct gendisk *disk)
571 const struct block_device_operations *bdops = disk->fops;
573 if (bdops->unlock_native_capacity &&
574 !(disk->flags & GENHD_FL_NATIVE_CAPACITY)) {
575 printk(KERN_CONT "enabling native capacity\n");
576 bdops->unlock_native_capacity(disk);
577 disk->flags |= GENHD_FL_NATIVE_CAPACITY;
578 return true;
579 } else {
580 printk(KERN_CONT "truncated\n");
581 return false;
585 int rescan_partitions(struct gendisk *disk, struct block_device *bdev)
587 struct parsed_partitions *state = NULL;
588 struct disk_part_iter piter;
589 struct hd_struct *part;
590 int p, highest, res;
591 rescan:
592 if (state && !IS_ERR(state)) {
593 kfree(state);
594 state = NULL;
597 if (bdev->bd_part_count)
598 return -EBUSY;
599 res = invalidate_partition(disk, 0);
600 if (res)
601 return res;
603 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
604 while ((part = disk_part_iter_next(&piter)))
605 delete_partition(disk, part->partno);
606 disk_part_iter_exit(&piter);
608 if (disk->fops->revalidate_disk)
609 disk->fops->revalidate_disk(disk);
610 check_disk_size_change(disk, bdev);
611 bdev->bd_invalidated = 0;
612 if (!get_capacity(disk) || !(state = check_partition(disk, bdev)))
613 return 0;
614 if (IS_ERR(state)) {
616 * I/O error reading the partition table. If any
617 * partition code tried to read beyond EOD, retry
618 * after unlocking native capacity.
620 if (PTR_ERR(state) == -ENOSPC) {
621 printk(KERN_WARNING "%s: partition table beyond EOD, ",
622 disk->disk_name);
623 if (disk_unlock_native_capacity(disk))
624 goto rescan;
626 return -EIO;
629 * If any partition code tried to read beyond EOD, try
630 * unlocking native capacity even if partition table is
631 * sucessfully read as we could be missing some partitions.
633 if (state->access_beyond_eod) {
634 printk(KERN_WARNING
635 "%s: partition table partially beyond EOD, ",
636 disk->disk_name);
637 if (disk_unlock_native_capacity(disk))
638 goto rescan;
641 /* tell userspace that the media / partition table may have changed */
642 kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
644 /* Detect the highest partition number and preallocate
645 * disk->part_tbl. This is an optimization and not strictly
646 * necessary.
648 for (p = 1, highest = 0; p < state->limit; p++)
649 if (state->parts[p].size)
650 highest = p;
652 disk_expand_part_tbl(disk, highest);
654 /* add partitions */
655 for (p = 1; p < state->limit; p++) {
656 sector_t size, from;
657 struct partition_meta_info *info = NULL;
659 size = state->parts[p].size;
660 if (!size)
661 continue;
663 from = state->parts[p].from;
664 if (from >= get_capacity(disk)) {
665 printk(KERN_WARNING
666 "%s: p%d start %llu is beyond EOD, ",
667 disk->disk_name, p, (unsigned long long) from);
668 if (disk_unlock_native_capacity(disk))
669 goto rescan;
670 continue;
673 if (from + size > get_capacity(disk)) {
674 printk(KERN_WARNING
675 "%s: p%d size %llu extends beyond EOD, ",
676 disk->disk_name, p, (unsigned long long) size);
678 if (disk_unlock_native_capacity(disk)) {
679 /* free state and restart */
680 goto rescan;
681 } else {
683 * we can not ignore partitions of broken tables
684 * created by for example camera firmware, but
685 * we limit them to the end of the disk to avoid
686 * creating invalid block devices
688 size = get_capacity(disk) - from;
692 if (state->parts[p].has_info)
693 info = &state->parts[p].info;
694 part = add_partition(disk, p, from, size,
695 state->parts[p].flags,
696 &state->parts[p].info);
697 if (IS_ERR(part)) {
698 printk(KERN_ERR " %s: p%d could not be added: %ld\n",
699 disk->disk_name, p, -PTR_ERR(part));
700 continue;
702 #ifdef CONFIG_BLK_DEV_MD
703 if (state->parts[p].flags & ADDPART_FLAG_RAID)
704 md_autodetect_dev(part_to_dev(part)->devt);
705 #endif
707 kfree(state);
708 return 0;
711 unsigned char *read_dev_sector(struct block_device *bdev, sector_t n, Sector *p)
713 struct address_space *mapping = bdev->bd_inode->i_mapping;
714 struct page *page;
716 page = read_mapping_page(mapping, (pgoff_t)(n >> (PAGE_CACHE_SHIFT-9)),
717 NULL);
718 if (!IS_ERR(page)) {
719 if (PageError(page))
720 goto fail;
721 p->v = page;
722 return (unsigned char *)page_address(page) + ((n & ((1 << (PAGE_CACHE_SHIFT - 9)) - 1)) << 9);
723 fail:
724 page_cache_release(page);
726 p->v = NULL;
727 return NULL;
730 EXPORT_SYMBOL(read_dev_sector);
732 void del_gendisk(struct gendisk *disk)
734 struct disk_part_iter piter;
735 struct hd_struct *part;
737 /* invalidate stuff */
738 disk_part_iter_init(&piter, disk,
739 DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
740 while ((part = disk_part_iter_next(&piter))) {
741 invalidate_partition(disk, part->partno);
742 delete_partition(disk, part->partno);
744 disk_part_iter_exit(&piter);
746 invalidate_partition(disk, 0);
747 blk_free_devt(disk_to_dev(disk)->devt);
748 set_capacity(disk, 0);
749 disk->flags &= ~GENHD_FL_UP;
750 unlink_gendisk(disk);
751 part_stat_set_all(&disk->part0, 0);
752 disk->part0.stamp = 0;
754 kobject_put(disk->part0.holder_dir);
755 kobject_put(disk->slave_dir);
756 disk->driverfs_dev = NULL;
757 if (!sysfs_deprecated)
758 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
759 device_del(disk_to_dev(disk));