Merge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1...
[pandora-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 static 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 static 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 static 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", atomic_read(&p->in_flight[0]),
294                 atomic_read(&p->in_flight[1]));
295 }
296
297 #ifdef CONFIG_FAIL_MAKE_REQUEST
298 ssize_t part_fail_show(struct device *dev,
299                        struct device_attribute *attr, char *buf)
300 {
301         struct hd_struct *p = dev_to_part(dev);
302
303         return sprintf(buf, "%d\n", p->make_it_fail);
304 }
305
306 ssize_t part_fail_store(struct device *dev,
307                         struct device_attribute *attr,
308                         const char *buf, size_t count)
309 {
310         struct hd_struct *p = dev_to_part(dev);
311         int i;
312
313         if (count > 0 && sscanf(buf, "%d", &i) > 0)
314                 p->make_it_fail = (i == 0) ? 0 : 1;
315
316         return count;
317 }
318 #endif
319
320 static DEVICE_ATTR(partition, S_IRUGO, part_partition_show, NULL);
321 static DEVICE_ATTR(start, S_IRUGO, part_start_show, NULL);
322 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
323 static DEVICE_ATTR(ro, S_IRUGO, part_ro_show, NULL);
324 static DEVICE_ATTR(alignment_offset, S_IRUGO, part_alignment_offset_show, NULL);
325 static DEVICE_ATTR(discard_alignment, S_IRUGO, part_discard_alignment_show,
326                    NULL);
327 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
328 static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
329 #ifdef CONFIG_FAIL_MAKE_REQUEST
330 static struct device_attribute dev_attr_fail =
331         __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
332 #endif
333
334 static struct attribute *part_attrs[] = {
335         &dev_attr_partition.attr,
336         &dev_attr_start.attr,
337         &dev_attr_size.attr,
338         &dev_attr_ro.attr,
339         &dev_attr_alignment_offset.attr,
340         &dev_attr_discard_alignment.attr,
341         &dev_attr_stat.attr,
342         &dev_attr_inflight.attr,
343 #ifdef CONFIG_FAIL_MAKE_REQUEST
344         &dev_attr_fail.attr,
345 #endif
346         NULL
347 };
348
349 static struct attribute_group part_attr_group = {
350         .attrs = part_attrs,
351 };
352
353 static const struct attribute_group *part_attr_groups[] = {
354         &part_attr_group,
355 #ifdef CONFIG_BLK_DEV_IO_TRACE
356         &blk_trace_attr_group,
357 #endif
358         NULL
359 };
360
361 static void part_release(struct device *dev)
362 {
363         struct hd_struct *p = dev_to_part(dev);
364         free_part_stats(p);
365         free_part_info(p);
366         kfree(p);
367 }
368
369 struct device_type part_type = {
370         .name           = "partition",
371         .groups         = part_attr_groups,
372         .release        = part_release,
373 };
374
375 static void delete_partition_rcu_cb(struct rcu_head *head)
376 {
377         struct hd_struct *part = container_of(head, struct hd_struct, rcu_head);
378
379         part->start_sect = 0;
380         part->nr_sects = 0;
381         part_stat_set_all(part, 0);
382         put_device(part_to_dev(part));
383 }
384
385 void __delete_partition(struct hd_struct *part)
386 {
387         call_rcu(&part->rcu_head, delete_partition_rcu_cb);
388 }
389
390 void delete_partition(struct gendisk *disk, int partno)
391 {
392         struct disk_part_tbl *ptbl = disk->part_tbl;
393         struct hd_struct *part;
394
395         if (partno >= ptbl->len)
396                 return;
397
398         part = ptbl->part[partno];
399         if (!part)
400                 return;
401
402         blk_free_devt(part_devt(part));
403         rcu_assign_pointer(ptbl->part[partno], NULL);
404         rcu_assign_pointer(ptbl->last_lookup, NULL);
405         kobject_put(part->holder_dir);
406         device_del(part_to_dev(part));
407
408         hd_struct_put(part);
409 }
410
411 static ssize_t whole_disk_show(struct device *dev,
412                                struct device_attribute *attr, char *buf)
413 {
414         return 0;
415 }
416 static DEVICE_ATTR(whole_disk, S_IRUSR | S_IRGRP | S_IROTH,
417                    whole_disk_show, NULL);
418
419 struct hd_struct *add_partition(struct gendisk *disk, int partno,
420                                 sector_t start, sector_t len, int flags,
421                                 struct partition_meta_info *info)
422 {
423         struct hd_struct *p;
424         dev_t devt = MKDEV(0, 0);
425         struct device *ddev = disk_to_dev(disk);
426         struct device *pdev;
427         struct disk_part_tbl *ptbl;
428         const char *dname;
429         int err;
430
431         err = disk_expand_part_tbl(disk, partno);
432         if (err)
433                 return ERR_PTR(err);
434         ptbl = disk->part_tbl;
435
436         if (ptbl->part[partno])
437                 return ERR_PTR(-EBUSY);
438
439         p = kzalloc(sizeof(*p), GFP_KERNEL);
440         if (!p)
441                 return ERR_PTR(-EBUSY);
442
443         if (!init_part_stats(p)) {
444                 err = -ENOMEM;
445                 goto out_free;
446         }
447         pdev = part_to_dev(p);
448
449         p->start_sect = start;
450         p->alignment_offset =
451                 queue_limit_alignment_offset(&disk->queue->limits, start);
452         p->discard_alignment =
453                 queue_limit_discard_alignment(&disk->queue->limits, start);
454         p->nr_sects = len;
455         p->partno = partno;
456         p->policy = get_disk_ro(disk);
457
458         if (info) {
459                 struct partition_meta_info *pinfo = alloc_part_info(disk);
460                 if (!pinfo)
461                         goto out_free_stats;
462                 memcpy(pinfo, info, sizeof(*info));
463                 p->info = pinfo;
464         }
465
466         dname = dev_name(ddev);
467         if (isdigit(dname[strlen(dname) - 1]))
468                 dev_set_name(pdev, "%sp%d", dname, partno);
469         else
470                 dev_set_name(pdev, "%s%d", dname, partno);
471
472         device_initialize(pdev);
473         pdev->class = &block_class;
474         pdev->type = &part_type;
475         pdev->parent = ddev;
476
477         err = blk_alloc_devt(p, &devt);
478         if (err)
479                 goto out_free_info;
480         pdev->devt = devt;
481
482         /* delay uevent until 'holders' subdir is created */
483         dev_set_uevent_suppress(pdev, 1);
484         err = device_add(pdev);
485         if (err)
486                 goto out_put;
487
488         err = -ENOMEM;
489         p->holder_dir = kobject_create_and_add("holders", &pdev->kobj);
490         if (!p->holder_dir)
491                 goto out_del;
492
493         dev_set_uevent_suppress(pdev, 0);
494         if (flags & ADDPART_FLAG_WHOLEDISK) {
495                 err = device_create_file(pdev, &dev_attr_whole_disk);
496                 if (err)
497                         goto out_del;
498         }
499
500         /* everything is up and running, commence */
501         rcu_assign_pointer(ptbl->part[partno], p);
502
503         /* suppress uevent if the disk suppresses it */
504         if (!dev_get_uevent_suppress(ddev))
505                 kobject_uevent(&pdev->kobj, KOBJ_ADD);
506
507         hd_ref_init(p);
508         return p;
509
510 out_free_info:
511         free_part_info(p);
512 out_free_stats:
513         free_part_stats(p);
514 out_free:
515         kfree(p);
516         return ERR_PTR(err);
517 out_del:
518         kobject_put(p->holder_dir);
519         device_del(pdev);
520 out_put:
521         put_device(pdev);
522         blk_free_devt(devt);
523         return ERR_PTR(err);
524 }
525
526 static bool disk_unlock_native_capacity(struct gendisk *disk)
527 {
528         const struct block_device_operations *bdops = disk->fops;
529
530         if (bdops->unlock_native_capacity &&
531             !(disk->flags & GENHD_FL_NATIVE_CAPACITY)) {
532                 printk(KERN_CONT "enabling native capacity\n");
533                 bdops->unlock_native_capacity(disk);
534                 disk->flags |= GENHD_FL_NATIVE_CAPACITY;
535                 return true;
536         } else {
537                 printk(KERN_CONT "truncated\n");
538                 return false;
539         }
540 }
541
542 int rescan_partitions(struct gendisk *disk, struct block_device *bdev)
543 {
544         struct parsed_partitions *state = NULL;
545         struct disk_part_iter piter;
546         struct hd_struct *part;
547         int p, highest, res;
548 rescan:
549         if (state && !IS_ERR(state)) {
550                 kfree(state);
551                 state = NULL;
552         }
553
554         if (bdev->bd_part_count)
555                 return -EBUSY;
556         res = invalidate_partition(disk, 0);
557         if (res)
558                 return res;
559
560         disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
561         while ((part = disk_part_iter_next(&piter)))
562                 delete_partition(disk, part->partno);
563         disk_part_iter_exit(&piter);
564
565         if (disk->fops->revalidate_disk)
566                 disk->fops->revalidate_disk(disk);
567         check_disk_size_change(disk, bdev);
568         bdev->bd_invalidated = 0;
569         if (!get_capacity(disk) || !(state = check_partition(disk, bdev)))
570                 return 0;
571         if (IS_ERR(state)) {
572                 /*
573                  * I/O error reading the partition table.  If any
574                  * partition code tried to read beyond EOD, retry
575                  * after unlocking native capacity.
576                  */
577                 if (PTR_ERR(state) == -ENOSPC) {
578                         printk(KERN_WARNING "%s: partition table beyond EOD, ",
579                                disk->disk_name);
580                         if (disk_unlock_native_capacity(disk))
581                                 goto rescan;
582                 }
583                 return -EIO;
584         }
585         /*
586          * If any partition code tried to read beyond EOD, try
587          * unlocking native capacity even if partition table is
588          * successfully read as we could be missing some partitions.
589          */
590         if (state->access_beyond_eod) {
591                 printk(KERN_WARNING
592                        "%s: partition table partially beyond EOD, ",
593                        disk->disk_name);
594                 if (disk_unlock_native_capacity(disk))
595                         goto rescan;
596         }
597
598         /* tell userspace that the media / partition table may have changed */
599         kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
600
601         /* Detect the highest partition number and preallocate
602          * disk->part_tbl.  This is an optimization and not strictly
603          * necessary.
604          */
605         for (p = 1, highest = 0; p < state->limit; p++)
606                 if (state->parts[p].size)
607                         highest = p;
608
609         disk_expand_part_tbl(disk, highest);
610
611         /* add partitions */
612         for (p = 1; p < state->limit; p++) {
613                 sector_t size, from;
614                 struct partition_meta_info *info = NULL;
615
616                 size = state->parts[p].size;
617                 if (!size)
618                         continue;
619
620                 from = state->parts[p].from;
621                 if (from >= get_capacity(disk)) {
622                         printk(KERN_WARNING
623                                "%s: p%d start %llu is beyond EOD, ",
624                                disk->disk_name, p, (unsigned long long) from);
625                         if (disk_unlock_native_capacity(disk))
626                                 goto rescan;
627                         continue;
628                 }
629
630                 if (from + size > get_capacity(disk)) {
631                         printk(KERN_WARNING
632                                "%s: p%d size %llu extends beyond EOD, ",
633                                disk->disk_name, p, (unsigned long long) size);
634
635                         if (disk_unlock_native_capacity(disk)) {
636                                 /* free state and restart */
637                                 goto rescan;
638                         } else {
639                                 /*
640                                  * we can not ignore partitions of broken tables
641                                  * created by for example camera firmware, but
642                                  * we limit them to the end of the disk to avoid
643                                  * creating invalid block devices
644                                  */
645                                 size = get_capacity(disk) - from;
646                         }
647                 }
648
649                 if (state->parts[p].has_info)
650                         info = &state->parts[p].info;
651                 part = add_partition(disk, p, from, size,
652                                      state->parts[p].flags,
653                                      &state->parts[p].info);
654                 if (IS_ERR(part)) {
655                         printk(KERN_ERR " %s: p%d could not be added: %ld\n",
656                                disk->disk_name, p, -PTR_ERR(part));
657                         continue;
658                 }
659 #ifdef CONFIG_BLK_DEV_MD
660                 if (state->parts[p].flags & ADDPART_FLAG_RAID)
661                         md_autodetect_dev(part_to_dev(part)->devt);
662 #endif
663         }
664         kfree(state);
665         return 0;
666 }
667
668 unsigned char *read_dev_sector(struct block_device *bdev, sector_t n, Sector *p)
669 {
670         struct address_space *mapping = bdev->bd_inode->i_mapping;
671         struct page *page;
672
673         page = read_mapping_page(mapping, (pgoff_t)(n >> (PAGE_CACHE_SHIFT-9)),
674                                  NULL);
675         if (!IS_ERR(page)) {
676                 if (PageError(page))
677                         goto fail;
678                 p->v = page;
679                 return (unsigned char *)page_address(page) +  ((n & ((1 << (PAGE_CACHE_SHIFT - 9)) - 1)) << 9);
680 fail:
681                 page_cache_release(page);
682         }
683         p->v = NULL;
684         return NULL;
685 }
686
687 EXPORT_SYMBOL(read_dev_sector);