spi: tegra: moving to clk prepare APIs
[linux-2.6.git] / block / partition-generic.c
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/blktrace_api.h>
20
21 #include "partitions/check.h"
22
23 #ifdef CONFIG_BLK_DEV_MD
24 extern void md_autodetect_dev(dev_t dev);
25 #endif
26  
27 /*
28  * disk_name() is used by partition check code and the genhd driver.
29  * It formats the devicename of the indicated disk into
30  * the supplied buffer (of size at least 32), and returns
31  * a pointer to that same buffer (for convenience).
32  */
33
34 char *disk_name(struct gendisk *hd, int partno, char *buf)
35 {
36         if (!partno)
37                 snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name);
38         else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1]))
39                 snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno);
40         else
41                 snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno);
42
43         return buf;
44 }
45
46 const char *bdevname(struct block_device *bdev, char *buf)
47 {
48         return disk_name(bdev->bd_disk, bdev->bd_part->partno, buf);
49 }
50
51 EXPORT_SYMBOL(bdevname);
52
53 /*
54  * There's very little reason to use this, you should really
55  * have a struct block_device just about everywhere and use
56  * bdevname() instead.
57  */
58 const char *__bdevname(dev_t dev, char *buffer)
59 {
60         scnprintf(buffer, BDEVNAME_SIZE, "unknown-block(%u,%u)",
61                                 MAJOR(dev), MINOR(dev));
62         return buffer;
63 }
64
65 EXPORT_SYMBOL(__bdevname);
66
67 static ssize_t part_partition_show(struct device *dev,
68                                    struct device_attribute *attr, char *buf)
69 {
70         struct hd_struct *p = dev_to_part(dev);
71
72         return sprintf(buf, "%d\n", p->partno);
73 }
74
75 static ssize_t part_start_show(struct device *dev,
76                                struct device_attribute *attr, char *buf)
77 {
78         struct hd_struct *p = dev_to_part(dev);
79
80         return sprintf(buf, "%llu\n",(unsigned long long)p->start_sect);
81 }
82
83 ssize_t part_size_show(struct device *dev,
84                        struct device_attribute *attr, char *buf)
85 {
86         struct hd_struct *p = dev_to_part(dev);
87         return sprintf(buf, "%llu\n",(unsigned long long)p->nr_sects);
88 }
89
90 static ssize_t part_ro_show(struct device *dev,
91                             struct device_attribute *attr, char *buf)
92 {
93         struct hd_struct *p = dev_to_part(dev);
94         return sprintf(buf, "%d\n", p->policy ? 1 : 0);
95 }
96
97 static ssize_t part_alignment_offset_show(struct device *dev,
98                                           struct device_attribute *attr, char *buf)
99 {
100         struct hd_struct *p = dev_to_part(dev);
101         return sprintf(buf, "%llu\n", (unsigned long long)p->alignment_offset);
102 }
103
104 static ssize_t part_discard_alignment_show(struct device *dev,
105                                            struct device_attribute *attr, char *buf)
106 {
107         struct hd_struct *p = dev_to_part(dev);
108         return sprintf(buf, "%u\n", p->discard_alignment);
109 }
110
111 ssize_t part_stat_show(struct device *dev,
112                        struct device_attribute *attr, char *buf)
113 {
114         struct hd_struct *p = dev_to_part(dev);
115         int cpu;
116
117         cpu = part_stat_lock();
118         part_round_stats(cpu, p);
119         part_stat_unlock();
120         return sprintf(buf,
121                 "%8lu %8lu %8llu %8u "
122                 "%8lu %8lu %8llu %8u "
123                 "%8u %8u %8u"
124                 "\n",
125                 part_stat_read(p, ios[READ]),
126                 part_stat_read(p, merges[READ]),
127                 (unsigned long long)part_stat_read(p, sectors[READ]),
128                 jiffies_to_msecs(part_stat_read(p, ticks[READ])),
129                 part_stat_read(p, ios[WRITE]),
130                 part_stat_read(p, merges[WRITE]),
131                 (unsigned long long)part_stat_read(p, sectors[WRITE]),
132                 jiffies_to_msecs(part_stat_read(p, ticks[WRITE])),
133                 part_in_flight(p),
134                 jiffies_to_msecs(part_stat_read(p, io_ticks)),
135                 jiffies_to_msecs(part_stat_read(p, time_in_queue)));
136 }
137
138 ssize_t part_inflight_show(struct device *dev,
139                         struct device_attribute *attr, char *buf)
140 {
141         struct hd_struct *p = dev_to_part(dev);
142
143         return sprintf(buf, "%8u %8u\n", atomic_read(&p->in_flight[0]),
144                 atomic_read(&p->in_flight[1]));
145 }
146
147 #ifdef CONFIG_FAIL_MAKE_REQUEST
148 ssize_t part_fail_show(struct device *dev,
149                        struct device_attribute *attr, char *buf)
150 {
151         struct hd_struct *p = dev_to_part(dev);
152
153         return sprintf(buf, "%d\n", p->make_it_fail);
154 }
155
156 ssize_t part_fail_store(struct device *dev,
157                         struct device_attribute *attr,
158                         const char *buf, size_t count)
159 {
160         struct hd_struct *p = dev_to_part(dev);
161         int i;
162
163         if (count > 0 && sscanf(buf, "%d", &i) > 0)
164                 p->make_it_fail = (i == 0) ? 0 : 1;
165
166         return count;
167 }
168 #endif
169
170 static DEVICE_ATTR(partition, S_IRUGO, part_partition_show, NULL);
171 static DEVICE_ATTR(start, S_IRUGO, part_start_show, NULL);
172 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
173 static DEVICE_ATTR(ro, S_IRUGO, part_ro_show, NULL);
174 static DEVICE_ATTR(alignment_offset, S_IRUGO, part_alignment_offset_show, NULL);
175 static DEVICE_ATTR(discard_alignment, S_IRUGO, part_discard_alignment_show,
176                    NULL);
177 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
178 static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
179 #ifdef CONFIG_FAIL_MAKE_REQUEST
180 static struct device_attribute dev_attr_fail =
181         __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
182 #endif
183
184 static struct attribute *part_attrs[] = {
185         &dev_attr_partition.attr,
186         &dev_attr_start.attr,
187         &dev_attr_size.attr,
188         &dev_attr_ro.attr,
189         &dev_attr_alignment_offset.attr,
190         &dev_attr_discard_alignment.attr,
191         &dev_attr_stat.attr,
192         &dev_attr_inflight.attr,
193 #ifdef CONFIG_FAIL_MAKE_REQUEST
194         &dev_attr_fail.attr,
195 #endif
196         NULL
197 };
198
199 static struct attribute_group part_attr_group = {
200         .attrs = part_attrs,
201 };
202
203 static const struct attribute_group *part_attr_groups[] = {
204         &part_attr_group,
205 #ifdef CONFIG_BLK_DEV_IO_TRACE
206         &blk_trace_attr_group,
207 #endif
208         NULL
209 };
210
211 static void part_release(struct device *dev)
212 {
213         struct hd_struct *p = dev_to_part(dev);
214         free_part_stats(p);
215         free_part_info(p);
216         kfree(p);
217 }
218
219 static int part_uevent(struct device *dev, struct kobj_uevent_env *env)
220 {
221         struct hd_struct *part = dev_to_part(dev);
222
223         add_uevent_var(env, "PARTN=%u", part->partno);
224         if (part->info && part->info->volname[0])
225                 add_uevent_var(env, "PARTNAME=%s", part->info->volname);
226         return 0;
227 }
228
229 struct device_type part_type = {
230         .name           = "partition",
231         .groups         = part_attr_groups,
232         .release        = part_release,
233         .uevent         = part_uevent,
234 };
235
236 static void delete_partition_rcu_cb(struct rcu_head *head)
237 {
238         struct hd_struct *part = container_of(head, struct hd_struct, rcu_head);
239
240         part->start_sect = 0;
241         part->nr_sects = 0;
242         part_stat_set_all(part, 0);
243         put_device(part_to_dev(part));
244 }
245
246 void __delete_partition(struct hd_struct *part)
247 {
248         call_rcu(&part->rcu_head, delete_partition_rcu_cb);
249 }
250
251 void delete_partition(struct gendisk *disk, int partno)
252 {
253         struct disk_part_tbl *ptbl = disk->part_tbl;
254         struct hd_struct *part;
255
256         if (partno >= ptbl->len)
257                 return;
258
259         part = ptbl->part[partno];
260         if (!part)
261                 return;
262
263         blk_free_devt(part_devt(part));
264         rcu_assign_pointer(ptbl->part[partno], NULL);
265         rcu_assign_pointer(ptbl->last_lookup, NULL);
266         kobject_put(part->holder_dir);
267         device_del(part_to_dev(part));
268
269         hd_struct_put(part);
270 }
271
272 static ssize_t whole_disk_show(struct device *dev,
273                                struct device_attribute *attr, char *buf)
274 {
275         return 0;
276 }
277 static DEVICE_ATTR(whole_disk, S_IRUSR | S_IRGRP | S_IROTH,
278                    whole_disk_show, NULL);
279
280 struct hd_struct *add_partition(struct gendisk *disk, int partno,
281                                 sector_t start, sector_t len, int flags,
282                                 struct partition_meta_info *info)
283 {
284         struct hd_struct *p;
285         dev_t devt = MKDEV(0, 0);
286         struct device *ddev = disk_to_dev(disk);
287         struct device *pdev;
288         struct disk_part_tbl *ptbl;
289         const char *dname;
290         int err;
291
292         err = disk_expand_part_tbl(disk, partno);
293         if (err)
294                 return ERR_PTR(err);
295         ptbl = disk->part_tbl;
296
297         if (ptbl->part[partno])
298                 return ERR_PTR(-EBUSY);
299
300         p = kzalloc(sizeof(*p), GFP_KERNEL);
301         if (!p)
302                 return ERR_PTR(-EBUSY);
303
304         if (!init_part_stats(p)) {
305                 err = -ENOMEM;
306                 goto out_free;
307         }
308         pdev = part_to_dev(p);
309
310         p->start_sect = start;
311         p->alignment_offset =
312                 queue_limit_alignment_offset(&disk->queue->limits, start);
313         p->discard_alignment =
314                 queue_limit_discard_alignment(&disk->queue->limits, start);
315         p->nr_sects = len;
316         p->partno = partno;
317         p->policy = get_disk_ro(disk);
318
319         if (info) {
320                 struct partition_meta_info *pinfo = alloc_part_info(disk);
321                 if (!pinfo)
322                         goto out_free_stats;
323                 memcpy(pinfo, info, sizeof(*info));
324                 p->info = pinfo;
325         }
326
327         dname = dev_name(ddev);
328         if (isdigit(dname[strlen(dname) - 1]))
329                 dev_set_name(pdev, "%sp%d", dname, partno);
330         else
331                 dev_set_name(pdev, "%s%d", dname, partno);
332
333         device_initialize(pdev);
334         pdev->class = &block_class;
335         pdev->type = &part_type;
336         pdev->parent = ddev;
337
338         err = blk_alloc_devt(p, &devt);
339         if (err)
340                 goto out_free_info;
341         pdev->devt = devt;
342
343         /* delay uevent until 'holders' subdir is created */
344         dev_set_uevent_suppress(pdev, 1);
345         err = device_add(pdev);
346         if (err)
347                 goto out_put;
348
349         err = -ENOMEM;
350         p->holder_dir = kobject_create_and_add("holders", &pdev->kobj);
351         if (!p->holder_dir)
352                 goto out_del;
353
354         dev_set_uevent_suppress(pdev, 0);
355         if (flags & ADDPART_FLAG_WHOLEDISK) {
356                 err = device_create_file(pdev, &dev_attr_whole_disk);
357                 if (err)
358                         goto out_del;
359         }
360
361         /* everything is up and running, commence */
362         rcu_assign_pointer(ptbl->part[partno], p);
363
364         /* suppress uevent if the disk suppresses it */
365         if (!dev_get_uevent_suppress(ddev))
366                 kobject_uevent(&pdev->kobj, KOBJ_ADD);
367
368         hd_ref_init(p);
369         return p;
370
371 out_free_info:
372         free_part_info(p);
373 out_free_stats:
374         free_part_stats(p);
375 out_free:
376         kfree(p);
377         return ERR_PTR(err);
378 out_del:
379         kobject_put(p->holder_dir);
380         device_del(pdev);
381 out_put:
382         put_device(pdev);
383         blk_free_devt(devt);
384         return ERR_PTR(err);
385 }
386
387 static bool disk_unlock_native_capacity(struct gendisk *disk)
388 {
389         const struct block_device_operations *bdops = disk->fops;
390
391         if (bdops->unlock_native_capacity &&
392             !(disk->flags & GENHD_FL_NATIVE_CAPACITY)) {
393                 printk(KERN_CONT "enabling native capacity\n");
394                 bdops->unlock_native_capacity(disk);
395                 disk->flags |= GENHD_FL_NATIVE_CAPACITY;
396                 return true;
397         } else {
398                 printk(KERN_CONT "truncated\n");
399                 return false;
400         }
401 }
402
403 static int drop_partitions(struct gendisk *disk, struct block_device *bdev)
404 {
405         struct disk_part_iter piter;
406         struct hd_struct *part;
407         int res;
408
409         if (bdev->bd_part_count)
410                 return -EBUSY;
411         res = invalidate_partition(disk, 0);
412         if (res)
413                 return res;
414
415         disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
416         while ((part = disk_part_iter_next(&piter)))
417                 delete_partition(disk, part->partno);
418         disk_part_iter_exit(&piter);
419
420         return 0;
421 }
422
423 int rescan_partitions(struct gendisk *disk, struct block_device *bdev)
424 {
425         struct parsed_partitions *state = NULL;
426         struct hd_struct *part;
427         int p, highest, res;
428 rescan:
429         if (state && !IS_ERR(state)) {
430                 kfree(state);
431                 state = NULL;
432         }
433
434         res = drop_partitions(disk, bdev);
435         if (res)
436                 return res;
437
438         if (disk->fops->revalidate_disk)
439                 disk->fops->revalidate_disk(disk);
440         check_disk_size_change(disk, bdev);
441         bdev->bd_invalidated = 0;
442         if (!get_capacity(disk) || !(state = check_partition(disk, bdev)))
443                 return 0;
444         if (IS_ERR(state)) {
445                 /*
446                  * I/O error reading the partition table.  If any
447                  * partition code tried to read beyond EOD, retry
448                  * after unlocking native capacity.
449                  */
450                 if (PTR_ERR(state) == -ENOSPC) {
451                         printk(KERN_WARNING "%s: partition table beyond EOD, ",
452                                disk->disk_name);
453                         if (disk_unlock_native_capacity(disk))
454                                 goto rescan;
455                 }
456                 return -EIO;
457         }
458         /*
459          * If any partition code tried to read beyond EOD, try
460          * unlocking native capacity even if partition table is
461          * successfully read as we could be missing some partitions.
462          */
463         if (state->access_beyond_eod) {
464                 printk(KERN_WARNING
465                        "%s: partition table partially beyond EOD, ",
466                        disk->disk_name);
467                 if (disk_unlock_native_capacity(disk))
468                         goto rescan;
469         }
470
471         /* tell userspace that the media / partition table may have changed */
472         kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
473
474         /* Detect the highest partition number and preallocate
475          * disk->part_tbl.  This is an optimization and not strictly
476          * necessary.
477          */
478         for (p = 1, highest = 0; p < state->limit; p++)
479                 if (state->parts[p].size)
480                         highest = p;
481
482         disk_expand_part_tbl(disk, highest);
483
484         /* add partitions */
485         for (p = 1; p < state->limit; p++) {
486                 sector_t size, from;
487                 struct partition_meta_info *info = NULL;
488
489                 size = state->parts[p].size;
490                 if (!size)
491                         continue;
492
493                 from = state->parts[p].from;
494                 if (from >= get_capacity(disk)) {
495                         printk(KERN_WARNING
496                                "%s: p%d start %llu is beyond EOD, ",
497                                disk->disk_name, p, (unsigned long long) from);
498                         if (disk_unlock_native_capacity(disk))
499                                 goto rescan;
500                         continue;
501                 }
502
503                 if (from + size > get_capacity(disk)) {
504                         printk(KERN_WARNING
505                                "%s: p%d size %llu extends beyond EOD, ",
506                                disk->disk_name, p, (unsigned long long) size);
507
508                         if (disk_unlock_native_capacity(disk)) {
509                                 /* free state and restart */
510                                 goto rescan;
511                         } else {
512                                 /*
513                                  * we can not ignore partitions of broken tables
514                                  * created by for example camera firmware, but
515                                  * we limit them to the end of the disk to avoid
516                                  * creating invalid block devices
517                                  */
518                                 size = get_capacity(disk) - from;
519                         }
520                 }
521
522                 if (state->parts[p].has_info)
523                         info = &state->parts[p].info;
524                 part = add_partition(disk, p, from, size,
525                                      state->parts[p].flags,
526                                      &state->parts[p].info);
527                 if (IS_ERR(part)) {
528                         printk(KERN_ERR " %s: p%d could not be added: %ld\n",
529                                disk->disk_name, p, -PTR_ERR(part));
530                         continue;
531                 }
532 #ifdef CONFIG_BLK_DEV_MD
533                 if (state->parts[p].flags & ADDPART_FLAG_RAID)
534                         md_autodetect_dev(part_to_dev(part)->devt);
535 #endif
536         }
537         kfree(state);
538         return 0;
539 }
540
541 int invalidate_partitions(struct gendisk *disk, struct block_device *bdev)
542 {
543         int res;
544
545         if (!bdev->bd_invalidated)
546                 return 0;
547
548         res = drop_partitions(disk, bdev);
549         if (res)
550                 return res;
551
552         set_capacity(disk, 0);
553         check_disk_size_change(disk, bdev);
554         bdev->bd_invalidated = 0;
555         /* tell userspace that the media / partition table may have changed */
556         kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
557
558         return 0;
559 }
560
561 unsigned char *read_dev_sector(struct block_device *bdev, sector_t n, Sector *p)
562 {
563         struct address_space *mapping = bdev->bd_inode->i_mapping;
564         struct page *page;
565
566         page = read_mapping_page(mapping, (pgoff_t)(n >> (PAGE_CACHE_SHIFT-9)),
567                                  NULL);
568         if (!IS_ERR(page)) {
569                 if (PageError(page))
570                         goto fail;
571                 p->v = page;
572                 return (unsigned char *)page_address(page) +  ((n & ((1 << (PAGE_CACHE_SHIFT - 9)) - 1)) << 9);
573 fail:
574                 page_cache_release(page);
575         }
576         p->v = NULL;
577         return NULL;
578 }
579
580 EXPORT_SYMBOL(read_dev_sector);