blob: d42840cc0d1dfcd0fad18fca25041c04ac69661a [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * gendisk handling
3 */
4
5#include <linux/config.h>
6#include <linux/module.h>
7#include <linux/fs.h>
8#include <linux/genhd.h>
9#include <linux/kernel.h>
10#include <linux/blkdev.h>
11#include <linux/init.h>
12#include <linux/spinlock.h>
13#include <linux/seq_file.h>
14#include <linux/slab.h>
15#include <linux/kmod.h>
16#include <linux/kobj_map.h>
Christoph Hellwig2ef41632005-05-05 16:15:59 -070017#include <linux/buffer_head.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018
19#define MAX_PROBE_HASH 255 /* random */
20
21static struct subsystem block_subsys;
22
23static DECLARE_MUTEX(block_subsys_sem);
24
25/*
26 * Can be deleted altogether. Later.
27 *
28 */
29static struct blk_major_name {
30 struct blk_major_name *next;
31 int major;
32 char name[16];
33} *major_names[MAX_PROBE_HASH];
34
35/* index in the above - for now: assume no multimajor ranges */
36static inline int major_to_index(int major)
37{
38 return major % MAX_PROBE_HASH;
39}
40
41#ifdef CONFIG_PROC_FS
42/* get block device names in somewhat random order */
Neil Hormanac204272005-06-23 00:09:11 -070043int get_blkdev_list(char *p, int used)
Linus Torvalds1da177e2005-04-16 15:20:36 -070044{
45 struct blk_major_name *n;
46 int i, len;
47
Neil Hormanf62c6d02005-09-06 15:17:30 -070048 len = snprintf(p, (PAGE_SIZE-used), "\nBlock devices:\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -070049
50 down(&block_subsys_sem);
51 for (i = 0; i < ARRAY_SIZE(major_names); i++) {
Neil Hormanac204272005-06-23 00:09:11 -070052 for (n = major_names[i]; n; n = n->next) {
53 /*
54 * If the curent string plus the 5 extra characters
55 * in the line would run us off the page, then we're done
56 */
57 if ((len + used + strlen(n->name) + 5) >= PAGE_SIZE)
58 goto page_full;
Linus Torvalds1da177e2005-04-16 15:20:36 -070059 len += sprintf(p+len, "%3d %s\n",
60 n->major, n->name);
Neil Hormanac204272005-06-23 00:09:11 -070061 }
Linus Torvalds1da177e2005-04-16 15:20:36 -070062 }
Neil Hormanac204272005-06-23 00:09:11 -070063page_full:
Linus Torvalds1da177e2005-04-16 15:20:36 -070064 up(&block_subsys_sem);
65
66 return len;
67}
68#endif
69
70int register_blkdev(unsigned int major, const char *name)
71{
72 struct blk_major_name **n, *p;
73 int index, ret = 0;
74
75 down(&block_subsys_sem);
76
77 /* temporary */
78 if (major == 0) {
79 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
80 if (major_names[index] == NULL)
81 break;
82 }
83
84 if (index == 0) {
85 printk("register_blkdev: failed to get major for %s\n",
86 name);
87 ret = -EBUSY;
88 goto out;
89 }
90 major = index;
91 ret = major;
92 }
93
94 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
95 if (p == NULL) {
96 ret = -ENOMEM;
97 goto out;
98 }
99
100 p->major = major;
101 strlcpy(p->name, name, sizeof(p->name));
102 p->next = NULL;
103 index = major_to_index(major);
104
105 for (n = &major_names[index]; *n; n = &(*n)->next) {
106 if ((*n)->major == major)
107 break;
108 }
109 if (!*n)
110 *n = p;
111 else
112 ret = -EBUSY;
113
114 if (ret < 0) {
115 printk("register_blkdev: cannot get major %d for %s\n",
116 major, name);
117 kfree(p);
118 }
119out:
120 up(&block_subsys_sem);
121 return ret;
122}
123
124EXPORT_SYMBOL(register_blkdev);
125
126/* todo: make void - error printk here */
127int unregister_blkdev(unsigned int major, const char *name)
128{
129 struct blk_major_name **n;
130 struct blk_major_name *p = NULL;
131 int index = major_to_index(major);
132 int ret = 0;
133
134 down(&block_subsys_sem);
135 for (n = &major_names[index]; *n; n = &(*n)->next)
136 if ((*n)->major == major)
137 break;
138 if (!*n || strcmp((*n)->name, name))
139 ret = -EINVAL;
140 else {
141 p = *n;
142 *n = p->next;
143 }
144 up(&block_subsys_sem);
145 kfree(p);
146
147 return ret;
148}
149
150EXPORT_SYMBOL(unregister_blkdev);
151
152static struct kobj_map *bdev_map;
153
154/*
155 * Register device numbers dev..(dev+range-1)
156 * range must be nonzero
157 * The hash chain is sorted on range, so that subranges can override.
158 */
159void blk_register_region(dev_t dev, unsigned long range, struct module *module,
160 struct kobject *(*probe)(dev_t, int *, void *),
161 int (*lock)(dev_t, void *), void *data)
162{
163 kobj_map(bdev_map, dev, range, module, probe, lock, data);
164}
165
166EXPORT_SYMBOL(blk_register_region);
167
168void blk_unregister_region(dev_t dev, unsigned long range)
169{
170 kobj_unmap(bdev_map, dev, range);
171}
172
173EXPORT_SYMBOL(blk_unregister_region);
174
175static struct kobject *exact_match(dev_t dev, int *part, void *data)
176{
177 struct gendisk *p = data;
178 return &p->kobj;
179}
180
181static int exact_lock(dev_t dev, void *data)
182{
183 struct gendisk *p = data;
184
185 if (!get_disk(p))
186 return -1;
187 return 0;
188}
189
190/**
191 * add_disk - add partitioning information to kernel list
192 * @disk: per-device partitioning information
193 *
194 * This function registers the partitioning information in @disk
195 * with the kernel.
196 */
197void add_disk(struct gendisk *disk)
198{
199 disk->flags |= GENHD_FL_UP;
200 blk_register_region(MKDEV(disk->major, disk->first_minor),
201 disk->minors, NULL, exact_match, exact_lock, disk);
202 register_disk(disk);
203 blk_register_queue(disk);
204}
205
206EXPORT_SYMBOL(add_disk);
207EXPORT_SYMBOL(del_gendisk); /* in partitions/check.c */
208
209void unlink_gendisk(struct gendisk *disk)
210{
211 blk_unregister_queue(disk);
212 blk_unregister_region(MKDEV(disk->major, disk->first_minor),
213 disk->minors);
214}
215
216#define to_disk(obj) container_of(obj,struct gendisk,kobj)
217
218/**
219 * get_gendisk - get partitioning information for a given device
220 * @dev: device to get partitioning information for
221 *
222 * This function gets the structure containing partitioning
223 * information for the given device @dev.
224 */
225struct gendisk *get_gendisk(dev_t dev, int *part)
226{
227 struct kobject *kobj = kobj_lookup(bdev_map, dev, part);
228 return kobj ? to_disk(kobj) : NULL;
229}
230
231#ifdef CONFIG_PROC_FS
232/* iterator */
233static void *part_start(struct seq_file *part, loff_t *pos)
234{
235 struct list_head *p;
236 loff_t l = *pos;
237
238 down(&block_subsys_sem);
239 list_for_each(p, &block_subsys.kset.list)
240 if (!l--)
241 return list_entry(p, struct gendisk, kobj.entry);
242 return NULL;
243}
244
245static void *part_next(struct seq_file *part, void *v, loff_t *pos)
246{
247 struct list_head *p = ((struct gendisk *)v)->kobj.entry.next;
248 ++*pos;
249 return p==&block_subsys.kset.list ? NULL :
250 list_entry(p, struct gendisk, kobj.entry);
251}
252
253static void part_stop(struct seq_file *part, void *v)
254{
255 up(&block_subsys_sem);
256}
257
258static int show_partition(struct seq_file *part, void *v)
259{
260 struct gendisk *sgp = v;
261 int n;
262 char buf[BDEVNAME_SIZE];
263
264 if (&sgp->kobj.entry == block_subsys.kset.list.next)
265 seq_puts(part, "major minor #blocks name\n\n");
266
267 /* Don't show non-partitionable removeable devices or empty devices */
268 if (!get_capacity(sgp) ||
269 (sgp->minors == 1 && (sgp->flags & GENHD_FL_REMOVABLE)))
270 return 0;
271 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
272 return 0;
273
274 /* show the full disk and all non-0 size partitions of it */
275 seq_printf(part, "%4d %4d %10llu %s\n",
276 sgp->major, sgp->first_minor,
277 (unsigned long long)get_capacity(sgp) >> 1,
278 disk_name(sgp, 0, buf));
279 for (n = 0; n < sgp->minors - 1; n++) {
280 if (!sgp->part[n])
281 continue;
282 if (sgp->part[n]->nr_sects == 0)
283 continue;
284 seq_printf(part, "%4d %4d %10llu %s\n",
285 sgp->major, n + 1 + sgp->first_minor,
286 (unsigned long long)sgp->part[n]->nr_sects >> 1 ,
287 disk_name(sgp, n + 1, buf));
288 }
289
290 return 0;
291}
292
293struct seq_operations partitions_op = {
294 .start =part_start,
295 .next = part_next,
296 .stop = part_stop,
297 .show = show_partition
298};
299#endif
300
301
302extern int blk_dev_init(void);
303
304static struct kobject *base_probe(dev_t dev, int *part, void *data)
305{
306 if (request_module("block-major-%d-%d", MAJOR(dev), MINOR(dev)) > 0)
307 /* Make old-style 2.4 aliases work */
308 request_module("block-major-%d", MAJOR(dev));
309 return NULL;
310}
311
312static int __init genhd_device_init(void)
313{
314 bdev_map = kobj_map_init(base_probe, &block_subsys_sem);
315 blk_dev_init();
316 subsystem_register(&block_subsys);
317 return 0;
318}
319
320subsys_initcall(genhd_device_init);
321
322
323
324/*
325 * kobject & sysfs bindings for block devices
326 */
327static ssize_t disk_attr_show(struct kobject *kobj, struct attribute *attr,
328 char *page)
329{
330 struct gendisk *disk = to_disk(kobj);
331 struct disk_attribute *disk_attr =
332 container_of(attr,struct disk_attribute,attr);
Dmitry Torokhov6c1852a2005-04-29 01:26:06 -0500333 ssize_t ret = -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334
335 if (disk_attr->show)
336 ret = disk_attr->show(disk,page);
337 return ret;
338}
339
340static struct sysfs_ops disk_sysfs_ops = {
341 .show = &disk_attr_show,
342};
343
344static ssize_t disk_dev_read(struct gendisk * disk, char *page)
345{
346 dev_t base = MKDEV(disk->major, disk->first_minor);
347 return print_dev_t(page, base);
348}
349static ssize_t disk_range_read(struct gendisk * disk, char *page)
350{
351 return sprintf(page, "%d\n", disk->minors);
352}
353static ssize_t disk_removable_read(struct gendisk * disk, char *page)
354{
355 return sprintf(page, "%d\n",
356 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
357
358}
359static ssize_t disk_size_read(struct gendisk * disk, char *page)
360{
361 return sprintf(page, "%llu\n", (unsigned long long)get_capacity(disk));
362}
363
364static ssize_t disk_stats_read(struct gendisk * disk, char *page)
365{
366 preempt_disable();
367 disk_round_stats(disk);
368 preempt_enable();
369 return sprintf(page,
370 "%8u %8u %8llu %8u "
371 "%8u %8u %8llu %8u "
372 "%8u %8u %8u"
373 "\n",
374 disk_stat_read(disk, reads), disk_stat_read(disk, read_merges),
375 (unsigned long long)disk_stat_read(disk, read_sectors),
376 jiffies_to_msecs(disk_stat_read(disk, read_ticks)),
377 disk_stat_read(disk, writes),
378 disk_stat_read(disk, write_merges),
379 (unsigned long long)disk_stat_read(disk, write_sectors),
380 jiffies_to_msecs(disk_stat_read(disk, write_ticks)),
381 disk->in_flight,
382 jiffies_to_msecs(disk_stat_read(disk, io_ticks)),
383 jiffies_to_msecs(disk_stat_read(disk, time_in_queue)));
384}
385static struct disk_attribute disk_attr_dev = {
386 .attr = {.name = "dev", .mode = S_IRUGO },
387 .show = disk_dev_read
388};
389static struct disk_attribute disk_attr_range = {
390 .attr = {.name = "range", .mode = S_IRUGO },
391 .show = disk_range_read
392};
393static struct disk_attribute disk_attr_removable = {
394 .attr = {.name = "removable", .mode = S_IRUGO },
395 .show = disk_removable_read
396};
397static struct disk_attribute disk_attr_size = {
398 .attr = {.name = "size", .mode = S_IRUGO },
399 .show = disk_size_read
400};
401static struct disk_attribute disk_attr_stat = {
402 .attr = {.name = "stat", .mode = S_IRUGO },
403 .show = disk_stats_read
404};
405
406static struct attribute * default_attrs[] = {
407 &disk_attr_dev.attr,
408 &disk_attr_range.attr,
409 &disk_attr_removable.attr,
410 &disk_attr_size.attr,
411 &disk_attr_stat.attr,
412 NULL,
413};
414
415static void disk_release(struct kobject * kobj)
416{
417 struct gendisk *disk = to_disk(kobj);
418 kfree(disk->random);
419 kfree(disk->part);
420 free_disk_stats(disk);
421 kfree(disk);
422}
423
424static struct kobj_type ktype_block = {
425 .release = disk_release,
426 .sysfs_ops = &disk_sysfs_ops,
427 .default_attrs = default_attrs,
428};
429
430extern struct kobj_type ktype_part;
431
432static int block_hotplug_filter(struct kset *kset, struct kobject *kobj)
433{
434 struct kobj_type *ktype = get_ktype(kobj);
435
436 return ((ktype == &ktype_block) || (ktype == &ktype_part));
437}
438
439static int block_hotplug(struct kset *kset, struct kobject *kobj, char **envp,
440 int num_envp, char *buffer, int buffer_size)
441{
442 struct kobj_type *ktype = get_ktype(kobj);
443 struct device *physdev;
444 struct gendisk *disk;
445 struct hd_struct *part;
446 int length = 0;
447 int i = 0;
448
449 if (ktype == &ktype_block) {
450 disk = container_of(kobj, struct gendisk, kobj);
451 add_hotplug_env_var(envp, num_envp, &i, buffer, buffer_size,
452 &length, "MINOR=%u", disk->first_minor);
453 } else if (ktype == &ktype_part) {
454 disk = container_of(kobj->parent, struct gendisk, kobj);
455 part = container_of(kobj, struct hd_struct, kobj);
456 add_hotplug_env_var(envp, num_envp, &i, buffer, buffer_size,
457 &length, "MINOR=%u",
458 disk->first_minor + part->partno);
459 } else
460 return 0;
461
462 add_hotplug_env_var(envp, num_envp, &i, buffer, buffer_size, &length,
463 "MAJOR=%u", disk->major);
464
465 /* add physical device, backing this device */
466 physdev = disk->driverfs_dev;
467 if (physdev) {
468 char *path = kobject_get_path(&physdev->kobj, GFP_KERNEL);
469
470 add_hotplug_env_var(envp, num_envp, &i, buffer, buffer_size,
471 &length, "PHYSDEVPATH=%s", path);
472 kfree(path);
473
474 if (physdev->bus)
475 add_hotplug_env_var(envp, num_envp, &i,
476 buffer, buffer_size, &length,
477 "PHYSDEVBUS=%s",
478 physdev->bus->name);
479
480 if (physdev->driver)
481 add_hotplug_env_var(envp, num_envp, &i,
482 buffer, buffer_size, &length,
483 "PHYSDEVDRIVER=%s",
484 physdev->driver->name);
485 }
486
487 /* terminate, set to next free slot, shrink available space */
488 envp[i] = NULL;
489 envp = &envp[i];
490 num_envp -= i;
491 buffer = &buffer[length];
492 buffer_size -= length;
493
494 return 0;
495}
496
497static struct kset_hotplug_ops block_hotplug_ops = {
498 .filter = block_hotplug_filter,
499 .hotplug = block_hotplug,
500};
501
502/* declare block_subsys. */
503static decl_subsys(block, &ktype_block, &block_hotplug_ops);
504
505
506/*
507 * aggregate disk stat collector. Uses the same stats that the sysfs
508 * entries do, above, but makes them available through one seq_file.
509 * Watching a few disks may be efficient through sysfs, but watching
510 * all of them will be more efficient through this interface.
511 *
512 * The output looks suspiciously like /proc/partitions with a bunch of
513 * extra fields.
514 */
515
516/* iterator */
517static void *diskstats_start(struct seq_file *part, loff_t *pos)
518{
519 loff_t k = *pos;
520 struct list_head *p;
521
522 down(&block_subsys_sem);
523 list_for_each(p, &block_subsys.kset.list)
524 if (!k--)
525 return list_entry(p, struct gendisk, kobj.entry);
526 return NULL;
527}
528
529static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos)
530{
531 struct list_head *p = ((struct gendisk *)v)->kobj.entry.next;
532 ++*pos;
533 return p==&block_subsys.kset.list ? NULL :
534 list_entry(p, struct gendisk, kobj.entry);
535}
536
537static void diskstats_stop(struct seq_file *part, void *v)
538{
539 up(&block_subsys_sem);
540}
541
542static int diskstats_show(struct seq_file *s, void *v)
543{
544 struct gendisk *gp = v;
545 char buf[BDEVNAME_SIZE];
546 int n = 0;
547
548 /*
549 if (&sgp->kobj.entry == block_subsys.kset.list.next)
550 seq_puts(s, "major minor name"
551 " rio rmerge rsect ruse wio wmerge "
552 "wsect wuse running use aveq"
553 "\n\n");
554 */
555
556 preempt_disable();
557 disk_round_stats(gp);
558 preempt_enable();
559 seq_printf(s, "%4d %4d %s %u %u %llu %u %u %u %llu %u %u %u %u\n",
560 gp->major, n + gp->first_minor, disk_name(gp, n, buf),
561 disk_stat_read(gp, reads), disk_stat_read(gp, read_merges),
562 (unsigned long long)disk_stat_read(gp, read_sectors),
563 jiffies_to_msecs(disk_stat_read(gp, read_ticks)),
564 disk_stat_read(gp, writes), disk_stat_read(gp, write_merges),
565 (unsigned long long)disk_stat_read(gp, write_sectors),
566 jiffies_to_msecs(disk_stat_read(gp, write_ticks)),
567 gp->in_flight,
568 jiffies_to_msecs(disk_stat_read(gp, io_ticks)),
569 jiffies_to_msecs(disk_stat_read(gp, time_in_queue)));
570
571 /* now show all non-0 size partitions of it */
572 for (n = 0; n < gp->minors - 1; n++) {
573 struct hd_struct *hd = gp->part[n];
574
575 if (hd && hd->nr_sects)
576 seq_printf(s, "%4d %4d %s %u %u %u %u\n",
577 gp->major, n + gp->first_minor + 1,
578 disk_name(gp, n + 1, buf),
579 hd->reads, hd->read_sectors,
580 hd->writes, hd->write_sectors);
581 }
582
583 return 0;
584}
585
586struct seq_operations diskstats_op = {
587 .start = diskstats_start,
588 .next = diskstats_next,
589 .stop = diskstats_stop,
590 .show = diskstats_show
591};
592
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593struct gendisk *alloc_disk(int minors)
594{
Christoph Lameter19460892005-06-23 00:08:19 -0700595 return alloc_disk_node(minors, -1);
596}
597
598struct gendisk *alloc_disk_node(int minors, int node_id)
599{
600 struct gendisk *disk;
601
602 disk = kmalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 if (disk) {
604 memset(disk, 0, sizeof(struct gendisk));
605 if (!init_disk_stats(disk)) {
606 kfree(disk);
607 return NULL;
608 }
609 if (minors > 1) {
610 int size = (minors - 1) * sizeof(struct hd_struct *);
Christoph Lameter19460892005-06-23 00:08:19 -0700611 disk->part = kmalloc_node(size, GFP_KERNEL, node_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612 if (!disk->part) {
613 kfree(disk);
614 return NULL;
615 }
616 memset(disk->part, 0, size);
617 }
618 disk->minors = minors;
619 kobj_set_kset_s(disk,block_subsys);
620 kobject_init(&disk->kobj);
621 rand_initialize_disk(disk);
622 }
623 return disk;
624}
625
626EXPORT_SYMBOL(alloc_disk);
Christoph Lameter19460892005-06-23 00:08:19 -0700627EXPORT_SYMBOL(alloc_disk_node);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628
629struct kobject *get_disk(struct gendisk *disk)
630{
631 struct module *owner;
632 struct kobject *kobj;
633
634 if (!disk->fops)
635 return NULL;
636 owner = disk->fops->owner;
637 if (owner && !try_module_get(owner))
638 return NULL;
639 kobj = kobject_get(&disk->kobj);
640 if (kobj == NULL) {
641 module_put(owner);
642 return NULL;
643 }
644 return kobj;
645
646}
647
648EXPORT_SYMBOL(get_disk);
649
650void put_disk(struct gendisk *disk)
651{
652 if (disk)
653 kobject_put(&disk->kobj);
654}
655
656EXPORT_SYMBOL(put_disk);
657
658void set_device_ro(struct block_device *bdev, int flag)
659{
660 if (bdev->bd_contains != bdev)
661 bdev->bd_part->policy = flag;
662 else
663 bdev->bd_disk->policy = flag;
664}
665
666EXPORT_SYMBOL(set_device_ro);
667
668void set_disk_ro(struct gendisk *disk, int flag)
669{
670 int i;
671 disk->policy = flag;
672 for (i = 0; i < disk->minors - 1; i++)
673 if (disk->part[i]) disk->part[i]->policy = flag;
674}
675
676EXPORT_SYMBOL(set_disk_ro);
677
678int bdev_read_only(struct block_device *bdev)
679{
680 if (!bdev)
681 return 0;
682 else if (bdev->bd_contains != bdev)
683 return bdev->bd_part->policy;
684 else
685 return bdev->bd_disk->policy;
686}
687
688EXPORT_SYMBOL(bdev_read_only);
689
690int invalidate_partition(struct gendisk *disk, int index)
691{
692 int res = 0;
693 struct block_device *bdev = bdget_disk(disk, index);
694 if (bdev) {
Christoph Hellwig2ef41632005-05-05 16:15:59 -0700695 fsync_bdev(bdev);
696 res = __invalidate_device(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697 bdput(bdev);
698 }
699 return res;
700}
701
702EXPORT_SYMBOL(invalidate_partition);