blob: 863a8c0623ed824e27d85f32a52cf10ea39d1c02 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * gendisk handling
3 */
4
Linus Torvalds1da177e2005-04-16 15:20:36 -07005#include <linux/module.h>
6#include <linux/fs.h>
7#include <linux/genhd.h>
Andrew Mortonb446b602007-02-20 13:57:48 -08008#include <linux/kdev_t.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#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>
Jes Sorensen58383af2006-02-06 14:12:43 -080018#include <linux/mutex.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019
Greg Kroah-Hartman823bccf2007-04-13 13:15:19 -070020struct kset block_subsys;
Jes Sorensen58383af2006-02-06 14:12:43 -080021static DEFINE_MUTEX(block_subsys_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070022
23/*
24 * Can be deleted altogether. Later.
25 *
26 */
27static struct blk_major_name {
28 struct blk_major_name *next;
29 int major;
30 char name[16];
Joe Korty68eef3b2006-03-31 02:30:32 -080031} *major_names[BLKDEV_MAJOR_HASH_SIZE];
Linus Torvalds1da177e2005-04-16 15:20:36 -070032
33/* index in the above - for now: assume no multimajor ranges */
34static inline int major_to_index(int major)
35{
Joe Korty68eef3b2006-03-31 02:30:32 -080036 return major % BLKDEV_MAJOR_HASH_SIZE;
Linus Torvalds1da177e2005-04-16 15:20:36 -070037}
38
Joe Korty68eef3b2006-03-31 02:30:32 -080039#ifdef CONFIG_PROC_FS
Neil Horman7170be52006-01-14 13:20:38 -080040
Joe Korty68eef3b2006-03-31 02:30:32 -080041void blkdev_show(struct seq_file *f, off_t offset)
Neil Horman7170be52006-01-14 13:20:38 -080042{
Joe Korty68eef3b2006-03-31 02:30:32 -080043 struct blk_major_name *dp;
Neil Horman7170be52006-01-14 13:20:38 -080044
Joe Korty68eef3b2006-03-31 02:30:32 -080045 if (offset < BLKDEV_MAJOR_HASH_SIZE) {
46 mutex_lock(&block_subsys_lock);
47 for (dp = major_names[offset]; dp; dp = dp->next)
48 seq_printf(f, "%3d %s\n", dp->major, dp->name);
49 mutex_unlock(&block_subsys_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070050 }
Linus Torvalds1da177e2005-04-16 15:20:36 -070051}
Neil Horman7170be52006-01-14 13:20:38 -080052
Joe Korty68eef3b2006-03-31 02:30:32 -080053#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -070054
55int register_blkdev(unsigned int major, const char *name)
56{
57 struct blk_major_name **n, *p;
58 int index, ret = 0;
59
Jes Sorensen58383af2006-02-06 14:12:43 -080060 mutex_lock(&block_subsys_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070061
62 /* temporary */
63 if (major == 0) {
64 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
65 if (major_names[index] == NULL)
66 break;
67 }
68
69 if (index == 0) {
70 printk("register_blkdev: failed to get major for %s\n",
71 name);
72 ret = -EBUSY;
73 goto out;
74 }
75 major = index;
76 ret = major;
77 }
78
79 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
80 if (p == NULL) {
81 ret = -ENOMEM;
82 goto out;
83 }
84
85 p->major = major;
86 strlcpy(p->name, name, sizeof(p->name));
87 p->next = NULL;
88 index = major_to_index(major);
89
90 for (n = &major_names[index]; *n; n = &(*n)->next) {
91 if ((*n)->major == major)
92 break;
93 }
94 if (!*n)
95 *n = p;
96 else
97 ret = -EBUSY;
98
99 if (ret < 0) {
100 printk("register_blkdev: cannot get major %d for %s\n",
101 major, name);
102 kfree(p);
103 }
104out:
Jes Sorensen58383af2006-02-06 14:12:43 -0800105 mutex_unlock(&block_subsys_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106 return ret;
107}
108
109EXPORT_SYMBOL(register_blkdev);
110
111/* todo: make void - error printk here */
112int unregister_blkdev(unsigned int major, const char *name)
113{
114 struct blk_major_name **n;
115 struct blk_major_name *p = NULL;
116 int index = major_to_index(major);
117 int ret = 0;
118
Jes Sorensen58383af2006-02-06 14:12:43 -0800119 mutex_lock(&block_subsys_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120 for (n = &major_names[index]; *n; n = &(*n)->next)
121 if ((*n)->major == major)
122 break;
123 if (!*n || strcmp((*n)->name, name))
124 ret = -EINVAL;
125 else {
126 p = *n;
127 *n = p->next;
128 }
Jes Sorensen58383af2006-02-06 14:12:43 -0800129 mutex_unlock(&block_subsys_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 kfree(p);
131
132 return ret;
133}
134
135EXPORT_SYMBOL(unregister_blkdev);
136
137static struct kobj_map *bdev_map;
138
139/*
140 * Register device numbers dev..(dev+range-1)
141 * range must be nonzero
142 * The hash chain is sorted on range, so that subranges can override.
143 */
144void blk_register_region(dev_t dev, unsigned long range, struct module *module,
145 struct kobject *(*probe)(dev_t, int *, void *),
146 int (*lock)(dev_t, void *), void *data)
147{
148 kobj_map(bdev_map, dev, range, module, probe, lock, data);
149}
150
151EXPORT_SYMBOL(blk_register_region);
152
153void blk_unregister_region(dev_t dev, unsigned long range)
154{
155 kobj_unmap(bdev_map, dev, range);
156}
157
158EXPORT_SYMBOL(blk_unregister_region);
159
160static struct kobject *exact_match(dev_t dev, int *part, void *data)
161{
162 struct gendisk *p = data;
163 return &p->kobj;
164}
165
166static int exact_lock(dev_t dev, void *data)
167{
168 struct gendisk *p = data;
169
170 if (!get_disk(p))
171 return -1;
172 return 0;
173}
174
175/**
176 * add_disk - add partitioning information to kernel list
177 * @disk: per-device partitioning information
178 *
179 * This function registers the partitioning information in @disk
180 * with the kernel.
181 */
182void add_disk(struct gendisk *disk)
183{
184 disk->flags |= GENHD_FL_UP;
185 blk_register_region(MKDEV(disk->major, disk->first_minor),
186 disk->minors, NULL, exact_match, exact_lock, disk);
187 register_disk(disk);
188 blk_register_queue(disk);
189}
190
191EXPORT_SYMBOL(add_disk);
192EXPORT_SYMBOL(del_gendisk); /* in partitions/check.c */
193
194void unlink_gendisk(struct gendisk *disk)
195{
196 blk_unregister_queue(disk);
197 blk_unregister_region(MKDEV(disk->major, disk->first_minor),
198 disk->minors);
199}
200
201#define to_disk(obj) container_of(obj,struct gendisk,kobj)
202
203/**
204 * get_gendisk - get partitioning information for a given device
205 * @dev: device to get partitioning information for
206 *
207 * This function gets the structure containing partitioning
208 * information for the given device @dev.
209 */
210struct gendisk *get_gendisk(dev_t dev, int *part)
211{
212 struct kobject *kobj = kobj_lookup(bdev_map, dev, part);
213 return kobj ? to_disk(kobj) : NULL;
214}
215
Dave Gilbertdd2a3452007-05-09 02:33:24 -0700216/*
217 * print a full list of all partitions - intended for places where the root
218 * filesystem can't be mounted and thus to give the victim some idea of what
219 * went wrong
220 */
221void __init printk_all_partitions(void)
222{
223 int n;
224 struct gendisk *sgp;
225
226 mutex_lock(&block_subsys_lock);
227 /* For each block device... */
228 list_for_each_entry(sgp, &block_subsys.list, kobj.entry) {
229 char buf[BDEVNAME_SIZE];
230 /*
231 * Don't show empty devices or things that have been surpressed
232 */
233 if (get_capacity(sgp) == 0 ||
234 (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
235 continue;
236
237 /*
238 * Note, unlike /proc/partitions, I am showing the numbers in
239 * hex - the same format as the root= option takes.
240 */
241 printk("%02x%02x %10llu %s",
242 sgp->major, sgp->first_minor,
243 (unsigned long long)get_capacity(sgp) >> 1,
244 disk_name(sgp, 0, buf));
245 if (sgp->driverfs_dev != NULL &&
246 sgp->driverfs_dev->driver != NULL)
247 printk(" driver: %s\n",
248 sgp->driverfs_dev->driver->name);
249 else
250 printk(" (driver?)\n");
251
252 /* now show the partitions */
253 for (n = 0; n < sgp->minors - 1; ++n) {
254 if (sgp->part[n] == NULL)
255 continue;
256 if (sgp->part[n]->nr_sects == 0)
257 continue;
258 printk(" %02x%02x %10llu %s\n",
259 sgp->major, n + 1 + sgp->first_minor,
260 (unsigned long long)sgp->part[n]->nr_sects >> 1,
261 disk_name(sgp, n + 1, buf));
262 } /* partition subloop */
263 } /* Block device loop */
264
265 mutex_unlock(&block_subsys_lock);
266 return;
267}
268
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269#ifdef CONFIG_PROC_FS
270/* iterator */
271static void *part_start(struct seq_file *part, loff_t *pos)
272{
273 struct list_head *p;
274 loff_t l = *pos;
275
Jes Sorensen58383af2006-02-06 14:12:43 -0800276 mutex_lock(&block_subsys_lock);
Greg Kroah-Hartman823bccf2007-04-13 13:15:19 -0700277 list_for_each(p, &block_subsys.list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278 if (!l--)
279 return list_entry(p, struct gendisk, kobj.entry);
280 return NULL;
281}
282
283static void *part_next(struct seq_file *part, void *v, loff_t *pos)
284{
285 struct list_head *p = ((struct gendisk *)v)->kobj.entry.next;
286 ++*pos;
Greg Kroah-Hartman823bccf2007-04-13 13:15:19 -0700287 return p==&block_subsys.list ? NULL :
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 list_entry(p, struct gendisk, kobj.entry);
289}
290
291static void part_stop(struct seq_file *part, void *v)
292{
Jes Sorensen58383af2006-02-06 14:12:43 -0800293 mutex_unlock(&block_subsys_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294}
295
296static int show_partition(struct seq_file *part, void *v)
297{
298 struct gendisk *sgp = v;
299 int n;
300 char buf[BDEVNAME_SIZE];
301
Greg Kroah-Hartman823bccf2007-04-13 13:15:19 -0700302 if (&sgp->kobj.entry == block_subsys.list.next)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303 seq_puts(part, "major minor #blocks name\n\n");
304
305 /* Don't show non-partitionable removeable devices or empty devices */
306 if (!get_capacity(sgp) ||
307 (sgp->minors == 1 && (sgp->flags & GENHD_FL_REMOVABLE)))
308 return 0;
309 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
310 return 0;
311
312 /* show the full disk and all non-0 size partitions of it */
313 seq_printf(part, "%4d %4d %10llu %s\n",
314 sgp->major, sgp->first_minor,
315 (unsigned long long)get_capacity(sgp) >> 1,
316 disk_name(sgp, 0, buf));
317 for (n = 0; n < sgp->minors - 1; n++) {
318 if (!sgp->part[n])
319 continue;
320 if (sgp->part[n]->nr_sects == 0)
321 continue;
322 seq_printf(part, "%4d %4d %10llu %s\n",
323 sgp->major, n + 1 + sgp->first_minor,
324 (unsigned long long)sgp->part[n]->nr_sects >> 1 ,
325 disk_name(sgp, n + 1, buf));
326 }
327
328 return 0;
329}
330
331struct seq_operations partitions_op = {
332 .start =part_start,
333 .next = part_next,
334 .stop = part_stop,
335 .show = show_partition
336};
337#endif
338
339
340extern int blk_dev_init(void);
341
342static struct kobject *base_probe(dev_t dev, int *part, void *data)
343{
344 if (request_module("block-major-%d-%d", MAJOR(dev), MINOR(dev)) > 0)
345 /* Make old-style 2.4 aliases work */
346 request_module("block-major-%d", MAJOR(dev));
347 return NULL;
348}
349
350static int __init genhd_device_init(void)
351{
Randy Dunlap87a57262006-09-29 01:58:56 -0700352 int err;
353
Jes Sorensen58383af2006-02-06 14:12:43 -0800354 bdev_map = kobj_map_init(base_probe, &block_subsys_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 blk_dev_init();
Randy Dunlap87a57262006-09-29 01:58:56 -0700356 err = subsystem_register(&block_subsys);
357 if (err < 0)
358 printk(KERN_WARNING "%s: subsystem_register error: %d\n",
359 __FUNCTION__, err);
360 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361}
362
363subsys_initcall(genhd_device_init);
364
365
366
367/*
368 * kobject & sysfs bindings for block devices
369 */
370static ssize_t disk_attr_show(struct kobject *kobj, struct attribute *attr,
371 char *page)
372{
373 struct gendisk *disk = to_disk(kobj);
374 struct disk_attribute *disk_attr =
375 container_of(attr,struct disk_attribute,attr);
Dmitry Torokhov6c1852a2005-04-29 01:26:06 -0500376 ssize_t ret = -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377
378 if (disk_attr->show)
379 ret = disk_attr->show(disk,page);
380 return ret;
381}
382
Kay Sieversa7fd6702005-10-01 14:49:43 +0200383static ssize_t disk_attr_store(struct kobject * kobj, struct attribute * attr,
384 const char *page, size_t count)
385{
386 struct gendisk *disk = to_disk(kobj);
387 struct disk_attribute *disk_attr =
388 container_of(attr,struct disk_attribute,attr);
389 ssize_t ret = 0;
390
391 if (disk_attr->store)
392 ret = disk_attr->store(disk, page, count);
393 return ret;
394}
395
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396static struct sysfs_ops disk_sysfs_ops = {
397 .show = &disk_attr_show,
Kay Sieversa7fd6702005-10-01 14:49:43 +0200398 .store = &disk_attr_store,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399};
400
Kay Sieversa7fd6702005-10-01 14:49:43 +0200401static ssize_t disk_uevent_store(struct gendisk * disk,
402 const char *buf, size_t count)
403{
Kay Sievers312c0042005-11-16 09:00:00 +0100404 kobject_uevent(&disk->kobj, KOBJ_ADD);
Kay Sieversa7fd6702005-10-01 14:49:43 +0200405 return count;
406}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407static ssize_t disk_dev_read(struct gendisk * disk, char *page)
408{
409 dev_t base = MKDEV(disk->major, disk->first_minor);
410 return print_dev_t(page, base);
411}
412static ssize_t disk_range_read(struct gendisk * disk, char *page)
413{
414 return sprintf(page, "%d\n", disk->minors);
415}
416static ssize_t disk_removable_read(struct gendisk * disk, char *page)
417{
418 return sprintf(page, "%d\n",
419 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
420
421}
422static ssize_t disk_size_read(struct gendisk * disk, char *page)
423{
424 return sprintf(page, "%llu\n", (unsigned long long)get_capacity(disk));
425}
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -0700426static ssize_t disk_capability_read(struct gendisk *disk, char *page)
427{
428 return sprintf(page, "%x\n", disk->flags);
429}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430static ssize_t disk_stats_read(struct gendisk * disk, char *page)
431{
432 preempt_disable();
433 disk_round_stats(disk);
434 preempt_enable();
435 return sprintf(page,
Ben Woodard837c7872006-03-22 08:09:31 +0100436 "%8lu %8lu %8llu %8u "
437 "%8lu %8lu %8llu %8u "
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 "%8u %8u %8u"
439 "\n",
Jens Axboe47a00412005-11-09 13:38:47 +0100440 disk_stat_read(disk, ios[READ]),
441 disk_stat_read(disk, merges[READ]),
442 (unsigned long long)disk_stat_read(disk, sectors[READ]),
443 jiffies_to_msecs(disk_stat_read(disk, ticks[READ])),
444 disk_stat_read(disk, ios[WRITE]),
445 disk_stat_read(disk, merges[WRITE]),
446 (unsigned long long)disk_stat_read(disk, sectors[WRITE]),
447 jiffies_to_msecs(disk_stat_read(disk, ticks[WRITE])),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 disk->in_flight,
449 jiffies_to_msecs(disk_stat_read(disk, io_ticks)),
450 jiffies_to_msecs(disk_stat_read(disk, time_in_queue)));
451}
Kay Sieversa7fd6702005-10-01 14:49:43 +0200452static struct disk_attribute disk_attr_uevent = {
453 .attr = {.name = "uevent", .mode = S_IWUSR },
454 .store = disk_uevent_store
455};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456static struct disk_attribute disk_attr_dev = {
457 .attr = {.name = "dev", .mode = S_IRUGO },
458 .show = disk_dev_read
459};
460static struct disk_attribute disk_attr_range = {
461 .attr = {.name = "range", .mode = S_IRUGO },
462 .show = disk_range_read
463};
464static struct disk_attribute disk_attr_removable = {
465 .attr = {.name = "removable", .mode = S_IRUGO },
466 .show = disk_removable_read
467};
468static struct disk_attribute disk_attr_size = {
469 .attr = {.name = "size", .mode = S_IRUGO },
470 .show = disk_size_read
471};
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -0700472static struct disk_attribute disk_attr_capability = {
473 .attr = {.name = "capability", .mode = S_IRUGO },
474 .show = disk_capability_read
475};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476static struct disk_attribute disk_attr_stat = {
477 .attr = {.name = "stat", .mode = S_IRUGO },
478 .show = disk_stats_read
479};
480
Akinobu Mitac17bb492006-12-08 02:39:46 -0800481#ifdef CONFIG_FAIL_MAKE_REQUEST
482
483static ssize_t disk_fail_store(struct gendisk * disk,
484 const char *buf, size_t count)
485{
486 int i;
487
488 if (count > 0 && sscanf(buf, "%d", &i) > 0) {
489 if (i == 0)
490 disk->flags &= ~GENHD_FL_FAIL;
491 else
492 disk->flags |= GENHD_FL_FAIL;
493 }
494
495 return count;
496}
497static ssize_t disk_fail_read(struct gendisk * disk, char *page)
498{
499 return sprintf(page, "%d\n", disk->flags & GENHD_FL_FAIL ? 1 : 0);
500}
501static struct disk_attribute disk_attr_fail = {
502 .attr = {.name = "make-it-fail", .mode = S_IRUGO | S_IWUSR },
503 .store = disk_fail_store,
504 .show = disk_fail_read
505};
506
507#endif
508
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509static struct attribute * default_attrs[] = {
Kay Sieversa7fd6702005-10-01 14:49:43 +0200510 &disk_attr_uevent.attr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 &disk_attr_dev.attr,
512 &disk_attr_range.attr,
513 &disk_attr_removable.attr,
514 &disk_attr_size.attr,
515 &disk_attr_stat.attr,
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -0700516 &disk_attr_capability.attr,
Akinobu Mitac17bb492006-12-08 02:39:46 -0800517#ifdef CONFIG_FAIL_MAKE_REQUEST
518 &disk_attr_fail.attr,
519#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 NULL,
521};
522
523static void disk_release(struct kobject * kobj)
524{
525 struct gendisk *disk = to_disk(kobj);
526 kfree(disk->random);
527 kfree(disk->part);
528 free_disk_stats(disk);
529 kfree(disk);
530}
531
532static struct kobj_type ktype_block = {
533 .release = disk_release,
534 .sysfs_ops = &disk_sysfs_ops,
535 .default_attrs = default_attrs,
536};
537
538extern struct kobj_type ktype_part;
539
Kay Sievers312c0042005-11-16 09:00:00 +0100540static int block_uevent_filter(struct kset *kset, struct kobject *kobj)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541{
542 struct kobj_type *ktype = get_ktype(kobj);
543
544 return ((ktype == &ktype_block) || (ktype == &ktype_part));
545}
546
Kay Sievers312c0042005-11-16 09:00:00 +0100547static int block_uevent(struct kset *kset, struct kobject *kobj, char **envp,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548 int num_envp, char *buffer, int buffer_size)
549{
550 struct kobj_type *ktype = get_ktype(kobj);
551 struct device *physdev;
552 struct gendisk *disk;
553 struct hd_struct *part;
554 int length = 0;
555 int i = 0;
556
557 if (ktype == &ktype_block) {
558 disk = container_of(kobj, struct gendisk, kobj);
Kay Sievers312c0042005-11-16 09:00:00 +0100559 add_uevent_var(envp, num_envp, &i, buffer, buffer_size,
560 &length, "MINOR=%u", disk->first_minor);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 } else if (ktype == &ktype_part) {
562 disk = container_of(kobj->parent, struct gendisk, kobj);
563 part = container_of(kobj, struct hd_struct, kobj);
Kay Sievers312c0042005-11-16 09:00:00 +0100564 add_uevent_var(envp, num_envp, &i, buffer, buffer_size,
565 &length, "MINOR=%u",
566 disk->first_minor + part->partno);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 } else
568 return 0;
569
Kay Sievers312c0042005-11-16 09:00:00 +0100570 add_uevent_var(envp, num_envp, &i, buffer, buffer_size, &length,
571 "MAJOR=%u", disk->major);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572
573 /* add physical device, backing this device */
574 physdev = disk->driverfs_dev;
575 if (physdev) {
576 char *path = kobject_get_path(&physdev->kobj, GFP_KERNEL);
577
Kay Sievers312c0042005-11-16 09:00:00 +0100578 add_uevent_var(envp, num_envp, &i, buffer, buffer_size,
579 &length, "PHYSDEVPATH=%s", path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 kfree(path);
581
582 if (physdev->bus)
Kay Sievers312c0042005-11-16 09:00:00 +0100583 add_uevent_var(envp, num_envp, &i,
584 buffer, buffer_size, &length,
585 "PHYSDEVBUS=%s",
586 physdev->bus->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587
588 if (physdev->driver)
Kay Sievers312c0042005-11-16 09:00:00 +0100589 add_uevent_var(envp, num_envp, &i,
590 buffer, buffer_size, &length,
591 "PHYSDEVDRIVER=%s",
592 physdev->driver->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593 }
594
595 /* terminate, set to next free slot, shrink available space */
596 envp[i] = NULL;
597 envp = &envp[i];
598 num_envp -= i;
599 buffer = &buffer[length];
600 buffer_size -= length;
601
602 return 0;
603}
604
Kay Sievers312c0042005-11-16 09:00:00 +0100605static struct kset_uevent_ops block_uevent_ops = {
606 .filter = block_uevent_filter,
607 .uevent = block_uevent,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608};
609
Kay Sieversb9d9c822006-06-15 15:31:56 +0200610decl_subsys(block, &ktype_block, &block_uevent_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611
612/*
613 * aggregate disk stat collector. Uses the same stats that the sysfs
614 * entries do, above, but makes them available through one seq_file.
615 * Watching a few disks may be efficient through sysfs, but watching
616 * all of them will be more efficient through this interface.
617 *
618 * The output looks suspiciously like /proc/partitions with a bunch of
619 * extra fields.
620 */
621
622/* iterator */
623static void *diskstats_start(struct seq_file *part, loff_t *pos)
624{
625 loff_t k = *pos;
626 struct list_head *p;
627
Jes Sorensen58383af2006-02-06 14:12:43 -0800628 mutex_lock(&block_subsys_lock);
Greg Kroah-Hartman823bccf2007-04-13 13:15:19 -0700629 list_for_each(p, &block_subsys.list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 if (!k--)
631 return list_entry(p, struct gendisk, kobj.entry);
632 return NULL;
633}
634
635static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos)
636{
637 struct list_head *p = ((struct gendisk *)v)->kobj.entry.next;
638 ++*pos;
Greg Kroah-Hartman823bccf2007-04-13 13:15:19 -0700639 return p==&block_subsys.list ? NULL :
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640 list_entry(p, struct gendisk, kobj.entry);
641}
642
643static void diskstats_stop(struct seq_file *part, void *v)
644{
Jes Sorensen58383af2006-02-06 14:12:43 -0800645 mutex_unlock(&block_subsys_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646}
647
648static int diskstats_show(struct seq_file *s, void *v)
649{
650 struct gendisk *gp = v;
651 char buf[BDEVNAME_SIZE];
652 int n = 0;
653
654 /*
655 if (&sgp->kobj.entry == block_subsys.kset.list.next)
656 seq_puts(s, "major minor name"
657 " rio rmerge rsect ruse wio wmerge "
658 "wsect wuse running use aveq"
659 "\n\n");
660 */
661
662 preempt_disable();
663 disk_round_stats(gp);
664 preempt_enable();
Ben Woodard837c7872006-03-22 08:09:31 +0100665 seq_printf(s, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666 gp->major, n + gp->first_minor, disk_name(gp, n, buf),
Jens Axboea3623572005-11-01 09:26:16 +0100667 disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]),
668 (unsigned long long)disk_stat_read(gp, sectors[0]),
669 jiffies_to_msecs(disk_stat_read(gp, ticks[0])),
670 disk_stat_read(gp, ios[1]), disk_stat_read(gp, merges[1]),
671 (unsigned long long)disk_stat_read(gp, sectors[1]),
672 jiffies_to_msecs(disk_stat_read(gp, ticks[1])),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 gp->in_flight,
674 jiffies_to_msecs(disk_stat_read(gp, io_ticks)),
675 jiffies_to_msecs(disk_stat_read(gp, time_in_queue)));
676
677 /* now show all non-0 size partitions of it */
678 for (n = 0; n < gp->minors - 1; n++) {
679 struct hd_struct *hd = gp->part[n];
680
681 if (hd && hd->nr_sects)
682 seq_printf(s, "%4d %4d %s %u %u %u %u\n",
683 gp->major, n + gp->first_minor + 1,
684 disk_name(gp, n + 1, buf),
Jens Axboea3623572005-11-01 09:26:16 +0100685 hd->ios[0], hd->sectors[0],
686 hd->ios[1], hd->sectors[1]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 }
688
689 return 0;
690}
691
692struct seq_operations diskstats_op = {
693 .start = diskstats_start,
694 .next = diskstats_next,
695 .stop = diskstats_stop,
696 .show = diskstats_show
697};
698
Kristen Carlson Accardi8ce7ad72007-05-23 13:57:38 -0700699static void media_change_notify_thread(struct work_struct *work)
700{
701 struct gendisk *gd = container_of(work, struct gendisk, async_notify);
702 char event[] = "MEDIA_CHANGE=1";
703 char *envp[] = { event, NULL };
704
705 /*
706 * set enviroment vars to indicate which event this is for
707 * so that user space will know to go check the media status.
708 */
709 kobject_uevent_env(&gd->kobj, KOBJ_CHANGE, envp);
710 put_device(gd->driverfs_dev);
711}
712
713void genhd_media_change_notify(struct gendisk *disk)
714{
715 get_device(disk->driverfs_dev);
716 schedule_work(&disk->async_notify);
717}
718EXPORT_SYMBOL_GPL(genhd_media_change_notify);
719
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720struct gendisk *alloc_disk(int minors)
721{
Christoph Lameter19460892005-06-23 00:08:19 -0700722 return alloc_disk_node(minors, -1);
723}
724
725struct gendisk *alloc_disk_node(int minors, int node_id)
726{
727 struct gendisk *disk;
728
729 disk = kmalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 if (disk) {
731 memset(disk, 0, sizeof(struct gendisk));
732 if (!init_disk_stats(disk)) {
733 kfree(disk);
734 return NULL;
735 }
736 if (minors > 1) {
737 int size = (minors - 1) * sizeof(struct hd_struct *);
Christoph Lameter19460892005-06-23 00:08:19 -0700738 disk->part = kmalloc_node(size, GFP_KERNEL, node_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739 if (!disk->part) {
740 kfree(disk);
741 return NULL;
742 }
743 memset(disk->part, 0, size);
744 }
745 disk->minors = minors;
746 kobj_set_kset_s(disk,block_subsys);
747 kobject_init(&disk->kobj);
748 rand_initialize_disk(disk);
Kristen Carlson Accardi8ce7ad72007-05-23 13:57:38 -0700749 INIT_WORK(&disk->async_notify,
750 media_change_notify_thread);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751 }
752 return disk;
753}
754
755EXPORT_SYMBOL(alloc_disk);
Christoph Lameter19460892005-06-23 00:08:19 -0700756EXPORT_SYMBOL(alloc_disk_node);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757
758struct kobject *get_disk(struct gendisk *disk)
759{
760 struct module *owner;
761 struct kobject *kobj;
762
763 if (!disk->fops)
764 return NULL;
765 owner = disk->fops->owner;
766 if (owner && !try_module_get(owner))
767 return NULL;
768 kobj = kobject_get(&disk->kobj);
769 if (kobj == NULL) {
770 module_put(owner);
771 return NULL;
772 }
773 return kobj;
774
775}
776
777EXPORT_SYMBOL(get_disk);
778
779void put_disk(struct gendisk *disk)
780{
781 if (disk)
782 kobject_put(&disk->kobj);
783}
784
785EXPORT_SYMBOL(put_disk);
786
787void set_device_ro(struct block_device *bdev, int flag)
788{
789 if (bdev->bd_contains != bdev)
790 bdev->bd_part->policy = flag;
791 else
792 bdev->bd_disk->policy = flag;
793}
794
795EXPORT_SYMBOL(set_device_ro);
796
797void set_disk_ro(struct gendisk *disk, int flag)
798{
799 int i;
800 disk->policy = flag;
801 for (i = 0; i < disk->minors - 1; i++)
802 if (disk->part[i]) disk->part[i]->policy = flag;
803}
804
805EXPORT_SYMBOL(set_disk_ro);
806
807int bdev_read_only(struct block_device *bdev)
808{
809 if (!bdev)
810 return 0;
811 else if (bdev->bd_contains != bdev)
812 return bdev->bd_part->policy;
813 else
814 return bdev->bd_disk->policy;
815}
816
817EXPORT_SYMBOL(bdev_read_only);
818
819int invalidate_partition(struct gendisk *disk, int index)
820{
821 int res = 0;
822 struct block_device *bdev = bdget_disk(disk, index);
823 if (bdev) {
Christoph Hellwig2ef41632005-05-05 16:15:59 -0700824 fsync_bdev(bdev);
825 res = __invalidate_device(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826 bdput(bdev);
827 }
828 return res;
829}
830
831EXPORT_SYMBOL(invalidate_partition);