5 #include <linux/config.h>
6 #include <linux/module.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>
17 #include <linux/buffer_head.h>
18 #include <linux/mutex.h>
20 static struct subsystem block_subsys
;
22 static DEFINE_MUTEX(block_subsys_lock
);
25 * Can be deleted altogether. Later.
28 static struct blk_major_name
{
29 struct blk_major_name
*next
;
32 } *major_names
[BLKDEV_MAJOR_HASH_SIZE
];
34 /* index in the above - for now: assume no multimajor ranges */
35 static inline int major_to_index(int major
)
37 return major
% BLKDEV_MAJOR_HASH_SIZE
;
42 void blkdev_show(struct seq_file
*f
, off_t offset
)
44 struct blk_major_name
*dp
;
46 if (offset
< BLKDEV_MAJOR_HASH_SIZE
) {
47 mutex_lock(&block_subsys_lock
);
48 for (dp
= major_names
[offset
]; dp
; dp
= dp
->next
)
49 seq_printf(f
, "%3d %s\n", dp
->major
, dp
->name
);
50 mutex_unlock(&block_subsys_lock
);
54 #endif /* CONFIG_PROC_FS */
56 int register_blkdev(unsigned int major
, const char *name
)
58 struct blk_major_name
**n
, *p
;
61 mutex_lock(&block_subsys_lock
);
65 for (index
= ARRAY_SIZE(major_names
)-1; index
> 0; index
--) {
66 if (major_names
[index
] == NULL
)
71 printk("register_blkdev: failed to get major for %s\n",
80 p
= kmalloc(sizeof(struct blk_major_name
), GFP_KERNEL
);
87 strlcpy(p
->name
, name
, sizeof(p
->name
));
89 index
= major_to_index(major
);
91 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
) {
92 if ((*n
)->major
== major
)
101 printk("register_blkdev: cannot get major %d for %s\n",
106 mutex_unlock(&block_subsys_lock
);
110 EXPORT_SYMBOL(register_blkdev
);
112 /* todo: make void - error printk here */
113 int unregister_blkdev(unsigned int major
, const char *name
)
115 struct blk_major_name
**n
;
116 struct blk_major_name
*p
= NULL
;
117 int index
= major_to_index(major
);
120 mutex_lock(&block_subsys_lock
);
121 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
)
122 if ((*n
)->major
== major
)
124 if (!*n
|| strcmp((*n
)->name
, name
))
130 mutex_unlock(&block_subsys_lock
);
136 EXPORT_SYMBOL(unregister_blkdev
);
138 static struct kobj_map
*bdev_map
;
141 * Register device numbers dev..(dev+range-1)
142 * range must be nonzero
143 * The hash chain is sorted on range, so that subranges can override.
145 void blk_register_region(dev_t dev
, unsigned long range
, struct module
*module
,
146 struct kobject
*(*probe
)(dev_t
, int *, void *),
147 int (*lock
)(dev_t
, void *), void *data
)
149 kobj_map(bdev_map
, dev
, range
, module
, probe
, lock
, data
);
152 EXPORT_SYMBOL(blk_register_region
);
154 void blk_unregister_region(dev_t dev
, unsigned long range
)
156 kobj_unmap(bdev_map
, dev
, range
);
159 EXPORT_SYMBOL(blk_unregister_region
);
161 static struct kobject
*exact_match(dev_t dev
, int *part
, void *data
)
163 struct gendisk
*p
= data
;
167 static int exact_lock(dev_t dev
, void *data
)
169 struct gendisk
*p
= data
;
177 * add_disk - add partitioning information to kernel list
178 * @disk: per-device partitioning information
180 * This function registers the partitioning information in @disk
183 void add_disk(struct gendisk
*disk
)
185 get_device(disk
->driverfs_dev
);
186 disk
->flags
|= GENHD_FL_UP
;
187 blk_register_region(MKDEV(disk
->major
, disk
->first_minor
),
188 disk
->minors
, NULL
, exact_match
, exact_lock
, disk
);
190 blk_register_queue(disk
);
193 EXPORT_SYMBOL(add_disk
);
194 EXPORT_SYMBOL(del_gendisk
); /* in partitions/check.c */
196 void unlink_gendisk(struct gendisk
*disk
)
198 blk_unregister_queue(disk
);
199 blk_unregister_region(MKDEV(disk
->major
, disk
->first_minor
),
203 #define to_disk(obj) container_of(obj,struct gendisk,kobj)
206 * get_gendisk - get partitioning information for a given device
207 * @dev: device to get partitioning information for
209 * This function gets the structure containing partitioning
210 * information for the given device @dev.
212 struct gendisk
*get_gendisk(dev_t dev
, int *part
)
214 struct kobject
*kobj
= kobj_lookup(bdev_map
, dev
, part
);
215 return kobj
? to_disk(kobj
) : NULL
;
218 #ifdef CONFIG_PROC_FS
220 static void *part_start(struct seq_file
*part
, loff_t
*pos
)
225 mutex_lock(&block_subsys_lock
);
226 list_for_each(p
, &block_subsys
.kset
.list
)
228 return list_entry(p
, struct gendisk
, kobj
.entry
);
232 static void *part_next(struct seq_file
*part
, void *v
, loff_t
*pos
)
234 struct list_head
*p
= ((struct gendisk
*)v
)->kobj
.entry
.next
;
236 return p
==&block_subsys
.kset
.list
? NULL
:
237 list_entry(p
, struct gendisk
, kobj
.entry
);
240 static void part_stop(struct seq_file
*part
, void *v
)
242 mutex_unlock(&block_subsys_lock
);
245 static int show_partition(struct seq_file
*part
, void *v
)
247 struct gendisk
*sgp
= v
;
249 char buf
[BDEVNAME_SIZE
];
251 if (&sgp
->kobj
.entry
== block_subsys
.kset
.list
.next
)
252 seq_puts(part
, "major minor #blocks name\n\n");
254 /* Don't show non-partitionable removeable devices or empty devices */
255 if (!get_capacity(sgp
) ||
256 (sgp
->minors
== 1 && (sgp
->flags
& GENHD_FL_REMOVABLE
)))
258 if (sgp
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
)
261 /* show the full disk and all non-0 size partitions of it */
262 seq_printf(part
, "%4d %4d %10llu %s\n",
263 sgp
->major
, sgp
->first_minor
,
264 (unsigned long long)get_capacity(sgp
) >> 1,
265 disk_name(sgp
, 0, buf
));
266 for (n
= 0; n
< sgp
->minors
- 1; n
++) {
269 if (sgp
->part
[n
]->nr_sects
== 0)
271 seq_printf(part
, "%4d %4d %10llu %s\n",
272 sgp
->major
, n
+ 1 + sgp
->first_minor
,
273 (unsigned long long)sgp
->part
[n
]->nr_sects
>> 1 ,
274 disk_name(sgp
, n
+ 1, buf
));
280 struct seq_operations partitions_op
= {
284 .show
= show_partition
289 extern int blk_dev_init(void);
291 static struct kobject
*base_probe(dev_t dev
, int *part
, void *data
)
293 if (request_module("block-major-%d-%d", MAJOR(dev
), MINOR(dev
)) > 0)
294 /* Make old-style 2.4 aliases work */
295 request_module("block-major-%d", MAJOR(dev
));
299 static int __init
genhd_device_init(void)
301 bdev_map
= kobj_map_init(base_probe
, &block_subsys_lock
);
303 subsystem_register(&block_subsys
);
307 subsys_initcall(genhd_device_init
);
312 * kobject & sysfs bindings for block devices
314 static ssize_t
disk_attr_show(struct kobject
*kobj
, struct attribute
*attr
,
317 struct gendisk
*disk
= to_disk(kobj
);
318 struct disk_attribute
*disk_attr
=
319 container_of(attr
,struct disk_attribute
,attr
);
323 ret
= disk_attr
->show(disk
,page
);
327 static ssize_t
disk_attr_store(struct kobject
* kobj
, struct attribute
* attr
,
328 const char *page
, size_t count
)
330 struct gendisk
*disk
= to_disk(kobj
);
331 struct disk_attribute
*disk_attr
=
332 container_of(attr
,struct disk_attribute
,attr
);
335 if (disk_attr
->store
)
336 ret
= disk_attr
->store(disk
, page
, count
);
340 static struct sysfs_ops disk_sysfs_ops
= {
341 .show
= &disk_attr_show
,
342 .store
= &disk_attr_store
,
345 static ssize_t
disk_uevent_store(struct gendisk
* disk
,
346 const char *buf
, size_t count
)
348 kobject_uevent(&disk
->kobj
, KOBJ_ADD
);
351 static ssize_t
disk_dev_read(struct gendisk
* disk
, char *page
)
353 dev_t base
= MKDEV(disk
->major
, disk
->first_minor
);
354 return print_dev_t(page
, base
);
356 static ssize_t
disk_range_read(struct gendisk
* disk
, char *page
)
358 return sprintf(page
, "%d\n", disk
->minors
);
360 static ssize_t
disk_removable_read(struct gendisk
* disk
, char *page
)
362 return sprintf(page
, "%d\n",
363 (disk
->flags
& GENHD_FL_REMOVABLE
? 1 : 0));
366 static ssize_t
disk_size_read(struct gendisk
* disk
, char *page
)
368 return sprintf(page
, "%llu\n", (unsigned long long)get_capacity(disk
));
371 static ssize_t
disk_stats_read(struct gendisk
* disk
, char *page
)
374 disk_round_stats(disk
);
377 "%8lu %8lu %8llu %8u "
378 "%8lu %8lu %8llu %8u "
381 disk_stat_read(disk
, ios
[READ
]),
382 disk_stat_read(disk
, merges
[READ
]),
383 (unsigned long long)disk_stat_read(disk
, sectors
[READ
]),
384 jiffies_to_msecs(disk_stat_read(disk
, ticks
[READ
])),
385 disk_stat_read(disk
, ios
[WRITE
]),
386 disk_stat_read(disk
, merges
[WRITE
]),
387 (unsigned long long)disk_stat_read(disk
, sectors
[WRITE
]),
388 jiffies_to_msecs(disk_stat_read(disk
, ticks
[WRITE
])),
390 jiffies_to_msecs(disk_stat_read(disk
, io_ticks
)),
391 jiffies_to_msecs(disk_stat_read(disk
, time_in_queue
)));
393 static struct disk_attribute disk_attr_uevent
= {
394 .attr
= {.name
= "uevent", .mode
= S_IWUSR
},
395 .store
= disk_uevent_store
397 static struct disk_attribute disk_attr_dev
= {
398 .attr
= {.name
= "dev", .mode
= S_IRUGO
},
399 .show
= disk_dev_read
401 static struct disk_attribute disk_attr_range
= {
402 .attr
= {.name
= "range", .mode
= S_IRUGO
},
403 .show
= disk_range_read
405 static struct disk_attribute disk_attr_removable
= {
406 .attr
= {.name
= "removable", .mode
= S_IRUGO
},
407 .show
= disk_removable_read
409 static struct disk_attribute disk_attr_size
= {
410 .attr
= {.name
= "size", .mode
= S_IRUGO
},
411 .show
= disk_size_read
413 static struct disk_attribute disk_attr_stat
= {
414 .attr
= {.name
= "stat", .mode
= S_IRUGO
},
415 .show
= disk_stats_read
418 static struct attribute
* default_attrs
[] = {
419 &disk_attr_uevent
.attr
,
421 &disk_attr_range
.attr
,
422 &disk_attr_removable
.attr
,
423 &disk_attr_size
.attr
,
424 &disk_attr_stat
.attr
,
428 static void disk_release(struct kobject
* kobj
)
430 struct gendisk
*disk
= to_disk(kobj
);
431 put_device(disk
->driverfs_dev
);
434 free_disk_stats(disk
);
438 static struct kobj_type ktype_block
= {
439 .release
= disk_release
,
440 .sysfs_ops
= &disk_sysfs_ops
,
441 .default_attrs
= default_attrs
,
444 extern struct kobj_type ktype_part
;
446 static int block_uevent_filter(struct kset
*kset
, struct kobject
*kobj
)
448 struct kobj_type
*ktype
= get_ktype(kobj
);
450 return ((ktype
== &ktype_block
) || (ktype
== &ktype_part
));
453 static int block_uevent(struct kset
*kset
, struct kobject
*kobj
, char **envp
,
454 int num_envp
, char *buffer
, int buffer_size
)
456 struct kobj_type
*ktype
= get_ktype(kobj
);
457 struct device
*physdev
;
458 struct gendisk
*disk
;
459 struct hd_struct
*part
;
463 if (ktype
== &ktype_block
) {
464 disk
= container_of(kobj
, struct gendisk
, kobj
);
465 add_uevent_var(envp
, num_envp
, &i
, buffer
, buffer_size
,
466 &length
, "MINOR=%u", disk
->first_minor
);
467 } else if (ktype
== &ktype_part
) {
468 disk
= container_of(kobj
->parent
, struct gendisk
, kobj
);
469 part
= container_of(kobj
, struct hd_struct
, kobj
);
470 add_uevent_var(envp
, num_envp
, &i
, buffer
, buffer_size
,
472 disk
->first_minor
+ part
->partno
);
476 add_uevent_var(envp
, num_envp
, &i
, buffer
, buffer_size
, &length
,
477 "MAJOR=%u", disk
->major
);
479 /* add physical device, backing this device */
480 physdev
= disk
->driverfs_dev
;
482 char *path
= kobject_get_path(&physdev
->kobj
, GFP_KERNEL
);
484 add_uevent_var(envp
, num_envp
, &i
, buffer
, buffer_size
,
485 &length
, "PHYSDEVPATH=%s", path
);
489 add_uevent_var(envp
, num_envp
, &i
,
490 buffer
, buffer_size
, &length
,
495 add_uevent_var(envp
, num_envp
, &i
,
496 buffer
, buffer_size
, &length
,
498 physdev
->driver
->name
);
501 /* terminate, set to next free slot, shrink available space */
505 buffer
= &buffer
[length
];
506 buffer_size
-= length
;
511 static struct kset_uevent_ops block_uevent_ops
= {
512 .filter
= block_uevent_filter
,
513 .uevent
= block_uevent
,
516 /* declare block_subsys. */
517 static decl_subsys(block
, &ktype_block
, &block_uevent_ops
);
521 * aggregate disk stat collector. Uses the same stats that the sysfs
522 * entries do, above, but makes them available through one seq_file.
523 * Watching a few disks may be efficient through sysfs, but watching
524 * all of them will be more efficient through this interface.
526 * The output looks suspiciously like /proc/partitions with a bunch of
531 static void *diskstats_start(struct seq_file
*part
, loff_t
*pos
)
536 mutex_lock(&block_subsys_lock
);
537 list_for_each(p
, &block_subsys
.kset
.list
)
539 return list_entry(p
, struct gendisk
, kobj
.entry
);
543 static void *diskstats_next(struct seq_file
*part
, void *v
, loff_t
*pos
)
545 struct list_head
*p
= ((struct gendisk
*)v
)->kobj
.entry
.next
;
547 return p
==&block_subsys
.kset
.list
? NULL
:
548 list_entry(p
, struct gendisk
, kobj
.entry
);
551 static void diskstats_stop(struct seq_file
*part
, void *v
)
553 mutex_unlock(&block_subsys_lock
);
556 static int diskstats_show(struct seq_file
*s
, void *v
)
558 struct gendisk
*gp
= v
;
559 char buf
[BDEVNAME_SIZE
];
563 if (&sgp->kobj.entry == block_subsys.kset.list.next)
564 seq_puts(s, "major minor name"
565 " rio rmerge rsect ruse wio wmerge "
566 "wsect wuse running use aveq"
571 disk_round_stats(gp
);
573 seq_printf(s
, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n",
574 gp
->major
, n
+ gp
->first_minor
, disk_name(gp
, n
, buf
),
575 disk_stat_read(gp
, ios
[0]), disk_stat_read(gp
, merges
[0]),
576 (unsigned long long)disk_stat_read(gp
, sectors
[0]),
577 jiffies_to_msecs(disk_stat_read(gp
, ticks
[0])),
578 disk_stat_read(gp
, ios
[1]), disk_stat_read(gp
, merges
[1]),
579 (unsigned long long)disk_stat_read(gp
, sectors
[1]),
580 jiffies_to_msecs(disk_stat_read(gp
, ticks
[1])),
582 jiffies_to_msecs(disk_stat_read(gp
, io_ticks
)),
583 jiffies_to_msecs(disk_stat_read(gp
, time_in_queue
)));
585 /* now show all non-0 size partitions of it */
586 for (n
= 0; n
< gp
->minors
- 1; n
++) {
587 struct hd_struct
*hd
= gp
->part
[n
];
589 if (hd
&& hd
->nr_sects
)
590 seq_printf(s
, "%4d %4d %s %u %u %u %u\n",
591 gp
->major
, n
+ gp
->first_minor
+ 1,
592 disk_name(gp
, n
+ 1, buf
),
593 hd
->ios
[0], hd
->sectors
[0],
594 hd
->ios
[1], hd
->sectors
[1]);
600 struct seq_operations diskstats_op
= {
601 .start
= diskstats_start
,
602 .next
= diskstats_next
,
603 .stop
= diskstats_stop
,
604 .show
= diskstats_show
607 struct gendisk
*alloc_disk(int minors
)
609 return alloc_disk_node(minors
, -1);
612 struct gendisk
*alloc_disk_node(int minors
, int node_id
)
614 struct gendisk
*disk
;
616 disk
= kmalloc_node(sizeof(struct gendisk
), GFP_KERNEL
, node_id
);
618 memset(disk
, 0, sizeof(struct gendisk
));
619 if (!init_disk_stats(disk
)) {
624 int size
= (minors
- 1) * sizeof(struct hd_struct
*);
625 disk
->part
= kmalloc_node(size
, GFP_KERNEL
, node_id
);
630 memset(disk
->part
, 0, size
);
632 disk
->minors
= minors
;
633 kobj_set_kset_s(disk
,block_subsys
);
634 kobject_init(&disk
->kobj
);
635 rand_initialize_disk(disk
);
640 EXPORT_SYMBOL(alloc_disk
);
641 EXPORT_SYMBOL(alloc_disk_node
);
643 struct kobject
*get_disk(struct gendisk
*disk
)
645 struct module
*owner
;
646 struct kobject
*kobj
;
650 owner
= disk
->fops
->owner
;
651 if (owner
&& !try_module_get(owner
))
653 kobj
= kobject_get(&disk
->kobj
);
662 EXPORT_SYMBOL(get_disk
);
664 void put_disk(struct gendisk
*disk
)
667 kobject_put(&disk
->kobj
);
670 EXPORT_SYMBOL(put_disk
);
672 void set_device_ro(struct block_device
*bdev
, int flag
)
674 if (bdev
->bd_contains
!= bdev
)
675 bdev
->bd_part
->policy
= flag
;
677 bdev
->bd_disk
->policy
= flag
;
680 EXPORT_SYMBOL(set_device_ro
);
682 void set_disk_ro(struct gendisk
*disk
, int flag
)
686 for (i
= 0; i
< disk
->minors
- 1; i
++)
687 if (disk
->part
[i
]) disk
->part
[i
]->policy
= flag
;
690 EXPORT_SYMBOL(set_disk_ro
);
692 int bdev_read_only(struct block_device
*bdev
)
696 else if (bdev
->bd_contains
!= bdev
)
697 return bdev
->bd_part
->policy
;
699 return bdev
->bd_disk
->policy
;
702 EXPORT_SYMBOL(bdev_read_only
);
704 int invalidate_partition(struct gendisk
*disk
, int index
)
707 struct block_device
*bdev
= bdget_disk(disk
, index
);
710 res
= __invalidate_device(bdev
);
716 EXPORT_SYMBOL(invalidate_partition
);