5 #include <linux/module.h>
7 #include <linux/genhd.h>
8 #include <linux/kdev_t.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 struct kset block_subsys
;
21 static DEFINE_MUTEX(block_subsys_lock
);
24 * Can be deleted altogether. Later.
27 static struct blk_major_name
{
28 struct blk_major_name
*next
;
31 } *major_names
[BLKDEV_MAJOR_HASH_SIZE
];
33 /* index in the above - for now: assume no multimajor ranges */
34 static inline int major_to_index(int major
)
36 return major
% BLKDEV_MAJOR_HASH_SIZE
;
41 void blkdev_show(struct seq_file
*f
, off_t offset
)
43 struct blk_major_name
*dp
;
45 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
);
53 #endif /* CONFIG_PROC_FS */
55 int register_blkdev(unsigned int major
, const char *name
)
57 struct blk_major_name
**n
, *p
;
60 mutex_lock(&block_subsys_lock
);
64 for (index
= ARRAY_SIZE(major_names
)-1; index
> 0; index
--) {
65 if (major_names
[index
] == NULL
)
70 printk("register_blkdev: failed to get major for %s\n",
79 p
= kmalloc(sizeof(struct blk_major_name
), GFP_KERNEL
);
86 strlcpy(p
->name
, name
, sizeof(p
->name
));
88 index
= major_to_index(major
);
90 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
) {
91 if ((*n
)->major
== major
)
100 printk("register_blkdev: cannot get major %d for %s\n",
105 mutex_unlock(&block_subsys_lock
);
109 EXPORT_SYMBOL(register_blkdev
);
111 void unregister_blkdev(unsigned int major
, const char *name
)
113 struct blk_major_name
**n
;
114 struct blk_major_name
*p
= NULL
;
115 int index
= major_to_index(major
);
117 mutex_lock(&block_subsys_lock
);
118 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
)
119 if ((*n
)->major
== major
)
121 if (!*n
|| strcmp((*n
)->name
, name
)) {
127 mutex_unlock(&block_subsys_lock
);
131 EXPORT_SYMBOL(unregister_blkdev
);
133 static struct kobj_map
*bdev_map
;
136 * Register device numbers dev..(dev+range-1)
137 * range must be nonzero
138 * The hash chain is sorted on range, so that subranges can override.
140 void blk_register_region(dev_t dev
, unsigned long range
, struct module
*module
,
141 struct kobject
*(*probe
)(dev_t
, int *, void *),
142 int (*lock
)(dev_t
, void *), void *data
)
144 kobj_map(bdev_map
, dev
, range
, module
, probe
, lock
, data
);
147 EXPORT_SYMBOL(blk_register_region
);
149 void blk_unregister_region(dev_t dev
, unsigned long range
)
151 kobj_unmap(bdev_map
, dev
, range
);
154 EXPORT_SYMBOL(blk_unregister_region
);
156 static struct kobject
*exact_match(dev_t dev
, int *part
, void *data
)
158 struct gendisk
*p
= data
;
162 static int exact_lock(dev_t dev
, void *data
)
164 struct gendisk
*p
= data
;
172 * add_disk - add partitioning information to kernel list
173 * @disk: per-device partitioning information
175 * This function registers the partitioning information in @disk
178 void add_disk(struct gendisk
*disk
)
180 disk
->flags
|= GENHD_FL_UP
;
181 blk_register_region(MKDEV(disk
->major
, disk
->first_minor
),
182 disk
->minors
, NULL
, exact_match
, exact_lock
, disk
);
184 blk_register_queue(disk
);
187 EXPORT_SYMBOL(add_disk
);
188 EXPORT_SYMBOL(del_gendisk
); /* in partitions/check.c */
190 void unlink_gendisk(struct gendisk
*disk
)
192 blk_unregister_queue(disk
);
193 blk_unregister_region(MKDEV(disk
->major
, disk
->first_minor
),
197 #define to_disk(obj) container_of(obj,struct gendisk,kobj)
200 * get_gendisk - get partitioning information for a given device
201 * @dev: device to get partitioning information for
203 * This function gets the structure containing partitioning
204 * information for the given device @dev.
206 struct gendisk
*get_gendisk(dev_t dev
, int *part
)
208 struct kobject
*kobj
= kobj_lookup(bdev_map
, dev
, part
);
209 return kobj
? to_disk(kobj
) : NULL
;
213 * print a full list of all partitions - intended for places where the root
214 * filesystem can't be mounted and thus to give the victim some idea of what
217 void __init
printk_all_partitions(void)
222 mutex_lock(&block_subsys_lock
);
223 /* For each block device... */
224 list_for_each_entry(sgp
, &block_subsys
.list
, kobj
.entry
) {
225 char buf
[BDEVNAME_SIZE
];
227 * Don't show empty devices or things that have been surpressed
229 if (get_capacity(sgp
) == 0 ||
230 (sgp
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
))
234 * Note, unlike /proc/partitions, I am showing the numbers in
235 * hex - the same format as the root= option takes.
237 printk("%02x%02x %10llu %s",
238 sgp
->major
, sgp
->first_minor
,
239 (unsigned long long)get_capacity(sgp
) >> 1,
240 disk_name(sgp
, 0, buf
));
241 if (sgp
->driverfs_dev
!= NULL
&&
242 sgp
->driverfs_dev
->driver
!= NULL
)
243 printk(" driver: %s\n",
244 sgp
->driverfs_dev
->driver
->name
);
246 printk(" (driver?)\n");
248 /* now show the partitions */
249 for (n
= 0; n
< sgp
->minors
- 1; ++n
) {
250 if (sgp
->part
[n
] == NULL
)
252 if (sgp
->part
[n
]->nr_sects
== 0)
254 printk(" %02x%02x %10llu %s\n",
255 sgp
->major
, n
+ 1 + sgp
->first_minor
,
256 (unsigned long long)sgp
->part
[n
]->nr_sects
>> 1,
257 disk_name(sgp
, n
+ 1, buf
));
258 } /* partition subloop */
259 } /* Block device loop */
261 mutex_unlock(&block_subsys_lock
);
265 #ifdef CONFIG_PROC_FS
267 static void *part_start(struct seq_file
*part
, loff_t
*pos
)
272 mutex_lock(&block_subsys_lock
);
273 list_for_each(p
, &block_subsys
.list
)
275 return list_entry(p
, struct gendisk
, kobj
.entry
);
279 static void *part_next(struct seq_file
*part
, void *v
, loff_t
*pos
)
281 struct list_head
*p
= ((struct gendisk
*)v
)->kobj
.entry
.next
;
283 return p
==&block_subsys
.list
? NULL
:
284 list_entry(p
, struct gendisk
, kobj
.entry
);
287 static void part_stop(struct seq_file
*part
, void *v
)
289 mutex_unlock(&block_subsys_lock
);
292 static int show_partition(struct seq_file
*part
, void *v
)
294 struct gendisk
*sgp
= v
;
296 char buf
[BDEVNAME_SIZE
];
298 if (&sgp
->kobj
.entry
== block_subsys
.list
.next
)
299 seq_puts(part
, "major minor #blocks name\n\n");
301 /* Don't show non-partitionable removeable devices or empty devices */
302 if (!get_capacity(sgp
) ||
303 (sgp
->minors
== 1 && (sgp
->flags
& GENHD_FL_REMOVABLE
)))
305 if (sgp
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
)
308 /* show the full disk and all non-0 size partitions of it */
309 seq_printf(part
, "%4d %4d %10llu %s\n",
310 sgp
->major
, sgp
->first_minor
,
311 (unsigned long long)get_capacity(sgp
) >> 1,
312 disk_name(sgp
, 0, buf
));
313 for (n
= 0; n
< sgp
->minors
- 1; n
++) {
316 if (sgp
->part
[n
]->nr_sects
== 0)
318 seq_printf(part
, "%4d %4d %10llu %s\n",
319 sgp
->major
, n
+ 1 + sgp
->first_minor
,
320 (unsigned long long)sgp
->part
[n
]->nr_sects
>> 1 ,
321 disk_name(sgp
, n
+ 1, buf
));
327 struct seq_operations partitions_op
= {
331 .show
= show_partition
336 extern int blk_dev_init(void);
338 static struct kobject
*base_probe(dev_t dev
, int *part
, void *data
)
340 if (request_module("block-major-%d-%d", MAJOR(dev
), MINOR(dev
)) > 0)
341 /* Make old-style 2.4 aliases work */
342 request_module("block-major-%d", MAJOR(dev
));
346 static int __init
genhd_device_init(void)
350 bdev_map
= kobj_map_init(base_probe
, &block_subsys_lock
);
352 err
= subsystem_register(&block_subsys
);
354 printk(KERN_WARNING
"%s: subsystem_register error: %d\n",
359 subsys_initcall(genhd_device_init
);
364 * kobject & sysfs bindings for block devices
366 static ssize_t
disk_attr_show(struct kobject
*kobj
, struct attribute
*attr
,
369 struct gendisk
*disk
= to_disk(kobj
);
370 struct disk_attribute
*disk_attr
=
371 container_of(attr
,struct disk_attribute
,attr
);
375 ret
= disk_attr
->show(disk
,page
);
379 static ssize_t
disk_attr_store(struct kobject
* kobj
, struct attribute
* attr
,
380 const char *page
, size_t count
)
382 struct gendisk
*disk
= to_disk(kobj
);
383 struct disk_attribute
*disk_attr
=
384 container_of(attr
,struct disk_attribute
,attr
);
387 if (disk_attr
->store
)
388 ret
= disk_attr
->store(disk
, page
, count
);
392 static struct sysfs_ops disk_sysfs_ops
= {
393 .show
= &disk_attr_show
,
394 .store
= &disk_attr_store
,
397 static ssize_t
disk_uevent_store(struct gendisk
* disk
,
398 const char *buf
, size_t count
)
400 kobject_uevent(&disk
->kobj
, KOBJ_ADD
);
403 static ssize_t
disk_dev_read(struct gendisk
* disk
, char *page
)
405 dev_t base
= MKDEV(disk
->major
, disk
->first_minor
);
406 return print_dev_t(page
, base
);
408 static ssize_t
disk_range_read(struct gendisk
* disk
, char *page
)
410 return sprintf(page
, "%d\n", disk
->minors
);
412 static ssize_t
disk_removable_read(struct gendisk
* disk
, char *page
)
414 return sprintf(page
, "%d\n",
415 (disk
->flags
& GENHD_FL_REMOVABLE
? 1 : 0));
418 static ssize_t
disk_size_read(struct gendisk
* disk
, char *page
)
420 return sprintf(page
, "%llu\n", (unsigned long long)get_capacity(disk
));
422 static ssize_t
disk_capability_read(struct gendisk
*disk
, char *page
)
424 return sprintf(page
, "%x\n", disk
->flags
);
426 static ssize_t
disk_stats_read(struct gendisk
* disk
, char *page
)
429 disk_round_stats(disk
);
432 "%8lu %8lu %8llu %8u "
433 "%8lu %8lu %8llu %8u "
436 disk_stat_read(disk
, ios
[READ
]),
437 disk_stat_read(disk
, merges
[READ
]),
438 (unsigned long long)disk_stat_read(disk
, sectors
[READ
]),
439 jiffies_to_msecs(disk_stat_read(disk
, ticks
[READ
])),
440 disk_stat_read(disk
, ios
[WRITE
]),
441 disk_stat_read(disk
, merges
[WRITE
]),
442 (unsigned long long)disk_stat_read(disk
, sectors
[WRITE
]),
443 jiffies_to_msecs(disk_stat_read(disk
, ticks
[WRITE
])),
445 jiffies_to_msecs(disk_stat_read(disk
, io_ticks
)),
446 jiffies_to_msecs(disk_stat_read(disk
, time_in_queue
)));
448 static struct disk_attribute disk_attr_uevent
= {
449 .attr
= {.name
= "uevent", .mode
= S_IWUSR
},
450 .store
= disk_uevent_store
452 static struct disk_attribute disk_attr_dev
= {
453 .attr
= {.name
= "dev", .mode
= S_IRUGO
},
454 .show
= disk_dev_read
456 static struct disk_attribute disk_attr_range
= {
457 .attr
= {.name
= "range", .mode
= S_IRUGO
},
458 .show
= disk_range_read
460 static struct disk_attribute disk_attr_removable
= {
461 .attr
= {.name
= "removable", .mode
= S_IRUGO
},
462 .show
= disk_removable_read
464 static struct disk_attribute disk_attr_size
= {
465 .attr
= {.name
= "size", .mode
= S_IRUGO
},
466 .show
= disk_size_read
468 static struct disk_attribute disk_attr_capability
= {
469 .attr
= {.name
= "capability", .mode
= S_IRUGO
},
470 .show
= disk_capability_read
472 static struct disk_attribute disk_attr_stat
= {
473 .attr
= {.name
= "stat", .mode
= S_IRUGO
},
474 .show
= disk_stats_read
477 #ifdef CONFIG_FAIL_MAKE_REQUEST
479 static ssize_t
disk_fail_store(struct gendisk
* disk
,
480 const char *buf
, size_t count
)
484 if (count
> 0 && sscanf(buf
, "%d", &i
) > 0) {
486 disk
->flags
&= ~GENHD_FL_FAIL
;
488 disk
->flags
|= GENHD_FL_FAIL
;
493 static ssize_t
disk_fail_read(struct gendisk
* disk
, char *page
)
495 return sprintf(page
, "%d\n", disk
->flags
& GENHD_FL_FAIL
? 1 : 0);
497 static struct disk_attribute disk_attr_fail
= {
498 .attr
= {.name
= "make-it-fail", .mode
= S_IRUGO
| S_IWUSR
},
499 .store
= disk_fail_store
,
500 .show
= disk_fail_read
505 static struct attribute
* default_attrs
[] = {
506 &disk_attr_uevent
.attr
,
508 &disk_attr_range
.attr
,
509 &disk_attr_removable
.attr
,
510 &disk_attr_size
.attr
,
511 &disk_attr_stat
.attr
,
512 &disk_attr_capability
.attr
,
513 #ifdef CONFIG_FAIL_MAKE_REQUEST
514 &disk_attr_fail
.attr
,
519 static void disk_release(struct kobject
* kobj
)
521 struct gendisk
*disk
= to_disk(kobj
);
524 free_disk_stats(disk
);
528 static struct kobj_type ktype_block
= {
529 .release
= disk_release
,
530 .sysfs_ops
= &disk_sysfs_ops
,
531 .default_attrs
= default_attrs
,
534 extern struct kobj_type ktype_part
;
536 static int block_uevent_filter(struct kset
*kset
, struct kobject
*kobj
)
538 struct kobj_type
*ktype
= get_ktype(kobj
);
540 return ((ktype
== &ktype_block
) || (ktype
== &ktype_part
));
543 static int block_uevent(struct kset
*kset
, struct kobject
*kobj
, char **envp
,
544 int num_envp
, char *buffer
, int buffer_size
)
546 struct kobj_type
*ktype
= get_ktype(kobj
);
547 struct device
*physdev
;
548 struct gendisk
*disk
;
549 struct hd_struct
*part
;
553 if (ktype
== &ktype_block
) {
554 disk
= container_of(kobj
, struct gendisk
, kobj
);
555 add_uevent_var(envp
, num_envp
, &i
, buffer
, buffer_size
,
556 &length
, "MINOR=%u", disk
->first_minor
);
557 } else if (ktype
== &ktype_part
) {
558 disk
= container_of(kobj
->parent
, struct gendisk
, kobj
);
559 part
= container_of(kobj
, struct hd_struct
, kobj
);
560 add_uevent_var(envp
, num_envp
, &i
, buffer
, buffer_size
,
562 disk
->first_minor
+ part
->partno
);
566 add_uevent_var(envp
, num_envp
, &i
, buffer
, buffer_size
, &length
,
567 "MAJOR=%u", disk
->major
);
569 /* add physical device, backing this device */
570 physdev
= disk
->driverfs_dev
;
572 char *path
= kobject_get_path(&physdev
->kobj
, GFP_KERNEL
);
574 add_uevent_var(envp
, num_envp
, &i
, buffer
, buffer_size
,
575 &length
, "PHYSDEVPATH=%s", path
);
579 add_uevent_var(envp
, num_envp
, &i
,
580 buffer
, buffer_size
, &length
,
585 add_uevent_var(envp
, num_envp
, &i
,
586 buffer
, buffer_size
, &length
,
588 physdev
->driver
->name
);
591 /* terminate, set to next free slot, shrink available space */
595 buffer
= &buffer
[length
];
596 buffer_size
-= length
;
601 static struct kset_uevent_ops block_uevent_ops
= {
602 .filter
= block_uevent_filter
,
603 .uevent
= block_uevent
,
606 decl_subsys(block
, &ktype_block
, &block_uevent_ops
);
609 * aggregate disk stat collector. Uses the same stats that the sysfs
610 * entries do, above, but makes them available through one seq_file.
611 * Watching a few disks may be efficient through sysfs, but watching
612 * all of them will be more efficient through this interface.
614 * The output looks suspiciously like /proc/partitions with a bunch of
619 static void *diskstats_start(struct seq_file
*part
, loff_t
*pos
)
624 mutex_lock(&block_subsys_lock
);
625 list_for_each(p
, &block_subsys
.list
)
627 return list_entry(p
, struct gendisk
, kobj
.entry
);
631 static void *diskstats_next(struct seq_file
*part
, void *v
, loff_t
*pos
)
633 struct list_head
*p
= ((struct gendisk
*)v
)->kobj
.entry
.next
;
635 return p
==&block_subsys
.list
? NULL
:
636 list_entry(p
, struct gendisk
, kobj
.entry
);
639 static void diskstats_stop(struct seq_file
*part
, void *v
)
641 mutex_unlock(&block_subsys_lock
);
644 static int diskstats_show(struct seq_file
*s
, void *v
)
646 struct gendisk
*gp
= v
;
647 char buf
[BDEVNAME_SIZE
];
651 if (&sgp->kobj.entry == block_subsys.kset.list.next)
652 seq_puts(s, "major minor name"
653 " rio rmerge rsect ruse wio wmerge "
654 "wsect wuse running use aveq"
659 disk_round_stats(gp
);
661 seq_printf(s
, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n",
662 gp
->major
, n
+ gp
->first_minor
, disk_name(gp
, n
, buf
),
663 disk_stat_read(gp
, ios
[0]), disk_stat_read(gp
, merges
[0]),
664 (unsigned long long)disk_stat_read(gp
, sectors
[0]),
665 jiffies_to_msecs(disk_stat_read(gp
, ticks
[0])),
666 disk_stat_read(gp
, ios
[1]), disk_stat_read(gp
, merges
[1]),
667 (unsigned long long)disk_stat_read(gp
, sectors
[1]),
668 jiffies_to_msecs(disk_stat_read(gp
, ticks
[1])),
670 jiffies_to_msecs(disk_stat_read(gp
, io_ticks
)),
671 jiffies_to_msecs(disk_stat_read(gp
, time_in_queue
)));
673 /* now show all non-0 size partitions of it */
674 for (n
= 0; n
< gp
->minors
- 1; n
++) {
675 struct hd_struct
*hd
= gp
->part
[n
];
677 if (hd
&& hd
->nr_sects
)
678 seq_printf(s
, "%4d %4d %s %u %u %u %u\n",
679 gp
->major
, n
+ gp
->first_minor
+ 1,
680 disk_name(gp
, n
+ 1, buf
),
681 hd
->ios
[0], hd
->sectors
[0],
682 hd
->ios
[1], hd
->sectors
[1]);
688 struct seq_operations diskstats_op
= {
689 .start
= diskstats_start
,
690 .next
= diskstats_next
,
691 .stop
= diskstats_stop
,
692 .show
= diskstats_show
695 static void media_change_notify_thread(struct work_struct
*work
)
697 struct gendisk
*gd
= container_of(work
, struct gendisk
, async_notify
);
698 char event
[] = "MEDIA_CHANGE=1";
699 char *envp
[] = { event
, NULL
};
702 * set enviroment vars to indicate which event this is for
703 * so that user space will know to go check the media status.
705 kobject_uevent_env(&gd
->kobj
, KOBJ_CHANGE
, envp
);
706 put_device(gd
->driverfs_dev
);
709 void genhd_media_change_notify(struct gendisk
*disk
)
711 get_device(disk
->driverfs_dev
);
712 schedule_work(&disk
->async_notify
);
714 EXPORT_SYMBOL_GPL(genhd_media_change_notify
);
716 struct gendisk
*alloc_disk(int minors
)
718 return alloc_disk_node(minors
, -1);
721 struct gendisk
*alloc_disk_node(int minors
, int node_id
)
723 struct gendisk
*disk
;
725 disk
= kmalloc_node(sizeof(struct gendisk
),
726 GFP_KERNEL
| __GFP_ZERO
, node_id
);
728 if (!init_disk_stats(disk
)) {
733 int size
= (minors
- 1) * sizeof(struct hd_struct
*);
734 disk
->part
= kmalloc_node(size
,
735 GFP_KERNEL
| __GFP_ZERO
, node_id
);
741 disk
->minors
= minors
;
742 kobj_set_kset_s(disk
,block_subsys
);
743 kobject_init(&disk
->kobj
);
744 rand_initialize_disk(disk
);
745 INIT_WORK(&disk
->async_notify
,
746 media_change_notify_thread
);
751 EXPORT_SYMBOL(alloc_disk
);
752 EXPORT_SYMBOL(alloc_disk_node
);
754 struct kobject
*get_disk(struct gendisk
*disk
)
756 struct module
*owner
;
757 struct kobject
*kobj
;
761 owner
= disk
->fops
->owner
;
762 if (owner
&& !try_module_get(owner
))
764 kobj
= kobject_get(&disk
->kobj
);
773 EXPORT_SYMBOL(get_disk
);
775 void put_disk(struct gendisk
*disk
)
778 kobject_put(&disk
->kobj
);
781 EXPORT_SYMBOL(put_disk
);
783 void set_device_ro(struct block_device
*bdev
, int flag
)
785 if (bdev
->bd_contains
!= bdev
)
786 bdev
->bd_part
->policy
= flag
;
788 bdev
->bd_disk
->policy
= flag
;
791 EXPORT_SYMBOL(set_device_ro
);
793 void set_disk_ro(struct gendisk
*disk
, int flag
)
797 for (i
= 0; i
< disk
->minors
- 1; i
++)
798 if (disk
->part
[i
]) disk
->part
[i
]->policy
= flag
;
801 EXPORT_SYMBOL(set_disk_ro
);
803 int bdev_read_only(struct block_device
*bdev
)
807 else if (bdev
->bd_contains
!= bdev
)
808 return bdev
->bd_part
->policy
;
810 return bdev
->bd_disk
->policy
;
813 EXPORT_SYMBOL(bdev_read_only
);
815 int invalidate_partition(struct gendisk
*disk
, int index
)
818 struct block_device
*bdev
= bdget_disk(disk
, index
);
821 res
= __invalidate_device(bdev
);
827 EXPORT_SYMBOL(invalidate_partition
);