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>
18 #define MAX_PROBE_HASH 255 /* random */
20 static struct subsystem block_subsys;
23 * Can be deleted altogether. Later.
25 * Modified under both block_subsys.rwsem and major_names_lock.
27 static struct blk_major_name {
28 struct blk_major_name *next;
31 } *major_names[MAX_PROBE_HASH];
33 static spinlock_t major_names_lock = SPIN_LOCK_UNLOCKED;
35 /* index in the above - for now: assume no multimajor ranges */
36 static inline int major_to_index(int major)
38 return major % MAX_PROBE_HASH;
41 /* get block device names in somewhat random order */
42 int get_blkdev_list(char *p)
44 struct blk_major_name *n;
47 len = sprintf(p, "\nBlock devices:\n");
49 down_read(&block_subsys.rwsem);
50 for (i = 0; i < ARRAY_SIZE(major_names); i++) {
51 for (n = major_names[i]; n; n = n->next)
52 len += sprintf(p+len, "%3d %s\n",
55 up_read(&block_subsys.rwsem);
60 int register_blkdev(unsigned int major, const char *name)
62 struct blk_major_name **n, *p;
66 down_write(&block_subsys.rwsem);
70 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
71 if (major_names[index] == NULL)
76 printk("register_blkdev: failed to get major for %s\n",
85 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
92 strlcpy(p->name, name, sizeof(p->name));
94 index = major_to_index(major);
96 spin_lock_irqsave(&major_names_lock, flags);
97 for (n = &major_names[index]; *n; n = &(*n)->next) {
98 if ((*n)->major == major)
105 spin_unlock_irqrestore(&major_names_lock, flags);
108 printk("register_blkdev: cannot get major %d for %s\n",
113 up_write(&block_subsys.rwsem);
117 /* todo: make void - error printk here */
118 int unregister_blkdev(unsigned int major, const char *name)
120 struct blk_major_name **n;
121 struct blk_major_name *p = NULL;
122 int index = major_to_index(major);
126 down_write(&block_subsys.rwsem);
127 spin_lock_irqsave(&major_names_lock, flags);
128 for (n = &major_names[index]; *n; n = &(*n)->next)
129 if ((*n)->major == major)
131 if (!*n || strcmp((*n)->name, name))
137 spin_unlock_irqrestore(&major_names_lock, flags);
138 up_write(&block_subsys.rwsem);
144 static struct kobj_map *bdev_map;
147 * Register device numbers dev..(dev+range-1)
148 * range must be nonzero
149 * The hash chain is sorted on range, so that subranges can override.
151 void blk_register_region(dev_t dev, unsigned long range, struct module *module,
152 struct kobject *(*probe)(dev_t, int *, void *),
153 int (*lock)(dev_t, void *), void *data)
155 kobj_map(bdev_map, dev, range, module, probe, lock, data);
158 void blk_unregister_region(dev_t dev, unsigned long range)
160 kobj_unmap(bdev_map, dev, range);
163 EXPORT_SYMBOL(blk_register_region);
164 EXPORT_SYMBOL(blk_unregister_region);
166 static struct kobject *exact_match(dev_t dev, int *part, void *data)
168 struct gendisk *p = data;
172 static int exact_lock(dev_t dev, void *data)
174 struct gendisk *p = data;
182 * add_gendisk - add partitioning information to kernel list
183 * @disk: per-device partitioning information
185 * This function registers the partitioning information in @disk
188 void add_disk(struct gendisk *disk)
190 disk->flags |= GENHD_FL_UP;
191 blk_register_region(MKDEV(disk->major, disk->first_minor),
192 disk->minors, NULL, exact_match, exact_lock, disk);
194 blk_register_queue(disk);
197 EXPORT_SYMBOL(add_disk);
198 EXPORT_SYMBOL(del_gendisk); /* in partitions/check.c */
200 void unlink_gendisk(struct gendisk *disk)
202 blk_unregister_queue(disk);
203 blk_unregister_region(MKDEV(disk->major, disk->first_minor),
207 #define to_disk(obj) container_of(obj,struct gendisk,kobj)
210 * get_gendisk - get partitioning information for a given device
211 * @dev: device to get partitioning information for
213 * This function gets the structure containing partitioning
214 * information for the given device @dev.
216 struct gendisk *get_gendisk(dev_t dev, int *part)
218 struct kobject *kobj = kobj_lookup(bdev_map, dev, part);
219 return kobj ? to_disk(kobj) : NULL;
222 #ifdef CONFIG_PROC_FS
224 static void *part_start(struct seq_file *part, loff_t *pos)
229 down_read(&block_subsys.rwsem);
230 list_for_each(p, &block_subsys.kset.list)
232 return list_entry(p, struct gendisk, kobj.entry);
236 static void *part_next(struct seq_file *part, void *v, loff_t *pos)
238 struct list_head *p = ((struct gendisk *)v)->kobj.entry.next;
240 return p==&block_subsys.kset.list ? NULL :
241 list_entry(p, struct gendisk, kobj.entry);
244 static void part_stop(struct seq_file *part, void *v)
246 up_read(&block_subsys.rwsem);
249 static int show_partition(struct seq_file *part, void *v)
251 struct gendisk *sgp = v;
255 if (&sgp->kobj.entry == block_subsys.kset.list.next)
256 seq_puts(part, "major minor #blocks name\n\n");
258 /* Don't show non-partitionable devices or empty devices */
259 if (!get_capacity(sgp) || sgp->minors == 1)
262 /* show the full disk and all non-0 size partitions of it */
263 seq_printf(part, "%4d %4d %10llu %s\n",
264 sgp->major, sgp->first_minor,
265 (unsigned long long)get_capacity(sgp) >> 1,
266 disk_name(sgp, 0, buf));
267 for (n = 0; n < sgp->minors - 1; n++) {
270 if (sgp->part[n]->nr_sects == 0)
272 seq_printf(part, "%4d %4d %10llu %s\n",
273 sgp->major, n + 1 + sgp->first_minor,
274 (unsigned long long)sgp->part[n]->nr_sects >> 1 ,
275 disk_name(sgp, n + 1, buf));
281 struct seq_operations partitions_op = {
285 .show = show_partition
290 extern int blk_dev_init(void);
292 static struct kobject *base_probe(dev_t dev, int *part, void *data)
294 request_module("block-major-%d", MAJOR(dev));
298 int __init device_init(void)
300 bdev_map = kobj_map_init(base_probe, &block_subsys);
302 subsystem_register(&block_subsys);
306 subsys_initcall(device_init);
311 * kobject & sysfs bindings for block devices
314 struct disk_attribute {
315 struct attribute attr;
316 ssize_t (*show)(struct gendisk *, char *);
319 static ssize_t disk_attr_show(struct kobject *kobj, struct attribute *attr,
322 struct gendisk *disk = to_disk(kobj);
323 struct disk_attribute *disk_attr =
324 container_of(attr,struct disk_attribute,attr);
328 ret = disk_attr->show(disk,page);
332 static struct sysfs_ops disk_sysfs_ops = {
333 .show = &disk_attr_show,
336 static ssize_t disk_dev_read(struct gendisk * disk, char *page)
338 dev_t base = MKDEV(disk->major, disk->first_minor);
339 return print_dev_t(page, base);
341 static ssize_t disk_range_read(struct gendisk * disk, char *page)
343 return sprintf(page, "%d\n", disk->minors);
345 static ssize_t disk_size_read(struct gendisk * disk, char *page)
347 return sprintf(page, "%llu\n", (unsigned long long)get_capacity(disk));
350 static inline unsigned jiffies_to_msec(unsigned jif)
353 return jif * (1000 / HZ);
355 return jif / (HZ / 1000);
357 return (jif / HZ) * 1000 + (jif % HZ) * 1000 / HZ;
360 static ssize_t disk_stats_read(struct gendisk * disk, char *page)
362 disk_round_stats(disk);
368 disk_stat_read(disk, reads), disk_stat_read(disk, read_merges),
369 (unsigned long long)disk_stat_read(disk, read_sectors),
370 jiffies_to_msec(disk_stat_read(disk, read_ticks)),
371 disk_stat_read(disk, writes),
372 disk_stat_read(disk, write_merges),
373 (unsigned long long)disk_stat_read(disk, write_sectors),
374 jiffies_to_msec(disk_stat_read(disk, write_ticks)),
375 disk_stat_read(disk, in_flight),
376 jiffies_to_msec(disk_stat_read(disk, io_ticks)),
377 jiffies_to_msec(disk_stat_read(disk, time_in_queue)));
379 static struct disk_attribute disk_attr_dev = {
380 .attr = {.name = "dev", .mode = S_IRUGO },
381 .show = disk_dev_read
383 static struct disk_attribute disk_attr_range = {
384 .attr = {.name = "range", .mode = S_IRUGO },
385 .show = disk_range_read
387 static struct disk_attribute disk_attr_size = {
388 .attr = {.name = "size", .mode = S_IRUGO },
389 .show = disk_size_read
391 static struct disk_attribute disk_attr_stat = {
392 .attr = {.name = "stat", .mode = S_IRUGO },
393 .show = disk_stats_read
396 static struct attribute * default_attrs[] = {
398 &disk_attr_range.attr,
399 &disk_attr_size.attr,
400 &disk_attr_stat.attr,
404 static void disk_release(struct kobject * kobj)
406 struct gendisk *disk = to_disk(kobj);
409 free_disk_stats(disk);
413 static struct kobj_type ktype_block = {
414 .release = disk_release,
415 .sysfs_ops = &disk_sysfs_ops,
416 .default_attrs = default_attrs,
419 extern struct kobj_type ktype_part;
421 static int block_hotplug_filter(struct kset *kset, struct kobject *kobj)
423 struct kobj_type *ktype = get_ktype(kobj);
425 return ((ktype == &ktype_block) || (ktype == &ktype_part));
428 static struct kset_hotplug_ops block_hotplug_ops = {
429 .filter = block_hotplug_filter,
432 /* declare block_subsys. */
433 static decl_subsys(block, &ktype_block, &block_hotplug_ops);
437 * aggregate disk stat collector. Uses the same stats that the sysfs
438 * entries do, above, but makes them available through one seq_file.
439 * Watching a few disks may be efficient through sysfs, but watching
440 * all of them will be more efficient through this interface.
442 * The output looks suspiciously like /proc/partitions with a bunch of
447 static void *diskstats_start(struct seq_file *part, loff_t *pos)
452 down_read(&block_subsys.rwsem);
453 list_for_each(p, &block_subsys.kset.list)
455 return list_entry(p, struct gendisk, kobj.entry);
459 static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos)
461 struct list_head *p = ((struct gendisk *)v)->kobj.entry.next;
463 return p==&block_subsys.kset.list ? NULL :
464 list_entry(p, struct gendisk, kobj.entry);
467 static void diskstats_stop(struct seq_file *part, void *v)
469 up_read(&block_subsys.rwsem);
472 static int diskstats_show(struct seq_file *s, void *v)
474 struct gendisk *gp = v;
479 if (&sgp->kobj.entry == block_subsys.kset.list.next)
480 seq_puts(s, "major minor name"
481 " rio rmerge rsect ruse wio wmerge "
482 "wsect wuse running use aveq"
486 disk_round_stats(gp);
487 seq_printf(s, "%4d %4d %s %u %u %llu %u %u %u %llu %u %u %u %u\n",
488 gp->major, n + gp->first_minor, disk_name(gp, n, buf),
489 disk_stat_read(gp, reads), disk_stat_read(gp, read_merges),
490 (unsigned long long)disk_stat_read(gp, read_sectors),
491 jiffies_to_msec(disk_stat_read(gp, read_ticks)),
492 disk_stat_read(gp, writes), disk_stat_read(gp, write_merges),
493 (unsigned long long)disk_stat_read(gp, write_sectors),
494 jiffies_to_msec(disk_stat_read(gp, write_ticks)),
495 disk_stat_read(gp, in_flight),
496 jiffies_to_msec(disk_stat_read(gp, io_ticks)),
497 jiffies_to_msec(disk_stat_read(gp, time_in_queue)));
499 /* now show all non-0 size partitions of it */
500 for (n = 0; n < gp->minors - 1; n++) {
501 struct hd_struct *hd = gp->part[n];
503 if (hd && hd->nr_sects)
504 seq_printf(s, "%4d %4d %s %u %u %u %u\n",
505 gp->major, n + gp->first_minor + 1,
506 disk_name(gp, n + 1, buf),
507 hd->reads, hd->read_sectors,
508 hd->writes, hd->write_sectors);
514 struct seq_operations diskstats_op = {
515 .start = diskstats_start,
516 .next = diskstats_next,
517 .stop = diskstats_stop,
518 .show = diskstats_show
522 struct gendisk *alloc_disk(int minors)
524 struct gendisk *disk = kmalloc(sizeof(struct gendisk), GFP_KERNEL);
526 memset(disk, 0, sizeof(struct gendisk));
527 if (!init_disk_stats(disk)) {
532 int size = (minors - 1) * sizeof(struct hd_struct *);
533 disk->part = kmalloc(size, GFP_KERNEL);
538 memset(disk->part, 0, size);
540 disk->minors = minors;
541 kobj_set_kset_s(disk,block_subsys);
542 kobject_init(&disk->kobj);
543 rand_initialize_disk(disk);
548 struct kobject *get_disk(struct gendisk *disk)
550 struct module *owner;
551 struct kobject *kobj;
555 owner = disk->fops->owner;
556 if (owner && !try_module_get(owner))
558 kobj = kobject_get(&disk->kobj);
567 void put_disk(struct gendisk *disk)
570 kobject_put(&disk->kobj);
573 EXPORT_SYMBOL(alloc_disk);
574 EXPORT_SYMBOL(get_disk);
575 EXPORT_SYMBOL(put_disk);
577 void set_device_ro(struct block_device *bdev, int flag)
579 struct gendisk *disk = bdev->bd_disk;
580 if (bdev->bd_contains != bdev) {
581 int part = bdev->bd_dev - MKDEV(disk->major, disk->first_minor);
582 struct hd_struct *p = disk->part[part-1];
583 if (p) p->policy = flag;
588 void set_disk_ro(struct gendisk *disk, int flag)
592 for (i = 0; i < disk->minors - 1; i++)
593 if (disk->part[i]) disk->part[i]->policy = flag;
596 int bdev_read_only(struct block_device *bdev)
598 struct gendisk *disk;
601 disk = bdev->bd_disk;
602 if (bdev->bd_contains != bdev) {
603 int part = bdev->bd_dev - MKDEV(disk->major, disk->first_minor);
604 struct hd_struct *p = disk->part[part-1];
605 if (p) return p->policy;
611 int invalidate_partition(struct gendisk *disk, int index)
614 struct block_device *bdev = bdget_disk(disk, index);
616 res = __invalidate_device(bdev, 1);
621 EXPORT_SYMBOL(bdev_read_only);
622 EXPORT_SYMBOL(set_device_ro);
623 EXPORT_SYMBOL(set_disk_ro);
624 EXPORT_SYMBOL(invalidate_partition);