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 /* todo: make void - error printk here */
112 int unregister_blkdev(unsigned int major, const char *name)
114 struct blk_major_name **n;
115 struct blk_major_name *p = NULL;
116 int index = major_to_index(major);
119 mutex_lock(&block_subsys_lock);
120 for (n = &major_names[index]; *n; n = &(*n)->next)
121 if ((*n)->major == major)
123 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 disk->flags |= GENHD_FL_UP;
186 blk_register_region(MKDEV(disk->major, disk->first_minor),
187 disk->minors, NULL, exact_match, exact_lock, disk);
189 blk_register_queue(disk);
192 EXPORT_SYMBOL(add_disk);
193 EXPORT_SYMBOL(del_gendisk); /* in partitions/check.c */
195 void unlink_gendisk(struct gendisk *disk)
197 blk_unregister_queue(disk);
198 blk_unregister_region(MKDEV(disk->major, disk->first_minor),
202 #define to_disk(obj) container_of(obj,struct gendisk,kobj)
205 * get_gendisk - get partitioning information for a given device
206 * @dev: device to get partitioning information for
208 * This function gets the structure containing partitioning
209 * information for the given device @dev.
211 struct gendisk *get_gendisk(dev_t dev, int *part)
213 struct kobject *kobj = kobj_lookup(bdev_map, dev, part);
214 return kobj ? to_disk(kobj) : NULL;
218 * print a full list of all partitions - intended for places where the root
219 * filesystem can't be mounted and thus to give the victim some idea of what
222 void __init printk_all_partitions(void)
227 mutex_lock(&block_subsys_lock);
228 /* For each block device... */
229 list_for_each_entry(sgp, &block_subsys.list, kobj.entry) {
230 char buf[BDEVNAME_SIZE];
232 * Don't show empty devices or things that have been surpressed
234 if (get_capacity(sgp) == 0 ||
235 (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
239 * Note, unlike /proc/partitions, I am showing the numbers in
240 * hex - the same format as the root= option takes.
242 printk("%02x%02x %10llu %s",
243 sgp->major, sgp->first_minor,
244 (unsigned long long)get_capacity(sgp) >> 1,
245 disk_name(sgp, 0, buf));
246 if (sgp->driverfs_dev != NULL &&
247 sgp->driverfs_dev->driver != NULL)
248 printk(" driver: %s\n",
249 sgp->driverfs_dev->driver->name);
251 printk(" (driver?)\n");
253 /* now show the partitions */
254 for (n = 0; n < sgp->minors - 1; ++n) {
255 if (sgp->part[n] == NULL)
257 if (sgp->part[n]->nr_sects == 0)
259 printk(" %02x%02x %10llu %s\n",
260 sgp->major, n + 1 + sgp->first_minor,
261 (unsigned long long)sgp->part[n]->nr_sects >> 1,
262 disk_name(sgp, n + 1, buf));
263 } /* partition subloop */
264 } /* Block device loop */
266 mutex_unlock(&block_subsys_lock);
270 #ifdef CONFIG_PROC_FS
272 static void *part_start(struct seq_file *part, loff_t *pos)
277 mutex_lock(&block_subsys_lock);
278 list_for_each(p, &block_subsys.list)
280 return list_entry(p, struct gendisk, kobj.entry);
284 static void *part_next(struct seq_file *part, void *v, loff_t *pos)
286 struct list_head *p = ((struct gendisk *)v)->kobj.entry.next;
288 return p==&block_subsys.list ? NULL :
289 list_entry(p, struct gendisk, kobj.entry);
292 static void part_stop(struct seq_file *part, void *v)
294 mutex_unlock(&block_subsys_lock);
297 static int show_partition(struct seq_file *part, void *v)
299 struct gendisk *sgp = v;
301 char buf[BDEVNAME_SIZE];
303 if (&sgp->kobj.entry == block_subsys.list.next)
304 seq_puts(part, "major minor #blocks name\n\n");
306 /* Don't show non-partitionable removeable devices or empty devices */
307 if (!get_capacity(sgp) ||
308 (sgp->minors == 1 && (sgp->flags & GENHD_FL_REMOVABLE)))
310 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
313 /* show the full disk and all non-0 size partitions of it */
314 seq_printf(part, "%4d %4d %10llu %s\n",
315 sgp->major, sgp->first_minor,
316 (unsigned long long)get_capacity(sgp) >> 1,
317 disk_name(sgp, 0, buf));
318 for (n = 0; n < sgp->minors - 1; n++) {
321 if (sgp->part[n]->nr_sects == 0)
323 seq_printf(part, "%4d %4d %10llu %s\n",
324 sgp->major, n + 1 + sgp->first_minor,
325 (unsigned long long)sgp->part[n]->nr_sects >> 1 ,
326 disk_name(sgp, n + 1, buf));
332 struct seq_operations partitions_op = {
336 .show = show_partition
341 extern int blk_dev_init(void);
343 static struct kobject *base_probe(dev_t dev, int *part, void *data)
345 if (request_module("block-major-%d-%d", MAJOR(dev), MINOR(dev)) > 0)
346 /* Make old-style 2.4 aliases work */
347 request_module("block-major-%d", MAJOR(dev));
351 static int __init genhd_device_init(void)
355 bdev_map = kobj_map_init(base_probe, &block_subsys_lock);
357 err = subsystem_register(&block_subsys);
359 printk(KERN_WARNING "%s: subsystem_register error: %d\n",
364 subsys_initcall(genhd_device_init);
369 * kobject & sysfs bindings for block devices
371 static ssize_t disk_attr_show(struct kobject *kobj, struct attribute *attr,
374 struct gendisk *disk = to_disk(kobj);
375 struct disk_attribute *disk_attr =
376 container_of(attr,struct disk_attribute,attr);
380 ret = disk_attr->show(disk,page);
384 static ssize_t disk_attr_store(struct kobject * kobj, struct attribute * attr,
385 const char *page, size_t count)
387 struct gendisk *disk = to_disk(kobj);
388 struct disk_attribute *disk_attr =
389 container_of(attr,struct disk_attribute,attr);
392 if (disk_attr->store)
393 ret = disk_attr->store(disk, page, count);
397 static struct sysfs_ops disk_sysfs_ops = {
398 .show = &disk_attr_show,
399 .store = &disk_attr_store,
402 static ssize_t disk_uevent_store(struct gendisk * disk,
403 const char *buf, size_t count)
405 kobject_uevent(&disk->kobj, KOBJ_ADD);
408 static ssize_t disk_dev_read(struct gendisk * disk, char *page)
410 dev_t base = MKDEV(disk->major, disk->first_minor);
411 return print_dev_t(page, base);
413 static ssize_t disk_range_read(struct gendisk * disk, char *page)
415 return sprintf(page, "%d\n", disk->minors);
417 static ssize_t disk_removable_read(struct gendisk * disk, char *page)
419 return sprintf(page, "%d\n",
420 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
423 static ssize_t disk_size_read(struct gendisk * disk, char *page)
425 return sprintf(page, "%llu\n", (unsigned long long)get_capacity(disk));
427 static ssize_t disk_capability_read(struct gendisk *disk, char *page)
429 return sprintf(page, "%x\n", disk->flags);
431 static ssize_t disk_stats_read(struct gendisk * disk, char *page)
434 disk_round_stats(disk);
437 "%8lu %8lu %8llu %8u "
438 "%8lu %8lu %8llu %8u "
441 disk_stat_read(disk, ios[READ]),
442 disk_stat_read(disk, merges[READ]),
443 (unsigned long long)disk_stat_read(disk, sectors[READ]),
444 jiffies_to_msecs(disk_stat_read(disk, ticks[READ])),
445 disk_stat_read(disk, ios[WRITE]),
446 disk_stat_read(disk, merges[WRITE]),
447 (unsigned long long)disk_stat_read(disk, sectors[WRITE]),
448 jiffies_to_msecs(disk_stat_read(disk, ticks[WRITE])),
450 jiffies_to_msecs(disk_stat_read(disk, io_ticks)),
451 jiffies_to_msecs(disk_stat_read(disk, time_in_queue)));
453 static struct disk_attribute disk_attr_uevent = {
454 .attr = {.name = "uevent", .mode = S_IWUSR },
455 .store = disk_uevent_store
457 static struct disk_attribute disk_attr_dev = {
458 .attr = {.name = "dev", .mode = S_IRUGO },
459 .show = disk_dev_read
461 static struct disk_attribute disk_attr_range = {
462 .attr = {.name = "range", .mode = S_IRUGO },
463 .show = disk_range_read
465 static struct disk_attribute disk_attr_removable = {
466 .attr = {.name = "removable", .mode = S_IRUGO },
467 .show = disk_removable_read
469 static struct disk_attribute disk_attr_size = {
470 .attr = {.name = "size", .mode = S_IRUGO },
471 .show = disk_size_read
473 static struct disk_attribute disk_attr_capability = {
474 .attr = {.name = "capability", .mode = S_IRUGO },
475 .show = disk_capability_read
477 static struct disk_attribute disk_attr_stat = {
478 .attr = {.name = "stat", .mode = S_IRUGO },
479 .show = disk_stats_read
482 #ifdef CONFIG_FAIL_MAKE_REQUEST
484 static ssize_t disk_fail_store(struct gendisk * disk,
485 const char *buf, size_t count)
489 if (count > 0 && sscanf(buf, "%d", &i) > 0) {
491 disk->flags &= ~GENHD_FL_FAIL;
493 disk->flags |= GENHD_FL_FAIL;
498 static ssize_t disk_fail_read(struct gendisk * disk, char *page)
500 return sprintf(page, "%d\n", disk->flags & GENHD_FL_FAIL ? 1 : 0);
502 static struct disk_attribute disk_attr_fail = {
503 .attr = {.name = "make-it-fail", .mode = S_IRUGO | S_IWUSR },
504 .store = disk_fail_store,
505 .show = disk_fail_read
510 static struct attribute * default_attrs[] = {
511 &disk_attr_uevent.attr,
513 &disk_attr_range.attr,
514 &disk_attr_removable.attr,
515 &disk_attr_size.attr,
516 &disk_attr_stat.attr,
517 &disk_attr_capability.attr,
518 #ifdef CONFIG_FAIL_MAKE_REQUEST
519 &disk_attr_fail.attr,
524 static void disk_release(struct kobject * kobj)
526 struct gendisk *disk = to_disk(kobj);
529 free_disk_stats(disk);
533 static struct kobj_type ktype_block = {
534 .release = disk_release,
535 .sysfs_ops = &disk_sysfs_ops,
536 .default_attrs = default_attrs,
539 extern struct kobj_type ktype_part;
541 static int block_uevent_filter(struct kset *kset, struct kobject *kobj)
543 struct kobj_type *ktype = get_ktype(kobj);
545 return ((ktype == &ktype_block) || (ktype == &ktype_part));
548 static int block_uevent(struct kset *kset, struct kobject *kobj, char **envp,
549 int num_envp, char *buffer, int buffer_size)
551 struct kobj_type *ktype = get_ktype(kobj);
552 struct device *physdev;
553 struct gendisk *disk;
554 struct hd_struct *part;
558 if (ktype == &ktype_block) {
559 disk = container_of(kobj, struct gendisk, kobj);
560 add_uevent_var(envp, num_envp, &i, buffer, buffer_size,
561 &length, "MINOR=%u", disk->first_minor);
562 } else if (ktype == &ktype_part) {
563 disk = container_of(kobj->parent, struct gendisk, kobj);
564 part = container_of(kobj, struct hd_struct, kobj);
565 add_uevent_var(envp, num_envp, &i, buffer, buffer_size,
567 disk->first_minor + part->partno);
571 add_uevent_var(envp, num_envp, &i, buffer, buffer_size, &length,
572 "MAJOR=%u", disk->major);
574 /* add physical device, backing this device */
575 physdev = disk->driverfs_dev;
577 char *path = kobject_get_path(&physdev->kobj, GFP_KERNEL);
579 add_uevent_var(envp, num_envp, &i, buffer, buffer_size,
580 &length, "PHYSDEVPATH=%s", path);
584 add_uevent_var(envp, num_envp, &i,
585 buffer, buffer_size, &length,
590 add_uevent_var(envp, num_envp, &i,
591 buffer, buffer_size, &length,
593 physdev->driver->name);
596 /* terminate, set to next free slot, shrink available space */
600 buffer = &buffer[length];
601 buffer_size -= length;
606 static struct kset_uevent_ops block_uevent_ops = {
607 .filter = block_uevent_filter,
608 .uevent = block_uevent,
611 decl_subsys(block, &ktype_block, &block_uevent_ops);
614 * aggregate disk stat collector. Uses the same stats that the sysfs
615 * entries do, above, but makes them available through one seq_file.
616 * Watching a few disks may be efficient through sysfs, but watching
617 * all of them will be more efficient through this interface.
619 * The output looks suspiciously like /proc/partitions with a bunch of
624 static void *diskstats_start(struct seq_file *part, loff_t *pos)
629 mutex_lock(&block_subsys_lock);
630 list_for_each(p, &block_subsys.list)
632 return list_entry(p, struct gendisk, kobj.entry);
636 static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos)
638 struct list_head *p = ((struct gendisk *)v)->kobj.entry.next;
640 return p==&block_subsys.list ? NULL :
641 list_entry(p, struct gendisk, kobj.entry);
644 static void diskstats_stop(struct seq_file *part, void *v)
646 mutex_unlock(&block_subsys_lock);
649 static int diskstats_show(struct seq_file *s, void *v)
651 struct gendisk *gp = v;
652 char buf[BDEVNAME_SIZE];
656 if (&sgp->kobj.entry == block_subsys.kset.list.next)
657 seq_puts(s, "major minor name"
658 " rio rmerge rsect ruse wio wmerge "
659 "wsect wuse running use aveq"
664 disk_round_stats(gp);
666 seq_printf(s, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n",
667 gp->major, n + gp->first_minor, disk_name(gp, n, buf),
668 disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]),
669 (unsigned long long)disk_stat_read(gp, sectors[0]),
670 jiffies_to_msecs(disk_stat_read(gp, ticks[0])),
671 disk_stat_read(gp, ios[1]), disk_stat_read(gp, merges[1]),
672 (unsigned long long)disk_stat_read(gp, sectors[1]),
673 jiffies_to_msecs(disk_stat_read(gp, ticks[1])),
675 jiffies_to_msecs(disk_stat_read(gp, io_ticks)),
676 jiffies_to_msecs(disk_stat_read(gp, time_in_queue)));
678 /* now show all non-0 size partitions of it */
679 for (n = 0; n < gp->minors - 1; n++) {
680 struct hd_struct *hd = gp->part[n];
682 if (hd && hd->nr_sects)
683 seq_printf(s, "%4d %4d %s %u %u %u %u\n",
684 gp->major, n + gp->first_minor + 1,
685 disk_name(gp, n + 1, buf),
686 hd->ios[0], hd->sectors[0],
687 hd->ios[1], hd->sectors[1]);
693 struct seq_operations diskstats_op = {
694 .start = diskstats_start,
695 .next = diskstats_next,
696 .stop = diskstats_stop,
697 .show = diskstats_show
700 static void media_change_notify_thread(struct work_struct *work)
702 struct gendisk *gd = container_of(work, struct gendisk, async_notify);
703 char event[] = "MEDIA_CHANGE=1";
704 char *envp[] = { event, NULL };
707 * set enviroment vars to indicate which event this is for
708 * so that user space will know to go check the media status.
710 kobject_uevent_env(&gd->kobj, KOBJ_CHANGE, envp);
711 put_device(gd->driverfs_dev);
714 void genhd_media_change_notify(struct gendisk *disk)
716 get_device(disk->driverfs_dev);
717 schedule_work(&disk->async_notify);
719 EXPORT_SYMBOL_GPL(genhd_media_change_notify);
721 struct gendisk *alloc_disk(int minors)
723 return alloc_disk_node(minors, -1);
726 struct gendisk *alloc_disk_node(int minors, int node_id)
728 struct gendisk *disk;
730 disk = kmalloc_node(sizeof(struct gendisk),
731 GFP_KERNEL | __GFP_ZERO, node_id);
733 if (!init_disk_stats(disk)) {
738 int size = (minors - 1) * sizeof(struct hd_struct *);
739 disk->part = kmalloc_node(size,
740 GFP_KERNEL | __GFP_ZERO, node_id);
746 disk->minors = minors;
747 kobj_set_kset_s(disk,block_subsys);
748 kobject_init(&disk->kobj);
749 rand_initialize_disk(disk);
750 INIT_WORK(&disk->async_notify,
751 media_change_notify_thread);
756 EXPORT_SYMBOL(alloc_disk);
757 EXPORT_SYMBOL(alloc_disk_node);
759 struct kobject *get_disk(struct gendisk *disk)
761 struct module *owner;
762 struct kobject *kobj;
766 owner = disk->fops->owner;
767 if (owner && !try_module_get(owner))
769 kobj = kobject_get(&disk->kobj);
778 EXPORT_SYMBOL(get_disk);
780 void put_disk(struct gendisk *disk)
783 kobject_put(&disk->kobj);
786 EXPORT_SYMBOL(put_disk);
788 void set_device_ro(struct block_device *bdev, int flag)
790 if (bdev->bd_contains != bdev)
791 bdev->bd_part->policy = flag;
793 bdev->bd_disk->policy = flag;
796 EXPORT_SYMBOL(set_device_ro);
798 void set_disk_ro(struct gendisk *disk, int flag)
802 for (i = 0; i < disk->minors - 1; i++)
803 if (disk->part[i]) disk->part[i]->policy = flag;
806 EXPORT_SYMBOL(set_disk_ro);
808 int bdev_read_only(struct block_device *bdev)
812 else if (bdev->bd_contains != bdev)
813 return bdev->bd_part->policy;
815 return bdev->bd_disk->policy;
818 EXPORT_SYMBOL(bdev_read_only);
820 int invalidate_partition(struct gendisk *disk, int index)
823 struct block_device *bdev = bdget_disk(disk, index);
826 res = __invalidate_device(bdev);
832 EXPORT_SYMBOL(invalidate_partition);