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))
129 mutex_unlock(&block_subsys_lock);
135 EXPORT_SYMBOL(unregister_blkdev);
137 static struct kobj_map *bdev_map;
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.
144 void 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)
148 kobj_map(bdev_map, dev, range, module, probe, lock, data);
151 EXPORT_SYMBOL(blk_register_region);
153 void blk_unregister_region(dev_t dev, unsigned long range)
155 kobj_unmap(bdev_map, dev, range);
158 EXPORT_SYMBOL(blk_unregister_region);
160 static struct kobject *exact_match(dev_t dev, int *part, void *data)
162 struct gendisk *p = data;
166 static int exact_lock(dev_t dev, void *data)
168 struct gendisk *p = data;
176 * add_disk - add partitioning information to kernel list
177 * @disk: per-device partitioning information
179 * This function registers the partitioning information in @disk
182 void add_disk(struct gendisk *disk)
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);
188 blk_register_queue(disk);
191 EXPORT_SYMBOL(add_disk);
192 EXPORT_SYMBOL(del_gendisk); /* in partitions/check.c */
194 void unlink_gendisk(struct gendisk *disk)
196 blk_unregister_queue(disk);
197 blk_unregister_region(MKDEV(disk->major, disk->first_minor),
201 #define to_disk(obj) container_of(obj,struct gendisk,kobj)
204 * get_gendisk - get partitioning information for a given device
205 * @dev: device to get partitioning information for
207 * This function gets the structure containing partitioning
208 * information for the given device @dev.
210 struct gendisk *get_gendisk(dev_t dev, int *part)
212 struct kobject *kobj = kobj_lookup(bdev_map, dev, part);
213 return kobj ? to_disk(kobj) : NULL;
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
221 void __init printk_all_partitions(void)
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];
231 * Don't show empty devices or things that have been surpressed
233 if (get_capacity(sgp) == 0 ||
234 (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
238 * Note, unlike /proc/partitions, I am showing the numbers in
239 * hex - the same format as the root= option takes.
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);
250 printk(" (driver?)\n");
252 /* now show the partitions */
253 for (n = 0; n < sgp->minors - 1; ++n) {
254 if (sgp->part[n] == NULL)
256 if (sgp->part[n]->nr_sects == 0)
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 */
265 mutex_unlock(&block_subsys_lock);
269 #ifdef CONFIG_PROC_FS
271 static void *part_start(struct seq_file *part, loff_t *pos)
276 mutex_lock(&block_subsys_lock);
277 list_for_each(p, &block_subsys.list)
279 return list_entry(p, struct gendisk, kobj.entry);
283 static void *part_next(struct seq_file *part, void *v, loff_t *pos)
285 struct list_head *p = ((struct gendisk *)v)->kobj.entry.next;
287 return p==&block_subsys.list ? NULL :
288 list_entry(p, struct gendisk, kobj.entry);
291 static void part_stop(struct seq_file *part, void *v)
293 mutex_unlock(&block_subsys_lock);
296 static int show_partition(struct seq_file *part, void *v)
298 struct gendisk *sgp = v;
300 char buf[BDEVNAME_SIZE];
302 if (&sgp->kobj.entry == block_subsys.list.next)
303 seq_puts(part, "major minor #blocks name\n\n");
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)))
309 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
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++) {
320 if (sgp->part[n]->nr_sects == 0)
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));
331 struct seq_operations partitions_op = {
335 .show = show_partition
340 extern int blk_dev_init(void);
342 static struct kobject *base_probe(dev_t dev, int *part, void *data)
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));
350 static int __init genhd_device_init(void)
354 bdev_map = kobj_map_init(base_probe, &block_subsys_lock);
356 err = subsystem_register(&block_subsys);
358 printk(KERN_WARNING "%s: subsystem_register error: %d\n",
363 subsys_initcall(genhd_device_init);
368 * kobject & sysfs bindings for block devices
370 static ssize_t disk_attr_show(struct kobject *kobj, struct attribute *attr,
373 struct gendisk *disk = to_disk(kobj);
374 struct disk_attribute *disk_attr =
375 container_of(attr,struct disk_attribute,attr);
379 ret = disk_attr->show(disk,page);
383 static ssize_t disk_attr_store(struct kobject * kobj, struct attribute * attr,
384 const char *page, size_t count)
386 struct gendisk *disk = to_disk(kobj);
387 struct disk_attribute *disk_attr =
388 container_of(attr,struct disk_attribute,attr);
391 if (disk_attr->store)
392 ret = disk_attr->store(disk, page, count);
396 static struct sysfs_ops disk_sysfs_ops = {
397 .show = &disk_attr_show,
398 .store = &disk_attr_store,
401 static ssize_t disk_uevent_store(struct gendisk * disk,
402 const char *buf, size_t count)
404 kobject_uevent(&disk->kobj, KOBJ_ADD);
407 static ssize_t disk_dev_read(struct gendisk * disk, char *page)
409 dev_t base = MKDEV(disk->major, disk->first_minor);
410 return print_dev_t(page, base);
412 static ssize_t disk_range_read(struct gendisk * disk, char *page)
414 return sprintf(page, "%d\n", disk->minors);
416 static ssize_t disk_removable_read(struct gendisk * disk, char *page)
418 return sprintf(page, "%d\n",
419 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
422 static ssize_t disk_size_read(struct gendisk * disk, char *page)
424 return sprintf(page, "%llu\n", (unsigned long long)get_capacity(disk));
426 static ssize_t disk_capability_read(struct gendisk *disk, char *page)
428 return sprintf(page, "%x\n", disk->flags);
430 static ssize_t disk_stats_read(struct gendisk * disk, char *page)
433 disk_round_stats(disk);
436 "%8lu %8lu %8llu %8u "
437 "%8lu %8lu %8llu %8u "
440 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])),
449 jiffies_to_msecs(disk_stat_read(disk, io_ticks)),
450 jiffies_to_msecs(disk_stat_read(disk, time_in_queue)));
452 static struct disk_attribute disk_attr_uevent = {
453 .attr = {.name = "uevent", .mode = S_IWUSR },
454 .store = disk_uevent_store
456 static struct disk_attribute disk_attr_dev = {
457 .attr = {.name = "dev", .mode = S_IRUGO },
458 .show = disk_dev_read
460 static struct disk_attribute disk_attr_range = {
461 .attr = {.name = "range", .mode = S_IRUGO },
462 .show = disk_range_read
464 static struct disk_attribute disk_attr_removable = {
465 .attr = {.name = "removable", .mode = S_IRUGO },
466 .show = disk_removable_read
468 static struct disk_attribute disk_attr_size = {
469 .attr = {.name = "size", .mode = S_IRUGO },
470 .show = disk_size_read
472 static struct disk_attribute disk_attr_capability = {
473 .attr = {.name = "capability", .mode = S_IRUGO },
474 .show = disk_capability_read
476 static struct disk_attribute disk_attr_stat = {
477 .attr = {.name = "stat", .mode = S_IRUGO },
478 .show = disk_stats_read
481 #ifdef CONFIG_FAIL_MAKE_REQUEST
483 static ssize_t disk_fail_store(struct gendisk * disk,
484 const char *buf, size_t count)
488 if (count > 0 && sscanf(buf, "%d", &i) > 0) {
490 disk->flags &= ~GENHD_FL_FAIL;
492 disk->flags |= GENHD_FL_FAIL;
497 static ssize_t disk_fail_read(struct gendisk * disk, char *page)
499 return sprintf(page, "%d\n", disk->flags & GENHD_FL_FAIL ? 1 : 0);
501 static 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
509 static struct attribute * default_attrs[] = {
510 &disk_attr_uevent.attr,
512 &disk_attr_range.attr,
513 &disk_attr_removable.attr,
514 &disk_attr_size.attr,
515 &disk_attr_stat.attr,
516 &disk_attr_capability.attr,
517 #ifdef CONFIG_FAIL_MAKE_REQUEST
518 &disk_attr_fail.attr,
523 static void disk_release(struct kobject * kobj)
525 struct gendisk *disk = to_disk(kobj);
528 free_disk_stats(disk);
532 static struct kobj_type ktype_block = {
533 .release = disk_release,
534 .sysfs_ops = &disk_sysfs_ops,
535 .default_attrs = default_attrs,
538 extern struct kobj_type ktype_part;
540 static int block_uevent_filter(struct kset *kset, struct kobject *kobj)
542 struct kobj_type *ktype = get_ktype(kobj);
544 return ((ktype == &ktype_block) || (ktype == &ktype_part));
547 static int block_uevent(struct kset *kset, struct kobject *kobj, char **envp,
548 int num_envp, char *buffer, int buffer_size)
550 struct kobj_type *ktype = get_ktype(kobj);
551 struct device *physdev;
552 struct gendisk *disk;
553 struct hd_struct *part;
557 if (ktype == &ktype_block) {
558 disk = container_of(kobj, struct gendisk, kobj);
559 add_uevent_var(envp, num_envp, &i, buffer, buffer_size,
560 &length, "MINOR=%u", disk->first_minor);
561 } else if (ktype == &ktype_part) {
562 disk = container_of(kobj->parent, struct gendisk, kobj);
563 part = container_of(kobj, struct hd_struct, kobj);
564 add_uevent_var(envp, num_envp, &i, buffer, buffer_size,
566 disk->first_minor + part->partno);
570 add_uevent_var(envp, num_envp, &i, buffer, buffer_size, &length,
571 "MAJOR=%u", disk->major);
573 /* add physical device, backing this device */
574 physdev = disk->driverfs_dev;
576 char *path = kobject_get_path(&physdev->kobj, GFP_KERNEL);
578 add_uevent_var(envp, num_envp, &i, buffer, buffer_size,
579 &length, "PHYSDEVPATH=%s", path);
583 add_uevent_var(envp, num_envp, &i,
584 buffer, buffer_size, &length,
589 add_uevent_var(envp, num_envp, &i,
590 buffer, buffer_size, &length,
592 physdev->driver->name);
595 /* terminate, set to next free slot, shrink available space */
599 buffer = &buffer[length];
600 buffer_size -= length;
605 static struct kset_uevent_ops block_uevent_ops = {
606 .filter = block_uevent_filter,
607 .uevent = block_uevent,
610 decl_subsys(block, &ktype_block, &block_uevent_ops);
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.
618 * The output looks suspiciously like /proc/partitions with a bunch of
623 static void *diskstats_start(struct seq_file *part, loff_t *pos)
628 mutex_lock(&block_subsys_lock);
629 list_for_each(p, &block_subsys.list)
631 return list_entry(p, struct gendisk, kobj.entry);
635 static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos)
637 struct list_head *p = ((struct gendisk *)v)->kobj.entry.next;
639 return p==&block_subsys.list ? NULL :
640 list_entry(p, struct gendisk, kobj.entry);
643 static void diskstats_stop(struct seq_file *part, void *v)
645 mutex_unlock(&block_subsys_lock);
648 static int diskstats_show(struct seq_file *s, void *v)
650 struct gendisk *gp = v;
651 char buf[BDEVNAME_SIZE];
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"
663 disk_round_stats(gp);
665 seq_printf(s, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n",
666 gp->major, n + gp->first_minor, disk_name(gp, n, buf),
667 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])),
674 jiffies_to_msecs(disk_stat_read(gp, io_ticks)),
675 jiffies_to_msecs(disk_stat_read(gp, time_in_queue)));
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];
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),
685 hd->ios[0], hd->sectors[0],
686 hd->ios[1], hd->sectors[1]);
692 struct seq_operations diskstats_op = {
693 .start = diskstats_start,
694 .next = diskstats_next,
695 .stop = diskstats_stop,
696 .show = diskstats_show
699 static void media_change_notify_thread(struct work_struct *work)
701 struct gendisk *gd = container_of(work, struct gendisk, async_notify);
702 char event[] = "MEDIA_CHANGE=1";
703 char *envp[] = { event, NULL };
706 * set enviroment vars to indicate which event this is for
707 * so that user space will know to go check the media status.
709 kobject_uevent_env(&gd->kobj, KOBJ_CHANGE, envp);
710 put_device(gd->driverfs_dev);
713 void genhd_media_change_notify(struct gendisk *disk)
715 get_device(disk->driverfs_dev);
716 schedule_work(&disk->async_notify);
718 EXPORT_SYMBOL_GPL(genhd_media_change_notify);
720 struct gendisk *alloc_disk(int minors)
722 return alloc_disk_node(minors, -1);
725 struct gendisk *alloc_disk_node(int minors, int node_id)
727 struct gendisk *disk;
729 disk = kmalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
731 memset(disk, 0, sizeof(struct gendisk));
732 if (!init_disk_stats(disk)) {
737 int size = (minors - 1) * sizeof(struct hd_struct *);
738 disk->part = kmalloc_node(size, GFP_KERNEL, node_id);
743 memset(disk->part, 0, size);
745 disk->minors = minors;
746 kobj_set_kset_s(disk,block_subsys);
747 kobject_init(&disk->kobj);
748 rand_initialize_disk(disk);
749 INIT_WORK(&disk->async_notify,
750 media_change_notify_thread);
755 EXPORT_SYMBOL(alloc_disk);
756 EXPORT_SYMBOL(alloc_disk_node);
758 struct kobject *get_disk(struct gendisk *disk)
760 struct module *owner;
761 struct kobject *kobj;
765 owner = disk->fops->owner;
766 if (owner && !try_module_get(owner))
768 kobj = kobject_get(&disk->kobj);
777 EXPORT_SYMBOL(get_disk);
779 void put_disk(struct gendisk *disk)
782 kobject_put(&disk->kobj);
785 EXPORT_SYMBOL(put_disk);
787 void set_device_ro(struct block_device *bdev, int flag)
789 if (bdev->bd_contains != bdev)
790 bdev->bd_part->policy = flag;
792 bdev->bd_disk->policy = flag;
795 EXPORT_SYMBOL(set_device_ro);
797 void set_disk_ro(struct gendisk *disk, int flag)
801 for (i = 0; i < disk->minors - 1; i++)
802 if (disk->part[i]) disk->part[i]->policy = flag;
805 EXPORT_SYMBOL(set_disk_ro);
807 int bdev_read_only(struct block_device *bdev)
811 else if (bdev->bd_contains != bdev)
812 return bdev->bd_part->policy;
814 return bdev->bd_disk->policy;
817 EXPORT_SYMBOL(bdev_read_only);
819 int invalidate_partition(struct gendisk *disk, int index)
822 struct block_device *bdev = bdget_disk(disk, index);
825 res = __invalidate_device(bdev);
831 EXPORT_SYMBOL(invalidate_partition);