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>
22 static DEFINE_MUTEX(block_class_lock);
23 #ifndef CONFIG_SYSFS_DEPRECATED
24 struct kobject *block_depr;
27 static struct device_type disk_type;
30 * Can be deleted altogether. Later.
33 static struct blk_major_name {
34 struct blk_major_name *next;
37 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
39 /* index in the above - for now: assume no multimajor ranges */
40 static inline int major_to_index(int major)
42 return major % BLKDEV_MAJOR_HASH_SIZE;
46 void blkdev_show(struct seq_file *f, off_t offset)
48 struct blk_major_name *dp;
50 if (offset < BLKDEV_MAJOR_HASH_SIZE) {
51 mutex_lock(&block_class_lock);
52 for (dp = major_names[offset]; dp; dp = dp->next)
53 seq_printf(f, "%3d %s\n", dp->major, dp->name);
54 mutex_unlock(&block_class_lock);
57 #endif /* CONFIG_PROC_FS */
59 int register_blkdev(unsigned int major, const char *name)
61 struct blk_major_name **n, *p;
64 mutex_lock(&block_class_lock);
68 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
69 if (major_names[index] == NULL)
74 printk("register_blkdev: failed to get major for %s\n",
83 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
90 strlcpy(p->name, name, sizeof(p->name));
92 index = major_to_index(major);
94 for (n = &major_names[index]; *n; n = &(*n)->next) {
95 if ((*n)->major == major)
104 printk("register_blkdev: cannot get major %d for %s\n",
109 mutex_unlock(&block_class_lock);
113 EXPORT_SYMBOL(register_blkdev);
115 void unregister_blkdev(unsigned int major, const char *name)
117 struct blk_major_name **n;
118 struct blk_major_name *p = NULL;
119 int index = major_to_index(major);
121 mutex_lock(&block_class_lock);
122 for (n = &major_names[index]; *n; n = &(*n)->next)
123 if ((*n)->major == major)
125 if (!*n || strcmp((*n)->name, name)) {
131 mutex_unlock(&block_class_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 devt, 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, devt, range, module, probe, lock, data);
151 EXPORT_SYMBOL(blk_register_region);
153 void blk_unregister_region(dev_t devt, unsigned long range)
155 kobj_unmap(bdev_map, devt, range);
158 EXPORT_SYMBOL(blk_unregister_region);
160 static struct kobject *exact_match(dev_t devt, int *part, void *data)
162 struct gendisk *p = data;
167 static int exact_lock(dev_t devt, 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 struct backing_dev_info *bdi;
188 disk->flags |= GENHD_FL_UP;
189 blk_register_region(MKDEV(disk->major, disk->first_minor),
190 disk->minors, NULL, exact_match, exact_lock, disk);
192 blk_register_queue(disk);
194 bdi = &disk->queue->backing_dev_info;
195 bdi_register_dev(bdi, MKDEV(disk->major, disk->first_minor));
196 retval = sysfs_create_link(&disk->dev.kobj, &bdi->dev->kobj, "bdi");
200 EXPORT_SYMBOL(add_disk);
201 EXPORT_SYMBOL(del_gendisk); /* in partitions/check.c */
203 void unlink_gendisk(struct gendisk *disk)
205 sysfs_remove_link(&disk->dev.kobj, "bdi");
206 bdi_unregister(&disk->queue->backing_dev_info);
207 blk_unregister_queue(disk);
208 blk_unregister_region(MKDEV(disk->major, disk->first_minor),
213 * get_gendisk - get partitioning information for a given device
214 * @devt: device to get partitioning information for
215 * @part: returned partition index
217 * This function gets the structure containing partitioning
218 * information for the given device @devt.
220 struct gendisk *get_gendisk(dev_t devt, int *part)
222 struct kobject *kobj = kobj_lookup(bdev_map, devt, part);
223 struct device *dev = kobj_to_dev(kobj);
225 return kobj ? dev_to_disk(dev) : NULL;
229 * print a partitions - intended for places where the root filesystem can't be
230 * mounted and thus to give the victim some idea of what went wrong
232 static int printk_partition(struct device *dev, void *data)
235 char buf[BDEVNAME_SIZE];
238 if (dev->type != &disk_type)
241 sgp = dev_to_disk(dev);
243 * Don't show empty devices or things that have been surpressed
245 if (get_capacity(sgp) == 0 ||
246 (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
250 * Note, unlike /proc/partitions, I am showing the numbers in
251 * hex - the same format as the root= option takes.
253 printk("%02x%02x %10llu %s",
254 sgp->major, sgp->first_minor,
255 (unsigned long long)get_capacity(sgp) >> 1,
256 disk_name(sgp, 0, buf));
257 if (sgp->driverfs_dev != NULL &&
258 sgp->driverfs_dev->driver != NULL)
259 printk(" driver: %s\n",
260 sgp->driverfs_dev->driver->name);
262 printk(" (driver?)\n");
264 /* now show the partitions */
265 for (n = 0; n < sgp->minors - 1; ++n) {
266 if (sgp->part[n] == NULL)
268 if (sgp->part[n]->nr_sects == 0)
270 printk(" %02x%02x %10llu %s\n",
271 sgp->major, n + 1 + sgp->first_minor,
272 (unsigned long long)sgp->part[n]->nr_sects >> 1,
273 disk_name(sgp, n + 1, buf));
280 * print a full list of all partitions - intended for places where the root
281 * filesystem can't be mounted and thus to give the victim some idea of what
284 void __init printk_all_partitions(void)
286 mutex_lock(&block_class_lock);
287 class_for_each_device(&block_class, NULL, NULL, printk_partition);
288 mutex_unlock(&block_class_lock);
291 #ifdef CONFIG_PROC_FS
293 static int find_start(struct device *dev, void *data)
297 if (dev->type != &disk_type)
305 static void *part_start(struct seq_file *part, loff_t *pos)
311 part->private = (void *)1LU; /* tell show to print header */
313 mutex_lock(&block_class_lock);
314 dev = class_find_device(&block_class, NULL, &k, find_start);
317 return dev_to_disk(dev);
322 static int find_next(struct device *dev, void *data)
324 if (dev->type == &disk_type)
329 static void *part_next(struct seq_file *part, void *v, loff_t *pos)
331 struct gendisk *gp = v;
334 dev = class_find_device(&block_class, &gp->dev, NULL, find_next);
337 return dev_to_disk(dev);
342 static void part_stop(struct seq_file *part, void *v)
344 mutex_unlock(&block_class_lock);
347 static int show_partition(struct seq_file *part, void *v)
349 struct gendisk *sgp = v;
351 char buf[BDEVNAME_SIZE];
354 * Print header if start told us to do. This is to preserve
355 * the original behavior of not printing header if no
356 * partition exists. This hackery will be removed later with
357 * class iteration clean up.
360 seq_puts(part, "major minor #blocks name\n\n");
361 part->private = NULL;
364 /* Don't show non-partitionable removeable devices or empty devices */
365 if (!get_capacity(sgp) ||
366 (sgp->minors == 1 && (sgp->flags & GENHD_FL_REMOVABLE)))
368 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
371 /* show the full disk and all non-0 size partitions of it */
372 seq_printf(part, "%4d %4d %10llu %s\n",
373 sgp->major, sgp->first_minor,
374 (unsigned long long)get_capacity(sgp) >> 1,
375 disk_name(sgp, 0, buf));
376 for (n = 0; n < sgp->minors - 1; n++) {
379 if (sgp->part[n]->nr_sects == 0)
381 seq_printf(part, "%4d %4d %10llu %s\n",
382 sgp->major, n + 1 + sgp->first_minor,
383 (unsigned long long)sgp->part[n]->nr_sects >> 1 ,
384 disk_name(sgp, n + 1, buf));
390 const struct seq_operations partitions_op = {
394 .show = show_partition
399 static struct kobject *base_probe(dev_t devt, int *part, void *data)
401 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
402 /* Make old-style 2.4 aliases work */
403 request_module("block-major-%d", MAJOR(devt));
407 static int __init genhd_device_init(void)
411 block_class.dev_kobj = sysfs_dev_block_kobj;
412 error = class_register(&block_class);
415 bdev_map = kobj_map_init(base_probe, &block_class_lock);
418 #ifndef CONFIG_SYSFS_DEPRECATED
419 /* create top-level block dir */
420 block_depr = kobject_create_and_add("block", NULL);
425 subsys_initcall(genhd_device_init);
427 static ssize_t disk_range_show(struct device *dev,
428 struct device_attribute *attr, char *buf)
430 struct gendisk *disk = dev_to_disk(dev);
432 return sprintf(buf, "%d\n", disk->minors);
435 static ssize_t disk_removable_show(struct device *dev,
436 struct device_attribute *attr, char *buf)
438 struct gendisk *disk = dev_to_disk(dev);
440 return sprintf(buf, "%d\n",
441 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
444 static ssize_t disk_ro_show(struct device *dev,
445 struct device_attribute *attr, char *buf)
447 struct gendisk *disk = dev_to_disk(dev);
449 return sprintf(buf, "%d\n", disk->policy ? 1 : 0);
452 static ssize_t disk_size_show(struct device *dev,
453 struct device_attribute *attr, char *buf)
455 struct gendisk *disk = dev_to_disk(dev);
457 return sprintf(buf, "%llu\n", (unsigned long long)get_capacity(disk));
460 static ssize_t disk_capability_show(struct device *dev,
461 struct device_attribute *attr, char *buf)
463 struct gendisk *disk = dev_to_disk(dev);
465 return sprintf(buf, "%x\n", disk->flags);
468 static ssize_t disk_stat_show(struct device *dev,
469 struct device_attribute *attr, char *buf)
471 struct gendisk *disk = dev_to_disk(dev);
474 disk_round_stats(disk);
477 "%8lu %8lu %8llu %8u "
478 "%8lu %8lu %8llu %8u "
481 disk_stat_read(disk, ios[READ]),
482 disk_stat_read(disk, merges[READ]),
483 (unsigned long long)disk_stat_read(disk, sectors[READ]),
484 jiffies_to_msecs(disk_stat_read(disk, ticks[READ])),
485 disk_stat_read(disk, ios[WRITE]),
486 disk_stat_read(disk, merges[WRITE]),
487 (unsigned long long)disk_stat_read(disk, sectors[WRITE]),
488 jiffies_to_msecs(disk_stat_read(disk, ticks[WRITE])),
490 jiffies_to_msecs(disk_stat_read(disk, io_ticks)),
491 jiffies_to_msecs(disk_stat_read(disk, time_in_queue)));
494 #ifdef CONFIG_FAIL_MAKE_REQUEST
495 static ssize_t disk_fail_show(struct device *dev,
496 struct device_attribute *attr, char *buf)
498 struct gendisk *disk = dev_to_disk(dev);
500 return sprintf(buf, "%d\n", disk->flags & GENHD_FL_FAIL ? 1 : 0);
503 static ssize_t disk_fail_store(struct device *dev,
504 struct device_attribute *attr,
505 const char *buf, size_t count)
507 struct gendisk *disk = dev_to_disk(dev);
510 if (count > 0 && sscanf(buf, "%d", &i) > 0) {
512 disk->flags &= ~GENHD_FL_FAIL;
514 disk->flags |= GENHD_FL_FAIL;
522 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
523 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
524 static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
525 static DEVICE_ATTR(size, S_IRUGO, disk_size_show, NULL);
526 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
527 static DEVICE_ATTR(stat, S_IRUGO, disk_stat_show, NULL);
528 #ifdef CONFIG_FAIL_MAKE_REQUEST
529 static struct device_attribute dev_attr_fail =
530 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, disk_fail_show, disk_fail_store);
533 static struct attribute *disk_attrs[] = {
534 &dev_attr_range.attr,
535 &dev_attr_removable.attr,
538 &dev_attr_capability.attr,
540 #ifdef CONFIG_FAIL_MAKE_REQUEST
546 static struct attribute_group disk_attr_group = {
550 static struct attribute_group *disk_attr_groups[] = {
555 static void disk_release(struct device *dev)
557 struct gendisk *disk = dev_to_disk(dev);
561 free_disk_stats(disk);
564 struct class block_class = {
568 static struct device_type disk_type = {
570 .groups = disk_attr_groups,
571 .release = disk_release,
574 #ifdef CONFIG_PROC_FS
576 * aggregate disk stat collector. Uses the same stats that the sysfs
577 * entries do, above, but makes them available through one seq_file.
579 * The output looks suspiciously like /proc/partitions with a bunch of
583 static void *diskstats_start(struct seq_file *part, loff_t *pos)
588 mutex_lock(&block_class_lock);
589 dev = class_find_device(&block_class, NULL, &k, find_start);
592 return dev_to_disk(dev);
597 static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos)
599 struct gendisk *gp = v;
603 dev = class_find_device(&block_class, &gp->dev, NULL, find_next);
606 return dev_to_disk(dev);
611 static void diskstats_stop(struct seq_file *part, void *v)
613 mutex_unlock(&block_class_lock);
616 static int diskstats_show(struct seq_file *s, void *v)
618 struct gendisk *gp = v;
619 char buf[BDEVNAME_SIZE];
623 if (&gp->dev.kobj.entry == block_class.devices.next)
624 seq_puts(s, "major minor name"
625 " rio rmerge rsect ruse wio wmerge "
626 "wsect wuse running use aveq"
631 disk_round_stats(gp);
633 seq_printf(s, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n",
634 gp->major, n + gp->first_minor, disk_name(gp, n, buf),
635 disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]),
636 (unsigned long long)disk_stat_read(gp, sectors[0]),
637 jiffies_to_msecs(disk_stat_read(gp, ticks[0])),
638 disk_stat_read(gp, ios[1]), disk_stat_read(gp, merges[1]),
639 (unsigned long long)disk_stat_read(gp, sectors[1]),
640 jiffies_to_msecs(disk_stat_read(gp, ticks[1])),
642 jiffies_to_msecs(disk_stat_read(gp, io_ticks)),
643 jiffies_to_msecs(disk_stat_read(gp, time_in_queue)));
645 /* now show all non-0 size partitions of it */
646 for (n = 0; n < gp->minors - 1; n++) {
647 struct hd_struct *hd = gp->part[n];
649 if (!hd || !hd->nr_sects)
653 part_round_stats(hd);
655 seq_printf(s, "%4d %4d %s %lu %lu %llu "
656 "%u %lu %lu %llu %u %u %u %u\n",
657 gp->major, n + gp->first_minor + 1,
658 disk_name(gp, n + 1, buf),
659 part_stat_read(hd, ios[0]),
660 part_stat_read(hd, merges[0]),
661 (unsigned long long)part_stat_read(hd, sectors[0]),
662 jiffies_to_msecs(part_stat_read(hd, ticks[0])),
663 part_stat_read(hd, ios[1]),
664 part_stat_read(hd, merges[1]),
665 (unsigned long long)part_stat_read(hd, sectors[1]),
666 jiffies_to_msecs(part_stat_read(hd, ticks[1])),
668 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
669 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
676 const struct seq_operations diskstats_op = {
677 .start = diskstats_start,
678 .next = diskstats_next,
679 .stop = diskstats_stop,
680 .show = diskstats_show
682 #endif /* CONFIG_PROC_FS */
684 static void media_change_notify_thread(struct work_struct *work)
686 struct gendisk *gd = container_of(work, struct gendisk, async_notify);
687 char event[] = "MEDIA_CHANGE=1";
688 char *envp[] = { event, NULL };
691 * set enviroment vars to indicate which event this is for
692 * so that user space will know to go check the media status.
694 kobject_uevent_env(&gd->dev.kobj, KOBJ_CHANGE, envp);
695 put_device(gd->driverfs_dev);
699 void genhd_media_change_notify(struct gendisk *disk)
701 get_device(disk->driverfs_dev);
702 schedule_work(&disk->async_notify);
704 EXPORT_SYMBOL_GPL(genhd_media_change_notify);
712 static int match_id(struct device *dev, void *data)
714 struct find_block *find = data;
716 if (dev->type != &disk_type)
718 if (strcmp(dev->bus_id, find->name) == 0) {
719 struct gendisk *disk = dev_to_disk(dev);
720 if (find->part < disk->minors)
726 dev_t blk_lookup_devt(const char *name, int part)
729 dev_t devt = MKDEV(0, 0);
730 struct find_block find;
732 mutex_lock(&block_class_lock);
735 dev = class_find_device(&block_class, NULL, &find, match_id);
738 devt = MKDEV(MAJOR(dev->devt),
739 MINOR(dev->devt) + part);
741 mutex_unlock(&block_class_lock);
745 EXPORT_SYMBOL(blk_lookup_devt);
747 struct gendisk *alloc_disk(int minors)
749 return alloc_disk_node(minors, -1);
752 struct gendisk *alloc_disk_node(int minors, int node_id)
754 struct gendisk *disk;
756 disk = kmalloc_node(sizeof(struct gendisk),
757 GFP_KERNEL | __GFP_ZERO, node_id);
759 if (!init_disk_stats(disk)) {
764 int size = (minors - 1) * sizeof(struct hd_struct *);
765 disk->part = kmalloc_node(size,
766 GFP_KERNEL | __GFP_ZERO, node_id);
768 free_disk_stats(disk);
773 disk->minors = minors;
774 rand_initialize_disk(disk);
775 disk->dev.class = &block_class;
776 disk->dev.type = &disk_type;
777 device_initialize(&disk->dev);
778 INIT_WORK(&disk->async_notify,
779 media_change_notify_thread);
784 EXPORT_SYMBOL(alloc_disk);
785 EXPORT_SYMBOL(alloc_disk_node);
787 struct kobject *get_disk(struct gendisk *disk)
789 struct module *owner;
790 struct kobject *kobj;
794 owner = disk->fops->owner;
795 if (owner && !try_module_get(owner))
797 kobj = kobject_get(&disk->dev.kobj);
806 EXPORT_SYMBOL(get_disk);
808 void put_disk(struct gendisk *disk)
811 kobject_put(&disk->dev.kobj);
814 EXPORT_SYMBOL(put_disk);
816 void set_device_ro(struct block_device *bdev, int flag)
818 if (bdev->bd_contains != bdev)
819 bdev->bd_part->policy = flag;
821 bdev->bd_disk->policy = flag;
824 EXPORT_SYMBOL(set_device_ro);
826 void set_disk_ro(struct gendisk *disk, int flag)
830 for (i = 0; i < disk->minors - 1; i++)
831 if (disk->part[i]) disk->part[i]->policy = flag;
834 EXPORT_SYMBOL(set_disk_ro);
836 int bdev_read_only(struct block_device *bdev)
840 else if (bdev->bd_contains != bdev)
841 return bdev->bd_part->policy;
843 return bdev->bd_disk->policy;
846 EXPORT_SYMBOL(bdev_read_only);
848 int invalidate_partition(struct gendisk *disk, int index)
851 struct block_device *bdev = bdget_disk(disk, index);
854 res = __invalidate_device(bdev);
860 EXPORT_SYMBOL(invalidate_partition);