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);
193 blk_register_filter(disk);
195 bdi = &disk->queue->backing_dev_info;
196 bdi_register_dev(bdi, MKDEV(disk->major, disk->first_minor));
197 retval = sysfs_create_link(&disk->dev.kobj, &bdi->dev->kobj, "bdi");
201 EXPORT_SYMBOL(add_disk);
202 EXPORT_SYMBOL(del_gendisk); /* in partitions/check.c */
204 void unlink_gendisk(struct gendisk *disk)
206 blk_unregister_filter(disk);
207 sysfs_remove_link(&disk->dev.kobj, "bdi");
208 bdi_unregister(&disk->queue->backing_dev_info);
209 blk_unregister_queue(disk);
210 blk_unregister_region(MKDEV(disk->major, disk->first_minor),
215 * get_gendisk - get partitioning information for a given device
216 * @dev: device to get partitioning information for
218 * This function gets the structure containing partitioning
219 * information for the given device @dev.
221 struct gendisk *get_gendisk(dev_t devt, int *part)
223 struct kobject *kobj = kobj_lookup(bdev_map, devt, part);
224 struct device *dev = kobj_to_dev(kobj);
226 return kobj ? dev_to_disk(dev) : NULL;
230 * print a full list of all partitions - intended for places where the root
231 * filesystem can't be mounted and thus to give the victim some idea of what
234 void __init printk_all_partitions(void)
238 char buf[BDEVNAME_SIZE];
241 mutex_lock(&block_class_lock);
242 /* For each block device... */
243 list_for_each_entry(dev, &block_class.devices, node) {
244 if (dev->type != &disk_type)
246 sgp = dev_to_disk(dev);
248 * Don't show empty devices or things that have been surpressed
250 if (get_capacity(sgp) == 0 ||
251 (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
255 * Note, unlike /proc/partitions, I am showing the numbers in
256 * hex - the same format as the root= option takes.
258 printk("%02x%02x %10llu %s",
259 sgp->major, sgp->first_minor,
260 (unsigned long long)get_capacity(sgp) >> 1,
261 disk_name(sgp, 0, buf));
262 if (sgp->driverfs_dev != NULL &&
263 sgp->driverfs_dev->driver != NULL)
264 printk(" driver: %s\n",
265 sgp->driverfs_dev->driver->name);
267 printk(" (driver?)\n");
269 /* now show the partitions */
270 for (n = 0; n < sgp->minors - 1; ++n) {
271 if (sgp->part[n] == NULL)
273 if (sgp->part[n]->nr_sects == 0)
275 printk(" %02x%02x %10llu %s\n",
276 sgp->major, n + 1 + sgp->first_minor,
277 (unsigned long long)sgp->part[n]->nr_sects >> 1,
278 disk_name(sgp, n + 1, buf));
282 mutex_unlock(&block_class_lock);
285 #ifdef CONFIG_PROC_FS
287 static void *part_start(struct seq_file *part, loff_t *pos)
292 mutex_lock(&block_class_lock);
293 list_for_each_entry(dev, &block_class.devices, node) {
294 if (dev->type != &disk_type)
297 return dev_to_disk(dev);
302 static void *part_next(struct seq_file *part, void *v, loff_t *pos)
304 struct gendisk *gp = v;
307 list_for_each_entry(dev, &gp->dev.node, node) {
308 if (&dev->node == &block_class.devices)
310 if (dev->type == &disk_type)
311 return dev_to_disk(dev);
316 static void part_stop(struct seq_file *part, void *v)
318 mutex_unlock(&block_class_lock);
321 static int show_partition(struct seq_file *part, void *v)
323 struct gendisk *sgp = v;
325 char buf[BDEVNAME_SIZE];
327 if (&sgp->dev.node == block_class.devices.next)
328 seq_puts(part, "major minor #blocks name\n\n");
330 /* Don't show non-partitionable removeable devices or empty devices */
331 if (!get_capacity(sgp) ||
332 (sgp->minors == 1 && (sgp->flags & GENHD_FL_REMOVABLE)))
334 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
337 /* show the full disk and all non-0 size partitions of it */
338 seq_printf(part, "%4d %4d %10llu %s\n",
339 sgp->major, sgp->first_minor,
340 (unsigned long long)get_capacity(sgp) >> 1,
341 disk_name(sgp, 0, buf));
342 for (n = 0; n < sgp->minors - 1; n++) {
345 if (sgp->part[n]->nr_sects == 0)
347 seq_printf(part, "%4d %4d %10llu %s\n",
348 sgp->major, n + 1 + sgp->first_minor,
349 (unsigned long long)sgp->part[n]->nr_sects >> 1 ,
350 disk_name(sgp, n + 1, buf));
356 const struct seq_operations partitions_op = {
360 .show = show_partition
365 static struct kobject *base_probe(dev_t devt, int *part, void *data)
367 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
368 /* Make old-style 2.4 aliases work */
369 request_module("block-major-%d", MAJOR(devt));
373 static int __init genhd_device_init(void)
377 block_class.dev_kobj = sysfs_dev_block_kobj;
378 error = class_register(&block_class);
381 bdev_map = kobj_map_init(base_probe, &block_class_lock);
384 #ifndef CONFIG_SYSFS_DEPRECATED
385 /* create top-level block dir */
386 block_depr = kobject_create_and_add("block", NULL);
391 subsys_initcall(genhd_device_init);
393 static ssize_t disk_range_show(struct device *dev,
394 struct device_attribute *attr, char *buf)
396 struct gendisk *disk = dev_to_disk(dev);
398 return sprintf(buf, "%d\n", disk->minors);
401 static ssize_t disk_removable_show(struct device *dev,
402 struct device_attribute *attr, char *buf)
404 struct gendisk *disk = dev_to_disk(dev);
406 return sprintf(buf, "%d\n",
407 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
410 static ssize_t disk_ro_show(struct device *dev,
411 struct device_attribute *attr, char *buf)
413 struct gendisk *disk = dev_to_disk(dev);
415 return sprintf(buf, "%d\n", disk->policy ? 1 : 0);
418 static ssize_t disk_size_show(struct device *dev,
419 struct device_attribute *attr, char *buf)
421 struct gendisk *disk = dev_to_disk(dev);
423 return sprintf(buf, "%llu\n", (unsigned long long)get_capacity(disk));
426 static ssize_t disk_capability_show(struct device *dev,
427 struct device_attribute *attr, char *buf)
429 struct gendisk *disk = dev_to_disk(dev);
431 return sprintf(buf, "%x\n", disk->flags);
434 static ssize_t disk_stat_show(struct device *dev,
435 struct device_attribute *attr, char *buf)
437 struct gendisk *disk = dev_to_disk(dev);
440 disk_round_stats(disk);
443 "%8lu %8lu %8llu %8u "
444 "%8lu %8lu %8llu %8u "
447 disk_stat_read(disk, ios[READ]),
448 disk_stat_read(disk, merges[READ]),
449 (unsigned long long)disk_stat_read(disk, sectors[READ]),
450 jiffies_to_msecs(disk_stat_read(disk, ticks[READ])),
451 disk_stat_read(disk, ios[WRITE]),
452 disk_stat_read(disk, merges[WRITE]),
453 (unsigned long long)disk_stat_read(disk, sectors[WRITE]),
454 jiffies_to_msecs(disk_stat_read(disk, ticks[WRITE])),
456 jiffies_to_msecs(disk_stat_read(disk, io_ticks)),
457 jiffies_to_msecs(disk_stat_read(disk, time_in_queue)));
460 #ifdef CONFIG_FAIL_MAKE_REQUEST
461 static ssize_t disk_fail_show(struct device *dev,
462 struct device_attribute *attr, char *buf)
464 struct gendisk *disk = dev_to_disk(dev);
466 return sprintf(buf, "%d\n", disk->flags & GENHD_FL_FAIL ? 1 : 0);
469 static ssize_t disk_fail_store(struct device *dev,
470 struct device_attribute *attr,
471 const char *buf, size_t count)
473 struct gendisk *disk = dev_to_disk(dev);
476 if (count > 0 && sscanf(buf, "%d", &i) > 0) {
478 disk->flags &= ~GENHD_FL_FAIL;
480 disk->flags |= GENHD_FL_FAIL;
488 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
489 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
490 static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
491 static DEVICE_ATTR(size, S_IRUGO, disk_size_show, NULL);
492 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
493 static DEVICE_ATTR(stat, S_IRUGO, disk_stat_show, NULL);
494 #ifdef CONFIG_FAIL_MAKE_REQUEST
495 static struct device_attribute dev_attr_fail =
496 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, disk_fail_show, disk_fail_store);
499 static struct attribute *disk_attrs[] = {
500 &dev_attr_range.attr,
501 &dev_attr_removable.attr,
504 &dev_attr_capability.attr,
506 #ifdef CONFIG_FAIL_MAKE_REQUEST
512 static struct attribute_group disk_attr_group = {
516 static struct attribute_group *disk_attr_groups[] = {
521 static void disk_release(struct device *dev)
523 struct gendisk *disk = dev_to_disk(dev);
527 free_disk_stats(disk);
530 struct class block_class = {
534 static struct device_type disk_type = {
536 .groups = disk_attr_groups,
537 .release = disk_release,
541 * aggregate disk stat collector. Uses the same stats that the sysfs
542 * entries do, above, but makes them available through one seq_file.
544 * The output looks suspiciously like /proc/partitions with a bunch of
548 static void *diskstats_start(struct seq_file *part, loff_t *pos)
553 mutex_lock(&block_class_lock);
554 list_for_each_entry(dev, &block_class.devices, node) {
555 if (dev->type != &disk_type)
558 return dev_to_disk(dev);
563 static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos)
565 struct gendisk *gp = v;
569 list_for_each_entry(dev, &gp->dev.node, node) {
570 if (&dev->node == &block_class.devices)
572 if (dev->type == &disk_type)
573 return dev_to_disk(dev);
578 static void diskstats_stop(struct seq_file *part, void *v)
580 mutex_unlock(&block_class_lock);
583 static int diskstats_show(struct seq_file *s, void *v)
585 struct gendisk *gp = v;
586 char buf[BDEVNAME_SIZE];
590 if (&gp->dev.kobj.entry == block_class.devices.next)
591 seq_puts(s, "major minor name"
592 " rio rmerge rsect ruse wio wmerge "
593 "wsect wuse running use aveq"
598 disk_round_stats(gp);
600 seq_printf(s, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n",
601 gp->major, n + gp->first_minor, disk_name(gp, n, buf),
602 disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]),
603 (unsigned long long)disk_stat_read(gp, sectors[0]),
604 jiffies_to_msecs(disk_stat_read(gp, ticks[0])),
605 disk_stat_read(gp, ios[1]), disk_stat_read(gp, merges[1]),
606 (unsigned long long)disk_stat_read(gp, sectors[1]),
607 jiffies_to_msecs(disk_stat_read(gp, ticks[1])),
609 jiffies_to_msecs(disk_stat_read(gp, io_ticks)),
610 jiffies_to_msecs(disk_stat_read(gp, time_in_queue)));
612 /* now show all non-0 size partitions of it */
613 for (n = 0; n < gp->minors - 1; n++) {
614 struct hd_struct *hd = gp->part[n];
616 if (!hd || !hd->nr_sects)
620 part_round_stats(hd);
622 seq_printf(s, "%4d %4d %s %lu %lu %llu "
623 "%u %lu %lu %llu %u %u %u %u\n",
624 gp->major, n + gp->first_minor + 1,
625 disk_name(gp, n + 1, buf),
626 part_stat_read(hd, ios[0]),
627 part_stat_read(hd, merges[0]),
628 (unsigned long long)part_stat_read(hd, sectors[0]),
629 jiffies_to_msecs(part_stat_read(hd, ticks[0])),
630 part_stat_read(hd, ios[1]),
631 part_stat_read(hd, merges[1]),
632 (unsigned long long)part_stat_read(hd, sectors[1]),
633 jiffies_to_msecs(part_stat_read(hd, ticks[1])),
635 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
636 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
643 const struct seq_operations diskstats_op = {
644 .start = diskstats_start,
645 .next = diskstats_next,
646 .stop = diskstats_stop,
647 .show = diskstats_show
650 static void media_change_notify_thread(struct work_struct *work)
652 struct gendisk *gd = container_of(work, struct gendisk, async_notify);
653 char event[] = "MEDIA_CHANGE=1";
654 char *envp[] = { event, NULL };
657 * set enviroment vars to indicate which event this is for
658 * so that user space will know to go check the media status.
660 kobject_uevent_env(&gd->dev.kobj, KOBJ_CHANGE, envp);
661 put_device(gd->driverfs_dev);
665 void genhd_media_change_notify(struct gendisk *disk)
667 get_device(disk->driverfs_dev);
668 schedule_work(&disk->async_notify);
670 EXPORT_SYMBOL_GPL(genhd_media_change_notify);
673 dev_t blk_lookup_devt(const char *name, int part)
676 dev_t devt = MKDEV(0, 0);
678 mutex_lock(&block_class_lock);
679 list_for_each_entry(dev, &block_class.devices, node) {
680 if (dev->type != &disk_type)
682 if (strcmp(dev->bus_id, name) == 0) {
683 struct gendisk *disk = dev_to_disk(dev);
685 if (part < disk->minors)
686 devt = MKDEV(MAJOR(dev->devt),
687 MINOR(dev->devt) + part);
691 mutex_unlock(&block_class_lock);
695 EXPORT_SYMBOL(blk_lookup_devt);
697 struct gendisk *alloc_disk(int minors)
699 return alloc_disk_node(minors, -1);
702 struct gendisk *alloc_disk_node(int minors, int node_id)
704 struct gendisk *disk;
706 disk = kmalloc_node(sizeof(struct gendisk),
707 GFP_KERNEL | __GFP_ZERO, node_id);
709 if (!init_disk_stats(disk)) {
714 int size = (minors - 1) * sizeof(struct hd_struct *);
715 disk->part = kmalloc_node(size,
716 GFP_KERNEL | __GFP_ZERO, node_id);
718 free_disk_stats(disk);
723 disk->minors = minors;
724 rand_initialize_disk(disk);
725 disk->dev.class = &block_class;
726 disk->dev.type = &disk_type;
727 device_initialize(&disk->dev);
728 INIT_WORK(&disk->async_notify,
729 media_change_notify_thread);
734 EXPORT_SYMBOL(alloc_disk);
735 EXPORT_SYMBOL(alloc_disk_node);
737 struct kobject *get_disk(struct gendisk *disk)
739 struct module *owner;
740 struct kobject *kobj;
744 owner = disk->fops->owner;
745 if (owner && !try_module_get(owner))
747 kobj = kobject_get(&disk->dev.kobj);
756 EXPORT_SYMBOL(get_disk);
758 void put_disk(struct gendisk *disk)
761 kobject_put(&disk->dev.kobj);
764 EXPORT_SYMBOL(put_disk);
766 void set_device_ro(struct block_device *bdev, int flag)
768 if (bdev->bd_contains != bdev)
769 bdev->bd_part->policy = flag;
771 bdev->bd_disk->policy = flag;
774 EXPORT_SYMBOL(set_device_ro);
776 void set_disk_ro(struct gendisk *disk, int flag)
780 for (i = 0; i < disk->minors - 1; i++)
781 if (disk->part[i]) disk->part[i]->policy = flag;
784 EXPORT_SYMBOL(set_disk_ro);
786 int bdev_read_only(struct block_device *bdev)
790 else if (bdev->bd_contains != bdev)
791 return bdev->bd_part->policy;
793 return bdev->bd_disk->policy;
796 EXPORT_SYMBOL(bdev_read_only);
798 int invalidate_partition(struct gendisk *disk, int index)
801 struct block_device *bdev = bdget_disk(disk, index);
804 res = __invalidate_device(bdev);
810 EXPORT_SYMBOL(invalidate_partition);