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 partitions - intended for places where the root filesystem can't be
231 * mounted and thus to give the victim some idea of what went wrong
233 static int printk_partition(struct device *dev, void *data)
236 char buf[BDEVNAME_SIZE];
239 if (dev->type != &disk_type)
242 sgp = dev_to_disk(dev);
244 * Don't show empty devices or things that have been surpressed
246 if (get_capacity(sgp) == 0 ||
247 (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
251 * Note, unlike /proc/partitions, I am showing the numbers in
252 * hex - the same format as the root= option takes.
254 printk("%02x%02x %10llu %s",
255 sgp->major, sgp->first_minor,
256 (unsigned long long)get_capacity(sgp) >> 1,
257 disk_name(sgp, 0, buf));
258 if (sgp->driverfs_dev != NULL &&
259 sgp->driverfs_dev->driver != NULL)
260 printk(" driver: %s\n",
261 sgp->driverfs_dev->driver->name);
263 printk(" (driver?)\n");
265 /* now show the partitions */
266 for (n = 0; n < sgp->minors - 1; ++n) {
267 if (sgp->part[n] == NULL)
269 if (sgp->part[n]->nr_sects == 0)
271 printk(" %02x%02x %10llu %s\n",
272 sgp->major, n + 1 + sgp->first_minor,
273 (unsigned long long)sgp->part[n]->nr_sects >> 1,
274 disk_name(sgp, n + 1, buf));
281 * print a full list of all partitions - intended for places where the root
282 * filesystem can't be mounted and thus to give the victim some idea of what
285 void __init printk_all_partitions(void)
287 mutex_lock(&block_class_lock);
288 class_for_each_device(&block_class, NULL, NULL, printk_partition);
289 mutex_unlock(&block_class_lock);
292 #ifdef CONFIG_PROC_FS
294 static void *part_start(struct seq_file *part, loff_t *pos)
299 mutex_lock(&block_class_lock);
300 list_for_each_entry(dev, &block_class.devices, node) {
301 if (dev->type != &disk_type)
304 return dev_to_disk(dev);
309 static void *part_next(struct seq_file *part, void *v, loff_t *pos)
311 struct gendisk *gp = v;
314 list_for_each_entry(dev, &gp->dev.node, node) {
315 if (&dev->node == &block_class.devices)
317 if (dev->type == &disk_type)
318 return dev_to_disk(dev);
323 static void part_stop(struct seq_file *part, void *v)
325 mutex_unlock(&block_class_lock);
328 static int show_partition(struct seq_file *part, void *v)
330 struct gendisk *sgp = v;
332 char buf[BDEVNAME_SIZE];
334 if (&sgp->dev.node == block_class.devices.next)
335 seq_puts(part, "major minor #blocks name\n\n");
337 /* Don't show non-partitionable removeable devices or empty devices */
338 if (!get_capacity(sgp) ||
339 (sgp->minors == 1 && (sgp->flags & GENHD_FL_REMOVABLE)))
341 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
344 /* show the full disk and all non-0 size partitions of it */
345 seq_printf(part, "%4d %4d %10llu %s\n",
346 sgp->major, sgp->first_minor,
347 (unsigned long long)get_capacity(sgp) >> 1,
348 disk_name(sgp, 0, buf));
349 for (n = 0; n < sgp->minors - 1; n++) {
352 if (sgp->part[n]->nr_sects == 0)
354 seq_printf(part, "%4d %4d %10llu %s\n",
355 sgp->major, n + 1 + sgp->first_minor,
356 (unsigned long long)sgp->part[n]->nr_sects >> 1 ,
357 disk_name(sgp, n + 1, buf));
363 const struct seq_operations partitions_op = {
367 .show = show_partition
372 static struct kobject *base_probe(dev_t devt, int *part, void *data)
374 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
375 /* Make old-style 2.4 aliases work */
376 request_module("block-major-%d", MAJOR(devt));
380 static int __init genhd_device_init(void)
384 block_class.dev_kobj = sysfs_dev_block_kobj;
385 error = class_register(&block_class);
388 bdev_map = kobj_map_init(base_probe, &block_class_lock);
391 #ifndef CONFIG_SYSFS_DEPRECATED
392 /* create top-level block dir */
393 block_depr = kobject_create_and_add("block", NULL);
398 subsys_initcall(genhd_device_init);
400 static ssize_t disk_range_show(struct device *dev,
401 struct device_attribute *attr, char *buf)
403 struct gendisk *disk = dev_to_disk(dev);
405 return sprintf(buf, "%d\n", disk->minors);
408 static ssize_t disk_removable_show(struct device *dev,
409 struct device_attribute *attr, char *buf)
411 struct gendisk *disk = dev_to_disk(dev);
413 return sprintf(buf, "%d\n",
414 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
417 static ssize_t disk_ro_show(struct device *dev,
418 struct device_attribute *attr, char *buf)
420 struct gendisk *disk = dev_to_disk(dev);
422 return sprintf(buf, "%d\n", disk->policy ? 1 : 0);
425 static ssize_t disk_size_show(struct device *dev,
426 struct device_attribute *attr, char *buf)
428 struct gendisk *disk = dev_to_disk(dev);
430 return sprintf(buf, "%llu\n", (unsigned long long)get_capacity(disk));
433 static ssize_t disk_capability_show(struct device *dev,
434 struct device_attribute *attr, char *buf)
436 struct gendisk *disk = dev_to_disk(dev);
438 return sprintf(buf, "%x\n", disk->flags);
441 static ssize_t disk_stat_show(struct device *dev,
442 struct device_attribute *attr, char *buf)
444 struct gendisk *disk = dev_to_disk(dev);
447 disk_round_stats(disk);
450 "%8lu %8lu %8llu %8u "
451 "%8lu %8lu %8llu %8u "
454 disk_stat_read(disk, ios[READ]),
455 disk_stat_read(disk, merges[READ]),
456 (unsigned long long)disk_stat_read(disk, sectors[READ]),
457 jiffies_to_msecs(disk_stat_read(disk, ticks[READ])),
458 disk_stat_read(disk, ios[WRITE]),
459 disk_stat_read(disk, merges[WRITE]),
460 (unsigned long long)disk_stat_read(disk, sectors[WRITE]),
461 jiffies_to_msecs(disk_stat_read(disk, ticks[WRITE])),
463 jiffies_to_msecs(disk_stat_read(disk, io_ticks)),
464 jiffies_to_msecs(disk_stat_read(disk, time_in_queue)));
467 #ifdef CONFIG_FAIL_MAKE_REQUEST
468 static ssize_t disk_fail_show(struct device *dev,
469 struct device_attribute *attr, char *buf)
471 struct gendisk *disk = dev_to_disk(dev);
473 return sprintf(buf, "%d\n", disk->flags & GENHD_FL_FAIL ? 1 : 0);
476 static ssize_t disk_fail_store(struct device *dev,
477 struct device_attribute *attr,
478 const char *buf, size_t count)
480 struct gendisk *disk = dev_to_disk(dev);
483 if (count > 0 && sscanf(buf, "%d", &i) > 0) {
485 disk->flags &= ~GENHD_FL_FAIL;
487 disk->flags |= GENHD_FL_FAIL;
495 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
496 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
497 static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
498 static DEVICE_ATTR(size, S_IRUGO, disk_size_show, NULL);
499 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
500 static DEVICE_ATTR(stat, S_IRUGO, disk_stat_show, NULL);
501 #ifdef CONFIG_FAIL_MAKE_REQUEST
502 static struct device_attribute dev_attr_fail =
503 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, disk_fail_show, disk_fail_store);
506 static struct attribute *disk_attrs[] = {
507 &dev_attr_range.attr,
508 &dev_attr_removable.attr,
511 &dev_attr_capability.attr,
513 #ifdef CONFIG_FAIL_MAKE_REQUEST
519 static struct attribute_group disk_attr_group = {
523 static struct attribute_group *disk_attr_groups[] = {
528 static void disk_release(struct device *dev)
530 struct gendisk *disk = dev_to_disk(dev);
534 free_disk_stats(disk);
537 struct class block_class = {
541 static struct device_type disk_type = {
543 .groups = disk_attr_groups,
544 .release = disk_release,
548 * aggregate disk stat collector. Uses the same stats that the sysfs
549 * entries do, above, but makes them available through one seq_file.
551 * The output looks suspiciously like /proc/partitions with a bunch of
555 static void *diskstats_start(struct seq_file *part, loff_t *pos)
560 mutex_lock(&block_class_lock);
561 list_for_each_entry(dev, &block_class.devices, node) {
562 if (dev->type != &disk_type)
565 return dev_to_disk(dev);
570 static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos)
572 struct gendisk *gp = v;
576 list_for_each_entry(dev, &gp->dev.node, node) {
577 if (&dev->node == &block_class.devices)
579 if (dev->type == &disk_type)
580 return dev_to_disk(dev);
585 static void diskstats_stop(struct seq_file *part, void *v)
587 mutex_unlock(&block_class_lock);
590 static int diskstats_show(struct seq_file *s, void *v)
592 struct gendisk *gp = v;
593 char buf[BDEVNAME_SIZE];
597 if (&gp->dev.kobj.entry == block_class.devices.next)
598 seq_puts(s, "major minor name"
599 " rio rmerge rsect ruse wio wmerge "
600 "wsect wuse running use aveq"
605 disk_round_stats(gp);
607 seq_printf(s, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n",
608 gp->major, n + gp->first_minor, disk_name(gp, n, buf),
609 disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]),
610 (unsigned long long)disk_stat_read(gp, sectors[0]),
611 jiffies_to_msecs(disk_stat_read(gp, ticks[0])),
612 disk_stat_read(gp, ios[1]), disk_stat_read(gp, merges[1]),
613 (unsigned long long)disk_stat_read(gp, sectors[1]),
614 jiffies_to_msecs(disk_stat_read(gp, ticks[1])),
616 jiffies_to_msecs(disk_stat_read(gp, io_ticks)),
617 jiffies_to_msecs(disk_stat_read(gp, time_in_queue)));
619 /* now show all non-0 size partitions of it */
620 for (n = 0; n < gp->minors - 1; n++) {
621 struct hd_struct *hd = gp->part[n];
623 if (!hd || !hd->nr_sects)
627 part_round_stats(hd);
629 seq_printf(s, "%4d %4d %s %lu %lu %llu "
630 "%u %lu %lu %llu %u %u %u %u\n",
631 gp->major, n + gp->first_minor + 1,
632 disk_name(gp, n + 1, buf),
633 part_stat_read(hd, ios[0]),
634 part_stat_read(hd, merges[0]),
635 (unsigned long long)part_stat_read(hd, sectors[0]),
636 jiffies_to_msecs(part_stat_read(hd, ticks[0])),
637 part_stat_read(hd, ios[1]),
638 part_stat_read(hd, merges[1]),
639 (unsigned long long)part_stat_read(hd, sectors[1]),
640 jiffies_to_msecs(part_stat_read(hd, ticks[1])),
642 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
643 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
650 const struct seq_operations diskstats_op = {
651 .start = diskstats_start,
652 .next = diskstats_next,
653 .stop = diskstats_stop,
654 .show = diskstats_show
657 static void media_change_notify_thread(struct work_struct *work)
659 struct gendisk *gd = container_of(work, struct gendisk, async_notify);
660 char event[] = "MEDIA_CHANGE=1";
661 char *envp[] = { event, NULL };
664 * set enviroment vars to indicate which event this is for
665 * so that user space will know to go check the media status.
667 kobject_uevent_env(&gd->dev.kobj, KOBJ_CHANGE, envp);
668 put_device(gd->driverfs_dev);
672 void genhd_media_change_notify(struct gendisk *disk)
674 get_device(disk->driverfs_dev);
675 schedule_work(&disk->async_notify);
677 EXPORT_SYMBOL_GPL(genhd_media_change_notify);
680 dev_t blk_lookup_devt(const char *name, int part)
683 dev_t devt = MKDEV(0, 0);
685 mutex_lock(&block_class_lock);
686 list_for_each_entry(dev, &block_class.devices, node) {
687 if (dev->type != &disk_type)
689 if (strcmp(dev->bus_id, name) == 0) {
690 struct gendisk *disk = dev_to_disk(dev);
692 if (part < disk->minors)
693 devt = MKDEV(MAJOR(dev->devt),
694 MINOR(dev->devt) + part);
698 mutex_unlock(&block_class_lock);
702 EXPORT_SYMBOL(blk_lookup_devt);
704 struct gendisk *alloc_disk(int minors)
706 return alloc_disk_node(minors, -1);
709 struct gendisk *alloc_disk_node(int minors, int node_id)
711 struct gendisk *disk;
713 disk = kmalloc_node(sizeof(struct gendisk),
714 GFP_KERNEL | __GFP_ZERO, node_id);
716 if (!init_disk_stats(disk)) {
721 int size = (minors - 1) * sizeof(struct hd_struct *);
722 disk->part = kmalloc_node(size,
723 GFP_KERNEL | __GFP_ZERO, node_id);
725 free_disk_stats(disk);
730 disk->minors = minors;
731 rand_initialize_disk(disk);
732 disk->dev.class = &block_class;
733 disk->dev.type = &disk_type;
734 device_initialize(&disk->dev);
735 INIT_WORK(&disk->async_notify,
736 media_change_notify_thread);
741 EXPORT_SYMBOL(alloc_disk);
742 EXPORT_SYMBOL(alloc_disk_node);
744 struct kobject *get_disk(struct gendisk *disk)
746 struct module *owner;
747 struct kobject *kobj;
751 owner = disk->fops->owner;
752 if (owner && !try_module_get(owner))
754 kobj = kobject_get(&disk->dev.kobj);
763 EXPORT_SYMBOL(get_disk);
765 void put_disk(struct gendisk *disk)
768 kobject_put(&disk->dev.kobj);
771 EXPORT_SYMBOL(put_disk);
773 void set_device_ro(struct block_device *bdev, int flag)
775 if (bdev->bd_contains != bdev)
776 bdev->bd_part->policy = flag;
778 bdev->bd_disk->policy = flag;
781 EXPORT_SYMBOL(set_device_ro);
783 void set_disk_ro(struct gendisk *disk, int flag)
787 for (i = 0; i < disk->minors - 1; i++)
788 if (disk->part[i]) disk->part[i]->policy = flag;
791 EXPORT_SYMBOL(set_disk_ro);
793 int bdev_read_only(struct block_device *bdev)
797 else if (bdev->bd_contains != bdev)
798 return bdev->bd_part->policy;
800 return bdev->bd_disk->policy;
803 EXPORT_SYMBOL(bdev_read_only);
805 int invalidate_partition(struct gendisk *disk, int index)
808 struct block_device *bdev = bdget_disk(disk, index);
811 res = __invalidate_device(bdev);
817 EXPORT_SYMBOL(invalidate_partition);