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 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),
203 * get_gendisk - get partitioning information for a given device
204 * @dev: device to get partitioning information for
206 * This function gets the structure containing partitioning
207 * information for the given device @dev.
209 struct gendisk *get_gendisk(dev_t devt, int *part)
211 struct kobject *kobj = kobj_lookup(bdev_map, devt, part);
212 struct device *dev = kobj_to_dev(kobj);
214 return kobj ? dev_to_disk(dev) : 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)
226 char buf[BDEVNAME_SIZE];
229 mutex_lock(&block_class_lock);
230 /* For each block device... */
231 list_for_each_entry(dev, &block_class.devices, node) {
232 if (dev->type != &disk_type)
234 sgp = dev_to_disk(dev);
236 * Don't show empty devices or things that have been surpressed
238 if (get_capacity(sgp) == 0 ||
239 (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
243 * Note, unlike /proc/partitions, I am showing the numbers in
244 * hex - the same format as the root= option takes.
246 printk("%02x%02x %10llu %s",
247 sgp->major, sgp->first_minor,
248 (unsigned long long)get_capacity(sgp) >> 1,
249 disk_name(sgp, 0, buf));
250 if (sgp->driverfs_dev != NULL &&
251 sgp->driverfs_dev->driver != NULL)
252 printk(" driver: %s\n",
253 sgp->driverfs_dev->driver->name);
255 printk(" (driver?)\n");
257 /* now show the partitions */
258 for (n = 0; n < sgp->minors - 1; ++n) {
259 if (sgp->part[n] == NULL)
261 if (sgp->part[n]->nr_sects == 0)
263 printk(" %02x%02x %10llu %s\n",
264 sgp->major, n + 1 + sgp->first_minor,
265 (unsigned long long)sgp->part[n]->nr_sects >> 1,
266 disk_name(sgp, n + 1, buf));
270 mutex_unlock(&block_class_lock);
273 #ifdef CONFIG_PROC_FS
275 static void *part_start(struct seq_file *part, loff_t *pos)
280 mutex_lock(&block_class_lock);
281 list_for_each_entry(dev, &block_class.devices, node) {
282 if (dev->type != &disk_type)
285 return dev_to_disk(dev);
290 static void *part_next(struct seq_file *part, void *v, loff_t *pos)
292 struct gendisk *gp = v;
295 list_for_each_entry(dev, &gp->dev.node, node) {
296 if (&dev->node == &block_class.devices)
298 if (dev->type == &disk_type)
299 return dev_to_disk(dev);
304 static void part_stop(struct seq_file *part, void *v)
306 mutex_unlock(&block_class_lock);
309 static int show_partition(struct seq_file *part, void *v)
311 struct gendisk *sgp = v;
313 char buf[BDEVNAME_SIZE];
315 if (&sgp->dev.node == block_class.devices.next)
316 seq_puts(part, "major minor #blocks name\n\n");
318 /* Don't show non-partitionable removeable devices or empty devices */
319 if (!get_capacity(sgp) ||
320 (sgp->minors == 1 && (sgp->flags & GENHD_FL_REMOVABLE)))
322 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
325 /* show the full disk and all non-0 size partitions of it */
326 seq_printf(part, "%4d %4d %10llu %s\n",
327 sgp->major, sgp->first_minor,
328 (unsigned long long)get_capacity(sgp) >> 1,
329 disk_name(sgp, 0, buf));
330 for (n = 0; n < sgp->minors - 1; n++) {
333 if (sgp->part[n]->nr_sects == 0)
335 seq_printf(part, "%4d %4d %10llu %s\n",
336 sgp->major, n + 1 + sgp->first_minor,
337 (unsigned long long)sgp->part[n]->nr_sects >> 1 ,
338 disk_name(sgp, n + 1, buf));
344 const struct seq_operations partitions_op = {
348 .show = show_partition
353 static struct kobject *base_probe(dev_t devt, int *part, void *data)
355 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
356 /* Make old-style 2.4 aliases work */
357 request_module("block-major-%d", MAJOR(devt));
361 static int __init genhd_device_init(void)
363 class_register(&block_class);
364 bdev_map = kobj_map_init(base_probe, &block_class_lock);
367 #ifndef CONFIG_SYSFS_DEPRECATED
368 /* create top-level block dir */
369 block_depr = kobject_create_and_add("block", NULL);
374 subsys_initcall(genhd_device_init);
376 static ssize_t disk_range_show(struct device *dev,
377 struct device_attribute *attr, char *buf)
379 struct gendisk *disk = dev_to_disk(dev);
381 return sprintf(buf, "%d\n", disk->minors);
384 static ssize_t disk_removable_show(struct device *dev,
385 struct device_attribute *attr, char *buf)
387 struct gendisk *disk = dev_to_disk(dev);
389 return sprintf(buf, "%d\n",
390 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
393 static ssize_t disk_size_show(struct device *dev,
394 struct device_attribute *attr, char *buf)
396 struct gendisk *disk = dev_to_disk(dev);
398 return sprintf(buf, "%llu\n", (unsigned long long)get_capacity(disk));
401 static ssize_t disk_capability_show(struct device *dev,
402 struct device_attribute *attr, char *buf)
404 struct gendisk *disk = dev_to_disk(dev);
406 return sprintf(buf, "%x\n", disk->flags);
409 static ssize_t disk_stat_show(struct device *dev,
410 struct device_attribute *attr, char *buf)
412 struct gendisk *disk = dev_to_disk(dev);
415 disk_round_stats(disk);
418 "%8lu %8lu %8llu %8u "
419 "%8lu %8lu %8llu %8u "
422 disk_stat_read(disk, ios[READ]),
423 disk_stat_read(disk, merges[READ]),
424 (unsigned long long)disk_stat_read(disk, sectors[READ]),
425 jiffies_to_msecs(disk_stat_read(disk, ticks[READ])),
426 disk_stat_read(disk, ios[WRITE]),
427 disk_stat_read(disk, merges[WRITE]),
428 (unsigned long long)disk_stat_read(disk, sectors[WRITE]),
429 jiffies_to_msecs(disk_stat_read(disk, ticks[WRITE])),
431 jiffies_to_msecs(disk_stat_read(disk, io_ticks)),
432 jiffies_to_msecs(disk_stat_read(disk, time_in_queue)));
435 #ifdef CONFIG_FAIL_MAKE_REQUEST
436 static ssize_t disk_fail_show(struct device *dev,
437 struct device_attribute *attr, char *buf)
439 struct gendisk *disk = dev_to_disk(dev);
441 return sprintf(buf, "%d\n", disk->flags & GENHD_FL_FAIL ? 1 : 0);
444 static ssize_t disk_fail_store(struct device *dev,
445 struct device_attribute *attr,
446 const char *buf, size_t count)
448 struct gendisk *disk = dev_to_disk(dev);
451 if (count > 0 && sscanf(buf, "%d", &i) > 0) {
453 disk->flags &= ~GENHD_FL_FAIL;
455 disk->flags |= GENHD_FL_FAIL;
463 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
464 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
465 static DEVICE_ATTR(size, S_IRUGO, disk_size_show, NULL);
466 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
467 static DEVICE_ATTR(stat, S_IRUGO, disk_stat_show, NULL);
468 #ifdef CONFIG_FAIL_MAKE_REQUEST
469 static struct device_attribute dev_attr_fail =
470 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, disk_fail_show, disk_fail_store);
473 static struct attribute *disk_attrs[] = {
474 &dev_attr_range.attr,
475 &dev_attr_removable.attr,
477 &dev_attr_capability.attr,
479 #ifdef CONFIG_FAIL_MAKE_REQUEST
485 static struct attribute_group disk_attr_group = {
489 static struct attribute_group *disk_attr_groups[] = {
494 static void disk_release(struct device *dev)
496 struct gendisk *disk = dev_to_disk(dev);
500 free_disk_stats(disk);
503 struct class block_class = {
507 static struct device_type disk_type = {
509 .groups = disk_attr_groups,
510 .release = disk_release,
514 * aggregate disk stat collector. Uses the same stats that the sysfs
515 * entries do, above, but makes them available through one seq_file.
517 * The output looks suspiciously like /proc/partitions with a bunch of
521 static void *diskstats_start(struct seq_file *part, loff_t *pos)
526 mutex_lock(&block_class_lock);
527 list_for_each_entry(dev, &block_class.devices, node) {
528 if (dev->type != &disk_type)
531 return dev_to_disk(dev);
536 static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos)
538 struct gendisk *gp = v;
542 list_for_each_entry(dev, &gp->dev.node, node) {
543 if (&dev->node == &block_class.devices)
545 if (dev->type == &disk_type)
546 return dev_to_disk(dev);
551 static void diskstats_stop(struct seq_file *part, void *v)
553 mutex_unlock(&block_class_lock);
556 static int diskstats_show(struct seq_file *s, void *v)
558 struct gendisk *gp = v;
559 char buf[BDEVNAME_SIZE];
563 if (&gp->dev.kobj.entry == block_class.devices.next)
564 seq_puts(s, "major minor name"
565 " rio rmerge rsect ruse wio wmerge "
566 "wsect wuse running use aveq"
571 disk_round_stats(gp);
573 seq_printf(s, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n",
574 gp->major, n + gp->first_minor, disk_name(gp, n, buf),
575 disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]),
576 (unsigned long long)disk_stat_read(gp, sectors[0]),
577 jiffies_to_msecs(disk_stat_read(gp, ticks[0])),
578 disk_stat_read(gp, ios[1]), disk_stat_read(gp, merges[1]),
579 (unsigned long long)disk_stat_read(gp, sectors[1]),
580 jiffies_to_msecs(disk_stat_read(gp, ticks[1])),
582 jiffies_to_msecs(disk_stat_read(gp, io_ticks)),
583 jiffies_to_msecs(disk_stat_read(gp, time_in_queue)));
585 /* now show all non-0 size partitions of it */
586 for (n = 0; n < gp->minors - 1; n++) {
587 struct hd_struct *hd = gp->part[n];
589 if (!hd || !hd->nr_sects)
593 part_round_stats(hd);
595 seq_printf(s, "%4d %4d %s %lu %lu %llu "
596 "%u %lu %lu %llu %u %u %u %u\n",
597 gp->major, n + gp->first_minor + 1,
598 disk_name(gp, n + 1, buf),
599 part_stat_read(hd, ios[0]),
600 part_stat_read(hd, merges[0]),
601 (unsigned long long)part_stat_read(hd, sectors[0]),
602 jiffies_to_msecs(part_stat_read(hd, ticks[0])),
603 part_stat_read(hd, ios[1]),
604 part_stat_read(hd, merges[1]),
605 (unsigned long long)part_stat_read(hd, sectors[1]),
606 jiffies_to_msecs(part_stat_read(hd, ticks[1])),
608 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
609 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
616 const struct seq_operations diskstats_op = {
617 .start = diskstats_start,
618 .next = diskstats_next,
619 .stop = diskstats_stop,
620 .show = diskstats_show
623 static void media_change_notify_thread(struct work_struct *work)
625 struct gendisk *gd = container_of(work, struct gendisk, async_notify);
626 char event[] = "MEDIA_CHANGE=1";
627 char *envp[] = { event, NULL };
630 * set enviroment vars to indicate which event this is for
631 * so that user space will know to go check the media status.
633 kobject_uevent_env(&gd->dev.kobj, KOBJ_CHANGE, envp);
634 put_device(gd->driverfs_dev);
638 void genhd_media_change_notify(struct gendisk *disk)
640 get_device(disk->driverfs_dev);
641 schedule_work(&disk->async_notify);
643 EXPORT_SYMBOL_GPL(genhd_media_change_notify);
646 dev_t blk_lookup_devt(const char *name)
649 dev_t devt = MKDEV(0, 0);
651 mutex_lock(&block_class_lock);
652 list_for_each_entry(dev, &block_class.devices, node) {
653 if (strcmp(dev->bus_id, name) == 0) {
658 mutex_unlock(&block_class_lock);
663 EXPORT_SYMBOL(blk_lookup_devt);
665 struct gendisk *alloc_disk(int minors)
667 return alloc_disk_node(minors, -1);
670 struct gendisk *alloc_disk_node(int minors, int node_id)
672 struct gendisk *disk;
674 disk = kmalloc_node(sizeof(struct gendisk),
675 GFP_KERNEL | __GFP_ZERO, node_id);
677 if (!init_disk_stats(disk)) {
682 int size = (minors - 1) * sizeof(struct hd_struct *);
683 disk->part = kmalloc_node(size,
684 GFP_KERNEL | __GFP_ZERO, node_id);
686 free_disk_stats(disk);
691 disk->minors = minors;
692 rand_initialize_disk(disk);
693 disk->dev.class = &block_class;
694 disk->dev.type = &disk_type;
695 device_initialize(&disk->dev);
696 INIT_WORK(&disk->async_notify,
697 media_change_notify_thread);
702 EXPORT_SYMBOL(alloc_disk);
703 EXPORT_SYMBOL(alloc_disk_node);
705 struct kobject *get_disk(struct gendisk *disk)
707 struct module *owner;
708 struct kobject *kobj;
712 owner = disk->fops->owner;
713 if (owner && !try_module_get(owner))
715 kobj = kobject_get(&disk->dev.kobj);
724 EXPORT_SYMBOL(get_disk);
726 void put_disk(struct gendisk *disk)
729 kobject_put(&disk->dev.kobj);
732 EXPORT_SYMBOL(put_disk);
734 void set_device_ro(struct block_device *bdev, int flag)
736 if (bdev->bd_contains != bdev)
737 bdev->bd_part->policy = flag;
739 bdev->bd_disk->policy = flag;
742 EXPORT_SYMBOL(set_device_ro);
744 void set_disk_ro(struct gendisk *disk, int flag)
748 for (i = 0; i < disk->minors - 1; i++)
749 if (disk->part[i]) disk->part[i]->policy = flag;
752 EXPORT_SYMBOL(set_disk_ro);
754 int bdev_read_only(struct block_device *bdev)
758 else if (bdev->bd_contains != bdev)
759 return bdev->bd_part->policy;
761 return bdev->bd_disk->policy;
764 EXPORT_SYMBOL(bdev_read_only);
766 int invalidate_partition(struct gendisk *disk, int index)
769 struct block_device *bdev = bdget_disk(disk, index);
772 res = __invalidate_device(bdev);
778 EXPORT_SYMBOL(invalidate_partition);