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/proc_fs.h>
14 #include <linux/seq_file.h>
15 #include <linux/slab.h>
16 #include <linux/kmod.h>
17 #include <linux/kobj_map.h>
18 #include <linux/buffer_head.h>
19 #include <linux/mutex.h>
20 #include <linux/idr.h>
24 static DEFINE_MUTEX(block_class_lock);
25 #ifndef CONFIG_SYSFS_DEPRECATED
26 struct kobject *block_depr;
29 /* for extended dynamic devt allocation, currently only one major is used */
30 #define MAX_EXT_DEVT (1 << MINORBITS)
32 /* For extended devt allocation. ext_devt_mutex prevents look up
33 * results from going away underneath its user.
35 static DEFINE_MUTEX(ext_devt_mutex);
36 static DEFINE_IDR(ext_devt_idr);
38 static struct device_type disk_type;
41 * disk_get_part - get partition
42 * @disk: disk to look partition from
43 * @partno: partition number
45 * Look for partition @partno from @disk. If found, increment
46 * reference count and return it.
52 * Pointer to the found partition on success, NULL if not found.
54 struct hd_struct *disk_get_part(struct gendisk *disk, int partno)
56 struct hd_struct *part = NULL;
57 struct disk_part_tbl *ptbl;
59 if (unlikely(partno < 0))
64 ptbl = rcu_dereference(disk->part_tbl);
65 if (likely(partno < ptbl->len)) {
66 part = rcu_dereference(ptbl->part[partno]);
68 get_device(part_to_dev(part));
75 EXPORT_SYMBOL_GPL(disk_get_part);
78 * disk_part_iter_init - initialize partition iterator
79 * @piter: iterator to initialize
80 * @disk: disk to iterate over
81 * @flags: DISK_PITER_* flags
83 * Initialize @piter so that it iterates over partitions of @disk.
88 void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk,
91 struct disk_part_tbl *ptbl;
94 ptbl = rcu_dereference(disk->part_tbl);
99 if (flags & DISK_PITER_REVERSE)
100 piter->idx = ptbl->len - 1;
101 else if (flags & DISK_PITER_INCL_PART0)
106 piter->flags = flags;
110 EXPORT_SYMBOL_GPL(disk_part_iter_init);
113 * disk_part_iter_next - proceed iterator to the next partition and return it
114 * @piter: iterator of interest
116 * Proceed @piter to the next partition and return it.
121 struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter)
123 struct disk_part_tbl *ptbl;
126 /* put the last partition */
127 disk_put_part(piter->part);
132 ptbl = rcu_dereference(piter->disk->part_tbl);
134 /* determine iteration parameters */
135 if (piter->flags & DISK_PITER_REVERSE) {
137 if (piter->flags & DISK_PITER_INCL_PART0)
146 /* iterate to the next partition */
147 for (; piter->idx != end; piter->idx += inc) {
148 struct hd_struct *part;
150 part = rcu_dereference(ptbl->part[piter->idx]);
153 if (!(piter->flags & DISK_PITER_INCL_EMPTY) && !part->nr_sects)
156 get_device(part_to_dev(part));
166 EXPORT_SYMBOL_GPL(disk_part_iter_next);
169 * disk_part_iter_exit - finish up partition iteration
170 * @piter: iter of interest
172 * Called when iteration is over. Cleans up @piter.
177 void disk_part_iter_exit(struct disk_part_iter *piter)
179 disk_put_part(piter->part);
182 EXPORT_SYMBOL_GPL(disk_part_iter_exit);
184 static inline int sector_in_part(struct hd_struct *part, sector_t sector)
186 return part->start_sect <= sector &&
187 sector < part->start_sect + part->nr_sects;
191 * disk_map_sector_rcu - map sector to partition
192 * @disk: gendisk of interest
193 * @sector: sector to map
195 * Find out which partition @sector maps to on @disk. This is
196 * primarily used for stats accounting.
199 * RCU read locked. The returned partition pointer is valid only
200 * while preemption is disabled.
203 * Found partition on success, part0 is returned if no partition matches
205 struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector)
207 struct disk_part_tbl *ptbl;
208 struct hd_struct *part;
211 ptbl = rcu_dereference(disk->part_tbl);
213 part = rcu_dereference(ptbl->last_lookup);
214 if (part && sector_in_part(part, sector))
217 for (i = 1; i < ptbl->len; i++) {
218 part = rcu_dereference(ptbl->part[i]);
220 if (part && sector_in_part(part, sector)) {
221 rcu_assign_pointer(ptbl->last_lookup, part);
227 EXPORT_SYMBOL_GPL(disk_map_sector_rcu);
230 * Can be deleted altogether. Later.
233 static struct blk_major_name {
234 struct blk_major_name *next;
237 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
239 /* index in the above - for now: assume no multimajor ranges */
240 static inline int major_to_index(int major)
242 return major % BLKDEV_MAJOR_HASH_SIZE;
245 #ifdef CONFIG_PROC_FS
246 void blkdev_show(struct seq_file *seqf, off_t offset)
248 struct blk_major_name *dp;
250 if (offset < BLKDEV_MAJOR_HASH_SIZE) {
251 mutex_lock(&block_class_lock);
252 for (dp = major_names[offset]; dp; dp = dp->next)
253 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
254 mutex_unlock(&block_class_lock);
257 #endif /* CONFIG_PROC_FS */
260 * register_blkdev - register a new block device
262 * @major: the requested major device number [1..255]. If @major=0, try to
263 * allocate any unused major number.
264 * @name: the name of the new block device as a zero terminated string
266 * The @name must be unique within the system.
268 * The return value depends on the @major input parameter.
269 * - if a major device number was requested in range [1..255] then the
270 * function returns zero on success, or a negative error code
271 * - if any unused major number was requested with @major=0 parameter
272 * then the return value is the allocated major number in range
273 * [1..255] or a negative error code otherwise
275 int register_blkdev(unsigned int major, const char *name)
277 struct blk_major_name **n, *p;
280 mutex_lock(&block_class_lock);
284 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
285 if (major_names[index] == NULL)
290 printk("register_blkdev: failed to get major for %s\n",
299 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
306 strlcpy(p->name, name, sizeof(p->name));
308 index = major_to_index(major);
310 for (n = &major_names[index]; *n; n = &(*n)->next) {
311 if ((*n)->major == major)
320 printk("register_blkdev: cannot get major %d for %s\n",
325 mutex_unlock(&block_class_lock);
329 EXPORT_SYMBOL(register_blkdev);
331 void unregister_blkdev(unsigned int major, const char *name)
333 struct blk_major_name **n;
334 struct blk_major_name *p = NULL;
335 int index = major_to_index(major);
337 mutex_lock(&block_class_lock);
338 for (n = &major_names[index]; *n; n = &(*n)->next)
339 if ((*n)->major == major)
341 if (!*n || strcmp((*n)->name, name)) {
347 mutex_unlock(&block_class_lock);
351 EXPORT_SYMBOL(unregister_blkdev);
353 static struct kobj_map *bdev_map;
356 * blk_mangle_minor - scatter minor numbers apart
357 * @minor: minor number to mangle
359 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
360 * is enabled. Mangling twice gives the original value.
368 static int blk_mangle_minor(int minor)
370 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
373 for (i = 0; i < MINORBITS / 2; i++) {
374 int low = minor & (1 << i);
375 int high = minor & (1 << (MINORBITS - 1 - i));
376 int distance = MINORBITS - 1 - 2 * i;
378 minor ^= low | high; /* clear both bits */
379 low <<= distance; /* swap the positions */
381 minor |= low | high; /* and set */
388 * blk_alloc_devt - allocate a dev_t for a partition
389 * @part: partition to allocate dev_t for
390 * @devt: out parameter for resulting dev_t
392 * Allocate a dev_t for block device.
395 * 0 on success, allocated dev_t is returned in *@devt. -errno on
401 int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
403 struct gendisk *disk = part_to_disk(part);
406 /* in consecutive minor range? */
407 if (part->partno < disk->minors) {
408 *devt = MKDEV(disk->major, disk->first_minor + part->partno);
412 /* allocate ext devt */
414 if (!idr_pre_get(&ext_devt_idr, GFP_KERNEL))
416 rc = idr_get_new(&ext_devt_idr, part, &idx);
417 } while (rc == -EAGAIN);
422 if (idx > MAX_EXT_DEVT) {
423 idr_remove(&ext_devt_idr, idx);
427 *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
432 * blk_free_devt - free a dev_t
433 * @devt: dev_t to free
435 * Free @devt which was allocated using blk_alloc_devt().
440 void blk_free_devt(dev_t devt)
444 if (devt == MKDEV(0, 0))
447 if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
448 mutex_lock(&ext_devt_mutex);
449 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
450 mutex_unlock(&ext_devt_mutex);
454 static char *bdevt_str(dev_t devt, char *buf)
456 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
457 char tbuf[BDEVT_SIZE];
458 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
459 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
461 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
467 * Register device numbers dev..(dev+range-1)
468 * range must be nonzero
469 * The hash chain is sorted on range, so that subranges can override.
471 void blk_register_region(dev_t devt, unsigned long range, struct module *module,
472 struct kobject *(*probe)(dev_t, int *, void *),
473 int (*lock)(dev_t, void *), void *data)
475 kobj_map(bdev_map, devt, range, module, probe, lock, data);
478 EXPORT_SYMBOL(blk_register_region);
480 void blk_unregister_region(dev_t devt, unsigned long range)
482 kobj_unmap(bdev_map, devt, range);
485 EXPORT_SYMBOL(blk_unregister_region);
487 static struct kobject *exact_match(dev_t devt, int *partno, void *data)
489 struct gendisk *p = data;
491 return &disk_to_dev(p)->kobj;
494 static int exact_lock(dev_t devt, void *data)
496 struct gendisk *p = data;
504 * add_disk - add partitioning information to kernel list
505 * @disk: per-device partitioning information
507 * This function registers the partitioning information in @disk
510 * FIXME: error handling
512 void add_disk(struct gendisk *disk)
514 struct backing_dev_info *bdi;
518 /* minors == 0 indicates to use ext devt from part0 and should
519 * be accompanied with EXT_DEVT flag. Make sure all
520 * parameters make sense.
522 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
523 WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT));
525 disk->flags |= GENHD_FL_UP;
527 retval = blk_alloc_devt(&disk->part0, &devt);
532 disk_to_dev(disk)->devt = devt;
534 /* ->major and ->first_minor aren't supposed to be
535 * dereferenced from here on, but set them just in case.
537 disk->major = MAJOR(devt);
538 disk->first_minor = MINOR(devt);
540 blk_register_region(disk_devt(disk), disk->minors, NULL,
541 exact_match, exact_lock, disk);
543 blk_register_queue(disk);
545 bdi = &disk->queue->backing_dev_info;
546 bdi_register_dev(bdi, disk_devt(disk));
547 retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
552 EXPORT_SYMBOL(add_disk);
553 EXPORT_SYMBOL(del_gendisk); /* in partitions/check.c */
555 void unlink_gendisk(struct gendisk *disk)
557 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
558 bdi_unregister(&disk->queue->backing_dev_info);
559 blk_unregister_queue(disk);
560 blk_unregister_region(disk_devt(disk), disk->minors);
564 * get_gendisk - get partitioning information for a given device
565 * @devt: device to get partitioning information for
566 * @partno: returned partition index
568 * This function gets the structure containing partitioning
569 * information for the given device @devt.
571 struct gendisk *get_gendisk(dev_t devt, int *partno)
573 struct gendisk *disk = NULL;
575 if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
576 struct kobject *kobj;
578 kobj = kobj_lookup(bdev_map, devt, partno);
580 disk = dev_to_disk(kobj_to_dev(kobj));
582 struct hd_struct *part;
584 mutex_lock(&ext_devt_mutex);
585 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
586 if (part && get_disk(part_to_disk(part))) {
587 *partno = part->partno;
588 disk = part_to_disk(part);
590 mutex_unlock(&ext_devt_mutex);
597 * bdget_disk - do bdget() by gendisk and partition number
598 * @disk: gendisk of interest
599 * @partno: partition number
601 * Find partition @partno from @disk, do bdget() on it.
607 * Resulting block_device on success, NULL on failure.
609 struct block_device *bdget_disk(struct gendisk *disk, int partno)
611 struct hd_struct *part;
612 struct block_device *bdev = NULL;
614 part = disk_get_part(disk, partno);
616 bdev = bdget(part_devt(part));
621 EXPORT_SYMBOL(bdget_disk);
624 * print a full list of all partitions - intended for places where the root
625 * filesystem can't be mounted and thus to give the victim some idea of what
628 void __init printk_all_partitions(void)
630 struct class_dev_iter iter;
633 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
634 while ((dev = class_dev_iter_next(&iter))) {
635 struct gendisk *disk = dev_to_disk(dev);
636 struct disk_part_iter piter;
637 struct hd_struct *part;
638 char name_buf[BDEVNAME_SIZE];
639 char devt_buf[BDEVT_SIZE];
642 * Don't show empty devices or things that have been
645 if (get_capacity(disk) == 0 ||
646 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
650 * Note, unlike /proc/partitions, I am showing the
651 * numbers in hex - the same format as the root=
654 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
655 while ((part = disk_part_iter_next(&piter))) {
656 bool is_part0 = part == &disk->part0;
658 printk("%s%s %10llu %s", is_part0 ? "" : " ",
659 bdevt_str(part_devt(part), devt_buf),
660 (unsigned long long)part->nr_sects >> 1,
661 disk_name(disk, part->partno, name_buf));
663 if (disk->driverfs_dev != NULL &&
664 disk->driverfs_dev->driver != NULL)
665 printk(" driver: %s\n",
666 disk->driverfs_dev->driver->name);
668 printk(" (driver?)\n");
672 disk_part_iter_exit(&piter);
674 class_dev_iter_exit(&iter);
677 #ifdef CONFIG_PROC_FS
679 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
682 struct class_dev_iter *iter;
685 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
687 return ERR_PTR(-ENOMEM);
689 seqf->private = iter;
690 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
692 dev = class_dev_iter_next(iter);
697 return dev_to_disk(dev);
700 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
705 dev = class_dev_iter_next(seqf->private);
707 return dev_to_disk(dev);
712 static void disk_seqf_stop(struct seq_file *seqf, void *v)
714 struct class_dev_iter *iter = seqf->private;
716 /* stop is called even after start failed :-( */
718 class_dev_iter_exit(iter);
723 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
727 p = disk_seqf_start(seqf, pos);
728 if (!IS_ERR(p) && p && !*pos)
729 seq_puts(seqf, "major minor #blocks name\n\n");
733 static int show_partition(struct seq_file *seqf, void *v)
735 struct gendisk *sgp = v;
736 struct disk_part_iter piter;
737 struct hd_struct *part;
738 char buf[BDEVNAME_SIZE];
740 /* Don't show non-partitionable removeable devices or empty devices */
741 if (!get_capacity(sgp) || (!disk_partitionable(sgp) &&
742 (sgp->flags & GENHD_FL_REMOVABLE)))
744 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
747 /* show the full disk and all non-0 size partitions of it */
748 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
749 while ((part = disk_part_iter_next(&piter)))
750 seq_printf(seqf, "%4d %7d %10llu %s\n",
751 MAJOR(part_devt(part)), MINOR(part_devt(part)),
752 (unsigned long long)part->nr_sects >> 1,
753 disk_name(sgp, part->partno, buf));
754 disk_part_iter_exit(&piter);
759 static const struct seq_operations partitions_op = {
760 .start = show_partition_start,
761 .next = disk_seqf_next,
762 .stop = disk_seqf_stop,
763 .show = show_partition
766 static int partitions_open(struct inode *inode, struct file *file)
768 return seq_open(file, &partitions_op);
771 static const struct file_operations proc_partitions_operations = {
772 .open = partitions_open,
775 .release = seq_release,
780 static struct kobject *base_probe(dev_t devt, int *partno, void *data)
782 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
783 /* Make old-style 2.4 aliases work */
784 request_module("block-major-%d", MAJOR(devt));
788 static int __init genhd_device_init(void)
792 block_class.dev_kobj = sysfs_dev_block_kobj;
793 error = class_register(&block_class);
796 bdev_map = kobj_map_init(base_probe, &block_class_lock);
799 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
801 #ifndef CONFIG_SYSFS_DEPRECATED
802 /* create top-level block dir */
803 block_depr = kobject_create_and_add("block", NULL);
808 subsys_initcall(genhd_device_init);
810 static ssize_t disk_range_show(struct device *dev,
811 struct device_attribute *attr, char *buf)
813 struct gendisk *disk = dev_to_disk(dev);
815 return sprintf(buf, "%d\n", disk->minors);
818 static ssize_t disk_ext_range_show(struct device *dev,
819 struct device_attribute *attr, char *buf)
821 struct gendisk *disk = dev_to_disk(dev);
823 return sprintf(buf, "%d\n", disk_max_parts(disk));
826 static ssize_t disk_removable_show(struct device *dev,
827 struct device_attribute *attr, char *buf)
829 struct gendisk *disk = dev_to_disk(dev);
831 return sprintf(buf, "%d\n",
832 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
835 static ssize_t disk_ro_show(struct device *dev,
836 struct device_attribute *attr, char *buf)
838 struct gendisk *disk = dev_to_disk(dev);
840 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
843 static ssize_t disk_capability_show(struct device *dev,
844 struct device_attribute *attr, char *buf)
846 struct gendisk *disk = dev_to_disk(dev);
848 return sprintf(buf, "%x\n", disk->flags);
851 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
852 static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
853 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
854 static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
855 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
856 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
857 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
858 #ifdef CONFIG_FAIL_MAKE_REQUEST
859 static struct device_attribute dev_attr_fail =
860 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
862 #ifdef CONFIG_FAIL_IO_TIMEOUT
863 static struct device_attribute dev_attr_fail_timeout =
864 __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show,
868 static struct attribute *disk_attrs[] = {
869 &dev_attr_range.attr,
870 &dev_attr_ext_range.attr,
871 &dev_attr_removable.attr,
874 &dev_attr_capability.attr,
876 #ifdef CONFIG_FAIL_MAKE_REQUEST
879 #ifdef CONFIG_FAIL_IO_TIMEOUT
880 &dev_attr_fail_timeout.attr,
885 static struct attribute_group disk_attr_group = {
889 static struct attribute_group *disk_attr_groups[] = {
894 static void disk_free_ptbl_rcu_cb(struct rcu_head *head)
896 struct disk_part_tbl *ptbl =
897 container_of(head, struct disk_part_tbl, rcu_head);
903 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
904 * @disk: disk to replace part_tbl for
905 * @new_ptbl: new part_tbl to install
907 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
908 * original ptbl is freed using RCU callback.
911 * Matching bd_mutx locked.
913 static void disk_replace_part_tbl(struct gendisk *disk,
914 struct disk_part_tbl *new_ptbl)
916 struct disk_part_tbl *old_ptbl = disk->part_tbl;
918 rcu_assign_pointer(disk->part_tbl, new_ptbl);
921 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
922 call_rcu(&old_ptbl->rcu_head, disk_free_ptbl_rcu_cb);
927 * disk_expand_part_tbl - expand disk->part_tbl
928 * @disk: disk to expand part_tbl for
929 * @partno: expand such that this partno can fit in
931 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
932 * uses RCU to allow unlocked dereferencing for stats and other stuff.
935 * Matching bd_mutex locked, might sleep.
938 * 0 on success, -errno on failure.
940 int disk_expand_part_tbl(struct gendisk *disk, int partno)
942 struct disk_part_tbl *old_ptbl = disk->part_tbl;
943 struct disk_part_tbl *new_ptbl;
944 int len = old_ptbl ? old_ptbl->len : 0;
945 int target = partno + 1;
949 /* disk_max_parts() is zero during initialization, ignore if so */
950 if (disk_max_parts(disk) && target > disk_max_parts(disk))
956 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
957 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
961 INIT_RCU_HEAD(&new_ptbl->rcu_head);
962 new_ptbl->len = target;
964 for (i = 0; i < len; i++)
965 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
967 disk_replace_part_tbl(disk, new_ptbl);
971 static void disk_release(struct device *dev)
973 struct gendisk *disk = dev_to_disk(dev);
976 disk_replace_part_tbl(disk, NULL);
977 free_part_stats(&disk->part0);
980 struct class block_class = {
984 static struct device_type disk_type = {
986 .groups = disk_attr_groups,
987 .release = disk_release,
990 #ifdef CONFIG_PROC_FS
992 * aggregate disk stat collector. Uses the same stats that the sysfs
993 * entries do, above, but makes them available through one seq_file.
995 * The output looks suspiciously like /proc/partitions with a bunch of
998 static int diskstats_show(struct seq_file *seqf, void *v)
1000 struct gendisk *gp = v;
1001 struct disk_part_iter piter;
1002 struct hd_struct *hd;
1003 char buf[BDEVNAME_SIZE];
1007 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1008 seq_puts(seqf, "major minor name"
1009 " rio rmerge rsect ruse wio wmerge "
1010 "wsect wuse running use aveq"
1014 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_PART0);
1015 while ((hd = disk_part_iter_next(&piter))) {
1016 cpu = part_stat_lock();
1017 part_round_stats(cpu, hd);
1019 seq_printf(seqf, "%4d %7d %s %lu %lu %llu "
1020 "%u %lu %lu %llu %u %u %u %u\n",
1021 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1022 disk_name(gp, hd->partno, buf),
1023 part_stat_read(hd, ios[0]),
1024 part_stat_read(hd, merges[0]),
1025 (unsigned long long)part_stat_read(hd, sectors[0]),
1026 jiffies_to_msecs(part_stat_read(hd, ticks[0])),
1027 part_stat_read(hd, ios[1]),
1028 part_stat_read(hd, merges[1]),
1029 (unsigned long long)part_stat_read(hd, sectors[1]),
1030 jiffies_to_msecs(part_stat_read(hd, ticks[1])),
1032 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1033 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1036 disk_part_iter_exit(&piter);
1041 static const struct seq_operations diskstats_op = {
1042 .start = disk_seqf_start,
1043 .next = disk_seqf_next,
1044 .stop = disk_seqf_stop,
1045 .show = diskstats_show
1048 static int diskstats_open(struct inode *inode, struct file *file)
1050 return seq_open(file, &diskstats_op);
1053 static const struct file_operations proc_diskstats_operations = {
1054 .open = diskstats_open,
1056 .llseek = seq_lseek,
1057 .release = seq_release,
1060 static int __init proc_genhd_init(void)
1062 proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
1063 proc_create("partitions", 0, NULL, &proc_partitions_operations);
1066 module_init(proc_genhd_init);
1067 #endif /* CONFIG_PROC_FS */
1069 static void media_change_notify_thread(struct work_struct *work)
1071 struct gendisk *gd = container_of(work, struct gendisk, async_notify);
1072 char event[] = "MEDIA_CHANGE=1";
1073 char *envp[] = { event, NULL };
1076 * set enviroment vars to indicate which event this is for
1077 * so that user space will know to go check the media status.
1079 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1080 put_device(gd->driverfs_dev);
1084 void genhd_media_change_notify(struct gendisk *disk)
1086 get_device(disk->driverfs_dev);
1087 schedule_work(&disk->async_notify);
1089 EXPORT_SYMBOL_GPL(genhd_media_change_notify);
1092 dev_t blk_lookup_devt(const char *name, int partno)
1094 dev_t devt = MKDEV(0, 0);
1095 struct class_dev_iter iter;
1098 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1099 while ((dev = class_dev_iter_next(&iter))) {
1100 struct gendisk *disk = dev_to_disk(dev);
1101 struct hd_struct *part;
1103 if (strcmp(dev_name(dev), name))
1106 if (partno < disk->minors) {
1107 /* We need to return the right devno, even
1108 * if the partition doesn't exist yet.
1110 devt = MKDEV(MAJOR(dev->devt),
1111 MINOR(dev->devt) + partno);
1114 part = disk_get_part(disk, partno);
1116 devt = part_devt(part);
1117 disk_put_part(part);
1120 disk_put_part(part);
1122 class_dev_iter_exit(&iter);
1125 EXPORT_SYMBOL(blk_lookup_devt);
1127 struct gendisk *alloc_disk(int minors)
1129 return alloc_disk_node(minors, -1);
1131 EXPORT_SYMBOL(alloc_disk);
1133 struct gendisk *alloc_disk_node(int minors, int node_id)
1135 struct gendisk *disk;
1137 disk = kmalloc_node(sizeof(struct gendisk),
1138 GFP_KERNEL | __GFP_ZERO, node_id);
1140 if (!init_part_stats(&disk->part0)) {
1144 disk->node_id = node_id;
1145 if (disk_expand_part_tbl(disk, 0)) {
1146 free_part_stats(&disk->part0);
1150 disk->part_tbl->part[0] = &disk->part0;
1152 disk->minors = minors;
1153 rand_initialize_disk(disk);
1154 disk_to_dev(disk)->class = &block_class;
1155 disk_to_dev(disk)->type = &disk_type;
1156 device_initialize(disk_to_dev(disk));
1157 INIT_WORK(&disk->async_notify,
1158 media_change_notify_thread);
1162 EXPORT_SYMBOL(alloc_disk_node);
1164 struct kobject *get_disk(struct gendisk *disk)
1166 struct module *owner;
1167 struct kobject *kobj;
1171 owner = disk->fops->owner;
1172 if (owner && !try_module_get(owner))
1174 kobj = kobject_get(&disk_to_dev(disk)->kobj);
1183 EXPORT_SYMBOL(get_disk);
1185 void put_disk(struct gendisk *disk)
1188 kobject_put(&disk_to_dev(disk)->kobj);
1191 EXPORT_SYMBOL(put_disk);
1193 void set_device_ro(struct block_device *bdev, int flag)
1195 bdev->bd_part->policy = flag;
1198 EXPORT_SYMBOL(set_device_ro);
1200 void set_disk_ro(struct gendisk *disk, int flag)
1202 struct disk_part_iter piter;
1203 struct hd_struct *part;
1205 disk_part_iter_init(&piter, disk,
1206 DISK_PITER_INCL_EMPTY | DISK_PITER_INCL_PART0);
1207 while ((part = disk_part_iter_next(&piter)))
1208 part->policy = flag;
1209 disk_part_iter_exit(&piter);
1212 EXPORT_SYMBOL(set_disk_ro);
1214 int bdev_read_only(struct block_device *bdev)
1218 return bdev->bd_part->policy;
1221 EXPORT_SYMBOL(bdev_read_only);
1223 int invalidate_partition(struct gendisk *disk, int partno)
1226 struct block_device *bdev = bdget_disk(disk, partno);
1229 res = __invalidate_device(bdev);
1235 EXPORT_SYMBOL(invalidate_partition);