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 struct kobject *block_depr;
27 /* for extended dynamic devt allocation, currently only one major is used */
28 #define MAX_EXT_DEVT (1 << MINORBITS)
30 /* For extended devt allocation. ext_devt_mutex prevents look up
31 * results from going away underneath its user.
33 static DEFINE_MUTEX(ext_devt_mutex);
34 static DEFINE_IDR(ext_devt_idr);
36 static struct device_type disk_type;
39 * disk_get_part - get partition
40 * @disk: disk to look partition from
41 * @partno: partition number
43 * Look for partition @partno from @disk. If found, increment
44 * reference count and return it.
50 * Pointer to the found partition on success, NULL if not found.
52 struct hd_struct *disk_get_part(struct gendisk *disk, int partno)
54 struct hd_struct *part = NULL;
55 struct disk_part_tbl *ptbl;
57 if (unlikely(partno < 0))
62 ptbl = rcu_dereference(disk->part_tbl);
63 if (likely(partno < ptbl->len)) {
64 part = rcu_dereference(ptbl->part[partno]);
66 get_device(part_to_dev(part));
73 EXPORT_SYMBOL_GPL(disk_get_part);
76 * disk_part_iter_init - initialize partition iterator
77 * @piter: iterator to initialize
78 * @disk: disk to iterate over
79 * @flags: DISK_PITER_* flags
81 * Initialize @piter so that it iterates over partitions of @disk.
86 void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk,
89 struct disk_part_tbl *ptbl;
92 ptbl = rcu_dereference(disk->part_tbl);
97 if (flags & DISK_PITER_REVERSE)
98 piter->idx = ptbl->len - 1;
99 else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0))
104 piter->flags = flags;
108 EXPORT_SYMBOL_GPL(disk_part_iter_init);
111 * disk_part_iter_next - proceed iterator to the next partition and return it
112 * @piter: iterator of interest
114 * Proceed @piter to the next partition and return it.
119 struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter)
121 struct disk_part_tbl *ptbl;
124 /* put the last partition */
125 disk_put_part(piter->part);
130 ptbl = rcu_dereference(piter->disk->part_tbl);
132 /* determine iteration parameters */
133 if (piter->flags & DISK_PITER_REVERSE) {
135 if (piter->flags & (DISK_PITER_INCL_PART0 |
136 DISK_PITER_INCL_EMPTY_PART0))
145 /* iterate to the next partition */
146 for (; piter->idx != end; piter->idx += inc) {
147 struct hd_struct *part;
149 part = rcu_dereference(ptbl->part[piter->idx]);
152 if (!part->nr_sects &&
153 !(piter->flags & DISK_PITER_INCL_EMPTY) &&
154 !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 &&
158 get_device(part_to_dev(part));
168 EXPORT_SYMBOL_GPL(disk_part_iter_next);
171 * disk_part_iter_exit - finish up partition iteration
172 * @piter: iter of interest
174 * Called when iteration is over. Cleans up @piter.
179 void disk_part_iter_exit(struct disk_part_iter *piter)
181 disk_put_part(piter->part);
184 EXPORT_SYMBOL_GPL(disk_part_iter_exit);
186 static inline int sector_in_part(struct hd_struct *part, sector_t sector)
188 return part->start_sect <= sector &&
189 sector < part->start_sect + part->nr_sects;
193 * disk_map_sector_rcu - map sector to partition
194 * @disk: gendisk of interest
195 * @sector: sector to map
197 * Find out which partition @sector maps to on @disk. This is
198 * primarily used for stats accounting.
201 * RCU read locked. The returned partition pointer is valid only
202 * while preemption is disabled.
205 * Found partition on success, part0 is returned if no partition matches
207 struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector)
209 struct disk_part_tbl *ptbl;
210 struct hd_struct *part;
213 ptbl = rcu_dereference(disk->part_tbl);
215 part = rcu_dereference(ptbl->last_lookup);
216 if (part && sector_in_part(part, sector))
219 for (i = 1; i < ptbl->len; i++) {
220 part = rcu_dereference(ptbl->part[i]);
222 if (part && sector_in_part(part, sector)) {
223 rcu_assign_pointer(ptbl->last_lookup, part);
229 EXPORT_SYMBOL_GPL(disk_map_sector_rcu);
232 * Can be deleted altogether. Later.
235 static struct blk_major_name {
236 struct blk_major_name *next;
239 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
241 /* index in the above - for now: assume no multimajor ranges */
242 static inline int major_to_index(unsigned major)
244 return major % BLKDEV_MAJOR_HASH_SIZE;
247 #ifdef CONFIG_PROC_FS
248 void blkdev_show(struct seq_file *seqf, off_t offset)
250 struct blk_major_name *dp;
252 if (offset < BLKDEV_MAJOR_HASH_SIZE) {
253 mutex_lock(&block_class_lock);
254 for (dp = major_names[offset]; dp; dp = dp->next)
255 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
256 mutex_unlock(&block_class_lock);
259 #endif /* CONFIG_PROC_FS */
262 * register_blkdev - register a new block device
264 * @major: the requested major device number [1..255]. If @major=0, try to
265 * allocate any unused major number.
266 * @name: the name of the new block device as a zero terminated string
268 * The @name must be unique within the system.
270 * The return value depends on the @major input parameter.
271 * - if a major device number was requested in range [1..255] then the
272 * function returns zero on success, or a negative error code
273 * - if any unused major number was requested with @major=0 parameter
274 * then the return value is the allocated major number in range
275 * [1..255] or a negative error code otherwise
277 int register_blkdev(unsigned int major, const char *name)
279 struct blk_major_name **n, *p;
282 mutex_lock(&block_class_lock);
286 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
287 if (major_names[index] == NULL)
292 printk("register_blkdev: failed to get major for %s\n",
301 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
308 strlcpy(p->name, name, sizeof(p->name));
310 index = major_to_index(major);
312 for (n = &major_names[index]; *n; n = &(*n)->next) {
313 if ((*n)->major == major)
322 printk("register_blkdev: cannot get major %d for %s\n",
327 mutex_unlock(&block_class_lock);
331 EXPORT_SYMBOL(register_blkdev);
333 void unregister_blkdev(unsigned int major, const char *name)
335 struct blk_major_name **n;
336 struct blk_major_name *p = NULL;
337 int index = major_to_index(major);
339 mutex_lock(&block_class_lock);
340 for (n = &major_names[index]; *n; n = &(*n)->next)
341 if ((*n)->major == major)
343 if (!*n || strcmp((*n)->name, name)) {
349 mutex_unlock(&block_class_lock);
353 EXPORT_SYMBOL(unregister_blkdev);
355 static struct kobj_map *bdev_map;
358 * blk_mangle_minor - scatter minor numbers apart
359 * @minor: minor number to mangle
361 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
362 * is enabled. Mangling twice gives the original value.
370 static int blk_mangle_minor(int minor)
372 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
375 for (i = 0; i < MINORBITS / 2; i++) {
376 int low = minor & (1 << i);
377 int high = minor & (1 << (MINORBITS - 1 - i));
378 int distance = MINORBITS - 1 - 2 * i;
380 minor ^= low | high; /* clear both bits */
381 low <<= distance; /* swap the positions */
383 minor |= low | high; /* and set */
390 * blk_alloc_devt - allocate a dev_t for a partition
391 * @part: partition to allocate dev_t for
392 * @devt: out parameter for resulting dev_t
394 * Allocate a dev_t for block device.
397 * 0 on success, allocated dev_t is returned in *@devt. -errno on
403 int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
405 struct gendisk *disk = part_to_disk(part);
408 /* in consecutive minor range? */
409 if (part->partno < disk->minors) {
410 *devt = MKDEV(disk->major, disk->first_minor + part->partno);
414 /* allocate ext devt */
416 if (!idr_pre_get(&ext_devt_idr, GFP_KERNEL))
418 rc = idr_get_new(&ext_devt_idr, part, &idx);
419 } while (rc == -EAGAIN);
424 if (idx > MAX_EXT_DEVT) {
425 idr_remove(&ext_devt_idr, idx);
429 *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
434 * blk_free_devt - free a dev_t
435 * @devt: dev_t to free
437 * Free @devt which was allocated using blk_alloc_devt().
442 void blk_free_devt(dev_t devt)
446 if (devt == MKDEV(0, 0))
449 if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
450 mutex_lock(&ext_devt_mutex);
451 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
452 mutex_unlock(&ext_devt_mutex);
456 static char *bdevt_str(dev_t devt, char *buf)
458 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
459 char tbuf[BDEVT_SIZE];
460 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
461 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
463 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
469 * Register device numbers dev..(dev+range-1)
470 * range must be nonzero
471 * The hash chain is sorted on range, so that subranges can override.
473 void blk_register_region(dev_t devt, unsigned long range, struct module *module,
474 struct kobject *(*probe)(dev_t, int *, void *),
475 int (*lock)(dev_t, void *), void *data)
477 kobj_map(bdev_map, devt, range, module, probe, lock, data);
480 EXPORT_SYMBOL(blk_register_region);
482 void blk_unregister_region(dev_t devt, unsigned long range)
484 kobj_unmap(bdev_map, devt, range);
487 EXPORT_SYMBOL(blk_unregister_region);
489 static struct kobject *exact_match(dev_t devt, int *partno, void *data)
491 struct gendisk *p = data;
493 return &disk_to_dev(p)->kobj;
496 static int exact_lock(dev_t devt, void *data)
498 struct gendisk *p = data;
506 * add_disk - add partitioning information to kernel list
507 * @disk: per-device partitioning information
509 * This function registers the partitioning information in @disk
512 * FIXME: error handling
514 void add_disk(struct gendisk *disk)
516 struct backing_dev_info *bdi;
520 /* minors == 0 indicates to use ext devt from part0 and should
521 * be accompanied with EXT_DEVT flag. Make sure all
522 * parameters make sense.
524 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
525 WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT));
527 disk->flags |= GENHD_FL_UP;
529 retval = blk_alloc_devt(&disk->part0, &devt);
534 disk_to_dev(disk)->devt = devt;
536 /* ->major and ->first_minor aren't supposed to be
537 * dereferenced from here on, but set them just in case.
539 disk->major = MAJOR(devt);
540 disk->first_minor = MINOR(devt);
542 /* Register BDI before referencing it from bdev */
543 bdi = &disk->queue->backing_dev_info;
544 bdi_register_dev(bdi, disk_devt(disk));
546 blk_register_region(disk_devt(disk), disk->minors, NULL,
547 exact_match, exact_lock, disk);
549 blk_register_queue(disk);
551 retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
556 EXPORT_SYMBOL(add_disk);
557 EXPORT_SYMBOL(del_gendisk); /* in partitions/check.c */
559 void unlink_gendisk(struct gendisk *disk)
561 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
562 bdi_unregister(&disk->queue->backing_dev_info);
563 blk_unregister_queue(disk);
564 blk_unregister_region(disk_devt(disk), disk->minors);
568 * get_gendisk - get partitioning information for a given device
569 * @devt: device to get partitioning information for
570 * @partno: returned partition index
572 * This function gets the structure containing partitioning
573 * information for the given device @devt.
575 struct gendisk *get_gendisk(dev_t devt, int *partno)
577 struct gendisk *disk = NULL;
579 if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
580 struct kobject *kobj;
582 kobj = kobj_lookup(bdev_map, devt, partno);
584 disk = dev_to_disk(kobj_to_dev(kobj));
586 struct hd_struct *part;
588 mutex_lock(&ext_devt_mutex);
589 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
590 if (part && get_disk(part_to_disk(part))) {
591 *partno = part->partno;
592 disk = part_to_disk(part);
594 mutex_unlock(&ext_devt_mutex);
599 EXPORT_SYMBOL(get_gendisk);
602 * bdget_disk - do bdget() by gendisk and partition number
603 * @disk: gendisk of interest
604 * @partno: partition number
606 * Find partition @partno from @disk, do bdget() on it.
612 * Resulting block_device on success, NULL on failure.
614 struct block_device *bdget_disk(struct gendisk *disk, int partno)
616 struct hd_struct *part;
617 struct block_device *bdev = NULL;
619 part = disk_get_part(disk, partno);
621 bdev = bdget(part_devt(part));
626 EXPORT_SYMBOL(bdget_disk);
629 * print a full list of all partitions - intended for places where the root
630 * filesystem can't be mounted and thus to give the victim some idea of what
633 void __init printk_all_partitions(void)
635 struct class_dev_iter iter;
638 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
639 while ((dev = class_dev_iter_next(&iter))) {
640 struct gendisk *disk = dev_to_disk(dev);
641 struct disk_part_iter piter;
642 struct hd_struct *part;
643 char name_buf[BDEVNAME_SIZE];
644 char devt_buf[BDEVT_SIZE];
645 u8 uuid[PARTITION_META_INFO_UUIDLTH * 2 + 1];
648 * Don't show empty devices or things that have been
651 if (get_capacity(disk) == 0 ||
652 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
656 * Note, unlike /proc/partitions, I am showing the
657 * numbers in hex - the same format as the root=
660 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
661 while ((part = disk_part_iter_next(&piter))) {
662 bool is_part0 = part == &disk->part0;
666 part_unpack_uuid(part->info->uuid, uuid);
668 printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
669 bdevt_str(part_devt(part), devt_buf),
670 (unsigned long long)part->nr_sects >> 1,
671 disk_name(disk, part->partno, name_buf), uuid);
673 if (disk->driverfs_dev != NULL &&
674 disk->driverfs_dev->driver != NULL)
675 printk(" driver: %s\n",
676 disk->driverfs_dev->driver->name);
678 printk(" (driver?)\n");
682 disk_part_iter_exit(&piter);
684 class_dev_iter_exit(&iter);
687 #ifdef CONFIG_PROC_FS
689 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
692 struct class_dev_iter *iter;
695 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
697 return ERR_PTR(-ENOMEM);
699 seqf->private = iter;
700 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
702 dev = class_dev_iter_next(iter);
707 return dev_to_disk(dev);
710 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
715 dev = class_dev_iter_next(seqf->private);
717 return dev_to_disk(dev);
722 static void disk_seqf_stop(struct seq_file *seqf, void *v)
724 struct class_dev_iter *iter = seqf->private;
726 /* stop is called even after start failed :-( */
728 class_dev_iter_exit(iter);
733 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
737 p = disk_seqf_start(seqf, pos);
738 if (!IS_ERR_OR_NULL(p) && !*pos)
739 seq_puts(seqf, "major minor #blocks name\n\n");
743 static int show_partition(struct seq_file *seqf, void *v)
745 struct gendisk *sgp = v;
746 struct disk_part_iter piter;
747 struct hd_struct *part;
748 char buf[BDEVNAME_SIZE];
750 /* Don't show non-partitionable removeable devices or empty devices */
751 if (!get_capacity(sgp) || (!disk_partitionable(sgp) &&
752 (sgp->flags & GENHD_FL_REMOVABLE)))
754 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
757 /* show the full disk and all non-0 size partitions of it */
758 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
759 while ((part = disk_part_iter_next(&piter)))
760 seq_printf(seqf, "%4d %7d %10llu %s\n",
761 MAJOR(part_devt(part)), MINOR(part_devt(part)),
762 (unsigned long long)part->nr_sects >> 1,
763 disk_name(sgp, part->partno, buf));
764 disk_part_iter_exit(&piter);
769 static const struct seq_operations partitions_op = {
770 .start = show_partition_start,
771 .next = disk_seqf_next,
772 .stop = disk_seqf_stop,
773 .show = show_partition
776 static int partitions_open(struct inode *inode, struct file *file)
778 return seq_open(file, &partitions_op);
781 static const struct file_operations proc_partitions_operations = {
782 .open = partitions_open,
785 .release = seq_release,
790 static struct kobject *base_probe(dev_t devt, int *partno, void *data)
792 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
793 /* Make old-style 2.4 aliases work */
794 request_module("block-major-%d", MAJOR(devt));
798 static int __init genhd_device_init(void)
802 block_class.dev_kobj = sysfs_dev_block_kobj;
803 error = class_register(&block_class);
806 bdev_map = kobj_map_init(base_probe, &block_class_lock);
809 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
811 /* create top-level block dir */
812 if (!sysfs_deprecated)
813 block_depr = kobject_create_and_add("block", NULL);
817 subsys_initcall(genhd_device_init);
819 static ssize_t disk_range_show(struct device *dev,
820 struct device_attribute *attr, char *buf)
822 struct gendisk *disk = dev_to_disk(dev);
824 return sprintf(buf, "%d\n", disk->minors);
827 static ssize_t disk_ext_range_show(struct device *dev,
828 struct device_attribute *attr, char *buf)
830 struct gendisk *disk = dev_to_disk(dev);
832 return sprintf(buf, "%d\n", disk_max_parts(disk));
835 static ssize_t disk_removable_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",
841 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
844 static ssize_t disk_ro_show(struct device *dev,
845 struct device_attribute *attr, char *buf)
847 struct gendisk *disk = dev_to_disk(dev);
849 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
852 static ssize_t disk_capability_show(struct device *dev,
853 struct device_attribute *attr, char *buf)
855 struct gendisk *disk = dev_to_disk(dev);
857 return sprintf(buf, "%x\n", disk->flags);
860 static ssize_t disk_alignment_offset_show(struct device *dev,
861 struct device_attribute *attr,
864 struct gendisk *disk = dev_to_disk(dev);
866 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
869 static ssize_t disk_discard_alignment_show(struct device *dev,
870 struct device_attribute *attr,
873 struct gendisk *disk = dev_to_disk(dev);
875 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
878 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
879 static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
880 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
881 static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
882 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
883 static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL);
884 static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show,
886 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
887 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
888 static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
889 #ifdef CONFIG_FAIL_MAKE_REQUEST
890 static struct device_attribute dev_attr_fail =
891 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
893 #ifdef CONFIG_FAIL_IO_TIMEOUT
894 static struct device_attribute dev_attr_fail_timeout =
895 __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show,
899 static struct attribute *disk_attrs[] = {
900 &dev_attr_range.attr,
901 &dev_attr_ext_range.attr,
902 &dev_attr_removable.attr,
905 &dev_attr_alignment_offset.attr,
906 &dev_attr_discard_alignment.attr,
907 &dev_attr_capability.attr,
909 &dev_attr_inflight.attr,
910 #ifdef CONFIG_FAIL_MAKE_REQUEST
913 #ifdef CONFIG_FAIL_IO_TIMEOUT
914 &dev_attr_fail_timeout.attr,
919 static struct attribute_group disk_attr_group = {
923 static const struct attribute_group *disk_attr_groups[] = {
928 static void disk_free_ptbl_rcu_cb(struct rcu_head *head)
930 struct disk_part_tbl *ptbl =
931 container_of(head, struct disk_part_tbl, rcu_head);
937 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
938 * @disk: disk to replace part_tbl for
939 * @new_ptbl: new part_tbl to install
941 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
942 * original ptbl is freed using RCU callback.
945 * Matching bd_mutx locked.
947 static void disk_replace_part_tbl(struct gendisk *disk,
948 struct disk_part_tbl *new_ptbl)
950 struct disk_part_tbl *old_ptbl = disk->part_tbl;
952 rcu_assign_pointer(disk->part_tbl, new_ptbl);
955 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
956 call_rcu(&old_ptbl->rcu_head, disk_free_ptbl_rcu_cb);
961 * disk_expand_part_tbl - expand disk->part_tbl
962 * @disk: disk to expand part_tbl for
963 * @partno: expand such that this partno can fit in
965 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
966 * uses RCU to allow unlocked dereferencing for stats and other stuff.
969 * Matching bd_mutex locked, might sleep.
972 * 0 on success, -errno on failure.
974 int disk_expand_part_tbl(struct gendisk *disk, int partno)
976 struct disk_part_tbl *old_ptbl = disk->part_tbl;
977 struct disk_part_tbl *new_ptbl;
978 int len = old_ptbl ? old_ptbl->len : 0;
979 int target = partno + 1;
983 /* disk_max_parts() is zero during initialization, ignore if so */
984 if (disk_max_parts(disk) && target > disk_max_parts(disk))
990 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
991 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
995 new_ptbl->len = target;
997 for (i = 0; i < len; i++)
998 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
1000 disk_replace_part_tbl(disk, new_ptbl);
1004 static void disk_release(struct device *dev)
1006 struct gendisk *disk = dev_to_disk(dev);
1008 kfree(disk->random);
1009 disk_replace_part_tbl(disk, NULL);
1010 free_part_stats(&disk->part0);
1011 free_part_info(&disk->part0);
1014 struct class block_class = {
1018 static char *block_devnode(struct device *dev, mode_t *mode)
1020 struct gendisk *disk = dev_to_disk(dev);
1023 return disk->devnode(disk, mode);
1027 static struct device_type disk_type = {
1029 .groups = disk_attr_groups,
1030 .release = disk_release,
1031 .devnode = block_devnode,
1034 #ifdef CONFIG_PROC_FS
1036 * aggregate disk stat collector. Uses the same stats that the sysfs
1037 * entries do, above, but makes them available through one seq_file.
1039 * The output looks suspiciously like /proc/partitions with a bunch of
1042 static int diskstats_show(struct seq_file *seqf, void *v)
1044 struct gendisk *gp = v;
1045 struct disk_part_iter piter;
1046 struct hd_struct *hd;
1047 char buf[BDEVNAME_SIZE];
1051 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1052 seq_puts(seqf, "major minor name"
1053 " rio rmerge rsect ruse wio wmerge "
1054 "wsect wuse running use aveq"
1058 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
1059 while ((hd = disk_part_iter_next(&piter))) {
1060 cpu = part_stat_lock();
1061 part_round_stats(cpu, hd);
1063 seq_printf(seqf, "%4d %7d %s %lu %lu %llu "
1064 "%u %lu %lu %llu %u %u %u %u\n",
1065 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1066 disk_name(gp, hd->partno, buf),
1067 part_stat_read(hd, ios[0]),
1068 part_stat_read(hd, merges[0]),
1069 (unsigned long long)part_stat_read(hd, sectors[0]),
1070 jiffies_to_msecs(part_stat_read(hd, ticks[0])),
1071 part_stat_read(hd, ios[1]),
1072 part_stat_read(hd, merges[1]),
1073 (unsigned long long)part_stat_read(hd, sectors[1]),
1074 jiffies_to_msecs(part_stat_read(hd, ticks[1])),
1076 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1077 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1080 disk_part_iter_exit(&piter);
1085 static const struct seq_operations diskstats_op = {
1086 .start = disk_seqf_start,
1087 .next = disk_seqf_next,
1088 .stop = disk_seqf_stop,
1089 .show = diskstats_show
1092 static int diskstats_open(struct inode *inode, struct file *file)
1094 return seq_open(file, &diskstats_op);
1097 static const struct file_operations proc_diskstats_operations = {
1098 .open = diskstats_open,
1100 .llseek = seq_lseek,
1101 .release = seq_release,
1104 static int __init proc_genhd_init(void)
1106 proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
1107 proc_create("partitions", 0, NULL, &proc_partitions_operations);
1110 module_init(proc_genhd_init);
1111 #endif /* CONFIG_PROC_FS */
1113 static void media_change_notify_thread(struct work_struct *work)
1115 struct gendisk *gd = container_of(work, struct gendisk, async_notify);
1116 char event[] = "MEDIA_CHANGE=1";
1117 char *envp[] = { event, NULL };
1120 * set enviroment vars to indicate which event this is for
1121 * so that user space will know to go check the media status.
1123 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1124 put_device(gd->driverfs_dev);
1128 void genhd_media_change_notify(struct gendisk *disk)
1130 get_device(disk->driverfs_dev);
1131 schedule_work(&disk->async_notify);
1133 EXPORT_SYMBOL_GPL(genhd_media_change_notify);
1136 dev_t blk_lookup_devt(const char *name, int partno)
1138 dev_t devt = MKDEV(0, 0);
1139 struct class_dev_iter iter;
1142 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1143 while ((dev = class_dev_iter_next(&iter))) {
1144 struct gendisk *disk = dev_to_disk(dev);
1145 struct hd_struct *part;
1147 if (strcmp(dev_name(dev), name))
1150 if (partno < disk->minors) {
1151 /* We need to return the right devno, even
1152 * if the partition doesn't exist yet.
1154 devt = MKDEV(MAJOR(dev->devt),
1155 MINOR(dev->devt) + partno);
1158 part = disk_get_part(disk, partno);
1160 devt = part_devt(part);
1161 disk_put_part(part);
1164 disk_put_part(part);
1166 class_dev_iter_exit(&iter);
1169 EXPORT_SYMBOL(blk_lookup_devt);
1171 struct gendisk *alloc_disk(int minors)
1173 return alloc_disk_node(minors, -1);
1175 EXPORT_SYMBOL(alloc_disk);
1177 struct gendisk *alloc_disk_node(int minors, int node_id)
1179 struct gendisk *disk;
1181 disk = kmalloc_node(sizeof(struct gendisk),
1182 GFP_KERNEL | __GFP_ZERO, node_id);
1184 if (!init_part_stats(&disk->part0)) {
1188 disk->node_id = node_id;
1189 if (disk_expand_part_tbl(disk, 0)) {
1190 free_part_stats(&disk->part0);
1194 disk->part_tbl->part[0] = &disk->part0;
1196 hd_ref_init(&disk->part0);
1198 disk->minors = minors;
1199 rand_initialize_disk(disk);
1200 disk_to_dev(disk)->class = &block_class;
1201 disk_to_dev(disk)->type = &disk_type;
1202 device_initialize(disk_to_dev(disk));
1203 INIT_WORK(&disk->async_notify,
1204 media_change_notify_thread);
1208 EXPORT_SYMBOL(alloc_disk_node);
1210 struct kobject *get_disk(struct gendisk *disk)
1212 struct module *owner;
1213 struct kobject *kobj;
1217 owner = disk->fops->owner;
1218 if (owner && !try_module_get(owner))
1220 kobj = kobject_get(&disk_to_dev(disk)->kobj);
1229 EXPORT_SYMBOL(get_disk);
1231 void put_disk(struct gendisk *disk)
1234 kobject_put(&disk_to_dev(disk)->kobj);
1237 EXPORT_SYMBOL(put_disk);
1239 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1241 char event[] = "DISK_RO=1";
1242 char *envp[] = { event, NULL };
1246 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1249 void set_device_ro(struct block_device *bdev, int flag)
1251 bdev->bd_part->policy = flag;
1254 EXPORT_SYMBOL(set_device_ro);
1256 void set_disk_ro(struct gendisk *disk, int flag)
1258 struct disk_part_iter piter;
1259 struct hd_struct *part;
1261 if (disk->part0.policy != flag) {
1262 set_disk_ro_uevent(disk, flag);
1263 disk->part0.policy = flag;
1266 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
1267 while ((part = disk_part_iter_next(&piter)))
1268 part->policy = flag;
1269 disk_part_iter_exit(&piter);
1272 EXPORT_SYMBOL(set_disk_ro);
1274 int bdev_read_only(struct block_device *bdev)
1278 return bdev->bd_part->policy;
1281 EXPORT_SYMBOL(bdev_read_only);
1283 int invalidate_partition(struct gendisk *disk, int partno)
1286 struct block_device *bdev = bdget_disk(disk, partno);
1289 res = __invalidate_device(bdev);
1295 EXPORT_SYMBOL(invalidate_partition);