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 | DISK_PITER_INCL_EMPTY_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 |
138 DISK_PITER_INCL_EMPTY_PART0))
147 /* iterate to the next partition */
148 for (; piter->idx != end; piter->idx += inc) {
149 struct hd_struct *part;
151 part = rcu_dereference(ptbl->part[piter->idx]);
154 if (!part->nr_sects &&
155 !(piter->flags & DISK_PITER_INCL_EMPTY) &&
156 !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 &&
160 get_device(part_to_dev(part));
170 EXPORT_SYMBOL_GPL(disk_part_iter_next);
173 * disk_part_iter_exit - finish up partition iteration
174 * @piter: iter of interest
176 * Called when iteration is over. Cleans up @piter.
181 void disk_part_iter_exit(struct disk_part_iter *piter)
183 disk_put_part(piter->part);
186 EXPORT_SYMBOL_GPL(disk_part_iter_exit);
188 static inline int sector_in_part(struct hd_struct *part, sector_t sector)
190 return part->start_sect <= sector &&
191 sector < part->start_sect + part->nr_sects;
195 * disk_map_sector_rcu - map sector to partition
196 * @disk: gendisk of interest
197 * @sector: sector to map
199 * Find out which partition @sector maps to on @disk. This is
200 * primarily used for stats accounting.
203 * RCU read locked. The returned partition pointer is valid only
204 * while preemption is disabled.
207 * Found partition on success, part0 is returned if no partition matches
209 struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector)
211 struct disk_part_tbl *ptbl;
212 struct hd_struct *part;
215 ptbl = rcu_dereference(disk->part_tbl);
217 part = rcu_dereference(ptbl->last_lookup);
218 if (part && sector_in_part(part, sector))
221 for (i = 1; i < ptbl->len; i++) {
222 part = rcu_dereference(ptbl->part[i]);
224 if (part && sector_in_part(part, sector)) {
225 rcu_assign_pointer(ptbl->last_lookup, part);
231 EXPORT_SYMBOL_GPL(disk_map_sector_rcu);
234 * Can be deleted altogether. Later.
237 static struct blk_major_name {
238 struct blk_major_name *next;
241 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
243 /* index in the above - for now: assume no multimajor ranges */
244 static inline int major_to_index(int major)
246 return major % BLKDEV_MAJOR_HASH_SIZE;
249 #ifdef CONFIG_PROC_FS
250 void blkdev_show(struct seq_file *seqf, off_t offset)
252 struct blk_major_name *dp;
254 if (offset < BLKDEV_MAJOR_HASH_SIZE) {
255 mutex_lock(&block_class_lock);
256 for (dp = major_names[offset]; dp; dp = dp->next)
257 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
258 mutex_unlock(&block_class_lock);
261 #endif /* CONFIG_PROC_FS */
264 * register_blkdev - register a new block device
266 * @major: the requested major device number [1..255]. If @major=0, try to
267 * allocate any unused major number.
268 * @name: the name of the new block device as a zero terminated string
270 * The @name must be unique within the system.
272 * The return value depends on the @major input parameter.
273 * - if a major device number was requested in range [1..255] then the
274 * function returns zero on success, or a negative error code
275 * - if any unused major number was requested with @major=0 parameter
276 * then the return value is the allocated major number in range
277 * [1..255] or a negative error code otherwise
279 int register_blkdev(unsigned int major, const char *name)
281 struct blk_major_name **n, *p;
284 mutex_lock(&block_class_lock);
288 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
289 if (major_names[index] == NULL)
294 printk("register_blkdev: failed to get major for %s\n",
303 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
310 strlcpy(p->name, name, sizeof(p->name));
312 index = major_to_index(major);
314 for (n = &major_names[index]; *n; n = &(*n)->next) {
315 if ((*n)->major == major)
324 printk("register_blkdev: cannot get major %d for %s\n",
329 mutex_unlock(&block_class_lock);
333 EXPORT_SYMBOL(register_blkdev);
335 void unregister_blkdev(unsigned int major, const char *name)
337 struct blk_major_name **n;
338 struct blk_major_name *p = NULL;
339 int index = major_to_index(major);
341 mutex_lock(&block_class_lock);
342 for (n = &major_names[index]; *n; n = &(*n)->next)
343 if ((*n)->major == major)
345 if (!*n || strcmp((*n)->name, name)) {
351 mutex_unlock(&block_class_lock);
355 EXPORT_SYMBOL(unregister_blkdev);
357 static struct kobj_map *bdev_map;
360 * blk_mangle_minor - scatter minor numbers apart
361 * @minor: minor number to mangle
363 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
364 * is enabled. Mangling twice gives the original value.
372 static int blk_mangle_minor(int minor)
374 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
377 for (i = 0; i < MINORBITS / 2; i++) {
378 int low = minor & (1 << i);
379 int high = minor & (1 << (MINORBITS - 1 - i));
380 int distance = MINORBITS - 1 - 2 * i;
382 minor ^= low | high; /* clear both bits */
383 low <<= distance; /* swap the positions */
385 minor |= low | high; /* and set */
392 * blk_alloc_devt - allocate a dev_t for a partition
393 * @part: partition to allocate dev_t for
394 * @devt: out parameter for resulting dev_t
396 * Allocate a dev_t for block device.
399 * 0 on success, allocated dev_t is returned in *@devt. -errno on
405 int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
407 struct gendisk *disk = part_to_disk(part);
410 /* in consecutive minor range? */
411 if (part->partno < disk->minors) {
412 *devt = MKDEV(disk->major, disk->first_minor + part->partno);
416 /* allocate ext devt */
418 if (!idr_pre_get(&ext_devt_idr, GFP_KERNEL))
420 rc = idr_get_new(&ext_devt_idr, part, &idx);
421 } while (rc == -EAGAIN);
426 if (idx > MAX_EXT_DEVT) {
427 idr_remove(&ext_devt_idr, idx);
431 *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
436 * blk_free_devt - free a dev_t
437 * @devt: dev_t to free
439 * Free @devt which was allocated using blk_alloc_devt().
444 void blk_free_devt(dev_t devt)
448 if (devt == MKDEV(0, 0))
451 if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
452 mutex_lock(&ext_devt_mutex);
453 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
454 mutex_unlock(&ext_devt_mutex);
458 static char *bdevt_str(dev_t devt, char *buf)
460 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
461 char tbuf[BDEVT_SIZE];
462 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
463 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
465 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
471 * Register device numbers dev..(dev+range-1)
472 * range must be nonzero
473 * The hash chain is sorted on range, so that subranges can override.
475 void blk_register_region(dev_t devt, unsigned long range, struct module *module,
476 struct kobject *(*probe)(dev_t, int *, void *),
477 int (*lock)(dev_t, void *), void *data)
479 kobj_map(bdev_map, devt, range, module, probe, lock, data);
482 EXPORT_SYMBOL(blk_register_region);
484 void blk_unregister_region(dev_t devt, unsigned long range)
486 kobj_unmap(bdev_map, devt, range);
489 EXPORT_SYMBOL(blk_unregister_region);
491 static struct kobject *exact_match(dev_t devt, int *partno, void *data)
493 struct gendisk *p = data;
495 return &disk_to_dev(p)->kobj;
498 static int exact_lock(dev_t devt, void *data)
500 struct gendisk *p = data;
508 * add_disk - add partitioning information to kernel list
509 * @disk: per-device partitioning information
511 * This function registers the partitioning information in @disk
514 * FIXME: error handling
516 void add_disk(struct gendisk *disk)
518 struct backing_dev_info *bdi;
522 /* minors == 0 indicates to use ext devt from part0 and should
523 * be accompanied with EXT_DEVT flag. Make sure all
524 * parameters make sense.
526 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
527 WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT));
529 disk->flags |= GENHD_FL_UP;
531 retval = blk_alloc_devt(&disk->part0, &devt);
536 disk_to_dev(disk)->devt = devt;
538 /* ->major and ->first_minor aren't supposed to be
539 * dereferenced from here on, but set them just in case.
541 disk->major = MAJOR(devt);
542 disk->first_minor = MINOR(devt);
544 blk_register_region(disk_devt(disk), disk->minors, NULL,
545 exact_match, exact_lock, disk);
547 blk_register_queue(disk);
549 bdi = &disk->queue->backing_dev_info;
550 bdi_register_dev(bdi, disk_devt(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);
601 * bdget_disk - do bdget() by gendisk and partition number
602 * @disk: gendisk of interest
603 * @partno: partition number
605 * Find partition @partno from @disk, do bdget() on it.
611 * Resulting block_device on success, NULL on failure.
613 struct block_device *bdget_disk(struct gendisk *disk, int partno)
615 struct hd_struct *part;
616 struct block_device *bdev = NULL;
618 part = disk_get_part(disk, partno);
620 bdev = bdget(part_devt(part));
625 EXPORT_SYMBOL(bdget_disk);
628 * print a full list of all partitions - intended for places where the root
629 * filesystem can't be mounted and thus to give the victim some idea of what
632 void __init printk_all_partitions(void)
634 struct class_dev_iter iter;
637 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
638 while ((dev = class_dev_iter_next(&iter))) {
639 struct gendisk *disk = dev_to_disk(dev);
640 struct disk_part_iter piter;
641 struct hd_struct *part;
642 char name_buf[BDEVNAME_SIZE];
643 char devt_buf[BDEVT_SIZE];
646 * Don't show empty devices or things that have been
649 if (get_capacity(disk) == 0 ||
650 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
654 * Note, unlike /proc/partitions, I am showing the
655 * numbers in hex - the same format as the root=
658 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
659 while ((part = disk_part_iter_next(&piter))) {
660 bool is_part0 = part == &disk->part0;
662 printk("%s%s %10llu %s", is_part0 ? "" : " ",
663 bdevt_str(part_devt(part), devt_buf),
664 (unsigned long long)part->nr_sects >> 1,
665 disk_name(disk, part->partno, name_buf));
667 if (disk->driverfs_dev != NULL &&
668 disk->driverfs_dev->driver != NULL)
669 printk(" driver: %s\n",
670 disk->driverfs_dev->driver->name);
672 printk(" (driver?)\n");
676 disk_part_iter_exit(&piter);
678 class_dev_iter_exit(&iter);
681 #ifdef CONFIG_PROC_FS
683 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
686 struct class_dev_iter *iter;
689 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
691 return ERR_PTR(-ENOMEM);
693 seqf->private = iter;
694 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
696 dev = class_dev_iter_next(iter);
701 return dev_to_disk(dev);
704 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
709 dev = class_dev_iter_next(seqf->private);
711 return dev_to_disk(dev);
716 static void disk_seqf_stop(struct seq_file *seqf, void *v)
718 struct class_dev_iter *iter = seqf->private;
720 /* stop is called even after start failed :-( */
722 class_dev_iter_exit(iter);
727 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
731 p = disk_seqf_start(seqf, pos);
732 if (!IS_ERR(p) && p && !*pos)
733 seq_puts(seqf, "major minor #blocks name\n\n");
737 static int show_partition(struct seq_file *seqf, void *v)
739 struct gendisk *sgp = v;
740 struct disk_part_iter piter;
741 struct hd_struct *part;
742 char buf[BDEVNAME_SIZE];
744 /* Don't show non-partitionable removeable devices or empty devices */
745 if (!get_capacity(sgp) || (!disk_partitionable(sgp) &&
746 (sgp->flags & GENHD_FL_REMOVABLE)))
748 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
751 /* show the full disk and all non-0 size partitions of it */
752 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
753 while ((part = disk_part_iter_next(&piter)))
754 seq_printf(seqf, "%4d %7d %10llu %s\n",
755 MAJOR(part_devt(part)), MINOR(part_devt(part)),
756 (unsigned long long)part->nr_sects >> 1,
757 disk_name(sgp, part->partno, buf));
758 disk_part_iter_exit(&piter);
763 static const struct seq_operations partitions_op = {
764 .start = show_partition_start,
765 .next = disk_seqf_next,
766 .stop = disk_seqf_stop,
767 .show = show_partition
770 static int partitions_open(struct inode *inode, struct file *file)
772 return seq_open(file, &partitions_op);
775 static const struct file_operations proc_partitions_operations = {
776 .open = partitions_open,
779 .release = seq_release,
784 static struct kobject *base_probe(dev_t devt, int *partno, void *data)
786 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
787 /* Make old-style 2.4 aliases work */
788 request_module("block-major-%d", MAJOR(devt));
792 static int __init genhd_device_init(void)
796 block_class.dev_kobj = sysfs_dev_block_kobj;
797 error = class_register(&block_class);
800 bdev_map = kobj_map_init(base_probe, &block_class_lock);
803 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
805 #ifndef CONFIG_SYSFS_DEPRECATED
806 /* create top-level block dir */
807 block_depr = kobject_create_and_add("block", NULL);
812 subsys_initcall(genhd_device_init);
814 static ssize_t disk_range_show(struct device *dev,
815 struct device_attribute *attr, char *buf)
817 struct gendisk *disk = dev_to_disk(dev);
819 return sprintf(buf, "%d\n", disk->minors);
822 static ssize_t disk_ext_range_show(struct device *dev,
823 struct device_attribute *attr, char *buf)
825 struct gendisk *disk = dev_to_disk(dev);
827 return sprintf(buf, "%d\n", disk_max_parts(disk));
830 static ssize_t disk_removable_show(struct device *dev,
831 struct device_attribute *attr, char *buf)
833 struct gendisk *disk = dev_to_disk(dev);
835 return sprintf(buf, "%d\n",
836 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
839 static ssize_t disk_ro_show(struct device *dev,
840 struct device_attribute *attr, char *buf)
842 struct gendisk *disk = dev_to_disk(dev);
844 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
847 static ssize_t disk_capability_show(struct device *dev,
848 struct device_attribute *attr, char *buf)
850 struct gendisk *disk = dev_to_disk(dev);
852 return sprintf(buf, "%x\n", disk->flags);
855 static ssize_t disk_alignment_offset_show(struct device *dev,
856 struct device_attribute *attr,
859 struct gendisk *disk = dev_to_disk(dev);
861 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
864 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
865 static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
866 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
867 static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
868 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
869 static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL);
870 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
871 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
872 static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
873 #ifdef CONFIG_FAIL_MAKE_REQUEST
874 static struct device_attribute dev_attr_fail =
875 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
877 #ifdef CONFIG_FAIL_IO_TIMEOUT
878 static struct device_attribute dev_attr_fail_timeout =
879 __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show,
883 static struct attribute *disk_attrs[] = {
884 &dev_attr_range.attr,
885 &dev_attr_ext_range.attr,
886 &dev_attr_removable.attr,
889 &dev_attr_alignment_offset.attr,
890 &dev_attr_capability.attr,
892 &dev_attr_inflight.attr,
893 #ifdef CONFIG_FAIL_MAKE_REQUEST
896 #ifdef CONFIG_FAIL_IO_TIMEOUT
897 &dev_attr_fail_timeout.attr,
902 static struct attribute_group disk_attr_group = {
906 static const struct attribute_group *disk_attr_groups[] = {
911 static void disk_free_ptbl_rcu_cb(struct rcu_head *head)
913 struct disk_part_tbl *ptbl =
914 container_of(head, struct disk_part_tbl, rcu_head);
920 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
921 * @disk: disk to replace part_tbl for
922 * @new_ptbl: new part_tbl to install
924 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
925 * original ptbl is freed using RCU callback.
928 * Matching bd_mutx locked.
930 static void disk_replace_part_tbl(struct gendisk *disk,
931 struct disk_part_tbl *new_ptbl)
933 struct disk_part_tbl *old_ptbl = disk->part_tbl;
935 rcu_assign_pointer(disk->part_tbl, new_ptbl);
938 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
939 call_rcu(&old_ptbl->rcu_head, disk_free_ptbl_rcu_cb);
944 * disk_expand_part_tbl - expand disk->part_tbl
945 * @disk: disk to expand part_tbl for
946 * @partno: expand such that this partno can fit in
948 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
949 * uses RCU to allow unlocked dereferencing for stats and other stuff.
952 * Matching bd_mutex locked, might sleep.
955 * 0 on success, -errno on failure.
957 int disk_expand_part_tbl(struct gendisk *disk, int partno)
959 struct disk_part_tbl *old_ptbl = disk->part_tbl;
960 struct disk_part_tbl *new_ptbl;
961 int len = old_ptbl ? old_ptbl->len : 0;
962 int target = partno + 1;
966 /* disk_max_parts() is zero during initialization, ignore if so */
967 if (disk_max_parts(disk) && target > disk_max_parts(disk))
973 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
974 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
978 INIT_RCU_HEAD(&new_ptbl->rcu_head);
979 new_ptbl->len = target;
981 for (i = 0; i < len; i++)
982 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
984 disk_replace_part_tbl(disk, new_ptbl);
988 static void disk_release(struct device *dev)
990 struct gendisk *disk = dev_to_disk(dev);
993 disk_replace_part_tbl(disk, NULL);
994 free_part_stats(&disk->part0);
997 struct class block_class = {
1001 static char *block_devnode(struct device *dev, mode_t *mode)
1003 struct gendisk *disk = dev_to_disk(dev);
1006 return disk->devnode(disk, mode);
1010 static struct device_type disk_type = {
1012 .groups = disk_attr_groups,
1013 .release = disk_release,
1014 .devnode = block_devnode,
1017 #ifdef CONFIG_PROC_FS
1019 * aggregate disk stat collector. Uses the same stats that the sysfs
1020 * entries do, above, but makes them available through one seq_file.
1022 * The output looks suspiciously like /proc/partitions with a bunch of
1025 static int diskstats_show(struct seq_file *seqf, void *v)
1027 struct gendisk *gp = v;
1028 struct disk_part_iter piter;
1029 struct hd_struct *hd;
1030 char buf[BDEVNAME_SIZE];
1034 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1035 seq_puts(seqf, "major minor name"
1036 " rio rmerge rsect ruse wio wmerge "
1037 "wsect wuse running use aveq"
1041 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
1042 while ((hd = disk_part_iter_next(&piter))) {
1043 cpu = part_stat_lock();
1044 part_round_stats(cpu, hd);
1046 seq_printf(seqf, "%4d %7d %s %lu %lu %llu "
1047 "%u %lu %lu %llu %u %u %u %u\n",
1048 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1049 disk_name(gp, hd->partno, buf),
1050 part_stat_read(hd, ios[0]),
1051 part_stat_read(hd, merges[0]),
1052 (unsigned long long)part_stat_read(hd, sectors[0]),
1053 jiffies_to_msecs(part_stat_read(hd, ticks[0])),
1054 part_stat_read(hd, ios[1]),
1055 part_stat_read(hd, merges[1]),
1056 (unsigned long long)part_stat_read(hd, sectors[1]),
1057 jiffies_to_msecs(part_stat_read(hd, ticks[1])),
1059 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1060 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1063 disk_part_iter_exit(&piter);
1068 static const struct seq_operations diskstats_op = {
1069 .start = disk_seqf_start,
1070 .next = disk_seqf_next,
1071 .stop = disk_seqf_stop,
1072 .show = diskstats_show
1075 static int diskstats_open(struct inode *inode, struct file *file)
1077 return seq_open(file, &diskstats_op);
1080 static const struct file_operations proc_diskstats_operations = {
1081 .open = diskstats_open,
1083 .llseek = seq_lseek,
1084 .release = seq_release,
1087 static int __init proc_genhd_init(void)
1089 proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
1090 proc_create("partitions", 0, NULL, &proc_partitions_operations);
1093 module_init(proc_genhd_init);
1094 #endif /* CONFIG_PROC_FS */
1096 static void media_change_notify_thread(struct work_struct *work)
1098 struct gendisk *gd = container_of(work, struct gendisk, async_notify);
1099 char event[] = "MEDIA_CHANGE=1";
1100 char *envp[] = { event, NULL };
1103 * set enviroment vars to indicate which event this is for
1104 * so that user space will know to go check the media status.
1106 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1107 put_device(gd->driverfs_dev);
1111 void genhd_media_change_notify(struct gendisk *disk)
1113 get_device(disk->driverfs_dev);
1114 schedule_work(&disk->async_notify);
1116 EXPORT_SYMBOL_GPL(genhd_media_change_notify);
1119 dev_t blk_lookup_devt(const char *name, int partno)
1121 dev_t devt = MKDEV(0, 0);
1122 struct class_dev_iter iter;
1125 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1126 while ((dev = class_dev_iter_next(&iter))) {
1127 struct gendisk *disk = dev_to_disk(dev);
1128 struct hd_struct *part;
1130 if (strcmp(dev_name(dev), name))
1133 if (partno < disk->minors) {
1134 /* We need to return the right devno, even
1135 * if the partition doesn't exist yet.
1137 devt = MKDEV(MAJOR(dev->devt),
1138 MINOR(dev->devt) + partno);
1141 part = disk_get_part(disk, partno);
1143 devt = part_devt(part);
1144 disk_put_part(part);
1147 disk_put_part(part);
1149 class_dev_iter_exit(&iter);
1152 EXPORT_SYMBOL(blk_lookup_devt);
1154 struct gendisk *alloc_disk(int minors)
1156 return alloc_disk_node(minors, -1);
1158 EXPORT_SYMBOL(alloc_disk);
1160 struct gendisk *alloc_disk_node(int minors, int node_id)
1162 struct gendisk *disk;
1164 disk = kmalloc_node(sizeof(struct gendisk),
1165 GFP_KERNEL | __GFP_ZERO, node_id);
1167 if (!init_part_stats(&disk->part0)) {
1171 disk->node_id = node_id;
1172 if (disk_expand_part_tbl(disk, 0)) {
1173 free_part_stats(&disk->part0);
1177 disk->part_tbl->part[0] = &disk->part0;
1179 disk->minors = minors;
1180 rand_initialize_disk(disk);
1181 disk_to_dev(disk)->class = &block_class;
1182 disk_to_dev(disk)->type = &disk_type;
1183 device_initialize(disk_to_dev(disk));
1184 INIT_WORK(&disk->async_notify,
1185 media_change_notify_thread);
1189 EXPORT_SYMBOL(alloc_disk_node);
1191 struct kobject *get_disk(struct gendisk *disk)
1193 struct module *owner;
1194 struct kobject *kobj;
1198 owner = disk->fops->owner;
1199 if (owner && !try_module_get(owner))
1201 kobj = kobject_get(&disk_to_dev(disk)->kobj);
1210 EXPORT_SYMBOL(get_disk);
1212 void put_disk(struct gendisk *disk)
1215 kobject_put(&disk_to_dev(disk)->kobj);
1218 EXPORT_SYMBOL(put_disk);
1220 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1222 char event[] = "DISK_RO=1";
1223 char *envp[] = { event, NULL };
1227 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1230 void set_device_ro(struct block_device *bdev, int flag)
1232 bdev->bd_part->policy = flag;
1235 EXPORT_SYMBOL(set_device_ro);
1237 void set_disk_ro(struct gendisk *disk, int flag)
1239 struct disk_part_iter piter;
1240 struct hd_struct *part;
1242 if (disk->part0.policy != flag) {
1243 set_disk_ro_uevent(disk, flag);
1244 disk->part0.policy = flag;
1247 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
1248 while ((part = disk_part_iter_next(&piter)))
1249 part->policy = flag;
1250 disk_part_iter_exit(&piter);
1253 EXPORT_SYMBOL(set_disk_ro);
1255 int bdev_read_only(struct block_device *bdev)
1259 return bdev->bd_part->policy;
1262 EXPORT_SYMBOL(bdev_read_only);
1264 int invalidate_partition(struct gendisk *disk, int partno)
1267 struct block_device *bdev = bdget_disk(disk, partno);
1270 res = __invalidate_device(bdev);
1276 EXPORT_SYMBOL(invalidate_partition);