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(int 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 blk_register_region(disk_devt(disk), disk->minors, NULL,
543 exact_match, exact_lock, disk);
545 blk_register_queue(disk);
547 bdi = &disk->queue->backing_dev_info;
548 bdi_register_dev(bdi, disk_devt(disk));
549 retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
554 EXPORT_SYMBOL(add_disk);
555 EXPORT_SYMBOL(del_gendisk); /* in partitions/check.c */
557 void unlink_gendisk(struct gendisk *disk)
559 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
560 bdi_unregister(&disk->queue->backing_dev_info);
561 blk_unregister_queue(disk);
562 blk_unregister_region(disk_devt(disk), disk->minors);
566 * get_gendisk - get partitioning information for a given device
567 * @devt: device to get partitioning information for
568 * @partno: returned partition index
570 * This function gets the structure containing partitioning
571 * information for the given device @devt.
573 struct gendisk *get_gendisk(dev_t devt, int *partno)
575 struct gendisk *disk = NULL;
577 if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
578 struct kobject *kobj;
580 kobj = kobj_lookup(bdev_map, devt, partno);
582 disk = dev_to_disk(kobj_to_dev(kobj));
584 struct hd_struct *part;
586 mutex_lock(&ext_devt_mutex);
587 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
588 if (part && get_disk(part_to_disk(part))) {
589 *partno = part->partno;
590 disk = part_to_disk(part);
592 mutex_unlock(&ext_devt_mutex);
597 EXPORT_SYMBOL(get_gendisk);
600 * bdget_disk - do bdget() by gendisk and partition number
601 * @disk: gendisk of interest
602 * @partno: partition number
604 * Find partition @partno from @disk, do bdget() on it.
610 * Resulting block_device on success, NULL on failure.
612 struct block_device *bdget_disk(struct gendisk *disk, int partno)
614 struct hd_struct *part;
615 struct block_device *bdev = NULL;
617 part = disk_get_part(disk, partno);
619 bdev = bdget(part_devt(part));
624 EXPORT_SYMBOL(bdget_disk);
627 * print a full list of all partitions - intended for places where the root
628 * filesystem can't be mounted and thus to give the victim some idea of what
631 void __init printk_all_partitions(void)
633 struct class_dev_iter iter;
636 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
637 while ((dev = class_dev_iter_next(&iter))) {
638 struct gendisk *disk = dev_to_disk(dev);
639 struct disk_part_iter piter;
640 struct hd_struct *part;
641 char name_buf[BDEVNAME_SIZE];
642 char devt_buf[BDEVT_SIZE];
645 * Don't show empty devices or things that have been
648 if (get_capacity(disk) == 0 ||
649 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
653 * Note, unlike /proc/partitions, I am showing the
654 * numbers in hex - the same format as the root=
657 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
658 while ((part = disk_part_iter_next(&piter))) {
659 bool is_part0 = part == &disk->part0;
661 printk("%s%s %10llu %s", is_part0 ? "" : " ",
662 bdevt_str(part_devt(part), devt_buf),
663 (unsigned long long)part->nr_sects >> 1,
664 disk_name(disk, part->partno, name_buf));
666 if (disk->driverfs_dev != NULL &&
667 disk->driverfs_dev->driver != NULL)
668 printk(" driver: %s\n",
669 disk->driverfs_dev->driver->name);
671 printk(" (driver?)\n");
675 disk_part_iter_exit(&piter);
677 class_dev_iter_exit(&iter);
680 #ifdef CONFIG_PROC_FS
682 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
685 struct class_dev_iter *iter;
688 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
690 return ERR_PTR(-ENOMEM);
692 seqf->private = iter;
693 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
695 dev = class_dev_iter_next(iter);
700 return dev_to_disk(dev);
703 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
708 dev = class_dev_iter_next(seqf->private);
710 return dev_to_disk(dev);
715 static void disk_seqf_stop(struct seq_file *seqf, void *v)
717 struct class_dev_iter *iter = seqf->private;
719 /* stop is called even after start failed :-( */
721 class_dev_iter_exit(iter);
726 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
730 p = disk_seqf_start(seqf, pos);
731 if (!IS_ERR(p) && p && !*pos)
732 seq_puts(seqf, "major minor #blocks name\n\n");
736 static int show_partition(struct seq_file *seqf, void *v)
738 struct gendisk *sgp = v;
739 struct disk_part_iter piter;
740 struct hd_struct *part;
741 char buf[BDEVNAME_SIZE];
743 /* Don't show non-partitionable removeable devices or empty devices */
744 if (!get_capacity(sgp) || (!disk_partitionable(sgp) &&
745 (sgp->flags & GENHD_FL_REMOVABLE)))
747 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
750 /* show the full disk and all non-0 size partitions of it */
751 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
752 while ((part = disk_part_iter_next(&piter)))
753 seq_printf(seqf, "%4d %7d %10llu %s\n",
754 MAJOR(part_devt(part)), MINOR(part_devt(part)),
755 (unsigned long long)part->nr_sects >> 1,
756 disk_name(sgp, part->partno, buf));
757 disk_part_iter_exit(&piter);
762 static const struct seq_operations partitions_op = {
763 .start = show_partition_start,
764 .next = disk_seqf_next,
765 .stop = disk_seqf_stop,
766 .show = show_partition
769 static int partitions_open(struct inode *inode, struct file *file)
771 return seq_open(file, &partitions_op);
774 static const struct file_operations proc_partitions_operations = {
775 .open = partitions_open,
778 .release = seq_release,
783 static struct kobject *base_probe(dev_t devt, int *partno, void *data)
785 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
786 /* Make old-style 2.4 aliases work */
787 request_module("block-major-%d", MAJOR(devt));
791 static int __init genhd_device_init(void)
795 block_class.dev_kobj = sysfs_dev_block_kobj;
796 error = class_register(&block_class);
799 bdev_map = kobj_map_init(base_probe, &block_class_lock);
802 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
804 /* create top-level block dir */
805 if (!sysfs_deprecated)
806 block_depr = kobject_create_and_add("block", NULL);
810 subsys_initcall(genhd_device_init);
812 static ssize_t disk_range_show(struct device *dev,
813 struct device_attribute *attr, char *buf)
815 struct gendisk *disk = dev_to_disk(dev);
817 return sprintf(buf, "%d\n", disk->minors);
820 static ssize_t disk_ext_range_show(struct device *dev,
821 struct device_attribute *attr, char *buf)
823 struct gendisk *disk = dev_to_disk(dev);
825 return sprintf(buf, "%d\n", disk_max_parts(disk));
828 static ssize_t disk_removable_show(struct device *dev,
829 struct device_attribute *attr, char *buf)
831 struct gendisk *disk = dev_to_disk(dev);
833 return sprintf(buf, "%d\n",
834 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
837 static ssize_t disk_ro_show(struct device *dev,
838 struct device_attribute *attr, char *buf)
840 struct gendisk *disk = dev_to_disk(dev);
842 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
845 static ssize_t disk_capability_show(struct device *dev,
846 struct device_attribute *attr, char *buf)
848 struct gendisk *disk = dev_to_disk(dev);
850 return sprintf(buf, "%x\n", disk->flags);
853 static ssize_t disk_alignment_offset_show(struct device *dev,
854 struct device_attribute *attr,
857 struct gendisk *disk = dev_to_disk(dev);
859 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
862 static ssize_t disk_discard_alignment_show(struct device *dev,
863 struct device_attribute *attr,
866 struct gendisk *disk = dev_to_disk(dev);
868 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
871 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
872 static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
873 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
874 static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
875 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
876 static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL);
877 static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show,
879 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
880 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
881 static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
882 #ifdef CONFIG_FAIL_MAKE_REQUEST
883 static struct device_attribute dev_attr_fail =
884 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
886 #ifdef CONFIG_FAIL_IO_TIMEOUT
887 static struct device_attribute dev_attr_fail_timeout =
888 __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show,
892 static struct attribute *disk_attrs[] = {
893 &dev_attr_range.attr,
894 &dev_attr_ext_range.attr,
895 &dev_attr_removable.attr,
898 &dev_attr_alignment_offset.attr,
899 &dev_attr_discard_alignment.attr,
900 &dev_attr_capability.attr,
902 &dev_attr_inflight.attr,
903 #ifdef CONFIG_FAIL_MAKE_REQUEST
906 #ifdef CONFIG_FAIL_IO_TIMEOUT
907 &dev_attr_fail_timeout.attr,
912 static struct attribute_group disk_attr_group = {
916 static const struct attribute_group *disk_attr_groups[] = {
921 static void disk_free_ptbl_rcu_cb(struct rcu_head *head)
923 struct disk_part_tbl *ptbl =
924 container_of(head, struct disk_part_tbl, rcu_head);
930 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
931 * @disk: disk to replace part_tbl for
932 * @new_ptbl: new part_tbl to install
934 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
935 * original ptbl is freed using RCU callback.
938 * Matching bd_mutx locked.
940 static void disk_replace_part_tbl(struct gendisk *disk,
941 struct disk_part_tbl *new_ptbl)
943 struct disk_part_tbl *old_ptbl = disk->part_tbl;
945 rcu_assign_pointer(disk->part_tbl, new_ptbl);
948 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
949 call_rcu(&old_ptbl->rcu_head, disk_free_ptbl_rcu_cb);
954 * disk_expand_part_tbl - expand disk->part_tbl
955 * @disk: disk to expand part_tbl for
956 * @partno: expand such that this partno can fit in
958 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
959 * uses RCU to allow unlocked dereferencing for stats and other stuff.
962 * Matching bd_mutex locked, might sleep.
965 * 0 on success, -errno on failure.
967 int disk_expand_part_tbl(struct gendisk *disk, int partno)
969 struct disk_part_tbl *old_ptbl = disk->part_tbl;
970 struct disk_part_tbl *new_ptbl;
971 int len = old_ptbl ? old_ptbl->len : 0;
972 int target = partno + 1;
976 /* disk_max_parts() is zero during initialization, ignore if so */
977 if (disk_max_parts(disk) && target > disk_max_parts(disk))
983 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
984 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
988 new_ptbl->len = target;
990 for (i = 0; i < len; i++)
991 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
993 disk_replace_part_tbl(disk, new_ptbl);
997 static void disk_release(struct device *dev)
999 struct gendisk *disk = dev_to_disk(dev);
1001 kfree(disk->random);
1002 disk_replace_part_tbl(disk, NULL);
1003 free_part_stats(&disk->part0);
1006 struct class block_class = {
1010 static char *block_devnode(struct device *dev, mode_t *mode)
1012 struct gendisk *disk = dev_to_disk(dev);
1015 return disk->devnode(disk, mode);
1019 static struct device_type disk_type = {
1021 .groups = disk_attr_groups,
1022 .release = disk_release,
1023 .devnode = block_devnode,
1026 #ifdef CONFIG_PROC_FS
1028 * aggregate disk stat collector. Uses the same stats that the sysfs
1029 * entries do, above, but makes them available through one seq_file.
1031 * The output looks suspiciously like /proc/partitions with a bunch of
1034 static int diskstats_show(struct seq_file *seqf, void *v)
1036 struct gendisk *gp = v;
1037 struct disk_part_iter piter;
1038 struct hd_struct *hd;
1039 char buf[BDEVNAME_SIZE];
1043 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1044 seq_puts(seqf, "major minor name"
1045 " rio rmerge rsect ruse wio wmerge "
1046 "wsect wuse running use aveq"
1050 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
1051 while ((hd = disk_part_iter_next(&piter))) {
1052 cpu = part_stat_lock();
1053 part_round_stats(cpu, hd);
1055 seq_printf(seqf, "%4d %7d %s %lu %lu %llu "
1056 "%u %lu %lu %llu %u %u %u %u\n",
1057 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1058 disk_name(gp, hd->partno, buf),
1059 part_stat_read(hd, ios[0]),
1060 part_stat_read(hd, merges[0]),
1061 (unsigned long long)part_stat_read(hd, sectors[0]),
1062 jiffies_to_msecs(part_stat_read(hd, ticks[0])),
1063 part_stat_read(hd, ios[1]),
1064 part_stat_read(hd, merges[1]),
1065 (unsigned long long)part_stat_read(hd, sectors[1]),
1066 jiffies_to_msecs(part_stat_read(hd, ticks[1])),
1068 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1069 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1072 disk_part_iter_exit(&piter);
1077 static const struct seq_operations diskstats_op = {
1078 .start = disk_seqf_start,
1079 .next = disk_seqf_next,
1080 .stop = disk_seqf_stop,
1081 .show = diskstats_show
1084 static int diskstats_open(struct inode *inode, struct file *file)
1086 return seq_open(file, &diskstats_op);
1089 static const struct file_operations proc_diskstats_operations = {
1090 .open = diskstats_open,
1092 .llseek = seq_lseek,
1093 .release = seq_release,
1096 static int __init proc_genhd_init(void)
1098 proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
1099 proc_create("partitions", 0, NULL, &proc_partitions_operations);
1102 module_init(proc_genhd_init);
1103 #endif /* CONFIG_PROC_FS */
1105 static void media_change_notify_thread(struct work_struct *work)
1107 struct gendisk *gd = container_of(work, struct gendisk, async_notify);
1108 char event[] = "MEDIA_CHANGE=1";
1109 char *envp[] = { event, NULL };
1112 * set enviroment vars to indicate which event this is for
1113 * so that user space will know to go check the media status.
1115 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1116 put_device(gd->driverfs_dev);
1120 void genhd_media_change_notify(struct gendisk *disk)
1122 get_device(disk->driverfs_dev);
1123 schedule_work(&disk->async_notify);
1125 EXPORT_SYMBOL_GPL(genhd_media_change_notify);
1128 dev_t blk_lookup_devt(const char *name, int partno)
1130 dev_t devt = MKDEV(0, 0);
1131 struct class_dev_iter iter;
1134 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1135 while ((dev = class_dev_iter_next(&iter))) {
1136 struct gendisk *disk = dev_to_disk(dev);
1137 struct hd_struct *part;
1139 if (strcmp(dev_name(dev), name))
1142 if (partno < disk->minors) {
1143 /* We need to return the right devno, even
1144 * if the partition doesn't exist yet.
1146 devt = MKDEV(MAJOR(dev->devt),
1147 MINOR(dev->devt) + partno);
1150 part = disk_get_part(disk, partno);
1152 devt = part_devt(part);
1153 disk_put_part(part);
1156 disk_put_part(part);
1158 class_dev_iter_exit(&iter);
1161 EXPORT_SYMBOL(blk_lookup_devt);
1163 struct gendisk *alloc_disk(int minors)
1165 return alloc_disk_node(minors, -1);
1167 EXPORT_SYMBOL(alloc_disk);
1169 struct gendisk *alloc_disk_node(int minors, int node_id)
1171 struct gendisk *disk;
1173 disk = kmalloc_node(sizeof(struct gendisk),
1174 GFP_KERNEL | __GFP_ZERO, node_id);
1176 if (!init_part_stats(&disk->part0)) {
1180 disk->node_id = node_id;
1181 if (disk_expand_part_tbl(disk, 0)) {
1182 free_part_stats(&disk->part0);
1186 disk->part_tbl->part[0] = &disk->part0;
1188 disk->minors = minors;
1189 rand_initialize_disk(disk);
1190 disk_to_dev(disk)->class = &block_class;
1191 disk_to_dev(disk)->type = &disk_type;
1192 device_initialize(disk_to_dev(disk));
1193 INIT_WORK(&disk->async_notify,
1194 media_change_notify_thread);
1198 EXPORT_SYMBOL(alloc_disk_node);
1200 struct kobject *get_disk(struct gendisk *disk)
1202 struct module *owner;
1203 struct kobject *kobj;
1207 owner = disk->fops->owner;
1208 if (owner && !try_module_get(owner))
1210 kobj = kobject_get(&disk_to_dev(disk)->kobj);
1219 EXPORT_SYMBOL(get_disk);
1221 void put_disk(struct gendisk *disk)
1224 kobject_put(&disk_to_dev(disk)->kobj);
1227 EXPORT_SYMBOL(put_disk);
1229 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1231 char event[] = "DISK_RO=1";
1232 char *envp[] = { event, NULL };
1236 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1239 void set_device_ro(struct block_device *bdev, int flag)
1241 bdev->bd_part->policy = flag;
1244 EXPORT_SYMBOL(set_device_ro);
1246 void set_disk_ro(struct gendisk *disk, int flag)
1248 struct disk_part_iter piter;
1249 struct hd_struct *part;
1251 if (disk->part0.policy != flag) {
1252 set_disk_ro_uevent(disk, flag);
1253 disk->part0.policy = flag;
1256 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
1257 while ((part = disk_part_iter_next(&piter)))
1258 part->policy = flag;
1259 disk_part_iter_exit(&piter);
1262 EXPORT_SYMBOL(set_disk_ro);
1264 int bdev_read_only(struct block_device *bdev)
1268 return bdev->bd_part->policy;
1271 EXPORT_SYMBOL(bdev_read_only);
1273 int invalidate_partition(struct gendisk *disk, int partno)
1276 struct block_device *bdev = bdget_disk(disk, partno);
1279 res = __invalidate_device(bdev);
1285 EXPORT_SYMBOL(invalidate_partition);