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/backing-dev.h>
12 #include <linux/init.h>
13 #include <linux/spinlock.h>
14 #include <linux/proc_fs.h>
15 #include <linux/seq_file.h>
16 #include <linux/slab.h>
17 #include <linux/kmod.h>
18 #include <linux/kobj_map.h>
19 #include <linux/mutex.h>
20 #include <linux/idr.h>
21 #include <linux/log2.h>
22 #include <linux/pm_runtime.h>
26 static DEFINE_MUTEX(block_class_lock);
27 struct kobject *block_depr;
29 /* for extended dynamic devt allocation, currently only one major is used */
30 #define NR_EXT_DEVT (1 << MINORBITS)
32 /* For extended devt allocation. ext_devt_lock prevents look up
33 * results from going away underneath its user.
35 static DEFINE_SPINLOCK(ext_devt_lock);
36 static DEFINE_IDR(ext_devt_idr);
38 static struct device_type disk_type;
40 static void disk_check_events(struct disk_events *ev,
41 unsigned int *clearing_ptr);
42 static void disk_alloc_events(struct gendisk *disk);
43 static void disk_add_events(struct gendisk *disk);
44 static void disk_del_events(struct gendisk *disk);
45 static void disk_release_events(struct gendisk *disk);
48 * disk_get_part - get partition
49 * @disk: disk to look partition from
50 * @partno: partition number
52 * Look for partition @partno from @disk. If found, increment
53 * reference count and return it.
59 * Pointer to the found partition on success, NULL if not found.
61 struct hd_struct *disk_get_part(struct gendisk *disk, int partno)
63 struct hd_struct *part = NULL;
64 struct disk_part_tbl *ptbl;
66 if (unlikely(partno < 0))
71 ptbl = rcu_dereference(disk->part_tbl);
72 if (likely(partno < ptbl->len)) {
73 part = rcu_dereference(ptbl->part[partno]);
75 get_device(part_to_dev(part));
82 EXPORT_SYMBOL_GPL(disk_get_part);
85 * disk_part_iter_init - initialize partition iterator
86 * @piter: iterator to initialize
87 * @disk: disk to iterate over
88 * @flags: DISK_PITER_* flags
90 * Initialize @piter so that it iterates over partitions of @disk.
95 void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk,
98 struct disk_part_tbl *ptbl;
101 ptbl = rcu_dereference(disk->part_tbl);
106 if (flags & DISK_PITER_REVERSE)
107 piter->idx = ptbl->len - 1;
108 else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0))
113 piter->flags = flags;
117 EXPORT_SYMBOL_GPL(disk_part_iter_init);
120 * disk_part_iter_next - proceed iterator to the next partition and return it
121 * @piter: iterator of interest
123 * Proceed @piter to the next partition and return it.
128 struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter)
130 struct disk_part_tbl *ptbl;
133 /* put the last partition */
134 disk_put_part(piter->part);
139 ptbl = rcu_dereference(piter->disk->part_tbl);
141 /* determine iteration parameters */
142 if (piter->flags & DISK_PITER_REVERSE) {
144 if (piter->flags & (DISK_PITER_INCL_PART0 |
145 DISK_PITER_INCL_EMPTY_PART0))
154 /* iterate to the next partition */
155 for (; piter->idx != end; piter->idx += inc) {
156 struct hd_struct *part;
158 part = rcu_dereference(ptbl->part[piter->idx]);
161 if (!part_nr_sects_read(part) &&
162 !(piter->flags & DISK_PITER_INCL_EMPTY) &&
163 !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 &&
167 get_device(part_to_dev(part));
177 EXPORT_SYMBOL_GPL(disk_part_iter_next);
180 * disk_part_iter_exit - finish up partition iteration
181 * @piter: iter of interest
183 * Called when iteration is over. Cleans up @piter.
188 void disk_part_iter_exit(struct disk_part_iter *piter)
190 disk_put_part(piter->part);
193 EXPORT_SYMBOL_GPL(disk_part_iter_exit);
195 static inline int sector_in_part(struct hd_struct *part, sector_t sector)
197 return part->start_sect <= sector &&
198 sector < part->start_sect + part_nr_sects_read(part);
202 * disk_map_sector_rcu - map sector to partition
203 * @disk: gendisk of interest
204 * @sector: sector to map
206 * Find out which partition @sector maps to on @disk. This is
207 * primarily used for stats accounting.
210 * RCU read locked. The returned partition pointer is valid only
211 * while preemption is disabled.
214 * Found partition on success, part0 is returned if no partition matches
216 struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector)
218 struct disk_part_tbl *ptbl;
219 struct hd_struct *part;
222 ptbl = rcu_dereference(disk->part_tbl);
224 part = rcu_dereference(ptbl->last_lookup);
225 if (part && sector_in_part(part, sector))
228 for (i = 1; i < ptbl->len; i++) {
229 part = rcu_dereference(ptbl->part[i]);
231 if (part && sector_in_part(part, sector)) {
232 rcu_assign_pointer(ptbl->last_lookup, part);
238 EXPORT_SYMBOL_GPL(disk_map_sector_rcu);
241 * Can be deleted altogether. Later.
244 static struct blk_major_name {
245 struct blk_major_name *next;
248 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
250 /* index in the above - for now: assume no multimajor ranges */
251 static inline int major_to_index(unsigned major)
253 return major % BLKDEV_MAJOR_HASH_SIZE;
256 #ifdef CONFIG_PROC_FS
257 void blkdev_show(struct seq_file *seqf, off_t offset)
259 struct blk_major_name *dp;
261 if (offset < BLKDEV_MAJOR_HASH_SIZE) {
262 mutex_lock(&block_class_lock);
263 for (dp = major_names[offset]; dp; dp = dp->next)
264 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
265 mutex_unlock(&block_class_lock);
268 #endif /* CONFIG_PROC_FS */
271 * register_blkdev - register a new block device
273 * @major: the requested major device number [1..255]. If @major=0, try to
274 * allocate any unused major number.
275 * @name: the name of the new block device as a zero terminated string
277 * The @name must be unique within the system.
279 * The return value depends on the @major input parameter.
280 * - if a major device number was requested in range [1..255] then the
281 * function returns zero on success, or a negative error code
282 * - if any unused major number was requested with @major=0 parameter
283 * then the return value is the allocated major number in range
284 * [1..255] or a negative error code otherwise
286 int register_blkdev(unsigned int major, const char *name)
288 struct blk_major_name **n, *p;
291 mutex_lock(&block_class_lock);
295 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
296 if (major_names[index] == NULL)
301 printk("register_blkdev: failed to get major for %s\n",
310 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
317 strlcpy(p->name, name, sizeof(p->name));
319 index = major_to_index(major);
321 for (n = &major_names[index]; *n; n = &(*n)->next) {
322 if ((*n)->major == major)
331 printk("register_blkdev: cannot get major %d for %s\n",
336 mutex_unlock(&block_class_lock);
340 EXPORT_SYMBOL(register_blkdev);
342 void unregister_blkdev(unsigned int major, const char *name)
344 struct blk_major_name **n;
345 struct blk_major_name *p = NULL;
346 int index = major_to_index(major);
348 mutex_lock(&block_class_lock);
349 for (n = &major_names[index]; *n; n = &(*n)->next)
350 if ((*n)->major == major)
352 if (!*n || strcmp((*n)->name, name)) {
358 mutex_unlock(&block_class_lock);
362 EXPORT_SYMBOL(unregister_blkdev);
364 static struct kobj_map *bdev_map;
367 * blk_mangle_minor - scatter minor numbers apart
368 * @minor: minor number to mangle
370 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
371 * is enabled. Mangling twice gives the original value.
379 static int blk_mangle_minor(int minor)
381 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
384 for (i = 0; i < MINORBITS / 2; i++) {
385 int low = minor & (1 << i);
386 int high = minor & (1 << (MINORBITS - 1 - i));
387 int distance = MINORBITS - 1 - 2 * i;
389 minor ^= low | high; /* clear both bits */
390 low <<= distance; /* swap the positions */
392 minor |= low | high; /* and set */
399 * blk_alloc_devt - allocate a dev_t for a partition
400 * @part: partition to allocate dev_t for
401 * @devt: out parameter for resulting dev_t
403 * Allocate a dev_t for block device.
406 * 0 on success, allocated dev_t is returned in *@devt. -errno on
412 int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
414 struct gendisk *disk = part_to_disk(part);
417 /* in consecutive minor range? */
418 if (part->partno < disk->minors) {
419 *devt = MKDEV(disk->major, disk->first_minor + part->partno);
423 /* allocate ext devt */
424 idr_preload(GFP_KERNEL);
426 spin_lock_bh(&ext_devt_lock);
427 idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_NOWAIT);
428 spin_unlock_bh(&ext_devt_lock);
432 return idx == -ENOSPC ? -EBUSY : idx;
434 *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
439 * blk_free_devt - free a dev_t
440 * @devt: dev_t to free
442 * Free @devt which was allocated using blk_alloc_devt().
447 void blk_free_devt(dev_t devt)
449 if (devt == MKDEV(0, 0))
452 if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
453 spin_lock_bh(&ext_devt_lock);
454 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
455 spin_unlock_bh(&ext_devt_lock);
459 static char *bdevt_str(dev_t devt, char *buf)
461 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
462 char tbuf[BDEVT_SIZE];
463 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
464 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
466 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
472 * Register device numbers dev..(dev+range-1)
473 * range must be nonzero
474 * The hash chain is sorted on range, so that subranges can override.
476 void blk_register_region(dev_t devt, unsigned long range, struct module *module,
477 struct kobject *(*probe)(dev_t, int *, void *),
478 int (*lock)(dev_t, void *), void *data)
480 kobj_map(bdev_map, devt, range, module, probe, lock, data);
483 EXPORT_SYMBOL(blk_register_region);
485 void blk_unregister_region(dev_t devt, unsigned long range)
487 kobj_unmap(bdev_map, devt, range);
490 EXPORT_SYMBOL(blk_unregister_region);
492 static struct kobject *exact_match(dev_t devt, int *partno, void *data)
494 struct gendisk *p = data;
496 return &disk_to_dev(p)->kobj;
499 static int exact_lock(dev_t devt, void *data)
501 struct gendisk *p = data;
508 static void register_disk(struct gendisk *disk)
510 struct device *ddev = disk_to_dev(disk);
511 struct block_device *bdev;
512 struct disk_part_iter piter;
513 struct hd_struct *part;
516 ddev->parent = disk->driverfs_dev;
518 dev_set_name(ddev, "%s", disk->disk_name);
520 /* delay uevents, until we scanned partition table */
521 dev_set_uevent_suppress(ddev, 1);
523 if (device_add(ddev))
525 if (!sysfs_deprecated) {
526 err = sysfs_create_link(block_depr, &ddev->kobj,
527 kobject_name(&ddev->kobj));
535 * avoid probable deadlock caused by allocating memory with
536 * GFP_KERNEL in runtime_resume callback of its all ancestor
539 pm_runtime_set_memalloc_noio(ddev, true);
541 disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj);
542 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
544 /* No minors to use for partitions */
545 if (!disk_part_scan_enabled(disk))
548 /* No such device (e.g., media were just removed) */
549 if (!get_capacity(disk))
552 bdev = bdget_disk(disk, 0);
556 bdev->bd_invalidated = 1;
557 err = blkdev_get(bdev, FMODE_READ, NULL);
560 blkdev_put(bdev, FMODE_READ);
563 /* announce disk after possible partitions are created */
564 dev_set_uevent_suppress(ddev, 0);
565 kobject_uevent(&ddev->kobj, KOBJ_ADD);
567 /* announce possible partitions */
568 disk_part_iter_init(&piter, disk, 0);
569 while ((part = disk_part_iter_next(&piter)))
570 kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD);
571 disk_part_iter_exit(&piter);
575 * add_disk - add partitioning information to kernel list
576 * @disk: per-device partitioning information
578 * This function registers the partitioning information in @disk
581 * FIXME: error handling
583 void add_disk(struct gendisk *disk)
585 struct backing_dev_info *bdi;
589 /* minors == 0 indicates to use ext devt from part0 and should
590 * be accompanied with EXT_DEVT flag. Make sure all
591 * parameters make sense.
593 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
594 WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT));
596 disk->flags |= GENHD_FL_UP;
598 retval = blk_alloc_devt(&disk->part0, &devt);
603 disk_to_dev(disk)->devt = devt;
605 /* ->major and ->first_minor aren't supposed to be
606 * dereferenced from here on, but set them just in case.
608 disk->major = MAJOR(devt);
609 disk->first_minor = MINOR(devt);
611 disk_alloc_events(disk);
613 /* Register BDI before referencing it from bdev */
614 bdi = &disk->queue->backing_dev_info;
615 bdi_register_owner(bdi, disk_to_dev(disk));
617 blk_register_region(disk_devt(disk), disk->minors, NULL,
618 exact_match, exact_lock, disk);
620 blk_register_queue(disk);
623 * Take an extra ref on queue which will be put on disk_release()
624 * so that it sticks around as long as @disk is there.
626 WARN_ON_ONCE(!blk_get_queue(disk->queue));
628 retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
632 disk_add_events(disk);
633 blk_integrity_add(disk);
635 EXPORT_SYMBOL(add_disk);
637 void del_gendisk(struct gendisk *disk)
639 struct disk_part_iter piter;
640 struct hd_struct *part;
642 blk_integrity_del(disk);
643 disk_del_events(disk);
645 /* invalidate stuff */
646 disk_part_iter_init(&piter, disk,
647 DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
648 while ((part = disk_part_iter_next(&piter))) {
649 invalidate_partition(disk, part->partno);
650 delete_partition(disk, part->partno);
652 disk_part_iter_exit(&piter);
654 invalidate_partition(disk, 0);
655 set_capacity(disk, 0);
656 disk->flags &= ~GENHD_FL_UP;
658 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
659 blk_unregister_queue(disk);
660 blk_unregister_region(disk_devt(disk), disk->minors);
662 part_stat_set_all(&disk->part0, 0);
663 disk->part0.stamp = 0;
665 kobject_put(disk->part0.holder_dir);
666 kobject_put(disk->slave_dir);
667 disk->driverfs_dev = NULL;
668 if (!sysfs_deprecated)
669 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
670 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
671 device_del(disk_to_dev(disk));
673 EXPORT_SYMBOL(del_gendisk);
676 * get_gendisk - get partitioning information for a given device
677 * @devt: device to get partitioning information for
678 * @partno: returned partition index
680 * This function gets the structure containing partitioning
681 * information for the given device @devt.
683 struct gendisk *get_gendisk(dev_t devt, int *partno)
685 struct gendisk *disk = NULL;
687 if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
688 struct kobject *kobj;
690 kobj = kobj_lookup(bdev_map, devt, partno);
692 disk = dev_to_disk(kobj_to_dev(kobj));
694 struct hd_struct *part;
696 spin_lock_bh(&ext_devt_lock);
697 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
698 if (part && get_disk(part_to_disk(part))) {
699 *partno = part->partno;
700 disk = part_to_disk(part);
702 spin_unlock_bh(&ext_devt_lock);
707 EXPORT_SYMBOL(get_gendisk);
710 * bdget_disk - do bdget() by gendisk and partition number
711 * @disk: gendisk of interest
712 * @partno: partition number
714 * Find partition @partno from @disk, do bdget() on it.
720 * Resulting block_device on success, NULL on failure.
722 struct block_device *bdget_disk(struct gendisk *disk, int partno)
724 struct hd_struct *part;
725 struct block_device *bdev = NULL;
727 part = disk_get_part(disk, partno);
729 bdev = bdget(part_devt(part));
734 EXPORT_SYMBOL(bdget_disk);
737 * print a full list of all partitions - intended for places where the root
738 * filesystem can't be mounted and thus to give the victim some idea of what
741 void __init printk_all_partitions(void)
743 struct class_dev_iter iter;
746 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
747 while ((dev = class_dev_iter_next(&iter))) {
748 struct gendisk *disk = dev_to_disk(dev);
749 struct disk_part_iter piter;
750 struct hd_struct *part;
751 char name_buf[BDEVNAME_SIZE];
752 char devt_buf[BDEVT_SIZE];
755 * Don't show empty devices or things that have been
758 if (get_capacity(disk) == 0 ||
759 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
763 * Note, unlike /proc/partitions, I am showing the
764 * numbers in hex - the same format as the root=
767 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
768 while ((part = disk_part_iter_next(&piter))) {
769 bool is_part0 = part == &disk->part0;
771 printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
772 bdevt_str(part_devt(part), devt_buf),
773 (unsigned long long)part_nr_sects_read(part) >> 1
774 , disk_name(disk, part->partno, name_buf),
775 part->info ? part->info->uuid : "");
777 if (disk->driverfs_dev != NULL &&
778 disk->driverfs_dev->driver != NULL)
779 printk(" driver: %s\n",
780 disk->driverfs_dev->driver->name);
782 printk(" (driver?)\n");
786 disk_part_iter_exit(&piter);
788 class_dev_iter_exit(&iter);
791 #ifdef CONFIG_PROC_FS
793 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
796 struct class_dev_iter *iter;
799 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
801 return ERR_PTR(-ENOMEM);
803 seqf->private = iter;
804 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
806 dev = class_dev_iter_next(iter);
811 return dev_to_disk(dev);
814 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
819 dev = class_dev_iter_next(seqf->private);
821 return dev_to_disk(dev);
826 static void disk_seqf_stop(struct seq_file *seqf, void *v)
828 struct class_dev_iter *iter = seqf->private;
830 /* stop is called even after start failed :-( */
832 class_dev_iter_exit(iter);
834 seqf->private = NULL;
838 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
842 p = disk_seqf_start(seqf, pos);
843 if (!IS_ERR_OR_NULL(p) && !*pos)
844 seq_puts(seqf, "major minor #blocks name\n\n");
848 static int show_partition(struct seq_file *seqf, void *v)
850 struct gendisk *sgp = v;
851 struct disk_part_iter piter;
852 struct hd_struct *part;
853 char buf[BDEVNAME_SIZE];
855 /* Don't show non-partitionable removeable devices or empty devices */
856 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
857 (sgp->flags & GENHD_FL_REMOVABLE)))
859 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
862 /* show the full disk and all non-0 size partitions of it */
863 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
864 while ((part = disk_part_iter_next(&piter)))
865 seq_printf(seqf, "%4d %7d %10llu %s\n",
866 MAJOR(part_devt(part)), MINOR(part_devt(part)),
867 (unsigned long long)part_nr_sects_read(part) >> 1,
868 disk_name(sgp, part->partno, buf));
869 disk_part_iter_exit(&piter);
874 static const struct seq_operations partitions_op = {
875 .start = show_partition_start,
876 .next = disk_seqf_next,
877 .stop = disk_seqf_stop,
878 .show = show_partition
881 static int partitions_open(struct inode *inode, struct file *file)
883 return seq_open(file, &partitions_op);
886 static const struct file_operations proc_partitions_operations = {
887 .open = partitions_open,
890 .release = seq_release,
895 static struct kobject *base_probe(dev_t devt, int *partno, void *data)
897 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
898 /* Make old-style 2.4 aliases work */
899 request_module("block-major-%d", MAJOR(devt));
903 static int __init genhd_device_init(void)
907 block_class.dev_kobj = sysfs_dev_block_kobj;
908 error = class_register(&block_class);
911 bdev_map = kobj_map_init(base_probe, &block_class_lock);
914 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
916 /* create top-level block dir */
917 if (!sysfs_deprecated)
918 block_depr = kobject_create_and_add("block", NULL);
922 subsys_initcall(genhd_device_init);
924 static ssize_t disk_range_show(struct device *dev,
925 struct device_attribute *attr, char *buf)
927 struct gendisk *disk = dev_to_disk(dev);
929 return sprintf(buf, "%d\n", disk->minors);
932 static ssize_t disk_ext_range_show(struct device *dev,
933 struct device_attribute *attr, char *buf)
935 struct gendisk *disk = dev_to_disk(dev);
937 return sprintf(buf, "%d\n", disk_max_parts(disk));
940 static ssize_t disk_removable_show(struct device *dev,
941 struct device_attribute *attr, char *buf)
943 struct gendisk *disk = dev_to_disk(dev);
945 return sprintf(buf, "%d\n",
946 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
949 static ssize_t disk_ro_show(struct device *dev,
950 struct device_attribute *attr, char *buf)
952 struct gendisk *disk = dev_to_disk(dev);
954 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
957 static ssize_t disk_capability_show(struct device *dev,
958 struct device_attribute *attr, char *buf)
960 struct gendisk *disk = dev_to_disk(dev);
962 return sprintf(buf, "%x\n", disk->flags);
965 static ssize_t disk_alignment_offset_show(struct device *dev,
966 struct device_attribute *attr,
969 struct gendisk *disk = dev_to_disk(dev);
971 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
974 static ssize_t disk_discard_alignment_show(struct device *dev,
975 struct device_attribute *attr,
978 struct gendisk *disk = dev_to_disk(dev);
980 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
983 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
984 static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
985 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
986 static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
987 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
988 static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL);
989 static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show,
991 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
992 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
993 static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
994 #ifdef CONFIG_FAIL_MAKE_REQUEST
995 static struct device_attribute dev_attr_fail =
996 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
998 #ifdef CONFIG_FAIL_IO_TIMEOUT
999 static struct device_attribute dev_attr_fail_timeout =
1000 __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show,
1001 part_timeout_store);
1004 static struct attribute *disk_attrs[] = {
1005 &dev_attr_range.attr,
1006 &dev_attr_ext_range.attr,
1007 &dev_attr_removable.attr,
1009 &dev_attr_size.attr,
1010 &dev_attr_alignment_offset.attr,
1011 &dev_attr_discard_alignment.attr,
1012 &dev_attr_capability.attr,
1013 &dev_attr_stat.attr,
1014 &dev_attr_inflight.attr,
1015 #ifdef CONFIG_FAIL_MAKE_REQUEST
1016 &dev_attr_fail.attr,
1018 #ifdef CONFIG_FAIL_IO_TIMEOUT
1019 &dev_attr_fail_timeout.attr,
1024 static struct attribute_group disk_attr_group = {
1025 .attrs = disk_attrs,
1028 static const struct attribute_group *disk_attr_groups[] = {
1034 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1035 * @disk: disk to replace part_tbl for
1036 * @new_ptbl: new part_tbl to install
1038 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1039 * original ptbl is freed using RCU callback.
1042 * Matching bd_mutx locked.
1044 static void disk_replace_part_tbl(struct gendisk *disk,
1045 struct disk_part_tbl *new_ptbl)
1047 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1049 rcu_assign_pointer(disk->part_tbl, new_ptbl);
1052 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
1053 kfree_rcu(old_ptbl, rcu_head);
1058 * disk_expand_part_tbl - expand disk->part_tbl
1059 * @disk: disk to expand part_tbl for
1060 * @partno: expand such that this partno can fit in
1062 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1063 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1066 * Matching bd_mutex locked, might sleep.
1069 * 0 on success, -errno on failure.
1071 int disk_expand_part_tbl(struct gendisk *disk, int partno)
1073 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1074 struct disk_part_tbl *new_ptbl;
1075 int len = old_ptbl ? old_ptbl->len : 0;
1080 * check for int overflow, since we can get here from blkpg_ioctl()
1081 * with a user passed 'partno'.
1083 target = partno + 1;
1087 /* disk_max_parts() is zero during initialization, ignore if so */
1088 if (disk_max_parts(disk) && target > disk_max_parts(disk))
1094 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
1095 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
1099 new_ptbl->len = target;
1101 for (i = 0; i < len; i++)
1102 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
1104 disk_replace_part_tbl(disk, new_ptbl);
1108 static void disk_release(struct device *dev)
1110 struct gendisk *disk = dev_to_disk(dev);
1112 blk_free_devt(dev->devt);
1113 disk_release_events(disk);
1114 kfree(disk->random);
1115 disk_replace_part_tbl(disk, NULL);
1116 hd_free_part(&disk->part0);
1118 blk_put_queue(disk->queue);
1122 static int disk_uevent(struct device *dev, struct kobj_uevent_env *env)
1124 struct gendisk *disk = dev_to_disk(dev);
1125 struct disk_part_iter piter;
1126 struct hd_struct *part;
1129 disk_part_iter_init(&piter, disk, 0);
1130 while((part = disk_part_iter_next(&piter)))
1132 disk_part_iter_exit(&piter);
1133 add_uevent_var(env, "NPARTS=%u", cnt);
1137 struct class block_class = {
1141 static char *block_devnode(struct device *dev, umode_t *mode,
1142 kuid_t *uid, kgid_t *gid)
1144 struct gendisk *disk = dev_to_disk(dev);
1147 return disk->devnode(disk, mode);
1151 static struct device_type disk_type = {
1153 .groups = disk_attr_groups,
1154 .release = disk_release,
1155 .devnode = block_devnode,
1156 .uevent = disk_uevent,
1159 #ifdef CONFIG_PROC_FS
1161 * aggregate disk stat collector. Uses the same stats that the sysfs
1162 * entries do, above, but makes them available through one seq_file.
1164 * The output looks suspiciously like /proc/partitions with a bunch of
1167 static int diskstats_show(struct seq_file *seqf, void *v)
1169 struct gendisk *gp = v;
1170 struct disk_part_iter piter;
1171 struct hd_struct *hd;
1172 char buf[BDEVNAME_SIZE];
1176 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1177 seq_puts(seqf, "major minor name"
1178 " rio rmerge rsect ruse wio wmerge "
1179 "wsect wuse running use aveq"
1183 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
1184 while ((hd = disk_part_iter_next(&piter))) {
1185 cpu = part_stat_lock();
1186 part_round_stats(cpu, hd);
1188 seq_printf(seqf, "%4d %7d %s %lu %lu %lu "
1189 "%u %lu %lu %lu %u %u %u %u\n",
1190 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1191 disk_name(gp, hd->partno, buf),
1192 part_stat_read(hd, ios[READ]),
1193 part_stat_read(hd, merges[READ]),
1194 part_stat_read(hd, sectors[READ]),
1195 jiffies_to_msecs(part_stat_read(hd, ticks[READ])),
1196 part_stat_read(hd, ios[WRITE]),
1197 part_stat_read(hd, merges[WRITE]),
1198 part_stat_read(hd, sectors[WRITE]),
1199 jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])),
1201 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1202 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1205 disk_part_iter_exit(&piter);
1210 static const struct seq_operations diskstats_op = {
1211 .start = disk_seqf_start,
1212 .next = disk_seqf_next,
1213 .stop = disk_seqf_stop,
1214 .show = diskstats_show
1217 static int diskstats_open(struct inode *inode, struct file *file)
1219 return seq_open(file, &diskstats_op);
1222 static const struct file_operations proc_diskstats_operations = {
1223 .open = diskstats_open,
1225 .llseek = seq_lseek,
1226 .release = seq_release,
1229 static int __init proc_genhd_init(void)
1231 proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
1232 proc_create("partitions", 0, NULL, &proc_partitions_operations);
1235 module_init(proc_genhd_init);
1236 #endif /* CONFIG_PROC_FS */
1238 dev_t blk_lookup_devt(const char *name, int partno)
1240 dev_t devt = MKDEV(0, 0);
1241 struct class_dev_iter iter;
1244 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1245 while ((dev = class_dev_iter_next(&iter))) {
1246 struct gendisk *disk = dev_to_disk(dev);
1247 struct hd_struct *part;
1249 if (strcmp(dev_name(dev), name))
1252 if (partno < disk->minors) {
1253 /* We need to return the right devno, even
1254 * if the partition doesn't exist yet.
1256 devt = MKDEV(MAJOR(dev->devt),
1257 MINOR(dev->devt) + partno);
1260 part = disk_get_part(disk, partno);
1262 devt = part_devt(part);
1263 disk_put_part(part);
1266 disk_put_part(part);
1268 class_dev_iter_exit(&iter);
1271 EXPORT_SYMBOL(blk_lookup_devt);
1273 struct gendisk *alloc_disk(int minors)
1275 return alloc_disk_node(minors, NUMA_NO_NODE);
1277 EXPORT_SYMBOL(alloc_disk);
1279 struct gendisk *alloc_disk_node(int minors, int node_id)
1281 struct gendisk *disk;
1283 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1285 if (!init_part_stats(&disk->part0)) {
1289 disk->node_id = node_id;
1290 if (disk_expand_part_tbl(disk, 0)) {
1291 free_part_stats(&disk->part0);
1295 disk->part_tbl->part[0] = &disk->part0;
1298 * set_capacity() and get_capacity() currently don't use
1299 * seqcounter to read/update the part0->nr_sects. Still init
1300 * the counter as we can read the sectors in IO submission
1301 * patch using seqence counters.
1303 * TODO: Ideally set_capacity() and get_capacity() should be
1304 * converted to make use of bd_mutex and sequence counters.
1306 seqcount_init(&disk->part0.nr_sects_seq);
1307 if (hd_ref_init(&disk->part0)) {
1308 hd_free_part(&disk->part0);
1313 disk->minors = minors;
1314 rand_initialize_disk(disk);
1315 disk_to_dev(disk)->class = &block_class;
1316 disk_to_dev(disk)->type = &disk_type;
1317 device_initialize(disk_to_dev(disk));
1321 EXPORT_SYMBOL(alloc_disk_node);
1323 struct kobject *get_disk(struct gendisk *disk)
1325 struct module *owner;
1326 struct kobject *kobj;
1330 owner = disk->fops->owner;
1331 if (owner && !try_module_get(owner))
1333 kobj = kobject_get(&disk_to_dev(disk)->kobj);
1342 EXPORT_SYMBOL(get_disk);
1344 void put_disk(struct gendisk *disk)
1347 kobject_put(&disk_to_dev(disk)->kobj);
1350 EXPORT_SYMBOL(put_disk);
1352 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1354 char event[] = "DISK_RO=1";
1355 char *envp[] = { event, NULL };
1359 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1362 void set_device_ro(struct block_device *bdev, int flag)
1364 bdev->bd_part->policy = flag;
1367 EXPORT_SYMBOL(set_device_ro);
1369 void set_disk_ro(struct gendisk *disk, int flag)
1371 struct disk_part_iter piter;
1372 struct hd_struct *part;
1374 if (disk->part0.policy != flag) {
1375 set_disk_ro_uevent(disk, flag);
1376 disk->part0.policy = flag;
1379 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
1380 while ((part = disk_part_iter_next(&piter)))
1381 part->policy = flag;
1382 disk_part_iter_exit(&piter);
1385 EXPORT_SYMBOL(set_disk_ro);
1387 int bdev_read_only(struct block_device *bdev)
1391 return bdev->bd_part->policy;
1394 EXPORT_SYMBOL(bdev_read_only);
1396 int invalidate_partition(struct gendisk *disk, int partno)
1399 struct block_device *bdev = bdget_disk(disk, partno);
1402 res = __invalidate_device(bdev, true);
1408 EXPORT_SYMBOL(invalidate_partition);
1411 * Disk events - monitor disk events like media change and eject request.
1413 struct disk_events {
1414 struct list_head node; /* all disk_event's */
1415 struct gendisk *disk; /* the associated disk */
1418 struct mutex block_mutex; /* protects blocking */
1419 int block; /* event blocking depth */
1420 unsigned int pending; /* events already sent out */
1421 unsigned int clearing; /* events being cleared */
1423 long poll_msecs; /* interval, -1 for default */
1424 struct delayed_work dwork;
1427 static const char *disk_events_strs[] = {
1428 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1429 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1432 static char *disk_uevents[] = {
1433 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1434 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1437 /* list of all disk_events */
1438 static DEFINE_MUTEX(disk_events_mutex);
1439 static LIST_HEAD(disk_events);
1441 /* disable in-kernel polling by default */
1442 static unsigned long disk_events_dfl_poll_msecs = 0;
1444 static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1446 struct disk_events *ev = disk->ev;
1447 long intv_msecs = 0;
1450 * If device-specific poll interval is set, always use it. If
1451 * the default is being used, poll iff there are events which
1452 * can't be monitored asynchronously.
1454 if (ev->poll_msecs >= 0)
1455 intv_msecs = ev->poll_msecs;
1456 else if (disk->events & ~disk->async_events)
1457 intv_msecs = disk_events_dfl_poll_msecs;
1459 return msecs_to_jiffies(intv_msecs);
1463 * disk_block_events - block and flush disk event checking
1464 * @disk: disk to block events for
1466 * On return from this function, it is guaranteed that event checking
1467 * isn't in progress and won't happen until unblocked by
1468 * disk_unblock_events(). Events blocking is counted and the actual
1469 * unblocking happens after the matching number of unblocks are done.
1471 * Note that this intentionally does not block event checking from
1472 * disk_clear_events().
1477 void disk_block_events(struct gendisk *disk)
1479 struct disk_events *ev = disk->ev;
1480 unsigned long flags;
1487 * Outer mutex ensures that the first blocker completes canceling
1488 * the event work before further blockers are allowed to finish.
1490 mutex_lock(&ev->block_mutex);
1492 spin_lock_irqsave(&ev->lock, flags);
1493 cancel = !ev->block++;
1494 spin_unlock_irqrestore(&ev->lock, flags);
1497 cancel_delayed_work_sync(&disk->ev->dwork);
1499 mutex_unlock(&ev->block_mutex);
1502 static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1504 struct disk_events *ev = disk->ev;
1506 unsigned long flags;
1508 spin_lock_irqsave(&ev->lock, flags);
1510 if (WARN_ON_ONCE(ev->block <= 0))
1517 * Not exactly a latency critical operation, set poll timer
1518 * slack to 25% and kick event check.
1520 intv = disk_events_poll_jiffies(disk);
1521 set_timer_slack(&ev->dwork.timer, intv / 4);
1523 queue_delayed_work(system_freezable_power_efficient_wq,
1526 queue_delayed_work(system_freezable_power_efficient_wq,
1529 spin_unlock_irqrestore(&ev->lock, flags);
1533 * disk_unblock_events - unblock disk event checking
1534 * @disk: disk to unblock events for
1536 * Undo disk_block_events(). When the block count reaches zero, it
1537 * starts events polling if configured.
1540 * Don't care. Safe to call from irq context.
1542 void disk_unblock_events(struct gendisk *disk)
1545 __disk_unblock_events(disk, false);
1549 * disk_flush_events - schedule immediate event checking and flushing
1550 * @disk: disk to check and flush events for
1551 * @mask: events to flush
1553 * Schedule immediate event checking on @disk if not blocked. Events in
1554 * @mask are scheduled to be cleared from the driver. Note that this
1555 * doesn't clear the events from @disk->ev.
1558 * If @mask is non-zero must be called with bdev->bd_mutex held.
1560 void disk_flush_events(struct gendisk *disk, unsigned int mask)
1562 struct disk_events *ev = disk->ev;
1567 spin_lock_irq(&ev->lock);
1568 ev->clearing |= mask;
1570 mod_delayed_work(system_freezable_power_efficient_wq,
1572 spin_unlock_irq(&ev->lock);
1576 * disk_clear_events - synchronously check, clear and return pending events
1577 * @disk: disk to fetch and clear events from
1578 * @mask: mask of events to be fetched and cleared
1580 * Disk events are synchronously checked and pending events in @mask
1581 * are cleared and returned. This ignores the block count.
1586 unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1588 const struct block_device_operations *bdops = disk->fops;
1589 struct disk_events *ev = disk->ev;
1590 unsigned int pending;
1591 unsigned int clearing = mask;
1594 /* for drivers still using the old ->media_changed method */
1595 if ((mask & DISK_EVENT_MEDIA_CHANGE) &&
1596 bdops->media_changed && bdops->media_changed(disk))
1597 return DISK_EVENT_MEDIA_CHANGE;
1601 disk_block_events(disk);
1604 * store the union of mask and ev->clearing on the stack so that the
1605 * race with disk_flush_events does not cause ambiguity (ev->clearing
1606 * can still be modified even if events are blocked).
1608 spin_lock_irq(&ev->lock);
1609 clearing |= ev->clearing;
1611 spin_unlock_irq(&ev->lock);
1613 disk_check_events(ev, &clearing);
1615 * if ev->clearing is not 0, the disk_flush_events got called in the
1616 * middle of this function, so we want to run the workfn without delay.
1618 __disk_unblock_events(disk, ev->clearing ? true : false);
1620 /* then, fetch and clear pending events */
1621 spin_lock_irq(&ev->lock);
1622 pending = ev->pending & mask;
1623 ev->pending &= ~mask;
1624 spin_unlock_irq(&ev->lock);
1625 WARN_ON_ONCE(clearing & mask);
1631 * Separate this part out so that a different pointer for clearing_ptr can be
1632 * passed in for disk_clear_events.
1634 static void disk_events_workfn(struct work_struct *work)
1636 struct delayed_work *dwork = to_delayed_work(work);
1637 struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
1639 disk_check_events(ev, &ev->clearing);
1642 static void disk_check_events(struct disk_events *ev,
1643 unsigned int *clearing_ptr)
1645 struct gendisk *disk = ev->disk;
1646 char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
1647 unsigned int clearing = *clearing_ptr;
1648 unsigned int events;
1650 int nr_events = 0, i;
1653 events = disk->fops->check_events(disk, clearing);
1655 /* accumulate pending events and schedule next poll if necessary */
1656 spin_lock_irq(&ev->lock);
1658 events &= ~ev->pending;
1659 ev->pending |= events;
1660 *clearing_ptr &= ~clearing;
1662 intv = disk_events_poll_jiffies(disk);
1663 if (!ev->block && intv)
1664 queue_delayed_work(system_freezable_power_efficient_wq,
1667 spin_unlock_irq(&ev->lock);
1670 * Tell userland about new events. Only the events listed in
1671 * @disk->events are reported. Unlisted events are processed the
1672 * same internally but never get reported to userland.
1674 for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
1675 if (events & disk->events & (1 << i))
1676 envp[nr_events++] = disk_uevents[i];
1679 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1683 * A disk events enabled device has the following sysfs nodes under
1684 * its /sys/block/X/ directory.
1686 * events : list of all supported events
1687 * events_async : list of events which can be detected w/o polling
1688 * events_poll_msecs : polling interval, 0: disable, -1: system default
1690 static ssize_t __disk_events_show(unsigned int events, char *buf)
1692 const char *delim = "";
1696 for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1697 if (events & (1 << i)) {
1698 pos += sprintf(buf + pos, "%s%s",
1699 delim, disk_events_strs[i]);
1703 pos += sprintf(buf + pos, "\n");
1707 static ssize_t disk_events_show(struct device *dev,
1708 struct device_attribute *attr, char *buf)
1710 struct gendisk *disk = dev_to_disk(dev);
1712 return __disk_events_show(disk->events, buf);
1715 static ssize_t disk_events_async_show(struct device *dev,
1716 struct device_attribute *attr, char *buf)
1718 struct gendisk *disk = dev_to_disk(dev);
1720 return __disk_events_show(disk->async_events, buf);
1723 static ssize_t disk_events_poll_msecs_show(struct device *dev,
1724 struct device_attribute *attr,
1727 struct gendisk *disk = dev_to_disk(dev);
1729 return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1732 static ssize_t disk_events_poll_msecs_store(struct device *dev,
1733 struct device_attribute *attr,
1734 const char *buf, size_t count)
1736 struct gendisk *disk = dev_to_disk(dev);
1739 if (!count || !sscanf(buf, "%ld", &intv))
1742 if (intv < 0 && intv != -1)
1745 disk_block_events(disk);
1746 disk->ev->poll_msecs = intv;
1747 __disk_unblock_events(disk, true);
1752 static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL);
1753 static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL);
1754 static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR,
1755 disk_events_poll_msecs_show,
1756 disk_events_poll_msecs_store);
1758 static const struct attribute *disk_events_attrs[] = {
1759 &dev_attr_events.attr,
1760 &dev_attr_events_async.attr,
1761 &dev_attr_events_poll_msecs.attr,
1766 * The default polling interval can be specified by the kernel
1767 * parameter block.events_dfl_poll_msecs which defaults to 0
1768 * (disable). This can also be modified runtime by writing to
1769 * /sys/module/block/events_dfl_poll_msecs.
1771 static int disk_events_set_dfl_poll_msecs(const char *val,
1772 const struct kernel_param *kp)
1774 struct disk_events *ev;
1777 ret = param_set_ulong(val, kp);
1781 mutex_lock(&disk_events_mutex);
1783 list_for_each_entry(ev, &disk_events, node)
1784 disk_flush_events(ev->disk, 0);
1786 mutex_unlock(&disk_events_mutex);
1791 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1792 .set = disk_events_set_dfl_poll_msecs,
1793 .get = param_get_ulong,
1796 #undef MODULE_PARAM_PREFIX
1797 #define MODULE_PARAM_PREFIX "block."
1799 module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
1800 &disk_events_dfl_poll_msecs, 0644);
1803 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
1805 static void disk_alloc_events(struct gendisk *disk)
1807 struct disk_events *ev;
1809 if (!disk->fops->check_events)
1812 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1814 pr_warn("%s: failed to initialize events\n", disk->disk_name);
1818 INIT_LIST_HEAD(&ev->node);
1820 spin_lock_init(&ev->lock);
1821 mutex_init(&ev->block_mutex);
1823 ev->poll_msecs = -1;
1824 INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
1829 static void disk_add_events(struct gendisk *disk)
1834 /* FIXME: error handling */
1835 if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0)
1836 pr_warn("%s: failed to create sysfs files for events\n",
1839 mutex_lock(&disk_events_mutex);
1840 list_add_tail(&disk->ev->node, &disk_events);
1841 mutex_unlock(&disk_events_mutex);
1844 * Block count is initialized to 1 and the following initial
1845 * unblock kicks it into action.
1847 __disk_unblock_events(disk, true);
1850 static void disk_del_events(struct gendisk *disk)
1855 disk_block_events(disk);
1857 mutex_lock(&disk_events_mutex);
1858 list_del_init(&disk->ev->node);
1859 mutex_unlock(&disk_events_mutex);
1861 sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
1864 static void disk_release_events(struct gendisk *disk)
1866 /* the block count should be 1 from disk_del_events() */
1867 WARN_ON_ONCE(disk->ev && disk->ev->block != 1);