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>
23 #include <linux/badblocks.h>
27 static DEFINE_MUTEX(block_class_lock);
28 struct kobject *block_depr;
30 /* for extended dynamic devt allocation, currently only one major is used */
31 #define NR_EXT_DEVT (1 << MINORBITS)
33 /* For extended devt allocation. ext_devt_lock prevents look up
34 * results from going away underneath its user.
36 static DEFINE_SPINLOCK(ext_devt_lock);
37 static DEFINE_IDR(ext_devt_idr);
39 static const struct device_type disk_type;
41 static void disk_check_events(struct disk_events *ev,
42 unsigned int *clearing_ptr);
43 static void disk_alloc_events(struct gendisk *disk);
44 static void disk_add_events(struct gendisk *disk);
45 static void disk_del_events(struct gendisk *disk);
46 static void disk_release_events(struct gendisk *disk);
48 void part_inc_in_flight(struct request_queue *q, struct hd_struct *part, int rw)
53 atomic_inc(&part->in_flight[rw]);
55 atomic_inc(&part_to_disk(part)->part0.in_flight[rw]);
58 void part_dec_in_flight(struct request_queue *q, struct hd_struct *part, int rw)
63 atomic_dec(&part->in_flight[rw]);
65 atomic_dec(&part_to_disk(part)->part0.in_flight[rw]);
68 void part_in_flight(struct request_queue *q, struct hd_struct *part,
69 unsigned int inflight[2])
72 blk_mq_in_flight(q, part, inflight);
76 inflight[0] = atomic_read(&part->in_flight[0]) +
77 atomic_read(&part->in_flight[1]);
79 part = &part_to_disk(part)->part0;
80 inflight[1] = atomic_read(&part->in_flight[0]) +
81 atomic_read(&part->in_flight[1]);
85 void part_in_flight_rw(struct request_queue *q, struct hd_struct *part,
86 unsigned int inflight[2])
89 blk_mq_in_flight_rw(q, part, inflight);
93 inflight[0] = atomic_read(&part->in_flight[0]);
94 inflight[1] = atomic_read(&part->in_flight[1]);
97 struct hd_struct *__disk_get_part(struct gendisk *disk, int partno)
99 struct disk_part_tbl *ptbl = rcu_dereference(disk->part_tbl);
101 if (unlikely(partno < 0 || partno >= ptbl->len))
103 return rcu_dereference(ptbl->part[partno]);
107 * disk_get_part - get partition
108 * @disk: disk to look partition from
109 * @partno: partition number
111 * Look for partition @partno from @disk. If found, increment
112 * reference count and return it.
118 * Pointer to the found partition on success, NULL if not found.
120 struct hd_struct *disk_get_part(struct gendisk *disk, int partno)
122 struct hd_struct *part;
125 part = __disk_get_part(disk, partno);
127 get_device(part_to_dev(part));
132 EXPORT_SYMBOL_GPL(disk_get_part);
135 * disk_part_iter_init - initialize partition iterator
136 * @piter: iterator to initialize
137 * @disk: disk to iterate over
138 * @flags: DISK_PITER_* flags
140 * Initialize @piter so that it iterates over partitions of @disk.
145 void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk,
148 struct disk_part_tbl *ptbl;
151 ptbl = rcu_dereference(disk->part_tbl);
156 if (flags & DISK_PITER_REVERSE)
157 piter->idx = ptbl->len - 1;
158 else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0))
163 piter->flags = flags;
167 EXPORT_SYMBOL_GPL(disk_part_iter_init);
170 * disk_part_iter_next - proceed iterator to the next partition and return it
171 * @piter: iterator of interest
173 * Proceed @piter to the next partition and return it.
178 struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter)
180 struct disk_part_tbl *ptbl;
183 /* put the last partition */
184 disk_put_part(piter->part);
189 ptbl = rcu_dereference(piter->disk->part_tbl);
191 /* determine iteration parameters */
192 if (piter->flags & DISK_PITER_REVERSE) {
194 if (piter->flags & (DISK_PITER_INCL_PART0 |
195 DISK_PITER_INCL_EMPTY_PART0))
204 /* iterate to the next partition */
205 for (; piter->idx != end; piter->idx += inc) {
206 struct hd_struct *part;
208 part = rcu_dereference(ptbl->part[piter->idx]);
211 if (!part_nr_sects_read(part) &&
212 !(piter->flags & DISK_PITER_INCL_EMPTY) &&
213 !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 &&
217 get_device(part_to_dev(part));
227 EXPORT_SYMBOL_GPL(disk_part_iter_next);
230 * disk_part_iter_exit - finish up partition iteration
231 * @piter: iter of interest
233 * Called when iteration is over. Cleans up @piter.
238 void disk_part_iter_exit(struct disk_part_iter *piter)
240 disk_put_part(piter->part);
243 EXPORT_SYMBOL_GPL(disk_part_iter_exit);
245 static inline int sector_in_part(struct hd_struct *part, sector_t sector)
247 return part->start_sect <= sector &&
248 sector < part->start_sect + part_nr_sects_read(part);
252 * disk_map_sector_rcu - map sector to partition
253 * @disk: gendisk of interest
254 * @sector: sector to map
256 * Find out which partition @sector maps to on @disk. This is
257 * primarily used for stats accounting.
260 * RCU read locked. The returned partition pointer is valid only
261 * while preemption is disabled.
264 * Found partition on success, part0 is returned if no partition matches
266 struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector)
268 struct disk_part_tbl *ptbl;
269 struct hd_struct *part;
272 ptbl = rcu_dereference(disk->part_tbl);
274 part = rcu_dereference(ptbl->last_lookup);
275 if (part && sector_in_part(part, sector))
278 for (i = 1; i < ptbl->len; i++) {
279 part = rcu_dereference(ptbl->part[i]);
281 if (part && sector_in_part(part, sector)) {
282 rcu_assign_pointer(ptbl->last_lookup, part);
288 EXPORT_SYMBOL_GPL(disk_map_sector_rcu);
291 * Can be deleted altogether. Later.
294 #define BLKDEV_MAJOR_HASH_SIZE 255
295 static struct blk_major_name {
296 struct blk_major_name *next;
299 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
301 /* index in the above - for now: assume no multimajor ranges */
302 static inline int major_to_index(unsigned major)
304 return major % BLKDEV_MAJOR_HASH_SIZE;
307 #ifdef CONFIG_PROC_FS
308 void blkdev_show(struct seq_file *seqf, off_t offset)
310 struct blk_major_name *dp;
312 mutex_lock(&block_class_lock);
313 for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
314 if (dp->major == offset)
315 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
316 mutex_unlock(&block_class_lock);
318 #endif /* CONFIG_PROC_FS */
321 * register_blkdev - register a new block device
323 * @major: the requested major device number [1..255]. If @major = 0, try to
324 * allocate any unused major number.
325 * @name: the name of the new block device as a zero terminated string
327 * The @name must be unique within the system.
329 * The return value depends on the @major input parameter:
331 * - if a major device number was requested in range [1..255] then the
332 * function returns zero on success, or a negative error code
333 * - if any unused major number was requested with @major = 0 parameter
334 * then the return value is the allocated major number in range
335 * [1..255] or a negative error code otherwise
337 int register_blkdev(unsigned int major, const char *name)
339 struct blk_major_name **n, *p;
342 mutex_lock(&block_class_lock);
346 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
347 if (major_names[index] == NULL)
352 printk("register_blkdev: failed to get major for %s\n",
361 if (major >= BLKDEV_MAJOR_MAX) {
362 pr_err("register_blkdev: major requested (%d) is greater than the maximum (%d) for %s\n",
363 major, BLKDEV_MAJOR_MAX, name);
369 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
376 strlcpy(p->name, name, sizeof(p->name));
378 index = major_to_index(major);
380 for (n = &major_names[index]; *n; n = &(*n)->next) {
381 if ((*n)->major == major)
390 printk("register_blkdev: cannot get major %d for %s\n",
395 mutex_unlock(&block_class_lock);
399 EXPORT_SYMBOL(register_blkdev);
401 void unregister_blkdev(unsigned int major, const char *name)
403 struct blk_major_name **n;
404 struct blk_major_name *p = NULL;
405 int index = major_to_index(major);
407 mutex_lock(&block_class_lock);
408 for (n = &major_names[index]; *n; n = &(*n)->next)
409 if ((*n)->major == major)
411 if (!*n || strcmp((*n)->name, name)) {
417 mutex_unlock(&block_class_lock);
421 EXPORT_SYMBOL(unregister_blkdev);
423 static struct kobj_map *bdev_map;
426 * blk_mangle_minor - scatter minor numbers apart
427 * @minor: minor number to mangle
429 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
430 * is enabled. Mangling twice gives the original value.
438 static int blk_mangle_minor(int minor)
440 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
443 for (i = 0; i < MINORBITS / 2; i++) {
444 int low = minor & (1 << i);
445 int high = minor & (1 << (MINORBITS - 1 - i));
446 int distance = MINORBITS - 1 - 2 * i;
448 minor ^= low | high; /* clear both bits */
449 low <<= distance; /* swap the positions */
451 minor |= low | high; /* and set */
458 * blk_alloc_devt - allocate a dev_t for a partition
459 * @part: partition to allocate dev_t for
460 * @devt: out parameter for resulting dev_t
462 * Allocate a dev_t for block device.
465 * 0 on success, allocated dev_t is returned in *@devt. -errno on
471 int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
473 struct gendisk *disk = part_to_disk(part);
476 /* in consecutive minor range? */
477 if (part->partno < disk->minors) {
478 *devt = MKDEV(disk->major, disk->first_minor + part->partno);
482 /* allocate ext devt */
483 idr_preload(GFP_KERNEL);
485 spin_lock_bh(&ext_devt_lock);
486 idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_NOWAIT);
487 spin_unlock_bh(&ext_devt_lock);
491 return idx == -ENOSPC ? -EBUSY : idx;
493 *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
498 * blk_free_devt - free a dev_t
499 * @devt: dev_t to free
501 * Free @devt which was allocated using blk_alloc_devt().
506 void blk_free_devt(dev_t devt)
508 if (devt == MKDEV(0, 0))
511 if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
512 spin_lock_bh(&ext_devt_lock);
513 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
514 spin_unlock_bh(&ext_devt_lock);
518 static char *bdevt_str(dev_t devt, char *buf)
520 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
521 char tbuf[BDEVT_SIZE];
522 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
523 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
525 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
531 * Register device numbers dev..(dev+range-1)
532 * range must be nonzero
533 * The hash chain is sorted on range, so that subranges can override.
535 void blk_register_region(dev_t devt, unsigned long range, struct module *module,
536 struct kobject *(*probe)(dev_t, int *, void *),
537 int (*lock)(dev_t, void *), void *data)
539 kobj_map(bdev_map, devt, range, module, probe, lock, data);
542 EXPORT_SYMBOL(blk_register_region);
544 void blk_unregister_region(dev_t devt, unsigned long range)
546 kobj_unmap(bdev_map, devt, range);
549 EXPORT_SYMBOL(blk_unregister_region);
551 static struct kobject *exact_match(dev_t devt, int *partno, void *data)
553 struct gendisk *p = data;
555 return &disk_to_dev(p)->kobj;
558 static int exact_lock(dev_t devt, void *data)
560 struct gendisk *p = data;
567 static void register_disk(struct device *parent, struct gendisk *disk)
569 struct device *ddev = disk_to_dev(disk);
570 struct block_device *bdev;
571 struct disk_part_iter piter;
572 struct hd_struct *part;
575 ddev->parent = parent;
577 dev_set_name(ddev, "%s", disk->disk_name);
579 /* delay uevents, until we scanned partition table */
580 dev_set_uevent_suppress(ddev, 1);
582 if (device_add(ddev))
584 if (!sysfs_deprecated) {
585 err = sysfs_create_link(block_depr, &ddev->kobj,
586 kobject_name(&ddev->kobj));
594 * avoid probable deadlock caused by allocating memory with
595 * GFP_KERNEL in runtime_resume callback of its all ancestor
598 pm_runtime_set_memalloc_noio(ddev, true);
600 disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj);
601 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
603 /* No minors to use for partitions */
604 if (!disk_part_scan_enabled(disk))
607 /* No such device (e.g., media were just removed) */
608 if (!get_capacity(disk))
611 bdev = bdget_disk(disk, 0);
615 bdev->bd_invalidated = 1;
616 err = blkdev_get(bdev, FMODE_READ, NULL);
619 blkdev_put(bdev, FMODE_READ);
622 /* announce disk after possible partitions are created */
623 dev_set_uevent_suppress(ddev, 0);
624 kobject_uevent(&ddev->kobj, KOBJ_ADD);
626 /* announce possible partitions */
627 disk_part_iter_init(&piter, disk, 0);
628 while ((part = disk_part_iter_next(&piter)))
629 kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD);
630 disk_part_iter_exit(&piter);
634 * device_add_disk - add partitioning information to kernel list
635 * @parent: parent device for the disk
636 * @disk: per-device partitioning information
638 * This function registers the partitioning information in @disk
641 * FIXME: error handling
643 void device_add_disk(struct device *parent, struct gendisk *disk)
645 struct backing_dev_info *bdi;
649 /* minors == 0 indicates to use ext devt from part0 and should
650 * be accompanied with EXT_DEVT flag. Make sure all
651 * parameters make sense.
653 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
654 WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT));
656 disk->flags |= GENHD_FL_UP;
658 retval = blk_alloc_devt(&disk->part0, &devt);
663 disk_to_dev(disk)->devt = devt;
665 /* ->major and ->first_minor aren't supposed to be
666 * dereferenced from here on, but set them just in case.
668 disk->major = MAJOR(devt);
669 disk->first_minor = MINOR(devt);
671 disk_alloc_events(disk);
673 /* Register BDI before referencing it from bdev */
674 bdi = disk->queue->backing_dev_info;
675 bdi_register_owner(bdi, disk_to_dev(disk));
677 blk_register_region(disk_devt(disk), disk->minors, NULL,
678 exact_match, exact_lock, disk);
679 register_disk(parent, disk);
680 blk_register_queue(disk);
683 * Take an extra ref on queue which will be put on disk_release()
684 * so that it sticks around as long as @disk is there.
686 WARN_ON_ONCE(!blk_get_queue(disk->queue));
688 retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
692 disk_add_events(disk);
693 blk_integrity_add(disk);
695 EXPORT_SYMBOL(device_add_disk);
697 void del_gendisk(struct gendisk *disk)
699 struct disk_part_iter piter;
700 struct hd_struct *part;
702 blk_integrity_del(disk);
703 disk_del_events(disk);
705 /* invalidate stuff */
706 disk_part_iter_init(&piter, disk,
707 DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
708 while ((part = disk_part_iter_next(&piter))) {
709 invalidate_partition(disk, part->partno);
710 bdev_unhash_inode(part_devt(part));
711 delete_partition(disk, part->partno);
713 disk_part_iter_exit(&piter);
715 invalidate_partition(disk, 0);
716 bdev_unhash_inode(disk_devt(disk));
717 set_capacity(disk, 0);
718 disk->flags &= ~GENHD_FL_UP;
720 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
723 * Unregister bdi before releasing device numbers (as they can
724 * get reused and we'd get clashes in sysfs).
726 bdi_unregister(disk->queue->backing_dev_info);
727 blk_unregister_queue(disk);
731 blk_unregister_region(disk_devt(disk), disk->minors);
733 part_stat_set_all(&disk->part0, 0);
734 disk->part0.stamp = 0;
736 kobject_put(disk->part0.holder_dir);
737 kobject_put(disk->slave_dir);
738 if (!sysfs_deprecated)
739 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
740 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
741 device_del(disk_to_dev(disk));
743 EXPORT_SYMBOL(del_gendisk);
745 /* sysfs access to bad-blocks list. */
746 static ssize_t disk_badblocks_show(struct device *dev,
747 struct device_attribute *attr,
750 struct gendisk *disk = dev_to_disk(dev);
753 return sprintf(page, "\n");
755 return badblocks_show(disk->bb, page, 0);
758 static ssize_t disk_badblocks_store(struct device *dev,
759 struct device_attribute *attr,
760 const char *page, size_t len)
762 struct gendisk *disk = dev_to_disk(dev);
767 return badblocks_store(disk->bb, page, len, 0);
771 * get_gendisk - get partitioning information for a given device
772 * @devt: device to get partitioning information for
773 * @partno: returned partition index
775 * This function gets the structure containing partitioning
776 * information for the given device @devt.
778 struct gendisk *get_gendisk(dev_t devt, int *partno)
780 struct gendisk *disk = NULL;
782 if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
783 struct kobject *kobj;
785 kobj = kobj_lookup(bdev_map, devt, partno);
787 disk = dev_to_disk(kobj_to_dev(kobj));
789 struct hd_struct *part;
791 spin_lock_bh(&ext_devt_lock);
792 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
793 if (part && get_disk(part_to_disk(part))) {
794 *partno = part->partno;
795 disk = part_to_disk(part);
797 spin_unlock_bh(&ext_devt_lock);
802 EXPORT_SYMBOL(get_gendisk);
805 * bdget_disk - do bdget() by gendisk and partition number
806 * @disk: gendisk of interest
807 * @partno: partition number
809 * Find partition @partno from @disk, do bdget() on it.
815 * Resulting block_device on success, NULL on failure.
817 struct block_device *bdget_disk(struct gendisk *disk, int partno)
819 struct hd_struct *part;
820 struct block_device *bdev = NULL;
822 part = disk_get_part(disk, partno);
824 bdev = bdget(part_devt(part));
829 EXPORT_SYMBOL(bdget_disk);
832 * print a full list of all partitions - intended for places where the root
833 * filesystem can't be mounted and thus to give the victim some idea of what
836 void __init printk_all_partitions(void)
838 struct class_dev_iter iter;
841 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
842 while ((dev = class_dev_iter_next(&iter))) {
843 struct gendisk *disk = dev_to_disk(dev);
844 struct disk_part_iter piter;
845 struct hd_struct *part;
846 char name_buf[BDEVNAME_SIZE];
847 char devt_buf[BDEVT_SIZE];
850 * Don't show empty devices or things that have been
853 if (get_capacity(disk) == 0 ||
854 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
858 * Note, unlike /proc/partitions, I am showing the
859 * numbers in hex - the same format as the root=
862 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
863 while ((part = disk_part_iter_next(&piter))) {
864 bool is_part0 = part == &disk->part0;
866 printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
867 bdevt_str(part_devt(part), devt_buf),
868 (unsigned long long)part_nr_sects_read(part) >> 1
869 , disk_name(disk, part->partno, name_buf),
870 part->info ? part->info->uuid : "");
872 if (dev->parent && dev->parent->driver)
873 printk(" driver: %s\n",
874 dev->parent->driver->name);
876 printk(" (driver?)\n");
880 disk_part_iter_exit(&piter);
882 class_dev_iter_exit(&iter);
885 #ifdef CONFIG_PROC_FS
887 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
890 struct class_dev_iter *iter;
893 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
895 return ERR_PTR(-ENOMEM);
897 seqf->private = iter;
898 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
900 dev = class_dev_iter_next(iter);
905 return dev_to_disk(dev);
908 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
913 dev = class_dev_iter_next(seqf->private);
915 return dev_to_disk(dev);
920 static void disk_seqf_stop(struct seq_file *seqf, void *v)
922 struct class_dev_iter *iter = seqf->private;
924 /* stop is called even after start failed :-( */
926 class_dev_iter_exit(iter);
928 seqf->private = NULL;
932 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
936 p = disk_seqf_start(seqf, pos);
937 if (!IS_ERR_OR_NULL(p) && !*pos)
938 seq_puts(seqf, "major minor #blocks name\n\n");
942 static int show_partition(struct seq_file *seqf, void *v)
944 struct gendisk *sgp = v;
945 struct disk_part_iter piter;
946 struct hd_struct *part;
947 char buf[BDEVNAME_SIZE];
949 /* Don't show non-partitionable removeable devices or empty devices */
950 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
951 (sgp->flags & GENHD_FL_REMOVABLE)))
953 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
956 /* show the full disk and all non-0 size partitions of it */
957 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
958 while ((part = disk_part_iter_next(&piter)))
959 seq_printf(seqf, "%4d %7d %10llu %s\n",
960 MAJOR(part_devt(part)), MINOR(part_devt(part)),
961 (unsigned long long)part_nr_sects_read(part) >> 1,
962 disk_name(sgp, part->partno, buf));
963 disk_part_iter_exit(&piter);
968 static const struct seq_operations partitions_op = {
969 .start = show_partition_start,
970 .next = disk_seqf_next,
971 .stop = disk_seqf_stop,
972 .show = show_partition
975 static int partitions_open(struct inode *inode, struct file *file)
977 return seq_open(file, &partitions_op);
980 static const struct file_operations proc_partitions_operations = {
981 .open = partitions_open,
984 .release = seq_release,
989 static struct kobject *base_probe(dev_t devt, int *partno, void *data)
991 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
992 /* Make old-style 2.4 aliases work */
993 request_module("block-major-%d", MAJOR(devt));
997 static int __init genhd_device_init(void)
1001 block_class.dev_kobj = sysfs_dev_block_kobj;
1002 error = class_register(&block_class);
1003 if (unlikely(error))
1005 bdev_map = kobj_map_init(base_probe, &block_class_lock);
1008 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
1010 /* create top-level block dir */
1011 if (!sysfs_deprecated)
1012 block_depr = kobject_create_and_add("block", NULL);
1016 subsys_initcall(genhd_device_init);
1018 static ssize_t disk_range_show(struct device *dev,
1019 struct device_attribute *attr, char *buf)
1021 struct gendisk *disk = dev_to_disk(dev);
1023 return sprintf(buf, "%d\n", disk->minors);
1026 static ssize_t disk_ext_range_show(struct device *dev,
1027 struct device_attribute *attr, char *buf)
1029 struct gendisk *disk = dev_to_disk(dev);
1031 return sprintf(buf, "%d\n", disk_max_parts(disk));
1034 static ssize_t disk_removable_show(struct device *dev,
1035 struct device_attribute *attr, char *buf)
1037 struct gendisk *disk = dev_to_disk(dev);
1039 return sprintf(buf, "%d\n",
1040 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
1043 static ssize_t disk_ro_show(struct device *dev,
1044 struct device_attribute *attr, char *buf)
1046 struct gendisk *disk = dev_to_disk(dev);
1048 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
1051 static ssize_t disk_capability_show(struct device *dev,
1052 struct device_attribute *attr, char *buf)
1054 struct gendisk *disk = dev_to_disk(dev);
1056 return sprintf(buf, "%x\n", disk->flags);
1059 static ssize_t disk_alignment_offset_show(struct device *dev,
1060 struct device_attribute *attr,
1063 struct gendisk *disk = dev_to_disk(dev);
1065 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
1068 static ssize_t disk_discard_alignment_show(struct device *dev,
1069 struct device_attribute *attr,
1072 struct gendisk *disk = dev_to_disk(dev);
1074 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
1077 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
1078 static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
1079 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
1080 static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
1081 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
1082 static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL);
1083 static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show,
1085 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
1086 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
1087 static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
1088 static DEVICE_ATTR(badblocks, S_IRUGO | S_IWUSR, disk_badblocks_show,
1089 disk_badblocks_store);
1090 #ifdef CONFIG_FAIL_MAKE_REQUEST
1091 static struct device_attribute dev_attr_fail =
1092 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
1094 #ifdef CONFIG_FAIL_IO_TIMEOUT
1095 static struct device_attribute dev_attr_fail_timeout =
1096 __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show,
1097 part_timeout_store);
1100 static struct attribute *disk_attrs[] = {
1101 &dev_attr_range.attr,
1102 &dev_attr_ext_range.attr,
1103 &dev_attr_removable.attr,
1105 &dev_attr_size.attr,
1106 &dev_attr_alignment_offset.attr,
1107 &dev_attr_discard_alignment.attr,
1108 &dev_attr_capability.attr,
1109 &dev_attr_stat.attr,
1110 &dev_attr_inflight.attr,
1111 &dev_attr_badblocks.attr,
1112 #ifdef CONFIG_FAIL_MAKE_REQUEST
1113 &dev_attr_fail.attr,
1115 #ifdef CONFIG_FAIL_IO_TIMEOUT
1116 &dev_attr_fail_timeout.attr,
1121 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1123 struct device *dev = container_of(kobj, typeof(*dev), kobj);
1124 struct gendisk *disk = dev_to_disk(dev);
1126 if (a == &dev_attr_badblocks.attr && !disk->bb)
1131 static struct attribute_group disk_attr_group = {
1132 .attrs = disk_attrs,
1133 .is_visible = disk_visible,
1136 static const struct attribute_group *disk_attr_groups[] = {
1142 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1143 * @disk: disk to replace part_tbl for
1144 * @new_ptbl: new part_tbl to install
1146 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1147 * original ptbl is freed using RCU callback.
1150 * Matching bd_mutex locked or the caller is the only user of @disk.
1152 static void disk_replace_part_tbl(struct gendisk *disk,
1153 struct disk_part_tbl *new_ptbl)
1155 struct disk_part_tbl *old_ptbl =
1156 rcu_dereference_protected(disk->part_tbl, 1);
1158 rcu_assign_pointer(disk->part_tbl, new_ptbl);
1161 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
1162 kfree_rcu(old_ptbl, rcu_head);
1167 * disk_expand_part_tbl - expand disk->part_tbl
1168 * @disk: disk to expand part_tbl for
1169 * @partno: expand such that this partno can fit in
1171 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1172 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1175 * Matching bd_mutex locked or the caller is the only user of @disk.
1179 * 0 on success, -errno on failure.
1181 int disk_expand_part_tbl(struct gendisk *disk, int partno)
1183 struct disk_part_tbl *old_ptbl =
1184 rcu_dereference_protected(disk->part_tbl, 1);
1185 struct disk_part_tbl *new_ptbl;
1186 int len = old_ptbl ? old_ptbl->len : 0;
1191 * check for int overflow, since we can get here from blkpg_ioctl()
1192 * with a user passed 'partno'.
1194 target = partno + 1;
1198 /* disk_max_parts() is zero during initialization, ignore if so */
1199 if (disk_max_parts(disk) && target > disk_max_parts(disk))
1205 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
1206 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
1210 new_ptbl->len = target;
1212 for (i = 0; i < len; i++)
1213 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
1215 disk_replace_part_tbl(disk, new_ptbl);
1219 static void disk_release(struct device *dev)
1221 struct gendisk *disk = dev_to_disk(dev);
1223 blk_free_devt(dev->devt);
1224 disk_release_events(disk);
1225 kfree(disk->random);
1226 disk_replace_part_tbl(disk, NULL);
1227 hd_free_part(&disk->part0);
1229 blk_put_queue(disk->queue);
1232 struct class block_class = {
1236 static char *block_devnode(struct device *dev, umode_t *mode,
1237 kuid_t *uid, kgid_t *gid)
1239 struct gendisk *disk = dev_to_disk(dev);
1242 return disk->devnode(disk, mode);
1246 static const struct device_type disk_type = {
1248 .groups = disk_attr_groups,
1249 .release = disk_release,
1250 .devnode = block_devnode,
1253 #ifdef CONFIG_PROC_FS
1255 * aggregate disk stat collector. Uses the same stats that the sysfs
1256 * entries do, above, but makes them available through one seq_file.
1258 * The output looks suspiciously like /proc/partitions with a bunch of
1261 static int diskstats_show(struct seq_file *seqf, void *v)
1263 struct gendisk *gp = v;
1264 struct disk_part_iter piter;
1265 struct hd_struct *hd;
1266 char buf[BDEVNAME_SIZE];
1267 unsigned int inflight[2];
1271 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1272 seq_puts(seqf, "major minor name"
1273 " rio rmerge rsect ruse wio wmerge "
1274 "wsect wuse running use aveq"
1278 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
1279 while ((hd = disk_part_iter_next(&piter))) {
1280 cpu = part_stat_lock();
1281 part_round_stats(gp->queue, cpu, hd);
1283 part_in_flight(gp->queue, hd, inflight);
1284 seq_printf(seqf, "%4d %7d %s %lu %lu %lu "
1285 "%u %lu %lu %lu %u %u %u %u\n",
1286 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1287 disk_name(gp, hd->partno, buf),
1288 part_stat_read(hd, ios[READ]),
1289 part_stat_read(hd, merges[READ]),
1290 part_stat_read(hd, sectors[READ]),
1291 jiffies_to_msecs(part_stat_read(hd, ticks[READ])),
1292 part_stat_read(hd, ios[WRITE]),
1293 part_stat_read(hd, merges[WRITE]),
1294 part_stat_read(hd, sectors[WRITE]),
1295 jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])),
1297 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1298 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1301 disk_part_iter_exit(&piter);
1306 static const struct seq_operations diskstats_op = {
1307 .start = disk_seqf_start,
1308 .next = disk_seqf_next,
1309 .stop = disk_seqf_stop,
1310 .show = diskstats_show
1313 static int diskstats_open(struct inode *inode, struct file *file)
1315 return seq_open(file, &diskstats_op);
1318 static const struct file_operations proc_diskstats_operations = {
1319 .open = diskstats_open,
1321 .llseek = seq_lseek,
1322 .release = seq_release,
1325 static int __init proc_genhd_init(void)
1327 proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
1328 proc_create("partitions", 0, NULL, &proc_partitions_operations);
1331 module_init(proc_genhd_init);
1332 #endif /* CONFIG_PROC_FS */
1334 dev_t blk_lookup_devt(const char *name, int partno)
1336 dev_t devt = MKDEV(0, 0);
1337 struct class_dev_iter iter;
1340 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1341 while ((dev = class_dev_iter_next(&iter))) {
1342 struct gendisk *disk = dev_to_disk(dev);
1343 struct hd_struct *part;
1345 if (strcmp(dev_name(dev), name))
1348 if (partno < disk->minors) {
1349 /* We need to return the right devno, even
1350 * if the partition doesn't exist yet.
1352 devt = MKDEV(MAJOR(dev->devt),
1353 MINOR(dev->devt) + partno);
1356 part = disk_get_part(disk, partno);
1358 devt = part_devt(part);
1359 disk_put_part(part);
1362 disk_put_part(part);
1364 class_dev_iter_exit(&iter);
1367 EXPORT_SYMBOL(blk_lookup_devt);
1369 struct gendisk *alloc_disk(int minors)
1371 return alloc_disk_node(minors, NUMA_NO_NODE);
1373 EXPORT_SYMBOL(alloc_disk);
1375 struct gendisk *alloc_disk_node(int minors, int node_id)
1377 struct gendisk *disk;
1378 struct disk_part_tbl *ptbl;
1380 if (minors > DISK_MAX_PARTS) {
1382 "block: can't allocated more than %d partitions\n",
1384 minors = DISK_MAX_PARTS;
1387 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1389 if (!init_part_stats(&disk->part0)) {
1393 disk->node_id = node_id;
1394 if (disk_expand_part_tbl(disk, 0)) {
1395 free_part_stats(&disk->part0);
1399 ptbl = rcu_dereference_protected(disk->part_tbl, 1);
1400 rcu_assign_pointer(ptbl->part[0], &disk->part0);
1403 * set_capacity() and get_capacity() currently don't use
1404 * seqcounter to read/update the part0->nr_sects. Still init
1405 * the counter as we can read the sectors in IO submission
1406 * patch using seqence counters.
1408 * TODO: Ideally set_capacity() and get_capacity() should be
1409 * converted to make use of bd_mutex and sequence counters.
1411 seqcount_init(&disk->part0.nr_sects_seq);
1412 if (hd_ref_init(&disk->part0)) {
1413 hd_free_part(&disk->part0);
1418 disk->minors = minors;
1419 rand_initialize_disk(disk);
1420 disk_to_dev(disk)->class = &block_class;
1421 disk_to_dev(disk)->type = &disk_type;
1422 device_initialize(disk_to_dev(disk));
1426 EXPORT_SYMBOL(alloc_disk_node);
1428 struct kobject *get_disk(struct gendisk *disk)
1430 struct module *owner;
1431 struct kobject *kobj;
1435 owner = disk->fops->owner;
1436 if (owner && !try_module_get(owner))
1438 kobj = kobject_get_unless_zero(&disk_to_dev(disk)->kobj);
1447 EXPORT_SYMBOL(get_disk);
1449 void put_disk(struct gendisk *disk)
1452 kobject_put(&disk_to_dev(disk)->kobj);
1455 EXPORT_SYMBOL(put_disk);
1457 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1459 char event[] = "DISK_RO=1";
1460 char *envp[] = { event, NULL };
1464 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1467 void set_device_ro(struct block_device *bdev, int flag)
1469 bdev->bd_part->policy = flag;
1472 EXPORT_SYMBOL(set_device_ro);
1474 void set_disk_ro(struct gendisk *disk, int flag)
1476 struct disk_part_iter piter;
1477 struct hd_struct *part;
1479 if (disk->part0.policy != flag) {
1480 set_disk_ro_uevent(disk, flag);
1481 disk->part0.policy = flag;
1484 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
1485 while ((part = disk_part_iter_next(&piter)))
1486 part->policy = flag;
1487 disk_part_iter_exit(&piter);
1490 EXPORT_SYMBOL(set_disk_ro);
1492 int bdev_read_only(struct block_device *bdev)
1496 return bdev->bd_part->policy;
1499 EXPORT_SYMBOL(bdev_read_only);
1501 int invalidate_partition(struct gendisk *disk, int partno)
1504 struct block_device *bdev = bdget_disk(disk, partno);
1507 res = __invalidate_device(bdev, true);
1513 EXPORT_SYMBOL(invalidate_partition);
1516 * Disk events - monitor disk events like media change and eject request.
1518 struct disk_events {
1519 struct list_head node; /* all disk_event's */
1520 struct gendisk *disk; /* the associated disk */
1523 struct mutex block_mutex; /* protects blocking */
1524 int block; /* event blocking depth */
1525 unsigned int pending; /* events already sent out */
1526 unsigned int clearing; /* events being cleared */
1528 long poll_msecs; /* interval, -1 for default */
1529 struct delayed_work dwork;
1532 static const char *disk_events_strs[] = {
1533 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1534 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1537 static char *disk_uevents[] = {
1538 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1539 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1542 /* list of all disk_events */
1543 static DEFINE_MUTEX(disk_events_mutex);
1544 static LIST_HEAD(disk_events);
1546 /* disable in-kernel polling by default */
1547 static unsigned long disk_events_dfl_poll_msecs;
1549 static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1551 struct disk_events *ev = disk->ev;
1552 long intv_msecs = 0;
1555 * If device-specific poll interval is set, always use it. If
1556 * the default is being used, poll iff there are events which
1557 * can't be monitored asynchronously.
1559 if (ev->poll_msecs >= 0)
1560 intv_msecs = ev->poll_msecs;
1561 else if (disk->events & ~disk->async_events)
1562 intv_msecs = disk_events_dfl_poll_msecs;
1564 return msecs_to_jiffies(intv_msecs);
1568 * disk_block_events - block and flush disk event checking
1569 * @disk: disk to block events for
1571 * On return from this function, it is guaranteed that event checking
1572 * isn't in progress and won't happen until unblocked by
1573 * disk_unblock_events(). Events blocking is counted and the actual
1574 * unblocking happens after the matching number of unblocks are done.
1576 * Note that this intentionally does not block event checking from
1577 * disk_clear_events().
1582 void disk_block_events(struct gendisk *disk)
1584 struct disk_events *ev = disk->ev;
1585 unsigned long flags;
1592 * Outer mutex ensures that the first blocker completes canceling
1593 * the event work before further blockers are allowed to finish.
1595 mutex_lock(&ev->block_mutex);
1597 spin_lock_irqsave(&ev->lock, flags);
1598 cancel = !ev->block++;
1599 spin_unlock_irqrestore(&ev->lock, flags);
1602 cancel_delayed_work_sync(&disk->ev->dwork);
1604 mutex_unlock(&ev->block_mutex);
1607 static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1609 struct disk_events *ev = disk->ev;
1611 unsigned long flags;
1613 spin_lock_irqsave(&ev->lock, flags);
1615 if (WARN_ON_ONCE(ev->block <= 0))
1621 intv = disk_events_poll_jiffies(disk);
1623 queue_delayed_work(system_freezable_power_efficient_wq,
1626 queue_delayed_work(system_freezable_power_efficient_wq,
1629 spin_unlock_irqrestore(&ev->lock, flags);
1633 * disk_unblock_events - unblock disk event checking
1634 * @disk: disk to unblock events for
1636 * Undo disk_block_events(). When the block count reaches zero, it
1637 * starts events polling if configured.
1640 * Don't care. Safe to call from irq context.
1642 void disk_unblock_events(struct gendisk *disk)
1645 __disk_unblock_events(disk, false);
1649 * disk_flush_events - schedule immediate event checking and flushing
1650 * @disk: disk to check and flush events for
1651 * @mask: events to flush
1653 * Schedule immediate event checking on @disk if not blocked. Events in
1654 * @mask are scheduled to be cleared from the driver. Note that this
1655 * doesn't clear the events from @disk->ev.
1658 * If @mask is non-zero must be called with bdev->bd_mutex held.
1660 void disk_flush_events(struct gendisk *disk, unsigned int mask)
1662 struct disk_events *ev = disk->ev;
1667 spin_lock_irq(&ev->lock);
1668 ev->clearing |= mask;
1670 mod_delayed_work(system_freezable_power_efficient_wq,
1672 spin_unlock_irq(&ev->lock);
1676 * disk_clear_events - synchronously check, clear and return pending events
1677 * @disk: disk to fetch and clear events from
1678 * @mask: mask of events to be fetched and cleared
1680 * Disk events are synchronously checked and pending events in @mask
1681 * are cleared and returned. This ignores the block count.
1686 unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1688 const struct block_device_operations *bdops = disk->fops;
1689 struct disk_events *ev = disk->ev;
1690 unsigned int pending;
1691 unsigned int clearing = mask;
1694 /* for drivers still using the old ->media_changed method */
1695 if ((mask & DISK_EVENT_MEDIA_CHANGE) &&
1696 bdops->media_changed && bdops->media_changed(disk))
1697 return DISK_EVENT_MEDIA_CHANGE;
1701 disk_block_events(disk);
1704 * store the union of mask and ev->clearing on the stack so that the
1705 * race with disk_flush_events does not cause ambiguity (ev->clearing
1706 * can still be modified even if events are blocked).
1708 spin_lock_irq(&ev->lock);
1709 clearing |= ev->clearing;
1711 spin_unlock_irq(&ev->lock);
1713 disk_check_events(ev, &clearing);
1715 * if ev->clearing is not 0, the disk_flush_events got called in the
1716 * middle of this function, so we want to run the workfn without delay.
1718 __disk_unblock_events(disk, ev->clearing ? true : false);
1720 /* then, fetch and clear pending events */
1721 spin_lock_irq(&ev->lock);
1722 pending = ev->pending & mask;
1723 ev->pending &= ~mask;
1724 spin_unlock_irq(&ev->lock);
1725 WARN_ON_ONCE(clearing & mask);
1731 * Separate this part out so that a different pointer for clearing_ptr can be
1732 * passed in for disk_clear_events.
1734 static void disk_events_workfn(struct work_struct *work)
1736 struct delayed_work *dwork = to_delayed_work(work);
1737 struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
1739 disk_check_events(ev, &ev->clearing);
1742 static void disk_check_events(struct disk_events *ev,
1743 unsigned int *clearing_ptr)
1745 struct gendisk *disk = ev->disk;
1746 char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
1747 unsigned int clearing = *clearing_ptr;
1748 unsigned int events;
1750 int nr_events = 0, i;
1753 events = disk->fops->check_events(disk, clearing);
1755 /* accumulate pending events and schedule next poll if necessary */
1756 spin_lock_irq(&ev->lock);
1758 events &= ~ev->pending;
1759 ev->pending |= events;
1760 *clearing_ptr &= ~clearing;
1762 intv = disk_events_poll_jiffies(disk);
1763 if (!ev->block && intv)
1764 queue_delayed_work(system_freezable_power_efficient_wq,
1767 spin_unlock_irq(&ev->lock);
1770 * Tell userland about new events. Only the events listed in
1771 * @disk->events are reported. Unlisted events are processed the
1772 * same internally but never get reported to userland.
1774 for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
1775 if (events & disk->events & (1 << i))
1776 envp[nr_events++] = disk_uevents[i];
1779 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1783 * A disk events enabled device has the following sysfs nodes under
1784 * its /sys/block/X/ directory.
1786 * events : list of all supported events
1787 * events_async : list of events which can be detected w/o polling
1788 * events_poll_msecs : polling interval, 0: disable, -1: system default
1790 static ssize_t __disk_events_show(unsigned int events, char *buf)
1792 const char *delim = "";
1796 for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1797 if (events & (1 << i)) {
1798 pos += sprintf(buf + pos, "%s%s",
1799 delim, disk_events_strs[i]);
1803 pos += sprintf(buf + pos, "\n");
1807 static ssize_t disk_events_show(struct device *dev,
1808 struct device_attribute *attr, char *buf)
1810 struct gendisk *disk = dev_to_disk(dev);
1812 return __disk_events_show(disk->events, buf);
1815 static ssize_t disk_events_async_show(struct device *dev,
1816 struct device_attribute *attr, char *buf)
1818 struct gendisk *disk = dev_to_disk(dev);
1820 return __disk_events_show(disk->async_events, buf);
1823 static ssize_t disk_events_poll_msecs_show(struct device *dev,
1824 struct device_attribute *attr,
1827 struct gendisk *disk = dev_to_disk(dev);
1829 return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1832 static ssize_t disk_events_poll_msecs_store(struct device *dev,
1833 struct device_attribute *attr,
1834 const char *buf, size_t count)
1836 struct gendisk *disk = dev_to_disk(dev);
1839 if (!count || !sscanf(buf, "%ld", &intv))
1842 if (intv < 0 && intv != -1)
1845 disk_block_events(disk);
1846 disk->ev->poll_msecs = intv;
1847 __disk_unblock_events(disk, true);
1852 static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL);
1853 static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL);
1854 static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR,
1855 disk_events_poll_msecs_show,
1856 disk_events_poll_msecs_store);
1858 static const struct attribute *disk_events_attrs[] = {
1859 &dev_attr_events.attr,
1860 &dev_attr_events_async.attr,
1861 &dev_attr_events_poll_msecs.attr,
1866 * The default polling interval can be specified by the kernel
1867 * parameter block.events_dfl_poll_msecs which defaults to 0
1868 * (disable). This can also be modified runtime by writing to
1869 * /sys/module/block/events_dfl_poll_msecs.
1871 static int disk_events_set_dfl_poll_msecs(const char *val,
1872 const struct kernel_param *kp)
1874 struct disk_events *ev;
1877 ret = param_set_ulong(val, kp);
1881 mutex_lock(&disk_events_mutex);
1883 list_for_each_entry(ev, &disk_events, node)
1884 disk_flush_events(ev->disk, 0);
1886 mutex_unlock(&disk_events_mutex);
1891 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1892 .set = disk_events_set_dfl_poll_msecs,
1893 .get = param_get_ulong,
1896 #undef MODULE_PARAM_PREFIX
1897 #define MODULE_PARAM_PREFIX "block."
1899 module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
1900 &disk_events_dfl_poll_msecs, 0644);
1903 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
1905 static void disk_alloc_events(struct gendisk *disk)
1907 struct disk_events *ev;
1909 if (!disk->fops->check_events)
1912 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1914 pr_warn("%s: failed to initialize events\n", disk->disk_name);
1918 INIT_LIST_HEAD(&ev->node);
1920 spin_lock_init(&ev->lock);
1921 mutex_init(&ev->block_mutex);
1923 ev->poll_msecs = -1;
1924 INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
1929 static void disk_add_events(struct gendisk *disk)
1934 /* FIXME: error handling */
1935 if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0)
1936 pr_warn("%s: failed to create sysfs files for events\n",
1939 mutex_lock(&disk_events_mutex);
1940 list_add_tail(&disk->ev->node, &disk_events);
1941 mutex_unlock(&disk_events_mutex);
1944 * Block count is initialized to 1 and the following initial
1945 * unblock kicks it into action.
1947 __disk_unblock_events(disk, true);
1950 static void disk_del_events(struct gendisk *disk)
1955 disk_block_events(disk);
1957 mutex_lock(&disk_events_mutex);
1958 list_del_init(&disk->ev->node);
1959 mutex_unlock(&disk_events_mutex);
1961 sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
1964 static void disk_release_events(struct gendisk *disk)
1966 /* the block count should be 1 from disk_del_events() */
1967 WARN_ON_ONCE(disk->ev && disk->ev->block != 1);