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..BLKDEV_MAJOR_MAX-1]. If
324 * @major = 0, try to 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..BLKDEV_MAJOR_MAX-1]
332 * then the 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..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
337 * See Documentation/admin-guide/devices.txt for the list of allocated
340 int register_blkdev(unsigned int major, const char *name)
342 struct blk_major_name **n, *p;
345 mutex_lock(&block_class_lock);
349 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
350 if (major_names[index] == NULL)
355 printk("register_blkdev: failed to get major for %s\n",
364 if (major >= BLKDEV_MAJOR_MAX) {
365 pr_err("register_blkdev: major requested (%u) is greater than the maximum (%u) for %s\n",
366 major, BLKDEV_MAJOR_MAX-1, name);
372 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
379 strlcpy(p->name, name, sizeof(p->name));
381 index = major_to_index(major);
383 for (n = &major_names[index]; *n; n = &(*n)->next) {
384 if ((*n)->major == major)
393 printk("register_blkdev: cannot get major %u for %s\n",
398 mutex_unlock(&block_class_lock);
402 EXPORT_SYMBOL(register_blkdev);
404 void unregister_blkdev(unsigned int major, const char *name)
406 struct blk_major_name **n;
407 struct blk_major_name *p = NULL;
408 int index = major_to_index(major);
410 mutex_lock(&block_class_lock);
411 for (n = &major_names[index]; *n; n = &(*n)->next)
412 if ((*n)->major == major)
414 if (!*n || strcmp((*n)->name, name)) {
420 mutex_unlock(&block_class_lock);
424 EXPORT_SYMBOL(unregister_blkdev);
426 static struct kobj_map *bdev_map;
429 * blk_mangle_minor - scatter minor numbers apart
430 * @minor: minor number to mangle
432 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
433 * is enabled. Mangling twice gives the original value.
441 static int blk_mangle_minor(int minor)
443 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
446 for (i = 0; i < MINORBITS / 2; i++) {
447 int low = minor & (1 << i);
448 int high = minor & (1 << (MINORBITS - 1 - i));
449 int distance = MINORBITS - 1 - 2 * i;
451 minor ^= low | high; /* clear both bits */
452 low <<= distance; /* swap the positions */
454 minor |= low | high; /* and set */
461 * blk_alloc_devt - allocate a dev_t for a partition
462 * @part: partition to allocate dev_t for
463 * @devt: out parameter for resulting dev_t
465 * Allocate a dev_t for block device.
468 * 0 on success, allocated dev_t is returned in *@devt. -errno on
474 int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
476 struct gendisk *disk = part_to_disk(part);
479 /* in consecutive minor range? */
480 if (part->partno < disk->minors) {
481 *devt = MKDEV(disk->major, disk->first_minor + part->partno);
485 /* allocate ext devt */
486 idr_preload(GFP_KERNEL);
488 spin_lock_bh(&ext_devt_lock);
489 idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_NOWAIT);
490 spin_unlock_bh(&ext_devt_lock);
494 return idx == -ENOSPC ? -EBUSY : idx;
496 *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
501 * blk_free_devt - free a dev_t
502 * @devt: dev_t to free
504 * Free @devt which was allocated using blk_alloc_devt().
509 void blk_free_devt(dev_t devt)
511 if (devt == MKDEV(0, 0))
514 if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
515 spin_lock_bh(&ext_devt_lock);
516 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
517 spin_unlock_bh(&ext_devt_lock);
521 static char *bdevt_str(dev_t devt, char *buf)
523 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
524 char tbuf[BDEVT_SIZE];
525 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
526 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
528 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
534 * Register device numbers dev..(dev+range-1)
535 * range must be nonzero
536 * The hash chain is sorted on range, so that subranges can override.
538 void blk_register_region(dev_t devt, unsigned long range, struct module *module,
539 struct kobject *(*probe)(dev_t, int *, void *),
540 int (*lock)(dev_t, void *), void *data)
542 kobj_map(bdev_map, devt, range, module, probe, lock, data);
545 EXPORT_SYMBOL(blk_register_region);
547 void blk_unregister_region(dev_t devt, unsigned long range)
549 kobj_unmap(bdev_map, devt, range);
552 EXPORT_SYMBOL(blk_unregister_region);
554 static struct kobject *exact_match(dev_t devt, int *partno, void *data)
556 struct gendisk *p = data;
558 return &disk_to_dev(p)->kobj;
561 static int exact_lock(dev_t devt, void *data)
563 struct gendisk *p = data;
565 if (!get_disk_and_module(p))
570 static void register_disk(struct device *parent, struct gendisk *disk,
571 const struct attribute_group **groups)
573 struct device *ddev = disk_to_dev(disk);
574 struct block_device *bdev;
575 struct disk_part_iter piter;
576 struct hd_struct *part;
579 ddev->parent = parent;
581 dev_set_name(ddev, "%s", disk->disk_name);
583 /* delay uevents, until we scanned partition table */
584 dev_set_uevent_suppress(ddev, 1);
587 WARN_ON(ddev->groups);
588 ddev->groups = groups;
590 if (device_add(ddev))
592 if (!sysfs_deprecated) {
593 err = sysfs_create_link(block_depr, &ddev->kobj,
594 kobject_name(&ddev->kobj));
602 * avoid probable deadlock caused by allocating memory with
603 * GFP_KERNEL in runtime_resume callback of its all ancestor
606 pm_runtime_set_memalloc_noio(ddev, true);
608 disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj);
609 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
611 if (disk->flags & GENHD_FL_HIDDEN) {
612 dev_set_uevent_suppress(ddev, 0);
616 /* No minors to use for partitions */
617 if (!disk_part_scan_enabled(disk))
620 /* No such device (e.g., media were just removed) */
621 if (!get_capacity(disk))
624 bdev = bdget_disk(disk, 0);
628 bdev->bd_invalidated = 1;
629 err = blkdev_get(bdev, FMODE_READ, NULL);
632 blkdev_put(bdev, FMODE_READ);
635 /* announce disk after possible partitions are created */
636 dev_set_uevent_suppress(ddev, 0);
637 kobject_uevent(&ddev->kobj, KOBJ_ADD);
639 /* announce possible partitions */
640 disk_part_iter_init(&piter, disk, 0);
641 while ((part = disk_part_iter_next(&piter)))
642 kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD);
643 disk_part_iter_exit(&piter);
645 err = sysfs_create_link(&ddev->kobj,
646 &disk->queue->backing_dev_info->dev->kobj,
652 * __device_add_disk - add disk information to kernel list
653 * @parent: parent device for the disk
654 * @disk: per-device partitioning information
655 * @groups: Additional per-device sysfs groups
656 * @register_queue: register the queue if set to true
658 * This function registers the partitioning information in @disk
661 * FIXME: error handling
663 static void __device_add_disk(struct device *parent, struct gendisk *disk,
664 const struct attribute_group **groups,
670 /* minors == 0 indicates to use ext devt from part0 and should
671 * be accompanied with EXT_DEVT flag. Make sure all
672 * parameters make sense.
674 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
675 WARN_ON(!disk->minors &&
676 !(disk->flags & (GENHD_FL_EXT_DEVT | GENHD_FL_HIDDEN)));
678 disk->flags |= GENHD_FL_UP;
680 retval = blk_alloc_devt(&disk->part0, &devt);
685 disk->major = MAJOR(devt);
686 disk->first_minor = MINOR(devt);
688 disk_alloc_events(disk);
690 if (disk->flags & GENHD_FL_HIDDEN) {
692 * Don't let hidden disks show up in /proc/partitions,
693 * and don't bother scanning for partitions either.
695 disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
696 disk->flags |= GENHD_FL_NO_PART_SCAN;
700 /* Register BDI before referencing it from bdev */
701 disk_to_dev(disk)->devt = devt;
702 ret = bdi_register_owner(disk->queue->backing_dev_info,
705 blk_register_region(disk_devt(disk), disk->minors, NULL,
706 exact_match, exact_lock, disk);
708 register_disk(parent, disk, groups);
710 blk_register_queue(disk);
713 * Take an extra ref on queue which will be put on disk_release()
714 * so that it sticks around as long as @disk is there.
716 WARN_ON_ONCE(!blk_get_queue(disk->queue));
718 disk_add_events(disk);
719 blk_integrity_add(disk);
722 void device_add_disk(struct device *parent, struct gendisk *disk,
723 const struct attribute_group **groups)
726 __device_add_disk(parent, disk, groups, true);
728 EXPORT_SYMBOL(device_add_disk);
730 void device_add_disk_no_queue_reg(struct device *parent, struct gendisk *disk)
732 __device_add_disk(parent, disk, NULL, false);
734 EXPORT_SYMBOL(device_add_disk_no_queue_reg);
736 void del_gendisk(struct gendisk *disk)
738 struct disk_part_iter piter;
739 struct hd_struct *part;
741 blk_integrity_del(disk);
742 disk_del_events(disk);
745 * Block lookups of the disk until all bdevs are unhashed and the
746 * disk is marked as dead (GENHD_FL_UP cleared).
748 down_write(&disk->lookup_sem);
749 /* invalidate stuff */
750 disk_part_iter_init(&piter, disk,
751 DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
752 while ((part = disk_part_iter_next(&piter))) {
753 invalidate_partition(disk, part->partno);
754 bdev_unhash_inode(part_devt(part));
755 delete_partition(disk, part->partno);
757 disk_part_iter_exit(&piter);
759 invalidate_partition(disk, 0);
760 bdev_unhash_inode(disk_devt(disk));
761 set_capacity(disk, 0);
762 disk->flags &= ~GENHD_FL_UP;
763 up_write(&disk->lookup_sem);
765 if (!(disk->flags & GENHD_FL_HIDDEN))
766 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
769 * Unregister bdi before releasing device numbers (as they can
770 * get reused and we'd get clashes in sysfs).
772 if (!(disk->flags & GENHD_FL_HIDDEN))
773 bdi_unregister(disk->queue->backing_dev_info);
774 blk_unregister_queue(disk);
779 if (!(disk->flags & GENHD_FL_HIDDEN))
780 blk_unregister_region(disk_devt(disk), disk->minors);
782 kobject_put(disk->part0.holder_dir);
783 kobject_put(disk->slave_dir);
785 part_stat_set_all(&disk->part0, 0);
786 disk->part0.stamp = 0;
787 if (!sysfs_deprecated)
788 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
789 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
790 device_del(disk_to_dev(disk));
792 EXPORT_SYMBOL(del_gendisk);
794 /* sysfs access to bad-blocks list. */
795 static ssize_t disk_badblocks_show(struct device *dev,
796 struct device_attribute *attr,
799 struct gendisk *disk = dev_to_disk(dev);
802 return sprintf(page, "\n");
804 return badblocks_show(disk->bb, page, 0);
807 static ssize_t disk_badblocks_store(struct device *dev,
808 struct device_attribute *attr,
809 const char *page, size_t len)
811 struct gendisk *disk = dev_to_disk(dev);
816 return badblocks_store(disk->bb, page, len, 0);
820 * get_gendisk - get partitioning information for a given device
821 * @devt: device to get partitioning information for
822 * @partno: returned partition index
824 * This function gets the structure containing partitioning
825 * information for the given device @devt.
827 struct gendisk *get_gendisk(dev_t devt, int *partno)
829 struct gendisk *disk = NULL;
831 if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
832 struct kobject *kobj;
834 kobj = kobj_lookup(bdev_map, devt, partno);
836 disk = dev_to_disk(kobj_to_dev(kobj));
838 struct hd_struct *part;
840 spin_lock_bh(&ext_devt_lock);
841 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
842 if (part && get_disk_and_module(part_to_disk(part))) {
843 *partno = part->partno;
844 disk = part_to_disk(part);
846 spin_unlock_bh(&ext_devt_lock);
853 * Synchronize with del_gendisk() to not return disk that is being
856 down_read(&disk->lookup_sem);
857 if (unlikely((disk->flags & GENHD_FL_HIDDEN) ||
858 !(disk->flags & GENHD_FL_UP))) {
859 up_read(&disk->lookup_sem);
860 put_disk_and_module(disk);
863 up_read(&disk->lookup_sem);
867 EXPORT_SYMBOL(get_gendisk);
870 * bdget_disk - do bdget() by gendisk and partition number
871 * @disk: gendisk of interest
872 * @partno: partition number
874 * Find partition @partno from @disk, do bdget() on it.
880 * Resulting block_device on success, NULL on failure.
882 struct block_device *bdget_disk(struct gendisk *disk, int partno)
884 struct hd_struct *part;
885 struct block_device *bdev = NULL;
887 part = disk_get_part(disk, partno);
889 bdev = bdget(part_devt(part));
894 EXPORT_SYMBOL(bdget_disk);
897 * print a full list of all partitions - intended for places where the root
898 * filesystem can't be mounted and thus to give the victim some idea of what
901 void __init printk_all_partitions(void)
903 struct class_dev_iter iter;
906 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
907 while ((dev = class_dev_iter_next(&iter))) {
908 struct gendisk *disk = dev_to_disk(dev);
909 struct disk_part_iter piter;
910 struct hd_struct *part;
911 char name_buf[BDEVNAME_SIZE];
912 char devt_buf[BDEVT_SIZE];
915 * Don't show empty devices or things that have been
918 if (get_capacity(disk) == 0 ||
919 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
923 * Note, unlike /proc/partitions, I am showing the
924 * numbers in hex - the same format as the root=
927 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
928 while ((part = disk_part_iter_next(&piter))) {
929 bool is_part0 = part == &disk->part0;
931 printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
932 bdevt_str(part_devt(part), devt_buf),
933 (unsigned long long)part_nr_sects_read(part) >> 1
934 , disk_name(disk, part->partno, name_buf),
935 part->info ? part->info->uuid : "");
937 if (dev->parent && dev->parent->driver)
938 printk(" driver: %s\n",
939 dev->parent->driver->name);
941 printk(" (driver?)\n");
945 disk_part_iter_exit(&piter);
947 class_dev_iter_exit(&iter);
950 #ifdef CONFIG_PROC_FS
952 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
955 struct class_dev_iter *iter;
958 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
960 return ERR_PTR(-ENOMEM);
962 seqf->private = iter;
963 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
965 dev = class_dev_iter_next(iter);
970 return dev_to_disk(dev);
973 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
978 dev = class_dev_iter_next(seqf->private);
980 return dev_to_disk(dev);
985 static void disk_seqf_stop(struct seq_file *seqf, void *v)
987 struct class_dev_iter *iter = seqf->private;
989 /* stop is called even after start failed :-( */
991 class_dev_iter_exit(iter);
993 seqf->private = NULL;
997 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
1001 p = disk_seqf_start(seqf, pos);
1002 if (!IS_ERR_OR_NULL(p) && !*pos)
1003 seq_puts(seqf, "major minor #blocks name\n\n");
1007 static int show_partition(struct seq_file *seqf, void *v)
1009 struct gendisk *sgp = v;
1010 struct disk_part_iter piter;
1011 struct hd_struct *part;
1012 char buf[BDEVNAME_SIZE];
1014 /* Don't show non-partitionable removeable devices or empty devices */
1015 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
1016 (sgp->flags & GENHD_FL_REMOVABLE)))
1018 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
1021 /* show the full disk and all non-0 size partitions of it */
1022 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
1023 while ((part = disk_part_iter_next(&piter)))
1024 seq_printf(seqf, "%4d %7d %10llu %s\n",
1025 MAJOR(part_devt(part)), MINOR(part_devt(part)),
1026 (unsigned long long)part_nr_sects_read(part) >> 1,
1027 disk_name(sgp, part->partno, buf));
1028 disk_part_iter_exit(&piter);
1033 static const struct seq_operations partitions_op = {
1034 .start = show_partition_start,
1035 .next = disk_seqf_next,
1036 .stop = disk_seqf_stop,
1037 .show = show_partition
1042 static struct kobject *base_probe(dev_t devt, int *partno, void *data)
1044 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
1045 /* Make old-style 2.4 aliases work */
1046 request_module("block-major-%d", MAJOR(devt));
1050 static int __init genhd_device_init(void)
1054 block_class.dev_kobj = sysfs_dev_block_kobj;
1055 error = class_register(&block_class);
1056 if (unlikely(error))
1058 bdev_map = kobj_map_init(base_probe, &block_class_lock);
1061 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
1063 /* create top-level block dir */
1064 if (!sysfs_deprecated)
1065 block_depr = kobject_create_and_add("block", NULL);
1069 subsys_initcall(genhd_device_init);
1071 static ssize_t disk_range_show(struct device *dev,
1072 struct device_attribute *attr, char *buf)
1074 struct gendisk *disk = dev_to_disk(dev);
1076 return sprintf(buf, "%d\n", disk->minors);
1079 static ssize_t disk_ext_range_show(struct device *dev,
1080 struct device_attribute *attr, char *buf)
1082 struct gendisk *disk = dev_to_disk(dev);
1084 return sprintf(buf, "%d\n", disk_max_parts(disk));
1087 static ssize_t disk_removable_show(struct device *dev,
1088 struct device_attribute *attr, char *buf)
1090 struct gendisk *disk = dev_to_disk(dev);
1092 return sprintf(buf, "%d\n",
1093 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
1096 static ssize_t disk_hidden_show(struct device *dev,
1097 struct device_attribute *attr, char *buf)
1099 struct gendisk *disk = dev_to_disk(dev);
1101 return sprintf(buf, "%d\n",
1102 (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
1105 static ssize_t disk_ro_show(struct device *dev,
1106 struct device_attribute *attr, char *buf)
1108 struct gendisk *disk = dev_to_disk(dev);
1110 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
1113 static ssize_t disk_capability_show(struct device *dev,
1114 struct device_attribute *attr, char *buf)
1116 struct gendisk *disk = dev_to_disk(dev);
1118 return sprintf(buf, "%x\n", disk->flags);
1121 static ssize_t disk_alignment_offset_show(struct device *dev,
1122 struct device_attribute *attr,
1125 struct gendisk *disk = dev_to_disk(dev);
1127 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
1130 static ssize_t disk_discard_alignment_show(struct device *dev,
1131 struct device_attribute *attr,
1134 struct gendisk *disk = dev_to_disk(dev);
1136 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
1139 static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
1140 static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
1141 static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
1142 static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
1143 static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
1144 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
1145 static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
1146 static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
1147 static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
1148 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
1149 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
1150 static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
1151 #ifdef CONFIG_FAIL_MAKE_REQUEST
1152 static struct device_attribute dev_attr_fail =
1153 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
1155 #ifdef CONFIG_FAIL_IO_TIMEOUT
1156 static struct device_attribute dev_attr_fail_timeout =
1157 __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
1160 static struct attribute *disk_attrs[] = {
1161 &dev_attr_range.attr,
1162 &dev_attr_ext_range.attr,
1163 &dev_attr_removable.attr,
1164 &dev_attr_hidden.attr,
1166 &dev_attr_size.attr,
1167 &dev_attr_alignment_offset.attr,
1168 &dev_attr_discard_alignment.attr,
1169 &dev_attr_capability.attr,
1170 &dev_attr_stat.attr,
1171 &dev_attr_inflight.attr,
1172 &dev_attr_badblocks.attr,
1173 #ifdef CONFIG_FAIL_MAKE_REQUEST
1174 &dev_attr_fail.attr,
1176 #ifdef CONFIG_FAIL_IO_TIMEOUT
1177 &dev_attr_fail_timeout.attr,
1182 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1184 struct device *dev = container_of(kobj, typeof(*dev), kobj);
1185 struct gendisk *disk = dev_to_disk(dev);
1187 if (a == &dev_attr_badblocks.attr && !disk->bb)
1192 static struct attribute_group disk_attr_group = {
1193 .attrs = disk_attrs,
1194 .is_visible = disk_visible,
1197 static const struct attribute_group *disk_attr_groups[] = {
1203 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1204 * @disk: disk to replace part_tbl for
1205 * @new_ptbl: new part_tbl to install
1207 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1208 * original ptbl is freed using RCU callback.
1211 * Matching bd_mutex locked or the caller is the only user of @disk.
1213 static void disk_replace_part_tbl(struct gendisk *disk,
1214 struct disk_part_tbl *new_ptbl)
1216 struct disk_part_tbl *old_ptbl =
1217 rcu_dereference_protected(disk->part_tbl, 1);
1219 rcu_assign_pointer(disk->part_tbl, new_ptbl);
1222 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
1223 kfree_rcu(old_ptbl, rcu_head);
1228 * disk_expand_part_tbl - expand disk->part_tbl
1229 * @disk: disk to expand part_tbl for
1230 * @partno: expand such that this partno can fit in
1232 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1233 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1236 * Matching bd_mutex locked or the caller is the only user of @disk.
1240 * 0 on success, -errno on failure.
1242 int disk_expand_part_tbl(struct gendisk *disk, int partno)
1244 struct disk_part_tbl *old_ptbl =
1245 rcu_dereference_protected(disk->part_tbl, 1);
1246 struct disk_part_tbl *new_ptbl;
1247 int len = old_ptbl ? old_ptbl->len : 0;
1252 * check for int overflow, since we can get here from blkpg_ioctl()
1253 * with a user passed 'partno'.
1255 target = partno + 1;
1259 /* disk_max_parts() is zero during initialization, ignore if so */
1260 if (disk_max_parts(disk) && target > disk_max_parts(disk))
1266 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
1267 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
1271 new_ptbl->len = target;
1273 for (i = 0; i < len; i++)
1274 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
1276 disk_replace_part_tbl(disk, new_ptbl);
1280 static void disk_release(struct device *dev)
1282 struct gendisk *disk = dev_to_disk(dev);
1284 blk_free_devt(dev->devt);
1285 disk_release_events(disk);
1286 kfree(disk->random);
1287 disk_replace_part_tbl(disk, NULL);
1288 hd_free_part(&disk->part0);
1290 blk_put_queue(disk->queue);
1293 struct class block_class = {
1297 static char *block_devnode(struct device *dev, umode_t *mode,
1298 kuid_t *uid, kgid_t *gid)
1300 struct gendisk *disk = dev_to_disk(dev);
1303 return disk->devnode(disk, mode);
1307 static const struct device_type disk_type = {
1309 .groups = disk_attr_groups,
1310 .release = disk_release,
1311 .devnode = block_devnode,
1314 #ifdef CONFIG_PROC_FS
1316 * aggregate disk stat collector. Uses the same stats that the sysfs
1317 * entries do, above, but makes them available through one seq_file.
1319 * The output looks suspiciously like /proc/partitions with a bunch of
1322 static int diskstats_show(struct seq_file *seqf, void *v)
1324 struct gendisk *gp = v;
1325 struct disk_part_iter piter;
1326 struct hd_struct *hd;
1327 char buf[BDEVNAME_SIZE];
1328 unsigned int inflight[2];
1331 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1332 seq_puts(seqf, "major minor name"
1333 " rio rmerge rsect ruse wio wmerge "
1334 "wsect wuse running use aveq"
1338 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
1339 while ((hd = disk_part_iter_next(&piter))) {
1340 part_in_flight(gp->queue, hd, inflight);
1341 seq_printf(seqf, "%4d %7d %s "
1346 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1347 disk_name(gp, hd->partno, buf),
1348 part_stat_read(hd, ios[STAT_READ]),
1349 part_stat_read(hd, merges[STAT_READ]),
1350 part_stat_read(hd, sectors[STAT_READ]),
1351 (unsigned int)part_stat_read_msecs(hd, STAT_READ),
1352 part_stat_read(hd, ios[STAT_WRITE]),
1353 part_stat_read(hd, merges[STAT_WRITE]),
1354 part_stat_read(hd, sectors[STAT_WRITE]),
1355 (unsigned int)part_stat_read_msecs(hd, STAT_WRITE),
1357 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1358 jiffies_to_msecs(part_stat_read(hd, time_in_queue)),
1359 part_stat_read(hd, ios[STAT_DISCARD]),
1360 part_stat_read(hd, merges[STAT_DISCARD]),
1361 part_stat_read(hd, sectors[STAT_DISCARD]),
1362 (unsigned int)part_stat_read_msecs(hd, STAT_DISCARD)
1365 disk_part_iter_exit(&piter);
1370 static const struct seq_operations diskstats_op = {
1371 .start = disk_seqf_start,
1372 .next = disk_seqf_next,
1373 .stop = disk_seqf_stop,
1374 .show = diskstats_show
1377 static int __init proc_genhd_init(void)
1379 proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1380 proc_create_seq("partitions", 0, NULL, &partitions_op);
1383 module_init(proc_genhd_init);
1384 #endif /* CONFIG_PROC_FS */
1386 dev_t blk_lookup_devt(const char *name, int partno)
1388 dev_t devt = MKDEV(0, 0);
1389 struct class_dev_iter iter;
1392 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1393 while ((dev = class_dev_iter_next(&iter))) {
1394 struct gendisk *disk = dev_to_disk(dev);
1395 struct hd_struct *part;
1397 if (strcmp(dev_name(dev), name))
1400 if (partno < disk->minors) {
1401 /* We need to return the right devno, even
1402 * if the partition doesn't exist yet.
1404 devt = MKDEV(MAJOR(dev->devt),
1405 MINOR(dev->devt) + partno);
1408 part = disk_get_part(disk, partno);
1410 devt = part_devt(part);
1411 disk_put_part(part);
1414 disk_put_part(part);
1416 class_dev_iter_exit(&iter);
1419 EXPORT_SYMBOL(blk_lookup_devt);
1421 struct gendisk *__alloc_disk_node(int minors, int node_id)
1423 struct gendisk *disk;
1424 struct disk_part_tbl *ptbl;
1426 if (minors > DISK_MAX_PARTS) {
1428 "block: can't allocate more than %d partitions\n",
1430 minors = DISK_MAX_PARTS;
1433 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1435 if (!init_part_stats(&disk->part0)) {
1439 init_rwsem(&disk->lookup_sem);
1440 disk->node_id = node_id;
1441 if (disk_expand_part_tbl(disk, 0)) {
1442 free_part_stats(&disk->part0);
1446 ptbl = rcu_dereference_protected(disk->part_tbl, 1);
1447 rcu_assign_pointer(ptbl->part[0], &disk->part0);
1450 * set_capacity() and get_capacity() currently don't use
1451 * seqcounter to read/update the part0->nr_sects. Still init
1452 * the counter as we can read the sectors in IO submission
1453 * patch using seqence counters.
1455 * TODO: Ideally set_capacity() and get_capacity() should be
1456 * converted to make use of bd_mutex and sequence counters.
1458 seqcount_init(&disk->part0.nr_sects_seq);
1459 if (hd_ref_init(&disk->part0)) {
1460 hd_free_part(&disk->part0);
1465 disk->minors = minors;
1466 rand_initialize_disk(disk);
1467 disk_to_dev(disk)->class = &block_class;
1468 disk_to_dev(disk)->type = &disk_type;
1469 device_initialize(disk_to_dev(disk));
1473 EXPORT_SYMBOL(__alloc_disk_node);
1475 struct kobject *get_disk_and_module(struct gendisk *disk)
1477 struct module *owner;
1478 struct kobject *kobj;
1482 owner = disk->fops->owner;
1483 if (owner && !try_module_get(owner))
1485 kobj = kobject_get_unless_zero(&disk_to_dev(disk)->kobj);
1493 EXPORT_SYMBOL(get_disk_and_module);
1495 void put_disk(struct gendisk *disk)
1498 kobject_put(&disk_to_dev(disk)->kobj);
1500 EXPORT_SYMBOL(put_disk);
1503 * This is a counterpart of get_disk_and_module() and thus also of
1506 void put_disk_and_module(struct gendisk *disk)
1509 struct module *owner = disk->fops->owner;
1515 EXPORT_SYMBOL(put_disk_and_module);
1517 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1519 char event[] = "DISK_RO=1";
1520 char *envp[] = { event, NULL };
1524 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1527 void set_device_ro(struct block_device *bdev, int flag)
1529 bdev->bd_part->policy = flag;
1532 EXPORT_SYMBOL(set_device_ro);
1534 void set_disk_ro(struct gendisk *disk, int flag)
1536 struct disk_part_iter piter;
1537 struct hd_struct *part;
1539 if (disk->part0.policy != flag) {
1540 set_disk_ro_uevent(disk, flag);
1541 disk->part0.policy = flag;
1544 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
1545 while ((part = disk_part_iter_next(&piter)))
1546 part->policy = flag;
1547 disk_part_iter_exit(&piter);
1550 EXPORT_SYMBOL(set_disk_ro);
1552 int bdev_read_only(struct block_device *bdev)
1556 return bdev->bd_part->policy;
1559 EXPORT_SYMBOL(bdev_read_only);
1561 int invalidate_partition(struct gendisk *disk, int partno)
1564 struct block_device *bdev = bdget_disk(disk, partno);
1567 res = __invalidate_device(bdev, true);
1573 EXPORT_SYMBOL(invalidate_partition);
1576 * Disk events - monitor disk events like media change and eject request.
1578 struct disk_events {
1579 struct list_head node; /* all disk_event's */
1580 struct gendisk *disk; /* the associated disk */
1583 struct mutex block_mutex; /* protects blocking */
1584 int block; /* event blocking depth */
1585 unsigned int pending; /* events already sent out */
1586 unsigned int clearing; /* events being cleared */
1588 long poll_msecs; /* interval, -1 for default */
1589 struct delayed_work dwork;
1592 static const char *disk_events_strs[] = {
1593 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1594 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1597 static char *disk_uevents[] = {
1598 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1599 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1602 /* list of all disk_events */
1603 static DEFINE_MUTEX(disk_events_mutex);
1604 static LIST_HEAD(disk_events);
1606 /* disable in-kernel polling by default */
1607 static unsigned long disk_events_dfl_poll_msecs;
1609 static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1611 struct disk_events *ev = disk->ev;
1612 long intv_msecs = 0;
1615 * If device-specific poll interval is set, always use it. If
1616 * the default is being used, poll iff there are events which
1617 * can't be monitored asynchronously.
1619 if (ev->poll_msecs >= 0)
1620 intv_msecs = ev->poll_msecs;
1621 else if (disk->events & ~disk->async_events)
1622 intv_msecs = disk_events_dfl_poll_msecs;
1624 return msecs_to_jiffies(intv_msecs);
1628 * disk_block_events - block and flush disk event checking
1629 * @disk: disk to block events for
1631 * On return from this function, it is guaranteed that event checking
1632 * isn't in progress and won't happen until unblocked by
1633 * disk_unblock_events(). Events blocking is counted and the actual
1634 * unblocking happens after the matching number of unblocks are done.
1636 * Note that this intentionally does not block event checking from
1637 * disk_clear_events().
1642 void disk_block_events(struct gendisk *disk)
1644 struct disk_events *ev = disk->ev;
1645 unsigned long flags;
1652 * Outer mutex ensures that the first blocker completes canceling
1653 * the event work before further blockers are allowed to finish.
1655 mutex_lock(&ev->block_mutex);
1657 spin_lock_irqsave(&ev->lock, flags);
1658 cancel = !ev->block++;
1659 spin_unlock_irqrestore(&ev->lock, flags);
1662 cancel_delayed_work_sync(&disk->ev->dwork);
1664 mutex_unlock(&ev->block_mutex);
1667 static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1669 struct disk_events *ev = disk->ev;
1671 unsigned long flags;
1673 spin_lock_irqsave(&ev->lock, flags);
1675 if (WARN_ON_ONCE(ev->block <= 0))
1681 intv = disk_events_poll_jiffies(disk);
1683 queue_delayed_work(system_freezable_power_efficient_wq,
1686 queue_delayed_work(system_freezable_power_efficient_wq,
1689 spin_unlock_irqrestore(&ev->lock, flags);
1693 * disk_unblock_events - unblock disk event checking
1694 * @disk: disk to unblock events for
1696 * Undo disk_block_events(). When the block count reaches zero, it
1697 * starts events polling if configured.
1700 * Don't care. Safe to call from irq context.
1702 void disk_unblock_events(struct gendisk *disk)
1705 __disk_unblock_events(disk, false);
1709 * disk_flush_events - schedule immediate event checking and flushing
1710 * @disk: disk to check and flush events for
1711 * @mask: events to flush
1713 * Schedule immediate event checking on @disk if not blocked. Events in
1714 * @mask are scheduled to be cleared from the driver. Note that this
1715 * doesn't clear the events from @disk->ev.
1718 * If @mask is non-zero must be called with bdev->bd_mutex held.
1720 void disk_flush_events(struct gendisk *disk, unsigned int mask)
1722 struct disk_events *ev = disk->ev;
1727 spin_lock_irq(&ev->lock);
1728 ev->clearing |= mask;
1730 mod_delayed_work(system_freezable_power_efficient_wq,
1732 spin_unlock_irq(&ev->lock);
1736 * disk_clear_events - synchronously check, clear and return pending events
1737 * @disk: disk to fetch and clear events from
1738 * @mask: mask of events to be fetched and cleared
1740 * Disk events are synchronously checked and pending events in @mask
1741 * are cleared and returned. This ignores the block count.
1746 unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1748 const struct block_device_operations *bdops = disk->fops;
1749 struct disk_events *ev = disk->ev;
1750 unsigned int pending;
1751 unsigned int clearing = mask;
1754 /* for drivers still using the old ->media_changed method */
1755 if ((mask & DISK_EVENT_MEDIA_CHANGE) &&
1756 bdops->media_changed && bdops->media_changed(disk))
1757 return DISK_EVENT_MEDIA_CHANGE;
1761 disk_block_events(disk);
1764 * store the union of mask and ev->clearing on the stack so that the
1765 * race with disk_flush_events does not cause ambiguity (ev->clearing
1766 * can still be modified even if events are blocked).
1768 spin_lock_irq(&ev->lock);
1769 clearing |= ev->clearing;
1771 spin_unlock_irq(&ev->lock);
1773 disk_check_events(ev, &clearing);
1775 * if ev->clearing is not 0, the disk_flush_events got called in the
1776 * middle of this function, so we want to run the workfn without delay.
1778 __disk_unblock_events(disk, ev->clearing ? true : false);
1780 /* then, fetch and clear pending events */
1781 spin_lock_irq(&ev->lock);
1782 pending = ev->pending & mask;
1783 ev->pending &= ~mask;
1784 spin_unlock_irq(&ev->lock);
1785 WARN_ON_ONCE(clearing & mask);
1791 * Separate this part out so that a different pointer for clearing_ptr can be
1792 * passed in for disk_clear_events.
1794 static void disk_events_workfn(struct work_struct *work)
1796 struct delayed_work *dwork = to_delayed_work(work);
1797 struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
1799 disk_check_events(ev, &ev->clearing);
1802 static void disk_check_events(struct disk_events *ev,
1803 unsigned int *clearing_ptr)
1805 struct gendisk *disk = ev->disk;
1806 char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
1807 unsigned int clearing = *clearing_ptr;
1808 unsigned int events;
1810 int nr_events = 0, i;
1813 events = disk->fops->check_events(disk, clearing);
1815 /* accumulate pending events and schedule next poll if necessary */
1816 spin_lock_irq(&ev->lock);
1818 events &= ~ev->pending;
1819 ev->pending |= events;
1820 *clearing_ptr &= ~clearing;
1822 intv = disk_events_poll_jiffies(disk);
1823 if (!ev->block && intv)
1824 queue_delayed_work(system_freezable_power_efficient_wq,
1827 spin_unlock_irq(&ev->lock);
1830 * Tell userland about new events. Only the events listed in
1831 * @disk->events are reported. Unlisted events are processed the
1832 * same internally but never get reported to userland.
1834 for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
1835 if (events & disk->events & (1 << i))
1836 envp[nr_events++] = disk_uevents[i];
1839 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1843 * A disk events enabled device has the following sysfs nodes under
1844 * its /sys/block/X/ directory.
1846 * events : list of all supported events
1847 * events_async : list of events which can be detected w/o polling
1848 * events_poll_msecs : polling interval, 0: disable, -1: system default
1850 static ssize_t __disk_events_show(unsigned int events, char *buf)
1852 const char *delim = "";
1856 for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1857 if (events & (1 << i)) {
1858 pos += sprintf(buf + pos, "%s%s",
1859 delim, disk_events_strs[i]);
1863 pos += sprintf(buf + pos, "\n");
1867 static ssize_t disk_events_show(struct device *dev,
1868 struct device_attribute *attr, char *buf)
1870 struct gendisk *disk = dev_to_disk(dev);
1872 return __disk_events_show(disk->events, buf);
1875 static ssize_t disk_events_async_show(struct device *dev,
1876 struct device_attribute *attr, char *buf)
1878 struct gendisk *disk = dev_to_disk(dev);
1880 return __disk_events_show(disk->async_events, buf);
1883 static ssize_t disk_events_poll_msecs_show(struct device *dev,
1884 struct device_attribute *attr,
1887 struct gendisk *disk = dev_to_disk(dev);
1889 return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1892 static ssize_t disk_events_poll_msecs_store(struct device *dev,
1893 struct device_attribute *attr,
1894 const char *buf, size_t count)
1896 struct gendisk *disk = dev_to_disk(dev);
1899 if (!count || !sscanf(buf, "%ld", &intv))
1902 if (intv < 0 && intv != -1)
1905 disk_block_events(disk);
1906 disk->ev->poll_msecs = intv;
1907 __disk_unblock_events(disk, true);
1912 static const DEVICE_ATTR(events, 0444, disk_events_show, NULL);
1913 static const DEVICE_ATTR(events_async, 0444, disk_events_async_show, NULL);
1914 static const DEVICE_ATTR(events_poll_msecs, 0644,
1915 disk_events_poll_msecs_show,
1916 disk_events_poll_msecs_store);
1918 static const struct attribute *disk_events_attrs[] = {
1919 &dev_attr_events.attr,
1920 &dev_attr_events_async.attr,
1921 &dev_attr_events_poll_msecs.attr,
1926 * The default polling interval can be specified by the kernel
1927 * parameter block.events_dfl_poll_msecs which defaults to 0
1928 * (disable). This can also be modified runtime by writing to
1929 * /sys/module/block/events_dfl_poll_msecs.
1931 static int disk_events_set_dfl_poll_msecs(const char *val,
1932 const struct kernel_param *kp)
1934 struct disk_events *ev;
1937 ret = param_set_ulong(val, kp);
1941 mutex_lock(&disk_events_mutex);
1943 list_for_each_entry(ev, &disk_events, node)
1944 disk_flush_events(ev->disk, 0);
1946 mutex_unlock(&disk_events_mutex);
1951 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1952 .set = disk_events_set_dfl_poll_msecs,
1953 .get = param_get_ulong,
1956 #undef MODULE_PARAM_PREFIX
1957 #define MODULE_PARAM_PREFIX "block."
1959 module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
1960 &disk_events_dfl_poll_msecs, 0644);
1963 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
1965 static void disk_alloc_events(struct gendisk *disk)
1967 struct disk_events *ev;
1969 if (!disk->fops->check_events)
1972 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1974 pr_warn("%s: failed to initialize events\n", disk->disk_name);
1978 INIT_LIST_HEAD(&ev->node);
1980 spin_lock_init(&ev->lock);
1981 mutex_init(&ev->block_mutex);
1983 ev->poll_msecs = -1;
1984 INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
1989 static void disk_add_events(struct gendisk *disk)
1994 /* FIXME: error handling */
1995 if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0)
1996 pr_warn("%s: failed to create sysfs files for events\n",
1999 mutex_lock(&disk_events_mutex);
2000 list_add_tail(&disk->ev->node, &disk_events);
2001 mutex_unlock(&disk_events_mutex);
2004 * Block count is initialized to 1 and the following initial
2005 * unblock kicks it into action.
2007 __disk_unblock_events(disk, true);
2010 static void disk_del_events(struct gendisk *disk)
2015 disk_block_events(disk);
2017 mutex_lock(&disk_events_mutex);
2018 list_del_init(&disk->ev->node);
2019 mutex_unlock(&disk_events_mutex);
2021 sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
2024 static void disk_release_events(struct gendisk *disk)
2026 /* the block count should be 1 from disk_del_events() */
2027 WARN_ON_ONCE(disk->ev && disk->ev->block != 1);