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 get_device(part_to_dev(part));
213 if (!part_nr_sects_read(part) &&
214 !(piter->flags & DISK_PITER_INCL_EMPTY) &&
215 !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 &&
217 put_device(part_to_dev(part));
230 EXPORT_SYMBOL_GPL(disk_part_iter_next);
233 * disk_part_iter_exit - finish up partition iteration
234 * @piter: iter of interest
236 * Called when iteration is over. Cleans up @piter.
241 void disk_part_iter_exit(struct disk_part_iter *piter)
243 disk_put_part(piter->part);
246 EXPORT_SYMBOL_GPL(disk_part_iter_exit);
248 static inline int sector_in_part(struct hd_struct *part, sector_t sector)
250 return part->start_sect <= sector &&
251 sector < part->start_sect + part_nr_sects_read(part);
255 * disk_map_sector_rcu - map sector to partition
256 * @disk: gendisk of interest
257 * @sector: sector to map
259 * Find out which partition @sector maps to on @disk. This is
260 * primarily used for stats accounting.
263 * RCU read locked. The returned partition pointer is valid only
264 * while preemption is disabled.
267 * Found partition on success, part0 is returned if no partition matches
269 struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector)
271 struct disk_part_tbl *ptbl;
272 struct hd_struct *part;
275 ptbl = rcu_dereference(disk->part_tbl);
277 part = rcu_dereference(ptbl->last_lookup);
278 if (part && sector_in_part(part, sector))
281 for (i = 1; i < ptbl->len; i++) {
282 part = rcu_dereference(ptbl->part[i]);
284 if (part && sector_in_part(part, sector)) {
285 rcu_assign_pointer(ptbl->last_lookup, part);
291 EXPORT_SYMBOL_GPL(disk_map_sector_rcu);
294 * Can be deleted altogether. Later.
297 #define BLKDEV_MAJOR_HASH_SIZE 255
298 static struct blk_major_name {
299 struct blk_major_name *next;
302 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
304 /* index in the above - for now: assume no multimajor ranges */
305 static inline int major_to_index(unsigned major)
307 return major % BLKDEV_MAJOR_HASH_SIZE;
310 #ifdef CONFIG_PROC_FS
311 void blkdev_show(struct seq_file *seqf, off_t offset)
313 struct blk_major_name *dp;
315 mutex_lock(&block_class_lock);
316 for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
317 if (dp->major == offset)
318 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
319 mutex_unlock(&block_class_lock);
321 #endif /* CONFIG_PROC_FS */
324 * register_blkdev - register a new block device
326 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
327 * @major = 0, try to allocate any unused major number.
328 * @name: the name of the new block device as a zero terminated string
330 * The @name must be unique within the system.
332 * The return value depends on the @major input parameter:
334 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
335 * then the function returns zero on success, or a negative error code
336 * - if any unused major number was requested with @major = 0 parameter
337 * then the return value is the allocated major number in range
338 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
340 * See Documentation/admin-guide/devices.txt for the list of allocated
343 int register_blkdev(unsigned int major, const char *name)
345 struct blk_major_name **n, *p;
348 mutex_lock(&block_class_lock);
352 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
353 if (major_names[index] == NULL)
358 printk("register_blkdev: failed to get major for %s\n",
367 if (major >= BLKDEV_MAJOR_MAX) {
368 pr_err("register_blkdev: major requested (%u) is greater than the maximum (%u) for %s\n",
369 major, BLKDEV_MAJOR_MAX-1, name);
375 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
382 strlcpy(p->name, name, sizeof(p->name));
384 index = major_to_index(major);
386 for (n = &major_names[index]; *n; n = &(*n)->next) {
387 if ((*n)->major == major)
396 printk("register_blkdev: cannot get major %u for %s\n",
401 mutex_unlock(&block_class_lock);
405 EXPORT_SYMBOL(register_blkdev);
407 void unregister_blkdev(unsigned int major, const char *name)
409 struct blk_major_name **n;
410 struct blk_major_name *p = NULL;
411 int index = major_to_index(major);
413 mutex_lock(&block_class_lock);
414 for (n = &major_names[index]; *n; n = &(*n)->next)
415 if ((*n)->major == major)
417 if (!*n || strcmp((*n)->name, name)) {
423 mutex_unlock(&block_class_lock);
427 EXPORT_SYMBOL(unregister_blkdev);
429 static struct kobj_map *bdev_map;
432 * blk_mangle_minor - scatter minor numbers apart
433 * @minor: minor number to mangle
435 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
436 * is enabled. Mangling twice gives the original value.
444 static int blk_mangle_minor(int minor)
446 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
449 for (i = 0; i < MINORBITS / 2; i++) {
450 int low = minor & (1 << i);
451 int high = minor & (1 << (MINORBITS - 1 - i));
452 int distance = MINORBITS - 1 - 2 * i;
454 minor ^= low | high; /* clear both bits */
455 low <<= distance; /* swap the positions */
457 minor |= low | high; /* and set */
464 * blk_alloc_devt - allocate a dev_t for a partition
465 * @part: partition to allocate dev_t for
466 * @devt: out parameter for resulting dev_t
468 * Allocate a dev_t for block device.
471 * 0 on success, allocated dev_t is returned in *@devt. -errno on
477 int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
479 struct gendisk *disk = part_to_disk(part);
482 /* in consecutive minor range? */
483 if (part->partno < disk->minors) {
484 *devt = MKDEV(disk->major, disk->first_minor + part->partno);
488 /* allocate ext devt */
489 idr_preload(GFP_KERNEL);
491 spin_lock_bh(&ext_devt_lock);
492 idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_NOWAIT);
493 spin_unlock_bh(&ext_devt_lock);
497 return idx == -ENOSPC ? -EBUSY : idx;
499 *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
504 * blk_free_devt - free a dev_t
505 * @devt: dev_t to free
507 * Free @devt which was allocated using blk_alloc_devt().
512 void blk_free_devt(dev_t devt)
514 if (devt == MKDEV(0, 0))
517 if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
518 spin_lock_bh(&ext_devt_lock);
519 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
520 spin_unlock_bh(&ext_devt_lock);
525 * We invalidate devt by assigning NULL pointer for devt in idr.
527 void blk_invalidate_devt(dev_t devt)
529 if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
530 spin_lock_bh(&ext_devt_lock);
531 idr_replace(&ext_devt_idr, NULL, blk_mangle_minor(MINOR(devt)));
532 spin_unlock_bh(&ext_devt_lock);
536 static char *bdevt_str(dev_t devt, char *buf)
538 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
539 char tbuf[BDEVT_SIZE];
540 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
541 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
543 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
549 * Register device numbers dev..(dev+range-1)
550 * range must be nonzero
551 * The hash chain is sorted on range, so that subranges can override.
553 void blk_register_region(dev_t devt, unsigned long range, struct module *module,
554 struct kobject *(*probe)(dev_t, int *, void *),
555 int (*lock)(dev_t, void *), void *data)
557 kobj_map(bdev_map, devt, range, module, probe, lock, data);
560 EXPORT_SYMBOL(blk_register_region);
562 void blk_unregister_region(dev_t devt, unsigned long range)
564 kobj_unmap(bdev_map, devt, range);
567 EXPORT_SYMBOL(blk_unregister_region);
569 static struct kobject *exact_match(dev_t devt, int *partno, void *data)
571 struct gendisk *p = data;
573 return &disk_to_dev(p)->kobj;
576 static int exact_lock(dev_t devt, void *data)
578 struct gendisk *p = data;
580 if (!get_disk_and_module(p))
585 static void register_disk(struct device *parent, struct gendisk *disk,
586 const struct attribute_group **groups)
588 struct device *ddev = disk_to_dev(disk);
589 struct block_device *bdev;
590 struct disk_part_iter piter;
591 struct hd_struct *part;
594 ddev->parent = parent;
596 dev_set_name(ddev, "%s", disk->disk_name);
598 /* delay uevents, until we scanned partition table */
599 dev_set_uevent_suppress(ddev, 1);
602 WARN_ON(ddev->groups);
603 ddev->groups = groups;
605 if (device_add(ddev))
607 if (!sysfs_deprecated) {
608 err = sysfs_create_link(block_depr, &ddev->kobj,
609 kobject_name(&ddev->kobj));
617 * avoid probable deadlock caused by allocating memory with
618 * GFP_KERNEL in runtime_resume callback of its all ancestor
621 pm_runtime_set_memalloc_noio(ddev, true);
623 disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj);
624 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
626 if (disk->flags & GENHD_FL_HIDDEN)
629 /* No minors to use for partitions */
630 if (!disk_part_scan_enabled(disk))
633 /* No such device (e.g., media were just removed) */
634 if (!get_capacity(disk))
637 bdev = bdget_disk(disk, 0);
641 bdev->bd_invalidated = 1;
642 err = blkdev_get(bdev, FMODE_READ, NULL);
645 blkdev_put(bdev, FMODE_READ);
648 /* announce disk after possible partitions are created */
649 dev_set_uevent_suppress(ddev, 0);
650 kobject_uevent(&ddev->kobj, KOBJ_ADD);
652 /* announce possible partitions */
653 disk_part_iter_init(&piter, disk, 0);
654 while ((part = disk_part_iter_next(&piter)))
655 kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD);
656 disk_part_iter_exit(&piter);
658 if (disk->queue->backing_dev_info->dev) {
659 err = sysfs_create_link(&ddev->kobj,
660 &disk->queue->backing_dev_info->dev->kobj,
667 * __device_add_disk - add disk information to kernel list
668 * @parent: parent device for the disk
669 * @disk: per-device partitioning information
670 * @groups: Additional per-device sysfs groups
671 * @register_queue: register the queue if set to true
673 * This function registers the partitioning information in @disk
676 * FIXME: error handling
678 static void __device_add_disk(struct device *parent, struct gendisk *disk,
679 const struct attribute_group **groups,
685 /* minors == 0 indicates to use ext devt from part0 and should
686 * be accompanied with EXT_DEVT flag. Make sure all
687 * parameters make sense.
689 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
690 WARN_ON(!disk->minors &&
691 !(disk->flags & (GENHD_FL_EXT_DEVT | GENHD_FL_HIDDEN)));
693 disk->flags |= GENHD_FL_UP;
695 retval = blk_alloc_devt(&disk->part0, &devt);
700 disk->major = MAJOR(devt);
701 disk->first_minor = MINOR(devt);
703 disk_alloc_events(disk);
705 if (disk->flags & GENHD_FL_HIDDEN) {
707 * Don't let hidden disks show up in /proc/partitions,
708 * and don't bother scanning for partitions either.
710 disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
711 disk->flags |= GENHD_FL_NO_PART_SCAN;
715 /* Register BDI before referencing it from bdev */
716 disk_to_dev(disk)->devt = devt;
717 ret = bdi_register_owner(disk->queue->backing_dev_info,
720 blk_register_region(disk_devt(disk), disk->minors, NULL,
721 exact_match, exact_lock, disk);
723 register_disk(parent, disk, groups);
725 blk_register_queue(disk);
728 * Take an extra ref on queue which will be put on disk_release()
729 * so that it sticks around as long as @disk is there.
731 WARN_ON_ONCE(!blk_get_queue(disk->queue));
733 disk_add_events(disk);
734 blk_integrity_add(disk);
737 void device_add_disk(struct device *parent, struct gendisk *disk,
738 const struct attribute_group **groups)
741 __device_add_disk(parent, disk, groups, true);
743 EXPORT_SYMBOL(device_add_disk);
745 void device_add_disk_no_queue_reg(struct device *parent, struct gendisk *disk)
747 __device_add_disk(parent, disk, NULL, false);
749 EXPORT_SYMBOL(device_add_disk_no_queue_reg);
751 void del_gendisk(struct gendisk *disk)
753 struct disk_part_iter piter;
754 struct hd_struct *part;
756 blk_integrity_del(disk);
757 disk_del_events(disk);
760 * Block lookups of the disk until all bdevs are unhashed and the
761 * disk is marked as dead (GENHD_FL_UP cleared).
763 down_write(&disk->lookup_sem);
764 /* invalidate stuff */
765 disk_part_iter_init(&piter, disk,
766 DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
767 while ((part = disk_part_iter_next(&piter))) {
768 invalidate_partition(disk, part->partno);
769 bdev_unhash_inode(part_devt(part));
770 delete_partition(disk, part->partno);
772 disk_part_iter_exit(&piter);
774 invalidate_partition(disk, 0);
775 bdev_unhash_inode(disk_devt(disk));
776 set_capacity(disk, 0);
777 disk->flags &= ~GENHD_FL_UP;
778 up_write(&disk->lookup_sem);
780 if (!(disk->flags & GENHD_FL_HIDDEN))
781 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
784 * Unregister bdi before releasing device numbers (as they can
785 * get reused and we'd get clashes in sysfs).
787 if (!(disk->flags & GENHD_FL_HIDDEN))
788 bdi_unregister(disk->queue->backing_dev_info);
789 blk_unregister_queue(disk);
794 if (!(disk->flags & GENHD_FL_HIDDEN))
795 blk_unregister_region(disk_devt(disk), disk->minors);
797 * Remove gendisk pointer from idr so that it cannot be looked up
798 * while RCU period before freeing gendisk is running to prevent
799 * use-after-free issues. Note that the device number stays
800 * "in-use" until we really free the gendisk.
802 blk_invalidate_devt(disk_devt(disk));
804 kobject_put(disk->part0.holder_dir);
805 kobject_put(disk->slave_dir);
807 part_stat_set_all(&disk->part0, 0);
808 disk->part0.stamp = 0;
809 if (!sysfs_deprecated)
810 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
811 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
812 device_del(disk_to_dev(disk));
814 EXPORT_SYMBOL(del_gendisk);
816 /* sysfs access to bad-blocks list. */
817 static ssize_t disk_badblocks_show(struct device *dev,
818 struct device_attribute *attr,
821 struct gendisk *disk = dev_to_disk(dev);
824 return sprintf(page, "\n");
826 return badblocks_show(disk->bb, page, 0);
829 static ssize_t disk_badblocks_store(struct device *dev,
830 struct device_attribute *attr,
831 const char *page, size_t len)
833 struct gendisk *disk = dev_to_disk(dev);
838 return badblocks_store(disk->bb, page, len, 0);
842 * get_gendisk - get partitioning information for a given device
843 * @devt: device to get partitioning information for
844 * @partno: returned partition index
846 * This function gets the structure containing partitioning
847 * information for the given device @devt.
849 struct gendisk *get_gendisk(dev_t devt, int *partno)
851 struct gendisk *disk = NULL;
853 if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
854 struct kobject *kobj;
856 kobj = kobj_lookup(bdev_map, devt, partno);
858 disk = dev_to_disk(kobj_to_dev(kobj));
860 struct hd_struct *part;
862 spin_lock_bh(&ext_devt_lock);
863 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
864 if (part && get_disk_and_module(part_to_disk(part))) {
865 *partno = part->partno;
866 disk = part_to_disk(part);
868 spin_unlock_bh(&ext_devt_lock);
875 * Synchronize with del_gendisk() to not return disk that is being
878 down_read(&disk->lookup_sem);
879 if (unlikely((disk->flags & GENHD_FL_HIDDEN) ||
880 !(disk->flags & GENHD_FL_UP))) {
881 up_read(&disk->lookup_sem);
882 put_disk_and_module(disk);
885 up_read(&disk->lookup_sem);
889 EXPORT_SYMBOL(get_gendisk);
892 * bdget_disk - do bdget() by gendisk and partition number
893 * @disk: gendisk of interest
894 * @partno: partition number
896 * Find partition @partno from @disk, do bdget() on it.
902 * Resulting block_device on success, NULL on failure.
904 struct block_device *bdget_disk(struct gendisk *disk, int partno)
906 struct hd_struct *part;
907 struct block_device *bdev = NULL;
909 part = disk_get_part(disk, partno);
911 bdev = bdget(part_devt(part));
916 EXPORT_SYMBOL(bdget_disk);
919 * print a full list of all partitions - intended for places where the root
920 * filesystem can't be mounted and thus to give the victim some idea of what
923 void __init printk_all_partitions(void)
925 struct class_dev_iter iter;
928 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
929 while ((dev = class_dev_iter_next(&iter))) {
930 struct gendisk *disk = dev_to_disk(dev);
931 struct disk_part_iter piter;
932 struct hd_struct *part;
933 char name_buf[BDEVNAME_SIZE];
934 char devt_buf[BDEVT_SIZE];
937 * Don't show empty devices or things that have been
940 if (get_capacity(disk) == 0 ||
941 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
945 * Note, unlike /proc/partitions, I am showing the
946 * numbers in hex - the same format as the root=
949 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
950 while ((part = disk_part_iter_next(&piter))) {
951 bool is_part0 = part == &disk->part0;
953 printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
954 bdevt_str(part_devt(part), devt_buf),
955 (unsigned long long)part_nr_sects_read(part) >> 1
956 , disk_name(disk, part->partno, name_buf),
957 part->info ? part->info->uuid : "");
959 if (dev->parent && dev->parent->driver)
960 printk(" driver: %s\n",
961 dev->parent->driver->name);
963 printk(" (driver?)\n");
967 disk_part_iter_exit(&piter);
969 class_dev_iter_exit(&iter);
972 #ifdef CONFIG_PROC_FS
974 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
977 struct class_dev_iter *iter;
980 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
982 return ERR_PTR(-ENOMEM);
984 seqf->private = iter;
985 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
987 dev = class_dev_iter_next(iter);
992 return dev_to_disk(dev);
995 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
1000 dev = class_dev_iter_next(seqf->private);
1002 return dev_to_disk(dev);
1007 static void disk_seqf_stop(struct seq_file *seqf, void *v)
1009 struct class_dev_iter *iter = seqf->private;
1011 /* stop is called even after start failed :-( */
1013 class_dev_iter_exit(iter);
1015 seqf->private = NULL;
1019 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
1023 p = disk_seqf_start(seqf, pos);
1024 if (!IS_ERR_OR_NULL(p) && !*pos)
1025 seq_puts(seqf, "major minor #blocks name\n\n");
1029 static int show_partition(struct seq_file *seqf, void *v)
1031 struct gendisk *sgp = v;
1032 struct disk_part_iter piter;
1033 struct hd_struct *part;
1034 char buf[BDEVNAME_SIZE];
1036 /* Don't show non-partitionable removeable devices or empty devices */
1037 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
1038 (sgp->flags & GENHD_FL_REMOVABLE)))
1040 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
1043 /* show the full disk and all non-0 size partitions of it */
1044 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
1045 while ((part = disk_part_iter_next(&piter)))
1046 seq_printf(seqf, "%4d %7d %10llu %s\n",
1047 MAJOR(part_devt(part)), MINOR(part_devt(part)),
1048 (unsigned long long)part_nr_sects_read(part) >> 1,
1049 disk_name(sgp, part->partno, buf));
1050 disk_part_iter_exit(&piter);
1055 static const struct seq_operations partitions_op = {
1056 .start = show_partition_start,
1057 .next = disk_seqf_next,
1058 .stop = disk_seqf_stop,
1059 .show = show_partition
1064 static struct kobject *base_probe(dev_t devt, int *partno, void *data)
1066 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
1067 /* Make old-style 2.4 aliases work */
1068 request_module("block-major-%d", MAJOR(devt));
1072 static int __init genhd_device_init(void)
1076 block_class.dev_kobj = sysfs_dev_block_kobj;
1077 error = class_register(&block_class);
1078 if (unlikely(error))
1080 bdev_map = kobj_map_init(base_probe, &block_class_lock);
1083 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
1085 /* create top-level block dir */
1086 if (!sysfs_deprecated)
1087 block_depr = kobject_create_and_add("block", NULL);
1091 subsys_initcall(genhd_device_init);
1093 static ssize_t disk_range_show(struct device *dev,
1094 struct device_attribute *attr, char *buf)
1096 struct gendisk *disk = dev_to_disk(dev);
1098 return sprintf(buf, "%d\n", disk->minors);
1101 static ssize_t disk_ext_range_show(struct device *dev,
1102 struct device_attribute *attr, char *buf)
1104 struct gendisk *disk = dev_to_disk(dev);
1106 return sprintf(buf, "%d\n", disk_max_parts(disk));
1109 static ssize_t disk_removable_show(struct device *dev,
1110 struct device_attribute *attr, char *buf)
1112 struct gendisk *disk = dev_to_disk(dev);
1114 return sprintf(buf, "%d\n",
1115 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
1118 static ssize_t disk_hidden_show(struct device *dev,
1119 struct device_attribute *attr, char *buf)
1121 struct gendisk *disk = dev_to_disk(dev);
1123 return sprintf(buf, "%d\n",
1124 (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
1127 static ssize_t disk_ro_show(struct device *dev,
1128 struct device_attribute *attr, char *buf)
1130 struct gendisk *disk = dev_to_disk(dev);
1132 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
1135 static ssize_t disk_capability_show(struct device *dev,
1136 struct device_attribute *attr, char *buf)
1138 struct gendisk *disk = dev_to_disk(dev);
1140 return sprintf(buf, "%x\n", disk->flags);
1143 static ssize_t disk_alignment_offset_show(struct device *dev,
1144 struct device_attribute *attr,
1147 struct gendisk *disk = dev_to_disk(dev);
1149 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
1152 static ssize_t disk_discard_alignment_show(struct device *dev,
1153 struct device_attribute *attr,
1156 struct gendisk *disk = dev_to_disk(dev);
1158 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
1161 static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
1162 static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
1163 static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
1164 static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
1165 static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
1166 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
1167 static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
1168 static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
1169 static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
1170 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
1171 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
1172 static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
1173 #ifdef CONFIG_FAIL_MAKE_REQUEST
1174 static struct device_attribute dev_attr_fail =
1175 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
1177 #ifdef CONFIG_FAIL_IO_TIMEOUT
1178 static struct device_attribute dev_attr_fail_timeout =
1179 __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
1182 static struct attribute *disk_attrs[] = {
1183 &dev_attr_range.attr,
1184 &dev_attr_ext_range.attr,
1185 &dev_attr_removable.attr,
1186 &dev_attr_hidden.attr,
1188 &dev_attr_size.attr,
1189 &dev_attr_alignment_offset.attr,
1190 &dev_attr_discard_alignment.attr,
1191 &dev_attr_capability.attr,
1192 &dev_attr_stat.attr,
1193 &dev_attr_inflight.attr,
1194 &dev_attr_badblocks.attr,
1195 #ifdef CONFIG_FAIL_MAKE_REQUEST
1196 &dev_attr_fail.attr,
1198 #ifdef CONFIG_FAIL_IO_TIMEOUT
1199 &dev_attr_fail_timeout.attr,
1204 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1206 struct device *dev = container_of(kobj, typeof(*dev), kobj);
1207 struct gendisk *disk = dev_to_disk(dev);
1209 if (a == &dev_attr_badblocks.attr && !disk->bb)
1214 static struct attribute_group disk_attr_group = {
1215 .attrs = disk_attrs,
1216 .is_visible = disk_visible,
1219 static const struct attribute_group *disk_attr_groups[] = {
1225 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1226 * @disk: disk to replace part_tbl for
1227 * @new_ptbl: new part_tbl to install
1229 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1230 * original ptbl is freed using RCU callback.
1233 * Matching bd_mutex locked or the caller is the only user of @disk.
1235 static void disk_replace_part_tbl(struct gendisk *disk,
1236 struct disk_part_tbl *new_ptbl)
1238 struct disk_part_tbl *old_ptbl =
1239 rcu_dereference_protected(disk->part_tbl, 1);
1241 rcu_assign_pointer(disk->part_tbl, new_ptbl);
1244 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
1245 kfree_rcu(old_ptbl, rcu_head);
1250 * disk_expand_part_tbl - expand disk->part_tbl
1251 * @disk: disk to expand part_tbl for
1252 * @partno: expand such that this partno can fit in
1254 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1255 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1258 * Matching bd_mutex locked or the caller is the only user of @disk.
1262 * 0 on success, -errno on failure.
1264 int disk_expand_part_tbl(struct gendisk *disk, int partno)
1266 struct disk_part_tbl *old_ptbl =
1267 rcu_dereference_protected(disk->part_tbl, 1);
1268 struct disk_part_tbl *new_ptbl;
1269 int len = old_ptbl ? old_ptbl->len : 0;
1274 * check for int overflow, since we can get here from blkpg_ioctl()
1275 * with a user passed 'partno'.
1277 target = partno + 1;
1281 /* disk_max_parts() is zero during initialization, ignore if so */
1282 if (disk_max_parts(disk) && target > disk_max_parts(disk))
1288 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
1289 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
1293 new_ptbl->len = target;
1295 for (i = 0; i < len; i++)
1296 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
1298 disk_replace_part_tbl(disk, new_ptbl);
1302 static void disk_release(struct device *dev)
1304 struct gendisk *disk = dev_to_disk(dev);
1306 blk_free_devt(dev->devt);
1307 disk_release_events(disk);
1308 kfree(disk->random);
1309 disk_replace_part_tbl(disk, NULL);
1310 hd_free_part(&disk->part0);
1312 blk_put_queue(disk->queue);
1315 struct class block_class = {
1319 static char *block_devnode(struct device *dev, umode_t *mode,
1320 kuid_t *uid, kgid_t *gid)
1322 struct gendisk *disk = dev_to_disk(dev);
1325 return disk->devnode(disk, mode);
1329 static const struct device_type disk_type = {
1331 .groups = disk_attr_groups,
1332 .release = disk_release,
1333 .devnode = block_devnode,
1336 #ifdef CONFIG_PROC_FS
1338 * aggregate disk stat collector. Uses the same stats that the sysfs
1339 * entries do, above, but makes them available through one seq_file.
1341 * The output looks suspiciously like /proc/partitions with a bunch of
1344 static int diskstats_show(struct seq_file *seqf, void *v)
1346 struct gendisk *gp = v;
1347 struct disk_part_iter piter;
1348 struct hd_struct *hd;
1349 char buf[BDEVNAME_SIZE];
1350 unsigned int inflight[2];
1354 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1355 seq_puts(seqf, "major minor name"
1356 " rio rmerge rsect ruse wio wmerge "
1357 "wsect wuse running use aveq"
1361 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
1362 while ((hd = disk_part_iter_next(&piter))) {
1363 cpu = part_stat_lock();
1364 part_round_stats(gp->queue, cpu, hd);
1366 part_in_flight(gp->queue, hd, inflight);
1367 seq_printf(seqf, "%4d %7d %s "
1372 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1373 disk_name(gp, hd->partno, buf),
1374 part_stat_read(hd, ios[STAT_READ]),
1375 part_stat_read(hd, merges[STAT_READ]),
1376 part_stat_read(hd, sectors[STAT_READ]),
1377 (unsigned int)part_stat_read_msecs(hd, STAT_READ),
1378 part_stat_read(hd, ios[STAT_WRITE]),
1379 part_stat_read(hd, merges[STAT_WRITE]),
1380 part_stat_read(hd, sectors[STAT_WRITE]),
1381 (unsigned int)part_stat_read_msecs(hd, STAT_WRITE),
1383 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1384 jiffies_to_msecs(part_stat_read(hd, time_in_queue)),
1385 part_stat_read(hd, ios[STAT_DISCARD]),
1386 part_stat_read(hd, merges[STAT_DISCARD]),
1387 part_stat_read(hd, sectors[STAT_DISCARD]),
1388 (unsigned int)part_stat_read_msecs(hd, STAT_DISCARD)
1391 disk_part_iter_exit(&piter);
1396 static const struct seq_operations diskstats_op = {
1397 .start = disk_seqf_start,
1398 .next = disk_seqf_next,
1399 .stop = disk_seqf_stop,
1400 .show = diskstats_show
1403 static int __init proc_genhd_init(void)
1405 proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1406 proc_create_seq("partitions", 0, NULL, &partitions_op);
1409 module_init(proc_genhd_init);
1410 #endif /* CONFIG_PROC_FS */
1412 dev_t blk_lookup_devt(const char *name, int partno)
1414 dev_t devt = MKDEV(0, 0);
1415 struct class_dev_iter iter;
1418 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1419 while ((dev = class_dev_iter_next(&iter))) {
1420 struct gendisk *disk = dev_to_disk(dev);
1421 struct hd_struct *part;
1423 if (strcmp(dev_name(dev), name))
1426 if (partno < disk->minors) {
1427 /* We need to return the right devno, even
1428 * if the partition doesn't exist yet.
1430 devt = MKDEV(MAJOR(dev->devt),
1431 MINOR(dev->devt) + partno);
1434 part = disk_get_part(disk, partno);
1436 devt = part_devt(part);
1437 disk_put_part(part);
1440 disk_put_part(part);
1442 class_dev_iter_exit(&iter);
1445 EXPORT_SYMBOL(blk_lookup_devt);
1447 struct gendisk *__alloc_disk_node(int minors, int node_id)
1449 struct gendisk *disk;
1450 struct disk_part_tbl *ptbl;
1452 if (minors > DISK_MAX_PARTS) {
1454 "block: can't allocate more than %d partitions\n",
1456 minors = DISK_MAX_PARTS;
1459 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1461 if (!init_part_stats(&disk->part0)) {
1465 init_rwsem(&disk->lookup_sem);
1466 disk->node_id = node_id;
1467 if (disk_expand_part_tbl(disk, 0)) {
1468 free_part_stats(&disk->part0);
1472 ptbl = rcu_dereference_protected(disk->part_tbl, 1);
1473 rcu_assign_pointer(ptbl->part[0], &disk->part0);
1476 * set_capacity() and get_capacity() currently don't use
1477 * seqcounter to read/update the part0->nr_sects. Still init
1478 * the counter as we can read the sectors in IO submission
1479 * patch using seqence counters.
1481 * TODO: Ideally set_capacity() and get_capacity() should be
1482 * converted to make use of bd_mutex and sequence counters.
1484 seqcount_init(&disk->part0.nr_sects_seq);
1485 if (hd_ref_init(&disk->part0)) {
1486 hd_free_part(&disk->part0);
1491 disk->minors = minors;
1492 rand_initialize_disk(disk);
1493 disk_to_dev(disk)->class = &block_class;
1494 disk_to_dev(disk)->type = &disk_type;
1495 device_initialize(disk_to_dev(disk));
1499 EXPORT_SYMBOL(__alloc_disk_node);
1501 struct kobject *get_disk_and_module(struct gendisk *disk)
1503 struct module *owner;
1504 struct kobject *kobj;
1508 owner = disk->fops->owner;
1509 if (owner && !try_module_get(owner))
1511 kobj = kobject_get_unless_zero(&disk_to_dev(disk)->kobj);
1519 EXPORT_SYMBOL(get_disk_and_module);
1521 void put_disk(struct gendisk *disk)
1524 kobject_put(&disk_to_dev(disk)->kobj);
1526 EXPORT_SYMBOL(put_disk);
1529 * This is a counterpart of get_disk_and_module() and thus also of
1532 void put_disk_and_module(struct gendisk *disk)
1535 struct module *owner = disk->fops->owner;
1541 EXPORT_SYMBOL(put_disk_and_module);
1543 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1545 char event[] = "DISK_RO=1";
1546 char *envp[] = { event, NULL };
1550 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1553 void set_device_ro(struct block_device *bdev, int flag)
1555 bdev->bd_part->policy = flag;
1558 EXPORT_SYMBOL(set_device_ro);
1560 void set_disk_ro(struct gendisk *disk, int flag)
1562 struct disk_part_iter piter;
1563 struct hd_struct *part;
1565 if (disk->part0.policy != flag) {
1566 set_disk_ro_uevent(disk, flag);
1567 disk->part0.policy = flag;
1570 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
1571 while ((part = disk_part_iter_next(&piter)))
1572 part->policy = flag;
1573 disk_part_iter_exit(&piter);
1576 EXPORT_SYMBOL(set_disk_ro);
1578 int bdev_read_only(struct block_device *bdev)
1582 return bdev->bd_part->policy;
1585 EXPORT_SYMBOL(bdev_read_only);
1587 int invalidate_partition(struct gendisk *disk, int partno)
1590 struct block_device *bdev = bdget_disk(disk, partno);
1593 res = __invalidate_device(bdev, true);
1599 EXPORT_SYMBOL(invalidate_partition);
1602 * Disk events - monitor disk events like media change and eject request.
1604 struct disk_events {
1605 struct list_head node; /* all disk_event's */
1606 struct gendisk *disk; /* the associated disk */
1609 struct mutex block_mutex; /* protects blocking */
1610 int block; /* event blocking depth */
1611 unsigned int pending; /* events already sent out */
1612 unsigned int clearing; /* events being cleared */
1614 long poll_msecs; /* interval, -1 for default */
1615 struct delayed_work dwork;
1618 static const char *disk_events_strs[] = {
1619 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1620 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1623 static char *disk_uevents[] = {
1624 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1625 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1628 /* list of all disk_events */
1629 static DEFINE_MUTEX(disk_events_mutex);
1630 static LIST_HEAD(disk_events);
1632 /* disable in-kernel polling by default */
1633 static unsigned long disk_events_dfl_poll_msecs;
1635 static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1637 struct disk_events *ev = disk->ev;
1638 long intv_msecs = 0;
1641 * If device-specific poll interval is set, always use it. If
1642 * the default is being used, poll iff there are events which
1643 * can't be monitored asynchronously.
1645 if (ev->poll_msecs >= 0)
1646 intv_msecs = ev->poll_msecs;
1647 else if (disk->events & ~disk->async_events)
1648 intv_msecs = disk_events_dfl_poll_msecs;
1650 return msecs_to_jiffies(intv_msecs);
1654 * disk_block_events - block and flush disk event checking
1655 * @disk: disk to block events for
1657 * On return from this function, it is guaranteed that event checking
1658 * isn't in progress and won't happen until unblocked by
1659 * disk_unblock_events(). Events blocking is counted and the actual
1660 * unblocking happens after the matching number of unblocks are done.
1662 * Note that this intentionally does not block event checking from
1663 * disk_clear_events().
1668 void disk_block_events(struct gendisk *disk)
1670 struct disk_events *ev = disk->ev;
1671 unsigned long flags;
1678 * Outer mutex ensures that the first blocker completes canceling
1679 * the event work before further blockers are allowed to finish.
1681 mutex_lock(&ev->block_mutex);
1683 spin_lock_irqsave(&ev->lock, flags);
1684 cancel = !ev->block++;
1685 spin_unlock_irqrestore(&ev->lock, flags);
1688 cancel_delayed_work_sync(&disk->ev->dwork);
1690 mutex_unlock(&ev->block_mutex);
1693 static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1695 struct disk_events *ev = disk->ev;
1697 unsigned long flags;
1699 spin_lock_irqsave(&ev->lock, flags);
1701 if (WARN_ON_ONCE(ev->block <= 0))
1707 intv = disk_events_poll_jiffies(disk);
1709 queue_delayed_work(system_freezable_power_efficient_wq,
1712 queue_delayed_work(system_freezable_power_efficient_wq,
1715 spin_unlock_irqrestore(&ev->lock, flags);
1719 * disk_unblock_events - unblock disk event checking
1720 * @disk: disk to unblock events for
1722 * Undo disk_block_events(). When the block count reaches zero, it
1723 * starts events polling if configured.
1726 * Don't care. Safe to call from irq context.
1728 void disk_unblock_events(struct gendisk *disk)
1731 __disk_unblock_events(disk, false);
1735 * disk_flush_events - schedule immediate event checking and flushing
1736 * @disk: disk to check and flush events for
1737 * @mask: events to flush
1739 * Schedule immediate event checking on @disk if not blocked. Events in
1740 * @mask are scheduled to be cleared from the driver. Note that this
1741 * doesn't clear the events from @disk->ev.
1744 * If @mask is non-zero must be called with bdev->bd_mutex held.
1746 void disk_flush_events(struct gendisk *disk, unsigned int mask)
1748 struct disk_events *ev = disk->ev;
1753 spin_lock_irq(&ev->lock);
1754 ev->clearing |= mask;
1756 mod_delayed_work(system_freezable_power_efficient_wq,
1758 spin_unlock_irq(&ev->lock);
1762 * disk_clear_events - synchronously check, clear and return pending events
1763 * @disk: disk to fetch and clear events from
1764 * @mask: mask of events to be fetched and cleared
1766 * Disk events are synchronously checked and pending events in @mask
1767 * are cleared and returned. This ignores the block count.
1772 unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1774 const struct block_device_operations *bdops = disk->fops;
1775 struct disk_events *ev = disk->ev;
1776 unsigned int pending;
1777 unsigned int clearing = mask;
1780 /* for drivers still using the old ->media_changed method */
1781 if ((mask & DISK_EVENT_MEDIA_CHANGE) &&
1782 bdops->media_changed && bdops->media_changed(disk))
1783 return DISK_EVENT_MEDIA_CHANGE;
1787 disk_block_events(disk);
1790 * store the union of mask and ev->clearing on the stack so that the
1791 * race with disk_flush_events does not cause ambiguity (ev->clearing
1792 * can still be modified even if events are blocked).
1794 spin_lock_irq(&ev->lock);
1795 clearing |= ev->clearing;
1797 spin_unlock_irq(&ev->lock);
1799 disk_check_events(ev, &clearing);
1801 * if ev->clearing is not 0, the disk_flush_events got called in the
1802 * middle of this function, so we want to run the workfn without delay.
1804 __disk_unblock_events(disk, ev->clearing ? true : false);
1806 /* then, fetch and clear pending events */
1807 spin_lock_irq(&ev->lock);
1808 pending = ev->pending & mask;
1809 ev->pending &= ~mask;
1810 spin_unlock_irq(&ev->lock);
1811 WARN_ON_ONCE(clearing & mask);
1817 * Separate this part out so that a different pointer for clearing_ptr can be
1818 * passed in for disk_clear_events.
1820 static void disk_events_workfn(struct work_struct *work)
1822 struct delayed_work *dwork = to_delayed_work(work);
1823 struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
1825 disk_check_events(ev, &ev->clearing);
1828 static void disk_check_events(struct disk_events *ev,
1829 unsigned int *clearing_ptr)
1831 struct gendisk *disk = ev->disk;
1832 char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
1833 unsigned int clearing = *clearing_ptr;
1834 unsigned int events;
1836 int nr_events = 0, i;
1839 events = disk->fops->check_events(disk, clearing);
1841 /* accumulate pending events and schedule next poll if necessary */
1842 spin_lock_irq(&ev->lock);
1844 events &= ~ev->pending;
1845 ev->pending |= events;
1846 *clearing_ptr &= ~clearing;
1848 intv = disk_events_poll_jiffies(disk);
1849 if (!ev->block && intv)
1850 queue_delayed_work(system_freezable_power_efficient_wq,
1853 spin_unlock_irq(&ev->lock);
1856 * Tell userland about new events. Only the events listed in
1857 * @disk->events are reported. Unlisted events are processed the
1858 * same internally but never get reported to userland.
1860 for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
1861 if (events & disk->events & (1 << i))
1862 envp[nr_events++] = disk_uevents[i];
1865 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1869 * A disk events enabled device has the following sysfs nodes under
1870 * its /sys/block/X/ directory.
1872 * events : list of all supported events
1873 * events_async : list of events which can be detected w/o polling
1874 * events_poll_msecs : polling interval, 0: disable, -1: system default
1876 static ssize_t __disk_events_show(unsigned int events, char *buf)
1878 const char *delim = "";
1882 for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1883 if (events & (1 << i)) {
1884 pos += sprintf(buf + pos, "%s%s",
1885 delim, disk_events_strs[i]);
1889 pos += sprintf(buf + pos, "\n");
1893 static ssize_t disk_events_show(struct device *dev,
1894 struct device_attribute *attr, char *buf)
1896 struct gendisk *disk = dev_to_disk(dev);
1898 return __disk_events_show(disk->events, buf);
1901 static ssize_t disk_events_async_show(struct device *dev,
1902 struct device_attribute *attr, char *buf)
1904 struct gendisk *disk = dev_to_disk(dev);
1906 return __disk_events_show(disk->async_events, buf);
1909 static ssize_t disk_events_poll_msecs_show(struct device *dev,
1910 struct device_attribute *attr,
1913 struct gendisk *disk = dev_to_disk(dev);
1915 return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1918 static ssize_t disk_events_poll_msecs_store(struct device *dev,
1919 struct device_attribute *attr,
1920 const char *buf, size_t count)
1922 struct gendisk *disk = dev_to_disk(dev);
1925 if (!count || !sscanf(buf, "%ld", &intv))
1928 if (intv < 0 && intv != -1)
1931 disk_block_events(disk);
1932 disk->ev->poll_msecs = intv;
1933 __disk_unblock_events(disk, true);
1938 static const DEVICE_ATTR(events, 0444, disk_events_show, NULL);
1939 static const DEVICE_ATTR(events_async, 0444, disk_events_async_show, NULL);
1940 static const DEVICE_ATTR(events_poll_msecs, 0644,
1941 disk_events_poll_msecs_show,
1942 disk_events_poll_msecs_store);
1944 static const struct attribute *disk_events_attrs[] = {
1945 &dev_attr_events.attr,
1946 &dev_attr_events_async.attr,
1947 &dev_attr_events_poll_msecs.attr,
1952 * The default polling interval can be specified by the kernel
1953 * parameter block.events_dfl_poll_msecs which defaults to 0
1954 * (disable). This can also be modified runtime by writing to
1955 * /sys/module/block/events_dfl_poll_msecs.
1957 static int disk_events_set_dfl_poll_msecs(const char *val,
1958 const struct kernel_param *kp)
1960 struct disk_events *ev;
1963 ret = param_set_ulong(val, kp);
1967 mutex_lock(&disk_events_mutex);
1969 list_for_each_entry(ev, &disk_events, node)
1970 disk_flush_events(ev->disk, 0);
1972 mutex_unlock(&disk_events_mutex);
1977 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1978 .set = disk_events_set_dfl_poll_msecs,
1979 .get = param_get_ulong,
1982 #undef MODULE_PARAM_PREFIX
1983 #define MODULE_PARAM_PREFIX "block."
1985 module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
1986 &disk_events_dfl_poll_msecs, 0644);
1989 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
1991 static void disk_alloc_events(struct gendisk *disk)
1993 struct disk_events *ev;
1995 if (!disk->fops->check_events)
1998 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
2000 pr_warn("%s: failed to initialize events\n", disk->disk_name);
2004 INIT_LIST_HEAD(&ev->node);
2006 spin_lock_init(&ev->lock);
2007 mutex_init(&ev->block_mutex);
2009 ev->poll_msecs = -1;
2010 INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
2015 static void disk_add_events(struct gendisk *disk)
2020 /* FIXME: error handling */
2021 if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0)
2022 pr_warn("%s: failed to create sysfs files for events\n",
2025 mutex_lock(&disk_events_mutex);
2026 list_add_tail(&disk->ev->node, &disk_events);
2027 mutex_unlock(&disk_events_mutex);
2030 * Block count is initialized to 1 and the following initial
2031 * unblock kicks it into action.
2033 __disk_unblock_events(disk, true);
2036 static void disk_del_events(struct gendisk *disk)
2041 disk_block_events(disk);
2043 mutex_lock(&disk_events_mutex);
2044 list_del_init(&disk->ev->node);
2045 mutex_unlock(&disk_events_mutex);
2047 sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
2050 static void disk_release_events(struct gendisk *disk)
2052 /* the block count should be 1 from disk_del_events() */
2053 WARN_ON_ONCE(disk->ev && disk->ev->block != 1);