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)
572 struct device *ddev = disk_to_dev(disk);
573 struct block_device *bdev;
574 struct disk_part_iter piter;
575 struct hd_struct *part;
578 ddev->parent = parent;
580 dev_set_name(ddev, "%s", disk->disk_name);
582 /* delay uevents, until we scanned partition table */
583 dev_set_uevent_suppress(ddev, 1);
585 if (device_add(ddev))
587 if (!sysfs_deprecated) {
588 err = sysfs_create_link(block_depr, &ddev->kobj,
589 kobject_name(&ddev->kobj));
597 * avoid probable deadlock caused by allocating memory with
598 * GFP_KERNEL in runtime_resume callback of its all ancestor
601 pm_runtime_set_memalloc_noio(ddev, true);
603 disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj);
604 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
606 if (disk->flags & GENHD_FL_HIDDEN) {
607 dev_set_uevent_suppress(ddev, 0);
611 /* No minors to use for partitions */
612 if (!disk_part_scan_enabled(disk))
615 /* No such device (e.g., media were just removed) */
616 if (!get_capacity(disk))
619 bdev = bdget_disk(disk, 0);
623 bdev->bd_invalidated = 1;
624 err = blkdev_get(bdev, FMODE_READ, NULL);
627 blkdev_put(bdev, FMODE_READ);
630 /* announce disk after possible partitions are created */
631 dev_set_uevent_suppress(ddev, 0);
632 kobject_uevent(&ddev->kobj, KOBJ_ADD);
634 /* announce possible partitions */
635 disk_part_iter_init(&piter, disk, 0);
636 while ((part = disk_part_iter_next(&piter)))
637 kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD);
638 disk_part_iter_exit(&piter);
640 err = sysfs_create_link(&ddev->kobj,
641 &disk->queue->backing_dev_info->dev->kobj,
647 * __device_add_disk - add disk information to kernel list
648 * @parent: parent device for the disk
649 * @disk: per-device partitioning information
650 * @register_queue: register the queue if set to true
652 * This function registers the partitioning information in @disk
655 * FIXME: error handling
657 static void __device_add_disk(struct device *parent, struct gendisk *disk,
663 /* minors == 0 indicates to use ext devt from part0 and should
664 * be accompanied with EXT_DEVT flag. Make sure all
665 * parameters make sense.
667 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
668 WARN_ON(!disk->minors &&
669 !(disk->flags & (GENHD_FL_EXT_DEVT | GENHD_FL_HIDDEN)));
671 disk->flags |= GENHD_FL_UP;
673 retval = blk_alloc_devt(&disk->part0, &devt);
678 disk->major = MAJOR(devt);
679 disk->first_minor = MINOR(devt);
681 disk_alloc_events(disk);
683 if (disk->flags & GENHD_FL_HIDDEN) {
685 * Don't let hidden disks show up in /proc/partitions,
686 * and don't bother scanning for partitions either.
688 disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
689 disk->flags |= GENHD_FL_NO_PART_SCAN;
693 /* Register BDI before referencing it from bdev */
694 disk_to_dev(disk)->devt = devt;
695 ret = bdi_register_owner(disk->queue->backing_dev_info,
698 blk_register_region(disk_devt(disk), disk->minors, NULL,
699 exact_match, exact_lock, disk);
701 register_disk(parent, disk);
703 blk_register_queue(disk);
706 * Take an extra ref on queue which will be put on disk_release()
707 * so that it sticks around as long as @disk is there.
709 WARN_ON_ONCE(!blk_get_queue(disk->queue));
711 disk_add_events(disk);
712 blk_integrity_add(disk);
715 void device_add_disk(struct device *parent, struct gendisk *disk)
717 __device_add_disk(parent, disk, true);
719 EXPORT_SYMBOL(device_add_disk);
721 void device_add_disk_no_queue_reg(struct device *parent, struct gendisk *disk)
723 __device_add_disk(parent, disk, false);
725 EXPORT_SYMBOL(device_add_disk_no_queue_reg);
727 void del_gendisk(struct gendisk *disk)
729 struct disk_part_iter piter;
730 struct hd_struct *part;
732 blk_integrity_del(disk);
733 disk_del_events(disk);
736 * Block lookups of the disk until all bdevs are unhashed and the
737 * disk is marked as dead (GENHD_FL_UP cleared).
739 down_write(&disk->lookup_sem);
740 /* invalidate stuff */
741 disk_part_iter_init(&piter, disk,
742 DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
743 while ((part = disk_part_iter_next(&piter))) {
744 invalidate_partition(disk, part->partno);
745 bdev_unhash_inode(part_devt(part));
746 delete_partition(disk, part->partno);
748 disk_part_iter_exit(&piter);
750 invalidate_partition(disk, 0);
751 bdev_unhash_inode(disk_devt(disk));
752 set_capacity(disk, 0);
753 disk->flags &= ~GENHD_FL_UP;
754 up_write(&disk->lookup_sem);
756 if (!(disk->flags & GENHD_FL_HIDDEN))
757 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
760 * Unregister bdi before releasing device numbers (as they can
761 * get reused and we'd get clashes in sysfs).
763 if (!(disk->flags & GENHD_FL_HIDDEN))
764 bdi_unregister(disk->queue->backing_dev_info);
765 blk_unregister_queue(disk);
770 if (!(disk->flags & GENHD_FL_HIDDEN))
771 blk_unregister_region(disk_devt(disk), disk->minors);
773 kobject_put(disk->part0.holder_dir);
774 kobject_put(disk->slave_dir);
776 part_stat_set_all(&disk->part0, 0);
777 disk->part0.stamp = 0;
778 if (!sysfs_deprecated)
779 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
780 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
781 device_del(disk_to_dev(disk));
783 EXPORT_SYMBOL(del_gendisk);
785 /* sysfs access to bad-blocks list. */
786 static ssize_t disk_badblocks_show(struct device *dev,
787 struct device_attribute *attr,
790 struct gendisk *disk = dev_to_disk(dev);
793 return sprintf(page, "\n");
795 return badblocks_show(disk->bb, page, 0);
798 static ssize_t disk_badblocks_store(struct device *dev,
799 struct device_attribute *attr,
800 const char *page, size_t len)
802 struct gendisk *disk = dev_to_disk(dev);
807 return badblocks_store(disk->bb, page, len, 0);
811 * get_gendisk - get partitioning information for a given device
812 * @devt: device to get partitioning information for
813 * @partno: returned partition index
815 * This function gets the structure containing partitioning
816 * information for the given device @devt.
818 struct gendisk *get_gendisk(dev_t devt, int *partno)
820 struct gendisk *disk = NULL;
822 if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
823 struct kobject *kobj;
825 kobj = kobj_lookup(bdev_map, devt, partno);
827 disk = dev_to_disk(kobj_to_dev(kobj));
829 struct hd_struct *part;
831 spin_lock_bh(&ext_devt_lock);
832 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
833 if (part && get_disk_and_module(part_to_disk(part))) {
834 *partno = part->partno;
835 disk = part_to_disk(part);
837 spin_unlock_bh(&ext_devt_lock);
844 * Synchronize with del_gendisk() to not return disk that is being
847 down_read(&disk->lookup_sem);
848 if (unlikely((disk->flags & GENHD_FL_HIDDEN) ||
849 !(disk->flags & GENHD_FL_UP))) {
850 up_read(&disk->lookup_sem);
851 put_disk_and_module(disk);
854 up_read(&disk->lookup_sem);
858 EXPORT_SYMBOL(get_gendisk);
861 * bdget_disk - do bdget() by gendisk and partition number
862 * @disk: gendisk of interest
863 * @partno: partition number
865 * Find partition @partno from @disk, do bdget() on it.
871 * Resulting block_device on success, NULL on failure.
873 struct block_device *bdget_disk(struct gendisk *disk, int partno)
875 struct hd_struct *part;
876 struct block_device *bdev = NULL;
878 part = disk_get_part(disk, partno);
880 bdev = bdget(part_devt(part));
885 EXPORT_SYMBOL(bdget_disk);
888 * print a full list of all partitions - intended for places where the root
889 * filesystem can't be mounted and thus to give the victim some idea of what
892 void __init printk_all_partitions(void)
894 struct class_dev_iter iter;
897 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
898 while ((dev = class_dev_iter_next(&iter))) {
899 struct gendisk *disk = dev_to_disk(dev);
900 struct disk_part_iter piter;
901 struct hd_struct *part;
902 char name_buf[BDEVNAME_SIZE];
903 char devt_buf[BDEVT_SIZE];
906 * Don't show empty devices or things that have been
909 if (get_capacity(disk) == 0 ||
910 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
914 * Note, unlike /proc/partitions, I am showing the
915 * numbers in hex - the same format as the root=
918 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
919 while ((part = disk_part_iter_next(&piter))) {
920 bool is_part0 = part == &disk->part0;
922 printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
923 bdevt_str(part_devt(part), devt_buf),
924 (unsigned long long)part_nr_sects_read(part) >> 1
925 , disk_name(disk, part->partno, name_buf),
926 part->info ? part->info->uuid : "");
928 if (dev->parent && dev->parent->driver)
929 printk(" driver: %s\n",
930 dev->parent->driver->name);
932 printk(" (driver?)\n");
936 disk_part_iter_exit(&piter);
938 class_dev_iter_exit(&iter);
941 #ifdef CONFIG_PROC_FS
943 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
946 struct class_dev_iter *iter;
949 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
951 return ERR_PTR(-ENOMEM);
953 seqf->private = iter;
954 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
956 dev = class_dev_iter_next(iter);
961 return dev_to_disk(dev);
964 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
969 dev = class_dev_iter_next(seqf->private);
971 return dev_to_disk(dev);
976 static void disk_seqf_stop(struct seq_file *seqf, void *v)
978 struct class_dev_iter *iter = seqf->private;
980 /* stop is called even after start failed :-( */
982 class_dev_iter_exit(iter);
984 seqf->private = NULL;
988 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
992 p = disk_seqf_start(seqf, pos);
993 if (!IS_ERR_OR_NULL(p) && !*pos)
994 seq_puts(seqf, "major minor #blocks name\n\n");
998 static int show_partition(struct seq_file *seqf, void *v)
1000 struct gendisk *sgp = v;
1001 struct disk_part_iter piter;
1002 struct hd_struct *part;
1003 char buf[BDEVNAME_SIZE];
1005 /* Don't show non-partitionable removeable devices or empty devices */
1006 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
1007 (sgp->flags & GENHD_FL_REMOVABLE)))
1009 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
1012 /* show the full disk and all non-0 size partitions of it */
1013 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
1014 while ((part = disk_part_iter_next(&piter)))
1015 seq_printf(seqf, "%4d %7d %10llu %s\n",
1016 MAJOR(part_devt(part)), MINOR(part_devt(part)),
1017 (unsigned long long)part_nr_sects_read(part) >> 1,
1018 disk_name(sgp, part->partno, buf));
1019 disk_part_iter_exit(&piter);
1024 static const struct seq_operations partitions_op = {
1025 .start = show_partition_start,
1026 .next = disk_seqf_next,
1027 .stop = disk_seqf_stop,
1028 .show = show_partition
1031 static int partitions_open(struct inode *inode, struct file *file)
1033 return seq_open(file, &partitions_op);
1036 static const struct file_operations proc_partitions_operations = {
1037 .open = partitions_open,
1039 .llseek = seq_lseek,
1040 .release = seq_release,
1045 static struct kobject *base_probe(dev_t devt, int *partno, void *data)
1047 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
1048 /* Make old-style 2.4 aliases work */
1049 request_module("block-major-%d", MAJOR(devt));
1053 static int __init genhd_device_init(void)
1057 block_class.dev_kobj = sysfs_dev_block_kobj;
1058 error = class_register(&block_class);
1059 if (unlikely(error))
1061 bdev_map = kobj_map_init(base_probe, &block_class_lock);
1064 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
1066 /* create top-level block dir */
1067 if (!sysfs_deprecated)
1068 block_depr = kobject_create_and_add("block", NULL);
1072 subsys_initcall(genhd_device_init);
1074 static ssize_t disk_range_show(struct device *dev,
1075 struct device_attribute *attr, char *buf)
1077 struct gendisk *disk = dev_to_disk(dev);
1079 return sprintf(buf, "%d\n", disk->minors);
1082 static ssize_t disk_ext_range_show(struct device *dev,
1083 struct device_attribute *attr, char *buf)
1085 struct gendisk *disk = dev_to_disk(dev);
1087 return sprintf(buf, "%d\n", disk_max_parts(disk));
1090 static ssize_t disk_removable_show(struct device *dev,
1091 struct device_attribute *attr, char *buf)
1093 struct gendisk *disk = dev_to_disk(dev);
1095 return sprintf(buf, "%d\n",
1096 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
1099 static ssize_t disk_hidden_show(struct device *dev,
1100 struct device_attribute *attr, char *buf)
1102 struct gendisk *disk = dev_to_disk(dev);
1104 return sprintf(buf, "%d\n",
1105 (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
1108 static ssize_t disk_ro_show(struct device *dev,
1109 struct device_attribute *attr, char *buf)
1111 struct gendisk *disk = dev_to_disk(dev);
1113 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
1116 static ssize_t disk_capability_show(struct device *dev,
1117 struct device_attribute *attr, char *buf)
1119 struct gendisk *disk = dev_to_disk(dev);
1121 return sprintf(buf, "%x\n", disk->flags);
1124 static ssize_t disk_alignment_offset_show(struct device *dev,
1125 struct device_attribute *attr,
1128 struct gendisk *disk = dev_to_disk(dev);
1130 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
1133 static ssize_t disk_discard_alignment_show(struct device *dev,
1134 struct device_attribute *attr,
1137 struct gendisk *disk = dev_to_disk(dev);
1139 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
1142 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
1143 static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
1144 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
1145 static DEVICE_ATTR(hidden, S_IRUGO, disk_hidden_show, NULL);
1146 static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
1147 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
1148 static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL);
1149 static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show,
1151 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
1152 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
1153 static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
1154 static DEVICE_ATTR(badblocks, S_IRUGO | S_IWUSR, disk_badblocks_show,
1155 disk_badblocks_store);
1156 #ifdef CONFIG_FAIL_MAKE_REQUEST
1157 static struct device_attribute dev_attr_fail =
1158 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
1160 #ifdef CONFIG_FAIL_IO_TIMEOUT
1161 static struct device_attribute dev_attr_fail_timeout =
1162 __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show,
1163 part_timeout_store);
1166 static struct attribute *disk_attrs[] = {
1167 &dev_attr_range.attr,
1168 &dev_attr_ext_range.attr,
1169 &dev_attr_removable.attr,
1170 &dev_attr_hidden.attr,
1172 &dev_attr_size.attr,
1173 &dev_attr_alignment_offset.attr,
1174 &dev_attr_discard_alignment.attr,
1175 &dev_attr_capability.attr,
1176 &dev_attr_stat.attr,
1177 &dev_attr_inflight.attr,
1178 &dev_attr_badblocks.attr,
1179 #ifdef CONFIG_FAIL_MAKE_REQUEST
1180 &dev_attr_fail.attr,
1182 #ifdef CONFIG_FAIL_IO_TIMEOUT
1183 &dev_attr_fail_timeout.attr,
1188 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1190 struct device *dev = container_of(kobj, typeof(*dev), kobj);
1191 struct gendisk *disk = dev_to_disk(dev);
1193 if (a == &dev_attr_badblocks.attr && !disk->bb)
1198 static struct attribute_group disk_attr_group = {
1199 .attrs = disk_attrs,
1200 .is_visible = disk_visible,
1203 static const struct attribute_group *disk_attr_groups[] = {
1209 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1210 * @disk: disk to replace part_tbl for
1211 * @new_ptbl: new part_tbl to install
1213 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1214 * original ptbl is freed using RCU callback.
1217 * Matching bd_mutex locked or the caller is the only user of @disk.
1219 static void disk_replace_part_tbl(struct gendisk *disk,
1220 struct disk_part_tbl *new_ptbl)
1222 struct disk_part_tbl *old_ptbl =
1223 rcu_dereference_protected(disk->part_tbl, 1);
1225 rcu_assign_pointer(disk->part_tbl, new_ptbl);
1228 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
1229 kfree_rcu(old_ptbl, rcu_head);
1234 * disk_expand_part_tbl - expand disk->part_tbl
1235 * @disk: disk to expand part_tbl for
1236 * @partno: expand such that this partno can fit in
1238 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1239 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1242 * Matching bd_mutex locked or the caller is the only user of @disk.
1246 * 0 on success, -errno on failure.
1248 int disk_expand_part_tbl(struct gendisk *disk, int partno)
1250 struct disk_part_tbl *old_ptbl =
1251 rcu_dereference_protected(disk->part_tbl, 1);
1252 struct disk_part_tbl *new_ptbl;
1253 int len = old_ptbl ? old_ptbl->len : 0;
1258 * check for int overflow, since we can get here from blkpg_ioctl()
1259 * with a user passed 'partno'.
1261 target = partno + 1;
1265 /* disk_max_parts() is zero during initialization, ignore if so */
1266 if (disk_max_parts(disk) && target > disk_max_parts(disk))
1272 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
1273 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
1277 new_ptbl->len = target;
1279 for (i = 0; i < len; i++)
1280 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
1282 disk_replace_part_tbl(disk, new_ptbl);
1286 static void disk_release(struct device *dev)
1288 struct gendisk *disk = dev_to_disk(dev);
1290 blk_free_devt(dev->devt);
1291 disk_release_events(disk);
1292 kfree(disk->random);
1293 disk_replace_part_tbl(disk, NULL);
1294 hd_free_part(&disk->part0);
1296 blk_put_queue(disk->queue);
1299 struct class block_class = {
1303 static char *block_devnode(struct device *dev, umode_t *mode,
1304 kuid_t *uid, kgid_t *gid)
1306 struct gendisk *disk = dev_to_disk(dev);
1309 return disk->devnode(disk, mode);
1313 static const struct device_type disk_type = {
1315 .groups = disk_attr_groups,
1316 .release = disk_release,
1317 .devnode = block_devnode,
1320 #ifdef CONFIG_PROC_FS
1322 * aggregate disk stat collector. Uses the same stats that the sysfs
1323 * entries do, above, but makes them available through one seq_file.
1325 * The output looks suspiciously like /proc/partitions with a bunch of
1328 static int diskstats_show(struct seq_file *seqf, void *v)
1330 struct gendisk *gp = v;
1331 struct disk_part_iter piter;
1332 struct hd_struct *hd;
1333 char buf[BDEVNAME_SIZE];
1334 unsigned int inflight[2];
1338 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1339 seq_puts(seqf, "major minor name"
1340 " rio rmerge rsect ruse wio wmerge "
1341 "wsect wuse running use aveq"
1345 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
1346 while ((hd = disk_part_iter_next(&piter))) {
1347 cpu = part_stat_lock();
1348 part_round_stats(gp->queue, cpu, hd);
1350 part_in_flight(gp->queue, hd, inflight);
1351 seq_printf(seqf, "%4d %7d %s %lu %lu %lu "
1352 "%u %lu %lu %lu %u %u %u %u\n",
1353 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1354 disk_name(gp, hd->partno, buf),
1355 part_stat_read(hd, ios[READ]),
1356 part_stat_read(hd, merges[READ]),
1357 part_stat_read(hd, sectors[READ]),
1358 jiffies_to_msecs(part_stat_read(hd, ticks[READ])),
1359 part_stat_read(hd, ios[WRITE]),
1360 part_stat_read(hd, merges[WRITE]),
1361 part_stat_read(hd, sectors[WRITE]),
1362 jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])),
1364 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1365 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1368 disk_part_iter_exit(&piter);
1373 static const struct seq_operations diskstats_op = {
1374 .start = disk_seqf_start,
1375 .next = disk_seqf_next,
1376 .stop = disk_seqf_stop,
1377 .show = diskstats_show
1380 static int diskstats_open(struct inode *inode, struct file *file)
1382 return seq_open(file, &diskstats_op);
1385 static const struct file_operations proc_diskstats_operations = {
1386 .open = diskstats_open,
1388 .llseek = seq_lseek,
1389 .release = seq_release,
1392 static int __init proc_genhd_init(void)
1394 proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
1395 proc_create("partitions", 0, NULL, &proc_partitions_operations);
1398 module_init(proc_genhd_init);
1399 #endif /* CONFIG_PROC_FS */
1401 dev_t blk_lookup_devt(const char *name, int partno)
1403 dev_t devt = MKDEV(0, 0);
1404 struct class_dev_iter iter;
1407 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1408 while ((dev = class_dev_iter_next(&iter))) {
1409 struct gendisk *disk = dev_to_disk(dev);
1410 struct hd_struct *part;
1412 if (strcmp(dev_name(dev), name))
1415 if (partno < disk->minors) {
1416 /* We need to return the right devno, even
1417 * if the partition doesn't exist yet.
1419 devt = MKDEV(MAJOR(dev->devt),
1420 MINOR(dev->devt) + partno);
1423 part = disk_get_part(disk, partno);
1425 devt = part_devt(part);
1426 disk_put_part(part);
1429 disk_put_part(part);
1431 class_dev_iter_exit(&iter);
1434 EXPORT_SYMBOL(blk_lookup_devt);
1436 struct gendisk *__alloc_disk_node(int minors, int node_id)
1438 struct gendisk *disk;
1439 struct disk_part_tbl *ptbl;
1441 if (minors > DISK_MAX_PARTS) {
1443 "block: can't allocate more than %d partitions\n",
1445 minors = DISK_MAX_PARTS;
1448 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1450 if (!init_part_stats(&disk->part0)) {
1454 init_rwsem(&disk->lookup_sem);
1455 disk->node_id = node_id;
1456 if (disk_expand_part_tbl(disk, 0)) {
1457 free_part_stats(&disk->part0);
1461 ptbl = rcu_dereference_protected(disk->part_tbl, 1);
1462 rcu_assign_pointer(ptbl->part[0], &disk->part0);
1465 * set_capacity() and get_capacity() currently don't use
1466 * seqcounter to read/update the part0->nr_sects. Still init
1467 * the counter as we can read the sectors in IO submission
1468 * patch using seqence counters.
1470 * TODO: Ideally set_capacity() and get_capacity() should be
1471 * converted to make use of bd_mutex and sequence counters.
1473 seqcount_init(&disk->part0.nr_sects_seq);
1474 if (hd_ref_init(&disk->part0)) {
1475 hd_free_part(&disk->part0);
1480 disk->minors = minors;
1481 rand_initialize_disk(disk);
1482 disk_to_dev(disk)->class = &block_class;
1483 disk_to_dev(disk)->type = &disk_type;
1484 device_initialize(disk_to_dev(disk));
1488 EXPORT_SYMBOL(__alloc_disk_node);
1490 struct kobject *get_disk_and_module(struct gendisk *disk)
1492 struct module *owner;
1493 struct kobject *kobj;
1497 owner = disk->fops->owner;
1498 if (owner && !try_module_get(owner))
1500 kobj = kobject_get_unless_zero(&disk_to_dev(disk)->kobj);
1508 EXPORT_SYMBOL(get_disk_and_module);
1510 void put_disk(struct gendisk *disk)
1513 kobject_put(&disk_to_dev(disk)->kobj);
1515 EXPORT_SYMBOL(put_disk);
1518 * This is a counterpart of get_disk_and_module() and thus also of
1521 void put_disk_and_module(struct gendisk *disk)
1524 struct module *owner = disk->fops->owner;
1530 EXPORT_SYMBOL(put_disk_and_module);
1532 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1534 char event[] = "DISK_RO=1";
1535 char *envp[] = { event, NULL };
1539 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1542 void set_device_ro(struct block_device *bdev, int flag)
1544 bdev->bd_part->policy = flag;
1547 EXPORT_SYMBOL(set_device_ro);
1549 void set_disk_ro(struct gendisk *disk, int flag)
1551 struct disk_part_iter piter;
1552 struct hd_struct *part;
1554 if (disk->part0.policy != flag) {
1555 set_disk_ro_uevent(disk, flag);
1556 disk->part0.policy = flag;
1559 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
1560 while ((part = disk_part_iter_next(&piter)))
1561 part->policy = flag;
1562 disk_part_iter_exit(&piter);
1565 EXPORT_SYMBOL(set_disk_ro);
1567 int bdev_read_only(struct block_device *bdev)
1571 return bdev->bd_part->policy;
1574 EXPORT_SYMBOL(bdev_read_only);
1576 int invalidate_partition(struct gendisk *disk, int partno)
1579 struct block_device *bdev = bdget_disk(disk, partno);
1582 res = __invalidate_device(bdev, true);
1588 EXPORT_SYMBOL(invalidate_partition);
1591 * Disk events - monitor disk events like media change and eject request.
1593 struct disk_events {
1594 struct list_head node; /* all disk_event's */
1595 struct gendisk *disk; /* the associated disk */
1598 struct mutex block_mutex; /* protects blocking */
1599 int block; /* event blocking depth */
1600 unsigned int pending; /* events already sent out */
1601 unsigned int clearing; /* events being cleared */
1603 long poll_msecs; /* interval, -1 for default */
1604 struct delayed_work dwork;
1607 static const char *disk_events_strs[] = {
1608 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1609 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1612 static char *disk_uevents[] = {
1613 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1614 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1617 /* list of all disk_events */
1618 static DEFINE_MUTEX(disk_events_mutex);
1619 static LIST_HEAD(disk_events);
1621 /* disable in-kernel polling by default */
1622 static unsigned long disk_events_dfl_poll_msecs;
1624 static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1626 struct disk_events *ev = disk->ev;
1627 long intv_msecs = 0;
1630 * If device-specific poll interval is set, always use it. If
1631 * the default is being used, poll iff there are events which
1632 * can't be monitored asynchronously.
1634 if (ev->poll_msecs >= 0)
1635 intv_msecs = ev->poll_msecs;
1636 else if (disk->events & ~disk->async_events)
1637 intv_msecs = disk_events_dfl_poll_msecs;
1639 return msecs_to_jiffies(intv_msecs);
1643 * disk_block_events - block and flush disk event checking
1644 * @disk: disk to block events for
1646 * On return from this function, it is guaranteed that event checking
1647 * isn't in progress and won't happen until unblocked by
1648 * disk_unblock_events(). Events blocking is counted and the actual
1649 * unblocking happens after the matching number of unblocks are done.
1651 * Note that this intentionally does not block event checking from
1652 * disk_clear_events().
1657 void disk_block_events(struct gendisk *disk)
1659 struct disk_events *ev = disk->ev;
1660 unsigned long flags;
1667 * Outer mutex ensures that the first blocker completes canceling
1668 * the event work before further blockers are allowed to finish.
1670 mutex_lock(&ev->block_mutex);
1672 spin_lock_irqsave(&ev->lock, flags);
1673 cancel = !ev->block++;
1674 spin_unlock_irqrestore(&ev->lock, flags);
1677 cancel_delayed_work_sync(&disk->ev->dwork);
1679 mutex_unlock(&ev->block_mutex);
1682 static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1684 struct disk_events *ev = disk->ev;
1686 unsigned long flags;
1688 spin_lock_irqsave(&ev->lock, flags);
1690 if (WARN_ON_ONCE(ev->block <= 0))
1696 intv = disk_events_poll_jiffies(disk);
1698 queue_delayed_work(system_freezable_power_efficient_wq,
1701 queue_delayed_work(system_freezable_power_efficient_wq,
1704 spin_unlock_irqrestore(&ev->lock, flags);
1708 * disk_unblock_events - unblock disk event checking
1709 * @disk: disk to unblock events for
1711 * Undo disk_block_events(). When the block count reaches zero, it
1712 * starts events polling if configured.
1715 * Don't care. Safe to call from irq context.
1717 void disk_unblock_events(struct gendisk *disk)
1720 __disk_unblock_events(disk, false);
1724 * disk_flush_events - schedule immediate event checking and flushing
1725 * @disk: disk to check and flush events for
1726 * @mask: events to flush
1728 * Schedule immediate event checking on @disk if not blocked. Events in
1729 * @mask are scheduled to be cleared from the driver. Note that this
1730 * doesn't clear the events from @disk->ev.
1733 * If @mask is non-zero must be called with bdev->bd_mutex held.
1735 void disk_flush_events(struct gendisk *disk, unsigned int mask)
1737 struct disk_events *ev = disk->ev;
1742 spin_lock_irq(&ev->lock);
1743 ev->clearing |= mask;
1745 mod_delayed_work(system_freezable_power_efficient_wq,
1747 spin_unlock_irq(&ev->lock);
1751 * disk_clear_events - synchronously check, clear and return pending events
1752 * @disk: disk to fetch and clear events from
1753 * @mask: mask of events to be fetched and cleared
1755 * Disk events are synchronously checked and pending events in @mask
1756 * are cleared and returned. This ignores the block count.
1761 unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1763 const struct block_device_operations *bdops = disk->fops;
1764 struct disk_events *ev = disk->ev;
1765 unsigned int pending;
1766 unsigned int clearing = mask;
1769 /* for drivers still using the old ->media_changed method */
1770 if ((mask & DISK_EVENT_MEDIA_CHANGE) &&
1771 bdops->media_changed && bdops->media_changed(disk))
1772 return DISK_EVENT_MEDIA_CHANGE;
1776 disk_block_events(disk);
1779 * store the union of mask and ev->clearing on the stack so that the
1780 * race with disk_flush_events does not cause ambiguity (ev->clearing
1781 * can still be modified even if events are blocked).
1783 spin_lock_irq(&ev->lock);
1784 clearing |= ev->clearing;
1786 spin_unlock_irq(&ev->lock);
1788 disk_check_events(ev, &clearing);
1790 * if ev->clearing is not 0, the disk_flush_events got called in the
1791 * middle of this function, so we want to run the workfn without delay.
1793 __disk_unblock_events(disk, ev->clearing ? true : false);
1795 /* then, fetch and clear pending events */
1796 spin_lock_irq(&ev->lock);
1797 pending = ev->pending & mask;
1798 ev->pending &= ~mask;
1799 spin_unlock_irq(&ev->lock);
1800 WARN_ON_ONCE(clearing & mask);
1806 * Separate this part out so that a different pointer for clearing_ptr can be
1807 * passed in for disk_clear_events.
1809 static void disk_events_workfn(struct work_struct *work)
1811 struct delayed_work *dwork = to_delayed_work(work);
1812 struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
1814 disk_check_events(ev, &ev->clearing);
1817 static void disk_check_events(struct disk_events *ev,
1818 unsigned int *clearing_ptr)
1820 struct gendisk *disk = ev->disk;
1821 char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
1822 unsigned int clearing = *clearing_ptr;
1823 unsigned int events;
1825 int nr_events = 0, i;
1828 events = disk->fops->check_events(disk, clearing);
1830 /* accumulate pending events and schedule next poll if necessary */
1831 spin_lock_irq(&ev->lock);
1833 events &= ~ev->pending;
1834 ev->pending |= events;
1835 *clearing_ptr &= ~clearing;
1837 intv = disk_events_poll_jiffies(disk);
1838 if (!ev->block && intv)
1839 queue_delayed_work(system_freezable_power_efficient_wq,
1842 spin_unlock_irq(&ev->lock);
1845 * Tell userland about new events. Only the events listed in
1846 * @disk->events are reported. Unlisted events are processed the
1847 * same internally but never get reported to userland.
1849 for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
1850 if (events & disk->events & (1 << i))
1851 envp[nr_events++] = disk_uevents[i];
1854 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1858 * A disk events enabled device has the following sysfs nodes under
1859 * its /sys/block/X/ directory.
1861 * events : list of all supported events
1862 * events_async : list of events which can be detected w/o polling
1863 * events_poll_msecs : polling interval, 0: disable, -1: system default
1865 static ssize_t __disk_events_show(unsigned int events, char *buf)
1867 const char *delim = "";
1871 for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1872 if (events & (1 << i)) {
1873 pos += sprintf(buf + pos, "%s%s",
1874 delim, disk_events_strs[i]);
1878 pos += sprintf(buf + pos, "\n");
1882 static ssize_t disk_events_show(struct device *dev,
1883 struct device_attribute *attr, char *buf)
1885 struct gendisk *disk = dev_to_disk(dev);
1887 return __disk_events_show(disk->events, buf);
1890 static ssize_t disk_events_async_show(struct device *dev,
1891 struct device_attribute *attr, char *buf)
1893 struct gendisk *disk = dev_to_disk(dev);
1895 return __disk_events_show(disk->async_events, buf);
1898 static ssize_t disk_events_poll_msecs_show(struct device *dev,
1899 struct device_attribute *attr,
1902 struct gendisk *disk = dev_to_disk(dev);
1904 return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1907 static ssize_t disk_events_poll_msecs_store(struct device *dev,
1908 struct device_attribute *attr,
1909 const char *buf, size_t count)
1911 struct gendisk *disk = dev_to_disk(dev);
1914 if (!count || !sscanf(buf, "%ld", &intv))
1917 if (intv < 0 && intv != -1)
1920 disk_block_events(disk);
1921 disk->ev->poll_msecs = intv;
1922 __disk_unblock_events(disk, true);
1927 static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL);
1928 static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL);
1929 static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR,
1930 disk_events_poll_msecs_show,
1931 disk_events_poll_msecs_store);
1933 static const struct attribute *disk_events_attrs[] = {
1934 &dev_attr_events.attr,
1935 &dev_attr_events_async.attr,
1936 &dev_attr_events_poll_msecs.attr,
1941 * The default polling interval can be specified by the kernel
1942 * parameter block.events_dfl_poll_msecs which defaults to 0
1943 * (disable). This can also be modified runtime by writing to
1944 * /sys/module/block/events_dfl_poll_msecs.
1946 static int disk_events_set_dfl_poll_msecs(const char *val,
1947 const struct kernel_param *kp)
1949 struct disk_events *ev;
1952 ret = param_set_ulong(val, kp);
1956 mutex_lock(&disk_events_mutex);
1958 list_for_each_entry(ev, &disk_events, node)
1959 disk_flush_events(ev->disk, 0);
1961 mutex_unlock(&disk_events_mutex);
1966 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1967 .set = disk_events_set_dfl_poll_msecs,
1968 .get = param_get_ulong,
1971 #undef MODULE_PARAM_PREFIX
1972 #define MODULE_PARAM_PREFIX "block."
1974 module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
1975 &disk_events_dfl_poll_msecs, 0644);
1978 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
1980 static void disk_alloc_events(struct gendisk *disk)
1982 struct disk_events *ev;
1984 if (!disk->fops->check_events)
1987 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1989 pr_warn("%s: failed to initialize events\n", disk->disk_name);
1993 INIT_LIST_HEAD(&ev->node);
1995 spin_lock_init(&ev->lock);
1996 mutex_init(&ev->block_mutex);
1998 ev->poll_msecs = -1;
1999 INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
2004 static void disk_add_events(struct gendisk *disk)
2009 /* FIXME: error handling */
2010 if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0)
2011 pr_warn("%s: failed to create sysfs files for events\n",
2014 mutex_lock(&disk_events_mutex);
2015 list_add_tail(&disk->ev->node, &disk_events);
2016 mutex_unlock(&disk_events_mutex);
2019 * Block count is initialized to 1 and the following initial
2020 * unblock kicks it into action.
2022 __disk_unblock_events(disk, true);
2025 static void disk_del_events(struct gendisk *disk)
2030 disk_block_events(disk);
2032 mutex_lock(&disk_events_mutex);
2033 list_del_init(&disk->ev->node);
2034 mutex_unlock(&disk_events_mutex);
2036 sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
2039 static void disk_release_events(struct gendisk *disk)
2041 /* the block count should be 1 from disk_del_events() */
2042 WARN_ON_ONCE(disk->ev && disk->ev->block != 1);