1 // SPDX-License-Identifier: GPL-2.0
6 #include <linux/module.h>
8 #include <linux/genhd.h>
9 #include <linux/kdev_t.h>
10 #include <linux/kernel.h>
11 #include <linux/blkdev.h>
12 #include <linux/backing-dev.h>
13 #include <linux/init.h>
14 #include <linux/spinlock.h>
15 #include <linux/proc_fs.h>
16 #include <linux/seq_file.h>
17 #include <linux/slab.h>
18 #include <linux/kmod.h>
19 #include <linux/kobj_map.h>
20 #include <linux/mutex.h>
21 #include <linux/idr.h>
22 #include <linux/log2.h>
23 #include <linux/pm_runtime.h>
24 #include <linux/badblocks.h>
28 static DEFINE_MUTEX(block_class_lock);
29 struct kobject *block_depr;
31 /* for extended dynamic devt allocation, currently only one major is used */
32 #define NR_EXT_DEVT (1 << MINORBITS)
34 /* For extended devt allocation. ext_devt_lock prevents look up
35 * results from going away underneath its user.
37 static DEFINE_SPINLOCK(ext_devt_lock);
38 static DEFINE_IDR(ext_devt_idr);
40 static const struct device_type disk_type;
42 static void disk_check_events(struct disk_events *ev,
43 unsigned int *clearing_ptr);
44 static void disk_alloc_events(struct gendisk *disk);
45 static void disk_add_events(struct gendisk *disk);
46 static void disk_del_events(struct gendisk *disk);
47 static void disk_release_events(struct gendisk *disk);
49 void part_inc_in_flight(struct request_queue *q, struct hd_struct *part, int rw)
54 part_stat_local_inc(part, in_flight[rw]);
56 part_stat_local_inc(&part_to_disk(part)->part0, in_flight[rw]);
59 void part_dec_in_flight(struct request_queue *q, struct hd_struct *part, int rw)
64 part_stat_local_dec(part, in_flight[rw]);
66 part_stat_local_dec(&part_to_disk(part)->part0, in_flight[rw]);
69 unsigned int part_in_flight(struct request_queue *q, struct hd_struct *part)
72 unsigned int inflight;
75 return blk_mq_in_flight(q, part);
79 for_each_possible_cpu(cpu) {
80 inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) +
81 part_stat_local_read_cpu(part, in_flight[1], cpu);
83 if ((int)inflight < 0)
89 void part_in_flight_rw(struct request_queue *q, struct hd_struct *part,
90 unsigned int inflight[2])
95 blk_mq_in_flight_rw(q, part, inflight);
101 for_each_possible_cpu(cpu) {
102 inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu);
103 inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu);
105 if ((int)inflight[0] < 0)
107 if ((int)inflight[1] < 0)
111 struct hd_struct *__disk_get_part(struct gendisk *disk, int partno)
113 struct disk_part_tbl *ptbl = rcu_dereference(disk->part_tbl);
115 if (unlikely(partno < 0 || partno >= ptbl->len))
117 return rcu_dereference(ptbl->part[partno]);
121 * disk_get_part - get partition
122 * @disk: disk to look partition from
123 * @partno: partition number
125 * Look for partition @partno from @disk. If found, increment
126 * reference count and return it.
132 * Pointer to the found partition on success, NULL if not found.
134 struct hd_struct *disk_get_part(struct gendisk *disk, int partno)
136 struct hd_struct *part;
139 part = __disk_get_part(disk, partno);
141 get_device(part_to_dev(part));
146 EXPORT_SYMBOL_GPL(disk_get_part);
149 * disk_part_iter_init - initialize partition iterator
150 * @piter: iterator to initialize
151 * @disk: disk to iterate over
152 * @flags: DISK_PITER_* flags
154 * Initialize @piter so that it iterates over partitions of @disk.
159 void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk,
162 struct disk_part_tbl *ptbl;
165 ptbl = rcu_dereference(disk->part_tbl);
170 if (flags & DISK_PITER_REVERSE)
171 piter->idx = ptbl->len - 1;
172 else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0))
177 piter->flags = flags;
181 EXPORT_SYMBOL_GPL(disk_part_iter_init);
184 * disk_part_iter_next - proceed iterator to the next partition and return it
185 * @piter: iterator of interest
187 * Proceed @piter to the next partition and return it.
192 struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter)
194 struct disk_part_tbl *ptbl;
197 /* put the last partition */
198 disk_put_part(piter->part);
203 ptbl = rcu_dereference(piter->disk->part_tbl);
205 /* determine iteration parameters */
206 if (piter->flags & DISK_PITER_REVERSE) {
208 if (piter->flags & (DISK_PITER_INCL_PART0 |
209 DISK_PITER_INCL_EMPTY_PART0))
218 /* iterate to the next partition */
219 for (; piter->idx != end; piter->idx += inc) {
220 struct hd_struct *part;
222 part = rcu_dereference(ptbl->part[piter->idx]);
225 if (!part_nr_sects_read(part) &&
226 !(piter->flags & DISK_PITER_INCL_EMPTY) &&
227 !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 &&
231 get_device(part_to_dev(part));
241 EXPORT_SYMBOL_GPL(disk_part_iter_next);
244 * disk_part_iter_exit - finish up partition iteration
245 * @piter: iter of interest
247 * Called when iteration is over. Cleans up @piter.
252 void disk_part_iter_exit(struct disk_part_iter *piter)
254 disk_put_part(piter->part);
257 EXPORT_SYMBOL_GPL(disk_part_iter_exit);
259 static inline int sector_in_part(struct hd_struct *part, sector_t sector)
261 return part->start_sect <= sector &&
262 sector < part->start_sect + part_nr_sects_read(part);
266 * disk_map_sector_rcu - map sector to partition
267 * @disk: gendisk of interest
268 * @sector: sector to map
270 * Find out which partition @sector maps to on @disk. This is
271 * primarily used for stats accounting.
274 * RCU read locked. The returned partition pointer is valid only
275 * while preemption is disabled.
278 * Found partition on success, part0 is returned if no partition matches
280 struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector)
282 struct disk_part_tbl *ptbl;
283 struct hd_struct *part;
286 ptbl = rcu_dereference(disk->part_tbl);
288 part = rcu_dereference(ptbl->last_lookup);
289 if (part && sector_in_part(part, sector))
292 for (i = 1; i < ptbl->len; i++) {
293 part = rcu_dereference(ptbl->part[i]);
295 if (part && sector_in_part(part, sector)) {
296 rcu_assign_pointer(ptbl->last_lookup, part);
302 EXPORT_SYMBOL_GPL(disk_map_sector_rcu);
305 * Can be deleted altogether. Later.
308 #define BLKDEV_MAJOR_HASH_SIZE 255
309 static struct blk_major_name {
310 struct blk_major_name *next;
313 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
315 /* index in the above - for now: assume no multimajor ranges */
316 static inline int major_to_index(unsigned major)
318 return major % BLKDEV_MAJOR_HASH_SIZE;
321 #ifdef CONFIG_PROC_FS
322 void blkdev_show(struct seq_file *seqf, off_t offset)
324 struct blk_major_name *dp;
326 mutex_lock(&block_class_lock);
327 for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
328 if (dp->major == offset)
329 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
330 mutex_unlock(&block_class_lock);
332 #endif /* CONFIG_PROC_FS */
335 * register_blkdev - register a new block device
337 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
338 * @major = 0, try to allocate any unused major number.
339 * @name: the name of the new block device as a zero terminated string
341 * The @name must be unique within the system.
343 * The return value depends on the @major input parameter:
345 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
346 * then the function returns zero on success, or a negative error code
347 * - if any unused major number was requested with @major = 0 parameter
348 * then the return value is the allocated major number in range
349 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
351 * See Documentation/admin-guide/devices.txt for the list of allocated
354 int register_blkdev(unsigned int major, const char *name)
356 struct blk_major_name **n, *p;
359 mutex_lock(&block_class_lock);
363 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
364 if (major_names[index] == NULL)
369 printk("%s: failed to get major for %s\n",
378 if (major >= BLKDEV_MAJOR_MAX) {
379 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
380 __func__, major, BLKDEV_MAJOR_MAX-1, name);
386 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
393 strlcpy(p->name, name, sizeof(p->name));
395 index = major_to_index(major);
397 for (n = &major_names[index]; *n; n = &(*n)->next) {
398 if ((*n)->major == major)
407 printk("register_blkdev: cannot get major %u for %s\n",
412 mutex_unlock(&block_class_lock);
416 EXPORT_SYMBOL(register_blkdev);
418 void unregister_blkdev(unsigned int major, const char *name)
420 struct blk_major_name **n;
421 struct blk_major_name *p = NULL;
422 int index = major_to_index(major);
424 mutex_lock(&block_class_lock);
425 for (n = &major_names[index]; *n; n = &(*n)->next)
426 if ((*n)->major == major)
428 if (!*n || strcmp((*n)->name, name)) {
434 mutex_unlock(&block_class_lock);
438 EXPORT_SYMBOL(unregister_blkdev);
440 static struct kobj_map *bdev_map;
443 * blk_mangle_minor - scatter minor numbers apart
444 * @minor: minor number to mangle
446 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
447 * is enabled. Mangling twice gives the original value.
455 static int blk_mangle_minor(int minor)
457 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
460 for (i = 0; i < MINORBITS / 2; i++) {
461 int low = minor & (1 << i);
462 int high = minor & (1 << (MINORBITS - 1 - i));
463 int distance = MINORBITS - 1 - 2 * i;
465 minor ^= low | high; /* clear both bits */
466 low <<= distance; /* swap the positions */
468 minor |= low | high; /* and set */
475 * blk_alloc_devt - allocate a dev_t for a partition
476 * @part: partition to allocate dev_t for
477 * @devt: out parameter for resulting dev_t
479 * Allocate a dev_t for block device.
482 * 0 on success, allocated dev_t is returned in *@devt. -errno on
488 int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
490 struct gendisk *disk = part_to_disk(part);
493 /* in consecutive minor range? */
494 if (part->partno < disk->minors) {
495 *devt = MKDEV(disk->major, disk->first_minor + part->partno);
499 /* allocate ext devt */
500 idr_preload(GFP_KERNEL);
502 spin_lock_bh(&ext_devt_lock);
503 idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_NOWAIT);
504 spin_unlock_bh(&ext_devt_lock);
508 return idx == -ENOSPC ? -EBUSY : idx;
510 *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
515 * blk_free_devt - free a dev_t
516 * @devt: dev_t to free
518 * Free @devt which was allocated using blk_alloc_devt().
523 void blk_free_devt(dev_t devt)
525 if (devt == MKDEV(0, 0))
528 if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
529 spin_lock_bh(&ext_devt_lock);
530 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
531 spin_unlock_bh(&ext_devt_lock);
536 * We invalidate devt by assigning NULL pointer for devt in idr.
538 void blk_invalidate_devt(dev_t devt)
540 if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
541 spin_lock_bh(&ext_devt_lock);
542 idr_replace(&ext_devt_idr, NULL, blk_mangle_minor(MINOR(devt)));
543 spin_unlock_bh(&ext_devt_lock);
547 static char *bdevt_str(dev_t devt, char *buf)
549 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
550 char tbuf[BDEVT_SIZE];
551 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
552 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
554 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
560 * Register device numbers dev..(dev+range-1)
561 * range must be nonzero
562 * The hash chain is sorted on range, so that subranges can override.
564 void blk_register_region(dev_t devt, unsigned long range, struct module *module,
565 struct kobject *(*probe)(dev_t, int *, void *),
566 int (*lock)(dev_t, void *), void *data)
568 kobj_map(bdev_map, devt, range, module, probe, lock, data);
571 EXPORT_SYMBOL(blk_register_region);
573 void blk_unregister_region(dev_t devt, unsigned long range)
575 kobj_unmap(bdev_map, devt, range);
578 EXPORT_SYMBOL(blk_unregister_region);
580 static struct kobject *exact_match(dev_t devt, int *partno, void *data)
582 struct gendisk *p = data;
584 return &disk_to_dev(p)->kobj;
587 static int exact_lock(dev_t devt, void *data)
589 struct gendisk *p = data;
591 if (!get_disk_and_module(p))
596 static void register_disk(struct device *parent, struct gendisk *disk,
597 const struct attribute_group **groups)
599 struct device *ddev = disk_to_dev(disk);
600 struct block_device *bdev;
601 struct disk_part_iter piter;
602 struct hd_struct *part;
605 ddev->parent = parent;
607 dev_set_name(ddev, "%s", disk->disk_name);
609 /* delay uevents, until we scanned partition table */
610 dev_set_uevent_suppress(ddev, 1);
613 WARN_ON(ddev->groups);
614 ddev->groups = groups;
616 if (device_add(ddev))
618 if (!sysfs_deprecated) {
619 err = sysfs_create_link(block_depr, &ddev->kobj,
620 kobject_name(&ddev->kobj));
628 * avoid probable deadlock caused by allocating memory with
629 * GFP_KERNEL in runtime_resume callback of its all ancestor
632 pm_runtime_set_memalloc_noio(ddev, true);
634 disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj);
635 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
637 if (disk->flags & GENHD_FL_HIDDEN) {
638 dev_set_uevent_suppress(ddev, 0);
642 /* No minors to use for partitions */
643 if (!disk_part_scan_enabled(disk))
646 /* No such device (e.g., media were just removed) */
647 if (!get_capacity(disk))
650 bdev = bdget_disk(disk, 0);
654 bdev->bd_invalidated = 1;
655 err = blkdev_get(bdev, FMODE_READ, NULL);
658 blkdev_put(bdev, FMODE_READ);
661 /* announce disk after possible partitions are created */
662 dev_set_uevent_suppress(ddev, 0);
663 kobject_uevent(&ddev->kobj, KOBJ_ADD);
665 /* announce possible partitions */
666 disk_part_iter_init(&piter, disk, 0);
667 while ((part = disk_part_iter_next(&piter)))
668 kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD);
669 disk_part_iter_exit(&piter);
671 if (disk->queue->backing_dev_info->dev) {
672 err = sysfs_create_link(&ddev->kobj,
673 &disk->queue->backing_dev_info->dev->kobj,
680 * __device_add_disk - add disk information to kernel list
681 * @parent: parent device for the disk
682 * @disk: per-device partitioning information
683 * @groups: Additional per-device sysfs groups
684 * @register_queue: register the queue if set to true
686 * This function registers the partitioning information in @disk
689 * FIXME: error handling
691 static void __device_add_disk(struct device *parent, struct gendisk *disk,
692 const struct attribute_group **groups,
698 /* minors == 0 indicates to use ext devt from part0 and should
699 * be accompanied with EXT_DEVT flag. Make sure all
700 * parameters make sense.
702 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
703 WARN_ON(!disk->minors &&
704 !(disk->flags & (GENHD_FL_EXT_DEVT | GENHD_FL_HIDDEN)));
706 disk->flags |= GENHD_FL_UP;
708 retval = blk_alloc_devt(&disk->part0, &devt);
713 disk->major = MAJOR(devt);
714 disk->first_minor = MINOR(devt);
716 disk_alloc_events(disk);
718 if (disk->flags & GENHD_FL_HIDDEN) {
720 * Don't let hidden disks show up in /proc/partitions,
721 * and don't bother scanning for partitions either.
723 disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
724 disk->flags |= GENHD_FL_NO_PART_SCAN;
728 /* Register BDI before referencing it from bdev */
729 disk_to_dev(disk)->devt = devt;
730 ret = bdi_register_owner(disk->queue->backing_dev_info,
733 blk_register_region(disk_devt(disk), disk->minors, NULL,
734 exact_match, exact_lock, disk);
736 register_disk(parent, disk, groups);
738 blk_register_queue(disk);
741 * Take an extra ref on queue which will be put on disk_release()
742 * so that it sticks around as long as @disk is there.
744 WARN_ON_ONCE(!blk_get_queue(disk->queue));
746 disk_add_events(disk);
747 blk_integrity_add(disk);
750 void device_add_disk(struct device *parent, struct gendisk *disk,
751 const struct attribute_group **groups)
754 __device_add_disk(parent, disk, groups, true);
756 EXPORT_SYMBOL(device_add_disk);
758 void device_add_disk_no_queue_reg(struct device *parent, struct gendisk *disk)
760 __device_add_disk(parent, disk, NULL, false);
762 EXPORT_SYMBOL(device_add_disk_no_queue_reg);
764 void del_gendisk(struct gendisk *disk)
766 struct disk_part_iter piter;
767 struct hd_struct *part;
769 blk_integrity_del(disk);
770 disk_del_events(disk);
773 * Block lookups of the disk until all bdevs are unhashed and the
774 * disk is marked as dead (GENHD_FL_UP cleared).
776 down_write(&disk->lookup_sem);
777 /* invalidate stuff */
778 disk_part_iter_init(&piter, disk,
779 DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
780 while ((part = disk_part_iter_next(&piter))) {
781 invalidate_partition(disk, part->partno);
782 bdev_unhash_inode(part_devt(part));
783 delete_partition(disk, part->partno);
785 disk_part_iter_exit(&piter);
787 invalidate_partition(disk, 0);
788 bdev_unhash_inode(disk_devt(disk));
789 set_capacity(disk, 0);
790 disk->flags &= ~GENHD_FL_UP;
791 up_write(&disk->lookup_sem);
793 if (!(disk->flags & GENHD_FL_HIDDEN))
794 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
797 * Unregister bdi before releasing device numbers (as they can
798 * get reused and we'd get clashes in sysfs).
800 if (!(disk->flags & GENHD_FL_HIDDEN))
801 bdi_unregister(disk->queue->backing_dev_info);
802 blk_unregister_queue(disk);
807 if (!(disk->flags & GENHD_FL_HIDDEN))
808 blk_unregister_region(disk_devt(disk), disk->minors);
810 * Remove gendisk pointer from idr so that it cannot be looked up
811 * while RCU period before freeing gendisk is running to prevent
812 * use-after-free issues. Note that the device number stays
813 * "in-use" until we really free the gendisk.
815 blk_invalidate_devt(disk_devt(disk));
817 kobject_put(disk->part0.holder_dir);
818 kobject_put(disk->slave_dir);
820 part_stat_set_all(&disk->part0, 0);
821 disk->part0.stamp = 0;
822 if (!sysfs_deprecated)
823 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
824 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
825 device_del(disk_to_dev(disk));
827 EXPORT_SYMBOL(del_gendisk);
829 /* sysfs access to bad-blocks list. */
830 static ssize_t disk_badblocks_show(struct device *dev,
831 struct device_attribute *attr,
834 struct gendisk *disk = dev_to_disk(dev);
837 return sprintf(page, "\n");
839 return badblocks_show(disk->bb, page, 0);
842 static ssize_t disk_badblocks_store(struct device *dev,
843 struct device_attribute *attr,
844 const char *page, size_t len)
846 struct gendisk *disk = dev_to_disk(dev);
851 return badblocks_store(disk->bb, page, len, 0);
855 * get_gendisk - get partitioning information for a given device
856 * @devt: device to get partitioning information for
857 * @partno: returned partition index
859 * This function gets the structure containing partitioning
860 * information for the given device @devt.
862 struct gendisk *get_gendisk(dev_t devt, int *partno)
864 struct gendisk *disk = NULL;
866 if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
867 struct kobject *kobj;
869 kobj = kobj_lookup(bdev_map, devt, partno);
871 disk = dev_to_disk(kobj_to_dev(kobj));
873 struct hd_struct *part;
875 spin_lock_bh(&ext_devt_lock);
876 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
877 if (part && get_disk_and_module(part_to_disk(part))) {
878 *partno = part->partno;
879 disk = part_to_disk(part);
881 spin_unlock_bh(&ext_devt_lock);
888 * Synchronize with del_gendisk() to not return disk that is being
891 down_read(&disk->lookup_sem);
892 if (unlikely((disk->flags & GENHD_FL_HIDDEN) ||
893 !(disk->flags & GENHD_FL_UP))) {
894 up_read(&disk->lookup_sem);
895 put_disk_and_module(disk);
898 up_read(&disk->lookup_sem);
902 EXPORT_SYMBOL(get_gendisk);
905 * bdget_disk - do bdget() by gendisk and partition number
906 * @disk: gendisk of interest
907 * @partno: partition number
909 * Find partition @partno from @disk, do bdget() on it.
915 * Resulting block_device on success, NULL on failure.
917 struct block_device *bdget_disk(struct gendisk *disk, int partno)
919 struct hd_struct *part;
920 struct block_device *bdev = NULL;
922 part = disk_get_part(disk, partno);
924 bdev = bdget(part_devt(part));
929 EXPORT_SYMBOL(bdget_disk);
932 * print a full list of all partitions - intended for places where the root
933 * filesystem can't be mounted and thus to give the victim some idea of what
936 void __init printk_all_partitions(void)
938 struct class_dev_iter iter;
941 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
942 while ((dev = class_dev_iter_next(&iter))) {
943 struct gendisk *disk = dev_to_disk(dev);
944 struct disk_part_iter piter;
945 struct hd_struct *part;
946 char name_buf[BDEVNAME_SIZE];
947 char devt_buf[BDEVT_SIZE];
950 * Don't show empty devices or things that have been
953 if (get_capacity(disk) == 0 ||
954 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
958 * Note, unlike /proc/partitions, I am showing the
959 * numbers in hex - the same format as the root=
962 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
963 while ((part = disk_part_iter_next(&piter))) {
964 bool is_part0 = part == &disk->part0;
966 printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
967 bdevt_str(part_devt(part), devt_buf),
968 (unsigned long long)part_nr_sects_read(part) >> 1
969 , disk_name(disk, part->partno, name_buf),
970 part->info ? part->info->uuid : "");
972 if (dev->parent && dev->parent->driver)
973 printk(" driver: %s\n",
974 dev->parent->driver->name);
976 printk(" (driver?)\n");
980 disk_part_iter_exit(&piter);
982 class_dev_iter_exit(&iter);
985 #ifdef CONFIG_PROC_FS
987 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
990 struct class_dev_iter *iter;
993 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
995 return ERR_PTR(-ENOMEM);
997 seqf->private = iter;
998 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
1000 dev = class_dev_iter_next(iter);
1005 return dev_to_disk(dev);
1008 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
1013 dev = class_dev_iter_next(seqf->private);
1015 return dev_to_disk(dev);
1020 static void disk_seqf_stop(struct seq_file *seqf, void *v)
1022 struct class_dev_iter *iter = seqf->private;
1024 /* stop is called even after start failed :-( */
1026 class_dev_iter_exit(iter);
1028 seqf->private = NULL;
1032 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
1036 p = disk_seqf_start(seqf, pos);
1037 if (!IS_ERR_OR_NULL(p) && !*pos)
1038 seq_puts(seqf, "major minor #blocks name\n\n");
1042 static int show_partition(struct seq_file *seqf, void *v)
1044 struct gendisk *sgp = v;
1045 struct disk_part_iter piter;
1046 struct hd_struct *part;
1047 char buf[BDEVNAME_SIZE];
1049 /* Don't show non-partitionable removeable devices or empty devices */
1050 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
1051 (sgp->flags & GENHD_FL_REMOVABLE)))
1053 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
1056 /* show the full disk and all non-0 size partitions of it */
1057 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
1058 while ((part = disk_part_iter_next(&piter)))
1059 seq_printf(seqf, "%4d %7d %10llu %s\n",
1060 MAJOR(part_devt(part)), MINOR(part_devt(part)),
1061 (unsigned long long)part_nr_sects_read(part) >> 1,
1062 disk_name(sgp, part->partno, buf));
1063 disk_part_iter_exit(&piter);
1068 static const struct seq_operations partitions_op = {
1069 .start = show_partition_start,
1070 .next = disk_seqf_next,
1071 .stop = disk_seqf_stop,
1072 .show = show_partition
1077 static struct kobject *base_probe(dev_t devt, int *partno, void *data)
1079 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
1080 /* Make old-style 2.4 aliases work */
1081 request_module("block-major-%d", MAJOR(devt));
1085 static int __init genhd_device_init(void)
1089 block_class.dev_kobj = sysfs_dev_block_kobj;
1090 error = class_register(&block_class);
1091 if (unlikely(error))
1093 bdev_map = kobj_map_init(base_probe, &block_class_lock);
1096 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
1098 /* create top-level block dir */
1099 if (!sysfs_deprecated)
1100 block_depr = kobject_create_and_add("block", NULL);
1104 subsys_initcall(genhd_device_init);
1106 static ssize_t disk_range_show(struct device *dev,
1107 struct device_attribute *attr, char *buf)
1109 struct gendisk *disk = dev_to_disk(dev);
1111 return sprintf(buf, "%d\n", disk->minors);
1114 static ssize_t disk_ext_range_show(struct device *dev,
1115 struct device_attribute *attr, char *buf)
1117 struct gendisk *disk = dev_to_disk(dev);
1119 return sprintf(buf, "%d\n", disk_max_parts(disk));
1122 static ssize_t disk_removable_show(struct device *dev,
1123 struct device_attribute *attr, char *buf)
1125 struct gendisk *disk = dev_to_disk(dev);
1127 return sprintf(buf, "%d\n",
1128 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
1131 static ssize_t disk_hidden_show(struct device *dev,
1132 struct device_attribute *attr, char *buf)
1134 struct gendisk *disk = dev_to_disk(dev);
1136 return sprintf(buf, "%d\n",
1137 (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
1140 static ssize_t disk_ro_show(struct device *dev,
1141 struct device_attribute *attr, char *buf)
1143 struct gendisk *disk = dev_to_disk(dev);
1145 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
1148 static ssize_t disk_capability_show(struct device *dev,
1149 struct device_attribute *attr, char *buf)
1151 struct gendisk *disk = dev_to_disk(dev);
1153 return sprintf(buf, "%x\n", disk->flags);
1156 static ssize_t disk_alignment_offset_show(struct device *dev,
1157 struct device_attribute *attr,
1160 struct gendisk *disk = dev_to_disk(dev);
1162 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
1165 static ssize_t disk_discard_alignment_show(struct device *dev,
1166 struct device_attribute *attr,
1169 struct gendisk *disk = dev_to_disk(dev);
1171 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
1174 static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
1175 static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
1176 static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
1177 static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
1178 static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
1179 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
1180 static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
1181 static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
1182 static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
1183 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
1184 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
1185 static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
1186 #ifdef CONFIG_FAIL_MAKE_REQUEST
1187 static struct device_attribute dev_attr_fail =
1188 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
1190 #ifdef CONFIG_FAIL_IO_TIMEOUT
1191 static struct device_attribute dev_attr_fail_timeout =
1192 __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
1195 static struct attribute *disk_attrs[] = {
1196 &dev_attr_range.attr,
1197 &dev_attr_ext_range.attr,
1198 &dev_attr_removable.attr,
1199 &dev_attr_hidden.attr,
1201 &dev_attr_size.attr,
1202 &dev_attr_alignment_offset.attr,
1203 &dev_attr_discard_alignment.attr,
1204 &dev_attr_capability.attr,
1205 &dev_attr_stat.attr,
1206 &dev_attr_inflight.attr,
1207 &dev_attr_badblocks.attr,
1208 #ifdef CONFIG_FAIL_MAKE_REQUEST
1209 &dev_attr_fail.attr,
1211 #ifdef CONFIG_FAIL_IO_TIMEOUT
1212 &dev_attr_fail_timeout.attr,
1217 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1219 struct device *dev = container_of(kobj, typeof(*dev), kobj);
1220 struct gendisk *disk = dev_to_disk(dev);
1222 if (a == &dev_attr_badblocks.attr && !disk->bb)
1227 static struct attribute_group disk_attr_group = {
1228 .attrs = disk_attrs,
1229 .is_visible = disk_visible,
1232 static const struct attribute_group *disk_attr_groups[] = {
1238 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1239 * @disk: disk to replace part_tbl for
1240 * @new_ptbl: new part_tbl to install
1242 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1243 * original ptbl is freed using RCU callback.
1246 * Matching bd_mutex locked or the caller is the only user of @disk.
1248 static void disk_replace_part_tbl(struct gendisk *disk,
1249 struct disk_part_tbl *new_ptbl)
1251 struct disk_part_tbl *old_ptbl =
1252 rcu_dereference_protected(disk->part_tbl, 1);
1254 rcu_assign_pointer(disk->part_tbl, new_ptbl);
1257 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
1258 kfree_rcu(old_ptbl, rcu_head);
1263 * disk_expand_part_tbl - expand disk->part_tbl
1264 * @disk: disk to expand part_tbl for
1265 * @partno: expand such that this partno can fit in
1267 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1268 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1271 * Matching bd_mutex locked or the caller is the only user of @disk.
1275 * 0 on success, -errno on failure.
1277 int disk_expand_part_tbl(struct gendisk *disk, int partno)
1279 struct disk_part_tbl *old_ptbl =
1280 rcu_dereference_protected(disk->part_tbl, 1);
1281 struct disk_part_tbl *new_ptbl;
1282 int len = old_ptbl ? old_ptbl->len : 0;
1286 * check for int overflow, since we can get here from blkpg_ioctl()
1287 * with a user passed 'partno'.
1289 target = partno + 1;
1293 /* disk_max_parts() is zero during initialization, ignore if so */
1294 if (disk_max_parts(disk) && target > disk_max_parts(disk))
1300 new_ptbl = kzalloc_node(struct_size(new_ptbl, part, target), GFP_KERNEL,
1305 new_ptbl->len = target;
1307 for (i = 0; i < len; i++)
1308 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
1310 disk_replace_part_tbl(disk, new_ptbl);
1314 static void disk_release(struct device *dev)
1316 struct gendisk *disk = dev_to_disk(dev);
1318 blk_free_devt(dev->devt);
1319 disk_release_events(disk);
1320 kfree(disk->random);
1321 disk_replace_part_tbl(disk, NULL);
1322 hd_free_part(&disk->part0);
1324 blk_put_queue(disk->queue);
1327 struct class block_class = {
1331 static char *block_devnode(struct device *dev, umode_t *mode,
1332 kuid_t *uid, kgid_t *gid)
1334 struct gendisk *disk = dev_to_disk(dev);
1337 return disk->devnode(disk, mode);
1341 static const struct device_type disk_type = {
1343 .groups = disk_attr_groups,
1344 .release = disk_release,
1345 .devnode = block_devnode,
1348 #ifdef CONFIG_PROC_FS
1350 * aggregate disk stat collector. Uses the same stats that the sysfs
1351 * entries do, above, but makes them available through one seq_file.
1353 * The output looks suspiciously like /proc/partitions with a bunch of
1356 static int diskstats_show(struct seq_file *seqf, void *v)
1358 struct gendisk *gp = v;
1359 struct disk_part_iter piter;
1360 struct hd_struct *hd;
1361 char buf[BDEVNAME_SIZE];
1362 unsigned int inflight;
1365 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1366 seq_puts(seqf, "major minor name"
1367 " rio rmerge rsect ruse wio wmerge "
1368 "wsect wuse running use aveq"
1372 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
1373 while ((hd = disk_part_iter_next(&piter))) {
1374 inflight = part_in_flight(gp->queue, hd);
1375 seq_printf(seqf, "%4d %7d %s "
1380 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1381 disk_name(gp, hd->partno, buf),
1382 part_stat_read(hd, ios[STAT_READ]),
1383 part_stat_read(hd, merges[STAT_READ]),
1384 part_stat_read(hd, sectors[STAT_READ]),
1385 (unsigned int)part_stat_read_msecs(hd, STAT_READ),
1386 part_stat_read(hd, ios[STAT_WRITE]),
1387 part_stat_read(hd, merges[STAT_WRITE]),
1388 part_stat_read(hd, sectors[STAT_WRITE]),
1389 (unsigned int)part_stat_read_msecs(hd, STAT_WRITE),
1391 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1392 jiffies_to_msecs(part_stat_read(hd, time_in_queue)),
1393 part_stat_read(hd, ios[STAT_DISCARD]),
1394 part_stat_read(hd, merges[STAT_DISCARD]),
1395 part_stat_read(hd, sectors[STAT_DISCARD]),
1396 (unsigned int)part_stat_read_msecs(hd, STAT_DISCARD)
1399 disk_part_iter_exit(&piter);
1404 static const struct seq_operations diskstats_op = {
1405 .start = disk_seqf_start,
1406 .next = disk_seqf_next,
1407 .stop = disk_seqf_stop,
1408 .show = diskstats_show
1411 static int __init proc_genhd_init(void)
1413 proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1414 proc_create_seq("partitions", 0, NULL, &partitions_op);
1417 module_init(proc_genhd_init);
1418 #endif /* CONFIG_PROC_FS */
1420 dev_t blk_lookup_devt(const char *name, int partno)
1422 dev_t devt = MKDEV(0, 0);
1423 struct class_dev_iter iter;
1426 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1427 while ((dev = class_dev_iter_next(&iter))) {
1428 struct gendisk *disk = dev_to_disk(dev);
1429 struct hd_struct *part;
1431 if (strcmp(dev_name(dev), name))
1434 if (partno < disk->minors) {
1435 /* We need to return the right devno, even
1436 * if the partition doesn't exist yet.
1438 devt = MKDEV(MAJOR(dev->devt),
1439 MINOR(dev->devt) + partno);
1442 part = disk_get_part(disk, partno);
1444 devt = part_devt(part);
1445 disk_put_part(part);
1448 disk_put_part(part);
1450 class_dev_iter_exit(&iter);
1453 EXPORT_SYMBOL(blk_lookup_devt);
1455 struct gendisk *__alloc_disk_node(int minors, int node_id)
1457 struct gendisk *disk;
1458 struct disk_part_tbl *ptbl;
1460 if (minors > DISK_MAX_PARTS) {
1462 "block: can't allocate more than %d partitions\n",
1464 minors = DISK_MAX_PARTS;
1467 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1469 if (!init_part_stats(&disk->part0)) {
1473 init_rwsem(&disk->lookup_sem);
1474 disk->node_id = node_id;
1475 if (disk_expand_part_tbl(disk, 0)) {
1476 free_part_stats(&disk->part0);
1480 ptbl = rcu_dereference_protected(disk->part_tbl, 1);
1481 rcu_assign_pointer(ptbl->part[0], &disk->part0);
1484 * set_capacity() and get_capacity() currently don't use
1485 * seqcounter to read/update the part0->nr_sects. Still init
1486 * the counter as we can read the sectors in IO submission
1487 * patch using seqence counters.
1489 * TODO: Ideally set_capacity() and get_capacity() should be
1490 * converted to make use of bd_mutex and sequence counters.
1492 seqcount_init(&disk->part0.nr_sects_seq);
1493 if (hd_ref_init(&disk->part0)) {
1494 hd_free_part(&disk->part0);
1499 disk->minors = minors;
1500 rand_initialize_disk(disk);
1501 disk_to_dev(disk)->class = &block_class;
1502 disk_to_dev(disk)->type = &disk_type;
1503 device_initialize(disk_to_dev(disk));
1507 EXPORT_SYMBOL(__alloc_disk_node);
1509 struct kobject *get_disk_and_module(struct gendisk *disk)
1511 struct module *owner;
1512 struct kobject *kobj;
1516 owner = disk->fops->owner;
1517 if (owner && !try_module_get(owner))
1519 kobj = kobject_get_unless_zero(&disk_to_dev(disk)->kobj);
1527 EXPORT_SYMBOL(get_disk_and_module);
1529 void put_disk(struct gendisk *disk)
1532 kobject_put(&disk_to_dev(disk)->kobj);
1534 EXPORT_SYMBOL(put_disk);
1537 * This is a counterpart of get_disk_and_module() and thus also of
1540 void put_disk_and_module(struct gendisk *disk)
1543 struct module *owner = disk->fops->owner;
1549 EXPORT_SYMBOL(put_disk_and_module);
1551 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1553 char event[] = "DISK_RO=1";
1554 char *envp[] = { event, NULL };
1558 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1561 void set_device_ro(struct block_device *bdev, int flag)
1563 bdev->bd_part->policy = flag;
1566 EXPORT_SYMBOL(set_device_ro);
1568 void set_disk_ro(struct gendisk *disk, int flag)
1570 struct disk_part_iter piter;
1571 struct hd_struct *part;
1573 if (disk->part0.policy != flag) {
1574 set_disk_ro_uevent(disk, flag);
1575 disk->part0.policy = flag;
1578 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
1579 while ((part = disk_part_iter_next(&piter)))
1580 part->policy = flag;
1581 disk_part_iter_exit(&piter);
1584 EXPORT_SYMBOL(set_disk_ro);
1586 int bdev_read_only(struct block_device *bdev)
1590 return bdev->bd_part->policy;
1593 EXPORT_SYMBOL(bdev_read_only);
1595 int invalidate_partition(struct gendisk *disk, int partno)
1598 struct block_device *bdev = bdget_disk(disk, partno);
1601 res = __invalidate_device(bdev, true);
1607 EXPORT_SYMBOL(invalidate_partition);
1610 * Disk events - monitor disk events like media change and eject request.
1612 struct disk_events {
1613 struct list_head node; /* all disk_event's */
1614 struct gendisk *disk; /* the associated disk */
1617 struct mutex block_mutex; /* protects blocking */
1618 int block; /* event blocking depth */
1619 unsigned int pending; /* events already sent out */
1620 unsigned int clearing; /* events being cleared */
1622 long poll_msecs; /* interval, -1 for default */
1623 struct delayed_work dwork;
1626 static const char *disk_events_strs[] = {
1627 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1628 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1631 static char *disk_uevents[] = {
1632 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1633 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1636 /* list of all disk_events */
1637 static DEFINE_MUTEX(disk_events_mutex);
1638 static LIST_HEAD(disk_events);
1640 /* disable in-kernel polling by default */
1641 static unsigned long disk_events_dfl_poll_msecs;
1643 static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1645 struct disk_events *ev = disk->ev;
1646 long intv_msecs = 0;
1649 * If device-specific poll interval is set, always use it. If
1650 * the default is being used, poll if the POLL flag is set.
1652 if (ev->poll_msecs >= 0)
1653 intv_msecs = ev->poll_msecs;
1654 else if (disk->event_flags & DISK_EVENT_FLAG_POLL)
1655 intv_msecs = disk_events_dfl_poll_msecs;
1657 return msecs_to_jiffies(intv_msecs);
1661 * disk_block_events - block and flush disk event checking
1662 * @disk: disk to block events for
1664 * On return from this function, it is guaranteed that event checking
1665 * isn't in progress and won't happen until unblocked by
1666 * disk_unblock_events(). Events blocking is counted and the actual
1667 * unblocking happens after the matching number of unblocks are done.
1669 * Note that this intentionally does not block event checking from
1670 * disk_clear_events().
1675 void disk_block_events(struct gendisk *disk)
1677 struct disk_events *ev = disk->ev;
1678 unsigned long flags;
1685 * Outer mutex ensures that the first blocker completes canceling
1686 * the event work before further blockers are allowed to finish.
1688 mutex_lock(&ev->block_mutex);
1690 spin_lock_irqsave(&ev->lock, flags);
1691 cancel = !ev->block++;
1692 spin_unlock_irqrestore(&ev->lock, flags);
1695 cancel_delayed_work_sync(&disk->ev->dwork);
1697 mutex_unlock(&ev->block_mutex);
1700 static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1702 struct disk_events *ev = disk->ev;
1704 unsigned long flags;
1706 spin_lock_irqsave(&ev->lock, flags);
1708 if (WARN_ON_ONCE(ev->block <= 0))
1714 intv = disk_events_poll_jiffies(disk);
1716 queue_delayed_work(system_freezable_power_efficient_wq,
1719 queue_delayed_work(system_freezable_power_efficient_wq,
1722 spin_unlock_irqrestore(&ev->lock, flags);
1726 * disk_unblock_events - unblock disk event checking
1727 * @disk: disk to unblock events for
1729 * Undo disk_block_events(). When the block count reaches zero, it
1730 * starts events polling if configured.
1733 * Don't care. Safe to call from irq context.
1735 void disk_unblock_events(struct gendisk *disk)
1738 __disk_unblock_events(disk, false);
1742 * disk_flush_events - schedule immediate event checking and flushing
1743 * @disk: disk to check and flush events for
1744 * @mask: events to flush
1746 * Schedule immediate event checking on @disk if not blocked. Events in
1747 * @mask are scheduled to be cleared from the driver. Note that this
1748 * doesn't clear the events from @disk->ev.
1751 * If @mask is non-zero must be called with bdev->bd_mutex held.
1753 void disk_flush_events(struct gendisk *disk, unsigned int mask)
1755 struct disk_events *ev = disk->ev;
1760 spin_lock_irq(&ev->lock);
1761 ev->clearing |= mask;
1763 mod_delayed_work(system_freezable_power_efficient_wq,
1765 spin_unlock_irq(&ev->lock);
1769 * disk_clear_events - synchronously check, clear and return pending events
1770 * @disk: disk to fetch and clear events from
1771 * @mask: mask of events to be fetched and cleared
1773 * Disk events are synchronously checked and pending events in @mask
1774 * are cleared and returned. This ignores the block count.
1779 unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1781 const struct block_device_operations *bdops = disk->fops;
1782 struct disk_events *ev = disk->ev;
1783 unsigned int pending;
1784 unsigned int clearing = mask;
1787 /* for drivers still using the old ->media_changed method */
1788 if ((mask & DISK_EVENT_MEDIA_CHANGE) &&
1789 bdops->media_changed && bdops->media_changed(disk))
1790 return DISK_EVENT_MEDIA_CHANGE;
1794 disk_block_events(disk);
1797 * store the union of mask and ev->clearing on the stack so that the
1798 * race with disk_flush_events does not cause ambiguity (ev->clearing
1799 * can still be modified even if events are blocked).
1801 spin_lock_irq(&ev->lock);
1802 clearing |= ev->clearing;
1804 spin_unlock_irq(&ev->lock);
1806 disk_check_events(ev, &clearing);
1808 * if ev->clearing is not 0, the disk_flush_events got called in the
1809 * middle of this function, so we want to run the workfn without delay.
1811 __disk_unblock_events(disk, ev->clearing ? true : false);
1813 /* then, fetch and clear pending events */
1814 spin_lock_irq(&ev->lock);
1815 pending = ev->pending & mask;
1816 ev->pending &= ~mask;
1817 spin_unlock_irq(&ev->lock);
1818 WARN_ON_ONCE(clearing & mask);
1824 * Separate this part out so that a different pointer for clearing_ptr can be
1825 * passed in for disk_clear_events.
1827 static void disk_events_workfn(struct work_struct *work)
1829 struct delayed_work *dwork = to_delayed_work(work);
1830 struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
1832 disk_check_events(ev, &ev->clearing);
1835 static void disk_check_events(struct disk_events *ev,
1836 unsigned int *clearing_ptr)
1838 struct gendisk *disk = ev->disk;
1839 char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
1840 unsigned int clearing = *clearing_ptr;
1841 unsigned int events;
1843 int nr_events = 0, i;
1846 events = disk->fops->check_events(disk, clearing);
1848 /* accumulate pending events and schedule next poll if necessary */
1849 spin_lock_irq(&ev->lock);
1851 events &= ~ev->pending;
1852 ev->pending |= events;
1853 *clearing_ptr &= ~clearing;
1855 intv = disk_events_poll_jiffies(disk);
1856 if (!ev->block && intv)
1857 queue_delayed_work(system_freezable_power_efficient_wq,
1860 spin_unlock_irq(&ev->lock);
1863 * Tell userland about new events. Only the events listed in
1864 * @disk->events are reported, and only if DISK_EVENT_FLAG_UEVENT
1865 * is set. Otherwise, events are processed internally but never
1866 * get reported to userland.
1868 for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
1869 if ((events & disk->events & (1 << i)) &&
1870 (disk->event_flags & DISK_EVENT_FLAG_UEVENT))
1871 envp[nr_events++] = disk_uevents[i];
1874 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1878 * A disk events enabled device has the following sysfs nodes under
1879 * its /sys/block/X/ directory.
1881 * events : list of all supported events
1882 * events_async : list of events which can be detected w/o polling
1883 * (always empty, only for backwards compatibility)
1884 * events_poll_msecs : polling interval, 0: disable, -1: system default
1886 static ssize_t __disk_events_show(unsigned int events, char *buf)
1888 const char *delim = "";
1892 for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1893 if (events & (1 << i)) {
1894 pos += sprintf(buf + pos, "%s%s",
1895 delim, disk_events_strs[i]);
1899 pos += sprintf(buf + pos, "\n");
1903 static ssize_t disk_events_show(struct device *dev,
1904 struct device_attribute *attr, char *buf)
1906 struct gendisk *disk = dev_to_disk(dev);
1908 if (!(disk->event_flags & DISK_EVENT_FLAG_UEVENT))
1911 return __disk_events_show(disk->events, buf);
1914 static ssize_t disk_events_async_show(struct device *dev,
1915 struct device_attribute *attr, char *buf)
1920 static ssize_t disk_events_poll_msecs_show(struct device *dev,
1921 struct device_attribute *attr,
1924 struct gendisk *disk = dev_to_disk(dev);
1927 return sprintf(buf, "-1\n");
1929 return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1932 static ssize_t disk_events_poll_msecs_store(struct device *dev,
1933 struct device_attribute *attr,
1934 const char *buf, size_t count)
1936 struct gendisk *disk = dev_to_disk(dev);
1939 if (!count || !sscanf(buf, "%ld", &intv))
1942 if (intv < 0 && intv != -1)
1948 disk_block_events(disk);
1949 disk->ev->poll_msecs = intv;
1950 __disk_unblock_events(disk, true);
1955 static const DEVICE_ATTR(events, 0444, disk_events_show, NULL);
1956 static const DEVICE_ATTR(events_async, 0444, disk_events_async_show, NULL);
1957 static const DEVICE_ATTR(events_poll_msecs, 0644,
1958 disk_events_poll_msecs_show,
1959 disk_events_poll_msecs_store);
1961 static const struct attribute *disk_events_attrs[] = {
1962 &dev_attr_events.attr,
1963 &dev_attr_events_async.attr,
1964 &dev_attr_events_poll_msecs.attr,
1969 * The default polling interval can be specified by the kernel
1970 * parameter block.events_dfl_poll_msecs which defaults to 0
1971 * (disable). This can also be modified runtime by writing to
1972 * /sys/module/block/events_dfl_poll_msecs.
1974 static int disk_events_set_dfl_poll_msecs(const char *val,
1975 const struct kernel_param *kp)
1977 struct disk_events *ev;
1980 ret = param_set_ulong(val, kp);
1984 mutex_lock(&disk_events_mutex);
1986 list_for_each_entry(ev, &disk_events, node)
1987 disk_flush_events(ev->disk, 0);
1989 mutex_unlock(&disk_events_mutex);
1994 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1995 .set = disk_events_set_dfl_poll_msecs,
1996 .get = param_get_ulong,
1999 #undef MODULE_PARAM_PREFIX
2000 #define MODULE_PARAM_PREFIX "block."
2002 module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
2003 &disk_events_dfl_poll_msecs, 0644);
2006 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
2008 static void disk_alloc_events(struct gendisk *disk)
2010 struct disk_events *ev;
2012 if (!disk->fops->check_events || !disk->events)
2015 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
2017 pr_warn("%s: failed to initialize events\n", disk->disk_name);
2021 INIT_LIST_HEAD(&ev->node);
2023 spin_lock_init(&ev->lock);
2024 mutex_init(&ev->block_mutex);
2026 ev->poll_msecs = -1;
2027 INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
2032 static void disk_add_events(struct gendisk *disk)
2034 /* FIXME: error handling */
2035 if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0)
2036 pr_warn("%s: failed to create sysfs files for events\n",
2042 mutex_lock(&disk_events_mutex);
2043 list_add_tail(&disk->ev->node, &disk_events);
2044 mutex_unlock(&disk_events_mutex);
2047 * Block count is initialized to 1 and the following initial
2048 * unblock kicks it into action.
2050 __disk_unblock_events(disk, true);
2053 static void disk_del_events(struct gendisk *disk)
2056 disk_block_events(disk);
2058 mutex_lock(&disk_events_mutex);
2059 list_del_init(&disk->ev->node);
2060 mutex_unlock(&disk_events_mutex);
2063 sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
2066 static void disk_release_events(struct gendisk *disk)
2068 /* the block count should be 1 from disk_del_events() */
2069 WARN_ON_ONCE(disk->ev && disk->ev->block != 1);