1 // SPDX-License-Identifier: GPL-2.0
5 * Portions Copyright (C) 2020 Christoph Hellwig
8 #include <linux/module.h>
9 #include <linux/ctype.h>
11 #include <linux/genhd.h>
12 #include <linux/kdev_t.h>
13 #include <linux/kernel.h>
14 #include <linux/blkdev.h>
15 #include <linux/backing-dev.h>
16 #include <linux/init.h>
17 #include <linux/spinlock.h>
18 #include <linux/proc_fs.h>
19 #include <linux/seq_file.h>
20 #include <linux/slab.h>
21 #include <linux/kmod.h>
22 #include <linux/mutex.h>
23 #include <linux/idr.h>
24 #include <linux/log2.h>
25 #include <linux/pm_runtime.h>
26 #include <linux/badblocks.h>
30 static struct kobject *block_depr;
32 DECLARE_RWSEM(bdev_lookup_sem);
34 /* for extended dynamic devt allocation, currently only one major is used */
35 #define NR_EXT_DEVT (1 << MINORBITS)
36 static DEFINE_IDA(ext_devt_ida);
38 static void disk_check_events(struct disk_events *ev,
39 unsigned int *clearing_ptr);
40 static void disk_alloc_events(struct gendisk *disk);
41 static void disk_add_events(struct gendisk *disk);
42 static void disk_del_events(struct gendisk *disk);
43 static void disk_release_events(struct gendisk *disk);
45 void set_capacity(struct gendisk *disk, sector_t sectors)
47 struct block_device *bdev = disk->part0;
49 spin_lock(&bdev->bd_size_lock);
50 i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
51 spin_unlock(&bdev->bd_size_lock);
53 EXPORT_SYMBOL(set_capacity);
56 * Set disk capacity and notify if the size is not currently zero and will not
57 * be set to zero. Returns true if a uevent was sent, otherwise false.
59 bool set_capacity_and_notify(struct gendisk *disk, sector_t size)
61 sector_t capacity = get_capacity(disk);
62 char *envp[] = { "RESIZE=1", NULL };
64 set_capacity(disk, size);
67 * Only print a message and send a uevent if the gendisk is user visible
68 * and alive. This avoids spamming the log and udev when setting the
69 * initial capacity during probing.
71 if (size == capacity ||
72 (disk->flags & (GENHD_FL_UP | GENHD_FL_HIDDEN)) != GENHD_FL_UP)
75 pr_info("%s: detected capacity change from %lld to %lld\n",
76 disk->disk_name, capacity, size);
79 * Historically we did not send a uevent for changes to/from an empty
82 if (!capacity || !size)
84 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
87 EXPORT_SYMBOL_GPL(set_capacity_and_notify);
90 * Format the device name of the indicated disk into the supplied buffer and
91 * return a pointer to that same buffer for convenience.
93 char *disk_name(struct gendisk *hd, int partno, char *buf)
96 snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name);
97 else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1]))
98 snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno);
100 snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno);
105 const char *bdevname(struct block_device *bdev, char *buf)
107 return disk_name(bdev->bd_disk, bdev->bd_partno, buf);
109 EXPORT_SYMBOL(bdevname);
111 static void part_stat_read_all(struct block_device *part,
112 struct disk_stats *stat)
116 memset(stat, 0, sizeof(struct disk_stats));
117 for_each_possible_cpu(cpu) {
118 struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu);
121 for (group = 0; group < NR_STAT_GROUPS; group++) {
122 stat->nsecs[group] += ptr->nsecs[group];
123 stat->sectors[group] += ptr->sectors[group];
124 stat->ios[group] += ptr->ios[group];
125 stat->merges[group] += ptr->merges[group];
128 stat->io_ticks += ptr->io_ticks;
132 static unsigned int part_in_flight(struct block_device *part)
134 unsigned int inflight = 0;
137 for_each_possible_cpu(cpu) {
138 inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) +
139 part_stat_local_read_cpu(part, in_flight[1], cpu);
141 if ((int)inflight < 0)
147 static void part_in_flight_rw(struct block_device *part,
148 unsigned int inflight[2])
154 for_each_possible_cpu(cpu) {
155 inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu);
156 inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu);
158 if ((int)inflight[0] < 0)
160 if ((int)inflight[1] < 0)
165 * disk_part_iter_init - initialize partition iterator
166 * @piter: iterator to initialize
167 * @disk: disk to iterate over
168 * @flags: DISK_PITER_* flags
170 * Initialize @piter so that it iterates over partitions of @disk.
175 void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk,
180 if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0))
184 piter->flags = flags;
188 * disk_part_iter_next - proceed iterator to the next partition and return it
189 * @piter: iterator of interest
191 * Proceed @piter to the next partition and return it.
196 struct block_device *disk_part_iter_next(struct disk_part_iter *piter)
198 struct block_device *part;
201 /* put the last partition */
202 disk_part_iter_exit(piter);
205 xa_for_each_start(&piter->disk->part_tbl, idx, part, piter->idx) {
206 if (!bdev_nr_sectors(part) &&
207 !(piter->flags & DISK_PITER_INCL_EMPTY) &&
208 !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 &&
212 piter->part = bdgrab(part);
215 piter->idx = idx + 1;
224 * disk_part_iter_exit - finish up partition iteration
225 * @piter: iter of interest
227 * Called when iteration is over. Cleans up @piter.
232 void disk_part_iter_exit(struct disk_part_iter *piter)
240 * Can be deleted altogether. Later.
243 #define BLKDEV_MAJOR_HASH_SIZE 255
244 static struct blk_major_name {
245 struct blk_major_name *next;
248 void (*probe)(dev_t devt);
249 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
250 static DEFINE_MUTEX(major_names_lock);
252 /* index in the above - for now: assume no multimajor ranges */
253 static inline int major_to_index(unsigned major)
255 return major % BLKDEV_MAJOR_HASH_SIZE;
258 #ifdef CONFIG_PROC_FS
259 void blkdev_show(struct seq_file *seqf, off_t offset)
261 struct blk_major_name *dp;
263 mutex_lock(&major_names_lock);
264 for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
265 if (dp->major == offset)
266 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
267 mutex_unlock(&major_names_lock);
269 #endif /* CONFIG_PROC_FS */
272 * __register_blkdev - register a new block device
274 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
275 * @major = 0, try to allocate any unused major number.
276 * @name: the name of the new block device as a zero terminated string
277 * @probe: allback that is called on access to any minor number of @major
279 * The @name must be unique within the system.
281 * The return value depends on the @major input parameter:
283 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
284 * then the function returns zero on success, or a negative error code
285 * - if any unused major number was requested with @major = 0 parameter
286 * then the return value is the allocated major number in range
287 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
289 * See Documentation/admin-guide/devices.txt for the list of allocated
292 * Use register_blkdev instead for any new code.
294 int __register_blkdev(unsigned int major, const char *name,
295 void (*probe)(dev_t devt))
297 struct blk_major_name **n, *p;
300 mutex_lock(&major_names_lock);
304 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
305 if (major_names[index] == NULL)
310 printk("%s: failed to get major for %s\n",
319 if (major >= BLKDEV_MAJOR_MAX) {
320 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
321 __func__, major, BLKDEV_MAJOR_MAX-1, name);
327 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
335 strlcpy(p->name, name, sizeof(p->name));
337 index = major_to_index(major);
339 for (n = &major_names[index]; *n; n = &(*n)->next) {
340 if ((*n)->major == major)
349 printk("register_blkdev: cannot get major %u for %s\n",
354 mutex_unlock(&major_names_lock);
357 EXPORT_SYMBOL(__register_blkdev);
359 void unregister_blkdev(unsigned int major, const char *name)
361 struct blk_major_name **n;
362 struct blk_major_name *p = NULL;
363 int index = major_to_index(major);
365 mutex_lock(&major_names_lock);
366 for (n = &major_names[index]; *n; n = &(*n)->next)
367 if ((*n)->major == major)
369 if (!*n || strcmp((*n)->name, name)) {
375 mutex_unlock(&major_names_lock);
379 EXPORT_SYMBOL(unregister_blkdev);
382 * blk_mangle_minor - scatter minor numbers apart
383 * @minor: minor number to mangle
385 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
386 * is enabled. Mangling twice gives the original value.
394 static int blk_mangle_minor(int minor)
396 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
399 for (i = 0; i < MINORBITS / 2; i++) {
400 int low = minor & (1 << i);
401 int high = minor & (1 << (MINORBITS - 1 - i));
402 int distance = MINORBITS - 1 - 2 * i;
404 minor ^= low | high; /* clear both bits */
405 low <<= distance; /* swap the positions */
407 minor |= low | high; /* and set */
414 * blk_alloc_devt - allocate a dev_t for a block device
415 * @bdev: block device to allocate dev_t for
416 * @devt: out parameter for resulting dev_t
418 * Allocate a dev_t for block device.
421 * 0 on success, allocated dev_t is returned in *@devt. -errno on
427 int blk_alloc_devt(struct block_device *bdev, dev_t *devt)
429 struct gendisk *disk = bdev->bd_disk;
432 /* in consecutive minor range? */
433 if (bdev->bd_partno < disk->minors) {
434 *devt = MKDEV(disk->major, disk->first_minor + bdev->bd_partno);
438 idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT, GFP_KERNEL);
440 return idx == -ENOSPC ? -EBUSY : idx;
442 *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
447 * blk_free_devt - free a dev_t
448 * @devt: dev_t to free
450 * Free @devt which was allocated using blk_alloc_devt().
455 void blk_free_devt(dev_t devt)
457 if (MAJOR(devt) == BLOCK_EXT_MAJOR)
458 ida_free(&ext_devt_ida, blk_mangle_minor(MINOR(devt)));
461 static char *bdevt_str(dev_t devt, char *buf)
463 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
464 char tbuf[BDEVT_SIZE];
465 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
466 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
468 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
473 void disk_uevent(struct gendisk *disk, enum kobject_action action)
475 struct disk_part_iter piter;
476 struct block_device *part;
478 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY_PART0);
479 while ((part = disk_part_iter_next(&piter)))
480 kobject_uevent(bdev_kobj(part), action);
481 disk_part_iter_exit(&piter);
483 EXPORT_SYMBOL_GPL(disk_uevent);
485 static void disk_scan_partitions(struct gendisk *disk)
487 struct block_device *bdev;
489 if (!get_capacity(disk) || !disk_part_scan_enabled(disk))
492 set_bit(GD_NEED_PART_SCAN, &disk->state);
493 bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL);
495 blkdev_put(bdev, FMODE_READ);
498 static void register_disk(struct device *parent, struct gendisk *disk,
499 const struct attribute_group **groups)
501 struct device *ddev = disk_to_dev(disk);
504 ddev->parent = parent;
506 dev_set_name(ddev, "%s", disk->disk_name);
508 /* delay uevents, until we scanned partition table */
509 dev_set_uevent_suppress(ddev, 1);
512 WARN_ON(ddev->groups);
513 ddev->groups = groups;
515 if (device_add(ddev))
517 if (!sysfs_deprecated) {
518 err = sysfs_create_link(block_depr, &ddev->kobj,
519 kobject_name(&ddev->kobj));
527 * avoid probable deadlock caused by allocating memory with
528 * GFP_KERNEL in runtime_resume callback of its all ancestor
531 pm_runtime_set_memalloc_noio(ddev, true);
533 disk->part0->bd_holder_dir =
534 kobject_create_and_add("holders", &ddev->kobj);
535 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
537 if (disk->flags & GENHD_FL_HIDDEN)
540 disk_scan_partitions(disk);
542 /* announce the disk and partitions after all partitions are created */
543 dev_set_uevent_suppress(ddev, 0);
544 disk_uevent(disk, KOBJ_ADD);
546 if (disk->queue->backing_dev_info->dev) {
547 err = sysfs_create_link(&ddev->kobj,
548 &disk->queue->backing_dev_info->dev->kobj,
555 * __device_add_disk - add disk information to kernel list
556 * @parent: parent device for the disk
557 * @disk: per-device partitioning information
558 * @groups: Additional per-device sysfs groups
559 * @register_queue: register the queue if set to true
561 * This function registers the partitioning information in @disk
564 * FIXME: error handling
566 static void __device_add_disk(struct device *parent, struct gendisk *disk,
567 const struct attribute_group **groups,
574 * The disk queue should now be all set with enough information about
575 * the device for the elevator code to pick an adequate default
576 * elevator if one is needed, that is, for devices requesting queue
580 elevator_init_mq(disk->queue);
582 /* minors == 0 indicates to use ext devt from part0 and should
583 * be accompanied with EXT_DEVT flag. Make sure all
584 * parameters make sense.
586 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
587 WARN_ON(!disk->minors &&
588 !(disk->flags & (GENHD_FL_EXT_DEVT | GENHD_FL_HIDDEN)));
590 disk->flags |= GENHD_FL_UP;
592 retval = blk_alloc_devt(disk->part0, &devt);
597 disk->major = MAJOR(devt);
598 disk->first_minor = MINOR(devt);
600 disk_alloc_events(disk);
602 if (disk->flags & GENHD_FL_HIDDEN) {
604 * Don't let hidden disks show up in /proc/partitions,
605 * and don't bother scanning for partitions either.
607 disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
608 disk->flags |= GENHD_FL_NO_PART_SCAN;
610 struct backing_dev_info *bdi = disk->queue->backing_dev_info;
611 struct device *dev = disk_to_dev(disk);
614 /* Register BDI before referencing it from bdev */
616 ret = bdi_register(bdi, "%u:%u", MAJOR(devt), MINOR(devt));
618 bdi_set_owner(bdi, dev);
619 bdev_add(disk->part0, devt);
621 register_disk(parent, disk, groups);
623 blk_register_queue(disk);
626 * Take an extra ref on queue which will be put on disk_release()
627 * so that it sticks around as long as @disk is there.
629 WARN_ON_ONCE(!blk_get_queue(disk->queue));
631 disk_add_events(disk);
632 blk_integrity_add(disk);
635 void device_add_disk(struct device *parent, struct gendisk *disk,
636 const struct attribute_group **groups)
639 __device_add_disk(parent, disk, groups, true);
641 EXPORT_SYMBOL(device_add_disk);
643 void device_add_disk_no_queue_reg(struct device *parent, struct gendisk *disk)
645 __device_add_disk(parent, disk, NULL, false);
647 EXPORT_SYMBOL(device_add_disk_no_queue_reg);
650 * del_gendisk - remove the gendisk
651 * @disk: the struct gendisk to remove
653 * Removes the gendisk and all its associated resources. This deletes the
654 * partitions associated with the gendisk, and unregisters the associated
657 * This is the counter to the respective __device_add_disk() call.
659 * The final removal of the struct gendisk happens when its refcount reaches 0
660 * with put_disk(), which should be called after del_gendisk(), if
661 * __device_add_disk() was used.
663 * Drivers exist which depend on the release of the gendisk to be synchronous,
664 * it should not be deferred.
668 void del_gendisk(struct gendisk *disk)
672 if (WARN_ON_ONCE(!disk->queue))
675 blk_integrity_del(disk);
676 disk_del_events(disk);
679 * Block lookups of the disk until all bdevs are unhashed and the
680 * disk is marked as dead (GENHD_FL_UP cleared).
682 down_write(&bdev_lookup_sem);
684 mutex_lock(&disk->part0->bd_mutex);
685 blk_drop_partitions(disk);
686 mutex_unlock(&disk->part0->bd_mutex);
688 fsync_bdev(disk->part0);
689 __invalidate_device(disk->part0, true);
692 * Unhash the bdev inode for this device so that it can't be looked
693 * up any more even if openers still hold references to it.
695 remove_inode_hash(disk->part0->bd_inode);
697 set_capacity(disk, 0);
698 disk->flags &= ~GENHD_FL_UP;
699 up_write(&bdev_lookup_sem);
701 if (!(disk->flags & GENHD_FL_HIDDEN)) {
702 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
705 * Unregister bdi before releasing device numbers (as they can
706 * get reused and we'd get clashes in sysfs).
708 bdi_unregister(disk->queue->backing_dev_info);
711 blk_unregister_queue(disk);
713 kobject_put(disk->part0->bd_holder_dir);
714 kobject_put(disk->slave_dir);
716 part_stat_set_all(disk->part0, 0);
717 disk->part0->bd_stamp = 0;
718 if (!sysfs_deprecated)
719 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
720 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
721 device_del(disk_to_dev(disk));
723 EXPORT_SYMBOL(del_gendisk);
725 /* sysfs access to bad-blocks list. */
726 static ssize_t disk_badblocks_show(struct device *dev,
727 struct device_attribute *attr,
730 struct gendisk *disk = dev_to_disk(dev);
733 return sprintf(page, "\n");
735 return badblocks_show(disk->bb, page, 0);
738 static ssize_t disk_badblocks_store(struct device *dev,
739 struct device_attribute *attr,
740 const char *page, size_t len)
742 struct gendisk *disk = dev_to_disk(dev);
747 return badblocks_store(disk->bb, page, len, 0);
750 void blk_request_module(dev_t devt)
752 unsigned int major = MAJOR(devt);
753 struct blk_major_name **n;
755 mutex_lock(&major_names_lock);
756 for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) {
757 if ((*n)->major == major && (*n)->probe) {
759 mutex_unlock(&major_names_lock);
763 mutex_unlock(&major_names_lock);
765 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
766 /* Make old-style 2.4 aliases work */
767 request_module("block-major-%d", MAJOR(devt));
771 * bdget_disk - do bdget() by gendisk and partition number
772 * @disk: gendisk of interest
773 * @partno: partition number
775 * Find partition @partno from @disk, do bdget() on it.
781 * Resulting block_device on success, NULL on failure.
783 struct block_device *bdget_disk(struct gendisk *disk, int partno)
785 struct block_device *bdev = NULL;
788 bdev = xa_load(&disk->part_tbl, partno);
789 if (bdev && !bdgrab(bdev))
797 * print a full list of all partitions - intended for places where the root
798 * filesystem can't be mounted and thus to give the victim some idea of what
801 void __init printk_all_partitions(void)
803 struct class_dev_iter iter;
806 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
807 while ((dev = class_dev_iter_next(&iter))) {
808 struct gendisk *disk = dev_to_disk(dev);
809 struct block_device *part;
810 char name_buf[BDEVNAME_SIZE];
811 char devt_buf[BDEVT_SIZE];
815 * Don't show empty devices or things that have been
818 if (get_capacity(disk) == 0 ||
819 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
823 * Note, unlike /proc/partitions, I am showing the numbers in
824 * hex - the same format as the root= option takes.
827 xa_for_each(&disk->part_tbl, idx, part) {
828 if (!bdev_nr_sectors(part))
830 printk("%s%s %10llu %s %s",
831 bdev_is_partition(part) ? " " : "",
832 bdevt_str(part->bd_dev, devt_buf),
833 bdev_nr_sectors(part) >> 1,
834 disk_name(disk, part->bd_partno, name_buf),
836 part->bd_meta_info->uuid : "");
837 if (bdev_is_partition(part))
839 else if (dev->parent && dev->parent->driver)
840 printk(" driver: %s\n",
841 dev->parent->driver->name);
843 printk(" (driver?)\n");
847 class_dev_iter_exit(&iter);
850 #ifdef CONFIG_PROC_FS
852 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
855 struct class_dev_iter *iter;
858 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
860 return ERR_PTR(-ENOMEM);
862 seqf->private = iter;
863 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
865 dev = class_dev_iter_next(iter);
870 return dev_to_disk(dev);
873 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
878 dev = class_dev_iter_next(seqf->private);
880 return dev_to_disk(dev);
885 static void disk_seqf_stop(struct seq_file *seqf, void *v)
887 struct class_dev_iter *iter = seqf->private;
889 /* stop is called even after start failed :-( */
891 class_dev_iter_exit(iter);
893 seqf->private = NULL;
897 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
901 p = disk_seqf_start(seqf, pos);
902 if (!IS_ERR_OR_NULL(p) && !*pos)
903 seq_puts(seqf, "major minor #blocks name\n\n");
907 static int show_partition(struct seq_file *seqf, void *v)
909 struct gendisk *sgp = v;
910 struct block_device *part;
912 char buf[BDEVNAME_SIZE];
914 /* Don't show non-partitionable removeable devices or empty devices */
915 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
916 (sgp->flags & GENHD_FL_REMOVABLE)))
918 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
922 xa_for_each(&sgp->part_tbl, idx, part) {
923 if (!bdev_nr_sectors(part))
925 seq_printf(seqf, "%4d %7d %10llu %s\n",
926 MAJOR(part->bd_dev), MINOR(part->bd_dev),
927 bdev_nr_sectors(part) >> 1,
928 disk_name(sgp, part->bd_partno, buf));
934 static const struct seq_operations partitions_op = {
935 .start = show_partition_start,
936 .next = disk_seqf_next,
937 .stop = disk_seqf_stop,
938 .show = show_partition
942 static int __init genhd_device_init(void)
946 block_class.dev_kobj = sysfs_dev_block_kobj;
947 error = class_register(&block_class);
952 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
954 /* create top-level block dir */
955 if (!sysfs_deprecated)
956 block_depr = kobject_create_and_add("block", NULL);
960 subsys_initcall(genhd_device_init);
962 static ssize_t disk_range_show(struct device *dev,
963 struct device_attribute *attr, char *buf)
965 struct gendisk *disk = dev_to_disk(dev);
967 return sprintf(buf, "%d\n", disk->minors);
970 static ssize_t disk_ext_range_show(struct device *dev,
971 struct device_attribute *attr, char *buf)
973 struct gendisk *disk = dev_to_disk(dev);
975 return sprintf(buf, "%d\n", disk_max_parts(disk));
978 static ssize_t disk_removable_show(struct device *dev,
979 struct device_attribute *attr, char *buf)
981 struct gendisk *disk = dev_to_disk(dev);
983 return sprintf(buf, "%d\n",
984 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
987 static ssize_t disk_hidden_show(struct device *dev,
988 struct device_attribute *attr, char *buf)
990 struct gendisk *disk = dev_to_disk(dev);
992 return sprintf(buf, "%d\n",
993 (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
996 static ssize_t disk_ro_show(struct device *dev,
997 struct device_attribute *attr, char *buf)
999 struct gendisk *disk = dev_to_disk(dev);
1001 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
1004 ssize_t part_size_show(struct device *dev,
1005 struct device_attribute *attr, char *buf)
1007 return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev)));
1010 ssize_t part_stat_show(struct device *dev,
1011 struct device_attribute *attr, char *buf)
1013 struct block_device *bdev = dev_to_bdev(dev);
1014 struct request_queue *q = bdev->bd_disk->queue;
1015 struct disk_stats stat;
1016 unsigned int inflight;
1018 part_stat_read_all(bdev, &stat);
1020 inflight = blk_mq_in_flight(q, bdev);
1022 inflight = part_in_flight(bdev);
1025 "%8lu %8lu %8llu %8u "
1026 "%8lu %8lu %8llu %8u "
1028 "%8lu %8lu %8llu %8u "
1031 stat.ios[STAT_READ],
1032 stat.merges[STAT_READ],
1033 (unsigned long long)stat.sectors[STAT_READ],
1034 (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC),
1035 stat.ios[STAT_WRITE],
1036 stat.merges[STAT_WRITE],
1037 (unsigned long long)stat.sectors[STAT_WRITE],
1038 (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC),
1040 jiffies_to_msecs(stat.io_ticks),
1041 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1042 stat.nsecs[STAT_WRITE] +
1043 stat.nsecs[STAT_DISCARD] +
1044 stat.nsecs[STAT_FLUSH],
1046 stat.ios[STAT_DISCARD],
1047 stat.merges[STAT_DISCARD],
1048 (unsigned long long)stat.sectors[STAT_DISCARD],
1049 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC),
1050 stat.ios[STAT_FLUSH],
1051 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC));
1054 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
1057 struct block_device *bdev = dev_to_bdev(dev);
1058 struct request_queue *q = bdev->bd_disk->queue;
1059 unsigned int inflight[2];
1062 blk_mq_in_flight_rw(q, bdev, inflight);
1064 part_in_flight_rw(bdev, inflight);
1066 return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
1069 static ssize_t disk_capability_show(struct device *dev,
1070 struct device_attribute *attr, char *buf)
1072 struct gendisk *disk = dev_to_disk(dev);
1074 return sprintf(buf, "%x\n", disk->flags);
1077 static ssize_t disk_alignment_offset_show(struct device *dev,
1078 struct device_attribute *attr,
1081 struct gendisk *disk = dev_to_disk(dev);
1083 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
1086 static ssize_t disk_discard_alignment_show(struct device *dev,
1087 struct device_attribute *attr,
1090 struct gendisk *disk = dev_to_disk(dev);
1092 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
1095 static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
1096 static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
1097 static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
1098 static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
1099 static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
1100 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
1101 static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
1102 static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
1103 static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
1104 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
1105 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
1106 static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
1108 #ifdef CONFIG_FAIL_MAKE_REQUEST
1109 ssize_t part_fail_show(struct device *dev,
1110 struct device_attribute *attr, char *buf)
1112 return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail);
1115 ssize_t part_fail_store(struct device *dev,
1116 struct device_attribute *attr,
1117 const char *buf, size_t count)
1121 if (count > 0 && sscanf(buf, "%d", &i) > 0)
1122 dev_to_bdev(dev)->bd_make_it_fail = i;
1127 static struct device_attribute dev_attr_fail =
1128 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
1129 #endif /* CONFIG_FAIL_MAKE_REQUEST */
1131 #ifdef CONFIG_FAIL_IO_TIMEOUT
1132 static struct device_attribute dev_attr_fail_timeout =
1133 __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
1136 static struct attribute *disk_attrs[] = {
1137 &dev_attr_range.attr,
1138 &dev_attr_ext_range.attr,
1139 &dev_attr_removable.attr,
1140 &dev_attr_hidden.attr,
1142 &dev_attr_size.attr,
1143 &dev_attr_alignment_offset.attr,
1144 &dev_attr_discard_alignment.attr,
1145 &dev_attr_capability.attr,
1146 &dev_attr_stat.attr,
1147 &dev_attr_inflight.attr,
1148 &dev_attr_badblocks.attr,
1149 #ifdef CONFIG_FAIL_MAKE_REQUEST
1150 &dev_attr_fail.attr,
1152 #ifdef CONFIG_FAIL_IO_TIMEOUT
1153 &dev_attr_fail_timeout.attr,
1158 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1160 struct device *dev = container_of(kobj, typeof(*dev), kobj);
1161 struct gendisk *disk = dev_to_disk(dev);
1163 if (a == &dev_attr_badblocks.attr && !disk->bb)
1168 static struct attribute_group disk_attr_group = {
1169 .attrs = disk_attrs,
1170 .is_visible = disk_visible,
1173 static const struct attribute_group *disk_attr_groups[] = {
1179 * disk_release - releases all allocated resources of the gendisk
1180 * @dev: the device representing this disk
1182 * This function releases all allocated resources of the gendisk.
1184 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1185 * assigned. Since the request_queue sits on top of the gendisk for these
1186 * drivers we also call blk_put_queue() for them, and we expect the
1187 * request_queue refcount to reach 0 at this point, and so the request_queue
1188 * will also be freed prior to the disk.
1190 * Context: can sleep
1192 static void disk_release(struct device *dev)
1194 struct gendisk *disk = dev_to_disk(dev);
1198 blk_free_devt(dev->devt);
1199 disk_release_events(disk);
1200 kfree(disk->random);
1201 xa_destroy(&disk->part_tbl);
1204 blk_put_queue(disk->queue);
1207 struct class block_class = {
1211 static char *block_devnode(struct device *dev, umode_t *mode,
1212 kuid_t *uid, kgid_t *gid)
1214 struct gendisk *disk = dev_to_disk(dev);
1216 if (disk->fops->devnode)
1217 return disk->fops->devnode(disk, mode);
1221 const struct device_type disk_type = {
1223 .groups = disk_attr_groups,
1224 .release = disk_release,
1225 .devnode = block_devnode,
1228 #ifdef CONFIG_PROC_FS
1230 * aggregate disk stat collector. Uses the same stats that the sysfs
1231 * entries do, above, but makes them available through one seq_file.
1233 * The output looks suspiciously like /proc/partitions with a bunch of
1236 static int diskstats_show(struct seq_file *seqf, void *v)
1238 struct gendisk *gp = v;
1239 struct disk_part_iter piter;
1240 struct block_device *hd;
1241 char buf[BDEVNAME_SIZE];
1242 unsigned int inflight;
1243 struct disk_stats stat;
1246 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1247 seq_puts(seqf, "major minor name"
1248 " rio rmerge rsect ruse wio wmerge "
1249 "wsect wuse running use aveq"
1253 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
1254 while ((hd = disk_part_iter_next(&piter))) {
1255 part_stat_read_all(hd, &stat);
1256 if (queue_is_mq(gp->queue))
1257 inflight = blk_mq_in_flight(gp->queue, hd);
1259 inflight = part_in_flight(hd);
1261 seq_printf(seqf, "%4d %7d %s "
1268 MAJOR(hd->bd_dev), MINOR(hd->bd_dev),
1269 disk_name(gp, hd->bd_partno, buf),
1270 stat.ios[STAT_READ],
1271 stat.merges[STAT_READ],
1272 stat.sectors[STAT_READ],
1273 (unsigned int)div_u64(stat.nsecs[STAT_READ],
1275 stat.ios[STAT_WRITE],
1276 stat.merges[STAT_WRITE],
1277 stat.sectors[STAT_WRITE],
1278 (unsigned int)div_u64(stat.nsecs[STAT_WRITE],
1281 jiffies_to_msecs(stat.io_ticks),
1282 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1283 stat.nsecs[STAT_WRITE] +
1284 stat.nsecs[STAT_DISCARD] +
1285 stat.nsecs[STAT_FLUSH],
1287 stat.ios[STAT_DISCARD],
1288 stat.merges[STAT_DISCARD],
1289 stat.sectors[STAT_DISCARD],
1290 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD],
1292 stat.ios[STAT_FLUSH],
1293 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH],
1297 disk_part_iter_exit(&piter);
1302 static const struct seq_operations diskstats_op = {
1303 .start = disk_seqf_start,
1304 .next = disk_seqf_next,
1305 .stop = disk_seqf_stop,
1306 .show = diskstats_show
1309 static int __init proc_genhd_init(void)
1311 proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1312 proc_create_seq("partitions", 0, NULL, &partitions_op);
1315 module_init(proc_genhd_init);
1316 #endif /* CONFIG_PROC_FS */
1318 dev_t blk_lookup_devt(const char *name, int partno)
1320 dev_t devt = MKDEV(0, 0);
1321 struct class_dev_iter iter;
1324 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1325 while ((dev = class_dev_iter_next(&iter))) {
1326 struct gendisk *disk = dev_to_disk(dev);
1327 struct block_device *part;
1329 if (strcmp(dev_name(dev), name))
1332 if (partno < disk->minors) {
1333 /* We need to return the right devno, even
1334 * if the partition doesn't exist yet.
1336 devt = MKDEV(MAJOR(dev->devt),
1337 MINOR(dev->devt) + partno);
1340 part = bdget_disk(disk, partno);
1342 devt = part->bd_dev;
1347 class_dev_iter_exit(&iter);
1351 struct gendisk *__alloc_disk_node(int minors, int node_id)
1353 struct gendisk *disk;
1355 if (minors > DISK_MAX_PARTS) {
1357 "block: can't allocate more than %d partitions\n",
1359 minors = DISK_MAX_PARTS;
1362 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1366 disk->part0 = bdev_alloc(disk, 0);
1370 disk->node_id = node_id;
1371 xa_init(&disk->part_tbl);
1372 if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL))
1373 goto out_destroy_part_tbl;
1375 disk->minors = minors;
1376 rand_initialize_disk(disk);
1377 disk_to_dev(disk)->class = &block_class;
1378 disk_to_dev(disk)->type = &disk_type;
1379 device_initialize(disk_to_dev(disk));
1382 out_destroy_part_tbl:
1383 xa_destroy(&disk->part_tbl);
1389 EXPORT_SYMBOL(__alloc_disk_node);
1392 * put_disk - decrements the gendisk refcount
1393 * @disk: the struct gendisk to decrement the refcount for
1395 * This decrements the refcount for the struct gendisk. When this reaches 0
1396 * we'll have disk_release() called.
1398 * Context: Any context, but the last reference must not be dropped from
1401 void put_disk(struct gendisk *disk)
1404 put_device(disk_to_dev(disk));
1406 EXPORT_SYMBOL(put_disk);
1408 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1410 char event[] = "DISK_RO=1";
1411 char *envp[] = { event, NULL };
1415 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1419 * set_disk_ro - set a gendisk read-only
1420 * @disk: gendisk to operate on
1421 * @read_only: %true to set the disk read-only, %false set the disk read/write
1423 * This function is used to indicate whether a given disk device should have its
1424 * read-only flag set. set_disk_ro() is typically used by device drivers to
1425 * indicate whether the underlying physical device is write-protected.
1427 void set_disk_ro(struct gendisk *disk, bool read_only)
1430 if (test_and_set_bit(GD_READ_ONLY, &disk->state))
1433 if (!test_and_clear_bit(GD_READ_ONLY, &disk->state))
1436 set_disk_ro_uevent(disk, read_only);
1438 EXPORT_SYMBOL(set_disk_ro);
1440 int bdev_read_only(struct block_device *bdev)
1442 return bdev->bd_read_only || get_disk_ro(bdev->bd_disk);
1444 EXPORT_SYMBOL(bdev_read_only);
1447 * Disk events - monitor disk events like media change and eject request.
1449 struct disk_events {
1450 struct list_head node; /* all disk_event's */
1451 struct gendisk *disk; /* the associated disk */
1454 struct mutex block_mutex; /* protects blocking */
1455 int block; /* event blocking depth */
1456 unsigned int pending; /* events already sent out */
1457 unsigned int clearing; /* events being cleared */
1459 long poll_msecs; /* interval, -1 for default */
1460 struct delayed_work dwork;
1463 static const char *disk_events_strs[] = {
1464 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1465 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1468 static char *disk_uevents[] = {
1469 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1470 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1473 /* list of all disk_events */
1474 static DEFINE_MUTEX(disk_events_mutex);
1475 static LIST_HEAD(disk_events);
1477 /* disable in-kernel polling by default */
1478 static unsigned long disk_events_dfl_poll_msecs;
1480 static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1482 struct disk_events *ev = disk->ev;
1483 long intv_msecs = 0;
1486 * If device-specific poll interval is set, always use it. If
1487 * the default is being used, poll if the POLL flag is set.
1489 if (ev->poll_msecs >= 0)
1490 intv_msecs = ev->poll_msecs;
1491 else if (disk->event_flags & DISK_EVENT_FLAG_POLL)
1492 intv_msecs = disk_events_dfl_poll_msecs;
1494 return msecs_to_jiffies(intv_msecs);
1498 * disk_block_events - block and flush disk event checking
1499 * @disk: disk to block events for
1501 * On return from this function, it is guaranteed that event checking
1502 * isn't in progress and won't happen until unblocked by
1503 * disk_unblock_events(). Events blocking is counted and the actual
1504 * unblocking happens after the matching number of unblocks are done.
1506 * Note that this intentionally does not block event checking from
1507 * disk_clear_events().
1512 void disk_block_events(struct gendisk *disk)
1514 struct disk_events *ev = disk->ev;
1515 unsigned long flags;
1522 * Outer mutex ensures that the first blocker completes canceling
1523 * the event work before further blockers are allowed to finish.
1525 mutex_lock(&ev->block_mutex);
1527 spin_lock_irqsave(&ev->lock, flags);
1528 cancel = !ev->block++;
1529 spin_unlock_irqrestore(&ev->lock, flags);
1532 cancel_delayed_work_sync(&disk->ev->dwork);
1534 mutex_unlock(&ev->block_mutex);
1537 static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1539 struct disk_events *ev = disk->ev;
1541 unsigned long flags;
1543 spin_lock_irqsave(&ev->lock, flags);
1545 if (WARN_ON_ONCE(ev->block <= 0))
1551 intv = disk_events_poll_jiffies(disk);
1553 queue_delayed_work(system_freezable_power_efficient_wq,
1556 queue_delayed_work(system_freezable_power_efficient_wq,
1559 spin_unlock_irqrestore(&ev->lock, flags);
1563 * disk_unblock_events - unblock disk event checking
1564 * @disk: disk to unblock events for
1566 * Undo disk_block_events(). When the block count reaches zero, it
1567 * starts events polling if configured.
1570 * Don't care. Safe to call from irq context.
1572 void disk_unblock_events(struct gendisk *disk)
1575 __disk_unblock_events(disk, false);
1579 * disk_flush_events - schedule immediate event checking and flushing
1580 * @disk: disk to check and flush events for
1581 * @mask: events to flush
1583 * Schedule immediate event checking on @disk if not blocked. Events in
1584 * @mask are scheduled to be cleared from the driver. Note that this
1585 * doesn't clear the events from @disk->ev.
1588 * If @mask is non-zero must be called with bdev->bd_mutex held.
1590 void disk_flush_events(struct gendisk *disk, unsigned int mask)
1592 struct disk_events *ev = disk->ev;
1597 spin_lock_irq(&ev->lock);
1598 ev->clearing |= mask;
1600 mod_delayed_work(system_freezable_power_efficient_wq,
1602 spin_unlock_irq(&ev->lock);
1606 * disk_clear_events - synchronously check, clear and return pending events
1607 * @disk: disk to fetch and clear events from
1608 * @mask: mask of events to be fetched and cleared
1610 * Disk events are synchronously checked and pending events in @mask
1611 * are cleared and returned. This ignores the block count.
1616 static unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1618 struct disk_events *ev = disk->ev;
1619 unsigned int pending;
1620 unsigned int clearing = mask;
1625 disk_block_events(disk);
1628 * store the union of mask and ev->clearing on the stack so that the
1629 * race with disk_flush_events does not cause ambiguity (ev->clearing
1630 * can still be modified even if events are blocked).
1632 spin_lock_irq(&ev->lock);
1633 clearing |= ev->clearing;
1635 spin_unlock_irq(&ev->lock);
1637 disk_check_events(ev, &clearing);
1639 * if ev->clearing is not 0, the disk_flush_events got called in the
1640 * middle of this function, so we want to run the workfn without delay.
1642 __disk_unblock_events(disk, ev->clearing ? true : false);
1644 /* then, fetch and clear pending events */
1645 spin_lock_irq(&ev->lock);
1646 pending = ev->pending & mask;
1647 ev->pending &= ~mask;
1648 spin_unlock_irq(&ev->lock);
1649 WARN_ON_ONCE(clearing & mask);
1655 * bdev_check_media_change - check if a removable media has been changed
1656 * @bdev: block device to check
1658 * Check whether a removable media has been changed, and attempt to free all
1659 * dentries and inodes and invalidates all block device page cache entries in
1662 * Returns %true if the block device changed, or %false if not.
1664 bool bdev_check_media_change(struct block_device *bdev)
1666 unsigned int events;
1668 events = disk_clear_events(bdev->bd_disk, DISK_EVENT_MEDIA_CHANGE |
1669 DISK_EVENT_EJECT_REQUEST);
1670 if (!(events & DISK_EVENT_MEDIA_CHANGE))
1673 if (__invalidate_device(bdev, true))
1674 pr_warn("VFS: busy inodes on changed media %s\n",
1675 bdev->bd_disk->disk_name);
1676 set_bit(GD_NEED_PART_SCAN, &bdev->bd_disk->state);
1679 EXPORT_SYMBOL(bdev_check_media_change);
1682 * Separate this part out so that a different pointer for clearing_ptr can be
1683 * passed in for disk_clear_events.
1685 static void disk_events_workfn(struct work_struct *work)
1687 struct delayed_work *dwork = to_delayed_work(work);
1688 struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
1690 disk_check_events(ev, &ev->clearing);
1693 static void disk_check_events(struct disk_events *ev,
1694 unsigned int *clearing_ptr)
1696 struct gendisk *disk = ev->disk;
1697 char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
1698 unsigned int clearing = *clearing_ptr;
1699 unsigned int events;
1701 int nr_events = 0, i;
1704 events = disk->fops->check_events(disk, clearing);
1706 /* accumulate pending events and schedule next poll if necessary */
1707 spin_lock_irq(&ev->lock);
1709 events &= ~ev->pending;
1710 ev->pending |= events;
1711 *clearing_ptr &= ~clearing;
1713 intv = disk_events_poll_jiffies(disk);
1714 if (!ev->block && intv)
1715 queue_delayed_work(system_freezable_power_efficient_wq,
1718 spin_unlock_irq(&ev->lock);
1721 * Tell userland about new events. Only the events listed in
1722 * @disk->events are reported, and only if DISK_EVENT_FLAG_UEVENT
1723 * is set. Otherwise, events are processed internally but never
1724 * get reported to userland.
1726 for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
1727 if ((events & disk->events & (1 << i)) &&
1728 (disk->event_flags & DISK_EVENT_FLAG_UEVENT))
1729 envp[nr_events++] = disk_uevents[i];
1732 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1736 * A disk events enabled device has the following sysfs nodes under
1737 * its /sys/block/X/ directory.
1739 * events : list of all supported events
1740 * events_async : list of events which can be detected w/o polling
1741 * (always empty, only for backwards compatibility)
1742 * events_poll_msecs : polling interval, 0: disable, -1: system default
1744 static ssize_t __disk_events_show(unsigned int events, char *buf)
1746 const char *delim = "";
1750 for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1751 if (events & (1 << i)) {
1752 pos += sprintf(buf + pos, "%s%s",
1753 delim, disk_events_strs[i]);
1757 pos += sprintf(buf + pos, "\n");
1761 static ssize_t disk_events_show(struct device *dev,
1762 struct device_attribute *attr, char *buf)
1764 struct gendisk *disk = dev_to_disk(dev);
1766 if (!(disk->event_flags & DISK_EVENT_FLAG_UEVENT))
1769 return __disk_events_show(disk->events, buf);
1772 static ssize_t disk_events_async_show(struct device *dev,
1773 struct device_attribute *attr, char *buf)
1778 static ssize_t disk_events_poll_msecs_show(struct device *dev,
1779 struct device_attribute *attr,
1782 struct gendisk *disk = dev_to_disk(dev);
1785 return sprintf(buf, "-1\n");
1787 return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1790 static ssize_t disk_events_poll_msecs_store(struct device *dev,
1791 struct device_attribute *attr,
1792 const char *buf, size_t count)
1794 struct gendisk *disk = dev_to_disk(dev);
1797 if (!count || !sscanf(buf, "%ld", &intv))
1800 if (intv < 0 && intv != -1)
1806 disk_block_events(disk);
1807 disk->ev->poll_msecs = intv;
1808 __disk_unblock_events(disk, true);
1813 static const DEVICE_ATTR(events, 0444, disk_events_show, NULL);
1814 static const DEVICE_ATTR(events_async, 0444, disk_events_async_show, NULL);
1815 static const DEVICE_ATTR(events_poll_msecs, 0644,
1816 disk_events_poll_msecs_show,
1817 disk_events_poll_msecs_store);
1819 static const struct attribute *disk_events_attrs[] = {
1820 &dev_attr_events.attr,
1821 &dev_attr_events_async.attr,
1822 &dev_attr_events_poll_msecs.attr,
1827 * The default polling interval can be specified by the kernel
1828 * parameter block.events_dfl_poll_msecs which defaults to 0
1829 * (disable). This can also be modified runtime by writing to
1830 * /sys/module/block/parameters/events_dfl_poll_msecs.
1832 static int disk_events_set_dfl_poll_msecs(const char *val,
1833 const struct kernel_param *kp)
1835 struct disk_events *ev;
1838 ret = param_set_ulong(val, kp);
1842 mutex_lock(&disk_events_mutex);
1844 list_for_each_entry(ev, &disk_events, node)
1845 disk_flush_events(ev->disk, 0);
1847 mutex_unlock(&disk_events_mutex);
1852 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1853 .set = disk_events_set_dfl_poll_msecs,
1854 .get = param_get_ulong,
1857 #undef MODULE_PARAM_PREFIX
1858 #define MODULE_PARAM_PREFIX "block."
1860 module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
1861 &disk_events_dfl_poll_msecs, 0644);
1864 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
1866 static void disk_alloc_events(struct gendisk *disk)
1868 struct disk_events *ev;
1870 if (!disk->fops->check_events || !disk->events)
1873 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1875 pr_warn("%s: failed to initialize events\n", disk->disk_name);
1879 INIT_LIST_HEAD(&ev->node);
1881 spin_lock_init(&ev->lock);
1882 mutex_init(&ev->block_mutex);
1884 ev->poll_msecs = -1;
1885 INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
1890 static void disk_add_events(struct gendisk *disk)
1892 /* FIXME: error handling */
1893 if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0)
1894 pr_warn("%s: failed to create sysfs files for events\n",
1900 mutex_lock(&disk_events_mutex);
1901 list_add_tail(&disk->ev->node, &disk_events);
1902 mutex_unlock(&disk_events_mutex);
1905 * Block count is initialized to 1 and the following initial
1906 * unblock kicks it into action.
1908 __disk_unblock_events(disk, true);
1911 static void disk_del_events(struct gendisk *disk)
1914 disk_block_events(disk);
1916 mutex_lock(&disk_events_mutex);
1917 list_del_init(&disk->ev->node);
1918 mutex_unlock(&disk_events_mutex);
1921 sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
1924 static void disk_release_events(struct gendisk *disk)
1926 /* the block count should be 1 from disk_del_events() */
1927 WARN_ON_ONCE(disk->ev && disk->ev->block != 1);