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;
50 spin_lock_irqsave(&bdev->bd_size_lock, flags);
51 i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
52 spin_unlock_irqrestore(&bdev->bd_size_lock, flags);
54 EXPORT_SYMBOL(set_capacity);
57 * Set disk capacity and notify if the size is not currently zero and will not
58 * be set to zero. Returns true if a uevent was sent, otherwise false.
60 bool set_capacity_and_notify(struct gendisk *disk, sector_t size)
62 sector_t capacity = get_capacity(disk);
63 char *envp[] = { "RESIZE=1", NULL };
65 set_capacity(disk, size);
68 * Only print a message and send a uevent if the gendisk is user visible
69 * and alive. This avoids spamming the log and udev when setting the
70 * initial capacity during probing.
72 if (size == capacity ||
73 (disk->flags & (GENHD_FL_UP | GENHD_FL_HIDDEN)) != GENHD_FL_UP)
76 pr_info("%s: detected capacity change from %lld to %lld\n",
77 disk->disk_name, size, capacity);
80 * Historically we did not send a uevent for changes to/from an empty
83 if (!capacity || !size)
85 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
88 EXPORT_SYMBOL_GPL(set_capacity_and_notify);
91 * Format the device name of the indicated disk into the supplied buffer and
92 * return a pointer to that same buffer for convenience.
94 char *disk_name(struct gendisk *hd, int partno, char *buf)
97 snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name);
98 else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1]))
99 snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno);
101 snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno);
106 const char *bdevname(struct block_device *bdev, char *buf)
108 return disk_name(bdev->bd_disk, bdev->bd_partno, buf);
110 EXPORT_SYMBOL(bdevname);
112 static void part_stat_read_all(struct block_device *part,
113 struct disk_stats *stat)
117 memset(stat, 0, sizeof(struct disk_stats));
118 for_each_possible_cpu(cpu) {
119 struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu);
122 for (group = 0; group < NR_STAT_GROUPS; group++) {
123 stat->nsecs[group] += ptr->nsecs[group];
124 stat->sectors[group] += ptr->sectors[group];
125 stat->ios[group] += ptr->ios[group];
126 stat->merges[group] += ptr->merges[group];
129 stat->io_ticks += ptr->io_ticks;
133 static unsigned int part_in_flight(struct block_device *part)
135 unsigned int inflight = 0;
138 for_each_possible_cpu(cpu) {
139 inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) +
140 part_stat_local_read_cpu(part, in_flight[1], cpu);
142 if ((int)inflight < 0)
148 static void part_in_flight_rw(struct block_device *part,
149 unsigned int inflight[2])
155 for_each_possible_cpu(cpu) {
156 inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu);
157 inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu);
159 if ((int)inflight[0] < 0)
161 if ((int)inflight[1] < 0)
166 * disk_part_iter_init - initialize partition iterator
167 * @piter: iterator to initialize
168 * @disk: disk to iterate over
169 * @flags: DISK_PITER_* flags
171 * Initialize @piter so that it iterates over partitions of @disk.
176 void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk,
181 if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0))
185 piter->flags = flags;
189 * disk_part_iter_next - proceed iterator to the next partition and return it
190 * @piter: iterator of interest
192 * Proceed @piter to the next partition and return it.
197 struct block_device *disk_part_iter_next(struct disk_part_iter *piter)
199 struct block_device *part;
202 /* put the last partition */
203 disk_part_iter_exit(piter);
206 xa_for_each_start(&piter->disk->part_tbl, idx, part, piter->idx) {
207 if (!bdev_nr_sectors(part) &&
208 !(piter->flags & DISK_PITER_INCL_EMPTY) &&
209 !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 &&
213 piter->part = bdgrab(part);
216 piter->idx = idx + 1;
225 * disk_part_iter_exit - finish up partition iteration
226 * @piter: iter of interest
228 * Called when iteration is over. Cleans up @piter.
233 void disk_part_iter_exit(struct disk_part_iter *piter)
241 * Can be deleted altogether. Later.
244 #define BLKDEV_MAJOR_HASH_SIZE 255
245 static struct blk_major_name {
246 struct blk_major_name *next;
249 void (*probe)(dev_t devt);
250 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
251 static DEFINE_MUTEX(major_names_lock);
253 /* index in the above - for now: assume no multimajor ranges */
254 static inline int major_to_index(unsigned major)
256 return major % BLKDEV_MAJOR_HASH_SIZE;
259 #ifdef CONFIG_PROC_FS
260 void blkdev_show(struct seq_file *seqf, off_t offset)
262 struct blk_major_name *dp;
264 mutex_lock(&major_names_lock);
265 for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
266 if (dp->major == offset)
267 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
268 mutex_unlock(&major_names_lock);
270 #endif /* CONFIG_PROC_FS */
273 * __register_blkdev - register a new block device
275 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
276 * @major = 0, try to allocate any unused major number.
277 * @name: the name of the new block device as a zero terminated string
278 * @probe: allback that is called on access to any minor number of @major
280 * The @name must be unique within the system.
282 * The return value depends on the @major input parameter:
284 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
285 * then the function returns zero on success, or a negative error code
286 * - if any unused major number was requested with @major = 0 parameter
287 * then the return value is the allocated major number in range
288 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
290 * See Documentation/admin-guide/devices.txt for the list of allocated
293 * Use register_blkdev instead for any new code.
295 int __register_blkdev(unsigned int major, const char *name,
296 void (*probe)(dev_t devt))
298 struct blk_major_name **n, *p;
301 mutex_lock(&major_names_lock);
305 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
306 if (major_names[index] == NULL)
311 printk("%s: failed to get major for %s\n",
320 if (major >= BLKDEV_MAJOR_MAX) {
321 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
322 __func__, major, BLKDEV_MAJOR_MAX-1, name);
328 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
336 strlcpy(p->name, name, sizeof(p->name));
338 index = major_to_index(major);
340 for (n = &major_names[index]; *n; n = &(*n)->next) {
341 if ((*n)->major == major)
350 printk("register_blkdev: cannot get major %u for %s\n",
355 mutex_unlock(&major_names_lock);
358 EXPORT_SYMBOL(__register_blkdev);
360 void unregister_blkdev(unsigned int major, const char *name)
362 struct blk_major_name **n;
363 struct blk_major_name *p = NULL;
364 int index = major_to_index(major);
366 mutex_lock(&major_names_lock);
367 for (n = &major_names[index]; *n; n = &(*n)->next)
368 if ((*n)->major == major)
370 if (!*n || strcmp((*n)->name, name)) {
376 mutex_unlock(&major_names_lock);
380 EXPORT_SYMBOL(unregister_blkdev);
383 * blk_mangle_minor - scatter minor numbers apart
384 * @minor: minor number to mangle
386 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
387 * is enabled. Mangling twice gives the original value.
395 static int blk_mangle_minor(int minor)
397 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
400 for (i = 0; i < MINORBITS / 2; i++) {
401 int low = minor & (1 << i);
402 int high = minor & (1 << (MINORBITS - 1 - i));
403 int distance = MINORBITS - 1 - 2 * i;
405 minor ^= low | high; /* clear both bits */
406 low <<= distance; /* swap the positions */
408 minor |= low | high; /* and set */
415 * blk_alloc_devt - allocate a dev_t for a block device
416 * @bdev: block device to allocate dev_t for
417 * @devt: out parameter for resulting dev_t
419 * Allocate a dev_t for block device.
422 * 0 on success, allocated dev_t is returned in *@devt. -errno on
428 int blk_alloc_devt(struct block_device *bdev, dev_t *devt)
430 struct gendisk *disk = bdev->bd_disk;
433 /* in consecutive minor range? */
434 if (bdev->bd_partno < disk->minors) {
435 *devt = MKDEV(disk->major, disk->first_minor + bdev->bd_partno);
439 idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT, GFP_KERNEL);
441 return idx == -ENOSPC ? -EBUSY : idx;
443 *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
448 * blk_free_devt - free a dev_t
449 * @devt: dev_t to free
451 * Free @devt which was allocated using blk_alloc_devt().
456 void blk_free_devt(dev_t devt)
458 if (MAJOR(devt) == BLOCK_EXT_MAJOR)
459 ida_free(&ext_devt_ida, blk_mangle_minor(MINOR(devt)));
462 static char *bdevt_str(dev_t devt, char *buf)
464 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
465 char tbuf[BDEVT_SIZE];
466 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
467 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
469 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
474 void disk_uevent(struct gendisk *disk, enum kobject_action action)
476 struct disk_part_iter piter;
477 struct block_device *part;
479 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
480 while ((part = disk_part_iter_next(&piter)))
481 kobject_uevent(bdev_kobj(part), action);
482 disk_part_iter_exit(&piter);
484 EXPORT_SYMBOL_GPL(disk_uevent);
486 static void disk_scan_partitions(struct gendisk *disk)
488 struct block_device *bdev;
490 if (!get_capacity(disk) || !disk_part_scan_enabled(disk))
493 set_bit(GD_NEED_PART_SCAN, &disk->state);
494 bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL);
496 blkdev_put(bdev, FMODE_READ);
499 static void register_disk(struct device *parent, struct gendisk *disk,
500 const struct attribute_group **groups)
502 struct device *ddev = disk_to_dev(disk);
505 ddev->parent = parent;
507 dev_set_name(ddev, "%s", disk->disk_name);
509 /* delay uevents, until we scanned partition table */
510 dev_set_uevent_suppress(ddev, 1);
513 WARN_ON(ddev->groups);
514 ddev->groups = groups;
516 if (device_add(ddev))
518 if (!sysfs_deprecated) {
519 err = sysfs_create_link(block_depr, &ddev->kobj,
520 kobject_name(&ddev->kobj));
528 * avoid probable deadlock caused by allocating memory with
529 * GFP_KERNEL in runtime_resume callback of its all ancestor
532 pm_runtime_set_memalloc_noio(ddev, true);
534 disk->part0->bd_holder_dir =
535 kobject_create_and_add("holders", &ddev->kobj);
536 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
538 if (disk->flags & GENHD_FL_HIDDEN) {
539 dev_set_uevent_suppress(ddev, 0);
543 disk_scan_partitions(disk);
545 /* announce the disk and partitions after all partitions are created */
546 dev_set_uevent_suppress(ddev, 0);
547 disk_uevent(disk, KOBJ_ADD);
549 if (disk->queue->backing_dev_info->dev) {
550 err = sysfs_create_link(&ddev->kobj,
551 &disk->queue->backing_dev_info->dev->kobj,
558 * __device_add_disk - add disk information to kernel list
559 * @parent: parent device for the disk
560 * @disk: per-device partitioning information
561 * @groups: Additional per-device sysfs groups
562 * @register_queue: register the queue if set to true
564 * This function registers the partitioning information in @disk
567 * FIXME: error handling
569 static void __device_add_disk(struct device *parent, struct gendisk *disk,
570 const struct attribute_group **groups,
577 * The disk queue should now be all set with enough information about
578 * the device for the elevator code to pick an adequate default
579 * elevator if one is needed, that is, for devices requesting queue
583 elevator_init_mq(disk->queue);
585 /* minors == 0 indicates to use ext devt from part0 and should
586 * be accompanied with EXT_DEVT flag. Make sure all
587 * parameters make sense.
589 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
590 WARN_ON(!disk->minors &&
591 !(disk->flags & (GENHD_FL_EXT_DEVT | GENHD_FL_HIDDEN)));
593 disk->flags |= GENHD_FL_UP;
595 retval = blk_alloc_devt(disk->part0, &devt);
600 disk->major = MAJOR(devt);
601 disk->first_minor = MINOR(devt);
603 disk_alloc_events(disk);
605 if (disk->flags & GENHD_FL_HIDDEN) {
607 * Don't let hidden disks show up in /proc/partitions,
608 * and don't bother scanning for partitions either.
610 disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
611 disk->flags |= GENHD_FL_NO_PART_SCAN;
613 struct backing_dev_info *bdi = disk->queue->backing_dev_info;
614 struct device *dev = disk_to_dev(disk);
617 /* Register BDI before referencing it from bdev */
619 ret = bdi_register(bdi, "%u:%u", MAJOR(devt), MINOR(devt));
621 bdi_set_owner(bdi, dev);
622 bdev_add(disk->part0, devt);
624 register_disk(parent, disk, groups);
626 blk_register_queue(disk);
629 * Take an extra ref on queue which will be put on disk_release()
630 * so that it sticks around as long as @disk is there.
632 WARN_ON_ONCE(!blk_get_queue(disk->queue));
634 disk_add_events(disk);
635 blk_integrity_add(disk);
638 void device_add_disk(struct device *parent, struct gendisk *disk,
639 const struct attribute_group **groups)
642 __device_add_disk(parent, disk, groups, true);
644 EXPORT_SYMBOL(device_add_disk);
646 void device_add_disk_no_queue_reg(struct device *parent, struct gendisk *disk)
648 __device_add_disk(parent, disk, NULL, false);
650 EXPORT_SYMBOL(device_add_disk_no_queue_reg);
652 static void invalidate_partition(struct block_device *bdev)
655 __invalidate_device(bdev, true);
658 * Unhash the bdev inode for this device so that it can't be looked
659 * up any more even if openers still hold references to it.
661 remove_inode_hash(bdev->bd_inode);
665 * del_gendisk - remove the gendisk
666 * @disk: the struct gendisk to remove
668 * Removes the gendisk and all its associated resources. This deletes the
669 * partitions associated with the gendisk, and unregisters the associated
672 * This is the counter to the respective __device_add_disk() call.
674 * The final removal of the struct gendisk happens when its refcount reaches 0
675 * with put_disk(), which should be called after del_gendisk(), if
676 * __device_add_disk() was used.
678 * Drivers exist which depend on the release of the gendisk to be synchronous,
679 * it should not be deferred.
683 void del_gendisk(struct gendisk *disk)
685 struct disk_part_iter piter;
686 struct block_device *part;
690 if (WARN_ON_ONCE(!disk->queue))
693 blk_integrity_del(disk);
694 disk_del_events(disk);
697 * Block lookups of the disk until all bdevs are unhashed and the
698 * disk is marked as dead (GENHD_FL_UP cleared).
700 down_write(&bdev_lookup_sem);
702 /* invalidate stuff */
703 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
704 while ((part = disk_part_iter_next(&piter))) {
705 invalidate_partition(part);
706 delete_partition(part);
708 disk_part_iter_exit(&piter);
710 invalidate_partition(disk->part0);
711 set_capacity(disk, 0);
712 disk->flags &= ~GENHD_FL_UP;
713 up_write(&bdev_lookup_sem);
715 if (!(disk->flags & GENHD_FL_HIDDEN)) {
716 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
719 * Unregister bdi before releasing device numbers (as they can
720 * get reused and we'd get clashes in sysfs).
722 bdi_unregister(disk->queue->backing_dev_info);
725 blk_unregister_queue(disk);
727 kobject_put(disk->part0->bd_holder_dir);
728 kobject_put(disk->slave_dir);
730 part_stat_set_all(disk->part0, 0);
731 disk->part0->bd_stamp = 0;
732 if (!sysfs_deprecated)
733 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
734 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
735 device_del(disk_to_dev(disk));
737 EXPORT_SYMBOL(del_gendisk);
739 /* sysfs access to bad-blocks list. */
740 static ssize_t disk_badblocks_show(struct device *dev,
741 struct device_attribute *attr,
744 struct gendisk *disk = dev_to_disk(dev);
747 return sprintf(page, "\n");
749 return badblocks_show(disk->bb, page, 0);
752 static ssize_t disk_badblocks_store(struct device *dev,
753 struct device_attribute *attr,
754 const char *page, size_t len)
756 struct gendisk *disk = dev_to_disk(dev);
761 return badblocks_store(disk->bb, page, len, 0);
764 void blk_request_module(dev_t devt)
766 unsigned int major = MAJOR(devt);
767 struct blk_major_name **n;
769 mutex_lock(&major_names_lock);
770 for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) {
771 if ((*n)->major == major && (*n)->probe) {
773 mutex_unlock(&major_names_lock);
777 mutex_unlock(&major_names_lock);
779 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
780 /* Make old-style 2.4 aliases work */
781 request_module("block-major-%d", MAJOR(devt));
785 * bdget_disk - do bdget() by gendisk and partition number
786 * @disk: gendisk of interest
787 * @partno: partition number
789 * Find partition @partno from @disk, do bdget() on it.
795 * Resulting block_device on success, NULL on failure.
797 struct block_device *bdget_disk(struct gendisk *disk, int partno)
799 struct block_device *bdev = NULL;
802 bdev = xa_load(&disk->part_tbl, partno);
803 if (bdev && !bdgrab(bdev))
811 * print a full list of all partitions - intended for places where the root
812 * filesystem can't be mounted and thus to give the victim some idea of what
815 void __init printk_all_partitions(void)
817 struct class_dev_iter iter;
820 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
821 while ((dev = class_dev_iter_next(&iter))) {
822 struct gendisk *disk = dev_to_disk(dev);
823 struct disk_part_iter piter;
824 struct block_device *part;
825 char name_buf[BDEVNAME_SIZE];
826 char devt_buf[BDEVT_SIZE];
829 * Don't show empty devices or things that have been
832 if (get_capacity(disk) == 0 ||
833 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
837 * Note, unlike /proc/partitions, I am showing the
838 * numbers in hex - the same format as the root=
841 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
842 while ((part = disk_part_iter_next(&piter))) {
843 bool is_part0 = part == disk->part0;
845 printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
846 bdevt_str(part->bd_dev, devt_buf),
847 bdev_nr_sectors(part) >> 1,
848 disk_name(disk, part->bd_partno, name_buf),
850 part->bd_meta_info->uuid : "");
852 if (dev->parent && dev->parent->driver)
853 printk(" driver: %s\n",
854 dev->parent->driver->name);
856 printk(" (driver?)\n");
860 disk_part_iter_exit(&piter);
862 class_dev_iter_exit(&iter);
865 #ifdef CONFIG_PROC_FS
867 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
870 struct class_dev_iter *iter;
873 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
875 return ERR_PTR(-ENOMEM);
877 seqf->private = iter;
878 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
880 dev = class_dev_iter_next(iter);
885 return dev_to_disk(dev);
888 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
893 dev = class_dev_iter_next(seqf->private);
895 return dev_to_disk(dev);
900 static void disk_seqf_stop(struct seq_file *seqf, void *v)
902 struct class_dev_iter *iter = seqf->private;
904 /* stop is called even after start failed :-( */
906 class_dev_iter_exit(iter);
908 seqf->private = NULL;
912 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
916 p = disk_seqf_start(seqf, pos);
917 if (!IS_ERR_OR_NULL(p) && !*pos)
918 seq_puts(seqf, "major minor #blocks name\n\n");
922 static int show_partition(struct seq_file *seqf, void *v)
924 struct gendisk *sgp = v;
925 struct disk_part_iter piter;
926 struct block_device *part;
927 char buf[BDEVNAME_SIZE];
929 /* Don't show non-partitionable removeable devices or empty devices */
930 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
931 (sgp->flags & GENHD_FL_REMOVABLE)))
933 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
936 /* show the full disk and all non-0 size partitions of it */
937 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
938 while ((part = disk_part_iter_next(&piter)))
939 seq_printf(seqf, "%4d %7d %10llu %s\n",
940 MAJOR(part->bd_dev), MINOR(part->bd_dev),
941 bdev_nr_sectors(part) >> 1,
942 disk_name(sgp, part->bd_partno, buf));
943 disk_part_iter_exit(&piter);
948 static const struct seq_operations partitions_op = {
949 .start = show_partition_start,
950 .next = disk_seqf_next,
951 .stop = disk_seqf_stop,
952 .show = show_partition
956 static int __init genhd_device_init(void)
960 block_class.dev_kobj = sysfs_dev_block_kobj;
961 error = class_register(&block_class);
966 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
968 /* create top-level block dir */
969 if (!sysfs_deprecated)
970 block_depr = kobject_create_and_add("block", NULL);
974 subsys_initcall(genhd_device_init);
976 static ssize_t disk_range_show(struct device *dev,
977 struct device_attribute *attr, char *buf)
979 struct gendisk *disk = dev_to_disk(dev);
981 return sprintf(buf, "%d\n", disk->minors);
984 static ssize_t disk_ext_range_show(struct device *dev,
985 struct device_attribute *attr, char *buf)
987 struct gendisk *disk = dev_to_disk(dev);
989 return sprintf(buf, "%d\n", disk_max_parts(disk));
992 static ssize_t disk_removable_show(struct device *dev,
993 struct device_attribute *attr, char *buf)
995 struct gendisk *disk = dev_to_disk(dev);
997 return sprintf(buf, "%d\n",
998 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
1001 static ssize_t disk_hidden_show(struct device *dev,
1002 struct device_attribute *attr, char *buf)
1004 struct gendisk *disk = dev_to_disk(dev);
1006 return sprintf(buf, "%d\n",
1007 (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
1010 static ssize_t disk_ro_show(struct device *dev,
1011 struct device_attribute *attr, char *buf)
1013 struct gendisk *disk = dev_to_disk(dev);
1015 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
1018 ssize_t part_size_show(struct device *dev,
1019 struct device_attribute *attr, char *buf)
1021 return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev)));
1024 ssize_t part_stat_show(struct device *dev,
1025 struct device_attribute *attr, char *buf)
1027 struct block_device *bdev = dev_to_bdev(dev);
1028 struct request_queue *q = bdev->bd_disk->queue;
1029 struct disk_stats stat;
1030 unsigned int inflight;
1032 part_stat_read_all(bdev, &stat);
1034 inflight = blk_mq_in_flight(q, bdev);
1036 inflight = part_in_flight(bdev);
1039 "%8lu %8lu %8llu %8u "
1040 "%8lu %8lu %8llu %8u "
1042 "%8lu %8lu %8llu %8u "
1045 stat.ios[STAT_READ],
1046 stat.merges[STAT_READ],
1047 (unsigned long long)stat.sectors[STAT_READ],
1048 (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC),
1049 stat.ios[STAT_WRITE],
1050 stat.merges[STAT_WRITE],
1051 (unsigned long long)stat.sectors[STAT_WRITE],
1052 (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC),
1054 jiffies_to_msecs(stat.io_ticks),
1055 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1056 stat.nsecs[STAT_WRITE] +
1057 stat.nsecs[STAT_DISCARD] +
1058 stat.nsecs[STAT_FLUSH],
1060 stat.ios[STAT_DISCARD],
1061 stat.merges[STAT_DISCARD],
1062 (unsigned long long)stat.sectors[STAT_DISCARD],
1063 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC),
1064 stat.ios[STAT_FLUSH],
1065 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC));
1068 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
1071 struct block_device *bdev = dev_to_bdev(dev);
1072 struct request_queue *q = bdev->bd_disk->queue;
1073 unsigned int inflight[2];
1076 blk_mq_in_flight_rw(q, bdev, inflight);
1078 part_in_flight_rw(bdev, inflight);
1080 return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
1083 static ssize_t disk_capability_show(struct device *dev,
1084 struct device_attribute *attr, char *buf)
1086 struct gendisk *disk = dev_to_disk(dev);
1088 return sprintf(buf, "%x\n", disk->flags);
1091 static ssize_t disk_alignment_offset_show(struct device *dev,
1092 struct device_attribute *attr,
1095 struct gendisk *disk = dev_to_disk(dev);
1097 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
1100 static ssize_t disk_discard_alignment_show(struct device *dev,
1101 struct device_attribute *attr,
1104 struct gendisk *disk = dev_to_disk(dev);
1106 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
1109 static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
1110 static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
1111 static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
1112 static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
1113 static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
1114 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
1115 static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
1116 static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
1117 static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
1118 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
1119 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
1120 static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
1122 #ifdef CONFIG_FAIL_MAKE_REQUEST
1123 ssize_t part_fail_show(struct device *dev,
1124 struct device_attribute *attr, char *buf)
1126 return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail);
1129 ssize_t part_fail_store(struct device *dev,
1130 struct device_attribute *attr,
1131 const char *buf, size_t count)
1135 if (count > 0 && sscanf(buf, "%d", &i) > 0)
1136 dev_to_bdev(dev)->bd_make_it_fail = i;
1141 static struct device_attribute dev_attr_fail =
1142 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
1143 #endif /* CONFIG_FAIL_MAKE_REQUEST */
1145 #ifdef CONFIG_FAIL_IO_TIMEOUT
1146 static struct device_attribute dev_attr_fail_timeout =
1147 __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
1150 static struct attribute *disk_attrs[] = {
1151 &dev_attr_range.attr,
1152 &dev_attr_ext_range.attr,
1153 &dev_attr_removable.attr,
1154 &dev_attr_hidden.attr,
1156 &dev_attr_size.attr,
1157 &dev_attr_alignment_offset.attr,
1158 &dev_attr_discard_alignment.attr,
1159 &dev_attr_capability.attr,
1160 &dev_attr_stat.attr,
1161 &dev_attr_inflight.attr,
1162 &dev_attr_badblocks.attr,
1163 #ifdef CONFIG_FAIL_MAKE_REQUEST
1164 &dev_attr_fail.attr,
1166 #ifdef CONFIG_FAIL_IO_TIMEOUT
1167 &dev_attr_fail_timeout.attr,
1172 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1174 struct device *dev = container_of(kobj, typeof(*dev), kobj);
1175 struct gendisk *disk = dev_to_disk(dev);
1177 if (a == &dev_attr_badblocks.attr && !disk->bb)
1182 static struct attribute_group disk_attr_group = {
1183 .attrs = disk_attrs,
1184 .is_visible = disk_visible,
1187 static const struct attribute_group *disk_attr_groups[] = {
1193 * disk_release - releases all allocated resources of the gendisk
1194 * @dev: the device representing this disk
1196 * This function releases all allocated resources of the gendisk.
1198 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1199 * assigned. Since the request_queue sits on top of the gendisk for these
1200 * drivers we also call blk_put_queue() for them, and we expect the
1201 * request_queue refcount to reach 0 at this point, and so the request_queue
1202 * will also be freed prior to the disk.
1204 * Context: can sleep
1206 static void disk_release(struct device *dev)
1208 struct gendisk *disk = dev_to_disk(dev);
1212 blk_free_devt(dev->devt);
1213 disk_release_events(disk);
1214 kfree(disk->random);
1215 xa_destroy(&disk->part_tbl);
1218 blk_put_queue(disk->queue);
1221 struct class block_class = {
1225 static char *block_devnode(struct device *dev, umode_t *mode,
1226 kuid_t *uid, kgid_t *gid)
1228 struct gendisk *disk = dev_to_disk(dev);
1230 if (disk->fops->devnode)
1231 return disk->fops->devnode(disk, mode);
1235 const struct device_type disk_type = {
1237 .groups = disk_attr_groups,
1238 .release = disk_release,
1239 .devnode = block_devnode,
1242 #ifdef CONFIG_PROC_FS
1244 * aggregate disk stat collector. Uses the same stats that the sysfs
1245 * entries do, above, but makes them available through one seq_file.
1247 * The output looks suspiciously like /proc/partitions with a bunch of
1250 static int diskstats_show(struct seq_file *seqf, void *v)
1252 struct gendisk *gp = v;
1253 struct disk_part_iter piter;
1254 struct block_device *hd;
1255 char buf[BDEVNAME_SIZE];
1256 unsigned int inflight;
1257 struct disk_stats stat;
1260 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1261 seq_puts(seqf, "major minor name"
1262 " rio rmerge rsect ruse wio wmerge "
1263 "wsect wuse running use aveq"
1267 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
1268 while ((hd = disk_part_iter_next(&piter))) {
1269 part_stat_read_all(hd, &stat);
1270 if (queue_is_mq(gp->queue))
1271 inflight = blk_mq_in_flight(gp->queue, hd);
1273 inflight = part_in_flight(hd);
1275 seq_printf(seqf, "%4d %7d %s "
1282 MAJOR(hd->bd_dev), MINOR(hd->bd_dev),
1283 disk_name(gp, hd->bd_partno, buf),
1284 stat.ios[STAT_READ],
1285 stat.merges[STAT_READ],
1286 stat.sectors[STAT_READ],
1287 (unsigned int)div_u64(stat.nsecs[STAT_READ],
1289 stat.ios[STAT_WRITE],
1290 stat.merges[STAT_WRITE],
1291 stat.sectors[STAT_WRITE],
1292 (unsigned int)div_u64(stat.nsecs[STAT_WRITE],
1295 jiffies_to_msecs(stat.io_ticks),
1296 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1297 stat.nsecs[STAT_WRITE] +
1298 stat.nsecs[STAT_DISCARD] +
1299 stat.nsecs[STAT_FLUSH],
1301 stat.ios[STAT_DISCARD],
1302 stat.merges[STAT_DISCARD],
1303 stat.sectors[STAT_DISCARD],
1304 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD],
1306 stat.ios[STAT_FLUSH],
1307 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH],
1311 disk_part_iter_exit(&piter);
1316 static const struct seq_operations diskstats_op = {
1317 .start = disk_seqf_start,
1318 .next = disk_seqf_next,
1319 .stop = disk_seqf_stop,
1320 .show = diskstats_show
1323 static int __init proc_genhd_init(void)
1325 proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1326 proc_create_seq("partitions", 0, NULL, &partitions_op);
1329 module_init(proc_genhd_init);
1330 #endif /* CONFIG_PROC_FS */
1332 dev_t blk_lookup_devt(const char *name, int partno)
1334 dev_t devt = MKDEV(0, 0);
1335 struct class_dev_iter iter;
1338 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1339 while ((dev = class_dev_iter_next(&iter))) {
1340 struct gendisk *disk = dev_to_disk(dev);
1341 struct block_device *part;
1343 if (strcmp(dev_name(dev), name))
1346 if (partno < disk->minors) {
1347 /* We need to return the right devno, even
1348 * if the partition doesn't exist yet.
1350 devt = MKDEV(MAJOR(dev->devt),
1351 MINOR(dev->devt) + partno);
1354 part = bdget_disk(disk, partno);
1356 devt = part->bd_dev;
1361 class_dev_iter_exit(&iter);
1365 struct gendisk *__alloc_disk_node(int minors, int node_id)
1367 struct gendisk *disk;
1369 if (minors > DISK_MAX_PARTS) {
1371 "block: can't allocate more than %d partitions\n",
1373 minors = DISK_MAX_PARTS;
1376 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1380 disk->part0 = bdev_alloc(disk, 0);
1384 disk->node_id = node_id;
1385 xa_init(&disk->part_tbl);
1386 if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL))
1387 goto out_destroy_part_tbl;
1389 disk->minors = minors;
1390 rand_initialize_disk(disk);
1391 disk_to_dev(disk)->class = &block_class;
1392 disk_to_dev(disk)->type = &disk_type;
1393 device_initialize(disk_to_dev(disk));
1396 out_destroy_part_tbl:
1397 xa_destroy(&disk->part_tbl);
1403 EXPORT_SYMBOL(__alloc_disk_node);
1406 * put_disk - decrements the gendisk refcount
1407 * @disk: the struct gendisk to decrement the refcount for
1409 * This decrements the refcount for the struct gendisk. When this reaches 0
1410 * we'll have disk_release() called.
1412 * Context: Any context, but the last reference must not be dropped from
1415 void put_disk(struct gendisk *disk)
1418 put_device(disk_to_dev(disk));
1420 EXPORT_SYMBOL(put_disk);
1422 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1424 char event[] = "DISK_RO=1";
1425 char *envp[] = { event, NULL };
1429 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1433 * set_disk_ro - set a gendisk read-only
1434 * @disk: gendisk to operate on
1435 * @read_only: %true to set the disk read-only, %false set the disk read/write
1437 * This function is used to indicate whether a given disk device should have its
1438 * read-only flag set. set_disk_ro() is typically used by device drivers to
1439 * indicate whether the underlying physical device is write-protected.
1441 void set_disk_ro(struct gendisk *disk, bool read_only)
1444 if (test_and_set_bit(GD_READ_ONLY, &disk->state))
1447 if (!test_and_clear_bit(GD_READ_ONLY, &disk->state))
1450 set_disk_ro_uevent(disk, read_only);
1452 EXPORT_SYMBOL(set_disk_ro);
1454 int bdev_read_only(struct block_device *bdev)
1456 return bdev->bd_read_only || get_disk_ro(bdev->bd_disk);
1458 EXPORT_SYMBOL(bdev_read_only);
1461 * Disk events - monitor disk events like media change and eject request.
1463 struct disk_events {
1464 struct list_head node; /* all disk_event's */
1465 struct gendisk *disk; /* the associated disk */
1468 struct mutex block_mutex; /* protects blocking */
1469 int block; /* event blocking depth */
1470 unsigned int pending; /* events already sent out */
1471 unsigned int clearing; /* events being cleared */
1473 long poll_msecs; /* interval, -1 for default */
1474 struct delayed_work dwork;
1477 static const char *disk_events_strs[] = {
1478 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1479 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1482 static char *disk_uevents[] = {
1483 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1484 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1487 /* list of all disk_events */
1488 static DEFINE_MUTEX(disk_events_mutex);
1489 static LIST_HEAD(disk_events);
1491 /* disable in-kernel polling by default */
1492 static unsigned long disk_events_dfl_poll_msecs;
1494 static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1496 struct disk_events *ev = disk->ev;
1497 long intv_msecs = 0;
1500 * If device-specific poll interval is set, always use it. If
1501 * the default is being used, poll if the POLL flag is set.
1503 if (ev->poll_msecs >= 0)
1504 intv_msecs = ev->poll_msecs;
1505 else if (disk->event_flags & DISK_EVENT_FLAG_POLL)
1506 intv_msecs = disk_events_dfl_poll_msecs;
1508 return msecs_to_jiffies(intv_msecs);
1512 * disk_block_events - block and flush disk event checking
1513 * @disk: disk to block events for
1515 * On return from this function, it is guaranteed that event checking
1516 * isn't in progress and won't happen until unblocked by
1517 * disk_unblock_events(). Events blocking is counted and the actual
1518 * unblocking happens after the matching number of unblocks are done.
1520 * Note that this intentionally does not block event checking from
1521 * disk_clear_events().
1526 void disk_block_events(struct gendisk *disk)
1528 struct disk_events *ev = disk->ev;
1529 unsigned long flags;
1536 * Outer mutex ensures that the first blocker completes canceling
1537 * the event work before further blockers are allowed to finish.
1539 mutex_lock(&ev->block_mutex);
1541 spin_lock_irqsave(&ev->lock, flags);
1542 cancel = !ev->block++;
1543 spin_unlock_irqrestore(&ev->lock, flags);
1546 cancel_delayed_work_sync(&disk->ev->dwork);
1548 mutex_unlock(&ev->block_mutex);
1551 static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1553 struct disk_events *ev = disk->ev;
1555 unsigned long flags;
1557 spin_lock_irqsave(&ev->lock, flags);
1559 if (WARN_ON_ONCE(ev->block <= 0))
1565 intv = disk_events_poll_jiffies(disk);
1567 queue_delayed_work(system_freezable_power_efficient_wq,
1570 queue_delayed_work(system_freezable_power_efficient_wq,
1573 spin_unlock_irqrestore(&ev->lock, flags);
1577 * disk_unblock_events - unblock disk event checking
1578 * @disk: disk to unblock events for
1580 * Undo disk_block_events(). When the block count reaches zero, it
1581 * starts events polling if configured.
1584 * Don't care. Safe to call from irq context.
1586 void disk_unblock_events(struct gendisk *disk)
1589 __disk_unblock_events(disk, false);
1593 * disk_flush_events - schedule immediate event checking and flushing
1594 * @disk: disk to check and flush events for
1595 * @mask: events to flush
1597 * Schedule immediate event checking on @disk if not blocked. Events in
1598 * @mask are scheduled to be cleared from the driver. Note that this
1599 * doesn't clear the events from @disk->ev.
1602 * If @mask is non-zero must be called with bdev->bd_mutex held.
1604 void disk_flush_events(struct gendisk *disk, unsigned int mask)
1606 struct disk_events *ev = disk->ev;
1611 spin_lock_irq(&ev->lock);
1612 ev->clearing |= mask;
1614 mod_delayed_work(system_freezable_power_efficient_wq,
1616 spin_unlock_irq(&ev->lock);
1620 * disk_clear_events - synchronously check, clear and return pending events
1621 * @disk: disk to fetch and clear events from
1622 * @mask: mask of events to be fetched and cleared
1624 * Disk events are synchronously checked and pending events in @mask
1625 * are cleared and returned. This ignores the block count.
1630 static unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1632 struct disk_events *ev = disk->ev;
1633 unsigned int pending;
1634 unsigned int clearing = mask;
1639 disk_block_events(disk);
1642 * store the union of mask and ev->clearing on the stack so that the
1643 * race with disk_flush_events does not cause ambiguity (ev->clearing
1644 * can still be modified even if events are blocked).
1646 spin_lock_irq(&ev->lock);
1647 clearing |= ev->clearing;
1649 spin_unlock_irq(&ev->lock);
1651 disk_check_events(ev, &clearing);
1653 * if ev->clearing is not 0, the disk_flush_events got called in the
1654 * middle of this function, so we want to run the workfn without delay.
1656 __disk_unblock_events(disk, ev->clearing ? true : false);
1658 /* then, fetch and clear pending events */
1659 spin_lock_irq(&ev->lock);
1660 pending = ev->pending & mask;
1661 ev->pending &= ~mask;
1662 spin_unlock_irq(&ev->lock);
1663 WARN_ON_ONCE(clearing & mask);
1669 * bdev_check_media_change - check if a removable media has been changed
1670 * @bdev: block device to check
1672 * Check whether a removable media has been changed, and attempt to free all
1673 * dentries and inodes and invalidates all block device page cache entries in
1676 * Returns %true if the block device changed, or %false if not.
1678 bool bdev_check_media_change(struct block_device *bdev)
1680 unsigned int events;
1682 events = disk_clear_events(bdev->bd_disk, DISK_EVENT_MEDIA_CHANGE |
1683 DISK_EVENT_EJECT_REQUEST);
1684 if (!(events & DISK_EVENT_MEDIA_CHANGE))
1687 if (__invalidate_device(bdev, true))
1688 pr_warn("VFS: busy inodes on changed media %s\n",
1689 bdev->bd_disk->disk_name);
1690 set_bit(GD_NEED_PART_SCAN, &bdev->bd_disk->state);
1693 EXPORT_SYMBOL(bdev_check_media_change);
1696 * Separate this part out so that a different pointer for clearing_ptr can be
1697 * passed in for disk_clear_events.
1699 static void disk_events_workfn(struct work_struct *work)
1701 struct delayed_work *dwork = to_delayed_work(work);
1702 struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
1704 disk_check_events(ev, &ev->clearing);
1707 static void disk_check_events(struct disk_events *ev,
1708 unsigned int *clearing_ptr)
1710 struct gendisk *disk = ev->disk;
1711 char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
1712 unsigned int clearing = *clearing_ptr;
1713 unsigned int events;
1715 int nr_events = 0, i;
1718 events = disk->fops->check_events(disk, clearing);
1720 /* accumulate pending events and schedule next poll if necessary */
1721 spin_lock_irq(&ev->lock);
1723 events &= ~ev->pending;
1724 ev->pending |= events;
1725 *clearing_ptr &= ~clearing;
1727 intv = disk_events_poll_jiffies(disk);
1728 if (!ev->block && intv)
1729 queue_delayed_work(system_freezable_power_efficient_wq,
1732 spin_unlock_irq(&ev->lock);
1735 * Tell userland about new events. Only the events listed in
1736 * @disk->events are reported, and only if DISK_EVENT_FLAG_UEVENT
1737 * is set. Otherwise, events are processed internally but never
1738 * get reported to userland.
1740 for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
1741 if ((events & disk->events & (1 << i)) &&
1742 (disk->event_flags & DISK_EVENT_FLAG_UEVENT))
1743 envp[nr_events++] = disk_uevents[i];
1746 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1750 * A disk events enabled device has the following sysfs nodes under
1751 * its /sys/block/X/ directory.
1753 * events : list of all supported events
1754 * events_async : list of events which can be detected w/o polling
1755 * (always empty, only for backwards compatibility)
1756 * events_poll_msecs : polling interval, 0: disable, -1: system default
1758 static ssize_t __disk_events_show(unsigned int events, char *buf)
1760 const char *delim = "";
1764 for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1765 if (events & (1 << i)) {
1766 pos += sprintf(buf + pos, "%s%s",
1767 delim, disk_events_strs[i]);
1771 pos += sprintf(buf + pos, "\n");
1775 static ssize_t disk_events_show(struct device *dev,
1776 struct device_attribute *attr, char *buf)
1778 struct gendisk *disk = dev_to_disk(dev);
1780 if (!(disk->event_flags & DISK_EVENT_FLAG_UEVENT))
1783 return __disk_events_show(disk->events, buf);
1786 static ssize_t disk_events_async_show(struct device *dev,
1787 struct device_attribute *attr, char *buf)
1792 static ssize_t disk_events_poll_msecs_show(struct device *dev,
1793 struct device_attribute *attr,
1796 struct gendisk *disk = dev_to_disk(dev);
1799 return sprintf(buf, "-1\n");
1801 return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1804 static ssize_t disk_events_poll_msecs_store(struct device *dev,
1805 struct device_attribute *attr,
1806 const char *buf, size_t count)
1808 struct gendisk *disk = dev_to_disk(dev);
1811 if (!count || !sscanf(buf, "%ld", &intv))
1814 if (intv < 0 && intv != -1)
1820 disk_block_events(disk);
1821 disk->ev->poll_msecs = intv;
1822 __disk_unblock_events(disk, true);
1827 static const DEVICE_ATTR(events, 0444, disk_events_show, NULL);
1828 static const DEVICE_ATTR(events_async, 0444, disk_events_async_show, NULL);
1829 static const DEVICE_ATTR(events_poll_msecs, 0644,
1830 disk_events_poll_msecs_show,
1831 disk_events_poll_msecs_store);
1833 static const struct attribute *disk_events_attrs[] = {
1834 &dev_attr_events.attr,
1835 &dev_attr_events_async.attr,
1836 &dev_attr_events_poll_msecs.attr,
1841 * The default polling interval can be specified by the kernel
1842 * parameter block.events_dfl_poll_msecs which defaults to 0
1843 * (disable). This can also be modified runtime by writing to
1844 * /sys/module/block/parameters/events_dfl_poll_msecs.
1846 static int disk_events_set_dfl_poll_msecs(const char *val,
1847 const struct kernel_param *kp)
1849 struct disk_events *ev;
1852 ret = param_set_ulong(val, kp);
1856 mutex_lock(&disk_events_mutex);
1858 list_for_each_entry(ev, &disk_events, node)
1859 disk_flush_events(ev->disk, 0);
1861 mutex_unlock(&disk_events_mutex);
1866 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1867 .set = disk_events_set_dfl_poll_msecs,
1868 .get = param_get_ulong,
1871 #undef MODULE_PARAM_PREFIX
1872 #define MODULE_PARAM_PREFIX "block."
1874 module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
1875 &disk_events_dfl_poll_msecs, 0644);
1878 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
1880 static void disk_alloc_events(struct gendisk *disk)
1882 struct disk_events *ev;
1884 if (!disk->fops->check_events || !disk->events)
1887 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1889 pr_warn("%s: failed to initialize events\n", disk->disk_name);
1893 INIT_LIST_HEAD(&ev->node);
1895 spin_lock_init(&ev->lock);
1896 mutex_init(&ev->block_mutex);
1898 ev->poll_msecs = -1;
1899 INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
1904 static void disk_add_events(struct gendisk *disk)
1906 /* FIXME: error handling */
1907 if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0)
1908 pr_warn("%s: failed to create sysfs files for events\n",
1914 mutex_lock(&disk_events_mutex);
1915 list_add_tail(&disk->ev->node, &disk_events);
1916 mutex_unlock(&disk_events_mutex);
1919 * Block count is initialized to 1 and the following initial
1920 * unblock kicks it into action.
1922 __disk_unblock_events(disk, true);
1925 static void disk_del_events(struct gendisk *disk)
1928 disk_block_events(disk);
1930 mutex_lock(&disk_events_mutex);
1931 list_del_init(&disk->ev->node);
1932 mutex_unlock(&disk_events_mutex);
1935 sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
1938 static void disk_release_events(struct gendisk *disk)
1940 /* the block count should be 1 from disk_del_events() */
1941 WARN_ON_ONCE(disk->ev && disk->ev->block != 1);