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>
29 #include "blk-rq-qos.h"
31 static struct kobject *block_depr;
34 * Unique, monotonically increasing sequential number associated with block
35 * devices instances (i.e. incremented each time a device is attached).
36 * Associating uevents with block devices in userspace is difficult and racy:
37 * the uevent netlink socket is lossy, and on slow and overloaded systems has
38 * a very high latency.
39 * Block devices do not have exclusive owners in userspace, any process can set
40 * one up (e.g. loop devices). Moreover, device names can be reused (e.g. loop0
41 * can be reused again and again).
42 * A userspace process setting up a block device and watching for its events
43 * cannot thus reliably tell whether an event relates to the device it just set
44 * up or another earlier instance with the same name.
45 * This sequential number allows userspace processes to solve this problem, and
46 * uniquely associate an uevent to the lifetime to a device.
48 static atomic64_t diskseq;
50 /* for extended dynamic devt allocation, currently only one major is used */
51 #define NR_EXT_DEVT (1 << MINORBITS)
52 static DEFINE_IDA(ext_devt_ida);
54 void set_capacity(struct gendisk *disk, sector_t sectors)
56 struct block_device *bdev = disk->part0;
58 spin_lock(&bdev->bd_size_lock);
59 i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
60 spin_unlock(&bdev->bd_size_lock);
62 EXPORT_SYMBOL(set_capacity);
65 * Set disk capacity and notify if the size is not currently zero and will not
66 * be set to zero. Returns true if a uevent was sent, otherwise false.
68 bool set_capacity_and_notify(struct gendisk *disk, sector_t size)
70 sector_t capacity = get_capacity(disk);
71 char *envp[] = { "RESIZE=1", NULL };
73 set_capacity(disk, size);
76 * Only print a message and send a uevent if the gendisk is user visible
77 * and alive. This avoids spamming the log and udev when setting the
78 * initial capacity during probing.
80 if (size == capacity ||
82 (disk->flags & GENHD_FL_HIDDEN))
85 pr_info("%s: detected capacity change from %lld to %lld\n",
86 disk->disk_name, capacity, size);
89 * Historically we did not send a uevent for changes to/from an empty
92 if (!capacity || !size)
94 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
97 EXPORT_SYMBOL_GPL(set_capacity_and_notify);
100 * Format the device name of the indicated block device into the supplied buffer
101 * and return a pointer to that same buffer for convenience.
103 * Note: do not use this in new code, use the %pg specifier to sprintf and
106 const char *bdevname(struct block_device *bdev, char *buf)
108 struct gendisk *hd = bdev->bd_disk;
109 int partno = bdev->bd_partno;
112 snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name);
113 else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1]))
114 snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno);
116 snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno);
120 EXPORT_SYMBOL(bdevname);
122 static void part_stat_read_all(struct block_device *part,
123 struct disk_stats *stat)
127 memset(stat, 0, sizeof(struct disk_stats));
128 for_each_possible_cpu(cpu) {
129 struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu);
132 for (group = 0; group < NR_STAT_GROUPS; group++) {
133 stat->nsecs[group] += ptr->nsecs[group];
134 stat->sectors[group] += ptr->sectors[group];
135 stat->ios[group] += ptr->ios[group];
136 stat->merges[group] += ptr->merges[group];
139 stat->io_ticks += ptr->io_ticks;
143 static unsigned int part_in_flight(struct block_device *part)
145 unsigned int inflight = 0;
148 for_each_possible_cpu(cpu) {
149 inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) +
150 part_stat_local_read_cpu(part, in_flight[1], cpu);
152 if ((int)inflight < 0)
158 static void part_in_flight_rw(struct block_device *part,
159 unsigned int inflight[2])
165 for_each_possible_cpu(cpu) {
166 inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu);
167 inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu);
169 if ((int)inflight[0] < 0)
171 if ((int)inflight[1] < 0)
176 * Can be deleted altogether. Later.
179 #define BLKDEV_MAJOR_HASH_SIZE 255
180 static struct blk_major_name {
181 struct blk_major_name *next;
184 void (*probe)(dev_t devt);
185 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
186 static DEFINE_MUTEX(major_names_lock);
187 static DEFINE_SPINLOCK(major_names_spinlock);
189 /* index in the above - for now: assume no multimajor ranges */
190 static inline int major_to_index(unsigned major)
192 return major % BLKDEV_MAJOR_HASH_SIZE;
195 #ifdef CONFIG_PROC_FS
196 void blkdev_show(struct seq_file *seqf, off_t offset)
198 struct blk_major_name *dp;
200 spin_lock(&major_names_spinlock);
201 for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
202 if (dp->major == offset)
203 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
204 spin_unlock(&major_names_spinlock);
206 #endif /* CONFIG_PROC_FS */
209 * __register_blkdev - register a new block device
211 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
212 * @major = 0, try to allocate any unused major number.
213 * @name: the name of the new block device as a zero terminated string
214 * @probe: allback that is called on access to any minor number of @major
216 * The @name must be unique within the system.
218 * The return value depends on the @major input parameter:
220 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
221 * then the function returns zero on success, or a negative error code
222 * - if any unused major number was requested with @major = 0 parameter
223 * then the return value is the allocated major number in range
224 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
226 * See Documentation/admin-guide/devices.txt for the list of allocated
229 * Use register_blkdev instead for any new code.
231 int __register_blkdev(unsigned int major, const char *name,
232 void (*probe)(dev_t devt))
234 struct blk_major_name **n, *p;
237 mutex_lock(&major_names_lock);
241 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
242 if (major_names[index] == NULL)
247 printk("%s: failed to get major for %s\n",
256 if (major >= BLKDEV_MAJOR_MAX) {
257 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
258 __func__, major, BLKDEV_MAJOR_MAX-1, name);
264 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
272 strlcpy(p->name, name, sizeof(p->name));
274 index = major_to_index(major);
276 spin_lock(&major_names_spinlock);
277 for (n = &major_names[index]; *n; n = &(*n)->next) {
278 if ((*n)->major == major)
285 spin_unlock(&major_names_spinlock);
288 printk("register_blkdev: cannot get major %u for %s\n",
293 mutex_unlock(&major_names_lock);
296 EXPORT_SYMBOL(__register_blkdev);
298 void unregister_blkdev(unsigned int major, const char *name)
300 struct blk_major_name **n;
301 struct blk_major_name *p = NULL;
302 int index = major_to_index(major);
304 mutex_lock(&major_names_lock);
305 spin_lock(&major_names_spinlock);
306 for (n = &major_names[index]; *n; n = &(*n)->next)
307 if ((*n)->major == major)
309 if (!*n || strcmp((*n)->name, name)) {
315 spin_unlock(&major_names_spinlock);
316 mutex_unlock(&major_names_lock);
320 EXPORT_SYMBOL(unregister_blkdev);
322 int blk_alloc_ext_minor(void)
326 idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT, GFP_KERNEL);
332 void blk_free_ext_minor(unsigned int minor)
334 ida_free(&ext_devt_ida, minor);
337 static char *bdevt_str(dev_t devt, char *buf)
339 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
340 char tbuf[BDEVT_SIZE];
341 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
342 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
344 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
349 void disk_uevent(struct gendisk *disk, enum kobject_action action)
351 struct block_device *part;
355 xa_for_each(&disk->part_tbl, idx, part) {
356 if (bdev_is_partition(part) && !bdev_nr_sectors(part))
358 if (!kobject_get_unless_zero(&part->bd_device.kobj))
362 kobject_uevent(bdev_kobj(part), action);
363 put_device(&part->bd_device);
368 EXPORT_SYMBOL_GPL(disk_uevent);
370 static void disk_scan_partitions(struct gendisk *disk)
372 struct block_device *bdev;
374 if (!get_capacity(disk) || !disk_part_scan_enabled(disk))
377 set_bit(GD_NEED_PART_SCAN, &disk->state);
378 bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL);
380 blkdev_put(bdev, FMODE_READ);
384 * device_add_disk - add disk information to kernel list
385 * @parent: parent device for the disk
386 * @disk: per-device partitioning information
387 * @groups: Additional per-device sysfs groups
389 * This function registers the partitioning information in @disk
392 int device_add_disk(struct device *parent, struct gendisk *disk,
393 const struct attribute_group **groups)
396 struct device *ddev = disk_to_dev(disk);
400 * The disk queue should now be all set with enough information about
401 * the device for the elevator code to pick an adequate default
402 * elevator if one is needed, that is, for devices requesting queue
405 elevator_init_mq(disk->queue);
408 * If the driver provides an explicit major number it also must provide
409 * the number of minors numbers supported, and those will be used to
411 * Otherwise just allocate the device numbers for both the whole device
412 * and all partitions from the extended dev_t space.
415 if (WARN_ON(!disk->minors))
418 if (disk->minors > DISK_MAX_PARTS) {
419 pr_err("block: can't allocate more than %d partitions\n",
421 disk->minors = DISK_MAX_PARTS;
424 if (WARN_ON(disk->minors))
427 ret = blk_alloc_ext_minor();
430 disk->major = BLOCK_EXT_MAJOR;
431 disk->first_minor = ret;
432 disk->flags |= GENHD_FL_EXT_DEVT;
435 /* delay uevents, until we scanned partition table */
436 dev_set_uevent_suppress(ddev, 1);
438 ddev->parent = parent;
439 ddev->groups = groups;
440 dev_set_name(ddev, "%s", disk->disk_name);
441 if (!(disk->flags & GENHD_FL_HIDDEN))
442 ddev->devt = MKDEV(disk->major, disk->first_minor);
443 ret = device_add(ddev);
445 goto out_free_ext_minor;
447 ret = disk_alloc_events(disk);
451 if (!sysfs_deprecated) {
452 ret = sysfs_create_link(block_depr, &ddev->kobj,
453 kobject_name(&ddev->kobj));
459 * avoid probable deadlock caused by allocating memory with
460 * GFP_KERNEL in runtime_resume callback of its all ancestor
463 pm_runtime_set_memalloc_noio(ddev, true);
465 ret = blk_integrity_add(disk);
467 goto out_del_block_link;
469 disk->part0->bd_holder_dir =
470 kobject_create_and_add("holders", &ddev->kobj);
471 if (!disk->part0->bd_holder_dir) {
473 goto out_del_integrity;
475 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
476 if (!disk->slave_dir) {
478 goto out_put_holder_dir;
481 ret = bd_register_pending_holders(disk);
483 goto out_put_slave_dir;
485 ret = blk_register_queue(disk);
487 goto out_put_slave_dir;
489 if (disk->flags & GENHD_FL_HIDDEN) {
491 * Don't let hidden disks show up in /proc/partitions,
492 * and don't bother scanning for partitions either.
494 disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
495 disk->flags |= GENHD_FL_NO_PART_SCAN;
497 ret = bdi_register(disk->bdi, "%u:%u",
498 disk->major, disk->first_minor);
500 goto out_unregister_queue;
501 bdi_set_owner(disk->bdi, ddev);
502 ret = sysfs_create_link(&ddev->kobj,
503 &disk->bdi->dev->kobj, "bdi");
505 goto out_unregister_bdi;
507 bdev_add(disk->part0, ddev->devt);
508 disk_scan_partitions(disk);
511 * Announce the disk and partitions after all partitions are
512 * created. (for hidden disks uevents remain suppressed forever)
514 dev_set_uevent_suppress(ddev, 0);
515 disk_uevent(disk, KOBJ_ADD);
518 disk_update_readahead(disk);
519 disk_add_events(disk);
523 if (!(disk->flags & GENHD_FL_HIDDEN))
524 bdi_unregister(disk->bdi);
525 out_unregister_queue:
526 blk_unregister_queue(disk);
528 kobject_put(disk->slave_dir);
530 kobject_put(disk->part0->bd_holder_dir);
532 blk_integrity_del(disk);
534 if (!sysfs_deprecated)
535 sysfs_remove_link(block_depr, dev_name(ddev));
539 if (disk->major == BLOCK_EXT_MAJOR)
540 blk_free_ext_minor(disk->first_minor);
541 return WARN_ON_ONCE(ret); /* keep until all callers handle errors */
543 EXPORT_SYMBOL(device_add_disk);
546 * del_gendisk - remove the gendisk
547 * @disk: the struct gendisk to remove
549 * Removes the gendisk and all its associated resources. This deletes the
550 * partitions associated with the gendisk, and unregisters the associated
553 * This is the counter to the respective __device_add_disk() call.
555 * The final removal of the struct gendisk happens when its refcount reaches 0
556 * with put_disk(), which should be called after del_gendisk(), if
557 * __device_add_disk() was used.
559 * Drivers exist which depend on the release of the gendisk to be synchronous,
560 * it should not be deferred.
564 void del_gendisk(struct gendisk *disk)
566 struct request_queue *q = disk->queue;
570 if (WARN_ON_ONCE(!disk_live(disk) && !(disk->flags & GENHD_FL_HIDDEN)))
573 blk_integrity_del(disk);
574 disk_del_events(disk);
576 mutex_lock(&disk->open_mutex);
577 remove_inode_hash(disk->part0->bd_inode);
578 blk_drop_partitions(disk);
579 mutex_unlock(&disk->open_mutex);
581 fsync_bdev(disk->part0);
582 __invalidate_device(disk->part0, true);
587 set_bit(GD_DEAD, &disk->state);
588 set_capacity(disk, 0);
591 * Prevent new I/O from crossing bio_queue_enter().
593 blk_queue_start_drain(q);
595 if (!(disk->flags & GENHD_FL_HIDDEN)) {
596 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
599 * Unregister bdi before releasing device numbers (as they can
600 * get reused and we'd get clashes in sysfs).
602 bdi_unregister(disk->bdi);
605 blk_unregister_queue(disk);
607 kobject_put(disk->part0->bd_holder_dir);
608 kobject_put(disk->slave_dir);
610 part_stat_set_all(disk->part0, 0);
611 disk->part0->bd_stamp = 0;
612 if (!sysfs_deprecated)
613 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
614 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
615 device_del(disk_to_dev(disk));
617 blk_mq_freeze_queue_wait(q);
621 blk_flush_integrity();
623 * Allow using passthrough request again after the queue is torn down.
625 blk_queue_flag_clear(QUEUE_FLAG_INIT_DONE, q);
626 __blk_mq_unfreeze_queue(q, true);
629 EXPORT_SYMBOL(del_gendisk);
631 /* sysfs access to bad-blocks list. */
632 static ssize_t disk_badblocks_show(struct device *dev,
633 struct device_attribute *attr,
636 struct gendisk *disk = dev_to_disk(dev);
639 return sprintf(page, "\n");
641 return badblocks_show(disk->bb, page, 0);
644 static ssize_t disk_badblocks_store(struct device *dev,
645 struct device_attribute *attr,
646 const char *page, size_t len)
648 struct gendisk *disk = dev_to_disk(dev);
653 return badblocks_store(disk->bb, page, len, 0);
656 void blk_request_module(dev_t devt)
658 unsigned int major = MAJOR(devt);
659 struct blk_major_name **n;
661 mutex_lock(&major_names_lock);
662 for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) {
663 if ((*n)->major == major && (*n)->probe) {
665 mutex_unlock(&major_names_lock);
669 mutex_unlock(&major_names_lock);
671 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
672 /* Make old-style 2.4 aliases work */
673 request_module("block-major-%d", MAJOR(devt));
677 * print a full list of all partitions - intended for places where the root
678 * filesystem can't be mounted and thus to give the victim some idea of what
681 void __init printk_all_partitions(void)
683 struct class_dev_iter iter;
686 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
687 while ((dev = class_dev_iter_next(&iter))) {
688 struct gendisk *disk = dev_to_disk(dev);
689 struct block_device *part;
690 char devt_buf[BDEVT_SIZE];
694 * Don't show empty devices or things that have been
697 if (get_capacity(disk) == 0 ||
698 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
702 * Note, unlike /proc/partitions, I am showing the numbers in
703 * hex - the same format as the root= option takes.
706 xa_for_each(&disk->part_tbl, idx, part) {
707 if (!bdev_nr_sectors(part))
709 printk("%s%s %10llu %pg %s",
710 bdev_is_partition(part) ? " " : "",
711 bdevt_str(part->bd_dev, devt_buf),
712 bdev_nr_sectors(part) >> 1, part,
714 part->bd_meta_info->uuid : "");
715 if (bdev_is_partition(part))
717 else if (dev->parent && dev->parent->driver)
718 printk(" driver: %s\n",
719 dev->parent->driver->name);
721 printk(" (driver?)\n");
725 class_dev_iter_exit(&iter);
728 #ifdef CONFIG_PROC_FS
730 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
733 struct class_dev_iter *iter;
736 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
738 return ERR_PTR(-ENOMEM);
740 seqf->private = iter;
741 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
743 dev = class_dev_iter_next(iter);
748 return dev_to_disk(dev);
751 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
756 dev = class_dev_iter_next(seqf->private);
758 return dev_to_disk(dev);
763 static void disk_seqf_stop(struct seq_file *seqf, void *v)
765 struct class_dev_iter *iter = seqf->private;
767 /* stop is called even after start failed :-( */
769 class_dev_iter_exit(iter);
771 seqf->private = NULL;
775 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
779 p = disk_seqf_start(seqf, pos);
780 if (!IS_ERR_OR_NULL(p) && !*pos)
781 seq_puts(seqf, "major minor #blocks name\n\n");
785 static int show_partition(struct seq_file *seqf, void *v)
787 struct gendisk *sgp = v;
788 struct block_device *part;
791 /* Don't show non-partitionable removeable devices or empty devices */
792 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
793 (sgp->flags & GENHD_FL_REMOVABLE)))
795 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
799 xa_for_each(&sgp->part_tbl, idx, part) {
800 if (!bdev_nr_sectors(part))
802 seq_printf(seqf, "%4d %7d %10llu %pg\n",
803 MAJOR(part->bd_dev), MINOR(part->bd_dev),
804 bdev_nr_sectors(part) >> 1, part);
810 static const struct seq_operations partitions_op = {
811 .start = show_partition_start,
812 .next = disk_seqf_next,
813 .stop = disk_seqf_stop,
814 .show = show_partition
818 static int __init genhd_device_init(void)
822 block_class.dev_kobj = sysfs_dev_block_kobj;
823 error = class_register(&block_class);
828 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
830 /* create top-level block dir */
831 if (!sysfs_deprecated)
832 block_depr = kobject_create_and_add("block", NULL);
836 subsys_initcall(genhd_device_init);
838 static ssize_t disk_range_show(struct device *dev,
839 struct device_attribute *attr, char *buf)
841 struct gendisk *disk = dev_to_disk(dev);
843 return sprintf(buf, "%d\n", disk->minors);
846 static ssize_t disk_ext_range_show(struct device *dev,
847 struct device_attribute *attr, char *buf)
849 struct gendisk *disk = dev_to_disk(dev);
851 return sprintf(buf, "%d\n", disk_max_parts(disk));
854 static ssize_t disk_removable_show(struct device *dev,
855 struct device_attribute *attr, char *buf)
857 struct gendisk *disk = dev_to_disk(dev);
859 return sprintf(buf, "%d\n",
860 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
863 static ssize_t disk_hidden_show(struct device *dev,
864 struct device_attribute *attr, char *buf)
866 struct gendisk *disk = dev_to_disk(dev);
868 return sprintf(buf, "%d\n",
869 (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
872 static ssize_t disk_ro_show(struct device *dev,
873 struct device_attribute *attr, char *buf)
875 struct gendisk *disk = dev_to_disk(dev);
877 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
880 ssize_t part_size_show(struct device *dev,
881 struct device_attribute *attr, char *buf)
883 return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev)));
886 ssize_t part_stat_show(struct device *dev,
887 struct device_attribute *attr, char *buf)
889 struct block_device *bdev = dev_to_bdev(dev);
890 struct request_queue *q = bdev->bd_disk->queue;
891 struct disk_stats stat;
892 unsigned int inflight;
894 part_stat_read_all(bdev, &stat);
896 inflight = blk_mq_in_flight(q, bdev);
898 inflight = part_in_flight(bdev);
901 "%8lu %8lu %8llu %8u "
902 "%8lu %8lu %8llu %8u "
904 "%8lu %8lu %8llu %8u "
908 stat.merges[STAT_READ],
909 (unsigned long long)stat.sectors[STAT_READ],
910 (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC),
911 stat.ios[STAT_WRITE],
912 stat.merges[STAT_WRITE],
913 (unsigned long long)stat.sectors[STAT_WRITE],
914 (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC),
916 jiffies_to_msecs(stat.io_ticks),
917 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
918 stat.nsecs[STAT_WRITE] +
919 stat.nsecs[STAT_DISCARD] +
920 stat.nsecs[STAT_FLUSH],
922 stat.ios[STAT_DISCARD],
923 stat.merges[STAT_DISCARD],
924 (unsigned long long)stat.sectors[STAT_DISCARD],
925 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC),
926 stat.ios[STAT_FLUSH],
927 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC));
930 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
933 struct block_device *bdev = dev_to_bdev(dev);
934 struct request_queue *q = bdev->bd_disk->queue;
935 unsigned int inflight[2];
938 blk_mq_in_flight_rw(q, bdev, inflight);
940 part_in_flight_rw(bdev, inflight);
942 return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
945 static ssize_t disk_capability_show(struct device *dev,
946 struct device_attribute *attr, char *buf)
948 struct gendisk *disk = dev_to_disk(dev);
950 return sprintf(buf, "%x\n", disk->flags);
953 static ssize_t disk_alignment_offset_show(struct device *dev,
954 struct device_attribute *attr,
957 struct gendisk *disk = dev_to_disk(dev);
959 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
962 static ssize_t disk_discard_alignment_show(struct device *dev,
963 struct device_attribute *attr,
966 struct gendisk *disk = dev_to_disk(dev);
968 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
971 static ssize_t diskseq_show(struct device *dev,
972 struct device_attribute *attr, char *buf)
974 struct gendisk *disk = dev_to_disk(dev);
976 return sprintf(buf, "%llu\n", disk->diskseq);
979 static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
980 static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
981 static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
982 static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
983 static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
984 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
985 static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
986 static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
987 static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
988 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
989 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
990 static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
991 static DEVICE_ATTR(diskseq, 0444, diskseq_show, NULL);
993 #ifdef CONFIG_FAIL_MAKE_REQUEST
994 ssize_t part_fail_show(struct device *dev,
995 struct device_attribute *attr, char *buf)
997 return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail);
1000 ssize_t part_fail_store(struct device *dev,
1001 struct device_attribute *attr,
1002 const char *buf, size_t count)
1006 if (count > 0 && sscanf(buf, "%d", &i) > 0)
1007 dev_to_bdev(dev)->bd_make_it_fail = i;
1012 static struct device_attribute dev_attr_fail =
1013 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
1014 #endif /* CONFIG_FAIL_MAKE_REQUEST */
1016 #ifdef CONFIG_FAIL_IO_TIMEOUT
1017 static struct device_attribute dev_attr_fail_timeout =
1018 __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
1021 static struct attribute *disk_attrs[] = {
1022 &dev_attr_range.attr,
1023 &dev_attr_ext_range.attr,
1024 &dev_attr_removable.attr,
1025 &dev_attr_hidden.attr,
1027 &dev_attr_size.attr,
1028 &dev_attr_alignment_offset.attr,
1029 &dev_attr_discard_alignment.attr,
1030 &dev_attr_capability.attr,
1031 &dev_attr_stat.attr,
1032 &dev_attr_inflight.attr,
1033 &dev_attr_badblocks.attr,
1034 &dev_attr_events.attr,
1035 &dev_attr_events_async.attr,
1036 &dev_attr_events_poll_msecs.attr,
1037 &dev_attr_diskseq.attr,
1038 #ifdef CONFIG_FAIL_MAKE_REQUEST
1039 &dev_attr_fail.attr,
1041 #ifdef CONFIG_FAIL_IO_TIMEOUT
1042 &dev_attr_fail_timeout.attr,
1047 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1049 struct device *dev = container_of(kobj, typeof(*dev), kobj);
1050 struct gendisk *disk = dev_to_disk(dev);
1052 if (a == &dev_attr_badblocks.attr && !disk->bb)
1057 static struct attribute_group disk_attr_group = {
1058 .attrs = disk_attrs,
1059 .is_visible = disk_visible,
1062 static const struct attribute_group *disk_attr_groups[] = {
1068 * disk_release - releases all allocated resources of the gendisk
1069 * @dev: the device representing this disk
1071 * This function releases all allocated resources of the gendisk.
1073 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1074 * assigned. Since the request_queue sits on top of the gendisk for these
1075 * drivers we also call blk_put_queue() for them, and we expect the
1076 * request_queue refcount to reach 0 at this point, and so the request_queue
1077 * will also be freed prior to the disk.
1079 * Context: can sleep
1081 static void disk_release(struct device *dev)
1083 struct gendisk *disk = dev_to_disk(dev);
1086 WARN_ON_ONCE(disk_live(disk));
1088 blk_mq_cancel_work_sync(disk->queue);
1090 disk_release_events(disk);
1091 kfree(disk->random);
1092 xa_destroy(&disk->part_tbl);
1093 disk->queue->disk = NULL;
1094 blk_put_queue(disk->queue);
1095 iput(disk->part0->bd_inode); /* frees the disk */
1098 static int block_uevent(struct device *dev, struct kobj_uevent_env *env)
1100 struct gendisk *disk = dev_to_disk(dev);
1102 return add_uevent_var(env, "DISKSEQ=%llu", disk->diskseq);
1105 struct class block_class = {
1107 .dev_uevent = block_uevent,
1110 static char *block_devnode(struct device *dev, umode_t *mode,
1111 kuid_t *uid, kgid_t *gid)
1113 struct gendisk *disk = dev_to_disk(dev);
1115 if (disk->fops->devnode)
1116 return disk->fops->devnode(disk, mode);
1120 const struct device_type disk_type = {
1122 .groups = disk_attr_groups,
1123 .release = disk_release,
1124 .devnode = block_devnode,
1127 #ifdef CONFIG_PROC_FS
1129 * aggregate disk stat collector. Uses the same stats that the sysfs
1130 * entries do, above, but makes them available through one seq_file.
1132 * The output looks suspiciously like /proc/partitions with a bunch of
1135 static int diskstats_show(struct seq_file *seqf, void *v)
1137 struct gendisk *gp = v;
1138 struct block_device *hd;
1139 unsigned int inflight;
1140 struct disk_stats stat;
1144 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1145 seq_puts(seqf, "major minor name"
1146 " rio rmerge rsect ruse wio wmerge "
1147 "wsect wuse running use aveq"
1152 xa_for_each(&gp->part_tbl, idx, hd) {
1153 if (bdev_is_partition(hd) && !bdev_nr_sectors(hd))
1155 part_stat_read_all(hd, &stat);
1156 if (queue_is_mq(gp->queue))
1157 inflight = blk_mq_in_flight(gp->queue, hd);
1159 inflight = part_in_flight(hd);
1161 seq_printf(seqf, "%4d %7d %pg "
1168 MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd,
1169 stat.ios[STAT_READ],
1170 stat.merges[STAT_READ],
1171 stat.sectors[STAT_READ],
1172 (unsigned int)div_u64(stat.nsecs[STAT_READ],
1174 stat.ios[STAT_WRITE],
1175 stat.merges[STAT_WRITE],
1176 stat.sectors[STAT_WRITE],
1177 (unsigned int)div_u64(stat.nsecs[STAT_WRITE],
1180 jiffies_to_msecs(stat.io_ticks),
1181 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1182 stat.nsecs[STAT_WRITE] +
1183 stat.nsecs[STAT_DISCARD] +
1184 stat.nsecs[STAT_FLUSH],
1186 stat.ios[STAT_DISCARD],
1187 stat.merges[STAT_DISCARD],
1188 stat.sectors[STAT_DISCARD],
1189 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD],
1191 stat.ios[STAT_FLUSH],
1192 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH],
1201 static const struct seq_operations diskstats_op = {
1202 .start = disk_seqf_start,
1203 .next = disk_seqf_next,
1204 .stop = disk_seqf_stop,
1205 .show = diskstats_show
1208 static int __init proc_genhd_init(void)
1210 proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1211 proc_create_seq("partitions", 0, NULL, &partitions_op);
1214 module_init(proc_genhd_init);
1215 #endif /* CONFIG_PROC_FS */
1217 dev_t part_devt(struct gendisk *disk, u8 partno)
1219 struct block_device *part;
1223 part = xa_load(&disk->part_tbl, partno);
1225 devt = part->bd_dev;
1231 dev_t blk_lookup_devt(const char *name, int partno)
1233 dev_t devt = MKDEV(0, 0);
1234 struct class_dev_iter iter;
1237 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1238 while ((dev = class_dev_iter_next(&iter))) {
1239 struct gendisk *disk = dev_to_disk(dev);
1241 if (strcmp(dev_name(dev), name))
1244 if (partno < disk->minors) {
1245 /* We need to return the right devno, even
1246 * if the partition doesn't exist yet.
1248 devt = MKDEV(MAJOR(dev->devt),
1249 MINOR(dev->devt) + partno);
1251 devt = part_devt(disk, partno);
1256 class_dev_iter_exit(&iter);
1260 struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id,
1261 struct lock_class_key *lkclass)
1263 struct gendisk *disk;
1265 if (!blk_get_queue(q))
1268 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1272 disk->bdi = bdi_alloc(node_id);
1276 disk->part0 = bdev_alloc(disk, 0);
1280 disk->node_id = node_id;
1281 mutex_init(&disk->open_mutex);
1282 xa_init(&disk->part_tbl);
1283 if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL))
1284 goto out_destroy_part_tbl;
1286 rand_initialize_disk(disk);
1287 disk_to_dev(disk)->class = &block_class;
1288 disk_to_dev(disk)->type = &disk_type;
1289 device_initialize(disk_to_dev(disk));
1293 lockdep_init_map(&disk->lockdep_map, "(bio completion)", lkclass, 0);
1294 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
1295 INIT_LIST_HEAD(&disk->slave_bdevs);
1299 out_destroy_part_tbl:
1300 xa_destroy(&disk->part_tbl);
1301 disk->part0->bd_disk = NULL;
1302 iput(disk->part0->bd_inode);
1311 EXPORT_SYMBOL(__alloc_disk_node);
1313 struct gendisk *__blk_alloc_disk(int node, struct lock_class_key *lkclass)
1315 struct request_queue *q;
1316 struct gendisk *disk;
1318 q = blk_alloc_queue(node);
1322 disk = __alloc_disk_node(q, node, lkclass);
1324 blk_cleanup_queue(q);
1329 EXPORT_SYMBOL(__blk_alloc_disk);
1332 * put_disk - decrements the gendisk refcount
1333 * @disk: the struct gendisk to decrement the refcount for
1335 * This decrements the refcount for the struct gendisk. When this reaches 0
1336 * we'll have disk_release() called.
1338 * Context: Any context, but the last reference must not be dropped from
1341 void put_disk(struct gendisk *disk)
1344 put_device(disk_to_dev(disk));
1346 EXPORT_SYMBOL(put_disk);
1349 * blk_cleanup_disk - shutdown a gendisk allocated by blk_alloc_disk
1350 * @disk: gendisk to shutdown
1352 * Mark the queue hanging off @disk DYING, drain all pending requests, then mark
1353 * the queue DEAD, destroy and put it and the gendisk structure.
1355 * Context: can sleep
1357 void blk_cleanup_disk(struct gendisk *disk)
1359 blk_cleanup_queue(disk->queue);
1362 EXPORT_SYMBOL(blk_cleanup_disk);
1364 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1366 char event[] = "DISK_RO=1";
1367 char *envp[] = { event, NULL };
1371 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1375 * set_disk_ro - set a gendisk read-only
1376 * @disk: gendisk to operate on
1377 * @read_only: %true to set the disk read-only, %false set the disk read/write
1379 * This function is used to indicate whether a given disk device should have its
1380 * read-only flag set. set_disk_ro() is typically used by device drivers to
1381 * indicate whether the underlying physical device is write-protected.
1383 void set_disk_ro(struct gendisk *disk, bool read_only)
1386 if (test_and_set_bit(GD_READ_ONLY, &disk->state))
1389 if (!test_and_clear_bit(GD_READ_ONLY, &disk->state))
1392 set_disk_ro_uevent(disk, read_only);
1394 EXPORT_SYMBOL(set_disk_ro);
1396 int bdev_read_only(struct block_device *bdev)
1398 return bdev->bd_read_only || get_disk_ro(bdev->bd_disk);
1400 EXPORT_SYMBOL(bdev_read_only);
1402 void inc_diskseq(struct gendisk *disk)
1404 disk->diskseq = atomic64_inc_return(&diskseq);