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/major.h>
23 #include <linux/mutex.h>
24 #include <linux/idr.h>
25 #include <linux/log2.h>
26 #include <linux/pm_runtime.h>
27 #include <linux/badblocks.h>
30 #include "blk-rq-qos.h"
32 static struct kobject *block_depr;
35 * Unique, monotonically increasing sequential number associated with block
36 * devices instances (i.e. incremented each time a device is attached).
37 * Associating uevents with block devices in userspace is difficult and racy:
38 * the uevent netlink socket is lossy, and on slow and overloaded systems has
39 * a very high latency.
40 * Block devices do not have exclusive owners in userspace, any process can set
41 * one up (e.g. loop devices). Moreover, device names can be reused (e.g. loop0
42 * can be reused again and again).
43 * A userspace process setting up a block device and watching for its events
44 * cannot thus reliably tell whether an event relates to the device it just set
45 * up or another earlier instance with the same name.
46 * This sequential number allows userspace processes to solve this problem, and
47 * uniquely associate an uevent to the lifetime to a device.
49 static atomic64_t diskseq;
51 /* for extended dynamic devt allocation, currently only one major is used */
52 #define NR_EXT_DEVT (1 << MINORBITS)
53 static DEFINE_IDA(ext_devt_ida);
55 void set_capacity(struct gendisk *disk, sector_t sectors)
57 struct block_device *bdev = disk->part0;
59 spin_lock(&bdev->bd_size_lock);
60 i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
61 bdev->bd_nr_sectors = sectors;
62 spin_unlock(&bdev->bd_size_lock);
64 EXPORT_SYMBOL(set_capacity);
67 * Set disk capacity and notify if the size is not currently zero and will not
68 * be set to zero. Returns true if a uevent was sent, otherwise false.
70 bool set_capacity_and_notify(struct gendisk *disk, sector_t size)
72 sector_t capacity = get_capacity(disk);
73 char *envp[] = { "RESIZE=1", NULL };
75 set_capacity(disk, size);
78 * Only print a message and send a uevent if the gendisk is user visible
79 * and alive. This avoids spamming the log and udev when setting the
80 * initial capacity during probing.
82 if (size == capacity ||
84 (disk->flags & GENHD_FL_HIDDEN))
87 pr_info("%s: detected capacity change from %lld to %lld\n",
88 disk->disk_name, capacity, size);
91 * Historically we did not send a uevent for changes to/from an empty
94 if (!capacity || !size)
96 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
99 EXPORT_SYMBOL_GPL(set_capacity_and_notify);
102 * Format the device name of the indicated block device into the supplied buffer
103 * and return a pointer to that same buffer for convenience.
105 * Note: do not use this in new code, use the %pg specifier to sprintf and
108 const char *bdevname(struct block_device *bdev, char *buf)
110 struct gendisk *hd = bdev->bd_disk;
111 int partno = bdev->bd_partno;
114 snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name);
115 else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1]))
116 snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno);
118 snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno);
122 EXPORT_SYMBOL(bdevname);
124 static void part_stat_read_all(struct block_device *part,
125 struct disk_stats *stat)
129 memset(stat, 0, sizeof(struct disk_stats));
130 for_each_possible_cpu(cpu) {
131 struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu);
134 for (group = 0; group < NR_STAT_GROUPS; group++) {
135 stat->nsecs[group] += ptr->nsecs[group];
136 stat->sectors[group] += ptr->sectors[group];
137 stat->ios[group] += ptr->ios[group];
138 stat->merges[group] += ptr->merges[group];
141 stat->io_ticks += ptr->io_ticks;
145 static unsigned int part_in_flight(struct block_device *part)
147 unsigned int inflight = 0;
150 for_each_possible_cpu(cpu) {
151 inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) +
152 part_stat_local_read_cpu(part, in_flight[1], cpu);
154 if ((int)inflight < 0)
160 static void part_in_flight_rw(struct block_device *part,
161 unsigned int inflight[2])
167 for_each_possible_cpu(cpu) {
168 inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu);
169 inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu);
171 if ((int)inflight[0] < 0)
173 if ((int)inflight[1] < 0)
178 * Can be deleted altogether. Later.
181 #define BLKDEV_MAJOR_HASH_SIZE 255
182 static struct blk_major_name {
183 struct blk_major_name *next;
186 void (*probe)(dev_t devt);
187 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
188 static DEFINE_MUTEX(major_names_lock);
189 static DEFINE_SPINLOCK(major_names_spinlock);
191 /* index in the above - for now: assume no multimajor ranges */
192 static inline int major_to_index(unsigned major)
194 return major % BLKDEV_MAJOR_HASH_SIZE;
197 #ifdef CONFIG_PROC_FS
198 void blkdev_show(struct seq_file *seqf, off_t offset)
200 struct blk_major_name *dp;
202 spin_lock(&major_names_spinlock);
203 for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
204 if (dp->major == offset)
205 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
206 spin_unlock(&major_names_spinlock);
208 #endif /* CONFIG_PROC_FS */
211 * __register_blkdev - register a new block device
213 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
214 * @major = 0, try to allocate any unused major number.
215 * @name: the name of the new block device as a zero terminated string
216 * @probe: allback that is called on access to any minor number of @major
218 * The @name must be unique within the system.
220 * The return value depends on the @major input parameter:
222 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
223 * then the function returns zero on success, or a negative error code
224 * - if any unused major number was requested with @major = 0 parameter
225 * then the return value is the allocated major number in range
226 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
228 * See Documentation/admin-guide/devices.txt for the list of allocated
231 * Use register_blkdev instead for any new code.
233 int __register_blkdev(unsigned int major, const char *name,
234 void (*probe)(dev_t devt))
236 struct blk_major_name **n, *p;
239 mutex_lock(&major_names_lock);
243 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
244 if (major_names[index] == NULL)
249 printk("%s: failed to get major for %s\n",
258 if (major >= BLKDEV_MAJOR_MAX) {
259 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
260 __func__, major, BLKDEV_MAJOR_MAX-1, name);
266 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
274 strlcpy(p->name, name, sizeof(p->name));
276 index = major_to_index(major);
278 spin_lock(&major_names_spinlock);
279 for (n = &major_names[index]; *n; n = &(*n)->next) {
280 if ((*n)->major == major)
287 spin_unlock(&major_names_spinlock);
290 printk("register_blkdev: cannot get major %u for %s\n",
295 mutex_unlock(&major_names_lock);
298 EXPORT_SYMBOL(__register_blkdev);
300 void unregister_blkdev(unsigned int major, const char *name)
302 struct blk_major_name **n;
303 struct blk_major_name *p = NULL;
304 int index = major_to_index(major);
306 mutex_lock(&major_names_lock);
307 spin_lock(&major_names_spinlock);
308 for (n = &major_names[index]; *n; n = &(*n)->next)
309 if ((*n)->major == major)
311 if (!*n || strcmp((*n)->name, name)) {
317 spin_unlock(&major_names_spinlock);
318 mutex_unlock(&major_names_lock);
322 EXPORT_SYMBOL(unregister_blkdev);
324 int blk_alloc_ext_minor(void)
328 idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT, GFP_KERNEL);
334 void blk_free_ext_minor(unsigned int minor)
336 ida_free(&ext_devt_ida, minor);
339 static char *bdevt_str(dev_t devt, char *buf)
341 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
342 char tbuf[BDEVT_SIZE];
343 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
344 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
346 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
351 void disk_uevent(struct gendisk *disk, enum kobject_action action)
353 struct block_device *part;
357 xa_for_each(&disk->part_tbl, idx, part) {
358 if (bdev_is_partition(part) && !bdev_nr_sectors(part))
360 if (!kobject_get_unless_zero(&part->bd_device.kobj))
364 kobject_uevent(bdev_kobj(part), action);
365 put_device(&part->bd_device);
370 EXPORT_SYMBOL_GPL(disk_uevent);
372 static void disk_scan_partitions(struct gendisk *disk)
374 struct block_device *bdev;
376 if (!get_capacity(disk) || !disk_part_scan_enabled(disk))
379 set_bit(GD_NEED_PART_SCAN, &disk->state);
380 bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL);
382 blkdev_put(bdev, FMODE_READ);
386 * device_add_disk - add disk information to kernel list
387 * @parent: parent device for the disk
388 * @disk: per-device partitioning information
389 * @groups: Additional per-device sysfs groups
391 * This function registers the partitioning information in @disk
394 int device_add_disk(struct device *parent, struct gendisk *disk,
395 const struct attribute_group **groups)
398 struct device *ddev = disk_to_dev(disk);
402 * The disk queue should now be all set with enough information about
403 * the device for the elevator code to pick an adequate default
404 * elevator if one is needed, that is, for devices requesting queue
407 elevator_init_mq(disk->queue);
410 * If the driver provides an explicit major number it also must provide
411 * the number of minors numbers supported, and those will be used to
413 * Otherwise just allocate the device numbers for both the whole device
414 * and all partitions from the extended dev_t space.
417 if (WARN_ON(!disk->minors))
420 if (disk->minors > DISK_MAX_PARTS) {
421 pr_err("block: can't allocate more than %d partitions\n",
423 disk->minors = DISK_MAX_PARTS;
426 if (WARN_ON(disk->minors))
429 ret = blk_alloc_ext_minor();
432 disk->major = BLOCK_EXT_MAJOR;
433 disk->first_minor = ret;
434 disk->flags |= GENHD_FL_EXT_DEVT;
437 ret = disk_alloc_events(disk);
439 goto out_free_ext_minor;
441 /* delay uevents, until we scanned partition table */
442 dev_set_uevent_suppress(ddev, 1);
444 ddev->parent = parent;
445 ddev->groups = groups;
446 dev_set_name(ddev, "%s", disk->disk_name);
447 if (!(disk->flags & GENHD_FL_HIDDEN))
448 ddev->devt = MKDEV(disk->major, disk->first_minor);
449 ret = device_add(ddev);
451 goto out_disk_release_events;
452 if (!sysfs_deprecated) {
453 ret = sysfs_create_link(block_depr, &ddev->kobj,
454 kobject_name(&ddev->kobj));
460 * avoid probable deadlock caused by allocating memory with
461 * GFP_KERNEL in runtime_resume callback of its all ancestor
464 pm_runtime_set_memalloc_noio(ddev, true);
466 ret = blk_integrity_add(disk);
468 goto out_del_block_link;
470 disk->part0->bd_holder_dir =
471 kobject_create_and_add("holders", &ddev->kobj);
472 if (!disk->part0->bd_holder_dir) {
474 goto out_del_integrity;
476 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
477 if (!disk->slave_dir) {
479 goto out_put_holder_dir;
482 ret = bd_register_pending_holders(disk);
484 goto out_put_slave_dir;
486 ret = blk_register_queue(disk);
488 goto out_put_slave_dir;
490 if (disk->flags & GENHD_FL_HIDDEN) {
492 * Don't let hidden disks show up in /proc/partitions,
493 * and don't bother scanning for partitions either.
495 disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
496 disk->flags |= GENHD_FL_NO_PART_SCAN;
498 ret = bdi_register(disk->bdi, "%u:%u",
499 disk->major, disk->first_minor);
501 goto out_unregister_queue;
502 bdi_set_owner(disk->bdi, ddev);
503 ret = sysfs_create_link(&ddev->kobj,
504 &disk->bdi->dev->kobj, "bdi");
506 goto out_unregister_bdi;
508 bdev_add(disk->part0, ddev->devt);
509 disk_scan_partitions(disk);
512 * Announce the disk and partitions after all partitions are
513 * created. (for hidden disks uevents remain suppressed forever)
515 dev_set_uevent_suppress(ddev, 0);
516 disk_uevent(disk, KOBJ_ADD);
519 disk_update_readahead(disk);
520 disk_add_events(disk);
524 if (!(disk->flags & GENHD_FL_HIDDEN))
525 bdi_unregister(disk->bdi);
526 out_unregister_queue:
527 blk_unregister_queue(disk);
529 kobject_put(disk->slave_dir);
531 kobject_put(disk->part0->bd_holder_dir);
533 blk_integrity_del(disk);
535 if (!sysfs_deprecated)
536 sysfs_remove_link(block_depr, dev_name(ddev));
539 out_disk_release_events:
540 disk_release_events(disk);
542 if (disk->major == BLOCK_EXT_MAJOR)
543 blk_free_ext_minor(disk->first_minor);
544 return WARN_ON_ONCE(ret); /* keep until all callers handle errors */
546 EXPORT_SYMBOL(device_add_disk);
549 * del_gendisk - remove the gendisk
550 * @disk: the struct gendisk to remove
552 * Removes the gendisk and all its associated resources. This deletes the
553 * partitions associated with the gendisk, and unregisters the associated
556 * This is the counter to the respective __device_add_disk() call.
558 * The final removal of the struct gendisk happens when its refcount reaches 0
559 * with put_disk(), which should be called after del_gendisk(), if
560 * __device_add_disk() was used.
562 * Drivers exist which depend on the release of the gendisk to be synchronous,
563 * it should not be deferred.
567 void del_gendisk(struct gendisk *disk)
569 struct request_queue *q = disk->queue;
573 if (WARN_ON_ONCE(!disk_live(disk) && !(disk->flags & GENHD_FL_HIDDEN)))
576 blk_integrity_del(disk);
577 disk_del_events(disk);
579 mutex_lock(&disk->open_mutex);
580 remove_inode_hash(disk->part0->bd_inode);
581 blk_drop_partitions(disk);
582 mutex_unlock(&disk->open_mutex);
584 fsync_bdev(disk->part0);
585 __invalidate_device(disk->part0, true);
590 set_bit(GD_DEAD, &disk->state);
591 set_capacity(disk, 0);
594 * Prevent new I/O from crossing bio_queue_enter().
596 blk_queue_start_drain(q);
598 if (!(disk->flags & GENHD_FL_HIDDEN)) {
599 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
602 * Unregister bdi before releasing device numbers (as they can
603 * get reused and we'd get clashes in sysfs).
605 bdi_unregister(disk->bdi);
608 blk_unregister_queue(disk);
610 kobject_put(disk->part0->bd_holder_dir);
611 kobject_put(disk->slave_dir);
613 part_stat_set_all(disk->part0, 0);
614 disk->part0->bd_stamp = 0;
615 if (!sysfs_deprecated)
616 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
617 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
618 device_del(disk_to_dev(disk));
620 blk_mq_freeze_queue_wait(q);
624 blk_flush_integrity();
626 * Allow using passthrough request again after the queue is torn down.
628 blk_queue_flag_clear(QUEUE_FLAG_INIT_DONE, q);
629 __blk_mq_unfreeze_queue(q, true);
632 EXPORT_SYMBOL(del_gendisk);
635 * invalidate_disk - invalidate the disk
636 * @disk: the struct gendisk to invalidate
638 * A helper to invalidates the disk. It will clean the disk's associated
639 * buffer/page caches and reset its internal states so that the disk
640 * can be reused by the drivers.
644 void invalidate_disk(struct gendisk *disk)
646 struct block_device *bdev = disk->part0;
648 invalidate_bdev(bdev);
649 bdev->bd_inode->i_mapping->wb_err = 0;
650 set_capacity(disk, 0);
652 EXPORT_SYMBOL(invalidate_disk);
654 /* sysfs access to bad-blocks list. */
655 static ssize_t disk_badblocks_show(struct device *dev,
656 struct device_attribute *attr,
659 struct gendisk *disk = dev_to_disk(dev);
662 return sprintf(page, "\n");
664 return badblocks_show(disk->bb, page, 0);
667 static ssize_t disk_badblocks_store(struct device *dev,
668 struct device_attribute *attr,
669 const char *page, size_t len)
671 struct gendisk *disk = dev_to_disk(dev);
676 return badblocks_store(disk->bb, page, len, 0);
679 void blk_request_module(dev_t devt)
681 unsigned int major = MAJOR(devt);
682 struct blk_major_name **n;
684 mutex_lock(&major_names_lock);
685 for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) {
686 if ((*n)->major == major && (*n)->probe) {
688 mutex_unlock(&major_names_lock);
692 mutex_unlock(&major_names_lock);
694 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
695 /* Make old-style 2.4 aliases work */
696 request_module("block-major-%d", MAJOR(devt));
700 * print a full list of all partitions - intended for places where the root
701 * filesystem can't be mounted and thus to give the victim some idea of what
704 void __init printk_all_partitions(void)
706 struct class_dev_iter iter;
709 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
710 while ((dev = class_dev_iter_next(&iter))) {
711 struct gendisk *disk = dev_to_disk(dev);
712 struct block_device *part;
713 char devt_buf[BDEVT_SIZE];
717 * Don't show empty devices or things that have been
720 if (get_capacity(disk) == 0 ||
721 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
725 * Note, unlike /proc/partitions, I am showing the numbers in
726 * hex - the same format as the root= option takes.
729 xa_for_each(&disk->part_tbl, idx, part) {
730 if (!bdev_nr_sectors(part))
732 printk("%s%s %10llu %pg %s",
733 bdev_is_partition(part) ? " " : "",
734 bdevt_str(part->bd_dev, devt_buf),
735 bdev_nr_sectors(part) >> 1, part,
737 part->bd_meta_info->uuid : "");
738 if (bdev_is_partition(part))
740 else if (dev->parent && dev->parent->driver)
741 printk(" driver: %s\n",
742 dev->parent->driver->name);
744 printk(" (driver?)\n");
748 class_dev_iter_exit(&iter);
751 #ifdef CONFIG_PROC_FS
753 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
756 struct class_dev_iter *iter;
759 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
761 return ERR_PTR(-ENOMEM);
763 seqf->private = iter;
764 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
766 dev = class_dev_iter_next(iter);
771 return dev_to_disk(dev);
774 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
779 dev = class_dev_iter_next(seqf->private);
781 return dev_to_disk(dev);
786 static void disk_seqf_stop(struct seq_file *seqf, void *v)
788 struct class_dev_iter *iter = seqf->private;
790 /* stop is called even after start failed :-( */
792 class_dev_iter_exit(iter);
794 seqf->private = NULL;
798 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
802 p = disk_seqf_start(seqf, pos);
803 if (!IS_ERR_OR_NULL(p) && !*pos)
804 seq_puts(seqf, "major minor #blocks name\n\n");
808 static int show_partition(struct seq_file *seqf, void *v)
810 struct gendisk *sgp = v;
811 struct block_device *part;
814 /* Don't show non-partitionable removeable devices or empty devices */
815 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
816 (sgp->flags & GENHD_FL_REMOVABLE)))
818 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
822 xa_for_each(&sgp->part_tbl, idx, part) {
823 if (!bdev_nr_sectors(part))
825 seq_printf(seqf, "%4d %7d %10llu %pg\n",
826 MAJOR(part->bd_dev), MINOR(part->bd_dev),
827 bdev_nr_sectors(part) >> 1, part);
833 static const struct seq_operations partitions_op = {
834 .start = show_partition_start,
835 .next = disk_seqf_next,
836 .stop = disk_seqf_stop,
837 .show = show_partition
841 static int __init genhd_device_init(void)
845 block_class.dev_kobj = sysfs_dev_block_kobj;
846 error = class_register(&block_class);
851 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
853 /* create top-level block dir */
854 if (!sysfs_deprecated)
855 block_depr = kobject_create_and_add("block", NULL);
859 subsys_initcall(genhd_device_init);
861 static ssize_t disk_range_show(struct device *dev,
862 struct device_attribute *attr, char *buf)
864 struct gendisk *disk = dev_to_disk(dev);
866 return sprintf(buf, "%d\n", disk->minors);
869 static ssize_t disk_ext_range_show(struct device *dev,
870 struct device_attribute *attr, char *buf)
872 struct gendisk *disk = dev_to_disk(dev);
874 return sprintf(buf, "%d\n", disk_max_parts(disk));
877 static ssize_t disk_removable_show(struct device *dev,
878 struct device_attribute *attr, char *buf)
880 struct gendisk *disk = dev_to_disk(dev);
882 return sprintf(buf, "%d\n",
883 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
886 static ssize_t disk_hidden_show(struct device *dev,
887 struct device_attribute *attr, char *buf)
889 struct gendisk *disk = dev_to_disk(dev);
891 return sprintf(buf, "%d\n",
892 (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
895 static ssize_t disk_ro_show(struct device *dev,
896 struct device_attribute *attr, char *buf)
898 struct gendisk *disk = dev_to_disk(dev);
900 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
903 ssize_t part_size_show(struct device *dev,
904 struct device_attribute *attr, char *buf)
906 return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev)));
909 ssize_t part_stat_show(struct device *dev,
910 struct device_attribute *attr, char *buf)
912 struct block_device *bdev = dev_to_bdev(dev);
913 struct request_queue *q = bdev_get_queue(bdev);
914 struct disk_stats stat;
915 unsigned int inflight;
917 part_stat_read_all(bdev, &stat);
919 inflight = blk_mq_in_flight(q, bdev);
921 inflight = part_in_flight(bdev);
924 "%8lu %8lu %8llu %8u "
925 "%8lu %8lu %8llu %8u "
927 "%8lu %8lu %8llu %8u "
931 stat.merges[STAT_READ],
932 (unsigned long long)stat.sectors[STAT_READ],
933 (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC),
934 stat.ios[STAT_WRITE],
935 stat.merges[STAT_WRITE],
936 (unsigned long long)stat.sectors[STAT_WRITE],
937 (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC),
939 jiffies_to_msecs(stat.io_ticks),
940 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
941 stat.nsecs[STAT_WRITE] +
942 stat.nsecs[STAT_DISCARD] +
943 stat.nsecs[STAT_FLUSH],
945 stat.ios[STAT_DISCARD],
946 stat.merges[STAT_DISCARD],
947 (unsigned long long)stat.sectors[STAT_DISCARD],
948 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC),
949 stat.ios[STAT_FLUSH],
950 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC));
953 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
956 struct block_device *bdev = dev_to_bdev(dev);
957 struct request_queue *q = bdev_get_queue(bdev);
958 unsigned int inflight[2];
961 blk_mq_in_flight_rw(q, bdev, inflight);
963 part_in_flight_rw(bdev, inflight);
965 return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
968 static ssize_t disk_capability_show(struct device *dev,
969 struct device_attribute *attr, char *buf)
971 struct gendisk *disk = dev_to_disk(dev);
973 return sprintf(buf, "%x\n", disk->flags);
976 static ssize_t disk_alignment_offset_show(struct device *dev,
977 struct device_attribute *attr,
980 struct gendisk *disk = dev_to_disk(dev);
982 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
985 static ssize_t disk_discard_alignment_show(struct device *dev,
986 struct device_attribute *attr,
989 struct gendisk *disk = dev_to_disk(dev);
991 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
994 static ssize_t diskseq_show(struct device *dev,
995 struct device_attribute *attr, char *buf)
997 struct gendisk *disk = dev_to_disk(dev);
999 return sprintf(buf, "%llu\n", disk->diskseq);
1002 static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
1003 static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
1004 static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
1005 static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
1006 static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
1007 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
1008 static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
1009 static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
1010 static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
1011 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
1012 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
1013 static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
1014 static DEVICE_ATTR(diskseq, 0444, diskseq_show, NULL);
1016 #ifdef CONFIG_FAIL_MAKE_REQUEST
1017 ssize_t part_fail_show(struct device *dev,
1018 struct device_attribute *attr, char *buf)
1020 return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail);
1023 ssize_t part_fail_store(struct device *dev,
1024 struct device_attribute *attr,
1025 const char *buf, size_t count)
1029 if (count > 0 && sscanf(buf, "%d", &i) > 0)
1030 dev_to_bdev(dev)->bd_make_it_fail = i;
1035 static struct device_attribute dev_attr_fail =
1036 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
1037 #endif /* CONFIG_FAIL_MAKE_REQUEST */
1039 #ifdef CONFIG_FAIL_IO_TIMEOUT
1040 static struct device_attribute dev_attr_fail_timeout =
1041 __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
1044 static struct attribute *disk_attrs[] = {
1045 &dev_attr_range.attr,
1046 &dev_attr_ext_range.attr,
1047 &dev_attr_removable.attr,
1048 &dev_attr_hidden.attr,
1050 &dev_attr_size.attr,
1051 &dev_attr_alignment_offset.attr,
1052 &dev_attr_discard_alignment.attr,
1053 &dev_attr_capability.attr,
1054 &dev_attr_stat.attr,
1055 &dev_attr_inflight.attr,
1056 &dev_attr_badblocks.attr,
1057 &dev_attr_events.attr,
1058 &dev_attr_events_async.attr,
1059 &dev_attr_events_poll_msecs.attr,
1060 &dev_attr_diskseq.attr,
1061 #ifdef CONFIG_FAIL_MAKE_REQUEST
1062 &dev_attr_fail.attr,
1064 #ifdef CONFIG_FAIL_IO_TIMEOUT
1065 &dev_attr_fail_timeout.attr,
1070 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1072 struct device *dev = container_of(kobj, typeof(*dev), kobj);
1073 struct gendisk *disk = dev_to_disk(dev);
1075 if (a == &dev_attr_badblocks.attr && !disk->bb)
1080 static struct attribute_group disk_attr_group = {
1081 .attrs = disk_attrs,
1082 .is_visible = disk_visible,
1085 static const struct attribute_group *disk_attr_groups[] = {
1091 * disk_release - releases all allocated resources of the gendisk
1092 * @dev: the device representing this disk
1094 * This function releases all allocated resources of the gendisk.
1096 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1097 * assigned. Since the request_queue sits on top of the gendisk for these
1098 * drivers we also call blk_put_queue() for them, and we expect the
1099 * request_queue refcount to reach 0 at this point, and so the request_queue
1100 * will also be freed prior to the disk.
1102 * Context: can sleep
1104 static void disk_release(struct device *dev)
1106 struct gendisk *disk = dev_to_disk(dev);
1109 WARN_ON_ONCE(disk_live(disk));
1111 disk_release_events(disk);
1112 kfree(disk->random);
1113 xa_destroy(&disk->part_tbl);
1114 disk->queue->disk = NULL;
1115 blk_put_queue(disk->queue);
1116 iput(disk->part0->bd_inode); /* frees the disk */
1119 static int block_uevent(struct device *dev, struct kobj_uevent_env *env)
1121 struct gendisk *disk = dev_to_disk(dev);
1123 return add_uevent_var(env, "DISKSEQ=%llu", disk->diskseq);
1126 struct class block_class = {
1128 .dev_uevent = block_uevent,
1131 static char *block_devnode(struct device *dev, umode_t *mode,
1132 kuid_t *uid, kgid_t *gid)
1134 struct gendisk *disk = dev_to_disk(dev);
1136 if (disk->fops->devnode)
1137 return disk->fops->devnode(disk, mode);
1141 const struct device_type disk_type = {
1143 .groups = disk_attr_groups,
1144 .release = disk_release,
1145 .devnode = block_devnode,
1148 #ifdef CONFIG_PROC_FS
1150 * aggregate disk stat collector. Uses the same stats that the sysfs
1151 * entries do, above, but makes them available through one seq_file.
1153 * The output looks suspiciously like /proc/partitions with a bunch of
1156 static int diskstats_show(struct seq_file *seqf, void *v)
1158 struct gendisk *gp = v;
1159 struct block_device *hd;
1160 unsigned int inflight;
1161 struct disk_stats stat;
1165 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1166 seq_puts(seqf, "major minor name"
1167 " rio rmerge rsect ruse wio wmerge "
1168 "wsect wuse running use aveq"
1173 xa_for_each(&gp->part_tbl, idx, hd) {
1174 if (bdev_is_partition(hd) && !bdev_nr_sectors(hd))
1176 part_stat_read_all(hd, &stat);
1177 if (queue_is_mq(gp->queue))
1178 inflight = blk_mq_in_flight(gp->queue, hd);
1180 inflight = part_in_flight(hd);
1182 seq_printf(seqf, "%4d %7d %pg "
1189 MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd,
1190 stat.ios[STAT_READ],
1191 stat.merges[STAT_READ],
1192 stat.sectors[STAT_READ],
1193 (unsigned int)div_u64(stat.nsecs[STAT_READ],
1195 stat.ios[STAT_WRITE],
1196 stat.merges[STAT_WRITE],
1197 stat.sectors[STAT_WRITE],
1198 (unsigned int)div_u64(stat.nsecs[STAT_WRITE],
1201 jiffies_to_msecs(stat.io_ticks),
1202 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1203 stat.nsecs[STAT_WRITE] +
1204 stat.nsecs[STAT_DISCARD] +
1205 stat.nsecs[STAT_FLUSH],
1207 stat.ios[STAT_DISCARD],
1208 stat.merges[STAT_DISCARD],
1209 stat.sectors[STAT_DISCARD],
1210 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD],
1212 stat.ios[STAT_FLUSH],
1213 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH],
1222 static const struct seq_operations diskstats_op = {
1223 .start = disk_seqf_start,
1224 .next = disk_seqf_next,
1225 .stop = disk_seqf_stop,
1226 .show = diskstats_show
1229 static int __init proc_genhd_init(void)
1231 proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1232 proc_create_seq("partitions", 0, NULL, &partitions_op);
1235 module_init(proc_genhd_init);
1236 #endif /* CONFIG_PROC_FS */
1238 dev_t part_devt(struct gendisk *disk, u8 partno)
1240 struct block_device *part;
1244 part = xa_load(&disk->part_tbl, partno);
1246 devt = part->bd_dev;
1252 dev_t blk_lookup_devt(const char *name, int partno)
1254 dev_t devt = MKDEV(0, 0);
1255 struct class_dev_iter iter;
1258 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1259 while ((dev = class_dev_iter_next(&iter))) {
1260 struct gendisk *disk = dev_to_disk(dev);
1262 if (strcmp(dev_name(dev), name))
1265 if (partno < disk->minors) {
1266 /* We need to return the right devno, even
1267 * if the partition doesn't exist yet.
1269 devt = MKDEV(MAJOR(dev->devt),
1270 MINOR(dev->devt) + partno);
1272 devt = part_devt(disk, partno);
1277 class_dev_iter_exit(&iter);
1281 struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id,
1282 struct lock_class_key *lkclass)
1284 struct gendisk *disk;
1286 if (!blk_get_queue(q))
1289 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1293 disk->bdi = bdi_alloc(node_id);
1297 /* bdev_alloc() might need the queue, set before the first call */
1300 disk->part0 = bdev_alloc(disk, 0);
1304 disk->node_id = node_id;
1305 mutex_init(&disk->open_mutex);
1306 xa_init(&disk->part_tbl);
1307 if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL))
1308 goto out_destroy_part_tbl;
1310 rand_initialize_disk(disk);
1311 disk_to_dev(disk)->class = &block_class;
1312 disk_to_dev(disk)->type = &disk_type;
1313 device_initialize(disk_to_dev(disk));
1316 lockdep_init_map(&disk->lockdep_map, "(bio completion)", lkclass, 0);
1317 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
1318 INIT_LIST_HEAD(&disk->slave_bdevs);
1322 out_destroy_part_tbl:
1323 xa_destroy(&disk->part_tbl);
1324 disk->part0->bd_disk = NULL;
1325 iput(disk->part0->bd_inode);
1334 EXPORT_SYMBOL(__alloc_disk_node);
1336 struct gendisk *__blk_alloc_disk(int node, struct lock_class_key *lkclass)
1338 struct request_queue *q;
1339 struct gendisk *disk;
1341 q = blk_alloc_queue(node);
1345 disk = __alloc_disk_node(q, node, lkclass);
1347 blk_cleanup_queue(q);
1352 EXPORT_SYMBOL(__blk_alloc_disk);
1355 * put_disk - decrements the gendisk refcount
1356 * @disk: the struct gendisk to decrement the refcount for
1358 * This decrements the refcount for the struct gendisk. When this reaches 0
1359 * we'll have disk_release() called.
1361 * Context: Any context, but the last reference must not be dropped from
1364 void put_disk(struct gendisk *disk)
1367 put_device(disk_to_dev(disk));
1369 EXPORT_SYMBOL(put_disk);
1372 * blk_cleanup_disk - shutdown a gendisk allocated by blk_alloc_disk
1373 * @disk: gendisk to shutdown
1375 * Mark the queue hanging off @disk DYING, drain all pending requests, then mark
1376 * the queue DEAD, destroy and put it and the gendisk structure.
1378 * Context: can sleep
1380 void blk_cleanup_disk(struct gendisk *disk)
1382 blk_cleanup_queue(disk->queue);
1385 EXPORT_SYMBOL(blk_cleanup_disk);
1387 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1389 char event[] = "DISK_RO=1";
1390 char *envp[] = { event, NULL };
1394 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1398 * set_disk_ro - set a gendisk read-only
1399 * @disk: gendisk to operate on
1400 * @read_only: %true to set the disk read-only, %false set the disk read/write
1402 * This function is used to indicate whether a given disk device should have its
1403 * read-only flag set. set_disk_ro() is typically used by device drivers to
1404 * indicate whether the underlying physical device is write-protected.
1406 void set_disk_ro(struct gendisk *disk, bool read_only)
1409 if (test_and_set_bit(GD_READ_ONLY, &disk->state))
1412 if (!test_and_clear_bit(GD_READ_ONLY, &disk->state))
1415 set_disk_ro_uevent(disk, read_only);
1417 EXPORT_SYMBOL(set_disk_ro);
1419 void inc_diskseq(struct gendisk *disk)
1421 disk->diskseq = atomic64_inc_return(&diskseq);