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;
33 * Unique, monotonically increasing sequential number associated with block
34 * devices instances (i.e. incremented each time a device is attached).
35 * Associating uevents with block devices in userspace is difficult and racy:
36 * the uevent netlink socket is lossy, and on slow and overloaded systems has
37 * a very high latency.
38 * Block devices do not have exclusive owners in userspace, any process can set
39 * one up (e.g. loop devices). Moreover, device names can be reused (e.g. loop0
40 * can be reused again and again).
41 * A userspace process setting up a block device and watching for its events
42 * cannot thus reliably tell whether an event relates to the device it just set
43 * up or another earlier instance with the same name.
44 * This sequential number allows userspace processes to solve this problem, and
45 * uniquely associate an uevent to the lifetime to a device.
47 static atomic64_t diskseq;
49 /* for extended dynamic devt allocation, currently only one major is used */
50 #define NR_EXT_DEVT (1 << MINORBITS)
51 static DEFINE_IDA(ext_devt_ida);
53 void set_capacity(struct gendisk *disk, sector_t sectors)
55 struct block_device *bdev = disk->part0;
57 spin_lock(&bdev->bd_size_lock);
58 i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
59 spin_unlock(&bdev->bd_size_lock);
61 EXPORT_SYMBOL(set_capacity);
64 * Set disk capacity and notify if the size is not currently zero and will not
65 * be set to zero. Returns true if a uevent was sent, otherwise false.
67 bool set_capacity_and_notify(struct gendisk *disk, sector_t size)
69 sector_t capacity = get_capacity(disk);
70 char *envp[] = { "RESIZE=1", NULL };
72 set_capacity(disk, size);
75 * Only print a message and send a uevent if the gendisk is user visible
76 * and alive. This avoids spamming the log and udev when setting the
77 * initial capacity during probing.
79 if (size == capacity ||
81 (disk->flags & GENHD_FL_HIDDEN))
84 pr_info("%s: detected capacity change from %lld to %lld\n",
85 disk->disk_name, capacity, size);
88 * Historically we did not send a uevent for changes to/from an empty
91 if (!capacity || !size)
93 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
96 EXPORT_SYMBOL_GPL(set_capacity_and_notify);
99 * Format the device name of the indicated block device into the supplied buffer
100 * and return a pointer to that same buffer for convenience.
102 * Note: do not use this in new code, use the %pg specifier to sprintf and
105 const char *bdevname(struct block_device *bdev, char *buf)
107 struct gendisk *hd = bdev->bd_disk;
108 int partno = bdev->bd_partno;
111 snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name);
112 else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1]))
113 snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno);
115 snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno);
119 EXPORT_SYMBOL(bdevname);
121 static void part_stat_read_all(struct block_device *part,
122 struct disk_stats *stat)
126 memset(stat, 0, sizeof(struct disk_stats));
127 for_each_possible_cpu(cpu) {
128 struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu);
131 for (group = 0; group < NR_STAT_GROUPS; group++) {
132 stat->nsecs[group] += ptr->nsecs[group];
133 stat->sectors[group] += ptr->sectors[group];
134 stat->ios[group] += ptr->ios[group];
135 stat->merges[group] += ptr->merges[group];
138 stat->io_ticks += ptr->io_ticks;
142 static unsigned int part_in_flight(struct block_device *part)
144 unsigned int inflight = 0;
147 for_each_possible_cpu(cpu) {
148 inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) +
149 part_stat_local_read_cpu(part, in_flight[1], cpu);
151 if ((int)inflight < 0)
157 static void part_in_flight_rw(struct block_device *part,
158 unsigned int inflight[2])
164 for_each_possible_cpu(cpu) {
165 inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu);
166 inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu);
168 if ((int)inflight[0] < 0)
170 if ((int)inflight[1] < 0)
175 * Can be deleted altogether. Later.
178 #define BLKDEV_MAJOR_HASH_SIZE 255
179 static struct blk_major_name {
180 struct blk_major_name *next;
183 void (*probe)(dev_t devt);
184 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
185 static DEFINE_MUTEX(major_names_lock);
186 static DEFINE_SPINLOCK(major_names_spinlock);
188 /* index in the above - for now: assume no multimajor ranges */
189 static inline int major_to_index(unsigned major)
191 return major % BLKDEV_MAJOR_HASH_SIZE;
194 #ifdef CONFIG_PROC_FS
195 void blkdev_show(struct seq_file *seqf, off_t offset)
197 struct blk_major_name *dp;
199 spin_lock(&major_names_spinlock);
200 for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
201 if (dp->major == offset)
202 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
203 spin_unlock(&major_names_spinlock);
205 #endif /* CONFIG_PROC_FS */
208 * __register_blkdev - register a new block device
210 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
211 * @major = 0, try to allocate any unused major number.
212 * @name: the name of the new block device as a zero terminated string
213 * @probe: allback that is called on access to any minor number of @major
215 * The @name must be unique within the system.
217 * The return value depends on the @major input parameter:
219 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
220 * then the function returns zero on success, or a negative error code
221 * - if any unused major number was requested with @major = 0 parameter
222 * then the return value is the allocated major number in range
223 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
225 * See Documentation/admin-guide/devices.txt for the list of allocated
228 * Use register_blkdev instead for any new code.
230 int __register_blkdev(unsigned int major, const char *name,
231 void (*probe)(dev_t devt))
233 struct blk_major_name **n, *p;
236 mutex_lock(&major_names_lock);
240 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
241 if (major_names[index] == NULL)
246 printk("%s: failed to get major for %s\n",
255 if (major >= BLKDEV_MAJOR_MAX) {
256 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
257 __func__, major, BLKDEV_MAJOR_MAX-1, name);
263 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
271 strlcpy(p->name, name, sizeof(p->name));
273 index = major_to_index(major);
275 spin_lock(&major_names_spinlock);
276 for (n = &major_names[index]; *n; n = &(*n)->next) {
277 if ((*n)->major == major)
284 spin_unlock(&major_names_spinlock);
287 printk("register_blkdev: cannot get major %u for %s\n",
292 mutex_unlock(&major_names_lock);
295 EXPORT_SYMBOL(__register_blkdev);
297 void unregister_blkdev(unsigned int major, const char *name)
299 struct blk_major_name **n;
300 struct blk_major_name *p = NULL;
301 int index = major_to_index(major);
303 mutex_lock(&major_names_lock);
304 spin_lock(&major_names_spinlock);
305 for (n = &major_names[index]; *n; n = &(*n)->next)
306 if ((*n)->major == major)
308 if (!*n || strcmp((*n)->name, name)) {
314 spin_unlock(&major_names_spinlock);
315 mutex_unlock(&major_names_lock);
319 EXPORT_SYMBOL(unregister_blkdev);
321 int blk_alloc_ext_minor(void)
325 idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT, GFP_KERNEL);
331 void blk_free_ext_minor(unsigned int minor)
333 ida_free(&ext_devt_ida, minor);
336 static char *bdevt_str(dev_t devt, char *buf)
338 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
339 char tbuf[BDEVT_SIZE];
340 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
341 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
343 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
348 void disk_uevent(struct gendisk *disk, enum kobject_action action)
350 struct block_device *part;
354 xa_for_each(&disk->part_tbl, idx, part) {
355 if (bdev_is_partition(part) && !bdev_nr_sectors(part))
357 if (!kobject_get_unless_zero(&part->bd_device.kobj))
361 kobject_uevent(bdev_kobj(part), action);
362 put_device(&part->bd_device);
367 EXPORT_SYMBOL_GPL(disk_uevent);
369 static void disk_scan_partitions(struct gendisk *disk)
371 struct block_device *bdev;
373 if (!get_capacity(disk) || !disk_part_scan_enabled(disk))
376 set_bit(GD_NEED_PART_SCAN, &disk->state);
377 bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL);
379 blkdev_put(bdev, FMODE_READ);
383 * device_add_disk - add disk information to kernel list
384 * @parent: parent device for the disk
385 * @disk: per-device partitioning information
386 * @groups: Additional per-device sysfs groups
388 * This function registers the partitioning information in @disk
391 int device_add_disk(struct device *parent, struct gendisk *disk,
392 const struct attribute_group **groups)
395 struct device *ddev = disk_to_dev(disk);
399 * The disk queue should now be all set with enough information about
400 * the device for the elevator code to pick an adequate default
401 * elevator if one is needed, that is, for devices requesting queue
404 elevator_init_mq(disk->queue);
407 * If the driver provides an explicit major number it also must provide
408 * the number of minors numbers supported, and those will be used to
410 * Otherwise just allocate the device numbers for both the whole device
411 * and all partitions from the extended dev_t space.
414 if (WARN_ON(!disk->minors))
417 if (disk->minors > DISK_MAX_PARTS) {
418 pr_err("block: can't allocate more than %d partitions\n",
420 disk->minors = DISK_MAX_PARTS;
423 if (WARN_ON(disk->minors))
426 ret = blk_alloc_ext_minor();
429 disk->major = BLOCK_EXT_MAJOR;
430 disk->first_minor = ret;
431 disk->flags |= GENHD_FL_EXT_DEVT;
434 ret = disk_alloc_events(disk);
436 goto out_free_ext_minor;
438 /* delay uevents, until we scanned partition table */
439 dev_set_uevent_suppress(ddev, 1);
441 ddev->parent = parent;
442 ddev->groups = groups;
443 dev_set_name(ddev, "%s", disk->disk_name);
444 if (!(disk->flags & GENHD_FL_HIDDEN))
445 ddev->devt = MKDEV(disk->major, disk->first_minor);
446 ret = device_add(ddev);
448 goto out_disk_release_events;
449 if (!sysfs_deprecated) {
450 ret = sysfs_create_link(block_depr, &ddev->kobj,
451 kobject_name(&ddev->kobj));
457 * avoid probable deadlock caused by allocating memory with
458 * GFP_KERNEL in runtime_resume callback of its all ancestor
461 pm_runtime_set_memalloc_noio(ddev, true);
463 ret = blk_integrity_add(disk);
465 goto out_del_block_link;
467 disk->part0->bd_holder_dir =
468 kobject_create_and_add("holders", &ddev->kobj);
469 if (!disk->part0->bd_holder_dir)
470 goto out_del_integrity;
471 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
472 if (!disk->slave_dir)
473 goto out_put_holder_dir;
475 ret = bd_register_pending_holders(disk);
477 goto out_put_slave_dir;
479 ret = blk_register_queue(disk);
481 goto out_put_slave_dir;
483 if (disk->flags & GENHD_FL_HIDDEN) {
485 * Don't let hidden disks show up in /proc/partitions,
486 * and don't bother scanning for partitions either.
488 disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
489 disk->flags |= GENHD_FL_NO_PART_SCAN;
491 ret = bdi_register(disk->bdi, "%u:%u",
492 disk->major, disk->first_minor);
494 goto out_unregister_queue;
495 bdi_set_owner(disk->bdi, ddev);
496 ret = sysfs_create_link(&ddev->kobj,
497 &disk->bdi->dev->kobj, "bdi");
499 goto out_unregister_bdi;
501 bdev_add(disk->part0, ddev->devt);
502 disk_scan_partitions(disk);
505 * Announce the disk and partitions after all partitions are
506 * created. (for hidden disks uevents remain suppressed forever)
508 dev_set_uevent_suppress(ddev, 0);
509 disk_uevent(disk, KOBJ_ADD);
512 disk_update_readahead(disk);
513 disk_add_events(disk);
517 if (!(disk->flags & GENHD_FL_HIDDEN))
518 bdi_unregister(disk->bdi);
519 out_unregister_queue:
520 blk_unregister_queue(disk);
522 kobject_put(disk->slave_dir);
524 kobject_put(disk->part0->bd_holder_dir);
526 blk_integrity_del(disk);
528 if (!sysfs_deprecated)
529 sysfs_remove_link(block_depr, dev_name(ddev));
532 out_disk_release_events:
533 disk_release_events(disk);
535 if (disk->major == BLOCK_EXT_MAJOR)
536 blk_free_ext_minor(disk->first_minor);
537 return WARN_ON_ONCE(ret); /* keep until all callers handle errors */
539 EXPORT_SYMBOL(device_add_disk);
542 * del_gendisk - remove the gendisk
543 * @disk: the struct gendisk to remove
545 * Removes the gendisk and all its associated resources. This deletes the
546 * partitions associated with the gendisk, and unregisters the associated
549 * This is the counter to the respective __device_add_disk() call.
551 * The final removal of the struct gendisk happens when its refcount reaches 0
552 * with put_disk(), which should be called after del_gendisk(), if
553 * __device_add_disk() was used.
555 * Drivers exist which depend on the release of the gendisk to be synchronous,
556 * it should not be deferred.
560 void del_gendisk(struct gendisk *disk)
564 if (WARN_ON_ONCE(!disk_live(disk) && !(disk->flags & GENHD_FL_HIDDEN)))
567 blk_integrity_del(disk);
568 disk_del_events(disk);
570 mutex_lock(&disk->open_mutex);
571 remove_inode_hash(disk->part0->bd_inode);
572 blk_drop_partitions(disk);
573 mutex_unlock(&disk->open_mutex);
575 fsync_bdev(disk->part0);
576 __invalidate_device(disk->part0, true);
578 set_capacity(disk, 0);
580 if (!(disk->flags & GENHD_FL_HIDDEN)) {
581 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
584 * Unregister bdi before releasing device numbers (as they can
585 * get reused and we'd get clashes in sysfs).
587 bdi_unregister(disk->bdi);
590 blk_unregister_queue(disk);
592 kobject_put(disk->part0->bd_holder_dir);
593 kobject_put(disk->slave_dir);
595 part_stat_set_all(disk->part0, 0);
596 disk->part0->bd_stamp = 0;
597 if (!sysfs_deprecated)
598 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
599 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
600 device_del(disk_to_dev(disk));
602 EXPORT_SYMBOL(del_gendisk);
604 /* sysfs access to bad-blocks list. */
605 static ssize_t disk_badblocks_show(struct device *dev,
606 struct device_attribute *attr,
609 struct gendisk *disk = dev_to_disk(dev);
612 return sprintf(page, "\n");
614 return badblocks_show(disk->bb, page, 0);
617 static ssize_t disk_badblocks_store(struct device *dev,
618 struct device_attribute *attr,
619 const char *page, size_t len)
621 struct gendisk *disk = dev_to_disk(dev);
626 return badblocks_store(disk->bb, page, len, 0);
629 void blk_request_module(dev_t devt)
631 unsigned int major = MAJOR(devt);
632 struct blk_major_name **n;
634 mutex_lock(&major_names_lock);
635 for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) {
636 if ((*n)->major == major && (*n)->probe) {
638 mutex_unlock(&major_names_lock);
642 mutex_unlock(&major_names_lock);
644 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
645 /* Make old-style 2.4 aliases work */
646 request_module("block-major-%d", MAJOR(devt));
650 * print a full list of all partitions - intended for places where the root
651 * filesystem can't be mounted and thus to give the victim some idea of what
654 void __init printk_all_partitions(void)
656 struct class_dev_iter iter;
659 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
660 while ((dev = class_dev_iter_next(&iter))) {
661 struct gendisk *disk = dev_to_disk(dev);
662 struct block_device *part;
663 char devt_buf[BDEVT_SIZE];
667 * Don't show empty devices or things that have been
670 if (get_capacity(disk) == 0 ||
671 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
675 * Note, unlike /proc/partitions, I am showing the numbers in
676 * hex - the same format as the root= option takes.
679 xa_for_each(&disk->part_tbl, idx, part) {
680 if (!bdev_nr_sectors(part))
682 printk("%s%s %10llu %pg %s",
683 bdev_is_partition(part) ? " " : "",
684 bdevt_str(part->bd_dev, devt_buf),
685 bdev_nr_sectors(part) >> 1, part,
687 part->bd_meta_info->uuid : "");
688 if (bdev_is_partition(part))
690 else if (dev->parent && dev->parent->driver)
691 printk(" driver: %s\n",
692 dev->parent->driver->name);
694 printk(" (driver?)\n");
698 class_dev_iter_exit(&iter);
701 #ifdef CONFIG_PROC_FS
703 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
706 struct class_dev_iter *iter;
709 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
711 return ERR_PTR(-ENOMEM);
713 seqf->private = iter;
714 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
716 dev = class_dev_iter_next(iter);
721 return dev_to_disk(dev);
724 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
729 dev = class_dev_iter_next(seqf->private);
731 return dev_to_disk(dev);
736 static void disk_seqf_stop(struct seq_file *seqf, void *v)
738 struct class_dev_iter *iter = seqf->private;
740 /* stop is called even after start failed :-( */
742 class_dev_iter_exit(iter);
744 seqf->private = NULL;
748 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
752 p = disk_seqf_start(seqf, pos);
753 if (!IS_ERR_OR_NULL(p) && !*pos)
754 seq_puts(seqf, "major minor #blocks name\n\n");
758 static int show_partition(struct seq_file *seqf, void *v)
760 struct gendisk *sgp = v;
761 struct block_device *part;
764 /* Don't show non-partitionable removeable devices or empty devices */
765 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
766 (sgp->flags & GENHD_FL_REMOVABLE)))
768 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
772 xa_for_each(&sgp->part_tbl, idx, part) {
773 if (!bdev_nr_sectors(part))
775 seq_printf(seqf, "%4d %7d %10llu %pg\n",
776 MAJOR(part->bd_dev), MINOR(part->bd_dev),
777 bdev_nr_sectors(part) >> 1, part);
783 static const struct seq_operations partitions_op = {
784 .start = show_partition_start,
785 .next = disk_seqf_next,
786 .stop = disk_seqf_stop,
787 .show = show_partition
791 static int __init genhd_device_init(void)
795 block_class.dev_kobj = sysfs_dev_block_kobj;
796 error = class_register(&block_class);
801 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
803 /* create top-level block dir */
804 if (!sysfs_deprecated)
805 block_depr = kobject_create_and_add("block", NULL);
809 subsys_initcall(genhd_device_init);
811 static ssize_t disk_range_show(struct device *dev,
812 struct device_attribute *attr, char *buf)
814 struct gendisk *disk = dev_to_disk(dev);
816 return sprintf(buf, "%d\n", disk->minors);
819 static ssize_t disk_ext_range_show(struct device *dev,
820 struct device_attribute *attr, char *buf)
822 struct gendisk *disk = dev_to_disk(dev);
824 return sprintf(buf, "%d\n", disk_max_parts(disk));
827 static ssize_t disk_removable_show(struct device *dev,
828 struct device_attribute *attr, char *buf)
830 struct gendisk *disk = dev_to_disk(dev);
832 return sprintf(buf, "%d\n",
833 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
836 static ssize_t disk_hidden_show(struct device *dev,
837 struct device_attribute *attr, char *buf)
839 struct gendisk *disk = dev_to_disk(dev);
841 return sprintf(buf, "%d\n",
842 (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
845 static ssize_t disk_ro_show(struct device *dev,
846 struct device_attribute *attr, char *buf)
848 struct gendisk *disk = dev_to_disk(dev);
850 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
853 ssize_t part_size_show(struct device *dev,
854 struct device_attribute *attr, char *buf)
856 return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev)));
859 ssize_t part_stat_show(struct device *dev,
860 struct device_attribute *attr, char *buf)
862 struct block_device *bdev = dev_to_bdev(dev);
863 struct request_queue *q = bdev->bd_disk->queue;
864 struct disk_stats stat;
865 unsigned int inflight;
867 part_stat_read_all(bdev, &stat);
869 inflight = blk_mq_in_flight(q, bdev);
871 inflight = part_in_flight(bdev);
874 "%8lu %8lu %8llu %8u "
875 "%8lu %8lu %8llu %8u "
877 "%8lu %8lu %8llu %8u "
881 stat.merges[STAT_READ],
882 (unsigned long long)stat.sectors[STAT_READ],
883 (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC),
884 stat.ios[STAT_WRITE],
885 stat.merges[STAT_WRITE],
886 (unsigned long long)stat.sectors[STAT_WRITE],
887 (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC),
889 jiffies_to_msecs(stat.io_ticks),
890 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
891 stat.nsecs[STAT_WRITE] +
892 stat.nsecs[STAT_DISCARD] +
893 stat.nsecs[STAT_FLUSH],
895 stat.ios[STAT_DISCARD],
896 stat.merges[STAT_DISCARD],
897 (unsigned long long)stat.sectors[STAT_DISCARD],
898 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC),
899 stat.ios[STAT_FLUSH],
900 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC));
903 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
906 struct block_device *bdev = dev_to_bdev(dev);
907 struct request_queue *q = bdev->bd_disk->queue;
908 unsigned int inflight[2];
911 blk_mq_in_flight_rw(q, bdev, inflight);
913 part_in_flight_rw(bdev, inflight);
915 return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
918 static ssize_t disk_capability_show(struct device *dev,
919 struct device_attribute *attr, char *buf)
921 struct gendisk *disk = dev_to_disk(dev);
923 return sprintf(buf, "%x\n", disk->flags);
926 static ssize_t disk_alignment_offset_show(struct device *dev,
927 struct device_attribute *attr,
930 struct gendisk *disk = dev_to_disk(dev);
932 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
935 static ssize_t disk_discard_alignment_show(struct device *dev,
936 struct device_attribute *attr,
939 struct gendisk *disk = dev_to_disk(dev);
941 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
944 static ssize_t diskseq_show(struct device *dev,
945 struct device_attribute *attr, char *buf)
947 struct gendisk *disk = dev_to_disk(dev);
949 return sprintf(buf, "%llu\n", disk->diskseq);
952 static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
953 static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
954 static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
955 static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
956 static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
957 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
958 static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
959 static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
960 static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
961 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
962 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
963 static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
964 static DEVICE_ATTR(diskseq, 0444, diskseq_show, NULL);
966 #ifdef CONFIG_FAIL_MAKE_REQUEST
967 ssize_t part_fail_show(struct device *dev,
968 struct device_attribute *attr, char *buf)
970 return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail);
973 ssize_t part_fail_store(struct device *dev,
974 struct device_attribute *attr,
975 const char *buf, size_t count)
979 if (count > 0 && sscanf(buf, "%d", &i) > 0)
980 dev_to_bdev(dev)->bd_make_it_fail = i;
985 static struct device_attribute dev_attr_fail =
986 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
987 #endif /* CONFIG_FAIL_MAKE_REQUEST */
989 #ifdef CONFIG_FAIL_IO_TIMEOUT
990 static struct device_attribute dev_attr_fail_timeout =
991 __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
994 static struct attribute *disk_attrs[] = {
995 &dev_attr_range.attr,
996 &dev_attr_ext_range.attr,
997 &dev_attr_removable.attr,
998 &dev_attr_hidden.attr,
1000 &dev_attr_size.attr,
1001 &dev_attr_alignment_offset.attr,
1002 &dev_attr_discard_alignment.attr,
1003 &dev_attr_capability.attr,
1004 &dev_attr_stat.attr,
1005 &dev_attr_inflight.attr,
1006 &dev_attr_badblocks.attr,
1007 &dev_attr_events.attr,
1008 &dev_attr_events_async.attr,
1009 &dev_attr_events_poll_msecs.attr,
1010 &dev_attr_diskseq.attr,
1011 #ifdef CONFIG_FAIL_MAKE_REQUEST
1012 &dev_attr_fail.attr,
1014 #ifdef CONFIG_FAIL_IO_TIMEOUT
1015 &dev_attr_fail_timeout.attr,
1020 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1022 struct device *dev = container_of(kobj, typeof(*dev), kobj);
1023 struct gendisk *disk = dev_to_disk(dev);
1025 if (a == &dev_attr_badblocks.attr && !disk->bb)
1030 static struct attribute_group disk_attr_group = {
1031 .attrs = disk_attrs,
1032 .is_visible = disk_visible,
1035 static const struct attribute_group *disk_attr_groups[] = {
1041 * disk_release - releases all allocated resources of the gendisk
1042 * @dev: the device representing this disk
1044 * This function releases all allocated resources of the gendisk.
1046 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1047 * assigned. Since the request_queue sits on top of the gendisk for these
1048 * drivers we also call blk_put_queue() for them, and we expect the
1049 * request_queue refcount to reach 0 at this point, and so the request_queue
1050 * will also be freed prior to the disk.
1052 * Context: can sleep
1054 static void disk_release(struct device *dev)
1056 struct gendisk *disk = dev_to_disk(dev);
1060 disk_release_events(disk);
1061 kfree(disk->random);
1062 xa_destroy(&disk->part_tbl);
1063 disk->queue->disk = NULL;
1064 blk_put_queue(disk->queue);
1065 iput(disk->part0->bd_inode); /* frees the disk */
1068 static int block_uevent(struct device *dev, struct kobj_uevent_env *env)
1070 struct gendisk *disk = dev_to_disk(dev);
1072 return add_uevent_var(env, "DISKSEQ=%llu", disk->diskseq);
1075 struct class block_class = {
1077 .dev_uevent = block_uevent,
1080 static char *block_devnode(struct device *dev, umode_t *mode,
1081 kuid_t *uid, kgid_t *gid)
1083 struct gendisk *disk = dev_to_disk(dev);
1085 if (disk->fops->devnode)
1086 return disk->fops->devnode(disk, mode);
1090 const struct device_type disk_type = {
1092 .groups = disk_attr_groups,
1093 .release = disk_release,
1094 .devnode = block_devnode,
1097 #ifdef CONFIG_PROC_FS
1099 * aggregate disk stat collector. Uses the same stats that the sysfs
1100 * entries do, above, but makes them available through one seq_file.
1102 * The output looks suspiciously like /proc/partitions with a bunch of
1105 static int diskstats_show(struct seq_file *seqf, void *v)
1107 struct gendisk *gp = v;
1108 struct block_device *hd;
1109 unsigned int inflight;
1110 struct disk_stats stat;
1114 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1115 seq_puts(seqf, "major minor name"
1116 " rio rmerge rsect ruse wio wmerge "
1117 "wsect wuse running use aveq"
1122 xa_for_each(&gp->part_tbl, idx, hd) {
1123 if (bdev_is_partition(hd) && !bdev_nr_sectors(hd))
1125 part_stat_read_all(hd, &stat);
1126 if (queue_is_mq(gp->queue))
1127 inflight = blk_mq_in_flight(gp->queue, hd);
1129 inflight = part_in_flight(hd);
1131 seq_printf(seqf, "%4d %7d %pg "
1138 MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd,
1139 stat.ios[STAT_READ],
1140 stat.merges[STAT_READ],
1141 stat.sectors[STAT_READ],
1142 (unsigned int)div_u64(stat.nsecs[STAT_READ],
1144 stat.ios[STAT_WRITE],
1145 stat.merges[STAT_WRITE],
1146 stat.sectors[STAT_WRITE],
1147 (unsigned int)div_u64(stat.nsecs[STAT_WRITE],
1150 jiffies_to_msecs(stat.io_ticks),
1151 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1152 stat.nsecs[STAT_WRITE] +
1153 stat.nsecs[STAT_DISCARD] +
1154 stat.nsecs[STAT_FLUSH],
1156 stat.ios[STAT_DISCARD],
1157 stat.merges[STAT_DISCARD],
1158 stat.sectors[STAT_DISCARD],
1159 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD],
1161 stat.ios[STAT_FLUSH],
1162 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH],
1171 static const struct seq_operations diskstats_op = {
1172 .start = disk_seqf_start,
1173 .next = disk_seqf_next,
1174 .stop = disk_seqf_stop,
1175 .show = diskstats_show
1178 static int __init proc_genhd_init(void)
1180 proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1181 proc_create_seq("partitions", 0, NULL, &partitions_op);
1184 module_init(proc_genhd_init);
1185 #endif /* CONFIG_PROC_FS */
1187 dev_t part_devt(struct gendisk *disk, u8 partno)
1189 struct block_device *part;
1193 part = xa_load(&disk->part_tbl, partno);
1195 devt = part->bd_dev;
1201 dev_t blk_lookup_devt(const char *name, int partno)
1203 dev_t devt = MKDEV(0, 0);
1204 struct class_dev_iter iter;
1207 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1208 while ((dev = class_dev_iter_next(&iter))) {
1209 struct gendisk *disk = dev_to_disk(dev);
1211 if (strcmp(dev_name(dev), name))
1214 if (partno < disk->minors) {
1215 /* We need to return the right devno, even
1216 * if the partition doesn't exist yet.
1218 devt = MKDEV(MAJOR(dev->devt),
1219 MINOR(dev->devt) + partno);
1221 devt = part_devt(disk, partno);
1226 class_dev_iter_exit(&iter);
1230 struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id,
1231 struct lock_class_key *lkclass)
1233 struct gendisk *disk;
1235 if (!blk_get_queue(q))
1238 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1242 disk->bdi = bdi_alloc(node_id);
1246 disk->part0 = bdev_alloc(disk, 0);
1250 disk->node_id = node_id;
1251 mutex_init(&disk->open_mutex);
1252 xa_init(&disk->part_tbl);
1253 if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL))
1254 goto out_destroy_part_tbl;
1256 rand_initialize_disk(disk);
1257 disk_to_dev(disk)->class = &block_class;
1258 disk_to_dev(disk)->type = &disk_type;
1259 device_initialize(disk_to_dev(disk));
1263 lockdep_init_map(&disk->lockdep_map, "(bio completion)", lkclass, 0);
1264 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
1265 INIT_LIST_HEAD(&disk->slave_bdevs);
1269 out_destroy_part_tbl:
1270 xa_destroy(&disk->part_tbl);
1271 iput(disk->part0->bd_inode);
1280 EXPORT_SYMBOL(__alloc_disk_node);
1282 struct gendisk *__blk_alloc_disk(int node, struct lock_class_key *lkclass)
1284 struct request_queue *q;
1285 struct gendisk *disk;
1287 q = blk_alloc_queue(node);
1291 disk = __alloc_disk_node(q, node, lkclass);
1293 blk_cleanup_queue(q);
1298 EXPORT_SYMBOL(__blk_alloc_disk);
1301 * put_disk - decrements the gendisk refcount
1302 * @disk: the struct gendisk to decrement the refcount for
1304 * This decrements the refcount for the struct gendisk. When this reaches 0
1305 * we'll have disk_release() called.
1307 * Context: Any context, but the last reference must not be dropped from
1310 void put_disk(struct gendisk *disk)
1313 put_device(disk_to_dev(disk));
1315 EXPORT_SYMBOL(put_disk);
1318 * blk_cleanup_disk - shutdown a gendisk allocated by blk_alloc_disk
1319 * @disk: gendisk to shutdown
1321 * Mark the queue hanging off @disk DYING, drain all pending requests, then mark
1322 * the queue DEAD, destroy and put it and the gendisk structure.
1324 * Context: can sleep
1326 void blk_cleanup_disk(struct gendisk *disk)
1328 blk_cleanup_queue(disk->queue);
1331 EXPORT_SYMBOL(blk_cleanup_disk);
1333 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1335 char event[] = "DISK_RO=1";
1336 char *envp[] = { event, NULL };
1340 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1344 * set_disk_ro - set a gendisk read-only
1345 * @disk: gendisk to operate on
1346 * @read_only: %true to set the disk read-only, %false set the disk read/write
1348 * This function is used to indicate whether a given disk device should have its
1349 * read-only flag set. set_disk_ro() is typically used by device drivers to
1350 * indicate whether the underlying physical device is write-protected.
1352 void set_disk_ro(struct gendisk *disk, bool read_only)
1355 if (test_and_set_bit(GD_READ_ONLY, &disk->state))
1358 if (!test_and_clear_bit(GD_READ_ONLY, &disk->state))
1361 set_disk_ro_uevent(disk, read_only);
1363 EXPORT_SYMBOL(set_disk_ro);
1365 int bdev_read_only(struct block_device *bdev)
1367 return bdev->bd_read_only || get_disk_ro(bdev->bd_disk);
1369 EXPORT_SYMBOL(bdev_read_only);
1371 void inc_diskseq(struct gendisk *disk)
1373 disk->diskseq = atomic64_inc_return(&diskseq);