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);
187 /* index in the above - for now: assume no multimajor ranges */
188 static inline int major_to_index(unsigned major)
190 return major % BLKDEV_MAJOR_HASH_SIZE;
193 #ifdef CONFIG_PROC_FS
194 void blkdev_show(struct seq_file *seqf, off_t offset)
196 struct blk_major_name *dp;
198 mutex_lock(&major_names_lock);
199 for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
200 if (dp->major == offset)
201 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
202 mutex_unlock(&major_names_lock);
204 #endif /* CONFIG_PROC_FS */
207 * __register_blkdev - register a new block device
209 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
210 * @major = 0, try to allocate any unused major number.
211 * @name: the name of the new block device as a zero terminated string
212 * @probe: allback that is called on access to any minor number of @major
214 * The @name must be unique within the system.
216 * The return value depends on the @major input parameter:
218 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
219 * then the function returns zero on success, or a negative error code
220 * - if any unused major number was requested with @major = 0 parameter
221 * then the return value is the allocated major number in range
222 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
224 * See Documentation/admin-guide/devices.txt for the list of allocated
227 * Use register_blkdev instead for any new code.
229 int __register_blkdev(unsigned int major, const char *name,
230 void (*probe)(dev_t devt))
232 struct blk_major_name **n, *p;
235 mutex_lock(&major_names_lock);
239 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
240 if (major_names[index] == NULL)
245 printk("%s: failed to get major for %s\n",
254 if (major >= BLKDEV_MAJOR_MAX) {
255 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
256 __func__, major, BLKDEV_MAJOR_MAX-1, name);
262 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
270 strlcpy(p->name, name, sizeof(p->name));
272 index = major_to_index(major);
274 for (n = &major_names[index]; *n; n = &(*n)->next) {
275 if ((*n)->major == major)
284 printk("register_blkdev: cannot get major %u for %s\n",
289 mutex_unlock(&major_names_lock);
292 EXPORT_SYMBOL(__register_blkdev);
294 void unregister_blkdev(unsigned int major, const char *name)
296 struct blk_major_name **n;
297 struct blk_major_name *p = NULL;
298 int index = major_to_index(major);
300 mutex_lock(&major_names_lock);
301 for (n = &major_names[index]; *n; n = &(*n)->next)
302 if ((*n)->major == major)
304 if (!*n || strcmp((*n)->name, name)) {
310 mutex_unlock(&major_names_lock);
314 EXPORT_SYMBOL(unregister_blkdev);
316 int blk_alloc_ext_minor(void)
320 idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT, GFP_KERNEL);
326 void blk_free_ext_minor(unsigned int minor)
328 ida_free(&ext_devt_ida, minor);
331 static char *bdevt_str(dev_t devt, char *buf)
333 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
334 char tbuf[BDEVT_SIZE];
335 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
336 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
338 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
343 void disk_uevent(struct gendisk *disk, enum kobject_action action)
345 struct block_device *part;
349 xa_for_each(&disk->part_tbl, idx, part) {
350 if (bdev_is_partition(part) && !bdev_nr_sectors(part))
352 if (!kobject_get_unless_zero(&part->bd_device.kobj))
356 kobject_uevent(bdev_kobj(part), action);
357 put_device(&part->bd_device);
362 EXPORT_SYMBOL_GPL(disk_uevent);
364 static void disk_scan_partitions(struct gendisk *disk)
366 struct block_device *bdev;
368 if (!get_capacity(disk) || !disk_part_scan_enabled(disk))
371 set_bit(GD_NEED_PART_SCAN, &disk->state);
372 bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL);
374 blkdev_put(bdev, FMODE_READ);
378 * device_add_disk - add disk information to kernel list
379 * @parent: parent device for the disk
380 * @disk: per-device partitioning information
381 * @groups: Additional per-device sysfs groups
383 * This function registers the partitioning information in @disk
386 int device_add_disk(struct device *parent, struct gendisk *disk,
387 const struct attribute_group **groups)
390 struct device *ddev = disk_to_dev(disk);
394 * The disk queue should now be all set with enough information about
395 * the device for the elevator code to pick an adequate default
396 * elevator if one is needed, that is, for devices requesting queue
399 elevator_init_mq(disk->queue);
402 * If the driver provides an explicit major number it also must provide
403 * the number of minors numbers supported, and those will be used to
405 * Otherwise just allocate the device numbers for both the whole device
406 * and all partitions from the extended dev_t space.
409 if (WARN_ON(!disk->minors))
412 if (disk->minors > DISK_MAX_PARTS) {
413 pr_err("block: can't allocate more than %d partitions\n",
415 disk->minors = DISK_MAX_PARTS;
418 if (WARN_ON(disk->minors))
421 ret = blk_alloc_ext_minor();
424 disk->major = BLOCK_EXT_MAJOR;
425 disk->first_minor = ret;
426 disk->flags |= GENHD_FL_EXT_DEVT;
429 ret = disk_alloc_events(disk);
431 goto out_free_ext_minor;
433 /* delay uevents, until we scanned partition table */
434 dev_set_uevent_suppress(ddev, 1);
436 ddev->parent = parent;
437 ddev->groups = groups;
438 dev_set_name(ddev, "%s", disk->disk_name);
439 if (!(disk->flags & GENHD_FL_HIDDEN))
440 ddev->devt = MKDEV(disk->major, disk->first_minor);
441 ret = device_add(ddev);
443 goto out_disk_release_events;
444 if (!sysfs_deprecated) {
445 ret = sysfs_create_link(block_depr, &ddev->kobj,
446 kobject_name(&ddev->kobj));
452 * avoid probable deadlock caused by allocating memory with
453 * GFP_KERNEL in runtime_resume callback of its all ancestor
456 pm_runtime_set_memalloc_noio(ddev, true);
458 ret = blk_integrity_add(disk);
460 goto out_del_block_link;
462 disk->part0->bd_holder_dir =
463 kobject_create_and_add("holders", &ddev->kobj);
464 if (!disk->part0->bd_holder_dir)
465 goto out_del_integrity;
466 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
467 if (!disk->slave_dir)
468 goto out_put_holder_dir;
470 ret = bd_register_pending_holders(disk);
472 goto out_put_slave_dir;
474 ret = blk_register_queue(disk);
476 goto out_put_slave_dir;
478 if (disk->flags & GENHD_FL_HIDDEN) {
480 * Don't let hidden disks show up in /proc/partitions,
481 * and don't bother scanning for partitions either.
483 disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
484 disk->flags |= GENHD_FL_NO_PART_SCAN;
486 ret = bdi_register(disk->bdi, "%u:%u",
487 disk->major, disk->first_minor);
489 goto out_unregister_queue;
490 bdi_set_owner(disk->bdi, ddev);
491 ret = sysfs_create_link(&ddev->kobj,
492 &disk->bdi->dev->kobj, "bdi");
494 goto out_unregister_bdi;
496 bdev_add(disk->part0, ddev->devt);
497 disk_scan_partitions(disk);
500 * Announce the disk and partitions after all partitions are
501 * created. (for hidden disks uevents remain suppressed forever)
503 dev_set_uevent_suppress(ddev, 0);
504 disk_uevent(disk, KOBJ_ADD);
507 disk_update_readahead(disk);
508 disk_add_events(disk);
512 if (!(disk->flags & GENHD_FL_HIDDEN))
513 bdi_unregister(disk->bdi);
514 out_unregister_queue:
515 blk_unregister_queue(disk);
517 kobject_put(disk->slave_dir);
519 kobject_put(disk->part0->bd_holder_dir);
521 blk_integrity_del(disk);
523 if (!sysfs_deprecated)
524 sysfs_remove_link(block_depr, dev_name(ddev));
527 out_disk_release_events:
528 disk_release_events(disk);
530 if (disk->major == BLOCK_EXT_MAJOR)
531 blk_free_ext_minor(disk->first_minor);
532 return WARN_ON_ONCE(ret); /* keep until all callers handle errors */
534 EXPORT_SYMBOL(device_add_disk);
537 * del_gendisk - remove the gendisk
538 * @disk: the struct gendisk to remove
540 * Removes the gendisk and all its associated resources. This deletes the
541 * partitions associated with the gendisk, and unregisters the associated
544 * This is the counter to the respective __device_add_disk() call.
546 * The final removal of the struct gendisk happens when its refcount reaches 0
547 * with put_disk(), which should be called after del_gendisk(), if
548 * __device_add_disk() was used.
550 * Drivers exist which depend on the release of the gendisk to be synchronous,
551 * it should not be deferred.
555 void del_gendisk(struct gendisk *disk)
559 if (WARN_ON_ONCE(!disk_live(disk) && !(disk->flags & GENHD_FL_HIDDEN)))
562 blk_integrity_del(disk);
563 disk_del_events(disk);
565 mutex_lock(&disk->open_mutex);
566 remove_inode_hash(disk->part0->bd_inode);
567 blk_drop_partitions(disk);
568 mutex_unlock(&disk->open_mutex);
570 fsync_bdev(disk->part0);
571 __invalidate_device(disk->part0, true);
573 set_capacity(disk, 0);
575 if (!(disk->flags & GENHD_FL_HIDDEN)) {
576 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
579 * Unregister bdi before releasing device numbers (as they can
580 * get reused and we'd get clashes in sysfs).
582 bdi_unregister(disk->bdi);
585 blk_unregister_queue(disk);
587 kobject_put(disk->part0->bd_holder_dir);
588 kobject_put(disk->slave_dir);
590 part_stat_set_all(disk->part0, 0);
591 disk->part0->bd_stamp = 0;
592 if (!sysfs_deprecated)
593 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
594 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
595 device_del(disk_to_dev(disk));
597 EXPORT_SYMBOL(del_gendisk);
599 /* sysfs access to bad-blocks list. */
600 static ssize_t disk_badblocks_show(struct device *dev,
601 struct device_attribute *attr,
604 struct gendisk *disk = dev_to_disk(dev);
607 return sprintf(page, "\n");
609 return badblocks_show(disk->bb, page, 0);
612 static ssize_t disk_badblocks_store(struct device *dev,
613 struct device_attribute *attr,
614 const char *page, size_t len)
616 struct gendisk *disk = dev_to_disk(dev);
621 return badblocks_store(disk->bb, page, len, 0);
624 void blk_request_module(dev_t devt)
626 unsigned int major = MAJOR(devt);
627 struct blk_major_name **n;
629 mutex_lock(&major_names_lock);
630 for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) {
631 if ((*n)->major == major && (*n)->probe) {
633 mutex_unlock(&major_names_lock);
637 mutex_unlock(&major_names_lock);
639 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
640 /* Make old-style 2.4 aliases work */
641 request_module("block-major-%d", MAJOR(devt));
645 * print a full list of all partitions - intended for places where the root
646 * filesystem can't be mounted and thus to give the victim some idea of what
649 void __init printk_all_partitions(void)
651 struct class_dev_iter iter;
654 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
655 while ((dev = class_dev_iter_next(&iter))) {
656 struct gendisk *disk = dev_to_disk(dev);
657 struct block_device *part;
658 char devt_buf[BDEVT_SIZE];
662 * Don't show empty devices or things that have been
665 if (get_capacity(disk) == 0 ||
666 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
670 * Note, unlike /proc/partitions, I am showing the numbers in
671 * hex - the same format as the root= option takes.
674 xa_for_each(&disk->part_tbl, idx, part) {
675 if (!bdev_nr_sectors(part))
677 printk("%s%s %10llu %pg %s",
678 bdev_is_partition(part) ? " " : "",
679 bdevt_str(part->bd_dev, devt_buf),
680 bdev_nr_sectors(part) >> 1, part,
682 part->bd_meta_info->uuid : "");
683 if (bdev_is_partition(part))
685 else if (dev->parent && dev->parent->driver)
686 printk(" driver: %s\n",
687 dev->parent->driver->name);
689 printk(" (driver?)\n");
693 class_dev_iter_exit(&iter);
696 #ifdef CONFIG_PROC_FS
698 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
701 struct class_dev_iter *iter;
704 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
706 return ERR_PTR(-ENOMEM);
708 seqf->private = iter;
709 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
711 dev = class_dev_iter_next(iter);
716 return dev_to_disk(dev);
719 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
724 dev = class_dev_iter_next(seqf->private);
726 return dev_to_disk(dev);
731 static void disk_seqf_stop(struct seq_file *seqf, void *v)
733 struct class_dev_iter *iter = seqf->private;
735 /* stop is called even after start failed :-( */
737 class_dev_iter_exit(iter);
739 seqf->private = NULL;
743 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
747 p = disk_seqf_start(seqf, pos);
748 if (!IS_ERR_OR_NULL(p) && !*pos)
749 seq_puts(seqf, "major minor #blocks name\n\n");
753 static int show_partition(struct seq_file *seqf, void *v)
755 struct gendisk *sgp = v;
756 struct block_device *part;
759 /* Don't show non-partitionable removeable devices or empty devices */
760 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
761 (sgp->flags & GENHD_FL_REMOVABLE)))
763 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
767 xa_for_each(&sgp->part_tbl, idx, part) {
768 if (!bdev_nr_sectors(part))
770 seq_printf(seqf, "%4d %7d %10llu %pg\n",
771 MAJOR(part->bd_dev), MINOR(part->bd_dev),
772 bdev_nr_sectors(part) >> 1, part);
778 static const struct seq_operations partitions_op = {
779 .start = show_partition_start,
780 .next = disk_seqf_next,
781 .stop = disk_seqf_stop,
782 .show = show_partition
786 static int __init genhd_device_init(void)
790 block_class.dev_kobj = sysfs_dev_block_kobj;
791 error = class_register(&block_class);
796 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
798 /* create top-level block dir */
799 if (!sysfs_deprecated)
800 block_depr = kobject_create_and_add("block", NULL);
804 subsys_initcall(genhd_device_init);
806 static ssize_t disk_range_show(struct device *dev,
807 struct device_attribute *attr, char *buf)
809 struct gendisk *disk = dev_to_disk(dev);
811 return sprintf(buf, "%d\n", disk->minors);
814 static ssize_t disk_ext_range_show(struct device *dev,
815 struct device_attribute *attr, char *buf)
817 struct gendisk *disk = dev_to_disk(dev);
819 return sprintf(buf, "%d\n", disk_max_parts(disk));
822 static ssize_t disk_removable_show(struct device *dev,
823 struct device_attribute *attr, char *buf)
825 struct gendisk *disk = dev_to_disk(dev);
827 return sprintf(buf, "%d\n",
828 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
831 static ssize_t disk_hidden_show(struct device *dev,
832 struct device_attribute *attr, char *buf)
834 struct gendisk *disk = dev_to_disk(dev);
836 return sprintf(buf, "%d\n",
837 (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
840 static ssize_t disk_ro_show(struct device *dev,
841 struct device_attribute *attr, char *buf)
843 struct gendisk *disk = dev_to_disk(dev);
845 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
848 ssize_t part_size_show(struct device *dev,
849 struct device_attribute *attr, char *buf)
851 return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev)));
854 ssize_t part_stat_show(struct device *dev,
855 struct device_attribute *attr, char *buf)
857 struct block_device *bdev = dev_to_bdev(dev);
858 struct request_queue *q = bdev->bd_disk->queue;
859 struct disk_stats stat;
860 unsigned int inflight;
862 part_stat_read_all(bdev, &stat);
864 inflight = blk_mq_in_flight(q, bdev);
866 inflight = part_in_flight(bdev);
869 "%8lu %8lu %8llu %8u "
870 "%8lu %8lu %8llu %8u "
872 "%8lu %8lu %8llu %8u "
876 stat.merges[STAT_READ],
877 (unsigned long long)stat.sectors[STAT_READ],
878 (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC),
879 stat.ios[STAT_WRITE],
880 stat.merges[STAT_WRITE],
881 (unsigned long long)stat.sectors[STAT_WRITE],
882 (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC),
884 jiffies_to_msecs(stat.io_ticks),
885 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
886 stat.nsecs[STAT_WRITE] +
887 stat.nsecs[STAT_DISCARD] +
888 stat.nsecs[STAT_FLUSH],
890 stat.ios[STAT_DISCARD],
891 stat.merges[STAT_DISCARD],
892 (unsigned long long)stat.sectors[STAT_DISCARD],
893 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC),
894 stat.ios[STAT_FLUSH],
895 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC));
898 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
901 struct block_device *bdev = dev_to_bdev(dev);
902 struct request_queue *q = bdev->bd_disk->queue;
903 unsigned int inflight[2];
906 blk_mq_in_flight_rw(q, bdev, inflight);
908 part_in_flight_rw(bdev, inflight);
910 return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
913 static ssize_t disk_capability_show(struct device *dev,
914 struct device_attribute *attr, char *buf)
916 struct gendisk *disk = dev_to_disk(dev);
918 return sprintf(buf, "%x\n", disk->flags);
921 static ssize_t disk_alignment_offset_show(struct device *dev,
922 struct device_attribute *attr,
925 struct gendisk *disk = dev_to_disk(dev);
927 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
930 static ssize_t disk_discard_alignment_show(struct device *dev,
931 struct device_attribute *attr,
934 struct gendisk *disk = dev_to_disk(dev);
936 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
939 static ssize_t diskseq_show(struct device *dev,
940 struct device_attribute *attr, char *buf)
942 struct gendisk *disk = dev_to_disk(dev);
944 return sprintf(buf, "%llu\n", disk->diskseq);
947 static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
948 static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
949 static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
950 static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
951 static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
952 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
953 static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
954 static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
955 static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
956 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
957 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
958 static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
959 static DEVICE_ATTR(diskseq, 0444, diskseq_show, NULL);
961 #ifdef CONFIG_FAIL_MAKE_REQUEST
962 ssize_t part_fail_show(struct device *dev,
963 struct device_attribute *attr, char *buf)
965 return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail);
968 ssize_t part_fail_store(struct device *dev,
969 struct device_attribute *attr,
970 const char *buf, size_t count)
974 if (count > 0 && sscanf(buf, "%d", &i) > 0)
975 dev_to_bdev(dev)->bd_make_it_fail = i;
980 static struct device_attribute dev_attr_fail =
981 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
982 #endif /* CONFIG_FAIL_MAKE_REQUEST */
984 #ifdef CONFIG_FAIL_IO_TIMEOUT
985 static struct device_attribute dev_attr_fail_timeout =
986 __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
989 static struct attribute *disk_attrs[] = {
990 &dev_attr_range.attr,
991 &dev_attr_ext_range.attr,
992 &dev_attr_removable.attr,
993 &dev_attr_hidden.attr,
996 &dev_attr_alignment_offset.attr,
997 &dev_attr_discard_alignment.attr,
998 &dev_attr_capability.attr,
1000 &dev_attr_inflight.attr,
1001 &dev_attr_badblocks.attr,
1002 &dev_attr_events.attr,
1003 &dev_attr_events_async.attr,
1004 &dev_attr_events_poll_msecs.attr,
1005 &dev_attr_diskseq.attr,
1006 #ifdef CONFIG_FAIL_MAKE_REQUEST
1007 &dev_attr_fail.attr,
1009 #ifdef CONFIG_FAIL_IO_TIMEOUT
1010 &dev_attr_fail_timeout.attr,
1015 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1017 struct device *dev = container_of(kobj, typeof(*dev), kobj);
1018 struct gendisk *disk = dev_to_disk(dev);
1020 if (a == &dev_attr_badblocks.attr && !disk->bb)
1025 static struct attribute_group disk_attr_group = {
1026 .attrs = disk_attrs,
1027 .is_visible = disk_visible,
1030 static const struct attribute_group *disk_attr_groups[] = {
1036 * disk_release - releases all allocated resources of the gendisk
1037 * @dev: the device representing this disk
1039 * This function releases all allocated resources of the gendisk.
1041 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1042 * assigned. Since the request_queue sits on top of the gendisk for these
1043 * drivers we also call blk_put_queue() for them, and we expect the
1044 * request_queue refcount to reach 0 at this point, and so the request_queue
1045 * will also be freed prior to the disk.
1047 * Context: can sleep
1049 static void disk_release(struct device *dev)
1051 struct gendisk *disk = dev_to_disk(dev);
1055 disk_release_events(disk);
1056 kfree(disk->random);
1057 xa_destroy(&disk->part_tbl);
1058 disk->queue->disk = NULL;
1059 blk_put_queue(disk->queue);
1060 iput(disk->part0->bd_inode); /* frees the disk */
1063 static int block_uevent(struct device *dev, struct kobj_uevent_env *env)
1065 struct gendisk *disk = dev_to_disk(dev);
1067 return add_uevent_var(env, "DISKSEQ=%llu", disk->diskseq);
1070 struct class block_class = {
1072 .dev_uevent = block_uevent,
1075 static char *block_devnode(struct device *dev, umode_t *mode,
1076 kuid_t *uid, kgid_t *gid)
1078 struct gendisk *disk = dev_to_disk(dev);
1080 if (disk->fops->devnode)
1081 return disk->fops->devnode(disk, mode);
1085 const struct device_type disk_type = {
1087 .groups = disk_attr_groups,
1088 .release = disk_release,
1089 .devnode = block_devnode,
1092 #ifdef CONFIG_PROC_FS
1094 * aggregate disk stat collector. Uses the same stats that the sysfs
1095 * entries do, above, but makes them available through one seq_file.
1097 * The output looks suspiciously like /proc/partitions with a bunch of
1100 static int diskstats_show(struct seq_file *seqf, void *v)
1102 struct gendisk *gp = v;
1103 struct block_device *hd;
1104 unsigned int inflight;
1105 struct disk_stats stat;
1109 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1110 seq_puts(seqf, "major minor name"
1111 " rio rmerge rsect ruse wio wmerge "
1112 "wsect wuse running use aveq"
1117 xa_for_each(&gp->part_tbl, idx, hd) {
1118 if (bdev_is_partition(hd) && !bdev_nr_sectors(hd))
1120 part_stat_read_all(hd, &stat);
1121 if (queue_is_mq(gp->queue))
1122 inflight = blk_mq_in_flight(gp->queue, hd);
1124 inflight = part_in_flight(hd);
1126 seq_printf(seqf, "%4d %7d %pg "
1133 MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd,
1134 stat.ios[STAT_READ],
1135 stat.merges[STAT_READ],
1136 stat.sectors[STAT_READ],
1137 (unsigned int)div_u64(stat.nsecs[STAT_READ],
1139 stat.ios[STAT_WRITE],
1140 stat.merges[STAT_WRITE],
1141 stat.sectors[STAT_WRITE],
1142 (unsigned int)div_u64(stat.nsecs[STAT_WRITE],
1145 jiffies_to_msecs(stat.io_ticks),
1146 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1147 stat.nsecs[STAT_WRITE] +
1148 stat.nsecs[STAT_DISCARD] +
1149 stat.nsecs[STAT_FLUSH],
1151 stat.ios[STAT_DISCARD],
1152 stat.merges[STAT_DISCARD],
1153 stat.sectors[STAT_DISCARD],
1154 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD],
1156 stat.ios[STAT_FLUSH],
1157 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH],
1166 static const struct seq_operations diskstats_op = {
1167 .start = disk_seqf_start,
1168 .next = disk_seqf_next,
1169 .stop = disk_seqf_stop,
1170 .show = diskstats_show
1173 static int __init proc_genhd_init(void)
1175 proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1176 proc_create_seq("partitions", 0, NULL, &partitions_op);
1179 module_init(proc_genhd_init);
1180 #endif /* CONFIG_PROC_FS */
1182 dev_t part_devt(struct gendisk *disk, u8 partno)
1184 struct block_device *part;
1188 part = xa_load(&disk->part_tbl, partno);
1190 devt = part->bd_dev;
1196 dev_t blk_lookup_devt(const char *name, int partno)
1198 dev_t devt = MKDEV(0, 0);
1199 struct class_dev_iter iter;
1202 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1203 while ((dev = class_dev_iter_next(&iter))) {
1204 struct gendisk *disk = dev_to_disk(dev);
1206 if (strcmp(dev_name(dev), name))
1209 if (partno < disk->minors) {
1210 /* We need to return the right devno, even
1211 * if the partition doesn't exist yet.
1213 devt = MKDEV(MAJOR(dev->devt),
1214 MINOR(dev->devt) + partno);
1216 devt = part_devt(disk, partno);
1221 class_dev_iter_exit(&iter);
1225 struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id,
1226 struct lock_class_key *lkclass)
1228 struct gendisk *disk;
1230 if (!blk_get_queue(q))
1233 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1237 disk->bdi = bdi_alloc(node_id);
1241 disk->part0 = bdev_alloc(disk, 0);
1245 disk->node_id = node_id;
1246 mutex_init(&disk->open_mutex);
1247 xa_init(&disk->part_tbl);
1248 if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL))
1249 goto out_destroy_part_tbl;
1251 rand_initialize_disk(disk);
1252 disk_to_dev(disk)->class = &block_class;
1253 disk_to_dev(disk)->type = &disk_type;
1254 device_initialize(disk_to_dev(disk));
1258 lockdep_init_map(&disk->lockdep_map, "(bio completion)", lkclass, 0);
1259 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
1260 INIT_LIST_HEAD(&disk->slave_bdevs);
1264 out_destroy_part_tbl:
1265 xa_destroy(&disk->part_tbl);
1266 iput(disk->part0->bd_inode);
1275 EXPORT_SYMBOL(__alloc_disk_node);
1277 struct gendisk *__blk_alloc_disk(int node, struct lock_class_key *lkclass)
1279 struct request_queue *q;
1280 struct gendisk *disk;
1282 q = blk_alloc_queue(node);
1286 disk = __alloc_disk_node(q, node, lkclass);
1288 blk_cleanup_queue(q);
1293 EXPORT_SYMBOL(__blk_alloc_disk);
1296 * put_disk - decrements the gendisk refcount
1297 * @disk: the struct gendisk to decrement the refcount for
1299 * This decrements the refcount for the struct gendisk. When this reaches 0
1300 * we'll have disk_release() called.
1302 * Context: Any context, but the last reference must not be dropped from
1305 void put_disk(struct gendisk *disk)
1308 put_device(disk_to_dev(disk));
1310 EXPORT_SYMBOL(put_disk);
1313 * blk_cleanup_disk - shutdown a gendisk allocated by blk_alloc_disk
1314 * @disk: gendisk to shutdown
1316 * Mark the queue hanging off @disk DYING, drain all pending requests, then mark
1317 * the queue DEAD, destroy and put it and the gendisk structure.
1319 * Context: can sleep
1321 void blk_cleanup_disk(struct gendisk *disk)
1323 blk_cleanup_queue(disk->queue);
1326 EXPORT_SYMBOL(blk_cleanup_disk);
1328 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1330 char event[] = "DISK_RO=1";
1331 char *envp[] = { event, NULL };
1335 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1339 * set_disk_ro - set a gendisk read-only
1340 * @disk: gendisk to operate on
1341 * @read_only: %true to set the disk read-only, %false set the disk read/write
1343 * This function is used to indicate whether a given disk device should have its
1344 * read-only flag set. set_disk_ro() is typically used by device drivers to
1345 * indicate whether the underlying physical device is write-protected.
1347 void set_disk_ro(struct gendisk *disk, bool read_only)
1350 if (test_and_set_bit(GD_READ_ONLY, &disk->state))
1353 if (!test_and_clear_bit(GD_READ_ONLY, &disk->state))
1356 set_disk_ro_uevent(disk, read_only);
1358 EXPORT_SYMBOL(set_disk_ro);
1360 int bdev_read_only(struct block_device *bdev)
1362 return bdev->bd_read_only || get_disk_ro(bdev->bd_disk);
1364 EXPORT_SYMBOL(bdev_read_only);
1366 void inc_diskseq(struct gendisk *disk)
1368 disk->diskseq = atomic64_inc_return(&diskseq);