1 // SPDX-License-Identifier: GPL-2.0
5 * Portions Copyright (C) 2020 Christoph Hellwig
8 #include <linux/module.h>
9 #include <linux/ctype.h>
11 #include <linux/genhd.h>
12 #include <linux/kdev_t.h>
13 #include <linux/kernel.h>
14 #include <linux/blkdev.h>
15 #include <linux/backing-dev.h>
16 #include <linux/init.h>
17 #include <linux/spinlock.h>
18 #include <linux/proc_fs.h>
19 #include <linux/seq_file.h>
20 #include <linux/slab.h>
21 #include <linux/kmod.h>
22 #include <linux/mutex.h>
23 #include <linux/idr.h>
24 #include <linux/log2.h>
25 #include <linux/pm_runtime.h>
26 #include <linux/badblocks.h>
29 #include "blk-rq-qos.h"
31 static struct kobject *block_depr;
34 * Unique, monotonically increasing sequential number associated with block
35 * devices instances (i.e. incremented each time a device is attached).
36 * Associating uevents with block devices in userspace is difficult and racy:
37 * the uevent netlink socket is lossy, and on slow and overloaded systems has
38 * a very high latency.
39 * Block devices do not have exclusive owners in userspace, any process can set
40 * one up (e.g. loop devices). Moreover, device names can be reused (e.g. loop0
41 * can be reused again and again).
42 * A userspace process setting up a block device and watching for its events
43 * cannot thus reliably tell whether an event relates to the device it just set
44 * up or another earlier instance with the same name.
45 * This sequential number allows userspace processes to solve this problem, and
46 * uniquely associate an uevent to the lifetime to a device.
48 static atomic64_t diskseq;
50 /* for extended dynamic devt allocation, currently only one major is used */
51 #define NR_EXT_DEVT (1 << MINORBITS)
52 static DEFINE_IDA(ext_devt_ida);
54 void set_capacity(struct gendisk *disk, sector_t sectors)
56 struct block_device *bdev = disk->part0;
58 spin_lock(&bdev->bd_size_lock);
59 i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
60 spin_unlock(&bdev->bd_size_lock);
62 EXPORT_SYMBOL(set_capacity);
65 * Set disk capacity and notify if the size is not currently zero and will not
66 * be set to zero. Returns true if a uevent was sent, otherwise false.
68 bool set_capacity_and_notify(struct gendisk *disk, sector_t size)
70 sector_t capacity = get_capacity(disk);
71 char *envp[] = { "RESIZE=1", NULL };
73 set_capacity(disk, size);
76 * Only print a message and send a uevent if the gendisk is user visible
77 * and alive. This avoids spamming the log and udev when setting the
78 * initial capacity during probing.
80 if (size == capacity ||
82 (disk->flags & GENHD_FL_HIDDEN))
85 pr_info("%s: detected capacity change from %lld to %lld\n",
86 disk->disk_name, capacity, size);
89 * Historically we did not send a uevent for changes to/from an empty
92 if (!capacity || !size)
94 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
97 EXPORT_SYMBOL_GPL(set_capacity_and_notify);
100 * Format the device name of the indicated block device into the supplied buffer
101 * and return a pointer to that same buffer for convenience.
103 * Note: do not use this in new code, use the %pg specifier to sprintf and
106 const char *bdevname(struct block_device *bdev, char *buf)
108 struct gendisk *hd = bdev->bd_disk;
109 int partno = bdev->bd_partno;
112 snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name);
113 else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1]))
114 snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno);
116 snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno);
120 EXPORT_SYMBOL(bdevname);
122 static void part_stat_read_all(struct block_device *part,
123 struct disk_stats *stat)
127 memset(stat, 0, sizeof(struct disk_stats));
128 for_each_possible_cpu(cpu) {
129 struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu);
132 for (group = 0; group < NR_STAT_GROUPS; group++) {
133 stat->nsecs[group] += ptr->nsecs[group];
134 stat->sectors[group] += ptr->sectors[group];
135 stat->ios[group] += ptr->ios[group];
136 stat->merges[group] += ptr->merges[group];
139 stat->io_ticks += ptr->io_ticks;
143 static unsigned int part_in_flight(struct block_device *part)
145 unsigned int inflight = 0;
148 for_each_possible_cpu(cpu) {
149 inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) +
150 part_stat_local_read_cpu(part, in_flight[1], cpu);
152 if ((int)inflight < 0)
158 static void part_in_flight_rw(struct block_device *part,
159 unsigned int inflight[2])
165 for_each_possible_cpu(cpu) {
166 inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu);
167 inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu);
169 if ((int)inflight[0] < 0)
171 if ((int)inflight[1] < 0)
176 * Can be deleted altogether. Later.
179 #define BLKDEV_MAJOR_HASH_SIZE 255
180 static struct blk_major_name {
181 struct blk_major_name *next;
184 void (*probe)(dev_t devt);
185 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
186 static DEFINE_MUTEX(major_names_lock);
187 static DEFINE_SPINLOCK(major_names_spinlock);
189 /* index in the above - for now: assume no multimajor ranges */
190 static inline int major_to_index(unsigned major)
192 return major % BLKDEV_MAJOR_HASH_SIZE;
195 #ifdef CONFIG_PROC_FS
196 void blkdev_show(struct seq_file *seqf, off_t offset)
198 struct blk_major_name *dp;
200 spin_lock(&major_names_spinlock);
201 for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
202 if (dp->major == offset)
203 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
204 spin_unlock(&major_names_spinlock);
206 #endif /* CONFIG_PROC_FS */
209 * __register_blkdev - register a new block device
211 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
212 * @major = 0, try to allocate any unused major number.
213 * @name: the name of the new block device as a zero terminated string
214 * @probe: allback that is called on access to any minor number of @major
216 * The @name must be unique within the system.
218 * The return value depends on the @major input parameter:
220 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
221 * then the function returns zero on success, or a negative error code
222 * - if any unused major number was requested with @major = 0 parameter
223 * then the return value is the allocated major number in range
224 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
226 * See Documentation/admin-guide/devices.txt for the list of allocated
229 * Use register_blkdev instead for any new code.
231 int __register_blkdev(unsigned int major, const char *name,
232 void (*probe)(dev_t devt))
234 struct blk_major_name **n, *p;
237 mutex_lock(&major_names_lock);
241 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
242 if (major_names[index] == NULL)
247 printk("%s: failed to get major for %s\n",
256 if (major >= BLKDEV_MAJOR_MAX) {
257 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
258 __func__, major, BLKDEV_MAJOR_MAX-1, name);
264 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
272 strlcpy(p->name, name, sizeof(p->name));
274 index = major_to_index(major);
276 spin_lock(&major_names_spinlock);
277 for (n = &major_names[index]; *n; n = &(*n)->next) {
278 if ((*n)->major == major)
285 spin_unlock(&major_names_spinlock);
288 printk("register_blkdev: cannot get major %u for %s\n",
293 mutex_unlock(&major_names_lock);
296 EXPORT_SYMBOL(__register_blkdev);
298 void unregister_blkdev(unsigned int major, const char *name)
300 struct blk_major_name **n;
301 struct blk_major_name *p = NULL;
302 int index = major_to_index(major);
304 mutex_lock(&major_names_lock);
305 spin_lock(&major_names_spinlock);
306 for (n = &major_names[index]; *n; n = &(*n)->next)
307 if ((*n)->major == major)
309 if (!*n || strcmp((*n)->name, name)) {
315 spin_unlock(&major_names_spinlock);
316 mutex_unlock(&major_names_lock);
320 EXPORT_SYMBOL(unregister_blkdev);
322 int blk_alloc_ext_minor(void)
326 idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT, GFP_KERNEL);
332 void blk_free_ext_minor(unsigned int minor)
334 ida_free(&ext_devt_ida, minor);
337 static char *bdevt_str(dev_t devt, char *buf)
339 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
340 char tbuf[BDEVT_SIZE];
341 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
342 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
344 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
349 void disk_uevent(struct gendisk *disk, enum kobject_action action)
351 struct block_device *part;
355 xa_for_each(&disk->part_tbl, idx, part) {
356 if (bdev_is_partition(part) && !bdev_nr_sectors(part))
358 if (!kobject_get_unless_zero(&part->bd_device.kobj))
362 kobject_uevent(bdev_kobj(part), action);
363 put_device(&part->bd_device);
368 EXPORT_SYMBOL_GPL(disk_uevent);
370 static void disk_scan_partitions(struct gendisk *disk)
372 struct block_device *bdev;
374 if (!get_capacity(disk) || !disk_part_scan_enabled(disk))
377 set_bit(GD_NEED_PART_SCAN, &disk->state);
378 bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL);
380 blkdev_put(bdev, FMODE_READ);
384 * device_add_disk - add disk information to kernel list
385 * @parent: parent device for the disk
386 * @disk: per-device partitioning information
387 * @groups: Additional per-device sysfs groups
389 * This function registers the partitioning information in @disk
392 int device_add_disk(struct device *parent, struct gendisk *disk,
393 const struct attribute_group **groups)
396 struct device *ddev = disk_to_dev(disk);
400 * The disk queue should now be all set with enough information about
401 * the device for the elevator code to pick an adequate default
402 * elevator if one is needed, that is, for devices requesting queue
405 elevator_init_mq(disk->queue);
408 * If the driver provides an explicit major number it also must provide
409 * the number of minors numbers supported, and those will be used to
411 * Otherwise just allocate the device numbers for both the whole device
412 * and all partitions from the extended dev_t space.
415 if (WARN_ON(!disk->minors))
418 if (disk->minors > DISK_MAX_PARTS) {
419 pr_err("block: can't allocate more than %d partitions\n",
421 disk->minors = DISK_MAX_PARTS;
424 if (WARN_ON(disk->minors))
427 ret = blk_alloc_ext_minor();
430 disk->major = BLOCK_EXT_MAJOR;
431 disk->first_minor = ret;
432 disk->flags |= GENHD_FL_EXT_DEVT;
435 ret = disk_alloc_events(disk);
437 goto out_free_ext_minor;
439 /* delay uevents, until we scanned partition table */
440 dev_set_uevent_suppress(ddev, 1);
442 ddev->parent = parent;
443 ddev->groups = groups;
444 dev_set_name(ddev, "%s", disk->disk_name);
445 if (!(disk->flags & GENHD_FL_HIDDEN))
446 ddev->devt = MKDEV(disk->major, disk->first_minor);
447 ret = device_add(ddev);
449 goto out_disk_release_events;
450 if (!sysfs_deprecated) {
451 ret = sysfs_create_link(block_depr, &ddev->kobj,
452 kobject_name(&ddev->kobj));
458 * avoid probable deadlock caused by allocating memory with
459 * GFP_KERNEL in runtime_resume callback of its all ancestor
462 pm_runtime_set_memalloc_noio(ddev, true);
464 ret = blk_integrity_add(disk);
466 goto out_del_block_link;
468 disk->part0->bd_holder_dir =
469 kobject_create_and_add("holders", &ddev->kobj);
470 if (!disk->part0->bd_holder_dir)
471 goto out_del_integrity;
472 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
473 if (!disk->slave_dir)
474 goto out_put_holder_dir;
476 ret = bd_register_pending_holders(disk);
478 goto out_put_slave_dir;
480 ret = blk_register_queue(disk);
482 goto out_put_slave_dir;
484 if (disk->flags & GENHD_FL_HIDDEN) {
486 * Don't let hidden disks show up in /proc/partitions,
487 * and don't bother scanning for partitions either.
489 disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
490 disk->flags |= GENHD_FL_NO_PART_SCAN;
492 ret = bdi_register(disk->bdi, "%u:%u",
493 disk->major, disk->first_minor);
495 goto out_unregister_queue;
496 bdi_set_owner(disk->bdi, ddev);
497 ret = sysfs_create_link(&ddev->kobj,
498 &disk->bdi->dev->kobj, "bdi");
500 goto out_unregister_bdi;
502 bdev_add(disk->part0, ddev->devt);
503 disk_scan_partitions(disk);
506 * Announce the disk and partitions after all partitions are
507 * created. (for hidden disks uevents remain suppressed forever)
509 dev_set_uevent_suppress(ddev, 0);
510 disk_uevent(disk, KOBJ_ADD);
513 disk_update_readahead(disk);
514 disk_add_events(disk);
518 if (!(disk->flags & GENHD_FL_HIDDEN))
519 bdi_unregister(disk->bdi);
520 out_unregister_queue:
521 blk_unregister_queue(disk);
523 kobject_put(disk->slave_dir);
525 kobject_put(disk->part0->bd_holder_dir);
527 blk_integrity_del(disk);
529 if (!sysfs_deprecated)
530 sysfs_remove_link(block_depr, dev_name(ddev));
533 out_disk_release_events:
534 disk_release_events(disk);
536 if (disk->major == BLOCK_EXT_MAJOR)
537 blk_free_ext_minor(disk->first_minor);
538 return WARN_ON_ONCE(ret); /* keep until all callers handle errors */
540 EXPORT_SYMBOL(device_add_disk);
543 * del_gendisk - remove the gendisk
544 * @disk: the struct gendisk to remove
546 * Removes the gendisk and all its associated resources. This deletes the
547 * partitions associated with the gendisk, and unregisters the associated
550 * This is the counter to the respective __device_add_disk() call.
552 * The final removal of the struct gendisk happens when its refcount reaches 0
553 * with put_disk(), which should be called after del_gendisk(), if
554 * __device_add_disk() was used.
556 * Drivers exist which depend on the release of the gendisk to be synchronous,
557 * it should not be deferred.
561 void del_gendisk(struct gendisk *disk)
563 struct request_queue *q = disk->queue;
567 if (WARN_ON_ONCE(!disk_live(disk) && !(disk->flags & GENHD_FL_HIDDEN)))
570 blk_integrity_del(disk);
571 disk_del_events(disk);
573 mutex_lock(&disk->open_mutex);
574 remove_inode_hash(disk->part0->bd_inode);
575 blk_drop_partitions(disk);
576 mutex_unlock(&disk->open_mutex);
578 fsync_bdev(disk->part0);
579 __invalidate_device(disk->part0, true);
584 set_bit(GD_DEAD, &disk->state);
585 set_capacity(disk, 0);
588 * Prevent new I/O from crossing bio_queue_enter().
590 blk_queue_start_drain(q);
591 blk_mq_freeze_queue_wait(q);
595 blk_flush_integrity();
597 * Allow using passthrough request again after the queue is torn down.
599 blk_queue_flag_clear(QUEUE_FLAG_INIT_DONE, q);
600 __blk_mq_unfreeze_queue(q, true);
602 if (!(disk->flags & GENHD_FL_HIDDEN)) {
603 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
606 * Unregister bdi before releasing device numbers (as they can
607 * get reused and we'd get clashes in sysfs).
609 bdi_unregister(disk->bdi);
612 blk_unregister_queue(disk);
614 kobject_put(disk->part0->bd_holder_dir);
615 kobject_put(disk->slave_dir);
617 part_stat_set_all(disk->part0, 0);
618 disk->part0->bd_stamp = 0;
619 if (!sysfs_deprecated)
620 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
621 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
622 device_del(disk_to_dev(disk));
624 EXPORT_SYMBOL(del_gendisk);
626 /* sysfs access to bad-blocks list. */
627 static ssize_t disk_badblocks_show(struct device *dev,
628 struct device_attribute *attr,
631 struct gendisk *disk = dev_to_disk(dev);
634 return sprintf(page, "\n");
636 return badblocks_show(disk->bb, page, 0);
639 static ssize_t disk_badblocks_store(struct device *dev,
640 struct device_attribute *attr,
641 const char *page, size_t len)
643 struct gendisk *disk = dev_to_disk(dev);
648 return badblocks_store(disk->bb, page, len, 0);
651 void blk_request_module(dev_t devt)
653 unsigned int major = MAJOR(devt);
654 struct blk_major_name **n;
656 mutex_lock(&major_names_lock);
657 for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) {
658 if ((*n)->major == major && (*n)->probe) {
660 mutex_unlock(&major_names_lock);
664 mutex_unlock(&major_names_lock);
666 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
667 /* Make old-style 2.4 aliases work */
668 request_module("block-major-%d", MAJOR(devt));
672 * print a full list of all partitions - intended for places where the root
673 * filesystem can't be mounted and thus to give the victim some idea of what
676 void __init printk_all_partitions(void)
678 struct class_dev_iter iter;
681 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
682 while ((dev = class_dev_iter_next(&iter))) {
683 struct gendisk *disk = dev_to_disk(dev);
684 struct block_device *part;
685 char devt_buf[BDEVT_SIZE];
689 * Don't show empty devices or things that have been
692 if (get_capacity(disk) == 0 ||
693 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
697 * Note, unlike /proc/partitions, I am showing the numbers in
698 * hex - the same format as the root= option takes.
701 xa_for_each(&disk->part_tbl, idx, part) {
702 if (!bdev_nr_sectors(part))
704 printk("%s%s %10llu %pg %s",
705 bdev_is_partition(part) ? " " : "",
706 bdevt_str(part->bd_dev, devt_buf),
707 bdev_nr_sectors(part) >> 1, part,
709 part->bd_meta_info->uuid : "");
710 if (bdev_is_partition(part))
712 else if (dev->parent && dev->parent->driver)
713 printk(" driver: %s\n",
714 dev->parent->driver->name);
716 printk(" (driver?)\n");
720 class_dev_iter_exit(&iter);
723 #ifdef CONFIG_PROC_FS
725 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
728 struct class_dev_iter *iter;
731 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
733 return ERR_PTR(-ENOMEM);
735 seqf->private = iter;
736 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
738 dev = class_dev_iter_next(iter);
743 return dev_to_disk(dev);
746 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
751 dev = class_dev_iter_next(seqf->private);
753 return dev_to_disk(dev);
758 static void disk_seqf_stop(struct seq_file *seqf, void *v)
760 struct class_dev_iter *iter = seqf->private;
762 /* stop is called even after start failed :-( */
764 class_dev_iter_exit(iter);
766 seqf->private = NULL;
770 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
774 p = disk_seqf_start(seqf, pos);
775 if (!IS_ERR_OR_NULL(p) && !*pos)
776 seq_puts(seqf, "major minor #blocks name\n\n");
780 static int show_partition(struct seq_file *seqf, void *v)
782 struct gendisk *sgp = v;
783 struct block_device *part;
786 /* Don't show non-partitionable removeable devices or empty devices */
787 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
788 (sgp->flags & GENHD_FL_REMOVABLE)))
790 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
794 xa_for_each(&sgp->part_tbl, idx, part) {
795 if (!bdev_nr_sectors(part))
797 seq_printf(seqf, "%4d %7d %10llu %pg\n",
798 MAJOR(part->bd_dev), MINOR(part->bd_dev),
799 bdev_nr_sectors(part) >> 1, part);
805 static const struct seq_operations partitions_op = {
806 .start = show_partition_start,
807 .next = disk_seqf_next,
808 .stop = disk_seqf_stop,
809 .show = show_partition
813 static int __init genhd_device_init(void)
817 block_class.dev_kobj = sysfs_dev_block_kobj;
818 error = class_register(&block_class);
823 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
825 /* create top-level block dir */
826 if (!sysfs_deprecated)
827 block_depr = kobject_create_and_add("block", NULL);
831 subsys_initcall(genhd_device_init);
833 static ssize_t disk_range_show(struct device *dev,
834 struct device_attribute *attr, char *buf)
836 struct gendisk *disk = dev_to_disk(dev);
838 return sprintf(buf, "%d\n", disk->minors);
841 static ssize_t disk_ext_range_show(struct device *dev,
842 struct device_attribute *attr, char *buf)
844 struct gendisk *disk = dev_to_disk(dev);
846 return sprintf(buf, "%d\n", disk_max_parts(disk));
849 static ssize_t disk_removable_show(struct device *dev,
850 struct device_attribute *attr, char *buf)
852 struct gendisk *disk = dev_to_disk(dev);
854 return sprintf(buf, "%d\n",
855 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
858 static ssize_t disk_hidden_show(struct device *dev,
859 struct device_attribute *attr, char *buf)
861 struct gendisk *disk = dev_to_disk(dev);
863 return sprintf(buf, "%d\n",
864 (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
867 static ssize_t disk_ro_show(struct device *dev,
868 struct device_attribute *attr, char *buf)
870 struct gendisk *disk = dev_to_disk(dev);
872 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
875 ssize_t part_size_show(struct device *dev,
876 struct device_attribute *attr, char *buf)
878 return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev)));
881 ssize_t part_stat_show(struct device *dev,
882 struct device_attribute *attr, char *buf)
884 struct block_device *bdev = dev_to_bdev(dev);
885 struct request_queue *q = bdev->bd_disk->queue;
886 struct disk_stats stat;
887 unsigned int inflight;
889 part_stat_read_all(bdev, &stat);
891 inflight = blk_mq_in_flight(q, bdev);
893 inflight = part_in_flight(bdev);
896 "%8lu %8lu %8llu %8u "
897 "%8lu %8lu %8llu %8u "
899 "%8lu %8lu %8llu %8u "
903 stat.merges[STAT_READ],
904 (unsigned long long)stat.sectors[STAT_READ],
905 (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC),
906 stat.ios[STAT_WRITE],
907 stat.merges[STAT_WRITE],
908 (unsigned long long)stat.sectors[STAT_WRITE],
909 (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC),
911 jiffies_to_msecs(stat.io_ticks),
912 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
913 stat.nsecs[STAT_WRITE] +
914 stat.nsecs[STAT_DISCARD] +
915 stat.nsecs[STAT_FLUSH],
917 stat.ios[STAT_DISCARD],
918 stat.merges[STAT_DISCARD],
919 (unsigned long long)stat.sectors[STAT_DISCARD],
920 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC),
921 stat.ios[STAT_FLUSH],
922 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC));
925 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
928 struct block_device *bdev = dev_to_bdev(dev);
929 struct request_queue *q = bdev->bd_disk->queue;
930 unsigned int inflight[2];
933 blk_mq_in_flight_rw(q, bdev, inflight);
935 part_in_flight_rw(bdev, inflight);
937 return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
940 static ssize_t disk_capability_show(struct device *dev,
941 struct device_attribute *attr, char *buf)
943 struct gendisk *disk = dev_to_disk(dev);
945 return sprintf(buf, "%x\n", disk->flags);
948 static ssize_t disk_alignment_offset_show(struct device *dev,
949 struct device_attribute *attr,
952 struct gendisk *disk = dev_to_disk(dev);
954 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
957 static ssize_t disk_discard_alignment_show(struct device *dev,
958 struct device_attribute *attr,
961 struct gendisk *disk = dev_to_disk(dev);
963 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
966 static ssize_t diskseq_show(struct device *dev,
967 struct device_attribute *attr, char *buf)
969 struct gendisk *disk = dev_to_disk(dev);
971 return sprintf(buf, "%llu\n", disk->diskseq);
974 static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
975 static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
976 static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
977 static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
978 static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
979 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
980 static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
981 static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
982 static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
983 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
984 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
985 static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
986 static DEVICE_ATTR(diskseq, 0444, diskseq_show, NULL);
988 #ifdef CONFIG_FAIL_MAKE_REQUEST
989 ssize_t part_fail_show(struct device *dev,
990 struct device_attribute *attr, char *buf)
992 return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail);
995 ssize_t part_fail_store(struct device *dev,
996 struct device_attribute *attr,
997 const char *buf, size_t count)
1001 if (count > 0 && sscanf(buf, "%d", &i) > 0)
1002 dev_to_bdev(dev)->bd_make_it_fail = i;
1007 static struct device_attribute dev_attr_fail =
1008 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
1009 #endif /* CONFIG_FAIL_MAKE_REQUEST */
1011 #ifdef CONFIG_FAIL_IO_TIMEOUT
1012 static struct device_attribute dev_attr_fail_timeout =
1013 __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
1016 static struct attribute *disk_attrs[] = {
1017 &dev_attr_range.attr,
1018 &dev_attr_ext_range.attr,
1019 &dev_attr_removable.attr,
1020 &dev_attr_hidden.attr,
1022 &dev_attr_size.attr,
1023 &dev_attr_alignment_offset.attr,
1024 &dev_attr_discard_alignment.attr,
1025 &dev_attr_capability.attr,
1026 &dev_attr_stat.attr,
1027 &dev_attr_inflight.attr,
1028 &dev_attr_badblocks.attr,
1029 &dev_attr_events.attr,
1030 &dev_attr_events_async.attr,
1031 &dev_attr_events_poll_msecs.attr,
1032 &dev_attr_diskseq.attr,
1033 #ifdef CONFIG_FAIL_MAKE_REQUEST
1034 &dev_attr_fail.attr,
1036 #ifdef CONFIG_FAIL_IO_TIMEOUT
1037 &dev_attr_fail_timeout.attr,
1042 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1044 struct device *dev = container_of(kobj, typeof(*dev), kobj);
1045 struct gendisk *disk = dev_to_disk(dev);
1047 if (a == &dev_attr_badblocks.attr && !disk->bb)
1052 static struct attribute_group disk_attr_group = {
1053 .attrs = disk_attrs,
1054 .is_visible = disk_visible,
1057 static const struct attribute_group *disk_attr_groups[] = {
1063 * disk_release - releases all allocated resources of the gendisk
1064 * @dev: the device representing this disk
1066 * This function releases all allocated resources of the gendisk.
1068 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1069 * assigned. Since the request_queue sits on top of the gendisk for these
1070 * drivers we also call blk_put_queue() for them, and we expect the
1071 * request_queue refcount to reach 0 at this point, and so the request_queue
1072 * will also be freed prior to the disk.
1074 * Context: can sleep
1076 static void disk_release(struct device *dev)
1078 struct gendisk *disk = dev_to_disk(dev);
1081 WARN_ON_ONCE(disk_live(disk));
1083 disk_release_events(disk);
1084 kfree(disk->random);
1085 xa_destroy(&disk->part_tbl);
1086 disk->queue->disk = NULL;
1087 blk_put_queue(disk->queue);
1088 iput(disk->part0->bd_inode); /* frees the disk */
1091 static int block_uevent(struct device *dev, struct kobj_uevent_env *env)
1093 struct gendisk *disk = dev_to_disk(dev);
1095 return add_uevent_var(env, "DISKSEQ=%llu", disk->diskseq);
1098 struct class block_class = {
1100 .dev_uevent = block_uevent,
1103 static char *block_devnode(struct device *dev, umode_t *mode,
1104 kuid_t *uid, kgid_t *gid)
1106 struct gendisk *disk = dev_to_disk(dev);
1108 if (disk->fops->devnode)
1109 return disk->fops->devnode(disk, mode);
1113 const struct device_type disk_type = {
1115 .groups = disk_attr_groups,
1116 .release = disk_release,
1117 .devnode = block_devnode,
1120 #ifdef CONFIG_PROC_FS
1122 * aggregate disk stat collector. Uses the same stats that the sysfs
1123 * entries do, above, but makes them available through one seq_file.
1125 * The output looks suspiciously like /proc/partitions with a bunch of
1128 static int diskstats_show(struct seq_file *seqf, void *v)
1130 struct gendisk *gp = v;
1131 struct block_device *hd;
1132 unsigned int inflight;
1133 struct disk_stats stat;
1137 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1138 seq_puts(seqf, "major minor name"
1139 " rio rmerge rsect ruse wio wmerge "
1140 "wsect wuse running use aveq"
1145 xa_for_each(&gp->part_tbl, idx, hd) {
1146 if (bdev_is_partition(hd) && !bdev_nr_sectors(hd))
1148 part_stat_read_all(hd, &stat);
1149 if (queue_is_mq(gp->queue))
1150 inflight = blk_mq_in_flight(gp->queue, hd);
1152 inflight = part_in_flight(hd);
1154 seq_printf(seqf, "%4d %7d %pg "
1161 MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd,
1162 stat.ios[STAT_READ],
1163 stat.merges[STAT_READ],
1164 stat.sectors[STAT_READ],
1165 (unsigned int)div_u64(stat.nsecs[STAT_READ],
1167 stat.ios[STAT_WRITE],
1168 stat.merges[STAT_WRITE],
1169 stat.sectors[STAT_WRITE],
1170 (unsigned int)div_u64(stat.nsecs[STAT_WRITE],
1173 jiffies_to_msecs(stat.io_ticks),
1174 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1175 stat.nsecs[STAT_WRITE] +
1176 stat.nsecs[STAT_DISCARD] +
1177 stat.nsecs[STAT_FLUSH],
1179 stat.ios[STAT_DISCARD],
1180 stat.merges[STAT_DISCARD],
1181 stat.sectors[STAT_DISCARD],
1182 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD],
1184 stat.ios[STAT_FLUSH],
1185 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH],
1194 static const struct seq_operations diskstats_op = {
1195 .start = disk_seqf_start,
1196 .next = disk_seqf_next,
1197 .stop = disk_seqf_stop,
1198 .show = diskstats_show
1201 static int __init proc_genhd_init(void)
1203 proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1204 proc_create_seq("partitions", 0, NULL, &partitions_op);
1207 module_init(proc_genhd_init);
1208 #endif /* CONFIG_PROC_FS */
1210 dev_t part_devt(struct gendisk *disk, u8 partno)
1212 struct block_device *part;
1216 part = xa_load(&disk->part_tbl, partno);
1218 devt = part->bd_dev;
1224 dev_t blk_lookup_devt(const char *name, int partno)
1226 dev_t devt = MKDEV(0, 0);
1227 struct class_dev_iter iter;
1230 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1231 while ((dev = class_dev_iter_next(&iter))) {
1232 struct gendisk *disk = dev_to_disk(dev);
1234 if (strcmp(dev_name(dev), name))
1237 if (partno < disk->minors) {
1238 /* We need to return the right devno, even
1239 * if the partition doesn't exist yet.
1241 devt = MKDEV(MAJOR(dev->devt),
1242 MINOR(dev->devt) + partno);
1244 devt = part_devt(disk, partno);
1249 class_dev_iter_exit(&iter);
1253 struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id,
1254 struct lock_class_key *lkclass)
1256 struct gendisk *disk;
1258 if (!blk_get_queue(q))
1261 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1265 disk->bdi = bdi_alloc(node_id);
1269 disk->part0 = bdev_alloc(disk, 0);
1273 disk->node_id = node_id;
1274 mutex_init(&disk->open_mutex);
1275 xa_init(&disk->part_tbl);
1276 if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL))
1277 goto out_destroy_part_tbl;
1279 rand_initialize_disk(disk);
1280 disk_to_dev(disk)->class = &block_class;
1281 disk_to_dev(disk)->type = &disk_type;
1282 device_initialize(disk_to_dev(disk));
1286 lockdep_init_map(&disk->lockdep_map, "(bio completion)", lkclass, 0);
1287 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
1288 INIT_LIST_HEAD(&disk->slave_bdevs);
1292 out_destroy_part_tbl:
1293 xa_destroy(&disk->part_tbl);
1294 disk->part0->bd_disk = NULL;
1295 iput(disk->part0->bd_inode);
1304 EXPORT_SYMBOL(__alloc_disk_node);
1306 struct gendisk *__blk_alloc_disk(int node, struct lock_class_key *lkclass)
1308 struct request_queue *q;
1309 struct gendisk *disk;
1311 q = blk_alloc_queue(node);
1315 disk = __alloc_disk_node(q, node, lkclass);
1317 blk_cleanup_queue(q);
1322 EXPORT_SYMBOL(__blk_alloc_disk);
1325 * put_disk - decrements the gendisk refcount
1326 * @disk: the struct gendisk to decrement the refcount for
1328 * This decrements the refcount for the struct gendisk. When this reaches 0
1329 * we'll have disk_release() called.
1331 * Context: Any context, but the last reference must not be dropped from
1334 void put_disk(struct gendisk *disk)
1337 put_device(disk_to_dev(disk));
1339 EXPORT_SYMBOL(put_disk);
1342 * blk_cleanup_disk - shutdown a gendisk allocated by blk_alloc_disk
1343 * @disk: gendisk to shutdown
1345 * Mark the queue hanging off @disk DYING, drain all pending requests, then mark
1346 * the queue DEAD, destroy and put it and the gendisk structure.
1348 * Context: can sleep
1350 void blk_cleanup_disk(struct gendisk *disk)
1352 blk_cleanup_queue(disk->queue);
1355 EXPORT_SYMBOL(blk_cleanup_disk);
1357 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1359 char event[] = "DISK_RO=1";
1360 char *envp[] = { event, NULL };
1364 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1368 * set_disk_ro - set a gendisk read-only
1369 * @disk: gendisk to operate on
1370 * @read_only: %true to set the disk read-only, %false set the disk read/write
1372 * This function is used to indicate whether a given disk device should have its
1373 * read-only flag set. set_disk_ro() is typically used by device drivers to
1374 * indicate whether the underlying physical device is write-protected.
1376 void set_disk_ro(struct gendisk *disk, bool read_only)
1379 if (test_and_set_bit(GD_READ_ONLY, &disk->state))
1382 if (!test_and_clear_bit(GD_READ_ONLY, &disk->state))
1385 set_disk_ro_uevent(disk, read_only);
1387 EXPORT_SYMBOL(set_disk_ro);
1389 int bdev_read_only(struct block_device *bdev)
1391 return bdev->bd_read_only || get_disk_ro(bdev->bd_disk);
1393 EXPORT_SYMBOL(bdev_read_only);
1395 void inc_diskseq(struct gendisk *disk)
1397 disk->diskseq = atomic64_inc_return(&diskseq);