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);
317 * blk_mangle_minor - scatter minor numbers apart
318 * @minor: minor number to mangle
320 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
321 * is enabled. Mangling twice gives the original value.
329 static int blk_mangle_minor(int minor)
331 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
334 for (i = 0; i < MINORBITS / 2; i++) {
335 int low = minor & (1 << i);
336 int high = minor & (1 << (MINORBITS - 1 - i));
337 int distance = MINORBITS - 1 - 2 * i;
339 minor ^= low | high; /* clear both bits */
340 low <<= distance; /* swap the positions */
342 minor |= low | high; /* and set */
348 int blk_alloc_ext_minor(void)
352 idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT, GFP_KERNEL);
358 return blk_mangle_minor(idx);
361 void blk_free_ext_minor(unsigned int minor)
363 ida_free(&ext_devt_ida, blk_mangle_minor(minor));
366 static char *bdevt_str(dev_t devt, char *buf)
368 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
369 char tbuf[BDEVT_SIZE];
370 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
371 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
373 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
378 void disk_uevent(struct gendisk *disk, enum kobject_action action)
380 struct block_device *part;
384 xa_for_each(&disk->part_tbl, idx, part) {
385 if (bdev_is_partition(part) && !bdev_nr_sectors(part))
387 if (!kobject_get_unless_zero(&part->bd_device.kobj))
391 kobject_uevent(bdev_kobj(part), action);
392 put_device(&part->bd_device);
397 EXPORT_SYMBOL_GPL(disk_uevent);
399 static void disk_scan_partitions(struct gendisk *disk)
401 struct block_device *bdev;
403 if (!get_capacity(disk) || !disk_part_scan_enabled(disk))
406 set_bit(GD_NEED_PART_SCAN, &disk->state);
407 bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL);
409 blkdev_put(bdev, FMODE_READ);
413 * device_add_disk - add disk information to kernel list
414 * @parent: parent device for the disk
415 * @disk: per-device partitioning information
416 * @groups: Additional per-device sysfs groups
418 * This function registers the partitioning information in @disk
421 int device_add_disk(struct device *parent, struct gendisk *disk,
422 const struct attribute_group **groups)
425 struct device *ddev = disk_to_dev(disk);
429 * The disk queue should now be all set with enough information about
430 * the device for the elevator code to pick an adequate default
431 * elevator if one is needed, that is, for devices requesting queue
434 elevator_init_mq(disk->queue);
437 * If the driver provides an explicit major number it also must provide
438 * the number of minors numbers supported, and those will be used to
440 * Otherwise just allocate the device numbers for both the whole device
441 * and all partitions from the extended dev_t space.
444 if (WARN_ON(!disk->minors))
447 if (disk->minors > DISK_MAX_PARTS) {
448 pr_err("block: can't allocate more than %d partitions\n",
450 disk->minors = DISK_MAX_PARTS;
453 if (WARN_ON(disk->minors))
456 ret = blk_alloc_ext_minor();
459 disk->major = BLOCK_EXT_MAJOR;
460 disk->first_minor = MINOR(ret);
461 disk->flags |= GENHD_FL_EXT_DEVT;
464 ret = disk_alloc_events(disk);
466 goto out_free_ext_minor;
468 /* delay uevents, until we scanned partition table */
469 dev_set_uevent_suppress(ddev, 1);
471 ddev->parent = parent;
472 ddev->groups = groups;
473 dev_set_name(ddev, "%s", disk->disk_name);
474 if (!(disk->flags & GENHD_FL_HIDDEN))
475 ddev->devt = MKDEV(disk->major, disk->first_minor);
476 ret = device_add(ddev);
478 goto out_disk_release_events;
479 if (!sysfs_deprecated) {
480 ret = sysfs_create_link(block_depr, &ddev->kobj,
481 kobject_name(&ddev->kobj));
487 * avoid probable deadlock caused by allocating memory with
488 * GFP_KERNEL in runtime_resume callback of its all ancestor
491 pm_runtime_set_memalloc_noio(ddev, true);
493 ret = blk_integrity_add(disk);
495 goto out_del_block_link;
497 disk->part0->bd_holder_dir =
498 kobject_create_and_add("holders", &ddev->kobj);
499 if (!disk->part0->bd_holder_dir)
500 goto out_del_integrity;
501 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
502 if (!disk->slave_dir)
503 goto out_put_holder_dir;
505 ret = bd_register_pending_holders(disk);
507 goto out_put_slave_dir;
509 ret = blk_register_queue(disk);
511 goto out_put_slave_dir;
513 if (disk->flags & GENHD_FL_HIDDEN) {
515 * Don't let hidden disks show up in /proc/partitions,
516 * and don't bother scanning for partitions either.
518 disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
519 disk->flags |= GENHD_FL_NO_PART_SCAN;
521 ret = bdi_register(disk->bdi, "%u:%u",
522 disk->major, disk->first_minor);
524 goto out_unregister_queue;
525 bdi_set_owner(disk->bdi, ddev);
526 ret = sysfs_create_link(&ddev->kobj,
527 &disk->bdi->dev->kobj, "bdi");
529 goto out_unregister_bdi;
531 bdev_add(disk->part0, ddev->devt);
532 disk_scan_partitions(disk);
535 * Announce the disk and partitions after all partitions are
536 * created. (for hidden disks uevents remain suppressed forever)
538 dev_set_uevent_suppress(ddev, 0);
539 disk_uevent(disk, KOBJ_ADD);
542 disk_update_readahead(disk);
543 disk_add_events(disk);
547 if (!(disk->flags & GENHD_FL_HIDDEN))
548 bdi_unregister(disk->bdi);
549 out_unregister_queue:
550 blk_unregister_queue(disk);
552 kobject_put(disk->slave_dir);
554 kobject_put(disk->part0->bd_holder_dir);
556 blk_integrity_del(disk);
558 if (!sysfs_deprecated)
559 sysfs_remove_link(block_depr, dev_name(ddev));
562 out_disk_release_events:
563 disk_release_events(disk);
565 if (disk->major == BLOCK_EXT_MAJOR)
566 blk_free_ext_minor(disk->first_minor);
567 return WARN_ON_ONCE(ret); /* keep until all callers handle errors */
569 EXPORT_SYMBOL(device_add_disk);
572 * del_gendisk - remove the gendisk
573 * @disk: the struct gendisk to remove
575 * Removes the gendisk and all its associated resources. This deletes the
576 * partitions associated with the gendisk, and unregisters the associated
579 * This is the counter to the respective __device_add_disk() call.
581 * The final removal of the struct gendisk happens when its refcount reaches 0
582 * with put_disk(), which should be called after del_gendisk(), if
583 * __device_add_disk() was used.
585 * Drivers exist which depend on the release of the gendisk to be synchronous,
586 * it should not be deferred.
590 void del_gendisk(struct gendisk *disk)
594 if (WARN_ON_ONCE(!disk_live(disk)))
597 blk_integrity_del(disk);
598 disk_del_events(disk);
600 mutex_lock(&disk->open_mutex);
601 remove_inode_hash(disk->part0->bd_inode);
602 blk_drop_partitions(disk);
603 mutex_unlock(&disk->open_mutex);
605 fsync_bdev(disk->part0);
606 __invalidate_device(disk->part0, true);
608 set_capacity(disk, 0);
610 if (!(disk->flags & GENHD_FL_HIDDEN)) {
611 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
614 * Unregister bdi before releasing device numbers (as they can
615 * get reused and we'd get clashes in sysfs).
617 bdi_unregister(disk->bdi);
620 blk_unregister_queue(disk);
622 kobject_put(disk->part0->bd_holder_dir);
623 kobject_put(disk->slave_dir);
625 part_stat_set_all(disk->part0, 0);
626 disk->part0->bd_stamp = 0;
627 if (!sysfs_deprecated)
628 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
629 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
630 device_del(disk_to_dev(disk));
632 EXPORT_SYMBOL(del_gendisk);
634 /* sysfs access to bad-blocks list. */
635 static ssize_t disk_badblocks_show(struct device *dev,
636 struct device_attribute *attr,
639 struct gendisk *disk = dev_to_disk(dev);
642 return sprintf(page, "\n");
644 return badblocks_show(disk->bb, page, 0);
647 static ssize_t disk_badblocks_store(struct device *dev,
648 struct device_attribute *attr,
649 const char *page, size_t len)
651 struct gendisk *disk = dev_to_disk(dev);
656 return badblocks_store(disk->bb, page, len, 0);
659 void blk_request_module(dev_t devt)
661 unsigned int major = MAJOR(devt);
662 struct blk_major_name **n;
664 mutex_lock(&major_names_lock);
665 for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) {
666 if ((*n)->major == major && (*n)->probe) {
668 mutex_unlock(&major_names_lock);
672 mutex_unlock(&major_names_lock);
674 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
675 /* Make old-style 2.4 aliases work */
676 request_module("block-major-%d", MAJOR(devt));
680 * print a full list of all partitions - intended for places where the root
681 * filesystem can't be mounted and thus to give the victim some idea of what
684 void __init printk_all_partitions(void)
686 struct class_dev_iter iter;
689 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
690 while ((dev = class_dev_iter_next(&iter))) {
691 struct gendisk *disk = dev_to_disk(dev);
692 struct block_device *part;
693 char devt_buf[BDEVT_SIZE];
697 * Don't show empty devices or things that have been
700 if (get_capacity(disk) == 0 ||
701 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
705 * Note, unlike /proc/partitions, I am showing the numbers in
706 * hex - the same format as the root= option takes.
709 xa_for_each(&disk->part_tbl, idx, part) {
710 if (!bdev_nr_sectors(part))
712 printk("%s%s %10llu %pg %s",
713 bdev_is_partition(part) ? " " : "",
714 bdevt_str(part->bd_dev, devt_buf),
715 bdev_nr_sectors(part) >> 1, part,
717 part->bd_meta_info->uuid : "");
718 if (bdev_is_partition(part))
720 else if (dev->parent && dev->parent->driver)
721 printk(" driver: %s\n",
722 dev->parent->driver->name);
724 printk(" (driver?)\n");
728 class_dev_iter_exit(&iter);
731 #ifdef CONFIG_PROC_FS
733 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
736 struct class_dev_iter *iter;
739 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
741 return ERR_PTR(-ENOMEM);
743 seqf->private = iter;
744 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
746 dev = class_dev_iter_next(iter);
751 return dev_to_disk(dev);
754 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
759 dev = class_dev_iter_next(seqf->private);
761 return dev_to_disk(dev);
766 static void disk_seqf_stop(struct seq_file *seqf, void *v)
768 struct class_dev_iter *iter = seqf->private;
770 /* stop is called even after start failed :-( */
772 class_dev_iter_exit(iter);
774 seqf->private = NULL;
778 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
782 p = disk_seqf_start(seqf, pos);
783 if (!IS_ERR_OR_NULL(p) && !*pos)
784 seq_puts(seqf, "major minor #blocks name\n\n");
788 static int show_partition(struct seq_file *seqf, void *v)
790 struct gendisk *sgp = v;
791 struct block_device *part;
794 /* Don't show non-partitionable removeable devices or empty devices */
795 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
796 (sgp->flags & GENHD_FL_REMOVABLE)))
798 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
802 xa_for_each(&sgp->part_tbl, idx, part) {
803 if (!bdev_nr_sectors(part))
805 seq_printf(seqf, "%4d %7d %10llu %pg\n",
806 MAJOR(part->bd_dev), MINOR(part->bd_dev),
807 bdev_nr_sectors(part) >> 1, part);
813 static const struct seq_operations partitions_op = {
814 .start = show_partition_start,
815 .next = disk_seqf_next,
816 .stop = disk_seqf_stop,
817 .show = show_partition
821 static int __init genhd_device_init(void)
825 block_class.dev_kobj = sysfs_dev_block_kobj;
826 error = class_register(&block_class);
831 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
833 /* create top-level block dir */
834 if (!sysfs_deprecated)
835 block_depr = kobject_create_and_add("block", NULL);
839 subsys_initcall(genhd_device_init);
841 static ssize_t disk_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->minors);
849 static ssize_t disk_ext_range_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", disk_max_parts(disk));
857 static ssize_t disk_removable_show(struct device *dev,
858 struct device_attribute *attr, char *buf)
860 struct gendisk *disk = dev_to_disk(dev);
862 return sprintf(buf, "%d\n",
863 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
866 static ssize_t disk_hidden_show(struct device *dev,
867 struct device_attribute *attr, char *buf)
869 struct gendisk *disk = dev_to_disk(dev);
871 return sprintf(buf, "%d\n",
872 (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
875 static ssize_t disk_ro_show(struct device *dev,
876 struct device_attribute *attr, char *buf)
878 struct gendisk *disk = dev_to_disk(dev);
880 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
883 ssize_t part_size_show(struct device *dev,
884 struct device_attribute *attr, char *buf)
886 return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev)));
889 ssize_t part_stat_show(struct device *dev,
890 struct device_attribute *attr, char *buf)
892 struct block_device *bdev = dev_to_bdev(dev);
893 struct request_queue *q = bdev->bd_disk->queue;
894 struct disk_stats stat;
895 unsigned int inflight;
897 part_stat_read_all(bdev, &stat);
899 inflight = blk_mq_in_flight(q, bdev);
901 inflight = part_in_flight(bdev);
904 "%8lu %8lu %8llu %8u "
905 "%8lu %8lu %8llu %8u "
907 "%8lu %8lu %8llu %8u "
911 stat.merges[STAT_READ],
912 (unsigned long long)stat.sectors[STAT_READ],
913 (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC),
914 stat.ios[STAT_WRITE],
915 stat.merges[STAT_WRITE],
916 (unsigned long long)stat.sectors[STAT_WRITE],
917 (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC),
919 jiffies_to_msecs(stat.io_ticks),
920 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
921 stat.nsecs[STAT_WRITE] +
922 stat.nsecs[STAT_DISCARD] +
923 stat.nsecs[STAT_FLUSH],
925 stat.ios[STAT_DISCARD],
926 stat.merges[STAT_DISCARD],
927 (unsigned long long)stat.sectors[STAT_DISCARD],
928 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC),
929 stat.ios[STAT_FLUSH],
930 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC));
933 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
936 struct block_device *bdev = dev_to_bdev(dev);
937 struct request_queue *q = bdev->bd_disk->queue;
938 unsigned int inflight[2];
941 blk_mq_in_flight_rw(q, bdev, inflight);
943 part_in_flight_rw(bdev, inflight);
945 return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
948 static ssize_t disk_capability_show(struct device *dev,
949 struct device_attribute *attr, char *buf)
951 struct gendisk *disk = dev_to_disk(dev);
953 return sprintf(buf, "%x\n", disk->flags);
956 static ssize_t disk_alignment_offset_show(struct device *dev,
957 struct device_attribute *attr,
960 struct gendisk *disk = dev_to_disk(dev);
962 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
965 static ssize_t disk_discard_alignment_show(struct device *dev,
966 struct device_attribute *attr,
969 struct gendisk *disk = dev_to_disk(dev);
971 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
974 static ssize_t diskseq_show(struct device *dev,
975 struct device_attribute *attr, char *buf)
977 struct gendisk *disk = dev_to_disk(dev);
979 return sprintf(buf, "%llu\n", disk->diskseq);
982 static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
983 static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
984 static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
985 static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
986 static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
987 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
988 static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
989 static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
990 static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
991 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
992 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
993 static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
994 static DEVICE_ATTR(diskseq, 0444, diskseq_show, NULL);
996 #ifdef CONFIG_FAIL_MAKE_REQUEST
997 ssize_t part_fail_show(struct device *dev,
998 struct device_attribute *attr, char *buf)
1000 return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail);
1003 ssize_t part_fail_store(struct device *dev,
1004 struct device_attribute *attr,
1005 const char *buf, size_t count)
1009 if (count > 0 && sscanf(buf, "%d", &i) > 0)
1010 dev_to_bdev(dev)->bd_make_it_fail = i;
1015 static struct device_attribute dev_attr_fail =
1016 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
1017 #endif /* CONFIG_FAIL_MAKE_REQUEST */
1019 #ifdef CONFIG_FAIL_IO_TIMEOUT
1020 static struct device_attribute dev_attr_fail_timeout =
1021 __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
1024 static struct attribute *disk_attrs[] = {
1025 &dev_attr_range.attr,
1026 &dev_attr_ext_range.attr,
1027 &dev_attr_removable.attr,
1028 &dev_attr_hidden.attr,
1030 &dev_attr_size.attr,
1031 &dev_attr_alignment_offset.attr,
1032 &dev_attr_discard_alignment.attr,
1033 &dev_attr_capability.attr,
1034 &dev_attr_stat.attr,
1035 &dev_attr_inflight.attr,
1036 &dev_attr_badblocks.attr,
1037 &dev_attr_events.attr,
1038 &dev_attr_events_async.attr,
1039 &dev_attr_events_poll_msecs.attr,
1040 &dev_attr_diskseq.attr,
1041 #ifdef CONFIG_FAIL_MAKE_REQUEST
1042 &dev_attr_fail.attr,
1044 #ifdef CONFIG_FAIL_IO_TIMEOUT
1045 &dev_attr_fail_timeout.attr,
1050 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1052 struct device *dev = container_of(kobj, typeof(*dev), kobj);
1053 struct gendisk *disk = dev_to_disk(dev);
1055 if (a == &dev_attr_badblocks.attr && !disk->bb)
1060 static struct attribute_group disk_attr_group = {
1061 .attrs = disk_attrs,
1062 .is_visible = disk_visible,
1065 static const struct attribute_group *disk_attr_groups[] = {
1071 * disk_release - releases all allocated resources of the gendisk
1072 * @dev: the device representing this disk
1074 * This function releases all allocated resources of the gendisk.
1076 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1077 * assigned. Since the request_queue sits on top of the gendisk for these
1078 * drivers we also call blk_put_queue() for them, and we expect the
1079 * request_queue refcount to reach 0 at this point, and so the request_queue
1080 * will also be freed prior to the disk.
1082 * Context: can sleep
1084 static void disk_release(struct device *dev)
1086 struct gendisk *disk = dev_to_disk(dev);
1090 disk_release_events(disk);
1091 kfree(disk->random);
1092 xa_destroy(&disk->part_tbl);
1093 disk->queue->disk = NULL;
1094 blk_put_queue(disk->queue);
1095 iput(disk->part0->bd_inode); /* frees the disk */
1098 static int block_uevent(struct device *dev, struct kobj_uevent_env *env)
1100 struct gendisk *disk = dev_to_disk(dev);
1102 return add_uevent_var(env, "DISKSEQ=%llu", disk->diskseq);
1105 struct class block_class = {
1107 .dev_uevent = block_uevent,
1110 static char *block_devnode(struct device *dev, umode_t *mode,
1111 kuid_t *uid, kgid_t *gid)
1113 struct gendisk *disk = dev_to_disk(dev);
1115 if (disk->fops->devnode)
1116 return disk->fops->devnode(disk, mode);
1120 const struct device_type disk_type = {
1122 .groups = disk_attr_groups,
1123 .release = disk_release,
1124 .devnode = block_devnode,
1127 #ifdef CONFIG_PROC_FS
1129 * aggregate disk stat collector. Uses the same stats that the sysfs
1130 * entries do, above, but makes them available through one seq_file.
1132 * The output looks suspiciously like /proc/partitions with a bunch of
1135 static int diskstats_show(struct seq_file *seqf, void *v)
1137 struct gendisk *gp = v;
1138 struct block_device *hd;
1139 unsigned int inflight;
1140 struct disk_stats stat;
1144 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1145 seq_puts(seqf, "major minor name"
1146 " rio rmerge rsect ruse wio wmerge "
1147 "wsect wuse running use aveq"
1152 xa_for_each(&gp->part_tbl, idx, hd) {
1153 if (bdev_is_partition(hd) && !bdev_nr_sectors(hd))
1155 part_stat_read_all(hd, &stat);
1156 if (queue_is_mq(gp->queue))
1157 inflight = blk_mq_in_flight(gp->queue, hd);
1159 inflight = part_in_flight(hd);
1161 seq_printf(seqf, "%4d %7d %pg "
1168 MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd,
1169 stat.ios[STAT_READ],
1170 stat.merges[STAT_READ],
1171 stat.sectors[STAT_READ],
1172 (unsigned int)div_u64(stat.nsecs[STAT_READ],
1174 stat.ios[STAT_WRITE],
1175 stat.merges[STAT_WRITE],
1176 stat.sectors[STAT_WRITE],
1177 (unsigned int)div_u64(stat.nsecs[STAT_WRITE],
1180 jiffies_to_msecs(stat.io_ticks),
1181 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1182 stat.nsecs[STAT_WRITE] +
1183 stat.nsecs[STAT_DISCARD] +
1184 stat.nsecs[STAT_FLUSH],
1186 stat.ios[STAT_DISCARD],
1187 stat.merges[STAT_DISCARD],
1188 stat.sectors[STAT_DISCARD],
1189 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD],
1191 stat.ios[STAT_FLUSH],
1192 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH],
1201 static const struct seq_operations diskstats_op = {
1202 .start = disk_seqf_start,
1203 .next = disk_seqf_next,
1204 .stop = disk_seqf_stop,
1205 .show = diskstats_show
1208 static int __init proc_genhd_init(void)
1210 proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1211 proc_create_seq("partitions", 0, NULL, &partitions_op);
1214 module_init(proc_genhd_init);
1215 #endif /* CONFIG_PROC_FS */
1217 dev_t part_devt(struct gendisk *disk, u8 partno)
1219 struct block_device *part;
1223 part = xa_load(&disk->part_tbl, partno);
1225 devt = part->bd_dev;
1231 dev_t blk_lookup_devt(const char *name, int partno)
1233 dev_t devt = MKDEV(0, 0);
1234 struct class_dev_iter iter;
1237 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1238 while ((dev = class_dev_iter_next(&iter))) {
1239 struct gendisk *disk = dev_to_disk(dev);
1241 if (strcmp(dev_name(dev), name))
1244 if (partno < disk->minors) {
1245 /* We need to return the right devno, even
1246 * if the partition doesn't exist yet.
1248 devt = MKDEV(MAJOR(dev->devt),
1249 MINOR(dev->devt) + partno);
1251 devt = part_devt(disk, partno);
1256 class_dev_iter_exit(&iter);
1260 struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id,
1261 struct lock_class_key *lkclass)
1263 struct gendisk *disk;
1265 if (!blk_get_queue(q))
1268 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1272 disk->bdi = bdi_alloc(node_id);
1276 disk->part0 = bdev_alloc(disk, 0);
1280 disk->node_id = node_id;
1281 mutex_init(&disk->open_mutex);
1282 xa_init(&disk->part_tbl);
1283 if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL))
1284 goto out_destroy_part_tbl;
1286 rand_initialize_disk(disk);
1287 disk_to_dev(disk)->class = &block_class;
1288 disk_to_dev(disk)->type = &disk_type;
1289 device_initialize(disk_to_dev(disk));
1293 lockdep_init_map(&disk->lockdep_map, "(bio completion)", lkclass, 0);
1294 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
1295 INIT_LIST_HEAD(&disk->slave_bdevs);
1299 out_destroy_part_tbl:
1300 xa_destroy(&disk->part_tbl);
1301 iput(disk->part0->bd_inode);
1310 EXPORT_SYMBOL(__alloc_disk_node);
1312 struct gendisk *__blk_alloc_disk(int node, struct lock_class_key *lkclass)
1314 struct request_queue *q;
1315 struct gendisk *disk;
1317 q = blk_alloc_queue(node);
1321 disk = __alloc_disk_node(q, node, lkclass);
1323 blk_cleanup_queue(q);
1328 EXPORT_SYMBOL(__blk_alloc_disk);
1331 * put_disk - decrements the gendisk refcount
1332 * @disk: the struct gendisk to decrement the refcount for
1334 * This decrements the refcount for the struct gendisk. When this reaches 0
1335 * we'll have disk_release() called.
1337 * Context: Any context, but the last reference must not be dropped from
1340 void put_disk(struct gendisk *disk)
1343 put_device(disk_to_dev(disk));
1345 EXPORT_SYMBOL(put_disk);
1348 * blk_cleanup_disk - shutdown a gendisk allocated by blk_alloc_disk
1349 * @disk: gendisk to shutdown
1351 * Mark the queue hanging off @disk DYING, drain all pending requests, then mark
1352 * the queue DEAD, destroy and put it and the gendisk structure.
1354 * Context: can sleep
1356 void blk_cleanup_disk(struct gendisk *disk)
1358 blk_cleanup_queue(disk->queue);
1361 EXPORT_SYMBOL(blk_cleanup_disk);
1363 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1365 char event[] = "DISK_RO=1";
1366 char *envp[] = { event, NULL };
1370 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1374 * set_disk_ro - set a gendisk read-only
1375 * @disk: gendisk to operate on
1376 * @read_only: %true to set the disk read-only, %false set the disk read/write
1378 * This function is used to indicate whether a given disk device should have its
1379 * read-only flag set. set_disk_ro() is typically used by device drivers to
1380 * indicate whether the underlying physical device is write-protected.
1382 void set_disk_ro(struct gendisk *disk, bool read_only)
1385 if (test_and_set_bit(GD_READ_ONLY, &disk->state))
1388 if (!test_and_clear_bit(GD_READ_ONLY, &disk->state))
1391 set_disk_ro_uevent(disk, read_only);
1393 EXPORT_SYMBOL(set_disk_ro);
1395 int bdev_read_only(struct block_device *bdev)
1397 return bdev->bd_read_only || get_disk_ro(bdev->bd_disk);
1399 EXPORT_SYMBOL(bdev_read_only);
1401 void inc_diskseq(struct gendisk *disk)
1403 disk->diskseq = atomic64_inc_return(&diskseq);