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;
32 /* for extended dynamic devt allocation, currently only one major is used */
33 #define NR_EXT_DEVT (1 << MINORBITS)
34 static DEFINE_IDA(ext_devt_ida);
36 void set_capacity(struct gendisk *disk, sector_t sectors)
38 struct block_device *bdev = disk->part0;
40 spin_lock(&bdev->bd_size_lock);
41 i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
42 spin_unlock(&bdev->bd_size_lock);
44 EXPORT_SYMBOL(set_capacity);
47 * Set disk capacity and notify if the size is not currently zero and will not
48 * be set to zero. Returns true if a uevent was sent, otherwise false.
50 bool set_capacity_and_notify(struct gendisk *disk, sector_t size)
52 sector_t capacity = get_capacity(disk);
53 char *envp[] = { "RESIZE=1", NULL };
55 set_capacity(disk, size);
58 * Only print a message and send a uevent if the gendisk is user visible
59 * and alive. This avoids spamming the log and udev when setting the
60 * initial capacity during probing.
62 if (size == capacity ||
63 (disk->flags & (GENHD_FL_UP | GENHD_FL_HIDDEN)) != GENHD_FL_UP)
66 pr_info("%s: detected capacity change from %lld to %lld\n",
67 disk->disk_name, capacity, size);
70 * Historically we did not send a uevent for changes to/from an empty
73 if (!capacity || !size)
75 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
78 EXPORT_SYMBOL_GPL(set_capacity_and_notify);
81 * Format the device name of the indicated disk into the supplied buffer and
82 * return a pointer to that same buffer for convenience.
84 char *disk_name(struct gendisk *hd, int partno, char *buf)
87 snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name);
88 else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1]))
89 snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno);
91 snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno);
96 const char *bdevname(struct block_device *bdev, char *buf)
98 return disk_name(bdev->bd_disk, bdev->bd_partno, buf);
100 EXPORT_SYMBOL(bdevname);
102 static void part_stat_read_all(struct block_device *part,
103 struct disk_stats *stat)
107 memset(stat, 0, sizeof(struct disk_stats));
108 for_each_possible_cpu(cpu) {
109 struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu);
112 for (group = 0; group < NR_STAT_GROUPS; group++) {
113 stat->nsecs[group] += ptr->nsecs[group];
114 stat->sectors[group] += ptr->sectors[group];
115 stat->ios[group] += ptr->ios[group];
116 stat->merges[group] += ptr->merges[group];
119 stat->io_ticks += ptr->io_ticks;
123 static unsigned int part_in_flight(struct block_device *part)
125 unsigned int inflight = 0;
128 for_each_possible_cpu(cpu) {
129 inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) +
130 part_stat_local_read_cpu(part, in_flight[1], cpu);
132 if ((int)inflight < 0)
138 static void part_in_flight_rw(struct block_device *part,
139 unsigned int inflight[2])
145 for_each_possible_cpu(cpu) {
146 inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu);
147 inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu);
149 if ((int)inflight[0] < 0)
151 if ((int)inflight[1] < 0)
156 * Can be deleted altogether. Later.
159 #define BLKDEV_MAJOR_HASH_SIZE 255
160 static struct blk_major_name {
161 struct blk_major_name *next;
164 void (*probe)(dev_t devt);
165 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
166 static DEFINE_MUTEX(major_names_lock);
168 /* index in the above - for now: assume no multimajor ranges */
169 static inline int major_to_index(unsigned major)
171 return major % BLKDEV_MAJOR_HASH_SIZE;
174 #ifdef CONFIG_PROC_FS
175 void blkdev_show(struct seq_file *seqf, off_t offset)
177 struct blk_major_name *dp;
179 mutex_lock(&major_names_lock);
180 for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
181 if (dp->major == offset)
182 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
183 mutex_unlock(&major_names_lock);
185 #endif /* CONFIG_PROC_FS */
188 * __register_blkdev - register a new block device
190 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
191 * @major = 0, try to allocate any unused major number.
192 * @name: the name of the new block device as a zero terminated string
193 * @probe: allback that is called on access to any minor number of @major
195 * The @name must be unique within the system.
197 * The return value depends on the @major input parameter:
199 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
200 * then the function returns zero on success, or a negative error code
201 * - if any unused major number was requested with @major = 0 parameter
202 * then the return value is the allocated major number in range
203 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
205 * See Documentation/admin-guide/devices.txt for the list of allocated
208 * Use register_blkdev instead for any new code.
210 int __register_blkdev(unsigned int major, const char *name,
211 void (*probe)(dev_t devt))
213 struct blk_major_name **n, *p;
216 mutex_lock(&major_names_lock);
220 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
221 if (major_names[index] == NULL)
226 printk("%s: failed to get major for %s\n",
235 if (major >= BLKDEV_MAJOR_MAX) {
236 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
237 __func__, major, BLKDEV_MAJOR_MAX-1, name);
243 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
251 strlcpy(p->name, name, sizeof(p->name));
253 index = major_to_index(major);
255 for (n = &major_names[index]; *n; n = &(*n)->next) {
256 if ((*n)->major == major)
265 printk("register_blkdev: cannot get major %u for %s\n",
270 mutex_unlock(&major_names_lock);
273 EXPORT_SYMBOL(__register_blkdev);
275 void unregister_blkdev(unsigned int major, const char *name)
277 struct blk_major_name **n;
278 struct blk_major_name *p = NULL;
279 int index = major_to_index(major);
281 mutex_lock(&major_names_lock);
282 for (n = &major_names[index]; *n; n = &(*n)->next)
283 if ((*n)->major == major)
285 if (!*n || strcmp((*n)->name, name)) {
291 mutex_unlock(&major_names_lock);
295 EXPORT_SYMBOL(unregister_blkdev);
298 * blk_mangle_minor - scatter minor numbers apart
299 * @minor: minor number to mangle
301 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
302 * is enabled. Mangling twice gives the original value.
310 static int blk_mangle_minor(int minor)
312 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
315 for (i = 0; i < MINORBITS / 2; i++) {
316 int low = minor & (1 << i);
317 int high = minor & (1 << (MINORBITS - 1 - i));
318 int distance = MINORBITS - 1 - 2 * i;
320 minor ^= low | high; /* clear both bits */
321 low <<= distance; /* swap the positions */
323 minor |= low | high; /* and set */
329 int blk_alloc_ext_minor(void)
333 idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT, GFP_KERNEL);
339 return blk_mangle_minor(idx);
342 void blk_free_ext_minor(unsigned int minor)
344 ida_free(&ext_devt_ida, blk_mangle_minor(minor));
347 static char *bdevt_str(dev_t devt, char *buf)
349 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
350 char tbuf[BDEVT_SIZE];
351 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
352 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
354 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
359 void disk_uevent(struct gendisk *disk, enum kobject_action action)
361 struct block_device *part;
365 xa_for_each(&disk->part_tbl, idx, part) {
366 if (bdev_is_partition(part) && !bdev_nr_sectors(part))
368 if (!kobject_get_unless_zero(&part->bd_device.kobj))
372 kobject_uevent(bdev_kobj(part), action);
373 put_device(&part->bd_device);
378 EXPORT_SYMBOL_GPL(disk_uevent);
380 static void disk_scan_partitions(struct gendisk *disk)
382 struct block_device *bdev;
384 if (!get_capacity(disk) || !disk_part_scan_enabled(disk))
387 set_bit(GD_NEED_PART_SCAN, &disk->state);
388 bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL);
390 blkdev_put(bdev, FMODE_READ);
393 static void register_disk(struct device *parent, struct gendisk *disk,
394 const struct attribute_group **groups)
396 struct device *ddev = disk_to_dev(disk);
399 ddev->parent = parent;
401 dev_set_name(ddev, "%s", disk->disk_name);
403 /* delay uevents, until we scanned partition table */
404 dev_set_uevent_suppress(ddev, 1);
407 WARN_ON(ddev->groups);
408 ddev->groups = groups;
410 if (device_add(ddev))
412 if (!sysfs_deprecated) {
413 err = sysfs_create_link(block_depr, &ddev->kobj,
414 kobject_name(&ddev->kobj));
422 * avoid probable deadlock caused by allocating memory with
423 * GFP_KERNEL in runtime_resume callback of its all ancestor
426 pm_runtime_set_memalloc_noio(ddev, true);
428 disk->part0->bd_holder_dir =
429 kobject_create_and_add("holders", &ddev->kobj);
430 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
432 if (disk->flags & GENHD_FL_HIDDEN)
435 disk_scan_partitions(disk);
437 /* announce the disk and partitions after all partitions are created */
438 dev_set_uevent_suppress(ddev, 0);
439 disk_uevent(disk, KOBJ_ADD);
441 if (disk->queue->backing_dev_info->dev) {
442 err = sysfs_create_link(&ddev->kobj,
443 &disk->queue->backing_dev_info->dev->kobj,
450 * __device_add_disk - add disk information to kernel list
451 * @parent: parent device for the disk
452 * @disk: per-device partitioning information
453 * @groups: Additional per-device sysfs groups
454 * @register_queue: register the queue if set to true
456 * This function registers the partitioning information in @disk
459 * FIXME: error handling
461 static void __device_add_disk(struct device *parent, struct gendisk *disk,
462 const struct attribute_group **groups,
468 * The disk queue should now be all set with enough information about
469 * the device for the elevator code to pick an adequate default
470 * elevator if one is needed, that is, for devices requesting queue
474 elevator_init_mq(disk->queue);
477 * If the driver provides an explicit major number it also must provide
478 * the number of minors numbers supported, and those will be used to
480 * Otherwise just allocate the device numbers for both the whole device
481 * and all partitions from the extended dev_t space.
484 WARN_ON(!disk->minors);
486 if (disk->minors > DISK_MAX_PARTS) {
487 pr_err("block: can't allocate more than %d partitions\n",
489 disk->minors = DISK_MAX_PARTS;
492 WARN_ON(disk->minors);
494 ret = blk_alloc_ext_minor();
499 disk->major = BLOCK_EXT_MAJOR;
500 disk->first_minor = MINOR(ret);
501 disk->flags |= GENHD_FL_EXT_DEVT;
504 disk->flags |= GENHD_FL_UP;
506 disk_alloc_events(disk);
508 if (disk->flags & GENHD_FL_HIDDEN) {
510 * Don't let hidden disks show up in /proc/partitions,
511 * and don't bother scanning for partitions either.
513 disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
514 disk->flags |= GENHD_FL_NO_PART_SCAN;
516 struct backing_dev_info *bdi = disk->queue->backing_dev_info;
517 struct device *dev = disk_to_dev(disk);
519 /* Register BDI before referencing it from bdev */
520 dev->devt = MKDEV(disk->major, disk->first_minor);
521 ret = bdi_register(bdi, "%u:%u",
522 disk->major, disk->first_minor);
524 bdi_set_owner(bdi, dev);
525 bdev_add(disk->part0, dev->devt);
527 register_disk(parent, disk, groups);
529 blk_register_queue(disk);
532 * Take an extra ref on queue which will be put on disk_release()
533 * so that it sticks around as long as @disk is there.
535 if (blk_get_queue(disk->queue))
536 set_bit(GD_QUEUE_REF, &disk->state);
540 disk_add_events(disk);
541 blk_integrity_add(disk);
544 void device_add_disk(struct device *parent, struct gendisk *disk,
545 const struct attribute_group **groups)
548 __device_add_disk(parent, disk, groups, true);
550 EXPORT_SYMBOL(device_add_disk);
552 void device_add_disk_no_queue_reg(struct device *parent, struct gendisk *disk)
554 __device_add_disk(parent, disk, NULL, false);
556 EXPORT_SYMBOL(device_add_disk_no_queue_reg);
559 * del_gendisk - remove the gendisk
560 * @disk: the struct gendisk to remove
562 * Removes the gendisk and all its associated resources. This deletes the
563 * partitions associated with the gendisk, and unregisters the associated
566 * This is the counter to the respective __device_add_disk() call.
568 * The final removal of the struct gendisk happens when its refcount reaches 0
569 * with put_disk(), which should be called after del_gendisk(), if
570 * __device_add_disk() was used.
572 * Drivers exist which depend on the release of the gendisk to be synchronous,
573 * it should not be deferred.
577 void del_gendisk(struct gendisk *disk)
581 if (WARN_ON_ONCE(!disk->queue))
584 blk_integrity_del(disk);
585 disk_del_events(disk);
587 mutex_lock(&disk->open_mutex);
588 remove_inode_hash(disk->part0->bd_inode);
589 disk->flags &= ~GENHD_FL_UP;
590 blk_drop_partitions(disk);
591 mutex_unlock(&disk->open_mutex);
593 fsync_bdev(disk->part0);
594 __invalidate_device(disk->part0, true);
596 set_capacity(disk, 0);
598 if (!(disk->flags & GENHD_FL_HIDDEN)) {
599 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
602 * Unregister bdi before releasing device numbers (as they can
603 * get reused and we'd get clashes in sysfs).
605 bdi_unregister(disk->queue->backing_dev_info);
608 blk_unregister_queue(disk);
610 kobject_put(disk->part0->bd_holder_dir);
611 kobject_put(disk->slave_dir);
613 part_stat_set_all(disk->part0, 0);
614 disk->part0->bd_stamp = 0;
615 if (!sysfs_deprecated)
616 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
617 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
618 device_del(disk_to_dev(disk));
620 EXPORT_SYMBOL(del_gendisk);
622 /* sysfs access to bad-blocks list. */
623 static ssize_t disk_badblocks_show(struct device *dev,
624 struct device_attribute *attr,
627 struct gendisk *disk = dev_to_disk(dev);
630 return sprintf(page, "\n");
632 return badblocks_show(disk->bb, page, 0);
635 static ssize_t disk_badblocks_store(struct device *dev,
636 struct device_attribute *attr,
637 const char *page, size_t len)
639 struct gendisk *disk = dev_to_disk(dev);
644 return badblocks_store(disk->bb, page, len, 0);
647 void blk_request_module(dev_t devt)
649 unsigned int major = MAJOR(devt);
650 struct blk_major_name **n;
652 mutex_lock(&major_names_lock);
653 for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) {
654 if ((*n)->major == major && (*n)->probe) {
656 mutex_unlock(&major_names_lock);
660 mutex_unlock(&major_names_lock);
662 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
663 /* Make old-style 2.4 aliases work */
664 request_module("block-major-%d", MAJOR(devt));
668 * print a full list of all partitions - intended for places where the root
669 * filesystem can't be mounted and thus to give the victim some idea of what
672 void __init printk_all_partitions(void)
674 struct class_dev_iter iter;
677 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
678 while ((dev = class_dev_iter_next(&iter))) {
679 struct gendisk *disk = dev_to_disk(dev);
680 struct block_device *part;
681 char devt_buf[BDEVT_SIZE];
685 * Don't show empty devices or things that have been
688 if (get_capacity(disk) == 0 ||
689 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
693 * Note, unlike /proc/partitions, I am showing the numbers in
694 * hex - the same format as the root= option takes.
697 xa_for_each(&disk->part_tbl, idx, part) {
698 if (!bdev_nr_sectors(part))
700 printk("%s%s %10llu %pg %s",
701 bdev_is_partition(part) ? " " : "",
702 bdevt_str(part->bd_dev, devt_buf),
703 bdev_nr_sectors(part) >> 1, part,
705 part->bd_meta_info->uuid : "");
706 if (bdev_is_partition(part))
708 else if (dev->parent && dev->parent->driver)
709 printk(" driver: %s\n",
710 dev->parent->driver->name);
712 printk(" (driver?)\n");
716 class_dev_iter_exit(&iter);
719 #ifdef CONFIG_PROC_FS
721 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
724 struct class_dev_iter *iter;
727 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
729 return ERR_PTR(-ENOMEM);
731 seqf->private = iter;
732 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
734 dev = class_dev_iter_next(iter);
739 return dev_to_disk(dev);
742 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
747 dev = class_dev_iter_next(seqf->private);
749 return dev_to_disk(dev);
754 static void disk_seqf_stop(struct seq_file *seqf, void *v)
756 struct class_dev_iter *iter = seqf->private;
758 /* stop is called even after start failed :-( */
760 class_dev_iter_exit(iter);
762 seqf->private = NULL;
766 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
770 p = disk_seqf_start(seqf, pos);
771 if (!IS_ERR_OR_NULL(p) && !*pos)
772 seq_puts(seqf, "major minor #blocks name\n\n");
776 static int show_partition(struct seq_file *seqf, void *v)
778 struct gendisk *sgp = v;
779 struct block_device *part;
782 /* Don't show non-partitionable removeable devices or empty devices */
783 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
784 (sgp->flags & GENHD_FL_REMOVABLE)))
786 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
790 xa_for_each(&sgp->part_tbl, idx, part) {
791 if (!bdev_nr_sectors(part))
793 seq_printf(seqf, "%4d %7d %10llu %pg\n",
794 MAJOR(part->bd_dev), MINOR(part->bd_dev),
795 bdev_nr_sectors(part) >> 1, part);
801 static const struct seq_operations partitions_op = {
802 .start = show_partition_start,
803 .next = disk_seqf_next,
804 .stop = disk_seqf_stop,
805 .show = show_partition
809 static int __init genhd_device_init(void)
813 block_class.dev_kobj = sysfs_dev_block_kobj;
814 error = class_register(&block_class);
819 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
821 /* create top-level block dir */
822 if (!sysfs_deprecated)
823 block_depr = kobject_create_and_add("block", NULL);
827 subsys_initcall(genhd_device_init);
829 static ssize_t disk_range_show(struct device *dev,
830 struct device_attribute *attr, char *buf)
832 struct gendisk *disk = dev_to_disk(dev);
834 return sprintf(buf, "%d\n", disk->minors);
837 static ssize_t disk_ext_range_show(struct device *dev,
838 struct device_attribute *attr, char *buf)
840 struct gendisk *disk = dev_to_disk(dev);
842 return sprintf(buf, "%d\n", disk_max_parts(disk));
845 static ssize_t disk_removable_show(struct device *dev,
846 struct device_attribute *attr, char *buf)
848 struct gendisk *disk = dev_to_disk(dev);
850 return sprintf(buf, "%d\n",
851 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
854 static ssize_t disk_hidden_show(struct device *dev,
855 struct device_attribute *attr, char *buf)
857 struct gendisk *disk = dev_to_disk(dev);
859 return sprintf(buf, "%d\n",
860 (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
863 static ssize_t disk_ro_show(struct device *dev,
864 struct device_attribute *attr, char *buf)
866 struct gendisk *disk = dev_to_disk(dev);
868 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
871 ssize_t part_size_show(struct device *dev,
872 struct device_attribute *attr, char *buf)
874 return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev)));
877 ssize_t part_stat_show(struct device *dev,
878 struct device_attribute *attr, char *buf)
880 struct block_device *bdev = dev_to_bdev(dev);
881 struct request_queue *q = bdev->bd_disk->queue;
882 struct disk_stats stat;
883 unsigned int inflight;
885 part_stat_read_all(bdev, &stat);
887 inflight = blk_mq_in_flight(q, bdev);
889 inflight = part_in_flight(bdev);
892 "%8lu %8lu %8llu %8u "
893 "%8lu %8lu %8llu %8u "
895 "%8lu %8lu %8llu %8u "
899 stat.merges[STAT_READ],
900 (unsigned long long)stat.sectors[STAT_READ],
901 (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC),
902 stat.ios[STAT_WRITE],
903 stat.merges[STAT_WRITE],
904 (unsigned long long)stat.sectors[STAT_WRITE],
905 (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC),
907 jiffies_to_msecs(stat.io_ticks),
908 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
909 stat.nsecs[STAT_WRITE] +
910 stat.nsecs[STAT_DISCARD] +
911 stat.nsecs[STAT_FLUSH],
913 stat.ios[STAT_DISCARD],
914 stat.merges[STAT_DISCARD],
915 (unsigned long long)stat.sectors[STAT_DISCARD],
916 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC),
917 stat.ios[STAT_FLUSH],
918 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC));
921 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
924 struct block_device *bdev = dev_to_bdev(dev);
925 struct request_queue *q = bdev->bd_disk->queue;
926 unsigned int inflight[2];
929 blk_mq_in_flight_rw(q, bdev, inflight);
931 part_in_flight_rw(bdev, inflight);
933 return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
936 static ssize_t disk_capability_show(struct device *dev,
937 struct device_attribute *attr, char *buf)
939 struct gendisk *disk = dev_to_disk(dev);
941 return sprintf(buf, "%x\n", disk->flags);
944 static ssize_t disk_alignment_offset_show(struct device *dev,
945 struct device_attribute *attr,
948 struct gendisk *disk = dev_to_disk(dev);
950 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
953 static ssize_t disk_discard_alignment_show(struct device *dev,
954 struct device_attribute *attr,
957 struct gendisk *disk = dev_to_disk(dev);
959 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
962 static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
963 static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
964 static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
965 static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
966 static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
967 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
968 static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
969 static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
970 static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
971 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
972 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
973 static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
975 #ifdef CONFIG_FAIL_MAKE_REQUEST
976 ssize_t part_fail_show(struct device *dev,
977 struct device_attribute *attr, char *buf)
979 return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail);
982 ssize_t part_fail_store(struct device *dev,
983 struct device_attribute *attr,
984 const char *buf, size_t count)
988 if (count > 0 && sscanf(buf, "%d", &i) > 0)
989 dev_to_bdev(dev)->bd_make_it_fail = i;
994 static struct device_attribute dev_attr_fail =
995 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
996 #endif /* CONFIG_FAIL_MAKE_REQUEST */
998 #ifdef CONFIG_FAIL_IO_TIMEOUT
999 static struct device_attribute dev_attr_fail_timeout =
1000 __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
1003 static struct attribute *disk_attrs[] = {
1004 &dev_attr_range.attr,
1005 &dev_attr_ext_range.attr,
1006 &dev_attr_removable.attr,
1007 &dev_attr_hidden.attr,
1009 &dev_attr_size.attr,
1010 &dev_attr_alignment_offset.attr,
1011 &dev_attr_discard_alignment.attr,
1012 &dev_attr_capability.attr,
1013 &dev_attr_stat.attr,
1014 &dev_attr_inflight.attr,
1015 &dev_attr_badblocks.attr,
1016 &dev_attr_events.attr,
1017 &dev_attr_events_async.attr,
1018 &dev_attr_events_poll_msecs.attr,
1019 #ifdef CONFIG_FAIL_MAKE_REQUEST
1020 &dev_attr_fail.attr,
1022 #ifdef CONFIG_FAIL_IO_TIMEOUT
1023 &dev_attr_fail_timeout.attr,
1028 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1030 struct device *dev = container_of(kobj, typeof(*dev), kobj);
1031 struct gendisk *disk = dev_to_disk(dev);
1033 if (a == &dev_attr_badblocks.attr && !disk->bb)
1038 static struct attribute_group disk_attr_group = {
1039 .attrs = disk_attrs,
1040 .is_visible = disk_visible,
1043 static const struct attribute_group *disk_attr_groups[] = {
1049 * disk_release - releases all allocated resources of the gendisk
1050 * @dev: the device representing this disk
1052 * This function releases all allocated resources of the gendisk.
1054 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1055 * assigned. Since the request_queue sits on top of the gendisk for these
1056 * drivers we also call blk_put_queue() for them, and we expect the
1057 * request_queue refcount to reach 0 at this point, and so the request_queue
1058 * will also be freed prior to the disk.
1060 * Context: can sleep
1062 static void disk_release(struct device *dev)
1064 struct gendisk *disk = dev_to_disk(dev);
1068 if (MAJOR(dev->devt) == BLOCK_EXT_MAJOR)
1069 blk_free_ext_minor(MINOR(dev->devt));
1070 disk_release_events(disk);
1071 kfree(disk->random);
1072 xa_destroy(&disk->part_tbl);
1073 if (test_bit(GD_QUEUE_REF, &disk->state) && disk->queue)
1074 blk_put_queue(disk->queue);
1075 iput(disk->part0->bd_inode); /* frees the disk */
1077 struct class block_class = {
1081 static char *block_devnode(struct device *dev, umode_t *mode,
1082 kuid_t *uid, kgid_t *gid)
1084 struct gendisk *disk = dev_to_disk(dev);
1086 if (disk->fops->devnode)
1087 return disk->fops->devnode(disk, mode);
1091 const struct device_type disk_type = {
1093 .groups = disk_attr_groups,
1094 .release = disk_release,
1095 .devnode = block_devnode,
1098 #ifdef CONFIG_PROC_FS
1100 * aggregate disk stat collector. Uses the same stats that the sysfs
1101 * entries do, above, but makes them available through one seq_file.
1103 * The output looks suspiciously like /proc/partitions with a bunch of
1106 static int diskstats_show(struct seq_file *seqf, void *v)
1108 struct gendisk *gp = v;
1109 struct block_device *hd;
1110 unsigned int inflight;
1111 struct disk_stats stat;
1115 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1116 seq_puts(seqf, "major minor name"
1117 " rio rmerge rsect ruse wio wmerge "
1118 "wsect wuse running use aveq"
1123 xa_for_each(&gp->part_tbl, idx, hd) {
1124 if (bdev_is_partition(hd) && !bdev_nr_sectors(hd))
1126 part_stat_read_all(hd, &stat);
1127 if (queue_is_mq(gp->queue))
1128 inflight = blk_mq_in_flight(gp->queue, hd);
1130 inflight = part_in_flight(hd);
1132 seq_printf(seqf, "%4d %7d %pg "
1139 MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd,
1140 stat.ios[STAT_READ],
1141 stat.merges[STAT_READ],
1142 stat.sectors[STAT_READ],
1143 (unsigned int)div_u64(stat.nsecs[STAT_READ],
1145 stat.ios[STAT_WRITE],
1146 stat.merges[STAT_WRITE],
1147 stat.sectors[STAT_WRITE],
1148 (unsigned int)div_u64(stat.nsecs[STAT_WRITE],
1151 jiffies_to_msecs(stat.io_ticks),
1152 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1153 stat.nsecs[STAT_WRITE] +
1154 stat.nsecs[STAT_DISCARD] +
1155 stat.nsecs[STAT_FLUSH],
1157 stat.ios[STAT_DISCARD],
1158 stat.merges[STAT_DISCARD],
1159 stat.sectors[STAT_DISCARD],
1160 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD],
1162 stat.ios[STAT_FLUSH],
1163 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH],
1172 static const struct seq_operations diskstats_op = {
1173 .start = disk_seqf_start,
1174 .next = disk_seqf_next,
1175 .stop = disk_seqf_stop,
1176 .show = diskstats_show
1179 static int __init proc_genhd_init(void)
1181 proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1182 proc_create_seq("partitions", 0, NULL, &partitions_op);
1185 module_init(proc_genhd_init);
1186 #endif /* CONFIG_PROC_FS */
1188 dev_t part_devt(struct gendisk *disk, u8 partno)
1190 struct block_device *part;
1194 part = xa_load(&disk->part_tbl, partno);
1196 devt = part->bd_dev;
1202 dev_t blk_lookup_devt(const char *name, int partno)
1204 dev_t devt = MKDEV(0, 0);
1205 struct class_dev_iter iter;
1208 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1209 while ((dev = class_dev_iter_next(&iter))) {
1210 struct gendisk *disk = dev_to_disk(dev);
1212 if (strcmp(dev_name(dev), name))
1215 if (partno < disk->minors) {
1216 /* We need to return the right devno, even
1217 * if the partition doesn't exist yet.
1219 devt = MKDEV(MAJOR(dev->devt),
1220 MINOR(dev->devt) + partno);
1222 devt = part_devt(disk, partno);
1227 class_dev_iter_exit(&iter);
1231 struct gendisk *__alloc_disk_node(int minors, int node_id)
1233 struct gendisk *disk;
1235 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1239 disk->part0 = bdev_alloc(disk, 0);
1243 disk->node_id = node_id;
1244 mutex_init(&disk->open_mutex);
1245 xa_init(&disk->part_tbl);
1246 if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL))
1247 goto out_destroy_part_tbl;
1249 disk->minors = minors;
1250 rand_initialize_disk(disk);
1251 disk_to_dev(disk)->class = &block_class;
1252 disk_to_dev(disk)->type = &disk_type;
1253 device_initialize(disk_to_dev(disk));
1256 out_destroy_part_tbl:
1257 xa_destroy(&disk->part_tbl);
1258 iput(disk->part0->bd_inode);
1263 EXPORT_SYMBOL(__alloc_disk_node);
1265 struct gendisk *__blk_alloc_disk(int node)
1267 struct request_queue *q;
1268 struct gendisk *disk;
1270 q = blk_alloc_queue(node);
1274 disk = __alloc_disk_node(0, node);
1276 blk_cleanup_queue(q);
1282 EXPORT_SYMBOL(__blk_alloc_disk);
1285 * put_disk - decrements the gendisk refcount
1286 * @disk: the struct gendisk to decrement the refcount for
1288 * This decrements the refcount for the struct gendisk. When this reaches 0
1289 * we'll have disk_release() called.
1291 * Context: Any context, but the last reference must not be dropped from
1294 void put_disk(struct gendisk *disk)
1297 put_device(disk_to_dev(disk));
1299 EXPORT_SYMBOL(put_disk);
1302 * blk_cleanup_disk - shutdown a gendisk allocated by blk_alloc_disk
1303 * @disk: gendisk to shutdown
1305 * Mark the queue hanging off @disk DYING, drain all pending requests, then mark
1306 * the queue DEAD, destroy and put it and the gendisk structure.
1308 * Context: can sleep
1310 void blk_cleanup_disk(struct gendisk *disk)
1312 blk_cleanup_queue(disk->queue);
1315 EXPORT_SYMBOL(blk_cleanup_disk);
1317 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1319 char event[] = "DISK_RO=1";
1320 char *envp[] = { event, NULL };
1324 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1328 * set_disk_ro - set a gendisk read-only
1329 * @disk: gendisk to operate on
1330 * @read_only: %true to set the disk read-only, %false set the disk read/write
1332 * This function is used to indicate whether a given disk device should have its
1333 * read-only flag set. set_disk_ro() is typically used by device drivers to
1334 * indicate whether the underlying physical device is write-protected.
1336 void set_disk_ro(struct gendisk *disk, bool read_only)
1339 if (test_and_set_bit(GD_READ_ONLY, &disk->state))
1342 if (!test_and_clear_bit(GD_READ_ONLY, &disk->state))
1345 set_disk_ro_uevent(disk, read_only);
1347 EXPORT_SYMBOL(set_disk_ro);
1349 int bdev_read_only(struct block_device *bdev)
1351 return bdev->bd_read_only || get_disk_ro(bdev->bd_disk);
1353 EXPORT_SYMBOL(bdev_read_only);