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 * FIXME: error handling
424 void device_add_disk(struct device *parent, struct gendisk *disk,
425 const struct attribute_group **groups)
428 struct device *ddev = disk_to_dev(disk);
432 * The disk queue should now be all set with enough information about
433 * the device for the elevator code to pick an adequate default
434 * elevator if one is needed, that is, for devices requesting queue
437 elevator_init_mq(disk->queue);
440 * If the driver provides an explicit major number it also must provide
441 * the number of minors numbers supported, and those will be used to
443 * Otherwise just allocate the device numbers for both the whole device
444 * and all partitions from the extended dev_t space.
447 WARN_ON(!disk->minors);
449 if (disk->minors > DISK_MAX_PARTS) {
450 pr_err("block: can't allocate more than %d partitions\n",
452 disk->minors = DISK_MAX_PARTS;
455 WARN_ON(disk->minors);
457 ret = blk_alloc_ext_minor();
462 disk->major = BLOCK_EXT_MAJOR;
463 disk->first_minor = MINOR(ret);
464 disk->flags |= GENHD_FL_EXT_DEVT;
467 disk_alloc_events(disk);
469 if (disk->flags & GENHD_FL_HIDDEN) {
471 * Don't let hidden disks show up in /proc/partitions,
472 * and don't bother scanning for partitions either.
474 disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
475 disk->flags |= GENHD_FL_NO_PART_SCAN;
477 /* Register BDI before referencing it from bdev */
478 ddev->devt = MKDEV(disk->major, disk->first_minor);
479 ret = bdi_register(disk->bdi, "%u:%u",
480 disk->major, disk->first_minor);
482 bdi_set_owner(disk->bdi, ddev);
483 bdev_add(disk->part0, ddev->devt);
486 /* delay uevents, until we scanned partition table */
487 dev_set_uevent_suppress(ddev, 1);
489 ddev->parent = parent;
490 ddev->groups = groups;
491 dev_set_name(ddev, "%s", disk->disk_name);
492 if (device_add(ddev))
494 if (!sysfs_deprecated) {
495 ret = sysfs_create_link(block_depr, &ddev->kobj,
496 kobject_name(&ddev->kobj));
504 * avoid probable deadlock caused by allocating memory with
505 * GFP_KERNEL in runtime_resume callback of its all ancestor
508 pm_runtime_set_memalloc_noio(ddev, true);
510 disk->part0->bd_holder_dir =
511 kobject_create_and_add("holders", &ddev->kobj);
512 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
515 * XXX: this is a mess, can't wait for real error handling in add_disk.
516 * Make sure ->slave_dir is NULL if we failed some of the registration
517 * so that the cleanup in bd_unlink_disk_holder works properly.
519 if (bd_register_pending_holders(disk) < 0) {
520 kobject_put(disk->slave_dir);
521 disk->slave_dir = NULL;
524 if (!(disk->flags & GENHD_FL_HIDDEN)) {
525 disk_scan_partitions(disk);
528 * Announce the disk and partitions after all partitions are
531 dev_set_uevent_suppress(ddev, 0);
532 disk_uevent(disk, KOBJ_ADD);
534 if (disk->bdi->dev) {
535 ret = sysfs_create_link(&ddev->kobj,
536 &disk->bdi->dev->kobj, "bdi");
541 blk_register_queue(disk);
543 disk_add_events(disk);
544 blk_integrity_add(disk);
546 EXPORT_SYMBOL(device_add_disk);
549 * del_gendisk - remove the gendisk
550 * @disk: the struct gendisk to remove
552 * Removes the gendisk and all its associated resources. This deletes the
553 * partitions associated with the gendisk, and unregisters the associated
556 * This is the counter to the respective __device_add_disk() call.
558 * The final removal of the struct gendisk happens when its refcount reaches 0
559 * with put_disk(), which should be called after del_gendisk(), if
560 * __device_add_disk() was used.
562 * Drivers exist which depend on the release of the gendisk to be synchronous,
563 * it should not be deferred.
567 void del_gendisk(struct gendisk *disk)
571 if (WARN_ON_ONCE(!disk_live(disk)))
574 blk_integrity_del(disk);
575 disk_del_events(disk);
577 mutex_lock(&disk->open_mutex);
578 remove_inode_hash(disk->part0->bd_inode);
579 blk_drop_partitions(disk);
580 mutex_unlock(&disk->open_mutex);
582 fsync_bdev(disk->part0);
583 __invalidate_device(disk->part0, true);
585 set_capacity(disk, 0);
587 if (!(disk->flags & GENHD_FL_HIDDEN)) {
588 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
591 * Unregister bdi before releasing device numbers (as they can
592 * get reused and we'd get clashes in sysfs).
594 bdi_unregister(disk->bdi);
597 blk_unregister_queue(disk);
599 kobject_put(disk->part0->bd_holder_dir);
600 kobject_put(disk->slave_dir);
602 part_stat_set_all(disk->part0, 0);
603 disk->part0->bd_stamp = 0;
604 if (!sysfs_deprecated)
605 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
606 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
607 device_del(disk_to_dev(disk));
609 EXPORT_SYMBOL(del_gendisk);
611 /* sysfs access to bad-blocks list. */
612 static ssize_t disk_badblocks_show(struct device *dev,
613 struct device_attribute *attr,
616 struct gendisk *disk = dev_to_disk(dev);
619 return sprintf(page, "\n");
621 return badblocks_show(disk->bb, page, 0);
624 static ssize_t disk_badblocks_store(struct device *dev,
625 struct device_attribute *attr,
626 const char *page, size_t len)
628 struct gendisk *disk = dev_to_disk(dev);
633 return badblocks_store(disk->bb, page, len, 0);
636 void blk_request_module(dev_t devt)
638 unsigned int major = MAJOR(devt);
639 struct blk_major_name **n;
641 mutex_lock(&major_names_lock);
642 for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) {
643 if ((*n)->major == major && (*n)->probe) {
645 mutex_unlock(&major_names_lock);
649 mutex_unlock(&major_names_lock);
651 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
652 /* Make old-style 2.4 aliases work */
653 request_module("block-major-%d", MAJOR(devt));
657 * print a full list of all partitions - intended for places where the root
658 * filesystem can't be mounted and thus to give the victim some idea of what
661 void __init printk_all_partitions(void)
663 struct class_dev_iter iter;
666 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
667 while ((dev = class_dev_iter_next(&iter))) {
668 struct gendisk *disk = dev_to_disk(dev);
669 struct block_device *part;
670 char devt_buf[BDEVT_SIZE];
674 * Don't show empty devices or things that have been
677 if (get_capacity(disk) == 0 ||
678 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
682 * Note, unlike /proc/partitions, I am showing the numbers in
683 * hex - the same format as the root= option takes.
686 xa_for_each(&disk->part_tbl, idx, part) {
687 if (!bdev_nr_sectors(part))
689 printk("%s%s %10llu %pg %s",
690 bdev_is_partition(part) ? " " : "",
691 bdevt_str(part->bd_dev, devt_buf),
692 bdev_nr_sectors(part) >> 1, part,
694 part->bd_meta_info->uuid : "");
695 if (bdev_is_partition(part))
697 else if (dev->parent && dev->parent->driver)
698 printk(" driver: %s\n",
699 dev->parent->driver->name);
701 printk(" (driver?)\n");
705 class_dev_iter_exit(&iter);
708 #ifdef CONFIG_PROC_FS
710 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
713 struct class_dev_iter *iter;
716 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
718 return ERR_PTR(-ENOMEM);
720 seqf->private = iter;
721 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
723 dev = class_dev_iter_next(iter);
728 return dev_to_disk(dev);
731 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
736 dev = class_dev_iter_next(seqf->private);
738 return dev_to_disk(dev);
743 static void disk_seqf_stop(struct seq_file *seqf, void *v)
745 struct class_dev_iter *iter = seqf->private;
747 /* stop is called even after start failed :-( */
749 class_dev_iter_exit(iter);
751 seqf->private = NULL;
755 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
759 p = disk_seqf_start(seqf, pos);
760 if (!IS_ERR_OR_NULL(p) && !*pos)
761 seq_puts(seqf, "major minor #blocks name\n\n");
765 static int show_partition(struct seq_file *seqf, void *v)
767 struct gendisk *sgp = v;
768 struct block_device *part;
771 /* Don't show non-partitionable removeable devices or empty devices */
772 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
773 (sgp->flags & GENHD_FL_REMOVABLE)))
775 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
779 xa_for_each(&sgp->part_tbl, idx, part) {
780 if (!bdev_nr_sectors(part))
782 seq_printf(seqf, "%4d %7d %10llu %pg\n",
783 MAJOR(part->bd_dev), MINOR(part->bd_dev),
784 bdev_nr_sectors(part) >> 1, part);
790 static const struct seq_operations partitions_op = {
791 .start = show_partition_start,
792 .next = disk_seqf_next,
793 .stop = disk_seqf_stop,
794 .show = show_partition
798 static int __init genhd_device_init(void)
802 block_class.dev_kobj = sysfs_dev_block_kobj;
803 error = class_register(&block_class);
808 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
810 /* create top-level block dir */
811 if (!sysfs_deprecated)
812 block_depr = kobject_create_and_add("block", NULL);
816 subsys_initcall(genhd_device_init);
818 static ssize_t disk_range_show(struct device *dev,
819 struct device_attribute *attr, char *buf)
821 struct gendisk *disk = dev_to_disk(dev);
823 return sprintf(buf, "%d\n", disk->minors);
826 static ssize_t disk_ext_range_show(struct device *dev,
827 struct device_attribute *attr, char *buf)
829 struct gendisk *disk = dev_to_disk(dev);
831 return sprintf(buf, "%d\n", disk_max_parts(disk));
834 static ssize_t disk_removable_show(struct device *dev,
835 struct device_attribute *attr, char *buf)
837 struct gendisk *disk = dev_to_disk(dev);
839 return sprintf(buf, "%d\n",
840 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
843 static ssize_t disk_hidden_show(struct device *dev,
844 struct device_attribute *attr, char *buf)
846 struct gendisk *disk = dev_to_disk(dev);
848 return sprintf(buf, "%d\n",
849 (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
852 static ssize_t disk_ro_show(struct device *dev,
853 struct device_attribute *attr, char *buf)
855 struct gendisk *disk = dev_to_disk(dev);
857 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
860 ssize_t part_size_show(struct device *dev,
861 struct device_attribute *attr, char *buf)
863 return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev)));
866 ssize_t part_stat_show(struct device *dev,
867 struct device_attribute *attr, char *buf)
869 struct block_device *bdev = dev_to_bdev(dev);
870 struct request_queue *q = bdev->bd_disk->queue;
871 struct disk_stats stat;
872 unsigned int inflight;
874 part_stat_read_all(bdev, &stat);
876 inflight = blk_mq_in_flight(q, bdev);
878 inflight = part_in_flight(bdev);
881 "%8lu %8lu %8llu %8u "
882 "%8lu %8lu %8llu %8u "
884 "%8lu %8lu %8llu %8u "
888 stat.merges[STAT_READ],
889 (unsigned long long)stat.sectors[STAT_READ],
890 (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC),
891 stat.ios[STAT_WRITE],
892 stat.merges[STAT_WRITE],
893 (unsigned long long)stat.sectors[STAT_WRITE],
894 (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC),
896 jiffies_to_msecs(stat.io_ticks),
897 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
898 stat.nsecs[STAT_WRITE] +
899 stat.nsecs[STAT_DISCARD] +
900 stat.nsecs[STAT_FLUSH],
902 stat.ios[STAT_DISCARD],
903 stat.merges[STAT_DISCARD],
904 (unsigned long long)stat.sectors[STAT_DISCARD],
905 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC),
906 stat.ios[STAT_FLUSH],
907 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC));
910 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
913 struct block_device *bdev = dev_to_bdev(dev);
914 struct request_queue *q = bdev->bd_disk->queue;
915 unsigned int inflight[2];
918 blk_mq_in_flight_rw(q, bdev, inflight);
920 part_in_flight_rw(bdev, inflight);
922 return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
925 static ssize_t disk_capability_show(struct device *dev,
926 struct device_attribute *attr, char *buf)
928 struct gendisk *disk = dev_to_disk(dev);
930 return sprintf(buf, "%x\n", disk->flags);
933 static ssize_t disk_alignment_offset_show(struct device *dev,
934 struct device_attribute *attr,
937 struct gendisk *disk = dev_to_disk(dev);
939 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
942 static ssize_t disk_discard_alignment_show(struct device *dev,
943 struct device_attribute *attr,
946 struct gendisk *disk = dev_to_disk(dev);
948 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
951 static ssize_t diskseq_show(struct device *dev,
952 struct device_attribute *attr, char *buf)
954 struct gendisk *disk = dev_to_disk(dev);
956 return sprintf(buf, "%llu\n", disk->diskseq);
959 static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
960 static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
961 static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
962 static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
963 static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
964 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
965 static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
966 static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
967 static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
968 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
969 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
970 static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
971 static DEVICE_ATTR(diskseq, 0444, diskseq_show, NULL);
973 #ifdef CONFIG_FAIL_MAKE_REQUEST
974 ssize_t part_fail_show(struct device *dev,
975 struct device_attribute *attr, char *buf)
977 return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail);
980 ssize_t part_fail_store(struct device *dev,
981 struct device_attribute *attr,
982 const char *buf, size_t count)
986 if (count > 0 && sscanf(buf, "%d", &i) > 0)
987 dev_to_bdev(dev)->bd_make_it_fail = i;
992 static struct device_attribute dev_attr_fail =
993 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
994 #endif /* CONFIG_FAIL_MAKE_REQUEST */
996 #ifdef CONFIG_FAIL_IO_TIMEOUT
997 static struct device_attribute dev_attr_fail_timeout =
998 __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
1001 static struct attribute *disk_attrs[] = {
1002 &dev_attr_range.attr,
1003 &dev_attr_ext_range.attr,
1004 &dev_attr_removable.attr,
1005 &dev_attr_hidden.attr,
1007 &dev_attr_size.attr,
1008 &dev_attr_alignment_offset.attr,
1009 &dev_attr_discard_alignment.attr,
1010 &dev_attr_capability.attr,
1011 &dev_attr_stat.attr,
1012 &dev_attr_inflight.attr,
1013 &dev_attr_badblocks.attr,
1014 &dev_attr_events.attr,
1015 &dev_attr_events_async.attr,
1016 &dev_attr_events_poll_msecs.attr,
1017 &dev_attr_diskseq.attr,
1018 #ifdef CONFIG_FAIL_MAKE_REQUEST
1019 &dev_attr_fail.attr,
1021 #ifdef CONFIG_FAIL_IO_TIMEOUT
1022 &dev_attr_fail_timeout.attr,
1027 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1029 struct device *dev = container_of(kobj, typeof(*dev), kobj);
1030 struct gendisk *disk = dev_to_disk(dev);
1032 if (a == &dev_attr_badblocks.attr && !disk->bb)
1037 static struct attribute_group disk_attr_group = {
1038 .attrs = disk_attrs,
1039 .is_visible = disk_visible,
1042 static const struct attribute_group *disk_attr_groups[] = {
1048 * disk_release - releases all allocated resources of the gendisk
1049 * @dev: the device representing this disk
1051 * This function releases all allocated resources of the gendisk.
1053 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1054 * assigned. Since the request_queue sits on top of the gendisk for these
1055 * drivers we also call blk_put_queue() for them, and we expect the
1056 * request_queue refcount to reach 0 at this point, and so the request_queue
1057 * will also be freed prior to the disk.
1059 * Context: can sleep
1061 static void disk_release(struct device *dev)
1063 struct gendisk *disk = dev_to_disk(dev);
1067 disk_release_events(disk);
1068 kfree(disk->random);
1069 xa_destroy(&disk->part_tbl);
1070 disk->queue->disk = NULL;
1071 blk_put_queue(disk->queue);
1072 iput(disk->part0->bd_inode); /* frees the disk */
1075 static int block_uevent(struct device *dev, struct kobj_uevent_env *env)
1077 struct gendisk *disk = dev_to_disk(dev);
1079 return add_uevent_var(env, "DISKSEQ=%llu", disk->diskseq);
1082 struct class block_class = {
1084 .dev_uevent = block_uevent,
1087 static char *block_devnode(struct device *dev, umode_t *mode,
1088 kuid_t *uid, kgid_t *gid)
1090 struct gendisk *disk = dev_to_disk(dev);
1092 if (disk->fops->devnode)
1093 return disk->fops->devnode(disk, mode);
1097 const struct device_type disk_type = {
1099 .groups = disk_attr_groups,
1100 .release = disk_release,
1101 .devnode = block_devnode,
1104 #ifdef CONFIG_PROC_FS
1106 * aggregate disk stat collector. Uses the same stats that the sysfs
1107 * entries do, above, but makes them available through one seq_file.
1109 * The output looks suspiciously like /proc/partitions with a bunch of
1112 static int diskstats_show(struct seq_file *seqf, void *v)
1114 struct gendisk *gp = v;
1115 struct block_device *hd;
1116 unsigned int inflight;
1117 struct disk_stats stat;
1121 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1122 seq_puts(seqf, "major minor name"
1123 " rio rmerge rsect ruse wio wmerge "
1124 "wsect wuse running use aveq"
1129 xa_for_each(&gp->part_tbl, idx, hd) {
1130 if (bdev_is_partition(hd) && !bdev_nr_sectors(hd))
1132 part_stat_read_all(hd, &stat);
1133 if (queue_is_mq(gp->queue))
1134 inflight = blk_mq_in_flight(gp->queue, hd);
1136 inflight = part_in_flight(hd);
1138 seq_printf(seqf, "%4d %7d %pg "
1145 MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd,
1146 stat.ios[STAT_READ],
1147 stat.merges[STAT_READ],
1148 stat.sectors[STAT_READ],
1149 (unsigned int)div_u64(stat.nsecs[STAT_READ],
1151 stat.ios[STAT_WRITE],
1152 stat.merges[STAT_WRITE],
1153 stat.sectors[STAT_WRITE],
1154 (unsigned int)div_u64(stat.nsecs[STAT_WRITE],
1157 jiffies_to_msecs(stat.io_ticks),
1158 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1159 stat.nsecs[STAT_WRITE] +
1160 stat.nsecs[STAT_DISCARD] +
1161 stat.nsecs[STAT_FLUSH],
1163 stat.ios[STAT_DISCARD],
1164 stat.merges[STAT_DISCARD],
1165 stat.sectors[STAT_DISCARD],
1166 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD],
1168 stat.ios[STAT_FLUSH],
1169 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH],
1178 static const struct seq_operations diskstats_op = {
1179 .start = disk_seqf_start,
1180 .next = disk_seqf_next,
1181 .stop = disk_seqf_stop,
1182 .show = diskstats_show
1185 static int __init proc_genhd_init(void)
1187 proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1188 proc_create_seq("partitions", 0, NULL, &partitions_op);
1191 module_init(proc_genhd_init);
1192 #endif /* CONFIG_PROC_FS */
1194 dev_t part_devt(struct gendisk *disk, u8 partno)
1196 struct block_device *part;
1200 part = xa_load(&disk->part_tbl, partno);
1202 devt = part->bd_dev;
1208 dev_t blk_lookup_devt(const char *name, int partno)
1210 dev_t devt = MKDEV(0, 0);
1211 struct class_dev_iter iter;
1214 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1215 while ((dev = class_dev_iter_next(&iter))) {
1216 struct gendisk *disk = dev_to_disk(dev);
1218 if (strcmp(dev_name(dev), name))
1221 if (partno < disk->minors) {
1222 /* We need to return the right devno, even
1223 * if the partition doesn't exist yet.
1225 devt = MKDEV(MAJOR(dev->devt),
1226 MINOR(dev->devt) + partno);
1228 devt = part_devt(disk, partno);
1233 class_dev_iter_exit(&iter);
1237 struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id,
1238 struct lock_class_key *lkclass)
1240 struct gendisk *disk;
1242 if (!blk_get_queue(q))
1245 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1249 disk->bdi = bdi_alloc(node_id);
1253 disk->part0 = bdev_alloc(disk, 0);
1257 disk->node_id = node_id;
1258 mutex_init(&disk->open_mutex);
1259 xa_init(&disk->part_tbl);
1260 if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL))
1261 goto out_destroy_part_tbl;
1263 rand_initialize_disk(disk);
1264 disk_to_dev(disk)->class = &block_class;
1265 disk_to_dev(disk)->type = &disk_type;
1266 device_initialize(disk_to_dev(disk));
1270 lockdep_init_map(&disk->lockdep_map, "(bio completion)", lkclass, 0);
1271 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
1272 INIT_LIST_HEAD(&disk->slave_bdevs);
1276 out_destroy_part_tbl:
1277 xa_destroy(&disk->part_tbl);
1278 iput(disk->part0->bd_inode);
1287 EXPORT_SYMBOL(__alloc_disk_node);
1289 struct gendisk *__blk_alloc_disk(int node, struct lock_class_key *lkclass)
1291 struct request_queue *q;
1292 struct gendisk *disk;
1294 q = blk_alloc_queue(node);
1298 disk = __alloc_disk_node(q, node, lkclass);
1300 blk_cleanup_queue(q);
1305 EXPORT_SYMBOL(__blk_alloc_disk);
1308 * put_disk - decrements the gendisk refcount
1309 * @disk: the struct gendisk to decrement the refcount for
1311 * This decrements the refcount for the struct gendisk. When this reaches 0
1312 * we'll have disk_release() called.
1314 * Context: Any context, but the last reference must not be dropped from
1317 void put_disk(struct gendisk *disk)
1320 put_device(disk_to_dev(disk));
1322 EXPORT_SYMBOL(put_disk);
1325 * blk_cleanup_disk - shutdown a gendisk allocated by blk_alloc_disk
1326 * @disk: gendisk to shutdown
1328 * Mark the queue hanging off @disk DYING, drain all pending requests, then mark
1329 * the queue DEAD, destroy and put it and the gendisk structure.
1331 * Context: can sleep
1333 void blk_cleanup_disk(struct gendisk *disk)
1335 blk_cleanup_queue(disk->queue);
1338 EXPORT_SYMBOL(blk_cleanup_disk);
1340 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1342 char event[] = "DISK_RO=1";
1343 char *envp[] = { event, NULL };
1347 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1351 * set_disk_ro - set a gendisk read-only
1352 * @disk: gendisk to operate on
1353 * @read_only: %true to set the disk read-only, %false set the disk read/write
1355 * This function is used to indicate whether a given disk device should have its
1356 * read-only flag set. set_disk_ro() is typically used by device drivers to
1357 * indicate whether the underlying physical device is write-protected.
1359 void set_disk_ro(struct gendisk *disk, bool read_only)
1362 if (test_and_set_bit(GD_READ_ONLY, &disk->state))
1365 if (!test_and_clear_bit(GD_READ_ONLY, &disk->state))
1368 set_disk_ro_uevent(disk, read_only);
1370 EXPORT_SYMBOL(set_disk_ro);
1372 int bdev_read_only(struct block_device *bdev)
1374 return bdev->bd_read_only || get_disk_ro(bdev->bd_disk);
1376 EXPORT_SYMBOL(bdev_read_only);
1378 void inc_diskseq(struct gendisk *disk)
1380 disk->diskseq = atomic64_inc_return(&diskseq);