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 ||
80 (disk->flags & (GENHD_FL_UP | GENHD_FL_HIDDEN)) != GENHD_FL_UP)
83 pr_info("%s: detected capacity change from %lld to %lld\n",
84 disk->disk_name, capacity, size);
87 * Historically we did not send a uevent for changes to/from an empty
90 if (!capacity || !size)
92 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
95 EXPORT_SYMBOL_GPL(set_capacity_and_notify);
98 * Format the device name of the indicated block device into the supplied buffer
99 * and return a pointer to that same buffer for convenience.
101 * Note: do not use this in new code, use the %pg specifier to sprintf and
104 const char *bdevname(struct block_device *bdev, char *buf)
106 struct gendisk *hd = bdev->bd_disk;
107 int partno = bdev->bd_partno;
110 snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name);
111 else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1]))
112 snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno);
114 snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno);
118 EXPORT_SYMBOL(bdevname);
120 static void part_stat_read_all(struct block_device *part,
121 struct disk_stats *stat)
125 memset(stat, 0, sizeof(struct disk_stats));
126 for_each_possible_cpu(cpu) {
127 struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu);
130 for (group = 0; group < NR_STAT_GROUPS; group++) {
131 stat->nsecs[group] += ptr->nsecs[group];
132 stat->sectors[group] += ptr->sectors[group];
133 stat->ios[group] += ptr->ios[group];
134 stat->merges[group] += ptr->merges[group];
137 stat->io_ticks += ptr->io_ticks;
141 static unsigned int part_in_flight(struct block_device *part)
143 unsigned int inflight = 0;
146 for_each_possible_cpu(cpu) {
147 inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) +
148 part_stat_local_read_cpu(part, in_flight[1], cpu);
150 if ((int)inflight < 0)
156 static void part_in_flight_rw(struct block_device *part,
157 unsigned int inflight[2])
163 for_each_possible_cpu(cpu) {
164 inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu);
165 inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu);
167 if ((int)inflight[0] < 0)
169 if ((int)inflight[1] < 0)
174 * Can be deleted altogether. Later.
177 #define BLKDEV_MAJOR_HASH_SIZE 255
178 static struct blk_major_name {
179 struct blk_major_name *next;
182 void (*probe)(dev_t devt);
183 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
184 static DEFINE_MUTEX(major_names_lock);
186 /* index in the above - for now: assume no multimajor ranges */
187 static inline int major_to_index(unsigned major)
189 return major % BLKDEV_MAJOR_HASH_SIZE;
192 #ifdef CONFIG_PROC_FS
193 void blkdev_show(struct seq_file *seqf, off_t offset)
195 struct blk_major_name *dp;
197 mutex_lock(&major_names_lock);
198 for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
199 if (dp->major == offset)
200 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
201 mutex_unlock(&major_names_lock);
203 #endif /* CONFIG_PROC_FS */
206 * __register_blkdev - register a new block device
208 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
209 * @major = 0, try to allocate any unused major number.
210 * @name: the name of the new block device as a zero terminated string
211 * @probe: allback that is called on access to any minor number of @major
213 * The @name must be unique within the system.
215 * The return value depends on the @major input parameter:
217 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
218 * then the function returns zero on success, or a negative error code
219 * - if any unused major number was requested with @major = 0 parameter
220 * then the return value is the allocated major number in range
221 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
223 * See Documentation/admin-guide/devices.txt for the list of allocated
226 * Use register_blkdev instead for any new code.
228 int __register_blkdev(unsigned int major, const char *name,
229 void (*probe)(dev_t devt))
231 struct blk_major_name **n, *p;
234 mutex_lock(&major_names_lock);
238 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
239 if (major_names[index] == NULL)
244 printk("%s: failed to get major for %s\n",
253 if (major >= BLKDEV_MAJOR_MAX) {
254 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
255 __func__, major, BLKDEV_MAJOR_MAX-1, name);
261 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
269 strlcpy(p->name, name, sizeof(p->name));
271 index = major_to_index(major);
273 for (n = &major_names[index]; *n; n = &(*n)->next) {
274 if ((*n)->major == major)
283 printk("register_blkdev: cannot get major %u for %s\n",
288 mutex_unlock(&major_names_lock);
291 EXPORT_SYMBOL(__register_blkdev);
293 void unregister_blkdev(unsigned int major, const char *name)
295 struct blk_major_name **n;
296 struct blk_major_name *p = NULL;
297 int index = major_to_index(major);
299 mutex_lock(&major_names_lock);
300 for (n = &major_names[index]; *n; n = &(*n)->next)
301 if ((*n)->major == major)
303 if (!*n || strcmp((*n)->name, name)) {
309 mutex_unlock(&major_names_lock);
313 EXPORT_SYMBOL(unregister_blkdev);
316 * blk_mangle_minor - scatter minor numbers apart
317 * @minor: minor number to mangle
319 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
320 * is enabled. Mangling twice gives the original value.
328 static int blk_mangle_minor(int minor)
330 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
333 for (i = 0; i < MINORBITS / 2; i++) {
334 int low = minor & (1 << i);
335 int high = minor & (1 << (MINORBITS - 1 - i));
336 int distance = MINORBITS - 1 - 2 * i;
338 minor ^= low | high; /* clear both bits */
339 low <<= distance; /* swap the positions */
341 minor |= low | high; /* and set */
347 int blk_alloc_ext_minor(void)
351 idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT, GFP_KERNEL);
357 return blk_mangle_minor(idx);
360 void blk_free_ext_minor(unsigned int minor)
362 ida_free(&ext_devt_ida, blk_mangle_minor(minor));
365 static char *bdevt_str(dev_t devt, char *buf)
367 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
368 char tbuf[BDEVT_SIZE];
369 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
370 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
372 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
377 void disk_uevent(struct gendisk *disk, enum kobject_action action)
379 struct block_device *part;
383 xa_for_each(&disk->part_tbl, idx, part) {
384 if (bdev_is_partition(part) && !bdev_nr_sectors(part))
386 if (!kobject_get_unless_zero(&part->bd_device.kobj))
390 kobject_uevent(bdev_kobj(part), action);
391 put_device(&part->bd_device);
396 EXPORT_SYMBOL_GPL(disk_uevent);
398 static void disk_scan_partitions(struct gendisk *disk)
400 struct block_device *bdev;
402 if (!get_capacity(disk) || !disk_part_scan_enabled(disk))
405 set_bit(GD_NEED_PART_SCAN, &disk->state);
406 bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL);
408 blkdev_put(bdev, FMODE_READ);
411 static void register_disk(struct device *parent, struct gendisk *disk,
412 const struct attribute_group **groups)
414 struct device *ddev = disk_to_dev(disk);
417 ddev->parent = parent;
419 dev_set_name(ddev, "%s", disk->disk_name);
421 /* delay uevents, until we scanned partition table */
422 dev_set_uevent_suppress(ddev, 1);
425 WARN_ON(ddev->groups);
426 ddev->groups = groups;
428 if (device_add(ddev))
430 if (!sysfs_deprecated) {
431 err = sysfs_create_link(block_depr, &ddev->kobj,
432 kobject_name(&ddev->kobj));
440 * avoid probable deadlock caused by allocating memory with
441 * GFP_KERNEL in runtime_resume callback of its all ancestor
444 pm_runtime_set_memalloc_noio(ddev, true);
446 disk->part0->bd_holder_dir =
447 kobject_create_and_add("holders", &ddev->kobj);
448 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
451 * XXX: this is a mess, can't wait for real error handling in add_disk.
452 * Make sure ->slave_dir is NULL if we failed some of the registration
453 * so that the cleanup in bd_unlink_disk_holder works properly.
455 if (bd_register_pending_holders(disk) < 0) {
456 kobject_put(disk->slave_dir);
457 disk->slave_dir = NULL;
460 if (disk->flags & GENHD_FL_HIDDEN)
463 disk_scan_partitions(disk);
465 /* announce the disk and partitions after all partitions are created */
466 dev_set_uevent_suppress(ddev, 0);
467 disk_uevent(disk, KOBJ_ADD);
469 if (disk->queue->backing_dev_info->dev) {
470 err = sysfs_create_link(&ddev->kobj,
471 &disk->queue->backing_dev_info->dev->kobj,
478 * __device_add_disk - add disk information to kernel list
479 * @parent: parent device for the disk
480 * @disk: per-device partitioning information
481 * @groups: Additional per-device sysfs groups
482 * @register_queue: register the queue if set to true
484 * This function registers the partitioning information in @disk
487 * FIXME: error handling
489 static void __device_add_disk(struct device *parent, struct gendisk *disk,
490 const struct attribute_group **groups,
496 * The disk queue should now be all set with enough information about
497 * the device for the elevator code to pick an adequate default
498 * elevator if one is needed, that is, for devices requesting queue
502 elevator_init_mq(disk->queue);
505 * If the driver provides an explicit major number it also must provide
506 * the number of minors numbers supported, and those will be used to
508 * Otherwise just allocate the device numbers for both the whole device
509 * and all partitions from the extended dev_t space.
512 WARN_ON(!disk->minors);
514 if (disk->minors > DISK_MAX_PARTS) {
515 pr_err("block: can't allocate more than %d partitions\n",
517 disk->minors = DISK_MAX_PARTS;
520 WARN_ON(disk->minors);
522 ret = blk_alloc_ext_minor();
527 disk->major = BLOCK_EXT_MAJOR;
528 disk->first_minor = MINOR(ret);
529 disk->flags |= GENHD_FL_EXT_DEVT;
532 disk->flags |= GENHD_FL_UP;
534 disk_alloc_events(disk);
536 if (disk->flags & GENHD_FL_HIDDEN) {
538 * Don't let hidden disks show up in /proc/partitions,
539 * and don't bother scanning for partitions either.
541 disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
542 disk->flags |= GENHD_FL_NO_PART_SCAN;
544 struct backing_dev_info *bdi = disk->queue->backing_dev_info;
545 struct device *dev = disk_to_dev(disk);
547 /* Register BDI before referencing it from bdev */
548 dev->devt = MKDEV(disk->major, disk->first_minor);
549 ret = bdi_register(bdi, "%u:%u",
550 disk->major, disk->first_minor);
552 bdi_set_owner(bdi, dev);
553 bdev_add(disk->part0, dev->devt);
555 register_disk(parent, disk, groups);
557 blk_register_queue(disk);
560 * Take an extra ref on queue which will be put on disk_release()
561 * so that it sticks around as long as @disk is there.
563 if (blk_get_queue(disk->queue))
564 set_bit(GD_QUEUE_REF, &disk->state);
568 disk_add_events(disk);
569 blk_integrity_add(disk);
572 void device_add_disk(struct device *parent, struct gendisk *disk,
573 const struct attribute_group **groups)
576 __device_add_disk(parent, disk, groups, true);
578 EXPORT_SYMBOL(device_add_disk);
580 void device_add_disk_no_queue_reg(struct device *parent, struct gendisk *disk)
582 __device_add_disk(parent, disk, NULL, false);
584 EXPORT_SYMBOL(device_add_disk_no_queue_reg);
587 * del_gendisk - remove the gendisk
588 * @disk: the struct gendisk to remove
590 * Removes the gendisk and all its associated resources. This deletes the
591 * partitions associated with the gendisk, and unregisters the associated
594 * This is the counter to the respective __device_add_disk() call.
596 * The final removal of the struct gendisk happens when its refcount reaches 0
597 * with put_disk(), which should be called after del_gendisk(), if
598 * __device_add_disk() was used.
600 * Drivers exist which depend on the release of the gendisk to be synchronous,
601 * it should not be deferred.
605 void del_gendisk(struct gendisk *disk)
609 if (WARN_ON_ONCE(!disk->queue))
612 blk_integrity_del(disk);
613 disk_del_events(disk);
615 mutex_lock(&disk->open_mutex);
616 remove_inode_hash(disk->part0->bd_inode);
617 disk->flags &= ~GENHD_FL_UP;
618 blk_drop_partitions(disk);
619 mutex_unlock(&disk->open_mutex);
621 fsync_bdev(disk->part0);
622 __invalidate_device(disk->part0, true);
624 set_capacity(disk, 0);
626 if (!(disk->flags & GENHD_FL_HIDDEN)) {
627 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
630 * Unregister bdi before releasing device numbers (as they can
631 * get reused and we'd get clashes in sysfs).
633 bdi_unregister(disk->queue->backing_dev_info);
636 blk_unregister_queue(disk);
638 kobject_put(disk->part0->bd_holder_dir);
639 kobject_put(disk->slave_dir);
641 part_stat_set_all(disk->part0, 0);
642 disk->part0->bd_stamp = 0;
643 if (!sysfs_deprecated)
644 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
645 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
646 device_del(disk_to_dev(disk));
648 EXPORT_SYMBOL(del_gendisk);
650 /* sysfs access to bad-blocks list. */
651 static ssize_t disk_badblocks_show(struct device *dev,
652 struct device_attribute *attr,
655 struct gendisk *disk = dev_to_disk(dev);
658 return sprintf(page, "\n");
660 return badblocks_show(disk->bb, page, 0);
663 static ssize_t disk_badblocks_store(struct device *dev,
664 struct device_attribute *attr,
665 const char *page, size_t len)
667 struct gendisk *disk = dev_to_disk(dev);
672 return badblocks_store(disk->bb, page, len, 0);
675 void blk_request_module(dev_t devt)
677 unsigned int major = MAJOR(devt);
678 struct blk_major_name **n;
680 mutex_lock(&major_names_lock);
681 for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) {
682 if ((*n)->major == major && (*n)->probe) {
684 mutex_unlock(&major_names_lock);
688 mutex_unlock(&major_names_lock);
690 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
691 /* Make old-style 2.4 aliases work */
692 request_module("block-major-%d", MAJOR(devt));
696 * print a full list of all partitions - intended for places where the root
697 * filesystem can't be mounted and thus to give the victim some idea of what
700 void __init printk_all_partitions(void)
702 struct class_dev_iter iter;
705 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
706 while ((dev = class_dev_iter_next(&iter))) {
707 struct gendisk *disk = dev_to_disk(dev);
708 struct block_device *part;
709 char devt_buf[BDEVT_SIZE];
713 * Don't show empty devices or things that have been
716 if (get_capacity(disk) == 0 ||
717 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
721 * Note, unlike /proc/partitions, I am showing the numbers in
722 * hex - the same format as the root= option takes.
725 xa_for_each(&disk->part_tbl, idx, part) {
726 if (!bdev_nr_sectors(part))
728 printk("%s%s %10llu %pg %s",
729 bdev_is_partition(part) ? " " : "",
730 bdevt_str(part->bd_dev, devt_buf),
731 bdev_nr_sectors(part) >> 1, part,
733 part->bd_meta_info->uuid : "");
734 if (bdev_is_partition(part))
736 else if (dev->parent && dev->parent->driver)
737 printk(" driver: %s\n",
738 dev->parent->driver->name);
740 printk(" (driver?)\n");
744 class_dev_iter_exit(&iter);
747 #ifdef CONFIG_PROC_FS
749 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
752 struct class_dev_iter *iter;
755 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
757 return ERR_PTR(-ENOMEM);
759 seqf->private = iter;
760 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
762 dev = class_dev_iter_next(iter);
767 return dev_to_disk(dev);
770 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
775 dev = class_dev_iter_next(seqf->private);
777 return dev_to_disk(dev);
782 static void disk_seqf_stop(struct seq_file *seqf, void *v)
784 struct class_dev_iter *iter = seqf->private;
786 /* stop is called even after start failed :-( */
788 class_dev_iter_exit(iter);
790 seqf->private = NULL;
794 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
798 p = disk_seqf_start(seqf, pos);
799 if (!IS_ERR_OR_NULL(p) && !*pos)
800 seq_puts(seqf, "major minor #blocks name\n\n");
804 static int show_partition(struct seq_file *seqf, void *v)
806 struct gendisk *sgp = v;
807 struct block_device *part;
810 /* Don't show non-partitionable removeable devices or empty devices */
811 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
812 (sgp->flags & GENHD_FL_REMOVABLE)))
814 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
818 xa_for_each(&sgp->part_tbl, idx, part) {
819 if (!bdev_nr_sectors(part))
821 seq_printf(seqf, "%4d %7d %10llu %pg\n",
822 MAJOR(part->bd_dev), MINOR(part->bd_dev),
823 bdev_nr_sectors(part) >> 1, part);
829 static const struct seq_operations partitions_op = {
830 .start = show_partition_start,
831 .next = disk_seqf_next,
832 .stop = disk_seqf_stop,
833 .show = show_partition
837 static int __init genhd_device_init(void)
841 block_class.dev_kobj = sysfs_dev_block_kobj;
842 error = class_register(&block_class);
847 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
849 /* create top-level block dir */
850 if (!sysfs_deprecated)
851 block_depr = kobject_create_and_add("block", NULL);
855 subsys_initcall(genhd_device_init);
857 static ssize_t disk_range_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", disk->minors);
865 static ssize_t disk_ext_range_show(struct device *dev,
866 struct device_attribute *attr, char *buf)
868 struct gendisk *disk = dev_to_disk(dev);
870 return sprintf(buf, "%d\n", disk_max_parts(disk));
873 static ssize_t disk_removable_show(struct device *dev,
874 struct device_attribute *attr, char *buf)
876 struct gendisk *disk = dev_to_disk(dev);
878 return sprintf(buf, "%d\n",
879 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
882 static ssize_t disk_hidden_show(struct device *dev,
883 struct device_attribute *attr, char *buf)
885 struct gendisk *disk = dev_to_disk(dev);
887 return sprintf(buf, "%d\n",
888 (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
891 static ssize_t disk_ro_show(struct device *dev,
892 struct device_attribute *attr, char *buf)
894 struct gendisk *disk = dev_to_disk(dev);
896 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
899 ssize_t part_size_show(struct device *dev,
900 struct device_attribute *attr, char *buf)
902 return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev)));
905 ssize_t part_stat_show(struct device *dev,
906 struct device_attribute *attr, char *buf)
908 struct block_device *bdev = dev_to_bdev(dev);
909 struct request_queue *q = bdev->bd_disk->queue;
910 struct disk_stats stat;
911 unsigned int inflight;
913 part_stat_read_all(bdev, &stat);
915 inflight = blk_mq_in_flight(q, bdev);
917 inflight = part_in_flight(bdev);
920 "%8lu %8lu %8llu %8u "
921 "%8lu %8lu %8llu %8u "
923 "%8lu %8lu %8llu %8u "
927 stat.merges[STAT_READ],
928 (unsigned long long)stat.sectors[STAT_READ],
929 (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC),
930 stat.ios[STAT_WRITE],
931 stat.merges[STAT_WRITE],
932 (unsigned long long)stat.sectors[STAT_WRITE],
933 (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC),
935 jiffies_to_msecs(stat.io_ticks),
936 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
937 stat.nsecs[STAT_WRITE] +
938 stat.nsecs[STAT_DISCARD] +
939 stat.nsecs[STAT_FLUSH],
941 stat.ios[STAT_DISCARD],
942 stat.merges[STAT_DISCARD],
943 (unsigned long long)stat.sectors[STAT_DISCARD],
944 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC),
945 stat.ios[STAT_FLUSH],
946 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC));
949 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
952 struct block_device *bdev = dev_to_bdev(dev);
953 struct request_queue *q = bdev->bd_disk->queue;
954 unsigned int inflight[2];
957 blk_mq_in_flight_rw(q, bdev, inflight);
959 part_in_flight_rw(bdev, inflight);
961 return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
964 static ssize_t disk_capability_show(struct device *dev,
965 struct device_attribute *attr, char *buf)
967 struct gendisk *disk = dev_to_disk(dev);
969 return sprintf(buf, "%x\n", disk->flags);
972 static ssize_t disk_alignment_offset_show(struct device *dev,
973 struct device_attribute *attr,
976 struct gendisk *disk = dev_to_disk(dev);
978 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
981 static ssize_t disk_discard_alignment_show(struct device *dev,
982 struct device_attribute *attr,
985 struct gendisk *disk = dev_to_disk(dev);
987 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
990 static ssize_t diskseq_show(struct device *dev,
991 struct device_attribute *attr, char *buf)
993 struct gendisk *disk = dev_to_disk(dev);
995 return sprintf(buf, "%llu\n", disk->diskseq);
998 static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
999 static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
1000 static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
1001 static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
1002 static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
1003 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
1004 static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
1005 static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
1006 static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
1007 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
1008 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
1009 static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
1010 static DEVICE_ATTR(diskseq, 0444, diskseq_show, NULL);
1012 #ifdef CONFIG_FAIL_MAKE_REQUEST
1013 ssize_t part_fail_show(struct device *dev,
1014 struct device_attribute *attr, char *buf)
1016 return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail);
1019 ssize_t part_fail_store(struct device *dev,
1020 struct device_attribute *attr,
1021 const char *buf, size_t count)
1025 if (count > 0 && sscanf(buf, "%d", &i) > 0)
1026 dev_to_bdev(dev)->bd_make_it_fail = i;
1031 static struct device_attribute dev_attr_fail =
1032 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
1033 #endif /* CONFIG_FAIL_MAKE_REQUEST */
1035 #ifdef CONFIG_FAIL_IO_TIMEOUT
1036 static struct device_attribute dev_attr_fail_timeout =
1037 __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
1040 static struct attribute *disk_attrs[] = {
1041 &dev_attr_range.attr,
1042 &dev_attr_ext_range.attr,
1043 &dev_attr_removable.attr,
1044 &dev_attr_hidden.attr,
1046 &dev_attr_size.attr,
1047 &dev_attr_alignment_offset.attr,
1048 &dev_attr_discard_alignment.attr,
1049 &dev_attr_capability.attr,
1050 &dev_attr_stat.attr,
1051 &dev_attr_inflight.attr,
1052 &dev_attr_badblocks.attr,
1053 &dev_attr_events.attr,
1054 &dev_attr_events_async.attr,
1055 &dev_attr_events_poll_msecs.attr,
1056 &dev_attr_diskseq.attr,
1057 #ifdef CONFIG_FAIL_MAKE_REQUEST
1058 &dev_attr_fail.attr,
1060 #ifdef CONFIG_FAIL_IO_TIMEOUT
1061 &dev_attr_fail_timeout.attr,
1066 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1068 struct device *dev = container_of(kobj, typeof(*dev), kobj);
1069 struct gendisk *disk = dev_to_disk(dev);
1071 if (a == &dev_attr_badblocks.attr && !disk->bb)
1076 static struct attribute_group disk_attr_group = {
1077 .attrs = disk_attrs,
1078 .is_visible = disk_visible,
1081 static const struct attribute_group *disk_attr_groups[] = {
1087 * disk_release - releases all allocated resources of the gendisk
1088 * @dev: the device representing this disk
1090 * This function releases all allocated resources of the gendisk.
1092 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1093 * assigned. Since the request_queue sits on top of the gendisk for these
1094 * drivers we also call blk_put_queue() for them, and we expect the
1095 * request_queue refcount to reach 0 at this point, and so the request_queue
1096 * will also be freed prior to the disk.
1098 * Context: can sleep
1100 static void disk_release(struct device *dev)
1102 struct gendisk *disk = dev_to_disk(dev);
1106 if (MAJOR(dev->devt) == BLOCK_EXT_MAJOR)
1107 blk_free_ext_minor(MINOR(dev->devt));
1108 disk_release_events(disk);
1109 kfree(disk->random);
1110 xa_destroy(&disk->part_tbl);
1111 if (test_bit(GD_QUEUE_REF, &disk->state) && disk->queue)
1112 blk_put_queue(disk->queue);
1113 iput(disk->part0->bd_inode); /* frees the disk */
1116 static int block_uevent(struct device *dev, struct kobj_uevent_env *env)
1118 struct gendisk *disk = dev_to_disk(dev);
1120 return add_uevent_var(env, "DISKSEQ=%llu", disk->diskseq);
1123 struct class block_class = {
1125 .dev_uevent = block_uevent,
1128 static char *block_devnode(struct device *dev, umode_t *mode,
1129 kuid_t *uid, kgid_t *gid)
1131 struct gendisk *disk = dev_to_disk(dev);
1133 if (disk->fops->devnode)
1134 return disk->fops->devnode(disk, mode);
1138 const struct device_type disk_type = {
1140 .groups = disk_attr_groups,
1141 .release = disk_release,
1142 .devnode = block_devnode,
1145 #ifdef CONFIG_PROC_FS
1147 * aggregate disk stat collector. Uses the same stats that the sysfs
1148 * entries do, above, but makes them available through one seq_file.
1150 * The output looks suspiciously like /proc/partitions with a bunch of
1153 static int diskstats_show(struct seq_file *seqf, void *v)
1155 struct gendisk *gp = v;
1156 struct block_device *hd;
1157 unsigned int inflight;
1158 struct disk_stats stat;
1162 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1163 seq_puts(seqf, "major minor name"
1164 " rio rmerge rsect ruse wio wmerge "
1165 "wsect wuse running use aveq"
1170 xa_for_each(&gp->part_tbl, idx, hd) {
1171 if (bdev_is_partition(hd) && !bdev_nr_sectors(hd))
1173 part_stat_read_all(hd, &stat);
1174 if (queue_is_mq(gp->queue))
1175 inflight = blk_mq_in_flight(gp->queue, hd);
1177 inflight = part_in_flight(hd);
1179 seq_printf(seqf, "%4d %7d %pg "
1186 MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd,
1187 stat.ios[STAT_READ],
1188 stat.merges[STAT_READ],
1189 stat.sectors[STAT_READ],
1190 (unsigned int)div_u64(stat.nsecs[STAT_READ],
1192 stat.ios[STAT_WRITE],
1193 stat.merges[STAT_WRITE],
1194 stat.sectors[STAT_WRITE],
1195 (unsigned int)div_u64(stat.nsecs[STAT_WRITE],
1198 jiffies_to_msecs(stat.io_ticks),
1199 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1200 stat.nsecs[STAT_WRITE] +
1201 stat.nsecs[STAT_DISCARD] +
1202 stat.nsecs[STAT_FLUSH],
1204 stat.ios[STAT_DISCARD],
1205 stat.merges[STAT_DISCARD],
1206 stat.sectors[STAT_DISCARD],
1207 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD],
1209 stat.ios[STAT_FLUSH],
1210 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH],
1219 static const struct seq_operations diskstats_op = {
1220 .start = disk_seqf_start,
1221 .next = disk_seqf_next,
1222 .stop = disk_seqf_stop,
1223 .show = diskstats_show
1226 static int __init proc_genhd_init(void)
1228 proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1229 proc_create_seq("partitions", 0, NULL, &partitions_op);
1232 module_init(proc_genhd_init);
1233 #endif /* CONFIG_PROC_FS */
1235 dev_t part_devt(struct gendisk *disk, u8 partno)
1237 struct block_device *part;
1241 part = xa_load(&disk->part_tbl, partno);
1243 devt = part->bd_dev;
1249 dev_t blk_lookup_devt(const char *name, int partno)
1251 dev_t devt = MKDEV(0, 0);
1252 struct class_dev_iter iter;
1255 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1256 while ((dev = class_dev_iter_next(&iter))) {
1257 struct gendisk *disk = dev_to_disk(dev);
1259 if (strcmp(dev_name(dev), name))
1262 if (partno < disk->minors) {
1263 /* We need to return the right devno, even
1264 * if the partition doesn't exist yet.
1266 devt = MKDEV(MAJOR(dev->devt),
1267 MINOR(dev->devt) + partno);
1269 devt = part_devt(disk, partno);
1274 class_dev_iter_exit(&iter);
1278 struct gendisk *__alloc_disk_node(int minors, int node_id)
1280 struct gendisk *disk;
1282 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1286 disk->part0 = bdev_alloc(disk, 0);
1290 disk->node_id = node_id;
1291 mutex_init(&disk->open_mutex);
1292 xa_init(&disk->part_tbl);
1293 if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL))
1294 goto out_destroy_part_tbl;
1296 disk->minors = minors;
1297 rand_initialize_disk(disk);
1298 disk_to_dev(disk)->class = &block_class;
1299 disk_to_dev(disk)->type = &disk_type;
1300 device_initialize(disk_to_dev(disk));
1302 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
1303 INIT_LIST_HEAD(&disk->slave_bdevs);
1307 out_destroy_part_tbl:
1308 xa_destroy(&disk->part_tbl);
1309 iput(disk->part0->bd_inode);
1314 EXPORT_SYMBOL(__alloc_disk_node);
1316 struct gendisk *__blk_alloc_disk(int node)
1318 struct request_queue *q;
1319 struct gendisk *disk;
1321 q = blk_alloc_queue(node);
1325 disk = __alloc_disk_node(0, node);
1327 blk_cleanup_queue(q);
1333 EXPORT_SYMBOL(__blk_alloc_disk);
1336 * put_disk - decrements the gendisk refcount
1337 * @disk: the struct gendisk to decrement the refcount for
1339 * This decrements the refcount for the struct gendisk. When this reaches 0
1340 * we'll have disk_release() called.
1342 * Context: Any context, but the last reference must not be dropped from
1345 void put_disk(struct gendisk *disk)
1348 put_device(disk_to_dev(disk));
1350 EXPORT_SYMBOL(put_disk);
1353 * blk_cleanup_disk - shutdown a gendisk allocated by blk_alloc_disk
1354 * @disk: gendisk to shutdown
1356 * Mark the queue hanging off @disk DYING, drain all pending requests, then mark
1357 * the queue DEAD, destroy and put it and the gendisk structure.
1359 * Context: can sleep
1361 void blk_cleanup_disk(struct gendisk *disk)
1363 blk_cleanup_queue(disk->queue);
1366 EXPORT_SYMBOL(blk_cleanup_disk);
1368 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1370 char event[] = "DISK_RO=1";
1371 char *envp[] = { event, NULL };
1375 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1379 * set_disk_ro - set a gendisk read-only
1380 * @disk: gendisk to operate on
1381 * @read_only: %true to set the disk read-only, %false set the disk read/write
1383 * This function is used to indicate whether a given disk device should have its
1384 * read-only flag set. set_disk_ro() is typically used by device drivers to
1385 * indicate whether the underlying physical device is write-protected.
1387 void set_disk_ro(struct gendisk *disk, bool read_only)
1390 if (test_and_set_bit(GD_READ_ONLY, &disk->state))
1393 if (!test_and_clear_bit(GD_READ_ONLY, &disk->state))
1396 set_disk_ro_uevent(disk, read_only);
1398 EXPORT_SYMBOL(set_disk_ro);
1400 int bdev_read_only(struct block_device *bdev)
1402 return bdev->bd_read_only || get_disk_ro(bdev->bd_disk);
1404 EXPORT_SYMBOL(bdev_read_only);
1406 void inc_diskseq(struct gendisk *disk)
1408 disk->diskseq = atomic64_inc_return(&diskseq);