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 block device into the supplied buffer
82 * and return a pointer to that same buffer for convenience.
84 * Note: do not use this in new code, use the %pg specifier to sprintf and
87 const char *bdevname(struct block_device *bdev, char *buf)
89 struct gendisk *hd = bdev->bd_disk;
90 int partno = bdev->bd_partno;
93 snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name);
94 else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1]))
95 snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno);
97 snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno);
101 EXPORT_SYMBOL(bdevname);
103 static void part_stat_read_all(struct block_device *part,
104 struct disk_stats *stat)
108 memset(stat, 0, sizeof(struct disk_stats));
109 for_each_possible_cpu(cpu) {
110 struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu);
113 for (group = 0; group < NR_STAT_GROUPS; group++) {
114 stat->nsecs[group] += ptr->nsecs[group];
115 stat->sectors[group] += ptr->sectors[group];
116 stat->ios[group] += ptr->ios[group];
117 stat->merges[group] += ptr->merges[group];
120 stat->io_ticks += ptr->io_ticks;
124 static unsigned int part_in_flight(struct block_device *part)
126 unsigned int inflight = 0;
129 for_each_possible_cpu(cpu) {
130 inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) +
131 part_stat_local_read_cpu(part, in_flight[1], cpu);
133 if ((int)inflight < 0)
139 static void part_in_flight_rw(struct block_device *part,
140 unsigned int inflight[2])
146 for_each_possible_cpu(cpu) {
147 inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu);
148 inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu);
150 if ((int)inflight[0] < 0)
152 if ((int)inflight[1] < 0)
157 * Can be deleted altogether. Later.
160 #define BLKDEV_MAJOR_HASH_SIZE 255
161 static struct blk_major_name {
162 struct blk_major_name *next;
165 void (*probe)(dev_t devt);
166 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
167 static DEFINE_MUTEX(major_names_lock);
169 /* index in the above - for now: assume no multimajor ranges */
170 static inline int major_to_index(unsigned major)
172 return major % BLKDEV_MAJOR_HASH_SIZE;
175 #ifdef CONFIG_PROC_FS
176 void blkdev_show(struct seq_file *seqf, off_t offset)
178 struct blk_major_name *dp;
180 mutex_lock(&major_names_lock);
181 for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
182 if (dp->major == offset)
183 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
184 mutex_unlock(&major_names_lock);
186 #endif /* CONFIG_PROC_FS */
189 * __register_blkdev - register a new block device
191 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
192 * @major = 0, try to allocate any unused major number.
193 * @name: the name of the new block device as a zero terminated string
194 * @probe: allback that is called on access to any minor number of @major
196 * The @name must be unique within the system.
198 * The return value depends on the @major input parameter:
200 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
201 * then the function returns zero on success, or a negative error code
202 * - if any unused major number was requested with @major = 0 parameter
203 * then the return value is the allocated major number in range
204 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
206 * See Documentation/admin-guide/devices.txt for the list of allocated
209 * Use register_blkdev instead for any new code.
211 int __register_blkdev(unsigned int major, const char *name,
212 void (*probe)(dev_t devt))
214 struct blk_major_name **n, *p;
217 mutex_lock(&major_names_lock);
221 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
222 if (major_names[index] == NULL)
227 printk("%s: failed to get major for %s\n",
236 if (major >= BLKDEV_MAJOR_MAX) {
237 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
238 __func__, major, BLKDEV_MAJOR_MAX-1, name);
244 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
252 strlcpy(p->name, name, sizeof(p->name));
254 index = major_to_index(major);
256 for (n = &major_names[index]; *n; n = &(*n)->next) {
257 if ((*n)->major == major)
266 printk("register_blkdev: cannot get major %u for %s\n",
271 mutex_unlock(&major_names_lock);
274 EXPORT_SYMBOL(__register_blkdev);
276 void unregister_blkdev(unsigned int major, const char *name)
278 struct blk_major_name **n;
279 struct blk_major_name *p = NULL;
280 int index = major_to_index(major);
282 mutex_lock(&major_names_lock);
283 for (n = &major_names[index]; *n; n = &(*n)->next)
284 if ((*n)->major == major)
286 if (!*n || strcmp((*n)->name, name)) {
292 mutex_unlock(&major_names_lock);
296 EXPORT_SYMBOL(unregister_blkdev);
299 * blk_mangle_minor - scatter minor numbers apart
300 * @minor: minor number to mangle
302 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
303 * is enabled. Mangling twice gives the original value.
311 static int blk_mangle_minor(int minor)
313 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
316 for (i = 0; i < MINORBITS / 2; i++) {
317 int low = minor & (1 << i);
318 int high = minor & (1 << (MINORBITS - 1 - i));
319 int distance = MINORBITS - 1 - 2 * i;
321 minor ^= low | high; /* clear both bits */
322 low <<= distance; /* swap the positions */
324 minor |= low | high; /* and set */
330 int blk_alloc_ext_minor(void)
334 idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT, GFP_KERNEL);
340 return blk_mangle_minor(idx);
343 void blk_free_ext_minor(unsigned int minor)
345 ida_free(&ext_devt_ida, blk_mangle_minor(minor));
348 static char *bdevt_str(dev_t devt, char *buf)
350 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
351 char tbuf[BDEVT_SIZE];
352 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
353 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
355 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
360 void disk_uevent(struct gendisk *disk, enum kobject_action action)
362 struct block_device *part;
366 xa_for_each(&disk->part_tbl, idx, part) {
367 if (bdev_is_partition(part) && !bdev_nr_sectors(part))
369 if (!kobject_get_unless_zero(&part->bd_device.kobj))
373 kobject_uevent(bdev_kobj(part), action);
374 put_device(&part->bd_device);
379 EXPORT_SYMBOL_GPL(disk_uevent);
381 static void disk_scan_partitions(struct gendisk *disk)
383 struct block_device *bdev;
385 if (!get_capacity(disk) || !disk_part_scan_enabled(disk))
388 set_bit(GD_NEED_PART_SCAN, &disk->state);
389 bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL);
391 blkdev_put(bdev, FMODE_READ);
394 static void register_disk(struct device *parent, struct gendisk *disk,
395 const struct attribute_group **groups)
397 struct device *ddev = disk_to_dev(disk);
400 ddev->parent = parent;
402 dev_set_name(ddev, "%s", disk->disk_name);
404 /* delay uevents, until we scanned partition table */
405 dev_set_uevent_suppress(ddev, 1);
408 WARN_ON(ddev->groups);
409 ddev->groups = groups;
411 if (device_add(ddev))
413 if (!sysfs_deprecated) {
414 err = sysfs_create_link(block_depr, &ddev->kobj,
415 kobject_name(&ddev->kobj));
423 * avoid probable deadlock caused by allocating memory with
424 * GFP_KERNEL in runtime_resume callback of its all ancestor
427 pm_runtime_set_memalloc_noio(ddev, true);
429 disk->part0->bd_holder_dir =
430 kobject_create_and_add("holders", &ddev->kobj);
431 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
433 if (disk->flags & GENHD_FL_HIDDEN)
436 disk_scan_partitions(disk);
438 /* announce the disk and partitions after all partitions are created */
439 dev_set_uevent_suppress(ddev, 0);
440 disk_uevent(disk, KOBJ_ADD);
442 if (disk->queue->backing_dev_info->dev) {
443 err = sysfs_create_link(&ddev->kobj,
444 &disk->queue->backing_dev_info->dev->kobj,
451 * __device_add_disk - add disk information to kernel list
452 * @parent: parent device for the disk
453 * @disk: per-device partitioning information
454 * @groups: Additional per-device sysfs groups
455 * @register_queue: register the queue if set to true
457 * This function registers the partitioning information in @disk
460 * FIXME: error handling
462 static void __device_add_disk(struct device *parent, struct gendisk *disk,
463 const struct attribute_group **groups,
469 * The disk queue should now be all set with enough information about
470 * the device for the elevator code to pick an adequate default
471 * elevator if one is needed, that is, for devices requesting queue
475 elevator_init_mq(disk->queue);
478 * If the driver provides an explicit major number it also must provide
479 * the number of minors numbers supported, and those will be used to
481 * Otherwise just allocate the device numbers for both the whole device
482 * and all partitions from the extended dev_t space.
485 WARN_ON(!disk->minors);
487 if (disk->minors > DISK_MAX_PARTS) {
488 pr_err("block: can't allocate more than %d partitions\n",
490 disk->minors = DISK_MAX_PARTS;
493 WARN_ON(disk->minors);
495 ret = blk_alloc_ext_minor();
500 disk->major = BLOCK_EXT_MAJOR;
501 disk->first_minor = MINOR(ret);
502 disk->flags |= GENHD_FL_EXT_DEVT;
505 disk->flags |= GENHD_FL_UP;
507 disk_alloc_events(disk);
509 if (disk->flags & GENHD_FL_HIDDEN) {
511 * Don't let hidden disks show up in /proc/partitions,
512 * and don't bother scanning for partitions either.
514 disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
515 disk->flags |= GENHD_FL_NO_PART_SCAN;
517 struct backing_dev_info *bdi = disk->queue->backing_dev_info;
518 struct device *dev = disk_to_dev(disk);
520 /* Register BDI before referencing it from bdev */
521 dev->devt = MKDEV(disk->major, disk->first_minor);
522 ret = bdi_register(bdi, "%u:%u",
523 disk->major, disk->first_minor);
525 bdi_set_owner(bdi, dev);
526 bdev_add(disk->part0, dev->devt);
528 register_disk(parent, disk, groups);
530 blk_register_queue(disk);
533 * Take an extra ref on queue which will be put on disk_release()
534 * so that it sticks around as long as @disk is there.
536 if (blk_get_queue(disk->queue))
537 set_bit(GD_QUEUE_REF, &disk->state);
541 disk_add_events(disk);
542 blk_integrity_add(disk);
545 void device_add_disk(struct device *parent, struct gendisk *disk,
546 const struct attribute_group **groups)
549 __device_add_disk(parent, disk, groups, true);
551 EXPORT_SYMBOL(device_add_disk);
553 void device_add_disk_no_queue_reg(struct device *parent, struct gendisk *disk)
555 __device_add_disk(parent, disk, NULL, false);
557 EXPORT_SYMBOL(device_add_disk_no_queue_reg);
560 * del_gendisk - remove the gendisk
561 * @disk: the struct gendisk to remove
563 * Removes the gendisk and all its associated resources. This deletes the
564 * partitions associated with the gendisk, and unregisters the associated
567 * This is the counter to the respective __device_add_disk() call.
569 * The final removal of the struct gendisk happens when its refcount reaches 0
570 * with put_disk(), which should be called after del_gendisk(), if
571 * __device_add_disk() was used.
573 * Drivers exist which depend on the release of the gendisk to be synchronous,
574 * it should not be deferred.
578 void del_gendisk(struct gendisk *disk)
582 if (WARN_ON_ONCE(!disk->queue))
585 blk_integrity_del(disk);
586 disk_del_events(disk);
588 mutex_lock(&disk->open_mutex);
589 remove_inode_hash(disk->part0->bd_inode);
590 disk->flags &= ~GENHD_FL_UP;
591 blk_drop_partitions(disk);
592 mutex_unlock(&disk->open_mutex);
594 fsync_bdev(disk->part0);
595 __invalidate_device(disk->part0, true);
597 set_capacity(disk, 0);
599 if (!(disk->flags & GENHD_FL_HIDDEN)) {
600 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
603 * Unregister bdi before releasing device numbers (as they can
604 * get reused and we'd get clashes in sysfs).
606 bdi_unregister(disk->queue->backing_dev_info);
609 blk_unregister_queue(disk);
611 kobject_put(disk->part0->bd_holder_dir);
612 kobject_put(disk->slave_dir);
614 part_stat_set_all(disk->part0, 0);
615 disk->part0->bd_stamp = 0;
616 if (!sysfs_deprecated)
617 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
618 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
619 device_del(disk_to_dev(disk));
621 EXPORT_SYMBOL(del_gendisk);
623 /* sysfs access to bad-blocks list. */
624 static ssize_t disk_badblocks_show(struct device *dev,
625 struct device_attribute *attr,
628 struct gendisk *disk = dev_to_disk(dev);
631 return sprintf(page, "\n");
633 return badblocks_show(disk->bb, page, 0);
636 static ssize_t disk_badblocks_store(struct device *dev,
637 struct device_attribute *attr,
638 const char *page, size_t len)
640 struct gendisk *disk = dev_to_disk(dev);
645 return badblocks_store(disk->bb, page, len, 0);
648 void blk_request_module(dev_t devt)
650 unsigned int major = MAJOR(devt);
651 struct blk_major_name **n;
653 mutex_lock(&major_names_lock);
654 for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) {
655 if ((*n)->major == major && (*n)->probe) {
657 mutex_unlock(&major_names_lock);
661 mutex_unlock(&major_names_lock);
663 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
664 /* Make old-style 2.4 aliases work */
665 request_module("block-major-%d", MAJOR(devt));
669 * print a full list of all partitions - intended for places where the root
670 * filesystem can't be mounted and thus to give the victim some idea of what
673 void __init printk_all_partitions(void)
675 struct class_dev_iter iter;
678 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
679 while ((dev = class_dev_iter_next(&iter))) {
680 struct gendisk *disk = dev_to_disk(dev);
681 struct block_device *part;
682 char devt_buf[BDEVT_SIZE];
686 * Don't show empty devices or things that have been
689 if (get_capacity(disk) == 0 ||
690 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
694 * Note, unlike /proc/partitions, I am showing the numbers in
695 * hex - the same format as the root= option takes.
698 xa_for_each(&disk->part_tbl, idx, part) {
699 if (!bdev_nr_sectors(part))
701 printk("%s%s %10llu %pg %s",
702 bdev_is_partition(part) ? " " : "",
703 bdevt_str(part->bd_dev, devt_buf),
704 bdev_nr_sectors(part) >> 1, part,
706 part->bd_meta_info->uuid : "");
707 if (bdev_is_partition(part))
709 else if (dev->parent && dev->parent->driver)
710 printk(" driver: %s\n",
711 dev->parent->driver->name);
713 printk(" (driver?)\n");
717 class_dev_iter_exit(&iter);
720 #ifdef CONFIG_PROC_FS
722 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
725 struct class_dev_iter *iter;
728 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
730 return ERR_PTR(-ENOMEM);
732 seqf->private = iter;
733 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
735 dev = class_dev_iter_next(iter);
740 return dev_to_disk(dev);
743 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
748 dev = class_dev_iter_next(seqf->private);
750 return dev_to_disk(dev);
755 static void disk_seqf_stop(struct seq_file *seqf, void *v)
757 struct class_dev_iter *iter = seqf->private;
759 /* stop is called even after start failed :-( */
761 class_dev_iter_exit(iter);
763 seqf->private = NULL;
767 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
771 p = disk_seqf_start(seqf, pos);
772 if (!IS_ERR_OR_NULL(p) && !*pos)
773 seq_puts(seqf, "major minor #blocks name\n\n");
777 static int show_partition(struct seq_file *seqf, void *v)
779 struct gendisk *sgp = v;
780 struct block_device *part;
783 /* Don't show non-partitionable removeable devices or empty devices */
784 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
785 (sgp->flags & GENHD_FL_REMOVABLE)))
787 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
791 xa_for_each(&sgp->part_tbl, idx, part) {
792 if (!bdev_nr_sectors(part))
794 seq_printf(seqf, "%4d %7d %10llu %pg\n",
795 MAJOR(part->bd_dev), MINOR(part->bd_dev),
796 bdev_nr_sectors(part) >> 1, part);
802 static const struct seq_operations partitions_op = {
803 .start = show_partition_start,
804 .next = disk_seqf_next,
805 .stop = disk_seqf_stop,
806 .show = show_partition
810 static int __init genhd_device_init(void)
814 block_class.dev_kobj = sysfs_dev_block_kobj;
815 error = class_register(&block_class);
820 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
822 /* create top-level block dir */
823 if (!sysfs_deprecated)
824 block_depr = kobject_create_and_add("block", NULL);
828 subsys_initcall(genhd_device_init);
830 static ssize_t disk_range_show(struct device *dev,
831 struct device_attribute *attr, char *buf)
833 struct gendisk *disk = dev_to_disk(dev);
835 return sprintf(buf, "%d\n", disk->minors);
838 static ssize_t disk_ext_range_show(struct device *dev,
839 struct device_attribute *attr, char *buf)
841 struct gendisk *disk = dev_to_disk(dev);
843 return sprintf(buf, "%d\n", disk_max_parts(disk));
846 static ssize_t disk_removable_show(struct device *dev,
847 struct device_attribute *attr, char *buf)
849 struct gendisk *disk = dev_to_disk(dev);
851 return sprintf(buf, "%d\n",
852 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
855 static ssize_t disk_hidden_show(struct device *dev,
856 struct device_attribute *attr, char *buf)
858 struct gendisk *disk = dev_to_disk(dev);
860 return sprintf(buf, "%d\n",
861 (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
864 static ssize_t disk_ro_show(struct device *dev,
865 struct device_attribute *attr, char *buf)
867 struct gendisk *disk = dev_to_disk(dev);
869 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
872 ssize_t part_size_show(struct device *dev,
873 struct device_attribute *attr, char *buf)
875 return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev)));
878 ssize_t part_stat_show(struct device *dev,
879 struct device_attribute *attr, char *buf)
881 struct block_device *bdev = dev_to_bdev(dev);
882 struct request_queue *q = bdev->bd_disk->queue;
883 struct disk_stats stat;
884 unsigned int inflight;
886 part_stat_read_all(bdev, &stat);
888 inflight = blk_mq_in_flight(q, bdev);
890 inflight = part_in_flight(bdev);
893 "%8lu %8lu %8llu %8u "
894 "%8lu %8lu %8llu %8u "
896 "%8lu %8lu %8llu %8u "
900 stat.merges[STAT_READ],
901 (unsigned long long)stat.sectors[STAT_READ],
902 (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC),
903 stat.ios[STAT_WRITE],
904 stat.merges[STAT_WRITE],
905 (unsigned long long)stat.sectors[STAT_WRITE],
906 (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC),
908 jiffies_to_msecs(stat.io_ticks),
909 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
910 stat.nsecs[STAT_WRITE] +
911 stat.nsecs[STAT_DISCARD] +
912 stat.nsecs[STAT_FLUSH],
914 stat.ios[STAT_DISCARD],
915 stat.merges[STAT_DISCARD],
916 (unsigned long long)stat.sectors[STAT_DISCARD],
917 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC),
918 stat.ios[STAT_FLUSH],
919 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC));
922 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
925 struct block_device *bdev = dev_to_bdev(dev);
926 struct request_queue *q = bdev->bd_disk->queue;
927 unsigned int inflight[2];
930 blk_mq_in_flight_rw(q, bdev, inflight);
932 part_in_flight_rw(bdev, inflight);
934 return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
937 static ssize_t disk_capability_show(struct device *dev,
938 struct device_attribute *attr, char *buf)
940 struct gendisk *disk = dev_to_disk(dev);
942 return sprintf(buf, "%x\n", disk->flags);
945 static ssize_t disk_alignment_offset_show(struct device *dev,
946 struct device_attribute *attr,
949 struct gendisk *disk = dev_to_disk(dev);
951 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
954 static ssize_t disk_discard_alignment_show(struct device *dev,
955 struct device_attribute *attr,
958 struct gendisk *disk = dev_to_disk(dev);
960 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
963 static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
964 static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
965 static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
966 static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
967 static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
968 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
969 static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
970 static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
971 static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
972 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
973 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
974 static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
976 #ifdef CONFIG_FAIL_MAKE_REQUEST
977 ssize_t part_fail_show(struct device *dev,
978 struct device_attribute *attr, char *buf)
980 return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail);
983 ssize_t part_fail_store(struct device *dev,
984 struct device_attribute *attr,
985 const char *buf, size_t count)
989 if (count > 0 && sscanf(buf, "%d", &i) > 0)
990 dev_to_bdev(dev)->bd_make_it_fail = i;
995 static struct device_attribute dev_attr_fail =
996 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
997 #endif /* CONFIG_FAIL_MAKE_REQUEST */
999 #ifdef CONFIG_FAIL_IO_TIMEOUT
1000 static struct device_attribute dev_attr_fail_timeout =
1001 __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
1004 static struct attribute *disk_attrs[] = {
1005 &dev_attr_range.attr,
1006 &dev_attr_ext_range.attr,
1007 &dev_attr_removable.attr,
1008 &dev_attr_hidden.attr,
1010 &dev_attr_size.attr,
1011 &dev_attr_alignment_offset.attr,
1012 &dev_attr_discard_alignment.attr,
1013 &dev_attr_capability.attr,
1014 &dev_attr_stat.attr,
1015 &dev_attr_inflight.attr,
1016 &dev_attr_badblocks.attr,
1017 &dev_attr_events.attr,
1018 &dev_attr_events_async.attr,
1019 &dev_attr_events_poll_msecs.attr,
1020 #ifdef CONFIG_FAIL_MAKE_REQUEST
1021 &dev_attr_fail.attr,
1023 #ifdef CONFIG_FAIL_IO_TIMEOUT
1024 &dev_attr_fail_timeout.attr,
1029 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1031 struct device *dev = container_of(kobj, typeof(*dev), kobj);
1032 struct gendisk *disk = dev_to_disk(dev);
1034 if (a == &dev_attr_badblocks.attr && !disk->bb)
1039 static struct attribute_group disk_attr_group = {
1040 .attrs = disk_attrs,
1041 .is_visible = disk_visible,
1044 static const struct attribute_group *disk_attr_groups[] = {
1050 * disk_release - releases all allocated resources of the gendisk
1051 * @dev: the device representing this disk
1053 * This function releases all allocated resources of the gendisk.
1055 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1056 * assigned. Since the request_queue sits on top of the gendisk for these
1057 * drivers we also call blk_put_queue() for them, and we expect the
1058 * request_queue refcount to reach 0 at this point, and so the request_queue
1059 * will also be freed prior to the disk.
1061 * Context: can sleep
1063 static void disk_release(struct device *dev)
1065 struct gendisk *disk = dev_to_disk(dev);
1069 if (MAJOR(dev->devt) == BLOCK_EXT_MAJOR)
1070 blk_free_ext_minor(MINOR(dev->devt));
1071 disk_release_events(disk);
1072 kfree(disk->random);
1073 xa_destroy(&disk->part_tbl);
1074 if (test_bit(GD_QUEUE_REF, &disk->state) && disk->queue)
1075 blk_put_queue(disk->queue);
1076 iput(disk->part0->bd_inode); /* frees the disk */
1078 struct class block_class = {
1082 static char *block_devnode(struct device *dev, umode_t *mode,
1083 kuid_t *uid, kgid_t *gid)
1085 struct gendisk *disk = dev_to_disk(dev);
1087 if (disk->fops->devnode)
1088 return disk->fops->devnode(disk, mode);
1092 const struct device_type disk_type = {
1094 .groups = disk_attr_groups,
1095 .release = disk_release,
1096 .devnode = block_devnode,
1099 #ifdef CONFIG_PROC_FS
1101 * aggregate disk stat collector. Uses the same stats that the sysfs
1102 * entries do, above, but makes them available through one seq_file.
1104 * The output looks suspiciously like /proc/partitions with a bunch of
1107 static int diskstats_show(struct seq_file *seqf, void *v)
1109 struct gendisk *gp = v;
1110 struct block_device *hd;
1111 unsigned int inflight;
1112 struct disk_stats stat;
1116 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1117 seq_puts(seqf, "major minor name"
1118 " rio rmerge rsect ruse wio wmerge "
1119 "wsect wuse running use aveq"
1124 xa_for_each(&gp->part_tbl, idx, hd) {
1125 if (bdev_is_partition(hd) && !bdev_nr_sectors(hd))
1127 part_stat_read_all(hd, &stat);
1128 if (queue_is_mq(gp->queue))
1129 inflight = blk_mq_in_flight(gp->queue, hd);
1131 inflight = part_in_flight(hd);
1133 seq_printf(seqf, "%4d %7d %pg "
1140 MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd,
1141 stat.ios[STAT_READ],
1142 stat.merges[STAT_READ],
1143 stat.sectors[STAT_READ],
1144 (unsigned int)div_u64(stat.nsecs[STAT_READ],
1146 stat.ios[STAT_WRITE],
1147 stat.merges[STAT_WRITE],
1148 stat.sectors[STAT_WRITE],
1149 (unsigned int)div_u64(stat.nsecs[STAT_WRITE],
1152 jiffies_to_msecs(stat.io_ticks),
1153 (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1154 stat.nsecs[STAT_WRITE] +
1155 stat.nsecs[STAT_DISCARD] +
1156 stat.nsecs[STAT_FLUSH],
1158 stat.ios[STAT_DISCARD],
1159 stat.merges[STAT_DISCARD],
1160 stat.sectors[STAT_DISCARD],
1161 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD],
1163 stat.ios[STAT_FLUSH],
1164 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH],
1173 static const struct seq_operations diskstats_op = {
1174 .start = disk_seqf_start,
1175 .next = disk_seqf_next,
1176 .stop = disk_seqf_stop,
1177 .show = diskstats_show
1180 static int __init proc_genhd_init(void)
1182 proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1183 proc_create_seq("partitions", 0, NULL, &partitions_op);
1186 module_init(proc_genhd_init);
1187 #endif /* CONFIG_PROC_FS */
1189 dev_t part_devt(struct gendisk *disk, u8 partno)
1191 struct block_device *part;
1195 part = xa_load(&disk->part_tbl, partno);
1197 devt = part->bd_dev;
1203 dev_t blk_lookup_devt(const char *name, int partno)
1205 dev_t devt = MKDEV(0, 0);
1206 struct class_dev_iter iter;
1209 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1210 while ((dev = class_dev_iter_next(&iter))) {
1211 struct gendisk *disk = dev_to_disk(dev);
1213 if (strcmp(dev_name(dev), name))
1216 if (partno < disk->minors) {
1217 /* We need to return the right devno, even
1218 * if the partition doesn't exist yet.
1220 devt = MKDEV(MAJOR(dev->devt),
1221 MINOR(dev->devt) + partno);
1223 devt = part_devt(disk, partno);
1228 class_dev_iter_exit(&iter);
1232 struct gendisk *__alloc_disk_node(int minors, int node_id)
1234 struct gendisk *disk;
1236 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1240 disk->part0 = bdev_alloc(disk, 0);
1244 disk->node_id = node_id;
1245 mutex_init(&disk->open_mutex);
1246 xa_init(&disk->part_tbl);
1247 if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL))
1248 goto out_destroy_part_tbl;
1250 disk->minors = minors;
1251 rand_initialize_disk(disk);
1252 disk_to_dev(disk)->class = &block_class;
1253 disk_to_dev(disk)->type = &disk_type;
1254 device_initialize(disk_to_dev(disk));
1257 out_destroy_part_tbl:
1258 xa_destroy(&disk->part_tbl);
1259 iput(disk->part0->bd_inode);
1264 EXPORT_SYMBOL(__alloc_disk_node);
1266 struct gendisk *__blk_alloc_disk(int node)
1268 struct request_queue *q;
1269 struct gendisk *disk;
1271 q = blk_alloc_queue(node);
1275 disk = __alloc_disk_node(0, node);
1277 blk_cleanup_queue(q);
1283 EXPORT_SYMBOL(__blk_alloc_disk);
1286 * put_disk - decrements the gendisk refcount
1287 * @disk: the struct gendisk to decrement the refcount for
1289 * This decrements the refcount for the struct gendisk. When this reaches 0
1290 * we'll have disk_release() called.
1292 * Context: Any context, but the last reference must not be dropped from
1295 void put_disk(struct gendisk *disk)
1298 put_device(disk_to_dev(disk));
1300 EXPORT_SYMBOL(put_disk);
1303 * blk_cleanup_disk - shutdown a gendisk allocated by blk_alloc_disk
1304 * @disk: gendisk to shutdown
1306 * Mark the queue hanging off @disk DYING, drain all pending requests, then mark
1307 * the queue DEAD, destroy and put it and the gendisk structure.
1309 * Context: can sleep
1311 void blk_cleanup_disk(struct gendisk *disk)
1313 blk_cleanup_queue(disk->queue);
1316 EXPORT_SYMBOL(blk_cleanup_disk);
1318 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1320 char event[] = "DISK_RO=1";
1321 char *envp[] = { event, NULL };
1325 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1329 * set_disk_ro - set a gendisk read-only
1330 * @disk: gendisk to operate on
1331 * @read_only: %true to set the disk read-only, %false set the disk read/write
1333 * This function is used to indicate whether a given disk device should have its
1334 * read-only flag set. set_disk_ro() is typically used by device drivers to
1335 * indicate whether the underlying physical device is write-protected.
1337 void set_disk_ro(struct gendisk *disk, bool read_only)
1340 if (test_and_set_bit(GD_READ_ONLY, &disk->state))
1343 if (!test_and_clear_bit(GD_READ_ONLY, &disk->state))
1346 set_disk_ro_uevent(disk, read_only);
1348 EXPORT_SYMBOL(set_disk_ro);
1350 int bdev_read_only(struct block_device *bdev)
1352 return bdev->bd_read_only || get_disk_ro(bdev->bd_disk);
1354 EXPORT_SYMBOL(bdev_read_only);