1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *  gendisk handling
4  *
5  * Portions Copyright (C) 2020 Christoph Hellwig
6  */
7 
8 #include <linux/module.h>
9 #include <linux/ctype.h>
10 #include <linux/fs.h>
11 #include <linux/kdev_t.h>
12 #include <linux/kernel.h>
13 #include <linux/blkdev.h>
14 #include <linux/backing-dev.h>
15 #include <linux/init.h>
16 #include <linux/spinlock.h>
17 #include <linux/proc_fs.h>
18 #include <linux/seq_file.h>
19 #include <linux/slab.h>
20 #include <linux/kmod.h>
21 #include <linux/major.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>
27 #include <linux/part_stat.h>
28 #include "blk-throttle.h"
29 
30 #include "blk.h"
31 #include "blk-mq-sched.h"
32 #include "blk-rq-qos.h"
33 #include "blk-cgroup.h"
34 
35 static struct kobject *block_depr;
36 
37 /*
38  * Unique, monotonically increasing sequential number associated with block
39  * devices instances (i.e. incremented each time a device is attached).
40  * Associating uevents with block devices in userspace is difficult and racy:
41  * the uevent netlink socket is lossy, and on slow and overloaded systems has
42  * a very high latency.
43  * Block devices do not have exclusive owners in userspace, any process can set
44  * one up (e.g. loop devices). Moreover, device names can be reused (e.g. loop0
45  * can be reused again and again).
46  * A userspace process setting up a block device and watching for its events
47  * cannot thus reliably tell whether an event relates to the device it just set
48  * up or another earlier instance with the same name.
49  * This sequential number allows userspace processes to solve this problem, and
50  * uniquely associate an uevent to the lifetime to a device.
51  */
52 static atomic64_t diskseq;
53 
54 /* for extended dynamic devt allocation, currently only one major is used */
55 #define NR_EXT_DEVT		(1 << MINORBITS)
56 static DEFINE_IDA(ext_devt_ida);
57 
set_capacity(struct gendisk * disk,sector_t sectors)58 void set_capacity(struct gendisk *disk, sector_t sectors)
59 {
60 	struct block_device *bdev = disk->part0;
61 
62 	spin_lock(&bdev->bd_size_lock);
63 	i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
64 	bdev->bd_nr_sectors = sectors;
65 	spin_unlock(&bdev->bd_size_lock);
66 }
67 EXPORT_SYMBOL(set_capacity);
68 
69 /*
70  * Set disk capacity and notify if the size is not currently zero and will not
71  * be set to zero.  Returns true if a uevent was sent, otherwise false.
72  */
set_capacity_and_notify(struct gendisk * disk,sector_t size)73 bool set_capacity_and_notify(struct gendisk *disk, sector_t size)
74 {
75 	sector_t capacity = get_capacity(disk);
76 	char *envp[] = { "RESIZE=1", NULL };
77 
78 	set_capacity(disk, size);
79 
80 	/*
81 	 * Only print a message and send a uevent if the gendisk is user visible
82 	 * and alive.  This avoids spamming the log and udev when setting the
83 	 * initial capacity during probing.
84 	 */
85 	if (size == capacity ||
86 	    !disk_live(disk) ||
87 	    (disk->flags & GENHD_FL_HIDDEN))
88 		return false;
89 
90 	pr_info("%s: detected capacity change from %lld to %lld\n",
91 		disk->disk_name, capacity, size);
92 
93 	/*
94 	 * Historically we did not send a uevent for changes to/from an empty
95 	 * device.
96 	 */
97 	if (!capacity || !size)
98 		return false;
99 	kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
100 	return true;
101 }
102 EXPORT_SYMBOL_GPL(set_capacity_and_notify);
103 
104 /*
105  * Format the device name of the indicated block device into the supplied buffer
106  * and return a pointer to that same buffer for convenience.
107  *
108  * Note: do not use this in new code, use the %pg specifier to sprintf and
109  * printk insted.
110  */
bdevname(struct block_device * bdev,char * buf)111 const char *bdevname(struct block_device *bdev, char *buf)
112 {
113 	struct gendisk *hd = bdev->bd_disk;
114 	int partno = bdev->bd_partno;
115 
116 	if (!partno)
117 		snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name);
118 	else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1]))
119 		snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno);
120 	else
121 		snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno);
122 
123 	return buf;
124 }
125 EXPORT_SYMBOL(bdevname);
126 
part_stat_read_all(struct block_device * part,struct disk_stats * stat)127 static void part_stat_read_all(struct block_device *part,
128 		struct disk_stats *stat)
129 {
130 	int cpu;
131 
132 	memset(stat, 0, sizeof(struct disk_stats));
133 	for_each_possible_cpu(cpu) {
134 		struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu);
135 		int group;
136 
137 		for (group = 0; group < NR_STAT_GROUPS; group++) {
138 			stat->nsecs[group] += ptr->nsecs[group];
139 			stat->sectors[group] += ptr->sectors[group];
140 			stat->ios[group] += ptr->ios[group];
141 			stat->merges[group] += ptr->merges[group];
142 		}
143 
144 		stat->io_ticks += ptr->io_ticks;
145 	}
146 }
147 
part_in_flight(struct block_device * part)148 static unsigned int part_in_flight(struct block_device *part)
149 {
150 	unsigned int inflight = 0;
151 	int cpu;
152 
153 	for_each_possible_cpu(cpu) {
154 		inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) +
155 			    part_stat_local_read_cpu(part, in_flight[1], cpu);
156 	}
157 	if ((int)inflight < 0)
158 		inflight = 0;
159 
160 	return inflight;
161 }
162 
part_in_flight_rw(struct block_device * part,unsigned int inflight[2])163 static void part_in_flight_rw(struct block_device *part,
164 		unsigned int inflight[2])
165 {
166 	int cpu;
167 
168 	inflight[0] = 0;
169 	inflight[1] = 0;
170 	for_each_possible_cpu(cpu) {
171 		inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu);
172 		inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu);
173 	}
174 	if ((int)inflight[0] < 0)
175 		inflight[0] = 0;
176 	if ((int)inflight[1] < 0)
177 		inflight[1] = 0;
178 }
179 
180 /*
181  * Can be deleted altogether. Later.
182  *
183  */
184 #define BLKDEV_MAJOR_HASH_SIZE 255
185 static struct blk_major_name {
186 	struct blk_major_name *next;
187 	int major;
188 	char name[16];
189 #ifdef CONFIG_BLOCK_LEGACY_AUTOLOAD
190 	void (*probe)(dev_t devt);
191 #endif
192 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
193 static DEFINE_MUTEX(major_names_lock);
194 static DEFINE_SPINLOCK(major_names_spinlock);
195 
196 /* index in the above - for now: assume no multimajor ranges */
major_to_index(unsigned major)197 static inline int major_to_index(unsigned major)
198 {
199 	return major % BLKDEV_MAJOR_HASH_SIZE;
200 }
201 
202 #ifdef CONFIG_PROC_FS
blkdev_show(struct seq_file * seqf,off_t offset)203 void blkdev_show(struct seq_file *seqf, off_t offset)
204 {
205 	struct blk_major_name *dp;
206 
207 	spin_lock(&major_names_spinlock);
208 	for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
209 		if (dp->major == offset)
210 			seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
211 	spin_unlock(&major_names_spinlock);
212 }
213 #endif /* CONFIG_PROC_FS */
214 
215 /**
216  * __register_blkdev - register a new block device
217  *
218  * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
219  *         @major = 0, try to allocate any unused major number.
220  * @name: the name of the new block device as a zero terminated string
221  * @probe: pre-devtmpfs / pre-udev callback used to create disks when their
222  *	   pre-created device node is accessed. When a probe call uses
223  *	   add_disk() and it fails the driver must cleanup resources. This
224  *	   interface may soon be removed.
225  *
226  * The @name must be unique within the system.
227  *
228  * The return value depends on the @major input parameter:
229  *
230  *  - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
231  *    then the function returns zero on success, or a negative error code
232  *  - if any unused major number was requested with @major = 0 parameter
233  *    then the return value is the allocated major number in range
234  *    [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
235  *
236  * See Documentation/admin-guide/devices.txt for the list of allocated
237  * major numbers.
238  *
239  * Use register_blkdev instead for any new code.
240  */
__register_blkdev(unsigned int major,const char * name,void (* probe)(dev_t devt))241 int __register_blkdev(unsigned int major, const char *name,
242 		void (*probe)(dev_t devt))
243 {
244 	struct blk_major_name **n, *p;
245 	int index, ret = 0;
246 
247 	mutex_lock(&major_names_lock);
248 
249 	/* temporary */
250 	if (major == 0) {
251 		for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
252 			if (major_names[index] == NULL)
253 				break;
254 		}
255 
256 		if (index == 0) {
257 			printk("%s: failed to get major for %s\n",
258 			       __func__, name);
259 			ret = -EBUSY;
260 			goto out;
261 		}
262 		major = index;
263 		ret = major;
264 	}
265 
266 	if (major >= BLKDEV_MAJOR_MAX) {
267 		pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
268 		       __func__, major, BLKDEV_MAJOR_MAX-1, name);
269 
270 		ret = -EINVAL;
271 		goto out;
272 	}
273 
274 	p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
275 	if (p == NULL) {
276 		ret = -ENOMEM;
277 		goto out;
278 	}
279 
280 	p->major = major;
281 #ifdef CONFIG_BLOCK_LEGACY_AUTOLOAD
282 	p->probe = probe;
283 #endif
284 	strlcpy(p->name, name, sizeof(p->name));
285 	p->next = NULL;
286 	index = major_to_index(major);
287 
288 	spin_lock(&major_names_spinlock);
289 	for (n = &major_names[index]; *n; n = &(*n)->next) {
290 		if ((*n)->major == major)
291 			break;
292 	}
293 	if (!*n)
294 		*n = p;
295 	else
296 		ret = -EBUSY;
297 	spin_unlock(&major_names_spinlock);
298 
299 	if (ret < 0) {
300 		printk("register_blkdev: cannot get major %u for %s\n",
301 		       major, name);
302 		kfree(p);
303 	}
304 out:
305 	mutex_unlock(&major_names_lock);
306 	return ret;
307 }
308 EXPORT_SYMBOL(__register_blkdev);
309 
unregister_blkdev(unsigned int major,const char * name)310 void unregister_blkdev(unsigned int major, const char *name)
311 {
312 	struct blk_major_name **n;
313 	struct blk_major_name *p = NULL;
314 	int index = major_to_index(major);
315 
316 	mutex_lock(&major_names_lock);
317 	spin_lock(&major_names_spinlock);
318 	for (n = &major_names[index]; *n; n = &(*n)->next)
319 		if ((*n)->major == major)
320 			break;
321 	if (!*n || strcmp((*n)->name, name)) {
322 		WARN_ON(1);
323 	} else {
324 		p = *n;
325 		*n = p->next;
326 	}
327 	spin_unlock(&major_names_spinlock);
328 	mutex_unlock(&major_names_lock);
329 	kfree(p);
330 }
331 
332 EXPORT_SYMBOL(unregister_blkdev);
333 
blk_alloc_ext_minor(void)334 int blk_alloc_ext_minor(void)
335 {
336 	int idx;
337 
338 	idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT - 1, GFP_KERNEL);
339 	if (idx == -ENOSPC)
340 		return -EBUSY;
341 	return idx;
342 }
343 
blk_free_ext_minor(unsigned int minor)344 void blk_free_ext_minor(unsigned int minor)
345 {
346 	ida_free(&ext_devt_ida, minor);
347 }
348 
bdevt_str(dev_t devt,char * buf)349 static char *bdevt_str(dev_t devt, char *buf)
350 {
351 	if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
352 		char tbuf[BDEVT_SIZE];
353 		snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
354 		snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
355 	} else
356 		snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
357 
358 	return buf;
359 }
360 
disk_uevent(struct gendisk * disk,enum kobject_action action)361 void disk_uevent(struct gendisk *disk, enum kobject_action action)
362 {
363 	struct block_device *part;
364 	unsigned long idx;
365 
366 	rcu_read_lock();
367 	xa_for_each(&disk->part_tbl, idx, part) {
368 		if (bdev_is_partition(part) && !bdev_nr_sectors(part))
369 			continue;
370 		if (!kobject_get_unless_zero(&part->bd_device.kobj))
371 			continue;
372 
373 		rcu_read_unlock();
374 		kobject_uevent(bdev_kobj(part), action);
375 		put_device(&part->bd_device);
376 		rcu_read_lock();
377 	}
378 	rcu_read_unlock();
379 }
380 EXPORT_SYMBOL_GPL(disk_uevent);
381 
disk_scan_partitions(struct gendisk * disk,fmode_t mode)382 int disk_scan_partitions(struct gendisk *disk, fmode_t mode)
383 {
384 	struct block_device *bdev;
385 
386 	if (disk->flags & (GENHD_FL_NO_PART | GENHD_FL_HIDDEN))
387 		return -EINVAL;
388 	if (test_bit(GD_SUPPRESS_PART_SCAN, &disk->state))
389 		return -EINVAL;
390 	if (disk->open_partitions)
391 		return -EBUSY;
392 
393 	set_bit(GD_NEED_PART_SCAN, &disk->state);
394 	bdev = blkdev_get_by_dev(disk_devt(disk), mode, NULL);
395 	if (IS_ERR(bdev))
396 		return PTR_ERR(bdev);
397 	blkdev_put(bdev, mode);
398 	return 0;
399 }
400 
401 /**
402  * device_add_disk - add disk information to kernel list
403  * @parent: parent device for the disk
404  * @disk: per-device partitioning information
405  * @groups: Additional per-device sysfs groups
406  *
407  * This function registers the partitioning information in @disk
408  * with the kernel.
409  */
device_add_disk(struct device * parent,struct gendisk * disk,const struct attribute_group ** groups)410 int __must_check device_add_disk(struct device *parent, struct gendisk *disk,
411 				 const struct attribute_group **groups)
412 
413 {
414 	struct device *ddev = disk_to_dev(disk);
415 	int ret;
416 
417 	/* Only makes sense for bio-based to set ->poll_bio */
418 	if (queue_is_mq(disk->queue) && disk->fops->poll_bio)
419 		return -EINVAL;
420 
421 	/*
422 	 * The disk queue should now be all set with enough information about
423 	 * the device for the elevator code to pick an adequate default
424 	 * elevator if one is needed, that is, for devices requesting queue
425 	 * registration.
426 	 */
427 	elevator_init_mq(disk->queue);
428 
429 	/*
430 	 * If the driver provides an explicit major number it also must provide
431 	 * the number of minors numbers supported, and those will be used to
432 	 * setup the gendisk.
433 	 * Otherwise just allocate the device numbers for both the whole device
434 	 * and all partitions from the extended dev_t space.
435 	 */
436 	if (disk->major) {
437 		if (WARN_ON(!disk->minors))
438 			return -EINVAL;
439 
440 		if (disk->minors > DISK_MAX_PARTS) {
441 			pr_err("block: can't allocate more than %d partitions\n",
442 				DISK_MAX_PARTS);
443 			disk->minors = DISK_MAX_PARTS;
444 		}
445 		if (disk->first_minor + disk->minors > MINORMASK + 1)
446 			return -EINVAL;
447 	} else {
448 		if (WARN_ON(disk->minors))
449 			return -EINVAL;
450 
451 		ret = blk_alloc_ext_minor();
452 		if (ret < 0)
453 			return ret;
454 		disk->major = BLOCK_EXT_MAJOR;
455 		disk->first_minor = ret;
456 	}
457 
458 	/* delay uevents, until we scanned partition table */
459 	dev_set_uevent_suppress(ddev, 1);
460 
461 	ddev->parent = parent;
462 	ddev->groups = groups;
463 	dev_set_name(ddev, "%s", disk->disk_name);
464 	if (!(disk->flags & GENHD_FL_HIDDEN))
465 		ddev->devt = MKDEV(disk->major, disk->first_minor);
466 	ret = device_add(ddev);
467 	if (ret)
468 		goto out_free_ext_minor;
469 
470 	ret = disk_alloc_events(disk);
471 	if (ret)
472 		goto out_device_del;
473 
474 	if (!sysfs_deprecated) {
475 		ret = sysfs_create_link(block_depr, &ddev->kobj,
476 					kobject_name(&ddev->kobj));
477 		if (ret)
478 			goto out_device_del;
479 	}
480 
481 	/*
482 	 * avoid probable deadlock caused by allocating memory with
483 	 * GFP_KERNEL in runtime_resume callback of its all ancestor
484 	 * devices
485 	 */
486 	pm_runtime_set_memalloc_noio(ddev, true);
487 
488 	ret = blk_integrity_add(disk);
489 	if (ret)
490 		goto out_del_block_link;
491 
492 	disk->part0->bd_holder_dir =
493 		kobject_create_and_add("holders", &ddev->kobj);
494 	if (!disk->part0->bd_holder_dir) {
495 		ret = -ENOMEM;
496 		goto out_del_integrity;
497 	}
498 	disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
499 	if (!disk->slave_dir) {
500 		ret = -ENOMEM;
501 		goto out_put_holder_dir;
502 	}
503 
504 	ret = bd_register_pending_holders(disk);
505 	if (ret < 0)
506 		goto out_put_slave_dir;
507 
508 	ret = blk_register_queue(disk);
509 	if (ret)
510 		goto out_put_slave_dir;
511 
512 	if (!(disk->flags & GENHD_FL_HIDDEN)) {
513 		ret = bdi_register(disk->bdi, "%u:%u",
514 				   disk->major, disk->first_minor);
515 		if (ret)
516 			goto out_unregister_queue;
517 		bdi_set_owner(disk->bdi, ddev);
518 		ret = sysfs_create_link(&ddev->kobj,
519 					&disk->bdi->dev->kobj, "bdi");
520 		if (ret)
521 			goto out_unregister_bdi;
522 
523 		bdev_add(disk->part0, ddev->devt);
524 		if (get_capacity(disk))
525 			disk_scan_partitions(disk, FMODE_READ);
526 
527 		/*
528 		 * Announce the disk and partitions after all partitions are
529 		 * created. (for hidden disks uevents remain suppressed forever)
530 		 */
531 		dev_set_uevent_suppress(ddev, 0);
532 		disk_uevent(disk, KOBJ_ADD);
533 	}
534 
535 	disk_update_readahead(disk);
536 	disk_add_events(disk);
537 	set_bit(GD_ADDED, &disk->state);
538 	return 0;
539 
540 out_unregister_bdi:
541 	if (!(disk->flags & GENHD_FL_HIDDEN))
542 		bdi_unregister(disk->bdi);
543 out_unregister_queue:
544 	blk_unregister_queue(disk);
545 out_put_slave_dir:
546 	kobject_put(disk->slave_dir);
547 out_put_holder_dir:
548 	kobject_put(disk->part0->bd_holder_dir);
549 out_del_integrity:
550 	blk_integrity_del(disk);
551 out_del_block_link:
552 	if (!sysfs_deprecated)
553 		sysfs_remove_link(block_depr, dev_name(ddev));
554 out_device_del:
555 	device_del(ddev);
556 out_free_ext_minor:
557 	if (disk->major == BLOCK_EXT_MAJOR)
558 		blk_free_ext_minor(disk->first_minor);
559 	return ret;
560 }
561 EXPORT_SYMBOL(device_add_disk);
562 
563 /**
564  * blk_mark_disk_dead - mark a disk as dead
565  * @disk: disk to mark as dead
566  *
567  * Mark as disk as dead (e.g. surprise removed) and don't accept any new I/O
568  * to this disk.
569  */
blk_mark_disk_dead(struct gendisk * disk)570 void blk_mark_disk_dead(struct gendisk *disk)
571 {
572 	set_bit(GD_DEAD, &disk->state);
573 	blk_queue_start_drain(disk->queue);
574 }
575 EXPORT_SYMBOL_GPL(blk_mark_disk_dead);
576 
577 /**
578  * del_gendisk - remove the gendisk
579  * @disk: the struct gendisk to remove
580  *
581  * Removes the gendisk and all its associated resources. This deletes the
582  * partitions associated with the gendisk, and unregisters the associated
583  * request_queue.
584  *
585  * This is the counter to the respective __device_add_disk() call.
586  *
587  * The final removal of the struct gendisk happens when its refcount reaches 0
588  * with put_disk(), which should be called after del_gendisk(), if
589  * __device_add_disk() was used.
590  *
591  * Drivers exist which depend on the release of the gendisk to be synchronous,
592  * it should not be deferred.
593  *
594  * Context: can sleep
595  */
del_gendisk(struct gendisk * disk)596 void del_gendisk(struct gendisk *disk)
597 {
598 	struct request_queue *q = disk->queue;
599 
600 	might_sleep();
601 
602 	if (WARN_ON_ONCE(!disk_live(disk) && !(disk->flags & GENHD_FL_HIDDEN)))
603 		return;
604 
605 	blk_integrity_del(disk);
606 	disk_del_events(disk);
607 
608 	mutex_lock(&disk->open_mutex);
609 	remove_inode_hash(disk->part0->bd_inode);
610 	blk_drop_partitions(disk);
611 	mutex_unlock(&disk->open_mutex);
612 
613 	fsync_bdev(disk->part0);
614 	__invalidate_device(disk->part0, true);
615 
616 	/*
617 	 * Fail any new I/O.
618 	 */
619 	set_bit(GD_DEAD, &disk->state);
620 	set_capacity(disk, 0);
621 
622 	/*
623 	 * Prevent new I/O from crossing bio_queue_enter().
624 	 */
625 	blk_queue_start_drain(q);
626 	blk_mq_freeze_queue_wait(q);
627 
628 	if (!(disk->flags & GENHD_FL_HIDDEN)) {
629 		sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
630 
631 		/*
632 		 * Unregister bdi before releasing device numbers (as they can
633 		 * get reused and we'd get clashes in sysfs).
634 		 */
635 		bdi_unregister(disk->bdi);
636 	}
637 
638 	blk_unregister_queue(disk);
639 
640 	kobject_put(disk->part0->bd_holder_dir);
641 	kobject_put(disk->slave_dir);
642 
643 	part_stat_set_all(disk->part0, 0);
644 	disk->part0->bd_stamp = 0;
645 	if (!sysfs_deprecated)
646 		sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
647 	pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
648 	device_del(disk_to_dev(disk));
649 
650 	blk_throtl_cancel_bios(disk->queue);
651 
652 	blk_sync_queue(q);
653 	blk_flush_integrity();
654 	blk_mq_cancel_work_sync(q);
655 
656 	blk_mq_quiesce_queue(q);
657 	if (q->elevator) {
658 		mutex_lock(&q->sysfs_lock);
659 		elevator_exit(q);
660 		mutex_unlock(&q->sysfs_lock);
661 	}
662 	rq_qos_exit(q);
663 	blk_mq_unquiesce_queue(q);
664 
665 	/*
666 	 * Allow using passthrough request again after the queue is torn down.
667 	 */
668 	blk_queue_flag_clear(QUEUE_FLAG_INIT_DONE, q);
669 	__blk_mq_unfreeze_queue(q, true);
670 
671 }
672 EXPORT_SYMBOL(del_gendisk);
673 
674 /**
675  * invalidate_disk - invalidate the disk
676  * @disk: the struct gendisk to invalidate
677  *
678  * A helper to invalidates the disk. It will clean the disk's associated
679  * buffer/page caches and reset its internal states so that the disk
680  * can be reused by the drivers.
681  *
682  * Context: can sleep
683  */
invalidate_disk(struct gendisk * disk)684 void invalidate_disk(struct gendisk *disk)
685 {
686 	struct block_device *bdev = disk->part0;
687 
688 	invalidate_bdev(bdev);
689 	bdev->bd_inode->i_mapping->wb_err = 0;
690 	set_capacity(disk, 0);
691 }
692 EXPORT_SYMBOL(invalidate_disk);
693 
694 /* sysfs access to bad-blocks list. */
disk_badblocks_show(struct device * dev,struct device_attribute * attr,char * page)695 static ssize_t disk_badblocks_show(struct device *dev,
696 					struct device_attribute *attr,
697 					char *page)
698 {
699 	struct gendisk *disk = dev_to_disk(dev);
700 
701 	if (!disk->bb)
702 		return sprintf(page, "\n");
703 
704 	return badblocks_show(disk->bb, page, 0);
705 }
706 
disk_badblocks_store(struct device * dev,struct device_attribute * attr,const char * page,size_t len)707 static ssize_t disk_badblocks_store(struct device *dev,
708 					struct device_attribute *attr,
709 					const char *page, size_t len)
710 {
711 	struct gendisk *disk = dev_to_disk(dev);
712 
713 	if (!disk->bb)
714 		return -ENXIO;
715 
716 	return badblocks_store(disk->bb, page, len, 0);
717 }
718 
719 #ifdef CONFIG_BLOCK_LEGACY_AUTOLOAD
blk_request_module(dev_t devt)720 void blk_request_module(dev_t devt)
721 {
722 	unsigned int major = MAJOR(devt);
723 	struct blk_major_name **n;
724 
725 	mutex_lock(&major_names_lock);
726 	for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) {
727 		if ((*n)->major == major && (*n)->probe) {
728 			(*n)->probe(devt);
729 			mutex_unlock(&major_names_lock);
730 			return;
731 		}
732 	}
733 	mutex_unlock(&major_names_lock);
734 
735 	if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
736 		/* Make old-style 2.4 aliases work */
737 		request_module("block-major-%d", MAJOR(devt));
738 }
739 #endif /* CONFIG_BLOCK_LEGACY_AUTOLOAD */
740 
741 /*
742  * print a full list of all partitions - intended for places where the root
743  * filesystem can't be mounted and thus to give the victim some idea of what
744  * went wrong
745  */
printk_all_partitions(void)746 void __init printk_all_partitions(void)
747 {
748 	struct class_dev_iter iter;
749 	struct device *dev;
750 
751 	class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
752 	while ((dev = class_dev_iter_next(&iter))) {
753 		struct gendisk *disk = dev_to_disk(dev);
754 		struct block_device *part;
755 		char devt_buf[BDEVT_SIZE];
756 		unsigned long idx;
757 
758 		/*
759 		 * Don't show empty devices or things that have been
760 		 * suppressed
761 		 */
762 		if (get_capacity(disk) == 0 || (disk->flags & GENHD_FL_HIDDEN))
763 			continue;
764 
765 		/*
766 		 * Note, unlike /proc/partitions, I am showing the numbers in
767 		 * hex - the same format as the root= option takes.
768 		 */
769 		rcu_read_lock();
770 		xa_for_each(&disk->part_tbl, idx, part) {
771 			if (!bdev_nr_sectors(part))
772 				continue;
773 			printk("%s%s %10llu %pg %s",
774 			       bdev_is_partition(part) ? "  " : "",
775 			       bdevt_str(part->bd_dev, devt_buf),
776 			       bdev_nr_sectors(part) >> 1, part,
777 			       part->bd_meta_info ?
778 					part->bd_meta_info->uuid : "");
779 			if (bdev_is_partition(part))
780 				printk("\n");
781 			else if (dev->parent && dev->parent->driver)
782 				printk(" driver: %s\n",
783 					dev->parent->driver->name);
784 			else
785 				printk(" (driver?)\n");
786 		}
787 		rcu_read_unlock();
788 	}
789 	class_dev_iter_exit(&iter);
790 }
791 
792 #ifdef CONFIG_PROC_FS
793 /* iterator */
disk_seqf_start(struct seq_file * seqf,loff_t * pos)794 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
795 {
796 	loff_t skip = *pos;
797 	struct class_dev_iter *iter;
798 	struct device *dev;
799 
800 	iter = kmalloc(sizeof(*iter), GFP_KERNEL);
801 	if (!iter)
802 		return ERR_PTR(-ENOMEM);
803 
804 	seqf->private = iter;
805 	class_dev_iter_init(iter, &block_class, NULL, &disk_type);
806 	do {
807 		dev = class_dev_iter_next(iter);
808 		if (!dev)
809 			return NULL;
810 	} while (skip--);
811 
812 	return dev_to_disk(dev);
813 }
814 
disk_seqf_next(struct seq_file * seqf,void * v,loff_t * pos)815 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
816 {
817 	struct device *dev;
818 
819 	(*pos)++;
820 	dev = class_dev_iter_next(seqf->private);
821 	if (dev)
822 		return dev_to_disk(dev);
823 
824 	return NULL;
825 }
826 
disk_seqf_stop(struct seq_file * seqf,void * v)827 static void disk_seqf_stop(struct seq_file *seqf, void *v)
828 {
829 	struct class_dev_iter *iter = seqf->private;
830 
831 	/* stop is called even after start failed :-( */
832 	if (iter) {
833 		class_dev_iter_exit(iter);
834 		kfree(iter);
835 		seqf->private = NULL;
836 	}
837 }
838 
show_partition_start(struct seq_file * seqf,loff_t * pos)839 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
840 {
841 	void *p;
842 
843 	p = disk_seqf_start(seqf, pos);
844 	if (!IS_ERR_OR_NULL(p) && !*pos)
845 		seq_puts(seqf, "major minor  #blocks  name\n\n");
846 	return p;
847 }
848 
show_partition(struct seq_file * seqf,void * v)849 static int show_partition(struct seq_file *seqf, void *v)
850 {
851 	struct gendisk *sgp = v;
852 	struct block_device *part;
853 	unsigned long idx;
854 
855 	if (!get_capacity(sgp) || (sgp->flags & GENHD_FL_HIDDEN))
856 		return 0;
857 
858 	rcu_read_lock();
859 	xa_for_each(&sgp->part_tbl, idx, part) {
860 		if (!bdev_nr_sectors(part))
861 			continue;
862 		seq_printf(seqf, "%4d  %7d %10llu %pg\n",
863 			   MAJOR(part->bd_dev), MINOR(part->bd_dev),
864 			   bdev_nr_sectors(part) >> 1, part);
865 	}
866 	rcu_read_unlock();
867 	return 0;
868 }
869 
870 static const struct seq_operations partitions_op = {
871 	.start	= show_partition_start,
872 	.next	= disk_seqf_next,
873 	.stop	= disk_seqf_stop,
874 	.show	= show_partition
875 };
876 #endif
877 
genhd_device_init(void)878 static int __init genhd_device_init(void)
879 {
880 	int error;
881 
882 	block_class.dev_kobj = sysfs_dev_block_kobj;
883 	error = class_register(&block_class);
884 	if (unlikely(error))
885 		return error;
886 	blk_dev_init();
887 
888 	register_blkdev(BLOCK_EXT_MAJOR, "blkext");
889 
890 	/* create top-level block dir */
891 	if (!sysfs_deprecated)
892 		block_depr = kobject_create_and_add("block", NULL);
893 	return 0;
894 }
895 
896 subsys_initcall(genhd_device_init);
897 
disk_range_show(struct device * dev,struct device_attribute * attr,char * buf)898 static ssize_t disk_range_show(struct device *dev,
899 			       struct device_attribute *attr, char *buf)
900 {
901 	struct gendisk *disk = dev_to_disk(dev);
902 
903 	return sprintf(buf, "%d\n", disk->minors);
904 }
905 
disk_ext_range_show(struct device * dev,struct device_attribute * attr,char * buf)906 static ssize_t disk_ext_range_show(struct device *dev,
907 				   struct device_attribute *attr, char *buf)
908 {
909 	struct gendisk *disk = dev_to_disk(dev);
910 
911 	return sprintf(buf, "%d\n",
912 		(disk->flags & GENHD_FL_NO_PART) ? 1 : DISK_MAX_PARTS);
913 }
914 
disk_removable_show(struct device * dev,struct device_attribute * attr,char * buf)915 static ssize_t disk_removable_show(struct device *dev,
916 				   struct device_attribute *attr, char *buf)
917 {
918 	struct gendisk *disk = dev_to_disk(dev);
919 
920 	return sprintf(buf, "%d\n",
921 		       (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
922 }
923 
disk_hidden_show(struct device * dev,struct device_attribute * attr,char * buf)924 static ssize_t disk_hidden_show(struct device *dev,
925 				   struct device_attribute *attr, char *buf)
926 {
927 	struct gendisk *disk = dev_to_disk(dev);
928 
929 	return sprintf(buf, "%d\n",
930 		       (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
931 }
932 
disk_ro_show(struct device * dev,struct device_attribute * attr,char * buf)933 static ssize_t disk_ro_show(struct device *dev,
934 				   struct device_attribute *attr, char *buf)
935 {
936 	struct gendisk *disk = dev_to_disk(dev);
937 
938 	return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
939 }
940 
part_size_show(struct device * dev,struct device_attribute * attr,char * buf)941 ssize_t part_size_show(struct device *dev,
942 		       struct device_attribute *attr, char *buf)
943 {
944 	return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev)));
945 }
946 
part_stat_show(struct device * dev,struct device_attribute * attr,char * buf)947 ssize_t part_stat_show(struct device *dev,
948 		       struct device_attribute *attr, char *buf)
949 {
950 	struct block_device *bdev = dev_to_bdev(dev);
951 	struct request_queue *q = bdev_get_queue(bdev);
952 	struct disk_stats stat;
953 	unsigned int inflight;
954 
955 	if (queue_is_mq(q))
956 		inflight = blk_mq_in_flight(q, bdev);
957 	else
958 		inflight = part_in_flight(bdev);
959 
960 	if (inflight) {
961 		part_stat_lock();
962 		update_io_ticks(bdev, jiffies, true);
963 		part_stat_unlock();
964 	}
965 	part_stat_read_all(bdev, &stat);
966 	return sprintf(buf,
967 		"%8lu %8lu %8llu %8u "
968 		"%8lu %8lu %8llu %8u "
969 		"%8u %8u %8u "
970 		"%8lu %8lu %8llu %8u "
971 		"%8lu %8u"
972 		"\n",
973 		stat.ios[STAT_READ],
974 		stat.merges[STAT_READ],
975 		(unsigned long long)stat.sectors[STAT_READ],
976 		(unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC),
977 		stat.ios[STAT_WRITE],
978 		stat.merges[STAT_WRITE],
979 		(unsigned long long)stat.sectors[STAT_WRITE],
980 		(unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC),
981 		inflight,
982 		jiffies_to_msecs(stat.io_ticks),
983 		(unsigned int)div_u64(stat.nsecs[STAT_READ] +
984 				      stat.nsecs[STAT_WRITE] +
985 				      stat.nsecs[STAT_DISCARD] +
986 				      stat.nsecs[STAT_FLUSH],
987 						NSEC_PER_MSEC),
988 		stat.ios[STAT_DISCARD],
989 		stat.merges[STAT_DISCARD],
990 		(unsigned long long)stat.sectors[STAT_DISCARD],
991 		(unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC),
992 		stat.ios[STAT_FLUSH],
993 		(unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC));
994 }
995 
part_inflight_show(struct device * dev,struct device_attribute * attr,char * buf)996 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
997 			   char *buf)
998 {
999 	struct block_device *bdev = dev_to_bdev(dev);
1000 	struct request_queue *q = bdev_get_queue(bdev);
1001 	unsigned int inflight[2];
1002 
1003 	if (queue_is_mq(q))
1004 		blk_mq_in_flight_rw(q, bdev, inflight);
1005 	else
1006 		part_in_flight_rw(bdev, inflight);
1007 
1008 	return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
1009 }
1010 
disk_capability_show(struct device * dev,struct device_attribute * attr,char * buf)1011 static ssize_t disk_capability_show(struct device *dev,
1012 				    struct device_attribute *attr, char *buf)
1013 {
1014 	struct gendisk *disk = dev_to_disk(dev);
1015 
1016 	return sprintf(buf, "%x\n", disk->flags);
1017 }
1018 
disk_alignment_offset_show(struct device * dev,struct device_attribute * attr,char * buf)1019 static ssize_t disk_alignment_offset_show(struct device *dev,
1020 					  struct device_attribute *attr,
1021 					  char *buf)
1022 {
1023 	struct gendisk *disk = dev_to_disk(dev);
1024 
1025 	return sprintf(buf, "%d\n", bdev_alignment_offset(disk->part0));
1026 }
1027 
disk_discard_alignment_show(struct device * dev,struct device_attribute * attr,char * buf)1028 static ssize_t disk_discard_alignment_show(struct device *dev,
1029 					   struct device_attribute *attr,
1030 					   char *buf)
1031 {
1032 	struct gendisk *disk = dev_to_disk(dev);
1033 
1034 	return sprintf(buf, "%d\n", bdev_alignment_offset(disk->part0));
1035 }
1036 
diskseq_show(struct device * dev,struct device_attribute * attr,char * buf)1037 static ssize_t diskseq_show(struct device *dev,
1038 			    struct device_attribute *attr, char *buf)
1039 {
1040 	struct gendisk *disk = dev_to_disk(dev);
1041 
1042 	return sprintf(buf, "%llu\n", disk->diskseq);
1043 }
1044 
1045 static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
1046 static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
1047 static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
1048 static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
1049 static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
1050 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
1051 static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
1052 static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
1053 static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
1054 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
1055 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
1056 static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
1057 static DEVICE_ATTR(diskseq, 0444, diskseq_show, NULL);
1058 
1059 #ifdef CONFIG_FAIL_MAKE_REQUEST
part_fail_show(struct device * dev,struct device_attribute * attr,char * buf)1060 ssize_t part_fail_show(struct device *dev,
1061 		       struct device_attribute *attr, char *buf)
1062 {
1063 	return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail);
1064 }
1065 
part_fail_store(struct device * dev,struct device_attribute * attr,const char * buf,size_t count)1066 ssize_t part_fail_store(struct device *dev,
1067 			struct device_attribute *attr,
1068 			const char *buf, size_t count)
1069 {
1070 	int i;
1071 
1072 	if (count > 0 && sscanf(buf, "%d", &i) > 0)
1073 		dev_to_bdev(dev)->bd_make_it_fail = i;
1074 
1075 	return count;
1076 }
1077 
1078 static struct device_attribute dev_attr_fail =
1079 	__ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
1080 #endif /* CONFIG_FAIL_MAKE_REQUEST */
1081 
1082 #ifdef CONFIG_FAIL_IO_TIMEOUT
1083 static struct device_attribute dev_attr_fail_timeout =
1084 	__ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
1085 #endif
1086 
1087 static struct attribute *disk_attrs[] = {
1088 	&dev_attr_range.attr,
1089 	&dev_attr_ext_range.attr,
1090 	&dev_attr_removable.attr,
1091 	&dev_attr_hidden.attr,
1092 	&dev_attr_ro.attr,
1093 	&dev_attr_size.attr,
1094 	&dev_attr_alignment_offset.attr,
1095 	&dev_attr_discard_alignment.attr,
1096 	&dev_attr_capability.attr,
1097 	&dev_attr_stat.attr,
1098 	&dev_attr_inflight.attr,
1099 	&dev_attr_badblocks.attr,
1100 	&dev_attr_events.attr,
1101 	&dev_attr_events_async.attr,
1102 	&dev_attr_events_poll_msecs.attr,
1103 	&dev_attr_diskseq.attr,
1104 #ifdef CONFIG_FAIL_MAKE_REQUEST
1105 	&dev_attr_fail.attr,
1106 #endif
1107 #ifdef CONFIG_FAIL_IO_TIMEOUT
1108 	&dev_attr_fail_timeout.attr,
1109 #endif
1110 	NULL
1111 };
1112 
disk_visible(struct kobject * kobj,struct attribute * a,int n)1113 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1114 {
1115 	struct device *dev = container_of(kobj, typeof(*dev), kobj);
1116 	struct gendisk *disk = dev_to_disk(dev);
1117 
1118 	if (a == &dev_attr_badblocks.attr && !disk->bb)
1119 		return 0;
1120 	return a->mode;
1121 }
1122 
1123 static struct attribute_group disk_attr_group = {
1124 	.attrs = disk_attrs,
1125 	.is_visible = disk_visible,
1126 };
1127 
1128 static const struct attribute_group *disk_attr_groups[] = {
1129 	&disk_attr_group,
1130 	NULL
1131 };
1132 
1133 /**
1134  * disk_release - releases all allocated resources of the gendisk
1135  * @dev: the device representing this disk
1136  *
1137  * This function releases all allocated resources of the gendisk.
1138  *
1139  * Drivers which used __device_add_disk() have a gendisk with a request_queue
1140  * assigned. Since the request_queue sits on top of the gendisk for these
1141  * drivers we also call blk_put_queue() for them, and we expect the
1142  * request_queue refcount to reach 0 at this point, and so the request_queue
1143  * will also be freed prior to the disk.
1144  *
1145  * Context: can sleep
1146  */
disk_release(struct device * dev)1147 static void disk_release(struct device *dev)
1148 {
1149 	struct gendisk *disk = dev_to_disk(dev);
1150 
1151 	might_sleep();
1152 	WARN_ON_ONCE(disk_live(disk));
1153 
1154 	blkcg_exit_queue(disk->queue);
1155 
1156 	disk_release_events(disk);
1157 	kfree(disk->random);
1158 	xa_destroy(&disk->part_tbl);
1159 
1160 	disk->queue->disk = NULL;
1161 	blk_put_queue(disk->queue);
1162 
1163 	if (test_bit(GD_ADDED, &disk->state) && disk->fops->free_disk)
1164 		disk->fops->free_disk(disk);
1165 
1166 	iput(disk->part0->bd_inode);	/* frees the disk */
1167 }
1168 
block_uevent(struct device * dev,struct kobj_uevent_env * env)1169 static int block_uevent(struct device *dev, struct kobj_uevent_env *env)
1170 {
1171 	struct gendisk *disk = dev_to_disk(dev);
1172 
1173 	return add_uevent_var(env, "DISKSEQ=%llu", disk->diskseq);
1174 }
1175 
1176 struct class block_class = {
1177 	.name		= "block",
1178 	.dev_uevent	= block_uevent,
1179 };
1180 
block_devnode(struct device * dev,umode_t * mode,kuid_t * uid,kgid_t * gid)1181 static char *block_devnode(struct device *dev, umode_t *mode,
1182 			   kuid_t *uid, kgid_t *gid)
1183 {
1184 	struct gendisk *disk = dev_to_disk(dev);
1185 
1186 	if (disk->fops->devnode)
1187 		return disk->fops->devnode(disk, mode);
1188 	return NULL;
1189 }
1190 
1191 const struct device_type disk_type = {
1192 	.name		= "disk",
1193 	.groups		= disk_attr_groups,
1194 	.release	= disk_release,
1195 	.devnode	= block_devnode,
1196 };
1197 
1198 #ifdef CONFIG_PROC_FS
1199 /*
1200  * aggregate disk stat collector.  Uses the same stats that the sysfs
1201  * entries do, above, but makes them available through one seq_file.
1202  *
1203  * The output looks suspiciously like /proc/partitions with a bunch of
1204  * extra fields.
1205  */
diskstats_show(struct seq_file * seqf,void * v)1206 static int diskstats_show(struct seq_file *seqf, void *v)
1207 {
1208 	struct gendisk *gp = v;
1209 	struct block_device *hd;
1210 	unsigned int inflight;
1211 	struct disk_stats stat;
1212 	unsigned long idx;
1213 
1214 	/*
1215 	if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1216 		seq_puts(seqf,	"major minor name"
1217 				"     rio rmerge rsect ruse wio wmerge "
1218 				"wsect wuse running use aveq"
1219 				"\n\n");
1220 	*/
1221 
1222 	rcu_read_lock();
1223 	xa_for_each(&gp->part_tbl, idx, hd) {
1224 		if (bdev_is_partition(hd) && !bdev_nr_sectors(hd))
1225 			continue;
1226 		if (queue_is_mq(gp->queue))
1227 			inflight = blk_mq_in_flight(gp->queue, hd);
1228 		else
1229 			inflight = part_in_flight(hd);
1230 
1231 		if (inflight) {
1232 			part_stat_lock();
1233 			update_io_ticks(hd, jiffies, true);
1234 			part_stat_unlock();
1235 		}
1236 		part_stat_read_all(hd, &stat);
1237 		seq_printf(seqf, "%4d %7d %pg "
1238 			   "%lu %lu %lu %u "
1239 			   "%lu %lu %lu %u "
1240 			   "%u %u %u "
1241 			   "%lu %lu %lu %u "
1242 			   "%lu %u"
1243 			   "\n",
1244 			   MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd,
1245 			   stat.ios[STAT_READ],
1246 			   stat.merges[STAT_READ],
1247 			   stat.sectors[STAT_READ],
1248 			   (unsigned int)div_u64(stat.nsecs[STAT_READ],
1249 							NSEC_PER_MSEC),
1250 			   stat.ios[STAT_WRITE],
1251 			   stat.merges[STAT_WRITE],
1252 			   stat.sectors[STAT_WRITE],
1253 			   (unsigned int)div_u64(stat.nsecs[STAT_WRITE],
1254 							NSEC_PER_MSEC),
1255 			   inflight,
1256 			   jiffies_to_msecs(stat.io_ticks),
1257 			   (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1258 						 stat.nsecs[STAT_WRITE] +
1259 						 stat.nsecs[STAT_DISCARD] +
1260 						 stat.nsecs[STAT_FLUSH],
1261 							NSEC_PER_MSEC),
1262 			   stat.ios[STAT_DISCARD],
1263 			   stat.merges[STAT_DISCARD],
1264 			   stat.sectors[STAT_DISCARD],
1265 			   (unsigned int)div_u64(stat.nsecs[STAT_DISCARD],
1266 						 NSEC_PER_MSEC),
1267 			   stat.ios[STAT_FLUSH],
1268 			   (unsigned int)div_u64(stat.nsecs[STAT_FLUSH],
1269 						 NSEC_PER_MSEC)
1270 			);
1271 	}
1272 	rcu_read_unlock();
1273 
1274 	return 0;
1275 }
1276 
1277 static const struct seq_operations diskstats_op = {
1278 	.start	= disk_seqf_start,
1279 	.next	= disk_seqf_next,
1280 	.stop	= disk_seqf_stop,
1281 	.show	= diskstats_show
1282 };
1283 
proc_genhd_init(void)1284 static int __init proc_genhd_init(void)
1285 {
1286 	proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1287 	proc_create_seq("partitions", 0, NULL, &partitions_op);
1288 	return 0;
1289 }
1290 module_init(proc_genhd_init);
1291 #endif /* CONFIG_PROC_FS */
1292 
part_devt(struct gendisk * disk,u8 partno)1293 dev_t part_devt(struct gendisk *disk, u8 partno)
1294 {
1295 	struct block_device *part;
1296 	dev_t devt = 0;
1297 
1298 	rcu_read_lock();
1299 	part = xa_load(&disk->part_tbl, partno);
1300 	if (part)
1301 		devt = part->bd_dev;
1302 	rcu_read_unlock();
1303 
1304 	return devt;
1305 }
1306 
blk_lookup_devt(const char * name,int partno)1307 dev_t blk_lookup_devt(const char *name, int partno)
1308 {
1309 	dev_t devt = MKDEV(0, 0);
1310 	struct class_dev_iter iter;
1311 	struct device *dev;
1312 
1313 	class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1314 	while ((dev = class_dev_iter_next(&iter))) {
1315 		struct gendisk *disk = dev_to_disk(dev);
1316 
1317 		if (strcmp(dev_name(dev), name))
1318 			continue;
1319 
1320 		if (partno < disk->minors) {
1321 			/* We need to return the right devno, even
1322 			 * if the partition doesn't exist yet.
1323 			 */
1324 			devt = MKDEV(MAJOR(dev->devt),
1325 				     MINOR(dev->devt) + partno);
1326 		} else {
1327 			devt = part_devt(disk, partno);
1328 			if (devt)
1329 				break;
1330 		}
1331 	}
1332 	class_dev_iter_exit(&iter);
1333 	return devt;
1334 }
1335 
__alloc_disk_node(struct request_queue * q,int node_id,struct lock_class_key * lkclass)1336 struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id,
1337 		struct lock_class_key *lkclass)
1338 {
1339 	struct gendisk *disk;
1340 
1341 	if (!blk_get_queue(q))
1342 		return NULL;
1343 
1344 	disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1345 	if (!disk)
1346 		goto out_put_queue;
1347 
1348 	disk->bdi = bdi_alloc(node_id);
1349 	if (!disk->bdi)
1350 		goto out_free_disk;
1351 
1352 	/* bdev_alloc() might need the queue, set before the first call */
1353 	disk->queue = q;
1354 
1355 	disk->part0 = bdev_alloc(disk, 0);
1356 	if (!disk->part0)
1357 		goto out_free_bdi;
1358 
1359 	disk->node_id = node_id;
1360 	mutex_init(&disk->open_mutex);
1361 	xa_init(&disk->part_tbl);
1362 	if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL))
1363 		goto out_destroy_part_tbl;
1364 
1365 	if (blkcg_init_queue(q))
1366 		goto out_erase_part0;
1367 
1368 	rand_initialize_disk(disk);
1369 	disk_to_dev(disk)->class = &block_class;
1370 	disk_to_dev(disk)->type = &disk_type;
1371 	device_initialize(disk_to_dev(disk));
1372 	inc_diskseq(disk);
1373 	q->disk = disk;
1374 	lockdep_init_map(&disk->lockdep_map, "(bio completion)", lkclass, 0);
1375 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
1376 	INIT_LIST_HEAD(&disk->slave_bdevs);
1377 #endif
1378 	return disk;
1379 
1380 out_erase_part0:
1381 	xa_erase(&disk->part_tbl, 0);
1382 out_destroy_part_tbl:
1383 	xa_destroy(&disk->part_tbl);
1384 	disk->part0->bd_disk = NULL;
1385 	iput(disk->part0->bd_inode);
1386 out_free_bdi:
1387 	bdi_put(disk->bdi);
1388 out_free_disk:
1389 	kfree(disk);
1390 out_put_queue:
1391 	blk_put_queue(q);
1392 	return NULL;
1393 }
1394 EXPORT_SYMBOL(__alloc_disk_node);
1395 
__blk_alloc_disk(int node,struct lock_class_key * lkclass)1396 struct gendisk *__blk_alloc_disk(int node, struct lock_class_key *lkclass)
1397 {
1398 	struct request_queue *q;
1399 	struct gendisk *disk;
1400 
1401 	q = blk_alloc_queue(node, false);
1402 	if (!q)
1403 		return NULL;
1404 
1405 	disk = __alloc_disk_node(q, node, lkclass);
1406 	if (!disk) {
1407 		blk_cleanup_queue(q);
1408 		return NULL;
1409 	}
1410 	return disk;
1411 }
1412 EXPORT_SYMBOL(__blk_alloc_disk);
1413 
1414 /**
1415  * put_disk - decrements the gendisk refcount
1416  * @disk: the struct gendisk to decrement the refcount for
1417  *
1418  * This decrements the refcount for the struct gendisk. When this reaches 0
1419  * we'll have disk_release() called.
1420  *
1421  * Context: Any context, but the last reference must not be dropped from
1422  *          atomic context.
1423  */
put_disk(struct gendisk * disk)1424 void put_disk(struct gendisk *disk)
1425 {
1426 	if (disk)
1427 		put_device(disk_to_dev(disk));
1428 }
1429 EXPORT_SYMBOL(put_disk);
1430 
1431 /**
1432  * blk_cleanup_disk - shutdown a gendisk allocated by blk_alloc_disk
1433  * @disk: gendisk to shutdown
1434  *
1435  * Mark the queue hanging off @disk DYING, drain all pending requests, then mark
1436  * the queue DEAD, destroy and put it and the gendisk structure.
1437  *
1438  * Context: can sleep
1439  */
blk_cleanup_disk(struct gendisk * disk)1440 void blk_cleanup_disk(struct gendisk *disk)
1441 {
1442 	blk_cleanup_queue(disk->queue);
1443 	put_disk(disk);
1444 }
1445 EXPORT_SYMBOL(blk_cleanup_disk);
1446 
set_disk_ro_uevent(struct gendisk * gd,int ro)1447 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1448 {
1449 	char event[] = "DISK_RO=1";
1450 	char *envp[] = { event, NULL };
1451 
1452 	if (!ro)
1453 		event[8] = '0';
1454 	kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1455 }
1456 
1457 /**
1458  * set_disk_ro - set a gendisk read-only
1459  * @disk:	gendisk to operate on
1460  * @read_only:	%true to set the disk read-only, %false set the disk read/write
1461  *
1462  * This function is used to indicate whether a given disk device should have its
1463  * read-only flag set. set_disk_ro() is typically used by device drivers to
1464  * indicate whether the underlying physical device is write-protected.
1465  */
set_disk_ro(struct gendisk * disk,bool read_only)1466 void set_disk_ro(struct gendisk *disk, bool read_only)
1467 {
1468 	if (read_only) {
1469 		if (test_and_set_bit(GD_READ_ONLY, &disk->state))
1470 			return;
1471 	} else {
1472 		if (!test_and_clear_bit(GD_READ_ONLY, &disk->state))
1473 			return;
1474 	}
1475 	set_disk_ro_uevent(disk, read_only);
1476 }
1477 EXPORT_SYMBOL(set_disk_ro);
1478 
inc_diskseq(struct gendisk * disk)1479 void inc_diskseq(struct gendisk *disk)
1480 {
1481 	disk->diskseq = atomic64_inc_return(&diskseq);
1482 }
1483