xref: /openbmc/linux/block/genhd.c (revision 827beb77)
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/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/major.h>
23 #include <linux/mutex.h>
24 #include <linux/idr.h>
25 #include <linux/log2.h>
26 #include <linux/pm_runtime.h>
27 #include <linux/badblocks.h>
28 
29 #include "blk.h"
30 #include "blk-rq-qos.h"
31 
32 static struct kobject *block_depr;
33 
34 /*
35  * Unique, monotonically increasing sequential number associated with block
36  * devices instances (i.e. incremented each time a device is attached).
37  * Associating uevents with block devices in userspace is difficult and racy:
38  * the uevent netlink socket is lossy, and on slow and overloaded systems has
39  * a very high latency.
40  * Block devices do not have exclusive owners in userspace, any process can set
41  * one up (e.g. loop devices). Moreover, device names can be reused (e.g. loop0
42  * can be reused again and again).
43  * A userspace process setting up a block device and watching for its events
44  * cannot thus reliably tell whether an event relates to the device it just set
45  * up or another earlier instance with the same name.
46  * This sequential number allows userspace processes to solve this problem, and
47  * uniquely associate an uevent to the lifetime to a device.
48  */
49 static atomic64_t diskseq;
50 
51 /* for extended dynamic devt allocation, currently only one major is used */
52 #define NR_EXT_DEVT		(1 << MINORBITS)
53 static DEFINE_IDA(ext_devt_ida);
54 
55 void set_capacity(struct gendisk *disk, sector_t sectors)
56 {
57 	struct block_device *bdev = disk->part0;
58 
59 	spin_lock(&bdev->bd_size_lock);
60 	i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
61 	bdev->bd_nr_sectors = sectors;
62 	spin_unlock(&bdev->bd_size_lock);
63 }
64 EXPORT_SYMBOL(set_capacity);
65 
66 /*
67  * Set disk capacity and notify if the size is not currently zero and will not
68  * be set to zero.  Returns true if a uevent was sent, otherwise false.
69  */
70 bool set_capacity_and_notify(struct gendisk *disk, sector_t size)
71 {
72 	sector_t capacity = get_capacity(disk);
73 	char *envp[] = { "RESIZE=1", NULL };
74 
75 	set_capacity(disk, size);
76 
77 	/*
78 	 * Only print a message and send a uevent if the gendisk is user visible
79 	 * and alive.  This avoids spamming the log and udev when setting the
80 	 * initial capacity during probing.
81 	 */
82 	if (size == capacity ||
83 	    !disk_live(disk) ||
84 	    (disk->flags & GENHD_FL_HIDDEN))
85 		return false;
86 
87 	pr_info("%s: detected capacity change from %lld to %lld\n",
88 		disk->disk_name, capacity, size);
89 
90 	/*
91 	 * Historically we did not send a uevent for changes to/from an empty
92 	 * device.
93 	 */
94 	if (!capacity || !size)
95 		return false;
96 	kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
97 	return true;
98 }
99 EXPORT_SYMBOL_GPL(set_capacity_and_notify);
100 
101 /*
102  * Format the device name of the indicated block device into the supplied buffer
103  * and return a pointer to that same buffer for convenience.
104  *
105  * Note: do not use this in new code, use the %pg specifier to sprintf and
106  * printk insted.
107  */
108 const char *bdevname(struct block_device *bdev, char *buf)
109 {
110 	struct gendisk *hd = bdev->bd_disk;
111 	int partno = bdev->bd_partno;
112 
113 	if (!partno)
114 		snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name);
115 	else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1]))
116 		snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno);
117 	else
118 		snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno);
119 
120 	return buf;
121 }
122 EXPORT_SYMBOL(bdevname);
123 
124 static void part_stat_read_all(struct block_device *part,
125 		struct disk_stats *stat)
126 {
127 	int cpu;
128 
129 	memset(stat, 0, sizeof(struct disk_stats));
130 	for_each_possible_cpu(cpu) {
131 		struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu);
132 		int group;
133 
134 		for (group = 0; group < NR_STAT_GROUPS; group++) {
135 			stat->nsecs[group] += ptr->nsecs[group];
136 			stat->sectors[group] += ptr->sectors[group];
137 			stat->ios[group] += ptr->ios[group];
138 			stat->merges[group] += ptr->merges[group];
139 		}
140 
141 		stat->io_ticks += ptr->io_ticks;
142 	}
143 }
144 
145 static unsigned int part_in_flight(struct block_device *part)
146 {
147 	unsigned int inflight = 0;
148 	int cpu;
149 
150 	for_each_possible_cpu(cpu) {
151 		inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) +
152 			    part_stat_local_read_cpu(part, in_flight[1], cpu);
153 	}
154 	if ((int)inflight < 0)
155 		inflight = 0;
156 
157 	return inflight;
158 }
159 
160 static void part_in_flight_rw(struct block_device *part,
161 		unsigned int inflight[2])
162 {
163 	int cpu;
164 
165 	inflight[0] = 0;
166 	inflight[1] = 0;
167 	for_each_possible_cpu(cpu) {
168 		inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu);
169 		inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu);
170 	}
171 	if ((int)inflight[0] < 0)
172 		inflight[0] = 0;
173 	if ((int)inflight[1] < 0)
174 		inflight[1] = 0;
175 }
176 
177 /*
178  * Can be deleted altogether. Later.
179  *
180  */
181 #define BLKDEV_MAJOR_HASH_SIZE 255
182 static struct blk_major_name {
183 	struct blk_major_name *next;
184 	int major;
185 	char name[16];
186 	void (*probe)(dev_t devt);
187 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
188 static DEFINE_MUTEX(major_names_lock);
189 static DEFINE_SPINLOCK(major_names_spinlock);
190 
191 /* index in the above - for now: assume no multimajor ranges */
192 static inline int major_to_index(unsigned major)
193 {
194 	return major % BLKDEV_MAJOR_HASH_SIZE;
195 }
196 
197 #ifdef CONFIG_PROC_FS
198 void blkdev_show(struct seq_file *seqf, off_t offset)
199 {
200 	struct blk_major_name *dp;
201 
202 	spin_lock(&major_names_spinlock);
203 	for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
204 		if (dp->major == offset)
205 			seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
206 	spin_unlock(&major_names_spinlock);
207 }
208 #endif /* CONFIG_PROC_FS */
209 
210 /**
211  * __register_blkdev - register a new block device
212  *
213  * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
214  *         @major = 0, try to allocate any unused major number.
215  * @name: the name of the new block device as a zero terminated string
216  * @probe: allback that is called on access to any minor number of @major
217  *
218  * The @name must be unique within the system.
219  *
220  * The return value depends on the @major input parameter:
221  *
222  *  - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
223  *    then the function returns zero on success, or a negative error code
224  *  - if any unused major number was requested with @major = 0 parameter
225  *    then the return value is the allocated major number in range
226  *    [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
227  *
228  * See Documentation/admin-guide/devices.txt for the list of allocated
229  * major numbers.
230  *
231  * Use register_blkdev instead for any new code.
232  */
233 int __register_blkdev(unsigned int major, const char *name,
234 		void (*probe)(dev_t devt))
235 {
236 	struct blk_major_name **n, *p;
237 	int index, ret = 0;
238 
239 	mutex_lock(&major_names_lock);
240 
241 	/* temporary */
242 	if (major == 0) {
243 		for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
244 			if (major_names[index] == NULL)
245 				break;
246 		}
247 
248 		if (index == 0) {
249 			printk("%s: failed to get major for %s\n",
250 			       __func__, name);
251 			ret = -EBUSY;
252 			goto out;
253 		}
254 		major = index;
255 		ret = major;
256 	}
257 
258 	if (major >= BLKDEV_MAJOR_MAX) {
259 		pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
260 		       __func__, major, BLKDEV_MAJOR_MAX-1, name);
261 
262 		ret = -EINVAL;
263 		goto out;
264 	}
265 
266 	p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
267 	if (p == NULL) {
268 		ret = -ENOMEM;
269 		goto out;
270 	}
271 
272 	p->major = major;
273 	p->probe = probe;
274 	strlcpy(p->name, name, sizeof(p->name));
275 	p->next = NULL;
276 	index = major_to_index(major);
277 
278 	spin_lock(&major_names_spinlock);
279 	for (n = &major_names[index]; *n; n = &(*n)->next) {
280 		if ((*n)->major == major)
281 			break;
282 	}
283 	if (!*n)
284 		*n = p;
285 	else
286 		ret = -EBUSY;
287 	spin_unlock(&major_names_spinlock);
288 
289 	if (ret < 0) {
290 		printk("register_blkdev: cannot get major %u for %s\n",
291 		       major, name);
292 		kfree(p);
293 	}
294 out:
295 	mutex_unlock(&major_names_lock);
296 	return ret;
297 }
298 EXPORT_SYMBOL(__register_blkdev);
299 
300 void unregister_blkdev(unsigned int major, const char *name)
301 {
302 	struct blk_major_name **n;
303 	struct blk_major_name *p = NULL;
304 	int index = major_to_index(major);
305 
306 	mutex_lock(&major_names_lock);
307 	spin_lock(&major_names_spinlock);
308 	for (n = &major_names[index]; *n; n = &(*n)->next)
309 		if ((*n)->major == major)
310 			break;
311 	if (!*n || strcmp((*n)->name, name)) {
312 		WARN_ON(1);
313 	} else {
314 		p = *n;
315 		*n = p->next;
316 	}
317 	spin_unlock(&major_names_spinlock);
318 	mutex_unlock(&major_names_lock);
319 	kfree(p);
320 }
321 
322 EXPORT_SYMBOL(unregister_blkdev);
323 
324 int blk_alloc_ext_minor(void)
325 {
326 	int idx;
327 
328 	idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT, GFP_KERNEL);
329 	if (idx == -ENOSPC)
330 		return -EBUSY;
331 	return idx;
332 }
333 
334 void blk_free_ext_minor(unsigned int minor)
335 {
336 	ida_free(&ext_devt_ida, minor);
337 }
338 
339 static char *bdevt_str(dev_t devt, char *buf)
340 {
341 	if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
342 		char tbuf[BDEVT_SIZE];
343 		snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
344 		snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
345 	} else
346 		snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
347 
348 	return buf;
349 }
350 
351 void disk_uevent(struct gendisk *disk, enum kobject_action action)
352 {
353 	struct block_device *part;
354 	unsigned long idx;
355 
356 	rcu_read_lock();
357 	xa_for_each(&disk->part_tbl, idx, part) {
358 		if (bdev_is_partition(part) && !bdev_nr_sectors(part))
359 			continue;
360 		if (!kobject_get_unless_zero(&part->bd_device.kobj))
361 			continue;
362 
363 		rcu_read_unlock();
364 		kobject_uevent(bdev_kobj(part), action);
365 		put_device(&part->bd_device);
366 		rcu_read_lock();
367 	}
368 	rcu_read_unlock();
369 }
370 EXPORT_SYMBOL_GPL(disk_uevent);
371 
372 static void disk_scan_partitions(struct gendisk *disk)
373 {
374 	struct block_device *bdev;
375 
376 	if (!get_capacity(disk) || !disk_part_scan_enabled(disk))
377 		return;
378 
379 	set_bit(GD_NEED_PART_SCAN, &disk->state);
380 	bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL);
381 	if (!IS_ERR(bdev))
382 		blkdev_put(bdev, FMODE_READ);
383 }
384 
385 /**
386  * device_add_disk - add disk information to kernel list
387  * @parent: parent device for the disk
388  * @disk: per-device partitioning information
389  * @groups: Additional per-device sysfs groups
390  *
391  * This function registers the partitioning information in @disk
392  * with the kernel.
393  */
394 int device_add_disk(struct device *parent, struct gendisk *disk,
395 		     const struct attribute_group **groups)
396 
397 {
398 	struct device *ddev = disk_to_dev(disk);
399 	int ret;
400 
401 	/*
402 	 * The disk queue should now be all set with enough information about
403 	 * the device for the elevator code to pick an adequate default
404 	 * elevator if one is needed, that is, for devices requesting queue
405 	 * registration.
406 	 */
407 	elevator_init_mq(disk->queue);
408 
409 	/*
410 	 * If the driver provides an explicit major number it also must provide
411 	 * the number of minors numbers supported, and those will be used to
412 	 * setup the gendisk.
413 	 * Otherwise just allocate the device numbers for both the whole device
414 	 * and all partitions from the extended dev_t space.
415 	 */
416 	if (disk->major) {
417 		if (WARN_ON(!disk->minors))
418 			return -EINVAL;
419 
420 		if (disk->minors > DISK_MAX_PARTS) {
421 			pr_err("block: can't allocate more than %d partitions\n",
422 				DISK_MAX_PARTS);
423 			disk->minors = DISK_MAX_PARTS;
424 		}
425 	} else {
426 		if (WARN_ON(disk->minors))
427 			return -EINVAL;
428 
429 		ret = blk_alloc_ext_minor();
430 		if (ret < 0)
431 			return ret;
432 		disk->major = BLOCK_EXT_MAJOR;
433 		disk->first_minor = ret;
434 		disk->flags |= GENHD_FL_EXT_DEVT;
435 	}
436 
437 	ret = disk_alloc_events(disk);
438 	if (ret)
439 		goto out_free_ext_minor;
440 
441 	/* delay uevents, until we scanned partition table */
442 	dev_set_uevent_suppress(ddev, 1);
443 
444 	ddev->parent = parent;
445 	ddev->groups = groups;
446 	dev_set_name(ddev, "%s", disk->disk_name);
447 	if (!(disk->flags & GENHD_FL_HIDDEN))
448 		ddev->devt = MKDEV(disk->major, disk->first_minor);
449 	ret = device_add(ddev);
450 	if (ret)
451 		goto out_disk_release_events;
452 	if (!sysfs_deprecated) {
453 		ret = sysfs_create_link(block_depr, &ddev->kobj,
454 					kobject_name(&ddev->kobj));
455 		if (ret)
456 			goto out_device_del;
457 	}
458 
459 	/*
460 	 * avoid probable deadlock caused by allocating memory with
461 	 * GFP_KERNEL in runtime_resume callback of its all ancestor
462 	 * devices
463 	 */
464 	pm_runtime_set_memalloc_noio(ddev, true);
465 
466 	ret = blk_integrity_add(disk);
467 	if (ret)
468 		goto out_del_block_link;
469 
470 	disk->part0->bd_holder_dir =
471 		kobject_create_and_add("holders", &ddev->kobj);
472 	if (!disk->part0->bd_holder_dir)
473 		goto out_del_integrity;
474 	disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
475 	if (!disk->slave_dir)
476 		goto out_put_holder_dir;
477 
478 	ret = bd_register_pending_holders(disk);
479 	if (ret < 0)
480 		goto out_put_slave_dir;
481 
482 	ret = blk_register_queue(disk);
483 	if (ret)
484 		goto out_put_slave_dir;
485 
486 	if (disk->flags & GENHD_FL_HIDDEN) {
487 		/*
488 		 * Don't let hidden disks show up in /proc/partitions,
489 		 * and don't bother scanning for partitions either.
490 		 */
491 		disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
492 		disk->flags |= GENHD_FL_NO_PART_SCAN;
493 	} else {
494 		ret = bdi_register(disk->bdi, "%u:%u",
495 				   disk->major, disk->first_minor);
496 		if (ret)
497 			goto out_unregister_queue;
498 		bdi_set_owner(disk->bdi, ddev);
499 		ret = sysfs_create_link(&ddev->kobj,
500 					&disk->bdi->dev->kobj, "bdi");
501 		if (ret)
502 			goto out_unregister_bdi;
503 
504 		bdev_add(disk->part0, ddev->devt);
505 		disk_scan_partitions(disk);
506 
507 		/*
508 		 * Announce the disk and partitions after all partitions are
509 		 * created. (for hidden disks uevents remain suppressed forever)
510 		 */
511 		dev_set_uevent_suppress(ddev, 0);
512 		disk_uevent(disk, KOBJ_ADD);
513 	}
514 
515 	disk_update_readahead(disk);
516 	disk_add_events(disk);
517 	return 0;
518 
519 out_unregister_bdi:
520 	if (!(disk->flags & GENHD_FL_HIDDEN))
521 		bdi_unregister(disk->bdi);
522 out_unregister_queue:
523 	blk_unregister_queue(disk);
524 out_put_slave_dir:
525 	kobject_put(disk->slave_dir);
526 out_put_holder_dir:
527 	kobject_put(disk->part0->bd_holder_dir);
528 out_del_integrity:
529 	blk_integrity_del(disk);
530 out_del_block_link:
531 	if (!sysfs_deprecated)
532 		sysfs_remove_link(block_depr, dev_name(ddev));
533 out_device_del:
534 	device_del(ddev);
535 out_disk_release_events:
536 	disk_release_events(disk);
537 out_free_ext_minor:
538 	if (disk->major == BLOCK_EXT_MAJOR)
539 		blk_free_ext_minor(disk->first_minor);
540 	return WARN_ON_ONCE(ret); /* keep until all callers handle errors */
541 }
542 EXPORT_SYMBOL(device_add_disk);
543 
544 /**
545  * del_gendisk - remove the gendisk
546  * @disk: the struct gendisk to remove
547  *
548  * Removes the gendisk and all its associated resources. This deletes the
549  * partitions associated with the gendisk, and unregisters the associated
550  * request_queue.
551  *
552  * This is the counter to the respective __device_add_disk() call.
553  *
554  * The final removal of the struct gendisk happens when its refcount reaches 0
555  * with put_disk(), which should be called after del_gendisk(), if
556  * __device_add_disk() was used.
557  *
558  * Drivers exist which depend on the release of the gendisk to be synchronous,
559  * it should not be deferred.
560  *
561  * Context: can sleep
562  */
563 void del_gendisk(struct gendisk *disk)
564 {
565 	struct request_queue *q = disk->queue;
566 
567 	might_sleep();
568 
569 	if (WARN_ON_ONCE(!disk_live(disk) && !(disk->flags & GENHD_FL_HIDDEN)))
570 		return;
571 
572 	blk_integrity_del(disk);
573 	disk_del_events(disk);
574 
575 	mutex_lock(&disk->open_mutex);
576 	remove_inode_hash(disk->part0->bd_inode);
577 	blk_drop_partitions(disk);
578 	mutex_unlock(&disk->open_mutex);
579 
580 	fsync_bdev(disk->part0);
581 	__invalidate_device(disk->part0, true);
582 
583 	/*
584 	 * Fail any new I/O.
585 	 */
586 	set_bit(GD_DEAD, &disk->state);
587 	set_capacity(disk, 0);
588 
589 	/*
590 	 * Prevent new I/O from crossing bio_queue_enter().
591 	 */
592 	blk_queue_start_drain(q);
593 
594 	if (!(disk->flags & GENHD_FL_HIDDEN)) {
595 		sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
596 
597 		/*
598 		 * Unregister bdi before releasing device numbers (as they can
599 		 * get reused and we'd get clashes in sysfs).
600 		 */
601 		bdi_unregister(disk->bdi);
602 	}
603 
604 	blk_unregister_queue(disk);
605 
606 	kobject_put(disk->part0->bd_holder_dir);
607 	kobject_put(disk->slave_dir);
608 
609 	part_stat_set_all(disk->part0, 0);
610 	disk->part0->bd_stamp = 0;
611 	if (!sysfs_deprecated)
612 		sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
613 	pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
614 	device_del(disk_to_dev(disk));
615 
616 	blk_mq_freeze_queue_wait(q);
617 
618 	rq_qos_exit(q);
619 	blk_sync_queue(q);
620 	blk_flush_integrity();
621 	/*
622 	 * Allow using passthrough request again after the queue is torn down.
623 	 */
624 	blk_queue_flag_clear(QUEUE_FLAG_INIT_DONE, q);
625 	__blk_mq_unfreeze_queue(q, true);
626 
627 }
628 EXPORT_SYMBOL(del_gendisk);
629 
630 /**
631  * invalidate_disk - invalidate the disk
632  * @disk: the struct gendisk to invalidate
633  *
634  * A helper to invalidates the disk. It will clean the disk's associated
635  * buffer/page caches and reset its internal states so that the disk
636  * can be reused by the drivers.
637  *
638  * Context: can sleep
639  */
640 void invalidate_disk(struct gendisk *disk)
641 {
642 	struct block_device *bdev = disk->part0;
643 
644 	invalidate_bdev(bdev);
645 	bdev->bd_inode->i_mapping->wb_err = 0;
646 	set_capacity(disk, 0);
647 }
648 EXPORT_SYMBOL(invalidate_disk);
649 
650 /* sysfs access to bad-blocks list. */
651 static ssize_t disk_badblocks_show(struct device *dev,
652 					struct device_attribute *attr,
653 					char *page)
654 {
655 	struct gendisk *disk = dev_to_disk(dev);
656 
657 	if (!disk->bb)
658 		return sprintf(page, "\n");
659 
660 	return badblocks_show(disk->bb, page, 0);
661 }
662 
663 static ssize_t disk_badblocks_store(struct device *dev,
664 					struct device_attribute *attr,
665 					const char *page, size_t len)
666 {
667 	struct gendisk *disk = dev_to_disk(dev);
668 
669 	if (!disk->bb)
670 		return -ENXIO;
671 
672 	return badblocks_store(disk->bb, page, len, 0);
673 }
674 
675 void blk_request_module(dev_t devt)
676 {
677 	unsigned int major = MAJOR(devt);
678 	struct blk_major_name **n;
679 
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) {
683 			(*n)->probe(devt);
684 			mutex_unlock(&major_names_lock);
685 			return;
686 		}
687 	}
688 	mutex_unlock(&major_names_lock);
689 
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));
693 }
694 
695 /*
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
698  * went wrong
699  */
700 void __init printk_all_partitions(void)
701 {
702 	struct class_dev_iter iter;
703 	struct device *dev;
704 
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];
710 		unsigned long idx;
711 
712 		/*
713 		 * Don't show empty devices or things that have been
714 		 * suppressed
715 		 */
716 		if (get_capacity(disk) == 0 ||
717 		    (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
718 			continue;
719 
720 		/*
721 		 * Note, unlike /proc/partitions, I am showing the numbers in
722 		 * hex - the same format as the root= option takes.
723 		 */
724 		rcu_read_lock();
725 		xa_for_each(&disk->part_tbl, idx, part) {
726 			if (!bdev_nr_sectors(part))
727 				continue;
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,
732 			       part->bd_meta_info ?
733 					part->bd_meta_info->uuid : "");
734 			if (bdev_is_partition(part))
735 				printk("\n");
736 			else if (dev->parent && dev->parent->driver)
737 				printk(" driver: %s\n",
738 					dev->parent->driver->name);
739 			else
740 				printk(" (driver?)\n");
741 		}
742 		rcu_read_unlock();
743 	}
744 	class_dev_iter_exit(&iter);
745 }
746 
747 #ifdef CONFIG_PROC_FS
748 /* iterator */
749 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
750 {
751 	loff_t skip = *pos;
752 	struct class_dev_iter *iter;
753 	struct device *dev;
754 
755 	iter = kmalloc(sizeof(*iter), GFP_KERNEL);
756 	if (!iter)
757 		return ERR_PTR(-ENOMEM);
758 
759 	seqf->private = iter;
760 	class_dev_iter_init(iter, &block_class, NULL, &disk_type);
761 	do {
762 		dev = class_dev_iter_next(iter);
763 		if (!dev)
764 			return NULL;
765 	} while (skip--);
766 
767 	return dev_to_disk(dev);
768 }
769 
770 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
771 {
772 	struct device *dev;
773 
774 	(*pos)++;
775 	dev = class_dev_iter_next(seqf->private);
776 	if (dev)
777 		return dev_to_disk(dev);
778 
779 	return NULL;
780 }
781 
782 static void disk_seqf_stop(struct seq_file *seqf, void *v)
783 {
784 	struct class_dev_iter *iter = seqf->private;
785 
786 	/* stop is called even after start failed :-( */
787 	if (iter) {
788 		class_dev_iter_exit(iter);
789 		kfree(iter);
790 		seqf->private = NULL;
791 	}
792 }
793 
794 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
795 {
796 	void *p;
797 
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");
801 	return p;
802 }
803 
804 static int show_partition(struct seq_file *seqf, void *v)
805 {
806 	struct gendisk *sgp = v;
807 	struct block_device *part;
808 	unsigned long idx;
809 
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)))
813 		return 0;
814 	if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
815 		return 0;
816 
817 	rcu_read_lock();
818 	xa_for_each(&sgp->part_tbl, idx, part) {
819 		if (!bdev_nr_sectors(part))
820 			continue;
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);
824 	}
825 	rcu_read_unlock();
826 	return 0;
827 }
828 
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
834 };
835 #endif
836 
837 static int __init genhd_device_init(void)
838 {
839 	int error;
840 
841 	block_class.dev_kobj = sysfs_dev_block_kobj;
842 	error = class_register(&block_class);
843 	if (unlikely(error))
844 		return error;
845 	blk_dev_init();
846 
847 	register_blkdev(BLOCK_EXT_MAJOR, "blkext");
848 
849 	/* create top-level block dir */
850 	if (!sysfs_deprecated)
851 		block_depr = kobject_create_and_add("block", NULL);
852 	return 0;
853 }
854 
855 subsys_initcall(genhd_device_init);
856 
857 static ssize_t disk_range_show(struct device *dev,
858 			       struct device_attribute *attr, char *buf)
859 {
860 	struct gendisk *disk = dev_to_disk(dev);
861 
862 	return sprintf(buf, "%d\n", disk->minors);
863 }
864 
865 static ssize_t disk_ext_range_show(struct device *dev,
866 				   struct device_attribute *attr, char *buf)
867 {
868 	struct gendisk *disk = dev_to_disk(dev);
869 
870 	return sprintf(buf, "%d\n", disk_max_parts(disk));
871 }
872 
873 static ssize_t disk_removable_show(struct device *dev,
874 				   struct device_attribute *attr, char *buf)
875 {
876 	struct gendisk *disk = dev_to_disk(dev);
877 
878 	return sprintf(buf, "%d\n",
879 		       (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
880 }
881 
882 static ssize_t disk_hidden_show(struct device *dev,
883 				   struct device_attribute *attr, char *buf)
884 {
885 	struct gendisk *disk = dev_to_disk(dev);
886 
887 	return sprintf(buf, "%d\n",
888 		       (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
889 }
890 
891 static ssize_t disk_ro_show(struct device *dev,
892 				   struct device_attribute *attr, char *buf)
893 {
894 	struct gendisk *disk = dev_to_disk(dev);
895 
896 	return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
897 }
898 
899 ssize_t part_size_show(struct device *dev,
900 		       struct device_attribute *attr, char *buf)
901 {
902 	return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev)));
903 }
904 
905 ssize_t part_stat_show(struct device *dev,
906 		       struct device_attribute *attr, char *buf)
907 {
908 	struct block_device *bdev = dev_to_bdev(dev);
909 	struct request_queue *q = bdev_get_queue(bdev);
910 	struct disk_stats stat;
911 	unsigned int inflight;
912 
913 	part_stat_read_all(bdev, &stat);
914 	if (queue_is_mq(q))
915 		inflight = blk_mq_in_flight(q, bdev);
916 	else
917 		inflight = part_in_flight(bdev);
918 
919 	return sprintf(buf,
920 		"%8lu %8lu %8llu %8u "
921 		"%8lu %8lu %8llu %8u "
922 		"%8u %8u %8u "
923 		"%8lu %8lu %8llu %8u "
924 		"%8lu %8u"
925 		"\n",
926 		stat.ios[STAT_READ],
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),
934 		inflight,
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],
940 						NSEC_PER_MSEC),
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));
947 }
948 
949 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
950 			   char *buf)
951 {
952 	struct block_device *bdev = dev_to_bdev(dev);
953 	struct request_queue *q = bdev_get_queue(bdev);
954 	unsigned int inflight[2];
955 
956 	if (queue_is_mq(q))
957 		blk_mq_in_flight_rw(q, bdev, inflight);
958 	else
959 		part_in_flight_rw(bdev, inflight);
960 
961 	return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
962 }
963 
964 static ssize_t disk_capability_show(struct device *dev,
965 				    struct device_attribute *attr, char *buf)
966 {
967 	struct gendisk *disk = dev_to_disk(dev);
968 
969 	return sprintf(buf, "%x\n", disk->flags);
970 }
971 
972 static ssize_t disk_alignment_offset_show(struct device *dev,
973 					  struct device_attribute *attr,
974 					  char *buf)
975 {
976 	struct gendisk *disk = dev_to_disk(dev);
977 
978 	return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
979 }
980 
981 static ssize_t disk_discard_alignment_show(struct device *dev,
982 					   struct device_attribute *attr,
983 					   char *buf)
984 {
985 	struct gendisk *disk = dev_to_disk(dev);
986 
987 	return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
988 }
989 
990 static ssize_t diskseq_show(struct device *dev,
991 			    struct device_attribute *attr, char *buf)
992 {
993 	struct gendisk *disk = dev_to_disk(dev);
994 
995 	return sprintf(buf, "%llu\n", disk->diskseq);
996 }
997 
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);
1011 
1012 #ifdef CONFIG_FAIL_MAKE_REQUEST
1013 ssize_t part_fail_show(struct device *dev,
1014 		       struct device_attribute *attr, char *buf)
1015 {
1016 	return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail);
1017 }
1018 
1019 ssize_t part_fail_store(struct device *dev,
1020 			struct device_attribute *attr,
1021 			const char *buf, size_t count)
1022 {
1023 	int i;
1024 
1025 	if (count > 0 && sscanf(buf, "%d", &i) > 0)
1026 		dev_to_bdev(dev)->bd_make_it_fail = i;
1027 
1028 	return count;
1029 }
1030 
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 */
1034 
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);
1038 #endif
1039 
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,
1045 	&dev_attr_ro.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,
1059 #endif
1060 #ifdef CONFIG_FAIL_IO_TIMEOUT
1061 	&dev_attr_fail_timeout.attr,
1062 #endif
1063 	NULL
1064 };
1065 
1066 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1067 {
1068 	struct device *dev = container_of(kobj, typeof(*dev), kobj);
1069 	struct gendisk *disk = dev_to_disk(dev);
1070 
1071 	if (a == &dev_attr_badblocks.attr && !disk->bb)
1072 		return 0;
1073 	return a->mode;
1074 }
1075 
1076 static struct attribute_group disk_attr_group = {
1077 	.attrs = disk_attrs,
1078 	.is_visible = disk_visible,
1079 };
1080 
1081 static const struct attribute_group *disk_attr_groups[] = {
1082 	&disk_attr_group,
1083 	NULL
1084 };
1085 
1086 /**
1087  * disk_release - releases all allocated resources of the gendisk
1088  * @dev: the device representing this disk
1089  *
1090  * This function releases all allocated resources of the gendisk.
1091  *
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.
1097  *
1098  * Context: can sleep
1099  */
1100 static void disk_release(struct device *dev)
1101 {
1102 	struct gendisk *disk = dev_to_disk(dev);
1103 
1104 	might_sleep();
1105 	WARN_ON_ONCE(disk_live(disk));
1106 
1107 	disk_release_events(disk);
1108 	kfree(disk->random);
1109 	xa_destroy(&disk->part_tbl);
1110 	disk->queue->disk = NULL;
1111 	blk_put_queue(disk->queue);
1112 	iput(disk->part0->bd_inode);	/* frees the disk */
1113 }
1114 
1115 static int block_uevent(struct device *dev, struct kobj_uevent_env *env)
1116 {
1117 	struct gendisk *disk = dev_to_disk(dev);
1118 
1119 	return add_uevent_var(env, "DISKSEQ=%llu", disk->diskseq);
1120 }
1121 
1122 struct class block_class = {
1123 	.name		= "block",
1124 	.dev_uevent	= block_uevent,
1125 };
1126 
1127 static char *block_devnode(struct device *dev, umode_t *mode,
1128 			   kuid_t *uid, kgid_t *gid)
1129 {
1130 	struct gendisk *disk = dev_to_disk(dev);
1131 
1132 	if (disk->fops->devnode)
1133 		return disk->fops->devnode(disk, mode);
1134 	return NULL;
1135 }
1136 
1137 const struct device_type disk_type = {
1138 	.name		= "disk",
1139 	.groups		= disk_attr_groups,
1140 	.release	= disk_release,
1141 	.devnode	= block_devnode,
1142 };
1143 
1144 #ifdef CONFIG_PROC_FS
1145 /*
1146  * aggregate disk stat collector.  Uses the same stats that the sysfs
1147  * entries do, above, but makes them available through one seq_file.
1148  *
1149  * The output looks suspiciously like /proc/partitions with a bunch of
1150  * extra fields.
1151  */
1152 static int diskstats_show(struct seq_file *seqf, void *v)
1153 {
1154 	struct gendisk *gp = v;
1155 	struct block_device *hd;
1156 	unsigned int inflight;
1157 	struct disk_stats stat;
1158 	unsigned long idx;
1159 
1160 	/*
1161 	if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1162 		seq_puts(seqf,	"major minor name"
1163 				"     rio rmerge rsect ruse wio wmerge "
1164 				"wsect wuse running use aveq"
1165 				"\n\n");
1166 	*/
1167 
1168 	rcu_read_lock();
1169 	xa_for_each(&gp->part_tbl, idx, hd) {
1170 		if (bdev_is_partition(hd) && !bdev_nr_sectors(hd))
1171 			continue;
1172 		part_stat_read_all(hd, &stat);
1173 		if (queue_is_mq(gp->queue))
1174 			inflight = blk_mq_in_flight(gp->queue, hd);
1175 		else
1176 			inflight = part_in_flight(hd);
1177 
1178 		seq_printf(seqf, "%4d %7d %pg "
1179 			   "%lu %lu %lu %u "
1180 			   "%lu %lu %lu %u "
1181 			   "%u %u %u "
1182 			   "%lu %lu %lu %u "
1183 			   "%lu %u"
1184 			   "\n",
1185 			   MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd,
1186 			   stat.ios[STAT_READ],
1187 			   stat.merges[STAT_READ],
1188 			   stat.sectors[STAT_READ],
1189 			   (unsigned int)div_u64(stat.nsecs[STAT_READ],
1190 							NSEC_PER_MSEC),
1191 			   stat.ios[STAT_WRITE],
1192 			   stat.merges[STAT_WRITE],
1193 			   stat.sectors[STAT_WRITE],
1194 			   (unsigned int)div_u64(stat.nsecs[STAT_WRITE],
1195 							NSEC_PER_MSEC),
1196 			   inflight,
1197 			   jiffies_to_msecs(stat.io_ticks),
1198 			   (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1199 						 stat.nsecs[STAT_WRITE] +
1200 						 stat.nsecs[STAT_DISCARD] +
1201 						 stat.nsecs[STAT_FLUSH],
1202 							NSEC_PER_MSEC),
1203 			   stat.ios[STAT_DISCARD],
1204 			   stat.merges[STAT_DISCARD],
1205 			   stat.sectors[STAT_DISCARD],
1206 			   (unsigned int)div_u64(stat.nsecs[STAT_DISCARD],
1207 						 NSEC_PER_MSEC),
1208 			   stat.ios[STAT_FLUSH],
1209 			   (unsigned int)div_u64(stat.nsecs[STAT_FLUSH],
1210 						 NSEC_PER_MSEC)
1211 			);
1212 	}
1213 	rcu_read_unlock();
1214 
1215 	return 0;
1216 }
1217 
1218 static const struct seq_operations diskstats_op = {
1219 	.start	= disk_seqf_start,
1220 	.next	= disk_seqf_next,
1221 	.stop	= disk_seqf_stop,
1222 	.show	= diskstats_show
1223 };
1224 
1225 static int __init proc_genhd_init(void)
1226 {
1227 	proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1228 	proc_create_seq("partitions", 0, NULL, &partitions_op);
1229 	return 0;
1230 }
1231 module_init(proc_genhd_init);
1232 #endif /* CONFIG_PROC_FS */
1233 
1234 dev_t part_devt(struct gendisk *disk, u8 partno)
1235 {
1236 	struct block_device *part;
1237 	dev_t devt = 0;
1238 
1239 	rcu_read_lock();
1240 	part = xa_load(&disk->part_tbl, partno);
1241 	if (part)
1242 		devt = part->bd_dev;
1243 	rcu_read_unlock();
1244 
1245 	return devt;
1246 }
1247 
1248 dev_t blk_lookup_devt(const char *name, int partno)
1249 {
1250 	dev_t devt = MKDEV(0, 0);
1251 	struct class_dev_iter iter;
1252 	struct device *dev;
1253 
1254 	class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1255 	while ((dev = class_dev_iter_next(&iter))) {
1256 		struct gendisk *disk = dev_to_disk(dev);
1257 
1258 		if (strcmp(dev_name(dev), name))
1259 			continue;
1260 
1261 		if (partno < disk->minors) {
1262 			/* We need to return the right devno, even
1263 			 * if the partition doesn't exist yet.
1264 			 */
1265 			devt = MKDEV(MAJOR(dev->devt),
1266 				     MINOR(dev->devt) + partno);
1267 		} else {
1268 			devt = part_devt(disk, partno);
1269 			if (devt)
1270 				break;
1271 		}
1272 	}
1273 	class_dev_iter_exit(&iter);
1274 	return devt;
1275 }
1276 
1277 struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id,
1278 		struct lock_class_key *lkclass)
1279 {
1280 	struct gendisk *disk;
1281 
1282 	if (!blk_get_queue(q))
1283 		return NULL;
1284 
1285 	disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1286 	if (!disk)
1287 		goto out_put_queue;
1288 
1289 	disk->bdi = bdi_alloc(node_id);
1290 	if (!disk->bdi)
1291 		goto out_free_disk;
1292 
1293 	/* bdev_alloc() might need the queue, set before the first call */
1294 	disk->queue = q;
1295 
1296 	disk->part0 = bdev_alloc(disk, 0);
1297 	if (!disk->part0)
1298 		goto out_free_bdi;
1299 
1300 	disk->node_id = node_id;
1301 	mutex_init(&disk->open_mutex);
1302 	xa_init(&disk->part_tbl);
1303 	if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL))
1304 		goto out_destroy_part_tbl;
1305 
1306 	rand_initialize_disk(disk);
1307 	disk_to_dev(disk)->class = &block_class;
1308 	disk_to_dev(disk)->type = &disk_type;
1309 	device_initialize(disk_to_dev(disk));
1310 	inc_diskseq(disk);
1311 	q->disk = disk;
1312 	lockdep_init_map(&disk->lockdep_map, "(bio completion)", lkclass, 0);
1313 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
1314 	INIT_LIST_HEAD(&disk->slave_bdevs);
1315 #endif
1316 	return disk;
1317 
1318 out_destroy_part_tbl:
1319 	xa_destroy(&disk->part_tbl);
1320 	disk->part0->bd_disk = NULL;
1321 	iput(disk->part0->bd_inode);
1322 out_free_bdi:
1323 	bdi_put(disk->bdi);
1324 out_free_disk:
1325 	kfree(disk);
1326 out_put_queue:
1327 	blk_put_queue(q);
1328 	return NULL;
1329 }
1330 EXPORT_SYMBOL(__alloc_disk_node);
1331 
1332 struct gendisk *__blk_alloc_disk(int node, struct lock_class_key *lkclass)
1333 {
1334 	struct request_queue *q;
1335 	struct gendisk *disk;
1336 
1337 	q = blk_alloc_queue(node);
1338 	if (!q)
1339 		return NULL;
1340 
1341 	disk = __alloc_disk_node(q, node, lkclass);
1342 	if (!disk) {
1343 		blk_cleanup_queue(q);
1344 		return NULL;
1345 	}
1346 	return disk;
1347 }
1348 EXPORT_SYMBOL(__blk_alloc_disk);
1349 
1350 /**
1351  * put_disk - decrements the gendisk refcount
1352  * @disk: the struct gendisk to decrement the refcount for
1353  *
1354  * This decrements the refcount for the struct gendisk. When this reaches 0
1355  * we'll have disk_release() called.
1356  *
1357  * Context: Any context, but the last reference must not be dropped from
1358  *          atomic context.
1359  */
1360 void put_disk(struct gendisk *disk)
1361 {
1362 	if (disk)
1363 		put_device(disk_to_dev(disk));
1364 }
1365 EXPORT_SYMBOL(put_disk);
1366 
1367 /**
1368  * blk_cleanup_disk - shutdown a gendisk allocated by blk_alloc_disk
1369  * @disk: gendisk to shutdown
1370  *
1371  * Mark the queue hanging off @disk DYING, drain all pending requests, then mark
1372  * the queue DEAD, destroy and put it and the gendisk structure.
1373  *
1374  * Context: can sleep
1375  */
1376 void blk_cleanup_disk(struct gendisk *disk)
1377 {
1378 	blk_cleanup_queue(disk->queue);
1379 	put_disk(disk);
1380 }
1381 EXPORT_SYMBOL(blk_cleanup_disk);
1382 
1383 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1384 {
1385 	char event[] = "DISK_RO=1";
1386 	char *envp[] = { event, NULL };
1387 
1388 	if (!ro)
1389 		event[8] = '0';
1390 	kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1391 }
1392 
1393 /**
1394  * set_disk_ro - set a gendisk read-only
1395  * @disk:	gendisk to operate on
1396  * @read_only:	%true to set the disk read-only, %false set the disk read/write
1397  *
1398  * This function is used to indicate whether a given disk device should have its
1399  * read-only flag set. set_disk_ro() is typically used by device drivers to
1400  * indicate whether the underlying physical device is write-protected.
1401  */
1402 void set_disk_ro(struct gendisk *disk, bool read_only)
1403 {
1404 	if (read_only) {
1405 		if (test_and_set_bit(GD_READ_ONLY, &disk->state))
1406 			return;
1407 	} else {
1408 		if (!test_and_clear_bit(GD_READ_ONLY, &disk->state))
1409 			return;
1410 	}
1411 	set_disk_ro_uevent(disk, read_only);
1412 }
1413 EXPORT_SYMBOL(set_disk_ro);
1414 
1415 void inc_diskseq(struct gendisk *disk)
1416 {
1417 	disk->diskseq = atomic64_inc_return(&diskseq);
1418 }
1419