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 
part_stat_read_all(struct block_device * part,struct disk_stats * stat)104 static void part_stat_read_all(struct block_device *part,
105 		struct disk_stats *stat)
106 {
107 	int cpu;
108 
109 	memset(stat, 0, sizeof(struct disk_stats));
110 	for_each_possible_cpu(cpu) {
111 		struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu);
112 		int group;
113 
114 		for (group = 0; group < NR_STAT_GROUPS; group++) {
115 			stat->nsecs[group] += ptr->nsecs[group];
116 			stat->sectors[group] += ptr->sectors[group];
117 			stat->ios[group] += ptr->ios[group];
118 			stat->merges[group] += ptr->merges[group];
119 		}
120 
121 		stat->io_ticks += ptr->io_ticks;
122 	}
123 }
124 
part_in_flight(struct block_device * part)125 static unsigned int part_in_flight(struct block_device *part)
126 {
127 	unsigned int inflight = 0;
128 	int cpu;
129 
130 	for_each_possible_cpu(cpu) {
131 		inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) +
132 			    part_stat_local_read_cpu(part, in_flight[1], cpu);
133 	}
134 	if ((int)inflight < 0)
135 		inflight = 0;
136 
137 	return inflight;
138 }
139 
part_in_flight_rw(struct block_device * part,unsigned int inflight[2])140 static void part_in_flight_rw(struct block_device *part,
141 		unsigned int inflight[2])
142 {
143 	int cpu;
144 
145 	inflight[0] = 0;
146 	inflight[1] = 0;
147 	for_each_possible_cpu(cpu) {
148 		inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu);
149 		inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu);
150 	}
151 	if ((int)inflight[0] < 0)
152 		inflight[0] = 0;
153 	if ((int)inflight[1] < 0)
154 		inflight[1] = 0;
155 }
156 
157 /*
158  * Can be deleted altogether. Later.
159  *
160  */
161 #define BLKDEV_MAJOR_HASH_SIZE 255
162 static struct blk_major_name {
163 	struct blk_major_name *next;
164 	int major;
165 	char name[16];
166 #ifdef CONFIG_BLOCK_LEGACY_AUTOLOAD
167 	void (*probe)(dev_t devt);
168 #endif
169 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
170 static DEFINE_MUTEX(major_names_lock);
171 static DEFINE_SPINLOCK(major_names_spinlock);
172 
173 /* index in the above - for now: assume no multimajor ranges */
major_to_index(unsigned major)174 static inline int major_to_index(unsigned major)
175 {
176 	return major % BLKDEV_MAJOR_HASH_SIZE;
177 }
178 
179 #ifdef CONFIG_PROC_FS
blkdev_show(struct seq_file * seqf,off_t offset)180 void blkdev_show(struct seq_file *seqf, off_t offset)
181 {
182 	struct blk_major_name *dp;
183 
184 	spin_lock(&major_names_spinlock);
185 	for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
186 		if (dp->major == offset)
187 			seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
188 	spin_unlock(&major_names_spinlock);
189 }
190 #endif /* CONFIG_PROC_FS */
191 
192 /**
193  * __register_blkdev - register a new block device
194  *
195  * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
196  *         @major = 0, try to allocate any unused major number.
197  * @name: the name of the new block device as a zero terminated string
198  * @probe: pre-devtmpfs / pre-udev callback used to create disks when their
199  *	   pre-created device node is accessed. When a probe call uses
200  *	   add_disk() and it fails the driver must cleanup resources. This
201  *	   interface may soon be removed.
202  *
203  * The @name must be unique within the system.
204  *
205  * The return value depends on the @major input parameter:
206  *
207  *  - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
208  *    then the function returns zero on success, or a negative error code
209  *  - if any unused major number was requested with @major = 0 parameter
210  *    then the return value is the allocated major number in range
211  *    [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
212  *
213  * See Documentation/admin-guide/devices.txt for the list of allocated
214  * major numbers.
215  *
216  * Use register_blkdev instead for any new code.
217  */
__register_blkdev(unsigned int major,const char * name,void (* probe)(dev_t devt))218 int __register_blkdev(unsigned int major, const char *name,
219 		void (*probe)(dev_t devt))
220 {
221 	struct blk_major_name **n, *p;
222 	int index, ret = 0;
223 
224 	mutex_lock(&major_names_lock);
225 
226 	/* temporary */
227 	if (major == 0) {
228 		for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
229 			if (major_names[index] == NULL)
230 				break;
231 		}
232 
233 		if (index == 0) {
234 			printk("%s: failed to get major for %s\n",
235 			       __func__, name);
236 			ret = -EBUSY;
237 			goto out;
238 		}
239 		major = index;
240 		ret = major;
241 	}
242 
243 	if (major >= BLKDEV_MAJOR_MAX) {
244 		pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
245 		       __func__, major, BLKDEV_MAJOR_MAX-1, name);
246 
247 		ret = -EINVAL;
248 		goto out;
249 	}
250 
251 	p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
252 	if (p == NULL) {
253 		ret = -ENOMEM;
254 		goto out;
255 	}
256 
257 	p->major = major;
258 #ifdef CONFIG_BLOCK_LEGACY_AUTOLOAD
259 	p->probe = probe;
260 #endif
261 	strlcpy(p->name, name, sizeof(p->name));
262 	p->next = NULL;
263 	index = major_to_index(major);
264 
265 	spin_lock(&major_names_spinlock);
266 	for (n = &major_names[index]; *n; n = &(*n)->next) {
267 		if ((*n)->major == major)
268 			break;
269 	}
270 	if (!*n)
271 		*n = p;
272 	else
273 		ret = -EBUSY;
274 	spin_unlock(&major_names_spinlock);
275 
276 	if (ret < 0) {
277 		printk("register_blkdev: cannot get major %u for %s\n",
278 		       major, name);
279 		kfree(p);
280 	}
281 out:
282 	mutex_unlock(&major_names_lock);
283 	return ret;
284 }
285 EXPORT_SYMBOL(__register_blkdev);
286 
unregister_blkdev(unsigned int major,const char * name)287 void unregister_blkdev(unsigned int major, const char *name)
288 {
289 	struct blk_major_name **n;
290 	struct blk_major_name *p = NULL;
291 	int index = major_to_index(major);
292 
293 	mutex_lock(&major_names_lock);
294 	spin_lock(&major_names_spinlock);
295 	for (n = &major_names[index]; *n; n = &(*n)->next)
296 		if ((*n)->major == major)
297 			break;
298 	if (!*n || strcmp((*n)->name, name)) {
299 		WARN_ON(1);
300 	} else {
301 		p = *n;
302 		*n = p->next;
303 	}
304 	spin_unlock(&major_names_spinlock);
305 	mutex_unlock(&major_names_lock);
306 	kfree(p);
307 }
308 
309 EXPORT_SYMBOL(unregister_blkdev);
310 
blk_alloc_ext_minor(void)311 int blk_alloc_ext_minor(void)
312 {
313 	int idx;
314 
315 	idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT - 1, GFP_KERNEL);
316 	if (idx == -ENOSPC)
317 		return -EBUSY;
318 	return idx;
319 }
320 
blk_free_ext_minor(unsigned int minor)321 void blk_free_ext_minor(unsigned int minor)
322 {
323 	ida_free(&ext_devt_ida, minor);
324 }
325 
bdevt_str(dev_t devt,char * buf)326 static char *bdevt_str(dev_t devt, char *buf)
327 {
328 	if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
329 		char tbuf[BDEVT_SIZE];
330 		snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
331 		snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
332 	} else
333 		snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
334 
335 	return buf;
336 }
337 
disk_uevent(struct gendisk * disk,enum kobject_action action)338 void disk_uevent(struct gendisk *disk, enum kobject_action action)
339 {
340 	struct block_device *part;
341 	unsigned long idx;
342 
343 	rcu_read_lock();
344 	xa_for_each(&disk->part_tbl, idx, part) {
345 		if (bdev_is_partition(part) && !bdev_nr_sectors(part))
346 			continue;
347 		if (!kobject_get_unless_zero(&part->bd_device.kobj))
348 			continue;
349 
350 		rcu_read_unlock();
351 		kobject_uevent(bdev_kobj(part), action);
352 		put_device(&part->bd_device);
353 		rcu_read_lock();
354 	}
355 	rcu_read_unlock();
356 }
357 EXPORT_SYMBOL_GPL(disk_uevent);
358 
disk_scan_partitions(struct gendisk * disk,fmode_t mode,void * owner)359 int disk_scan_partitions(struct gendisk *disk, fmode_t mode, void *owner)
360 {
361 	struct block_device *bdev;
362 
363 	if (disk->flags & (GENHD_FL_NO_PART | GENHD_FL_HIDDEN))
364 		return -EINVAL;
365 	if (test_bit(GD_SUPPRESS_PART_SCAN, &disk->state))
366 		return -EINVAL;
367 	if (disk->open_partitions)
368 		return -EBUSY;
369 	/* Someone else has bdev exclusively open? */
370 	if (disk->part0->bd_holder && disk->part0->bd_holder != owner)
371 		return -EBUSY;
372 
373 	set_bit(GD_NEED_PART_SCAN, &disk->state);
374 	bdev = blkdev_get_by_dev(disk_devt(disk), mode, NULL);
375 	if (IS_ERR(bdev))
376 		return PTR_ERR(bdev);
377 	blkdev_put(bdev, mode);
378 	return 0;
379 }
380 
381 /**
382  * device_add_disk - add disk information to kernel list
383  * @parent: parent device for the disk
384  * @disk: per-device partitioning information
385  * @groups: Additional per-device sysfs groups
386  *
387  * This function registers the partitioning information in @disk
388  * with the kernel.
389  */
device_add_disk(struct device * parent,struct gendisk * disk,const struct attribute_group ** groups)390 int __must_check device_add_disk(struct device *parent, struct gendisk *disk,
391 				 const struct attribute_group **groups)
392 
393 {
394 	struct device *ddev = disk_to_dev(disk);
395 	int ret;
396 
397 	/* Only makes sense for bio-based to set ->poll_bio */
398 	if (queue_is_mq(disk->queue) && disk->fops->poll_bio)
399 		return -EINVAL;
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 	ret = -EINVAL;
417 	if (disk->major) {
418 		if (WARN_ON(!disk->minors))
419 			goto out_exit_elevator;
420 
421 		if (disk->minors > DISK_MAX_PARTS) {
422 			pr_err("block: can't allocate more than %d partitions\n",
423 				DISK_MAX_PARTS);
424 			disk->minors = DISK_MAX_PARTS;
425 		}
426 		if (disk->first_minor + disk->minors > MINORMASK + 1)
427 			goto out_exit_elevator;
428 	} else {
429 		if (WARN_ON(disk->minors))
430 			goto out_exit_elevator;
431 
432 		ret = blk_alloc_ext_minor();
433 		if (ret < 0)
434 			goto out_exit_elevator;
435 		disk->major = BLOCK_EXT_MAJOR;
436 		disk->first_minor = ret;
437 	}
438 
439 	/* delay uevents, until we scanned partition table */
440 	dev_set_uevent_suppress(ddev, 1);
441 
442 	ddev->parent = parent;
443 	ddev->groups = groups;
444 	dev_set_name(ddev, "%s", disk->disk_name);
445 	if (!(disk->flags & GENHD_FL_HIDDEN))
446 		ddev->devt = MKDEV(disk->major, disk->first_minor);
447 	ret = device_add(ddev);
448 	if (ret)
449 		goto out_free_ext_minor;
450 
451 	ret = disk_alloc_events(disk);
452 	if (ret)
453 		goto out_device_del;
454 
455 	if (!sysfs_deprecated) {
456 		ret = sysfs_create_link(block_depr, &ddev->kobj,
457 					kobject_name(&ddev->kobj));
458 		if (ret)
459 			goto out_device_del;
460 	}
461 
462 	/*
463 	 * avoid probable deadlock caused by allocating memory with
464 	 * GFP_KERNEL in runtime_resume callback of its all ancestor
465 	 * devices
466 	 */
467 	pm_runtime_set_memalloc_noio(ddev, true);
468 
469 	ret = blk_integrity_add(disk);
470 	if (ret)
471 		goto out_del_block_link;
472 
473 	disk->part0->bd_holder_dir =
474 		kobject_create_and_add("holders", &ddev->kobj);
475 	if (!disk->part0->bd_holder_dir) {
476 		ret = -ENOMEM;
477 		goto out_del_integrity;
478 	}
479 	disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
480 	if (!disk->slave_dir) {
481 		ret = -ENOMEM;
482 		goto out_put_holder_dir;
483 	}
484 
485 	ret = bd_register_pending_holders(disk);
486 	if (ret < 0)
487 		goto out_put_slave_dir;
488 
489 	ret = blk_register_queue(disk);
490 	if (ret)
491 		goto out_put_slave_dir;
492 
493 	if (!(disk->flags & GENHD_FL_HIDDEN)) {
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 		if (get_capacity(disk))
506 			disk_scan_partitions(disk, FMODE_READ, NULL);
507 
508 		/*
509 		 * Announce the disk and partitions after all partitions are
510 		 * created. (for hidden disks uevents remain suppressed forever)
511 		 */
512 		dev_set_uevent_suppress(ddev, 0);
513 		disk_uevent(disk, KOBJ_ADD);
514 	} else {
515 		/*
516 		 * Even if the block_device for a hidden gendisk is not
517 		 * registered, it needs to have a valid bd_dev so that the
518 		 * freeing of the dynamic major works.
519 		 */
520 		disk->part0->bd_dev = MKDEV(disk->major, disk->first_minor);
521 	}
522 
523 	disk_update_readahead(disk);
524 	disk_add_events(disk);
525 	set_bit(GD_ADDED, &disk->state);
526 	return 0;
527 
528 out_unregister_bdi:
529 	if (!(disk->flags & GENHD_FL_HIDDEN))
530 		bdi_unregister(disk->bdi);
531 out_unregister_queue:
532 	blk_unregister_queue(disk);
533 	rq_qos_exit(disk->queue);
534 out_put_slave_dir:
535 	kobject_put(disk->slave_dir);
536 	disk->slave_dir = NULL;
537 out_put_holder_dir:
538 	kobject_put(disk->part0->bd_holder_dir);
539 out_del_integrity:
540 	blk_integrity_del(disk);
541 out_del_block_link:
542 	if (!sysfs_deprecated)
543 		sysfs_remove_link(block_depr, dev_name(ddev));
544 out_device_del:
545 	device_del(ddev);
546 out_free_ext_minor:
547 	if (disk->major == BLOCK_EXT_MAJOR)
548 		blk_free_ext_minor(disk->first_minor);
549 out_exit_elevator:
550 	if (disk->queue->elevator)
551 		elevator_exit(disk->queue);
552 	return ret;
553 }
554 EXPORT_SYMBOL(device_add_disk);
555 
556 /**
557  * blk_mark_disk_dead - mark a disk as dead
558  * @disk: disk to mark as dead
559  *
560  * Mark as disk as dead (e.g. surprise removed) and don't accept any new I/O
561  * to this disk.
562  */
blk_mark_disk_dead(struct gendisk * disk)563 void blk_mark_disk_dead(struct gendisk *disk)
564 {
565 	set_bit(GD_DEAD, &disk->state);
566 	blk_queue_start_drain(disk->queue);
567 }
568 EXPORT_SYMBOL_GPL(blk_mark_disk_dead);
569 
570 /**
571  * del_gendisk - remove the gendisk
572  * @disk: the struct gendisk to remove
573  *
574  * Removes the gendisk and all its associated resources. This deletes the
575  * partitions associated with the gendisk, and unregisters the associated
576  * request_queue.
577  *
578  * This is the counter to the respective __device_add_disk() call.
579  *
580  * The final removal of the struct gendisk happens when its refcount reaches 0
581  * with put_disk(), which should be called after del_gendisk(), if
582  * __device_add_disk() was used.
583  *
584  * Drivers exist which depend on the release of the gendisk to be synchronous,
585  * it should not be deferred.
586  *
587  * Context: can sleep
588  */
del_gendisk(struct gendisk * disk)589 void del_gendisk(struct gendisk *disk)
590 {
591 	struct request_queue *q = disk->queue;
592 
593 	might_sleep();
594 
595 	if (WARN_ON_ONCE(!disk_live(disk) && !(disk->flags & GENHD_FL_HIDDEN)))
596 		return;
597 
598 	blk_integrity_del(disk);
599 	disk_del_events(disk);
600 
601 	mutex_lock(&disk->open_mutex);
602 	remove_inode_hash(disk->part0->bd_inode);
603 	blk_drop_partitions(disk);
604 	mutex_unlock(&disk->open_mutex);
605 
606 	fsync_bdev(disk->part0);
607 	__invalidate_device(disk->part0, true);
608 
609 	/*
610 	 * Fail any new I/O.
611 	 */
612 	set_bit(GD_DEAD, &disk->state);
613 	if (test_bit(GD_OWNS_QUEUE, &disk->state))
614 		blk_queue_flag_set(QUEUE_FLAG_DYING, q);
615 	set_capacity(disk, 0);
616 
617 	/*
618 	 * Prevent new I/O from crossing bio_queue_enter().
619 	 */
620 	blk_queue_start_drain(q);
621 
622 	if (!(disk->flags & GENHD_FL_HIDDEN)) {
623 		sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
624 
625 		/*
626 		 * Unregister bdi before releasing device numbers (as they can
627 		 * get reused and we'd get clashes in sysfs).
628 		 */
629 		bdi_unregister(disk->bdi);
630 	}
631 
632 	blk_unregister_queue(disk);
633 
634 	kobject_put(disk->part0->bd_holder_dir);
635 	kobject_put(disk->slave_dir);
636 	disk->slave_dir = NULL;
637 
638 	part_stat_set_all(disk->part0, 0);
639 	disk->part0->bd_stamp = 0;
640 	if (!sysfs_deprecated)
641 		sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
642 	pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
643 	device_del(disk_to_dev(disk));
644 
645 	blk_mq_freeze_queue_wait(q);
646 
647 	blk_throtl_cancel_bios(disk);
648 
649 	blk_sync_queue(q);
650 	blk_flush_integrity();
651 	blk_mq_cancel_work_sync(q);
652 
653 	blk_mq_quiesce_queue(q);
654 	if (q->elevator) {
655 		mutex_lock(&q->sysfs_lock);
656 		elevator_exit(q);
657 		mutex_unlock(&q->sysfs_lock);
658 	}
659 	rq_qos_exit(q);
660 	blk_mq_unquiesce_queue(q);
661 
662 	/*
663 	 * If the disk does not own the queue, allow using passthrough requests
664 	 * again.  Else leave the queue frozen to fail all I/O.
665 	 */
666 	if (!test_bit(GD_OWNS_QUEUE, &disk->state)) {
667 		blk_queue_flag_clear(QUEUE_FLAG_INIT_DONE, q);
668 		__blk_mq_unfreeze_queue(q, true);
669 	} else {
670 		if (queue_is_mq(q))
671 			blk_mq_exit_queue(q);
672 	}
673 }
674 EXPORT_SYMBOL(del_gendisk);
675 
676 /**
677  * invalidate_disk - invalidate the disk
678  * @disk: the struct gendisk to invalidate
679  *
680  * A helper to invalidates the disk. It will clean the disk's associated
681  * buffer/page caches and reset its internal states so that the disk
682  * can be reused by the drivers.
683  *
684  * Context: can sleep
685  */
invalidate_disk(struct gendisk * disk)686 void invalidate_disk(struct gendisk *disk)
687 {
688 	struct block_device *bdev = disk->part0;
689 
690 	invalidate_bdev(bdev);
691 	bdev->bd_inode->i_mapping->wb_err = 0;
692 	set_capacity(disk, 0);
693 }
694 EXPORT_SYMBOL(invalidate_disk);
695 
696 /* sysfs access to bad-blocks list. */
disk_badblocks_show(struct device * dev,struct device_attribute * attr,char * page)697 static ssize_t disk_badblocks_show(struct device *dev,
698 					struct device_attribute *attr,
699 					char *page)
700 {
701 	struct gendisk *disk = dev_to_disk(dev);
702 
703 	if (!disk->bb)
704 		return sprintf(page, "\n");
705 
706 	return badblocks_show(disk->bb, page, 0);
707 }
708 
disk_badblocks_store(struct device * dev,struct device_attribute * attr,const char * page,size_t len)709 static ssize_t disk_badblocks_store(struct device *dev,
710 					struct device_attribute *attr,
711 					const char *page, size_t len)
712 {
713 	struct gendisk *disk = dev_to_disk(dev);
714 
715 	if (!disk->bb)
716 		return -ENXIO;
717 
718 	return badblocks_store(disk->bb, page, len, 0);
719 }
720 
721 #ifdef CONFIG_BLOCK_LEGACY_AUTOLOAD
blk_request_module(dev_t devt)722 void blk_request_module(dev_t devt)
723 {
724 	unsigned int major = MAJOR(devt);
725 	struct blk_major_name **n;
726 
727 	mutex_lock(&major_names_lock);
728 	for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) {
729 		if ((*n)->major == major && (*n)->probe) {
730 			(*n)->probe(devt);
731 			mutex_unlock(&major_names_lock);
732 			return;
733 		}
734 	}
735 	mutex_unlock(&major_names_lock);
736 
737 	if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
738 		/* Make old-style 2.4 aliases work */
739 		request_module("block-major-%d", MAJOR(devt));
740 }
741 #endif /* CONFIG_BLOCK_LEGACY_AUTOLOAD */
742 
743 /*
744  * print a full list of all partitions - intended for places where the root
745  * filesystem can't be mounted and thus to give the victim some idea of what
746  * went wrong
747  */
printk_all_partitions(void)748 void __init printk_all_partitions(void)
749 {
750 	struct class_dev_iter iter;
751 	struct device *dev;
752 
753 	class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
754 	while ((dev = class_dev_iter_next(&iter))) {
755 		struct gendisk *disk = dev_to_disk(dev);
756 		struct block_device *part;
757 		char devt_buf[BDEVT_SIZE];
758 		unsigned long idx;
759 
760 		/*
761 		 * Don't show empty devices or things that have been
762 		 * suppressed
763 		 */
764 		if (get_capacity(disk) == 0 || (disk->flags & GENHD_FL_HIDDEN))
765 			continue;
766 
767 		/*
768 		 * Note, unlike /proc/partitions, I am showing the numbers in
769 		 * hex - the same format as the root= option takes.
770 		 */
771 		rcu_read_lock();
772 		xa_for_each(&disk->part_tbl, idx, part) {
773 			if (!bdev_nr_sectors(part))
774 				continue;
775 			printk("%s%s %10llu %pg %s",
776 			       bdev_is_partition(part) ? "  " : "",
777 			       bdevt_str(part->bd_dev, devt_buf),
778 			       bdev_nr_sectors(part) >> 1, part,
779 			       part->bd_meta_info ?
780 					part->bd_meta_info->uuid : "");
781 			if (bdev_is_partition(part))
782 				printk("\n");
783 			else if (dev->parent && dev->parent->driver)
784 				printk(" driver: %s\n",
785 					dev->parent->driver->name);
786 			else
787 				printk(" (driver?)\n");
788 		}
789 		rcu_read_unlock();
790 	}
791 	class_dev_iter_exit(&iter);
792 }
793 
794 #ifdef CONFIG_PROC_FS
795 /* iterator */
disk_seqf_start(struct seq_file * seqf,loff_t * pos)796 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
797 {
798 	loff_t skip = *pos;
799 	struct class_dev_iter *iter;
800 	struct device *dev;
801 
802 	iter = kmalloc(sizeof(*iter), GFP_KERNEL);
803 	if (!iter)
804 		return ERR_PTR(-ENOMEM);
805 
806 	seqf->private = iter;
807 	class_dev_iter_init(iter, &block_class, NULL, &disk_type);
808 	do {
809 		dev = class_dev_iter_next(iter);
810 		if (!dev)
811 			return NULL;
812 	} while (skip--);
813 
814 	return dev_to_disk(dev);
815 }
816 
disk_seqf_next(struct seq_file * seqf,void * v,loff_t * pos)817 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
818 {
819 	struct device *dev;
820 
821 	(*pos)++;
822 	dev = class_dev_iter_next(seqf->private);
823 	if (dev)
824 		return dev_to_disk(dev);
825 
826 	return NULL;
827 }
828 
disk_seqf_stop(struct seq_file * seqf,void * v)829 static void disk_seqf_stop(struct seq_file *seqf, void *v)
830 {
831 	struct class_dev_iter *iter = seqf->private;
832 
833 	/* stop is called even after start failed :-( */
834 	if (iter) {
835 		class_dev_iter_exit(iter);
836 		kfree(iter);
837 		seqf->private = NULL;
838 	}
839 }
840 
show_partition_start(struct seq_file * seqf,loff_t * pos)841 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
842 {
843 	void *p;
844 
845 	p = disk_seqf_start(seqf, pos);
846 	if (!IS_ERR_OR_NULL(p) && !*pos)
847 		seq_puts(seqf, "major minor  #blocks  name\n\n");
848 	return p;
849 }
850 
show_partition(struct seq_file * seqf,void * v)851 static int show_partition(struct seq_file *seqf, void *v)
852 {
853 	struct gendisk *sgp = v;
854 	struct block_device *part;
855 	unsigned long idx;
856 
857 	if (!get_capacity(sgp) || (sgp->flags & GENHD_FL_HIDDEN))
858 		return 0;
859 
860 	rcu_read_lock();
861 	xa_for_each(&sgp->part_tbl, idx, part) {
862 		if (!bdev_nr_sectors(part))
863 			continue;
864 		seq_printf(seqf, "%4d  %7d %10llu %pg\n",
865 			   MAJOR(part->bd_dev), MINOR(part->bd_dev),
866 			   bdev_nr_sectors(part) >> 1, part);
867 	}
868 	rcu_read_unlock();
869 	return 0;
870 }
871 
872 static const struct seq_operations partitions_op = {
873 	.start	= show_partition_start,
874 	.next	= disk_seqf_next,
875 	.stop	= disk_seqf_stop,
876 	.show	= show_partition
877 };
878 #endif
879 
genhd_device_init(void)880 static int __init genhd_device_init(void)
881 {
882 	int error;
883 
884 	block_class.dev_kobj = sysfs_dev_block_kobj;
885 	error = class_register(&block_class);
886 	if (unlikely(error))
887 		return error;
888 	blk_dev_init();
889 
890 	register_blkdev(BLOCK_EXT_MAJOR, "blkext");
891 
892 	/* create top-level block dir */
893 	if (!sysfs_deprecated)
894 		block_depr = kobject_create_and_add("block", NULL);
895 	return 0;
896 }
897 
898 subsys_initcall(genhd_device_init);
899 
disk_range_show(struct device * dev,struct device_attribute * attr,char * buf)900 static ssize_t disk_range_show(struct device *dev,
901 			       struct device_attribute *attr, char *buf)
902 {
903 	struct gendisk *disk = dev_to_disk(dev);
904 
905 	return sprintf(buf, "%d\n", disk->minors);
906 }
907 
disk_ext_range_show(struct device * dev,struct device_attribute * attr,char * buf)908 static ssize_t disk_ext_range_show(struct device *dev,
909 				   struct device_attribute *attr, char *buf)
910 {
911 	struct gendisk *disk = dev_to_disk(dev);
912 
913 	return sprintf(buf, "%d\n",
914 		(disk->flags & GENHD_FL_NO_PART) ? 1 : DISK_MAX_PARTS);
915 }
916 
disk_removable_show(struct device * dev,struct device_attribute * attr,char * buf)917 static ssize_t disk_removable_show(struct device *dev,
918 				   struct device_attribute *attr, char *buf)
919 {
920 	struct gendisk *disk = dev_to_disk(dev);
921 
922 	return sprintf(buf, "%d\n",
923 		       (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
924 }
925 
disk_hidden_show(struct device * dev,struct device_attribute * attr,char * buf)926 static ssize_t disk_hidden_show(struct device *dev,
927 				   struct device_attribute *attr, char *buf)
928 {
929 	struct gendisk *disk = dev_to_disk(dev);
930 
931 	return sprintf(buf, "%d\n",
932 		       (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
933 }
934 
disk_ro_show(struct device * dev,struct device_attribute * attr,char * buf)935 static ssize_t disk_ro_show(struct device *dev,
936 				   struct device_attribute *attr, char *buf)
937 {
938 	struct gendisk *disk = dev_to_disk(dev);
939 
940 	return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
941 }
942 
part_size_show(struct device * dev,struct device_attribute * attr,char * buf)943 ssize_t part_size_show(struct device *dev,
944 		       struct device_attribute *attr, char *buf)
945 {
946 	return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev)));
947 }
948 
part_stat_show(struct device * dev,struct device_attribute * attr,char * buf)949 ssize_t part_stat_show(struct device *dev,
950 		       struct device_attribute *attr, char *buf)
951 {
952 	struct block_device *bdev = dev_to_bdev(dev);
953 	struct request_queue *q = bdev_get_queue(bdev);
954 	struct disk_stats stat;
955 	unsigned int inflight;
956 
957 	if (queue_is_mq(q))
958 		inflight = blk_mq_in_flight(q, bdev);
959 	else
960 		inflight = part_in_flight(bdev);
961 
962 	if (inflight) {
963 		part_stat_lock();
964 		update_io_ticks(bdev, jiffies, true);
965 		part_stat_unlock();
966 	}
967 	part_stat_read_all(bdev, &stat);
968 	return sprintf(buf,
969 		"%8lu %8lu %8llu %8u "
970 		"%8lu %8lu %8llu %8u "
971 		"%8u %8u %8u "
972 		"%8lu %8lu %8llu %8u "
973 		"%8lu %8u"
974 		"\n",
975 		stat.ios[STAT_READ],
976 		stat.merges[STAT_READ],
977 		(unsigned long long)stat.sectors[STAT_READ],
978 		(unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC),
979 		stat.ios[STAT_WRITE],
980 		stat.merges[STAT_WRITE],
981 		(unsigned long long)stat.sectors[STAT_WRITE],
982 		(unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC),
983 		inflight,
984 		jiffies_to_msecs(stat.io_ticks),
985 		(unsigned int)div_u64(stat.nsecs[STAT_READ] +
986 				      stat.nsecs[STAT_WRITE] +
987 				      stat.nsecs[STAT_DISCARD] +
988 				      stat.nsecs[STAT_FLUSH],
989 						NSEC_PER_MSEC),
990 		stat.ios[STAT_DISCARD],
991 		stat.merges[STAT_DISCARD],
992 		(unsigned long long)stat.sectors[STAT_DISCARD],
993 		(unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC),
994 		stat.ios[STAT_FLUSH],
995 		(unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC));
996 }
997 
part_inflight_show(struct device * dev,struct device_attribute * attr,char * buf)998 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
999 			   char *buf)
1000 {
1001 	struct block_device *bdev = dev_to_bdev(dev);
1002 	struct request_queue *q = bdev_get_queue(bdev);
1003 	unsigned int inflight[2];
1004 
1005 	if (queue_is_mq(q))
1006 		blk_mq_in_flight_rw(q, bdev, inflight);
1007 	else
1008 		part_in_flight_rw(bdev, inflight);
1009 
1010 	return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
1011 }
1012 
disk_capability_show(struct device * dev,struct device_attribute * attr,char * buf)1013 static ssize_t disk_capability_show(struct device *dev,
1014 				    struct device_attribute *attr, char *buf)
1015 {
1016 	struct gendisk *disk = dev_to_disk(dev);
1017 
1018 	return sprintf(buf, "%x\n", disk->flags);
1019 }
1020 
disk_alignment_offset_show(struct device * dev,struct device_attribute * attr,char * buf)1021 static ssize_t disk_alignment_offset_show(struct device *dev,
1022 					  struct device_attribute *attr,
1023 					  char *buf)
1024 {
1025 	struct gendisk *disk = dev_to_disk(dev);
1026 
1027 	return sprintf(buf, "%d\n", bdev_alignment_offset(disk->part0));
1028 }
1029 
disk_discard_alignment_show(struct device * dev,struct device_attribute * attr,char * buf)1030 static ssize_t disk_discard_alignment_show(struct device *dev,
1031 					   struct device_attribute *attr,
1032 					   char *buf)
1033 {
1034 	struct gendisk *disk = dev_to_disk(dev);
1035 
1036 	return sprintf(buf, "%d\n", bdev_alignment_offset(disk->part0));
1037 }
1038 
diskseq_show(struct device * dev,struct device_attribute * attr,char * buf)1039 static ssize_t diskseq_show(struct device *dev,
1040 			    struct device_attribute *attr, char *buf)
1041 {
1042 	struct gendisk *disk = dev_to_disk(dev);
1043 
1044 	return sprintf(buf, "%llu\n", disk->diskseq);
1045 }
1046 
1047 static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
1048 static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
1049 static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
1050 static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
1051 static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
1052 static DEVICE_ATTR(size, 0444, part_size_show, NULL);
1053 static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
1054 static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
1055 static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
1056 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
1057 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
1058 static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
1059 static DEVICE_ATTR(diskseq, 0444, diskseq_show, NULL);
1060 
1061 #ifdef CONFIG_FAIL_MAKE_REQUEST
part_fail_show(struct device * dev,struct device_attribute * attr,char * buf)1062 ssize_t part_fail_show(struct device *dev,
1063 		       struct device_attribute *attr, char *buf)
1064 {
1065 	return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail);
1066 }
1067 
part_fail_store(struct device * dev,struct device_attribute * attr,const char * buf,size_t count)1068 ssize_t part_fail_store(struct device *dev,
1069 			struct device_attribute *attr,
1070 			const char *buf, size_t count)
1071 {
1072 	int i;
1073 
1074 	if (count > 0 && sscanf(buf, "%d", &i) > 0)
1075 		dev_to_bdev(dev)->bd_make_it_fail = i;
1076 
1077 	return count;
1078 }
1079 
1080 static struct device_attribute dev_attr_fail =
1081 	__ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
1082 #endif /* CONFIG_FAIL_MAKE_REQUEST */
1083 
1084 #ifdef CONFIG_FAIL_IO_TIMEOUT
1085 static struct device_attribute dev_attr_fail_timeout =
1086 	__ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
1087 #endif
1088 
1089 static struct attribute *disk_attrs[] = {
1090 	&dev_attr_range.attr,
1091 	&dev_attr_ext_range.attr,
1092 	&dev_attr_removable.attr,
1093 	&dev_attr_hidden.attr,
1094 	&dev_attr_ro.attr,
1095 	&dev_attr_size.attr,
1096 	&dev_attr_alignment_offset.attr,
1097 	&dev_attr_discard_alignment.attr,
1098 	&dev_attr_capability.attr,
1099 	&dev_attr_stat.attr,
1100 	&dev_attr_inflight.attr,
1101 	&dev_attr_badblocks.attr,
1102 	&dev_attr_events.attr,
1103 	&dev_attr_events_async.attr,
1104 	&dev_attr_events_poll_msecs.attr,
1105 	&dev_attr_diskseq.attr,
1106 #ifdef CONFIG_FAIL_MAKE_REQUEST
1107 	&dev_attr_fail.attr,
1108 #endif
1109 #ifdef CONFIG_FAIL_IO_TIMEOUT
1110 	&dev_attr_fail_timeout.attr,
1111 #endif
1112 	NULL
1113 };
1114 
disk_visible(struct kobject * kobj,struct attribute * a,int n)1115 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1116 {
1117 	struct device *dev = container_of(kobj, typeof(*dev), kobj);
1118 	struct gendisk *disk = dev_to_disk(dev);
1119 
1120 	if (a == &dev_attr_badblocks.attr && !disk->bb)
1121 		return 0;
1122 	return a->mode;
1123 }
1124 
1125 static struct attribute_group disk_attr_group = {
1126 	.attrs = disk_attrs,
1127 	.is_visible = disk_visible,
1128 };
1129 
1130 static const struct attribute_group *disk_attr_groups[] = {
1131 	&disk_attr_group,
1132 #ifdef CONFIG_BLK_DEV_IO_TRACE
1133 	&blk_trace_attr_group,
1134 #endif
1135 	NULL
1136 };
1137 
1138 /**
1139  * disk_release - releases all allocated resources of the gendisk
1140  * @dev: the device representing this disk
1141  *
1142  * This function releases all allocated resources of the gendisk.
1143  *
1144  * Drivers which used __device_add_disk() have a gendisk with a request_queue
1145  * assigned. Since the request_queue sits on top of the gendisk for these
1146  * drivers we also call blk_put_queue() for them, and we expect the
1147  * request_queue refcount to reach 0 at this point, and so the request_queue
1148  * will also be freed prior to the disk.
1149  *
1150  * Context: can sleep
1151  */
disk_release(struct device * dev)1152 static void disk_release(struct device *dev)
1153 {
1154 	struct gendisk *disk = dev_to_disk(dev);
1155 
1156 	might_sleep();
1157 	WARN_ON_ONCE(disk_live(disk));
1158 
1159 	/*
1160 	 * To undo the all initialization from blk_mq_init_allocated_queue in
1161 	 * case of a probe failure where add_disk is never called we have to
1162 	 * call blk_mq_exit_queue here. We can't do this for the more common
1163 	 * teardown case (yet) as the tagset can be gone by the time the disk
1164 	 * is released once it was added.
1165 	 */
1166 	if (queue_is_mq(disk->queue) &&
1167 	    test_bit(GD_OWNS_QUEUE, &disk->state) &&
1168 	    !test_bit(GD_ADDED, &disk->state))
1169 		blk_mq_exit_queue(disk->queue);
1170 
1171 	blkcg_exit_disk(disk);
1172 
1173 	bioset_exit(&disk->bio_split);
1174 
1175 	disk_release_events(disk);
1176 	kfree(disk->random);
1177 	disk_free_zone_bitmaps(disk);
1178 	xa_destroy(&disk->part_tbl);
1179 
1180 	disk->queue->disk = NULL;
1181 	blk_put_queue(disk->queue);
1182 
1183 	if (test_bit(GD_ADDED, &disk->state) && disk->fops->free_disk)
1184 		disk->fops->free_disk(disk);
1185 
1186 	iput(disk->part0->bd_inode);	/* frees the disk */
1187 }
1188 
block_uevent(struct device * dev,struct kobj_uevent_env * env)1189 static int block_uevent(struct device *dev, struct kobj_uevent_env *env)
1190 {
1191 	struct gendisk *disk = dev_to_disk(dev);
1192 
1193 	return add_uevent_var(env, "DISKSEQ=%llu", disk->diskseq);
1194 }
1195 
1196 struct class block_class = {
1197 	.name		= "block",
1198 	.dev_uevent	= block_uevent,
1199 };
1200 
block_devnode(struct device * dev,umode_t * mode,kuid_t * uid,kgid_t * gid)1201 static char *block_devnode(struct device *dev, umode_t *mode,
1202 			   kuid_t *uid, kgid_t *gid)
1203 {
1204 	struct gendisk *disk = dev_to_disk(dev);
1205 
1206 	if (disk->fops->devnode)
1207 		return disk->fops->devnode(disk, mode);
1208 	return NULL;
1209 }
1210 
1211 const struct device_type disk_type = {
1212 	.name		= "disk",
1213 	.groups		= disk_attr_groups,
1214 	.release	= disk_release,
1215 	.devnode	= block_devnode,
1216 };
1217 
1218 #ifdef CONFIG_PROC_FS
1219 /*
1220  * aggregate disk stat collector.  Uses the same stats that the sysfs
1221  * entries do, above, but makes them available through one seq_file.
1222  *
1223  * The output looks suspiciously like /proc/partitions with a bunch of
1224  * extra fields.
1225  */
diskstats_show(struct seq_file * seqf,void * v)1226 static int diskstats_show(struct seq_file *seqf, void *v)
1227 {
1228 	struct gendisk *gp = v;
1229 	struct block_device *hd;
1230 	unsigned int inflight;
1231 	struct disk_stats stat;
1232 	unsigned long idx;
1233 
1234 	/*
1235 	if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1236 		seq_puts(seqf,	"major minor name"
1237 				"     rio rmerge rsect ruse wio wmerge "
1238 				"wsect wuse running use aveq"
1239 				"\n\n");
1240 	*/
1241 
1242 	rcu_read_lock();
1243 	xa_for_each(&gp->part_tbl, idx, hd) {
1244 		if (bdev_is_partition(hd) && !bdev_nr_sectors(hd))
1245 			continue;
1246 		if (queue_is_mq(gp->queue))
1247 			inflight = blk_mq_in_flight(gp->queue, hd);
1248 		else
1249 			inflight = part_in_flight(hd);
1250 
1251 		if (inflight) {
1252 			part_stat_lock();
1253 			update_io_ticks(hd, jiffies, true);
1254 			part_stat_unlock();
1255 		}
1256 		part_stat_read_all(hd, &stat);
1257 		seq_printf(seqf, "%4d %7d %pg "
1258 			   "%lu %lu %lu %u "
1259 			   "%lu %lu %lu %u "
1260 			   "%u %u %u "
1261 			   "%lu %lu %lu %u "
1262 			   "%lu %u"
1263 			   "\n",
1264 			   MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd,
1265 			   stat.ios[STAT_READ],
1266 			   stat.merges[STAT_READ],
1267 			   stat.sectors[STAT_READ],
1268 			   (unsigned int)div_u64(stat.nsecs[STAT_READ],
1269 							NSEC_PER_MSEC),
1270 			   stat.ios[STAT_WRITE],
1271 			   stat.merges[STAT_WRITE],
1272 			   stat.sectors[STAT_WRITE],
1273 			   (unsigned int)div_u64(stat.nsecs[STAT_WRITE],
1274 							NSEC_PER_MSEC),
1275 			   inflight,
1276 			   jiffies_to_msecs(stat.io_ticks),
1277 			   (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1278 						 stat.nsecs[STAT_WRITE] +
1279 						 stat.nsecs[STAT_DISCARD] +
1280 						 stat.nsecs[STAT_FLUSH],
1281 							NSEC_PER_MSEC),
1282 			   stat.ios[STAT_DISCARD],
1283 			   stat.merges[STAT_DISCARD],
1284 			   stat.sectors[STAT_DISCARD],
1285 			   (unsigned int)div_u64(stat.nsecs[STAT_DISCARD],
1286 						 NSEC_PER_MSEC),
1287 			   stat.ios[STAT_FLUSH],
1288 			   (unsigned int)div_u64(stat.nsecs[STAT_FLUSH],
1289 						 NSEC_PER_MSEC)
1290 			);
1291 	}
1292 	rcu_read_unlock();
1293 
1294 	return 0;
1295 }
1296 
1297 static const struct seq_operations diskstats_op = {
1298 	.start	= disk_seqf_start,
1299 	.next	= disk_seqf_next,
1300 	.stop	= disk_seqf_stop,
1301 	.show	= diskstats_show
1302 };
1303 
proc_genhd_init(void)1304 static int __init proc_genhd_init(void)
1305 {
1306 	proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1307 	proc_create_seq("partitions", 0, NULL, &partitions_op);
1308 	return 0;
1309 }
1310 module_init(proc_genhd_init);
1311 #endif /* CONFIG_PROC_FS */
1312 
part_devt(struct gendisk * disk,u8 partno)1313 dev_t part_devt(struct gendisk *disk, u8 partno)
1314 {
1315 	struct block_device *part;
1316 	dev_t devt = 0;
1317 
1318 	rcu_read_lock();
1319 	part = xa_load(&disk->part_tbl, partno);
1320 	if (part)
1321 		devt = part->bd_dev;
1322 	rcu_read_unlock();
1323 
1324 	return devt;
1325 }
1326 
blk_lookup_devt(const char * name,int partno)1327 dev_t blk_lookup_devt(const char *name, int partno)
1328 {
1329 	dev_t devt = MKDEV(0, 0);
1330 	struct class_dev_iter iter;
1331 	struct device *dev;
1332 
1333 	class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1334 	while ((dev = class_dev_iter_next(&iter))) {
1335 		struct gendisk *disk = dev_to_disk(dev);
1336 
1337 		if (strcmp(dev_name(dev), name))
1338 			continue;
1339 
1340 		if (partno < disk->minors) {
1341 			/* We need to return the right devno, even
1342 			 * if the partition doesn't exist yet.
1343 			 */
1344 			devt = MKDEV(MAJOR(dev->devt),
1345 				     MINOR(dev->devt) + partno);
1346 		} else {
1347 			devt = part_devt(disk, partno);
1348 			if (devt)
1349 				break;
1350 		}
1351 	}
1352 	class_dev_iter_exit(&iter);
1353 	return devt;
1354 }
1355 
__alloc_disk_node(struct request_queue * q,int node_id,struct lock_class_key * lkclass)1356 struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id,
1357 		struct lock_class_key *lkclass)
1358 {
1359 	struct gendisk *disk;
1360 
1361 	disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1362 	if (!disk)
1363 		return NULL;
1364 
1365 	if (bioset_init(&disk->bio_split, BIO_POOL_SIZE, 0, 0))
1366 		goto out_free_disk;
1367 
1368 	disk->bdi = bdi_alloc(node_id);
1369 	if (!disk->bdi)
1370 		goto out_free_bioset;
1371 
1372 	/* bdev_alloc() might need the queue, set before the first call */
1373 	disk->queue = q;
1374 
1375 	disk->part0 = bdev_alloc(disk, 0);
1376 	if (!disk->part0)
1377 		goto out_free_bdi;
1378 
1379 	disk->node_id = node_id;
1380 	mutex_init(&disk->open_mutex);
1381 	xa_init(&disk->part_tbl);
1382 	if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL))
1383 		goto out_destroy_part_tbl;
1384 
1385 	if (blkcg_init_disk(disk))
1386 		goto out_erase_part0;
1387 
1388 	rand_initialize_disk(disk);
1389 	disk_to_dev(disk)->class = &block_class;
1390 	disk_to_dev(disk)->type = &disk_type;
1391 	device_initialize(disk_to_dev(disk));
1392 	inc_diskseq(disk);
1393 	q->disk = disk;
1394 	lockdep_init_map(&disk->lockdep_map, "(bio completion)", lkclass, 0);
1395 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
1396 	INIT_LIST_HEAD(&disk->slave_bdevs);
1397 #endif
1398 	return disk;
1399 
1400 out_erase_part0:
1401 	xa_erase(&disk->part_tbl, 0);
1402 out_destroy_part_tbl:
1403 	xa_destroy(&disk->part_tbl);
1404 	disk->part0->bd_disk = NULL;
1405 	iput(disk->part0->bd_inode);
1406 out_free_bdi:
1407 	bdi_put(disk->bdi);
1408 out_free_bioset:
1409 	bioset_exit(&disk->bio_split);
1410 out_free_disk:
1411 	kfree(disk);
1412 	return NULL;
1413 }
1414 
__blk_alloc_disk(int node,struct lock_class_key * lkclass)1415 struct gendisk *__blk_alloc_disk(int node, struct lock_class_key *lkclass)
1416 {
1417 	struct request_queue *q;
1418 	struct gendisk *disk;
1419 
1420 	q = blk_alloc_queue(node, false);
1421 	if (!q)
1422 		return NULL;
1423 
1424 	disk = __alloc_disk_node(q, node, lkclass);
1425 	if (!disk) {
1426 		blk_put_queue(q);
1427 		return NULL;
1428 	}
1429 	set_bit(GD_OWNS_QUEUE, &disk->state);
1430 	return disk;
1431 }
1432 EXPORT_SYMBOL(__blk_alloc_disk);
1433 
1434 /**
1435  * put_disk - decrements the gendisk refcount
1436  * @disk: the struct gendisk to decrement the refcount for
1437  *
1438  * This decrements the refcount for the struct gendisk. When this reaches 0
1439  * we'll have disk_release() called.
1440  *
1441  * Note: for blk-mq disk put_disk must be called before freeing the tag_set
1442  * when handling probe errors (that is before add_disk() is called).
1443  *
1444  * Context: Any context, but the last reference must not be dropped from
1445  *          atomic context.
1446  */
put_disk(struct gendisk * disk)1447 void put_disk(struct gendisk *disk)
1448 {
1449 	if (disk)
1450 		put_device(disk_to_dev(disk));
1451 }
1452 EXPORT_SYMBOL(put_disk);
1453 
set_disk_ro_uevent(struct gendisk * gd,int ro)1454 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1455 {
1456 	char event[] = "DISK_RO=1";
1457 	char *envp[] = { event, NULL };
1458 
1459 	if (!ro)
1460 		event[8] = '0';
1461 	kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1462 }
1463 
1464 /**
1465  * set_disk_ro - set a gendisk read-only
1466  * @disk:	gendisk to operate on
1467  * @read_only:	%true to set the disk read-only, %false set the disk read/write
1468  *
1469  * This function is used to indicate whether a given disk device should have its
1470  * read-only flag set. set_disk_ro() is typically used by device drivers to
1471  * indicate whether the underlying physical device is write-protected.
1472  */
set_disk_ro(struct gendisk * disk,bool read_only)1473 void set_disk_ro(struct gendisk *disk, bool read_only)
1474 {
1475 	if (read_only) {
1476 		if (test_and_set_bit(GD_READ_ONLY, &disk->state))
1477 			return;
1478 	} else {
1479 		if (!test_and_clear_bit(GD_READ_ONLY, &disk->state))
1480 			return;
1481 	}
1482 	set_disk_ro_uevent(disk, read_only);
1483 }
1484 EXPORT_SYMBOL(set_disk_ro);
1485 
inc_diskseq(struct gendisk * disk)1486 void inc_diskseq(struct gendisk *disk)
1487 {
1488 	disk->diskseq = atomic64_inc_return(&diskseq);
1489 }
1490