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