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 "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 :
58 0 : void set_capacity(struct gendisk *disk, sector_t sectors)
59 : {
60 0 : struct block_device *bdev = disk->part0;
61 :
62 0 : spin_lock(&bdev->bd_size_lock);
63 0 : i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
64 0 : bdev->bd_nr_sectors = sectors;
65 0 : spin_unlock(&bdev->bd_size_lock);
66 0 : }
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 : */
73 0 : bool set_capacity_and_notify(struct gendisk *disk, sector_t size)
74 : {
75 0 : sector_t capacity = get_capacity(disk);
76 0 : char *envp[] = { "RESIZE=1", NULL };
77 :
78 0 : 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 0 : if (size == capacity ||
86 0 : !disk_live(disk) ||
87 0 : (disk->flags & GENHD_FL_HIDDEN))
88 : return false;
89 :
90 0 : 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 0 : if (!capacity || !size)
98 : return false;
99 0 : kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
100 0 : return true;
101 : }
102 : EXPORT_SYMBOL_GPL(set_capacity_and_notify);
103 :
104 : /*
105 : * Format the device name of the indicated block device into the supplied buffer
106 : * and return a pointer to that same buffer for convenience.
107 : *
108 : * Note: do not use this in new code, use the %pg specifier to sprintf and
109 : * printk insted.
110 : */
111 0 : const char *bdevname(struct block_device *bdev, char *buf)
112 : {
113 0 : struct gendisk *hd = bdev->bd_disk;
114 0 : int partno = bdev->bd_partno;
115 :
116 0 : if (!partno)
117 0 : snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name);
118 0 : else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1]))
119 0 : snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno);
120 : else
121 0 : snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno);
122 :
123 0 : return buf;
124 : }
125 : EXPORT_SYMBOL(bdevname);
126 :
127 0 : static void part_stat_read_all(struct block_device *part,
128 : struct disk_stats *stat)
129 : {
130 : int cpu;
131 :
132 0 : memset(stat, 0, sizeof(struct disk_stats));
133 0 : for_each_possible_cpu(cpu) {
134 0 : struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu);
135 : int group;
136 :
137 0 : for (group = 0; group < NR_STAT_GROUPS; group++) {
138 0 : stat->nsecs[group] += ptr->nsecs[group];
139 0 : stat->sectors[group] += ptr->sectors[group];
140 0 : stat->ios[group] += ptr->ios[group];
141 0 : stat->merges[group] += ptr->merges[group];
142 : }
143 :
144 0 : stat->io_ticks += ptr->io_ticks;
145 : }
146 0 : }
147 :
148 : static unsigned int part_in_flight(struct block_device *part)
149 : {
150 : unsigned int inflight = 0;
151 : int cpu;
152 :
153 0 : for_each_possible_cpu(cpu) {
154 0 : inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) +
155 0 : part_stat_local_read_cpu(part, in_flight[1], cpu);
156 : }
157 0 : if ((int)inflight < 0)
158 0 : inflight = 0;
159 :
160 : return inflight;
161 : }
162 :
163 : static void part_in_flight_rw(struct block_device *part,
164 : unsigned int inflight[2])
165 : {
166 : int cpu;
167 :
168 0 : inflight[0] = 0;
169 0 : inflight[1] = 0;
170 0 : for_each_possible_cpu(cpu) {
171 0 : inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu);
172 0 : inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu);
173 : }
174 0 : if ((int)inflight[0] < 0)
175 0 : inflight[0] = 0;
176 0 : if ((int)inflight[1] < 0)
177 0 : inflight[1] = 0;
178 : }
179 :
180 : /*
181 : * Can be deleted altogether. Later.
182 : *
183 : */
184 : #define BLKDEV_MAJOR_HASH_SIZE 255
185 : static struct blk_major_name {
186 : struct blk_major_name *next;
187 : int major;
188 : char name[16];
189 : #ifdef CONFIG_BLOCK_LEGACY_AUTOLOAD
190 : void (*probe)(dev_t devt);
191 : #endif
192 : } *major_names[BLKDEV_MAJOR_HASH_SIZE];
193 : static DEFINE_MUTEX(major_names_lock);
194 : static DEFINE_SPINLOCK(major_names_spinlock);
195 :
196 : /* index in the above - for now: assume no multimajor ranges */
197 : static inline int major_to_index(unsigned major)
198 : {
199 1 : return major % BLKDEV_MAJOR_HASH_SIZE;
200 : }
201 :
202 : #ifdef CONFIG_PROC_FS
203 0 : void blkdev_show(struct seq_file *seqf, off_t offset)
204 : {
205 : struct blk_major_name *dp;
206 :
207 0 : spin_lock(&major_names_spinlock);
208 0 : for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
209 0 : if (dp->major == offset)
210 0 : seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
211 0 : spin_unlock(&major_names_spinlock);
212 0 : }
213 : #endif /* CONFIG_PROC_FS */
214 :
215 : /**
216 : * __register_blkdev - register a new block device
217 : *
218 : * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
219 : * @major = 0, try to allocate any unused major number.
220 : * @name: the name of the new block device as a zero terminated string
221 : * @probe: pre-devtmpfs / pre-udev callback used to create disks when their
222 : * pre-created device node is accessed. When a probe call uses
223 : * add_disk() and it fails the driver must cleanup resources. This
224 : * interface may soon be removed.
225 : *
226 : * The @name must be unique within the system.
227 : *
228 : * The return value depends on the @major input parameter:
229 : *
230 : * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
231 : * then the function returns zero on success, or a negative error code
232 : * - if any unused major number was requested with @major = 0 parameter
233 : * then the return value is the allocated major number in range
234 : * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
235 : *
236 : * See Documentation/admin-guide/devices.txt for the list of allocated
237 : * major numbers.
238 : *
239 : * Use register_blkdev instead for any new code.
240 : */
241 1 : int __register_blkdev(unsigned int major, const char *name,
242 : void (*probe)(dev_t devt))
243 : {
244 : struct blk_major_name **n, *p;
245 1 : int index, ret = 0;
246 :
247 1 : mutex_lock(&major_names_lock);
248 :
249 : /* temporary */
250 1 : if (major == 0) {
251 0 : for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
252 0 : if (major_names[index] == NULL)
253 : break;
254 : }
255 :
256 0 : if (index == 0) {
257 0 : printk("%s: failed to get major for %s\n",
258 : __func__, name);
259 0 : ret = -EBUSY;
260 0 : goto out;
261 : }
262 0 : major = index;
263 0 : ret = major;
264 : }
265 :
266 1 : if (major >= BLKDEV_MAJOR_MAX) {
267 0 : pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
268 : __func__, major, BLKDEV_MAJOR_MAX-1, name);
269 :
270 0 : ret = -EINVAL;
271 0 : goto out;
272 : }
273 :
274 1 : p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
275 1 : if (p == NULL) {
276 : ret = -ENOMEM;
277 : goto out;
278 : }
279 :
280 1 : p->major = major;
281 : #ifdef CONFIG_BLOCK_LEGACY_AUTOLOAD
282 1 : p->probe = probe;
283 : #endif
284 1 : strlcpy(p->name, name, sizeof(p->name));
285 1 : p->next = NULL;
286 1 : index = major_to_index(major);
287 :
288 1 : spin_lock(&major_names_spinlock);
289 1 : for (n = &major_names[index]; *n; n = &(*n)->next) {
290 0 : if ((*n)->major == major)
291 : break;
292 : }
293 1 : if (!*n)
294 1 : *n = p;
295 : else
296 : ret = -EBUSY;
297 1 : spin_unlock(&major_names_spinlock);
298 :
299 1 : if (ret < 0) {
300 0 : printk("register_blkdev: cannot get major %u for %s\n",
301 : major, name);
302 0 : kfree(p);
303 : }
304 : out:
305 1 : mutex_unlock(&major_names_lock);
306 1 : return ret;
307 : }
308 : EXPORT_SYMBOL(__register_blkdev);
309 :
310 0 : void unregister_blkdev(unsigned int major, const char *name)
311 : {
312 : struct blk_major_name **n;
313 0 : struct blk_major_name *p = NULL;
314 0 : int index = major_to_index(major);
315 :
316 0 : mutex_lock(&major_names_lock);
317 0 : spin_lock(&major_names_spinlock);
318 0 : for (n = &major_names[index]; *n; n = &(*n)->next)
319 0 : if ((*n)->major == major)
320 : break;
321 0 : if (!*n || strcmp((*n)->name, name)) {
322 0 : WARN_ON(1);
323 : } else {
324 0 : p = *n;
325 0 : *n = p->next;
326 : }
327 0 : spin_unlock(&major_names_spinlock);
328 0 : mutex_unlock(&major_names_lock);
329 0 : kfree(p);
330 0 : }
331 :
332 : EXPORT_SYMBOL(unregister_blkdev);
333 :
334 0 : int blk_alloc_ext_minor(void)
335 : {
336 : int idx;
337 :
338 0 : idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT - 1, GFP_KERNEL);
339 0 : if (idx == -ENOSPC)
340 : return -EBUSY;
341 0 : return idx;
342 : }
343 :
344 0 : void blk_free_ext_minor(unsigned int minor)
345 : {
346 0 : ida_free(&ext_devt_ida, minor);
347 0 : }
348 :
349 0 : static char *bdevt_str(dev_t devt, char *buf)
350 : {
351 0 : if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
352 : char tbuf[BDEVT_SIZE];
353 0 : snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
354 0 : snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
355 : } else
356 0 : snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
357 :
358 0 : return buf;
359 : }
360 :
361 0 : void disk_uevent(struct gendisk *disk, enum kobject_action action)
362 : {
363 : struct block_device *part;
364 : unsigned long idx;
365 :
366 : rcu_read_lock();
367 0 : xa_for_each(&disk->part_tbl, idx, part) {
368 0 : if (bdev_is_partition(part) && !bdev_nr_sectors(part))
369 0 : continue;
370 0 : if (!kobject_get_unless_zero(&part->bd_device.kobj))
371 0 : continue;
372 :
373 : rcu_read_unlock();
374 0 : kobject_uevent(bdev_kobj(part), action);
375 0 : put_device(&part->bd_device);
376 : rcu_read_lock();
377 : }
378 : rcu_read_unlock();
379 0 : }
380 : EXPORT_SYMBOL_GPL(disk_uevent);
381 :
382 0 : int disk_scan_partitions(struct gendisk *disk, fmode_t mode)
383 : {
384 : struct block_device *bdev;
385 :
386 0 : if (disk->flags & (GENHD_FL_NO_PART | GENHD_FL_HIDDEN))
387 : return -EINVAL;
388 0 : if (disk->open_partitions)
389 : return -EBUSY;
390 :
391 0 : set_bit(GD_NEED_PART_SCAN, &disk->state);
392 0 : bdev = blkdev_get_by_dev(disk_devt(disk), mode, NULL);
393 0 : if (IS_ERR(bdev))
394 0 : return PTR_ERR(bdev);
395 0 : blkdev_put(bdev, mode);
396 0 : return 0;
397 : }
398 :
399 : /**
400 : * device_add_disk - add disk information to kernel list
401 : * @parent: parent device for the disk
402 : * @disk: per-device partitioning information
403 : * @groups: Additional per-device sysfs groups
404 : *
405 : * This function registers the partitioning information in @disk
406 : * with the kernel.
407 : */
408 0 : int __must_check device_add_disk(struct device *parent, struct gendisk *disk,
409 : const struct attribute_group **groups)
410 :
411 : {
412 0 : struct device *ddev = disk_to_dev(disk);
413 : int ret;
414 :
415 : /* Only makes sense for bio-based to set ->poll_bio */
416 0 : if (queue_is_mq(disk->queue) && disk->fops->poll_bio)
417 : return -EINVAL;
418 :
419 : /*
420 : * The disk queue should now be all set with enough information about
421 : * the device for the elevator code to pick an adequate default
422 : * elevator if one is needed, that is, for devices requesting queue
423 : * registration.
424 : */
425 0 : elevator_init_mq(disk->queue);
426 :
427 : /*
428 : * If the driver provides an explicit major number it also must provide
429 : * the number of minors numbers supported, and those will be used to
430 : * setup the gendisk.
431 : * Otherwise just allocate the device numbers for both the whole device
432 : * and all partitions from the extended dev_t space.
433 : */
434 0 : if (disk->major) {
435 0 : if (WARN_ON(!disk->minors))
436 : return -EINVAL;
437 :
438 0 : if (disk->minors > DISK_MAX_PARTS) {
439 0 : pr_err("block: can't allocate more than %d partitions\n",
440 : DISK_MAX_PARTS);
441 0 : disk->minors = DISK_MAX_PARTS;
442 : }
443 0 : if (disk->first_minor + disk->minors > MINORMASK + 1)
444 : return -EINVAL;
445 : } else {
446 0 : if (WARN_ON(disk->minors))
447 : return -EINVAL;
448 :
449 0 : ret = blk_alloc_ext_minor();
450 0 : if (ret < 0)
451 : return ret;
452 0 : disk->major = BLOCK_EXT_MAJOR;
453 0 : disk->first_minor = ret;
454 : }
455 :
456 : /* delay uevents, until we scanned partition table */
457 0 : dev_set_uevent_suppress(ddev, 1);
458 :
459 0 : ddev->parent = parent;
460 0 : ddev->groups = groups;
461 0 : dev_set_name(ddev, "%s", disk->disk_name);
462 0 : if (!(disk->flags & GENHD_FL_HIDDEN))
463 0 : ddev->devt = MKDEV(disk->major, disk->first_minor);
464 0 : ret = device_add(ddev);
465 0 : if (ret)
466 : goto out_free_ext_minor;
467 :
468 0 : ret = disk_alloc_events(disk);
469 0 : if (ret)
470 : goto out_device_del;
471 :
472 : if (!sysfs_deprecated) {
473 0 : ret = sysfs_create_link(block_depr, &ddev->kobj,
474 0 : kobject_name(&ddev->kobj));
475 0 : if (ret)
476 : goto out_device_del;
477 : }
478 :
479 : /*
480 : * avoid probable deadlock caused by allocating memory with
481 : * GFP_KERNEL in runtime_resume callback of its all ancestor
482 : * devices
483 : */
484 0 : pm_runtime_set_memalloc_noio(ddev, true);
485 :
486 0 : ret = blk_integrity_add(disk);
487 : if (ret)
488 : goto out_del_block_link;
489 :
490 0 : disk->part0->bd_holder_dir =
491 0 : kobject_create_and_add("holders", &ddev->kobj);
492 0 : if (!disk->part0->bd_holder_dir) {
493 : ret = -ENOMEM;
494 : goto out_del_integrity;
495 : }
496 0 : disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
497 0 : if (!disk->slave_dir) {
498 : ret = -ENOMEM;
499 : goto out_put_holder_dir;
500 : }
501 :
502 0 : ret = bd_register_pending_holders(disk);
503 : if (ret < 0)
504 : goto out_put_slave_dir;
505 :
506 0 : ret = blk_register_queue(disk);
507 0 : if (ret)
508 : goto out_put_slave_dir;
509 :
510 0 : if (!(disk->flags & GENHD_FL_HIDDEN)) {
511 0 : ret = bdi_register(disk->bdi, "%u:%u",
512 : disk->major, disk->first_minor);
513 0 : if (ret)
514 : goto out_unregister_queue;
515 0 : bdi_set_owner(disk->bdi, ddev);
516 0 : ret = sysfs_create_link(&ddev->kobj,
517 0 : &disk->bdi->dev->kobj, "bdi");
518 0 : if (ret)
519 : goto out_unregister_bdi;
520 :
521 0 : bdev_add(disk->part0, ddev->devt);
522 0 : if (get_capacity(disk))
523 0 : disk_scan_partitions(disk, FMODE_READ);
524 :
525 : /*
526 : * Announce the disk and partitions after all partitions are
527 : * created. (for hidden disks uevents remain suppressed forever)
528 : */
529 0 : dev_set_uevent_suppress(ddev, 0);
530 0 : disk_uevent(disk, KOBJ_ADD);
531 : }
532 :
533 0 : disk_update_readahead(disk);
534 0 : disk_add_events(disk);
535 0 : set_bit(GD_ADDED, &disk->state);
536 0 : return 0;
537 :
538 : out_unregister_bdi:
539 0 : if (!(disk->flags & GENHD_FL_HIDDEN))
540 0 : bdi_unregister(disk->bdi);
541 : out_unregister_queue:
542 0 : blk_unregister_queue(disk);
543 : out_put_slave_dir:
544 0 : kobject_put(disk->slave_dir);
545 : out_put_holder_dir:
546 0 : kobject_put(disk->part0->bd_holder_dir);
547 : out_del_integrity:
548 0 : blk_integrity_del(disk);
549 : out_del_block_link:
550 : if (!sysfs_deprecated)
551 0 : sysfs_remove_link(block_depr, dev_name(ddev));
552 : out_device_del:
553 0 : device_del(ddev);
554 : out_free_ext_minor:
555 0 : if (disk->major == BLOCK_EXT_MAJOR)
556 0 : blk_free_ext_minor(disk->first_minor);
557 : return ret;
558 : }
559 : EXPORT_SYMBOL(device_add_disk);
560 :
561 : /**
562 : * blk_mark_disk_dead - mark a disk as dead
563 : * @disk: disk to mark as dead
564 : *
565 : * Mark as disk as dead (e.g. surprise removed) and don't accept any new I/O
566 : * to this disk.
567 : */
568 0 : void blk_mark_disk_dead(struct gendisk *disk)
569 : {
570 0 : set_bit(GD_DEAD, &disk->state);
571 0 : blk_queue_start_drain(disk->queue);
572 0 : }
573 : EXPORT_SYMBOL_GPL(blk_mark_disk_dead);
574 :
575 : /**
576 : * del_gendisk - remove the gendisk
577 : * @disk: the struct gendisk to remove
578 : *
579 : * Removes the gendisk and all its associated resources. This deletes the
580 : * partitions associated with the gendisk, and unregisters the associated
581 : * request_queue.
582 : *
583 : * This is the counter to the respective __device_add_disk() call.
584 : *
585 : * The final removal of the struct gendisk happens when its refcount reaches 0
586 : * with put_disk(), which should be called after del_gendisk(), if
587 : * __device_add_disk() was used.
588 : *
589 : * Drivers exist which depend on the release of the gendisk to be synchronous,
590 : * it should not be deferred.
591 : *
592 : * Context: can sleep
593 : */
594 0 : void del_gendisk(struct gendisk *disk)
595 : {
596 0 : struct request_queue *q = disk->queue;
597 :
598 : might_sleep();
599 :
600 0 : if (WARN_ON_ONCE(!disk_live(disk) && !(disk->flags & GENHD_FL_HIDDEN)))
601 : return;
602 :
603 0 : blk_integrity_del(disk);
604 0 : disk_del_events(disk);
605 :
606 0 : mutex_lock(&disk->open_mutex);
607 0 : remove_inode_hash(disk->part0->bd_inode);
608 0 : blk_drop_partitions(disk);
609 0 : mutex_unlock(&disk->open_mutex);
610 :
611 0 : fsync_bdev(disk->part0);
612 0 : __invalidate_device(disk->part0, true);
613 :
614 : /*
615 : * Fail any new I/O.
616 : */
617 0 : set_bit(GD_DEAD, &disk->state);
618 0 : set_capacity(disk, 0);
619 :
620 : /*
621 : * Prevent new I/O from crossing bio_queue_enter().
622 : */
623 0 : blk_queue_start_drain(q);
624 :
625 0 : if (!(disk->flags & GENHD_FL_HIDDEN)) {
626 0 : sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
627 :
628 : /*
629 : * Unregister bdi before releasing device numbers (as they can
630 : * get reused and we'd get clashes in sysfs).
631 : */
632 0 : bdi_unregister(disk->bdi);
633 : }
634 :
635 0 : blk_unregister_queue(disk);
636 :
637 0 : kobject_put(disk->part0->bd_holder_dir);
638 0 : kobject_put(disk->slave_dir);
639 :
640 0 : part_stat_set_all(disk->part0, 0);
641 0 : disk->part0->bd_stamp = 0;
642 : if (!sysfs_deprecated)
643 0 : sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
644 0 : pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
645 0 : device_del(disk_to_dev(disk));
646 :
647 0 : blk_mq_freeze_queue_wait(q);
648 :
649 0 : blk_throtl_cancel_bios(disk->queue);
650 :
651 0 : blk_sync_queue(q);
652 : blk_flush_integrity();
653 : /*
654 : * Allow using passthrough request again after the queue is torn down.
655 : */
656 0 : blk_queue_flag_clear(QUEUE_FLAG_INIT_DONE, q);
657 0 : __blk_mq_unfreeze_queue(q, true);
658 :
659 : }
660 : EXPORT_SYMBOL(del_gendisk);
661 :
662 : /**
663 : * invalidate_disk - invalidate the disk
664 : * @disk: the struct gendisk to invalidate
665 : *
666 : * A helper to invalidates the disk. It will clean the disk's associated
667 : * buffer/page caches and reset its internal states so that the disk
668 : * can be reused by the drivers.
669 : *
670 : * Context: can sleep
671 : */
672 0 : void invalidate_disk(struct gendisk *disk)
673 : {
674 0 : struct block_device *bdev = disk->part0;
675 :
676 0 : invalidate_bdev(bdev);
677 0 : bdev->bd_inode->i_mapping->wb_err = 0;
678 0 : set_capacity(disk, 0);
679 0 : }
680 : EXPORT_SYMBOL(invalidate_disk);
681 :
682 : /* sysfs access to bad-blocks list. */
683 0 : static ssize_t disk_badblocks_show(struct device *dev,
684 : struct device_attribute *attr,
685 : char *page)
686 : {
687 0 : struct gendisk *disk = dev_to_disk(dev);
688 :
689 0 : if (!disk->bb)
690 0 : return sprintf(page, "\n");
691 :
692 0 : return badblocks_show(disk->bb, page, 0);
693 : }
694 :
695 0 : static ssize_t disk_badblocks_store(struct device *dev,
696 : struct device_attribute *attr,
697 : const char *page, size_t len)
698 : {
699 0 : struct gendisk *disk = dev_to_disk(dev);
700 :
701 0 : if (!disk->bb)
702 : return -ENXIO;
703 :
704 0 : return badblocks_store(disk->bb, page, len, 0);
705 : }
706 :
707 : #ifdef CONFIG_BLOCK_LEGACY_AUTOLOAD
708 0 : void blk_request_module(dev_t devt)
709 : {
710 0 : unsigned int major = MAJOR(devt);
711 : struct blk_major_name **n;
712 :
713 0 : mutex_lock(&major_names_lock);
714 0 : for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) {
715 0 : if ((*n)->major == major && (*n)->probe) {
716 0 : (*n)->probe(devt);
717 0 : mutex_unlock(&major_names_lock);
718 0 : return;
719 : }
720 : }
721 0 : mutex_unlock(&major_names_lock);
722 :
723 0 : if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
724 : /* Make old-style 2.4 aliases work */
725 : request_module("block-major-%d", MAJOR(devt));
726 : }
727 : #endif /* CONFIG_BLOCK_LEGACY_AUTOLOAD */
728 :
729 : /*
730 : * print a full list of all partitions - intended for places where the root
731 : * filesystem can't be mounted and thus to give the victim some idea of what
732 : * went wrong
733 : */
734 0 : void __init printk_all_partitions(void)
735 : {
736 : struct class_dev_iter iter;
737 : struct device *dev;
738 :
739 0 : class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
740 0 : while ((dev = class_dev_iter_next(&iter))) {
741 0 : struct gendisk *disk = dev_to_disk(dev);
742 : struct block_device *part;
743 : char devt_buf[BDEVT_SIZE];
744 : unsigned long idx;
745 :
746 : /*
747 : * Don't show empty devices or things that have been
748 : * suppressed
749 : */
750 0 : if (get_capacity(disk) == 0 || (disk->flags & GENHD_FL_HIDDEN))
751 0 : continue;
752 :
753 : /*
754 : * Note, unlike /proc/partitions, I am showing the numbers in
755 : * hex - the same format as the root= option takes.
756 : */
757 : rcu_read_lock();
758 0 : xa_for_each(&disk->part_tbl, idx, part) {
759 0 : if (!bdev_nr_sectors(part))
760 0 : continue;
761 0 : printk("%s%s %10llu %pg %s",
762 : bdev_is_partition(part) ? " " : "",
763 : bdevt_str(part->bd_dev, devt_buf),
764 : bdev_nr_sectors(part) >> 1, part,
765 : part->bd_meta_info ?
766 : part->bd_meta_info->uuid : "");
767 0 : if (bdev_is_partition(part))
768 0 : printk("\n");
769 0 : else if (dev->parent && dev->parent->driver)
770 0 : printk(" driver: %s\n",
771 : dev->parent->driver->name);
772 : else
773 0 : printk(" (driver?)\n");
774 : }
775 : rcu_read_unlock();
776 : }
777 0 : class_dev_iter_exit(&iter);
778 0 : }
779 :
780 : #ifdef CONFIG_PROC_FS
781 : /* iterator */
782 0 : static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
783 : {
784 0 : loff_t skip = *pos;
785 : struct class_dev_iter *iter;
786 : struct device *dev;
787 :
788 0 : iter = kmalloc(sizeof(*iter), GFP_KERNEL);
789 0 : if (!iter)
790 : return ERR_PTR(-ENOMEM);
791 :
792 0 : seqf->private = iter;
793 0 : class_dev_iter_init(iter, &block_class, NULL, &disk_type);
794 : do {
795 0 : dev = class_dev_iter_next(iter);
796 0 : if (!dev)
797 : return NULL;
798 0 : } while (skip--);
799 :
800 0 : return dev_to_disk(dev);
801 : }
802 :
803 0 : static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
804 : {
805 : struct device *dev;
806 :
807 0 : (*pos)++;
808 0 : dev = class_dev_iter_next(seqf->private);
809 0 : if (dev)
810 0 : return dev_to_disk(dev);
811 :
812 : return NULL;
813 : }
814 :
815 0 : static void disk_seqf_stop(struct seq_file *seqf, void *v)
816 : {
817 0 : struct class_dev_iter *iter = seqf->private;
818 :
819 : /* stop is called even after start failed :-( */
820 0 : if (iter) {
821 0 : class_dev_iter_exit(iter);
822 0 : kfree(iter);
823 0 : seqf->private = NULL;
824 : }
825 0 : }
826 :
827 0 : static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
828 : {
829 : void *p;
830 :
831 0 : p = disk_seqf_start(seqf, pos);
832 0 : if (!IS_ERR_OR_NULL(p) && !*pos)
833 0 : seq_puts(seqf, "major minor #blocks name\n\n");
834 0 : return p;
835 : }
836 :
837 0 : static int show_partition(struct seq_file *seqf, void *v)
838 : {
839 0 : struct gendisk *sgp = v;
840 : struct block_device *part;
841 : unsigned long idx;
842 :
843 0 : if (!get_capacity(sgp) || (sgp->flags & GENHD_FL_HIDDEN))
844 : return 0;
845 :
846 : rcu_read_lock();
847 0 : xa_for_each(&sgp->part_tbl, idx, part) {
848 0 : if (!bdev_nr_sectors(part))
849 0 : continue;
850 0 : seq_printf(seqf, "%4d %7d %10llu %pg\n",
851 0 : MAJOR(part->bd_dev), MINOR(part->bd_dev),
852 0 : bdev_nr_sectors(part) >> 1, part);
853 : }
854 : rcu_read_unlock();
855 0 : return 0;
856 : }
857 :
858 : static const struct seq_operations partitions_op = {
859 : .start = show_partition_start,
860 : .next = disk_seqf_next,
861 : .stop = disk_seqf_stop,
862 : .show = show_partition
863 : };
864 : #endif
865 :
866 1 : static int __init genhd_device_init(void)
867 : {
868 : int error;
869 :
870 1 : block_class.dev_kobj = sysfs_dev_block_kobj;
871 1 : error = class_register(&block_class);
872 1 : if (unlikely(error))
873 : return error;
874 1 : blk_dev_init();
875 :
876 1 : register_blkdev(BLOCK_EXT_MAJOR, "blkext");
877 :
878 : /* create top-level block dir */
879 : if (!sysfs_deprecated)
880 1 : block_depr = kobject_create_and_add("block", NULL);
881 1 : return 0;
882 : }
883 :
884 : subsys_initcall(genhd_device_init);
885 :
886 0 : static ssize_t disk_range_show(struct device *dev,
887 : struct device_attribute *attr, char *buf)
888 : {
889 0 : struct gendisk *disk = dev_to_disk(dev);
890 :
891 0 : return sprintf(buf, "%d\n", disk->minors);
892 : }
893 :
894 0 : static ssize_t disk_ext_range_show(struct device *dev,
895 : struct device_attribute *attr, char *buf)
896 : {
897 0 : struct gendisk *disk = dev_to_disk(dev);
898 :
899 0 : return sprintf(buf, "%d\n",
900 0 : (disk->flags & GENHD_FL_NO_PART) ? 1 : DISK_MAX_PARTS);
901 : }
902 :
903 0 : static ssize_t disk_removable_show(struct device *dev,
904 : struct device_attribute *attr, char *buf)
905 : {
906 0 : struct gendisk *disk = dev_to_disk(dev);
907 :
908 0 : return sprintf(buf, "%d\n",
909 0 : (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
910 : }
911 :
912 0 : static ssize_t disk_hidden_show(struct device *dev,
913 : struct device_attribute *attr, char *buf)
914 : {
915 0 : struct gendisk *disk = dev_to_disk(dev);
916 :
917 0 : return sprintf(buf, "%d\n",
918 0 : (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
919 : }
920 :
921 0 : static ssize_t disk_ro_show(struct device *dev,
922 : struct device_attribute *attr, char *buf)
923 : {
924 0 : struct gendisk *disk = dev_to_disk(dev);
925 :
926 0 : return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
927 : }
928 :
929 0 : ssize_t part_size_show(struct device *dev,
930 : struct device_attribute *attr, char *buf)
931 : {
932 0 : return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev)));
933 : }
934 :
935 0 : ssize_t part_stat_show(struct device *dev,
936 : struct device_attribute *attr, char *buf)
937 : {
938 0 : struct block_device *bdev = dev_to_bdev(dev);
939 0 : struct request_queue *q = bdev_get_queue(bdev);
940 : struct disk_stats stat;
941 : unsigned int inflight;
942 :
943 0 : if (queue_is_mq(q))
944 0 : inflight = blk_mq_in_flight(q, bdev);
945 : else
946 : inflight = part_in_flight(bdev);
947 :
948 0 : if (inflight) {
949 0 : part_stat_lock();
950 0 : update_io_ticks(bdev, jiffies, true);
951 0 : part_stat_unlock();
952 : }
953 0 : part_stat_read_all(bdev, &stat);
954 0 : return sprintf(buf,
955 : "%8lu %8lu %8llu %8u "
956 : "%8lu %8lu %8llu %8u "
957 : "%8u %8u %8u "
958 : "%8lu %8lu %8llu %8u "
959 : "%8lu %8u"
960 : "\n",
961 : stat.ios[STAT_READ],
962 : stat.merges[STAT_READ],
963 0 : (unsigned long long)stat.sectors[STAT_READ],
964 0 : (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC),
965 : stat.ios[STAT_WRITE],
966 : stat.merges[STAT_WRITE],
967 0 : (unsigned long long)stat.sectors[STAT_WRITE],
968 0 : (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC),
969 : inflight,
970 : jiffies_to_msecs(stat.io_ticks),
971 0 : (unsigned int)div_u64(stat.nsecs[STAT_READ] +
972 0 : stat.nsecs[STAT_WRITE] +
973 : stat.nsecs[STAT_DISCARD] +
974 : stat.nsecs[STAT_FLUSH],
975 : NSEC_PER_MSEC),
976 : stat.ios[STAT_DISCARD],
977 : stat.merges[STAT_DISCARD],
978 0 : (unsigned long long)stat.sectors[STAT_DISCARD],
979 0 : (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC),
980 : stat.ios[STAT_FLUSH],
981 0 : (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC));
982 : }
983 :
984 0 : ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
985 : char *buf)
986 : {
987 0 : struct block_device *bdev = dev_to_bdev(dev);
988 0 : struct request_queue *q = bdev_get_queue(bdev);
989 : unsigned int inflight[2];
990 :
991 0 : if (queue_is_mq(q))
992 0 : blk_mq_in_flight_rw(q, bdev, inflight);
993 : else
994 0 : part_in_flight_rw(bdev, inflight);
995 :
996 0 : return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
997 : }
998 :
999 0 : static ssize_t disk_capability_show(struct device *dev,
1000 : struct device_attribute *attr, char *buf)
1001 : {
1002 0 : struct gendisk *disk = dev_to_disk(dev);
1003 :
1004 0 : return sprintf(buf, "%x\n", disk->flags);
1005 : }
1006 :
1007 0 : static ssize_t disk_alignment_offset_show(struct device *dev,
1008 : struct device_attribute *attr,
1009 : char *buf)
1010 : {
1011 0 : struct gendisk *disk = dev_to_disk(dev);
1012 :
1013 0 : return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
1014 : }
1015 :
1016 0 : static ssize_t disk_discard_alignment_show(struct device *dev,
1017 : struct device_attribute *attr,
1018 : char *buf)
1019 : {
1020 0 : struct gendisk *disk = dev_to_disk(dev);
1021 :
1022 0 : return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
1023 : }
1024 :
1025 0 : static ssize_t diskseq_show(struct device *dev,
1026 : struct device_attribute *attr, char *buf)
1027 : {
1028 0 : struct gendisk *disk = dev_to_disk(dev);
1029 :
1030 0 : return sprintf(buf, "%llu\n", disk->diskseq);
1031 : }
1032 :
1033 : static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
1034 : static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
1035 : static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
1036 : static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
1037 : static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
1038 : static DEVICE_ATTR(size, 0444, part_size_show, NULL);
1039 : static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
1040 : static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
1041 : static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
1042 : static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
1043 : static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
1044 : static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
1045 : static DEVICE_ATTR(diskseq, 0444, diskseq_show, NULL);
1046 :
1047 : #ifdef CONFIG_FAIL_MAKE_REQUEST
1048 : ssize_t part_fail_show(struct device *dev,
1049 : struct device_attribute *attr, char *buf)
1050 : {
1051 : return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail);
1052 : }
1053 :
1054 : ssize_t part_fail_store(struct device *dev,
1055 : struct device_attribute *attr,
1056 : const char *buf, size_t count)
1057 : {
1058 : int i;
1059 :
1060 : if (count > 0 && sscanf(buf, "%d", &i) > 0)
1061 : dev_to_bdev(dev)->bd_make_it_fail = i;
1062 :
1063 : return count;
1064 : }
1065 :
1066 : static struct device_attribute dev_attr_fail =
1067 : __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
1068 : #endif /* CONFIG_FAIL_MAKE_REQUEST */
1069 :
1070 : #ifdef CONFIG_FAIL_IO_TIMEOUT
1071 : static struct device_attribute dev_attr_fail_timeout =
1072 : __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
1073 : #endif
1074 :
1075 : static struct attribute *disk_attrs[] = {
1076 : &dev_attr_range.attr,
1077 : &dev_attr_ext_range.attr,
1078 : &dev_attr_removable.attr,
1079 : &dev_attr_hidden.attr,
1080 : &dev_attr_ro.attr,
1081 : &dev_attr_size.attr,
1082 : &dev_attr_alignment_offset.attr,
1083 : &dev_attr_discard_alignment.attr,
1084 : &dev_attr_capability.attr,
1085 : &dev_attr_stat.attr,
1086 : &dev_attr_inflight.attr,
1087 : &dev_attr_badblocks.attr,
1088 : &dev_attr_events.attr,
1089 : &dev_attr_events_async.attr,
1090 : &dev_attr_events_poll_msecs.attr,
1091 : &dev_attr_diskseq.attr,
1092 : #ifdef CONFIG_FAIL_MAKE_REQUEST
1093 : &dev_attr_fail.attr,
1094 : #endif
1095 : #ifdef CONFIG_FAIL_IO_TIMEOUT
1096 : &dev_attr_fail_timeout.attr,
1097 : #endif
1098 : NULL
1099 : };
1100 :
1101 0 : static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1102 : {
1103 0 : struct device *dev = container_of(kobj, typeof(*dev), kobj);
1104 0 : struct gendisk *disk = dev_to_disk(dev);
1105 :
1106 0 : if (a == &dev_attr_badblocks.attr && !disk->bb)
1107 : return 0;
1108 0 : return a->mode;
1109 : }
1110 :
1111 : static struct attribute_group disk_attr_group = {
1112 : .attrs = disk_attrs,
1113 : .is_visible = disk_visible,
1114 : };
1115 :
1116 : static const struct attribute_group *disk_attr_groups[] = {
1117 : &disk_attr_group,
1118 : NULL
1119 : };
1120 :
1121 0 : static void disk_release_mq(struct request_queue *q)
1122 : {
1123 0 : blk_mq_cancel_work_sync(q);
1124 :
1125 : /*
1126 : * There can't be any non non-passthrough bios in flight here, but
1127 : * requests stay around longer, including passthrough ones so we
1128 : * still need to freeze the queue here.
1129 : */
1130 0 : blk_mq_freeze_queue(q);
1131 :
1132 : /*
1133 : * Since the I/O scheduler exit code may access cgroup information,
1134 : * perform I/O scheduler exit before disassociating from the block
1135 : * cgroup controller.
1136 : */
1137 0 : if (q->elevator) {
1138 0 : mutex_lock(&q->sysfs_lock);
1139 0 : elevator_exit(q);
1140 0 : mutex_unlock(&q->sysfs_lock);
1141 : }
1142 0 : rq_qos_exit(q);
1143 0 : __blk_mq_unfreeze_queue(q, true);
1144 0 : }
1145 :
1146 : /**
1147 : * disk_release - releases all allocated resources of the gendisk
1148 : * @dev: the device representing this disk
1149 : *
1150 : * This function releases all allocated resources of the gendisk.
1151 : *
1152 : * Drivers which used __device_add_disk() have a gendisk with a request_queue
1153 : * assigned. Since the request_queue sits on top of the gendisk for these
1154 : * drivers we also call blk_put_queue() for them, and we expect the
1155 : * request_queue refcount to reach 0 at this point, and so the request_queue
1156 : * will also be freed prior to the disk.
1157 : *
1158 : * Context: can sleep
1159 : */
1160 0 : static void disk_release(struct device *dev)
1161 : {
1162 0 : struct gendisk *disk = dev_to_disk(dev);
1163 :
1164 : might_sleep();
1165 0 : WARN_ON_ONCE(disk_live(disk));
1166 :
1167 0 : if (queue_is_mq(disk->queue))
1168 0 : disk_release_mq(disk->queue);
1169 :
1170 0 : blkcg_exit_queue(disk->queue);
1171 :
1172 0 : disk_release_events(disk);
1173 0 : kfree(disk->random);
1174 0 : xa_destroy(&disk->part_tbl);
1175 :
1176 0 : disk->queue->disk = NULL;
1177 0 : blk_put_queue(disk->queue);
1178 :
1179 0 : if (test_bit(GD_ADDED, &disk->state) && disk->fops->free_disk)
1180 0 : disk->fops->free_disk(disk);
1181 :
1182 0 : iput(disk->part0->bd_inode); /* frees the disk */
1183 0 : }
1184 :
1185 0 : static int block_uevent(struct device *dev, struct kobj_uevent_env *env)
1186 : {
1187 0 : struct gendisk *disk = dev_to_disk(dev);
1188 :
1189 0 : return add_uevent_var(env, "DISKSEQ=%llu", disk->diskseq);
1190 : }
1191 :
1192 : struct class block_class = {
1193 : .name = "block",
1194 : .dev_uevent = block_uevent,
1195 : };
1196 :
1197 0 : static char *block_devnode(struct device *dev, umode_t *mode,
1198 : kuid_t *uid, kgid_t *gid)
1199 : {
1200 0 : struct gendisk *disk = dev_to_disk(dev);
1201 :
1202 0 : if (disk->fops->devnode)
1203 0 : return disk->fops->devnode(disk, mode);
1204 : return NULL;
1205 : }
1206 :
1207 : const struct device_type disk_type = {
1208 : .name = "disk",
1209 : .groups = disk_attr_groups,
1210 : .release = disk_release,
1211 : .devnode = block_devnode,
1212 : };
1213 :
1214 : #ifdef CONFIG_PROC_FS
1215 : /*
1216 : * aggregate disk stat collector. Uses the same stats that the sysfs
1217 : * entries do, above, but makes them available through one seq_file.
1218 : *
1219 : * The output looks suspiciously like /proc/partitions with a bunch of
1220 : * extra fields.
1221 : */
1222 0 : static int diskstats_show(struct seq_file *seqf, void *v)
1223 : {
1224 0 : struct gendisk *gp = v;
1225 : struct block_device *hd;
1226 : unsigned int inflight;
1227 : struct disk_stats stat;
1228 : unsigned long idx;
1229 :
1230 : /*
1231 : if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1232 : seq_puts(seqf, "major minor name"
1233 : " rio rmerge rsect ruse wio wmerge "
1234 : "wsect wuse running use aveq"
1235 : "\n\n");
1236 : */
1237 :
1238 : rcu_read_lock();
1239 0 : xa_for_each(&gp->part_tbl, idx, hd) {
1240 0 : if (bdev_is_partition(hd) && !bdev_nr_sectors(hd))
1241 0 : continue;
1242 0 : if (queue_is_mq(gp->queue))
1243 0 : inflight = blk_mq_in_flight(gp->queue, hd);
1244 : else
1245 : inflight = part_in_flight(hd);
1246 :
1247 0 : if (inflight) {
1248 0 : part_stat_lock();
1249 0 : update_io_ticks(hd, jiffies, true);
1250 0 : part_stat_unlock();
1251 : }
1252 0 : part_stat_read_all(hd, &stat);
1253 0 : seq_printf(seqf, "%4d %7d %pg "
1254 : "%lu %lu %lu %u "
1255 : "%lu %lu %lu %u "
1256 : "%u %u %u "
1257 : "%lu %lu %lu %u "
1258 : "%lu %u"
1259 : "\n",
1260 0 : MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd,
1261 : stat.ios[STAT_READ],
1262 : stat.merges[STAT_READ],
1263 : stat.sectors[STAT_READ],
1264 0 : (unsigned int)div_u64(stat.nsecs[STAT_READ],
1265 : NSEC_PER_MSEC),
1266 : stat.ios[STAT_WRITE],
1267 : stat.merges[STAT_WRITE],
1268 : stat.sectors[STAT_WRITE],
1269 0 : (unsigned int)div_u64(stat.nsecs[STAT_WRITE],
1270 : NSEC_PER_MSEC),
1271 : inflight,
1272 : jiffies_to_msecs(stat.io_ticks),
1273 0 : (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1274 0 : stat.nsecs[STAT_WRITE] +
1275 : stat.nsecs[STAT_DISCARD] +
1276 : stat.nsecs[STAT_FLUSH],
1277 : NSEC_PER_MSEC),
1278 : stat.ios[STAT_DISCARD],
1279 : stat.merges[STAT_DISCARD],
1280 : stat.sectors[STAT_DISCARD],
1281 0 : (unsigned int)div_u64(stat.nsecs[STAT_DISCARD],
1282 : NSEC_PER_MSEC),
1283 : stat.ios[STAT_FLUSH],
1284 0 : (unsigned int)div_u64(stat.nsecs[STAT_FLUSH],
1285 : NSEC_PER_MSEC)
1286 : );
1287 : }
1288 : rcu_read_unlock();
1289 :
1290 0 : return 0;
1291 : }
1292 :
1293 : static const struct seq_operations diskstats_op = {
1294 : .start = disk_seqf_start,
1295 : .next = disk_seqf_next,
1296 : .stop = disk_seqf_stop,
1297 : .show = diskstats_show
1298 : };
1299 :
1300 1 : static int __init proc_genhd_init(void)
1301 : {
1302 1 : proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1303 1 : proc_create_seq("partitions", 0, NULL, &partitions_op);
1304 1 : return 0;
1305 : }
1306 : module_init(proc_genhd_init);
1307 : #endif /* CONFIG_PROC_FS */
1308 :
1309 0 : dev_t part_devt(struct gendisk *disk, u8 partno)
1310 : {
1311 : struct block_device *part;
1312 0 : dev_t devt = 0;
1313 :
1314 : rcu_read_lock();
1315 0 : part = xa_load(&disk->part_tbl, partno);
1316 0 : if (part)
1317 0 : devt = part->bd_dev;
1318 : rcu_read_unlock();
1319 :
1320 0 : return devt;
1321 : }
1322 :
1323 0 : dev_t blk_lookup_devt(const char *name, int partno)
1324 : {
1325 0 : dev_t devt = MKDEV(0, 0);
1326 : struct class_dev_iter iter;
1327 : struct device *dev;
1328 :
1329 0 : class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1330 0 : while ((dev = class_dev_iter_next(&iter))) {
1331 0 : struct gendisk *disk = dev_to_disk(dev);
1332 :
1333 0 : if (strcmp(dev_name(dev), name))
1334 0 : continue;
1335 :
1336 0 : if (partno < disk->minors) {
1337 : /* We need to return the right devno, even
1338 : * if the partition doesn't exist yet.
1339 : */
1340 0 : devt = MKDEV(MAJOR(dev->devt),
1341 : MINOR(dev->devt) + partno);
1342 : } else {
1343 0 : devt = part_devt(disk, partno);
1344 0 : if (devt)
1345 : break;
1346 : }
1347 : }
1348 0 : class_dev_iter_exit(&iter);
1349 0 : return devt;
1350 : }
1351 :
1352 0 : struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id,
1353 : struct lock_class_key *lkclass)
1354 : {
1355 : struct gendisk *disk;
1356 :
1357 0 : if (!blk_get_queue(q))
1358 : return NULL;
1359 :
1360 0 : disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1361 0 : if (!disk)
1362 : goto out_put_queue;
1363 :
1364 0 : disk->bdi = bdi_alloc(node_id);
1365 0 : if (!disk->bdi)
1366 : goto out_free_disk;
1367 :
1368 : /* bdev_alloc() might need the queue, set before the first call */
1369 0 : disk->queue = q;
1370 :
1371 0 : disk->part0 = bdev_alloc(disk, 0);
1372 0 : if (!disk->part0)
1373 : goto out_free_bdi;
1374 :
1375 0 : disk->node_id = node_id;
1376 0 : mutex_init(&disk->open_mutex);
1377 0 : xa_init(&disk->part_tbl);
1378 0 : if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL))
1379 : goto out_destroy_part_tbl;
1380 :
1381 0 : if (blkcg_init_queue(q))
1382 : goto out_erase_part0;
1383 :
1384 0 : rand_initialize_disk(disk);
1385 0 : disk_to_dev(disk)->class = &block_class;
1386 0 : disk_to_dev(disk)->type = &disk_type;
1387 0 : device_initialize(disk_to_dev(disk));
1388 0 : inc_diskseq(disk);
1389 0 : q->disk = disk;
1390 : lockdep_init_map(&disk->lockdep_map, "(bio completion)", lkclass, 0);
1391 : #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
1392 : INIT_LIST_HEAD(&disk->slave_bdevs);
1393 : #endif
1394 0 : return disk;
1395 :
1396 : out_erase_part0:
1397 : xa_erase(&disk->part_tbl, 0);
1398 : out_destroy_part_tbl:
1399 0 : xa_destroy(&disk->part_tbl);
1400 0 : disk->part0->bd_disk = NULL;
1401 0 : iput(disk->part0->bd_inode);
1402 : out_free_bdi:
1403 0 : bdi_put(disk->bdi);
1404 : out_free_disk:
1405 0 : kfree(disk);
1406 : out_put_queue:
1407 0 : blk_put_queue(q);
1408 0 : return NULL;
1409 : }
1410 : EXPORT_SYMBOL(__alloc_disk_node);
1411 :
1412 0 : struct gendisk *__blk_alloc_disk(int node, struct lock_class_key *lkclass)
1413 : {
1414 : struct request_queue *q;
1415 : struct gendisk *disk;
1416 :
1417 0 : q = blk_alloc_queue(node, false);
1418 0 : if (!q)
1419 : return NULL;
1420 :
1421 0 : disk = __alloc_disk_node(q, node, lkclass);
1422 0 : if (!disk) {
1423 0 : blk_cleanup_queue(q);
1424 0 : return NULL;
1425 : }
1426 : return disk;
1427 : }
1428 : EXPORT_SYMBOL(__blk_alloc_disk);
1429 :
1430 : /**
1431 : * put_disk - decrements the gendisk refcount
1432 : * @disk: the struct gendisk to decrement the refcount for
1433 : *
1434 : * This decrements the refcount for the struct gendisk. When this reaches 0
1435 : * we'll have disk_release() called.
1436 : *
1437 : * Context: Any context, but the last reference must not be dropped from
1438 : * atomic context.
1439 : */
1440 0 : void put_disk(struct gendisk *disk)
1441 : {
1442 0 : if (disk)
1443 0 : put_device(disk_to_dev(disk));
1444 0 : }
1445 : EXPORT_SYMBOL(put_disk);
1446 :
1447 : /**
1448 : * blk_cleanup_disk - shutdown a gendisk allocated by blk_alloc_disk
1449 : * @disk: gendisk to shutdown
1450 : *
1451 : * Mark the queue hanging off @disk DYING, drain all pending requests, then mark
1452 : * the queue DEAD, destroy and put it and the gendisk structure.
1453 : *
1454 : * Context: can sleep
1455 : */
1456 0 : void blk_cleanup_disk(struct gendisk *disk)
1457 : {
1458 0 : blk_cleanup_queue(disk->queue);
1459 0 : put_disk(disk);
1460 0 : }
1461 : EXPORT_SYMBOL(blk_cleanup_disk);
1462 :
1463 0 : static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1464 : {
1465 0 : char event[] = "DISK_RO=1";
1466 0 : char *envp[] = { event, NULL };
1467 :
1468 0 : if (!ro)
1469 0 : event[8] = '0';
1470 0 : kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1471 0 : }
1472 :
1473 : /**
1474 : * set_disk_ro - set a gendisk read-only
1475 : * @disk: gendisk to operate on
1476 : * @read_only: %true to set the disk read-only, %false set the disk read/write
1477 : *
1478 : * This function is used to indicate whether a given disk device should have its
1479 : * read-only flag set. set_disk_ro() is typically used by device drivers to
1480 : * indicate whether the underlying physical device is write-protected.
1481 : */
1482 0 : void set_disk_ro(struct gendisk *disk, bool read_only)
1483 : {
1484 0 : if (read_only) {
1485 0 : if (test_and_set_bit(GD_READ_ONLY, &disk->state))
1486 : return;
1487 : } else {
1488 0 : if (!test_and_clear_bit(GD_READ_ONLY, &disk->state))
1489 : return;
1490 : }
1491 0 : set_disk_ro_uevent(disk, read_only);
1492 : }
1493 : EXPORT_SYMBOL(set_disk_ro);
1494 :
1495 0 : void inc_diskseq(struct gendisk *disk)
1496 : {
1497 0 : disk->diskseq = atomic64_inc_return(&diskseq);
1498 0 : }
|
