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 0 : static void part_stat_read_all(struct block_device *part,
105 : struct disk_stats *stat)
106 : {
107 : int cpu;
108 :
109 0 : memset(stat, 0, sizeof(struct disk_stats));
110 0 : for_each_possible_cpu(cpu) {
111 0 : struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu);
112 : int group;
113 :
114 0 : for (group = 0; group < NR_STAT_GROUPS; group++) {
115 0 : stat->nsecs[group] += ptr->nsecs[group];
116 0 : stat->sectors[group] += ptr->sectors[group];
117 0 : stat->ios[group] += ptr->ios[group];
118 0 : stat->merges[group] += ptr->merges[group];
119 : }
120 :
121 0 : stat->io_ticks += ptr->io_ticks;
122 : }
123 0 : }
124 :
125 : static unsigned int part_in_flight(struct block_device *part)
126 : {
127 : unsigned int inflight = 0;
128 : int cpu;
129 :
130 0 : for_each_possible_cpu(cpu) {
131 0 : inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) +
132 0 : part_stat_local_read_cpu(part, in_flight[1], cpu);
133 : }
134 0 : if ((int)inflight < 0)
135 0 : inflight = 0;
136 :
137 : return inflight;
138 : }
139 :
140 : static void part_in_flight_rw(struct block_device *part,
141 : unsigned int inflight[2])
142 : {
143 : int cpu;
144 :
145 0 : inflight[0] = 0;
146 0 : inflight[1] = 0;
147 0 : for_each_possible_cpu(cpu) {
148 0 : inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu);
149 0 : inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu);
150 : }
151 0 : if ((int)inflight[0] < 0)
152 0 : inflight[0] = 0;
153 0 : if ((int)inflight[1] < 0)
154 0 : inflight[1] = 0;
155 : }
156 :
157 : /*
158 : * Can be deleted altogether. Later.
159 : *
160 : */
161 : #define BLKDEV_MAJOR_HASH_SIZE 255
162 : static struct blk_major_name {
163 : struct blk_major_name *next;
164 : int major;
165 : char name[16];
166 : #ifdef CONFIG_BLOCK_LEGACY_AUTOLOAD
167 : void (*probe)(dev_t devt);
168 : #endif
169 : } *major_names[BLKDEV_MAJOR_HASH_SIZE];
170 : static DEFINE_MUTEX(major_names_lock);
171 : static DEFINE_SPINLOCK(major_names_spinlock);
172 :
173 : /* index in the above - for now: assume no multimajor ranges */
174 : static inline int major_to_index(unsigned major)
175 : {
176 1 : return major % BLKDEV_MAJOR_HASH_SIZE;
177 : }
178 :
179 : #ifdef CONFIG_PROC_FS
180 0 : void blkdev_show(struct seq_file *seqf, off_t offset)
181 : {
182 : struct blk_major_name *dp;
183 :
184 0 : spin_lock(&major_names_spinlock);
185 0 : for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
186 0 : if (dp->major == offset)
187 0 : seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
188 0 : spin_unlock(&major_names_spinlock);
189 0 : }
190 : #endif /* CONFIG_PROC_FS */
191 :
192 : /**
193 : * __register_blkdev - register a new block device
194 : *
195 : * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
196 : * @major = 0, try to allocate any unused major number.
197 : * @name: the name of the new block device as a zero terminated string
198 : * @probe: pre-devtmpfs / pre-udev callback used to create disks when their
199 : * pre-created device node is accessed. When a probe call uses
200 : * add_disk() and it fails the driver must cleanup resources. This
201 : * interface may soon be removed.
202 : *
203 : * The @name must be unique within the system.
204 : *
205 : * The return value depends on the @major input parameter:
206 : *
207 : * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
208 : * then the function returns zero on success, or a negative error code
209 : * - if any unused major number was requested with @major = 0 parameter
210 : * then the return value is the allocated major number in range
211 : * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
212 : *
213 : * See Documentation/admin-guide/devices.txt for the list of allocated
214 : * major numbers.
215 : *
216 : * Use register_blkdev instead for any new code.
217 : */
218 1 : int __register_blkdev(unsigned int major, const char *name,
219 : void (*probe)(dev_t devt))
220 : {
221 : struct blk_major_name **n, *p;
222 1 : int index, ret = 0;
223 :
224 1 : mutex_lock(&major_names_lock);
225 :
226 : /* temporary */
227 1 : if (major == 0) {
228 0 : for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
229 0 : if (major_names[index] == NULL)
230 : break;
231 : }
232 :
233 0 : if (index == 0) {
234 0 : printk("%s: failed to get major for %s\n",
235 : __func__, name);
236 0 : ret = -EBUSY;
237 0 : goto out;
238 : }
239 0 : major = index;
240 0 : ret = major;
241 : }
242 :
243 1 : if (major >= BLKDEV_MAJOR_MAX) {
244 0 : pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
245 : __func__, major, BLKDEV_MAJOR_MAX-1, name);
246 :
247 0 : ret = -EINVAL;
248 0 : goto out;
249 : }
250 :
251 1 : p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
252 1 : if (p == NULL) {
253 : ret = -ENOMEM;
254 : goto out;
255 : }
256 :
257 1 : p->major = major;
258 : #ifdef CONFIG_BLOCK_LEGACY_AUTOLOAD
259 1 : p->probe = probe;
260 : #endif
261 1 : strlcpy(p->name, name, sizeof(p->name));
262 1 : p->next = NULL;
263 1 : index = major_to_index(major);
264 :
265 1 : spin_lock(&major_names_spinlock);
266 1 : for (n = &major_names[index]; *n; n = &(*n)->next) {
267 0 : if ((*n)->major == major)
268 : break;
269 : }
270 1 : if (!*n)
271 1 : *n = p;
272 : else
273 : ret = -EBUSY;
274 1 : spin_unlock(&major_names_spinlock);
275 :
276 1 : if (ret < 0) {
277 0 : printk("register_blkdev: cannot get major %u for %s\n",
278 : major, name);
279 0 : kfree(p);
280 : }
281 : out:
282 1 : mutex_unlock(&major_names_lock);
283 1 : return ret;
284 : }
285 : EXPORT_SYMBOL(__register_blkdev);
286 :
287 0 : void unregister_blkdev(unsigned int major, const char *name)
288 : {
289 : struct blk_major_name **n;
290 0 : struct blk_major_name *p = NULL;
291 0 : int index = major_to_index(major);
292 :
293 0 : mutex_lock(&major_names_lock);
294 0 : spin_lock(&major_names_spinlock);
295 0 : for (n = &major_names[index]; *n; n = &(*n)->next)
296 0 : if ((*n)->major == major)
297 : break;
298 0 : if (!*n || strcmp((*n)->name, name)) {
299 0 : WARN_ON(1);
300 : } else {
301 0 : p = *n;
302 0 : *n = p->next;
303 : }
304 0 : spin_unlock(&major_names_spinlock);
305 0 : mutex_unlock(&major_names_lock);
306 0 : kfree(p);
307 0 : }
308 :
309 : EXPORT_SYMBOL(unregister_blkdev);
310 :
311 0 : int blk_alloc_ext_minor(void)
312 : {
313 : int idx;
314 :
315 0 : idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT - 1, GFP_KERNEL);
316 0 : if (idx == -ENOSPC)
317 : return -EBUSY;
318 0 : return idx;
319 : }
320 :
321 0 : void blk_free_ext_minor(unsigned int minor)
322 : {
323 0 : ida_free(&ext_devt_ida, minor);
324 0 : }
325 :
326 0 : static char *bdevt_str(dev_t devt, char *buf)
327 : {
328 0 : if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
329 : char tbuf[BDEVT_SIZE];
330 0 : snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
331 0 : snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
332 : } else
333 0 : snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
334 :
335 0 : return buf;
336 : }
337 :
338 0 : void disk_uevent(struct gendisk *disk, enum kobject_action action)
339 : {
340 : struct block_device *part;
341 : unsigned long idx;
342 :
343 : rcu_read_lock();
344 0 : xa_for_each(&disk->part_tbl, idx, part) {
345 0 : if (bdev_is_partition(part) && !bdev_nr_sectors(part))
346 0 : continue;
347 0 : if (!kobject_get_unless_zero(&part->bd_device.kobj))
348 0 : continue;
349 :
350 : rcu_read_unlock();
351 0 : kobject_uevent(bdev_kobj(part), action);
352 0 : put_device(&part->bd_device);
353 : rcu_read_lock();
354 : }
355 : rcu_read_unlock();
356 0 : }
357 : EXPORT_SYMBOL_GPL(disk_uevent);
358 :
359 0 : int disk_scan_partitions(struct gendisk *disk, fmode_t mode)
360 : {
361 : struct block_device *bdev;
362 0 : int ret = 0;
363 :
364 0 : if (disk->flags & (GENHD_FL_NO_PART | GENHD_FL_HIDDEN))
365 : return -EINVAL;
366 0 : if (test_bit(GD_SUPPRESS_PART_SCAN, &disk->state))
367 : return -EINVAL;
368 0 : if (disk->open_partitions)
369 : return -EBUSY;
370 :
371 0 : set_bit(GD_NEED_PART_SCAN, &disk->state);
372 : /*
373 : * If the device is opened exclusively by current thread already, it's
374 : * safe to scan partitons, otherwise, use bd_prepare_to_claim() to
375 : * synchronize with other exclusive openers and other partition
376 : * scanners.
377 : */
378 0 : if (!(mode & FMODE_EXCL)) {
379 0 : ret = bd_prepare_to_claim(disk->part0, disk_scan_partitions);
380 0 : if (ret)
381 : return ret;
382 : }
383 :
384 0 : bdev = blkdev_get_by_dev(disk_devt(disk), mode & ~FMODE_EXCL, NULL);
385 0 : if (IS_ERR(bdev))
386 0 : ret = PTR_ERR(bdev);
387 : else
388 0 : blkdev_put(bdev, mode);
389 :
390 0 : if (!(mode & FMODE_EXCL))
391 0 : bd_abort_claiming(disk->part0, disk_scan_partitions);
392 : return ret;
393 : }
394 :
395 : /**
396 : * device_add_disk - add disk information to kernel list
397 : * @parent: parent device for the disk
398 : * @disk: per-device partitioning information
399 : * @groups: Additional per-device sysfs groups
400 : *
401 : * This function registers the partitioning information in @disk
402 : * with the kernel.
403 : */
404 0 : int __must_check device_add_disk(struct device *parent, struct gendisk *disk,
405 : const struct attribute_group **groups)
406 :
407 : {
408 0 : struct device *ddev = disk_to_dev(disk);
409 : int ret;
410 :
411 : /* Only makes sense for bio-based to set ->poll_bio */
412 0 : if (queue_is_mq(disk->queue) && disk->fops->poll_bio)
413 : return -EINVAL;
414 :
415 : /*
416 : * The disk queue should now be all set with enough information about
417 : * the device for the elevator code to pick an adequate default
418 : * elevator if one is needed, that is, for devices requesting queue
419 : * registration.
420 : */
421 0 : elevator_init_mq(disk->queue);
422 :
423 : /*
424 : * If the driver provides an explicit major number it also must provide
425 : * the number of minors numbers supported, and those will be used to
426 : * setup the gendisk.
427 : * Otherwise just allocate the device numbers for both the whole device
428 : * and all partitions from the extended dev_t space.
429 : */
430 0 : ret = -EINVAL;
431 0 : if (disk->major) {
432 0 : if (WARN_ON(!disk->minors))
433 : goto out_exit_elevator;
434 :
435 0 : if (disk->minors > DISK_MAX_PARTS) {
436 0 : pr_err("block: can't allocate more than %d partitions\n",
437 : DISK_MAX_PARTS);
438 0 : disk->minors = DISK_MAX_PARTS;
439 : }
440 0 : if (disk->first_minor + disk->minors > MINORMASK + 1)
441 : goto out_exit_elevator;
442 : } else {
443 0 : if (WARN_ON(disk->minors))
444 : goto out_exit_elevator;
445 :
446 0 : ret = blk_alloc_ext_minor();
447 0 : if (ret < 0)
448 : goto out_exit_elevator;
449 0 : disk->major = BLOCK_EXT_MAJOR;
450 0 : disk->first_minor = ret;
451 : }
452 :
453 : /* delay uevents, until we scanned partition table */
454 0 : dev_set_uevent_suppress(ddev, 1);
455 :
456 0 : ddev->parent = parent;
457 0 : ddev->groups = groups;
458 0 : dev_set_name(ddev, "%s", disk->disk_name);
459 0 : if (!(disk->flags & GENHD_FL_HIDDEN))
460 0 : ddev->devt = MKDEV(disk->major, disk->first_minor);
461 0 : ret = device_add(ddev);
462 0 : if (ret)
463 : goto out_free_ext_minor;
464 :
465 0 : ret = disk_alloc_events(disk);
466 0 : if (ret)
467 : goto out_device_del;
468 :
469 : if (!sysfs_deprecated) {
470 0 : ret = sysfs_create_link(block_depr, &ddev->kobj,
471 0 : kobject_name(&ddev->kobj));
472 0 : if (ret)
473 : goto out_device_del;
474 : }
475 :
476 : /*
477 : * avoid probable deadlock caused by allocating memory with
478 : * GFP_KERNEL in runtime_resume callback of its all ancestor
479 : * devices
480 : */
481 0 : pm_runtime_set_memalloc_noio(ddev, true);
482 :
483 0 : ret = blk_integrity_add(disk);
484 : if (ret)
485 : goto out_del_block_link;
486 :
487 0 : disk->part0->bd_holder_dir =
488 0 : kobject_create_and_add("holders", &ddev->kobj);
489 0 : if (!disk->part0->bd_holder_dir) {
490 : ret = -ENOMEM;
491 : goto out_del_integrity;
492 : }
493 0 : disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
494 0 : if (!disk->slave_dir) {
495 : ret = -ENOMEM;
496 : goto out_put_holder_dir;
497 : }
498 :
499 0 : ret = blk_register_queue(disk);
500 0 : if (ret)
501 : goto out_put_slave_dir;
502 :
503 0 : if (!(disk->flags & GENHD_FL_HIDDEN)) {
504 0 : ret = bdi_register(disk->bdi, "%u:%u",
505 : disk->major, disk->first_minor);
506 0 : if (ret)
507 : goto out_unregister_queue;
508 0 : bdi_set_owner(disk->bdi, ddev);
509 0 : ret = sysfs_create_link(&ddev->kobj,
510 0 : &disk->bdi->dev->kobj, "bdi");
511 0 : if (ret)
512 : goto out_unregister_bdi;
513 :
514 : /* Make sure the first partition scan will be proceed */
515 0 : if (get_capacity(disk) && !(disk->flags & GENHD_FL_NO_PART) &&
516 0 : !test_bit(GD_SUPPRESS_PART_SCAN, &disk->state))
517 0 : set_bit(GD_NEED_PART_SCAN, &disk->state);
518 :
519 0 : bdev_add(disk->part0, ddev->devt);
520 0 : if (get_capacity(disk))
521 0 : disk_scan_partitions(disk, FMODE_READ);
522 :
523 : /*
524 : * Announce the disk and partitions after all partitions are
525 : * created. (for hidden disks uevents remain suppressed forever)
526 : */
527 0 : dev_set_uevent_suppress(ddev, 0);
528 0 : disk_uevent(disk, KOBJ_ADD);
529 : } else {
530 : /*
531 : * Even if the block_device for a hidden gendisk is not
532 : * registered, it needs to have a valid bd_dev so that the
533 : * freeing of the dynamic major works.
534 : */
535 0 : disk->part0->bd_dev = MKDEV(disk->major, disk->first_minor);
536 : }
537 :
538 0 : disk_update_readahead(disk);
539 0 : disk_add_events(disk);
540 0 : set_bit(GD_ADDED, &disk->state);
541 0 : return 0;
542 :
543 : out_unregister_bdi:
544 0 : if (!(disk->flags & GENHD_FL_HIDDEN))
545 0 : bdi_unregister(disk->bdi);
546 : out_unregister_queue:
547 0 : blk_unregister_queue(disk);
548 0 : rq_qos_exit(disk->queue);
549 : out_put_slave_dir:
550 0 : kobject_put(disk->slave_dir);
551 0 : disk->slave_dir = NULL;
552 : out_put_holder_dir:
553 0 : kobject_put(disk->part0->bd_holder_dir);
554 : out_del_integrity:
555 0 : blk_integrity_del(disk);
556 : out_del_block_link:
557 : if (!sysfs_deprecated)
558 0 : sysfs_remove_link(block_depr, dev_name(ddev));
559 : out_device_del:
560 0 : device_del(ddev);
561 : out_free_ext_minor:
562 0 : if (disk->major == BLOCK_EXT_MAJOR)
563 0 : blk_free_ext_minor(disk->first_minor);
564 : out_exit_elevator:
565 0 : if (disk->queue->elevator)
566 0 : elevator_exit(disk->queue);
567 : return ret;
568 : }
569 : EXPORT_SYMBOL(device_add_disk);
570 :
571 : /**
572 : * blk_mark_disk_dead - mark a disk as dead
573 : * @disk: disk to mark as dead
574 : *
575 : * Mark as disk as dead (e.g. surprise removed) and don't accept any new I/O
576 : * to this disk.
577 : */
578 0 : void blk_mark_disk_dead(struct gendisk *disk)
579 : {
580 0 : set_bit(GD_DEAD, &disk->state);
581 0 : blk_queue_start_drain(disk->queue);
582 :
583 : /*
584 : * Stop buffered writers from dirtying pages that can't be written out.
585 : */
586 0 : set_capacity_and_notify(disk, 0);
587 0 : }
588 : EXPORT_SYMBOL_GPL(blk_mark_disk_dead);
589 :
590 : /**
591 : * del_gendisk - remove the gendisk
592 : * @disk: the struct gendisk to remove
593 : *
594 : * Removes the gendisk and all its associated resources. This deletes the
595 : * partitions associated with the gendisk, and unregisters the associated
596 : * request_queue.
597 : *
598 : * This is the counter to the respective __device_add_disk() call.
599 : *
600 : * The final removal of the struct gendisk happens when its refcount reaches 0
601 : * with put_disk(), which should be called after del_gendisk(), if
602 : * __device_add_disk() was used.
603 : *
604 : * Drivers exist which depend on the release of the gendisk to be synchronous,
605 : * it should not be deferred.
606 : *
607 : * Context: can sleep
608 : */
609 0 : void del_gendisk(struct gendisk *disk)
610 : {
611 0 : struct request_queue *q = disk->queue;
612 :
613 : might_sleep();
614 :
615 0 : if (WARN_ON_ONCE(!disk_live(disk) && !(disk->flags & GENHD_FL_HIDDEN)))
616 : return;
617 :
618 0 : blk_integrity_del(disk);
619 0 : disk_del_events(disk);
620 :
621 0 : mutex_lock(&disk->open_mutex);
622 0 : remove_inode_hash(disk->part0->bd_inode);
623 0 : blk_drop_partitions(disk);
624 0 : mutex_unlock(&disk->open_mutex);
625 :
626 0 : fsync_bdev(disk->part0);
627 0 : __invalidate_device(disk->part0, true);
628 :
629 : /*
630 : * Fail any new I/O.
631 : */
632 0 : set_bit(GD_DEAD, &disk->state);
633 0 : if (test_bit(GD_OWNS_QUEUE, &disk->state))
634 0 : blk_queue_flag_set(QUEUE_FLAG_DYING, q);
635 0 : set_capacity(disk, 0);
636 :
637 : /*
638 : * Prevent new I/O from crossing bio_queue_enter().
639 : */
640 0 : blk_queue_start_drain(q);
641 :
642 0 : if (!(disk->flags & GENHD_FL_HIDDEN)) {
643 0 : sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
644 :
645 : /*
646 : * Unregister bdi before releasing device numbers (as they can
647 : * get reused and we'd get clashes in sysfs).
648 : */
649 0 : bdi_unregister(disk->bdi);
650 : }
651 :
652 0 : blk_unregister_queue(disk);
653 :
654 0 : kobject_put(disk->part0->bd_holder_dir);
655 0 : kobject_put(disk->slave_dir);
656 0 : disk->slave_dir = NULL;
657 :
658 0 : part_stat_set_all(disk->part0, 0);
659 0 : disk->part0->bd_stamp = 0;
660 : if (!sysfs_deprecated)
661 0 : sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
662 0 : pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
663 0 : device_del(disk_to_dev(disk));
664 :
665 0 : blk_mq_freeze_queue_wait(q);
666 :
667 0 : blk_throtl_cancel_bios(disk);
668 :
669 0 : blk_sync_queue(q);
670 : blk_flush_integrity();
671 :
672 0 : if (queue_is_mq(q))
673 0 : blk_mq_cancel_work_sync(q);
674 :
675 0 : blk_mq_quiesce_queue(q);
676 0 : if (q->elevator) {
677 0 : mutex_lock(&q->sysfs_lock);
678 0 : elevator_exit(q);
679 0 : mutex_unlock(&q->sysfs_lock);
680 : }
681 0 : rq_qos_exit(q);
682 0 : blk_mq_unquiesce_queue(q);
683 :
684 : /*
685 : * If the disk does not own the queue, allow using passthrough requests
686 : * again. Else leave the queue frozen to fail all I/O.
687 : */
688 0 : if (!test_bit(GD_OWNS_QUEUE, &disk->state)) {
689 0 : blk_queue_flag_clear(QUEUE_FLAG_INIT_DONE, q);
690 0 : __blk_mq_unfreeze_queue(q, true);
691 : } else {
692 0 : if (queue_is_mq(q))
693 0 : blk_mq_exit_queue(q);
694 : }
695 : }
696 : EXPORT_SYMBOL(del_gendisk);
697 :
698 : /**
699 : * invalidate_disk - invalidate the disk
700 : * @disk: the struct gendisk to invalidate
701 : *
702 : * A helper to invalidates the disk. It will clean the disk's associated
703 : * buffer/page caches and reset its internal states so that the disk
704 : * can be reused by the drivers.
705 : *
706 : * Context: can sleep
707 : */
708 0 : void invalidate_disk(struct gendisk *disk)
709 : {
710 0 : struct block_device *bdev = disk->part0;
711 :
712 0 : invalidate_bdev(bdev);
713 0 : bdev->bd_inode->i_mapping->wb_err = 0;
714 0 : set_capacity(disk, 0);
715 0 : }
716 : EXPORT_SYMBOL(invalidate_disk);
717 :
718 : /* sysfs access to bad-blocks list. */
719 0 : static ssize_t disk_badblocks_show(struct device *dev,
720 : struct device_attribute *attr,
721 : char *page)
722 : {
723 0 : struct gendisk *disk = dev_to_disk(dev);
724 :
725 0 : if (!disk->bb)
726 0 : return sprintf(page, "\n");
727 :
728 0 : return badblocks_show(disk->bb, page, 0);
729 : }
730 :
731 0 : static ssize_t disk_badblocks_store(struct device *dev,
732 : struct device_attribute *attr,
733 : const char *page, size_t len)
734 : {
735 0 : struct gendisk *disk = dev_to_disk(dev);
736 :
737 0 : if (!disk->bb)
738 : return -ENXIO;
739 :
740 0 : return badblocks_store(disk->bb, page, len, 0);
741 : }
742 :
743 : #ifdef CONFIG_BLOCK_LEGACY_AUTOLOAD
744 0 : void blk_request_module(dev_t devt)
745 : {
746 0 : unsigned int major = MAJOR(devt);
747 : struct blk_major_name **n;
748 :
749 0 : mutex_lock(&major_names_lock);
750 0 : for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) {
751 0 : if ((*n)->major == major && (*n)->probe) {
752 0 : (*n)->probe(devt);
753 0 : mutex_unlock(&major_names_lock);
754 0 : return;
755 : }
756 : }
757 0 : mutex_unlock(&major_names_lock);
758 :
759 0 : if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
760 : /* Make old-style 2.4 aliases work */
761 : request_module("block-major-%d", MAJOR(devt));
762 : }
763 : #endif /* CONFIG_BLOCK_LEGACY_AUTOLOAD */
764 :
765 : /*
766 : * print a full list of all partitions - intended for places where the root
767 : * filesystem can't be mounted and thus to give the victim some idea of what
768 : * went wrong
769 : */
770 0 : void __init printk_all_partitions(void)
771 : {
772 : struct class_dev_iter iter;
773 : struct device *dev;
774 :
775 0 : class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
776 0 : while ((dev = class_dev_iter_next(&iter))) {
777 0 : struct gendisk *disk = dev_to_disk(dev);
778 : struct block_device *part;
779 : char devt_buf[BDEVT_SIZE];
780 : unsigned long idx;
781 :
782 : /*
783 : * Don't show empty devices or things that have been
784 : * suppressed
785 : */
786 0 : if (get_capacity(disk) == 0 || (disk->flags & GENHD_FL_HIDDEN))
787 0 : continue;
788 :
789 : /*
790 : * Note, unlike /proc/partitions, I am showing the numbers in
791 : * hex - the same format as the root= option takes.
792 : */
793 : rcu_read_lock();
794 0 : xa_for_each(&disk->part_tbl, idx, part) {
795 0 : if (!bdev_nr_sectors(part))
796 0 : continue;
797 0 : printk("%s%s %10llu %pg %s",
798 : bdev_is_partition(part) ? " " : "",
799 : bdevt_str(part->bd_dev, devt_buf),
800 : bdev_nr_sectors(part) >> 1, part,
801 : part->bd_meta_info ?
802 : part->bd_meta_info->uuid : "");
803 0 : if (bdev_is_partition(part))
804 0 : printk("\n");
805 0 : else if (dev->parent && dev->parent->driver)
806 0 : printk(" driver: %s\n",
807 : dev->parent->driver->name);
808 : else
809 0 : printk(" (driver?)\n");
810 : }
811 : rcu_read_unlock();
812 : }
813 0 : class_dev_iter_exit(&iter);
814 0 : }
815 :
816 : #ifdef CONFIG_PROC_FS
817 : /* iterator */
818 0 : static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
819 : {
820 0 : loff_t skip = *pos;
821 : struct class_dev_iter *iter;
822 : struct device *dev;
823 :
824 0 : iter = kmalloc(sizeof(*iter), GFP_KERNEL);
825 0 : if (!iter)
826 : return ERR_PTR(-ENOMEM);
827 :
828 0 : seqf->private = iter;
829 0 : class_dev_iter_init(iter, &block_class, NULL, &disk_type);
830 : do {
831 0 : dev = class_dev_iter_next(iter);
832 0 : if (!dev)
833 : return NULL;
834 0 : } while (skip--);
835 :
836 0 : return dev_to_disk(dev);
837 : }
838 :
839 0 : static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
840 : {
841 : struct device *dev;
842 :
843 0 : (*pos)++;
844 0 : dev = class_dev_iter_next(seqf->private);
845 0 : if (dev)
846 0 : return dev_to_disk(dev);
847 :
848 : return NULL;
849 : }
850 :
851 0 : static void disk_seqf_stop(struct seq_file *seqf, void *v)
852 : {
853 0 : struct class_dev_iter *iter = seqf->private;
854 :
855 : /* stop is called even after start failed :-( */
856 0 : if (iter) {
857 0 : class_dev_iter_exit(iter);
858 0 : kfree(iter);
859 0 : seqf->private = NULL;
860 : }
861 0 : }
862 :
863 0 : static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
864 : {
865 : void *p;
866 :
867 0 : p = disk_seqf_start(seqf, pos);
868 0 : if (!IS_ERR_OR_NULL(p) && !*pos)
869 0 : seq_puts(seqf, "major minor #blocks name\n\n");
870 0 : return p;
871 : }
872 :
873 0 : static int show_partition(struct seq_file *seqf, void *v)
874 : {
875 0 : struct gendisk *sgp = v;
876 : struct block_device *part;
877 : unsigned long idx;
878 :
879 0 : if (!get_capacity(sgp) || (sgp->flags & GENHD_FL_HIDDEN))
880 : return 0;
881 :
882 : rcu_read_lock();
883 0 : xa_for_each(&sgp->part_tbl, idx, part) {
884 0 : if (!bdev_nr_sectors(part))
885 0 : continue;
886 0 : seq_printf(seqf, "%4d %7d %10llu %pg\n",
887 0 : MAJOR(part->bd_dev), MINOR(part->bd_dev),
888 0 : bdev_nr_sectors(part) >> 1, part);
889 : }
890 : rcu_read_unlock();
891 0 : return 0;
892 : }
893 :
894 : static const struct seq_operations partitions_op = {
895 : .start = show_partition_start,
896 : .next = disk_seqf_next,
897 : .stop = disk_seqf_stop,
898 : .show = show_partition
899 : };
900 : #endif
901 :
902 1 : static int __init genhd_device_init(void)
903 : {
904 : int error;
905 :
906 1 : block_class.dev_kobj = sysfs_dev_block_kobj;
907 1 : error = class_register(&block_class);
908 1 : if (unlikely(error))
909 : return error;
910 1 : blk_dev_init();
911 :
912 1 : register_blkdev(BLOCK_EXT_MAJOR, "blkext");
913 :
914 : /* create top-level block dir */
915 : if (!sysfs_deprecated)
916 1 : block_depr = kobject_create_and_add("block", NULL);
917 1 : return 0;
918 : }
919 :
920 : subsys_initcall(genhd_device_init);
921 :
922 0 : static ssize_t disk_range_show(struct device *dev,
923 : struct device_attribute *attr, char *buf)
924 : {
925 0 : struct gendisk *disk = dev_to_disk(dev);
926 :
927 0 : return sprintf(buf, "%d\n", disk->minors);
928 : }
929 :
930 0 : static ssize_t disk_ext_range_show(struct device *dev,
931 : struct device_attribute *attr, char *buf)
932 : {
933 0 : struct gendisk *disk = dev_to_disk(dev);
934 :
935 0 : return sprintf(buf, "%d\n",
936 0 : (disk->flags & GENHD_FL_NO_PART) ? 1 : DISK_MAX_PARTS);
937 : }
938 :
939 0 : static ssize_t disk_removable_show(struct device *dev,
940 : struct device_attribute *attr, char *buf)
941 : {
942 0 : struct gendisk *disk = dev_to_disk(dev);
943 :
944 0 : return sprintf(buf, "%d\n",
945 0 : (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
946 : }
947 :
948 0 : static ssize_t disk_hidden_show(struct device *dev,
949 : struct device_attribute *attr, char *buf)
950 : {
951 0 : struct gendisk *disk = dev_to_disk(dev);
952 :
953 0 : return sprintf(buf, "%d\n",
954 0 : (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
955 : }
956 :
957 0 : static ssize_t disk_ro_show(struct device *dev,
958 : struct device_attribute *attr, char *buf)
959 : {
960 0 : struct gendisk *disk = dev_to_disk(dev);
961 :
962 0 : return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
963 : }
964 :
965 0 : ssize_t part_size_show(struct device *dev,
966 : struct device_attribute *attr, char *buf)
967 : {
968 0 : return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev)));
969 : }
970 :
971 0 : ssize_t part_stat_show(struct device *dev,
972 : struct device_attribute *attr, char *buf)
973 : {
974 0 : struct block_device *bdev = dev_to_bdev(dev);
975 0 : struct request_queue *q = bdev_get_queue(bdev);
976 : struct disk_stats stat;
977 : unsigned int inflight;
978 :
979 0 : if (queue_is_mq(q))
980 0 : inflight = blk_mq_in_flight(q, bdev);
981 : else
982 : inflight = part_in_flight(bdev);
983 :
984 0 : if (inflight) {
985 0 : part_stat_lock();
986 0 : update_io_ticks(bdev, jiffies, true);
987 0 : part_stat_unlock();
988 : }
989 0 : part_stat_read_all(bdev, &stat);
990 0 : return sprintf(buf,
991 : "%8lu %8lu %8llu %8u "
992 : "%8lu %8lu %8llu %8u "
993 : "%8u %8u %8u "
994 : "%8lu %8lu %8llu %8u "
995 : "%8lu %8u"
996 : "\n",
997 : stat.ios[STAT_READ],
998 : stat.merges[STAT_READ],
999 0 : (unsigned long long)stat.sectors[STAT_READ],
1000 0 : (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC),
1001 : stat.ios[STAT_WRITE],
1002 : stat.merges[STAT_WRITE],
1003 0 : (unsigned long long)stat.sectors[STAT_WRITE],
1004 0 : (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC),
1005 : inflight,
1006 : jiffies_to_msecs(stat.io_ticks),
1007 0 : (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1008 0 : stat.nsecs[STAT_WRITE] +
1009 : stat.nsecs[STAT_DISCARD] +
1010 : stat.nsecs[STAT_FLUSH],
1011 : NSEC_PER_MSEC),
1012 : stat.ios[STAT_DISCARD],
1013 : stat.merges[STAT_DISCARD],
1014 0 : (unsigned long long)stat.sectors[STAT_DISCARD],
1015 0 : (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC),
1016 : stat.ios[STAT_FLUSH],
1017 0 : (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC));
1018 : }
1019 :
1020 0 : ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
1021 : char *buf)
1022 : {
1023 0 : struct block_device *bdev = dev_to_bdev(dev);
1024 0 : struct request_queue *q = bdev_get_queue(bdev);
1025 : unsigned int inflight[2];
1026 :
1027 0 : if (queue_is_mq(q))
1028 0 : blk_mq_in_flight_rw(q, bdev, inflight);
1029 : else
1030 0 : part_in_flight_rw(bdev, inflight);
1031 :
1032 0 : return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
1033 : }
1034 :
1035 0 : static ssize_t disk_capability_show(struct device *dev,
1036 : struct device_attribute *attr, char *buf)
1037 : {
1038 0 : dev_warn_once(dev, "the capability attribute has been deprecated.\n");
1039 0 : return sprintf(buf, "0\n");
1040 : }
1041 :
1042 0 : static ssize_t disk_alignment_offset_show(struct device *dev,
1043 : struct device_attribute *attr,
1044 : char *buf)
1045 : {
1046 0 : struct gendisk *disk = dev_to_disk(dev);
1047 :
1048 0 : return sprintf(buf, "%d\n", bdev_alignment_offset(disk->part0));
1049 : }
1050 :
1051 0 : static ssize_t disk_discard_alignment_show(struct device *dev,
1052 : struct device_attribute *attr,
1053 : char *buf)
1054 : {
1055 0 : struct gendisk *disk = dev_to_disk(dev);
1056 :
1057 0 : return sprintf(buf, "%d\n", bdev_alignment_offset(disk->part0));
1058 : }
1059 :
1060 0 : static ssize_t diskseq_show(struct device *dev,
1061 : struct device_attribute *attr, char *buf)
1062 : {
1063 0 : struct gendisk *disk = dev_to_disk(dev);
1064 :
1065 0 : return sprintf(buf, "%llu\n", disk->diskseq);
1066 : }
1067 :
1068 : static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
1069 : static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
1070 : static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
1071 : static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
1072 : static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
1073 : static DEVICE_ATTR(size, 0444, part_size_show, NULL);
1074 : static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
1075 : static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
1076 : static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
1077 : static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
1078 : static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
1079 : static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
1080 : static DEVICE_ATTR(diskseq, 0444, diskseq_show, NULL);
1081 :
1082 : #ifdef CONFIG_FAIL_MAKE_REQUEST
1083 : ssize_t part_fail_show(struct device *dev,
1084 : struct device_attribute *attr, char *buf)
1085 : {
1086 : return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail);
1087 : }
1088 :
1089 : ssize_t part_fail_store(struct device *dev,
1090 : struct device_attribute *attr,
1091 : const char *buf, size_t count)
1092 : {
1093 : int i;
1094 :
1095 : if (count > 0 && sscanf(buf, "%d", &i) > 0)
1096 : dev_to_bdev(dev)->bd_make_it_fail = i;
1097 :
1098 : return count;
1099 : }
1100 :
1101 : static struct device_attribute dev_attr_fail =
1102 : __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
1103 : #endif /* CONFIG_FAIL_MAKE_REQUEST */
1104 :
1105 : #ifdef CONFIG_FAIL_IO_TIMEOUT
1106 : static struct device_attribute dev_attr_fail_timeout =
1107 : __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
1108 : #endif
1109 :
1110 : static struct attribute *disk_attrs[] = {
1111 : &dev_attr_range.attr,
1112 : &dev_attr_ext_range.attr,
1113 : &dev_attr_removable.attr,
1114 : &dev_attr_hidden.attr,
1115 : &dev_attr_ro.attr,
1116 : &dev_attr_size.attr,
1117 : &dev_attr_alignment_offset.attr,
1118 : &dev_attr_discard_alignment.attr,
1119 : &dev_attr_capability.attr,
1120 : &dev_attr_stat.attr,
1121 : &dev_attr_inflight.attr,
1122 : &dev_attr_badblocks.attr,
1123 : &dev_attr_events.attr,
1124 : &dev_attr_events_async.attr,
1125 : &dev_attr_events_poll_msecs.attr,
1126 : &dev_attr_diskseq.attr,
1127 : #ifdef CONFIG_FAIL_MAKE_REQUEST
1128 : &dev_attr_fail.attr,
1129 : #endif
1130 : #ifdef CONFIG_FAIL_IO_TIMEOUT
1131 : &dev_attr_fail_timeout.attr,
1132 : #endif
1133 : NULL
1134 : };
1135 :
1136 0 : static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1137 : {
1138 0 : struct device *dev = container_of(kobj, typeof(*dev), kobj);
1139 0 : struct gendisk *disk = dev_to_disk(dev);
1140 :
1141 0 : if (a == &dev_attr_badblocks.attr && !disk->bb)
1142 : return 0;
1143 0 : return a->mode;
1144 : }
1145 :
1146 : static struct attribute_group disk_attr_group = {
1147 : .attrs = disk_attrs,
1148 : .is_visible = disk_visible,
1149 : };
1150 :
1151 : static const struct attribute_group *disk_attr_groups[] = {
1152 : &disk_attr_group,
1153 : #ifdef CONFIG_BLK_DEV_IO_TRACE
1154 : &blk_trace_attr_group,
1155 : #endif
1156 : NULL
1157 : };
1158 :
1159 : /**
1160 : * disk_release - releases all allocated resources of the gendisk
1161 : * @dev: the device representing this disk
1162 : *
1163 : * This function releases all allocated resources of the gendisk.
1164 : *
1165 : * Drivers which used __device_add_disk() have a gendisk with a request_queue
1166 : * assigned. Since the request_queue sits on top of the gendisk for these
1167 : * drivers we also call blk_put_queue() for them, and we expect the
1168 : * request_queue refcount to reach 0 at this point, and so the request_queue
1169 : * will also be freed prior to the disk.
1170 : *
1171 : * Context: can sleep
1172 : */
1173 0 : static void disk_release(struct device *dev)
1174 : {
1175 0 : struct gendisk *disk = dev_to_disk(dev);
1176 :
1177 : might_sleep();
1178 0 : WARN_ON_ONCE(disk_live(disk));
1179 :
1180 : /*
1181 : * To undo the all initialization from blk_mq_init_allocated_queue in
1182 : * case of a probe failure where add_disk is never called we have to
1183 : * call blk_mq_exit_queue here. We can't do this for the more common
1184 : * teardown case (yet) as the tagset can be gone by the time the disk
1185 : * is released once it was added.
1186 : */
1187 0 : if (queue_is_mq(disk->queue) &&
1188 0 : test_bit(GD_OWNS_QUEUE, &disk->state) &&
1189 0 : !test_bit(GD_ADDED, &disk->state))
1190 0 : blk_mq_exit_queue(disk->queue);
1191 :
1192 0 : blkcg_exit_disk(disk);
1193 :
1194 0 : bioset_exit(&disk->bio_split);
1195 :
1196 0 : disk_release_events(disk);
1197 0 : kfree(disk->random);
1198 0 : disk_free_zone_bitmaps(disk);
1199 0 : xa_destroy(&disk->part_tbl);
1200 :
1201 0 : disk->queue->disk = NULL;
1202 0 : blk_put_queue(disk->queue);
1203 :
1204 0 : if (test_bit(GD_ADDED, &disk->state) && disk->fops->free_disk)
1205 0 : disk->fops->free_disk(disk);
1206 :
1207 0 : iput(disk->part0->bd_inode); /* frees the disk */
1208 0 : }
1209 :
1210 0 : static int block_uevent(const struct device *dev, struct kobj_uevent_env *env)
1211 : {
1212 0 : const struct gendisk *disk = dev_to_disk(dev);
1213 :
1214 0 : return add_uevent_var(env, "DISKSEQ=%llu", disk->diskseq);
1215 : }
1216 :
1217 : struct class block_class = {
1218 : .name = "block",
1219 : .dev_uevent = block_uevent,
1220 : };
1221 :
1222 0 : static char *block_devnode(const struct device *dev, umode_t *mode,
1223 : kuid_t *uid, kgid_t *gid)
1224 : {
1225 0 : struct gendisk *disk = dev_to_disk(dev);
1226 :
1227 0 : if (disk->fops->devnode)
1228 0 : return disk->fops->devnode(disk, mode);
1229 : return NULL;
1230 : }
1231 :
1232 : const struct device_type disk_type = {
1233 : .name = "disk",
1234 : .groups = disk_attr_groups,
1235 : .release = disk_release,
1236 : .devnode = block_devnode,
1237 : };
1238 :
1239 : #ifdef CONFIG_PROC_FS
1240 : /*
1241 : * aggregate disk stat collector. Uses the same stats that the sysfs
1242 : * entries do, above, but makes them available through one seq_file.
1243 : *
1244 : * The output looks suspiciously like /proc/partitions with a bunch of
1245 : * extra fields.
1246 : */
1247 0 : static int diskstats_show(struct seq_file *seqf, void *v)
1248 : {
1249 0 : struct gendisk *gp = v;
1250 : struct block_device *hd;
1251 : unsigned int inflight;
1252 : struct disk_stats stat;
1253 : unsigned long idx;
1254 :
1255 : /*
1256 : if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1257 : seq_puts(seqf, "major minor name"
1258 : " rio rmerge rsect ruse wio wmerge "
1259 : "wsect wuse running use aveq"
1260 : "\n\n");
1261 : */
1262 :
1263 : rcu_read_lock();
1264 0 : xa_for_each(&gp->part_tbl, idx, hd) {
1265 0 : if (bdev_is_partition(hd) && !bdev_nr_sectors(hd))
1266 0 : continue;
1267 0 : if (queue_is_mq(gp->queue))
1268 0 : inflight = blk_mq_in_flight(gp->queue, hd);
1269 : else
1270 : inflight = part_in_flight(hd);
1271 :
1272 0 : if (inflight) {
1273 0 : part_stat_lock();
1274 0 : update_io_ticks(hd, jiffies, true);
1275 0 : part_stat_unlock();
1276 : }
1277 0 : part_stat_read_all(hd, &stat);
1278 0 : seq_printf(seqf, "%4d %7d %pg "
1279 : "%lu %lu %lu %u "
1280 : "%lu %lu %lu %u "
1281 : "%u %u %u "
1282 : "%lu %lu %lu %u "
1283 : "%lu %u"
1284 : "\n",
1285 0 : MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd,
1286 : stat.ios[STAT_READ],
1287 : stat.merges[STAT_READ],
1288 : stat.sectors[STAT_READ],
1289 0 : (unsigned int)div_u64(stat.nsecs[STAT_READ],
1290 : NSEC_PER_MSEC),
1291 : stat.ios[STAT_WRITE],
1292 : stat.merges[STAT_WRITE],
1293 : stat.sectors[STAT_WRITE],
1294 0 : (unsigned int)div_u64(stat.nsecs[STAT_WRITE],
1295 : NSEC_PER_MSEC),
1296 : inflight,
1297 : jiffies_to_msecs(stat.io_ticks),
1298 0 : (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1299 0 : stat.nsecs[STAT_WRITE] +
1300 : stat.nsecs[STAT_DISCARD] +
1301 : stat.nsecs[STAT_FLUSH],
1302 : NSEC_PER_MSEC),
1303 : stat.ios[STAT_DISCARD],
1304 : stat.merges[STAT_DISCARD],
1305 : stat.sectors[STAT_DISCARD],
1306 0 : (unsigned int)div_u64(stat.nsecs[STAT_DISCARD],
1307 : NSEC_PER_MSEC),
1308 : stat.ios[STAT_FLUSH],
1309 0 : (unsigned int)div_u64(stat.nsecs[STAT_FLUSH],
1310 : NSEC_PER_MSEC)
1311 : );
1312 : }
1313 : rcu_read_unlock();
1314 :
1315 0 : return 0;
1316 : }
1317 :
1318 : static const struct seq_operations diskstats_op = {
1319 : .start = disk_seqf_start,
1320 : .next = disk_seqf_next,
1321 : .stop = disk_seqf_stop,
1322 : .show = diskstats_show
1323 : };
1324 :
1325 1 : static int __init proc_genhd_init(void)
1326 : {
1327 1 : proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1328 1 : proc_create_seq("partitions", 0, NULL, &partitions_op);
1329 1 : return 0;
1330 : }
1331 : module_init(proc_genhd_init);
1332 : #endif /* CONFIG_PROC_FS */
1333 :
1334 0 : dev_t part_devt(struct gendisk *disk, u8 partno)
1335 : {
1336 : struct block_device *part;
1337 0 : dev_t devt = 0;
1338 :
1339 : rcu_read_lock();
1340 0 : part = xa_load(&disk->part_tbl, partno);
1341 0 : if (part)
1342 0 : devt = part->bd_dev;
1343 : rcu_read_unlock();
1344 :
1345 0 : return devt;
1346 : }
1347 :
1348 0 : dev_t blk_lookup_devt(const char *name, int partno)
1349 : {
1350 0 : dev_t devt = MKDEV(0, 0);
1351 : struct class_dev_iter iter;
1352 : struct device *dev;
1353 :
1354 0 : class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1355 0 : while ((dev = class_dev_iter_next(&iter))) {
1356 0 : struct gendisk *disk = dev_to_disk(dev);
1357 :
1358 0 : if (strcmp(dev_name(dev), name))
1359 0 : continue;
1360 :
1361 0 : if (partno < disk->minors) {
1362 : /* We need to return the right devno, even
1363 : * if the partition doesn't exist yet.
1364 : */
1365 0 : devt = MKDEV(MAJOR(dev->devt),
1366 : MINOR(dev->devt) + partno);
1367 : } else {
1368 0 : devt = part_devt(disk, partno);
1369 0 : if (devt)
1370 : break;
1371 : }
1372 : }
1373 0 : class_dev_iter_exit(&iter);
1374 0 : return devt;
1375 : }
1376 :
1377 0 : struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id,
1378 : struct lock_class_key *lkclass)
1379 : {
1380 : struct gendisk *disk;
1381 :
1382 0 : disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1383 0 : if (!disk)
1384 : return NULL;
1385 :
1386 0 : if (bioset_init(&disk->bio_split, BIO_POOL_SIZE, 0, 0))
1387 : goto out_free_disk;
1388 :
1389 0 : disk->bdi = bdi_alloc(node_id);
1390 0 : if (!disk->bdi)
1391 : goto out_free_bioset;
1392 :
1393 : /* bdev_alloc() might need the queue, set before the first call */
1394 0 : disk->queue = q;
1395 :
1396 0 : disk->part0 = bdev_alloc(disk, 0);
1397 0 : if (!disk->part0)
1398 : goto out_free_bdi;
1399 :
1400 0 : disk->node_id = node_id;
1401 0 : mutex_init(&disk->open_mutex);
1402 0 : xa_init(&disk->part_tbl);
1403 0 : if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL))
1404 : goto out_destroy_part_tbl;
1405 :
1406 0 : if (blkcg_init_disk(disk))
1407 : goto out_erase_part0;
1408 :
1409 0 : rand_initialize_disk(disk);
1410 0 : disk_to_dev(disk)->class = &block_class;
1411 0 : disk_to_dev(disk)->type = &disk_type;
1412 0 : device_initialize(disk_to_dev(disk));
1413 0 : inc_diskseq(disk);
1414 0 : q->disk = disk;
1415 : lockdep_init_map(&disk->lockdep_map, "(bio completion)", lkclass, 0);
1416 : #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
1417 : INIT_LIST_HEAD(&disk->slave_bdevs);
1418 : #endif
1419 0 : return disk;
1420 :
1421 : out_erase_part0:
1422 : xa_erase(&disk->part_tbl, 0);
1423 : out_destroy_part_tbl:
1424 0 : xa_destroy(&disk->part_tbl);
1425 0 : disk->part0->bd_disk = NULL;
1426 0 : iput(disk->part0->bd_inode);
1427 : out_free_bdi:
1428 0 : bdi_put(disk->bdi);
1429 : out_free_bioset:
1430 0 : bioset_exit(&disk->bio_split);
1431 : out_free_disk:
1432 0 : kfree(disk);
1433 0 : return NULL;
1434 : }
1435 :
1436 0 : struct gendisk *__blk_alloc_disk(int node, struct lock_class_key *lkclass)
1437 : {
1438 : struct request_queue *q;
1439 : struct gendisk *disk;
1440 :
1441 0 : q = blk_alloc_queue(node);
1442 0 : if (!q)
1443 : return NULL;
1444 :
1445 0 : disk = __alloc_disk_node(q, node, lkclass);
1446 0 : if (!disk) {
1447 0 : blk_put_queue(q);
1448 0 : return NULL;
1449 : }
1450 0 : set_bit(GD_OWNS_QUEUE, &disk->state);
1451 0 : return disk;
1452 : }
1453 : EXPORT_SYMBOL(__blk_alloc_disk);
1454 :
1455 : /**
1456 : * put_disk - decrements the gendisk refcount
1457 : * @disk: the struct gendisk to decrement the refcount for
1458 : *
1459 : * This decrements the refcount for the struct gendisk. When this reaches 0
1460 : * we'll have disk_release() called.
1461 : *
1462 : * Note: for blk-mq disk put_disk must be called before freeing the tag_set
1463 : * when handling probe errors (that is before add_disk() is called).
1464 : *
1465 : * Context: Any context, but the last reference must not be dropped from
1466 : * atomic context.
1467 : */
1468 0 : void put_disk(struct gendisk *disk)
1469 : {
1470 0 : if (disk)
1471 0 : put_device(disk_to_dev(disk));
1472 0 : }
1473 : EXPORT_SYMBOL(put_disk);
1474 :
1475 0 : static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1476 : {
1477 0 : char event[] = "DISK_RO=1";
1478 0 : char *envp[] = { event, NULL };
1479 :
1480 0 : if (!ro)
1481 0 : event[8] = '0';
1482 0 : kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1483 0 : }
1484 :
1485 : /**
1486 : * set_disk_ro - set a gendisk read-only
1487 : * @disk: gendisk to operate on
1488 : * @read_only: %true to set the disk read-only, %false set the disk read/write
1489 : *
1490 : * This function is used to indicate whether a given disk device should have its
1491 : * read-only flag set. set_disk_ro() is typically used by device drivers to
1492 : * indicate whether the underlying physical device is write-protected.
1493 : */
1494 0 : void set_disk_ro(struct gendisk *disk, bool read_only)
1495 : {
1496 0 : if (read_only) {
1497 0 : if (test_and_set_bit(GD_READ_ONLY, &disk->state))
1498 : return;
1499 : } else {
1500 0 : if (!test_and_clear_bit(GD_READ_ONLY, &disk->state))
1501 : return;
1502 : }
1503 0 : set_disk_ro_uevent(disk, read_only);
1504 : }
1505 : EXPORT_SYMBOL(set_disk_ro);
1506 :
1507 0 : void inc_diskseq(struct gendisk *disk)
1508 : {
1509 0 : disk->diskseq = atomic64_inc_return(&diskseq);
1510 0 : }
|
