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/genhd.h>
12 : #include <linux/kdev_t.h>
13 : #include <linux/kernel.h>
14 : #include <linux/blkdev.h>
15 : #include <linux/backing-dev.h>
16 : #include <linux/init.h>
17 : #include <linux/spinlock.h>
18 : #include <linux/proc_fs.h>
19 : #include <linux/seq_file.h>
20 : #include <linux/slab.h>
21 : #include <linux/kmod.h>
22 : #include <linux/mutex.h>
23 : #include <linux/idr.h>
24 : #include <linux/log2.h>
25 : #include <linux/pm_runtime.h>
26 : #include <linux/badblocks.h>
27 :
28 : #include "blk.h"
29 :
30 : static struct kobject *block_depr;
31 :
32 : DECLARE_RWSEM(bdev_lookup_sem);
33 :
34 : /* for extended dynamic devt allocation, currently only one major is used */
35 : #define NR_EXT_DEVT (1 << MINORBITS)
36 : static DEFINE_IDA(ext_devt_ida);
37 :
38 : static void disk_check_events(struct disk_events *ev,
39 : unsigned int *clearing_ptr);
40 : static void disk_alloc_events(struct gendisk *disk);
41 : static void disk_add_events(struct gendisk *disk);
42 : static void disk_del_events(struct gendisk *disk);
43 : static void disk_release_events(struct gendisk *disk);
44 :
45 1 : void set_capacity(struct gendisk *disk, sector_t sectors)
46 : {
47 1 : struct block_device *bdev = disk->part0;
48 :
49 1 : spin_lock(&bdev->bd_size_lock);
50 1 : i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
51 1 : spin_unlock(&bdev->bd_size_lock);
52 1 : }
53 : EXPORT_SYMBOL(set_capacity);
54 :
55 : /*
56 : * Set disk capacity and notify if the size is not currently zero and will not
57 : * be set to zero. Returns true if a uevent was sent, otherwise false.
58 : */
59 1 : bool set_capacity_and_notify(struct gendisk *disk, sector_t size)
60 : {
61 1 : sector_t capacity = get_capacity(disk);
62 1 : char *envp[] = { "RESIZE=1", NULL };
63 :
64 1 : set_capacity(disk, size);
65 :
66 : /*
67 : * Only print a message and send a uevent if the gendisk is user visible
68 : * and alive. This avoids spamming the log and udev when setting the
69 : * initial capacity during probing.
70 : */
71 1 : if (size == capacity ||
72 1 : (disk->flags & (GENHD_FL_UP | GENHD_FL_HIDDEN)) != GENHD_FL_UP)
73 : return false;
74 :
75 0 : pr_info("%s: detected capacity change from %lld to %lld\n",
76 : disk->disk_name, capacity, size);
77 :
78 : /*
79 : * Historically we did not send a uevent for changes to/from an empty
80 : * device.
81 : */
82 0 : if (!capacity || !size)
83 : return false;
84 0 : kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
85 0 : return true;
86 : }
87 : EXPORT_SYMBOL_GPL(set_capacity_and_notify);
88 :
89 : /*
90 : * Format the device name of the indicated disk into the supplied buffer and
91 : * return a pointer to that same buffer for convenience.
92 : */
93 4 : char *disk_name(struct gendisk *hd, int partno, char *buf)
94 : {
95 4 : if (!partno)
96 2 : snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name);
97 2 : else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1]))
98 0 : snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno);
99 : else
100 2 : snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno);
101 :
102 4 : return buf;
103 : }
104 :
105 1 : const char *bdevname(struct block_device *bdev, char *buf)
106 : {
107 1 : return disk_name(bdev->bd_disk, bdev->bd_partno, buf);
108 : }
109 : EXPORT_SYMBOL(bdevname);
110 :
111 0 : static void part_stat_read_all(struct block_device *part,
112 : struct disk_stats *stat)
113 : {
114 0 : int cpu;
115 :
116 0 : memset(stat, 0, sizeof(struct disk_stats));
117 0 : for_each_possible_cpu(cpu) {
118 0 : struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu);
119 0 : int group;
120 :
121 0 : for (group = 0; group < NR_STAT_GROUPS; group++) {
122 0 : stat->nsecs[group] += ptr->nsecs[group];
123 0 : stat->sectors[group] += ptr->sectors[group];
124 0 : stat->ios[group] += ptr->ios[group];
125 0 : stat->merges[group] += ptr->merges[group];
126 : }
127 :
128 0 : stat->io_ticks += ptr->io_ticks;
129 : }
130 0 : }
131 :
132 0 : static unsigned int part_in_flight(struct block_device *part)
133 : {
134 0 : unsigned int inflight = 0;
135 0 : int cpu;
136 :
137 0 : for_each_possible_cpu(cpu) {
138 0 : inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) +
139 0 : part_stat_local_read_cpu(part, in_flight[1], cpu);
140 : }
141 0 : if ((int)inflight < 0)
142 0 : inflight = 0;
143 :
144 0 : return inflight;
145 : }
146 :
147 0 : static void part_in_flight_rw(struct block_device *part,
148 : unsigned int inflight[2])
149 : {
150 0 : int cpu;
151 :
152 0 : inflight[0] = 0;
153 0 : inflight[1] = 0;
154 0 : for_each_possible_cpu(cpu) {
155 0 : inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu);
156 0 : inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu);
157 : }
158 0 : if ((int)inflight[0] < 0)
159 0 : inflight[0] = 0;
160 0 : if ((int)inflight[1] < 0)
161 0 : inflight[1] = 0;
162 0 : }
163 :
164 : /**
165 : * disk_part_iter_init - initialize partition iterator
166 : * @piter: iterator to initialize
167 : * @disk: disk to iterate over
168 : * @flags: DISK_PITER_* flags
169 : *
170 : * Initialize @piter so that it iterates over partitions of @disk.
171 : *
172 : * CONTEXT:
173 : * Don't care.
174 : */
175 11 : void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk,
176 : unsigned int flags)
177 : {
178 11 : piter->disk = disk;
179 11 : piter->part = NULL;
180 1 : if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0))
181 0 : piter->idx = 0;
182 : else
183 1 : piter->idx = 1;
184 11 : piter->flags = flags;
185 11 : }
186 :
187 : /**
188 : * disk_part_iter_next - proceed iterator to the next partition and return it
189 : * @piter: iterator of interest
190 : *
191 : * Proceed @piter to the next partition and return it.
192 : *
193 : * CONTEXT:
194 : * Don't care.
195 : */
196 23 : struct block_device *disk_part_iter_next(struct disk_part_iter *piter)
197 : {
198 23 : struct block_device *part;
199 23 : unsigned long idx;
200 :
201 : /* put the last partition */
202 46 : disk_part_iter_exit(piter);
203 :
204 23 : rcu_read_lock();
205 23 : xa_for_each_start(&piter->disk->part_tbl, idx, part, piter->idx) {
206 12 : if (!bdev_nr_sectors(part) &&
207 8 : !(piter->flags & DISK_PITER_INCL_EMPTY) &&
208 8 : !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 &&
209 8 : piter->idx == 0))
210 0 : continue;
211 :
212 12 : piter->part = bdgrab(part);
213 12 : if (!piter->part)
214 0 : continue;
215 12 : piter->idx = idx + 1;
216 12 : break;
217 : }
218 23 : rcu_read_unlock();
219 :
220 23 : return piter->part;
221 : }
222 :
223 : /**
224 : * disk_part_iter_exit - finish up partition iteration
225 : * @piter: iter of interest
226 : *
227 : * Called when iteration is over. Cleans up @piter.
228 : *
229 : * CONTEXT:
230 : * Don't care.
231 : */
232 34 : void disk_part_iter_exit(struct disk_part_iter *piter)
233 : {
234 24 : if (piter->part)
235 12 : bdput(piter->part);
236 23 : piter->part = NULL;
237 1 : }
238 :
239 : /*
240 : * Can be deleted altogether. Later.
241 : *
242 : */
243 : #define BLKDEV_MAJOR_HASH_SIZE 255
244 : static struct blk_major_name {
245 : struct blk_major_name *next;
246 : int major;
247 : char name[16];
248 : void (*probe)(dev_t devt);
249 : } *major_names[BLKDEV_MAJOR_HASH_SIZE];
250 : static DEFINE_MUTEX(major_names_lock);
251 :
252 : /* index in the above - for now: assume no multimajor ranges */
253 21 : static inline int major_to_index(unsigned major)
254 : {
255 21 : return major % BLKDEV_MAJOR_HASH_SIZE;
256 : }
257 :
258 : #ifdef CONFIG_PROC_FS
259 0 : void blkdev_show(struct seq_file *seqf, off_t offset)
260 : {
261 0 : struct blk_major_name *dp;
262 :
263 0 : mutex_lock(&major_names_lock);
264 0 : for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next)
265 0 : if (dp->major == offset)
266 0 : seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
267 0 : mutex_unlock(&major_names_lock);
268 0 : }
269 : #endif /* CONFIG_PROC_FS */
270 :
271 : /**
272 : * __register_blkdev - register a new block device
273 : *
274 : * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
275 : * @major = 0, try to allocate any unused major number.
276 : * @name: the name of the new block device as a zero terminated string
277 : * @probe: allback that is called on access to any minor number of @major
278 : *
279 : * The @name must be unique within the system.
280 : *
281 : * The return value depends on the @major input parameter:
282 : *
283 : * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
284 : * then the function returns zero on success, or a negative error code
285 : * - if any unused major number was requested with @major = 0 parameter
286 : * then the return value is the allocated major number in range
287 : * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
288 : *
289 : * See Documentation/admin-guide/devices.txt for the list of allocated
290 : * major numbers.
291 : *
292 : * Use register_blkdev instead for any new code.
293 : */
294 21 : int __register_blkdev(unsigned int major, const char *name,
295 : void (*probe)(dev_t devt))
296 : {
297 21 : struct blk_major_name **n, *p;
298 21 : int index, ret = 0;
299 :
300 21 : mutex_lock(&major_names_lock);
301 :
302 : /* temporary */
303 21 : if (major == 0) {
304 3 : for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
305 3 : if (major_names[index] == NULL)
306 : break;
307 : }
308 :
309 2 : if (index == 0) {
310 0 : printk("%s: failed to get major for %s\n",
311 : __func__, name);
312 0 : ret = -EBUSY;
313 0 : goto out;
314 : }
315 2 : major = index;
316 2 : ret = major;
317 : }
318 :
319 21 : if (major >= BLKDEV_MAJOR_MAX) {
320 0 : pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
321 : __func__, major, BLKDEV_MAJOR_MAX-1, name);
322 :
323 0 : ret = -EINVAL;
324 0 : goto out;
325 : }
326 :
327 21 : p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
328 21 : if (p == NULL) {
329 0 : ret = -ENOMEM;
330 0 : goto out;
331 : }
332 :
333 21 : p->major = major;
334 21 : p->probe = probe;
335 21 : strlcpy(p->name, name, sizeof(p->name));
336 21 : p->next = NULL;
337 21 : index = major_to_index(major);
338 :
339 21 : for (n = &major_names[index]; *n; n = &(*n)->next) {
340 0 : if ((*n)->major == major)
341 : break;
342 : }
343 21 : if (!*n)
344 21 : *n = p;
345 : else
346 : ret = -EBUSY;
347 :
348 21 : if (ret < 0) {
349 0 : printk("register_blkdev: cannot get major %u for %s\n",
350 : major, name);
351 0 : kfree(p);
352 : }
353 21 : out:
354 21 : mutex_unlock(&major_names_lock);
355 21 : return ret;
356 : }
357 : EXPORT_SYMBOL(__register_blkdev);
358 :
359 0 : void unregister_blkdev(unsigned int major, const char *name)
360 : {
361 0 : struct blk_major_name **n;
362 0 : struct blk_major_name *p = NULL;
363 0 : int index = major_to_index(major);
364 :
365 0 : mutex_lock(&major_names_lock);
366 0 : for (n = &major_names[index]; *n; n = &(*n)->next)
367 0 : if ((*n)->major == major)
368 : break;
369 0 : if (!*n || strcmp((*n)->name, name)) {
370 0 : WARN_ON(1);
371 : } else {
372 0 : p = *n;
373 0 : *n = p->next;
374 : }
375 0 : mutex_unlock(&major_names_lock);
376 0 : kfree(p);
377 0 : }
378 :
379 : EXPORT_SYMBOL(unregister_blkdev);
380 :
381 : /**
382 : * blk_mangle_minor - scatter minor numbers apart
383 : * @minor: minor number to mangle
384 : *
385 : * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
386 : * is enabled. Mangling twice gives the original value.
387 : *
388 : * RETURNS:
389 : * Mangled value.
390 : *
391 : * CONTEXT:
392 : * Don't care.
393 : */
394 0 : static int blk_mangle_minor(int minor)
395 : {
396 : #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
397 : int i;
398 :
399 : for (i = 0; i < MINORBITS / 2; i++) {
400 : int low = minor & (1 << i);
401 : int high = minor & (1 << (MINORBITS - 1 - i));
402 : int distance = MINORBITS - 1 - 2 * i;
403 :
404 : minor ^= low | high; /* clear both bits */
405 : low <<= distance; /* swap the positions */
406 : high >>= distance;
407 : minor |= low | high; /* and set */
408 : }
409 : #endif
410 0 : return minor;
411 : }
412 :
413 : /**
414 : * blk_alloc_devt - allocate a dev_t for a block device
415 : * @bdev: block device to allocate dev_t for
416 : * @devt: out parameter for resulting dev_t
417 : *
418 : * Allocate a dev_t for block device.
419 : *
420 : * RETURNS:
421 : * 0 on success, allocated dev_t is returned in *@devt. -errno on
422 : * failure.
423 : *
424 : * CONTEXT:
425 : * Might sleep.
426 : */
427 10 : int blk_alloc_devt(struct block_device *bdev, dev_t *devt)
428 : {
429 10 : struct gendisk *disk = bdev->bd_disk;
430 10 : int idx;
431 :
432 : /* in consecutive minor range? */
433 10 : if (bdev->bd_partno < disk->minors) {
434 10 : *devt = MKDEV(disk->major, disk->first_minor + bdev->bd_partno);
435 10 : return 0;
436 : }
437 :
438 0 : idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT, GFP_KERNEL);
439 0 : if (idx < 0)
440 0 : return idx == -ENOSPC ? -EBUSY : idx;
441 :
442 0 : *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
443 0 : return 0;
444 : }
445 :
446 : /**
447 : * blk_free_devt - free a dev_t
448 : * @devt: dev_t to free
449 : *
450 : * Free @devt which was allocated using blk_alloc_devt().
451 : *
452 : * CONTEXT:
453 : * Might sleep.
454 : */
455 0 : void blk_free_devt(dev_t devt)
456 : {
457 0 : if (MAJOR(devt) == BLOCK_EXT_MAJOR)
458 0 : ida_free(&ext_devt_ida, blk_mangle_minor(MINOR(devt)));
459 0 : }
460 :
461 0 : static char *bdevt_str(dev_t devt, char *buf)
462 : {
463 0 : if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
464 0 : char tbuf[BDEVT_SIZE];
465 0 : snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
466 0 : snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
467 : } else
468 0 : snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
469 :
470 0 : return buf;
471 : }
472 :
473 9 : void disk_uevent(struct gendisk *disk, enum kobject_action action)
474 : {
475 9 : struct disk_part_iter piter;
476 9 : struct block_device *part;
477 :
478 9 : disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY_PART0);
479 19 : while ((part = disk_part_iter_next(&piter)))
480 10 : kobject_uevent(bdev_kobj(part), action);
481 9 : disk_part_iter_exit(&piter);
482 9 : }
483 : EXPORT_SYMBOL_GPL(disk_uevent);
484 :
485 9 : static void disk_scan_partitions(struct gendisk *disk)
486 : {
487 9 : struct block_device *bdev;
488 :
489 10 : if (!get_capacity(disk) || !disk_part_scan_enabled(disk))
490 : return;
491 :
492 1 : set_bit(GD_NEED_PART_SCAN, &disk->state);
493 1 : bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL);
494 1 : if (!IS_ERR(bdev))
495 1 : blkdev_put(bdev, FMODE_READ);
496 : }
497 :
498 9 : static void register_disk(struct device *parent, struct gendisk *disk,
499 : const struct attribute_group **groups)
500 : {
501 9 : struct device *ddev = disk_to_dev(disk);
502 9 : int err;
503 :
504 9 : ddev->parent = parent;
505 :
506 9 : dev_set_name(ddev, "%s", disk->disk_name);
507 :
508 : /* delay uevents, until we scanned partition table */
509 9 : dev_set_uevent_suppress(ddev, 1);
510 :
511 9 : if (groups) {
512 1 : WARN_ON(ddev->groups);
513 1 : ddev->groups = groups;
514 : }
515 9 : if (device_add(ddev))
516 : return;
517 9 : if (!sysfs_deprecated) {
518 18 : err = sysfs_create_link(block_depr, &ddev->kobj,
519 9 : kobject_name(&ddev->kobj));
520 9 : if (err) {
521 0 : device_del(ddev);
522 0 : return;
523 : }
524 : }
525 :
526 : /*
527 : * avoid probable deadlock caused by allocating memory with
528 : * GFP_KERNEL in runtime_resume callback of its all ancestor
529 : * devices
530 : */
531 9 : pm_runtime_set_memalloc_noio(ddev, true);
532 :
533 18 : disk->part0->bd_holder_dir =
534 9 : kobject_create_and_add("holders", &ddev->kobj);
535 9 : disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
536 :
537 9 : if (disk->flags & GENHD_FL_HIDDEN)
538 : return;
539 :
540 9 : disk_scan_partitions(disk);
541 :
542 : /* announce the disk and partitions after all partitions are created */
543 9 : dev_set_uevent_suppress(ddev, 0);
544 9 : disk_uevent(disk, KOBJ_ADD);
545 :
546 9 : if (disk->queue->backing_dev_info->dev) {
547 9 : err = sysfs_create_link(&ddev->kobj,
548 : &disk->queue->backing_dev_info->dev->kobj,
549 : "bdi");
550 9 : WARN_ON(err);
551 : }
552 : }
553 :
554 : /**
555 : * __device_add_disk - add disk information to kernel list
556 : * @parent: parent device for the disk
557 : * @disk: per-device partitioning information
558 : * @groups: Additional per-device sysfs groups
559 : * @register_queue: register the queue if set to true
560 : *
561 : * This function registers the partitioning information in @disk
562 : * with the kernel.
563 : *
564 : * FIXME: error handling
565 : */
566 9 : static void __device_add_disk(struct device *parent, struct gendisk *disk,
567 : const struct attribute_group **groups,
568 : bool register_queue)
569 : {
570 9 : dev_t devt;
571 9 : int retval;
572 :
573 : /*
574 : * The disk queue should now be all set with enough information about
575 : * the device for the elevator code to pick an adequate default
576 : * elevator if one is needed, that is, for devices requesting queue
577 : * registration.
578 : */
579 9 : if (register_queue)
580 9 : elevator_init_mq(disk->queue);
581 :
582 : /* minors == 0 indicates to use ext devt from part0 and should
583 : * be accompanied with EXT_DEVT flag. Make sure all
584 : * parameters make sense.
585 : */
586 18 : WARN_ON(disk->minors && !(disk->major || disk->first_minor));
587 18 : WARN_ON(!disk->minors &&
588 : !(disk->flags & (GENHD_FL_EXT_DEVT | GENHD_FL_HIDDEN)));
589 :
590 9 : disk->flags |= GENHD_FL_UP;
591 :
592 9 : retval = blk_alloc_devt(disk->part0, &devt);
593 9 : if (retval) {
594 0 : WARN_ON(1);
595 0 : return;
596 : }
597 9 : disk->major = MAJOR(devt);
598 9 : disk->first_minor = MINOR(devt);
599 :
600 9 : disk_alloc_events(disk);
601 :
602 9 : if (disk->flags & GENHD_FL_HIDDEN) {
603 : /*
604 : * Don't let hidden disks show up in /proc/partitions,
605 : * and don't bother scanning for partitions either.
606 : */
607 0 : disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
608 0 : disk->flags |= GENHD_FL_NO_PART_SCAN;
609 : } else {
610 9 : struct backing_dev_info *bdi = disk->queue->backing_dev_info;
611 9 : struct device *dev = disk_to_dev(disk);
612 9 : int ret;
613 :
614 : /* Register BDI before referencing it from bdev */
615 9 : dev->devt = devt;
616 9 : ret = bdi_register(bdi, "%u:%u", MAJOR(devt), MINOR(devt));
617 9 : WARN_ON(ret);
618 9 : bdi_set_owner(bdi, dev);
619 9 : bdev_add(disk->part0, devt);
620 : }
621 9 : register_disk(parent, disk, groups);
622 9 : if (register_queue)
623 9 : blk_register_queue(disk);
624 :
625 : /*
626 : * Take an extra ref on queue which will be put on disk_release()
627 : * so that it sticks around as long as @disk is there.
628 : */
629 9 : WARN_ON_ONCE(!blk_get_queue(disk->queue));
630 :
631 9 : disk_add_events(disk);
632 9 : blk_integrity_add(disk);
633 : }
634 :
635 9 : void device_add_disk(struct device *parent, struct gendisk *disk,
636 : const struct attribute_group **groups)
637 :
638 : {
639 9 : __device_add_disk(parent, disk, groups, true);
640 9 : }
641 : EXPORT_SYMBOL(device_add_disk);
642 :
643 0 : void device_add_disk_no_queue_reg(struct device *parent, struct gendisk *disk)
644 : {
645 0 : __device_add_disk(parent, disk, NULL, false);
646 0 : }
647 : EXPORT_SYMBOL(device_add_disk_no_queue_reg);
648 :
649 0 : static void invalidate_partition(struct block_device *bdev)
650 : {
651 0 : fsync_bdev(bdev);
652 0 : __invalidate_device(bdev, true);
653 :
654 : /*
655 : * Unhash the bdev inode for this device so that it can't be looked
656 : * up any more even if openers still hold references to it.
657 : */
658 0 : remove_inode_hash(bdev->bd_inode);
659 0 : }
660 :
661 : /**
662 : * del_gendisk - remove the gendisk
663 : * @disk: the struct gendisk to remove
664 : *
665 : * Removes the gendisk and all its associated resources. This deletes the
666 : * partitions associated with the gendisk, and unregisters the associated
667 : * request_queue.
668 : *
669 : * This is the counter to the respective __device_add_disk() call.
670 : *
671 : * The final removal of the struct gendisk happens when its refcount reaches 0
672 : * with put_disk(), which should be called after del_gendisk(), if
673 : * __device_add_disk() was used.
674 : *
675 : * Drivers exist which depend on the release of the gendisk to be synchronous,
676 : * it should not be deferred.
677 : *
678 : * Context: can sleep
679 : */
680 0 : void del_gendisk(struct gendisk *disk)
681 : {
682 0 : struct disk_part_iter piter;
683 0 : struct block_device *part;
684 :
685 0 : might_sleep();
686 :
687 0 : if (WARN_ON_ONCE(!disk->queue))
688 0 : return;
689 :
690 0 : blk_integrity_del(disk);
691 0 : disk_del_events(disk);
692 :
693 : /*
694 : * Block lookups of the disk until all bdevs are unhashed and the
695 : * disk is marked as dead (GENHD_FL_UP cleared).
696 : */
697 0 : down_write(&bdev_lookup_sem);
698 :
699 : /* invalidate stuff */
700 0 : disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
701 0 : while ((part = disk_part_iter_next(&piter))) {
702 0 : invalidate_partition(part);
703 0 : delete_partition(part);
704 : }
705 0 : disk_part_iter_exit(&piter);
706 :
707 0 : invalidate_partition(disk->part0);
708 0 : set_capacity(disk, 0);
709 0 : disk->flags &= ~GENHD_FL_UP;
710 0 : up_write(&bdev_lookup_sem);
711 :
712 0 : if (!(disk->flags & GENHD_FL_HIDDEN)) {
713 0 : sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
714 :
715 : /*
716 : * Unregister bdi before releasing device numbers (as they can
717 : * get reused and we'd get clashes in sysfs).
718 : */
719 0 : bdi_unregister(disk->queue->backing_dev_info);
720 : }
721 :
722 0 : blk_unregister_queue(disk);
723 :
724 0 : kobject_put(disk->part0->bd_holder_dir);
725 0 : kobject_put(disk->slave_dir);
726 :
727 0 : part_stat_set_all(disk->part0, 0);
728 0 : disk->part0->bd_stamp = 0;
729 0 : if (!sysfs_deprecated)
730 0 : sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
731 0 : pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
732 0 : device_del(disk_to_dev(disk));
733 : }
734 : EXPORT_SYMBOL(del_gendisk);
735 :
736 : /* sysfs access to bad-blocks list. */
737 0 : static ssize_t disk_badblocks_show(struct device *dev,
738 : struct device_attribute *attr,
739 : char *page)
740 : {
741 0 : struct gendisk *disk = dev_to_disk(dev);
742 :
743 0 : if (!disk->bb)
744 0 : return sprintf(page, "\n");
745 :
746 0 : return badblocks_show(disk->bb, page, 0);
747 : }
748 :
749 0 : static ssize_t disk_badblocks_store(struct device *dev,
750 : struct device_attribute *attr,
751 : const char *page, size_t len)
752 : {
753 0 : struct gendisk *disk = dev_to_disk(dev);
754 :
755 0 : if (!disk->bb)
756 : return -ENXIO;
757 :
758 0 : return badblocks_store(disk->bb, page, len, 0);
759 : }
760 :
761 0 : void blk_request_module(dev_t devt)
762 : {
763 0 : unsigned int major = MAJOR(devt);
764 0 : struct blk_major_name **n;
765 :
766 0 : mutex_lock(&major_names_lock);
767 0 : for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) {
768 0 : if ((*n)->major == major && (*n)->probe) {
769 0 : (*n)->probe(devt);
770 0 : mutex_unlock(&major_names_lock);
771 0 : return;
772 : }
773 : }
774 0 : mutex_unlock(&major_names_lock);
775 :
776 0 : if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
777 : /* Make old-style 2.4 aliases work */
778 0 : request_module("block-major-%d", MAJOR(devt));
779 : }
780 :
781 : /**
782 : * bdget_disk - do bdget() by gendisk and partition number
783 : * @disk: gendisk of interest
784 : * @partno: partition number
785 : *
786 : * Find partition @partno from @disk, do bdget() on it.
787 : *
788 : * CONTEXT:
789 : * Don't care.
790 : *
791 : * RETURNS:
792 : * Resulting block_device on success, NULL on failure.
793 : */
794 0 : struct block_device *bdget_disk(struct gendisk *disk, int partno)
795 : {
796 0 : struct block_device *bdev = NULL;
797 :
798 0 : rcu_read_lock();
799 0 : bdev = xa_load(&disk->part_tbl, partno);
800 0 : if (bdev && !bdgrab(bdev))
801 0 : bdev = NULL;
802 0 : rcu_read_unlock();
803 :
804 0 : return bdev;
805 : }
806 :
807 : /*
808 : * print a full list of all partitions - intended for places where the root
809 : * filesystem can't be mounted and thus to give the victim some idea of what
810 : * went wrong
811 : */
812 0 : void __init printk_all_partitions(void)
813 : {
814 0 : struct class_dev_iter iter;
815 0 : struct device *dev;
816 :
817 0 : class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
818 0 : while ((dev = class_dev_iter_next(&iter))) {
819 0 : struct gendisk *disk = dev_to_disk(dev);
820 0 : struct disk_part_iter piter;
821 0 : struct block_device *part;
822 0 : char name_buf[BDEVNAME_SIZE];
823 0 : char devt_buf[BDEVT_SIZE];
824 :
825 : /*
826 : * Don't show empty devices or things that have been
827 : * suppressed
828 : */
829 0 : if (get_capacity(disk) == 0 ||
830 0 : (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
831 0 : continue;
832 :
833 : /*
834 : * Note, unlike /proc/partitions, I am showing the
835 : * numbers in hex - the same format as the root=
836 : * option takes.
837 : */
838 0 : disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
839 0 : while ((part = disk_part_iter_next(&piter))) {
840 0 : bool is_part0 = part == disk->part0;
841 :
842 0 : printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
843 : bdevt_str(part->bd_dev, devt_buf),
844 0 : bdev_nr_sectors(part) >> 1,
845 0 : disk_name(disk, part->bd_partno, name_buf),
846 0 : part->bd_meta_info ?
847 : part->bd_meta_info->uuid : "");
848 0 : if (is_part0) {
849 0 : if (dev->parent && dev->parent->driver)
850 0 : printk(" driver: %s\n",
851 : dev->parent->driver->name);
852 : else
853 0 : printk(" (driver?)\n");
854 : } else
855 0 : printk("\n");
856 : }
857 0 : disk_part_iter_exit(&piter);
858 : }
859 0 : class_dev_iter_exit(&iter);
860 0 : }
861 :
862 : #ifdef CONFIG_PROC_FS
863 : /* iterator */
864 2 : static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
865 : {
866 2 : loff_t skip = *pos;
867 2 : struct class_dev_iter *iter;
868 2 : struct device *dev;
869 :
870 2 : iter = kmalloc(sizeof(*iter), GFP_KERNEL);
871 2 : if (!iter)
872 2 : return ERR_PTR(-ENOMEM);
873 :
874 2 : seqf->private = iter;
875 2 : class_dev_iter_init(iter, &block_class, NULL, &disk_type);
876 11 : do {
877 11 : dev = class_dev_iter_next(iter);
878 11 : if (!dev)
879 : return NULL;
880 10 : } while (skip--);
881 :
882 1 : return dev_to_disk(dev);
883 : }
884 :
885 9 : static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
886 : {
887 9 : struct device *dev;
888 :
889 9 : (*pos)++;
890 9 : dev = class_dev_iter_next(seqf->private);
891 9 : if (dev)
892 8 : return dev_to_disk(dev);
893 :
894 : return NULL;
895 : }
896 :
897 2 : static void disk_seqf_stop(struct seq_file *seqf, void *v)
898 : {
899 2 : struct class_dev_iter *iter = seqf->private;
900 :
901 : /* stop is called even after start failed :-( */
902 2 : if (iter) {
903 2 : class_dev_iter_exit(iter);
904 2 : kfree(iter);
905 2 : seqf->private = NULL;
906 : }
907 2 : }
908 :
909 2 : static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
910 : {
911 2 : void *p;
912 :
913 2 : p = disk_seqf_start(seqf, pos);
914 3 : if (!IS_ERR_OR_NULL(p) && !*pos)
915 1 : seq_puts(seqf, "major minor #blocks name\n\n");
916 2 : return p;
917 : }
918 :
919 9 : static int show_partition(struct seq_file *seqf, void *v)
920 : {
921 9 : struct gendisk *sgp = v;
922 9 : struct disk_part_iter piter;
923 9 : struct block_device *part;
924 9 : char buf[BDEVNAME_SIZE];
925 :
926 : /* Don't show non-partitionable removeable devices or empty devices */
927 9 : if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
928 0 : (sgp->flags & GENHD_FL_REMOVABLE)))
929 : return 0;
930 1 : if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
931 : return 0;
932 :
933 : /* show the full disk and all non-0 size partitions of it */
934 1 : disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
935 3 : while ((part = disk_part_iter_next(&piter)))
936 2 : seq_printf(seqf, "%4d %7d %10llu %s\n",
937 2 : MAJOR(part->bd_dev), MINOR(part->bd_dev),
938 2 : bdev_nr_sectors(part) >> 1,
939 2 : disk_name(sgp, part->bd_partno, buf));
940 1 : disk_part_iter_exit(&piter);
941 :
942 1 : return 0;
943 : }
944 :
945 : static const struct seq_operations partitions_op = {
946 : .start = show_partition_start,
947 : .next = disk_seqf_next,
948 : .stop = disk_seqf_stop,
949 : .show = show_partition
950 : };
951 : #endif
952 :
953 1 : static int __init genhd_device_init(void)
954 : {
955 1 : int error;
956 :
957 1 : block_class.dev_kobj = sysfs_dev_block_kobj;
958 1 : error = class_register(&block_class);
959 1 : if (unlikely(error))
960 : return error;
961 1 : blk_dev_init();
962 :
963 1 : register_blkdev(BLOCK_EXT_MAJOR, "blkext");
964 :
965 : /* create top-level block dir */
966 1 : if (!sysfs_deprecated)
967 1 : block_depr = kobject_create_and_add("block", NULL);
968 1 : return 0;
969 : }
970 :
971 : subsys_initcall(genhd_device_init);
972 :
973 0 : static ssize_t disk_range_show(struct device *dev,
974 : struct device_attribute *attr, char *buf)
975 : {
976 0 : struct gendisk *disk = dev_to_disk(dev);
977 :
978 0 : return sprintf(buf, "%d\n", disk->minors);
979 : }
980 :
981 0 : static ssize_t disk_ext_range_show(struct device *dev,
982 : struct device_attribute *attr, char *buf)
983 : {
984 0 : struct gendisk *disk = dev_to_disk(dev);
985 :
986 0 : return sprintf(buf, "%d\n", disk_max_parts(disk));
987 : }
988 :
989 0 : static ssize_t disk_removable_show(struct device *dev,
990 : struct device_attribute *attr, char *buf)
991 : {
992 0 : struct gendisk *disk = dev_to_disk(dev);
993 :
994 0 : return sprintf(buf, "%d\n",
995 0 : (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
996 : }
997 :
998 0 : static ssize_t disk_hidden_show(struct device *dev,
999 : struct device_attribute *attr, char *buf)
1000 : {
1001 0 : struct gendisk *disk = dev_to_disk(dev);
1002 :
1003 0 : return sprintf(buf, "%d\n",
1004 0 : (disk->flags & GENHD_FL_HIDDEN ? 1 : 0));
1005 : }
1006 :
1007 0 : static ssize_t disk_ro_show(struct device *dev,
1008 : struct device_attribute *attr, char *buf)
1009 : {
1010 0 : struct gendisk *disk = dev_to_disk(dev);
1011 :
1012 0 : return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
1013 : }
1014 :
1015 3 : ssize_t part_size_show(struct device *dev,
1016 : struct device_attribute *attr, char *buf)
1017 : {
1018 3 : return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev)));
1019 : }
1020 :
1021 0 : ssize_t part_stat_show(struct device *dev,
1022 : struct device_attribute *attr, char *buf)
1023 : {
1024 0 : struct block_device *bdev = dev_to_bdev(dev);
1025 0 : struct request_queue *q = bdev->bd_disk->queue;
1026 0 : struct disk_stats stat;
1027 0 : unsigned int inflight;
1028 :
1029 0 : part_stat_read_all(bdev, &stat);
1030 0 : if (queue_is_mq(q))
1031 0 : inflight = blk_mq_in_flight(q, bdev);
1032 : else
1033 0 : inflight = part_in_flight(bdev);
1034 :
1035 0 : return sprintf(buf,
1036 : "%8lu %8lu %8llu %8u "
1037 : "%8lu %8lu %8llu %8u "
1038 : "%8u %8u %8u "
1039 : "%8lu %8lu %8llu %8u "
1040 : "%8lu %8u"
1041 : "\n",
1042 : stat.ios[STAT_READ],
1043 : stat.merges[STAT_READ],
1044 0 : (unsigned long long)stat.sectors[STAT_READ],
1045 0 : (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC),
1046 : stat.ios[STAT_WRITE],
1047 : stat.merges[STAT_WRITE],
1048 0 : (unsigned long long)stat.sectors[STAT_WRITE],
1049 0 : (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC),
1050 : inflight,
1051 : jiffies_to_msecs(stat.io_ticks),
1052 0 : (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1053 0 : stat.nsecs[STAT_WRITE] +
1054 : stat.nsecs[STAT_DISCARD] +
1055 : stat.nsecs[STAT_FLUSH],
1056 : NSEC_PER_MSEC),
1057 : stat.ios[STAT_DISCARD],
1058 : stat.merges[STAT_DISCARD],
1059 0 : (unsigned long long)stat.sectors[STAT_DISCARD],
1060 0 : (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC),
1061 : stat.ios[STAT_FLUSH],
1062 0 : (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC));
1063 : }
1064 :
1065 0 : ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
1066 : char *buf)
1067 : {
1068 0 : struct block_device *bdev = dev_to_bdev(dev);
1069 0 : struct request_queue *q = bdev->bd_disk->queue;
1070 0 : unsigned int inflight[2];
1071 :
1072 0 : if (queue_is_mq(q))
1073 0 : blk_mq_in_flight_rw(q, bdev, inflight);
1074 : else
1075 0 : part_in_flight_rw(bdev, inflight);
1076 :
1077 0 : return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
1078 : }
1079 :
1080 0 : static ssize_t disk_capability_show(struct device *dev,
1081 : struct device_attribute *attr, char *buf)
1082 : {
1083 0 : struct gendisk *disk = dev_to_disk(dev);
1084 :
1085 0 : return sprintf(buf, "%x\n", disk->flags);
1086 : }
1087 :
1088 0 : static ssize_t disk_alignment_offset_show(struct device *dev,
1089 : struct device_attribute *attr,
1090 : char *buf)
1091 : {
1092 0 : struct gendisk *disk = dev_to_disk(dev);
1093 :
1094 0 : return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
1095 : }
1096 :
1097 0 : static ssize_t disk_discard_alignment_show(struct device *dev,
1098 : struct device_attribute *attr,
1099 : char *buf)
1100 : {
1101 0 : struct gendisk *disk = dev_to_disk(dev);
1102 :
1103 0 : return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
1104 : }
1105 :
1106 : static DEVICE_ATTR(range, 0444, disk_range_show, NULL);
1107 : static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL);
1108 : static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL);
1109 : static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL);
1110 : static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL);
1111 : static DEVICE_ATTR(size, 0444, part_size_show, NULL);
1112 : static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL);
1113 : static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL);
1114 : static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL);
1115 : static DEVICE_ATTR(stat, 0444, part_stat_show, NULL);
1116 : static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL);
1117 : static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store);
1118 :
1119 : #ifdef CONFIG_FAIL_MAKE_REQUEST
1120 : ssize_t part_fail_show(struct device *dev,
1121 : struct device_attribute *attr, char *buf)
1122 : {
1123 : return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail);
1124 : }
1125 :
1126 : ssize_t part_fail_store(struct device *dev,
1127 : struct device_attribute *attr,
1128 : const char *buf, size_t count)
1129 : {
1130 : int i;
1131 :
1132 : if (count > 0 && sscanf(buf, "%d", &i) > 0)
1133 : dev_to_bdev(dev)->bd_make_it_fail = i;
1134 :
1135 : return count;
1136 : }
1137 :
1138 : static struct device_attribute dev_attr_fail =
1139 : __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store);
1140 : #endif /* CONFIG_FAIL_MAKE_REQUEST */
1141 :
1142 : #ifdef CONFIG_FAIL_IO_TIMEOUT
1143 : static struct device_attribute dev_attr_fail_timeout =
1144 : __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store);
1145 : #endif
1146 :
1147 : static struct attribute *disk_attrs[] = {
1148 : &dev_attr_range.attr,
1149 : &dev_attr_ext_range.attr,
1150 : &dev_attr_removable.attr,
1151 : &dev_attr_hidden.attr,
1152 : &dev_attr_ro.attr,
1153 : &dev_attr_size.attr,
1154 : &dev_attr_alignment_offset.attr,
1155 : &dev_attr_discard_alignment.attr,
1156 : &dev_attr_capability.attr,
1157 : &dev_attr_stat.attr,
1158 : &dev_attr_inflight.attr,
1159 : &dev_attr_badblocks.attr,
1160 : #ifdef CONFIG_FAIL_MAKE_REQUEST
1161 : &dev_attr_fail.attr,
1162 : #endif
1163 : #ifdef CONFIG_FAIL_IO_TIMEOUT
1164 : &dev_attr_fail_timeout.attr,
1165 : #endif
1166 : NULL
1167 : };
1168 :
1169 108 : static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n)
1170 : {
1171 108 : struct device *dev = container_of(kobj, typeof(*dev), kobj);
1172 108 : struct gendisk *disk = dev_to_disk(dev);
1173 :
1174 108 : if (a == &dev_attr_badblocks.attr && !disk->bb)
1175 : return 0;
1176 99 : return a->mode;
1177 : }
1178 :
1179 : static struct attribute_group disk_attr_group = {
1180 : .attrs = disk_attrs,
1181 : .is_visible = disk_visible,
1182 : };
1183 :
1184 : static const struct attribute_group *disk_attr_groups[] = {
1185 : &disk_attr_group,
1186 : NULL
1187 : };
1188 :
1189 : /**
1190 : * disk_release - releases all allocated resources of the gendisk
1191 : * @dev: the device representing this disk
1192 : *
1193 : * This function releases all allocated resources of the gendisk.
1194 : *
1195 : * Drivers which used __device_add_disk() have a gendisk with a request_queue
1196 : * assigned. Since the request_queue sits on top of the gendisk for these
1197 : * drivers we also call blk_put_queue() for them, and we expect the
1198 : * request_queue refcount to reach 0 at this point, and so the request_queue
1199 : * will also be freed prior to the disk.
1200 : *
1201 : * Context: can sleep
1202 : */
1203 0 : static void disk_release(struct device *dev)
1204 : {
1205 0 : struct gendisk *disk = dev_to_disk(dev);
1206 :
1207 0 : might_sleep();
1208 :
1209 0 : blk_free_devt(dev->devt);
1210 0 : disk_release_events(disk);
1211 0 : kfree(disk->random);
1212 0 : xa_destroy(&disk->part_tbl);
1213 0 : bdput(disk->part0);
1214 0 : if (disk->queue)
1215 0 : blk_put_queue(disk->queue);
1216 0 : kfree(disk);
1217 0 : }
1218 : struct class block_class = {
1219 : .name = "block",
1220 : };
1221 :
1222 56 : static char *block_devnode(struct device *dev, umode_t *mode,
1223 : kuid_t *uid, kgid_t *gid)
1224 : {
1225 56 : struct gendisk *disk = dev_to_disk(dev);
1226 :
1227 56 : 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 0 : struct disk_part_iter piter;
1251 0 : struct block_device *hd;
1252 0 : char buf[BDEVNAME_SIZE];
1253 0 : unsigned int inflight;
1254 0 : struct disk_stats stat;
1255 :
1256 : /*
1257 : if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1258 : seq_puts(seqf, "major minor name"
1259 : " rio rmerge rsect ruse wio wmerge "
1260 : "wsect wuse running use aveq"
1261 : "\n\n");
1262 : */
1263 :
1264 0 : disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
1265 0 : while ((hd = disk_part_iter_next(&piter))) {
1266 0 : part_stat_read_all(hd, &stat);
1267 0 : if (queue_is_mq(gp->queue))
1268 0 : inflight = blk_mq_in_flight(gp->queue, hd);
1269 : else
1270 0 : inflight = part_in_flight(hd);
1271 :
1272 0 : seq_printf(seqf, "%4d %7d %s "
1273 : "%lu %lu %lu %u "
1274 : "%lu %lu %lu %u "
1275 : "%u %u %u "
1276 : "%lu %lu %lu %u "
1277 : "%lu %u"
1278 : "\n",
1279 0 : MAJOR(hd->bd_dev), MINOR(hd->bd_dev),
1280 0 : disk_name(gp, hd->bd_partno, buf),
1281 : stat.ios[STAT_READ],
1282 : stat.merges[STAT_READ],
1283 : stat.sectors[STAT_READ],
1284 0 : (unsigned int)div_u64(stat.nsecs[STAT_READ],
1285 : NSEC_PER_MSEC),
1286 : stat.ios[STAT_WRITE],
1287 : stat.merges[STAT_WRITE],
1288 : stat.sectors[STAT_WRITE],
1289 0 : (unsigned int)div_u64(stat.nsecs[STAT_WRITE],
1290 : NSEC_PER_MSEC),
1291 : inflight,
1292 : jiffies_to_msecs(stat.io_ticks),
1293 0 : (unsigned int)div_u64(stat.nsecs[STAT_READ] +
1294 0 : stat.nsecs[STAT_WRITE] +
1295 : stat.nsecs[STAT_DISCARD] +
1296 : stat.nsecs[STAT_FLUSH],
1297 : NSEC_PER_MSEC),
1298 : stat.ios[STAT_DISCARD],
1299 : stat.merges[STAT_DISCARD],
1300 : stat.sectors[STAT_DISCARD],
1301 0 : (unsigned int)div_u64(stat.nsecs[STAT_DISCARD],
1302 : NSEC_PER_MSEC),
1303 : stat.ios[STAT_FLUSH],
1304 0 : (unsigned int)div_u64(stat.nsecs[STAT_FLUSH],
1305 : NSEC_PER_MSEC)
1306 : );
1307 : }
1308 0 : disk_part_iter_exit(&piter);
1309 :
1310 0 : return 0;
1311 : }
1312 :
1313 : static const struct seq_operations diskstats_op = {
1314 : .start = disk_seqf_start,
1315 : .next = disk_seqf_next,
1316 : .stop = disk_seqf_stop,
1317 : .show = diskstats_show
1318 : };
1319 :
1320 1 : static int __init proc_genhd_init(void)
1321 : {
1322 1 : proc_create_seq("diskstats", 0, NULL, &diskstats_op);
1323 1 : proc_create_seq("partitions", 0, NULL, &partitions_op);
1324 1 : return 0;
1325 : }
1326 : module_init(proc_genhd_init);
1327 : #endif /* CONFIG_PROC_FS */
1328 :
1329 2 : dev_t blk_lookup_devt(const char *name, int partno)
1330 : {
1331 2 : dev_t devt = MKDEV(0, 0);
1332 2 : struct class_dev_iter iter;
1333 2 : struct device *dev;
1334 :
1335 2 : class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1336 19 : while ((dev = class_dev_iter_next(&iter))) {
1337 18 : struct gendisk *disk = dev_to_disk(dev);
1338 18 : struct block_device *part;
1339 :
1340 36 : if (strcmp(dev_name(dev), name))
1341 17 : continue;
1342 :
1343 1 : if (partno < disk->minors) {
1344 : /* We need to return the right devno, even
1345 : * if the partition doesn't exist yet.
1346 : */
1347 1 : devt = MKDEV(MAJOR(dev->devt),
1348 : MINOR(dev->devt) + partno);
1349 1 : break;
1350 : }
1351 0 : part = bdget_disk(disk, partno);
1352 0 : if (part) {
1353 0 : devt = part->bd_dev;
1354 0 : bdput(part);
1355 0 : break;
1356 : }
1357 : }
1358 2 : class_dev_iter_exit(&iter);
1359 2 : return devt;
1360 : }
1361 :
1362 9 : struct gendisk *__alloc_disk_node(int minors, int node_id)
1363 : {
1364 9 : struct gendisk *disk;
1365 :
1366 9 : if (minors > DISK_MAX_PARTS) {
1367 0 : printk(KERN_ERR
1368 : "block: can't allocate more than %d partitions\n",
1369 : DISK_MAX_PARTS);
1370 0 : minors = DISK_MAX_PARTS;
1371 : }
1372 :
1373 9 : disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
1374 9 : if (!disk)
1375 : return NULL;
1376 :
1377 9 : disk->part0 = bdev_alloc(disk, 0);
1378 9 : if (!disk->part0)
1379 0 : goto out_free_disk;
1380 :
1381 9 : disk->node_id = node_id;
1382 9 : xa_init(&disk->part_tbl);
1383 9 : if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL))
1384 0 : goto out_destroy_part_tbl;
1385 :
1386 9 : disk->minors = minors;
1387 9 : rand_initialize_disk(disk);
1388 9 : disk_to_dev(disk)->class = &block_class;
1389 9 : disk_to_dev(disk)->type = &disk_type;
1390 9 : device_initialize(disk_to_dev(disk));
1391 9 : return disk;
1392 :
1393 0 : out_destroy_part_tbl:
1394 0 : xa_destroy(&disk->part_tbl);
1395 0 : bdput(disk->part0);
1396 0 : out_free_disk:
1397 0 : kfree(disk);
1398 0 : return NULL;
1399 : }
1400 : EXPORT_SYMBOL(__alloc_disk_node);
1401 :
1402 : /**
1403 : * put_disk - decrements the gendisk refcount
1404 : * @disk: the struct gendisk to decrement the refcount for
1405 : *
1406 : * This decrements the refcount for the struct gendisk. When this reaches 0
1407 : * we'll have disk_release() called.
1408 : *
1409 : * Context: Any context, but the last reference must not be dropped from
1410 : * atomic context.
1411 : */
1412 31 : void put_disk(struct gendisk *disk)
1413 : {
1414 31 : if (disk)
1415 31 : put_device(disk_to_dev(disk));
1416 31 : }
1417 : EXPORT_SYMBOL(put_disk);
1418 :
1419 0 : static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1420 : {
1421 0 : char event[] = "DISK_RO=1";
1422 0 : char *envp[] = { event, NULL };
1423 :
1424 0 : if (!ro)
1425 0 : event[8] = '0';
1426 0 : kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1427 0 : }
1428 :
1429 : /**
1430 : * set_disk_ro - set a gendisk read-only
1431 : * @disk: gendisk to operate on
1432 : * @read_only: %true to set the disk read-only, %false set the disk read/write
1433 : *
1434 : * This function is used to indicate whether a given disk device should have its
1435 : * read-only flag set. set_disk_ro() is typically used by device drivers to
1436 : * indicate whether the underlying physical device is write-protected.
1437 : */
1438 0 : void set_disk_ro(struct gendisk *disk, bool read_only)
1439 : {
1440 0 : if (read_only) {
1441 0 : if (test_and_set_bit(GD_READ_ONLY, &disk->state))
1442 : return;
1443 : } else {
1444 0 : if (!test_and_clear_bit(GD_READ_ONLY, &disk->state))
1445 : return;
1446 : }
1447 0 : set_disk_ro_uevent(disk, read_only);
1448 : }
1449 : EXPORT_SYMBOL(set_disk_ro);
1450 :
1451 3079 : int bdev_read_only(struct block_device *bdev)
1452 : {
1453 3079 : return bdev->bd_read_only || get_disk_ro(bdev->bd_disk);
1454 : }
1455 : EXPORT_SYMBOL(bdev_read_only);
1456 :
1457 : /*
1458 : * Disk events - monitor disk events like media change and eject request.
1459 : */
1460 : struct disk_events {
1461 : struct list_head node; /* all disk_event's */
1462 : struct gendisk *disk; /* the associated disk */
1463 : spinlock_t lock;
1464 :
1465 : struct mutex block_mutex; /* protects blocking */
1466 : int block; /* event blocking depth */
1467 : unsigned int pending; /* events already sent out */
1468 : unsigned int clearing; /* events being cleared */
1469 :
1470 : long poll_msecs; /* interval, -1 for default */
1471 : struct delayed_work dwork;
1472 : };
1473 :
1474 : static const char *disk_events_strs[] = {
1475 : [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1476 : [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1477 : };
1478 :
1479 : static char *disk_uevents[] = {
1480 : [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1481 : [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1482 : };
1483 :
1484 : /* list of all disk_events */
1485 : static DEFINE_MUTEX(disk_events_mutex);
1486 : static LIST_HEAD(disk_events);
1487 :
1488 : /* disable in-kernel polling by default */
1489 : static unsigned long disk_events_dfl_poll_msecs;
1490 :
1491 0 : static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1492 : {
1493 0 : struct disk_events *ev = disk->ev;
1494 0 : long intv_msecs = 0;
1495 :
1496 : /*
1497 : * If device-specific poll interval is set, always use it. If
1498 : * the default is being used, poll if the POLL flag is set.
1499 : */
1500 0 : if (ev->poll_msecs >= 0)
1501 : intv_msecs = ev->poll_msecs;
1502 0 : else if (disk->event_flags & DISK_EVENT_FLAG_POLL)
1503 0 : intv_msecs = disk_events_dfl_poll_msecs;
1504 :
1505 0 : return msecs_to_jiffies(intv_msecs);
1506 : }
1507 :
1508 : /**
1509 : * disk_block_events - block and flush disk event checking
1510 : * @disk: disk to block events for
1511 : *
1512 : * On return from this function, it is guaranteed that event checking
1513 : * isn't in progress and won't happen until unblocked by
1514 : * disk_unblock_events(). Events blocking is counted and the actual
1515 : * unblocking happens after the matching number of unblocks are done.
1516 : *
1517 : * Note that this intentionally does not block event checking from
1518 : * disk_clear_events().
1519 : *
1520 : * CONTEXT:
1521 : * Might sleep.
1522 : */
1523 31 : void disk_block_events(struct gendisk *disk)
1524 : {
1525 31 : struct disk_events *ev = disk->ev;
1526 31 : unsigned long flags;
1527 31 : bool cancel;
1528 :
1529 31 : if (!ev)
1530 : return;
1531 :
1532 : /*
1533 : * Outer mutex ensures that the first blocker completes canceling
1534 : * the event work before further blockers are allowed to finish.
1535 : */
1536 0 : mutex_lock(&ev->block_mutex);
1537 :
1538 0 : spin_lock_irqsave(&ev->lock, flags);
1539 0 : cancel = !ev->block++;
1540 0 : spin_unlock_irqrestore(&ev->lock, flags);
1541 :
1542 0 : if (cancel)
1543 0 : cancel_delayed_work_sync(&disk->ev->dwork);
1544 :
1545 0 : mutex_unlock(&ev->block_mutex);
1546 : }
1547 :
1548 0 : static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1549 : {
1550 0 : struct disk_events *ev = disk->ev;
1551 0 : unsigned long intv;
1552 0 : unsigned long flags;
1553 :
1554 0 : spin_lock_irqsave(&ev->lock, flags);
1555 :
1556 0 : if (WARN_ON_ONCE(ev->block <= 0))
1557 0 : goto out_unlock;
1558 :
1559 0 : if (--ev->block)
1560 0 : goto out_unlock;
1561 :
1562 0 : intv = disk_events_poll_jiffies(disk);
1563 0 : if (check_now)
1564 0 : queue_delayed_work(system_freezable_power_efficient_wq,
1565 : &ev->dwork, 0);
1566 0 : else if (intv)
1567 0 : queue_delayed_work(system_freezable_power_efficient_wq,
1568 : &ev->dwork, intv);
1569 0 : out_unlock:
1570 0 : spin_unlock_irqrestore(&ev->lock, flags);
1571 0 : }
1572 :
1573 : /**
1574 : * disk_unblock_events - unblock disk event checking
1575 : * @disk: disk to unblock events for
1576 : *
1577 : * Undo disk_block_events(). When the block count reaches zero, it
1578 : * starts events polling if configured.
1579 : *
1580 : * CONTEXT:
1581 : * Don't care. Safe to call from irq context.
1582 : */
1583 31 : void disk_unblock_events(struct gendisk *disk)
1584 : {
1585 31 : if (disk->ev)
1586 0 : __disk_unblock_events(disk, false);
1587 31 : }
1588 :
1589 : /**
1590 : * disk_flush_events - schedule immediate event checking and flushing
1591 : * @disk: disk to check and flush events for
1592 : * @mask: events to flush
1593 : *
1594 : * Schedule immediate event checking on @disk if not blocked. Events in
1595 : * @mask are scheduled to be cleared from the driver. Note that this
1596 : * doesn't clear the events from @disk->ev.
1597 : *
1598 : * CONTEXT:
1599 : * If @mask is non-zero must be called with bdev->bd_mutex held.
1600 : */
1601 30 : void disk_flush_events(struct gendisk *disk, unsigned int mask)
1602 : {
1603 30 : struct disk_events *ev = disk->ev;
1604 :
1605 30 : if (!ev)
1606 : return;
1607 :
1608 0 : spin_lock_irq(&ev->lock);
1609 0 : ev->clearing |= mask;
1610 0 : if (!ev->block)
1611 0 : mod_delayed_work(system_freezable_power_efficient_wq,
1612 : &ev->dwork, 0);
1613 0 : spin_unlock_irq(&ev->lock);
1614 : }
1615 :
1616 : /**
1617 : * disk_clear_events - synchronously check, clear and return pending events
1618 : * @disk: disk to fetch and clear events from
1619 : * @mask: mask of events to be fetched and cleared
1620 : *
1621 : * Disk events are synchronously checked and pending events in @mask
1622 : * are cleared and returned. This ignores the block count.
1623 : *
1624 : * CONTEXT:
1625 : * Might sleep.
1626 : */
1627 0 : static unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1628 : {
1629 0 : struct disk_events *ev = disk->ev;
1630 0 : unsigned int pending;
1631 0 : unsigned int clearing = mask;
1632 :
1633 0 : if (!ev)
1634 : return 0;
1635 :
1636 0 : disk_block_events(disk);
1637 :
1638 : /*
1639 : * store the union of mask and ev->clearing on the stack so that the
1640 : * race with disk_flush_events does not cause ambiguity (ev->clearing
1641 : * can still be modified even if events are blocked).
1642 : */
1643 0 : spin_lock_irq(&ev->lock);
1644 0 : clearing |= ev->clearing;
1645 0 : ev->clearing = 0;
1646 0 : spin_unlock_irq(&ev->lock);
1647 :
1648 0 : disk_check_events(ev, &clearing);
1649 : /*
1650 : * if ev->clearing is not 0, the disk_flush_events got called in the
1651 : * middle of this function, so we want to run the workfn without delay.
1652 : */
1653 0 : __disk_unblock_events(disk, ev->clearing ? true : false);
1654 :
1655 : /* then, fetch and clear pending events */
1656 0 : spin_lock_irq(&ev->lock);
1657 0 : pending = ev->pending & mask;
1658 0 : ev->pending &= ~mask;
1659 0 : spin_unlock_irq(&ev->lock);
1660 0 : WARN_ON_ONCE(clearing & mask);
1661 :
1662 : return pending;
1663 : }
1664 :
1665 : /**
1666 : * bdev_check_media_change - check if a removable media has been changed
1667 : * @bdev: block device to check
1668 : *
1669 : * Check whether a removable media has been changed, and attempt to free all
1670 : * dentries and inodes and invalidates all block device page cache entries in
1671 : * that case.
1672 : *
1673 : * Returns %true if the block device changed, or %false if not.
1674 : */
1675 0 : bool bdev_check_media_change(struct block_device *bdev)
1676 : {
1677 0 : unsigned int events;
1678 :
1679 0 : events = disk_clear_events(bdev->bd_disk, DISK_EVENT_MEDIA_CHANGE |
1680 : DISK_EVENT_EJECT_REQUEST);
1681 0 : if (!(events & DISK_EVENT_MEDIA_CHANGE))
1682 : return false;
1683 :
1684 0 : if (__invalidate_device(bdev, true))
1685 0 : pr_warn("VFS: busy inodes on changed media %s\n",
1686 : bdev->bd_disk->disk_name);
1687 0 : set_bit(GD_NEED_PART_SCAN, &bdev->bd_disk->state);
1688 0 : return true;
1689 : }
1690 : EXPORT_SYMBOL(bdev_check_media_change);
1691 :
1692 : /*
1693 : * Separate this part out so that a different pointer for clearing_ptr can be
1694 : * passed in for disk_clear_events.
1695 : */
1696 0 : static void disk_events_workfn(struct work_struct *work)
1697 : {
1698 0 : struct delayed_work *dwork = to_delayed_work(work);
1699 0 : struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
1700 :
1701 0 : disk_check_events(ev, &ev->clearing);
1702 0 : }
1703 :
1704 0 : static void disk_check_events(struct disk_events *ev,
1705 : unsigned int *clearing_ptr)
1706 : {
1707 0 : struct gendisk *disk = ev->disk;
1708 0 : char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
1709 0 : unsigned int clearing = *clearing_ptr;
1710 0 : unsigned int events;
1711 0 : unsigned long intv;
1712 0 : int nr_events = 0, i;
1713 :
1714 : /* check events */
1715 0 : events = disk->fops->check_events(disk, clearing);
1716 :
1717 : /* accumulate pending events and schedule next poll if necessary */
1718 0 : spin_lock_irq(&ev->lock);
1719 :
1720 0 : events &= ~ev->pending;
1721 0 : ev->pending |= events;
1722 0 : *clearing_ptr &= ~clearing;
1723 :
1724 0 : intv = disk_events_poll_jiffies(disk);
1725 0 : if (!ev->block && intv)
1726 0 : queue_delayed_work(system_freezable_power_efficient_wq,
1727 : &ev->dwork, intv);
1728 :
1729 0 : spin_unlock_irq(&ev->lock);
1730 :
1731 : /*
1732 : * Tell userland about new events. Only the events listed in
1733 : * @disk->events are reported, and only if DISK_EVENT_FLAG_UEVENT
1734 : * is set. Otherwise, events are processed internally but never
1735 : * get reported to userland.
1736 : */
1737 0 : for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
1738 0 : if ((events & disk->events & (1 << i)) &&
1739 0 : (disk->event_flags & DISK_EVENT_FLAG_UEVENT))
1740 0 : envp[nr_events++] = disk_uevents[i];
1741 :
1742 0 : if (nr_events)
1743 0 : kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1744 0 : }
1745 :
1746 : /*
1747 : * A disk events enabled device has the following sysfs nodes under
1748 : * its /sys/block/X/ directory.
1749 : *
1750 : * events : list of all supported events
1751 : * events_async : list of events which can be detected w/o polling
1752 : * (always empty, only for backwards compatibility)
1753 : * events_poll_msecs : polling interval, 0: disable, -1: system default
1754 : */
1755 0 : static ssize_t __disk_events_show(unsigned int events, char *buf)
1756 : {
1757 0 : const char *delim = "";
1758 0 : ssize_t pos = 0;
1759 0 : int i;
1760 :
1761 0 : for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1762 0 : if (events & (1 << i)) {
1763 0 : pos += sprintf(buf + pos, "%s%s",
1764 : delim, disk_events_strs[i]);
1765 0 : delim = " ";
1766 : }
1767 0 : if (pos)
1768 0 : pos += sprintf(buf + pos, "\n");
1769 0 : return pos;
1770 : }
1771 :
1772 0 : static ssize_t disk_events_show(struct device *dev,
1773 : struct device_attribute *attr, char *buf)
1774 : {
1775 0 : struct gendisk *disk = dev_to_disk(dev);
1776 :
1777 0 : if (!(disk->event_flags & DISK_EVENT_FLAG_UEVENT))
1778 : return 0;
1779 :
1780 0 : return __disk_events_show(disk->events, buf);
1781 : }
1782 :
1783 0 : static ssize_t disk_events_async_show(struct device *dev,
1784 : struct device_attribute *attr, char *buf)
1785 : {
1786 0 : return 0;
1787 : }
1788 :
1789 0 : static ssize_t disk_events_poll_msecs_show(struct device *dev,
1790 : struct device_attribute *attr,
1791 : char *buf)
1792 : {
1793 0 : struct gendisk *disk = dev_to_disk(dev);
1794 :
1795 0 : if (!disk->ev)
1796 0 : return sprintf(buf, "-1\n");
1797 :
1798 0 : return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1799 : }
1800 :
1801 0 : static ssize_t disk_events_poll_msecs_store(struct device *dev,
1802 : struct device_attribute *attr,
1803 : const char *buf, size_t count)
1804 : {
1805 0 : struct gendisk *disk = dev_to_disk(dev);
1806 0 : long intv;
1807 :
1808 0 : if (!count || !sscanf(buf, "%ld", &intv))
1809 0 : return -EINVAL;
1810 :
1811 0 : if (intv < 0 && intv != -1)
1812 : return -EINVAL;
1813 :
1814 0 : if (!disk->ev)
1815 : return -ENODEV;
1816 :
1817 0 : disk_block_events(disk);
1818 0 : disk->ev->poll_msecs = intv;
1819 0 : __disk_unblock_events(disk, true);
1820 :
1821 0 : return count;
1822 : }
1823 :
1824 : static const DEVICE_ATTR(events, 0444, disk_events_show, NULL);
1825 : static const DEVICE_ATTR(events_async, 0444, disk_events_async_show, NULL);
1826 : static const DEVICE_ATTR(events_poll_msecs, 0644,
1827 : disk_events_poll_msecs_show,
1828 : disk_events_poll_msecs_store);
1829 :
1830 : static const struct attribute *disk_events_attrs[] = {
1831 : &dev_attr_events.attr,
1832 : &dev_attr_events_async.attr,
1833 : &dev_attr_events_poll_msecs.attr,
1834 : NULL,
1835 : };
1836 :
1837 : /*
1838 : * The default polling interval can be specified by the kernel
1839 : * parameter block.events_dfl_poll_msecs which defaults to 0
1840 : * (disable). This can also be modified runtime by writing to
1841 : * /sys/module/block/parameters/events_dfl_poll_msecs.
1842 : */
1843 0 : static int disk_events_set_dfl_poll_msecs(const char *val,
1844 : const struct kernel_param *kp)
1845 : {
1846 0 : struct disk_events *ev;
1847 0 : int ret;
1848 :
1849 0 : ret = param_set_ulong(val, kp);
1850 0 : if (ret < 0)
1851 : return ret;
1852 :
1853 0 : mutex_lock(&disk_events_mutex);
1854 :
1855 0 : list_for_each_entry(ev, &disk_events, node)
1856 0 : disk_flush_events(ev->disk, 0);
1857 :
1858 0 : mutex_unlock(&disk_events_mutex);
1859 :
1860 0 : return 0;
1861 : }
1862 :
1863 : static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1864 : .set = disk_events_set_dfl_poll_msecs,
1865 : .get = param_get_ulong,
1866 : };
1867 :
1868 : #undef MODULE_PARAM_PREFIX
1869 : #define MODULE_PARAM_PREFIX "block."
1870 :
1871 : module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
1872 : &disk_events_dfl_poll_msecs, 0644);
1873 :
1874 : /*
1875 : * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
1876 : */
1877 9 : static void disk_alloc_events(struct gendisk *disk)
1878 : {
1879 9 : struct disk_events *ev;
1880 :
1881 9 : if (!disk->fops->check_events || !disk->events)
1882 : return;
1883 :
1884 0 : ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1885 0 : if (!ev) {
1886 0 : pr_warn("%s: failed to initialize events\n", disk->disk_name);
1887 0 : return;
1888 : }
1889 :
1890 0 : INIT_LIST_HEAD(&ev->node);
1891 0 : ev->disk = disk;
1892 0 : spin_lock_init(&ev->lock);
1893 0 : mutex_init(&ev->block_mutex);
1894 0 : ev->block = 1;
1895 0 : ev->poll_msecs = -1;
1896 0 : INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
1897 :
1898 0 : disk->ev = ev;
1899 : }
1900 :
1901 9 : static void disk_add_events(struct gendisk *disk)
1902 : {
1903 : /* FIXME: error handling */
1904 9 : if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0)
1905 0 : pr_warn("%s: failed to create sysfs files for events\n",
1906 : disk->disk_name);
1907 :
1908 9 : if (!disk->ev)
1909 : return;
1910 :
1911 0 : mutex_lock(&disk_events_mutex);
1912 0 : list_add_tail(&disk->ev->node, &disk_events);
1913 0 : mutex_unlock(&disk_events_mutex);
1914 :
1915 : /*
1916 : * Block count is initialized to 1 and the following initial
1917 : * unblock kicks it into action.
1918 : */
1919 0 : __disk_unblock_events(disk, true);
1920 : }
1921 :
1922 0 : static void disk_del_events(struct gendisk *disk)
1923 : {
1924 0 : if (disk->ev) {
1925 0 : disk_block_events(disk);
1926 :
1927 0 : mutex_lock(&disk_events_mutex);
1928 0 : list_del_init(&disk->ev->node);
1929 0 : mutex_unlock(&disk_events_mutex);
1930 : }
1931 :
1932 0 : sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
1933 0 : }
1934 :
1935 0 : static void disk_release_events(struct gendisk *disk)
1936 : {
1937 : /* the block count should be 1 from disk_del_events() */
1938 0 : WARN_ON_ONCE(disk->ev && disk->ev->block != 1);
1939 0 : kfree(disk->ev);
1940 0 : }
|
