]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * gendisk handling | |
3 | */ | |
4 | ||
5 | #include <linux/module.h> | |
6 | #include <linux/fs.h> | |
7 | #include <linux/genhd.h> | |
8 | #include <linux/kdev_t.h> | |
9 | #include <linux/kernel.h> | |
10 | #include <linux/blkdev.h> | |
11 | #include <linux/backing-dev.h> | |
12 | #include <linux/init.h> | |
13 | #include <linux/spinlock.h> | |
14 | #include <linux/proc_fs.h> | |
15 | #include <linux/seq_file.h> | |
16 | #include <linux/slab.h> | |
17 | #include <linux/kmod.h> | |
18 | #include <linux/kobj_map.h> | |
19 | #include <linux/mutex.h> | |
20 | #include <linux/idr.h> | |
21 | #include <linux/log2.h> | |
22 | #include <linux/pm_runtime.h> | |
23 | #include <linux/badblocks.h> | |
24 | ||
25 | #include "blk.h" | |
26 | ||
27 | static DEFINE_MUTEX(block_class_lock); | |
28 | struct kobject *block_depr; | |
29 | ||
30 | /* for extended dynamic devt allocation, currently only one major is used */ | |
31 | #define NR_EXT_DEVT (1 << MINORBITS) | |
32 | ||
33 | /* For extended devt allocation. ext_devt_lock prevents look up | |
34 | * results from going away underneath its user. | |
35 | */ | |
36 | static DEFINE_SPINLOCK(ext_devt_lock); | |
37 | static DEFINE_IDR(ext_devt_idr); | |
38 | ||
39 | static const struct device_type disk_type; | |
40 | ||
41 | static void disk_check_events(struct disk_events *ev, | |
42 | unsigned int *clearing_ptr); | |
43 | static void disk_alloc_events(struct gendisk *disk); | |
44 | static void disk_add_events(struct gendisk *disk); | |
45 | static void disk_del_events(struct gendisk *disk); | |
46 | static void disk_release_events(struct gendisk *disk); | |
47 | ||
48 | void part_inc_in_flight(struct request_queue *q, struct hd_struct *part, int rw) | |
49 | { | |
50 | if (q->mq_ops) | |
51 | return; | |
52 | ||
53 | atomic_inc(&part->in_flight[rw]); | |
54 | if (part->partno) | |
55 | atomic_inc(&part_to_disk(part)->part0.in_flight[rw]); | |
56 | } | |
57 | ||
58 | void part_dec_in_flight(struct request_queue *q, struct hd_struct *part, int rw) | |
59 | { | |
60 | if (q->mq_ops) | |
61 | return; | |
62 | ||
63 | atomic_dec(&part->in_flight[rw]); | |
64 | if (part->partno) | |
65 | atomic_dec(&part_to_disk(part)->part0.in_flight[rw]); | |
66 | } | |
67 | ||
68 | void part_in_flight(struct request_queue *q, struct hd_struct *part, | |
69 | unsigned int inflight[2]) | |
70 | { | |
71 | if (q->mq_ops) { | |
72 | blk_mq_in_flight(q, part, inflight); | |
73 | return; | |
74 | } | |
75 | ||
76 | inflight[0] = atomic_read(&part->in_flight[0]) + | |
77 | atomic_read(&part->in_flight[1]); | |
78 | if (part->partno) { | |
79 | part = &part_to_disk(part)->part0; | |
80 | inflight[1] = atomic_read(&part->in_flight[0]) + | |
81 | atomic_read(&part->in_flight[1]); | |
82 | } | |
83 | } | |
84 | ||
85 | struct hd_struct *__disk_get_part(struct gendisk *disk, int partno) | |
86 | { | |
87 | struct disk_part_tbl *ptbl = rcu_dereference(disk->part_tbl); | |
88 | ||
89 | if (unlikely(partno < 0 || partno >= ptbl->len)) | |
90 | return NULL; | |
91 | return rcu_dereference(ptbl->part[partno]); | |
92 | } | |
93 | ||
94 | /** | |
95 | * disk_get_part - get partition | |
96 | * @disk: disk to look partition from | |
97 | * @partno: partition number | |
98 | * | |
99 | * Look for partition @partno from @disk. If found, increment | |
100 | * reference count and return it. | |
101 | * | |
102 | * CONTEXT: | |
103 | * Don't care. | |
104 | * | |
105 | * RETURNS: | |
106 | * Pointer to the found partition on success, NULL if not found. | |
107 | */ | |
108 | struct hd_struct *disk_get_part(struct gendisk *disk, int partno) | |
109 | { | |
110 | struct hd_struct *part; | |
111 | ||
112 | rcu_read_lock(); | |
113 | part = __disk_get_part(disk, partno); | |
114 | if (part) | |
115 | get_device(part_to_dev(part)); | |
116 | rcu_read_unlock(); | |
117 | ||
118 | return part; | |
119 | } | |
120 | EXPORT_SYMBOL_GPL(disk_get_part); | |
121 | ||
122 | /** | |
123 | * disk_part_iter_init - initialize partition iterator | |
124 | * @piter: iterator to initialize | |
125 | * @disk: disk to iterate over | |
126 | * @flags: DISK_PITER_* flags | |
127 | * | |
128 | * Initialize @piter so that it iterates over partitions of @disk. | |
129 | * | |
130 | * CONTEXT: | |
131 | * Don't care. | |
132 | */ | |
133 | void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk, | |
134 | unsigned int flags) | |
135 | { | |
136 | struct disk_part_tbl *ptbl; | |
137 | ||
138 | rcu_read_lock(); | |
139 | ptbl = rcu_dereference(disk->part_tbl); | |
140 | ||
141 | piter->disk = disk; | |
142 | piter->part = NULL; | |
143 | ||
144 | if (flags & DISK_PITER_REVERSE) | |
145 | piter->idx = ptbl->len - 1; | |
146 | else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0)) | |
147 | piter->idx = 0; | |
148 | else | |
149 | piter->idx = 1; | |
150 | ||
151 | piter->flags = flags; | |
152 | ||
153 | rcu_read_unlock(); | |
154 | } | |
155 | EXPORT_SYMBOL_GPL(disk_part_iter_init); | |
156 | ||
157 | /** | |
158 | * disk_part_iter_next - proceed iterator to the next partition and return it | |
159 | * @piter: iterator of interest | |
160 | * | |
161 | * Proceed @piter to the next partition and return it. | |
162 | * | |
163 | * CONTEXT: | |
164 | * Don't care. | |
165 | */ | |
166 | struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter) | |
167 | { | |
168 | struct disk_part_tbl *ptbl; | |
169 | int inc, end; | |
170 | ||
171 | /* put the last partition */ | |
172 | disk_put_part(piter->part); | |
173 | piter->part = NULL; | |
174 | ||
175 | /* get part_tbl */ | |
176 | rcu_read_lock(); | |
177 | ptbl = rcu_dereference(piter->disk->part_tbl); | |
178 | ||
179 | /* determine iteration parameters */ | |
180 | if (piter->flags & DISK_PITER_REVERSE) { | |
181 | inc = -1; | |
182 | if (piter->flags & (DISK_PITER_INCL_PART0 | | |
183 | DISK_PITER_INCL_EMPTY_PART0)) | |
184 | end = -1; | |
185 | else | |
186 | end = 0; | |
187 | } else { | |
188 | inc = 1; | |
189 | end = ptbl->len; | |
190 | } | |
191 | ||
192 | /* iterate to the next partition */ | |
193 | for (; piter->idx != end; piter->idx += inc) { | |
194 | struct hd_struct *part; | |
195 | ||
196 | part = rcu_dereference(ptbl->part[piter->idx]); | |
197 | if (!part) | |
198 | continue; | |
199 | if (!part_nr_sects_read(part) && | |
200 | !(piter->flags & DISK_PITER_INCL_EMPTY) && | |
201 | !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 && | |
202 | piter->idx == 0)) | |
203 | continue; | |
204 | ||
205 | get_device(part_to_dev(part)); | |
206 | piter->part = part; | |
207 | piter->idx += inc; | |
208 | break; | |
209 | } | |
210 | ||
211 | rcu_read_unlock(); | |
212 | ||
213 | return piter->part; | |
214 | } | |
215 | EXPORT_SYMBOL_GPL(disk_part_iter_next); | |
216 | ||
217 | /** | |
218 | * disk_part_iter_exit - finish up partition iteration | |
219 | * @piter: iter of interest | |
220 | * | |
221 | * Called when iteration is over. Cleans up @piter. | |
222 | * | |
223 | * CONTEXT: | |
224 | * Don't care. | |
225 | */ | |
226 | void disk_part_iter_exit(struct disk_part_iter *piter) | |
227 | { | |
228 | disk_put_part(piter->part); | |
229 | piter->part = NULL; | |
230 | } | |
231 | EXPORT_SYMBOL_GPL(disk_part_iter_exit); | |
232 | ||
233 | static inline int sector_in_part(struct hd_struct *part, sector_t sector) | |
234 | { | |
235 | return part->start_sect <= sector && | |
236 | sector < part->start_sect + part_nr_sects_read(part); | |
237 | } | |
238 | ||
239 | /** | |
240 | * disk_map_sector_rcu - map sector to partition | |
241 | * @disk: gendisk of interest | |
242 | * @sector: sector to map | |
243 | * | |
244 | * Find out which partition @sector maps to on @disk. This is | |
245 | * primarily used for stats accounting. | |
246 | * | |
247 | * CONTEXT: | |
248 | * RCU read locked. The returned partition pointer is valid only | |
249 | * while preemption is disabled. | |
250 | * | |
251 | * RETURNS: | |
252 | * Found partition on success, part0 is returned if no partition matches | |
253 | */ | |
254 | struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector) | |
255 | { | |
256 | struct disk_part_tbl *ptbl; | |
257 | struct hd_struct *part; | |
258 | int i; | |
259 | ||
260 | ptbl = rcu_dereference(disk->part_tbl); | |
261 | ||
262 | part = rcu_dereference(ptbl->last_lookup); | |
263 | if (part && sector_in_part(part, sector)) | |
264 | return part; | |
265 | ||
266 | for (i = 1; i < ptbl->len; i++) { | |
267 | part = rcu_dereference(ptbl->part[i]); | |
268 | ||
269 | if (part && sector_in_part(part, sector)) { | |
270 | rcu_assign_pointer(ptbl->last_lookup, part); | |
271 | return part; | |
272 | } | |
273 | } | |
274 | return &disk->part0; | |
275 | } | |
276 | EXPORT_SYMBOL_GPL(disk_map_sector_rcu); | |
277 | ||
278 | /* | |
279 | * Can be deleted altogether. Later. | |
280 | * | |
281 | */ | |
282 | #define BLKDEV_MAJOR_HASH_SIZE 255 | |
283 | static struct blk_major_name { | |
284 | struct blk_major_name *next; | |
285 | int major; | |
286 | char name[16]; | |
287 | } *major_names[BLKDEV_MAJOR_HASH_SIZE]; | |
288 | ||
289 | /* index in the above - for now: assume no multimajor ranges */ | |
290 | static inline int major_to_index(unsigned major) | |
291 | { | |
292 | return major % BLKDEV_MAJOR_HASH_SIZE; | |
293 | } | |
294 | ||
295 | #ifdef CONFIG_PROC_FS | |
296 | void blkdev_show(struct seq_file *seqf, off_t offset) | |
297 | { | |
298 | struct blk_major_name *dp; | |
299 | ||
300 | mutex_lock(&block_class_lock); | |
301 | for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next) | |
302 | if (dp->major == offset) | |
303 | seq_printf(seqf, "%3d %s\n", dp->major, dp->name); | |
304 | mutex_unlock(&block_class_lock); | |
305 | } | |
306 | #endif /* CONFIG_PROC_FS */ | |
307 | ||
308 | /** | |
309 | * register_blkdev - register a new block device | |
310 | * | |
311 | * @major: the requested major device number [1..255]. If @major = 0, try to | |
312 | * allocate any unused major number. | |
313 | * @name: the name of the new block device as a zero terminated string | |
314 | * | |
315 | * The @name must be unique within the system. | |
316 | * | |
317 | * The return value depends on the @major input parameter: | |
318 | * | |
319 | * - if a major device number was requested in range [1..255] then the | |
320 | * function returns zero on success, or a negative error code | |
321 | * - if any unused major number was requested with @major = 0 parameter | |
322 | * then the return value is the allocated major number in range | |
323 | * [1..255] or a negative error code otherwise | |
324 | */ | |
325 | int register_blkdev(unsigned int major, const char *name) | |
326 | { | |
327 | struct blk_major_name **n, *p; | |
328 | int index, ret = 0; | |
329 | ||
330 | mutex_lock(&block_class_lock); | |
331 | ||
332 | /* temporary */ | |
333 | if (major == 0) { | |
334 | for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) { | |
335 | if (major_names[index] == NULL) | |
336 | break; | |
337 | } | |
338 | ||
339 | if (index == 0) { | |
340 | printk("register_blkdev: failed to get major for %s\n", | |
341 | name); | |
342 | ret = -EBUSY; | |
343 | goto out; | |
344 | } | |
345 | major = index; | |
346 | ret = major; | |
347 | } | |
348 | ||
349 | if (major >= BLKDEV_MAJOR_MAX) { | |
350 | pr_err("register_blkdev: major requested (%d) is greater than the maximum (%d) for %s\n", | |
351 | major, BLKDEV_MAJOR_MAX, name); | |
352 | ||
353 | ret = -EINVAL; | |
354 | goto out; | |
355 | } | |
356 | ||
357 | p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL); | |
358 | if (p == NULL) { | |
359 | ret = -ENOMEM; | |
360 | goto out; | |
361 | } | |
362 | ||
363 | p->major = major; | |
364 | strlcpy(p->name, name, sizeof(p->name)); | |
365 | p->next = NULL; | |
366 | index = major_to_index(major); | |
367 | ||
368 | for (n = &major_names[index]; *n; n = &(*n)->next) { | |
369 | if ((*n)->major == major) | |
370 | break; | |
371 | } | |
372 | if (!*n) | |
373 | *n = p; | |
374 | else | |
375 | ret = -EBUSY; | |
376 | ||
377 | if (ret < 0) { | |
378 | printk("register_blkdev: cannot get major %d for %s\n", | |
379 | major, name); | |
380 | kfree(p); | |
381 | } | |
382 | out: | |
383 | mutex_unlock(&block_class_lock); | |
384 | return ret; | |
385 | } | |
386 | ||
387 | EXPORT_SYMBOL(register_blkdev); | |
388 | ||
389 | void unregister_blkdev(unsigned int major, const char *name) | |
390 | { | |
391 | struct blk_major_name **n; | |
392 | struct blk_major_name *p = NULL; | |
393 | int index = major_to_index(major); | |
394 | ||
395 | mutex_lock(&block_class_lock); | |
396 | for (n = &major_names[index]; *n; n = &(*n)->next) | |
397 | if ((*n)->major == major) | |
398 | break; | |
399 | if (!*n || strcmp((*n)->name, name)) { | |
400 | WARN_ON(1); | |
401 | } else { | |
402 | p = *n; | |
403 | *n = p->next; | |
404 | } | |
405 | mutex_unlock(&block_class_lock); | |
406 | kfree(p); | |
407 | } | |
408 | ||
409 | EXPORT_SYMBOL(unregister_blkdev); | |
410 | ||
411 | static struct kobj_map *bdev_map; | |
412 | ||
413 | /** | |
414 | * blk_mangle_minor - scatter minor numbers apart | |
415 | * @minor: minor number to mangle | |
416 | * | |
417 | * Scatter consecutively allocated @minor number apart if MANGLE_DEVT | |
418 | * is enabled. Mangling twice gives the original value. | |
419 | * | |
420 | * RETURNS: | |
421 | * Mangled value. | |
422 | * | |
423 | * CONTEXT: | |
424 | * Don't care. | |
425 | */ | |
426 | static int blk_mangle_minor(int minor) | |
427 | { | |
428 | #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT | |
429 | int i; | |
430 | ||
431 | for (i = 0; i < MINORBITS / 2; i++) { | |
432 | int low = minor & (1 << i); | |
433 | int high = minor & (1 << (MINORBITS - 1 - i)); | |
434 | int distance = MINORBITS - 1 - 2 * i; | |
435 | ||
436 | minor ^= low | high; /* clear both bits */ | |
437 | low <<= distance; /* swap the positions */ | |
438 | high >>= distance; | |
439 | minor |= low | high; /* and set */ | |
440 | } | |
441 | #endif | |
442 | return minor; | |
443 | } | |
444 | ||
445 | /** | |
446 | * blk_alloc_devt - allocate a dev_t for a partition | |
447 | * @part: partition to allocate dev_t for | |
448 | * @devt: out parameter for resulting dev_t | |
449 | * | |
450 | * Allocate a dev_t for block device. | |
451 | * | |
452 | * RETURNS: | |
453 | * 0 on success, allocated dev_t is returned in *@devt. -errno on | |
454 | * failure. | |
455 | * | |
456 | * CONTEXT: | |
457 | * Might sleep. | |
458 | */ | |
459 | int blk_alloc_devt(struct hd_struct *part, dev_t *devt) | |
460 | { | |
461 | struct gendisk *disk = part_to_disk(part); | |
462 | int idx; | |
463 | ||
464 | /* in consecutive minor range? */ | |
465 | if (part->partno < disk->minors) { | |
466 | *devt = MKDEV(disk->major, disk->first_minor + part->partno); | |
467 | return 0; | |
468 | } | |
469 | ||
470 | /* allocate ext devt */ | |
471 | idr_preload(GFP_KERNEL); | |
472 | ||
473 | spin_lock_bh(&ext_devt_lock); | |
474 | idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_NOWAIT); | |
475 | spin_unlock_bh(&ext_devt_lock); | |
476 | ||
477 | idr_preload_end(); | |
478 | if (idx < 0) | |
479 | return idx == -ENOSPC ? -EBUSY : idx; | |
480 | ||
481 | *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx)); | |
482 | return 0; | |
483 | } | |
484 | ||
485 | /** | |
486 | * blk_free_devt - free a dev_t | |
487 | * @devt: dev_t to free | |
488 | * | |
489 | * Free @devt which was allocated using blk_alloc_devt(). | |
490 | * | |
491 | * CONTEXT: | |
492 | * Might sleep. | |
493 | */ | |
494 | void blk_free_devt(dev_t devt) | |
495 | { | |
496 | if (devt == MKDEV(0, 0)) | |
497 | return; | |
498 | ||
499 | if (MAJOR(devt) == BLOCK_EXT_MAJOR) { | |
500 | spin_lock_bh(&ext_devt_lock); | |
501 | idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); | |
502 | spin_unlock_bh(&ext_devt_lock); | |
503 | } | |
504 | } | |
505 | ||
506 | static char *bdevt_str(dev_t devt, char *buf) | |
507 | { | |
508 | if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) { | |
509 | char tbuf[BDEVT_SIZE]; | |
510 | snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt)); | |
511 | snprintf(buf, BDEVT_SIZE, "%-9s", tbuf); | |
512 | } else | |
513 | snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt)); | |
514 | ||
515 | return buf; | |
516 | } | |
517 | ||
518 | /* | |
519 | * Register device numbers dev..(dev+range-1) | |
520 | * range must be nonzero | |
521 | * The hash chain is sorted on range, so that subranges can override. | |
522 | */ | |
523 | void blk_register_region(dev_t devt, unsigned long range, struct module *module, | |
524 | struct kobject *(*probe)(dev_t, int *, void *), | |
525 | int (*lock)(dev_t, void *), void *data) | |
526 | { | |
527 | kobj_map(bdev_map, devt, range, module, probe, lock, data); | |
528 | } | |
529 | ||
530 | EXPORT_SYMBOL(blk_register_region); | |
531 | ||
532 | void blk_unregister_region(dev_t devt, unsigned long range) | |
533 | { | |
534 | kobj_unmap(bdev_map, devt, range); | |
535 | } | |
536 | ||
537 | EXPORT_SYMBOL(blk_unregister_region); | |
538 | ||
539 | static struct kobject *exact_match(dev_t devt, int *partno, void *data) | |
540 | { | |
541 | struct gendisk *p = data; | |
542 | ||
543 | return &disk_to_dev(p)->kobj; | |
544 | } | |
545 | ||
546 | static int exact_lock(dev_t devt, void *data) | |
547 | { | |
548 | struct gendisk *p = data; | |
549 | ||
550 | if (!get_disk(p)) | |
551 | return -1; | |
552 | return 0; | |
553 | } | |
554 | ||
555 | static void register_disk(struct device *parent, struct gendisk *disk) | |
556 | { | |
557 | struct device *ddev = disk_to_dev(disk); | |
558 | struct block_device *bdev; | |
559 | struct disk_part_iter piter; | |
560 | struct hd_struct *part; | |
561 | int err; | |
562 | ||
563 | ddev->parent = parent; | |
564 | ||
565 | dev_set_name(ddev, "%s", disk->disk_name); | |
566 | ||
567 | /* delay uevents, until we scanned partition table */ | |
568 | dev_set_uevent_suppress(ddev, 1); | |
569 | ||
570 | if (device_add(ddev)) | |
571 | return; | |
572 | if (!sysfs_deprecated) { | |
573 | err = sysfs_create_link(block_depr, &ddev->kobj, | |
574 | kobject_name(&ddev->kobj)); | |
575 | if (err) { | |
576 | device_del(ddev); | |
577 | return; | |
578 | } | |
579 | } | |
580 | ||
581 | /* | |
582 | * avoid probable deadlock caused by allocating memory with | |
583 | * GFP_KERNEL in runtime_resume callback of its all ancestor | |
584 | * devices | |
585 | */ | |
586 | pm_runtime_set_memalloc_noio(ddev, true); | |
587 | ||
588 | disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj); | |
589 | disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj); | |
590 | ||
591 | /* No minors to use for partitions */ | |
592 | if (!disk_part_scan_enabled(disk)) | |
593 | goto exit; | |
594 | ||
595 | /* No such device (e.g., media were just removed) */ | |
596 | if (!get_capacity(disk)) | |
597 | goto exit; | |
598 | ||
599 | bdev = bdget_disk(disk, 0); | |
600 | if (!bdev) | |
601 | goto exit; | |
602 | ||
603 | bdev->bd_invalidated = 1; | |
604 | err = blkdev_get(bdev, FMODE_READ, NULL); | |
605 | if (err < 0) | |
606 | goto exit; | |
607 | blkdev_put(bdev, FMODE_READ); | |
608 | ||
609 | exit: | |
610 | /* announce disk after possible partitions are created */ | |
611 | dev_set_uevent_suppress(ddev, 0); | |
612 | kobject_uevent(&ddev->kobj, KOBJ_ADD); | |
613 | ||
614 | /* announce possible partitions */ | |
615 | disk_part_iter_init(&piter, disk, 0); | |
616 | while ((part = disk_part_iter_next(&piter))) | |
617 | kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD); | |
618 | disk_part_iter_exit(&piter); | |
619 | } | |
620 | ||
621 | /** | |
622 | * device_add_disk - add partitioning information to kernel list | |
623 | * @parent: parent device for the disk | |
624 | * @disk: per-device partitioning information | |
625 | * | |
626 | * This function registers the partitioning information in @disk | |
627 | * with the kernel. | |
628 | * | |
629 | * FIXME: error handling | |
630 | */ | |
631 | void device_add_disk(struct device *parent, struct gendisk *disk) | |
632 | { | |
633 | struct backing_dev_info *bdi; | |
634 | dev_t devt; | |
635 | int retval; | |
636 | ||
637 | /* minors == 0 indicates to use ext devt from part0 and should | |
638 | * be accompanied with EXT_DEVT flag. Make sure all | |
639 | * parameters make sense. | |
640 | */ | |
641 | WARN_ON(disk->minors && !(disk->major || disk->first_minor)); | |
642 | WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT)); | |
643 | ||
644 | disk->flags |= GENHD_FL_UP; | |
645 | ||
646 | retval = blk_alloc_devt(&disk->part0, &devt); | |
647 | if (retval) { | |
648 | WARN_ON(1); | |
649 | return; | |
650 | } | |
651 | disk_to_dev(disk)->devt = devt; | |
652 | ||
653 | /* ->major and ->first_minor aren't supposed to be | |
654 | * dereferenced from here on, but set them just in case. | |
655 | */ | |
656 | disk->major = MAJOR(devt); | |
657 | disk->first_minor = MINOR(devt); | |
658 | ||
659 | disk_alloc_events(disk); | |
660 | ||
661 | /* Register BDI before referencing it from bdev */ | |
662 | bdi = disk->queue->backing_dev_info; | |
663 | bdi_register_owner(bdi, disk_to_dev(disk)); | |
664 | ||
665 | blk_register_region(disk_devt(disk), disk->minors, NULL, | |
666 | exact_match, exact_lock, disk); | |
667 | register_disk(parent, disk); | |
668 | blk_register_queue(disk); | |
669 | ||
670 | /* | |
671 | * Take an extra ref on queue which will be put on disk_release() | |
672 | * so that it sticks around as long as @disk is there. | |
673 | */ | |
674 | WARN_ON_ONCE(!blk_get_queue(disk->queue)); | |
675 | ||
676 | retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj, | |
677 | "bdi"); | |
678 | WARN_ON(retval); | |
679 | ||
680 | disk_add_events(disk); | |
681 | blk_integrity_add(disk); | |
682 | } | |
683 | EXPORT_SYMBOL(device_add_disk); | |
684 | ||
685 | void del_gendisk(struct gendisk *disk) | |
686 | { | |
687 | struct disk_part_iter piter; | |
688 | struct hd_struct *part; | |
689 | ||
690 | blk_integrity_del(disk); | |
691 | disk_del_events(disk); | |
692 | ||
693 | /* invalidate stuff */ | |
694 | disk_part_iter_init(&piter, disk, | |
695 | DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE); | |
696 | while ((part = disk_part_iter_next(&piter))) { | |
697 | invalidate_partition(disk, part->partno); | |
698 | bdev_unhash_inode(part_devt(part)); | |
699 | delete_partition(disk, part->partno); | |
700 | } | |
701 | disk_part_iter_exit(&piter); | |
702 | ||
703 | invalidate_partition(disk, 0); | |
704 | bdev_unhash_inode(disk_devt(disk)); | |
705 | set_capacity(disk, 0); | |
706 | disk->flags &= ~GENHD_FL_UP; | |
707 | ||
708 | sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi"); | |
709 | if (disk->queue) { | |
710 | /* | |
711 | * Unregister bdi before releasing device numbers (as they can | |
712 | * get reused and we'd get clashes in sysfs). | |
713 | */ | |
714 | bdi_unregister(disk->queue->backing_dev_info); | |
715 | blk_unregister_queue(disk); | |
716 | } else { | |
717 | WARN_ON(1); | |
718 | } | |
719 | blk_unregister_region(disk_devt(disk), disk->minors); | |
720 | ||
721 | part_stat_set_all(&disk->part0, 0); | |
722 | disk->part0.stamp = 0; | |
723 | ||
724 | kobject_put(disk->part0.holder_dir); | |
725 | kobject_put(disk->slave_dir); | |
726 | if (!sysfs_deprecated) | |
727 | sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk))); | |
728 | pm_runtime_set_memalloc_noio(disk_to_dev(disk), false); | |
729 | device_del(disk_to_dev(disk)); | |
730 | } | |
731 | EXPORT_SYMBOL(del_gendisk); | |
732 | ||
733 | /* sysfs access to bad-blocks list. */ | |
734 | static ssize_t disk_badblocks_show(struct device *dev, | |
735 | struct device_attribute *attr, | |
736 | char *page) | |
737 | { | |
738 | struct gendisk *disk = dev_to_disk(dev); | |
739 | ||
740 | if (!disk->bb) | |
741 | return sprintf(page, "\n"); | |
742 | ||
743 | return badblocks_show(disk->bb, page, 0); | |
744 | } | |
745 | ||
746 | static ssize_t disk_badblocks_store(struct device *dev, | |
747 | struct device_attribute *attr, | |
748 | const char *page, size_t len) | |
749 | { | |
750 | struct gendisk *disk = dev_to_disk(dev); | |
751 | ||
752 | if (!disk->bb) | |
753 | return -ENXIO; | |
754 | ||
755 | return badblocks_store(disk->bb, page, len, 0); | |
756 | } | |
757 | ||
758 | /** | |
759 | * get_gendisk - get partitioning information for a given device | |
760 | * @devt: device to get partitioning information for | |
761 | * @partno: returned partition index | |
762 | * | |
763 | * This function gets the structure containing partitioning | |
764 | * information for the given device @devt. | |
765 | */ | |
766 | struct gendisk *get_gendisk(dev_t devt, int *partno) | |
767 | { | |
768 | struct gendisk *disk = NULL; | |
769 | ||
770 | if (MAJOR(devt) != BLOCK_EXT_MAJOR) { | |
771 | struct kobject *kobj; | |
772 | ||
773 | kobj = kobj_lookup(bdev_map, devt, partno); | |
774 | if (kobj) | |
775 | disk = dev_to_disk(kobj_to_dev(kobj)); | |
776 | } else { | |
777 | struct hd_struct *part; | |
778 | ||
779 | spin_lock_bh(&ext_devt_lock); | |
780 | part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); | |
781 | if (part && get_disk(part_to_disk(part))) { | |
782 | *partno = part->partno; | |
783 | disk = part_to_disk(part); | |
784 | } | |
785 | spin_unlock_bh(&ext_devt_lock); | |
786 | } | |
787 | ||
788 | return disk; | |
789 | } | |
790 | EXPORT_SYMBOL(get_gendisk); | |
791 | ||
792 | /** | |
793 | * bdget_disk - do bdget() by gendisk and partition number | |
794 | * @disk: gendisk of interest | |
795 | * @partno: partition number | |
796 | * | |
797 | * Find partition @partno from @disk, do bdget() on it. | |
798 | * | |
799 | * CONTEXT: | |
800 | * Don't care. | |
801 | * | |
802 | * RETURNS: | |
803 | * Resulting block_device on success, NULL on failure. | |
804 | */ | |
805 | struct block_device *bdget_disk(struct gendisk *disk, int partno) | |
806 | { | |
807 | struct hd_struct *part; | |
808 | struct block_device *bdev = NULL; | |
809 | ||
810 | part = disk_get_part(disk, partno); | |
811 | if (part) | |
812 | bdev = bdget(part_devt(part)); | |
813 | disk_put_part(part); | |
814 | ||
815 | return bdev; | |
816 | } | |
817 | EXPORT_SYMBOL(bdget_disk); | |
818 | ||
819 | /* | |
820 | * print a full list of all partitions - intended for places where the root | |
821 | * filesystem can't be mounted and thus to give the victim some idea of what | |
822 | * went wrong | |
823 | */ | |
824 | void __init printk_all_partitions(void) | |
825 | { | |
826 | struct class_dev_iter iter; | |
827 | struct device *dev; | |
828 | ||
829 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); | |
830 | while ((dev = class_dev_iter_next(&iter))) { | |
831 | struct gendisk *disk = dev_to_disk(dev); | |
832 | struct disk_part_iter piter; | |
833 | struct hd_struct *part; | |
834 | char name_buf[BDEVNAME_SIZE]; | |
835 | char devt_buf[BDEVT_SIZE]; | |
836 | ||
837 | /* | |
838 | * Don't show empty devices or things that have been | |
839 | * suppressed | |
840 | */ | |
841 | if (get_capacity(disk) == 0 || | |
842 | (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)) | |
843 | continue; | |
844 | ||
845 | /* | |
846 | * Note, unlike /proc/partitions, I am showing the | |
847 | * numbers in hex - the same format as the root= | |
848 | * option takes. | |
849 | */ | |
850 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0); | |
851 | while ((part = disk_part_iter_next(&piter))) { | |
852 | bool is_part0 = part == &disk->part0; | |
853 | ||
854 | printk("%s%s %10llu %s %s", is_part0 ? "" : " ", | |
855 | bdevt_str(part_devt(part), devt_buf), | |
856 | (unsigned long long)part_nr_sects_read(part) >> 1 | |
857 | , disk_name(disk, part->partno, name_buf), | |
858 | part->info ? part->info->uuid : ""); | |
859 | if (is_part0) { | |
860 | if (dev->parent && dev->parent->driver) | |
861 | printk(" driver: %s\n", | |
862 | dev->parent->driver->name); | |
863 | else | |
864 | printk(" (driver?)\n"); | |
865 | } else | |
866 | printk("\n"); | |
867 | } | |
868 | disk_part_iter_exit(&piter); | |
869 | } | |
870 | class_dev_iter_exit(&iter); | |
871 | } | |
872 | ||
873 | #ifdef CONFIG_PROC_FS | |
874 | /* iterator */ | |
875 | static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos) | |
876 | { | |
877 | loff_t skip = *pos; | |
878 | struct class_dev_iter *iter; | |
879 | struct device *dev; | |
880 | ||
881 | iter = kmalloc(sizeof(*iter), GFP_KERNEL); | |
882 | if (!iter) | |
883 | return ERR_PTR(-ENOMEM); | |
884 | ||
885 | seqf->private = iter; | |
886 | class_dev_iter_init(iter, &block_class, NULL, &disk_type); | |
887 | do { | |
888 | dev = class_dev_iter_next(iter); | |
889 | if (!dev) | |
890 | return NULL; | |
891 | } while (skip--); | |
892 | ||
893 | return dev_to_disk(dev); | |
894 | } | |
895 | ||
896 | static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos) | |
897 | { | |
898 | struct device *dev; | |
899 | ||
900 | (*pos)++; | |
901 | dev = class_dev_iter_next(seqf->private); | |
902 | if (dev) | |
903 | return dev_to_disk(dev); | |
904 | ||
905 | return NULL; | |
906 | } | |
907 | ||
908 | static void disk_seqf_stop(struct seq_file *seqf, void *v) | |
909 | { | |
910 | struct class_dev_iter *iter = seqf->private; | |
911 | ||
912 | /* stop is called even after start failed :-( */ | |
913 | if (iter) { | |
914 | class_dev_iter_exit(iter); | |
915 | kfree(iter); | |
916 | seqf->private = NULL; | |
917 | } | |
918 | } | |
919 | ||
920 | static void *show_partition_start(struct seq_file *seqf, loff_t *pos) | |
921 | { | |
922 | void *p; | |
923 | ||
924 | p = disk_seqf_start(seqf, pos); | |
925 | if (!IS_ERR_OR_NULL(p) && !*pos) | |
926 | seq_puts(seqf, "major minor #blocks name\n\n"); | |
927 | return p; | |
928 | } | |
929 | ||
930 | static int show_partition(struct seq_file *seqf, void *v) | |
931 | { | |
932 | struct gendisk *sgp = v; | |
933 | struct disk_part_iter piter; | |
934 | struct hd_struct *part; | |
935 | char buf[BDEVNAME_SIZE]; | |
936 | ||
937 | /* Don't show non-partitionable removeable devices or empty devices */ | |
938 | if (!get_capacity(sgp) || (!disk_max_parts(sgp) && | |
939 | (sgp->flags & GENHD_FL_REMOVABLE))) | |
940 | return 0; | |
941 | if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO) | |
942 | return 0; | |
943 | ||
944 | /* show the full disk and all non-0 size partitions of it */ | |
945 | disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0); | |
946 | while ((part = disk_part_iter_next(&piter))) | |
947 | seq_printf(seqf, "%4d %7d %10llu %s\n", | |
948 | MAJOR(part_devt(part)), MINOR(part_devt(part)), | |
949 | (unsigned long long)part_nr_sects_read(part) >> 1, | |
950 | disk_name(sgp, part->partno, buf)); | |
951 | disk_part_iter_exit(&piter); | |
952 | ||
953 | return 0; | |
954 | } | |
955 | ||
956 | static const struct seq_operations partitions_op = { | |
957 | .start = show_partition_start, | |
958 | .next = disk_seqf_next, | |
959 | .stop = disk_seqf_stop, | |
960 | .show = show_partition | |
961 | }; | |
962 | ||
963 | static int partitions_open(struct inode *inode, struct file *file) | |
964 | { | |
965 | return seq_open(file, &partitions_op); | |
966 | } | |
967 | ||
968 | static const struct file_operations proc_partitions_operations = { | |
969 | .open = partitions_open, | |
970 | .read = seq_read, | |
971 | .llseek = seq_lseek, | |
972 | .release = seq_release, | |
973 | }; | |
974 | #endif | |
975 | ||
976 | ||
977 | static struct kobject *base_probe(dev_t devt, int *partno, void *data) | |
978 | { | |
979 | if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0) | |
980 | /* Make old-style 2.4 aliases work */ | |
981 | request_module("block-major-%d", MAJOR(devt)); | |
982 | return NULL; | |
983 | } | |
984 | ||
985 | static int __init genhd_device_init(void) | |
986 | { | |
987 | int error; | |
988 | ||
989 | block_class.dev_kobj = sysfs_dev_block_kobj; | |
990 | error = class_register(&block_class); | |
991 | if (unlikely(error)) | |
992 | return error; | |
993 | bdev_map = kobj_map_init(base_probe, &block_class_lock); | |
994 | blk_dev_init(); | |
995 | ||
996 | register_blkdev(BLOCK_EXT_MAJOR, "blkext"); | |
997 | ||
998 | /* create top-level block dir */ | |
999 | if (!sysfs_deprecated) | |
1000 | block_depr = kobject_create_and_add("block", NULL); | |
1001 | return 0; | |
1002 | } | |
1003 | ||
1004 | subsys_initcall(genhd_device_init); | |
1005 | ||
1006 | static ssize_t disk_range_show(struct device *dev, | |
1007 | struct device_attribute *attr, char *buf) | |
1008 | { | |
1009 | struct gendisk *disk = dev_to_disk(dev); | |
1010 | ||
1011 | return sprintf(buf, "%d\n", disk->minors); | |
1012 | } | |
1013 | ||
1014 | static ssize_t disk_ext_range_show(struct device *dev, | |
1015 | struct device_attribute *attr, char *buf) | |
1016 | { | |
1017 | struct gendisk *disk = dev_to_disk(dev); | |
1018 | ||
1019 | return sprintf(buf, "%d\n", disk_max_parts(disk)); | |
1020 | } | |
1021 | ||
1022 | static ssize_t disk_removable_show(struct device *dev, | |
1023 | struct device_attribute *attr, char *buf) | |
1024 | { | |
1025 | struct gendisk *disk = dev_to_disk(dev); | |
1026 | ||
1027 | return sprintf(buf, "%d\n", | |
1028 | (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0)); | |
1029 | } | |
1030 | ||
1031 | static ssize_t disk_ro_show(struct device *dev, | |
1032 | struct device_attribute *attr, char *buf) | |
1033 | { | |
1034 | struct gendisk *disk = dev_to_disk(dev); | |
1035 | ||
1036 | return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0); | |
1037 | } | |
1038 | ||
1039 | static ssize_t disk_capability_show(struct device *dev, | |
1040 | struct device_attribute *attr, char *buf) | |
1041 | { | |
1042 | struct gendisk *disk = dev_to_disk(dev); | |
1043 | ||
1044 | return sprintf(buf, "%x\n", disk->flags); | |
1045 | } | |
1046 | ||
1047 | static ssize_t disk_alignment_offset_show(struct device *dev, | |
1048 | struct device_attribute *attr, | |
1049 | char *buf) | |
1050 | { | |
1051 | struct gendisk *disk = dev_to_disk(dev); | |
1052 | ||
1053 | return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue)); | |
1054 | } | |
1055 | ||
1056 | static ssize_t disk_discard_alignment_show(struct device *dev, | |
1057 | struct device_attribute *attr, | |
1058 | char *buf) | |
1059 | { | |
1060 | struct gendisk *disk = dev_to_disk(dev); | |
1061 | ||
1062 | return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue)); | |
1063 | } | |
1064 | ||
1065 | static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL); | |
1066 | static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL); | |
1067 | static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL); | |
1068 | static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL); | |
1069 | static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL); | |
1070 | static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL); | |
1071 | static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show, | |
1072 | NULL); | |
1073 | static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL); | |
1074 | static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL); | |
1075 | static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL); | |
1076 | static DEVICE_ATTR(badblocks, S_IRUGO | S_IWUSR, disk_badblocks_show, | |
1077 | disk_badblocks_store); | |
1078 | #ifdef CONFIG_FAIL_MAKE_REQUEST | |
1079 | static struct device_attribute dev_attr_fail = | |
1080 | __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store); | |
1081 | #endif | |
1082 | #ifdef CONFIG_FAIL_IO_TIMEOUT | |
1083 | static struct device_attribute dev_attr_fail_timeout = | |
1084 | __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show, | |
1085 | part_timeout_store); | |
1086 | #endif | |
1087 | ||
1088 | static struct attribute *disk_attrs[] = { | |
1089 | &dev_attr_range.attr, | |
1090 | &dev_attr_ext_range.attr, | |
1091 | &dev_attr_removable.attr, | |
1092 | &dev_attr_ro.attr, | |
1093 | &dev_attr_size.attr, | |
1094 | &dev_attr_alignment_offset.attr, | |
1095 | &dev_attr_discard_alignment.attr, | |
1096 | &dev_attr_capability.attr, | |
1097 | &dev_attr_stat.attr, | |
1098 | &dev_attr_inflight.attr, | |
1099 | &dev_attr_badblocks.attr, | |
1100 | #ifdef CONFIG_FAIL_MAKE_REQUEST | |
1101 | &dev_attr_fail.attr, | |
1102 | #endif | |
1103 | #ifdef CONFIG_FAIL_IO_TIMEOUT | |
1104 | &dev_attr_fail_timeout.attr, | |
1105 | #endif | |
1106 | NULL | |
1107 | }; | |
1108 | ||
1109 | static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n) | |
1110 | { | |
1111 | struct device *dev = container_of(kobj, typeof(*dev), kobj); | |
1112 | struct gendisk *disk = dev_to_disk(dev); | |
1113 | ||
1114 | if (a == &dev_attr_badblocks.attr && !disk->bb) | |
1115 | return 0; | |
1116 | return a->mode; | |
1117 | } | |
1118 | ||
1119 | static struct attribute_group disk_attr_group = { | |
1120 | .attrs = disk_attrs, | |
1121 | .is_visible = disk_visible, | |
1122 | }; | |
1123 | ||
1124 | static const struct attribute_group *disk_attr_groups[] = { | |
1125 | &disk_attr_group, | |
1126 | NULL | |
1127 | }; | |
1128 | ||
1129 | /** | |
1130 | * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way | |
1131 | * @disk: disk to replace part_tbl for | |
1132 | * @new_ptbl: new part_tbl to install | |
1133 | * | |
1134 | * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The | |
1135 | * original ptbl is freed using RCU callback. | |
1136 | * | |
1137 | * LOCKING: | |
1138 | * Matching bd_mutex locked or the caller is the only user of @disk. | |
1139 | */ | |
1140 | static void disk_replace_part_tbl(struct gendisk *disk, | |
1141 | struct disk_part_tbl *new_ptbl) | |
1142 | { | |
1143 | struct disk_part_tbl *old_ptbl = | |
1144 | rcu_dereference_protected(disk->part_tbl, 1); | |
1145 | ||
1146 | rcu_assign_pointer(disk->part_tbl, new_ptbl); | |
1147 | ||
1148 | if (old_ptbl) { | |
1149 | rcu_assign_pointer(old_ptbl->last_lookup, NULL); | |
1150 | kfree_rcu(old_ptbl, rcu_head); | |
1151 | } | |
1152 | } | |
1153 | ||
1154 | /** | |
1155 | * disk_expand_part_tbl - expand disk->part_tbl | |
1156 | * @disk: disk to expand part_tbl for | |
1157 | * @partno: expand such that this partno can fit in | |
1158 | * | |
1159 | * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl | |
1160 | * uses RCU to allow unlocked dereferencing for stats and other stuff. | |
1161 | * | |
1162 | * LOCKING: | |
1163 | * Matching bd_mutex locked or the caller is the only user of @disk. | |
1164 | * Might sleep. | |
1165 | * | |
1166 | * RETURNS: | |
1167 | * 0 on success, -errno on failure. | |
1168 | */ | |
1169 | int disk_expand_part_tbl(struct gendisk *disk, int partno) | |
1170 | { | |
1171 | struct disk_part_tbl *old_ptbl = | |
1172 | rcu_dereference_protected(disk->part_tbl, 1); | |
1173 | struct disk_part_tbl *new_ptbl; | |
1174 | int len = old_ptbl ? old_ptbl->len : 0; | |
1175 | int i, target; | |
1176 | size_t size; | |
1177 | ||
1178 | /* | |
1179 | * check for int overflow, since we can get here from blkpg_ioctl() | |
1180 | * with a user passed 'partno'. | |
1181 | */ | |
1182 | target = partno + 1; | |
1183 | if (target < 0) | |
1184 | return -EINVAL; | |
1185 | ||
1186 | /* disk_max_parts() is zero during initialization, ignore if so */ | |
1187 | if (disk_max_parts(disk) && target > disk_max_parts(disk)) | |
1188 | return -EINVAL; | |
1189 | ||
1190 | if (target <= len) | |
1191 | return 0; | |
1192 | ||
1193 | size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]); | |
1194 | new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id); | |
1195 | if (!new_ptbl) | |
1196 | return -ENOMEM; | |
1197 | ||
1198 | new_ptbl->len = target; | |
1199 | ||
1200 | for (i = 0; i < len; i++) | |
1201 | rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]); | |
1202 | ||
1203 | disk_replace_part_tbl(disk, new_ptbl); | |
1204 | return 0; | |
1205 | } | |
1206 | ||
1207 | static void disk_release(struct device *dev) | |
1208 | { | |
1209 | struct gendisk *disk = dev_to_disk(dev); | |
1210 | ||
1211 | blk_free_devt(dev->devt); | |
1212 | disk_release_events(disk); | |
1213 | kfree(disk->random); | |
1214 | disk_replace_part_tbl(disk, NULL); | |
1215 | hd_free_part(&disk->part0); | |
1216 | if (disk->queue) | |
1217 | blk_put_queue(disk->queue); | |
1218 | kfree(disk); | |
1219 | } | |
1220 | struct class block_class = { | |
1221 | .name = "block", | |
1222 | }; | |
1223 | ||
1224 | static char *block_devnode(struct device *dev, umode_t *mode, | |
1225 | kuid_t *uid, kgid_t *gid) | |
1226 | { | |
1227 | struct gendisk *disk = dev_to_disk(dev); | |
1228 | ||
1229 | if (disk->devnode) | |
1230 | return disk->devnode(disk, mode); | |
1231 | return NULL; | |
1232 | } | |
1233 | ||
1234 | static const struct device_type disk_type = { | |
1235 | .name = "disk", | |
1236 | .groups = disk_attr_groups, | |
1237 | .release = disk_release, | |
1238 | .devnode = block_devnode, | |
1239 | }; | |
1240 | ||
1241 | #ifdef CONFIG_PROC_FS | |
1242 | /* | |
1243 | * aggregate disk stat collector. Uses the same stats that the sysfs | |
1244 | * entries do, above, but makes them available through one seq_file. | |
1245 | * | |
1246 | * The output looks suspiciously like /proc/partitions with a bunch of | |
1247 | * extra fields. | |
1248 | */ | |
1249 | static int diskstats_show(struct seq_file *seqf, void *v) | |
1250 | { | |
1251 | struct gendisk *gp = v; | |
1252 | struct disk_part_iter piter; | |
1253 | struct hd_struct *hd; | |
1254 | char buf[BDEVNAME_SIZE]; | |
1255 | unsigned int inflight[2]; | |
1256 | int cpu; | |
1257 | ||
1258 | /* | |
1259 | if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next) | |
1260 | seq_puts(seqf, "major minor name" | |
1261 | " rio rmerge rsect ruse wio wmerge " | |
1262 | "wsect wuse running use aveq" | |
1263 | "\n\n"); | |
1264 | */ | |
1265 | ||
1266 | disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0); | |
1267 | while ((hd = disk_part_iter_next(&piter))) { | |
1268 | cpu = part_stat_lock(); | |
1269 | part_round_stats(gp->queue, cpu, hd); | |
1270 | part_stat_unlock(); | |
1271 | part_in_flight(gp->queue, hd, inflight); | |
1272 | seq_printf(seqf, "%4d %7d %s %lu %lu %lu " | |
1273 | "%u %lu %lu %lu %u %u %u %u\n", | |
1274 | MAJOR(part_devt(hd)), MINOR(part_devt(hd)), | |
1275 | disk_name(gp, hd->partno, buf), | |
1276 | part_stat_read(hd, ios[READ]), | |
1277 | part_stat_read(hd, merges[READ]), | |
1278 | part_stat_read(hd, sectors[READ]), | |
1279 | jiffies_to_msecs(part_stat_read(hd, ticks[READ])), | |
1280 | part_stat_read(hd, ios[WRITE]), | |
1281 | part_stat_read(hd, merges[WRITE]), | |
1282 | part_stat_read(hd, sectors[WRITE]), | |
1283 | jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])), | |
1284 | inflight[0], | |
1285 | jiffies_to_msecs(part_stat_read(hd, io_ticks)), | |
1286 | jiffies_to_msecs(part_stat_read(hd, time_in_queue)) | |
1287 | ); | |
1288 | } | |
1289 | disk_part_iter_exit(&piter); | |
1290 | ||
1291 | return 0; | |
1292 | } | |
1293 | ||
1294 | static const struct seq_operations diskstats_op = { | |
1295 | .start = disk_seqf_start, | |
1296 | .next = disk_seqf_next, | |
1297 | .stop = disk_seqf_stop, | |
1298 | .show = diskstats_show | |
1299 | }; | |
1300 | ||
1301 | static int diskstats_open(struct inode *inode, struct file *file) | |
1302 | { | |
1303 | return seq_open(file, &diskstats_op); | |
1304 | } | |
1305 | ||
1306 | static const struct file_operations proc_diskstats_operations = { | |
1307 | .open = diskstats_open, | |
1308 | .read = seq_read, | |
1309 | .llseek = seq_lseek, | |
1310 | .release = seq_release, | |
1311 | }; | |
1312 | ||
1313 | static int __init proc_genhd_init(void) | |
1314 | { | |
1315 | proc_create("diskstats", 0, NULL, &proc_diskstats_operations); | |
1316 | proc_create("partitions", 0, NULL, &proc_partitions_operations); | |
1317 | return 0; | |
1318 | } | |
1319 | module_init(proc_genhd_init); | |
1320 | #endif /* CONFIG_PROC_FS */ | |
1321 | ||
1322 | dev_t blk_lookup_devt(const char *name, int partno) | |
1323 | { | |
1324 | dev_t devt = MKDEV(0, 0); | |
1325 | struct class_dev_iter iter; | |
1326 | struct device *dev; | |
1327 | ||
1328 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); | |
1329 | while ((dev = class_dev_iter_next(&iter))) { | |
1330 | struct gendisk *disk = dev_to_disk(dev); | |
1331 | struct hd_struct *part; | |
1332 | ||
1333 | if (strcmp(dev_name(dev), name)) | |
1334 | continue; | |
1335 | ||
1336 | if (partno < disk->minors) { | |
1337 | /* We need to return the right devno, even | |
1338 | * if the partition doesn't exist yet. | |
1339 | */ | |
1340 | devt = MKDEV(MAJOR(dev->devt), | |
1341 | MINOR(dev->devt) + partno); | |
1342 | break; | |
1343 | } | |
1344 | part = disk_get_part(disk, partno); | |
1345 | if (part) { | |
1346 | devt = part_devt(part); | |
1347 | disk_put_part(part); | |
1348 | break; | |
1349 | } | |
1350 | disk_put_part(part); | |
1351 | } | |
1352 | class_dev_iter_exit(&iter); | |
1353 | return devt; | |
1354 | } | |
1355 | EXPORT_SYMBOL(blk_lookup_devt); | |
1356 | ||
1357 | struct gendisk *alloc_disk(int minors) | |
1358 | { | |
1359 | return alloc_disk_node(minors, NUMA_NO_NODE); | |
1360 | } | |
1361 | EXPORT_SYMBOL(alloc_disk); | |
1362 | ||
1363 | struct gendisk *alloc_disk_node(int minors, int node_id) | |
1364 | { | |
1365 | struct gendisk *disk; | |
1366 | struct disk_part_tbl *ptbl; | |
1367 | ||
1368 | if (minors > DISK_MAX_PARTS) { | |
1369 | printk(KERN_ERR | |
1370 | "block: can't allocated more than %d partitions\n", | |
1371 | DISK_MAX_PARTS); | |
1372 | minors = DISK_MAX_PARTS; | |
1373 | } | |
1374 | ||
1375 | disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id); | |
1376 | if (disk) { | |
1377 | if (!init_part_stats(&disk->part0)) { | |
1378 | kfree(disk); | |
1379 | return NULL; | |
1380 | } | |
1381 | disk->node_id = node_id; | |
1382 | if (disk_expand_part_tbl(disk, 0)) { | |
1383 | free_part_stats(&disk->part0); | |
1384 | kfree(disk); | |
1385 | return NULL; | |
1386 | } | |
1387 | ptbl = rcu_dereference_protected(disk->part_tbl, 1); | |
1388 | rcu_assign_pointer(ptbl->part[0], &disk->part0); | |
1389 | ||
1390 | /* | |
1391 | * set_capacity() and get_capacity() currently don't use | |
1392 | * seqcounter to read/update the part0->nr_sects. Still init | |
1393 | * the counter as we can read the sectors in IO submission | |
1394 | * patch using seqence counters. | |
1395 | * | |
1396 | * TODO: Ideally set_capacity() and get_capacity() should be | |
1397 | * converted to make use of bd_mutex and sequence counters. | |
1398 | */ | |
1399 | seqcount_init(&disk->part0.nr_sects_seq); | |
1400 | if (hd_ref_init(&disk->part0)) { | |
1401 | hd_free_part(&disk->part0); | |
1402 | kfree(disk); | |
1403 | return NULL; | |
1404 | } | |
1405 | ||
1406 | disk->minors = minors; | |
1407 | rand_initialize_disk(disk); | |
1408 | disk_to_dev(disk)->class = &block_class; | |
1409 | disk_to_dev(disk)->type = &disk_type; | |
1410 | device_initialize(disk_to_dev(disk)); | |
1411 | } | |
1412 | return disk; | |
1413 | } | |
1414 | EXPORT_SYMBOL(alloc_disk_node); | |
1415 | ||
1416 | struct kobject *get_disk(struct gendisk *disk) | |
1417 | { | |
1418 | struct module *owner; | |
1419 | struct kobject *kobj; | |
1420 | ||
1421 | if (!disk->fops) | |
1422 | return NULL; | |
1423 | owner = disk->fops->owner; | |
1424 | if (owner && !try_module_get(owner)) | |
1425 | return NULL; | |
1426 | kobj = kobject_get_unless_zero(&disk_to_dev(disk)->kobj); | |
1427 | if (kobj == NULL) { | |
1428 | module_put(owner); | |
1429 | return NULL; | |
1430 | } | |
1431 | return kobj; | |
1432 | ||
1433 | } | |
1434 | ||
1435 | EXPORT_SYMBOL(get_disk); | |
1436 | ||
1437 | void put_disk(struct gendisk *disk) | |
1438 | { | |
1439 | if (disk) | |
1440 | kobject_put(&disk_to_dev(disk)->kobj); | |
1441 | } | |
1442 | ||
1443 | EXPORT_SYMBOL(put_disk); | |
1444 | ||
1445 | static void set_disk_ro_uevent(struct gendisk *gd, int ro) | |
1446 | { | |
1447 | char event[] = "DISK_RO=1"; | |
1448 | char *envp[] = { event, NULL }; | |
1449 | ||
1450 | if (!ro) | |
1451 | event[8] = '0'; | |
1452 | kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); | |
1453 | } | |
1454 | ||
1455 | void set_device_ro(struct block_device *bdev, int flag) | |
1456 | { | |
1457 | bdev->bd_part->policy = flag; | |
1458 | } | |
1459 | ||
1460 | EXPORT_SYMBOL(set_device_ro); | |
1461 | ||
1462 | void set_disk_ro(struct gendisk *disk, int flag) | |
1463 | { | |
1464 | struct disk_part_iter piter; | |
1465 | struct hd_struct *part; | |
1466 | ||
1467 | if (disk->part0.policy != flag) { | |
1468 | set_disk_ro_uevent(disk, flag); | |
1469 | disk->part0.policy = flag; | |
1470 | } | |
1471 | ||
1472 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY); | |
1473 | while ((part = disk_part_iter_next(&piter))) | |
1474 | part->policy = flag; | |
1475 | disk_part_iter_exit(&piter); | |
1476 | } | |
1477 | ||
1478 | EXPORT_SYMBOL(set_disk_ro); | |
1479 | ||
1480 | int bdev_read_only(struct block_device *bdev) | |
1481 | { | |
1482 | if (!bdev) | |
1483 | return 0; | |
1484 | return bdev->bd_part->policy; | |
1485 | } | |
1486 | ||
1487 | EXPORT_SYMBOL(bdev_read_only); | |
1488 | ||
1489 | int invalidate_partition(struct gendisk *disk, int partno) | |
1490 | { | |
1491 | int res = 0; | |
1492 | struct block_device *bdev = bdget_disk(disk, partno); | |
1493 | if (bdev) { | |
1494 | fsync_bdev(bdev); | |
1495 | res = __invalidate_device(bdev, true); | |
1496 | bdput(bdev); | |
1497 | } | |
1498 | return res; | |
1499 | } | |
1500 | ||
1501 | EXPORT_SYMBOL(invalidate_partition); | |
1502 | ||
1503 | /* | |
1504 | * Disk events - monitor disk events like media change and eject request. | |
1505 | */ | |
1506 | struct disk_events { | |
1507 | struct list_head node; /* all disk_event's */ | |
1508 | struct gendisk *disk; /* the associated disk */ | |
1509 | spinlock_t lock; | |
1510 | ||
1511 | struct mutex block_mutex; /* protects blocking */ | |
1512 | int block; /* event blocking depth */ | |
1513 | unsigned int pending; /* events already sent out */ | |
1514 | unsigned int clearing; /* events being cleared */ | |
1515 | ||
1516 | long poll_msecs; /* interval, -1 for default */ | |
1517 | struct delayed_work dwork; | |
1518 | }; | |
1519 | ||
1520 | static const char *disk_events_strs[] = { | |
1521 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change", | |
1522 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request", | |
1523 | }; | |
1524 | ||
1525 | static char *disk_uevents[] = { | |
1526 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1", | |
1527 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1", | |
1528 | }; | |
1529 | ||
1530 | /* list of all disk_events */ | |
1531 | static DEFINE_MUTEX(disk_events_mutex); | |
1532 | static LIST_HEAD(disk_events); | |
1533 | ||
1534 | /* disable in-kernel polling by default */ | |
1535 | static unsigned long disk_events_dfl_poll_msecs; | |
1536 | ||
1537 | static unsigned long disk_events_poll_jiffies(struct gendisk *disk) | |
1538 | { | |
1539 | struct disk_events *ev = disk->ev; | |
1540 | long intv_msecs = 0; | |
1541 | ||
1542 | /* | |
1543 | * If device-specific poll interval is set, always use it. If | |
1544 | * the default is being used, poll iff there are events which | |
1545 | * can't be monitored asynchronously. | |
1546 | */ | |
1547 | if (ev->poll_msecs >= 0) | |
1548 | intv_msecs = ev->poll_msecs; | |
1549 | else if (disk->events & ~disk->async_events) | |
1550 | intv_msecs = disk_events_dfl_poll_msecs; | |
1551 | ||
1552 | return msecs_to_jiffies(intv_msecs); | |
1553 | } | |
1554 | ||
1555 | /** | |
1556 | * disk_block_events - block and flush disk event checking | |
1557 | * @disk: disk to block events for | |
1558 | * | |
1559 | * On return from this function, it is guaranteed that event checking | |
1560 | * isn't in progress and won't happen until unblocked by | |
1561 | * disk_unblock_events(). Events blocking is counted and the actual | |
1562 | * unblocking happens after the matching number of unblocks are done. | |
1563 | * | |
1564 | * Note that this intentionally does not block event checking from | |
1565 | * disk_clear_events(). | |
1566 | * | |
1567 | * CONTEXT: | |
1568 | * Might sleep. | |
1569 | */ | |
1570 | void disk_block_events(struct gendisk *disk) | |
1571 | { | |
1572 | struct disk_events *ev = disk->ev; | |
1573 | unsigned long flags; | |
1574 | bool cancel; | |
1575 | ||
1576 | if (!ev) | |
1577 | return; | |
1578 | ||
1579 | /* | |
1580 | * Outer mutex ensures that the first blocker completes canceling | |
1581 | * the event work before further blockers are allowed to finish. | |
1582 | */ | |
1583 | mutex_lock(&ev->block_mutex); | |
1584 | ||
1585 | spin_lock_irqsave(&ev->lock, flags); | |
1586 | cancel = !ev->block++; | |
1587 | spin_unlock_irqrestore(&ev->lock, flags); | |
1588 | ||
1589 | if (cancel) | |
1590 | cancel_delayed_work_sync(&disk->ev->dwork); | |
1591 | ||
1592 | mutex_unlock(&ev->block_mutex); | |
1593 | } | |
1594 | ||
1595 | static void __disk_unblock_events(struct gendisk *disk, bool check_now) | |
1596 | { | |
1597 | struct disk_events *ev = disk->ev; | |
1598 | unsigned long intv; | |
1599 | unsigned long flags; | |
1600 | ||
1601 | spin_lock_irqsave(&ev->lock, flags); | |
1602 | ||
1603 | if (WARN_ON_ONCE(ev->block <= 0)) | |
1604 | goto out_unlock; | |
1605 | ||
1606 | if (--ev->block) | |
1607 | goto out_unlock; | |
1608 | ||
1609 | intv = disk_events_poll_jiffies(disk); | |
1610 | if (check_now) | |
1611 | queue_delayed_work(system_freezable_power_efficient_wq, | |
1612 | &ev->dwork, 0); | |
1613 | else if (intv) | |
1614 | queue_delayed_work(system_freezable_power_efficient_wq, | |
1615 | &ev->dwork, intv); | |
1616 | out_unlock: | |
1617 | spin_unlock_irqrestore(&ev->lock, flags); | |
1618 | } | |
1619 | ||
1620 | /** | |
1621 | * disk_unblock_events - unblock disk event checking | |
1622 | * @disk: disk to unblock events for | |
1623 | * | |
1624 | * Undo disk_block_events(). When the block count reaches zero, it | |
1625 | * starts events polling if configured. | |
1626 | * | |
1627 | * CONTEXT: | |
1628 | * Don't care. Safe to call from irq context. | |
1629 | */ | |
1630 | void disk_unblock_events(struct gendisk *disk) | |
1631 | { | |
1632 | if (disk->ev) | |
1633 | __disk_unblock_events(disk, false); | |
1634 | } | |
1635 | ||
1636 | /** | |
1637 | * disk_flush_events - schedule immediate event checking and flushing | |
1638 | * @disk: disk to check and flush events for | |
1639 | * @mask: events to flush | |
1640 | * | |
1641 | * Schedule immediate event checking on @disk if not blocked. Events in | |
1642 | * @mask are scheduled to be cleared from the driver. Note that this | |
1643 | * doesn't clear the events from @disk->ev. | |
1644 | * | |
1645 | * CONTEXT: | |
1646 | * If @mask is non-zero must be called with bdev->bd_mutex held. | |
1647 | */ | |
1648 | void disk_flush_events(struct gendisk *disk, unsigned int mask) | |
1649 | { | |
1650 | struct disk_events *ev = disk->ev; | |
1651 | ||
1652 | if (!ev) | |
1653 | return; | |
1654 | ||
1655 | spin_lock_irq(&ev->lock); | |
1656 | ev->clearing |= mask; | |
1657 | if (!ev->block) | |
1658 | mod_delayed_work(system_freezable_power_efficient_wq, | |
1659 | &ev->dwork, 0); | |
1660 | spin_unlock_irq(&ev->lock); | |
1661 | } | |
1662 | ||
1663 | /** | |
1664 | * disk_clear_events - synchronously check, clear and return pending events | |
1665 | * @disk: disk to fetch and clear events from | |
1666 | * @mask: mask of events to be fetched and cleared | |
1667 | * | |
1668 | * Disk events are synchronously checked and pending events in @mask | |
1669 | * are cleared and returned. This ignores the block count. | |
1670 | * | |
1671 | * CONTEXT: | |
1672 | * Might sleep. | |
1673 | */ | |
1674 | unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask) | |
1675 | { | |
1676 | const struct block_device_operations *bdops = disk->fops; | |
1677 | struct disk_events *ev = disk->ev; | |
1678 | unsigned int pending; | |
1679 | unsigned int clearing = mask; | |
1680 | ||
1681 | if (!ev) { | |
1682 | /* for drivers still using the old ->media_changed method */ | |
1683 | if ((mask & DISK_EVENT_MEDIA_CHANGE) && | |
1684 | bdops->media_changed && bdops->media_changed(disk)) | |
1685 | return DISK_EVENT_MEDIA_CHANGE; | |
1686 | return 0; | |
1687 | } | |
1688 | ||
1689 | disk_block_events(disk); | |
1690 | ||
1691 | /* | |
1692 | * store the union of mask and ev->clearing on the stack so that the | |
1693 | * race with disk_flush_events does not cause ambiguity (ev->clearing | |
1694 | * can still be modified even if events are blocked). | |
1695 | */ | |
1696 | spin_lock_irq(&ev->lock); | |
1697 | clearing |= ev->clearing; | |
1698 | ev->clearing = 0; | |
1699 | spin_unlock_irq(&ev->lock); | |
1700 | ||
1701 | disk_check_events(ev, &clearing); | |
1702 | /* | |
1703 | * if ev->clearing is not 0, the disk_flush_events got called in the | |
1704 | * middle of this function, so we want to run the workfn without delay. | |
1705 | */ | |
1706 | __disk_unblock_events(disk, ev->clearing ? true : false); | |
1707 | ||
1708 | /* then, fetch and clear pending events */ | |
1709 | spin_lock_irq(&ev->lock); | |
1710 | pending = ev->pending & mask; | |
1711 | ev->pending &= ~mask; | |
1712 | spin_unlock_irq(&ev->lock); | |
1713 | WARN_ON_ONCE(clearing & mask); | |
1714 | ||
1715 | return pending; | |
1716 | } | |
1717 | ||
1718 | /* | |
1719 | * Separate this part out so that a different pointer for clearing_ptr can be | |
1720 | * passed in for disk_clear_events. | |
1721 | */ | |
1722 | static void disk_events_workfn(struct work_struct *work) | |
1723 | { | |
1724 | struct delayed_work *dwork = to_delayed_work(work); | |
1725 | struct disk_events *ev = container_of(dwork, struct disk_events, dwork); | |
1726 | ||
1727 | disk_check_events(ev, &ev->clearing); | |
1728 | } | |
1729 | ||
1730 | static void disk_check_events(struct disk_events *ev, | |
1731 | unsigned int *clearing_ptr) | |
1732 | { | |
1733 | struct gendisk *disk = ev->disk; | |
1734 | char *envp[ARRAY_SIZE(disk_uevents) + 1] = { }; | |
1735 | unsigned int clearing = *clearing_ptr; | |
1736 | unsigned int events; | |
1737 | unsigned long intv; | |
1738 | int nr_events = 0, i; | |
1739 | ||
1740 | /* check events */ | |
1741 | events = disk->fops->check_events(disk, clearing); | |
1742 | ||
1743 | /* accumulate pending events and schedule next poll if necessary */ | |
1744 | spin_lock_irq(&ev->lock); | |
1745 | ||
1746 | events &= ~ev->pending; | |
1747 | ev->pending |= events; | |
1748 | *clearing_ptr &= ~clearing; | |
1749 | ||
1750 | intv = disk_events_poll_jiffies(disk); | |
1751 | if (!ev->block && intv) | |
1752 | queue_delayed_work(system_freezable_power_efficient_wq, | |
1753 | &ev->dwork, intv); | |
1754 | ||
1755 | spin_unlock_irq(&ev->lock); | |
1756 | ||
1757 | /* | |
1758 | * Tell userland about new events. Only the events listed in | |
1759 | * @disk->events are reported. Unlisted events are processed the | |
1760 | * same internally but never get reported to userland. | |
1761 | */ | |
1762 | for (i = 0; i < ARRAY_SIZE(disk_uevents); i++) | |
1763 | if (events & disk->events & (1 << i)) | |
1764 | envp[nr_events++] = disk_uevents[i]; | |
1765 | ||
1766 | if (nr_events) | |
1767 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
1768 | } | |
1769 | ||
1770 | /* | |
1771 | * A disk events enabled device has the following sysfs nodes under | |
1772 | * its /sys/block/X/ directory. | |
1773 | * | |
1774 | * events : list of all supported events | |
1775 | * events_async : list of events which can be detected w/o polling | |
1776 | * events_poll_msecs : polling interval, 0: disable, -1: system default | |
1777 | */ | |
1778 | static ssize_t __disk_events_show(unsigned int events, char *buf) | |
1779 | { | |
1780 | const char *delim = ""; | |
1781 | ssize_t pos = 0; | |
1782 | int i; | |
1783 | ||
1784 | for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++) | |
1785 | if (events & (1 << i)) { | |
1786 | pos += sprintf(buf + pos, "%s%s", | |
1787 | delim, disk_events_strs[i]); | |
1788 | delim = " "; | |
1789 | } | |
1790 | if (pos) | |
1791 | pos += sprintf(buf + pos, "\n"); | |
1792 | return pos; | |
1793 | } | |
1794 | ||
1795 | static ssize_t disk_events_show(struct device *dev, | |
1796 | struct device_attribute *attr, char *buf) | |
1797 | { | |
1798 | struct gendisk *disk = dev_to_disk(dev); | |
1799 | ||
1800 | return __disk_events_show(disk->events, buf); | |
1801 | } | |
1802 | ||
1803 | static ssize_t disk_events_async_show(struct device *dev, | |
1804 | struct device_attribute *attr, char *buf) | |
1805 | { | |
1806 | struct gendisk *disk = dev_to_disk(dev); | |
1807 | ||
1808 | return __disk_events_show(disk->async_events, buf); | |
1809 | } | |
1810 | ||
1811 | static ssize_t disk_events_poll_msecs_show(struct device *dev, | |
1812 | struct device_attribute *attr, | |
1813 | char *buf) | |
1814 | { | |
1815 | struct gendisk *disk = dev_to_disk(dev); | |
1816 | ||
1817 | return sprintf(buf, "%ld\n", disk->ev->poll_msecs); | |
1818 | } | |
1819 | ||
1820 | static ssize_t disk_events_poll_msecs_store(struct device *dev, | |
1821 | struct device_attribute *attr, | |
1822 | const char *buf, size_t count) | |
1823 | { | |
1824 | struct gendisk *disk = dev_to_disk(dev); | |
1825 | long intv; | |
1826 | ||
1827 | if (!count || !sscanf(buf, "%ld", &intv)) | |
1828 | return -EINVAL; | |
1829 | ||
1830 | if (intv < 0 && intv != -1) | |
1831 | return -EINVAL; | |
1832 | ||
1833 | disk_block_events(disk); | |
1834 | disk->ev->poll_msecs = intv; | |
1835 | __disk_unblock_events(disk, true); | |
1836 | ||
1837 | return count; | |
1838 | } | |
1839 | ||
1840 | static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL); | |
1841 | static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL); | |
1842 | static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR, | |
1843 | disk_events_poll_msecs_show, | |
1844 | disk_events_poll_msecs_store); | |
1845 | ||
1846 | static const struct attribute *disk_events_attrs[] = { | |
1847 | &dev_attr_events.attr, | |
1848 | &dev_attr_events_async.attr, | |
1849 | &dev_attr_events_poll_msecs.attr, | |
1850 | NULL, | |
1851 | }; | |
1852 | ||
1853 | /* | |
1854 | * The default polling interval can be specified by the kernel | |
1855 | * parameter block.events_dfl_poll_msecs which defaults to 0 | |
1856 | * (disable). This can also be modified runtime by writing to | |
1857 | * /sys/module/block/events_dfl_poll_msecs. | |
1858 | */ | |
1859 | static int disk_events_set_dfl_poll_msecs(const char *val, | |
1860 | const struct kernel_param *kp) | |
1861 | { | |
1862 | struct disk_events *ev; | |
1863 | int ret; | |
1864 | ||
1865 | ret = param_set_ulong(val, kp); | |
1866 | if (ret < 0) | |
1867 | return ret; | |
1868 | ||
1869 | mutex_lock(&disk_events_mutex); | |
1870 | ||
1871 | list_for_each_entry(ev, &disk_events, node) | |
1872 | disk_flush_events(ev->disk, 0); | |
1873 | ||
1874 | mutex_unlock(&disk_events_mutex); | |
1875 | ||
1876 | return 0; | |
1877 | } | |
1878 | ||
1879 | static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = { | |
1880 | .set = disk_events_set_dfl_poll_msecs, | |
1881 | .get = param_get_ulong, | |
1882 | }; | |
1883 | ||
1884 | #undef MODULE_PARAM_PREFIX | |
1885 | #define MODULE_PARAM_PREFIX "block." | |
1886 | ||
1887 | module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops, | |
1888 | &disk_events_dfl_poll_msecs, 0644); | |
1889 | ||
1890 | /* | |
1891 | * disk_{alloc|add|del|release}_events - initialize and destroy disk_events. | |
1892 | */ | |
1893 | static void disk_alloc_events(struct gendisk *disk) | |
1894 | { | |
1895 | struct disk_events *ev; | |
1896 | ||
1897 | if (!disk->fops->check_events) | |
1898 | return; | |
1899 | ||
1900 | ev = kzalloc(sizeof(*ev), GFP_KERNEL); | |
1901 | if (!ev) { | |
1902 | pr_warn("%s: failed to initialize events\n", disk->disk_name); | |
1903 | return; | |
1904 | } | |
1905 | ||
1906 | INIT_LIST_HEAD(&ev->node); | |
1907 | ev->disk = disk; | |
1908 | spin_lock_init(&ev->lock); | |
1909 | mutex_init(&ev->block_mutex); | |
1910 | ev->block = 1; | |
1911 | ev->poll_msecs = -1; | |
1912 | INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn); | |
1913 | ||
1914 | disk->ev = ev; | |
1915 | } | |
1916 | ||
1917 | static void disk_add_events(struct gendisk *disk) | |
1918 | { | |
1919 | if (!disk->ev) | |
1920 | return; | |
1921 | ||
1922 | /* FIXME: error handling */ | |
1923 | if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0) | |
1924 | pr_warn("%s: failed to create sysfs files for events\n", | |
1925 | disk->disk_name); | |
1926 | ||
1927 | mutex_lock(&disk_events_mutex); | |
1928 | list_add_tail(&disk->ev->node, &disk_events); | |
1929 | mutex_unlock(&disk_events_mutex); | |
1930 | ||
1931 | /* | |
1932 | * Block count is initialized to 1 and the following initial | |
1933 | * unblock kicks it into action. | |
1934 | */ | |
1935 | __disk_unblock_events(disk, true); | |
1936 | } | |
1937 | ||
1938 | static void disk_del_events(struct gendisk *disk) | |
1939 | { | |
1940 | if (!disk->ev) | |
1941 | return; | |
1942 | ||
1943 | disk_block_events(disk); | |
1944 | ||
1945 | mutex_lock(&disk_events_mutex); | |
1946 | list_del_init(&disk->ev->node); | |
1947 | mutex_unlock(&disk_events_mutex); | |
1948 | ||
1949 | sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs); | |
1950 | } | |
1951 | ||
1952 | static void disk_release_events(struct gendisk *disk) | |
1953 | { | |
1954 | /* the block count should be 1 from disk_del_events() */ | |
1955 | WARN_ON_ONCE(disk->ev && disk->ev->block != 1); | |
1956 | kfree(disk->ev); | |
1957 | } |