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