]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * gendisk handling | |
3 | */ | |
4 | ||
1da177e4 LT |
5 | #include <linux/module.h> |
6 | #include <linux/fs.h> | |
7 | #include <linux/genhd.h> | |
b446b60e | 8 | #include <linux/kdev_t.h> |
1da177e4 LT |
9 | #include <linux/kernel.h> |
10 | #include <linux/blkdev.h> | |
11 | #include <linux/init.h> | |
12 | #include <linux/spinlock.h> | |
f500975a | 13 | #include <linux/proc_fs.h> |
1da177e4 LT |
14 | #include <linux/seq_file.h> |
15 | #include <linux/slab.h> | |
16 | #include <linux/kmod.h> | |
17 | #include <linux/kobj_map.h> | |
2ef41634 | 18 | #include <linux/buffer_head.h> |
58383af6 | 19 | #include <linux/mutex.h> |
bcce3de1 | 20 | #include <linux/idr.h> |
77ea887e | 21 | #include <linux/log2.h> |
a72c5e5e | 22 | #include <linux/ctype.h> |
1da177e4 | 23 | |
ff88972c AB |
24 | #include "blk.h" |
25 | ||
edfaa7c3 | 26 | static DEFINE_MUTEX(block_class_lock); |
edfaa7c3 | 27 | struct kobject *block_depr; |
1da177e4 | 28 | |
bcce3de1 TH |
29 | /* for extended dynamic devt allocation, currently only one major is used */ |
30 | #define MAX_EXT_DEVT (1 << MINORBITS) | |
31 | ||
32 | /* For extended devt allocation. ext_devt_mutex prevents look up | |
33 | * results from going away underneath its user. | |
34 | */ | |
35 | static DEFINE_MUTEX(ext_devt_mutex); | |
36 | static DEFINE_IDR(ext_devt_idr); | |
37 | ||
1826eadf AB |
38 | static struct device_type disk_type; |
39 | ||
77ea887e TH |
40 | static void disk_add_events(struct gendisk *disk); |
41 | static void disk_del_events(struct gendisk *disk); | |
42 | static void disk_release_events(struct gendisk *disk); | |
43 | ||
e71bf0d0 TH |
44 | /** |
45 | * disk_get_part - get partition | |
46 | * @disk: disk to look partition from | |
47 | * @partno: partition number | |
48 | * | |
49 | * Look for partition @partno from @disk. If found, increment | |
50 | * reference count and return it. | |
51 | * | |
52 | * CONTEXT: | |
53 | * Don't care. | |
54 | * | |
55 | * RETURNS: | |
56 | * Pointer to the found partition on success, NULL if not found. | |
57 | */ | |
58 | struct hd_struct *disk_get_part(struct gendisk *disk, int partno) | |
59 | { | |
540eed56 TH |
60 | struct hd_struct *part = NULL; |
61 | struct disk_part_tbl *ptbl; | |
e71bf0d0 | 62 | |
540eed56 | 63 | if (unlikely(partno < 0)) |
e71bf0d0 | 64 | return NULL; |
540eed56 | 65 | |
e71bf0d0 | 66 | rcu_read_lock(); |
540eed56 TH |
67 | |
68 | ptbl = rcu_dereference(disk->part_tbl); | |
69 | if (likely(partno < ptbl->len)) { | |
70 | part = rcu_dereference(ptbl->part[partno]); | |
71 | if (part) | |
72 | get_device(part_to_dev(part)); | |
73 | } | |
74 | ||
e71bf0d0 TH |
75 | rcu_read_unlock(); |
76 | ||
77 | return part; | |
78 | } | |
79 | EXPORT_SYMBOL_GPL(disk_get_part); | |
80 | ||
81 | /** | |
82 | * disk_part_iter_init - initialize partition iterator | |
83 | * @piter: iterator to initialize | |
84 | * @disk: disk to iterate over | |
85 | * @flags: DISK_PITER_* flags | |
86 | * | |
87 | * Initialize @piter so that it iterates over partitions of @disk. | |
88 | * | |
89 | * CONTEXT: | |
90 | * Don't care. | |
91 | */ | |
92 | void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk, | |
93 | unsigned int flags) | |
94 | { | |
540eed56 TH |
95 | struct disk_part_tbl *ptbl; |
96 | ||
97 | rcu_read_lock(); | |
98 | ptbl = rcu_dereference(disk->part_tbl); | |
99 | ||
e71bf0d0 TH |
100 | piter->disk = disk; |
101 | piter->part = NULL; | |
102 | ||
103 | if (flags & DISK_PITER_REVERSE) | |
540eed56 | 104 | piter->idx = ptbl->len - 1; |
71982a40 | 105 | else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0)) |
e71bf0d0 | 106 | piter->idx = 0; |
b5d0b9df TH |
107 | else |
108 | piter->idx = 1; | |
e71bf0d0 TH |
109 | |
110 | piter->flags = flags; | |
540eed56 TH |
111 | |
112 | rcu_read_unlock(); | |
e71bf0d0 TH |
113 | } |
114 | EXPORT_SYMBOL_GPL(disk_part_iter_init); | |
115 | ||
116 | /** | |
117 | * disk_part_iter_next - proceed iterator to the next partition and return it | |
118 | * @piter: iterator of interest | |
119 | * | |
120 | * Proceed @piter to the next partition and return it. | |
121 | * | |
122 | * CONTEXT: | |
123 | * Don't care. | |
124 | */ | |
125 | struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter) | |
126 | { | |
540eed56 | 127 | struct disk_part_tbl *ptbl; |
e71bf0d0 TH |
128 | int inc, end; |
129 | ||
130 | /* put the last partition */ | |
131 | disk_put_part(piter->part); | |
132 | piter->part = NULL; | |
133 | ||
540eed56 | 134 | /* get part_tbl */ |
e71bf0d0 | 135 | rcu_read_lock(); |
540eed56 | 136 | ptbl = rcu_dereference(piter->disk->part_tbl); |
e71bf0d0 TH |
137 | |
138 | /* determine iteration parameters */ | |
139 | if (piter->flags & DISK_PITER_REVERSE) { | |
140 | inc = -1; | |
71982a40 TH |
141 | if (piter->flags & (DISK_PITER_INCL_PART0 | |
142 | DISK_PITER_INCL_EMPTY_PART0)) | |
b5d0b9df TH |
143 | end = -1; |
144 | else | |
145 | end = 0; | |
e71bf0d0 TH |
146 | } else { |
147 | inc = 1; | |
540eed56 | 148 | end = ptbl->len; |
e71bf0d0 TH |
149 | } |
150 | ||
151 | /* iterate to the next partition */ | |
152 | for (; piter->idx != end; piter->idx += inc) { | |
153 | struct hd_struct *part; | |
154 | ||
540eed56 | 155 | part = rcu_dereference(ptbl->part[piter->idx]); |
e71bf0d0 TH |
156 | if (!part) |
157 | continue; | |
71982a40 TH |
158 | if (!part->nr_sects && |
159 | !(piter->flags & DISK_PITER_INCL_EMPTY) && | |
160 | !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 && | |
161 | piter->idx == 0)) | |
e71bf0d0 TH |
162 | continue; |
163 | ||
ed9e1982 | 164 | get_device(part_to_dev(part)); |
e71bf0d0 TH |
165 | piter->part = part; |
166 | piter->idx += inc; | |
167 | break; | |
168 | } | |
169 | ||
170 | rcu_read_unlock(); | |
171 | ||
172 | return piter->part; | |
173 | } | |
174 | EXPORT_SYMBOL_GPL(disk_part_iter_next); | |
175 | ||
176 | /** | |
177 | * disk_part_iter_exit - finish up partition iteration | |
178 | * @piter: iter of interest | |
179 | * | |
180 | * Called when iteration is over. Cleans up @piter. | |
181 | * | |
182 | * CONTEXT: | |
183 | * Don't care. | |
184 | */ | |
185 | void disk_part_iter_exit(struct disk_part_iter *piter) | |
186 | { | |
187 | disk_put_part(piter->part); | |
188 | piter->part = NULL; | |
189 | } | |
190 | EXPORT_SYMBOL_GPL(disk_part_iter_exit); | |
191 | ||
a6f23657 JA |
192 | static inline int sector_in_part(struct hd_struct *part, sector_t sector) |
193 | { | |
194 | return part->start_sect <= sector && | |
195 | sector < part->start_sect + part->nr_sects; | |
196 | } | |
197 | ||
e71bf0d0 TH |
198 | /** |
199 | * disk_map_sector_rcu - map sector to partition | |
200 | * @disk: gendisk of interest | |
201 | * @sector: sector to map | |
202 | * | |
203 | * Find out which partition @sector maps to on @disk. This is | |
204 | * primarily used for stats accounting. | |
205 | * | |
206 | * CONTEXT: | |
207 | * RCU read locked. The returned partition pointer is valid only | |
208 | * while preemption is disabled. | |
209 | * | |
210 | * RETURNS: | |
074a7aca | 211 | * Found partition on success, part0 is returned if no partition matches |
e71bf0d0 TH |
212 | */ |
213 | struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector) | |
214 | { | |
540eed56 | 215 | struct disk_part_tbl *ptbl; |
a6f23657 | 216 | struct hd_struct *part; |
e71bf0d0 TH |
217 | int i; |
218 | ||
540eed56 TH |
219 | ptbl = rcu_dereference(disk->part_tbl); |
220 | ||
a6f23657 JA |
221 | part = rcu_dereference(ptbl->last_lookup); |
222 | if (part && sector_in_part(part, sector)) | |
223 | return part; | |
224 | ||
540eed56 | 225 | for (i = 1; i < ptbl->len; i++) { |
a6f23657 | 226 | part = rcu_dereference(ptbl->part[i]); |
e71bf0d0 | 227 | |
a6f23657 JA |
228 | if (part && sector_in_part(part, sector)) { |
229 | rcu_assign_pointer(ptbl->last_lookup, part); | |
e71bf0d0 | 230 | return part; |
a6f23657 | 231 | } |
e71bf0d0 | 232 | } |
074a7aca | 233 | return &disk->part0; |
e71bf0d0 TH |
234 | } |
235 | EXPORT_SYMBOL_GPL(disk_map_sector_rcu); | |
236 | ||
1da177e4 LT |
237 | /* |
238 | * Can be deleted altogether. Later. | |
239 | * | |
240 | */ | |
241 | static struct blk_major_name { | |
242 | struct blk_major_name *next; | |
243 | int major; | |
244 | char name[16]; | |
68eef3b4 | 245 | } *major_names[BLKDEV_MAJOR_HASH_SIZE]; |
1da177e4 LT |
246 | |
247 | /* index in the above - for now: assume no multimajor ranges */ | |
e61eb2e9 | 248 | static inline int major_to_index(unsigned major) |
1da177e4 | 249 | { |
68eef3b4 | 250 | return major % BLKDEV_MAJOR_HASH_SIZE; |
7170be5f NH |
251 | } |
252 | ||
68eef3b4 | 253 | #ifdef CONFIG_PROC_FS |
cf771cb5 | 254 | void blkdev_show(struct seq_file *seqf, off_t offset) |
7170be5f | 255 | { |
68eef3b4 | 256 | struct blk_major_name *dp; |
7170be5f | 257 | |
68eef3b4 | 258 | if (offset < BLKDEV_MAJOR_HASH_SIZE) { |
edfaa7c3 | 259 | mutex_lock(&block_class_lock); |
68eef3b4 | 260 | for (dp = major_names[offset]; dp; dp = dp->next) |
cf771cb5 | 261 | seq_printf(seqf, "%3d %s\n", dp->major, dp->name); |
edfaa7c3 | 262 | mutex_unlock(&block_class_lock); |
1da177e4 | 263 | } |
1da177e4 | 264 | } |
68eef3b4 | 265 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 266 | |
9e8c0bcc MN |
267 | /** |
268 | * register_blkdev - register a new block device | |
269 | * | |
270 | * @major: the requested major device number [1..255]. If @major=0, try to | |
271 | * allocate any unused major number. | |
272 | * @name: the name of the new block device as a zero terminated string | |
273 | * | |
274 | * The @name must be unique within the system. | |
275 | * | |
276 | * The return value depends on the @major input parameter. | |
277 | * - if a major device number was requested in range [1..255] then the | |
278 | * function returns zero on success, or a negative error code | |
279 | * - if any unused major number was requested with @major=0 parameter | |
280 | * then the return value is the allocated major number in range | |
281 | * [1..255] or a negative error code otherwise | |
282 | */ | |
1da177e4 LT |
283 | int register_blkdev(unsigned int major, const char *name) |
284 | { | |
285 | struct blk_major_name **n, *p; | |
286 | int index, ret = 0; | |
287 | ||
edfaa7c3 | 288 | mutex_lock(&block_class_lock); |
1da177e4 LT |
289 | |
290 | /* temporary */ | |
291 | if (major == 0) { | |
292 | for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) { | |
293 | if (major_names[index] == NULL) | |
294 | break; | |
295 | } | |
296 | ||
297 | if (index == 0) { | |
298 | printk("register_blkdev: failed to get major for %s\n", | |
299 | name); | |
300 | ret = -EBUSY; | |
301 | goto out; | |
302 | } | |
303 | major = index; | |
304 | ret = major; | |
305 | } | |
306 | ||
307 | p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL); | |
308 | if (p == NULL) { | |
309 | ret = -ENOMEM; | |
310 | goto out; | |
311 | } | |
312 | ||
313 | p->major = major; | |
314 | strlcpy(p->name, name, sizeof(p->name)); | |
315 | p->next = NULL; | |
316 | index = major_to_index(major); | |
317 | ||
318 | for (n = &major_names[index]; *n; n = &(*n)->next) { | |
319 | if ((*n)->major == major) | |
320 | break; | |
321 | } | |
322 | if (!*n) | |
323 | *n = p; | |
324 | else | |
325 | ret = -EBUSY; | |
326 | ||
327 | if (ret < 0) { | |
328 | printk("register_blkdev: cannot get major %d for %s\n", | |
329 | major, name); | |
330 | kfree(p); | |
331 | } | |
332 | out: | |
edfaa7c3 | 333 | mutex_unlock(&block_class_lock); |
1da177e4 LT |
334 | return ret; |
335 | } | |
336 | ||
337 | EXPORT_SYMBOL(register_blkdev); | |
338 | ||
f4480240 | 339 | void unregister_blkdev(unsigned int major, const char *name) |
1da177e4 LT |
340 | { |
341 | struct blk_major_name **n; | |
342 | struct blk_major_name *p = NULL; | |
343 | int index = major_to_index(major); | |
1da177e4 | 344 | |
edfaa7c3 | 345 | mutex_lock(&block_class_lock); |
1da177e4 LT |
346 | for (n = &major_names[index]; *n; n = &(*n)->next) |
347 | if ((*n)->major == major) | |
348 | break; | |
294462a5 AM |
349 | if (!*n || strcmp((*n)->name, name)) { |
350 | WARN_ON(1); | |
294462a5 | 351 | } else { |
1da177e4 LT |
352 | p = *n; |
353 | *n = p->next; | |
354 | } | |
edfaa7c3 | 355 | mutex_unlock(&block_class_lock); |
1da177e4 | 356 | kfree(p); |
1da177e4 LT |
357 | } |
358 | ||
359 | EXPORT_SYMBOL(unregister_blkdev); | |
360 | ||
361 | static struct kobj_map *bdev_map; | |
362 | ||
870d6656 TH |
363 | /** |
364 | * blk_mangle_minor - scatter minor numbers apart | |
365 | * @minor: minor number to mangle | |
366 | * | |
367 | * Scatter consecutively allocated @minor number apart if MANGLE_DEVT | |
368 | * is enabled. Mangling twice gives the original value. | |
369 | * | |
370 | * RETURNS: | |
371 | * Mangled value. | |
372 | * | |
373 | * CONTEXT: | |
374 | * Don't care. | |
375 | */ | |
376 | static int blk_mangle_minor(int minor) | |
377 | { | |
378 | #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT | |
379 | int i; | |
380 | ||
381 | for (i = 0; i < MINORBITS / 2; i++) { | |
382 | int low = minor & (1 << i); | |
383 | int high = minor & (1 << (MINORBITS - 1 - i)); | |
384 | int distance = MINORBITS - 1 - 2 * i; | |
385 | ||
386 | minor ^= low | high; /* clear both bits */ | |
387 | low <<= distance; /* swap the positions */ | |
388 | high >>= distance; | |
389 | minor |= low | high; /* and set */ | |
390 | } | |
391 | #endif | |
392 | return minor; | |
393 | } | |
394 | ||
bcce3de1 TH |
395 | /** |
396 | * blk_alloc_devt - allocate a dev_t for a partition | |
397 | * @part: partition to allocate dev_t for | |
bcce3de1 TH |
398 | * @devt: out parameter for resulting dev_t |
399 | * | |
400 | * Allocate a dev_t for block device. | |
401 | * | |
402 | * RETURNS: | |
403 | * 0 on success, allocated dev_t is returned in *@devt. -errno on | |
404 | * failure. | |
405 | * | |
406 | * CONTEXT: | |
407 | * Might sleep. | |
408 | */ | |
409 | int blk_alloc_devt(struct hd_struct *part, dev_t *devt) | |
410 | { | |
411 | struct gendisk *disk = part_to_disk(part); | |
412 | int idx, rc; | |
413 | ||
414 | /* in consecutive minor range? */ | |
415 | if (part->partno < disk->minors) { | |
416 | *devt = MKDEV(disk->major, disk->first_minor + part->partno); | |
417 | return 0; | |
418 | } | |
419 | ||
420 | /* allocate ext devt */ | |
421 | do { | |
422 | if (!idr_pre_get(&ext_devt_idr, GFP_KERNEL)) | |
423 | return -ENOMEM; | |
424 | rc = idr_get_new(&ext_devt_idr, part, &idx); | |
425 | } while (rc == -EAGAIN); | |
426 | ||
427 | if (rc) | |
428 | return rc; | |
429 | ||
430 | if (idx > MAX_EXT_DEVT) { | |
431 | idr_remove(&ext_devt_idr, idx); | |
432 | return -EBUSY; | |
433 | } | |
434 | ||
870d6656 | 435 | *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx)); |
bcce3de1 TH |
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 | might_sleep(); | |
451 | ||
452 | if (devt == MKDEV(0, 0)) | |
453 | return; | |
454 | ||
455 | if (MAJOR(devt) == BLOCK_EXT_MAJOR) { | |
456 | mutex_lock(&ext_devt_mutex); | |
870d6656 | 457 | idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); |
bcce3de1 TH |
458 | mutex_unlock(&ext_devt_mutex); |
459 | } | |
460 | } | |
461 | ||
1f014290 TH |
462 | static char *bdevt_str(dev_t devt, char *buf) |
463 | { | |
464 | if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) { | |
465 | char tbuf[BDEVT_SIZE]; | |
466 | snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt)); | |
467 | snprintf(buf, BDEVT_SIZE, "%-9s", tbuf); | |
468 | } else | |
469 | snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt)); | |
470 | ||
471 | return buf; | |
472 | } | |
473 | ||
1da177e4 LT |
474 | /* |
475 | * Register device numbers dev..(dev+range-1) | |
476 | * range must be nonzero | |
477 | * The hash chain is sorted on range, so that subranges can override. | |
478 | */ | |
edfaa7c3 | 479 | void blk_register_region(dev_t devt, unsigned long range, struct module *module, |
1da177e4 LT |
480 | struct kobject *(*probe)(dev_t, int *, void *), |
481 | int (*lock)(dev_t, void *), void *data) | |
482 | { | |
edfaa7c3 | 483 | kobj_map(bdev_map, devt, range, module, probe, lock, data); |
1da177e4 LT |
484 | } |
485 | ||
486 | EXPORT_SYMBOL(blk_register_region); | |
487 | ||
edfaa7c3 | 488 | void blk_unregister_region(dev_t devt, unsigned long range) |
1da177e4 | 489 | { |
edfaa7c3 | 490 | kobj_unmap(bdev_map, devt, range); |
1da177e4 LT |
491 | } |
492 | ||
493 | EXPORT_SYMBOL(blk_unregister_region); | |
494 | ||
cf771cb5 | 495 | static struct kobject *exact_match(dev_t devt, int *partno, void *data) |
1da177e4 LT |
496 | { |
497 | struct gendisk *p = data; | |
edfaa7c3 | 498 | |
ed9e1982 | 499 | return &disk_to_dev(p)->kobj; |
1da177e4 LT |
500 | } |
501 | ||
edfaa7c3 | 502 | static int exact_lock(dev_t devt, void *data) |
1da177e4 LT |
503 | { |
504 | struct gendisk *p = data; | |
505 | ||
506 | if (!get_disk(p)) | |
507 | return -1; | |
508 | return 0; | |
509 | } | |
510 | ||
d2bf1b67 TH |
511 | void register_disk(struct gendisk *disk) |
512 | { | |
513 | struct device *ddev = disk_to_dev(disk); | |
514 | struct block_device *bdev; | |
515 | struct disk_part_iter piter; | |
516 | struct hd_struct *part; | |
517 | int err; | |
518 | ||
519 | ddev->parent = disk->driverfs_dev; | |
520 | ||
521 | dev_set_name(ddev, disk->disk_name); | |
522 | ||
523 | /* delay uevents, until we scanned partition table */ | |
524 | dev_set_uevent_suppress(ddev, 1); | |
525 | ||
526 | if (device_add(ddev)) | |
527 | return; | |
528 | if (!sysfs_deprecated) { | |
529 | err = sysfs_create_link(block_depr, &ddev->kobj, | |
530 | kobject_name(&ddev->kobj)); | |
531 | if (err) { | |
532 | device_del(ddev); | |
533 | return; | |
534 | } | |
535 | } | |
536 | disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj); | |
537 | disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj); | |
538 | ||
539 | /* No minors to use for partitions */ | |
d27769ec | 540 | if (!disk_part_scan_enabled(disk)) |
d2bf1b67 TH |
541 | goto exit; |
542 | ||
543 | /* No such device (e.g., media were just removed) */ | |
544 | if (!get_capacity(disk)) | |
545 | goto exit; | |
546 | ||
547 | bdev = bdget_disk(disk, 0); | |
548 | if (!bdev) | |
549 | goto exit; | |
550 | ||
551 | bdev->bd_invalidated = 1; | |
552 | err = blkdev_get(bdev, FMODE_READ, NULL); | |
553 | if (err < 0) | |
554 | goto exit; | |
555 | blkdev_put(bdev, FMODE_READ); | |
556 | ||
557 | exit: | |
558 | /* announce disk after possible partitions are created */ | |
559 | dev_set_uevent_suppress(ddev, 0); | |
560 | kobject_uevent(&ddev->kobj, KOBJ_ADD); | |
561 | ||
562 | /* announce possible partitions */ | |
563 | disk_part_iter_init(&piter, disk, 0); | |
564 | while ((part = disk_part_iter_next(&piter))) | |
565 | kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD); | |
566 | disk_part_iter_exit(&piter); | |
567 | } | |
568 | ||
1da177e4 LT |
569 | /** |
570 | * add_disk - add partitioning information to kernel list | |
571 | * @disk: per-device partitioning information | |
572 | * | |
573 | * This function registers the partitioning information in @disk | |
574 | * with the kernel. | |
3e1a7ff8 TH |
575 | * |
576 | * FIXME: error handling | |
1da177e4 LT |
577 | */ |
578 | void add_disk(struct gendisk *disk) | |
579 | { | |
cf0ca9fe | 580 | struct backing_dev_info *bdi; |
3e1a7ff8 | 581 | dev_t devt; |
6ffeea77 | 582 | int retval; |
cf0ca9fe | 583 | |
3e1a7ff8 TH |
584 | /* minors == 0 indicates to use ext devt from part0 and should |
585 | * be accompanied with EXT_DEVT flag. Make sure all | |
586 | * parameters make sense. | |
587 | */ | |
588 | WARN_ON(disk->minors && !(disk->major || disk->first_minor)); | |
589 | WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT)); | |
590 | ||
1da177e4 | 591 | disk->flags |= GENHD_FL_UP; |
3e1a7ff8 TH |
592 | |
593 | retval = blk_alloc_devt(&disk->part0, &devt); | |
594 | if (retval) { | |
595 | WARN_ON(1); | |
596 | return; | |
597 | } | |
598 | disk_to_dev(disk)->devt = devt; | |
599 | ||
600 | /* ->major and ->first_minor aren't supposed to be | |
601 | * dereferenced from here on, but set them just in case. | |
602 | */ | |
603 | disk->major = MAJOR(devt); | |
604 | disk->first_minor = MINOR(devt); | |
605 | ||
9f5e4865 | 606 | /* Register BDI before referencing it from bdev */ |
01ea5063 SJK |
607 | bdi = &disk->queue->backing_dev_info; |
608 | bdi_register_dev(bdi, disk_devt(disk)); | |
609 | ||
f331c029 TH |
610 | blk_register_region(disk_devt(disk), disk->minors, NULL, |
611 | exact_match, exact_lock, disk); | |
1da177e4 LT |
612 | register_disk(disk); |
613 | blk_register_queue(disk); | |
cf0ca9fe | 614 | |
523e1d39 TH |
615 | /* |
616 | * Take an extra ref on queue which will be put on disk_release() | |
617 | * so that it sticks around as long as @disk is there. | |
618 | */ | |
619 | WARN_ON_ONCE(blk_get_queue(disk->queue)); | |
620 | ||
ed9e1982 TH |
621 | retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj, |
622 | "bdi"); | |
6ffeea77 | 623 | WARN_ON(retval); |
1da177e4 | 624 | |
77ea887e | 625 | disk_add_events(disk); |
1da177e4 | 626 | } |
1da177e4 | 627 | EXPORT_SYMBOL(add_disk); |
1da177e4 | 628 | |
d2bf1b67 | 629 | void del_gendisk(struct gendisk *disk) |
1da177e4 | 630 | { |
d2bf1b67 TH |
631 | struct disk_part_iter piter; |
632 | struct hd_struct *part; | |
633 | ||
77ea887e TH |
634 | disk_del_events(disk); |
635 | ||
d2bf1b67 TH |
636 | /* invalidate stuff */ |
637 | disk_part_iter_init(&piter, disk, | |
638 | DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE); | |
639 | while ((part = disk_part_iter_next(&piter))) { | |
640 | invalidate_partition(disk, part->partno); | |
641 | delete_partition(disk, part->partno); | |
642 | } | |
643 | disk_part_iter_exit(&piter); | |
644 | ||
645 | invalidate_partition(disk, 0); | |
646 | blk_free_devt(disk_to_dev(disk)->devt); | |
647 | set_capacity(disk, 0); | |
648 | disk->flags &= ~GENHD_FL_UP; | |
649 | ||
ed9e1982 | 650 | sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi"); |
cf0ca9fe | 651 | bdi_unregister(&disk->queue->backing_dev_info); |
1da177e4 | 652 | blk_unregister_queue(disk); |
f331c029 | 653 | blk_unregister_region(disk_devt(disk), disk->minors); |
d2bf1b67 TH |
654 | |
655 | part_stat_set_all(&disk->part0, 0); | |
656 | disk->part0.stamp = 0; | |
657 | ||
658 | kobject_put(disk->part0.holder_dir); | |
659 | kobject_put(disk->slave_dir); | |
660 | disk->driverfs_dev = NULL; | |
661 | if (!sysfs_deprecated) | |
662 | sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk))); | |
663 | device_del(disk_to_dev(disk)); | |
1da177e4 | 664 | } |
d2bf1b67 | 665 | EXPORT_SYMBOL(del_gendisk); |
1da177e4 | 666 | |
1da177e4 LT |
667 | /** |
668 | * get_gendisk - get partitioning information for a given device | |
710027a4 | 669 | * @devt: device to get partitioning information for |
496aa8a9 | 670 | * @partno: returned partition index |
1da177e4 LT |
671 | * |
672 | * This function gets the structure containing partitioning | |
710027a4 | 673 | * information for the given device @devt. |
1da177e4 | 674 | */ |
cf771cb5 | 675 | struct gendisk *get_gendisk(dev_t devt, int *partno) |
1da177e4 | 676 | { |
bcce3de1 TH |
677 | struct gendisk *disk = NULL; |
678 | ||
679 | if (MAJOR(devt) != BLOCK_EXT_MAJOR) { | |
680 | struct kobject *kobj; | |
681 | ||
682 | kobj = kobj_lookup(bdev_map, devt, partno); | |
683 | if (kobj) | |
684 | disk = dev_to_disk(kobj_to_dev(kobj)); | |
685 | } else { | |
686 | struct hd_struct *part; | |
687 | ||
688 | mutex_lock(&ext_devt_mutex); | |
870d6656 | 689 | part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); |
bcce3de1 TH |
690 | if (part && get_disk(part_to_disk(part))) { |
691 | *partno = part->partno; | |
692 | disk = part_to_disk(part); | |
693 | } | |
694 | mutex_unlock(&ext_devt_mutex); | |
695 | } | |
edfaa7c3 | 696 | |
bcce3de1 | 697 | return disk; |
1da177e4 | 698 | } |
b6ac23af | 699 | EXPORT_SYMBOL(get_gendisk); |
1da177e4 | 700 | |
f331c029 TH |
701 | /** |
702 | * bdget_disk - do bdget() by gendisk and partition number | |
703 | * @disk: gendisk of interest | |
704 | * @partno: partition number | |
705 | * | |
706 | * Find partition @partno from @disk, do bdget() on it. | |
707 | * | |
708 | * CONTEXT: | |
709 | * Don't care. | |
710 | * | |
711 | * RETURNS: | |
712 | * Resulting block_device on success, NULL on failure. | |
713 | */ | |
aeb3d3a8 | 714 | struct block_device *bdget_disk(struct gendisk *disk, int partno) |
f331c029 | 715 | { |
548b10eb TH |
716 | struct hd_struct *part; |
717 | struct block_device *bdev = NULL; | |
f331c029 | 718 | |
548b10eb | 719 | part = disk_get_part(disk, partno); |
2bbedcb4 | 720 | if (part) |
548b10eb TH |
721 | bdev = bdget(part_devt(part)); |
722 | disk_put_part(part); | |
f331c029 | 723 | |
548b10eb | 724 | return bdev; |
f331c029 TH |
725 | } |
726 | EXPORT_SYMBOL(bdget_disk); | |
727 | ||
5c6f35c5 GKH |
728 | /* |
729 | * print a full list of all partitions - intended for places where the root | |
730 | * filesystem can't be mounted and thus to give the victim some idea of what | |
731 | * went wrong | |
732 | */ | |
733 | void __init printk_all_partitions(void) | |
734 | { | |
def4e38d TH |
735 | struct class_dev_iter iter; |
736 | struct device *dev; | |
737 | ||
738 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); | |
739 | while ((dev = class_dev_iter_next(&iter))) { | |
740 | struct gendisk *disk = dev_to_disk(dev); | |
e71bf0d0 TH |
741 | struct disk_part_iter piter; |
742 | struct hd_struct *part; | |
1f014290 TH |
743 | char name_buf[BDEVNAME_SIZE]; |
744 | char devt_buf[BDEVT_SIZE]; | |
b5af921e | 745 | u8 uuid[PARTITION_META_INFO_UUIDLTH * 2 + 1]; |
def4e38d TH |
746 | |
747 | /* | |
748 | * Don't show empty devices or things that have been | |
25985edc | 749 | * suppressed |
def4e38d TH |
750 | */ |
751 | if (get_capacity(disk) == 0 || | |
752 | (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)) | |
753 | continue; | |
754 | ||
755 | /* | |
756 | * Note, unlike /proc/partitions, I am showing the | |
757 | * numbers in hex - the same format as the root= | |
758 | * option takes. | |
759 | */ | |
074a7aca TH |
760 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0); |
761 | while ((part = disk_part_iter_next(&piter))) { | |
762 | bool is_part0 = part == &disk->part0; | |
def4e38d | 763 | |
b5af921e WD |
764 | uuid[0] = 0; |
765 | if (part->info) | |
766 | part_unpack_uuid(part->info->uuid, uuid); | |
767 | ||
768 | printk("%s%s %10llu %s %s", is_part0 ? "" : " ", | |
1f014290 | 769 | bdevt_str(part_devt(part), devt_buf), |
f331c029 | 770 | (unsigned long long)part->nr_sects >> 1, |
b5af921e | 771 | disk_name(disk, part->partno, name_buf), uuid); |
074a7aca TH |
772 | if (is_part0) { |
773 | if (disk->driverfs_dev != NULL && | |
774 | disk->driverfs_dev->driver != NULL) | |
775 | printk(" driver: %s\n", | |
776 | disk->driverfs_dev->driver->name); | |
777 | else | |
778 | printk(" (driver?)\n"); | |
779 | } else | |
780 | printk("\n"); | |
781 | } | |
e71bf0d0 | 782 | disk_part_iter_exit(&piter); |
def4e38d TH |
783 | } |
784 | class_dev_iter_exit(&iter); | |
dd2a345f DG |
785 | } |
786 | ||
1da177e4 LT |
787 | #ifdef CONFIG_PROC_FS |
788 | /* iterator */ | |
def4e38d | 789 | static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos) |
68c4d4a7 | 790 | { |
def4e38d TH |
791 | loff_t skip = *pos; |
792 | struct class_dev_iter *iter; | |
793 | struct device *dev; | |
68c4d4a7 | 794 | |
aeb3d3a8 | 795 | iter = kmalloc(sizeof(*iter), GFP_KERNEL); |
def4e38d TH |
796 | if (!iter) |
797 | return ERR_PTR(-ENOMEM); | |
798 | ||
799 | seqf->private = iter; | |
800 | class_dev_iter_init(iter, &block_class, NULL, &disk_type); | |
801 | do { | |
802 | dev = class_dev_iter_next(iter); | |
803 | if (!dev) | |
804 | return NULL; | |
805 | } while (skip--); | |
806 | ||
807 | return dev_to_disk(dev); | |
68c4d4a7 GKH |
808 | } |
809 | ||
def4e38d | 810 | static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos) |
1da177e4 | 811 | { |
edfaa7c3 | 812 | struct device *dev; |
1da177e4 | 813 | |
def4e38d TH |
814 | (*pos)++; |
815 | dev = class_dev_iter_next(seqf->private); | |
2ac3cee5 | 816 | if (dev) |
68c4d4a7 | 817 | return dev_to_disk(dev); |
2ac3cee5 | 818 | |
1da177e4 LT |
819 | return NULL; |
820 | } | |
821 | ||
def4e38d | 822 | static void disk_seqf_stop(struct seq_file *seqf, void *v) |
27f30251 | 823 | { |
def4e38d | 824 | struct class_dev_iter *iter = seqf->private; |
27f30251 | 825 | |
def4e38d TH |
826 | /* stop is called even after start failed :-( */ |
827 | if (iter) { | |
828 | class_dev_iter_exit(iter); | |
829 | kfree(iter); | |
5c0ef6d0 | 830 | } |
1da177e4 LT |
831 | } |
832 | ||
def4e38d | 833 | static void *show_partition_start(struct seq_file *seqf, loff_t *pos) |
1da177e4 | 834 | { |
def4e38d TH |
835 | static void *p; |
836 | ||
837 | p = disk_seqf_start(seqf, pos); | |
b9f985b6 | 838 | if (!IS_ERR_OR_NULL(p) && !*pos) |
def4e38d TH |
839 | seq_puts(seqf, "major minor #blocks name\n\n"); |
840 | return p; | |
1da177e4 LT |
841 | } |
842 | ||
cf771cb5 | 843 | static int show_partition(struct seq_file *seqf, void *v) |
1da177e4 LT |
844 | { |
845 | struct gendisk *sgp = v; | |
e71bf0d0 TH |
846 | struct disk_part_iter piter; |
847 | struct hd_struct *part; | |
1da177e4 LT |
848 | char buf[BDEVNAME_SIZE]; |
849 | ||
1da177e4 | 850 | /* Don't show non-partitionable removeable devices or empty devices */ |
d27769ec | 851 | if (!get_capacity(sgp) || (!disk_max_parts(sgp) && |
f331c029 | 852 | (sgp->flags & GENHD_FL_REMOVABLE))) |
1da177e4 LT |
853 | return 0; |
854 | if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO) | |
855 | return 0; | |
856 | ||
857 | /* show the full disk and all non-0 size partitions of it */ | |
074a7aca | 858 | disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0); |
e71bf0d0 | 859 | while ((part = disk_part_iter_next(&piter))) |
1f014290 | 860 | seq_printf(seqf, "%4d %7d %10llu %s\n", |
f331c029 TH |
861 | MAJOR(part_devt(part)), MINOR(part_devt(part)), |
862 | (unsigned long long)part->nr_sects >> 1, | |
863 | disk_name(sgp, part->partno, buf)); | |
e71bf0d0 | 864 | disk_part_iter_exit(&piter); |
1da177e4 LT |
865 | |
866 | return 0; | |
867 | } | |
868 | ||
f500975a | 869 | static const struct seq_operations partitions_op = { |
def4e38d TH |
870 | .start = show_partition_start, |
871 | .next = disk_seqf_next, | |
872 | .stop = disk_seqf_stop, | |
edfaa7c3 | 873 | .show = show_partition |
1da177e4 | 874 | }; |
f500975a AD |
875 | |
876 | static int partitions_open(struct inode *inode, struct file *file) | |
877 | { | |
878 | return seq_open(file, &partitions_op); | |
879 | } | |
880 | ||
881 | static const struct file_operations proc_partitions_operations = { | |
882 | .open = partitions_open, | |
883 | .read = seq_read, | |
884 | .llseek = seq_lseek, | |
885 | .release = seq_release, | |
886 | }; | |
1da177e4 LT |
887 | #endif |
888 | ||
889 | ||
cf771cb5 | 890 | static struct kobject *base_probe(dev_t devt, int *partno, void *data) |
1da177e4 | 891 | { |
edfaa7c3 | 892 | if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0) |
1da177e4 | 893 | /* Make old-style 2.4 aliases work */ |
edfaa7c3 | 894 | request_module("block-major-%d", MAJOR(devt)); |
1da177e4 LT |
895 | return NULL; |
896 | } | |
897 | ||
898 | static int __init genhd_device_init(void) | |
899 | { | |
e105b8bf DW |
900 | int error; |
901 | ||
902 | block_class.dev_kobj = sysfs_dev_block_kobj; | |
903 | error = class_register(&block_class); | |
ee27a558 RM |
904 | if (unlikely(error)) |
905 | return error; | |
edfaa7c3 | 906 | bdev_map = kobj_map_init(base_probe, &block_class_lock); |
1da177e4 | 907 | blk_dev_init(); |
edfaa7c3 | 908 | |
561ec68e ZY |
909 | register_blkdev(BLOCK_EXT_MAJOR, "blkext"); |
910 | ||
edfaa7c3 | 911 | /* create top-level block dir */ |
e52eec13 AK |
912 | if (!sysfs_deprecated) |
913 | block_depr = kobject_create_and_add("block", NULL); | |
830d3cfb | 914 | return 0; |
1da177e4 LT |
915 | } |
916 | ||
917 | subsys_initcall(genhd_device_init); | |
918 | ||
a72c5e5e NN |
919 | static ssize_t alias_show(struct device *dev, |
920 | struct device_attribute *attr, char *buf) | |
921 | { | |
922 | struct gendisk *disk = dev_to_disk(dev); | |
923 | ssize_t ret = 0; | |
924 | ||
925 | if (disk->alias) | |
926 | ret = snprintf(buf, ALIAS_LEN, "%s\n", disk->alias); | |
927 | return ret; | |
928 | } | |
929 | ||
930 | static ssize_t alias_store(struct device *dev, struct device_attribute *attr, | |
931 | const char *buf, size_t count) | |
932 | { | |
933 | struct gendisk *disk = dev_to_disk(dev); | |
934 | char *alias; | |
935 | char *envp[] = { NULL, NULL }; | |
936 | unsigned char c; | |
937 | int i; | |
938 | ssize_t ret = count; | |
939 | ||
940 | if (!count) | |
941 | return -EINVAL; | |
942 | ||
943 | if (count >= ALIAS_LEN) { | |
944 | printk(KERN_ERR "alias: alias is too long\n"); | |
945 | return -EINVAL; | |
946 | } | |
947 | ||
948 | /* Validation check */ | |
949 | for (i = 0; i < count; i++) { | |
950 | c = buf[i]; | |
951 | if (i == count - 1 && c == '\n') | |
952 | break; | |
953 | if (!isalnum(c) && c != '_' && c != '-') { | |
954 | printk(KERN_ERR "alias: invalid alias\n"); | |
955 | return -EINVAL; | |
956 | } | |
957 | } | |
958 | ||
959 | if (disk->alias) { | |
960 | printk(KERN_INFO "alias: %s is already assigned (%s)\n", | |
961 | disk->disk_name, disk->alias); | |
962 | return -EINVAL; | |
963 | } | |
964 | ||
965 | alias = kasprintf(GFP_KERNEL, "%s", buf); | |
966 | if (!alias) | |
967 | return -ENOMEM; | |
968 | ||
969 | if (alias[count - 1] == '\n') | |
970 | alias[count - 1] = '\0'; | |
971 | ||
972 | envp[0] = kasprintf(GFP_KERNEL, "ALIAS=%s", alias); | |
973 | if (!envp[0]) { | |
974 | kfree(alias); | |
975 | return -ENOMEM; | |
976 | } | |
977 | ||
978 | disk->alias = alias; | |
979 | printk(KERN_INFO "alias: assigned %s to %s\n", alias, disk->disk_name); | |
980 | ||
981 | kobject_uevent_env(&dev->kobj, KOBJ_ADD, envp); | |
982 | ||
983 | kfree(envp[0]); | |
984 | return ret; | |
985 | } | |
986 | ||
edfaa7c3 KS |
987 | static ssize_t disk_range_show(struct device *dev, |
988 | struct device_attribute *attr, char *buf) | |
1da177e4 | 989 | { |
edfaa7c3 | 990 | struct gendisk *disk = dev_to_disk(dev); |
1da177e4 | 991 | |
edfaa7c3 | 992 | return sprintf(buf, "%d\n", disk->minors); |
1da177e4 LT |
993 | } |
994 | ||
1f014290 TH |
995 | static ssize_t disk_ext_range_show(struct device *dev, |
996 | struct device_attribute *attr, char *buf) | |
997 | { | |
998 | struct gendisk *disk = dev_to_disk(dev); | |
999 | ||
b5d0b9df | 1000 | return sprintf(buf, "%d\n", disk_max_parts(disk)); |
1f014290 TH |
1001 | } |
1002 | ||
edfaa7c3 KS |
1003 | static ssize_t disk_removable_show(struct device *dev, |
1004 | struct device_attribute *attr, char *buf) | |
a7fd6706 | 1005 | { |
edfaa7c3 | 1006 | struct gendisk *disk = dev_to_disk(dev); |
a7fd6706 | 1007 | |
edfaa7c3 KS |
1008 | return sprintf(buf, "%d\n", |
1009 | (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0)); | |
a7fd6706 KS |
1010 | } |
1011 | ||
1c9ce527 KS |
1012 | static ssize_t disk_ro_show(struct device *dev, |
1013 | struct device_attribute *attr, char *buf) | |
1014 | { | |
1015 | struct gendisk *disk = dev_to_disk(dev); | |
1016 | ||
b7db9956 | 1017 | return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0); |
1c9ce527 KS |
1018 | } |
1019 | ||
edfaa7c3 KS |
1020 | static ssize_t disk_capability_show(struct device *dev, |
1021 | struct device_attribute *attr, char *buf) | |
86ce18d7 | 1022 | { |
edfaa7c3 KS |
1023 | struct gendisk *disk = dev_to_disk(dev); |
1024 | ||
1025 | return sprintf(buf, "%x\n", disk->flags); | |
86ce18d7 | 1026 | } |
edfaa7c3 | 1027 | |
c72758f3 MP |
1028 | static ssize_t disk_alignment_offset_show(struct device *dev, |
1029 | struct device_attribute *attr, | |
1030 | char *buf) | |
1031 | { | |
1032 | struct gendisk *disk = dev_to_disk(dev); | |
1033 | ||
1034 | return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue)); | |
1035 | } | |
1036 | ||
86b37281 MP |
1037 | static ssize_t disk_discard_alignment_show(struct device *dev, |
1038 | struct device_attribute *attr, | |
1039 | char *buf) | |
1040 | { | |
1041 | struct gendisk *disk = dev_to_disk(dev); | |
1042 | ||
dd3d145d | 1043 | return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue)); |
86b37281 MP |
1044 | } |
1045 | ||
a72c5e5e | 1046 | static DEVICE_ATTR(alias, S_IRUGO|S_IWUSR, alias_show, alias_store); |
edfaa7c3 | 1047 | static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL); |
1f014290 | 1048 | static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL); |
edfaa7c3 | 1049 | static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL); |
1c9ce527 | 1050 | static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL); |
e5610521 | 1051 | static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL); |
c72758f3 | 1052 | static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL); |
86b37281 MP |
1053 | static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show, |
1054 | NULL); | |
edfaa7c3 | 1055 | static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL); |
074a7aca | 1056 | static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL); |
316d315b | 1057 | static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL); |
c17bb495 | 1058 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
edfaa7c3 | 1059 | static struct device_attribute dev_attr_fail = |
eddb2e26 | 1060 | __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store); |
c17bb495 | 1061 | #endif |
581d4e28 JA |
1062 | #ifdef CONFIG_FAIL_IO_TIMEOUT |
1063 | static struct device_attribute dev_attr_fail_timeout = | |
1064 | __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show, | |
1065 | part_timeout_store); | |
1066 | #endif | |
edfaa7c3 KS |
1067 | |
1068 | static struct attribute *disk_attrs[] = { | |
a72c5e5e | 1069 | &dev_attr_alias.attr, |
edfaa7c3 | 1070 | &dev_attr_range.attr, |
1f014290 | 1071 | &dev_attr_ext_range.attr, |
edfaa7c3 | 1072 | &dev_attr_removable.attr, |
1c9ce527 | 1073 | &dev_attr_ro.attr, |
edfaa7c3 | 1074 | &dev_attr_size.attr, |
c72758f3 | 1075 | &dev_attr_alignment_offset.attr, |
86b37281 | 1076 | &dev_attr_discard_alignment.attr, |
edfaa7c3 KS |
1077 | &dev_attr_capability.attr, |
1078 | &dev_attr_stat.attr, | |
316d315b | 1079 | &dev_attr_inflight.attr, |
edfaa7c3 KS |
1080 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
1081 | &dev_attr_fail.attr, | |
581d4e28 JA |
1082 | #endif |
1083 | #ifdef CONFIG_FAIL_IO_TIMEOUT | |
1084 | &dev_attr_fail_timeout.attr, | |
edfaa7c3 KS |
1085 | #endif |
1086 | NULL | |
1087 | }; | |
1088 | ||
1089 | static struct attribute_group disk_attr_group = { | |
1090 | .attrs = disk_attrs, | |
1091 | }; | |
1092 | ||
a4dbd674 | 1093 | static const struct attribute_group *disk_attr_groups[] = { |
edfaa7c3 KS |
1094 | &disk_attr_group, |
1095 | NULL | |
1da177e4 LT |
1096 | }; |
1097 | ||
540eed56 TH |
1098 | /** |
1099 | * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way | |
1100 | * @disk: disk to replace part_tbl for | |
1101 | * @new_ptbl: new part_tbl to install | |
1102 | * | |
1103 | * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The | |
1104 | * original ptbl is freed using RCU callback. | |
1105 | * | |
1106 | * LOCKING: | |
1107 | * Matching bd_mutx locked. | |
1108 | */ | |
1109 | static void disk_replace_part_tbl(struct gendisk *disk, | |
1110 | struct disk_part_tbl *new_ptbl) | |
1111 | { | |
1112 | struct disk_part_tbl *old_ptbl = disk->part_tbl; | |
1113 | ||
1114 | rcu_assign_pointer(disk->part_tbl, new_ptbl); | |
a6f23657 JA |
1115 | |
1116 | if (old_ptbl) { | |
1117 | rcu_assign_pointer(old_ptbl->last_lookup, NULL); | |
57bdfbf9 | 1118 | kfree_rcu(old_ptbl, rcu_head); |
a6f23657 | 1119 | } |
540eed56 TH |
1120 | } |
1121 | ||
1122 | /** | |
1123 | * disk_expand_part_tbl - expand disk->part_tbl | |
1124 | * @disk: disk to expand part_tbl for | |
1125 | * @partno: expand such that this partno can fit in | |
1126 | * | |
1127 | * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl | |
1128 | * uses RCU to allow unlocked dereferencing for stats and other stuff. | |
1129 | * | |
1130 | * LOCKING: | |
1131 | * Matching bd_mutex locked, might sleep. | |
1132 | * | |
1133 | * RETURNS: | |
1134 | * 0 on success, -errno on failure. | |
1135 | */ | |
1136 | int disk_expand_part_tbl(struct gendisk *disk, int partno) | |
1137 | { | |
1138 | struct disk_part_tbl *old_ptbl = disk->part_tbl; | |
1139 | struct disk_part_tbl *new_ptbl; | |
1140 | int len = old_ptbl ? old_ptbl->len : 0; | |
1141 | int target = partno + 1; | |
1142 | size_t size; | |
1143 | int i; | |
1144 | ||
1145 | /* disk_max_parts() is zero during initialization, ignore if so */ | |
1146 | if (disk_max_parts(disk) && target > disk_max_parts(disk)) | |
1147 | return -EINVAL; | |
1148 | ||
1149 | if (target <= len) | |
1150 | return 0; | |
1151 | ||
1152 | size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]); | |
1153 | new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id); | |
1154 | if (!new_ptbl) | |
1155 | return -ENOMEM; | |
1156 | ||
540eed56 TH |
1157 | new_ptbl->len = target; |
1158 | ||
1159 | for (i = 0; i < len; i++) | |
1160 | rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]); | |
1161 | ||
1162 | disk_replace_part_tbl(disk, new_ptbl); | |
1163 | return 0; | |
1164 | } | |
1165 | ||
edfaa7c3 | 1166 | static void disk_release(struct device *dev) |
1da177e4 | 1167 | { |
edfaa7c3 KS |
1168 | struct gendisk *disk = dev_to_disk(dev); |
1169 | ||
77ea887e | 1170 | disk_release_events(disk); |
1da177e4 | 1171 | kfree(disk->random); |
540eed56 | 1172 | disk_replace_part_tbl(disk, NULL); |
074a7aca | 1173 | free_part_stats(&disk->part0); |
6d1d8050 | 1174 | free_part_info(&disk->part0); |
523e1d39 TH |
1175 | if (disk->queue) |
1176 | blk_put_queue(disk->queue); | |
1da177e4 LT |
1177 | kfree(disk); |
1178 | } | |
edfaa7c3 KS |
1179 | struct class block_class = { |
1180 | .name = "block", | |
1da177e4 LT |
1181 | }; |
1182 | ||
e454cea2 | 1183 | static char *block_devnode(struct device *dev, mode_t *mode) |
b03f38b6 KS |
1184 | { |
1185 | struct gendisk *disk = dev_to_disk(dev); | |
1186 | ||
e454cea2 KS |
1187 | if (disk->devnode) |
1188 | return disk->devnode(disk, mode); | |
b03f38b6 KS |
1189 | return NULL; |
1190 | } | |
1191 | ||
1826eadf | 1192 | static struct device_type disk_type = { |
edfaa7c3 KS |
1193 | .name = "disk", |
1194 | .groups = disk_attr_groups, | |
1195 | .release = disk_release, | |
e454cea2 | 1196 | .devnode = block_devnode, |
1da177e4 LT |
1197 | }; |
1198 | ||
a6e2ba88 | 1199 | #ifdef CONFIG_PROC_FS |
cf771cb5 TH |
1200 | /* |
1201 | * aggregate disk stat collector. Uses the same stats that the sysfs | |
1202 | * entries do, above, but makes them available through one seq_file. | |
1203 | * | |
1204 | * The output looks suspiciously like /proc/partitions with a bunch of | |
1205 | * extra fields. | |
1206 | */ | |
1207 | static int diskstats_show(struct seq_file *seqf, void *v) | |
1da177e4 LT |
1208 | { |
1209 | struct gendisk *gp = v; | |
e71bf0d0 TH |
1210 | struct disk_part_iter piter; |
1211 | struct hd_struct *hd; | |
1da177e4 | 1212 | char buf[BDEVNAME_SIZE]; |
c9959059 | 1213 | int cpu; |
1da177e4 LT |
1214 | |
1215 | /* | |
ed9e1982 | 1216 | if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next) |
cf771cb5 | 1217 | seq_puts(seqf, "major minor name" |
1da177e4 LT |
1218 | " rio rmerge rsect ruse wio wmerge " |
1219 | "wsect wuse running use aveq" | |
1220 | "\n\n"); | |
1221 | */ | |
9f5e4865 | 1222 | |
71982a40 | 1223 | disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0); |
e71bf0d0 | 1224 | while ((hd = disk_part_iter_next(&piter))) { |
074a7aca | 1225 | cpu = part_stat_lock(); |
c9959059 | 1226 | part_round_stats(cpu, hd); |
074a7aca | 1227 | part_stat_unlock(); |
f95fe9cf HP |
1228 | seq_printf(seqf, "%4d %7d %s %lu %lu %lu " |
1229 | "%u %lu %lu %lu %u %u %u %u\n", | |
f331c029 TH |
1230 | MAJOR(part_devt(hd)), MINOR(part_devt(hd)), |
1231 | disk_name(gp, hd->partno, buf), | |
53f22956 LY |
1232 | part_stat_read(hd, ios[READ]), |
1233 | part_stat_read(hd, merges[READ]), | |
f95fe9cf | 1234 | part_stat_read(hd, sectors[READ]), |
53f22956 LY |
1235 | jiffies_to_msecs(part_stat_read(hd, ticks[READ])), |
1236 | part_stat_read(hd, ios[WRITE]), | |
1237 | part_stat_read(hd, merges[WRITE]), | |
f95fe9cf | 1238 | part_stat_read(hd, sectors[WRITE]), |
53f22956 | 1239 | jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])), |
316d315b | 1240 | part_in_flight(hd), |
28f39d55 JM |
1241 | jiffies_to_msecs(part_stat_read(hd, io_ticks)), |
1242 | jiffies_to_msecs(part_stat_read(hd, time_in_queue)) | |
1243 | ); | |
1da177e4 | 1244 | } |
e71bf0d0 | 1245 | disk_part_iter_exit(&piter); |
9f5e4865 | 1246 | |
1da177e4 LT |
1247 | return 0; |
1248 | } | |
1249 | ||
31d85ab2 | 1250 | static const struct seq_operations diskstats_op = { |
def4e38d TH |
1251 | .start = disk_seqf_start, |
1252 | .next = disk_seqf_next, | |
1253 | .stop = disk_seqf_stop, | |
1da177e4 LT |
1254 | .show = diskstats_show |
1255 | }; | |
f500975a | 1256 | |
31d85ab2 AD |
1257 | static int diskstats_open(struct inode *inode, struct file *file) |
1258 | { | |
1259 | return seq_open(file, &diskstats_op); | |
1260 | } | |
1261 | ||
1262 | static const struct file_operations proc_diskstats_operations = { | |
1263 | .open = diskstats_open, | |
1264 | .read = seq_read, | |
1265 | .llseek = seq_lseek, | |
1266 | .release = seq_release, | |
1267 | }; | |
1268 | ||
f500975a AD |
1269 | static int __init proc_genhd_init(void) |
1270 | { | |
31d85ab2 | 1271 | proc_create("diskstats", 0, NULL, &proc_diskstats_operations); |
f500975a AD |
1272 | proc_create("partitions", 0, NULL, &proc_partitions_operations); |
1273 | return 0; | |
1274 | } | |
1275 | module_init(proc_genhd_init); | |
a6e2ba88 | 1276 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 1277 | |
cf771cb5 | 1278 | dev_t blk_lookup_devt(const char *name, int partno) |
a142be85 | 1279 | { |
def4e38d TH |
1280 | dev_t devt = MKDEV(0, 0); |
1281 | struct class_dev_iter iter; | |
1282 | struct device *dev; | |
a142be85 | 1283 | |
def4e38d TH |
1284 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); |
1285 | while ((dev = class_dev_iter_next(&iter))) { | |
a142be85 | 1286 | struct gendisk *disk = dev_to_disk(dev); |
548b10eb | 1287 | struct hd_struct *part; |
a142be85 | 1288 | |
3ada8b7e | 1289 | if (strcmp(dev_name(dev), name)) |
f331c029 | 1290 | continue; |
f331c029 | 1291 | |
41b8c853 NB |
1292 | if (partno < disk->minors) { |
1293 | /* We need to return the right devno, even | |
1294 | * if the partition doesn't exist yet. | |
1295 | */ | |
1296 | devt = MKDEV(MAJOR(dev->devt), | |
1297 | MINOR(dev->devt) + partno); | |
1298 | break; | |
1299 | } | |
548b10eb | 1300 | part = disk_get_part(disk, partno); |
2bbedcb4 | 1301 | if (part) { |
f331c029 | 1302 | devt = part_devt(part); |
e71bf0d0 | 1303 | disk_put_part(part); |
548b10eb | 1304 | break; |
def4e38d | 1305 | } |
548b10eb | 1306 | disk_put_part(part); |
5c0ef6d0 | 1307 | } |
def4e38d | 1308 | class_dev_iter_exit(&iter); |
edfaa7c3 KS |
1309 | return devt; |
1310 | } | |
edfaa7c3 KS |
1311 | EXPORT_SYMBOL(blk_lookup_devt); |
1312 | ||
1da177e4 LT |
1313 | struct gendisk *alloc_disk(int minors) |
1314 | { | |
1946089a CL |
1315 | return alloc_disk_node(minors, -1); |
1316 | } | |
689d6fac | 1317 | EXPORT_SYMBOL(alloc_disk); |
1946089a CL |
1318 | |
1319 | struct gendisk *alloc_disk_node(int minors, int node_id) | |
1320 | { | |
1321 | struct gendisk *disk; | |
1322 | ||
94f6030c CL |
1323 | disk = kmalloc_node(sizeof(struct gendisk), |
1324 | GFP_KERNEL | __GFP_ZERO, node_id); | |
1da177e4 | 1325 | if (disk) { |
074a7aca | 1326 | if (!init_part_stats(&disk->part0)) { |
1da177e4 LT |
1327 | kfree(disk); |
1328 | return NULL; | |
1329 | } | |
bf91db18 | 1330 | disk->node_id = node_id; |
540eed56 TH |
1331 | if (disk_expand_part_tbl(disk, 0)) { |
1332 | free_part_stats(&disk->part0); | |
b5d0b9df TH |
1333 | kfree(disk); |
1334 | return NULL; | |
1da177e4 | 1335 | } |
540eed56 | 1336 | disk->part_tbl->part[0] = &disk->part0; |
6c23a968 JA |
1337 | |
1338 | hd_ref_init(&disk->part0); | |
b5d0b9df | 1339 | |
1da177e4 | 1340 | disk->minors = minors; |
1da177e4 | 1341 | rand_initialize_disk(disk); |
ed9e1982 TH |
1342 | disk_to_dev(disk)->class = &block_class; |
1343 | disk_to_dev(disk)->type = &disk_type; | |
1344 | device_initialize(disk_to_dev(disk)); | |
1da177e4 LT |
1345 | } |
1346 | return disk; | |
1347 | } | |
1946089a | 1348 | EXPORT_SYMBOL(alloc_disk_node); |
1da177e4 LT |
1349 | |
1350 | struct kobject *get_disk(struct gendisk *disk) | |
1351 | { | |
1352 | struct module *owner; | |
1353 | struct kobject *kobj; | |
1354 | ||
1355 | if (!disk->fops) | |
1356 | return NULL; | |
1357 | owner = disk->fops->owner; | |
1358 | if (owner && !try_module_get(owner)) | |
1359 | return NULL; | |
ed9e1982 | 1360 | kobj = kobject_get(&disk_to_dev(disk)->kobj); |
1da177e4 LT |
1361 | if (kobj == NULL) { |
1362 | module_put(owner); | |
1363 | return NULL; | |
1364 | } | |
1365 | return kobj; | |
1366 | ||
1367 | } | |
1368 | ||
1369 | EXPORT_SYMBOL(get_disk); | |
1370 | ||
1371 | void put_disk(struct gendisk *disk) | |
1372 | { | |
1373 | if (disk) | |
ed9e1982 | 1374 | kobject_put(&disk_to_dev(disk)->kobj); |
1da177e4 LT |
1375 | } |
1376 | ||
1377 | EXPORT_SYMBOL(put_disk); | |
1378 | ||
e3264a4d HR |
1379 | static void set_disk_ro_uevent(struct gendisk *gd, int ro) |
1380 | { | |
1381 | char event[] = "DISK_RO=1"; | |
1382 | char *envp[] = { event, NULL }; | |
1383 | ||
1384 | if (!ro) | |
1385 | event[8] = '0'; | |
1386 | kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); | |
1387 | } | |
1388 | ||
1da177e4 LT |
1389 | void set_device_ro(struct block_device *bdev, int flag) |
1390 | { | |
b7db9956 | 1391 | bdev->bd_part->policy = flag; |
1da177e4 LT |
1392 | } |
1393 | ||
1394 | EXPORT_SYMBOL(set_device_ro); | |
1395 | ||
1396 | void set_disk_ro(struct gendisk *disk, int flag) | |
1397 | { | |
e71bf0d0 TH |
1398 | struct disk_part_iter piter; |
1399 | struct hd_struct *part; | |
1400 | ||
e3264a4d HR |
1401 | if (disk->part0.policy != flag) { |
1402 | set_disk_ro_uevent(disk, flag); | |
1403 | disk->part0.policy = flag; | |
1404 | } | |
1405 | ||
1406 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY); | |
e71bf0d0 TH |
1407 | while ((part = disk_part_iter_next(&piter))) |
1408 | part->policy = flag; | |
1409 | disk_part_iter_exit(&piter); | |
1da177e4 LT |
1410 | } |
1411 | ||
1412 | EXPORT_SYMBOL(set_disk_ro); | |
1413 | ||
1414 | int bdev_read_only(struct block_device *bdev) | |
1415 | { | |
1416 | if (!bdev) | |
1417 | return 0; | |
b7db9956 | 1418 | return bdev->bd_part->policy; |
1da177e4 LT |
1419 | } |
1420 | ||
1421 | EXPORT_SYMBOL(bdev_read_only); | |
1422 | ||
cf771cb5 | 1423 | int invalidate_partition(struct gendisk *disk, int partno) |
1da177e4 LT |
1424 | { |
1425 | int res = 0; | |
cf771cb5 | 1426 | struct block_device *bdev = bdget_disk(disk, partno); |
1da177e4 | 1427 | if (bdev) { |
2ef41634 | 1428 | fsync_bdev(bdev); |
93b270f7 | 1429 | res = __invalidate_device(bdev, true); |
1da177e4 LT |
1430 | bdput(bdev); |
1431 | } | |
1432 | return res; | |
1433 | } | |
1434 | ||
1435 | EXPORT_SYMBOL(invalidate_partition); | |
77ea887e TH |
1436 | |
1437 | /* | |
1438 | * Disk events - monitor disk events like media change and eject request. | |
1439 | */ | |
1440 | struct disk_events { | |
1441 | struct list_head node; /* all disk_event's */ | |
1442 | struct gendisk *disk; /* the associated disk */ | |
1443 | spinlock_t lock; | |
1444 | ||
fdd514e1 | 1445 | struct mutex block_mutex; /* protects blocking */ |
77ea887e TH |
1446 | int block; /* event blocking depth */ |
1447 | unsigned int pending; /* events already sent out */ | |
1448 | unsigned int clearing; /* events being cleared */ | |
1449 | ||
1450 | long poll_msecs; /* interval, -1 for default */ | |
1451 | struct delayed_work dwork; | |
1452 | }; | |
1453 | ||
1454 | static const char *disk_events_strs[] = { | |
1455 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change", | |
1456 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request", | |
1457 | }; | |
1458 | ||
1459 | static char *disk_uevents[] = { | |
1460 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1", | |
1461 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1", | |
1462 | }; | |
1463 | ||
1464 | /* list of all disk_events */ | |
1465 | static DEFINE_MUTEX(disk_events_mutex); | |
1466 | static LIST_HEAD(disk_events); | |
1467 | ||
1468 | /* disable in-kernel polling by default */ | |
1469 | static unsigned long disk_events_dfl_poll_msecs = 0; | |
1470 | ||
1471 | static unsigned long disk_events_poll_jiffies(struct gendisk *disk) | |
1472 | { | |
1473 | struct disk_events *ev = disk->ev; | |
1474 | long intv_msecs = 0; | |
1475 | ||
1476 | /* | |
1477 | * If device-specific poll interval is set, always use it. If | |
1478 | * the default is being used, poll iff there are events which | |
1479 | * can't be monitored asynchronously. | |
1480 | */ | |
1481 | if (ev->poll_msecs >= 0) | |
1482 | intv_msecs = ev->poll_msecs; | |
1483 | else if (disk->events & ~disk->async_events) | |
1484 | intv_msecs = disk_events_dfl_poll_msecs; | |
1485 | ||
1486 | return msecs_to_jiffies(intv_msecs); | |
1487 | } | |
1488 | ||
c3af54af TH |
1489 | /** |
1490 | * disk_block_events - block and flush disk event checking | |
1491 | * @disk: disk to block events for | |
1492 | * | |
1493 | * On return from this function, it is guaranteed that event checking | |
1494 | * isn't in progress and won't happen until unblocked by | |
1495 | * disk_unblock_events(). Events blocking is counted and the actual | |
1496 | * unblocking happens after the matching number of unblocks are done. | |
1497 | * | |
1498 | * Note that this intentionally does not block event checking from | |
1499 | * disk_clear_events(). | |
1500 | * | |
1501 | * CONTEXT: | |
1502 | * Might sleep. | |
1503 | */ | |
1504 | void disk_block_events(struct gendisk *disk) | |
77ea887e TH |
1505 | { |
1506 | struct disk_events *ev = disk->ev; | |
1507 | unsigned long flags; | |
1508 | bool cancel; | |
1509 | ||
c3af54af TH |
1510 | if (!ev) |
1511 | return; | |
1512 | ||
fdd514e1 TH |
1513 | /* |
1514 | * Outer mutex ensures that the first blocker completes canceling | |
1515 | * the event work before further blockers are allowed to finish. | |
1516 | */ | |
1517 | mutex_lock(&ev->block_mutex); | |
1518 | ||
77ea887e TH |
1519 | spin_lock_irqsave(&ev->lock, flags); |
1520 | cancel = !ev->block++; | |
1521 | spin_unlock_irqrestore(&ev->lock, flags); | |
1522 | ||
c3af54af TH |
1523 | if (cancel) |
1524 | cancel_delayed_work_sync(&disk->ev->dwork); | |
fdd514e1 TH |
1525 | |
1526 | mutex_unlock(&ev->block_mutex); | |
77ea887e TH |
1527 | } |
1528 | ||
1529 | static void __disk_unblock_events(struct gendisk *disk, bool check_now) | |
1530 | { | |
1531 | struct disk_events *ev = disk->ev; | |
1532 | unsigned long intv; | |
1533 | unsigned long flags; | |
1534 | ||
1535 | spin_lock_irqsave(&ev->lock, flags); | |
1536 | ||
1537 | if (WARN_ON_ONCE(ev->block <= 0)) | |
1538 | goto out_unlock; | |
1539 | ||
1540 | if (--ev->block) | |
1541 | goto out_unlock; | |
1542 | ||
1543 | /* | |
1544 | * Not exactly a latency critical operation, set poll timer | |
1545 | * slack to 25% and kick event check. | |
1546 | */ | |
1547 | intv = disk_events_poll_jiffies(disk); | |
1548 | set_timer_slack(&ev->dwork.timer, intv / 4); | |
1549 | if (check_now) | |
1550 | queue_delayed_work(system_nrt_wq, &ev->dwork, 0); | |
1551 | else if (intv) | |
1552 | queue_delayed_work(system_nrt_wq, &ev->dwork, intv); | |
1553 | out_unlock: | |
1554 | spin_unlock_irqrestore(&ev->lock, flags); | |
1555 | } | |
1556 | ||
77ea887e TH |
1557 | /** |
1558 | * disk_unblock_events - unblock disk event checking | |
1559 | * @disk: disk to unblock events for | |
1560 | * | |
1561 | * Undo disk_block_events(). When the block count reaches zero, it | |
1562 | * starts events polling if configured. | |
1563 | * | |
1564 | * CONTEXT: | |
1565 | * Don't care. Safe to call from irq context. | |
1566 | */ | |
1567 | void disk_unblock_events(struct gendisk *disk) | |
1568 | { | |
1569 | if (disk->ev) | |
facc31dd | 1570 | __disk_unblock_events(disk, false); |
77ea887e TH |
1571 | } |
1572 | ||
1573 | /** | |
85ef06d1 TH |
1574 | * disk_flush_events - schedule immediate event checking and flushing |
1575 | * @disk: disk to check and flush events for | |
1576 | * @mask: events to flush | |
77ea887e | 1577 | * |
85ef06d1 TH |
1578 | * Schedule immediate event checking on @disk if not blocked. Events in |
1579 | * @mask are scheduled to be cleared from the driver. Note that this | |
1580 | * doesn't clear the events from @disk->ev. | |
77ea887e TH |
1581 | * |
1582 | * CONTEXT: | |
85ef06d1 | 1583 | * If @mask is non-zero must be called with bdev->bd_mutex held. |
77ea887e | 1584 | */ |
85ef06d1 | 1585 | void disk_flush_events(struct gendisk *disk, unsigned int mask) |
77ea887e | 1586 | { |
a9dce2a3 | 1587 | struct disk_events *ev = disk->ev; |
a9dce2a3 TH |
1588 | |
1589 | if (!ev) | |
1590 | return; | |
1591 | ||
85ef06d1 TH |
1592 | spin_lock_irq(&ev->lock); |
1593 | ev->clearing |= mask; | |
a9dce2a3 TH |
1594 | if (!ev->block) { |
1595 | cancel_delayed_work(&ev->dwork); | |
1596 | queue_delayed_work(system_nrt_wq, &ev->dwork, 0); | |
77ea887e | 1597 | } |
85ef06d1 | 1598 | spin_unlock_irq(&ev->lock); |
77ea887e | 1599 | } |
77ea887e TH |
1600 | |
1601 | /** | |
1602 | * disk_clear_events - synchronously check, clear and return pending events | |
1603 | * @disk: disk to fetch and clear events from | |
1604 | * @mask: mask of events to be fetched and clearted | |
1605 | * | |
1606 | * Disk events are synchronously checked and pending events in @mask | |
1607 | * are cleared and returned. This ignores the block count. | |
1608 | * | |
1609 | * CONTEXT: | |
1610 | * Might sleep. | |
1611 | */ | |
1612 | unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask) | |
1613 | { | |
1614 | const struct block_device_operations *bdops = disk->fops; | |
1615 | struct disk_events *ev = disk->ev; | |
1616 | unsigned int pending; | |
1617 | ||
1618 | if (!ev) { | |
1619 | /* for drivers still using the old ->media_changed method */ | |
1620 | if ((mask & DISK_EVENT_MEDIA_CHANGE) && | |
1621 | bdops->media_changed && bdops->media_changed(disk)) | |
1622 | return DISK_EVENT_MEDIA_CHANGE; | |
1623 | return 0; | |
1624 | } | |
1625 | ||
1626 | /* tell the workfn about the events being cleared */ | |
1627 | spin_lock_irq(&ev->lock); | |
1628 | ev->clearing |= mask; | |
1629 | spin_unlock_irq(&ev->lock); | |
1630 | ||
1631 | /* uncondtionally schedule event check and wait for it to finish */ | |
c3af54af | 1632 | disk_block_events(disk); |
77ea887e TH |
1633 | queue_delayed_work(system_nrt_wq, &ev->dwork, 0); |
1634 | flush_delayed_work(&ev->dwork); | |
1635 | __disk_unblock_events(disk, false); | |
1636 | ||
1637 | /* then, fetch and clear pending events */ | |
1638 | spin_lock_irq(&ev->lock); | |
1639 | WARN_ON_ONCE(ev->clearing & mask); /* cleared by workfn */ | |
1640 | pending = ev->pending & mask; | |
1641 | ev->pending &= ~mask; | |
1642 | spin_unlock_irq(&ev->lock); | |
1643 | ||
1644 | return pending; | |
1645 | } | |
1646 | ||
1647 | static void disk_events_workfn(struct work_struct *work) | |
1648 | { | |
1649 | struct delayed_work *dwork = to_delayed_work(work); | |
1650 | struct disk_events *ev = container_of(dwork, struct disk_events, dwork); | |
1651 | struct gendisk *disk = ev->disk; | |
1652 | char *envp[ARRAY_SIZE(disk_uevents) + 1] = { }; | |
1653 | unsigned int clearing = ev->clearing; | |
1654 | unsigned int events; | |
1655 | unsigned long intv; | |
1656 | int nr_events = 0, i; | |
1657 | ||
1658 | /* check events */ | |
1659 | events = disk->fops->check_events(disk, clearing); | |
1660 | ||
1661 | /* accumulate pending events and schedule next poll if necessary */ | |
1662 | spin_lock_irq(&ev->lock); | |
1663 | ||
1664 | events &= ~ev->pending; | |
1665 | ev->pending |= events; | |
1666 | ev->clearing &= ~clearing; | |
1667 | ||
1668 | intv = disk_events_poll_jiffies(disk); | |
1669 | if (!ev->block && intv) | |
1670 | queue_delayed_work(system_nrt_wq, &ev->dwork, intv); | |
1671 | ||
1672 | spin_unlock_irq(&ev->lock); | |
1673 | ||
7c88a168 TH |
1674 | /* |
1675 | * Tell userland about new events. Only the events listed in | |
1676 | * @disk->events are reported. Unlisted events are processed the | |
1677 | * same internally but never get reported to userland. | |
1678 | */ | |
77ea887e | 1679 | for (i = 0; i < ARRAY_SIZE(disk_uevents); i++) |
7c88a168 | 1680 | if (events & disk->events & (1 << i)) |
77ea887e TH |
1681 | envp[nr_events++] = disk_uevents[i]; |
1682 | ||
1683 | if (nr_events) | |
1684 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
1685 | } | |
1686 | ||
1687 | /* | |
1688 | * A disk events enabled device has the following sysfs nodes under | |
1689 | * its /sys/block/X/ directory. | |
1690 | * | |
1691 | * events : list of all supported events | |
1692 | * events_async : list of events which can be detected w/o polling | |
1693 | * events_poll_msecs : polling interval, 0: disable, -1: system default | |
1694 | */ | |
1695 | static ssize_t __disk_events_show(unsigned int events, char *buf) | |
1696 | { | |
1697 | const char *delim = ""; | |
1698 | ssize_t pos = 0; | |
1699 | int i; | |
1700 | ||
1701 | for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++) | |
1702 | if (events & (1 << i)) { | |
1703 | pos += sprintf(buf + pos, "%s%s", | |
1704 | delim, disk_events_strs[i]); | |
1705 | delim = " "; | |
1706 | } | |
1707 | if (pos) | |
1708 | pos += sprintf(buf + pos, "\n"); | |
1709 | return pos; | |
1710 | } | |
1711 | ||
1712 | static ssize_t disk_events_show(struct device *dev, | |
1713 | struct device_attribute *attr, char *buf) | |
1714 | { | |
1715 | struct gendisk *disk = dev_to_disk(dev); | |
1716 | ||
1717 | return __disk_events_show(disk->events, buf); | |
1718 | } | |
1719 | ||
1720 | static ssize_t disk_events_async_show(struct device *dev, | |
1721 | struct device_attribute *attr, char *buf) | |
1722 | { | |
1723 | struct gendisk *disk = dev_to_disk(dev); | |
1724 | ||
1725 | return __disk_events_show(disk->async_events, buf); | |
1726 | } | |
1727 | ||
1728 | static ssize_t disk_events_poll_msecs_show(struct device *dev, | |
1729 | struct device_attribute *attr, | |
1730 | char *buf) | |
1731 | { | |
1732 | struct gendisk *disk = dev_to_disk(dev); | |
1733 | ||
1734 | return sprintf(buf, "%ld\n", disk->ev->poll_msecs); | |
1735 | } | |
1736 | ||
1737 | static ssize_t disk_events_poll_msecs_store(struct device *dev, | |
1738 | struct device_attribute *attr, | |
1739 | const char *buf, size_t count) | |
1740 | { | |
1741 | struct gendisk *disk = dev_to_disk(dev); | |
1742 | long intv; | |
1743 | ||
1744 | if (!count || !sscanf(buf, "%ld", &intv)) | |
1745 | return -EINVAL; | |
1746 | ||
1747 | if (intv < 0 && intv != -1) | |
1748 | return -EINVAL; | |
1749 | ||
c3af54af | 1750 | disk_block_events(disk); |
77ea887e TH |
1751 | disk->ev->poll_msecs = intv; |
1752 | __disk_unblock_events(disk, true); | |
1753 | ||
1754 | return count; | |
1755 | } | |
1756 | ||
1757 | static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL); | |
1758 | static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL); | |
1759 | static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR, | |
1760 | disk_events_poll_msecs_show, | |
1761 | disk_events_poll_msecs_store); | |
1762 | ||
1763 | static const struct attribute *disk_events_attrs[] = { | |
1764 | &dev_attr_events.attr, | |
1765 | &dev_attr_events_async.attr, | |
1766 | &dev_attr_events_poll_msecs.attr, | |
1767 | NULL, | |
1768 | }; | |
1769 | ||
1770 | /* | |
1771 | * The default polling interval can be specified by the kernel | |
1772 | * parameter block.events_dfl_poll_msecs which defaults to 0 | |
1773 | * (disable). This can also be modified runtime by writing to | |
1774 | * /sys/module/block/events_dfl_poll_msecs. | |
1775 | */ | |
1776 | static int disk_events_set_dfl_poll_msecs(const char *val, | |
1777 | const struct kernel_param *kp) | |
1778 | { | |
1779 | struct disk_events *ev; | |
1780 | int ret; | |
1781 | ||
1782 | ret = param_set_ulong(val, kp); | |
1783 | if (ret < 0) | |
1784 | return ret; | |
1785 | ||
1786 | mutex_lock(&disk_events_mutex); | |
1787 | ||
1788 | list_for_each_entry(ev, &disk_events, node) | |
85ef06d1 | 1789 | disk_flush_events(ev->disk, 0); |
77ea887e TH |
1790 | |
1791 | mutex_unlock(&disk_events_mutex); | |
1792 | ||
1793 | return 0; | |
1794 | } | |
1795 | ||
1796 | static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = { | |
1797 | .set = disk_events_set_dfl_poll_msecs, | |
1798 | .get = param_get_ulong, | |
1799 | }; | |
1800 | ||
1801 | #undef MODULE_PARAM_PREFIX | |
1802 | #define MODULE_PARAM_PREFIX "block." | |
1803 | ||
1804 | module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops, | |
1805 | &disk_events_dfl_poll_msecs, 0644); | |
1806 | ||
1807 | /* | |
1808 | * disk_{add|del|release}_events - initialize and destroy disk_events. | |
1809 | */ | |
1810 | static void disk_add_events(struct gendisk *disk) | |
1811 | { | |
1812 | struct disk_events *ev; | |
1813 | ||
75e3f3ee | 1814 | if (!disk->fops->check_events) |
77ea887e TH |
1815 | return; |
1816 | ||
1817 | ev = kzalloc(sizeof(*ev), GFP_KERNEL); | |
1818 | if (!ev) { | |
1819 | pr_warn("%s: failed to initialize events\n", disk->disk_name); | |
1820 | return; | |
1821 | } | |
1822 | ||
1823 | if (sysfs_create_files(&disk_to_dev(disk)->kobj, | |
1824 | disk_events_attrs) < 0) { | |
1825 | pr_warn("%s: failed to create sysfs files for events\n", | |
1826 | disk->disk_name); | |
1827 | kfree(ev); | |
1828 | return; | |
1829 | } | |
1830 | ||
1831 | disk->ev = ev; | |
1832 | ||
1833 | INIT_LIST_HEAD(&ev->node); | |
1834 | ev->disk = disk; | |
1835 | spin_lock_init(&ev->lock); | |
fdd514e1 | 1836 | mutex_init(&ev->block_mutex); |
77ea887e TH |
1837 | ev->block = 1; |
1838 | ev->poll_msecs = -1; | |
1839 | INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn); | |
1840 | ||
1841 | mutex_lock(&disk_events_mutex); | |
1842 | list_add_tail(&ev->node, &disk_events); | |
1843 | mutex_unlock(&disk_events_mutex); | |
1844 | ||
1845 | /* | |
1846 | * Block count is initialized to 1 and the following initial | |
1847 | * unblock kicks it into action. | |
1848 | */ | |
1849 | __disk_unblock_events(disk, true); | |
1850 | } | |
1851 | ||
1852 | static void disk_del_events(struct gendisk *disk) | |
1853 | { | |
1854 | if (!disk->ev) | |
1855 | return; | |
1856 | ||
c3af54af | 1857 | disk_block_events(disk); |
77ea887e TH |
1858 | |
1859 | mutex_lock(&disk_events_mutex); | |
1860 | list_del_init(&disk->ev->node); | |
1861 | mutex_unlock(&disk_events_mutex); | |
1862 | ||
1863 | sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs); | |
1864 | } | |
1865 | ||
1866 | static void disk_release_events(struct gendisk *disk) | |
1867 | { | |
1868 | /* the block count should be 1 from disk_del_events() */ | |
1869 | WARN_ON_ONCE(disk->ev && disk->ev->block != 1); | |
1870 | kfree(disk->ev); | |
1871 | } |