]>
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> | |
66114cad | 11 | #include <linux/backing-dev.h> |
1da177e4 LT |
12 | #include <linux/init.h> |
13 | #include <linux/spinlock.h> | |
f500975a | 14 | #include <linux/proc_fs.h> |
1da177e4 LT |
15 | #include <linux/seq_file.h> |
16 | #include <linux/slab.h> | |
17 | #include <linux/kmod.h> | |
18 | #include <linux/kobj_map.h> | |
58383af6 | 19 | #include <linux/mutex.h> |
bcce3de1 | 20 | #include <linux/idr.h> |
77ea887e | 21 | #include <linux/log2.h> |
25e823c8 | 22 | #include <linux/pm_runtime.h> |
99e6608c | 23 | #include <linux/badblocks.h> |
1da177e4 | 24 | |
ff88972c AB |
25 | #include "blk.h" |
26 | ||
edfaa7c3 | 27 | static DEFINE_MUTEX(block_class_lock); |
edfaa7c3 | 28 | struct kobject *block_depr; |
1da177e4 | 29 | |
bcce3de1 | 30 | /* for extended dynamic devt allocation, currently only one major is used */ |
ce23bba8 | 31 | #define NR_EXT_DEVT (1 << MINORBITS) |
bcce3de1 | 32 | |
2da78092 | 33 | /* For extended devt allocation. ext_devt_lock prevents look up |
bcce3de1 TH |
34 | * results from going away underneath its user. |
35 | */ | |
2da78092 | 36 | static DEFINE_SPINLOCK(ext_devt_lock); |
bcce3de1 TH |
37 | static DEFINE_IDR(ext_devt_idr); |
38 | ||
edf8ff55 | 39 | static const struct device_type disk_type; |
1826eadf | 40 | |
12c2bdb2 DB |
41 | static void disk_check_events(struct disk_events *ev, |
42 | unsigned int *clearing_ptr); | |
9f53d2fe | 43 | static void disk_alloc_events(struct gendisk *disk); |
77ea887e TH |
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 | ||
f299b7c7 JA |
48 | void part_inc_in_flight(struct request_queue *q, struct hd_struct *part, int rw) |
49 | { | |
50 | if (q->mq_ops) | |
51 | return; | |
52 | ||
53 | atomic_inc(&part->in_flight[rw]); | |
54 | if (part->partno) | |
55 | atomic_inc(&part_to_disk(part)->part0.in_flight[rw]); | |
56 | } | |
57 | ||
58 | void part_dec_in_flight(struct request_queue *q, struct hd_struct *part, int rw) | |
59 | { | |
60 | if (q->mq_ops) | |
61 | return; | |
62 | ||
63 | atomic_dec(&part->in_flight[rw]); | |
64 | if (part->partno) | |
65 | atomic_dec(&part_to_disk(part)->part0.in_flight[rw]); | |
66 | } | |
67 | ||
68 | void part_in_flight(struct request_queue *q, struct hd_struct *part, | |
69 | unsigned int inflight[2]) | |
70 | { | |
71 | if (q->mq_ops) { | |
72 | blk_mq_in_flight(q, part, inflight); | |
73 | return; | |
74 | } | |
75 | ||
76 | inflight[0] = atomic_read(&part->in_flight[0]) + | |
77 | atomic_read(&part->in_flight[1]); | |
78 | if (part->partno) { | |
79 | part = &part_to_disk(part)->part0; | |
80 | inflight[1] = atomic_read(&part->in_flight[0]) + | |
81 | atomic_read(&part->in_flight[1]); | |
82 | } | |
83 | } | |
84 | ||
416d3aa7 OS |
85 | void part_in_flight_rw(struct request_queue *q, struct hd_struct *part, |
86 | unsigned int inflight[2]) | |
87 | { | |
88 | if (q->mq_ops) { | |
89 | blk_mq_in_flight_rw(q, part, inflight); | |
90 | return; | |
91 | } | |
92 | ||
93 | inflight[0] = atomic_read(&part->in_flight[0]); | |
94 | inflight[1] = atomic_read(&part->in_flight[1]); | |
95 | } | |
96 | ||
807d4af2 CH |
97 | struct hd_struct *__disk_get_part(struct gendisk *disk, int partno) |
98 | { | |
99 | struct disk_part_tbl *ptbl = rcu_dereference(disk->part_tbl); | |
100 | ||
101 | if (unlikely(partno < 0 || partno >= ptbl->len)) | |
102 | return NULL; | |
103 | return rcu_dereference(ptbl->part[partno]); | |
104 | } | |
105 | ||
e71bf0d0 TH |
106 | /** |
107 | * disk_get_part - get partition | |
108 | * @disk: disk to look partition from | |
109 | * @partno: partition number | |
110 | * | |
111 | * Look for partition @partno from @disk. If found, increment | |
112 | * reference count and return it. | |
113 | * | |
114 | * CONTEXT: | |
115 | * Don't care. | |
116 | * | |
117 | * RETURNS: | |
118 | * Pointer to the found partition on success, NULL if not found. | |
119 | */ | |
120 | struct hd_struct *disk_get_part(struct gendisk *disk, int partno) | |
121 | { | |
807d4af2 | 122 | struct hd_struct *part; |
540eed56 | 123 | |
e71bf0d0 | 124 | rcu_read_lock(); |
807d4af2 CH |
125 | part = __disk_get_part(disk, partno); |
126 | if (part) | |
127 | get_device(part_to_dev(part)); | |
e71bf0d0 TH |
128 | rcu_read_unlock(); |
129 | ||
130 | return part; | |
131 | } | |
132 | EXPORT_SYMBOL_GPL(disk_get_part); | |
133 | ||
134 | /** | |
135 | * disk_part_iter_init - initialize partition iterator | |
136 | * @piter: iterator to initialize | |
137 | * @disk: disk to iterate over | |
138 | * @flags: DISK_PITER_* flags | |
139 | * | |
140 | * Initialize @piter so that it iterates over partitions of @disk. | |
141 | * | |
142 | * CONTEXT: | |
143 | * Don't care. | |
144 | */ | |
145 | void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk, | |
146 | unsigned int flags) | |
147 | { | |
540eed56 TH |
148 | struct disk_part_tbl *ptbl; |
149 | ||
150 | rcu_read_lock(); | |
151 | ptbl = rcu_dereference(disk->part_tbl); | |
152 | ||
e71bf0d0 TH |
153 | piter->disk = disk; |
154 | piter->part = NULL; | |
155 | ||
156 | if (flags & DISK_PITER_REVERSE) | |
540eed56 | 157 | piter->idx = ptbl->len - 1; |
71982a40 | 158 | else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0)) |
e71bf0d0 | 159 | piter->idx = 0; |
b5d0b9df TH |
160 | else |
161 | piter->idx = 1; | |
e71bf0d0 TH |
162 | |
163 | piter->flags = flags; | |
540eed56 TH |
164 | |
165 | rcu_read_unlock(); | |
e71bf0d0 TH |
166 | } |
167 | EXPORT_SYMBOL_GPL(disk_part_iter_init); | |
168 | ||
169 | /** | |
170 | * disk_part_iter_next - proceed iterator to the next partition and return it | |
171 | * @piter: iterator of interest | |
172 | * | |
173 | * Proceed @piter to the next partition and return it. | |
174 | * | |
175 | * CONTEXT: | |
176 | * Don't care. | |
177 | */ | |
178 | struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter) | |
179 | { | |
540eed56 | 180 | struct disk_part_tbl *ptbl; |
e71bf0d0 TH |
181 | int inc, end; |
182 | ||
183 | /* put the last partition */ | |
184 | disk_put_part(piter->part); | |
185 | piter->part = NULL; | |
186 | ||
540eed56 | 187 | /* get part_tbl */ |
e71bf0d0 | 188 | rcu_read_lock(); |
540eed56 | 189 | ptbl = rcu_dereference(piter->disk->part_tbl); |
e71bf0d0 TH |
190 | |
191 | /* determine iteration parameters */ | |
192 | if (piter->flags & DISK_PITER_REVERSE) { | |
193 | inc = -1; | |
71982a40 TH |
194 | if (piter->flags & (DISK_PITER_INCL_PART0 | |
195 | DISK_PITER_INCL_EMPTY_PART0)) | |
b5d0b9df TH |
196 | end = -1; |
197 | else | |
198 | end = 0; | |
e71bf0d0 TH |
199 | } else { |
200 | inc = 1; | |
540eed56 | 201 | end = ptbl->len; |
e71bf0d0 TH |
202 | } |
203 | ||
204 | /* iterate to the next partition */ | |
205 | for (; piter->idx != end; piter->idx += inc) { | |
206 | struct hd_struct *part; | |
207 | ||
540eed56 | 208 | part = rcu_dereference(ptbl->part[piter->idx]); |
e71bf0d0 TH |
209 | if (!part) |
210 | continue; | |
c83f6bf9 | 211 | if (!part_nr_sects_read(part) && |
71982a40 TH |
212 | !(piter->flags & DISK_PITER_INCL_EMPTY) && |
213 | !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 && | |
214 | piter->idx == 0)) | |
e71bf0d0 TH |
215 | continue; |
216 | ||
ed9e1982 | 217 | get_device(part_to_dev(part)); |
e71bf0d0 TH |
218 | piter->part = part; |
219 | piter->idx += inc; | |
220 | break; | |
221 | } | |
222 | ||
223 | rcu_read_unlock(); | |
224 | ||
225 | return piter->part; | |
226 | } | |
227 | EXPORT_SYMBOL_GPL(disk_part_iter_next); | |
228 | ||
229 | /** | |
230 | * disk_part_iter_exit - finish up partition iteration | |
231 | * @piter: iter of interest | |
232 | * | |
233 | * Called when iteration is over. Cleans up @piter. | |
234 | * | |
235 | * CONTEXT: | |
236 | * Don't care. | |
237 | */ | |
238 | void disk_part_iter_exit(struct disk_part_iter *piter) | |
239 | { | |
240 | disk_put_part(piter->part); | |
241 | piter->part = NULL; | |
242 | } | |
243 | EXPORT_SYMBOL_GPL(disk_part_iter_exit); | |
244 | ||
a6f23657 JA |
245 | static inline int sector_in_part(struct hd_struct *part, sector_t sector) |
246 | { | |
247 | return part->start_sect <= sector && | |
c83f6bf9 | 248 | sector < part->start_sect + part_nr_sects_read(part); |
a6f23657 JA |
249 | } |
250 | ||
e71bf0d0 TH |
251 | /** |
252 | * disk_map_sector_rcu - map sector to partition | |
253 | * @disk: gendisk of interest | |
254 | * @sector: sector to map | |
255 | * | |
256 | * Find out which partition @sector maps to on @disk. This is | |
257 | * primarily used for stats accounting. | |
258 | * | |
259 | * CONTEXT: | |
260 | * RCU read locked. The returned partition pointer is valid only | |
261 | * while preemption is disabled. | |
262 | * | |
263 | * RETURNS: | |
074a7aca | 264 | * Found partition on success, part0 is returned if no partition matches |
e71bf0d0 TH |
265 | */ |
266 | struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector) | |
267 | { | |
540eed56 | 268 | struct disk_part_tbl *ptbl; |
a6f23657 | 269 | struct hd_struct *part; |
e71bf0d0 TH |
270 | int i; |
271 | ||
540eed56 TH |
272 | ptbl = rcu_dereference(disk->part_tbl); |
273 | ||
a6f23657 JA |
274 | part = rcu_dereference(ptbl->last_lookup); |
275 | if (part && sector_in_part(part, sector)) | |
276 | return part; | |
277 | ||
540eed56 | 278 | for (i = 1; i < ptbl->len; i++) { |
a6f23657 | 279 | part = rcu_dereference(ptbl->part[i]); |
e71bf0d0 | 280 | |
a6f23657 JA |
281 | if (part && sector_in_part(part, sector)) { |
282 | rcu_assign_pointer(ptbl->last_lookup, part); | |
e71bf0d0 | 283 | return part; |
a6f23657 | 284 | } |
e71bf0d0 | 285 | } |
074a7aca | 286 | return &disk->part0; |
e71bf0d0 TH |
287 | } |
288 | EXPORT_SYMBOL_GPL(disk_map_sector_rcu); | |
289 | ||
1da177e4 LT |
290 | /* |
291 | * Can be deleted altogether. Later. | |
292 | * | |
293 | */ | |
133d55cd | 294 | #define BLKDEV_MAJOR_HASH_SIZE 255 |
1da177e4 LT |
295 | static struct blk_major_name { |
296 | struct blk_major_name *next; | |
297 | int major; | |
298 | char name[16]; | |
68eef3b4 | 299 | } *major_names[BLKDEV_MAJOR_HASH_SIZE]; |
1da177e4 LT |
300 | |
301 | /* index in the above - for now: assume no multimajor ranges */ | |
e61eb2e9 | 302 | static inline int major_to_index(unsigned major) |
1da177e4 | 303 | { |
68eef3b4 | 304 | return major % BLKDEV_MAJOR_HASH_SIZE; |
7170be5f NH |
305 | } |
306 | ||
68eef3b4 | 307 | #ifdef CONFIG_PROC_FS |
cf771cb5 | 308 | void blkdev_show(struct seq_file *seqf, off_t offset) |
7170be5f | 309 | { |
68eef3b4 | 310 | struct blk_major_name *dp; |
7170be5f | 311 | |
133d55cd LG |
312 | mutex_lock(&block_class_lock); |
313 | for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next) | |
314 | if (dp->major == offset) | |
cf771cb5 | 315 | seq_printf(seqf, "%3d %s\n", dp->major, dp->name); |
133d55cd | 316 | mutex_unlock(&block_class_lock); |
1da177e4 | 317 | } |
68eef3b4 | 318 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 319 | |
9e8c0bcc MN |
320 | /** |
321 | * register_blkdev - register a new block device | |
322 | * | |
0e056eb5 | 323 | * @major: the requested major device number [1..255]. If @major = 0, try to |
9e8c0bcc MN |
324 | * allocate any unused major number. |
325 | * @name: the name of the new block device as a zero terminated string | |
326 | * | |
327 | * The @name must be unique within the system. | |
328 | * | |
0e056eb5 MCC |
329 | * The return value depends on the @major input parameter: |
330 | * | |
9e8c0bcc MN |
331 | * - if a major device number was requested in range [1..255] then the |
332 | * function returns zero on success, or a negative error code | |
0e056eb5 | 333 | * - if any unused major number was requested with @major = 0 parameter |
9e8c0bcc MN |
334 | * then the return value is the allocated major number in range |
335 | * [1..255] or a negative error code otherwise | |
336 | */ | |
1da177e4 LT |
337 | int register_blkdev(unsigned int major, const char *name) |
338 | { | |
339 | struct blk_major_name **n, *p; | |
340 | int index, ret = 0; | |
341 | ||
edfaa7c3 | 342 | mutex_lock(&block_class_lock); |
1da177e4 LT |
343 | |
344 | /* temporary */ | |
345 | if (major == 0) { | |
346 | for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) { | |
347 | if (major_names[index] == NULL) | |
348 | break; | |
349 | } | |
350 | ||
351 | if (index == 0) { | |
352 | printk("register_blkdev: failed to get major for %s\n", | |
353 | name); | |
354 | ret = -EBUSY; | |
355 | goto out; | |
356 | } | |
357 | major = index; | |
358 | ret = major; | |
359 | } | |
360 | ||
133d55cd LG |
361 | if (major >= BLKDEV_MAJOR_MAX) { |
362 | pr_err("register_blkdev: major requested (%d) is greater than the maximum (%d) for %s\n", | |
363 | major, BLKDEV_MAJOR_MAX, name); | |
364 | ||
365 | ret = -EINVAL; | |
366 | goto out; | |
367 | } | |
368 | ||
1da177e4 LT |
369 | p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL); |
370 | if (p == NULL) { | |
371 | ret = -ENOMEM; | |
372 | goto out; | |
373 | } | |
374 | ||
375 | p->major = major; | |
376 | strlcpy(p->name, name, sizeof(p->name)); | |
377 | p->next = NULL; | |
378 | index = major_to_index(major); | |
379 | ||
380 | for (n = &major_names[index]; *n; n = &(*n)->next) { | |
381 | if ((*n)->major == major) | |
382 | break; | |
383 | } | |
384 | if (!*n) | |
385 | *n = p; | |
386 | else | |
387 | ret = -EBUSY; | |
388 | ||
389 | if (ret < 0) { | |
390 | printk("register_blkdev: cannot get major %d for %s\n", | |
391 | major, name); | |
392 | kfree(p); | |
393 | } | |
394 | out: | |
edfaa7c3 | 395 | mutex_unlock(&block_class_lock); |
1da177e4 LT |
396 | return ret; |
397 | } | |
398 | ||
399 | EXPORT_SYMBOL(register_blkdev); | |
400 | ||
f4480240 | 401 | void unregister_blkdev(unsigned int major, const char *name) |
1da177e4 LT |
402 | { |
403 | struct blk_major_name **n; | |
404 | struct blk_major_name *p = NULL; | |
405 | int index = major_to_index(major); | |
1da177e4 | 406 | |
edfaa7c3 | 407 | mutex_lock(&block_class_lock); |
1da177e4 LT |
408 | for (n = &major_names[index]; *n; n = &(*n)->next) |
409 | if ((*n)->major == major) | |
410 | break; | |
294462a5 AM |
411 | if (!*n || strcmp((*n)->name, name)) { |
412 | WARN_ON(1); | |
294462a5 | 413 | } else { |
1da177e4 LT |
414 | p = *n; |
415 | *n = p->next; | |
416 | } | |
edfaa7c3 | 417 | mutex_unlock(&block_class_lock); |
1da177e4 | 418 | kfree(p); |
1da177e4 LT |
419 | } |
420 | ||
421 | EXPORT_SYMBOL(unregister_blkdev); | |
422 | ||
423 | static struct kobj_map *bdev_map; | |
424 | ||
870d6656 TH |
425 | /** |
426 | * blk_mangle_minor - scatter minor numbers apart | |
427 | * @minor: minor number to mangle | |
428 | * | |
429 | * Scatter consecutively allocated @minor number apart if MANGLE_DEVT | |
430 | * is enabled. Mangling twice gives the original value. | |
431 | * | |
432 | * RETURNS: | |
433 | * Mangled value. | |
434 | * | |
435 | * CONTEXT: | |
436 | * Don't care. | |
437 | */ | |
438 | static int blk_mangle_minor(int minor) | |
439 | { | |
440 | #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT | |
441 | int i; | |
442 | ||
443 | for (i = 0; i < MINORBITS / 2; i++) { | |
444 | int low = minor & (1 << i); | |
445 | int high = minor & (1 << (MINORBITS - 1 - i)); | |
446 | int distance = MINORBITS - 1 - 2 * i; | |
447 | ||
448 | minor ^= low | high; /* clear both bits */ | |
449 | low <<= distance; /* swap the positions */ | |
450 | high >>= distance; | |
451 | minor |= low | high; /* and set */ | |
452 | } | |
453 | #endif | |
454 | return minor; | |
455 | } | |
456 | ||
bcce3de1 TH |
457 | /** |
458 | * blk_alloc_devt - allocate a dev_t for a partition | |
459 | * @part: partition to allocate dev_t for | |
bcce3de1 TH |
460 | * @devt: out parameter for resulting dev_t |
461 | * | |
462 | * Allocate a dev_t for block device. | |
463 | * | |
464 | * RETURNS: | |
465 | * 0 on success, allocated dev_t is returned in *@devt. -errno on | |
466 | * failure. | |
467 | * | |
468 | * CONTEXT: | |
469 | * Might sleep. | |
470 | */ | |
471 | int blk_alloc_devt(struct hd_struct *part, dev_t *devt) | |
472 | { | |
473 | struct gendisk *disk = part_to_disk(part); | |
bab998d6 | 474 | int idx; |
bcce3de1 TH |
475 | |
476 | /* in consecutive minor range? */ | |
477 | if (part->partno < disk->minors) { | |
478 | *devt = MKDEV(disk->major, disk->first_minor + part->partno); | |
479 | return 0; | |
480 | } | |
481 | ||
482 | /* allocate ext devt */ | |
2da78092 KB |
483 | idr_preload(GFP_KERNEL); |
484 | ||
4d66e5e9 | 485 | spin_lock_bh(&ext_devt_lock); |
2da78092 | 486 | idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_NOWAIT); |
4d66e5e9 | 487 | spin_unlock_bh(&ext_devt_lock); |
2da78092 KB |
488 | |
489 | idr_preload_end(); | |
bab998d6 TH |
490 | if (idx < 0) |
491 | return idx == -ENOSPC ? -EBUSY : idx; | |
bcce3de1 | 492 | |
870d6656 | 493 | *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx)); |
bcce3de1 TH |
494 | return 0; |
495 | } | |
496 | ||
497 | /** | |
498 | * blk_free_devt - free a dev_t | |
499 | * @devt: dev_t to free | |
500 | * | |
501 | * Free @devt which was allocated using blk_alloc_devt(). | |
502 | * | |
503 | * CONTEXT: | |
504 | * Might sleep. | |
505 | */ | |
506 | void blk_free_devt(dev_t devt) | |
507 | { | |
bcce3de1 TH |
508 | if (devt == MKDEV(0, 0)) |
509 | return; | |
510 | ||
511 | if (MAJOR(devt) == BLOCK_EXT_MAJOR) { | |
4d66e5e9 | 512 | spin_lock_bh(&ext_devt_lock); |
870d6656 | 513 | idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); |
4d66e5e9 | 514 | spin_unlock_bh(&ext_devt_lock); |
bcce3de1 TH |
515 | } |
516 | } | |
517 | ||
fde48572 YY |
518 | /** |
519 | * We invalidate devt by assigning NULL pointer for devt in idr. | |
520 | */ | |
521 | void blk_invalidate_devt(dev_t devt) | |
522 | { | |
523 | if (MAJOR(devt) == BLOCK_EXT_MAJOR) { | |
524 | spin_lock_bh(&ext_devt_lock); | |
525 | idr_replace(&ext_devt_idr, NULL, blk_mangle_minor(MINOR(devt))); | |
526 | spin_unlock_bh(&ext_devt_lock); | |
527 | } | |
528 | } | |
529 | ||
1f014290 TH |
530 | static char *bdevt_str(dev_t devt, char *buf) |
531 | { | |
532 | if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) { | |
533 | char tbuf[BDEVT_SIZE]; | |
534 | snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt)); | |
535 | snprintf(buf, BDEVT_SIZE, "%-9s", tbuf); | |
536 | } else | |
537 | snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt)); | |
538 | ||
539 | return buf; | |
540 | } | |
541 | ||
1da177e4 LT |
542 | /* |
543 | * Register device numbers dev..(dev+range-1) | |
544 | * range must be nonzero | |
545 | * The hash chain is sorted on range, so that subranges can override. | |
546 | */ | |
edfaa7c3 | 547 | void blk_register_region(dev_t devt, unsigned long range, struct module *module, |
1da177e4 LT |
548 | struct kobject *(*probe)(dev_t, int *, void *), |
549 | int (*lock)(dev_t, void *), void *data) | |
550 | { | |
edfaa7c3 | 551 | kobj_map(bdev_map, devt, range, module, probe, lock, data); |
1da177e4 LT |
552 | } |
553 | ||
554 | EXPORT_SYMBOL(blk_register_region); | |
555 | ||
edfaa7c3 | 556 | void blk_unregister_region(dev_t devt, unsigned long range) |
1da177e4 | 557 | { |
edfaa7c3 | 558 | kobj_unmap(bdev_map, devt, range); |
1da177e4 LT |
559 | } |
560 | ||
561 | EXPORT_SYMBOL(blk_unregister_region); | |
562 | ||
cf771cb5 | 563 | static struct kobject *exact_match(dev_t devt, int *partno, void *data) |
1da177e4 LT |
564 | { |
565 | struct gendisk *p = data; | |
edfaa7c3 | 566 | |
ed9e1982 | 567 | return &disk_to_dev(p)->kobj; |
1da177e4 LT |
568 | } |
569 | ||
edfaa7c3 | 570 | static int exact_lock(dev_t devt, void *data) |
1da177e4 LT |
571 | { |
572 | struct gendisk *p = data; | |
573 | ||
574 | if (!get_disk(p)) | |
575 | return -1; | |
576 | return 0; | |
577 | } | |
578 | ||
e63a46be | 579 | static void register_disk(struct device *parent, struct gendisk *disk) |
d2bf1b67 TH |
580 | { |
581 | struct device *ddev = disk_to_dev(disk); | |
582 | struct block_device *bdev; | |
583 | struct disk_part_iter piter; | |
584 | struct hd_struct *part; | |
585 | int err; | |
586 | ||
e63a46be | 587 | ddev->parent = parent; |
d2bf1b67 | 588 | |
ffc8b308 | 589 | dev_set_name(ddev, "%s", disk->disk_name); |
d2bf1b67 TH |
590 | |
591 | /* delay uevents, until we scanned partition table */ | |
592 | dev_set_uevent_suppress(ddev, 1); | |
593 | ||
594 | if (device_add(ddev)) | |
595 | return; | |
596 | if (!sysfs_deprecated) { | |
597 | err = sysfs_create_link(block_depr, &ddev->kobj, | |
598 | kobject_name(&ddev->kobj)); | |
599 | if (err) { | |
600 | device_del(ddev); | |
601 | return; | |
602 | } | |
603 | } | |
25e823c8 ML |
604 | |
605 | /* | |
606 | * avoid probable deadlock caused by allocating memory with | |
607 | * GFP_KERNEL in runtime_resume callback of its all ancestor | |
608 | * devices | |
609 | */ | |
610 | pm_runtime_set_memalloc_noio(ddev, true); | |
611 | ||
d2bf1b67 TH |
612 | disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj); |
613 | disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj); | |
614 | ||
8ddcd653 CH |
615 | if (disk->flags & GENHD_FL_HIDDEN) { |
616 | dev_set_uevent_suppress(ddev, 0); | |
617 | return; | |
618 | } | |
619 | ||
d2bf1b67 | 620 | /* No minors to use for partitions */ |
d27769ec | 621 | if (!disk_part_scan_enabled(disk)) |
d2bf1b67 TH |
622 | goto exit; |
623 | ||
624 | /* No such device (e.g., media were just removed) */ | |
625 | if (!get_capacity(disk)) | |
626 | goto exit; | |
627 | ||
628 | bdev = bdget_disk(disk, 0); | |
629 | if (!bdev) | |
630 | goto exit; | |
631 | ||
632 | bdev->bd_invalidated = 1; | |
633 | err = blkdev_get(bdev, FMODE_READ, NULL); | |
634 | if (err < 0) | |
635 | goto exit; | |
636 | blkdev_put(bdev, FMODE_READ); | |
637 | ||
638 | exit: | |
639 | /* announce disk after possible partitions are created */ | |
640 | dev_set_uevent_suppress(ddev, 0); | |
641 | kobject_uevent(&ddev->kobj, KOBJ_ADD); | |
642 | ||
643 | /* announce possible partitions */ | |
644 | disk_part_iter_init(&piter, disk, 0); | |
645 | while ((part = disk_part_iter_next(&piter))) | |
646 | kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD); | |
647 | disk_part_iter_exit(&piter); | |
8ddcd653 CH |
648 | |
649 | err = sysfs_create_link(&ddev->kobj, | |
650 | &disk->queue->backing_dev_info->dev->kobj, | |
651 | "bdi"); | |
652 | WARN_ON(err); | |
d2bf1b67 TH |
653 | } |
654 | ||
1da177e4 | 655 | /** |
e63a46be DW |
656 | * device_add_disk - add partitioning information to kernel list |
657 | * @parent: parent device for the disk | |
1da177e4 LT |
658 | * @disk: per-device partitioning information |
659 | * | |
660 | * This function registers the partitioning information in @disk | |
661 | * with the kernel. | |
3e1a7ff8 TH |
662 | * |
663 | * FIXME: error handling | |
1da177e4 | 664 | */ |
e63a46be | 665 | void device_add_disk(struct device *parent, struct gendisk *disk) |
1da177e4 | 666 | { |
3e1a7ff8 | 667 | dev_t devt; |
6ffeea77 | 668 | int retval; |
cf0ca9fe | 669 | |
3e1a7ff8 TH |
670 | /* minors == 0 indicates to use ext devt from part0 and should |
671 | * be accompanied with EXT_DEVT flag. Make sure all | |
672 | * parameters make sense. | |
673 | */ | |
674 | WARN_ON(disk->minors && !(disk->major || disk->first_minor)); | |
8ddcd653 CH |
675 | WARN_ON(!disk->minors && |
676 | !(disk->flags & (GENHD_FL_EXT_DEVT | GENHD_FL_HIDDEN))); | |
3e1a7ff8 | 677 | |
1da177e4 | 678 | disk->flags |= GENHD_FL_UP; |
3e1a7ff8 TH |
679 | |
680 | retval = blk_alloc_devt(&disk->part0, &devt); | |
681 | if (retval) { | |
682 | WARN_ON(1); | |
683 | return; | |
684 | } | |
3e1a7ff8 TH |
685 | disk->major = MAJOR(devt); |
686 | disk->first_minor = MINOR(devt); | |
687 | ||
9f53d2fe SG |
688 | disk_alloc_events(disk); |
689 | ||
8ddcd653 CH |
690 | if (disk->flags & GENHD_FL_HIDDEN) { |
691 | /* | |
692 | * Don't let hidden disks show up in /proc/partitions, | |
693 | * and don't bother scanning for partitions either. | |
694 | */ | |
695 | disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO; | |
696 | disk->flags |= GENHD_FL_NO_PART_SCAN; | |
697 | } else { | |
3a92168b | 698 | int ret; |
699 | ||
8ddcd653 CH |
700 | /* Register BDI before referencing it from bdev */ |
701 | disk_to_dev(disk)->devt = devt; | |
3a92168b | 702 | ret = bdi_register_owner(disk->queue->backing_dev_info, |
703 | disk_to_dev(disk)); | |
704 | WARN_ON(ret); | |
8ddcd653 CH |
705 | blk_register_region(disk_devt(disk), disk->minors, NULL, |
706 | exact_match, exact_lock, disk); | |
707 | } | |
e63a46be | 708 | register_disk(parent, disk); |
1da177e4 | 709 | blk_register_queue(disk); |
cf0ca9fe | 710 | |
523e1d39 TH |
711 | /* |
712 | * Take an extra ref on queue which will be put on disk_release() | |
713 | * so that it sticks around as long as @disk is there. | |
714 | */ | |
09ac46c4 | 715 | WARN_ON_ONCE(!blk_get_queue(disk->queue)); |
523e1d39 | 716 | |
77ea887e | 717 | disk_add_events(disk); |
25520d55 | 718 | blk_integrity_add(disk); |
1da177e4 | 719 | } |
e63a46be | 720 | EXPORT_SYMBOL(device_add_disk); |
1da177e4 | 721 | |
d2bf1b67 | 722 | void del_gendisk(struct gendisk *disk) |
1da177e4 | 723 | { |
d2bf1b67 TH |
724 | struct disk_part_iter piter; |
725 | struct hd_struct *part; | |
726 | ||
25520d55 | 727 | blk_integrity_del(disk); |
77ea887e TH |
728 | disk_del_events(disk); |
729 | ||
d2bf1b67 TH |
730 | /* invalidate stuff */ |
731 | disk_part_iter_init(&piter, disk, | |
732 | DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE); | |
733 | while ((part = disk_part_iter_next(&piter))) { | |
734 | invalidate_partition(disk, part->partno); | |
4b8c861a | 735 | bdev_unhash_inode(part_devt(part)); |
d2bf1b67 TH |
736 | delete_partition(disk, part->partno); |
737 | } | |
738 | disk_part_iter_exit(&piter); | |
739 | ||
740 | invalidate_partition(disk, 0); | |
d06e05c0 | 741 | bdev_unhash_inode(disk_devt(disk)); |
d2bf1b67 TH |
742 | set_capacity(disk, 0); |
743 | disk->flags &= ~GENHD_FL_UP; | |
744 | ||
8ddcd653 CH |
745 | if (!(disk->flags & GENHD_FL_HIDDEN)) |
746 | sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi"); | |
90f16fdd JK |
747 | if (disk->queue) { |
748 | /* | |
749 | * Unregister bdi before releasing device numbers (as they can | |
750 | * get reused and we'd get clashes in sysfs). | |
751 | */ | |
752 | bdi_unregister(disk->queue->backing_dev_info); | |
753 | blk_unregister_queue(disk); | |
754 | } else { | |
755 | WARN_ON(1); | |
756 | } | |
d2bf1b67 | 757 | |
17eac099 | 758 | if (!(disk->flags & GENHD_FL_HIDDEN)) |
8ddcd653 | 759 | blk_unregister_region(disk_devt(disk), disk->minors); |
fde48572 YY |
760 | /* |
761 | * Remove gendisk pointer from idr so that it cannot be looked up | |
762 | * while RCU period before freeing gendisk is running to prevent | |
763 | * use-after-free issues. Note that the device number stays | |
764 | * "in-use" until we really free the gendisk. | |
765 | */ | |
766 | blk_invalidate_devt(disk_devt(disk)); | |
d2bf1b67 TH |
767 | |
768 | kobject_put(disk->part0.holder_dir); | |
769 | kobject_put(disk->slave_dir); | |
d2bf1b67 TH |
770 | |
771 | part_stat_set_all(&disk->part0, 0); | |
772 | disk->part0.stamp = 0; | |
d2bf1b67 TH |
773 | if (!sysfs_deprecated) |
774 | sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk))); | |
25e823c8 | 775 | pm_runtime_set_memalloc_noio(disk_to_dev(disk), false); |
d2bf1b67 | 776 | device_del(disk_to_dev(disk)); |
1da177e4 | 777 | } |
d2bf1b67 | 778 | EXPORT_SYMBOL(del_gendisk); |
1da177e4 | 779 | |
99e6608c VV |
780 | /* sysfs access to bad-blocks list. */ |
781 | static ssize_t disk_badblocks_show(struct device *dev, | |
782 | struct device_attribute *attr, | |
783 | char *page) | |
784 | { | |
785 | struct gendisk *disk = dev_to_disk(dev); | |
786 | ||
787 | if (!disk->bb) | |
788 | return sprintf(page, "\n"); | |
789 | ||
790 | return badblocks_show(disk->bb, page, 0); | |
791 | } | |
792 | ||
793 | static ssize_t disk_badblocks_store(struct device *dev, | |
794 | struct device_attribute *attr, | |
795 | const char *page, size_t len) | |
796 | { | |
797 | struct gendisk *disk = dev_to_disk(dev); | |
798 | ||
799 | if (!disk->bb) | |
800 | return -ENXIO; | |
801 | ||
802 | return badblocks_store(disk->bb, page, len, 0); | |
803 | } | |
804 | ||
1da177e4 LT |
805 | /** |
806 | * get_gendisk - get partitioning information for a given device | |
710027a4 | 807 | * @devt: device to get partitioning information for |
496aa8a9 | 808 | * @partno: returned partition index |
1da177e4 LT |
809 | * |
810 | * This function gets the structure containing partitioning | |
710027a4 | 811 | * information for the given device @devt. |
1da177e4 | 812 | */ |
cf771cb5 | 813 | struct gendisk *get_gendisk(dev_t devt, int *partno) |
1da177e4 | 814 | { |
bcce3de1 TH |
815 | struct gendisk *disk = NULL; |
816 | ||
817 | if (MAJOR(devt) != BLOCK_EXT_MAJOR) { | |
818 | struct kobject *kobj; | |
819 | ||
820 | kobj = kobj_lookup(bdev_map, devt, partno); | |
821 | if (kobj) | |
822 | disk = dev_to_disk(kobj_to_dev(kobj)); | |
823 | } else { | |
824 | struct hd_struct *part; | |
825 | ||
4d66e5e9 | 826 | spin_lock_bh(&ext_devt_lock); |
870d6656 | 827 | part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); |
bcce3de1 TH |
828 | if (part && get_disk(part_to_disk(part))) { |
829 | *partno = part->partno; | |
830 | disk = part_to_disk(part); | |
831 | } | |
4d66e5e9 | 832 | spin_unlock_bh(&ext_devt_lock); |
bcce3de1 | 833 | } |
edfaa7c3 | 834 | |
f0fba398 | 835 | if (disk && unlikely(disk->flags & GENHD_FL_HIDDEN)) { |
8ddcd653 CH |
836 | put_disk(disk); |
837 | disk = NULL; | |
838 | } | |
bcce3de1 | 839 | return disk; |
1da177e4 | 840 | } |
b6ac23af | 841 | EXPORT_SYMBOL(get_gendisk); |
1da177e4 | 842 | |
f331c029 TH |
843 | /** |
844 | * bdget_disk - do bdget() by gendisk and partition number | |
845 | * @disk: gendisk of interest | |
846 | * @partno: partition number | |
847 | * | |
848 | * Find partition @partno from @disk, do bdget() on it. | |
849 | * | |
850 | * CONTEXT: | |
851 | * Don't care. | |
852 | * | |
853 | * RETURNS: | |
854 | * Resulting block_device on success, NULL on failure. | |
855 | */ | |
aeb3d3a8 | 856 | struct block_device *bdget_disk(struct gendisk *disk, int partno) |
f331c029 | 857 | { |
548b10eb TH |
858 | struct hd_struct *part; |
859 | struct block_device *bdev = NULL; | |
f331c029 | 860 | |
548b10eb | 861 | part = disk_get_part(disk, partno); |
2bbedcb4 | 862 | if (part) |
548b10eb TH |
863 | bdev = bdget(part_devt(part)); |
864 | disk_put_part(part); | |
f331c029 | 865 | |
548b10eb | 866 | return bdev; |
f331c029 TH |
867 | } |
868 | EXPORT_SYMBOL(bdget_disk); | |
869 | ||
5c6f35c5 GKH |
870 | /* |
871 | * print a full list of all partitions - intended for places where the root | |
872 | * filesystem can't be mounted and thus to give the victim some idea of what | |
873 | * went wrong | |
874 | */ | |
875 | void __init printk_all_partitions(void) | |
876 | { | |
def4e38d TH |
877 | struct class_dev_iter iter; |
878 | struct device *dev; | |
879 | ||
880 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); | |
881 | while ((dev = class_dev_iter_next(&iter))) { | |
882 | struct gendisk *disk = dev_to_disk(dev); | |
e71bf0d0 TH |
883 | struct disk_part_iter piter; |
884 | struct hd_struct *part; | |
1f014290 TH |
885 | char name_buf[BDEVNAME_SIZE]; |
886 | char devt_buf[BDEVT_SIZE]; | |
def4e38d TH |
887 | |
888 | /* | |
889 | * Don't show empty devices or things that have been | |
25985edc | 890 | * suppressed |
def4e38d TH |
891 | */ |
892 | if (get_capacity(disk) == 0 || | |
893 | (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)) | |
894 | continue; | |
895 | ||
896 | /* | |
897 | * Note, unlike /proc/partitions, I am showing the | |
898 | * numbers in hex - the same format as the root= | |
899 | * option takes. | |
900 | */ | |
074a7aca TH |
901 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0); |
902 | while ((part = disk_part_iter_next(&piter))) { | |
903 | bool is_part0 = part == &disk->part0; | |
def4e38d | 904 | |
b5af921e | 905 | printk("%s%s %10llu %s %s", is_part0 ? "" : " ", |
1f014290 | 906 | bdevt_str(part_devt(part), devt_buf), |
c83f6bf9 VG |
907 | (unsigned long long)part_nr_sects_read(part) >> 1 |
908 | , disk_name(disk, part->partno, name_buf), | |
1ad7e899 | 909 | part->info ? part->info->uuid : ""); |
074a7aca | 910 | if (is_part0) { |
52c44d93 | 911 | if (dev->parent && dev->parent->driver) |
074a7aca | 912 | printk(" driver: %s\n", |
52c44d93 | 913 | dev->parent->driver->name); |
074a7aca TH |
914 | else |
915 | printk(" (driver?)\n"); | |
916 | } else | |
917 | printk("\n"); | |
918 | } | |
e71bf0d0 | 919 | disk_part_iter_exit(&piter); |
def4e38d TH |
920 | } |
921 | class_dev_iter_exit(&iter); | |
dd2a345f DG |
922 | } |
923 | ||
1da177e4 LT |
924 | #ifdef CONFIG_PROC_FS |
925 | /* iterator */ | |
def4e38d | 926 | static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos) |
68c4d4a7 | 927 | { |
def4e38d TH |
928 | loff_t skip = *pos; |
929 | struct class_dev_iter *iter; | |
930 | struct device *dev; | |
68c4d4a7 | 931 | |
aeb3d3a8 | 932 | iter = kmalloc(sizeof(*iter), GFP_KERNEL); |
def4e38d TH |
933 | if (!iter) |
934 | return ERR_PTR(-ENOMEM); | |
935 | ||
936 | seqf->private = iter; | |
937 | class_dev_iter_init(iter, &block_class, NULL, &disk_type); | |
938 | do { | |
939 | dev = class_dev_iter_next(iter); | |
940 | if (!dev) | |
941 | return NULL; | |
942 | } while (skip--); | |
943 | ||
944 | return dev_to_disk(dev); | |
68c4d4a7 GKH |
945 | } |
946 | ||
def4e38d | 947 | static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos) |
1da177e4 | 948 | { |
edfaa7c3 | 949 | struct device *dev; |
1da177e4 | 950 | |
def4e38d TH |
951 | (*pos)++; |
952 | dev = class_dev_iter_next(seqf->private); | |
2ac3cee5 | 953 | if (dev) |
68c4d4a7 | 954 | return dev_to_disk(dev); |
2ac3cee5 | 955 | |
1da177e4 LT |
956 | return NULL; |
957 | } | |
958 | ||
def4e38d | 959 | static void disk_seqf_stop(struct seq_file *seqf, void *v) |
27f30251 | 960 | { |
def4e38d | 961 | struct class_dev_iter *iter = seqf->private; |
27f30251 | 962 | |
def4e38d TH |
963 | /* stop is called even after start failed :-( */ |
964 | if (iter) { | |
965 | class_dev_iter_exit(iter); | |
966 | kfree(iter); | |
77da1605 | 967 | seqf->private = NULL; |
5c0ef6d0 | 968 | } |
1da177e4 LT |
969 | } |
970 | ||
def4e38d | 971 | static void *show_partition_start(struct seq_file *seqf, loff_t *pos) |
1da177e4 | 972 | { |
06768067 | 973 | void *p; |
def4e38d TH |
974 | |
975 | p = disk_seqf_start(seqf, pos); | |
b9f985b6 | 976 | if (!IS_ERR_OR_NULL(p) && !*pos) |
def4e38d TH |
977 | seq_puts(seqf, "major minor #blocks name\n\n"); |
978 | return p; | |
1da177e4 LT |
979 | } |
980 | ||
cf771cb5 | 981 | static int show_partition(struct seq_file *seqf, void *v) |
1da177e4 LT |
982 | { |
983 | struct gendisk *sgp = v; | |
e71bf0d0 TH |
984 | struct disk_part_iter piter; |
985 | struct hd_struct *part; | |
1da177e4 LT |
986 | char buf[BDEVNAME_SIZE]; |
987 | ||
1da177e4 | 988 | /* Don't show non-partitionable removeable devices or empty devices */ |
d27769ec | 989 | if (!get_capacity(sgp) || (!disk_max_parts(sgp) && |
f331c029 | 990 | (sgp->flags & GENHD_FL_REMOVABLE))) |
1da177e4 LT |
991 | return 0; |
992 | if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO) | |
993 | return 0; | |
994 | ||
995 | /* show the full disk and all non-0 size partitions of it */ | |
074a7aca | 996 | disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0); |
e71bf0d0 | 997 | while ((part = disk_part_iter_next(&piter))) |
1f014290 | 998 | seq_printf(seqf, "%4d %7d %10llu %s\n", |
f331c029 | 999 | MAJOR(part_devt(part)), MINOR(part_devt(part)), |
c83f6bf9 | 1000 | (unsigned long long)part_nr_sects_read(part) >> 1, |
f331c029 | 1001 | disk_name(sgp, part->partno, buf)); |
e71bf0d0 | 1002 | disk_part_iter_exit(&piter); |
1da177e4 LT |
1003 | |
1004 | return 0; | |
1005 | } | |
1006 | ||
f500975a | 1007 | static const struct seq_operations partitions_op = { |
def4e38d TH |
1008 | .start = show_partition_start, |
1009 | .next = disk_seqf_next, | |
1010 | .stop = disk_seqf_stop, | |
edfaa7c3 | 1011 | .show = show_partition |
1da177e4 | 1012 | }; |
f500975a AD |
1013 | |
1014 | static int partitions_open(struct inode *inode, struct file *file) | |
1015 | { | |
1016 | return seq_open(file, &partitions_op); | |
1017 | } | |
1018 | ||
1019 | static const struct file_operations proc_partitions_operations = { | |
1020 | .open = partitions_open, | |
1021 | .read = seq_read, | |
1022 | .llseek = seq_lseek, | |
1023 | .release = seq_release, | |
1024 | }; | |
1da177e4 LT |
1025 | #endif |
1026 | ||
1027 | ||
cf771cb5 | 1028 | static struct kobject *base_probe(dev_t devt, int *partno, void *data) |
1da177e4 | 1029 | { |
edfaa7c3 | 1030 | if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0) |
1da177e4 | 1031 | /* Make old-style 2.4 aliases work */ |
edfaa7c3 | 1032 | request_module("block-major-%d", MAJOR(devt)); |
1da177e4 LT |
1033 | return NULL; |
1034 | } | |
1035 | ||
1036 | static int __init genhd_device_init(void) | |
1037 | { | |
e105b8bf DW |
1038 | int error; |
1039 | ||
1040 | block_class.dev_kobj = sysfs_dev_block_kobj; | |
1041 | error = class_register(&block_class); | |
ee27a558 RM |
1042 | if (unlikely(error)) |
1043 | return error; | |
edfaa7c3 | 1044 | bdev_map = kobj_map_init(base_probe, &block_class_lock); |
1da177e4 | 1045 | blk_dev_init(); |
edfaa7c3 | 1046 | |
561ec68e ZY |
1047 | register_blkdev(BLOCK_EXT_MAJOR, "blkext"); |
1048 | ||
edfaa7c3 | 1049 | /* create top-level block dir */ |
e52eec13 AK |
1050 | if (!sysfs_deprecated) |
1051 | block_depr = kobject_create_and_add("block", NULL); | |
830d3cfb | 1052 | return 0; |
1da177e4 LT |
1053 | } |
1054 | ||
1055 | subsys_initcall(genhd_device_init); | |
1056 | ||
edfaa7c3 KS |
1057 | static ssize_t disk_range_show(struct device *dev, |
1058 | struct device_attribute *attr, char *buf) | |
1da177e4 | 1059 | { |
edfaa7c3 | 1060 | struct gendisk *disk = dev_to_disk(dev); |
1da177e4 | 1061 | |
edfaa7c3 | 1062 | return sprintf(buf, "%d\n", disk->minors); |
1da177e4 LT |
1063 | } |
1064 | ||
1f014290 TH |
1065 | static ssize_t disk_ext_range_show(struct device *dev, |
1066 | struct device_attribute *attr, char *buf) | |
1067 | { | |
1068 | struct gendisk *disk = dev_to_disk(dev); | |
1069 | ||
b5d0b9df | 1070 | return sprintf(buf, "%d\n", disk_max_parts(disk)); |
1f014290 TH |
1071 | } |
1072 | ||
edfaa7c3 KS |
1073 | static ssize_t disk_removable_show(struct device *dev, |
1074 | struct device_attribute *attr, char *buf) | |
a7fd6706 | 1075 | { |
edfaa7c3 | 1076 | struct gendisk *disk = dev_to_disk(dev); |
a7fd6706 | 1077 | |
edfaa7c3 KS |
1078 | return sprintf(buf, "%d\n", |
1079 | (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0)); | |
a7fd6706 KS |
1080 | } |
1081 | ||
8ddcd653 CH |
1082 | static ssize_t disk_hidden_show(struct device *dev, |
1083 | struct device_attribute *attr, char *buf) | |
1084 | { | |
1085 | struct gendisk *disk = dev_to_disk(dev); | |
1086 | ||
1087 | return sprintf(buf, "%d\n", | |
1088 | (disk->flags & GENHD_FL_HIDDEN ? 1 : 0)); | |
1089 | } | |
1090 | ||
1c9ce527 KS |
1091 | static ssize_t disk_ro_show(struct device *dev, |
1092 | struct device_attribute *attr, char *buf) | |
1093 | { | |
1094 | struct gendisk *disk = dev_to_disk(dev); | |
1095 | ||
b7db9956 | 1096 | return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0); |
1c9ce527 KS |
1097 | } |
1098 | ||
edfaa7c3 KS |
1099 | static ssize_t disk_capability_show(struct device *dev, |
1100 | struct device_attribute *attr, char *buf) | |
86ce18d7 | 1101 | { |
edfaa7c3 KS |
1102 | struct gendisk *disk = dev_to_disk(dev); |
1103 | ||
1104 | return sprintf(buf, "%x\n", disk->flags); | |
86ce18d7 | 1105 | } |
edfaa7c3 | 1106 | |
c72758f3 MP |
1107 | static ssize_t disk_alignment_offset_show(struct device *dev, |
1108 | struct device_attribute *attr, | |
1109 | char *buf) | |
1110 | { | |
1111 | struct gendisk *disk = dev_to_disk(dev); | |
1112 | ||
1113 | return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue)); | |
1114 | } | |
1115 | ||
86b37281 MP |
1116 | static ssize_t disk_discard_alignment_show(struct device *dev, |
1117 | struct device_attribute *attr, | |
1118 | char *buf) | |
1119 | { | |
1120 | struct gendisk *disk = dev_to_disk(dev); | |
1121 | ||
dd3d145d | 1122 | return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue)); |
86b37281 MP |
1123 | } |
1124 | ||
edfaa7c3 | 1125 | static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL); |
1f014290 | 1126 | static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL); |
edfaa7c3 | 1127 | static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL); |
8ddcd653 | 1128 | static DEVICE_ATTR(hidden, S_IRUGO, disk_hidden_show, NULL); |
1c9ce527 | 1129 | static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL); |
e5610521 | 1130 | static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL); |
c72758f3 | 1131 | static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL); |
86b37281 MP |
1132 | static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show, |
1133 | NULL); | |
edfaa7c3 | 1134 | static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL); |
074a7aca | 1135 | static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL); |
316d315b | 1136 | static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL); |
99e6608c VV |
1137 | static DEVICE_ATTR(badblocks, S_IRUGO | S_IWUSR, disk_badblocks_show, |
1138 | disk_badblocks_store); | |
c17bb495 | 1139 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
edfaa7c3 | 1140 | static struct device_attribute dev_attr_fail = |
eddb2e26 | 1141 | __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store); |
c17bb495 | 1142 | #endif |
581d4e28 JA |
1143 | #ifdef CONFIG_FAIL_IO_TIMEOUT |
1144 | static struct device_attribute dev_attr_fail_timeout = | |
1145 | __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show, | |
1146 | part_timeout_store); | |
1147 | #endif | |
edfaa7c3 KS |
1148 | |
1149 | static struct attribute *disk_attrs[] = { | |
1150 | &dev_attr_range.attr, | |
1f014290 | 1151 | &dev_attr_ext_range.attr, |
edfaa7c3 | 1152 | &dev_attr_removable.attr, |
8ddcd653 | 1153 | &dev_attr_hidden.attr, |
1c9ce527 | 1154 | &dev_attr_ro.attr, |
edfaa7c3 | 1155 | &dev_attr_size.attr, |
c72758f3 | 1156 | &dev_attr_alignment_offset.attr, |
86b37281 | 1157 | &dev_attr_discard_alignment.attr, |
edfaa7c3 KS |
1158 | &dev_attr_capability.attr, |
1159 | &dev_attr_stat.attr, | |
316d315b | 1160 | &dev_attr_inflight.attr, |
99e6608c | 1161 | &dev_attr_badblocks.attr, |
edfaa7c3 KS |
1162 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
1163 | &dev_attr_fail.attr, | |
581d4e28 JA |
1164 | #endif |
1165 | #ifdef CONFIG_FAIL_IO_TIMEOUT | |
1166 | &dev_attr_fail_timeout.attr, | |
edfaa7c3 KS |
1167 | #endif |
1168 | NULL | |
1169 | }; | |
1170 | ||
9438b3e0 DW |
1171 | static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n) |
1172 | { | |
1173 | struct device *dev = container_of(kobj, typeof(*dev), kobj); | |
1174 | struct gendisk *disk = dev_to_disk(dev); | |
1175 | ||
1176 | if (a == &dev_attr_badblocks.attr && !disk->bb) | |
1177 | return 0; | |
1178 | return a->mode; | |
1179 | } | |
1180 | ||
edfaa7c3 KS |
1181 | static struct attribute_group disk_attr_group = { |
1182 | .attrs = disk_attrs, | |
9438b3e0 | 1183 | .is_visible = disk_visible, |
edfaa7c3 KS |
1184 | }; |
1185 | ||
a4dbd674 | 1186 | static const struct attribute_group *disk_attr_groups[] = { |
edfaa7c3 KS |
1187 | &disk_attr_group, |
1188 | NULL | |
1da177e4 LT |
1189 | }; |
1190 | ||
540eed56 TH |
1191 | /** |
1192 | * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way | |
1193 | * @disk: disk to replace part_tbl for | |
1194 | * @new_ptbl: new part_tbl to install | |
1195 | * | |
1196 | * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The | |
1197 | * original ptbl is freed using RCU callback. | |
1198 | * | |
1199 | * LOCKING: | |
6d2cf6f2 | 1200 | * Matching bd_mutex locked or the caller is the only user of @disk. |
540eed56 TH |
1201 | */ |
1202 | static void disk_replace_part_tbl(struct gendisk *disk, | |
1203 | struct disk_part_tbl *new_ptbl) | |
1204 | { | |
6d2cf6f2 BVA |
1205 | struct disk_part_tbl *old_ptbl = |
1206 | rcu_dereference_protected(disk->part_tbl, 1); | |
540eed56 TH |
1207 | |
1208 | rcu_assign_pointer(disk->part_tbl, new_ptbl); | |
a6f23657 JA |
1209 | |
1210 | if (old_ptbl) { | |
1211 | rcu_assign_pointer(old_ptbl->last_lookup, NULL); | |
57bdfbf9 | 1212 | kfree_rcu(old_ptbl, rcu_head); |
a6f23657 | 1213 | } |
540eed56 TH |
1214 | } |
1215 | ||
1216 | /** | |
1217 | * disk_expand_part_tbl - expand disk->part_tbl | |
1218 | * @disk: disk to expand part_tbl for | |
1219 | * @partno: expand such that this partno can fit in | |
1220 | * | |
1221 | * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl | |
1222 | * uses RCU to allow unlocked dereferencing for stats and other stuff. | |
1223 | * | |
1224 | * LOCKING: | |
6d2cf6f2 BVA |
1225 | * Matching bd_mutex locked or the caller is the only user of @disk. |
1226 | * Might sleep. | |
540eed56 TH |
1227 | * |
1228 | * RETURNS: | |
1229 | * 0 on success, -errno on failure. | |
1230 | */ | |
1231 | int disk_expand_part_tbl(struct gendisk *disk, int partno) | |
1232 | { | |
6d2cf6f2 BVA |
1233 | struct disk_part_tbl *old_ptbl = |
1234 | rcu_dereference_protected(disk->part_tbl, 1); | |
540eed56 TH |
1235 | struct disk_part_tbl *new_ptbl; |
1236 | int len = old_ptbl ? old_ptbl->len : 0; | |
5fabcb4c | 1237 | int i, target; |
540eed56 | 1238 | size_t size; |
5fabcb4c JA |
1239 | |
1240 | /* | |
1241 | * check for int overflow, since we can get here from blkpg_ioctl() | |
1242 | * with a user passed 'partno'. | |
1243 | */ | |
1244 | target = partno + 1; | |
1245 | if (target < 0) | |
1246 | return -EINVAL; | |
540eed56 TH |
1247 | |
1248 | /* disk_max_parts() is zero during initialization, ignore if so */ | |
1249 | if (disk_max_parts(disk) && target > disk_max_parts(disk)) | |
1250 | return -EINVAL; | |
1251 | ||
1252 | if (target <= len) | |
1253 | return 0; | |
1254 | ||
1255 | size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]); | |
1256 | new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id); | |
1257 | if (!new_ptbl) | |
1258 | return -ENOMEM; | |
1259 | ||
540eed56 TH |
1260 | new_ptbl->len = target; |
1261 | ||
1262 | for (i = 0; i < len; i++) | |
1263 | rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]); | |
1264 | ||
1265 | disk_replace_part_tbl(disk, new_ptbl); | |
1266 | return 0; | |
1267 | } | |
1268 | ||
edfaa7c3 | 1269 | static void disk_release(struct device *dev) |
1da177e4 | 1270 | { |
edfaa7c3 KS |
1271 | struct gendisk *disk = dev_to_disk(dev); |
1272 | ||
2da78092 | 1273 | blk_free_devt(dev->devt); |
77ea887e | 1274 | disk_release_events(disk); |
1da177e4 | 1275 | kfree(disk->random); |
540eed56 | 1276 | disk_replace_part_tbl(disk, NULL); |
b54e5ed8 | 1277 | hd_free_part(&disk->part0); |
523e1d39 TH |
1278 | if (disk->queue) |
1279 | blk_put_queue(disk->queue); | |
1da177e4 LT |
1280 | kfree(disk); |
1281 | } | |
edfaa7c3 KS |
1282 | struct class block_class = { |
1283 | .name = "block", | |
1da177e4 LT |
1284 | }; |
1285 | ||
3c2670e6 | 1286 | static char *block_devnode(struct device *dev, umode_t *mode, |
4e4098a3 | 1287 | kuid_t *uid, kgid_t *gid) |
b03f38b6 KS |
1288 | { |
1289 | struct gendisk *disk = dev_to_disk(dev); | |
1290 | ||
e454cea2 KS |
1291 | if (disk->devnode) |
1292 | return disk->devnode(disk, mode); | |
b03f38b6 KS |
1293 | return NULL; |
1294 | } | |
1295 | ||
edf8ff55 | 1296 | static const struct device_type disk_type = { |
edfaa7c3 KS |
1297 | .name = "disk", |
1298 | .groups = disk_attr_groups, | |
1299 | .release = disk_release, | |
e454cea2 | 1300 | .devnode = block_devnode, |
1da177e4 LT |
1301 | }; |
1302 | ||
a6e2ba88 | 1303 | #ifdef CONFIG_PROC_FS |
cf771cb5 TH |
1304 | /* |
1305 | * aggregate disk stat collector. Uses the same stats that the sysfs | |
1306 | * entries do, above, but makes them available through one seq_file. | |
1307 | * | |
1308 | * The output looks suspiciously like /proc/partitions with a bunch of | |
1309 | * extra fields. | |
1310 | */ | |
1311 | static int diskstats_show(struct seq_file *seqf, void *v) | |
1da177e4 LT |
1312 | { |
1313 | struct gendisk *gp = v; | |
e71bf0d0 TH |
1314 | struct disk_part_iter piter; |
1315 | struct hd_struct *hd; | |
1da177e4 | 1316 | char buf[BDEVNAME_SIZE]; |
0609e0ef | 1317 | unsigned int inflight[2]; |
c9959059 | 1318 | int cpu; |
1da177e4 LT |
1319 | |
1320 | /* | |
ed9e1982 | 1321 | if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next) |
cf771cb5 | 1322 | seq_puts(seqf, "major minor name" |
1da177e4 LT |
1323 | " rio rmerge rsect ruse wio wmerge " |
1324 | "wsect wuse running use aveq" | |
1325 | "\n\n"); | |
1326 | */ | |
9f5e4865 | 1327 | |
71982a40 | 1328 | disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0); |
e71bf0d0 | 1329 | while ((hd = disk_part_iter_next(&piter))) { |
074a7aca | 1330 | cpu = part_stat_lock(); |
d62e26b3 | 1331 | part_round_stats(gp->queue, cpu, hd); |
074a7aca | 1332 | part_stat_unlock(); |
0609e0ef | 1333 | part_in_flight(gp->queue, hd, inflight); |
f95fe9cf HP |
1334 | seq_printf(seqf, "%4d %7d %s %lu %lu %lu " |
1335 | "%u %lu %lu %lu %u %u %u %u\n", | |
f331c029 TH |
1336 | MAJOR(part_devt(hd)), MINOR(part_devt(hd)), |
1337 | disk_name(gp, hd->partno, buf), | |
53f22956 LY |
1338 | part_stat_read(hd, ios[READ]), |
1339 | part_stat_read(hd, merges[READ]), | |
f95fe9cf | 1340 | part_stat_read(hd, sectors[READ]), |
53f22956 LY |
1341 | jiffies_to_msecs(part_stat_read(hd, ticks[READ])), |
1342 | part_stat_read(hd, ios[WRITE]), | |
1343 | part_stat_read(hd, merges[WRITE]), | |
f95fe9cf | 1344 | part_stat_read(hd, sectors[WRITE]), |
53f22956 | 1345 | jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])), |
0609e0ef | 1346 | inflight[0], |
28f39d55 JM |
1347 | jiffies_to_msecs(part_stat_read(hd, io_ticks)), |
1348 | jiffies_to_msecs(part_stat_read(hd, time_in_queue)) | |
1349 | ); | |
1da177e4 | 1350 | } |
e71bf0d0 | 1351 | disk_part_iter_exit(&piter); |
9f5e4865 | 1352 | |
1da177e4 LT |
1353 | return 0; |
1354 | } | |
1355 | ||
31d85ab2 | 1356 | static const struct seq_operations diskstats_op = { |
def4e38d TH |
1357 | .start = disk_seqf_start, |
1358 | .next = disk_seqf_next, | |
1359 | .stop = disk_seqf_stop, | |
1da177e4 LT |
1360 | .show = diskstats_show |
1361 | }; | |
f500975a | 1362 | |
31d85ab2 AD |
1363 | static int diskstats_open(struct inode *inode, struct file *file) |
1364 | { | |
1365 | return seq_open(file, &diskstats_op); | |
1366 | } | |
1367 | ||
1368 | static const struct file_operations proc_diskstats_operations = { | |
1369 | .open = diskstats_open, | |
1370 | .read = seq_read, | |
1371 | .llseek = seq_lseek, | |
1372 | .release = seq_release, | |
1373 | }; | |
1374 | ||
f500975a AD |
1375 | static int __init proc_genhd_init(void) |
1376 | { | |
31d85ab2 | 1377 | proc_create("diskstats", 0, NULL, &proc_diskstats_operations); |
f500975a AD |
1378 | proc_create("partitions", 0, NULL, &proc_partitions_operations); |
1379 | return 0; | |
1380 | } | |
1381 | module_init(proc_genhd_init); | |
a6e2ba88 | 1382 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 1383 | |
cf771cb5 | 1384 | dev_t blk_lookup_devt(const char *name, int partno) |
a142be85 | 1385 | { |
def4e38d TH |
1386 | dev_t devt = MKDEV(0, 0); |
1387 | struct class_dev_iter iter; | |
1388 | struct device *dev; | |
a142be85 | 1389 | |
def4e38d TH |
1390 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); |
1391 | while ((dev = class_dev_iter_next(&iter))) { | |
a142be85 | 1392 | struct gendisk *disk = dev_to_disk(dev); |
548b10eb | 1393 | struct hd_struct *part; |
a142be85 | 1394 | |
3ada8b7e | 1395 | if (strcmp(dev_name(dev), name)) |
f331c029 | 1396 | continue; |
f331c029 | 1397 | |
41b8c853 NB |
1398 | if (partno < disk->minors) { |
1399 | /* We need to return the right devno, even | |
1400 | * if the partition doesn't exist yet. | |
1401 | */ | |
1402 | devt = MKDEV(MAJOR(dev->devt), | |
1403 | MINOR(dev->devt) + partno); | |
1404 | break; | |
1405 | } | |
548b10eb | 1406 | part = disk_get_part(disk, partno); |
2bbedcb4 | 1407 | if (part) { |
f331c029 | 1408 | devt = part_devt(part); |
e71bf0d0 | 1409 | disk_put_part(part); |
548b10eb | 1410 | break; |
def4e38d | 1411 | } |
548b10eb | 1412 | disk_put_part(part); |
5c0ef6d0 | 1413 | } |
def4e38d | 1414 | class_dev_iter_exit(&iter); |
edfaa7c3 KS |
1415 | return devt; |
1416 | } | |
edfaa7c3 KS |
1417 | EXPORT_SYMBOL(blk_lookup_devt); |
1418 | ||
e319e1fb | 1419 | struct gendisk *__alloc_disk_node(int minors, int node_id) |
1946089a CL |
1420 | { |
1421 | struct gendisk *disk; | |
6d2cf6f2 | 1422 | struct disk_part_tbl *ptbl; |
1946089a | 1423 | |
de65b012 CH |
1424 | if (minors > DISK_MAX_PARTS) { |
1425 | printk(KERN_ERR | |
7fb52621 | 1426 | "block: can't allocate more than %d partitions\n", |
de65b012 CH |
1427 | DISK_MAX_PARTS); |
1428 | minors = DISK_MAX_PARTS; | |
1429 | } | |
1946089a | 1430 | |
c1b511eb | 1431 | disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id); |
1da177e4 | 1432 | if (disk) { |
074a7aca | 1433 | if (!init_part_stats(&disk->part0)) { |
1da177e4 LT |
1434 | kfree(disk); |
1435 | return NULL; | |
1436 | } | |
bf91db18 | 1437 | disk->node_id = node_id; |
540eed56 TH |
1438 | if (disk_expand_part_tbl(disk, 0)) { |
1439 | free_part_stats(&disk->part0); | |
b5d0b9df TH |
1440 | kfree(disk); |
1441 | return NULL; | |
1da177e4 | 1442 | } |
6d2cf6f2 BVA |
1443 | ptbl = rcu_dereference_protected(disk->part_tbl, 1); |
1444 | rcu_assign_pointer(ptbl->part[0], &disk->part0); | |
6c23a968 | 1445 | |
c83f6bf9 VG |
1446 | /* |
1447 | * set_capacity() and get_capacity() currently don't use | |
1448 | * seqcounter to read/update the part0->nr_sects. Still init | |
1449 | * the counter as we can read the sectors in IO submission | |
1450 | * patch using seqence counters. | |
1451 | * | |
1452 | * TODO: Ideally set_capacity() and get_capacity() should be | |
1453 | * converted to make use of bd_mutex and sequence counters. | |
1454 | */ | |
1455 | seqcount_init(&disk->part0.nr_sects_seq); | |
6c71013e ML |
1456 | if (hd_ref_init(&disk->part0)) { |
1457 | hd_free_part(&disk->part0); | |
1458 | kfree(disk); | |
1459 | return NULL; | |
1460 | } | |
b5d0b9df | 1461 | |
1da177e4 | 1462 | disk->minors = minors; |
1da177e4 | 1463 | rand_initialize_disk(disk); |
ed9e1982 TH |
1464 | disk_to_dev(disk)->class = &block_class; |
1465 | disk_to_dev(disk)->type = &disk_type; | |
1466 | device_initialize(disk_to_dev(disk)); | |
1da177e4 LT |
1467 | } |
1468 | return disk; | |
1469 | } | |
e319e1fb | 1470 | EXPORT_SYMBOL(__alloc_disk_node); |
1da177e4 LT |
1471 | |
1472 | struct kobject *get_disk(struct gendisk *disk) | |
1473 | { | |
1474 | struct module *owner; | |
1475 | struct kobject *kobj; | |
1476 | ||
1477 | if (!disk->fops) | |
1478 | return NULL; | |
1479 | owner = disk->fops->owner; | |
1480 | if (owner && !try_module_get(owner)) | |
1481 | return NULL; | |
d01b2dcb | 1482 | kobj = kobject_get_unless_zero(&disk_to_dev(disk)->kobj); |
1da177e4 LT |
1483 | if (kobj == NULL) { |
1484 | module_put(owner); | |
1485 | return NULL; | |
1486 | } | |
1487 | return kobj; | |
1488 | ||
1489 | } | |
1490 | ||
1491 | EXPORT_SYMBOL(get_disk); | |
1492 | ||
1493 | void put_disk(struct gendisk *disk) | |
1494 | { | |
1495 | if (disk) | |
ed9e1982 | 1496 | kobject_put(&disk_to_dev(disk)->kobj); |
1da177e4 LT |
1497 | } |
1498 | ||
1499 | EXPORT_SYMBOL(put_disk); | |
1500 | ||
e3264a4d HR |
1501 | static void set_disk_ro_uevent(struct gendisk *gd, int ro) |
1502 | { | |
1503 | char event[] = "DISK_RO=1"; | |
1504 | char *envp[] = { event, NULL }; | |
1505 | ||
1506 | if (!ro) | |
1507 | event[8] = '0'; | |
1508 | kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); | |
1509 | } | |
1510 | ||
1da177e4 LT |
1511 | void set_device_ro(struct block_device *bdev, int flag) |
1512 | { | |
b7db9956 | 1513 | bdev->bd_part->policy = flag; |
1da177e4 LT |
1514 | } |
1515 | ||
1516 | EXPORT_SYMBOL(set_device_ro); | |
1517 | ||
1518 | void set_disk_ro(struct gendisk *disk, int flag) | |
1519 | { | |
e71bf0d0 TH |
1520 | struct disk_part_iter piter; |
1521 | struct hd_struct *part; | |
1522 | ||
e3264a4d HR |
1523 | if (disk->part0.policy != flag) { |
1524 | set_disk_ro_uevent(disk, flag); | |
1525 | disk->part0.policy = flag; | |
1526 | } | |
1527 | ||
1528 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY); | |
e71bf0d0 TH |
1529 | while ((part = disk_part_iter_next(&piter))) |
1530 | part->policy = flag; | |
1531 | disk_part_iter_exit(&piter); | |
1da177e4 LT |
1532 | } |
1533 | ||
1534 | EXPORT_SYMBOL(set_disk_ro); | |
1535 | ||
1536 | int bdev_read_only(struct block_device *bdev) | |
1537 | { | |
1538 | if (!bdev) | |
1539 | return 0; | |
b7db9956 | 1540 | return bdev->bd_part->policy; |
1da177e4 LT |
1541 | } |
1542 | ||
1543 | EXPORT_SYMBOL(bdev_read_only); | |
1544 | ||
cf771cb5 | 1545 | int invalidate_partition(struct gendisk *disk, int partno) |
1da177e4 LT |
1546 | { |
1547 | int res = 0; | |
cf771cb5 | 1548 | struct block_device *bdev = bdget_disk(disk, partno); |
1da177e4 | 1549 | if (bdev) { |
2ef41634 | 1550 | fsync_bdev(bdev); |
93b270f7 | 1551 | res = __invalidate_device(bdev, true); |
1da177e4 LT |
1552 | bdput(bdev); |
1553 | } | |
1554 | return res; | |
1555 | } | |
1556 | ||
1557 | EXPORT_SYMBOL(invalidate_partition); | |
77ea887e TH |
1558 | |
1559 | /* | |
1560 | * Disk events - monitor disk events like media change and eject request. | |
1561 | */ | |
1562 | struct disk_events { | |
1563 | struct list_head node; /* all disk_event's */ | |
1564 | struct gendisk *disk; /* the associated disk */ | |
1565 | spinlock_t lock; | |
1566 | ||
fdd514e1 | 1567 | struct mutex block_mutex; /* protects blocking */ |
77ea887e TH |
1568 | int block; /* event blocking depth */ |
1569 | unsigned int pending; /* events already sent out */ | |
1570 | unsigned int clearing; /* events being cleared */ | |
1571 | ||
1572 | long poll_msecs; /* interval, -1 for default */ | |
1573 | struct delayed_work dwork; | |
1574 | }; | |
1575 | ||
1576 | static const char *disk_events_strs[] = { | |
1577 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change", | |
1578 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request", | |
1579 | }; | |
1580 | ||
1581 | static char *disk_uevents[] = { | |
1582 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1", | |
1583 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1", | |
1584 | }; | |
1585 | ||
1586 | /* list of all disk_events */ | |
1587 | static DEFINE_MUTEX(disk_events_mutex); | |
1588 | static LIST_HEAD(disk_events); | |
1589 | ||
1590 | /* disable in-kernel polling by default */ | |
1fe8f348 | 1591 | static unsigned long disk_events_dfl_poll_msecs; |
77ea887e TH |
1592 | |
1593 | static unsigned long disk_events_poll_jiffies(struct gendisk *disk) | |
1594 | { | |
1595 | struct disk_events *ev = disk->ev; | |
1596 | long intv_msecs = 0; | |
1597 | ||
1598 | /* | |
1599 | * If device-specific poll interval is set, always use it. If | |
1600 | * the default is being used, poll iff there are events which | |
1601 | * can't be monitored asynchronously. | |
1602 | */ | |
1603 | if (ev->poll_msecs >= 0) | |
1604 | intv_msecs = ev->poll_msecs; | |
1605 | else if (disk->events & ~disk->async_events) | |
1606 | intv_msecs = disk_events_dfl_poll_msecs; | |
1607 | ||
1608 | return msecs_to_jiffies(intv_msecs); | |
1609 | } | |
1610 | ||
c3af54af TH |
1611 | /** |
1612 | * disk_block_events - block and flush disk event checking | |
1613 | * @disk: disk to block events for | |
1614 | * | |
1615 | * On return from this function, it is guaranteed that event checking | |
1616 | * isn't in progress and won't happen until unblocked by | |
1617 | * disk_unblock_events(). Events blocking is counted and the actual | |
1618 | * unblocking happens after the matching number of unblocks are done. | |
1619 | * | |
1620 | * Note that this intentionally does not block event checking from | |
1621 | * disk_clear_events(). | |
1622 | * | |
1623 | * CONTEXT: | |
1624 | * Might sleep. | |
1625 | */ | |
1626 | void disk_block_events(struct gendisk *disk) | |
77ea887e TH |
1627 | { |
1628 | struct disk_events *ev = disk->ev; | |
1629 | unsigned long flags; | |
1630 | bool cancel; | |
1631 | ||
c3af54af TH |
1632 | if (!ev) |
1633 | return; | |
1634 | ||
fdd514e1 TH |
1635 | /* |
1636 | * Outer mutex ensures that the first blocker completes canceling | |
1637 | * the event work before further blockers are allowed to finish. | |
1638 | */ | |
1639 | mutex_lock(&ev->block_mutex); | |
1640 | ||
77ea887e TH |
1641 | spin_lock_irqsave(&ev->lock, flags); |
1642 | cancel = !ev->block++; | |
1643 | spin_unlock_irqrestore(&ev->lock, flags); | |
1644 | ||
c3af54af TH |
1645 | if (cancel) |
1646 | cancel_delayed_work_sync(&disk->ev->dwork); | |
fdd514e1 TH |
1647 | |
1648 | mutex_unlock(&ev->block_mutex); | |
77ea887e TH |
1649 | } |
1650 | ||
1651 | static void __disk_unblock_events(struct gendisk *disk, bool check_now) | |
1652 | { | |
1653 | struct disk_events *ev = disk->ev; | |
1654 | unsigned long intv; | |
1655 | unsigned long flags; | |
1656 | ||
1657 | spin_lock_irqsave(&ev->lock, flags); | |
1658 | ||
1659 | if (WARN_ON_ONCE(ev->block <= 0)) | |
1660 | goto out_unlock; | |
1661 | ||
1662 | if (--ev->block) | |
1663 | goto out_unlock; | |
1664 | ||
77ea887e | 1665 | intv = disk_events_poll_jiffies(disk); |
77ea887e | 1666 | if (check_now) |
695588f9 VK |
1667 | queue_delayed_work(system_freezable_power_efficient_wq, |
1668 | &ev->dwork, 0); | |
77ea887e | 1669 | else if (intv) |
695588f9 VK |
1670 | queue_delayed_work(system_freezable_power_efficient_wq, |
1671 | &ev->dwork, intv); | |
77ea887e TH |
1672 | out_unlock: |
1673 | spin_unlock_irqrestore(&ev->lock, flags); | |
1674 | } | |
1675 | ||
77ea887e TH |
1676 | /** |
1677 | * disk_unblock_events - unblock disk event checking | |
1678 | * @disk: disk to unblock events for | |
1679 | * | |
1680 | * Undo disk_block_events(). When the block count reaches zero, it | |
1681 | * starts events polling if configured. | |
1682 | * | |
1683 | * CONTEXT: | |
1684 | * Don't care. Safe to call from irq context. | |
1685 | */ | |
1686 | void disk_unblock_events(struct gendisk *disk) | |
1687 | { | |
1688 | if (disk->ev) | |
facc31dd | 1689 | __disk_unblock_events(disk, false); |
77ea887e TH |
1690 | } |
1691 | ||
1692 | /** | |
85ef06d1 TH |
1693 | * disk_flush_events - schedule immediate event checking and flushing |
1694 | * @disk: disk to check and flush events for | |
1695 | * @mask: events to flush | |
77ea887e | 1696 | * |
85ef06d1 TH |
1697 | * Schedule immediate event checking on @disk if not blocked. Events in |
1698 | * @mask are scheduled to be cleared from the driver. Note that this | |
1699 | * doesn't clear the events from @disk->ev. | |
77ea887e TH |
1700 | * |
1701 | * CONTEXT: | |
85ef06d1 | 1702 | * If @mask is non-zero must be called with bdev->bd_mutex held. |
77ea887e | 1703 | */ |
85ef06d1 | 1704 | void disk_flush_events(struct gendisk *disk, unsigned int mask) |
77ea887e | 1705 | { |
a9dce2a3 | 1706 | struct disk_events *ev = disk->ev; |
a9dce2a3 TH |
1707 | |
1708 | if (!ev) | |
1709 | return; | |
1710 | ||
85ef06d1 TH |
1711 | spin_lock_irq(&ev->lock); |
1712 | ev->clearing |= mask; | |
41f63c53 | 1713 | if (!ev->block) |
695588f9 VK |
1714 | mod_delayed_work(system_freezable_power_efficient_wq, |
1715 | &ev->dwork, 0); | |
85ef06d1 | 1716 | spin_unlock_irq(&ev->lock); |
77ea887e | 1717 | } |
77ea887e TH |
1718 | |
1719 | /** | |
1720 | * disk_clear_events - synchronously check, clear and return pending events | |
1721 | * @disk: disk to fetch and clear events from | |
da3dae54 | 1722 | * @mask: mask of events to be fetched and cleared |
77ea887e TH |
1723 | * |
1724 | * Disk events are synchronously checked and pending events in @mask | |
1725 | * are cleared and returned. This ignores the block count. | |
1726 | * | |
1727 | * CONTEXT: | |
1728 | * Might sleep. | |
1729 | */ | |
1730 | unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask) | |
1731 | { | |
1732 | const struct block_device_operations *bdops = disk->fops; | |
1733 | struct disk_events *ev = disk->ev; | |
1734 | unsigned int pending; | |
12c2bdb2 | 1735 | unsigned int clearing = mask; |
77ea887e TH |
1736 | |
1737 | if (!ev) { | |
1738 | /* for drivers still using the old ->media_changed method */ | |
1739 | if ((mask & DISK_EVENT_MEDIA_CHANGE) && | |
1740 | bdops->media_changed && bdops->media_changed(disk)) | |
1741 | return DISK_EVENT_MEDIA_CHANGE; | |
1742 | return 0; | |
1743 | } | |
1744 | ||
12c2bdb2 DB |
1745 | disk_block_events(disk); |
1746 | ||
1747 | /* | |
1748 | * store the union of mask and ev->clearing on the stack so that the | |
1749 | * race with disk_flush_events does not cause ambiguity (ev->clearing | |
1750 | * can still be modified even if events are blocked). | |
1751 | */ | |
77ea887e | 1752 | spin_lock_irq(&ev->lock); |
12c2bdb2 DB |
1753 | clearing |= ev->clearing; |
1754 | ev->clearing = 0; | |
77ea887e TH |
1755 | spin_unlock_irq(&ev->lock); |
1756 | ||
12c2bdb2 | 1757 | disk_check_events(ev, &clearing); |
aea24a8b | 1758 | /* |
12c2bdb2 DB |
1759 | * if ev->clearing is not 0, the disk_flush_events got called in the |
1760 | * middle of this function, so we want to run the workfn without delay. | |
aea24a8b | 1761 | */ |
12c2bdb2 | 1762 | __disk_unblock_events(disk, ev->clearing ? true : false); |
77ea887e TH |
1763 | |
1764 | /* then, fetch and clear pending events */ | |
1765 | spin_lock_irq(&ev->lock); | |
77ea887e TH |
1766 | pending = ev->pending & mask; |
1767 | ev->pending &= ~mask; | |
1768 | spin_unlock_irq(&ev->lock); | |
12c2bdb2 | 1769 | WARN_ON_ONCE(clearing & mask); |
77ea887e TH |
1770 | |
1771 | return pending; | |
1772 | } | |
1773 | ||
12c2bdb2 DB |
1774 | /* |
1775 | * Separate this part out so that a different pointer for clearing_ptr can be | |
1776 | * passed in for disk_clear_events. | |
1777 | */ | |
77ea887e TH |
1778 | static void disk_events_workfn(struct work_struct *work) |
1779 | { | |
1780 | struct delayed_work *dwork = to_delayed_work(work); | |
1781 | struct disk_events *ev = container_of(dwork, struct disk_events, dwork); | |
12c2bdb2 DB |
1782 | |
1783 | disk_check_events(ev, &ev->clearing); | |
1784 | } | |
1785 | ||
1786 | static void disk_check_events(struct disk_events *ev, | |
1787 | unsigned int *clearing_ptr) | |
1788 | { | |
77ea887e TH |
1789 | struct gendisk *disk = ev->disk; |
1790 | char *envp[ARRAY_SIZE(disk_uevents) + 1] = { }; | |
12c2bdb2 | 1791 | unsigned int clearing = *clearing_ptr; |
77ea887e TH |
1792 | unsigned int events; |
1793 | unsigned long intv; | |
1794 | int nr_events = 0, i; | |
1795 | ||
1796 | /* check events */ | |
1797 | events = disk->fops->check_events(disk, clearing); | |
1798 | ||
1799 | /* accumulate pending events and schedule next poll if necessary */ | |
1800 | spin_lock_irq(&ev->lock); | |
1801 | ||
1802 | events &= ~ev->pending; | |
1803 | ev->pending |= events; | |
12c2bdb2 | 1804 | *clearing_ptr &= ~clearing; |
77ea887e TH |
1805 | |
1806 | intv = disk_events_poll_jiffies(disk); | |
1807 | if (!ev->block && intv) | |
695588f9 VK |
1808 | queue_delayed_work(system_freezable_power_efficient_wq, |
1809 | &ev->dwork, intv); | |
77ea887e TH |
1810 | |
1811 | spin_unlock_irq(&ev->lock); | |
1812 | ||
7c88a168 TH |
1813 | /* |
1814 | * Tell userland about new events. Only the events listed in | |
1815 | * @disk->events are reported. Unlisted events are processed the | |
1816 | * same internally but never get reported to userland. | |
1817 | */ | |
77ea887e | 1818 | for (i = 0; i < ARRAY_SIZE(disk_uevents); i++) |
7c88a168 | 1819 | if (events & disk->events & (1 << i)) |
77ea887e TH |
1820 | envp[nr_events++] = disk_uevents[i]; |
1821 | ||
1822 | if (nr_events) | |
1823 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
1824 | } | |
1825 | ||
1826 | /* | |
1827 | * A disk events enabled device has the following sysfs nodes under | |
1828 | * its /sys/block/X/ directory. | |
1829 | * | |
1830 | * events : list of all supported events | |
1831 | * events_async : list of events which can be detected w/o polling | |
1832 | * events_poll_msecs : polling interval, 0: disable, -1: system default | |
1833 | */ | |
1834 | static ssize_t __disk_events_show(unsigned int events, char *buf) | |
1835 | { | |
1836 | const char *delim = ""; | |
1837 | ssize_t pos = 0; | |
1838 | int i; | |
1839 | ||
1840 | for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++) | |
1841 | if (events & (1 << i)) { | |
1842 | pos += sprintf(buf + pos, "%s%s", | |
1843 | delim, disk_events_strs[i]); | |
1844 | delim = " "; | |
1845 | } | |
1846 | if (pos) | |
1847 | pos += sprintf(buf + pos, "\n"); | |
1848 | return pos; | |
1849 | } | |
1850 | ||
1851 | static ssize_t disk_events_show(struct device *dev, | |
1852 | struct device_attribute *attr, char *buf) | |
1853 | { | |
1854 | struct gendisk *disk = dev_to_disk(dev); | |
1855 | ||
1856 | return __disk_events_show(disk->events, buf); | |
1857 | } | |
1858 | ||
1859 | static ssize_t disk_events_async_show(struct device *dev, | |
1860 | struct device_attribute *attr, char *buf) | |
1861 | { | |
1862 | struct gendisk *disk = dev_to_disk(dev); | |
1863 | ||
1864 | return __disk_events_show(disk->async_events, buf); | |
1865 | } | |
1866 | ||
1867 | static ssize_t disk_events_poll_msecs_show(struct device *dev, | |
1868 | struct device_attribute *attr, | |
1869 | char *buf) | |
1870 | { | |
1871 | struct gendisk *disk = dev_to_disk(dev); | |
1872 | ||
1873 | return sprintf(buf, "%ld\n", disk->ev->poll_msecs); | |
1874 | } | |
1875 | ||
1876 | static ssize_t disk_events_poll_msecs_store(struct device *dev, | |
1877 | struct device_attribute *attr, | |
1878 | const char *buf, size_t count) | |
1879 | { | |
1880 | struct gendisk *disk = dev_to_disk(dev); | |
1881 | long intv; | |
1882 | ||
1883 | if (!count || !sscanf(buf, "%ld", &intv)) | |
1884 | return -EINVAL; | |
1885 | ||
1886 | if (intv < 0 && intv != -1) | |
1887 | return -EINVAL; | |
1888 | ||
c3af54af | 1889 | disk_block_events(disk); |
77ea887e TH |
1890 | disk->ev->poll_msecs = intv; |
1891 | __disk_unblock_events(disk, true); | |
1892 | ||
1893 | return count; | |
1894 | } | |
1895 | ||
1896 | static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL); | |
1897 | static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL); | |
1898 | static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR, | |
1899 | disk_events_poll_msecs_show, | |
1900 | disk_events_poll_msecs_store); | |
1901 | ||
1902 | static const struct attribute *disk_events_attrs[] = { | |
1903 | &dev_attr_events.attr, | |
1904 | &dev_attr_events_async.attr, | |
1905 | &dev_attr_events_poll_msecs.attr, | |
1906 | NULL, | |
1907 | }; | |
1908 | ||
1909 | /* | |
1910 | * The default polling interval can be specified by the kernel | |
1911 | * parameter block.events_dfl_poll_msecs which defaults to 0 | |
1912 | * (disable). This can also be modified runtime by writing to | |
1913 | * /sys/module/block/events_dfl_poll_msecs. | |
1914 | */ | |
1915 | static int disk_events_set_dfl_poll_msecs(const char *val, | |
1916 | const struct kernel_param *kp) | |
1917 | { | |
1918 | struct disk_events *ev; | |
1919 | int ret; | |
1920 | ||
1921 | ret = param_set_ulong(val, kp); | |
1922 | if (ret < 0) | |
1923 | return ret; | |
1924 | ||
1925 | mutex_lock(&disk_events_mutex); | |
1926 | ||
1927 | list_for_each_entry(ev, &disk_events, node) | |
85ef06d1 | 1928 | disk_flush_events(ev->disk, 0); |
77ea887e TH |
1929 | |
1930 | mutex_unlock(&disk_events_mutex); | |
1931 | ||
1932 | return 0; | |
1933 | } | |
1934 | ||
1935 | static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = { | |
1936 | .set = disk_events_set_dfl_poll_msecs, | |
1937 | .get = param_get_ulong, | |
1938 | }; | |
1939 | ||
1940 | #undef MODULE_PARAM_PREFIX | |
1941 | #define MODULE_PARAM_PREFIX "block." | |
1942 | ||
1943 | module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops, | |
1944 | &disk_events_dfl_poll_msecs, 0644); | |
1945 | ||
1946 | /* | |
9f53d2fe | 1947 | * disk_{alloc|add|del|release}_events - initialize and destroy disk_events. |
77ea887e | 1948 | */ |
9f53d2fe | 1949 | static void disk_alloc_events(struct gendisk *disk) |
77ea887e TH |
1950 | { |
1951 | struct disk_events *ev; | |
1952 | ||
75e3f3ee | 1953 | if (!disk->fops->check_events) |
77ea887e TH |
1954 | return; |
1955 | ||
1956 | ev = kzalloc(sizeof(*ev), GFP_KERNEL); | |
1957 | if (!ev) { | |
1958 | pr_warn("%s: failed to initialize events\n", disk->disk_name); | |
1959 | return; | |
1960 | } | |
1961 | ||
77ea887e TH |
1962 | INIT_LIST_HEAD(&ev->node); |
1963 | ev->disk = disk; | |
1964 | spin_lock_init(&ev->lock); | |
fdd514e1 | 1965 | mutex_init(&ev->block_mutex); |
77ea887e TH |
1966 | ev->block = 1; |
1967 | ev->poll_msecs = -1; | |
1968 | INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn); | |
1969 | ||
9f53d2fe SG |
1970 | disk->ev = ev; |
1971 | } | |
1972 | ||
1973 | static void disk_add_events(struct gendisk *disk) | |
1974 | { | |
1975 | if (!disk->ev) | |
1976 | return; | |
1977 | ||
1978 | /* FIXME: error handling */ | |
1979 | if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0) | |
1980 | pr_warn("%s: failed to create sysfs files for events\n", | |
1981 | disk->disk_name); | |
1982 | ||
77ea887e | 1983 | mutex_lock(&disk_events_mutex); |
9f53d2fe | 1984 | list_add_tail(&disk->ev->node, &disk_events); |
77ea887e TH |
1985 | mutex_unlock(&disk_events_mutex); |
1986 | ||
1987 | /* | |
1988 | * Block count is initialized to 1 and the following initial | |
1989 | * unblock kicks it into action. | |
1990 | */ | |
1991 | __disk_unblock_events(disk, true); | |
1992 | } | |
1993 | ||
1994 | static void disk_del_events(struct gendisk *disk) | |
1995 | { | |
1996 | if (!disk->ev) | |
1997 | return; | |
1998 | ||
c3af54af | 1999 | disk_block_events(disk); |
77ea887e TH |
2000 | |
2001 | mutex_lock(&disk_events_mutex); | |
2002 | list_del_init(&disk->ev->node); | |
2003 | mutex_unlock(&disk_events_mutex); | |
2004 | ||
2005 | sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs); | |
2006 | } | |
2007 | ||
2008 | static void disk_release_events(struct gendisk *disk) | |
2009 | { | |
2010 | /* the block count should be 1 from disk_del_events() */ | |
2011 | WARN_ON_ONCE(disk->ev && disk->ev->block != 1); | |
2012 | kfree(disk->ev); | |
2013 | } |