Commit | Line | Data |
---|---|---|
3dcf60bc | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * gendisk handling | |
4 | */ | |
5 | ||
1da177e4 | 6 | #include <linux/module.h> |
3ad5cee5 | 7 | #include <linux/ctype.h> |
1da177e4 LT |
8 | #include <linux/fs.h> |
9 | #include <linux/genhd.h> | |
b446b60e | 10 | #include <linux/kdev_t.h> |
1da177e4 LT |
11 | #include <linux/kernel.h> |
12 | #include <linux/blkdev.h> | |
66114cad | 13 | #include <linux/backing-dev.h> |
1da177e4 LT |
14 | #include <linux/init.h> |
15 | #include <linux/spinlock.h> | |
f500975a | 16 | #include <linux/proc_fs.h> |
1da177e4 LT |
17 | #include <linux/seq_file.h> |
18 | #include <linux/slab.h> | |
19 | #include <linux/kmod.h> | |
20 | #include <linux/kobj_map.h> | |
58383af6 | 21 | #include <linux/mutex.h> |
bcce3de1 | 22 | #include <linux/idr.h> |
77ea887e | 23 | #include <linux/log2.h> |
25e823c8 | 24 | #include <linux/pm_runtime.h> |
99e6608c | 25 | #include <linux/badblocks.h> |
1da177e4 | 26 | |
ff88972c AB |
27 | #include "blk.h" |
28 | ||
edfaa7c3 | 29 | static DEFINE_MUTEX(block_class_lock); |
31eb6186 | 30 | static struct kobject *block_depr; |
1da177e4 | 31 | |
bcce3de1 | 32 | /* for extended dynamic devt allocation, currently only one major is used */ |
ce23bba8 | 33 | #define NR_EXT_DEVT (1 << MINORBITS) |
bcce3de1 | 34 | |
2da78092 | 35 | /* For extended devt allocation. ext_devt_lock prevents look up |
bcce3de1 TH |
36 | * results from going away underneath its user. |
37 | */ | |
2da78092 | 38 | static DEFINE_SPINLOCK(ext_devt_lock); |
bcce3de1 TH |
39 | static DEFINE_IDR(ext_devt_idr); |
40 | ||
edf8ff55 | 41 | static const struct device_type disk_type; |
1826eadf | 42 | |
12c2bdb2 DB |
43 | static void disk_check_events(struct disk_events *ev, |
44 | unsigned int *clearing_ptr); | |
9f53d2fe | 45 | static void disk_alloc_events(struct gendisk *disk); |
77ea887e TH |
46 | static void disk_add_events(struct gendisk *disk); |
47 | static void disk_del_events(struct gendisk *disk); | |
48 | static void disk_release_events(struct gendisk *disk); | |
49 | ||
e598a72f BS |
50 | /* |
51 | * Set disk capacity and notify if the size is not currently | |
52 | * zero and will not be set to zero | |
53 | */ | |
54 | void set_capacity_revalidate_and_notify(struct gendisk *disk, sector_t size, | |
55 | bool revalidate) | |
56 | { | |
57 | sector_t capacity = get_capacity(disk); | |
58 | ||
59 | set_capacity(disk, size); | |
60 | ||
61 | if (revalidate) | |
62 | revalidate_disk(disk); | |
63 | ||
64 | if (capacity != size && capacity != 0 && size != 0) { | |
65 | char *envp[] = { "RESIZE=1", NULL }; | |
66 | ||
67 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
68 | } | |
69 | } | |
70 | ||
71 | EXPORT_SYMBOL_GPL(set_capacity_revalidate_and_notify); | |
72 | ||
5cbd28e3 CH |
73 | /* |
74 | * Format the device name of the indicated disk into the supplied buffer and | |
75 | * return a pointer to that same buffer for convenience. | |
76 | */ | |
77 | char *disk_name(struct gendisk *hd, int partno, char *buf) | |
78 | { | |
79 | if (!partno) | |
80 | snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name); | |
81 | else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1])) | |
82 | snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno); | |
83 | else | |
84 | snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno); | |
85 | ||
86 | return buf; | |
87 | } | |
88 | ||
89 | const char *bdevname(struct block_device *bdev, char *buf) | |
90 | { | |
91 | return disk_name(bdev->bd_disk, bdev->bd_part->partno, buf); | |
92 | } | |
93 | EXPORT_SYMBOL(bdevname); | |
e598a72f | 94 | |
ea18e0f0 KK |
95 | static void part_stat_read_all(struct hd_struct *part, struct disk_stats *stat) |
96 | { | |
97 | int cpu; | |
98 | ||
99 | memset(stat, 0, sizeof(struct disk_stats)); | |
100 | for_each_possible_cpu(cpu) { | |
101 | struct disk_stats *ptr = per_cpu_ptr(part->dkstats, cpu); | |
102 | int group; | |
103 | ||
104 | for (group = 0; group < NR_STAT_GROUPS; group++) { | |
105 | stat->nsecs[group] += ptr->nsecs[group]; | |
106 | stat->sectors[group] += ptr->sectors[group]; | |
107 | stat->ios[group] += ptr->ios[group]; | |
108 | stat->merges[group] += ptr->merges[group]; | |
109 | } | |
110 | ||
111 | stat->io_ticks += ptr->io_ticks; | |
ea18e0f0 KK |
112 | } |
113 | } | |
ea18e0f0 | 114 | |
6005771c CH |
115 | static unsigned int part_in_flight(struct request_queue *q, |
116 | struct hd_struct *part) | |
f299b7c7 | 117 | { |
b2f609e1 | 118 | unsigned int inflight = 0; |
1226b8dd | 119 | int cpu; |
f299b7c7 | 120 | |
1226b8dd | 121 | for_each_possible_cpu(cpu) { |
e016b782 MP |
122 | inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) + |
123 | part_stat_local_read_cpu(part, in_flight[1], cpu); | |
1226b8dd | 124 | } |
e016b782 MP |
125 | if ((int)inflight < 0) |
126 | inflight = 0; | |
1226b8dd | 127 | |
e016b782 | 128 | return inflight; |
f299b7c7 JA |
129 | } |
130 | ||
6005771c CH |
131 | static void part_in_flight_rw(struct request_queue *q, struct hd_struct *part, |
132 | unsigned int inflight[2]) | |
bf0ddaba | 133 | { |
1226b8dd MP |
134 | int cpu; |
135 | ||
1226b8dd MP |
136 | inflight[0] = 0; |
137 | inflight[1] = 0; | |
138 | for_each_possible_cpu(cpu) { | |
139 | inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu); | |
140 | inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu); | |
141 | } | |
142 | if ((int)inflight[0] < 0) | |
143 | inflight[0] = 0; | |
144 | if ((int)inflight[1] < 0) | |
145 | inflight[1] = 0; | |
bf0ddaba OS |
146 | } |
147 | ||
807d4af2 CH |
148 | struct hd_struct *__disk_get_part(struct gendisk *disk, int partno) |
149 | { | |
150 | struct disk_part_tbl *ptbl = rcu_dereference(disk->part_tbl); | |
151 | ||
152 | if (unlikely(partno < 0 || partno >= ptbl->len)) | |
153 | return NULL; | |
154 | return rcu_dereference(ptbl->part[partno]); | |
155 | } | |
156 | ||
e71bf0d0 TH |
157 | /** |
158 | * disk_get_part - get partition | |
159 | * @disk: disk to look partition from | |
160 | * @partno: partition number | |
161 | * | |
162 | * Look for partition @partno from @disk. If found, increment | |
163 | * reference count and return it. | |
164 | * | |
165 | * CONTEXT: | |
166 | * Don't care. | |
167 | * | |
168 | * RETURNS: | |
169 | * Pointer to the found partition on success, NULL if not found. | |
170 | */ | |
171 | struct hd_struct *disk_get_part(struct gendisk *disk, int partno) | |
172 | { | |
807d4af2 | 173 | struct hd_struct *part; |
540eed56 | 174 | |
e71bf0d0 | 175 | rcu_read_lock(); |
807d4af2 CH |
176 | part = __disk_get_part(disk, partno); |
177 | if (part) | |
178 | get_device(part_to_dev(part)); | |
e71bf0d0 TH |
179 | rcu_read_unlock(); |
180 | ||
181 | return part; | |
182 | } | |
e71bf0d0 TH |
183 | |
184 | /** | |
185 | * disk_part_iter_init - initialize partition iterator | |
186 | * @piter: iterator to initialize | |
187 | * @disk: disk to iterate over | |
188 | * @flags: DISK_PITER_* flags | |
189 | * | |
190 | * Initialize @piter so that it iterates over partitions of @disk. | |
191 | * | |
192 | * CONTEXT: | |
193 | * Don't care. | |
194 | */ | |
195 | void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk, | |
196 | unsigned int flags) | |
197 | { | |
540eed56 TH |
198 | struct disk_part_tbl *ptbl; |
199 | ||
200 | rcu_read_lock(); | |
201 | ptbl = rcu_dereference(disk->part_tbl); | |
202 | ||
e71bf0d0 TH |
203 | piter->disk = disk; |
204 | piter->part = NULL; | |
205 | ||
206 | if (flags & DISK_PITER_REVERSE) | |
540eed56 | 207 | piter->idx = ptbl->len - 1; |
71982a40 | 208 | else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0)) |
e71bf0d0 | 209 | piter->idx = 0; |
b5d0b9df TH |
210 | else |
211 | piter->idx = 1; | |
e71bf0d0 TH |
212 | |
213 | piter->flags = flags; | |
540eed56 TH |
214 | |
215 | rcu_read_unlock(); | |
e71bf0d0 TH |
216 | } |
217 | EXPORT_SYMBOL_GPL(disk_part_iter_init); | |
218 | ||
219 | /** | |
220 | * disk_part_iter_next - proceed iterator to the next partition and return it | |
221 | * @piter: iterator of interest | |
222 | * | |
223 | * Proceed @piter to the next partition and return it. | |
224 | * | |
225 | * CONTEXT: | |
226 | * Don't care. | |
227 | */ | |
228 | struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter) | |
229 | { | |
540eed56 | 230 | struct disk_part_tbl *ptbl; |
e71bf0d0 TH |
231 | int inc, end; |
232 | ||
233 | /* put the last partition */ | |
234 | disk_put_part(piter->part); | |
235 | piter->part = NULL; | |
236 | ||
540eed56 | 237 | /* get part_tbl */ |
e71bf0d0 | 238 | rcu_read_lock(); |
540eed56 | 239 | ptbl = rcu_dereference(piter->disk->part_tbl); |
e71bf0d0 TH |
240 | |
241 | /* determine iteration parameters */ | |
242 | if (piter->flags & DISK_PITER_REVERSE) { | |
243 | inc = -1; | |
71982a40 TH |
244 | if (piter->flags & (DISK_PITER_INCL_PART0 | |
245 | DISK_PITER_INCL_EMPTY_PART0)) | |
b5d0b9df TH |
246 | end = -1; |
247 | else | |
248 | end = 0; | |
e71bf0d0 TH |
249 | } else { |
250 | inc = 1; | |
540eed56 | 251 | end = ptbl->len; |
e71bf0d0 TH |
252 | } |
253 | ||
254 | /* iterate to the next partition */ | |
255 | for (; piter->idx != end; piter->idx += inc) { | |
256 | struct hd_struct *part; | |
257 | ||
540eed56 | 258 | part = rcu_dereference(ptbl->part[piter->idx]); |
e71bf0d0 TH |
259 | if (!part) |
260 | continue; | |
c83f6bf9 | 261 | if (!part_nr_sects_read(part) && |
71982a40 TH |
262 | !(piter->flags & DISK_PITER_INCL_EMPTY) && |
263 | !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 && | |
264 | piter->idx == 0)) | |
e71bf0d0 TH |
265 | continue; |
266 | ||
ed9e1982 | 267 | get_device(part_to_dev(part)); |
e71bf0d0 TH |
268 | piter->part = part; |
269 | piter->idx += inc; | |
270 | break; | |
271 | } | |
272 | ||
273 | rcu_read_unlock(); | |
274 | ||
275 | return piter->part; | |
276 | } | |
277 | EXPORT_SYMBOL_GPL(disk_part_iter_next); | |
278 | ||
279 | /** | |
280 | * disk_part_iter_exit - finish up partition iteration | |
281 | * @piter: iter of interest | |
282 | * | |
283 | * Called when iteration is over. Cleans up @piter. | |
284 | * | |
285 | * CONTEXT: | |
286 | * Don't care. | |
287 | */ | |
288 | void disk_part_iter_exit(struct disk_part_iter *piter) | |
289 | { | |
290 | disk_put_part(piter->part); | |
291 | piter->part = NULL; | |
292 | } | |
293 | EXPORT_SYMBOL_GPL(disk_part_iter_exit); | |
294 | ||
a6f23657 JA |
295 | static inline int sector_in_part(struct hd_struct *part, sector_t sector) |
296 | { | |
297 | return part->start_sect <= sector && | |
c83f6bf9 | 298 | sector < part->start_sect + part_nr_sects_read(part); |
a6f23657 JA |
299 | } |
300 | ||
e71bf0d0 TH |
301 | /** |
302 | * disk_map_sector_rcu - map sector to partition | |
303 | * @disk: gendisk of interest | |
304 | * @sector: sector to map | |
305 | * | |
306 | * Find out which partition @sector maps to on @disk. This is | |
307 | * primarily used for stats accounting. | |
308 | * | |
309 | * CONTEXT: | |
b7d6c303 | 310 | * RCU read locked. The returned partition pointer is always valid |
27eb3af9 ML |
311 | * because its refcount is grabbed except for part0, which lifetime |
312 | * is same with the disk. | |
e71bf0d0 TH |
313 | * |
314 | * RETURNS: | |
074a7aca | 315 | * Found partition on success, part0 is returned if no partition matches |
b7d6c303 | 316 | * or the matched partition is being deleted. |
e71bf0d0 TH |
317 | */ |
318 | struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector) | |
319 | { | |
540eed56 | 320 | struct disk_part_tbl *ptbl; |
a6f23657 | 321 | struct hd_struct *part; |
e71bf0d0 TH |
322 | int i; |
323 | ||
8ab1d40a | 324 | rcu_read_lock(); |
540eed56 TH |
325 | ptbl = rcu_dereference(disk->part_tbl); |
326 | ||
a6f23657 | 327 | part = rcu_dereference(ptbl->last_lookup); |
b7d6c303 | 328 | if (part && sector_in_part(part, sector) && hd_struct_try_get(part)) |
8ab1d40a | 329 | goto out_unlock; |
a6f23657 | 330 | |
540eed56 | 331 | for (i = 1; i < ptbl->len; i++) { |
a6f23657 | 332 | part = rcu_dereference(ptbl->part[i]); |
e71bf0d0 | 333 | |
a6f23657 | 334 | if (part && sector_in_part(part, sector)) { |
b7d6c303 ML |
335 | /* |
336 | * only live partition can be cached for lookup, | |
337 | * so use-after-free on cached & deleting partition | |
338 | * can be avoided | |
339 | */ | |
340 | if (!hd_struct_try_get(part)) | |
341 | break; | |
a6f23657 | 342 | rcu_assign_pointer(ptbl->last_lookup, part); |
8ab1d40a | 343 | goto out_unlock; |
a6f23657 | 344 | } |
e71bf0d0 | 345 | } |
8ab1d40a KK |
346 | |
347 | part = &disk->part0; | |
348 | out_unlock: | |
349 | rcu_read_unlock(); | |
350 | return part; | |
e71bf0d0 | 351 | } |
e71bf0d0 | 352 | |
b53df2e7 SK |
353 | /** |
354 | * disk_has_partitions | |
355 | * @disk: gendisk of interest | |
356 | * | |
357 | * Walk through the partition table and check if valid partition exists. | |
358 | * | |
359 | * CONTEXT: | |
360 | * Don't care. | |
361 | * | |
362 | * RETURNS: | |
363 | * True if the gendisk has at least one valid non-zero size partition. | |
364 | * Otherwise false. | |
365 | */ | |
366 | bool disk_has_partitions(struct gendisk *disk) | |
367 | { | |
368 | struct disk_part_tbl *ptbl; | |
369 | int i; | |
370 | bool ret = false; | |
371 | ||
372 | rcu_read_lock(); | |
373 | ptbl = rcu_dereference(disk->part_tbl); | |
374 | ||
375 | /* Iterate partitions skipping the whole device at index 0 */ | |
376 | for (i = 1; i < ptbl->len; i++) { | |
377 | if (rcu_dereference(ptbl->part[i])) { | |
378 | ret = true; | |
379 | break; | |
380 | } | |
381 | } | |
382 | ||
383 | rcu_read_unlock(); | |
384 | ||
385 | return ret; | |
386 | } | |
387 | EXPORT_SYMBOL_GPL(disk_has_partitions); | |
388 | ||
1da177e4 LT |
389 | /* |
390 | * Can be deleted altogether. Later. | |
391 | * | |
392 | */ | |
133d55cd | 393 | #define BLKDEV_MAJOR_HASH_SIZE 255 |
1da177e4 LT |
394 | static struct blk_major_name { |
395 | struct blk_major_name *next; | |
396 | int major; | |
397 | char name[16]; | |
68eef3b4 | 398 | } *major_names[BLKDEV_MAJOR_HASH_SIZE]; |
1da177e4 LT |
399 | |
400 | /* index in the above - for now: assume no multimajor ranges */ | |
e61eb2e9 | 401 | static inline int major_to_index(unsigned major) |
1da177e4 | 402 | { |
68eef3b4 | 403 | return major % BLKDEV_MAJOR_HASH_SIZE; |
7170be5f NH |
404 | } |
405 | ||
68eef3b4 | 406 | #ifdef CONFIG_PROC_FS |
cf771cb5 | 407 | void blkdev_show(struct seq_file *seqf, off_t offset) |
7170be5f | 408 | { |
68eef3b4 | 409 | struct blk_major_name *dp; |
7170be5f | 410 | |
133d55cd LG |
411 | mutex_lock(&block_class_lock); |
412 | for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next) | |
413 | if (dp->major == offset) | |
cf771cb5 | 414 | seq_printf(seqf, "%3d %s\n", dp->major, dp->name); |
133d55cd | 415 | mutex_unlock(&block_class_lock); |
1da177e4 | 416 | } |
68eef3b4 | 417 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 418 | |
9e8c0bcc MN |
419 | /** |
420 | * register_blkdev - register a new block device | |
421 | * | |
f33ff110 SB |
422 | * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If |
423 | * @major = 0, try to allocate any unused major number. | |
9e8c0bcc MN |
424 | * @name: the name of the new block device as a zero terminated string |
425 | * | |
426 | * The @name must be unique within the system. | |
427 | * | |
0e056eb5 | 428 | * The return value depends on the @major input parameter: |
429 | * | |
f33ff110 SB |
430 | * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1] |
431 | * then the function returns zero on success, or a negative error code | |
0e056eb5 | 432 | * - if any unused major number was requested with @major = 0 parameter |
9e8c0bcc | 433 | * then the return value is the allocated major number in range |
f33ff110 SB |
434 | * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise |
435 | * | |
436 | * See Documentation/admin-guide/devices.txt for the list of allocated | |
437 | * major numbers. | |
9e8c0bcc | 438 | */ |
1da177e4 LT |
439 | int register_blkdev(unsigned int major, const char *name) |
440 | { | |
441 | struct blk_major_name **n, *p; | |
442 | int index, ret = 0; | |
443 | ||
edfaa7c3 | 444 | mutex_lock(&block_class_lock); |
1da177e4 LT |
445 | |
446 | /* temporary */ | |
447 | if (major == 0) { | |
448 | for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) { | |
449 | if (major_names[index] == NULL) | |
450 | break; | |
451 | } | |
452 | ||
453 | if (index == 0) { | |
dfc76d11 KP |
454 | printk("%s: failed to get major for %s\n", |
455 | __func__, name); | |
1da177e4 LT |
456 | ret = -EBUSY; |
457 | goto out; | |
458 | } | |
459 | major = index; | |
460 | ret = major; | |
461 | } | |
462 | ||
133d55cd | 463 | if (major >= BLKDEV_MAJOR_MAX) { |
dfc76d11 KP |
464 | pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n", |
465 | __func__, major, BLKDEV_MAJOR_MAX-1, name); | |
133d55cd LG |
466 | |
467 | ret = -EINVAL; | |
468 | goto out; | |
469 | } | |
470 | ||
1da177e4 LT |
471 | p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL); |
472 | if (p == NULL) { | |
473 | ret = -ENOMEM; | |
474 | goto out; | |
475 | } | |
476 | ||
477 | p->major = major; | |
478 | strlcpy(p->name, name, sizeof(p->name)); | |
479 | p->next = NULL; | |
480 | index = major_to_index(major); | |
481 | ||
482 | for (n = &major_names[index]; *n; n = &(*n)->next) { | |
483 | if ((*n)->major == major) | |
484 | break; | |
485 | } | |
486 | if (!*n) | |
487 | *n = p; | |
488 | else | |
489 | ret = -EBUSY; | |
490 | ||
491 | if (ret < 0) { | |
f33ff110 | 492 | printk("register_blkdev: cannot get major %u for %s\n", |
1da177e4 LT |
493 | major, name); |
494 | kfree(p); | |
495 | } | |
496 | out: | |
edfaa7c3 | 497 | mutex_unlock(&block_class_lock); |
1da177e4 LT |
498 | return ret; |
499 | } | |
500 | ||
501 | EXPORT_SYMBOL(register_blkdev); | |
502 | ||
f4480240 | 503 | void unregister_blkdev(unsigned int major, const char *name) |
1da177e4 LT |
504 | { |
505 | struct blk_major_name **n; | |
506 | struct blk_major_name *p = NULL; | |
507 | int index = major_to_index(major); | |
1da177e4 | 508 | |
edfaa7c3 | 509 | mutex_lock(&block_class_lock); |
1da177e4 LT |
510 | for (n = &major_names[index]; *n; n = &(*n)->next) |
511 | if ((*n)->major == major) | |
512 | break; | |
294462a5 AM |
513 | if (!*n || strcmp((*n)->name, name)) { |
514 | WARN_ON(1); | |
294462a5 | 515 | } else { |
1da177e4 LT |
516 | p = *n; |
517 | *n = p->next; | |
518 | } | |
edfaa7c3 | 519 | mutex_unlock(&block_class_lock); |
1da177e4 | 520 | kfree(p); |
1da177e4 LT |
521 | } |
522 | ||
523 | EXPORT_SYMBOL(unregister_blkdev); | |
524 | ||
525 | static struct kobj_map *bdev_map; | |
526 | ||
870d6656 TH |
527 | /** |
528 | * blk_mangle_minor - scatter minor numbers apart | |
529 | * @minor: minor number to mangle | |
530 | * | |
531 | * Scatter consecutively allocated @minor number apart if MANGLE_DEVT | |
532 | * is enabled. Mangling twice gives the original value. | |
533 | * | |
534 | * RETURNS: | |
535 | * Mangled value. | |
536 | * | |
537 | * CONTEXT: | |
538 | * Don't care. | |
539 | */ | |
540 | static int blk_mangle_minor(int minor) | |
541 | { | |
542 | #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT | |
543 | int i; | |
544 | ||
545 | for (i = 0; i < MINORBITS / 2; i++) { | |
546 | int low = minor & (1 << i); | |
547 | int high = minor & (1 << (MINORBITS - 1 - i)); | |
548 | int distance = MINORBITS - 1 - 2 * i; | |
549 | ||
550 | minor ^= low | high; /* clear both bits */ | |
551 | low <<= distance; /* swap the positions */ | |
552 | high >>= distance; | |
553 | minor |= low | high; /* and set */ | |
554 | } | |
555 | #endif | |
556 | return minor; | |
557 | } | |
558 | ||
bcce3de1 TH |
559 | /** |
560 | * blk_alloc_devt - allocate a dev_t for a partition | |
561 | * @part: partition to allocate dev_t for | |
bcce3de1 TH |
562 | * @devt: out parameter for resulting dev_t |
563 | * | |
564 | * Allocate a dev_t for block device. | |
565 | * | |
566 | * RETURNS: | |
567 | * 0 on success, allocated dev_t is returned in *@devt. -errno on | |
568 | * failure. | |
569 | * | |
570 | * CONTEXT: | |
571 | * Might sleep. | |
572 | */ | |
573 | int blk_alloc_devt(struct hd_struct *part, dev_t *devt) | |
574 | { | |
575 | struct gendisk *disk = part_to_disk(part); | |
bab998d6 | 576 | int idx; |
bcce3de1 TH |
577 | |
578 | /* in consecutive minor range? */ | |
579 | if (part->partno < disk->minors) { | |
580 | *devt = MKDEV(disk->major, disk->first_minor + part->partno); | |
581 | return 0; | |
582 | } | |
583 | ||
584 | /* allocate ext devt */ | |
2da78092 KB |
585 | idr_preload(GFP_KERNEL); |
586 | ||
4d66e5e9 | 587 | spin_lock_bh(&ext_devt_lock); |
2da78092 | 588 | idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_NOWAIT); |
4d66e5e9 | 589 | spin_unlock_bh(&ext_devt_lock); |
2da78092 KB |
590 | |
591 | idr_preload_end(); | |
bab998d6 TH |
592 | if (idx < 0) |
593 | return idx == -ENOSPC ? -EBUSY : idx; | |
bcce3de1 | 594 | |
870d6656 | 595 | *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx)); |
bcce3de1 TH |
596 | return 0; |
597 | } | |
598 | ||
599 | /** | |
600 | * blk_free_devt - free a dev_t | |
601 | * @devt: dev_t to free | |
602 | * | |
603 | * Free @devt which was allocated using blk_alloc_devt(). | |
604 | * | |
605 | * CONTEXT: | |
606 | * Might sleep. | |
607 | */ | |
608 | void blk_free_devt(dev_t devt) | |
609 | { | |
bcce3de1 TH |
610 | if (devt == MKDEV(0, 0)) |
611 | return; | |
612 | ||
613 | if (MAJOR(devt) == BLOCK_EXT_MAJOR) { | |
4d66e5e9 | 614 | spin_lock_bh(&ext_devt_lock); |
870d6656 | 615 | idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); |
4d66e5e9 | 616 | spin_unlock_bh(&ext_devt_lock); |
bcce3de1 TH |
617 | } |
618 | } | |
619 | ||
33c826ef BVA |
620 | /* |
621 | * We invalidate devt by assigning NULL pointer for devt in idr. | |
6fcc44d1 YY |
622 | */ |
623 | void blk_invalidate_devt(dev_t devt) | |
624 | { | |
625 | if (MAJOR(devt) == BLOCK_EXT_MAJOR) { | |
626 | spin_lock_bh(&ext_devt_lock); | |
627 | idr_replace(&ext_devt_idr, NULL, blk_mangle_minor(MINOR(devt))); | |
628 | spin_unlock_bh(&ext_devt_lock); | |
629 | } | |
630 | } | |
631 | ||
1f014290 TH |
632 | static char *bdevt_str(dev_t devt, char *buf) |
633 | { | |
634 | if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) { | |
635 | char tbuf[BDEVT_SIZE]; | |
636 | snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt)); | |
637 | snprintf(buf, BDEVT_SIZE, "%-9s", tbuf); | |
638 | } else | |
639 | snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt)); | |
640 | ||
641 | return buf; | |
642 | } | |
643 | ||
1da177e4 LT |
644 | /* |
645 | * Register device numbers dev..(dev+range-1) | |
646 | * range must be nonzero | |
647 | * The hash chain is sorted on range, so that subranges can override. | |
648 | */ | |
edfaa7c3 | 649 | void blk_register_region(dev_t devt, unsigned long range, struct module *module, |
1da177e4 LT |
650 | struct kobject *(*probe)(dev_t, int *, void *), |
651 | int (*lock)(dev_t, void *), void *data) | |
652 | { | |
edfaa7c3 | 653 | kobj_map(bdev_map, devt, range, module, probe, lock, data); |
1da177e4 LT |
654 | } |
655 | ||
656 | EXPORT_SYMBOL(blk_register_region); | |
657 | ||
edfaa7c3 | 658 | void blk_unregister_region(dev_t devt, unsigned long range) |
1da177e4 | 659 | { |
edfaa7c3 | 660 | kobj_unmap(bdev_map, devt, range); |
1da177e4 LT |
661 | } |
662 | ||
663 | EXPORT_SYMBOL(blk_unregister_region); | |
664 | ||
cf771cb5 | 665 | static struct kobject *exact_match(dev_t devt, int *partno, void *data) |
1da177e4 LT |
666 | { |
667 | struct gendisk *p = data; | |
edfaa7c3 | 668 | |
ed9e1982 | 669 | return &disk_to_dev(p)->kobj; |
1da177e4 LT |
670 | } |
671 | ||
edfaa7c3 | 672 | static int exact_lock(dev_t devt, void *data) |
1da177e4 LT |
673 | { |
674 | struct gendisk *p = data; | |
675 | ||
3079c22e | 676 | if (!get_disk_and_module(p)) |
1da177e4 LT |
677 | return -1; |
678 | return 0; | |
679 | } | |
680 | ||
fef912bf HR |
681 | static void register_disk(struct device *parent, struct gendisk *disk, |
682 | const struct attribute_group **groups) | |
d2bf1b67 TH |
683 | { |
684 | struct device *ddev = disk_to_dev(disk); | |
685 | struct block_device *bdev; | |
686 | struct disk_part_iter piter; | |
687 | struct hd_struct *part; | |
688 | int err; | |
689 | ||
e63a46be | 690 | ddev->parent = parent; |
d2bf1b67 | 691 | |
ffc8b308 | 692 | dev_set_name(ddev, "%s", disk->disk_name); |
d2bf1b67 TH |
693 | |
694 | /* delay uevents, until we scanned partition table */ | |
695 | dev_set_uevent_suppress(ddev, 1); | |
696 | ||
fef912bf HR |
697 | if (groups) { |
698 | WARN_ON(ddev->groups); | |
699 | ddev->groups = groups; | |
700 | } | |
d2bf1b67 TH |
701 | if (device_add(ddev)) |
702 | return; | |
703 | if (!sysfs_deprecated) { | |
704 | err = sysfs_create_link(block_depr, &ddev->kobj, | |
705 | kobject_name(&ddev->kobj)); | |
706 | if (err) { | |
707 | device_del(ddev); | |
708 | return; | |
709 | } | |
710 | } | |
25e823c8 ML |
711 | |
712 | /* | |
713 | * avoid probable deadlock caused by allocating memory with | |
714 | * GFP_KERNEL in runtime_resume callback of its all ancestor | |
715 | * devices | |
716 | */ | |
717 | pm_runtime_set_memalloc_noio(ddev, true); | |
718 | ||
d2bf1b67 TH |
719 | disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj); |
720 | disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj); | |
721 | ||
8ddcd653 CH |
722 | if (disk->flags & GENHD_FL_HIDDEN) { |
723 | dev_set_uevent_suppress(ddev, 0); | |
724 | return; | |
725 | } | |
726 | ||
d2bf1b67 | 727 | /* No minors to use for partitions */ |
d27769ec | 728 | if (!disk_part_scan_enabled(disk)) |
d2bf1b67 TH |
729 | goto exit; |
730 | ||
731 | /* No such device (e.g., media were just removed) */ | |
732 | if (!get_capacity(disk)) | |
733 | goto exit; | |
734 | ||
735 | bdev = bdget_disk(disk, 0); | |
736 | if (!bdev) | |
737 | goto exit; | |
738 | ||
739 | bdev->bd_invalidated = 1; | |
740 | err = blkdev_get(bdev, FMODE_READ, NULL); | |
741 | if (err < 0) | |
742 | goto exit; | |
743 | blkdev_put(bdev, FMODE_READ); | |
744 | ||
745 | exit: | |
746 | /* announce disk after possible partitions are created */ | |
747 | dev_set_uevent_suppress(ddev, 0); | |
748 | kobject_uevent(&ddev->kobj, KOBJ_ADD); | |
749 | ||
750 | /* announce possible partitions */ | |
751 | disk_part_iter_init(&piter, disk, 0); | |
752 | while ((part = disk_part_iter_next(&piter))) | |
753 | kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD); | |
754 | disk_part_iter_exit(&piter); | |
8ddcd653 | 755 | |
4d7c1d3f | 756 | if (disk->queue->backing_dev_info->dev) { |
757 | err = sysfs_create_link(&ddev->kobj, | |
758 | &disk->queue->backing_dev_info->dev->kobj, | |
759 | "bdi"); | |
760 | WARN_ON(err); | |
761 | } | |
d2bf1b67 TH |
762 | } |
763 | ||
1da177e4 | 764 | /** |
fa70d2e2 | 765 | * __device_add_disk - add disk information to kernel list |
e63a46be | 766 | * @parent: parent device for the disk |
1da177e4 | 767 | * @disk: per-device partitioning information |
fef912bf | 768 | * @groups: Additional per-device sysfs groups |
fa70d2e2 | 769 | * @register_queue: register the queue if set to true |
1da177e4 LT |
770 | * |
771 | * This function registers the partitioning information in @disk | |
772 | * with the kernel. | |
3e1a7ff8 TH |
773 | * |
774 | * FIXME: error handling | |
1da177e4 | 775 | */ |
fa70d2e2 | 776 | static void __device_add_disk(struct device *parent, struct gendisk *disk, |
fef912bf | 777 | const struct attribute_group **groups, |
fa70d2e2 | 778 | bool register_queue) |
1da177e4 | 779 | { |
3e1a7ff8 | 780 | dev_t devt; |
6ffeea77 | 781 | int retval; |
cf0ca9fe | 782 | |
737eb78e DLM |
783 | /* |
784 | * The disk queue should now be all set with enough information about | |
785 | * the device for the elevator code to pick an adequate default | |
786 | * elevator if one is needed, that is, for devices requesting queue | |
787 | * registration. | |
788 | */ | |
789 | if (register_queue) | |
790 | elevator_init_mq(disk->queue); | |
791 | ||
3e1a7ff8 TH |
792 | /* minors == 0 indicates to use ext devt from part0 and should |
793 | * be accompanied with EXT_DEVT flag. Make sure all | |
794 | * parameters make sense. | |
795 | */ | |
796 | WARN_ON(disk->minors && !(disk->major || disk->first_minor)); | |
8ddcd653 CH |
797 | WARN_ON(!disk->minors && |
798 | !(disk->flags & (GENHD_FL_EXT_DEVT | GENHD_FL_HIDDEN))); | |
3e1a7ff8 | 799 | |
1da177e4 | 800 | disk->flags |= GENHD_FL_UP; |
3e1a7ff8 TH |
801 | |
802 | retval = blk_alloc_devt(&disk->part0, &devt); | |
803 | if (retval) { | |
804 | WARN_ON(1); | |
805 | return; | |
806 | } | |
3e1a7ff8 TH |
807 | disk->major = MAJOR(devt); |
808 | disk->first_minor = MINOR(devt); | |
809 | ||
9f53d2fe SG |
810 | disk_alloc_events(disk); |
811 | ||
8ddcd653 CH |
812 | if (disk->flags & GENHD_FL_HIDDEN) { |
813 | /* | |
814 | * Don't let hidden disks show up in /proc/partitions, | |
815 | * and don't bother scanning for partitions either. | |
816 | */ | |
817 | disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO; | |
818 | disk->flags |= GENHD_FL_NO_PART_SCAN; | |
819 | } else { | |
3c5d202b CH |
820 | struct backing_dev_info *bdi = disk->queue->backing_dev_info; |
821 | struct device *dev = disk_to_dev(disk); | |
3a92168b | 822 | int ret; |
823 | ||
8ddcd653 | 824 | /* Register BDI before referencing it from bdev */ |
3c5d202b CH |
825 | dev->devt = devt; |
826 | ret = bdi_register(bdi, "%u:%u", MAJOR(devt), MINOR(devt)); | |
3a92168b | 827 | WARN_ON(ret); |
3c5d202b | 828 | bdi_set_owner(bdi, dev); |
8ddcd653 CH |
829 | blk_register_region(disk_devt(disk), disk->minors, NULL, |
830 | exact_match, exact_lock, disk); | |
831 | } | |
fef912bf | 832 | register_disk(parent, disk, groups); |
fa70d2e2 MS |
833 | if (register_queue) |
834 | blk_register_queue(disk); | |
cf0ca9fe | 835 | |
523e1d39 TH |
836 | /* |
837 | * Take an extra ref on queue which will be put on disk_release() | |
838 | * so that it sticks around as long as @disk is there. | |
839 | */ | |
09ac46c4 | 840 | WARN_ON_ONCE(!blk_get_queue(disk->queue)); |
523e1d39 | 841 | |
77ea887e | 842 | disk_add_events(disk); |
25520d55 | 843 | blk_integrity_add(disk); |
1da177e4 | 844 | } |
fa70d2e2 | 845 | |
fef912bf HR |
846 | void device_add_disk(struct device *parent, struct gendisk *disk, |
847 | const struct attribute_group **groups) | |
848 | ||
fa70d2e2 | 849 | { |
fef912bf | 850 | __device_add_disk(parent, disk, groups, true); |
fa70d2e2 | 851 | } |
e63a46be | 852 | EXPORT_SYMBOL(device_add_disk); |
1da177e4 | 853 | |
fa70d2e2 MS |
854 | void device_add_disk_no_queue_reg(struct device *parent, struct gendisk *disk) |
855 | { | |
fef912bf | 856 | __device_add_disk(parent, disk, NULL, false); |
fa70d2e2 MS |
857 | } |
858 | EXPORT_SYMBOL(device_add_disk_no_queue_reg); | |
859 | ||
02d33b67 CH |
860 | static void invalidate_partition(struct gendisk *disk, int partno) |
861 | { | |
862 | struct block_device *bdev; | |
863 | ||
864 | bdev = bdget_disk(disk, partno); | |
865 | if (!bdev) | |
866 | return; | |
867 | ||
868 | fsync_bdev(bdev); | |
869 | __invalidate_device(bdev, true); | |
9bc5c397 CH |
870 | |
871 | /* | |
872 | * Unhash the bdev inode for this device so that it gets evicted as soon | |
873 | * as last inode reference is dropped. | |
874 | */ | |
875 | remove_inode_hash(bdev->bd_inode); | |
02d33b67 CH |
876 | bdput(bdev); |
877 | } | |
878 | ||
d2bf1b67 | 879 | void del_gendisk(struct gendisk *disk) |
1da177e4 | 880 | { |
d2bf1b67 TH |
881 | struct disk_part_iter piter; |
882 | struct hd_struct *part; | |
883 | ||
25520d55 | 884 | blk_integrity_del(disk); |
77ea887e TH |
885 | disk_del_events(disk); |
886 | ||
56c0908c JK |
887 | /* |
888 | * Block lookups of the disk until all bdevs are unhashed and the | |
889 | * disk is marked as dead (GENHD_FL_UP cleared). | |
890 | */ | |
891 | down_write(&disk->lookup_sem); | |
d2bf1b67 TH |
892 | /* invalidate stuff */ |
893 | disk_part_iter_init(&piter, disk, | |
894 | DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE); | |
895 | while ((part = disk_part_iter_next(&piter))) { | |
896 | invalidate_partition(disk, part->partno); | |
cddae808 | 897 | delete_partition(disk, part); |
d2bf1b67 TH |
898 | } |
899 | disk_part_iter_exit(&piter); | |
900 | ||
901 | invalidate_partition(disk, 0); | |
d2bf1b67 TH |
902 | set_capacity(disk, 0); |
903 | disk->flags &= ~GENHD_FL_UP; | |
56c0908c | 904 | up_write(&disk->lookup_sem); |
d2bf1b67 | 905 | |
8ddcd653 CH |
906 | if (!(disk->flags & GENHD_FL_HIDDEN)) |
907 | sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi"); | |
90f16fdd JK |
908 | if (disk->queue) { |
909 | /* | |
910 | * Unregister bdi before releasing device numbers (as they can | |
911 | * get reused and we'd get clashes in sysfs). | |
912 | */ | |
bc8d062c MS |
913 | if (!(disk->flags & GENHD_FL_HIDDEN)) |
914 | bdi_unregister(disk->queue->backing_dev_info); | |
90f16fdd JK |
915 | blk_unregister_queue(disk); |
916 | } else { | |
917 | WARN_ON(1); | |
918 | } | |
d2bf1b67 | 919 | |
17eac099 | 920 | if (!(disk->flags & GENHD_FL_HIDDEN)) |
8ddcd653 | 921 | blk_unregister_region(disk_devt(disk), disk->minors); |
6fcc44d1 YY |
922 | /* |
923 | * Remove gendisk pointer from idr so that it cannot be looked up | |
924 | * while RCU period before freeing gendisk is running to prevent | |
925 | * use-after-free issues. Note that the device number stays | |
926 | * "in-use" until we really free the gendisk. | |
927 | */ | |
928 | blk_invalidate_devt(disk_devt(disk)); | |
d2bf1b67 TH |
929 | |
930 | kobject_put(disk->part0.holder_dir); | |
931 | kobject_put(disk->slave_dir); | |
d2bf1b67 TH |
932 | |
933 | part_stat_set_all(&disk->part0, 0); | |
934 | disk->part0.stamp = 0; | |
d2bf1b67 TH |
935 | if (!sysfs_deprecated) |
936 | sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk))); | |
25e823c8 | 937 | pm_runtime_set_memalloc_noio(disk_to_dev(disk), false); |
d2bf1b67 | 938 | device_del(disk_to_dev(disk)); |
1da177e4 | 939 | } |
d2bf1b67 | 940 | EXPORT_SYMBOL(del_gendisk); |
1da177e4 | 941 | |
99e6608c VV |
942 | /* sysfs access to bad-blocks list. */ |
943 | static ssize_t disk_badblocks_show(struct device *dev, | |
944 | struct device_attribute *attr, | |
945 | char *page) | |
946 | { | |
947 | struct gendisk *disk = dev_to_disk(dev); | |
948 | ||
949 | if (!disk->bb) | |
950 | return sprintf(page, "\n"); | |
951 | ||
952 | return badblocks_show(disk->bb, page, 0); | |
953 | } | |
954 | ||
955 | static ssize_t disk_badblocks_store(struct device *dev, | |
956 | struct device_attribute *attr, | |
957 | const char *page, size_t len) | |
958 | { | |
959 | struct gendisk *disk = dev_to_disk(dev); | |
960 | ||
961 | if (!disk->bb) | |
962 | return -ENXIO; | |
963 | ||
964 | return badblocks_store(disk->bb, page, len, 0); | |
965 | } | |
966 | ||
1da177e4 LT |
967 | /** |
968 | * get_gendisk - get partitioning information for a given device | |
710027a4 | 969 | * @devt: device to get partitioning information for |
496aa8a9 | 970 | * @partno: returned partition index |
1da177e4 LT |
971 | * |
972 | * This function gets the structure containing partitioning | |
710027a4 | 973 | * information for the given device @devt. |
1da177e4 | 974 | */ |
cf771cb5 | 975 | struct gendisk *get_gendisk(dev_t devt, int *partno) |
1da177e4 | 976 | { |
bcce3de1 TH |
977 | struct gendisk *disk = NULL; |
978 | ||
979 | if (MAJOR(devt) != BLOCK_EXT_MAJOR) { | |
980 | struct kobject *kobj; | |
981 | ||
982 | kobj = kobj_lookup(bdev_map, devt, partno); | |
983 | if (kobj) | |
984 | disk = dev_to_disk(kobj_to_dev(kobj)); | |
985 | } else { | |
986 | struct hd_struct *part; | |
987 | ||
4d66e5e9 | 988 | spin_lock_bh(&ext_devt_lock); |
870d6656 | 989 | part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); |
3079c22e | 990 | if (part && get_disk_and_module(part_to_disk(part))) { |
bcce3de1 TH |
991 | *partno = part->partno; |
992 | disk = part_to_disk(part); | |
993 | } | |
4d66e5e9 | 994 | spin_unlock_bh(&ext_devt_lock); |
bcce3de1 | 995 | } |
edfaa7c3 | 996 | |
56c0908c JK |
997 | if (!disk) |
998 | return NULL; | |
999 | ||
1000 | /* | |
1001 | * Synchronize with del_gendisk() to not return disk that is being | |
1002 | * destroyed. | |
1003 | */ | |
1004 | down_read(&disk->lookup_sem); | |
1005 | if (unlikely((disk->flags & GENHD_FL_HIDDEN) || | |
1006 | !(disk->flags & GENHD_FL_UP))) { | |
1007 | up_read(&disk->lookup_sem); | |
9df6c299 | 1008 | put_disk_and_module(disk); |
8ddcd653 | 1009 | disk = NULL; |
56c0908c JK |
1010 | } else { |
1011 | up_read(&disk->lookup_sem); | |
8ddcd653 | 1012 | } |
bcce3de1 | 1013 | return disk; |
1da177e4 LT |
1014 | } |
1015 | ||
f331c029 TH |
1016 | /** |
1017 | * bdget_disk - do bdget() by gendisk and partition number | |
1018 | * @disk: gendisk of interest | |
1019 | * @partno: partition number | |
1020 | * | |
1021 | * Find partition @partno from @disk, do bdget() on it. | |
1022 | * | |
1023 | * CONTEXT: | |
1024 | * Don't care. | |
1025 | * | |
1026 | * RETURNS: | |
1027 | * Resulting block_device on success, NULL on failure. | |
1028 | */ | |
aeb3d3a8 | 1029 | struct block_device *bdget_disk(struct gendisk *disk, int partno) |
f331c029 | 1030 | { |
548b10eb TH |
1031 | struct hd_struct *part; |
1032 | struct block_device *bdev = NULL; | |
f331c029 | 1033 | |
548b10eb | 1034 | part = disk_get_part(disk, partno); |
2bbedcb4 | 1035 | if (part) |
548b10eb TH |
1036 | bdev = bdget(part_devt(part)); |
1037 | disk_put_part(part); | |
f331c029 | 1038 | |
548b10eb | 1039 | return bdev; |
f331c029 TH |
1040 | } |
1041 | EXPORT_SYMBOL(bdget_disk); | |
1042 | ||
5c6f35c5 GKH |
1043 | /* |
1044 | * print a full list of all partitions - intended for places where the root | |
1045 | * filesystem can't be mounted and thus to give the victim some idea of what | |
1046 | * went wrong | |
1047 | */ | |
1048 | void __init printk_all_partitions(void) | |
1049 | { | |
def4e38d TH |
1050 | struct class_dev_iter iter; |
1051 | struct device *dev; | |
1052 | ||
1053 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); | |
1054 | while ((dev = class_dev_iter_next(&iter))) { | |
1055 | struct gendisk *disk = dev_to_disk(dev); | |
e71bf0d0 TH |
1056 | struct disk_part_iter piter; |
1057 | struct hd_struct *part; | |
1f014290 TH |
1058 | char name_buf[BDEVNAME_SIZE]; |
1059 | char devt_buf[BDEVT_SIZE]; | |
def4e38d TH |
1060 | |
1061 | /* | |
1062 | * Don't show empty devices or things that have been | |
25985edc | 1063 | * suppressed |
def4e38d TH |
1064 | */ |
1065 | if (get_capacity(disk) == 0 || | |
1066 | (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)) | |
1067 | continue; | |
1068 | ||
1069 | /* | |
1070 | * Note, unlike /proc/partitions, I am showing the | |
1071 | * numbers in hex - the same format as the root= | |
1072 | * option takes. | |
1073 | */ | |
074a7aca TH |
1074 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0); |
1075 | while ((part = disk_part_iter_next(&piter))) { | |
1076 | bool is_part0 = part == &disk->part0; | |
def4e38d | 1077 | |
b5af921e | 1078 | printk("%s%s %10llu %s %s", is_part0 ? "" : " ", |
1f014290 | 1079 | bdevt_str(part_devt(part), devt_buf), |
c83f6bf9 VG |
1080 | (unsigned long long)part_nr_sects_read(part) >> 1 |
1081 | , disk_name(disk, part->partno, name_buf), | |
1ad7e899 | 1082 | part->info ? part->info->uuid : ""); |
074a7aca | 1083 | if (is_part0) { |
52c44d93 | 1084 | if (dev->parent && dev->parent->driver) |
074a7aca | 1085 | printk(" driver: %s\n", |
52c44d93 | 1086 | dev->parent->driver->name); |
074a7aca TH |
1087 | else |
1088 | printk(" (driver?)\n"); | |
1089 | } else | |
1090 | printk("\n"); | |
1091 | } | |
e71bf0d0 | 1092 | disk_part_iter_exit(&piter); |
def4e38d TH |
1093 | } |
1094 | class_dev_iter_exit(&iter); | |
dd2a345f DG |
1095 | } |
1096 | ||
1da177e4 LT |
1097 | #ifdef CONFIG_PROC_FS |
1098 | /* iterator */ | |
def4e38d | 1099 | static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos) |
68c4d4a7 | 1100 | { |
def4e38d TH |
1101 | loff_t skip = *pos; |
1102 | struct class_dev_iter *iter; | |
1103 | struct device *dev; | |
68c4d4a7 | 1104 | |
aeb3d3a8 | 1105 | iter = kmalloc(sizeof(*iter), GFP_KERNEL); |
def4e38d TH |
1106 | if (!iter) |
1107 | return ERR_PTR(-ENOMEM); | |
1108 | ||
1109 | seqf->private = iter; | |
1110 | class_dev_iter_init(iter, &block_class, NULL, &disk_type); | |
1111 | do { | |
1112 | dev = class_dev_iter_next(iter); | |
1113 | if (!dev) | |
1114 | return NULL; | |
1115 | } while (skip--); | |
1116 | ||
1117 | return dev_to_disk(dev); | |
68c4d4a7 GKH |
1118 | } |
1119 | ||
def4e38d | 1120 | static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos) |
1da177e4 | 1121 | { |
edfaa7c3 | 1122 | struct device *dev; |
1da177e4 | 1123 | |
def4e38d TH |
1124 | (*pos)++; |
1125 | dev = class_dev_iter_next(seqf->private); | |
2ac3cee5 | 1126 | if (dev) |
68c4d4a7 | 1127 | return dev_to_disk(dev); |
2ac3cee5 | 1128 | |
1da177e4 LT |
1129 | return NULL; |
1130 | } | |
1131 | ||
def4e38d | 1132 | static void disk_seqf_stop(struct seq_file *seqf, void *v) |
27f30251 | 1133 | { |
def4e38d | 1134 | struct class_dev_iter *iter = seqf->private; |
27f30251 | 1135 | |
def4e38d TH |
1136 | /* stop is called even after start failed :-( */ |
1137 | if (iter) { | |
1138 | class_dev_iter_exit(iter); | |
1139 | kfree(iter); | |
77da1605 | 1140 | seqf->private = NULL; |
5c0ef6d0 | 1141 | } |
1da177e4 LT |
1142 | } |
1143 | ||
def4e38d | 1144 | static void *show_partition_start(struct seq_file *seqf, loff_t *pos) |
1da177e4 | 1145 | { |
06768067 | 1146 | void *p; |
def4e38d TH |
1147 | |
1148 | p = disk_seqf_start(seqf, pos); | |
b9f985b6 | 1149 | if (!IS_ERR_OR_NULL(p) && !*pos) |
def4e38d TH |
1150 | seq_puts(seqf, "major minor #blocks name\n\n"); |
1151 | return p; | |
1da177e4 LT |
1152 | } |
1153 | ||
cf771cb5 | 1154 | static int show_partition(struct seq_file *seqf, void *v) |
1da177e4 LT |
1155 | { |
1156 | struct gendisk *sgp = v; | |
e71bf0d0 TH |
1157 | struct disk_part_iter piter; |
1158 | struct hd_struct *part; | |
1da177e4 LT |
1159 | char buf[BDEVNAME_SIZE]; |
1160 | ||
1da177e4 | 1161 | /* Don't show non-partitionable removeable devices or empty devices */ |
d27769ec | 1162 | if (!get_capacity(sgp) || (!disk_max_parts(sgp) && |
f331c029 | 1163 | (sgp->flags & GENHD_FL_REMOVABLE))) |
1da177e4 LT |
1164 | return 0; |
1165 | if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO) | |
1166 | return 0; | |
1167 | ||
1168 | /* show the full disk and all non-0 size partitions of it */ | |
074a7aca | 1169 | disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0); |
e71bf0d0 | 1170 | while ((part = disk_part_iter_next(&piter))) |
1f014290 | 1171 | seq_printf(seqf, "%4d %7d %10llu %s\n", |
f331c029 | 1172 | MAJOR(part_devt(part)), MINOR(part_devt(part)), |
c83f6bf9 | 1173 | (unsigned long long)part_nr_sects_read(part) >> 1, |
f331c029 | 1174 | disk_name(sgp, part->partno, buf)); |
e71bf0d0 | 1175 | disk_part_iter_exit(&piter); |
1da177e4 LT |
1176 | |
1177 | return 0; | |
1178 | } | |
1179 | ||
f500975a | 1180 | static const struct seq_operations partitions_op = { |
def4e38d TH |
1181 | .start = show_partition_start, |
1182 | .next = disk_seqf_next, | |
1183 | .stop = disk_seqf_stop, | |
edfaa7c3 | 1184 | .show = show_partition |
1da177e4 LT |
1185 | }; |
1186 | #endif | |
1187 | ||
1188 | ||
cf771cb5 | 1189 | static struct kobject *base_probe(dev_t devt, int *partno, void *data) |
1da177e4 | 1190 | { |
edfaa7c3 | 1191 | if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0) |
1da177e4 | 1192 | /* Make old-style 2.4 aliases work */ |
edfaa7c3 | 1193 | request_module("block-major-%d", MAJOR(devt)); |
1da177e4 LT |
1194 | return NULL; |
1195 | } | |
1196 | ||
1197 | static int __init genhd_device_init(void) | |
1198 | { | |
e105b8bf DW |
1199 | int error; |
1200 | ||
1201 | block_class.dev_kobj = sysfs_dev_block_kobj; | |
1202 | error = class_register(&block_class); | |
ee27a558 RM |
1203 | if (unlikely(error)) |
1204 | return error; | |
edfaa7c3 | 1205 | bdev_map = kobj_map_init(base_probe, &block_class_lock); |
1da177e4 | 1206 | blk_dev_init(); |
edfaa7c3 | 1207 | |
561ec68e ZY |
1208 | register_blkdev(BLOCK_EXT_MAJOR, "blkext"); |
1209 | ||
edfaa7c3 | 1210 | /* create top-level block dir */ |
e52eec13 AK |
1211 | if (!sysfs_deprecated) |
1212 | block_depr = kobject_create_and_add("block", NULL); | |
830d3cfb | 1213 | return 0; |
1da177e4 LT |
1214 | } |
1215 | ||
1216 | subsys_initcall(genhd_device_init); | |
1217 | ||
edfaa7c3 KS |
1218 | static ssize_t disk_range_show(struct device *dev, |
1219 | struct device_attribute *attr, char *buf) | |
1da177e4 | 1220 | { |
edfaa7c3 | 1221 | struct gendisk *disk = dev_to_disk(dev); |
1da177e4 | 1222 | |
edfaa7c3 | 1223 | return sprintf(buf, "%d\n", disk->minors); |
1da177e4 LT |
1224 | } |
1225 | ||
1f014290 TH |
1226 | static ssize_t disk_ext_range_show(struct device *dev, |
1227 | struct device_attribute *attr, char *buf) | |
1228 | { | |
1229 | struct gendisk *disk = dev_to_disk(dev); | |
1230 | ||
b5d0b9df | 1231 | return sprintf(buf, "%d\n", disk_max_parts(disk)); |
1f014290 TH |
1232 | } |
1233 | ||
edfaa7c3 KS |
1234 | static ssize_t disk_removable_show(struct device *dev, |
1235 | struct device_attribute *attr, char *buf) | |
a7fd6706 | 1236 | { |
edfaa7c3 | 1237 | struct gendisk *disk = dev_to_disk(dev); |
a7fd6706 | 1238 | |
edfaa7c3 KS |
1239 | return sprintf(buf, "%d\n", |
1240 | (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0)); | |
a7fd6706 KS |
1241 | } |
1242 | ||
8ddcd653 CH |
1243 | static ssize_t disk_hidden_show(struct device *dev, |
1244 | struct device_attribute *attr, char *buf) | |
1245 | { | |
1246 | struct gendisk *disk = dev_to_disk(dev); | |
1247 | ||
1248 | return sprintf(buf, "%d\n", | |
1249 | (disk->flags & GENHD_FL_HIDDEN ? 1 : 0)); | |
1250 | } | |
1251 | ||
1c9ce527 KS |
1252 | static ssize_t disk_ro_show(struct device *dev, |
1253 | struct device_attribute *attr, char *buf) | |
1254 | { | |
1255 | struct gendisk *disk = dev_to_disk(dev); | |
1256 | ||
b7db9956 | 1257 | return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0); |
1c9ce527 KS |
1258 | } |
1259 | ||
3ad5cee5 CH |
1260 | ssize_t part_size_show(struct device *dev, |
1261 | struct device_attribute *attr, char *buf) | |
1262 | { | |
1263 | struct hd_struct *p = dev_to_part(dev); | |
1264 | ||
1265 | return sprintf(buf, "%llu\n", | |
1266 | (unsigned long long)part_nr_sects_read(p)); | |
1267 | } | |
1268 | ||
1269 | ssize_t part_stat_show(struct device *dev, | |
1270 | struct device_attribute *attr, char *buf) | |
1271 | { | |
1272 | struct hd_struct *p = dev_to_part(dev); | |
1273 | struct request_queue *q = part_to_disk(p)->queue; | |
ea18e0f0 | 1274 | struct disk_stats stat; |
3ad5cee5 CH |
1275 | unsigned int inflight; |
1276 | ||
ea18e0f0 | 1277 | part_stat_read_all(p, &stat); |
b2f609e1 CH |
1278 | if (queue_is_mq(q)) |
1279 | inflight = blk_mq_in_flight(q, p); | |
1280 | else | |
1281 | inflight = part_in_flight(q, p); | |
ea18e0f0 | 1282 | |
3ad5cee5 CH |
1283 | return sprintf(buf, |
1284 | "%8lu %8lu %8llu %8u " | |
1285 | "%8lu %8lu %8llu %8u " | |
1286 | "%8u %8u %8u " | |
1287 | "%8lu %8lu %8llu %8u " | |
1288 | "%8lu %8u" | |
1289 | "\n", | |
ea18e0f0 KK |
1290 | stat.ios[STAT_READ], |
1291 | stat.merges[STAT_READ], | |
1292 | (unsigned long long)stat.sectors[STAT_READ], | |
1293 | (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC), | |
1294 | stat.ios[STAT_WRITE], | |
1295 | stat.merges[STAT_WRITE], | |
1296 | (unsigned long long)stat.sectors[STAT_WRITE], | |
1297 | (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC), | |
3ad5cee5 | 1298 | inflight, |
ea18e0f0 | 1299 | jiffies_to_msecs(stat.io_ticks), |
8cd5b8fc KK |
1300 | (unsigned int)div_u64(stat.nsecs[STAT_READ] + |
1301 | stat.nsecs[STAT_WRITE] + | |
1302 | stat.nsecs[STAT_DISCARD] + | |
1303 | stat.nsecs[STAT_FLUSH], | |
1304 | NSEC_PER_MSEC), | |
ea18e0f0 KK |
1305 | stat.ios[STAT_DISCARD], |
1306 | stat.merges[STAT_DISCARD], | |
1307 | (unsigned long long)stat.sectors[STAT_DISCARD], | |
1308 | (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC), | |
1309 | stat.ios[STAT_FLUSH], | |
1310 | (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC)); | |
3ad5cee5 CH |
1311 | } |
1312 | ||
1313 | ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr, | |
1314 | char *buf) | |
1315 | { | |
1316 | struct hd_struct *p = dev_to_part(dev); | |
1317 | struct request_queue *q = part_to_disk(p)->queue; | |
1318 | unsigned int inflight[2]; | |
1319 | ||
b2f609e1 CH |
1320 | if (queue_is_mq(q)) |
1321 | blk_mq_in_flight_rw(q, p, inflight); | |
1322 | else | |
1323 | part_in_flight_rw(q, p, inflight); | |
1324 | ||
3ad5cee5 CH |
1325 | return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]); |
1326 | } | |
1327 | ||
edfaa7c3 KS |
1328 | static ssize_t disk_capability_show(struct device *dev, |
1329 | struct device_attribute *attr, char *buf) | |
86ce18d7 | 1330 | { |
edfaa7c3 KS |
1331 | struct gendisk *disk = dev_to_disk(dev); |
1332 | ||
1333 | return sprintf(buf, "%x\n", disk->flags); | |
86ce18d7 | 1334 | } |
edfaa7c3 | 1335 | |
c72758f3 MP |
1336 | static ssize_t disk_alignment_offset_show(struct device *dev, |
1337 | struct device_attribute *attr, | |
1338 | char *buf) | |
1339 | { | |
1340 | struct gendisk *disk = dev_to_disk(dev); | |
1341 | ||
1342 | return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue)); | |
1343 | } | |
1344 | ||
86b37281 MP |
1345 | static ssize_t disk_discard_alignment_show(struct device *dev, |
1346 | struct device_attribute *attr, | |
1347 | char *buf) | |
1348 | { | |
1349 | struct gendisk *disk = dev_to_disk(dev); | |
1350 | ||
dd3d145d | 1351 | return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue)); |
86b37281 MP |
1352 | } |
1353 | ||
5657a819 JP |
1354 | static DEVICE_ATTR(range, 0444, disk_range_show, NULL); |
1355 | static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL); | |
1356 | static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL); | |
1357 | static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL); | |
1358 | static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL); | |
1359 | static DEVICE_ATTR(size, 0444, part_size_show, NULL); | |
1360 | static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL); | |
1361 | static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL); | |
1362 | static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL); | |
1363 | static DEVICE_ATTR(stat, 0444, part_stat_show, NULL); | |
1364 | static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL); | |
1365 | static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store); | |
3ad5cee5 | 1366 | |
c17bb495 | 1367 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
3ad5cee5 CH |
1368 | ssize_t part_fail_show(struct device *dev, |
1369 | struct device_attribute *attr, char *buf) | |
1370 | { | |
1371 | struct hd_struct *p = dev_to_part(dev); | |
1372 | ||
1373 | return sprintf(buf, "%d\n", p->make_it_fail); | |
1374 | } | |
1375 | ||
1376 | ssize_t part_fail_store(struct device *dev, | |
1377 | struct device_attribute *attr, | |
1378 | const char *buf, size_t count) | |
1379 | { | |
1380 | struct hd_struct *p = dev_to_part(dev); | |
1381 | int i; | |
1382 | ||
1383 | if (count > 0 && sscanf(buf, "%d", &i) > 0) | |
1384 | p->make_it_fail = (i == 0) ? 0 : 1; | |
1385 | ||
1386 | return count; | |
1387 | } | |
1388 | ||
edfaa7c3 | 1389 | static struct device_attribute dev_attr_fail = |
5657a819 | 1390 | __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store); |
3ad5cee5 CH |
1391 | #endif /* CONFIG_FAIL_MAKE_REQUEST */ |
1392 | ||
581d4e28 JA |
1393 | #ifdef CONFIG_FAIL_IO_TIMEOUT |
1394 | static struct device_attribute dev_attr_fail_timeout = | |
5657a819 | 1395 | __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store); |
581d4e28 | 1396 | #endif |
edfaa7c3 KS |
1397 | |
1398 | static struct attribute *disk_attrs[] = { | |
1399 | &dev_attr_range.attr, | |
1f014290 | 1400 | &dev_attr_ext_range.attr, |
edfaa7c3 | 1401 | &dev_attr_removable.attr, |
8ddcd653 | 1402 | &dev_attr_hidden.attr, |
1c9ce527 | 1403 | &dev_attr_ro.attr, |
edfaa7c3 | 1404 | &dev_attr_size.attr, |
c72758f3 | 1405 | &dev_attr_alignment_offset.attr, |
86b37281 | 1406 | &dev_attr_discard_alignment.attr, |
edfaa7c3 KS |
1407 | &dev_attr_capability.attr, |
1408 | &dev_attr_stat.attr, | |
316d315b | 1409 | &dev_attr_inflight.attr, |
99e6608c | 1410 | &dev_attr_badblocks.attr, |
edfaa7c3 KS |
1411 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
1412 | &dev_attr_fail.attr, | |
581d4e28 JA |
1413 | #endif |
1414 | #ifdef CONFIG_FAIL_IO_TIMEOUT | |
1415 | &dev_attr_fail_timeout.attr, | |
edfaa7c3 KS |
1416 | #endif |
1417 | NULL | |
1418 | }; | |
1419 | ||
9438b3e0 DW |
1420 | static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n) |
1421 | { | |
1422 | struct device *dev = container_of(kobj, typeof(*dev), kobj); | |
1423 | struct gendisk *disk = dev_to_disk(dev); | |
1424 | ||
1425 | if (a == &dev_attr_badblocks.attr && !disk->bb) | |
1426 | return 0; | |
1427 | return a->mode; | |
1428 | } | |
1429 | ||
edfaa7c3 KS |
1430 | static struct attribute_group disk_attr_group = { |
1431 | .attrs = disk_attrs, | |
9438b3e0 | 1432 | .is_visible = disk_visible, |
edfaa7c3 KS |
1433 | }; |
1434 | ||
a4dbd674 | 1435 | static const struct attribute_group *disk_attr_groups[] = { |
edfaa7c3 KS |
1436 | &disk_attr_group, |
1437 | NULL | |
1da177e4 LT |
1438 | }; |
1439 | ||
540eed56 TH |
1440 | /** |
1441 | * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way | |
1442 | * @disk: disk to replace part_tbl for | |
1443 | * @new_ptbl: new part_tbl to install | |
1444 | * | |
1445 | * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The | |
1446 | * original ptbl is freed using RCU callback. | |
1447 | * | |
1448 | * LOCKING: | |
6d2cf6f2 | 1449 | * Matching bd_mutex locked or the caller is the only user of @disk. |
540eed56 TH |
1450 | */ |
1451 | static void disk_replace_part_tbl(struct gendisk *disk, | |
1452 | struct disk_part_tbl *new_ptbl) | |
1453 | { | |
6d2cf6f2 BVA |
1454 | struct disk_part_tbl *old_ptbl = |
1455 | rcu_dereference_protected(disk->part_tbl, 1); | |
540eed56 TH |
1456 | |
1457 | rcu_assign_pointer(disk->part_tbl, new_ptbl); | |
a6f23657 JA |
1458 | |
1459 | if (old_ptbl) { | |
1460 | rcu_assign_pointer(old_ptbl->last_lookup, NULL); | |
57bdfbf9 | 1461 | kfree_rcu(old_ptbl, rcu_head); |
a6f23657 | 1462 | } |
540eed56 TH |
1463 | } |
1464 | ||
1465 | /** | |
1466 | * disk_expand_part_tbl - expand disk->part_tbl | |
1467 | * @disk: disk to expand part_tbl for | |
1468 | * @partno: expand such that this partno can fit in | |
1469 | * | |
1470 | * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl | |
1471 | * uses RCU to allow unlocked dereferencing for stats and other stuff. | |
1472 | * | |
1473 | * LOCKING: | |
6d2cf6f2 BVA |
1474 | * Matching bd_mutex locked or the caller is the only user of @disk. |
1475 | * Might sleep. | |
540eed56 TH |
1476 | * |
1477 | * RETURNS: | |
1478 | * 0 on success, -errno on failure. | |
1479 | */ | |
1480 | int disk_expand_part_tbl(struct gendisk *disk, int partno) | |
1481 | { | |
6d2cf6f2 BVA |
1482 | struct disk_part_tbl *old_ptbl = |
1483 | rcu_dereference_protected(disk->part_tbl, 1); | |
540eed56 TH |
1484 | struct disk_part_tbl *new_ptbl; |
1485 | int len = old_ptbl ? old_ptbl->len : 0; | |
5fabcb4c | 1486 | int i, target; |
5fabcb4c JA |
1487 | |
1488 | /* | |
1489 | * check for int overflow, since we can get here from blkpg_ioctl() | |
1490 | * with a user passed 'partno'. | |
1491 | */ | |
1492 | target = partno + 1; | |
1493 | if (target < 0) | |
1494 | return -EINVAL; | |
540eed56 TH |
1495 | |
1496 | /* disk_max_parts() is zero during initialization, ignore if so */ | |
1497 | if (disk_max_parts(disk) && target > disk_max_parts(disk)) | |
1498 | return -EINVAL; | |
1499 | ||
1500 | if (target <= len) | |
1501 | return 0; | |
1502 | ||
78b90a2c GS |
1503 | new_ptbl = kzalloc_node(struct_size(new_ptbl, part, target), GFP_KERNEL, |
1504 | disk->node_id); | |
540eed56 TH |
1505 | if (!new_ptbl) |
1506 | return -ENOMEM; | |
1507 | ||
540eed56 TH |
1508 | new_ptbl->len = target; |
1509 | ||
1510 | for (i = 0; i < len; i++) | |
1511 | rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]); | |
1512 | ||
1513 | disk_replace_part_tbl(disk, new_ptbl); | |
1514 | return 0; | |
1515 | } | |
1516 | ||
edfaa7c3 | 1517 | static void disk_release(struct device *dev) |
1da177e4 | 1518 | { |
edfaa7c3 KS |
1519 | struct gendisk *disk = dev_to_disk(dev); |
1520 | ||
2da78092 | 1521 | blk_free_devt(dev->devt); |
77ea887e | 1522 | disk_release_events(disk); |
1da177e4 | 1523 | kfree(disk->random); |
540eed56 | 1524 | disk_replace_part_tbl(disk, NULL); |
b54e5ed8 | 1525 | hd_free_part(&disk->part0); |
523e1d39 TH |
1526 | if (disk->queue) |
1527 | blk_put_queue(disk->queue); | |
1da177e4 LT |
1528 | kfree(disk); |
1529 | } | |
edfaa7c3 KS |
1530 | struct class block_class = { |
1531 | .name = "block", | |
1da177e4 LT |
1532 | }; |
1533 | ||
3c2670e6 | 1534 | static char *block_devnode(struct device *dev, umode_t *mode, |
4e4098a3 | 1535 | kuid_t *uid, kgid_t *gid) |
b03f38b6 KS |
1536 | { |
1537 | struct gendisk *disk = dev_to_disk(dev); | |
1538 | ||
348e114b CH |
1539 | if (disk->fops->devnode) |
1540 | return disk->fops->devnode(disk, mode); | |
b03f38b6 KS |
1541 | return NULL; |
1542 | } | |
1543 | ||
edf8ff55 | 1544 | static const struct device_type disk_type = { |
edfaa7c3 KS |
1545 | .name = "disk", |
1546 | .groups = disk_attr_groups, | |
1547 | .release = disk_release, | |
e454cea2 | 1548 | .devnode = block_devnode, |
1da177e4 LT |
1549 | }; |
1550 | ||
a6e2ba88 | 1551 | #ifdef CONFIG_PROC_FS |
cf771cb5 TH |
1552 | /* |
1553 | * aggregate disk stat collector. Uses the same stats that the sysfs | |
1554 | * entries do, above, but makes them available through one seq_file. | |
1555 | * | |
1556 | * The output looks suspiciously like /proc/partitions with a bunch of | |
1557 | * extra fields. | |
1558 | */ | |
1559 | static int diskstats_show(struct seq_file *seqf, void *v) | |
1da177e4 LT |
1560 | { |
1561 | struct gendisk *gp = v; | |
e71bf0d0 TH |
1562 | struct disk_part_iter piter; |
1563 | struct hd_struct *hd; | |
1da177e4 | 1564 | char buf[BDEVNAME_SIZE]; |
e016b782 | 1565 | unsigned int inflight; |
ea18e0f0 | 1566 | struct disk_stats stat; |
1da177e4 LT |
1567 | |
1568 | /* | |
ed9e1982 | 1569 | if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next) |
cf771cb5 | 1570 | seq_puts(seqf, "major minor name" |
1da177e4 LT |
1571 | " rio rmerge rsect ruse wio wmerge " |
1572 | "wsect wuse running use aveq" | |
1573 | "\n\n"); | |
1574 | */ | |
9f5e4865 | 1575 | |
71982a40 | 1576 | disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0); |
e71bf0d0 | 1577 | while ((hd = disk_part_iter_next(&piter))) { |
ea18e0f0 | 1578 | part_stat_read_all(hd, &stat); |
b2f609e1 CH |
1579 | if (queue_is_mq(gp->queue)) |
1580 | inflight = blk_mq_in_flight(gp->queue, hd); | |
1581 | else | |
1582 | inflight = part_in_flight(gp->queue, hd); | |
ea18e0f0 | 1583 | |
bdca3c87 MC |
1584 | seq_printf(seqf, "%4d %7d %s " |
1585 | "%lu %lu %lu %u " | |
1586 | "%lu %lu %lu %u " | |
1587 | "%u %u %u " | |
b6866318 KK |
1588 | "%lu %lu %lu %u " |
1589 | "%lu %u" | |
1590 | "\n", | |
f331c029 TH |
1591 | MAJOR(part_devt(hd)), MINOR(part_devt(hd)), |
1592 | disk_name(gp, hd->partno, buf), | |
ea18e0f0 KK |
1593 | stat.ios[STAT_READ], |
1594 | stat.merges[STAT_READ], | |
1595 | stat.sectors[STAT_READ], | |
1596 | (unsigned int)div_u64(stat.nsecs[STAT_READ], | |
1597 | NSEC_PER_MSEC), | |
1598 | stat.ios[STAT_WRITE], | |
1599 | stat.merges[STAT_WRITE], | |
1600 | stat.sectors[STAT_WRITE], | |
1601 | (unsigned int)div_u64(stat.nsecs[STAT_WRITE], | |
1602 | NSEC_PER_MSEC), | |
e016b782 | 1603 | inflight, |
ea18e0f0 | 1604 | jiffies_to_msecs(stat.io_ticks), |
8cd5b8fc KK |
1605 | (unsigned int)div_u64(stat.nsecs[STAT_READ] + |
1606 | stat.nsecs[STAT_WRITE] + | |
1607 | stat.nsecs[STAT_DISCARD] + | |
1608 | stat.nsecs[STAT_FLUSH], | |
1609 | NSEC_PER_MSEC), | |
ea18e0f0 KK |
1610 | stat.ios[STAT_DISCARD], |
1611 | stat.merges[STAT_DISCARD], | |
1612 | stat.sectors[STAT_DISCARD], | |
1613 | (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], | |
1614 | NSEC_PER_MSEC), | |
1615 | stat.ios[STAT_FLUSH], | |
1616 | (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], | |
1617 | NSEC_PER_MSEC) | |
28f39d55 | 1618 | ); |
1da177e4 | 1619 | } |
e71bf0d0 | 1620 | disk_part_iter_exit(&piter); |
9f5e4865 | 1621 | |
1da177e4 LT |
1622 | return 0; |
1623 | } | |
1624 | ||
31d85ab2 | 1625 | static const struct seq_operations diskstats_op = { |
def4e38d TH |
1626 | .start = disk_seqf_start, |
1627 | .next = disk_seqf_next, | |
1628 | .stop = disk_seqf_stop, | |
1da177e4 LT |
1629 | .show = diskstats_show |
1630 | }; | |
f500975a AD |
1631 | |
1632 | static int __init proc_genhd_init(void) | |
1633 | { | |
fddda2b7 CH |
1634 | proc_create_seq("diskstats", 0, NULL, &diskstats_op); |
1635 | proc_create_seq("partitions", 0, NULL, &partitions_op); | |
f500975a AD |
1636 | return 0; |
1637 | } | |
1638 | module_init(proc_genhd_init); | |
a6e2ba88 | 1639 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 1640 | |
cf771cb5 | 1641 | dev_t blk_lookup_devt(const char *name, int partno) |
a142be85 | 1642 | { |
def4e38d TH |
1643 | dev_t devt = MKDEV(0, 0); |
1644 | struct class_dev_iter iter; | |
1645 | struct device *dev; | |
a142be85 | 1646 | |
def4e38d TH |
1647 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); |
1648 | while ((dev = class_dev_iter_next(&iter))) { | |
a142be85 | 1649 | struct gendisk *disk = dev_to_disk(dev); |
548b10eb | 1650 | struct hd_struct *part; |
a142be85 | 1651 | |
3ada8b7e | 1652 | if (strcmp(dev_name(dev), name)) |
f331c029 | 1653 | continue; |
f331c029 | 1654 | |
41b8c853 NB |
1655 | if (partno < disk->minors) { |
1656 | /* We need to return the right devno, even | |
1657 | * if the partition doesn't exist yet. | |
1658 | */ | |
1659 | devt = MKDEV(MAJOR(dev->devt), | |
1660 | MINOR(dev->devt) + partno); | |
1661 | break; | |
1662 | } | |
548b10eb | 1663 | part = disk_get_part(disk, partno); |
2bbedcb4 | 1664 | if (part) { |
f331c029 | 1665 | devt = part_devt(part); |
e71bf0d0 | 1666 | disk_put_part(part); |
548b10eb | 1667 | break; |
def4e38d | 1668 | } |
548b10eb | 1669 | disk_put_part(part); |
5c0ef6d0 | 1670 | } |
def4e38d | 1671 | class_dev_iter_exit(&iter); |
edfaa7c3 KS |
1672 | return devt; |
1673 | } | |
edfaa7c3 | 1674 | |
e319e1fb | 1675 | struct gendisk *__alloc_disk_node(int minors, int node_id) |
1946089a CL |
1676 | { |
1677 | struct gendisk *disk; | |
6d2cf6f2 | 1678 | struct disk_part_tbl *ptbl; |
1946089a | 1679 | |
de65b012 CH |
1680 | if (minors > DISK_MAX_PARTS) { |
1681 | printk(KERN_ERR | |
7fb52621 | 1682 | "block: can't allocate more than %d partitions\n", |
de65b012 CH |
1683 | DISK_MAX_PARTS); |
1684 | minors = DISK_MAX_PARTS; | |
1685 | } | |
1946089a | 1686 | |
c1b511eb | 1687 | disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id); |
1da177e4 | 1688 | if (disk) { |
58d4f14f CH |
1689 | disk->part0.dkstats = alloc_percpu(struct disk_stats); |
1690 | if (!disk->part0.dkstats) { | |
1da177e4 LT |
1691 | kfree(disk); |
1692 | return NULL; | |
1693 | } | |
56c0908c | 1694 | init_rwsem(&disk->lookup_sem); |
bf91db18 | 1695 | disk->node_id = node_id; |
540eed56 | 1696 | if (disk_expand_part_tbl(disk, 0)) { |
58d4f14f | 1697 | free_percpu(disk->part0.dkstats); |
b5d0b9df TH |
1698 | kfree(disk); |
1699 | return NULL; | |
1da177e4 | 1700 | } |
6d2cf6f2 BVA |
1701 | ptbl = rcu_dereference_protected(disk->part_tbl, 1); |
1702 | rcu_assign_pointer(ptbl->part[0], &disk->part0); | |
6c23a968 | 1703 | |
c83f6bf9 VG |
1704 | /* |
1705 | * set_capacity() and get_capacity() currently don't use | |
1706 | * seqcounter to read/update the part0->nr_sects. Still init | |
1707 | * the counter as we can read the sectors in IO submission | |
1708 | * patch using seqence counters. | |
1709 | * | |
1710 | * TODO: Ideally set_capacity() and get_capacity() should be | |
1711 | * converted to make use of bd_mutex and sequence counters. | |
1712 | */ | |
07c4e1e8 | 1713 | hd_sects_seq_init(&disk->part0); |
6c71013e ML |
1714 | if (hd_ref_init(&disk->part0)) { |
1715 | hd_free_part(&disk->part0); | |
1716 | kfree(disk); | |
1717 | return NULL; | |
1718 | } | |
b5d0b9df | 1719 | |
1da177e4 | 1720 | disk->minors = minors; |
1da177e4 | 1721 | rand_initialize_disk(disk); |
ed9e1982 TH |
1722 | disk_to_dev(disk)->class = &block_class; |
1723 | disk_to_dev(disk)->type = &disk_type; | |
1724 | device_initialize(disk_to_dev(disk)); | |
1da177e4 LT |
1725 | } |
1726 | return disk; | |
1727 | } | |
e319e1fb | 1728 | EXPORT_SYMBOL(__alloc_disk_node); |
1da177e4 | 1729 | |
3079c22e | 1730 | struct kobject *get_disk_and_module(struct gendisk *disk) |
1da177e4 LT |
1731 | { |
1732 | struct module *owner; | |
1733 | struct kobject *kobj; | |
1734 | ||
1735 | if (!disk->fops) | |
1736 | return NULL; | |
1737 | owner = disk->fops->owner; | |
1738 | if (owner && !try_module_get(owner)) | |
1739 | return NULL; | |
d01b2dcb | 1740 | kobj = kobject_get_unless_zero(&disk_to_dev(disk)->kobj); |
1da177e4 LT |
1741 | if (kobj == NULL) { |
1742 | module_put(owner); | |
1743 | return NULL; | |
1744 | } | |
1745 | return kobj; | |
1746 | ||
1747 | } | |
3079c22e | 1748 | EXPORT_SYMBOL(get_disk_and_module); |
1da177e4 LT |
1749 | |
1750 | void put_disk(struct gendisk *disk) | |
1751 | { | |
1752 | if (disk) | |
ed9e1982 | 1753 | kobject_put(&disk_to_dev(disk)->kobj); |
1da177e4 | 1754 | } |
1da177e4 LT |
1755 | EXPORT_SYMBOL(put_disk); |
1756 | ||
9df6c299 JK |
1757 | /* |
1758 | * This is a counterpart of get_disk_and_module() and thus also of | |
1759 | * get_gendisk(). | |
1760 | */ | |
1761 | void put_disk_and_module(struct gendisk *disk) | |
1762 | { | |
1763 | if (disk) { | |
1764 | struct module *owner = disk->fops->owner; | |
1765 | ||
1766 | put_disk(disk); | |
1767 | module_put(owner); | |
1768 | } | |
1769 | } | |
1770 | EXPORT_SYMBOL(put_disk_and_module); | |
1771 | ||
e3264a4d HR |
1772 | static void set_disk_ro_uevent(struct gendisk *gd, int ro) |
1773 | { | |
1774 | char event[] = "DISK_RO=1"; | |
1775 | char *envp[] = { event, NULL }; | |
1776 | ||
1777 | if (!ro) | |
1778 | event[8] = '0'; | |
1779 | kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); | |
1780 | } | |
1781 | ||
1da177e4 LT |
1782 | void set_device_ro(struct block_device *bdev, int flag) |
1783 | { | |
b7db9956 | 1784 | bdev->bd_part->policy = flag; |
1da177e4 LT |
1785 | } |
1786 | ||
1787 | EXPORT_SYMBOL(set_device_ro); | |
1788 | ||
1789 | void set_disk_ro(struct gendisk *disk, int flag) | |
1790 | { | |
e71bf0d0 TH |
1791 | struct disk_part_iter piter; |
1792 | struct hd_struct *part; | |
1793 | ||
e3264a4d HR |
1794 | if (disk->part0.policy != flag) { |
1795 | set_disk_ro_uevent(disk, flag); | |
1796 | disk->part0.policy = flag; | |
1797 | } | |
1798 | ||
1799 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY); | |
e71bf0d0 TH |
1800 | while ((part = disk_part_iter_next(&piter))) |
1801 | part->policy = flag; | |
1802 | disk_part_iter_exit(&piter); | |
1da177e4 LT |
1803 | } |
1804 | ||
1805 | EXPORT_SYMBOL(set_disk_ro); | |
1806 | ||
1807 | int bdev_read_only(struct block_device *bdev) | |
1808 | { | |
1809 | if (!bdev) | |
1810 | return 0; | |
b7db9956 | 1811 | return bdev->bd_part->policy; |
1da177e4 LT |
1812 | } |
1813 | ||
1814 | EXPORT_SYMBOL(bdev_read_only); | |
1815 | ||
77ea887e TH |
1816 | /* |
1817 | * Disk events - monitor disk events like media change and eject request. | |
1818 | */ | |
1819 | struct disk_events { | |
1820 | struct list_head node; /* all disk_event's */ | |
1821 | struct gendisk *disk; /* the associated disk */ | |
1822 | spinlock_t lock; | |
1823 | ||
fdd514e1 | 1824 | struct mutex block_mutex; /* protects blocking */ |
77ea887e TH |
1825 | int block; /* event blocking depth */ |
1826 | unsigned int pending; /* events already sent out */ | |
1827 | unsigned int clearing; /* events being cleared */ | |
1828 | ||
1829 | long poll_msecs; /* interval, -1 for default */ | |
1830 | struct delayed_work dwork; | |
1831 | }; | |
1832 | ||
1833 | static const char *disk_events_strs[] = { | |
1834 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change", | |
1835 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request", | |
1836 | }; | |
1837 | ||
1838 | static char *disk_uevents[] = { | |
1839 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1", | |
1840 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1", | |
1841 | }; | |
1842 | ||
1843 | /* list of all disk_events */ | |
1844 | static DEFINE_MUTEX(disk_events_mutex); | |
1845 | static LIST_HEAD(disk_events); | |
1846 | ||
1847 | /* disable in-kernel polling by default */ | |
1fe8f348 | 1848 | static unsigned long disk_events_dfl_poll_msecs; |
77ea887e TH |
1849 | |
1850 | static unsigned long disk_events_poll_jiffies(struct gendisk *disk) | |
1851 | { | |
1852 | struct disk_events *ev = disk->ev; | |
1853 | long intv_msecs = 0; | |
1854 | ||
1855 | /* | |
1856 | * If device-specific poll interval is set, always use it. If | |
673387a9 | 1857 | * the default is being used, poll if the POLL flag is set. |
77ea887e TH |
1858 | */ |
1859 | if (ev->poll_msecs >= 0) | |
1860 | intv_msecs = ev->poll_msecs; | |
c92e2f04 | 1861 | else if (disk->event_flags & DISK_EVENT_FLAG_POLL) |
77ea887e TH |
1862 | intv_msecs = disk_events_dfl_poll_msecs; |
1863 | ||
1864 | return msecs_to_jiffies(intv_msecs); | |
1865 | } | |
1866 | ||
c3af54af TH |
1867 | /** |
1868 | * disk_block_events - block and flush disk event checking | |
1869 | * @disk: disk to block events for | |
1870 | * | |
1871 | * On return from this function, it is guaranteed that event checking | |
1872 | * isn't in progress and won't happen until unblocked by | |
1873 | * disk_unblock_events(). Events blocking is counted and the actual | |
1874 | * unblocking happens after the matching number of unblocks are done. | |
1875 | * | |
1876 | * Note that this intentionally does not block event checking from | |
1877 | * disk_clear_events(). | |
1878 | * | |
1879 | * CONTEXT: | |
1880 | * Might sleep. | |
1881 | */ | |
1882 | void disk_block_events(struct gendisk *disk) | |
77ea887e TH |
1883 | { |
1884 | struct disk_events *ev = disk->ev; | |
1885 | unsigned long flags; | |
1886 | bool cancel; | |
1887 | ||
c3af54af TH |
1888 | if (!ev) |
1889 | return; | |
1890 | ||
fdd514e1 TH |
1891 | /* |
1892 | * Outer mutex ensures that the first blocker completes canceling | |
1893 | * the event work before further blockers are allowed to finish. | |
1894 | */ | |
1895 | mutex_lock(&ev->block_mutex); | |
1896 | ||
77ea887e TH |
1897 | spin_lock_irqsave(&ev->lock, flags); |
1898 | cancel = !ev->block++; | |
1899 | spin_unlock_irqrestore(&ev->lock, flags); | |
1900 | ||
c3af54af TH |
1901 | if (cancel) |
1902 | cancel_delayed_work_sync(&disk->ev->dwork); | |
fdd514e1 TH |
1903 | |
1904 | mutex_unlock(&ev->block_mutex); | |
77ea887e TH |
1905 | } |
1906 | ||
1907 | static void __disk_unblock_events(struct gendisk *disk, bool check_now) | |
1908 | { | |
1909 | struct disk_events *ev = disk->ev; | |
1910 | unsigned long intv; | |
1911 | unsigned long flags; | |
1912 | ||
1913 | spin_lock_irqsave(&ev->lock, flags); | |
1914 | ||
1915 | if (WARN_ON_ONCE(ev->block <= 0)) | |
1916 | goto out_unlock; | |
1917 | ||
1918 | if (--ev->block) | |
1919 | goto out_unlock; | |
1920 | ||
77ea887e | 1921 | intv = disk_events_poll_jiffies(disk); |
77ea887e | 1922 | if (check_now) |
695588f9 VK |
1923 | queue_delayed_work(system_freezable_power_efficient_wq, |
1924 | &ev->dwork, 0); | |
77ea887e | 1925 | else if (intv) |
695588f9 VK |
1926 | queue_delayed_work(system_freezable_power_efficient_wq, |
1927 | &ev->dwork, intv); | |
77ea887e TH |
1928 | out_unlock: |
1929 | spin_unlock_irqrestore(&ev->lock, flags); | |
1930 | } | |
1931 | ||
77ea887e TH |
1932 | /** |
1933 | * disk_unblock_events - unblock disk event checking | |
1934 | * @disk: disk to unblock events for | |
1935 | * | |
1936 | * Undo disk_block_events(). When the block count reaches zero, it | |
1937 | * starts events polling if configured. | |
1938 | * | |
1939 | * CONTEXT: | |
1940 | * Don't care. Safe to call from irq context. | |
1941 | */ | |
1942 | void disk_unblock_events(struct gendisk *disk) | |
1943 | { | |
1944 | if (disk->ev) | |
facc31dd | 1945 | __disk_unblock_events(disk, false); |
77ea887e TH |
1946 | } |
1947 | ||
1948 | /** | |
85ef06d1 TH |
1949 | * disk_flush_events - schedule immediate event checking and flushing |
1950 | * @disk: disk to check and flush events for | |
1951 | * @mask: events to flush | |
77ea887e | 1952 | * |
85ef06d1 TH |
1953 | * Schedule immediate event checking on @disk if not blocked. Events in |
1954 | * @mask are scheduled to be cleared from the driver. Note that this | |
1955 | * doesn't clear the events from @disk->ev. | |
77ea887e TH |
1956 | * |
1957 | * CONTEXT: | |
85ef06d1 | 1958 | * If @mask is non-zero must be called with bdev->bd_mutex held. |
77ea887e | 1959 | */ |
85ef06d1 | 1960 | void disk_flush_events(struct gendisk *disk, unsigned int mask) |
77ea887e | 1961 | { |
a9dce2a3 | 1962 | struct disk_events *ev = disk->ev; |
a9dce2a3 TH |
1963 | |
1964 | if (!ev) | |
1965 | return; | |
1966 | ||
85ef06d1 TH |
1967 | spin_lock_irq(&ev->lock); |
1968 | ev->clearing |= mask; | |
41f63c53 | 1969 | if (!ev->block) |
695588f9 VK |
1970 | mod_delayed_work(system_freezable_power_efficient_wq, |
1971 | &ev->dwork, 0); | |
85ef06d1 | 1972 | spin_unlock_irq(&ev->lock); |
77ea887e | 1973 | } |
77ea887e TH |
1974 | |
1975 | /** | |
1976 | * disk_clear_events - synchronously check, clear and return pending events | |
1977 | * @disk: disk to fetch and clear events from | |
da3dae54 | 1978 | * @mask: mask of events to be fetched and cleared |
77ea887e TH |
1979 | * |
1980 | * Disk events are synchronously checked and pending events in @mask | |
1981 | * are cleared and returned. This ignores the block count. | |
1982 | * | |
1983 | * CONTEXT: | |
1984 | * Might sleep. | |
1985 | */ | |
1986 | unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask) | |
1987 | { | |
1988 | const struct block_device_operations *bdops = disk->fops; | |
1989 | struct disk_events *ev = disk->ev; | |
1990 | unsigned int pending; | |
12c2bdb2 | 1991 | unsigned int clearing = mask; |
77ea887e TH |
1992 | |
1993 | if (!ev) { | |
1994 | /* for drivers still using the old ->media_changed method */ | |
1995 | if ((mask & DISK_EVENT_MEDIA_CHANGE) && | |
1996 | bdops->media_changed && bdops->media_changed(disk)) | |
1997 | return DISK_EVENT_MEDIA_CHANGE; | |
1998 | return 0; | |
1999 | } | |
2000 | ||
12c2bdb2 DB |
2001 | disk_block_events(disk); |
2002 | ||
2003 | /* | |
2004 | * store the union of mask and ev->clearing on the stack so that the | |
2005 | * race with disk_flush_events does not cause ambiguity (ev->clearing | |
2006 | * can still be modified even if events are blocked). | |
2007 | */ | |
77ea887e | 2008 | spin_lock_irq(&ev->lock); |
12c2bdb2 DB |
2009 | clearing |= ev->clearing; |
2010 | ev->clearing = 0; | |
77ea887e TH |
2011 | spin_unlock_irq(&ev->lock); |
2012 | ||
12c2bdb2 | 2013 | disk_check_events(ev, &clearing); |
aea24a8b | 2014 | /* |
12c2bdb2 DB |
2015 | * if ev->clearing is not 0, the disk_flush_events got called in the |
2016 | * middle of this function, so we want to run the workfn without delay. | |
aea24a8b | 2017 | */ |
12c2bdb2 | 2018 | __disk_unblock_events(disk, ev->clearing ? true : false); |
77ea887e TH |
2019 | |
2020 | /* then, fetch and clear pending events */ | |
2021 | spin_lock_irq(&ev->lock); | |
77ea887e TH |
2022 | pending = ev->pending & mask; |
2023 | ev->pending &= ~mask; | |
2024 | spin_unlock_irq(&ev->lock); | |
12c2bdb2 | 2025 | WARN_ON_ONCE(clearing & mask); |
77ea887e TH |
2026 | |
2027 | return pending; | |
2028 | } | |
2029 | ||
12c2bdb2 DB |
2030 | /* |
2031 | * Separate this part out so that a different pointer for clearing_ptr can be | |
2032 | * passed in for disk_clear_events. | |
2033 | */ | |
77ea887e TH |
2034 | static void disk_events_workfn(struct work_struct *work) |
2035 | { | |
2036 | struct delayed_work *dwork = to_delayed_work(work); | |
2037 | struct disk_events *ev = container_of(dwork, struct disk_events, dwork); | |
12c2bdb2 DB |
2038 | |
2039 | disk_check_events(ev, &ev->clearing); | |
2040 | } | |
2041 | ||
2042 | static void disk_check_events(struct disk_events *ev, | |
2043 | unsigned int *clearing_ptr) | |
2044 | { | |
77ea887e TH |
2045 | struct gendisk *disk = ev->disk; |
2046 | char *envp[ARRAY_SIZE(disk_uevents) + 1] = { }; | |
12c2bdb2 | 2047 | unsigned int clearing = *clearing_ptr; |
77ea887e TH |
2048 | unsigned int events; |
2049 | unsigned long intv; | |
2050 | int nr_events = 0, i; | |
2051 | ||
2052 | /* check events */ | |
2053 | events = disk->fops->check_events(disk, clearing); | |
2054 | ||
2055 | /* accumulate pending events and schedule next poll if necessary */ | |
2056 | spin_lock_irq(&ev->lock); | |
2057 | ||
2058 | events &= ~ev->pending; | |
2059 | ev->pending |= events; | |
12c2bdb2 | 2060 | *clearing_ptr &= ~clearing; |
77ea887e TH |
2061 | |
2062 | intv = disk_events_poll_jiffies(disk); | |
2063 | if (!ev->block && intv) | |
695588f9 VK |
2064 | queue_delayed_work(system_freezable_power_efficient_wq, |
2065 | &ev->dwork, intv); | |
77ea887e TH |
2066 | |
2067 | spin_unlock_irq(&ev->lock); | |
2068 | ||
7c88a168 TH |
2069 | /* |
2070 | * Tell userland about new events. Only the events listed in | |
c92e2f04 MW |
2071 | * @disk->events are reported, and only if DISK_EVENT_FLAG_UEVENT |
2072 | * is set. Otherwise, events are processed internally but never | |
2073 | * get reported to userland. | |
7c88a168 | 2074 | */ |
77ea887e | 2075 | for (i = 0; i < ARRAY_SIZE(disk_uevents); i++) |
c92e2f04 MW |
2076 | if ((events & disk->events & (1 << i)) && |
2077 | (disk->event_flags & DISK_EVENT_FLAG_UEVENT)) | |
77ea887e TH |
2078 | envp[nr_events++] = disk_uevents[i]; |
2079 | ||
2080 | if (nr_events) | |
2081 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
2082 | } | |
2083 | ||
2084 | /* | |
2085 | * A disk events enabled device has the following sysfs nodes under | |
2086 | * its /sys/block/X/ directory. | |
2087 | * | |
2088 | * events : list of all supported events | |
2089 | * events_async : list of events which can be detected w/o polling | |
673387a9 | 2090 | * (always empty, only for backwards compatibility) |
77ea887e TH |
2091 | * events_poll_msecs : polling interval, 0: disable, -1: system default |
2092 | */ | |
2093 | static ssize_t __disk_events_show(unsigned int events, char *buf) | |
2094 | { | |
2095 | const char *delim = ""; | |
2096 | ssize_t pos = 0; | |
2097 | int i; | |
2098 | ||
2099 | for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++) | |
2100 | if (events & (1 << i)) { | |
2101 | pos += sprintf(buf + pos, "%s%s", | |
2102 | delim, disk_events_strs[i]); | |
2103 | delim = " "; | |
2104 | } | |
2105 | if (pos) | |
2106 | pos += sprintf(buf + pos, "\n"); | |
2107 | return pos; | |
2108 | } | |
2109 | ||
2110 | static ssize_t disk_events_show(struct device *dev, | |
2111 | struct device_attribute *attr, char *buf) | |
2112 | { | |
2113 | struct gendisk *disk = dev_to_disk(dev); | |
2114 | ||
c92e2f04 MW |
2115 | if (!(disk->event_flags & DISK_EVENT_FLAG_UEVENT)) |
2116 | return 0; | |
2117 | ||
77ea887e TH |
2118 | return __disk_events_show(disk->events, buf); |
2119 | } | |
2120 | ||
2121 | static ssize_t disk_events_async_show(struct device *dev, | |
2122 | struct device_attribute *attr, char *buf) | |
2123 | { | |
673387a9 | 2124 | return 0; |
77ea887e TH |
2125 | } |
2126 | ||
2127 | static ssize_t disk_events_poll_msecs_show(struct device *dev, | |
2128 | struct device_attribute *attr, | |
2129 | char *buf) | |
2130 | { | |
2131 | struct gendisk *disk = dev_to_disk(dev); | |
2132 | ||
cdf3e3de MW |
2133 | if (!disk->ev) |
2134 | return sprintf(buf, "-1\n"); | |
2135 | ||
77ea887e TH |
2136 | return sprintf(buf, "%ld\n", disk->ev->poll_msecs); |
2137 | } | |
2138 | ||
2139 | static ssize_t disk_events_poll_msecs_store(struct device *dev, | |
2140 | struct device_attribute *attr, | |
2141 | const char *buf, size_t count) | |
2142 | { | |
2143 | struct gendisk *disk = dev_to_disk(dev); | |
2144 | long intv; | |
2145 | ||
2146 | if (!count || !sscanf(buf, "%ld", &intv)) | |
2147 | return -EINVAL; | |
2148 | ||
2149 | if (intv < 0 && intv != -1) | |
2150 | return -EINVAL; | |
2151 | ||
cdf3e3de MW |
2152 | if (!disk->ev) |
2153 | return -ENODEV; | |
2154 | ||
c3af54af | 2155 | disk_block_events(disk); |
77ea887e TH |
2156 | disk->ev->poll_msecs = intv; |
2157 | __disk_unblock_events(disk, true); | |
2158 | ||
2159 | return count; | |
2160 | } | |
2161 | ||
5657a819 JP |
2162 | static const DEVICE_ATTR(events, 0444, disk_events_show, NULL); |
2163 | static const DEVICE_ATTR(events_async, 0444, disk_events_async_show, NULL); | |
2164 | static const DEVICE_ATTR(events_poll_msecs, 0644, | |
77ea887e TH |
2165 | disk_events_poll_msecs_show, |
2166 | disk_events_poll_msecs_store); | |
2167 | ||
2168 | static const struct attribute *disk_events_attrs[] = { | |
2169 | &dev_attr_events.attr, | |
2170 | &dev_attr_events_async.attr, | |
2171 | &dev_attr_events_poll_msecs.attr, | |
2172 | NULL, | |
2173 | }; | |
2174 | ||
2175 | /* | |
2176 | * The default polling interval can be specified by the kernel | |
2177 | * parameter block.events_dfl_poll_msecs which defaults to 0 | |
2178 | * (disable). This can also be modified runtime by writing to | |
1624b0b2 | 2179 | * /sys/module/block/parameters/events_dfl_poll_msecs. |
77ea887e TH |
2180 | */ |
2181 | static int disk_events_set_dfl_poll_msecs(const char *val, | |
2182 | const struct kernel_param *kp) | |
2183 | { | |
2184 | struct disk_events *ev; | |
2185 | int ret; | |
2186 | ||
2187 | ret = param_set_ulong(val, kp); | |
2188 | if (ret < 0) | |
2189 | return ret; | |
2190 | ||
2191 | mutex_lock(&disk_events_mutex); | |
2192 | ||
2193 | list_for_each_entry(ev, &disk_events, node) | |
85ef06d1 | 2194 | disk_flush_events(ev->disk, 0); |
77ea887e TH |
2195 | |
2196 | mutex_unlock(&disk_events_mutex); | |
2197 | ||
2198 | return 0; | |
2199 | } | |
2200 | ||
2201 | static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = { | |
2202 | .set = disk_events_set_dfl_poll_msecs, | |
2203 | .get = param_get_ulong, | |
2204 | }; | |
2205 | ||
2206 | #undef MODULE_PARAM_PREFIX | |
2207 | #define MODULE_PARAM_PREFIX "block." | |
2208 | ||
2209 | module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops, | |
2210 | &disk_events_dfl_poll_msecs, 0644); | |
2211 | ||
2212 | /* | |
9f53d2fe | 2213 | * disk_{alloc|add|del|release}_events - initialize and destroy disk_events. |
77ea887e | 2214 | */ |
9f53d2fe | 2215 | static void disk_alloc_events(struct gendisk *disk) |
77ea887e TH |
2216 | { |
2217 | struct disk_events *ev; | |
2218 | ||
cdf3e3de | 2219 | if (!disk->fops->check_events || !disk->events) |
77ea887e TH |
2220 | return; |
2221 | ||
2222 | ev = kzalloc(sizeof(*ev), GFP_KERNEL); | |
2223 | if (!ev) { | |
2224 | pr_warn("%s: failed to initialize events\n", disk->disk_name); | |
2225 | return; | |
2226 | } | |
2227 | ||
77ea887e TH |
2228 | INIT_LIST_HEAD(&ev->node); |
2229 | ev->disk = disk; | |
2230 | spin_lock_init(&ev->lock); | |
fdd514e1 | 2231 | mutex_init(&ev->block_mutex); |
77ea887e TH |
2232 | ev->block = 1; |
2233 | ev->poll_msecs = -1; | |
2234 | INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn); | |
2235 | ||
9f53d2fe SG |
2236 | disk->ev = ev; |
2237 | } | |
2238 | ||
2239 | static void disk_add_events(struct gendisk *disk) | |
2240 | { | |
9f53d2fe SG |
2241 | /* FIXME: error handling */ |
2242 | if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0) | |
2243 | pr_warn("%s: failed to create sysfs files for events\n", | |
2244 | disk->disk_name); | |
2245 | ||
cdf3e3de MW |
2246 | if (!disk->ev) |
2247 | return; | |
2248 | ||
77ea887e | 2249 | mutex_lock(&disk_events_mutex); |
9f53d2fe | 2250 | list_add_tail(&disk->ev->node, &disk_events); |
77ea887e TH |
2251 | mutex_unlock(&disk_events_mutex); |
2252 | ||
2253 | /* | |
2254 | * Block count is initialized to 1 and the following initial | |
2255 | * unblock kicks it into action. | |
2256 | */ | |
2257 | __disk_unblock_events(disk, true); | |
2258 | } | |
2259 | ||
2260 | static void disk_del_events(struct gendisk *disk) | |
2261 | { | |
cdf3e3de MW |
2262 | if (disk->ev) { |
2263 | disk_block_events(disk); | |
77ea887e | 2264 | |
cdf3e3de MW |
2265 | mutex_lock(&disk_events_mutex); |
2266 | list_del_init(&disk->ev->node); | |
2267 | mutex_unlock(&disk_events_mutex); | |
2268 | } | |
77ea887e TH |
2269 | |
2270 | sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs); | |
2271 | } | |
2272 | ||
2273 | static void disk_release_events(struct gendisk *disk) | |
2274 | { | |
2275 | /* the block count should be 1 from disk_del_events() */ | |
2276 | WARN_ON_ONCE(disk->ev && disk->ev->block != 1); | |
2277 | kfree(disk->ev); | |
2278 | } |