]>
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 | ||
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 | ||
e598a72f BS |
48 | /* |
49 | * Set disk capacity and notify if the size is not currently | |
50 | * zero and will not be set to zero | |
51 | */ | |
52 | void set_capacity_revalidate_and_notify(struct gendisk *disk, sector_t size, | |
b8086d3f | 53 | bool update_bdev) |
e598a72f BS |
54 | { |
55 | sector_t capacity = get_capacity(disk); | |
56 | ||
57 | set_capacity(disk, size); | |
b8086d3f CH |
58 | if (update_bdev) |
59 | revalidate_disk_size(disk, true); | |
e598a72f BS |
60 | |
61 | if (capacity != size && capacity != 0 && size != 0) { | |
62 | char *envp[] = { "RESIZE=1", NULL }; | |
63 | ||
64 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
65 | } | |
66 | } | |
67 | ||
68 | EXPORT_SYMBOL_GPL(set_capacity_revalidate_and_notify); | |
69 | ||
5cbd28e3 CH |
70 | /* |
71 | * Format the device name of the indicated disk into the supplied buffer and | |
72 | * return a pointer to that same buffer for convenience. | |
73 | */ | |
74 | char *disk_name(struct gendisk *hd, int partno, char *buf) | |
75 | { | |
76 | if (!partno) | |
77 | snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name); | |
78 | else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1])) | |
79 | snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno); | |
80 | else | |
81 | snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno); | |
82 | ||
83 | return buf; | |
84 | } | |
85 | ||
86 | const char *bdevname(struct block_device *bdev, char *buf) | |
87 | { | |
88 | return disk_name(bdev->bd_disk, bdev->bd_part->partno, buf); | |
89 | } | |
90 | EXPORT_SYMBOL(bdevname); | |
e598a72f | 91 | |
ea18e0f0 KK |
92 | static void part_stat_read_all(struct hd_struct *part, struct disk_stats *stat) |
93 | { | |
94 | int cpu; | |
95 | ||
96 | memset(stat, 0, sizeof(struct disk_stats)); | |
97 | for_each_possible_cpu(cpu) { | |
98 | struct disk_stats *ptr = per_cpu_ptr(part->dkstats, cpu); | |
99 | int group; | |
100 | ||
101 | for (group = 0; group < NR_STAT_GROUPS; group++) { | |
102 | stat->nsecs[group] += ptr->nsecs[group]; | |
103 | stat->sectors[group] += ptr->sectors[group]; | |
104 | stat->ios[group] += ptr->ios[group]; | |
105 | stat->merges[group] += ptr->merges[group]; | |
106 | } | |
107 | ||
108 | stat->io_ticks += ptr->io_ticks; | |
ea18e0f0 KK |
109 | } |
110 | } | |
ea18e0f0 | 111 | |
1f06959b | 112 | static unsigned int part_in_flight(struct hd_struct *part) |
f299b7c7 | 113 | { |
b2f609e1 | 114 | unsigned int inflight = 0; |
1226b8dd | 115 | int cpu; |
f299b7c7 | 116 | |
1226b8dd | 117 | for_each_possible_cpu(cpu) { |
e016b782 MP |
118 | inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) + |
119 | part_stat_local_read_cpu(part, in_flight[1], cpu); | |
1226b8dd | 120 | } |
e016b782 MP |
121 | if ((int)inflight < 0) |
122 | inflight = 0; | |
1226b8dd | 123 | |
e016b782 | 124 | return inflight; |
f299b7c7 JA |
125 | } |
126 | ||
1f06959b | 127 | static void part_in_flight_rw(struct hd_struct *part, unsigned int inflight[2]) |
bf0ddaba | 128 | { |
1226b8dd MP |
129 | int cpu; |
130 | ||
1226b8dd MP |
131 | inflight[0] = 0; |
132 | inflight[1] = 0; | |
133 | for_each_possible_cpu(cpu) { | |
134 | inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu); | |
135 | inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu); | |
136 | } | |
137 | if ((int)inflight[0] < 0) | |
138 | inflight[0] = 0; | |
139 | if ((int)inflight[1] < 0) | |
140 | inflight[1] = 0; | |
bf0ddaba OS |
141 | } |
142 | ||
807d4af2 CH |
143 | struct hd_struct *__disk_get_part(struct gendisk *disk, int partno) |
144 | { | |
145 | struct disk_part_tbl *ptbl = rcu_dereference(disk->part_tbl); | |
146 | ||
147 | if (unlikely(partno < 0 || partno >= ptbl->len)) | |
148 | return NULL; | |
149 | return rcu_dereference(ptbl->part[partno]); | |
150 | } | |
151 | ||
e71bf0d0 TH |
152 | /** |
153 | * disk_get_part - get partition | |
154 | * @disk: disk to look partition from | |
155 | * @partno: partition number | |
156 | * | |
157 | * Look for partition @partno from @disk. If found, increment | |
158 | * reference count and return it. | |
159 | * | |
160 | * CONTEXT: | |
161 | * Don't care. | |
162 | * | |
163 | * RETURNS: | |
164 | * Pointer to the found partition on success, NULL if not found. | |
165 | */ | |
166 | struct hd_struct *disk_get_part(struct gendisk *disk, int partno) | |
167 | { | |
807d4af2 | 168 | struct hd_struct *part; |
540eed56 | 169 | |
e71bf0d0 | 170 | rcu_read_lock(); |
807d4af2 CH |
171 | part = __disk_get_part(disk, partno); |
172 | if (part) | |
173 | get_device(part_to_dev(part)); | |
e71bf0d0 TH |
174 | rcu_read_unlock(); |
175 | ||
176 | return part; | |
177 | } | |
e71bf0d0 TH |
178 | |
179 | /** | |
180 | * disk_part_iter_init - initialize partition iterator | |
181 | * @piter: iterator to initialize | |
182 | * @disk: disk to iterate over | |
183 | * @flags: DISK_PITER_* flags | |
184 | * | |
185 | * Initialize @piter so that it iterates over partitions of @disk. | |
186 | * | |
187 | * CONTEXT: | |
188 | * Don't care. | |
189 | */ | |
190 | void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk, | |
191 | unsigned int flags) | |
192 | { | |
540eed56 TH |
193 | struct disk_part_tbl *ptbl; |
194 | ||
195 | rcu_read_lock(); | |
196 | ptbl = rcu_dereference(disk->part_tbl); | |
197 | ||
e71bf0d0 TH |
198 | piter->disk = disk; |
199 | piter->part = NULL; | |
200 | ||
201 | if (flags & DISK_PITER_REVERSE) | |
540eed56 | 202 | piter->idx = ptbl->len - 1; |
71982a40 | 203 | else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0)) |
e71bf0d0 | 204 | piter->idx = 0; |
b5d0b9df TH |
205 | else |
206 | piter->idx = 1; | |
e71bf0d0 TH |
207 | |
208 | piter->flags = flags; | |
540eed56 TH |
209 | |
210 | rcu_read_unlock(); | |
e71bf0d0 TH |
211 | } |
212 | EXPORT_SYMBOL_GPL(disk_part_iter_init); | |
213 | ||
214 | /** | |
215 | * disk_part_iter_next - proceed iterator to the next partition and return it | |
216 | * @piter: iterator of interest | |
217 | * | |
218 | * Proceed @piter to the next partition and return it. | |
219 | * | |
220 | * CONTEXT: | |
221 | * Don't care. | |
222 | */ | |
223 | struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter) | |
224 | { | |
540eed56 | 225 | struct disk_part_tbl *ptbl; |
e71bf0d0 TH |
226 | int inc, end; |
227 | ||
228 | /* put the last partition */ | |
229 | disk_put_part(piter->part); | |
230 | piter->part = NULL; | |
231 | ||
540eed56 | 232 | /* get part_tbl */ |
e71bf0d0 | 233 | rcu_read_lock(); |
540eed56 | 234 | ptbl = rcu_dereference(piter->disk->part_tbl); |
e71bf0d0 TH |
235 | |
236 | /* determine iteration parameters */ | |
237 | if (piter->flags & DISK_PITER_REVERSE) { | |
238 | inc = -1; | |
71982a40 TH |
239 | if (piter->flags & (DISK_PITER_INCL_PART0 | |
240 | DISK_PITER_INCL_EMPTY_PART0)) | |
b5d0b9df TH |
241 | end = -1; |
242 | else | |
243 | end = 0; | |
e71bf0d0 TH |
244 | } else { |
245 | inc = 1; | |
540eed56 | 246 | end = ptbl->len; |
e71bf0d0 TH |
247 | } |
248 | ||
249 | /* iterate to the next partition */ | |
250 | for (; piter->idx != end; piter->idx += inc) { | |
251 | struct hd_struct *part; | |
252 | ||
540eed56 | 253 | part = rcu_dereference(ptbl->part[piter->idx]); |
e71bf0d0 TH |
254 | if (!part) |
255 | continue; | |
c83f6bf9 | 256 | if (!part_nr_sects_read(part) && |
71982a40 TH |
257 | !(piter->flags & DISK_PITER_INCL_EMPTY) && |
258 | !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 && | |
259 | piter->idx == 0)) | |
e71bf0d0 TH |
260 | continue; |
261 | ||
ed9e1982 | 262 | get_device(part_to_dev(part)); |
e71bf0d0 TH |
263 | piter->part = part; |
264 | piter->idx += inc; | |
265 | break; | |
266 | } | |
267 | ||
268 | rcu_read_unlock(); | |
269 | ||
270 | return piter->part; | |
271 | } | |
272 | EXPORT_SYMBOL_GPL(disk_part_iter_next); | |
273 | ||
274 | /** | |
275 | * disk_part_iter_exit - finish up partition iteration | |
276 | * @piter: iter of interest | |
277 | * | |
278 | * Called when iteration is over. Cleans up @piter. | |
279 | * | |
280 | * CONTEXT: | |
281 | * Don't care. | |
282 | */ | |
283 | void disk_part_iter_exit(struct disk_part_iter *piter) | |
284 | { | |
285 | disk_put_part(piter->part); | |
286 | piter->part = NULL; | |
287 | } | |
288 | EXPORT_SYMBOL_GPL(disk_part_iter_exit); | |
289 | ||
a6f23657 JA |
290 | static inline int sector_in_part(struct hd_struct *part, sector_t sector) |
291 | { | |
292 | return part->start_sect <= sector && | |
c83f6bf9 | 293 | sector < part->start_sect + part_nr_sects_read(part); |
a6f23657 JA |
294 | } |
295 | ||
e71bf0d0 TH |
296 | /** |
297 | * disk_map_sector_rcu - map sector to partition | |
298 | * @disk: gendisk of interest | |
299 | * @sector: sector to map | |
300 | * | |
301 | * Find out which partition @sector maps to on @disk. This is | |
302 | * primarily used for stats accounting. | |
303 | * | |
304 | * CONTEXT: | |
b7d6c303 | 305 | * RCU read locked. The returned partition pointer is always valid |
27eb3af9 ML |
306 | * because its refcount is grabbed except for part0, which lifetime |
307 | * is same with the disk. | |
e71bf0d0 TH |
308 | * |
309 | * RETURNS: | |
074a7aca | 310 | * Found partition on success, part0 is returned if no partition matches |
b7d6c303 | 311 | * or the matched partition is being deleted. |
e71bf0d0 TH |
312 | */ |
313 | struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector) | |
314 | { | |
540eed56 | 315 | struct disk_part_tbl *ptbl; |
a6f23657 | 316 | struct hd_struct *part; |
e71bf0d0 TH |
317 | int i; |
318 | ||
8ab1d40a | 319 | rcu_read_lock(); |
540eed56 TH |
320 | ptbl = rcu_dereference(disk->part_tbl); |
321 | ||
a6f23657 | 322 | part = rcu_dereference(ptbl->last_lookup); |
b7d6c303 | 323 | if (part && sector_in_part(part, sector) && hd_struct_try_get(part)) |
8ab1d40a | 324 | goto out_unlock; |
a6f23657 | 325 | |
540eed56 | 326 | for (i = 1; i < ptbl->len; i++) { |
a6f23657 | 327 | part = rcu_dereference(ptbl->part[i]); |
e71bf0d0 | 328 | |
a6f23657 | 329 | if (part && sector_in_part(part, sector)) { |
b7d6c303 ML |
330 | /* |
331 | * only live partition can be cached for lookup, | |
332 | * so use-after-free on cached & deleting partition | |
333 | * can be avoided | |
334 | */ | |
335 | if (!hd_struct_try_get(part)) | |
336 | break; | |
a6f23657 | 337 | rcu_assign_pointer(ptbl->last_lookup, part); |
8ab1d40a | 338 | goto out_unlock; |
a6f23657 | 339 | } |
e71bf0d0 | 340 | } |
8ab1d40a KK |
341 | |
342 | part = &disk->part0; | |
343 | out_unlock: | |
344 | rcu_read_unlock(); | |
345 | return part; | |
e71bf0d0 | 346 | } |
e71bf0d0 | 347 | |
b53df2e7 SK |
348 | /** |
349 | * disk_has_partitions | |
350 | * @disk: gendisk of interest | |
351 | * | |
352 | * Walk through the partition table and check if valid partition exists. | |
353 | * | |
354 | * CONTEXT: | |
355 | * Don't care. | |
356 | * | |
357 | * RETURNS: | |
358 | * True if the gendisk has at least one valid non-zero size partition. | |
359 | * Otherwise false. | |
360 | */ | |
361 | bool disk_has_partitions(struct gendisk *disk) | |
362 | { | |
363 | struct disk_part_tbl *ptbl; | |
364 | int i; | |
365 | bool ret = false; | |
366 | ||
367 | rcu_read_lock(); | |
368 | ptbl = rcu_dereference(disk->part_tbl); | |
369 | ||
370 | /* Iterate partitions skipping the whole device at index 0 */ | |
371 | for (i = 1; i < ptbl->len; i++) { | |
372 | if (rcu_dereference(ptbl->part[i])) { | |
373 | ret = true; | |
374 | break; | |
375 | } | |
376 | } | |
377 | ||
378 | rcu_read_unlock(); | |
379 | ||
380 | return ret; | |
381 | } | |
382 | EXPORT_SYMBOL_GPL(disk_has_partitions); | |
383 | ||
1da177e4 LT |
384 | /* |
385 | * Can be deleted altogether. Later. | |
386 | * | |
387 | */ | |
133d55cd | 388 | #define BLKDEV_MAJOR_HASH_SIZE 255 |
1da177e4 LT |
389 | static struct blk_major_name { |
390 | struct blk_major_name *next; | |
391 | int major; | |
392 | char name[16]; | |
68eef3b4 | 393 | } *major_names[BLKDEV_MAJOR_HASH_SIZE]; |
1da177e4 LT |
394 | |
395 | /* index in the above - for now: assume no multimajor ranges */ | |
e61eb2e9 | 396 | static inline int major_to_index(unsigned major) |
1da177e4 | 397 | { |
68eef3b4 | 398 | return major % BLKDEV_MAJOR_HASH_SIZE; |
7170be5f NH |
399 | } |
400 | ||
68eef3b4 | 401 | #ifdef CONFIG_PROC_FS |
cf771cb5 | 402 | void blkdev_show(struct seq_file *seqf, off_t offset) |
7170be5f | 403 | { |
68eef3b4 | 404 | struct blk_major_name *dp; |
7170be5f | 405 | |
133d55cd LG |
406 | mutex_lock(&block_class_lock); |
407 | for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next) | |
408 | if (dp->major == offset) | |
cf771cb5 | 409 | seq_printf(seqf, "%3d %s\n", dp->major, dp->name); |
133d55cd | 410 | mutex_unlock(&block_class_lock); |
1da177e4 | 411 | } |
68eef3b4 | 412 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 413 | |
9e8c0bcc MN |
414 | /** |
415 | * register_blkdev - register a new block device | |
416 | * | |
f33ff110 SB |
417 | * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If |
418 | * @major = 0, try to allocate any unused major number. | |
9e8c0bcc MN |
419 | * @name: the name of the new block device as a zero terminated string |
420 | * | |
421 | * The @name must be unique within the system. | |
422 | * | |
0e056eb5 MCC |
423 | * The return value depends on the @major input parameter: |
424 | * | |
f33ff110 SB |
425 | * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1] |
426 | * then the function returns zero on success, or a negative error code | |
0e056eb5 | 427 | * - if any unused major number was requested with @major = 0 parameter |
9e8c0bcc | 428 | * then the return value is the allocated major number in range |
f33ff110 SB |
429 | * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise |
430 | * | |
431 | * See Documentation/admin-guide/devices.txt for the list of allocated | |
432 | * major numbers. | |
9e8c0bcc | 433 | */ |
1da177e4 LT |
434 | int register_blkdev(unsigned int major, const char *name) |
435 | { | |
436 | struct blk_major_name **n, *p; | |
437 | int index, ret = 0; | |
438 | ||
edfaa7c3 | 439 | mutex_lock(&block_class_lock); |
1da177e4 LT |
440 | |
441 | /* temporary */ | |
442 | if (major == 0) { | |
443 | for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) { | |
444 | if (major_names[index] == NULL) | |
445 | break; | |
446 | } | |
447 | ||
448 | if (index == 0) { | |
dfc76d11 KP |
449 | printk("%s: failed to get major for %s\n", |
450 | __func__, name); | |
1da177e4 LT |
451 | ret = -EBUSY; |
452 | goto out; | |
453 | } | |
454 | major = index; | |
455 | ret = major; | |
456 | } | |
457 | ||
133d55cd | 458 | if (major >= BLKDEV_MAJOR_MAX) { |
dfc76d11 KP |
459 | pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n", |
460 | __func__, major, BLKDEV_MAJOR_MAX-1, name); | |
133d55cd LG |
461 | |
462 | ret = -EINVAL; | |
463 | goto out; | |
464 | } | |
465 | ||
1da177e4 LT |
466 | p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL); |
467 | if (p == NULL) { | |
468 | ret = -ENOMEM; | |
469 | goto out; | |
470 | } | |
471 | ||
472 | p->major = major; | |
473 | strlcpy(p->name, name, sizeof(p->name)); | |
474 | p->next = NULL; | |
475 | index = major_to_index(major); | |
476 | ||
477 | for (n = &major_names[index]; *n; n = &(*n)->next) { | |
478 | if ((*n)->major == major) | |
479 | break; | |
480 | } | |
481 | if (!*n) | |
482 | *n = p; | |
483 | else | |
484 | ret = -EBUSY; | |
485 | ||
486 | if (ret < 0) { | |
f33ff110 | 487 | printk("register_blkdev: cannot get major %u for %s\n", |
1da177e4 LT |
488 | major, name); |
489 | kfree(p); | |
490 | } | |
491 | out: | |
edfaa7c3 | 492 | mutex_unlock(&block_class_lock); |
1da177e4 LT |
493 | return ret; |
494 | } | |
495 | ||
496 | EXPORT_SYMBOL(register_blkdev); | |
497 | ||
f4480240 | 498 | void unregister_blkdev(unsigned int major, const char *name) |
1da177e4 LT |
499 | { |
500 | struct blk_major_name **n; | |
501 | struct blk_major_name *p = NULL; | |
502 | int index = major_to_index(major); | |
1da177e4 | 503 | |
edfaa7c3 | 504 | mutex_lock(&block_class_lock); |
1da177e4 LT |
505 | for (n = &major_names[index]; *n; n = &(*n)->next) |
506 | if ((*n)->major == major) | |
507 | break; | |
294462a5 AM |
508 | if (!*n || strcmp((*n)->name, name)) { |
509 | WARN_ON(1); | |
294462a5 | 510 | } else { |
1da177e4 LT |
511 | p = *n; |
512 | *n = p->next; | |
513 | } | |
edfaa7c3 | 514 | mutex_unlock(&block_class_lock); |
1da177e4 | 515 | kfree(p); |
1da177e4 LT |
516 | } |
517 | ||
518 | EXPORT_SYMBOL(unregister_blkdev); | |
519 | ||
520 | static struct kobj_map *bdev_map; | |
521 | ||
870d6656 TH |
522 | /** |
523 | * blk_mangle_minor - scatter minor numbers apart | |
524 | * @minor: minor number to mangle | |
525 | * | |
526 | * Scatter consecutively allocated @minor number apart if MANGLE_DEVT | |
527 | * is enabled. Mangling twice gives the original value. | |
528 | * | |
529 | * RETURNS: | |
530 | * Mangled value. | |
531 | * | |
532 | * CONTEXT: | |
533 | * Don't care. | |
534 | */ | |
535 | static int blk_mangle_minor(int minor) | |
536 | { | |
537 | #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT | |
538 | int i; | |
539 | ||
540 | for (i = 0; i < MINORBITS / 2; i++) { | |
541 | int low = minor & (1 << i); | |
542 | int high = minor & (1 << (MINORBITS - 1 - i)); | |
543 | int distance = MINORBITS - 1 - 2 * i; | |
544 | ||
545 | minor ^= low | high; /* clear both bits */ | |
546 | low <<= distance; /* swap the positions */ | |
547 | high >>= distance; | |
548 | minor |= low | high; /* and set */ | |
549 | } | |
550 | #endif | |
551 | return minor; | |
552 | } | |
553 | ||
bcce3de1 TH |
554 | /** |
555 | * blk_alloc_devt - allocate a dev_t for a partition | |
556 | * @part: partition to allocate dev_t for | |
bcce3de1 TH |
557 | * @devt: out parameter for resulting dev_t |
558 | * | |
559 | * Allocate a dev_t for block device. | |
560 | * | |
561 | * RETURNS: | |
562 | * 0 on success, allocated dev_t is returned in *@devt. -errno on | |
563 | * failure. | |
564 | * | |
565 | * CONTEXT: | |
566 | * Might sleep. | |
567 | */ | |
568 | int blk_alloc_devt(struct hd_struct *part, dev_t *devt) | |
569 | { | |
570 | struct gendisk *disk = part_to_disk(part); | |
bab998d6 | 571 | int idx; |
bcce3de1 TH |
572 | |
573 | /* in consecutive minor range? */ | |
574 | if (part->partno < disk->minors) { | |
575 | *devt = MKDEV(disk->major, disk->first_minor + part->partno); | |
576 | return 0; | |
577 | } | |
578 | ||
579 | /* allocate ext devt */ | |
2da78092 KB |
580 | idr_preload(GFP_KERNEL); |
581 | ||
4d66e5e9 | 582 | spin_lock_bh(&ext_devt_lock); |
2da78092 | 583 | idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_NOWAIT); |
4d66e5e9 | 584 | spin_unlock_bh(&ext_devt_lock); |
2da78092 KB |
585 | |
586 | idr_preload_end(); | |
bab998d6 TH |
587 | if (idx < 0) |
588 | return idx == -ENOSPC ? -EBUSY : idx; | |
bcce3de1 | 589 | |
870d6656 | 590 | *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx)); |
bcce3de1 TH |
591 | return 0; |
592 | } | |
593 | ||
594 | /** | |
595 | * blk_free_devt - free a dev_t | |
596 | * @devt: dev_t to free | |
597 | * | |
598 | * Free @devt which was allocated using blk_alloc_devt(). | |
599 | * | |
600 | * CONTEXT: | |
601 | * Might sleep. | |
602 | */ | |
603 | void blk_free_devt(dev_t devt) | |
604 | { | |
bcce3de1 TH |
605 | if (devt == MKDEV(0, 0)) |
606 | return; | |
607 | ||
608 | if (MAJOR(devt) == BLOCK_EXT_MAJOR) { | |
4d66e5e9 | 609 | spin_lock_bh(&ext_devt_lock); |
870d6656 | 610 | idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); |
4d66e5e9 | 611 | spin_unlock_bh(&ext_devt_lock); |
bcce3de1 TH |
612 | } |
613 | } | |
614 | ||
33c826ef BVA |
615 | /* |
616 | * We invalidate devt by assigning NULL pointer for devt in idr. | |
6fcc44d1 YY |
617 | */ |
618 | void blk_invalidate_devt(dev_t devt) | |
619 | { | |
620 | if (MAJOR(devt) == BLOCK_EXT_MAJOR) { | |
621 | spin_lock_bh(&ext_devt_lock); | |
622 | idr_replace(&ext_devt_idr, NULL, blk_mangle_minor(MINOR(devt))); | |
623 | spin_unlock_bh(&ext_devt_lock); | |
624 | } | |
625 | } | |
626 | ||
1f014290 TH |
627 | static char *bdevt_str(dev_t devt, char *buf) |
628 | { | |
629 | if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) { | |
630 | char tbuf[BDEVT_SIZE]; | |
631 | snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt)); | |
632 | snprintf(buf, BDEVT_SIZE, "%-9s", tbuf); | |
633 | } else | |
634 | snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt)); | |
635 | ||
636 | return buf; | |
637 | } | |
638 | ||
1da177e4 LT |
639 | /* |
640 | * Register device numbers dev..(dev+range-1) | |
641 | * range must be nonzero | |
642 | * The hash chain is sorted on range, so that subranges can override. | |
643 | */ | |
edfaa7c3 | 644 | void blk_register_region(dev_t devt, unsigned long range, struct module *module, |
1da177e4 LT |
645 | struct kobject *(*probe)(dev_t, int *, void *), |
646 | int (*lock)(dev_t, void *), void *data) | |
647 | { | |
edfaa7c3 | 648 | kobj_map(bdev_map, devt, range, module, probe, lock, data); |
1da177e4 LT |
649 | } |
650 | ||
651 | EXPORT_SYMBOL(blk_register_region); | |
652 | ||
edfaa7c3 | 653 | void blk_unregister_region(dev_t devt, unsigned long range) |
1da177e4 | 654 | { |
edfaa7c3 | 655 | kobj_unmap(bdev_map, devt, range); |
1da177e4 LT |
656 | } |
657 | ||
658 | EXPORT_SYMBOL(blk_unregister_region); | |
659 | ||
cf771cb5 | 660 | static struct kobject *exact_match(dev_t devt, int *partno, void *data) |
1da177e4 LT |
661 | { |
662 | struct gendisk *p = data; | |
edfaa7c3 | 663 | |
ed9e1982 | 664 | return &disk_to_dev(p)->kobj; |
1da177e4 LT |
665 | } |
666 | ||
edfaa7c3 | 667 | static int exact_lock(dev_t devt, void *data) |
1da177e4 LT |
668 | { |
669 | struct gendisk *p = data; | |
670 | ||
3079c22e | 671 | if (!get_disk_and_module(p)) |
1da177e4 LT |
672 | return -1; |
673 | return 0; | |
674 | } | |
675 | ||
fef912bf HR |
676 | static void register_disk(struct device *parent, struct gendisk *disk, |
677 | const struct attribute_group **groups) | |
d2bf1b67 TH |
678 | { |
679 | struct device *ddev = disk_to_dev(disk); | |
680 | struct block_device *bdev; | |
681 | struct disk_part_iter piter; | |
682 | struct hd_struct *part; | |
683 | int err; | |
684 | ||
e63a46be | 685 | ddev->parent = parent; |
d2bf1b67 | 686 | |
ffc8b308 | 687 | dev_set_name(ddev, "%s", disk->disk_name); |
d2bf1b67 TH |
688 | |
689 | /* delay uevents, until we scanned partition table */ | |
690 | dev_set_uevent_suppress(ddev, 1); | |
691 | ||
fef912bf HR |
692 | if (groups) { |
693 | WARN_ON(ddev->groups); | |
694 | ddev->groups = groups; | |
695 | } | |
d2bf1b67 TH |
696 | if (device_add(ddev)) |
697 | return; | |
698 | if (!sysfs_deprecated) { | |
699 | err = sysfs_create_link(block_depr, &ddev->kobj, | |
700 | kobject_name(&ddev->kobj)); | |
701 | if (err) { | |
702 | device_del(ddev); | |
703 | return; | |
704 | } | |
705 | } | |
25e823c8 ML |
706 | |
707 | /* | |
708 | * avoid probable deadlock caused by allocating memory with | |
709 | * GFP_KERNEL in runtime_resume callback of its all ancestor | |
710 | * devices | |
711 | */ | |
712 | pm_runtime_set_memalloc_noio(ddev, true); | |
713 | ||
d2bf1b67 TH |
714 | disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj); |
715 | disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj); | |
716 | ||
8ddcd653 CH |
717 | if (disk->flags & GENHD_FL_HIDDEN) { |
718 | dev_set_uevent_suppress(ddev, 0); | |
719 | return; | |
720 | } | |
721 | ||
d2bf1b67 | 722 | /* No minors to use for partitions */ |
d27769ec | 723 | if (!disk_part_scan_enabled(disk)) |
d2bf1b67 TH |
724 | goto exit; |
725 | ||
726 | /* No such device (e.g., media were just removed) */ | |
727 | if (!get_capacity(disk)) | |
728 | goto exit; | |
729 | ||
730 | bdev = bdget_disk(disk, 0); | |
731 | if (!bdev) | |
732 | goto exit; | |
733 | ||
f4ad06f2 | 734 | set_bit(BDEV_NEED_PART_SCAN, &bdev->bd_flags); |
d2bf1b67 TH |
735 | err = blkdev_get(bdev, FMODE_READ, NULL); |
736 | if (err < 0) | |
737 | goto exit; | |
738 | blkdev_put(bdev, FMODE_READ); | |
739 | ||
740 | exit: | |
741 | /* announce disk after possible partitions are created */ | |
742 | dev_set_uevent_suppress(ddev, 0); | |
743 | kobject_uevent(&ddev->kobj, KOBJ_ADD); | |
744 | ||
745 | /* announce possible partitions */ | |
746 | disk_part_iter_init(&piter, disk, 0); | |
747 | while ((part = disk_part_iter_next(&piter))) | |
748 | kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD); | |
749 | disk_part_iter_exit(&piter); | |
8ddcd653 | 750 | |
4d7c1d3f | 751 | if (disk->queue->backing_dev_info->dev) { |
752 | err = sysfs_create_link(&ddev->kobj, | |
753 | &disk->queue->backing_dev_info->dev->kobj, | |
754 | "bdi"); | |
755 | WARN_ON(err); | |
756 | } | |
d2bf1b67 TH |
757 | } |
758 | ||
1da177e4 | 759 | /** |
fa70d2e2 | 760 | * __device_add_disk - add disk information to kernel list |
e63a46be | 761 | * @parent: parent device for the disk |
1da177e4 | 762 | * @disk: per-device partitioning information |
fef912bf | 763 | * @groups: Additional per-device sysfs groups |
fa70d2e2 | 764 | * @register_queue: register the queue if set to true |
1da177e4 LT |
765 | * |
766 | * This function registers the partitioning information in @disk | |
767 | * with the kernel. | |
3e1a7ff8 TH |
768 | * |
769 | * FIXME: error handling | |
1da177e4 | 770 | */ |
fa70d2e2 | 771 | static void __device_add_disk(struct device *parent, struct gendisk *disk, |
fef912bf | 772 | const struct attribute_group **groups, |
fa70d2e2 | 773 | bool register_queue) |
1da177e4 | 774 | { |
3e1a7ff8 | 775 | dev_t devt; |
6ffeea77 | 776 | int retval; |
cf0ca9fe | 777 | |
737eb78e DLM |
778 | /* |
779 | * The disk queue should now be all set with enough information about | |
780 | * the device for the elevator code to pick an adequate default | |
781 | * elevator if one is needed, that is, for devices requesting queue | |
782 | * registration. | |
783 | */ | |
784 | if (register_queue) | |
785 | elevator_init_mq(disk->queue); | |
786 | ||
3e1a7ff8 TH |
787 | /* minors == 0 indicates to use ext devt from part0 and should |
788 | * be accompanied with EXT_DEVT flag. Make sure all | |
789 | * parameters make sense. | |
790 | */ | |
791 | WARN_ON(disk->minors && !(disk->major || disk->first_minor)); | |
8ddcd653 CH |
792 | WARN_ON(!disk->minors && |
793 | !(disk->flags & (GENHD_FL_EXT_DEVT | GENHD_FL_HIDDEN))); | |
3e1a7ff8 | 794 | |
1da177e4 | 795 | disk->flags |= GENHD_FL_UP; |
3e1a7ff8 TH |
796 | |
797 | retval = blk_alloc_devt(&disk->part0, &devt); | |
798 | if (retval) { | |
799 | WARN_ON(1); | |
800 | return; | |
801 | } | |
3e1a7ff8 TH |
802 | disk->major = MAJOR(devt); |
803 | disk->first_minor = MINOR(devt); | |
804 | ||
9f53d2fe SG |
805 | disk_alloc_events(disk); |
806 | ||
8ddcd653 CH |
807 | if (disk->flags & GENHD_FL_HIDDEN) { |
808 | /* | |
809 | * Don't let hidden disks show up in /proc/partitions, | |
810 | * and don't bother scanning for partitions either. | |
811 | */ | |
812 | disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO; | |
813 | disk->flags |= GENHD_FL_NO_PART_SCAN; | |
814 | } else { | |
3c5d202b CH |
815 | struct backing_dev_info *bdi = disk->queue->backing_dev_info; |
816 | struct device *dev = disk_to_dev(disk); | |
3a92168b | 817 | int ret; |
818 | ||
8ddcd653 | 819 | /* Register BDI before referencing it from bdev */ |
3c5d202b CH |
820 | dev->devt = devt; |
821 | ret = bdi_register(bdi, "%u:%u", MAJOR(devt), MINOR(devt)); | |
3a92168b | 822 | WARN_ON(ret); |
3c5d202b | 823 | bdi_set_owner(bdi, dev); |
8ddcd653 CH |
824 | blk_register_region(disk_devt(disk), disk->minors, NULL, |
825 | exact_match, exact_lock, disk); | |
826 | } | |
fef912bf | 827 | register_disk(parent, disk, groups); |
fa70d2e2 MS |
828 | if (register_queue) |
829 | blk_register_queue(disk); | |
cf0ca9fe | 830 | |
523e1d39 TH |
831 | /* |
832 | * Take an extra ref on queue which will be put on disk_release() | |
833 | * so that it sticks around as long as @disk is there. | |
834 | */ | |
09ac46c4 | 835 | WARN_ON_ONCE(!blk_get_queue(disk->queue)); |
523e1d39 | 836 | |
77ea887e | 837 | disk_add_events(disk); |
25520d55 | 838 | blk_integrity_add(disk); |
1da177e4 | 839 | } |
fa70d2e2 | 840 | |
fef912bf HR |
841 | void device_add_disk(struct device *parent, struct gendisk *disk, |
842 | const struct attribute_group **groups) | |
843 | ||
fa70d2e2 | 844 | { |
fef912bf | 845 | __device_add_disk(parent, disk, groups, true); |
fa70d2e2 | 846 | } |
e63a46be | 847 | EXPORT_SYMBOL(device_add_disk); |
1da177e4 | 848 | |
fa70d2e2 MS |
849 | void device_add_disk_no_queue_reg(struct device *parent, struct gendisk *disk) |
850 | { | |
fef912bf | 851 | __device_add_disk(parent, disk, NULL, false); |
fa70d2e2 MS |
852 | } |
853 | EXPORT_SYMBOL(device_add_disk_no_queue_reg); | |
854 | ||
02d33b67 CH |
855 | static void invalidate_partition(struct gendisk *disk, int partno) |
856 | { | |
857 | struct block_device *bdev; | |
858 | ||
859 | bdev = bdget_disk(disk, partno); | |
860 | if (!bdev) | |
861 | return; | |
862 | ||
863 | fsync_bdev(bdev); | |
864 | __invalidate_device(bdev, true); | |
9bc5c397 CH |
865 | |
866 | /* | |
867 | * Unhash the bdev inode for this device so that it gets evicted as soon | |
868 | * as last inode reference is dropped. | |
869 | */ | |
870 | remove_inode_hash(bdev->bd_inode); | |
02d33b67 CH |
871 | bdput(bdev); |
872 | } | |
873 | ||
b5bd357c LC |
874 | /** |
875 | * del_gendisk - remove the gendisk | |
876 | * @disk: the struct gendisk to remove | |
877 | * | |
878 | * Removes the gendisk and all its associated resources. This deletes the | |
879 | * partitions associated with the gendisk, and unregisters the associated | |
880 | * request_queue. | |
881 | * | |
882 | * This is the counter to the respective __device_add_disk() call. | |
883 | * | |
884 | * The final removal of the struct gendisk happens when its refcount reaches 0 | |
885 | * with put_disk(), which should be called after del_gendisk(), if | |
886 | * __device_add_disk() was used. | |
e8c7d14a LC |
887 | * |
888 | * Drivers exist which depend on the release of the gendisk to be synchronous, | |
889 | * it should not be deferred. | |
890 | * | |
891 | * Context: can sleep | |
b5bd357c | 892 | */ |
d2bf1b67 | 893 | void del_gendisk(struct gendisk *disk) |
1da177e4 | 894 | { |
d2bf1b67 TH |
895 | struct disk_part_iter piter; |
896 | struct hd_struct *part; | |
897 | ||
e8c7d14a LC |
898 | might_sleep(); |
899 | ||
25520d55 | 900 | blk_integrity_del(disk); |
77ea887e TH |
901 | disk_del_events(disk); |
902 | ||
56c0908c JK |
903 | /* |
904 | * Block lookups of the disk until all bdevs are unhashed and the | |
905 | * disk is marked as dead (GENHD_FL_UP cleared). | |
906 | */ | |
907 | down_write(&disk->lookup_sem); | |
d2bf1b67 TH |
908 | /* invalidate stuff */ |
909 | disk_part_iter_init(&piter, disk, | |
910 | DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE); | |
911 | while ((part = disk_part_iter_next(&piter))) { | |
912 | invalidate_partition(disk, part->partno); | |
8328eb28 | 913 | delete_partition(part); |
d2bf1b67 TH |
914 | } |
915 | disk_part_iter_exit(&piter); | |
916 | ||
917 | invalidate_partition(disk, 0); | |
d2bf1b67 TH |
918 | set_capacity(disk, 0); |
919 | disk->flags &= ~GENHD_FL_UP; | |
56c0908c | 920 | up_write(&disk->lookup_sem); |
d2bf1b67 | 921 | |
8ddcd653 CH |
922 | if (!(disk->flags & GENHD_FL_HIDDEN)) |
923 | sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi"); | |
90f16fdd JK |
924 | if (disk->queue) { |
925 | /* | |
926 | * Unregister bdi before releasing device numbers (as they can | |
927 | * get reused and we'd get clashes in sysfs). | |
928 | */ | |
bc8d062c MS |
929 | if (!(disk->flags & GENHD_FL_HIDDEN)) |
930 | bdi_unregister(disk->queue->backing_dev_info); | |
90f16fdd JK |
931 | blk_unregister_queue(disk); |
932 | } else { | |
933 | WARN_ON(1); | |
934 | } | |
d2bf1b67 | 935 | |
17eac099 | 936 | if (!(disk->flags & GENHD_FL_HIDDEN)) |
8ddcd653 | 937 | blk_unregister_region(disk_devt(disk), disk->minors); |
6fcc44d1 YY |
938 | /* |
939 | * Remove gendisk pointer from idr so that it cannot be looked up | |
940 | * while RCU period before freeing gendisk is running to prevent | |
941 | * use-after-free issues. Note that the device number stays | |
942 | * "in-use" until we really free the gendisk. | |
943 | */ | |
944 | blk_invalidate_devt(disk_devt(disk)); | |
d2bf1b67 TH |
945 | |
946 | kobject_put(disk->part0.holder_dir); | |
947 | kobject_put(disk->slave_dir); | |
d2bf1b67 TH |
948 | |
949 | part_stat_set_all(&disk->part0, 0); | |
950 | disk->part0.stamp = 0; | |
d2bf1b67 TH |
951 | if (!sysfs_deprecated) |
952 | sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk))); | |
25e823c8 | 953 | pm_runtime_set_memalloc_noio(disk_to_dev(disk), false); |
d2bf1b67 | 954 | device_del(disk_to_dev(disk)); |
1da177e4 | 955 | } |
d2bf1b67 | 956 | EXPORT_SYMBOL(del_gendisk); |
1da177e4 | 957 | |
99e6608c VV |
958 | /* sysfs access to bad-blocks list. */ |
959 | static ssize_t disk_badblocks_show(struct device *dev, | |
960 | struct device_attribute *attr, | |
961 | char *page) | |
962 | { | |
963 | struct gendisk *disk = dev_to_disk(dev); | |
964 | ||
965 | if (!disk->bb) | |
966 | return sprintf(page, "\n"); | |
967 | ||
968 | return badblocks_show(disk->bb, page, 0); | |
969 | } | |
970 | ||
971 | static ssize_t disk_badblocks_store(struct device *dev, | |
972 | struct device_attribute *attr, | |
973 | const char *page, size_t len) | |
974 | { | |
975 | struct gendisk *disk = dev_to_disk(dev); | |
976 | ||
977 | if (!disk->bb) | |
978 | return -ENXIO; | |
979 | ||
980 | return badblocks_store(disk->bb, page, len, 0); | |
981 | } | |
982 | ||
1da177e4 LT |
983 | /** |
984 | * get_gendisk - get partitioning information for a given device | |
710027a4 | 985 | * @devt: device to get partitioning information for |
496aa8a9 | 986 | * @partno: returned partition index |
1da177e4 LT |
987 | * |
988 | * This function gets the structure containing partitioning | |
710027a4 | 989 | * information for the given device @devt. |
763b5892 LC |
990 | * |
991 | * Context: can sleep | |
1da177e4 | 992 | */ |
cf771cb5 | 993 | struct gendisk *get_gendisk(dev_t devt, int *partno) |
1da177e4 | 994 | { |
bcce3de1 TH |
995 | struct gendisk *disk = NULL; |
996 | ||
763b5892 LC |
997 | might_sleep(); |
998 | ||
bcce3de1 TH |
999 | if (MAJOR(devt) != BLOCK_EXT_MAJOR) { |
1000 | struct kobject *kobj; | |
1001 | ||
1002 | kobj = kobj_lookup(bdev_map, devt, partno); | |
1003 | if (kobj) | |
1004 | disk = dev_to_disk(kobj_to_dev(kobj)); | |
1005 | } else { | |
1006 | struct hd_struct *part; | |
1007 | ||
4d66e5e9 | 1008 | spin_lock_bh(&ext_devt_lock); |
870d6656 | 1009 | part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); |
3079c22e | 1010 | if (part && get_disk_and_module(part_to_disk(part))) { |
bcce3de1 TH |
1011 | *partno = part->partno; |
1012 | disk = part_to_disk(part); | |
1013 | } | |
4d66e5e9 | 1014 | spin_unlock_bh(&ext_devt_lock); |
bcce3de1 | 1015 | } |
edfaa7c3 | 1016 | |
56c0908c JK |
1017 | if (!disk) |
1018 | return NULL; | |
1019 | ||
1020 | /* | |
1021 | * Synchronize with del_gendisk() to not return disk that is being | |
1022 | * destroyed. | |
1023 | */ | |
1024 | down_read(&disk->lookup_sem); | |
1025 | if (unlikely((disk->flags & GENHD_FL_HIDDEN) || | |
1026 | !(disk->flags & GENHD_FL_UP))) { | |
1027 | up_read(&disk->lookup_sem); | |
9df6c299 | 1028 | put_disk_and_module(disk); |
8ddcd653 | 1029 | disk = NULL; |
56c0908c JK |
1030 | } else { |
1031 | up_read(&disk->lookup_sem); | |
8ddcd653 | 1032 | } |
bcce3de1 | 1033 | return disk; |
1da177e4 LT |
1034 | } |
1035 | ||
f331c029 TH |
1036 | /** |
1037 | * bdget_disk - do bdget() by gendisk and partition number | |
1038 | * @disk: gendisk of interest | |
1039 | * @partno: partition number | |
1040 | * | |
1041 | * Find partition @partno from @disk, do bdget() on it. | |
1042 | * | |
1043 | * CONTEXT: | |
1044 | * Don't care. | |
1045 | * | |
1046 | * RETURNS: | |
1047 | * Resulting block_device on success, NULL on failure. | |
1048 | */ | |
aeb3d3a8 | 1049 | struct block_device *bdget_disk(struct gendisk *disk, int partno) |
f331c029 | 1050 | { |
548b10eb TH |
1051 | struct hd_struct *part; |
1052 | struct block_device *bdev = NULL; | |
f331c029 | 1053 | |
548b10eb | 1054 | part = disk_get_part(disk, partno); |
2bbedcb4 | 1055 | if (part) |
548b10eb TH |
1056 | bdev = bdget(part_devt(part)); |
1057 | disk_put_part(part); | |
f331c029 | 1058 | |
548b10eb | 1059 | return bdev; |
f331c029 TH |
1060 | } |
1061 | EXPORT_SYMBOL(bdget_disk); | |
1062 | ||
5c6f35c5 GKH |
1063 | /* |
1064 | * print a full list of all partitions - intended for places where the root | |
1065 | * filesystem can't be mounted and thus to give the victim some idea of what | |
1066 | * went wrong | |
1067 | */ | |
1068 | void __init printk_all_partitions(void) | |
1069 | { | |
def4e38d TH |
1070 | struct class_dev_iter iter; |
1071 | struct device *dev; | |
1072 | ||
1073 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); | |
1074 | while ((dev = class_dev_iter_next(&iter))) { | |
1075 | struct gendisk *disk = dev_to_disk(dev); | |
e71bf0d0 TH |
1076 | struct disk_part_iter piter; |
1077 | struct hd_struct *part; | |
1f014290 TH |
1078 | char name_buf[BDEVNAME_SIZE]; |
1079 | char devt_buf[BDEVT_SIZE]; | |
def4e38d TH |
1080 | |
1081 | /* | |
1082 | * Don't show empty devices or things that have been | |
25985edc | 1083 | * suppressed |
def4e38d TH |
1084 | */ |
1085 | if (get_capacity(disk) == 0 || | |
1086 | (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)) | |
1087 | continue; | |
1088 | ||
1089 | /* | |
1090 | * Note, unlike /proc/partitions, I am showing the | |
1091 | * numbers in hex - the same format as the root= | |
1092 | * option takes. | |
1093 | */ | |
074a7aca TH |
1094 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0); |
1095 | while ((part = disk_part_iter_next(&piter))) { | |
1096 | bool is_part0 = part == &disk->part0; | |
def4e38d | 1097 | |
b5af921e | 1098 | printk("%s%s %10llu %s %s", is_part0 ? "" : " ", |
1f014290 | 1099 | bdevt_str(part_devt(part), devt_buf), |
c83f6bf9 VG |
1100 | (unsigned long long)part_nr_sects_read(part) >> 1 |
1101 | , disk_name(disk, part->partno, name_buf), | |
1ad7e899 | 1102 | part->info ? part->info->uuid : ""); |
074a7aca | 1103 | if (is_part0) { |
52c44d93 | 1104 | if (dev->parent && dev->parent->driver) |
074a7aca | 1105 | printk(" driver: %s\n", |
52c44d93 | 1106 | dev->parent->driver->name); |
074a7aca TH |
1107 | else |
1108 | printk(" (driver?)\n"); | |
1109 | } else | |
1110 | printk("\n"); | |
1111 | } | |
e71bf0d0 | 1112 | disk_part_iter_exit(&piter); |
def4e38d TH |
1113 | } |
1114 | class_dev_iter_exit(&iter); | |
dd2a345f DG |
1115 | } |
1116 | ||
1da177e4 LT |
1117 | #ifdef CONFIG_PROC_FS |
1118 | /* iterator */ | |
def4e38d | 1119 | static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos) |
68c4d4a7 | 1120 | { |
def4e38d TH |
1121 | loff_t skip = *pos; |
1122 | struct class_dev_iter *iter; | |
1123 | struct device *dev; | |
68c4d4a7 | 1124 | |
aeb3d3a8 | 1125 | iter = kmalloc(sizeof(*iter), GFP_KERNEL); |
def4e38d TH |
1126 | if (!iter) |
1127 | return ERR_PTR(-ENOMEM); | |
1128 | ||
1129 | seqf->private = iter; | |
1130 | class_dev_iter_init(iter, &block_class, NULL, &disk_type); | |
1131 | do { | |
1132 | dev = class_dev_iter_next(iter); | |
1133 | if (!dev) | |
1134 | return NULL; | |
1135 | } while (skip--); | |
1136 | ||
1137 | return dev_to_disk(dev); | |
68c4d4a7 GKH |
1138 | } |
1139 | ||
def4e38d | 1140 | static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos) |
1da177e4 | 1141 | { |
edfaa7c3 | 1142 | struct device *dev; |
1da177e4 | 1143 | |
def4e38d TH |
1144 | (*pos)++; |
1145 | dev = class_dev_iter_next(seqf->private); | |
2ac3cee5 | 1146 | if (dev) |
68c4d4a7 | 1147 | return dev_to_disk(dev); |
2ac3cee5 | 1148 | |
1da177e4 LT |
1149 | return NULL; |
1150 | } | |
1151 | ||
def4e38d | 1152 | static void disk_seqf_stop(struct seq_file *seqf, void *v) |
27f30251 | 1153 | { |
def4e38d | 1154 | struct class_dev_iter *iter = seqf->private; |
27f30251 | 1155 | |
def4e38d TH |
1156 | /* stop is called even after start failed :-( */ |
1157 | if (iter) { | |
1158 | class_dev_iter_exit(iter); | |
1159 | kfree(iter); | |
77da1605 | 1160 | seqf->private = NULL; |
5c0ef6d0 | 1161 | } |
1da177e4 LT |
1162 | } |
1163 | ||
def4e38d | 1164 | static void *show_partition_start(struct seq_file *seqf, loff_t *pos) |
1da177e4 | 1165 | { |
06768067 | 1166 | void *p; |
def4e38d TH |
1167 | |
1168 | p = disk_seqf_start(seqf, pos); | |
b9f985b6 | 1169 | if (!IS_ERR_OR_NULL(p) && !*pos) |
def4e38d TH |
1170 | seq_puts(seqf, "major minor #blocks name\n\n"); |
1171 | return p; | |
1da177e4 LT |
1172 | } |
1173 | ||
cf771cb5 | 1174 | static int show_partition(struct seq_file *seqf, void *v) |
1da177e4 LT |
1175 | { |
1176 | struct gendisk *sgp = v; | |
e71bf0d0 TH |
1177 | struct disk_part_iter piter; |
1178 | struct hd_struct *part; | |
1da177e4 LT |
1179 | char buf[BDEVNAME_SIZE]; |
1180 | ||
1da177e4 | 1181 | /* Don't show non-partitionable removeable devices or empty devices */ |
d27769ec | 1182 | if (!get_capacity(sgp) || (!disk_max_parts(sgp) && |
f331c029 | 1183 | (sgp->flags & GENHD_FL_REMOVABLE))) |
1da177e4 LT |
1184 | return 0; |
1185 | if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO) | |
1186 | return 0; | |
1187 | ||
1188 | /* show the full disk and all non-0 size partitions of it */ | |
074a7aca | 1189 | disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0); |
e71bf0d0 | 1190 | while ((part = disk_part_iter_next(&piter))) |
1f014290 | 1191 | seq_printf(seqf, "%4d %7d %10llu %s\n", |
f331c029 | 1192 | MAJOR(part_devt(part)), MINOR(part_devt(part)), |
c83f6bf9 | 1193 | (unsigned long long)part_nr_sects_read(part) >> 1, |
f331c029 | 1194 | disk_name(sgp, part->partno, buf)); |
e71bf0d0 | 1195 | disk_part_iter_exit(&piter); |
1da177e4 LT |
1196 | |
1197 | return 0; | |
1198 | } | |
1199 | ||
f500975a | 1200 | static const struct seq_operations partitions_op = { |
def4e38d TH |
1201 | .start = show_partition_start, |
1202 | .next = disk_seqf_next, | |
1203 | .stop = disk_seqf_stop, | |
edfaa7c3 | 1204 | .show = show_partition |
1da177e4 LT |
1205 | }; |
1206 | #endif | |
1207 | ||
1208 | ||
cf771cb5 | 1209 | static struct kobject *base_probe(dev_t devt, int *partno, void *data) |
1da177e4 | 1210 | { |
edfaa7c3 | 1211 | if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0) |
1da177e4 | 1212 | /* Make old-style 2.4 aliases work */ |
edfaa7c3 | 1213 | request_module("block-major-%d", MAJOR(devt)); |
1da177e4 LT |
1214 | return NULL; |
1215 | } | |
1216 | ||
1217 | static int __init genhd_device_init(void) | |
1218 | { | |
e105b8bf DW |
1219 | int error; |
1220 | ||
1221 | block_class.dev_kobj = sysfs_dev_block_kobj; | |
1222 | error = class_register(&block_class); | |
ee27a558 RM |
1223 | if (unlikely(error)) |
1224 | return error; | |
edfaa7c3 | 1225 | bdev_map = kobj_map_init(base_probe, &block_class_lock); |
1da177e4 | 1226 | blk_dev_init(); |
edfaa7c3 | 1227 | |
561ec68e ZY |
1228 | register_blkdev(BLOCK_EXT_MAJOR, "blkext"); |
1229 | ||
edfaa7c3 | 1230 | /* create top-level block dir */ |
e52eec13 AK |
1231 | if (!sysfs_deprecated) |
1232 | block_depr = kobject_create_and_add("block", NULL); | |
830d3cfb | 1233 | return 0; |
1da177e4 LT |
1234 | } |
1235 | ||
1236 | subsys_initcall(genhd_device_init); | |
1237 | ||
edfaa7c3 KS |
1238 | static ssize_t disk_range_show(struct device *dev, |
1239 | struct device_attribute *attr, char *buf) | |
1da177e4 | 1240 | { |
edfaa7c3 | 1241 | struct gendisk *disk = dev_to_disk(dev); |
1da177e4 | 1242 | |
edfaa7c3 | 1243 | return sprintf(buf, "%d\n", disk->minors); |
1da177e4 LT |
1244 | } |
1245 | ||
1f014290 TH |
1246 | static ssize_t disk_ext_range_show(struct device *dev, |
1247 | struct device_attribute *attr, char *buf) | |
1248 | { | |
1249 | struct gendisk *disk = dev_to_disk(dev); | |
1250 | ||
b5d0b9df | 1251 | return sprintf(buf, "%d\n", disk_max_parts(disk)); |
1f014290 TH |
1252 | } |
1253 | ||
edfaa7c3 KS |
1254 | static ssize_t disk_removable_show(struct device *dev, |
1255 | struct device_attribute *attr, char *buf) | |
a7fd6706 | 1256 | { |
edfaa7c3 | 1257 | struct gendisk *disk = dev_to_disk(dev); |
a7fd6706 | 1258 | |
edfaa7c3 KS |
1259 | return sprintf(buf, "%d\n", |
1260 | (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0)); | |
a7fd6706 KS |
1261 | } |
1262 | ||
8ddcd653 CH |
1263 | static ssize_t disk_hidden_show(struct device *dev, |
1264 | struct device_attribute *attr, char *buf) | |
1265 | { | |
1266 | struct gendisk *disk = dev_to_disk(dev); | |
1267 | ||
1268 | return sprintf(buf, "%d\n", | |
1269 | (disk->flags & GENHD_FL_HIDDEN ? 1 : 0)); | |
1270 | } | |
1271 | ||
1c9ce527 KS |
1272 | static ssize_t disk_ro_show(struct device *dev, |
1273 | struct device_attribute *attr, char *buf) | |
1274 | { | |
1275 | struct gendisk *disk = dev_to_disk(dev); | |
1276 | ||
b7db9956 | 1277 | return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0); |
1c9ce527 KS |
1278 | } |
1279 | ||
3ad5cee5 CH |
1280 | ssize_t part_size_show(struct device *dev, |
1281 | struct device_attribute *attr, char *buf) | |
1282 | { | |
1283 | struct hd_struct *p = dev_to_part(dev); | |
1284 | ||
1285 | return sprintf(buf, "%llu\n", | |
1286 | (unsigned long long)part_nr_sects_read(p)); | |
1287 | } | |
1288 | ||
1289 | ssize_t part_stat_show(struct device *dev, | |
1290 | struct device_attribute *attr, char *buf) | |
1291 | { | |
1292 | struct hd_struct *p = dev_to_part(dev); | |
1293 | struct request_queue *q = part_to_disk(p)->queue; | |
ea18e0f0 | 1294 | struct disk_stats stat; |
3ad5cee5 CH |
1295 | unsigned int inflight; |
1296 | ||
ea18e0f0 | 1297 | part_stat_read_all(p, &stat); |
b2f609e1 CH |
1298 | if (queue_is_mq(q)) |
1299 | inflight = blk_mq_in_flight(q, p); | |
1300 | else | |
1f06959b | 1301 | inflight = part_in_flight(p); |
ea18e0f0 | 1302 | |
3ad5cee5 CH |
1303 | return sprintf(buf, |
1304 | "%8lu %8lu %8llu %8u " | |
1305 | "%8lu %8lu %8llu %8u " | |
1306 | "%8u %8u %8u " | |
1307 | "%8lu %8lu %8llu %8u " | |
1308 | "%8lu %8u" | |
1309 | "\n", | |
ea18e0f0 KK |
1310 | stat.ios[STAT_READ], |
1311 | stat.merges[STAT_READ], | |
1312 | (unsigned long long)stat.sectors[STAT_READ], | |
1313 | (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC), | |
1314 | stat.ios[STAT_WRITE], | |
1315 | stat.merges[STAT_WRITE], | |
1316 | (unsigned long long)stat.sectors[STAT_WRITE], | |
1317 | (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC), | |
3ad5cee5 | 1318 | inflight, |
ea18e0f0 | 1319 | jiffies_to_msecs(stat.io_ticks), |
8cd5b8fc KK |
1320 | (unsigned int)div_u64(stat.nsecs[STAT_READ] + |
1321 | stat.nsecs[STAT_WRITE] + | |
1322 | stat.nsecs[STAT_DISCARD] + | |
1323 | stat.nsecs[STAT_FLUSH], | |
1324 | NSEC_PER_MSEC), | |
ea18e0f0 KK |
1325 | stat.ios[STAT_DISCARD], |
1326 | stat.merges[STAT_DISCARD], | |
1327 | (unsigned long long)stat.sectors[STAT_DISCARD], | |
1328 | (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC), | |
1329 | stat.ios[STAT_FLUSH], | |
1330 | (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC)); | |
3ad5cee5 CH |
1331 | } |
1332 | ||
1333 | ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr, | |
1334 | char *buf) | |
1335 | { | |
1336 | struct hd_struct *p = dev_to_part(dev); | |
1337 | struct request_queue *q = part_to_disk(p)->queue; | |
1338 | unsigned int inflight[2]; | |
1339 | ||
b2f609e1 CH |
1340 | if (queue_is_mq(q)) |
1341 | blk_mq_in_flight_rw(q, p, inflight); | |
1342 | else | |
1f06959b | 1343 | part_in_flight_rw(p, inflight); |
b2f609e1 | 1344 | |
3ad5cee5 CH |
1345 | return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]); |
1346 | } | |
1347 | ||
edfaa7c3 KS |
1348 | static ssize_t disk_capability_show(struct device *dev, |
1349 | struct device_attribute *attr, char *buf) | |
86ce18d7 | 1350 | { |
edfaa7c3 KS |
1351 | struct gendisk *disk = dev_to_disk(dev); |
1352 | ||
1353 | return sprintf(buf, "%x\n", disk->flags); | |
86ce18d7 | 1354 | } |
edfaa7c3 | 1355 | |
c72758f3 MP |
1356 | static ssize_t disk_alignment_offset_show(struct device *dev, |
1357 | struct device_attribute *attr, | |
1358 | char *buf) | |
1359 | { | |
1360 | struct gendisk *disk = dev_to_disk(dev); | |
1361 | ||
1362 | return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue)); | |
1363 | } | |
1364 | ||
86b37281 MP |
1365 | static ssize_t disk_discard_alignment_show(struct device *dev, |
1366 | struct device_attribute *attr, | |
1367 | char *buf) | |
1368 | { | |
1369 | struct gendisk *disk = dev_to_disk(dev); | |
1370 | ||
dd3d145d | 1371 | return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue)); |
86b37281 MP |
1372 | } |
1373 | ||
5657a819 JP |
1374 | static DEVICE_ATTR(range, 0444, disk_range_show, NULL); |
1375 | static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL); | |
1376 | static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL); | |
1377 | static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL); | |
1378 | static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL); | |
1379 | static DEVICE_ATTR(size, 0444, part_size_show, NULL); | |
1380 | static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL); | |
1381 | static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL); | |
1382 | static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL); | |
1383 | static DEVICE_ATTR(stat, 0444, part_stat_show, NULL); | |
1384 | static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL); | |
1385 | static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store); | |
3ad5cee5 | 1386 | |
c17bb495 | 1387 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
3ad5cee5 CH |
1388 | ssize_t part_fail_show(struct device *dev, |
1389 | struct device_attribute *attr, char *buf) | |
1390 | { | |
1391 | struct hd_struct *p = dev_to_part(dev); | |
1392 | ||
1393 | return sprintf(buf, "%d\n", p->make_it_fail); | |
1394 | } | |
1395 | ||
1396 | ssize_t part_fail_store(struct device *dev, | |
1397 | struct device_attribute *attr, | |
1398 | const char *buf, size_t count) | |
1399 | { | |
1400 | struct hd_struct *p = dev_to_part(dev); | |
1401 | int i; | |
1402 | ||
1403 | if (count > 0 && sscanf(buf, "%d", &i) > 0) | |
1404 | p->make_it_fail = (i == 0) ? 0 : 1; | |
1405 | ||
1406 | return count; | |
1407 | } | |
1408 | ||
edfaa7c3 | 1409 | static struct device_attribute dev_attr_fail = |
5657a819 | 1410 | __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store); |
3ad5cee5 CH |
1411 | #endif /* CONFIG_FAIL_MAKE_REQUEST */ |
1412 | ||
581d4e28 JA |
1413 | #ifdef CONFIG_FAIL_IO_TIMEOUT |
1414 | static struct device_attribute dev_attr_fail_timeout = | |
5657a819 | 1415 | __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store); |
581d4e28 | 1416 | #endif |
edfaa7c3 KS |
1417 | |
1418 | static struct attribute *disk_attrs[] = { | |
1419 | &dev_attr_range.attr, | |
1f014290 | 1420 | &dev_attr_ext_range.attr, |
edfaa7c3 | 1421 | &dev_attr_removable.attr, |
8ddcd653 | 1422 | &dev_attr_hidden.attr, |
1c9ce527 | 1423 | &dev_attr_ro.attr, |
edfaa7c3 | 1424 | &dev_attr_size.attr, |
c72758f3 | 1425 | &dev_attr_alignment_offset.attr, |
86b37281 | 1426 | &dev_attr_discard_alignment.attr, |
edfaa7c3 KS |
1427 | &dev_attr_capability.attr, |
1428 | &dev_attr_stat.attr, | |
316d315b | 1429 | &dev_attr_inflight.attr, |
99e6608c | 1430 | &dev_attr_badblocks.attr, |
edfaa7c3 KS |
1431 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
1432 | &dev_attr_fail.attr, | |
581d4e28 JA |
1433 | #endif |
1434 | #ifdef CONFIG_FAIL_IO_TIMEOUT | |
1435 | &dev_attr_fail_timeout.attr, | |
edfaa7c3 KS |
1436 | #endif |
1437 | NULL | |
1438 | }; | |
1439 | ||
9438b3e0 DW |
1440 | static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n) |
1441 | { | |
1442 | struct device *dev = container_of(kobj, typeof(*dev), kobj); | |
1443 | struct gendisk *disk = dev_to_disk(dev); | |
1444 | ||
1445 | if (a == &dev_attr_badblocks.attr && !disk->bb) | |
1446 | return 0; | |
1447 | return a->mode; | |
1448 | } | |
1449 | ||
edfaa7c3 KS |
1450 | static struct attribute_group disk_attr_group = { |
1451 | .attrs = disk_attrs, | |
9438b3e0 | 1452 | .is_visible = disk_visible, |
edfaa7c3 KS |
1453 | }; |
1454 | ||
a4dbd674 | 1455 | static const struct attribute_group *disk_attr_groups[] = { |
edfaa7c3 KS |
1456 | &disk_attr_group, |
1457 | NULL | |
1da177e4 LT |
1458 | }; |
1459 | ||
540eed56 TH |
1460 | /** |
1461 | * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way | |
1462 | * @disk: disk to replace part_tbl for | |
1463 | * @new_ptbl: new part_tbl to install | |
1464 | * | |
1465 | * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The | |
1466 | * original ptbl is freed using RCU callback. | |
1467 | * | |
1468 | * LOCKING: | |
6d2cf6f2 | 1469 | * Matching bd_mutex locked or the caller is the only user of @disk. |
540eed56 TH |
1470 | */ |
1471 | static void disk_replace_part_tbl(struct gendisk *disk, | |
1472 | struct disk_part_tbl *new_ptbl) | |
1473 | { | |
6d2cf6f2 BVA |
1474 | struct disk_part_tbl *old_ptbl = |
1475 | rcu_dereference_protected(disk->part_tbl, 1); | |
540eed56 TH |
1476 | |
1477 | rcu_assign_pointer(disk->part_tbl, new_ptbl); | |
a6f23657 JA |
1478 | |
1479 | if (old_ptbl) { | |
1480 | rcu_assign_pointer(old_ptbl->last_lookup, NULL); | |
57bdfbf9 | 1481 | kfree_rcu(old_ptbl, rcu_head); |
a6f23657 | 1482 | } |
540eed56 TH |
1483 | } |
1484 | ||
1485 | /** | |
1486 | * disk_expand_part_tbl - expand disk->part_tbl | |
1487 | * @disk: disk to expand part_tbl for | |
1488 | * @partno: expand such that this partno can fit in | |
1489 | * | |
1490 | * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl | |
1491 | * uses RCU to allow unlocked dereferencing for stats and other stuff. | |
1492 | * | |
1493 | * LOCKING: | |
6d2cf6f2 BVA |
1494 | * Matching bd_mutex locked or the caller is the only user of @disk. |
1495 | * Might sleep. | |
540eed56 TH |
1496 | * |
1497 | * RETURNS: | |
1498 | * 0 on success, -errno on failure. | |
1499 | */ | |
1500 | int disk_expand_part_tbl(struct gendisk *disk, int partno) | |
1501 | { | |
6d2cf6f2 BVA |
1502 | struct disk_part_tbl *old_ptbl = |
1503 | rcu_dereference_protected(disk->part_tbl, 1); | |
540eed56 TH |
1504 | struct disk_part_tbl *new_ptbl; |
1505 | int len = old_ptbl ? old_ptbl->len : 0; | |
5fabcb4c | 1506 | int i, target; |
5fabcb4c JA |
1507 | |
1508 | /* | |
1509 | * check for int overflow, since we can get here from blkpg_ioctl() | |
1510 | * with a user passed 'partno'. | |
1511 | */ | |
1512 | target = partno + 1; | |
1513 | if (target < 0) | |
1514 | return -EINVAL; | |
540eed56 TH |
1515 | |
1516 | /* disk_max_parts() is zero during initialization, ignore if so */ | |
1517 | if (disk_max_parts(disk) && target > disk_max_parts(disk)) | |
1518 | return -EINVAL; | |
1519 | ||
1520 | if (target <= len) | |
1521 | return 0; | |
1522 | ||
78b90a2c GS |
1523 | new_ptbl = kzalloc_node(struct_size(new_ptbl, part, target), GFP_KERNEL, |
1524 | disk->node_id); | |
540eed56 TH |
1525 | if (!new_ptbl) |
1526 | return -ENOMEM; | |
1527 | ||
540eed56 TH |
1528 | new_ptbl->len = target; |
1529 | ||
1530 | for (i = 0; i < len; i++) | |
1531 | rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]); | |
1532 | ||
1533 | disk_replace_part_tbl(disk, new_ptbl); | |
1534 | return 0; | |
1535 | } | |
1536 | ||
b5bd357c LC |
1537 | /** |
1538 | * disk_release - releases all allocated resources of the gendisk | |
1539 | * @dev: the device representing this disk | |
1540 | * | |
1541 | * This function releases all allocated resources of the gendisk. | |
1542 | * | |
1543 | * The struct gendisk refcount is incremented with get_gendisk() or | |
1544 | * get_disk_and_module(), and its refcount is decremented with | |
1545 | * put_disk_and_module() or put_disk(). Once the refcount reaches 0 this | |
1546 | * function is called. | |
1547 | * | |
1548 | * Drivers which used __device_add_disk() have a gendisk with a request_queue | |
1549 | * assigned. Since the request_queue sits on top of the gendisk for these | |
1550 | * drivers we also call blk_put_queue() for them, and we expect the | |
1551 | * request_queue refcount to reach 0 at this point, and so the request_queue | |
1552 | * will also be freed prior to the disk. | |
e8c7d14a LC |
1553 | * |
1554 | * Context: can sleep | |
b5bd357c | 1555 | */ |
edfaa7c3 | 1556 | static void disk_release(struct device *dev) |
1da177e4 | 1557 | { |
edfaa7c3 KS |
1558 | struct gendisk *disk = dev_to_disk(dev); |
1559 | ||
e8c7d14a LC |
1560 | might_sleep(); |
1561 | ||
2da78092 | 1562 | blk_free_devt(dev->devt); |
77ea887e | 1563 | disk_release_events(disk); |
1da177e4 | 1564 | kfree(disk->random); |
540eed56 | 1565 | disk_replace_part_tbl(disk, NULL); |
b54e5ed8 | 1566 | hd_free_part(&disk->part0); |
523e1d39 TH |
1567 | if (disk->queue) |
1568 | blk_put_queue(disk->queue); | |
1da177e4 LT |
1569 | kfree(disk); |
1570 | } | |
edfaa7c3 KS |
1571 | struct class block_class = { |
1572 | .name = "block", | |
1da177e4 LT |
1573 | }; |
1574 | ||
3c2670e6 | 1575 | static char *block_devnode(struct device *dev, umode_t *mode, |
4e4098a3 | 1576 | kuid_t *uid, kgid_t *gid) |
b03f38b6 KS |
1577 | { |
1578 | struct gendisk *disk = dev_to_disk(dev); | |
1579 | ||
348e114b CH |
1580 | if (disk->fops->devnode) |
1581 | return disk->fops->devnode(disk, mode); | |
b03f38b6 KS |
1582 | return NULL; |
1583 | } | |
1584 | ||
ef45fe47 | 1585 | const struct device_type disk_type = { |
edfaa7c3 KS |
1586 | .name = "disk", |
1587 | .groups = disk_attr_groups, | |
1588 | .release = disk_release, | |
e454cea2 | 1589 | .devnode = block_devnode, |
1da177e4 LT |
1590 | }; |
1591 | ||
a6e2ba88 | 1592 | #ifdef CONFIG_PROC_FS |
cf771cb5 TH |
1593 | /* |
1594 | * aggregate disk stat collector. Uses the same stats that the sysfs | |
1595 | * entries do, above, but makes them available through one seq_file. | |
1596 | * | |
1597 | * The output looks suspiciously like /proc/partitions with a bunch of | |
1598 | * extra fields. | |
1599 | */ | |
1600 | static int diskstats_show(struct seq_file *seqf, void *v) | |
1da177e4 LT |
1601 | { |
1602 | struct gendisk *gp = v; | |
e71bf0d0 TH |
1603 | struct disk_part_iter piter; |
1604 | struct hd_struct *hd; | |
1da177e4 | 1605 | char buf[BDEVNAME_SIZE]; |
e016b782 | 1606 | unsigned int inflight; |
ea18e0f0 | 1607 | struct disk_stats stat; |
1da177e4 LT |
1608 | |
1609 | /* | |
ed9e1982 | 1610 | if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next) |
cf771cb5 | 1611 | seq_puts(seqf, "major minor name" |
1da177e4 LT |
1612 | " rio rmerge rsect ruse wio wmerge " |
1613 | "wsect wuse running use aveq" | |
1614 | "\n\n"); | |
1615 | */ | |
9f5e4865 | 1616 | |
71982a40 | 1617 | disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0); |
e71bf0d0 | 1618 | while ((hd = disk_part_iter_next(&piter))) { |
ea18e0f0 | 1619 | part_stat_read_all(hd, &stat); |
b2f609e1 CH |
1620 | if (queue_is_mq(gp->queue)) |
1621 | inflight = blk_mq_in_flight(gp->queue, hd); | |
1622 | else | |
1f06959b | 1623 | inflight = part_in_flight(hd); |
ea18e0f0 | 1624 | |
bdca3c87 MC |
1625 | seq_printf(seqf, "%4d %7d %s " |
1626 | "%lu %lu %lu %u " | |
1627 | "%lu %lu %lu %u " | |
1628 | "%u %u %u " | |
b6866318 KK |
1629 | "%lu %lu %lu %u " |
1630 | "%lu %u" | |
1631 | "\n", | |
f331c029 TH |
1632 | MAJOR(part_devt(hd)), MINOR(part_devt(hd)), |
1633 | disk_name(gp, hd->partno, buf), | |
ea18e0f0 KK |
1634 | stat.ios[STAT_READ], |
1635 | stat.merges[STAT_READ], | |
1636 | stat.sectors[STAT_READ], | |
1637 | (unsigned int)div_u64(stat.nsecs[STAT_READ], | |
1638 | NSEC_PER_MSEC), | |
1639 | stat.ios[STAT_WRITE], | |
1640 | stat.merges[STAT_WRITE], | |
1641 | stat.sectors[STAT_WRITE], | |
1642 | (unsigned int)div_u64(stat.nsecs[STAT_WRITE], | |
1643 | NSEC_PER_MSEC), | |
e016b782 | 1644 | inflight, |
ea18e0f0 | 1645 | jiffies_to_msecs(stat.io_ticks), |
8cd5b8fc KK |
1646 | (unsigned int)div_u64(stat.nsecs[STAT_READ] + |
1647 | stat.nsecs[STAT_WRITE] + | |
1648 | stat.nsecs[STAT_DISCARD] + | |
1649 | stat.nsecs[STAT_FLUSH], | |
1650 | NSEC_PER_MSEC), | |
ea18e0f0 KK |
1651 | stat.ios[STAT_DISCARD], |
1652 | stat.merges[STAT_DISCARD], | |
1653 | stat.sectors[STAT_DISCARD], | |
1654 | (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], | |
1655 | NSEC_PER_MSEC), | |
1656 | stat.ios[STAT_FLUSH], | |
1657 | (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], | |
1658 | NSEC_PER_MSEC) | |
28f39d55 | 1659 | ); |
1da177e4 | 1660 | } |
e71bf0d0 | 1661 | disk_part_iter_exit(&piter); |
9f5e4865 | 1662 | |
1da177e4 LT |
1663 | return 0; |
1664 | } | |
1665 | ||
31d85ab2 | 1666 | static const struct seq_operations diskstats_op = { |
def4e38d TH |
1667 | .start = disk_seqf_start, |
1668 | .next = disk_seqf_next, | |
1669 | .stop = disk_seqf_stop, | |
1da177e4 LT |
1670 | .show = diskstats_show |
1671 | }; | |
f500975a AD |
1672 | |
1673 | static int __init proc_genhd_init(void) | |
1674 | { | |
fddda2b7 CH |
1675 | proc_create_seq("diskstats", 0, NULL, &diskstats_op); |
1676 | proc_create_seq("partitions", 0, NULL, &partitions_op); | |
f500975a AD |
1677 | return 0; |
1678 | } | |
1679 | module_init(proc_genhd_init); | |
a6e2ba88 | 1680 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 1681 | |
cf771cb5 | 1682 | dev_t blk_lookup_devt(const char *name, int partno) |
a142be85 | 1683 | { |
def4e38d TH |
1684 | dev_t devt = MKDEV(0, 0); |
1685 | struct class_dev_iter iter; | |
1686 | struct device *dev; | |
a142be85 | 1687 | |
def4e38d TH |
1688 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); |
1689 | while ((dev = class_dev_iter_next(&iter))) { | |
a142be85 | 1690 | struct gendisk *disk = dev_to_disk(dev); |
548b10eb | 1691 | struct hd_struct *part; |
a142be85 | 1692 | |
3ada8b7e | 1693 | if (strcmp(dev_name(dev), name)) |
f331c029 | 1694 | continue; |
f331c029 | 1695 | |
41b8c853 NB |
1696 | if (partno < disk->minors) { |
1697 | /* We need to return the right devno, even | |
1698 | * if the partition doesn't exist yet. | |
1699 | */ | |
1700 | devt = MKDEV(MAJOR(dev->devt), | |
1701 | MINOR(dev->devt) + partno); | |
1702 | break; | |
1703 | } | |
548b10eb | 1704 | part = disk_get_part(disk, partno); |
2bbedcb4 | 1705 | if (part) { |
f331c029 | 1706 | devt = part_devt(part); |
e71bf0d0 | 1707 | disk_put_part(part); |
548b10eb | 1708 | break; |
def4e38d | 1709 | } |
548b10eb | 1710 | disk_put_part(part); |
5c0ef6d0 | 1711 | } |
def4e38d | 1712 | class_dev_iter_exit(&iter); |
edfaa7c3 KS |
1713 | return devt; |
1714 | } | |
edfaa7c3 | 1715 | |
e319e1fb | 1716 | struct gendisk *__alloc_disk_node(int minors, int node_id) |
1946089a CL |
1717 | { |
1718 | struct gendisk *disk; | |
6d2cf6f2 | 1719 | struct disk_part_tbl *ptbl; |
1946089a | 1720 | |
de65b012 CH |
1721 | if (minors > DISK_MAX_PARTS) { |
1722 | printk(KERN_ERR | |
7fb52621 | 1723 | "block: can't allocate more than %d partitions\n", |
de65b012 CH |
1724 | DISK_MAX_PARTS); |
1725 | minors = DISK_MAX_PARTS; | |
1726 | } | |
1946089a | 1727 | |
c1b511eb | 1728 | disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id); |
f93af2a4 CH |
1729 | if (!disk) |
1730 | return NULL; | |
6c23a968 | 1731 | |
f93af2a4 CH |
1732 | disk->part0.dkstats = alloc_percpu(struct disk_stats); |
1733 | if (!disk->part0.dkstats) | |
1734 | goto out_free_disk; | |
b5d0b9df | 1735 | |
f93af2a4 CH |
1736 | init_rwsem(&disk->lookup_sem); |
1737 | disk->node_id = node_id; | |
1738 | if (disk_expand_part_tbl(disk, 0)) { | |
1739 | free_percpu(disk->part0.dkstats); | |
1740 | goto out_free_disk; | |
1da177e4 | 1741 | } |
f93af2a4 CH |
1742 | |
1743 | ptbl = rcu_dereference_protected(disk->part_tbl, 1); | |
1744 | rcu_assign_pointer(ptbl->part[0], &disk->part0); | |
1745 | ||
1746 | /* | |
1747 | * set_capacity() and get_capacity() currently don't use | |
1748 | * seqcounter to read/update the part0->nr_sects. Still init | |
1749 | * the counter as we can read the sectors in IO submission | |
1750 | * patch using seqence counters. | |
1751 | * | |
1752 | * TODO: Ideally set_capacity() and get_capacity() should be | |
1753 | * converted to make use of bd_mutex and sequence counters. | |
1754 | */ | |
1755 | hd_sects_seq_init(&disk->part0); | |
1756 | if (hd_ref_init(&disk->part0)) | |
1757 | goto out_free_part0; | |
1758 | ||
1759 | disk->minors = minors; | |
1760 | rand_initialize_disk(disk); | |
1761 | disk_to_dev(disk)->class = &block_class; | |
1762 | disk_to_dev(disk)->type = &disk_type; | |
1763 | device_initialize(disk_to_dev(disk)); | |
1da177e4 | 1764 | return disk; |
f93af2a4 CH |
1765 | |
1766 | out_free_part0: | |
1767 | hd_free_part(&disk->part0); | |
1768 | out_free_disk: | |
1769 | kfree(disk); | |
1770 | return NULL; | |
1da177e4 | 1771 | } |
e319e1fb | 1772 | EXPORT_SYMBOL(__alloc_disk_node); |
1da177e4 | 1773 | |
b5bd357c LC |
1774 | /** |
1775 | * get_disk_and_module - increments the gendisk and gendisk fops module refcount | |
0d20dcc2 | 1776 | * @disk: the struct gendisk to increment the refcount for |
b5bd357c LC |
1777 | * |
1778 | * This increments the refcount for the struct gendisk, and the gendisk's | |
1779 | * fops module owner. | |
763b5892 LC |
1780 | * |
1781 | * Context: Any context. | |
b5bd357c | 1782 | */ |
3079c22e | 1783 | struct kobject *get_disk_and_module(struct gendisk *disk) |
1da177e4 LT |
1784 | { |
1785 | struct module *owner; | |
1786 | struct kobject *kobj; | |
1787 | ||
1788 | if (!disk->fops) | |
1789 | return NULL; | |
1790 | owner = disk->fops->owner; | |
1791 | if (owner && !try_module_get(owner)) | |
1792 | return NULL; | |
d01b2dcb | 1793 | kobj = kobject_get_unless_zero(&disk_to_dev(disk)->kobj); |
1da177e4 LT |
1794 | if (kobj == NULL) { |
1795 | module_put(owner); | |
1796 | return NULL; | |
1797 | } | |
1798 | return kobj; | |
1799 | ||
1800 | } | |
3079c22e | 1801 | EXPORT_SYMBOL(get_disk_and_module); |
1da177e4 | 1802 | |
b5bd357c LC |
1803 | /** |
1804 | * put_disk - decrements the gendisk refcount | |
0d20dcc2 | 1805 | * @disk: the struct gendisk to decrement the refcount for |
b5bd357c LC |
1806 | * |
1807 | * This decrements the refcount for the struct gendisk. When this reaches 0 | |
1808 | * we'll have disk_release() called. | |
e8c7d14a LC |
1809 | * |
1810 | * Context: Any context, but the last reference must not be dropped from | |
1811 | * atomic context. | |
b5bd357c | 1812 | */ |
1da177e4 LT |
1813 | void put_disk(struct gendisk *disk) |
1814 | { | |
1815 | if (disk) | |
ed9e1982 | 1816 | kobject_put(&disk_to_dev(disk)->kobj); |
1da177e4 | 1817 | } |
1da177e4 LT |
1818 | EXPORT_SYMBOL(put_disk); |
1819 | ||
b5bd357c LC |
1820 | /** |
1821 | * put_disk_and_module - decrements the module and gendisk refcount | |
0d20dcc2 | 1822 | * @disk: the struct gendisk to decrement the refcount for |
b5bd357c | 1823 | * |
9df6c299 JK |
1824 | * This is a counterpart of get_disk_and_module() and thus also of |
1825 | * get_gendisk(). | |
e8c7d14a LC |
1826 | * |
1827 | * Context: Any context, but the last reference must not be dropped from | |
1828 | * atomic context. | |
9df6c299 JK |
1829 | */ |
1830 | void put_disk_and_module(struct gendisk *disk) | |
1831 | { | |
1832 | if (disk) { | |
1833 | struct module *owner = disk->fops->owner; | |
1834 | ||
1835 | put_disk(disk); | |
1836 | module_put(owner); | |
1837 | } | |
1838 | } | |
1839 | EXPORT_SYMBOL(put_disk_and_module); | |
1840 | ||
e3264a4d HR |
1841 | static void set_disk_ro_uevent(struct gendisk *gd, int ro) |
1842 | { | |
1843 | char event[] = "DISK_RO=1"; | |
1844 | char *envp[] = { event, NULL }; | |
1845 | ||
1846 | if (!ro) | |
1847 | event[8] = '0'; | |
1848 | kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); | |
1849 | } | |
1850 | ||
1da177e4 LT |
1851 | void set_device_ro(struct block_device *bdev, int flag) |
1852 | { | |
b7db9956 | 1853 | bdev->bd_part->policy = flag; |
1da177e4 LT |
1854 | } |
1855 | ||
1856 | EXPORT_SYMBOL(set_device_ro); | |
1857 | ||
1858 | void set_disk_ro(struct gendisk *disk, int flag) | |
1859 | { | |
e71bf0d0 TH |
1860 | struct disk_part_iter piter; |
1861 | struct hd_struct *part; | |
1862 | ||
e3264a4d HR |
1863 | if (disk->part0.policy != flag) { |
1864 | set_disk_ro_uevent(disk, flag); | |
1865 | disk->part0.policy = flag; | |
1866 | } | |
1867 | ||
1868 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY); | |
e71bf0d0 TH |
1869 | while ((part = disk_part_iter_next(&piter))) |
1870 | part->policy = flag; | |
1871 | disk_part_iter_exit(&piter); | |
1da177e4 LT |
1872 | } |
1873 | ||
1874 | EXPORT_SYMBOL(set_disk_ro); | |
1875 | ||
1876 | int bdev_read_only(struct block_device *bdev) | |
1877 | { | |
1878 | if (!bdev) | |
1879 | return 0; | |
b7db9956 | 1880 | return bdev->bd_part->policy; |
1da177e4 LT |
1881 | } |
1882 | ||
1883 | EXPORT_SYMBOL(bdev_read_only); | |
1884 | ||
77ea887e TH |
1885 | /* |
1886 | * Disk events - monitor disk events like media change and eject request. | |
1887 | */ | |
1888 | struct disk_events { | |
1889 | struct list_head node; /* all disk_event's */ | |
1890 | struct gendisk *disk; /* the associated disk */ | |
1891 | spinlock_t lock; | |
1892 | ||
fdd514e1 | 1893 | struct mutex block_mutex; /* protects blocking */ |
77ea887e TH |
1894 | int block; /* event blocking depth */ |
1895 | unsigned int pending; /* events already sent out */ | |
1896 | unsigned int clearing; /* events being cleared */ | |
1897 | ||
1898 | long poll_msecs; /* interval, -1 for default */ | |
1899 | struct delayed_work dwork; | |
1900 | }; | |
1901 | ||
1902 | static const char *disk_events_strs[] = { | |
1903 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change", | |
1904 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request", | |
1905 | }; | |
1906 | ||
1907 | static char *disk_uevents[] = { | |
1908 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1", | |
1909 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1", | |
1910 | }; | |
1911 | ||
1912 | /* list of all disk_events */ | |
1913 | static DEFINE_MUTEX(disk_events_mutex); | |
1914 | static LIST_HEAD(disk_events); | |
1915 | ||
1916 | /* disable in-kernel polling by default */ | |
1fe8f348 | 1917 | static unsigned long disk_events_dfl_poll_msecs; |
77ea887e TH |
1918 | |
1919 | static unsigned long disk_events_poll_jiffies(struct gendisk *disk) | |
1920 | { | |
1921 | struct disk_events *ev = disk->ev; | |
1922 | long intv_msecs = 0; | |
1923 | ||
1924 | /* | |
1925 | * If device-specific poll interval is set, always use it. If | |
673387a9 | 1926 | * the default is being used, poll if the POLL flag is set. |
77ea887e TH |
1927 | */ |
1928 | if (ev->poll_msecs >= 0) | |
1929 | intv_msecs = ev->poll_msecs; | |
c92e2f04 | 1930 | else if (disk->event_flags & DISK_EVENT_FLAG_POLL) |
77ea887e TH |
1931 | intv_msecs = disk_events_dfl_poll_msecs; |
1932 | ||
1933 | return msecs_to_jiffies(intv_msecs); | |
1934 | } | |
1935 | ||
c3af54af TH |
1936 | /** |
1937 | * disk_block_events - block and flush disk event checking | |
1938 | * @disk: disk to block events for | |
1939 | * | |
1940 | * On return from this function, it is guaranteed that event checking | |
1941 | * isn't in progress and won't happen until unblocked by | |
1942 | * disk_unblock_events(). Events blocking is counted and the actual | |
1943 | * unblocking happens after the matching number of unblocks are done. | |
1944 | * | |
1945 | * Note that this intentionally does not block event checking from | |
1946 | * disk_clear_events(). | |
1947 | * | |
1948 | * CONTEXT: | |
1949 | * Might sleep. | |
1950 | */ | |
1951 | void disk_block_events(struct gendisk *disk) | |
77ea887e TH |
1952 | { |
1953 | struct disk_events *ev = disk->ev; | |
1954 | unsigned long flags; | |
1955 | bool cancel; | |
1956 | ||
c3af54af TH |
1957 | if (!ev) |
1958 | return; | |
1959 | ||
fdd514e1 TH |
1960 | /* |
1961 | * Outer mutex ensures that the first blocker completes canceling | |
1962 | * the event work before further blockers are allowed to finish. | |
1963 | */ | |
1964 | mutex_lock(&ev->block_mutex); | |
1965 | ||
77ea887e TH |
1966 | spin_lock_irqsave(&ev->lock, flags); |
1967 | cancel = !ev->block++; | |
1968 | spin_unlock_irqrestore(&ev->lock, flags); | |
1969 | ||
c3af54af TH |
1970 | if (cancel) |
1971 | cancel_delayed_work_sync(&disk->ev->dwork); | |
fdd514e1 TH |
1972 | |
1973 | mutex_unlock(&ev->block_mutex); | |
77ea887e TH |
1974 | } |
1975 | ||
1976 | static void __disk_unblock_events(struct gendisk *disk, bool check_now) | |
1977 | { | |
1978 | struct disk_events *ev = disk->ev; | |
1979 | unsigned long intv; | |
1980 | unsigned long flags; | |
1981 | ||
1982 | spin_lock_irqsave(&ev->lock, flags); | |
1983 | ||
1984 | if (WARN_ON_ONCE(ev->block <= 0)) | |
1985 | goto out_unlock; | |
1986 | ||
1987 | if (--ev->block) | |
1988 | goto out_unlock; | |
1989 | ||
77ea887e | 1990 | intv = disk_events_poll_jiffies(disk); |
77ea887e | 1991 | if (check_now) |
695588f9 VK |
1992 | queue_delayed_work(system_freezable_power_efficient_wq, |
1993 | &ev->dwork, 0); | |
77ea887e | 1994 | else if (intv) |
695588f9 VK |
1995 | queue_delayed_work(system_freezable_power_efficient_wq, |
1996 | &ev->dwork, intv); | |
77ea887e TH |
1997 | out_unlock: |
1998 | spin_unlock_irqrestore(&ev->lock, flags); | |
1999 | } | |
2000 | ||
77ea887e TH |
2001 | /** |
2002 | * disk_unblock_events - unblock disk event checking | |
2003 | * @disk: disk to unblock events for | |
2004 | * | |
2005 | * Undo disk_block_events(). When the block count reaches zero, it | |
2006 | * starts events polling if configured. | |
2007 | * | |
2008 | * CONTEXT: | |
2009 | * Don't care. Safe to call from irq context. | |
2010 | */ | |
2011 | void disk_unblock_events(struct gendisk *disk) | |
2012 | { | |
2013 | if (disk->ev) | |
facc31dd | 2014 | __disk_unblock_events(disk, false); |
77ea887e TH |
2015 | } |
2016 | ||
2017 | /** | |
85ef06d1 TH |
2018 | * disk_flush_events - schedule immediate event checking and flushing |
2019 | * @disk: disk to check and flush events for | |
2020 | * @mask: events to flush | |
77ea887e | 2021 | * |
85ef06d1 TH |
2022 | * Schedule immediate event checking on @disk if not blocked. Events in |
2023 | * @mask are scheduled to be cleared from the driver. Note that this | |
2024 | * doesn't clear the events from @disk->ev. | |
77ea887e TH |
2025 | * |
2026 | * CONTEXT: | |
85ef06d1 | 2027 | * If @mask is non-zero must be called with bdev->bd_mutex held. |
77ea887e | 2028 | */ |
85ef06d1 | 2029 | void disk_flush_events(struct gendisk *disk, unsigned int mask) |
77ea887e | 2030 | { |
a9dce2a3 | 2031 | struct disk_events *ev = disk->ev; |
a9dce2a3 TH |
2032 | |
2033 | if (!ev) | |
2034 | return; | |
2035 | ||
85ef06d1 TH |
2036 | spin_lock_irq(&ev->lock); |
2037 | ev->clearing |= mask; | |
41f63c53 | 2038 | if (!ev->block) |
695588f9 VK |
2039 | mod_delayed_work(system_freezable_power_efficient_wq, |
2040 | &ev->dwork, 0); | |
85ef06d1 | 2041 | spin_unlock_irq(&ev->lock); |
77ea887e | 2042 | } |
77ea887e TH |
2043 | |
2044 | /** | |
2045 | * disk_clear_events - synchronously check, clear and return pending events | |
2046 | * @disk: disk to fetch and clear events from | |
da3dae54 | 2047 | * @mask: mask of events to be fetched and cleared |
77ea887e TH |
2048 | * |
2049 | * Disk events are synchronously checked and pending events in @mask | |
2050 | * are cleared and returned. This ignores the block count. | |
2051 | * | |
2052 | * CONTEXT: | |
2053 | * Might sleep. | |
2054 | */ | |
2055 | unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask) | |
2056 | { | |
77ea887e TH |
2057 | struct disk_events *ev = disk->ev; |
2058 | unsigned int pending; | |
12c2bdb2 | 2059 | unsigned int clearing = mask; |
77ea887e | 2060 | |
a564e23f | 2061 | if (!ev) |
77ea887e | 2062 | return 0; |
77ea887e | 2063 | |
12c2bdb2 DB |
2064 | disk_block_events(disk); |
2065 | ||
2066 | /* | |
2067 | * store the union of mask and ev->clearing on the stack so that the | |
2068 | * race with disk_flush_events does not cause ambiguity (ev->clearing | |
2069 | * can still be modified even if events are blocked). | |
2070 | */ | |
77ea887e | 2071 | spin_lock_irq(&ev->lock); |
12c2bdb2 DB |
2072 | clearing |= ev->clearing; |
2073 | ev->clearing = 0; | |
77ea887e TH |
2074 | spin_unlock_irq(&ev->lock); |
2075 | ||
12c2bdb2 | 2076 | disk_check_events(ev, &clearing); |
aea24a8b | 2077 | /* |
12c2bdb2 DB |
2078 | * if ev->clearing is not 0, the disk_flush_events got called in the |
2079 | * middle of this function, so we want to run the workfn without delay. | |
aea24a8b | 2080 | */ |
12c2bdb2 | 2081 | __disk_unblock_events(disk, ev->clearing ? true : false); |
77ea887e TH |
2082 | |
2083 | /* then, fetch and clear pending events */ | |
2084 | spin_lock_irq(&ev->lock); | |
77ea887e TH |
2085 | pending = ev->pending & mask; |
2086 | ev->pending &= ~mask; | |
2087 | spin_unlock_irq(&ev->lock); | |
12c2bdb2 | 2088 | WARN_ON_ONCE(clearing & mask); |
77ea887e TH |
2089 | |
2090 | return pending; | |
2091 | } | |
2092 | ||
12c2bdb2 DB |
2093 | /* |
2094 | * Separate this part out so that a different pointer for clearing_ptr can be | |
2095 | * passed in for disk_clear_events. | |
2096 | */ | |
77ea887e TH |
2097 | static void disk_events_workfn(struct work_struct *work) |
2098 | { | |
2099 | struct delayed_work *dwork = to_delayed_work(work); | |
2100 | struct disk_events *ev = container_of(dwork, struct disk_events, dwork); | |
12c2bdb2 DB |
2101 | |
2102 | disk_check_events(ev, &ev->clearing); | |
2103 | } | |
2104 | ||
2105 | static void disk_check_events(struct disk_events *ev, | |
2106 | unsigned int *clearing_ptr) | |
2107 | { | |
77ea887e TH |
2108 | struct gendisk *disk = ev->disk; |
2109 | char *envp[ARRAY_SIZE(disk_uevents) + 1] = { }; | |
12c2bdb2 | 2110 | unsigned int clearing = *clearing_ptr; |
77ea887e TH |
2111 | unsigned int events; |
2112 | unsigned long intv; | |
2113 | int nr_events = 0, i; | |
2114 | ||
2115 | /* check events */ | |
2116 | events = disk->fops->check_events(disk, clearing); | |
2117 | ||
2118 | /* accumulate pending events and schedule next poll if necessary */ | |
2119 | spin_lock_irq(&ev->lock); | |
2120 | ||
2121 | events &= ~ev->pending; | |
2122 | ev->pending |= events; | |
12c2bdb2 | 2123 | *clearing_ptr &= ~clearing; |
77ea887e TH |
2124 | |
2125 | intv = disk_events_poll_jiffies(disk); | |
2126 | if (!ev->block && intv) | |
695588f9 VK |
2127 | queue_delayed_work(system_freezable_power_efficient_wq, |
2128 | &ev->dwork, intv); | |
77ea887e TH |
2129 | |
2130 | spin_unlock_irq(&ev->lock); | |
2131 | ||
7c88a168 TH |
2132 | /* |
2133 | * Tell userland about new events. Only the events listed in | |
c92e2f04 MW |
2134 | * @disk->events are reported, and only if DISK_EVENT_FLAG_UEVENT |
2135 | * is set. Otherwise, events are processed internally but never | |
2136 | * get reported to userland. | |
7c88a168 | 2137 | */ |
77ea887e | 2138 | for (i = 0; i < ARRAY_SIZE(disk_uevents); i++) |
c92e2f04 MW |
2139 | if ((events & disk->events & (1 << i)) && |
2140 | (disk->event_flags & DISK_EVENT_FLAG_UEVENT)) | |
77ea887e TH |
2141 | envp[nr_events++] = disk_uevents[i]; |
2142 | ||
2143 | if (nr_events) | |
2144 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
2145 | } | |
2146 | ||
2147 | /* | |
2148 | * A disk events enabled device has the following sysfs nodes under | |
2149 | * its /sys/block/X/ directory. | |
2150 | * | |
2151 | * events : list of all supported events | |
2152 | * events_async : list of events which can be detected w/o polling | |
673387a9 | 2153 | * (always empty, only for backwards compatibility) |
77ea887e TH |
2154 | * events_poll_msecs : polling interval, 0: disable, -1: system default |
2155 | */ | |
2156 | static ssize_t __disk_events_show(unsigned int events, char *buf) | |
2157 | { | |
2158 | const char *delim = ""; | |
2159 | ssize_t pos = 0; | |
2160 | int i; | |
2161 | ||
2162 | for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++) | |
2163 | if (events & (1 << i)) { | |
2164 | pos += sprintf(buf + pos, "%s%s", | |
2165 | delim, disk_events_strs[i]); | |
2166 | delim = " "; | |
2167 | } | |
2168 | if (pos) | |
2169 | pos += sprintf(buf + pos, "\n"); | |
2170 | return pos; | |
2171 | } | |
2172 | ||
2173 | static ssize_t disk_events_show(struct device *dev, | |
2174 | struct device_attribute *attr, char *buf) | |
2175 | { | |
2176 | struct gendisk *disk = dev_to_disk(dev); | |
2177 | ||
c92e2f04 MW |
2178 | if (!(disk->event_flags & DISK_EVENT_FLAG_UEVENT)) |
2179 | return 0; | |
2180 | ||
77ea887e TH |
2181 | return __disk_events_show(disk->events, buf); |
2182 | } | |
2183 | ||
2184 | static ssize_t disk_events_async_show(struct device *dev, | |
2185 | struct device_attribute *attr, char *buf) | |
2186 | { | |
673387a9 | 2187 | return 0; |
77ea887e TH |
2188 | } |
2189 | ||
2190 | static ssize_t disk_events_poll_msecs_show(struct device *dev, | |
2191 | struct device_attribute *attr, | |
2192 | char *buf) | |
2193 | { | |
2194 | struct gendisk *disk = dev_to_disk(dev); | |
2195 | ||
cdf3e3de MW |
2196 | if (!disk->ev) |
2197 | return sprintf(buf, "-1\n"); | |
2198 | ||
77ea887e TH |
2199 | return sprintf(buf, "%ld\n", disk->ev->poll_msecs); |
2200 | } | |
2201 | ||
2202 | static ssize_t disk_events_poll_msecs_store(struct device *dev, | |
2203 | struct device_attribute *attr, | |
2204 | const char *buf, size_t count) | |
2205 | { | |
2206 | struct gendisk *disk = dev_to_disk(dev); | |
2207 | long intv; | |
2208 | ||
2209 | if (!count || !sscanf(buf, "%ld", &intv)) | |
2210 | return -EINVAL; | |
2211 | ||
2212 | if (intv < 0 && intv != -1) | |
2213 | return -EINVAL; | |
2214 | ||
cdf3e3de MW |
2215 | if (!disk->ev) |
2216 | return -ENODEV; | |
2217 | ||
c3af54af | 2218 | disk_block_events(disk); |
77ea887e TH |
2219 | disk->ev->poll_msecs = intv; |
2220 | __disk_unblock_events(disk, true); | |
2221 | ||
2222 | return count; | |
2223 | } | |
2224 | ||
5657a819 JP |
2225 | static const DEVICE_ATTR(events, 0444, disk_events_show, NULL); |
2226 | static const DEVICE_ATTR(events_async, 0444, disk_events_async_show, NULL); | |
2227 | static const DEVICE_ATTR(events_poll_msecs, 0644, | |
77ea887e TH |
2228 | disk_events_poll_msecs_show, |
2229 | disk_events_poll_msecs_store); | |
2230 | ||
2231 | static const struct attribute *disk_events_attrs[] = { | |
2232 | &dev_attr_events.attr, | |
2233 | &dev_attr_events_async.attr, | |
2234 | &dev_attr_events_poll_msecs.attr, | |
2235 | NULL, | |
2236 | }; | |
2237 | ||
2238 | /* | |
2239 | * The default polling interval can be specified by the kernel | |
2240 | * parameter block.events_dfl_poll_msecs which defaults to 0 | |
2241 | * (disable). This can also be modified runtime by writing to | |
1624b0b2 | 2242 | * /sys/module/block/parameters/events_dfl_poll_msecs. |
77ea887e TH |
2243 | */ |
2244 | static int disk_events_set_dfl_poll_msecs(const char *val, | |
2245 | const struct kernel_param *kp) | |
2246 | { | |
2247 | struct disk_events *ev; | |
2248 | int ret; | |
2249 | ||
2250 | ret = param_set_ulong(val, kp); | |
2251 | if (ret < 0) | |
2252 | return ret; | |
2253 | ||
2254 | mutex_lock(&disk_events_mutex); | |
2255 | ||
2256 | list_for_each_entry(ev, &disk_events, node) | |
85ef06d1 | 2257 | disk_flush_events(ev->disk, 0); |
77ea887e TH |
2258 | |
2259 | mutex_unlock(&disk_events_mutex); | |
2260 | ||
2261 | return 0; | |
2262 | } | |
2263 | ||
2264 | static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = { | |
2265 | .set = disk_events_set_dfl_poll_msecs, | |
2266 | .get = param_get_ulong, | |
2267 | }; | |
2268 | ||
2269 | #undef MODULE_PARAM_PREFIX | |
2270 | #define MODULE_PARAM_PREFIX "block." | |
2271 | ||
2272 | module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops, | |
2273 | &disk_events_dfl_poll_msecs, 0644); | |
2274 | ||
2275 | /* | |
9f53d2fe | 2276 | * disk_{alloc|add|del|release}_events - initialize and destroy disk_events. |
77ea887e | 2277 | */ |
9f53d2fe | 2278 | static void disk_alloc_events(struct gendisk *disk) |
77ea887e TH |
2279 | { |
2280 | struct disk_events *ev; | |
2281 | ||
cdf3e3de | 2282 | if (!disk->fops->check_events || !disk->events) |
77ea887e TH |
2283 | return; |
2284 | ||
2285 | ev = kzalloc(sizeof(*ev), GFP_KERNEL); | |
2286 | if (!ev) { | |
2287 | pr_warn("%s: failed to initialize events\n", disk->disk_name); | |
2288 | return; | |
2289 | } | |
2290 | ||
77ea887e TH |
2291 | INIT_LIST_HEAD(&ev->node); |
2292 | ev->disk = disk; | |
2293 | spin_lock_init(&ev->lock); | |
fdd514e1 | 2294 | mutex_init(&ev->block_mutex); |
77ea887e TH |
2295 | ev->block = 1; |
2296 | ev->poll_msecs = -1; | |
2297 | INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn); | |
2298 | ||
9f53d2fe SG |
2299 | disk->ev = ev; |
2300 | } | |
2301 | ||
2302 | static void disk_add_events(struct gendisk *disk) | |
2303 | { | |
9f53d2fe SG |
2304 | /* FIXME: error handling */ |
2305 | if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0) | |
2306 | pr_warn("%s: failed to create sysfs files for events\n", | |
2307 | disk->disk_name); | |
2308 | ||
cdf3e3de MW |
2309 | if (!disk->ev) |
2310 | return; | |
2311 | ||
77ea887e | 2312 | mutex_lock(&disk_events_mutex); |
9f53d2fe | 2313 | list_add_tail(&disk->ev->node, &disk_events); |
77ea887e TH |
2314 | mutex_unlock(&disk_events_mutex); |
2315 | ||
2316 | /* | |
2317 | * Block count is initialized to 1 and the following initial | |
2318 | * unblock kicks it into action. | |
2319 | */ | |
2320 | __disk_unblock_events(disk, true); | |
2321 | } | |
2322 | ||
2323 | static void disk_del_events(struct gendisk *disk) | |
2324 | { | |
cdf3e3de MW |
2325 | if (disk->ev) { |
2326 | disk_block_events(disk); | |
77ea887e | 2327 | |
cdf3e3de MW |
2328 | mutex_lock(&disk_events_mutex); |
2329 | list_del_init(&disk->ev->node); | |
2330 | mutex_unlock(&disk_events_mutex); | |
2331 | } | |
77ea887e TH |
2332 | |
2333 | sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs); | |
2334 | } | |
2335 | ||
2336 | static void disk_release_events(struct gendisk *disk) | |
2337 | { | |
2338 | /* the block count should be 1 from disk_del_events() */ | |
2339 | WARN_ON_ONCE(disk->ev && disk->ev->block != 1); | |
2340 | kfree(disk->ev); | |
2341 | } |