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> | |
58383af6 | 20 | #include <linux/mutex.h> |
bcce3de1 | 21 | #include <linux/idr.h> |
77ea887e | 22 | #include <linux/log2.h> |
25e823c8 | 23 | #include <linux/pm_runtime.h> |
99e6608c | 24 | #include <linux/badblocks.h> |
1da177e4 | 25 | |
ff88972c AB |
26 | #include "blk.h" |
27 | ||
31eb6186 | 28 | static struct kobject *block_depr; |
1da177e4 | 29 | |
e418de3a | 30 | static DEFINE_XARRAY(bdev_map); |
e49fbbbf | 31 | static DEFINE_MUTEX(bdev_map_lock); |
62b508f8 | 32 | |
bcce3de1 | 33 | /* for extended dynamic devt allocation, currently only one major is used */ |
ce23bba8 | 34 | #define NR_EXT_DEVT (1 << MINORBITS) |
bcce3de1 | 35 | |
2da78092 | 36 | /* For extended devt allocation. ext_devt_lock prevents look up |
bcce3de1 TH |
37 | * results from going away underneath its user. |
38 | */ | |
2da78092 | 39 | static DEFINE_SPINLOCK(ext_devt_lock); |
bcce3de1 TH |
40 | static DEFINE_IDR(ext_devt_idr); |
41 | ||
12c2bdb2 DB |
42 | static void disk_check_events(struct disk_events *ev, |
43 | unsigned int *clearing_ptr); | |
9f53d2fe | 44 | static void disk_alloc_events(struct gendisk *disk); |
77ea887e TH |
45 | static void disk_add_events(struct gendisk *disk); |
46 | static void disk_del_events(struct gendisk *disk); | |
47 | static void disk_release_events(struct gendisk *disk); | |
48 | ||
e598a72f | 49 | /* |
449f4ec9 CH |
50 | * Set disk capacity and notify if the size is not currently zero and will not |
51 | * be set to zero. Returns true if a uevent was sent, otherwise false. | |
e598a72f | 52 | */ |
449f4ec9 | 53 | bool set_capacity_and_notify(struct gendisk *disk, sector_t size) |
e598a72f BS |
54 | { |
55 | sector_t capacity = get_capacity(disk); | |
56 | ||
57 | set_capacity(disk, size); | |
449f4ec9 | 58 | revalidate_disk_size(disk, true); |
e598a72f BS |
59 | |
60 | if (capacity != size && capacity != 0 && size != 0) { | |
61 | char *envp[] = { "RESIZE=1", NULL }; | |
62 | ||
63 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
7e890c37 | 64 | return true; |
e598a72f | 65 | } |
7e890c37 CH |
66 | |
67 | return false; | |
e598a72f | 68 | } |
449f4ec9 | 69 | EXPORT_SYMBOL_GPL(set_capacity_and_notify); |
e598a72f | 70 | |
5cbd28e3 CH |
71 | /* |
72 | * Format the device name of the indicated disk into the supplied buffer and | |
73 | * return a pointer to that same buffer for convenience. | |
74 | */ | |
75 | char *disk_name(struct gendisk *hd, int partno, char *buf) | |
76 | { | |
77 | if (!partno) | |
78 | snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name); | |
79 | else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1])) | |
80 | snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno); | |
81 | else | |
82 | snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno); | |
83 | ||
84 | return buf; | |
85 | } | |
86 | ||
87 | const char *bdevname(struct block_device *bdev, char *buf) | |
88 | { | |
8a63a86e | 89 | return disk_name(bdev->bd_disk, bdev->bd_partno, buf); |
5cbd28e3 CH |
90 | } |
91 | EXPORT_SYMBOL(bdevname); | |
e598a72f | 92 | |
ea18e0f0 KK |
93 | static void part_stat_read_all(struct hd_struct *part, struct disk_stats *stat) |
94 | { | |
95 | int cpu; | |
96 | ||
97 | memset(stat, 0, sizeof(struct disk_stats)); | |
98 | for_each_possible_cpu(cpu) { | |
99 | struct disk_stats *ptr = per_cpu_ptr(part->dkstats, cpu); | |
100 | int group; | |
101 | ||
102 | for (group = 0; group < NR_STAT_GROUPS; group++) { | |
103 | stat->nsecs[group] += ptr->nsecs[group]; | |
104 | stat->sectors[group] += ptr->sectors[group]; | |
105 | stat->ios[group] += ptr->ios[group]; | |
106 | stat->merges[group] += ptr->merges[group]; | |
107 | } | |
108 | ||
109 | stat->io_ticks += ptr->io_ticks; | |
ea18e0f0 KK |
110 | } |
111 | } | |
ea18e0f0 | 112 | |
1f06959b | 113 | static unsigned int part_in_flight(struct hd_struct *part) |
f299b7c7 | 114 | { |
b2f609e1 | 115 | unsigned int inflight = 0; |
1226b8dd | 116 | int cpu; |
f299b7c7 | 117 | |
1226b8dd | 118 | for_each_possible_cpu(cpu) { |
e016b782 MP |
119 | inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) + |
120 | part_stat_local_read_cpu(part, in_flight[1], cpu); | |
1226b8dd | 121 | } |
e016b782 MP |
122 | if ((int)inflight < 0) |
123 | inflight = 0; | |
1226b8dd | 124 | |
e016b782 | 125 | return inflight; |
f299b7c7 JA |
126 | } |
127 | ||
1f06959b | 128 | static void part_in_flight_rw(struct hd_struct *part, unsigned int inflight[2]) |
bf0ddaba | 129 | { |
1226b8dd MP |
130 | int cpu; |
131 | ||
1226b8dd MP |
132 | inflight[0] = 0; |
133 | inflight[1] = 0; | |
134 | for_each_possible_cpu(cpu) { | |
135 | inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu); | |
136 | inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu); | |
137 | } | |
138 | if ((int)inflight[0] < 0) | |
139 | inflight[0] = 0; | |
140 | if ((int)inflight[1] < 0) | |
141 | inflight[1] = 0; | |
bf0ddaba OS |
142 | } |
143 | ||
807d4af2 CH |
144 | struct hd_struct *__disk_get_part(struct gendisk *disk, int partno) |
145 | { | |
146 | struct disk_part_tbl *ptbl = rcu_dereference(disk->part_tbl); | |
147 | ||
148 | if (unlikely(partno < 0 || partno >= ptbl->len)) | |
149 | return NULL; | |
150 | return rcu_dereference(ptbl->part[partno]); | |
151 | } | |
152 | ||
e71bf0d0 TH |
153 | /** |
154 | * disk_get_part - get partition | |
155 | * @disk: disk to look partition from | |
156 | * @partno: partition number | |
157 | * | |
158 | * Look for partition @partno from @disk. If found, increment | |
159 | * reference count and return it. | |
160 | * | |
161 | * CONTEXT: | |
162 | * Don't care. | |
163 | * | |
164 | * RETURNS: | |
165 | * Pointer to the found partition on success, NULL if not found. | |
166 | */ | |
167 | struct hd_struct *disk_get_part(struct gendisk *disk, int partno) | |
168 | { | |
807d4af2 | 169 | struct hd_struct *part; |
540eed56 | 170 | |
e71bf0d0 | 171 | rcu_read_lock(); |
807d4af2 CH |
172 | part = __disk_get_part(disk, partno); |
173 | if (part) | |
174 | get_device(part_to_dev(part)); | |
e71bf0d0 TH |
175 | rcu_read_unlock(); |
176 | ||
177 | return part; | |
178 | } | |
e71bf0d0 TH |
179 | |
180 | /** | |
181 | * disk_part_iter_init - initialize partition iterator | |
182 | * @piter: iterator to initialize | |
183 | * @disk: disk to iterate over | |
184 | * @flags: DISK_PITER_* flags | |
185 | * | |
186 | * Initialize @piter so that it iterates over partitions of @disk. | |
187 | * | |
188 | * CONTEXT: | |
189 | * Don't care. | |
190 | */ | |
191 | void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk, | |
192 | unsigned int flags) | |
193 | { | |
540eed56 TH |
194 | struct disk_part_tbl *ptbl; |
195 | ||
196 | rcu_read_lock(); | |
197 | ptbl = rcu_dereference(disk->part_tbl); | |
198 | ||
e71bf0d0 TH |
199 | piter->disk = disk; |
200 | piter->part = NULL; | |
201 | ||
202 | if (flags & DISK_PITER_REVERSE) | |
540eed56 | 203 | piter->idx = ptbl->len - 1; |
71982a40 | 204 | else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0)) |
e71bf0d0 | 205 | piter->idx = 0; |
b5d0b9df TH |
206 | else |
207 | piter->idx = 1; | |
e71bf0d0 TH |
208 | |
209 | piter->flags = flags; | |
540eed56 TH |
210 | |
211 | rcu_read_unlock(); | |
e71bf0d0 TH |
212 | } |
213 | EXPORT_SYMBOL_GPL(disk_part_iter_init); | |
214 | ||
215 | /** | |
216 | * disk_part_iter_next - proceed iterator to the next partition and return it | |
217 | * @piter: iterator of interest | |
218 | * | |
219 | * Proceed @piter to the next partition and return it. | |
220 | * | |
221 | * CONTEXT: | |
222 | * Don't care. | |
223 | */ | |
224 | struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter) | |
225 | { | |
540eed56 | 226 | struct disk_part_tbl *ptbl; |
e71bf0d0 TH |
227 | int inc, end; |
228 | ||
229 | /* put the last partition */ | |
e79319af | 230 | disk_part_iter_exit(piter); |
e71bf0d0 | 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]; | |
a160c615 | 393 | void (*probe)(dev_t devt); |
68eef3b4 | 394 | } *major_names[BLKDEV_MAJOR_HASH_SIZE]; |
e49fbbbf | 395 | static DEFINE_MUTEX(major_names_lock); |
1da177e4 LT |
396 | |
397 | /* index in the above - for now: assume no multimajor ranges */ | |
e61eb2e9 | 398 | static inline int major_to_index(unsigned major) |
1da177e4 | 399 | { |
68eef3b4 | 400 | return major % BLKDEV_MAJOR_HASH_SIZE; |
7170be5f NH |
401 | } |
402 | ||
68eef3b4 | 403 | #ifdef CONFIG_PROC_FS |
cf771cb5 | 404 | void blkdev_show(struct seq_file *seqf, off_t offset) |
7170be5f | 405 | { |
68eef3b4 | 406 | struct blk_major_name *dp; |
7170be5f | 407 | |
e49fbbbf | 408 | mutex_lock(&major_names_lock); |
133d55cd LG |
409 | for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next) |
410 | if (dp->major == offset) | |
cf771cb5 | 411 | seq_printf(seqf, "%3d %s\n", dp->major, dp->name); |
e49fbbbf | 412 | mutex_unlock(&major_names_lock); |
1da177e4 | 413 | } |
68eef3b4 | 414 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 415 | |
9e8c0bcc | 416 | /** |
e2b6b301 | 417 | * __register_blkdev - register a new block device |
9e8c0bcc | 418 | * |
f33ff110 SB |
419 | * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If |
420 | * @major = 0, try to allocate any unused major number. | |
9e8c0bcc | 421 | * @name: the name of the new block device as a zero terminated string |
e2b6b301 | 422 | * @probe: allback that is called on access to any minor number of @major |
9e8c0bcc MN |
423 | * |
424 | * The @name must be unique within the system. | |
425 | * | |
0e056eb5 | 426 | * The return value depends on the @major input parameter: |
427 | * | |
f33ff110 SB |
428 | * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1] |
429 | * then the function returns zero on success, or a negative error code | |
0e056eb5 | 430 | * - if any unused major number was requested with @major = 0 parameter |
9e8c0bcc | 431 | * then the return value is the allocated major number in range |
f33ff110 SB |
432 | * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise |
433 | * | |
434 | * See Documentation/admin-guide/devices.txt for the list of allocated | |
435 | * major numbers. | |
e2b6b301 CH |
436 | * |
437 | * Use register_blkdev instead for any new code. | |
9e8c0bcc | 438 | */ |
a160c615 CH |
439 | int __register_blkdev(unsigned int major, const char *name, |
440 | void (*probe)(dev_t devt)) | |
1da177e4 LT |
441 | { |
442 | struct blk_major_name **n, *p; | |
443 | int index, ret = 0; | |
444 | ||
e49fbbbf | 445 | mutex_lock(&major_names_lock); |
1da177e4 LT |
446 | |
447 | /* temporary */ | |
448 | if (major == 0) { | |
449 | for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) { | |
450 | if (major_names[index] == NULL) | |
451 | break; | |
452 | } | |
453 | ||
454 | if (index == 0) { | |
dfc76d11 KP |
455 | printk("%s: failed to get major for %s\n", |
456 | __func__, name); | |
1da177e4 LT |
457 | ret = -EBUSY; |
458 | goto out; | |
459 | } | |
460 | major = index; | |
461 | ret = major; | |
462 | } | |
463 | ||
133d55cd | 464 | if (major >= BLKDEV_MAJOR_MAX) { |
dfc76d11 KP |
465 | pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n", |
466 | __func__, major, BLKDEV_MAJOR_MAX-1, name); | |
133d55cd LG |
467 | |
468 | ret = -EINVAL; | |
469 | goto out; | |
470 | } | |
471 | ||
1da177e4 LT |
472 | p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL); |
473 | if (p == NULL) { | |
474 | ret = -ENOMEM; | |
475 | goto out; | |
476 | } | |
477 | ||
478 | p->major = major; | |
a160c615 | 479 | p->probe = probe; |
1da177e4 LT |
480 | strlcpy(p->name, name, sizeof(p->name)); |
481 | p->next = NULL; | |
482 | index = major_to_index(major); | |
483 | ||
484 | for (n = &major_names[index]; *n; n = &(*n)->next) { | |
485 | if ((*n)->major == major) | |
486 | break; | |
487 | } | |
488 | if (!*n) | |
489 | *n = p; | |
490 | else | |
491 | ret = -EBUSY; | |
492 | ||
493 | if (ret < 0) { | |
f33ff110 | 494 | printk("register_blkdev: cannot get major %u for %s\n", |
1da177e4 LT |
495 | major, name); |
496 | kfree(p); | |
497 | } | |
498 | out: | |
e49fbbbf | 499 | mutex_unlock(&major_names_lock); |
1da177e4 LT |
500 | return ret; |
501 | } | |
a160c615 | 502 | EXPORT_SYMBOL(__register_blkdev); |
1da177e4 | 503 | |
f4480240 | 504 | void unregister_blkdev(unsigned int major, const char *name) |
1da177e4 LT |
505 | { |
506 | struct blk_major_name **n; | |
507 | struct blk_major_name *p = NULL; | |
508 | int index = major_to_index(major); | |
1da177e4 | 509 | |
e49fbbbf | 510 | mutex_lock(&major_names_lock); |
1da177e4 LT |
511 | for (n = &major_names[index]; *n; n = &(*n)->next) |
512 | if ((*n)->major == major) | |
513 | break; | |
294462a5 AM |
514 | if (!*n || strcmp((*n)->name, name)) { |
515 | WARN_ON(1); | |
294462a5 | 516 | } else { |
1da177e4 LT |
517 | p = *n; |
518 | *n = p->next; | |
519 | } | |
e49fbbbf | 520 | mutex_unlock(&major_names_lock); |
1da177e4 | 521 | kfree(p); |
1da177e4 LT |
522 | } |
523 | ||
524 | EXPORT_SYMBOL(unregister_blkdev); | |
525 | ||
870d6656 TH |
526 | /** |
527 | * blk_mangle_minor - scatter minor numbers apart | |
528 | * @minor: minor number to mangle | |
529 | * | |
530 | * Scatter consecutively allocated @minor number apart if MANGLE_DEVT | |
531 | * is enabled. Mangling twice gives the original value. | |
532 | * | |
533 | * RETURNS: | |
534 | * Mangled value. | |
535 | * | |
536 | * CONTEXT: | |
537 | * Don't care. | |
538 | */ | |
539 | static int blk_mangle_minor(int minor) | |
540 | { | |
541 | #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT | |
542 | int i; | |
543 | ||
544 | for (i = 0; i < MINORBITS / 2; i++) { | |
545 | int low = minor & (1 << i); | |
546 | int high = minor & (1 << (MINORBITS - 1 - i)); | |
547 | int distance = MINORBITS - 1 - 2 * i; | |
548 | ||
549 | minor ^= low | high; /* clear both bits */ | |
550 | low <<= distance; /* swap the positions */ | |
551 | high >>= distance; | |
552 | minor |= low | high; /* and set */ | |
553 | } | |
554 | #endif | |
555 | return minor; | |
556 | } | |
557 | ||
bcce3de1 TH |
558 | /** |
559 | * blk_alloc_devt - allocate a dev_t for a partition | |
560 | * @part: partition to allocate dev_t for | |
bcce3de1 TH |
561 | * @devt: out parameter for resulting dev_t |
562 | * | |
563 | * Allocate a dev_t for block device. | |
564 | * | |
565 | * RETURNS: | |
566 | * 0 on success, allocated dev_t is returned in *@devt. -errno on | |
567 | * failure. | |
568 | * | |
569 | * CONTEXT: | |
570 | * Might sleep. | |
571 | */ | |
572 | int blk_alloc_devt(struct hd_struct *part, dev_t *devt) | |
573 | { | |
574 | struct gendisk *disk = part_to_disk(part); | |
bab998d6 | 575 | int idx; |
bcce3de1 TH |
576 | |
577 | /* in consecutive minor range? */ | |
578 | if (part->partno < disk->minors) { | |
579 | *devt = MKDEV(disk->major, disk->first_minor + part->partno); | |
580 | return 0; | |
581 | } | |
582 | ||
583 | /* allocate ext devt */ | |
2da78092 KB |
584 | idr_preload(GFP_KERNEL); |
585 | ||
4d66e5e9 | 586 | spin_lock_bh(&ext_devt_lock); |
2da78092 | 587 | idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_NOWAIT); |
4d66e5e9 | 588 | spin_unlock_bh(&ext_devt_lock); |
2da78092 KB |
589 | |
590 | idr_preload_end(); | |
bab998d6 TH |
591 | if (idx < 0) |
592 | return idx == -ENOSPC ? -EBUSY : idx; | |
bcce3de1 | 593 | |
870d6656 | 594 | *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx)); |
bcce3de1 TH |
595 | return 0; |
596 | } | |
597 | ||
598 | /** | |
599 | * blk_free_devt - free a dev_t | |
600 | * @devt: dev_t to free | |
601 | * | |
602 | * Free @devt which was allocated using blk_alloc_devt(). | |
603 | * | |
604 | * CONTEXT: | |
605 | * Might sleep. | |
606 | */ | |
607 | void blk_free_devt(dev_t devt) | |
608 | { | |
bcce3de1 TH |
609 | if (devt == MKDEV(0, 0)) |
610 | return; | |
611 | ||
612 | if (MAJOR(devt) == BLOCK_EXT_MAJOR) { | |
4d66e5e9 | 613 | spin_lock_bh(&ext_devt_lock); |
870d6656 | 614 | idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); |
4d66e5e9 | 615 | spin_unlock_bh(&ext_devt_lock); |
bcce3de1 TH |
616 | } |
617 | } | |
618 | ||
33c826ef BVA |
619 | /* |
620 | * We invalidate devt by assigning NULL pointer for devt in idr. | |
6fcc44d1 YY |
621 | */ |
622 | void blk_invalidate_devt(dev_t devt) | |
623 | { | |
624 | if (MAJOR(devt) == BLOCK_EXT_MAJOR) { | |
625 | spin_lock_bh(&ext_devt_lock); | |
626 | idr_replace(&ext_devt_idr, NULL, blk_mangle_minor(MINOR(devt))); | |
627 | spin_unlock_bh(&ext_devt_lock); | |
628 | } | |
629 | } | |
630 | ||
1f014290 TH |
631 | static char *bdevt_str(dev_t devt, char *buf) |
632 | { | |
633 | if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) { | |
634 | char tbuf[BDEVT_SIZE]; | |
635 | snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt)); | |
636 | snprintf(buf, BDEVT_SIZE, "%-9s", tbuf); | |
637 | } else | |
638 | snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt)); | |
639 | ||
640 | return buf; | |
641 | } | |
642 | ||
e418de3a CH |
643 | static void blk_register_region(struct gendisk *disk) |
644 | { | |
645 | int i; | |
1da177e4 | 646 | |
e49fbbbf | 647 | mutex_lock(&bdev_map_lock); |
e418de3a CH |
648 | for (i = 0; i < disk->minors; i++) { |
649 | if (xa_insert(&bdev_map, disk_devt(disk) + i, disk, GFP_KERNEL)) | |
650 | WARN_ON_ONCE(1); | |
62b508f8 | 651 | } |
e49fbbbf | 652 | mutex_unlock(&bdev_map_lock); |
62b508f8 | 653 | } |
1da177e4 | 654 | |
e418de3a | 655 | static void blk_unregister_region(struct gendisk *disk) |
1da177e4 | 656 | { |
e418de3a | 657 | int i; |
1da177e4 | 658 | |
e49fbbbf | 659 | mutex_lock(&bdev_map_lock); |
e418de3a CH |
660 | for (i = 0; i < disk->minors; i++) |
661 | xa_erase(&bdev_map, disk_devt(disk) + i); | |
e49fbbbf | 662 | mutex_unlock(&bdev_map_lock); |
1da177e4 LT |
663 | } |
664 | ||
9301fe73 CH |
665 | static void disk_scan_partitions(struct gendisk *disk) |
666 | { | |
667 | struct block_device *bdev; | |
668 | ||
669 | if (!get_capacity(disk) || !disk_part_scan_enabled(disk)) | |
670 | return; | |
671 | ||
672 | set_bit(GD_NEED_PART_SCAN, &disk->state); | |
673 | bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL); | |
674 | if (!IS_ERR(bdev)) | |
675 | blkdev_put(bdev, FMODE_READ); | |
676 | } | |
677 | ||
fef912bf HR |
678 | static void register_disk(struct device *parent, struct gendisk *disk, |
679 | const struct attribute_group **groups) | |
d2bf1b67 TH |
680 | { |
681 | struct device *ddev = disk_to_dev(disk); | |
d2bf1b67 TH |
682 | struct disk_part_iter piter; |
683 | struct hd_struct *part; | |
684 | int err; | |
685 | ||
e63a46be | 686 | ddev->parent = parent; |
d2bf1b67 | 687 | |
ffc8b308 | 688 | dev_set_name(ddev, "%s", disk->disk_name); |
d2bf1b67 TH |
689 | |
690 | /* delay uevents, until we scanned partition table */ | |
691 | dev_set_uevent_suppress(ddev, 1); | |
692 | ||
fef912bf HR |
693 | if (groups) { |
694 | WARN_ON(ddev->groups); | |
695 | ddev->groups = groups; | |
696 | } | |
d2bf1b67 TH |
697 | if (device_add(ddev)) |
698 | return; | |
699 | if (!sysfs_deprecated) { | |
700 | err = sysfs_create_link(block_depr, &ddev->kobj, | |
701 | kobject_name(&ddev->kobj)); | |
702 | if (err) { | |
703 | device_del(ddev); | |
704 | return; | |
705 | } | |
706 | } | |
25e823c8 ML |
707 | |
708 | /* | |
709 | * avoid probable deadlock caused by allocating memory with | |
710 | * GFP_KERNEL in runtime_resume callback of its all ancestor | |
711 | * devices | |
712 | */ | |
713 | pm_runtime_set_memalloc_noio(ddev, true); | |
714 | ||
d2bf1b67 TH |
715 | disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj); |
716 | disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj); | |
717 | ||
8ddcd653 CH |
718 | if (disk->flags & GENHD_FL_HIDDEN) { |
719 | dev_set_uevent_suppress(ddev, 0); | |
720 | return; | |
721 | } | |
722 | ||
9301fe73 | 723 | disk_scan_partitions(disk); |
d2bf1b67 | 724 | |
d2bf1b67 TH |
725 | /* announce disk after possible partitions are created */ |
726 | dev_set_uevent_suppress(ddev, 0); | |
727 | kobject_uevent(&ddev->kobj, KOBJ_ADD); | |
728 | ||
729 | /* announce possible partitions */ | |
730 | disk_part_iter_init(&piter, disk, 0); | |
731 | while ((part = disk_part_iter_next(&piter))) | |
732 | kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD); | |
733 | disk_part_iter_exit(&piter); | |
8ddcd653 | 734 | |
4d7c1d3f | 735 | if (disk->queue->backing_dev_info->dev) { |
736 | err = sysfs_create_link(&ddev->kobj, | |
737 | &disk->queue->backing_dev_info->dev->kobj, | |
738 | "bdi"); | |
739 | WARN_ON(err); | |
740 | } | |
d2bf1b67 TH |
741 | } |
742 | ||
1da177e4 | 743 | /** |
fa70d2e2 | 744 | * __device_add_disk - add disk information to kernel list |
e63a46be | 745 | * @parent: parent device for the disk |
1da177e4 | 746 | * @disk: per-device partitioning information |
fef912bf | 747 | * @groups: Additional per-device sysfs groups |
fa70d2e2 | 748 | * @register_queue: register the queue if set to true |
1da177e4 LT |
749 | * |
750 | * This function registers the partitioning information in @disk | |
751 | * with the kernel. | |
3e1a7ff8 TH |
752 | * |
753 | * FIXME: error handling | |
1da177e4 | 754 | */ |
fa70d2e2 | 755 | static void __device_add_disk(struct device *parent, struct gendisk *disk, |
fef912bf | 756 | const struct attribute_group **groups, |
fa70d2e2 | 757 | bool register_queue) |
1da177e4 | 758 | { |
3e1a7ff8 | 759 | dev_t devt; |
6ffeea77 | 760 | int retval; |
cf0ca9fe | 761 | |
737eb78e DLM |
762 | /* |
763 | * The disk queue should now be all set with enough information about | |
764 | * the device for the elevator code to pick an adequate default | |
765 | * elevator if one is needed, that is, for devices requesting queue | |
766 | * registration. | |
767 | */ | |
768 | if (register_queue) | |
769 | elevator_init_mq(disk->queue); | |
770 | ||
3e1a7ff8 TH |
771 | /* minors == 0 indicates to use ext devt from part0 and should |
772 | * be accompanied with EXT_DEVT flag. Make sure all | |
773 | * parameters make sense. | |
774 | */ | |
775 | WARN_ON(disk->minors && !(disk->major || disk->first_minor)); | |
8ddcd653 CH |
776 | WARN_ON(!disk->minors && |
777 | !(disk->flags & (GENHD_FL_EXT_DEVT | GENHD_FL_HIDDEN))); | |
3e1a7ff8 | 778 | |
1da177e4 | 779 | disk->flags |= GENHD_FL_UP; |
3e1a7ff8 TH |
780 | |
781 | retval = blk_alloc_devt(&disk->part0, &devt); | |
782 | if (retval) { | |
783 | WARN_ON(1); | |
784 | return; | |
785 | } | |
3e1a7ff8 TH |
786 | disk->major = MAJOR(devt); |
787 | disk->first_minor = MINOR(devt); | |
788 | ||
9f53d2fe SG |
789 | disk_alloc_events(disk); |
790 | ||
8ddcd653 CH |
791 | if (disk->flags & GENHD_FL_HIDDEN) { |
792 | /* | |
793 | * Don't let hidden disks show up in /proc/partitions, | |
794 | * and don't bother scanning for partitions either. | |
795 | */ | |
796 | disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO; | |
797 | disk->flags |= GENHD_FL_NO_PART_SCAN; | |
798 | } else { | |
3c5d202b CH |
799 | struct backing_dev_info *bdi = disk->queue->backing_dev_info; |
800 | struct device *dev = disk_to_dev(disk); | |
3a92168b | 801 | int ret; |
802 | ||
8ddcd653 | 803 | /* Register BDI before referencing it from bdev */ |
3c5d202b CH |
804 | dev->devt = devt; |
805 | ret = bdi_register(bdi, "%u:%u", MAJOR(devt), MINOR(devt)); | |
3a92168b | 806 | WARN_ON(ret); |
3c5d202b | 807 | bdi_set_owner(bdi, dev); |
e418de3a | 808 | blk_register_region(disk); |
8ddcd653 | 809 | } |
fef912bf | 810 | register_disk(parent, disk, groups); |
fa70d2e2 MS |
811 | if (register_queue) |
812 | blk_register_queue(disk); | |
cf0ca9fe | 813 | |
523e1d39 TH |
814 | /* |
815 | * Take an extra ref on queue which will be put on disk_release() | |
816 | * so that it sticks around as long as @disk is there. | |
817 | */ | |
09ac46c4 | 818 | WARN_ON_ONCE(!blk_get_queue(disk->queue)); |
523e1d39 | 819 | |
77ea887e | 820 | disk_add_events(disk); |
25520d55 | 821 | blk_integrity_add(disk); |
1da177e4 | 822 | } |
fa70d2e2 | 823 | |
fef912bf HR |
824 | void device_add_disk(struct device *parent, struct gendisk *disk, |
825 | const struct attribute_group **groups) | |
826 | ||
fa70d2e2 | 827 | { |
fef912bf | 828 | __device_add_disk(parent, disk, groups, true); |
fa70d2e2 | 829 | } |
e63a46be | 830 | EXPORT_SYMBOL(device_add_disk); |
1da177e4 | 831 | |
fa70d2e2 MS |
832 | void device_add_disk_no_queue_reg(struct device *parent, struct gendisk *disk) |
833 | { | |
fef912bf | 834 | __device_add_disk(parent, disk, NULL, false); |
fa70d2e2 MS |
835 | } |
836 | EXPORT_SYMBOL(device_add_disk_no_queue_reg); | |
837 | ||
02d33b67 CH |
838 | static void invalidate_partition(struct gendisk *disk, int partno) |
839 | { | |
840 | struct block_device *bdev; | |
841 | ||
842 | bdev = bdget_disk(disk, partno); | |
843 | if (!bdev) | |
844 | return; | |
845 | ||
846 | fsync_bdev(bdev); | |
847 | __invalidate_device(bdev, true); | |
9bc5c397 CH |
848 | |
849 | /* | |
850 | * Unhash the bdev inode for this device so that it gets evicted as soon | |
851 | * as last inode reference is dropped. | |
852 | */ | |
853 | remove_inode_hash(bdev->bd_inode); | |
02d33b67 CH |
854 | bdput(bdev); |
855 | } | |
856 | ||
b5bd357c LC |
857 | /** |
858 | * del_gendisk - remove the gendisk | |
859 | * @disk: the struct gendisk to remove | |
860 | * | |
861 | * Removes the gendisk and all its associated resources. This deletes the | |
862 | * partitions associated with the gendisk, and unregisters the associated | |
863 | * request_queue. | |
864 | * | |
865 | * This is the counter to the respective __device_add_disk() call. | |
866 | * | |
867 | * The final removal of the struct gendisk happens when its refcount reaches 0 | |
868 | * with put_disk(), which should be called after del_gendisk(), if | |
869 | * __device_add_disk() was used. | |
e8c7d14a LC |
870 | * |
871 | * Drivers exist which depend on the release of the gendisk to be synchronous, | |
872 | * it should not be deferred. | |
873 | * | |
874 | * Context: can sleep | |
b5bd357c | 875 | */ |
d2bf1b67 | 876 | void del_gendisk(struct gendisk *disk) |
1da177e4 | 877 | { |
d2bf1b67 TH |
878 | struct disk_part_iter piter; |
879 | struct hd_struct *part; | |
880 | ||
e8c7d14a LC |
881 | might_sleep(); |
882 | ||
6b3ba976 CH |
883 | if (WARN_ON_ONCE(!disk->queue)) |
884 | return; | |
885 | ||
25520d55 | 886 | blk_integrity_del(disk); |
77ea887e TH |
887 | disk_del_events(disk); |
888 | ||
56c0908c JK |
889 | /* |
890 | * Block lookups of the disk until all bdevs are unhashed and the | |
891 | * disk is marked as dead (GENHD_FL_UP cleared). | |
892 | */ | |
893 | down_write(&disk->lookup_sem); | |
d2bf1b67 TH |
894 | /* invalidate stuff */ |
895 | disk_part_iter_init(&piter, disk, | |
896 | DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE); | |
897 | while ((part = disk_part_iter_next(&piter))) { | |
898 | invalidate_partition(disk, part->partno); | |
8328eb28 | 899 | delete_partition(part); |
d2bf1b67 TH |
900 | } |
901 | disk_part_iter_exit(&piter); | |
902 | ||
903 | invalidate_partition(disk, 0); | |
d2bf1b67 TH |
904 | set_capacity(disk, 0); |
905 | disk->flags &= ~GENHD_FL_UP; | |
56c0908c | 906 | up_write(&disk->lookup_sem); |
d2bf1b67 | 907 | |
6b3ba976 | 908 | if (!(disk->flags & GENHD_FL_HIDDEN)) { |
8ddcd653 | 909 | sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi"); |
6b3ba976 | 910 | |
90f16fdd JK |
911 | /* |
912 | * Unregister bdi before releasing device numbers (as they can | |
913 | * get reused and we'd get clashes in sysfs). | |
914 | */ | |
6b3ba976 | 915 | bdi_unregister(disk->queue->backing_dev_info); |
90f16fdd | 916 | } |
d2bf1b67 | 917 | |
6b3ba976 CH |
918 | blk_unregister_queue(disk); |
919 | ||
17eac099 | 920 | if (!(disk->flags & GENHD_FL_HIDDEN)) |
e418de3a | 921 | blk_unregister_region(disk); |
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 | ||
bd8eff3b CH |
967 | static void request_gendisk_module(dev_t devt) |
968 | { | |
a160c615 CH |
969 | unsigned int major = MAJOR(devt); |
970 | struct blk_major_name **n; | |
971 | ||
972 | mutex_lock(&major_names_lock); | |
973 | for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) { | |
974 | if ((*n)->major == major && (*n)->probe) { | |
975 | (*n)->probe(devt); | |
976 | mutex_unlock(&major_names_lock); | |
977 | return; | |
978 | } | |
979 | } | |
980 | mutex_unlock(&major_names_lock); | |
981 | ||
bd8eff3b CH |
982 | if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0) |
983 | /* Make old-style 2.4 aliases work */ | |
984 | request_module("block-major-%d", MAJOR(devt)); | |
985 | } | |
986 | ||
e418de3a | 987 | static bool get_disk_and_module(struct gendisk *disk) |
62b508f8 | 988 | { |
e418de3a | 989 | struct module *owner; |
bd8eff3b | 990 | |
e418de3a CH |
991 | if (!disk->fops) |
992 | return false; | |
993 | owner = disk->fops->owner; | |
994 | if (owner && !try_module_get(owner)) | |
995 | return false; | |
996 | if (!kobject_get_unless_zero(&disk_to_dev(disk)->kobj)) { | |
62b508f8 | 997 | module_put(owner); |
e418de3a | 998 | return false; |
62b508f8 | 999 | } |
e418de3a | 1000 | return true; |
62b508f8 | 1001 | |
e418de3a | 1002 | } |
62b508f8 | 1003 | |
1da177e4 LT |
1004 | /** |
1005 | * get_gendisk - get partitioning information for a given device | |
710027a4 | 1006 | * @devt: device to get partitioning information for |
496aa8a9 | 1007 | * @partno: returned partition index |
1da177e4 LT |
1008 | * |
1009 | * This function gets the structure containing partitioning | |
710027a4 | 1010 | * information for the given device @devt. |
763b5892 LC |
1011 | * |
1012 | * Context: can sleep | |
1da177e4 | 1013 | */ |
cf771cb5 | 1014 | struct gendisk *get_gendisk(dev_t devt, int *partno) |
1da177e4 | 1015 | { |
bcce3de1 TH |
1016 | struct gendisk *disk = NULL; |
1017 | ||
763b5892 LC |
1018 | might_sleep(); |
1019 | ||
bcce3de1 | 1020 | if (MAJOR(devt) != BLOCK_EXT_MAJOR) { |
e418de3a CH |
1021 | mutex_lock(&bdev_map_lock); |
1022 | disk = xa_load(&bdev_map, devt); | |
1023 | if (!disk) { | |
1024 | mutex_unlock(&bdev_map_lock); | |
1025 | request_gendisk_module(devt); | |
1026 | mutex_lock(&bdev_map_lock); | |
1027 | disk = xa_load(&bdev_map, devt); | |
1028 | } | |
1029 | if (disk && !get_disk_and_module(disk)) | |
1030 | disk = NULL; | |
1031 | if (disk) | |
1032 | *partno = devt - disk_devt(disk); | |
1033 | mutex_unlock(&bdev_map_lock); | |
bcce3de1 TH |
1034 | } else { |
1035 | struct hd_struct *part; | |
1036 | ||
4d66e5e9 | 1037 | spin_lock_bh(&ext_devt_lock); |
870d6656 | 1038 | part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); |
3079c22e | 1039 | if (part && get_disk_and_module(part_to_disk(part))) { |
bcce3de1 TH |
1040 | *partno = part->partno; |
1041 | disk = part_to_disk(part); | |
1042 | } | |
4d66e5e9 | 1043 | spin_unlock_bh(&ext_devt_lock); |
bcce3de1 | 1044 | } |
edfaa7c3 | 1045 | |
56c0908c JK |
1046 | if (!disk) |
1047 | return NULL; | |
1048 | ||
1049 | /* | |
1050 | * Synchronize with del_gendisk() to not return disk that is being | |
1051 | * destroyed. | |
1052 | */ | |
1053 | down_read(&disk->lookup_sem); | |
1054 | if (unlikely((disk->flags & GENHD_FL_HIDDEN) || | |
1055 | !(disk->flags & GENHD_FL_UP))) { | |
1056 | up_read(&disk->lookup_sem); | |
9df6c299 | 1057 | put_disk_and_module(disk); |
8ddcd653 | 1058 | disk = NULL; |
56c0908c JK |
1059 | } else { |
1060 | up_read(&disk->lookup_sem); | |
8ddcd653 | 1061 | } |
bcce3de1 | 1062 | return disk; |
1da177e4 LT |
1063 | } |
1064 | ||
f331c029 TH |
1065 | /** |
1066 | * bdget_disk - do bdget() by gendisk and partition number | |
1067 | * @disk: gendisk of interest | |
1068 | * @partno: partition number | |
1069 | * | |
1070 | * Find partition @partno from @disk, do bdget() on it. | |
1071 | * | |
1072 | * CONTEXT: | |
1073 | * Don't care. | |
1074 | * | |
1075 | * RETURNS: | |
1076 | * Resulting block_device on success, NULL on failure. | |
1077 | */ | |
aeb3d3a8 | 1078 | struct block_device *bdget_disk(struct gendisk *disk, int partno) |
f331c029 | 1079 | { |
548b10eb TH |
1080 | struct hd_struct *part; |
1081 | struct block_device *bdev = NULL; | |
f331c029 | 1082 | |
548b10eb | 1083 | part = disk_get_part(disk, partno); |
2bbedcb4 | 1084 | if (part) |
10ed1666 | 1085 | bdev = bdget_part(part); |
548b10eb | 1086 | disk_put_part(part); |
f331c029 | 1087 | |
548b10eb | 1088 | return bdev; |
f331c029 TH |
1089 | } |
1090 | EXPORT_SYMBOL(bdget_disk); | |
1091 | ||
5c6f35c5 GKH |
1092 | /* |
1093 | * print a full list of all partitions - intended for places where the root | |
1094 | * filesystem can't be mounted and thus to give the victim some idea of what | |
1095 | * went wrong | |
1096 | */ | |
1097 | void __init printk_all_partitions(void) | |
1098 | { | |
def4e38d TH |
1099 | struct class_dev_iter iter; |
1100 | struct device *dev; | |
1101 | ||
1102 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); | |
1103 | while ((dev = class_dev_iter_next(&iter))) { | |
1104 | struct gendisk *disk = dev_to_disk(dev); | |
e71bf0d0 TH |
1105 | struct disk_part_iter piter; |
1106 | struct hd_struct *part; | |
1f014290 TH |
1107 | char name_buf[BDEVNAME_SIZE]; |
1108 | char devt_buf[BDEVT_SIZE]; | |
def4e38d TH |
1109 | |
1110 | /* | |
1111 | * Don't show empty devices or things that have been | |
25985edc | 1112 | * suppressed |
def4e38d TH |
1113 | */ |
1114 | if (get_capacity(disk) == 0 || | |
1115 | (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)) | |
1116 | continue; | |
1117 | ||
1118 | /* | |
1119 | * Note, unlike /proc/partitions, I am showing the | |
1120 | * numbers in hex - the same format as the root= | |
1121 | * option takes. | |
1122 | */ | |
074a7aca TH |
1123 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0); |
1124 | while ((part = disk_part_iter_next(&piter))) { | |
1125 | bool is_part0 = part == &disk->part0; | |
def4e38d | 1126 | |
b5af921e | 1127 | printk("%s%s %10llu %s %s", is_part0 ? "" : " ", |
1f014290 | 1128 | bdevt_str(part_devt(part), devt_buf), |
c83f6bf9 VG |
1129 | (unsigned long long)part_nr_sects_read(part) >> 1 |
1130 | , disk_name(disk, part->partno, name_buf), | |
1ad7e899 | 1131 | part->info ? part->info->uuid : ""); |
074a7aca | 1132 | if (is_part0) { |
52c44d93 | 1133 | if (dev->parent && dev->parent->driver) |
074a7aca | 1134 | printk(" driver: %s\n", |
52c44d93 | 1135 | dev->parent->driver->name); |
074a7aca TH |
1136 | else |
1137 | printk(" (driver?)\n"); | |
1138 | } else | |
1139 | printk("\n"); | |
1140 | } | |
e71bf0d0 | 1141 | disk_part_iter_exit(&piter); |
def4e38d TH |
1142 | } |
1143 | class_dev_iter_exit(&iter); | |
dd2a345f DG |
1144 | } |
1145 | ||
1da177e4 LT |
1146 | #ifdef CONFIG_PROC_FS |
1147 | /* iterator */ | |
def4e38d | 1148 | static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos) |
68c4d4a7 | 1149 | { |
def4e38d TH |
1150 | loff_t skip = *pos; |
1151 | struct class_dev_iter *iter; | |
1152 | struct device *dev; | |
68c4d4a7 | 1153 | |
aeb3d3a8 | 1154 | iter = kmalloc(sizeof(*iter), GFP_KERNEL); |
def4e38d TH |
1155 | if (!iter) |
1156 | return ERR_PTR(-ENOMEM); | |
1157 | ||
1158 | seqf->private = iter; | |
1159 | class_dev_iter_init(iter, &block_class, NULL, &disk_type); | |
1160 | do { | |
1161 | dev = class_dev_iter_next(iter); | |
1162 | if (!dev) | |
1163 | return NULL; | |
1164 | } while (skip--); | |
1165 | ||
1166 | return dev_to_disk(dev); | |
68c4d4a7 GKH |
1167 | } |
1168 | ||
def4e38d | 1169 | static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos) |
1da177e4 | 1170 | { |
edfaa7c3 | 1171 | struct device *dev; |
1da177e4 | 1172 | |
def4e38d TH |
1173 | (*pos)++; |
1174 | dev = class_dev_iter_next(seqf->private); | |
2ac3cee5 | 1175 | if (dev) |
68c4d4a7 | 1176 | return dev_to_disk(dev); |
2ac3cee5 | 1177 | |
1da177e4 LT |
1178 | return NULL; |
1179 | } | |
1180 | ||
def4e38d | 1181 | static void disk_seqf_stop(struct seq_file *seqf, void *v) |
27f30251 | 1182 | { |
def4e38d | 1183 | struct class_dev_iter *iter = seqf->private; |
27f30251 | 1184 | |
def4e38d TH |
1185 | /* stop is called even after start failed :-( */ |
1186 | if (iter) { | |
1187 | class_dev_iter_exit(iter); | |
1188 | kfree(iter); | |
77da1605 | 1189 | seqf->private = NULL; |
5c0ef6d0 | 1190 | } |
1da177e4 LT |
1191 | } |
1192 | ||
def4e38d | 1193 | static void *show_partition_start(struct seq_file *seqf, loff_t *pos) |
1da177e4 | 1194 | { |
06768067 | 1195 | void *p; |
def4e38d TH |
1196 | |
1197 | p = disk_seqf_start(seqf, pos); | |
b9f985b6 | 1198 | if (!IS_ERR_OR_NULL(p) && !*pos) |
def4e38d TH |
1199 | seq_puts(seqf, "major minor #blocks name\n\n"); |
1200 | return p; | |
1da177e4 LT |
1201 | } |
1202 | ||
cf771cb5 | 1203 | static int show_partition(struct seq_file *seqf, void *v) |
1da177e4 LT |
1204 | { |
1205 | struct gendisk *sgp = v; | |
e71bf0d0 TH |
1206 | struct disk_part_iter piter; |
1207 | struct hd_struct *part; | |
1da177e4 LT |
1208 | char buf[BDEVNAME_SIZE]; |
1209 | ||
1da177e4 | 1210 | /* Don't show non-partitionable removeable devices or empty devices */ |
d27769ec | 1211 | if (!get_capacity(sgp) || (!disk_max_parts(sgp) && |
f331c029 | 1212 | (sgp->flags & GENHD_FL_REMOVABLE))) |
1da177e4 LT |
1213 | return 0; |
1214 | if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO) | |
1215 | return 0; | |
1216 | ||
1217 | /* show the full disk and all non-0 size partitions of it */ | |
074a7aca | 1218 | disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0); |
e71bf0d0 | 1219 | while ((part = disk_part_iter_next(&piter))) |
1f014290 | 1220 | seq_printf(seqf, "%4d %7d %10llu %s\n", |
f331c029 | 1221 | MAJOR(part_devt(part)), MINOR(part_devt(part)), |
c83f6bf9 | 1222 | (unsigned long long)part_nr_sects_read(part) >> 1, |
f331c029 | 1223 | disk_name(sgp, part->partno, buf)); |
e71bf0d0 | 1224 | disk_part_iter_exit(&piter); |
1da177e4 LT |
1225 | |
1226 | return 0; | |
1227 | } | |
1228 | ||
f500975a | 1229 | static const struct seq_operations partitions_op = { |
def4e38d TH |
1230 | .start = show_partition_start, |
1231 | .next = disk_seqf_next, | |
1232 | .stop = disk_seqf_stop, | |
edfaa7c3 | 1233 | .show = show_partition |
1da177e4 LT |
1234 | }; |
1235 | #endif | |
1236 | ||
1da177e4 LT |
1237 | static int __init genhd_device_init(void) |
1238 | { | |
e105b8bf DW |
1239 | int error; |
1240 | ||
1241 | block_class.dev_kobj = sysfs_dev_block_kobj; | |
1242 | error = class_register(&block_class); | |
ee27a558 RM |
1243 | if (unlikely(error)) |
1244 | return error; | |
1da177e4 | 1245 | blk_dev_init(); |
edfaa7c3 | 1246 | |
561ec68e ZY |
1247 | register_blkdev(BLOCK_EXT_MAJOR, "blkext"); |
1248 | ||
edfaa7c3 | 1249 | /* create top-level block dir */ |
e52eec13 AK |
1250 | if (!sysfs_deprecated) |
1251 | block_depr = kobject_create_and_add("block", NULL); | |
830d3cfb | 1252 | return 0; |
1da177e4 LT |
1253 | } |
1254 | ||
1255 | subsys_initcall(genhd_device_init); | |
1256 | ||
edfaa7c3 KS |
1257 | static ssize_t disk_range_show(struct device *dev, |
1258 | struct device_attribute *attr, char *buf) | |
1da177e4 | 1259 | { |
edfaa7c3 | 1260 | struct gendisk *disk = dev_to_disk(dev); |
1da177e4 | 1261 | |
edfaa7c3 | 1262 | return sprintf(buf, "%d\n", disk->minors); |
1da177e4 LT |
1263 | } |
1264 | ||
1f014290 TH |
1265 | static ssize_t disk_ext_range_show(struct device *dev, |
1266 | struct device_attribute *attr, char *buf) | |
1267 | { | |
1268 | struct gendisk *disk = dev_to_disk(dev); | |
1269 | ||
b5d0b9df | 1270 | return sprintf(buf, "%d\n", disk_max_parts(disk)); |
1f014290 TH |
1271 | } |
1272 | ||
edfaa7c3 KS |
1273 | static ssize_t disk_removable_show(struct device *dev, |
1274 | struct device_attribute *attr, char *buf) | |
a7fd6706 | 1275 | { |
edfaa7c3 | 1276 | struct gendisk *disk = dev_to_disk(dev); |
a7fd6706 | 1277 | |
edfaa7c3 KS |
1278 | return sprintf(buf, "%d\n", |
1279 | (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0)); | |
a7fd6706 KS |
1280 | } |
1281 | ||
8ddcd653 CH |
1282 | static ssize_t disk_hidden_show(struct device *dev, |
1283 | struct device_attribute *attr, char *buf) | |
1284 | { | |
1285 | struct gendisk *disk = dev_to_disk(dev); | |
1286 | ||
1287 | return sprintf(buf, "%d\n", | |
1288 | (disk->flags & GENHD_FL_HIDDEN ? 1 : 0)); | |
1289 | } | |
1290 | ||
1c9ce527 KS |
1291 | static ssize_t disk_ro_show(struct device *dev, |
1292 | struct device_attribute *attr, char *buf) | |
1293 | { | |
1294 | struct gendisk *disk = dev_to_disk(dev); | |
1295 | ||
b7db9956 | 1296 | return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0); |
1c9ce527 KS |
1297 | } |
1298 | ||
3ad5cee5 CH |
1299 | ssize_t part_size_show(struct device *dev, |
1300 | struct device_attribute *attr, char *buf) | |
1301 | { | |
1302 | struct hd_struct *p = dev_to_part(dev); | |
1303 | ||
1304 | return sprintf(buf, "%llu\n", | |
1305 | (unsigned long long)part_nr_sects_read(p)); | |
1306 | } | |
1307 | ||
1308 | ssize_t part_stat_show(struct device *dev, | |
1309 | struct device_attribute *attr, char *buf) | |
1310 | { | |
1311 | struct hd_struct *p = dev_to_part(dev); | |
1312 | struct request_queue *q = part_to_disk(p)->queue; | |
ea18e0f0 | 1313 | struct disk_stats stat; |
3ad5cee5 CH |
1314 | unsigned int inflight; |
1315 | ||
ea18e0f0 | 1316 | part_stat_read_all(p, &stat); |
b2f609e1 CH |
1317 | if (queue_is_mq(q)) |
1318 | inflight = blk_mq_in_flight(q, p); | |
1319 | else | |
1f06959b | 1320 | inflight = part_in_flight(p); |
ea18e0f0 | 1321 | |
3ad5cee5 CH |
1322 | return sprintf(buf, |
1323 | "%8lu %8lu %8llu %8u " | |
1324 | "%8lu %8lu %8llu %8u " | |
1325 | "%8u %8u %8u " | |
1326 | "%8lu %8lu %8llu %8u " | |
1327 | "%8lu %8u" | |
1328 | "\n", | |
ea18e0f0 KK |
1329 | stat.ios[STAT_READ], |
1330 | stat.merges[STAT_READ], | |
1331 | (unsigned long long)stat.sectors[STAT_READ], | |
1332 | (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC), | |
1333 | stat.ios[STAT_WRITE], | |
1334 | stat.merges[STAT_WRITE], | |
1335 | (unsigned long long)stat.sectors[STAT_WRITE], | |
1336 | (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC), | |
3ad5cee5 | 1337 | inflight, |
ea18e0f0 | 1338 | jiffies_to_msecs(stat.io_ticks), |
8cd5b8fc KK |
1339 | (unsigned int)div_u64(stat.nsecs[STAT_READ] + |
1340 | stat.nsecs[STAT_WRITE] + | |
1341 | stat.nsecs[STAT_DISCARD] + | |
1342 | stat.nsecs[STAT_FLUSH], | |
1343 | NSEC_PER_MSEC), | |
ea18e0f0 KK |
1344 | stat.ios[STAT_DISCARD], |
1345 | stat.merges[STAT_DISCARD], | |
1346 | (unsigned long long)stat.sectors[STAT_DISCARD], | |
1347 | (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC), | |
1348 | stat.ios[STAT_FLUSH], | |
1349 | (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC)); | |
3ad5cee5 CH |
1350 | } |
1351 | ||
1352 | ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr, | |
1353 | char *buf) | |
1354 | { | |
1355 | struct hd_struct *p = dev_to_part(dev); | |
1356 | struct request_queue *q = part_to_disk(p)->queue; | |
1357 | unsigned int inflight[2]; | |
1358 | ||
b2f609e1 CH |
1359 | if (queue_is_mq(q)) |
1360 | blk_mq_in_flight_rw(q, p, inflight); | |
1361 | else | |
1f06959b | 1362 | part_in_flight_rw(p, inflight); |
b2f609e1 | 1363 | |
3ad5cee5 CH |
1364 | return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]); |
1365 | } | |
1366 | ||
edfaa7c3 KS |
1367 | static ssize_t disk_capability_show(struct device *dev, |
1368 | struct device_attribute *attr, char *buf) | |
86ce18d7 | 1369 | { |
edfaa7c3 KS |
1370 | struct gendisk *disk = dev_to_disk(dev); |
1371 | ||
1372 | return sprintf(buf, "%x\n", disk->flags); | |
86ce18d7 | 1373 | } |
edfaa7c3 | 1374 | |
c72758f3 MP |
1375 | static ssize_t disk_alignment_offset_show(struct device *dev, |
1376 | struct device_attribute *attr, | |
1377 | char *buf) | |
1378 | { | |
1379 | struct gendisk *disk = dev_to_disk(dev); | |
1380 | ||
1381 | return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue)); | |
1382 | } | |
1383 | ||
86b37281 MP |
1384 | static ssize_t disk_discard_alignment_show(struct device *dev, |
1385 | struct device_attribute *attr, | |
1386 | char *buf) | |
1387 | { | |
1388 | struct gendisk *disk = dev_to_disk(dev); | |
1389 | ||
dd3d145d | 1390 | return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue)); |
86b37281 MP |
1391 | } |
1392 | ||
5657a819 JP |
1393 | static DEVICE_ATTR(range, 0444, disk_range_show, NULL); |
1394 | static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL); | |
1395 | static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL); | |
1396 | static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL); | |
1397 | static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL); | |
1398 | static DEVICE_ATTR(size, 0444, part_size_show, NULL); | |
1399 | static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL); | |
1400 | static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL); | |
1401 | static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL); | |
1402 | static DEVICE_ATTR(stat, 0444, part_stat_show, NULL); | |
1403 | static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL); | |
1404 | static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store); | |
3ad5cee5 | 1405 | |
c17bb495 | 1406 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
3ad5cee5 CH |
1407 | ssize_t part_fail_show(struct device *dev, |
1408 | struct device_attribute *attr, char *buf) | |
1409 | { | |
1410 | struct hd_struct *p = dev_to_part(dev); | |
1411 | ||
1412 | return sprintf(buf, "%d\n", p->make_it_fail); | |
1413 | } | |
1414 | ||
1415 | ssize_t part_fail_store(struct device *dev, | |
1416 | struct device_attribute *attr, | |
1417 | const char *buf, size_t count) | |
1418 | { | |
1419 | struct hd_struct *p = dev_to_part(dev); | |
1420 | int i; | |
1421 | ||
1422 | if (count > 0 && sscanf(buf, "%d", &i) > 0) | |
1423 | p->make_it_fail = (i == 0) ? 0 : 1; | |
1424 | ||
1425 | return count; | |
1426 | } | |
1427 | ||
edfaa7c3 | 1428 | static struct device_attribute dev_attr_fail = |
5657a819 | 1429 | __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store); |
3ad5cee5 CH |
1430 | #endif /* CONFIG_FAIL_MAKE_REQUEST */ |
1431 | ||
581d4e28 JA |
1432 | #ifdef CONFIG_FAIL_IO_TIMEOUT |
1433 | static struct device_attribute dev_attr_fail_timeout = | |
5657a819 | 1434 | __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store); |
581d4e28 | 1435 | #endif |
edfaa7c3 KS |
1436 | |
1437 | static struct attribute *disk_attrs[] = { | |
1438 | &dev_attr_range.attr, | |
1f014290 | 1439 | &dev_attr_ext_range.attr, |
edfaa7c3 | 1440 | &dev_attr_removable.attr, |
8ddcd653 | 1441 | &dev_attr_hidden.attr, |
1c9ce527 | 1442 | &dev_attr_ro.attr, |
edfaa7c3 | 1443 | &dev_attr_size.attr, |
c72758f3 | 1444 | &dev_attr_alignment_offset.attr, |
86b37281 | 1445 | &dev_attr_discard_alignment.attr, |
edfaa7c3 KS |
1446 | &dev_attr_capability.attr, |
1447 | &dev_attr_stat.attr, | |
316d315b | 1448 | &dev_attr_inflight.attr, |
99e6608c | 1449 | &dev_attr_badblocks.attr, |
edfaa7c3 KS |
1450 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
1451 | &dev_attr_fail.attr, | |
581d4e28 JA |
1452 | #endif |
1453 | #ifdef CONFIG_FAIL_IO_TIMEOUT | |
1454 | &dev_attr_fail_timeout.attr, | |
edfaa7c3 KS |
1455 | #endif |
1456 | NULL | |
1457 | }; | |
1458 | ||
9438b3e0 DW |
1459 | static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n) |
1460 | { | |
1461 | struct device *dev = container_of(kobj, typeof(*dev), kobj); | |
1462 | struct gendisk *disk = dev_to_disk(dev); | |
1463 | ||
1464 | if (a == &dev_attr_badblocks.attr && !disk->bb) | |
1465 | return 0; | |
1466 | return a->mode; | |
1467 | } | |
1468 | ||
edfaa7c3 KS |
1469 | static struct attribute_group disk_attr_group = { |
1470 | .attrs = disk_attrs, | |
9438b3e0 | 1471 | .is_visible = disk_visible, |
edfaa7c3 KS |
1472 | }; |
1473 | ||
a4dbd674 | 1474 | static const struct attribute_group *disk_attr_groups[] = { |
edfaa7c3 KS |
1475 | &disk_attr_group, |
1476 | NULL | |
1da177e4 LT |
1477 | }; |
1478 | ||
540eed56 TH |
1479 | /** |
1480 | * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way | |
1481 | * @disk: disk to replace part_tbl for | |
1482 | * @new_ptbl: new part_tbl to install | |
1483 | * | |
1484 | * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The | |
1485 | * original ptbl is freed using RCU callback. | |
1486 | * | |
1487 | * LOCKING: | |
6d2cf6f2 | 1488 | * Matching bd_mutex locked or the caller is the only user of @disk. |
540eed56 TH |
1489 | */ |
1490 | static void disk_replace_part_tbl(struct gendisk *disk, | |
1491 | struct disk_part_tbl *new_ptbl) | |
1492 | { | |
6d2cf6f2 BVA |
1493 | struct disk_part_tbl *old_ptbl = |
1494 | rcu_dereference_protected(disk->part_tbl, 1); | |
540eed56 TH |
1495 | |
1496 | rcu_assign_pointer(disk->part_tbl, new_ptbl); | |
a6f23657 JA |
1497 | |
1498 | if (old_ptbl) { | |
1499 | rcu_assign_pointer(old_ptbl->last_lookup, NULL); | |
57bdfbf9 | 1500 | kfree_rcu(old_ptbl, rcu_head); |
a6f23657 | 1501 | } |
540eed56 TH |
1502 | } |
1503 | ||
1504 | /** | |
1505 | * disk_expand_part_tbl - expand disk->part_tbl | |
1506 | * @disk: disk to expand part_tbl for | |
1507 | * @partno: expand such that this partno can fit in | |
1508 | * | |
1509 | * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl | |
1510 | * uses RCU to allow unlocked dereferencing for stats and other stuff. | |
1511 | * | |
1512 | * LOCKING: | |
6d2cf6f2 BVA |
1513 | * Matching bd_mutex locked or the caller is the only user of @disk. |
1514 | * Might sleep. | |
540eed56 TH |
1515 | * |
1516 | * RETURNS: | |
1517 | * 0 on success, -errno on failure. | |
1518 | */ | |
1519 | int disk_expand_part_tbl(struct gendisk *disk, int partno) | |
1520 | { | |
6d2cf6f2 BVA |
1521 | struct disk_part_tbl *old_ptbl = |
1522 | rcu_dereference_protected(disk->part_tbl, 1); | |
540eed56 TH |
1523 | struct disk_part_tbl *new_ptbl; |
1524 | int len = old_ptbl ? old_ptbl->len : 0; | |
5fabcb4c | 1525 | int i, target; |
5fabcb4c JA |
1526 | |
1527 | /* | |
1528 | * check for int overflow, since we can get here from blkpg_ioctl() | |
1529 | * with a user passed 'partno'. | |
1530 | */ | |
1531 | target = partno + 1; | |
1532 | if (target < 0) | |
1533 | return -EINVAL; | |
540eed56 TH |
1534 | |
1535 | /* disk_max_parts() is zero during initialization, ignore if so */ | |
1536 | if (disk_max_parts(disk) && target > disk_max_parts(disk)) | |
1537 | return -EINVAL; | |
1538 | ||
1539 | if (target <= len) | |
1540 | return 0; | |
1541 | ||
78b90a2c GS |
1542 | new_ptbl = kzalloc_node(struct_size(new_ptbl, part, target), GFP_KERNEL, |
1543 | disk->node_id); | |
540eed56 TH |
1544 | if (!new_ptbl) |
1545 | return -ENOMEM; | |
1546 | ||
540eed56 TH |
1547 | new_ptbl->len = target; |
1548 | ||
1549 | for (i = 0; i < len; i++) | |
1550 | rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]); | |
1551 | ||
1552 | disk_replace_part_tbl(disk, new_ptbl); | |
1553 | return 0; | |
1554 | } | |
1555 | ||
b5bd357c LC |
1556 | /** |
1557 | * disk_release - releases all allocated resources of the gendisk | |
1558 | * @dev: the device representing this disk | |
1559 | * | |
1560 | * This function releases all allocated resources of the gendisk. | |
1561 | * | |
1562 | * The struct gendisk refcount is incremented with get_gendisk() or | |
1563 | * get_disk_and_module(), and its refcount is decremented with | |
1564 | * put_disk_and_module() or put_disk(). Once the refcount reaches 0 this | |
1565 | * function is called. | |
1566 | * | |
1567 | * Drivers which used __device_add_disk() have a gendisk with a request_queue | |
1568 | * assigned. Since the request_queue sits on top of the gendisk for these | |
1569 | * drivers we also call blk_put_queue() for them, and we expect the | |
1570 | * request_queue refcount to reach 0 at this point, and so the request_queue | |
1571 | * will also be freed prior to the disk. | |
e8c7d14a LC |
1572 | * |
1573 | * Context: can sleep | |
b5bd357c | 1574 | */ |
edfaa7c3 | 1575 | static void disk_release(struct device *dev) |
1da177e4 | 1576 | { |
edfaa7c3 KS |
1577 | struct gendisk *disk = dev_to_disk(dev); |
1578 | ||
e8c7d14a LC |
1579 | might_sleep(); |
1580 | ||
2da78092 | 1581 | blk_free_devt(dev->devt); |
77ea887e | 1582 | disk_release_events(disk); |
1da177e4 | 1583 | kfree(disk->random); |
540eed56 | 1584 | disk_replace_part_tbl(disk, NULL); |
b54e5ed8 | 1585 | hd_free_part(&disk->part0); |
523e1d39 TH |
1586 | if (disk->queue) |
1587 | blk_put_queue(disk->queue); | |
1da177e4 LT |
1588 | kfree(disk); |
1589 | } | |
edfaa7c3 KS |
1590 | struct class block_class = { |
1591 | .name = "block", | |
1da177e4 LT |
1592 | }; |
1593 | ||
3c2670e6 | 1594 | static char *block_devnode(struct device *dev, umode_t *mode, |
4e4098a3 | 1595 | kuid_t *uid, kgid_t *gid) |
b03f38b6 KS |
1596 | { |
1597 | struct gendisk *disk = dev_to_disk(dev); | |
1598 | ||
348e114b CH |
1599 | if (disk->fops->devnode) |
1600 | return disk->fops->devnode(disk, mode); | |
b03f38b6 KS |
1601 | return NULL; |
1602 | } | |
1603 | ||
ef45fe47 | 1604 | const struct device_type disk_type = { |
edfaa7c3 KS |
1605 | .name = "disk", |
1606 | .groups = disk_attr_groups, | |
1607 | .release = disk_release, | |
e454cea2 | 1608 | .devnode = block_devnode, |
1da177e4 LT |
1609 | }; |
1610 | ||
a6e2ba88 | 1611 | #ifdef CONFIG_PROC_FS |
cf771cb5 TH |
1612 | /* |
1613 | * aggregate disk stat collector. Uses the same stats that the sysfs | |
1614 | * entries do, above, but makes them available through one seq_file. | |
1615 | * | |
1616 | * The output looks suspiciously like /proc/partitions with a bunch of | |
1617 | * extra fields. | |
1618 | */ | |
1619 | static int diskstats_show(struct seq_file *seqf, void *v) | |
1da177e4 LT |
1620 | { |
1621 | struct gendisk *gp = v; | |
e71bf0d0 TH |
1622 | struct disk_part_iter piter; |
1623 | struct hd_struct *hd; | |
1da177e4 | 1624 | char buf[BDEVNAME_SIZE]; |
e016b782 | 1625 | unsigned int inflight; |
ea18e0f0 | 1626 | struct disk_stats stat; |
1da177e4 LT |
1627 | |
1628 | /* | |
ed9e1982 | 1629 | if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next) |
cf771cb5 | 1630 | seq_puts(seqf, "major minor name" |
1da177e4 LT |
1631 | " rio rmerge rsect ruse wio wmerge " |
1632 | "wsect wuse running use aveq" | |
1633 | "\n\n"); | |
1634 | */ | |
9f5e4865 | 1635 | |
71982a40 | 1636 | disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0); |
e71bf0d0 | 1637 | while ((hd = disk_part_iter_next(&piter))) { |
ea18e0f0 | 1638 | part_stat_read_all(hd, &stat); |
b2f609e1 CH |
1639 | if (queue_is_mq(gp->queue)) |
1640 | inflight = blk_mq_in_flight(gp->queue, hd); | |
1641 | else | |
1f06959b | 1642 | inflight = part_in_flight(hd); |
ea18e0f0 | 1643 | |
bdca3c87 MC |
1644 | seq_printf(seqf, "%4d %7d %s " |
1645 | "%lu %lu %lu %u " | |
1646 | "%lu %lu %lu %u " | |
1647 | "%u %u %u " | |
b6866318 KK |
1648 | "%lu %lu %lu %u " |
1649 | "%lu %u" | |
1650 | "\n", | |
f331c029 TH |
1651 | MAJOR(part_devt(hd)), MINOR(part_devt(hd)), |
1652 | disk_name(gp, hd->partno, buf), | |
ea18e0f0 KK |
1653 | stat.ios[STAT_READ], |
1654 | stat.merges[STAT_READ], | |
1655 | stat.sectors[STAT_READ], | |
1656 | (unsigned int)div_u64(stat.nsecs[STAT_READ], | |
1657 | NSEC_PER_MSEC), | |
1658 | stat.ios[STAT_WRITE], | |
1659 | stat.merges[STAT_WRITE], | |
1660 | stat.sectors[STAT_WRITE], | |
1661 | (unsigned int)div_u64(stat.nsecs[STAT_WRITE], | |
1662 | NSEC_PER_MSEC), | |
e016b782 | 1663 | inflight, |
ea18e0f0 | 1664 | jiffies_to_msecs(stat.io_ticks), |
8cd5b8fc KK |
1665 | (unsigned int)div_u64(stat.nsecs[STAT_READ] + |
1666 | stat.nsecs[STAT_WRITE] + | |
1667 | stat.nsecs[STAT_DISCARD] + | |
1668 | stat.nsecs[STAT_FLUSH], | |
1669 | NSEC_PER_MSEC), | |
ea18e0f0 KK |
1670 | stat.ios[STAT_DISCARD], |
1671 | stat.merges[STAT_DISCARD], | |
1672 | stat.sectors[STAT_DISCARD], | |
1673 | (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], | |
1674 | NSEC_PER_MSEC), | |
1675 | stat.ios[STAT_FLUSH], | |
1676 | (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], | |
1677 | NSEC_PER_MSEC) | |
28f39d55 | 1678 | ); |
1da177e4 | 1679 | } |
e71bf0d0 | 1680 | disk_part_iter_exit(&piter); |
9f5e4865 | 1681 | |
1da177e4 LT |
1682 | return 0; |
1683 | } | |
1684 | ||
31d85ab2 | 1685 | static const struct seq_operations diskstats_op = { |
def4e38d TH |
1686 | .start = disk_seqf_start, |
1687 | .next = disk_seqf_next, | |
1688 | .stop = disk_seqf_stop, | |
1da177e4 LT |
1689 | .show = diskstats_show |
1690 | }; | |
f500975a AD |
1691 | |
1692 | static int __init proc_genhd_init(void) | |
1693 | { | |
fddda2b7 CH |
1694 | proc_create_seq("diskstats", 0, NULL, &diskstats_op); |
1695 | proc_create_seq("partitions", 0, NULL, &partitions_op); | |
f500975a AD |
1696 | return 0; |
1697 | } | |
1698 | module_init(proc_genhd_init); | |
a6e2ba88 | 1699 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 1700 | |
cf771cb5 | 1701 | dev_t blk_lookup_devt(const char *name, int partno) |
a142be85 | 1702 | { |
def4e38d TH |
1703 | dev_t devt = MKDEV(0, 0); |
1704 | struct class_dev_iter iter; | |
1705 | struct device *dev; | |
a142be85 | 1706 | |
def4e38d TH |
1707 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); |
1708 | while ((dev = class_dev_iter_next(&iter))) { | |
a142be85 | 1709 | struct gendisk *disk = dev_to_disk(dev); |
548b10eb | 1710 | struct hd_struct *part; |
a142be85 | 1711 | |
3ada8b7e | 1712 | if (strcmp(dev_name(dev), name)) |
f331c029 | 1713 | continue; |
f331c029 | 1714 | |
41b8c853 NB |
1715 | if (partno < disk->minors) { |
1716 | /* We need to return the right devno, even | |
1717 | * if the partition doesn't exist yet. | |
1718 | */ | |
1719 | devt = MKDEV(MAJOR(dev->devt), | |
1720 | MINOR(dev->devt) + partno); | |
1721 | break; | |
1722 | } | |
548b10eb | 1723 | part = disk_get_part(disk, partno); |
2bbedcb4 | 1724 | if (part) { |
f331c029 | 1725 | devt = part_devt(part); |
e71bf0d0 | 1726 | disk_put_part(part); |
548b10eb | 1727 | break; |
def4e38d | 1728 | } |
548b10eb | 1729 | disk_put_part(part); |
5c0ef6d0 | 1730 | } |
def4e38d | 1731 | class_dev_iter_exit(&iter); |
edfaa7c3 KS |
1732 | return devt; |
1733 | } | |
edfaa7c3 | 1734 | |
e319e1fb | 1735 | struct gendisk *__alloc_disk_node(int minors, int node_id) |
1946089a CL |
1736 | { |
1737 | struct gendisk *disk; | |
6d2cf6f2 | 1738 | struct disk_part_tbl *ptbl; |
1946089a | 1739 | |
de65b012 CH |
1740 | if (minors > DISK_MAX_PARTS) { |
1741 | printk(KERN_ERR | |
7fb52621 | 1742 | "block: can't allocate more than %d partitions\n", |
de65b012 CH |
1743 | DISK_MAX_PARTS); |
1744 | minors = DISK_MAX_PARTS; | |
1745 | } | |
1946089a | 1746 | |
c1b511eb | 1747 | disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id); |
f93af2a4 CH |
1748 | if (!disk) |
1749 | return NULL; | |
6c23a968 | 1750 | |
f93af2a4 CH |
1751 | disk->part0.dkstats = alloc_percpu(struct disk_stats); |
1752 | if (!disk->part0.dkstats) | |
1753 | goto out_free_disk; | |
b5d0b9df | 1754 | |
f93af2a4 CH |
1755 | init_rwsem(&disk->lookup_sem); |
1756 | disk->node_id = node_id; | |
1757 | if (disk_expand_part_tbl(disk, 0)) { | |
1758 | free_percpu(disk->part0.dkstats); | |
1759 | goto out_free_disk; | |
1da177e4 | 1760 | } |
f93af2a4 CH |
1761 | |
1762 | ptbl = rcu_dereference_protected(disk->part_tbl, 1); | |
1763 | rcu_assign_pointer(ptbl->part[0], &disk->part0); | |
1764 | ||
1765 | /* | |
1766 | * set_capacity() and get_capacity() currently don't use | |
1767 | * seqcounter to read/update the part0->nr_sects. Still init | |
1768 | * the counter as we can read the sectors in IO submission | |
1769 | * patch using seqence counters. | |
1770 | * | |
1771 | * TODO: Ideally set_capacity() and get_capacity() should be | |
1772 | * converted to make use of bd_mutex and sequence counters. | |
1773 | */ | |
1774 | hd_sects_seq_init(&disk->part0); | |
1775 | if (hd_ref_init(&disk->part0)) | |
1776 | goto out_free_part0; | |
1777 | ||
1778 | disk->minors = minors; | |
1779 | rand_initialize_disk(disk); | |
1780 | disk_to_dev(disk)->class = &block_class; | |
1781 | disk_to_dev(disk)->type = &disk_type; | |
1782 | device_initialize(disk_to_dev(disk)); | |
1da177e4 | 1783 | return disk; |
f93af2a4 CH |
1784 | |
1785 | out_free_part0: | |
1786 | hd_free_part(&disk->part0); | |
1787 | out_free_disk: | |
1788 | kfree(disk); | |
1789 | return NULL; | |
1da177e4 | 1790 | } |
e319e1fb | 1791 | EXPORT_SYMBOL(__alloc_disk_node); |
1da177e4 | 1792 | |
b5bd357c LC |
1793 | /** |
1794 | * put_disk - decrements the gendisk refcount | |
0d20dcc2 | 1795 | * @disk: the struct gendisk to decrement the refcount for |
b5bd357c LC |
1796 | * |
1797 | * This decrements the refcount for the struct gendisk. When this reaches 0 | |
1798 | * we'll have disk_release() called. | |
e8c7d14a LC |
1799 | * |
1800 | * Context: Any context, but the last reference must not be dropped from | |
1801 | * atomic context. | |
b5bd357c | 1802 | */ |
1da177e4 LT |
1803 | void put_disk(struct gendisk *disk) |
1804 | { | |
1805 | if (disk) | |
efdc41c8 | 1806 | put_device(disk_to_dev(disk)); |
1da177e4 | 1807 | } |
1da177e4 LT |
1808 | EXPORT_SYMBOL(put_disk); |
1809 | ||
b5bd357c LC |
1810 | /** |
1811 | * put_disk_and_module - decrements the module and gendisk refcount | |
0d20dcc2 | 1812 | * @disk: the struct gendisk to decrement the refcount for |
b5bd357c | 1813 | * |
9df6c299 JK |
1814 | * This is a counterpart of get_disk_and_module() and thus also of |
1815 | * get_gendisk(). | |
e8c7d14a LC |
1816 | * |
1817 | * Context: Any context, but the last reference must not be dropped from | |
1818 | * atomic context. | |
9df6c299 JK |
1819 | */ |
1820 | void put_disk_and_module(struct gendisk *disk) | |
1821 | { | |
1822 | if (disk) { | |
1823 | struct module *owner = disk->fops->owner; | |
1824 | ||
1825 | put_disk(disk); | |
1826 | module_put(owner); | |
1827 | } | |
1828 | } | |
9df6c299 | 1829 | |
e3264a4d HR |
1830 | static void set_disk_ro_uevent(struct gendisk *gd, int ro) |
1831 | { | |
1832 | char event[] = "DISK_RO=1"; | |
1833 | char *envp[] = { event, NULL }; | |
1834 | ||
1835 | if (!ro) | |
1836 | event[8] = '0'; | |
1837 | kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); | |
1838 | } | |
1839 | ||
1da177e4 LT |
1840 | void set_disk_ro(struct gendisk *disk, int flag) |
1841 | { | |
e71bf0d0 TH |
1842 | struct disk_part_iter piter; |
1843 | struct hd_struct *part; | |
1844 | ||
e3264a4d HR |
1845 | if (disk->part0.policy != flag) { |
1846 | set_disk_ro_uevent(disk, flag); | |
1847 | disk->part0.policy = flag; | |
1848 | } | |
1849 | ||
1850 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY); | |
e71bf0d0 TH |
1851 | while ((part = disk_part_iter_next(&piter))) |
1852 | part->policy = flag; | |
1853 | disk_part_iter_exit(&piter); | |
1da177e4 LT |
1854 | } |
1855 | ||
1856 | EXPORT_SYMBOL(set_disk_ro); | |
1857 | ||
1858 | int bdev_read_only(struct block_device *bdev) | |
1859 | { | |
1860 | if (!bdev) | |
1861 | return 0; | |
b7db9956 | 1862 | return bdev->bd_part->policy; |
1da177e4 LT |
1863 | } |
1864 | ||
1865 | EXPORT_SYMBOL(bdev_read_only); | |
1866 | ||
77ea887e TH |
1867 | /* |
1868 | * Disk events - monitor disk events like media change and eject request. | |
1869 | */ | |
1870 | struct disk_events { | |
1871 | struct list_head node; /* all disk_event's */ | |
1872 | struct gendisk *disk; /* the associated disk */ | |
1873 | spinlock_t lock; | |
1874 | ||
fdd514e1 | 1875 | struct mutex block_mutex; /* protects blocking */ |
77ea887e TH |
1876 | int block; /* event blocking depth */ |
1877 | unsigned int pending; /* events already sent out */ | |
1878 | unsigned int clearing; /* events being cleared */ | |
1879 | ||
1880 | long poll_msecs; /* interval, -1 for default */ | |
1881 | struct delayed_work dwork; | |
1882 | }; | |
1883 | ||
1884 | static const char *disk_events_strs[] = { | |
1885 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change", | |
1886 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request", | |
1887 | }; | |
1888 | ||
1889 | static char *disk_uevents[] = { | |
1890 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1", | |
1891 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1", | |
1892 | }; | |
1893 | ||
1894 | /* list of all disk_events */ | |
1895 | static DEFINE_MUTEX(disk_events_mutex); | |
1896 | static LIST_HEAD(disk_events); | |
1897 | ||
1898 | /* disable in-kernel polling by default */ | |
1fe8f348 | 1899 | static unsigned long disk_events_dfl_poll_msecs; |
77ea887e TH |
1900 | |
1901 | static unsigned long disk_events_poll_jiffies(struct gendisk *disk) | |
1902 | { | |
1903 | struct disk_events *ev = disk->ev; | |
1904 | long intv_msecs = 0; | |
1905 | ||
1906 | /* | |
1907 | * If device-specific poll interval is set, always use it. If | |
673387a9 | 1908 | * the default is being used, poll if the POLL flag is set. |
77ea887e TH |
1909 | */ |
1910 | if (ev->poll_msecs >= 0) | |
1911 | intv_msecs = ev->poll_msecs; | |
c92e2f04 | 1912 | else if (disk->event_flags & DISK_EVENT_FLAG_POLL) |
77ea887e TH |
1913 | intv_msecs = disk_events_dfl_poll_msecs; |
1914 | ||
1915 | return msecs_to_jiffies(intv_msecs); | |
1916 | } | |
1917 | ||
c3af54af TH |
1918 | /** |
1919 | * disk_block_events - block and flush disk event checking | |
1920 | * @disk: disk to block events for | |
1921 | * | |
1922 | * On return from this function, it is guaranteed that event checking | |
1923 | * isn't in progress and won't happen until unblocked by | |
1924 | * disk_unblock_events(). Events blocking is counted and the actual | |
1925 | * unblocking happens after the matching number of unblocks are done. | |
1926 | * | |
1927 | * Note that this intentionally does not block event checking from | |
1928 | * disk_clear_events(). | |
1929 | * | |
1930 | * CONTEXT: | |
1931 | * Might sleep. | |
1932 | */ | |
1933 | void disk_block_events(struct gendisk *disk) | |
77ea887e TH |
1934 | { |
1935 | struct disk_events *ev = disk->ev; | |
1936 | unsigned long flags; | |
1937 | bool cancel; | |
1938 | ||
c3af54af TH |
1939 | if (!ev) |
1940 | return; | |
1941 | ||
fdd514e1 TH |
1942 | /* |
1943 | * Outer mutex ensures that the first blocker completes canceling | |
1944 | * the event work before further blockers are allowed to finish. | |
1945 | */ | |
1946 | mutex_lock(&ev->block_mutex); | |
1947 | ||
77ea887e TH |
1948 | spin_lock_irqsave(&ev->lock, flags); |
1949 | cancel = !ev->block++; | |
1950 | spin_unlock_irqrestore(&ev->lock, flags); | |
1951 | ||
c3af54af TH |
1952 | if (cancel) |
1953 | cancel_delayed_work_sync(&disk->ev->dwork); | |
fdd514e1 TH |
1954 | |
1955 | mutex_unlock(&ev->block_mutex); | |
77ea887e TH |
1956 | } |
1957 | ||
1958 | static void __disk_unblock_events(struct gendisk *disk, bool check_now) | |
1959 | { | |
1960 | struct disk_events *ev = disk->ev; | |
1961 | unsigned long intv; | |
1962 | unsigned long flags; | |
1963 | ||
1964 | spin_lock_irqsave(&ev->lock, flags); | |
1965 | ||
1966 | if (WARN_ON_ONCE(ev->block <= 0)) | |
1967 | goto out_unlock; | |
1968 | ||
1969 | if (--ev->block) | |
1970 | goto out_unlock; | |
1971 | ||
77ea887e | 1972 | intv = disk_events_poll_jiffies(disk); |
77ea887e | 1973 | if (check_now) |
695588f9 VK |
1974 | queue_delayed_work(system_freezable_power_efficient_wq, |
1975 | &ev->dwork, 0); | |
77ea887e | 1976 | else if (intv) |
695588f9 VK |
1977 | queue_delayed_work(system_freezable_power_efficient_wq, |
1978 | &ev->dwork, intv); | |
77ea887e TH |
1979 | out_unlock: |
1980 | spin_unlock_irqrestore(&ev->lock, flags); | |
1981 | } | |
1982 | ||
77ea887e TH |
1983 | /** |
1984 | * disk_unblock_events - unblock disk event checking | |
1985 | * @disk: disk to unblock events for | |
1986 | * | |
1987 | * Undo disk_block_events(). When the block count reaches zero, it | |
1988 | * starts events polling if configured. | |
1989 | * | |
1990 | * CONTEXT: | |
1991 | * Don't care. Safe to call from irq context. | |
1992 | */ | |
1993 | void disk_unblock_events(struct gendisk *disk) | |
1994 | { | |
1995 | if (disk->ev) | |
facc31dd | 1996 | __disk_unblock_events(disk, false); |
77ea887e TH |
1997 | } |
1998 | ||
1999 | /** | |
85ef06d1 TH |
2000 | * disk_flush_events - schedule immediate event checking and flushing |
2001 | * @disk: disk to check and flush events for | |
2002 | * @mask: events to flush | |
77ea887e | 2003 | * |
85ef06d1 TH |
2004 | * Schedule immediate event checking on @disk if not blocked. Events in |
2005 | * @mask are scheduled to be cleared from the driver. Note that this | |
2006 | * doesn't clear the events from @disk->ev. | |
77ea887e TH |
2007 | * |
2008 | * CONTEXT: | |
85ef06d1 | 2009 | * If @mask is non-zero must be called with bdev->bd_mutex held. |
77ea887e | 2010 | */ |
85ef06d1 | 2011 | void disk_flush_events(struct gendisk *disk, unsigned int mask) |
77ea887e | 2012 | { |
a9dce2a3 | 2013 | struct disk_events *ev = disk->ev; |
a9dce2a3 TH |
2014 | |
2015 | if (!ev) | |
2016 | return; | |
2017 | ||
85ef06d1 TH |
2018 | spin_lock_irq(&ev->lock); |
2019 | ev->clearing |= mask; | |
41f63c53 | 2020 | if (!ev->block) |
695588f9 VK |
2021 | mod_delayed_work(system_freezable_power_efficient_wq, |
2022 | &ev->dwork, 0); | |
85ef06d1 | 2023 | spin_unlock_irq(&ev->lock); |
77ea887e | 2024 | } |
77ea887e TH |
2025 | |
2026 | /** | |
2027 | * disk_clear_events - synchronously check, clear and return pending events | |
2028 | * @disk: disk to fetch and clear events from | |
da3dae54 | 2029 | * @mask: mask of events to be fetched and cleared |
77ea887e TH |
2030 | * |
2031 | * Disk events are synchronously checked and pending events in @mask | |
2032 | * are cleared and returned. This ignores the block count. | |
2033 | * | |
2034 | * CONTEXT: | |
2035 | * Might sleep. | |
2036 | */ | |
95f6f3a4 | 2037 | static unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask) |
77ea887e | 2038 | { |
77ea887e TH |
2039 | struct disk_events *ev = disk->ev; |
2040 | unsigned int pending; | |
12c2bdb2 | 2041 | unsigned int clearing = mask; |
77ea887e | 2042 | |
a564e23f | 2043 | if (!ev) |
77ea887e | 2044 | return 0; |
77ea887e | 2045 | |
12c2bdb2 DB |
2046 | disk_block_events(disk); |
2047 | ||
2048 | /* | |
2049 | * store the union of mask and ev->clearing on the stack so that the | |
2050 | * race with disk_flush_events does not cause ambiguity (ev->clearing | |
2051 | * can still be modified even if events are blocked). | |
2052 | */ | |
77ea887e | 2053 | spin_lock_irq(&ev->lock); |
12c2bdb2 DB |
2054 | clearing |= ev->clearing; |
2055 | ev->clearing = 0; | |
77ea887e TH |
2056 | spin_unlock_irq(&ev->lock); |
2057 | ||
12c2bdb2 | 2058 | disk_check_events(ev, &clearing); |
aea24a8b | 2059 | /* |
12c2bdb2 DB |
2060 | * if ev->clearing is not 0, the disk_flush_events got called in the |
2061 | * middle of this function, so we want to run the workfn without delay. | |
aea24a8b | 2062 | */ |
12c2bdb2 | 2063 | __disk_unblock_events(disk, ev->clearing ? true : false); |
77ea887e TH |
2064 | |
2065 | /* then, fetch and clear pending events */ | |
2066 | spin_lock_irq(&ev->lock); | |
77ea887e TH |
2067 | pending = ev->pending & mask; |
2068 | ev->pending &= ~mask; | |
2069 | spin_unlock_irq(&ev->lock); | |
12c2bdb2 | 2070 | WARN_ON_ONCE(clearing & mask); |
77ea887e TH |
2071 | |
2072 | return pending; | |
2073 | } | |
2074 | ||
95f6f3a4 CH |
2075 | /** |
2076 | * bdev_check_media_change - check if a removable media has been changed | |
2077 | * @bdev: block device to check | |
2078 | * | |
2079 | * Check whether a removable media has been changed, and attempt to free all | |
2080 | * dentries and inodes and invalidates all block device page cache entries in | |
2081 | * that case. | |
2082 | * | |
2083 | * Returns %true if the block device changed, or %false if not. | |
2084 | */ | |
2085 | bool bdev_check_media_change(struct block_device *bdev) | |
2086 | { | |
2087 | unsigned int events; | |
2088 | ||
2089 | events = disk_clear_events(bdev->bd_disk, DISK_EVENT_MEDIA_CHANGE | | |
2090 | DISK_EVENT_EJECT_REQUEST); | |
2091 | if (!(events & DISK_EVENT_MEDIA_CHANGE)) | |
2092 | return false; | |
2093 | ||
2094 | if (__invalidate_device(bdev, true)) | |
2095 | pr_warn("VFS: busy inodes on changed media %s\n", | |
2096 | bdev->bd_disk->disk_name); | |
38430f08 | 2097 | set_bit(GD_NEED_PART_SCAN, &bdev->bd_disk->state); |
95f6f3a4 CH |
2098 | return true; |
2099 | } | |
2100 | EXPORT_SYMBOL(bdev_check_media_change); | |
2101 | ||
12c2bdb2 DB |
2102 | /* |
2103 | * Separate this part out so that a different pointer for clearing_ptr can be | |
2104 | * passed in for disk_clear_events. | |
2105 | */ | |
77ea887e TH |
2106 | static void disk_events_workfn(struct work_struct *work) |
2107 | { | |
2108 | struct delayed_work *dwork = to_delayed_work(work); | |
2109 | struct disk_events *ev = container_of(dwork, struct disk_events, dwork); | |
12c2bdb2 DB |
2110 | |
2111 | disk_check_events(ev, &ev->clearing); | |
2112 | } | |
2113 | ||
2114 | static void disk_check_events(struct disk_events *ev, | |
2115 | unsigned int *clearing_ptr) | |
2116 | { | |
77ea887e TH |
2117 | struct gendisk *disk = ev->disk; |
2118 | char *envp[ARRAY_SIZE(disk_uevents) + 1] = { }; | |
12c2bdb2 | 2119 | unsigned int clearing = *clearing_ptr; |
77ea887e TH |
2120 | unsigned int events; |
2121 | unsigned long intv; | |
2122 | int nr_events = 0, i; | |
2123 | ||
2124 | /* check events */ | |
2125 | events = disk->fops->check_events(disk, clearing); | |
2126 | ||
2127 | /* accumulate pending events and schedule next poll if necessary */ | |
2128 | spin_lock_irq(&ev->lock); | |
2129 | ||
2130 | events &= ~ev->pending; | |
2131 | ev->pending |= events; | |
12c2bdb2 | 2132 | *clearing_ptr &= ~clearing; |
77ea887e TH |
2133 | |
2134 | intv = disk_events_poll_jiffies(disk); | |
2135 | if (!ev->block && intv) | |
695588f9 VK |
2136 | queue_delayed_work(system_freezable_power_efficient_wq, |
2137 | &ev->dwork, intv); | |
77ea887e TH |
2138 | |
2139 | spin_unlock_irq(&ev->lock); | |
2140 | ||
7c88a168 TH |
2141 | /* |
2142 | * Tell userland about new events. Only the events listed in | |
c92e2f04 MW |
2143 | * @disk->events are reported, and only if DISK_EVENT_FLAG_UEVENT |
2144 | * is set. Otherwise, events are processed internally but never | |
2145 | * get reported to userland. | |
7c88a168 | 2146 | */ |
77ea887e | 2147 | for (i = 0; i < ARRAY_SIZE(disk_uevents); i++) |
c92e2f04 MW |
2148 | if ((events & disk->events & (1 << i)) && |
2149 | (disk->event_flags & DISK_EVENT_FLAG_UEVENT)) | |
77ea887e TH |
2150 | envp[nr_events++] = disk_uevents[i]; |
2151 | ||
2152 | if (nr_events) | |
2153 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
2154 | } | |
2155 | ||
2156 | /* | |
2157 | * A disk events enabled device has the following sysfs nodes under | |
2158 | * its /sys/block/X/ directory. | |
2159 | * | |
2160 | * events : list of all supported events | |
2161 | * events_async : list of events which can be detected w/o polling | |
673387a9 | 2162 | * (always empty, only for backwards compatibility) |
77ea887e TH |
2163 | * events_poll_msecs : polling interval, 0: disable, -1: system default |
2164 | */ | |
2165 | static ssize_t __disk_events_show(unsigned int events, char *buf) | |
2166 | { | |
2167 | const char *delim = ""; | |
2168 | ssize_t pos = 0; | |
2169 | int i; | |
2170 | ||
2171 | for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++) | |
2172 | if (events & (1 << i)) { | |
2173 | pos += sprintf(buf + pos, "%s%s", | |
2174 | delim, disk_events_strs[i]); | |
2175 | delim = " "; | |
2176 | } | |
2177 | if (pos) | |
2178 | pos += sprintf(buf + pos, "\n"); | |
2179 | return pos; | |
2180 | } | |
2181 | ||
2182 | static ssize_t disk_events_show(struct device *dev, | |
2183 | struct device_attribute *attr, char *buf) | |
2184 | { | |
2185 | struct gendisk *disk = dev_to_disk(dev); | |
2186 | ||
c92e2f04 MW |
2187 | if (!(disk->event_flags & DISK_EVENT_FLAG_UEVENT)) |
2188 | return 0; | |
2189 | ||
77ea887e TH |
2190 | return __disk_events_show(disk->events, buf); |
2191 | } | |
2192 | ||
2193 | static ssize_t disk_events_async_show(struct device *dev, | |
2194 | struct device_attribute *attr, char *buf) | |
2195 | { | |
673387a9 | 2196 | return 0; |
77ea887e TH |
2197 | } |
2198 | ||
2199 | static ssize_t disk_events_poll_msecs_show(struct device *dev, | |
2200 | struct device_attribute *attr, | |
2201 | char *buf) | |
2202 | { | |
2203 | struct gendisk *disk = dev_to_disk(dev); | |
2204 | ||
cdf3e3de MW |
2205 | if (!disk->ev) |
2206 | return sprintf(buf, "-1\n"); | |
2207 | ||
77ea887e TH |
2208 | return sprintf(buf, "%ld\n", disk->ev->poll_msecs); |
2209 | } | |
2210 | ||
2211 | static ssize_t disk_events_poll_msecs_store(struct device *dev, | |
2212 | struct device_attribute *attr, | |
2213 | const char *buf, size_t count) | |
2214 | { | |
2215 | struct gendisk *disk = dev_to_disk(dev); | |
2216 | long intv; | |
2217 | ||
2218 | if (!count || !sscanf(buf, "%ld", &intv)) | |
2219 | return -EINVAL; | |
2220 | ||
2221 | if (intv < 0 && intv != -1) | |
2222 | return -EINVAL; | |
2223 | ||
cdf3e3de MW |
2224 | if (!disk->ev) |
2225 | return -ENODEV; | |
2226 | ||
c3af54af | 2227 | disk_block_events(disk); |
77ea887e TH |
2228 | disk->ev->poll_msecs = intv; |
2229 | __disk_unblock_events(disk, true); | |
2230 | ||
2231 | return count; | |
2232 | } | |
2233 | ||
5657a819 JP |
2234 | static const DEVICE_ATTR(events, 0444, disk_events_show, NULL); |
2235 | static const DEVICE_ATTR(events_async, 0444, disk_events_async_show, NULL); | |
2236 | static const DEVICE_ATTR(events_poll_msecs, 0644, | |
77ea887e TH |
2237 | disk_events_poll_msecs_show, |
2238 | disk_events_poll_msecs_store); | |
2239 | ||
2240 | static const struct attribute *disk_events_attrs[] = { | |
2241 | &dev_attr_events.attr, | |
2242 | &dev_attr_events_async.attr, | |
2243 | &dev_attr_events_poll_msecs.attr, | |
2244 | NULL, | |
2245 | }; | |
2246 | ||
2247 | /* | |
2248 | * The default polling interval can be specified by the kernel | |
2249 | * parameter block.events_dfl_poll_msecs which defaults to 0 | |
2250 | * (disable). This can also be modified runtime by writing to | |
1624b0b2 | 2251 | * /sys/module/block/parameters/events_dfl_poll_msecs. |
77ea887e TH |
2252 | */ |
2253 | static int disk_events_set_dfl_poll_msecs(const char *val, | |
2254 | const struct kernel_param *kp) | |
2255 | { | |
2256 | struct disk_events *ev; | |
2257 | int ret; | |
2258 | ||
2259 | ret = param_set_ulong(val, kp); | |
2260 | if (ret < 0) | |
2261 | return ret; | |
2262 | ||
2263 | mutex_lock(&disk_events_mutex); | |
2264 | ||
2265 | list_for_each_entry(ev, &disk_events, node) | |
85ef06d1 | 2266 | disk_flush_events(ev->disk, 0); |
77ea887e TH |
2267 | |
2268 | mutex_unlock(&disk_events_mutex); | |
2269 | ||
2270 | return 0; | |
2271 | } | |
2272 | ||
2273 | static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = { | |
2274 | .set = disk_events_set_dfl_poll_msecs, | |
2275 | .get = param_get_ulong, | |
2276 | }; | |
2277 | ||
2278 | #undef MODULE_PARAM_PREFIX | |
2279 | #define MODULE_PARAM_PREFIX "block." | |
2280 | ||
2281 | module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops, | |
2282 | &disk_events_dfl_poll_msecs, 0644); | |
2283 | ||
2284 | /* | |
9f53d2fe | 2285 | * disk_{alloc|add|del|release}_events - initialize and destroy disk_events. |
77ea887e | 2286 | */ |
9f53d2fe | 2287 | static void disk_alloc_events(struct gendisk *disk) |
77ea887e TH |
2288 | { |
2289 | struct disk_events *ev; | |
2290 | ||
cdf3e3de | 2291 | if (!disk->fops->check_events || !disk->events) |
77ea887e TH |
2292 | return; |
2293 | ||
2294 | ev = kzalloc(sizeof(*ev), GFP_KERNEL); | |
2295 | if (!ev) { | |
2296 | pr_warn("%s: failed to initialize events\n", disk->disk_name); | |
2297 | return; | |
2298 | } | |
2299 | ||
77ea887e TH |
2300 | INIT_LIST_HEAD(&ev->node); |
2301 | ev->disk = disk; | |
2302 | spin_lock_init(&ev->lock); | |
fdd514e1 | 2303 | mutex_init(&ev->block_mutex); |
77ea887e TH |
2304 | ev->block = 1; |
2305 | ev->poll_msecs = -1; | |
2306 | INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn); | |
2307 | ||
9f53d2fe SG |
2308 | disk->ev = ev; |
2309 | } | |
2310 | ||
2311 | static void disk_add_events(struct gendisk *disk) | |
2312 | { | |
9f53d2fe SG |
2313 | /* FIXME: error handling */ |
2314 | if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0) | |
2315 | pr_warn("%s: failed to create sysfs files for events\n", | |
2316 | disk->disk_name); | |
2317 | ||
cdf3e3de MW |
2318 | if (!disk->ev) |
2319 | return; | |
2320 | ||
77ea887e | 2321 | mutex_lock(&disk_events_mutex); |
9f53d2fe | 2322 | list_add_tail(&disk->ev->node, &disk_events); |
77ea887e TH |
2323 | mutex_unlock(&disk_events_mutex); |
2324 | ||
2325 | /* | |
2326 | * Block count is initialized to 1 and the following initial | |
2327 | * unblock kicks it into action. | |
2328 | */ | |
2329 | __disk_unblock_events(disk, true); | |
2330 | } | |
2331 | ||
2332 | static void disk_del_events(struct gendisk *disk) | |
2333 | { | |
cdf3e3de MW |
2334 | if (disk->ev) { |
2335 | disk_block_events(disk); | |
77ea887e | 2336 | |
cdf3e3de MW |
2337 | mutex_lock(&disk_events_mutex); |
2338 | list_del_init(&disk->ev->node); | |
2339 | mutex_unlock(&disk_events_mutex); | |
2340 | } | |
77ea887e TH |
2341 | |
2342 | sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs); | |
2343 | } | |
2344 | ||
2345 | static void disk_release_events(struct gendisk *disk) | |
2346 | { | |
2347 | /* the block count should be 1 from disk_del_events() */ | |
2348 | WARN_ON_ONCE(disk->ev && disk->ev->block != 1); | |
2349 | kfree(disk->ev); | |
2350 | } |