]>
Commit | Line | Data |
---|---|---|
3dcf60bc | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * gendisk handling | |
4 | */ | |
5 | ||
1da177e4 | 6 | #include <linux/module.h> |
3ad5cee5 | 7 | #include <linux/ctype.h> |
1da177e4 LT |
8 | #include <linux/fs.h> |
9 | #include <linux/genhd.h> | |
b446b60e | 10 | #include <linux/kdev_t.h> |
1da177e4 LT |
11 | #include <linux/kernel.h> |
12 | #include <linux/blkdev.h> | |
66114cad | 13 | #include <linux/backing-dev.h> |
1da177e4 LT |
14 | #include <linux/init.h> |
15 | #include <linux/spinlock.h> | |
f500975a | 16 | #include <linux/proc_fs.h> |
1da177e4 LT |
17 | #include <linux/seq_file.h> |
18 | #include <linux/slab.h> | |
19 | #include <linux/kmod.h> | |
20 | #include <linux/kobj_map.h> | |
58383af6 | 21 | #include <linux/mutex.h> |
bcce3de1 | 22 | #include <linux/idr.h> |
77ea887e | 23 | #include <linux/log2.h> |
25e823c8 | 24 | #include <linux/pm_runtime.h> |
99e6608c | 25 | #include <linux/badblocks.h> |
1da177e4 | 26 | |
ff88972c AB |
27 | #include "blk.h" |
28 | ||
edfaa7c3 | 29 | static DEFINE_MUTEX(block_class_lock); |
31eb6186 | 30 | static struct kobject *block_depr; |
1da177e4 | 31 | |
bcce3de1 | 32 | /* for extended dynamic devt allocation, currently only one major is used */ |
ce23bba8 | 33 | #define NR_EXT_DEVT (1 << MINORBITS) |
bcce3de1 | 34 | |
2da78092 | 35 | /* For extended devt allocation. ext_devt_lock prevents look up |
bcce3de1 TH |
36 | * results from going away underneath its user. |
37 | */ | |
2da78092 | 38 | static DEFINE_SPINLOCK(ext_devt_lock); |
bcce3de1 TH |
39 | static DEFINE_IDR(ext_devt_idr); |
40 | ||
edf8ff55 | 41 | static const struct device_type disk_type; |
1826eadf | 42 | |
12c2bdb2 DB |
43 | static void disk_check_events(struct disk_events *ev, |
44 | unsigned int *clearing_ptr); | |
9f53d2fe | 45 | static void disk_alloc_events(struct gendisk *disk); |
77ea887e TH |
46 | static void disk_add_events(struct gendisk *disk); |
47 | static void disk_del_events(struct gendisk *disk); | |
48 | static void disk_release_events(struct gendisk *disk); | |
49 | ||
e598a72f BS |
50 | /* |
51 | * Set disk capacity and notify if the size is not currently | |
52 | * zero and will not be set to zero | |
53 | */ | |
54 | void set_capacity_revalidate_and_notify(struct gendisk *disk, sector_t size, | |
55 | bool revalidate) | |
56 | { | |
57 | sector_t capacity = get_capacity(disk); | |
58 | ||
59 | set_capacity(disk, size); | |
60 | ||
61 | if (revalidate) | |
62 | revalidate_disk(disk); | |
63 | ||
64 | if (capacity != size && capacity != 0 && size != 0) { | |
65 | char *envp[] = { "RESIZE=1", NULL }; | |
66 | ||
67 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
68 | } | |
69 | } | |
70 | ||
71 | EXPORT_SYMBOL_GPL(set_capacity_revalidate_and_notify); | |
72 | ||
5cbd28e3 CH |
73 | /* |
74 | * Format the device name of the indicated disk into the supplied buffer and | |
75 | * return a pointer to that same buffer for convenience. | |
76 | */ | |
77 | char *disk_name(struct gendisk *hd, int partno, char *buf) | |
78 | { | |
79 | if (!partno) | |
80 | snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name); | |
81 | else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1])) | |
82 | snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno); | |
83 | else | |
84 | snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno); | |
85 | ||
86 | return buf; | |
87 | } | |
88 | ||
89 | const char *bdevname(struct block_device *bdev, char *buf) | |
90 | { | |
91 | return disk_name(bdev->bd_disk, bdev->bd_part->partno, buf); | |
92 | } | |
93 | EXPORT_SYMBOL(bdevname); | |
e598a72f | 94 | |
ea18e0f0 KK |
95 | #ifdef CONFIG_SMP |
96 | static void part_stat_read_all(struct hd_struct *part, struct disk_stats *stat) | |
97 | { | |
98 | int cpu; | |
99 | ||
100 | memset(stat, 0, sizeof(struct disk_stats)); | |
101 | for_each_possible_cpu(cpu) { | |
102 | struct disk_stats *ptr = per_cpu_ptr(part->dkstats, cpu); | |
103 | int group; | |
104 | ||
105 | for (group = 0; group < NR_STAT_GROUPS; group++) { | |
106 | stat->nsecs[group] += ptr->nsecs[group]; | |
107 | stat->sectors[group] += ptr->sectors[group]; | |
108 | stat->ios[group] += ptr->ios[group]; | |
109 | stat->merges[group] += ptr->merges[group]; | |
110 | } | |
111 | ||
112 | stat->io_ticks += ptr->io_ticks; | |
ea18e0f0 KK |
113 | } |
114 | } | |
115 | #else /* CONFIG_SMP */ | |
116 | static void part_stat_read_all(struct hd_struct *part, struct disk_stats *stat) | |
117 | { | |
118 | memcpy(stat, &part->dkstats, sizeof(struct disk_stats)); | |
119 | } | |
120 | #endif /* CONFIG_SMP */ | |
121 | ||
f299b7c7 JA |
122 | void part_inc_in_flight(struct request_queue *q, struct hd_struct *part, int rw) |
123 | { | |
344e9ffc | 124 | if (queue_is_mq(q)) |
f299b7c7 JA |
125 | return; |
126 | ||
1226b8dd | 127 | part_stat_local_inc(part, in_flight[rw]); |
f299b7c7 | 128 | if (part->partno) |
1226b8dd | 129 | part_stat_local_inc(&part_to_disk(part)->part0, in_flight[rw]); |
f299b7c7 JA |
130 | } |
131 | ||
132 | void part_dec_in_flight(struct request_queue *q, struct hd_struct *part, int rw) | |
133 | { | |
344e9ffc | 134 | if (queue_is_mq(q)) |
f299b7c7 JA |
135 | return; |
136 | ||
1226b8dd | 137 | part_stat_local_dec(part, in_flight[rw]); |
f299b7c7 | 138 | if (part->partno) |
1226b8dd | 139 | part_stat_local_dec(&part_to_disk(part)->part0, in_flight[rw]); |
f299b7c7 JA |
140 | } |
141 | ||
6005771c CH |
142 | static unsigned int part_in_flight(struct request_queue *q, |
143 | struct hd_struct *part) | |
f299b7c7 | 144 | { |
1226b8dd | 145 | int cpu; |
e016b782 | 146 | unsigned int inflight; |
1226b8dd | 147 | |
344e9ffc | 148 | if (queue_is_mq(q)) { |
e016b782 | 149 | return blk_mq_in_flight(q, part); |
f299b7c7 JA |
150 | } |
151 | ||
e016b782 | 152 | inflight = 0; |
1226b8dd | 153 | for_each_possible_cpu(cpu) { |
e016b782 MP |
154 | inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) + |
155 | part_stat_local_read_cpu(part, in_flight[1], cpu); | |
1226b8dd | 156 | } |
e016b782 MP |
157 | if ((int)inflight < 0) |
158 | inflight = 0; | |
1226b8dd | 159 | |
e016b782 | 160 | return inflight; |
f299b7c7 JA |
161 | } |
162 | ||
6005771c CH |
163 | static void part_in_flight_rw(struct request_queue *q, struct hd_struct *part, |
164 | unsigned int inflight[2]) | |
bf0ddaba | 165 | { |
1226b8dd MP |
166 | int cpu; |
167 | ||
344e9ffc | 168 | if (queue_is_mq(q)) { |
bf0ddaba OS |
169 | blk_mq_in_flight_rw(q, part, inflight); |
170 | return; | |
171 | } | |
172 | ||
1226b8dd MP |
173 | inflight[0] = 0; |
174 | inflight[1] = 0; | |
175 | for_each_possible_cpu(cpu) { | |
176 | inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu); | |
177 | inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu); | |
178 | } | |
179 | if ((int)inflight[0] < 0) | |
180 | inflight[0] = 0; | |
181 | if ((int)inflight[1] < 0) | |
182 | inflight[1] = 0; | |
bf0ddaba OS |
183 | } |
184 | ||
807d4af2 CH |
185 | struct hd_struct *__disk_get_part(struct gendisk *disk, int partno) |
186 | { | |
187 | struct disk_part_tbl *ptbl = rcu_dereference(disk->part_tbl); | |
188 | ||
189 | if (unlikely(partno < 0 || partno >= ptbl->len)) | |
190 | return NULL; | |
191 | return rcu_dereference(ptbl->part[partno]); | |
192 | } | |
193 | ||
e71bf0d0 TH |
194 | /** |
195 | * disk_get_part - get partition | |
196 | * @disk: disk to look partition from | |
197 | * @partno: partition number | |
198 | * | |
199 | * Look for partition @partno from @disk. If found, increment | |
200 | * reference count and return it. | |
201 | * | |
202 | * CONTEXT: | |
203 | * Don't care. | |
204 | * | |
205 | * RETURNS: | |
206 | * Pointer to the found partition on success, NULL if not found. | |
207 | */ | |
208 | struct hd_struct *disk_get_part(struct gendisk *disk, int partno) | |
209 | { | |
807d4af2 | 210 | struct hd_struct *part; |
540eed56 | 211 | |
e71bf0d0 | 212 | rcu_read_lock(); |
807d4af2 CH |
213 | part = __disk_get_part(disk, partno); |
214 | if (part) | |
215 | get_device(part_to_dev(part)); | |
e71bf0d0 TH |
216 | rcu_read_unlock(); |
217 | ||
218 | return part; | |
219 | } | |
e71bf0d0 TH |
220 | |
221 | /** | |
222 | * disk_part_iter_init - initialize partition iterator | |
223 | * @piter: iterator to initialize | |
224 | * @disk: disk to iterate over | |
225 | * @flags: DISK_PITER_* flags | |
226 | * | |
227 | * Initialize @piter so that it iterates over partitions of @disk. | |
228 | * | |
229 | * CONTEXT: | |
230 | * Don't care. | |
231 | */ | |
232 | void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk, | |
233 | unsigned int flags) | |
234 | { | |
540eed56 TH |
235 | struct disk_part_tbl *ptbl; |
236 | ||
237 | rcu_read_lock(); | |
238 | ptbl = rcu_dereference(disk->part_tbl); | |
239 | ||
e71bf0d0 TH |
240 | piter->disk = disk; |
241 | piter->part = NULL; | |
242 | ||
243 | if (flags & DISK_PITER_REVERSE) | |
540eed56 | 244 | piter->idx = ptbl->len - 1; |
71982a40 | 245 | else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0)) |
e71bf0d0 | 246 | piter->idx = 0; |
b5d0b9df TH |
247 | else |
248 | piter->idx = 1; | |
e71bf0d0 TH |
249 | |
250 | piter->flags = flags; | |
540eed56 TH |
251 | |
252 | rcu_read_unlock(); | |
e71bf0d0 TH |
253 | } |
254 | EXPORT_SYMBOL_GPL(disk_part_iter_init); | |
255 | ||
256 | /** | |
257 | * disk_part_iter_next - proceed iterator to the next partition and return it | |
258 | * @piter: iterator of interest | |
259 | * | |
260 | * Proceed @piter to the next partition and return it. | |
261 | * | |
262 | * CONTEXT: | |
263 | * Don't care. | |
264 | */ | |
265 | struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter) | |
266 | { | |
540eed56 | 267 | struct disk_part_tbl *ptbl; |
e71bf0d0 TH |
268 | int inc, end; |
269 | ||
270 | /* put the last partition */ | |
271 | disk_put_part(piter->part); | |
272 | piter->part = NULL; | |
273 | ||
540eed56 | 274 | /* get part_tbl */ |
e71bf0d0 | 275 | rcu_read_lock(); |
540eed56 | 276 | ptbl = rcu_dereference(piter->disk->part_tbl); |
e71bf0d0 TH |
277 | |
278 | /* determine iteration parameters */ | |
279 | if (piter->flags & DISK_PITER_REVERSE) { | |
280 | inc = -1; | |
71982a40 TH |
281 | if (piter->flags & (DISK_PITER_INCL_PART0 | |
282 | DISK_PITER_INCL_EMPTY_PART0)) | |
b5d0b9df TH |
283 | end = -1; |
284 | else | |
285 | end = 0; | |
e71bf0d0 TH |
286 | } else { |
287 | inc = 1; | |
540eed56 | 288 | end = ptbl->len; |
e71bf0d0 TH |
289 | } |
290 | ||
291 | /* iterate to the next partition */ | |
292 | for (; piter->idx != end; piter->idx += inc) { | |
293 | struct hd_struct *part; | |
294 | ||
540eed56 | 295 | part = rcu_dereference(ptbl->part[piter->idx]); |
e71bf0d0 TH |
296 | if (!part) |
297 | continue; | |
c83f6bf9 | 298 | if (!part_nr_sects_read(part) && |
71982a40 TH |
299 | !(piter->flags & DISK_PITER_INCL_EMPTY) && |
300 | !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 && | |
301 | piter->idx == 0)) | |
e71bf0d0 TH |
302 | continue; |
303 | ||
ed9e1982 | 304 | get_device(part_to_dev(part)); |
e71bf0d0 TH |
305 | piter->part = part; |
306 | piter->idx += inc; | |
307 | break; | |
308 | } | |
309 | ||
310 | rcu_read_unlock(); | |
311 | ||
312 | return piter->part; | |
313 | } | |
314 | EXPORT_SYMBOL_GPL(disk_part_iter_next); | |
315 | ||
316 | /** | |
317 | * disk_part_iter_exit - finish up partition iteration | |
318 | * @piter: iter of interest | |
319 | * | |
320 | * Called when iteration is over. Cleans up @piter. | |
321 | * | |
322 | * CONTEXT: | |
323 | * Don't care. | |
324 | */ | |
325 | void disk_part_iter_exit(struct disk_part_iter *piter) | |
326 | { | |
327 | disk_put_part(piter->part); | |
328 | piter->part = NULL; | |
329 | } | |
330 | EXPORT_SYMBOL_GPL(disk_part_iter_exit); | |
331 | ||
a6f23657 JA |
332 | static inline int sector_in_part(struct hd_struct *part, sector_t sector) |
333 | { | |
334 | return part->start_sect <= sector && | |
c83f6bf9 | 335 | sector < part->start_sect + part_nr_sects_read(part); |
a6f23657 JA |
336 | } |
337 | ||
e71bf0d0 TH |
338 | /** |
339 | * disk_map_sector_rcu - map sector to partition | |
340 | * @disk: gendisk of interest | |
341 | * @sector: sector to map | |
342 | * | |
343 | * Find out which partition @sector maps to on @disk. This is | |
344 | * primarily used for stats accounting. | |
345 | * | |
346 | * CONTEXT: | |
347 | * RCU read locked. The returned partition pointer is valid only | |
348 | * while preemption is disabled. | |
349 | * | |
350 | * RETURNS: | |
074a7aca | 351 | * Found partition on success, part0 is returned if no partition matches |
e71bf0d0 TH |
352 | */ |
353 | struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector) | |
354 | { | |
540eed56 | 355 | struct disk_part_tbl *ptbl; |
a6f23657 | 356 | struct hd_struct *part; |
e71bf0d0 TH |
357 | int i; |
358 | ||
540eed56 TH |
359 | ptbl = rcu_dereference(disk->part_tbl); |
360 | ||
a6f23657 JA |
361 | part = rcu_dereference(ptbl->last_lookup); |
362 | if (part && sector_in_part(part, sector)) | |
363 | return part; | |
364 | ||
540eed56 | 365 | for (i = 1; i < ptbl->len; i++) { |
a6f23657 | 366 | part = rcu_dereference(ptbl->part[i]); |
e71bf0d0 | 367 | |
a6f23657 JA |
368 | if (part && sector_in_part(part, sector)) { |
369 | rcu_assign_pointer(ptbl->last_lookup, part); | |
e71bf0d0 | 370 | return part; |
a6f23657 | 371 | } |
e71bf0d0 | 372 | } |
074a7aca | 373 | return &disk->part0; |
e71bf0d0 | 374 | } |
e71bf0d0 | 375 | |
b53df2e7 SK |
376 | /** |
377 | * disk_has_partitions | |
378 | * @disk: gendisk of interest | |
379 | * | |
380 | * Walk through the partition table and check if valid partition exists. | |
381 | * | |
382 | * CONTEXT: | |
383 | * Don't care. | |
384 | * | |
385 | * RETURNS: | |
386 | * True if the gendisk has at least one valid non-zero size partition. | |
387 | * Otherwise false. | |
388 | */ | |
389 | bool disk_has_partitions(struct gendisk *disk) | |
390 | { | |
391 | struct disk_part_tbl *ptbl; | |
392 | int i; | |
393 | bool ret = false; | |
394 | ||
395 | rcu_read_lock(); | |
396 | ptbl = rcu_dereference(disk->part_tbl); | |
397 | ||
398 | /* Iterate partitions skipping the whole device at index 0 */ | |
399 | for (i = 1; i < ptbl->len; i++) { | |
400 | if (rcu_dereference(ptbl->part[i])) { | |
401 | ret = true; | |
402 | break; | |
403 | } | |
404 | } | |
405 | ||
406 | rcu_read_unlock(); | |
407 | ||
408 | return ret; | |
409 | } | |
410 | EXPORT_SYMBOL_GPL(disk_has_partitions); | |
411 | ||
1da177e4 LT |
412 | /* |
413 | * Can be deleted altogether. Later. | |
414 | * | |
415 | */ | |
133d55cd | 416 | #define BLKDEV_MAJOR_HASH_SIZE 255 |
1da177e4 LT |
417 | static struct blk_major_name { |
418 | struct blk_major_name *next; | |
419 | int major; | |
420 | char name[16]; | |
68eef3b4 | 421 | } *major_names[BLKDEV_MAJOR_HASH_SIZE]; |
1da177e4 LT |
422 | |
423 | /* index in the above - for now: assume no multimajor ranges */ | |
e61eb2e9 | 424 | static inline int major_to_index(unsigned major) |
1da177e4 | 425 | { |
68eef3b4 | 426 | return major % BLKDEV_MAJOR_HASH_SIZE; |
7170be5f NH |
427 | } |
428 | ||
68eef3b4 | 429 | #ifdef CONFIG_PROC_FS |
cf771cb5 | 430 | void blkdev_show(struct seq_file *seqf, off_t offset) |
7170be5f | 431 | { |
68eef3b4 | 432 | struct blk_major_name *dp; |
7170be5f | 433 | |
133d55cd LG |
434 | mutex_lock(&block_class_lock); |
435 | for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next) | |
436 | if (dp->major == offset) | |
cf771cb5 | 437 | seq_printf(seqf, "%3d %s\n", dp->major, dp->name); |
133d55cd | 438 | mutex_unlock(&block_class_lock); |
1da177e4 | 439 | } |
68eef3b4 | 440 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 441 | |
9e8c0bcc MN |
442 | /** |
443 | * register_blkdev - register a new block device | |
444 | * | |
f33ff110 SB |
445 | * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If |
446 | * @major = 0, try to allocate any unused major number. | |
9e8c0bcc MN |
447 | * @name: the name of the new block device as a zero terminated string |
448 | * | |
449 | * The @name must be unique within the system. | |
450 | * | |
0e056eb5 MCC |
451 | * The return value depends on the @major input parameter: |
452 | * | |
f33ff110 SB |
453 | * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1] |
454 | * then the function returns zero on success, or a negative error code | |
0e056eb5 | 455 | * - if any unused major number was requested with @major = 0 parameter |
9e8c0bcc | 456 | * then the return value is the allocated major number in range |
f33ff110 SB |
457 | * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise |
458 | * | |
459 | * See Documentation/admin-guide/devices.txt for the list of allocated | |
460 | * major numbers. | |
9e8c0bcc | 461 | */ |
1da177e4 LT |
462 | int register_blkdev(unsigned int major, const char *name) |
463 | { | |
464 | struct blk_major_name **n, *p; | |
465 | int index, ret = 0; | |
466 | ||
edfaa7c3 | 467 | mutex_lock(&block_class_lock); |
1da177e4 LT |
468 | |
469 | /* temporary */ | |
470 | if (major == 0) { | |
471 | for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) { | |
472 | if (major_names[index] == NULL) | |
473 | break; | |
474 | } | |
475 | ||
476 | if (index == 0) { | |
dfc76d11 KP |
477 | printk("%s: failed to get major for %s\n", |
478 | __func__, name); | |
1da177e4 LT |
479 | ret = -EBUSY; |
480 | goto out; | |
481 | } | |
482 | major = index; | |
483 | ret = major; | |
484 | } | |
485 | ||
133d55cd | 486 | if (major >= BLKDEV_MAJOR_MAX) { |
dfc76d11 KP |
487 | pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n", |
488 | __func__, major, BLKDEV_MAJOR_MAX-1, name); | |
133d55cd LG |
489 | |
490 | ret = -EINVAL; | |
491 | goto out; | |
492 | } | |
493 | ||
1da177e4 LT |
494 | p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL); |
495 | if (p == NULL) { | |
496 | ret = -ENOMEM; | |
497 | goto out; | |
498 | } | |
499 | ||
500 | p->major = major; | |
501 | strlcpy(p->name, name, sizeof(p->name)); | |
502 | p->next = NULL; | |
503 | index = major_to_index(major); | |
504 | ||
505 | for (n = &major_names[index]; *n; n = &(*n)->next) { | |
506 | if ((*n)->major == major) | |
507 | break; | |
508 | } | |
509 | if (!*n) | |
510 | *n = p; | |
511 | else | |
512 | ret = -EBUSY; | |
513 | ||
514 | if (ret < 0) { | |
f33ff110 | 515 | printk("register_blkdev: cannot get major %u for %s\n", |
1da177e4 LT |
516 | major, name); |
517 | kfree(p); | |
518 | } | |
519 | out: | |
edfaa7c3 | 520 | mutex_unlock(&block_class_lock); |
1da177e4 LT |
521 | return ret; |
522 | } | |
523 | ||
524 | EXPORT_SYMBOL(register_blkdev); | |
525 | ||
f4480240 | 526 | void unregister_blkdev(unsigned int major, const char *name) |
1da177e4 LT |
527 | { |
528 | struct blk_major_name **n; | |
529 | struct blk_major_name *p = NULL; | |
530 | int index = major_to_index(major); | |
1da177e4 | 531 | |
edfaa7c3 | 532 | mutex_lock(&block_class_lock); |
1da177e4 LT |
533 | for (n = &major_names[index]; *n; n = &(*n)->next) |
534 | if ((*n)->major == major) | |
535 | break; | |
294462a5 AM |
536 | if (!*n || strcmp((*n)->name, name)) { |
537 | WARN_ON(1); | |
294462a5 | 538 | } else { |
1da177e4 LT |
539 | p = *n; |
540 | *n = p->next; | |
541 | } | |
edfaa7c3 | 542 | mutex_unlock(&block_class_lock); |
1da177e4 | 543 | kfree(p); |
1da177e4 LT |
544 | } |
545 | ||
546 | EXPORT_SYMBOL(unregister_blkdev); | |
547 | ||
548 | static struct kobj_map *bdev_map; | |
549 | ||
870d6656 TH |
550 | /** |
551 | * blk_mangle_minor - scatter minor numbers apart | |
552 | * @minor: minor number to mangle | |
553 | * | |
554 | * Scatter consecutively allocated @minor number apart if MANGLE_DEVT | |
555 | * is enabled. Mangling twice gives the original value. | |
556 | * | |
557 | * RETURNS: | |
558 | * Mangled value. | |
559 | * | |
560 | * CONTEXT: | |
561 | * Don't care. | |
562 | */ | |
563 | static int blk_mangle_minor(int minor) | |
564 | { | |
565 | #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT | |
566 | int i; | |
567 | ||
568 | for (i = 0; i < MINORBITS / 2; i++) { | |
569 | int low = minor & (1 << i); | |
570 | int high = minor & (1 << (MINORBITS - 1 - i)); | |
571 | int distance = MINORBITS - 1 - 2 * i; | |
572 | ||
573 | minor ^= low | high; /* clear both bits */ | |
574 | low <<= distance; /* swap the positions */ | |
575 | high >>= distance; | |
576 | minor |= low | high; /* and set */ | |
577 | } | |
578 | #endif | |
579 | return minor; | |
580 | } | |
581 | ||
bcce3de1 TH |
582 | /** |
583 | * blk_alloc_devt - allocate a dev_t for a partition | |
584 | * @part: partition to allocate dev_t for | |
bcce3de1 TH |
585 | * @devt: out parameter for resulting dev_t |
586 | * | |
587 | * Allocate a dev_t for block device. | |
588 | * | |
589 | * RETURNS: | |
590 | * 0 on success, allocated dev_t is returned in *@devt. -errno on | |
591 | * failure. | |
592 | * | |
593 | * CONTEXT: | |
594 | * Might sleep. | |
595 | */ | |
596 | int blk_alloc_devt(struct hd_struct *part, dev_t *devt) | |
597 | { | |
598 | struct gendisk *disk = part_to_disk(part); | |
bab998d6 | 599 | int idx; |
bcce3de1 TH |
600 | |
601 | /* in consecutive minor range? */ | |
602 | if (part->partno < disk->minors) { | |
603 | *devt = MKDEV(disk->major, disk->first_minor + part->partno); | |
604 | return 0; | |
605 | } | |
606 | ||
607 | /* allocate ext devt */ | |
2da78092 KB |
608 | idr_preload(GFP_KERNEL); |
609 | ||
4d66e5e9 | 610 | spin_lock_bh(&ext_devt_lock); |
2da78092 | 611 | idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_NOWAIT); |
4d66e5e9 | 612 | spin_unlock_bh(&ext_devt_lock); |
2da78092 KB |
613 | |
614 | idr_preload_end(); | |
bab998d6 TH |
615 | if (idx < 0) |
616 | return idx == -ENOSPC ? -EBUSY : idx; | |
bcce3de1 | 617 | |
870d6656 | 618 | *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx)); |
bcce3de1 TH |
619 | return 0; |
620 | } | |
621 | ||
622 | /** | |
623 | * blk_free_devt - free a dev_t | |
624 | * @devt: dev_t to free | |
625 | * | |
626 | * Free @devt which was allocated using blk_alloc_devt(). | |
627 | * | |
628 | * CONTEXT: | |
629 | * Might sleep. | |
630 | */ | |
631 | void blk_free_devt(dev_t devt) | |
632 | { | |
bcce3de1 TH |
633 | if (devt == MKDEV(0, 0)) |
634 | return; | |
635 | ||
636 | if (MAJOR(devt) == BLOCK_EXT_MAJOR) { | |
4d66e5e9 | 637 | spin_lock_bh(&ext_devt_lock); |
870d6656 | 638 | idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); |
4d66e5e9 | 639 | spin_unlock_bh(&ext_devt_lock); |
bcce3de1 TH |
640 | } |
641 | } | |
642 | ||
33c826ef BVA |
643 | /* |
644 | * We invalidate devt by assigning NULL pointer for devt in idr. | |
6fcc44d1 YY |
645 | */ |
646 | void blk_invalidate_devt(dev_t devt) | |
647 | { | |
648 | if (MAJOR(devt) == BLOCK_EXT_MAJOR) { | |
649 | spin_lock_bh(&ext_devt_lock); | |
650 | idr_replace(&ext_devt_idr, NULL, blk_mangle_minor(MINOR(devt))); | |
651 | spin_unlock_bh(&ext_devt_lock); | |
652 | } | |
653 | } | |
654 | ||
1f014290 TH |
655 | static char *bdevt_str(dev_t devt, char *buf) |
656 | { | |
657 | if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) { | |
658 | char tbuf[BDEVT_SIZE]; | |
659 | snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt)); | |
660 | snprintf(buf, BDEVT_SIZE, "%-9s", tbuf); | |
661 | } else | |
662 | snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt)); | |
663 | ||
664 | return buf; | |
665 | } | |
666 | ||
1da177e4 LT |
667 | /* |
668 | * Register device numbers dev..(dev+range-1) | |
669 | * range must be nonzero | |
670 | * The hash chain is sorted on range, so that subranges can override. | |
671 | */ | |
edfaa7c3 | 672 | void blk_register_region(dev_t devt, unsigned long range, struct module *module, |
1da177e4 LT |
673 | struct kobject *(*probe)(dev_t, int *, void *), |
674 | int (*lock)(dev_t, void *), void *data) | |
675 | { | |
edfaa7c3 | 676 | kobj_map(bdev_map, devt, range, module, probe, lock, data); |
1da177e4 LT |
677 | } |
678 | ||
679 | EXPORT_SYMBOL(blk_register_region); | |
680 | ||
edfaa7c3 | 681 | void blk_unregister_region(dev_t devt, unsigned long range) |
1da177e4 | 682 | { |
edfaa7c3 | 683 | kobj_unmap(bdev_map, devt, range); |
1da177e4 LT |
684 | } |
685 | ||
686 | EXPORT_SYMBOL(blk_unregister_region); | |
687 | ||
cf771cb5 | 688 | static struct kobject *exact_match(dev_t devt, int *partno, void *data) |
1da177e4 LT |
689 | { |
690 | struct gendisk *p = data; | |
edfaa7c3 | 691 | |
ed9e1982 | 692 | return &disk_to_dev(p)->kobj; |
1da177e4 LT |
693 | } |
694 | ||
edfaa7c3 | 695 | static int exact_lock(dev_t devt, void *data) |
1da177e4 LT |
696 | { |
697 | struct gendisk *p = data; | |
698 | ||
3079c22e | 699 | if (!get_disk_and_module(p)) |
1da177e4 LT |
700 | return -1; |
701 | return 0; | |
702 | } | |
703 | ||
fef912bf HR |
704 | static void register_disk(struct device *parent, struct gendisk *disk, |
705 | const struct attribute_group **groups) | |
d2bf1b67 TH |
706 | { |
707 | struct device *ddev = disk_to_dev(disk); | |
708 | struct block_device *bdev; | |
709 | struct disk_part_iter piter; | |
710 | struct hd_struct *part; | |
711 | int err; | |
712 | ||
e63a46be | 713 | ddev->parent = parent; |
d2bf1b67 | 714 | |
ffc8b308 | 715 | dev_set_name(ddev, "%s", disk->disk_name); |
d2bf1b67 TH |
716 | |
717 | /* delay uevents, until we scanned partition table */ | |
718 | dev_set_uevent_suppress(ddev, 1); | |
719 | ||
fef912bf HR |
720 | if (groups) { |
721 | WARN_ON(ddev->groups); | |
722 | ddev->groups = groups; | |
723 | } | |
d2bf1b67 TH |
724 | if (device_add(ddev)) |
725 | return; | |
726 | if (!sysfs_deprecated) { | |
727 | err = sysfs_create_link(block_depr, &ddev->kobj, | |
728 | kobject_name(&ddev->kobj)); | |
729 | if (err) { | |
730 | device_del(ddev); | |
731 | return; | |
732 | } | |
733 | } | |
25e823c8 ML |
734 | |
735 | /* | |
736 | * avoid probable deadlock caused by allocating memory with | |
737 | * GFP_KERNEL in runtime_resume callback of its all ancestor | |
738 | * devices | |
739 | */ | |
740 | pm_runtime_set_memalloc_noio(ddev, true); | |
741 | ||
d2bf1b67 TH |
742 | disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj); |
743 | disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj); | |
744 | ||
8ddcd653 CH |
745 | if (disk->flags & GENHD_FL_HIDDEN) { |
746 | dev_set_uevent_suppress(ddev, 0); | |
747 | return; | |
748 | } | |
749 | ||
d2bf1b67 | 750 | /* No minors to use for partitions */ |
d27769ec | 751 | if (!disk_part_scan_enabled(disk)) |
d2bf1b67 TH |
752 | goto exit; |
753 | ||
754 | /* No such device (e.g., media were just removed) */ | |
755 | if (!get_capacity(disk)) | |
756 | goto exit; | |
757 | ||
758 | bdev = bdget_disk(disk, 0); | |
759 | if (!bdev) | |
760 | goto exit; | |
761 | ||
762 | bdev->bd_invalidated = 1; | |
763 | err = blkdev_get(bdev, FMODE_READ, NULL); | |
764 | if (err < 0) | |
765 | goto exit; | |
766 | blkdev_put(bdev, FMODE_READ); | |
767 | ||
768 | exit: | |
769 | /* announce disk after possible partitions are created */ | |
770 | dev_set_uevent_suppress(ddev, 0); | |
771 | kobject_uevent(&ddev->kobj, KOBJ_ADD); | |
772 | ||
773 | /* announce possible partitions */ | |
774 | disk_part_iter_init(&piter, disk, 0); | |
775 | while ((part = disk_part_iter_next(&piter))) | |
776 | kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD); | |
777 | disk_part_iter_exit(&piter); | |
8ddcd653 | 778 | |
4d7c1d3f | 779 | if (disk->queue->backing_dev_info->dev) { |
780 | err = sysfs_create_link(&ddev->kobj, | |
781 | &disk->queue->backing_dev_info->dev->kobj, | |
782 | "bdi"); | |
783 | WARN_ON(err); | |
784 | } | |
d2bf1b67 TH |
785 | } |
786 | ||
1da177e4 | 787 | /** |
fa70d2e2 | 788 | * __device_add_disk - add disk information to kernel list |
e63a46be | 789 | * @parent: parent device for the disk |
1da177e4 | 790 | * @disk: per-device partitioning information |
fef912bf | 791 | * @groups: Additional per-device sysfs groups |
fa70d2e2 | 792 | * @register_queue: register the queue if set to true |
1da177e4 LT |
793 | * |
794 | * This function registers the partitioning information in @disk | |
795 | * with the kernel. | |
3e1a7ff8 TH |
796 | * |
797 | * FIXME: error handling | |
1da177e4 | 798 | */ |
fa70d2e2 | 799 | static void __device_add_disk(struct device *parent, struct gendisk *disk, |
fef912bf | 800 | const struct attribute_group **groups, |
fa70d2e2 | 801 | bool register_queue) |
1da177e4 | 802 | { |
3e1a7ff8 | 803 | dev_t devt; |
6ffeea77 | 804 | int retval; |
cf0ca9fe | 805 | |
737eb78e DLM |
806 | /* |
807 | * The disk queue should now be all set with enough information about | |
808 | * the device for the elevator code to pick an adequate default | |
809 | * elevator if one is needed, that is, for devices requesting queue | |
810 | * registration. | |
811 | */ | |
812 | if (register_queue) | |
813 | elevator_init_mq(disk->queue); | |
814 | ||
3e1a7ff8 TH |
815 | /* minors == 0 indicates to use ext devt from part0 and should |
816 | * be accompanied with EXT_DEVT flag. Make sure all | |
817 | * parameters make sense. | |
818 | */ | |
819 | WARN_ON(disk->minors && !(disk->major || disk->first_minor)); | |
8ddcd653 CH |
820 | WARN_ON(!disk->minors && |
821 | !(disk->flags & (GENHD_FL_EXT_DEVT | GENHD_FL_HIDDEN))); | |
3e1a7ff8 | 822 | |
1da177e4 | 823 | disk->flags |= GENHD_FL_UP; |
3e1a7ff8 TH |
824 | |
825 | retval = blk_alloc_devt(&disk->part0, &devt); | |
826 | if (retval) { | |
827 | WARN_ON(1); | |
828 | return; | |
829 | } | |
3e1a7ff8 TH |
830 | disk->major = MAJOR(devt); |
831 | disk->first_minor = MINOR(devt); | |
832 | ||
9f53d2fe SG |
833 | disk_alloc_events(disk); |
834 | ||
8ddcd653 CH |
835 | if (disk->flags & GENHD_FL_HIDDEN) { |
836 | /* | |
837 | * Don't let hidden disks show up in /proc/partitions, | |
838 | * and don't bother scanning for partitions either. | |
839 | */ | |
840 | disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO; | |
841 | disk->flags |= GENHD_FL_NO_PART_SCAN; | |
842 | } else { | |
3a92168b | 843 | int ret; |
844 | ||
8ddcd653 CH |
845 | /* Register BDI before referencing it from bdev */ |
846 | disk_to_dev(disk)->devt = devt; | |
3a92168b | 847 | ret = bdi_register_owner(disk->queue->backing_dev_info, |
848 | disk_to_dev(disk)); | |
849 | WARN_ON(ret); | |
8ddcd653 CH |
850 | blk_register_region(disk_devt(disk), disk->minors, NULL, |
851 | exact_match, exact_lock, disk); | |
852 | } | |
fef912bf | 853 | register_disk(parent, disk, groups); |
fa70d2e2 MS |
854 | if (register_queue) |
855 | blk_register_queue(disk); | |
cf0ca9fe | 856 | |
523e1d39 TH |
857 | /* |
858 | * Take an extra ref on queue which will be put on disk_release() | |
859 | * so that it sticks around as long as @disk is there. | |
860 | */ | |
09ac46c4 | 861 | WARN_ON_ONCE(!blk_get_queue(disk->queue)); |
523e1d39 | 862 | |
77ea887e | 863 | disk_add_events(disk); |
25520d55 | 864 | blk_integrity_add(disk); |
1da177e4 | 865 | } |
fa70d2e2 | 866 | |
fef912bf HR |
867 | void device_add_disk(struct device *parent, struct gendisk *disk, |
868 | const struct attribute_group **groups) | |
869 | ||
fa70d2e2 | 870 | { |
fef912bf | 871 | __device_add_disk(parent, disk, groups, true); |
fa70d2e2 | 872 | } |
e63a46be | 873 | EXPORT_SYMBOL(device_add_disk); |
1da177e4 | 874 | |
fa70d2e2 MS |
875 | void device_add_disk_no_queue_reg(struct device *parent, struct gendisk *disk) |
876 | { | |
fef912bf | 877 | __device_add_disk(parent, disk, NULL, false); |
fa70d2e2 MS |
878 | } |
879 | EXPORT_SYMBOL(device_add_disk_no_queue_reg); | |
880 | ||
02d33b67 CH |
881 | static void invalidate_partition(struct gendisk *disk, int partno) |
882 | { | |
883 | struct block_device *bdev; | |
884 | ||
885 | bdev = bdget_disk(disk, partno); | |
886 | if (!bdev) | |
887 | return; | |
888 | ||
889 | fsync_bdev(bdev); | |
890 | __invalidate_device(bdev, true); | |
9bc5c397 CH |
891 | |
892 | /* | |
893 | * Unhash the bdev inode for this device so that it gets evicted as soon | |
894 | * as last inode reference is dropped. | |
895 | */ | |
896 | remove_inode_hash(bdev->bd_inode); | |
02d33b67 CH |
897 | bdput(bdev); |
898 | } | |
899 | ||
d2bf1b67 | 900 | void del_gendisk(struct gendisk *disk) |
1da177e4 | 901 | { |
d2bf1b67 TH |
902 | struct disk_part_iter piter; |
903 | struct hd_struct *part; | |
904 | ||
25520d55 | 905 | blk_integrity_del(disk); |
77ea887e TH |
906 | disk_del_events(disk); |
907 | ||
56c0908c JK |
908 | /* |
909 | * Block lookups of the disk until all bdevs are unhashed and the | |
910 | * disk is marked as dead (GENHD_FL_UP cleared). | |
911 | */ | |
912 | down_write(&disk->lookup_sem); | |
d2bf1b67 TH |
913 | /* invalidate stuff */ |
914 | disk_part_iter_init(&piter, disk, | |
915 | DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE); | |
916 | while ((part = disk_part_iter_next(&piter))) { | |
917 | invalidate_partition(disk, part->partno); | |
cddae808 | 918 | delete_partition(disk, part); |
d2bf1b67 TH |
919 | } |
920 | disk_part_iter_exit(&piter); | |
921 | ||
922 | invalidate_partition(disk, 0); | |
d2bf1b67 TH |
923 | set_capacity(disk, 0); |
924 | disk->flags &= ~GENHD_FL_UP; | |
56c0908c | 925 | up_write(&disk->lookup_sem); |
d2bf1b67 | 926 | |
8ddcd653 CH |
927 | if (!(disk->flags & GENHD_FL_HIDDEN)) |
928 | sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi"); | |
90f16fdd JK |
929 | if (disk->queue) { |
930 | /* | |
931 | * Unregister bdi before releasing device numbers (as they can | |
932 | * get reused and we'd get clashes in sysfs). | |
933 | */ | |
bc8d062c MS |
934 | if (!(disk->flags & GENHD_FL_HIDDEN)) |
935 | bdi_unregister(disk->queue->backing_dev_info); | |
90f16fdd JK |
936 | blk_unregister_queue(disk); |
937 | } else { | |
938 | WARN_ON(1); | |
939 | } | |
d2bf1b67 | 940 | |
17eac099 | 941 | if (!(disk->flags & GENHD_FL_HIDDEN)) |
8ddcd653 | 942 | blk_unregister_region(disk_devt(disk), disk->minors); |
6fcc44d1 YY |
943 | /* |
944 | * Remove gendisk pointer from idr so that it cannot be looked up | |
945 | * while RCU period before freeing gendisk is running to prevent | |
946 | * use-after-free issues. Note that the device number stays | |
947 | * "in-use" until we really free the gendisk. | |
948 | */ | |
949 | blk_invalidate_devt(disk_devt(disk)); | |
d2bf1b67 TH |
950 | |
951 | kobject_put(disk->part0.holder_dir); | |
952 | kobject_put(disk->slave_dir); | |
d2bf1b67 TH |
953 | |
954 | part_stat_set_all(&disk->part0, 0); | |
955 | disk->part0.stamp = 0; | |
d2bf1b67 TH |
956 | if (!sysfs_deprecated) |
957 | sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk))); | |
25e823c8 | 958 | pm_runtime_set_memalloc_noio(disk_to_dev(disk), false); |
d2bf1b67 | 959 | device_del(disk_to_dev(disk)); |
1da177e4 | 960 | } |
d2bf1b67 | 961 | EXPORT_SYMBOL(del_gendisk); |
1da177e4 | 962 | |
99e6608c VV |
963 | /* sysfs access to bad-blocks list. */ |
964 | static ssize_t disk_badblocks_show(struct device *dev, | |
965 | struct device_attribute *attr, | |
966 | char *page) | |
967 | { | |
968 | struct gendisk *disk = dev_to_disk(dev); | |
969 | ||
970 | if (!disk->bb) | |
971 | return sprintf(page, "\n"); | |
972 | ||
973 | return badblocks_show(disk->bb, page, 0); | |
974 | } | |
975 | ||
976 | static ssize_t disk_badblocks_store(struct device *dev, | |
977 | struct device_attribute *attr, | |
978 | const char *page, size_t len) | |
979 | { | |
980 | struct gendisk *disk = dev_to_disk(dev); | |
981 | ||
982 | if (!disk->bb) | |
983 | return -ENXIO; | |
984 | ||
985 | return badblocks_store(disk->bb, page, len, 0); | |
986 | } | |
987 | ||
1da177e4 LT |
988 | /** |
989 | * get_gendisk - get partitioning information for a given device | |
710027a4 | 990 | * @devt: device to get partitioning information for |
496aa8a9 | 991 | * @partno: returned partition index |
1da177e4 LT |
992 | * |
993 | * This function gets the structure containing partitioning | |
710027a4 | 994 | * information for the given device @devt. |
1da177e4 | 995 | */ |
cf771cb5 | 996 | struct gendisk *get_gendisk(dev_t devt, int *partno) |
1da177e4 | 997 | { |
bcce3de1 TH |
998 | struct gendisk *disk = NULL; |
999 | ||
1000 | if (MAJOR(devt) != BLOCK_EXT_MAJOR) { | |
1001 | struct kobject *kobj; | |
1002 | ||
1003 | kobj = kobj_lookup(bdev_map, devt, partno); | |
1004 | if (kobj) | |
1005 | disk = dev_to_disk(kobj_to_dev(kobj)); | |
1006 | } else { | |
1007 | struct hd_struct *part; | |
1008 | ||
4d66e5e9 | 1009 | spin_lock_bh(&ext_devt_lock); |
870d6656 | 1010 | part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); |
3079c22e | 1011 | if (part && get_disk_and_module(part_to_disk(part))) { |
bcce3de1 TH |
1012 | *partno = part->partno; |
1013 | disk = part_to_disk(part); | |
1014 | } | |
4d66e5e9 | 1015 | spin_unlock_bh(&ext_devt_lock); |
bcce3de1 | 1016 | } |
edfaa7c3 | 1017 | |
56c0908c JK |
1018 | if (!disk) |
1019 | return NULL; | |
1020 | ||
1021 | /* | |
1022 | * Synchronize with del_gendisk() to not return disk that is being | |
1023 | * destroyed. | |
1024 | */ | |
1025 | down_read(&disk->lookup_sem); | |
1026 | if (unlikely((disk->flags & GENHD_FL_HIDDEN) || | |
1027 | !(disk->flags & GENHD_FL_UP))) { | |
1028 | up_read(&disk->lookup_sem); | |
9df6c299 | 1029 | put_disk_and_module(disk); |
8ddcd653 | 1030 | disk = NULL; |
56c0908c JK |
1031 | } else { |
1032 | up_read(&disk->lookup_sem); | |
8ddcd653 | 1033 | } |
bcce3de1 | 1034 | return disk; |
1da177e4 LT |
1035 | } |
1036 | ||
f331c029 TH |
1037 | /** |
1038 | * bdget_disk - do bdget() by gendisk and partition number | |
1039 | * @disk: gendisk of interest | |
1040 | * @partno: partition number | |
1041 | * | |
1042 | * Find partition @partno from @disk, do bdget() on it. | |
1043 | * | |
1044 | * CONTEXT: | |
1045 | * Don't care. | |
1046 | * | |
1047 | * RETURNS: | |
1048 | * Resulting block_device on success, NULL on failure. | |
1049 | */ | |
aeb3d3a8 | 1050 | struct block_device *bdget_disk(struct gendisk *disk, int partno) |
f331c029 | 1051 | { |
548b10eb TH |
1052 | struct hd_struct *part; |
1053 | struct block_device *bdev = NULL; | |
f331c029 | 1054 | |
548b10eb | 1055 | part = disk_get_part(disk, partno); |
2bbedcb4 | 1056 | if (part) |
548b10eb TH |
1057 | bdev = bdget(part_devt(part)); |
1058 | disk_put_part(part); | |
f331c029 | 1059 | |
548b10eb | 1060 | return bdev; |
f331c029 TH |
1061 | } |
1062 | EXPORT_SYMBOL(bdget_disk); | |
1063 | ||
5c6f35c5 GKH |
1064 | /* |
1065 | * print a full list of all partitions - intended for places where the root | |
1066 | * filesystem can't be mounted and thus to give the victim some idea of what | |
1067 | * went wrong | |
1068 | */ | |
1069 | void __init printk_all_partitions(void) | |
1070 | { | |
def4e38d TH |
1071 | struct class_dev_iter iter; |
1072 | struct device *dev; | |
1073 | ||
1074 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); | |
1075 | while ((dev = class_dev_iter_next(&iter))) { | |
1076 | struct gendisk *disk = dev_to_disk(dev); | |
e71bf0d0 TH |
1077 | struct disk_part_iter piter; |
1078 | struct hd_struct *part; | |
1f014290 TH |
1079 | char name_buf[BDEVNAME_SIZE]; |
1080 | char devt_buf[BDEVT_SIZE]; | |
def4e38d TH |
1081 | |
1082 | /* | |
1083 | * Don't show empty devices or things that have been | |
25985edc | 1084 | * suppressed |
def4e38d TH |
1085 | */ |
1086 | if (get_capacity(disk) == 0 || | |
1087 | (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)) | |
1088 | continue; | |
1089 | ||
1090 | /* | |
1091 | * Note, unlike /proc/partitions, I am showing the | |
1092 | * numbers in hex - the same format as the root= | |
1093 | * option takes. | |
1094 | */ | |
074a7aca TH |
1095 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0); |
1096 | while ((part = disk_part_iter_next(&piter))) { | |
1097 | bool is_part0 = part == &disk->part0; | |
def4e38d | 1098 | |
b5af921e | 1099 | printk("%s%s %10llu %s %s", is_part0 ? "" : " ", |
1f014290 | 1100 | bdevt_str(part_devt(part), devt_buf), |
c83f6bf9 VG |
1101 | (unsigned long long)part_nr_sects_read(part) >> 1 |
1102 | , disk_name(disk, part->partno, name_buf), | |
1ad7e899 | 1103 | part->info ? part->info->uuid : ""); |
074a7aca | 1104 | if (is_part0) { |
52c44d93 | 1105 | if (dev->parent && dev->parent->driver) |
074a7aca | 1106 | printk(" driver: %s\n", |
52c44d93 | 1107 | dev->parent->driver->name); |
074a7aca TH |
1108 | else |
1109 | printk(" (driver?)\n"); | |
1110 | } else | |
1111 | printk("\n"); | |
1112 | } | |
e71bf0d0 | 1113 | disk_part_iter_exit(&piter); |
def4e38d TH |
1114 | } |
1115 | class_dev_iter_exit(&iter); | |
dd2a345f DG |
1116 | } |
1117 | ||
1da177e4 LT |
1118 | #ifdef CONFIG_PROC_FS |
1119 | /* iterator */ | |
def4e38d | 1120 | static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos) |
68c4d4a7 | 1121 | { |
def4e38d TH |
1122 | loff_t skip = *pos; |
1123 | struct class_dev_iter *iter; | |
1124 | struct device *dev; | |
68c4d4a7 | 1125 | |
aeb3d3a8 | 1126 | iter = kmalloc(sizeof(*iter), GFP_KERNEL); |
def4e38d TH |
1127 | if (!iter) |
1128 | return ERR_PTR(-ENOMEM); | |
1129 | ||
1130 | seqf->private = iter; | |
1131 | class_dev_iter_init(iter, &block_class, NULL, &disk_type); | |
1132 | do { | |
1133 | dev = class_dev_iter_next(iter); | |
1134 | if (!dev) | |
1135 | return NULL; | |
1136 | } while (skip--); | |
1137 | ||
1138 | return dev_to_disk(dev); | |
68c4d4a7 GKH |
1139 | } |
1140 | ||
def4e38d | 1141 | static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos) |
1da177e4 | 1142 | { |
edfaa7c3 | 1143 | struct device *dev; |
1da177e4 | 1144 | |
def4e38d TH |
1145 | (*pos)++; |
1146 | dev = class_dev_iter_next(seqf->private); | |
2ac3cee5 | 1147 | if (dev) |
68c4d4a7 | 1148 | return dev_to_disk(dev); |
2ac3cee5 | 1149 | |
1da177e4 LT |
1150 | return NULL; |
1151 | } | |
1152 | ||
def4e38d | 1153 | static void disk_seqf_stop(struct seq_file *seqf, void *v) |
27f30251 | 1154 | { |
def4e38d | 1155 | struct class_dev_iter *iter = seqf->private; |
27f30251 | 1156 | |
def4e38d TH |
1157 | /* stop is called even after start failed :-( */ |
1158 | if (iter) { | |
1159 | class_dev_iter_exit(iter); | |
1160 | kfree(iter); | |
77da1605 | 1161 | seqf->private = NULL; |
5c0ef6d0 | 1162 | } |
1da177e4 LT |
1163 | } |
1164 | ||
def4e38d | 1165 | static void *show_partition_start(struct seq_file *seqf, loff_t *pos) |
1da177e4 | 1166 | { |
06768067 | 1167 | void *p; |
def4e38d TH |
1168 | |
1169 | p = disk_seqf_start(seqf, pos); | |
b9f985b6 | 1170 | if (!IS_ERR_OR_NULL(p) && !*pos) |
def4e38d TH |
1171 | seq_puts(seqf, "major minor #blocks name\n\n"); |
1172 | return p; | |
1da177e4 LT |
1173 | } |
1174 | ||
cf771cb5 | 1175 | static int show_partition(struct seq_file *seqf, void *v) |
1da177e4 LT |
1176 | { |
1177 | struct gendisk *sgp = v; | |
e71bf0d0 TH |
1178 | struct disk_part_iter piter; |
1179 | struct hd_struct *part; | |
1da177e4 LT |
1180 | char buf[BDEVNAME_SIZE]; |
1181 | ||
1da177e4 | 1182 | /* Don't show non-partitionable removeable devices or empty devices */ |
d27769ec | 1183 | if (!get_capacity(sgp) || (!disk_max_parts(sgp) && |
f331c029 | 1184 | (sgp->flags & GENHD_FL_REMOVABLE))) |
1da177e4 LT |
1185 | return 0; |
1186 | if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO) | |
1187 | return 0; | |
1188 | ||
1189 | /* show the full disk and all non-0 size partitions of it */ | |
074a7aca | 1190 | disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0); |
e71bf0d0 | 1191 | while ((part = disk_part_iter_next(&piter))) |
1f014290 | 1192 | seq_printf(seqf, "%4d %7d %10llu %s\n", |
f331c029 | 1193 | MAJOR(part_devt(part)), MINOR(part_devt(part)), |
c83f6bf9 | 1194 | (unsigned long long)part_nr_sects_read(part) >> 1, |
f331c029 | 1195 | disk_name(sgp, part->partno, buf)); |
e71bf0d0 | 1196 | disk_part_iter_exit(&piter); |
1da177e4 LT |
1197 | |
1198 | return 0; | |
1199 | } | |
1200 | ||
f500975a | 1201 | static const struct seq_operations partitions_op = { |
def4e38d TH |
1202 | .start = show_partition_start, |
1203 | .next = disk_seqf_next, | |
1204 | .stop = disk_seqf_stop, | |
edfaa7c3 | 1205 | .show = show_partition |
1da177e4 LT |
1206 | }; |
1207 | #endif | |
1208 | ||
1209 | ||
cf771cb5 | 1210 | static struct kobject *base_probe(dev_t devt, int *partno, void *data) |
1da177e4 | 1211 | { |
edfaa7c3 | 1212 | if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0) |
1da177e4 | 1213 | /* Make old-style 2.4 aliases work */ |
edfaa7c3 | 1214 | request_module("block-major-%d", MAJOR(devt)); |
1da177e4 LT |
1215 | return NULL; |
1216 | } | |
1217 | ||
1218 | static int __init genhd_device_init(void) | |
1219 | { | |
e105b8bf DW |
1220 | int error; |
1221 | ||
1222 | block_class.dev_kobj = sysfs_dev_block_kobj; | |
1223 | error = class_register(&block_class); | |
ee27a558 RM |
1224 | if (unlikely(error)) |
1225 | return error; | |
edfaa7c3 | 1226 | bdev_map = kobj_map_init(base_probe, &block_class_lock); |
1da177e4 | 1227 | blk_dev_init(); |
edfaa7c3 | 1228 | |
561ec68e ZY |
1229 | register_blkdev(BLOCK_EXT_MAJOR, "blkext"); |
1230 | ||
edfaa7c3 | 1231 | /* create top-level block dir */ |
e52eec13 AK |
1232 | if (!sysfs_deprecated) |
1233 | block_depr = kobject_create_and_add("block", NULL); | |
830d3cfb | 1234 | return 0; |
1da177e4 LT |
1235 | } |
1236 | ||
1237 | subsys_initcall(genhd_device_init); | |
1238 | ||
edfaa7c3 KS |
1239 | static ssize_t disk_range_show(struct device *dev, |
1240 | struct device_attribute *attr, char *buf) | |
1da177e4 | 1241 | { |
edfaa7c3 | 1242 | struct gendisk *disk = dev_to_disk(dev); |
1da177e4 | 1243 | |
edfaa7c3 | 1244 | return sprintf(buf, "%d\n", disk->minors); |
1da177e4 LT |
1245 | } |
1246 | ||
1f014290 TH |
1247 | static ssize_t disk_ext_range_show(struct device *dev, |
1248 | struct device_attribute *attr, char *buf) | |
1249 | { | |
1250 | struct gendisk *disk = dev_to_disk(dev); | |
1251 | ||
b5d0b9df | 1252 | return sprintf(buf, "%d\n", disk_max_parts(disk)); |
1f014290 TH |
1253 | } |
1254 | ||
edfaa7c3 KS |
1255 | static ssize_t disk_removable_show(struct device *dev, |
1256 | struct device_attribute *attr, char *buf) | |
a7fd6706 | 1257 | { |
edfaa7c3 | 1258 | struct gendisk *disk = dev_to_disk(dev); |
a7fd6706 | 1259 | |
edfaa7c3 KS |
1260 | return sprintf(buf, "%d\n", |
1261 | (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0)); | |
a7fd6706 KS |
1262 | } |
1263 | ||
8ddcd653 CH |
1264 | static ssize_t disk_hidden_show(struct device *dev, |
1265 | struct device_attribute *attr, char *buf) | |
1266 | { | |
1267 | struct gendisk *disk = dev_to_disk(dev); | |
1268 | ||
1269 | return sprintf(buf, "%d\n", | |
1270 | (disk->flags & GENHD_FL_HIDDEN ? 1 : 0)); | |
1271 | } | |
1272 | ||
1c9ce527 KS |
1273 | static ssize_t disk_ro_show(struct device *dev, |
1274 | struct device_attribute *attr, char *buf) | |
1275 | { | |
1276 | struct gendisk *disk = dev_to_disk(dev); | |
1277 | ||
b7db9956 | 1278 | return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0); |
1c9ce527 KS |
1279 | } |
1280 | ||
3ad5cee5 CH |
1281 | ssize_t part_size_show(struct device *dev, |
1282 | struct device_attribute *attr, char *buf) | |
1283 | { | |
1284 | struct hd_struct *p = dev_to_part(dev); | |
1285 | ||
1286 | return sprintf(buf, "%llu\n", | |
1287 | (unsigned long long)part_nr_sects_read(p)); | |
1288 | } | |
1289 | ||
1290 | ssize_t part_stat_show(struct device *dev, | |
1291 | struct device_attribute *attr, char *buf) | |
1292 | { | |
1293 | struct hd_struct *p = dev_to_part(dev); | |
1294 | struct request_queue *q = part_to_disk(p)->queue; | |
ea18e0f0 | 1295 | struct disk_stats stat; |
3ad5cee5 CH |
1296 | unsigned int inflight; |
1297 | ||
ea18e0f0 | 1298 | part_stat_read_all(p, &stat); |
3ad5cee5 | 1299 | inflight = part_in_flight(q, p); |
ea18e0f0 | 1300 | |
3ad5cee5 CH |
1301 | return sprintf(buf, |
1302 | "%8lu %8lu %8llu %8u " | |
1303 | "%8lu %8lu %8llu %8u " | |
1304 | "%8u %8u %8u " | |
1305 | "%8lu %8lu %8llu %8u " | |
1306 | "%8lu %8u" | |
1307 | "\n", | |
ea18e0f0 KK |
1308 | stat.ios[STAT_READ], |
1309 | stat.merges[STAT_READ], | |
1310 | (unsigned long long)stat.sectors[STAT_READ], | |
1311 | (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC), | |
1312 | stat.ios[STAT_WRITE], | |
1313 | stat.merges[STAT_WRITE], | |
1314 | (unsigned long long)stat.sectors[STAT_WRITE], | |
1315 | (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC), | |
3ad5cee5 | 1316 | inflight, |
ea18e0f0 | 1317 | jiffies_to_msecs(stat.io_ticks), |
8cd5b8fc KK |
1318 | (unsigned int)div_u64(stat.nsecs[STAT_READ] + |
1319 | stat.nsecs[STAT_WRITE] + | |
1320 | stat.nsecs[STAT_DISCARD] + | |
1321 | stat.nsecs[STAT_FLUSH], | |
1322 | NSEC_PER_MSEC), | |
ea18e0f0 KK |
1323 | stat.ios[STAT_DISCARD], |
1324 | stat.merges[STAT_DISCARD], | |
1325 | (unsigned long long)stat.sectors[STAT_DISCARD], | |
1326 | (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC), | |
1327 | stat.ios[STAT_FLUSH], | |
1328 | (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC)); | |
3ad5cee5 CH |
1329 | } |
1330 | ||
1331 | ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr, | |
1332 | char *buf) | |
1333 | { | |
1334 | struct hd_struct *p = dev_to_part(dev); | |
1335 | struct request_queue *q = part_to_disk(p)->queue; | |
1336 | unsigned int inflight[2]; | |
1337 | ||
1338 | part_in_flight_rw(q, p, inflight); | |
1339 | return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]); | |
1340 | } | |
1341 | ||
edfaa7c3 KS |
1342 | static ssize_t disk_capability_show(struct device *dev, |
1343 | struct device_attribute *attr, char *buf) | |
86ce18d7 | 1344 | { |
edfaa7c3 KS |
1345 | struct gendisk *disk = dev_to_disk(dev); |
1346 | ||
1347 | return sprintf(buf, "%x\n", disk->flags); | |
86ce18d7 | 1348 | } |
edfaa7c3 | 1349 | |
c72758f3 MP |
1350 | static ssize_t disk_alignment_offset_show(struct device *dev, |
1351 | struct device_attribute *attr, | |
1352 | char *buf) | |
1353 | { | |
1354 | struct gendisk *disk = dev_to_disk(dev); | |
1355 | ||
1356 | return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue)); | |
1357 | } | |
1358 | ||
86b37281 MP |
1359 | static ssize_t disk_discard_alignment_show(struct device *dev, |
1360 | struct device_attribute *attr, | |
1361 | char *buf) | |
1362 | { | |
1363 | struct gendisk *disk = dev_to_disk(dev); | |
1364 | ||
dd3d145d | 1365 | return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue)); |
86b37281 MP |
1366 | } |
1367 | ||
5657a819 JP |
1368 | static DEVICE_ATTR(range, 0444, disk_range_show, NULL); |
1369 | static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL); | |
1370 | static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL); | |
1371 | static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL); | |
1372 | static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL); | |
1373 | static DEVICE_ATTR(size, 0444, part_size_show, NULL); | |
1374 | static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL); | |
1375 | static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL); | |
1376 | static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL); | |
1377 | static DEVICE_ATTR(stat, 0444, part_stat_show, NULL); | |
1378 | static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL); | |
1379 | static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store); | |
3ad5cee5 | 1380 | |
c17bb495 | 1381 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
3ad5cee5 CH |
1382 | ssize_t part_fail_show(struct device *dev, |
1383 | struct device_attribute *attr, char *buf) | |
1384 | { | |
1385 | struct hd_struct *p = dev_to_part(dev); | |
1386 | ||
1387 | return sprintf(buf, "%d\n", p->make_it_fail); | |
1388 | } | |
1389 | ||
1390 | ssize_t part_fail_store(struct device *dev, | |
1391 | struct device_attribute *attr, | |
1392 | const char *buf, size_t count) | |
1393 | { | |
1394 | struct hd_struct *p = dev_to_part(dev); | |
1395 | int i; | |
1396 | ||
1397 | if (count > 0 && sscanf(buf, "%d", &i) > 0) | |
1398 | p->make_it_fail = (i == 0) ? 0 : 1; | |
1399 | ||
1400 | return count; | |
1401 | } | |
1402 | ||
edfaa7c3 | 1403 | static struct device_attribute dev_attr_fail = |
5657a819 | 1404 | __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store); |
3ad5cee5 CH |
1405 | #endif /* CONFIG_FAIL_MAKE_REQUEST */ |
1406 | ||
581d4e28 JA |
1407 | #ifdef CONFIG_FAIL_IO_TIMEOUT |
1408 | static struct device_attribute dev_attr_fail_timeout = | |
5657a819 | 1409 | __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store); |
581d4e28 | 1410 | #endif |
edfaa7c3 KS |
1411 | |
1412 | static struct attribute *disk_attrs[] = { | |
1413 | &dev_attr_range.attr, | |
1f014290 | 1414 | &dev_attr_ext_range.attr, |
edfaa7c3 | 1415 | &dev_attr_removable.attr, |
8ddcd653 | 1416 | &dev_attr_hidden.attr, |
1c9ce527 | 1417 | &dev_attr_ro.attr, |
edfaa7c3 | 1418 | &dev_attr_size.attr, |
c72758f3 | 1419 | &dev_attr_alignment_offset.attr, |
86b37281 | 1420 | &dev_attr_discard_alignment.attr, |
edfaa7c3 KS |
1421 | &dev_attr_capability.attr, |
1422 | &dev_attr_stat.attr, | |
316d315b | 1423 | &dev_attr_inflight.attr, |
99e6608c | 1424 | &dev_attr_badblocks.attr, |
edfaa7c3 KS |
1425 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
1426 | &dev_attr_fail.attr, | |
581d4e28 JA |
1427 | #endif |
1428 | #ifdef CONFIG_FAIL_IO_TIMEOUT | |
1429 | &dev_attr_fail_timeout.attr, | |
edfaa7c3 KS |
1430 | #endif |
1431 | NULL | |
1432 | }; | |
1433 | ||
9438b3e0 DW |
1434 | static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n) |
1435 | { | |
1436 | struct device *dev = container_of(kobj, typeof(*dev), kobj); | |
1437 | struct gendisk *disk = dev_to_disk(dev); | |
1438 | ||
1439 | if (a == &dev_attr_badblocks.attr && !disk->bb) | |
1440 | return 0; | |
1441 | return a->mode; | |
1442 | } | |
1443 | ||
edfaa7c3 KS |
1444 | static struct attribute_group disk_attr_group = { |
1445 | .attrs = disk_attrs, | |
9438b3e0 | 1446 | .is_visible = disk_visible, |
edfaa7c3 KS |
1447 | }; |
1448 | ||
a4dbd674 | 1449 | static const struct attribute_group *disk_attr_groups[] = { |
edfaa7c3 KS |
1450 | &disk_attr_group, |
1451 | NULL | |
1da177e4 LT |
1452 | }; |
1453 | ||
540eed56 TH |
1454 | /** |
1455 | * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way | |
1456 | * @disk: disk to replace part_tbl for | |
1457 | * @new_ptbl: new part_tbl to install | |
1458 | * | |
1459 | * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The | |
1460 | * original ptbl is freed using RCU callback. | |
1461 | * | |
1462 | * LOCKING: | |
6d2cf6f2 | 1463 | * Matching bd_mutex locked or the caller is the only user of @disk. |
540eed56 TH |
1464 | */ |
1465 | static void disk_replace_part_tbl(struct gendisk *disk, | |
1466 | struct disk_part_tbl *new_ptbl) | |
1467 | { | |
6d2cf6f2 BVA |
1468 | struct disk_part_tbl *old_ptbl = |
1469 | rcu_dereference_protected(disk->part_tbl, 1); | |
540eed56 TH |
1470 | |
1471 | rcu_assign_pointer(disk->part_tbl, new_ptbl); | |
a6f23657 JA |
1472 | |
1473 | if (old_ptbl) { | |
1474 | rcu_assign_pointer(old_ptbl->last_lookup, NULL); | |
57bdfbf9 | 1475 | kfree_rcu(old_ptbl, rcu_head); |
a6f23657 | 1476 | } |
540eed56 TH |
1477 | } |
1478 | ||
1479 | /** | |
1480 | * disk_expand_part_tbl - expand disk->part_tbl | |
1481 | * @disk: disk to expand part_tbl for | |
1482 | * @partno: expand such that this partno can fit in | |
1483 | * | |
1484 | * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl | |
1485 | * uses RCU to allow unlocked dereferencing for stats and other stuff. | |
1486 | * | |
1487 | * LOCKING: | |
6d2cf6f2 BVA |
1488 | * Matching bd_mutex locked or the caller is the only user of @disk. |
1489 | * Might sleep. | |
540eed56 TH |
1490 | * |
1491 | * RETURNS: | |
1492 | * 0 on success, -errno on failure. | |
1493 | */ | |
1494 | int disk_expand_part_tbl(struct gendisk *disk, int partno) | |
1495 | { | |
6d2cf6f2 BVA |
1496 | struct disk_part_tbl *old_ptbl = |
1497 | rcu_dereference_protected(disk->part_tbl, 1); | |
540eed56 TH |
1498 | struct disk_part_tbl *new_ptbl; |
1499 | int len = old_ptbl ? old_ptbl->len : 0; | |
5fabcb4c | 1500 | int i, target; |
5fabcb4c JA |
1501 | |
1502 | /* | |
1503 | * check for int overflow, since we can get here from blkpg_ioctl() | |
1504 | * with a user passed 'partno'. | |
1505 | */ | |
1506 | target = partno + 1; | |
1507 | if (target < 0) | |
1508 | return -EINVAL; | |
540eed56 TH |
1509 | |
1510 | /* disk_max_parts() is zero during initialization, ignore if so */ | |
1511 | if (disk_max_parts(disk) && target > disk_max_parts(disk)) | |
1512 | return -EINVAL; | |
1513 | ||
1514 | if (target <= len) | |
1515 | return 0; | |
1516 | ||
78b90a2c GS |
1517 | new_ptbl = kzalloc_node(struct_size(new_ptbl, part, target), GFP_KERNEL, |
1518 | disk->node_id); | |
540eed56 TH |
1519 | if (!new_ptbl) |
1520 | return -ENOMEM; | |
1521 | ||
540eed56 TH |
1522 | new_ptbl->len = target; |
1523 | ||
1524 | for (i = 0; i < len; i++) | |
1525 | rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]); | |
1526 | ||
1527 | disk_replace_part_tbl(disk, new_ptbl); | |
1528 | return 0; | |
1529 | } | |
1530 | ||
edfaa7c3 | 1531 | static void disk_release(struct device *dev) |
1da177e4 | 1532 | { |
edfaa7c3 KS |
1533 | struct gendisk *disk = dev_to_disk(dev); |
1534 | ||
2da78092 | 1535 | blk_free_devt(dev->devt); |
77ea887e | 1536 | disk_release_events(disk); |
1da177e4 | 1537 | kfree(disk->random); |
540eed56 | 1538 | disk_replace_part_tbl(disk, NULL); |
b54e5ed8 | 1539 | hd_free_part(&disk->part0); |
523e1d39 TH |
1540 | if (disk->queue) |
1541 | blk_put_queue(disk->queue); | |
1da177e4 LT |
1542 | kfree(disk); |
1543 | } | |
edfaa7c3 KS |
1544 | struct class block_class = { |
1545 | .name = "block", | |
1da177e4 LT |
1546 | }; |
1547 | ||
3c2670e6 | 1548 | static char *block_devnode(struct device *dev, umode_t *mode, |
4e4098a3 | 1549 | kuid_t *uid, kgid_t *gid) |
b03f38b6 KS |
1550 | { |
1551 | struct gendisk *disk = dev_to_disk(dev); | |
1552 | ||
348e114b CH |
1553 | if (disk->fops->devnode) |
1554 | return disk->fops->devnode(disk, mode); | |
b03f38b6 KS |
1555 | return NULL; |
1556 | } | |
1557 | ||
edf8ff55 | 1558 | static const struct device_type disk_type = { |
edfaa7c3 KS |
1559 | .name = "disk", |
1560 | .groups = disk_attr_groups, | |
1561 | .release = disk_release, | |
e454cea2 | 1562 | .devnode = block_devnode, |
1da177e4 LT |
1563 | }; |
1564 | ||
a6e2ba88 | 1565 | #ifdef CONFIG_PROC_FS |
cf771cb5 TH |
1566 | /* |
1567 | * aggregate disk stat collector. Uses the same stats that the sysfs | |
1568 | * entries do, above, but makes them available through one seq_file. | |
1569 | * | |
1570 | * The output looks suspiciously like /proc/partitions with a bunch of | |
1571 | * extra fields. | |
1572 | */ | |
1573 | static int diskstats_show(struct seq_file *seqf, void *v) | |
1da177e4 LT |
1574 | { |
1575 | struct gendisk *gp = v; | |
e71bf0d0 TH |
1576 | struct disk_part_iter piter; |
1577 | struct hd_struct *hd; | |
1da177e4 | 1578 | char buf[BDEVNAME_SIZE]; |
e016b782 | 1579 | unsigned int inflight; |
ea18e0f0 | 1580 | struct disk_stats stat; |
1da177e4 LT |
1581 | |
1582 | /* | |
ed9e1982 | 1583 | if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next) |
cf771cb5 | 1584 | seq_puts(seqf, "major minor name" |
1da177e4 LT |
1585 | " rio rmerge rsect ruse wio wmerge " |
1586 | "wsect wuse running use aveq" | |
1587 | "\n\n"); | |
1588 | */ | |
9f5e4865 | 1589 | |
71982a40 | 1590 | disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0); |
e71bf0d0 | 1591 | while ((hd = disk_part_iter_next(&piter))) { |
ea18e0f0 | 1592 | part_stat_read_all(hd, &stat); |
e016b782 | 1593 | inflight = part_in_flight(gp->queue, hd); |
ea18e0f0 | 1594 | |
bdca3c87 MC |
1595 | seq_printf(seqf, "%4d %7d %s " |
1596 | "%lu %lu %lu %u " | |
1597 | "%lu %lu %lu %u " | |
1598 | "%u %u %u " | |
b6866318 KK |
1599 | "%lu %lu %lu %u " |
1600 | "%lu %u" | |
1601 | "\n", | |
f331c029 TH |
1602 | MAJOR(part_devt(hd)), MINOR(part_devt(hd)), |
1603 | disk_name(gp, hd->partno, buf), | |
ea18e0f0 KK |
1604 | stat.ios[STAT_READ], |
1605 | stat.merges[STAT_READ], | |
1606 | stat.sectors[STAT_READ], | |
1607 | (unsigned int)div_u64(stat.nsecs[STAT_READ], | |
1608 | NSEC_PER_MSEC), | |
1609 | stat.ios[STAT_WRITE], | |
1610 | stat.merges[STAT_WRITE], | |
1611 | stat.sectors[STAT_WRITE], | |
1612 | (unsigned int)div_u64(stat.nsecs[STAT_WRITE], | |
1613 | NSEC_PER_MSEC), | |
e016b782 | 1614 | inflight, |
ea18e0f0 | 1615 | jiffies_to_msecs(stat.io_ticks), |
8cd5b8fc KK |
1616 | (unsigned int)div_u64(stat.nsecs[STAT_READ] + |
1617 | stat.nsecs[STAT_WRITE] + | |
1618 | stat.nsecs[STAT_DISCARD] + | |
1619 | stat.nsecs[STAT_FLUSH], | |
1620 | NSEC_PER_MSEC), | |
ea18e0f0 KK |
1621 | stat.ios[STAT_DISCARD], |
1622 | stat.merges[STAT_DISCARD], | |
1623 | stat.sectors[STAT_DISCARD], | |
1624 | (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], | |
1625 | NSEC_PER_MSEC), | |
1626 | stat.ios[STAT_FLUSH], | |
1627 | (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], | |
1628 | NSEC_PER_MSEC) | |
28f39d55 | 1629 | ); |
1da177e4 | 1630 | } |
e71bf0d0 | 1631 | disk_part_iter_exit(&piter); |
9f5e4865 | 1632 | |
1da177e4 LT |
1633 | return 0; |
1634 | } | |
1635 | ||
31d85ab2 | 1636 | static const struct seq_operations diskstats_op = { |
def4e38d TH |
1637 | .start = disk_seqf_start, |
1638 | .next = disk_seqf_next, | |
1639 | .stop = disk_seqf_stop, | |
1da177e4 LT |
1640 | .show = diskstats_show |
1641 | }; | |
f500975a AD |
1642 | |
1643 | static int __init proc_genhd_init(void) | |
1644 | { | |
fddda2b7 CH |
1645 | proc_create_seq("diskstats", 0, NULL, &diskstats_op); |
1646 | proc_create_seq("partitions", 0, NULL, &partitions_op); | |
f500975a AD |
1647 | return 0; |
1648 | } | |
1649 | module_init(proc_genhd_init); | |
a6e2ba88 | 1650 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 1651 | |
cf771cb5 | 1652 | dev_t blk_lookup_devt(const char *name, int partno) |
a142be85 | 1653 | { |
def4e38d TH |
1654 | dev_t devt = MKDEV(0, 0); |
1655 | struct class_dev_iter iter; | |
1656 | struct device *dev; | |
a142be85 | 1657 | |
def4e38d TH |
1658 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); |
1659 | while ((dev = class_dev_iter_next(&iter))) { | |
a142be85 | 1660 | struct gendisk *disk = dev_to_disk(dev); |
548b10eb | 1661 | struct hd_struct *part; |
a142be85 | 1662 | |
3ada8b7e | 1663 | if (strcmp(dev_name(dev), name)) |
f331c029 | 1664 | continue; |
f331c029 | 1665 | |
41b8c853 NB |
1666 | if (partno < disk->minors) { |
1667 | /* We need to return the right devno, even | |
1668 | * if the partition doesn't exist yet. | |
1669 | */ | |
1670 | devt = MKDEV(MAJOR(dev->devt), | |
1671 | MINOR(dev->devt) + partno); | |
1672 | break; | |
1673 | } | |
548b10eb | 1674 | part = disk_get_part(disk, partno); |
2bbedcb4 | 1675 | if (part) { |
f331c029 | 1676 | devt = part_devt(part); |
e71bf0d0 | 1677 | disk_put_part(part); |
548b10eb | 1678 | break; |
def4e38d | 1679 | } |
548b10eb | 1680 | disk_put_part(part); |
5c0ef6d0 | 1681 | } |
def4e38d | 1682 | class_dev_iter_exit(&iter); |
edfaa7c3 KS |
1683 | return devt; |
1684 | } | |
edfaa7c3 | 1685 | |
e319e1fb | 1686 | struct gendisk *__alloc_disk_node(int minors, int node_id) |
1946089a CL |
1687 | { |
1688 | struct gendisk *disk; | |
6d2cf6f2 | 1689 | struct disk_part_tbl *ptbl; |
1946089a | 1690 | |
de65b012 CH |
1691 | if (minors > DISK_MAX_PARTS) { |
1692 | printk(KERN_ERR | |
7fb52621 | 1693 | "block: can't allocate more than %d partitions\n", |
de65b012 CH |
1694 | DISK_MAX_PARTS); |
1695 | minors = DISK_MAX_PARTS; | |
1696 | } | |
1946089a | 1697 | |
c1b511eb | 1698 | disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id); |
1da177e4 | 1699 | if (disk) { |
074a7aca | 1700 | if (!init_part_stats(&disk->part0)) { |
1da177e4 LT |
1701 | kfree(disk); |
1702 | return NULL; | |
1703 | } | |
56c0908c | 1704 | init_rwsem(&disk->lookup_sem); |
bf91db18 | 1705 | disk->node_id = node_id; |
540eed56 TH |
1706 | if (disk_expand_part_tbl(disk, 0)) { |
1707 | free_part_stats(&disk->part0); | |
b5d0b9df TH |
1708 | kfree(disk); |
1709 | return NULL; | |
1da177e4 | 1710 | } |
6d2cf6f2 BVA |
1711 | ptbl = rcu_dereference_protected(disk->part_tbl, 1); |
1712 | rcu_assign_pointer(ptbl->part[0], &disk->part0); | |
6c23a968 | 1713 | |
c83f6bf9 VG |
1714 | /* |
1715 | * set_capacity() and get_capacity() currently don't use | |
1716 | * seqcounter to read/update the part0->nr_sects. Still init | |
1717 | * the counter as we can read the sectors in IO submission | |
1718 | * patch using seqence counters. | |
1719 | * | |
1720 | * TODO: Ideally set_capacity() and get_capacity() should be | |
1721 | * converted to make use of bd_mutex and sequence counters. | |
1722 | */ | |
1723 | seqcount_init(&disk->part0.nr_sects_seq); | |
6c71013e ML |
1724 | if (hd_ref_init(&disk->part0)) { |
1725 | hd_free_part(&disk->part0); | |
1726 | kfree(disk); | |
1727 | return NULL; | |
1728 | } | |
b5d0b9df | 1729 | |
1da177e4 | 1730 | disk->minors = minors; |
1da177e4 | 1731 | rand_initialize_disk(disk); |
ed9e1982 TH |
1732 | disk_to_dev(disk)->class = &block_class; |
1733 | disk_to_dev(disk)->type = &disk_type; | |
1734 | device_initialize(disk_to_dev(disk)); | |
1da177e4 LT |
1735 | } |
1736 | return disk; | |
1737 | } | |
e319e1fb | 1738 | EXPORT_SYMBOL(__alloc_disk_node); |
1da177e4 | 1739 | |
3079c22e | 1740 | struct kobject *get_disk_and_module(struct gendisk *disk) |
1da177e4 LT |
1741 | { |
1742 | struct module *owner; | |
1743 | struct kobject *kobj; | |
1744 | ||
1745 | if (!disk->fops) | |
1746 | return NULL; | |
1747 | owner = disk->fops->owner; | |
1748 | if (owner && !try_module_get(owner)) | |
1749 | return NULL; | |
d01b2dcb | 1750 | kobj = kobject_get_unless_zero(&disk_to_dev(disk)->kobj); |
1da177e4 LT |
1751 | if (kobj == NULL) { |
1752 | module_put(owner); | |
1753 | return NULL; | |
1754 | } | |
1755 | return kobj; | |
1756 | ||
1757 | } | |
3079c22e | 1758 | EXPORT_SYMBOL(get_disk_and_module); |
1da177e4 LT |
1759 | |
1760 | void put_disk(struct gendisk *disk) | |
1761 | { | |
1762 | if (disk) | |
ed9e1982 | 1763 | kobject_put(&disk_to_dev(disk)->kobj); |
1da177e4 | 1764 | } |
1da177e4 LT |
1765 | EXPORT_SYMBOL(put_disk); |
1766 | ||
9df6c299 JK |
1767 | /* |
1768 | * This is a counterpart of get_disk_and_module() and thus also of | |
1769 | * get_gendisk(). | |
1770 | */ | |
1771 | void put_disk_and_module(struct gendisk *disk) | |
1772 | { | |
1773 | if (disk) { | |
1774 | struct module *owner = disk->fops->owner; | |
1775 | ||
1776 | put_disk(disk); | |
1777 | module_put(owner); | |
1778 | } | |
1779 | } | |
1780 | EXPORT_SYMBOL(put_disk_and_module); | |
1781 | ||
e3264a4d HR |
1782 | static void set_disk_ro_uevent(struct gendisk *gd, int ro) |
1783 | { | |
1784 | char event[] = "DISK_RO=1"; | |
1785 | char *envp[] = { event, NULL }; | |
1786 | ||
1787 | if (!ro) | |
1788 | event[8] = '0'; | |
1789 | kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); | |
1790 | } | |
1791 | ||
1da177e4 LT |
1792 | void set_device_ro(struct block_device *bdev, int flag) |
1793 | { | |
b7db9956 | 1794 | bdev->bd_part->policy = flag; |
1da177e4 LT |
1795 | } |
1796 | ||
1797 | EXPORT_SYMBOL(set_device_ro); | |
1798 | ||
1799 | void set_disk_ro(struct gendisk *disk, int flag) | |
1800 | { | |
e71bf0d0 TH |
1801 | struct disk_part_iter piter; |
1802 | struct hd_struct *part; | |
1803 | ||
e3264a4d HR |
1804 | if (disk->part0.policy != flag) { |
1805 | set_disk_ro_uevent(disk, flag); | |
1806 | disk->part0.policy = flag; | |
1807 | } | |
1808 | ||
1809 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY); | |
e71bf0d0 TH |
1810 | while ((part = disk_part_iter_next(&piter))) |
1811 | part->policy = flag; | |
1812 | disk_part_iter_exit(&piter); | |
1da177e4 LT |
1813 | } |
1814 | ||
1815 | EXPORT_SYMBOL(set_disk_ro); | |
1816 | ||
1817 | int bdev_read_only(struct block_device *bdev) | |
1818 | { | |
1819 | if (!bdev) | |
1820 | return 0; | |
b7db9956 | 1821 | return bdev->bd_part->policy; |
1da177e4 LT |
1822 | } |
1823 | ||
1824 | EXPORT_SYMBOL(bdev_read_only); | |
1825 | ||
77ea887e TH |
1826 | /* |
1827 | * Disk events - monitor disk events like media change and eject request. | |
1828 | */ | |
1829 | struct disk_events { | |
1830 | struct list_head node; /* all disk_event's */ | |
1831 | struct gendisk *disk; /* the associated disk */ | |
1832 | spinlock_t lock; | |
1833 | ||
fdd514e1 | 1834 | struct mutex block_mutex; /* protects blocking */ |
77ea887e TH |
1835 | int block; /* event blocking depth */ |
1836 | unsigned int pending; /* events already sent out */ | |
1837 | unsigned int clearing; /* events being cleared */ | |
1838 | ||
1839 | long poll_msecs; /* interval, -1 for default */ | |
1840 | struct delayed_work dwork; | |
1841 | }; | |
1842 | ||
1843 | static const char *disk_events_strs[] = { | |
1844 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change", | |
1845 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request", | |
1846 | }; | |
1847 | ||
1848 | static char *disk_uevents[] = { | |
1849 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1", | |
1850 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1", | |
1851 | }; | |
1852 | ||
1853 | /* list of all disk_events */ | |
1854 | static DEFINE_MUTEX(disk_events_mutex); | |
1855 | static LIST_HEAD(disk_events); | |
1856 | ||
1857 | /* disable in-kernel polling by default */ | |
1fe8f348 | 1858 | static unsigned long disk_events_dfl_poll_msecs; |
77ea887e TH |
1859 | |
1860 | static unsigned long disk_events_poll_jiffies(struct gendisk *disk) | |
1861 | { | |
1862 | struct disk_events *ev = disk->ev; | |
1863 | long intv_msecs = 0; | |
1864 | ||
1865 | /* | |
1866 | * If device-specific poll interval is set, always use it. If | |
673387a9 | 1867 | * the default is being used, poll if the POLL flag is set. |
77ea887e TH |
1868 | */ |
1869 | if (ev->poll_msecs >= 0) | |
1870 | intv_msecs = ev->poll_msecs; | |
c92e2f04 | 1871 | else if (disk->event_flags & DISK_EVENT_FLAG_POLL) |
77ea887e TH |
1872 | intv_msecs = disk_events_dfl_poll_msecs; |
1873 | ||
1874 | return msecs_to_jiffies(intv_msecs); | |
1875 | } | |
1876 | ||
c3af54af TH |
1877 | /** |
1878 | * disk_block_events - block and flush disk event checking | |
1879 | * @disk: disk to block events for | |
1880 | * | |
1881 | * On return from this function, it is guaranteed that event checking | |
1882 | * isn't in progress and won't happen until unblocked by | |
1883 | * disk_unblock_events(). Events blocking is counted and the actual | |
1884 | * unblocking happens after the matching number of unblocks are done. | |
1885 | * | |
1886 | * Note that this intentionally does not block event checking from | |
1887 | * disk_clear_events(). | |
1888 | * | |
1889 | * CONTEXT: | |
1890 | * Might sleep. | |
1891 | */ | |
1892 | void disk_block_events(struct gendisk *disk) | |
77ea887e TH |
1893 | { |
1894 | struct disk_events *ev = disk->ev; | |
1895 | unsigned long flags; | |
1896 | bool cancel; | |
1897 | ||
c3af54af TH |
1898 | if (!ev) |
1899 | return; | |
1900 | ||
fdd514e1 TH |
1901 | /* |
1902 | * Outer mutex ensures that the first blocker completes canceling | |
1903 | * the event work before further blockers are allowed to finish. | |
1904 | */ | |
1905 | mutex_lock(&ev->block_mutex); | |
1906 | ||
77ea887e TH |
1907 | spin_lock_irqsave(&ev->lock, flags); |
1908 | cancel = !ev->block++; | |
1909 | spin_unlock_irqrestore(&ev->lock, flags); | |
1910 | ||
c3af54af TH |
1911 | if (cancel) |
1912 | cancel_delayed_work_sync(&disk->ev->dwork); | |
fdd514e1 TH |
1913 | |
1914 | mutex_unlock(&ev->block_mutex); | |
77ea887e TH |
1915 | } |
1916 | ||
1917 | static void __disk_unblock_events(struct gendisk *disk, bool check_now) | |
1918 | { | |
1919 | struct disk_events *ev = disk->ev; | |
1920 | unsigned long intv; | |
1921 | unsigned long flags; | |
1922 | ||
1923 | spin_lock_irqsave(&ev->lock, flags); | |
1924 | ||
1925 | if (WARN_ON_ONCE(ev->block <= 0)) | |
1926 | goto out_unlock; | |
1927 | ||
1928 | if (--ev->block) | |
1929 | goto out_unlock; | |
1930 | ||
77ea887e | 1931 | intv = disk_events_poll_jiffies(disk); |
77ea887e | 1932 | if (check_now) |
695588f9 VK |
1933 | queue_delayed_work(system_freezable_power_efficient_wq, |
1934 | &ev->dwork, 0); | |
77ea887e | 1935 | else if (intv) |
695588f9 VK |
1936 | queue_delayed_work(system_freezable_power_efficient_wq, |
1937 | &ev->dwork, intv); | |
77ea887e TH |
1938 | out_unlock: |
1939 | spin_unlock_irqrestore(&ev->lock, flags); | |
1940 | } | |
1941 | ||
77ea887e TH |
1942 | /** |
1943 | * disk_unblock_events - unblock disk event checking | |
1944 | * @disk: disk to unblock events for | |
1945 | * | |
1946 | * Undo disk_block_events(). When the block count reaches zero, it | |
1947 | * starts events polling if configured. | |
1948 | * | |
1949 | * CONTEXT: | |
1950 | * Don't care. Safe to call from irq context. | |
1951 | */ | |
1952 | void disk_unblock_events(struct gendisk *disk) | |
1953 | { | |
1954 | if (disk->ev) | |
facc31dd | 1955 | __disk_unblock_events(disk, false); |
77ea887e TH |
1956 | } |
1957 | ||
1958 | /** | |
85ef06d1 TH |
1959 | * disk_flush_events - schedule immediate event checking and flushing |
1960 | * @disk: disk to check and flush events for | |
1961 | * @mask: events to flush | |
77ea887e | 1962 | * |
85ef06d1 TH |
1963 | * Schedule immediate event checking on @disk if not blocked. Events in |
1964 | * @mask are scheduled to be cleared from the driver. Note that this | |
1965 | * doesn't clear the events from @disk->ev. | |
77ea887e TH |
1966 | * |
1967 | * CONTEXT: | |
85ef06d1 | 1968 | * If @mask is non-zero must be called with bdev->bd_mutex held. |
77ea887e | 1969 | */ |
85ef06d1 | 1970 | void disk_flush_events(struct gendisk *disk, unsigned int mask) |
77ea887e | 1971 | { |
a9dce2a3 | 1972 | struct disk_events *ev = disk->ev; |
a9dce2a3 TH |
1973 | |
1974 | if (!ev) | |
1975 | return; | |
1976 | ||
85ef06d1 TH |
1977 | spin_lock_irq(&ev->lock); |
1978 | ev->clearing |= mask; | |
41f63c53 | 1979 | if (!ev->block) |
695588f9 VK |
1980 | mod_delayed_work(system_freezable_power_efficient_wq, |
1981 | &ev->dwork, 0); | |
85ef06d1 | 1982 | spin_unlock_irq(&ev->lock); |
77ea887e | 1983 | } |
77ea887e TH |
1984 | |
1985 | /** | |
1986 | * disk_clear_events - synchronously check, clear and return pending events | |
1987 | * @disk: disk to fetch and clear events from | |
da3dae54 | 1988 | * @mask: mask of events to be fetched and cleared |
77ea887e TH |
1989 | * |
1990 | * Disk events are synchronously checked and pending events in @mask | |
1991 | * are cleared and returned. This ignores the block count. | |
1992 | * | |
1993 | * CONTEXT: | |
1994 | * Might sleep. | |
1995 | */ | |
1996 | unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask) | |
1997 | { | |
1998 | const struct block_device_operations *bdops = disk->fops; | |
1999 | struct disk_events *ev = disk->ev; | |
2000 | unsigned int pending; | |
12c2bdb2 | 2001 | unsigned int clearing = mask; |
77ea887e TH |
2002 | |
2003 | if (!ev) { | |
2004 | /* for drivers still using the old ->media_changed method */ | |
2005 | if ((mask & DISK_EVENT_MEDIA_CHANGE) && | |
2006 | bdops->media_changed && bdops->media_changed(disk)) | |
2007 | return DISK_EVENT_MEDIA_CHANGE; | |
2008 | return 0; | |
2009 | } | |
2010 | ||
12c2bdb2 DB |
2011 | disk_block_events(disk); |
2012 | ||
2013 | /* | |
2014 | * store the union of mask and ev->clearing on the stack so that the | |
2015 | * race with disk_flush_events does not cause ambiguity (ev->clearing | |
2016 | * can still be modified even if events are blocked). | |
2017 | */ | |
77ea887e | 2018 | spin_lock_irq(&ev->lock); |
12c2bdb2 DB |
2019 | clearing |= ev->clearing; |
2020 | ev->clearing = 0; | |
77ea887e TH |
2021 | spin_unlock_irq(&ev->lock); |
2022 | ||
12c2bdb2 | 2023 | disk_check_events(ev, &clearing); |
aea24a8b | 2024 | /* |
12c2bdb2 DB |
2025 | * if ev->clearing is not 0, the disk_flush_events got called in the |
2026 | * middle of this function, so we want to run the workfn without delay. | |
aea24a8b | 2027 | */ |
12c2bdb2 | 2028 | __disk_unblock_events(disk, ev->clearing ? true : false); |
77ea887e TH |
2029 | |
2030 | /* then, fetch and clear pending events */ | |
2031 | spin_lock_irq(&ev->lock); | |
77ea887e TH |
2032 | pending = ev->pending & mask; |
2033 | ev->pending &= ~mask; | |
2034 | spin_unlock_irq(&ev->lock); | |
12c2bdb2 | 2035 | WARN_ON_ONCE(clearing & mask); |
77ea887e TH |
2036 | |
2037 | return pending; | |
2038 | } | |
2039 | ||
12c2bdb2 DB |
2040 | /* |
2041 | * Separate this part out so that a different pointer for clearing_ptr can be | |
2042 | * passed in for disk_clear_events. | |
2043 | */ | |
77ea887e TH |
2044 | static void disk_events_workfn(struct work_struct *work) |
2045 | { | |
2046 | struct delayed_work *dwork = to_delayed_work(work); | |
2047 | struct disk_events *ev = container_of(dwork, struct disk_events, dwork); | |
12c2bdb2 DB |
2048 | |
2049 | disk_check_events(ev, &ev->clearing); | |
2050 | } | |
2051 | ||
2052 | static void disk_check_events(struct disk_events *ev, | |
2053 | unsigned int *clearing_ptr) | |
2054 | { | |
77ea887e TH |
2055 | struct gendisk *disk = ev->disk; |
2056 | char *envp[ARRAY_SIZE(disk_uevents) + 1] = { }; | |
12c2bdb2 | 2057 | unsigned int clearing = *clearing_ptr; |
77ea887e TH |
2058 | unsigned int events; |
2059 | unsigned long intv; | |
2060 | int nr_events = 0, i; | |
2061 | ||
2062 | /* check events */ | |
2063 | events = disk->fops->check_events(disk, clearing); | |
2064 | ||
2065 | /* accumulate pending events and schedule next poll if necessary */ | |
2066 | spin_lock_irq(&ev->lock); | |
2067 | ||
2068 | events &= ~ev->pending; | |
2069 | ev->pending |= events; | |
12c2bdb2 | 2070 | *clearing_ptr &= ~clearing; |
77ea887e TH |
2071 | |
2072 | intv = disk_events_poll_jiffies(disk); | |
2073 | if (!ev->block && intv) | |
695588f9 VK |
2074 | queue_delayed_work(system_freezable_power_efficient_wq, |
2075 | &ev->dwork, intv); | |
77ea887e TH |
2076 | |
2077 | spin_unlock_irq(&ev->lock); | |
2078 | ||
7c88a168 TH |
2079 | /* |
2080 | * Tell userland about new events. Only the events listed in | |
c92e2f04 MW |
2081 | * @disk->events are reported, and only if DISK_EVENT_FLAG_UEVENT |
2082 | * is set. Otherwise, events are processed internally but never | |
2083 | * get reported to userland. | |
7c88a168 | 2084 | */ |
77ea887e | 2085 | for (i = 0; i < ARRAY_SIZE(disk_uevents); i++) |
c92e2f04 MW |
2086 | if ((events & disk->events & (1 << i)) && |
2087 | (disk->event_flags & DISK_EVENT_FLAG_UEVENT)) | |
77ea887e TH |
2088 | envp[nr_events++] = disk_uevents[i]; |
2089 | ||
2090 | if (nr_events) | |
2091 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
2092 | } | |
2093 | ||
2094 | /* | |
2095 | * A disk events enabled device has the following sysfs nodes under | |
2096 | * its /sys/block/X/ directory. | |
2097 | * | |
2098 | * events : list of all supported events | |
2099 | * events_async : list of events which can be detected w/o polling | |
673387a9 | 2100 | * (always empty, only for backwards compatibility) |
77ea887e TH |
2101 | * events_poll_msecs : polling interval, 0: disable, -1: system default |
2102 | */ | |
2103 | static ssize_t __disk_events_show(unsigned int events, char *buf) | |
2104 | { | |
2105 | const char *delim = ""; | |
2106 | ssize_t pos = 0; | |
2107 | int i; | |
2108 | ||
2109 | for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++) | |
2110 | if (events & (1 << i)) { | |
2111 | pos += sprintf(buf + pos, "%s%s", | |
2112 | delim, disk_events_strs[i]); | |
2113 | delim = " "; | |
2114 | } | |
2115 | if (pos) | |
2116 | pos += sprintf(buf + pos, "\n"); | |
2117 | return pos; | |
2118 | } | |
2119 | ||
2120 | static ssize_t disk_events_show(struct device *dev, | |
2121 | struct device_attribute *attr, char *buf) | |
2122 | { | |
2123 | struct gendisk *disk = dev_to_disk(dev); | |
2124 | ||
c92e2f04 MW |
2125 | if (!(disk->event_flags & DISK_EVENT_FLAG_UEVENT)) |
2126 | return 0; | |
2127 | ||
77ea887e TH |
2128 | return __disk_events_show(disk->events, buf); |
2129 | } | |
2130 | ||
2131 | static ssize_t disk_events_async_show(struct device *dev, | |
2132 | struct device_attribute *attr, char *buf) | |
2133 | { | |
673387a9 | 2134 | return 0; |
77ea887e TH |
2135 | } |
2136 | ||
2137 | static ssize_t disk_events_poll_msecs_show(struct device *dev, | |
2138 | struct device_attribute *attr, | |
2139 | char *buf) | |
2140 | { | |
2141 | struct gendisk *disk = dev_to_disk(dev); | |
2142 | ||
cdf3e3de MW |
2143 | if (!disk->ev) |
2144 | return sprintf(buf, "-1\n"); | |
2145 | ||
77ea887e TH |
2146 | return sprintf(buf, "%ld\n", disk->ev->poll_msecs); |
2147 | } | |
2148 | ||
2149 | static ssize_t disk_events_poll_msecs_store(struct device *dev, | |
2150 | struct device_attribute *attr, | |
2151 | const char *buf, size_t count) | |
2152 | { | |
2153 | struct gendisk *disk = dev_to_disk(dev); | |
2154 | long intv; | |
2155 | ||
2156 | if (!count || !sscanf(buf, "%ld", &intv)) | |
2157 | return -EINVAL; | |
2158 | ||
2159 | if (intv < 0 && intv != -1) | |
2160 | return -EINVAL; | |
2161 | ||
cdf3e3de MW |
2162 | if (!disk->ev) |
2163 | return -ENODEV; | |
2164 | ||
c3af54af | 2165 | disk_block_events(disk); |
77ea887e TH |
2166 | disk->ev->poll_msecs = intv; |
2167 | __disk_unblock_events(disk, true); | |
2168 | ||
2169 | return count; | |
2170 | } | |
2171 | ||
5657a819 JP |
2172 | static const DEVICE_ATTR(events, 0444, disk_events_show, NULL); |
2173 | static const DEVICE_ATTR(events_async, 0444, disk_events_async_show, NULL); | |
2174 | static const DEVICE_ATTR(events_poll_msecs, 0644, | |
77ea887e TH |
2175 | disk_events_poll_msecs_show, |
2176 | disk_events_poll_msecs_store); | |
2177 | ||
2178 | static const struct attribute *disk_events_attrs[] = { | |
2179 | &dev_attr_events.attr, | |
2180 | &dev_attr_events_async.attr, | |
2181 | &dev_attr_events_poll_msecs.attr, | |
2182 | NULL, | |
2183 | }; | |
2184 | ||
2185 | /* | |
2186 | * The default polling interval can be specified by the kernel | |
2187 | * parameter block.events_dfl_poll_msecs which defaults to 0 | |
2188 | * (disable). This can also be modified runtime by writing to | |
1624b0b2 | 2189 | * /sys/module/block/parameters/events_dfl_poll_msecs. |
77ea887e TH |
2190 | */ |
2191 | static int disk_events_set_dfl_poll_msecs(const char *val, | |
2192 | const struct kernel_param *kp) | |
2193 | { | |
2194 | struct disk_events *ev; | |
2195 | int ret; | |
2196 | ||
2197 | ret = param_set_ulong(val, kp); | |
2198 | if (ret < 0) | |
2199 | return ret; | |
2200 | ||
2201 | mutex_lock(&disk_events_mutex); | |
2202 | ||
2203 | list_for_each_entry(ev, &disk_events, node) | |
85ef06d1 | 2204 | disk_flush_events(ev->disk, 0); |
77ea887e TH |
2205 | |
2206 | mutex_unlock(&disk_events_mutex); | |
2207 | ||
2208 | return 0; | |
2209 | } | |
2210 | ||
2211 | static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = { | |
2212 | .set = disk_events_set_dfl_poll_msecs, | |
2213 | .get = param_get_ulong, | |
2214 | }; | |
2215 | ||
2216 | #undef MODULE_PARAM_PREFIX | |
2217 | #define MODULE_PARAM_PREFIX "block." | |
2218 | ||
2219 | module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops, | |
2220 | &disk_events_dfl_poll_msecs, 0644); | |
2221 | ||
2222 | /* | |
9f53d2fe | 2223 | * disk_{alloc|add|del|release}_events - initialize and destroy disk_events. |
77ea887e | 2224 | */ |
9f53d2fe | 2225 | static void disk_alloc_events(struct gendisk *disk) |
77ea887e TH |
2226 | { |
2227 | struct disk_events *ev; | |
2228 | ||
cdf3e3de | 2229 | if (!disk->fops->check_events || !disk->events) |
77ea887e TH |
2230 | return; |
2231 | ||
2232 | ev = kzalloc(sizeof(*ev), GFP_KERNEL); | |
2233 | if (!ev) { | |
2234 | pr_warn("%s: failed to initialize events\n", disk->disk_name); | |
2235 | return; | |
2236 | } | |
2237 | ||
77ea887e TH |
2238 | INIT_LIST_HEAD(&ev->node); |
2239 | ev->disk = disk; | |
2240 | spin_lock_init(&ev->lock); | |
fdd514e1 | 2241 | mutex_init(&ev->block_mutex); |
77ea887e TH |
2242 | ev->block = 1; |
2243 | ev->poll_msecs = -1; | |
2244 | INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn); | |
2245 | ||
9f53d2fe SG |
2246 | disk->ev = ev; |
2247 | } | |
2248 | ||
2249 | static void disk_add_events(struct gendisk *disk) | |
2250 | { | |
9f53d2fe SG |
2251 | /* FIXME: error handling */ |
2252 | if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0) | |
2253 | pr_warn("%s: failed to create sysfs files for events\n", | |
2254 | disk->disk_name); | |
2255 | ||
cdf3e3de MW |
2256 | if (!disk->ev) |
2257 | return; | |
2258 | ||
77ea887e | 2259 | mutex_lock(&disk_events_mutex); |
9f53d2fe | 2260 | list_add_tail(&disk->ev->node, &disk_events); |
77ea887e TH |
2261 | mutex_unlock(&disk_events_mutex); |
2262 | ||
2263 | /* | |
2264 | * Block count is initialized to 1 and the following initial | |
2265 | * unblock kicks it into action. | |
2266 | */ | |
2267 | __disk_unblock_events(disk, true); | |
2268 | } | |
2269 | ||
2270 | static void disk_del_events(struct gendisk *disk) | |
2271 | { | |
cdf3e3de MW |
2272 | if (disk->ev) { |
2273 | disk_block_events(disk); | |
77ea887e | 2274 | |
cdf3e3de MW |
2275 | mutex_lock(&disk_events_mutex); |
2276 | list_del_init(&disk->ev->node); | |
2277 | mutex_unlock(&disk_events_mutex); | |
2278 | } | |
77ea887e TH |
2279 | |
2280 | sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs); | |
2281 | } | |
2282 | ||
2283 | static void disk_release_events(struct gendisk *disk) | |
2284 | { | |
2285 | /* the block count should be 1 from disk_del_events() */ | |
2286 | WARN_ON_ONCE(disk->ev && disk->ev->block != 1); | |
2287 | kfree(disk->ev); | |
2288 | } |