]>
Commit | Line | Data |
---|---|---|
3dcf60bc | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * gendisk handling | |
7b51e703 CH |
4 | * |
5 | * Portions Copyright (C) 2020 Christoph Hellwig | |
1da177e4 LT |
6 | */ |
7 | ||
1da177e4 | 8 | #include <linux/module.h> |
3ad5cee5 | 9 | #include <linux/ctype.h> |
1da177e4 LT |
10 | #include <linux/fs.h> |
11 | #include <linux/genhd.h> | |
b446b60e | 12 | #include <linux/kdev_t.h> |
1da177e4 LT |
13 | #include <linux/kernel.h> |
14 | #include <linux/blkdev.h> | |
66114cad | 15 | #include <linux/backing-dev.h> |
1da177e4 LT |
16 | #include <linux/init.h> |
17 | #include <linux/spinlock.h> | |
f500975a | 18 | #include <linux/proc_fs.h> |
1da177e4 LT |
19 | #include <linux/seq_file.h> |
20 | #include <linux/slab.h> | |
21 | #include <linux/kmod.h> | |
58383af6 | 22 | #include <linux/mutex.h> |
bcce3de1 | 23 | #include <linux/idr.h> |
77ea887e | 24 | #include <linux/log2.h> |
25e823c8 | 25 | #include <linux/pm_runtime.h> |
99e6608c | 26 | #include <linux/badblocks.h> |
1da177e4 | 27 | |
ff88972c AB |
28 | #include "blk.h" |
29 | ||
31eb6186 | 30 | static struct kobject *block_depr; |
1da177e4 | 31 | |
22ae8ce8 | 32 | DECLARE_RWSEM(bdev_lookup_sem); |
62b508f8 | 33 | |
bcce3de1 | 34 | /* for extended dynamic devt allocation, currently only one major is used */ |
ce23bba8 | 35 | #define NR_EXT_DEVT (1 << MINORBITS) |
22ae8ce8 | 36 | static DEFINE_IDA(ext_devt_ida); |
bcce3de1 | 37 | |
12c2bdb2 DB |
38 | static void disk_check_events(struct disk_events *ev, |
39 | unsigned int *clearing_ptr); | |
9f53d2fe | 40 | static void disk_alloc_events(struct gendisk *disk); |
77ea887e TH |
41 | static void disk_add_events(struct gendisk *disk); |
42 | static void disk_del_events(struct gendisk *disk); | |
43 | static void disk_release_events(struct gendisk *disk); | |
44 | ||
a782483c CH |
45 | void set_capacity(struct gendisk *disk, sector_t sectors) |
46 | { | |
cb8432d6 | 47 | struct block_device *bdev = disk->part0; |
a782483c CH |
48 | |
49 | spin_lock(&bdev->bd_size_lock); | |
50 | i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT); | |
51 | spin_unlock(&bdev->bd_size_lock); | |
52 | } | |
53 | EXPORT_SYMBOL(set_capacity); | |
54 | ||
e598a72f | 55 | /* |
449f4ec9 CH |
56 | * Set disk capacity and notify if the size is not currently zero and will not |
57 | * be set to zero. Returns true if a uevent was sent, otherwise false. | |
e598a72f | 58 | */ |
449f4ec9 | 59 | bool set_capacity_and_notify(struct gendisk *disk, sector_t size) |
e598a72f BS |
60 | { |
61 | sector_t capacity = get_capacity(disk); | |
a782483c | 62 | char *envp[] = { "RESIZE=1", NULL }; |
e598a72f BS |
63 | |
64 | set_capacity(disk, size); | |
e598a72f | 65 | |
a782483c CH |
66 | /* |
67 | * Only print a message and send a uevent if the gendisk is user visible | |
68 | * and alive. This avoids spamming the log and udev when setting the | |
69 | * initial capacity during probing. | |
70 | */ | |
71 | if (size == capacity || | |
72 | (disk->flags & (GENHD_FL_UP | GENHD_FL_HIDDEN)) != GENHD_FL_UP) | |
73 | return false; | |
e598a72f | 74 | |
a782483c CH |
75 | pr_info("%s: detected capacity change from %lld to %lld\n", |
76 | disk->disk_name, size, capacity); | |
7e890c37 | 77 | |
a782483c CH |
78 | /* |
79 | * Historically we did not send a uevent for changes to/from an empty | |
80 | * device. | |
81 | */ | |
82 | if (!capacity || !size) | |
83 | return false; | |
84 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
85 | return true; | |
e598a72f | 86 | } |
449f4ec9 | 87 | EXPORT_SYMBOL_GPL(set_capacity_and_notify); |
e598a72f | 88 | |
5cbd28e3 CH |
89 | /* |
90 | * Format the device name of the indicated disk into the supplied buffer and | |
91 | * return a pointer to that same buffer for convenience. | |
92 | */ | |
93 | char *disk_name(struct gendisk *hd, int partno, char *buf) | |
94 | { | |
95 | if (!partno) | |
96 | snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name); | |
97 | else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1])) | |
98 | snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno); | |
99 | else | |
100 | snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno); | |
101 | ||
102 | return buf; | |
103 | } | |
104 | ||
105 | const char *bdevname(struct block_device *bdev, char *buf) | |
106 | { | |
8a63a86e | 107 | return disk_name(bdev->bd_disk, bdev->bd_partno, buf); |
5cbd28e3 CH |
108 | } |
109 | EXPORT_SYMBOL(bdevname); | |
e598a72f | 110 | |
0d02129e CH |
111 | static void part_stat_read_all(struct block_device *part, |
112 | struct disk_stats *stat) | |
ea18e0f0 KK |
113 | { |
114 | int cpu; | |
115 | ||
116 | memset(stat, 0, sizeof(struct disk_stats)); | |
117 | for_each_possible_cpu(cpu) { | |
0d02129e | 118 | struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu); |
ea18e0f0 KK |
119 | int group; |
120 | ||
121 | for (group = 0; group < NR_STAT_GROUPS; group++) { | |
122 | stat->nsecs[group] += ptr->nsecs[group]; | |
123 | stat->sectors[group] += ptr->sectors[group]; | |
124 | stat->ios[group] += ptr->ios[group]; | |
125 | stat->merges[group] += ptr->merges[group]; | |
126 | } | |
127 | ||
128 | stat->io_ticks += ptr->io_ticks; | |
ea18e0f0 KK |
129 | } |
130 | } | |
ea18e0f0 | 131 | |
8446fe92 | 132 | static unsigned int part_in_flight(struct block_device *part) |
f299b7c7 | 133 | { |
b2f609e1 | 134 | unsigned int inflight = 0; |
1226b8dd | 135 | int cpu; |
f299b7c7 | 136 | |
1226b8dd | 137 | for_each_possible_cpu(cpu) { |
e016b782 MP |
138 | inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) + |
139 | part_stat_local_read_cpu(part, in_flight[1], cpu); | |
1226b8dd | 140 | } |
e016b782 MP |
141 | if ((int)inflight < 0) |
142 | inflight = 0; | |
1226b8dd | 143 | |
e016b782 | 144 | return inflight; |
f299b7c7 JA |
145 | } |
146 | ||
8446fe92 CH |
147 | static void part_in_flight_rw(struct block_device *part, |
148 | unsigned int inflight[2]) | |
bf0ddaba | 149 | { |
1226b8dd MP |
150 | int cpu; |
151 | ||
1226b8dd MP |
152 | inflight[0] = 0; |
153 | inflight[1] = 0; | |
154 | for_each_possible_cpu(cpu) { | |
155 | inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu); | |
156 | inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu); | |
157 | } | |
158 | if ((int)inflight[0] < 0) | |
159 | inflight[0] = 0; | |
160 | if ((int)inflight[1] < 0) | |
161 | inflight[1] = 0; | |
bf0ddaba OS |
162 | } |
163 | ||
8446fe92 | 164 | struct block_device *__disk_get_part(struct gendisk *disk, int partno) |
807d4af2 CH |
165 | { |
166 | struct disk_part_tbl *ptbl = rcu_dereference(disk->part_tbl); | |
167 | ||
168 | if (unlikely(partno < 0 || partno >= ptbl->len)) | |
169 | return NULL; | |
170 | return rcu_dereference(ptbl->part[partno]); | |
171 | } | |
172 | ||
e71bf0d0 TH |
173 | /** |
174 | * disk_part_iter_init - initialize partition iterator | |
175 | * @piter: iterator to initialize | |
176 | * @disk: disk to iterate over | |
177 | * @flags: DISK_PITER_* flags | |
178 | * | |
179 | * Initialize @piter so that it iterates over partitions of @disk. | |
180 | * | |
181 | * CONTEXT: | |
182 | * Don't care. | |
183 | */ | |
184 | void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk, | |
185 | unsigned int flags) | |
186 | { | |
540eed56 TH |
187 | struct disk_part_tbl *ptbl; |
188 | ||
189 | rcu_read_lock(); | |
190 | ptbl = rcu_dereference(disk->part_tbl); | |
191 | ||
e71bf0d0 TH |
192 | piter->disk = disk; |
193 | piter->part = NULL; | |
194 | ||
195 | if (flags & DISK_PITER_REVERSE) | |
540eed56 | 196 | piter->idx = ptbl->len - 1; |
71982a40 | 197 | else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0)) |
e71bf0d0 | 198 | piter->idx = 0; |
b5d0b9df TH |
199 | else |
200 | piter->idx = 1; | |
e71bf0d0 TH |
201 | |
202 | piter->flags = flags; | |
540eed56 TH |
203 | |
204 | rcu_read_unlock(); | |
e71bf0d0 TH |
205 | } |
206 | EXPORT_SYMBOL_GPL(disk_part_iter_init); | |
207 | ||
208 | /** | |
209 | * disk_part_iter_next - proceed iterator to the next partition and return it | |
210 | * @piter: iterator of interest | |
211 | * | |
212 | * Proceed @piter to the next partition and return it. | |
213 | * | |
214 | * CONTEXT: | |
215 | * Don't care. | |
216 | */ | |
ad1eaa53 | 217 | struct block_device *disk_part_iter_next(struct disk_part_iter *piter) |
e71bf0d0 | 218 | { |
540eed56 | 219 | struct disk_part_tbl *ptbl; |
e71bf0d0 TH |
220 | int inc, end; |
221 | ||
222 | /* put the last partition */ | |
e79319af | 223 | disk_part_iter_exit(piter); |
e71bf0d0 | 224 | |
540eed56 | 225 | /* get part_tbl */ |
e71bf0d0 | 226 | rcu_read_lock(); |
540eed56 | 227 | ptbl = rcu_dereference(piter->disk->part_tbl); |
e71bf0d0 TH |
228 | |
229 | /* determine iteration parameters */ | |
230 | if (piter->flags & DISK_PITER_REVERSE) { | |
231 | inc = -1; | |
71982a40 TH |
232 | if (piter->flags & (DISK_PITER_INCL_PART0 | |
233 | DISK_PITER_INCL_EMPTY_PART0)) | |
b5d0b9df TH |
234 | end = -1; |
235 | else | |
236 | end = 0; | |
e71bf0d0 TH |
237 | } else { |
238 | inc = 1; | |
540eed56 | 239 | end = ptbl->len; |
e71bf0d0 TH |
240 | } |
241 | ||
242 | /* iterate to the next partition */ | |
243 | for (; piter->idx != end; piter->idx += inc) { | |
8446fe92 | 244 | struct block_device *part; |
e71bf0d0 | 245 | |
540eed56 | 246 | part = rcu_dereference(ptbl->part[piter->idx]); |
e71bf0d0 TH |
247 | if (!part) |
248 | continue; | |
8446fe92 | 249 | if (!bdev_nr_sectors(part) && |
71982a40 TH |
250 | !(piter->flags & DISK_PITER_INCL_EMPTY) && |
251 | !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 && | |
252 | piter->idx == 0)) | |
e71bf0d0 TH |
253 | continue; |
254 | ||
ad1eaa53 CH |
255 | piter->part = bdgrab(part); |
256 | if (!piter->part) | |
257 | continue; | |
e71bf0d0 TH |
258 | piter->idx += inc; |
259 | break; | |
260 | } | |
261 | ||
262 | rcu_read_unlock(); | |
263 | ||
264 | return piter->part; | |
265 | } | |
266 | EXPORT_SYMBOL_GPL(disk_part_iter_next); | |
267 | ||
268 | /** | |
269 | * disk_part_iter_exit - finish up partition iteration | |
270 | * @piter: iter of interest | |
271 | * | |
272 | * Called when iteration is over. Cleans up @piter. | |
273 | * | |
274 | * CONTEXT: | |
275 | * Don't care. | |
276 | */ | |
277 | void disk_part_iter_exit(struct disk_part_iter *piter) | |
278 | { | |
ad1eaa53 CH |
279 | if (piter->part) |
280 | bdput(piter->part); | |
e71bf0d0 TH |
281 | piter->part = NULL; |
282 | } | |
283 | EXPORT_SYMBOL_GPL(disk_part_iter_exit); | |
284 | ||
8446fe92 | 285 | static inline int sector_in_part(struct block_device *part, sector_t sector) |
a6f23657 | 286 | { |
8446fe92 CH |
287 | return part->bd_start_sect <= sector && |
288 | sector < part->bd_start_sect + bdev_nr_sectors(part); | |
a6f23657 JA |
289 | } |
290 | ||
e71bf0d0 TH |
291 | /** |
292 | * disk_map_sector_rcu - map sector to partition | |
293 | * @disk: gendisk of interest | |
294 | * @sector: sector to map | |
295 | * | |
296 | * Find out which partition @sector maps to on @disk. This is | |
297 | * primarily used for stats accounting. | |
298 | * | |
299 | * CONTEXT: | |
cb8432d6 | 300 | * RCU read locked. |
e71bf0d0 TH |
301 | * |
302 | * RETURNS: | |
074a7aca | 303 | * Found partition on success, part0 is returned if no partition matches |
b7d6c303 | 304 | * or the matched partition is being deleted. |
e71bf0d0 | 305 | */ |
8446fe92 | 306 | struct block_device *disk_map_sector_rcu(struct gendisk *disk, sector_t sector) |
e71bf0d0 | 307 | { |
540eed56 | 308 | struct disk_part_tbl *ptbl; |
8446fe92 | 309 | struct block_device *part; |
e71bf0d0 TH |
310 | int i; |
311 | ||
8ab1d40a | 312 | rcu_read_lock(); |
540eed56 TH |
313 | ptbl = rcu_dereference(disk->part_tbl); |
314 | ||
a6f23657 | 315 | part = rcu_dereference(ptbl->last_lookup); |
cb8432d6 | 316 | if (part && sector_in_part(part, sector)) |
8ab1d40a | 317 | goto out_unlock; |
a6f23657 | 318 | |
540eed56 | 319 | for (i = 1; i < ptbl->len; i++) { |
a6f23657 | 320 | part = rcu_dereference(ptbl->part[i]); |
a6f23657 JA |
321 | if (part && sector_in_part(part, sector)) { |
322 | rcu_assign_pointer(ptbl->last_lookup, part); | |
8ab1d40a | 323 | goto out_unlock; |
a6f23657 | 324 | } |
e71bf0d0 | 325 | } |
8ab1d40a | 326 | |
8446fe92 | 327 | part = disk->part0; |
8ab1d40a KK |
328 | out_unlock: |
329 | rcu_read_unlock(); | |
330 | return part; | |
e71bf0d0 | 331 | } |
e71bf0d0 | 332 | |
b53df2e7 SK |
333 | /** |
334 | * disk_has_partitions | |
335 | * @disk: gendisk of interest | |
336 | * | |
337 | * Walk through the partition table and check if valid partition exists. | |
338 | * | |
339 | * CONTEXT: | |
340 | * Don't care. | |
341 | * | |
342 | * RETURNS: | |
343 | * True if the gendisk has at least one valid non-zero size partition. | |
344 | * Otherwise false. | |
345 | */ | |
346 | bool disk_has_partitions(struct gendisk *disk) | |
347 | { | |
348 | struct disk_part_tbl *ptbl; | |
349 | int i; | |
350 | bool ret = false; | |
351 | ||
352 | rcu_read_lock(); | |
353 | ptbl = rcu_dereference(disk->part_tbl); | |
354 | ||
355 | /* Iterate partitions skipping the whole device at index 0 */ | |
356 | for (i = 1; i < ptbl->len; i++) { | |
357 | if (rcu_dereference(ptbl->part[i])) { | |
358 | ret = true; | |
359 | break; | |
360 | } | |
361 | } | |
362 | ||
363 | rcu_read_unlock(); | |
364 | ||
365 | return ret; | |
366 | } | |
367 | EXPORT_SYMBOL_GPL(disk_has_partitions); | |
368 | ||
1da177e4 LT |
369 | /* |
370 | * Can be deleted altogether. Later. | |
371 | * | |
372 | */ | |
133d55cd | 373 | #define BLKDEV_MAJOR_HASH_SIZE 255 |
1da177e4 LT |
374 | static struct blk_major_name { |
375 | struct blk_major_name *next; | |
376 | int major; | |
377 | char name[16]; | |
a160c615 | 378 | void (*probe)(dev_t devt); |
68eef3b4 | 379 | } *major_names[BLKDEV_MAJOR_HASH_SIZE]; |
e49fbbbf | 380 | static DEFINE_MUTEX(major_names_lock); |
1da177e4 LT |
381 | |
382 | /* index in the above - for now: assume no multimajor ranges */ | |
e61eb2e9 | 383 | static inline int major_to_index(unsigned major) |
1da177e4 | 384 | { |
68eef3b4 | 385 | return major % BLKDEV_MAJOR_HASH_SIZE; |
7170be5f NH |
386 | } |
387 | ||
68eef3b4 | 388 | #ifdef CONFIG_PROC_FS |
cf771cb5 | 389 | void blkdev_show(struct seq_file *seqf, off_t offset) |
7170be5f | 390 | { |
68eef3b4 | 391 | struct blk_major_name *dp; |
7170be5f | 392 | |
e49fbbbf | 393 | mutex_lock(&major_names_lock); |
133d55cd LG |
394 | for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next) |
395 | if (dp->major == offset) | |
cf771cb5 | 396 | seq_printf(seqf, "%3d %s\n", dp->major, dp->name); |
e49fbbbf | 397 | mutex_unlock(&major_names_lock); |
1da177e4 | 398 | } |
68eef3b4 | 399 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 400 | |
9e8c0bcc | 401 | /** |
e2b6b301 | 402 | * __register_blkdev - register a new block device |
9e8c0bcc | 403 | * |
f33ff110 SB |
404 | * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If |
405 | * @major = 0, try to allocate any unused major number. | |
9e8c0bcc | 406 | * @name: the name of the new block device as a zero terminated string |
e2b6b301 | 407 | * @probe: allback that is called on access to any minor number of @major |
9e8c0bcc MN |
408 | * |
409 | * The @name must be unique within the system. | |
410 | * | |
0e056eb5 MCC |
411 | * The return value depends on the @major input parameter: |
412 | * | |
f33ff110 SB |
413 | * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1] |
414 | * then the function returns zero on success, or a negative error code | |
0e056eb5 | 415 | * - if any unused major number was requested with @major = 0 parameter |
9e8c0bcc | 416 | * then the return value is the allocated major number in range |
f33ff110 SB |
417 | * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise |
418 | * | |
419 | * See Documentation/admin-guide/devices.txt for the list of allocated | |
420 | * major numbers. | |
e2b6b301 CH |
421 | * |
422 | * Use register_blkdev instead for any new code. | |
9e8c0bcc | 423 | */ |
a160c615 CH |
424 | int __register_blkdev(unsigned int major, const char *name, |
425 | void (*probe)(dev_t devt)) | |
1da177e4 LT |
426 | { |
427 | struct blk_major_name **n, *p; | |
428 | int index, ret = 0; | |
429 | ||
e49fbbbf | 430 | mutex_lock(&major_names_lock); |
1da177e4 LT |
431 | |
432 | /* temporary */ | |
433 | if (major == 0) { | |
434 | for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) { | |
435 | if (major_names[index] == NULL) | |
436 | break; | |
437 | } | |
438 | ||
439 | if (index == 0) { | |
dfc76d11 KP |
440 | printk("%s: failed to get major for %s\n", |
441 | __func__, name); | |
1da177e4 LT |
442 | ret = -EBUSY; |
443 | goto out; | |
444 | } | |
445 | major = index; | |
446 | ret = major; | |
447 | } | |
448 | ||
133d55cd | 449 | if (major >= BLKDEV_MAJOR_MAX) { |
dfc76d11 KP |
450 | pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n", |
451 | __func__, major, BLKDEV_MAJOR_MAX-1, name); | |
133d55cd LG |
452 | |
453 | ret = -EINVAL; | |
454 | goto out; | |
455 | } | |
456 | ||
1da177e4 LT |
457 | p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL); |
458 | if (p == NULL) { | |
459 | ret = -ENOMEM; | |
460 | goto out; | |
461 | } | |
462 | ||
463 | p->major = major; | |
a160c615 | 464 | p->probe = probe; |
1da177e4 LT |
465 | strlcpy(p->name, name, sizeof(p->name)); |
466 | p->next = NULL; | |
467 | index = major_to_index(major); | |
468 | ||
469 | for (n = &major_names[index]; *n; n = &(*n)->next) { | |
470 | if ((*n)->major == major) | |
471 | break; | |
472 | } | |
473 | if (!*n) | |
474 | *n = p; | |
475 | else | |
476 | ret = -EBUSY; | |
477 | ||
478 | if (ret < 0) { | |
f33ff110 | 479 | printk("register_blkdev: cannot get major %u for %s\n", |
1da177e4 LT |
480 | major, name); |
481 | kfree(p); | |
482 | } | |
483 | out: | |
e49fbbbf | 484 | mutex_unlock(&major_names_lock); |
1da177e4 LT |
485 | return ret; |
486 | } | |
a160c615 | 487 | EXPORT_SYMBOL(__register_blkdev); |
1da177e4 | 488 | |
f4480240 | 489 | void unregister_blkdev(unsigned int major, const char *name) |
1da177e4 LT |
490 | { |
491 | struct blk_major_name **n; | |
492 | struct blk_major_name *p = NULL; | |
493 | int index = major_to_index(major); | |
1da177e4 | 494 | |
e49fbbbf | 495 | mutex_lock(&major_names_lock); |
1da177e4 LT |
496 | for (n = &major_names[index]; *n; n = &(*n)->next) |
497 | if ((*n)->major == major) | |
498 | break; | |
294462a5 AM |
499 | if (!*n || strcmp((*n)->name, name)) { |
500 | WARN_ON(1); | |
294462a5 | 501 | } else { |
1da177e4 LT |
502 | p = *n; |
503 | *n = p->next; | |
504 | } | |
e49fbbbf | 505 | mutex_unlock(&major_names_lock); |
1da177e4 | 506 | kfree(p); |
1da177e4 LT |
507 | } |
508 | ||
509 | EXPORT_SYMBOL(unregister_blkdev); | |
510 | ||
870d6656 TH |
511 | /** |
512 | * blk_mangle_minor - scatter minor numbers apart | |
513 | * @minor: minor number to mangle | |
514 | * | |
515 | * Scatter consecutively allocated @minor number apart if MANGLE_DEVT | |
516 | * is enabled. Mangling twice gives the original value. | |
517 | * | |
518 | * RETURNS: | |
519 | * Mangled value. | |
520 | * | |
521 | * CONTEXT: | |
522 | * Don't care. | |
523 | */ | |
524 | static int blk_mangle_minor(int minor) | |
525 | { | |
526 | #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT | |
527 | int i; | |
528 | ||
529 | for (i = 0; i < MINORBITS / 2; i++) { | |
530 | int low = minor & (1 << i); | |
531 | int high = minor & (1 << (MINORBITS - 1 - i)); | |
532 | int distance = MINORBITS - 1 - 2 * i; | |
533 | ||
534 | minor ^= low | high; /* clear both bits */ | |
535 | low <<= distance; /* swap the positions */ | |
536 | high >>= distance; | |
537 | minor |= low | high; /* and set */ | |
538 | } | |
539 | #endif | |
540 | return minor; | |
541 | } | |
542 | ||
bcce3de1 | 543 | /** |
9fc995a6 CH |
544 | * blk_alloc_devt - allocate a dev_t for a block device |
545 | * @bdev: block device to allocate dev_t for | |
bcce3de1 TH |
546 | * @devt: out parameter for resulting dev_t |
547 | * | |
548 | * Allocate a dev_t for block device. | |
549 | * | |
550 | * RETURNS: | |
551 | * 0 on success, allocated dev_t is returned in *@devt. -errno on | |
552 | * failure. | |
553 | * | |
554 | * CONTEXT: | |
555 | * Might sleep. | |
556 | */ | |
9fc995a6 | 557 | int blk_alloc_devt(struct block_device *bdev, dev_t *devt) |
bcce3de1 | 558 | { |
9fc995a6 | 559 | struct gendisk *disk = bdev->bd_disk; |
bab998d6 | 560 | int idx; |
bcce3de1 TH |
561 | |
562 | /* in consecutive minor range? */ | |
9fc995a6 CH |
563 | if (bdev->bd_partno < disk->minors) { |
564 | *devt = MKDEV(disk->major, disk->first_minor + bdev->bd_partno); | |
bcce3de1 TH |
565 | return 0; |
566 | } | |
567 | ||
22ae8ce8 | 568 | idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT, GFP_KERNEL); |
bab998d6 TH |
569 | if (idx < 0) |
570 | return idx == -ENOSPC ? -EBUSY : idx; | |
bcce3de1 | 571 | |
870d6656 | 572 | *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx)); |
bcce3de1 TH |
573 | return 0; |
574 | } | |
575 | ||
576 | /** | |
577 | * blk_free_devt - free a dev_t | |
578 | * @devt: dev_t to free | |
579 | * | |
580 | * Free @devt which was allocated using blk_alloc_devt(). | |
581 | * | |
582 | * CONTEXT: | |
583 | * Might sleep. | |
584 | */ | |
585 | void blk_free_devt(dev_t devt) | |
586 | { | |
22ae8ce8 CH |
587 | if (MAJOR(devt) == BLOCK_EXT_MAJOR) |
588 | ida_free(&ext_devt_ida, blk_mangle_minor(MINOR(devt))); | |
6fcc44d1 YY |
589 | } |
590 | ||
1f014290 TH |
591 | static char *bdevt_str(dev_t devt, char *buf) |
592 | { | |
593 | if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) { | |
594 | char tbuf[BDEVT_SIZE]; | |
595 | snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt)); | |
596 | snprintf(buf, BDEVT_SIZE, "%-9s", tbuf); | |
597 | } else | |
598 | snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt)); | |
599 | ||
600 | return buf; | |
601 | } | |
602 | ||
9301fe73 CH |
603 | static void disk_scan_partitions(struct gendisk *disk) |
604 | { | |
605 | struct block_device *bdev; | |
606 | ||
607 | if (!get_capacity(disk) || !disk_part_scan_enabled(disk)) | |
608 | return; | |
609 | ||
610 | set_bit(GD_NEED_PART_SCAN, &disk->state); | |
611 | bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL); | |
612 | if (!IS_ERR(bdev)) | |
613 | blkdev_put(bdev, FMODE_READ); | |
614 | } | |
615 | ||
fef912bf HR |
616 | static void register_disk(struct device *parent, struct gendisk *disk, |
617 | const struct attribute_group **groups) | |
d2bf1b67 TH |
618 | { |
619 | struct device *ddev = disk_to_dev(disk); | |
d2bf1b67 | 620 | struct disk_part_iter piter; |
ad1eaa53 | 621 | struct block_device *part; |
d2bf1b67 TH |
622 | int err; |
623 | ||
e63a46be | 624 | ddev->parent = parent; |
d2bf1b67 | 625 | |
ffc8b308 | 626 | dev_set_name(ddev, "%s", disk->disk_name); |
d2bf1b67 TH |
627 | |
628 | /* delay uevents, until we scanned partition table */ | |
629 | dev_set_uevent_suppress(ddev, 1); | |
630 | ||
fef912bf HR |
631 | if (groups) { |
632 | WARN_ON(ddev->groups); | |
633 | ddev->groups = groups; | |
634 | } | |
d2bf1b67 TH |
635 | if (device_add(ddev)) |
636 | return; | |
637 | if (!sysfs_deprecated) { | |
638 | err = sysfs_create_link(block_depr, &ddev->kobj, | |
639 | kobject_name(&ddev->kobj)); | |
640 | if (err) { | |
641 | device_del(ddev); | |
642 | return; | |
643 | } | |
644 | } | |
25e823c8 ML |
645 | |
646 | /* | |
647 | * avoid probable deadlock caused by allocating memory with | |
648 | * GFP_KERNEL in runtime_resume callback of its all ancestor | |
649 | * devices | |
650 | */ | |
651 | pm_runtime_set_memalloc_noio(ddev, true); | |
652 | ||
cb8432d6 CH |
653 | disk->part0->bd_holder_dir = |
654 | kobject_create_and_add("holders", &ddev->kobj); | |
d2bf1b67 TH |
655 | disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj); |
656 | ||
8ddcd653 CH |
657 | if (disk->flags & GENHD_FL_HIDDEN) { |
658 | dev_set_uevent_suppress(ddev, 0); | |
659 | return; | |
660 | } | |
661 | ||
9301fe73 | 662 | disk_scan_partitions(disk); |
d2bf1b67 | 663 | |
d2bf1b67 TH |
664 | /* announce disk after possible partitions are created */ |
665 | dev_set_uevent_suppress(ddev, 0); | |
666 | kobject_uevent(&ddev->kobj, KOBJ_ADD); | |
667 | ||
668 | /* announce possible partitions */ | |
669 | disk_part_iter_init(&piter, disk, 0); | |
670 | while ((part = disk_part_iter_next(&piter))) | |
ad1eaa53 | 671 | kobject_uevent(bdev_kobj(part), KOBJ_ADD); |
d2bf1b67 | 672 | disk_part_iter_exit(&piter); |
8ddcd653 | 673 | |
4d7c1d3f | 674 | if (disk->queue->backing_dev_info->dev) { |
675 | err = sysfs_create_link(&ddev->kobj, | |
676 | &disk->queue->backing_dev_info->dev->kobj, | |
677 | "bdi"); | |
678 | WARN_ON(err); | |
679 | } | |
d2bf1b67 TH |
680 | } |
681 | ||
1da177e4 | 682 | /** |
fa70d2e2 | 683 | * __device_add_disk - add disk information to kernel list |
e63a46be | 684 | * @parent: parent device for the disk |
1da177e4 | 685 | * @disk: per-device partitioning information |
fef912bf | 686 | * @groups: Additional per-device sysfs groups |
fa70d2e2 | 687 | * @register_queue: register the queue if set to true |
1da177e4 LT |
688 | * |
689 | * This function registers the partitioning information in @disk | |
690 | * with the kernel. | |
3e1a7ff8 TH |
691 | * |
692 | * FIXME: error handling | |
1da177e4 | 693 | */ |
fa70d2e2 | 694 | static void __device_add_disk(struct device *parent, struct gendisk *disk, |
fef912bf | 695 | const struct attribute_group **groups, |
fa70d2e2 | 696 | bool register_queue) |
1da177e4 | 697 | { |
3e1a7ff8 | 698 | dev_t devt; |
6ffeea77 | 699 | int retval; |
cf0ca9fe | 700 | |
737eb78e DLM |
701 | /* |
702 | * The disk queue should now be all set with enough information about | |
703 | * the device for the elevator code to pick an adequate default | |
704 | * elevator if one is needed, that is, for devices requesting queue | |
705 | * registration. | |
706 | */ | |
707 | if (register_queue) | |
708 | elevator_init_mq(disk->queue); | |
709 | ||
3e1a7ff8 TH |
710 | /* minors == 0 indicates to use ext devt from part0 and should |
711 | * be accompanied with EXT_DEVT flag. Make sure all | |
712 | * parameters make sense. | |
713 | */ | |
714 | WARN_ON(disk->minors && !(disk->major || disk->first_minor)); | |
8ddcd653 CH |
715 | WARN_ON(!disk->minors && |
716 | !(disk->flags & (GENHD_FL_EXT_DEVT | GENHD_FL_HIDDEN))); | |
3e1a7ff8 | 717 | |
1da177e4 | 718 | disk->flags |= GENHD_FL_UP; |
3e1a7ff8 | 719 | |
9fc995a6 | 720 | retval = blk_alloc_devt(disk->part0, &devt); |
3e1a7ff8 TH |
721 | if (retval) { |
722 | WARN_ON(1); | |
723 | return; | |
724 | } | |
3e1a7ff8 TH |
725 | disk->major = MAJOR(devt); |
726 | disk->first_minor = MINOR(devt); | |
727 | ||
9f53d2fe SG |
728 | disk_alloc_events(disk); |
729 | ||
8ddcd653 CH |
730 | if (disk->flags & GENHD_FL_HIDDEN) { |
731 | /* | |
732 | * Don't let hidden disks show up in /proc/partitions, | |
733 | * and don't bother scanning for partitions either. | |
734 | */ | |
735 | disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO; | |
736 | disk->flags |= GENHD_FL_NO_PART_SCAN; | |
737 | } else { | |
3c5d202b CH |
738 | struct backing_dev_info *bdi = disk->queue->backing_dev_info; |
739 | struct device *dev = disk_to_dev(disk); | |
3a92168b | 740 | int ret; |
741 | ||
8ddcd653 | 742 | /* Register BDI before referencing it from bdev */ |
3c5d202b CH |
743 | dev->devt = devt; |
744 | ret = bdi_register(bdi, "%u:%u", MAJOR(devt), MINOR(devt)); | |
3a92168b | 745 | WARN_ON(ret); |
3c5d202b | 746 | bdi_set_owner(bdi, dev); |
cb8432d6 | 747 | bdev_add(disk->part0, devt); |
8ddcd653 | 748 | } |
fef912bf | 749 | register_disk(parent, disk, groups); |
fa70d2e2 MS |
750 | if (register_queue) |
751 | blk_register_queue(disk); | |
cf0ca9fe | 752 | |
523e1d39 TH |
753 | /* |
754 | * Take an extra ref on queue which will be put on disk_release() | |
755 | * so that it sticks around as long as @disk is there. | |
756 | */ | |
09ac46c4 | 757 | WARN_ON_ONCE(!blk_get_queue(disk->queue)); |
523e1d39 | 758 | |
77ea887e | 759 | disk_add_events(disk); |
25520d55 | 760 | blk_integrity_add(disk); |
1da177e4 | 761 | } |
fa70d2e2 | 762 | |
fef912bf HR |
763 | void device_add_disk(struct device *parent, struct gendisk *disk, |
764 | const struct attribute_group **groups) | |
765 | ||
fa70d2e2 | 766 | { |
fef912bf | 767 | __device_add_disk(parent, disk, groups, true); |
fa70d2e2 | 768 | } |
e63a46be | 769 | EXPORT_SYMBOL(device_add_disk); |
1da177e4 | 770 | |
fa70d2e2 MS |
771 | void device_add_disk_no_queue_reg(struct device *parent, struct gendisk *disk) |
772 | { | |
fef912bf | 773 | __device_add_disk(parent, disk, NULL, false); |
fa70d2e2 MS |
774 | } |
775 | EXPORT_SYMBOL(device_add_disk_no_queue_reg); | |
776 | ||
71773cf7 | 777 | static void invalidate_partition(struct block_device *bdev) |
02d33b67 | 778 | { |
02d33b67 CH |
779 | fsync_bdev(bdev); |
780 | __invalidate_device(bdev, true); | |
9bc5c397 CH |
781 | |
782 | /* | |
22ae8ce8 CH |
783 | * Unhash the bdev inode for this device so that it can't be looked |
784 | * up any more even if openers still hold references to it. | |
9bc5c397 CH |
785 | */ |
786 | remove_inode_hash(bdev->bd_inode); | |
02d33b67 CH |
787 | } |
788 | ||
b5bd357c LC |
789 | /** |
790 | * del_gendisk - remove the gendisk | |
791 | * @disk: the struct gendisk to remove | |
792 | * | |
793 | * Removes the gendisk and all its associated resources. This deletes the | |
794 | * partitions associated with the gendisk, and unregisters the associated | |
795 | * request_queue. | |
796 | * | |
797 | * This is the counter to the respective __device_add_disk() call. | |
798 | * | |
799 | * The final removal of the struct gendisk happens when its refcount reaches 0 | |
800 | * with put_disk(), which should be called after del_gendisk(), if | |
801 | * __device_add_disk() was used. | |
e8c7d14a LC |
802 | * |
803 | * Drivers exist which depend on the release of the gendisk to be synchronous, | |
804 | * it should not be deferred. | |
805 | * | |
806 | * Context: can sleep | |
b5bd357c | 807 | */ |
d2bf1b67 | 808 | void del_gendisk(struct gendisk *disk) |
1da177e4 | 809 | { |
d2bf1b67 | 810 | struct disk_part_iter piter; |
ad1eaa53 | 811 | struct block_device *part; |
d2bf1b67 | 812 | |
e8c7d14a LC |
813 | might_sleep(); |
814 | ||
6b3ba976 CH |
815 | if (WARN_ON_ONCE(!disk->queue)) |
816 | return; | |
817 | ||
25520d55 | 818 | blk_integrity_del(disk); |
77ea887e TH |
819 | disk_del_events(disk); |
820 | ||
56c0908c JK |
821 | /* |
822 | * Block lookups of the disk until all bdevs are unhashed and the | |
823 | * disk is marked as dead (GENHD_FL_UP cleared). | |
824 | */ | |
22ae8ce8 CH |
825 | down_write(&bdev_lookup_sem); |
826 | ||
d2bf1b67 TH |
827 | /* invalidate stuff */ |
828 | disk_part_iter_init(&piter, disk, | |
829 | DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE); | |
830 | while ((part = disk_part_iter_next(&piter))) { | |
ad1eaa53 | 831 | invalidate_partition(part); |
0d02129e | 832 | delete_partition(part); |
d2bf1b67 TH |
833 | } |
834 | disk_part_iter_exit(&piter); | |
835 | ||
71773cf7 | 836 | invalidate_partition(disk->part0); |
d2bf1b67 TH |
837 | set_capacity(disk, 0); |
838 | disk->flags &= ~GENHD_FL_UP; | |
22ae8ce8 | 839 | up_write(&bdev_lookup_sem); |
d2bf1b67 | 840 | |
6b3ba976 | 841 | if (!(disk->flags & GENHD_FL_HIDDEN)) { |
8ddcd653 | 842 | sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi"); |
6b3ba976 | 843 | |
90f16fdd JK |
844 | /* |
845 | * Unregister bdi before releasing device numbers (as they can | |
846 | * get reused and we'd get clashes in sysfs). | |
847 | */ | |
6b3ba976 | 848 | bdi_unregister(disk->queue->backing_dev_info); |
90f16fdd | 849 | } |
d2bf1b67 | 850 | |
6b3ba976 | 851 | blk_unregister_queue(disk); |
d2bf1b67 | 852 | |
cb8432d6 | 853 | kobject_put(disk->part0->bd_holder_dir); |
d2bf1b67 | 854 | kobject_put(disk->slave_dir); |
d2bf1b67 | 855 | |
8446fe92 | 856 | part_stat_set_all(disk->part0, 0); |
cb8432d6 | 857 | disk->part0->bd_stamp = 0; |
d2bf1b67 TH |
858 | if (!sysfs_deprecated) |
859 | sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk))); | |
25e823c8 | 860 | pm_runtime_set_memalloc_noio(disk_to_dev(disk), false); |
d2bf1b67 | 861 | device_del(disk_to_dev(disk)); |
1da177e4 | 862 | } |
d2bf1b67 | 863 | EXPORT_SYMBOL(del_gendisk); |
1da177e4 | 864 | |
99e6608c VV |
865 | /* sysfs access to bad-blocks list. */ |
866 | static ssize_t disk_badblocks_show(struct device *dev, | |
867 | struct device_attribute *attr, | |
868 | char *page) | |
869 | { | |
870 | struct gendisk *disk = dev_to_disk(dev); | |
871 | ||
872 | if (!disk->bb) | |
873 | return sprintf(page, "\n"); | |
874 | ||
875 | return badblocks_show(disk->bb, page, 0); | |
876 | } | |
877 | ||
878 | static ssize_t disk_badblocks_store(struct device *dev, | |
879 | struct device_attribute *attr, | |
880 | const char *page, size_t len) | |
881 | { | |
882 | struct gendisk *disk = dev_to_disk(dev); | |
883 | ||
884 | if (!disk->bb) | |
885 | return -ENXIO; | |
886 | ||
887 | return badblocks_store(disk->bb, page, len, 0); | |
888 | } | |
889 | ||
22ae8ce8 | 890 | void blk_request_module(dev_t devt) |
bd8eff3b | 891 | { |
a160c615 CH |
892 | unsigned int major = MAJOR(devt); |
893 | struct blk_major_name **n; | |
894 | ||
895 | mutex_lock(&major_names_lock); | |
896 | for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) { | |
897 | if ((*n)->major == major && (*n)->probe) { | |
898 | (*n)->probe(devt); | |
899 | mutex_unlock(&major_names_lock); | |
900 | return; | |
901 | } | |
902 | } | |
903 | mutex_unlock(&major_names_lock); | |
904 | ||
bd8eff3b CH |
905 | if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0) |
906 | /* Make old-style 2.4 aliases work */ | |
907 | request_module("block-major-%d", MAJOR(devt)); | |
908 | } | |
909 | ||
f331c029 TH |
910 | /** |
911 | * bdget_disk - do bdget() by gendisk and partition number | |
912 | * @disk: gendisk of interest | |
913 | * @partno: partition number | |
914 | * | |
915 | * Find partition @partno from @disk, do bdget() on it. | |
916 | * | |
917 | * CONTEXT: | |
918 | * Don't care. | |
919 | * | |
920 | * RETURNS: | |
921 | * Resulting block_device on success, NULL on failure. | |
922 | */ | |
aeb3d3a8 | 923 | struct block_device *bdget_disk(struct gendisk *disk, int partno) |
f331c029 | 924 | { |
548b10eb | 925 | struct block_device *bdev = NULL; |
f331c029 | 926 | |
0d02129e CH |
927 | rcu_read_lock(); |
928 | bdev = __disk_get_part(disk, partno); | |
929 | if (bdev && !bdgrab(bdev)) | |
930 | bdev = NULL; | |
931 | rcu_read_unlock(); | |
f331c029 | 932 | |
548b10eb | 933 | return bdev; |
f331c029 | 934 | } |
f331c029 | 935 | |
5c6f35c5 GKH |
936 | /* |
937 | * print a full list of all partitions - intended for places where the root | |
938 | * filesystem can't be mounted and thus to give the victim some idea of what | |
939 | * went wrong | |
940 | */ | |
941 | void __init printk_all_partitions(void) | |
942 | { | |
def4e38d TH |
943 | struct class_dev_iter iter; |
944 | struct device *dev; | |
945 | ||
946 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); | |
947 | while ((dev = class_dev_iter_next(&iter))) { | |
948 | struct gendisk *disk = dev_to_disk(dev); | |
e71bf0d0 | 949 | struct disk_part_iter piter; |
ad1eaa53 | 950 | struct block_device *part; |
1f014290 TH |
951 | char name_buf[BDEVNAME_SIZE]; |
952 | char devt_buf[BDEVT_SIZE]; | |
def4e38d TH |
953 | |
954 | /* | |
955 | * Don't show empty devices or things that have been | |
25985edc | 956 | * suppressed |
def4e38d TH |
957 | */ |
958 | if (get_capacity(disk) == 0 || | |
959 | (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)) | |
960 | continue; | |
961 | ||
962 | /* | |
963 | * Note, unlike /proc/partitions, I am showing the | |
964 | * numbers in hex - the same format as the root= | |
965 | * option takes. | |
966 | */ | |
074a7aca TH |
967 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0); |
968 | while ((part = disk_part_iter_next(&piter))) { | |
ad1eaa53 | 969 | bool is_part0 = part == disk->part0; |
def4e38d | 970 | |
b5af921e | 971 | printk("%s%s %10llu %s %s", is_part0 ? "" : " ", |
ad1eaa53 CH |
972 | bdevt_str(part->bd_dev, devt_buf), |
973 | bdev_nr_sectors(part) >> 1, | |
974 | disk_name(disk, part->bd_partno, name_buf), | |
975 | part->bd_meta_info ? | |
976 | part->bd_meta_info->uuid : ""); | |
074a7aca | 977 | if (is_part0) { |
52c44d93 | 978 | if (dev->parent && dev->parent->driver) |
074a7aca | 979 | printk(" driver: %s\n", |
52c44d93 | 980 | dev->parent->driver->name); |
074a7aca TH |
981 | else |
982 | printk(" (driver?)\n"); | |
983 | } else | |
984 | printk("\n"); | |
985 | } | |
e71bf0d0 | 986 | disk_part_iter_exit(&piter); |
def4e38d TH |
987 | } |
988 | class_dev_iter_exit(&iter); | |
dd2a345f DG |
989 | } |
990 | ||
1da177e4 LT |
991 | #ifdef CONFIG_PROC_FS |
992 | /* iterator */ | |
def4e38d | 993 | static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos) |
68c4d4a7 | 994 | { |
def4e38d TH |
995 | loff_t skip = *pos; |
996 | struct class_dev_iter *iter; | |
997 | struct device *dev; | |
68c4d4a7 | 998 | |
aeb3d3a8 | 999 | iter = kmalloc(sizeof(*iter), GFP_KERNEL); |
def4e38d TH |
1000 | if (!iter) |
1001 | return ERR_PTR(-ENOMEM); | |
1002 | ||
1003 | seqf->private = iter; | |
1004 | class_dev_iter_init(iter, &block_class, NULL, &disk_type); | |
1005 | do { | |
1006 | dev = class_dev_iter_next(iter); | |
1007 | if (!dev) | |
1008 | return NULL; | |
1009 | } while (skip--); | |
1010 | ||
1011 | return dev_to_disk(dev); | |
68c4d4a7 GKH |
1012 | } |
1013 | ||
def4e38d | 1014 | static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos) |
1da177e4 | 1015 | { |
edfaa7c3 | 1016 | struct device *dev; |
1da177e4 | 1017 | |
def4e38d TH |
1018 | (*pos)++; |
1019 | dev = class_dev_iter_next(seqf->private); | |
2ac3cee5 | 1020 | if (dev) |
68c4d4a7 | 1021 | return dev_to_disk(dev); |
2ac3cee5 | 1022 | |
1da177e4 LT |
1023 | return NULL; |
1024 | } | |
1025 | ||
def4e38d | 1026 | static void disk_seqf_stop(struct seq_file *seqf, void *v) |
27f30251 | 1027 | { |
def4e38d | 1028 | struct class_dev_iter *iter = seqf->private; |
27f30251 | 1029 | |
def4e38d TH |
1030 | /* stop is called even after start failed :-( */ |
1031 | if (iter) { | |
1032 | class_dev_iter_exit(iter); | |
1033 | kfree(iter); | |
77da1605 | 1034 | seqf->private = NULL; |
5c0ef6d0 | 1035 | } |
1da177e4 LT |
1036 | } |
1037 | ||
def4e38d | 1038 | static void *show_partition_start(struct seq_file *seqf, loff_t *pos) |
1da177e4 | 1039 | { |
06768067 | 1040 | void *p; |
def4e38d TH |
1041 | |
1042 | p = disk_seqf_start(seqf, pos); | |
b9f985b6 | 1043 | if (!IS_ERR_OR_NULL(p) && !*pos) |
def4e38d TH |
1044 | seq_puts(seqf, "major minor #blocks name\n\n"); |
1045 | return p; | |
1da177e4 LT |
1046 | } |
1047 | ||
cf771cb5 | 1048 | static int show_partition(struct seq_file *seqf, void *v) |
1da177e4 LT |
1049 | { |
1050 | struct gendisk *sgp = v; | |
e71bf0d0 | 1051 | struct disk_part_iter piter; |
ad1eaa53 | 1052 | struct block_device *part; |
1da177e4 LT |
1053 | char buf[BDEVNAME_SIZE]; |
1054 | ||
1da177e4 | 1055 | /* Don't show non-partitionable removeable devices or empty devices */ |
d27769ec | 1056 | if (!get_capacity(sgp) || (!disk_max_parts(sgp) && |
f331c029 | 1057 | (sgp->flags & GENHD_FL_REMOVABLE))) |
1da177e4 LT |
1058 | return 0; |
1059 | if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO) | |
1060 | return 0; | |
1061 | ||
1062 | /* show the full disk and all non-0 size partitions of it */ | |
074a7aca | 1063 | disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0); |
e71bf0d0 | 1064 | while ((part = disk_part_iter_next(&piter))) |
1f014290 | 1065 | seq_printf(seqf, "%4d %7d %10llu %s\n", |
ad1eaa53 CH |
1066 | MAJOR(part->bd_dev), MINOR(part->bd_dev), |
1067 | bdev_nr_sectors(part) >> 1, | |
1068 | disk_name(sgp, part->bd_partno, buf)); | |
e71bf0d0 | 1069 | disk_part_iter_exit(&piter); |
1da177e4 LT |
1070 | |
1071 | return 0; | |
1072 | } | |
1073 | ||
f500975a | 1074 | static const struct seq_operations partitions_op = { |
def4e38d TH |
1075 | .start = show_partition_start, |
1076 | .next = disk_seqf_next, | |
1077 | .stop = disk_seqf_stop, | |
edfaa7c3 | 1078 | .show = show_partition |
1da177e4 LT |
1079 | }; |
1080 | #endif | |
1081 | ||
1da177e4 LT |
1082 | static int __init genhd_device_init(void) |
1083 | { | |
e105b8bf DW |
1084 | int error; |
1085 | ||
1086 | block_class.dev_kobj = sysfs_dev_block_kobj; | |
1087 | error = class_register(&block_class); | |
ee27a558 RM |
1088 | if (unlikely(error)) |
1089 | return error; | |
1da177e4 | 1090 | blk_dev_init(); |
edfaa7c3 | 1091 | |
561ec68e ZY |
1092 | register_blkdev(BLOCK_EXT_MAJOR, "blkext"); |
1093 | ||
edfaa7c3 | 1094 | /* create top-level block dir */ |
e52eec13 AK |
1095 | if (!sysfs_deprecated) |
1096 | block_depr = kobject_create_and_add("block", NULL); | |
830d3cfb | 1097 | return 0; |
1da177e4 LT |
1098 | } |
1099 | ||
1100 | subsys_initcall(genhd_device_init); | |
1101 | ||
edfaa7c3 KS |
1102 | static ssize_t disk_range_show(struct device *dev, |
1103 | struct device_attribute *attr, char *buf) | |
1da177e4 | 1104 | { |
edfaa7c3 | 1105 | struct gendisk *disk = dev_to_disk(dev); |
1da177e4 | 1106 | |
edfaa7c3 | 1107 | return sprintf(buf, "%d\n", disk->minors); |
1da177e4 LT |
1108 | } |
1109 | ||
1f014290 TH |
1110 | static ssize_t disk_ext_range_show(struct device *dev, |
1111 | struct device_attribute *attr, char *buf) | |
1112 | { | |
1113 | struct gendisk *disk = dev_to_disk(dev); | |
1114 | ||
b5d0b9df | 1115 | return sprintf(buf, "%d\n", disk_max_parts(disk)); |
1f014290 TH |
1116 | } |
1117 | ||
edfaa7c3 KS |
1118 | static ssize_t disk_removable_show(struct device *dev, |
1119 | struct device_attribute *attr, char *buf) | |
a7fd6706 | 1120 | { |
edfaa7c3 | 1121 | struct gendisk *disk = dev_to_disk(dev); |
a7fd6706 | 1122 | |
edfaa7c3 KS |
1123 | return sprintf(buf, "%d\n", |
1124 | (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0)); | |
a7fd6706 KS |
1125 | } |
1126 | ||
8ddcd653 CH |
1127 | static ssize_t disk_hidden_show(struct device *dev, |
1128 | struct device_attribute *attr, char *buf) | |
1129 | { | |
1130 | struct gendisk *disk = dev_to_disk(dev); | |
1131 | ||
1132 | return sprintf(buf, "%d\n", | |
1133 | (disk->flags & GENHD_FL_HIDDEN ? 1 : 0)); | |
1134 | } | |
1135 | ||
1c9ce527 KS |
1136 | static ssize_t disk_ro_show(struct device *dev, |
1137 | struct device_attribute *attr, char *buf) | |
1138 | { | |
1139 | struct gendisk *disk = dev_to_disk(dev); | |
1140 | ||
b7db9956 | 1141 | return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0); |
1c9ce527 KS |
1142 | } |
1143 | ||
3ad5cee5 CH |
1144 | ssize_t part_size_show(struct device *dev, |
1145 | struct device_attribute *attr, char *buf) | |
1146 | { | |
0d02129e | 1147 | return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev))); |
3ad5cee5 CH |
1148 | } |
1149 | ||
1150 | ssize_t part_stat_show(struct device *dev, | |
1151 | struct device_attribute *attr, char *buf) | |
1152 | { | |
0d02129e CH |
1153 | struct block_device *bdev = dev_to_bdev(dev); |
1154 | struct request_queue *q = bdev->bd_disk->queue; | |
ea18e0f0 | 1155 | struct disk_stats stat; |
3ad5cee5 CH |
1156 | unsigned int inflight; |
1157 | ||
0d02129e | 1158 | part_stat_read_all(bdev, &stat); |
b2f609e1 | 1159 | if (queue_is_mq(q)) |
0d02129e | 1160 | inflight = blk_mq_in_flight(q, bdev); |
b2f609e1 | 1161 | else |
0d02129e | 1162 | inflight = part_in_flight(bdev); |
ea18e0f0 | 1163 | |
3ad5cee5 CH |
1164 | return sprintf(buf, |
1165 | "%8lu %8lu %8llu %8u " | |
1166 | "%8lu %8lu %8llu %8u " | |
1167 | "%8u %8u %8u " | |
1168 | "%8lu %8lu %8llu %8u " | |
1169 | "%8lu %8u" | |
1170 | "\n", | |
ea18e0f0 KK |
1171 | stat.ios[STAT_READ], |
1172 | stat.merges[STAT_READ], | |
1173 | (unsigned long long)stat.sectors[STAT_READ], | |
1174 | (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC), | |
1175 | stat.ios[STAT_WRITE], | |
1176 | stat.merges[STAT_WRITE], | |
1177 | (unsigned long long)stat.sectors[STAT_WRITE], | |
1178 | (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC), | |
3ad5cee5 | 1179 | inflight, |
ea18e0f0 | 1180 | jiffies_to_msecs(stat.io_ticks), |
8cd5b8fc KK |
1181 | (unsigned int)div_u64(stat.nsecs[STAT_READ] + |
1182 | stat.nsecs[STAT_WRITE] + | |
1183 | stat.nsecs[STAT_DISCARD] + | |
1184 | stat.nsecs[STAT_FLUSH], | |
1185 | NSEC_PER_MSEC), | |
ea18e0f0 KK |
1186 | stat.ios[STAT_DISCARD], |
1187 | stat.merges[STAT_DISCARD], | |
1188 | (unsigned long long)stat.sectors[STAT_DISCARD], | |
1189 | (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC), | |
1190 | stat.ios[STAT_FLUSH], | |
1191 | (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC)); | |
3ad5cee5 CH |
1192 | } |
1193 | ||
1194 | ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr, | |
1195 | char *buf) | |
1196 | { | |
0d02129e CH |
1197 | struct block_device *bdev = dev_to_bdev(dev); |
1198 | struct request_queue *q = bdev->bd_disk->queue; | |
3ad5cee5 CH |
1199 | unsigned int inflight[2]; |
1200 | ||
b2f609e1 | 1201 | if (queue_is_mq(q)) |
0d02129e | 1202 | blk_mq_in_flight_rw(q, bdev, inflight); |
b2f609e1 | 1203 | else |
0d02129e | 1204 | part_in_flight_rw(bdev, inflight); |
b2f609e1 | 1205 | |
3ad5cee5 CH |
1206 | return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]); |
1207 | } | |
1208 | ||
edfaa7c3 KS |
1209 | static ssize_t disk_capability_show(struct device *dev, |
1210 | struct device_attribute *attr, char *buf) | |
86ce18d7 | 1211 | { |
edfaa7c3 KS |
1212 | struct gendisk *disk = dev_to_disk(dev); |
1213 | ||
1214 | return sprintf(buf, "%x\n", disk->flags); | |
86ce18d7 | 1215 | } |
edfaa7c3 | 1216 | |
c72758f3 MP |
1217 | static ssize_t disk_alignment_offset_show(struct device *dev, |
1218 | struct device_attribute *attr, | |
1219 | char *buf) | |
1220 | { | |
1221 | struct gendisk *disk = dev_to_disk(dev); | |
1222 | ||
1223 | return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue)); | |
1224 | } | |
1225 | ||
86b37281 MP |
1226 | static ssize_t disk_discard_alignment_show(struct device *dev, |
1227 | struct device_attribute *attr, | |
1228 | char *buf) | |
1229 | { | |
1230 | struct gendisk *disk = dev_to_disk(dev); | |
1231 | ||
dd3d145d | 1232 | return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue)); |
86b37281 MP |
1233 | } |
1234 | ||
5657a819 JP |
1235 | static DEVICE_ATTR(range, 0444, disk_range_show, NULL); |
1236 | static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL); | |
1237 | static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL); | |
1238 | static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL); | |
1239 | static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL); | |
1240 | static DEVICE_ATTR(size, 0444, part_size_show, NULL); | |
1241 | static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL); | |
1242 | static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL); | |
1243 | static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL); | |
1244 | static DEVICE_ATTR(stat, 0444, part_stat_show, NULL); | |
1245 | static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL); | |
1246 | static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store); | |
3ad5cee5 | 1247 | |
c17bb495 | 1248 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
3ad5cee5 CH |
1249 | ssize_t part_fail_show(struct device *dev, |
1250 | struct device_attribute *attr, char *buf) | |
1251 | { | |
0d02129e | 1252 | return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail); |
3ad5cee5 CH |
1253 | } |
1254 | ||
1255 | ssize_t part_fail_store(struct device *dev, | |
1256 | struct device_attribute *attr, | |
1257 | const char *buf, size_t count) | |
1258 | { | |
3ad5cee5 CH |
1259 | int i; |
1260 | ||
1261 | if (count > 0 && sscanf(buf, "%d", &i) > 0) | |
0d02129e | 1262 | dev_to_bdev(dev)->bd_make_it_fail = i; |
3ad5cee5 CH |
1263 | |
1264 | return count; | |
1265 | } | |
1266 | ||
edfaa7c3 | 1267 | static struct device_attribute dev_attr_fail = |
5657a819 | 1268 | __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store); |
3ad5cee5 CH |
1269 | #endif /* CONFIG_FAIL_MAKE_REQUEST */ |
1270 | ||
581d4e28 JA |
1271 | #ifdef CONFIG_FAIL_IO_TIMEOUT |
1272 | static struct device_attribute dev_attr_fail_timeout = | |
5657a819 | 1273 | __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store); |
581d4e28 | 1274 | #endif |
edfaa7c3 KS |
1275 | |
1276 | static struct attribute *disk_attrs[] = { | |
1277 | &dev_attr_range.attr, | |
1f014290 | 1278 | &dev_attr_ext_range.attr, |
edfaa7c3 | 1279 | &dev_attr_removable.attr, |
8ddcd653 | 1280 | &dev_attr_hidden.attr, |
1c9ce527 | 1281 | &dev_attr_ro.attr, |
edfaa7c3 | 1282 | &dev_attr_size.attr, |
c72758f3 | 1283 | &dev_attr_alignment_offset.attr, |
86b37281 | 1284 | &dev_attr_discard_alignment.attr, |
edfaa7c3 KS |
1285 | &dev_attr_capability.attr, |
1286 | &dev_attr_stat.attr, | |
316d315b | 1287 | &dev_attr_inflight.attr, |
99e6608c | 1288 | &dev_attr_badblocks.attr, |
edfaa7c3 KS |
1289 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
1290 | &dev_attr_fail.attr, | |
581d4e28 JA |
1291 | #endif |
1292 | #ifdef CONFIG_FAIL_IO_TIMEOUT | |
1293 | &dev_attr_fail_timeout.attr, | |
edfaa7c3 KS |
1294 | #endif |
1295 | NULL | |
1296 | }; | |
1297 | ||
9438b3e0 DW |
1298 | static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n) |
1299 | { | |
1300 | struct device *dev = container_of(kobj, typeof(*dev), kobj); | |
1301 | struct gendisk *disk = dev_to_disk(dev); | |
1302 | ||
1303 | if (a == &dev_attr_badblocks.attr && !disk->bb) | |
1304 | return 0; | |
1305 | return a->mode; | |
1306 | } | |
1307 | ||
edfaa7c3 KS |
1308 | static struct attribute_group disk_attr_group = { |
1309 | .attrs = disk_attrs, | |
9438b3e0 | 1310 | .is_visible = disk_visible, |
edfaa7c3 KS |
1311 | }; |
1312 | ||
a4dbd674 | 1313 | static const struct attribute_group *disk_attr_groups[] = { |
edfaa7c3 KS |
1314 | &disk_attr_group, |
1315 | NULL | |
1da177e4 LT |
1316 | }; |
1317 | ||
540eed56 TH |
1318 | /** |
1319 | * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way | |
1320 | * @disk: disk to replace part_tbl for | |
1321 | * @new_ptbl: new part_tbl to install | |
1322 | * | |
1323 | * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The | |
1324 | * original ptbl is freed using RCU callback. | |
1325 | * | |
1326 | * LOCKING: | |
6d2cf6f2 | 1327 | * Matching bd_mutex locked or the caller is the only user of @disk. |
540eed56 TH |
1328 | */ |
1329 | static void disk_replace_part_tbl(struct gendisk *disk, | |
1330 | struct disk_part_tbl *new_ptbl) | |
1331 | { | |
6d2cf6f2 BVA |
1332 | struct disk_part_tbl *old_ptbl = |
1333 | rcu_dereference_protected(disk->part_tbl, 1); | |
540eed56 TH |
1334 | |
1335 | rcu_assign_pointer(disk->part_tbl, new_ptbl); | |
a6f23657 JA |
1336 | |
1337 | if (old_ptbl) { | |
1338 | rcu_assign_pointer(old_ptbl->last_lookup, NULL); | |
57bdfbf9 | 1339 | kfree_rcu(old_ptbl, rcu_head); |
a6f23657 | 1340 | } |
540eed56 TH |
1341 | } |
1342 | ||
1343 | /** | |
1344 | * disk_expand_part_tbl - expand disk->part_tbl | |
1345 | * @disk: disk to expand part_tbl for | |
1346 | * @partno: expand such that this partno can fit in | |
1347 | * | |
1348 | * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl | |
1349 | * uses RCU to allow unlocked dereferencing for stats and other stuff. | |
1350 | * | |
1351 | * LOCKING: | |
6d2cf6f2 BVA |
1352 | * Matching bd_mutex locked or the caller is the only user of @disk. |
1353 | * Might sleep. | |
540eed56 TH |
1354 | * |
1355 | * RETURNS: | |
1356 | * 0 on success, -errno on failure. | |
1357 | */ | |
1358 | int disk_expand_part_tbl(struct gendisk *disk, int partno) | |
1359 | { | |
6d2cf6f2 BVA |
1360 | struct disk_part_tbl *old_ptbl = |
1361 | rcu_dereference_protected(disk->part_tbl, 1); | |
540eed56 TH |
1362 | struct disk_part_tbl *new_ptbl; |
1363 | int len = old_ptbl ? old_ptbl->len : 0; | |
5fabcb4c | 1364 | int i, target; |
5fabcb4c JA |
1365 | |
1366 | /* | |
1367 | * check for int overflow, since we can get here from blkpg_ioctl() | |
1368 | * with a user passed 'partno'. | |
1369 | */ | |
1370 | target = partno + 1; | |
1371 | if (target < 0) | |
1372 | return -EINVAL; | |
540eed56 TH |
1373 | |
1374 | /* disk_max_parts() is zero during initialization, ignore if so */ | |
1375 | if (disk_max_parts(disk) && target > disk_max_parts(disk)) | |
1376 | return -EINVAL; | |
1377 | ||
1378 | if (target <= len) | |
1379 | return 0; | |
1380 | ||
78b90a2c GS |
1381 | new_ptbl = kzalloc_node(struct_size(new_ptbl, part, target), GFP_KERNEL, |
1382 | disk->node_id); | |
540eed56 TH |
1383 | if (!new_ptbl) |
1384 | return -ENOMEM; | |
1385 | ||
540eed56 TH |
1386 | new_ptbl->len = target; |
1387 | ||
1388 | for (i = 0; i < len; i++) | |
1389 | rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]); | |
1390 | ||
1391 | disk_replace_part_tbl(disk, new_ptbl); | |
1392 | return 0; | |
1393 | } | |
1394 | ||
b5bd357c LC |
1395 | /** |
1396 | * disk_release - releases all allocated resources of the gendisk | |
1397 | * @dev: the device representing this disk | |
1398 | * | |
1399 | * This function releases all allocated resources of the gendisk. | |
1400 | * | |
b5bd357c LC |
1401 | * Drivers which used __device_add_disk() have a gendisk with a request_queue |
1402 | * assigned. Since the request_queue sits on top of the gendisk for these | |
1403 | * drivers we also call blk_put_queue() for them, and we expect the | |
1404 | * request_queue refcount to reach 0 at this point, and so the request_queue | |
1405 | * will also be freed prior to the disk. | |
e8c7d14a LC |
1406 | * |
1407 | * Context: can sleep | |
b5bd357c | 1408 | */ |
edfaa7c3 | 1409 | static void disk_release(struct device *dev) |
1da177e4 | 1410 | { |
edfaa7c3 KS |
1411 | struct gendisk *disk = dev_to_disk(dev); |
1412 | ||
e8c7d14a LC |
1413 | might_sleep(); |
1414 | ||
2da78092 | 1415 | blk_free_devt(dev->devt); |
77ea887e | 1416 | disk_release_events(disk); |
1da177e4 | 1417 | kfree(disk->random); |
540eed56 | 1418 | disk_replace_part_tbl(disk, NULL); |
cb8432d6 | 1419 | bdput(disk->part0); |
523e1d39 TH |
1420 | if (disk->queue) |
1421 | blk_put_queue(disk->queue); | |
1da177e4 LT |
1422 | kfree(disk); |
1423 | } | |
edfaa7c3 KS |
1424 | struct class block_class = { |
1425 | .name = "block", | |
1da177e4 LT |
1426 | }; |
1427 | ||
3c2670e6 | 1428 | static char *block_devnode(struct device *dev, umode_t *mode, |
4e4098a3 | 1429 | kuid_t *uid, kgid_t *gid) |
b03f38b6 KS |
1430 | { |
1431 | struct gendisk *disk = dev_to_disk(dev); | |
1432 | ||
348e114b CH |
1433 | if (disk->fops->devnode) |
1434 | return disk->fops->devnode(disk, mode); | |
b03f38b6 KS |
1435 | return NULL; |
1436 | } | |
1437 | ||
ef45fe47 | 1438 | const struct device_type disk_type = { |
edfaa7c3 KS |
1439 | .name = "disk", |
1440 | .groups = disk_attr_groups, | |
1441 | .release = disk_release, | |
e454cea2 | 1442 | .devnode = block_devnode, |
1da177e4 LT |
1443 | }; |
1444 | ||
a6e2ba88 | 1445 | #ifdef CONFIG_PROC_FS |
cf771cb5 TH |
1446 | /* |
1447 | * aggregate disk stat collector. Uses the same stats that the sysfs | |
1448 | * entries do, above, but makes them available through one seq_file. | |
1449 | * | |
1450 | * The output looks suspiciously like /proc/partitions with a bunch of | |
1451 | * extra fields. | |
1452 | */ | |
1453 | static int diskstats_show(struct seq_file *seqf, void *v) | |
1da177e4 LT |
1454 | { |
1455 | struct gendisk *gp = v; | |
e71bf0d0 | 1456 | struct disk_part_iter piter; |
ad1eaa53 | 1457 | struct block_device *hd; |
1da177e4 | 1458 | char buf[BDEVNAME_SIZE]; |
e016b782 | 1459 | unsigned int inflight; |
ea18e0f0 | 1460 | struct disk_stats stat; |
1da177e4 LT |
1461 | |
1462 | /* | |
ed9e1982 | 1463 | if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next) |
cf771cb5 | 1464 | seq_puts(seqf, "major minor name" |
1da177e4 LT |
1465 | " rio rmerge rsect ruse wio wmerge " |
1466 | "wsect wuse running use aveq" | |
1467 | "\n\n"); | |
1468 | */ | |
9f5e4865 | 1469 | |
71982a40 | 1470 | disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0); |
e71bf0d0 | 1471 | while ((hd = disk_part_iter_next(&piter))) { |
0d02129e | 1472 | part_stat_read_all(hd, &stat); |
b2f609e1 | 1473 | if (queue_is_mq(gp->queue)) |
ad1eaa53 | 1474 | inflight = blk_mq_in_flight(gp->queue, hd); |
b2f609e1 | 1475 | else |
ad1eaa53 | 1476 | inflight = part_in_flight(hd); |
ea18e0f0 | 1477 | |
bdca3c87 MC |
1478 | seq_printf(seqf, "%4d %7d %s " |
1479 | "%lu %lu %lu %u " | |
1480 | "%lu %lu %lu %u " | |
1481 | "%u %u %u " | |
b6866318 KK |
1482 | "%lu %lu %lu %u " |
1483 | "%lu %u" | |
1484 | "\n", | |
ad1eaa53 CH |
1485 | MAJOR(hd->bd_dev), MINOR(hd->bd_dev), |
1486 | disk_name(gp, hd->bd_partno, buf), | |
ea18e0f0 KK |
1487 | stat.ios[STAT_READ], |
1488 | stat.merges[STAT_READ], | |
1489 | stat.sectors[STAT_READ], | |
1490 | (unsigned int)div_u64(stat.nsecs[STAT_READ], | |
1491 | NSEC_PER_MSEC), | |
1492 | stat.ios[STAT_WRITE], | |
1493 | stat.merges[STAT_WRITE], | |
1494 | stat.sectors[STAT_WRITE], | |
1495 | (unsigned int)div_u64(stat.nsecs[STAT_WRITE], | |
1496 | NSEC_PER_MSEC), | |
e016b782 | 1497 | inflight, |
ea18e0f0 | 1498 | jiffies_to_msecs(stat.io_ticks), |
8cd5b8fc KK |
1499 | (unsigned int)div_u64(stat.nsecs[STAT_READ] + |
1500 | stat.nsecs[STAT_WRITE] + | |
1501 | stat.nsecs[STAT_DISCARD] + | |
1502 | stat.nsecs[STAT_FLUSH], | |
1503 | NSEC_PER_MSEC), | |
ea18e0f0 KK |
1504 | stat.ios[STAT_DISCARD], |
1505 | stat.merges[STAT_DISCARD], | |
1506 | stat.sectors[STAT_DISCARD], | |
1507 | (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], | |
1508 | NSEC_PER_MSEC), | |
1509 | stat.ios[STAT_FLUSH], | |
1510 | (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], | |
1511 | NSEC_PER_MSEC) | |
28f39d55 | 1512 | ); |
1da177e4 | 1513 | } |
e71bf0d0 | 1514 | disk_part_iter_exit(&piter); |
9f5e4865 | 1515 | |
1da177e4 LT |
1516 | return 0; |
1517 | } | |
1518 | ||
31d85ab2 | 1519 | static const struct seq_operations diskstats_op = { |
def4e38d TH |
1520 | .start = disk_seqf_start, |
1521 | .next = disk_seqf_next, | |
1522 | .stop = disk_seqf_stop, | |
1da177e4 LT |
1523 | .show = diskstats_show |
1524 | }; | |
f500975a AD |
1525 | |
1526 | static int __init proc_genhd_init(void) | |
1527 | { | |
fddda2b7 CH |
1528 | proc_create_seq("diskstats", 0, NULL, &diskstats_op); |
1529 | proc_create_seq("partitions", 0, NULL, &partitions_op); | |
f500975a AD |
1530 | return 0; |
1531 | } | |
1532 | module_init(proc_genhd_init); | |
a6e2ba88 | 1533 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 1534 | |
cf771cb5 | 1535 | dev_t blk_lookup_devt(const char *name, int partno) |
a142be85 | 1536 | { |
def4e38d TH |
1537 | dev_t devt = MKDEV(0, 0); |
1538 | struct class_dev_iter iter; | |
1539 | struct device *dev; | |
a142be85 | 1540 | |
def4e38d TH |
1541 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); |
1542 | while ((dev = class_dev_iter_next(&iter))) { | |
a142be85 | 1543 | struct gendisk *disk = dev_to_disk(dev); |
0d02129e | 1544 | struct block_device *part; |
a142be85 | 1545 | |
3ada8b7e | 1546 | if (strcmp(dev_name(dev), name)) |
f331c029 | 1547 | continue; |
f331c029 | 1548 | |
41b8c853 NB |
1549 | if (partno < disk->minors) { |
1550 | /* We need to return the right devno, even | |
1551 | * if the partition doesn't exist yet. | |
1552 | */ | |
1553 | devt = MKDEV(MAJOR(dev->devt), | |
1554 | MINOR(dev->devt) + partno); | |
1555 | break; | |
1556 | } | |
0d02129e | 1557 | part = bdget_disk(disk, partno); |
2bbedcb4 | 1558 | if (part) { |
0d02129e CH |
1559 | devt = part->bd_dev; |
1560 | bdput(part); | |
548b10eb | 1561 | break; |
def4e38d | 1562 | } |
5c0ef6d0 | 1563 | } |
def4e38d | 1564 | class_dev_iter_exit(&iter); |
edfaa7c3 KS |
1565 | return devt; |
1566 | } | |
edfaa7c3 | 1567 | |
e319e1fb | 1568 | struct gendisk *__alloc_disk_node(int minors, int node_id) |
1946089a CL |
1569 | { |
1570 | struct gendisk *disk; | |
6d2cf6f2 | 1571 | struct disk_part_tbl *ptbl; |
1946089a | 1572 | |
de65b012 CH |
1573 | if (minors > DISK_MAX_PARTS) { |
1574 | printk(KERN_ERR | |
7fb52621 | 1575 | "block: can't allocate more than %d partitions\n", |
de65b012 CH |
1576 | DISK_MAX_PARTS); |
1577 | minors = DISK_MAX_PARTS; | |
1578 | } | |
1946089a | 1579 | |
c1b511eb | 1580 | disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id); |
f93af2a4 CH |
1581 | if (!disk) |
1582 | return NULL; | |
6c23a968 | 1583 | |
cb8432d6 CH |
1584 | disk->part0 = bdev_alloc(disk, 0); |
1585 | if (!disk->part0) | |
22ae8ce8 CH |
1586 | goto out_free_disk; |
1587 | ||
f93af2a4 | 1588 | disk->node_id = node_id; |
22ae8ce8 | 1589 | if (disk_expand_part_tbl(disk, 0)) |
15e3d2c5 | 1590 | goto out_bdput; |
f93af2a4 CH |
1591 | |
1592 | ptbl = rcu_dereference_protected(disk->part_tbl, 1); | |
8446fe92 | 1593 | rcu_assign_pointer(ptbl->part[0], disk->part0); |
f93af2a4 CH |
1594 | |
1595 | disk->minors = minors; | |
1596 | rand_initialize_disk(disk); | |
1597 | disk_to_dev(disk)->class = &block_class; | |
1598 | disk_to_dev(disk)->type = &disk_type; | |
1599 | device_initialize(disk_to_dev(disk)); | |
1da177e4 | 1600 | return disk; |
f93af2a4 | 1601 | |
22ae8ce8 | 1602 | out_bdput: |
cb8432d6 | 1603 | bdput(disk->part0); |
f93af2a4 CH |
1604 | out_free_disk: |
1605 | kfree(disk); | |
1606 | return NULL; | |
1da177e4 | 1607 | } |
e319e1fb | 1608 | EXPORT_SYMBOL(__alloc_disk_node); |
1da177e4 | 1609 | |
b5bd357c LC |
1610 | /** |
1611 | * put_disk - decrements the gendisk refcount | |
0d20dcc2 | 1612 | * @disk: the struct gendisk to decrement the refcount for |
b5bd357c LC |
1613 | * |
1614 | * This decrements the refcount for the struct gendisk. When this reaches 0 | |
1615 | * we'll have disk_release() called. | |
e8c7d14a LC |
1616 | * |
1617 | * Context: Any context, but the last reference must not be dropped from | |
1618 | * atomic context. | |
b5bd357c | 1619 | */ |
1da177e4 LT |
1620 | void put_disk(struct gendisk *disk) |
1621 | { | |
1622 | if (disk) | |
efdc41c8 | 1623 | put_device(disk_to_dev(disk)); |
1da177e4 | 1624 | } |
1da177e4 LT |
1625 | EXPORT_SYMBOL(put_disk); |
1626 | ||
e3264a4d HR |
1627 | static void set_disk_ro_uevent(struct gendisk *gd, int ro) |
1628 | { | |
1629 | char event[] = "DISK_RO=1"; | |
1630 | char *envp[] = { event, NULL }; | |
1631 | ||
1632 | if (!ro) | |
1633 | event[8] = '0'; | |
1634 | kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); | |
1635 | } | |
1636 | ||
1da177e4 LT |
1637 | void set_disk_ro(struct gendisk *disk, int flag) |
1638 | { | |
e71bf0d0 | 1639 | struct disk_part_iter piter; |
ad1eaa53 | 1640 | struct block_device *part; |
e71bf0d0 | 1641 | |
cb8432d6 | 1642 | if (disk->part0->bd_read_only != flag) { |
e3264a4d | 1643 | set_disk_ro_uevent(disk, flag); |
cb8432d6 | 1644 | disk->part0->bd_read_only = flag; |
e3264a4d HR |
1645 | } |
1646 | ||
1647 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY); | |
e71bf0d0 | 1648 | while ((part = disk_part_iter_next(&piter))) |
ad1eaa53 | 1649 | part->bd_read_only = flag; |
e71bf0d0 | 1650 | disk_part_iter_exit(&piter); |
1da177e4 LT |
1651 | } |
1652 | ||
1653 | EXPORT_SYMBOL(set_disk_ro); | |
1654 | ||
1655 | int bdev_read_only(struct block_device *bdev) | |
1656 | { | |
1657 | if (!bdev) | |
1658 | return 0; | |
83950d35 | 1659 | return bdev->bd_read_only; |
1da177e4 LT |
1660 | } |
1661 | ||
1662 | EXPORT_SYMBOL(bdev_read_only); | |
1663 | ||
77ea887e TH |
1664 | /* |
1665 | * Disk events - monitor disk events like media change and eject request. | |
1666 | */ | |
1667 | struct disk_events { | |
1668 | struct list_head node; /* all disk_event's */ | |
1669 | struct gendisk *disk; /* the associated disk */ | |
1670 | spinlock_t lock; | |
1671 | ||
fdd514e1 | 1672 | struct mutex block_mutex; /* protects blocking */ |
77ea887e TH |
1673 | int block; /* event blocking depth */ |
1674 | unsigned int pending; /* events already sent out */ | |
1675 | unsigned int clearing; /* events being cleared */ | |
1676 | ||
1677 | long poll_msecs; /* interval, -1 for default */ | |
1678 | struct delayed_work dwork; | |
1679 | }; | |
1680 | ||
1681 | static const char *disk_events_strs[] = { | |
1682 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change", | |
1683 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request", | |
1684 | }; | |
1685 | ||
1686 | static char *disk_uevents[] = { | |
1687 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1", | |
1688 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1", | |
1689 | }; | |
1690 | ||
1691 | /* list of all disk_events */ | |
1692 | static DEFINE_MUTEX(disk_events_mutex); | |
1693 | static LIST_HEAD(disk_events); | |
1694 | ||
1695 | /* disable in-kernel polling by default */ | |
1fe8f348 | 1696 | static unsigned long disk_events_dfl_poll_msecs; |
77ea887e TH |
1697 | |
1698 | static unsigned long disk_events_poll_jiffies(struct gendisk *disk) | |
1699 | { | |
1700 | struct disk_events *ev = disk->ev; | |
1701 | long intv_msecs = 0; | |
1702 | ||
1703 | /* | |
1704 | * If device-specific poll interval is set, always use it. If | |
673387a9 | 1705 | * the default is being used, poll if the POLL flag is set. |
77ea887e TH |
1706 | */ |
1707 | if (ev->poll_msecs >= 0) | |
1708 | intv_msecs = ev->poll_msecs; | |
c92e2f04 | 1709 | else if (disk->event_flags & DISK_EVENT_FLAG_POLL) |
77ea887e TH |
1710 | intv_msecs = disk_events_dfl_poll_msecs; |
1711 | ||
1712 | return msecs_to_jiffies(intv_msecs); | |
1713 | } | |
1714 | ||
c3af54af TH |
1715 | /** |
1716 | * disk_block_events - block and flush disk event checking | |
1717 | * @disk: disk to block events for | |
1718 | * | |
1719 | * On return from this function, it is guaranteed that event checking | |
1720 | * isn't in progress and won't happen until unblocked by | |
1721 | * disk_unblock_events(). Events blocking is counted and the actual | |
1722 | * unblocking happens after the matching number of unblocks are done. | |
1723 | * | |
1724 | * Note that this intentionally does not block event checking from | |
1725 | * disk_clear_events(). | |
1726 | * | |
1727 | * CONTEXT: | |
1728 | * Might sleep. | |
1729 | */ | |
1730 | void disk_block_events(struct gendisk *disk) | |
77ea887e TH |
1731 | { |
1732 | struct disk_events *ev = disk->ev; | |
1733 | unsigned long flags; | |
1734 | bool cancel; | |
1735 | ||
c3af54af TH |
1736 | if (!ev) |
1737 | return; | |
1738 | ||
fdd514e1 TH |
1739 | /* |
1740 | * Outer mutex ensures that the first blocker completes canceling | |
1741 | * the event work before further blockers are allowed to finish. | |
1742 | */ | |
1743 | mutex_lock(&ev->block_mutex); | |
1744 | ||
77ea887e TH |
1745 | spin_lock_irqsave(&ev->lock, flags); |
1746 | cancel = !ev->block++; | |
1747 | spin_unlock_irqrestore(&ev->lock, flags); | |
1748 | ||
c3af54af TH |
1749 | if (cancel) |
1750 | cancel_delayed_work_sync(&disk->ev->dwork); | |
fdd514e1 TH |
1751 | |
1752 | mutex_unlock(&ev->block_mutex); | |
77ea887e TH |
1753 | } |
1754 | ||
1755 | static void __disk_unblock_events(struct gendisk *disk, bool check_now) | |
1756 | { | |
1757 | struct disk_events *ev = disk->ev; | |
1758 | unsigned long intv; | |
1759 | unsigned long flags; | |
1760 | ||
1761 | spin_lock_irqsave(&ev->lock, flags); | |
1762 | ||
1763 | if (WARN_ON_ONCE(ev->block <= 0)) | |
1764 | goto out_unlock; | |
1765 | ||
1766 | if (--ev->block) | |
1767 | goto out_unlock; | |
1768 | ||
77ea887e | 1769 | intv = disk_events_poll_jiffies(disk); |
77ea887e | 1770 | if (check_now) |
695588f9 VK |
1771 | queue_delayed_work(system_freezable_power_efficient_wq, |
1772 | &ev->dwork, 0); | |
77ea887e | 1773 | else if (intv) |
695588f9 VK |
1774 | queue_delayed_work(system_freezable_power_efficient_wq, |
1775 | &ev->dwork, intv); | |
77ea887e TH |
1776 | out_unlock: |
1777 | spin_unlock_irqrestore(&ev->lock, flags); | |
1778 | } | |
1779 | ||
77ea887e TH |
1780 | /** |
1781 | * disk_unblock_events - unblock disk event checking | |
1782 | * @disk: disk to unblock events for | |
1783 | * | |
1784 | * Undo disk_block_events(). When the block count reaches zero, it | |
1785 | * starts events polling if configured. | |
1786 | * | |
1787 | * CONTEXT: | |
1788 | * Don't care. Safe to call from irq context. | |
1789 | */ | |
1790 | void disk_unblock_events(struct gendisk *disk) | |
1791 | { | |
1792 | if (disk->ev) | |
facc31dd | 1793 | __disk_unblock_events(disk, false); |
77ea887e TH |
1794 | } |
1795 | ||
1796 | /** | |
85ef06d1 TH |
1797 | * disk_flush_events - schedule immediate event checking and flushing |
1798 | * @disk: disk to check and flush events for | |
1799 | * @mask: events to flush | |
77ea887e | 1800 | * |
85ef06d1 TH |
1801 | * Schedule immediate event checking on @disk if not blocked. Events in |
1802 | * @mask are scheduled to be cleared from the driver. Note that this | |
1803 | * doesn't clear the events from @disk->ev. | |
77ea887e TH |
1804 | * |
1805 | * CONTEXT: | |
85ef06d1 | 1806 | * If @mask is non-zero must be called with bdev->bd_mutex held. |
77ea887e | 1807 | */ |
85ef06d1 | 1808 | void disk_flush_events(struct gendisk *disk, unsigned int mask) |
77ea887e | 1809 | { |
a9dce2a3 | 1810 | struct disk_events *ev = disk->ev; |
a9dce2a3 TH |
1811 | |
1812 | if (!ev) | |
1813 | return; | |
1814 | ||
85ef06d1 TH |
1815 | spin_lock_irq(&ev->lock); |
1816 | ev->clearing |= mask; | |
41f63c53 | 1817 | if (!ev->block) |
695588f9 VK |
1818 | mod_delayed_work(system_freezable_power_efficient_wq, |
1819 | &ev->dwork, 0); | |
85ef06d1 | 1820 | spin_unlock_irq(&ev->lock); |
77ea887e | 1821 | } |
77ea887e TH |
1822 | |
1823 | /** | |
1824 | * disk_clear_events - synchronously check, clear and return pending events | |
1825 | * @disk: disk to fetch and clear events from | |
da3dae54 | 1826 | * @mask: mask of events to be fetched and cleared |
77ea887e TH |
1827 | * |
1828 | * Disk events are synchronously checked and pending events in @mask | |
1829 | * are cleared and returned. This ignores the block count. | |
1830 | * | |
1831 | * CONTEXT: | |
1832 | * Might sleep. | |
1833 | */ | |
95f6f3a4 | 1834 | static unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask) |
77ea887e | 1835 | { |
77ea887e TH |
1836 | struct disk_events *ev = disk->ev; |
1837 | unsigned int pending; | |
12c2bdb2 | 1838 | unsigned int clearing = mask; |
77ea887e | 1839 | |
a564e23f | 1840 | if (!ev) |
77ea887e | 1841 | return 0; |
77ea887e | 1842 | |
12c2bdb2 DB |
1843 | disk_block_events(disk); |
1844 | ||
1845 | /* | |
1846 | * store the union of mask and ev->clearing on the stack so that the | |
1847 | * race with disk_flush_events does not cause ambiguity (ev->clearing | |
1848 | * can still be modified even if events are blocked). | |
1849 | */ | |
77ea887e | 1850 | spin_lock_irq(&ev->lock); |
12c2bdb2 DB |
1851 | clearing |= ev->clearing; |
1852 | ev->clearing = 0; | |
77ea887e TH |
1853 | spin_unlock_irq(&ev->lock); |
1854 | ||
12c2bdb2 | 1855 | disk_check_events(ev, &clearing); |
aea24a8b | 1856 | /* |
12c2bdb2 DB |
1857 | * if ev->clearing is not 0, the disk_flush_events got called in the |
1858 | * middle of this function, so we want to run the workfn without delay. | |
aea24a8b | 1859 | */ |
12c2bdb2 | 1860 | __disk_unblock_events(disk, ev->clearing ? true : false); |
77ea887e TH |
1861 | |
1862 | /* then, fetch and clear pending events */ | |
1863 | spin_lock_irq(&ev->lock); | |
77ea887e TH |
1864 | pending = ev->pending & mask; |
1865 | ev->pending &= ~mask; | |
1866 | spin_unlock_irq(&ev->lock); | |
12c2bdb2 | 1867 | WARN_ON_ONCE(clearing & mask); |
77ea887e TH |
1868 | |
1869 | return pending; | |
1870 | } | |
1871 | ||
95f6f3a4 CH |
1872 | /** |
1873 | * bdev_check_media_change - check if a removable media has been changed | |
1874 | * @bdev: block device to check | |
1875 | * | |
1876 | * Check whether a removable media has been changed, and attempt to free all | |
1877 | * dentries and inodes and invalidates all block device page cache entries in | |
1878 | * that case. | |
1879 | * | |
1880 | * Returns %true if the block device changed, or %false if not. | |
1881 | */ | |
1882 | bool bdev_check_media_change(struct block_device *bdev) | |
1883 | { | |
1884 | unsigned int events; | |
1885 | ||
1886 | events = disk_clear_events(bdev->bd_disk, DISK_EVENT_MEDIA_CHANGE | | |
1887 | DISK_EVENT_EJECT_REQUEST); | |
1888 | if (!(events & DISK_EVENT_MEDIA_CHANGE)) | |
1889 | return false; | |
1890 | ||
1891 | if (__invalidate_device(bdev, true)) | |
1892 | pr_warn("VFS: busy inodes on changed media %s\n", | |
1893 | bdev->bd_disk->disk_name); | |
38430f08 | 1894 | set_bit(GD_NEED_PART_SCAN, &bdev->bd_disk->state); |
95f6f3a4 CH |
1895 | return true; |
1896 | } | |
1897 | EXPORT_SYMBOL(bdev_check_media_change); | |
1898 | ||
12c2bdb2 DB |
1899 | /* |
1900 | * Separate this part out so that a different pointer for clearing_ptr can be | |
1901 | * passed in for disk_clear_events. | |
1902 | */ | |
77ea887e TH |
1903 | static void disk_events_workfn(struct work_struct *work) |
1904 | { | |
1905 | struct delayed_work *dwork = to_delayed_work(work); | |
1906 | struct disk_events *ev = container_of(dwork, struct disk_events, dwork); | |
12c2bdb2 DB |
1907 | |
1908 | disk_check_events(ev, &ev->clearing); | |
1909 | } | |
1910 | ||
1911 | static void disk_check_events(struct disk_events *ev, | |
1912 | unsigned int *clearing_ptr) | |
1913 | { | |
77ea887e TH |
1914 | struct gendisk *disk = ev->disk; |
1915 | char *envp[ARRAY_SIZE(disk_uevents) + 1] = { }; | |
12c2bdb2 | 1916 | unsigned int clearing = *clearing_ptr; |
77ea887e TH |
1917 | unsigned int events; |
1918 | unsigned long intv; | |
1919 | int nr_events = 0, i; | |
1920 | ||
1921 | /* check events */ | |
1922 | events = disk->fops->check_events(disk, clearing); | |
1923 | ||
1924 | /* accumulate pending events and schedule next poll if necessary */ | |
1925 | spin_lock_irq(&ev->lock); | |
1926 | ||
1927 | events &= ~ev->pending; | |
1928 | ev->pending |= events; | |
12c2bdb2 | 1929 | *clearing_ptr &= ~clearing; |
77ea887e TH |
1930 | |
1931 | intv = disk_events_poll_jiffies(disk); | |
1932 | if (!ev->block && intv) | |
695588f9 VK |
1933 | queue_delayed_work(system_freezable_power_efficient_wq, |
1934 | &ev->dwork, intv); | |
77ea887e TH |
1935 | |
1936 | spin_unlock_irq(&ev->lock); | |
1937 | ||
7c88a168 TH |
1938 | /* |
1939 | * Tell userland about new events. Only the events listed in | |
c92e2f04 MW |
1940 | * @disk->events are reported, and only if DISK_EVENT_FLAG_UEVENT |
1941 | * is set. Otherwise, events are processed internally but never | |
1942 | * get reported to userland. | |
7c88a168 | 1943 | */ |
77ea887e | 1944 | for (i = 0; i < ARRAY_SIZE(disk_uevents); i++) |
c92e2f04 MW |
1945 | if ((events & disk->events & (1 << i)) && |
1946 | (disk->event_flags & DISK_EVENT_FLAG_UEVENT)) | |
77ea887e TH |
1947 | envp[nr_events++] = disk_uevents[i]; |
1948 | ||
1949 | if (nr_events) | |
1950 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
1951 | } | |
1952 | ||
1953 | /* | |
1954 | * A disk events enabled device has the following sysfs nodes under | |
1955 | * its /sys/block/X/ directory. | |
1956 | * | |
1957 | * events : list of all supported events | |
1958 | * events_async : list of events which can be detected w/o polling | |
673387a9 | 1959 | * (always empty, only for backwards compatibility) |
77ea887e TH |
1960 | * events_poll_msecs : polling interval, 0: disable, -1: system default |
1961 | */ | |
1962 | static ssize_t __disk_events_show(unsigned int events, char *buf) | |
1963 | { | |
1964 | const char *delim = ""; | |
1965 | ssize_t pos = 0; | |
1966 | int i; | |
1967 | ||
1968 | for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++) | |
1969 | if (events & (1 << i)) { | |
1970 | pos += sprintf(buf + pos, "%s%s", | |
1971 | delim, disk_events_strs[i]); | |
1972 | delim = " "; | |
1973 | } | |
1974 | if (pos) | |
1975 | pos += sprintf(buf + pos, "\n"); | |
1976 | return pos; | |
1977 | } | |
1978 | ||
1979 | static ssize_t disk_events_show(struct device *dev, | |
1980 | struct device_attribute *attr, char *buf) | |
1981 | { | |
1982 | struct gendisk *disk = dev_to_disk(dev); | |
1983 | ||
c92e2f04 MW |
1984 | if (!(disk->event_flags & DISK_EVENT_FLAG_UEVENT)) |
1985 | return 0; | |
1986 | ||
77ea887e TH |
1987 | return __disk_events_show(disk->events, buf); |
1988 | } | |
1989 | ||
1990 | static ssize_t disk_events_async_show(struct device *dev, | |
1991 | struct device_attribute *attr, char *buf) | |
1992 | { | |
673387a9 | 1993 | return 0; |
77ea887e TH |
1994 | } |
1995 | ||
1996 | static ssize_t disk_events_poll_msecs_show(struct device *dev, | |
1997 | struct device_attribute *attr, | |
1998 | char *buf) | |
1999 | { | |
2000 | struct gendisk *disk = dev_to_disk(dev); | |
2001 | ||
cdf3e3de MW |
2002 | if (!disk->ev) |
2003 | return sprintf(buf, "-1\n"); | |
2004 | ||
77ea887e TH |
2005 | return sprintf(buf, "%ld\n", disk->ev->poll_msecs); |
2006 | } | |
2007 | ||
2008 | static ssize_t disk_events_poll_msecs_store(struct device *dev, | |
2009 | struct device_attribute *attr, | |
2010 | const char *buf, size_t count) | |
2011 | { | |
2012 | struct gendisk *disk = dev_to_disk(dev); | |
2013 | long intv; | |
2014 | ||
2015 | if (!count || !sscanf(buf, "%ld", &intv)) | |
2016 | return -EINVAL; | |
2017 | ||
2018 | if (intv < 0 && intv != -1) | |
2019 | return -EINVAL; | |
2020 | ||
cdf3e3de MW |
2021 | if (!disk->ev) |
2022 | return -ENODEV; | |
2023 | ||
c3af54af | 2024 | disk_block_events(disk); |
77ea887e TH |
2025 | disk->ev->poll_msecs = intv; |
2026 | __disk_unblock_events(disk, true); | |
2027 | ||
2028 | return count; | |
2029 | } | |
2030 | ||
5657a819 JP |
2031 | static const DEVICE_ATTR(events, 0444, disk_events_show, NULL); |
2032 | static const DEVICE_ATTR(events_async, 0444, disk_events_async_show, NULL); | |
2033 | static const DEVICE_ATTR(events_poll_msecs, 0644, | |
77ea887e TH |
2034 | disk_events_poll_msecs_show, |
2035 | disk_events_poll_msecs_store); | |
2036 | ||
2037 | static const struct attribute *disk_events_attrs[] = { | |
2038 | &dev_attr_events.attr, | |
2039 | &dev_attr_events_async.attr, | |
2040 | &dev_attr_events_poll_msecs.attr, | |
2041 | NULL, | |
2042 | }; | |
2043 | ||
2044 | /* | |
2045 | * The default polling interval can be specified by the kernel | |
2046 | * parameter block.events_dfl_poll_msecs which defaults to 0 | |
2047 | * (disable). This can also be modified runtime by writing to | |
1624b0b2 | 2048 | * /sys/module/block/parameters/events_dfl_poll_msecs. |
77ea887e TH |
2049 | */ |
2050 | static int disk_events_set_dfl_poll_msecs(const char *val, | |
2051 | const struct kernel_param *kp) | |
2052 | { | |
2053 | struct disk_events *ev; | |
2054 | int ret; | |
2055 | ||
2056 | ret = param_set_ulong(val, kp); | |
2057 | if (ret < 0) | |
2058 | return ret; | |
2059 | ||
2060 | mutex_lock(&disk_events_mutex); | |
2061 | ||
2062 | list_for_each_entry(ev, &disk_events, node) | |
85ef06d1 | 2063 | disk_flush_events(ev->disk, 0); |
77ea887e TH |
2064 | |
2065 | mutex_unlock(&disk_events_mutex); | |
2066 | ||
2067 | return 0; | |
2068 | } | |
2069 | ||
2070 | static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = { | |
2071 | .set = disk_events_set_dfl_poll_msecs, | |
2072 | .get = param_get_ulong, | |
2073 | }; | |
2074 | ||
2075 | #undef MODULE_PARAM_PREFIX | |
2076 | #define MODULE_PARAM_PREFIX "block." | |
2077 | ||
2078 | module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops, | |
2079 | &disk_events_dfl_poll_msecs, 0644); | |
2080 | ||
2081 | /* | |
9f53d2fe | 2082 | * disk_{alloc|add|del|release}_events - initialize and destroy disk_events. |
77ea887e | 2083 | */ |
9f53d2fe | 2084 | static void disk_alloc_events(struct gendisk *disk) |
77ea887e TH |
2085 | { |
2086 | struct disk_events *ev; | |
2087 | ||
cdf3e3de | 2088 | if (!disk->fops->check_events || !disk->events) |
77ea887e TH |
2089 | return; |
2090 | ||
2091 | ev = kzalloc(sizeof(*ev), GFP_KERNEL); | |
2092 | if (!ev) { | |
2093 | pr_warn("%s: failed to initialize events\n", disk->disk_name); | |
2094 | return; | |
2095 | } | |
2096 | ||
77ea887e TH |
2097 | INIT_LIST_HEAD(&ev->node); |
2098 | ev->disk = disk; | |
2099 | spin_lock_init(&ev->lock); | |
fdd514e1 | 2100 | mutex_init(&ev->block_mutex); |
77ea887e TH |
2101 | ev->block = 1; |
2102 | ev->poll_msecs = -1; | |
2103 | INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn); | |
2104 | ||
9f53d2fe SG |
2105 | disk->ev = ev; |
2106 | } | |
2107 | ||
2108 | static void disk_add_events(struct gendisk *disk) | |
2109 | { | |
9f53d2fe SG |
2110 | /* FIXME: error handling */ |
2111 | if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0) | |
2112 | pr_warn("%s: failed to create sysfs files for events\n", | |
2113 | disk->disk_name); | |
2114 | ||
cdf3e3de MW |
2115 | if (!disk->ev) |
2116 | return; | |
2117 | ||
77ea887e | 2118 | mutex_lock(&disk_events_mutex); |
9f53d2fe | 2119 | list_add_tail(&disk->ev->node, &disk_events); |
77ea887e TH |
2120 | mutex_unlock(&disk_events_mutex); |
2121 | ||
2122 | /* | |
2123 | * Block count is initialized to 1 and the following initial | |
2124 | * unblock kicks it into action. | |
2125 | */ | |
2126 | __disk_unblock_events(disk, true); | |
2127 | } | |
2128 | ||
2129 | static void disk_del_events(struct gendisk *disk) | |
2130 | { | |
cdf3e3de MW |
2131 | if (disk->ev) { |
2132 | disk_block_events(disk); | |
77ea887e | 2133 | |
cdf3e3de MW |
2134 | mutex_lock(&disk_events_mutex); |
2135 | list_del_init(&disk->ev->node); | |
2136 | mutex_unlock(&disk_events_mutex); | |
2137 | } | |
77ea887e TH |
2138 | |
2139 | sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs); | |
2140 | } | |
2141 | ||
2142 | static void disk_release_events(struct gendisk *disk) | |
2143 | { | |
2144 | /* the block count should be 1 from disk_del_events() */ | |
2145 | WARN_ON_ONCE(disk->ev && disk->ev->block != 1); | |
2146 | kfree(disk->ev); | |
2147 | } |