]>
Commit | Line | Data |
---|---|---|
3dcf60bc | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * gendisk handling | |
4 | */ | |
5 | ||
1da177e4 | 6 | #include <linux/module.h> |
3ad5cee5 | 7 | #include <linux/ctype.h> |
1da177e4 LT |
8 | #include <linux/fs.h> |
9 | #include <linux/genhd.h> | |
b446b60e | 10 | #include <linux/kdev_t.h> |
1da177e4 LT |
11 | #include <linux/kernel.h> |
12 | #include <linux/blkdev.h> | |
66114cad | 13 | #include <linux/backing-dev.h> |
1da177e4 LT |
14 | #include <linux/init.h> |
15 | #include <linux/spinlock.h> | |
f500975a | 16 | #include <linux/proc_fs.h> |
1da177e4 LT |
17 | #include <linux/seq_file.h> |
18 | #include <linux/slab.h> | |
19 | #include <linux/kmod.h> | |
58383af6 | 20 | #include <linux/mutex.h> |
bcce3de1 | 21 | #include <linux/idr.h> |
77ea887e | 22 | #include <linux/log2.h> |
25e823c8 | 23 | #include <linux/pm_runtime.h> |
99e6608c | 24 | #include <linux/badblocks.h> |
1da177e4 | 25 | |
ff88972c AB |
26 | #include "blk.h" |
27 | ||
31eb6186 | 28 | static struct kobject *block_depr; |
1da177e4 | 29 | |
22ae8ce8 | 30 | DECLARE_RWSEM(bdev_lookup_sem); |
62b508f8 | 31 | |
bcce3de1 | 32 | /* for extended dynamic devt allocation, currently only one major is used */ |
ce23bba8 | 33 | #define NR_EXT_DEVT (1 << MINORBITS) |
22ae8ce8 | 34 | static DEFINE_IDA(ext_devt_ida); |
bcce3de1 | 35 | |
12c2bdb2 DB |
36 | static void disk_check_events(struct disk_events *ev, |
37 | unsigned int *clearing_ptr); | |
9f53d2fe | 38 | static void disk_alloc_events(struct gendisk *disk); |
77ea887e TH |
39 | static void disk_add_events(struct gendisk *disk); |
40 | static void disk_del_events(struct gendisk *disk); | |
41 | static void disk_release_events(struct gendisk *disk); | |
42 | ||
a782483c CH |
43 | void set_capacity(struct gendisk *disk, sector_t sectors) |
44 | { | |
45 | struct block_device *bdev = disk->part0.bdev; | |
46 | ||
47 | spin_lock(&bdev->bd_size_lock); | |
48 | i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT); | |
49 | spin_unlock(&bdev->bd_size_lock); | |
50 | } | |
51 | EXPORT_SYMBOL(set_capacity); | |
52 | ||
e598a72f | 53 | /* |
449f4ec9 CH |
54 | * Set disk capacity and notify if the size is not currently zero and will not |
55 | * be set to zero. Returns true if a uevent was sent, otherwise false. | |
e598a72f | 56 | */ |
449f4ec9 | 57 | bool set_capacity_and_notify(struct gendisk *disk, sector_t size) |
e598a72f BS |
58 | { |
59 | sector_t capacity = get_capacity(disk); | |
a782483c | 60 | char *envp[] = { "RESIZE=1", NULL }; |
e598a72f BS |
61 | |
62 | set_capacity(disk, size); | |
e598a72f | 63 | |
a782483c CH |
64 | /* |
65 | * Only print a message and send a uevent if the gendisk is user visible | |
66 | * and alive. This avoids spamming the log and udev when setting the | |
67 | * initial capacity during probing. | |
68 | */ | |
69 | if (size == capacity || | |
70 | (disk->flags & (GENHD_FL_UP | GENHD_FL_HIDDEN)) != GENHD_FL_UP) | |
71 | return false; | |
e598a72f | 72 | |
a782483c CH |
73 | pr_info("%s: detected capacity change from %lld to %lld\n", |
74 | disk->disk_name, size, capacity); | |
7e890c37 | 75 | |
a782483c CH |
76 | /* |
77 | * Historically we did not send a uevent for changes to/from an empty | |
78 | * device. | |
79 | */ | |
80 | if (!capacity || !size) | |
81 | return false; | |
82 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
83 | return true; | |
e598a72f | 84 | } |
449f4ec9 | 85 | EXPORT_SYMBOL_GPL(set_capacity_and_notify); |
e598a72f | 86 | |
5cbd28e3 CH |
87 | /* |
88 | * Format the device name of the indicated disk into the supplied buffer and | |
89 | * return a pointer to that same buffer for convenience. | |
90 | */ | |
91 | char *disk_name(struct gendisk *hd, int partno, char *buf) | |
92 | { | |
93 | if (!partno) | |
94 | snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name); | |
95 | else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1])) | |
96 | snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno); | |
97 | else | |
98 | snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno); | |
99 | ||
100 | return buf; | |
101 | } | |
102 | ||
103 | const char *bdevname(struct block_device *bdev, char *buf) | |
104 | { | |
8a63a86e | 105 | return disk_name(bdev->bd_disk, bdev->bd_partno, buf); |
5cbd28e3 CH |
106 | } |
107 | EXPORT_SYMBOL(bdevname); | |
e598a72f | 108 | |
ea18e0f0 KK |
109 | static void part_stat_read_all(struct hd_struct *part, struct disk_stats *stat) |
110 | { | |
111 | int cpu; | |
112 | ||
113 | memset(stat, 0, sizeof(struct disk_stats)); | |
114 | for_each_possible_cpu(cpu) { | |
115 | struct disk_stats *ptr = per_cpu_ptr(part->dkstats, cpu); | |
116 | int group; | |
117 | ||
118 | for (group = 0; group < NR_STAT_GROUPS; group++) { | |
119 | stat->nsecs[group] += ptr->nsecs[group]; | |
120 | stat->sectors[group] += ptr->sectors[group]; | |
121 | stat->ios[group] += ptr->ios[group]; | |
122 | stat->merges[group] += ptr->merges[group]; | |
123 | } | |
124 | ||
125 | stat->io_ticks += ptr->io_ticks; | |
ea18e0f0 KK |
126 | } |
127 | } | |
ea18e0f0 | 128 | |
1f06959b | 129 | static unsigned int part_in_flight(struct hd_struct *part) |
f299b7c7 | 130 | { |
b2f609e1 | 131 | unsigned int inflight = 0; |
1226b8dd | 132 | int cpu; |
f299b7c7 | 133 | |
1226b8dd | 134 | for_each_possible_cpu(cpu) { |
e016b782 MP |
135 | inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) + |
136 | part_stat_local_read_cpu(part, in_flight[1], cpu); | |
1226b8dd | 137 | } |
e016b782 MP |
138 | if ((int)inflight < 0) |
139 | inflight = 0; | |
1226b8dd | 140 | |
e016b782 | 141 | return inflight; |
f299b7c7 JA |
142 | } |
143 | ||
1f06959b | 144 | static void part_in_flight_rw(struct hd_struct *part, unsigned int inflight[2]) |
bf0ddaba | 145 | { |
1226b8dd MP |
146 | int cpu; |
147 | ||
1226b8dd MP |
148 | inflight[0] = 0; |
149 | inflight[1] = 0; | |
150 | for_each_possible_cpu(cpu) { | |
151 | inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu); | |
152 | inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu); | |
153 | } | |
154 | if ((int)inflight[0] < 0) | |
155 | inflight[0] = 0; | |
156 | if ((int)inflight[1] < 0) | |
157 | inflight[1] = 0; | |
bf0ddaba OS |
158 | } |
159 | ||
807d4af2 CH |
160 | struct hd_struct *__disk_get_part(struct gendisk *disk, int partno) |
161 | { | |
162 | struct disk_part_tbl *ptbl = rcu_dereference(disk->part_tbl); | |
163 | ||
164 | if (unlikely(partno < 0 || partno >= ptbl->len)) | |
165 | return NULL; | |
166 | return rcu_dereference(ptbl->part[partno]); | |
167 | } | |
168 | ||
e71bf0d0 TH |
169 | /** |
170 | * disk_get_part - get partition | |
171 | * @disk: disk to look partition from | |
172 | * @partno: partition number | |
173 | * | |
174 | * Look for partition @partno from @disk. If found, increment | |
175 | * reference count and return it. | |
176 | * | |
177 | * CONTEXT: | |
178 | * Don't care. | |
179 | * | |
180 | * RETURNS: | |
181 | * Pointer to the found partition on success, NULL if not found. | |
182 | */ | |
183 | struct hd_struct *disk_get_part(struct gendisk *disk, int partno) | |
184 | { | |
807d4af2 | 185 | struct hd_struct *part; |
540eed56 | 186 | |
e71bf0d0 | 187 | rcu_read_lock(); |
807d4af2 CH |
188 | part = __disk_get_part(disk, partno); |
189 | if (part) | |
190 | get_device(part_to_dev(part)); | |
e71bf0d0 TH |
191 | rcu_read_unlock(); |
192 | ||
193 | return part; | |
194 | } | |
e71bf0d0 TH |
195 | |
196 | /** | |
197 | * disk_part_iter_init - initialize partition iterator | |
198 | * @piter: iterator to initialize | |
199 | * @disk: disk to iterate over | |
200 | * @flags: DISK_PITER_* flags | |
201 | * | |
202 | * Initialize @piter so that it iterates over partitions of @disk. | |
203 | * | |
204 | * CONTEXT: | |
205 | * Don't care. | |
206 | */ | |
207 | void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk, | |
208 | unsigned int flags) | |
209 | { | |
540eed56 TH |
210 | struct disk_part_tbl *ptbl; |
211 | ||
212 | rcu_read_lock(); | |
213 | ptbl = rcu_dereference(disk->part_tbl); | |
214 | ||
e71bf0d0 TH |
215 | piter->disk = disk; |
216 | piter->part = NULL; | |
217 | ||
218 | if (flags & DISK_PITER_REVERSE) | |
540eed56 | 219 | piter->idx = ptbl->len - 1; |
71982a40 | 220 | else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0)) |
e71bf0d0 | 221 | piter->idx = 0; |
b5d0b9df TH |
222 | else |
223 | piter->idx = 1; | |
e71bf0d0 TH |
224 | |
225 | piter->flags = flags; | |
540eed56 TH |
226 | |
227 | rcu_read_unlock(); | |
e71bf0d0 TH |
228 | } |
229 | EXPORT_SYMBOL_GPL(disk_part_iter_init); | |
230 | ||
231 | /** | |
232 | * disk_part_iter_next - proceed iterator to the next partition and return it | |
233 | * @piter: iterator of interest | |
234 | * | |
235 | * Proceed @piter to the next partition and return it. | |
236 | * | |
237 | * CONTEXT: | |
238 | * Don't care. | |
239 | */ | |
240 | struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter) | |
241 | { | |
540eed56 | 242 | struct disk_part_tbl *ptbl; |
e71bf0d0 TH |
243 | int inc, end; |
244 | ||
245 | /* put the last partition */ | |
e79319af | 246 | disk_part_iter_exit(piter); |
e71bf0d0 | 247 | |
540eed56 | 248 | /* get part_tbl */ |
e71bf0d0 | 249 | rcu_read_lock(); |
540eed56 | 250 | ptbl = rcu_dereference(piter->disk->part_tbl); |
e71bf0d0 TH |
251 | |
252 | /* determine iteration parameters */ | |
253 | if (piter->flags & DISK_PITER_REVERSE) { | |
254 | inc = -1; | |
71982a40 TH |
255 | if (piter->flags & (DISK_PITER_INCL_PART0 | |
256 | DISK_PITER_INCL_EMPTY_PART0)) | |
b5d0b9df TH |
257 | end = -1; |
258 | else | |
259 | end = 0; | |
e71bf0d0 TH |
260 | } else { |
261 | inc = 1; | |
540eed56 | 262 | end = ptbl->len; |
e71bf0d0 TH |
263 | } |
264 | ||
265 | /* iterate to the next partition */ | |
266 | for (; piter->idx != end; piter->idx += inc) { | |
267 | struct hd_struct *part; | |
268 | ||
540eed56 | 269 | part = rcu_dereference(ptbl->part[piter->idx]); |
e71bf0d0 TH |
270 | if (!part) |
271 | continue; | |
a782483c | 272 | if (!bdev_nr_sectors(part->bdev) && |
71982a40 TH |
273 | !(piter->flags & DISK_PITER_INCL_EMPTY) && |
274 | !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 && | |
275 | piter->idx == 0)) | |
e71bf0d0 TH |
276 | continue; |
277 | ||
ed9e1982 | 278 | get_device(part_to_dev(part)); |
e71bf0d0 TH |
279 | piter->part = part; |
280 | piter->idx += inc; | |
281 | break; | |
282 | } | |
283 | ||
284 | rcu_read_unlock(); | |
285 | ||
286 | return piter->part; | |
287 | } | |
288 | EXPORT_SYMBOL_GPL(disk_part_iter_next); | |
289 | ||
290 | /** | |
291 | * disk_part_iter_exit - finish up partition iteration | |
292 | * @piter: iter of interest | |
293 | * | |
294 | * Called when iteration is over. Cleans up @piter. | |
295 | * | |
296 | * CONTEXT: | |
297 | * Don't care. | |
298 | */ | |
299 | void disk_part_iter_exit(struct disk_part_iter *piter) | |
300 | { | |
301 | disk_put_part(piter->part); | |
302 | piter->part = NULL; | |
303 | } | |
304 | EXPORT_SYMBOL_GPL(disk_part_iter_exit); | |
305 | ||
a6f23657 JA |
306 | static inline int sector_in_part(struct hd_struct *part, sector_t sector) |
307 | { | |
308 | return part->start_sect <= sector && | |
a782483c | 309 | sector < part->start_sect + bdev_nr_sectors(part->bdev); |
a6f23657 JA |
310 | } |
311 | ||
e71bf0d0 TH |
312 | /** |
313 | * disk_map_sector_rcu - map sector to partition | |
314 | * @disk: gendisk of interest | |
315 | * @sector: sector to map | |
316 | * | |
317 | * Find out which partition @sector maps to on @disk. This is | |
318 | * primarily used for stats accounting. | |
319 | * | |
320 | * CONTEXT: | |
b7d6c303 | 321 | * RCU read locked. The returned partition pointer is always valid |
27eb3af9 ML |
322 | * because its refcount is grabbed except for part0, which lifetime |
323 | * is same with the disk. | |
e71bf0d0 TH |
324 | * |
325 | * RETURNS: | |
074a7aca | 326 | * Found partition on success, part0 is returned if no partition matches |
b7d6c303 | 327 | * or the matched partition is being deleted. |
e71bf0d0 TH |
328 | */ |
329 | struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector) | |
330 | { | |
540eed56 | 331 | struct disk_part_tbl *ptbl; |
a6f23657 | 332 | struct hd_struct *part; |
e71bf0d0 TH |
333 | int i; |
334 | ||
8ab1d40a | 335 | rcu_read_lock(); |
540eed56 TH |
336 | ptbl = rcu_dereference(disk->part_tbl); |
337 | ||
a6f23657 | 338 | part = rcu_dereference(ptbl->last_lookup); |
b7d6c303 | 339 | if (part && sector_in_part(part, sector) && hd_struct_try_get(part)) |
8ab1d40a | 340 | goto out_unlock; |
a6f23657 | 341 | |
540eed56 | 342 | for (i = 1; i < ptbl->len; i++) { |
a6f23657 | 343 | part = rcu_dereference(ptbl->part[i]); |
e71bf0d0 | 344 | |
a6f23657 | 345 | if (part && sector_in_part(part, sector)) { |
b7d6c303 ML |
346 | /* |
347 | * only live partition can be cached for lookup, | |
348 | * so use-after-free on cached & deleting partition | |
349 | * can be avoided | |
350 | */ | |
351 | if (!hd_struct_try_get(part)) | |
352 | break; | |
a6f23657 | 353 | rcu_assign_pointer(ptbl->last_lookup, part); |
8ab1d40a | 354 | goto out_unlock; |
a6f23657 | 355 | } |
e71bf0d0 | 356 | } |
8ab1d40a KK |
357 | |
358 | part = &disk->part0; | |
359 | out_unlock: | |
360 | rcu_read_unlock(); | |
361 | return part; | |
e71bf0d0 | 362 | } |
e71bf0d0 | 363 | |
b53df2e7 SK |
364 | /** |
365 | * disk_has_partitions | |
366 | * @disk: gendisk of interest | |
367 | * | |
368 | * Walk through the partition table and check if valid partition exists. | |
369 | * | |
370 | * CONTEXT: | |
371 | * Don't care. | |
372 | * | |
373 | * RETURNS: | |
374 | * True if the gendisk has at least one valid non-zero size partition. | |
375 | * Otherwise false. | |
376 | */ | |
377 | bool disk_has_partitions(struct gendisk *disk) | |
378 | { | |
379 | struct disk_part_tbl *ptbl; | |
380 | int i; | |
381 | bool ret = false; | |
382 | ||
383 | rcu_read_lock(); | |
384 | ptbl = rcu_dereference(disk->part_tbl); | |
385 | ||
386 | /* Iterate partitions skipping the whole device at index 0 */ | |
387 | for (i = 1; i < ptbl->len; i++) { | |
388 | if (rcu_dereference(ptbl->part[i])) { | |
389 | ret = true; | |
390 | break; | |
391 | } | |
392 | } | |
393 | ||
394 | rcu_read_unlock(); | |
395 | ||
396 | return ret; | |
397 | } | |
398 | EXPORT_SYMBOL_GPL(disk_has_partitions); | |
399 | ||
1da177e4 LT |
400 | /* |
401 | * Can be deleted altogether. Later. | |
402 | * | |
403 | */ | |
133d55cd | 404 | #define BLKDEV_MAJOR_HASH_SIZE 255 |
1da177e4 LT |
405 | static struct blk_major_name { |
406 | struct blk_major_name *next; | |
407 | int major; | |
408 | char name[16]; | |
a160c615 | 409 | void (*probe)(dev_t devt); |
68eef3b4 | 410 | } *major_names[BLKDEV_MAJOR_HASH_SIZE]; |
e49fbbbf | 411 | static DEFINE_MUTEX(major_names_lock); |
1da177e4 LT |
412 | |
413 | /* index in the above - for now: assume no multimajor ranges */ | |
e61eb2e9 | 414 | static inline int major_to_index(unsigned major) |
1da177e4 | 415 | { |
68eef3b4 | 416 | return major % BLKDEV_MAJOR_HASH_SIZE; |
7170be5f NH |
417 | } |
418 | ||
68eef3b4 | 419 | #ifdef CONFIG_PROC_FS |
cf771cb5 | 420 | void blkdev_show(struct seq_file *seqf, off_t offset) |
7170be5f | 421 | { |
68eef3b4 | 422 | struct blk_major_name *dp; |
7170be5f | 423 | |
e49fbbbf | 424 | mutex_lock(&major_names_lock); |
133d55cd LG |
425 | for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next) |
426 | if (dp->major == offset) | |
cf771cb5 | 427 | seq_printf(seqf, "%3d %s\n", dp->major, dp->name); |
e49fbbbf | 428 | mutex_unlock(&major_names_lock); |
1da177e4 | 429 | } |
68eef3b4 | 430 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 431 | |
9e8c0bcc | 432 | /** |
e2b6b301 | 433 | * __register_blkdev - register a new block device |
9e8c0bcc | 434 | * |
f33ff110 SB |
435 | * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If |
436 | * @major = 0, try to allocate any unused major number. | |
9e8c0bcc | 437 | * @name: the name of the new block device as a zero terminated string |
e2b6b301 | 438 | * @probe: allback that is called on access to any minor number of @major |
9e8c0bcc MN |
439 | * |
440 | * The @name must be unique within the system. | |
441 | * | |
0e056eb5 MCC |
442 | * The return value depends on the @major input parameter: |
443 | * | |
f33ff110 SB |
444 | * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1] |
445 | * then the function returns zero on success, or a negative error code | |
0e056eb5 | 446 | * - if any unused major number was requested with @major = 0 parameter |
9e8c0bcc | 447 | * then the return value is the allocated major number in range |
f33ff110 SB |
448 | * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise |
449 | * | |
450 | * See Documentation/admin-guide/devices.txt for the list of allocated | |
451 | * major numbers. | |
e2b6b301 CH |
452 | * |
453 | * Use register_blkdev instead for any new code. | |
9e8c0bcc | 454 | */ |
a160c615 CH |
455 | int __register_blkdev(unsigned int major, const char *name, |
456 | void (*probe)(dev_t devt)) | |
1da177e4 LT |
457 | { |
458 | struct blk_major_name **n, *p; | |
459 | int index, ret = 0; | |
460 | ||
e49fbbbf | 461 | mutex_lock(&major_names_lock); |
1da177e4 LT |
462 | |
463 | /* temporary */ | |
464 | if (major == 0) { | |
465 | for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) { | |
466 | if (major_names[index] == NULL) | |
467 | break; | |
468 | } | |
469 | ||
470 | if (index == 0) { | |
dfc76d11 KP |
471 | printk("%s: failed to get major for %s\n", |
472 | __func__, name); | |
1da177e4 LT |
473 | ret = -EBUSY; |
474 | goto out; | |
475 | } | |
476 | major = index; | |
477 | ret = major; | |
478 | } | |
479 | ||
133d55cd | 480 | if (major >= BLKDEV_MAJOR_MAX) { |
dfc76d11 KP |
481 | pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n", |
482 | __func__, major, BLKDEV_MAJOR_MAX-1, name); | |
133d55cd LG |
483 | |
484 | ret = -EINVAL; | |
485 | goto out; | |
486 | } | |
487 | ||
1da177e4 LT |
488 | p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL); |
489 | if (p == NULL) { | |
490 | ret = -ENOMEM; | |
491 | goto out; | |
492 | } | |
493 | ||
494 | p->major = major; | |
a160c615 | 495 | p->probe = probe; |
1da177e4 LT |
496 | strlcpy(p->name, name, sizeof(p->name)); |
497 | p->next = NULL; | |
498 | index = major_to_index(major); | |
499 | ||
500 | for (n = &major_names[index]; *n; n = &(*n)->next) { | |
501 | if ((*n)->major == major) | |
502 | break; | |
503 | } | |
504 | if (!*n) | |
505 | *n = p; | |
506 | else | |
507 | ret = -EBUSY; | |
508 | ||
509 | if (ret < 0) { | |
f33ff110 | 510 | printk("register_blkdev: cannot get major %u for %s\n", |
1da177e4 LT |
511 | major, name); |
512 | kfree(p); | |
513 | } | |
514 | out: | |
e49fbbbf | 515 | mutex_unlock(&major_names_lock); |
1da177e4 LT |
516 | return ret; |
517 | } | |
a160c615 | 518 | EXPORT_SYMBOL(__register_blkdev); |
1da177e4 | 519 | |
f4480240 | 520 | void unregister_blkdev(unsigned int major, const char *name) |
1da177e4 LT |
521 | { |
522 | struct blk_major_name **n; | |
523 | struct blk_major_name *p = NULL; | |
524 | int index = major_to_index(major); | |
1da177e4 | 525 | |
e49fbbbf | 526 | mutex_lock(&major_names_lock); |
1da177e4 LT |
527 | for (n = &major_names[index]; *n; n = &(*n)->next) |
528 | if ((*n)->major == major) | |
529 | break; | |
294462a5 AM |
530 | if (!*n || strcmp((*n)->name, name)) { |
531 | WARN_ON(1); | |
294462a5 | 532 | } else { |
1da177e4 LT |
533 | p = *n; |
534 | *n = p->next; | |
535 | } | |
e49fbbbf | 536 | mutex_unlock(&major_names_lock); |
1da177e4 | 537 | kfree(p); |
1da177e4 LT |
538 | } |
539 | ||
540 | EXPORT_SYMBOL(unregister_blkdev); | |
541 | ||
870d6656 TH |
542 | /** |
543 | * blk_mangle_minor - scatter minor numbers apart | |
544 | * @minor: minor number to mangle | |
545 | * | |
546 | * Scatter consecutively allocated @minor number apart if MANGLE_DEVT | |
547 | * is enabled. Mangling twice gives the original value. | |
548 | * | |
549 | * RETURNS: | |
550 | * Mangled value. | |
551 | * | |
552 | * CONTEXT: | |
553 | * Don't care. | |
554 | */ | |
555 | static int blk_mangle_minor(int minor) | |
556 | { | |
557 | #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT | |
558 | int i; | |
559 | ||
560 | for (i = 0; i < MINORBITS / 2; i++) { | |
561 | int low = minor & (1 << i); | |
562 | int high = minor & (1 << (MINORBITS - 1 - i)); | |
563 | int distance = MINORBITS - 1 - 2 * i; | |
564 | ||
565 | minor ^= low | high; /* clear both bits */ | |
566 | low <<= distance; /* swap the positions */ | |
567 | high >>= distance; | |
568 | minor |= low | high; /* and set */ | |
569 | } | |
570 | #endif | |
571 | return minor; | |
572 | } | |
573 | ||
bcce3de1 TH |
574 | /** |
575 | * blk_alloc_devt - allocate a dev_t for a partition | |
576 | * @part: partition to allocate dev_t for | |
bcce3de1 TH |
577 | * @devt: out parameter for resulting dev_t |
578 | * | |
579 | * Allocate a dev_t for block device. | |
580 | * | |
581 | * RETURNS: | |
582 | * 0 on success, allocated dev_t is returned in *@devt. -errno on | |
583 | * failure. | |
584 | * | |
585 | * CONTEXT: | |
586 | * Might sleep. | |
587 | */ | |
588 | int blk_alloc_devt(struct hd_struct *part, dev_t *devt) | |
589 | { | |
590 | struct gendisk *disk = part_to_disk(part); | |
bab998d6 | 591 | int idx; |
bcce3de1 TH |
592 | |
593 | /* in consecutive minor range? */ | |
594 | if (part->partno < disk->minors) { | |
595 | *devt = MKDEV(disk->major, disk->first_minor + part->partno); | |
596 | return 0; | |
597 | } | |
598 | ||
22ae8ce8 | 599 | idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT, GFP_KERNEL); |
bab998d6 TH |
600 | if (idx < 0) |
601 | return idx == -ENOSPC ? -EBUSY : idx; | |
bcce3de1 | 602 | |
870d6656 | 603 | *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx)); |
bcce3de1 TH |
604 | return 0; |
605 | } | |
606 | ||
607 | /** | |
608 | * blk_free_devt - free a dev_t | |
609 | * @devt: dev_t to free | |
610 | * | |
611 | * Free @devt which was allocated using blk_alloc_devt(). | |
612 | * | |
613 | * CONTEXT: | |
614 | * Might sleep. | |
615 | */ | |
616 | void blk_free_devt(dev_t devt) | |
617 | { | |
22ae8ce8 CH |
618 | if (MAJOR(devt) == BLOCK_EXT_MAJOR) |
619 | ida_free(&ext_devt_ida, blk_mangle_minor(MINOR(devt))); | |
6fcc44d1 YY |
620 | } |
621 | ||
1f014290 TH |
622 | static char *bdevt_str(dev_t devt, char *buf) |
623 | { | |
624 | if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) { | |
625 | char tbuf[BDEVT_SIZE]; | |
626 | snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt)); | |
627 | snprintf(buf, BDEVT_SIZE, "%-9s", tbuf); | |
628 | } else | |
629 | snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt)); | |
630 | ||
631 | return buf; | |
632 | } | |
633 | ||
9301fe73 CH |
634 | static void disk_scan_partitions(struct gendisk *disk) |
635 | { | |
636 | struct block_device *bdev; | |
637 | ||
638 | if (!get_capacity(disk) || !disk_part_scan_enabled(disk)) | |
639 | return; | |
640 | ||
641 | set_bit(GD_NEED_PART_SCAN, &disk->state); | |
642 | bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL); | |
643 | if (!IS_ERR(bdev)) | |
644 | blkdev_put(bdev, FMODE_READ); | |
645 | } | |
646 | ||
fef912bf HR |
647 | static void register_disk(struct device *parent, struct gendisk *disk, |
648 | const struct attribute_group **groups) | |
d2bf1b67 TH |
649 | { |
650 | struct device *ddev = disk_to_dev(disk); | |
d2bf1b67 TH |
651 | struct disk_part_iter piter; |
652 | struct hd_struct *part; | |
653 | int err; | |
654 | ||
e63a46be | 655 | ddev->parent = parent; |
d2bf1b67 | 656 | |
ffc8b308 | 657 | dev_set_name(ddev, "%s", disk->disk_name); |
d2bf1b67 TH |
658 | |
659 | /* delay uevents, until we scanned partition table */ | |
660 | dev_set_uevent_suppress(ddev, 1); | |
661 | ||
fef912bf HR |
662 | if (groups) { |
663 | WARN_ON(ddev->groups); | |
664 | ddev->groups = groups; | |
665 | } | |
d2bf1b67 TH |
666 | if (device_add(ddev)) |
667 | return; | |
668 | if (!sysfs_deprecated) { | |
669 | err = sysfs_create_link(block_depr, &ddev->kobj, | |
670 | kobject_name(&ddev->kobj)); | |
671 | if (err) { | |
672 | device_del(ddev); | |
673 | return; | |
674 | } | |
675 | } | |
25e823c8 ML |
676 | |
677 | /* | |
678 | * avoid probable deadlock caused by allocating memory with | |
679 | * GFP_KERNEL in runtime_resume callback of its all ancestor | |
680 | * devices | |
681 | */ | |
682 | pm_runtime_set_memalloc_noio(ddev, true); | |
683 | ||
d2bf1b67 TH |
684 | disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj); |
685 | disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj); | |
686 | ||
8ddcd653 CH |
687 | if (disk->flags & GENHD_FL_HIDDEN) { |
688 | dev_set_uevent_suppress(ddev, 0); | |
689 | return; | |
690 | } | |
691 | ||
9301fe73 | 692 | disk_scan_partitions(disk); |
d2bf1b67 | 693 | |
d2bf1b67 TH |
694 | /* announce disk after possible partitions are created */ |
695 | dev_set_uevent_suppress(ddev, 0); | |
696 | kobject_uevent(&ddev->kobj, KOBJ_ADD); | |
697 | ||
698 | /* announce possible partitions */ | |
699 | disk_part_iter_init(&piter, disk, 0); | |
700 | while ((part = disk_part_iter_next(&piter))) | |
701 | kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD); | |
702 | disk_part_iter_exit(&piter); | |
8ddcd653 | 703 | |
4d7c1d3f | 704 | if (disk->queue->backing_dev_info->dev) { |
705 | err = sysfs_create_link(&ddev->kobj, | |
706 | &disk->queue->backing_dev_info->dev->kobj, | |
707 | "bdi"); | |
708 | WARN_ON(err); | |
709 | } | |
d2bf1b67 TH |
710 | } |
711 | ||
1da177e4 | 712 | /** |
fa70d2e2 | 713 | * __device_add_disk - add disk information to kernel list |
e63a46be | 714 | * @parent: parent device for the disk |
1da177e4 | 715 | * @disk: per-device partitioning information |
fef912bf | 716 | * @groups: Additional per-device sysfs groups |
fa70d2e2 | 717 | * @register_queue: register the queue if set to true |
1da177e4 LT |
718 | * |
719 | * This function registers the partitioning information in @disk | |
720 | * with the kernel. | |
3e1a7ff8 TH |
721 | * |
722 | * FIXME: error handling | |
1da177e4 | 723 | */ |
fa70d2e2 | 724 | static void __device_add_disk(struct device *parent, struct gendisk *disk, |
fef912bf | 725 | const struct attribute_group **groups, |
fa70d2e2 | 726 | bool register_queue) |
1da177e4 | 727 | { |
3e1a7ff8 | 728 | dev_t devt; |
6ffeea77 | 729 | int retval; |
cf0ca9fe | 730 | |
737eb78e DLM |
731 | /* |
732 | * The disk queue should now be all set with enough information about | |
733 | * the device for the elevator code to pick an adequate default | |
734 | * elevator if one is needed, that is, for devices requesting queue | |
735 | * registration. | |
736 | */ | |
737 | if (register_queue) | |
738 | elevator_init_mq(disk->queue); | |
739 | ||
3e1a7ff8 TH |
740 | /* minors == 0 indicates to use ext devt from part0 and should |
741 | * be accompanied with EXT_DEVT flag. Make sure all | |
742 | * parameters make sense. | |
743 | */ | |
744 | WARN_ON(disk->minors && !(disk->major || disk->first_minor)); | |
8ddcd653 CH |
745 | WARN_ON(!disk->minors && |
746 | !(disk->flags & (GENHD_FL_EXT_DEVT | GENHD_FL_HIDDEN))); | |
3e1a7ff8 | 747 | |
1da177e4 | 748 | disk->flags |= GENHD_FL_UP; |
3e1a7ff8 TH |
749 | |
750 | retval = blk_alloc_devt(&disk->part0, &devt); | |
751 | if (retval) { | |
752 | WARN_ON(1); | |
753 | return; | |
754 | } | |
3e1a7ff8 TH |
755 | disk->major = MAJOR(devt); |
756 | disk->first_minor = MINOR(devt); | |
757 | ||
9f53d2fe SG |
758 | disk_alloc_events(disk); |
759 | ||
8ddcd653 CH |
760 | if (disk->flags & GENHD_FL_HIDDEN) { |
761 | /* | |
762 | * Don't let hidden disks show up in /proc/partitions, | |
763 | * and don't bother scanning for partitions either. | |
764 | */ | |
765 | disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO; | |
766 | disk->flags |= GENHD_FL_NO_PART_SCAN; | |
767 | } else { | |
3c5d202b CH |
768 | struct backing_dev_info *bdi = disk->queue->backing_dev_info; |
769 | struct device *dev = disk_to_dev(disk); | |
3a92168b | 770 | int ret; |
771 | ||
8ddcd653 | 772 | /* Register BDI before referencing it from bdev */ |
3c5d202b CH |
773 | dev->devt = devt; |
774 | ret = bdi_register(bdi, "%u:%u", MAJOR(devt), MINOR(devt)); | |
3a92168b | 775 | WARN_ON(ret); |
3c5d202b | 776 | bdi_set_owner(bdi, dev); |
22ae8ce8 | 777 | bdev_add(disk->part0.bdev, devt); |
8ddcd653 | 778 | } |
fef912bf | 779 | register_disk(parent, disk, groups); |
fa70d2e2 MS |
780 | if (register_queue) |
781 | blk_register_queue(disk); | |
cf0ca9fe | 782 | |
523e1d39 TH |
783 | /* |
784 | * Take an extra ref on queue which will be put on disk_release() | |
785 | * so that it sticks around as long as @disk is there. | |
786 | */ | |
09ac46c4 | 787 | WARN_ON_ONCE(!blk_get_queue(disk->queue)); |
523e1d39 | 788 | |
77ea887e | 789 | disk_add_events(disk); |
25520d55 | 790 | blk_integrity_add(disk); |
1da177e4 | 791 | } |
fa70d2e2 | 792 | |
fef912bf HR |
793 | void device_add_disk(struct device *parent, struct gendisk *disk, |
794 | const struct attribute_group **groups) | |
795 | ||
fa70d2e2 | 796 | { |
fef912bf | 797 | __device_add_disk(parent, disk, groups, true); |
fa70d2e2 | 798 | } |
e63a46be | 799 | EXPORT_SYMBOL(device_add_disk); |
1da177e4 | 800 | |
fa70d2e2 MS |
801 | void device_add_disk_no_queue_reg(struct device *parent, struct gendisk *disk) |
802 | { | |
fef912bf | 803 | __device_add_disk(parent, disk, NULL, false); |
fa70d2e2 MS |
804 | } |
805 | EXPORT_SYMBOL(device_add_disk_no_queue_reg); | |
806 | ||
02d33b67 CH |
807 | static void invalidate_partition(struct gendisk *disk, int partno) |
808 | { | |
809 | struct block_device *bdev; | |
810 | ||
811 | bdev = bdget_disk(disk, partno); | |
812 | if (!bdev) | |
813 | return; | |
814 | ||
815 | fsync_bdev(bdev); | |
816 | __invalidate_device(bdev, true); | |
9bc5c397 CH |
817 | |
818 | /* | |
22ae8ce8 CH |
819 | * Unhash the bdev inode for this device so that it can't be looked |
820 | * up any more even if openers still hold references to it. | |
9bc5c397 CH |
821 | */ |
822 | remove_inode_hash(bdev->bd_inode); | |
02d33b67 CH |
823 | bdput(bdev); |
824 | } | |
825 | ||
b5bd357c LC |
826 | /** |
827 | * del_gendisk - remove the gendisk | |
828 | * @disk: the struct gendisk to remove | |
829 | * | |
830 | * Removes the gendisk and all its associated resources. This deletes the | |
831 | * partitions associated with the gendisk, and unregisters the associated | |
832 | * request_queue. | |
833 | * | |
834 | * This is the counter to the respective __device_add_disk() call. | |
835 | * | |
836 | * The final removal of the struct gendisk happens when its refcount reaches 0 | |
837 | * with put_disk(), which should be called after del_gendisk(), if | |
838 | * __device_add_disk() was used. | |
e8c7d14a LC |
839 | * |
840 | * Drivers exist which depend on the release of the gendisk to be synchronous, | |
841 | * it should not be deferred. | |
842 | * | |
843 | * Context: can sleep | |
b5bd357c | 844 | */ |
d2bf1b67 | 845 | void del_gendisk(struct gendisk *disk) |
1da177e4 | 846 | { |
d2bf1b67 TH |
847 | struct disk_part_iter piter; |
848 | struct hd_struct *part; | |
849 | ||
e8c7d14a LC |
850 | might_sleep(); |
851 | ||
6b3ba976 CH |
852 | if (WARN_ON_ONCE(!disk->queue)) |
853 | return; | |
854 | ||
25520d55 | 855 | blk_integrity_del(disk); |
77ea887e TH |
856 | disk_del_events(disk); |
857 | ||
56c0908c JK |
858 | /* |
859 | * Block lookups of the disk until all bdevs are unhashed and the | |
860 | * disk is marked as dead (GENHD_FL_UP cleared). | |
861 | */ | |
22ae8ce8 CH |
862 | down_write(&bdev_lookup_sem); |
863 | ||
d2bf1b67 TH |
864 | /* invalidate stuff */ |
865 | disk_part_iter_init(&piter, disk, | |
866 | DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE); | |
867 | while ((part = disk_part_iter_next(&piter))) { | |
868 | invalidate_partition(disk, part->partno); | |
8328eb28 | 869 | delete_partition(part); |
d2bf1b67 TH |
870 | } |
871 | disk_part_iter_exit(&piter); | |
872 | ||
873 | invalidate_partition(disk, 0); | |
d2bf1b67 TH |
874 | set_capacity(disk, 0); |
875 | disk->flags &= ~GENHD_FL_UP; | |
22ae8ce8 | 876 | up_write(&bdev_lookup_sem); |
d2bf1b67 | 877 | |
6b3ba976 | 878 | if (!(disk->flags & GENHD_FL_HIDDEN)) { |
8ddcd653 | 879 | sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi"); |
6b3ba976 | 880 | |
90f16fdd JK |
881 | /* |
882 | * Unregister bdi before releasing device numbers (as they can | |
883 | * get reused and we'd get clashes in sysfs). | |
884 | */ | |
6b3ba976 | 885 | bdi_unregister(disk->queue->backing_dev_info); |
90f16fdd | 886 | } |
d2bf1b67 | 887 | |
6b3ba976 | 888 | blk_unregister_queue(disk); |
d2bf1b67 TH |
889 | |
890 | kobject_put(disk->part0.holder_dir); | |
891 | kobject_put(disk->slave_dir); | |
d2bf1b67 TH |
892 | |
893 | part_stat_set_all(&disk->part0, 0); | |
894 | disk->part0.stamp = 0; | |
d2bf1b67 TH |
895 | if (!sysfs_deprecated) |
896 | sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk))); | |
25e823c8 | 897 | pm_runtime_set_memalloc_noio(disk_to_dev(disk), false); |
d2bf1b67 | 898 | device_del(disk_to_dev(disk)); |
1da177e4 | 899 | } |
d2bf1b67 | 900 | EXPORT_SYMBOL(del_gendisk); |
1da177e4 | 901 | |
99e6608c VV |
902 | /* sysfs access to bad-blocks list. */ |
903 | static ssize_t disk_badblocks_show(struct device *dev, | |
904 | struct device_attribute *attr, | |
905 | char *page) | |
906 | { | |
907 | struct gendisk *disk = dev_to_disk(dev); | |
908 | ||
909 | if (!disk->bb) | |
910 | return sprintf(page, "\n"); | |
911 | ||
912 | return badblocks_show(disk->bb, page, 0); | |
913 | } | |
914 | ||
915 | static ssize_t disk_badblocks_store(struct device *dev, | |
916 | struct device_attribute *attr, | |
917 | const char *page, size_t len) | |
918 | { | |
919 | struct gendisk *disk = dev_to_disk(dev); | |
920 | ||
921 | if (!disk->bb) | |
922 | return -ENXIO; | |
923 | ||
924 | return badblocks_store(disk->bb, page, len, 0); | |
925 | } | |
926 | ||
22ae8ce8 | 927 | void blk_request_module(dev_t devt) |
bd8eff3b | 928 | { |
a160c615 CH |
929 | unsigned int major = MAJOR(devt); |
930 | struct blk_major_name **n; | |
931 | ||
932 | mutex_lock(&major_names_lock); | |
933 | for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) { | |
934 | if ((*n)->major == major && (*n)->probe) { | |
935 | (*n)->probe(devt); | |
936 | mutex_unlock(&major_names_lock); | |
937 | return; | |
938 | } | |
939 | } | |
940 | mutex_unlock(&major_names_lock); | |
941 | ||
bd8eff3b CH |
942 | if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0) |
943 | /* Make old-style 2.4 aliases work */ | |
944 | request_module("block-major-%d", MAJOR(devt)); | |
945 | } | |
946 | ||
f331c029 TH |
947 | /** |
948 | * bdget_disk - do bdget() by gendisk and partition number | |
949 | * @disk: gendisk of interest | |
950 | * @partno: partition number | |
951 | * | |
952 | * Find partition @partno from @disk, do bdget() on it. | |
953 | * | |
954 | * CONTEXT: | |
955 | * Don't care. | |
956 | * | |
957 | * RETURNS: | |
958 | * Resulting block_device on success, NULL on failure. | |
959 | */ | |
aeb3d3a8 | 960 | struct block_device *bdget_disk(struct gendisk *disk, int partno) |
f331c029 | 961 | { |
548b10eb TH |
962 | struct hd_struct *part; |
963 | struct block_device *bdev = NULL; | |
f331c029 | 964 | |
548b10eb | 965 | part = disk_get_part(disk, partno); |
2bbedcb4 | 966 | if (part) |
10ed1666 | 967 | bdev = bdget_part(part); |
548b10eb | 968 | disk_put_part(part); |
f331c029 | 969 | |
548b10eb | 970 | return bdev; |
f331c029 TH |
971 | } |
972 | EXPORT_SYMBOL(bdget_disk); | |
973 | ||
5c6f35c5 GKH |
974 | /* |
975 | * print a full list of all partitions - intended for places where the root | |
976 | * filesystem can't be mounted and thus to give the victim some idea of what | |
977 | * went wrong | |
978 | */ | |
979 | void __init printk_all_partitions(void) | |
980 | { | |
def4e38d TH |
981 | struct class_dev_iter iter; |
982 | struct device *dev; | |
983 | ||
984 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); | |
985 | while ((dev = class_dev_iter_next(&iter))) { | |
986 | struct gendisk *disk = dev_to_disk(dev); | |
e71bf0d0 TH |
987 | struct disk_part_iter piter; |
988 | struct hd_struct *part; | |
1f014290 TH |
989 | char name_buf[BDEVNAME_SIZE]; |
990 | char devt_buf[BDEVT_SIZE]; | |
def4e38d TH |
991 | |
992 | /* | |
993 | * Don't show empty devices or things that have been | |
25985edc | 994 | * suppressed |
def4e38d TH |
995 | */ |
996 | if (get_capacity(disk) == 0 || | |
997 | (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)) | |
998 | continue; | |
999 | ||
1000 | /* | |
1001 | * Note, unlike /proc/partitions, I am showing the | |
1002 | * numbers in hex - the same format as the root= | |
1003 | * option takes. | |
1004 | */ | |
074a7aca TH |
1005 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0); |
1006 | while ((part = disk_part_iter_next(&piter))) { | |
1007 | bool is_part0 = part == &disk->part0; | |
def4e38d | 1008 | |
b5af921e | 1009 | printk("%s%s %10llu %s %s", is_part0 ? "" : " ", |
1f014290 | 1010 | bdevt_str(part_devt(part), devt_buf), |
a782483c CH |
1011 | bdev_nr_sectors(part->bdev) >> 1, |
1012 | disk_name(disk, part->partno, name_buf), | |
1ad7e899 | 1013 | part->info ? part->info->uuid : ""); |
074a7aca | 1014 | if (is_part0) { |
52c44d93 | 1015 | if (dev->parent && dev->parent->driver) |
074a7aca | 1016 | printk(" driver: %s\n", |
52c44d93 | 1017 | dev->parent->driver->name); |
074a7aca TH |
1018 | else |
1019 | printk(" (driver?)\n"); | |
1020 | } else | |
1021 | printk("\n"); | |
1022 | } | |
e71bf0d0 | 1023 | disk_part_iter_exit(&piter); |
def4e38d TH |
1024 | } |
1025 | class_dev_iter_exit(&iter); | |
dd2a345f DG |
1026 | } |
1027 | ||
1da177e4 LT |
1028 | #ifdef CONFIG_PROC_FS |
1029 | /* iterator */ | |
def4e38d | 1030 | static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos) |
68c4d4a7 | 1031 | { |
def4e38d TH |
1032 | loff_t skip = *pos; |
1033 | struct class_dev_iter *iter; | |
1034 | struct device *dev; | |
68c4d4a7 | 1035 | |
aeb3d3a8 | 1036 | iter = kmalloc(sizeof(*iter), GFP_KERNEL); |
def4e38d TH |
1037 | if (!iter) |
1038 | return ERR_PTR(-ENOMEM); | |
1039 | ||
1040 | seqf->private = iter; | |
1041 | class_dev_iter_init(iter, &block_class, NULL, &disk_type); | |
1042 | do { | |
1043 | dev = class_dev_iter_next(iter); | |
1044 | if (!dev) | |
1045 | return NULL; | |
1046 | } while (skip--); | |
1047 | ||
1048 | return dev_to_disk(dev); | |
68c4d4a7 GKH |
1049 | } |
1050 | ||
def4e38d | 1051 | static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos) |
1da177e4 | 1052 | { |
edfaa7c3 | 1053 | struct device *dev; |
1da177e4 | 1054 | |
def4e38d TH |
1055 | (*pos)++; |
1056 | dev = class_dev_iter_next(seqf->private); | |
2ac3cee5 | 1057 | if (dev) |
68c4d4a7 | 1058 | return dev_to_disk(dev); |
2ac3cee5 | 1059 | |
1da177e4 LT |
1060 | return NULL; |
1061 | } | |
1062 | ||
def4e38d | 1063 | static void disk_seqf_stop(struct seq_file *seqf, void *v) |
27f30251 | 1064 | { |
def4e38d | 1065 | struct class_dev_iter *iter = seqf->private; |
27f30251 | 1066 | |
def4e38d TH |
1067 | /* stop is called even after start failed :-( */ |
1068 | if (iter) { | |
1069 | class_dev_iter_exit(iter); | |
1070 | kfree(iter); | |
77da1605 | 1071 | seqf->private = NULL; |
5c0ef6d0 | 1072 | } |
1da177e4 LT |
1073 | } |
1074 | ||
def4e38d | 1075 | static void *show_partition_start(struct seq_file *seqf, loff_t *pos) |
1da177e4 | 1076 | { |
06768067 | 1077 | void *p; |
def4e38d TH |
1078 | |
1079 | p = disk_seqf_start(seqf, pos); | |
b9f985b6 | 1080 | if (!IS_ERR_OR_NULL(p) && !*pos) |
def4e38d TH |
1081 | seq_puts(seqf, "major minor #blocks name\n\n"); |
1082 | return p; | |
1da177e4 LT |
1083 | } |
1084 | ||
cf771cb5 | 1085 | static int show_partition(struct seq_file *seqf, void *v) |
1da177e4 LT |
1086 | { |
1087 | struct gendisk *sgp = v; | |
e71bf0d0 TH |
1088 | struct disk_part_iter piter; |
1089 | struct hd_struct *part; | |
1da177e4 LT |
1090 | char buf[BDEVNAME_SIZE]; |
1091 | ||
1da177e4 | 1092 | /* Don't show non-partitionable removeable devices or empty devices */ |
d27769ec | 1093 | if (!get_capacity(sgp) || (!disk_max_parts(sgp) && |
f331c029 | 1094 | (sgp->flags & GENHD_FL_REMOVABLE))) |
1da177e4 LT |
1095 | return 0; |
1096 | if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO) | |
1097 | return 0; | |
1098 | ||
1099 | /* show the full disk and all non-0 size partitions of it */ | |
074a7aca | 1100 | disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0); |
e71bf0d0 | 1101 | while ((part = disk_part_iter_next(&piter))) |
1f014290 | 1102 | seq_printf(seqf, "%4d %7d %10llu %s\n", |
f331c029 | 1103 | MAJOR(part_devt(part)), MINOR(part_devt(part)), |
a782483c | 1104 | bdev_nr_sectors(part->bdev) >> 1, |
f331c029 | 1105 | disk_name(sgp, part->partno, buf)); |
e71bf0d0 | 1106 | disk_part_iter_exit(&piter); |
1da177e4 LT |
1107 | |
1108 | return 0; | |
1109 | } | |
1110 | ||
f500975a | 1111 | static const struct seq_operations partitions_op = { |
def4e38d TH |
1112 | .start = show_partition_start, |
1113 | .next = disk_seqf_next, | |
1114 | .stop = disk_seqf_stop, | |
edfaa7c3 | 1115 | .show = show_partition |
1da177e4 LT |
1116 | }; |
1117 | #endif | |
1118 | ||
1da177e4 LT |
1119 | static int __init genhd_device_init(void) |
1120 | { | |
e105b8bf DW |
1121 | int error; |
1122 | ||
1123 | block_class.dev_kobj = sysfs_dev_block_kobj; | |
1124 | error = class_register(&block_class); | |
ee27a558 RM |
1125 | if (unlikely(error)) |
1126 | return error; | |
1da177e4 | 1127 | blk_dev_init(); |
edfaa7c3 | 1128 | |
561ec68e ZY |
1129 | register_blkdev(BLOCK_EXT_MAJOR, "blkext"); |
1130 | ||
edfaa7c3 | 1131 | /* create top-level block dir */ |
e52eec13 AK |
1132 | if (!sysfs_deprecated) |
1133 | block_depr = kobject_create_and_add("block", NULL); | |
830d3cfb | 1134 | return 0; |
1da177e4 LT |
1135 | } |
1136 | ||
1137 | subsys_initcall(genhd_device_init); | |
1138 | ||
edfaa7c3 KS |
1139 | static ssize_t disk_range_show(struct device *dev, |
1140 | struct device_attribute *attr, char *buf) | |
1da177e4 | 1141 | { |
edfaa7c3 | 1142 | struct gendisk *disk = dev_to_disk(dev); |
1da177e4 | 1143 | |
edfaa7c3 | 1144 | return sprintf(buf, "%d\n", disk->minors); |
1da177e4 LT |
1145 | } |
1146 | ||
1f014290 TH |
1147 | static ssize_t disk_ext_range_show(struct device *dev, |
1148 | struct device_attribute *attr, char *buf) | |
1149 | { | |
1150 | struct gendisk *disk = dev_to_disk(dev); | |
1151 | ||
b5d0b9df | 1152 | return sprintf(buf, "%d\n", disk_max_parts(disk)); |
1f014290 TH |
1153 | } |
1154 | ||
edfaa7c3 KS |
1155 | static ssize_t disk_removable_show(struct device *dev, |
1156 | struct device_attribute *attr, char *buf) | |
a7fd6706 | 1157 | { |
edfaa7c3 | 1158 | struct gendisk *disk = dev_to_disk(dev); |
a7fd6706 | 1159 | |
edfaa7c3 KS |
1160 | return sprintf(buf, "%d\n", |
1161 | (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0)); | |
a7fd6706 KS |
1162 | } |
1163 | ||
8ddcd653 CH |
1164 | static ssize_t disk_hidden_show(struct device *dev, |
1165 | struct device_attribute *attr, char *buf) | |
1166 | { | |
1167 | struct gendisk *disk = dev_to_disk(dev); | |
1168 | ||
1169 | return sprintf(buf, "%d\n", | |
1170 | (disk->flags & GENHD_FL_HIDDEN ? 1 : 0)); | |
1171 | } | |
1172 | ||
1c9ce527 KS |
1173 | static ssize_t disk_ro_show(struct device *dev, |
1174 | struct device_attribute *attr, char *buf) | |
1175 | { | |
1176 | struct gendisk *disk = dev_to_disk(dev); | |
1177 | ||
b7db9956 | 1178 | return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0); |
1c9ce527 KS |
1179 | } |
1180 | ||
3ad5cee5 CH |
1181 | ssize_t part_size_show(struct device *dev, |
1182 | struct device_attribute *attr, char *buf) | |
1183 | { | |
1184 | struct hd_struct *p = dev_to_part(dev); | |
1185 | ||
a782483c | 1186 | return sprintf(buf, "%llu\n", bdev_nr_sectors(p->bdev)); |
3ad5cee5 CH |
1187 | } |
1188 | ||
1189 | ssize_t part_stat_show(struct device *dev, | |
1190 | struct device_attribute *attr, char *buf) | |
1191 | { | |
1192 | struct hd_struct *p = dev_to_part(dev); | |
1193 | struct request_queue *q = part_to_disk(p)->queue; | |
ea18e0f0 | 1194 | struct disk_stats stat; |
3ad5cee5 CH |
1195 | unsigned int inflight; |
1196 | ||
ea18e0f0 | 1197 | part_stat_read_all(p, &stat); |
b2f609e1 CH |
1198 | if (queue_is_mq(q)) |
1199 | inflight = blk_mq_in_flight(q, p); | |
1200 | else | |
1f06959b | 1201 | inflight = part_in_flight(p); |
ea18e0f0 | 1202 | |
3ad5cee5 CH |
1203 | return sprintf(buf, |
1204 | "%8lu %8lu %8llu %8u " | |
1205 | "%8lu %8lu %8llu %8u " | |
1206 | "%8u %8u %8u " | |
1207 | "%8lu %8lu %8llu %8u " | |
1208 | "%8lu %8u" | |
1209 | "\n", | |
ea18e0f0 KK |
1210 | stat.ios[STAT_READ], |
1211 | stat.merges[STAT_READ], | |
1212 | (unsigned long long)stat.sectors[STAT_READ], | |
1213 | (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC), | |
1214 | stat.ios[STAT_WRITE], | |
1215 | stat.merges[STAT_WRITE], | |
1216 | (unsigned long long)stat.sectors[STAT_WRITE], | |
1217 | (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC), | |
3ad5cee5 | 1218 | inflight, |
ea18e0f0 | 1219 | jiffies_to_msecs(stat.io_ticks), |
8cd5b8fc KK |
1220 | (unsigned int)div_u64(stat.nsecs[STAT_READ] + |
1221 | stat.nsecs[STAT_WRITE] + | |
1222 | stat.nsecs[STAT_DISCARD] + | |
1223 | stat.nsecs[STAT_FLUSH], | |
1224 | NSEC_PER_MSEC), | |
ea18e0f0 KK |
1225 | stat.ios[STAT_DISCARD], |
1226 | stat.merges[STAT_DISCARD], | |
1227 | (unsigned long long)stat.sectors[STAT_DISCARD], | |
1228 | (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC), | |
1229 | stat.ios[STAT_FLUSH], | |
1230 | (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC)); | |
3ad5cee5 CH |
1231 | } |
1232 | ||
1233 | ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr, | |
1234 | char *buf) | |
1235 | { | |
1236 | struct hd_struct *p = dev_to_part(dev); | |
1237 | struct request_queue *q = part_to_disk(p)->queue; | |
1238 | unsigned int inflight[2]; | |
1239 | ||
b2f609e1 CH |
1240 | if (queue_is_mq(q)) |
1241 | blk_mq_in_flight_rw(q, p, inflight); | |
1242 | else | |
1f06959b | 1243 | part_in_flight_rw(p, inflight); |
b2f609e1 | 1244 | |
3ad5cee5 CH |
1245 | return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]); |
1246 | } | |
1247 | ||
edfaa7c3 KS |
1248 | static ssize_t disk_capability_show(struct device *dev, |
1249 | struct device_attribute *attr, char *buf) | |
86ce18d7 | 1250 | { |
edfaa7c3 KS |
1251 | struct gendisk *disk = dev_to_disk(dev); |
1252 | ||
1253 | return sprintf(buf, "%x\n", disk->flags); | |
86ce18d7 | 1254 | } |
edfaa7c3 | 1255 | |
c72758f3 MP |
1256 | static ssize_t disk_alignment_offset_show(struct device *dev, |
1257 | struct device_attribute *attr, | |
1258 | char *buf) | |
1259 | { | |
1260 | struct gendisk *disk = dev_to_disk(dev); | |
1261 | ||
1262 | return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue)); | |
1263 | } | |
1264 | ||
86b37281 MP |
1265 | static ssize_t disk_discard_alignment_show(struct device *dev, |
1266 | struct device_attribute *attr, | |
1267 | char *buf) | |
1268 | { | |
1269 | struct gendisk *disk = dev_to_disk(dev); | |
1270 | ||
dd3d145d | 1271 | return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue)); |
86b37281 MP |
1272 | } |
1273 | ||
5657a819 JP |
1274 | static DEVICE_ATTR(range, 0444, disk_range_show, NULL); |
1275 | static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL); | |
1276 | static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL); | |
1277 | static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL); | |
1278 | static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL); | |
1279 | static DEVICE_ATTR(size, 0444, part_size_show, NULL); | |
1280 | static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL); | |
1281 | static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL); | |
1282 | static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL); | |
1283 | static DEVICE_ATTR(stat, 0444, part_stat_show, NULL); | |
1284 | static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL); | |
1285 | static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store); | |
3ad5cee5 | 1286 | |
c17bb495 | 1287 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
3ad5cee5 CH |
1288 | ssize_t part_fail_show(struct device *dev, |
1289 | struct device_attribute *attr, char *buf) | |
1290 | { | |
1291 | struct hd_struct *p = dev_to_part(dev); | |
1292 | ||
1293 | return sprintf(buf, "%d\n", p->make_it_fail); | |
1294 | } | |
1295 | ||
1296 | ssize_t part_fail_store(struct device *dev, | |
1297 | struct device_attribute *attr, | |
1298 | const char *buf, size_t count) | |
1299 | { | |
1300 | struct hd_struct *p = dev_to_part(dev); | |
1301 | int i; | |
1302 | ||
1303 | if (count > 0 && sscanf(buf, "%d", &i) > 0) | |
1304 | p->make_it_fail = (i == 0) ? 0 : 1; | |
1305 | ||
1306 | return count; | |
1307 | } | |
1308 | ||
edfaa7c3 | 1309 | static struct device_attribute dev_attr_fail = |
5657a819 | 1310 | __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store); |
3ad5cee5 CH |
1311 | #endif /* CONFIG_FAIL_MAKE_REQUEST */ |
1312 | ||
581d4e28 JA |
1313 | #ifdef CONFIG_FAIL_IO_TIMEOUT |
1314 | static struct device_attribute dev_attr_fail_timeout = | |
5657a819 | 1315 | __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store); |
581d4e28 | 1316 | #endif |
edfaa7c3 KS |
1317 | |
1318 | static struct attribute *disk_attrs[] = { | |
1319 | &dev_attr_range.attr, | |
1f014290 | 1320 | &dev_attr_ext_range.attr, |
edfaa7c3 | 1321 | &dev_attr_removable.attr, |
8ddcd653 | 1322 | &dev_attr_hidden.attr, |
1c9ce527 | 1323 | &dev_attr_ro.attr, |
edfaa7c3 | 1324 | &dev_attr_size.attr, |
c72758f3 | 1325 | &dev_attr_alignment_offset.attr, |
86b37281 | 1326 | &dev_attr_discard_alignment.attr, |
edfaa7c3 KS |
1327 | &dev_attr_capability.attr, |
1328 | &dev_attr_stat.attr, | |
316d315b | 1329 | &dev_attr_inflight.attr, |
99e6608c | 1330 | &dev_attr_badblocks.attr, |
edfaa7c3 KS |
1331 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
1332 | &dev_attr_fail.attr, | |
581d4e28 JA |
1333 | #endif |
1334 | #ifdef CONFIG_FAIL_IO_TIMEOUT | |
1335 | &dev_attr_fail_timeout.attr, | |
edfaa7c3 KS |
1336 | #endif |
1337 | NULL | |
1338 | }; | |
1339 | ||
9438b3e0 DW |
1340 | static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n) |
1341 | { | |
1342 | struct device *dev = container_of(kobj, typeof(*dev), kobj); | |
1343 | struct gendisk *disk = dev_to_disk(dev); | |
1344 | ||
1345 | if (a == &dev_attr_badblocks.attr && !disk->bb) | |
1346 | return 0; | |
1347 | return a->mode; | |
1348 | } | |
1349 | ||
edfaa7c3 KS |
1350 | static struct attribute_group disk_attr_group = { |
1351 | .attrs = disk_attrs, | |
9438b3e0 | 1352 | .is_visible = disk_visible, |
edfaa7c3 KS |
1353 | }; |
1354 | ||
a4dbd674 | 1355 | static const struct attribute_group *disk_attr_groups[] = { |
edfaa7c3 KS |
1356 | &disk_attr_group, |
1357 | NULL | |
1da177e4 LT |
1358 | }; |
1359 | ||
540eed56 TH |
1360 | /** |
1361 | * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way | |
1362 | * @disk: disk to replace part_tbl for | |
1363 | * @new_ptbl: new part_tbl to install | |
1364 | * | |
1365 | * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The | |
1366 | * original ptbl is freed using RCU callback. | |
1367 | * | |
1368 | * LOCKING: | |
6d2cf6f2 | 1369 | * Matching bd_mutex locked or the caller is the only user of @disk. |
540eed56 TH |
1370 | */ |
1371 | static void disk_replace_part_tbl(struct gendisk *disk, | |
1372 | struct disk_part_tbl *new_ptbl) | |
1373 | { | |
6d2cf6f2 BVA |
1374 | struct disk_part_tbl *old_ptbl = |
1375 | rcu_dereference_protected(disk->part_tbl, 1); | |
540eed56 TH |
1376 | |
1377 | rcu_assign_pointer(disk->part_tbl, new_ptbl); | |
a6f23657 JA |
1378 | |
1379 | if (old_ptbl) { | |
1380 | rcu_assign_pointer(old_ptbl->last_lookup, NULL); | |
57bdfbf9 | 1381 | kfree_rcu(old_ptbl, rcu_head); |
a6f23657 | 1382 | } |
540eed56 TH |
1383 | } |
1384 | ||
1385 | /** | |
1386 | * disk_expand_part_tbl - expand disk->part_tbl | |
1387 | * @disk: disk to expand part_tbl for | |
1388 | * @partno: expand such that this partno can fit in | |
1389 | * | |
1390 | * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl | |
1391 | * uses RCU to allow unlocked dereferencing for stats and other stuff. | |
1392 | * | |
1393 | * LOCKING: | |
6d2cf6f2 BVA |
1394 | * Matching bd_mutex locked or the caller is the only user of @disk. |
1395 | * Might sleep. | |
540eed56 TH |
1396 | * |
1397 | * RETURNS: | |
1398 | * 0 on success, -errno on failure. | |
1399 | */ | |
1400 | int disk_expand_part_tbl(struct gendisk *disk, int partno) | |
1401 | { | |
6d2cf6f2 BVA |
1402 | struct disk_part_tbl *old_ptbl = |
1403 | rcu_dereference_protected(disk->part_tbl, 1); | |
540eed56 TH |
1404 | struct disk_part_tbl *new_ptbl; |
1405 | int len = old_ptbl ? old_ptbl->len : 0; | |
5fabcb4c | 1406 | int i, target; |
5fabcb4c JA |
1407 | |
1408 | /* | |
1409 | * check for int overflow, since we can get here from blkpg_ioctl() | |
1410 | * with a user passed 'partno'. | |
1411 | */ | |
1412 | target = partno + 1; | |
1413 | if (target < 0) | |
1414 | return -EINVAL; | |
540eed56 TH |
1415 | |
1416 | /* disk_max_parts() is zero during initialization, ignore if so */ | |
1417 | if (disk_max_parts(disk) && target > disk_max_parts(disk)) | |
1418 | return -EINVAL; | |
1419 | ||
1420 | if (target <= len) | |
1421 | return 0; | |
1422 | ||
78b90a2c GS |
1423 | new_ptbl = kzalloc_node(struct_size(new_ptbl, part, target), GFP_KERNEL, |
1424 | disk->node_id); | |
540eed56 TH |
1425 | if (!new_ptbl) |
1426 | return -ENOMEM; | |
1427 | ||
540eed56 TH |
1428 | new_ptbl->len = target; |
1429 | ||
1430 | for (i = 0; i < len; i++) | |
1431 | rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]); | |
1432 | ||
1433 | disk_replace_part_tbl(disk, new_ptbl); | |
1434 | return 0; | |
1435 | } | |
1436 | ||
b5bd357c LC |
1437 | /** |
1438 | * disk_release - releases all allocated resources of the gendisk | |
1439 | * @dev: the device representing this disk | |
1440 | * | |
1441 | * This function releases all allocated resources of the gendisk. | |
1442 | * | |
b5bd357c LC |
1443 | * Drivers which used __device_add_disk() have a gendisk with a request_queue |
1444 | * assigned. Since the request_queue sits on top of the gendisk for these | |
1445 | * drivers we also call blk_put_queue() for them, and we expect the | |
1446 | * request_queue refcount to reach 0 at this point, and so the request_queue | |
1447 | * will also be freed prior to the disk. | |
e8c7d14a LC |
1448 | * |
1449 | * Context: can sleep | |
b5bd357c | 1450 | */ |
edfaa7c3 | 1451 | static void disk_release(struct device *dev) |
1da177e4 | 1452 | { |
edfaa7c3 KS |
1453 | struct gendisk *disk = dev_to_disk(dev); |
1454 | ||
e8c7d14a LC |
1455 | might_sleep(); |
1456 | ||
2da78092 | 1457 | blk_free_devt(dev->devt); |
77ea887e | 1458 | disk_release_events(disk); |
1da177e4 | 1459 | kfree(disk->random); |
540eed56 | 1460 | disk_replace_part_tbl(disk, NULL); |
b54e5ed8 | 1461 | hd_free_part(&disk->part0); |
523e1d39 TH |
1462 | if (disk->queue) |
1463 | blk_put_queue(disk->queue); | |
1da177e4 LT |
1464 | kfree(disk); |
1465 | } | |
edfaa7c3 KS |
1466 | struct class block_class = { |
1467 | .name = "block", | |
1da177e4 LT |
1468 | }; |
1469 | ||
3c2670e6 | 1470 | static char *block_devnode(struct device *dev, umode_t *mode, |
4e4098a3 | 1471 | kuid_t *uid, kgid_t *gid) |
b03f38b6 KS |
1472 | { |
1473 | struct gendisk *disk = dev_to_disk(dev); | |
1474 | ||
348e114b CH |
1475 | if (disk->fops->devnode) |
1476 | return disk->fops->devnode(disk, mode); | |
b03f38b6 KS |
1477 | return NULL; |
1478 | } | |
1479 | ||
ef45fe47 | 1480 | const struct device_type disk_type = { |
edfaa7c3 KS |
1481 | .name = "disk", |
1482 | .groups = disk_attr_groups, | |
1483 | .release = disk_release, | |
e454cea2 | 1484 | .devnode = block_devnode, |
1da177e4 LT |
1485 | }; |
1486 | ||
a6e2ba88 | 1487 | #ifdef CONFIG_PROC_FS |
cf771cb5 TH |
1488 | /* |
1489 | * aggregate disk stat collector. Uses the same stats that the sysfs | |
1490 | * entries do, above, but makes them available through one seq_file. | |
1491 | * | |
1492 | * The output looks suspiciously like /proc/partitions with a bunch of | |
1493 | * extra fields. | |
1494 | */ | |
1495 | static int diskstats_show(struct seq_file *seqf, void *v) | |
1da177e4 LT |
1496 | { |
1497 | struct gendisk *gp = v; | |
e71bf0d0 TH |
1498 | struct disk_part_iter piter; |
1499 | struct hd_struct *hd; | |
1da177e4 | 1500 | char buf[BDEVNAME_SIZE]; |
e016b782 | 1501 | unsigned int inflight; |
ea18e0f0 | 1502 | struct disk_stats stat; |
1da177e4 LT |
1503 | |
1504 | /* | |
ed9e1982 | 1505 | if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next) |
cf771cb5 | 1506 | seq_puts(seqf, "major minor name" |
1da177e4 LT |
1507 | " rio rmerge rsect ruse wio wmerge " |
1508 | "wsect wuse running use aveq" | |
1509 | "\n\n"); | |
1510 | */ | |
9f5e4865 | 1511 | |
71982a40 | 1512 | disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0); |
e71bf0d0 | 1513 | while ((hd = disk_part_iter_next(&piter))) { |
ea18e0f0 | 1514 | part_stat_read_all(hd, &stat); |
b2f609e1 CH |
1515 | if (queue_is_mq(gp->queue)) |
1516 | inflight = blk_mq_in_flight(gp->queue, hd); | |
1517 | else | |
1f06959b | 1518 | inflight = part_in_flight(hd); |
ea18e0f0 | 1519 | |
bdca3c87 MC |
1520 | seq_printf(seqf, "%4d %7d %s " |
1521 | "%lu %lu %lu %u " | |
1522 | "%lu %lu %lu %u " | |
1523 | "%u %u %u " | |
b6866318 KK |
1524 | "%lu %lu %lu %u " |
1525 | "%lu %u" | |
1526 | "\n", | |
f331c029 TH |
1527 | MAJOR(part_devt(hd)), MINOR(part_devt(hd)), |
1528 | disk_name(gp, hd->partno, buf), | |
ea18e0f0 KK |
1529 | stat.ios[STAT_READ], |
1530 | stat.merges[STAT_READ], | |
1531 | stat.sectors[STAT_READ], | |
1532 | (unsigned int)div_u64(stat.nsecs[STAT_READ], | |
1533 | NSEC_PER_MSEC), | |
1534 | stat.ios[STAT_WRITE], | |
1535 | stat.merges[STAT_WRITE], | |
1536 | stat.sectors[STAT_WRITE], | |
1537 | (unsigned int)div_u64(stat.nsecs[STAT_WRITE], | |
1538 | NSEC_PER_MSEC), | |
e016b782 | 1539 | inflight, |
ea18e0f0 | 1540 | jiffies_to_msecs(stat.io_ticks), |
8cd5b8fc KK |
1541 | (unsigned int)div_u64(stat.nsecs[STAT_READ] + |
1542 | stat.nsecs[STAT_WRITE] + | |
1543 | stat.nsecs[STAT_DISCARD] + | |
1544 | stat.nsecs[STAT_FLUSH], | |
1545 | NSEC_PER_MSEC), | |
ea18e0f0 KK |
1546 | stat.ios[STAT_DISCARD], |
1547 | stat.merges[STAT_DISCARD], | |
1548 | stat.sectors[STAT_DISCARD], | |
1549 | (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], | |
1550 | NSEC_PER_MSEC), | |
1551 | stat.ios[STAT_FLUSH], | |
1552 | (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], | |
1553 | NSEC_PER_MSEC) | |
28f39d55 | 1554 | ); |
1da177e4 | 1555 | } |
e71bf0d0 | 1556 | disk_part_iter_exit(&piter); |
9f5e4865 | 1557 | |
1da177e4 LT |
1558 | return 0; |
1559 | } | |
1560 | ||
31d85ab2 | 1561 | static const struct seq_operations diskstats_op = { |
def4e38d TH |
1562 | .start = disk_seqf_start, |
1563 | .next = disk_seqf_next, | |
1564 | .stop = disk_seqf_stop, | |
1da177e4 LT |
1565 | .show = diskstats_show |
1566 | }; | |
f500975a AD |
1567 | |
1568 | static int __init proc_genhd_init(void) | |
1569 | { | |
fddda2b7 CH |
1570 | proc_create_seq("diskstats", 0, NULL, &diskstats_op); |
1571 | proc_create_seq("partitions", 0, NULL, &partitions_op); | |
f500975a AD |
1572 | return 0; |
1573 | } | |
1574 | module_init(proc_genhd_init); | |
a6e2ba88 | 1575 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 1576 | |
cf771cb5 | 1577 | dev_t blk_lookup_devt(const char *name, int partno) |
a142be85 | 1578 | { |
def4e38d TH |
1579 | dev_t devt = MKDEV(0, 0); |
1580 | struct class_dev_iter iter; | |
1581 | struct device *dev; | |
a142be85 | 1582 | |
def4e38d TH |
1583 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); |
1584 | while ((dev = class_dev_iter_next(&iter))) { | |
a142be85 | 1585 | struct gendisk *disk = dev_to_disk(dev); |
548b10eb | 1586 | struct hd_struct *part; |
a142be85 | 1587 | |
3ada8b7e | 1588 | if (strcmp(dev_name(dev), name)) |
f331c029 | 1589 | continue; |
f331c029 | 1590 | |
41b8c853 NB |
1591 | if (partno < disk->minors) { |
1592 | /* We need to return the right devno, even | |
1593 | * if the partition doesn't exist yet. | |
1594 | */ | |
1595 | devt = MKDEV(MAJOR(dev->devt), | |
1596 | MINOR(dev->devt) + partno); | |
1597 | break; | |
1598 | } | |
548b10eb | 1599 | part = disk_get_part(disk, partno); |
2bbedcb4 | 1600 | if (part) { |
f331c029 | 1601 | devt = part_devt(part); |
e71bf0d0 | 1602 | disk_put_part(part); |
548b10eb | 1603 | break; |
def4e38d | 1604 | } |
548b10eb | 1605 | disk_put_part(part); |
5c0ef6d0 | 1606 | } |
def4e38d | 1607 | class_dev_iter_exit(&iter); |
edfaa7c3 KS |
1608 | return devt; |
1609 | } | |
edfaa7c3 | 1610 | |
e319e1fb | 1611 | struct gendisk *__alloc_disk_node(int minors, int node_id) |
1946089a CL |
1612 | { |
1613 | struct gendisk *disk; | |
6d2cf6f2 | 1614 | struct disk_part_tbl *ptbl; |
1946089a | 1615 | |
de65b012 CH |
1616 | if (minors > DISK_MAX_PARTS) { |
1617 | printk(KERN_ERR | |
7fb52621 | 1618 | "block: can't allocate more than %d partitions\n", |
de65b012 CH |
1619 | DISK_MAX_PARTS); |
1620 | minors = DISK_MAX_PARTS; | |
1621 | } | |
1946089a | 1622 | |
c1b511eb | 1623 | disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id); |
f93af2a4 CH |
1624 | if (!disk) |
1625 | return NULL; | |
6c23a968 | 1626 | |
22ae8ce8 CH |
1627 | disk->part0.bdev = bdev_alloc(disk, 0); |
1628 | if (!disk->part0.bdev) | |
1629 | goto out_free_disk; | |
1630 | ||
f93af2a4 CH |
1631 | disk->part0.dkstats = alloc_percpu(struct disk_stats); |
1632 | if (!disk->part0.dkstats) | |
22ae8ce8 | 1633 | goto out_bdput; |
b5d0b9df | 1634 | |
f93af2a4 | 1635 | disk->node_id = node_id; |
22ae8ce8 CH |
1636 | if (disk_expand_part_tbl(disk, 0)) |
1637 | goto out_free_bdstats; | |
f93af2a4 CH |
1638 | |
1639 | ptbl = rcu_dereference_protected(disk->part_tbl, 1); | |
1640 | rcu_assign_pointer(ptbl->part[0], &disk->part0); | |
1641 | ||
f93af2a4 | 1642 | if (hd_ref_init(&disk->part0)) |
22ae8ce8 | 1643 | goto out_free_bdstats; |
f93af2a4 CH |
1644 | |
1645 | disk->minors = minors; | |
1646 | rand_initialize_disk(disk); | |
1647 | disk_to_dev(disk)->class = &block_class; | |
1648 | disk_to_dev(disk)->type = &disk_type; | |
1649 | device_initialize(disk_to_dev(disk)); | |
1da177e4 | 1650 | return disk; |
f93af2a4 | 1651 | |
22ae8ce8 CH |
1652 | out_free_bdstats: |
1653 | free_percpu(disk->part0.dkstats); | |
1654 | out_bdput: | |
1655 | bdput(disk->part0.bdev); | |
f93af2a4 CH |
1656 | out_free_disk: |
1657 | kfree(disk); | |
1658 | return NULL; | |
1da177e4 | 1659 | } |
e319e1fb | 1660 | EXPORT_SYMBOL(__alloc_disk_node); |
1da177e4 | 1661 | |
b5bd357c LC |
1662 | /** |
1663 | * put_disk - decrements the gendisk refcount | |
0d20dcc2 | 1664 | * @disk: the struct gendisk to decrement the refcount for |
b5bd357c LC |
1665 | * |
1666 | * This decrements the refcount for the struct gendisk. When this reaches 0 | |
1667 | * we'll have disk_release() called. | |
e8c7d14a LC |
1668 | * |
1669 | * Context: Any context, but the last reference must not be dropped from | |
1670 | * atomic context. | |
b5bd357c | 1671 | */ |
1da177e4 LT |
1672 | void put_disk(struct gendisk *disk) |
1673 | { | |
1674 | if (disk) | |
efdc41c8 | 1675 | put_device(disk_to_dev(disk)); |
1da177e4 | 1676 | } |
1da177e4 LT |
1677 | EXPORT_SYMBOL(put_disk); |
1678 | ||
e3264a4d HR |
1679 | static void set_disk_ro_uevent(struct gendisk *gd, int ro) |
1680 | { | |
1681 | char event[] = "DISK_RO=1"; | |
1682 | char *envp[] = { event, NULL }; | |
1683 | ||
1684 | if (!ro) | |
1685 | event[8] = '0'; | |
1686 | kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); | |
1687 | } | |
1688 | ||
1da177e4 LT |
1689 | void set_disk_ro(struct gendisk *disk, int flag) |
1690 | { | |
e71bf0d0 TH |
1691 | struct disk_part_iter piter; |
1692 | struct hd_struct *part; | |
1693 | ||
e3264a4d HR |
1694 | if (disk->part0.policy != flag) { |
1695 | set_disk_ro_uevent(disk, flag); | |
1696 | disk->part0.policy = flag; | |
1697 | } | |
1698 | ||
1699 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY); | |
e71bf0d0 TH |
1700 | while ((part = disk_part_iter_next(&piter))) |
1701 | part->policy = flag; | |
1702 | disk_part_iter_exit(&piter); | |
1da177e4 LT |
1703 | } |
1704 | ||
1705 | EXPORT_SYMBOL(set_disk_ro); | |
1706 | ||
1707 | int bdev_read_only(struct block_device *bdev) | |
1708 | { | |
1709 | if (!bdev) | |
1710 | return 0; | |
b7db9956 | 1711 | return bdev->bd_part->policy; |
1da177e4 LT |
1712 | } |
1713 | ||
1714 | EXPORT_SYMBOL(bdev_read_only); | |
1715 | ||
77ea887e TH |
1716 | /* |
1717 | * Disk events - monitor disk events like media change and eject request. | |
1718 | */ | |
1719 | struct disk_events { | |
1720 | struct list_head node; /* all disk_event's */ | |
1721 | struct gendisk *disk; /* the associated disk */ | |
1722 | spinlock_t lock; | |
1723 | ||
fdd514e1 | 1724 | struct mutex block_mutex; /* protects blocking */ |
77ea887e TH |
1725 | int block; /* event blocking depth */ |
1726 | unsigned int pending; /* events already sent out */ | |
1727 | unsigned int clearing; /* events being cleared */ | |
1728 | ||
1729 | long poll_msecs; /* interval, -1 for default */ | |
1730 | struct delayed_work dwork; | |
1731 | }; | |
1732 | ||
1733 | static const char *disk_events_strs[] = { | |
1734 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change", | |
1735 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request", | |
1736 | }; | |
1737 | ||
1738 | static char *disk_uevents[] = { | |
1739 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1", | |
1740 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1", | |
1741 | }; | |
1742 | ||
1743 | /* list of all disk_events */ | |
1744 | static DEFINE_MUTEX(disk_events_mutex); | |
1745 | static LIST_HEAD(disk_events); | |
1746 | ||
1747 | /* disable in-kernel polling by default */ | |
1fe8f348 | 1748 | static unsigned long disk_events_dfl_poll_msecs; |
77ea887e TH |
1749 | |
1750 | static unsigned long disk_events_poll_jiffies(struct gendisk *disk) | |
1751 | { | |
1752 | struct disk_events *ev = disk->ev; | |
1753 | long intv_msecs = 0; | |
1754 | ||
1755 | /* | |
1756 | * If device-specific poll interval is set, always use it. If | |
673387a9 | 1757 | * the default is being used, poll if the POLL flag is set. |
77ea887e TH |
1758 | */ |
1759 | if (ev->poll_msecs >= 0) | |
1760 | intv_msecs = ev->poll_msecs; | |
c92e2f04 | 1761 | else if (disk->event_flags & DISK_EVENT_FLAG_POLL) |
77ea887e TH |
1762 | intv_msecs = disk_events_dfl_poll_msecs; |
1763 | ||
1764 | return msecs_to_jiffies(intv_msecs); | |
1765 | } | |
1766 | ||
c3af54af TH |
1767 | /** |
1768 | * disk_block_events - block and flush disk event checking | |
1769 | * @disk: disk to block events for | |
1770 | * | |
1771 | * On return from this function, it is guaranteed that event checking | |
1772 | * isn't in progress and won't happen until unblocked by | |
1773 | * disk_unblock_events(). Events blocking is counted and the actual | |
1774 | * unblocking happens after the matching number of unblocks are done. | |
1775 | * | |
1776 | * Note that this intentionally does not block event checking from | |
1777 | * disk_clear_events(). | |
1778 | * | |
1779 | * CONTEXT: | |
1780 | * Might sleep. | |
1781 | */ | |
1782 | void disk_block_events(struct gendisk *disk) | |
77ea887e TH |
1783 | { |
1784 | struct disk_events *ev = disk->ev; | |
1785 | unsigned long flags; | |
1786 | bool cancel; | |
1787 | ||
c3af54af TH |
1788 | if (!ev) |
1789 | return; | |
1790 | ||
fdd514e1 TH |
1791 | /* |
1792 | * Outer mutex ensures that the first blocker completes canceling | |
1793 | * the event work before further blockers are allowed to finish. | |
1794 | */ | |
1795 | mutex_lock(&ev->block_mutex); | |
1796 | ||
77ea887e TH |
1797 | spin_lock_irqsave(&ev->lock, flags); |
1798 | cancel = !ev->block++; | |
1799 | spin_unlock_irqrestore(&ev->lock, flags); | |
1800 | ||
c3af54af TH |
1801 | if (cancel) |
1802 | cancel_delayed_work_sync(&disk->ev->dwork); | |
fdd514e1 TH |
1803 | |
1804 | mutex_unlock(&ev->block_mutex); | |
77ea887e TH |
1805 | } |
1806 | ||
1807 | static void __disk_unblock_events(struct gendisk *disk, bool check_now) | |
1808 | { | |
1809 | struct disk_events *ev = disk->ev; | |
1810 | unsigned long intv; | |
1811 | unsigned long flags; | |
1812 | ||
1813 | spin_lock_irqsave(&ev->lock, flags); | |
1814 | ||
1815 | if (WARN_ON_ONCE(ev->block <= 0)) | |
1816 | goto out_unlock; | |
1817 | ||
1818 | if (--ev->block) | |
1819 | goto out_unlock; | |
1820 | ||
77ea887e | 1821 | intv = disk_events_poll_jiffies(disk); |
77ea887e | 1822 | if (check_now) |
695588f9 VK |
1823 | queue_delayed_work(system_freezable_power_efficient_wq, |
1824 | &ev->dwork, 0); | |
77ea887e | 1825 | else if (intv) |
695588f9 VK |
1826 | queue_delayed_work(system_freezable_power_efficient_wq, |
1827 | &ev->dwork, intv); | |
77ea887e TH |
1828 | out_unlock: |
1829 | spin_unlock_irqrestore(&ev->lock, flags); | |
1830 | } | |
1831 | ||
77ea887e TH |
1832 | /** |
1833 | * disk_unblock_events - unblock disk event checking | |
1834 | * @disk: disk to unblock events for | |
1835 | * | |
1836 | * Undo disk_block_events(). When the block count reaches zero, it | |
1837 | * starts events polling if configured. | |
1838 | * | |
1839 | * CONTEXT: | |
1840 | * Don't care. Safe to call from irq context. | |
1841 | */ | |
1842 | void disk_unblock_events(struct gendisk *disk) | |
1843 | { | |
1844 | if (disk->ev) | |
facc31dd | 1845 | __disk_unblock_events(disk, false); |
77ea887e TH |
1846 | } |
1847 | ||
1848 | /** | |
85ef06d1 TH |
1849 | * disk_flush_events - schedule immediate event checking and flushing |
1850 | * @disk: disk to check and flush events for | |
1851 | * @mask: events to flush | |
77ea887e | 1852 | * |
85ef06d1 TH |
1853 | * Schedule immediate event checking on @disk if not blocked. Events in |
1854 | * @mask are scheduled to be cleared from the driver. Note that this | |
1855 | * doesn't clear the events from @disk->ev. | |
77ea887e TH |
1856 | * |
1857 | * CONTEXT: | |
85ef06d1 | 1858 | * If @mask is non-zero must be called with bdev->bd_mutex held. |
77ea887e | 1859 | */ |
85ef06d1 | 1860 | void disk_flush_events(struct gendisk *disk, unsigned int mask) |
77ea887e | 1861 | { |
a9dce2a3 | 1862 | struct disk_events *ev = disk->ev; |
a9dce2a3 TH |
1863 | |
1864 | if (!ev) | |
1865 | return; | |
1866 | ||
85ef06d1 TH |
1867 | spin_lock_irq(&ev->lock); |
1868 | ev->clearing |= mask; | |
41f63c53 | 1869 | if (!ev->block) |
695588f9 VK |
1870 | mod_delayed_work(system_freezable_power_efficient_wq, |
1871 | &ev->dwork, 0); | |
85ef06d1 | 1872 | spin_unlock_irq(&ev->lock); |
77ea887e | 1873 | } |
77ea887e TH |
1874 | |
1875 | /** | |
1876 | * disk_clear_events - synchronously check, clear and return pending events | |
1877 | * @disk: disk to fetch and clear events from | |
da3dae54 | 1878 | * @mask: mask of events to be fetched and cleared |
77ea887e TH |
1879 | * |
1880 | * Disk events are synchronously checked and pending events in @mask | |
1881 | * are cleared and returned. This ignores the block count. | |
1882 | * | |
1883 | * CONTEXT: | |
1884 | * Might sleep. | |
1885 | */ | |
95f6f3a4 | 1886 | static unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask) |
77ea887e | 1887 | { |
77ea887e TH |
1888 | struct disk_events *ev = disk->ev; |
1889 | unsigned int pending; | |
12c2bdb2 | 1890 | unsigned int clearing = mask; |
77ea887e | 1891 | |
a564e23f | 1892 | if (!ev) |
77ea887e | 1893 | return 0; |
77ea887e | 1894 | |
12c2bdb2 DB |
1895 | disk_block_events(disk); |
1896 | ||
1897 | /* | |
1898 | * store the union of mask and ev->clearing on the stack so that the | |
1899 | * race with disk_flush_events does not cause ambiguity (ev->clearing | |
1900 | * can still be modified even if events are blocked). | |
1901 | */ | |
77ea887e | 1902 | spin_lock_irq(&ev->lock); |
12c2bdb2 DB |
1903 | clearing |= ev->clearing; |
1904 | ev->clearing = 0; | |
77ea887e TH |
1905 | spin_unlock_irq(&ev->lock); |
1906 | ||
12c2bdb2 | 1907 | disk_check_events(ev, &clearing); |
aea24a8b | 1908 | /* |
12c2bdb2 DB |
1909 | * if ev->clearing is not 0, the disk_flush_events got called in the |
1910 | * middle of this function, so we want to run the workfn without delay. | |
aea24a8b | 1911 | */ |
12c2bdb2 | 1912 | __disk_unblock_events(disk, ev->clearing ? true : false); |
77ea887e TH |
1913 | |
1914 | /* then, fetch and clear pending events */ | |
1915 | spin_lock_irq(&ev->lock); | |
77ea887e TH |
1916 | pending = ev->pending & mask; |
1917 | ev->pending &= ~mask; | |
1918 | spin_unlock_irq(&ev->lock); | |
12c2bdb2 | 1919 | WARN_ON_ONCE(clearing & mask); |
77ea887e TH |
1920 | |
1921 | return pending; | |
1922 | } | |
1923 | ||
95f6f3a4 CH |
1924 | /** |
1925 | * bdev_check_media_change - check if a removable media has been changed | |
1926 | * @bdev: block device to check | |
1927 | * | |
1928 | * Check whether a removable media has been changed, and attempt to free all | |
1929 | * dentries and inodes and invalidates all block device page cache entries in | |
1930 | * that case. | |
1931 | * | |
1932 | * Returns %true if the block device changed, or %false if not. | |
1933 | */ | |
1934 | bool bdev_check_media_change(struct block_device *bdev) | |
1935 | { | |
1936 | unsigned int events; | |
1937 | ||
1938 | events = disk_clear_events(bdev->bd_disk, DISK_EVENT_MEDIA_CHANGE | | |
1939 | DISK_EVENT_EJECT_REQUEST); | |
1940 | if (!(events & DISK_EVENT_MEDIA_CHANGE)) | |
1941 | return false; | |
1942 | ||
1943 | if (__invalidate_device(bdev, true)) | |
1944 | pr_warn("VFS: busy inodes on changed media %s\n", | |
1945 | bdev->bd_disk->disk_name); | |
38430f08 | 1946 | set_bit(GD_NEED_PART_SCAN, &bdev->bd_disk->state); |
95f6f3a4 CH |
1947 | return true; |
1948 | } | |
1949 | EXPORT_SYMBOL(bdev_check_media_change); | |
1950 | ||
12c2bdb2 DB |
1951 | /* |
1952 | * Separate this part out so that a different pointer for clearing_ptr can be | |
1953 | * passed in for disk_clear_events. | |
1954 | */ | |
77ea887e TH |
1955 | static void disk_events_workfn(struct work_struct *work) |
1956 | { | |
1957 | struct delayed_work *dwork = to_delayed_work(work); | |
1958 | struct disk_events *ev = container_of(dwork, struct disk_events, dwork); | |
12c2bdb2 DB |
1959 | |
1960 | disk_check_events(ev, &ev->clearing); | |
1961 | } | |
1962 | ||
1963 | static void disk_check_events(struct disk_events *ev, | |
1964 | unsigned int *clearing_ptr) | |
1965 | { | |
77ea887e TH |
1966 | struct gendisk *disk = ev->disk; |
1967 | char *envp[ARRAY_SIZE(disk_uevents) + 1] = { }; | |
12c2bdb2 | 1968 | unsigned int clearing = *clearing_ptr; |
77ea887e TH |
1969 | unsigned int events; |
1970 | unsigned long intv; | |
1971 | int nr_events = 0, i; | |
1972 | ||
1973 | /* check events */ | |
1974 | events = disk->fops->check_events(disk, clearing); | |
1975 | ||
1976 | /* accumulate pending events and schedule next poll if necessary */ | |
1977 | spin_lock_irq(&ev->lock); | |
1978 | ||
1979 | events &= ~ev->pending; | |
1980 | ev->pending |= events; | |
12c2bdb2 | 1981 | *clearing_ptr &= ~clearing; |
77ea887e TH |
1982 | |
1983 | intv = disk_events_poll_jiffies(disk); | |
1984 | if (!ev->block && intv) | |
695588f9 VK |
1985 | queue_delayed_work(system_freezable_power_efficient_wq, |
1986 | &ev->dwork, intv); | |
77ea887e TH |
1987 | |
1988 | spin_unlock_irq(&ev->lock); | |
1989 | ||
7c88a168 TH |
1990 | /* |
1991 | * Tell userland about new events. Only the events listed in | |
c92e2f04 MW |
1992 | * @disk->events are reported, and only if DISK_EVENT_FLAG_UEVENT |
1993 | * is set. Otherwise, events are processed internally but never | |
1994 | * get reported to userland. | |
7c88a168 | 1995 | */ |
77ea887e | 1996 | for (i = 0; i < ARRAY_SIZE(disk_uevents); i++) |
c92e2f04 MW |
1997 | if ((events & disk->events & (1 << i)) && |
1998 | (disk->event_flags & DISK_EVENT_FLAG_UEVENT)) | |
77ea887e TH |
1999 | envp[nr_events++] = disk_uevents[i]; |
2000 | ||
2001 | if (nr_events) | |
2002 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
2003 | } | |
2004 | ||
2005 | /* | |
2006 | * A disk events enabled device has the following sysfs nodes under | |
2007 | * its /sys/block/X/ directory. | |
2008 | * | |
2009 | * events : list of all supported events | |
2010 | * events_async : list of events which can be detected w/o polling | |
673387a9 | 2011 | * (always empty, only for backwards compatibility) |
77ea887e TH |
2012 | * events_poll_msecs : polling interval, 0: disable, -1: system default |
2013 | */ | |
2014 | static ssize_t __disk_events_show(unsigned int events, char *buf) | |
2015 | { | |
2016 | const char *delim = ""; | |
2017 | ssize_t pos = 0; | |
2018 | int i; | |
2019 | ||
2020 | for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++) | |
2021 | if (events & (1 << i)) { | |
2022 | pos += sprintf(buf + pos, "%s%s", | |
2023 | delim, disk_events_strs[i]); | |
2024 | delim = " "; | |
2025 | } | |
2026 | if (pos) | |
2027 | pos += sprintf(buf + pos, "\n"); | |
2028 | return pos; | |
2029 | } | |
2030 | ||
2031 | static ssize_t disk_events_show(struct device *dev, | |
2032 | struct device_attribute *attr, char *buf) | |
2033 | { | |
2034 | struct gendisk *disk = dev_to_disk(dev); | |
2035 | ||
c92e2f04 MW |
2036 | if (!(disk->event_flags & DISK_EVENT_FLAG_UEVENT)) |
2037 | return 0; | |
2038 | ||
77ea887e TH |
2039 | return __disk_events_show(disk->events, buf); |
2040 | } | |
2041 | ||
2042 | static ssize_t disk_events_async_show(struct device *dev, | |
2043 | struct device_attribute *attr, char *buf) | |
2044 | { | |
673387a9 | 2045 | return 0; |
77ea887e TH |
2046 | } |
2047 | ||
2048 | static ssize_t disk_events_poll_msecs_show(struct device *dev, | |
2049 | struct device_attribute *attr, | |
2050 | char *buf) | |
2051 | { | |
2052 | struct gendisk *disk = dev_to_disk(dev); | |
2053 | ||
cdf3e3de MW |
2054 | if (!disk->ev) |
2055 | return sprintf(buf, "-1\n"); | |
2056 | ||
77ea887e TH |
2057 | return sprintf(buf, "%ld\n", disk->ev->poll_msecs); |
2058 | } | |
2059 | ||
2060 | static ssize_t disk_events_poll_msecs_store(struct device *dev, | |
2061 | struct device_attribute *attr, | |
2062 | const char *buf, size_t count) | |
2063 | { | |
2064 | struct gendisk *disk = dev_to_disk(dev); | |
2065 | long intv; | |
2066 | ||
2067 | if (!count || !sscanf(buf, "%ld", &intv)) | |
2068 | return -EINVAL; | |
2069 | ||
2070 | if (intv < 0 && intv != -1) | |
2071 | return -EINVAL; | |
2072 | ||
cdf3e3de MW |
2073 | if (!disk->ev) |
2074 | return -ENODEV; | |
2075 | ||
c3af54af | 2076 | disk_block_events(disk); |
77ea887e TH |
2077 | disk->ev->poll_msecs = intv; |
2078 | __disk_unblock_events(disk, true); | |
2079 | ||
2080 | return count; | |
2081 | } | |
2082 | ||
5657a819 JP |
2083 | static const DEVICE_ATTR(events, 0444, disk_events_show, NULL); |
2084 | static const DEVICE_ATTR(events_async, 0444, disk_events_async_show, NULL); | |
2085 | static const DEVICE_ATTR(events_poll_msecs, 0644, | |
77ea887e TH |
2086 | disk_events_poll_msecs_show, |
2087 | disk_events_poll_msecs_store); | |
2088 | ||
2089 | static const struct attribute *disk_events_attrs[] = { | |
2090 | &dev_attr_events.attr, | |
2091 | &dev_attr_events_async.attr, | |
2092 | &dev_attr_events_poll_msecs.attr, | |
2093 | NULL, | |
2094 | }; | |
2095 | ||
2096 | /* | |
2097 | * The default polling interval can be specified by the kernel | |
2098 | * parameter block.events_dfl_poll_msecs which defaults to 0 | |
2099 | * (disable). This can also be modified runtime by writing to | |
1624b0b2 | 2100 | * /sys/module/block/parameters/events_dfl_poll_msecs. |
77ea887e TH |
2101 | */ |
2102 | static int disk_events_set_dfl_poll_msecs(const char *val, | |
2103 | const struct kernel_param *kp) | |
2104 | { | |
2105 | struct disk_events *ev; | |
2106 | int ret; | |
2107 | ||
2108 | ret = param_set_ulong(val, kp); | |
2109 | if (ret < 0) | |
2110 | return ret; | |
2111 | ||
2112 | mutex_lock(&disk_events_mutex); | |
2113 | ||
2114 | list_for_each_entry(ev, &disk_events, node) | |
85ef06d1 | 2115 | disk_flush_events(ev->disk, 0); |
77ea887e TH |
2116 | |
2117 | mutex_unlock(&disk_events_mutex); | |
2118 | ||
2119 | return 0; | |
2120 | } | |
2121 | ||
2122 | static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = { | |
2123 | .set = disk_events_set_dfl_poll_msecs, | |
2124 | .get = param_get_ulong, | |
2125 | }; | |
2126 | ||
2127 | #undef MODULE_PARAM_PREFIX | |
2128 | #define MODULE_PARAM_PREFIX "block." | |
2129 | ||
2130 | module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops, | |
2131 | &disk_events_dfl_poll_msecs, 0644); | |
2132 | ||
2133 | /* | |
9f53d2fe | 2134 | * disk_{alloc|add|del|release}_events - initialize and destroy disk_events. |
77ea887e | 2135 | */ |
9f53d2fe | 2136 | static void disk_alloc_events(struct gendisk *disk) |
77ea887e TH |
2137 | { |
2138 | struct disk_events *ev; | |
2139 | ||
cdf3e3de | 2140 | if (!disk->fops->check_events || !disk->events) |
77ea887e TH |
2141 | return; |
2142 | ||
2143 | ev = kzalloc(sizeof(*ev), GFP_KERNEL); | |
2144 | if (!ev) { | |
2145 | pr_warn("%s: failed to initialize events\n", disk->disk_name); | |
2146 | return; | |
2147 | } | |
2148 | ||
77ea887e TH |
2149 | INIT_LIST_HEAD(&ev->node); |
2150 | ev->disk = disk; | |
2151 | spin_lock_init(&ev->lock); | |
fdd514e1 | 2152 | mutex_init(&ev->block_mutex); |
77ea887e TH |
2153 | ev->block = 1; |
2154 | ev->poll_msecs = -1; | |
2155 | INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn); | |
2156 | ||
9f53d2fe SG |
2157 | disk->ev = ev; |
2158 | } | |
2159 | ||
2160 | static void disk_add_events(struct gendisk *disk) | |
2161 | { | |
9f53d2fe SG |
2162 | /* FIXME: error handling */ |
2163 | if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0) | |
2164 | pr_warn("%s: failed to create sysfs files for events\n", | |
2165 | disk->disk_name); | |
2166 | ||
cdf3e3de MW |
2167 | if (!disk->ev) |
2168 | return; | |
2169 | ||
77ea887e | 2170 | mutex_lock(&disk_events_mutex); |
9f53d2fe | 2171 | list_add_tail(&disk->ev->node, &disk_events); |
77ea887e TH |
2172 | mutex_unlock(&disk_events_mutex); |
2173 | ||
2174 | /* | |
2175 | * Block count is initialized to 1 and the following initial | |
2176 | * unblock kicks it into action. | |
2177 | */ | |
2178 | __disk_unblock_events(disk, true); | |
2179 | } | |
2180 | ||
2181 | static void disk_del_events(struct gendisk *disk) | |
2182 | { | |
cdf3e3de MW |
2183 | if (disk->ev) { |
2184 | disk_block_events(disk); | |
77ea887e | 2185 | |
cdf3e3de MW |
2186 | mutex_lock(&disk_events_mutex); |
2187 | list_del_init(&disk->ev->node); | |
2188 | mutex_unlock(&disk_events_mutex); | |
2189 | } | |
77ea887e TH |
2190 | |
2191 | sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs); | |
2192 | } | |
2193 | ||
2194 | static void disk_release_events(struct gendisk *disk) | |
2195 | { | |
2196 | /* the block count should be 1 from disk_del_events() */ | |
2197 | WARN_ON_ONCE(disk->ev && disk->ev->block != 1); | |
2198 | kfree(disk->ev); | |
2199 | } |