Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * gendisk handling | |
3 | */ | |
4 | ||
1da177e4 LT |
5 | #include <linux/module.h> |
6 | #include <linux/fs.h> | |
7 | #include <linux/genhd.h> | |
b446b60e | 8 | #include <linux/kdev_t.h> |
1da177e4 LT |
9 | #include <linux/kernel.h> |
10 | #include <linux/blkdev.h> | |
66114cad | 11 | #include <linux/backing-dev.h> |
1da177e4 LT |
12 | #include <linux/init.h> |
13 | #include <linux/spinlock.h> | |
f500975a | 14 | #include <linux/proc_fs.h> |
1da177e4 LT |
15 | #include <linux/seq_file.h> |
16 | #include <linux/slab.h> | |
17 | #include <linux/kmod.h> | |
18 | #include <linux/kobj_map.h> | |
58383af6 | 19 | #include <linux/mutex.h> |
bcce3de1 | 20 | #include <linux/idr.h> |
77ea887e | 21 | #include <linux/log2.h> |
25e823c8 | 22 | #include <linux/pm_runtime.h> |
99e6608c | 23 | #include <linux/badblocks.h> |
1da177e4 | 24 | |
ff88972c AB |
25 | #include "blk.h" |
26 | ||
edfaa7c3 | 27 | static DEFINE_MUTEX(block_class_lock); |
edfaa7c3 | 28 | struct kobject *block_depr; |
1da177e4 | 29 | |
bcce3de1 | 30 | /* for extended dynamic devt allocation, currently only one major is used */ |
ce23bba8 | 31 | #define NR_EXT_DEVT (1 << MINORBITS) |
bcce3de1 | 32 | |
2da78092 | 33 | /* For extended devt allocation. ext_devt_lock prevents look up |
bcce3de1 TH |
34 | * results from going away underneath its user. |
35 | */ | |
2da78092 | 36 | static DEFINE_SPINLOCK(ext_devt_lock); |
bcce3de1 TH |
37 | static DEFINE_IDR(ext_devt_idr); |
38 | ||
edf8ff55 | 39 | static const struct device_type disk_type; |
1826eadf | 40 | |
12c2bdb2 DB |
41 | static void disk_check_events(struct disk_events *ev, |
42 | unsigned int *clearing_ptr); | |
9f53d2fe | 43 | static void disk_alloc_events(struct gendisk *disk); |
77ea887e TH |
44 | static void disk_add_events(struct gendisk *disk); |
45 | static void disk_del_events(struct gendisk *disk); | |
46 | static void disk_release_events(struct gendisk *disk); | |
47 | ||
f299b7c7 JA |
48 | void part_inc_in_flight(struct request_queue *q, struct hd_struct *part, int rw) |
49 | { | |
50 | if (q->mq_ops) | |
51 | return; | |
52 | ||
53 | atomic_inc(&part->in_flight[rw]); | |
54 | if (part->partno) | |
55 | atomic_inc(&part_to_disk(part)->part0.in_flight[rw]); | |
56 | } | |
57 | ||
58 | void part_dec_in_flight(struct request_queue *q, struct hd_struct *part, int rw) | |
59 | { | |
60 | if (q->mq_ops) | |
61 | return; | |
62 | ||
63 | atomic_dec(&part->in_flight[rw]); | |
64 | if (part->partno) | |
65 | atomic_dec(&part_to_disk(part)->part0.in_flight[rw]); | |
66 | } | |
67 | ||
68 | void part_in_flight(struct request_queue *q, struct hd_struct *part, | |
69 | unsigned int inflight[2]) | |
70 | { | |
71 | if (q->mq_ops) { | |
72 | blk_mq_in_flight(q, part, inflight); | |
73 | return; | |
74 | } | |
75 | ||
76 | inflight[0] = atomic_read(&part->in_flight[0]) + | |
77 | atomic_read(&part->in_flight[1]); | |
78 | if (part->partno) { | |
79 | part = &part_to_disk(part)->part0; | |
80 | inflight[1] = atomic_read(&part->in_flight[0]) + | |
81 | atomic_read(&part->in_flight[1]); | |
82 | } | |
83 | } | |
84 | ||
e71bf0d0 TH |
85 | /** |
86 | * disk_get_part - get partition | |
87 | * @disk: disk to look partition from | |
88 | * @partno: partition number | |
89 | * | |
90 | * Look for partition @partno from @disk. If found, increment | |
91 | * reference count and return it. | |
92 | * | |
93 | * CONTEXT: | |
94 | * Don't care. | |
95 | * | |
96 | * RETURNS: | |
97 | * Pointer to the found partition on success, NULL if not found. | |
98 | */ | |
99 | struct hd_struct *disk_get_part(struct gendisk *disk, int partno) | |
100 | { | |
540eed56 TH |
101 | struct hd_struct *part = NULL; |
102 | struct disk_part_tbl *ptbl; | |
e71bf0d0 | 103 | |
540eed56 | 104 | if (unlikely(partno < 0)) |
e71bf0d0 | 105 | return NULL; |
540eed56 | 106 | |
e71bf0d0 | 107 | rcu_read_lock(); |
540eed56 TH |
108 | |
109 | ptbl = rcu_dereference(disk->part_tbl); | |
110 | if (likely(partno < ptbl->len)) { | |
111 | part = rcu_dereference(ptbl->part[partno]); | |
112 | if (part) | |
113 | get_device(part_to_dev(part)); | |
114 | } | |
115 | ||
e71bf0d0 TH |
116 | rcu_read_unlock(); |
117 | ||
118 | return part; | |
119 | } | |
120 | EXPORT_SYMBOL_GPL(disk_get_part); | |
121 | ||
122 | /** | |
123 | * disk_part_iter_init - initialize partition iterator | |
124 | * @piter: iterator to initialize | |
125 | * @disk: disk to iterate over | |
126 | * @flags: DISK_PITER_* flags | |
127 | * | |
128 | * Initialize @piter so that it iterates over partitions of @disk. | |
129 | * | |
130 | * CONTEXT: | |
131 | * Don't care. | |
132 | */ | |
133 | void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk, | |
134 | unsigned int flags) | |
135 | { | |
540eed56 TH |
136 | struct disk_part_tbl *ptbl; |
137 | ||
138 | rcu_read_lock(); | |
139 | ptbl = rcu_dereference(disk->part_tbl); | |
140 | ||
e71bf0d0 TH |
141 | piter->disk = disk; |
142 | piter->part = NULL; | |
143 | ||
144 | if (flags & DISK_PITER_REVERSE) | |
540eed56 | 145 | piter->idx = ptbl->len - 1; |
71982a40 | 146 | else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0)) |
e71bf0d0 | 147 | piter->idx = 0; |
b5d0b9df TH |
148 | else |
149 | piter->idx = 1; | |
e71bf0d0 TH |
150 | |
151 | piter->flags = flags; | |
540eed56 TH |
152 | |
153 | rcu_read_unlock(); | |
e71bf0d0 TH |
154 | } |
155 | EXPORT_SYMBOL_GPL(disk_part_iter_init); | |
156 | ||
157 | /** | |
158 | * disk_part_iter_next - proceed iterator to the next partition and return it | |
159 | * @piter: iterator of interest | |
160 | * | |
161 | * Proceed @piter to the next partition and return it. | |
162 | * | |
163 | * CONTEXT: | |
164 | * Don't care. | |
165 | */ | |
166 | struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter) | |
167 | { | |
540eed56 | 168 | struct disk_part_tbl *ptbl; |
e71bf0d0 TH |
169 | int inc, end; |
170 | ||
171 | /* put the last partition */ | |
172 | disk_put_part(piter->part); | |
173 | piter->part = NULL; | |
174 | ||
540eed56 | 175 | /* get part_tbl */ |
e71bf0d0 | 176 | rcu_read_lock(); |
540eed56 | 177 | ptbl = rcu_dereference(piter->disk->part_tbl); |
e71bf0d0 TH |
178 | |
179 | /* determine iteration parameters */ | |
180 | if (piter->flags & DISK_PITER_REVERSE) { | |
181 | inc = -1; | |
71982a40 TH |
182 | if (piter->flags & (DISK_PITER_INCL_PART0 | |
183 | DISK_PITER_INCL_EMPTY_PART0)) | |
b5d0b9df TH |
184 | end = -1; |
185 | else | |
186 | end = 0; | |
e71bf0d0 TH |
187 | } else { |
188 | inc = 1; | |
540eed56 | 189 | end = ptbl->len; |
e71bf0d0 TH |
190 | } |
191 | ||
192 | /* iterate to the next partition */ | |
193 | for (; piter->idx != end; piter->idx += inc) { | |
194 | struct hd_struct *part; | |
195 | ||
540eed56 | 196 | part = rcu_dereference(ptbl->part[piter->idx]); |
e71bf0d0 TH |
197 | if (!part) |
198 | continue; | |
c83f6bf9 | 199 | if (!part_nr_sects_read(part) && |
71982a40 TH |
200 | !(piter->flags & DISK_PITER_INCL_EMPTY) && |
201 | !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 && | |
202 | piter->idx == 0)) | |
e71bf0d0 TH |
203 | continue; |
204 | ||
ed9e1982 | 205 | get_device(part_to_dev(part)); |
e71bf0d0 TH |
206 | piter->part = part; |
207 | piter->idx += inc; | |
208 | break; | |
209 | } | |
210 | ||
211 | rcu_read_unlock(); | |
212 | ||
213 | return piter->part; | |
214 | } | |
215 | EXPORT_SYMBOL_GPL(disk_part_iter_next); | |
216 | ||
217 | /** | |
218 | * disk_part_iter_exit - finish up partition iteration | |
219 | * @piter: iter of interest | |
220 | * | |
221 | * Called when iteration is over. Cleans up @piter. | |
222 | * | |
223 | * CONTEXT: | |
224 | * Don't care. | |
225 | */ | |
226 | void disk_part_iter_exit(struct disk_part_iter *piter) | |
227 | { | |
228 | disk_put_part(piter->part); | |
229 | piter->part = NULL; | |
230 | } | |
231 | EXPORT_SYMBOL_GPL(disk_part_iter_exit); | |
232 | ||
a6f23657 JA |
233 | static inline int sector_in_part(struct hd_struct *part, sector_t sector) |
234 | { | |
235 | return part->start_sect <= sector && | |
c83f6bf9 | 236 | sector < part->start_sect + part_nr_sects_read(part); |
a6f23657 JA |
237 | } |
238 | ||
e71bf0d0 TH |
239 | /** |
240 | * disk_map_sector_rcu - map sector to partition | |
241 | * @disk: gendisk of interest | |
242 | * @sector: sector to map | |
243 | * | |
244 | * Find out which partition @sector maps to on @disk. This is | |
245 | * primarily used for stats accounting. | |
246 | * | |
247 | * CONTEXT: | |
248 | * RCU read locked. The returned partition pointer is valid only | |
249 | * while preemption is disabled. | |
250 | * | |
251 | * RETURNS: | |
074a7aca | 252 | * Found partition on success, part0 is returned if no partition matches |
e71bf0d0 TH |
253 | */ |
254 | struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector) | |
255 | { | |
540eed56 | 256 | struct disk_part_tbl *ptbl; |
a6f23657 | 257 | struct hd_struct *part; |
e71bf0d0 TH |
258 | int i; |
259 | ||
540eed56 TH |
260 | ptbl = rcu_dereference(disk->part_tbl); |
261 | ||
a6f23657 JA |
262 | part = rcu_dereference(ptbl->last_lookup); |
263 | if (part && sector_in_part(part, sector)) | |
264 | return part; | |
265 | ||
540eed56 | 266 | for (i = 1; i < ptbl->len; i++) { |
a6f23657 | 267 | part = rcu_dereference(ptbl->part[i]); |
e71bf0d0 | 268 | |
a6f23657 JA |
269 | if (part && sector_in_part(part, sector)) { |
270 | rcu_assign_pointer(ptbl->last_lookup, part); | |
e71bf0d0 | 271 | return part; |
a6f23657 | 272 | } |
e71bf0d0 | 273 | } |
074a7aca | 274 | return &disk->part0; |
e71bf0d0 TH |
275 | } |
276 | EXPORT_SYMBOL_GPL(disk_map_sector_rcu); | |
277 | ||
1da177e4 LT |
278 | /* |
279 | * Can be deleted altogether. Later. | |
280 | * | |
281 | */ | |
282 | static struct blk_major_name { | |
283 | struct blk_major_name *next; | |
284 | int major; | |
285 | char name[16]; | |
68eef3b4 | 286 | } *major_names[BLKDEV_MAJOR_HASH_SIZE]; |
1da177e4 LT |
287 | |
288 | /* index in the above - for now: assume no multimajor ranges */ | |
e61eb2e9 | 289 | static inline int major_to_index(unsigned major) |
1da177e4 | 290 | { |
68eef3b4 | 291 | return major % BLKDEV_MAJOR_HASH_SIZE; |
7170be5f NH |
292 | } |
293 | ||
68eef3b4 | 294 | #ifdef CONFIG_PROC_FS |
cf771cb5 | 295 | void blkdev_show(struct seq_file *seqf, off_t offset) |
7170be5f | 296 | { |
68eef3b4 | 297 | struct blk_major_name *dp; |
7170be5f | 298 | |
68eef3b4 | 299 | if (offset < BLKDEV_MAJOR_HASH_SIZE) { |
edfaa7c3 | 300 | mutex_lock(&block_class_lock); |
68eef3b4 | 301 | for (dp = major_names[offset]; dp; dp = dp->next) |
cf771cb5 | 302 | seq_printf(seqf, "%3d %s\n", dp->major, dp->name); |
edfaa7c3 | 303 | mutex_unlock(&block_class_lock); |
1da177e4 | 304 | } |
1da177e4 | 305 | } |
68eef3b4 | 306 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 307 | |
9e8c0bcc MN |
308 | /** |
309 | * register_blkdev - register a new block device | |
310 | * | |
0e056eb5 | 311 | * @major: the requested major device number [1..255]. If @major = 0, try to |
9e8c0bcc MN |
312 | * allocate any unused major number. |
313 | * @name: the name of the new block device as a zero terminated string | |
314 | * | |
315 | * The @name must be unique within the system. | |
316 | * | |
0e056eb5 | 317 | * The return value depends on the @major input parameter: |
318 | * | |
9e8c0bcc MN |
319 | * - if a major device number was requested in range [1..255] then the |
320 | * function returns zero on success, or a negative error code | |
0e056eb5 | 321 | * - if any unused major number was requested with @major = 0 parameter |
9e8c0bcc MN |
322 | * then the return value is the allocated major number in range |
323 | * [1..255] or a negative error code otherwise | |
324 | */ | |
1da177e4 LT |
325 | int register_blkdev(unsigned int major, const char *name) |
326 | { | |
327 | struct blk_major_name **n, *p; | |
328 | int index, ret = 0; | |
329 | ||
edfaa7c3 | 330 | mutex_lock(&block_class_lock); |
1da177e4 LT |
331 | |
332 | /* temporary */ | |
333 | if (major == 0) { | |
334 | for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) { | |
335 | if (major_names[index] == NULL) | |
336 | break; | |
337 | } | |
338 | ||
339 | if (index == 0) { | |
340 | printk("register_blkdev: failed to get major for %s\n", | |
341 | name); | |
342 | ret = -EBUSY; | |
343 | goto out; | |
344 | } | |
345 | major = index; | |
346 | ret = major; | |
347 | } | |
348 | ||
349 | p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL); | |
350 | if (p == NULL) { | |
351 | ret = -ENOMEM; | |
352 | goto out; | |
353 | } | |
354 | ||
355 | p->major = major; | |
356 | strlcpy(p->name, name, sizeof(p->name)); | |
357 | p->next = NULL; | |
358 | index = major_to_index(major); | |
359 | ||
360 | for (n = &major_names[index]; *n; n = &(*n)->next) { | |
361 | if ((*n)->major == major) | |
362 | break; | |
363 | } | |
364 | if (!*n) | |
365 | *n = p; | |
366 | else | |
367 | ret = -EBUSY; | |
368 | ||
369 | if (ret < 0) { | |
370 | printk("register_blkdev: cannot get major %d for %s\n", | |
371 | major, name); | |
372 | kfree(p); | |
373 | } | |
374 | out: | |
edfaa7c3 | 375 | mutex_unlock(&block_class_lock); |
1da177e4 LT |
376 | return ret; |
377 | } | |
378 | ||
379 | EXPORT_SYMBOL(register_blkdev); | |
380 | ||
f4480240 | 381 | void unregister_blkdev(unsigned int major, const char *name) |
1da177e4 LT |
382 | { |
383 | struct blk_major_name **n; | |
384 | struct blk_major_name *p = NULL; | |
385 | int index = major_to_index(major); | |
1da177e4 | 386 | |
edfaa7c3 | 387 | mutex_lock(&block_class_lock); |
1da177e4 LT |
388 | for (n = &major_names[index]; *n; n = &(*n)->next) |
389 | if ((*n)->major == major) | |
390 | break; | |
294462a5 AM |
391 | if (!*n || strcmp((*n)->name, name)) { |
392 | WARN_ON(1); | |
294462a5 | 393 | } else { |
1da177e4 LT |
394 | p = *n; |
395 | *n = p->next; | |
396 | } | |
edfaa7c3 | 397 | mutex_unlock(&block_class_lock); |
1da177e4 | 398 | kfree(p); |
1da177e4 LT |
399 | } |
400 | ||
401 | EXPORT_SYMBOL(unregister_blkdev); | |
402 | ||
403 | static struct kobj_map *bdev_map; | |
404 | ||
870d6656 TH |
405 | /** |
406 | * blk_mangle_minor - scatter minor numbers apart | |
407 | * @minor: minor number to mangle | |
408 | * | |
409 | * Scatter consecutively allocated @minor number apart if MANGLE_DEVT | |
410 | * is enabled. Mangling twice gives the original value. | |
411 | * | |
412 | * RETURNS: | |
413 | * Mangled value. | |
414 | * | |
415 | * CONTEXT: | |
416 | * Don't care. | |
417 | */ | |
418 | static int blk_mangle_minor(int minor) | |
419 | { | |
420 | #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT | |
421 | int i; | |
422 | ||
423 | for (i = 0; i < MINORBITS / 2; i++) { | |
424 | int low = minor & (1 << i); | |
425 | int high = minor & (1 << (MINORBITS - 1 - i)); | |
426 | int distance = MINORBITS - 1 - 2 * i; | |
427 | ||
428 | minor ^= low | high; /* clear both bits */ | |
429 | low <<= distance; /* swap the positions */ | |
430 | high >>= distance; | |
431 | minor |= low | high; /* and set */ | |
432 | } | |
433 | #endif | |
434 | return minor; | |
435 | } | |
436 | ||
bcce3de1 TH |
437 | /** |
438 | * blk_alloc_devt - allocate a dev_t for a partition | |
439 | * @part: partition to allocate dev_t for | |
bcce3de1 TH |
440 | * @devt: out parameter for resulting dev_t |
441 | * | |
442 | * Allocate a dev_t for block device. | |
443 | * | |
444 | * RETURNS: | |
445 | * 0 on success, allocated dev_t is returned in *@devt. -errno on | |
446 | * failure. | |
447 | * | |
448 | * CONTEXT: | |
449 | * Might sleep. | |
450 | */ | |
451 | int blk_alloc_devt(struct hd_struct *part, dev_t *devt) | |
452 | { | |
453 | struct gendisk *disk = part_to_disk(part); | |
bab998d6 | 454 | int idx; |
bcce3de1 TH |
455 | |
456 | /* in consecutive minor range? */ | |
457 | if (part->partno < disk->minors) { | |
458 | *devt = MKDEV(disk->major, disk->first_minor + part->partno); | |
459 | return 0; | |
460 | } | |
461 | ||
462 | /* allocate ext devt */ | |
2da78092 KB |
463 | idr_preload(GFP_KERNEL); |
464 | ||
4d66e5e9 | 465 | spin_lock_bh(&ext_devt_lock); |
2da78092 | 466 | idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_NOWAIT); |
4d66e5e9 | 467 | spin_unlock_bh(&ext_devt_lock); |
2da78092 KB |
468 | |
469 | idr_preload_end(); | |
bab998d6 TH |
470 | if (idx < 0) |
471 | return idx == -ENOSPC ? -EBUSY : idx; | |
bcce3de1 | 472 | |
870d6656 | 473 | *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx)); |
bcce3de1 TH |
474 | return 0; |
475 | } | |
476 | ||
477 | /** | |
478 | * blk_free_devt - free a dev_t | |
479 | * @devt: dev_t to free | |
480 | * | |
481 | * Free @devt which was allocated using blk_alloc_devt(). | |
482 | * | |
483 | * CONTEXT: | |
484 | * Might sleep. | |
485 | */ | |
486 | void blk_free_devt(dev_t devt) | |
487 | { | |
bcce3de1 TH |
488 | if (devt == MKDEV(0, 0)) |
489 | return; | |
490 | ||
491 | if (MAJOR(devt) == BLOCK_EXT_MAJOR) { | |
4d66e5e9 | 492 | spin_lock_bh(&ext_devt_lock); |
870d6656 | 493 | idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); |
4d66e5e9 | 494 | spin_unlock_bh(&ext_devt_lock); |
bcce3de1 TH |
495 | } |
496 | } | |
497 | ||
1f014290 TH |
498 | static char *bdevt_str(dev_t devt, char *buf) |
499 | { | |
500 | if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) { | |
501 | char tbuf[BDEVT_SIZE]; | |
502 | snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt)); | |
503 | snprintf(buf, BDEVT_SIZE, "%-9s", tbuf); | |
504 | } else | |
505 | snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt)); | |
506 | ||
507 | return buf; | |
508 | } | |
509 | ||
1da177e4 LT |
510 | /* |
511 | * Register device numbers dev..(dev+range-1) | |
512 | * range must be nonzero | |
513 | * The hash chain is sorted on range, so that subranges can override. | |
514 | */ | |
edfaa7c3 | 515 | void blk_register_region(dev_t devt, unsigned long range, struct module *module, |
1da177e4 LT |
516 | struct kobject *(*probe)(dev_t, int *, void *), |
517 | int (*lock)(dev_t, void *), void *data) | |
518 | { | |
edfaa7c3 | 519 | kobj_map(bdev_map, devt, range, module, probe, lock, data); |
1da177e4 LT |
520 | } |
521 | ||
522 | EXPORT_SYMBOL(blk_register_region); | |
523 | ||
edfaa7c3 | 524 | void blk_unregister_region(dev_t devt, unsigned long range) |
1da177e4 | 525 | { |
edfaa7c3 | 526 | kobj_unmap(bdev_map, devt, range); |
1da177e4 LT |
527 | } |
528 | ||
529 | EXPORT_SYMBOL(blk_unregister_region); | |
530 | ||
cf771cb5 | 531 | static struct kobject *exact_match(dev_t devt, int *partno, void *data) |
1da177e4 LT |
532 | { |
533 | struct gendisk *p = data; | |
edfaa7c3 | 534 | |
ed9e1982 | 535 | return &disk_to_dev(p)->kobj; |
1da177e4 LT |
536 | } |
537 | ||
edfaa7c3 | 538 | static int exact_lock(dev_t devt, void *data) |
1da177e4 LT |
539 | { |
540 | struct gendisk *p = data; | |
541 | ||
542 | if (!get_disk(p)) | |
543 | return -1; | |
544 | return 0; | |
545 | } | |
546 | ||
e63a46be | 547 | static void register_disk(struct device *parent, struct gendisk *disk) |
d2bf1b67 TH |
548 | { |
549 | struct device *ddev = disk_to_dev(disk); | |
550 | struct block_device *bdev; | |
551 | struct disk_part_iter piter; | |
552 | struct hd_struct *part; | |
553 | int err; | |
554 | ||
e63a46be | 555 | ddev->parent = parent; |
d2bf1b67 | 556 | |
ffc8b308 | 557 | dev_set_name(ddev, "%s", disk->disk_name); |
d2bf1b67 TH |
558 | |
559 | /* delay uevents, until we scanned partition table */ | |
560 | dev_set_uevent_suppress(ddev, 1); | |
561 | ||
562 | if (device_add(ddev)) | |
563 | return; | |
564 | if (!sysfs_deprecated) { | |
565 | err = sysfs_create_link(block_depr, &ddev->kobj, | |
566 | kobject_name(&ddev->kobj)); | |
567 | if (err) { | |
568 | device_del(ddev); | |
569 | return; | |
570 | } | |
571 | } | |
25e823c8 ML |
572 | |
573 | /* | |
574 | * avoid probable deadlock caused by allocating memory with | |
575 | * GFP_KERNEL in runtime_resume callback of its all ancestor | |
576 | * devices | |
577 | */ | |
578 | pm_runtime_set_memalloc_noio(ddev, true); | |
579 | ||
d2bf1b67 TH |
580 | disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj); |
581 | disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj); | |
582 | ||
583 | /* No minors to use for partitions */ | |
d27769ec | 584 | if (!disk_part_scan_enabled(disk)) |
d2bf1b67 TH |
585 | goto exit; |
586 | ||
587 | /* No such device (e.g., media were just removed) */ | |
588 | if (!get_capacity(disk)) | |
589 | goto exit; | |
590 | ||
591 | bdev = bdget_disk(disk, 0); | |
592 | if (!bdev) | |
593 | goto exit; | |
594 | ||
595 | bdev->bd_invalidated = 1; | |
596 | err = blkdev_get(bdev, FMODE_READ, NULL); | |
597 | if (err < 0) | |
598 | goto exit; | |
599 | blkdev_put(bdev, FMODE_READ); | |
600 | ||
601 | exit: | |
602 | /* announce disk after possible partitions are created */ | |
603 | dev_set_uevent_suppress(ddev, 0); | |
604 | kobject_uevent(&ddev->kobj, KOBJ_ADD); | |
605 | ||
606 | /* announce possible partitions */ | |
607 | disk_part_iter_init(&piter, disk, 0); | |
608 | while ((part = disk_part_iter_next(&piter))) | |
609 | kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD); | |
610 | disk_part_iter_exit(&piter); | |
611 | } | |
612 | ||
1da177e4 | 613 | /** |
e63a46be DW |
614 | * device_add_disk - add partitioning information to kernel list |
615 | * @parent: parent device for the disk | |
1da177e4 LT |
616 | * @disk: per-device partitioning information |
617 | * | |
618 | * This function registers the partitioning information in @disk | |
619 | * with the kernel. | |
3e1a7ff8 TH |
620 | * |
621 | * FIXME: error handling | |
1da177e4 | 622 | */ |
e63a46be | 623 | void device_add_disk(struct device *parent, struct gendisk *disk) |
1da177e4 | 624 | { |
cf0ca9fe | 625 | struct backing_dev_info *bdi; |
3e1a7ff8 | 626 | dev_t devt; |
6ffeea77 | 627 | int retval; |
cf0ca9fe | 628 | |
3e1a7ff8 TH |
629 | /* minors == 0 indicates to use ext devt from part0 and should |
630 | * be accompanied with EXT_DEVT flag. Make sure all | |
631 | * parameters make sense. | |
632 | */ | |
633 | WARN_ON(disk->minors && !(disk->major || disk->first_minor)); | |
634 | WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT)); | |
635 | ||
1da177e4 | 636 | disk->flags |= GENHD_FL_UP; |
3e1a7ff8 TH |
637 | |
638 | retval = blk_alloc_devt(&disk->part0, &devt); | |
639 | if (retval) { | |
640 | WARN_ON(1); | |
641 | return; | |
642 | } | |
643 | disk_to_dev(disk)->devt = devt; | |
644 | ||
645 | /* ->major and ->first_minor aren't supposed to be | |
646 | * dereferenced from here on, but set them just in case. | |
647 | */ | |
648 | disk->major = MAJOR(devt); | |
649 | disk->first_minor = MINOR(devt); | |
650 | ||
9f53d2fe SG |
651 | disk_alloc_events(disk); |
652 | ||
9f5e4865 | 653 | /* Register BDI before referencing it from bdev */ |
dc3b17cc | 654 | bdi = disk->queue->backing_dev_info; |
df08c32c | 655 | bdi_register_owner(bdi, disk_to_dev(disk)); |
01ea5063 | 656 | |
f331c029 TH |
657 | blk_register_region(disk_devt(disk), disk->minors, NULL, |
658 | exact_match, exact_lock, disk); | |
e63a46be | 659 | register_disk(parent, disk); |
1da177e4 | 660 | blk_register_queue(disk); |
cf0ca9fe | 661 | |
523e1d39 TH |
662 | /* |
663 | * Take an extra ref on queue which will be put on disk_release() | |
664 | * so that it sticks around as long as @disk is there. | |
665 | */ | |
09ac46c4 | 666 | WARN_ON_ONCE(!blk_get_queue(disk->queue)); |
523e1d39 | 667 | |
ed9e1982 TH |
668 | retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj, |
669 | "bdi"); | |
6ffeea77 | 670 | WARN_ON(retval); |
1da177e4 | 671 | |
77ea887e | 672 | disk_add_events(disk); |
25520d55 | 673 | blk_integrity_add(disk); |
1da177e4 | 674 | } |
e63a46be | 675 | EXPORT_SYMBOL(device_add_disk); |
1da177e4 | 676 | |
d2bf1b67 | 677 | void del_gendisk(struct gendisk *disk) |
1da177e4 | 678 | { |
d2bf1b67 TH |
679 | struct disk_part_iter piter; |
680 | struct hd_struct *part; | |
681 | ||
25520d55 | 682 | blk_integrity_del(disk); |
77ea887e TH |
683 | disk_del_events(disk); |
684 | ||
d2bf1b67 TH |
685 | /* invalidate stuff */ |
686 | disk_part_iter_init(&piter, disk, | |
687 | DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE); | |
688 | while ((part = disk_part_iter_next(&piter))) { | |
689 | invalidate_partition(disk, part->partno); | |
4b8c861a | 690 | bdev_unhash_inode(part_devt(part)); |
d2bf1b67 TH |
691 | delete_partition(disk, part->partno); |
692 | } | |
693 | disk_part_iter_exit(&piter); | |
694 | ||
695 | invalidate_partition(disk, 0); | |
d06e05c0 | 696 | bdev_unhash_inode(disk_devt(disk)); |
d2bf1b67 TH |
697 | set_capacity(disk, 0); |
698 | disk->flags &= ~GENHD_FL_UP; | |
699 | ||
ed9e1982 | 700 | sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi"); |
90f16fdd JK |
701 | if (disk->queue) { |
702 | /* | |
703 | * Unregister bdi before releasing device numbers (as they can | |
704 | * get reused and we'd get clashes in sysfs). | |
705 | */ | |
706 | bdi_unregister(disk->queue->backing_dev_info); | |
707 | blk_unregister_queue(disk); | |
708 | } else { | |
709 | WARN_ON(1); | |
710 | } | |
f331c029 | 711 | blk_unregister_region(disk_devt(disk), disk->minors); |
d2bf1b67 TH |
712 | |
713 | part_stat_set_all(&disk->part0, 0); | |
714 | disk->part0.stamp = 0; | |
715 | ||
716 | kobject_put(disk->part0.holder_dir); | |
717 | kobject_put(disk->slave_dir); | |
d2bf1b67 TH |
718 | if (!sysfs_deprecated) |
719 | sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk))); | |
25e823c8 | 720 | pm_runtime_set_memalloc_noio(disk_to_dev(disk), false); |
d2bf1b67 | 721 | device_del(disk_to_dev(disk)); |
1da177e4 | 722 | } |
d2bf1b67 | 723 | EXPORT_SYMBOL(del_gendisk); |
1da177e4 | 724 | |
99e6608c VV |
725 | /* sysfs access to bad-blocks list. */ |
726 | static ssize_t disk_badblocks_show(struct device *dev, | |
727 | struct device_attribute *attr, | |
728 | char *page) | |
729 | { | |
730 | struct gendisk *disk = dev_to_disk(dev); | |
731 | ||
732 | if (!disk->bb) | |
733 | return sprintf(page, "\n"); | |
734 | ||
735 | return badblocks_show(disk->bb, page, 0); | |
736 | } | |
737 | ||
738 | static ssize_t disk_badblocks_store(struct device *dev, | |
739 | struct device_attribute *attr, | |
740 | const char *page, size_t len) | |
741 | { | |
742 | struct gendisk *disk = dev_to_disk(dev); | |
743 | ||
744 | if (!disk->bb) | |
745 | return -ENXIO; | |
746 | ||
747 | return badblocks_store(disk->bb, page, len, 0); | |
748 | } | |
749 | ||
1da177e4 LT |
750 | /** |
751 | * get_gendisk - get partitioning information for a given device | |
710027a4 | 752 | * @devt: device to get partitioning information for |
496aa8a9 | 753 | * @partno: returned partition index |
1da177e4 LT |
754 | * |
755 | * This function gets the structure containing partitioning | |
710027a4 | 756 | * information for the given device @devt. |
1da177e4 | 757 | */ |
cf771cb5 | 758 | struct gendisk *get_gendisk(dev_t devt, int *partno) |
1da177e4 | 759 | { |
bcce3de1 TH |
760 | struct gendisk *disk = NULL; |
761 | ||
762 | if (MAJOR(devt) != BLOCK_EXT_MAJOR) { | |
763 | struct kobject *kobj; | |
764 | ||
765 | kobj = kobj_lookup(bdev_map, devt, partno); | |
766 | if (kobj) | |
767 | disk = dev_to_disk(kobj_to_dev(kobj)); | |
768 | } else { | |
769 | struct hd_struct *part; | |
770 | ||
4d66e5e9 | 771 | spin_lock_bh(&ext_devt_lock); |
870d6656 | 772 | part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); |
bcce3de1 TH |
773 | if (part && get_disk(part_to_disk(part))) { |
774 | *partno = part->partno; | |
775 | disk = part_to_disk(part); | |
776 | } | |
4d66e5e9 | 777 | spin_unlock_bh(&ext_devt_lock); |
bcce3de1 | 778 | } |
edfaa7c3 | 779 | |
bcce3de1 | 780 | return disk; |
1da177e4 | 781 | } |
b6ac23af | 782 | EXPORT_SYMBOL(get_gendisk); |
1da177e4 | 783 | |
f331c029 TH |
784 | /** |
785 | * bdget_disk - do bdget() by gendisk and partition number | |
786 | * @disk: gendisk of interest | |
787 | * @partno: partition number | |
788 | * | |
789 | * Find partition @partno from @disk, do bdget() on it. | |
790 | * | |
791 | * CONTEXT: | |
792 | * Don't care. | |
793 | * | |
794 | * RETURNS: | |
795 | * Resulting block_device on success, NULL on failure. | |
796 | */ | |
aeb3d3a8 | 797 | struct block_device *bdget_disk(struct gendisk *disk, int partno) |
f331c029 | 798 | { |
548b10eb TH |
799 | struct hd_struct *part; |
800 | struct block_device *bdev = NULL; | |
f331c029 | 801 | |
548b10eb | 802 | part = disk_get_part(disk, partno); |
2bbedcb4 | 803 | if (part) |
548b10eb TH |
804 | bdev = bdget(part_devt(part)); |
805 | disk_put_part(part); | |
f331c029 | 806 | |
548b10eb | 807 | return bdev; |
f331c029 TH |
808 | } |
809 | EXPORT_SYMBOL(bdget_disk); | |
810 | ||
5c6f35c5 GKH |
811 | /* |
812 | * print a full list of all partitions - intended for places where the root | |
813 | * filesystem can't be mounted and thus to give the victim some idea of what | |
814 | * went wrong | |
815 | */ | |
816 | void __init printk_all_partitions(void) | |
817 | { | |
def4e38d TH |
818 | struct class_dev_iter iter; |
819 | struct device *dev; | |
820 | ||
821 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); | |
822 | while ((dev = class_dev_iter_next(&iter))) { | |
823 | struct gendisk *disk = dev_to_disk(dev); | |
e71bf0d0 TH |
824 | struct disk_part_iter piter; |
825 | struct hd_struct *part; | |
1f014290 TH |
826 | char name_buf[BDEVNAME_SIZE]; |
827 | char devt_buf[BDEVT_SIZE]; | |
def4e38d TH |
828 | |
829 | /* | |
830 | * Don't show empty devices or things that have been | |
25985edc | 831 | * suppressed |
def4e38d TH |
832 | */ |
833 | if (get_capacity(disk) == 0 || | |
834 | (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)) | |
835 | continue; | |
836 | ||
837 | /* | |
838 | * Note, unlike /proc/partitions, I am showing the | |
839 | * numbers in hex - the same format as the root= | |
840 | * option takes. | |
841 | */ | |
074a7aca TH |
842 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0); |
843 | while ((part = disk_part_iter_next(&piter))) { | |
844 | bool is_part0 = part == &disk->part0; | |
def4e38d | 845 | |
b5af921e | 846 | printk("%s%s %10llu %s %s", is_part0 ? "" : " ", |
1f014290 | 847 | bdevt_str(part_devt(part), devt_buf), |
c83f6bf9 VG |
848 | (unsigned long long)part_nr_sects_read(part) >> 1 |
849 | , disk_name(disk, part->partno, name_buf), | |
1ad7e899 | 850 | part->info ? part->info->uuid : ""); |
074a7aca | 851 | if (is_part0) { |
52c44d93 | 852 | if (dev->parent && dev->parent->driver) |
074a7aca | 853 | printk(" driver: %s\n", |
52c44d93 | 854 | dev->parent->driver->name); |
074a7aca TH |
855 | else |
856 | printk(" (driver?)\n"); | |
857 | } else | |
858 | printk("\n"); | |
859 | } | |
e71bf0d0 | 860 | disk_part_iter_exit(&piter); |
def4e38d TH |
861 | } |
862 | class_dev_iter_exit(&iter); | |
dd2a345f DG |
863 | } |
864 | ||
1da177e4 LT |
865 | #ifdef CONFIG_PROC_FS |
866 | /* iterator */ | |
def4e38d | 867 | static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos) |
68c4d4a7 | 868 | { |
def4e38d TH |
869 | loff_t skip = *pos; |
870 | struct class_dev_iter *iter; | |
871 | struct device *dev; | |
68c4d4a7 | 872 | |
aeb3d3a8 | 873 | iter = kmalloc(sizeof(*iter), GFP_KERNEL); |
def4e38d TH |
874 | if (!iter) |
875 | return ERR_PTR(-ENOMEM); | |
876 | ||
877 | seqf->private = iter; | |
878 | class_dev_iter_init(iter, &block_class, NULL, &disk_type); | |
879 | do { | |
880 | dev = class_dev_iter_next(iter); | |
881 | if (!dev) | |
882 | return NULL; | |
883 | } while (skip--); | |
884 | ||
885 | return dev_to_disk(dev); | |
68c4d4a7 GKH |
886 | } |
887 | ||
def4e38d | 888 | static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos) |
1da177e4 | 889 | { |
edfaa7c3 | 890 | struct device *dev; |
1da177e4 | 891 | |
def4e38d TH |
892 | (*pos)++; |
893 | dev = class_dev_iter_next(seqf->private); | |
2ac3cee5 | 894 | if (dev) |
68c4d4a7 | 895 | return dev_to_disk(dev); |
2ac3cee5 | 896 | |
1da177e4 LT |
897 | return NULL; |
898 | } | |
899 | ||
def4e38d | 900 | static void disk_seqf_stop(struct seq_file *seqf, void *v) |
27f30251 | 901 | { |
def4e38d | 902 | struct class_dev_iter *iter = seqf->private; |
27f30251 | 903 | |
def4e38d TH |
904 | /* stop is called even after start failed :-( */ |
905 | if (iter) { | |
906 | class_dev_iter_exit(iter); | |
907 | kfree(iter); | |
77da1605 | 908 | seqf->private = NULL; |
5c0ef6d0 | 909 | } |
1da177e4 LT |
910 | } |
911 | ||
def4e38d | 912 | static void *show_partition_start(struct seq_file *seqf, loff_t *pos) |
1da177e4 | 913 | { |
06768067 | 914 | void *p; |
def4e38d TH |
915 | |
916 | p = disk_seqf_start(seqf, pos); | |
b9f985b6 | 917 | if (!IS_ERR_OR_NULL(p) && !*pos) |
def4e38d TH |
918 | seq_puts(seqf, "major minor #blocks name\n\n"); |
919 | return p; | |
1da177e4 LT |
920 | } |
921 | ||
cf771cb5 | 922 | static int show_partition(struct seq_file *seqf, void *v) |
1da177e4 LT |
923 | { |
924 | struct gendisk *sgp = v; | |
e71bf0d0 TH |
925 | struct disk_part_iter piter; |
926 | struct hd_struct *part; | |
1da177e4 LT |
927 | char buf[BDEVNAME_SIZE]; |
928 | ||
1da177e4 | 929 | /* Don't show non-partitionable removeable devices or empty devices */ |
d27769ec | 930 | if (!get_capacity(sgp) || (!disk_max_parts(sgp) && |
f331c029 | 931 | (sgp->flags & GENHD_FL_REMOVABLE))) |
1da177e4 LT |
932 | return 0; |
933 | if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO) | |
934 | return 0; | |
935 | ||
936 | /* show the full disk and all non-0 size partitions of it */ | |
074a7aca | 937 | disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0); |
e71bf0d0 | 938 | while ((part = disk_part_iter_next(&piter))) |
1f014290 | 939 | seq_printf(seqf, "%4d %7d %10llu %s\n", |
f331c029 | 940 | MAJOR(part_devt(part)), MINOR(part_devt(part)), |
c83f6bf9 | 941 | (unsigned long long)part_nr_sects_read(part) >> 1, |
f331c029 | 942 | disk_name(sgp, part->partno, buf)); |
e71bf0d0 | 943 | disk_part_iter_exit(&piter); |
1da177e4 LT |
944 | |
945 | return 0; | |
946 | } | |
947 | ||
f500975a | 948 | static const struct seq_operations partitions_op = { |
def4e38d TH |
949 | .start = show_partition_start, |
950 | .next = disk_seqf_next, | |
951 | .stop = disk_seqf_stop, | |
edfaa7c3 | 952 | .show = show_partition |
1da177e4 | 953 | }; |
f500975a AD |
954 | |
955 | static int partitions_open(struct inode *inode, struct file *file) | |
956 | { | |
957 | return seq_open(file, &partitions_op); | |
958 | } | |
959 | ||
960 | static const struct file_operations proc_partitions_operations = { | |
961 | .open = partitions_open, | |
962 | .read = seq_read, | |
963 | .llseek = seq_lseek, | |
964 | .release = seq_release, | |
965 | }; | |
1da177e4 LT |
966 | #endif |
967 | ||
968 | ||
cf771cb5 | 969 | static struct kobject *base_probe(dev_t devt, int *partno, void *data) |
1da177e4 | 970 | { |
edfaa7c3 | 971 | if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0) |
1da177e4 | 972 | /* Make old-style 2.4 aliases work */ |
edfaa7c3 | 973 | request_module("block-major-%d", MAJOR(devt)); |
1da177e4 LT |
974 | return NULL; |
975 | } | |
976 | ||
977 | static int __init genhd_device_init(void) | |
978 | { | |
e105b8bf DW |
979 | int error; |
980 | ||
981 | block_class.dev_kobj = sysfs_dev_block_kobj; | |
982 | error = class_register(&block_class); | |
ee27a558 RM |
983 | if (unlikely(error)) |
984 | return error; | |
edfaa7c3 | 985 | bdev_map = kobj_map_init(base_probe, &block_class_lock); |
1da177e4 | 986 | blk_dev_init(); |
edfaa7c3 | 987 | |
561ec68e ZY |
988 | register_blkdev(BLOCK_EXT_MAJOR, "blkext"); |
989 | ||
edfaa7c3 | 990 | /* create top-level block dir */ |
e52eec13 AK |
991 | if (!sysfs_deprecated) |
992 | block_depr = kobject_create_and_add("block", NULL); | |
830d3cfb | 993 | return 0; |
1da177e4 LT |
994 | } |
995 | ||
996 | subsys_initcall(genhd_device_init); | |
997 | ||
edfaa7c3 KS |
998 | static ssize_t disk_range_show(struct device *dev, |
999 | struct device_attribute *attr, char *buf) | |
1da177e4 | 1000 | { |
edfaa7c3 | 1001 | struct gendisk *disk = dev_to_disk(dev); |
1da177e4 | 1002 | |
edfaa7c3 | 1003 | return sprintf(buf, "%d\n", disk->minors); |
1da177e4 LT |
1004 | } |
1005 | ||
1f014290 TH |
1006 | static ssize_t disk_ext_range_show(struct device *dev, |
1007 | struct device_attribute *attr, char *buf) | |
1008 | { | |
1009 | struct gendisk *disk = dev_to_disk(dev); | |
1010 | ||
b5d0b9df | 1011 | return sprintf(buf, "%d\n", disk_max_parts(disk)); |
1f014290 TH |
1012 | } |
1013 | ||
edfaa7c3 KS |
1014 | static ssize_t disk_removable_show(struct device *dev, |
1015 | struct device_attribute *attr, char *buf) | |
a7fd6706 | 1016 | { |
edfaa7c3 | 1017 | struct gendisk *disk = dev_to_disk(dev); |
a7fd6706 | 1018 | |
edfaa7c3 KS |
1019 | return sprintf(buf, "%d\n", |
1020 | (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0)); | |
a7fd6706 KS |
1021 | } |
1022 | ||
1c9ce527 KS |
1023 | static ssize_t disk_ro_show(struct device *dev, |
1024 | struct device_attribute *attr, char *buf) | |
1025 | { | |
1026 | struct gendisk *disk = dev_to_disk(dev); | |
1027 | ||
b7db9956 | 1028 | return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0); |
1c9ce527 KS |
1029 | } |
1030 | ||
edfaa7c3 KS |
1031 | static ssize_t disk_capability_show(struct device *dev, |
1032 | struct device_attribute *attr, char *buf) | |
86ce18d7 | 1033 | { |
edfaa7c3 KS |
1034 | struct gendisk *disk = dev_to_disk(dev); |
1035 | ||
1036 | return sprintf(buf, "%x\n", disk->flags); | |
86ce18d7 | 1037 | } |
edfaa7c3 | 1038 | |
c72758f3 MP |
1039 | static ssize_t disk_alignment_offset_show(struct device *dev, |
1040 | struct device_attribute *attr, | |
1041 | char *buf) | |
1042 | { | |
1043 | struct gendisk *disk = dev_to_disk(dev); | |
1044 | ||
1045 | return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue)); | |
1046 | } | |
1047 | ||
86b37281 MP |
1048 | static ssize_t disk_discard_alignment_show(struct device *dev, |
1049 | struct device_attribute *attr, | |
1050 | char *buf) | |
1051 | { | |
1052 | struct gendisk *disk = dev_to_disk(dev); | |
1053 | ||
dd3d145d | 1054 | return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue)); |
86b37281 MP |
1055 | } |
1056 | ||
edfaa7c3 | 1057 | static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL); |
1f014290 | 1058 | static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL); |
edfaa7c3 | 1059 | static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL); |
1c9ce527 | 1060 | static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL); |
e5610521 | 1061 | static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL); |
c72758f3 | 1062 | static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL); |
86b37281 MP |
1063 | static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show, |
1064 | NULL); | |
edfaa7c3 | 1065 | static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL); |
074a7aca | 1066 | static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL); |
316d315b | 1067 | static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL); |
99e6608c VV |
1068 | static DEVICE_ATTR(badblocks, S_IRUGO | S_IWUSR, disk_badblocks_show, |
1069 | disk_badblocks_store); | |
c17bb495 | 1070 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
edfaa7c3 | 1071 | static struct device_attribute dev_attr_fail = |
eddb2e26 | 1072 | __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store); |
c17bb495 | 1073 | #endif |
581d4e28 JA |
1074 | #ifdef CONFIG_FAIL_IO_TIMEOUT |
1075 | static struct device_attribute dev_attr_fail_timeout = | |
1076 | __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show, | |
1077 | part_timeout_store); | |
1078 | #endif | |
edfaa7c3 KS |
1079 | |
1080 | static struct attribute *disk_attrs[] = { | |
1081 | &dev_attr_range.attr, | |
1f014290 | 1082 | &dev_attr_ext_range.attr, |
edfaa7c3 | 1083 | &dev_attr_removable.attr, |
1c9ce527 | 1084 | &dev_attr_ro.attr, |
edfaa7c3 | 1085 | &dev_attr_size.attr, |
c72758f3 | 1086 | &dev_attr_alignment_offset.attr, |
86b37281 | 1087 | &dev_attr_discard_alignment.attr, |
edfaa7c3 KS |
1088 | &dev_attr_capability.attr, |
1089 | &dev_attr_stat.attr, | |
316d315b | 1090 | &dev_attr_inflight.attr, |
99e6608c | 1091 | &dev_attr_badblocks.attr, |
edfaa7c3 KS |
1092 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
1093 | &dev_attr_fail.attr, | |
581d4e28 JA |
1094 | #endif |
1095 | #ifdef CONFIG_FAIL_IO_TIMEOUT | |
1096 | &dev_attr_fail_timeout.attr, | |
edfaa7c3 KS |
1097 | #endif |
1098 | NULL | |
1099 | }; | |
1100 | ||
9438b3e0 DW |
1101 | static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n) |
1102 | { | |
1103 | struct device *dev = container_of(kobj, typeof(*dev), kobj); | |
1104 | struct gendisk *disk = dev_to_disk(dev); | |
1105 | ||
1106 | if (a == &dev_attr_badblocks.attr && !disk->bb) | |
1107 | return 0; | |
1108 | return a->mode; | |
1109 | } | |
1110 | ||
edfaa7c3 KS |
1111 | static struct attribute_group disk_attr_group = { |
1112 | .attrs = disk_attrs, | |
9438b3e0 | 1113 | .is_visible = disk_visible, |
edfaa7c3 KS |
1114 | }; |
1115 | ||
a4dbd674 | 1116 | static const struct attribute_group *disk_attr_groups[] = { |
edfaa7c3 KS |
1117 | &disk_attr_group, |
1118 | NULL | |
1da177e4 LT |
1119 | }; |
1120 | ||
540eed56 TH |
1121 | /** |
1122 | * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way | |
1123 | * @disk: disk to replace part_tbl for | |
1124 | * @new_ptbl: new part_tbl to install | |
1125 | * | |
1126 | * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The | |
1127 | * original ptbl is freed using RCU callback. | |
1128 | * | |
1129 | * LOCKING: | |
6d2cf6f2 | 1130 | * Matching bd_mutex locked or the caller is the only user of @disk. |
540eed56 TH |
1131 | */ |
1132 | static void disk_replace_part_tbl(struct gendisk *disk, | |
1133 | struct disk_part_tbl *new_ptbl) | |
1134 | { | |
6d2cf6f2 BVA |
1135 | struct disk_part_tbl *old_ptbl = |
1136 | rcu_dereference_protected(disk->part_tbl, 1); | |
540eed56 TH |
1137 | |
1138 | rcu_assign_pointer(disk->part_tbl, new_ptbl); | |
a6f23657 JA |
1139 | |
1140 | if (old_ptbl) { | |
1141 | rcu_assign_pointer(old_ptbl->last_lookup, NULL); | |
57bdfbf9 | 1142 | kfree_rcu(old_ptbl, rcu_head); |
a6f23657 | 1143 | } |
540eed56 TH |
1144 | } |
1145 | ||
1146 | /** | |
1147 | * disk_expand_part_tbl - expand disk->part_tbl | |
1148 | * @disk: disk to expand part_tbl for | |
1149 | * @partno: expand such that this partno can fit in | |
1150 | * | |
1151 | * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl | |
1152 | * uses RCU to allow unlocked dereferencing for stats and other stuff. | |
1153 | * | |
1154 | * LOCKING: | |
6d2cf6f2 BVA |
1155 | * Matching bd_mutex locked or the caller is the only user of @disk. |
1156 | * Might sleep. | |
540eed56 TH |
1157 | * |
1158 | * RETURNS: | |
1159 | * 0 on success, -errno on failure. | |
1160 | */ | |
1161 | int disk_expand_part_tbl(struct gendisk *disk, int partno) | |
1162 | { | |
6d2cf6f2 BVA |
1163 | struct disk_part_tbl *old_ptbl = |
1164 | rcu_dereference_protected(disk->part_tbl, 1); | |
540eed56 TH |
1165 | struct disk_part_tbl *new_ptbl; |
1166 | int len = old_ptbl ? old_ptbl->len : 0; | |
5fabcb4c | 1167 | int i, target; |
540eed56 | 1168 | size_t size; |
5fabcb4c JA |
1169 | |
1170 | /* | |
1171 | * check for int overflow, since we can get here from blkpg_ioctl() | |
1172 | * with a user passed 'partno'. | |
1173 | */ | |
1174 | target = partno + 1; | |
1175 | if (target < 0) | |
1176 | return -EINVAL; | |
540eed56 TH |
1177 | |
1178 | /* disk_max_parts() is zero during initialization, ignore if so */ | |
1179 | if (disk_max_parts(disk) && target > disk_max_parts(disk)) | |
1180 | return -EINVAL; | |
1181 | ||
1182 | if (target <= len) | |
1183 | return 0; | |
1184 | ||
1185 | size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]); | |
1186 | new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id); | |
1187 | if (!new_ptbl) | |
1188 | return -ENOMEM; | |
1189 | ||
540eed56 TH |
1190 | new_ptbl->len = target; |
1191 | ||
1192 | for (i = 0; i < len; i++) | |
1193 | rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]); | |
1194 | ||
1195 | disk_replace_part_tbl(disk, new_ptbl); | |
1196 | return 0; | |
1197 | } | |
1198 | ||
edfaa7c3 | 1199 | static void disk_release(struct device *dev) |
1da177e4 | 1200 | { |
edfaa7c3 KS |
1201 | struct gendisk *disk = dev_to_disk(dev); |
1202 | ||
2da78092 | 1203 | blk_free_devt(dev->devt); |
77ea887e | 1204 | disk_release_events(disk); |
1da177e4 | 1205 | kfree(disk->random); |
540eed56 | 1206 | disk_replace_part_tbl(disk, NULL); |
b54e5ed8 | 1207 | hd_free_part(&disk->part0); |
523e1d39 TH |
1208 | if (disk->queue) |
1209 | blk_put_queue(disk->queue); | |
1da177e4 LT |
1210 | kfree(disk); |
1211 | } | |
edfaa7c3 KS |
1212 | struct class block_class = { |
1213 | .name = "block", | |
1da177e4 LT |
1214 | }; |
1215 | ||
3c2670e6 | 1216 | static char *block_devnode(struct device *dev, umode_t *mode, |
4e4098a3 | 1217 | kuid_t *uid, kgid_t *gid) |
b03f38b6 KS |
1218 | { |
1219 | struct gendisk *disk = dev_to_disk(dev); | |
1220 | ||
e454cea2 KS |
1221 | if (disk->devnode) |
1222 | return disk->devnode(disk, mode); | |
b03f38b6 KS |
1223 | return NULL; |
1224 | } | |
1225 | ||
edf8ff55 | 1226 | static const struct device_type disk_type = { |
edfaa7c3 KS |
1227 | .name = "disk", |
1228 | .groups = disk_attr_groups, | |
1229 | .release = disk_release, | |
e454cea2 | 1230 | .devnode = block_devnode, |
1da177e4 LT |
1231 | }; |
1232 | ||
a6e2ba88 | 1233 | #ifdef CONFIG_PROC_FS |
cf771cb5 TH |
1234 | /* |
1235 | * aggregate disk stat collector. Uses the same stats that the sysfs | |
1236 | * entries do, above, but makes them available through one seq_file. | |
1237 | * | |
1238 | * The output looks suspiciously like /proc/partitions with a bunch of | |
1239 | * extra fields. | |
1240 | */ | |
1241 | static int diskstats_show(struct seq_file *seqf, void *v) | |
1da177e4 LT |
1242 | { |
1243 | struct gendisk *gp = v; | |
e71bf0d0 TH |
1244 | struct disk_part_iter piter; |
1245 | struct hd_struct *hd; | |
1da177e4 | 1246 | char buf[BDEVNAME_SIZE]; |
0609e0ef | 1247 | unsigned int inflight[2]; |
c9959059 | 1248 | int cpu; |
1da177e4 LT |
1249 | |
1250 | /* | |
ed9e1982 | 1251 | if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next) |
cf771cb5 | 1252 | seq_puts(seqf, "major minor name" |
1da177e4 LT |
1253 | " rio rmerge rsect ruse wio wmerge " |
1254 | "wsect wuse running use aveq" | |
1255 | "\n\n"); | |
1256 | */ | |
9f5e4865 | 1257 | |
71982a40 | 1258 | disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0); |
e71bf0d0 | 1259 | while ((hd = disk_part_iter_next(&piter))) { |
074a7aca | 1260 | cpu = part_stat_lock(); |
d62e26b3 | 1261 | part_round_stats(gp->queue, cpu, hd); |
074a7aca | 1262 | part_stat_unlock(); |
0609e0ef | 1263 | part_in_flight(gp->queue, hd, inflight); |
f95fe9cf HP |
1264 | seq_printf(seqf, "%4d %7d %s %lu %lu %lu " |
1265 | "%u %lu %lu %lu %u %u %u %u\n", | |
f331c029 TH |
1266 | MAJOR(part_devt(hd)), MINOR(part_devt(hd)), |
1267 | disk_name(gp, hd->partno, buf), | |
53f22956 LY |
1268 | part_stat_read(hd, ios[READ]), |
1269 | part_stat_read(hd, merges[READ]), | |
f95fe9cf | 1270 | part_stat_read(hd, sectors[READ]), |
53f22956 LY |
1271 | jiffies_to_msecs(part_stat_read(hd, ticks[READ])), |
1272 | part_stat_read(hd, ios[WRITE]), | |
1273 | part_stat_read(hd, merges[WRITE]), | |
f95fe9cf | 1274 | part_stat_read(hd, sectors[WRITE]), |
53f22956 | 1275 | jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])), |
0609e0ef | 1276 | inflight[0], |
28f39d55 JM |
1277 | jiffies_to_msecs(part_stat_read(hd, io_ticks)), |
1278 | jiffies_to_msecs(part_stat_read(hd, time_in_queue)) | |
1279 | ); | |
1da177e4 | 1280 | } |
e71bf0d0 | 1281 | disk_part_iter_exit(&piter); |
9f5e4865 | 1282 | |
1da177e4 LT |
1283 | return 0; |
1284 | } | |
1285 | ||
31d85ab2 | 1286 | static const struct seq_operations diskstats_op = { |
def4e38d TH |
1287 | .start = disk_seqf_start, |
1288 | .next = disk_seqf_next, | |
1289 | .stop = disk_seqf_stop, | |
1da177e4 LT |
1290 | .show = diskstats_show |
1291 | }; | |
f500975a | 1292 | |
31d85ab2 AD |
1293 | static int diskstats_open(struct inode *inode, struct file *file) |
1294 | { | |
1295 | return seq_open(file, &diskstats_op); | |
1296 | } | |
1297 | ||
1298 | static const struct file_operations proc_diskstats_operations = { | |
1299 | .open = diskstats_open, | |
1300 | .read = seq_read, | |
1301 | .llseek = seq_lseek, | |
1302 | .release = seq_release, | |
1303 | }; | |
1304 | ||
f500975a AD |
1305 | static int __init proc_genhd_init(void) |
1306 | { | |
31d85ab2 | 1307 | proc_create("diskstats", 0, NULL, &proc_diskstats_operations); |
f500975a AD |
1308 | proc_create("partitions", 0, NULL, &proc_partitions_operations); |
1309 | return 0; | |
1310 | } | |
1311 | module_init(proc_genhd_init); | |
a6e2ba88 | 1312 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 1313 | |
cf771cb5 | 1314 | dev_t blk_lookup_devt(const char *name, int partno) |
a142be85 | 1315 | { |
def4e38d TH |
1316 | dev_t devt = MKDEV(0, 0); |
1317 | struct class_dev_iter iter; | |
1318 | struct device *dev; | |
a142be85 | 1319 | |
def4e38d TH |
1320 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); |
1321 | while ((dev = class_dev_iter_next(&iter))) { | |
a142be85 | 1322 | struct gendisk *disk = dev_to_disk(dev); |
548b10eb | 1323 | struct hd_struct *part; |
a142be85 | 1324 | |
3ada8b7e | 1325 | if (strcmp(dev_name(dev), name)) |
f331c029 | 1326 | continue; |
f331c029 | 1327 | |
41b8c853 NB |
1328 | if (partno < disk->minors) { |
1329 | /* We need to return the right devno, even | |
1330 | * if the partition doesn't exist yet. | |
1331 | */ | |
1332 | devt = MKDEV(MAJOR(dev->devt), | |
1333 | MINOR(dev->devt) + partno); | |
1334 | break; | |
1335 | } | |
548b10eb | 1336 | part = disk_get_part(disk, partno); |
2bbedcb4 | 1337 | if (part) { |
f331c029 | 1338 | devt = part_devt(part); |
e71bf0d0 | 1339 | disk_put_part(part); |
548b10eb | 1340 | break; |
def4e38d | 1341 | } |
548b10eb | 1342 | disk_put_part(part); |
5c0ef6d0 | 1343 | } |
def4e38d | 1344 | class_dev_iter_exit(&iter); |
edfaa7c3 KS |
1345 | return devt; |
1346 | } | |
edfaa7c3 KS |
1347 | EXPORT_SYMBOL(blk_lookup_devt); |
1348 | ||
1da177e4 LT |
1349 | struct gendisk *alloc_disk(int minors) |
1350 | { | |
c304a51b | 1351 | return alloc_disk_node(minors, NUMA_NO_NODE); |
1946089a | 1352 | } |
689d6fac | 1353 | EXPORT_SYMBOL(alloc_disk); |
1946089a CL |
1354 | |
1355 | struct gendisk *alloc_disk_node(int minors, int node_id) | |
1356 | { | |
1357 | struct gendisk *disk; | |
6d2cf6f2 | 1358 | struct disk_part_tbl *ptbl; |
1946089a | 1359 | |
de65b012 CH |
1360 | if (minors > DISK_MAX_PARTS) { |
1361 | printk(KERN_ERR | |
1362 | "block: can't allocated more than %d partitions\n", | |
1363 | DISK_MAX_PARTS); | |
1364 | minors = DISK_MAX_PARTS; | |
1365 | } | |
1366 | ||
c1b511eb | 1367 | disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id); |
1da177e4 | 1368 | if (disk) { |
074a7aca | 1369 | if (!init_part_stats(&disk->part0)) { |
1da177e4 LT |
1370 | kfree(disk); |
1371 | return NULL; | |
1372 | } | |
bf91db18 | 1373 | disk->node_id = node_id; |
540eed56 TH |
1374 | if (disk_expand_part_tbl(disk, 0)) { |
1375 | free_part_stats(&disk->part0); | |
b5d0b9df TH |
1376 | kfree(disk); |
1377 | return NULL; | |
1da177e4 | 1378 | } |
6d2cf6f2 BVA |
1379 | ptbl = rcu_dereference_protected(disk->part_tbl, 1); |
1380 | rcu_assign_pointer(ptbl->part[0], &disk->part0); | |
6c23a968 | 1381 | |
c83f6bf9 VG |
1382 | /* |
1383 | * set_capacity() and get_capacity() currently don't use | |
1384 | * seqcounter to read/update the part0->nr_sects. Still init | |
1385 | * the counter as we can read the sectors in IO submission | |
1386 | * patch using seqence counters. | |
1387 | * | |
1388 | * TODO: Ideally set_capacity() and get_capacity() should be | |
1389 | * converted to make use of bd_mutex and sequence counters. | |
1390 | */ | |
1391 | seqcount_init(&disk->part0.nr_sects_seq); | |
6c71013e ML |
1392 | if (hd_ref_init(&disk->part0)) { |
1393 | hd_free_part(&disk->part0); | |
1394 | kfree(disk); | |
1395 | return NULL; | |
1396 | } | |
b5d0b9df | 1397 | |
1da177e4 | 1398 | disk->minors = minors; |
1da177e4 | 1399 | rand_initialize_disk(disk); |
ed9e1982 TH |
1400 | disk_to_dev(disk)->class = &block_class; |
1401 | disk_to_dev(disk)->type = &disk_type; | |
1402 | device_initialize(disk_to_dev(disk)); | |
1da177e4 LT |
1403 | } |
1404 | return disk; | |
1405 | } | |
1946089a | 1406 | EXPORT_SYMBOL(alloc_disk_node); |
1da177e4 LT |
1407 | |
1408 | struct kobject *get_disk(struct gendisk *disk) | |
1409 | { | |
1410 | struct module *owner; | |
1411 | struct kobject *kobj; | |
1412 | ||
1413 | if (!disk->fops) | |
1414 | return NULL; | |
1415 | owner = disk->fops->owner; | |
1416 | if (owner && !try_module_get(owner)) | |
1417 | return NULL; | |
d01b2dcb | 1418 | kobj = kobject_get_unless_zero(&disk_to_dev(disk)->kobj); |
1da177e4 LT |
1419 | if (kobj == NULL) { |
1420 | module_put(owner); | |
1421 | return NULL; | |
1422 | } | |
1423 | return kobj; | |
1424 | ||
1425 | } | |
1426 | ||
1427 | EXPORT_SYMBOL(get_disk); | |
1428 | ||
1429 | void put_disk(struct gendisk *disk) | |
1430 | { | |
1431 | if (disk) | |
ed9e1982 | 1432 | kobject_put(&disk_to_dev(disk)->kobj); |
1da177e4 LT |
1433 | } |
1434 | ||
1435 | EXPORT_SYMBOL(put_disk); | |
1436 | ||
e3264a4d HR |
1437 | static void set_disk_ro_uevent(struct gendisk *gd, int ro) |
1438 | { | |
1439 | char event[] = "DISK_RO=1"; | |
1440 | char *envp[] = { event, NULL }; | |
1441 | ||
1442 | if (!ro) | |
1443 | event[8] = '0'; | |
1444 | kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); | |
1445 | } | |
1446 | ||
1da177e4 LT |
1447 | void set_device_ro(struct block_device *bdev, int flag) |
1448 | { | |
b7db9956 | 1449 | bdev->bd_part->policy = flag; |
1da177e4 LT |
1450 | } |
1451 | ||
1452 | EXPORT_SYMBOL(set_device_ro); | |
1453 | ||
1454 | void set_disk_ro(struct gendisk *disk, int flag) | |
1455 | { | |
e71bf0d0 TH |
1456 | struct disk_part_iter piter; |
1457 | struct hd_struct *part; | |
1458 | ||
e3264a4d HR |
1459 | if (disk->part0.policy != flag) { |
1460 | set_disk_ro_uevent(disk, flag); | |
1461 | disk->part0.policy = flag; | |
1462 | } | |
1463 | ||
1464 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY); | |
e71bf0d0 TH |
1465 | while ((part = disk_part_iter_next(&piter))) |
1466 | part->policy = flag; | |
1467 | disk_part_iter_exit(&piter); | |
1da177e4 LT |
1468 | } |
1469 | ||
1470 | EXPORT_SYMBOL(set_disk_ro); | |
1471 | ||
1472 | int bdev_read_only(struct block_device *bdev) | |
1473 | { | |
1474 | if (!bdev) | |
1475 | return 0; | |
b7db9956 | 1476 | return bdev->bd_part->policy; |
1da177e4 LT |
1477 | } |
1478 | ||
1479 | EXPORT_SYMBOL(bdev_read_only); | |
1480 | ||
cf771cb5 | 1481 | int invalidate_partition(struct gendisk *disk, int partno) |
1da177e4 LT |
1482 | { |
1483 | int res = 0; | |
cf771cb5 | 1484 | struct block_device *bdev = bdget_disk(disk, partno); |
1da177e4 | 1485 | if (bdev) { |
2ef41634 | 1486 | fsync_bdev(bdev); |
93b270f7 | 1487 | res = __invalidate_device(bdev, true); |
1da177e4 LT |
1488 | bdput(bdev); |
1489 | } | |
1490 | return res; | |
1491 | } | |
1492 | ||
1493 | EXPORT_SYMBOL(invalidate_partition); | |
77ea887e TH |
1494 | |
1495 | /* | |
1496 | * Disk events - monitor disk events like media change and eject request. | |
1497 | */ | |
1498 | struct disk_events { | |
1499 | struct list_head node; /* all disk_event's */ | |
1500 | struct gendisk *disk; /* the associated disk */ | |
1501 | spinlock_t lock; | |
1502 | ||
fdd514e1 | 1503 | struct mutex block_mutex; /* protects blocking */ |
77ea887e TH |
1504 | int block; /* event blocking depth */ |
1505 | unsigned int pending; /* events already sent out */ | |
1506 | unsigned int clearing; /* events being cleared */ | |
1507 | ||
1508 | long poll_msecs; /* interval, -1 for default */ | |
1509 | struct delayed_work dwork; | |
1510 | }; | |
1511 | ||
1512 | static const char *disk_events_strs[] = { | |
1513 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change", | |
1514 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request", | |
1515 | }; | |
1516 | ||
1517 | static char *disk_uevents[] = { | |
1518 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1", | |
1519 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1", | |
1520 | }; | |
1521 | ||
1522 | /* list of all disk_events */ | |
1523 | static DEFINE_MUTEX(disk_events_mutex); | |
1524 | static LIST_HEAD(disk_events); | |
1525 | ||
1526 | /* disable in-kernel polling by default */ | |
1fe8f348 | 1527 | static unsigned long disk_events_dfl_poll_msecs; |
77ea887e TH |
1528 | |
1529 | static unsigned long disk_events_poll_jiffies(struct gendisk *disk) | |
1530 | { | |
1531 | struct disk_events *ev = disk->ev; | |
1532 | long intv_msecs = 0; | |
1533 | ||
1534 | /* | |
1535 | * If device-specific poll interval is set, always use it. If | |
1536 | * the default is being used, poll iff there are events which | |
1537 | * can't be monitored asynchronously. | |
1538 | */ | |
1539 | if (ev->poll_msecs >= 0) | |
1540 | intv_msecs = ev->poll_msecs; | |
1541 | else if (disk->events & ~disk->async_events) | |
1542 | intv_msecs = disk_events_dfl_poll_msecs; | |
1543 | ||
1544 | return msecs_to_jiffies(intv_msecs); | |
1545 | } | |
1546 | ||
c3af54af TH |
1547 | /** |
1548 | * disk_block_events - block and flush disk event checking | |
1549 | * @disk: disk to block events for | |
1550 | * | |
1551 | * On return from this function, it is guaranteed that event checking | |
1552 | * isn't in progress and won't happen until unblocked by | |
1553 | * disk_unblock_events(). Events blocking is counted and the actual | |
1554 | * unblocking happens after the matching number of unblocks are done. | |
1555 | * | |
1556 | * Note that this intentionally does not block event checking from | |
1557 | * disk_clear_events(). | |
1558 | * | |
1559 | * CONTEXT: | |
1560 | * Might sleep. | |
1561 | */ | |
1562 | void disk_block_events(struct gendisk *disk) | |
77ea887e TH |
1563 | { |
1564 | struct disk_events *ev = disk->ev; | |
1565 | unsigned long flags; | |
1566 | bool cancel; | |
1567 | ||
c3af54af TH |
1568 | if (!ev) |
1569 | return; | |
1570 | ||
fdd514e1 TH |
1571 | /* |
1572 | * Outer mutex ensures that the first blocker completes canceling | |
1573 | * the event work before further blockers are allowed to finish. | |
1574 | */ | |
1575 | mutex_lock(&ev->block_mutex); | |
1576 | ||
77ea887e TH |
1577 | spin_lock_irqsave(&ev->lock, flags); |
1578 | cancel = !ev->block++; | |
1579 | spin_unlock_irqrestore(&ev->lock, flags); | |
1580 | ||
c3af54af TH |
1581 | if (cancel) |
1582 | cancel_delayed_work_sync(&disk->ev->dwork); | |
fdd514e1 TH |
1583 | |
1584 | mutex_unlock(&ev->block_mutex); | |
77ea887e TH |
1585 | } |
1586 | ||
1587 | static void __disk_unblock_events(struct gendisk *disk, bool check_now) | |
1588 | { | |
1589 | struct disk_events *ev = disk->ev; | |
1590 | unsigned long intv; | |
1591 | unsigned long flags; | |
1592 | ||
1593 | spin_lock_irqsave(&ev->lock, flags); | |
1594 | ||
1595 | if (WARN_ON_ONCE(ev->block <= 0)) | |
1596 | goto out_unlock; | |
1597 | ||
1598 | if (--ev->block) | |
1599 | goto out_unlock; | |
1600 | ||
77ea887e | 1601 | intv = disk_events_poll_jiffies(disk); |
77ea887e | 1602 | if (check_now) |
695588f9 VK |
1603 | queue_delayed_work(system_freezable_power_efficient_wq, |
1604 | &ev->dwork, 0); | |
77ea887e | 1605 | else if (intv) |
695588f9 VK |
1606 | queue_delayed_work(system_freezable_power_efficient_wq, |
1607 | &ev->dwork, intv); | |
77ea887e TH |
1608 | out_unlock: |
1609 | spin_unlock_irqrestore(&ev->lock, flags); | |
1610 | } | |
1611 | ||
77ea887e TH |
1612 | /** |
1613 | * disk_unblock_events - unblock disk event checking | |
1614 | * @disk: disk to unblock events for | |
1615 | * | |
1616 | * Undo disk_block_events(). When the block count reaches zero, it | |
1617 | * starts events polling if configured. | |
1618 | * | |
1619 | * CONTEXT: | |
1620 | * Don't care. Safe to call from irq context. | |
1621 | */ | |
1622 | void disk_unblock_events(struct gendisk *disk) | |
1623 | { | |
1624 | if (disk->ev) | |
facc31dd | 1625 | __disk_unblock_events(disk, false); |
77ea887e TH |
1626 | } |
1627 | ||
1628 | /** | |
85ef06d1 TH |
1629 | * disk_flush_events - schedule immediate event checking and flushing |
1630 | * @disk: disk to check and flush events for | |
1631 | * @mask: events to flush | |
77ea887e | 1632 | * |
85ef06d1 TH |
1633 | * Schedule immediate event checking on @disk if not blocked. Events in |
1634 | * @mask are scheduled to be cleared from the driver. Note that this | |
1635 | * doesn't clear the events from @disk->ev. | |
77ea887e TH |
1636 | * |
1637 | * CONTEXT: | |
85ef06d1 | 1638 | * If @mask is non-zero must be called with bdev->bd_mutex held. |
77ea887e | 1639 | */ |
85ef06d1 | 1640 | void disk_flush_events(struct gendisk *disk, unsigned int mask) |
77ea887e | 1641 | { |
a9dce2a3 | 1642 | struct disk_events *ev = disk->ev; |
a9dce2a3 TH |
1643 | |
1644 | if (!ev) | |
1645 | return; | |
1646 | ||
85ef06d1 TH |
1647 | spin_lock_irq(&ev->lock); |
1648 | ev->clearing |= mask; | |
41f63c53 | 1649 | if (!ev->block) |
695588f9 VK |
1650 | mod_delayed_work(system_freezable_power_efficient_wq, |
1651 | &ev->dwork, 0); | |
85ef06d1 | 1652 | spin_unlock_irq(&ev->lock); |
77ea887e | 1653 | } |
77ea887e TH |
1654 | |
1655 | /** | |
1656 | * disk_clear_events - synchronously check, clear and return pending events | |
1657 | * @disk: disk to fetch and clear events from | |
da3dae54 | 1658 | * @mask: mask of events to be fetched and cleared |
77ea887e TH |
1659 | * |
1660 | * Disk events are synchronously checked and pending events in @mask | |
1661 | * are cleared and returned. This ignores the block count. | |
1662 | * | |
1663 | * CONTEXT: | |
1664 | * Might sleep. | |
1665 | */ | |
1666 | unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask) | |
1667 | { | |
1668 | const struct block_device_operations *bdops = disk->fops; | |
1669 | struct disk_events *ev = disk->ev; | |
1670 | unsigned int pending; | |
12c2bdb2 | 1671 | unsigned int clearing = mask; |
77ea887e TH |
1672 | |
1673 | if (!ev) { | |
1674 | /* for drivers still using the old ->media_changed method */ | |
1675 | if ((mask & DISK_EVENT_MEDIA_CHANGE) && | |
1676 | bdops->media_changed && bdops->media_changed(disk)) | |
1677 | return DISK_EVENT_MEDIA_CHANGE; | |
1678 | return 0; | |
1679 | } | |
1680 | ||
12c2bdb2 DB |
1681 | disk_block_events(disk); |
1682 | ||
1683 | /* | |
1684 | * store the union of mask and ev->clearing on the stack so that the | |
1685 | * race with disk_flush_events does not cause ambiguity (ev->clearing | |
1686 | * can still be modified even if events are blocked). | |
1687 | */ | |
77ea887e | 1688 | spin_lock_irq(&ev->lock); |
12c2bdb2 DB |
1689 | clearing |= ev->clearing; |
1690 | ev->clearing = 0; | |
77ea887e TH |
1691 | spin_unlock_irq(&ev->lock); |
1692 | ||
12c2bdb2 | 1693 | disk_check_events(ev, &clearing); |
aea24a8b | 1694 | /* |
12c2bdb2 DB |
1695 | * if ev->clearing is not 0, the disk_flush_events got called in the |
1696 | * middle of this function, so we want to run the workfn without delay. | |
aea24a8b | 1697 | */ |
12c2bdb2 | 1698 | __disk_unblock_events(disk, ev->clearing ? true : false); |
77ea887e TH |
1699 | |
1700 | /* then, fetch and clear pending events */ | |
1701 | spin_lock_irq(&ev->lock); | |
77ea887e TH |
1702 | pending = ev->pending & mask; |
1703 | ev->pending &= ~mask; | |
1704 | spin_unlock_irq(&ev->lock); | |
12c2bdb2 | 1705 | WARN_ON_ONCE(clearing & mask); |
77ea887e TH |
1706 | |
1707 | return pending; | |
1708 | } | |
1709 | ||
12c2bdb2 DB |
1710 | /* |
1711 | * Separate this part out so that a different pointer for clearing_ptr can be | |
1712 | * passed in for disk_clear_events. | |
1713 | */ | |
77ea887e TH |
1714 | static void disk_events_workfn(struct work_struct *work) |
1715 | { | |
1716 | struct delayed_work *dwork = to_delayed_work(work); | |
1717 | struct disk_events *ev = container_of(dwork, struct disk_events, dwork); | |
12c2bdb2 DB |
1718 | |
1719 | disk_check_events(ev, &ev->clearing); | |
1720 | } | |
1721 | ||
1722 | static void disk_check_events(struct disk_events *ev, | |
1723 | unsigned int *clearing_ptr) | |
1724 | { | |
77ea887e TH |
1725 | struct gendisk *disk = ev->disk; |
1726 | char *envp[ARRAY_SIZE(disk_uevents) + 1] = { }; | |
12c2bdb2 | 1727 | unsigned int clearing = *clearing_ptr; |
77ea887e TH |
1728 | unsigned int events; |
1729 | unsigned long intv; | |
1730 | int nr_events = 0, i; | |
1731 | ||
1732 | /* check events */ | |
1733 | events = disk->fops->check_events(disk, clearing); | |
1734 | ||
1735 | /* accumulate pending events and schedule next poll if necessary */ | |
1736 | spin_lock_irq(&ev->lock); | |
1737 | ||
1738 | events &= ~ev->pending; | |
1739 | ev->pending |= events; | |
12c2bdb2 | 1740 | *clearing_ptr &= ~clearing; |
77ea887e TH |
1741 | |
1742 | intv = disk_events_poll_jiffies(disk); | |
1743 | if (!ev->block && intv) | |
695588f9 VK |
1744 | queue_delayed_work(system_freezable_power_efficient_wq, |
1745 | &ev->dwork, intv); | |
77ea887e TH |
1746 | |
1747 | spin_unlock_irq(&ev->lock); | |
1748 | ||
7c88a168 TH |
1749 | /* |
1750 | * Tell userland about new events. Only the events listed in | |
1751 | * @disk->events are reported. Unlisted events are processed the | |
1752 | * same internally but never get reported to userland. | |
1753 | */ | |
77ea887e | 1754 | for (i = 0; i < ARRAY_SIZE(disk_uevents); i++) |
7c88a168 | 1755 | if (events & disk->events & (1 << i)) |
77ea887e TH |
1756 | envp[nr_events++] = disk_uevents[i]; |
1757 | ||
1758 | if (nr_events) | |
1759 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
1760 | } | |
1761 | ||
1762 | /* | |
1763 | * A disk events enabled device has the following sysfs nodes under | |
1764 | * its /sys/block/X/ directory. | |
1765 | * | |
1766 | * events : list of all supported events | |
1767 | * events_async : list of events which can be detected w/o polling | |
1768 | * events_poll_msecs : polling interval, 0: disable, -1: system default | |
1769 | */ | |
1770 | static ssize_t __disk_events_show(unsigned int events, char *buf) | |
1771 | { | |
1772 | const char *delim = ""; | |
1773 | ssize_t pos = 0; | |
1774 | int i; | |
1775 | ||
1776 | for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++) | |
1777 | if (events & (1 << i)) { | |
1778 | pos += sprintf(buf + pos, "%s%s", | |
1779 | delim, disk_events_strs[i]); | |
1780 | delim = " "; | |
1781 | } | |
1782 | if (pos) | |
1783 | pos += sprintf(buf + pos, "\n"); | |
1784 | return pos; | |
1785 | } | |
1786 | ||
1787 | static ssize_t disk_events_show(struct device *dev, | |
1788 | struct device_attribute *attr, char *buf) | |
1789 | { | |
1790 | struct gendisk *disk = dev_to_disk(dev); | |
1791 | ||
1792 | return __disk_events_show(disk->events, buf); | |
1793 | } | |
1794 | ||
1795 | static ssize_t disk_events_async_show(struct device *dev, | |
1796 | struct device_attribute *attr, char *buf) | |
1797 | { | |
1798 | struct gendisk *disk = dev_to_disk(dev); | |
1799 | ||
1800 | return __disk_events_show(disk->async_events, buf); | |
1801 | } | |
1802 | ||
1803 | static ssize_t disk_events_poll_msecs_show(struct device *dev, | |
1804 | struct device_attribute *attr, | |
1805 | char *buf) | |
1806 | { | |
1807 | struct gendisk *disk = dev_to_disk(dev); | |
1808 | ||
1809 | return sprintf(buf, "%ld\n", disk->ev->poll_msecs); | |
1810 | } | |
1811 | ||
1812 | static ssize_t disk_events_poll_msecs_store(struct device *dev, | |
1813 | struct device_attribute *attr, | |
1814 | const char *buf, size_t count) | |
1815 | { | |
1816 | struct gendisk *disk = dev_to_disk(dev); | |
1817 | long intv; | |
1818 | ||
1819 | if (!count || !sscanf(buf, "%ld", &intv)) | |
1820 | return -EINVAL; | |
1821 | ||
1822 | if (intv < 0 && intv != -1) | |
1823 | return -EINVAL; | |
1824 | ||
c3af54af | 1825 | disk_block_events(disk); |
77ea887e TH |
1826 | disk->ev->poll_msecs = intv; |
1827 | __disk_unblock_events(disk, true); | |
1828 | ||
1829 | return count; | |
1830 | } | |
1831 | ||
1832 | static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL); | |
1833 | static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL); | |
1834 | static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR, | |
1835 | disk_events_poll_msecs_show, | |
1836 | disk_events_poll_msecs_store); | |
1837 | ||
1838 | static const struct attribute *disk_events_attrs[] = { | |
1839 | &dev_attr_events.attr, | |
1840 | &dev_attr_events_async.attr, | |
1841 | &dev_attr_events_poll_msecs.attr, | |
1842 | NULL, | |
1843 | }; | |
1844 | ||
1845 | /* | |
1846 | * The default polling interval can be specified by the kernel | |
1847 | * parameter block.events_dfl_poll_msecs which defaults to 0 | |
1848 | * (disable). This can also be modified runtime by writing to | |
1849 | * /sys/module/block/events_dfl_poll_msecs. | |
1850 | */ | |
1851 | static int disk_events_set_dfl_poll_msecs(const char *val, | |
1852 | const struct kernel_param *kp) | |
1853 | { | |
1854 | struct disk_events *ev; | |
1855 | int ret; | |
1856 | ||
1857 | ret = param_set_ulong(val, kp); | |
1858 | if (ret < 0) | |
1859 | return ret; | |
1860 | ||
1861 | mutex_lock(&disk_events_mutex); | |
1862 | ||
1863 | list_for_each_entry(ev, &disk_events, node) | |
85ef06d1 | 1864 | disk_flush_events(ev->disk, 0); |
77ea887e TH |
1865 | |
1866 | mutex_unlock(&disk_events_mutex); | |
1867 | ||
1868 | return 0; | |
1869 | } | |
1870 | ||
1871 | static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = { | |
1872 | .set = disk_events_set_dfl_poll_msecs, | |
1873 | .get = param_get_ulong, | |
1874 | }; | |
1875 | ||
1876 | #undef MODULE_PARAM_PREFIX | |
1877 | #define MODULE_PARAM_PREFIX "block." | |
1878 | ||
1879 | module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops, | |
1880 | &disk_events_dfl_poll_msecs, 0644); | |
1881 | ||
1882 | /* | |
9f53d2fe | 1883 | * disk_{alloc|add|del|release}_events - initialize and destroy disk_events. |
77ea887e | 1884 | */ |
9f53d2fe | 1885 | static void disk_alloc_events(struct gendisk *disk) |
77ea887e TH |
1886 | { |
1887 | struct disk_events *ev; | |
1888 | ||
75e3f3ee | 1889 | if (!disk->fops->check_events) |
77ea887e TH |
1890 | return; |
1891 | ||
1892 | ev = kzalloc(sizeof(*ev), GFP_KERNEL); | |
1893 | if (!ev) { | |
1894 | pr_warn("%s: failed to initialize events\n", disk->disk_name); | |
1895 | return; | |
1896 | } | |
1897 | ||
77ea887e TH |
1898 | INIT_LIST_HEAD(&ev->node); |
1899 | ev->disk = disk; | |
1900 | spin_lock_init(&ev->lock); | |
fdd514e1 | 1901 | mutex_init(&ev->block_mutex); |
77ea887e TH |
1902 | ev->block = 1; |
1903 | ev->poll_msecs = -1; | |
1904 | INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn); | |
1905 | ||
9f53d2fe SG |
1906 | disk->ev = ev; |
1907 | } | |
1908 | ||
1909 | static void disk_add_events(struct gendisk *disk) | |
1910 | { | |
1911 | if (!disk->ev) | |
1912 | return; | |
1913 | ||
1914 | /* FIXME: error handling */ | |
1915 | if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0) | |
1916 | pr_warn("%s: failed to create sysfs files for events\n", | |
1917 | disk->disk_name); | |
1918 | ||
77ea887e | 1919 | mutex_lock(&disk_events_mutex); |
9f53d2fe | 1920 | list_add_tail(&disk->ev->node, &disk_events); |
77ea887e TH |
1921 | mutex_unlock(&disk_events_mutex); |
1922 | ||
1923 | /* | |
1924 | * Block count is initialized to 1 and the following initial | |
1925 | * unblock kicks it into action. | |
1926 | */ | |
1927 | __disk_unblock_events(disk, true); | |
1928 | } | |
1929 | ||
1930 | static void disk_del_events(struct gendisk *disk) | |
1931 | { | |
1932 | if (!disk->ev) | |
1933 | return; | |
1934 | ||
c3af54af | 1935 | disk_block_events(disk); |
77ea887e TH |
1936 | |
1937 | mutex_lock(&disk_events_mutex); | |
1938 | list_del_init(&disk->ev->node); | |
1939 | mutex_unlock(&disk_events_mutex); | |
1940 | ||
1941 | sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs); | |
1942 | } | |
1943 | ||
1944 | static void disk_release_events(struct gendisk *disk) | |
1945 | { | |
1946 | /* the block count should be 1 from disk_del_events() */ | |
1947 | WARN_ON_ONCE(disk->ev && disk->ev->block != 1); | |
1948 | kfree(disk->ev); | |
1949 | } |