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 | ||
1826eadf AB |
39 | static struct device_type disk_type; |
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 | ||
e71bf0d0 TH |
48 | /** |
49 | * disk_get_part - get partition | |
50 | * @disk: disk to look partition from | |
51 | * @partno: partition number | |
52 | * | |
53 | * Look for partition @partno from @disk. If found, increment | |
54 | * reference count and return it. | |
55 | * | |
56 | * CONTEXT: | |
57 | * Don't care. | |
58 | * | |
59 | * RETURNS: | |
60 | * Pointer to the found partition on success, NULL if not found. | |
61 | */ | |
62 | struct hd_struct *disk_get_part(struct gendisk *disk, int partno) | |
63 | { | |
540eed56 TH |
64 | struct hd_struct *part = NULL; |
65 | struct disk_part_tbl *ptbl; | |
e71bf0d0 | 66 | |
540eed56 | 67 | if (unlikely(partno < 0)) |
e71bf0d0 | 68 | return NULL; |
540eed56 | 69 | |
e71bf0d0 | 70 | rcu_read_lock(); |
540eed56 TH |
71 | |
72 | ptbl = rcu_dereference(disk->part_tbl); | |
73 | if (likely(partno < ptbl->len)) { | |
74 | part = rcu_dereference(ptbl->part[partno]); | |
75 | if (part) | |
76 | get_device(part_to_dev(part)); | |
77 | } | |
78 | ||
e71bf0d0 TH |
79 | rcu_read_unlock(); |
80 | ||
81 | return part; | |
82 | } | |
83 | EXPORT_SYMBOL_GPL(disk_get_part); | |
84 | ||
85 | /** | |
86 | * disk_part_iter_init - initialize partition iterator | |
87 | * @piter: iterator to initialize | |
88 | * @disk: disk to iterate over | |
89 | * @flags: DISK_PITER_* flags | |
90 | * | |
91 | * Initialize @piter so that it iterates over partitions of @disk. | |
92 | * | |
93 | * CONTEXT: | |
94 | * Don't care. | |
95 | */ | |
96 | void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk, | |
97 | unsigned int flags) | |
98 | { | |
540eed56 TH |
99 | struct disk_part_tbl *ptbl; |
100 | ||
101 | rcu_read_lock(); | |
102 | ptbl = rcu_dereference(disk->part_tbl); | |
103 | ||
e71bf0d0 TH |
104 | piter->disk = disk; |
105 | piter->part = NULL; | |
106 | ||
107 | if (flags & DISK_PITER_REVERSE) | |
540eed56 | 108 | piter->idx = ptbl->len - 1; |
71982a40 | 109 | else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0)) |
e71bf0d0 | 110 | piter->idx = 0; |
b5d0b9df TH |
111 | else |
112 | piter->idx = 1; | |
e71bf0d0 TH |
113 | |
114 | piter->flags = flags; | |
540eed56 TH |
115 | |
116 | rcu_read_unlock(); | |
e71bf0d0 TH |
117 | } |
118 | EXPORT_SYMBOL_GPL(disk_part_iter_init); | |
119 | ||
120 | /** | |
121 | * disk_part_iter_next - proceed iterator to the next partition and return it | |
122 | * @piter: iterator of interest | |
123 | * | |
124 | * Proceed @piter to the next partition and return it. | |
125 | * | |
126 | * CONTEXT: | |
127 | * Don't care. | |
128 | */ | |
129 | struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter) | |
130 | { | |
540eed56 | 131 | struct disk_part_tbl *ptbl; |
e71bf0d0 TH |
132 | int inc, end; |
133 | ||
134 | /* put the last partition */ | |
135 | disk_put_part(piter->part); | |
136 | piter->part = NULL; | |
137 | ||
540eed56 | 138 | /* get part_tbl */ |
e71bf0d0 | 139 | rcu_read_lock(); |
540eed56 | 140 | ptbl = rcu_dereference(piter->disk->part_tbl); |
e71bf0d0 TH |
141 | |
142 | /* determine iteration parameters */ | |
143 | if (piter->flags & DISK_PITER_REVERSE) { | |
144 | inc = -1; | |
71982a40 TH |
145 | if (piter->flags & (DISK_PITER_INCL_PART0 | |
146 | DISK_PITER_INCL_EMPTY_PART0)) | |
b5d0b9df TH |
147 | end = -1; |
148 | else | |
149 | end = 0; | |
e71bf0d0 TH |
150 | } else { |
151 | inc = 1; | |
540eed56 | 152 | end = ptbl->len; |
e71bf0d0 TH |
153 | } |
154 | ||
155 | /* iterate to the next partition */ | |
156 | for (; piter->idx != end; piter->idx += inc) { | |
157 | struct hd_struct *part; | |
158 | ||
540eed56 | 159 | part = rcu_dereference(ptbl->part[piter->idx]); |
e71bf0d0 TH |
160 | if (!part) |
161 | continue; | |
c83f6bf9 | 162 | if (!part_nr_sects_read(part) && |
71982a40 TH |
163 | !(piter->flags & DISK_PITER_INCL_EMPTY) && |
164 | !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 && | |
165 | piter->idx == 0)) | |
e71bf0d0 TH |
166 | continue; |
167 | ||
ed9e1982 | 168 | get_device(part_to_dev(part)); |
e71bf0d0 TH |
169 | piter->part = part; |
170 | piter->idx += inc; | |
171 | break; | |
172 | } | |
173 | ||
174 | rcu_read_unlock(); | |
175 | ||
176 | return piter->part; | |
177 | } | |
178 | EXPORT_SYMBOL_GPL(disk_part_iter_next); | |
179 | ||
180 | /** | |
181 | * disk_part_iter_exit - finish up partition iteration | |
182 | * @piter: iter of interest | |
183 | * | |
184 | * Called when iteration is over. Cleans up @piter. | |
185 | * | |
186 | * CONTEXT: | |
187 | * Don't care. | |
188 | */ | |
189 | void disk_part_iter_exit(struct disk_part_iter *piter) | |
190 | { | |
191 | disk_put_part(piter->part); | |
192 | piter->part = NULL; | |
193 | } | |
194 | EXPORT_SYMBOL_GPL(disk_part_iter_exit); | |
195 | ||
a6f23657 JA |
196 | static inline int sector_in_part(struct hd_struct *part, sector_t sector) |
197 | { | |
198 | return part->start_sect <= sector && | |
c83f6bf9 | 199 | sector < part->start_sect + part_nr_sects_read(part); |
a6f23657 JA |
200 | } |
201 | ||
e71bf0d0 TH |
202 | /** |
203 | * disk_map_sector_rcu - map sector to partition | |
204 | * @disk: gendisk of interest | |
205 | * @sector: sector to map | |
206 | * | |
207 | * Find out which partition @sector maps to on @disk. This is | |
208 | * primarily used for stats accounting. | |
209 | * | |
210 | * CONTEXT: | |
211 | * RCU read locked. The returned partition pointer is valid only | |
212 | * while preemption is disabled. | |
213 | * | |
214 | * RETURNS: | |
074a7aca | 215 | * Found partition on success, part0 is returned if no partition matches |
e71bf0d0 TH |
216 | */ |
217 | struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector) | |
218 | { | |
540eed56 | 219 | struct disk_part_tbl *ptbl; |
a6f23657 | 220 | struct hd_struct *part; |
e71bf0d0 TH |
221 | int i; |
222 | ||
540eed56 TH |
223 | ptbl = rcu_dereference(disk->part_tbl); |
224 | ||
a6f23657 JA |
225 | part = rcu_dereference(ptbl->last_lookup); |
226 | if (part && sector_in_part(part, sector)) | |
227 | return part; | |
228 | ||
540eed56 | 229 | for (i = 1; i < ptbl->len; i++) { |
a6f23657 | 230 | part = rcu_dereference(ptbl->part[i]); |
e71bf0d0 | 231 | |
a6f23657 JA |
232 | if (part && sector_in_part(part, sector)) { |
233 | rcu_assign_pointer(ptbl->last_lookup, part); | |
e71bf0d0 | 234 | return part; |
a6f23657 | 235 | } |
e71bf0d0 | 236 | } |
074a7aca | 237 | return &disk->part0; |
e71bf0d0 TH |
238 | } |
239 | EXPORT_SYMBOL_GPL(disk_map_sector_rcu); | |
240 | ||
1da177e4 LT |
241 | /* |
242 | * Can be deleted altogether. Later. | |
243 | * | |
244 | */ | |
245 | static struct blk_major_name { | |
246 | struct blk_major_name *next; | |
247 | int major; | |
248 | char name[16]; | |
68eef3b4 | 249 | } *major_names[BLKDEV_MAJOR_HASH_SIZE]; |
1da177e4 LT |
250 | |
251 | /* index in the above - for now: assume no multimajor ranges */ | |
e61eb2e9 | 252 | static inline int major_to_index(unsigned major) |
1da177e4 | 253 | { |
68eef3b4 | 254 | return major % BLKDEV_MAJOR_HASH_SIZE; |
7170be5f NH |
255 | } |
256 | ||
68eef3b4 | 257 | #ifdef CONFIG_PROC_FS |
cf771cb5 | 258 | void blkdev_show(struct seq_file *seqf, off_t offset) |
7170be5f | 259 | { |
68eef3b4 | 260 | struct blk_major_name *dp; |
7170be5f | 261 | |
68eef3b4 | 262 | if (offset < BLKDEV_MAJOR_HASH_SIZE) { |
edfaa7c3 | 263 | mutex_lock(&block_class_lock); |
68eef3b4 | 264 | for (dp = major_names[offset]; dp; dp = dp->next) |
cf771cb5 | 265 | seq_printf(seqf, "%3d %s\n", dp->major, dp->name); |
edfaa7c3 | 266 | mutex_unlock(&block_class_lock); |
1da177e4 | 267 | } |
1da177e4 | 268 | } |
68eef3b4 | 269 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 270 | |
9e8c0bcc MN |
271 | /** |
272 | * register_blkdev - register a new block device | |
273 | * | |
274 | * @major: the requested major device number [1..255]. If @major=0, try to | |
275 | * allocate any unused major number. | |
276 | * @name: the name of the new block device as a zero terminated string | |
277 | * | |
278 | * The @name must be unique within the system. | |
279 | * | |
280 | * The return value depends on the @major input parameter. | |
281 | * - if a major device number was requested in range [1..255] then the | |
282 | * function returns zero on success, or a negative error code | |
283 | * - if any unused major number was requested with @major=0 parameter | |
284 | * then the return value is the allocated major number in range | |
285 | * [1..255] or a negative error code otherwise | |
286 | */ | |
1da177e4 LT |
287 | int register_blkdev(unsigned int major, const char *name) |
288 | { | |
289 | struct blk_major_name **n, *p; | |
290 | int index, ret = 0; | |
291 | ||
edfaa7c3 | 292 | mutex_lock(&block_class_lock); |
1da177e4 LT |
293 | |
294 | /* temporary */ | |
295 | if (major == 0) { | |
296 | for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) { | |
297 | if (major_names[index] == NULL) | |
298 | break; | |
299 | } | |
300 | ||
301 | if (index == 0) { | |
302 | printk("register_blkdev: failed to get major for %s\n", | |
303 | name); | |
304 | ret = -EBUSY; | |
305 | goto out; | |
306 | } | |
307 | major = index; | |
308 | ret = major; | |
309 | } | |
310 | ||
311 | p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL); | |
312 | if (p == NULL) { | |
313 | ret = -ENOMEM; | |
314 | goto out; | |
315 | } | |
316 | ||
317 | p->major = major; | |
318 | strlcpy(p->name, name, sizeof(p->name)); | |
319 | p->next = NULL; | |
320 | index = major_to_index(major); | |
321 | ||
322 | for (n = &major_names[index]; *n; n = &(*n)->next) { | |
323 | if ((*n)->major == major) | |
324 | break; | |
325 | } | |
326 | if (!*n) | |
327 | *n = p; | |
328 | else | |
329 | ret = -EBUSY; | |
330 | ||
331 | if (ret < 0) { | |
332 | printk("register_blkdev: cannot get major %d for %s\n", | |
333 | major, name); | |
334 | kfree(p); | |
335 | } | |
336 | out: | |
edfaa7c3 | 337 | mutex_unlock(&block_class_lock); |
1da177e4 LT |
338 | return ret; |
339 | } | |
340 | ||
341 | EXPORT_SYMBOL(register_blkdev); | |
342 | ||
f4480240 | 343 | void unregister_blkdev(unsigned int major, const char *name) |
1da177e4 LT |
344 | { |
345 | struct blk_major_name **n; | |
346 | struct blk_major_name *p = NULL; | |
347 | int index = major_to_index(major); | |
1da177e4 | 348 | |
edfaa7c3 | 349 | mutex_lock(&block_class_lock); |
1da177e4 LT |
350 | for (n = &major_names[index]; *n; n = &(*n)->next) |
351 | if ((*n)->major == major) | |
352 | break; | |
294462a5 AM |
353 | if (!*n || strcmp((*n)->name, name)) { |
354 | WARN_ON(1); | |
294462a5 | 355 | } else { |
1da177e4 LT |
356 | p = *n; |
357 | *n = p->next; | |
358 | } | |
edfaa7c3 | 359 | mutex_unlock(&block_class_lock); |
1da177e4 | 360 | kfree(p); |
1da177e4 LT |
361 | } |
362 | ||
363 | EXPORT_SYMBOL(unregister_blkdev); | |
364 | ||
365 | static struct kobj_map *bdev_map; | |
366 | ||
870d6656 TH |
367 | /** |
368 | * blk_mangle_minor - scatter minor numbers apart | |
369 | * @minor: minor number to mangle | |
370 | * | |
371 | * Scatter consecutively allocated @minor number apart if MANGLE_DEVT | |
372 | * is enabled. Mangling twice gives the original value. | |
373 | * | |
374 | * RETURNS: | |
375 | * Mangled value. | |
376 | * | |
377 | * CONTEXT: | |
378 | * Don't care. | |
379 | */ | |
380 | static int blk_mangle_minor(int minor) | |
381 | { | |
382 | #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT | |
383 | int i; | |
384 | ||
385 | for (i = 0; i < MINORBITS / 2; i++) { | |
386 | int low = minor & (1 << i); | |
387 | int high = minor & (1 << (MINORBITS - 1 - i)); | |
388 | int distance = MINORBITS - 1 - 2 * i; | |
389 | ||
390 | minor ^= low | high; /* clear both bits */ | |
391 | low <<= distance; /* swap the positions */ | |
392 | high >>= distance; | |
393 | minor |= low | high; /* and set */ | |
394 | } | |
395 | #endif | |
396 | return minor; | |
397 | } | |
398 | ||
bcce3de1 TH |
399 | /** |
400 | * blk_alloc_devt - allocate a dev_t for a partition | |
401 | * @part: partition to allocate dev_t for | |
bcce3de1 TH |
402 | * @devt: out parameter for resulting dev_t |
403 | * | |
404 | * Allocate a dev_t for block device. | |
405 | * | |
406 | * RETURNS: | |
407 | * 0 on success, allocated dev_t is returned in *@devt. -errno on | |
408 | * failure. | |
409 | * | |
410 | * CONTEXT: | |
411 | * Might sleep. | |
412 | */ | |
413 | int blk_alloc_devt(struct hd_struct *part, dev_t *devt) | |
414 | { | |
415 | struct gendisk *disk = part_to_disk(part); | |
bab998d6 | 416 | int idx; |
bcce3de1 TH |
417 | |
418 | /* in consecutive minor range? */ | |
419 | if (part->partno < disk->minors) { | |
420 | *devt = MKDEV(disk->major, disk->first_minor + part->partno); | |
421 | return 0; | |
422 | } | |
423 | ||
424 | /* allocate ext devt */ | |
2da78092 KB |
425 | idr_preload(GFP_KERNEL); |
426 | ||
4d66e5e9 | 427 | spin_lock_bh(&ext_devt_lock); |
2da78092 | 428 | idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_NOWAIT); |
4d66e5e9 | 429 | spin_unlock_bh(&ext_devt_lock); |
2da78092 KB |
430 | |
431 | idr_preload_end(); | |
bab998d6 TH |
432 | if (idx < 0) |
433 | return idx == -ENOSPC ? -EBUSY : idx; | |
bcce3de1 | 434 | |
870d6656 | 435 | *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx)); |
bcce3de1 TH |
436 | return 0; |
437 | } | |
438 | ||
439 | /** | |
440 | * blk_free_devt - free a dev_t | |
441 | * @devt: dev_t to free | |
442 | * | |
443 | * Free @devt which was allocated using blk_alloc_devt(). | |
444 | * | |
445 | * CONTEXT: | |
446 | * Might sleep. | |
447 | */ | |
448 | void blk_free_devt(dev_t devt) | |
449 | { | |
bcce3de1 TH |
450 | if (devt == MKDEV(0, 0)) |
451 | return; | |
452 | ||
453 | if (MAJOR(devt) == BLOCK_EXT_MAJOR) { | |
4d66e5e9 | 454 | spin_lock_bh(&ext_devt_lock); |
870d6656 | 455 | idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); |
4d66e5e9 | 456 | spin_unlock_bh(&ext_devt_lock); |
bcce3de1 TH |
457 | } |
458 | } | |
459 | ||
1f014290 TH |
460 | static char *bdevt_str(dev_t devt, char *buf) |
461 | { | |
462 | if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) { | |
463 | char tbuf[BDEVT_SIZE]; | |
464 | snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt)); | |
465 | snprintf(buf, BDEVT_SIZE, "%-9s", tbuf); | |
466 | } else | |
467 | snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt)); | |
468 | ||
469 | return buf; | |
470 | } | |
471 | ||
1da177e4 LT |
472 | /* |
473 | * Register device numbers dev..(dev+range-1) | |
474 | * range must be nonzero | |
475 | * The hash chain is sorted on range, so that subranges can override. | |
476 | */ | |
edfaa7c3 | 477 | void blk_register_region(dev_t devt, unsigned long range, struct module *module, |
1da177e4 LT |
478 | struct kobject *(*probe)(dev_t, int *, void *), |
479 | int (*lock)(dev_t, void *), void *data) | |
480 | { | |
edfaa7c3 | 481 | kobj_map(bdev_map, devt, range, module, probe, lock, data); |
1da177e4 LT |
482 | } |
483 | ||
484 | EXPORT_SYMBOL(blk_register_region); | |
485 | ||
edfaa7c3 | 486 | void blk_unregister_region(dev_t devt, unsigned long range) |
1da177e4 | 487 | { |
edfaa7c3 | 488 | kobj_unmap(bdev_map, devt, range); |
1da177e4 LT |
489 | } |
490 | ||
491 | EXPORT_SYMBOL(blk_unregister_region); | |
492 | ||
cf771cb5 | 493 | static struct kobject *exact_match(dev_t devt, int *partno, void *data) |
1da177e4 LT |
494 | { |
495 | struct gendisk *p = data; | |
edfaa7c3 | 496 | |
ed9e1982 | 497 | return &disk_to_dev(p)->kobj; |
1da177e4 LT |
498 | } |
499 | ||
edfaa7c3 | 500 | static int exact_lock(dev_t devt, void *data) |
1da177e4 LT |
501 | { |
502 | struct gendisk *p = data; | |
503 | ||
504 | if (!get_disk(p)) | |
505 | return -1; | |
506 | return 0; | |
507 | } | |
508 | ||
e63a46be | 509 | static void register_disk(struct device *parent, struct gendisk *disk) |
d2bf1b67 TH |
510 | { |
511 | struct device *ddev = disk_to_dev(disk); | |
512 | struct block_device *bdev; | |
513 | struct disk_part_iter piter; | |
514 | struct hd_struct *part; | |
515 | int err; | |
516 | ||
e63a46be | 517 | ddev->parent = parent; |
d2bf1b67 | 518 | |
ffc8b308 | 519 | dev_set_name(ddev, "%s", disk->disk_name); |
d2bf1b67 TH |
520 | |
521 | /* delay uevents, until we scanned partition table */ | |
522 | dev_set_uevent_suppress(ddev, 1); | |
523 | ||
524 | if (device_add(ddev)) | |
525 | return; | |
526 | if (!sysfs_deprecated) { | |
527 | err = sysfs_create_link(block_depr, &ddev->kobj, | |
528 | kobject_name(&ddev->kobj)); | |
529 | if (err) { | |
530 | device_del(ddev); | |
531 | return; | |
532 | } | |
533 | } | |
25e823c8 ML |
534 | |
535 | /* | |
536 | * avoid probable deadlock caused by allocating memory with | |
537 | * GFP_KERNEL in runtime_resume callback of its all ancestor | |
538 | * devices | |
539 | */ | |
540 | pm_runtime_set_memalloc_noio(ddev, true); | |
541 | ||
d2bf1b67 TH |
542 | disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj); |
543 | disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj); | |
544 | ||
545 | /* No minors to use for partitions */ | |
d27769ec | 546 | if (!disk_part_scan_enabled(disk)) |
d2bf1b67 TH |
547 | goto exit; |
548 | ||
549 | /* No such device (e.g., media were just removed) */ | |
550 | if (!get_capacity(disk)) | |
551 | goto exit; | |
552 | ||
553 | bdev = bdget_disk(disk, 0); | |
554 | if (!bdev) | |
555 | goto exit; | |
556 | ||
557 | bdev->bd_invalidated = 1; | |
558 | err = blkdev_get(bdev, FMODE_READ, NULL); | |
559 | if (err < 0) | |
560 | goto exit; | |
561 | blkdev_put(bdev, FMODE_READ); | |
562 | ||
563 | exit: | |
564 | /* announce disk after possible partitions are created */ | |
565 | dev_set_uevent_suppress(ddev, 0); | |
566 | kobject_uevent(&ddev->kobj, KOBJ_ADD); | |
567 | ||
568 | /* announce possible partitions */ | |
569 | disk_part_iter_init(&piter, disk, 0); | |
570 | while ((part = disk_part_iter_next(&piter))) | |
571 | kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD); | |
572 | disk_part_iter_exit(&piter); | |
573 | } | |
574 | ||
0dba1314 DW |
575 | void put_disk_devt(struct disk_devt *disk_devt) |
576 | { | |
577 | if (disk_devt && atomic_dec_and_test(&disk_devt->count)) | |
578 | disk_devt->release(disk_devt); | |
579 | } | |
580 | EXPORT_SYMBOL(put_disk_devt); | |
581 | ||
582 | void get_disk_devt(struct disk_devt *disk_devt) | |
583 | { | |
584 | if (disk_devt) | |
585 | atomic_inc(&disk_devt->count); | |
586 | } | |
587 | EXPORT_SYMBOL(get_disk_devt); | |
588 | ||
1da177e4 | 589 | /** |
e63a46be DW |
590 | * device_add_disk - add partitioning information to kernel list |
591 | * @parent: parent device for the disk | |
1da177e4 LT |
592 | * @disk: per-device partitioning information |
593 | * | |
594 | * This function registers the partitioning information in @disk | |
595 | * with the kernel. | |
3e1a7ff8 TH |
596 | * |
597 | * FIXME: error handling | |
1da177e4 | 598 | */ |
e63a46be | 599 | void device_add_disk(struct device *parent, struct gendisk *disk) |
1da177e4 | 600 | { |
cf0ca9fe | 601 | struct backing_dev_info *bdi; |
3e1a7ff8 | 602 | dev_t devt; |
6ffeea77 | 603 | int retval; |
cf0ca9fe | 604 | |
3e1a7ff8 TH |
605 | /* minors == 0 indicates to use ext devt from part0 and should |
606 | * be accompanied with EXT_DEVT flag. Make sure all | |
607 | * parameters make sense. | |
608 | */ | |
609 | WARN_ON(disk->minors && !(disk->major || disk->first_minor)); | |
610 | WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT)); | |
611 | ||
1da177e4 | 612 | disk->flags |= GENHD_FL_UP; |
3e1a7ff8 TH |
613 | |
614 | retval = blk_alloc_devt(&disk->part0, &devt); | |
615 | if (retval) { | |
616 | WARN_ON(1); | |
617 | return; | |
618 | } | |
619 | disk_to_dev(disk)->devt = devt; | |
620 | ||
621 | /* ->major and ->first_minor aren't supposed to be | |
622 | * dereferenced from here on, but set them just in case. | |
623 | */ | |
624 | disk->major = MAJOR(devt); | |
625 | disk->first_minor = MINOR(devt); | |
626 | ||
9f53d2fe SG |
627 | disk_alloc_events(disk); |
628 | ||
0dba1314 DW |
629 | /* |
630 | * Take a reference on the devt and assign it to queue since it | |
631 | * must not be reallocated while the bdi is registered | |
632 | */ | |
633 | disk->queue->disk_devt = disk->disk_devt; | |
634 | get_disk_devt(disk->disk_devt); | |
635 | ||
9f5e4865 | 636 | /* Register BDI before referencing it from bdev */ |
dc3b17cc | 637 | bdi = disk->queue->backing_dev_info; |
df08c32c | 638 | bdi_register_owner(bdi, disk_to_dev(disk)); |
01ea5063 | 639 | |
f331c029 TH |
640 | blk_register_region(disk_devt(disk), disk->minors, NULL, |
641 | exact_match, exact_lock, disk); | |
e63a46be | 642 | register_disk(parent, disk); |
1da177e4 | 643 | blk_register_queue(disk); |
cf0ca9fe | 644 | |
523e1d39 TH |
645 | /* |
646 | * Take an extra ref on queue which will be put on disk_release() | |
647 | * so that it sticks around as long as @disk is there. | |
648 | */ | |
09ac46c4 | 649 | WARN_ON_ONCE(!blk_get_queue(disk->queue)); |
523e1d39 | 650 | |
ed9e1982 TH |
651 | retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj, |
652 | "bdi"); | |
6ffeea77 | 653 | WARN_ON(retval); |
1da177e4 | 654 | |
77ea887e | 655 | disk_add_events(disk); |
25520d55 | 656 | blk_integrity_add(disk); |
1da177e4 | 657 | } |
e63a46be | 658 | EXPORT_SYMBOL(device_add_disk); |
1da177e4 | 659 | |
d2bf1b67 | 660 | void del_gendisk(struct gendisk *disk) |
1da177e4 | 661 | { |
d2bf1b67 TH |
662 | struct disk_part_iter piter; |
663 | struct hd_struct *part; | |
664 | ||
25520d55 | 665 | blk_integrity_del(disk); |
77ea887e TH |
666 | disk_del_events(disk); |
667 | ||
d2bf1b67 TH |
668 | /* invalidate stuff */ |
669 | disk_part_iter_init(&piter, disk, | |
670 | DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE); | |
671 | while ((part = disk_part_iter_next(&piter))) { | |
672 | invalidate_partition(disk, part->partno); | |
4b8c861a | 673 | bdev_unhash_inode(part_devt(part)); |
d2bf1b67 TH |
674 | delete_partition(disk, part->partno); |
675 | } | |
676 | disk_part_iter_exit(&piter); | |
677 | ||
678 | invalidate_partition(disk, 0); | |
d06e05c0 | 679 | bdev_unhash_inode(disk_devt(disk)); |
d2bf1b67 TH |
680 | set_capacity(disk, 0); |
681 | disk->flags &= ~GENHD_FL_UP; | |
682 | ||
ed9e1982 | 683 | sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi"); |
165a5e22 JK |
684 | /* |
685 | * Unregister bdi before releasing device numbers (as they can get | |
686 | * reused and we'd get clashes in sysfs). | |
687 | */ | |
688 | bdi_unregister(disk->queue->backing_dev_info); | |
1da177e4 | 689 | blk_unregister_queue(disk); |
f331c029 | 690 | blk_unregister_region(disk_devt(disk), disk->minors); |
d2bf1b67 TH |
691 | |
692 | part_stat_set_all(&disk->part0, 0); | |
693 | disk->part0.stamp = 0; | |
694 | ||
695 | kobject_put(disk->part0.holder_dir); | |
696 | kobject_put(disk->slave_dir); | |
d2bf1b67 TH |
697 | if (!sysfs_deprecated) |
698 | sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk))); | |
25e823c8 | 699 | pm_runtime_set_memalloc_noio(disk_to_dev(disk), false); |
d2bf1b67 | 700 | device_del(disk_to_dev(disk)); |
1da177e4 | 701 | } |
d2bf1b67 | 702 | EXPORT_SYMBOL(del_gendisk); |
1da177e4 | 703 | |
99e6608c VV |
704 | /* sysfs access to bad-blocks list. */ |
705 | static ssize_t disk_badblocks_show(struct device *dev, | |
706 | struct device_attribute *attr, | |
707 | char *page) | |
708 | { | |
709 | struct gendisk *disk = dev_to_disk(dev); | |
710 | ||
711 | if (!disk->bb) | |
712 | return sprintf(page, "\n"); | |
713 | ||
714 | return badblocks_show(disk->bb, page, 0); | |
715 | } | |
716 | ||
717 | static ssize_t disk_badblocks_store(struct device *dev, | |
718 | struct device_attribute *attr, | |
719 | const char *page, size_t len) | |
720 | { | |
721 | struct gendisk *disk = dev_to_disk(dev); | |
722 | ||
723 | if (!disk->bb) | |
724 | return -ENXIO; | |
725 | ||
726 | return badblocks_store(disk->bb, page, len, 0); | |
727 | } | |
728 | ||
1da177e4 LT |
729 | /** |
730 | * get_gendisk - get partitioning information for a given device | |
710027a4 | 731 | * @devt: device to get partitioning information for |
496aa8a9 | 732 | * @partno: returned partition index |
1da177e4 LT |
733 | * |
734 | * This function gets the structure containing partitioning | |
710027a4 | 735 | * information for the given device @devt. |
1da177e4 | 736 | */ |
cf771cb5 | 737 | struct gendisk *get_gendisk(dev_t devt, int *partno) |
1da177e4 | 738 | { |
bcce3de1 TH |
739 | struct gendisk *disk = NULL; |
740 | ||
741 | if (MAJOR(devt) != BLOCK_EXT_MAJOR) { | |
742 | struct kobject *kobj; | |
743 | ||
744 | kobj = kobj_lookup(bdev_map, devt, partno); | |
745 | if (kobj) | |
746 | disk = dev_to_disk(kobj_to_dev(kobj)); | |
747 | } else { | |
748 | struct hd_struct *part; | |
749 | ||
4d66e5e9 | 750 | spin_lock_bh(&ext_devt_lock); |
870d6656 | 751 | part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); |
bcce3de1 TH |
752 | if (part && get_disk(part_to_disk(part))) { |
753 | *partno = part->partno; | |
754 | disk = part_to_disk(part); | |
755 | } | |
4d66e5e9 | 756 | spin_unlock_bh(&ext_devt_lock); |
bcce3de1 | 757 | } |
edfaa7c3 | 758 | |
bcce3de1 | 759 | return disk; |
1da177e4 | 760 | } |
b6ac23af | 761 | EXPORT_SYMBOL(get_gendisk); |
1da177e4 | 762 | |
f331c029 TH |
763 | /** |
764 | * bdget_disk - do bdget() by gendisk and partition number | |
765 | * @disk: gendisk of interest | |
766 | * @partno: partition number | |
767 | * | |
768 | * Find partition @partno from @disk, do bdget() on it. | |
769 | * | |
770 | * CONTEXT: | |
771 | * Don't care. | |
772 | * | |
773 | * RETURNS: | |
774 | * Resulting block_device on success, NULL on failure. | |
775 | */ | |
aeb3d3a8 | 776 | struct block_device *bdget_disk(struct gendisk *disk, int partno) |
f331c029 | 777 | { |
548b10eb TH |
778 | struct hd_struct *part; |
779 | struct block_device *bdev = NULL; | |
f331c029 | 780 | |
548b10eb | 781 | part = disk_get_part(disk, partno); |
2bbedcb4 | 782 | if (part) |
548b10eb TH |
783 | bdev = bdget(part_devt(part)); |
784 | disk_put_part(part); | |
f331c029 | 785 | |
548b10eb | 786 | return bdev; |
f331c029 TH |
787 | } |
788 | EXPORT_SYMBOL(bdget_disk); | |
789 | ||
5c6f35c5 GKH |
790 | /* |
791 | * print a full list of all partitions - intended for places where the root | |
792 | * filesystem can't be mounted and thus to give the victim some idea of what | |
793 | * went wrong | |
794 | */ | |
795 | void __init printk_all_partitions(void) | |
796 | { | |
def4e38d TH |
797 | struct class_dev_iter iter; |
798 | struct device *dev; | |
799 | ||
800 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); | |
801 | while ((dev = class_dev_iter_next(&iter))) { | |
802 | struct gendisk *disk = dev_to_disk(dev); | |
e71bf0d0 TH |
803 | struct disk_part_iter piter; |
804 | struct hd_struct *part; | |
1f014290 TH |
805 | char name_buf[BDEVNAME_SIZE]; |
806 | char devt_buf[BDEVT_SIZE]; | |
def4e38d TH |
807 | |
808 | /* | |
809 | * Don't show empty devices or things that have been | |
25985edc | 810 | * suppressed |
def4e38d TH |
811 | */ |
812 | if (get_capacity(disk) == 0 || | |
813 | (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)) | |
814 | continue; | |
815 | ||
816 | /* | |
817 | * Note, unlike /proc/partitions, I am showing the | |
818 | * numbers in hex - the same format as the root= | |
819 | * option takes. | |
820 | */ | |
074a7aca TH |
821 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0); |
822 | while ((part = disk_part_iter_next(&piter))) { | |
823 | bool is_part0 = part == &disk->part0; | |
def4e38d | 824 | |
b5af921e | 825 | printk("%s%s %10llu %s %s", is_part0 ? "" : " ", |
1f014290 | 826 | bdevt_str(part_devt(part), devt_buf), |
c83f6bf9 VG |
827 | (unsigned long long)part_nr_sects_read(part) >> 1 |
828 | , disk_name(disk, part->partno, name_buf), | |
1ad7e899 | 829 | part->info ? part->info->uuid : ""); |
074a7aca | 830 | if (is_part0) { |
52c44d93 | 831 | if (dev->parent && dev->parent->driver) |
074a7aca | 832 | printk(" driver: %s\n", |
52c44d93 | 833 | dev->parent->driver->name); |
074a7aca TH |
834 | else |
835 | printk(" (driver?)\n"); | |
836 | } else | |
837 | printk("\n"); | |
838 | } | |
e71bf0d0 | 839 | disk_part_iter_exit(&piter); |
def4e38d TH |
840 | } |
841 | class_dev_iter_exit(&iter); | |
dd2a345f DG |
842 | } |
843 | ||
1da177e4 LT |
844 | #ifdef CONFIG_PROC_FS |
845 | /* iterator */ | |
def4e38d | 846 | static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos) |
68c4d4a7 | 847 | { |
def4e38d TH |
848 | loff_t skip = *pos; |
849 | struct class_dev_iter *iter; | |
850 | struct device *dev; | |
68c4d4a7 | 851 | |
aeb3d3a8 | 852 | iter = kmalloc(sizeof(*iter), GFP_KERNEL); |
def4e38d TH |
853 | if (!iter) |
854 | return ERR_PTR(-ENOMEM); | |
855 | ||
856 | seqf->private = iter; | |
857 | class_dev_iter_init(iter, &block_class, NULL, &disk_type); | |
858 | do { | |
859 | dev = class_dev_iter_next(iter); | |
860 | if (!dev) | |
861 | return NULL; | |
862 | } while (skip--); | |
863 | ||
864 | return dev_to_disk(dev); | |
68c4d4a7 GKH |
865 | } |
866 | ||
def4e38d | 867 | static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos) |
1da177e4 | 868 | { |
edfaa7c3 | 869 | struct device *dev; |
1da177e4 | 870 | |
def4e38d TH |
871 | (*pos)++; |
872 | dev = class_dev_iter_next(seqf->private); | |
2ac3cee5 | 873 | if (dev) |
68c4d4a7 | 874 | return dev_to_disk(dev); |
2ac3cee5 | 875 | |
1da177e4 LT |
876 | return NULL; |
877 | } | |
878 | ||
def4e38d | 879 | static void disk_seqf_stop(struct seq_file *seqf, void *v) |
27f30251 | 880 | { |
def4e38d | 881 | struct class_dev_iter *iter = seqf->private; |
27f30251 | 882 | |
def4e38d TH |
883 | /* stop is called even after start failed :-( */ |
884 | if (iter) { | |
885 | class_dev_iter_exit(iter); | |
886 | kfree(iter); | |
77da1605 | 887 | seqf->private = NULL; |
5c0ef6d0 | 888 | } |
1da177e4 LT |
889 | } |
890 | ||
def4e38d | 891 | static void *show_partition_start(struct seq_file *seqf, loff_t *pos) |
1da177e4 | 892 | { |
06768067 | 893 | void *p; |
def4e38d TH |
894 | |
895 | p = disk_seqf_start(seqf, pos); | |
b9f985b6 | 896 | if (!IS_ERR_OR_NULL(p) && !*pos) |
def4e38d TH |
897 | seq_puts(seqf, "major minor #blocks name\n\n"); |
898 | return p; | |
1da177e4 LT |
899 | } |
900 | ||
cf771cb5 | 901 | static int show_partition(struct seq_file *seqf, void *v) |
1da177e4 LT |
902 | { |
903 | struct gendisk *sgp = v; | |
e71bf0d0 TH |
904 | struct disk_part_iter piter; |
905 | struct hd_struct *part; | |
1da177e4 LT |
906 | char buf[BDEVNAME_SIZE]; |
907 | ||
1da177e4 | 908 | /* Don't show non-partitionable removeable devices or empty devices */ |
d27769ec | 909 | if (!get_capacity(sgp) || (!disk_max_parts(sgp) && |
f331c029 | 910 | (sgp->flags & GENHD_FL_REMOVABLE))) |
1da177e4 LT |
911 | return 0; |
912 | if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO) | |
913 | return 0; | |
914 | ||
915 | /* show the full disk and all non-0 size partitions of it */ | |
074a7aca | 916 | disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0); |
e71bf0d0 | 917 | while ((part = disk_part_iter_next(&piter))) |
1f014290 | 918 | seq_printf(seqf, "%4d %7d %10llu %s\n", |
f331c029 | 919 | MAJOR(part_devt(part)), MINOR(part_devt(part)), |
c83f6bf9 | 920 | (unsigned long long)part_nr_sects_read(part) >> 1, |
f331c029 | 921 | disk_name(sgp, part->partno, buf)); |
e71bf0d0 | 922 | disk_part_iter_exit(&piter); |
1da177e4 LT |
923 | |
924 | return 0; | |
925 | } | |
926 | ||
f500975a | 927 | static const struct seq_operations partitions_op = { |
def4e38d TH |
928 | .start = show_partition_start, |
929 | .next = disk_seqf_next, | |
930 | .stop = disk_seqf_stop, | |
edfaa7c3 | 931 | .show = show_partition |
1da177e4 | 932 | }; |
f500975a AD |
933 | |
934 | static int partitions_open(struct inode *inode, struct file *file) | |
935 | { | |
936 | return seq_open(file, &partitions_op); | |
937 | } | |
938 | ||
939 | static const struct file_operations proc_partitions_operations = { | |
940 | .open = partitions_open, | |
941 | .read = seq_read, | |
942 | .llseek = seq_lseek, | |
943 | .release = seq_release, | |
944 | }; | |
1da177e4 LT |
945 | #endif |
946 | ||
947 | ||
cf771cb5 | 948 | static struct kobject *base_probe(dev_t devt, int *partno, void *data) |
1da177e4 | 949 | { |
edfaa7c3 | 950 | if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0) |
1da177e4 | 951 | /* Make old-style 2.4 aliases work */ |
edfaa7c3 | 952 | request_module("block-major-%d", MAJOR(devt)); |
1da177e4 LT |
953 | return NULL; |
954 | } | |
955 | ||
956 | static int __init genhd_device_init(void) | |
957 | { | |
e105b8bf DW |
958 | int error; |
959 | ||
960 | block_class.dev_kobj = sysfs_dev_block_kobj; | |
961 | error = class_register(&block_class); | |
ee27a558 RM |
962 | if (unlikely(error)) |
963 | return error; | |
edfaa7c3 | 964 | bdev_map = kobj_map_init(base_probe, &block_class_lock); |
1da177e4 | 965 | blk_dev_init(); |
edfaa7c3 | 966 | |
561ec68e ZY |
967 | register_blkdev(BLOCK_EXT_MAJOR, "blkext"); |
968 | ||
edfaa7c3 | 969 | /* create top-level block dir */ |
e52eec13 AK |
970 | if (!sysfs_deprecated) |
971 | block_depr = kobject_create_and_add("block", NULL); | |
830d3cfb | 972 | return 0; |
1da177e4 LT |
973 | } |
974 | ||
975 | subsys_initcall(genhd_device_init); | |
976 | ||
edfaa7c3 KS |
977 | static ssize_t disk_range_show(struct device *dev, |
978 | struct device_attribute *attr, char *buf) | |
1da177e4 | 979 | { |
edfaa7c3 | 980 | struct gendisk *disk = dev_to_disk(dev); |
1da177e4 | 981 | |
edfaa7c3 | 982 | return sprintf(buf, "%d\n", disk->minors); |
1da177e4 LT |
983 | } |
984 | ||
1f014290 TH |
985 | static ssize_t disk_ext_range_show(struct device *dev, |
986 | struct device_attribute *attr, char *buf) | |
987 | { | |
988 | struct gendisk *disk = dev_to_disk(dev); | |
989 | ||
b5d0b9df | 990 | return sprintf(buf, "%d\n", disk_max_parts(disk)); |
1f014290 TH |
991 | } |
992 | ||
edfaa7c3 KS |
993 | static ssize_t disk_removable_show(struct device *dev, |
994 | struct device_attribute *attr, char *buf) | |
a7fd6706 | 995 | { |
edfaa7c3 | 996 | struct gendisk *disk = dev_to_disk(dev); |
a7fd6706 | 997 | |
edfaa7c3 KS |
998 | return sprintf(buf, "%d\n", |
999 | (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0)); | |
a7fd6706 KS |
1000 | } |
1001 | ||
1c9ce527 KS |
1002 | static ssize_t disk_ro_show(struct device *dev, |
1003 | struct device_attribute *attr, char *buf) | |
1004 | { | |
1005 | struct gendisk *disk = dev_to_disk(dev); | |
1006 | ||
b7db9956 | 1007 | return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0); |
1c9ce527 KS |
1008 | } |
1009 | ||
edfaa7c3 KS |
1010 | static ssize_t disk_capability_show(struct device *dev, |
1011 | struct device_attribute *attr, char *buf) | |
86ce18d7 | 1012 | { |
edfaa7c3 KS |
1013 | struct gendisk *disk = dev_to_disk(dev); |
1014 | ||
1015 | return sprintf(buf, "%x\n", disk->flags); | |
86ce18d7 | 1016 | } |
edfaa7c3 | 1017 | |
c72758f3 MP |
1018 | static ssize_t disk_alignment_offset_show(struct device *dev, |
1019 | struct device_attribute *attr, | |
1020 | char *buf) | |
1021 | { | |
1022 | struct gendisk *disk = dev_to_disk(dev); | |
1023 | ||
1024 | return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue)); | |
1025 | } | |
1026 | ||
86b37281 MP |
1027 | static ssize_t disk_discard_alignment_show(struct device *dev, |
1028 | struct device_attribute *attr, | |
1029 | char *buf) | |
1030 | { | |
1031 | struct gendisk *disk = dev_to_disk(dev); | |
1032 | ||
dd3d145d | 1033 | return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue)); |
86b37281 MP |
1034 | } |
1035 | ||
edfaa7c3 | 1036 | static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL); |
1f014290 | 1037 | static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL); |
edfaa7c3 | 1038 | static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL); |
1c9ce527 | 1039 | static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL); |
e5610521 | 1040 | static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL); |
c72758f3 | 1041 | static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL); |
86b37281 MP |
1042 | static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show, |
1043 | NULL); | |
edfaa7c3 | 1044 | static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL); |
074a7aca | 1045 | static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL); |
316d315b | 1046 | static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL); |
99e6608c VV |
1047 | static DEVICE_ATTR(badblocks, S_IRUGO | S_IWUSR, disk_badblocks_show, |
1048 | disk_badblocks_store); | |
c17bb495 | 1049 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
edfaa7c3 | 1050 | static struct device_attribute dev_attr_fail = |
eddb2e26 | 1051 | __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store); |
c17bb495 | 1052 | #endif |
581d4e28 JA |
1053 | #ifdef CONFIG_FAIL_IO_TIMEOUT |
1054 | static struct device_attribute dev_attr_fail_timeout = | |
1055 | __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show, | |
1056 | part_timeout_store); | |
1057 | #endif | |
edfaa7c3 KS |
1058 | |
1059 | static struct attribute *disk_attrs[] = { | |
1060 | &dev_attr_range.attr, | |
1f014290 | 1061 | &dev_attr_ext_range.attr, |
edfaa7c3 | 1062 | &dev_attr_removable.attr, |
1c9ce527 | 1063 | &dev_attr_ro.attr, |
edfaa7c3 | 1064 | &dev_attr_size.attr, |
c72758f3 | 1065 | &dev_attr_alignment_offset.attr, |
86b37281 | 1066 | &dev_attr_discard_alignment.attr, |
edfaa7c3 KS |
1067 | &dev_attr_capability.attr, |
1068 | &dev_attr_stat.attr, | |
316d315b | 1069 | &dev_attr_inflight.attr, |
99e6608c | 1070 | &dev_attr_badblocks.attr, |
edfaa7c3 KS |
1071 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
1072 | &dev_attr_fail.attr, | |
581d4e28 JA |
1073 | #endif |
1074 | #ifdef CONFIG_FAIL_IO_TIMEOUT | |
1075 | &dev_attr_fail_timeout.attr, | |
edfaa7c3 KS |
1076 | #endif |
1077 | NULL | |
1078 | }; | |
1079 | ||
1080 | static struct attribute_group disk_attr_group = { | |
1081 | .attrs = disk_attrs, | |
1082 | }; | |
1083 | ||
a4dbd674 | 1084 | static const struct attribute_group *disk_attr_groups[] = { |
edfaa7c3 KS |
1085 | &disk_attr_group, |
1086 | NULL | |
1da177e4 LT |
1087 | }; |
1088 | ||
540eed56 TH |
1089 | /** |
1090 | * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way | |
1091 | * @disk: disk to replace part_tbl for | |
1092 | * @new_ptbl: new part_tbl to install | |
1093 | * | |
1094 | * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The | |
1095 | * original ptbl is freed using RCU callback. | |
1096 | * | |
1097 | * LOCKING: | |
1098 | * Matching bd_mutx locked. | |
1099 | */ | |
1100 | static void disk_replace_part_tbl(struct gendisk *disk, | |
1101 | struct disk_part_tbl *new_ptbl) | |
1102 | { | |
1103 | struct disk_part_tbl *old_ptbl = disk->part_tbl; | |
1104 | ||
1105 | rcu_assign_pointer(disk->part_tbl, new_ptbl); | |
a6f23657 JA |
1106 | |
1107 | if (old_ptbl) { | |
1108 | rcu_assign_pointer(old_ptbl->last_lookup, NULL); | |
57bdfbf9 | 1109 | kfree_rcu(old_ptbl, rcu_head); |
a6f23657 | 1110 | } |
540eed56 TH |
1111 | } |
1112 | ||
1113 | /** | |
1114 | * disk_expand_part_tbl - expand disk->part_tbl | |
1115 | * @disk: disk to expand part_tbl for | |
1116 | * @partno: expand such that this partno can fit in | |
1117 | * | |
1118 | * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl | |
1119 | * uses RCU to allow unlocked dereferencing for stats and other stuff. | |
1120 | * | |
1121 | * LOCKING: | |
1122 | * Matching bd_mutex locked, might sleep. | |
1123 | * | |
1124 | * RETURNS: | |
1125 | * 0 on success, -errno on failure. | |
1126 | */ | |
1127 | int disk_expand_part_tbl(struct gendisk *disk, int partno) | |
1128 | { | |
1129 | struct disk_part_tbl *old_ptbl = disk->part_tbl; | |
1130 | struct disk_part_tbl *new_ptbl; | |
1131 | int len = old_ptbl ? old_ptbl->len : 0; | |
5fabcb4c | 1132 | int i, target; |
540eed56 | 1133 | size_t size; |
5fabcb4c JA |
1134 | |
1135 | /* | |
1136 | * check for int overflow, since we can get here from blkpg_ioctl() | |
1137 | * with a user passed 'partno'. | |
1138 | */ | |
1139 | target = partno + 1; | |
1140 | if (target < 0) | |
1141 | return -EINVAL; | |
540eed56 TH |
1142 | |
1143 | /* disk_max_parts() is zero during initialization, ignore if so */ | |
1144 | if (disk_max_parts(disk) && target > disk_max_parts(disk)) | |
1145 | return -EINVAL; | |
1146 | ||
1147 | if (target <= len) | |
1148 | return 0; | |
1149 | ||
1150 | size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]); | |
1151 | new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id); | |
1152 | if (!new_ptbl) | |
1153 | return -ENOMEM; | |
1154 | ||
540eed56 TH |
1155 | new_ptbl->len = target; |
1156 | ||
1157 | for (i = 0; i < len; i++) | |
1158 | rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]); | |
1159 | ||
1160 | disk_replace_part_tbl(disk, new_ptbl); | |
1161 | return 0; | |
1162 | } | |
1163 | ||
edfaa7c3 | 1164 | static void disk_release(struct device *dev) |
1da177e4 | 1165 | { |
edfaa7c3 KS |
1166 | struct gendisk *disk = dev_to_disk(dev); |
1167 | ||
2da78092 | 1168 | blk_free_devt(dev->devt); |
77ea887e | 1169 | disk_release_events(disk); |
1da177e4 | 1170 | kfree(disk->random); |
540eed56 | 1171 | disk_replace_part_tbl(disk, NULL); |
b54e5ed8 | 1172 | hd_free_part(&disk->part0); |
523e1d39 TH |
1173 | if (disk->queue) |
1174 | blk_put_queue(disk->queue); | |
1da177e4 LT |
1175 | kfree(disk); |
1176 | } | |
edfaa7c3 KS |
1177 | struct class block_class = { |
1178 | .name = "block", | |
1da177e4 LT |
1179 | }; |
1180 | ||
3c2670e6 | 1181 | static char *block_devnode(struct device *dev, umode_t *mode, |
4e4098a3 | 1182 | kuid_t *uid, kgid_t *gid) |
b03f38b6 KS |
1183 | { |
1184 | struct gendisk *disk = dev_to_disk(dev); | |
1185 | ||
e454cea2 KS |
1186 | if (disk->devnode) |
1187 | return disk->devnode(disk, mode); | |
b03f38b6 KS |
1188 | return NULL; |
1189 | } | |
1190 | ||
1826eadf | 1191 | static struct device_type disk_type = { |
edfaa7c3 KS |
1192 | .name = "disk", |
1193 | .groups = disk_attr_groups, | |
1194 | .release = disk_release, | |
e454cea2 | 1195 | .devnode = block_devnode, |
1da177e4 LT |
1196 | }; |
1197 | ||
a6e2ba88 | 1198 | #ifdef CONFIG_PROC_FS |
cf771cb5 TH |
1199 | /* |
1200 | * aggregate disk stat collector. Uses the same stats that the sysfs | |
1201 | * entries do, above, but makes them available through one seq_file. | |
1202 | * | |
1203 | * The output looks suspiciously like /proc/partitions with a bunch of | |
1204 | * extra fields. | |
1205 | */ | |
1206 | static int diskstats_show(struct seq_file *seqf, void *v) | |
1da177e4 LT |
1207 | { |
1208 | struct gendisk *gp = v; | |
e71bf0d0 TH |
1209 | struct disk_part_iter piter; |
1210 | struct hd_struct *hd; | |
1da177e4 | 1211 | char buf[BDEVNAME_SIZE]; |
c9959059 | 1212 | int cpu; |
1da177e4 LT |
1213 | |
1214 | /* | |
ed9e1982 | 1215 | if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next) |
cf771cb5 | 1216 | seq_puts(seqf, "major minor name" |
1da177e4 LT |
1217 | " rio rmerge rsect ruse wio wmerge " |
1218 | "wsect wuse running use aveq" | |
1219 | "\n\n"); | |
1220 | */ | |
9f5e4865 | 1221 | |
71982a40 | 1222 | disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0); |
e71bf0d0 | 1223 | while ((hd = disk_part_iter_next(&piter))) { |
074a7aca | 1224 | cpu = part_stat_lock(); |
c9959059 | 1225 | part_round_stats(cpu, hd); |
074a7aca | 1226 | part_stat_unlock(); |
f95fe9cf HP |
1227 | seq_printf(seqf, "%4d %7d %s %lu %lu %lu " |
1228 | "%u %lu %lu %lu %u %u %u %u\n", | |
f331c029 TH |
1229 | MAJOR(part_devt(hd)), MINOR(part_devt(hd)), |
1230 | disk_name(gp, hd->partno, buf), | |
53f22956 LY |
1231 | part_stat_read(hd, ios[READ]), |
1232 | part_stat_read(hd, merges[READ]), | |
f95fe9cf | 1233 | part_stat_read(hd, sectors[READ]), |
53f22956 LY |
1234 | jiffies_to_msecs(part_stat_read(hd, ticks[READ])), |
1235 | part_stat_read(hd, ios[WRITE]), | |
1236 | part_stat_read(hd, merges[WRITE]), | |
f95fe9cf | 1237 | part_stat_read(hd, sectors[WRITE]), |
53f22956 | 1238 | jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])), |
316d315b | 1239 | part_in_flight(hd), |
28f39d55 JM |
1240 | jiffies_to_msecs(part_stat_read(hd, io_ticks)), |
1241 | jiffies_to_msecs(part_stat_read(hd, time_in_queue)) | |
1242 | ); | |
1da177e4 | 1243 | } |
e71bf0d0 | 1244 | disk_part_iter_exit(&piter); |
9f5e4865 | 1245 | |
1da177e4 LT |
1246 | return 0; |
1247 | } | |
1248 | ||
31d85ab2 | 1249 | static const struct seq_operations diskstats_op = { |
def4e38d TH |
1250 | .start = disk_seqf_start, |
1251 | .next = disk_seqf_next, | |
1252 | .stop = disk_seqf_stop, | |
1da177e4 LT |
1253 | .show = diskstats_show |
1254 | }; | |
f500975a | 1255 | |
31d85ab2 AD |
1256 | static int diskstats_open(struct inode *inode, struct file *file) |
1257 | { | |
1258 | return seq_open(file, &diskstats_op); | |
1259 | } | |
1260 | ||
1261 | static const struct file_operations proc_diskstats_operations = { | |
1262 | .open = diskstats_open, | |
1263 | .read = seq_read, | |
1264 | .llseek = seq_lseek, | |
1265 | .release = seq_release, | |
1266 | }; | |
1267 | ||
f500975a AD |
1268 | static int __init proc_genhd_init(void) |
1269 | { | |
31d85ab2 | 1270 | proc_create("diskstats", 0, NULL, &proc_diskstats_operations); |
f500975a AD |
1271 | proc_create("partitions", 0, NULL, &proc_partitions_operations); |
1272 | return 0; | |
1273 | } | |
1274 | module_init(proc_genhd_init); | |
a6e2ba88 | 1275 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 1276 | |
cf771cb5 | 1277 | dev_t blk_lookup_devt(const char *name, int partno) |
a142be85 | 1278 | { |
def4e38d TH |
1279 | dev_t devt = MKDEV(0, 0); |
1280 | struct class_dev_iter iter; | |
1281 | struct device *dev; | |
a142be85 | 1282 | |
def4e38d TH |
1283 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); |
1284 | while ((dev = class_dev_iter_next(&iter))) { | |
a142be85 | 1285 | struct gendisk *disk = dev_to_disk(dev); |
548b10eb | 1286 | struct hd_struct *part; |
a142be85 | 1287 | |
3ada8b7e | 1288 | if (strcmp(dev_name(dev), name)) |
f331c029 | 1289 | continue; |
f331c029 | 1290 | |
41b8c853 NB |
1291 | if (partno < disk->minors) { |
1292 | /* We need to return the right devno, even | |
1293 | * if the partition doesn't exist yet. | |
1294 | */ | |
1295 | devt = MKDEV(MAJOR(dev->devt), | |
1296 | MINOR(dev->devt) + partno); | |
1297 | break; | |
1298 | } | |
548b10eb | 1299 | part = disk_get_part(disk, partno); |
2bbedcb4 | 1300 | if (part) { |
f331c029 | 1301 | devt = part_devt(part); |
e71bf0d0 | 1302 | disk_put_part(part); |
548b10eb | 1303 | break; |
def4e38d | 1304 | } |
548b10eb | 1305 | disk_put_part(part); |
5c0ef6d0 | 1306 | } |
def4e38d | 1307 | class_dev_iter_exit(&iter); |
edfaa7c3 KS |
1308 | return devt; |
1309 | } | |
edfaa7c3 KS |
1310 | EXPORT_SYMBOL(blk_lookup_devt); |
1311 | ||
1da177e4 LT |
1312 | struct gendisk *alloc_disk(int minors) |
1313 | { | |
c304a51b | 1314 | return alloc_disk_node(minors, NUMA_NO_NODE); |
1946089a | 1315 | } |
689d6fac | 1316 | EXPORT_SYMBOL(alloc_disk); |
1946089a CL |
1317 | |
1318 | struct gendisk *alloc_disk_node(int minors, int node_id) | |
1319 | { | |
1320 | struct gendisk *disk; | |
1321 | ||
c1b511eb | 1322 | disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id); |
1da177e4 | 1323 | if (disk) { |
074a7aca | 1324 | if (!init_part_stats(&disk->part0)) { |
1da177e4 LT |
1325 | kfree(disk); |
1326 | return NULL; | |
1327 | } | |
bf91db18 | 1328 | disk->node_id = node_id; |
540eed56 TH |
1329 | if (disk_expand_part_tbl(disk, 0)) { |
1330 | free_part_stats(&disk->part0); | |
b5d0b9df TH |
1331 | kfree(disk); |
1332 | return NULL; | |
1da177e4 | 1333 | } |
540eed56 | 1334 | disk->part_tbl->part[0] = &disk->part0; |
6c23a968 | 1335 | |
c83f6bf9 VG |
1336 | /* |
1337 | * set_capacity() and get_capacity() currently don't use | |
1338 | * seqcounter to read/update the part0->nr_sects. Still init | |
1339 | * the counter as we can read the sectors in IO submission | |
1340 | * patch using seqence counters. | |
1341 | * | |
1342 | * TODO: Ideally set_capacity() and get_capacity() should be | |
1343 | * converted to make use of bd_mutex and sequence counters. | |
1344 | */ | |
1345 | seqcount_init(&disk->part0.nr_sects_seq); | |
6c71013e ML |
1346 | if (hd_ref_init(&disk->part0)) { |
1347 | hd_free_part(&disk->part0); | |
1348 | kfree(disk); | |
1349 | return NULL; | |
1350 | } | |
b5d0b9df | 1351 | |
1da177e4 | 1352 | disk->minors = minors; |
1da177e4 | 1353 | rand_initialize_disk(disk); |
ed9e1982 TH |
1354 | disk_to_dev(disk)->class = &block_class; |
1355 | disk_to_dev(disk)->type = &disk_type; | |
1356 | device_initialize(disk_to_dev(disk)); | |
1da177e4 LT |
1357 | } |
1358 | return disk; | |
1359 | } | |
1946089a | 1360 | EXPORT_SYMBOL(alloc_disk_node); |
1da177e4 LT |
1361 | |
1362 | struct kobject *get_disk(struct gendisk *disk) | |
1363 | { | |
1364 | struct module *owner; | |
1365 | struct kobject *kobj; | |
1366 | ||
1367 | if (!disk->fops) | |
1368 | return NULL; | |
1369 | owner = disk->fops->owner; | |
1370 | if (owner && !try_module_get(owner)) | |
1371 | return NULL; | |
ed9e1982 | 1372 | kobj = kobject_get(&disk_to_dev(disk)->kobj); |
1da177e4 LT |
1373 | if (kobj == NULL) { |
1374 | module_put(owner); | |
1375 | return NULL; | |
1376 | } | |
1377 | return kobj; | |
1378 | ||
1379 | } | |
1380 | ||
1381 | EXPORT_SYMBOL(get_disk); | |
1382 | ||
1383 | void put_disk(struct gendisk *disk) | |
1384 | { | |
1385 | if (disk) | |
ed9e1982 | 1386 | kobject_put(&disk_to_dev(disk)->kobj); |
1da177e4 LT |
1387 | } |
1388 | ||
1389 | EXPORT_SYMBOL(put_disk); | |
1390 | ||
e3264a4d HR |
1391 | static void set_disk_ro_uevent(struct gendisk *gd, int ro) |
1392 | { | |
1393 | char event[] = "DISK_RO=1"; | |
1394 | char *envp[] = { event, NULL }; | |
1395 | ||
1396 | if (!ro) | |
1397 | event[8] = '0'; | |
1398 | kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); | |
1399 | } | |
1400 | ||
1da177e4 LT |
1401 | void set_device_ro(struct block_device *bdev, int flag) |
1402 | { | |
b7db9956 | 1403 | bdev->bd_part->policy = flag; |
1da177e4 LT |
1404 | } |
1405 | ||
1406 | EXPORT_SYMBOL(set_device_ro); | |
1407 | ||
1408 | void set_disk_ro(struct gendisk *disk, int flag) | |
1409 | { | |
e71bf0d0 TH |
1410 | struct disk_part_iter piter; |
1411 | struct hd_struct *part; | |
1412 | ||
e3264a4d HR |
1413 | if (disk->part0.policy != flag) { |
1414 | set_disk_ro_uevent(disk, flag); | |
1415 | disk->part0.policy = flag; | |
1416 | } | |
1417 | ||
1418 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY); | |
e71bf0d0 TH |
1419 | while ((part = disk_part_iter_next(&piter))) |
1420 | part->policy = flag; | |
1421 | disk_part_iter_exit(&piter); | |
1da177e4 LT |
1422 | } |
1423 | ||
1424 | EXPORT_SYMBOL(set_disk_ro); | |
1425 | ||
1426 | int bdev_read_only(struct block_device *bdev) | |
1427 | { | |
1428 | if (!bdev) | |
1429 | return 0; | |
b7db9956 | 1430 | return bdev->bd_part->policy; |
1da177e4 LT |
1431 | } |
1432 | ||
1433 | EXPORT_SYMBOL(bdev_read_only); | |
1434 | ||
cf771cb5 | 1435 | int invalidate_partition(struct gendisk *disk, int partno) |
1da177e4 LT |
1436 | { |
1437 | int res = 0; | |
cf771cb5 | 1438 | struct block_device *bdev = bdget_disk(disk, partno); |
1da177e4 | 1439 | if (bdev) { |
2ef41634 | 1440 | fsync_bdev(bdev); |
93b270f7 | 1441 | res = __invalidate_device(bdev, true); |
1da177e4 LT |
1442 | bdput(bdev); |
1443 | } | |
1444 | return res; | |
1445 | } | |
1446 | ||
1447 | EXPORT_SYMBOL(invalidate_partition); | |
77ea887e TH |
1448 | |
1449 | /* | |
1450 | * Disk events - monitor disk events like media change and eject request. | |
1451 | */ | |
1452 | struct disk_events { | |
1453 | struct list_head node; /* all disk_event's */ | |
1454 | struct gendisk *disk; /* the associated disk */ | |
1455 | spinlock_t lock; | |
1456 | ||
fdd514e1 | 1457 | struct mutex block_mutex; /* protects blocking */ |
77ea887e TH |
1458 | int block; /* event blocking depth */ |
1459 | unsigned int pending; /* events already sent out */ | |
1460 | unsigned int clearing; /* events being cleared */ | |
1461 | ||
1462 | long poll_msecs; /* interval, -1 for default */ | |
1463 | struct delayed_work dwork; | |
1464 | }; | |
1465 | ||
1466 | static const char *disk_events_strs[] = { | |
1467 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change", | |
1468 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request", | |
1469 | }; | |
1470 | ||
1471 | static char *disk_uevents[] = { | |
1472 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1", | |
1473 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1", | |
1474 | }; | |
1475 | ||
1476 | /* list of all disk_events */ | |
1477 | static DEFINE_MUTEX(disk_events_mutex); | |
1478 | static LIST_HEAD(disk_events); | |
1479 | ||
1480 | /* disable in-kernel polling by default */ | |
1fe8f348 | 1481 | static unsigned long disk_events_dfl_poll_msecs; |
77ea887e TH |
1482 | |
1483 | static unsigned long disk_events_poll_jiffies(struct gendisk *disk) | |
1484 | { | |
1485 | struct disk_events *ev = disk->ev; | |
1486 | long intv_msecs = 0; | |
1487 | ||
1488 | /* | |
1489 | * If device-specific poll interval is set, always use it. If | |
1490 | * the default is being used, poll iff there are events which | |
1491 | * can't be monitored asynchronously. | |
1492 | */ | |
1493 | if (ev->poll_msecs >= 0) | |
1494 | intv_msecs = ev->poll_msecs; | |
1495 | else if (disk->events & ~disk->async_events) | |
1496 | intv_msecs = disk_events_dfl_poll_msecs; | |
1497 | ||
1498 | return msecs_to_jiffies(intv_msecs); | |
1499 | } | |
1500 | ||
c3af54af TH |
1501 | /** |
1502 | * disk_block_events - block and flush disk event checking | |
1503 | * @disk: disk to block events for | |
1504 | * | |
1505 | * On return from this function, it is guaranteed that event checking | |
1506 | * isn't in progress and won't happen until unblocked by | |
1507 | * disk_unblock_events(). Events blocking is counted and the actual | |
1508 | * unblocking happens after the matching number of unblocks are done. | |
1509 | * | |
1510 | * Note that this intentionally does not block event checking from | |
1511 | * disk_clear_events(). | |
1512 | * | |
1513 | * CONTEXT: | |
1514 | * Might sleep. | |
1515 | */ | |
1516 | void disk_block_events(struct gendisk *disk) | |
77ea887e TH |
1517 | { |
1518 | struct disk_events *ev = disk->ev; | |
1519 | unsigned long flags; | |
1520 | bool cancel; | |
1521 | ||
c3af54af TH |
1522 | if (!ev) |
1523 | return; | |
1524 | ||
fdd514e1 TH |
1525 | /* |
1526 | * Outer mutex ensures that the first blocker completes canceling | |
1527 | * the event work before further blockers are allowed to finish. | |
1528 | */ | |
1529 | mutex_lock(&ev->block_mutex); | |
1530 | ||
77ea887e TH |
1531 | spin_lock_irqsave(&ev->lock, flags); |
1532 | cancel = !ev->block++; | |
1533 | spin_unlock_irqrestore(&ev->lock, flags); | |
1534 | ||
c3af54af TH |
1535 | if (cancel) |
1536 | cancel_delayed_work_sync(&disk->ev->dwork); | |
fdd514e1 TH |
1537 | |
1538 | mutex_unlock(&ev->block_mutex); | |
77ea887e TH |
1539 | } |
1540 | ||
1541 | static void __disk_unblock_events(struct gendisk *disk, bool check_now) | |
1542 | { | |
1543 | struct disk_events *ev = disk->ev; | |
1544 | unsigned long intv; | |
1545 | unsigned long flags; | |
1546 | ||
1547 | spin_lock_irqsave(&ev->lock, flags); | |
1548 | ||
1549 | if (WARN_ON_ONCE(ev->block <= 0)) | |
1550 | goto out_unlock; | |
1551 | ||
1552 | if (--ev->block) | |
1553 | goto out_unlock; | |
1554 | ||
77ea887e | 1555 | intv = disk_events_poll_jiffies(disk); |
77ea887e | 1556 | if (check_now) |
695588f9 VK |
1557 | queue_delayed_work(system_freezable_power_efficient_wq, |
1558 | &ev->dwork, 0); | |
77ea887e | 1559 | else if (intv) |
695588f9 VK |
1560 | queue_delayed_work(system_freezable_power_efficient_wq, |
1561 | &ev->dwork, intv); | |
77ea887e TH |
1562 | out_unlock: |
1563 | spin_unlock_irqrestore(&ev->lock, flags); | |
1564 | } | |
1565 | ||
77ea887e TH |
1566 | /** |
1567 | * disk_unblock_events - unblock disk event checking | |
1568 | * @disk: disk to unblock events for | |
1569 | * | |
1570 | * Undo disk_block_events(). When the block count reaches zero, it | |
1571 | * starts events polling if configured. | |
1572 | * | |
1573 | * CONTEXT: | |
1574 | * Don't care. Safe to call from irq context. | |
1575 | */ | |
1576 | void disk_unblock_events(struct gendisk *disk) | |
1577 | { | |
1578 | if (disk->ev) | |
facc31dd | 1579 | __disk_unblock_events(disk, false); |
77ea887e TH |
1580 | } |
1581 | ||
1582 | /** | |
85ef06d1 TH |
1583 | * disk_flush_events - schedule immediate event checking and flushing |
1584 | * @disk: disk to check and flush events for | |
1585 | * @mask: events to flush | |
77ea887e | 1586 | * |
85ef06d1 TH |
1587 | * Schedule immediate event checking on @disk if not blocked. Events in |
1588 | * @mask are scheduled to be cleared from the driver. Note that this | |
1589 | * doesn't clear the events from @disk->ev. | |
77ea887e TH |
1590 | * |
1591 | * CONTEXT: | |
85ef06d1 | 1592 | * If @mask is non-zero must be called with bdev->bd_mutex held. |
77ea887e | 1593 | */ |
85ef06d1 | 1594 | void disk_flush_events(struct gendisk *disk, unsigned int mask) |
77ea887e | 1595 | { |
a9dce2a3 | 1596 | struct disk_events *ev = disk->ev; |
a9dce2a3 TH |
1597 | |
1598 | if (!ev) | |
1599 | return; | |
1600 | ||
85ef06d1 TH |
1601 | spin_lock_irq(&ev->lock); |
1602 | ev->clearing |= mask; | |
41f63c53 | 1603 | if (!ev->block) |
695588f9 VK |
1604 | mod_delayed_work(system_freezable_power_efficient_wq, |
1605 | &ev->dwork, 0); | |
85ef06d1 | 1606 | spin_unlock_irq(&ev->lock); |
77ea887e | 1607 | } |
77ea887e TH |
1608 | |
1609 | /** | |
1610 | * disk_clear_events - synchronously check, clear and return pending events | |
1611 | * @disk: disk to fetch and clear events from | |
da3dae54 | 1612 | * @mask: mask of events to be fetched and cleared |
77ea887e TH |
1613 | * |
1614 | * Disk events are synchronously checked and pending events in @mask | |
1615 | * are cleared and returned. This ignores the block count. | |
1616 | * | |
1617 | * CONTEXT: | |
1618 | * Might sleep. | |
1619 | */ | |
1620 | unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask) | |
1621 | { | |
1622 | const struct block_device_operations *bdops = disk->fops; | |
1623 | struct disk_events *ev = disk->ev; | |
1624 | unsigned int pending; | |
12c2bdb2 | 1625 | unsigned int clearing = mask; |
77ea887e TH |
1626 | |
1627 | if (!ev) { | |
1628 | /* for drivers still using the old ->media_changed method */ | |
1629 | if ((mask & DISK_EVENT_MEDIA_CHANGE) && | |
1630 | bdops->media_changed && bdops->media_changed(disk)) | |
1631 | return DISK_EVENT_MEDIA_CHANGE; | |
1632 | return 0; | |
1633 | } | |
1634 | ||
12c2bdb2 DB |
1635 | disk_block_events(disk); |
1636 | ||
1637 | /* | |
1638 | * store the union of mask and ev->clearing on the stack so that the | |
1639 | * race with disk_flush_events does not cause ambiguity (ev->clearing | |
1640 | * can still be modified even if events are blocked). | |
1641 | */ | |
77ea887e | 1642 | spin_lock_irq(&ev->lock); |
12c2bdb2 DB |
1643 | clearing |= ev->clearing; |
1644 | ev->clearing = 0; | |
77ea887e TH |
1645 | spin_unlock_irq(&ev->lock); |
1646 | ||
12c2bdb2 | 1647 | disk_check_events(ev, &clearing); |
aea24a8b | 1648 | /* |
12c2bdb2 DB |
1649 | * if ev->clearing is not 0, the disk_flush_events got called in the |
1650 | * middle of this function, so we want to run the workfn without delay. | |
aea24a8b | 1651 | */ |
12c2bdb2 | 1652 | __disk_unblock_events(disk, ev->clearing ? true : false); |
77ea887e TH |
1653 | |
1654 | /* then, fetch and clear pending events */ | |
1655 | spin_lock_irq(&ev->lock); | |
77ea887e TH |
1656 | pending = ev->pending & mask; |
1657 | ev->pending &= ~mask; | |
1658 | spin_unlock_irq(&ev->lock); | |
12c2bdb2 | 1659 | WARN_ON_ONCE(clearing & mask); |
77ea887e TH |
1660 | |
1661 | return pending; | |
1662 | } | |
1663 | ||
12c2bdb2 DB |
1664 | /* |
1665 | * Separate this part out so that a different pointer for clearing_ptr can be | |
1666 | * passed in for disk_clear_events. | |
1667 | */ | |
77ea887e TH |
1668 | static void disk_events_workfn(struct work_struct *work) |
1669 | { | |
1670 | struct delayed_work *dwork = to_delayed_work(work); | |
1671 | struct disk_events *ev = container_of(dwork, struct disk_events, dwork); | |
12c2bdb2 DB |
1672 | |
1673 | disk_check_events(ev, &ev->clearing); | |
1674 | } | |
1675 | ||
1676 | static void disk_check_events(struct disk_events *ev, | |
1677 | unsigned int *clearing_ptr) | |
1678 | { | |
77ea887e TH |
1679 | struct gendisk *disk = ev->disk; |
1680 | char *envp[ARRAY_SIZE(disk_uevents) + 1] = { }; | |
12c2bdb2 | 1681 | unsigned int clearing = *clearing_ptr; |
77ea887e TH |
1682 | unsigned int events; |
1683 | unsigned long intv; | |
1684 | int nr_events = 0, i; | |
1685 | ||
1686 | /* check events */ | |
1687 | events = disk->fops->check_events(disk, clearing); | |
1688 | ||
1689 | /* accumulate pending events and schedule next poll if necessary */ | |
1690 | spin_lock_irq(&ev->lock); | |
1691 | ||
1692 | events &= ~ev->pending; | |
1693 | ev->pending |= events; | |
12c2bdb2 | 1694 | *clearing_ptr &= ~clearing; |
77ea887e TH |
1695 | |
1696 | intv = disk_events_poll_jiffies(disk); | |
1697 | if (!ev->block && intv) | |
695588f9 VK |
1698 | queue_delayed_work(system_freezable_power_efficient_wq, |
1699 | &ev->dwork, intv); | |
77ea887e TH |
1700 | |
1701 | spin_unlock_irq(&ev->lock); | |
1702 | ||
7c88a168 TH |
1703 | /* |
1704 | * Tell userland about new events. Only the events listed in | |
1705 | * @disk->events are reported. Unlisted events are processed the | |
1706 | * same internally but never get reported to userland. | |
1707 | */ | |
77ea887e | 1708 | for (i = 0; i < ARRAY_SIZE(disk_uevents); i++) |
7c88a168 | 1709 | if (events & disk->events & (1 << i)) |
77ea887e TH |
1710 | envp[nr_events++] = disk_uevents[i]; |
1711 | ||
1712 | if (nr_events) | |
1713 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
1714 | } | |
1715 | ||
1716 | /* | |
1717 | * A disk events enabled device has the following sysfs nodes under | |
1718 | * its /sys/block/X/ directory. | |
1719 | * | |
1720 | * events : list of all supported events | |
1721 | * events_async : list of events which can be detected w/o polling | |
1722 | * events_poll_msecs : polling interval, 0: disable, -1: system default | |
1723 | */ | |
1724 | static ssize_t __disk_events_show(unsigned int events, char *buf) | |
1725 | { | |
1726 | const char *delim = ""; | |
1727 | ssize_t pos = 0; | |
1728 | int i; | |
1729 | ||
1730 | for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++) | |
1731 | if (events & (1 << i)) { | |
1732 | pos += sprintf(buf + pos, "%s%s", | |
1733 | delim, disk_events_strs[i]); | |
1734 | delim = " "; | |
1735 | } | |
1736 | if (pos) | |
1737 | pos += sprintf(buf + pos, "\n"); | |
1738 | return pos; | |
1739 | } | |
1740 | ||
1741 | static ssize_t disk_events_show(struct device *dev, | |
1742 | struct device_attribute *attr, char *buf) | |
1743 | { | |
1744 | struct gendisk *disk = dev_to_disk(dev); | |
1745 | ||
1746 | return __disk_events_show(disk->events, buf); | |
1747 | } | |
1748 | ||
1749 | static ssize_t disk_events_async_show(struct device *dev, | |
1750 | struct device_attribute *attr, char *buf) | |
1751 | { | |
1752 | struct gendisk *disk = dev_to_disk(dev); | |
1753 | ||
1754 | return __disk_events_show(disk->async_events, buf); | |
1755 | } | |
1756 | ||
1757 | static ssize_t disk_events_poll_msecs_show(struct device *dev, | |
1758 | struct device_attribute *attr, | |
1759 | char *buf) | |
1760 | { | |
1761 | struct gendisk *disk = dev_to_disk(dev); | |
1762 | ||
1763 | return sprintf(buf, "%ld\n", disk->ev->poll_msecs); | |
1764 | } | |
1765 | ||
1766 | static ssize_t disk_events_poll_msecs_store(struct device *dev, | |
1767 | struct device_attribute *attr, | |
1768 | const char *buf, size_t count) | |
1769 | { | |
1770 | struct gendisk *disk = dev_to_disk(dev); | |
1771 | long intv; | |
1772 | ||
1773 | if (!count || !sscanf(buf, "%ld", &intv)) | |
1774 | return -EINVAL; | |
1775 | ||
1776 | if (intv < 0 && intv != -1) | |
1777 | return -EINVAL; | |
1778 | ||
c3af54af | 1779 | disk_block_events(disk); |
77ea887e TH |
1780 | disk->ev->poll_msecs = intv; |
1781 | __disk_unblock_events(disk, true); | |
1782 | ||
1783 | return count; | |
1784 | } | |
1785 | ||
1786 | static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL); | |
1787 | static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL); | |
1788 | static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR, | |
1789 | disk_events_poll_msecs_show, | |
1790 | disk_events_poll_msecs_store); | |
1791 | ||
1792 | static const struct attribute *disk_events_attrs[] = { | |
1793 | &dev_attr_events.attr, | |
1794 | &dev_attr_events_async.attr, | |
1795 | &dev_attr_events_poll_msecs.attr, | |
1796 | NULL, | |
1797 | }; | |
1798 | ||
1799 | /* | |
1800 | * The default polling interval can be specified by the kernel | |
1801 | * parameter block.events_dfl_poll_msecs which defaults to 0 | |
1802 | * (disable). This can also be modified runtime by writing to | |
1803 | * /sys/module/block/events_dfl_poll_msecs. | |
1804 | */ | |
1805 | static int disk_events_set_dfl_poll_msecs(const char *val, | |
1806 | const struct kernel_param *kp) | |
1807 | { | |
1808 | struct disk_events *ev; | |
1809 | int ret; | |
1810 | ||
1811 | ret = param_set_ulong(val, kp); | |
1812 | if (ret < 0) | |
1813 | return ret; | |
1814 | ||
1815 | mutex_lock(&disk_events_mutex); | |
1816 | ||
1817 | list_for_each_entry(ev, &disk_events, node) | |
85ef06d1 | 1818 | disk_flush_events(ev->disk, 0); |
77ea887e TH |
1819 | |
1820 | mutex_unlock(&disk_events_mutex); | |
1821 | ||
1822 | return 0; | |
1823 | } | |
1824 | ||
1825 | static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = { | |
1826 | .set = disk_events_set_dfl_poll_msecs, | |
1827 | .get = param_get_ulong, | |
1828 | }; | |
1829 | ||
1830 | #undef MODULE_PARAM_PREFIX | |
1831 | #define MODULE_PARAM_PREFIX "block." | |
1832 | ||
1833 | module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops, | |
1834 | &disk_events_dfl_poll_msecs, 0644); | |
1835 | ||
1836 | /* | |
9f53d2fe | 1837 | * disk_{alloc|add|del|release}_events - initialize and destroy disk_events. |
77ea887e | 1838 | */ |
9f53d2fe | 1839 | static void disk_alloc_events(struct gendisk *disk) |
77ea887e TH |
1840 | { |
1841 | struct disk_events *ev; | |
1842 | ||
75e3f3ee | 1843 | if (!disk->fops->check_events) |
77ea887e TH |
1844 | return; |
1845 | ||
1846 | ev = kzalloc(sizeof(*ev), GFP_KERNEL); | |
1847 | if (!ev) { | |
1848 | pr_warn("%s: failed to initialize events\n", disk->disk_name); | |
1849 | return; | |
1850 | } | |
1851 | ||
77ea887e TH |
1852 | INIT_LIST_HEAD(&ev->node); |
1853 | ev->disk = disk; | |
1854 | spin_lock_init(&ev->lock); | |
fdd514e1 | 1855 | mutex_init(&ev->block_mutex); |
77ea887e TH |
1856 | ev->block = 1; |
1857 | ev->poll_msecs = -1; | |
1858 | INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn); | |
1859 | ||
9f53d2fe SG |
1860 | disk->ev = ev; |
1861 | } | |
1862 | ||
1863 | static void disk_add_events(struct gendisk *disk) | |
1864 | { | |
1865 | if (!disk->ev) | |
1866 | return; | |
1867 | ||
1868 | /* FIXME: error handling */ | |
1869 | if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0) | |
1870 | pr_warn("%s: failed to create sysfs files for events\n", | |
1871 | disk->disk_name); | |
1872 | ||
77ea887e | 1873 | mutex_lock(&disk_events_mutex); |
9f53d2fe | 1874 | list_add_tail(&disk->ev->node, &disk_events); |
77ea887e TH |
1875 | mutex_unlock(&disk_events_mutex); |
1876 | ||
1877 | /* | |
1878 | * Block count is initialized to 1 and the following initial | |
1879 | * unblock kicks it into action. | |
1880 | */ | |
1881 | __disk_unblock_events(disk, true); | |
1882 | } | |
1883 | ||
1884 | static void disk_del_events(struct gendisk *disk) | |
1885 | { | |
1886 | if (!disk->ev) | |
1887 | return; | |
1888 | ||
c3af54af | 1889 | disk_block_events(disk); |
77ea887e TH |
1890 | |
1891 | mutex_lock(&disk_events_mutex); | |
1892 | list_del_init(&disk->ev->node); | |
1893 | mutex_unlock(&disk_events_mutex); | |
1894 | ||
1895 | sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs); | |
1896 | } | |
1897 | ||
1898 | static void disk_release_events(struct gendisk *disk) | |
1899 | { | |
1900 | /* the block count should be 1 from disk_del_events() */ | |
1901 | WARN_ON_ONCE(disk->ev && disk->ev->block != 1); | |
1902 | kfree(disk->ev); | |
1903 | } |