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); | |
d2bf1b67 TH |
679 | set_capacity(disk, 0); |
680 | disk->flags &= ~GENHD_FL_UP; | |
681 | ||
ed9e1982 | 682 | sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi"); |
1da177e4 | 683 | blk_unregister_queue(disk); |
f331c029 | 684 | blk_unregister_region(disk_devt(disk), disk->minors); |
d2bf1b67 TH |
685 | |
686 | part_stat_set_all(&disk->part0, 0); | |
687 | disk->part0.stamp = 0; | |
688 | ||
689 | kobject_put(disk->part0.holder_dir); | |
690 | kobject_put(disk->slave_dir); | |
d2bf1b67 TH |
691 | if (!sysfs_deprecated) |
692 | sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk))); | |
25e823c8 | 693 | pm_runtime_set_memalloc_noio(disk_to_dev(disk), false); |
d2bf1b67 | 694 | device_del(disk_to_dev(disk)); |
1da177e4 | 695 | } |
d2bf1b67 | 696 | EXPORT_SYMBOL(del_gendisk); |
1da177e4 | 697 | |
99e6608c VV |
698 | /* sysfs access to bad-blocks list. */ |
699 | static ssize_t disk_badblocks_show(struct device *dev, | |
700 | struct device_attribute *attr, | |
701 | char *page) | |
702 | { | |
703 | struct gendisk *disk = dev_to_disk(dev); | |
704 | ||
705 | if (!disk->bb) | |
706 | return sprintf(page, "\n"); | |
707 | ||
708 | return badblocks_show(disk->bb, page, 0); | |
709 | } | |
710 | ||
711 | static ssize_t disk_badblocks_store(struct device *dev, | |
712 | struct device_attribute *attr, | |
713 | const char *page, size_t len) | |
714 | { | |
715 | struct gendisk *disk = dev_to_disk(dev); | |
716 | ||
717 | if (!disk->bb) | |
718 | return -ENXIO; | |
719 | ||
720 | return badblocks_store(disk->bb, page, len, 0); | |
721 | } | |
722 | ||
1da177e4 LT |
723 | /** |
724 | * get_gendisk - get partitioning information for a given device | |
710027a4 | 725 | * @devt: device to get partitioning information for |
496aa8a9 | 726 | * @partno: returned partition index |
1da177e4 LT |
727 | * |
728 | * This function gets the structure containing partitioning | |
710027a4 | 729 | * information for the given device @devt. |
1da177e4 | 730 | */ |
cf771cb5 | 731 | struct gendisk *get_gendisk(dev_t devt, int *partno) |
1da177e4 | 732 | { |
bcce3de1 TH |
733 | struct gendisk *disk = NULL; |
734 | ||
735 | if (MAJOR(devt) != BLOCK_EXT_MAJOR) { | |
736 | struct kobject *kobj; | |
737 | ||
738 | kobj = kobj_lookup(bdev_map, devt, partno); | |
739 | if (kobj) | |
740 | disk = dev_to_disk(kobj_to_dev(kobj)); | |
741 | } else { | |
742 | struct hd_struct *part; | |
743 | ||
4d66e5e9 | 744 | spin_lock_bh(&ext_devt_lock); |
870d6656 | 745 | part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); |
bcce3de1 TH |
746 | if (part && get_disk(part_to_disk(part))) { |
747 | *partno = part->partno; | |
748 | disk = part_to_disk(part); | |
749 | } | |
4d66e5e9 | 750 | spin_unlock_bh(&ext_devt_lock); |
bcce3de1 | 751 | } |
edfaa7c3 | 752 | |
bcce3de1 | 753 | return disk; |
1da177e4 | 754 | } |
b6ac23af | 755 | EXPORT_SYMBOL(get_gendisk); |
1da177e4 | 756 | |
f331c029 TH |
757 | /** |
758 | * bdget_disk - do bdget() by gendisk and partition number | |
759 | * @disk: gendisk of interest | |
760 | * @partno: partition number | |
761 | * | |
762 | * Find partition @partno from @disk, do bdget() on it. | |
763 | * | |
764 | * CONTEXT: | |
765 | * Don't care. | |
766 | * | |
767 | * RETURNS: | |
768 | * Resulting block_device on success, NULL on failure. | |
769 | */ | |
aeb3d3a8 | 770 | struct block_device *bdget_disk(struct gendisk *disk, int partno) |
f331c029 | 771 | { |
548b10eb TH |
772 | struct hd_struct *part; |
773 | struct block_device *bdev = NULL; | |
f331c029 | 774 | |
548b10eb | 775 | part = disk_get_part(disk, partno); |
2bbedcb4 | 776 | if (part) |
548b10eb TH |
777 | bdev = bdget(part_devt(part)); |
778 | disk_put_part(part); | |
f331c029 | 779 | |
548b10eb | 780 | return bdev; |
f331c029 TH |
781 | } |
782 | EXPORT_SYMBOL(bdget_disk); | |
783 | ||
5c6f35c5 GKH |
784 | /* |
785 | * print a full list of all partitions - intended for places where the root | |
786 | * filesystem can't be mounted and thus to give the victim some idea of what | |
787 | * went wrong | |
788 | */ | |
789 | void __init printk_all_partitions(void) | |
790 | { | |
def4e38d TH |
791 | struct class_dev_iter iter; |
792 | struct device *dev; | |
793 | ||
794 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); | |
795 | while ((dev = class_dev_iter_next(&iter))) { | |
796 | struct gendisk *disk = dev_to_disk(dev); | |
e71bf0d0 TH |
797 | struct disk_part_iter piter; |
798 | struct hd_struct *part; | |
1f014290 TH |
799 | char name_buf[BDEVNAME_SIZE]; |
800 | char devt_buf[BDEVT_SIZE]; | |
def4e38d TH |
801 | |
802 | /* | |
803 | * Don't show empty devices or things that have been | |
25985edc | 804 | * suppressed |
def4e38d TH |
805 | */ |
806 | if (get_capacity(disk) == 0 || | |
807 | (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)) | |
808 | continue; | |
809 | ||
810 | /* | |
811 | * Note, unlike /proc/partitions, I am showing the | |
812 | * numbers in hex - the same format as the root= | |
813 | * option takes. | |
814 | */ | |
074a7aca TH |
815 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0); |
816 | while ((part = disk_part_iter_next(&piter))) { | |
817 | bool is_part0 = part == &disk->part0; | |
def4e38d | 818 | |
b5af921e | 819 | printk("%s%s %10llu %s %s", is_part0 ? "" : " ", |
1f014290 | 820 | bdevt_str(part_devt(part), devt_buf), |
c83f6bf9 VG |
821 | (unsigned long long)part_nr_sects_read(part) >> 1 |
822 | , disk_name(disk, part->partno, name_buf), | |
1ad7e899 | 823 | part->info ? part->info->uuid : ""); |
074a7aca | 824 | if (is_part0) { |
52c44d93 | 825 | if (dev->parent && dev->parent->driver) |
074a7aca | 826 | printk(" driver: %s\n", |
52c44d93 | 827 | dev->parent->driver->name); |
074a7aca TH |
828 | else |
829 | printk(" (driver?)\n"); | |
830 | } else | |
831 | printk("\n"); | |
832 | } | |
e71bf0d0 | 833 | disk_part_iter_exit(&piter); |
def4e38d TH |
834 | } |
835 | class_dev_iter_exit(&iter); | |
dd2a345f DG |
836 | } |
837 | ||
1da177e4 LT |
838 | #ifdef CONFIG_PROC_FS |
839 | /* iterator */ | |
def4e38d | 840 | static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos) |
68c4d4a7 | 841 | { |
def4e38d TH |
842 | loff_t skip = *pos; |
843 | struct class_dev_iter *iter; | |
844 | struct device *dev; | |
68c4d4a7 | 845 | |
aeb3d3a8 | 846 | iter = kmalloc(sizeof(*iter), GFP_KERNEL); |
def4e38d TH |
847 | if (!iter) |
848 | return ERR_PTR(-ENOMEM); | |
849 | ||
850 | seqf->private = iter; | |
851 | class_dev_iter_init(iter, &block_class, NULL, &disk_type); | |
852 | do { | |
853 | dev = class_dev_iter_next(iter); | |
854 | if (!dev) | |
855 | return NULL; | |
856 | } while (skip--); | |
857 | ||
858 | return dev_to_disk(dev); | |
68c4d4a7 GKH |
859 | } |
860 | ||
def4e38d | 861 | static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos) |
1da177e4 | 862 | { |
edfaa7c3 | 863 | struct device *dev; |
1da177e4 | 864 | |
def4e38d TH |
865 | (*pos)++; |
866 | dev = class_dev_iter_next(seqf->private); | |
2ac3cee5 | 867 | if (dev) |
68c4d4a7 | 868 | return dev_to_disk(dev); |
2ac3cee5 | 869 | |
1da177e4 LT |
870 | return NULL; |
871 | } | |
872 | ||
def4e38d | 873 | static void disk_seqf_stop(struct seq_file *seqf, void *v) |
27f30251 | 874 | { |
def4e38d | 875 | struct class_dev_iter *iter = seqf->private; |
27f30251 | 876 | |
def4e38d TH |
877 | /* stop is called even after start failed :-( */ |
878 | if (iter) { | |
879 | class_dev_iter_exit(iter); | |
880 | kfree(iter); | |
77da1605 | 881 | seqf->private = NULL; |
5c0ef6d0 | 882 | } |
1da177e4 LT |
883 | } |
884 | ||
def4e38d | 885 | static void *show_partition_start(struct seq_file *seqf, loff_t *pos) |
1da177e4 | 886 | { |
06768067 | 887 | void *p; |
def4e38d TH |
888 | |
889 | p = disk_seqf_start(seqf, pos); | |
b9f985b6 | 890 | if (!IS_ERR_OR_NULL(p) && !*pos) |
def4e38d TH |
891 | seq_puts(seqf, "major minor #blocks name\n\n"); |
892 | return p; | |
1da177e4 LT |
893 | } |
894 | ||
cf771cb5 | 895 | static int show_partition(struct seq_file *seqf, void *v) |
1da177e4 LT |
896 | { |
897 | struct gendisk *sgp = v; | |
e71bf0d0 TH |
898 | struct disk_part_iter piter; |
899 | struct hd_struct *part; | |
1da177e4 LT |
900 | char buf[BDEVNAME_SIZE]; |
901 | ||
1da177e4 | 902 | /* Don't show non-partitionable removeable devices or empty devices */ |
d27769ec | 903 | if (!get_capacity(sgp) || (!disk_max_parts(sgp) && |
f331c029 | 904 | (sgp->flags & GENHD_FL_REMOVABLE))) |
1da177e4 LT |
905 | return 0; |
906 | if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO) | |
907 | return 0; | |
908 | ||
909 | /* show the full disk and all non-0 size partitions of it */ | |
074a7aca | 910 | disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0); |
e71bf0d0 | 911 | while ((part = disk_part_iter_next(&piter))) |
1f014290 | 912 | seq_printf(seqf, "%4d %7d %10llu %s\n", |
f331c029 | 913 | MAJOR(part_devt(part)), MINOR(part_devt(part)), |
c83f6bf9 | 914 | (unsigned long long)part_nr_sects_read(part) >> 1, |
f331c029 | 915 | disk_name(sgp, part->partno, buf)); |
e71bf0d0 | 916 | disk_part_iter_exit(&piter); |
1da177e4 LT |
917 | |
918 | return 0; | |
919 | } | |
920 | ||
f500975a | 921 | static const struct seq_operations partitions_op = { |
def4e38d TH |
922 | .start = show_partition_start, |
923 | .next = disk_seqf_next, | |
924 | .stop = disk_seqf_stop, | |
edfaa7c3 | 925 | .show = show_partition |
1da177e4 | 926 | }; |
f500975a AD |
927 | |
928 | static int partitions_open(struct inode *inode, struct file *file) | |
929 | { | |
930 | return seq_open(file, &partitions_op); | |
931 | } | |
932 | ||
933 | static const struct file_operations proc_partitions_operations = { | |
934 | .open = partitions_open, | |
935 | .read = seq_read, | |
936 | .llseek = seq_lseek, | |
937 | .release = seq_release, | |
938 | }; | |
1da177e4 LT |
939 | #endif |
940 | ||
941 | ||
cf771cb5 | 942 | static struct kobject *base_probe(dev_t devt, int *partno, void *data) |
1da177e4 | 943 | { |
edfaa7c3 | 944 | if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0) |
1da177e4 | 945 | /* Make old-style 2.4 aliases work */ |
edfaa7c3 | 946 | request_module("block-major-%d", MAJOR(devt)); |
1da177e4 LT |
947 | return NULL; |
948 | } | |
949 | ||
950 | static int __init genhd_device_init(void) | |
951 | { | |
e105b8bf DW |
952 | int error; |
953 | ||
954 | block_class.dev_kobj = sysfs_dev_block_kobj; | |
955 | error = class_register(&block_class); | |
ee27a558 RM |
956 | if (unlikely(error)) |
957 | return error; | |
edfaa7c3 | 958 | bdev_map = kobj_map_init(base_probe, &block_class_lock); |
1da177e4 | 959 | blk_dev_init(); |
edfaa7c3 | 960 | |
561ec68e ZY |
961 | register_blkdev(BLOCK_EXT_MAJOR, "blkext"); |
962 | ||
edfaa7c3 | 963 | /* create top-level block dir */ |
e52eec13 AK |
964 | if (!sysfs_deprecated) |
965 | block_depr = kobject_create_and_add("block", NULL); | |
830d3cfb | 966 | return 0; |
1da177e4 LT |
967 | } |
968 | ||
969 | subsys_initcall(genhd_device_init); | |
970 | ||
edfaa7c3 KS |
971 | static ssize_t disk_range_show(struct device *dev, |
972 | struct device_attribute *attr, char *buf) | |
1da177e4 | 973 | { |
edfaa7c3 | 974 | struct gendisk *disk = dev_to_disk(dev); |
1da177e4 | 975 | |
edfaa7c3 | 976 | return sprintf(buf, "%d\n", disk->minors); |
1da177e4 LT |
977 | } |
978 | ||
1f014290 TH |
979 | static ssize_t disk_ext_range_show(struct device *dev, |
980 | struct device_attribute *attr, char *buf) | |
981 | { | |
982 | struct gendisk *disk = dev_to_disk(dev); | |
983 | ||
b5d0b9df | 984 | return sprintf(buf, "%d\n", disk_max_parts(disk)); |
1f014290 TH |
985 | } |
986 | ||
edfaa7c3 KS |
987 | static ssize_t disk_removable_show(struct device *dev, |
988 | struct device_attribute *attr, char *buf) | |
a7fd6706 | 989 | { |
edfaa7c3 | 990 | struct gendisk *disk = dev_to_disk(dev); |
a7fd6706 | 991 | |
edfaa7c3 KS |
992 | return sprintf(buf, "%d\n", |
993 | (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0)); | |
a7fd6706 KS |
994 | } |
995 | ||
1c9ce527 KS |
996 | static ssize_t disk_ro_show(struct device *dev, |
997 | struct device_attribute *attr, char *buf) | |
998 | { | |
999 | struct gendisk *disk = dev_to_disk(dev); | |
1000 | ||
b7db9956 | 1001 | return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0); |
1c9ce527 KS |
1002 | } |
1003 | ||
edfaa7c3 KS |
1004 | static ssize_t disk_capability_show(struct device *dev, |
1005 | struct device_attribute *attr, char *buf) | |
86ce18d7 | 1006 | { |
edfaa7c3 KS |
1007 | struct gendisk *disk = dev_to_disk(dev); |
1008 | ||
1009 | return sprintf(buf, "%x\n", disk->flags); | |
86ce18d7 | 1010 | } |
edfaa7c3 | 1011 | |
c72758f3 MP |
1012 | static ssize_t disk_alignment_offset_show(struct device *dev, |
1013 | struct device_attribute *attr, | |
1014 | char *buf) | |
1015 | { | |
1016 | struct gendisk *disk = dev_to_disk(dev); | |
1017 | ||
1018 | return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue)); | |
1019 | } | |
1020 | ||
86b37281 MP |
1021 | static ssize_t disk_discard_alignment_show(struct device *dev, |
1022 | struct device_attribute *attr, | |
1023 | char *buf) | |
1024 | { | |
1025 | struct gendisk *disk = dev_to_disk(dev); | |
1026 | ||
dd3d145d | 1027 | return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue)); |
86b37281 MP |
1028 | } |
1029 | ||
edfaa7c3 | 1030 | static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL); |
1f014290 | 1031 | static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL); |
edfaa7c3 | 1032 | static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL); |
1c9ce527 | 1033 | static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL); |
e5610521 | 1034 | static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL); |
c72758f3 | 1035 | static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL); |
86b37281 MP |
1036 | static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show, |
1037 | NULL); | |
edfaa7c3 | 1038 | static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL); |
074a7aca | 1039 | static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL); |
316d315b | 1040 | static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL); |
99e6608c VV |
1041 | static DEVICE_ATTR(badblocks, S_IRUGO | S_IWUSR, disk_badblocks_show, |
1042 | disk_badblocks_store); | |
c17bb495 | 1043 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
edfaa7c3 | 1044 | static struct device_attribute dev_attr_fail = |
eddb2e26 | 1045 | __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store); |
c17bb495 | 1046 | #endif |
581d4e28 JA |
1047 | #ifdef CONFIG_FAIL_IO_TIMEOUT |
1048 | static struct device_attribute dev_attr_fail_timeout = | |
1049 | __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show, | |
1050 | part_timeout_store); | |
1051 | #endif | |
edfaa7c3 KS |
1052 | |
1053 | static struct attribute *disk_attrs[] = { | |
1054 | &dev_attr_range.attr, | |
1f014290 | 1055 | &dev_attr_ext_range.attr, |
edfaa7c3 | 1056 | &dev_attr_removable.attr, |
1c9ce527 | 1057 | &dev_attr_ro.attr, |
edfaa7c3 | 1058 | &dev_attr_size.attr, |
c72758f3 | 1059 | &dev_attr_alignment_offset.attr, |
86b37281 | 1060 | &dev_attr_discard_alignment.attr, |
edfaa7c3 KS |
1061 | &dev_attr_capability.attr, |
1062 | &dev_attr_stat.attr, | |
316d315b | 1063 | &dev_attr_inflight.attr, |
99e6608c | 1064 | &dev_attr_badblocks.attr, |
edfaa7c3 KS |
1065 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
1066 | &dev_attr_fail.attr, | |
581d4e28 JA |
1067 | #endif |
1068 | #ifdef CONFIG_FAIL_IO_TIMEOUT | |
1069 | &dev_attr_fail_timeout.attr, | |
edfaa7c3 KS |
1070 | #endif |
1071 | NULL | |
1072 | }; | |
1073 | ||
1074 | static struct attribute_group disk_attr_group = { | |
1075 | .attrs = disk_attrs, | |
1076 | }; | |
1077 | ||
a4dbd674 | 1078 | static const struct attribute_group *disk_attr_groups[] = { |
edfaa7c3 KS |
1079 | &disk_attr_group, |
1080 | NULL | |
1da177e4 LT |
1081 | }; |
1082 | ||
540eed56 TH |
1083 | /** |
1084 | * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way | |
1085 | * @disk: disk to replace part_tbl for | |
1086 | * @new_ptbl: new part_tbl to install | |
1087 | * | |
1088 | * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The | |
1089 | * original ptbl is freed using RCU callback. | |
1090 | * | |
1091 | * LOCKING: | |
1092 | * Matching bd_mutx locked. | |
1093 | */ | |
1094 | static void disk_replace_part_tbl(struct gendisk *disk, | |
1095 | struct disk_part_tbl *new_ptbl) | |
1096 | { | |
1097 | struct disk_part_tbl *old_ptbl = disk->part_tbl; | |
1098 | ||
1099 | rcu_assign_pointer(disk->part_tbl, new_ptbl); | |
a6f23657 JA |
1100 | |
1101 | if (old_ptbl) { | |
1102 | rcu_assign_pointer(old_ptbl->last_lookup, NULL); | |
57bdfbf9 | 1103 | kfree_rcu(old_ptbl, rcu_head); |
a6f23657 | 1104 | } |
540eed56 TH |
1105 | } |
1106 | ||
1107 | /** | |
1108 | * disk_expand_part_tbl - expand disk->part_tbl | |
1109 | * @disk: disk to expand part_tbl for | |
1110 | * @partno: expand such that this partno can fit in | |
1111 | * | |
1112 | * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl | |
1113 | * uses RCU to allow unlocked dereferencing for stats and other stuff. | |
1114 | * | |
1115 | * LOCKING: | |
1116 | * Matching bd_mutex locked, might sleep. | |
1117 | * | |
1118 | * RETURNS: | |
1119 | * 0 on success, -errno on failure. | |
1120 | */ | |
1121 | int disk_expand_part_tbl(struct gendisk *disk, int partno) | |
1122 | { | |
1123 | struct disk_part_tbl *old_ptbl = disk->part_tbl; | |
1124 | struct disk_part_tbl *new_ptbl; | |
1125 | int len = old_ptbl ? old_ptbl->len : 0; | |
5fabcb4c | 1126 | int i, target; |
540eed56 | 1127 | size_t size; |
5fabcb4c JA |
1128 | |
1129 | /* | |
1130 | * check for int overflow, since we can get here from blkpg_ioctl() | |
1131 | * with a user passed 'partno'. | |
1132 | */ | |
1133 | target = partno + 1; | |
1134 | if (target < 0) | |
1135 | return -EINVAL; | |
540eed56 TH |
1136 | |
1137 | /* disk_max_parts() is zero during initialization, ignore if so */ | |
1138 | if (disk_max_parts(disk) && target > disk_max_parts(disk)) | |
1139 | return -EINVAL; | |
1140 | ||
1141 | if (target <= len) | |
1142 | return 0; | |
1143 | ||
1144 | size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]); | |
1145 | new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id); | |
1146 | if (!new_ptbl) | |
1147 | return -ENOMEM; | |
1148 | ||
540eed56 TH |
1149 | new_ptbl->len = target; |
1150 | ||
1151 | for (i = 0; i < len; i++) | |
1152 | rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]); | |
1153 | ||
1154 | disk_replace_part_tbl(disk, new_ptbl); | |
1155 | return 0; | |
1156 | } | |
1157 | ||
edfaa7c3 | 1158 | static void disk_release(struct device *dev) |
1da177e4 | 1159 | { |
edfaa7c3 KS |
1160 | struct gendisk *disk = dev_to_disk(dev); |
1161 | ||
2da78092 | 1162 | blk_free_devt(dev->devt); |
77ea887e | 1163 | disk_release_events(disk); |
1da177e4 | 1164 | kfree(disk->random); |
540eed56 | 1165 | disk_replace_part_tbl(disk, NULL); |
b54e5ed8 | 1166 | hd_free_part(&disk->part0); |
523e1d39 TH |
1167 | if (disk->queue) |
1168 | blk_put_queue(disk->queue); | |
1da177e4 LT |
1169 | kfree(disk); |
1170 | } | |
edfaa7c3 KS |
1171 | struct class block_class = { |
1172 | .name = "block", | |
1da177e4 LT |
1173 | }; |
1174 | ||
3c2670e6 | 1175 | static char *block_devnode(struct device *dev, umode_t *mode, |
4e4098a3 | 1176 | kuid_t *uid, kgid_t *gid) |
b03f38b6 KS |
1177 | { |
1178 | struct gendisk *disk = dev_to_disk(dev); | |
1179 | ||
e454cea2 KS |
1180 | if (disk->devnode) |
1181 | return disk->devnode(disk, mode); | |
b03f38b6 KS |
1182 | return NULL; |
1183 | } | |
1184 | ||
1826eadf | 1185 | static struct device_type disk_type = { |
edfaa7c3 KS |
1186 | .name = "disk", |
1187 | .groups = disk_attr_groups, | |
1188 | .release = disk_release, | |
e454cea2 | 1189 | .devnode = block_devnode, |
1da177e4 LT |
1190 | }; |
1191 | ||
a6e2ba88 | 1192 | #ifdef CONFIG_PROC_FS |
cf771cb5 TH |
1193 | /* |
1194 | * aggregate disk stat collector. Uses the same stats that the sysfs | |
1195 | * entries do, above, but makes them available through one seq_file. | |
1196 | * | |
1197 | * The output looks suspiciously like /proc/partitions with a bunch of | |
1198 | * extra fields. | |
1199 | */ | |
1200 | static int diskstats_show(struct seq_file *seqf, void *v) | |
1da177e4 LT |
1201 | { |
1202 | struct gendisk *gp = v; | |
e71bf0d0 TH |
1203 | struct disk_part_iter piter; |
1204 | struct hd_struct *hd; | |
1da177e4 | 1205 | char buf[BDEVNAME_SIZE]; |
c9959059 | 1206 | int cpu; |
1da177e4 LT |
1207 | |
1208 | /* | |
ed9e1982 | 1209 | if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next) |
cf771cb5 | 1210 | seq_puts(seqf, "major minor name" |
1da177e4 LT |
1211 | " rio rmerge rsect ruse wio wmerge " |
1212 | "wsect wuse running use aveq" | |
1213 | "\n\n"); | |
1214 | */ | |
9f5e4865 | 1215 | |
71982a40 | 1216 | disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0); |
e71bf0d0 | 1217 | while ((hd = disk_part_iter_next(&piter))) { |
074a7aca | 1218 | cpu = part_stat_lock(); |
c9959059 | 1219 | part_round_stats(cpu, hd); |
074a7aca | 1220 | part_stat_unlock(); |
f95fe9cf HP |
1221 | seq_printf(seqf, "%4d %7d %s %lu %lu %lu " |
1222 | "%u %lu %lu %lu %u %u %u %u\n", | |
f331c029 TH |
1223 | MAJOR(part_devt(hd)), MINOR(part_devt(hd)), |
1224 | disk_name(gp, hd->partno, buf), | |
53f22956 LY |
1225 | part_stat_read(hd, ios[READ]), |
1226 | part_stat_read(hd, merges[READ]), | |
f95fe9cf | 1227 | part_stat_read(hd, sectors[READ]), |
53f22956 LY |
1228 | jiffies_to_msecs(part_stat_read(hd, ticks[READ])), |
1229 | part_stat_read(hd, ios[WRITE]), | |
1230 | part_stat_read(hd, merges[WRITE]), | |
f95fe9cf | 1231 | part_stat_read(hd, sectors[WRITE]), |
53f22956 | 1232 | jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])), |
316d315b | 1233 | part_in_flight(hd), |
28f39d55 JM |
1234 | jiffies_to_msecs(part_stat_read(hd, io_ticks)), |
1235 | jiffies_to_msecs(part_stat_read(hd, time_in_queue)) | |
1236 | ); | |
1da177e4 | 1237 | } |
e71bf0d0 | 1238 | disk_part_iter_exit(&piter); |
9f5e4865 | 1239 | |
1da177e4 LT |
1240 | return 0; |
1241 | } | |
1242 | ||
31d85ab2 | 1243 | static const struct seq_operations diskstats_op = { |
def4e38d TH |
1244 | .start = disk_seqf_start, |
1245 | .next = disk_seqf_next, | |
1246 | .stop = disk_seqf_stop, | |
1da177e4 LT |
1247 | .show = diskstats_show |
1248 | }; | |
f500975a | 1249 | |
31d85ab2 AD |
1250 | static int diskstats_open(struct inode *inode, struct file *file) |
1251 | { | |
1252 | return seq_open(file, &diskstats_op); | |
1253 | } | |
1254 | ||
1255 | static const struct file_operations proc_diskstats_operations = { | |
1256 | .open = diskstats_open, | |
1257 | .read = seq_read, | |
1258 | .llseek = seq_lseek, | |
1259 | .release = seq_release, | |
1260 | }; | |
1261 | ||
f500975a AD |
1262 | static int __init proc_genhd_init(void) |
1263 | { | |
31d85ab2 | 1264 | proc_create("diskstats", 0, NULL, &proc_diskstats_operations); |
f500975a AD |
1265 | proc_create("partitions", 0, NULL, &proc_partitions_operations); |
1266 | return 0; | |
1267 | } | |
1268 | module_init(proc_genhd_init); | |
a6e2ba88 | 1269 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 1270 | |
cf771cb5 | 1271 | dev_t blk_lookup_devt(const char *name, int partno) |
a142be85 | 1272 | { |
def4e38d TH |
1273 | dev_t devt = MKDEV(0, 0); |
1274 | struct class_dev_iter iter; | |
1275 | struct device *dev; | |
a142be85 | 1276 | |
def4e38d TH |
1277 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); |
1278 | while ((dev = class_dev_iter_next(&iter))) { | |
a142be85 | 1279 | struct gendisk *disk = dev_to_disk(dev); |
548b10eb | 1280 | struct hd_struct *part; |
a142be85 | 1281 | |
3ada8b7e | 1282 | if (strcmp(dev_name(dev), name)) |
f331c029 | 1283 | continue; |
f331c029 | 1284 | |
41b8c853 NB |
1285 | if (partno < disk->minors) { |
1286 | /* We need to return the right devno, even | |
1287 | * if the partition doesn't exist yet. | |
1288 | */ | |
1289 | devt = MKDEV(MAJOR(dev->devt), | |
1290 | MINOR(dev->devt) + partno); | |
1291 | break; | |
1292 | } | |
548b10eb | 1293 | part = disk_get_part(disk, partno); |
2bbedcb4 | 1294 | if (part) { |
f331c029 | 1295 | devt = part_devt(part); |
e71bf0d0 | 1296 | disk_put_part(part); |
548b10eb | 1297 | break; |
def4e38d | 1298 | } |
548b10eb | 1299 | disk_put_part(part); |
5c0ef6d0 | 1300 | } |
def4e38d | 1301 | class_dev_iter_exit(&iter); |
edfaa7c3 KS |
1302 | return devt; |
1303 | } | |
edfaa7c3 KS |
1304 | EXPORT_SYMBOL(blk_lookup_devt); |
1305 | ||
1da177e4 LT |
1306 | struct gendisk *alloc_disk(int minors) |
1307 | { | |
c304a51b | 1308 | return alloc_disk_node(minors, NUMA_NO_NODE); |
1946089a | 1309 | } |
689d6fac | 1310 | EXPORT_SYMBOL(alloc_disk); |
1946089a CL |
1311 | |
1312 | struct gendisk *alloc_disk_node(int minors, int node_id) | |
1313 | { | |
1314 | struct gendisk *disk; | |
1315 | ||
c1b511eb | 1316 | disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id); |
1da177e4 | 1317 | if (disk) { |
074a7aca | 1318 | if (!init_part_stats(&disk->part0)) { |
1da177e4 LT |
1319 | kfree(disk); |
1320 | return NULL; | |
1321 | } | |
bf91db18 | 1322 | disk->node_id = node_id; |
540eed56 TH |
1323 | if (disk_expand_part_tbl(disk, 0)) { |
1324 | free_part_stats(&disk->part0); | |
b5d0b9df TH |
1325 | kfree(disk); |
1326 | return NULL; | |
1da177e4 | 1327 | } |
540eed56 | 1328 | disk->part_tbl->part[0] = &disk->part0; |
6c23a968 | 1329 | |
c83f6bf9 VG |
1330 | /* |
1331 | * set_capacity() and get_capacity() currently don't use | |
1332 | * seqcounter to read/update the part0->nr_sects. Still init | |
1333 | * the counter as we can read the sectors in IO submission | |
1334 | * patch using seqence counters. | |
1335 | * | |
1336 | * TODO: Ideally set_capacity() and get_capacity() should be | |
1337 | * converted to make use of bd_mutex and sequence counters. | |
1338 | */ | |
1339 | seqcount_init(&disk->part0.nr_sects_seq); | |
6c71013e ML |
1340 | if (hd_ref_init(&disk->part0)) { |
1341 | hd_free_part(&disk->part0); | |
1342 | kfree(disk); | |
1343 | return NULL; | |
1344 | } | |
b5d0b9df | 1345 | |
1da177e4 | 1346 | disk->minors = minors; |
1da177e4 | 1347 | rand_initialize_disk(disk); |
ed9e1982 TH |
1348 | disk_to_dev(disk)->class = &block_class; |
1349 | disk_to_dev(disk)->type = &disk_type; | |
1350 | device_initialize(disk_to_dev(disk)); | |
1da177e4 LT |
1351 | } |
1352 | return disk; | |
1353 | } | |
1946089a | 1354 | EXPORT_SYMBOL(alloc_disk_node); |
1da177e4 LT |
1355 | |
1356 | struct kobject *get_disk(struct gendisk *disk) | |
1357 | { | |
1358 | struct module *owner; | |
1359 | struct kobject *kobj; | |
1360 | ||
1361 | if (!disk->fops) | |
1362 | return NULL; | |
1363 | owner = disk->fops->owner; | |
1364 | if (owner && !try_module_get(owner)) | |
1365 | return NULL; | |
ed9e1982 | 1366 | kobj = kobject_get(&disk_to_dev(disk)->kobj); |
1da177e4 LT |
1367 | if (kobj == NULL) { |
1368 | module_put(owner); | |
1369 | return NULL; | |
1370 | } | |
1371 | return kobj; | |
1372 | ||
1373 | } | |
1374 | ||
1375 | EXPORT_SYMBOL(get_disk); | |
1376 | ||
1377 | void put_disk(struct gendisk *disk) | |
1378 | { | |
1379 | if (disk) | |
ed9e1982 | 1380 | kobject_put(&disk_to_dev(disk)->kobj); |
1da177e4 LT |
1381 | } |
1382 | ||
1383 | EXPORT_SYMBOL(put_disk); | |
1384 | ||
e3264a4d HR |
1385 | static void set_disk_ro_uevent(struct gendisk *gd, int ro) |
1386 | { | |
1387 | char event[] = "DISK_RO=1"; | |
1388 | char *envp[] = { event, NULL }; | |
1389 | ||
1390 | if (!ro) | |
1391 | event[8] = '0'; | |
1392 | kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); | |
1393 | } | |
1394 | ||
1da177e4 LT |
1395 | void set_device_ro(struct block_device *bdev, int flag) |
1396 | { | |
b7db9956 | 1397 | bdev->bd_part->policy = flag; |
1da177e4 LT |
1398 | } |
1399 | ||
1400 | EXPORT_SYMBOL(set_device_ro); | |
1401 | ||
1402 | void set_disk_ro(struct gendisk *disk, int flag) | |
1403 | { | |
e71bf0d0 TH |
1404 | struct disk_part_iter piter; |
1405 | struct hd_struct *part; | |
1406 | ||
e3264a4d HR |
1407 | if (disk->part0.policy != flag) { |
1408 | set_disk_ro_uevent(disk, flag); | |
1409 | disk->part0.policy = flag; | |
1410 | } | |
1411 | ||
1412 | disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY); | |
e71bf0d0 TH |
1413 | while ((part = disk_part_iter_next(&piter))) |
1414 | part->policy = flag; | |
1415 | disk_part_iter_exit(&piter); | |
1da177e4 LT |
1416 | } |
1417 | ||
1418 | EXPORT_SYMBOL(set_disk_ro); | |
1419 | ||
1420 | int bdev_read_only(struct block_device *bdev) | |
1421 | { | |
1422 | if (!bdev) | |
1423 | return 0; | |
b7db9956 | 1424 | return bdev->bd_part->policy; |
1da177e4 LT |
1425 | } |
1426 | ||
1427 | EXPORT_SYMBOL(bdev_read_only); | |
1428 | ||
cf771cb5 | 1429 | int invalidate_partition(struct gendisk *disk, int partno) |
1da177e4 LT |
1430 | { |
1431 | int res = 0; | |
cf771cb5 | 1432 | struct block_device *bdev = bdget_disk(disk, partno); |
1da177e4 | 1433 | if (bdev) { |
2ef41634 | 1434 | fsync_bdev(bdev); |
93b270f7 | 1435 | res = __invalidate_device(bdev, true); |
1da177e4 LT |
1436 | bdput(bdev); |
1437 | } | |
1438 | return res; | |
1439 | } | |
1440 | ||
1441 | EXPORT_SYMBOL(invalidate_partition); | |
77ea887e TH |
1442 | |
1443 | /* | |
1444 | * Disk events - monitor disk events like media change and eject request. | |
1445 | */ | |
1446 | struct disk_events { | |
1447 | struct list_head node; /* all disk_event's */ | |
1448 | struct gendisk *disk; /* the associated disk */ | |
1449 | spinlock_t lock; | |
1450 | ||
fdd514e1 | 1451 | struct mutex block_mutex; /* protects blocking */ |
77ea887e TH |
1452 | int block; /* event blocking depth */ |
1453 | unsigned int pending; /* events already sent out */ | |
1454 | unsigned int clearing; /* events being cleared */ | |
1455 | ||
1456 | long poll_msecs; /* interval, -1 for default */ | |
1457 | struct delayed_work dwork; | |
1458 | }; | |
1459 | ||
1460 | static const char *disk_events_strs[] = { | |
1461 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change", | |
1462 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request", | |
1463 | }; | |
1464 | ||
1465 | static char *disk_uevents[] = { | |
1466 | [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1", | |
1467 | [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1", | |
1468 | }; | |
1469 | ||
1470 | /* list of all disk_events */ | |
1471 | static DEFINE_MUTEX(disk_events_mutex); | |
1472 | static LIST_HEAD(disk_events); | |
1473 | ||
1474 | /* disable in-kernel polling by default */ | |
1fe8f348 | 1475 | static unsigned long disk_events_dfl_poll_msecs; |
77ea887e TH |
1476 | |
1477 | static unsigned long disk_events_poll_jiffies(struct gendisk *disk) | |
1478 | { | |
1479 | struct disk_events *ev = disk->ev; | |
1480 | long intv_msecs = 0; | |
1481 | ||
1482 | /* | |
1483 | * If device-specific poll interval is set, always use it. If | |
1484 | * the default is being used, poll iff there are events which | |
1485 | * can't be monitored asynchronously. | |
1486 | */ | |
1487 | if (ev->poll_msecs >= 0) | |
1488 | intv_msecs = ev->poll_msecs; | |
1489 | else if (disk->events & ~disk->async_events) | |
1490 | intv_msecs = disk_events_dfl_poll_msecs; | |
1491 | ||
1492 | return msecs_to_jiffies(intv_msecs); | |
1493 | } | |
1494 | ||
c3af54af TH |
1495 | /** |
1496 | * disk_block_events - block and flush disk event checking | |
1497 | * @disk: disk to block events for | |
1498 | * | |
1499 | * On return from this function, it is guaranteed that event checking | |
1500 | * isn't in progress and won't happen until unblocked by | |
1501 | * disk_unblock_events(). Events blocking is counted and the actual | |
1502 | * unblocking happens after the matching number of unblocks are done. | |
1503 | * | |
1504 | * Note that this intentionally does not block event checking from | |
1505 | * disk_clear_events(). | |
1506 | * | |
1507 | * CONTEXT: | |
1508 | * Might sleep. | |
1509 | */ | |
1510 | void disk_block_events(struct gendisk *disk) | |
77ea887e TH |
1511 | { |
1512 | struct disk_events *ev = disk->ev; | |
1513 | unsigned long flags; | |
1514 | bool cancel; | |
1515 | ||
c3af54af TH |
1516 | if (!ev) |
1517 | return; | |
1518 | ||
fdd514e1 TH |
1519 | /* |
1520 | * Outer mutex ensures that the first blocker completes canceling | |
1521 | * the event work before further blockers are allowed to finish. | |
1522 | */ | |
1523 | mutex_lock(&ev->block_mutex); | |
1524 | ||
77ea887e TH |
1525 | spin_lock_irqsave(&ev->lock, flags); |
1526 | cancel = !ev->block++; | |
1527 | spin_unlock_irqrestore(&ev->lock, flags); | |
1528 | ||
c3af54af TH |
1529 | if (cancel) |
1530 | cancel_delayed_work_sync(&disk->ev->dwork); | |
fdd514e1 TH |
1531 | |
1532 | mutex_unlock(&ev->block_mutex); | |
77ea887e TH |
1533 | } |
1534 | ||
1535 | static void __disk_unblock_events(struct gendisk *disk, bool check_now) | |
1536 | { | |
1537 | struct disk_events *ev = disk->ev; | |
1538 | unsigned long intv; | |
1539 | unsigned long flags; | |
1540 | ||
1541 | spin_lock_irqsave(&ev->lock, flags); | |
1542 | ||
1543 | if (WARN_ON_ONCE(ev->block <= 0)) | |
1544 | goto out_unlock; | |
1545 | ||
1546 | if (--ev->block) | |
1547 | goto out_unlock; | |
1548 | ||
77ea887e | 1549 | intv = disk_events_poll_jiffies(disk); |
77ea887e | 1550 | if (check_now) |
695588f9 VK |
1551 | queue_delayed_work(system_freezable_power_efficient_wq, |
1552 | &ev->dwork, 0); | |
77ea887e | 1553 | else if (intv) |
695588f9 VK |
1554 | queue_delayed_work(system_freezable_power_efficient_wq, |
1555 | &ev->dwork, intv); | |
77ea887e TH |
1556 | out_unlock: |
1557 | spin_unlock_irqrestore(&ev->lock, flags); | |
1558 | } | |
1559 | ||
77ea887e TH |
1560 | /** |
1561 | * disk_unblock_events - unblock disk event checking | |
1562 | * @disk: disk to unblock events for | |
1563 | * | |
1564 | * Undo disk_block_events(). When the block count reaches zero, it | |
1565 | * starts events polling if configured. | |
1566 | * | |
1567 | * CONTEXT: | |
1568 | * Don't care. Safe to call from irq context. | |
1569 | */ | |
1570 | void disk_unblock_events(struct gendisk *disk) | |
1571 | { | |
1572 | if (disk->ev) | |
facc31dd | 1573 | __disk_unblock_events(disk, false); |
77ea887e TH |
1574 | } |
1575 | ||
1576 | /** | |
85ef06d1 TH |
1577 | * disk_flush_events - schedule immediate event checking and flushing |
1578 | * @disk: disk to check and flush events for | |
1579 | * @mask: events to flush | |
77ea887e | 1580 | * |
85ef06d1 TH |
1581 | * Schedule immediate event checking on @disk if not blocked. Events in |
1582 | * @mask are scheduled to be cleared from the driver. Note that this | |
1583 | * doesn't clear the events from @disk->ev. | |
77ea887e TH |
1584 | * |
1585 | * CONTEXT: | |
85ef06d1 | 1586 | * If @mask is non-zero must be called with bdev->bd_mutex held. |
77ea887e | 1587 | */ |
85ef06d1 | 1588 | void disk_flush_events(struct gendisk *disk, unsigned int mask) |
77ea887e | 1589 | { |
a9dce2a3 | 1590 | struct disk_events *ev = disk->ev; |
a9dce2a3 TH |
1591 | |
1592 | if (!ev) | |
1593 | return; | |
1594 | ||
85ef06d1 TH |
1595 | spin_lock_irq(&ev->lock); |
1596 | ev->clearing |= mask; | |
41f63c53 | 1597 | if (!ev->block) |
695588f9 VK |
1598 | mod_delayed_work(system_freezable_power_efficient_wq, |
1599 | &ev->dwork, 0); | |
85ef06d1 | 1600 | spin_unlock_irq(&ev->lock); |
77ea887e | 1601 | } |
77ea887e TH |
1602 | |
1603 | /** | |
1604 | * disk_clear_events - synchronously check, clear and return pending events | |
1605 | * @disk: disk to fetch and clear events from | |
da3dae54 | 1606 | * @mask: mask of events to be fetched and cleared |
77ea887e TH |
1607 | * |
1608 | * Disk events are synchronously checked and pending events in @mask | |
1609 | * are cleared and returned. This ignores the block count. | |
1610 | * | |
1611 | * CONTEXT: | |
1612 | * Might sleep. | |
1613 | */ | |
1614 | unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask) | |
1615 | { | |
1616 | const struct block_device_operations *bdops = disk->fops; | |
1617 | struct disk_events *ev = disk->ev; | |
1618 | unsigned int pending; | |
12c2bdb2 | 1619 | unsigned int clearing = mask; |
77ea887e TH |
1620 | |
1621 | if (!ev) { | |
1622 | /* for drivers still using the old ->media_changed method */ | |
1623 | if ((mask & DISK_EVENT_MEDIA_CHANGE) && | |
1624 | bdops->media_changed && bdops->media_changed(disk)) | |
1625 | return DISK_EVENT_MEDIA_CHANGE; | |
1626 | return 0; | |
1627 | } | |
1628 | ||
12c2bdb2 DB |
1629 | disk_block_events(disk); |
1630 | ||
1631 | /* | |
1632 | * store the union of mask and ev->clearing on the stack so that the | |
1633 | * race with disk_flush_events does not cause ambiguity (ev->clearing | |
1634 | * can still be modified even if events are blocked). | |
1635 | */ | |
77ea887e | 1636 | spin_lock_irq(&ev->lock); |
12c2bdb2 DB |
1637 | clearing |= ev->clearing; |
1638 | ev->clearing = 0; | |
77ea887e TH |
1639 | spin_unlock_irq(&ev->lock); |
1640 | ||
12c2bdb2 | 1641 | disk_check_events(ev, &clearing); |
aea24a8b | 1642 | /* |
12c2bdb2 DB |
1643 | * if ev->clearing is not 0, the disk_flush_events got called in the |
1644 | * middle of this function, so we want to run the workfn without delay. | |
aea24a8b | 1645 | */ |
12c2bdb2 | 1646 | __disk_unblock_events(disk, ev->clearing ? true : false); |
77ea887e TH |
1647 | |
1648 | /* then, fetch and clear pending events */ | |
1649 | spin_lock_irq(&ev->lock); | |
77ea887e TH |
1650 | pending = ev->pending & mask; |
1651 | ev->pending &= ~mask; | |
1652 | spin_unlock_irq(&ev->lock); | |
12c2bdb2 | 1653 | WARN_ON_ONCE(clearing & mask); |
77ea887e TH |
1654 | |
1655 | return pending; | |
1656 | } | |
1657 | ||
12c2bdb2 DB |
1658 | /* |
1659 | * Separate this part out so that a different pointer for clearing_ptr can be | |
1660 | * passed in for disk_clear_events. | |
1661 | */ | |
77ea887e TH |
1662 | static void disk_events_workfn(struct work_struct *work) |
1663 | { | |
1664 | struct delayed_work *dwork = to_delayed_work(work); | |
1665 | struct disk_events *ev = container_of(dwork, struct disk_events, dwork); | |
12c2bdb2 DB |
1666 | |
1667 | disk_check_events(ev, &ev->clearing); | |
1668 | } | |
1669 | ||
1670 | static void disk_check_events(struct disk_events *ev, | |
1671 | unsigned int *clearing_ptr) | |
1672 | { | |
77ea887e TH |
1673 | struct gendisk *disk = ev->disk; |
1674 | char *envp[ARRAY_SIZE(disk_uevents) + 1] = { }; | |
12c2bdb2 | 1675 | unsigned int clearing = *clearing_ptr; |
77ea887e TH |
1676 | unsigned int events; |
1677 | unsigned long intv; | |
1678 | int nr_events = 0, i; | |
1679 | ||
1680 | /* check events */ | |
1681 | events = disk->fops->check_events(disk, clearing); | |
1682 | ||
1683 | /* accumulate pending events and schedule next poll if necessary */ | |
1684 | spin_lock_irq(&ev->lock); | |
1685 | ||
1686 | events &= ~ev->pending; | |
1687 | ev->pending |= events; | |
12c2bdb2 | 1688 | *clearing_ptr &= ~clearing; |
77ea887e TH |
1689 | |
1690 | intv = disk_events_poll_jiffies(disk); | |
1691 | if (!ev->block && intv) | |
695588f9 VK |
1692 | queue_delayed_work(system_freezable_power_efficient_wq, |
1693 | &ev->dwork, intv); | |
77ea887e TH |
1694 | |
1695 | spin_unlock_irq(&ev->lock); | |
1696 | ||
7c88a168 TH |
1697 | /* |
1698 | * Tell userland about new events. Only the events listed in | |
1699 | * @disk->events are reported. Unlisted events are processed the | |
1700 | * same internally but never get reported to userland. | |
1701 | */ | |
77ea887e | 1702 | for (i = 0; i < ARRAY_SIZE(disk_uevents); i++) |
7c88a168 | 1703 | if (events & disk->events & (1 << i)) |
77ea887e TH |
1704 | envp[nr_events++] = disk_uevents[i]; |
1705 | ||
1706 | if (nr_events) | |
1707 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
1708 | } | |
1709 | ||
1710 | /* | |
1711 | * A disk events enabled device has the following sysfs nodes under | |
1712 | * its /sys/block/X/ directory. | |
1713 | * | |
1714 | * events : list of all supported events | |
1715 | * events_async : list of events which can be detected w/o polling | |
1716 | * events_poll_msecs : polling interval, 0: disable, -1: system default | |
1717 | */ | |
1718 | static ssize_t __disk_events_show(unsigned int events, char *buf) | |
1719 | { | |
1720 | const char *delim = ""; | |
1721 | ssize_t pos = 0; | |
1722 | int i; | |
1723 | ||
1724 | for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++) | |
1725 | if (events & (1 << i)) { | |
1726 | pos += sprintf(buf + pos, "%s%s", | |
1727 | delim, disk_events_strs[i]); | |
1728 | delim = " "; | |
1729 | } | |
1730 | if (pos) | |
1731 | pos += sprintf(buf + pos, "\n"); | |
1732 | return pos; | |
1733 | } | |
1734 | ||
1735 | static ssize_t disk_events_show(struct device *dev, | |
1736 | struct device_attribute *attr, char *buf) | |
1737 | { | |
1738 | struct gendisk *disk = dev_to_disk(dev); | |
1739 | ||
1740 | return __disk_events_show(disk->events, buf); | |
1741 | } | |
1742 | ||
1743 | static ssize_t disk_events_async_show(struct device *dev, | |
1744 | struct device_attribute *attr, char *buf) | |
1745 | { | |
1746 | struct gendisk *disk = dev_to_disk(dev); | |
1747 | ||
1748 | return __disk_events_show(disk->async_events, buf); | |
1749 | } | |
1750 | ||
1751 | static ssize_t disk_events_poll_msecs_show(struct device *dev, | |
1752 | struct device_attribute *attr, | |
1753 | char *buf) | |
1754 | { | |
1755 | struct gendisk *disk = dev_to_disk(dev); | |
1756 | ||
1757 | return sprintf(buf, "%ld\n", disk->ev->poll_msecs); | |
1758 | } | |
1759 | ||
1760 | static ssize_t disk_events_poll_msecs_store(struct device *dev, | |
1761 | struct device_attribute *attr, | |
1762 | const char *buf, size_t count) | |
1763 | { | |
1764 | struct gendisk *disk = dev_to_disk(dev); | |
1765 | long intv; | |
1766 | ||
1767 | if (!count || !sscanf(buf, "%ld", &intv)) | |
1768 | return -EINVAL; | |
1769 | ||
1770 | if (intv < 0 && intv != -1) | |
1771 | return -EINVAL; | |
1772 | ||
c3af54af | 1773 | disk_block_events(disk); |
77ea887e TH |
1774 | disk->ev->poll_msecs = intv; |
1775 | __disk_unblock_events(disk, true); | |
1776 | ||
1777 | return count; | |
1778 | } | |
1779 | ||
1780 | static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL); | |
1781 | static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL); | |
1782 | static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR, | |
1783 | disk_events_poll_msecs_show, | |
1784 | disk_events_poll_msecs_store); | |
1785 | ||
1786 | static const struct attribute *disk_events_attrs[] = { | |
1787 | &dev_attr_events.attr, | |
1788 | &dev_attr_events_async.attr, | |
1789 | &dev_attr_events_poll_msecs.attr, | |
1790 | NULL, | |
1791 | }; | |
1792 | ||
1793 | /* | |
1794 | * The default polling interval can be specified by the kernel | |
1795 | * parameter block.events_dfl_poll_msecs which defaults to 0 | |
1796 | * (disable). This can also be modified runtime by writing to | |
1797 | * /sys/module/block/events_dfl_poll_msecs. | |
1798 | */ | |
1799 | static int disk_events_set_dfl_poll_msecs(const char *val, | |
1800 | const struct kernel_param *kp) | |
1801 | { | |
1802 | struct disk_events *ev; | |
1803 | int ret; | |
1804 | ||
1805 | ret = param_set_ulong(val, kp); | |
1806 | if (ret < 0) | |
1807 | return ret; | |
1808 | ||
1809 | mutex_lock(&disk_events_mutex); | |
1810 | ||
1811 | list_for_each_entry(ev, &disk_events, node) | |
85ef06d1 | 1812 | disk_flush_events(ev->disk, 0); |
77ea887e TH |
1813 | |
1814 | mutex_unlock(&disk_events_mutex); | |
1815 | ||
1816 | return 0; | |
1817 | } | |
1818 | ||
1819 | static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = { | |
1820 | .set = disk_events_set_dfl_poll_msecs, | |
1821 | .get = param_get_ulong, | |
1822 | }; | |
1823 | ||
1824 | #undef MODULE_PARAM_PREFIX | |
1825 | #define MODULE_PARAM_PREFIX "block." | |
1826 | ||
1827 | module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops, | |
1828 | &disk_events_dfl_poll_msecs, 0644); | |
1829 | ||
1830 | /* | |
9f53d2fe | 1831 | * disk_{alloc|add|del|release}_events - initialize and destroy disk_events. |
77ea887e | 1832 | */ |
9f53d2fe | 1833 | static void disk_alloc_events(struct gendisk *disk) |
77ea887e TH |
1834 | { |
1835 | struct disk_events *ev; | |
1836 | ||
75e3f3ee | 1837 | if (!disk->fops->check_events) |
77ea887e TH |
1838 | return; |
1839 | ||
1840 | ev = kzalloc(sizeof(*ev), GFP_KERNEL); | |
1841 | if (!ev) { | |
1842 | pr_warn("%s: failed to initialize events\n", disk->disk_name); | |
1843 | return; | |
1844 | } | |
1845 | ||
77ea887e TH |
1846 | INIT_LIST_HEAD(&ev->node); |
1847 | ev->disk = disk; | |
1848 | spin_lock_init(&ev->lock); | |
fdd514e1 | 1849 | mutex_init(&ev->block_mutex); |
77ea887e TH |
1850 | ev->block = 1; |
1851 | ev->poll_msecs = -1; | |
1852 | INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn); | |
1853 | ||
9f53d2fe SG |
1854 | disk->ev = ev; |
1855 | } | |
1856 | ||
1857 | static void disk_add_events(struct gendisk *disk) | |
1858 | { | |
1859 | if (!disk->ev) | |
1860 | return; | |
1861 | ||
1862 | /* FIXME: error handling */ | |
1863 | if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0) | |
1864 | pr_warn("%s: failed to create sysfs files for events\n", | |
1865 | disk->disk_name); | |
1866 | ||
77ea887e | 1867 | mutex_lock(&disk_events_mutex); |
9f53d2fe | 1868 | list_add_tail(&disk->ev->node, &disk_events); |
77ea887e TH |
1869 | mutex_unlock(&disk_events_mutex); |
1870 | ||
1871 | /* | |
1872 | * Block count is initialized to 1 and the following initial | |
1873 | * unblock kicks it into action. | |
1874 | */ | |
1875 | __disk_unblock_events(disk, true); | |
1876 | } | |
1877 | ||
1878 | static void disk_del_events(struct gendisk *disk) | |
1879 | { | |
1880 | if (!disk->ev) | |
1881 | return; | |
1882 | ||
c3af54af | 1883 | disk_block_events(disk); |
77ea887e TH |
1884 | |
1885 | mutex_lock(&disk_events_mutex); | |
1886 | list_del_init(&disk->ev->node); | |
1887 | mutex_unlock(&disk_events_mutex); | |
1888 | ||
1889 | sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs); | |
1890 | } | |
1891 | ||
1892 | static void disk_release_events(struct gendisk *disk) | |
1893 | { | |
1894 | /* the block count should be 1 from disk_del_events() */ | |
1895 | WARN_ON_ONCE(disk->ev && disk->ev->block != 1); | |
1896 | kfree(disk->ev); | |
1897 | } |