]>
Commit | Line | Data |
---|---|---|
9db5579b NP |
1 | /* |
2 | * Ram backed block device driver. | |
3 | * | |
4 | * Copyright (C) 2007 Nick Piggin | |
5 | * Copyright (C) 2007 Novell Inc. | |
6 | * | |
7 | * Parts derived from drivers/block/rd.c, and drivers/block/loop.c, copyright | |
8 | * of their respective owners. | |
9 | */ | |
10 | ||
11 | #include <linux/init.h> | |
12 | #include <linux/module.h> | |
13 | #include <linux/moduleparam.h> | |
14 | #include <linux/major.h> | |
15 | #include <linux/blkdev.h> | |
16 | #include <linux/bio.h> | |
17 | #include <linux/highmem.h> | |
2a48fc0a | 18 | #include <linux/mutex.h> |
9db5579b | 19 | #include <linux/radix-tree.h> |
ff01bb48 | 20 | #include <linux/fs.h> |
5a0e3ad6 | 21 | #include <linux/slab.h> |
9db5579b NP |
22 | |
23 | #include <asm/uaccess.h> | |
24 | ||
25 | #define SECTOR_SHIFT 9 | |
26 | #define PAGE_SECTORS_SHIFT (PAGE_SHIFT - SECTOR_SHIFT) | |
27 | #define PAGE_SECTORS (1 << PAGE_SECTORS_SHIFT) | |
28 | ||
29 | /* | |
30 | * Each block ramdisk device has a radix_tree brd_pages of pages that stores | |
31 | * the pages containing the block device's contents. A brd page's ->index is | |
32 | * its offset in PAGE_SIZE units. This is similar to, but in no way connected | |
33 | * with, the kernel's pagecache or buffer cache (which sit above our block | |
34 | * device). | |
35 | */ | |
36 | struct brd_device { | |
37 | int brd_number; | |
9db5579b NP |
38 | |
39 | struct request_queue *brd_queue; | |
40 | struct gendisk *brd_disk; | |
41 | struct list_head brd_list; | |
42 | ||
43 | /* | |
44 | * Backing store of pages and lock to protect it. This is the contents | |
45 | * of the block device. | |
46 | */ | |
47 | spinlock_t brd_lock; | |
48 | struct radix_tree_root brd_pages; | |
49 | }; | |
50 | ||
51 | /* | |
52 | * Look up and return a brd's page for a given sector. | |
53 | */ | |
2a48fc0a | 54 | static DEFINE_MUTEX(brd_mutex); |
9db5579b NP |
55 | static struct page *brd_lookup_page(struct brd_device *brd, sector_t sector) |
56 | { | |
57 | pgoff_t idx; | |
58 | struct page *page; | |
59 | ||
60 | /* | |
61 | * The page lifetime is protected by the fact that we have opened the | |
62 | * device node -- brd pages will never be deleted under us, so we | |
63 | * don't need any further locking or refcounting. | |
64 | * | |
65 | * This is strictly true for the radix-tree nodes as well (ie. we | |
66 | * don't actually need the rcu_read_lock()), however that is not a | |
67 | * documented feature of the radix-tree API so it is better to be | |
68 | * safe here (we don't have total exclusion from radix tree updates | |
69 | * here, only deletes). | |
70 | */ | |
71 | rcu_read_lock(); | |
72 | idx = sector >> PAGE_SECTORS_SHIFT; /* sector to page index */ | |
73 | page = radix_tree_lookup(&brd->brd_pages, idx); | |
74 | rcu_read_unlock(); | |
75 | ||
76 | BUG_ON(page && page->index != idx); | |
77 | ||
78 | return page; | |
79 | } | |
80 | ||
81 | /* | |
82 | * Look up and return a brd's page for a given sector. | |
83 | * If one does not exist, allocate an empty page, and insert that. Then | |
84 | * return it. | |
85 | */ | |
86 | static struct page *brd_insert_page(struct brd_device *brd, sector_t sector) | |
87 | { | |
88 | pgoff_t idx; | |
89 | struct page *page; | |
75acb9cd | 90 | gfp_t gfp_flags; |
9db5579b NP |
91 | |
92 | page = brd_lookup_page(brd, sector); | |
93 | if (page) | |
94 | return page; | |
95 | ||
96 | /* | |
97 | * Must use NOIO because we don't want to recurse back into the | |
98 | * block or filesystem layers from page reclaim. | |
75acb9cd | 99 | * |
a7a97fc9 MW |
100 | * Cannot support DAX and highmem, because our ->direct_access |
101 | * routine for DAX must return memory that is always addressable. | |
102 | * If DAX was reworked to use pfns and kmap throughout, this | |
75acb9cd | 103 | * restriction might be able to be lifted. |
9db5579b | 104 | */ |
75acb9cd | 105 | gfp_flags = GFP_NOIO | __GFP_ZERO; |
a7a97fc9 | 106 | #ifndef CONFIG_BLK_DEV_RAM_DAX |
75acb9cd NP |
107 | gfp_flags |= __GFP_HIGHMEM; |
108 | #endif | |
26defe34 | 109 | page = alloc_page(gfp_flags); |
9db5579b NP |
110 | if (!page) |
111 | return NULL; | |
112 | ||
113 | if (radix_tree_preload(GFP_NOIO)) { | |
114 | __free_page(page); | |
115 | return NULL; | |
116 | } | |
117 | ||
118 | spin_lock(&brd->brd_lock); | |
119 | idx = sector >> PAGE_SECTORS_SHIFT; | |
dfd20b2b | 120 | page->index = idx; |
9db5579b NP |
121 | if (radix_tree_insert(&brd->brd_pages, idx, page)) { |
122 | __free_page(page); | |
123 | page = radix_tree_lookup(&brd->brd_pages, idx); | |
124 | BUG_ON(!page); | |
125 | BUG_ON(page->index != idx); | |
dfd20b2b | 126 | } |
9db5579b NP |
127 | spin_unlock(&brd->brd_lock); |
128 | ||
129 | radix_tree_preload_end(); | |
130 | ||
131 | return page; | |
132 | } | |
133 | ||
b7c33571 NP |
134 | static void brd_free_page(struct brd_device *brd, sector_t sector) |
135 | { | |
136 | struct page *page; | |
137 | pgoff_t idx; | |
138 | ||
139 | spin_lock(&brd->brd_lock); | |
140 | idx = sector >> PAGE_SECTORS_SHIFT; | |
141 | page = radix_tree_delete(&brd->brd_pages, idx); | |
142 | spin_unlock(&brd->brd_lock); | |
143 | if (page) | |
144 | __free_page(page); | |
145 | } | |
146 | ||
147 | static void brd_zero_page(struct brd_device *brd, sector_t sector) | |
148 | { | |
149 | struct page *page; | |
150 | ||
151 | page = brd_lookup_page(brd, sector); | |
152 | if (page) | |
153 | clear_highpage(page); | |
154 | } | |
155 | ||
9db5579b NP |
156 | /* |
157 | * Free all backing store pages and radix tree. This must only be called when | |
158 | * there are no other users of the device. | |
159 | */ | |
160 | #define FREE_BATCH 16 | |
161 | static void brd_free_pages(struct brd_device *brd) | |
162 | { | |
163 | unsigned long pos = 0; | |
164 | struct page *pages[FREE_BATCH]; | |
165 | int nr_pages; | |
166 | ||
167 | do { | |
168 | int i; | |
169 | ||
170 | nr_pages = radix_tree_gang_lookup(&brd->brd_pages, | |
171 | (void **)pages, pos, FREE_BATCH); | |
172 | ||
173 | for (i = 0; i < nr_pages; i++) { | |
174 | void *ret; | |
175 | ||
176 | BUG_ON(pages[i]->index < pos); | |
177 | pos = pages[i]->index; | |
178 | ret = radix_tree_delete(&brd->brd_pages, pos); | |
179 | BUG_ON(!ret || ret != pages[i]); | |
180 | __free_page(pages[i]); | |
181 | } | |
182 | ||
183 | pos++; | |
184 | ||
185 | /* | |
186 | * This assumes radix_tree_gang_lookup always returns as | |
187 | * many pages as possible. If the radix-tree code changes, | |
188 | * so will this have to. | |
189 | */ | |
190 | } while (nr_pages == FREE_BATCH); | |
191 | } | |
192 | ||
193 | /* | |
194 | * copy_to_brd_setup must be called before copy_to_brd. It may sleep. | |
195 | */ | |
196 | static int copy_to_brd_setup(struct brd_device *brd, sector_t sector, size_t n) | |
197 | { | |
198 | unsigned int offset = (sector & (PAGE_SECTORS-1)) << SECTOR_SHIFT; | |
199 | size_t copy; | |
200 | ||
201 | copy = min_t(size_t, n, PAGE_SIZE - offset); | |
202 | if (!brd_insert_page(brd, sector)) | |
96f8d8e0 | 203 | return -ENOSPC; |
9db5579b NP |
204 | if (copy < n) { |
205 | sector += copy >> SECTOR_SHIFT; | |
206 | if (!brd_insert_page(brd, sector)) | |
96f8d8e0 | 207 | return -ENOSPC; |
9db5579b NP |
208 | } |
209 | return 0; | |
210 | } | |
211 | ||
b7c33571 NP |
212 | static void discard_from_brd(struct brd_device *brd, |
213 | sector_t sector, size_t n) | |
214 | { | |
215 | while (n >= PAGE_SIZE) { | |
216 | /* | |
217 | * Don't want to actually discard pages here because | |
218 | * re-allocating the pages can result in writeback | |
219 | * deadlocks under heavy load. | |
220 | */ | |
221 | if (0) | |
222 | brd_free_page(brd, sector); | |
223 | else | |
224 | brd_zero_page(brd, sector); | |
225 | sector += PAGE_SIZE >> SECTOR_SHIFT; | |
226 | n -= PAGE_SIZE; | |
227 | } | |
228 | } | |
229 | ||
9db5579b NP |
230 | /* |
231 | * Copy n bytes from src to the brd starting at sector. Does not sleep. | |
232 | */ | |
233 | static void copy_to_brd(struct brd_device *brd, const void *src, | |
234 | sector_t sector, size_t n) | |
235 | { | |
236 | struct page *page; | |
237 | void *dst; | |
238 | unsigned int offset = (sector & (PAGE_SECTORS-1)) << SECTOR_SHIFT; | |
239 | size_t copy; | |
240 | ||
241 | copy = min_t(size_t, n, PAGE_SIZE - offset); | |
242 | page = brd_lookup_page(brd, sector); | |
243 | BUG_ON(!page); | |
244 | ||
cfd8005c | 245 | dst = kmap_atomic(page); |
9db5579b | 246 | memcpy(dst + offset, src, copy); |
cfd8005c | 247 | kunmap_atomic(dst); |
9db5579b NP |
248 | |
249 | if (copy < n) { | |
250 | src += copy; | |
251 | sector += copy >> SECTOR_SHIFT; | |
252 | copy = n - copy; | |
253 | page = brd_lookup_page(brd, sector); | |
254 | BUG_ON(!page); | |
255 | ||
cfd8005c | 256 | dst = kmap_atomic(page); |
9db5579b | 257 | memcpy(dst, src, copy); |
cfd8005c | 258 | kunmap_atomic(dst); |
9db5579b NP |
259 | } |
260 | } | |
261 | ||
262 | /* | |
263 | * Copy n bytes to dst from the brd starting at sector. Does not sleep. | |
264 | */ | |
265 | static void copy_from_brd(void *dst, struct brd_device *brd, | |
266 | sector_t sector, size_t n) | |
267 | { | |
268 | struct page *page; | |
269 | void *src; | |
270 | unsigned int offset = (sector & (PAGE_SECTORS-1)) << SECTOR_SHIFT; | |
271 | size_t copy; | |
272 | ||
273 | copy = min_t(size_t, n, PAGE_SIZE - offset); | |
274 | page = brd_lookup_page(brd, sector); | |
275 | if (page) { | |
cfd8005c | 276 | src = kmap_atomic(page); |
9db5579b | 277 | memcpy(dst, src + offset, copy); |
cfd8005c | 278 | kunmap_atomic(src); |
9db5579b NP |
279 | } else |
280 | memset(dst, 0, copy); | |
281 | ||
282 | if (copy < n) { | |
283 | dst += copy; | |
284 | sector += copy >> SECTOR_SHIFT; | |
285 | copy = n - copy; | |
286 | page = brd_lookup_page(brd, sector); | |
287 | if (page) { | |
cfd8005c | 288 | src = kmap_atomic(page); |
9db5579b | 289 | memcpy(dst, src, copy); |
cfd8005c | 290 | kunmap_atomic(src); |
9db5579b NP |
291 | } else |
292 | memset(dst, 0, copy); | |
293 | } | |
294 | } | |
295 | ||
296 | /* | |
297 | * Process a single bvec of a bio. | |
298 | */ | |
299 | static int brd_do_bvec(struct brd_device *brd, struct page *page, | |
300 | unsigned int len, unsigned int off, int rw, | |
301 | sector_t sector) | |
302 | { | |
303 | void *mem; | |
304 | int err = 0; | |
305 | ||
306 | if (rw != READ) { | |
307 | err = copy_to_brd_setup(brd, sector, len); | |
308 | if (err) | |
309 | goto out; | |
310 | } | |
311 | ||
cfd8005c | 312 | mem = kmap_atomic(page); |
9db5579b NP |
313 | if (rw == READ) { |
314 | copy_from_brd(mem + off, brd, sector, len); | |
315 | flush_dcache_page(page); | |
c2572f2b NP |
316 | } else { |
317 | flush_dcache_page(page); | |
9db5579b | 318 | copy_to_brd(brd, mem + off, sector, len); |
c2572f2b | 319 | } |
cfd8005c | 320 | kunmap_atomic(mem); |
9db5579b NP |
321 | |
322 | out: | |
323 | return err; | |
324 | } | |
325 | ||
5a7bbad2 | 326 | static void brd_make_request(struct request_queue *q, struct bio *bio) |
9db5579b NP |
327 | { |
328 | struct block_device *bdev = bio->bi_bdev; | |
329 | struct brd_device *brd = bdev->bd_disk->private_data; | |
330 | int rw; | |
7988613b | 331 | struct bio_vec bvec; |
9db5579b | 332 | sector_t sector; |
7988613b | 333 | struct bvec_iter iter; |
9db5579b | 334 | |
4f024f37 | 335 | sector = bio->bi_iter.bi_sector; |
f73a1c7d | 336 | if (bio_end_sector(bio) > get_capacity(bdev->bd_disk)) |
4246a0b6 | 337 | goto io_error; |
9db5579b | 338 | |
7b6d91da | 339 | if (unlikely(bio->bi_rw & REQ_DISCARD)) { |
4f024f37 | 340 | discard_from_brd(brd, sector, bio->bi_iter.bi_size); |
b7c33571 NP |
341 | goto out; |
342 | } | |
343 | ||
9db5579b NP |
344 | rw = bio_rw(bio); |
345 | if (rw == READA) | |
346 | rw = READ; | |
347 | ||
7988613b KO |
348 | bio_for_each_segment(bvec, bio, iter) { |
349 | unsigned int len = bvec.bv_len; | |
4246a0b6 CH |
350 | int err; |
351 | ||
7988613b KO |
352 | err = brd_do_bvec(brd, bvec.bv_page, len, |
353 | bvec.bv_offset, rw, sector); | |
9db5579b | 354 | if (err) |
4246a0b6 | 355 | goto io_error; |
9db5579b NP |
356 | sector += len >> SECTOR_SHIFT; |
357 | } | |
358 | ||
359 | out: | |
4246a0b6 CH |
360 | bio_endio(bio); |
361 | return; | |
362 | io_error: | |
363 | bio_io_error(bio); | |
9db5579b NP |
364 | } |
365 | ||
a72132c3 MW |
366 | static int brd_rw_page(struct block_device *bdev, sector_t sector, |
367 | struct page *page, int rw) | |
368 | { | |
369 | struct brd_device *brd = bdev->bd_disk->private_data; | |
370 | int err = brd_do_bvec(brd, page, PAGE_CACHE_SIZE, 0, rw, sector); | |
371 | page_endio(page, rw & WRITE, err); | |
372 | return err; | |
373 | } | |
374 | ||
a7a97fc9 | 375 | #ifdef CONFIG_BLK_DEV_RAM_DAX |
dd22f551 MW |
376 | static long brd_direct_access(struct block_device *bdev, sector_t sector, |
377 | void **kaddr, unsigned long *pfn, long size) | |
75acb9cd NP |
378 | { |
379 | struct brd_device *brd = bdev->bd_disk->private_data; | |
380 | struct page *page; | |
381 | ||
382 | if (!brd) | |
383 | return -ENODEV; | |
75acb9cd NP |
384 | page = brd_insert_page(brd, sector); |
385 | if (!page) | |
96f8d8e0 | 386 | return -ENOSPC; |
30afcb4b JH |
387 | *kaddr = page_address(page); |
388 | *pfn = page_to_pfn(page); | |
75acb9cd | 389 | |
dd22f551 MW |
390 | /* |
391 | * TODO: If size > PAGE_SIZE, we could look to see if the next page in | |
392 | * the file happens to be mapped to the next page of physical RAM. | |
393 | */ | |
394 | return PAGE_SIZE; | |
75acb9cd | 395 | } |
a7a97fc9 MW |
396 | #else |
397 | #define brd_direct_access NULL | |
75acb9cd NP |
398 | #endif |
399 | ||
2b9ecd03 | 400 | static int brd_ioctl(struct block_device *bdev, fmode_t mode, |
9db5579b NP |
401 | unsigned int cmd, unsigned long arg) |
402 | { | |
403 | int error; | |
9db5579b NP |
404 | struct brd_device *brd = bdev->bd_disk->private_data; |
405 | ||
406 | if (cmd != BLKFLSBUF) | |
407 | return -ENOTTY; | |
408 | ||
409 | /* | |
410 | * ram device BLKFLSBUF has special semantics, we want to actually | |
411 | * release and destroy the ramdisk data. | |
412 | */ | |
2a48fc0a | 413 | mutex_lock(&brd_mutex); |
9db5579b NP |
414 | mutex_lock(&bdev->bd_mutex); |
415 | error = -EBUSY; | |
416 | if (bdev->bd_openers <= 1) { | |
417 | /* | |
ff01bb48 AV |
418 | * Kill the cache first, so it isn't written back to the |
419 | * device. | |
9db5579b NP |
420 | * |
421 | * Another thread might instantiate more buffercache here, | |
422 | * but there is not much we can do to close that race. | |
423 | */ | |
ff01bb48 | 424 | kill_bdev(bdev); |
9db5579b NP |
425 | brd_free_pages(brd); |
426 | error = 0; | |
427 | } | |
428 | mutex_unlock(&bdev->bd_mutex); | |
2a48fc0a | 429 | mutex_unlock(&brd_mutex); |
9db5579b NP |
430 | |
431 | return error; | |
432 | } | |
433 | ||
83d5cde4 | 434 | static const struct block_device_operations brd_fops = { |
75acb9cd | 435 | .owner = THIS_MODULE, |
a72132c3 | 436 | .rw_page = brd_rw_page, |
8a6cfeb6 | 437 | .ioctl = brd_ioctl, |
75acb9cd | 438 | .direct_access = brd_direct_access, |
9db5579b NP |
439 | }; |
440 | ||
441 | /* | |
442 | * And now the modules code and kernel interface. | |
443 | */ | |
937af5ec | 444 | static int rd_nr = CONFIG_BLK_DEV_RAM_COUNT; |
8892cbaf | 445 | module_param(rd_nr, int, S_IRUGO); |
9db5579b | 446 | MODULE_PARM_DESC(rd_nr, "Maximum number of brd devices"); |
937af5ec BH |
447 | |
448 | int rd_size = CONFIG_BLK_DEV_RAM_SIZE; | |
8892cbaf | 449 | module_param(rd_size, int, S_IRUGO); |
9db5579b | 450 | MODULE_PARM_DESC(rd_size, "Size of each RAM disk in kbytes."); |
937af5ec BH |
451 | |
452 | static int max_part = 1; | |
8892cbaf | 453 | module_param(max_part, int, S_IRUGO); |
937af5ec BH |
454 | MODULE_PARM_DESC(max_part, "Num Minors to reserve between devices"); |
455 | ||
9db5579b NP |
456 | MODULE_LICENSE("GPL"); |
457 | MODULE_ALIAS_BLOCKDEV_MAJOR(RAMDISK_MAJOR); | |
efedf51c | 458 | MODULE_ALIAS("rd"); |
9db5579b NP |
459 | |
460 | #ifndef MODULE | |
461 | /* Legacy boot options - nonmodular */ | |
462 | static int __init ramdisk_size(char *str) | |
463 | { | |
464 | rd_size = simple_strtol(str, NULL, 0); | |
465 | return 1; | |
466 | } | |
1adbee50 | 467 | __setup("ramdisk_size=", ramdisk_size); |
9db5579b NP |
468 | #endif |
469 | ||
470 | /* | |
471 | * The device scheme is derived from loop.c. Keep them in synch where possible | |
472 | * (should share code eventually). | |
473 | */ | |
474 | static LIST_HEAD(brd_devices); | |
475 | static DEFINE_MUTEX(brd_devices_mutex); | |
476 | ||
477 | static struct brd_device *brd_alloc(int i) | |
478 | { | |
479 | struct brd_device *brd; | |
480 | struct gendisk *disk; | |
481 | ||
482 | brd = kzalloc(sizeof(*brd), GFP_KERNEL); | |
483 | if (!brd) | |
484 | goto out; | |
485 | brd->brd_number = i; | |
486 | spin_lock_init(&brd->brd_lock); | |
487 | INIT_RADIX_TREE(&brd->brd_pages, GFP_ATOMIC); | |
488 | ||
489 | brd->brd_queue = blk_alloc_queue(GFP_KERNEL); | |
490 | if (!brd->brd_queue) | |
491 | goto out_free_dev; | |
c8fa3173 | 492 | |
9db5579b | 493 | blk_queue_make_request(brd->brd_queue, brd_make_request); |
086fa5ff | 494 | blk_queue_max_hw_sectors(brd->brd_queue, 1024); |
9db5579b NP |
495 | blk_queue_bounce_limit(brd->brd_queue, BLK_BOUNCE_ANY); |
496 | ||
c8fa3173 BH |
497 | /* This is so fdisk will align partitions on 4k, because of |
498 | * direct_access API needing 4k alignment, returning a PFN | |
499 | * (This is only a problem on very small devices <= 4M, | |
500 | * otherwise fdisk will align on 1M. Regardless this call | |
501 | * is harmless) | |
502 | */ | |
503 | blk_queue_physical_block_size(brd->brd_queue, PAGE_SIZE); | |
504 | ||
b7c33571 | 505 | brd->brd_queue->limits.discard_granularity = PAGE_SIZE; |
2bb4cd5c | 506 | blk_queue_max_discard_sectors(brd->brd_queue, UINT_MAX); |
b7c33571 NP |
507 | brd->brd_queue->limits.discard_zeroes_data = 1; |
508 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, brd->brd_queue); | |
509 | ||
937af5ec | 510 | disk = brd->brd_disk = alloc_disk(max_part); |
9db5579b NP |
511 | if (!disk) |
512 | goto out_free_queue; | |
513 | disk->major = RAMDISK_MAJOR; | |
937af5ec | 514 | disk->first_minor = i * max_part; |
9db5579b NP |
515 | disk->fops = &brd_fops; |
516 | disk->private_data = brd; | |
517 | disk->queue = brd->brd_queue; | |
937af5ec | 518 | disk->flags = GENHD_FL_EXT_DEVT; |
9db5579b NP |
519 | sprintf(disk->disk_name, "ram%d", i); |
520 | set_capacity(disk, rd_size * 2); | |
521 | ||
522 | return brd; | |
523 | ||
524 | out_free_queue: | |
525 | blk_cleanup_queue(brd->brd_queue); | |
526 | out_free_dev: | |
527 | kfree(brd); | |
528 | out: | |
529 | return NULL; | |
530 | } | |
531 | ||
532 | static void brd_free(struct brd_device *brd) | |
533 | { | |
534 | put_disk(brd->brd_disk); | |
535 | blk_cleanup_queue(brd->brd_queue); | |
536 | brd_free_pages(brd); | |
537 | kfree(brd); | |
538 | } | |
539 | ||
937af5ec | 540 | static struct brd_device *brd_init_one(int i, bool *new) |
9db5579b NP |
541 | { |
542 | struct brd_device *brd; | |
543 | ||
937af5ec | 544 | *new = false; |
9db5579b NP |
545 | list_for_each_entry(brd, &brd_devices, brd_list) { |
546 | if (brd->brd_number == i) | |
547 | goto out; | |
548 | } | |
549 | ||
550 | brd = brd_alloc(i); | |
551 | if (brd) { | |
552 | add_disk(brd->brd_disk); | |
553 | list_add_tail(&brd->brd_list, &brd_devices); | |
554 | } | |
937af5ec | 555 | *new = true; |
9db5579b NP |
556 | out: |
557 | return brd; | |
558 | } | |
559 | ||
560 | static void brd_del_one(struct brd_device *brd) | |
561 | { | |
562 | list_del(&brd->brd_list); | |
563 | del_gendisk(brd->brd_disk); | |
564 | brd_free(brd); | |
565 | } | |
566 | ||
567 | static struct kobject *brd_probe(dev_t dev, int *part, void *data) | |
568 | { | |
569 | struct brd_device *brd; | |
570 | struct kobject *kobj; | |
937af5ec | 571 | bool new; |
9db5579b NP |
572 | |
573 | mutex_lock(&brd_devices_mutex); | |
937af5ec | 574 | brd = brd_init_one(MINOR(dev) / max_part, &new); |
a207f593 | 575 | kobj = brd ? get_disk(brd->brd_disk) : NULL; |
9db5579b NP |
576 | mutex_unlock(&brd_devices_mutex); |
577 | ||
937af5ec BH |
578 | if (new) |
579 | *part = 0; | |
580 | ||
9db5579b NP |
581 | return kobj; |
582 | } | |
583 | ||
584 | static int __init brd_init(void) | |
585 | { | |
9db5579b | 586 | struct brd_device *brd, *next; |
937af5ec | 587 | int i; |
9db5579b NP |
588 | |
589 | /* | |
590 | * brd module now has a feature to instantiate underlying device | |
591 | * structure on-demand, provided that there is an access dev node. | |
9db5579b | 592 | * |
937af5ec BH |
593 | * (1) if rd_nr is specified, create that many upfront. else |
594 | * it defaults to CONFIG_BLK_DEV_RAM_COUNT | |
595 | * (2) User can further extend brd devices by create dev node themselves | |
596 | * and have kernel automatically instantiate actual device | |
597 | * on-demand. Example: | |
598 | * mknod /path/devnod_name b 1 X # 1 is the rd major | |
599 | * fdisk -l /path/devnod_name | |
600 | * If (X / max_part) was not already created it will be created | |
601 | * dynamically. | |
9db5579b | 602 | */ |
d7853d1f | 603 | |
9db5579b NP |
604 | if (register_blkdev(RAMDISK_MAJOR, "ramdisk")) |
605 | return -EIO; | |
606 | ||
937af5ec BH |
607 | if (unlikely(!max_part)) |
608 | max_part = 1; | |
609 | ||
610 | for (i = 0; i < rd_nr; i++) { | |
9db5579b NP |
611 | brd = brd_alloc(i); |
612 | if (!brd) | |
613 | goto out_free; | |
614 | list_add_tail(&brd->brd_list, &brd_devices); | |
615 | } | |
616 | ||
617 | /* point of no return */ | |
618 | ||
619 | list_for_each_entry(brd, &brd_devices, brd_list) | |
620 | add_disk(brd->brd_disk); | |
621 | ||
937af5ec | 622 | blk_register_region(MKDEV(RAMDISK_MAJOR, 0), 1UL << MINORBITS, |
9db5579b NP |
623 | THIS_MODULE, brd_probe, NULL, NULL); |
624 | ||
937af5ec | 625 | pr_info("brd: module loaded\n"); |
9db5579b NP |
626 | return 0; |
627 | ||
628 | out_free: | |
629 | list_for_each_entry_safe(brd, next, &brd_devices, brd_list) { | |
630 | list_del(&brd->brd_list); | |
631 | brd_free(brd); | |
632 | } | |
c82f2966 | 633 | unregister_blkdev(RAMDISK_MAJOR, "ramdisk"); |
9db5579b | 634 | |
937af5ec | 635 | pr_info("brd: module NOT loaded !!!\n"); |
9db5579b NP |
636 | return -ENOMEM; |
637 | } | |
638 | ||
639 | static void __exit brd_exit(void) | |
640 | { | |
9db5579b NP |
641 | struct brd_device *brd, *next; |
642 | ||
9db5579b NP |
643 | list_for_each_entry_safe(brd, next, &brd_devices, brd_list) |
644 | brd_del_one(brd); | |
645 | ||
937af5ec | 646 | blk_unregister_region(MKDEV(RAMDISK_MAJOR, 0), 1UL << MINORBITS); |
9db5579b | 647 | unregister_blkdev(RAMDISK_MAJOR, "ramdisk"); |
937af5ec BH |
648 | |
649 | pr_info("brd: module unloaded\n"); | |
9db5579b NP |
650 | } |
651 | ||
652 | module_init(brd_init); | |
653 | module_exit(brd_exit); | |
654 |