Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/drivers/block/loop.c | |
3 | * | |
4 | * Written by Theodore Ts'o, 3/29/93 | |
5 | * | |
6 | * Copyright 1993 by Theodore Ts'o. Redistribution of this file is | |
7 | * permitted under the GNU General Public License. | |
8 | * | |
9 | * DES encryption plus some minor changes by Werner Almesberger, 30-MAY-1993 | |
10 | * more DES encryption plus IDEA encryption by Nicholas J. Leon, June 20, 1996 | |
11 | * | |
12 | * Modularized and updated for 1.1.16 kernel - Mitch Dsouza 28th May 1994 | |
13 | * Adapted for 1.3.59 kernel - Andries Brouwer, 1 Feb 1996 | |
14 | * | |
15 | * Fixed do_loop_request() re-entrancy - Vincent.Renardias@waw.com Mar 20, 1997 | |
16 | * | |
17 | * Added devfs support - Richard Gooch <rgooch@atnf.csiro.au> 16-Jan-1998 | |
18 | * | |
19 | * Handle sparse backing files correctly - Kenn Humborg, Jun 28, 1998 | |
20 | * | |
21 | * Loadable modules and other fixes by AK, 1998 | |
22 | * | |
23 | * Make real block number available to downstream transfer functions, enables | |
24 | * CBC (and relatives) mode encryption requiring unique IVs per data block. | |
25 | * Reed H. Petty, rhp@draper.net | |
26 | * | |
27 | * Maximum number of loop devices now dynamic via max_loop module parameter. | |
28 | * Russell Kroll <rkroll@exploits.org> 19990701 | |
29 | * | |
30 | * Maximum number of loop devices when compiled-in now selectable by passing | |
31 | * max_loop=<1-255> to the kernel on boot. | |
96de0e25 | 32 | * Erik I. Bolsø, <eriki@himolde.no>, Oct 31, 1999 |
1da177e4 LT |
33 | * |
34 | * Completely rewrite request handling to be make_request_fn style and | |
35 | * non blocking, pushing work to a helper thread. Lots of fixes from | |
36 | * Al Viro too. | |
37 | * Jens Axboe <axboe@suse.de>, Nov 2000 | |
38 | * | |
39 | * Support up to 256 loop devices | |
40 | * Heinz Mauelshagen <mge@sistina.com>, Feb 2002 | |
41 | * | |
42 | * Support for falling back on the write file operation when the address space | |
4e02ed4b | 43 | * operations write_begin is not available on the backing filesystem. |
1da177e4 LT |
44 | * Anton Altaparmakov, 16 Feb 2005 |
45 | * | |
46 | * Still To Fix: | |
47 | * - Advisory locking is ignored here. | |
48 | * - Should use an own CAP_* category instead of CAP_SYS_ADMIN | |
49 | * | |
50 | */ | |
51 | ||
1da177e4 LT |
52 | #include <linux/module.h> |
53 | #include <linux/moduleparam.h> | |
54 | #include <linux/sched.h> | |
55 | #include <linux/fs.h> | |
56 | #include <linux/file.h> | |
57 | #include <linux/stat.h> | |
58 | #include <linux/errno.h> | |
59 | #include <linux/major.h> | |
60 | #include <linux/wait.h> | |
61 | #include <linux/blkdev.h> | |
62 | #include <linux/blkpg.h> | |
63 | #include <linux/init.h> | |
1da177e4 LT |
64 | #include <linux/swap.h> |
65 | #include <linux/slab.h> | |
66 | #include <linux/loop.h> | |
863d5b82 | 67 | #include <linux/compat.h> |
1da177e4 | 68 | #include <linux/suspend.h> |
83144186 | 69 | #include <linux/freezer.h> |
2a48fc0a | 70 | #include <linux/mutex.h> |
1da177e4 LT |
71 | #include <linux/writeback.h> |
72 | #include <linux/buffer_head.h> /* for invalidate_bdev() */ | |
73 | #include <linux/completion.h> | |
74 | #include <linux/highmem.h> | |
6c997918 | 75 | #include <linux/kthread.h> |
d6b29d7c | 76 | #include <linux/splice.h> |
ee862730 | 77 | #include <linux/sysfs.h> |
770fe30a | 78 | #include <linux/miscdevice.h> |
1da177e4 LT |
79 | #include <asm/uaccess.h> |
80 | ||
34dd82af KS |
81 | static DEFINE_IDR(loop_index_idr); |
82 | static DEFINE_MUTEX(loop_index_mutex); | |
1da177e4 | 83 | |
476a4813 LV |
84 | static int max_part; |
85 | static int part_shift; | |
86 | ||
1da177e4 LT |
87 | /* |
88 | * Transfer functions | |
89 | */ | |
90 | static int transfer_none(struct loop_device *lo, int cmd, | |
91 | struct page *raw_page, unsigned raw_off, | |
92 | struct page *loop_page, unsigned loop_off, | |
93 | int size, sector_t real_block) | |
94 | { | |
95 | char *raw_buf = kmap_atomic(raw_page, KM_USER0) + raw_off; | |
96 | char *loop_buf = kmap_atomic(loop_page, KM_USER1) + loop_off; | |
97 | ||
98 | if (cmd == READ) | |
99 | memcpy(loop_buf, raw_buf, size); | |
100 | else | |
101 | memcpy(raw_buf, loop_buf, size); | |
102 | ||
1da177e4 | 103 | kunmap_atomic(loop_buf, KM_USER1); |
61ecdb80 | 104 | kunmap_atomic(raw_buf, KM_USER0); |
1da177e4 LT |
105 | cond_resched(); |
106 | return 0; | |
107 | } | |
108 | ||
109 | static int transfer_xor(struct loop_device *lo, int cmd, | |
110 | struct page *raw_page, unsigned raw_off, | |
111 | struct page *loop_page, unsigned loop_off, | |
112 | int size, sector_t real_block) | |
113 | { | |
114 | char *raw_buf = kmap_atomic(raw_page, KM_USER0) + raw_off; | |
115 | char *loop_buf = kmap_atomic(loop_page, KM_USER1) + loop_off; | |
116 | char *in, *out, *key; | |
117 | int i, keysize; | |
118 | ||
119 | if (cmd == READ) { | |
120 | in = raw_buf; | |
121 | out = loop_buf; | |
122 | } else { | |
123 | in = loop_buf; | |
124 | out = raw_buf; | |
125 | } | |
126 | ||
127 | key = lo->lo_encrypt_key; | |
128 | keysize = lo->lo_encrypt_key_size; | |
129 | for (i = 0; i < size; i++) | |
130 | *out++ = *in++ ^ key[(i & 511) % keysize]; | |
131 | ||
1da177e4 | 132 | kunmap_atomic(loop_buf, KM_USER1); |
61ecdb80 | 133 | kunmap_atomic(raw_buf, KM_USER0); |
1da177e4 LT |
134 | cond_resched(); |
135 | return 0; | |
136 | } | |
137 | ||
138 | static int xor_init(struct loop_device *lo, const struct loop_info64 *info) | |
139 | { | |
140 | if (unlikely(info->lo_encrypt_key_size <= 0)) | |
141 | return -EINVAL; | |
142 | return 0; | |
143 | } | |
144 | ||
145 | static struct loop_func_table none_funcs = { | |
146 | .number = LO_CRYPT_NONE, | |
147 | .transfer = transfer_none, | |
148 | }; | |
149 | ||
150 | static struct loop_func_table xor_funcs = { | |
151 | .number = LO_CRYPT_XOR, | |
152 | .transfer = transfer_xor, | |
153 | .init = xor_init | |
154 | }; | |
155 | ||
156 | /* xfer_funcs[0] is special - its release function is never called */ | |
157 | static struct loop_func_table *xfer_funcs[MAX_LO_CRYPT] = { | |
158 | &none_funcs, | |
159 | &xor_funcs | |
160 | }; | |
161 | ||
162 | static loff_t get_loop_size(struct loop_device *lo, struct file *file) | |
163 | { | |
164 | loff_t size, offset, loopsize; | |
165 | ||
166 | /* Compute loopsize in bytes */ | |
167 | size = i_size_read(file->f_mapping->host); | |
168 | offset = lo->lo_offset; | |
169 | loopsize = size - offset; | |
170 | if (lo->lo_sizelimit > 0 && lo->lo_sizelimit < loopsize) | |
171 | loopsize = lo->lo_sizelimit; | |
172 | ||
173 | /* | |
174 | * Unfortunately, if we want to do I/O on the device, | |
175 | * the number of 512-byte sectors has to fit into a sector_t. | |
176 | */ | |
177 | return loopsize >> 9; | |
178 | } | |
179 | ||
180 | static int | |
181 | figure_loop_size(struct loop_device *lo) | |
182 | { | |
183 | loff_t size = get_loop_size(lo, lo->lo_backing_file); | |
184 | sector_t x = (sector_t)size; | |
185 | ||
186 | if (unlikely((loff_t)x != size)) | |
187 | return -EFBIG; | |
188 | ||
73285082 | 189 | set_capacity(lo->lo_disk, x); |
1da177e4 LT |
190 | return 0; |
191 | } | |
192 | ||
193 | static inline int | |
194 | lo_do_transfer(struct loop_device *lo, int cmd, | |
195 | struct page *rpage, unsigned roffs, | |
196 | struct page *lpage, unsigned loffs, | |
197 | int size, sector_t rblock) | |
198 | { | |
199 | if (unlikely(!lo->transfer)) | |
200 | return 0; | |
201 | ||
202 | return lo->transfer(lo, cmd, rpage, roffs, lpage, loffs, size, rblock); | |
203 | } | |
204 | ||
205 | /** | |
206 | * do_lo_send_aops - helper for writing data to a loop device | |
207 | * | |
208 | * This is the fast version for backing filesystems which implement the address | |
afddba49 | 209 | * space operations write_begin and write_end. |
1da177e4 LT |
210 | */ |
211 | static int do_lo_send_aops(struct loop_device *lo, struct bio_vec *bvec, | |
511de73f | 212 | loff_t pos, struct page *unused) |
1da177e4 LT |
213 | { |
214 | struct file *file = lo->lo_backing_file; /* kudos to NFsckingS */ | |
215 | struct address_space *mapping = file->f_mapping; | |
1da177e4 LT |
216 | pgoff_t index; |
217 | unsigned offset, bv_offs; | |
994fc28c | 218 | int len, ret; |
1da177e4 | 219 | |
1b1dcc1b | 220 | mutex_lock(&mapping->host->i_mutex); |
1da177e4 LT |
221 | index = pos >> PAGE_CACHE_SHIFT; |
222 | offset = pos & ((pgoff_t)PAGE_CACHE_SIZE - 1); | |
223 | bv_offs = bvec->bv_offset; | |
224 | len = bvec->bv_len; | |
225 | while (len > 0) { | |
226 | sector_t IV; | |
afddba49 | 227 | unsigned size, copied; |
1da177e4 | 228 | int transfer_result; |
afddba49 NP |
229 | struct page *page; |
230 | void *fsdata; | |
1da177e4 LT |
231 | |
232 | IV = ((sector_t)index << (PAGE_CACHE_SHIFT - 9))+(offset >> 9); | |
233 | size = PAGE_CACHE_SIZE - offset; | |
234 | if (size > len) | |
235 | size = len; | |
afddba49 NP |
236 | |
237 | ret = pagecache_write_begin(file, mapping, pos, size, 0, | |
238 | &page, &fsdata); | |
239 | if (ret) | |
1da177e4 | 240 | goto fail; |
afddba49 | 241 | |
02246c41 NK |
242 | file_update_time(file); |
243 | ||
1da177e4 LT |
244 | transfer_result = lo_do_transfer(lo, WRITE, page, offset, |
245 | bvec->bv_page, bv_offs, size, IV); | |
afddba49 | 246 | copied = size; |
1da177e4 | 247 | if (unlikely(transfer_result)) |
afddba49 NP |
248 | copied = 0; |
249 | ||
250 | ret = pagecache_write_end(file, mapping, pos, size, copied, | |
251 | page, fsdata); | |
8268f5a7 | 252 | if (ret < 0 || ret != copied) |
afddba49 | 253 | goto fail; |
afddba49 NP |
254 | |
255 | if (unlikely(transfer_result)) | |
256 | goto fail; | |
257 | ||
258 | bv_offs += copied; | |
259 | len -= copied; | |
1da177e4 LT |
260 | offset = 0; |
261 | index++; | |
afddba49 | 262 | pos += copied; |
1da177e4 | 263 | } |
994fc28c | 264 | ret = 0; |
1da177e4 | 265 | out: |
1b1dcc1b | 266 | mutex_unlock(&mapping->host->i_mutex); |
1da177e4 | 267 | return ret; |
1da177e4 LT |
268 | fail: |
269 | ret = -1; | |
270 | goto out; | |
271 | } | |
272 | ||
273 | /** | |
274 | * __do_lo_send_write - helper for writing data to a loop device | |
275 | * | |
276 | * This helper just factors out common code between do_lo_send_direct_write() | |
277 | * and do_lo_send_write(). | |
278 | */ | |
858119e1 | 279 | static int __do_lo_send_write(struct file *file, |
98ae6ccd | 280 | u8 *buf, const int len, loff_t pos) |
1da177e4 LT |
281 | { |
282 | ssize_t bw; | |
283 | mm_segment_t old_fs = get_fs(); | |
284 | ||
285 | set_fs(get_ds()); | |
286 | bw = file->f_op->write(file, buf, len, &pos); | |
287 | set_fs(old_fs); | |
288 | if (likely(bw == len)) | |
289 | return 0; | |
290 | printk(KERN_ERR "loop: Write error at byte offset %llu, length %i.\n", | |
291 | (unsigned long long)pos, len); | |
292 | if (bw >= 0) | |
293 | bw = -EIO; | |
294 | return bw; | |
295 | } | |
296 | ||
297 | /** | |
298 | * do_lo_send_direct_write - helper for writing data to a loop device | |
299 | * | |
300 | * This is the fast, non-transforming version for backing filesystems which do | |
afddba49 | 301 | * not implement the address space operations write_begin and write_end. |
1da177e4 LT |
302 | * It uses the write file operation which should be present on all writeable |
303 | * filesystems. | |
304 | */ | |
305 | static int do_lo_send_direct_write(struct loop_device *lo, | |
511de73f | 306 | struct bio_vec *bvec, loff_t pos, struct page *page) |
1da177e4 LT |
307 | { |
308 | ssize_t bw = __do_lo_send_write(lo->lo_backing_file, | |
98ae6ccd | 309 | kmap(bvec->bv_page) + bvec->bv_offset, |
1da177e4 LT |
310 | bvec->bv_len, pos); |
311 | kunmap(bvec->bv_page); | |
312 | cond_resched(); | |
313 | return bw; | |
314 | } | |
315 | ||
316 | /** | |
317 | * do_lo_send_write - helper for writing data to a loop device | |
318 | * | |
319 | * This is the slow, transforming version for filesystems which do not | |
afddba49 | 320 | * implement the address space operations write_begin and write_end. It |
1da177e4 LT |
321 | * uses the write file operation which should be present on all writeable |
322 | * filesystems. | |
323 | * | |
324 | * Using fops->write is slower than using aops->{prepare,commit}_write in the | |
325 | * transforming case because we need to double buffer the data as we cannot do | |
326 | * the transformations in place as we do not have direct access to the | |
327 | * destination pages of the backing file. | |
328 | */ | |
329 | static int do_lo_send_write(struct loop_device *lo, struct bio_vec *bvec, | |
511de73f | 330 | loff_t pos, struct page *page) |
1da177e4 LT |
331 | { |
332 | int ret = lo_do_transfer(lo, WRITE, page, 0, bvec->bv_page, | |
333 | bvec->bv_offset, bvec->bv_len, pos >> 9); | |
334 | if (likely(!ret)) | |
335 | return __do_lo_send_write(lo->lo_backing_file, | |
98ae6ccd | 336 | page_address(page), bvec->bv_len, |
1da177e4 LT |
337 | pos); |
338 | printk(KERN_ERR "loop: Transfer error at byte offset %llu, " | |
339 | "length %i.\n", (unsigned long long)pos, bvec->bv_len); | |
340 | if (ret > 0) | |
341 | ret = -EIO; | |
342 | return ret; | |
343 | } | |
344 | ||
511de73f | 345 | static int lo_send(struct loop_device *lo, struct bio *bio, loff_t pos) |
1da177e4 | 346 | { |
511de73f | 347 | int (*do_lo_send)(struct loop_device *, struct bio_vec *, loff_t, |
1da177e4 LT |
348 | struct page *page); |
349 | struct bio_vec *bvec; | |
350 | struct page *page = NULL; | |
351 | int i, ret = 0; | |
352 | ||
353 | do_lo_send = do_lo_send_aops; | |
354 | if (!(lo->lo_flags & LO_FLAGS_USE_AOPS)) { | |
355 | do_lo_send = do_lo_send_direct_write; | |
356 | if (lo->transfer != transfer_none) { | |
357 | page = alloc_page(GFP_NOIO | __GFP_HIGHMEM); | |
358 | if (unlikely(!page)) | |
359 | goto fail; | |
360 | kmap(page); | |
361 | do_lo_send = do_lo_send_write; | |
362 | } | |
363 | } | |
364 | bio_for_each_segment(bvec, bio, i) { | |
511de73f | 365 | ret = do_lo_send(lo, bvec, pos, page); |
1da177e4 LT |
366 | if (ret < 0) |
367 | break; | |
368 | pos += bvec->bv_len; | |
369 | } | |
370 | if (page) { | |
371 | kunmap(page); | |
372 | __free_page(page); | |
373 | } | |
374 | out: | |
375 | return ret; | |
376 | fail: | |
377 | printk(KERN_ERR "loop: Failed to allocate temporary page for write.\n"); | |
378 | ret = -ENOMEM; | |
379 | goto out; | |
380 | } | |
381 | ||
382 | struct lo_read_data { | |
383 | struct loop_device *lo; | |
384 | struct page *page; | |
385 | unsigned offset; | |
386 | int bsize; | |
387 | }; | |
388 | ||
389 | static int | |
fd582140 JA |
390 | lo_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf, |
391 | struct splice_desc *sd) | |
1da177e4 | 392 | { |
fd582140 | 393 | struct lo_read_data *p = sd->u.data; |
1da177e4 | 394 | struct loop_device *lo = p->lo; |
fd582140 | 395 | struct page *page = buf->page; |
1da177e4 | 396 | sector_t IV; |
3603b8ea | 397 | int size; |
1da177e4 | 398 | |
fd582140 JA |
399 | IV = ((sector_t) page->index << (PAGE_CACHE_SHIFT - 9)) + |
400 | (buf->offset >> 9); | |
401 | size = sd->len; | |
402 | if (size > p->bsize) | |
403 | size = p->bsize; | |
1da177e4 | 404 | |
fd582140 | 405 | if (lo_do_transfer(lo, READ, page, buf->offset, p->page, p->offset, size, IV)) { |
1da177e4 LT |
406 | printk(KERN_ERR "loop: transfer error block %ld\n", |
407 | page->index); | |
fd582140 | 408 | size = -EINVAL; |
1da177e4 LT |
409 | } |
410 | ||
411 | flush_dcache_page(p->page); | |
412 | ||
fd582140 JA |
413 | if (size > 0) |
414 | p->offset += size; | |
415 | ||
1da177e4 LT |
416 | return size; |
417 | } | |
418 | ||
fd582140 JA |
419 | static int |
420 | lo_direct_splice_actor(struct pipe_inode_info *pipe, struct splice_desc *sd) | |
421 | { | |
422 | return __splice_from_pipe(pipe, sd, lo_splice_actor); | |
423 | } | |
424 | ||
1da177e4 LT |
425 | static int |
426 | do_lo_receive(struct loop_device *lo, | |
427 | struct bio_vec *bvec, int bsize, loff_t pos) | |
428 | { | |
429 | struct lo_read_data cookie; | |
fd582140 | 430 | struct splice_desc sd; |
1da177e4 | 431 | struct file *file; |
fd582140 | 432 | long retval; |
1da177e4 LT |
433 | |
434 | cookie.lo = lo; | |
435 | cookie.page = bvec->bv_page; | |
436 | cookie.offset = bvec->bv_offset; | |
437 | cookie.bsize = bsize; | |
fd582140 JA |
438 | |
439 | sd.len = 0; | |
440 | sd.total_len = bvec->bv_len; | |
441 | sd.flags = 0; | |
442 | sd.pos = pos; | |
443 | sd.u.data = &cookie; | |
444 | ||
1da177e4 | 445 | file = lo->lo_backing_file; |
fd582140 JA |
446 | retval = splice_direct_to_actor(file, &sd, lo_direct_splice_actor); |
447 | ||
448 | if (retval < 0) | |
449 | return retval; | |
450 | ||
451 | return 0; | |
1da177e4 LT |
452 | } |
453 | ||
454 | static int | |
455 | lo_receive(struct loop_device *lo, struct bio *bio, int bsize, loff_t pos) | |
456 | { | |
457 | struct bio_vec *bvec; | |
458 | int i, ret = 0; | |
459 | ||
460 | bio_for_each_segment(bvec, bio, i) { | |
461 | ret = do_lo_receive(lo, bvec, bsize, pos); | |
462 | if (ret < 0) | |
463 | break; | |
464 | pos += bvec->bv_len; | |
465 | } | |
466 | return ret; | |
467 | } | |
468 | ||
469 | static int do_bio_filebacked(struct loop_device *lo, struct bio *bio) | |
470 | { | |
471 | loff_t pos; | |
472 | int ret; | |
473 | ||
474 | pos = ((loff_t) bio->bi_sector << 9) + lo->lo_offset; | |
68db1961 NK |
475 | |
476 | if (bio_rw(bio) == WRITE) { | |
68db1961 NK |
477 | struct file *file = lo->lo_backing_file; |
478 | ||
6259f284 | 479 | if (bio->bi_rw & REQ_FLUSH) { |
8018ab05 | 480 | ret = vfs_fsync(file, 0); |
6259f284 | 481 | if (unlikely(ret && ret != -EINVAL)) { |
68db1961 NK |
482 | ret = -EIO; |
483 | goto out; | |
484 | } | |
485 | } | |
486 | ||
511de73f | 487 | ret = lo_send(lo, bio, pos); |
68db1961 | 488 | |
6259f284 | 489 | if ((bio->bi_rw & REQ_FUA) && !ret) { |
8018ab05 | 490 | ret = vfs_fsync(file, 0); |
6259f284 | 491 | if (unlikely(ret && ret != -EINVAL)) |
68db1961 NK |
492 | ret = -EIO; |
493 | } | |
494 | } else | |
1da177e4 | 495 | ret = lo_receive(lo, bio, lo->lo_blocksize, pos); |
68db1961 NK |
496 | |
497 | out: | |
1da177e4 LT |
498 | return ret; |
499 | } | |
500 | ||
501 | /* | |
502 | * Add bio to back of pending list | |
503 | */ | |
504 | static void loop_add_bio(struct loop_device *lo, struct bio *bio) | |
505 | { | |
e686307f | 506 | bio_list_add(&lo->lo_bio_list, bio); |
1da177e4 LT |
507 | } |
508 | ||
509 | /* | |
510 | * Grab first pending buffer | |
511 | */ | |
512 | static struct bio *loop_get_bio(struct loop_device *lo) | |
513 | { | |
e686307f | 514 | return bio_list_pop(&lo->lo_bio_list); |
1da177e4 LT |
515 | } |
516 | ||
165125e1 | 517 | static int loop_make_request(struct request_queue *q, struct bio *old_bio) |
1da177e4 LT |
518 | { |
519 | struct loop_device *lo = q->queuedata; | |
520 | int rw = bio_rw(old_bio); | |
521 | ||
35a82d1a NP |
522 | if (rw == READA) |
523 | rw = READ; | |
524 | ||
525 | BUG_ON(!lo || (rw != READ && rw != WRITE)); | |
1da177e4 LT |
526 | |
527 | spin_lock_irq(&lo->lo_lock); | |
528 | if (lo->lo_state != Lo_bound) | |
35a82d1a NP |
529 | goto out; |
530 | if (unlikely(rw == WRITE && (lo->lo_flags & LO_FLAGS_READ_ONLY))) | |
531 | goto out; | |
1da177e4 | 532 | loop_add_bio(lo, old_bio); |
6c997918 | 533 | wake_up(&lo->lo_event); |
35a82d1a | 534 | spin_unlock_irq(&lo->lo_lock); |
1da177e4 | 535 | return 0; |
35a82d1a | 536 | |
1da177e4 | 537 | out: |
35a82d1a | 538 | spin_unlock_irq(&lo->lo_lock); |
6712ecf8 | 539 | bio_io_error(old_bio); |
1da177e4 | 540 | return 0; |
1da177e4 LT |
541 | } |
542 | ||
1da177e4 LT |
543 | struct switch_request { |
544 | struct file *file; | |
545 | struct completion wait; | |
546 | }; | |
547 | ||
548 | static void do_loop_switch(struct loop_device *, struct switch_request *); | |
549 | ||
550 | static inline void loop_handle_bio(struct loop_device *lo, struct bio *bio) | |
551 | { | |
1da177e4 LT |
552 | if (unlikely(!bio->bi_bdev)) { |
553 | do_loop_switch(lo, bio->bi_private); | |
554 | bio_put(bio); | |
555 | } else { | |
35a82d1a | 556 | int ret = do_bio_filebacked(lo, bio); |
6712ecf8 | 557 | bio_endio(bio, ret); |
1da177e4 LT |
558 | } |
559 | } | |
560 | ||
561 | /* | |
562 | * worker thread that handles reads/writes to file backed loop devices, | |
563 | * to avoid blocking in our make_request_fn. it also does loop decrypting | |
564 | * on reads for block backed loop, as that is too heavy to do from | |
565 | * b_end_io context where irqs may be disabled. | |
6c997918 SH |
566 | * |
567 | * Loop explanation: loop_clr_fd() sets lo_state to Lo_rundown before | |
568 | * calling kthread_stop(). Therefore once kthread_should_stop() is | |
569 | * true, make_request will not place any more requests. Therefore | |
570 | * once kthread_should_stop() is true and lo_bio is NULL, we are | |
571 | * done with the loop. | |
1da177e4 LT |
572 | */ |
573 | static int loop_thread(void *data) | |
574 | { | |
575 | struct loop_device *lo = data; | |
576 | struct bio *bio; | |
577 | ||
1da177e4 LT |
578 | set_user_nice(current, -20); |
579 | ||
e686307f | 580 | while (!kthread_should_stop() || !bio_list_empty(&lo->lo_bio_list)) { |
09c0dc68 | 581 | |
6c997918 | 582 | wait_event_interruptible(lo->lo_event, |
e686307f AM |
583 | !bio_list_empty(&lo->lo_bio_list) || |
584 | kthread_should_stop()); | |
35a82d1a | 585 | |
e686307f | 586 | if (bio_list_empty(&lo->lo_bio_list)) |
35a82d1a | 587 | continue; |
35a82d1a | 588 | spin_lock_irq(&lo->lo_lock); |
1da177e4 | 589 | bio = loop_get_bio(lo); |
35a82d1a NP |
590 | spin_unlock_irq(&lo->lo_lock); |
591 | ||
592 | BUG_ON(!bio); | |
1da177e4 | 593 | loop_handle_bio(lo, bio); |
1da177e4 LT |
594 | } |
595 | ||
1da177e4 LT |
596 | return 0; |
597 | } | |
598 | ||
599 | /* | |
600 | * loop_switch performs the hard work of switching a backing store. | |
601 | * First it needs to flush existing IO, it does this by sending a magic | |
602 | * BIO down the pipe. The completion of this BIO does the actual switch. | |
603 | */ | |
604 | static int loop_switch(struct loop_device *lo, struct file *file) | |
605 | { | |
606 | struct switch_request w; | |
a24eab1e | 607 | struct bio *bio = bio_alloc(GFP_KERNEL, 0); |
1da177e4 LT |
608 | if (!bio) |
609 | return -ENOMEM; | |
610 | init_completion(&w.wait); | |
611 | w.file = file; | |
612 | bio->bi_private = &w; | |
613 | bio->bi_bdev = NULL; | |
614 | loop_make_request(lo->lo_queue, bio); | |
615 | wait_for_completion(&w.wait); | |
616 | return 0; | |
617 | } | |
618 | ||
14f27939 MB |
619 | /* |
620 | * Helper to flush the IOs in loop, but keeping loop thread running | |
621 | */ | |
622 | static int loop_flush(struct loop_device *lo) | |
623 | { | |
624 | /* loop not yet configured, no running thread, nothing to flush */ | |
625 | if (!lo->lo_thread) | |
626 | return 0; | |
627 | ||
628 | return loop_switch(lo, NULL); | |
629 | } | |
630 | ||
1da177e4 LT |
631 | /* |
632 | * Do the actual switch; called from the BIO completion routine | |
633 | */ | |
634 | static void do_loop_switch(struct loop_device *lo, struct switch_request *p) | |
635 | { | |
636 | struct file *file = p->file; | |
637 | struct file *old_file = lo->lo_backing_file; | |
14f27939 MB |
638 | struct address_space *mapping; |
639 | ||
640 | /* if no new file, only flush of queued bios requested */ | |
641 | if (!file) | |
642 | goto out; | |
1da177e4 | 643 | |
14f27939 | 644 | mapping = file->f_mapping; |
1da177e4 LT |
645 | mapping_set_gfp_mask(old_file->f_mapping, lo->old_gfp_mask); |
646 | lo->lo_backing_file = file; | |
ba52de12 TT |
647 | lo->lo_blocksize = S_ISBLK(mapping->host->i_mode) ? |
648 | mapping->host->i_bdev->bd_block_size : PAGE_SIZE; | |
1da177e4 LT |
649 | lo->old_gfp_mask = mapping_gfp_mask(mapping); |
650 | mapping_set_gfp_mask(mapping, lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS)); | |
14f27939 | 651 | out: |
1da177e4 LT |
652 | complete(&p->wait); |
653 | } | |
654 | ||
655 | ||
656 | /* | |
657 | * loop_change_fd switched the backing store of a loopback device to | |
658 | * a new file. This is useful for operating system installers to free up | |
659 | * the original file and in High Availability environments to switch to | |
660 | * an alternative location for the content in case of server meltdown. | |
661 | * This can only work if the loop device is used read-only, and if the | |
662 | * new backing store is the same size and type as the old backing store. | |
663 | */ | |
bb214884 AV |
664 | static int loop_change_fd(struct loop_device *lo, struct block_device *bdev, |
665 | unsigned int arg) | |
1da177e4 LT |
666 | { |
667 | struct file *file, *old_file; | |
668 | struct inode *inode; | |
669 | int error; | |
670 | ||
671 | error = -ENXIO; | |
672 | if (lo->lo_state != Lo_bound) | |
673 | goto out; | |
674 | ||
675 | /* the loop device has to be read-only */ | |
676 | error = -EINVAL; | |
677 | if (!(lo->lo_flags & LO_FLAGS_READ_ONLY)) | |
678 | goto out; | |
679 | ||
680 | error = -EBADF; | |
681 | file = fget(arg); | |
682 | if (!file) | |
683 | goto out; | |
684 | ||
685 | inode = file->f_mapping->host; | |
686 | old_file = lo->lo_backing_file; | |
687 | ||
688 | error = -EINVAL; | |
689 | ||
690 | if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode)) | |
691 | goto out_putf; | |
692 | ||
1da177e4 LT |
693 | /* size of the new backing store needs to be the same */ |
694 | if (get_loop_size(lo, file) != get_loop_size(lo, old_file)) | |
695 | goto out_putf; | |
696 | ||
697 | /* and ... switch */ | |
698 | error = loop_switch(lo, file); | |
699 | if (error) | |
700 | goto out_putf; | |
701 | ||
702 | fput(old_file); | |
476a4813 LV |
703 | if (max_part > 0) |
704 | ioctl_by_bdev(bdev, BLKRRPART, 0); | |
1da177e4 LT |
705 | return 0; |
706 | ||
707 | out_putf: | |
708 | fput(file); | |
709 | out: | |
710 | return error; | |
711 | } | |
712 | ||
713 | static inline int is_loop_device(struct file *file) | |
714 | { | |
715 | struct inode *i = file->f_mapping->host; | |
716 | ||
717 | return i && S_ISBLK(i->i_mode) && MAJOR(i->i_rdev) == LOOP_MAJOR; | |
718 | } | |
719 | ||
ee862730 MB |
720 | /* loop sysfs attributes */ |
721 | ||
722 | static ssize_t loop_attr_show(struct device *dev, char *page, | |
723 | ssize_t (*callback)(struct loop_device *, char *)) | |
724 | { | |
34dd82af KS |
725 | struct gendisk *disk = dev_to_disk(dev); |
726 | struct loop_device *lo = disk->private_data; | |
ee862730 | 727 | |
34dd82af | 728 | return callback(lo, page); |
ee862730 MB |
729 | } |
730 | ||
731 | #define LOOP_ATTR_RO(_name) \ | |
732 | static ssize_t loop_attr_##_name##_show(struct loop_device *, char *); \ | |
733 | static ssize_t loop_attr_do_show_##_name(struct device *d, \ | |
734 | struct device_attribute *attr, char *b) \ | |
735 | { \ | |
736 | return loop_attr_show(d, b, loop_attr_##_name##_show); \ | |
737 | } \ | |
738 | static struct device_attribute loop_attr_##_name = \ | |
739 | __ATTR(_name, S_IRUGO, loop_attr_do_show_##_name, NULL); | |
740 | ||
741 | static ssize_t loop_attr_backing_file_show(struct loop_device *lo, char *buf) | |
742 | { | |
743 | ssize_t ret; | |
744 | char *p = NULL; | |
745 | ||
05eb0f25 | 746 | spin_lock_irq(&lo->lo_lock); |
ee862730 MB |
747 | if (lo->lo_backing_file) |
748 | p = d_path(&lo->lo_backing_file->f_path, buf, PAGE_SIZE - 1); | |
05eb0f25 | 749 | spin_unlock_irq(&lo->lo_lock); |
ee862730 MB |
750 | |
751 | if (IS_ERR_OR_NULL(p)) | |
752 | ret = PTR_ERR(p); | |
753 | else { | |
754 | ret = strlen(p); | |
755 | memmove(buf, p, ret); | |
756 | buf[ret++] = '\n'; | |
757 | buf[ret] = 0; | |
758 | } | |
759 | ||
760 | return ret; | |
761 | } | |
762 | ||
763 | static ssize_t loop_attr_offset_show(struct loop_device *lo, char *buf) | |
764 | { | |
765 | return sprintf(buf, "%llu\n", (unsigned long long)lo->lo_offset); | |
766 | } | |
767 | ||
768 | static ssize_t loop_attr_sizelimit_show(struct loop_device *lo, char *buf) | |
769 | { | |
770 | return sprintf(buf, "%llu\n", (unsigned long long)lo->lo_sizelimit); | |
771 | } | |
772 | ||
773 | static ssize_t loop_attr_autoclear_show(struct loop_device *lo, char *buf) | |
774 | { | |
775 | int autoclear = (lo->lo_flags & LO_FLAGS_AUTOCLEAR); | |
776 | ||
777 | return sprintf(buf, "%s\n", autoclear ? "1" : "0"); | |
778 | } | |
779 | ||
780 | LOOP_ATTR_RO(backing_file); | |
781 | LOOP_ATTR_RO(offset); | |
782 | LOOP_ATTR_RO(sizelimit); | |
783 | LOOP_ATTR_RO(autoclear); | |
784 | ||
785 | static struct attribute *loop_attrs[] = { | |
786 | &loop_attr_backing_file.attr, | |
787 | &loop_attr_offset.attr, | |
788 | &loop_attr_sizelimit.attr, | |
789 | &loop_attr_autoclear.attr, | |
790 | NULL, | |
791 | }; | |
792 | ||
793 | static struct attribute_group loop_attribute_group = { | |
794 | .name = "loop", | |
795 | .attrs= loop_attrs, | |
796 | }; | |
797 | ||
798 | static int loop_sysfs_init(struct loop_device *lo) | |
799 | { | |
800 | return sysfs_create_group(&disk_to_dev(lo->lo_disk)->kobj, | |
801 | &loop_attribute_group); | |
802 | } | |
803 | ||
804 | static void loop_sysfs_exit(struct loop_device *lo) | |
805 | { | |
806 | sysfs_remove_group(&disk_to_dev(lo->lo_disk)->kobj, | |
807 | &loop_attribute_group); | |
808 | } | |
809 | ||
bb214884 | 810 | static int loop_set_fd(struct loop_device *lo, fmode_t mode, |
1da177e4 LT |
811 | struct block_device *bdev, unsigned int arg) |
812 | { | |
813 | struct file *file, *f; | |
814 | struct inode *inode; | |
815 | struct address_space *mapping; | |
816 | unsigned lo_blocksize; | |
817 | int lo_flags = 0; | |
818 | int error; | |
819 | loff_t size; | |
820 | ||
821 | /* This is safe, since we have a reference from open(). */ | |
822 | __module_get(THIS_MODULE); | |
823 | ||
824 | error = -EBADF; | |
825 | file = fget(arg); | |
826 | if (!file) | |
827 | goto out; | |
828 | ||
829 | error = -EBUSY; | |
830 | if (lo->lo_state != Lo_unbound) | |
831 | goto out_putf; | |
832 | ||
833 | /* Avoid recursion */ | |
834 | f = file; | |
835 | while (is_loop_device(f)) { | |
836 | struct loop_device *l; | |
837 | ||
bb214884 | 838 | if (f->f_mapping->host->i_bdev == bdev) |
1da177e4 LT |
839 | goto out_putf; |
840 | ||
841 | l = f->f_mapping->host->i_bdev->bd_disk->private_data; | |
842 | if (l->lo_state == Lo_unbound) { | |
843 | error = -EINVAL; | |
844 | goto out_putf; | |
845 | } | |
846 | f = l->lo_backing_file; | |
847 | } | |
848 | ||
849 | mapping = file->f_mapping; | |
850 | inode = mapping->host; | |
851 | ||
852 | if (!(file->f_mode & FMODE_WRITE)) | |
853 | lo_flags |= LO_FLAGS_READ_ONLY; | |
854 | ||
855 | error = -EINVAL; | |
856 | if (S_ISREG(inode->i_mode) || S_ISBLK(inode->i_mode)) { | |
f5e54d6e | 857 | const struct address_space_operations *aops = mapping->a_ops; |
6818173b | 858 | |
4e02ed4b | 859 | if (aops->write_begin) |
1da177e4 LT |
860 | lo_flags |= LO_FLAGS_USE_AOPS; |
861 | if (!(lo_flags & LO_FLAGS_USE_AOPS) && !file->f_op->write) | |
862 | lo_flags |= LO_FLAGS_READ_ONLY; | |
863 | ||
ba52de12 TT |
864 | lo_blocksize = S_ISBLK(inode->i_mode) ? |
865 | inode->i_bdev->bd_block_size : PAGE_SIZE; | |
866 | ||
1da177e4 LT |
867 | error = 0; |
868 | } else { | |
869 | goto out_putf; | |
870 | } | |
871 | ||
872 | size = get_loop_size(lo, file); | |
873 | ||
874 | if ((loff_t)(sector_t)size != size) { | |
875 | error = -EFBIG; | |
876 | goto out_putf; | |
877 | } | |
878 | ||
bb214884 | 879 | if (!(mode & FMODE_WRITE)) |
1da177e4 LT |
880 | lo_flags |= LO_FLAGS_READ_ONLY; |
881 | ||
882 | set_device_ro(bdev, (lo_flags & LO_FLAGS_READ_ONLY) != 0); | |
883 | ||
884 | lo->lo_blocksize = lo_blocksize; | |
885 | lo->lo_device = bdev; | |
886 | lo->lo_flags = lo_flags; | |
887 | lo->lo_backing_file = file; | |
eefe85ee | 888 | lo->transfer = transfer_none; |
1da177e4 LT |
889 | lo->ioctl = NULL; |
890 | lo->lo_sizelimit = 0; | |
891 | lo->old_gfp_mask = mapping_gfp_mask(mapping); | |
892 | mapping_set_gfp_mask(mapping, lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS)); | |
893 | ||
e686307f | 894 | bio_list_init(&lo->lo_bio_list); |
1da177e4 LT |
895 | |
896 | /* | |
897 | * set queue make_request_fn, and add limits based on lower level | |
898 | * device | |
899 | */ | |
900 | blk_queue_make_request(lo->lo_queue, loop_make_request); | |
901 | lo->lo_queue->queuedata = lo; | |
1da177e4 | 902 | |
68db1961 | 903 | if (!(lo_flags & LO_FLAGS_READ_ONLY) && file->f_op->fsync) |
4913efe4 | 904 | blk_queue_flush(lo->lo_queue, REQ_FLUSH); |
68db1961 | 905 | |
73285082 | 906 | set_capacity(lo->lo_disk, size); |
1da177e4 | 907 | bd_set_size(bdev, size << 9); |
ee862730 | 908 | loop_sysfs_init(lo); |
c3473c63 DZ |
909 | /* let user-space know about the new size */ |
910 | kobject_uevent(&disk_to_dev(bdev->bd_disk)->kobj, KOBJ_CHANGE); | |
1da177e4 LT |
911 | |
912 | set_blocksize(bdev, lo_blocksize); | |
913 | ||
6c997918 SH |
914 | lo->lo_thread = kthread_create(loop_thread, lo, "loop%d", |
915 | lo->lo_number); | |
916 | if (IS_ERR(lo->lo_thread)) { | |
917 | error = PTR_ERR(lo->lo_thread); | |
a7422bf8 | 918 | goto out_clr; |
6c997918 SH |
919 | } |
920 | lo->lo_state = Lo_bound; | |
921 | wake_up_process(lo->lo_thread); | |
476a4813 LV |
922 | if (max_part > 0) |
923 | ioctl_by_bdev(bdev, BLKRRPART, 0); | |
1da177e4 LT |
924 | return 0; |
925 | ||
a7422bf8 | 926 | out_clr: |
ee862730 | 927 | loop_sysfs_exit(lo); |
a7422bf8 SH |
928 | lo->lo_thread = NULL; |
929 | lo->lo_device = NULL; | |
930 | lo->lo_backing_file = NULL; | |
931 | lo->lo_flags = 0; | |
73285082 | 932 | set_capacity(lo->lo_disk, 0); |
f98393a6 | 933 | invalidate_bdev(bdev); |
a7422bf8 | 934 | bd_set_size(bdev, 0); |
c3473c63 | 935 | kobject_uevent(&disk_to_dev(bdev->bd_disk)->kobj, KOBJ_CHANGE); |
a7422bf8 SH |
936 | mapping_set_gfp_mask(mapping, lo->old_gfp_mask); |
937 | lo->lo_state = Lo_unbound; | |
1da177e4 LT |
938 | out_putf: |
939 | fput(file); | |
940 | out: | |
941 | /* This is safe: open() is still holding a reference. */ | |
942 | module_put(THIS_MODULE); | |
943 | return error; | |
944 | } | |
945 | ||
946 | static int | |
947 | loop_release_xfer(struct loop_device *lo) | |
948 | { | |
949 | int err = 0; | |
950 | struct loop_func_table *xfer = lo->lo_encryption; | |
951 | ||
952 | if (xfer) { | |
953 | if (xfer->release) | |
954 | err = xfer->release(lo); | |
955 | lo->transfer = NULL; | |
956 | lo->lo_encryption = NULL; | |
957 | module_put(xfer->owner); | |
958 | } | |
959 | return err; | |
960 | } | |
961 | ||
962 | static int | |
963 | loop_init_xfer(struct loop_device *lo, struct loop_func_table *xfer, | |
964 | const struct loop_info64 *i) | |
965 | { | |
966 | int err = 0; | |
967 | ||
968 | if (xfer) { | |
969 | struct module *owner = xfer->owner; | |
970 | ||
971 | if (!try_module_get(owner)) | |
972 | return -EINVAL; | |
973 | if (xfer->init) | |
974 | err = xfer->init(lo, i); | |
975 | if (err) | |
976 | module_put(owner); | |
977 | else | |
978 | lo->lo_encryption = xfer; | |
979 | } | |
980 | return err; | |
981 | } | |
982 | ||
983 | static int loop_clr_fd(struct loop_device *lo, struct block_device *bdev) | |
984 | { | |
985 | struct file *filp = lo->lo_backing_file; | |
b4e3ca1a | 986 | gfp_t gfp = lo->old_gfp_mask; |
1da177e4 LT |
987 | |
988 | if (lo->lo_state != Lo_bound) | |
989 | return -ENXIO; | |
990 | ||
991 | if (lo->lo_refcnt > 1) /* we needed one fd for the ioctl */ | |
992 | return -EBUSY; | |
993 | ||
994 | if (filp == NULL) | |
995 | return -EINVAL; | |
996 | ||
997 | spin_lock_irq(&lo->lo_lock); | |
998 | lo->lo_state = Lo_rundown; | |
1da177e4 LT |
999 | spin_unlock_irq(&lo->lo_lock); |
1000 | ||
6c997918 | 1001 | kthread_stop(lo->lo_thread); |
1da177e4 | 1002 | |
05eb0f25 | 1003 | spin_lock_irq(&lo->lo_lock); |
1da177e4 | 1004 | lo->lo_backing_file = NULL; |
05eb0f25 | 1005 | spin_unlock_irq(&lo->lo_lock); |
1da177e4 LT |
1006 | |
1007 | loop_release_xfer(lo); | |
1008 | lo->transfer = NULL; | |
1009 | lo->ioctl = NULL; | |
1010 | lo->lo_device = NULL; | |
1011 | lo->lo_encryption = NULL; | |
1012 | lo->lo_offset = 0; | |
1013 | lo->lo_sizelimit = 0; | |
1014 | lo->lo_encrypt_key_size = 0; | |
1015 | lo->lo_flags = 0; | |
6c997918 | 1016 | lo->lo_thread = NULL; |
1da177e4 LT |
1017 | memset(lo->lo_encrypt_key, 0, LO_KEY_SIZE); |
1018 | memset(lo->lo_crypt_name, 0, LO_NAME_SIZE); | |
1019 | memset(lo->lo_file_name, 0, LO_NAME_SIZE); | |
bb214884 AV |
1020 | if (bdev) |
1021 | invalidate_bdev(bdev); | |
73285082 | 1022 | set_capacity(lo->lo_disk, 0); |
51a0bb0c | 1023 | loop_sysfs_exit(lo); |
c3473c63 | 1024 | if (bdev) { |
bb214884 | 1025 | bd_set_size(bdev, 0); |
c3473c63 DZ |
1026 | /* let user-space know about this change */ |
1027 | kobject_uevent(&disk_to_dev(bdev->bd_disk)->kobj, KOBJ_CHANGE); | |
1028 | } | |
1da177e4 LT |
1029 | mapping_set_gfp_mask(filp->f_mapping, gfp); |
1030 | lo->lo_state = Lo_unbound; | |
1da177e4 LT |
1031 | /* This is safe: open() is still holding a reference. */ |
1032 | module_put(THIS_MODULE); | |
cf6e6932 | 1033 | if (max_part > 0 && bdev) |
476a4813 | 1034 | ioctl_by_bdev(bdev, BLKRRPART, 0); |
f028f3b2 NK |
1035 | mutex_unlock(&lo->lo_ctl_mutex); |
1036 | /* | |
1037 | * Need not hold lo_ctl_mutex to fput backing file. | |
1038 | * Calling fput holding lo_ctl_mutex triggers a circular | |
1039 | * lock dependency possibility warning as fput can take | |
1040 | * bd_mutex which is usually taken before lo_ctl_mutex. | |
1041 | */ | |
1042 | fput(filp); | |
1da177e4 LT |
1043 | return 0; |
1044 | } | |
1045 | ||
1046 | static int | |
1047 | loop_set_status(struct loop_device *lo, const struct loop_info64 *info) | |
1048 | { | |
1049 | int err; | |
1050 | struct loop_func_table *xfer; | |
b0fafa81 | 1051 | uid_t uid = current_uid(); |
1da177e4 | 1052 | |
b0fafa81 DH |
1053 | if (lo->lo_encrypt_key_size && |
1054 | lo->lo_key_owner != uid && | |
1da177e4 LT |
1055 | !capable(CAP_SYS_ADMIN)) |
1056 | return -EPERM; | |
1057 | if (lo->lo_state != Lo_bound) | |
1058 | return -ENXIO; | |
1059 | if ((unsigned int) info->lo_encrypt_key_size > LO_KEY_SIZE) | |
1060 | return -EINVAL; | |
1061 | ||
1062 | err = loop_release_xfer(lo); | |
1063 | if (err) | |
1064 | return err; | |
1065 | ||
1066 | if (info->lo_encrypt_type) { | |
1067 | unsigned int type = info->lo_encrypt_type; | |
1068 | ||
1069 | if (type >= MAX_LO_CRYPT) | |
1070 | return -EINVAL; | |
1071 | xfer = xfer_funcs[type]; | |
1072 | if (xfer == NULL) | |
1073 | return -EINVAL; | |
1074 | } else | |
1075 | xfer = NULL; | |
1076 | ||
1077 | err = loop_init_xfer(lo, xfer, info); | |
1078 | if (err) | |
1079 | return err; | |
1080 | ||
1081 | if (lo->lo_offset != info->lo_offset || | |
1082 | lo->lo_sizelimit != info->lo_sizelimit) { | |
1083 | lo->lo_offset = info->lo_offset; | |
1084 | lo->lo_sizelimit = info->lo_sizelimit; | |
1085 | if (figure_loop_size(lo)) | |
1086 | return -EFBIG; | |
1087 | } | |
1088 | ||
1089 | memcpy(lo->lo_file_name, info->lo_file_name, LO_NAME_SIZE); | |
1090 | memcpy(lo->lo_crypt_name, info->lo_crypt_name, LO_NAME_SIZE); | |
1091 | lo->lo_file_name[LO_NAME_SIZE-1] = 0; | |
1092 | lo->lo_crypt_name[LO_NAME_SIZE-1] = 0; | |
1093 | ||
1094 | if (!xfer) | |
1095 | xfer = &none_funcs; | |
1096 | lo->transfer = xfer->transfer; | |
1097 | lo->ioctl = xfer->ioctl; | |
1098 | ||
96c58655 DW |
1099 | if ((lo->lo_flags & LO_FLAGS_AUTOCLEAR) != |
1100 | (info->lo_flags & LO_FLAGS_AUTOCLEAR)) | |
1101 | lo->lo_flags ^= LO_FLAGS_AUTOCLEAR; | |
1102 | ||
1da177e4 LT |
1103 | lo->lo_encrypt_key_size = info->lo_encrypt_key_size; |
1104 | lo->lo_init[0] = info->lo_init[0]; | |
1105 | lo->lo_init[1] = info->lo_init[1]; | |
1106 | if (info->lo_encrypt_key_size) { | |
1107 | memcpy(lo->lo_encrypt_key, info->lo_encrypt_key, | |
1108 | info->lo_encrypt_key_size); | |
b0fafa81 | 1109 | lo->lo_key_owner = uid; |
1da177e4 LT |
1110 | } |
1111 | ||
1112 | return 0; | |
1113 | } | |
1114 | ||
1115 | static int | |
1116 | loop_get_status(struct loop_device *lo, struct loop_info64 *info) | |
1117 | { | |
1118 | struct file *file = lo->lo_backing_file; | |
1119 | struct kstat stat; | |
1120 | int error; | |
1121 | ||
1122 | if (lo->lo_state != Lo_bound) | |
1123 | return -ENXIO; | |
6c648be6 | 1124 | error = vfs_getattr(file->f_path.mnt, file->f_path.dentry, &stat); |
1da177e4 LT |
1125 | if (error) |
1126 | return error; | |
1127 | memset(info, 0, sizeof(*info)); | |
1128 | info->lo_number = lo->lo_number; | |
1129 | info->lo_device = huge_encode_dev(stat.dev); | |
1130 | info->lo_inode = stat.ino; | |
1131 | info->lo_rdevice = huge_encode_dev(lo->lo_device ? stat.rdev : stat.dev); | |
1132 | info->lo_offset = lo->lo_offset; | |
1133 | info->lo_sizelimit = lo->lo_sizelimit; | |
1134 | info->lo_flags = lo->lo_flags; | |
1135 | memcpy(info->lo_file_name, lo->lo_file_name, LO_NAME_SIZE); | |
1136 | memcpy(info->lo_crypt_name, lo->lo_crypt_name, LO_NAME_SIZE); | |
1137 | info->lo_encrypt_type = | |
1138 | lo->lo_encryption ? lo->lo_encryption->number : 0; | |
1139 | if (lo->lo_encrypt_key_size && capable(CAP_SYS_ADMIN)) { | |
1140 | info->lo_encrypt_key_size = lo->lo_encrypt_key_size; | |
1141 | memcpy(info->lo_encrypt_key, lo->lo_encrypt_key, | |
1142 | lo->lo_encrypt_key_size); | |
1143 | } | |
1144 | return 0; | |
1145 | } | |
1146 | ||
1147 | static void | |
1148 | loop_info64_from_old(const struct loop_info *info, struct loop_info64 *info64) | |
1149 | { | |
1150 | memset(info64, 0, sizeof(*info64)); | |
1151 | info64->lo_number = info->lo_number; | |
1152 | info64->lo_device = info->lo_device; | |
1153 | info64->lo_inode = info->lo_inode; | |
1154 | info64->lo_rdevice = info->lo_rdevice; | |
1155 | info64->lo_offset = info->lo_offset; | |
1156 | info64->lo_sizelimit = 0; | |
1157 | info64->lo_encrypt_type = info->lo_encrypt_type; | |
1158 | info64->lo_encrypt_key_size = info->lo_encrypt_key_size; | |
1159 | info64->lo_flags = info->lo_flags; | |
1160 | info64->lo_init[0] = info->lo_init[0]; | |
1161 | info64->lo_init[1] = info->lo_init[1]; | |
1162 | if (info->lo_encrypt_type == LO_CRYPT_CRYPTOAPI) | |
1163 | memcpy(info64->lo_crypt_name, info->lo_name, LO_NAME_SIZE); | |
1164 | else | |
1165 | memcpy(info64->lo_file_name, info->lo_name, LO_NAME_SIZE); | |
1166 | memcpy(info64->lo_encrypt_key, info->lo_encrypt_key, LO_KEY_SIZE); | |
1167 | } | |
1168 | ||
1169 | static int | |
1170 | loop_info64_to_old(const struct loop_info64 *info64, struct loop_info *info) | |
1171 | { | |
1172 | memset(info, 0, sizeof(*info)); | |
1173 | info->lo_number = info64->lo_number; | |
1174 | info->lo_device = info64->lo_device; | |
1175 | info->lo_inode = info64->lo_inode; | |
1176 | info->lo_rdevice = info64->lo_rdevice; | |
1177 | info->lo_offset = info64->lo_offset; | |
1178 | info->lo_encrypt_type = info64->lo_encrypt_type; | |
1179 | info->lo_encrypt_key_size = info64->lo_encrypt_key_size; | |
1180 | info->lo_flags = info64->lo_flags; | |
1181 | info->lo_init[0] = info64->lo_init[0]; | |
1182 | info->lo_init[1] = info64->lo_init[1]; | |
1183 | if (info->lo_encrypt_type == LO_CRYPT_CRYPTOAPI) | |
1184 | memcpy(info->lo_name, info64->lo_crypt_name, LO_NAME_SIZE); | |
1185 | else | |
1186 | memcpy(info->lo_name, info64->lo_file_name, LO_NAME_SIZE); | |
1187 | memcpy(info->lo_encrypt_key, info64->lo_encrypt_key, LO_KEY_SIZE); | |
1188 | ||
1189 | /* error in case values were truncated */ | |
1190 | if (info->lo_device != info64->lo_device || | |
1191 | info->lo_rdevice != info64->lo_rdevice || | |
1192 | info->lo_inode != info64->lo_inode || | |
1193 | info->lo_offset != info64->lo_offset) | |
1194 | return -EOVERFLOW; | |
1195 | ||
1196 | return 0; | |
1197 | } | |
1198 | ||
1199 | static int | |
1200 | loop_set_status_old(struct loop_device *lo, const struct loop_info __user *arg) | |
1201 | { | |
1202 | struct loop_info info; | |
1203 | struct loop_info64 info64; | |
1204 | ||
1205 | if (copy_from_user(&info, arg, sizeof (struct loop_info))) | |
1206 | return -EFAULT; | |
1207 | loop_info64_from_old(&info, &info64); | |
1208 | return loop_set_status(lo, &info64); | |
1209 | } | |
1210 | ||
1211 | static int | |
1212 | loop_set_status64(struct loop_device *lo, const struct loop_info64 __user *arg) | |
1213 | { | |
1214 | struct loop_info64 info64; | |
1215 | ||
1216 | if (copy_from_user(&info64, arg, sizeof (struct loop_info64))) | |
1217 | return -EFAULT; | |
1218 | return loop_set_status(lo, &info64); | |
1219 | } | |
1220 | ||
1221 | static int | |
1222 | loop_get_status_old(struct loop_device *lo, struct loop_info __user *arg) { | |
1223 | struct loop_info info; | |
1224 | struct loop_info64 info64; | |
1225 | int err = 0; | |
1226 | ||
1227 | if (!arg) | |
1228 | err = -EINVAL; | |
1229 | if (!err) | |
1230 | err = loop_get_status(lo, &info64); | |
1231 | if (!err) | |
1232 | err = loop_info64_to_old(&info64, &info); | |
1233 | if (!err && copy_to_user(arg, &info, sizeof(info))) | |
1234 | err = -EFAULT; | |
1235 | ||
1236 | return err; | |
1237 | } | |
1238 | ||
1239 | static int | |
1240 | loop_get_status64(struct loop_device *lo, struct loop_info64 __user *arg) { | |
1241 | struct loop_info64 info64; | |
1242 | int err = 0; | |
1243 | ||
1244 | if (!arg) | |
1245 | err = -EINVAL; | |
1246 | if (!err) | |
1247 | err = loop_get_status(lo, &info64); | |
1248 | if (!err && copy_to_user(arg, &info64, sizeof(info64))) | |
1249 | err = -EFAULT; | |
1250 | ||
1251 | return err; | |
1252 | } | |
1253 | ||
53d66608 O |
1254 | static int loop_set_capacity(struct loop_device *lo, struct block_device *bdev) |
1255 | { | |
1256 | int err; | |
1257 | sector_t sec; | |
1258 | loff_t sz; | |
1259 | ||
1260 | err = -ENXIO; | |
1261 | if (unlikely(lo->lo_state != Lo_bound)) | |
1262 | goto out; | |
1263 | err = figure_loop_size(lo); | |
1264 | if (unlikely(err)) | |
1265 | goto out; | |
1266 | sec = get_capacity(lo->lo_disk); | |
1267 | /* the width of sector_t may be narrow for bit-shift */ | |
1268 | sz = sec; | |
1269 | sz <<= 9; | |
1270 | mutex_lock(&bdev->bd_mutex); | |
1271 | bd_set_size(bdev, sz); | |
c3473c63 DZ |
1272 | /* let user-space know about the new size */ |
1273 | kobject_uevent(&disk_to_dev(bdev->bd_disk)->kobj, KOBJ_CHANGE); | |
53d66608 O |
1274 | mutex_unlock(&bdev->bd_mutex); |
1275 | ||
1276 | out: | |
1277 | return err; | |
1278 | } | |
1279 | ||
bb214884 | 1280 | static int lo_ioctl(struct block_device *bdev, fmode_t mode, |
1da177e4 LT |
1281 | unsigned int cmd, unsigned long arg) |
1282 | { | |
bb214884 | 1283 | struct loop_device *lo = bdev->bd_disk->private_data; |
1da177e4 LT |
1284 | int err; |
1285 | ||
f028f3b2 | 1286 | mutex_lock_nested(&lo->lo_ctl_mutex, 1); |
1da177e4 LT |
1287 | switch (cmd) { |
1288 | case LOOP_SET_FD: | |
bb214884 | 1289 | err = loop_set_fd(lo, mode, bdev, arg); |
1da177e4 LT |
1290 | break; |
1291 | case LOOP_CHANGE_FD: | |
bb214884 | 1292 | err = loop_change_fd(lo, bdev, arg); |
1da177e4 LT |
1293 | break; |
1294 | case LOOP_CLR_FD: | |
f028f3b2 | 1295 | /* loop_clr_fd would have unlocked lo_ctl_mutex on success */ |
bb214884 | 1296 | err = loop_clr_fd(lo, bdev); |
f028f3b2 NK |
1297 | if (!err) |
1298 | goto out_unlocked; | |
1da177e4 LT |
1299 | break; |
1300 | case LOOP_SET_STATUS: | |
1301 | err = loop_set_status_old(lo, (struct loop_info __user *) arg); | |
1302 | break; | |
1303 | case LOOP_GET_STATUS: | |
1304 | err = loop_get_status_old(lo, (struct loop_info __user *) arg); | |
1305 | break; | |
1306 | case LOOP_SET_STATUS64: | |
1307 | err = loop_set_status64(lo, (struct loop_info64 __user *) arg); | |
1308 | break; | |
1309 | case LOOP_GET_STATUS64: | |
1310 | err = loop_get_status64(lo, (struct loop_info64 __user *) arg); | |
1311 | break; | |
53d66608 O |
1312 | case LOOP_SET_CAPACITY: |
1313 | err = -EPERM; | |
1314 | if ((mode & FMODE_WRITE) || capable(CAP_SYS_ADMIN)) | |
1315 | err = loop_set_capacity(lo, bdev); | |
1316 | break; | |
1da177e4 LT |
1317 | default: |
1318 | err = lo->ioctl ? lo->ioctl(lo, cmd, arg) : -EINVAL; | |
1319 | } | |
f85221dd | 1320 | mutex_unlock(&lo->lo_ctl_mutex); |
f028f3b2 NK |
1321 | |
1322 | out_unlocked: | |
1da177e4 LT |
1323 | return err; |
1324 | } | |
1325 | ||
863d5b82 DH |
1326 | #ifdef CONFIG_COMPAT |
1327 | struct compat_loop_info { | |
1328 | compat_int_t lo_number; /* ioctl r/o */ | |
1329 | compat_dev_t lo_device; /* ioctl r/o */ | |
1330 | compat_ulong_t lo_inode; /* ioctl r/o */ | |
1331 | compat_dev_t lo_rdevice; /* ioctl r/o */ | |
1332 | compat_int_t lo_offset; | |
1333 | compat_int_t lo_encrypt_type; | |
1334 | compat_int_t lo_encrypt_key_size; /* ioctl w/o */ | |
1335 | compat_int_t lo_flags; /* ioctl r/o */ | |
1336 | char lo_name[LO_NAME_SIZE]; | |
1337 | unsigned char lo_encrypt_key[LO_KEY_SIZE]; /* ioctl w/o */ | |
1338 | compat_ulong_t lo_init[2]; | |
1339 | char reserved[4]; | |
1340 | }; | |
1341 | ||
1342 | /* | |
1343 | * Transfer 32-bit compatibility structure in userspace to 64-bit loop info | |
1344 | * - noinlined to reduce stack space usage in main part of driver | |
1345 | */ | |
1346 | static noinline int | |
ba674cfc | 1347 | loop_info64_from_compat(const struct compat_loop_info __user *arg, |
863d5b82 DH |
1348 | struct loop_info64 *info64) |
1349 | { | |
1350 | struct compat_loop_info info; | |
1351 | ||
1352 | if (copy_from_user(&info, arg, sizeof(info))) | |
1353 | return -EFAULT; | |
1354 | ||
1355 | memset(info64, 0, sizeof(*info64)); | |
1356 | info64->lo_number = info.lo_number; | |
1357 | info64->lo_device = info.lo_device; | |
1358 | info64->lo_inode = info.lo_inode; | |
1359 | info64->lo_rdevice = info.lo_rdevice; | |
1360 | info64->lo_offset = info.lo_offset; | |
1361 | info64->lo_sizelimit = 0; | |
1362 | info64->lo_encrypt_type = info.lo_encrypt_type; | |
1363 | info64->lo_encrypt_key_size = info.lo_encrypt_key_size; | |
1364 | info64->lo_flags = info.lo_flags; | |
1365 | info64->lo_init[0] = info.lo_init[0]; | |
1366 | info64->lo_init[1] = info.lo_init[1]; | |
1367 | if (info.lo_encrypt_type == LO_CRYPT_CRYPTOAPI) | |
1368 | memcpy(info64->lo_crypt_name, info.lo_name, LO_NAME_SIZE); | |
1369 | else | |
1370 | memcpy(info64->lo_file_name, info.lo_name, LO_NAME_SIZE); | |
1371 | memcpy(info64->lo_encrypt_key, info.lo_encrypt_key, LO_KEY_SIZE); | |
1372 | return 0; | |
1373 | } | |
1374 | ||
1375 | /* | |
1376 | * Transfer 64-bit loop info to 32-bit compatibility structure in userspace | |
1377 | * - noinlined to reduce stack space usage in main part of driver | |
1378 | */ | |
1379 | static noinline int | |
1380 | loop_info64_to_compat(const struct loop_info64 *info64, | |
1381 | struct compat_loop_info __user *arg) | |
1382 | { | |
1383 | struct compat_loop_info info; | |
1384 | ||
1385 | memset(&info, 0, sizeof(info)); | |
1386 | info.lo_number = info64->lo_number; | |
1387 | info.lo_device = info64->lo_device; | |
1388 | info.lo_inode = info64->lo_inode; | |
1389 | info.lo_rdevice = info64->lo_rdevice; | |
1390 | info.lo_offset = info64->lo_offset; | |
1391 | info.lo_encrypt_type = info64->lo_encrypt_type; | |
1392 | info.lo_encrypt_key_size = info64->lo_encrypt_key_size; | |
1393 | info.lo_flags = info64->lo_flags; | |
1394 | info.lo_init[0] = info64->lo_init[0]; | |
1395 | info.lo_init[1] = info64->lo_init[1]; | |
1396 | if (info.lo_encrypt_type == LO_CRYPT_CRYPTOAPI) | |
1397 | memcpy(info.lo_name, info64->lo_crypt_name, LO_NAME_SIZE); | |
1398 | else | |
1399 | memcpy(info.lo_name, info64->lo_file_name, LO_NAME_SIZE); | |
1400 | memcpy(info.lo_encrypt_key, info64->lo_encrypt_key, LO_KEY_SIZE); | |
1401 | ||
1402 | /* error in case values were truncated */ | |
1403 | if (info.lo_device != info64->lo_device || | |
1404 | info.lo_rdevice != info64->lo_rdevice || | |
1405 | info.lo_inode != info64->lo_inode || | |
1406 | info.lo_offset != info64->lo_offset || | |
1407 | info.lo_init[0] != info64->lo_init[0] || | |
1408 | info.lo_init[1] != info64->lo_init[1]) | |
1409 | return -EOVERFLOW; | |
1410 | ||
1411 | if (copy_to_user(arg, &info, sizeof(info))) | |
1412 | return -EFAULT; | |
1413 | return 0; | |
1414 | } | |
1415 | ||
1416 | static int | |
1417 | loop_set_status_compat(struct loop_device *lo, | |
1418 | const struct compat_loop_info __user *arg) | |
1419 | { | |
1420 | struct loop_info64 info64; | |
1421 | int ret; | |
1422 | ||
1423 | ret = loop_info64_from_compat(arg, &info64); | |
1424 | if (ret < 0) | |
1425 | return ret; | |
1426 | return loop_set_status(lo, &info64); | |
1427 | } | |
1428 | ||
1429 | static int | |
1430 | loop_get_status_compat(struct loop_device *lo, | |
1431 | struct compat_loop_info __user *arg) | |
1432 | { | |
1433 | struct loop_info64 info64; | |
1434 | int err = 0; | |
1435 | ||
1436 | if (!arg) | |
1437 | err = -EINVAL; | |
1438 | if (!err) | |
1439 | err = loop_get_status(lo, &info64); | |
1440 | if (!err) | |
1441 | err = loop_info64_to_compat(&info64, arg); | |
1442 | return err; | |
1443 | } | |
1444 | ||
bb214884 AV |
1445 | static int lo_compat_ioctl(struct block_device *bdev, fmode_t mode, |
1446 | unsigned int cmd, unsigned long arg) | |
863d5b82 | 1447 | { |
bb214884 | 1448 | struct loop_device *lo = bdev->bd_disk->private_data; |
863d5b82 DH |
1449 | int err; |
1450 | ||
863d5b82 DH |
1451 | switch(cmd) { |
1452 | case LOOP_SET_STATUS: | |
1453 | mutex_lock(&lo->lo_ctl_mutex); | |
1454 | err = loop_set_status_compat( | |
1455 | lo, (const struct compat_loop_info __user *) arg); | |
1456 | mutex_unlock(&lo->lo_ctl_mutex); | |
1457 | break; | |
1458 | case LOOP_GET_STATUS: | |
1459 | mutex_lock(&lo->lo_ctl_mutex); | |
1460 | err = loop_get_status_compat( | |
1461 | lo, (struct compat_loop_info __user *) arg); | |
1462 | mutex_unlock(&lo->lo_ctl_mutex); | |
1463 | break; | |
53d66608 | 1464 | case LOOP_SET_CAPACITY: |
863d5b82 DH |
1465 | case LOOP_CLR_FD: |
1466 | case LOOP_GET_STATUS64: | |
1467 | case LOOP_SET_STATUS64: | |
1468 | arg = (unsigned long) compat_ptr(arg); | |
1469 | case LOOP_SET_FD: | |
1470 | case LOOP_CHANGE_FD: | |
bb214884 | 1471 | err = lo_ioctl(bdev, mode, cmd, arg); |
863d5b82 DH |
1472 | break; |
1473 | default: | |
1474 | err = -ENOIOCTLCMD; | |
1475 | break; | |
1476 | } | |
863d5b82 DH |
1477 | return err; |
1478 | } | |
1479 | #endif | |
1480 | ||
bb214884 | 1481 | static int lo_open(struct block_device *bdev, fmode_t mode) |
1da177e4 | 1482 | { |
770fe30a KS |
1483 | struct loop_device *lo; |
1484 | int err = 0; | |
1485 | ||
1486 | mutex_lock(&loop_index_mutex); | |
1487 | lo = bdev->bd_disk->private_data; | |
1488 | if (!lo) { | |
1489 | err = -ENXIO; | |
1490 | goto out; | |
1491 | } | |
1da177e4 | 1492 | |
f85221dd | 1493 | mutex_lock(&lo->lo_ctl_mutex); |
1da177e4 | 1494 | lo->lo_refcnt++; |
f85221dd | 1495 | mutex_unlock(&lo->lo_ctl_mutex); |
770fe30a KS |
1496 | out: |
1497 | mutex_unlock(&loop_index_mutex); | |
1498 | return err; | |
1da177e4 LT |
1499 | } |
1500 | ||
bb214884 | 1501 | static int lo_release(struct gendisk *disk, fmode_t mode) |
1da177e4 | 1502 | { |
bb214884 | 1503 | struct loop_device *lo = disk->private_data; |
ffcd7dca | 1504 | int err; |
1da177e4 | 1505 | |
f85221dd | 1506 | mutex_lock(&lo->lo_ctl_mutex); |
96c58655 | 1507 | |
14f27939 MB |
1508 | if (--lo->lo_refcnt) |
1509 | goto out; | |
1510 | ||
1511 | if (lo->lo_flags & LO_FLAGS_AUTOCLEAR) { | |
1512 | /* | |
1513 | * In autoclear mode, stop the loop thread | |
1514 | * and remove configuration after last close. | |
1515 | */ | |
ffcd7dca AB |
1516 | err = loop_clr_fd(lo, NULL); |
1517 | if (!err) | |
1518 | goto out_unlocked; | |
14f27939 MB |
1519 | } else { |
1520 | /* | |
1521 | * Otherwise keep thread (if running) and config, | |
1522 | * but flush possible ongoing bios in thread. | |
1523 | */ | |
1524 | loop_flush(lo); | |
1525 | } | |
96c58655 | 1526 | |
14f27939 | 1527 | out: |
f85221dd | 1528 | mutex_unlock(&lo->lo_ctl_mutex); |
ffcd7dca | 1529 | out_unlocked: |
1da177e4 LT |
1530 | return 0; |
1531 | } | |
1532 | ||
83d5cde4 | 1533 | static const struct block_device_operations lo_fops = { |
1da177e4 | 1534 | .owner = THIS_MODULE, |
bb214884 AV |
1535 | .open = lo_open, |
1536 | .release = lo_release, | |
1537 | .ioctl = lo_ioctl, | |
863d5b82 | 1538 | #ifdef CONFIG_COMPAT |
bb214884 | 1539 | .compat_ioctl = lo_compat_ioctl, |
863d5b82 | 1540 | #endif |
1da177e4 LT |
1541 | }; |
1542 | ||
1543 | /* | |
1544 | * And now the modules code and kernel interface. | |
1545 | */ | |
73285082 | 1546 | static int max_loop; |
ac04fee0 | 1547 | module_param(max_loop, int, S_IRUGO); |
a47653fc | 1548 | MODULE_PARM_DESC(max_loop, "Maximum number of loop devices"); |
ac04fee0 | 1549 | module_param(max_part, int, S_IRUGO); |
476a4813 | 1550 | MODULE_PARM_DESC(max_part, "Maximum number of partitions per loop device"); |
1da177e4 LT |
1551 | MODULE_LICENSE("GPL"); |
1552 | MODULE_ALIAS_BLOCKDEV_MAJOR(LOOP_MAJOR); | |
1553 | ||
1554 | int loop_register_transfer(struct loop_func_table *funcs) | |
1555 | { | |
1556 | unsigned int n = funcs->number; | |
1557 | ||
1558 | if (n >= MAX_LO_CRYPT || xfer_funcs[n]) | |
1559 | return -EINVAL; | |
1560 | xfer_funcs[n] = funcs; | |
1561 | return 0; | |
1562 | } | |
1563 | ||
34dd82af KS |
1564 | static int unregister_transfer_cb(int id, void *ptr, void *data) |
1565 | { | |
1566 | struct loop_device *lo = ptr; | |
1567 | struct loop_func_table *xfer = data; | |
1568 | ||
1569 | mutex_lock(&lo->lo_ctl_mutex); | |
1570 | if (lo->lo_encryption == xfer) | |
1571 | loop_release_xfer(lo); | |
1572 | mutex_unlock(&lo->lo_ctl_mutex); | |
1573 | return 0; | |
1574 | } | |
1575 | ||
1da177e4 LT |
1576 | int loop_unregister_transfer(int number) |
1577 | { | |
1578 | unsigned int n = number; | |
1da177e4 LT |
1579 | struct loop_func_table *xfer; |
1580 | ||
1581 | if (n == 0 || n >= MAX_LO_CRYPT || (xfer = xfer_funcs[n]) == NULL) | |
1582 | return -EINVAL; | |
1583 | ||
1584 | xfer_funcs[n] = NULL; | |
34dd82af | 1585 | idr_for_each(&loop_index_idr, &unregister_transfer_cb, xfer); |
1da177e4 LT |
1586 | return 0; |
1587 | } | |
1588 | ||
1589 | EXPORT_SYMBOL(loop_register_transfer); | |
1590 | EXPORT_SYMBOL(loop_unregister_transfer); | |
1591 | ||
34dd82af | 1592 | static int loop_add(struct loop_device **l, int i) |
73285082 KC |
1593 | { |
1594 | struct loop_device *lo; | |
1595 | struct gendisk *disk; | |
34dd82af | 1596 | int err; |
73285082 KC |
1597 | |
1598 | lo = kzalloc(sizeof(*lo), GFP_KERNEL); | |
34dd82af KS |
1599 | if (!lo) { |
1600 | err = -ENOMEM; | |
73285082 | 1601 | goto out; |
34dd82af KS |
1602 | } |
1603 | ||
1604 | err = idr_pre_get(&loop_index_idr, GFP_KERNEL); | |
1605 | if (err < 0) | |
1606 | goto out_free_dev; | |
1607 | ||
1608 | if (i >= 0) { | |
1609 | int m; | |
1610 | ||
1611 | /* create specific i in the index */ | |
1612 | err = idr_get_new_above(&loop_index_idr, lo, i, &m); | |
1613 | if (err >= 0 && i != m) { | |
1614 | idr_remove(&loop_index_idr, m); | |
1615 | err = -EEXIST; | |
1616 | } | |
770fe30a KS |
1617 | } else if (i == -1) { |
1618 | int m; | |
1619 | ||
1620 | /* get next free nr */ | |
1621 | err = idr_get_new(&loop_index_idr, lo, &m); | |
1622 | if (err >= 0) | |
1623 | i = m; | |
34dd82af KS |
1624 | } else { |
1625 | err = -EINVAL; | |
1626 | } | |
1627 | if (err < 0) | |
1628 | goto out_free_dev; | |
73285082 KC |
1629 | |
1630 | lo->lo_queue = blk_alloc_queue(GFP_KERNEL); | |
1631 | if (!lo->lo_queue) | |
1632 | goto out_free_dev; | |
1633 | ||
476a4813 | 1634 | disk = lo->lo_disk = alloc_disk(1 << part_shift); |
73285082 KC |
1635 | if (!disk) |
1636 | goto out_free_queue; | |
1637 | ||
1638 | mutex_init(&lo->lo_ctl_mutex); | |
1639 | lo->lo_number = i; | |
1640 | lo->lo_thread = NULL; | |
1641 | init_waitqueue_head(&lo->lo_event); | |
1642 | spin_lock_init(&lo->lo_lock); | |
1643 | disk->major = LOOP_MAJOR; | |
476a4813 | 1644 | disk->first_minor = i << part_shift; |
73285082 KC |
1645 | disk->fops = &lo_fops; |
1646 | disk->private_data = lo; | |
1647 | disk->queue = lo->lo_queue; | |
1648 | sprintf(disk->disk_name, "loop%d", i); | |
34dd82af KS |
1649 | add_disk(disk); |
1650 | *l = lo; | |
1651 | return lo->lo_number; | |
73285082 KC |
1652 | |
1653 | out_free_queue: | |
1654 | blk_cleanup_queue(lo->lo_queue); | |
1655 | out_free_dev: | |
1656 | kfree(lo); | |
1657 | out: | |
34dd82af | 1658 | return err; |
73285082 KC |
1659 | } |
1660 | ||
34dd82af | 1661 | static void loop_remove(struct loop_device *lo) |
1da177e4 | 1662 | { |
34dd82af | 1663 | del_gendisk(lo->lo_disk); |
73285082 KC |
1664 | blk_cleanup_queue(lo->lo_queue); |
1665 | put_disk(lo->lo_disk); | |
73285082 KC |
1666 | kfree(lo); |
1667 | } | |
1da177e4 | 1668 | |
770fe30a KS |
1669 | static int find_free_cb(int id, void *ptr, void *data) |
1670 | { | |
1671 | struct loop_device *lo = ptr; | |
1672 | struct loop_device **l = data; | |
1673 | ||
1674 | if (lo->lo_state == Lo_unbound) { | |
1675 | *l = lo; | |
1676 | return 1; | |
1677 | } | |
1678 | return 0; | |
1679 | } | |
1680 | ||
34dd82af | 1681 | static int loop_lookup(struct loop_device **l, int i) |
a47653fc KC |
1682 | { |
1683 | struct loop_device *lo; | |
34dd82af | 1684 | int ret = -ENODEV; |
a47653fc | 1685 | |
770fe30a KS |
1686 | if (i < 0) { |
1687 | int err; | |
1688 | ||
1689 | err = idr_for_each(&loop_index_idr, &find_free_cb, &lo); | |
1690 | if (err == 1) { | |
1691 | *l = lo; | |
1692 | ret = lo->lo_number; | |
1693 | } | |
1694 | goto out; | |
1695 | } | |
1696 | ||
1697 | /* lookup and return a specific i */ | |
34dd82af | 1698 | lo = idr_find(&loop_index_idr, i); |
a47653fc | 1699 | if (lo) { |
34dd82af KS |
1700 | *l = lo; |
1701 | ret = lo->lo_number; | |
a47653fc | 1702 | } |
770fe30a | 1703 | out: |
34dd82af | 1704 | return ret; |
a47653fc KC |
1705 | } |
1706 | ||
73285082 KC |
1707 | static struct kobject *loop_probe(dev_t dev, int *part, void *data) |
1708 | { | |
705962cc | 1709 | struct loop_device *lo; |
07002e99 | 1710 | struct kobject *kobj; |
34dd82af | 1711 | int err; |
73285082 | 1712 | |
34dd82af KS |
1713 | mutex_lock(&loop_index_mutex); |
1714 | err = loop_lookup(&lo, MINOR(dev) >> part_shift); | |
1715 | if (err < 0) | |
1716 | err = loop_add(&lo, MINOR(dev) >> part_shift); | |
1717 | if (err < 0) | |
1718 | kobj = ERR_PTR(err); | |
1719 | else | |
1720 | kobj = get_disk(lo->lo_disk); | |
1721 | mutex_unlock(&loop_index_mutex); | |
73285082 KC |
1722 | |
1723 | *part = 0; | |
07002e99 | 1724 | return kobj; |
73285082 KC |
1725 | } |
1726 | ||
770fe30a KS |
1727 | static long loop_control_ioctl(struct file *file, unsigned int cmd, |
1728 | unsigned long parm) | |
1729 | { | |
1730 | struct loop_device *lo; | |
1731 | int ret = -ENOSYS; | |
1732 | ||
1733 | mutex_lock(&loop_index_mutex); | |
1734 | switch (cmd) { | |
1735 | case LOOP_CTL_ADD: | |
1736 | ret = loop_lookup(&lo, parm); | |
1737 | if (ret >= 0) { | |
1738 | ret = -EEXIST; | |
1739 | break; | |
1740 | } | |
1741 | ret = loop_add(&lo, parm); | |
1742 | break; | |
1743 | case LOOP_CTL_REMOVE: | |
1744 | ret = loop_lookup(&lo, parm); | |
1745 | if (ret < 0) | |
1746 | break; | |
1747 | mutex_lock(&lo->lo_ctl_mutex); | |
1748 | if (lo->lo_state != Lo_unbound) { | |
1749 | ret = -EBUSY; | |
1750 | mutex_unlock(&lo->lo_ctl_mutex); | |
1751 | break; | |
1752 | } | |
1753 | if (lo->lo_refcnt > 0) { | |
1754 | ret = -EBUSY; | |
1755 | mutex_unlock(&lo->lo_ctl_mutex); | |
1756 | break; | |
1757 | } | |
1758 | lo->lo_disk->private_data = NULL; | |
1759 | mutex_unlock(&lo->lo_ctl_mutex); | |
1760 | idr_remove(&loop_index_idr, lo->lo_number); | |
1761 | loop_remove(lo); | |
1762 | break; | |
1763 | case LOOP_CTL_GET_FREE: | |
1764 | ret = loop_lookup(&lo, -1); | |
1765 | if (ret >= 0) | |
1766 | break; | |
1767 | ret = loop_add(&lo, -1); | |
1768 | } | |
1769 | mutex_unlock(&loop_index_mutex); | |
1770 | ||
1771 | return ret; | |
1772 | } | |
1773 | ||
1774 | static const struct file_operations loop_ctl_fops = { | |
1775 | .open = nonseekable_open, | |
1776 | .unlocked_ioctl = loop_control_ioctl, | |
1777 | .compat_ioctl = loop_control_ioctl, | |
1778 | .owner = THIS_MODULE, | |
1779 | .llseek = noop_llseek, | |
1780 | }; | |
1781 | ||
1782 | static struct miscdevice loop_misc = { | |
1783 | .minor = LOOP_CTRL_MINOR, | |
1784 | .name = "loop-control", | |
1785 | .fops = &loop_ctl_fops, | |
1786 | }; | |
1787 | ||
1788 | MODULE_ALIAS_MISCDEV(LOOP_CTRL_MINOR); | |
1789 | MODULE_ALIAS("devname:loop-control"); | |
1790 | ||
73285082 KC |
1791 | static int __init loop_init(void) |
1792 | { | |
a47653fc KC |
1793 | int i, nr; |
1794 | unsigned long range; | |
34dd82af | 1795 | struct loop_device *lo; |
770fe30a | 1796 | int err; |
a47653fc | 1797 | |
770fe30a KS |
1798 | err = misc_register(&loop_misc); |
1799 | if (err < 0) | |
1800 | return err; | |
1801 | ||
476a4813 | 1802 | part_shift = 0; |
ac04fee0 | 1803 | if (max_part > 0) { |
476a4813 LV |
1804 | part_shift = fls(max_part); |
1805 | ||
ac04fee0 NK |
1806 | /* |
1807 | * Adjust max_part according to part_shift as it is exported | |
1808 | * to user space so that user can decide correct minor number | |
1809 | * if [s]he want to create more devices. | |
1810 | * | |
1811 | * Note that -1 is required because partition 0 is reserved | |
1812 | * for the whole disk. | |
1813 | */ | |
1814 | max_part = (1UL << part_shift) - 1; | |
1815 | } | |
1816 | ||
78f4bb36 NK |
1817 | if ((1UL << part_shift) > DISK_MAX_PARTS) |
1818 | return -EINVAL; | |
1819 | ||
476a4813 | 1820 | if (max_loop > 1UL << (MINORBITS - part_shift)) |
a47653fc | 1821 | return -EINVAL; |
1da177e4 | 1822 | |
d134b00b KS |
1823 | /* |
1824 | * If max_loop is specified, create that many devices upfront. | |
1825 | * This also becomes a hard limit. If max_loop is not specified, | |
1826 | * create CONFIG_BLK_DEV_LOOP_MIN_COUNT loop devices at module | |
1827 | * init time. Loop devices can be requested on-demand with the | |
1828 | * /dev/loop-control interface, or be instantiated by accessing | |
1829 | * a 'dead' device node. | |
1830 | */ | |
73285082 | 1831 | if (max_loop) { |
a47653fc | 1832 | nr = max_loop; |
a1c15c59 | 1833 | range = max_loop << part_shift; |
a47653fc | 1834 | } else { |
d134b00b | 1835 | nr = CONFIG_BLK_DEV_LOOP_MIN_COUNT; |
a1c15c59 | 1836 | range = 1UL << MINORBITS; |
a47653fc KC |
1837 | } |
1838 | ||
1839 | if (register_blkdev(LOOP_MAJOR, "loop")) | |
1840 | return -EIO; | |
1da177e4 | 1841 | |
a47653fc KC |
1842 | blk_register_region(MKDEV(LOOP_MAJOR, 0), range, |
1843 | THIS_MODULE, loop_probe, NULL, NULL); | |
1844 | ||
d134b00b | 1845 | /* pre-create number of devices given by config or max_loop */ |
34dd82af KS |
1846 | mutex_lock(&loop_index_mutex); |
1847 | for (i = 0; i < nr; i++) | |
1848 | loop_add(&lo, i); | |
1849 | mutex_unlock(&loop_index_mutex); | |
1850 | ||
73285082 | 1851 | printk(KERN_INFO "loop: module loaded\n"); |
1da177e4 | 1852 | return 0; |
34dd82af | 1853 | } |
a47653fc | 1854 | |
34dd82af KS |
1855 | static int loop_exit_cb(int id, void *ptr, void *data) |
1856 | { | |
1857 | struct loop_device *lo = ptr; | |
a47653fc | 1858 | |
34dd82af KS |
1859 | loop_remove(lo); |
1860 | return 0; | |
1da177e4 LT |
1861 | } |
1862 | ||
73285082 | 1863 | static void __exit loop_exit(void) |
1da177e4 | 1864 | { |
a47653fc | 1865 | unsigned long range; |
1da177e4 | 1866 | |
a1c15c59 | 1867 | range = max_loop ? max_loop << part_shift : 1UL << MINORBITS; |
a47653fc | 1868 | |
34dd82af KS |
1869 | idr_for_each(&loop_index_idr, &loop_exit_cb, NULL); |
1870 | idr_remove_all(&loop_index_idr); | |
1871 | idr_destroy(&loop_index_idr); | |
73285082 | 1872 | |
a47653fc | 1873 | blk_unregister_region(MKDEV(LOOP_MAJOR, 0), range); |
00d59405 | 1874 | unregister_blkdev(LOOP_MAJOR, "loop"); |
770fe30a KS |
1875 | |
1876 | misc_deregister(&loop_misc); | |
1da177e4 LT |
1877 | } |
1878 | ||
1879 | module_init(loop_init); | |
1880 | module_exit(loop_exit); | |
1881 | ||
1882 | #ifndef MODULE | |
1883 | static int __init max_loop_setup(char *str) | |
1884 | { | |
1885 | max_loop = simple_strtol(str, NULL, 0); | |
1886 | return 1; | |
1887 | } | |
1888 | ||
1889 | __setup("max_loop=", max_loop_setup); | |
1890 | #endif |