blob: 80dedde0de73df093123835d99dd37562a1dbba4
1 | /* |
2 | * bio-integrity.c - bio data integrity extensions |
3 | * |
4 | * Copyright (C) 2007, 2008, 2009 Oracle Corporation |
5 | * Written by: Martin K. Petersen <martin.petersen@oracle.com> |
6 | * |
7 | * This program is free software; you can redistribute it and/or |
8 | * modify it under the terms of the GNU General Public License version |
9 | * 2 as published by the Free Software Foundation. |
10 | * |
11 | * This program is distributed in the hope that it will be useful, but |
12 | * WITHOUT ANY WARRANTY; without even the implied warranty of |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
14 | * General Public License for more details. |
15 | * |
16 | * You should have received a copy of the GNU General Public License |
17 | * along with this program; see the file COPYING. If not, write to |
18 | * the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, |
19 | * USA. |
20 | * |
21 | */ |
22 | |
23 | #include <linux/blkdev.h> |
24 | #include <linux/mempool.h> |
25 | #include <linux/export.h> |
26 | #include <linux/bio.h> |
27 | #include <linux/workqueue.h> |
28 | #include <linux/slab.h> |
29 | #include "blk.h" |
30 | |
31 | #define BIP_INLINE_VECS 4 |
32 | |
33 | static struct kmem_cache *bip_slab; |
34 | static struct workqueue_struct *kintegrityd_wq; |
35 | |
36 | void blk_flush_integrity(void) |
37 | { |
38 | flush_workqueue(kintegrityd_wq); |
39 | } |
40 | |
41 | /** |
42 | * bio_integrity_alloc - Allocate integrity payload and attach it to bio |
43 | * @bio: bio to attach integrity metadata to |
44 | * @gfp_mask: Memory allocation mask |
45 | * @nr_vecs: Number of integrity metadata scatter-gather elements |
46 | * |
47 | * Description: This function prepares a bio for attaching integrity |
48 | * metadata. nr_vecs specifies the maximum number of pages containing |
49 | * integrity metadata that can be attached. |
50 | */ |
51 | struct bio_integrity_payload *bio_integrity_alloc(struct bio *bio, |
52 | gfp_t gfp_mask, |
53 | unsigned int nr_vecs) |
54 | { |
55 | struct bio_integrity_payload *bip; |
56 | struct bio_set *bs = bio->bi_pool; |
57 | unsigned inline_vecs; |
58 | |
59 | if (!bs || !bs->bio_integrity_pool) { |
60 | bip = kmalloc(sizeof(struct bio_integrity_payload) + |
61 | sizeof(struct bio_vec) * nr_vecs, gfp_mask); |
62 | inline_vecs = nr_vecs; |
63 | } else { |
64 | bip = mempool_alloc(bs->bio_integrity_pool, gfp_mask); |
65 | inline_vecs = BIP_INLINE_VECS; |
66 | } |
67 | |
68 | if (unlikely(!bip)) |
69 | return ERR_PTR(-ENOMEM); |
70 | |
71 | memset(bip, 0, sizeof(*bip)); |
72 | |
73 | if (nr_vecs > inline_vecs) { |
74 | unsigned long idx = 0; |
75 | |
76 | bip->bip_vec = bvec_alloc(gfp_mask, nr_vecs, &idx, |
77 | bs->bvec_integrity_pool); |
78 | if (!bip->bip_vec) |
79 | goto err; |
80 | bip->bip_max_vcnt = bvec_nr_vecs(idx); |
81 | bip->bip_slab = idx; |
82 | } else { |
83 | bip->bip_vec = bip->bip_inline_vecs; |
84 | bip->bip_max_vcnt = inline_vecs; |
85 | } |
86 | |
87 | bip->bip_bio = bio; |
88 | bio->bi_integrity = bip; |
89 | bio->bi_opf |= REQ_INTEGRITY; |
90 | |
91 | return bip; |
92 | err: |
93 | mempool_free(bip, bs->bio_integrity_pool); |
94 | return ERR_PTR(-ENOMEM); |
95 | } |
96 | EXPORT_SYMBOL(bio_integrity_alloc); |
97 | |
98 | /** |
99 | * bio_integrity_free - Free bio integrity payload |
100 | * @bio: bio containing bip to be freed |
101 | * |
102 | * Description: Used to free the integrity portion of a bio. Usually |
103 | * called from bio_free(). |
104 | */ |
105 | void bio_integrity_free(struct bio *bio) |
106 | { |
107 | struct bio_integrity_payload *bip = bio_integrity(bio); |
108 | struct bio_set *bs = bio->bi_pool; |
109 | |
110 | if (bip->bip_flags & BIP_BLOCK_INTEGRITY) |
111 | kfree(page_address(bip->bip_vec->bv_page) + |
112 | bip->bip_vec->bv_offset); |
113 | |
114 | if (bs && bs->bio_integrity_pool) { |
115 | bvec_free(bs->bvec_integrity_pool, bip->bip_vec, bip->bip_slab); |
116 | |
117 | mempool_free(bip, bs->bio_integrity_pool); |
118 | } else { |
119 | kfree(bip); |
120 | } |
121 | |
122 | bio->bi_integrity = NULL; |
123 | } |
124 | EXPORT_SYMBOL(bio_integrity_free); |
125 | |
126 | /** |
127 | * bio_integrity_add_page - Attach integrity metadata |
128 | * @bio: bio to update |
129 | * @page: page containing integrity metadata |
130 | * @len: number of bytes of integrity metadata in page |
131 | * @offset: start offset within page |
132 | * |
133 | * Description: Attach a page containing integrity metadata to bio. |
134 | */ |
135 | int bio_integrity_add_page(struct bio *bio, struct page *page, |
136 | unsigned int len, unsigned int offset) |
137 | { |
138 | struct bio_integrity_payload *bip = bio_integrity(bio); |
139 | struct bio_vec *iv; |
140 | |
141 | if (bip->bip_vcnt >= bip->bip_max_vcnt) { |
142 | printk(KERN_ERR "%s: bip_vec full\n", __func__); |
143 | return 0; |
144 | } |
145 | |
146 | iv = bip->bip_vec + bip->bip_vcnt; |
147 | |
148 | if (bip->bip_vcnt && |
149 | bvec_gap_to_prev(bdev_get_queue(bio->bi_bdev), |
150 | &bip->bip_vec[bip->bip_vcnt - 1], offset)) |
151 | return 0; |
152 | |
153 | iv->bv_page = page; |
154 | iv->bv_len = len; |
155 | iv->bv_offset = offset; |
156 | bip->bip_vcnt++; |
157 | |
158 | return len; |
159 | } |
160 | EXPORT_SYMBOL(bio_integrity_add_page); |
161 | |
162 | /** |
163 | * bio_integrity_enabled - Check whether integrity can be passed |
164 | * @bio: bio to check |
165 | * |
166 | * Description: Determines whether bio_integrity_prep() can be called |
167 | * on this bio or not. bio data direction and target device must be |
168 | * set prior to calling. The functions honors the write_generate and |
169 | * read_verify flags in sysfs. |
170 | */ |
171 | bool bio_integrity_enabled(struct bio *bio) |
172 | { |
173 | struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev); |
174 | |
175 | if (!bio_is_rw(bio)) |
176 | return false; |
177 | |
178 | if (!bio_sectors(bio)) |
179 | return false; |
180 | |
181 | /* Already protected? */ |
182 | if (bio_integrity(bio)) |
183 | return false; |
184 | |
185 | if (bi == NULL) |
186 | return false; |
187 | |
188 | if (bio_data_dir(bio) == READ && bi->profile->verify_fn != NULL && |
189 | (bi->flags & BLK_INTEGRITY_VERIFY)) |
190 | return true; |
191 | |
192 | if (bio_data_dir(bio) == WRITE && bi->profile->generate_fn != NULL && |
193 | (bi->flags & BLK_INTEGRITY_GENERATE)) |
194 | return true; |
195 | |
196 | return false; |
197 | } |
198 | EXPORT_SYMBOL(bio_integrity_enabled); |
199 | |
200 | /** |
201 | * bio_integrity_intervals - Return number of integrity intervals for a bio |
202 | * @bi: blk_integrity profile for device |
203 | * @sectors: Size of the bio in 512-byte sectors |
204 | * |
205 | * Description: The block layer calculates everything in 512 byte |
206 | * sectors but integrity metadata is done in terms of the data integrity |
207 | * interval size of the storage device. Convert the block layer sectors |
208 | * to the appropriate number of integrity intervals. |
209 | */ |
210 | static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi, |
211 | unsigned int sectors) |
212 | { |
213 | return sectors >> (bi->interval_exp - 9); |
214 | } |
215 | |
216 | static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi, |
217 | unsigned int sectors) |
218 | { |
219 | return bio_integrity_intervals(bi, sectors) * bi->tuple_size; |
220 | } |
221 | |
222 | /** |
223 | * bio_integrity_process - Process integrity metadata for a bio |
224 | * @bio: bio to generate/verify integrity metadata for |
225 | * @proc_fn: Pointer to the relevant processing function |
226 | */ |
227 | static int bio_integrity_process(struct bio *bio, |
228 | integrity_processing_fn *proc_fn) |
229 | { |
230 | struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev); |
231 | struct blk_integrity_iter iter; |
232 | struct bvec_iter bviter; |
233 | struct bio_vec bv; |
234 | struct bio_integrity_payload *bip = bio_integrity(bio); |
235 | unsigned int ret = 0; |
236 | void *prot_buf = page_address(bip->bip_vec->bv_page) + |
237 | bip->bip_vec->bv_offset; |
238 | |
239 | iter.disk_name = bio->bi_bdev->bd_disk->disk_name; |
240 | iter.interval = 1 << bi->interval_exp; |
241 | iter.seed = bip_get_seed(bip); |
242 | iter.prot_buf = prot_buf; |
243 | |
244 | bio_for_each_segment(bv, bio, bviter) { |
245 | void *kaddr = kmap_atomic(bv.bv_page); |
246 | |
247 | iter.data_buf = kaddr + bv.bv_offset; |
248 | iter.data_size = bv.bv_len; |
249 | |
250 | ret = proc_fn(&iter); |
251 | if (ret) { |
252 | kunmap_atomic(kaddr); |
253 | return ret; |
254 | } |
255 | |
256 | kunmap_atomic(kaddr); |
257 | } |
258 | return ret; |
259 | } |
260 | |
261 | /** |
262 | * bio_integrity_prep - Prepare bio for integrity I/O |
263 | * @bio: bio to prepare |
264 | * |
265 | * Description: Allocates a buffer for integrity metadata, maps the |
266 | * pages and attaches them to a bio. The bio must have data |
267 | * direction, target device and start sector set priot to calling. In |
268 | * the WRITE case, integrity metadata will be generated using the |
269 | * block device's integrity function. In the READ case, the buffer |
270 | * will be prepared for DMA and a suitable end_io handler set up. |
271 | */ |
272 | int bio_integrity_prep(struct bio *bio) |
273 | { |
274 | struct bio_integrity_payload *bip; |
275 | struct blk_integrity *bi; |
276 | struct request_queue *q; |
277 | void *buf; |
278 | unsigned long start, end; |
279 | unsigned int len, nr_pages; |
280 | unsigned int bytes, offset, i; |
281 | unsigned int intervals; |
282 | |
283 | bi = bdev_get_integrity(bio->bi_bdev); |
284 | q = bdev_get_queue(bio->bi_bdev); |
285 | BUG_ON(bi == NULL); |
286 | BUG_ON(bio_integrity(bio)); |
287 | |
288 | intervals = bio_integrity_intervals(bi, bio_sectors(bio)); |
289 | |
290 | /* Allocate kernel buffer for protection data */ |
291 | len = intervals * bi->tuple_size; |
292 | buf = kmalloc(len, GFP_NOIO | q->bounce_gfp); |
293 | if (unlikely(buf == NULL)) { |
294 | printk(KERN_ERR "could not allocate integrity buffer\n"); |
295 | return -ENOMEM; |
296 | } |
297 | |
298 | end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
299 | start = ((unsigned long) buf) >> PAGE_SHIFT; |
300 | nr_pages = end - start; |
301 | |
302 | /* Allocate bio integrity payload and integrity vectors */ |
303 | bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages); |
304 | if (IS_ERR(bip)) { |
305 | printk(KERN_ERR "could not allocate data integrity bioset\n"); |
306 | kfree(buf); |
307 | return PTR_ERR(bip); |
308 | } |
309 | |
310 | bip->bip_flags |= BIP_BLOCK_INTEGRITY; |
311 | bip->bip_iter.bi_size = len; |
312 | bip_set_seed(bip, bio->bi_iter.bi_sector); |
313 | |
314 | if (bi->flags & BLK_INTEGRITY_IP_CHECKSUM) |
315 | bip->bip_flags |= BIP_IP_CHECKSUM; |
316 | |
317 | /* Map it */ |
318 | offset = offset_in_page(buf); |
319 | for (i = 0 ; i < nr_pages ; i++) { |
320 | int ret; |
321 | bytes = PAGE_SIZE - offset; |
322 | |
323 | if (len <= 0) |
324 | break; |
325 | |
326 | if (bytes > len) |
327 | bytes = len; |
328 | |
329 | ret = bio_integrity_add_page(bio, virt_to_page(buf), |
330 | bytes, offset); |
331 | |
332 | if (ret == 0) |
333 | return 0; |
334 | |
335 | if (ret < bytes) |
336 | break; |
337 | |
338 | buf += bytes; |
339 | len -= bytes; |
340 | offset = 0; |
341 | } |
342 | |
343 | /* Install custom I/O completion handler if read verify is enabled */ |
344 | if (bio_data_dir(bio) == READ) { |
345 | bip->bip_end_io = bio->bi_end_io; |
346 | bio->bi_end_io = bio_integrity_endio; |
347 | } |
348 | |
349 | /* Auto-generate integrity metadata if this is a write */ |
350 | if (bio_data_dir(bio) == WRITE) |
351 | bio_integrity_process(bio, bi->profile->generate_fn); |
352 | |
353 | return 0; |
354 | } |
355 | EXPORT_SYMBOL(bio_integrity_prep); |
356 | |
357 | /** |
358 | * bio_integrity_verify_fn - Integrity I/O completion worker |
359 | * @work: Work struct stored in bio to be verified |
360 | * |
361 | * Description: This workqueue function is called to complete a READ |
362 | * request. The function verifies the transferred integrity metadata |
363 | * and then calls the original bio end_io function. |
364 | */ |
365 | static void bio_integrity_verify_fn(struct work_struct *work) |
366 | { |
367 | struct bio_integrity_payload *bip = |
368 | container_of(work, struct bio_integrity_payload, bip_work); |
369 | struct bio *bio = bip->bip_bio; |
370 | struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev); |
371 | |
372 | bio->bi_error = bio_integrity_process(bio, bi->profile->verify_fn); |
373 | |
374 | /* Restore original bio completion handler */ |
375 | bio->bi_end_io = bip->bip_end_io; |
376 | bio_endio(bio); |
377 | } |
378 | |
379 | /** |
380 | * bio_integrity_endio - Integrity I/O completion function |
381 | * @bio: Protected bio |
382 | * @error: Pointer to errno |
383 | * |
384 | * Description: Completion for integrity I/O |
385 | * |
386 | * Normally I/O completion is done in interrupt context. However, |
387 | * verifying I/O integrity is a time-consuming task which must be run |
388 | * in process context. This function postpones completion |
389 | * accordingly. |
390 | */ |
391 | void bio_integrity_endio(struct bio *bio) |
392 | { |
393 | struct bio_integrity_payload *bip = bio_integrity(bio); |
394 | |
395 | BUG_ON(bip->bip_bio != bio); |
396 | |
397 | /* In case of an I/O error there is no point in verifying the |
398 | * integrity metadata. Restore original bio end_io handler |
399 | * and run it. |
400 | */ |
401 | if (bio->bi_error) { |
402 | bio->bi_end_io = bip->bip_end_io; |
403 | bio_endio(bio); |
404 | |
405 | return; |
406 | } |
407 | |
408 | INIT_WORK(&bip->bip_work, bio_integrity_verify_fn); |
409 | queue_work(kintegrityd_wq, &bip->bip_work); |
410 | } |
411 | EXPORT_SYMBOL(bio_integrity_endio); |
412 | |
413 | /** |
414 | * bio_integrity_advance - Advance integrity vector |
415 | * @bio: bio whose integrity vector to update |
416 | * @bytes_done: number of data bytes that have been completed |
417 | * |
418 | * Description: This function calculates how many integrity bytes the |
419 | * number of completed data bytes correspond to and advances the |
420 | * integrity vector accordingly. |
421 | */ |
422 | void bio_integrity_advance(struct bio *bio, unsigned int bytes_done) |
423 | { |
424 | struct bio_integrity_payload *bip = bio_integrity(bio); |
425 | struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev); |
426 | unsigned bytes = bio_integrity_bytes(bi, bytes_done >> 9); |
427 | |
428 | bvec_iter_advance(bip->bip_vec, &bip->bip_iter, bytes); |
429 | } |
430 | EXPORT_SYMBOL(bio_integrity_advance); |
431 | |
432 | /** |
433 | * bio_integrity_trim - Trim integrity vector |
434 | * @bio: bio whose integrity vector to update |
435 | * @offset: offset to first data sector |
436 | * @sectors: number of data sectors |
437 | * |
438 | * Description: Used to trim the integrity vector in a cloned bio. |
439 | * The ivec will be advanced corresponding to 'offset' data sectors |
440 | * and the length will be truncated corresponding to 'len' data |
441 | * sectors. |
442 | */ |
443 | void bio_integrity_trim(struct bio *bio, unsigned int offset, |
444 | unsigned int sectors) |
445 | { |
446 | struct bio_integrity_payload *bip = bio_integrity(bio); |
447 | struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev); |
448 | |
449 | bio_integrity_advance(bio, offset << 9); |
450 | bip->bip_iter.bi_size = bio_integrity_bytes(bi, sectors); |
451 | } |
452 | EXPORT_SYMBOL(bio_integrity_trim); |
453 | |
454 | /** |
455 | * bio_integrity_clone - Callback for cloning bios with integrity metadata |
456 | * @bio: New bio |
457 | * @bio_src: Original bio |
458 | * @gfp_mask: Memory allocation mask |
459 | * |
460 | * Description: Called to allocate a bip when cloning a bio |
461 | */ |
462 | int bio_integrity_clone(struct bio *bio, struct bio *bio_src, |
463 | gfp_t gfp_mask) |
464 | { |
465 | struct bio_integrity_payload *bip_src = bio_integrity(bio_src); |
466 | struct bio_integrity_payload *bip; |
467 | |
468 | BUG_ON(bip_src == NULL); |
469 | |
470 | bip = bio_integrity_alloc(bio, gfp_mask, bip_src->bip_vcnt); |
471 | if (IS_ERR(bip)) |
472 | return PTR_ERR(bip); |
473 | |
474 | memcpy(bip->bip_vec, bip_src->bip_vec, |
475 | bip_src->bip_vcnt * sizeof(struct bio_vec)); |
476 | |
477 | bip->bip_vcnt = bip_src->bip_vcnt; |
478 | bip->bip_iter = bip_src->bip_iter; |
479 | |
480 | return 0; |
481 | } |
482 | EXPORT_SYMBOL(bio_integrity_clone); |
483 | |
484 | int bioset_integrity_create(struct bio_set *bs, int pool_size) |
485 | { |
486 | if (bs->bio_integrity_pool) |
487 | return 0; |
488 | |
489 | bs->bio_integrity_pool = mempool_create_slab_pool(pool_size, bip_slab); |
490 | if (!bs->bio_integrity_pool) |
491 | return -1; |
492 | |
493 | bs->bvec_integrity_pool = biovec_create_pool(pool_size); |
494 | if (!bs->bvec_integrity_pool) { |
495 | mempool_destroy(bs->bio_integrity_pool); |
496 | return -1; |
497 | } |
498 | |
499 | return 0; |
500 | } |
501 | EXPORT_SYMBOL(bioset_integrity_create); |
502 | |
503 | void bioset_integrity_free(struct bio_set *bs) |
504 | { |
505 | if (bs->bio_integrity_pool) |
506 | mempool_destroy(bs->bio_integrity_pool); |
507 | |
508 | if (bs->bvec_integrity_pool) |
509 | mempool_destroy(bs->bvec_integrity_pool); |
510 | } |
511 | EXPORT_SYMBOL(bioset_integrity_free); |
512 | |
513 | void __init bio_integrity_init(void) |
514 | { |
515 | /* |
516 | * kintegrityd won't block much but may burn a lot of CPU cycles. |
517 | * Make it highpri CPU intensive wq with max concurrency of 1. |
518 | */ |
519 | kintegrityd_wq = alloc_workqueue("kintegrityd", WQ_MEM_RECLAIM | |
520 | WQ_HIGHPRI | WQ_CPU_INTENSIVE, 1); |
521 | if (!kintegrityd_wq) |
522 | panic("Failed to create kintegrityd\n"); |
523 | |
524 | bip_slab = kmem_cache_create("bio_integrity_payload", |
525 | sizeof(struct bio_integrity_payload) + |
526 | sizeof(struct bio_vec) * BIP_INLINE_VECS, |
527 | 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
528 | } |
529 |