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
33static struct kmem_cache *bip_slab;
34static struct workqueue_struct *kintegrityd_wq;
35
36void 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 */
51struct 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 || !mempool_initialized(&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;
92err:
93 mempool_free(bip, &bs->bio_integrity_pool);
94 return ERR_PTR(-ENOMEM);
95}
96EXPORT_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 */
105static 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 && mempool_initialized(&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 bio->bi_opf &= ~REQ_INTEGRITY;
124}
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 */
135int 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(bio->bi_disk->queue,
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}
160EXPORT_SYMBOL(bio_integrity_add_page);
161
162/**
163 * bio_integrity_process - Process integrity metadata for a bio
164 * @bio: bio to generate/verify integrity metadata for
165 * @proc_iter: iterator to process
166 * @proc_fn: Pointer to the relevant processing function
167 */
168static blk_status_t bio_integrity_process(struct bio *bio,
169 struct bvec_iter *proc_iter, integrity_processing_fn *proc_fn)
170{
171 struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
172 struct blk_integrity_iter iter;
173 struct bvec_iter bviter;
174 struct bio_vec bv;
175 struct bio_integrity_payload *bip = bio_integrity(bio);
176 blk_status_t ret = BLK_STS_OK;
177 void *prot_buf = page_address(bip->bip_vec->bv_page) +
178 bip->bip_vec->bv_offset;
179
180 iter.disk_name = bio->bi_disk->disk_name;
181 iter.interval = 1 << bi->interval_exp;
182 iter.seed = proc_iter->bi_sector;
183 iter.prot_buf = prot_buf;
184
185 __bio_for_each_segment(bv, bio, bviter, *proc_iter) {
186 void *kaddr = kmap_atomic(bv.bv_page);
187
188 iter.data_buf = kaddr + bv.bv_offset;
189 iter.data_size = bv.bv_len;
190
191 ret = proc_fn(&iter);
192 if (ret) {
193 kunmap_atomic(kaddr);
194 return ret;
195 }
196
197 kunmap_atomic(kaddr);
198 }
199 return ret;
200}
201
202/**
203 * bio_integrity_prep - Prepare bio for integrity I/O
204 * @bio: bio to prepare
205 *
206 * Description: Checks if the bio already has an integrity payload attached.
207 * If it does, the payload has been generated by another kernel subsystem,
208 * and we just pass it through. Otherwise allocates integrity payload.
209 * The bio must have data direction, target device and start sector set priot
210 * to calling. In the WRITE case, integrity metadata will be generated using
211 * the block device's integrity function. In the READ case, the buffer
212 * will be prepared for DMA and a suitable end_io handler set up.
213 */
214bool bio_integrity_prep(struct bio *bio)
215{
216 struct bio_integrity_payload *bip;
217 struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
218 struct request_queue *q = bio->bi_disk->queue;
219 void *buf;
220 unsigned long start, end;
221 unsigned int len, nr_pages;
222 unsigned int bytes, offset, i;
223 unsigned int intervals;
224 blk_status_t status;
225
226 if (!bi)
227 return true;
228
229 if (bio_op(bio) != REQ_OP_READ && bio_op(bio) != REQ_OP_WRITE)
230 return true;
231
232 if (!bio_sectors(bio))
233 return true;
234
235 /* Already protected? */
236 if (bio_integrity(bio))
237 return true;
238
239 if (bio_data_dir(bio) == READ) {
240 if (!bi->profile->verify_fn ||
241 !(bi->flags & BLK_INTEGRITY_VERIFY))
242 return true;
243 } else {
244 if (!bi->profile->generate_fn ||
245 !(bi->flags & BLK_INTEGRITY_GENERATE))
246 return true;
247 }
248 intervals = bio_integrity_intervals(bi, bio_sectors(bio));
249
250 /* Allocate kernel buffer for protection data */
251 len = intervals * bi->tuple_size;
252 buf = kmalloc(len, GFP_NOIO | q->bounce_gfp);
253 status = BLK_STS_RESOURCE;
254 if (unlikely(buf == NULL)) {
255 printk(KERN_ERR "could not allocate integrity buffer\n");
256 goto err_end_io;
257 }
258
259 end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
260 start = ((unsigned long) buf) >> PAGE_SHIFT;
261 nr_pages = end - start;
262
263 /* Allocate bio integrity payload and integrity vectors */
264 bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages);
265 if (IS_ERR(bip)) {
266 printk(KERN_ERR "could not allocate data integrity bioset\n");
267 kfree(buf);
268 status = BLK_STS_RESOURCE;
269 goto err_end_io;
270 }
271
272 bip->bip_flags |= BIP_BLOCK_INTEGRITY;
273 bip->bip_iter.bi_size = len;
274 bip_set_seed(bip, bio->bi_iter.bi_sector);
275
276 if (bi->flags & BLK_INTEGRITY_IP_CHECKSUM)
277 bip->bip_flags |= BIP_IP_CHECKSUM;
278
279 /* Map it */
280 offset = offset_in_page(buf);
281 for (i = 0 ; i < nr_pages ; i++) {
282 int ret;
283 bytes = PAGE_SIZE - offset;
284
285 if (len <= 0)
286 break;
287
288 if (bytes > len)
289 bytes = len;
290
291 ret = bio_integrity_add_page(bio, virt_to_page(buf),
292 bytes, offset);
293
294 if (ret == 0)
295 return false;
296
297 if (ret < bytes)
298 break;
299
300 buf += bytes;
301 len -= bytes;
302 offset = 0;
303 }
304
305 /* Auto-generate integrity metadata if this is a write */
306 if (bio_data_dir(bio) == WRITE) {
307 bio_integrity_process(bio, &bio->bi_iter,
308 bi->profile->generate_fn);
309 } else {
310 bip->bio_iter = bio->bi_iter;
311 }
312 return true;
313
314err_end_io:
315 bio->bi_status = status;
316 bio_endio(bio);
317 return false;
318
319}
320EXPORT_SYMBOL(bio_integrity_prep);
321
322/**
323 * bio_integrity_verify_fn - Integrity I/O completion worker
324 * @work: Work struct stored in bio to be verified
325 *
326 * Description: This workqueue function is called to complete a READ
327 * request. The function verifies the transferred integrity metadata
328 * and then calls the original bio end_io function.
329 */
330static void bio_integrity_verify_fn(struct work_struct *work)
331{
332 struct bio_integrity_payload *bip =
333 container_of(work, struct bio_integrity_payload, bip_work);
334 struct bio *bio = bip->bip_bio;
335 struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
336
337 /*
338 * At the moment verify is called bio's iterator was advanced
339 * during split and completion, we need to rewind iterator to
340 * it's original position.
341 */
342 bio->bi_status = bio_integrity_process(bio, &bip->bio_iter,
343 bi->profile->verify_fn);
344 bio_integrity_free(bio);
345 bio_endio(bio);
346}
347
348/**
349 * __bio_integrity_endio - Integrity I/O completion function
350 * @bio: Protected bio
351 *
352 * Description: Completion for integrity I/O
353 *
354 * Normally I/O completion is done in interrupt context. However,
355 * verifying I/O integrity is a time-consuming task which must be run
356 * in process context. This function postpones completion
357 * accordingly.
358 */
359bool __bio_integrity_endio(struct bio *bio)
360{
361 struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
362 struct bio_integrity_payload *bip = bio_integrity(bio);
363
364 if (bio_op(bio) == REQ_OP_READ && !bio->bi_status &&
365 (bip->bip_flags & BIP_BLOCK_INTEGRITY) && bi->profile->verify_fn) {
366 INIT_WORK(&bip->bip_work, bio_integrity_verify_fn);
367 queue_work(kintegrityd_wq, &bip->bip_work);
368 return false;
369 }
370
371 bio_integrity_free(bio);
372 return true;
373}
374
375/**
376 * bio_integrity_advance - Advance integrity vector
377 * @bio: bio whose integrity vector to update
378 * @bytes_done: number of data bytes that have been completed
379 *
380 * Description: This function calculates how many integrity bytes the
381 * number of completed data bytes correspond to and advances the
382 * integrity vector accordingly.
383 */
384void bio_integrity_advance(struct bio *bio, unsigned int bytes_done)
385{
386 struct bio_integrity_payload *bip = bio_integrity(bio);
387 struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
388 unsigned bytes = bio_integrity_bytes(bi, bytes_done >> 9);
389
390 bip->bip_iter.bi_sector += bytes_done >> 9;
391 bvec_iter_advance(bip->bip_vec, &bip->bip_iter, bytes);
392}
393
394/**
395 * bio_integrity_trim - Trim integrity vector
396 * @bio: bio whose integrity vector to update
397 *
398 * Description: Used to trim the integrity vector in a cloned bio.
399 */
400void bio_integrity_trim(struct bio *bio)
401{
402 struct bio_integrity_payload *bip = bio_integrity(bio);
403 struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
404
405 bip->bip_iter.bi_size = bio_integrity_bytes(bi, bio_sectors(bio));
406}
407EXPORT_SYMBOL(bio_integrity_trim);
408
409/**
410 * bio_integrity_clone - Callback for cloning bios with integrity metadata
411 * @bio: New bio
412 * @bio_src: Original bio
413 * @gfp_mask: Memory allocation mask
414 *
415 * Description: Called to allocate a bip when cloning a bio
416 */
417int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
418 gfp_t gfp_mask)
419{
420 struct bio_integrity_payload *bip_src = bio_integrity(bio_src);
421 struct bio_integrity_payload *bip;
422
423 BUG_ON(bip_src == NULL);
424
425 bip = bio_integrity_alloc(bio, gfp_mask, bip_src->bip_vcnt);
426 if (IS_ERR(bip))
427 return PTR_ERR(bip);
428
429 memcpy(bip->bip_vec, bip_src->bip_vec,
430 bip_src->bip_vcnt * sizeof(struct bio_vec));
431
432 bip->bip_vcnt = bip_src->bip_vcnt;
433 bip->bip_iter = bip_src->bip_iter;
434
435 return 0;
436}
437EXPORT_SYMBOL(bio_integrity_clone);
438
439int bioset_integrity_create(struct bio_set *bs, int pool_size)
440{
441 if (mempool_initialized(&bs->bio_integrity_pool))
442 return 0;
443
444 if (mempool_init_slab_pool(&bs->bio_integrity_pool,
445 pool_size, bip_slab))
446 return -1;
447
448 if (biovec_init_pool(&bs->bvec_integrity_pool, pool_size)) {
449 mempool_exit(&bs->bio_integrity_pool);
450 return -1;
451 }
452
453 return 0;
454}
455EXPORT_SYMBOL(bioset_integrity_create);
456
457void bioset_integrity_free(struct bio_set *bs)
458{
459 mempool_exit(&bs->bio_integrity_pool);
460 mempool_exit(&bs->bvec_integrity_pool);
461}
462
463void __init bio_integrity_init(void)
464{
465 /*
466 * kintegrityd won't block much but may burn a lot of CPU cycles.
467 * Make it highpri CPU intensive wq with max concurrency of 1.
468 */
469 kintegrityd_wq = alloc_workqueue("kintegrityd", WQ_MEM_RECLAIM |
470 WQ_HIGHPRI | WQ_CPU_INTENSIVE, 1);
471 if (!kintegrityd_wq)
472 panic("Failed to create kintegrityd\n");
473
474 bip_slab = kmem_cache_create("bio_integrity_payload",
475 sizeof(struct bio_integrity_payload) +
476 sizeof(struct bio_vec) * BIP_INLINE_VECS,
477 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
478}
479