1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * bio-integrity.c - bio data integrity extensions
4 *
5 * Copyright (C) 2007, 2008, 2009 Oracle Corporation
6 * Written by: Martin K. Petersen <martin.petersen@oracle.com>
7 */
8
9 #include <linux/blkdev.h>
10 #include <linux/mempool.h>
11 #include <linux/export.h>
12 #include <linux/bio.h>
13 #include <linux/workqueue.h>
14 #include <linux/slab.h>
15 #include "blk.h"
16
17 #define BIP_INLINE_VECS 4
18
19 static struct kmem_cache *bip_slab;
20 static struct workqueue_struct *kintegrityd_wq;
21
blk_flush_integrity(void)22 void blk_flush_integrity(void)
23 {
24 flush_workqueue(kintegrityd_wq);
25 }
26
__bio_integrity_free(struct bio_set * bs,struct bio_integrity_payload * bip)27 static void __bio_integrity_free(struct bio_set *bs,
28 struct bio_integrity_payload *bip)
29 {
30 if (bs && mempool_initialized(&bs->bio_integrity_pool)) {
31 if (bip->bip_vec)
32 bvec_free(&bs->bvec_integrity_pool, bip->bip_vec,
33 bip->bip_slab);
34 mempool_free(bip, &bs->bio_integrity_pool);
35 } else {
36 kfree(bip);
37 }
38 }
39
40 /**
41 * bio_integrity_alloc - Allocate integrity payload and attach it to bio
42 * @bio: bio to attach integrity metadata to
43 * @gfp_mask: Memory allocation mask
44 * @nr_vecs: Number of integrity metadata scatter-gather elements
45 *
46 * Description: This function prepares a bio for attaching integrity
47 * metadata. nr_vecs specifies the maximum number of pages containing
48 * integrity metadata that can be attached.
49 */
bio_integrity_alloc(struct bio * bio,gfp_t gfp_mask,unsigned int nr_vecs)50 struct bio_integrity_payload *bio_integrity_alloc(struct bio *bio,
51 gfp_t gfp_mask,
52 unsigned int nr_vecs)
53 {
54 struct bio_integrity_payload *bip;
55 struct bio_set *bs = bio->bi_pool;
56 unsigned inline_vecs;
57
58 if (WARN_ON_ONCE(bio_has_crypt_ctx(bio)))
59 return ERR_PTR(-EOPNOTSUPP);
60
61 if (!bs || !mempool_initialized(&bs->bio_integrity_pool)) {
62 bip = kmalloc(struct_size(bip, bip_inline_vecs, nr_vecs), gfp_mask);
63 inline_vecs = nr_vecs;
64 } else {
65 bip = mempool_alloc(&bs->bio_integrity_pool, gfp_mask);
66 inline_vecs = BIP_INLINE_VECS;
67 }
68
69 if (unlikely(!bip))
70 return ERR_PTR(-ENOMEM);
71
72 memset(bip, 0, sizeof(*bip));
73
74 if (nr_vecs > inline_vecs) {
75 unsigned long idx = 0;
76
77 bip->bip_vec = bvec_alloc(gfp_mask, nr_vecs, &idx,
78 &bs->bvec_integrity_pool);
79 if (!bip->bip_vec)
80 goto err;
81 bip->bip_max_vcnt = bvec_nr_vecs(idx);
82 bip->bip_slab = idx;
83 } else {
84 bip->bip_vec = bip->bip_inline_vecs;
85 bip->bip_max_vcnt = inline_vecs;
86 }
87
88 bip->bip_bio = bio;
89 bio->bi_integrity = bip;
90 bio->bi_opf |= REQ_INTEGRITY;
91
92 return bip;
93 err:
94 __bio_integrity_free(bs, bip);
95 return ERR_PTR(-ENOMEM);
96 }
97 EXPORT_SYMBOL(bio_integrity_alloc);
98
99 /**
100 * bio_integrity_free - Free bio integrity payload
101 * @bio: bio containing bip to be freed
102 *
103 * Description: Used to free the integrity portion of a bio. Usually
104 * called from bio_free().
105 */
bio_integrity_free(struct bio * bio)106 void bio_integrity_free(struct bio *bio)
107 {
108 struct bio_integrity_payload *bip = bio_integrity(bio);
109 struct bio_set *bs = bio->bi_pool;
110
111 if (bip->bip_flags & BIP_BLOCK_INTEGRITY)
112 kfree(page_address(bip->bip_vec->bv_page) +
113 bip->bip_vec->bv_offset);
114
115 __bio_integrity_free(bs, bip);
116 bio->bi_integrity = NULL;
117 bio->bi_opf &= ~REQ_INTEGRITY;
118 }
119
120 /**
121 * bio_integrity_add_page - Attach integrity metadata
122 * @bio: bio to update
123 * @page: page containing integrity metadata
124 * @len: number of bytes of integrity metadata in page
125 * @offset: start offset within page
126 *
127 * Description: Attach a page containing integrity metadata to bio.
128 */
bio_integrity_add_page(struct bio * bio,struct page * page,unsigned int len,unsigned int offset)129 int bio_integrity_add_page(struct bio *bio, struct page *page,
130 unsigned int len, unsigned int offset)
131 {
132 struct bio_integrity_payload *bip = bio_integrity(bio);
133 struct bio_vec *iv;
134
135 if (bip->bip_vcnt >= bip->bip_max_vcnt) {
136 printk(KERN_ERR "%s: bip_vec full\n", __func__);
137 return 0;
138 }
139
140 iv = bip->bip_vec + bip->bip_vcnt;
141
142 if (bip->bip_vcnt &&
143 bvec_gap_to_prev(bio->bi_disk->queue,
144 &bip->bip_vec[bip->bip_vcnt - 1], offset))
145 return 0;
146
147 iv->bv_page = page;
148 iv->bv_len = len;
149 iv->bv_offset = offset;
150 bip->bip_vcnt++;
151
152 return len;
153 }
154 EXPORT_SYMBOL(bio_integrity_add_page);
155
156 /**
157 * bio_integrity_process - Process integrity metadata for a bio
158 * @bio: bio to generate/verify integrity metadata for
159 * @proc_iter: iterator to process
160 * @proc_fn: Pointer to the relevant processing function
161 */
bio_integrity_process(struct bio * bio,struct bvec_iter * proc_iter,integrity_processing_fn * proc_fn)162 static blk_status_t bio_integrity_process(struct bio *bio,
163 struct bvec_iter *proc_iter, integrity_processing_fn *proc_fn)
164 {
165 struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
166 struct blk_integrity_iter iter;
167 struct bvec_iter bviter;
168 struct bio_vec bv;
169 struct bio_integrity_payload *bip = bio_integrity(bio);
170 blk_status_t ret = BLK_STS_OK;
171 void *prot_buf = page_address(bip->bip_vec->bv_page) +
172 bip->bip_vec->bv_offset;
173
174 iter.disk_name = bio->bi_disk->disk_name;
175 iter.interval = 1 << bi->interval_exp;
176 iter.seed = proc_iter->bi_sector;
177 iter.prot_buf = prot_buf;
178
179 __bio_for_each_segment(bv, bio, bviter, *proc_iter) {
180 void *kaddr = kmap_atomic(bv.bv_page);
181
182 iter.data_buf = kaddr + bv.bv_offset;
183 iter.data_size = bv.bv_len;
184
185 ret = proc_fn(&iter);
186 if (ret) {
187 kunmap_atomic(kaddr);
188 return ret;
189 }
190
191 kunmap_atomic(kaddr);
192 }
193 return ret;
194 }
195
196 /**
197 * bio_integrity_prep - Prepare bio for integrity I/O
198 * @bio: bio to prepare
199 *
200 * Description: Checks if the bio already has an integrity payload attached.
201 * If it does, the payload has been generated by another kernel subsystem,
202 * and we just pass it through. Otherwise allocates integrity payload.
203 * The bio must have data direction, target device and start sector set priot
204 * to calling. In the WRITE case, integrity metadata will be generated using
205 * the block device's integrity function. In the READ case, the buffer
206 * will be prepared for DMA and a suitable end_io handler set up.
207 */
bio_integrity_prep(struct bio * bio)208 bool bio_integrity_prep(struct bio *bio)
209 {
210 struct bio_integrity_payload *bip;
211 struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
212 struct request_queue *q = bio->bi_disk->queue;
213 void *buf;
214 unsigned long start, end;
215 unsigned int len, nr_pages;
216 unsigned int bytes, offset, i;
217 unsigned int intervals;
218 blk_status_t status;
219
220 if (!bi)
221 return true;
222
223 if (bio_op(bio) != REQ_OP_READ && bio_op(bio) != REQ_OP_WRITE)
224 return true;
225
226 if (!bio_sectors(bio))
227 return true;
228
229 /* Already protected? */
230 if (bio_integrity(bio))
231 return true;
232
233 if (bio_data_dir(bio) == READ) {
234 if (!bi->profile->verify_fn ||
235 !(bi->flags & BLK_INTEGRITY_VERIFY))
236 return true;
237 } else {
238 if (!bi->profile->generate_fn ||
239 !(bi->flags & BLK_INTEGRITY_GENERATE))
240 return true;
241 }
242 intervals = bio_integrity_intervals(bi, bio_sectors(bio));
243
244 /* Allocate kernel buffer for protection data */
245 len = intervals * bi->tuple_size;
246 buf = kmalloc(len, GFP_NOIO | q->bounce_gfp);
247 status = BLK_STS_RESOURCE;
248 if (unlikely(buf == NULL)) {
249 printk(KERN_ERR "could not allocate integrity buffer\n");
250 goto err_end_io;
251 }
252
253 end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
254 start = ((unsigned long) buf) >> PAGE_SHIFT;
255 nr_pages = end - start;
256
257 /* Allocate bio integrity payload and integrity vectors */
258 bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages);
259 if (IS_ERR(bip)) {
260 printk(KERN_ERR "could not allocate data integrity bioset\n");
261 kfree(buf);
262 status = BLK_STS_RESOURCE;
263 goto err_end_io;
264 }
265
266 bip->bip_flags |= BIP_BLOCK_INTEGRITY;
267 bip->bip_iter.bi_size = len;
268 bip_set_seed(bip, bio->bi_iter.bi_sector);
269
270 if (bi->flags & BLK_INTEGRITY_IP_CHECKSUM)
271 bip->bip_flags |= BIP_IP_CHECKSUM;
272
273 /* Map it */
274 offset = offset_in_page(buf);
275 for (i = 0 ; i < nr_pages ; i++) {
276 int ret;
277 bytes = PAGE_SIZE - offset;
278
279 if (len <= 0)
280 break;
281
282 if (bytes > len)
283 bytes = len;
284
285 ret = bio_integrity_add_page(bio, virt_to_page(buf),
286 bytes, offset);
287
288 if (ret == 0) {
289 printk(KERN_ERR "could not attach integrity payload\n");
290 status = BLK_STS_RESOURCE;
291 goto err_end_io;
292 }
293
294 if (ret < bytes)
295 break;
296
297 buf += bytes;
298 len -= bytes;
299 offset = 0;
300 }
301
302 /* Auto-generate integrity metadata if this is a write */
303 if (bio_data_dir(bio) == WRITE) {
304 bio_integrity_process(bio, &bio->bi_iter,
305 bi->profile->generate_fn);
306 } else {
307 bip->bio_iter = bio->bi_iter;
308 }
309 return true;
310
311 err_end_io:
312 bio->bi_status = status;
313 bio_endio(bio);
314 return false;
315
316 }
317 EXPORT_SYMBOL(bio_integrity_prep);
318
319 /**
320 * bio_integrity_verify_fn - Integrity I/O completion worker
321 * @work: Work struct stored in bio to be verified
322 *
323 * Description: This workqueue function is called to complete a READ
324 * request. The function verifies the transferred integrity metadata
325 * and then calls the original bio end_io function.
326 */
bio_integrity_verify_fn(struct work_struct * work)327 static void bio_integrity_verify_fn(struct work_struct *work)
328 {
329 struct bio_integrity_payload *bip =
330 container_of(work, struct bio_integrity_payload, bip_work);
331 struct bio *bio = bip->bip_bio;
332 struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
333
334 /*
335 * At the moment verify is called bio's iterator was advanced
336 * during split and completion, we need to rewind iterator to
337 * it's original position.
338 */
339 bio->bi_status = bio_integrity_process(bio, &bip->bio_iter,
340 bi->profile->verify_fn);
341 bio_integrity_free(bio);
342 bio_endio(bio);
343 }
344
345 /**
346 * __bio_integrity_endio - Integrity I/O completion function
347 * @bio: Protected bio
348 *
349 * Description: Completion for integrity I/O
350 *
351 * Normally I/O completion is done in interrupt context. However,
352 * verifying I/O integrity is a time-consuming task which must be run
353 * in process context. This function postpones completion
354 * accordingly.
355 */
__bio_integrity_endio(struct bio * bio)356 bool __bio_integrity_endio(struct bio *bio)
357 {
358 struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
359 struct bio_integrity_payload *bip = bio_integrity(bio);
360
361 if (bio_op(bio) == REQ_OP_READ && !bio->bi_status &&
362 (bip->bip_flags & BIP_BLOCK_INTEGRITY) && bi->profile->verify_fn) {
363 INIT_WORK(&bip->bip_work, bio_integrity_verify_fn);
364 queue_work(kintegrityd_wq, &bip->bip_work);
365 return false;
366 }
367
368 bio_integrity_free(bio);
369 return true;
370 }
371
372 /**
373 * bio_integrity_advance - Advance integrity vector
374 * @bio: bio whose integrity vector to update
375 * @bytes_done: number of data bytes that have been completed
376 *
377 * Description: This function calculates how many integrity bytes the
378 * number of completed data bytes correspond to and advances the
379 * integrity vector accordingly.
380 */
bio_integrity_advance(struct bio * bio,unsigned int bytes_done)381 void bio_integrity_advance(struct bio *bio, unsigned int bytes_done)
382 {
383 struct bio_integrity_payload *bip = bio_integrity(bio);
384 struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
385 unsigned bytes = bio_integrity_bytes(bi, bytes_done >> 9);
386
387 bip->bip_iter.bi_sector += bytes_done >> 9;
388 bvec_iter_advance(bip->bip_vec, &bip->bip_iter, bytes);
389 }
390
391 /**
392 * bio_integrity_trim - Trim integrity vector
393 * @bio: bio whose integrity vector to update
394 *
395 * Description: Used to trim the integrity vector in a cloned bio.
396 */
bio_integrity_trim(struct bio * bio)397 void bio_integrity_trim(struct bio *bio)
398 {
399 struct bio_integrity_payload *bip = bio_integrity(bio);
400 struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
401
402 bip->bip_iter.bi_size = bio_integrity_bytes(bi, bio_sectors(bio));
403 }
404 EXPORT_SYMBOL(bio_integrity_trim);
405
406 /**
407 * bio_integrity_clone - Callback for cloning bios with integrity metadata
408 * @bio: New bio
409 * @bio_src: Original bio
410 * @gfp_mask: Memory allocation mask
411 *
412 * Description: Called to allocate a bip when cloning a bio
413 */
bio_integrity_clone(struct bio * bio,struct bio * bio_src,gfp_t gfp_mask)414 int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
415 gfp_t gfp_mask)
416 {
417 struct bio_integrity_payload *bip_src = bio_integrity(bio_src);
418 struct bio_integrity_payload *bip;
419
420 BUG_ON(bip_src == NULL);
421
422 bip = bio_integrity_alloc(bio, gfp_mask, bip_src->bip_vcnt);
423 if (IS_ERR(bip))
424 return PTR_ERR(bip);
425
426 memcpy(bip->bip_vec, bip_src->bip_vec,
427 bip_src->bip_vcnt * sizeof(struct bio_vec));
428
429 bip->bip_vcnt = bip_src->bip_vcnt;
430 bip->bip_iter = bip_src->bip_iter;
431
432 return 0;
433 }
434 EXPORT_SYMBOL(bio_integrity_clone);
435
bioset_integrity_create(struct bio_set * bs,int pool_size)436 int bioset_integrity_create(struct bio_set *bs, int pool_size)
437 {
438 if (mempool_initialized(&bs->bio_integrity_pool))
439 return 0;
440
441 if (mempool_init_slab_pool(&bs->bio_integrity_pool,
442 pool_size, bip_slab))
443 return -1;
444
445 if (biovec_init_pool(&bs->bvec_integrity_pool, pool_size)) {
446 mempool_exit(&bs->bio_integrity_pool);
447 return -1;
448 }
449
450 return 0;
451 }
452 EXPORT_SYMBOL(bioset_integrity_create);
453
bioset_integrity_free(struct bio_set * bs)454 void bioset_integrity_free(struct bio_set *bs)
455 {
456 mempool_exit(&bs->bio_integrity_pool);
457 mempool_exit(&bs->bvec_integrity_pool);
458 }
459
bio_integrity_init(void)460 void __init bio_integrity_init(void)
461 {
462 /*
463 * kintegrityd won't block much but may burn a lot of CPU cycles.
464 * Make it highpri CPU intensive wq with max concurrency of 1.
465 */
466 kintegrityd_wq = alloc_workqueue("kintegrityd", WQ_MEM_RECLAIM |
467 WQ_HIGHPRI | WQ_CPU_INTENSIVE, 1);
468 if (!kintegrityd_wq)
469 panic("Failed to create kintegrityd\n");
470
471 bip_slab = kmem_cache_create("bio_integrity_payload",
472 sizeof(struct bio_integrity_payload) +
473 sizeof(struct bio_vec) * BIP_INLINE_VECS,
474 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
475 }
476
