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