Loading...
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/blk-integrity.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
17static struct kmem_cache *bip_slab;
18static struct workqueue_struct *kintegrityd_wq;
19
20void blk_flush_integrity(void)
21{
22 flush_workqueue(kintegrityd_wq);
23}
24
25static void __bio_integrity_free(struct bio_set *bs,
26 struct bio_integrity_payload *bip)
27{
28 if (bs && mempool_initialized(&bs->bio_integrity_pool)) {
29 if (bip->bip_vec)
30 bvec_free(&bs->bvec_integrity_pool, bip->bip_vec,
31 bip->bip_max_vcnt);
32 mempool_free(bip, &bs->bio_integrity_pool);
33 } else {
34 kfree(bip);
35 }
36}
37
38/**
39 * bio_integrity_alloc - Allocate integrity payload and attach it to bio
40 * @bio: bio to attach integrity metadata to
41 * @gfp_mask: Memory allocation mask
42 * @nr_vecs: Number of integrity metadata scatter-gather elements
43 *
44 * Description: This function prepares a bio for attaching integrity
45 * metadata. nr_vecs specifies the maximum number of pages containing
46 * integrity metadata that can be attached.
47 */
48struct bio_integrity_payload *bio_integrity_alloc(struct bio *bio,
49 gfp_t gfp_mask,
50 unsigned int nr_vecs)
51{
52 struct bio_integrity_payload *bip;
53 struct bio_set *bs = bio->bi_pool;
54 unsigned inline_vecs;
55
56 if (WARN_ON_ONCE(bio_has_crypt_ctx(bio)))
57 return ERR_PTR(-EOPNOTSUPP);
58
59 if (!bs || !mempool_initialized(&bs->bio_integrity_pool)) {
60 bip = kmalloc(struct_size(bip, bip_inline_vecs, nr_vecs), gfp_mask);
61 inline_vecs = nr_vecs;
62 } else {
63 bip = mempool_alloc(&bs->bio_integrity_pool, gfp_mask);
64 inline_vecs = BIO_INLINE_VECS;
65 }
66
67 if (unlikely(!bip))
68 return ERR_PTR(-ENOMEM);
69
70 memset(bip, 0, sizeof(*bip));
71
72 if (nr_vecs > inline_vecs) {
73 bip->bip_max_vcnt = nr_vecs;
74 bip->bip_vec = bvec_alloc(&bs->bvec_integrity_pool,
75 &bip->bip_max_vcnt, gfp_mask);
76 if (!bip->bip_vec)
77 goto err;
78 } else {
79 bip->bip_vec = bip->bip_inline_vecs;
80 bip->bip_max_vcnt = inline_vecs;
81 }
82
83 bip->bip_bio = bio;
84 bio->bi_integrity = bip;
85 bio->bi_opf |= REQ_INTEGRITY;
86
87 return bip;
88err:
89 __bio_integrity_free(bs, bip);
90 return ERR_PTR(-ENOMEM);
91}
92EXPORT_SYMBOL(bio_integrity_alloc);
93
94/**
95 * bio_integrity_free - Free bio integrity payload
96 * @bio: bio containing bip to be freed
97 *
98 * Description: Used to free the integrity portion of a bio. Usually
99 * called from bio_free().
100 */
101void bio_integrity_free(struct bio *bio)
102{
103 struct bio_integrity_payload *bip = bio_integrity(bio);
104 struct bio_set *bs = bio->bi_pool;
105
106 if (bip->bip_flags & BIP_BLOCK_INTEGRITY)
107 kfree(bvec_virt(bip->bip_vec));
108
109 __bio_integrity_free(bs, bip);
110 bio->bi_integrity = NULL;
111 bio->bi_opf &= ~REQ_INTEGRITY;
112}
113
114/**
115 * bio_integrity_add_page - Attach integrity metadata
116 * @bio: bio to update
117 * @page: page containing integrity metadata
118 * @len: number of bytes of integrity metadata in page
119 * @offset: start offset within page
120 *
121 * Description: Attach a page containing integrity metadata to bio.
122 */
123int bio_integrity_add_page(struct bio *bio, struct page *page,
124 unsigned int len, unsigned int offset)
125{
126 struct bio_integrity_payload *bip = bio_integrity(bio);
127 struct bio_vec *iv;
128
129 if (bip->bip_vcnt >= bip->bip_max_vcnt) {
130 printk(KERN_ERR "%s: bip_vec full\n", __func__);
131 return 0;
132 }
133
134 iv = bip->bip_vec + bip->bip_vcnt;
135
136 if (bip->bip_vcnt &&
137 bvec_gap_to_prev(&bdev_get_queue(bio->bi_bdev)->limits,
138 &bip->bip_vec[bip->bip_vcnt - 1], offset))
139 return 0;
140
141 iv->bv_page = page;
142 iv->bv_len = len;
143 iv->bv_offset = offset;
144 bip->bip_vcnt++;
145
146 return len;
147}
148EXPORT_SYMBOL(bio_integrity_add_page);
149
150/**
151 * bio_integrity_process - Process integrity metadata for a bio
152 * @bio: bio to generate/verify integrity metadata for
153 * @proc_iter: iterator to process
154 * @proc_fn: Pointer to the relevant processing function
155 */
156static blk_status_t bio_integrity_process(struct bio *bio,
157 struct bvec_iter *proc_iter, integrity_processing_fn *proc_fn)
158{
159 struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk);
160 struct blk_integrity_iter iter;
161 struct bvec_iter bviter;
162 struct bio_vec bv;
163 struct bio_integrity_payload *bip = bio_integrity(bio);
164 blk_status_t ret = BLK_STS_OK;
165
166 iter.disk_name = bio->bi_bdev->bd_disk->disk_name;
167 iter.interval = 1 << bi->interval_exp;
168 iter.tuple_size = bi->tuple_size;
169 iter.seed = proc_iter->bi_sector;
170 iter.prot_buf = bvec_virt(bip->bip_vec);
171
172 __bio_for_each_segment(bv, bio, bviter, *proc_iter) {
173 void *kaddr = bvec_kmap_local(&bv);
174
175 iter.data_buf = kaddr;
176 iter.data_size = bv.bv_len;
177 ret = proc_fn(&iter);
178 kunmap_local(kaddr);
179
180 if (ret)
181 break;
182
183 }
184 return ret;
185}
186
187/**
188 * bio_integrity_prep - Prepare bio for integrity I/O
189 * @bio: bio to prepare
190 *
191 * Description: Checks if the bio already has an integrity payload attached.
192 * If it does, the payload has been generated by another kernel subsystem,
193 * and we just pass it through. Otherwise allocates integrity payload.
194 * The bio must have data direction, target device and start sector set priot
195 * to calling. In the WRITE case, integrity metadata will be generated using
196 * the block device's integrity function. In the READ case, the buffer
197 * will be prepared for DMA and a suitable end_io handler set up.
198 */
199bool bio_integrity_prep(struct bio *bio)
200{
201 struct bio_integrity_payload *bip;
202 struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk);
203 void *buf;
204 unsigned long start, end;
205 unsigned int len, nr_pages;
206 unsigned int bytes, offset, i;
207 unsigned int intervals;
208 blk_status_t status;
209
210 if (!bi)
211 return true;
212
213 if (bio_op(bio) != REQ_OP_READ && bio_op(bio) != REQ_OP_WRITE)
214 return true;
215
216 if (!bio_sectors(bio))
217 return true;
218
219 /* Already protected? */
220 if (bio_integrity(bio))
221 return true;
222
223 if (bio_data_dir(bio) == READ) {
224 if (!bi->profile->verify_fn ||
225 !(bi->flags & BLK_INTEGRITY_VERIFY))
226 return true;
227 } else {
228 if (!bi->profile->generate_fn ||
229 !(bi->flags & BLK_INTEGRITY_GENERATE))
230 return true;
231 }
232 intervals = bio_integrity_intervals(bi, bio_sectors(bio));
233
234 /* Allocate kernel buffer for protection data */
235 len = intervals * bi->tuple_size;
236 buf = kmalloc(len, GFP_NOIO);
237 status = BLK_STS_RESOURCE;
238 if (unlikely(buf == NULL)) {
239 printk(KERN_ERR "could not allocate integrity buffer\n");
240 goto err_end_io;
241 }
242
243 end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
244 start = ((unsigned long) buf) >> PAGE_SHIFT;
245 nr_pages = end - start;
246
247 /* Allocate bio integrity payload and integrity vectors */
248 bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages);
249 if (IS_ERR(bip)) {
250 printk(KERN_ERR "could not allocate data integrity bioset\n");
251 kfree(buf);
252 status = BLK_STS_RESOURCE;
253 goto err_end_io;
254 }
255
256 bip->bip_flags |= BIP_BLOCK_INTEGRITY;
257 bip->bip_iter.bi_size = len;
258 bip_set_seed(bip, bio->bi_iter.bi_sector);
259
260 if (bi->flags & BLK_INTEGRITY_IP_CHECKSUM)
261 bip->bip_flags |= BIP_IP_CHECKSUM;
262
263 /* Map it */
264 offset = offset_in_page(buf);
265 for (i = 0 ; i < nr_pages ; i++) {
266 int ret;
267 bytes = PAGE_SIZE - offset;
268
269 if (len <= 0)
270 break;
271
272 if (bytes > len)
273 bytes = len;
274
275 ret = bio_integrity_add_page(bio, virt_to_page(buf),
276 bytes, offset);
277
278 if (ret == 0) {
279 printk(KERN_ERR "could not attach integrity payload\n");
280 status = BLK_STS_RESOURCE;
281 goto err_end_io;
282 }
283
284 if (ret < bytes)
285 break;
286
287 buf += bytes;
288 len -= bytes;
289 offset = 0;
290 }
291
292 /* Auto-generate integrity metadata if this is a write */
293 if (bio_data_dir(bio) == WRITE) {
294 bio_integrity_process(bio, &bio->bi_iter,
295 bi->profile->generate_fn);
296 } else {
297 bip->bio_iter = bio->bi_iter;
298 }
299 return true;
300
301err_end_io:
302 bio->bi_status = status;
303 bio_endio(bio);
304 return false;
305
306}
307EXPORT_SYMBOL(bio_integrity_prep);
308
309/**
310 * bio_integrity_verify_fn - Integrity I/O completion worker
311 * @work: Work struct stored in bio to be verified
312 *
313 * Description: This workqueue function is called to complete a READ
314 * request. The function verifies the transferred integrity metadata
315 * and then calls the original bio end_io function.
316 */
317static void bio_integrity_verify_fn(struct work_struct *work)
318{
319 struct bio_integrity_payload *bip =
320 container_of(work, struct bio_integrity_payload, bip_work);
321 struct bio *bio = bip->bip_bio;
322 struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk);
323
324 /*
325 * At the moment verify is called bio's iterator was advanced
326 * during split and completion, we need to rewind iterator to
327 * it's original position.
328 */
329 bio->bi_status = bio_integrity_process(bio, &bip->bio_iter,
330 bi->profile->verify_fn);
331 bio_integrity_free(bio);
332 bio_endio(bio);
333}
334
335/**
336 * __bio_integrity_endio - Integrity I/O completion function
337 * @bio: Protected bio
338 *
339 * Description: Completion for integrity I/O
340 *
341 * Normally I/O completion is done in interrupt context. However,
342 * verifying I/O integrity is a time-consuming task which must be run
343 * in process context. This function postpones completion
344 * accordingly.
345 */
346bool __bio_integrity_endio(struct bio *bio)
347{
348 struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk);
349 struct bio_integrity_payload *bip = bio_integrity(bio);
350
351 if (bio_op(bio) == REQ_OP_READ && !bio->bi_status &&
352 (bip->bip_flags & BIP_BLOCK_INTEGRITY) && bi->profile->verify_fn) {
353 INIT_WORK(&bip->bip_work, bio_integrity_verify_fn);
354 queue_work(kintegrityd_wq, &bip->bip_work);
355 return false;
356 }
357
358 bio_integrity_free(bio);
359 return true;
360}
361
362/**
363 * bio_integrity_advance - Advance integrity vector
364 * @bio: bio whose integrity vector to update
365 * @bytes_done: number of data bytes that have been completed
366 *
367 * Description: This function calculates how many integrity bytes the
368 * number of completed data bytes correspond to and advances the
369 * integrity vector accordingly.
370 */
371void bio_integrity_advance(struct bio *bio, unsigned int bytes_done)
372{
373 struct bio_integrity_payload *bip = bio_integrity(bio);
374 struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk);
375 unsigned bytes = bio_integrity_bytes(bi, bytes_done >> 9);
376
377 bip->bip_iter.bi_sector += bio_integrity_intervals(bi, bytes_done >> 9);
378 bvec_iter_advance(bip->bip_vec, &bip->bip_iter, bytes);
379}
380
381/**
382 * bio_integrity_trim - Trim integrity vector
383 * @bio: bio whose integrity vector to update
384 *
385 * Description: Used to trim the integrity vector in a cloned bio.
386 */
387void bio_integrity_trim(struct bio *bio)
388{
389 struct bio_integrity_payload *bip = bio_integrity(bio);
390 struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk);
391
392 bip->bip_iter.bi_size = bio_integrity_bytes(bi, bio_sectors(bio));
393}
394EXPORT_SYMBOL(bio_integrity_trim);
395
396/**
397 * bio_integrity_clone - Callback for cloning bios with integrity metadata
398 * @bio: New bio
399 * @bio_src: Original bio
400 * @gfp_mask: Memory allocation mask
401 *
402 * Description: Called to allocate a bip when cloning a bio
403 */
404int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
405 gfp_t gfp_mask)
406{
407 struct bio_integrity_payload *bip_src = bio_integrity(bio_src);
408 struct bio_integrity_payload *bip;
409
410 BUG_ON(bip_src == NULL);
411
412 bip = bio_integrity_alloc(bio, gfp_mask, bip_src->bip_vcnt);
413 if (IS_ERR(bip))
414 return PTR_ERR(bip);
415
416 memcpy(bip->bip_vec, bip_src->bip_vec,
417 bip_src->bip_vcnt * sizeof(struct bio_vec));
418
419 bip->bip_vcnt = bip_src->bip_vcnt;
420 bip->bip_iter = bip_src->bip_iter;
421
422 return 0;
423}
424
425int bioset_integrity_create(struct bio_set *bs, int pool_size)
426{
427 if (mempool_initialized(&bs->bio_integrity_pool))
428 return 0;
429
430 if (mempool_init_slab_pool(&bs->bio_integrity_pool,
431 pool_size, bip_slab))
432 return -1;
433
434 if (biovec_init_pool(&bs->bvec_integrity_pool, pool_size)) {
435 mempool_exit(&bs->bio_integrity_pool);
436 return -1;
437 }
438
439 return 0;
440}
441EXPORT_SYMBOL(bioset_integrity_create);
442
443void bioset_integrity_free(struct bio_set *bs)
444{
445 mempool_exit(&bs->bio_integrity_pool);
446 mempool_exit(&bs->bvec_integrity_pool);
447}
448
449void __init bio_integrity_init(void)
450{
451 /*
452 * kintegrityd won't block much but may burn a lot of CPU cycles.
453 * Make it highpri CPU intensive wq with max concurrency of 1.
454 */
455 kintegrityd_wq = alloc_workqueue("kintegrityd", WQ_MEM_RECLAIM |
456 WQ_HIGHPRI | WQ_CPU_INTENSIVE, 1);
457 if (!kintegrityd_wq)
458 panic("Failed to create kintegrityd\n");
459
460 bip_slab = kmem_cache_create("bio_integrity_payload",
461 sizeof(struct bio_integrity_payload) +
462 sizeof(struct bio_vec) * BIO_INLINE_VECS,
463 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
464}
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/blk-integrity.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
17static struct kmem_cache *bip_slab;
18static struct workqueue_struct *kintegrityd_wq;
19
20void blk_flush_integrity(void)
21{
22 flush_workqueue(kintegrityd_wq);
23}
24
25static void __bio_integrity_free(struct bio_set *bs,
26 struct bio_integrity_payload *bip)
27{
28 if (bs && mempool_initialized(&bs->bio_integrity_pool)) {
29 if (bip->bip_vec)
30 bvec_free(&bs->bvec_integrity_pool, bip->bip_vec,
31 bip->bip_max_vcnt);
32 mempool_free(bip, &bs->bio_integrity_pool);
33 } else {
34 kfree(bip);
35 }
36}
37
38/**
39 * bio_integrity_alloc - Allocate integrity payload and attach it to bio
40 * @bio: bio to attach integrity metadata to
41 * @gfp_mask: Memory allocation mask
42 * @nr_vecs: Number of integrity metadata scatter-gather elements
43 *
44 * Description: This function prepares a bio for attaching integrity
45 * metadata. nr_vecs specifies the maximum number of pages containing
46 * integrity metadata that can be attached.
47 */
48struct bio_integrity_payload *bio_integrity_alloc(struct bio *bio,
49 gfp_t gfp_mask,
50 unsigned int nr_vecs)
51{
52 struct bio_integrity_payload *bip;
53 struct bio_set *bs = bio->bi_pool;
54 unsigned inline_vecs;
55
56 if (WARN_ON_ONCE(bio_has_crypt_ctx(bio)))
57 return ERR_PTR(-EOPNOTSUPP);
58
59 if (!bs || !mempool_initialized(&bs->bio_integrity_pool)) {
60 bip = kmalloc(struct_size(bip, bip_inline_vecs, nr_vecs), gfp_mask);
61 inline_vecs = nr_vecs;
62 } else {
63 bip = mempool_alloc(&bs->bio_integrity_pool, gfp_mask);
64 inline_vecs = BIO_INLINE_VECS;
65 }
66
67 if (unlikely(!bip))
68 return ERR_PTR(-ENOMEM);
69
70 memset(bip, 0, sizeof(*bip));
71
72 /* always report as many vecs as asked explicitly, not inline vecs */
73 bip->bip_max_vcnt = nr_vecs;
74 if (nr_vecs > inline_vecs) {
75 bip->bip_vec = bvec_alloc(&bs->bvec_integrity_pool,
76 &bip->bip_max_vcnt, gfp_mask);
77 if (!bip->bip_vec)
78 goto err;
79 } else {
80 bip->bip_vec = bip->bip_inline_vecs;
81 }
82
83 bip->bip_bio = bio;
84 bio->bi_integrity = bip;
85 bio->bi_opf |= REQ_INTEGRITY;
86
87 return bip;
88err:
89 __bio_integrity_free(bs, bip);
90 return ERR_PTR(-ENOMEM);
91}
92EXPORT_SYMBOL(bio_integrity_alloc);
93
94static void bio_integrity_unpin_bvec(struct bio_vec *bv, int nr_vecs,
95 bool dirty)
96{
97 int i;
98
99 for (i = 0; i < nr_vecs; i++) {
100 if (dirty && !PageCompound(bv[i].bv_page))
101 set_page_dirty_lock(bv[i].bv_page);
102 unpin_user_page(bv[i].bv_page);
103 }
104}
105
106static void bio_integrity_uncopy_user(struct bio_integrity_payload *bip)
107{
108 unsigned short nr_vecs = bip->bip_max_vcnt - 1;
109 struct bio_vec *copy = &bip->bip_vec[1];
110 size_t bytes = bip->bip_iter.bi_size;
111 struct iov_iter iter;
112 int ret;
113
114 iov_iter_bvec(&iter, ITER_DEST, copy, nr_vecs, bytes);
115 ret = copy_to_iter(bvec_virt(bip->bip_vec), bytes, &iter);
116 WARN_ON_ONCE(ret != bytes);
117
118 bio_integrity_unpin_bvec(copy, nr_vecs, true);
119}
120
121static void bio_integrity_unmap_user(struct bio_integrity_payload *bip)
122{
123 bool dirty = bio_data_dir(bip->bip_bio) == READ;
124
125 if (bip->bip_flags & BIP_COPY_USER) {
126 if (dirty)
127 bio_integrity_uncopy_user(bip);
128 kfree(bvec_virt(bip->bip_vec));
129 return;
130 }
131
132 bio_integrity_unpin_bvec(bip->bip_vec, bip->bip_max_vcnt, dirty);
133}
134
135/**
136 * bio_integrity_free - Free bio integrity payload
137 * @bio: bio containing bip to be freed
138 *
139 * Description: Used to free the integrity portion of a bio. Usually
140 * called from bio_free().
141 */
142void bio_integrity_free(struct bio *bio)
143{
144 struct bio_integrity_payload *bip = bio_integrity(bio);
145 struct bio_set *bs = bio->bi_pool;
146
147 if (bip->bip_flags & BIP_BLOCK_INTEGRITY)
148 kfree(bvec_virt(bip->bip_vec));
149 else if (bip->bip_flags & BIP_INTEGRITY_USER)
150 bio_integrity_unmap_user(bip);
151
152 __bio_integrity_free(bs, bip);
153 bio->bi_integrity = NULL;
154 bio->bi_opf &= ~REQ_INTEGRITY;
155}
156
157/**
158 * bio_integrity_add_page - Attach integrity metadata
159 * @bio: bio to update
160 * @page: page containing integrity metadata
161 * @len: number of bytes of integrity metadata in page
162 * @offset: start offset within page
163 *
164 * Description: Attach a page containing integrity metadata to bio.
165 */
166int bio_integrity_add_page(struct bio *bio, struct page *page,
167 unsigned int len, unsigned int offset)
168{
169 struct request_queue *q = bdev_get_queue(bio->bi_bdev);
170 struct bio_integrity_payload *bip = bio_integrity(bio);
171
172 if (((bip->bip_iter.bi_size + len) >> SECTOR_SHIFT) >
173 queue_max_hw_sectors(q))
174 return 0;
175
176 if (bip->bip_vcnt > 0) {
177 struct bio_vec *bv = &bip->bip_vec[bip->bip_vcnt - 1];
178 bool same_page = false;
179
180 if (bvec_try_merge_hw_page(q, bv, page, len, offset,
181 &same_page)) {
182 bip->bip_iter.bi_size += len;
183 return len;
184 }
185
186 if (bip->bip_vcnt >=
187 min(bip->bip_max_vcnt, queue_max_integrity_segments(q)))
188 return 0;
189
190 /*
191 * If the queue doesn't support SG gaps and adding this segment
192 * would create a gap, disallow it.
193 */
194 if (bvec_gap_to_prev(&q->limits, bv, offset))
195 return 0;
196 }
197
198 bvec_set_page(&bip->bip_vec[bip->bip_vcnt], page, len, offset);
199 bip->bip_vcnt++;
200 bip->bip_iter.bi_size += len;
201
202 return len;
203}
204EXPORT_SYMBOL(bio_integrity_add_page);
205
206static int bio_integrity_copy_user(struct bio *bio, struct bio_vec *bvec,
207 int nr_vecs, unsigned int len,
208 unsigned int direction, u32 seed)
209{
210 bool write = direction == ITER_SOURCE;
211 struct bio_integrity_payload *bip;
212 struct iov_iter iter;
213 void *buf;
214 int ret;
215
216 buf = kmalloc(len, GFP_KERNEL);
217 if (!buf)
218 return -ENOMEM;
219
220 if (write) {
221 iov_iter_bvec(&iter, direction, bvec, nr_vecs, len);
222 if (!copy_from_iter_full(buf, len, &iter)) {
223 ret = -EFAULT;
224 goto free_buf;
225 }
226
227 bip = bio_integrity_alloc(bio, GFP_KERNEL, 1);
228 } else {
229 memset(buf, 0, len);
230
231 /*
232 * We need to preserve the original bvec and the number of vecs
233 * in it for completion handling
234 */
235 bip = bio_integrity_alloc(bio, GFP_KERNEL, nr_vecs + 1);
236 }
237
238 if (IS_ERR(bip)) {
239 ret = PTR_ERR(bip);
240 goto free_buf;
241 }
242
243 if (write)
244 bio_integrity_unpin_bvec(bvec, nr_vecs, false);
245 else
246 memcpy(&bip->bip_vec[1], bvec, nr_vecs * sizeof(*bvec));
247
248 ret = bio_integrity_add_page(bio, virt_to_page(buf), len,
249 offset_in_page(buf));
250 if (ret != len) {
251 ret = -ENOMEM;
252 goto free_bip;
253 }
254
255 bip->bip_flags |= BIP_INTEGRITY_USER | BIP_COPY_USER;
256 bip->bip_iter.bi_sector = seed;
257 return 0;
258free_bip:
259 bio_integrity_free(bio);
260free_buf:
261 kfree(buf);
262 return ret;
263}
264
265static int bio_integrity_init_user(struct bio *bio, struct bio_vec *bvec,
266 int nr_vecs, unsigned int len, u32 seed)
267{
268 struct bio_integrity_payload *bip;
269
270 bip = bio_integrity_alloc(bio, GFP_KERNEL, nr_vecs);
271 if (IS_ERR(bip))
272 return PTR_ERR(bip);
273
274 memcpy(bip->bip_vec, bvec, nr_vecs * sizeof(*bvec));
275 bip->bip_flags |= BIP_INTEGRITY_USER;
276 bip->bip_iter.bi_sector = seed;
277 bip->bip_iter.bi_size = len;
278 return 0;
279}
280
281static unsigned int bvec_from_pages(struct bio_vec *bvec, struct page **pages,
282 int nr_vecs, ssize_t bytes, ssize_t offset)
283{
284 unsigned int nr_bvecs = 0;
285 int i, j;
286
287 for (i = 0; i < nr_vecs; i = j) {
288 size_t size = min_t(size_t, bytes, PAGE_SIZE - offset);
289 struct folio *folio = page_folio(pages[i]);
290
291 bytes -= size;
292 for (j = i + 1; j < nr_vecs; j++) {
293 size_t next = min_t(size_t, PAGE_SIZE, bytes);
294
295 if (page_folio(pages[j]) != folio ||
296 pages[j] != pages[j - 1] + 1)
297 break;
298 unpin_user_page(pages[j]);
299 size += next;
300 bytes -= next;
301 }
302
303 bvec_set_page(&bvec[nr_bvecs], pages[i], size, offset);
304 offset = 0;
305 nr_bvecs++;
306 }
307
308 return nr_bvecs;
309}
310
311int bio_integrity_map_user(struct bio *bio, void __user *ubuf, ssize_t bytes,
312 u32 seed)
313{
314 struct request_queue *q = bdev_get_queue(bio->bi_bdev);
315 unsigned int align = q->dma_pad_mask | queue_dma_alignment(q);
316 struct page *stack_pages[UIO_FASTIOV], **pages = stack_pages;
317 struct bio_vec stack_vec[UIO_FASTIOV], *bvec = stack_vec;
318 unsigned int direction, nr_bvecs;
319 struct iov_iter iter;
320 int ret, nr_vecs;
321 size_t offset;
322 bool copy;
323
324 if (bio_integrity(bio))
325 return -EINVAL;
326 if (bytes >> SECTOR_SHIFT > queue_max_hw_sectors(q))
327 return -E2BIG;
328
329 if (bio_data_dir(bio) == READ)
330 direction = ITER_DEST;
331 else
332 direction = ITER_SOURCE;
333
334 iov_iter_ubuf(&iter, direction, ubuf, bytes);
335 nr_vecs = iov_iter_npages(&iter, BIO_MAX_VECS + 1);
336 if (nr_vecs > BIO_MAX_VECS)
337 return -E2BIG;
338 if (nr_vecs > UIO_FASTIOV) {
339 bvec = kcalloc(nr_vecs, sizeof(*bvec), GFP_KERNEL);
340 if (!bvec)
341 return -ENOMEM;
342 pages = NULL;
343 }
344
345 copy = !iov_iter_is_aligned(&iter, align, align);
346 ret = iov_iter_extract_pages(&iter, &pages, bytes, nr_vecs, 0, &offset);
347 if (unlikely(ret < 0))
348 goto free_bvec;
349
350 nr_bvecs = bvec_from_pages(bvec, pages, nr_vecs, bytes, offset);
351 if (pages != stack_pages)
352 kvfree(pages);
353 if (nr_bvecs > queue_max_integrity_segments(q))
354 copy = true;
355
356 if (copy)
357 ret = bio_integrity_copy_user(bio, bvec, nr_bvecs, bytes,
358 direction, seed);
359 else
360 ret = bio_integrity_init_user(bio, bvec, nr_bvecs, bytes, seed);
361 if (ret)
362 goto release_pages;
363 if (bvec != stack_vec)
364 kfree(bvec);
365
366 return 0;
367
368release_pages:
369 bio_integrity_unpin_bvec(bvec, nr_bvecs, false);
370free_bvec:
371 if (bvec != stack_vec)
372 kfree(bvec);
373 return ret;
374}
375EXPORT_SYMBOL_GPL(bio_integrity_map_user);
376
377/**
378 * bio_integrity_process - Process integrity metadata for a bio
379 * @bio: bio to generate/verify integrity metadata for
380 * @proc_iter: iterator to process
381 * @proc_fn: Pointer to the relevant processing function
382 */
383static blk_status_t bio_integrity_process(struct bio *bio,
384 struct bvec_iter *proc_iter, integrity_processing_fn *proc_fn)
385{
386 struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk);
387 struct blk_integrity_iter iter;
388 struct bvec_iter bviter;
389 struct bio_vec bv;
390 struct bio_integrity_payload *bip = bio_integrity(bio);
391 blk_status_t ret = BLK_STS_OK;
392
393 iter.disk_name = bio->bi_bdev->bd_disk->disk_name;
394 iter.interval = 1 << bi->interval_exp;
395 iter.tuple_size = bi->tuple_size;
396 iter.seed = proc_iter->bi_sector;
397 iter.prot_buf = bvec_virt(bip->bip_vec);
398 iter.pi_offset = bi->pi_offset;
399
400 __bio_for_each_segment(bv, bio, bviter, *proc_iter) {
401 void *kaddr = bvec_kmap_local(&bv);
402
403 iter.data_buf = kaddr;
404 iter.data_size = bv.bv_len;
405 ret = proc_fn(&iter);
406 kunmap_local(kaddr);
407
408 if (ret)
409 break;
410
411 }
412 return ret;
413}
414
415/**
416 * bio_integrity_prep - Prepare bio for integrity I/O
417 * @bio: bio to prepare
418 *
419 * Description: Checks if the bio already has an integrity payload attached.
420 * If it does, the payload has been generated by another kernel subsystem,
421 * and we just pass it through. Otherwise allocates integrity payload.
422 * The bio must have data direction, target device and start sector set priot
423 * to calling. In the WRITE case, integrity metadata will be generated using
424 * the block device's integrity function. In the READ case, the buffer
425 * will be prepared for DMA and a suitable end_io handler set up.
426 */
427bool bio_integrity_prep(struct bio *bio)
428{
429 struct bio_integrity_payload *bip;
430 struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk);
431 void *buf;
432 unsigned long start, end;
433 unsigned int len, nr_pages;
434 unsigned int bytes, offset, i;
435
436 if (!bi)
437 return true;
438
439 if (bio_op(bio) != REQ_OP_READ && bio_op(bio) != REQ_OP_WRITE)
440 return true;
441
442 if (!bio_sectors(bio))
443 return true;
444
445 /* Already protected? */
446 if (bio_integrity(bio))
447 return true;
448
449 if (bio_data_dir(bio) == READ) {
450 if (!bi->profile->verify_fn ||
451 !(bi->flags & BLK_INTEGRITY_VERIFY))
452 return true;
453 } else {
454 if (!bi->profile->generate_fn ||
455 !(bi->flags & BLK_INTEGRITY_GENERATE))
456 return true;
457 }
458
459 /* Allocate kernel buffer for protection data */
460 len = bio_integrity_bytes(bi, bio_sectors(bio));
461 buf = kmalloc(len, GFP_NOIO);
462 if (unlikely(buf == NULL)) {
463 printk(KERN_ERR "could not allocate integrity buffer\n");
464 goto err_end_io;
465 }
466
467 end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
468 start = ((unsigned long) buf) >> PAGE_SHIFT;
469 nr_pages = end - start;
470
471 /* Allocate bio integrity payload and integrity vectors */
472 bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages);
473 if (IS_ERR(bip)) {
474 printk(KERN_ERR "could not allocate data integrity bioset\n");
475 kfree(buf);
476 goto err_end_io;
477 }
478
479 bip->bip_flags |= BIP_BLOCK_INTEGRITY;
480 bip_set_seed(bip, bio->bi_iter.bi_sector);
481
482 if (bi->flags & BLK_INTEGRITY_IP_CHECKSUM)
483 bip->bip_flags |= BIP_IP_CHECKSUM;
484
485 /* Map it */
486 offset = offset_in_page(buf);
487 for (i = 0; i < nr_pages && len > 0; i++) {
488 bytes = PAGE_SIZE - offset;
489
490 if (bytes > len)
491 bytes = len;
492
493 if (bio_integrity_add_page(bio, virt_to_page(buf),
494 bytes, offset) < bytes) {
495 printk(KERN_ERR "could not attach integrity payload\n");
496 goto err_end_io;
497 }
498
499 buf += bytes;
500 len -= bytes;
501 offset = 0;
502 }
503
504 /* Auto-generate integrity metadata if this is a write */
505 if (bio_data_dir(bio) == WRITE) {
506 bio_integrity_process(bio, &bio->bi_iter,
507 bi->profile->generate_fn);
508 } else {
509 bip->bio_iter = bio->bi_iter;
510 }
511 return true;
512
513err_end_io:
514 bio->bi_status = BLK_STS_RESOURCE;
515 bio_endio(bio);
516 return false;
517}
518EXPORT_SYMBOL(bio_integrity_prep);
519
520/**
521 * bio_integrity_verify_fn - Integrity I/O completion worker
522 * @work: Work struct stored in bio to be verified
523 *
524 * Description: This workqueue function is called to complete a READ
525 * request. The function verifies the transferred integrity metadata
526 * and then calls the original bio end_io function.
527 */
528static void bio_integrity_verify_fn(struct work_struct *work)
529{
530 struct bio_integrity_payload *bip =
531 container_of(work, struct bio_integrity_payload, bip_work);
532 struct bio *bio = bip->bip_bio;
533 struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk);
534
535 /*
536 * At the moment verify is called bio's iterator was advanced
537 * during split and completion, we need to rewind iterator to
538 * it's original position.
539 */
540 bio->bi_status = bio_integrity_process(bio, &bip->bio_iter,
541 bi->profile->verify_fn);
542 bio_integrity_free(bio);
543 bio_endio(bio);
544}
545
546/**
547 * __bio_integrity_endio - Integrity I/O completion function
548 * @bio: Protected bio
549 *
550 * Description: Completion for integrity I/O
551 *
552 * Normally I/O completion is done in interrupt context. However,
553 * verifying I/O integrity is a time-consuming task which must be run
554 * in process context. This function postpones completion
555 * accordingly.
556 */
557bool __bio_integrity_endio(struct bio *bio)
558{
559 struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk);
560 struct bio_integrity_payload *bip = bio_integrity(bio);
561
562 if (bio_op(bio) == REQ_OP_READ && !bio->bi_status &&
563 (bip->bip_flags & BIP_BLOCK_INTEGRITY) && bi->profile->verify_fn) {
564 INIT_WORK(&bip->bip_work, bio_integrity_verify_fn);
565 queue_work(kintegrityd_wq, &bip->bip_work);
566 return false;
567 }
568
569 bio_integrity_free(bio);
570 return true;
571}
572
573/**
574 * bio_integrity_advance - Advance integrity vector
575 * @bio: bio whose integrity vector to update
576 * @bytes_done: number of data bytes that have been completed
577 *
578 * Description: This function calculates how many integrity bytes the
579 * number of completed data bytes correspond to and advances the
580 * integrity vector accordingly.
581 */
582void bio_integrity_advance(struct bio *bio, unsigned int bytes_done)
583{
584 struct bio_integrity_payload *bip = bio_integrity(bio);
585 struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk);
586 unsigned bytes = bio_integrity_bytes(bi, bytes_done >> 9);
587
588 bip->bip_iter.bi_sector += bio_integrity_intervals(bi, bytes_done >> 9);
589 bvec_iter_advance(bip->bip_vec, &bip->bip_iter, bytes);
590}
591
592/**
593 * bio_integrity_trim - Trim integrity vector
594 * @bio: bio whose integrity vector to update
595 *
596 * Description: Used to trim the integrity vector in a cloned bio.
597 */
598void bio_integrity_trim(struct bio *bio)
599{
600 struct bio_integrity_payload *bip = bio_integrity(bio);
601 struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk);
602
603 bip->bip_iter.bi_size = bio_integrity_bytes(bi, bio_sectors(bio));
604}
605EXPORT_SYMBOL(bio_integrity_trim);
606
607/**
608 * bio_integrity_clone - Callback for cloning bios with integrity metadata
609 * @bio: New bio
610 * @bio_src: Original bio
611 * @gfp_mask: Memory allocation mask
612 *
613 * Description: Called to allocate a bip when cloning a bio
614 */
615int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
616 gfp_t gfp_mask)
617{
618 struct bio_integrity_payload *bip_src = bio_integrity(bio_src);
619 struct bio_integrity_payload *bip;
620
621 BUG_ON(bip_src == NULL);
622
623 bip = bio_integrity_alloc(bio, gfp_mask, bip_src->bip_vcnt);
624 if (IS_ERR(bip))
625 return PTR_ERR(bip);
626
627 memcpy(bip->bip_vec, bip_src->bip_vec,
628 bip_src->bip_vcnt * sizeof(struct bio_vec));
629
630 bip->bip_vcnt = bip_src->bip_vcnt;
631 bip->bip_iter = bip_src->bip_iter;
632 bip->bip_flags = bip_src->bip_flags & ~BIP_BLOCK_INTEGRITY;
633
634 return 0;
635}
636
637int bioset_integrity_create(struct bio_set *bs, int pool_size)
638{
639 if (mempool_initialized(&bs->bio_integrity_pool))
640 return 0;
641
642 if (mempool_init_slab_pool(&bs->bio_integrity_pool,
643 pool_size, bip_slab))
644 return -1;
645
646 if (biovec_init_pool(&bs->bvec_integrity_pool, pool_size)) {
647 mempool_exit(&bs->bio_integrity_pool);
648 return -1;
649 }
650
651 return 0;
652}
653EXPORT_SYMBOL(bioset_integrity_create);
654
655void bioset_integrity_free(struct bio_set *bs)
656{
657 mempool_exit(&bs->bio_integrity_pool);
658 mempool_exit(&bs->bvec_integrity_pool);
659}
660
661void __init bio_integrity_init(void)
662{
663 /*
664 * kintegrityd won't block much but may burn a lot of CPU cycles.
665 * Make it highpri CPU intensive wq with max concurrency of 1.
666 */
667 kintegrityd_wq = alloc_workqueue("kintegrityd", WQ_MEM_RECLAIM |
668 WQ_HIGHPRI | WQ_CPU_INTENSIVE, 1);
669 if (!kintegrityd_wq)
670 panic("Failed to create kintegrityd\n");
671
672 bip_slab = kmem_cache_create("bio_integrity_payload",
673 sizeof(struct bio_integrity_payload) +
674 sizeof(struct bio_vec) * BIO_INLINE_VECS,
675 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
676}