2 * bio-integrity.c - bio data integrity extensions
4 * Copyright (C) 2007, 2008 Oracle Corporation
5 * Written by: Martin K. Petersen <martin.petersen@oracle.com>
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.
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.
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,
23 #include <linux/blkdev.h>
24 #include <linux/mempool.h>
25 #include <linux/bio.h>
26 #include <linux/workqueue.h>
28 static struct kmem_cache *bio_integrity_slab __read_mostly;
29 static struct workqueue_struct *kintegrityd_wq;
32 * bio_integrity_alloc_bioset - Allocate integrity payload and attach it to bio
33 * @bio: bio to attach integrity metadata to
34 * @gfp_mask: Memory allocation mask
35 * @nr_vecs: Number of integrity metadata scatter-gather elements
36 * @bs: bio_set to allocate from
38 * Description: This function prepares a bio for attaching integrity
39 * metadata. nr_vecs specifies the maximum number of pages containing
40 * integrity metadata that can be attached.
42 struct bio_integrity_payload *bio_integrity_alloc_bioset(struct bio *bio,
47 struct bio_integrity_payload *bip;
53 bip = mempool_alloc(bs->bio_integrity_pool, gfp_mask);
54 if (unlikely(bip == NULL)) {
55 printk(KERN_ERR "%s: could not alloc bip\n", __func__);
59 memset(bip, 0, sizeof(*bip));
61 iv = bvec_alloc_bs(gfp_mask, nr_vecs, &idx, bs);
62 if (unlikely(iv == NULL)) {
63 printk(KERN_ERR "%s: could not alloc bip_vec\n", __func__);
64 mempool_free(bip, bs->bio_integrity_pool);
71 bio->bi_integrity = bip;
75 EXPORT_SYMBOL(bio_integrity_alloc_bioset);
78 * bio_integrity_alloc - Allocate integrity payload and attach it to bio
79 * @bio: bio to attach integrity metadata to
80 * @gfp_mask: Memory allocation mask
81 * @nr_vecs: Number of integrity metadata scatter-gather elements
83 * Description: This function prepares a bio for attaching integrity
84 * metadata. nr_vecs specifies the maximum number of pages containing
85 * integrity metadata that can be attached.
87 struct bio_integrity_payload *bio_integrity_alloc(struct bio *bio,
91 return bio_integrity_alloc_bioset(bio, gfp_mask, nr_vecs, fs_bio_set);
93 EXPORT_SYMBOL(bio_integrity_alloc);
96 * bio_integrity_free - Free bio integrity payload
97 * @bio: bio containing bip to be freed
98 * @bs: bio_set this bio was allocated from
100 * Description: Used to free the integrity portion of a bio. Usually
101 * called from bio_free().
103 void bio_integrity_free(struct bio *bio, struct bio_set *bs)
105 struct bio_integrity_payload *bip = bio->bi_integrity;
109 /* A cloned bio doesn't own the integrity metadata */
110 if (!bio_flagged(bio, BIO_CLONED) && bip->bip_buf != NULL)
113 mempool_free(bip->bip_vec, bs->bvec_pools[bip->bip_pool]);
114 mempool_free(bip, bs->bio_integrity_pool);
116 bio->bi_integrity = NULL;
118 EXPORT_SYMBOL(bio_integrity_free);
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
127 * Description: Attach a page containing integrity metadata to bio.
129 int bio_integrity_add_page(struct bio *bio, struct page *page,
130 unsigned int len, unsigned int offset)
132 struct bio_integrity_payload *bip = bio->bi_integrity;
135 if (bip->bip_vcnt >= bvec_nr_vecs(bip->bip_pool)) {
136 printk(KERN_ERR "%s: bip_vec full\n", __func__);
140 iv = bip_vec_idx(bip, bip->bip_vcnt);
142 BUG_ON(iv->bv_page != NULL);
146 iv->bv_offset = offset;
151 EXPORT_SYMBOL(bio_integrity_add_page);
154 * bio_integrity_enabled - Check whether integrity can be passed
157 * Description: Determines whether bio_integrity_prep() can be called
158 * on this bio or not. bio data direction and target device must be
159 * set prior to calling. The functions honors the write_generate and
160 * read_verify flags in sysfs.
162 int bio_integrity_enabled(struct bio *bio)
164 /* Already protected? */
165 if (bio_integrity(bio))
168 return bdev_integrity_enabled(bio->bi_bdev, bio_data_dir(bio));
170 EXPORT_SYMBOL(bio_integrity_enabled);
173 * bio_integrity_hw_sectors - Convert 512b sectors to hardware ditto
174 * @bi: blk_integrity profile for device
175 * @sectors: Number of 512 sectors to convert
177 * Description: The block layer calculates everything in 512 byte
178 * sectors but integrity metadata is done in terms of the hardware
179 * sector size of the storage device. Convert the block layer sectors
180 * to physical sectors.
182 static inline unsigned int bio_integrity_hw_sectors(struct blk_integrity *bi,
183 unsigned int sectors)
185 /* At this point there are only 512b or 4096b DIF/EPP devices */
186 if (bi->sector_size == 4096)
187 return sectors >>= 3;
193 * bio_integrity_tag_size - Retrieve integrity tag space
194 * @bio: bio to inspect
196 * Description: Returns the maximum number of tag bytes that can be
197 * attached to this bio. Filesystems can use this to determine how
198 * much metadata to attach to an I/O.
200 unsigned int bio_integrity_tag_size(struct bio *bio)
202 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
204 BUG_ON(bio->bi_size == 0);
206 return bi->tag_size * (bio->bi_size / bi->sector_size);
208 EXPORT_SYMBOL(bio_integrity_tag_size);
210 int bio_integrity_tag(struct bio *bio, void *tag_buf, unsigned int len, int set)
212 struct bio_integrity_payload *bip = bio->bi_integrity;
213 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
214 unsigned int nr_sectors;
216 BUG_ON(bip->bip_buf == NULL);
218 if (bi->tag_size == 0)
221 nr_sectors = bio_integrity_hw_sectors(bi,
222 DIV_ROUND_UP(len, bi->tag_size));
224 if (nr_sectors * bi->tuple_size > bip->bip_size) {
225 printk(KERN_ERR "%s: tag too big for bio: %u > %u\n",
226 __func__, nr_sectors * bi->tuple_size, bip->bip_size);
231 bi->set_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
233 bi->get_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
239 * bio_integrity_set_tag - Attach a tag buffer to a bio
240 * @bio: bio to attach buffer to
241 * @tag_buf: Pointer to a buffer containing tag data
242 * @len: Length of the included buffer
244 * Description: Use this function to tag a bio by leveraging the extra
245 * space provided by devices formatted with integrity protection. The
246 * size of the integrity buffer must be <= to the size reported by
247 * bio_integrity_tag_size().
249 int bio_integrity_set_tag(struct bio *bio, void *tag_buf, unsigned int len)
251 BUG_ON(bio_data_dir(bio) != WRITE);
253 return bio_integrity_tag(bio, tag_buf, len, 1);
255 EXPORT_SYMBOL(bio_integrity_set_tag);
258 * bio_integrity_get_tag - Retrieve a tag buffer from a bio
259 * @bio: bio to retrieve buffer from
260 * @tag_buf: Pointer to a buffer for the tag data
261 * @len: Length of the target buffer
263 * Description: Use this function to retrieve the tag buffer from a
264 * completed I/O. The size of the integrity buffer must be <= to the
265 * size reported by bio_integrity_tag_size().
267 int bio_integrity_get_tag(struct bio *bio, void *tag_buf, unsigned int len)
269 BUG_ON(bio_data_dir(bio) != READ);
271 return bio_integrity_tag(bio, tag_buf, len, 0);
273 EXPORT_SYMBOL(bio_integrity_get_tag);
276 * bio_integrity_generate - Generate integrity metadata for a bio
277 * @bio: bio to generate integrity metadata for
279 * Description: Generates integrity metadata for a bio by calling the
280 * block device's generation callback function. The bio must have a
281 * bip attached with enough room to accommodate the generated
282 * integrity metadata.
284 static void bio_integrity_generate(struct bio *bio)
286 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
287 struct blk_integrity_exchg bix;
289 sector_t sector = bio->bi_sector;
290 unsigned int i, sectors, total;
291 void *prot_buf = bio->bi_integrity->bip_buf;
294 bix.disk_name = bio->bi_bdev->bd_disk->disk_name;
295 bix.sector_size = bi->sector_size;
297 bio_for_each_segment(bv, bio, i) {
298 void *kaddr = kmap_atomic(bv->bv_page, KM_USER0);
299 bix.data_buf = kaddr + bv->bv_offset;
300 bix.data_size = bv->bv_len;
301 bix.prot_buf = prot_buf;
304 bi->generate_fn(&bix);
306 sectors = bv->bv_len / bi->sector_size;
308 prot_buf += sectors * bi->tuple_size;
309 total += sectors * bi->tuple_size;
310 BUG_ON(total > bio->bi_integrity->bip_size);
312 kunmap_atomic(kaddr, KM_USER0);
317 * bio_integrity_prep - Prepare bio for integrity I/O
318 * @bio: bio to prepare
320 * Description: Allocates a buffer for integrity metadata, maps the
321 * pages and attaches them to a bio. The bio must have data
322 * direction, target device and start sector set priot to calling. In
323 * the WRITE case, integrity metadata will be generated using the
324 * block device's integrity function. In the READ case, the buffer
325 * will be prepared for DMA and a suitable end_io handler set up.
327 int bio_integrity_prep(struct bio *bio)
329 struct bio_integrity_payload *bip;
330 struct blk_integrity *bi;
331 struct request_queue *q;
333 unsigned long start, end;
334 unsigned int len, nr_pages;
335 unsigned int bytes, offset, i;
336 unsigned int sectors;
338 bi = bdev_get_integrity(bio->bi_bdev);
339 q = bdev_get_queue(bio->bi_bdev);
341 BUG_ON(bio_integrity(bio));
343 sectors = bio_integrity_hw_sectors(bi, bio_sectors(bio));
345 /* Allocate kernel buffer for protection data */
346 len = sectors * blk_integrity_tuple_size(bi);
347 buf = kmalloc(len, GFP_NOIO | __GFP_NOFAIL | q->bounce_gfp);
348 if (unlikely(buf == NULL)) {
349 printk(KERN_ERR "could not allocate integrity buffer\n");
353 end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
354 start = ((unsigned long) buf) >> PAGE_SHIFT;
355 nr_pages = end - start;
357 /* Allocate bio integrity payload and integrity vectors */
358 bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages);
359 if (unlikely(bip == NULL)) {
360 printk(KERN_ERR "could not allocate data integrity bioset\n");
367 bip->bip_sector = bio->bi_sector;
370 offset = offset_in_page(buf);
371 for (i = 0 ; i < nr_pages ; i++) {
373 bytes = PAGE_SIZE - offset;
381 ret = bio_integrity_add_page(bio, virt_to_page(buf),
395 /* Install custom I/O completion handler if read verify is enabled */
396 if (bio_data_dir(bio) == READ) {
397 bip->bip_end_io = bio->bi_end_io;
398 bio->bi_end_io = bio_integrity_endio;
401 /* Auto-generate integrity metadata if this is a write */
402 if (bio_data_dir(bio) == WRITE)
403 bio_integrity_generate(bio);
407 EXPORT_SYMBOL(bio_integrity_prep);
410 * bio_integrity_verify - Verify integrity metadata for a bio
411 * @bio: bio to verify
413 * Description: This function is called to verify the integrity of a
414 * bio. The data in the bio io_vec is compared to the integrity
415 * metadata returned by the HBA.
417 static int bio_integrity_verify(struct bio *bio)
419 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
420 struct blk_integrity_exchg bix;
422 sector_t sector = bio->bi_integrity->bip_sector;
423 unsigned int i, sectors, total, ret;
424 void *prot_buf = bio->bi_integrity->bip_buf;
427 bix.disk_name = bio->bi_bdev->bd_disk->disk_name;
428 bix.sector_size = bi->sector_size;
430 bio_for_each_segment(bv, bio, i) {
431 void *kaddr = kmap_atomic(bv->bv_page, KM_USER0);
432 bix.data_buf = kaddr + bv->bv_offset;
433 bix.data_size = bv->bv_len;
434 bix.prot_buf = prot_buf;
437 ret = bi->verify_fn(&bix);
440 kunmap_atomic(kaddr, KM_USER0);
444 sectors = bv->bv_len / bi->sector_size;
446 prot_buf += sectors * bi->tuple_size;
447 total += sectors * bi->tuple_size;
448 BUG_ON(total > bio->bi_integrity->bip_size);
450 kunmap_atomic(kaddr, KM_USER0);
457 * bio_integrity_verify_fn - Integrity I/O completion worker
458 * @work: Work struct stored in bio to be verified
460 * Description: This workqueue function is called to complete a READ
461 * request. The function verifies the transferred integrity metadata
462 * and then calls the original bio end_io function.
464 static void bio_integrity_verify_fn(struct work_struct *work)
466 struct bio_integrity_payload *bip =
467 container_of(work, struct bio_integrity_payload, bip_work);
468 struct bio *bio = bip->bip_bio;
469 int error = bip->bip_error;
471 if (bio_integrity_verify(bio)) {
472 clear_bit(BIO_UPTODATE, &bio->bi_flags);
476 /* Restore original bio completion handler */
477 bio->bi_end_io = bip->bip_end_io;
480 bio->bi_end_io(bio, error);
484 * bio_integrity_endio - Integrity I/O completion function
485 * @bio: Protected bio
486 * @error: Pointer to errno
488 * Description: Completion for integrity I/O
490 * Normally I/O completion is done in interrupt context. However,
491 * verifying I/O integrity is a time-consuming task which must be run
492 * in process context. This function postpones completion
495 void bio_integrity_endio(struct bio *bio, int error)
497 struct bio_integrity_payload *bip = bio->bi_integrity;
499 BUG_ON(bip->bip_bio != bio);
501 bip->bip_error = error;
502 INIT_WORK(&bip->bip_work, bio_integrity_verify_fn);
503 queue_work(kintegrityd_wq, &bip->bip_work);
505 EXPORT_SYMBOL(bio_integrity_endio);
508 * bio_integrity_mark_head - Advance bip_vec skip bytes
509 * @bip: Integrity vector to advance
510 * @skip: Number of bytes to advance it
512 void bio_integrity_mark_head(struct bio_integrity_payload *bip,
518 bip_for_each_vec(iv, bip, i) {
522 } else if (skip >= iv->bv_len) {
524 } else { /* skip < iv->bv_len) */
525 iv->bv_offset += skip;
534 * bio_integrity_mark_tail - Truncate bip_vec to be len bytes long
535 * @bip: Integrity vector to truncate
536 * @len: New length of integrity vector
538 void bio_integrity_mark_tail(struct bio_integrity_payload *bip,
544 bip_for_each_vec(iv, bip, i) {
548 } else if (len >= iv->bv_len) {
550 } else { /* len < iv->bv_len) */
558 * bio_integrity_advance - Advance integrity vector
559 * @bio: bio whose integrity vector to update
560 * @bytes_done: number of data bytes that have been completed
562 * Description: This function calculates how many integrity bytes the
563 * number of completed data bytes correspond to and advances the
564 * integrity vector accordingly.
566 void bio_integrity_advance(struct bio *bio, unsigned int bytes_done)
568 struct bio_integrity_payload *bip = bio->bi_integrity;
569 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
570 unsigned int nr_sectors;
575 nr_sectors = bio_integrity_hw_sectors(bi, bytes_done >> 9);
576 bio_integrity_mark_head(bip, nr_sectors * bi->tuple_size);
578 EXPORT_SYMBOL(bio_integrity_advance);
581 * bio_integrity_trim - Trim integrity vector
582 * @bio: bio whose integrity vector to update
583 * @offset: offset to first data sector
584 * @sectors: number of data sectors
586 * Description: Used to trim the integrity vector in a cloned bio.
587 * The ivec will be advanced corresponding to 'offset' data sectors
588 * and the length will be truncated corresponding to 'len' data
591 void bio_integrity_trim(struct bio *bio, unsigned int offset,
592 unsigned int sectors)
594 struct bio_integrity_payload *bip = bio->bi_integrity;
595 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
596 unsigned int nr_sectors;
600 BUG_ON(!bio_flagged(bio, BIO_CLONED));
602 nr_sectors = bio_integrity_hw_sectors(bi, sectors);
603 bip->bip_sector = bip->bip_sector + offset;
604 bio_integrity_mark_head(bip, offset * bi->tuple_size);
605 bio_integrity_mark_tail(bip, sectors * bi->tuple_size);
607 EXPORT_SYMBOL(bio_integrity_trim);
610 * bio_integrity_split - Split integrity metadata
611 * @bio: Protected bio
612 * @bp: Resulting bio_pair
615 * Description: Splits an integrity page into a bio_pair.
617 void bio_integrity_split(struct bio *bio, struct bio_pair *bp, int sectors)
619 struct blk_integrity *bi;
620 struct bio_integrity_payload *bip = bio->bi_integrity;
621 unsigned int nr_sectors;
623 if (bio_integrity(bio) == 0)
626 bi = bdev_get_integrity(bio->bi_bdev);
628 BUG_ON(bip->bip_vcnt != 1);
630 nr_sectors = bio_integrity_hw_sectors(bi, sectors);
632 bp->bio1.bi_integrity = &bp->bip1;
633 bp->bio2.bi_integrity = &bp->bip2;
635 bp->iv1 = bip->bip_vec[0];
636 bp->iv2 = bip->bip_vec[0];
638 bp->bip1.bip_vec = &bp->iv1;
639 bp->bip2.bip_vec = &bp->iv2;
641 bp->iv1.bv_len = sectors * bi->tuple_size;
642 bp->iv2.bv_offset += sectors * bi->tuple_size;
643 bp->iv2.bv_len -= sectors * bi->tuple_size;
645 bp->bip1.bip_sector = bio->bi_integrity->bip_sector;
646 bp->bip2.bip_sector = bio->bi_integrity->bip_sector + nr_sectors;
648 bp->bip1.bip_vcnt = bp->bip2.bip_vcnt = 1;
649 bp->bip1.bip_idx = bp->bip2.bip_idx = 0;
651 EXPORT_SYMBOL(bio_integrity_split);
654 * bio_integrity_clone - Callback for cloning bios with integrity metadata
656 * @bio_src: Original bio
657 * @bs: bio_set to allocate bip from
659 * Description: Called to allocate a bip when cloning a bio
661 int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
664 struct bio_integrity_payload *bip_src = bio_src->bi_integrity;
665 struct bio_integrity_payload *bip;
667 BUG_ON(bip_src == NULL);
669 bip = bio_integrity_alloc_bioset(bio, GFP_NOIO, bip_src->bip_vcnt, bs);
674 memcpy(bip->bip_vec, bip_src->bip_vec,
675 bip_src->bip_vcnt * sizeof(struct bio_vec));
677 bip->bip_sector = bip_src->bip_sector;
678 bip->bip_vcnt = bip_src->bip_vcnt;
679 bip->bip_idx = bip_src->bip_idx;
683 EXPORT_SYMBOL(bio_integrity_clone);
685 int bioset_integrity_create(struct bio_set *bs, int pool_size)
687 bs->bio_integrity_pool = mempool_create_slab_pool(pool_size,
689 if (!bs->bio_integrity_pool)
694 EXPORT_SYMBOL(bioset_integrity_create);
696 void bioset_integrity_free(struct bio_set *bs)
698 if (bs->bio_integrity_pool)
699 mempool_destroy(bs->bio_integrity_pool);
701 EXPORT_SYMBOL(bioset_integrity_free);
703 void __init bio_integrity_init_slab(void)
705 bio_integrity_slab = KMEM_CACHE(bio_integrity_payload,
706 SLAB_HWCACHE_ALIGN|SLAB_PANIC);
709 static int __init integrity_init(void)
711 kintegrityd_wq = create_workqueue("kintegrityd");
714 panic("Failed to create kintegrityd\n");
718 subsys_initcall(integrity_init);