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) && !bio_flagged(bio, BIO_FS_INTEGRITY)
111 && bip->bip_buf != NULL)
114 bvec_free_bs(bs, bip->bip_vec, bip->bip_pool);
115 mempool_free(bip, bs->bio_integrity_pool);
117 bio->bi_integrity = NULL;
119 EXPORT_SYMBOL(bio_integrity_free);
122 * bio_integrity_add_page - Attach integrity metadata
123 * @bio: bio to update
124 * @page: page containing integrity metadata
125 * @len: number of bytes of integrity metadata in page
126 * @offset: start offset within page
128 * Description: Attach a page containing integrity metadata to bio.
130 int bio_integrity_add_page(struct bio *bio, struct page *page,
131 unsigned int len, unsigned int offset)
133 struct bio_integrity_payload *bip = bio->bi_integrity;
136 if (bip->bip_vcnt >= bvec_nr_vecs(bip->bip_pool)) {
137 printk(KERN_ERR "%s: bip_vec full\n", __func__);
141 iv = bip_vec_idx(bip, bip->bip_vcnt);
146 iv->bv_offset = offset;
151 EXPORT_SYMBOL(bio_integrity_add_page);
153 static int bdev_integrity_enabled(struct block_device *bdev, int rw)
155 struct blk_integrity *bi = bdev_get_integrity(bdev);
160 if (rw == READ && bi->verify_fn != NULL &&
161 (bi->flags & INTEGRITY_FLAG_READ))
164 if (rw == WRITE && bi->generate_fn != NULL &&
165 (bi->flags & INTEGRITY_FLAG_WRITE))
172 * bio_integrity_enabled - Check whether integrity can be passed
175 * Description: Determines whether bio_integrity_prep() can be called
176 * on this bio or not. bio data direction and target device must be
177 * set prior to calling. The functions honors the write_generate and
178 * read_verify flags in sysfs.
180 int bio_integrity_enabled(struct bio *bio)
182 /* Already protected? */
183 if (bio_integrity(bio))
186 return bdev_integrity_enabled(bio->bi_bdev, bio_data_dir(bio));
188 EXPORT_SYMBOL(bio_integrity_enabled);
191 * bio_integrity_hw_sectors - Convert 512b sectors to hardware ditto
192 * @bi: blk_integrity profile for device
193 * @sectors: Number of 512 sectors to convert
195 * Description: The block layer calculates everything in 512 byte
196 * sectors but integrity metadata is done in terms of the hardware
197 * sector size of the storage device. Convert the block layer sectors
198 * to physical sectors.
200 static inline unsigned int bio_integrity_hw_sectors(struct blk_integrity *bi,
201 unsigned int sectors)
203 /* At this point there are only 512b or 4096b DIF/EPP devices */
204 if (bi->sector_size == 4096)
205 return sectors >>= 3;
211 * bio_integrity_tag_size - Retrieve integrity tag space
212 * @bio: bio to inspect
214 * Description: Returns the maximum number of tag bytes that can be
215 * attached to this bio. Filesystems can use this to determine how
216 * much metadata to attach to an I/O.
218 unsigned int bio_integrity_tag_size(struct bio *bio)
220 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
222 BUG_ON(bio->bi_size == 0);
224 return bi->tag_size * (bio->bi_size / bi->sector_size);
226 EXPORT_SYMBOL(bio_integrity_tag_size);
228 int bio_integrity_tag(struct bio *bio, void *tag_buf, unsigned int len, int set)
230 struct bio_integrity_payload *bip = bio->bi_integrity;
231 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
232 unsigned int nr_sectors;
234 BUG_ON(bip->bip_buf == NULL);
236 if (bi->tag_size == 0)
239 nr_sectors = bio_integrity_hw_sectors(bi,
240 DIV_ROUND_UP(len, bi->tag_size));
242 if (nr_sectors * bi->tuple_size > bip->bip_size) {
243 printk(KERN_ERR "%s: tag too big for bio: %u > %u\n",
244 __func__, nr_sectors * bi->tuple_size, bip->bip_size);
249 bi->set_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
251 bi->get_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
257 * bio_integrity_set_tag - Attach a tag buffer to a bio
258 * @bio: bio to attach buffer to
259 * @tag_buf: Pointer to a buffer containing tag data
260 * @len: Length of the included buffer
262 * Description: Use this function to tag a bio by leveraging the extra
263 * space provided by devices formatted with integrity protection. The
264 * size of the integrity buffer must be <= to the size reported by
265 * bio_integrity_tag_size().
267 int bio_integrity_set_tag(struct bio *bio, void *tag_buf, unsigned int len)
269 BUG_ON(bio_data_dir(bio) != WRITE);
271 return bio_integrity_tag(bio, tag_buf, len, 1);
273 EXPORT_SYMBOL(bio_integrity_set_tag);
276 * bio_integrity_get_tag - Retrieve a tag buffer from a bio
277 * @bio: bio to retrieve buffer from
278 * @tag_buf: Pointer to a buffer for the tag data
279 * @len: Length of the target buffer
281 * Description: Use this function to retrieve the tag buffer from a
282 * completed I/O. The size of the integrity buffer must be <= to the
283 * size reported by bio_integrity_tag_size().
285 int bio_integrity_get_tag(struct bio *bio, void *tag_buf, unsigned int len)
287 BUG_ON(bio_data_dir(bio) != READ);
289 return bio_integrity_tag(bio, tag_buf, len, 0);
291 EXPORT_SYMBOL(bio_integrity_get_tag);
294 * bio_integrity_generate - Generate integrity metadata for a bio
295 * @bio: bio to generate integrity metadata for
297 * Description: Generates integrity metadata for a bio by calling the
298 * block device's generation callback function. The bio must have a
299 * bip attached with enough room to accommodate the generated
300 * integrity metadata.
302 static void bio_integrity_generate(struct bio *bio)
304 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
305 struct blk_integrity_exchg bix;
307 sector_t sector = bio->bi_sector;
308 unsigned int i, sectors, total;
309 void *prot_buf = bio->bi_integrity->bip_buf;
312 bix.disk_name = bio->bi_bdev->bd_disk->disk_name;
313 bix.sector_size = bi->sector_size;
315 bio_for_each_segment(bv, bio, i) {
316 void *kaddr = kmap_atomic(bv->bv_page, KM_USER0);
317 bix.data_buf = kaddr + bv->bv_offset;
318 bix.data_size = bv->bv_len;
319 bix.prot_buf = prot_buf;
322 bi->generate_fn(&bix);
324 sectors = bv->bv_len / bi->sector_size;
326 prot_buf += sectors * bi->tuple_size;
327 total += sectors * bi->tuple_size;
328 BUG_ON(total > bio->bi_integrity->bip_size);
330 kunmap_atomic(kaddr, KM_USER0);
334 static inline unsigned short blk_integrity_tuple_size(struct blk_integrity *bi)
337 return bi->tuple_size;
343 * bio_integrity_prep - Prepare bio for integrity I/O
344 * @bio: bio to prepare
346 * Description: Allocates a buffer for integrity metadata, maps the
347 * pages and attaches them to a bio. The bio must have data
348 * direction, target device and start sector set priot to calling. In
349 * the WRITE case, integrity metadata will be generated using the
350 * block device's integrity function. In the READ case, the buffer
351 * will be prepared for DMA and a suitable end_io handler set up.
353 int bio_integrity_prep(struct bio *bio)
355 struct bio_integrity_payload *bip;
356 struct blk_integrity *bi;
357 struct request_queue *q;
359 unsigned long start, end;
360 unsigned int len, nr_pages;
361 unsigned int bytes, offset, i;
362 unsigned int sectors;
364 bi = bdev_get_integrity(bio->bi_bdev);
365 q = bdev_get_queue(bio->bi_bdev);
367 BUG_ON(bio_integrity(bio));
369 sectors = bio_integrity_hw_sectors(bi, bio_sectors(bio));
371 /* Allocate kernel buffer for protection data */
372 len = sectors * blk_integrity_tuple_size(bi);
373 buf = kmalloc(len, GFP_NOIO | __GFP_NOFAIL | q->bounce_gfp);
374 if (unlikely(buf == NULL)) {
375 printk(KERN_ERR "could not allocate integrity buffer\n");
379 end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
380 start = ((unsigned long) buf) >> PAGE_SHIFT;
381 nr_pages = end - start;
383 /* Allocate bio integrity payload and integrity vectors */
384 bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages);
385 if (unlikely(bip == NULL)) {
386 printk(KERN_ERR "could not allocate data integrity bioset\n");
393 bip->bip_sector = bio->bi_sector;
396 offset = offset_in_page(buf);
397 for (i = 0 ; i < nr_pages ; i++) {
399 bytes = PAGE_SIZE - offset;
407 ret = bio_integrity_add_page(bio, virt_to_page(buf),
421 /* Install custom I/O completion handler if read verify is enabled */
422 if (bio_data_dir(bio) == READ) {
423 bip->bip_end_io = bio->bi_end_io;
424 bio->bi_end_io = bio_integrity_endio;
427 /* Auto-generate integrity metadata if this is a write */
428 if (bio_data_dir(bio) == WRITE)
429 bio_integrity_generate(bio);
433 EXPORT_SYMBOL(bio_integrity_prep);
436 * bio_integrity_verify - Verify integrity metadata for a bio
437 * @bio: bio to verify
439 * Description: This function is called to verify the integrity of a
440 * bio. The data in the bio io_vec is compared to the integrity
441 * metadata returned by the HBA.
443 static int bio_integrity_verify(struct bio *bio)
445 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
446 struct blk_integrity_exchg bix;
448 sector_t sector = bio->bi_integrity->bip_sector;
449 unsigned int i, sectors, total, ret;
450 void *prot_buf = bio->bi_integrity->bip_buf;
453 bix.disk_name = bio->bi_bdev->bd_disk->disk_name;
454 bix.sector_size = bi->sector_size;
456 bio_for_each_segment(bv, bio, i) {
457 void *kaddr = kmap_atomic(bv->bv_page, KM_USER0);
458 bix.data_buf = kaddr + bv->bv_offset;
459 bix.data_size = bv->bv_len;
460 bix.prot_buf = prot_buf;
463 ret = bi->verify_fn(&bix);
466 kunmap_atomic(kaddr, KM_USER0);
470 sectors = bv->bv_len / bi->sector_size;
472 prot_buf += sectors * bi->tuple_size;
473 total += sectors * bi->tuple_size;
474 BUG_ON(total > bio->bi_integrity->bip_size);
476 kunmap_atomic(kaddr, KM_USER0);
483 * bio_integrity_verify_fn - Integrity I/O completion worker
484 * @work: Work struct stored in bio to be verified
486 * Description: This workqueue function is called to complete a READ
487 * request. The function verifies the transferred integrity metadata
488 * and then calls the original bio end_io function.
490 static void bio_integrity_verify_fn(struct work_struct *work)
492 struct bio_integrity_payload *bip =
493 container_of(work, struct bio_integrity_payload, bip_work);
494 struct bio *bio = bip->bip_bio;
497 error = bio_integrity_verify(bio);
499 /* Restore original bio completion handler */
500 bio->bi_end_io = bip->bip_end_io;
501 bio_endio(bio, error);
505 * bio_integrity_endio - Integrity I/O completion function
506 * @bio: Protected bio
507 * @error: Pointer to errno
509 * Description: Completion for integrity I/O
511 * Normally I/O completion is done in interrupt context. However,
512 * verifying I/O integrity is a time-consuming task which must be run
513 * in process context. This function postpones completion
516 void bio_integrity_endio(struct bio *bio, int error)
518 struct bio_integrity_payload *bip = bio->bi_integrity;
520 BUG_ON(bip->bip_bio != bio);
522 /* In case of an I/O error there is no point in verifying the
523 * integrity metadata. Restore original bio end_io handler
527 bio->bi_end_io = bip->bip_end_io;
528 bio_endio(bio, error);
533 INIT_WORK(&bip->bip_work, bio_integrity_verify_fn);
534 queue_work(kintegrityd_wq, &bip->bip_work);
536 EXPORT_SYMBOL(bio_integrity_endio);
539 * bio_integrity_mark_head - Advance bip_vec skip bytes
540 * @bip: Integrity vector to advance
541 * @skip: Number of bytes to advance it
543 void bio_integrity_mark_head(struct bio_integrity_payload *bip,
549 bip_for_each_vec(iv, bip, i) {
553 } else if (skip >= iv->bv_len) {
555 } else { /* skip < iv->bv_len) */
556 iv->bv_offset += skip;
565 * bio_integrity_mark_tail - Truncate bip_vec to be len bytes long
566 * @bip: Integrity vector to truncate
567 * @len: New length of integrity vector
569 void bio_integrity_mark_tail(struct bio_integrity_payload *bip,
575 bip_for_each_vec(iv, bip, i) {
579 } else if (len >= iv->bv_len) {
581 } else { /* len < iv->bv_len) */
589 * bio_integrity_advance - Advance integrity vector
590 * @bio: bio whose integrity vector to update
591 * @bytes_done: number of data bytes that have been completed
593 * Description: This function calculates how many integrity bytes the
594 * number of completed data bytes correspond to and advances the
595 * integrity vector accordingly.
597 void bio_integrity_advance(struct bio *bio, unsigned int bytes_done)
599 struct bio_integrity_payload *bip = bio->bi_integrity;
600 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
601 unsigned int nr_sectors;
606 nr_sectors = bio_integrity_hw_sectors(bi, bytes_done >> 9);
607 bio_integrity_mark_head(bip, nr_sectors * bi->tuple_size);
609 EXPORT_SYMBOL(bio_integrity_advance);
612 * bio_integrity_trim - Trim integrity vector
613 * @bio: bio whose integrity vector to update
614 * @offset: offset to first data sector
615 * @sectors: number of data sectors
617 * Description: Used to trim the integrity vector in a cloned bio.
618 * The ivec will be advanced corresponding to 'offset' data sectors
619 * and the length will be truncated corresponding to 'len' data
622 void bio_integrity_trim(struct bio *bio, unsigned int offset,
623 unsigned int sectors)
625 struct bio_integrity_payload *bip = bio->bi_integrity;
626 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
627 unsigned int nr_sectors;
631 BUG_ON(!bio_flagged(bio, BIO_CLONED));
633 nr_sectors = bio_integrity_hw_sectors(bi, sectors);
634 bip->bip_sector = bip->bip_sector + offset;
635 bio_integrity_mark_head(bip, offset * bi->tuple_size);
636 bio_integrity_mark_tail(bip, sectors * bi->tuple_size);
638 EXPORT_SYMBOL(bio_integrity_trim);
641 * bio_integrity_split - Split integrity metadata
642 * @bio: Protected bio
643 * @bp: Resulting bio_pair
646 * Description: Splits an integrity page into a bio_pair.
648 void bio_integrity_split(struct bio *bio, struct bio_pair *bp, int sectors)
650 struct blk_integrity *bi;
651 struct bio_integrity_payload *bip = bio->bi_integrity;
652 unsigned int nr_sectors;
654 if (bio_integrity(bio) == 0)
657 bi = bdev_get_integrity(bio->bi_bdev);
659 BUG_ON(bip->bip_vcnt != 1);
661 nr_sectors = bio_integrity_hw_sectors(bi, sectors);
663 bp->bio1.bi_integrity = &bp->bip1;
664 bp->bio2.bi_integrity = &bp->bip2;
666 bp->iv1 = bip->bip_vec[0];
667 bp->iv2 = bip->bip_vec[0];
669 bp->bip1.bip_vec = &bp->iv1;
670 bp->bip2.bip_vec = &bp->iv2;
672 bp->iv1.bv_len = sectors * bi->tuple_size;
673 bp->iv2.bv_offset += sectors * bi->tuple_size;
674 bp->iv2.bv_len -= sectors * bi->tuple_size;
676 bp->bip1.bip_sector = bio->bi_integrity->bip_sector;
677 bp->bip2.bip_sector = bio->bi_integrity->bip_sector + nr_sectors;
679 bp->bip1.bip_vcnt = bp->bip2.bip_vcnt = 1;
680 bp->bip1.bip_idx = bp->bip2.bip_idx = 0;
682 EXPORT_SYMBOL(bio_integrity_split);
685 * bio_integrity_clone - Callback for cloning bios with integrity metadata
687 * @bio_src: Original bio
688 * @gfp_mask: Memory allocation mask
689 * @bs: bio_set to allocate bip from
691 * Description: Called to allocate a bip when cloning a bio
693 int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
694 gfp_t gfp_mask, struct bio_set *bs)
696 struct bio_integrity_payload *bip_src = bio_src->bi_integrity;
697 struct bio_integrity_payload *bip;
699 BUG_ON(bip_src == NULL);
701 bip = bio_integrity_alloc_bioset(bio, gfp_mask, bip_src->bip_vcnt, bs);
706 memcpy(bip->bip_vec, bip_src->bip_vec,
707 bip_src->bip_vcnt * sizeof(struct bio_vec));
709 bip->bip_sector = bip_src->bip_sector;
710 bip->bip_vcnt = bip_src->bip_vcnt;
711 bip->bip_idx = bip_src->bip_idx;
715 EXPORT_SYMBOL(bio_integrity_clone);
717 int bioset_integrity_create(struct bio_set *bs, int pool_size)
719 bs->bio_integrity_pool = mempool_create_slab_pool(pool_size,
721 if (!bs->bio_integrity_pool)
726 EXPORT_SYMBOL(bioset_integrity_create);
728 void bioset_integrity_free(struct bio_set *bs)
730 if (bs->bio_integrity_pool)
731 mempool_destroy(bs->bio_integrity_pool);
733 EXPORT_SYMBOL(bioset_integrity_free);
735 void __init bio_integrity_init_slab(void)
737 bio_integrity_slab = KMEM_CACHE(bio_integrity_payload,
738 SLAB_HWCACHE_ALIGN|SLAB_PANIC);
741 static int __init integrity_init(void)
743 kintegrityd_wq = create_workqueue("kintegrityd");
746 panic("Failed to create kintegrityd\n");
750 subsys_initcall(integrity_init);