2 * bio-integrity.c - bio data integrity extensions
4 * Copyright (C) 2007, 2008, 2009 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/export.h>
26 #include <linux/bio.h>
27 #include <linux/workqueue.h>
28 #include <linux/slab.h>
30 struct integrity_slab {
31 struct kmem_cache *slab;
32 unsigned short nr_vecs;
36 #define IS(x) { .nr_vecs = x, .name = "bip-"__stringify(x) }
37 struct integrity_slab bip_slab[BIOVEC_NR_POOLS] __read_mostly = {
38 IS(1), IS(4), IS(16), IS(64), IS(128), IS(BIO_MAX_PAGES),
42 static struct workqueue_struct *kintegrityd_wq;
44 static inline unsigned int vecs_to_idx(unsigned int nr)
57 case 129 ... BIO_MAX_PAGES:
64 static inline int use_bip_pool(unsigned int idx)
66 if (idx == BIOVEC_MAX_IDX)
73 * bio_integrity_alloc_bioset - Allocate integrity payload and attach it to bio
74 * @bio: bio to attach integrity metadata to
75 * @gfp_mask: Memory allocation mask
76 * @nr_vecs: Number of integrity metadata scatter-gather elements
77 * @bs: bio_set to allocate from
79 * Description: This function prepares a bio for attaching integrity
80 * metadata. nr_vecs specifies the maximum number of pages containing
81 * integrity metadata that can be attached.
83 struct bio_integrity_payload *bio_integrity_alloc_bioset(struct bio *bio,
88 struct bio_integrity_payload *bip;
89 unsigned int idx = vecs_to_idx(nr_vecs);
94 /* Lower order allocations come straight from slab */
95 if (!use_bip_pool(idx))
96 bip = kmem_cache_alloc(bip_slab[idx].slab, gfp_mask);
98 /* Use mempool if lower order alloc failed or max vecs were requested */
100 idx = BIOVEC_MAX_IDX; /* so we free the payload properly later */
101 bip = mempool_alloc(bs->bio_integrity_pool, gfp_mask);
103 if (unlikely(bip == NULL)) {
104 printk(KERN_ERR "%s: could not alloc bip\n", __func__);
109 memset(bip, 0, sizeof(*bip));
113 bio->bi_integrity = bip;
117 EXPORT_SYMBOL(bio_integrity_alloc_bioset);
120 * bio_integrity_alloc - Allocate integrity payload and attach it to bio
121 * @bio: bio to attach integrity metadata to
122 * @gfp_mask: Memory allocation mask
123 * @nr_vecs: Number of integrity metadata scatter-gather elements
125 * Description: This function prepares a bio for attaching integrity
126 * metadata. nr_vecs specifies the maximum number of pages containing
127 * integrity metadata that can be attached.
129 struct bio_integrity_payload *bio_integrity_alloc(struct bio *bio,
131 unsigned int nr_vecs)
133 return bio_integrity_alloc_bioset(bio, gfp_mask, nr_vecs, fs_bio_set);
135 EXPORT_SYMBOL(bio_integrity_alloc);
138 * bio_integrity_free - Free bio integrity payload
139 * @bio: bio containing bip to be freed
140 * @bs: bio_set this bio was allocated from
142 * Description: Used to free the integrity portion of a bio. Usually
143 * called from bio_free().
145 void bio_integrity_free(struct bio *bio, struct bio_set *bs)
147 struct bio_integrity_payload *bip = bio->bi_integrity;
151 /* A cloned bio doesn't own the integrity metadata */
152 if (!bio_flagged(bio, BIO_CLONED) && !bio_flagged(bio, BIO_FS_INTEGRITY)
153 && bip->bip_buf != NULL)
156 if (use_bip_pool(bip->bip_slab))
157 mempool_free(bip, bs->bio_integrity_pool);
159 kmem_cache_free(bip_slab[bip->bip_slab].slab, bip);
161 bio->bi_integrity = NULL;
163 EXPORT_SYMBOL(bio_integrity_free);
166 * bio_integrity_add_page - Attach integrity metadata
167 * @bio: bio to update
168 * @page: page containing integrity metadata
169 * @len: number of bytes of integrity metadata in page
170 * @offset: start offset within page
172 * Description: Attach a page containing integrity metadata to bio.
174 int bio_integrity_add_page(struct bio *bio, struct page *page,
175 unsigned int len, unsigned int offset)
177 struct bio_integrity_payload *bip = bio->bi_integrity;
180 if (bip->bip_vcnt >= bvec_nr_vecs(bip->bip_slab)) {
181 printk(KERN_ERR "%s: bip_vec full\n", __func__);
185 iv = bip_vec_idx(bip, bip->bip_vcnt);
190 iv->bv_offset = offset;
195 EXPORT_SYMBOL(bio_integrity_add_page);
197 static int bdev_integrity_enabled(struct block_device *bdev, int rw)
199 struct blk_integrity *bi = bdev_get_integrity(bdev);
204 if (rw == READ && bi->verify_fn != NULL &&
205 (bi->flags & INTEGRITY_FLAG_READ))
208 if (rw == WRITE && bi->generate_fn != NULL &&
209 (bi->flags & INTEGRITY_FLAG_WRITE))
216 * bio_integrity_enabled - Check whether integrity can be passed
219 * Description: Determines whether bio_integrity_prep() can be called
220 * on this bio or not. bio data direction and target device must be
221 * set prior to calling. The functions honors the write_generate and
222 * read_verify flags in sysfs.
224 int bio_integrity_enabled(struct bio *bio)
226 /* Already protected? */
227 if (bio_integrity(bio))
230 return bdev_integrity_enabled(bio->bi_bdev, bio_data_dir(bio));
232 EXPORT_SYMBOL(bio_integrity_enabled);
235 * bio_integrity_hw_sectors - Convert 512b sectors to hardware ditto
236 * @bi: blk_integrity profile for device
237 * @sectors: Number of 512 sectors to convert
239 * Description: The block layer calculates everything in 512 byte
240 * sectors but integrity metadata is done in terms of the hardware
241 * sector size of the storage device. Convert the block layer sectors
242 * to physical sectors.
244 static inline unsigned int bio_integrity_hw_sectors(struct blk_integrity *bi,
245 unsigned int sectors)
247 /* At this point there are only 512b or 4096b DIF/EPP devices */
248 if (bi->sector_size == 4096)
249 return sectors >>= 3;
255 * bio_integrity_tag_size - Retrieve integrity tag space
256 * @bio: bio to inspect
258 * Description: Returns the maximum number of tag bytes that can be
259 * attached to this bio. Filesystems can use this to determine how
260 * much metadata to attach to an I/O.
262 unsigned int bio_integrity_tag_size(struct bio *bio)
264 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
266 BUG_ON(bio->bi_size == 0);
268 return bi->tag_size * (bio->bi_size / bi->sector_size);
270 EXPORT_SYMBOL(bio_integrity_tag_size);
272 int bio_integrity_tag(struct bio *bio, void *tag_buf, unsigned int len, int set)
274 struct bio_integrity_payload *bip = bio->bi_integrity;
275 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
276 unsigned int nr_sectors;
278 BUG_ON(bip->bip_buf == NULL);
280 if (bi->tag_size == 0)
283 nr_sectors = bio_integrity_hw_sectors(bi,
284 DIV_ROUND_UP(len, bi->tag_size));
286 if (nr_sectors * bi->tuple_size > bip->bip_size) {
287 printk(KERN_ERR "%s: tag too big for bio: %u > %u\n",
288 __func__, nr_sectors * bi->tuple_size, bip->bip_size);
293 bi->set_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
295 bi->get_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
301 * bio_integrity_set_tag - Attach a tag buffer to a bio
302 * @bio: bio to attach buffer to
303 * @tag_buf: Pointer to a buffer containing tag data
304 * @len: Length of the included buffer
306 * Description: Use this function to tag a bio by leveraging the extra
307 * space provided by devices formatted with integrity protection. The
308 * size of the integrity buffer must be <= to the size reported by
309 * bio_integrity_tag_size().
311 int bio_integrity_set_tag(struct bio *bio, void *tag_buf, unsigned int len)
313 BUG_ON(bio_data_dir(bio) != WRITE);
315 return bio_integrity_tag(bio, tag_buf, len, 1);
317 EXPORT_SYMBOL(bio_integrity_set_tag);
320 * bio_integrity_get_tag - Retrieve a tag buffer from a bio
321 * @bio: bio to retrieve buffer from
322 * @tag_buf: Pointer to a buffer for the tag data
323 * @len: Length of the target buffer
325 * Description: Use this function to retrieve the tag buffer from a
326 * completed I/O. The size of the integrity buffer must be <= to the
327 * size reported by bio_integrity_tag_size().
329 int bio_integrity_get_tag(struct bio *bio, void *tag_buf, unsigned int len)
331 BUG_ON(bio_data_dir(bio) != READ);
333 return bio_integrity_tag(bio, tag_buf, len, 0);
335 EXPORT_SYMBOL(bio_integrity_get_tag);
338 * bio_integrity_generate - Generate integrity metadata for a bio
339 * @bio: bio to generate integrity metadata for
341 * Description: Generates integrity metadata for a bio by calling the
342 * block device's generation callback function. The bio must have a
343 * bip attached with enough room to accommodate the generated
344 * integrity metadata.
346 static void bio_integrity_generate(struct bio *bio)
348 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
349 struct blk_integrity_exchg bix;
351 sector_t sector = bio->bi_sector;
352 unsigned int i, sectors, total;
353 void *prot_buf = bio->bi_integrity->bip_buf;
356 bix.disk_name = bio->bi_bdev->bd_disk->disk_name;
357 bix.sector_size = bi->sector_size;
359 bio_for_each_segment(bv, bio, i) {
360 void *kaddr = kmap_atomic(bv->bv_page, KM_USER0);
361 bix.data_buf = kaddr + bv->bv_offset;
362 bix.data_size = bv->bv_len;
363 bix.prot_buf = prot_buf;
366 bi->generate_fn(&bix);
368 sectors = bv->bv_len / bi->sector_size;
370 prot_buf += sectors * bi->tuple_size;
371 total += sectors * bi->tuple_size;
372 BUG_ON(total > bio->bi_integrity->bip_size);
374 kunmap_atomic(kaddr, KM_USER0);
378 static inline unsigned short blk_integrity_tuple_size(struct blk_integrity *bi)
381 return bi->tuple_size;
387 * bio_integrity_prep - Prepare bio for integrity I/O
388 * @bio: bio to prepare
390 * Description: Allocates a buffer for integrity metadata, maps the
391 * pages and attaches them to a bio. The bio must have data
392 * direction, target device and start sector set priot to calling. In
393 * the WRITE case, integrity metadata will be generated using the
394 * block device's integrity function. In the READ case, the buffer
395 * will be prepared for DMA and a suitable end_io handler set up.
397 int bio_integrity_prep(struct bio *bio)
399 struct bio_integrity_payload *bip;
400 struct blk_integrity *bi;
401 struct request_queue *q;
403 unsigned long start, end;
404 unsigned int len, nr_pages;
405 unsigned int bytes, offset, i;
406 unsigned int sectors;
408 bi = bdev_get_integrity(bio->bi_bdev);
409 q = bdev_get_queue(bio->bi_bdev);
411 BUG_ON(bio_integrity(bio));
413 sectors = bio_integrity_hw_sectors(bi, bio_sectors(bio));
415 /* Allocate kernel buffer for protection data */
416 len = sectors * blk_integrity_tuple_size(bi);
417 buf = kmalloc(len, GFP_NOIO | q->bounce_gfp);
418 if (unlikely(buf == NULL)) {
419 printk(KERN_ERR "could not allocate integrity buffer\n");
423 end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
424 start = ((unsigned long) buf) >> PAGE_SHIFT;
425 nr_pages = end - start;
427 /* Allocate bio integrity payload and integrity vectors */
428 bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages);
429 if (unlikely(bip == NULL)) {
430 printk(KERN_ERR "could not allocate data integrity bioset\n");
437 bip->bip_sector = bio->bi_sector;
440 offset = offset_in_page(buf);
441 for (i = 0 ; i < nr_pages ; i++) {
443 bytes = PAGE_SIZE - offset;
451 ret = bio_integrity_add_page(bio, virt_to_page(buf),
465 /* Install custom I/O completion handler if read verify is enabled */
466 if (bio_data_dir(bio) == READ) {
467 bip->bip_end_io = bio->bi_end_io;
468 bio->bi_end_io = bio_integrity_endio;
471 /* Auto-generate integrity metadata if this is a write */
472 if (bio_data_dir(bio) == WRITE)
473 bio_integrity_generate(bio);
477 EXPORT_SYMBOL(bio_integrity_prep);
480 * bio_integrity_verify - Verify integrity metadata for a bio
481 * @bio: bio to verify
483 * Description: This function is called to verify the integrity of a
484 * bio. The data in the bio io_vec is compared to the integrity
485 * metadata returned by the HBA.
487 static int bio_integrity_verify(struct bio *bio)
489 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
490 struct blk_integrity_exchg bix;
492 sector_t sector = bio->bi_integrity->bip_sector;
493 unsigned int i, sectors, total, ret;
494 void *prot_buf = bio->bi_integrity->bip_buf;
497 bix.disk_name = bio->bi_bdev->bd_disk->disk_name;
498 bix.sector_size = bi->sector_size;
500 bio_for_each_segment(bv, bio, i) {
501 void *kaddr = kmap_atomic(bv->bv_page, KM_USER0);
502 bix.data_buf = kaddr + bv->bv_offset;
503 bix.data_size = bv->bv_len;
504 bix.prot_buf = prot_buf;
507 ret = bi->verify_fn(&bix);
510 kunmap_atomic(kaddr, KM_USER0);
514 sectors = bv->bv_len / bi->sector_size;
516 prot_buf += sectors * bi->tuple_size;
517 total += sectors * bi->tuple_size;
518 BUG_ON(total > bio->bi_integrity->bip_size);
520 kunmap_atomic(kaddr, KM_USER0);
527 * bio_integrity_verify_fn - Integrity I/O completion worker
528 * @work: Work struct stored in bio to be verified
530 * Description: This workqueue function is called to complete a READ
531 * request. The function verifies the transferred integrity metadata
532 * and then calls the original bio end_io function.
534 static void bio_integrity_verify_fn(struct work_struct *work)
536 struct bio_integrity_payload *bip =
537 container_of(work, struct bio_integrity_payload, bip_work);
538 struct bio *bio = bip->bip_bio;
541 error = bio_integrity_verify(bio);
543 /* Restore original bio completion handler */
544 bio->bi_end_io = bip->bip_end_io;
545 bio_endio(bio, error);
549 * bio_integrity_endio - Integrity I/O completion function
550 * @bio: Protected bio
551 * @error: Pointer to errno
553 * Description: Completion for integrity I/O
555 * Normally I/O completion is done in interrupt context. However,
556 * verifying I/O integrity is a time-consuming task which must be run
557 * in process context. This function postpones completion
560 void bio_integrity_endio(struct bio *bio, int error)
562 struct bio_integrity_payload *bip = bio->bi_integrity;
564 BUG_ON(bip->bip_bio != bio);
566 /* In case of an I/O error there is no point in verifying the
567 * integrity metadata. Restore original bio end_io handler
571 bio->bi_end_io = bip->bip_end_io;
572 bio_endio(bio, error);
577 INIT_WORK(&bip->bip_work, bio_integrity_verify_fn);
578 queue_work(kintegrityd_wq, &bip->bip_work);
580 EXPORT_SYMBOL(bio_integrity_endio);
583 * bio_integrity_mark_head - Advance bip_vec skip bytes
584 * @bip: Integrity vector to advance
585 * @skip: Number of bytes to advance it
587 void bio_integrity_mark_head(struct bio_integrity_payload *bip,
593 bip_for_each_vec(iv, bip, i) {
597 } else if (skip >= iv->bv_len) {
599 } else { /* skip < iv->bv_len) */
600 iv->bv_offset += skip;
609 * bio_integrity_mark_tail - Truncate bip_vec to be len bytes long
610 * @bip: Integrity vector to truncate
611 * @len: New length of integrity vector
613 void bio_integrity_mark_tail(struct bio_integrity_payload *bip,
619 bip_for_each_vec(iv, bip, i) {
623 } else if (len >= iv->bv_len) {
625 } else { /* len < iv->bv_len) */
633 * bio_integrity_advance - Advance integrity vector
634 * @bio: bio whose integrity vector to update
635 * @bytes_done: number of data bytes that have been completed
637 * Description: This function calculates how many integrity bytes the
638 * number of completed data bytes correspond to and advances the
639 * integrity vector accordingly.
641 void bio_integrity_advance(struct bio *bio, unsigned int bytes_done)
643 struct bio_integrity_payload *bip = bio->bi_integrity;
644 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
645 unsigned int nr_sectors;
650 nr_sectors = bio_integrity_hw_sectors(bi, bytes_done >> 9);
651 bio_integrity_mark_head(bip, nr_sectors * bi->tuple_size);
653 EXPORT_SYMBOL(bio_integrity_advance);
656 * bio_integrity_trim - Trim integrity vector
657 * @bio: bio whose integrity vector to update
658 * @offset: offset to first data sector
659 * @sectors: number of data sectors
661 * Description: Used to trim the integrity vector in a cloned bio.
662 * The ivec will be advanced corresponding to 'offset' data sectors
663 * and the length will be truncated corresponding to 'len' data
666 void bio_integrity_trim(struct bio *bio, unsigned int offset,
667 unsigned int sectors)
669 struct bio_integrity_payload *bip = bio->bi_integrity;
670 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
671 unsigned int nr_sectors;
675 BUG_ON(!bio_flagged(bio, BIO_CLONED));
677 nr_sectors = bio_integrity_hw_sectors(bi, sectors);
678 bip->bip_sector = bip->bip_sector + offset;
679 bio_integrity_mark_head(bip, offset * bi->tuple_size);
680 bio_integrity_mark_tail(bip, sectors * bi->tuple_size);
682 EXPORT_SYMBOL(bio_integrity_trim);
685 * bio_integrity_split - Split integrity metadata
686 * @bio: Protected bio
687 * @bp: Resulting bio_pair
690 * Description: Splits an integrity page into a bio_pair.
692 void bio_integrity_split(struct bio *bio, struct bio_pair *bp, int sectors)
694 struct blk_integrity *bi;
695 struct bio_integrity_payload *bip = bio->bi_integrity;
696 unsigned int nr_sectors;
698 if (bio_integrity(bio) == 0)
701 bi = bdev_get_integrity(bio->bi_bdev);
703 BUG_ON(bip->bip_vcnt != 1);
705 nr_sectors = bio_integrity_hw_sectors(bi, sectors);
707 bp->bio1.bi_integrity = &bp->bip1;
708 bp->bio2.bi_integrity = &bp->bip2;
710 bp->iv1 = bip->bip_vec[0];
711 bp->iv2 = bip->bip_vec[0];
713 bp->bip1.bip_vec[0] = bp->iv1;
714 bp->bip2.bip_vec[0] = bp->iv2;
716 bp->iv1.bv_len = sectors * bi->tuple_size;
717 bp->iv2.bv_offset += sectors * bi->tuple_size;
718 bp->iv2.bv_len -= sectors * bi->tuple_size;
720 bp->bip1.bip_sector = bio->bi_integrity->bip_sector;
721 bp->bip2.bip_sector = bio->bi_integrity->bip_sector + nr_sectors;
723 bp->bip1.bip_vcnt = bp->bip2.bip_vcnt = 1;
724 bp->bip1.bip_idx = bp->bip2.bip_idx = 0;
726 EXPORT_SYMBOL(bio_integrity_split);
729 * bio_integrity_clone - Callback for cloning bios with integrity metadata
731 * @bio_src: Original bio
732 * @gfp_mask: Memory allocation mask
733 * @bs: bio_set to allocate bip from
735 * Description: Called to allocate a bip when cloning a bio
737 int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
738 gfp_t gfp_mask, struct bio_set *bs)
740 struct bio_integrity_payload *bip_src = bio_src->bi_integrity;
741 struct bio_integrity_payload *bip;
743 BUG_ON(bip_src == NULL);
745 bip = bio_integrity_alloc_bioset(bio, gfp_mask, bip_src->bip_vcnt, bs);
750 memcpy(bip->bip_vec, bip_src->bip_vec,
751 bip_src->bip_vcnt * sizeof(struct bio_vec));
753 bip->bip_sector = bip_src->bip_sector;
754 bip->bip_vcnt = bip_src->bip_vcnt;
755 bip->bip_idx = bip_src->bip_idx;
759 EXPORT_SYMBOL(bio_integrity_clone);
761 int bioset_integrity_create(struct bio_set *bs, int pool_size)
763 unsigned int max_slab = vecs_to_idx(BIO_MAX_PAGES);
765 if (bs->bio_integrity_pool)
768 bs->bio_integrity_pool =
769 mempool_create_slab_pool(pool_size, bip_slab[max_slab].slab);
771 if (!bs->bio_integrity_pool)
776 EXPORT_SYMBOL(bioset_integrity_create);
778 void bioset_integrity_free(struct bio_set *bs)
780 if (bs->bio_integrity_pool)
781 mempool_destroy(bs->bio_integrity_pool);
783 EXPORT_SYMBOL(bioset_integrity_free);
785 void __init bio_integrity_init(void)
790 * kintegrityd won't block much but may burn a lot of CPU cycles.
791 * Make it highpri CPU intensive wq with max concurrency of 1.
793 kintegrityd_wq = alloc_workqueue("kintegrityd", WQ_MEM_RECLAIM |
794 WQ_HIGHPRI | WQ_CPU_INTENSIVE, 1);
796 panic("Failed to create kintegrityd\n");
798 for (i = 0 ; i < BIOVEC_NR_POOLS ; i++) {
801 size = sizeof(struct bio_integrity_payload)
802 + bip_slab[i].nr_vecs * sizeof(struct bio_vec);
805 kmem_cache_create(bip_slab[i].name, size, 0,
806 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);