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 - 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
78 * Description: This function prepares a bio for attaching integrity
79 * metadata. nr_vecs specifies the maximum number of pages containing
80 * integrity metadata that can be attached.
82 struct bio_integrity_payload *bio_integrity_alloc(struct bio *bio,
86 struct bio_integrity_payload *bip;
87 unsigned int idx = vecs_to_idx(nr_vecs);
88 struct bio_set *bs = bio->bi_pool;
96 /* Lower order allocations come straight from slab */
97 if (!use_bip_pool(idx))
98 bip = kmem_cache_alloc(bip_slab[idx].slab, gfp_mask);
100 /* Use mempool if lower order alloc failed or max vecs were requested */
102 idx = BIOVEC_MAX_IDX; /* so we free the payload properly later */
103 bip = mempool_alloc(bs->bio_integrity_pool, gfp_mask);
105 if (unlikely(bip == NULL)) {
106 printk(KERN_ERR "%s: could not alloc bip\n", __func__);
111 memset(bip, 0, sizeof(*bip));
115 bio->bi_integrity = bip;
119 EXPORT_SYMBOL(bio_integrity_alloc);
122 * bio_integrity_free - Free bio integrity payload
123 * @bio: bio containing bip to be freed
125 * Description: Used to free the integrity portion of a bio. Usually
126 * called from bio_free().
128 void bio_integrity_free(struct bio *bio)
130 struct bio_integrity_payload *bip = bio->bi_integrity;
131 struct bio_set *bs = bio->bi_pool;
138 /* A cloned bio doesn't own the integrity metadata */
139 if (!bio_flagged(bio, BIO_CLONED) && !bio_flagged(bio, BIO_FS_INTEGRITY)
140 && bip->bip_buf != NULL)
143 if (use_bip_pool(bip->bip_slab))
144 mempool_free(bip, bs->bio_integrity_pool);
146 kmem_cache_free(bip_slab[bip->bip_slab].slab, bip);
148 bio->bi_integrity = NULL;
150 EXPORT_SYMBOL(bio_integrity_free);
153 * bio_integrity_add_page - Attach integrity metadata
154 * @bio: bio to update
155 * @page: page containing integrity metadata
156 * @len: number of bytes of integrity metadata in page
157 * @offset: start offset within page
159 * Description: Attach a page containing integrity metadata to bio.
161 int bio_integrity_add_page(struct bio *bio, struct page *page,
162 unsigned int len, unsigned int offset)
164 struct bio_integrity_payload *bip = bio->bi_integrity;
167 if (bip->bip_vcnt >= bvec_nr_vecs(bip->bip_slab)) {
168 printk(KERN_ERR "%s: bip_vec full\n", __func__);
172 iv = bip_vec_idx(bip, bip->bip_vcnt);
177 iv->bv_offset = offset;
182 EXPORT_SYMBOL(bio_integrity_add_page);
184 static int bdev_integrity_enabled(struct block_device *bdev, int rw)
186 struct blk_integrity *bi = bdev_get_integrity(bdev);
191 if (rw == READ && bi->verify_fn != NULL &&
192 (bi->flags & INTEGRITY_FLAG_READ))
195 if (rw == WRITE && bi->generate_fn != NULL &&
196 (bi->flags & INTEGRITY_FLAG_WRITE))
203 * bio_integrity_enabled - Check whether integrity can be passed
206 * Description: Determines whether bio_integrity_prep() can be called
207 * on this bio or not. bio data direction and target device must be
208 * set prior to calling. The functions honors the write_generate and
209 * read_verify flags in sysfs.
211 int bio_integrity_enabled(struct bio *bio)
213 /* Already protected? */
214 if (bio_integrity(bio))
217 return bdev_integrity_enabled(bio->bi_bdev, bio_data_dir(bio));
219 EXPORT_SYMBOL(bio_integrity_enabled);
222 * bio_integrity_hw_sectors - Convert 512b sectors to hardware ditto
223 * @bi: blk_integrity profile for device
224 * @sectors: Number of 512 sectors to convert
226 * Description: The block layer calculates everything in 512 byte
227 * sectors but integrity metadata is done in terms of the hardware
228 * sector size of the storage device. Convert the block layer sectors
229 * to physical sectors.
231 static inline unsigned int bio_integrity_hw_sectors(struct blk_integrity *bi,
232 unsigned int sectors)
234 /* At this point there are only 512b or 4096b DIF/EPP devices */
235 if (bi->sector_size == 4096)
236 return sectors >>= 3;
242 * bio_integrity_tag_size - Retrieve integrity tag space
243 * @bio: bio to inspect
245 * Description: Returns the maximum number of tag bytes that can be
246 * attached to this bio. Filesystems can use this to determine how
247 * much metadata to attach to an I/O.
249 unsigned int bio_integrity_tag_size(struct bio *bio)
251 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
253 BUG_ON(bio->bi_size == 0);
255 return bi->tag_size * (bio->bi_size / bi->sector_size);
257 EXPORT_SYMBOL(bio_integrity_tag_size);
259 int bio_integrity_tag(struct bio *bio, void *tag_buf, unsigned int len, int set)
261 struct bio_integrity_payload *bip = bio->bi_integrity;
262 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
263 unsigned int nr_sectors;
265 BUG_ON(bip->bip_buf == NULL);
267 if (bi->tag_size == 0)
270 nr_sectors = bio_integrity_hw_sectors(bi,
271 DIV_ROUND_UP(len, bi->tag_size));
273 if (nr_sectors * bi->tuple_size > bip->bip_size) {
274 printk(KERN_ERR "%s: tag too big for bio: %u > %u\n",
275 __func__, nr_sectors * bi->tuple_size, bip->bip_size);
280 bi->set_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
282 bi->get_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
288 * bio_integrity_set_tag - Attach a tag buffer to a bio
289 * @bio: bio to attach buffer to
290 * @tag_buf: Pointer to a buffer containing tag data
291 * @len: Length of the included buffer
293 * Description: Use this function to tag a bio by leveraging the extra
294 * space provided by devices formatted with integrity protection. The
295 * size of the integrity buffer must be <= to the size reported by
296 * bio_integrity_tag_size().
298 int bio_integrity_set_tag(struct bio *bio, void *tag_buf, unsigned int len)
300 BUG_ON(bio_data_dir(bio) != WRITE);
302 return bio_integrity_tag(bio, tag_buf, len, 1);
304 EXPORT_SYMBOL(bio_integrity_set_tag);
307 * bio_integrity_get_tag - Retrieve a tag buffer from a bio
308 * @bio: bio to retrieve buffer from
309 * @tag_buf: Pointer to a buffer for the tag data
310 * @len: Length of the target buffer
312 * Description: Use this function to retrieve the tag buffer from a
313 * completed I/O. The size of the integrity buffer must be <= to the
314 * size reported by bio_integrity_tag_size().
316 int bio_integrity_get_tag(struct bio *bio, void *tag_buf, unsigned int len)
318 BUG_ON(bio_data_dir(bio) != READ);
320 return bio_integrity_tag(bio, tag_buf, len, 0);
322 EXPORT_SYMBOL(bio_integrity_get_tag);
325 * bio_integrity_generate - Generate integrity metadata for a bio
326 * @bio: bio to generate integrity metadata for
328 * Description: Generates integrity metadata for a bio by calling the
329 * block device's generation callback function. The bio must have a
330 * bip attached with enough room to accommodate the generated
331 * integrity metadata.
333 static void bio_integrity_generate(struct bio *bio)
335 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
336 struct blk_integrity_exchg bix;
338 sector_t sector = bio->bi_sector;
339 unsigned int i, sectors, total;
340 void *prot_buf = bio->bi_integrity->bip_buf;
343 bix.disk_name = bio->bi_bdev->bd_disk->disk_name;
344 bix.sector_size = bi->sector_size;
346 bio_for_each_segment(bv, bio, i) {
347 void *kaddr = kmap_atomic(bv->bv_page);
348 bix.data_buf = kaddr + bv->bv_offset;
349 bix.data_size = bv->bv_len;
350 bix.prot_buf = prot_buf;
353 bi->generate_fn(&bix);
355 sectors = bv->bv_len / bi->sector_size;
357 prot_buf += sectors * bi->tuple_size;
358 total += sectors * bi->tuple_size;
359 BUG_ON(total > bio->bi_integrity->bip_size);
361 kunmap_atomic(kaddr);
365 static inline unsigned short blk_integrity_tuple_size(struct blk_integrity *bi)
368 return bi->tuple_size;
374 * bio_integrity_prep - Prepare bio for integrity I/O
375 * @bio: bio to prepare
377 * Description: Allocates a buffer for integrity metadata, maps the
378 * pages and attaches them to a bio. The bio must have data
379 * direction, target device and start sector set priot to calling. In
380 * the WRITE case, integrity metadata will be generated using the
381 * block device's integrity function. In the READ case, the buffer
382 * will be prepared for DMA and a suitable end_io handler set up.
384 int bio_integrity_prep(struct bio *bio)
386 struct bio_integrity_payload *bip;
387 struct blk_integrity *bi;
388 struct request_queue *q;
390 unsigned long start, end;
391 unsigned int len, nr_pages;
392 unsigned int bytes, offset, i;
393 unsigned int sectors;
395 bi = bdev_get_integrity(bio->bi_bdev);
396 q = bdev_get_queue(bio->bi_bdev);
398 BUG_ON(bio_integrity(bio));
400 sectors = bio_integrity_hw_sectors(bi, bio_sectors(bio));
402 /* Allocate kernel buffer for protection data */
403 len = sectors * blk_integrity_tuple_size(bi);
404 buf = kmalloc(len, GFP_NOIO | q->bounce_gfp);
405 if (unlikely(buf == NULL)) {
406 printk(KERN_ERR "could not allocate integrity buffer\n");
410 end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
411 start = ((unsigned long) buf) >> PAGE_SHIFT;
412 nr_pages = end - start;
414 /* Allocate bio integrity payload and integrity vectors */
415 bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages);
416 if (unlikely(bip == NULL)) {
417 printk(KERN_ERR "could not allocate data integrity bioset\n");
424 bip->bip_sector = bio->bi_sector;
427 offset = offset_in_page(buf);
428 for (i = 0 ; i < nr_pages ; i++) {
430 bytes = PAGE_SIZE - offset;
438 ret = bio_integrity_add_page(bio, virt_to_page(buf),
452 /* Install custom I/O completion handler if read verify is enabled */
453 if (bio_data_dir(bio) == READ) {
454 bip->bip_end_io = bio->bi_end_io;
455 bio->bi_end_io = bio_integrity_endio;
458 /* Auto-generate integrity metadata if this is a write */
459 if (bio_data_dir(bio) == WRITE)
460 bio_integrity_generate(bio);
464 EXPORT_SYMBOL(bio_integrity_prep);
467 * bio_integrity_verify - Verify integrity metadata for a bio
468 * @bio: bio to verify
470 * Description: This function is called to verify the integrity of a
471 * bio. The data in the bio io_vec is compared to the integrity
472 * metadata returned by the HBA.
474 static int bio_integrity_verify(struct bio *bio)
476 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
477 struct blk_integrity_exchg bix;
479 sector_t sector = bio->bi_integrity->bip_sector;
480 unsigned int i, sectors, total, ret;
481 void *prot_buf = bio->bi_integrity->bip_buf;
484 bix.disk_name = bio->bi_bdev->bd_disk->disk_name;
485 bix.sector_size = bi->sector_size;
487 bio_for_each_segment(bv, bio, i) {
488 void *kaddr = kmap_atomic(bv->bv_page);
489 bix.data_buf = kaddr + bv->bv_offset;
490 bix.data_size = bv->bv_len;
491 bix.prot_buf = prot_buf;
494 ret = bi->verify_fn(&bix);
497 kunmap_atomic(kaddr);
501 sectors = bv->bv_len / bi->sector_size;
503 prot_buf += sectors * bi->tuple_size;
504 total += sectors * bi->tuple_size;
505 BUG_ON(total > bio->bi_integrity->bip_size);
507 kunmap_atomic(kaddr);
514 * bio_integrity_verify_fn - Integrity I/O completion worker
515 * @work: Work struct stored in bio to be verified
517 * Description: This workqueue function is called to complete a READ
518 * request. The function verifies the transferred integrity metadata
519 * and then calls the original bio end_io function.
521 static void bio_integrity_verify_fn(struct work_struct *work)
523 struct bio_integrity_payload *bip =
524 container_of(work, struct bio_integrity_payload, bip_work);
525 struct bio *bio = bip->bip_bio;
528 error = bio_integrity_verify(bio);
530 /* Restore original bio completion handler */
531 bio->bi_end_io = bip->bip_end_io;
532 bio_endio(bio, error);
536 * bio_integrity_endio - Integrity I/O completion function
537 * @bio: Protected bio
538 * @error: Pointer to errno
540 * Description: Completion for integrity I/O
542 * Normally I/O completion is done in interrupt context. However,
543 * verifying I/O integrity is a time-consuming task which must be run
544 * in process context. This function postpones completion
547 void bio_integrity_endio(struct bio *bio, int error)
549 struct bio_integrity_payload *bip = bio->bi_integrity;
551 BUG_ON(bip->bip_bio != bio);
553 /* In case of an I/O error there is no point in verifying the
554 * integrity metadata. Restore original bio end_io handler
558 bio->bi_end_io = bip->bip_end_io;
559 bio_endio(bio, error);
564 INIT_WORK(&bip->bip_work, bio_integrity_verify_fn);
565 queue_work(kintegrityd_wq, &bip->bip_work);
567 EXPORT_SYMBOL(bio_integrity_endio);
570 * bio_integrity_mark_head - Advance bip_vec skip bytes
571 * @bip: Integrity vector to advance
572 * @skip: Number of bytes to advance it
574 void bio_integrity_mark_head(struct bio_integrity_payload *bip,
580 bip_for_each_vec(iv, bip, i) {
584 } else if (skip >= iv->bv_len) {
586 } else { /* skip < iv->bv_len) */
587 iv->bv_offset += skip;
596 * bio_integrity_mark_tail - Truncate bip_vec to be len bytes long
597 * @bip: Integrity vector to truncate
598 * @len: New length of integrity vector
600 void bio_integrity_mark_tail(struct bio_integrity_payload *bip,
606 bip_for_each_vec(iv, bip, i) {
610 } else if (len >= iv->bv_len) {
612 } else { /* len < iv->bv_len) */
620 * bio_integrity_advance - Advance integrity vector
621 * @bio: bio whose integrity vector to update
622 * @bytes_done: number of data bytes that have been completed
624 * Description: This function calculates how many integrity bytes the
625 * number of completed data bytes correspond to and advances the
626 * integrity vector accordingly.
628 void bio_integrity_advance(struct bio *bio, unsigned int bytes_done)
630 struct bio_integrity_payload *bip = bio->bi_integrity;
631 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
632 unsigned int nr_sectors;
637 nr_sectors = bio_integrity_hw_sectors(bi, bytes_done >> 9);
638 bio_integrity_mark_head(bip, nr_sectors * bi->tuple_size);
640 EXPORT_SYMBOL(bio_integrity_advance);
643 * bio_integrity_trim - Trim integrity vector
644 * @bio: bio whose integrity vector to update
645 * @offset: offset to first data sector
646 * @sectors: number of data sectors
648 * Description: Used to trim the integrity vector in a cloned bio.
649 * The ivec will be advanced corresponding to 'offset' data sectors
650 * and the length will be truncated corresponding to 'len' data
653 void bio_integrity_trim(struct bio *bio, unsigned int offset,
654 unsigned int sectors)
656 struct bio_integrity_payload *bip = bio->bi_integrity;
657 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
658 unsigned int nr_sectors;
662 BUG_ON(!bio_flagged(bio, BIO_CLONED));
664 nr_sectors = bio_integrity_hw_sectors(bi, sectors);
665 bip->bip_sector = bip->bip_sector + offset;
666 bio_integrity_mark_head(bip, offset * bi->tuple_size);
667 bio_integrity_mark_tail(bip, sectors * bi->tuple_size);
669 EXPORT_SYMBOL(bio_integrity_trim);
672 * bio_integrity_split - Split integrity metadata
673 * @bio: Protected bio
674 * @bp: Resulting bio_pair
677 * Description: Splits an integrity page into a bio_pair.
679 void bio_integrity_split(struct bio *bio, struct bio_pair *bp, int sectors)
681 struct blk_integrity *bi;
682 struct bio_integrity_payload *bip = bio->bi_integrity;
683 unsigned int nr_sectors;
685 if (bio_integrity(bio) == 0)
688 bi = bdev_get_integrity(bio->bi_bdev);
690 BUG_ON(bip->bip_vcnt != 1);
692 nr_sectors = bio_integrity_hw_sectors(bi, sectors);
694 bp->bio1.bi_integrity = &bp->bip1;
695 bp->bio2.bi_integrity = &bp->bip2;
697 bp->iv1 = bip->bip_vec[0];
698 bp->iv2 = bip->bip_vec[0];
700 bp->bip1.bip_vec[0] = bp->iv1;
701 bp->bip2.bip_vec[0] = bp->iv2;
703 bp->iv1.bv_len = sectors * bi->tuple_size;
704 bp->iv2.bv_offset += sectors * bi->tuple_size;
705 bp->iv2.bv_len -= sectors * bi->tuple_size;
707 bp->bip1.bip_sector = bio->bi_integrity->bip_sector;
708 bp->bip2.bip_sector = bio->bi_integrity->bip_sector + nr_sectors;
710 bp->bip1.bip_vcnt = bp->bip2.bip_vcnt = 1;
711 bp->bip1.bip_idx = bp->bip2.bip_idx = 0;
713 EXPORT_SYMBOL(bio_integrity_split);
716 * bio_integrity_clone - Callback for cloning bios with integrity metadata
718 * @bio_src: Original bio
719 * @gfp_mask: Memory allocation mask
721 * Description: Called to allocate a bip when cloning a bio
723 int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
726 struct bio_integrity_payload *bip_src = bio_src->bi_integrity;
727 struct bio_integrity_payload *bip;
729 BUG_ON(bip_src == NULL);
731 bip = bio_integrity_alloc(bio, gfp_mask, bip_src->bip_vcnt);
736 memcpy(bip->bip_vec, bip_src->bip_vec,
737 bip_src->bip_vcnt * sizeof(struct bio_vec));
739 bip->bip_sector = bip_src->bip_sector;
740 bip->bip_vcnt = bip_src->bip_vcnt;
741 bip->bip_idx = bip_src->bip_idx;
745 EXPORT_SYMBOL(bio_integrity_clone);
747 int bioset_integrity_create(struct bio_set *bs, int pool_size)
749 unsigned int max_slab = vecs_to_idx(BIO_MAX_PAGES);
751 if (bs->bio_integrity_pool)
754 bs->bio_integrity_pool =
755 mempool_create_slab_pool(pool_size, bip_slab[max_slab].slab);
757 if (!bs->bio_integrity_pool)
762 EXPORT_SYMBOL(bioset_integrity_create);
764 void bioset_integrity_free(struct bio_set *bs)
766 if (bs->bio_integrity_pool)
767 mempool_destroy(bs->bio_integrity_pool);
769 EXPORT_SYMBOL(bioset_integrity_free);
771 void __init bio_integrity_init(void)
776 * kintegrityd won't block much but may burn a lot of CPU cycles.
777 * Make it highpri CPU intensive wq with max concurrency of 1.
779 kintegrityd_wq = alloc_workqueue("kintegrityd", WQ_MEM_RECLAIM |
780 WQ_HIGHPRI | WQ_CPU_INTENSIVE, 1);
782 panic("Failed to create kintegrityd\n");
784 for (i = 0 ; i < BIOVEC_NR_POOLS ; i++) {
787 size = sizeof(struct bio_integrity_payload)
788 + bip_slab[i].nr_vecs * sizeof(struct bio_vec);
791 kmem_cache_create(bip_slab[i].name, size, 0,
792 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);