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 #define BIP_INLINE_VECS 4
32 static struct kmem_cache *bip_slab;
33 static struct workqueue_struct *kintegrityd_wq;
36 * bio_integrity_alloc - Allocate integrity payload and attach it to bio
37 * @bio: bio to attach integrity metadata to
38 * @gfp_mask: Memory allocation mask
39 * @nr_vecs: Number of integrity metadata scatter-gather elements
41 * Description: This function prepares a bio for attaching integrity
42 * metadata. nr_vecs specifies the maximum number of pages containing
43 * integrity metadata that can be attached.
45 struct bio_integrity_payload *bio_integrity_alloc(struct bio *bio,
49 struct bio_integrity_payload *bip;
50 struct bio_set *bs = bio->bi_pool;
51 unsigned long idx = BIO_POOL_NONE;
55 bip = kmalloc(sizeof(struct bio_integrity_payload) +
56 sizeof(struct bio_vec) * nr_vecs, gfp_mask);
57 inline_vecs = nr_vecs;
59 bip = mempool_alloc(bs->bio_integrity_pool, gfp_mask);
60 inline_vecs = BIP_INLINE_VECS;
66 memset(bip, 0, sizeof(*bip));
68 if (nr_vecs > inline_vecs) {
69 bip->bip_vec = bvec_alloc(gfp_mask, nr_vecs, &idx,
70 bs->bvec_integrity_pool);
74 bip->bip_vec = bip->bip_inline_vecs;
79 bio->bi_integrity = bip;
83 mempool_free(bip, bs->bio_integrity_pool);
86 EXPORT_SYMBOL(bio_integrity_alloc);
89 * bio_integrity_free - Free bio integrity payload
90 * @bio: bio containing bip to be freed
92 * Description: Used to free the integrity portion of a bio. Usually
93 * called from bio_free().
95 void bio_integrity_free(struct bio *bio)
97 struct bio_integrity_payload *bip = bio->bi_integrity;
98 struct bio_set *bs = bio->bi_pool;
100 if (bip->bip_owns_buf)
104 if (bip->bip_slab != BIO_POOL_NONE)
105 bvec_free(bs->bvec_integrity_pool, bip->bip_vec,
108 mempool_free(bip, bs->bio_integrity_pool);
113 bio->bi_integrity = NULL;
115 EXPORT_SYMBOL(bio_integrity_free);
118 * bio_integrity_add_page - Attach integrity metadata
119 * @bio: bio to update
120 * @page: page containing integrity metadata
121 * @len: number of bytes of integrity metadata in page
122 * @offset: start offset within page
124 * Description: Attach a page containing integrity metadata to bio.
126 int bio_integrity_add_page(struct bio *bio, struct page *page,
127 unsigned int len, unsigned int offset)
129 struct bio_integrity_payload *bip = bio->bi_integrity;
132 if (bip->bip_vcnt >= bvec_nr_vecs(bip->bip_slab)) {
133 printk(KERN_ERR "%s: bip_vec full\n", __func__);
137 iv = bip_vec_idx(bip, bip->bip_vcnt);
142 iv->bv_offset = offset;
147 EXPORT_SYMBOL(bio_integrity_add_page);
149 static int bdev_integrity_enabled(struct block_device *bdev, int rw)
151 struct blk_integrity *bi = bdev_get_integrity(bdev);
156 if (rw == READ && bi->verify_fn != NULL &&
157 (bi->flags & INTEGRITY_FLAG_READ))
160 if (rw == WRITE && bi->generate_fn != NULL &&
161 (bi->flags & INTEGRITY_FLAG_WRITE))
168 * bio_integrity_enabled - Check whether integrity can be passed
171 * Description: Determines whether bio_integrity_prep() can be called
172 * on this bio or not. bio data direction and target device must be
173 * set prior to calling. The functions honors the write_generate and
174 * read_verify flags in sysfs.
176 int bio_integrity_enabled(struct bio *bio)
178 /* Already protected? */
179 if (bio_integrity(bio))
182 return bdev_integrity_enabled(bio->bi_bdev, bio_data_dir(bio));
184 EXPORT_SYMBOL(bio_integrity_enabled);
187 * bio_integrity_hw_sectors - Convert 512b sectors to hardware ditto
188 * @bi: blk_integrity profile for device
189 * @sectors: Number of 512 sectors to convert
191 * Description: The block layer calculates everything in 512 byte
192 * sectors but integrity metadata is done in terms of the hardware
193 * sector size of the storage device. Convert the block layer sectors
194 * to physical sectors.
196 static inline unsigned int bio_integrity_hw_sectors(struct blk_integrity *bi,
197 unsigned int sectors)
199 /* At this point there are only 512b or 4096b DIF/EPP devices */
200 if (bi->sector_size == 4096)
201 return sectors >>= 3;
207 * bio_integrity_tag_size - Retrieve integrity tag space
208 * @bio: bio to inspect
210 * Description: Returns the maximum number of tag bytes that can be
211 * attached to this bio. Filesystems can use this to determine how
212 * much metadata to attach to an I/O.
214 unsigned int bio_integrity_tag_size(struct bio *bio)
216 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
218 BUG_ON(bio->bi_size == 0);
220 return bi->tag_size * (bio->bi_size / bi->sector_size);
222 EXPORT_SYMBOL(bio_integrity_tag_size);
224 int bio_integrity_tag(struct bio *bio, void *tag_buf, unsigned int len, int set)
226 struct bio_integrity_payload *bip = bio->bi_integrity;
227 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
228 unsigned int nr_sectors;
230 BUG_ON(bip->bip_buf == NULL);
232 if (bi->tag_size == 0)
235 nr_sectors = bio_integrity_hw_sectors(bi,
236 DIV_ROUND_UP(len, bi->tag_size));
238 if (nr_sectors * bi->tuple_size > bip->bip_size) {
239 printk(KERN_ERR "%s: tag too big for bio: %u > %u\n",
240 __func__, nr_sectors * bi->tuple_size, bip->bip_size);
245 bi->set_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
247 bi->get_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
253 * bio_integrity_set_tag - Attach a tag buffer to a bio
254 * @bio: bio to attach buffer to
255 * @tag_buf: Pointer to a buffer containing tag data
256 * @len: Length of the included buffer
258 * Description: Use this function to tag a bio by leveraging the extra
259 * space provided by devices formatted with integrity protection. The
260 * size of the integrity buffer must be <= to the size reported by
261 * bio_integrity_tag_size().
263 int bio_integrity_set_tag(struct bio *bio, void *tag_buf, unsigned int len)
265 BUG_ON(bio_data_dir(bio) != WRITE);
267 return bio_integrity_tag(bio, tag_buf, len, 1);
269 EXPORT_SYMBOL(bio_integrity_set_tag);
272 * bio_integrity_get_tag - Retrieve a tag buffer from a bio
273 * @bio: bio to retrieve buffer from
274 * @tag_buf: Pointer to a buffer for the tag data
275 * @len: Length of the target buffer
277 * Description: Use this function to retrieve the tag buffer from a
278 * completed I/O. The size of the integrity buffer must be <= to the
279 * size reported by bio_integrity_tag_size().
281 int bio_integrity_get_tag(struct bio *bio, void *tag_buf, unsigned int len)
283 BUG_ON(bio_data_dir(bio) != READ);
285 return bio_integrity_tag(bio, tag_buf, len, 0);
287 EXPORT_SYMBOL(bio_integrity_get_tag);
290 * bio_integrity_generate - Generate integrity metadata for a bio
291 * @bio: bio to generate integrity metadata for
293 * Description: Generates integrity metadata for a bio by calling the
294 * block device's generation callback function. The bio must have a
295 * bip attached with enough room to accommodate the generated
296 * integrity metadata.
298 static void bio_integrity_generate(struct bio *bio)
300 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
301 struct blk_integrity_exchg bix;
303 sector_t sector = bio->bi_sector;
304 unsigned int i, sectors, total;
305 void *prot_buf = bio->bi_integrity->bip_buf;
308 bix.disk_name = bio->bi_bdev->bd_disk->disk_name;
309 bix.sector_size = bi->sector_size;
311 bio_for_each_segment(bv, bio, i) {
312 void *kaddr = kmap_atomic(bv->bv_page);
313 bix.data_buf = kaddr + bv->bv_offset;
314 bix.data_size = bv->bv_len;
315 bix.prot_buf = prot_buf;
318 bi->generate_fn(&bix);
320 sectors = bv->bv_len / bi->sector_size;
322 prot_buf += sectors * bi->tuple_size;
323 total += sectors * bi->tuple_size;
324 BUG_ON(total > bio->bi_integrity->bip_size);
326 kunmap_atomic(kaddr);
330 static inline unsigned short blk_integrity_tuple_size(struct blk_integrity *bi)
333 return bi->tuple_size;
339 * bio_integrity_prep - Prepare bio for integrity I/O
340 * @bio: bio to prepare
342 * Description: Allocates a buffer for integrity metadata, maps the
343 * pages and attaches them to a bio. The bio must have data
344 * direction, target device and start sector set priot to calling. In
345 * the WRITE case, integrity metadata will be generated using the
346 * block device's integrity function. In the READ case, the buffer
347 * will be prepared for DMA and a suitable end_io handler set up.
349 int bio_integrity_prep(struct bio *bio)
351 struct bio_integrity_payload *bip;
352 struct blk_integrity *bi;
353 struct request_queue *q;
355 unsigned long start, end;
356 unsigned int len, nr_pages;
357 unsigned int bytes, offset, i;
358 unsigned int sectors;
360 bi = bdev_get_integrity(bio->bi_bdev);
361 q = bdev_get_queue(bio->bi_bdev);
363 BUG_ON(bio_integrity(bio));
365 sectors = bio_integrity_hw_sectors(bi, bio_sectors(bio));
367 /* Allocate kernel buffer for protection data */
368 len = sectors * blk_integrity_tuple_size(bi);
369 buf = kmalloc(len, GFP_NOIO | q->bounce_gfp);
370 if (unlikely(buf == NULL)) {
371 printk(KERN_ERR "could not allocate integrity buffer\n");
375 end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
376 start = ((unsigned long) buf) >> PAGE_SHIFT;
377 nr_pages = end - start;
379 /* Allocate bio integrity payload and integrity vectors */
380 bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages);
381 if (unlikely(bip == NULL)) {
382 printk(KERN_ERR "could not allocate data integrity bioset\n");
387 bip->bip_owns_buf = 1;
390 bip->bip_sector = bio->bi_sector;
393 offset = offset_in_page(buf);
394 for (i = 0 ; i < nr_pages ; i++) {
396 bytes = PAGE_SIZE - offset;
404 ret = bio_integrity_add_page(bio, virt_to_page(buf),
418 /* Install custom I/O completion handler if read verify is enabled */
419 if (bio_data_dir(bio) == READ) {
420 bip->bip_end_io = bio->bi_end_io;
421 bio->bi_end_io = bio_integrity_endio;
424 /* Auto-generate integrity metadata if this is a write */
425 if (bio_data_dir(bio) == WRITE)
426 bio_integrity_generate(bio);
430 EXPORT_SYMBOL(bio_integrity_prep);
433 * bio_integrity_verify - Verify integrity metadata for a bio
434 * @bio: bio to verify
436 * Description: This function is called to verify the integrity of a
437 * bio. The data in the bio io_vec is compared to the integrity
438 * metadata returned by the HBA.
440 static int bio_integrity_verify(struct bio *bio)
442 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
443 struct blk_integrity_exchg bix;
445 sector_t sector = bio->bi_integrity->bip_sector;
446 unsigned int i, sectors, total, ret;
447 void *prot_buf = bio->bi_integrity->bip_buf;
450 bix.disk_name = bio->bi_bdev->bd_disk->disk_name;
451 bix.sector_size = bi->sector_size;
453 bio_for_each_segment(bv, bio, i) {
454 void *kaddr = kmap_atomic(bv->bv_page);
455 bix.data_buf = kaddr + bv->bv_offset;
456 bix.data_size = bv->bv_len;
457 bix.prot_buf = prot_buf;
460 ret = bi->verify_fn(&bix);
463 kunmap_atomic(kaddr);
467 sectors = bv->bv_len / bi->sector_size;
469 prot_buf += sectors * bi->tuple_size;
470 total += sectors * bi->tuple_size;
471 BUG_ON(total > bio->bi_integrity->bip_size);
473 kunmap_atomic(kaddr);
480 * bio_integrity_verify_fn - Integrity I/O completion worker
481 * @work: Work struct stored in bio to be verified
483 * Description: This workqueue function is called to complete a READ
484 * request. The function verifies the transferred integrity metadata
485 * and then calls the original bio end_io function.
487 static void bio_integrity_verify_fn(struct work_struct *work)
489 struct bio_integrity_payload *bip =
490 container_of(work, struct bio_integrity_payload, bip_work);
491 struct bio *bio = bip->bip_bio;
494 error = bio_integrity_verify(bio);
496 /* Restore original bio completion handler */
497 bio->bi_end_io = bip->bip_end_io;
498 bio_endio(bio, error);
502 * bio_integrity_endio - Integrity I/O completion function
503 * @bio: Protected bio
504 * @error: Pointer to errno
506 * Description: Completion for integrity I/O
508 * Normally I/O completion is done in interrupt context. However,
509 * verifying I/O integrity is a time-consuming task which must be run
510 * in process context. This function postpones completion
513 void bio_integrity_endio(struct bio *bio, int error)
515 struct bio_integrity_payload *bip = bio->bi_integrity;
517 BUG_ON(bip->bip_bio != bio);
519 /* In case of an I/O error there is no point in verifying the
520 * integrity metadata. Restore original bio end_io handler
524 bio->bi_end_io = bip->bip_end_io;
525 bio_endio(bio, error);
530 INIT_WORK(&bip->bip_work, bio_integrity_verify_fn);
531 queue_work(kintegrityd_wq, &bip->bip_work);
533 EXPORT_SYMBOL(bio_integrity_endio);
536 * bio_integrity_mark_head - Advance bip_vec skip bytes
537 * @bip: Integrity vector to advance
538 * @skip: Number of bytes to advance it
540 void bio_integrity_mark_head(struct bio_integrity_payload *bip,
546 bip_for_each_vec(iv, bip, i) {
550 } else if (skip >= iv->bv_len) {
552 } else { /* skip < iv->bv_len) */
553 iv->bv_offset += skip;
562 * bio_integrity_mark_tail - Truncate bip_vec to be len bytes long
563 * @bip: Integrity vector to truncate
564 * @len: New length of integrity vector
566 void bio_integrity_mark_tail(struct bio_integrity_payload *bip,
572 bip_for_each_vec(iv, bip, i) {
576 } else if (len >= iv->bv_len) {
578 } else { /* len < iv->bv_len) */
586 * bio_integrity_advance - Advance integrity vector
587 * @bio: bio whose integrity vector to update
588 * @bytes_done: number of data bytes that have been completed
590 * Description: This function calculates how many integrity bytes the
591 * number of completed data bytes correspond to and advances the
592 * integrity vector accordingly.
594 void bio_integrity_advance(struct bio *bio, unsigned int bytes_done)
596 struct bio_integrity_payload *bip = bio->bi_integrity;
597 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
598 unsigned int nr_sectors;
603 nr_sectors = bio_integrity_hw_sectors(bi, bytes_done >> 9);
604 bio_integrity_mark_head(bip, nr_sectors * bi->tuple_size);
606 EXPORT_SYMBOL(bio_integrity_advance);
609 * bio_integrity_trim - Trim integrity vector
610 * @bio: bio whose integrity vector to update
611 * @offset: offset to first data sector
612 * @sectors: number of data sectors
614 * Description: Used to trim the integrity vector in a cloned bio.
615 * The ivec will be advanced corresponding to 'offset' data sectors
616 * and the length will be truncated corresponding to 'len' data
619 void bio_integrity_trim(struct bio *bio, unsigned int offset,
620 unsigned int sectors)
622 struct bio_integrity_payload *bip = bio->bi_integrity;
623 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
624 unsigned int nr_sectors;
628 BUG_ON(!bio_flagged(bio, BIO_CLONED));
630 nr_sectors = bio_integrity_hw_sectors(bi, sectors);
631 bip->bip_sector = bip->bip_sector + offset;
632 bio_integrity_mark_head(bip, offset * bi->tuple_size);
633 bio_integrity_mark_tail(bip, sectors * bi->tuple_size);
635 EXPORT_SYMBOL(bio_integrity_trim);
638 * bio_integrity_split - Split integrity metadata
639 * @bio: Protected bio
640 * @bp: Resulting bio_pair
643 * Description: Splits an integrity page into a bio_pair.
645 void bio_integrity_split(struct bio *bio, struct bio_pair *bp, int sectors)
647 struct blk_integrity *bi;
648 struct bio_integrity_payload *bip = bio->bi_integrity;
649 unsigned int nr_sectors;
651 if (bio_integrity(bio) == 0)
654 bi = bdev_get_integrity(bio->bi_bdev);
656 BUG_ON(bip->bip_vcnt != 1);
658 nr_sectors = bio_integrity_hw_sectors(bi, sectors);
660 bp->bio1.bi_integrity = &bp->bip1;
661 bp->bio2.bi_integrity = &bp->bip2;
663 bp->iv1 = bip->bip_vec[bip->bip_idx];
664 bp->iv2 = bip->bip_vec[bip->bip_idx];
666 bp->bip1.bip_vec = &bp->iv1;
667 bp->bip2.bip_vec = &bp->iv2;
669 bp->iv1.bv_len = sectors * bi->tuple_size;
670 bp->iv2.bv_offset += sectors * bi->tuple_size;
671 bp->iv2.bv_len -= sectors * bi->tuple_size;
673 bp->bip1.bip_sector = bio->bi_integrity->bip_sector;
674 bp->bip2.bip_sector = bio->bi_integrity->bip_sector + nr_sectors;
676 bp->bip1.bip_vcnt = bp->bip2.bip_vcnt = 1;
677 bp->bip1.bip_idx = bp->bip2.bip_idx = 0;
679 EXPORT_SYMBOL(bio_integrity_split);
682 * bio_integrity_clone - Callback for cloning bios with integrity metadata
684 * @bio_src: Original bio
685 * @gfp_mask: Memory allocation mask
687 * Description: Called to allocate a bip when cloning a bio
689 int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
692 struct bio_integrity_payload *bip_src = bio_src->bi_integrity;
693 struct bio_integrity_payload *bip;
695 BUG_ON(bip_src == NULL);
697 bip = bio_integrity_alloc(bio, gfp_mask, bip_src->bip_vcnt);
702 memcpy(bip->bip_vec, bip_src->bip_vec,
703 bip_src->bip_vcnt * sizeof(struct bio_vec));
705 bip->bip_sector = bip_src->bip_sector;
706 bip->bip_vcnt = bip_src->bip_vcnt;
707 bip->bip_idx = bip_src->bip_idx;
711 EXPORT_SYMBOL(bio_integrity_clone);
713 int bioset_integrity_create(struct bio_set *bs, int pool_size)
715 if (bs->bio_integrity_pool)
718 bs->bio_integrity_pool = mempool_create_slab_pool(pool_size, bip_slab);
719 if (!bs->bio_integrity_pool)
722 bs->bvec_integrity_pool = biovec_create_pool(bs, pool_size);
723 if (!bs->bvec_integrity_pool) {
724 mempool_destroy(bs->bio_integrity_pool);
730 EXPORT_SYMBOL(bioset_integrity_create);
732 void bioset_integrity_free(struct bio_set *bs)
734 if (bs->bio_integrity_pool)
735 mempool_destroy(bs->bio_integrity_pool);
737 if (bs->bvec_integrity_pool)
738 mempool_destroy(bs->bvec_integrity_pool);
740 EXPORT_SYMBOL(bioset_integrity_free);
742 void __init bio_integrity_init(void)
745 * kintegrityd won't block much but may burn a lot of CPU cycles.
746 * Make it highpri CPU intensive wq with max concurrency of 1.
748 kintegrityd_wq = alloc_workqueue("kintegrityd", WQ_MEM_RECLAIM |
749 WQ_HIGHPRI | WQ_CPU_INTENSIVE, 1);
751 panic("Failed to create kintegrityd\n");
753 bip_slab = kmem_cache_create("bio_integrity_payload",
754 sizeof(struct bio_integrity_payload) +
755 sizeof(struct bio_vec) * BIP_INLINE_VECS,
756 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
758 panic("Failed to create slab\n");