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 mempool_t *bio_integrity_pool;
30 static struct bio_set *integrity_bio_set;
31 static struct workqueue_struct *kintegrityd_wq;
34 * bio_integrity_alloc - Allocate integrity payload and attach it to bio
35 * @bio: bio to attach integrity metadata to
36 * @gfp_mask: Memory allocation mask
37 * @nr_vecs: Number of integrity metadata scatter-gather elements
39 * Description: This function prepares a bio for attaching integrity
40 * metadata. nr_vecs specifies the maximum number of pages containing
41 * integrity metadata that can be attached.
43 struct bio_integrity_payload *bio_integrity_alloc(struct bio *bio,
47 struct bio_integrity_payload *bip;
53 bip = mempool_alloc(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, integrity_bio_set);
62 if (unlikely(iv == NULL)) {
63 printk(KERN_ERR "%s: could not alloc bip_vec\n", __func__);
64 mempool_free(bip, bio_integrity_pool);
71 bio->bi_integrity = bip;
75 EXPORT_SYMBOL(bio_integrity_alloc);
78 * bio_integrity_free - Free bio integrity payload
79 * @bio: bio containing bip to be freed
81 * Description: Used to free the integrity portion of a bio. Usually
82 * called from bio_free().
84 void bio_integrity_free(struct bio *bio)
86 struct bio_integrity_payload *bip = bio->bi_integrity;
90 /* A cloned bio doesn't own the integrity metadata */
91 if (!bio_flagged(bio, BIO_CLONED) && !bio_flagged(bio, BIO_FS_INTEGRITY)
92 && bip->bip_buf != NULL)
95 bvec_free_bs(integrity_bio_set, bip->bip_vec, bip->bip_pool);
96 mempool_free(bip, bio_integrity_pool);
98 bio->bi_integrity = NULL;
100 EXPORT_SYMBOL(bio_integrity_free);
103 * bio_integrity_add_page - Attach integrity metadata
104 * @bio: bio to update
105 * @page: page containing integrity metadata
106 * @len: number of bytes of integrity metadata in page
107 * @offset: start offset within page
109 * Description: Attach a page containing integrity metadata to bio.
111 int bio_integrity_add_page(struct bio *bio, struct page *page,
112 unsigned int len, unsigned int offset)
114 struct bio_integrity_payload *bip = bio->bi_integrity;
117 if (bip->bip_vcnt >= bvec_nr_vecs(bip->bip_pool)) {
118 printk(KERN_ERR "%s: bip_vec full\n", __func__);
122 iv = bip_vec_idx(bip, bip->bip_vcnt);
127 iv->bv_offset = offset;
132 EXPORT_SYMBOL(bio_integrity_add_page);
134 static int bdev_integrity_enabled(struct block_device *bdev, int rw)
136 struct blk_integrity *bi = bdev_get_integrity(bdev);
141 if (rw == READ && bi->verify_fn != NULL &&
142 (bi->flags & INTEGRITY_FLAG_READ))
145 if (rw == WRITE && bi->generate_fn != NULL &&
146 (bi->flags & INTEGRITY_FLAG_WRITE))
153 * bio_integrity_enabled - Check whether integrity can be passed
156 * Description: Determines whether bio_integrity_prep() can be called
157 * on this bio or not. bio data direction and target device must be
158 * set prior to calling. The functions honors the write_generate and
159 * read_verify flags in sysfs.
161 int bio_integrity_enabled(struct bio *bio)
163 /* Already protected? */
164 if (bio_integrity(bio))
167 return bdev_integrity_enabled(bio->bi_bdev, bio_data_dir(bio));
169 EXPORT_SYMBOL(bio_integrity_enabled);
172 * bio_integrity_hw_sectors - Convert 512b sectors to hardware ditto
173 * @bi: blk_integrity profile for device
174 * @sectors: Number of 512 sectors to convert
176 * Description: The block layer calculates everything in 512 byte
177 * sectors but integrity metadata is done in terms of the hardware
178 * sector size of the storage device. Convert the block layer sectors
179 * to physical sectors.
181 static inline unsigned int bio_integrity_hw_sectors(struct blk_integrity *bi,
182 unsigned int sectors)
184 /* At this point there are only 512b or 4096b DIF/EPP devices */
185 if (bi->sector_size == 4096)
186 return sectors >>= 3;
192 * bio_integrity_tag_size - Retrieve integrity tag space
193 * @bio: bio to inspect
195 * Description: Returns the maximum number of tag bytes that can be
196 * attached to this bio. Filesystems can use this to determine how
197 * much metadata to attach to an I/O.
199 unsigned int bio_integrity_tag_size(struct bio *bio)
201 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
203 BUG_ON(bio->bi_size == 0);
205 return bi->tag_size * (bio->bi_size / bi->sector_size);
207 EXPORT_SYMBOL(bio_integrity_tag_size);
209 int bio_integrity_tag(struct bio *bio, void *tag_buf, unsigned int len, int set)
211 struct bio_integrity_payload *bip = bio->bi_integrity;
212 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
213 unsigned int nr_sectors;
215 BUG_ON(bip->bip_buf == NULL);
217 if (bi->tag_size == 0)
220 nr_sectors = bio_integrity_hw_sectors(bi,
221 DIV_ROUND_UP(len, bi->tag_size));
223 if (nr_sectors * bi->tuple_size > bip->bip_size) {
224 printk(KERN_ERR "%s: tag too big for bio: %u > %u\n",
225 __func__, nr_sectors * bi->tuple_size, bip->bip_size);
230 bi->set_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
232 bi->get_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
238 * bio_integrity_set_tag - Attach a tag buffer to a bio
239 * @bio: bio to attach buffer to
240 * @tag_buf: Pointer to a buffer containing tag data
241 * @len: Length of the included buffer
243 * Description: Use this function to tag a bio by leveraging the extra
244 * space provided by devices formatted with integrity protection. The
245 * size of the integrity buffer must be <= to the size reported by
246 * bio_integrity_tag_size().
248 int bio_integrity_set_tag(struct bio *bio, void *tag_buf, unsigned int len)
250 BUG_ON(bio_data_dir(bio) != WRITE);
252 return bio_integrity_tag(bio, tag_buf, len, 1);
254 EXPORT_SYMBOL(bio_integrity_set_tag);
257 * bio_integrity_get_tag - Retrieve a tag buffer from a bio
258 * @bio: bio to retrieve buffer from
259 * @tag_buf: Pointer to a buffer for the tag data
260 * @len: Length of the target buffer
262 * Description: Use this function to retrieve the tag buffer from a
263 * completed I/O. The size of the integrity buffer must be <= to the
264 * size reported by bio_integrity_tag_size().
266 int bio_integrity_get_tag(struct bio *bio, void *tag_buf, unsigned int len)
268 BUG_ON(bio_data_dir(bio) != READ);
270 return bio_integrity_tag(bio, tag_buf, len, 0);
272 EXPORT_SYMBOL(bio_integrity_get_tag);
275 * bio_integrity_generate - Generate integrity metadata for a bio
276 * @bio: bio to generate integrity metadata for
278 * Description: Generates integrity metadata for a bio by calling the
279 * block device's generation callback function. The bio must have a
280 * bip attached with enough room to accommodate the generated
281 * integrity metadata.
283 static void bio_integrity_generate(struct bio *bio)
285 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
286 struct blk_integrity_exchg bix;
288 sector_t sector = bio->bi_sector;
289 unsigned int i, sectors, total;
290 void *prot_buf = bio->bi_integrity->bip_buf;
293 bix.disk_name = bio->bi_bdev->bd_disk->disk_name;
294 bix.sector_size = bi->sector_size;
296 bio_for_each_segment(bv, bio, i) {
297 void *kaddr = kmap_atomic(bv->bv_page, KM_USER0);
298 bix.data_buf = kaddr + bv->bv_offset;
299 bix.data_size = bv->bv_len;
300 bix.prot_buf = prot_buf;
303 bi->generate_fn(&bix);
305 sectors = bv->bv_len / bi->sector_size;
307 prot_buf += sectors * bi->tuple_size;
308 total += sectors * bi->tuple_size;
309 BUG_ON(total > bio->bi_integrity->bip_size);
311 kunmap_atomic(kaddr, KM_USER0);
315 static inline unsigned short blk_integrity_tuple_size(struct blk_integrity *bi)
318 return bi->tuple_size;
324 * bio_integrity_prep - Prepare bio for integrity I/O
325 * @bio: bio to prepare
327 * Description: Allocates a buffer for integrity metadata, maps the
328 * pages and attaches them to a bio. The bio must have data
329 * direction, target device and start sector set priot to calling. In
330 * the WRITE case, integrity metadata will be generated using the
331 * block device's integrity function. In the READ case, the buffer
332 * will be prepared for DMA and a suitable end_io handler set up.
334 int bio_integrity_prep(struct bio *bio)
336 struct bio_integrity_payload *bip;
337 struct blk_integrity *bi;
338 struct request_queue *q;
340 unsigned long start, end;
341 unsigned int len, nr_pages;
342 unsigned int bytes, offset, i;
343 unsigned int sectors;
345 bi = bdev_get_integrity(bio->bi_bdev);
346 q = bdev_get_queue(bio->bi_bdev);
348 BUG_ON(bio_integrity(bio));
350 sectors = bio_integrity_hw_sectors(bi, bio_sectors(bio));
352 /* Allocate kernel buffer for protection data */
353 len = sectors * blk_integrity_tuple_size(bi);
354 buf = kmalloc(len, GFP_NOIO | __GFP_NOFAIL | q->bounce_gfp);
355 if (unlikely(buf == NULL)) {
356 printk(KERN_ERR "could not allocate integrity buffer\n");
360 end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
361 start = ((unsigned long) buf) >> PAGE_SHIFT;
362 nr_pages = end - start;
364 /* Allocate bio integrity payload and integrity vectors */
365 bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages);
366 if (unlikely(bip == NULL)) {
367 printk(KERN_ERR "could not allocate data integrity bioset\n");
374 bip->bip_sector = bio->bi_sector;
377 offset = offset_in_page(buf);
378 for (i = 0 ; i < nr_pages ; i++) {
380 bytes = PAGE_SIZE - offset;
388 ret = bio_integrity_add_page(bio, virt_to_page(buf),
402 /* Install custom I/O completion handler if read verify is enabled */
403 if (bio_data_dir(bio) == READ) {
404 bip->bip_end_io = bio->bi_end_io;
405 bio->bi_end_io = bio_integrity_endio;
408 /* Auto-generate integrity metadata if this is a write */
409 if (bio_data_dir(bio) == WRITE)
410 bio_integrity_generate(bio);
414 EXPORT_SYMBOL(bio_integrity_prep);
417 * bio_integrity_verify - Verify integrity metadata for a bio
418 * @bio: bio to verify
420 * Description: This function is called to verify the integrity of a
421 * bio. The data in the bio io_vec is compared to the integrity
422 * metadata returned by the HBA.
424 static int bio_integrity_verify(struct bio *bio)
426 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
427 struct blk_integrity_exchg bix;
429 sector_t sector = bio->bi_integrity->bip_sector;
430 unsigned int i, sectors, total, ret;
431 void *prot_buf = bio->bi_integrity->bip_buf;
434 bix.disk_name = bio->bi_bdev->bd_disk->disk_name;
435 bix.sector_size = bi->sector_size;
437 bio_for_each_segment(bv, bio, i) {
438 void *kaddr = kmap_atomic(bv->bv_page, KM_USER0);
439 bix.data_buf = kaddr + bv->bv_offset;
440 bix.data_size = bv->bv_len;
441 bix.prot_buf = prot_buf;
444 ret = bi->verify_fn(&bix);
447 kunmap_atomic(kaddr, KM_USER0);
451 sectors = bv->bv_len / bi->sector_size;
453 prot_buf += sectors * bi->tuple_size;
454 total += sectors * bi->tuple_size;
455 BUG_ON(total > bio->bi_integrity->bip_size);
457 kunmap_atomic(kaddr, KM_USER0);
464 * bio_integrity_verify_fn - Integrity I/O completion worker
465 * @work: Work struct stored in bio to be verified
467 * Description: This workqueue function is called to complete a READ
468 * request. The function verifies the transferred integrity metadata
469 * and then calls the original bio end_io function.
471 static void bio_integrity_verify_fn(struct work_struct *work)
473 struct bio_integrity_payload *bip =
474 container_of(work, struct bio_integrity_payload, bip_work);
475 struct bio *bio = bip->bip_bio;
478 error = bio_integrity_verify(bio);
480 /* Restore original bio completion handler */
481 bio->bi_end_io = bip->bip_end_io;
482 bio_endio(bio, error);
486 * bio_integrity_endio - Integrity I/O completion function
487 * @bio: Protected bio
488 * @error: Pointer to errno
490 * Description: Completion for integrity I/O
492 * Normally I/O completion is done in interrupt context. However,
493 * verifying I/O integrity is a time-consuming task which must be run
494 * in process context. This function postpones completion
497 void bio_integrity_endio(struct bio *bio, int error)
499 struct bio_integrity_payload *bip = bio->bi_integrity;
501 BUG_ON(bip->bip_bio != bio);
503 /* In case of an I/O error there is no point in verifying the
504 * integrity metadata. Restore original bio end_io handler
508 bio->bi_end_io = bip->bip_end_io;
509 bio_endio(bio, error);
514 INIT_WORK(&bip->bip_work, bio_integrity_verify_fn);
515 queue_work(kintegrityd_wq, &bip->bip_work);
517 EXPORT_SYMBOL(bio_integrity_endio);
520 * bio_integrity_mark_head - Advance bip_vec skip bytes
521 * @bip: Integrity vector to advance
522 * @skip: Number of bytes to advance it
524 void bio_integrity_mark_head(struct bio_integrity_payload *bip,
530 bip_for_each_vec(iv, bip, i) {
534 } else if (skip >= iv->bv_len) {
536 } else { /* skip < iv->bv_len) */
537 iv->bv_offset += skip;
546 * bio_integrity_mark_tail - Truncate bip_vec to be len bytes long
547 * @bip: Integrity vector to truncate
548 * @len: New length of integrity vector
550 void bio_integrity_mark_tail(struct bio_integrity_payload *bip,
556 bip_for_each_vec(iv, bip, i) {
560 } else if (len >= iv->bv_len) {
562 } else { /* len < iv->bv_len) */
570 * bio_integrity_advance - Advance integrity vector
571 * @bio: bio whose integrity vector to update
572 * @bytes_done: number of data bytes that have been completed
574 * Description: This function calculates how many integrity bytes the
575 * number of completed data bytes correspond to and advances the
576 * integrity vector accordingly.
578 void bio_integrity_advance(struct bio *bio, unsigned int bytes_done)
580 struct bio_integrity_payload *bip = bio->bi_integrity;
581 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
582 unsigned int nr_sectors;
587 nr_sectors = bio_integrity_hw_sectors(bi, bytes_done >> 9);
588 bio_integrity_mark_head(bip, nr_sectors * bi->tuple_size);
590 EXPORT_SYMBOL(bio_integrity_advance);
593 * bio_integrity_trim - Trim integrity vector
594 * @bio: bio whose integrity vector to update
595 * @offset: offset to first data sector
596 * @sectors: number of data sectors
598 * Description: Used to trim the integrity vector in a cloned bio.
599 * The ivec will be advanced corresponding to 'offset' data sectors
600 * and the length will be truncated corresponding to 'len' data
603 void bio_integrity_trim(struct bio *bio, unsigned int offset,
604 unsigned int sectors)
606 struct bio_integrity_payload *bip = bio->bi_integrity;
607 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
608 unsigned int nr_sectors;
612 BUG_ON(!bio_flagged(bio, BIO_CLONED));
614 nr_sectors = bio_integrity_hw_sectors(bi, sectors);
615 bip->bip_sector = bip->bip_sector + offset;
616 bio_integrity_mark_head(bip, offset * bi->tuple_size);
617 bio_integrity_mark_tail(bip, sectors * bi->tuple_size);
619 EXPORT_SYMBOL(bio_integrity_trim);
622 * bio_integrity_split - Split integrity metadata
623 * @bio: Protected bio
624 * @bp: Resulting bio_pair
627 * Description: Splits an integrity page into a bio_pair.
629 void bio_integrity_split(struct bio *bio, struct bio_pair *bp, int sectors)
631 struct blk_integrity *bi;
632 struct bio_integrity_payload *bip = bio->bi_integrity;
633 unsigned int nr_sectors;
635 if (bio_integrity(bio) == 0)
638 bi = bdev_get_integrity(bio->bi_bdev);
640 BUG_ON(bip->bip_vcnt != 1);
642 nr_sectors = bio_integrity_hw_sectors(bi, sectors);
644 bp->bio1.bi_integrity = &bp->bip1;
645 bp->bio2.bi_integrity = &bp->bip2;
647 bp->iv1 = bip->bip_vec[0];
648 bp->iv2 = bip->bip_vec[0];
650 bp->bip1.bip_vec = &bp->iv1;
651 bp->bip2.bip_vec = &bp->iv2;
653 bp->iv1.bv_len = sectors * bi->tuple_size;
654 bp->iv2.bv_offset += sectors * bi->tuple_size;
655 bp->iv2.bv_len -= sectors * bi->tuple_size;
657 bp->bip1.bip_sector = bio->bi_integrity->bip_sector;
658 bp->bip2.bip_sector = bio->bi_integrity->bip_sector + nr_sectors;
660 bp->bip1.bip_vcnt = bp->bip2.bip_vcnt = 1;
661 bp->bip1.bip_idx = bp->bip2.bip_idx = 0;
663 EXPORT_SYMBOL(bio_integrity_split);
666 * bio_integrity_clone - Callback for cloning bios with integrity metadata
668 * @bio_src: Original bio
669 * @gfp_mask: Memory allocation mask
671 * Description: Called to allocate a bip when cloning a bio
673 int bio_integrity_clone(struct bio *bio, struct bio *bio_src, gfp_t gfp_mask)
675 struct bio_integrity_payload *bip_src = bio_src->bi_integrity;
676 struct bio_integrity_payload *bip;
678 BUG_ON(bip_src == NULL);
680 bip = bio_integrity_alloc(bio, gfp_mask, bip_src->bip_vcnt);
685 memcpy(bip->bip_vec, bip_src->bip_vec,
686 bip_src->bip_vcnt * sizeof(struct bio_vec));
688 bip->bip_sector = bip_src->bip_sector;
689 bip->bip_vcnt = bip_src->bip_vcnt;
690 bip->bip_idx = bip_src->bip_idx;
694 EXPORT_SYMBOL(bio_integrity_clone);
696 static int __init bio_integrity_init(void)
698 kintegrityd_wq = create_workqueue("kintegrityd");
701 panic("Failed to create kintegrityd\n");
703 bio_integrity_slab = KMEM_CACHE(bio_integrity_payload,
704 SLAB_HWCACHE_ALIGN|SLAB_PANIC);
706 bio_integrity_pool = mempool_create_slab_pool(BIO_POOL_SIZE,
708 if (!bio_integrity_pool)
709 panic("bio_integrity: can't allocate bip pool\n");
711 integrity_bio_set = bioset_create(BIO_POOL_SIZE, 0);
712 if (!integrity_bio_set)
713 panic("bio_integrity: can't allocate bio_set\n");
717 subsys_initcall(bio_integrity_init);