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);
73 bip->bip_max_vcnt = bvec_nr_vecs(idx);
75 bip->bip_vec = bip->bip_inline_vecs;
76 bip->bip_max_vcnt = inline_vecs;
81 bio->bi_integrity = bip;
82 bio->bi_rw |= REQ_INTEGRITY;
86 mempool_free(bip, bs->bio_integrity_pool);
89 EXPORT_SYMBOL(bio_integrity_alloc);
92 * bio_integrity_free - Free bio integrity payload
93 * @bio: bio containing bip to be freed
95 * Description: Used to free the integrity portion of a bio. Usually
96 * called from bio_free().
98 void bio_integrity_free(struct bio *bio)
100 struct bio_integrity_payload *bip = bio_integrity(bio);
101 struct bio_set *bs = bio->bi_pool;
103 if (bip->bip_owns_buf)
107 if (bip->bip_slab != BIO_POOL_NONE)
108 bvec_free(bs->bvec_integrity_pool, bip->bip_vec,
111 mempool_free(bip, bs->bio_integrity_pool);
116 bio->bi_integrity = NULL;
118 EXPORT_SYMBOL(bio_integrity_free);
121 * bio_integrity_add_page - Attach integrity metadata
122 * @bio: bio to update
123 * @page: page containing integrity metadata
124 * @len: number of bytes of integrity metadata in page
125 * @offset: start offset within page
127 * Description: Attach a page containing integrity metadata to bio.
129 int bio_integrity_add_page(struct bio *bio, struct page *page,
130 unsigned int len, unsigned int offset)
132 struct bio_integrity_payload *bip = bio_integrity(bio);
135 if (bip->bip_vcnt >= bip->bip_max_vcnt) {
136 printk(KERN_ERR "%s: bip_vec full\n", __func__);
140 iv = bip->bip_vec + bip->bip_vcnt;
144 iv->bv_offset = offset;
149 EXPORT_SYMBOL(bio_integrity_add_page);
152 * bio_integrity_enabled - Check whether integrity can be passed
155 * Description: Determines whether bio_integrity_prep() can be called
156 * on this bio or not. bio data direction and target device must be
157 * set prior to calling. The functions honors the write_generate and
158 * read_verify flags in sysfs.
160 bool bio_integrity_enabled(struct bio *bio)
162 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
167 /* Already protected? */
168 if (bio_integrity(bio))
174 if (bio_data_dir(bio) == READ && bi->verify_fn != NULL &&
175 (bi->flags & INTEGRITY_FLAG_READ))
178 if (bio_data_dir(bio) == WRITE && bi->generate_fn != NULL &&
179 (bi->flags & INTEGRITY_FLAG_WRITE))
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;
206 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
207 unsigned int sectors)
209 return bio_integrity_hw_sectors(bi, sectors) * bi->tuple_size;
213 * bio_integrity_tag_size - Retrieve integrity tag space
214 * @bio: bio to inspect
216 * Description: Returns the maximum number of tag bytes that can be
217 * attached to this bio. Filesystems can use this to determine how
218 * much metadata to attach to an I/O.
220 unsigned int bio_integrity_tag_size(struct bio *bio)
222 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
224 BUG_ON(bio->bi_iter.bi_size == 0);
226 return bi->tag_size * (bio->bi_iter.bi_size / bi->sector_size);
228 EXPORT_SYMBOL(bio_integrity_tag_size);
230 static int bio_integrity_tag(struct bio *bio, void *tag_buf, unsigned int len,
233 struct bio_integrity_payload *bip = bio_integrity(bio);
234 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
235 unsigned int nr_sectors;
237 BUG_ON(bip->bip_buf == NULL);
239 if (bi->tag_size == 0)
242 nr_sectors = bio_integrity_hw_sectors(bi,
243 DIV_ROUND_UP(len, bi->tag_size));
245 if (nr_sectors * bi->tuple_size > bip->bip_iter.bi_size) {
246 printk(KERN_ERR "%s: tag too big for bio: %u > %u\n", __func__,
247 nr_sectors * bi->tuple_size, bip->bip_iter.bi_size);
252 bi->set_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
254 bi->get_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
260 * bio_integrity_set_tag - Attach a tag buffer to a bio
261 * @bio: bio to attach buffer to
262 * @tag_buf: Pointer to a buffer containing tag data
263 * @len: Length of the included buffer
265 * Description: Use this function to tag a bio by leveraging the extra
266 * space provided by devices formatted with integrity protection. The
267 * size of the integrity buffer must be <= to the size reported by
268 * bio_integrity_tag_size().
270 int bio_integrity_set_tag(struct bio *bio, void *tag_buf, unsigned int len)
272 BUG_ON(bio_data_dir(bio) != WRITE);
274 return bio_integrity_tag(bio, tag_buf, len, 1);
276 EXPORT_SYMBOL(bio_integrity_set_tag);
279 * bio_integrity_get_tag - Retrieve a tag buffer from a bio
280 * @bio: bio to retrieve buffer from
281 * @tag_buf: Pointer to a buffer for the tag data
282 * @len: Length of the target buffer
284 * Description: Use this function to retrieve the tag buffer from a
285 * completed I/O. The size of the integrity buffer must be <= to the
286 * size reported by bio_integrity_tag_size().
288 int bio_integrity_get_tag(struct bio *bio, void *tag_buf, unsigned int len)
290 BUG_ON(bio_data_dir(bio) != READ);
292 return bio_integrity_tag(bio, tag_buf, len, 0);
294 EXPORT_SYMBOL(bio_integrity_get_tag);
297 * bio_integrity_generate_verify - Generate/verify integrity metadata for a bio
298 * @bio: bio to generate/verify integrity metadata for
299 * @operate: operate number, 1 for generate, 0 for verify
301 static int bio_integrity_generate_verify(struct bio *bio, int operate)
303 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
304 struct blk_integrity_exchg bix;
307 unsigned int sectors, ret = 0, i;
308 void *prot_buf = bio_integrity(bio)->bip_buf;
311 sector = bio->bi_iter.bi_sector;
313 sector = bio_integrity(bio)->bip_iter.bi_sector;
315 bix.disk_name = bio->bi_bdev->bd_disk->disk_name;
316 bix.sector_size = bi->sector_size;
318 bio_for_each_segment_all(bv, bio, i) {
319 void *kaddr = kmap_atomic(bv->bv_page);
320 bix.data_buf = kaddr + bv->bv_offset;
321 bix.data_size = bv->bv_len;
322 bix.prot_buf = prot_buf;
326 bi->generate_fn(&bix);
328 ret = bi->verify_fn(&bix);
330 kunmap_atomic(kaddr);
335 sectors = bv->bv_len / bi->sector_size;
337 prot_buf += sectors * bi->tuple_size;
339 kunmap_atomic(kaddr);
345 * bio_integrity_generate - Generate integrity metadata for a bio
346 * @bio: bio to generate integrity metadata for
348 * Description: Generates integrity metadata for a bio by calling the
349 * block device's generation callback function. The bio must have a
350 * bip attached with enough room to accommodate the generated
351 * integrity metadata.
353 static void bio_integrity_generate(struct bio *bio)
355 bio_integrity_generate_verify(bio, 1);
358 static inline unsigned short blk_integrity_tuple_size(struct blk_integrity *bi)
361 return bi->tuple_size;
367 * bio_integrity_prep - Prepare bio for integrity I/O
368 * @bio: bio to prepare
370 * Description: Allocates a buffer for integrity metadata, maps the
371 * pages and attaches them to a bio. The bio must have data
372 * direction, target device and start sector set priot to calling. In
373 * the WRITE case, integrity metadata will be generated using the
374 * block device's integrity function. In the READ case, the buffer
375 * will be prepared for DMA and a suitable end_io handler set up.
377 int bio_integrity_prep(struct bio *bio)
379 struct bio_integrity_payload *bip;
380 struct blk_integrity *bi;
381 struct request_queue *q;
383 unsigned long start, end;
384 unsigned int len, nr_pages;
385 unsigned int bytes, offset, i;
386 unsigned int sectors;
388 bi = bdev_get_integrity(bio->bi_bdev);
389 q = bdev_get_queue(bio->bi_bdev);
391 BUG_ON(bio_integrity(bio));
393 sectors = bio_integrity_hw_sectors(bi, bio_sectors(bio));
395 /* Allocate kernel buffer for protection data */
396 len = sectors * blk_integrity_tuple_size(bi);
397 buf = kmalloc(len, GFP_NOIO | q->bounce_gfp);
398 if (unlikely(buf == NULL)) {
399 printk(KERN_ERR "could not allocate integrity buffer\n");
403 end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
404 start = ((unsigned long) buf) >> PAGE_SHIFT;
405 nr_pages = end - start;
407 /* Allocate bio integrity payload and integrity vectors */
408 bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages);
409 if (unlikely(bip == NULL)) {
410 printk(KERN_ERR "could not allocate data integrity bioset\n");
415 bip->bip_owns_buf = 1;
417 bip->bip_iter.bi_size = len;
418 bip->bip_iter.bi_sector = bio->bi_iter.bi_sector;
421 offset = offset_in_page(buf);
422 for (i = 0 ; i < nr_pages ; i++) {
424 bytes = PAGE_SIZE - offset;
432 ret = bio_integrity_add_page(bio, virt_to_page(buf),
446 /* Install custom I/O completion handler if read verify is enabled */
447 if (bio_data_dir(bio) == READ) {
448 bip->bip_end_io = bio->bi_end_io;
449 bio->bi_end_io = bio_integrity_endio;
452 /* Auto-generate integrity metadata if this is a write */
453 if (bio_data_dir(bio) == WRITE)
454 bio_integrity_generate(bio);
458 EXPORT_SYMBOL(bio_integrity_prep);
461 * bio_integrity_verify - Verify integrity metadata for a bio
462 * @bio: bio to verify
464 * Description: This function is called to verify the integrity of a
465 * bio. The data in the bio io_vec is compared to the integrity
466 * metadata returned by the HBA.
468 static int bio_integrity_verify(struct bio *bio)
470 return bio_integrity_generate_verify(bio, 0);
474 * bio_integrity_verify_fn - Integrity I/O completion worker
475 * @work: Work struct stored in bio to be verified
477 * Description: This workqueue function is called to complete a READ
478 * request. The function verifies the transferred integrity metadata
479 * and then calls the original bio end_io function.
481 static void bio_integrity_verify_fn(struct work_struct *work)
483 struct bio_integrity_payload *bip =
484 container_of(work, struct bio_integrity_payload, bip_work);
485 struct bio *bio = bip->bip_bio;
488 error = bio_integrity_verify(bio);
490 /* Restore original bio completion handler */
491 bio->bi_end_io = bip->bip_end_io;
492 bio_endio_nodec(bio, error);
496 * bio_integrity_endio - Integrity I/O completion function
497 * @bio: Protected bio
498 * @error: Pointer to errno
500 * Description: Completion for integrity I/O
502 * Normally I/O completion is done in interrupt context. However,
503 * verifying I/O integrity is a time-consuming task which must be run
504 * in process context. This function postpones completion
507 void bio_integrity_endio(struct bio *bio, int error)
509 struct bio_integrity_payload *bip = bio_integrity(bio);
511 BUG_ON(bip->bip_bio != bio);
513 /* In case of an I/O error there is no point in verifying the
514 * integrity metadata. Restore original bio end_io handler
518 bio->bi_end_io = bip->bip_end_io;
519 bio_endio_nodec(bio, error);
524 INIT_WORK(&bip->bip_work, bio_integrity_verify_fn);
525 queue_work(kintegrityd_wq, &bip->bip_work);
527 EXPORT_SYMBOL(bio_integrity_endio);
530 * bio_integrity_advance - Advance integrity vector
531 * @bio: bio whose integrity vector to update
532 * @bytes_done: number of data bytes that have been completed
534 * Description: This function calculates how many integrity bytes the
535 * number of completed data bytes correspond to and advances the
536 * integrity vector accordingly.
538 void bio_integrity_advance(struct bio *bio, unsigned int bytes_done)
540 struct bio_integrity_payload *bip = bio_integrity(bio);
541 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
542 unsigned bytes = bio_integrity_bytes(bi, bytes_done >> 9);
544 bvec_iter_advance(bip->bip_vec, &bip->bip_iter, bytes);
546 EXPORT_SYMBOL(bio_integrity_advance);
549 * bio_integrity_trim - Trim integrity vector
550 * @bio: bio whose integrity vector to update
551 * @offset: offset to first data sector
552 * @sectors: number of data sectors
554 * Description: Used to trim the integrity vector in a cloned bio.
555 * The ivec will be advanced corresponding to 'offset' data sectors
556 * and the length will be truncated corresponding to 'len' data
559 void bio_integrity_trim(struct bio *bio, unsigned int offset,
560 unsigned int sectors)
562 struct bio_integrity_payload *bip = bio_integrity(bio);
563 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
565 bio_integrity_advance(bio, offset << 9);
566 bip->bip_iter.bi_size = bio_integrity_bytes(bi, sectors);
568 EXPORT_SYMBOL(bio_integrity_trim);
571 * bio_integrity_clone - Callback for cloning bios with integrity metadata
573 * @bio_src: Original bio
574 * @gfp_mask: Memory allocation mask
576 * Description: Called to allocate a bip when cloning a bio
578 int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
581 struct bio_integrity_payload *bip_src = bio_integrity(bio_src);
582 struct bio_integrity_payload *bip;
584 BUG_ON(bip_src == NULL);
586 bip = bio_integrity_alloc(bio, gfp_mask, bip_src->bip_vcnt);
591 memcpy(bip->bip_vec, bip_src->bip_vec,
592 bip_src->bip_vcnt * sizeof(struct bio_vec));
594 bip->bip_vcnt = bip_src->bip_vcnt;
595 bip->bip_iter = bip_src->bip_iter;
599 EXPORT_SYMBOL(bio_integrity_clone);
601 int bioset_integrity_create(struct bio_set *bs, int pool_size)
603 if (bs->bio_integrity_pool)
606 bs->bio_integrity_pool = mempool_create_slab_pool(pool_size, bip_slab);
607 if (!bs->bio_integrity_pool)
610 bs->bvec_integrity_pool = biovec_create_pool(pool_size);
611 if (!bs->bvec_integrity_pool) {
612 mempool_destroy(bs->bio_integrity_pool);
618 EXPORT_SYMBOL(bioset_integrity_create);
620 void bioset_integrity_free(struct bio_set *bs)
622 if (bs->bio_integrity_pool)
623 mempool_destroy(bs->bio_integrity_pool);
625 if (bs->bvec_integrity_pool)
626 mempool_destroy(bs->bvec_integrity_pool);
628 EXPORT_SYMBOL(bioset_integrity_free);
630 void __init bio_integrity_init(void)
633 * kintegrityd won't block much but may burn a lot of CPU cycles.
634 * Make it highpri CPU intensive wq with max concurrency of 1.
636 kintegrityd_wq = alloc_workqueue("kintegrityd", WQ_MEM_RECLAIM |
637 WQ_HIGHPRI | WQ_CPU_INTENSIVE, 1);
639 panic("Failed to create kintegrityd\n");
641 bip_slab = kmem_cache_create("bio_integrity_payload",
642 sizeof(struct bio_integrity_payload) +
643 sizeof(struct bio_vec) * BIP_INLINE_VECS,
644 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
646 panic("Failed to create slab\n");