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->bip_vec + bip->bip_vcnt;
141 iv->bv_offset = offset;
146 EXPORT_SYMBOL(bio_integrity_add_page);
148 static int bdev_integrity_enabled(struct block_device *bdev, int rw)
150 struct blk_integrity *bi = bdev_get_integrity(bdev);
155 if (rw == READ && bi->verify_fn != NULL &&
156 (bi->flags & INTEGRITY_FLAG_READ))
159 if (rw == WRITE && bi->generate_fn != NULL &&
160 (bi->flags & INTEGRITY_FLAG_WRITE))
167 * bio_integrity_enabled - Check whether integrity can be passed
170 * Description: Determines whether bio_integrity_prep() can be called
171 * on this bio or not. bio data direction and target device must be
172 * set prior to calling. The functions honors the write_generate and
173 * read_verify flags in sysfs.
175 int bio_integrity_enabled(struct bio *bio)
177 /* Already protected? */
178 if (bio_integrity(bio))
181 return bdev_integrity_enabled(bio->bi_bdev, bio_data_dir(bio));
183 EXPORT_SYMBOL(bio_integrity_enabled);
186 * bio_integrity_hw_sectors - Convert 512b sectors to hardware ditto
187 * @bi: blk_integrity profile for device
188 * @sectors: Number of 512 sectors to convert
190 * Description: The block layer calculates everything in 512 byte
191 * sectors but integrity metadata is done in terms of the hardware
192 * sector size of the storage device. Convert the block layer sectors
193 * to physical sectors.
195 static inline unsigned int bio_integrity_hw_sectors(struct blk_integrity *bi,
196 unsigned int sectors)
198 /* At this point there are only 512b or 4096b DIF/EPP devices */
199 if (bi->sector_size == 4096)
200 return sectors >>= 3;
205 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
206 unsigned int sectors)
208 return bio_integrity_hw_sectors(bi, sectors) * bi->tuple_size;
212 * bio_integrity_tag_size - Retrieve integrity tag space
213 * @bio: bio to inspect
215 * Description: Returns the maximum number of tag bytes that can be
216 * attached to this bio. Filesystems can use this to determine how
217 * much metadata to attach to an I/O.
219 unsigned int bio_integrity_tag_size(struct bio *bio)
221 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
223 BUG_ON(bio->bi_iter.bi_size == 0);
225 return bi->tag_size * (bio->bi_iter.bi_size / bi->sector_size);
227 EXPORT_SYMBOL(bio_integrity_tag_size);
229 int bio_integrity_tag(struct bio *bio, void *tag_buf, unsigned int len, int set)
231 struct bio_integrity_payload *bip = bio->bi_integrity;
232 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
233 unsigned int nr_sectors;
235 BUG_ON(bip->bip_buf == NULL);
237 if (bi->tag_size == 0)
240 nr_sectors = bio_integrity_hw_sectors(bi,
241 DIV_ROUND_UP(len, bi->tag_size));
243 if (nr_sectors * bi->tuple_size > bip->bip_iter.bi_size) {
244 printk(KERN_ERR "%s: tag too big for bio: %u > %u\n", __func__,
245 nr_sectors * bi->tuple_size, bip->bip_iter.bi_size);
250 bi->set_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
252 bi->get_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
258 * bio_integrity_set_tag - Attach a tag buffer to a bio
259 * @bio: bio to attach buffer to
260 * @tag_buf: Pointer to a buffer containing tag data
261 * @len: Length of the included buffer
263 * Description: Use this function to tag a bio by leveraging the extra
264 * space provided by devices formatted with integrity protection. The
265 * size of the integrity buffer must be <= to the size reported by
266 * bio_integrity_tag_size().
268 int bio_integrity_set_tag(struct bio *bio, void *tag_buf, unsigned int len)
270 BUG_ON(bio_data_dir(bio) != WRITE);
272 return bio_integrity_tag(bio, tag_buf, len, 1);
274 EXPORT_SYMBOL(bio_integrity_set_tag);
277 * bio_integrity_get_tag - Retrieve a tag buffer from a bio
278 * @bio: bio to retrieve buffer from
279 * @tag_buf: Pointer to a buffer for the tag data
280 * @len: Length of the target buffer
282 * Description: Use this function to retrieve the tag buffer from a
283 * completed I/O. The size of the integrity buffer must be <= to the
284 * size reported by bio_integrity_tag_size().
286 int bio_integrity_get_tag(struct bio *bio, void *tag_buf, unsigned int len)
288 BUG_ON(bio_data_dir(bio) != READ);
290 return bio_integrity_tag(bio, tag_buf, len, 0);
292 EXPORT_SYMBOL(bio_integrity_get_tag);
295 * bio_integrity_generate - Generate integrity metadata for a bio
296 * @bio: bio to generate integrity metadata for
298 * Description: Generates integrity metadata for a bio by calling the
299 * block device's generation callback function. The bio must have a
300 * bip attached with enough room to accommodate the generated
301 * integrity metadata.
303 static void bio_integrity_generate(struct bio *bio)
305 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
306 struct blk_integrity_exchg bix;
308 struct bvec_iter iter;
309 sector_t sector = bio->bi_iter.bi_sector;
310 unsigned int sectors, total;
311 void *prot_buf = bio->bi_integrity->bip_buf;
314 bix.disk_name = bio->bi_bdev->bd_disk->disk_name;
315 bix.sector_size = bi->sector_size;
317 bio_for_each_segment(bv, bio, iter) {
318 void *kaddr = kmap_atomic(bv.bv_page);
319 bix.data_buf = kaddr + bv.bv_offset;
320 bix.data_size = bv.bv_len;
321 bix.prot_buf = prot_buf;
324 bi->generate_fn(&bix);
326 sectors = bv.bv_len / bi->sector_size;
328 prot_buf += sectors * bi->tuple_size;
329 total += sectors * bi->tuple_size;
330 BUG_ON(total > bio->bi_integrity->bip_iter.bi_size);
332 kunmap_atomic(kaddr);
336 static inline unsigned short blk_integrity_tuple_size(struct blk_integrity *bi)
339 return bi->tuple_size;
345 * bio_integrity_prep - Prepare bio for integrity I/O
346 * @bio: bio to prepare
348 * Description: Allocates a buffer for integrity metadata, maps the
349 * pages and attaches them to a bio. The bio must have data
350 * direction, target device and start sector set priot to calling. In
351 * the WRITE case, integrity metadata will be generated using the
352 * block device's integrity function. In the READ case, the buffer
353 * will be prepared for DMA and a suitable end_io handler set up.
355 int bio_integrity_prep(struct bio *bio)
357 struct bio_integrity_payload *bip;
358 struct blk_integrity *bi;
359 struct request_queue *q;
361 unsigned long start, end;
362 unsigned int len, nr_pages;
363 unsigned int bytes, offset, i;
364 unsigned int sectors;
366 bi = bdev_get_integrity(bio->bi_bdev);
367 q = bdev_get_queue(bio->bi_bdev);
369 BUG_ON(bio_integrity(bio));
371 sectors = bio_integrity_hw_sectors(bi, bio_sectors(bio));
373 /* Allocate kernel buffer for protection data */
374 len = sectors * blk_integrity_tuple_size(bi);
375 buf = kmalloc(len, GFP_NOIO | q->bounce_gfp);
376 if (unlikely(buf == NULL)) {
377 printk(KERN_ERR "could not allocate integrity buffer\n");
381 end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
382 start = ((unsigned long) buf) >> PAGE_SHIFT;
383 nr_pages = end - start;
385 /* Allocate bio integrity payload and integrity vectors */
386 bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages);
387 if (unlikely(bip == NULL)) {
388 printk(KERN_ERR "could not allocate data integrity bioset\n");
393 bip->bip_owns_buf = 1;
395 bip->bip_iter.bi_size = len;
396 bip->bip_iter.bi_sector = bio->bi_iter.bi_sector;
399 offset = offset_in_page(buf);
400 for (i = 0 ; i < nr_pages ; i++) {
402 bytes = PAGE_SIZE - offset;
410 ret = bio_integrity_add_page(bio, virt_to_page(buf),
424 /* Install custom I/O completion handler if read verify is enabled */
425 if (bio_data_dir(bio) == READ) {
426 bip->bip_end_io = bio->bi_end_io;
427 bio->bi_end_io = bio_integrity_endio;
430 /* Auto-generate integrity metadata if this is a write */
431 if (bio_data_dir(bio) == WRITE)
432 bio_integrity_generate(bio);
436 EXPORT_SYMBOL(bio_integrity_prep);
439 * bio_integrity_verify - Verify integrity metadata for a bio
440 * @bio: bio to verify
442 * Description: This function is called to verify the integrity of a
443 * bio. The data in the bio io_vec is compared to the integrity
444 * metadata returned by the HBA.
446 static int bio_integrity_verify(struct bio *bio)
448 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
449 struct blk_integrity_exchg bix;
451 sector_t sector = bio->bi_integrity->bip_iter.bi_sector;
452 unsigned int sectors, total, ret;
453 void *prot_buf = bio->bi_integrity->bip_buf;
457 bix.disk_name = bio->bi_bdev->bd_disk->disk_name;
458 bix.sector_size = bi->sector_size;
460 bio_for_each_segment_all(bv, bio, i) {
461 void *kaddr = kmap_atomic(bv->bv_page);
463 bix.data_buf = kaddr + bv->bv_offset;
464 bix.data_size = bv->bv_len;
465 bix.prot_buf = prot_buf;
468 ret = bi->verify_fn(&bix);
471 kunmap_atomic(kaddr);
475 sectors = bv->bv_len / bi->sector_size;
477 prot_buf += sectors * bi->tuple_size;
478 total += sectors * bi->tuple_size;
479 BUG_ON(total > bio->bi_integrity->bip_iter.bi_size);
481 kunmap_atomic(kaddr);
488 * bio_integrity_verify_fn - Integrity I/O completion worker
489 * @work: Work struct stored in bio to be verified
491 * Description: This workqueue function is called to complete a READ
492 * request. The function verifies the transferred integrity metadata
493 * and then calls the original bio end_io function.
495 static void bio_integrity_verify_fn(struct work_struct *work)
497 struct bio_integrity_payload *bip =
498 container_of(work, struct bio_integrity_payload, bip_work);
499 struct bio *bio = bip->bip_bio;
502 error = bio_integrity_verify(bio);
504 /* Restore original bio completion handler */
505 bio->bi_end_io = bip->bip_end_io;
506 bio_endio_nodec(bio, error);
510 * bio_integrity_endio - Integrity I/O completion function
511 * @bio: Protected bio
512 * @error: Pointer to errno
514 * Description: Completion for integrity I/O
516 * Normally I/O completion is done in interrupt context. However,
517 * verifying I/O integrity is a time-consuming task which must be run
518 * in process context. This function postpones completion
521 void bio_integrity_endio(struct bio *bio, int error)
523 struct bio_integrity_payload *bip = bio->bi_integrity;
525 BUG_ON(bip->bip_bio != bio);
527 /* In case of an I/O error there is no point in verifying the
528 * integrity metadata. Restore original bio end_io handler
532 bio->bi_end_io = bip->bip_end_io;
533 bio_endio(bio, error);
538 INIT_WORK(&bip->bip_work, bio_integrity_verify_fn);
539 queue_work(kintegrityd_wq, &bip->bip_work);
541 EXPORT_SYMBOL(bio_integrity_endio);
544 * bio_integrity_advance - Advance integrity vector
545 * @bio: bio whose integrity vector to update
546 * @bytes_done: number of data bytes that have been completed
548 * Description: This function calculates how many integrity bytes the
549 * number of completed data bytes correspond to and advances the
550 * integrity vector accordingly.
552 void bio_integrity_advance(struct bio *bio, unsigned int bytes_done)
554 struct bio_integrity_payload *bip = bio->bi_integrity;
555 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
556 unsigned bytes = bio_integrity_bytes(bi, bytes_done >> 9);
558 bvec_iter_advance(bip->bip_vec, &bip->bip_iter, bytes);
560 EXPORT_SYMBOL(bio_integrity_advance);
563 * bio_integrity_trim - Trim integrity vector
564 * @bio: bio whose integrity vector to update
565 * @offset: offset to first data sector
566 * @sectors: number of data sectors
568 * Description: Used to trim the integrity vector in a cloned bio.
569 * The ivec will be advanced corresponding to 'offset' data sectors
570 * and the length will be truncated corresponding to 'len' data
573 void bio_integrity_trim(struct bio *bio, unsigned int offset,
574 unsigned int sectors)
576 struct bio_integrity_payload *bip = bio->bi_integrity;
577 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
579 bio_integrity_advance(bio, offset << 9);
580 bip->bip_iter.bi_size = bio_integrity_bytes(bi, sectors);
582 EXPORT_SYMBOL(bio_integrity_trim);
585 * bio_integrity_clone - Callback for cloning bios with integrity metadata
587 * @bio_src: Original bio
588 * @gfp_mask: Memory allocation mask
590 * Description: Called to allocate a bip when cloning a bio
592 int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
595 struct bio_integrity_payload *bip_src = bio_src->bi_integrity;
596 struct bio_integrity_payload *bip;
598 BUG_ON(bip_src == NULL);
600 bip = bio_integrity_alloc(bio, gfp_mask, bip_src->bip_vcnt);
605 memcpy(bip->bip_vec, bip_src->bip_vec,
606 bip_src->bip_vcnt * sizeof(struct bio_vec));
608 bip->bip_vcnt = bip_src->bip_vcnt;
609 bip->bip_iter = bip_src->bip_iter;
613 EXPORT_SYMBOL(bio_integrity_clone);
615 int bioset_integrity_create(struct bio_set *bs, int pool_size)
617 if (bs->bio_integrity_pool)
620 bs->bio_integrity_pool = mempool_create_slab_pool(pool_size, bip_slab);
621 if (!bs->bio_integrity_pool)
624 bs->bvec_integrity_pool = biovec_create_pool(bs, pool_size);
625 if (!bs->bvec_integrity_pool) {
626 mempool_destroy(bs->bio_integrity_pool);
632 EXPORT_SYMBOL(bioset_integrity_create);
634 void bioset_integrity_free(struct bio_set *bs)
636 if (bs->bio_integrity_pool)
637 mempool_destroy(bs->bio_integrity_pool);
639 if (bs->bvec_integrity_pool)
640 mempool_destroy(bs->bvec_integrity_pool);
642 EXPORT_SYMBOL(bioset_integrity_free);
644 void __init bio_integrity_init(void)
647 * kintegrityd won't block much but may burn a lot of CPU cycles.
648 * Make it highpri CPU intensive wq with max concurrency of 1.
650 kintegrityd_wq = alloc_workqueue("kintegrityd", WQ_MEM_RECLAIM |
651 WQ_HIGHPRI | WQ_CPU_INTENSIVE, 1);
653 panic("Failed to create kintegrityd\n");
655 bip_slab = kmem_cache_create("bio_integrity_payload",
656 sizeof(struct bio_integrity_payload) +
657 sizeof(struct bio_vec) * BIP_INLINE_VECS,
658 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
660 panic("Failed to create slab\n");