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;
85 mempool_free(bip, bs->bio_integrity_pool);
88 EXPORT_SYMBOL(bio_integrity_alloc);
91 * bio_integrity_free - Free bio integrity payload
92 * @bio: bio containing bip to be freed
94 * Description: Used to free the integrity portion of a bio. Usually
95 * called from bio_free().
97 void bio_integrity_free(struct bio *bio)
99 struct bio_integrity_payload *bip = bio->bi_integrity;
100 struct bio_set *bs = bio->bi_pool;
102 if (bip->bip_owns_buf)
106 if (bip->bip_slab != BIO_POOL_NONE)
107 bvec_free(bs->bvec_integrity_pool, bip->bip_vec,
110 mempool_free(bip, bs->bio_integrity_pool);
115 bio->bi_integrity = NULL;
117 EXPORT_SYMBOL(bio_integrity_free);
120 * bio_integrity_add_page - Attach integrity metadata
121 * @bio: bio to update
122 * @page: page containing integrity metadata
123 * @len: number of bytes of integrity metadata in page
124 * @offset: start offset within page
126 * Description: Attach a page containing integrity metadata to bio.
128 int bio_integrity_add_page(struct bio *bio, struct page *page,
129 unsigned int len, unsigned int offset)
131 struct bio_integrity_payload *bip = bio->bi_integrity;
134 if (bip->bip_vcnt >= bip->bip_max_vcnt) {
135 printk(KERN_ERR "%s: bip_vec full\n", __func__);
139 iv = bip->bip_vec + bip->bip_vcnt;
143 iv->bv_offset = offset;
148 EXPORT_SYMBOL(bio_integrity_add_page);
151 * bio_integrity_enabled - Check whether integrity can be passed
154 * Description: Determines whether bio_integrity_prep() can be called
155 * on this bio or not. bio data direction and target device must be
156 * set prior to calling. The functions honors the write_generate and
157 * read_verify flags in sysfs.
159 bool bio_integrity_enabled(struct bio *bio)
161 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
166 /* Already protected? */
167 if (bio_integrity(bio))
173 if (bio_data_dir(bio) == READ && bi->verify_fn != NULL &&
174 (bi->flags & INTEGRITY_FLAG_READ))
177 if (bio_data_dir(bio) == WRITE && bi->generate_fn != NULL &&
178 (bi->flags & INTEGRITY_FLAG_WRITE))
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 static int bio_integrity_tag(struct bio *bio, void *tag_buf, unsigned int len,
232 struct bio_integrity_payload *bip = bio->bi_integrity;
233 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
234 unsigned int nr_sectors;
236 BUG_ON(bip->bip_buf == NULL);
238 if (bi->tag_size == 0)
241 nr_sectors = bio_integrity_hw_sectors(bi,
242 DIV_ROUND_UP(len, bi->tag_size));
244 if (nr_sectors * bi->tuple_size > bip->bip_iter.bi_size) {
245 printk(KERN_ERR "%s: tag too big for bio: %u > %u\n", __func__,
246 nr_sectors * bi->tuple_size, bip->bip_iter.bi_size);
251 bi->set_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
253 bi->get_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
259 * bio_integrity_set_tag - Attach a tag buffer to a bio
260 * @bio: bio to attach buffer to
261 * @tag_buf: Pointer to a buffer containing tag data
262 * @len: Length of the included buffer
264 * Description: Use this function to tag a bio by leveraging the extra
265 * space provided by devices formatted with integrity protection. The
266 * size of the integrity buffer must be <= to the size reported by
267 * bio_integrity_tag_size().
269 int bio_integrity_set_tag(struct bio *bio, void *tag_buf, unsigned int len)
271 BUG_ON(bio_data_dir(bio) != WRITE);
273 return bio_integrity_tag(bio, tag_buf, len, 1);
275 EXPORT_SYMBOL(bio_integrity_set_tag);
278 * bio_integrity_get_tag - Retrieve a tag buffer from a bio
279 * @bio: bio to retrieve buffer from
280 * @tag_buf: Pointer to a buffer for the tag data
281 * @len: Length of the target buffer
283 * Description: Use this function to retrieve the tag buffer from a
284 * completed I/O. The size of the integrity buffer must be <= to the
285 * size reported by bio_integrity_tag_size().
287 int bio_integrity_get_tag(struct bio *bio, void *tag_buf, unsigned int len)
289 BUG_ON(bio_data_dir(bio) != READ);
291 return bio_integrity_tag(bio, tag_buf, len, 0);
293 EXPORT_SYMBOL(bio_integrity_get_tag);
296 * bio_integrity_generate_verify - Generate/verify integrity metadata for a bio
297 * @bio: bio to generate/verify integrity metadata for
298 * @operate: operate number, 1 for generate, 0 for verify
300 static int bio_integrity_generate_verify(struct bio *bio, int operate)
302 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
303 struct blk_integrity_exchg bix;
306 unsigned int sectors, ret = 0, i;
307 void *prot_buf = bio->bi_integrity->bip_buf;
310 sector = bio->bi_iter.bi_sector;
312 sector = bio->bi_integrity->bip_iter.bi_sector;
314 bix.disk_name = bio->bi_bdev->bd_disk->disk_name;
315 bix.sector_size = bi->sector_size;
317 bio_for_each_segment_all(bv, bio, i) {
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;
325 bi->generate_fn(&bix);
327 ret = bi->verify_fn(&bix);
329 kunmap_atomic(kaddr);
334 sectors = bv->bv_len / bi->sector_size;
336 prot_buf += sectors * bi->tuple_size;
338 kunmap_atomic(kaddr);
344 * bio_integrity_generate - Generate integrity metadata for a bio
345 * @bio: bio to generate integrity metadata for
347 * Description: Generates integrity metadata for a bio by calling the
348 * block device's generation callback function. The bio must have a
349 * bip attached with enough room to accommodate the generated
350 * integrity metadata.
352 static void bio_integrity_generate(struct bio *bio)
354 bio_integrity_generate_verify(bio, 1);
357 static inline unsigned short blk_integrity_tuple_size(struct blk_integrity *bi)
360 return bi->tuple_size;
366 * bio_integrity_prep - Prepare bio for integrity I/O
367 * @bio: bio to prepare
369 * Description: Allocates a buffer for integrity metadata, maps the
370 * pages and attaches them to a bio. The bio must have data
371 * direction, target device and start sector set priot to calling. In
372 * the WRITE case, integrity metadata will be generated using the
373 * block device's integrity function. In the READ case, the buffer
374 * will be prepared for DMA and a suitable end_io handler set up.
376 int bio_integrity_prep(struct bio *bio)
378 struct bio_integrity_payload *bip;
379 struct blk_integrity *bi;
380 struct request_queue *q;
382 unsigned long start, end;
383 unsigned int len, nr_pages;
384 unsigned int bytes, offset, i;
385 unsigned int sectors;
387 bi = bdev_get_integrity(bio->bi_bdev);
388 q = bdev_get_queue(bio->bi_bdev);
390 BUG_ON(bio_integrity(bio));
392 sectors = bio_integrity_hw_sectors(bi, bio_sectors(bio));
394 /* Allocate kernel buffer for protection data */
395 len = sectors * blk_integrity_tuple_size(bi);
396 buf = kmalloc(len, GFP_NOIO | q->bounce_gfp);
397 if (unlikely(buf == NULL)) {
398 printk(KERN_ERR "could not allocate integrity buffer\n");
402 end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
403 start = ((unsigned long) buf) >> PAGE_SHIFT;
404 nr_pages = end - start;
406 /* Allocate bio integrity payload and integrity vectors */
407 bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages);
408 if (unlikely(bip == NULL)) {
409 printk(KERN_ERR "could not allocate data integrity bioset\n");
414 bip->bip_owns_buf = 1;
416 bip->bip_iter.bi_size = len;
417 bip->bip_iter.bi_sector = bio->bi_iter.bi_sector;
420 offset = offset_in_page(buf);
421 for (i = 0 ; i < nr_pages ; i++) {
423 bytes = PAGE_SIZE - offset;
431 ret = bio_integrity_add_page(bio, virt_to_page(buf),
445 /* Install custom I/O completion handler if read verify is enabled */
446 if (bio_data_dir(bio) == READ) {
447 bip->bip_end_io = bio->bi_end_io;
448 bio->bi_end_io = bio_integrity_endio;
451 /* Auto-generate integrity metadata if this is a write */
452 if (bio_data_dir(bio) == WRITE)
453 bio_integrity_generate(bio);
457 EXPORT_SYMBOL(bio_integrity_prep);
460 * bio_integrity_verify - Verify integrity metadata for a bio
461 * @bio: bio to verify
463 * Description: This function is called to verify the integrity of a
464 * bio. The data in the bio io_vec is compared to the integrity
465 * metadata returned by the HBA.
467 static int bio_integrity_verify(struct bio *bio)
469 return bio_integrity_generate_verify(bio, 0);
473 * bio_integrity_verify_fn - Integrity I/O completion worker
474 * @work: Work struct stored in bio to be verified
476 * Description: This workqueue function is called to complete a READ
477 * request. The function verifies the transferred integrity metadata
478 * and then calls the original bio end_io function.
480 static void bio_integrity_verify_fn(struct work_struct *work)
482 struct bio_integrity_payload *bip =
483 container_of(work, struct bio_integrity_payload, bip_work);
484 struct bio *bio = bip->bip_bio;
487 error = bio_integrity_verify(bio);
489 /* Restore original bio completion handler */
490 bio->bi_end_io = bip->bip_end_io;
491 bio_endio_nodec(bio, error);
495 * bio_integrity_endio - Integrity I/O completion function
496 * @bio: Protected bio
497 * @error: Pointer to errno
499 * Description: Completion for integrity I/O
501 * Normally I/O completion is done in interrupt context. However,
502 * verifying I/O integrity is a time-consuming task which must be run
503 * in process context. This function postpones completion
506 void bio_integrity_endio(struct bio *bio, int error)
508 struct bio_integrity_payload *bip = bio->bi_integrity;
510 BUG_ON(bip->bip_bio != bio);
512 /* In case of an I/O error there is no point in verifying the
513 * integrity metadata. Restore original bio end_io handler
517 bio->bi_end_io = bip->bip_end_io;
518 bio_endio_nodec(bio, error);
523 INIT_WORK(&bip->bip_work, bio_integrity_verify_fn);
524 queue_work(kintegrityd_wq, &bip->bip_work);
526 EXPORT_SYMBOL(bio_integrity_endio);
529 * bio_integrity_advance - Advance integrity vector
530 * @bio: bio whose integrity vector to update
531 * @bytes_done: number of data bytes that have been completed
533 * Description: This function calculates how many integrity bytes the
534 * number of completed data bytes correspond to and advances the
535 * integrity vector accordingly.
537 void bio_integrity_advance(struct bio *bio, unsigned int bytes_done)
539 struct bio_integrity_payload *bip = bio->bi_integrity;
540 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
541 unsigned bytes = bio_integrity_bytes(bi, bytes_done >> 9);
543 bvec_iter_advance(bip->bip_vec, &bip->bip_iter, bytes);
545 EXPORT_SYMBOL(bio_integrity_advance);
548 * bio_integrity_trim - Trim integrity vector
549 * @bio: bio whose integrity vector to update
550 * @offset: offset to first data sector
551 * @sectors: number of data sectors
553 * Description: Used to trim the integrity vector in a cloned bio.
554 * The ivec will be advanced corresponding to 'offset' data sectors
555 * and the length will be truncated corresponding to 'len' data
558 void bio_integrity_trim(struct bio *bio, unsigned int offset,
559 unsigned int sectors)
561 struct bio_integrity_payload *bip = bio->bi_integrity;
562 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
564 bio_integrity_advance(bio, offset << 9);
565 bip->bip_iter.bi_size = bio_integrity_bytes(bi, sectors);
567 EXPORT_SYMBOL(bio_integrity_trim);
570 * bio_integrity_clone - Callback for cloning bios with integrity metadata
572 * @bio_src: Original bio
573 * @gfp_mask: Memory allocation mask
575 * Description: Called to allocate a bip when cloning a bio
577 int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
580 struct bio_integrity_payload *bip_src = bio_src->bi_integrity;
581 struct bio_integrity_payload *bip;
583 BUG_ON(bip_src == NULL);
585 bip = bio_integrity_alloc(bio, gfp_mask, bip_src->bip_vcnt);
590 memcpy(bip->bip_vec, bip_src->bip_vec,
591 bip_src->bip_vcnt * sizeof(struct bio_vec));
593 bip->bip_vcnt = bip_src->bip_vcnt;
594 bip->bip_iter = bip_src->bip_iter;
598 EXPORT_SYMBOL(bio_integrity_clone);
600 int bioset_integrity_create(struct bio_set *bs, int pool_size)
602 if (bs->bio_integrity_pool)
605 bs->bio_integrity_pool = mempool_create_slab_pool(pool_size, bip_slab);
606 if (!bs->bio_integrity_pool)
609 bs->bvec_integrity_pool = biovec_create_pool(pool_size);
610 if (!bs->bvec_integrity_pool) {
611 mempool_destroy(bs->bio_integrity_pool);
617 EXPORT_SYMBOL(bioset_integrity_create);
619 void bioset_integrity_free(struct bio_set *bs)
621 if (bs->bio_integrity_pool)
622 mempool_destroy(bs->bio_integrity_pool);
624 if (bs->bvec_integrity_pool)
625 mempool_destroy(bs->bvec_integrity_pool);
627 EXPORT_SYMBOL(bioset_integrity_free);
629 void __init bio_integrity_init(void)
632 * kintegrityd won't block much but may burn a lot of CPU cycles.
633 * Make it highpri CPU intensive wq with max concurrency of 1.
635 kintegrityd_wq = alloc_workqueue("kintegrityd", WQ_MEM_RECLAIM |
636 WQ_HIGHPRI | WQ_CPU_INTENSIVE, 1);
638 panic("Failed to create kintegrityd\n");
640 bip_slab = kmem_cache_create("bio_integrity_payload",
641 sizeof(struct bio_integrity_payload) +
642 sizeof(struct bio_vec) * BIP_INLINE_VECS,
643 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
645 panic("Failed to create slab\n");