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);
117 static inline unsigned int bip_integrity_vecs(struct bio_integrity_payload *bip)
119 if (bip->bip_slab == BIO_POOL_NONE)
120 return BIP_INLINE_VECS;
122 return bvec_nr_vecs(bip->bip_slab);
126 * bio_integrity_add_page - Attach integrity metadata
127 * @bio: bio to update
128 * @page: page containing integrity metadata
129 * @len: number of bytes of integrity metadata in page
130 * @offset: start offset within page
132 * Description: Attach a page containing integrity metadata to bio.
134 int bio_integrity_add_page(struct bio *bio, struct page *page,
135 unsigned int len, unsigned int offset)
137 struct bio_integrity_payload *bip = bio->bi_integrity;
140 if (bip->bip_vcnt >= bip_integrity_vecs(bip)) {
141 printk(KERN_ERR "%s: bip_vec full\n", __func__);
145 iv = bip->bip_vec + bip->bip_vcnt;
149 iv->bv_offset = offset;
154 EXPORT_SYMBOL(bio_integrity_add_page);
156 static int bdev_integrity_enabled(struct block_device *bdev, int rw)
158 struct blk_integrity *bi = bdev_get_integrity(bdev);
163 if (rw == READ && bi->verify_fn != NULL &&
164 (bi->flags & INTEGRITY_FLAG_READ))
167 if (rw == WRITE && bi->generate_fn != NULL &&
168 (bi->flags & INTEGRITY_FLAG_WRITE))
175 * bio_integrity_enabled - Check whether integrity can be passed
178 * Description: Determines whether bio_integrity_prep() can be called
179 * on this bio or not. bio data direction and target device must be
180 * set prior to calling. The functions honors the write_generate and
181 * read_verify flags in sysfs.
183 int bio_integrity_enabled(struct bio *bio)
185 /* Already protected? */
186 if (bio_integrity(bio))
189 return bdev_integrity_enabled(bio->bi_bdev, bio_data_dir(bio));
191 EXPORT_SYMBOL(bio_integrity_enabled);
194 * bio_integrity_hw_sectors - Convert 512b sectors to hardware ditto
195 * @bi: blk_integrity profile for device
196 * @sectors: Number of 512 sectors to convert
198 * Description: The block layer calculates everything in 512 byte
199 * sectors but integrity metadata is done in terms of the hardware
200 * sector size of the storage device. Convert the block layer sectors
201 * to physical sectors.
203 static inline unsigned int bio_integrity_hw_sectors(struct blk_integrity *bi,
204 unsigned int sectors)
206 /* At this point there are only 512b or 4096b DIF/EPP devices */
207 if (bi->sector_size == 4096)
208 return sectors >>= 3;
213 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
214 unsigned int sectors)
216 return bio_integrity_hw_sectors(bi, sectors) * bi->tuple_size;
220 * bio_integrity_tag_size - Retrieve integrity tag space
221 * @bio: bio to inspect
223 * Description: Returns the maximum number of tag bytes that can be
224 * attached to this bio. Filesystems can use this to determine how
225 * much metadata to attach to an I/O.
227 unsigned int bio_integrity_tag_size(struct bio *bio)
229 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
231 BUG_ON(bio->bi_iter.bi_size == 0);
233 return bi->tag_size * (bio->bi_iter.bi_size / bi->sector_size);
235 EXPORT_SYMBOL(bio_integrity_tag_size);
237 static int bio_integrity_tag(struct bio *bio, void *tag_buf, unsigned int len,
240 struct bio_integrity_payload *bip = bio->bi_integrity;
241 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
242 unsigned int nr_sectors;
244 BUG_ON(bip->bip_buf == NULL);
246 if (bi->tag_size == 0)
249 nr_sectors = bio_integrity_hw_sectors(bi,
250 DIV_ROUND_UP(len, bi->tag_size));
252 if (nr_sectors * bi->tuple_size > bip->bip_iter.bi_size) {
253 printk(KERN_ERR "%s: tag too big for bio: %u > %u\n", __func__,
254 nr_sectors * bi->tuple_size, bip->bip_iter.bi_size);
259 bi->set_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
261 bi->get_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
267 * bio_integrity_set_tag - Attach a tag buffer to a bio
268 * @bio: bio to attach buffer to
269 * @tag_buf: Pointer to a buffer containing tag data
270 * @len: Length of the included buffer
272 * Description: Use this function to tag a bio by leveraging the extra
273 * space provided by devices formatted with integrity protection. The
274 * size of the integrity buffer must be <= to the size reported by
275 * bio_integrity_tag_size().
277 int bio_integrity_set_tag(struct bio *bio, void *tag_buf, unsigned int len)
279 BUG_ON(bio_data_dir(bio) != WRITE);
281 return bio_integrity_tag(bio, tag_buf, len, 1);
283 EXPORT_SYMBOL(bio_integrity_set_tag);
286 * bio_integrity_get_tag - Retrieve a tag buffer from a bio
287 * @bio: bio to retrieve buffer from
288 * @tag_buf: Pointer to a buffer for the tag data
289 * @len: Length of the target buffer
291 * Description: Use this function to retrieve the tag buffer from a
292 * completed I/O. The size of the integrity buffer must be <= to the
293 * size reported by bio_integrity_tag_size().
295 int bio_integrity_get_tag(struct bio *bio, void *tag_buf, unsigned int len)
297 BUG_ON(bio_data_dir(bio) != READ);
299 return bio_integrity_tag(bio, tag_buf, len, 0);
301 EXPORT_SYMBOL(bio_integrity_get_tag);
304 * bio_integrity_generate_verify - Generate/verify integrity metadata for a bio
305 * @bio: bio to generate/verify integrity metadata for
306 * @operate: operate number, 1 for generate, 0 for verify
308 static int bio_integrity_generate_verify(struct bio *bio, int operate)
310 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
311 struct blk_integrity_exchg bix;
314 unsigned int sectors, ret = 0, i;
315 void *prot_buf = bio->bi_integrity->bip_buf;
318 sector = bio->bi_iter.bi_sector;
320 sector = bio->bi_integrity->bip_iter.bi_sector;
322 bix.disk_name = bio->bi_bdev->bd_disk->disk_name;
323 bix.sector_size = bi->sector_size;
325 bio_for_each_segment_all(bv, bio, i) {
326 void *kaddr = kmap_atomic(bv->bv_page);
327 bix.data_buf = kaddr + bv->bv_offset;
328 bix.data_size = bv->bv_len;
329 bix.prot_buf = prot_buf;
333 bi->generate_fn(&bix);
335 ret = bi->verify_fn(&bix);
337 kunmap_atomic(kaddr);
342 sectors = bv->bv_len / bi->sector_size;
344 prot_buf += sectors * bi->tuple_size;
346 kunmap_atomic(kaddr);
352 * bio_integrity_generate - Generate integrity metadata for a bio
353 * @bio: bio to generate integrity metadata for
355 * Description: Generates integrity metadata for a bio by calling the
356 * block device's generation callback function. The bio must have a
357 * bip attached with enough room to accommodate the generated
358 * integrity metadata.
360 static void bio_integrity_generate(struct bio *bio)
362 bio_integrity_generate_verify(bio, 1);
365 static inline unsigned short blk_integrity_tuple_size(struct blk_integrity *bi)
368 return bi->tuple_size;
374 * bio_integrity_prep - Prepare bio for integrity I/O
375 * @bio: bio to prepare
377 * Description: Allocates a buffer for integrity metadata, maps the
378 * pages and attaches them to a bio. The bio must have data
379 * direction, target device and start sector set priot to calling. In
380 * the WRITE case, integrity metadata will be generated using the
381 * block device's integrity function. In the READ case, the buffer
382 * will be prepared for DMA and a suitable end_io handler set up.
384 int bio_integrity_prep(struct bio *bio)
386 struct bio_integrity_payload *bip;
387 struct blk_integrity *bi;
388 struct request_queue *q;
390 unsigned long start, end;
391 unsigned int len, nr_pages;
392 unsigned int bytes, offset, i;
393 unsigned int sectors;
395 bi = bdev_get_integrity(bio->bi_bdev);
396 q = bdev_get_queue(bio->bi_bdev);
398 BUG_ON(bio_integrity(bio));
400 sectors = bio_integrity_hw_sectors(bi, bio_sectors(bio));
402 /* Allocate kernel buffer for protection data */
403 len = sectors * blk_integrity_tuple_size(bi);
404 buf = kmalloc(len, GFP_NOIO | q->bounce_gfp);
405 if (unlikely(buf == NULL)) {
406 printk(KERN_ERR "could not allocate integrity buffer\n");
410 end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
411 start = ((unsigned long) buf) >> PAGE_SHIFT;
412 nr_pages = end - start;
414 /* Allocate bio integrity payload and integrity vectors */
415 bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages);
416 if (unlikely(bip == NULL)) {
417 printk(KERN_ERR "could not allocate data integrity bioset\n");
422 bip->bip_owns_buf = 1;
424 bip->bip_iter.bi_size = len;
425 bip->bip_iter.bi_sector = bio->bi_iter.bi_sector;
428 offset = offset_in_page(buf);
429 for (i = 0 ; i < nr_pages ; i++) {
431 bytes = PAGE_SIZE - offset;
439 ret = bio_integrity_add_page(bio, virt_to_page(buf),
453 /* Install custom I/O completion handler if read verify is enabled */
454 if (bio_data_dir(bio) == READ) {
455 bip->bip_end_io = bio->bi_end_io;
456 bio->bi_end_io = bio_integrity_endio;
459 /* Auto-generate integrity metadata if this is a write */
460 if (bio_data_dir(bio) == WRITE)
461 bio_integrity_generate(bio);
465 EXPORT_SYMBOL(bio_integrity_prep);
468 * bio_integrity_verify - Verify integrity metadata for a bio
469 * @bio: bio to verify
471 * Description: This function is called to verify the integrity of a
472 * bio. The data in the bio io_vec is compared to the integrity
473 * metadata returned by the HBA.
475 static int bio_integrity_verify(struct bio *bio)
477 return bio_integrity_generate_verify(bio, 0);
481 * bio_integrity_verify_fn - Integrity I/O completion worker
482 * @work: Work struct stored in bio to be verified
484 * Description: This workqueue function is called to complete a READ
485 * request. The function verifies the transferred integrity metadata
486 * and then calls the original bio end_io function.
488 static void bio_integrity_verify_fn(struct work_struct *work)
490 struct bio_integrity_payload *bip =
491 container_of(work, struct bio_integrity_payload, bip_work);
492 struct bio *bio = bip->bip_bio;
495 error = bio_integrity_verify(bio);
497 /* Restore original bio completion handler */
498 bio->bi_end_io = bip->bip_end_io;
499 bio_endio_nodec(bio, error);
503 * bio_integrity_endio - Integrity I/O completion function
504 * @bio: Protected bio
505 * @error: Pointer to errno
507 * Description: Completion for integrity I/O
509 * Normally I/O completion is done in interrupt context. However,
510 * verifying I/O integrity is a time-consuming task which must be run
511 * in process context. This function postpones completion
514 void bio_integrity_endio(struct bio *bio, int error)
516 struct bio_integrity_payload *bip = bio->bi_integrity;
518 BUG_ON(bip->bip_bio != bio);
520 /* In case of an I/O error there is no point in verifying the
521 * integrity metadata. Restore original bio end_io handler
525 bio->bi_end_io = bip->bip_end_io;
526 bio_endio(bio, error);
531 INIT_WORK(&bip->bip_work, bio_integrity_verify_fn);
532 queue_work(kintegrityd_wq, &bip->bip_work);
534 EXPORT_SYMBOL(bio_integrity_endio);
537 * bio_integrity_advance - Advance integrity vector
538 * @bio: bio whose integrity vector to update
539 * @bytes_done: number of data bytes that have been completed
541 * Description: This function calculates how many integrity bytes the
542 * number of completed data bytes correspond to and advances the
543 * integrity vector accordingly.
545 void bio_integrity_advance(struct bio *bio, unsigned int bytes_done)
547 struct bio_integrity_payload *bip = bio->bi_integrity;
548 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
549 unsigned bytes = bio_integrity_bytes(bi, bytes_done >> 9);
551 bvec_iter_advance(bip->bip_vec, &bip->bip_iter, bytes);
553 EXPORT_SYMBOL(bio_integrity_advance);
556 * bio_integrity_trim - Trim integrity vector
557 * @bio: bio whose integrity vector to update
558 * @offset: offset to first data sector
559 * @sectors: number of data sectors
561 * Description: Used to trim the integrity vector in a cloned bio.
562 * The ivec will be advanced corresponding to 'offset' data sectors
563 * and the length will be truncated corresponding to 'len' data
566 void bio_integrity_trim(struct bio *bio, unsigned int offset,
567 unsigned int sectors)
569 struct bio_integrity_payload *bip = bio->bi_integrity;
570 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
572 bio_integrity_advance(bio, offset << 9);
573 bip->bip_iter.bi_size = bio_integrity_bytes(bi, sectors);
575 EXPORT_SYMBOL(bio_integrity_trim);
578 * bio_integrity_clone - Callback for cloning bios with integrity metadata
580 * @bio_src: Original bio
581 * @gfp_mask: Memory allocation mask
583 * Description: Called to allocate a bip when cloning a bio
585 int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
588 struct bio_integrity_payload *bip_src = bio_src->bi_integrity;
589 struct bio_integrity_payload *bip;
591 BUG_ON(bip_src == NULL);
593 bip = bio_integrity_alloc(bio, gfp_mask, bip_src->bip_vcnt);
598 memcpy(bip->bip_vec, bip_src->bip_vec,
599 bip_src->bip_vcnt * sizeof(struct bio_vec));
601 bip->bip_vcnt = bip_src->bip_vcnt;
602 bip->bip_iter = bip_src->bip_iter;
606 EXPORT_SYMBOL(bio_integrity_clone);
608 int bioset_integrity_create(struct bio_set *bs, int pool_size)
610 if (bs->bio_integrity_pool)
613 bs->bio_integrity_pool = mempool_create_slab_pool(pool_size, bip_slab);
614 if (!bs->bio_integrity_pool)
617 bs->bvec_integrity_pool = biovec_create_pool(bs, pool_size);
618 if (!bs->bvec_integrity_pool) {
619 mempool_destroy(bs->bio_integrity_pool);
625 EXPORT_SYMBOL(bioset_integrity_create);
627 void bioset_integrity_free(struct bio_set *bs)
629 if (bs->bio_integrity_pool)
630 mempool_destroy(bs->bio_integrity_pool);
632 if (bs->bvec_integrity_pool)
633 mempool_destroy(bs->bvec_integrity_pool);
635 EXPORT_SYMBOL(bioset_integrity_free);
637 void __init bio_integrity_init(void)
640 * kintegrityd won't block much but may burn a lot of CPU cycles.
641 * Make it highpri CPU intensive wq with max concurrency of 1.
643 kintegrityd_wq = alloc_workqueue("kintegrityd", WQ_MEM_RECLAIM |
644 WQ_HIGHPRI | WQ_CPU_INTENSIVE, 1);
646 panic("Failed to create kintegrityd\n");
648 bip_slab = kmem_cache_create("bio_integrity_payload",
649 sizeof(struct bio_integrity_payload) +
650 sizeof(struct bio_vec) * BIP_INLINE_VECS,
651 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
653 panic("Failed to create slab\n");