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>
31 #define BIP_INLINE_VECS 4
33 static struct kmem_cache *bip_slab;
34 static struct workqueue_struct *kintegrityd_wq;
36 void blk_flush_integrity(void)
38 flush_workqueue(kintegrityd_wq);
42 * bio_integrity_alloc - Allocate integrity payload and attach it to bio
43 * @bio: bio to attach integrity metadata to
44 * @gfp_mask: Memory allocation mask
45 * @nr_vecs: Number of integrity metadata scatter-gather elements
47 * Description: This function prepares a bio for attaching integrity
48 * metadata. nr_vecs specifies the maximum number of pages containing
49 * integrity metadata that can be attached.
51 struct bio_integrity_payload *bio_integrity_alloc(struct bio *bio,
55 struct bio_integrity_payload *bip;
56 struct bio_set *bs = bio->bi_pool;
57 unsigned long idx = BIO_POOL_NONE;
60 if (!bs || !bs->bio_integrity_pool) {
61 bip = kmalloc(sizeof(struct bio_integrity_payload) +
62 sizeof(struct bio_vec) * nr_vecs, gfp_mask);
63 inline_vecs = nr_vecs;
65 bip = mempool_alloc(bs->bio_integrity_pool, gfp_mask);
66 inline_vecs = BIP_INLINE_VECS;
70 return ERR_PTR(-ENOMEM);
72 memset(bip, 0, sizeof(*bip));
74 if (nr_vecs > inline_vecs) {
75 bip->bip_vec = bvec_alloc(gfp_mask, nr_vecs, &idx,
76 bs->bvec_integrity_pool);
79 bip->bip_max_vcnt = bvec_nr_vecs(idx);
81 bip->bip_vec = bip->bip_inline_vecs;
82 bip->bip_max_vcnt = inline_vecs;
87 bio->bi_integrity = bip;
88 bio->bi_rw |= REQ_INTEGRITY;
92 mempool_free(bip, bs->bio_integrity_pool);
93 return ERR_PTR(-ENOMEM);
95 EXPORT_SYMBOL(bio_integrity_alloc);
98 * bio_integrity_free - Free bio integrity payload
99 * @bio: bio containing bip to be freed
101 * Description: Used to free the integrity portion of a bio. Usually
102 * called from bio_free().
104 void bio_integrity_free(struct bio *bio)
106 struct bio_integrity_payload *bip = bio_integrity(bio);
107 struct bio_set *bs = bio->bi_pool;
109 if (bip->bip_flags & BIP_BLOCK_INTEGRITY)
110 kfree(page_address(bip->bip_vec->bv_page) +
111 bip->bip_vec->bv_offset);
113 if (bs && bs->bio_integrity_pool) {
114 if (bip->bip_slab != BIO_POOL_NONE)
115 bvec_free(bs->bvec_integrity_pool, bip->bip_vec,
118 mempool_free(bip, bs->bio_integrity_pool);
123 bio->bi_integrity = NULL;
125 EXPORT_SYMBOL(bio_integrity_free);
128 * bio_integrity_add_page - Attach integrity metadata
129 * @bio: bio to update
130 * @page: page containing integrity metadata
131 * @len: number of bytes of integrity metadata in page
132 * @offset: start offset within page
134 * Description: Attach a page containing integrity metadata to bio.
136 int bio_integrity_add_page(struct bio *bio, struct page *page,
137 unsigned int len, unsigned int offset)
139 struct bio_integrity_payload *bip = bio_integrity(bio);
142 if (bip->bip_vcnt >= bip->bip_max_vcnt) {
143 printk(KERN_ERR "%s: bip_vec full\n", __func__);
147 iv = bip->bip_vec + bip->bip_vcnt;
150 bvec_gap_to_prev(bdev_get_queue(bio->bi_bdev),
151 &bip->bip_vec[bip->bip_vcnt - 1], offset))
156 iv->bv_offset = offset;
161 EXPORT_SYMBOL(bio_integrity_add_page);
164 * bio_integrity_enabled - Check whether integrity can be passed
167 * Description: Determines whether bio_integrity_prep() can be called
168 * on this bio or not. bio data direction and target device must be
169 * set prior to calling. The functions honors the write_generate and
170 * read_verify flags in sysfs.
172 bool bio_integrity_enabled(struct bio *bio)
174 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
179 /* Already protected? */
180 if (bio_integrity(bio))
186 if (bio_data_dir(bio) == READ && bi->profile->verify_fn != NULL &&
187 (bi->flags & BLK_INTEGRITY_VERIFY))
190 if (bio_data_dir(bio) == WRITE && bi->profile->generate_fn != NULL &&
191 (bi->flags & BLK_INTEGRITY_GENERATE))
196 EXPORT_SYMBOL(bio_integrity_enabled);
199 * bio_integrity_intervals - Return number of integrity intervals for a bio
200 * @bi: blk_integrity profile for device
201 * @sectors: Size of the bio in 512-byte sectors
203 * Description: The block layer calculates everything in 512 byte
204 * sectors but integrity metadata is done in terms of the data integrity
205 * interval size of the storage device. Convert the block layer sectors
206 * to the appropriate number of integrity intervals.
208 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
209 unsigned int sectors)
211 return sectors >> (bi->interval_exp - 9);
214 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
215 unsigned int sectors)
217 return bio_integrity_intervals(bi, sectors) * bi->tuple_size;
221 * bio_integrity_process - Process integrity metadata for a bio
222 * @bio: bio to generate/verify integrity metadata for
223 * @proc_fn: Pointer to the relevant processing function
225 static int bio_integrity_process(struct bio *bio,
226 integrity_processing_fn *proc_fn)
228 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
229 struct blk_integrity_iter iter;
230 struct bvec_iter bviter;
232 struct bio_integrity_payload *bip = bio_integrity(bio);
233 unsigned int ret = 0;
234 void *prot_buf = page_address(bip->bip_vec->bv_page) +
235 bip->bip_vec->bv_offset;
237 iter.disk_name = bio->bi_bdev->bd_disk->disk_name;
238 iter.interval = 1 << bi->interval_exp;
239 iter.seed = bip_get_seed(bip);
240 iter.prot_buf = prot_buf;
242 bio_for_each_segment(bv, bio, bviter) {
243 void *kaddr = kmap_atomic(bv.bv_page);
245 iter.data_buf = kaddr + bv.bv_offset;
246 iter.data_size = bv.bv_len;
248 ret = proc_fn(&iter);
250 kunmap_atomic(kaddr);
254 kunmap_atomic(kaddr);
260 * bio_integrity_prep - Prepare bio for integrity I/O
261 * @bio: bio to prepare
263 * Description: Allocates a buffer for integrity metadata, maps the
264 * pages and attaches them to a bio. The bio must have data
265 * direction, target device and start sector set priot to calling. In
266 * the WRITE case, integrity metadata will be generated using the
267 * block device's integrity function. In the READ case, the buffer
268 * will be prepared for DMA and a suitable end_io handler set up.
270 int bio_integrity_prep(struct bio *bio)
272 struct bio_integrity_payload *bip;
273 struct blk_integrity *bi;
274 struct request_queue *q;
276 unsigned long start, end;
277 unsigned int len, nr_pages;
278 unsigned int bytes, offset, i;
279 unsigned int intervals;
281 bi = bdev_get_integrity(bio->bi_bdev);
282 q = bdev_get_queue(bio->bi_bdev);
284 BUG_ON(bio_integrity(bio));
286 intervals = bio_integrity_intervals(bi, bio_sectors(bio));
288 /* Allocate kernel buffer for protection data */
289 len = intervals * bi->tuple_size;
290 buf = kmalloc(len, GFP_NOIO | q->bounce_gfp);
291 if (unlikely(buf == NULL)) {
292 printk(KERN_ERR "could not allocate integrity buffer\n");
296 end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
297 start = ((unsigned long) buf) >> PAGE_SHIFT;
298 nr_pages = end - start;
300 /* Allocate bio integrity payload and integrity vectors */
301 bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages);
303 printk(KERN_ERR "could not allocate data integrity bioset\n");
308 bip->bip_flags |= BIP_BLOCK_INTEGRITY;
309 bip->bip_iter.bi_size = len;
310 bip_set_seed(bip, bio->bi_iter.bi_sector);
312 if (bi->flags & BLK_INTEGRITY_IP_CHECKSUM)
313 bip->bip_flags |= BIP_IP_CHECKSUM;
316 offset = offset_in_page(buf);
317 for (i = 0 ; i < nr_pages ; i++) {
319 bytes = PAGE_SIZE - offset;
327 ret = bio_integrity_add_page(bio, virt_to_page(buf),
341 /* Install custom I/O completion handler if read verify is enabled */
342 if (bio_data_dir(bio) == READ) {
343 bip->bip_end_io = bio->bi_end_io;
344 bio->bi_end_io = bio_integrity_endio;
347 /* Auto-generate integrity metadata if this is a write */
348 if (bio_data_dir(bio) == WRITE)
349 bio_integrity_process(bio, bi->profile->generate_fn);
353 EXPORT_SYMBOL(bio_integrity_prep);
356 * bio_integrity_verify_fn - Integrity I/O completion worker
357 * @work: Work struct stored in bio to be verified
359 * Description: This workqueue function is called to complete a READ
360 * request. The function verifies the transferred integrity metadata
361 * and then calls the original bio end_io function.
363 static void bio_integrity_verify_fn(struct work_struct *work)
365 struct bio_integrity_payload *bip =
366 container_of(work, struct bio_integrity_payload, bip_work);
367 struct bio *bio = bip->bip_bio;
368 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
370 bio->bi_error = bio_integrity_process(bio, bi->profile->verify_fn);
372 /* Restore original bio completion handler */
373 bio->bi_end_io = bip->bip_end_io;
378 * bio_integrity_endio - Integrity I/O completion function
379 * @bio: Protected bio
380 * @error: Pointer to errno
382 * Description: Completion for integrity I/O
384 * Normally I/O completion is done in interrupt context. However,
385 * verifying I/O integrity is a time-consuming task which must be run
386 * in process context. This function postpones completion
389 void bio_integrity_endio(struct bio *bio)
391 struct bio_integrity_payload *bip = bio_integrity(bio);
393 BUG_ON(bip->bip_bio != bio);
395 /* In case of an I/O error there is no point in verifying the
396 * integrity metadata. Restore original bio end_io handler
400 bio->bi_end_io = bip->bip_end_io;
406 INIT_WORK(&bip->bip_work, bio_integrity_verify_fn);
407 queue_work(kintegrityd_wq, &bip->bip_work);
409 EXPORT_SYMBOL(bio_integrity_endio);
412 * bio_integrity_advance - Advance integrity vector
413 * @bio: bio whose integrity vector to update
414 * @bytes_done: number of data bytes that have been completed
416 * Description: This function calculates how many integrity bytes the
417 * number of completed data bytes correspond to and advances the
418 * integrity vector accordingly.
420 void bio_integrity_advance(struct bio *bio, unsigned int bytes_done)
422 struct bio_integrity_payload *bip = bio_integrity(bio);
423 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
424 unsigned bytes = bio_integrity_bytes(bi, bytes_done >> 9);
426 bvec_iter_advance(bip->bip_vec, &bip->bip_iter, bytes);
428 EXPORT_SYMBOL(bio_integrity_advance);
431 * bio_integrity_trim - Trim integrity vector
432 * @bio: bio whose integrity vector to update
433 * @offset: offset to first data sector
434 * @sectors: number of data sectors
436 * Description: Used to trim the integrity vector in a cloned bio.
437 * The ivec will be advanced corresponding to 'offset' data sectors
438 * and the length will be truncated corresponding to 'len' data
441 void bio_integrity_trim(struct bio *bio, unsigned int offset,
442 unsigned int sectors)
444 struct bio_integrity_payload *bip = bio_integrity(bio);
445 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
447 bio_integrity_advance(bio, offset << 9);
448 bip->bip_iter.bi_size = bio_integrity_bytes(bi, sectors);
450 EXPORT_SYMBOL(bio_integrity_trim);
453 * bio_integrity_clone - Callback for cloning bios with integrity metadata
455 * @bio_src: Original bio
456 * @gfp_mask: Memory allocation mask
458 * Description: Called to allocate a bip when cloning a bio
460 int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
463 struct bio_integrity_payload *bip_src = bio_integrity(bio_src);
464 struct bio_integrity_payload *bip;
466 BUG_ON(bip_src == NULL);
468 bip = bio_integrity_alloc(bio, gfp_mask, bip_src->bip_vcnt);
472 memcpy(bip->bip_vec, bip_src->bip_vec,
473 bip_src->bip_vcnt * sizeof(struct bio_vec));
475 bip->bip_vcnt = bip_src->bip_vcnt;
476 bip->bip_iter = bip_src->bip_iter;
480 EXPORT_SYMBOL(bio_integrity_clone);
482 int bioset_integrity_create(struct bio_set *bs, int pool_size)
484 if (bs->bio_integrity_pool)
487 bs->bio_integrity_pool = mempool_create_slab_pool(pool_size, bip_slab);
488 if (!bs->bio_integrity_pool)
491 bs->bvec_integrity_pool = biovec_create_pool(pool_size);
492 if (!bs->bvec_integrity_pool) {
493 mempool_destroy(bs->bio_integrity_pool);
499 EXPORT_SYMBOL(bioset_integrity_create);
501 void bioset_integrity_free(struct bio_set *bs)
503 if (bs->bio_integrity_pool)
504 mempool_destroy(bs->bio_integrity_pool);
506 if (bs->bvec_integrity_pool)
507 mempool_destroy(bs->bvec_integrity_pool);
509 EXPORT_SYMBOL(bioset_integrity_free);
511 void __init bio_integrity_init(void)
514 * kintegrityd won't block much but may burn a lot of CPU cycles.
515 * Make it highpri CPU intensive wq with max concurrency of 1.
517 kintegrityd_wq = alloc_workqueue("kintegrityd", WQ_MEM_RECLAIM |
518 WQ_HIGHPRI | WQ_CPU_INTENSIVE, 1);
520 panic("Failed to create kintegrityd\n");
522 bip_slab = kmem_cache_create("bio_integrity_payload",
523 sizeof(struct bio_integrity_payload) +
524 sizeof(struct bio_vec) * BIP_INLINE_VECS,
525 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);