Root/block/bio-integrity.c

1/*
2 * bio-integrity.c - bio data integrity extensions
3 *
4 * Copyright (C) 2007, 2008, 2009 Oracle Corporation
5 * Written by: Martin K. Petersen <martin.petersen@oracle.com>
6 *
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.
10 *
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.
15 *
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,
19 * USA.
20 *
21 */
22
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>
29
30#define BIP_INLINE_VECS 4
31
32static struct kmem_cache *bip_slab;
33static struct workqueue_struct *kintegrityd_wq;
34
35/**
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
40 *
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.
44 */
45struct bio_integrity_payload *bio_integrity_alloc(struct bio *bio,
46                          gfp_t gfp_mask,
47                          unsigned int nr_vecs)
48{
49    struct bio_integrity_payload *bip;
50    struct bio_set *bs = bio->bi_pool;
51    unsigned long idx = BIO_POOL_NONE;
52    unsigned inline_vecs;
53
54    if (!bs) {
55        bip = kmalloc(sizeof(struct bio_integrity_payload) +
56                  sizeof(struct bio_vec) * nr_vecs, gfp_mask);
57        inline_vecs = nr_vecs;
58    } else {
59        bip = mempool_alloc(bs->bio_integrity_pool, gfp_mask);
60        inline_vecs = BIP_INLINE_VECS;
61    }
62
63    if (unlikely(!bip))
64        return NULL;
65
66    memset(bip, 0, sizeof(*bip));
67
68    if (nr_vecs > inline_vecs) {
69        bip->bip_vec = bvec_alloc(gfp_mask, nr_vecs, &idx,
70                      bs->bvec_integrity_pool);
71        if (!bip->bip_vec)
72            goto err;
73    } else {
74        bip->bip_vec = bip->bip_inline_vecs;
75    }
76
77    bip->bip_slab = idx;
78    bip->bip_bio = bio;
79    bio->bi_integrity = bip;
80
81    return bip;
82err:
83    mempool_free(bip, bs->bio_integrity_pool);
84    return NULL;
85}
86EXPORT_SYMBOL(bio_integrity_alloc);
87
88/**
89 * bio_integrity_free - Free bio integrity payload
90 * @bio: bio containing bip to be freed
91 *
92 * Description: Used to free the integrity portion of a bio. Usually
93 * called from bio_free().
94 */
95void bio_integrity_free(struct bio *bio)
96{
97    struct bio_integrity_payload *bip = bio->bi_integrity;
98    struct bio_set *bs = bio->bi_pool;
99
100    if (bip->bip_owns_buf)
101        kfree(bip->bip_buf);
102
103    if (bs) {
104        if (bip->bip_slab != BIO_POOL_NONE)
105            bvec_free(bs->bvec_integrity_pool, bip->bip_vec,
106                  bip->bip_slab);
107
108        mempool_free(bip, bs->bio_integrity_pool);
109    } else {
110        kfree(bip);
111    }
112
113    bio->bi_integrity = NULL;
114}
115EXPORT_SYMBOL(bio_integrity_free);
116
117static inline unsigned int bip_integrity_vecs(struct bio_integrity_payload *bip)
118{
119    if (bip->bip_slab == BIO_POOL_NONE)
120        return BIP_INLINE_VECS;
121
122    return bvec_nr_vecs(bip->bip_slab);
123}
124
125/**
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
131 *
132 * Description: Attach a page containing integrity metadata to bio.
133 */
134int bio_integrity_add_page(struct bio *bio, struct page *page,
135               unsigned int len, unsigned int offset)
136{
137    struct bio_integrity_payload *bip = bio->bi_integrity;
138    struct bio_vec *iv;
139
140    if (bip->bip_vcnt >= bip_integrity_vecs(bip)) {
141        printk(KERN_ERR "%s: bip_vec full\n", __func__);
142        return 0;
143    }
144
145    iv = bip->bip_vec + bip->bip_vcnt;
146
147    iv->bv_page = page;
148    iv->bv_len = len;
149    iv->bv_offset = offset;
150    bip->bip_vcnt++;
151
152    return len;
153}
154EXPORT_SYMBOL(bio_integrity_add_page);
155
156static int bdev_integrity_enabled(struct block_device *bdev, int rw)
157{
158    struct blk_integrity *bi = bdev_get_integrity(bdev);
159
160    if (bi == NULL)
161        return 0;
162
163    if (rw == READ && bi->verify_fn != NULL &&
164        (bi->flags & INTEGRITY_FLAG_READ))
165        return 1;
166
167    if (rw == WRITE && bi->generate_fn != NULL &&
168        (bi->flags & INTEGRITY_FLAG_WRITE))
169        return 1;
170
171    return 0;
172}
173
174/**
175 * bio_integrity_enabled - Check whether integrity can be passed
176 * @bio: bio to check
177 *
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.
182 */
183int bio_integrity_enabled(struct bio *bio)
184{
185    if (!bio_is_rw(bio))
186        return 0;
187
188    /* Already protected? */
189    if (bio_integrity(bio))
190        return 0;
191
192    return bdev_integrity_enabled(bio->bi_bdev, bio_data_dir(bio));
193}
194EXPORT_SYMBOL(bio_integrity_enabled);
195
196/**
197 * bio_integrity_hw_sectors - Convert 512b sectors to hardware ditto
198 * @bi: blk_integrity profile for device
199 * @sectors: Number of 512 sectors to convert
200 *
201 * Description: The block layer calculates everything in 512 byte
202 * sectors but integrity metadata is done in terms of the hardware
203 * sector size of the storage device. Convert the block layer sectors
204 * to physical sectors.
205 */
206static inline unsigned int bio_integrity_hw_sectors(struct blk_integrity *bi,
207                            unsigned int sectors)
208{
209    /* At this point there are only 512b or 4096b DIF/EPP devices */
210    if (bi->sector_size == 4096)
211        return sectors >>= 3;
212
213    return sectors;
214}
215
216static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
217                           unsigned int sectors)
218{
219    return bio_integrity_hw_sectors(bi, sectors) * bi->tuple_size;
220}
221
222/**
223 * bio_integrity_tag_size - Retrieve integrity tag space
224 * @bio: bio to inspect
225 *
226 * Description: Returns the maximum number of tag bytes that can be
227 * attached to this bio. Filesystems can use this to determine how
228 * much metadata to attach to an I/O.
229 */
230unsigned int bio_integrity_tag_size(struct bio *bio)
231{
232    struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
233
234    BUG_ON(bio->bi_iter.bi_size == 0);
235
236    return bi->tag_size * (bio->bi_iter.bi_size / bi->sector_size);
237}
238EXPORT_SYMBOL(bio_integrity_tag_size);
239
240static int bio_integrity_tag(struct bio *bio, void *tag_buf, unsigned int len,
241                 int set)
242{
243    struct bio_integrity_payload *bip = bio->bi_integrity;
244    struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
245    unsigned int nr_sectors;
246
247    BUG_ON(bip->bip_buf == NULL);
248
249    if (bi->tag_size == 0)
250        return -1;
251
252    nr_sectors = bio_integrity_hw_sectors(bi,
253                    DIV_ROUND_UP(len, bi->tag_size));
254
255    if (nr_sectors * bi->tuple_size > bip->bip_iter.bi_size) {
256        printk(KERN_ERR "%s: tag too big for bio: %u > %u\n", __func__,
257               nr_sectors * bi->tuple_size, bip->bip_iter.bi_size);
258        return -1;
259    }
260
261    if (set)
262        bi->set_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
263    else
264        bi->get_tag_fn(bip->bip_buf, tag_buf, nr_sectors);
265
266    return 0;
267}
268
269/**
270 * bio_integrity_set_tag - Attach a tag buffer to a bio
271 * @bio: bio to attach buffer to
272 * @tag_buf: Pointer to a buffer containing tag data
273 * @len: Length of the included buffer
274 *
275 * Description: Use this function to tag a bio by leveraging the extra
276 * space provided by devices formatted with integrity protection. The
277 * size of the integrity buffer must be <= to the size reported by
278 * bio_integrity_tag_size().
279 */
280int bio_integrity_set_tag(struct bio *bio, void *tag_buf, unsigned int len)
281{
282    BUG_ON(bio_data_dir(bio) != WRITE);
283
284    return bio_integrity_tag(bio, tag_buf, len, 1);
285}
286EXPORT_SYMBOL(bio_integrity_set_tag);
287
288/**
289 * bio_integrity_get_tag - Retrieve a tag buffer from a bio
290 * @bio: bio to retrieve buffer from
291 * @tag_buf: Pointer to a buffer for the tag data
292 * @len: Length of the target buffer
293 *
294 * Description: Use this function to retrieve the tag buffer from a
295 * completed I/O. The size of the integrity buffer must be <= to the
296 * size reported by bio_integrity_tag_size().
297 */
298int bio_integrity_get_tag(struct bio *bio, void *tag_buf, unsigned int len)
299{
300    BUG_ON(bio_data_dir(bio) != READ);
301
302    return bio_integrity_tag(bio, tag_buf, len, 0);
303}
304EXPORT_SYMBOL(bio_integrity_get_tag);
305
306/**
307 * bio_integrity_generate_verify - Generate/verify integrity metadata for a bio
308 * @bio: bio to generate/verify integrity metadata for
309 * @operate: operate number, 1 for generate, 0 for verify
310 */
311static int bio_integrity_generate_verify(struct bio *bio, int operate)
312{
313    struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
314    struct blk_integrity_exchg bix;
315    struct bio_vec *bv;
316    sector_t sector;
317    unsigned int sectors, ret = 0, i;
318    void *prot_buf = bio->bi_integrity->bip_buf;
319
320    if (operate)
321        sector = bio->bi_iter.bi_sector;
322    else
323        sector = bio->bi_integrity->bip_iter.bi_sector;
324
325    bix.disk_name = bio->bi_bdev->bd_disk->disk_name;
326    bix.sector_size = bi->sector_size;
327
328    bio_for_each_segment_all(bv, bio, i) {
329        void *kaddr = kmap_atomic(bv->bv_page);
330        bix.data_buf = kaddr + bv->bv_offset;
331        bix.data_size = bv->bv_len;
332        bix.prot_buf = prot_buf;
333        bix.sector = sector;
334
335        if (operate)
336            bi->generate_fn(&bix);
337        else {
338            ret = bi->verify_fn(&bix);
339            if (ret) {
340                kunmap_atomic(kaddr);
341                return ret;
342            }
343        }
344
345        sectors = bv->bv_len / bi->sector_size;
346        sector += sectors;
347        prot_buf += sectors * bi->tuple_size;
348
349        kunmap_atomic(kaddr);
350    }
351    return ret;
352}
353
354/**
355 * bio_integrity_generate - Generate integrity metadata for a bio
356 * @bio: bio to generate integrity metadata for
357 *
358 * Description: Generates integrity metadata for a bio by calling the
359 * block device's generation callback function. The bio must have a
360 * bip attached with enough room to accommodate the generated
361 * integrity metadata.
362 */
363static void bio_integrity_generate(struct bio *bio)
364{
365    bio_integrity_generate_verify(bio, 1);
366}
367
368static inline unsigned short blk_integrity_tuple_size(struct blk_integrity *bi)
369{
370    if (bi)
371        return bi->tuple_size;
372
373    return 0;
374}
375
376/**
377 * bio_integrity_prep - Prepare bio for integrity I/O
378 * @bio: bio to prepare
379 *
380 * Description: Allocates a buffer for integrity metadata, maps the
381 * pages and attaches them to a bio. The bio must have data
382 * direction, target device and start sector set priot to calling. In
383 * the WRITE case, integrity metadata will be generated using the
384 * block device's integrity function. In the READ case, the buffer
385 * will be prepared for DMA and a suitable end_io handler set up.
386 */
387int bio_integrity_prep(struct bio *bio)
388{
389    struct bio_integrity_payload *bip;
390    struct blk_integrity *bi;
391    struct request_queue *q;
392    void *buf;
393    unsigned long start, end;
394    unsigned int len, nr_pages;
395    unsigned int bytes, offset, i;
396    unsigned int sectors;
397
398    bi = bdev_get_integrity(bio->bi_bdev);
399    q = bdev_get_queue(bio->bi_bdev);
400    BUG_ON(bi == NULL);
401    BUG_ON(bio_integrity(bio));
402
403    sectors = bio_integrity_hw_sectors(bi, bio_sectors(bio));
404
405    /* Allocate kernel buffer for protection data */
406    len = sectors * blk_integrity_tuple_size(bi);
407    buf = kmalloc(len, GFP_NOIO | q->bounce_gfp);
408    if (unlikely(buf == NULL)) {
409        printk(KERN_ERR "could not allocate integrity buffer\n");
410        return -ENOMEM;
411    }
412
413    end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
414    start = ((unsigned long) buf) >> PAGE_SHIFT;
415    nr_pages = end - start;
416
417    /* Allocate bio integrity payload and integrity vectors */
418    bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages);
419    if (unlikely(bip == NULL)) {
420        printk(KERN_ERR "could not allocate data integrity bioset\n");
421        kfree(buf);
422        return -EIO;
423    }
424
425    bip->bip_owns_buf = 1;
426    bip->bip_buf = buf;
427    bip->bip_iter.bi_size = len;
428    bip->bip_iter.bi_sector = bio->bi_iter.bi_sector;
429
430    /* Map it */
431    offset = offset_in_page(buf);
432    for (i = 0 ; i < nr_pages ; i++) {
433        int ret;
434        bytes = PAGE_SIZE - offset;
435
436        if (len <= 0)
437            break;
438
439        if (bytes > len)
440            bytes = len;
441
442        ret = bio_integrity_add_page(bio, virt_to_page(buf),
443                         bytes, offset);
444
445        if (ret == 0)
446            return 0;
447
448        if (ret < bytes)
449            break;
450
451        buf += bytes;
452        len -= bytes;
453        offset = 0;
454    }
455
456    /* Install custom I/O completion handler if read verify is enabled */
457    if (bio_data_dir(bio) == READ) {
458        bip->bip_end_io = bio->bi_end_io;
459        bio->bi_end_io = bio_integrity_endio;
460    }
461
462    /* Auto-generate integrity metadata if this is a write */
463    if (bio_data_dir(bio) == WRITE)
464        bio_integrity_generate(bio);
465
466    return 0;
467}
468EXPORT_SYMBOL(bio_integrity_prep);
469
470/**
471 * bio_integrity_verify - Verify integrity metadata for a bio
472 * @bio: bio to verify
473 *
474 * Description: This function is called to verify the integrity of a
475 * bio. The data in the bio io_vec is compared to the integrity
476 * metadata returned by the HBA.
477 */
478static int bio_integrity_verify(struct bio *bio)
479{
480    return bio_integrity_generate_verify(bio, 0);
481}
482
483/**
484 * bio_integrity_verify_fn - Integrity I/O completion worker
485 * @work: Work struct stored in bio to be verified
486 *
487 * Description: This workqueue function is called to complete a READ
488 * request. The function verifies the transferred integrity metadata
489 * and then calls the original bio end_io function.
490 */
491static void bio_integrity_verify_fn(struct work_struct *work)
492{
493    struct bio_integrity_payload *bip =
494        container_of(work, struct bio_integrity_payload, bip_work);
495    struct bio *bio = bip->bip_bio;
496    int error;
497
498    error = bio_integrity_verify(bio);
499
500    /* Restore original bio completion handler */
501    bio->bi_end_io = bip->bip_end_io;
502    bio_endio_nodec(bio, error);
503}
504
505/**
506 * bio_integrity_endio - Integrity I/O completion function
507 * @bio: Protected bio
508 * @error: Pointer to errno
509 *
510 * Description: Completion for integrity I/O
511 *
512 * Normally I/O completion is done in interrupt context. However,
513 * verifying I/O integrity is a time-consuming task which must be run
514 * in process context. This function postpones completion
515 * accordingly.
516 */
517void bio_integrity_endio(struct bio *bio, int error)
518{
519    struct bio_integrity_payload *bip = bio->bi_integrity;
520
521    BUG_ON(bip->bip_bio != bio);
522
523    /* In case of an I/O error there is no point in verifying the
524     * integrity metadata. Restore original bio end_io handler
525     * and run it.
526     */
527    if (error) {
528        bio->bi_end_io = bip->bip_end_io;
529        bio_endio(bio, error);
530
531        return;
532    }
533
534    INIT_WORK(&bip->bip_work, bio_integrity_verify_fn);
535    queue_work(kintegrityd_wq, &bip->bip_work);
536}
537EXPORT_SYMBOL(bio_integrity_endio);
538
539/**
540 * bio_integrity_advance - Advance integrity vector
541 * @bio: bio whose integrity vector to update
542 * @bytes_done: number of data bytes that have been completed
543 *
544 * Description: This function calculates how many integrity bytes the
545 * number of completed data bytes correspond to and advances the
546 * integrity vector accordingly.
547 */
548void bio_integrity_advance(struct bio *bio, unsigned int bytes_done)
549{
550    struct bio_integrity_payload *bip = bio->bi_integrity;
551    struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
552    unsigned bytes = bio_integrity_bytes(bi, bytes_done >> 9);
553
554    bvec_iter_advance(bip->bip_vec, &bip->bip_iter, bytes);
555}
556EXPORT_SYMBOL(bio_integrity_advance);
557
558/**
559 * bio_integrity_trim - Trim integrity vector
560 * @bio: bio whose integrity vector to update
561 * @offset: offset to first data sector
562 * @sectors: number of data sectors
563 *
564 * Description: Used to trim the integrity vector in a cloned bio.
565 * The ivec will be advanced corresponding to 'offset' data sectors
566 * and the length will be truncated corresponding to 'len' data
567 * sectors.
568 */
569void bio_integrity_trim(struct bio *bio, unsigned int offset,
570            unsigned int sectors)
571{
572    struct bio_integrity_payload *bip = bio->bi_integrity;
573    struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev);
574
575    bio_integrity_advance(bio, offset << 9);
576    bip->bip_iter.bi_size = bio_integrity_bytes(bi, sectors);
577}
578EXPORT_SYMBOL(bio_integrity_trim);
579
580/**
581 * bio_integrity_clone - Callback for cloning bios with integrity metadata
582 * @bio: New bio
583 * @bio_src: Original bio
584 * @gfp_mask: Memory allocation mask
585 *
586 * Description: Called to allocate a bip when cloning a bio
587 */
588int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
589            gfp_t gfp_mask)
590{
591    struct bio_integrity_payload *bip_src = bio_src->bi_integrity;
592    struct bio_integrity_payload *bip;
593
594    BUG_ON(bip_src == NULL);
595
596    bip = bio_integrity_alloc(bio, gfp_mask, bip_src->bip_vcnt);
597
598    if (bip == NULL)
599        return -EIO;
600
601    memcpy(bip->bip_vec, bip_src->bip_vec,
602           bip_src->bip_vcnt * sizeof(struct bio_vec));
603
604    bip->bip_vcnt = bip_src->bip_vcnt;
605    bip->bip_iter = bip_src->bip_iter;
606
607    return 0;
608}
609EXPORT_SYMBOL(bio_integrity_clone);
610
611int bioset_integrity_create(struct bio_set *bs, int pool_size)
612{
613    if (bs->bio_integrity_pool)
614        return 0;
615
616    bs->bio_integrity_pool = mempool_create_slab_pool(pool_size, bip_slab);
617    if (!bs->bio_integrity_pool)
618        return -1;
619
620    bs->bvec_integrity_pool = biovec_create_pool(pool_size);
621    if (!bs->bvec_integrity_pool) {
622        mempool_destroy(bs->bio_integrity_pool);
623        return -1;
624    }
625
626    return 0;
627}
628EXPORT_SYMBOL(bioset_integrity_create);
629
630void bioset_integrity_free(struct bio_set *bs)
631{
632    if (bs->bio_integrity_pool)
633        mempool_destroy(bs->bio_integrity_pool);
634
635    if (bs->bvec_integrity_pool)
636        mempool_destroy(bs->bvec_integrity_pool);
637}
638EXPORT_SYMBOL(bioset_integrity_free);
639
640void __init bio_integrity_init(void)
641{
642    /*
643     * kintegrityd won't block much but may burn a lot of CPU cycles.
644     * Make it highpri CPU intensive wq with max concurrency of 1.
645     */
646    kintegrityd_wq = alloc_workqueue("kintegrityd", WQ_MEM_RECLAIM |
647                     WQ_HIGHPRI | WQ_CPU_INTENSIVE, 1);
648    if (!kintegrityd_wq)
649        panic("Failed to create kintegrityd\n");
650
651    bip_slab = kmem_cache_create("bio_integrity_payload",
652                     sizeof(struct bio_integrity_payload) +
653                     sizeof(struct bio_vec) * BIP_INLINE_VECS,
654                     0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
655    if (!bip_slab)
656        panic("Failed to create slab\n");
657}
658

Archive Download this file



interactive