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 32 static struct kmem_cache *bip_slab; 33 static 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 */ 45 struct 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 bip->bip_max_vcnt = bvec_nr_vecs(idx); 74 } else { 75 bip->bip_vec = bip->bip_inline_vecs; 76 bip->bip_max_vcnt = inline_vecs; 77 } 78 79 bip->bip_slab = idx; 80 bip->bip_bio = bio; 81 bio->bi_integrity = bip; 82 83 return bip; 84 err: 85 mempool_free(bip, bs->bio_integrity_pool); 86 return NULL; 87 } 88 EXPORT_SYMBOL(bio_integrity_alloc); 89 90 /** 91 * bio_integrity_free - Free bio integrity payload 92 * @bio: bio containing bip to be freed 93 * 94 * Description: Used to free the integrity portion of a bio. Usually 95 * called from bio_free(). 96 */ 97 void bio_integrity_free(struct bio *bio) 98 { 99 struct bio_integrity_payload *bip = bio->bi_integrity; 100 struct bio_set *bs = bio->bi_pool; 101 102 if (bip->bip_owns_buf) 103 kfree(bip->bip_buf); 104 105 if (bs) { 106 if (bip->bip_slab != BIO_POOL_NONE) 107 bvec_free(bs->bvec_integrity_pool, bip->bip_vec, 108 bip->bip_slab); 109 110 mempool_free(bip, bs->bio_integrity_pool); 111 } else { 112 kfree(bip); 113 } 114 115 bio->bi_integrity = NULL; 116 } 117 EXPORT_SYMBOL(bio_integrity_free); 118 119 /** 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 125 * 126 * Description: Attach a page containing integrity metadata to bio. 127 */ 128 int bio_integrity_add_page(struct bio *bio, struct page *page, 129 unsigned int len, unsigned int offset) 130 { 131 struct bio_integrity_payload *bip = bio->bi_integrity; 132 struct bio_vec *iv; 133 134 if (bip->bip_vcnt >= bip->bip_max_vcnt) { 135 printk(KERN_ERR "%s: bip_vec full\n", __func__); 136 return 0; 137 } 138 139 iv = bip->bip_vec + bip->bip_vcnt; 140 141 iv->bv_page = page; 142 iv->bv_len = len; 143 iv->bv_offset = offset; 144 bip->bip_vcnt++; 145 146 return len; 147 } 148 EXPORT_SYMBOL(bio_integrity_add_page); 149 150 static int bdev_integrity_enabled(struct block_device *bdev, int rw) 151 { 152 struct blk_integrity *bi = bdev_get_integrity(bdev); 153 154 if (bi == NULL) 155 return 0; 156 157 if (rw == READ && bi->verify_fn != NULL && 158 (bi->flags & INTEGRITY_FLAG_READ)) 159 return 1; 160 161 if (rw == WRITE && bi->generate_fn != NULL && 162 (bi->flags & INTEGRITY_FLAG_WRITE)) 163 return 1; 164 165 return 0; 166 } 167 168 /** 169 * bio_integrity_enabled - Check whether integrity can be passed 170 * @bio: bio to check 171 * 172 * Description: Determines whether bio_integrity_prep() can be called 173 * on this bio or not. bio data direction and target device must be 174 * set prior to calling. The functions honors the write_generate and 175 * read_verify flags in sysfs. 176 */ 177 int bio_integrity_enabled(struct bio *bio) 178 { 179 if (!bio_is_rw(bio)) 180 return 0; 181 182 /* Already protected? */ 183 if (bio_integrity(bio)) 184 return 0; 185 186 return bdev_integrity_enabled(bio->bi_bdev, bio_data_dir(bio)); 187 } 188 EXPORT_SYMBOL(bio_integrity_enabled); 189 190 /** 191 * bio_integrity_hw_sectors - Convert 512b sectors to hardware ditto 192 * @bi: blk_integrity profile for device 193 * @sectors: Number of 512 sectors to convert 194 * 195 * Description: The block layer calculates everything in 512 byte 196 * sectors but integrity metadata is done in terms of the hardware 197 * sector size of the storage device. Convert the block layer sectors 198 * to physical sectors. 199 */ 200 static inline unsigned int bio_integrity_hw_sectors(struct blk_integrity *bi, 201 unsigned int sectors) 202 { 203 /* At this point there are only 512b or 4096b DIF/EPP devices */ 204 if (bi->sector_size == 4096) 205 return sectors >>= 3; 206 207 return sectors; 208 } 209 210 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi, 211 unsigned int sectors) 212 { 213 return bio_integrity_hw_sectors(bi, sectors) * bi->tuple_size; 214 } 215 216 /** 217 * bio_integrity_tag_size - Retrieve integrity tag space 218 * @bio: bio to inspect 219 * 220 * Description: Returns the maximum number of tag bytes that can be 221 * attached to this bio. Filesystems can use this to determine how 222 * much metadata to attach to an I/O. 223 */ 224 unsigned int bio_integrity_tag_size(struct bio *bio) 225 { 226 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev); 227 228 BUG_ON(bio->bi_iter.bi_size == 0); 229 230 return bi->tag_size * (bio->bi_iter.bi_size / bi->sector_size); 231 } 232 EXPORT_SYMBOL(bio_integrity_tag_size); 233 234 static int bio_integrity_tag(struct bio *bio, void *tag_buf, unsigned int len, 235 int set) 236 { 237 struct bio_integrity_payload *bip = bio->bi_integrity; 238 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev); 239 unsigned int nr_sectors; 240 241 BUG_ON(bip->bip_buf == NULL); 242 243 if (bi->tag_size == 0) 244 return -1; 245 246 nr_sectors = bio_integrity_hw_sectors(bi, 247 DIV_ROUND_UP(len, bi->tag_size)); 248 249 if (nr_sectors * bi->tuple_size > bip->bip_iter.bi_size) { 250 printk(KERN_ERR "%s: tag too big for bio: %u > %u\n", __func__, 251 nr_sectors * bi->tuple_size, bip->bip_iter.bi_size); 252 return -1; 253 } 254 255 if (set) 256 bi->set_tag_fn(bip->bip_buf, tag_buf, nr_sectors); 257 else 258 bi->get_tag_fn(bip->bip_buf, tag_buf, nr_sectors); 259 260 return 0; 261 } 262 263 /** 264 * bio_integrity_set_tag - Attach a tag buffer to a bio 265 * @bio: bio to attach buffer to 266 * @tag_buf: Pointer to a buffer containing tag data 267 * @len: Length of the included buffer 268 * 269 * Description: Use this function to tag a bio by leveraging the extra 270 * space provided by devices formatted with integrity protection. The 271 * size of the integrity buffer must be <= to the size reported by 272 * bio_integrity_tag_size(). 273 */ 274 int bio_integrity_set_tag(struct bio *bio, void *tag_buf, unsigned int len) 275 { 276 BUG_ON(bio_data_dir(bio) != WRITE); 277 278 return bio_integrity_tag(bio, tag_buf, len, 1); 279 } 280 EXPORT_SYMBOL(bio_integrity_set_tag); 281 282 /** 283 * bio_integrity_get_tag - Retrieve a tag buffer from a bio 284 * @bio: bio to retrieve buffer from 285 * @tag_buf: Pointer to a buffer for the tag data 286 * @len: Length of the target buffer 287 * 288 * Description: Use this function to retrieve the tag buffer from a 289 * completed I/O. The size of the integrity buffer must be <= to the 290 * size reported by bio_integrity_tag_size(). 291 */ 292 int bio_integrity_get_tag(struct bio *bio, void *tag_buf, unsigned int len) 293 { 294 BUG_ON(bio_data_dir(bio) != READ); 295 296 return bio_integrity_tag(bio, tag_buf, len, 0); 297 } 298 EXPORT_SYMBOL(bio_integrity_get_tag); 299 300 /** 301 * bio_integrity_generate_verify - Generate/verify integrity metadata for a bio 302 * @bio: bio to generate/verify integrity metadata for 303 * @operate: operate number, 1 for generate, 0 for verify 304 */ 305 static int bio_integrity_generate_verify(struct bio *bio, int operate) 306 { 307 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev); 308 struct blk_integrity_exchg bix; 309 struct bio_vec *bv; 310 sector_t sector; 311 unsigned int sectors, ret = 0, i; 312 void *prot_buf = bio->bi_integrity->bip_buf; 313 314 if (operate) 315 sector = bio->bi_iter.bi_sector; 316 else 317 sector = bio->bi_integrity->bip_iter.bi_sector; 318 319 bix.disk_name = bio->bi_bdev->bd_disk->disk_name; 320 bix.sector_size = bi->sector_size; 321 322 bio_for_each_segment_all(bv, bio, i) { 323 void *kaddr = kmap_atomic(bv->bv_page); 324 bix.data_buf = kaddr + bv->bv_offset; 325 bix.data_size = bv->bv_len; 326 bix.prot_buf = prot_buf; 327 bix.sector = sector; 328 329 if (operate) 330 bi->generate_fn(&bix); 331 else { 332 ret = bi->verify_fn(&bix); 333 if (ret) { 334 kunmap_atomic(kaddr); 335 return ret; 336 } 337 } 338 339 sectors = bv->bv_len / bi->sector_size; 340 sector += sectors; 341 prot_buf += sectors * bi->tuple_size; 342 343 kunmap_atomic(kaddr); 344 } 345 return ret; 346 } 347 348 /** 349 * bio_integrity_generate - Generate integrity metadata for a bio 350 * @bio: bio to generate integrity metadata for 351 * 352 * Description: Generates integrity metadata for a bio by calling the 353 * block device's generation callback function. The bio must have a 354 * bip attached with enough room to accommodate the generated 355 * integrity metadata. 356 */ 357 static void bio_integrity_generate(struct bio *bio) 358 { 359 bio_integrity_generate_verify(bio, 1); 360 } 361 362 static inline unsigned short blk_integrity_tuple_size(struct blk_integrity *bi) 363 { 364 if (bi) 365 return bi->tuple_size; 366 367 return 0; 368 } 369 370 /** 371 * bio_integrity_prep - Prepare bio for integrity I/O 372 * @bio: bio to prepare 373 * 374 * Description: Allocates a buffer for integrity metadata, maps the 375 * pages and attaches them to a bio. The bio must have data 376 * direction, target device and start sector set priot to calling. In 377 * the WRITE case, integrity metadata will be generated using the 378 * block device's integrity function. In the READ case, the buffer 379 * will be prepared for DMA and a suitable end_io handler set up. 380 */ 381 int bio_integrity_prep(struct bio *bio) 382 { 383 struct bio_integrity_payload *bip; 384 struct blk_integrity *bi; 385 struct request_queue *q; 386 void *buf; 387 unsigned long start, end; 388 unsigned int len, nr_pages; 389 unsigned int bytes, offset, i; 390 unsigned int sectors; 391 392 bi = bdev_get_integrity(bio->bi_bdev); 393 q = bdev_get_queue(bio->bi_bdev); 394 BUG_ON(bi == NULL); 395 BUG_ON(bio_integrity(bio)); 396 397 sectors = bio_integrity_hw_sectors(bi, bio_sectors(bio)); 398 399 /* Allocate kernel buffer for protection data */ 400 len = sectors * blk_integrity_tuple_size(bi); 401 buf = kmalloc(len, GFP_NOIO | q->bounce_gfp); 402 if (unlikely(buf == NULL)) { 403 printk(KERN_ERR "could not allocate integrity buffer\n"); 404 return -ENOMEM; 405 } 406 407 end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT; 408 start = ((unsigned long) buf) >> PAGE_SHIFT; 409 nr_pages = end - start; 410 411 /* Allocate bio integrity payload and integrity vectors */ 412 bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages); 413 if (unlikely(bip == NULL)) { 414 printk(KERN_ERR "could not allocate data integrity bioset\n"); 415 kfree(buf); 416 return -EIO; 417 } 418 419 bip->bip_owns_buf = 1; 420 bip->bip_buf = buf; 421 bip->bip_iter.bi_size = len; 422 bip->bip_iter.bi_sector = bio->bi_iter.bi_sector; 423 424 /* Map it */ 425 offset = offset_in_page(buf); 426 for (i = 0 ; i < nr_pages ; i++) { 427 int ret; 428 bytes = PAGE_SIZE - offset; 429 430 if (len <= 0) 431 break; 432 433 if (bytes > len) 434 bytes = len; 435 436 ret = bio_integrity_add_page(bio, virt_to_page(buf), 437 bytes, offset); 438 439 if (ret == 0) 440 return 0; 441 442 if (ret < bytes) 443 break; 444 445 buf += bytes; 446 len -= bytes; 447 offset = 0; 448 } 449 450 /* Install custom I/O completion handler if read verify is enabled */ 451 if (bio_data_dir(bio) == READ) { 452 bip->bip_end_io = bio->bi_end_io; 453 bio->bi_end_io = bio_integrity_endio; 454 } 455 456 /* Auto-generate integrity metadata if this is a write */ 457 if (bio_data_dir(bio) == WRITE) 458 bio_integrity_generate(bio); 459 460 return 0; 461 } 462 EXPORT_SYMBOL(bio_integrity_prep); 463 464 /** 465 * bio_integrity_verify - Verify integrity metadata for a bio 466 * @bio: bio to verify 467 * 468 * Description: This function is called to verify the integrity of a 469 * bio. The data in the bio io_vec is compared to the integrity 470 * metadata returned by the HBA. 471 */ 472 static int bio_integrity_verify(struct bio *bio) 473 { 474 return bio_integrity_generate_verify(bio, 0); 475 } 476 477 /** 478 * bio_integrity_verify_fn - Integrity I/O completion worker 479 * @work: Work struct stored in bio to be verified 480 * 481 * Description: This workqueue function is called to complete a READ 482 * request. The function verifies the transferred integrity metadata 483 * and then calls the original bio end_io function. 484 */ 485 static void bio_integrity_verify_fn(struct work_struct *work) 486 { 487 struct bio_integrity_payload *bip = 488 container_of(work, struct bio_integrity_payload, bip_work); 489 struct bio *bio = bip->bip_bio; 490 int error; 491 492 error = bio_integrity_verify(bio); 493 494 /* Restore original bio completion handler */ 495 bio->bi_end_io = bip->bip_end_io; 496 bio_endio_nodec(bio, error); 497 } 498 499 /** 500 * bio_integrity_endio - Integrity I/O completion function 501 * @bio: Protected bio 502 * @error: Pointer to errno 503 * 504 * Description: Completion for integrity I/O 505 * 506 * Normally I/O completion is done in interrupt context. However, 507 * verifying I/O integrity is a time-consuming task which must be run 508 * in process context. This function postpones completion 509 * accordingly. 510 */ 511 void bio_integrity_endio(struct bio *bio, int error) 512 { 513 struct bio_integrity_payload *bip = bio->bi_integrity; 514 515 BUG_ON(bip->bip_bio != bio); 516 517 /* In case of an I/O error there is no point in verifying the 518 * integrity metadata. Restore original bio end_io handler 519 * and run it. 520 */ 521 if (error) { 522 bio->bi_end_io = bip->bip_end_io; 523 bio_endio(bio, error); 524 525 return; 526 } 527 528 INIT_WORK(&bip->bip_work, bio_integrity_verify_fn); 529 queue_work(kintegrityd_wq, &bip->bip_work); 530 } 531 EXPORT_SYMBOL(bio_integrity_endio); 532 533 /** 534 * bio_integrity_advance - Advance integrity vector 535 * @bio: bio whose integrity vector to update 536 * @bytes_done: number of data bytes that have been completed 537 * 538 * Description: This function calculates how many integrity bytes the 539 * number of completed data bytes correspond to and advances the 540 * integrity vector accordingly. 541 */ 542 void bio_integrity_advance(struct bio *bio, unsigned int bytes_done) 543 { 544 struct bio_integrity_payload *bip = bio->bi_integrity; 545 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev); 546 unsigned bytes = bio_integrity_bytes(bi, bytes_done >> 9); 547 548 bvec_iter_advance(bip->bip_vec, &bip->bip_iter, bytes); 549 } 550 EXPORT_SYMBOL(bio_integrity_advance); 551 552 /** 553 * bio_integrity_trim - Trim integrity vector 554 * @bio: bio whose integrity vector to update 555 * @offset: offset to first data sector 556 * @sectors: number of data sectors 557 * 558 * Description: Used to trim the integrity vector in a cloned bio. 559 * The ivec will be advanced corresponding to 'offset' data sectors 560 * and the length will be truncated corresponding to 'len' data 561 * sectors. 562 */ 563 void bio_integrity_trim(struct bio *bio, unsigned int offset, 564 unsigned int sectors) 565 { 566 struct bio_integrity_payload *bip = bio->bi_integrity; 567 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev); 568 569 bio_integrity_advance(bio, offset << 9); 570 bip->bip_iter.bi_size = bio_integrity_bytes(bi, sectors); 571 } 572 EXPORT_SYMBOL(bio_integrity_trim); 573 574 /** 575 * bio_integrity_clone - Callback for cloning bios with integrity metadata 576 * @bio: New bio 577 * @bio_src: Original bio 578 * @gfp_mask: Memory allocation mask 579 * 580 * Description: Called to allocate a bip when cloning a bio 581 */ 582 int bio_integrity_clone(struct bio *bio, struct bio *bio_src, 583 gfp_t gfp_mask) 584 { 585 struct bio_integrity_payload *bip_src = bio_src->bi_integrity; 586 struct bio_integrity_payload *bip; 587 588 BUG_ON(bip_src == NULL); 589 590 bip = bio_integrity_alloc(bio, gfp_mask, bip_src->bip_vcnt); 591 592 if (bip == NULL) 593 return -EIO; 594 595 memcpy(bip->bip_vec, bip_src->bip_vec, 596 bip_src->bip_vcnt * sizeof(struct bio_vec)); 597 598 bip->bip_vcnt = bip_src->bip_vcnt; 599 bip->bip_iter = bip_src->bip_iter; 600 601 return 0; 602 } 603 EXPORT_SYMBOL(bio_integrity_clone); 604 605 int bioset_integrity_create(struct bio_set *bs, int pool_size) 606 { 607 if (bs->bio_integrity_pool) 608 return 0; 609 610 bs->bio_integrity_pool = mempool_create_slab_pool(pool_size, bip_slab); 611 if (!bs->bio_integrity_pool) 612 return -1; 613 614 bs->bvec_integrity_pool = biovec_create_pool(pool_size); 615 if (!bs->bvec_integrity_pool) { 616 mempool_destroy(bs->bio_integrity_pool); 617 return -1; 618 } 619 620 return 0; 621 } 622 EXPORT_SYMBOL(bioset_integrity_create); 623 624 void bioset_integrity_free(struct bio_set *bs) 625 { 626 if (bs->bio_integrity_pool) 627 mempool_destroy(bs->bio_integrity_pool); 628 629 if (bs->bvec_integrity_pool) 630 mempool_destroy(bs->bvec_integrity_pool); 631 } 632 EXPORT_SYMBOL(bioset_integrity_free); 633 634 void __init bio_integrity_init(void) 635 { 636 /* 637 * kintegrityd won't block much but may burn a lot of CPU cycles. 638 * Make it highpri CPU intensive wq with max concurrency of 1. 639 */ 640 kintegrityd_wq = alloc_workqueue("kintegrityd", WQ_MEM_RECLAIM | 641 WQ_HIGHPRI | WQ_CPU_INTENSIVE, 1); 642 if (!kintegrityd_wq) 643 panic("Failed to create kintegrityd\n"); 644 645 bip_slab = kmem_cache_create("bio_integrity_payload", 646 sizeof(struct bio_integrity_payload) + 647 sizeof(struct bio_vec) * BIP_INLINE_VECS, 648 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); 649 if (!bip_slab) 650 panic("Failed to create slab\n"); 651 } 652