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 || !bs->bio_integrity_pool) { 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 bio->bi_rw |= REQ_INTEGRITY; 83 84 return bip; 85 err: 86 mempool_free(bip, bs->bio_integrity_pool); 87 return NULL; 88 } 89 EXPORT_SYMBOL(bio_integrity_alloc); 90 91 /** 92 * bio_integrity_free - Free bio integrity payload 93 * @bio: bio containing bip to be freed 94 * 95 * Description: Used to free the integrity portion of a bio. Usually 96 * called from bio_free(). 97 */ 98 void bio_integrity_free(struct bio *bio) 99 { 100 struct bio_integrity_payload *bip = bio_integrity(bio); 101 struct bio_set *bs = bio->bi_pool; 102 103 if (bip->bip_flags & BIP_BLOCK_INTEGRITY) 104 kfree(page_address(bip->bip_vec->bv_page) + 105 bip->bip_vec->bv_offset); 106 107 if (bs && bs->bio_integrity_pool) { 108 if (bip->bip_slab != BIO_POOL_NONE) 109 bvec_free(bs->bvec_integrity_pool, bip->bip_vec, 110 bip->bip_slab); 111 112 mempool_free(bip, bs->bio_integrity_pool); 113 } else { 114 kfree(bip); 115 } 116 117 bio->bi_integrity = NULL; 118 } 119 EXPORT_SYMBOL(bio_integrity_free); 120 121 /** 122 * bio_integrity_add_page - Attach integrity metadata 123 * @bio: bio to update 124 * @page: page containing integrity metadata 125 * @len: number of bytes of integrity metadata in page 126 * @offset: start offset within page 127 * 128 * Description: Attach a page containing integrity metadata to bio. 129 */ 130 int bio_integrity_add_page(struct bio *bio, struct page *page, 131 unsigned int len, unsigned int offset) 132 { 133 struct bio_integrity_payload *bip = bio_integrity(bio); 134 struct bio_vec *iv; 135 136 if (bip->bip_vcnt >= bip->bip_max_vcnt) { 137 printk(KERN_ERR "%s: bip_vec full\n", __func__); 138 return 0; 139 } 140 141 iv = bip->bip_vec + bip->bip_vcnt; 142 143 if (bip->bip_vcnt && 144 bvec_gap_to_prev(bdev_get_queue(bio->bi_bdev), 145 &bip->bip_vec[bip->bip_vcnt - 1], offset)) 146 return 0; 147 148 iv->bv_page = page; 149 iv->bv_len = len; 150 iv->bv_offset = offset; 151 bip->bip_vcnt++; 152 153 return len; 154 } 155 EXPORT_SYMBOL(bio_integrity_add_page); 156 157 /** 158 * bio_integrity_enabled - Check whether integrity can be passed 159 * @bio: bio to check 160 * 161 * Description: Determines whether bio_integrity_prep() can be called 162 * on this bio or not. bio data direction and target device must be 163 * set prior to calling. The functions honors the write_generate and 164 * read_verify flags in sysfs. 165 */ 166 bool bio_integrity_enabled(struct bio *bio) 167 { 168 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev); 169 170 if (!bio_is_rw(bio)) 171 return false; 172 173 /* Already protected? */ 174 if (bio_integrity(bio)) 175 return false; 176 177 if (bi == NULL) 178 return false; 179 180 if (bio_data_dir(bio) == READ && bi->verify_fn != NULL && 181 (bi->flags & BLK_INTEGRITY_VERIFY)) 182 return true; 183 184 if (bio_data_dir(bio) == WRITE && bi->generate_fn != NULL && 185 (bi->flags & BLK_INTEGRITY_GENERATE)) 186 return true; 187 188 return false; 189 } 190 EXPORT_SYMBOL(bio_integrity_enabled); 191 192 /** 193 * bio_integrity_intervals - Return number of integrity intervals for a bio 194 * @bi: blk_integrity profile for device 195 * @sectors: Size of the bio in 512-byte sectors 196 * 197 * Description: The block layer calculates everything in 512 byte 198 * sectors but integrity metadata is done in terms of the data integrity 199 * interval size of the storage device. Convert the block layer sectors 200 * to the appropriate number of integrity intervals. 201 */ 202 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi, 203 unsigned int sectors) 204 { 205 return sectors >> (ilog2(bi->interval) - 9); 206 } 207 208 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi, 209 unsigned int sectors) 210 { 211 return bio_integrity_intervals(bi, sectors) * bi->tuple_size; 212 } 213 214 /** 215 * bio_integrity_process - Process integrity metadata for a bio 216 * @bio: bio to generate/verify integrity metadata for 217 * @proc_fn: Pointer to the relevant processing function 218 */ 219 static int bio_integrity_process(struct bio *bio, 220 integrity_processing_fn *proc_fn) 221 { 222 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev); 223 struct blk_integrity_iter iter; 224 struct bvec_iter bviter; 225 struct bio_vec bv; 226 struct bio_integrity_payload *bip = bio_integrity(bio); 227 unsigned int ret = 0; 228 void *prot_buf = page_address(bip->bip_vec->bv_page) + 229 bip->bip_vec->bv_offset; 230 231 iter.disk_name = bio->bi_bdev->bd_disk->disk_name; 232 iter.interval = bi->interval; 233 iter.seed = bip_get_seed(bip); 234 iter.prot_buf = prot_buf; 235 236 bio_for_each_segment(bv, bio, bviter) { 237 void *kaddr = kmap_atomic(bv.bv_page); 238 239 iter.data_buf = kaddr + bv.bv_offset; 240 iter.data_size = bv.bv_len; 241 242 ret = proc_fn(&iter); 243 if (ret) { 244 kunmap_atomic(kaddr); 245 return ret; 246 } 247 248 kunmap_atomic(kaddr); 249 } 250 return ret; 251 } 252 253 /** 254 * bio_integrity_prep - Prepare bio for integrity I/O 255 * @bio: bio to prepare 256 * 257 * Description: Allocates a buffer for integrity metadata, maps the 258 * pages and attaches them to a bio. The bio must have data 259 * direction, target device and start sector set priot to calling. In 260 * the WRITE case, integrity metadata will be generated using the 261 * block device's integrity function. In the READ case, the buffer 262 * will be prepared for DMA and a suitable end_io handler set up. 263 */ 264 int bio_integrity_prep(struct bio *bio) 265 { 266 struct bio_integrity_payload *bip; 267 struct blk_integrity *bi; 268 struct request_queue *q; 269 void *buf; 270 unsigned long start, end; 271 unsigned int len, nr_pages; 272 unsigned int bytes, offset, i; 273 unsigned int intervals; 274 275 bi = bdev_get_integrity(bio->bi_bdev); 276 q = bdev_get_queue(bio->bi_bdev); 277 BUG_ON(bi == NULL); 278 BUG_ON(bio_integrity(bio)); 279 280 intervals = bio_integrity_intervals(bi, bio_sectors(bio)); 281 282 /* Allocate kernel buffer for protection data */ 283 len = intervals * bi->tuple_size; 284 buf = kmalloc(len, GFP_NOIO | q->bounce_gfp); 285 if (unlikely(buf == NULL)) { 286 printk(KERN_ERR "could not allocate integrity buffer\n"); 287 return -ENOMEM; 288 } 289 290 end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT; 291 start = ((unsigned long) buf) >> PAGE_SHIFT; 292 nr_pages = end - start; 293 294 /* Allocate bio integrity payload and integrity vectors */ 295 bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages); 296 if (unlikely(bip == NULL)) { 297 printk(KERN_ERR "could not allocate data integrity bioset\n"); 298 kfree(buf); 299 return -EIO; 300 } 301 302 bip->bip_flags |= BIP_BLOCK_INTEGRITY; 303 bip->bip_iter.bi_size = len; 304 bip_set_seed(bip, bio->bi_iter.bi_sector); 305 306 if (bi->flags & BLK_INTEGRITY_IP_CHECKSUM) 307 bip->bip_flags |= BIP_IP_CHECKSUM; 308 309 /* Map it */ 310 offset = offset_in_page(buf); 311 for (i = 0 ; i < nr_pages ; i++) { 312 int ret; 313 bytes = PAGE_SIZE - offset; 314 315 if (len <= 0) 316 break; 317 318 if (bytes > len) 319 bytes = len; 320 321 ret = bio_integrity_add_page(bio, virt_to_page(buf), 322 bytes, offset); 323 324 if (ret == 0) 325 return 0; 326 327 if (ret < bytes) 328 break; 329 330 buf += bytes; 331 len -= bytes; 332 offset = 0; 333 } 334 335 /* Install custom I/O completion handler if read verify is enabled */ 336 if (bio_data_dir(bio) == READ) { 337 bip->bip_end_io = bio->bi_end_io; 338 bio->bi_end_io = bio_integrity_endio; 339 } 340 341 /* Auto-generate integrity metadata if this is a write */ 342 if (bio_data_dir(bio) == WRITE) 343 bio_integrity_process(bio, bi->generate_fn); 344 345 return 0; 346 } 347 EXPORT_SYMBOL(bio_integrity_prep); 348 349 /** 350 * bio_integrity_verify_fn - Integrity I/O completion worker 351 * @work: Work struct stored in bio to be verified 352 * 353 * Description: This workqueue function is called to complete a READ 354 * request. The function verifies the transferred integrity metadata 355 * and then calls the original bio end_io function. 356 */ 357 static void bio_integrity_verify_fn(struct work_struct *work) 358 { 359 struct bio_integrity_payload *bip = 360 container_of(work, struct bio_integrity_payload, bip_work); 361 struct bio *bio = bip->bip_bio; 362 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev); 363 364 bio->bi_error = bio_integrity_process(bio, bi->verify_fn); 365 366 /* Restore original bio completion handler */ 367 bio->bi_end_io = bip->bip_end_io; 368 bio_endio(bio); 369 } 370 371 /** 372 * bio_integrity_endio - Integrity I/O completion function 373 * @bio: Protected bio 374 * @error: Pointer to errno 375 * 376 * Description: Completion for integrity I/O 377 * 378 * Normally I/O completion is done in interrupt context. However, 379 * verifying I/O integrity is a time-consuming task which must be run 380 * in process context. This function postpones completion 381 * accordingly. 382 */ 383 void bio_integrity_endio(struct bio *bio) 384 { 385 struct bio_integrity_payload *bip = bio_integrity(bio); 386 387 BUG_ON(bip->bip_bio != bio); 388 389 /* In case of an I/O error there is no point in verifying the 390 * integrity metadata. Restore original bio end_io handler 391 * and run it. 392 */ 393 if (bio->bi_error) { 394 bio->bi_end_io = bip->bip_end_io; 395 bio_endio(bio); 396 397 return; 398 } 399 400 INIT_WORK(&bip->bip_work, bio_integrity_verify_fn); 401 queue_work(kintegrityd_wq, &bip->bip_work); 402 } 403 EXPORT_SYMBOL(bio_integrity_endio); 404 405 /** 406 * bio_integrity_advance - Advance integrity vector 407 * @bio: bio whose integrity vector to update 408 * @bytes_done: number of data bytes that have been completed 409 * 410 * Description: This function calculates how many integrity bytes the 411 * number of completed data bytes correspond to and advances the 412 * integrity vector accordingly. 413 */ 414 void bio_integrity_advance(struct bio *bio, unsigned int bytes_done) 415 { 416 struct bio_integrity_payload *bip = bio_integrity(bio); 417 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev); 418 unsigned bytes = bio_integrity_bytes(bi, bytes_done >> 9); 419 420 bvec_iter_advance(bip->bip_vec, &bip->bip_iter, bytes); 421 } 422 EXPORT_SYMBOL(bio_integrity_advance); 423 424 /** 425 * bio_integrity_trim - Trim integrity vector 426 * @bio: bio whose integrity vector to update 427 * @offset: offset to first data sector 428 * @sectors: number of data sectors 429 * 430 * Description: Used to trim the integrity vector in a cloned bio. 431 * The ivec will be advanced corresponding to 'offset' data sectors 432 * and the length will be truncated corresponding to 'len' data 433 * sectors. 434 */ 435 void bio_integrity_trim(struct bio *bio, unsigned int offset, 436 unsigned int sectors) 437 { 438 struct bio_integrity_payload *bip = bio_integrity(bio); 439 struct blk_integrity *bi = bdev_get_integrity(bio->bi_bdev); 440 441 bio_integrity_advance(bio, offset << 9); 442 bip->bip_iter.bi_size = bio_integrity_bytes(bi, sectors); 443 } 444 EXPORT_SYMBOL(bio_integrity_trim); 445 446 /** 447 * bio_integrity_clone - Callback for cloning bios with integrity metadata 448 * @bio: New bio 449 * @bio_src: Original bio 450 * @gfp_mask: Memory allocation mask 451 * 452 * Description: Called to allocate a bip when cloning a bio 453 */ 454 int bio_integrity_clone(struct bio *bio, struct bio *bio_src, 455 gfp_t gfp_mask) 456 { 457 struct bio_integrity_payload *bip_src = bio_integrity(bio_src); 458 struct bio_integrity_payload *bip; 459 460 BUG_ON(bip_src == NULL); 461 462 bip = bio_integrity_alloc(bio, gfp_mask, bip_src->bip_vcnt); 463 464 if (bip == NULL) 465 return -EIO; 466 467 memcpy(bip->bip_vec, bip_src->bip_vec, 468 bip_src->bip_vcnt * sizeof(struct bio_vec)); 469 470 bip->bip_vcnt = bip_src->bip_vcnt; 471 bip->bip_iter = bip_src->bip_iter; 472 473 return 0; 474 } 475 EXPORT_SYMBOL(bio_integrity_clone); 476 477 int bioset_integrity_create(struct bio_set *bs, int pool_size) 478 { 479 if (bs->bio_integrity_pool) 480 return 0; 481 482 bs->bio_integrity_pool = mempool_create_slab_pool(pool_size, bip_slab); 483 if (!bs->bio_integrity_pool) 484 return -1; 485 486 bs->bvec_integrity_pool = biovec_create_pool(pool_size); 487 if (!bs->bvec_integrity_pool) { 488 mempool_destroy(bs->bio_integrity_pool); 489 return -1; 490 } 491 492 return 0; 493 } 494 EXPORT_SYMBOL(bioset_integrity_create); 495 496 void bioset_integrity_free(struct bio_set *bs) 497 { 498 if (bs->bio_integrity_pool) 499 mempool_destroy(bs->bio_integrity_pool); 500 501 if (bs->bvec_integrity_pool) 502 mempool_destroy(bs->bvec_integrity_pool); 503 } 504 EXPORT_SYMBOL(bioset_integrity_free); 505 506 void __init bio_integrity_init(void) 507 { 508 /* 509 * kintegrityd won't block much but may burn a lot of CPU cycles. 510 * Make it highpri CPU intensive wq with max concurrency of 1. 511 */ 512 kintegrityd_wq = alloc_workqueue("kintegrityd", WQ_MEM_RECLAIM | 513 WQ_HIGHPRI | WQ_CPU_INTENSIVE, 1); 514 if (!kintegrityd_wq) 515 panic("Failed to create kintegrityd\n"); 516 517 bip_slab = kmem_cache_create("bio_integrity_payload", 518 sizeof(struct bio_integrity_payload) + 519 sizeof(struct bio_vec) * BIP_INLINE_VECS, 520 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); 521 } 522