1 /* BlockDriver implementation for "raw" format driver 2 * 3 * Copyright (C) 2010-2016 Red Hat, Inc. 4 * Copyright (C) 2010, Blue Swirl <blauwirbel@gmail.com> 5 * Copyright (C) 2009, Anthony Liguori <aliguori@us.ibm.com> 6 * 7 * Author: 8 * Laszlo Ersek <lersek@redhat.com> 9 * 10 * Permission is hereby granted, free of charge, to any person obtaining a copy 11 * of this software and associated documentation files (the "Software"), to 12 * deal in the Software without restriction, including without limitation the 13 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or 14 * sell copies of the Software, and to permit persons to whom the Software is 15 * furnished to do so, subject to the following conditions: 16 * 17 * The above copyright notice and this permission notice shall be included in 18 * all copies or substantial portions of the Software. 19 * 20 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 21 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 22 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 23 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 24 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 25 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS 26 * IN THE SOFTWARE. 27 */ 28 29 #include "qemu/osdep.h" 30 #include "block/block_int.h" 31 #include "qapi/error.h" 32 #include "qemu/module.h" 33 #include "qemu/option.h" 34 35 typedef struct BDRVRawState { 36 uint64_t offset; 37 uint64_t size; 38 bool has_size; 39 } BDRVRawState; 40 41 static const char *const mutable_opts[] = { "offset", "size", NULL }; 42 43 static QemuOptsList raw_runtime_opts = { 44 .name = "raw", 45 .head = QTAILQ_HEAD_INITIALIZER(raw_runtime_opts.head), 46 .desc = { 47 { 48 .name = "offset", 49 .type = QEMU_OPT_SIZE, 50 .help = "offset in the disk where the image starts", 51 }, 52 { 53 .name = "size", 54 .type = QEMU_OPT_SIZE, 55 .help = "virtual disk size", 56 }, 57 { /* end of list */ } 58 }, 59 }; 60 61 static QemuOptsList raw_create_opts = { 62 .name = "raw-create-opts", 63 .head = QTAILQ_HEAD_INITIALIZER(raw_create_opts.head), 64 .desc = { 65 { 66 .name = BLOCK_OPT_SIZE, 67 .type = QEMU_OPT_SIZE, 68 .help = "Virtual disk size" 69 }, 70 { /* end of list */ } 71 } 72 }; 73 74 static int raw_read_options(QDict *options, BlockDriverState *bs, 75 BDRVRawState *s, Error **errp) 76 { 77 Error *local_err = NULL; 78 QemuOpts *opts = NULL; 79 int64_t real_size = 0; 80 int ret; 81 82 real_size = bdrv_getlength(bs->file->bs); 83 if (real_size < 0) { 84 error_setg_errno(errp, -real_size, "Could not get image size"); 85 return real_size; 86 } 87 88 opts = qemu_opts_create(&raw_runtime_opts, NULL, 0, &error_abort); 89 qemu_opts_absorb_qdict(opts, options, &local_err); 90 if (local_err) { 91 error_propagate(errp, local_err); 92 ret = -EINVAL; 93 goto end; 94 } 95 96 s->offset = qemu_opt_get_size(opts, "offset", 0); 97 if (s->offset > real_size) { 98 error_setg(errp, "Offset (%" PRIu64 ") cannot be greater than " 99 "size of the containing file (%" PRId64 ")", 100 s->offset, real_size); 101 ret = -EINVAL; 102 goto end; 103 } 104 105 if (qemu_opt_find(opts, "size") != NULL) { 106 s->size = qemu_opt_get_size(opts, "size", 0); 107 s->has_size = true; 108 } else { 109 s->has_size = false; 110 s->size = real_size - s->offset; 111 } 112 113 /* Check size and offset */ 114 if ((real_size - s->offset) < s->size) { 115 error_setg(errp, "The sum of offset (%" PRIu64 ") and size " 116 "(%" PRIu64 ") has to be smaller or equal to the " 117 " actual size of the containing file (%" PRId64 ")", 118 s->offset, s->size, real_size); 119 ret = -EINVAL; 120 goto end; 121 } 122 123 /* Make sure size is multiple of BDRV_SECTOR_SIZE to prevent rounding 124 * up and leaking out of the specified area. */ 125 if (s->has_size && !QEMU_IS_ALIGNED(s->size, BDRV_SECTOR_SIZE)) { 126 error_setg(errp, "Specified size is not multiple of %llu", 127 BDRV_SECTOR_SIZE); 128 ret = -EINVAL; 129 goto end; 130 } 131 132 ret = 0; 133 134 end: 135 136 qemu_opts_del(opts); 137 138 return ret; 139 } 140 141 static int raw_reopen_prepare(BDRVReopenState *reopen_state, 142 BlockReopenQueue *queue, Error **errp) 143 { 144 assert(reopen_state != NULL); 145 assert(reopen_state->bs != NULL); 146 147 reopen_state->opaque = g_new0(BDRVRawState, 1); 148 149 return raw_read_options( 150 reopen_state->options, 151 reopen_state->bs, 152 reopen_state->opaque, 153 errp); 154 } 155 156 static void raw_reopen_commit(BDRVReopenState *state) 157 { 158 BDRVRawState *new_s = state->opaque; 159 BDRVRawState *s = state->bs->opaque; 160 161 memcpy(s, new_s, sizeof(BDRVRawState)); 162 163 g_free(state->opaque); 164 state->opaque = NULL; 165 } 166 167 static void raw_reopen_abort(BDRVReopenState *state) 168 { 169 g_free(state->opaque); 170 state->opaque = NULL; 171 } 172 173 /* Check and adjust the offset, against 'offset' and 'size' options. */ 174 static inline int raw_adjust_offset(BlockDriverState *bs, uint64_t *offset, 175 uint64_t bytes, bool is_write) 176 { 177 BDRVRawState *s = bs->opaque; 178 179 if (s->has_size && (*offset > s->size || bytes > (s->size - *offset))) { 180 /* There's not enough space for the write, or the read request is 181 * out-of-range. Don't read/write anything to prevent leaking out of 182 * the size specified in options. */ 183 return is_write ? -ENOSPC : -EINVAL; 184 } 185 186 if (*offset > INT64_MAX - s->offset) { 187 return -EINVAL; 188 } 189 *offset += s->offset; 190 191 return 0; 192 } 193 194 static int coroutine_fn raw_co_preadv(BlockDriverState *bs, uint64_t offset, 195 uint64_t bytes, QEMUIOVector *qiov, 196 int flags) 197 { 198 int ret; 199 200 ret = raw_adjust_offset(bs, &offset, bytes, false); 201 if (ret) { 202 return ret; 203 } 204 205 BLKDBG_EVENT(bs->file, BLKDBG_READ_AIO); 206 return bdrv_co_preadv(bs->file, offset, bytes, qiov, flags); 207 } 208 209 static int coroutine_fn raw_co_pwritev(BlockDriverState *bs, uint64_t offset, 210 uint64_t bytes, QEMUIOVector *qiov, 211 int flags) 212 { 213 void *buf = NULL; 214 BlockDriver *drv; 215 QEMUIOVector local_qiov; 216 int ret; 217 218 if (bs->probed && offset < BLOCK_PROBE_BUF_SIZE && bytes) { 219 /* Handling partial writes would be a pain - so we just 220 * require that guests have 512-byte request alignment if 221 * probing occurred */ 222 QEMU_BUILD_BUG_ON(BLOCK_PROBE_BUF_SIZE != 512); 223 QEMU_BUILD_BUG_ON(BDRV_SECTOR_SIZE != 512); 224 assert(offset == 0 && bytes >= BLOCK_PROBE_BUF_SIZE); 225 226 buf = qemu_try_blockalign(bs->file->bs, 512); 227 if (!buf) { 228 ret = -ENOMEM; 229 goto fail; 230 } 231 232 ret = qemu_iovec_to_buf(qiov, 0, buf, 512); 233 if (ret != 512) { 234 ret = -EINVAL; 235 goto fail; 236 } 237 238 drv = bdrv_probe_all(buf, 512, NULL); 239 if (drv != bs->drv) { 240 ret = -EPERM; 241 goto fail; 242 } 243 244 /* Use the checked buffer, a malicious guest might be overwriting its 245 * original buffer in the background. */ 246 qemu_iovec_init(&local_qiov, qiov->niov + 1); 247 qemu_iovec_add(&local_qiov, buf, 512); 248 qemu_iovec_concat(&local_qiov, qiov, 512, qiov->size - 512); 249 qiov = &local_qiov; 250 } 251 252 ret = raw_adjust_offset(bs, &offset, bytes, true); 253 if (ret) { 254 goto fail; 255 } 256 257 BLKDBG_EVENT(bs->file, BLKDBG_WRITE_AIO); 258 ret = bdrv_co_pwritev(bs->file, offset, bytes, qiov, flags); 259 260 fail: 261 if (qiov == &local_qiov) { 262 qemu_iovec_destroy(&local_qiov); 263 } 264 qemu_vfree(buf); 265 return ret; 266 } 267 268 static int coroutine_fn raw_co_block_status(BlockDriverState *bs, 269 bool want_zero, int64_t offset, 270 int64_t bytes, int64_t *pnum, 271 int64_t *map, 272 BlockDriverState **file) 273 { 274 BDRVRawState *s = bs->opaque; 275 *pnum = bytes; 276 *file = bs->file->bs; 277 *map = offset + s->offset; 278 return BDRV_BLOCK_RAW | BDRV_BLOCK_OFFSET_VALID; 279 } 280 281 static int coroutine_fn raw_co_pwrite_zeroes(BlockDriverState *bs, 282 int64_t offset, int bytes, 283 BdrvRequestFlags flags) 284 { 285 int ret; 286 287 ret = raw_adjust_offset(bs, (uint64_t *)&offset, bytes, true); 288 if (ret) { 289 return ret; 290 } 291 return bdrv_co_pwrite_zeroes(bs->file, offset, bytes, flags); 292 } 293 294 static int coroutine_fn raw_co_pdiscard(BlockDriverState *bs, 295 int64_t offset, int bytes) 296 { 297 int ret; 298 299 ret = raw_adjust_offset(bs, (uint64_t *)&offset, bytes, true); 300 if (ret) { 301 return ret; 302 } 303 return bdrv_co_pdiscard(bs->file, offset, bytes); 304 } 305 306 static int64_t raw_getlength(BlockDriverState *bs) 307 { 308 int64_t len; 309 BDRVRawState *s = bs->opaque; 310 311 /* Update size. It should not change unless the file was externally 312 * modified. */ 313 len = bdrv_getlength(bs->file->bs); 314 if (len < 0) { 315 return len; 316 } 317 318 if (len < s->offset) { 319 s->size = 0; 320 } else { 321 if (s->has_size) { 322 /* Try to honour the size */ 323 s->size = MIN(s->size, len - s->offset); 324 } else { 325 s->size = len - s->offset; 326 } 327 } 328 329 return s->size; 330 } 331 332 static BlockMeasureInfo *raw_measure(QemuOpts *opts, BlockDriverState *in_bs, 333 Error **errp) 334 { 335 BlockMeasureInfo *info; 336 int64_t required; 337 338 if (in_bs) { 339 required = bdrv_getlength(in_bs); 340 if (required < 0) { 341 error_setg_errno(errp, -required, "Unable to get image size"); 342 return NULL; 343 } 344 } else { 345 required = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), 346 BDRV_SECTOR_SIZE); 347 } 348 349 info = g_new(BlockMeasureInfo, 1); 350 info->required = required; 351 352 /* Unallocated sectors count towards the file size in raw images */ 353 info->fully_allocated = info->required; 354 return info; 355 } 356 357 static int raw_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) 358 { 359 return bdrv_get_info(bs->file->bs, bdi); 360 } 361 362 static void raw_refresh_limits(BlockDriverState *bs, Error **errp) 363 { 364 if (bs->probed) { 365 /* To make it easier to protect the first sector, any probed 366 * image is restricted to read-modify-write on sub-sector 367 * operations. */ 368 bs->bl.request_alignment = BDRV_SECTOR_SIZE; 369 } 370 } 371 372 static int coroutine_fn raw_co_truncate(BlockDriverState *bs, int64_t offset, 373 bool exact, PreallocMode prealloc, 374 BdrvRequestFlags flags, Error **errp) 375 { 376 BDRVRawState *s = bs->opaque; 377 378 if (s->has_size) { 379 error_setg(errp, "Cannot resize fixed-size raw disks"); 380 return -ENOTSUP; 381 } 382 383 if (INT64_MAX - offset < s->offset) { 384 error_setg(errp, "Disk size too large for the chosen offset"); 385 return -EINVAL; 386 } 387 388 s->size = offset; 389 offset += s->offset; 390 return bdrv_co_truncate(bs->file, offset, exact, prealloc, flags, errp); 391 } 392 393 static void raw_eject(BlockDriverState *bs, bool eject_flag) 394 { 395 bdrv_eject(bs->file->bs, eject_flag); 396 } 397 398 static void raw_lock_medium(BlockDriverState *bs, bool locked) 399 { 400 bdrv_lock_medium(bs->file->bs, locked); 401 } 402 403 static int raw_co_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) 404 { 405 BDRVRawState *s = bs->opaque; 406 if (s->offset || s->has_size) { 407 return -ENOTSUP; 408 } 409 return bdrv_co_ioctl(bs->file->bs, req, buf); 410 } 411 412 static int raw_has_zero_init(BlockDriverState *bs) 413 { 414 return bdrv_has_zero_init(bs->file->bs); 415 } 416 417 static int coroutine_fn raw_co_create_opts(BlockDriver *drv, 418 const char *filename, 419 QemuOpts *opts, 420 Error **errp) 421 { 422 return bdrv_create_file(filename, opts, errp); 423 } 424 425 static int raw_open(BlockDriverState *bs, QDict *options, int flags, 426 Error **errp) 427 { 428 BDRVRawState *s = bs->opaque; 429 int ret; 430 431 bs->file = bdrv_open_child(NULL, options, "file", bs, &child_file, 432 false, errp); 433 if (!bs->file) { 434 return -EINVAL; 435 } 436 437 bs->sg = bs->file->bs->sg; 438 bs->supported_write_flags = BDRV_REQ_WRITE_UNCHANGED | 439 (BDRV_REQ_FUA & bs->file->bs->supported_write_flags); 440 bs->supported_zero_flags = BDRV_REQ_WRITE_UNCHANGED | 441 ((BDRV_REQ_FUA | BDRV_REQ_MAY_UNMAP | BDRV_REQ_NO_FALLBACK) & 442 bs->file->bs->supported_zero_flags); 443 bs->supported_truncate_flags = bs->file->bs->supported_truncate_flags & 444 BDRV_REQ_ZERO_WRITE; 445 446 if (bs->probed && !bdrv_is_read_only(bs)) { 447 bdrv_refresh_filename(bs->file->bs); 448 fprintf(stderr, 449 "WARNING: Image format was not specified for '%s' and probing " 450 "guessed raw.\n" 451 " Automatically detecting the format is dangerous for " 452 "raw images, write operations on block 0 will be restricted.\n" 453 " Specify the 'raw' format explicitly to remove the " 454 "restrictions.\n", 455 bs->file->bs->filename); 456 } 457 458 ret = raw_read_options(options, bs, s, errp); 459 if (ret < 0) { 460 return ret; 461 } 462 463 if (bs->sg && (s->offset || s->has_size)) { 464 error_setg(errp, "Cannot use offset/size with SCSI generic devices"); 465 return -EINVAL; 466 } 467 468 return 0; 469 } 470 471 static int raw_probe(const uint8_t *buf, int buf_size, const char *filename) 472 { 473 /* smallest possible positive score so that raw is used if and only if no 474 * other block driver works 475 */ 476 return 1; 477 } 478 479 static int raw_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz) 480 { 481 BDRVRawState *s = bs->opaque; 482 int ret; 483 484 ret = bdrv_probe_blocksizes(bs->file->bs, bsz); 485 if (ret < 0) { 486 return ret; 487 } 488 489 if (!QEMU_IS_ALIGNED(s->offset, MAX(bsz->log, bsz->phys))) { 490 return -ENOTSUP; 491 } 492 493 return 0; 494 } 495 496 static int raw_probe_geometry(BlockDriverState *bs, HDGeometry *geo) 497 { 498 BDRVRawState *s = bs->opaque; 499 if (s->offset || s->has_size) { 500 return -ENOTSUP; 501 } 502 return bdrv_probe_geometry(bs->file->bs, geo); 503 } 504 505 static int coroutine_fn raw_co_copy_range_from(BlockDriverState *bs, 506 BdrvChild *src, 507 uint64_t src_offset, 508 BdrvChild *dst, 509 uint64_t dst_offset, 510 uint64_t bytes, 511 BdrvRequestFlags read_flags, 512 BdrvRequestFlags write_flags) 513 { 514 int ret; 515 516 ret = raw_adjust_offset(bs, &src_offset, bytes, false); 517 if (ret) { 518 return ret; 519 } 520 return bdrv_co_copy_range_from(bs->file, src_offset, dst, dst_offset, 521 bytes, read_flags, write_flags); 522 } 523 524 static int coroutine_fn raw_co_copy_range_to(BlockDriverState *bs, 525 BdrvChild *src, 526 uint64_t src_offset, 527 BdrvChild *dst, 528 uint64_t dst_offset, 529 uint64_t bytes, 530 BdrvRequestFlags read_flags, 531 BdrvRequestFlags write_flags) 532 { 533 int ret; 534 535 ret = raw_adjust_offset(bs, &dst_offset, bytes, true); 536 if (ret) { 537 return ret; 538 } 539 return bdrv_co_copy_range_to(src, src_offset, bs->file, dst_offset, bytes, 540 read_flags, write_flags); 541 } 542 543 static const char *const raw_strong_runtime_opts[] = { 544 "offset", 545 "size", 546 547 NULL 548 }; 549 550 BlockDriver bdrv_raw = { 551 .format_name = "raw", 552 .instance_size = sizeof(BDRVRawState), 553 .bdrv_probe = &raw_probe, 554 .bdrv_reopen_prepare = &raw_reopen_prepare, 555 .bdrv_reopen_commit = &raw_reopen_commit, 556 .bdrv_reopen_abort = &raw_reopen_abort, 557 .bdrv_open = &raw_open, 558 .bdrv_child_perm = bdrv_filter_default_perms, 559 .bdrv_co_create_opts = &raw_co_create_opts, 560 .bdrv_co_preadv = &raw_co_preadv, 561 .bdrv_co_pwritev = &raw_co_pwritev, 562 .bdrv_co_pwrite_zeroes = &raw_co_pwrite_zeroes, 563 .bdrv_co_pdiscard = &raw_co_pdiscard, 564 .bdrv_co_block_status = &raw_co_block_status, 565 .bdrv_co_copy_range_from = &raw_co_copy_range_from, 566 .bdrv_co_copy_range_to = &raw_co_copy_range_to, 567 .bdrv_co_truncate = &raw_co_truncate, 568 .bdrv_getlength = &raw_getlength, 569 .has_variable_length = true, 570 .bdrv_measure = &raw_measure, 571 .bdrv_get_info = &raw_get_info, 572 .bdrv_refresh_limits = &raw_refresh_limits, 573 .bdrv_probe_blocksizes = &raw_probe_blocksizes, 574 .bdrv_probe_geometry = &raw_probe_geometry, 575 .bdrv_eject = &raw_eject, 576 .bdrv_lock_medium = &raw_lock_medium, 577 .bdrv_co_ioctl = &raw_co_ioctl, 578 .create_opts = &raw_create_opts, 579 .bdrv_has_zero_init = &raw_has_zero_init, 580 .strong_runtime_opts = raw_strong_runtime_opts, 581 .mutable_opts = mutable_opts, 582 }; 583 584 static void bdrv_raw_init(void) 585 { 586 bdrv_register(&bdrv_raw); 587 } 588 589 block_init(bdrv_raw_init); 590