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/option.h" 33 34 typedef struct BDRVRawState { 35 uint64_t offset; 36 uint64_t size; 37 bool has_size; 38 } BDRVRawState; 39 40 static QemuOptsList raw_runtime_opts = { 41 .name = "raw", 42 .head = QTAILQ_HEAD_INITIALIZER(raw_runtime_opts.head), 43 .desc = { 44 { 45 .name = "offset", 46 .type = QEMU_OPT_SIZE, 47 .help = "offset in the disk where the image starts", 48 }, 49 { 50 .name = "size", 51 .type = QEMU_OPT_SIZE, 52 .help = "virtual disk size", 53 }, 54 { /* end of list */ } 55 }, 56 }; 57 58 static QemuOptsList raw_create_opts = { 59 .name = "raw-create-opts", 60 .head = QTAILQ_HEAD_INITIALIZER(raw_create_opts.head), 61 .desc = { 62 { 63 .name = BLOCK_OPT_SIZE, 64 .type = QEMU_OPT_SIZE, 65 .help = "Virtual disk size" 66 }, 67 { /* end of list */ } 68 } 69 }; 70 71 static int raw_read_options(QDict *options, BlockDriverState *bs, 72 BDRVRawState *s, Error **errp) 73 { 74 Error *local_err = NULL; 75 QemuOpts *opts = NULL; 76 int64_t real_size = 0; 77 int ret; 78 79 real_size = bdrv_getlength(bs->file->bs); 80 if (real_size < 0) { 81 error_setg_errno(errp, -real_size, "Could not get image size"); 82 return real_size; 83 } 84 85 opts = qemu_opts_create(&raw_runtime_opts, NULL, 0, &error_abort); 86 qemu_opts_absorb_qdict(opts, options, &local_err); 87 if (local_err) { 88 error_propagate(errp, local_err); 89 ret = -EINVAL; 90 goto end; 91 } 92 93 s->offset = qemu_opt_get_size(opts, "offset", 0); 94 if (s->offset > real_size) { 95 error_setg(errp, "Offset (%" PRIu64 ") cannot be greater than " 96 "size of the containing file (%" PRId64 ")", 97 s->offset, real_size); 98 ret = -EINVAL; 99 goto end; 100 } 101 102 if (qemu_opt_find(opts, "size") != NULL) { 103 s->size = qemu_opt_get_size(opts, "size", 0); 104 s->has_size = true; 105 } else { 106 s->has_size = false; 107 s->size = real_size - s->offset; 108 } 109 110 /* Check size and offset */ 111 if ((real_size - s->offset) < s->size) { 112 error_setg(errp, "The sum of offset (%" PRIu64 ") and size " 113 "(%" PRIu64 ") has to be smaller or equal to the " 114 " actual size of the containing file (%" PRId64 ")", 115 s->offset, s->size, real_size); 116 ret = -EINVAL; 117 goto end; 118 } 119 120 /* Make sure size is multiple of BDRV_SECTOR_SIZE to prevent rounding 121 * up and leaking out of the specified area. */ 122 if (s->has_size && !QEMU_IS_ALIGNED(s->size, BDRV_SECTOR_SIZE)) { 123 error_setg(errp, "Specified size is not multiple of %llu", 124 BDRV_SECTOR_SIZE); 125 ret = -EINVAL; 126 goto end; 127 } 128 129 ret = 0; 130 131 end: 132 133 qemu_opts_del(opts); 134 135 return ret; 136 } 137 138 static int raw_reopen_prepare(BDRVReopenState *reopen_state, 139 BlockReopenQueue *queue, Error **errp) 140 { 141 assert(reopen_state != NULL); 142 assert(reopen_state->bs != NULL); 143 144 reopen_state->opaque = g_new0(BDRVRawState, 1); 145 146 return raw_read_options( 147 reopen_state->options, 148 reopen_state->bs, 149 reopen_state->opaque, 150 errp); 151 } 152 153 static void raw_reopen_commit(BDRVReopenState *state) 154 { 155 BDRVRawState *new_s = state->opaque; 156 BDRVRawState *s = state->bs->opaque; 157 158 memcpy(s, new_s, sizeof(BDRVRawState)); 159 160 g_free(state->opaque); 161 state->opaque = NULL; 162 } 163 164 static void raw_reopen_abort(BDRVReopenState *state) 165 { 166 g_free(state->opaque); 167 state->opaque = NULL; 168 } 169 170 static int coroutine_fn raw_co_preadv(BlockDriverState *bs, uint64_t offset, 171 uint64_t bytes, QEMUIOVector *qiov, 172 int flags) 173 { 174 BDRVRawState *s = bs->opaque; 175 176 if (offset > UINT64_MAX - s->offset) { 177 return -EINVAL; 178 } 179 offset += s->offset; 180 181 BLKDBG_EVENT(bs->file, BLKDBG_READ_AIO); 182 return bdrv_co_preadv(bs->file, offset, bytes, qiov, flags); 183 } 184 185 static int coroutine_fn raw_co_pwritev(BlockDriverState *bs, uint64_t offset, 186 uint64_t bytes, QEMUIOVector *qiov, 187 int flags) 188 { 189 BDRVRawState *s = bs->opaque; 190 void *buf = NULL; 191 BlockDriver *drv; 192 QEMUIOVector local_qiov; 193 int ret; 194 195 if (s->has_size && (offset > s->size || bytes > (s->size - offset))) { 196 /* There's not enough space for the data. Don't write anything and just 197 * fail to prevent leaking out of the size specified in options. */ 198 return -ENOSPC; 199 } 200 201 if (offset > UINT64_MAX - s->offset) { 202 ret = -EINVAL; 203 goto fail; 204 } 205 206 if (bs->probed && offset < BLOCK_PROBE_BUF_SIZE && bytes) { 207 /* Handling partial writes would be a pain - so we just 208 * require that guests have 512-byte request alignment if 209 * probing occurred */ 210 QEMU_BUILD_BUG_ON(BLOCK_PROBE_BUF_SIZE != 512); 211 QEMU_BUILD_BUG_ON(BDRV_SECTOR_SIZE != 512); 212 assert(offset == 0 && bytes >= BLOCK_PROBE_BUF_SIZE); 213 214 buf = qemu_try_blockalign(bs->file->bs, 512); 215 if (!buf) { 216 ret = -ENOMEM; 217 goto fail; 218 } 219 220 ret = qemu_iovec_to_buf(qiov, 0, buf, 512); 221 if (ret != 512) { 222 ret = -EINVAL; 223 goto fail; 224 } 225 226 drv = bdrv_probe_all(buf, 512, NULL); 227 if (drv != bs->drv) { 228 ret = -EPERM; 229 goto fail; 230 } 231 232 /* Use the checked buffer, a malicious guest might be overwriting its 233 * original buffer in the background. */ 234 qemu_iovec_init(&local_qiov, qiov->niov + 1); 235 qemu_iovec_add(&local_qiov, buf, 512); 236 qemu_iovec_concat(&local_qiov, qiov, 512, qiov->size - 512); 237 qiov = &local_qiov; 238 } 239 240 offset += s->offset; 241 242 BLKDBG_EVENT(bs->file, BLKDBG_WRITE_AIO); 243 ret = bdrv_co_pwritev(bs->file, offset, bytes, qiov, flags); 244 245 fail: 246 if (qiov == &local_qiov) { 247 qemu_iovec_destroy(&local_qiov); 248 } 249 qemu_vfree(buf); 250 return ret; 251 } 252 253 static int64_t coroutine_fn raw_co_get_block_status(BlockDriverState *bs, 254 int64_t sector_num, 255 int nb_sectors, int *pnum, 256 BlockDriverState **file) 257 { 258 BDRVRawState *s = bs->opaque; 259 *pnum = nb_sectors; 260 *file = bs->file->bs; 261 sector_num += s->offset / BDRV_SECTOR_SIZE; 262 return BDRV_BLOCK_RAW | BDRV_BLOCK_OFFSET_VALID | BDRV_BLOCK_DATA | 263 (sector_num << BDRV_SECTOR_BITS); 264 } 265 266 static int coroutine_fn raw_co_pwrite_zeroes(BlockDriverState *bs, 267 int64_t offset, int count, 268 BdrvRequestFlags flags) 269 { 270 BDRVRawState *s = bs->opaque; 271 if (offset > UINT64_MAX - s->offset) { 272 return -EINVAL; 273 } 274 offset += s->offset; 275 return bdrv_co_pwrite_zeroes(bs->file, offset, count, flags); 276 } 277 278 static int coroutine_fn raw_co_pdiscard(BlockDriverState *bs, 279 int64_t offset, int count) 280 { 281 BDRVRawState *s = bs->opaque; 282 if (offset > UINT64_MAX - s->offset) { 283 return -EINVAL; 284 } 285 offset += s->offset; 286 return bdrv_co_pdiscard(bs->file->bs, offset, count); 287 } 288 289 static int64_t raw_getlength(BlockDriverState *bs) 290 { 291 int64_t len; 292 BDRVRawState *s = bs->opaque; 293 294 /* Update size. It should not change unless the file was externally 295 * modified. */ 296 len = bdrv_getlength(bs->file->bs); 297 if (len < 0) { 298 return len; 299 } 300 301 if (len < s->offset) { 302 s->size = 0; 303 } else { 304 if (s->has_size) { 305 /* Try to honour the size */ 306 s->size = MIN(s->size, len - s->offset); 307 } else { 308 s->size = len - s->offset; 309 } 310 } 311 312 return s->size; 313 } 314 315 static int raw_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) 316 { 317 return bdrv_get_info(bs->file->bs, bdi); 318 } 319 320 static void raw_refresh_limits(BlockDriverState *bs, Error **errp) 321 { 322 if (bs->probed) { 323 /* To make it easier to protect the first sector, any probed 324 * image is restricted to read-modify-write on sub-sector 325 * operations. */ 326 bs->bl.request_alignment = BDRV_SECTOR_SIZE; 327 } 328 } 329 330 static int raw_truncate(BlockDriverState *bs, int64_t offset) 331 { 332 BDRVRawState *s = bs->opaque; 333 334 if (s->has_size) { 335 return -ENOTSUP; 336 } 337 338 if (INT64_MAX - offset < s->offset) { 339 return -EINVAL; 340 } 341 342 s->size = offset; 343 offset += s->offset; 344 return bdrv_truncate(bs->file, offset); 345 } 346 347 static int raw_media_changed(BlockDriverState *bs) 348 { 349 return bdrv_media_changed(bs->file->bs); 350 } 351 352 static void raw_eject(BlockDriverState *bs, bool eject_flag) 353 { 354 bdrv_eject(bs->file->bs, eject_flag); 355 } 356 357 static void raw_lock_medium(BlockDriverState *bs, bool locked) 358 { 359 bdrv_lock_medium(bs->file->bs, locked); 360 } 361 362 static int raw_co_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) 363 { 364 BDRVRawState *s = bs->opaque; 365 if (s->offset || s->has_size) { 366 return -ENOTSUP; 367 } 368 return bdrv_co_ioctl(bs->file->bs, req, buf); 369 } 370 371 static int raw_has_zero_init(BlockDriverState *bs) 372 { 373 return bdrv_has_zero_init(bs->file->bs); 374 } 375 376 static int raw_create(const char *filename, QemuOpts *opts, Error **errp) 377 { 378 return bdrv_create_file(filename, opts, errp); 379 } 380 381 static int raw_open(BlockDriverState *bs, QDict *options, int flags, 382 Error **errp) 383 { 384 BDRVRawState *s = bs->opaque; 385 int ret; 386 387 bs->file = bdrv_open_child(NULL, options, "file", bs, &child_file, 388 false, errp); 389 if (!bs->file) { 390 return -EINVAL; 391 } 392 393 bs->sg = bs->file->bs->sg; 394 bs->supported_write_flags = BDRV_REQ_FUA & 395 bs->file->bs->supported_write_flags; 396 bs->supported_zero_flags = (BDRV_REQ_FUA | BDRV_REQ_MAY_UNMAP) & 397 bs->file->bs->supported_zero_flags; 398 399 if (bs->probed && !bdrv_is_read_only(bs)) { 400 fprintf(stderr, 401 "WARNING: Image format was not specified for '%s' and probing " 402 "guessed raw.\n" 403 " Automatically detecting the format is dangerous for " 404 "raw images, write operations on block 0 will be restricted.\n" 405 " Specify the 'raw' format explicitly to remove the " 406 "restrictions.\n", 407 bs->file->bs->filename); 408 } 409 410 ret = raw_read_options(options, bs, s, errp); 411 if (ret < 0) { 412 return ret; 413 } 414 415 if (bs->sg && (s->offset || s->has_size)) { 416 error_setg(errp, "Cannot use offset/size with SCSI generic devices"); 417 return -EINVAL; 418 } 419 420 return 0; 421 } 422 423 static void raw_close(BlockDriverState *bs) 424 { 425 } 426 427 static int raw_probe(const uint8_t *buf, int buf_size, const char *filename) 428 { 429 /* smallest possible positive score so that raw is used if and only if no 430 * other block driver works 431 */ 432 return 1; 433 } 434 435 static int raw_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz) 436 { 437 BDRVRawState *s = bs->opaque; 438 int ret; 439 440 ret = bdrv_probe_blocksizes(bs->file->bs, bsz); 441 if (ret < 0) { 442 return ret; 443 } 444 445 if (!QEMU_IS_ALIGNED(s->offset, MAX(bsz->log, bsz->phys))) { 446 return -ENOTSUP; 447 } 448 449 return 0; 450 } 451 452 static int raw_probe_geometry(BlockDriverState *bs, HDGeometry *geo) 453 { 454 BDRVRawState *s = bs->opaque; 455 if (s->offset || s->has_size) { 456 return -ENOTSUP; 457 } 458 return bdrv_probe_geometry(bs->file->bs, geo); 459 } 460 461 BlockDriver bdrv_raw = { 462 .format_name = "raw", 463 .instance_size = sizeof(BDRVRawState), 464 .bdrv_probe = &raw_probe, 465 .bdrv_reopen_prepare = &raw_reopen_prepare, 466 .bdrv_reopen_commit = &raw_reopen_commit, 467 .bdrv_reopen_abort = &raw_reopen_abort, 468 .bdrv_open = &raw_open, 469 .bdrv_close = &raw_close, 470 .bdrv_create = &raw_create, 471 .bdrv_co_preadv = &raw_co_preadv, 472 .bdrv_co_pwritev = &raw_co_pwritev, 473 .bdrv_co_pwrite_zeroes = &raw_co_pwrite_zeroes, 474 .bdrv_co_pdiscard = &raw_co_pdiscard, 475 .bdrv_co_get_block_status = &raw_co_get_block_status, 476 .bdrv_truncate = &raw_truncate, 477 .bdrv_getlength = &raw_getlength, 478 .has_variable_length = true, 479 .bdrv_get_info = &raw_get_info, 480 .bdrv_refresh_limits = &raw_refresh_limits, 481 .bdrv_probe_blocksizes = &raw_probe_blocksizes, 482 .bdrv_probe_geometry = &raw_probe_geometry, 483 .bdrv_media_changed = &raw_media_changed, 484 .bdrv_eject = &raw_eject, 485 .bdrv_lock_medium = &raw_lock_medium, 486 .bdrv_co_ioctl = &raw_co_ioctl, 487 .create_opts = &raw_create_opts, 488 .bdrv_has_zero_init = &raw_has_zero_init 489 }; 490 491 static void bdrv_raw_init(void) 492 { 493 bdrv_register(&bdrv_raw); 494 } 495 496 block_init(bdrv_raw_init); 497