1 /* 2 * Block driver for Connectix / Microsoft Virtual PC images 3 * 4 * Copyright (c) 2005 Alex Beregszaszi 5 * Copyright (c) 2009 Kevin Wolf <kwolf@suse.de> 6 * 7 * Permission is hereby granted, free of charge, to any person obtaining a copy 8 * of this software and associated documentation files (the "Software"), to deal 9 * in the Software without restriction, including without limitation the rights 10 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 11 * copies of the Software, and to permit persons to whom the Software is 12 * furnished to do so, subject to the following conditions: 13 * 14 * The above copyright notice and this permission notice shall be included in 15 * all copies or substantial portions of the Software. 16 * 17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 20 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 21 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 22 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN 23 * THE SOFTWARE. 24 */ 25 #include "qemu-common.h" 26 #include "block_int.h" 27 #include "module.h" 28 29 /**************************************************************/ 30 31 #define HEADER_SIZE 512 32 33 //#define CACHE 34 35 enum vhd_type { 36 VHD_FIXED = 2, 37 VHD_DYNAMIC = 3, 38 VHD_DIFFERENCING = 4, 39 }; 40 41 // Seconds since Jan 1, 2000 0:00:00 (UTC) 42 #define VHD_TIMESTAMP_BASE 946684800 43 44 // always big-endian 45 struct vhd_footer { 46 char creator[8]; // "conectix" 47 uint32_t features; 48 uint32_t version; 49 50 // Offset of next header structure, 0xFFFFFFFF if none 51 uint64_t data_offset; 52 53 // Seconds since Jan 1, 2000 0:00:00 (UTC) 54 uint32_t timestamp; 55 56 char creator_app[4]; // "vpc " 57 uint16_t major; 58 uint16_t minor; 59 char creator_os[4]; // "Wi2k" 60 61 uint64_t orig_size; 62 uint64_t size; 63 64 uint16_t cyls; 65 uint8_t heads; 66 uint8_t secs_per_cyl; 67 68 uint32_t type; 69 70 // Checksum of the Hard Disk Footer ("one's complement of the sum of all 71 // the bytes in the footer without the checksum field") 72 uint32_t checksum; 73 74 // UUID used to identify a parent hard disk (backing file) 75 uint8_t uuid[16]; 76 77 uint8_t in_saved_state; 78 }; 79 80 struct vhd_dyndisk_header { 81 char magic[8]; // "cxsparse" 82 83 // Offset of next header structure, 0xFFFFFFFF if none 84 uint64_t data_offset; 85 86 // Offset of the Block Allocation Table (BAT) 87 uint64_t table_offset; 88 89 uint32_t version; 90 uint32_t max_table_entries; // 32bit/entry 91 92 // 2 MB by default, must be a power of two 93 uint32_t block_size; 94 95 uint32_t checksum; 96 uint8_t parent_uuid[16]; 97 uint32_t parent_timestamp; 98 uint32_t reserved; 99 100 // Backing file name (in UTF-16) 101 uint8_t parent_name[512]; 102 103 struct { 104 uint32_t platform; 105 uint32_t data_space; 106 uint32_t data_length; 107 uint32_t reserved; 108 uint64_t data_offset; 109 } parent_locator[8]; 110 }; 111 112 typedef struct BDRVVPCState { 113 uint8_t footer_buf[HEADER_SIZE]; 114 uint64_t free_data_block_offset; 115 int max_table_entries; 116 uint32_t *pagetable; 117 uint64_t bat_offset; 118 uint64_t last_bitmap_offset; 119 120 uint32_t block_size; 121 uint32_t bitmap_size; 122 123 #ifdef CACHE 124 uint8_t *pageentry_u8; 125 uint32_t *pageentry_u32; 126 uint16_t *pageentry_u16; 127 128 uint64_t last_bitmap; 129 #endif 130 } BDRVVPCState; 131 132 static uint32_t vpc_checksum(uint8_t* buf, size_t size) 133 { 134 uint32_t res = 0; 135 int i; 136 137 for (i = 0; i < size; i++) 138 res += buf[i]; 139 140 return ~res; 141 } 142 143 144 static int vpc_probe(const uint8_t *buf, int buf_size, const char *filename) 145 { 146 if (buf_size >= 8 && !strncmp((char *)buf, "conectix", 8)) 147 return 100; 148 return 0; 149 } 150 151 static int vpc_open(BlockDriverState *bs, int flags) 152 { 153 BDRVVPCState *s = bs->opaque; 154 int i; 155 struct vhd_footer* footer; 156 struct vhd_dyndisk_header* dyndisk_header; 157 uint8_t buf[HEADER_SIZE]; 158 uint32_t checksum; 159 int err = -1; 160 161 if (bdrv_pread(bs->file, 0, s->footer_buf, HEADER_SIZE) != HEADER_SIZE) 162 goto fail; 163 164 footer = (struct vhd_footer*) s->footer_buf; 165 if (strncmp(footer->creator, "conectix", 8)) 166 goto fail; 167 168 checksum = be32_to_cpu(footer->checksum); 169 footer->checksum = 0; 170 if (vpc_checksum(s->footer_buf, HEADER_SIZE) != checksum) 171 fprintf(stderr, "block-vpc: The header checksum of '%s' is " 172 "incorrect.\n", bs->filename); 173 174 // The visible size of a image in Virtual PC depends on the geometry 175 // rather than on the size stored in the footer (the size in the footer 176 // is too large usually) 177 bs->total_sectors = (int64_t) 178 be16_to_cpu(footer->cyls) * footer->heads * footer->secs_per_cyl; 179 180 if (bs->total_sectors >= 65535 * 16 * 255) { 181 err = -EFBIG; 182 goto fail; 183 } 184 185 if (bdrv_pread(bs->file, be64_to_cpu(footer->data_offset), buf, HEADER_SIZE) 186 != HEADER_SIZE) 187 goto fail; 188 189 dyndisk_header = (struct vhd_dyndisk_header*) buf; 190 191 if (strncmp(dyndisk_header->magic, "cxsparse", 8)) 192 goto fail; 193 194 195 s->block_size = be32_to_cpu(dyndisk_header->block_size); 196 s->bitmap_size = ((s->block_size / (8 * 512)) + 511) & ~511; 197 198 s->max_table_entries = be32_to_cpu(dyndisk_header->max_table_entries); 199 s->pagetable = g_malloc(s->max_table_entries * 4); 200 201 s->bat_offset = be64_to_cpu(dyndisk_header->table_offset); 202 if (bdrv_pread(bs->file, s->bat_offset, s->pagetable, 203 s->max_table_entries * 4) != s->max_table_entries * 4) 204 goto fail; 205 206 s->free_data_block_offset = 207 (s->bat_offset + (s->max_table_entries * 4) + 511) & ~511; 208 209 for (i = 0; i < s->max_table_entries; i++) { 210 be32_to_cpus(&s->pagetable[i]); 211 if (s->pagetable[i] != 0xFFFFFFFF) { 212 int64_t next = (512 * (int64_t) s->pagetable[i]) + 213 s->bitmap_size + s->block_size; 214 215 if (next> s->free_data_block_offset) 216 s->free_data_block_offset = next; 217 } 218 } 219 220 s->last_bitmap_offset = (int64_t) -1; 221 222 #ifdef CACHE 223 s->pageentry_u8 = g_malloc(512); 224 s->pageentry_u32 = s->pageentry_u8; 225 s->pageentry_u16 = s->pageentry_u8; 226 s->last_pagetable = -1; 227 #endif 228 229 return 0; 230 fail: 231 return err; 232 } 233 234 /* 235 * Returns the absolute byte offset of the given sector in the image file. 236 * If the sector is not allocated, -1 is returned instead. 237 * 238 * The parameter write must be 1 if the offset will be used for a write 239 * operation (the block bitmaps is updated then), 0 otherwise. 240 */ 241 static inline int64_t get_sector_offset(BlockDriverState *bs, 242 int64_t sector_num, int write) 243 { 244 BDRVVPCState *s = bs->opaque; 245 uint64_t offset = sector_num * 512; 246 uint64_t bitmap_offset, block_offset; 247 uint32_t pagetable_index, pageentry_index; 248 249 pagetable_index = offset / s->block_size; 250 pageentry_index = (offset % s->block_size) / 512; 251 252 if (pagetable_index >= s->max_table_entries || s->pagetable[pagetable_index] == 0xffffffff) 253 return -1; // not allocated 254 255 bitmap_offset = 512 * (uint64_t) s->pagetable[pagetable_index]; 256 block_offset = bitmap_offset + s->bitmap_size + (512 * pageentry_index); 257 258 // We must ensure that we don't write to any sectors which are marked as 259 // unused in the bitmap. We get away with setting all bits in the block 260 // bitmap each time we write to a new block. This might cause Virtual PC to 261 // miss sparse read optimization, but it's not a problem in terms of 262 // correctness. 263 if (write && (s->last_bitmap_offset != bitmap_offset)) { 264 uint8_t bitmap[s->bitmap_size]; 265 266 s->last_bitmap_offset = bitmap_offset; 267 memset(bitmap, 0xff, s->bitmap_size); 268 bdrv_pwrite_sync(bs->file, bitmap_offset, bitmap, s->bitmap_size); 269 } 270 271 // printf("sector: %" PRIx64 ", index: %x, offset: %x, bioff: %" PRIx64 ", bloff: %" PRIx64 "\n", 272 // sector_num, pagetable_index, pageentry_index, 273 // bitmap_offset, block_offset); 274 275 // disabled by reason 276 #if 0 277 #ifdef CACHE 278 if (bitmap_offset != s->last_bitmap) 279 { 280 lseek(s->fd, bitmap_offset, SEEK_SET); 281 282 s->last_bitmap = bitmap_offset; 283 284 // Scary! Bitmap is stored as big endian 32bit entries, 285 // while we used to look it up byte by byte 286 read(s->fd, s->pageentry_u8, 512); 287 for (i = 0; i < 128; i++) 288 be32_to_cpus(&s->pageentry_u32[i]); 289 } 290 291 if ((s->pageentry_u8[pageentry_index / 8] >> (pageentry_index % 8)) & 1) 292 return -1; 293 #else 294 lseek(s->fd, bitmap_offset + (pageentry_index / 8), SEEK_SET); 295 296 read(s->fd, &bitmap_entry, 1); 297 298 if ((bitmap_entry >> (pageentry_index % 8)) & 1) 299 return -1; // not allocated 300 #endif 301 #endif 302 303 return block_offset; 304 } 305 306 /* 307 * Writes the footer to the end of the image file. This is needed when the 308 * file grows as it overwrites the old footer 309 * 310 * Returns 0 on success and < 0 on error 311 */ 312 static int rewrite_footer(BlockDriverState* bs) 313 { 314 int ret; 315 BDRVVPCState *s = bs->opaque; 316 int64_t offset = s->free_data_block_offset; 317 318 ret = bdrv_pwrite_sync(bs->file, offset, s->footer_buf, HEADER_SIZE); 319 if (ret < 0) 320 return ret; 321 322 return 0; 323 } 324 325 /* 326 * Allocates a new block. This involves writing a new footer and updating 327 * the Block Allocation Table to use the space at the old end of the image 328 * file (overwriting the old footer) 329 * 330 * Returns the sectors' offset in the image file on success and < 0 on error 331 */ 332 static int64_t alloc_block(BlockDriverState* bs, int64_t sector_num) 333 { 334 BDRVVPCState *s = bs->opaque; 335 int64_t bat_offset; 336 uint32_t index, bat_value; 337 int ret; 338 uint8_t bitmap[s->bitmap_size]; 339 340 // Check if sector_num is valid 341 if ((sector_num < 0) || (sector_num > bs->total_sectors)) 342 return -1; 343 344 // Write entry into in-memory BAT 345 index = (sector_num * 512) / s->block_size; 346 if (s->pagetable[index] != 0xFFFFFFFF) 347 return -1; 348 349 s->pagetable[index] = s->free_data_block_offset / 512; 350 351 // Initialize the block's bitmap 352 memset(bitmap, 0xff, s->bitmap_size); 353 bdrv_pwrite_sync(bs->file, s->free_data_block_offset, bitmap, 354 s->bitmap_size); 355 356 // Write new footer (the old one will be overwritten) 357 s->free_data_block_offset += s->block_size + s->bitmap_size; 358 ret = rewrite_footer(bs); 359 if (ret < 0) 360 goto fail; 361 362 // Write BAT entry to disk 363 bat_offset = s->bat_offset + (4 * index); 364 bat_value = be32_to_cpu(s->pagetable[index]); 365 ret = bdrv_pwrite_sync(bs->file, bat_offset, &bat_value, 4); 366 if (ret < 0) 367 goto fail; 368 369 return get_sector_offset(bs, sector_num, 0); 370 371 fail: 372 s->free_data_block_offset -= (s->block_size + s->bitmap_size); 373 return -1; 374 } 375 376 static int vpc_read(BlockDriverState *bs, int64_t sector_num, 377 uint8_t *buf, int nb_sectors) 378 { 379 BDRVVPCState *s = bs->opaque; 380 int ret; 381 int64_t offset; 382 int64_t sectors, sectors_per_block; 383 384 while (nb_sectors > 0) { 385 offset = get_sector_offset(bs, sector_num, 0); 386 387 sectors_per_block = s->block_size >> BDRV_SECTOR_BITS; 388 sectors = sectors_per_block - (sector_num % sectors_per_block); 389 if (sectors > nb_sectors) { 390 sectors = nb_sectors; 391 } 392 393 if (offset == -1) { 394 memset(buf, 0, sectors * BDRV_SECTOR_SIZE); 395 } else { 396 ret = bdrv_pread(bs->file, offset, buf, 397 sectors * BDRV_SECTOR_SIZE); 398 if (ret != sectors * BDRV_SECTOR_SIZE) { 399 return -1; 400 } 401 } 402 403 nb_sectors -= sectors; 404 sector_num += sectors; 405 buf += sectors * BDRV_SECTOR_SIZE; 406 } 407 return 0; 408 } 409 410 static int vpc_write(BlockDriverState *bs, int64_t sector_num, 411 const uint8_t *buf, int nb_sectors) 412 { 413 BDRVVPCState *s = bs->opaque; 414 int64_t offset; 415 int64_t sectors, sectors_per_block; 416 int ret; 417 418 while (nb_sectors > 0) { 419 offset = get_sector_offset(bs, sector_num, 1); 420 421 sectors_per_block = s->block_size >> BDRV_SECTOR_BITS; 422 sectors = sectors_per_block - (sector_num % sectors_per_block); 423 if (sectors > nb_sectors) { 424 sectors = nb_sectors; 425 } 426 427 if (offset == -1) { 428 offset = alloc_block(bs, sector_num); 429 if (offset < 0) 430 return -1; 431 } 432 433 ret = bdrv_pwrite(bs->file, offset, buf, sectors * BDRV_SECTOR_SIZE); 434 if (ret != sectors * BDRV_SECTOR_SIZE) { 435 return -1; 436 } 437 438 nb_sectors -= sectors; 439 sector_num += sectors; 440 buf += sectors * BDRV_SECTOR_SIZE; 441 } 442 443 return 0; 444 } 445 446 static int vpc_flush(BlockDriverState *bs) 447 { 448 return bdrv_flush(bs->file); 449 } 450 451 /* 452 * Calculates the number of cylinders, heads and sectors per cylinder 453 * based on a given number of sectors. This is the algorithm described 454 * in the VHD specification. 455 * 456 * Note that the geometry doesn't always exactly match total_sectors but 457 * may round it down. 458 * 459 * Returns 0 on success, -EFBIG if the size is larger than 127 GB 460 */ 461 static int calculate_geometry(int64_t total_sectors, uint16_t* cyls, 462 uint8_t* heads, uint8_t* secs_per_cyl) 463 { 464 uint32_t cyls_times_heads; 465 466 if (total_sectors > 65535 * 16 * 255) 467 return -EFBIG; 468 469 if (total_sectors > 65535 * 16 * 63) { 470 *secs_per_cyl = 255; 471 *heads = 16; 472 cyls_times_heads = total_sectors / *secs_per_cyl; 473 } else { 474 *secs_per_cyl = 17; 475 cyls_times_heads = total_sectors / *secs_per_cyl; 476 *heads = (cyls_times_heads + 1023) / 1024; 477 478 if (*heads < 4) 479 *heads = 4; 480 481 if (cyls_times_heads >= (*heads * 1024) || *heads > 16) { 482 *secs_per_cyl = 31; 483 *heads = 16; 484 cyls_times_heads = total_sectors / *secs_per_cyl; 485 } 486 487 if (cyls_times_heads >= (*heads * 1024)) { 488 *secs_per_cyl = 63; 489 *heads = 16; 490 cyls_times_heads = total_sectors / *secs_per_cyl; 491 } 492 } 493 494 *cyls = cyls_times_heads / *heads; 495 496 return 0; 497 } 498 499 static int vpc_create(const char *filename, QEMUOptionParameter *options) 500 { 501 uint8_t buf[1024]; 502 struct vhd_footer* footer = (struct vhd_footer*) buf; 503 struct vhd_dyndisk_header* dyndisk_header = 504 (struct vhd_dyndisk_header*) buf; 505 int fd, i; 506 uint16_t cyls = 0; 507 uint8_t heads = 0; 508 uint8_t secs_per_cyl = 0; 509 size_t block_size, num_bat_entries; 510 int64_t total_sectors = 0; 511 int ret = -EIO; 512 513 // Read out options 514 total_sectors = get_option_parameter(options, BLOCK_OPT_SIZE)->value.n / 515 BDRV_SECTOR_SIZE; 516 517 // Create the file 518 fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0644); 519 if (fd < 0) 520 return -EIO; 521 522 /* Calculate matching total_size and geometry. Increase the number of 523 sectors requested until we get enough (or fail). */ 524 for (i = 0; total_sectors > (int64_t)cyls * heads * secs_per_cyl; i++) { 525 if (calculate_geometry(total_sectors + i, 526 &cyls, &heads, &secs_per_cyl)) { 527 ret = -EFBIG; 528 goto fail; 529 } 530 } 531 total_sectors = (int64_t) cyls * heads * secs_per_cyl; 532 533 // Prepare the Hard Disk Footer 534 memset(buf, 0, 1024); 535 536 memcpy(footer->creator, "conectix", 8); 537 // TODO Check if "qemu" creator_app is ok for VPC 538 memcpy(footer->creator_app, "qemu", 4); 539 memcpy(footer->creator_os, "Wi2k", 4); 540 541 footer->features = be32_to_cpu(0x02); 542 footer->version = be32_to_cpu(0x00010000); 543 footer->data_offset = be64_to_cpu(HEADER_SIZE); 544 footer->timestamp = be32_to_cpu(time(NULL) - VHD_TIMESTAMP_BASE); 545 546 // Version of Virtual PC 2007 547 footer->major = be16_to_cpu(0x0005); 548 footer->minor =be16_to_cpu(0x0003); 549 550 footer->orig_size = be64_to_cpu(total_sectors * 512); 551 footer->size = be64_to_cpu(total_sectors * 512); 552 553 footer->cyls = be16_to_cpu(cyls); 554 footer->heads = heads; 555 footer->secs_per_cyl = secs_per_cyl; 556 557 footer->type = be32_to_cpu(VHD_DYNAMIC); 558 559 // TODO uuid is missing 560 561 footer->checksum = be32_to_cpu(vpc_checksum(buf, HEADER_SIZE)); 562 563 // Write the footer (twice: at the beginning and at the end) 564 block_size = 0x200000; 565 num_bat_entries = (total_sectors + block_size / 512) / (block_size / 512); 566 567 if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE) { 568 goto fail; 569 } 570 571 if (lseek(fd, 1536 + ((num_bat_entries * 4 + 511) & ~511), SEEK_SET) < 0) { 572 goto fail; 573 } 574 if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE) { 575 goto fail; 576 } 577 578 // Write the initial BAT 579 if (lseek(fd, 3 * 512, SEEK_SET) < 0) { 580 goto fail; 581 } 582 583 memset(buf, 0xFF, 512); 584 for (i = 0; i < (num_bat_entries * 4 + 511) / 512; i++) { 585 if (write(fd, buf, 512) != 512) { 586 goto fail; 587 } 588 } 589 590 591 // Prepare the Dynamic Disk Header 592 memset(buf, 0, 1024); 593 594 memcpy(dyndisk_header->magic, "cxsparse", 8); 595 596 dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFF); 597 dyndisk_header->table_offset = be64_to_cpu(3 * 512); 598 dyndisk_header->version = be32_to_cpu(0x00010000); 599 dyndisk_header->block_size = be32_to_cpu(block_size); 600 dyndisk_header->max_table_entries = be32_to_cpu(num_bat_entries); 601 602 dyndisk_header->checksum = be32_to_cpu(vpc_checksum(buf, 1024)); 603 604 // Write the header 605 if (lseek(fd, 512, SEEK_SET) < 0) { 606 goto fail; 607 } 608 609 if (write(fd, buf, 1024) != 1024) { 610 goto fail; 611 } 612 ret = 0; 613 614 fail: 615 close(fd); 616 return ret; 617 } 618 619 static void vpc_close(BlockDriverState *bs) 620 { 621 BDRVVPCState *s = bs->opaque; 622 g_free(s->pagetable); 623 #ifdef CACHE 624 g_free(s->pageentry_u8); 625 #endif 626 } 627 628 static QEMUOptionParameter vpc_create_options[] = { 629 { 630 .name = BLOCK_OPT_SIZE, 631 .type = OPT_SIZE, 632 .help = "Virtual disk size" 633 }, 634 { NULL } 635 }; 636 637 static BlockDriver bdrv_vpc = { 638 .format_name = "vpc", 639 .instance_size = sizeof(BDRVVPCState), 640 .bdrv_probe = vpc_probe, 641 .bdrv_open = vpc_open, 642 .bdrv_read = vpc_read, 643 .bdrv_write = vpc_write, 644 .bdrv_flush = vpc_flush, 645 .bdrv_close = vpc_close, 646 .bdrv_create = vpc_create, 647 648 .create_options = vpc_create_options, 649 }; 650 651 static void bdrv_vpc_init(void) 652 { 653 bdrv_register(&bdrv_vpc); 654 } 655 656 block_init(bdrv_vpc_init); 657