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/block_int.h" 27 #include "qemu/module.h" 28 #include "migration/migration.h" 29 #if defined(CONFIG_UUID) 30 #include <uuid/uuid.h> 31 #endif 32 33 /**************************************************************/ 34 35 #define HEADER_SIZE 512 36 37 //#define CACHE 38 39 enum vhd_type { 40 VHD_FIXED = 2, 41 VHD_DYNAMIC = 3, 42 VHD_DIFFERENCING = 4, 43 }; 44 45 // Seconds since Jan 1, 2000 0:00:00 (UTC) 46 #define VHD_TIMESTAMP_BASE 946684800 47 48 #define VHD_MAX_SECTORS (65535LL * 255 * 255) 49 50 // always big-endian 51 typedef struct vhd_footer { 52 char creator[8]; // "conectix" 53 uint32_t features; 54 uint32_t version; 55 56 // Offset of next header structure, 0xFFFFFFFF if none 57 uint64_t data_offset; 58 59 // Seconds since Jan 1, 2000 0:00:00 (UTC) 60 uint32_t timestamp; 61 62 char creator_app[4]; // "vpc " 63 uint16_t major; 64 uint16_t minor; 65 char creator_os[4]; // "Wi2k" 66 67 uint64_t orig_size; 68 uint64_t size; 69 70 uint16_t cyls; 71 uint8_t heads; 72 uint8_t secs_per_cyl; 73 74 uint32_t type; 75 76 // Checksum of the Hard Disk Footer ("one's complement of the sum of all 77 // the bytes in the footer without the checksum field") 78 uint32_t checksum; 79 80 // UUID used to identify a parent hard disk (backing file) 81 uint8_t uuid[16]; 82 83 uint8_t in_saved_state; 84 } QEMU_PACKED VHDFooter; 85 86 typedef struct vhd_dyndisk_header { 87 char magic[8]; // "cxsparse" 88 89 // Offset of next header structure, 0xFFFFFFFF if none 90 uint64_t data_offset; 91 92 // Offset of the Block Allocation Table (BAT) 93 uint64_t table_offset; 94 95 uint32_t version; 96 uint32_t max_table_entries; // 32bit/entry 97 98 // 2 MB by default, must be a power of two 99 uint32_t block_size; 100 101 uint32_t checksum; 102 uint8_t parent_uuid[16]; 103 uint32_t parent_timestamp; 104 uint32_t reserved; 105 106 // Backing file name (in UTF-16) 107 uint8_t parent_name[512]; 108 109 struct { 110 uint32_t platform; 111 uint32_t data_space; 112 uint32_t data_length; 113 uint32_t reserved; 114 uint64_t data_offset; 115 } parent_locator[8]; 116 } QEMU_PACKED VHDDynDiskHeader; 117 118 typedef struct BDRVVPCState { 119 CoMutex lock; 120 uint8_t footer_buf[HEADER_SIZE]; 121 uint64_t free_data_block_offset; 122 int max_table_entries; 123 uint32_t *pagetable; 124 uint64_t bat_offset; 125 uint64_t last_bitmap_offset; 126 127 uint32_t block_size; 128 uint32_t bitmap_size; 129 130 #ifdef CACHE 131 uint8_t *pageentry_u8; 132 uint32_t *pageentry_u32; 133 uint16_t *pageentry_u16; 134 135 uint64_t last_bitmap; 136 #endif 137 138 Error *migration_blocker; 139 } BDRVVPCState; 140 141 static uint32_t vpc_checksum(uint8_t* buf, size_t size) 142 { 143 uint32_t res = 0; 144 int i; 145 146 for (i = 0; i < size; i++) 147 res += buf[i]; 148 149 return ~res; 150 } 151 152 153 static int vpc_probe(const uint8_t *buf, int buf_size, const char *filename) 154 { 155 if (buf_size >= 8 && !strncmp((char *)buf, "conectix", 8)) 156 return 100; 157 return 0; 158 } 159 160 static int vpc_open(BlockDriverState *bs, QDict *options, int flags, 161 Error **errp) 162 { 163 BDRVVPCState *s = bs->opaque; 164 int i; 165 VHDFooter *footer; 166 VHDDynDiskHeader *dyndisk_header; 167 uint8_t buf[HEADER_SIZE]; 168 uint32_t checksum; 169 uint64_t computed_size; 170 int disk_type = VHD_DYNAMIC; 171 int ret; 172 173 ret = bdrv_pread(bs->file, 0, s->footer_buf, HEADER_SIZE); 174 if (ret < 0) { 175 goto fail; 176 } 177 178 footer = (VHDFooter *) s->footer_buf; 179 if (strncmp(footer->creator, "conectix", 8)) { 180 int64_t offset = bdrv_getlength(bs->file); 181 if (offset < 0) { 182 ret = offset; 183 goto fail; 184 } else if (offset < HEADER_SIZE) { 185 ret = -EINVAL; 186 goto fail; 187 } 188 189 /* If a fixed disk, the footer is found only at the end of the file */ 190 ret = bdrv_pread(bs->file, offset-HEADER_SIZE, s->footer_buf, 191 HEADER_SIZE); 192 if (ret < 0) { 193 goto fail; 194 } 195 if (strncmp(footer->creator, "conectix", 8)) { 196 error_setg(errp, "invalid VPC image"); 197 ret = -EINVAL; 198 goto fail; 199 } 200 disk_type = VHD_FIXED; 201 } 202 203 checksum = be32_to_cpu(footer->checksum); 204 footer->checksum = 0; 205 if (vpc_checksum(s->footer_buf, HEADER_SIZE) != checksum) 206 fprintf(stderr, "block-vpc: The header checksum of '%s' is " 207 "incorrect.\n", bs->filename); 208 209 /* Write 'checksum' back to footer, or else will leave it with zero. */ 210 footer->checksum = cpu_to_be32(checksum); 211 212 // The visible size of a image in Virtual PC depends on the geometry 213 // rather than on the size stored in the footer (the size in the footer 214 // is too large usually) 215 bs->total_sectors = (int64_t) 216 be16_to_cpu(footer->cyls) * footer->heads * footer->secs_per_cyl; 217 218 /* images created with disk2vhd report a far higher virtual size 219 * than expected with the cyls * heads * sectors_per_cyl formula. 220 * use the footer->size instead if the image was created with 221 * disk2vhd. 222 */ 223 if (!strncmp(footer->creator_app, "d2v", 4)) { 224 bs->total_sectors = be64_to_cpu(footer->size) / BDRV_SECTOR_SIZE; 225 } 226 227 /* Allow a maximum disk size of approximately 2 TB */ 228 if (bs->total_sectors >= VHD_MAX_SECTORS) { 229 ret = -EFBIG; 230 goto fail; 231 } 232 233 if (disk_type == VHD_DYNAMIC) { 234 ret = bdrv_pread(bs->file, be64_to_cpu(footer->data_offset), buf, 235 HEADER_SIZE); 236 if (ret < 0) { 237 goto fail; 238 } 239 240 dyndisk_header = (VHDDynDiskHeader *) buf; 241 242 if (strncmp(dyndisk_header->magic, "cxsparse", 8)) { 243 ret = -EINVAL; 244 goto fail; 245 } 246 247 s->block_size = be32_to_cpu(dyndisk_header->block_size); 248 if (!is_power_of_2(s->block_size) || s->block_size < BDRV_SECTOR_SIZE) { 249 error_setg(errp, "Invalid block size %" PRIu32, s->block_size); 250 ret = -EINVAL; 251 goto fail; 252 } 253 s->bitmap_size = ((s->block_size / (8 * 512)) + 511) & ~511; 254 255 s->max_table_entries = be32_to_cpu(dyndisk_header->max_table_entries); 256 257 if ((bs->total_sectors * 512) / s->block_size > 0xffffffffU) { 258 ret = -EINVAL; 259 goto fail; 260 } 261 if (s->max_table_entries > (VHD_MAX_SECTORS * 512) / s->block_size) { 262 ret = -EINVAL; 263 goto fail; 264 } 265 266 computed_size = (uint64_t) s->max_table_entries * s->block_size; 267 if (computed_size < bs->total_sectors * 512) { 268 ret = -EINVAL; 269 goto fail; 270 } 271 272 s->pagetable = qemu_try_blockalign(bs->file, s->max_table_entries * 4); 273 if (s->pagetable == NULL) { 274 ret = -ENOMEM; 275 goto fail; 276 } 277 278 s->bat_offset = be64_to_cpu(dyndisk_header->table_offset); 279 280 ret = bdrv_pread(bs->file, s->bat_offset, s->pagetable, 281 s->max_table_entries * 4); 282 if (ret < 0) { 283 goto fail; 284 } 285 286 s->free_data_block_offset = 287 (s->bat_offset + (s->max_table_entries * 4) + 511) & ~511; 288 289 for (i = 0; i < s->max_table_entries; i++) { 290 be32_to_cpus(&s->pagetable[i]); 291 if (s->pagetable[i] != 0xFFFFFFFF) { 292 int64_t next = (512 * (int64_t) s->pagetable[i]) + 293 s->bitmap_size + s->block_size; 294 295 if (next > s->free_data_block_offset) { 296 s->free_data_block_offset = next; 297 } 298 } 299 } 300 301 if (s->free_data_block_offset > bdrv_getlength(bs->file)) { 302 error_setg(errp, "block-vpc: free_data_block_offset points after " 303 "the end of file. The image has been truncated."); 304 ret = -EINVAL; 305 goto fail; 306 } 307 308 s->last_bitmap_offset = (int64_t) -1; 309 310 #ifdef CACHE 311 s->pageentry_u8 = g_malloc(512); 312 s->pageentry_u32 = s->pageentry_u8; 313 s->pageentry_u16 = s->pageentry_u8; 314 s->last_pagetable = -1; 315 #endif 316 } 317 318 qemu_co_mutex_init(&s->lock); 319 320 /* Disable migration when VHD images are used */ 321 error_set(&s->migration_blocker, 322 QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED, 323 "vpc", bdrv_get_device_name(bs), "live migration"); 324 migrate_add_blocker(s->migration_blocker); 325 326 return 0; 327 328 fail: 329 qemu_vfree(s->pagetable); 330 #ifdef CACHE 331 g_free(s->pageentry_u8); 332 #endif 333 return ret; 334 } 335 336 static int vpc_reopen_prepare(BDRVReopenState *state, 337 BlockReopenQueue *queue, Error **errp) 338 { 339 return 0; 340 } 341 342 /* 343 * Returns the absolute byte offset of the given sector in the image file. 344 * If the sector is not allocated, -1 is returned instead. 345 * 346 * The parameter write must be 1 if the offset will be used for a write 347 * operation (the block bitmaps is updated then), 0 otherwise. 348 */ 349 static inline int64_t get_sector_offset(BlockDriverState *bs, 350 int64_t sector_num, int write) 351 { 352 BDRVVPCState *s = bs->opaque; 353 uint64_t offset = sector_num * 512; 354 uint64_t bitmap_offset, block_offset; 355 uint32_t pagetable_index, pageentry_index; 356 357 pagetable_index = offset / s->block_size; 358 pageentry_index = (offset % s->block_size) / 512; 359 360 if (pagetable_index >= s->max_table_entries || s->pagetable[pagetable_index] == 0xffffffff) 361 return -1; // not allocated 362 363 bitmap_offset = 512 * (uint64_t) s->pagetable[pagetable_index]; 364 block_offset = bitmap_offset + s->bitmap_size + (512 * pageentry_index); 365 366 // We must ensure that we don't write to any sectors which are marked as 367 // unused in the bitmap. We get away with setting all bits in the block 368 // bitmap each time we write to a new block. This might cause Virtual PC to 369 // miss sparse read optimization, but it's not a problem in terms of 370 // correctness. 371 if (write && (s->last_bitmap_offset != bitmap_offset)) { 372 uint8_t bitmap[s->bitmap_size]; 373 374 s->last_bitmap_offset = bitmap_offset; 375 memset(bitmap, 0xff, s->bitmap_size); 376 bdrv_pwrite_sync(bs->file, bitmap_offset, bitmap, s->bitmap_size); 377 } 378 379 // printf("sector: %" PRIx64 ", index: %x, offset: %x, bioff: %" PRIx64 ", bloff: %" PRIx64 "\n", 380 // sector_num, pagetable_index, pageentry_index, 381 // bitmap_offset, block_offset); 382 383 // disabled by reason 384 #if 0 385 #ifdef CACHE 386 if (bitmap_offset != s->last_bitmap) 387 { 388 lseek(s->fd, bitmap_offset, SEEK_SET); 389 390 s->last_bitmap = bitmap_offset; 391 392 // Scary! Bitmap is stored as big endian 32bit entries, 393 // while we used to look it up byte by byte 394 read(s->fd, s->pageentry_u8, 512); 395 for (i = 0; i < 128; i++) 396 be32_to_cpus(&s->pageentry_u32[i]); 397 } 398 399 if ((s->pageentry_u8[pageentry_index / 8] >> (pageentry_index % 8)) & 1) 400 return -1; 401 #else 402 lseek(s->fd, bitmap_offset + (pageentry_index / 8), SEEK_SET); 403 404 read(s->fd, &bitmap_entry, 1); 405 406 if ((bitmap_entry >> (pageentry_index % 8)) & 1) 407 return -1; // not allocated 408 #endif 409 #endif 410 411 return block_offset; 412 } 413 414 /* 415 * Writes the footer to the end of the image file. This is needed when the 416 * file grows as it overwrites the old footer 417 * 418 * Returns 0 on success and < 0 on error 419 */ 420 static int rewrite_footer(BlockDriverState* bs) 421 { 422 int ret; 423 BDRVVPCState *s = bs->opaque; 424 int64_t offset = s->free_data_block_offset; 425 426 ret = bdrv_pwrite_sync(bs->file, offset, s->footer_buf, HEADER_SIZE); 427 if (ret < 0) 428 return ret; 429 430 return 0; 431 } 432 433 /* 434 * Allocates a new block. This involves writing a new footer and updating 435 * the Block Allocation Table to use the space at the old end of the image 436 * file (overwriting the old footer) 437 * 438 * Returns the sectors' offset in the image file on success and < 0 on error 439 */ 440 static int64_t alloc_block(BlockDriverState* bs, int64_t sector_num) 441 { 442 BDRVVPCState *s = bs->opaque; 443 int64_t bat_offset; 444 uint32_t index, bat_value; 445 int ret; 446 uint8_t bitmap[s->bitmap_size]; 447 448 // Check if sector_num is valid 449 if ((sector_num < 0) || (sector_num > bs->total_sectors)) 450 return -1; 451 452 // Write entry into in-memory BAT 453 index = (sector_num * 512) / s->block_size; 454 if (s->pagetable[index] != 0xFFFFFFFF) 455 return -1; 456 457 s->pagetable[index] = s->free_data_block_offset / 512; 458 459 // Initialize the block's bitmap 460 memset(bitmap, 0xff, s->bitmap_size); 461 ret = bdrv_pwrite_sync(bs->file, s->free_data_block_offset, bitmap, 462 s->bitmap_size); 463 if (ret < 0) { 464 return ret; 465 } 466 467 // Write new footer (the old one will be overwritten) 468 s->free_data_block_offset += s->block_size + s->bitmap_size; 469 ret = rewrite_footer(bs); 470 if (ret < 0) 471 goto fail; 472 473 // Write BAT entry to disk 474 bat_offset = s->bat_offset + (4 * index); 475 bat_value = cpu_to_be32(s->pagetable[index]); 476 ret = bdrv_pwrite_sync(bs->file, bat_offset, &bat_value, 4); 477 if (ret < 0) 478 goto fail; 479 480 return get_sector_offset(bs, sector_num, 0); 481 482 fail: 483 s->free_data_block_offset -= (s->block_size + s->bitmap_size); 484 return -1; 485 } 486 487 static int vpc_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) 488 { 489 BDRVVPCState *s = (BDRVVPCState *)bs->opaque; 490 VHDFooter *footer = (VHDFooter *) s->footer_buf; 491 492 if (be32_to_cpu(footer->type) != VHD_FIXED) { 493 bdi->cluster_size = s->block_size; 494 } 495 496 bdi->unallocated_blocks_are_zero = true; 497 return 0; 498 } 499 500 static int vpc_read(BlockDriverState *bs, int64_t sector_num, 501 uint8_t *buf, int nb_sectors) 502 { 503 BDRVVPCState *s = bs->opaque; 504 int ret; 505 int64_t offset; 506 int64_t sectors, sectors_per_block; 507 VHDFooter *footer = (VHDFooter *) s->footer_buf; 508 509 if (be32_to_cpu(footer->type) == VHD_FIXED) { 510 return bdrv_read(bs->file, sector_num, buf, nb_sectors); 511 } 512 while (nb_sectors > 0) { 513 offset = get_sector_offset(bs, sector_num, 0); 514 515 sectors_per_block = s->block_size >> BDRV_SECTOR_BITS; 516 sectors = sectors_per_block - (sector_num % sectors_per_block); 517 if (sectors > nb_sectors) { 518 sectors = nb_sectors; 519 } 520 521 if (offset == -1) { 522 memset(buf, 0, sectors * BDRV_SECTOR_SIZE); 523 } else { 524 ret = bdrv_pread(bs->file, offset, buf, 525 sectors * BDRV_SECTOR_SIZE); 526 if (ret != sectors * BDRV_SECTOR_SIZE) { 527 return -1; 528 } 529 } 530 531 nb_sectors -= sectors; 532 sector_num += sectors; 533 buf += sectors * BDRV_SECTOR_SIZE; 534 } 535 return 0; 536 } 537 538 static coroutine_fn int vpc_co_read(BlockDriverState *bs, int64_t sector_num, 539 uint8_t *buf, int nb_sectors) 540 { 541 int ret; 542 BDRVVPCState *s = bs->opaque; 543 qemu_co_mutex_lock(&s->lock); 544 ret = vpc_read(bs, sector_num, buf, nb_sectors); 545 qemu_co_mutex_unlock(&s->lock); 546 return ret; 547 } 548 549 static int vpc_write(BlockDriverState *bs, int64_t sector_num, 550 const uint8_t *buf, int nb_sectors) 551 { 552 BDRVVPCState *s = bs->opaque; 553 int64_t offset; 554 int64_t sectors, sectors_per_block; 555 int ret; 556 VHDFooter *footer = (VHDFooter *) s->footer_buf; 557 558 if (be32_to_cpu(footer->type) == VHD_FIXED) { 559 return bdrv_write(bs->file, sector_num, buf, nb_sectors); 560 } 561 while (nb_sectors > 0) { 562 offset = get_sector_offset(bs, sector_num, 1); 563 564 sectors_per_block = s->block_size >> BDRV_SECTOR_BITS; 565 sectors = sectors_per_block - (sector_num % sectors_per_block); 566 if (sectors > nb_sectors) { 567 sectors = nb_sectors; 568 } 569 570 if (offset == -1) { 571 offset = alloc_block(bs, sector_num); 572 if (offset < 0) 573 return -1; 574 } 575 576 ret = bdrv_pwrite(bs->file, offset, buf, sectors * BDRV_SECTOR_SIZE); 577 if (ret != sectors * BDRV_SECTOR_SIZE) { 578 return -1; 579 } 580 581 nb_sectors -= sectors; 582 sector_num += sectors; 583 buf += sectors * BDRV_SECTOR_SIZE; 584 } 585 586 return 0; 587 } 588 589 static coroutine_fn int vpc_co_write(BlockDriverState *bs, int64_t sector_num, 590 const uint8_t *buf, int nb_sectors) 591 { 592 int ret; 593 BDRVVPCState *s = bs->opaque; 594 qemu_co_mutex_lock(&s->lock); 595 ret = vpc_write(bs, sector_num, buf, nb_sectors); 596 qemu_co_mutex_unlock(&s->lock); 597 return ret; 598 } 599 600 static int64_t coroutine_fn vpc_co_get_block_status(BlockDriverState *bs, 601 int64_t sector_num, int nb_sectors, int *pnum) 602 { 603 BDRVVPCState *s = bs->opaque; 604 VHDFooter *footer = (VHDFooter*) s->footer_buf; 605 int64_t start, offset, next; 606 bool allocated; 607 int n; 608 609 if (be32_to_cpu(footer->type) == VHD_FIXED) { 610 *pnum = nb_sectors; 611 return BDRV_BLOCK_RAW | BDRV_BLOCK_OFFSET_VALID | BDRV_BLOCK_DATA | 612 (sector_num << BDRV_SECTOR_BITS); 613 } 614 615 offset = get_sector_offset(bs, sector_num, 0); 616 start = offset; 617 allocated = (offset != -1); 618 *pnum = 0; 619 620 do { 621 /* All sectors in a block are contiguous (without using the bitmap) */ 622 n = ROUND_UP(sector_num + 1, s->block_size / BDRV_SECTOR_SIZE) 623 - sector_num; 624 n = MIN(n, nb_sectors); 625 626 *pnum += n; 627 sector_num += n; 628 nb_sectors -= n; 629 next = start + (*pnum * BDRV_SECTOR_SIZE); 630 631 if (nb_sectors == 0) { 632 break; 633 } 634 635 offset = get_sector_offset(bs, sector_num, 0); 636 } while ((allocated && offset == next) || (!allocated && offset == -1)); 637 638 if (allocated) { 639 return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID | start; 640 } else { 641 return 0; 642 } 643 } 644 645 /* 646 * Calculates the number of cylinders, heads and sectors per cylinder 647 * based on a given number of sectors. This is the algorithm described 648 * in the VHD specification. 649 * 650 * Note that the geometry doesn't always exactly match total_sectors but 651 * may round it down. 652 * 653 * Returns 0 on success, -EFBIG if the size is larger than ~2 TB. Override 654 * the hardware EIDE and ATA-2 limit of 16 heads (max disk size of 127 GB) 655 * and instead allow up to 255 heads. 656 */ 657 static int calculate_geometry(int64_t total_sectors, uint16_t* cyls, 658 uint8_t* heads, uint8_t* secs_per_cyl) 659 { 660 uint32_t cyls_times_heads; 661 662 /* Allow a maximum disk size of approximately 2 TB */ 663 if (total_sectors > 65535LL * 255 * 255) { 664 return -EFBIG; 665 } 666 667 if (total_sectors > 65535 * 16 * 63) { 668 *secs_per_cyl = 255; 669 if (total_sectors > 65535 * 16 * 255) { 670 *heads = 255; 671 } else { 672 *heads = 16; 673 } 674 cyls_times_heads = total_sectors / *secs_per_cyl; 675 } else { 676 *secs_per_cyl = 17; 677 cyls_times_heads = total_sectors / *secs_per_cyl; 678 *heads = (cyls_times_heads + 1023) / 1024; 679 680 if (*heads < 4) 681 *heads = 4; 682 683 if (cyls_times_heads >= (*heads * 1024) || *heads > 16) { 684 *secs_per_cyl = 31; 685 *heads = 16; 686 cyls_times_heads = total_sectors / *secs_per_cyl; 687 } 688 689 if (cyls_times_heads >= (*heads * 1024)) { 690 *secs_per_cyl = 63; 691 *heads = 16; 692 cyls_times_heads = total_sectors / *secs_per_cyl; 693 } 694 } 695 696 *cyls = cyls_times_heads / *heads; 697 698 return 0; 699 } 700 701 static int create_dynamic_disk(BlockDriverState *bs, uint8_t *buf, 702 int64_t total_sectors) 703 { 704 VHDDynDiskHeader *dyndisk_header = 705 (VHDDynDiskHeader *) buf; 706 size_t block_size, num_bat_entries; 707 int i; 708 int ret; 709 int64_t offset = 0; 710 711 // Write the footer (twice: at the beginning and at the end) 712 block_size = 0x200000; 713 num_bat_entries = (total_sectors + block_size / 512) / (block_size / 512); 714 715 ret = bdrv_pwrite_sync(bs, offset, buf, HEADER_SIZE); 716 if (ret) { 717 goto fail; 718 } 719 720 offset = 1536 + ((num_bat_entries * 4 + 511) & ~511); 721 ret = bdrv_pwrite_sync(bs, offset, buf, HEADER_SIZE); 722 if (ret < 0) { 723 goto fail; 724 } 725 726 // Write the initial BAT 727 offset = 3 * 512; 728 729 memset(buf, 0xFF, 512); 730 for (i = 0; i < (num_bat_entries * 4 + 511) / 512; i++) { 731 ret = bdrv_pwrite_sync(bs, offset, buf, 512); 732 if (ret < 0) { 733 goto fail; 734 } 735 offset += 512; 736 } 737 738 // Prepare the Dynamic Disk Header 739 memset(buf, 0, 1024); 740 741 memcpy(dyndisk_header->magic, "cxsparse", 8); 742 743 /* 744 * Note: The spec is actually wrong here for data_offset, it says 745 * 0xFFFFFFFF, but MS tools expect all 64 bits to be set. 746 */ 747 dyndisk_header->data_offset = cpu_to_be64(0xFFFFFFFFFFFFFFFFULL); 748 dyndisk_header->table_offset = cpu_to_be64(3 * 512); 749 dyndisk_header->version = cpu_to_be32(0x00010000); 750 dyndisk_header->block_size = cpu_to_be32(block_size); 751 dyndisk_header->max_table_entries = cpu_to_be32(num_bat_entries); 752 753 dyndisk_header->checksum = cpu_to_be32(vpc_checksum(buf, 1024)); 754 755 // Write the header 756 offset = 512; 757 758 ret = bdrv_pwrite_sync(bs, offset, buf, 1024); 759 if (ret < 0) { 760 goto fail; 761 } 762 763 fail: 764 return ret; 765 } 766 767 static int create_fixed_disk(BlockDriverState *bs, uint8_t *buf, 768 int64_t total_size) 769 { 770 int ret; 771 772 /* Add footer to total size */ 773 total_size += HEADER_SIZE; 774 775 ret = bdrv_truncate(bs, total_size); 776 if (ret < 0) { 777 return ret; 778 } 779 780 ret = bdrv_pwrite_sync(bs, total_size - HEADER_SIZE, buf, HEADER_SIZE); 781 if (ret < 0) { 782 return ret; 783 } 784 785 return ret; 786 } 787 788 static int vpc_create(const char *filename, QemuOpts *opts, Error **errp) 789 { 790 uint8_t buf[1024]; 791 VHDFooter *footer = (VHDFooter *) buf; 792 char *disk_type_param; 793 int i; 794 uint16_t cyls = 0; 795 uint8_t heads = 0; 796 uint8_t secs_per_cyl = 0; 797 int64_t total_sectors; 798 int64_t total_size; 799 int disk_type; 800 int ret = -EIO; 801 Error *local_err = NULL; 802 BlockDriverState *bs = NULL; 803 804 /* Read out options */ 805 total_size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), 806 BDRV_SECTOR_SIZE); 807 disk_type_param = qemu_opt_get_del(opts, BLOCK_OPT_SUBFMT); 808 if (disk_type_param) { 809 if (!strcmp(disk_type_param, "dynamic")) { 810 disk_type = VHD_DYNAMIC; 811 } else if (!strcmp(disk_type_param, "fixed")) { 812 disk_type = VHD_FIXED; 813 } else { 814 ret = -EINVAL; 815 goto out; 816 } 817 } else { 818 disk_type = VHD_DYNAMIC; 819 } 820 821 ret = bdrv_create_file(filename, opts, &local_err); 822 if (ret < 0) { 823 error_propagate(errp, local_err); 824 goto out; 825 } 826 ret = bdrv_open(&bs, filename, NULL, NULL, BDRV_O_RDWR | BDRV_O_PROTOCOL, 827 NULL, &local_err); 828 if (ret < 0) { 829 error_propagate(errp, local_err); 830 goto out; 831 } 832 833 /* 834 * Calculate matching total_size and geometry. Increase the number of 835 * sectors requested until we get enough (or fail). This ensures that 836 * qemu-img convert doesn't truncate images, but rather rounds up. 837 */ 838 total_sectors = total_size / BDRV_SECTOR_SIZE; 839 for (i = 0; total_sectors > (int64_t)cyls * heads * secs_per_cyl; i++) { 840 if (calculate_geometry(total_sectors + i, &cyls, &heads, 841 &secs_per_cyl)) 842 { 843 ret = -EFBIG; 844 goto out; 845 } 846 } 847 848 total_sectors = (int64_t) cyls * heads * secs_per_cyl; 849 total_size = total_sectors * BDRV_SECTOR_SIZE; 850 851 /* Prepare the Hard Disk Footer */ 852 memset(buf, 0, 1024); 853 854 memcpy(footer->creator, "conectix", 8); 855 /* TODO Check if "qemu" creator_app is ok for VPC */ 856 memcpy(footer->creator_app, "qemu", 4); 857 memcpy(footer->creator_os, "Wi2k", 4); 858 859 footer->features = cpu_to_be32(0x02); 860 footer->version = cpu_to_be32(0x00010000); 861 if (disk_type == VHD_DYNAMIC) { 862 footer->data_offset = cpu_to_be64(HEADER_SIZE); 863 } else { 864 footer->data_offset = cpu_to_be64(0xFFFFFFFFFFFFFFFFULL); 865 } 866 footer->timestamp = cpu_to_be32(time(NULL) - VHD_TIMESTAMP_BASE); 867 868 /* Version of Virtual PC 2007 */ 869 footer->major = cpu_to_be16(0x0005); 870 footer->minor = cpu_to_be16(0x0003); 871 footer->orig_size = cpu_to_be64(total_size); 872 footer->size = cpu_to_be64(total_size); 873 footer->cyls = cpu_to_be16(cyls); 874 footer->heads = heads; 875 footer->secs_per_cyl = secs_per_cyl; 876 877 footer->type = cpu_to_be32(disk_type); 878 879 #if defined(CONFIG_UUID) 880 uuid_generate(footer->uuid); 881 #endif 882 883 footer->checksum = cpu_to_be32(vpc_checksum(buf, HEADER_SIZE)); 884 885 if (disk_type == VHD_DYNAMIC) { 886 ret = create_dynamic_disk(bs, buf, total_sectors); 887 } else { 888 ret = create_fixed_disk(bs, buf, total_size); 889 } 890 891 out: 892 bdrv_unref(bs); 893 g_free(disk_type_param); 894 return ret; 895 } 896 897 static int vpc_has_zero_init(BlockDriverState *bs) 898 { 899 BDRVVPCState *s = bs->opaque; 900 VHDFooter *footer = (VHDFooter *) s->footer_buf; 901 902 if (be32_to_cpu(footer->type) == VHD_FIXED) { 903 return bdrv_has_zero_init(bs->file); 904 } else { 905 return 1; 906 } 907 } 908 909 static void vpc_close(BlockDriverState *bs) 910 { 911 BDRVVPCState *s = bs->opaque; 912 qemu_vfree(s->pagetable); 913 #ifdef CACHE 914 g_free(s->pageentry_u8); 915 #endif 916 917 migrate_del_blocker(s->migration_blocker); 918 error_free(s->migration_blocker); 919 } 920 921 static QemuOptsList vpc_create_opts = { 922 .name = "vpc-create-opts", 923 .head = QTAILQ_HEAD_INITIALIZER(vpc_create_opts.head), 924 .desc = { 925 { 926 .name = BLOCK_OPT_SIZE, 927 .type = QEMU_OPT_SIZE, 928 .help = "Virtual disk size" 929 }, 930 { 931 .name = BLOCK_OPT_SUBFMT, 932 .type = QEMU_OPT_STRING, 933 .help = 934 "Type of virtual hard disk format. Supported formats are " 935 "{dynamic (default) | fixed} " 936 }, 937 { /* end of list */ } 938 } 939 }; 940 941 static BlockDriver bdrv_vpc = { 942 .format_name = "vpc", 943 .instance_size = sizeof(BDRVVPCState), 944 945 .bdrv_probe = vpc_probe, 946 .bdrv_open = vpc_open, 947 .bdrv_close = vpc_close, 948 .bdrv_reopen_prepare = vpc_reopen_prepare, 949 .bdrv_create = vpc_create, 950 951 .bdrv_read = vpc_co_read, 952 .bdrv_write = vpc_co_write, 953 .bdrv_co_get_block_status = vpc_co_get_block_status, 954 955 .bdrv_get_info = vpc_get_info, 956 957 .create_opts = &vpc_create_opts, 958 .bdrv_has_zero_init = vpc_has_zero_init, 959 }; 960 961 static void bdrv_vpc_init(void) 962 { 963 bdrv_register(&bdrv_vpc); 964 } 965 966 block_init(bdrv_vpc_init); 967