qemu/block/vpc.c
<<
>>
Prefs
   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
  39enum 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// always big-endian
  49struct vhd_footer {
  50    char        creator[8]; // "conectix"
  51    uint32_t    features;
  52    uint32_t    version;
  53
  54    // Offset of next header structure, 0xFFFFFFFF if none
  55    uint64_t    data_offset;
  56
  57    // Seconds since Jan 1, 2000 0:00:00 (UTC)
  58    uint32_t    timestamp;
  59
  60    char        creator_app[4]; // "vpc "
  61    uint16_t    major;
  62    uint16_t    minor;
  63    char        creator_os[4]; // "Wi2k"
  64
  65    uint64_t    orig_size;
  66    uint64_t    size;
  67
  68    uint16_t    cyls;
  69    uint8_t     heads;
  70    uint8_t     secs_per_cyl;
  71
  72    uint32_t    type;
  73
  74    // Checksum of the Hard Disk Footer ("one's complement of the sum of all
  75    // the bytes in the footer without the checksum field")
  76    uint32_t    checksum;
  77
  78    // UUID used to identify a parent hard disk (backing file)
  79    uint8_t     uuid[16];
  80
  81    uint8_t     in_saved_state;
  82};
  83
  84struct vhd_dyndisk_header {
  85    char        magic[8]; // "cxsparse"
  86
  87    // Offset of next header structure, 0xFFFFFFFF if none
  88    uint64_t    data_offset;
  89
  90    // Offset of the Block Allocation Table (BAT)
  91    uint64_t    table_offset;
  92
  93    uint32_t    version;
  94    uint32_t    max_table_entries; // 32bit/entry
  95
  96    // 2 MB by default, must be a power of two
  97    uint32_t    block_size;
  98
  99    uint32_t    checksum;
 100    uint8_t     parent_uuid[16];
 101    uint32_t    parent_timestamp;
 102    uint32_t    reserved;
 103
 104    // Backing file name (in UTF-16)
 105    uint8_t     parent_name[512];
 106
 107    struct {
 108        uint32_t    platform;
 109        uint32_t    data_space;
 110        uint32_t    data_length;
 111        uint32_t    reserved;
 112        uint64_t    data_offset;
 113    } parent_locator[8];
 114};
 115
 116typedef struct BDRVVPCState {
 117    CoMutex lock;
 118    uint8_t footer_buf[HEADER_SIZE];
 119    uint64_t free_data_block_offset;
 120    int max_table_entries;
 121    uint32_t *pagetable;
 122    uint64_t bat_offset;
 123    uint64_t last_bitmap_offset;
 124
 125    uint32_t block_size;
 126    uint32_t bitmap_size;
 127
 128#ifdef CACHE
 129    uint8_t *pageentry_u8;
 130    uint32_t *pageentry_u32;
 131    uint16_t *pageentry_u16;
 132
 133    uint64_t last_bitmap;
 134#endif
 135
 136    Error *migration_blocker;
 137} BDRVVPCState;
 138
 139static uint32_t vpc_checksum(uint8_t* buf, size_t size)
 140{
 141    uint32_t res = 0;
 142    int i;
 143
 144    for (i = 0; i < size; i++)
 145        res += buf[i];
 146
 147    return ~res;
 148}
 149
 150
 151static int vpc_probe(const uint8_t *buf, int buf_size, const char *filename)
 152{
 153    if (buf_size >= 8 && !strncmp((char *)buf, "conectix", 8))
 154        return 100;
 155    return 0;
 156}
 157
 158static int vpc_open(BlockDriverState *bs, int flags)
 159{
 160    BDRVVPCState *s = bs->opaque;
 161    int i;
 162    struct vhd_footer* footer;
 163    struct vhd_dyndisk_header* dyndisk_header;
 164    uint8_t buf[HEADER_SIZE];
 165    uint32_t checksum;
 166    int disk_type = VHD_DYNAMIC;
 167    int ret;
 168
 169    ret = bdrv_pread(bs->file, 0, s->footer_buf, HEADER_SIZE);
 170    if (ret < 0) {
 171        goto fail;
 172    }
 173
 174    footer = (struct vhd_footer*) s->footer_buf;
 175    if (strncmp(footer->creator, "conectix", 8)) {
 176        int64_t offset = bdrv_getlength(bs->file);
 177        if (offset < 0) {
 178            ret = offset;
 179            goto fail;
 180        } else if (offset < HEADER_SIZE) {
 181            ret = -EINVAL;
 182            goto fail;
 183        }
 184
 185        /* If a fixed disk, the footer is found only at the end of the file */
 186        ret = bdrv_pread(bs->file, offset-HEADER_SIZE, s->footer_buf,
 187                         HEADER_SIZE);
 188        if (ret < 0) {
 189            goto fail;
 190        }
 191        if (strncmp(footer->creator, "conectix", 8)) {
 192            ret = -EMEDIUMTYPE;
 193            goto fail;
 194        }
 195        disk_type = VHD_FIXED;
 196    }
 197
 198    checksum = be32_to_cpu(footer->checksum);
 199    footer->checksum = 0;
 200    if (vpc_checksum(s->footer_buf, HEADER_SIZE) != checksum)
 201        fprintf(stderr, "block-vpc: The header checksum of '%s' is "
 202            "incorrect.\n", bs->filename);
 203
 204    /* Write 'checksum' back to footer, or else will leave it with zero. */
 205    footer->checksum = be32_to_cpu(checksum);
 206
 207    // The visible size of a image in Virtual PC depends on the geometry
 208    // rather than on the size stored in the footer (the size in the footer
 209    // is too large usually)
 210    bs->total_sectors = (int64_t)
 211        be16_to_cpu(footer->cyls) * footer->heads * footer->secs_per_cyl;
 212
 213    /* Allow a maximum disk size of approximately 2 TB */
 214    if (bs->total_sectors >= 65535LL * 255 * 255) {
 215        ret = -EFBIG;
 216        goto fail;
 217    }
 218
 219    if (disk_type == VHD_DYNAMIC) {
 220        ret = bdrv_pread(bs->file, be64_to_cpu(footer->data_offset), buf,
 221                         HEADER_SIZE);
 222        if (ret < 0) {
 223            goto fail;
 224        }
 225
 226        dyndisk_header = (struct vhd_dyndisk_header *) buf;
 227
 228        if (strncmp(dyndisk_header->magic, "cxsparse", 8)) {
 229            ret = -EINVAL;
 230            goto fail;
 231        }
 232
 233        s->block_size = be32_to_cpu(dyndisk_header->block_size);
 234        s->bitmap_size = ((s->block_size / (8 * 512)) + 511) & ~511;
 235
 236        s->max_table_entries = be32_to_cpu(dyndisk_header->max_table_entries);
 237        s->pagetable = g_malloc(s->max_table_entries * 4);
 238
 239        s->bat_offset = be64_to_cpu(dyndisk_header->table_offset);
 240
 241        ret = bdrv_pread(bs->file, s->bat_offset, s->pagetable,
 242                         s->max_table_entries * 4);
 243        if (ret < 0) {
 244            goto fail;
 245        }
 246
 247        s->free_data_block_offset =
 248            (s->bat_offset + (s->max_table_entries * 4) + 511) & ~511;
 249
 250        for (i = 0; i < s->max_table_entries; i++) {
 251            be32_to_cpus(&s->pagetable[i]);
 252            if (s->pagetable[i] != 0xFFFFFFFF) {
 253                int64_t next = (512 * (int64_t) s->pagetable[i]) +
 254                    s->bitmap_size + s->block_size;
 255
 256                if (next > s->free_data_block_offset) {
 257                    s->free_data_block_offset = next;
 258                }
 259            }
 260        }
 261
 262        s->last_bitmap_offset = (int64_t) -1;
 263
 264#ifdef CACHE
 265        s->pageentry_u8 = g_malloc(512);
 266        s->pageentry_u32 = s->pageentry_u8;
 267        s->pageentry_u16 = s->pageentry_u8;
 268        s->last_pagetable = -1;
 269#endif
 270    }
 271
 272    qemu_co_mutex_init(&s->lock);
 273
 274    /* Disable migration when VHD images are used */
 275    error_set(&s->migration_blocker,
 276              QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
 277              "vpc", bs->device_name, "live migration");
 278    migrate_add_blocker(s->migration_blocker);
 279
 280    return 0;
 281
 282fail:
 283    g_free(s->pagetable);
 284#ifdef CACHE
 285    g_free(s->pageentry_u8);
 286#endif
 287    return ret;
 288}
 289
 290static int vpc_reopen_prepare(BDRVReopenState *state,
 291                              BlockReopenQueue *queue, Error **errp)
 292{
 293    return 0;
 294}
 295
 296/*
 297 * Returns the absolute byte offset of the given sector in the image file.
 298 * If the sector is not allocated, -1 is returned instead.
 299 *
 300 * The parameter write must be 1 if the offset will be used for a write
 301 * operation (the block bitmaps is updated then), 0 otherwise.
 302 */
 303static inline int64_t get_sector_offset(BlockDriverState *bs,
 304    int64_t sector_num, int write)
 305{
 306    BDRVVPCState *s = bs->opaque;
 307    uint64_t offset = sector_num * 512;
 308    uint64_t bitmap_offset, block_offset;
 309    uint32_t pagetable_index, pageentry_index;
 310
 311    pagetable_index = offset / s->block_size;
 312    pageentry_index = (offset % s->block_size) / 512;
 313
 314    if (pagetable_index >= s->max_table_entries || s->pagetable[pagetable_index] == 0xffffffff)
 315        return -1; // not allocated
 316
 317    bitmap_offset = 512 * (uint64_t) s->pagetable[pagetable_index];
 318    block_offset = bitmap_offset + s->bitmap_size + (512 * pageentry_index);
 319
 320    // We must ensure that we don't write to any sectors which are marked as
 321    // unused in the bitmap. We get away with setting all bits in the block
 322    // bitmap each time we write to a new block. This might cause Virtual PC to
 323    // miss sparse read optimization, but it's not a problem in terms of
 324    // correctness.
 325    if (write && (s->last_bitmap_offset != bitmap_offset)) {
 326        uint8_t bitmap[s->bitmap_size];
 327
 328        s->last_bitmap_offset = bitmap_offset;
 329        memset(bitmap, 0xff, s->bitmap_size);
 330        bdrv_pwrite_sync(bs->file, bitmap_offset, bitmap, s->bitmap_size);
 331    }
 332
 333//    printf("sector: %" PRIx64 ", index: %x, offset: %x, bioff: %" PRIx64 ", bloff: %" PRIx64 "\n",
 334//      sector_num, pagetable_index, pageentry_index,
 335//      bitmap_offset, block_offset);
 336
 337// disabled by reason
 338#if 0
 339#ifdef CACHE
 340    if (bitmap_offset != s->last_bitmap)
 341    {
 342        lseek(s->fd, bitmap_offset, SEEK_SET);
 343
 344        s->last_bitmap = bitmap_offset;
 345
 346        // Scary! Bitmap is stored as big endian 32bit entries,
 347        // while we used to look it up byte by byte
 348        read(s->fd, s->pageentry_u8, 512);
 349        for (i = 0; i < 128; i++)
 350            be32_to_cpus(&s->pageentry_u32[i]);
 351    }
 352
 353    if ((s->pageentry_u8[pageentry_index / 8] >> (pageentry_index % 8)) & 1)
 354        return -1;
 355#else
 356    lseek(s->fd, bitmap_offset + (pageentry_index / 8), SEEK_SET);
 357
 358    read(s->fd, &bitmap_entry, 1);
 359
 360    if ((bitmap_entry >> (pageentry_index % 8)) & 1)
 361        return -1; // not allocated
 362#endif
 363#endif
 364
 365    return block_offset;
 366}
 367
 368/*
 369 * Writes the footer to the end of the image file. This is needed when the
 370 * file grows as it overwrites the old footer
 371 *
 372 * Returns 0 on success and < 0 on error
 373 */
 374static int rewrite_footer(BlockDriverState* bs)
 375{
 376    int ret;
 377    BDRVVPCState *s = bs->opaque;
 378    int64_t offset = s->free_data_block_offset;
 379
 380    ret = bdrv_pwrite_sync(bs->file, offset, s->footer_buf, HEADER_SIZE);
 381    if (ret < 0)
 382        return ret;
 383
 384    return 0;
 385}
 386
 387/*
 388 * Allocates a new block. This involves writing a new footer and updating
 389 * the Block Allocation Table to use the space at the old end of the image
 390 * file (overwriting the old footer)
 391 *
 392 * Returns the sectors' offset in the image file on success and < 0 on error
 393 */
 394static int64_t alloc_block(BlockDriverState* bs, int64_t sector_num)
 395{
 396    BDRVVPCState *s = bs->opaque;
 397    int64_t bat_offset;
 398    uint32_t index, bat_value;
 399    int ret;
 400    uint8_t bitmap[s->bitmap_size];
 401
 402    // Check if sector_num is valid
 403    if ((sector_num < 0) || (sector_num > bs->total_sectors))
 404        return -1;
 405
 406    // Write entry into in-memory BAT
 407    index = (sector_num * 512) / s->block_size;
 408    if (s->pagetable[index] != 0xFFFFFFFF)
 409        return -1;
 410
 411    s->pagetable[index] = s->free_data_block_offset / 512;
 412
 413    // Initialize the block's bitmap
 414    memset(bitmap, 0xff, s->bitmap_size);
 415    ret = bdrv_pwrite_sync(bs->file, s->free_data_block_offset, bitmap,
 416        s->bitmap_size);
 417    if (ret < 0) {
 418        return ret;
 419    }
 420
 421    // Write new footer (the old one will be overwritten)
 422    s->free_data_block_offset += s->block_size + s->bitmap_size;
 423    ret = rewrite_footer(bs);
 424    if (ret < 0)
 425        goto fail;
 426
 427    // Write BAT entry to disk
 428    bat_offset = s->bat_offset + (4 * index);
 429    bat_value = be32_to_cpu(s->pagetable[index]);
 430    ret = bdrv_pwrite_sync(bs->file, bat_offset, &bat_value, 4);
 431    if (ret < 0)
 432        goto fail;
 433
 434    return get_sector_offset(bs, sector_num, 0);
 435
 436fail:
 437    s->free_data_block_offset -= (s->block_size + s->bitmap_size);
 438    return -1;
 439}
 440
 441static int vpc_read(BlockDriverState *bs, int64_t sector_num,
 442                    uint8_t *buf, int nb_sectors)
 443{
 444    BDRVVPCState *s = bs->opaque;
 445    int ret;
 446    int64_t offset;
 447    int64_t sectors, sectors_per_block;
 448    struct vhd_footer *footer = (struct vhd_footer *) s->footer_buf;
 449
 450    if (cpu_to_be32(footer->type) == VHD_FIXED) {
 451        return bdrv_read(bs->file, sector_num, buf, nb_sectors);
 452    }
 453    while (nb_sectors > 0) {
 454        offset = get_sector_offset(bs, sector_num, 0);
 455
 456        sectors_per_block = s->block_size >> BDRV_SECTOR_BITS;
 457        sectors = sectors_per_block - (sector_num % sectors_per_block);
 458        if (sectors > nb_sectors) {
 459            sectors = nb_sectors;
 460        }
 461
 462        if (offset == -1) {
 463            memset(buf, 0, sectors * BDRV_SECTOR_SIZE);
 464        } else {
 465            ret = bdrv_pread(bs->file, offset, buf,
 466                sectors * BDRV_SECTOR_SIZE);
 467            if (ret != sectors * BDRV_SECTOR_SIZE) {
 468                return -1;
 469            }
 470        }
 471
 472        nb_sectors -= sectors;
 473        sector_num += sectors;
 474        buf += sectors * BDRV_SECTOR_SIZE;
 475    }
 476    return 0;
 477}
 478
 479static coroutine_fn int vpc_co_read(BlockDriverState *bs, int64_t sector_num,
 480                                    uint8_t *buf, int nb_sectors)
 481{
 482    int ret;
 483    BDRVVPCState *s = bs->opaque;
 484    qemu_co_mutex_lock(&s->lock);
 485    ret = vpc_read(bs, sector_num, buf, nb_sectors);
 486    qemu_co_mutex_unlock(&s->lock);
 487    return ret;
 488}
 489
 490static int vpc_write(BlockDriverState *bs, int64_t sector_num,
 491    const uint8_t *buf, int nb_sectors)
 492{
 493    BDRVVPCState *s = bs->opaque;
 494    int64_t offset;
 495    int64_t sectors, sectors_per_block;
 496    int ret;
 497    struct vhd_footer *footer =  (struct vhd_footer *) s->footer_buf;
 498
 499    if (cpu_to_be32(footer->type) == VHD_FIXED) {
 500        return bdrv_write(bs->file, sector_num, buf, nb_sectors);
 501    }
 502    while (nb_sectors > 0) {
 503        offset = get_sector_offset(bs, sector_num, 1);
 504
 505        sectors_per_block = s->block_size >> BDRV_SECTOR_BITS;
 506        sectors = sectors_per_block - (sector_num % sectors_per_block);
 507        if (sectors > nb_sectors) {
 508            sectors = nb_sectors;
 509        }
 510
 511        if (offset == -1) {
 512            offset = alloc_block(bs, sector_num);
 513            if (offset < 0)
 514                return -1;
 515        }
 516
 517        ret = bdrv_pwrite(bs->file, offset, buf, sectors * BDRV_SECTOR_SIZE);
 518        if (ret != sectors * BDRV_SECTOR_SIZE) {
 519            return -1;
 520        }
 521
 522        nb_sectors -= sectors;
 523        sector_num += sectors;
 524        buf += sectors * BDRV_SECTOR_SIZE;
 525    }
 526
 527    return 0;
 528}
 529
 530static coroutine_fn int vpc_co_write(BlockDriverState *bs, int64_t sector_num,
 531                                     const uint8_t *buf, int nb_sectors)
 532{
 533    int ret;
 534    BDRVVPCState *s = bs->opaque;
 535    qemu_co_mutex_lock(&s->lock);
 536    ret = vpc_write(bs, sector_num, buf, nb_sectors);
 537    qemu_co_mutex_unlock(&s->lock);
 538    return ret;
 539}
 540
 541/*
 542 * Calculates the number of cylinders, heads and sectors per cylinder
 543 * based on a given number of sectors. This is the algorithm described
 544 * in the VHD specification.
 545 *
 546 * Note that the geometry doesn't always exactly match total_sectors but
 547 * may round it down.
 548 *
 549 * Returns 0 on success, -EFBIG if the size is larger than ~2 TB. Override
 550 * the hardware EIDE and ATA-2 limit of 16 heads (max disk size of 127 GB)
 551 * and instead allow up to 255 heads.
 552 */
 553static int calculate_geometry(int64_t total_sectors, uint16_t* cyls,
 554    uint8_t* heads, uint8_t* secs_per_cyl)
 555{
 556    uint32_t cyls_times_heads;
 557
 558    /* Allow a maximum disk size of approximately 2 TB */
 559    if (total_sectors > 65535LL * 255 * 255) {
 560        return -EFBIG;
 561    }
 562
 563    if (total_sectors > 65535 * 16 * 63) {
 564        *secs_per_cyl = 255;
 565        if (total_sectors > 65535 * 16 * 255) {
 566            *heads = 255;
 567        } else {
 568            *heads = 16;
 569        }
 570        cyls_times_heads = total_sectors / *secs_per_cyl;
 571    } else {
 572        *secs_per_cyl = 17;
 573        cyls_times_heads = total_sectors / *secs_per_cyl;
 574        *heads = (cyls_times_heads + 1023) / 1024;
 575
 576        if (*heads < 4)
 577            *heads = 4;
 578
 579        if (cyls_times_heads >= (*heads * 1024) || *heads > 16) {
 580            *secs_per_cyl = 31;
 581            *heads = 16;
 582            cyls_times_heads = total_sectors / *secs_per_cyl;
 583        }
 584
 585        if (cyls_times_heads >= (*heads * 1024)) {
 586            *secs_per_cyl = 63;
 587            *heads = 16;
 588            cyls_times_heads = total_sectors / *secs_per_cyl;
 589        }
 590    }
 591
 592    *cyls = cyls_times_heads / *heads;
 593
 594    return 0;
 595}
 596
 597static int create_dynamic_disk(int fd, uint8_t *buf, int64_t total_sectors)
 598{
 599    struct vhd_dyndisk_header* dyndisk_header =
 600        (struct vhd_dyndisk_header*) buf;
 601    size_t block_size, num_bat_entries;
 602    int i;
 603    int ret = -EIO;
 604
 605    // Write the footer (twice: at the beginning and at the end)
 606    block_size = 0x200000;
 607    num_bat_entries = (total_sectors + block_size / 512) / (block_size / 512);
 608
 609    if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE) {
 610        goto fail;
 611    }
 612
 613    if (lseek(fd, 1536 + ((num_bat_entries * 4 + 511) & ~511), SEEK_SET) < 0) {
 614        goto fail;
 615    }
 616    if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE) {
 617        goto fail;
 618    }
 619
 620    // Write the initial BAT
 621    if (lseek(fd, 3 * 512, SEEK_SET) < 0) {
 622        goto fail;
 623    }
 624
 625    memset(buf, 0xFF, 512);
 626    for (i = 0; i < (num_bat_entries * 4 + 511) / 512; i++) {
 627        if (write(fd, buf, 512) != 512) {
 628            goto fail;
 629        }
 630    }
 631
 632    // Prepare the Dynamic Disk Header
 633    memset(buf, 0, 1024);
 634
 635    memcpy(dyndisk_header->magic, "cxsparse", 8);
 636
 637    /*
 638     * Note: The spec is actually wrong here for data_offset, it says
 639     * 0xFFFFFFFF, but MS tools expect all 64 bits to be set.
 640     */
 641    dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFFFFFFFFFFULL);
 642    dyndisk_header->table_offset = be64_to_cpu(3 * 512);
 643    dyndisk_header->version = be32_to_cpu(0x00010000);
 644    dyndisk_header->block_size = be32_to_cpu(block_size);
 645    dyndisk_header->max_table_entries = be32_to_cpu(num_bat_entries);
 646
 647    dyndisk_header->checksum = be32_to_cpu(vpc_checksum(buf, 1024));
 648
 649    // Write the header
 650    if (lseek(fd, 512, SEEK_SET) < 0) {
 651        goto fail;
 652    }
 653
 654    if (write(fd, buf, 1024) != 1024) {
 655        goto fail;
 656    }
 657    ret = 0;
 658
 659 fail:
 660    return ret;
 661}
 662
 663static int create_fixed_disk(int fd, uint8_t *buf, int64_t total_size)
 664{
 665    int ret = -EIO;
 666
 667    /* Add footer to total size */
 668    total_size += 512;
 669    if (ftruncate(fd, total_size) != 0) {
 670        ret = -errno;
 671        goto fail;
 672    }
 673    if (lseek(fd, -512, SEEK_END) < 0) {
 674        goto fail;
 675    }
 676    if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE) {
 677        goto fail;
 678    }
 679
 680    ret = 0;
 681
 682 fail:
 683    return ret;
 684}
 685
 686static int vpc_create(const char *filename, QEMUOptionParameter *options)
 687{
 688    uint8_t buf[1024];
 689    struct vhd_footer *footer = (struct vhd_footer *) buf;
 690    QEMUOptionParameter *disk_type_param;
 691    int fd, i;
 692    uint16_t cyls = 0;
 693    uint8_t heads = 0;
 694    uint8_t secs_per_cyl = 0;
 695    int64_t total_sectors;
 696    int64_t total_size;
 697    int disk_type;
 698    int ret = -EIO;
 699
 700    /* Read out options */
 701    total_size = get_option_parameter(options, BLOCK_OPT_SIZE)->value.n;
 702
 703    disk_type_param = get_option_parameter(options, BLOCK_OPT_SUBFMT);
 704    if (disk_type_param && disk_type_param->value.s) {
 705        if (!strcmp(disk_type_param->value.s, "dynamic")) {
 706            disk_type = VHD_DYNAMIC;
 707        } else if (!strcmp(disk_type_param->value.s, "fixed")) {
 708            disk_type = VHD_FIXED;
 709        } else {
 710            return -EINVAL;
 711        }
 712    } else {
 713        disk_type = VHD_DYNAMIC;
 714    }
 715
 716    /* Create the file */
 717    fd = qemu_open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0644);
 718    if (fd < 0) {
 719        return -EIO;
 720    }
 721
 722    /*
 723     * Calculate matching total_size and geometry. Increase the number of
 724     * sectors requested until we get enough (or fail). This ensures that
 725     * qemu-img convert doesn't truncate images, but rather rounds up.
 726     */
 727    total_sectors = total_size / BDRV_SECTOR_SIZE;
 728    for (i = 0; total_sectors > (int64_t)cyls * heads * secs_per_cyl; i++) {
 729        if (calculate_geometry(total_sectors + i, &cyls, &heads,
 730                               &secs_per_cyl))
 731        {
 732            ret = -EFBIG;
 733            goto fail;
 734        }
 735    }
 736
 737    total_sectors = (int64_t) cyls * heads * secs_per_cyl;
 738
 739    /* Prepare the Hard Disk Footer */
 740    memset(buf, 0, 1024);
 741
 742    memcpy(footer->creator, "conectix", 8);
 743    /* TODO Check if "qemu" creator_app is ok for VPC */
 744    memcpy(footer->creator_app, "qemu", 4);
 745    memcpy(footer->creator_os, "Wi2k", 4);
 746
 747    footer->features = be32_to_cpu(0x02);
 748    footer->version = be32_to_cpu(0x00010000);
 749    if (disk_type == VHD_DYNAMIC) {
 750        footer->data_offset = be64_to_cpu(HEADER_SIZE);
 751    } else {
 752        footer->data_offset = be64_to_cpu(0xFFFFFFFFFFFFFFFFULL);
 753    }
 754    footer->timestamp = be32_to_cpu(time(NULL) - VHD_TIMESTAMP_BASE);
 755
 756    /* Version of Virtual PC 2007 */
 757    footer->major = be16_to_cpu(0x0005);
 758    footer->minor = be16_to_cpu(0x0003);
 759    if (disk_type == VHD_DYNAMIC) {
 760        footer->orig_size = be64_to_cpu(total_sectors * 512);
 761        footer->size = be64_to_cpu(total_sectors * 512);
 762    } else {
 763        footer->orig_size = be64_to_cpu(total_size);
 764        footer->size = be64_to_cpu(total_size);
 765    }
 766    footer->cyls = be16_to_cpu(cyls);
 767    footer->heads = heads;
 768    footer->secs_per_cyl = secs_per_cyl;
 769
 770    footer->type = be32_to_cpu(disk_type);
 771
 772#if defined(CONFIG_UUID)
 773    uuid_generate(footer->uuid);
 774#endif
 775
 776    footer->checksum = be32_to_cpu(vpc_checksum(buf, HEADER_SIZE));
 777
 778    if (disk_type == VHD_DYNAMIC) {
 779        ret = create_dynamic_disk(fd, buf, total_sectors);
 780    } else {
 781        ret = create_fixed_disk(fd, buf, total_size);
 782    }
 783
 784 fail:
 785    qemu_close(fd);
 786    return ret;
 787}
 788
 789static void vpc_close(BlockDriverState *bs)
 790{
 791    BDRVVPCState *s = bs->opaque;
 792    g_free(s->pagetable);
 793#ifdef CACHE
 794    g_free(s->pageentry_u8);
 795#endif
 796
 797    migrate_del_blocker(s->migration_blocker);
 798    error_free(s->migration_blocker);
 799}
 800
 801static QEMUOptionParameter vpc_create_options[] = {
 802    {
 803        .name = BLOCK_OPT_SIZE,
 804        .type = OPT_SIZE,
 805        .help = "Virtual disk size"
 806    },
 807    {
 808        .name = BLOCK_OPT_SUBFMT,
 809        .type = OPT_STRING,
 810        .help =
 811            "Type of virtual hard disk format. Supported formats are "
 812            "{dynamic (default) | fixed} "
 813    },
 814    { NULL }
 815};
 816
 817static BlockDriver bdrv_vpc = {
 818    .format_name    = "vpc",
 819    .instance_size  = sizeof(BDRVVPCState),
 820
 821    .bdrv_probe     = vpc_probe,
 822    .bdrv_open      = vpc_open,
 823    .bdrv_close     = vpc_close,
 824    .bdrv_reopen_prepare = vpc_reopen_prepare,
 825    .bdrv_create    = vpc_create,
 826
 827    .bdrv_read              = vpc_co_read,
 828    .bdrv_write             = vpc_co_write,
 829
 830    .create_options = vpc_create_options,
 831};
 832
 833static void bdrv_vpc_init(void)
 834{
 835    bdrv_register(&bdrv_vpc);
 836}
 837
 838block_init(bdrv_vpc_init);
 839