vpc.c 26.8 KB
Newer Older
1
/*
S
Stefan Weil 已提交
2
 * Block driver for Connectix / Microsoft Virtual PC images
3
 *
4
 * Copyright (c) 2005 Alex Beregszaszi
A
aliguori 已提交
5
 * Copyright (c) 2009 Kevin Wolf <kwolf@suse.de>
6
 *
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this software and associated documentation files (the "Software"), to deal
 * in the Software without restriction, including without limitation the rights
 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
 * copies of the Software, and to permit persons to whom the Software is
 * furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
 * THE SOFTWARE.
 */
P
pbrook 已提交
25
#include "qemu-common.h"
26
#include "block/block_int.h"
27
#include "qemu/module.h"
28
#include "migration/migration.h"
29 30 31
#if defined(CONFIG_UUID)
#include <uuid/uuid.h>
#endif
32 33 34 35 36 37 38

/**************************************************************/

#define HEADER_SIZE 512

//#define CACHE

39 40 41 42 43 44
enum vhd_type {
    VHD_FIXED           = 2,
    VHD_DYNAMIC         = 3,
    VHD_DIFFERENCING    = 4,
};

A
aliguori 已提交
45 46 47
// Seconds since Jan 1, 2000 0:00:00 (UTC)
#define VHD_TIMESTAMP_BASE 946684800

48
#define VHD_MAX_SECTORS       (65535LL * 255 * 255)
49
#define VHD_MAX_GEOMETRY      (65535LL *  16 * 255)
50

51
// always big-endian
52
typedef struct vhd_footer {
53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68
    char        creator[8]; // "conectix"
    uint32_t    features;
    uint32_t    version;

    // Offset of next header structure, 0xFFFFFFFF if none
    uint64_t    data_offset;

    // Seconds since Jan 1, 2000 0:00:00 (UTC)
    uint32_t    timestamp;

    char        creator_app[4]; // "vpc "
    uint16_t    major;
    uint16_t    minor;
    char        creator_os[4]; // "Wi2k"

    uint64_t    orig_size;
69
    uint64_t    current_size;
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84

    uint16_t    cyls;
    uint8_t     heads;
    uint8_t     secs_per_cyl;

    uint32_t    type;

    // Checksum of the Hard Disk Footer ("one's complement of the sum of all
    // the bytes in the footer without the checksum field")
    uint32_t    checksum;

    // UUID used to identify a parent hard disk (backing file)
    uint8_t     uuid[16];

    uint8_t     in_saved_state;
85
} QEMU_PACKED VHDFooter;
86

87
typedef struct vhd_dyndisk_header {
88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116
    char        magic[8]; // "cxsparse"

    // Offset of next header structure, 0xFFFFFFFF if none
    uint64_t    data_offset;

    // Offset of the Block Allocation Table (BAT)
    uint64_t    table_offset;

    uint32_t    version;
    uint32_t    max_table_entries; // 32bit/entry

    // 2 MB by default, must be a power of two
    uint32_t    block_size;

    uint32_t    checksum;
    uint8_t     parent_uuid[16];
    uint32_t    parent_timestamp;
    uint32_t    reserved;

    // Backing file name (in UTF-16)
    uint8_t     parent_name[512];

    struct {
        uint32_t    platform;
        uint32_t    data_space;
        uint32_t    data_length;
        uint32_t    reserved;
        uint64_t    data_offset;
    } parent_locator[8];
117
} QEMU_PACKED VHDDynDiskHeader;
118 119

typedef struct BDRVVPCState {
120
    CoMutex lock;
A
aliguori 已提交
121 122
    uint8_t footer_buf[HEADER_SIZE];
    uint64_t free_data_block_offset;
123
    int max_table_entries;
124
    uint32_t *pagetable;
A
aliguori 已提交
125 126
    uint64_t bat_offset;
    uint64_t last_bitmap_offset;
127

128
    uint32_t block_size;
A
aliguori 已提交
129 130
    uint32_t bitmap_size;

131 132 133 134
#ifdef CACHE
    uint8_t *pageentry_u8;
    uint32_t *pageentry_u32;
    uint16_t *pageentry_u16;
135

136 137
    uint64_t last_bitmap;
#endif
K
Kevin Wolf 已提交
138 139

    Error *migration_blocker;
140 141
} BDRVVPCState;

A
aliguori 已提交
142 143 144 145 146 147 148 149 150 151 152 153
static uint32_t vpc_checksum(uint8_t* buf, size_t size)
{
    uint32_t res = 0;
    int i;

    for (i = 0; i < size; i++)
        res += buf[i];

    return ~res;
}


154 155
static int vpc_probe(const uint8_t *buf, int buf_size, const char *filename)
{
T
ths 已提交
156
    if (buf_size >= 8 && !strncmp((char *)buf, "conectix", 8))
157 158 159 160
	return 100;
    return 0;
}

M
Max Reitz 已提交
161 162
static int vpc_open(BlockDriverState *bs, QDict *options, int flags,
                    Error **errp)
163 164
{
    BDRVVPCState *s = bs->opaque;
165
    int i;
166 167
    VHDFooter *footer;
    VHDDynDiskHeader *dyndisk_header;
168
    uint8_t buf[HEADER_SIZE];
A
aliguori 已提交
169
    uint32_t checksum;
170
    uint64_t computed_size;
171
    int disk_type = VHD_DYNAMIC;
K
Kevin Wolf 已提交
172
    int ret;
173

K
Kevin Wolf 已提交
174 175
    ret = bdrv_pread(bs->file, 0, s->footer_buf, HEADER_SIZE);
    if (ret < 0) {
176
        goto fail;
K
Kevin Wolf 已提交
177
    }
178

179
    footer = (VHDFooter *) s->footer_buf;
180 181
    if (strncmp(footer->creator, "conectix", 8)) {
        int64_t offset = bdrv_getlength(bs->file);
K
Kevin Wolf 已提交
182 183 184 185 186
        if (offset < 0) {
            ret = offset;
            goto fail;
        } else if (offset < HEADER_SIZE) {
            ret = -EINVAL;
187 188
            goto fail;
        }
K
Kevin Wolf 已提交
189

190
        /* If a fixed disk, the footer is found only at the end of the file */
K
Kevin Wolf 已提交
191 192 193
        ret = bdrv_pread(bs->file, offset-HEADER_SIZE, s->footer_buf,
                         HEADER_SIZE);
        if (ret < 0) {
194 195 196
            goto fail;
        }
        if (strncmp(footer->creator, "conectix", 8)) {
P
Paolo Bonzini 已提交
197 198
            error_setg(errp, "invalid VPC image");
            ret = -EINVAL;
199 200 201 202
            goto fail;
        }
        disk_type = VHD_FIXED;
    }
203

A
aliguori 已提交
204 205 206 207
    checksum = be32_to_cpu(footer->checksum);
    footer->checksum = 0;
    if (vpc_checksum(s->footer_buf, HEADER_SIZE) != checksum)
        fprintf(stderr, "block-vpc: The header checksum of '%s' is "
208
            "incorrect.\n", bs->filename);
A
aliguori 已提交
209

210
    /* Write 'checksum' back to footer, or else will leave it with zero. */
211
    footer->checksum = cpu_to_be32(checksum);
212

213 214 215 216 217
    // The visible size of a image in Virtual PC depends on the geometry
    // rather than on the size stored in the footer (the size in the footer
    // is too large usually)
    bs->total_sectors = (int64_t)
        be16_to_cpu(footer->cyls) * footer->heads * footer->secs_per_cyl;
A
aliguori 已提交
218

219 220
    /* Images that have exactly the maximum geometry are probably bigger and
     * would be truncated if we adhered to the geometry for them. Rely on
221
     * footer->current_size for them. */
222
    if (bs->total_sectors == VHD_MAX_GEOMETRY) {
223 224
        bs->total_sectors = be64_to_cpu(footer->current_size) /
                            BDRV_SECTOR_SIZE;
225 226
    }

227
    /* Allow a maximum disk size of approximately 2 TB */
228
    if (bs->total_sectors >= VHD_MAX_SECTORS) {
K
Kevin Wolf 已提交
229
        ret = -EFBIG;
230 231 232
        goto fail;
    }

233
    if (disk_type == VHD_DYNAMIC) {
K
Kevin Wolf 已提交
234 235 236
        ret = bdrv_pread(bs->file, be64_to_cpu(footer->data_offset), buf,
                         HEADER_SIZE);
        if (ret < 0) {
237 238
            goto fail;
        }
239

240
        dyndisk_header = (VHDDynDiskHeader *) buf;
241

242
        if (strncmp(dyndisk_header->magic, "cxsparse", 8)) {
K
Kevin Wolf 已提交
243
            ret = -EINVAL;
244 245
            goto fail;
        }
246

247
        s->block_size = be32_to_cpu(dyndisk_header->block_size);
248 249 250 251 252
        if (!is_power_of_2(s->block_size) || s->block_size < BDRV_SECTOR_SIZE) {
            error_setg(errp, "Invalid block size %" PRIu32, s->block_size);
            ret = -EINVAL;
            goto fail;
        }
253
        s->bitmap_size = ((s->block_size / (8 * 512)) + 511) & ~511;
A
aliguori 已提交
254

255
        s->max_table_entries = be32_to_cpu(dyndisk_header->max_table_entries);
256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271

        if ((bs->total_sectors * 512) / s->block_size > 0xffffffffU) {
            ret = -EINVAL;
            goto fail;
        }
        if (s->max_table_entries > (VHD_MAX_SECTORS * 512) / s->block_size) {
            ret = -EINVAL;
            goto fail;
        }

        computed_size = (uint64_t) s->max_table_entries * s->block_size;
        if (computed_size < bs->total_sectors * 512) {
            ret = -EINVAL;
            goto fail;
        }

272 273 274 275 276
        s->pagetable = qemu_try_blockalign(bs->file, s->max_table_entries * 4);
        if (s->pagetable == NULL) {
            ret = -ENOMEM;
            goto fail;
        }
277

278
        s->bat_offset = be64_to_cpu(dyndisk_header->table_offset);
K
Kevin Wolf 已提交
279 280 281 282

        ret = bdrv_pread(bs->file, s->bat_offset, s->pagetable,
                         s->max_table_entries * 4);
        if (ret < 0) {
283 284
            goto fail;
        }
285

286 287
        s->free_data_block_offset =
            (s->bat_offset + (s->max_table_entries * 4) + 511) & ~511;
A
aliguori 已提交
288

289 290 291 292 293
        for (i = 0; i < s->max_table_entries; i++) {
            be32_to_cpus(&s->pagetable[i]);
            if (s->pagetable[i] != 0xFFFFFFFF) {
                int64_t next = (512 * (int64_t) s->pagetable[i]) +
                    s->bitmap_size + s->block_size;
A
aliguori 已提交
294

295 296 297 298
                if (next > s->free_data_block_offset) {
                    s->free_data_block_offset = next;
                }
            }
A
aliguori 已提交
299 300
        }

301 302 303 304 305 306 307
        if (s->free_data_block_offset > bdrv_getlength(bs->file)) {
            error_setg(errp, "block-vpc: free_data_block_offset points after "
                             "the end of file. The image has been truncated.");
            ret = -EINVAL;
            goto fail;
        }

308
        s->last_bitmap_offset = (int64_t) -1;
309 310

#ifdef CACHE
311 312 313 314
        s->pageentry_u8 = g_malloc(512);
        s->pageentry_u32 = s->pageentry_u8;
        s->pageentry_u16 = s->pageentry_u8;
        s->last_pagetable = -1;
315
#endif
316
    }
317

318
    qemu_co_mutex_init(&s->lock);
K
Kevin Wolf 已提交
319 320

    /* Disable migration when VHD images are used */
321 322 323
    error_setg(&s->migration_blocker, "The vpc format used by node '%s' "
               "does not support live migration",
               bdrv_get_device_or_node_name(bs));
K
Kevin Wolf 已提交
324 325
    migrate_add_blocker(s->migration_blocker);

326
    return 0;
K
Kevin Wolf 已提交
327 328

fail:
329
    qemu_vfree(s->pagetable);
K
Kevin Wolf 已提交
330 331 332 333
#ifdef CACHE
    g_free(s->pageentry_u8);
#endif
    return ret;
334 335
}

J
Jeff Cody 已提交
336 337 338 339 340 341
static int vpc_reopen_prepare(BDRVReopenState *state,
                              BlockReopenQueue *queue, Error **errp)
{
    return 0;
}

342 343 344
/*
 * Returns the absolute byte offset of the given sector in the image file.
 * If the sector is not allocated, -1 is returned instead.
A
aliguori 已提交
345 346 347
 *
 * The parameter write must be 1 if the offset will be used for a write
 * operation (the block bitmaps is updated then), 0 otherwise.
348
 */
A
aliguori 已提交
349 350
static inline int64_t get_sector_offset(BlockDriverState *bs,
    int64_t sector_num, int write)
351 352 353 354 355 356
{
    BDRVVPCState *s = bs->opaque;
    uint64_t offset = sector_num * 512;
    uint64_t bitmap_offset, block_offset;
    uint32_t pagetable_index, pageentry_index;

357 358
    pagetable_index = offset / s->block_size;
    pageentry_index = (offset % s->block_size) / 512;
359

A
aliguori 已提交
360 361
    if (pagetable_index >= s->max_table_entries || s->pagetable[pagetable_index] == 0xffffffff)
        return -1; // not allocated
362

363
    bitmap_offset = 512 * (uint64_t) s->pagetable[pagetable_index];
A
aliguori 已提交
364 365 366 367 368 369 370 371 372 373 374 375
    block_offset = bitmap_offset + s->bitmap_size + (512 * pageentry_index);

    // We must ensure that we don't write to any sectors which are marked as
    // unused in the bitmap. We get away with setting all bits in the block
    // bitmap each time we write to a new block. This might cause Virtual PC to
    // miss sparse read optimization, but it's not a problem in terms of
    // correctness.
    if (write && (s->last_bitmap_offset != bitmap_offset)) {
        uint8_t bitmap[s->bitmap_size];

        s->last_bitmap_offset = bitmap_offset;
        memset(bitmap, 0xff, s->bitmap_size);
376
        bdrv_pwrite_sync(bs->file, bitmap_offset, bitmap, s->bitmap_size);
A
aliguori 已提交
377
    }
378

379
    return block_offset;
380 381
}

A
aliguori 已提交
382 383 384 385 386 387 388 389 390 391 392 393
/*
 * Writes the footer to the end of the image file. This is needed when the
 * file grows as it overwrites the old footer
 *
 * Returns 0 on success and < 0 on error
 */
static int rewrite_footer(BlockDriverState* bs)
{
    int ret;
    BDRVVPCState *s = bs->opaque;
    int64_t offset = s->free_data_block_offset;

394
    ret = bdrv_pwrite_sync(bs->file, offset, s->footer_buf, HEADER_SIZE);
A
aliguori 已提交
395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428
    if (ret < 0)
        return ret;

    return 0;
}

/*
 * Allocates a new block. This involves writing a new footer and updating
 * the Block Allocation Table to use the space at the old end of the image
 * file (overwriting the old footer)
 *
 * Returns the sectors' offset in the image file on success and < 0 on error
 */
static int64_t alloc_block(BlockDriverState* bs, int64_t sector_num)
{
    BDRVVPCState *s = bs->opaque;
    int64_t bat_offset;
    uint32_t index, bat_value;
    int ret;
    uint8_t bitmap[s->bitmap_size];

    // Check if sector_num is valid
    if ((sector_num < 0) || (sector_num > bs->total_sectors))
        return -1;

    // Write entry into in-memory BAT
    index = (sector_num * 512) / s->block_size;
    if (s->pagetable[index] != 0xFFFFFFFF)
        return -1;

    s->pagetable[index] = s->free_data_block_offset / 512;

    // Initialize the block's bitmap
    memset(bitmap, 0xff, s->bitmap_size);
429
    ret = bdrv_pwrite_sync(bs->file, s->free_data_block_offset, bitmap,
430
        s->bitmap_size);
431 432 433
    if (ret < 0) {
        return ret;
    }
A
aliguori 已提交
434 435 436 437 438 439 440 441 442

    // Write new footer (the old one will be overwritten)
    s->free_data_block_offset += s->block_size + s->bitmap_size;
    ret = rewrite_footer(bs);
    if (ret < 0)
        goto fail;

    // Write BAT entry to disk
    bat_offset = s->bat_offset + (4 * index);
443
    bat_value = cpu_to_be32(s->pagetable[index]);
444
    ret = bdrv_pwrite_sync(bs->file, bat_offset, &bat_value, 4);
A
aliguori 已提交
445 446 447 448 449 450 451 452 453 454
    if (ret < 0)
        goto fail;

    return get_sector_offset(bs, sector_num, 0);

fail:
    s->free_data_block_offset -= (s->block_size + s->bitmap_size);
    return -1;
}

P
Paolo Bonzini 已提交
455 456 457 458 459
static int vpc_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
{
    BDRVVPCState *s = (BDRVVPCState *)bs->opaque;
    VHDFooter *footer = (VHDFooter *) s->footer_buf;

460
    if (be32_to_cpu(footer->type) != VHD_FIXED) {
P
Paolo Bonzini 已提交
461 462 463
        bdi->cluster_size = s->block_size;
    }

464
    bdi->unallocated_blocks_are_zero = true;
P
Paolo Bonzini 已提交
465 466 467
    return 0;
}

468
static int vpc_read(BlockDriverState *bs, int64_t sector_num,
469 470
                    uint8_t *buf, int nb_sectors)
{
471
    BDRVVPCState *s = bs->opaque;
472
    int ret;
473
    int64_t offset;
474
    int64_t sectors, sectors_per_block;
475
    VHDFooter *footer = (VHDFooter *) s->footer_buf;
476

477
    if (be32_to_cpu(footer->type) == VHD_FIXED) {
478 479
        return bdrv_read(bs->file, sector_num, buf, nb_sectors);
    }
480
    while (nb_sectors > 0) {
A
aliguori 已提交
481
        offset = get_sector_offset(bs, sector_num, 0);
482

483 484 485 486 487 488
        sectors_per_block = s->block_size >> BDRV_SECTOR_BITS;
        sectors = sectors_per_block - (sector_num % sectors_per_block);
        if (sectors > nb_sectors) {
            sectors = nb_sectors;
        }

489
        if (offset == -1) {
490
            memset(buf, 0, sectors * BDRV_SECTOR_SIZE);
491
        } else {
492 493 494
            ret = bdrv_pread(bs->file, offset, buf,
                sectors * BDRV_SECTOR_SIZE);
            if (ret != sectors * BDRV_SECTOR_SIZE) {
495
                return -1;
496
            }
497 498
        }

499 500 501
        nb_sectors -= sectors;
        sector_num += sectors;
        buf += sectors * BDRV_SECTOR_SIZE;
502 503 504 505
    }
    return 0;
}

506 507 508 509 510 511 512 513 514 515 516
static coroutine_fn int vpc_co_read(BlockDriverState *bs, int64_t sector_num,
                                    uint8_t *buf, int nb_sectors)
{
    int ret;
    BDRVVPCState *s = bs->opaque;
    qemu_co_mutex_lock(&s->lock);
    ret = vpc_read(bs, sector_num, buf, nb_sectors);
    qemu_co_mutex_unlock(&s->lock);
    return ret;
}

A
aliguori 已提交
517 518 519
static int vpc_write(BlockDriverState *bs, int64_t sector_num,
    const uint8_t *buf, int nb_sectors)
{
520
    BDRVVPCState *s = bs->opaque;
A
aliguori 已提交
521
    int64_t offset;
522
    int64_t sectors, sectors_per_block;
A
aliguori 已提交
523
    int ret;
524
    VHDFooter *footer =  (VHDFooter *) s->footer_buf;
A
aliguori 已提交
525

526
    if (be32_to_cpu(footer->type) == VHD_FIXED) {
527 528
        return bdrv_write(bs->file, sector_num, buf, nb_sectors);
    }
A
aliguori 已提交
529 530 531
    while (nb_sectors > 0) {
        offset = get_sector_offset(bs, sector_num, 1);

532 533 534 535 536 537
        sectors_per_block = s->block_size >> BDRV_SECTOR_BITS;
        sectors = sectors_per_block - (sector_num % sectors_per_block);
        if (sectors > nb_sectors) {
            sectors = nb_sectors;
        }

A
aliguori 已提交
538 539 540 541 542 543
        if (offset == -1) {
            offset = alloc_block(bs, sector_num);
            if (offset < 0)
                return -1;
        }

544 545
        ret = bdrv_pwrite(bs->file, offset, buf, sectors * BDRV_SECTOR_SIZE);
        if (ret != sectors * BDRV_SECTOR_SIZE) {
A
aliguori 已提交
546
            return -1;
547
        }
A
aliguori 已提交
548

549 550 551
        nb_sectors -= sectors;
        sector_num += sectors;
        buf += sectors * BDRV_SECTOR_SIZE;
A
aliguori 已提交
552 553 554 555 556
    }

    return 0;
}

557 558 559 560 561 562 563 564 565 566 567
static coroutine_fn int vpc_co_write(BlockDriverState *bs, int64_t sector_num,
                                     const uint8_t *buf, int nb_sectors)
{
    int ret;
    BDRVVPCState *s = bs->opaque;
    qemu_co_mutex_lock(&s->lock);
    ret = vpc_write(bs, sector_num, buf, nb_sectors);
    qemu_co_mutex_unlock(&s->lock);
    return ret;
}

568 569 570 571 572
static int64_t coroutine_fn vpc_co_get_block_status(BlockDriverState *bs,
        int64_t sector_num, int nb_sectors, int *pnum)
{
    BDRVVPCState *s = bs->opaque;
    VHDFooter *footer = (VHDFooter*) s->footer_buf;
573
    int64_t start, offset;
574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596
    bool allocated;
    int n;

    if (be32_to_cpu(footer->type) == VHD_FIXED) {
        *pnum = nb_sectors;
        return BDRV_BLOCK_RAW | BDRV_BLOCK_OFFSET_VALID | BDRV_BLOCK_DATA |
               (sector_num << BDRV_SECTOR_BITS);
    }

    offset = get_sector_offset(bs, sector_num, 0);
    start = offset;
    allocated = (offset != -1);
    *pnum = 0;

    do {
        /* All sectors in a block are contiguous (without using the bitmap) */
        n = ROUND_UP(sector_num + 1, s->block_size / BDRV_SECTOR_SIZE)
          - sector_num;
        n = MIN(n, nb_sectors);

        *pnum += n;
        sector_num += n;
        nb_sectors -= n;
597 598 599 600 601
        /* *pnum can't be greater than one block for allocated
         * sectors since there is always a bitmap in between. */
        if (allocated) {
            return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID | start;
        }
602 603 604 605
        if (nb_sectors == 0) {
            break;
        }
        offset = get_sector_offset(bs, sector_num, 0);
606
    } while (offset == -1);
607

608
    return 0;
609 610
}

A
aliguori 已提交
611 612 613 614 615 616 617
/*
 * Calculates the number of cylinders, heads and sectors per cylinder
 * based on a given number of sectors. This is the algorithm described
 * in the VHD specification.
 *
 * Note that the geometry doesn't always exactly match total_sectors but
 * may round it down.
618
 *
619 620 621
 * Returns 0 on success, -EFBIG if the size is larger than ~2 TB. Override
 * the hardware EIDE and ATA-2 limit of 16 heads (max disk size of 127 GB)
 * and instead allow up to 255 heads.
A
aliguori 已提交
622
 */
623
static int calculate_geometry(int64_t total_sectors, uint16_t* cyls,
A
aliguori 已提交
624 625 626 627
    uint8_t* heads, uint8_t* secs_per_cyl)
{
    uint32_t cyls_times_heads;

628
    total_sectors = MIN(total_sectors, VHD_MAX_GEOMETRY);
A
aliguori 已提交
629

630
    if (total_sectors >= 65535LL * 16 * 63) {
A
aliguori 已提交
631
        *secs_per_cyl = 255;
632
        *heads = 16;
A
aliguori 已提交
633 634 635 636 637 638
        cyls_times_heads = total_sectors / *secs_per_cyl;
    } else {
        *secs_per_cyl = 17;
        cyls_times_heads = total_sectors / *secs_per_cyl;
        *heads = (cyls_times_heads + 1023) / 1024;

639
        if (*heads < 4) {
A
aliguori 已提交
640
            *heads = 4;
641
        }
A
aliguori 已提交
642 643 644 645 646 647 648 649 650 651 652 653 654 655

        if (cyls_times_heads >= (*heads * 1024) || *heads > 16) {
            *secs_per_cyl = 31;
            *heads = 16;
            cyls_times_heads = total_sectors / *secs_per_cyl;
        }

        if (cyls_times_heads >= (*heads * 1024)) {
            *secs_per_cyl = 63;
            *heads = 16;
            cyls_times_heads = total_sectors / *secs_per_cyl;
        }
    }

656
    *cyls = cyls_times_heads / *heads;
657 658

    return 0;
A
aliguori 已提交
659 660
}

661 662
static int create_dynamic_disk(BlockDriverState *bs, uint8_t *buf,
                               int64_t total_sectors)
A
aliguori 已提交
663
{
664 665
    VHDDynDiskHeader *dyndisk_header =
        (VHDDynDiskHeader *) buf;
A
aliguori 已提交
666
    size_t block_size, num_bat_entries;
667
    int i;
668 669
    int ret;
    int64_t offset = 0;
A
aliguori 已提交
670 671 672 673 674

    // Write the footer (twice: at the beginning and at the end)
    block_size = 0x200000;
    num_bat_entries = (total_sectors + block_size / 512) / (block_size / 512);

675 676
    ret = bdrv_pwrite_sync(bs, offset, buf, HEADER_SIZE);
    if (ret) {
B
Blue Swirl 已提交
677 678
        goto fail;
    }
A
aliguori 已提交
679

680 681 682
    offset = 1536 + ((num_bat_entries * 4 + 511) & ~511);
    ret = bdrv_pwrite_sync(bs, offset, buf, HEADER_SIZE);
    if (ret < 0) {
B
Blue Swirl 已提交
683 684
        goto fail;
    }
A
aliguori 已提交
685 686

    // Write the initial BAT
687
    offset = 3 * 512;
A
aliguori 已提交
688 689

    memset(buf, 0xFF, 512);
B
Blue Swirl 已提交
690
    for (i = 0; i < (num_bat_entries * 4 + 511) / 512; i++) {
691 692
        ret = bdrv_pwrite_sync(bs, offset, buf, 512);
        if (ret < 0) {
B
Blue Swirl 已提交
693 694
            goto fail;
        }
695
        offset += 512;
B
Blue Swirl 已提交
696
    }
A
aliguori 已提交
697 698 699 700

    // Prepare the Dynamic Disk Header
    memset(buf, 0, 1024);

701
    memcpy(dyndisk_header->magic, "cxsparse", 8);
A
aliguori 已提交
702

703 704 705 706
    /*
     * Note: The spec is actually wrong here for data_offset, it says
     * 0xFFFFFFFF, but MS tools expect all 64 bits to be set.
     */
707 708 709 710 711
    dyndisk_header->data_offset = cpu_to_be64(0xFFFFFFFFFFFFFFFFULL);
    dyndisk_header->table_offset = cpu_to_be64(3 * 512);
    dyndisk_header->version = cpu_to_be32(0x00010000);
    dyndisk_header->block_size = cpu_to_be32(block_size);
    dyndisk_header->max_table_entries = cpu_to_be32(num_bat_entries);
A
aliguori 已提交
712

713
    dyndisk_header->checksum = cpu_to_be32(vpc_checksum(buf, 1024));
A
aliguori 已提交
714 715

    // Write the header
716
    offset = 512;
A
aliguori 已提交
717

718 719
    ret = bdrv_pwrite_sync(bs, offset, buf, 1024);
    if (ret < 0) {
B
Blue Swirl 已提交
720 721 722
        goto fail;
    }

723 724 725 726
 fail:
    return ret;
}

727 728
static int create_fixed_disk(BlockDriverState *bs, uint8_t *buf,
                             int64_t total_size)
729
{
730
    int ret;
731 732

    /* Add footer to total size */
733 734 735 736 737
    total_size += HEADER_SIZE;

    ret = bdrv_truncate(bs, total_size);
    if (ret < 0) {
        return ret;
738 739
    }

740 741 742 743
    ret = bdrv_pwrite_sync(bs, total_size - HEADER_SIZE, buf, HEADER_SIZE);
    if (ret < 0) {
        return ret;
    }
744 745 746 747

    return ret;
}

748
static int vpc_create(const char *filename, QemuOpts *opts, Error **errp)
749 750
{
    uint8_t buf[1024];
751
    VHDFooter *footer = (VHDFooter *) buf;
752
    char *disk_type_param;
753
    int i;
754 755 756 757 758 759 760
    uint16_t cyls = 0;
    uint8_t heads = 0;
    uint8_t secs_per_cyl = 0;
    int64_t total_sectors;
    int64_t total_size;
    int disk_type;
    int ret = -EIO;
761 762
    Error *local_err = NULL;
    BlockDriverState *bs = NULL;
763 764

    /* Read out options */
765 766
    total_size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0),
                          BDRV_SECTOR_SIZE);
767 768 769
    disk_type_param = qemu_opt_get_del(opts, BLOCK_OPT_SUBFMT);
    if (disk_type_param) {
        if (!strcmp(disk_type_param, "dynamic")) {
770
            disk_type = VHD_DYNAMIC;
771
        } else if (!strcmp(disk_type_param, "fixed")) {
772 773
            disk_type = VHD_FIXED;
        } else {
774 775
            ret = -EINVAL;
            goto out;
776 777 778 779 780
        }
    } else {
        disk_type = VHD_DYNAMIC;
    }

781 782 783
    ret = bdrv_create_file(filename, opts, &local_err);
    if (ret < 0) {
        error_propagate(errp, local_err);
784
        goto out;
785
    }
786 787 788 789 790
    ret = bdrv_open(&bs, filename, NULL, NULL, BDRV_O_RDWR | BDRV_O_PROTOCOL,
                    NULL, &local_err);
    if (ret < 0) {
        error_propagate(errp, local_err);
        goto out;
791 792
    }

793 794 795 796
    /*
     * Calculate matching total_size and geometry. Increase the number of
     * sectors requested until we get enough (or fail). This ensures that
     * qemu-img convert doesn't truncate images, but rather rounds up.
797 798 799 800
     *
     * If the image size can't be represented by a spec conform CHS geometry,
     * we set the geometry to 65535 x 16 x 255 (CxHxS) sectors and use
     * the image size from the VHD footer to calculate total_sectors.
801
     */
802
    total_sectors = MIN(VHD_MAX_GEOMETRY, total_size / BDRV_SECTOR_SIZE);
803
    for (i = 0; total_sectors > (int64_t)cyls * heads * secs_per_cyl; i++) {
804 805 806 807 808 809 810
        calculate_geometry(total_sectors + i, &cyls, &heads, &secs_per_cyl);
    }

    if ((int64_t)cyls * heads * secs_per_cyl == VHD_MAX_GEOMETRY) {
        total_sectors = total_size / BDRV_SECTOR_SIZE;
        /* Allow a maximum disk size of approximately 2 TB */
        if (total_sectors > VHD_MAX_SECTORS) {
811
            ret = -EFBIG;
812
            goto out;
813
        }
814 815 816
    } else {
        total_sectors = (int64_t)cyls * heads * secs_per_cyl;
        total_size = total_sectors * BDRV_SECTOR_SIZE;
817
    }
818

819 820 821 822 823 824 825 826
    /* Prepare the Hard Disk Footer */
    memset(buf, 0, 1024);

    memcpy(footer->creator, "conectix", 8);
    /* TODO Check if "qemu" creator_app is ok for VPC */
    memcpy(footer->creator_app, "qemu", 4);
    memcpy(footer->creator_os, "Wi2k", 4);

827 828
    footer->features = cpu_to_be32(0x02);
    footer->version = cpu_to_be32(0x00010000);
829
    if (disk_type == VHD_DYNAMIC) {
830
        footer->data_offset = cpu_to_be64(HEADER_SIZE);
831
    } else {
832
        footer->data_offset = cpu_to_be64(0xFFFFFFFFFFFFFFFFULL);
833
    }
834
    footer->timestamp = cpu_to_be32(time(NULL) - VHD_TIMESTAMP_BASE);
835 836

    /* Version of Virtual PC 2007 */
837 838
    footer->major = cpu_to_be16(0x0005);
    footer->minor = cpu_to_be16(0x0003);
839
    footer->orig_size = cpu_to_be64(total_size);
840
    footer->current_size = cpu_to_be64(total_size);
841
    footer->cyls = cpu_to_be16(cyls);
842 843 844
    footer->heads = heads;
    footer->secs_per_cyl = secs_per_cyl;

845
    footer->type = cpu_to_be32(disk_type);
846

847 848 849
#if defined(CONFIG_UUID)
    uuid_generate(footer->uuid);
#endif
850

851
    footer->checksum = cpu_to_be32(vpc_checksum(buf, HEADER_SIZE));
852 853

    if (disk_type == VHD_DYNAMIC) {
854
        ret = create_dynamic_disk(bs, buf, total_sectors);
855
    } else {
856
        ret = create_fixed_disk(bs, buf, total_size);
857 858
    }

859
out:
860
    bdrv_unref(bs);
861
    g_free(disk_type_param);
B
Blue Swirl 已提交
862
    return ret;
A
aliguori 已提交
863 864
}

K
Kevin Wolf 已提交
865 866 867
static int vpc_has_zero_init(BlockDriverState *bs)
{
    BDRVVPCState *s = bs->opaque;
868
    VHDFooter *footer =  (VHDFooter *) s->footer_buf;
K
Kevin Wolf 已提交
869

870
    if (be32_to_cpu(footer->type) == VHD_FIXED) {
K
Kevin Wolf 已提交
871 872 873 874 875 876
        return bdrv_has_zero_init(bs->file);
    } else {
        return 1;
    }
}

877 878 879
static void vpc_close(BlockDriverState *bs)
{
    BDRVVPCState *s = bs->opaque;
880
    qemu_vfree(s->pagetable);
881
#ifdef CACHE
882
    g_free(s->pageentry_u8);
883
#endif
K
Kevin Wolf 已提交
884 885 886

    migrate_del_blocker(s->migration_blocker);
    error_free(s->migration_blocker);
887 888
}

889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906
static QemuOptsList vpc_create_opts = {
    .name = "vpc-create-opts",
    .head = QTAILQ_HEAD_INITIALIZER(vpc_create_opts.head),
    .desc = {
        {
            .name = BLOCK_OPT_SIZE,
            .type = QEMU_OPT_SIZE,
            .help = "Virtual disk size"
        },
        {
            .name = BLOCK_OPT_SUBFMT,
            .type = QEMU_OPT_STRING,
            .help =
                "Type of virtual hard disk format. Supported formats are "
                "{dynamic (default) | fixed} "
        },
        { /* end of list */ }
    }
907 908
};

909
static BlockDriver bdrv_vpc = {
K
Kevin Wolf 已提交
910 911
    .format_name    = "vpc",
    .instance_size  = sizeof(BDRVVPCState),
912

K
Kevin Wolf 已提交
913 914 915 916
    .bdrv_probe             = vpc_probe,
    .bdrv_open              = vpc_open,
    .bdrv_close             = vpc_close,
    .bdrv_reopen_prepare    = vpc_reopen_prepare,
C
Chunyan Liu 已提交
917
    .bdrv_create            = vpc_create,
918

919 920 921
    .bdrv_read                  = vpc_co_read,
    .bdrv_write                 = vpc_co_write,
    .bdrv_co_get_block_status   = vpc_co_get_block_status,
922

P
Paolo Bonzini 已提交
923 924
    .bdrv_get_info          = vpc_get_info,

925
    .create_opts            = &vpc_create_opts,
K
Kevin Wolf 已提交
926
    .bdrv_has_zero_init     = vpc_has_zero_init,
927
};
928 929 930 931 932 933 934

static void bdrv_vpc_init(void)
{
    bdrv_register(&bdrv_vpc);
}

block_init(bdrv_vpc_init);