qcow2.c 109.4 KB
Newer Older
B
bellard 已提交
1 2
/*
 * Block driver for the QCOW version 2 format
3
 *
B
bellard 已提交
4
 * Copyright (c) 2004-2006 Fabrice Bellard
5
 *
B
bellard 已提交
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
 * 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
Peter Maydell 已提交
24
#include "qemu/osdep.h"
25
#include "block/block_int.h"
26
#include "sysemu/block-backend.h"
27
#include "qemu/module.h"
B
bellard 已提交
28
#include <zlib.h>
K
Kevin Wolf 已提交
29
#include "block/qcow2.h"
30
#include "qemu/error-report.h"
31
#include "qapi/qmp/qerror.h"
32
#include "qapi/qmp/qbool.h"
33
#include "qapi/util.h"
M
Max Reitz 已提交
34 35
#include "qapi/qmp/types.h"
#include "qapi-event.h"
K
Kevin Wolf 已提交
36
#include "trace.h"
37
#include "qemu/option_int.h"
38
#include "qemu/cutils.h"
39
#include "qemu/bswap.h"
B
bellard 已提交
40 41 42 43 44 45 46 47

/*
  Differences with QCOW:

  - Support for multiple incremental snapshots.
  - Memory management by reference counts.
  - Clusters which have a reference count of one have the bit
    QCOW_OFLAG_COPIED to optimize write performance.
48
  - Size of compressed clusters is stored in sectors to reduce bit usage
B
bellard 已提交
49 50
    in the cluster offsets.
  - Support for storing additional data (such as the VM state) in the
51
    snapshots.
B
bellard 已提交
52 53 54 55 56
  - If a backing store is used, the cluster size is not constrained
    (could be backported to QCOW).
  - L2 tables have always a size of one cluster.
*/

57 58 59 60

typedef struct {
    uint32_t magic;
    uint32_t len;
61
} QEMU_PACKED QCowExtension;
J
Jeff Cody 已提交
62

63 64
#define  QCOW2_EXT_MAGIC_END 0
#define  QCOW2_EXT_MAGIC_BACKING_FORMAT 0xE2792ACA
65
#define  QCOW2_EXT_MAGIC_FEATURE_TABLE 0x6803f857
66

67
static int qcow2_probe(const uint8_t *buf, int buf_size, const char *filename)
B
bellard 已提交
68 69
{
    const QCowHeader *cow_header = (const void *)buf;
70

B
bellard 已提交
71 72
    if (buf_size >= sizeof(QCowHeader) &&
        be32_to_cpu(cow_header->magic) == QCOW_MAGIC &&
K
Kevin Wolf 已提交
73
        be32_to_cpu(cow_header->version) >= 2)
B
bellard 已提交
74 75 76 77 78
        return 100;
    else
        return 0;
}

79 80 81 82 83 84 85 86

/* 
 * read qcow2 extension and fill bs
 * start reading from start_offset
 * finish reading upon magic of value 0 or when end_offset reached
 * unknown magic is skipped (future extension this version knows nothing about)
 * return 0 upon success, non-0 otherwise
 */
87
static int qcow2_read_extensions(BlockDriverState *bs, uint64_t start_offset,
M
Max Reitz 已提交
88 89
                                 uint64_t end_offset, void **p_feature_table,
                                 Error **errp)
90
{
91
    BDRVQcow2State *s = bs->opaque;
92 93
    QCowExtension ext;
    uint64_t offset;
94
    int ret;
95 96

#ifdef DEBUG_EXT
97
    printf("qcow2_read_extensions: start=%ld end=%ld\n", start_offset, end_offset);
98 99 100 101 102 103 104
#endif
    offset = start_offset;
    while (offset < end_offset) {

#ifdef DEBUG_EXT
        /* Sanity check */
        if (offset > s->cluster_size)
105
            printf("qcow2_read_extension: suspicious offset %lu\n", offset);
106

D
Dong Xu Wang 已提交
107
        printf("attempting to read extended header in offset %lu\n", offset);
108 109
#endif

110
        ret = bdrv_pread(bs->file, offset, &ext, sizeof(ext));
M
Max Reitz 已提交
111 112 113
        if (ret < 0) {
            error_setg_errno(errp, -ret, "qcow2_read_extension: ERROR: "
                             "pread fail from offset %" PRIu64, offset);
114 115 116 117 118 119 120 121
            return 1;
        }
        be32_to_cpus(&ext.magic);
        be32_to_cpus(&ext.len);
        offset += sizeof(ext);
#ifdef DEBUG_EXT
        printf("ext.magic = 0x%x\n", ext.magic);
#endif
122
        if (offset > end_offset || ext.len > end_offset - offset) {
M
Max Reitz 已提交
123
            error_setg(errp, "Header extension too large");
124 125 126
            return -EINVAL;
        }

127
        switch (ext.magic) {
128
        case QCOW2_EXT_MAGIC_END:
129
            return 0;
130

131
        case QCOW2_EXT_MAGIC_BACKING_FORMAT:
132
            if (ext.len >= sizeof(bs->backing_format)) {
133 134 135
                error_setg(errp, "ERROR: ext_backing_format: len=%" PRIu32
                           " too large (>=%zu)", ext.len,
                           sizeof(bs->backing_format));
136 137
                return 2;
            }
138
            ret = bdrv_pread(bs->file, offset, bs->backing_format, ext.len);
M
Max Reitz 已提交
139 140 141
            if (ret < 0) {
                error_setg_errno(errp, -ret, "ERROR: ext_backing_format: "
                                 "Could not read format name");
142
                return 3;
M
Max Reitz 已提交
143
            }
144
            bs->backing_format[ext.len] = '\0';
145
            s->image_backing_format = g_strdup(bs->backing_format);
146 147 148 149 150
#ifdef DEBUG_EXT
            printf("Qcow2: Got format extension %s\n", bs->backing_format);
#endif
            break;

151 152 153
        case QCOW2_EXT_MAGIC_FEATURE_TABLE:
            if (p_feature_table != NULL) {
                void* feature_table = g_malloc0(ext.len + 2 * sizeof(Qcow2Feature));
154
                ret = bdrv_pread(bs->file, offset , feature_table, ext.len);
155
                if (ret < 0) {
M
Max Reitz 已提交
156 157
                    error_setg_errno(errp, -ret, "ERROR: ext_feature_table: "
                                     "Could not read table");
158 159 160 161 162 163 164
                    return ret;
                }

                *p_feature_table = feature_table;
            }
            break;

165
        default:
166 167 168 169 170 171 172 173 174
            /* unknown magic - save it in case we need to rewrite the header */
            {
                Qcow2UnknownHeaderExtension *uext;

                uext = g_malloc0(sizeof(*uext)  + ext.len);
                uext->magic = ext.magic;
                uext->len = ext.len;
                QLIST_INSERT_HEAD(&s->unknown_header_ext, uext, next);

175
                ret = bdrv_pread(bs->file, offset , uext->data, uext->len);
176
                if (ret < 0) {
M
Max Reitz 已提交
177 178
                    error_setg_errno(errp, -ret, "ERROR: unknown extension: "
                                     "Could not read data");
179 180 181
                    return ret;
                }
            }
182 183
            break;
        }
184 185

        offset += ((ext.len + 7) & ~7);
186 187 188 189 190
    }

    return 0;
}

191 192
static void cleanup_unknown_header_ext(BlockDriverState *bs)
{
193
    BDRVQcow2State *s = bs->opaque;
194 195 196 197 198 199 200
    Qcow2UnknownHeaderExtension *uext, *next;

    QLIST_FOREACH_SAFE(uext, &s->unknown_header_ext, next, next) {
        QLIST_REMOVE(uext, next);
        g_free(uext);
    }
}
201

202 203
static void report_unsupported_feature(Error **errp, Qcow2Feature *table,
                                       uint64_t mask)
204
{
205 206 207
    char *features = g_strdup("");
    char *old;

208 209
    while (table && table->name[0] != '\0') {
        if (table->type == QCOW2_FEAT_TYPE_INCOMPATIBLE) {
210 211 212 213 214 215
            if (mask & (1ULL << table->bit)) {
                old = features;
                features = g_strdup_printf("%s%s%.46s", old, *old ? ", " : "",
                                           table->name);
                g_free(old);
                mask &= ~(1ULL << table->bit);
216 217 218 219 220 221
            }
        }
        table++;
    }

    if (mask) {
222 223 224 225
        old = features;
        features = g_strdup_printf("%s%sUnknown incompatible feature: %" PRIx64,
                                   old, *old ? ", " : "", mask);
        g_free(old);
226
    }
227

228
    error_setg(errp, "Unsupported qcow2 feature(s): %s", features);
229
    g_free(features);
230 231
}

232 233 234 235 236 237 238
/*
 * Sets the dirty bit and flushes afterwards if necessary.
 *
 * The incompatible_features bit is only set if the image file header was
 * updated successfully.  Therefore it is not required to check the return
 * value of this function.
 */
239
int qcow2_mark_dirty(BlockDriverState *bs)
240
{
241
    BDRVQcow2State *s = bs->opaque;
242 243 244 245 246 247 248 249 250 251
    uint64_t val;
    int ret;

    assert(s->qcow_version >= 3);

    if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) {
        return 0; /* already dirty */
    }

    val = cpu_to_be64(s->incompatible_features | QCOW2_INCOMPAT_DIRTY);
252
    ret = bdrv_pwrite(bs->file, offsetof(QCowHeader, incompatible_features),
253 254 255 256
                      &val, sizeof(val));
    if (ret < 0) {
        return ret;
    }
K
Kevin Wolf 已提交
257
    ret = bdrv_flush(bs->file->bs);
258 259 260 261 262 263 264 265 266
    if (ret < 0) {
        return ret;
    }

    /* Only treat image as dirty if the header was updated successfully */
    s->incompatible_features |= QCOW2_INCOMPAT_DIRTY;
    return 0;
}

S
Stefan Hajnoczi 已提交
267 268 269 270 271 272 273
/*
 * Clears the dirty bit and flushes before if necessary.  Only call this
 * function when there are no pending requests, it does not guard against
 * concurrent requests dirtying the image.
 */
static int qcow2_mark_clean(BlockDriverState *bs)
{
274
    BDRVQcow2State *s = bs->opaque;
S
Stefan Hajnoczi 已提交
275 276

    if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) {
277 278 279 280 281
        int ret;

        s->incompatible_features &= ~QCOW2_INCOMPAT_DIRTY;

        ret = bdrv_flush(bs);
S
Stefan Hajnoczi 已提交
282 283 284 285 286 287 288 289 290
        if (ret < 0) {
            return ret;
        }

        return qcow2_update_header(bs);
    }
    return 0;
}

M
Max Reitz 已提交
291 292 293 294 295
/*
 * Marks the image as corrupt.
 */
int qcow2_mark_corrupt(BlockDriverState *bs)
{
296
    BDRVQcow2State *s = bs->opaque;
M
Max Reitz 已提交
297 298 299 300 301 302 303 304 305 306 307

    s->incompatible_features |= QCOW2_INCOMPAT_CORRUPT;
    return qcow2_update_header(bs);
}

/*
 * Marks the image as consistent, i.e., unsets the corrupt bit, and flushes
 * before if necessary.
 */
int qcow2_mark_consistent(BlockDriverState *bs)
{
308
    BDRVQcow2State *s = bs->opaque;
M
Max Reitz 已提交
309 310 311 312 313 314 315 316 317 318 319 320 321

    if (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT) {
        int ret = bdrv_flush(bs);
        if (ret < 0) {
            return ret;
        }

        s->incompatible_features &= ~QCOW2_INCOMPAT_CORRUPT;
        return qcow2_update_header(bs);
    }
    return 0;
}

322 323 324 325 326 327 328 329 330
static int qcow2_check(BlockDriverState *bs, BdrvCheckResult *result,
                       BdrvCheckMode fix)
{
    int ret = qcow2_check_refcounts(bs, result, fix);
    if (ret < 0) {
        return ret;
    }

    if (fix && result->check_errors == 0 && result->corruptions == 0) {
M
Max Reitz 已提交
331 332 333 334 335
        ret = qcow2_mark_clean(bs);
        if (ret < 0) {
            return ret;
        }
        return qcow2_mark_consistent(bs);
336 337 338 339
    }
    return ret;
}

340 341 342
static int validate_table_offset(BlockDriverState *bs, uint64_t offset,
                                 uint64_t entries, size_t entry_len)
{
343
    BDRVQcow2State *s = bs->opaque;
344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365
    uint64_t size;

    /* Use signed INT64_MAX as the maximum even for uint64_t header fields,
     * because values will be passed to qemu functions taking int64_t. */
    if (entries > INT64_MAX / entry_len) {
        return -EINVAL;
    }

    size = entries * entry_len;

    if (INT64_MAX - size < offset) {
        return -EINVAL;
    }

    /* Tables must be cluster aligned */
    if (offset & (s->cluster_size - 1)) {
        return -EINVAL;
    }

    return 0;
}

366 367 368 369 370
static QemuOptsList qcow2_runtime_opts = {
    .name = "qcow2",
    .head = QTAILQ_HEAD_INITIALIZER(qcow2_runtime_opts.head),
    .desc = {
        {
371
            .name = QCOW2_OPT_LAZY_REFCOUNTS,
372 373 374
            .type = QEMU_OPT_BOOL,
            .help = "Postpone refcount updates",
        },
375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390
        {
            .name = QCOW2_OPT_DISCARD_REQUEST,
            .type = QEMU_OPT_BOOL,
            .help = "Pass guest discard requests to the layer below",
        },
        {
            .name = QCOW2_OPT_DISCARD_SNAPSHOT,
            .type = QEMU_OPT_BOOL,
            .help = "Generate discard requests when snapshot related space "
                    "is freed",
        },
        {
            .name = QCOW2_OPT_DISCARD_OTHER,
            .type = QEMU_OPT_BOOL,
            .help = "Generate discard requests when other clusters are freed",
        },
M
Max Reitz 已提交
391 392 393 394 395 396
        {
            .name = QCOW2_OPT_OVERLAP,
            .type = QEMU_OPT_STRING,
            .help = "Selects which overlap checks to perform from a range of "
                    "templates (none, constant, cached, all)",
        },
397 398 399 400 401 402
        {
            .name = QCOW2_OPT_OVERLAP_TEMPLATE,
            .type = QEMU_OPT_STRING,
            .help = "Selects which overlap checks to perform from a range of "
                    "templates (none, constant, cached, all)",
        },
M
Max Reitz 已提交
403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442
        {
            .name = QCOW2_OPT_OVERLAP_MAIN_HEADER,
            .type = QEMU_OPT_BOOL,
            .help = "Check for unintended writes into the main qcow2 header",
        },
        {
            .name = QCOW2_OPT_OVERLAP_ACTIVE_L1,
            .type = QEMU_OPT_BOOL,
            .help = "Check for unintended writes into the active L1 table",
        },
        {
            .name = QCOW2_OPT_OVERLAP_ACTIVE_L2,
            .type = QEMU_OPT_BOOL,
            .help = "Check for unintended writes into an active L2 table",
        },
        {
            .name = QCOW2_OPT_OVERLAP_REFCOUNT_TABLE,
            .type = QEMU_OPT_BOOL,
            .help = "Check for unintended writes into the refcount table",
        },
        {
            .name = QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK,
            .type = QEMU_OPT_BOOL,
            .help = "Check for unintended writes into a refcount block",
        },
        {
            .name = QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE,
            .type = QEMU_OPT_BOOL,
            .help = "Check for unintended writes into the snapshot table",
        },
        {
            .name = QCOW2_OPT_OVERLAP_INACTIVE_L1,
            .type = QEMU_OPT_BOOL,
            .help = "Check for unintended writes into an inactive L1 table",
        },
        {
            .name = QCOW2_OPT_OVERLAP_INACTIVE_L2,
            .type = QEMU_OPT_BOOL,
            .help = "Check for unintended writes into an inactive L2 table",
        },
443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458
        {
            .name = QCOW2_OPT_CACHE_SIZE,
            .type = QEMU_OPT_SIZE,
            .help = "Maximum combined metadata (L2 tables and refcount blocks) "
                    "cache size",
        },
        {
            .name = QCOW2_OPT_L2_CACHE_SIZE,
            .type = QEMU_OPT_SIZE,
            .help = "Maximum L2 table cache size",
        },
        {
            .name = QCOW2_OPT_REFCOUNT_CACHE_SIZE,
            .type = QEMU_OPT_SIZE,
            .help = "Maximum refcount block cache size",
        },
459 460 461 462 463
        {
            .name = QCOW2_OPT_CACHE_CLEAN_INTERVAL,
            .type = QEMU_OPT_NUMBER,
            .help = "Clean unused cache entries after this time (in seconds)",
        },
464 465 466 467
        { /* end of list */ }
    },
};

468 469 470 471 472 473 474 475 476 477 478
static const char *overlap_bool_option_names[QCOW2_OL_MAX_BITNR] = {
    [QCOW2_OL_MAIN_HEADER_BITNR]    = QCOW2_OPT_OVERLAP_MAIN_HEADER,
    [QCOW2_OL_ACTIVE_L1_BITNR]      = QCOW2_OPT_OVERLAP_ACTIVE_L1,
    [QCOW2_OL_ACTIVE_L2_BITNR]      = QCOW2_OPT_OVERLAP_ACTIVE_L2,
    [QCOW2_OL_REFCOUNT_TABLE_BITNR] = QCOW2_OPT_OVERLAP_REFCOUNT_TABLE,
    [QCOW2_OL_REFCOUNT_BLOCK_BITNR] = QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK,
    [QCOW2_OL_SNAPSHOT_TABLE_BITNR] = QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE,
    [QCOW2_OL_INACTIVE_L1_BITNR]    = QCOW2_OPT_OVERLAP_INACTIVE_L1,
    [QCOW2_OL_INACTIVE_L2_BITNR]    = QCOW2_OPT_OVERLAP_INACTIVE_L2,
};

479 480 481
static void cache_clean_timer_cb(void *opaque)
{
    BlockDriverState *bs = opaque;
482
    BDRVQcow2State *s = bs->opaque;
483 484 485 486 487 488 489 490
    qcow2_cache_clean_unused(bs, s->l2_table_cache);
    qcow2_cache_clean_unused(bs, s->refcount_block_cache);
    timer_mod(s->cache_clean_timer, qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) +
              (int64_t) s->cache_clean_interval * 1000);
}

static void cache_clean_timer_init(BlockDriverState *bs, AioContext *context)
{
491
    BDRVQcow2State *s = bs->opaque;
492 493 494 495 496 497 498 499 500 501 502
    if (s->cache_clean_interval > 0) {
        s->cache_clean_timer = aio_timer_new(context, QEMU_CLOCK_VIRTUAL,
                                             SCALE_MS, cache_clean_timer_cb,
                                             bs);
        timer_mod(s->cache_clean_timer, qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) +
                  (int64_t) s->cache_clean_interval * 1000);
    }
}

static void cache_clean_timer_del(BlockDriverState *bs)
{
503
    BDRVQcow2State *s = bs->opaque;
504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521
    if (s->cache_clean_timer) {
        timer_del(s->cache_clean_timer);
        timer_free(s->cache_clean_timer);
        s->cache_clean_timer = NULL;
    }
}

static void qcow2_detach_aio_context(BlockDriverState *bs)
{
    cache_clean_timer_del(bs);
}

static void qcow2_attach_aio_context(BlockDriverState *bs,
                                     AioContext *new_context)
{
    cache_clean_timer_init(bs, new_context);
}

M
Max Reitz 已提交
522 523
static void read_cache_sizes(BlockDriverState *bs, QemuOpts *opts,
                             uint64_t *l2_cache_size,
524 525
                             uint64_t *refcount_cache_size, Error **errp)
{
526
    BDRVQcow2State *s = bs->opaque;
527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565
    uint64_t combined_cache_size;
    bool l2_cache_size_set, refcount_cache_size_set, combined_cache_size_set;

    combined_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_CACHE_SIZE);
    l2_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_L2_CACHE_SIZE);
    refcount_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_REFCOUNT_CACHE_SIZE);

    combined_cache_size = qemu_opt_get_size(opts, QCOW2_OPT_CACHE_SIZE, 0);
    *l2_cache_size = qemu_opt_get_size(opts, QCOW2_OPT_L2_CACHE_SIZE, 0);
    *refcount_cache_size = qemu_opt_get_size(opts,
                                             QCOW2_OPT_REFCOUNT_CACHE_SIZE, 0);

    if (combined_cache_size_set) {
        if (l2_cache_size_set && refcount_cache_size_set) {
            error_setg(errp, QCOW2_OPT_CACHE_SIZE ", " QCOW2_OPT_L2_CACHE_SIZE
                       " and " QCOW2_OPT_REFCOUNT_CACHE_SIZE " may not be set "
                       "the same time");
            return;
        } else if (*l2_cache_size > combined_cache_size) {
            error_setg(errp, QCOW2_OPT_L2_CACHE_SIZE " may not exceed "
                       QCOW2_OPT_CACHE_SIZE);
            return;
        } else if (*refcount_cache_size > combined_cache_size) {
            error_setg(errp, QCOW2_OPT_REFCOUNT_CACHE_SIZE " may not exceed "
                       QCOW2_OPT_CACHE_SIZE);
            return;
        }

        if (l2_cache_size_set) {
            *refcount_cache_size = combined_cache_size - *l2_cache_size;
        } else if (refcount_cache_size_set) {
            *l2_cache_size = combined_cache_size - *refcount_cache_size;
        } else {
            *refcount_cache_size = combined_cache_size
                                 / (DEFAULT_L2_REFCOUNT_SIZE_RATIO + 1);
            *l2_cache_size = combined_cache_size - *refcount_cache_size;
        }
    } else {
        if (!l2_cache_size_set && !refcount_cache_size_set) {
M
Max Reitz 已提交
566 567 568
            *l2_cache_size = MAX(DEFAULT_L2_CACHE_BYTE_SIZE,
                                 (uint64_t)DEFAULT_L2_CACHE_CLUSTERS
                                 * s->cluster_size);
569 570 571 572 573 574 575 576 577 578 579 580
            *refcount_cache_size = *l2_cache_size
                                 / DEFAULT_L2_REFCOUNT_SIZE_RATIO;
        } else if (!l2_cache_size_set) {
            *l2_cache_size = *refcount_cache_size
                           * DEFAULT_L2_REFCOUNT_SIZE_RATIO;
        } else if (!refcount_cache_size_set) {
            *refcount_cache_size = *l2_cache_size
                                 / DEFAULT_L2_REFCOUNT_SIZE_RATIO;
        }
    }
}

581 582 583 584 585 586 587 588 589 590 591 592 593
typedef struct Qcow2ReopenState {
    Qcow2Cache *l2_table_cache;
    Qcow2Cache *refcount_block_cache;
    bool use_lazy_refcounts;
    int overlap_check;
    bool discard_passthrough[QCOW2_DISCARD_MAX];
    uint64_t cache_clean_interval;
} Qcow2ReopenState;

static int qcow2_update_options_prepare(BlockDriverState *bs,
                                        Qcow2ReopenState *r,
                                        QDict *options, int flags,
                                        Error **errp)
594 595
{
    BDRVQcow2State *s = bs->opaque;
596
    QemuOpts *opts = NULL;
597 598
    const char *opt_overlap_check, *opt_overlap_check_template;
    int overlap_check_template = 0;
599
    uint64_t l2_cache_size, refcount_cache_size;
600
    int i;
601
    Error *local_err = NULL;
602 603
    int ret;

604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640
    opts = qemu_opts_create(&qcow2_runtime_opts, NULL, 0, &error_abort);
    qemu_opts_absorb_qdict(opts, options, &local_err);
    if (local_err) {
        error_propagate(errp, local_err);
        ret = -EINVAL;
        goto fail;
    }

    /* get L2 table/refcount block cache size from command line options */
    read_cache_sizes(bs, opts, &l2_cache_size, &refcount_cache_size,
                     &local_err);
    if (local_err) {
        error_propagate(errp, local_err);
        ret = -EINVAL;
        goto fail;
    }

    l2_cache_size /= s->cluster_size;
    if (l2_cache_size < MIN_L2_CACHE_SIZE) {
        l2_cache_size = MIN_L2_CACHE_SIZE;
    }
    if (l2_cache_size > INT_MAX) {
        error_setg(errp, "L2 cache size too big");
        ret = -EINVAL;
        goto fail;
    }

    refcount_cache_size /= s->cluster_size;
    if (refcount_cache_size < MIN_REFCOUNT_CACHE_SIZE) {
        refcount_cache_size = MIN_REFCOUNT_CACHE_SIZE;
    }
    if (refcount_cache_size > INT_MAX) {
        error_setg(errp, "Refcount cache size too big");
        ret = -EINVAL;
        goto fail;
    }

641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658
    /* alloc new L2 table/refcount block cache, flush old one */
    if (s->l2_table_cache) {
        ret = qcow2_cache_flush(bs, s->l2_table_cache);
        if (ret) {
            error_setg_errno(errp, -ret, "Failed to flush the L2 table cache");
            goto fail;
        }
    }

    if (s->refcount_block_cache) {
        ret = qcow2_cache_flush(bs, s->refcount_block_cache);
        if (ret) {
            error_setg_errno(errp, -ret,
                             "Failed to flush the refcount block cache");
            goto fail;
        }
    }

659 660 661
    r->l2_table_cache = qcow2_cache_create(bs, l2_cache_size);
    r->refcount_block_cache = qcow2_cache_create(bs, refcount_cache_size);
    if (r->l2_table_cache == NULL || r->refcount_block_cache == NULL) {
662 663 664 665 666 667
        error_setg(errp, "Could not allocate metadata caches");
        ret = -ENOMEM;
        goto fail;
    }

    /* New interval for cache cleanup timer */
668
    r->cache_clean_interval =
669 670
        qemu_opt_get_number(opts, QCOW2_OPT_CACHE_CLEAN_INTERVAL,
                            s->cache_clean_interval);
671
    if (r->cache_clean_interval > UINT_MAX) {
672 673 674 675 676
        error_setg(errp, "Cache clean interval too big");
        ret = -EINVAL;
        goto fail;
    }

677
    /* lazy-refcounts; flush if going from enabled to disabled */
678
    r->use_lazy_refcounts = qemu_opt_get_bool(opts, QCOW2_OPT_LAZY_REFCOUNTS,
679
        (s->compatible_features & QCOW2_COMPAT_LAZY_REFCOUNTS));
680
    if (r->use_lazy_refcounts && s->qcow_version < 3) {
681 682 683 684 685
        error_setg(errp, "Lazy refcounts require a qcow2 image with at least "
                   "qemu 1.1 compatibility level");
        ret = -EINVAL;
        goto fail;
    }
686

687 688 689 690 691 692 693 694
    if (s->use_lazy_refcounts && !r->use_lazy_refcounts) {
        ret = qcow2_mark_clean(bs);
        if (ret < 0) {
            error_setg_errno(errp, -ret, "Failed to disable lazy refcounts");
            goto fail;
        }
    }

695
    /* Overlap check options */
696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726
    opt_overlap_check = qemu_opt_get(opts, QCOW2_OPT_OVERLAP);
    opt_overlap_check_template = qemu_opt_get(opts, QCOW2_OPT_OVERLAP_TEMPLATE);
    if (opt_overlap_check_template && opt_overlap_check &&
        strcmp(opt_overlap_check_template, opt_overlap_check))
    {
        error_setg(errp, "Conflicting values for qcow2 options '"
                   QCOW2_OPT_OVERLAP "' ('%s') and '" QCOW2_OPT_OVERLAP_TEMPLATE
                   "' ('%s')", opt_overlap_check, opt_overlap_check_template);
        ret = -EINVAL;
        goto fail;
    }
    if (!opt_overlap_check) {
        opt_overlap_check = opt_overlap_check_template ?: "cached";
    }

    if (!strcmp(opt_overlap_check, "none")) {
        overlap_check_template = 0;
    } else if (!strcmp(opt_overlap_check, "constant")) {
        overlap_check_template = QCOW2_OL_CONSTANT;
    } else if (!strcmp(opt_overlap_check, "cached")) {
        overlap_check_template = QCOW2_OL_CACHED;
    } else if (!strcmp(opt_overlap_check, "all")) {
        overlap_check_template = QCOW2_OL_ALL;
    } else {
        error_setg(errp, "Unsupported value '%s' for qcow2 option "
                   "'overlap-check'. Allowed are any of the following: "
                   "none, constant, cached, all", opt_overlap_check);
        ret = -EINVAL;
        goto fail;
    }

727
    r->overlap_check = 0;
728 729 730
    for (i = 0; i < QCOW2_OL_MAX_BITNR; i++) {
        /* overlap-check defines a template bitmask, but every flag may be
         * overwritten through the associated boolean option */
731
        r->overlap_check |=
732 733 734 735
            qemu_opt_get_bool(opts, overlap_bool_option_names[i],
                              overlap_check_template & (1 << i)) << i;
    }

736 737 738
    r->discard_passthrough[QCOW2_DISCARD_NEVER] = false;
    r->discard_passthrough[QCOW2_DISCARD_ALWAYS] = true;
    r->discard_passthrough[QCOW2_DISCARD_REQUEST] =
739 740
        qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_REQUEST,
                          flags & BDRV_O_UNMAP);
741
    r->discard_passthrough[QCOW2_DISCARD_SNAPSHOT] =
742
        qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_SNAPSHOT, true);
743
    r->discard_passthrough[QCOW2_DISCARD_OTHER] =
744 745
        qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_OTHER, false);

746 747
    ret = 0;
fail:
748 749
    qemu_opts_del(opts);
    opts = NULL;
750 751 752 753 754 755 756 757 758
    return ret;
}

static void qcow2_update_options_commit(BlockDriverState *bs,
                                        Qcow2ReopenState *r)
{
    BDRVQcow2State *s = bs->opaque;
    int i;

759 760 761 762 763 764
    if (s->l2_table_cache) {
        qcow2_cache_destroy(bs, s->l2_table_cache);
    }
    if (s->refcount_block_cache) {
        qcow2_cache_destroy(bs, s->refcount_block_cache);
    }
765 766 767 768 769 770 771 772 773 774
    s->l2_table_cache = r->l2_table_cache;
    s->refcount_block_cache = r->refcount_block_cache;

    s->overlap_check = r->overlap_check;
    s->use_lazy_refcounts = r->use_lazy_refcounts;

    for (i = 0; i < QCOW2_DISCARD_MAX; i++) {
        s->discard_passthrough[i] = r->discard_passthrough[i];
    }

775 776 777 778 779
    if (s->cache_clean_interval != r->cache_clean_interval) {
        cache_clean_timer_del(bs);
        s->cache_clean_interval = r->cache_clean_interval;
        cache_clean_timer_init(bs, bdrv_get_aio_context(bs));
    }
780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804
}

static void qcow2_update_options_abort(BlockDriverState *bs,
                                       Qcow2ReopenState *r)
{
    if (r->l2_table_cache) {
        qcow2_cache_destroy(bs, r->l2_table_cache);
    }
    if (r->refcount_block_cache) {
        qcow2_cache_destroy(bs, r->refcount_block_cache);
    }
}

static int qcow2_update_options(BlockDriverState *bs, QDict *options,
                                int flags, Error **errp)
{
    Qcow2ReopenState r = {};
    int ret;

    ret = qcow2_update_options_prepare(bs, &r, options, flags, errp);
    if (ret >= 0) {
        qcow2_update_options_commit(bs, &r);
    } else {
        qcow2_update_options_abort(bs, &r);
    }
805

806 807 808
    return ret;
}

M
Max Reitz 已提交
809 810
static int qcow2_open(BlockDriverState *bs, QDict *options, int flags,
                      Error **errp)
B
bellard 已提交
811
{
812
    BDRVQcow2State *s = bs->opaque;
813 814
    unsigned int len, i;
    int ret = 0;
B
bellard 已提交
815
    QCowHeader header;
816
    Error *local_err = NULL;
817
    uint64_t ext_end;
818
    uint64_t l1_vm_state_index;
B
bellard 已提交
819

820
    ret = bdrv_pread(bs->file, 0, &header, sizeof(header));
821
    if (ret < 0) {
M
Max Reitz 已提交
822
        error_setg_errno(errp, -ret, "Could not read qcow2 header");
B
bellard 已提交
823
        goto fail;
824
    }
B
bellard 已提交
825 826 827 828 829 830 831 832 833 834 835 836 837
    be32_to_cpus(&header.magic);
    be32_to_cpus(&header.version);
    be64_to_cpus(&header.backing_file_offset);
    be32_to_cpus(&header.backing_file_size);
    be64_to_cpus(&header.size);
    be32_to_cpus(&header.cluster_bits);
    be32_to_cpus(&header.crypt_method);
    be64_to_cpus(&header.l1_table_offset);
    be32_to_cpus(&header.l1_size);
    be64_to_cpus(&header.refcount_table_offset);
    be32_to_cpus(&header.refcount_table_clusters);
    be64_to_cpus(&header.snapshots_offset);
    be32_to_cpus(&header.nb_snapshots);
838

K
Kevin Wolf 已提交
839
    if (header.magic != QCOW_MAGIC) {
M
Max Reitz 已提交
840
        error_setg(errp, "Image is not in qcow2 format");
P
Paolo Bonzini 已提交
841
        ret = -EINVAL;
B
bellard 已提交
842
        goto fail;
843
    }
K
Kevin Wolf 已提交
844
    if (header.version < 2 || header.version > 3) {
845
        error_setg(errp, "Unsupported qcow2 version %" PRIu32, header.version);
K
Kevin Wolf 已提交
846 847 848 849 850 851
        ret = -ENOTSUP;
        goto fail;
    }

    s->qcow_version = header.version;

852 853 854
    /* Initialise cluster size */
    if (header.cluster_bits < MIN_CLUSTER_BITS ||
        header.cluster_bits > MAX_CLUSTER_BITS) {
855 856
        error_setg(errp, "Unsupported cluster size: 2^%" PRIu32,
                   header.cluster_bits);
857 858 859 860 861 862 863 864
        ret = -EINVAL;
        goto fail;
    }

    s->cluster_bits = header.cluster_bits;
    s->cluster_size = 1 << s->cluster_bits;
    s->cluster_sectors = 1 << (s->cluster_bits - 9);

K
Kevin Wolf 已提交
865 866 867 868 869 870 871 872 873 874 875 876 877
    /* Initialise version 3 header fields */
    if (header.version == 2) {
        header.incompatible_features    = 0;
        header.compatible_features      = 0;
        header.autoclear_features       = 0;
        header.refcount_order           = 4;
        header.header_length            = 72;
    } else {
        be64_to_cpus(&header.incompatible_features);
        be64_to_cpus(&header.compatible_features);
        be64_to_cpus(&header.autoclear_features);
        be32_to_cpus(&header.refcount_order);
        be32_to_cpus(&header.header_length);
878 879 880 881 882 883 884 885 886 887 888 889

        if (header.header_length < 104) {
            error_setg(errp, "qcow2 header too short");
            ret = -EINVAL;
            goto fail;
        }
    }

    if (header.header_length > s->cluster_size) {
        error_setg(errp, "qcow2 header exceeds cluster size");
        ret = -EINVAL;
        goto fail;
K
Kevin Wolf 已提交
890 891 892 893 894
    }

    if (header.header_length > sizeof(header)) {
        s->unknown_header_fields_size = header.header_length - sizeof(header);
        s->unknown_header_fields = g_malloc(s->unknown_header_fields_size);
895
        ret = bdrv_pread(bs->file, sizeof(header), s->unknown_header_fields,
K
Kevin Wolf 已提交
896 897
                         s->unknown_header_fields_size);
        if (ret < 0) {
M
Max Reitz 已提交
898 899
            error_setg_errno(errp, -ret, "Could not read unknown qcow2 header "
                             "fields");
K
Kevin Wolf 已提交
900 901 902 903
            goto fail;
        }
    }

904 905 906 907 908 909
    if (header.backing_file_offset > s->cluster_size) {
        error_setg(errp, "Invalid backing file offset");
        ret = -EINVAL;
        goto fail;
    }

910 911 912 913 914 915
    if (header.backing_file_offset) {
        ext_end = header.backing_file_offset;
    } else {
        ext_end = 1 << header.cluster_bits;
    }

K
Kevin Wolf 已提交
916 917 918 919 920
    /* Handle feature bits */
    s->incompatible_features    = header.incompatible_features;
    s->compatible_features      = header.compatible_features;
    s->autoclear_features       = header.autoclear_features;

S
Stefan Hajnoczi 已提交
921
    if (s->incompatible_features & ~QCOW2_INCOMPAT_MASK) {
922 923
        void *feature_table = NULL;
        qcow2_read_extensions(bs, header.header_length, ext_end,
M
Max Reitz 已提交
924
                              &feature_table, NULL);
925
        report_unsupported_feature(errp, feature_table,
S
Stefan Hajnoczi 已提交
926 927
                                   s->incompatible_features &
                                   ~QCOW2_INCOMPAT_MASK);
K
Kevin Wolf 已提交
928
        ret = -ENOTSUP;
929
        g_free(feature_table);
K
Kevin Wolf 已提交
930 931 932
        goto fail;
    }

M
Max Reitz 已提交
933 934 935 936
    if (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT) {
        /* Corrupt images may not be written to unless they are being repaired
         */
        if ((flags & BDRV_O_RDWR) && !(flags & BDRV_O_CHECK)) {
M
Max Reitz 已提交
937 938
            error_setg(errp, "qcow2: Image is corrupt; cannot be opened "
                       "read/write");
M
Max Reitz 已提交
939 940 941 942 943
            ret = -EACCES;
            goto fail;
        }
    }

K
Kevin Wolf 已提交
944
    /* Check support for various header values */
945 946 947 948
    if (header.refcount_order > 6) {
        error_setg(errp, "Reference count entry width too large; may not "
                   "exceed 64 bits");
        ret = -EINVAL;
K
Kevin Wolf 已提交
949 950
        goto fail;
    }
951
    s->refcount_order = header.refcount_order;
952 953 954
    s->refcount_bits = 1 << s->refcount_order;
    s->refcount_max = UINT64_C(1) << (s->refcount_bits - 1);
    s->refcount_max += s->refcount_max - 1;
K
Kevin Wolf 已提交
955

956
    if (header.crypt_method > QCOW_CRYPT_AES) {
957
        error_setg(errp, "Unsupported encryption method: %" PRIu32,
M
Max Reitz 已提交
958
                   header.crypt_method);
959
        ret = -EINVAL;
B
bellard 已提交
960
        goto fail;
961
    }
962 963
    if (!qcrypto_cipher_supports(QCRYPTO_CIPHER_ALG_AES_128,
                                 QCRYPTO_CIPHER_MODE_CBC)) {
964 965 966 967
        error_setg(errp, "AES cipher not available");
        ret = -EINVAL;
        goto fail;
    }
B
bellard 已提交
968
    s->crypt_method_header = header.crypt_method;
969
    if (s->crypt_method_header) {
970 971
        if (bdrv_uses_whitelist() &&
            s->crypt_method_header == QCOW_CRYPT_AES) {
972 973 974 975 976 977 978 979 980 981
            error_setg(errp,
                       "Use of AES-CBC encrypted qcow2 images is no longer "
                       "supported in system emulators");
            error_append_hint(errp,
                              "You can use 'qemu-img convert' to convert your "
                              "image to an alternative supported format, such "
                              "as unencrypted qcow2, or raw with the LUKS "
                              "format instead.\n");
            ret = -ENOSYS;
            goto fail;
982 983
        }

984
        bs->encrypted = true;
985
    }
986

B
bellard 已提交
987 988
    s->l2_bits = s->cluster_bits - 3; /* L2 is always one cluster */
    s->l2_size = 1 << s->l2_bits;
989 990 991
    /* 2^(s->refcount_order - 3) is the refcount width in bytes */
    s->refcount_block_bits = s->cluster_bits - (s->refcount_order - 3);
    s->refcount_block_size = 1 << s->refcount_block_bits;
B
bellard 已提交
992 993 994 995
    bs->total_sectors = header.size / 512;
    s->csize_shift = (62 - (s->cluster_bits - 8));
    s->csize_mask = (1 << (s->cluster_bits - 8)) - 1;
    s->cluster_offset_mask = (1LL << s->csize_shift) - 1;
996

B
bellard 已提交
997
    s->refcount_table_offset = header.refcount_table_offset;
998
    s->refcount_table_size =
B
bellard 已提交
999 1000
        header.refcount_table_clusters << (s->cluster_bits - 3);

1001
    if (header.refcount_table_clusters > qcow2_max_refcount_clusters(s)) {
1002 1003 1004 1005 1006
        error_setg(errp, "Reference count table too large");
        ret = -EINVAL;
        goto fail;
    }

1007 1008 1009 1010 1011 1012 1013
    ret = validate_table_offset(bs, s->refcount_table_offset,
                                s->refcount_table_size, sizeof(uint64_t));
    if (ret < 0) {
        error_setg(errp, "Invalid reference count table offset");
        goto fail;
    }

1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028
    /* Snapshot table offset/length */
    if (header.nb_snapshots > QCOW_MAX_SNAPSHOTS) {
        error_setg(errp, "Too many snapshots");
        ret = -EINVAL;
        goto fail;
    }

    ret = validate_table_offset(bs, header.snapshots_offset,
                                header.nb_snapshots,
                                sizeof(QCowSnapshotHeader));
    if (ret < 0) {
        error_setg(errp, "Invalid snapshot table offset");
        goto fail;
    }

B
bellard 已提交
1029
    /* read the level 1 table */
1030
    if (header.l1_size > QCOW_MAX_L1_SIZE / sizeof(uint64_t)) {
1031 1032 1033 1034
        error_setg(errp, "Active L1 table too large");
        ret = -EFBIG;
        goto fail;
    }
B
bellard 已提交
1035
    s->l1_size = header.l1_size;
1036 1037 1038

    l1_vm_state_index = size_to_l1(s, header.size);
    if (l1_vm_state_index > INT_MAX) {
M
Max Reitz 已提交
1039
        error_setg(errp, "Image is too big");
1040 1041 1042 1043 1044
        ret = -EFBIG;
        goto fail;
    }
    s->l1_vm_state_index = l1_vm_state_index;

B
bellard 已提交
1045 1046
    /* the L1 table must contain at least enough entries to put
       header.size bytes */
1047
    if (s->l1_size < s->l1_vm_state_index) {
M
Max Reitz 已提交
1048
        error_setg(errp, "L1 table is too small");
1049
        ret = -EINVAL;
B
bellard 已提交
1050
        goto fail;
1051
    }
1052 1053 1054 1055 1056 1057 1058

    ret = validate_table_offset(bs, header.l1_table_offset,
                                header.l1_size, sizeof(uint64_t));
    if (ret < 0) {
        error_setg(errp, "Invalid L1 table offset");
        goto fail;
    }
B
bellard 已提交
1059
    s->l1_table_offset = header.l1_table_offset;
1060 1061


1062
    if (s->l1_size > 0) {
K
Kevin Wolf 已提交
1063
        s->l1_table = qemu_try_blockalign(bs->file->bs,
1064
            align_offset(s->l1_size * sizeof(uint64_t), 512));
1065 1066 1067 1068 1069
        if (s->l1_table == NULL) {
            error_setg(errp, "Could not allocate L1 table");
            ret = -ENOMEM;
            goto fail;
        }
1070
        ret = bdrv_pread(bs->file, s->l1_table_offset, s->l1_table,
1071 1072
                         s->l1_size * sizeof(uint64_t));
        if (ret < 0) {
M
Max Reitz 已提交
1073
            error_setg_errno(errp, -ret, "Could not read L1 table");
1074
            goto fail;
1075
        }
1076 1077 1078
        for(i = 0;i < s->l1_size; i++) {
            be64_to_cpus(&s->l1_table[i]);
        }
B
bellard 已提交
1079
    }
K
Kevin Wolf 已提交
1080

1081 1082
    /* Parse driver-specific options */
    ret = qcow2_update_options(bs, options, flags, errp);
1083 1084 1085 1086
    if (ret < 0) {
        goto fail;
    }

1087
    s->cluster_cache = g_malloc(s->cluster_size);
B
bellard 已提交
1088
    /* one more sector for decompressed data alignment */
K
Kevin Wolf 已提交
1089
    s->cluster_data = qemu_try_blockalign(bs->file->bs, QCOW_MAX_CRYPT_CLUSTERS
1090 1091 1092 1093 1094 1095 1096
                                                    * s->cluster_size + 512);
    if (s->cluster_data == NULL) {
        error_setg(errp, "Could not allocate temporary cluster buffer");
        ret = -ENOMEM;
        goto fail;
    }

B
bellard 已提交
1097
    s->cluster_cache_offset = -1;
1098
    s->flags = flags;
1099

1100 1101
    ret = qcow2_refcount_init(bs);
    if (ret != 0) {
M
Max Reitz 已提交
1102
        error_setg_errno(errp, -ret, "Could not initialize refcount handling");
B
bellard 已提交
1103
        goto fail;
1104
    }
B
bellard 已提交
1105

B
Blue Swirl 已提交
1106
    QLIST_INIT(&s->cluster_allocs);
K
Kevin Wolf 已提交
1107
    QTAILQ_INIT(&s->discards);
1108

1109
    /* read qcow2 extensions */
M
Max Reitz 已提交
1110 1111 1112
    if (qcow2_read_extensions(bs, header.header_length, ext_end, NULL,
        &local_err)) {
        error_propagate(errp, local_err);
1113
        ret = -EINVAL;
1114
        goto fail;
1115
    }
1116

B
bellard 已提交
1117 1118 1119
    /* read the backing file name */
    if (header.backing_file_offset != 0) {
        len = header.backing_file_size;
1120
        if (len > MIN(1023, s->cluster_size - header.backing_file_offset) ||
1121
            len >= sizeof(bs->backing_file)) {
1122 1123 1124
            error_setg(errp, "Backing file name too long");
            ret = -EINVAL;
            goto fail;
1125
        }
1126
        ret = bdrv_pread(bs->file, header.backing_file_offset,
1127 1128
                         bs->backing_file, len);
        if (ret < 0) {
M
Max Reitz 已提交
1129
            error_setg_errno(errp, -ret, "Could not read backing file name");
B
bellard 已提交
1130
            goto fail;
1131
        }
B
bellard 已提交
1132
        bs->backing_file[len] = '\0';
1133
        s->image_backing_file = g_strdup(bs->backing_file);
B
bellard 已提交
1134
    }
1135

1136 1137 1138 1139
    /* Internal snapshots */
    s->snapshots_offset = header.snapshots_offset;
    s->nb_snapshots = header.nb_snapshots;

1140 1141
    ret = qcow2_read_snapshots(bs);
    if (ret < 0) {
M
Max Reitz 已提交
1142
        error_setg_errno(errp, -ret, "Could not read snapshots");
B
bellard 已提交
1143
        goto fail;
1144
    }
B
bellard 已提交
1145

1146
    /* Clear unknown autoclear feature bits */
1147
    if (!bs->read_only && !(flags & BDRV_O_INACTIVE) && s->autoclear_features) {
1148 1149 1150
        s->autoclear_features = 0;
        ret = qcow2_update_header(bs);
        if (ret < 0) {
M
Max Reitz 已提交
1151
            error_setg_errno(errp, -ret, "Could not update qcow2 header");
1152 1153 1154 1155
            goto fail;
        }
    }

K
Kevin Wolf 已提交
1156 1157
    /* Initialise locks */
    qemu_co_mutex_init(&s->lock);
F
Fam Zheng 已提交
1158
    bs->supported_zero_flags = BDRV_REQ_MAY_UNMAP;
K
Kevin Wolf 已提交
1159

S
Stefan Hajnoczi 已提交
1160
    /* Repair image if dirty */
1161
    if (!(flags & (BDRV_O_CHECK | BDRV_O_INACTIVE)) && !bs->read_only &&
1162
        (s->incompatible_features & QCOW2_INCOMPAT_DIRTY)) {
S
Stefan Hajnoczi 已提交
1163 1164
        BdrvCheckResult result = {0};

M
Max Reitz 已提交
1165
        ret = qcow2_check(bs, &result, BDRV_FIX_ERRORS | BDRV_FIX_LEAKS);
S
Stefan Hajnoczi 已提交
1166
        if (ret < 0) {
M
Max Reitz 已提交
1167
            error_setg_errno(errp, -ret, "Could not repair dirty image");
S
Stefan Hajnoczi 已提交
1168 1169 1170 1171
            goto fail;
        }
    }

B
bellard 已提交
1172
#ifdef DEBUG_ALLOC
P
Philipp Hahn 已提交
1173 1174
    {
        BdrvCheckResult result = {0};
1175
        qcow2_check_refcounts(bs, &result, 0);
P
Philipp Hahn 已提交
1176
    }
B
bellard 已提交
1177
#endif
1178
    return ret;
B
bellard 已提交
1179 1180

 fail:
K
Kevin Wolf 已提交
1181
    g_free(s->unknown_header_fields);
1182
    cleanup_unknown_header_ext(bs);
K
Kevin Wolf 已提交
1183 1184
    qcow2_free_snapshots(bs);
    qcow2_refcount_close(bs);
1185
    qemu_vfree(s->l1_table);
1186 1187
    /* else pre-write overlap checks in cache_destroy may crash */
    s->l1_table = NULL;
1188
    cache_clean_timer_del(bs);
K
Kevin Wolf 已提交
1189 1190 1191
    if (s->l2_table_cache) {
        qcow2_cache_destroy(bs, s->l2_table_cache);
    }
1192 1193 1194
    if (s->refcount_block_cache) {
        qcow2_cache_destroy(bs, s->refcount_block_cache);
    }
1195
    g_free(s->cluster_cache);
1196
    qemu_vfree(s->cluster_data);
1197
    return ret;
B
bellard 已提交
1198 1199
}

1200
static void qcow2_refresh_limits(BlockDriverState *bs, Error **errp)
1201
{
1202
    BDRVQcow2State *s = bs->opaque;
1203

1204 1205
    if (bs->encrypted) {
        /* Encryption works on a sector granularity */
1206
        bs->bl.request_alignment = BDRV_SECTOR_SIZE;
1207
    }
1208
    bs->bl.pwrite_zeroes_alignment = s->cluster_size;
1209
    bs->bl.pdiscard_alignment = s->cluster_size;
1210 1211
}

1212
static int qcow2_set_key(BlockDriverState *bs, const char *key)
B
bellard 已提交
1213
{
1214
    BDRVQcow2State *s = bs->opaque;
B
bellard 已提交
1215 1216
    uint8_t keybuf[16];
    int len, i;
1217
    Error *err = NULL;
1218

B
bellard 已提交
1219 1220 1221 1222 1223 1224 1225 1226 1227
    memset(keybuf, 0, 16);
    len = strlen(key);
    if (len > 16)
        len = 16;
    /* XXX: we could compress the chars to 7 bits to increase
       entropy */
    for(i = 0;i < len;i++) {
        keybuf[i] = key[i];
    }
1228
    assert(bs->encrypted);
B
bellard 已提交
1229

1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241
    qcrypto_cipher_free(s->cipher);
    s->cipher = qcrypto_cipher_new(
        QCRYPTO_CIPHER_ALG_AES_128,
        QCRYPTO_CIPHER_MODE_CBC,
        keybuf, G_N_ELEMENTS(keybuf),
        &err);

    if (!s->cipher) {
        /* XXX would be nice if errors in this method could
         * be properly propagate to the caller. Would need
         * the bdrv_set_key() API signature to be fixed. */
        error_free(err);
B
bellard 已提交
1242 1243 1244 1245 1246
        return -1;
    }
    return 0;
}

J
Jeff Cody 已提交
1247 1248 1249
static int qcow2_reopen_prepare(BDRVReopenState *state,
                                BlockReopenQueue *queue, Error **errp)
{
1250
    Qcow2ReopenState *r;
1251 1252
    int ret;

1253 1254 1255 1256 1257 1258 1259 1260 1261 1262
    r = g_new0(Qcow2ReopenState, 1);
    state->opaque = r;

    ret = qcow2_update_options_prepare(state->bs, r, state->options,
                                       state->flags, errp);
    if (ret < 0) {
        goto fail;
    }

    /* We need to write out any unwritten data if we reopen read-only. */
1263 1264 1265
    if ((state->flags & BDRV_O_RDWR) == 0) {
        ret = bdrv_flush(state->bs);
        if (ret < 0) {
1266
            goto fail;
1267 1268 1269 1270
        }

        ret = qcow2_mark_clean(state->bs);
        if (ret < 0) {
1271
            goto fail;
1272 1273 1274
        }
    }

J
Jeff Cody 已提交
1275
    return 0;
1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292

fail:
    qcow2_update_options_abort(state->bs, r);
    g_free(r);
    return ret;
}

static void qcow2_reopen_commit(BDRVReopenState *state)
{
    qcow2_update_options_commit(state->bs, state->opaque);
    g_free(state->opaque);
}

static void qcow2_reopen_abort(BDRVReopenState *state)
{
    qcow2_update_options_abort(state->bs, state->opaque);
    g_free(state->opaque);
J
Jeff Cody 已提交
1293 1294
}

1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340
static void qcow2_join_options(QDict *options, QDict *old_options)
{
    bool has_new_overlap_template =
        qdict_haskey(options, QCOW2_OPT_OVERLAP) ||
        qdict_haskey(options, QCOW2_OPT_OVERLAP_TEMPLATE);
    bool has_new_total_cache_size =
        qdict_haskey(options, QCOW2_OPT_CACHE_SIZE);
    bool has_all_cache_options;

    /* New overlap template overrides all old overlap options */
    if (has_new_overlap_template) {
        qdict_del(old_options, QCOW2_OPT_OVERLAP);
        qdict_del(old_options, QCOW2_OPT_OVERLAP_TEMPLATE);
        qdict_del(old_options, QCOW2_OPT_OVERLAP_MAIN_HEADER);
        qdict_del(old_options, QCOW2_OPT_OVERLAP_ACTIVE_L1);
        qdict_del(old_options, QCOW2_OPT_OVERLAP_ACTIVE_L2);
        qdict_del(old_options, QCOW2_OPT_OVERLAP_REFCOUNT_TABLE);
        qdict_del(old_options, QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK);
        qdict_del(old_options, QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE);
        qdict_del(old_options, QCOW2_OPT_OVERLAP_INACTIVE_L1);
        qdict_del(old_options, QCOW2_OPT_OVERLAP_INACTIVE_L2);
    }

    /* New total cache size overrides all old options */
    if (qdict_haskey(options, QCOW2_OPT_CACHE_SIZE)) {
        qdict_del(old_options, QCOW2_OPT_L2_CACHE_SIZE);
        qdict_del(old_options, QCOW2_OPT_REFCOUNT_CACHE_SIZE);
    }

    qdict_join(options, old_options, false);

    /*
     * If after merging all cache size options are set, an old total size is
     * overwritten. Do keep all options, however, if all three are new. The
     * resulting error message is what we want to happen.
     */
    has_all_cache_options =
        qdict_haskey(options, QCOW2_OPT_CACHE_SIZE) ||
        qdict_haskey(options, QCOW2_OPT_L2_CACHE_SIZE) ||
        qdict_haskey(options, QCOW2_OPT_REFCOUNT_CACHE_SIZE);

    if (has_all_cache_options && !has_new_total_cache_size) {
        qdict_del(options, QCOW2_OPT_CACHE_SIZE);
    }
}

1341
static int64_t coroutine_fn qcow2_co_get_block_status(BlockDriverState *bs,
1342
        int64_t sector_num, int nb_sectors, int *pnum, BlockDriverState **file)
B
bellard 已提交
1343
{
1344
    BDRVQcow2State *s = bs->opaque;
B
bellard 已提交
1345
    uint64_t cluster_offset;
1346
    int index_in_cluster, ret;
K
Kevin Wolf 已提交
1347
    unsigned int bytes;
1348
    int64_t status = 0;
B
bellard 已提交
1349

K
Kevin Wolf 已提交
1350
    bytes = MIN(INT_MAX, nb_sectors * BDRV_SECTOR_SIZE);
1351
    qemu_co_mutex_lock(&s->lock);
K
Kevin Wolf 已提交
1352 1353
    ret = qcow2_get_cluster_offset(bs, sector_num << 9, &bytes,
                                   &cluster_offset);
1354
    qemu_co_mutex_unlock(&s->lock);
1355
    if (ret < 0) {
1356
        return ret;
1357
    }
1358

K
Kevin Wolf 已提交
1359 1360
    *pnum = bytes >> BDRV_SECTOR_BITS;

1361
    if (cluster_offset != 0 && ret != QCOW2_CLUSTER_COMPRESSED &&
1362
        !s->cipher) {
1363 1364
        index_in_cluster = sector_num & (s->cluster_sectors - 1);
        cluster_offset |= (index_in_cluster << BDRV_SECTOR_BITS);
1365
        *file = bs->file->bs;
1366 1367 1368 1369 1370 1371 1372 1373
        status |= BDRV_BLOCK_OFFSET_VALID | cluster_offset;
    }
    if (ret == QCOW2_CLUSTER_ZERO) {
        status |= BDRV_BLOCK_ZERO;
    } else if (ret != QCOW2_CLUSTER_UNALLOCATED) {
        status |= BDRV_BLOCK_DATA;
    }
    return status;
B
bellard 已提交
1374 1375
}

B
bellard 已提交
1376
/* handle reading after the end of the backing file */
1377
int qcow2_backing_read1(BlockDriverState *bs, QEMUIOVector *qiov,
K
Kevin Wolf 已提交
1378
                        int64_t offset, int bytes)
B
bellard 已提交
1379
{
K
Kevin Wolf 已提交
1380
    uint64_t bs_size = bs->total_sectors * BDRV_SECTOR_SIZE;
B
bellard 已提交
1381
    int n1;
K
Kevin Wolf 已提交
1382 1383 1384 1385 1386 1387

    if ((offset + bytes) <= bs_size) {
        return bytes;
    }

    if (offset >= bs_size) {
B
bellard 已提交
1388
        n1 = 0;
K
Kevin Wolf 已提交
1389 1390 1391
    } else {
        n1 = bs_size - offset;
    }
1392

K
Kevin Wolf 已提交
1393
    qemu_iovec_memset(qiov, n1, 0, bytes - n1);
1394

B
bellard 已提交
1395 1396 1397
    return n1;
}

K
Kevin Wolf 已提交
1398 1399 1400
static coroutine_fn int qcow2_co_preadv(BlockDriverState *bs, uint64_t offset,
                                        uint64_t bytes, QEMUIOVector *qiov,
                                        int flags)
B
bellard 已提交
1401
{
1402
    BDRVQcow2State *s = bs->opaque;
K
Kevin Wolf 已提交
1403
    int offset_in_cluster, n1;
K
Kevin Wolf 已提交
1404
    int ret;
K
Kevin Wolf 已提交
1405
    unsigned int cur_bytes; /* number of bytes in current iteration */
1406
    uint64_t cluster_offset = 0;
1407 1408 1409
    uint64_t bytes_done = 0;
    QEMUIOVector hd_qiov;
    uint8_t *cluster_data = NULL;
B
bellard 已提交
1410

1411 1412 1413 1414
    qemu_iovec_init(&hd_qiov, qiov->niov);

    qemu_co_mutex_lock(&s->lock);

K
Kevin Wolf 已提交
1415
    while (bytes != 0) {
1416

1417
        /* prepare next request */
K
Kevin Wolf 已提交
1418
        cur_bytes = MIN(bytes, INT_MAX);
1419
        if (s->cipher) {
K
Kevin Wolf 已提交
1420 1421
            cur_bytes = MIN(cur_bytes,
                            QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size);
B
bellard 已提交
1422
        }
1423

K
Kevin Wolf 已提交
1424
        ret = qcow2_get_cluster_offset(bs, offset, &cur_bytes, &cluster_offset);
1425
        if (ret < 0) {
1426
            goto fail;
1427
        }
1428

K
Kevin Wolf 已提交
1429
        offset_in_cluster = offset_into_cluster(s, offset);
1430

1431
        qemu_iovec_reset(&hd_qiov);
K
Kevin Wolf 已提交
1432
        qemu_iovec_concat(&hd_qiov, qiov, bytes_done, cur_bytes);
1433

1434 1435
        switch (ret) {
        case QCOW2_CLUSTER_UNALLOCATED:
1436

1437
            if (bs->backing) {
1438
                /* read from the base image */
1439
                n1 = qcow2_backing_read1(bs->backing->bs, &hd_qiov,
K
Kevin Wolf 已提交
1440
                                         offset, cur_bytes);
1441
                if (n1 > 0) {
1442 1443 1444
                    QEMUIOVector local_qiov;

                    qemu_iovec_init(&local_qiov, hd_qiov.niov);
K
Kevin Wolf 已提交
1445
                    qemu_iovec_concat(&local_qiov, &hd_qiov, 0, n1);
1446

1447 1448
                    BLKDBG_EVENT(bs->file, BLKDBG_READ_BACKING_AIO);
                    qemu_co_mutex_unlock(&s->lock);
1449
                    ret = bdrv_co_preadv(bs->backing, offset, n1,
K
Kevin Wolf 已提交
1450
                                         &local_qiov, 0);
1451
                    qemu_co_mutex_lock(&s->lock);
1452 1453 1454

                    qemu_iovec_destroy(&local_qiov);

1455
                    if (ret < 0) {
1456
                        goto fail;
1457 1458 1459 1460
                    }
                }
            } else {
                /* Note: in this case, no need to wait */
K
Kevin Wolf 已提交
1461
                qemu_iovec_memset(&hd_qiov, 0, 0, cur_bytes);
1462
            }
1463 1464
            break;

1465
        case QCOW2_CLUSTER_ZERO:
K
Kevin Wolf 已提交
1466
            qemu_iovec_memset(&hd_qiov, 0, 0, cur_bytes);
1467 1468
            break;

1469
        case QCOW2_CLUSTER_COMPRESSED:
1470 1471 1472
            /* add AIO support for compressed blocks ? */
            ret = qcow2_decompress_cluster(bs, cluster_offset);
            if (ret < 0) {
1473
                goto fail;
1474 1475
            }

1476
            qemu_iovec_from_buf(&hd_qiov, 0,
K
Kevin Wolf 已提交
1477 1478
                                s->cluster_cache + offset_in_cluster,
                                cur_bytes);
1479 1480 1481
            break;

        case QCOW2_CLUSTER_NORMAL:
1482
            if ((cluster_offset & 511) != 0) {
1483 1484
                ret = -EIO;
                goto fail;
1485
            }
1486

1487
            if (bs->encrypted) {
1488
                assert(s->cipher);
1489

1490 1491 1492 1493
                /*
                 * For encrypted images, read everything into a temporary
                 * contiguous buffer on which the AES functions can work.
                 */
1494 1495
                if (!cluster_data) {
                    cluster_data =
K
Kevin Wolf 已提交
1496 1497 1498
                        qemu_try_blockalign(bs->file->bs,
                                            QCOW_MAX_CRYPT_CLUSTERS
                                            * s->cluster_size);
1499 1500 1501 1502
                    if (cluster_data == NULL) {
                        ret = -ENOMEM;
                        goto fail;
                    }
1503 1504
                }

K
Kevin Wolf 已提交
1505
                assert(cur_bytes <= QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size);
1506
                qemu_iovec_reset(&hd_qiov);
K
Kevin Wolf 已提交
1507
                qemu_iovec_add(&hd_qiov, cluster_data, cur_bytes);
1508 1509 1510 1511
            }

            BLKDBG_EVENT(bs->file, BLKDBG_READ_AIO);
            qemu_co_mutex_unlock(&s->lock);
1512
            ret = bdrv_co_preadv(bs->file,
K
Kevin Wolf 已提交
1513 1514
                                 cluster_offset + offset_in_cluster,
                                 cur_bytes, &hd_qiov, 0);
1515 1516
            qemu_co_mutex_lock(&s->lock);
            if (ret < 0) {
1517
                goto fail;
1518
            }
1519
            if (bs->encrypted) {
1520
                assert(s->cipher);
K
Kevin Wolf 已提交
1521 1522
                assert((offset & (BDRV_SECTOR_SIZE - 1)) == 0);
                assert((cur_bytes & (BDRV_SECTOR_SIZE - 1)) == 0);
1523
                Error *err = NULL;
K
Kevin Wolf 已提交
1524 1525 1526 1527
                if (qcow2_encrypt_sectors(s, offset >> BDRV_SECTOR_BITS,
                                          cluster_data, cluster_data,
                                          cur_bytes >> BDRV_SECTOR_BITS,
                                          false, &err) < 0) {
1528 1529 1530 1531
                    error_free(err);
                    ret = -EIO;
                    goto fail;
                }
K
Kevin Wolf 已提交
1532
                qemu_iovec_from_buf(qiov, bytes_done, cluster_data, cur_bytes);
1533
            }
1534 1535 1536 1537 1538 1539
            break;

        default:
            g_assert_not_reached();
            ret = -EIO;
            goto fail;
1540
        }
1541

K
Kevin Wolf 已提交
1542 1543 1544
        bytes -= cur_bytes;
        offset += cur_bytes;
        bytes_done += cur_bytes;
1545
    }
1546
    ret = 0;
1547

1548
fail:
K
Kevin Wolf 已提交
1549
    qemu_co_mutex_unlock(&s->lock);
K
Kevin Wolf 已提交
1550

1551
    qemu_iovec_destroy(&hd_qiov);
1552
    qemu_vfree(cluster_data);
K
Kevin Wolf 已提交
1553 1554

    return ret;
B
bellard 已提交
1555 1556
}

K
Kevin Wolf 已提交
1557 1558 1559
static coroutine_fn int qcow2_co_pwritev(BlockDriverState *bs, uint64_t offset,
                                         uint64_t bytes, QEMUIOVector *qiov,
                                         int flags)
B
bellard 已提交
1560
{
1561
    BDRVQcow2State *s = bs->opaque;
K
Kevin Wolf 已提交
1562
    int offset_in_cluster;
K
Kevin Wolf 已提交
1563
    int ret;
K
Kevin Wolf 已提交
1564
    unsigned int cur_bytes; /* number of sectors in current iteration */
1565
    uint64_t cluster_offset;
1566 1567 1568
    QEMUIOVector hd_qiov;
    uint64_t bytes_done = 0;
    uint8_t *cluster_data = NULL;
1569
    QCowL2Meta *l2meta = NULL;
1570

K
Kevin Wolf 已提交
1571
    trace_qcow2_writev_start_req(qemu_coroutine_self(), offset, bytes);
K
Kevin Wolf 已提交
1572

1573 1574 1575
    qemu_iovec_init(&hd_qiov, qiov->niov);

    s->cluster_cache_offset = -1; /* disable compressed cache */
1576

1577 1578
    qemu_co_mutex_lock(&s->lock);

K
Kevin Wolf 已提交
1579
    while (bytes != 0) {
1580

1581
        l2meta = NULL;
K
Kevin Wolf 已提交
1582

K
Kevin Wolf 已提交
1583
        trace_qcow2_writev_start_part(qemu_coroutine_self());
K
Kevin Wolf 已提交
1584 1585 1586 1587 1588 1589
        offset_in_cluster = offset_into_cluster(s, offset);
        cur_bytes = MIN(bytes, INT_MAX);
        if (bs->encrypted) {
            cur_bytes = MIN(cur_bytes,
                            QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size
                            - offset_in_cluster);
1590
        }
1591

K
Kevin Wolf 已提交
1592 1593
        ret = qcow2_alloc_cluster_offset(bs, offset, &cur_bytes,
                                         &cluster_offset, &l2meta);
1594
        if (ret < 0) {
1595
            goto fail;
1596
        }
1597

1598
        assert((cluster_offset & 511) == 0);
1599

1600
        qemu_iovec_reset(&hd_qiov);
K
Kevin Wolf 已提交
1601
        qemu_iovec_concat(&hd_qiov, qiov, bytes_done, cur_bytes);
1602

1603
        if (bs->encrypted) {
1604 1605
            Error *err = NULL;
            assert(s->cipher);
1606
            if (!cluster_data) {
K
Kevin Wolf 已提交
1607
                cluster_data = qemu_try_blockalign(bs->file->bs,
1608 1609 1610 1611 1612 1613
                                                   QCOW_MAX_CRYPT_CLUSTERS
                                                   * s->cluster_size);
                if (cluster_data == NULL) {
                    ret = -ENOMEM;
                    goto fail;
                }
1614
            }
1615

1616
            assert(hd_qiov.size <=
1617
                   QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size);
1618
            qemu_iovec_to_buf(&hd_qiov, 0, cluster_data, hd_qiov.size);
1619

K
Kevin Wolf 已提交
1620 1621 1622
            if (qcow2_encrypt_sectors(s, offset >> BDRV_SECTOR_BITS,
                                      cluster_data, cluster_data,
                                      cur_bytes >>BDRV_SECTOR_BITS,
1623 1624 1625 1626 1627
                                      true, &err) < 0) {
                error_free(err);
                ret = -EIO;
                goto fail;
            }
1628

1629
            qemu_iovec_reset(&hd_qiov);
K
Kevin Wolf 已提交
1630
            qemu_iovec_add(&hd_qiov, cluster_data, cur_bytes);
1631
        }
1632

M
Max Reitz 已提交
1633
        ret = qcow2_pre_write_overlap_check(bs, 0,
K
Kevin Wolf 已提交
1634
                cluster_offset + offset_in_cluster, cur_bytes);
1635 1636 1637 1638
        if (ret < 0) {
            goto fail;
        }

1639
        qemu_co_mutex_unlock(&s->lock);
1640
        BLKDBG_EVENT(bs->file, BLKDBG_WRITE_AIO);
K
Kevin Wolf 已提交
1641
        trace_qcow2_writev_data(qemu_coroutine_self(),
K
Kevin Wolf 已提交
1642
                                cluster_offset + offset_in_cluster);
1643
        ret = bdrv_co_pwritev(bs->file,
K
Kevin Wolf 已提交
1644 1645
                              cluster_offset + offset_in_cluster,
                              cur_bytes, &hd_qiov, 0);
1646 1647
        qemu_co_mutex_lock(&s->lock);
        if (ret < 0) {
1648
            goto fail;
1649
        }
1650

1651 1652 1653
        while (l2meta != NULL) {
            QCowL2Meta *next;

1654 1655 1656 1657
            ret = qcow2_alloc_cluster_link_l2(bs, l2meta);
            if (ret < 0) {
                goto fail;
            }
1658

1659 1660 1661 1662 1663 1664 1665
            /* Take the request off the list of running requests */
            if (l2meta->nb_clusters != 0) {
                QLIST_REMOVE(l2meta, next_in_flight);
            }

            qemu_co_queue_restart_all(&l2meta->dependent_requests);

1666
            next = l2meta->next;
1667
            g_free(l2meta);
1668
            l2meta = next;
1669
        }
1670

K
Kevin Wolf 已提交
1671 1672 1673 1674
        bytes -= cur_bytes;
        offset += cur_bytes;
        bytes_done += cur_bytes;
        trace_qcow2_writev_done_part(qemu_coroutine_self(), cur_bytes);
1675
    }
1676
    ret = 0;
1677

1678
fail:
1679 1680
    qemu_co_mutex_unlock(&s->lock);

1681 1682 1683
    while (l2meta != NULL) {
        QCowL2Meta *next;

1684 1685 1686 1687
        if (l2meta->nb_clusters != 0) {
            QLIST_REMOVE(l2meta, next_in_flight);
        }
        qemu_co_queue_restart_all(&l2meta->dependent_requests);
1688 1689

        next = l2meta->next;
K
Kevin Wolf 已提交
1690
        g_free(l2meta);
1691
        l2meta = next;
K
Kevin Wolf 已提交
1692
    }
1693

1694
    qemu_iovec_destroy(&hd_qiov);
1695
    qemu_vfree(cluster_data);
K
Kevin Wolf 已提交
1696
    trace_qcow2_writev_done_req(qemu_coroutine_self(), ret);
K
Kevin Wolf 已提交
1697

K
Kevin Wolf 已提交
1698
    return ret;
B
bellard 已提交
1699 1700
}

K
Kevin Wolf 已提交
1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726
static int qcow2_inactivate(BlockDriverState *bs)
{
    BDRVQcow2State *s = bs->opaque;
    int ret, result = 0;

    ret = qcow2_cache_flush(bs, s->l2_table_cache);
    if (ret) {
        result = ret;
        error_report("Failed to flush the L2 table cache: %s",
                     strerror(-ret));
    }

    ret = qcow2_cache_flush(bs, s->refcount_block_cache);
    if (ret) {
        result = ret;
        error_report("Failed to flush the refcount block cache: %s",
                     strerror(-ret));
    }

    if (result == 0) {
        qcow2_mark_clean(bs);
    }

    return result;
}

1727
static void qcow2_close(BlockDriverState *bs)
B
bellard 已提交
1728
{
1729
    BDRVQcow2State *s = bs->opaque;
1730
    qemu_vfree(s->l1_table);
1731 1732
    /* else pre-write overlap checks in cache_destroy may crash */
    s->l1_table = NULL;
K
Kevin Wolf 已提交
1733

1734
    if (!(s->flags & BDRV_O_INACTIVE)) {
K
Kevin Wolf 已提交
1735
        qcow2_inactivate(bs);
1736
    }
S
Stefan Hajnoczi 已提交
1737

1738
    cache_clean_timer_del(bs);
K
Kevin Wolf 已提交
1739 1740 1741
    qcow2_cache_destroy(bs, s->l2_table_cache);
    qcow2_cache_destroy(bs, s->refcount_block_cache);

1742 1743 1744
    qcrypto_cipher_free(s->cipher);
    s->cipher = NULL;

K
Kevin Wolf 已提交
1745
    g_free(s->unknown_header_fields);
1746
    cleanup_unknown_header_ext(bs);
K
Kevin Wolf 已提交
1747

1748 1749 1750
    g_free(s->image_backing_file);
    g_free(s->image_backing_format);

1751
    g_free(s->cluster_cache);
1752
    qemu_vfree(s->cluster_data);
K
Kevin Wolf 已提交
1753
    qcow2_refcount_close(bs);
1754
    qcow2_free_snapshots(bs);
B
bellard 已提交
1755 1756
}

1757
static void qcow2_invalidate_cache(BlockDriverState *bs, Error **errp)
1758
{
1759
    BDRVQcow2State *s = bs->opaque;
1760
    int flags = s->flags;
1761
    QCryptoCipher *cipher = NULL;
1762
    QDict *options;
1763 1764
    Error *local_err = NULL;
    int ret;
1765 1766 1767 1768 1769 1770

    /*
     * Backing files are read-only which makes all of their metadata immutable,
     * that means we don't have to worry about reopening them here.
     */

1771 1772
    cipher = s->cipher;
    s->cipher = NULL;
1773 1774 1775

    qcow2_close(bs);

1776
    memset(s, 0, sizeof(BDRVQcow2State));
1777
    options = qdict_clone_shallow(bs->options);
1778

1779
    flags &= ~BDRV_O_INACTIVE;
1780
    ret = qcow2_open(bs, options, flags, &local_err);
1781
    QDECREF(options);
1782
    if (local_err) {
1783 1784
        error_propagate(errp, local_err);
        error_prepend(errp, "Could not reopen qcow2 layer: ");
1785
        bs->drv = NULL;
1786 1787 1788
        return;
    } else if (ret < 0) {
        error_setg_errno(errp, -ret, "Could not reopen qcow2 layer");
1789
        bs->drv = NULL;
1790 1791
        return;
    }
1792

1793
    s->cipher = cipher;
1794 1795
}

K
Kevin Wolf 已提交
1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809
static size_t header_ext_add(char *buf, uint32_t magic, const void *s,
    size_t len, size_t buflen)
{
    QCowExtension *ext_backing_fmt = (QCowExtension*) buf;
    size_t ext_len = sizeof(QCowExtension) + ((len + 7) & ~7);

    if (buflen < ext_len) {
        return -ENOSPC;
    }

    *ext_backing_fmt = (QCowExtension) {
        .magic  = cpu_to_be32(magic),
        .len    = cpu_to_be32(len),
    };
1810 1811 1812 1813

    if (len) {
        memcpy(buf + sizeof(QCowExtension), s, len);
    }
K
Kevin Wolf 已提交
1814 1815 1816 1817

    return ext_len;
}

K
Kevin Wolf 已提交
1818
/*
K
Kevin Wolf 已提交
1819 1820 1821 1822
 * Updates the qcow2 header, including the variable length parts of it, i.e.
 * the backing file name and all extensions. qcow2 was not designed to allow
 * such changes, so if we run out of space (we can only use the first cluster)
 * this function may fail.
K
Kevin Wolf 已提交
1823 1824 1825
 *
 * Returns 0 on success, -errno in error cases.
 */
K
Kevin Wolf 已提交
1826
int qcow2_update_header(BlockDriverState *bs)
K
Kevin Wolf 已提交
1827
{
1828
    BDRVQcow2State *s = bs->opaque;
K
Kevin Wolf 已提交
1829 1830 1831
    QCowHeader *header;
    char *buf;
    size_t buflen = s->cluster_size;
K
Kevin Wolf 已提交
1832
    int ret;
K
Kevin Wolf 已提交
1833 1834
    uint64_t total_size;
    uint32_t refcount_table_clusters;
K
Kevin Wolf 已提交
1835
    size_t header_length;
1836
    Qcow2UnknownHeaderExtension *uext;
K
Kevin Wolf 已提交
1837

K
Kevin Wolf 已提交
1838
    buf = qemu_blockalign(bs, buflen);
K
Kevin Wolf 已提交
1839

K
Kevin Wolf 已提交
1840 1841
    /* Header structure */
    header = (QCowHeader*) buf;
K
Kevin Wolf 已提交
1842

K
Kevin Wolf 已提交
1843 1844 1845
    if (buflen < sizeof(*header)) {
        ret = -ENOSPC;
        goto fail;
K
Kevin Wolf 已提交
1846 1847
    }

K
Kevin Wolf 已提交
1848
    header_length = sizeof(*header) + s->unknown_header_fields_size;
K
Kevin Wolf 已提交
1849 1850 1851 1852
    total_size = bs->total_sectors * BDRV_SECTOR_SIZE;
    refcount_table_clusters = s->refcount_table_size >> (s->cluster_bits - 3);

    *header = (QCowHeader) {
K
Kevin Wolf 已提交
1853
        /* Version 2 fields */
K
Kevin Wolf 已提交
1854
        .magic                  = cpu_to_be32(QCOW_MAGIC),
K
Kevin Wolf 已提交
1855
        .version                = cpu_to_be32(s->qcow_version),
K
Kevin Wolf 已提交
1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866
        .backing_file_offset    = 0,
        .backing_file_size      = 0,
        .cluster_bits           = cpu_to_be32(s->cluster_bits),
        .size                   = cpu_to_be64(total_size),
        .crypt_method           = cpu_to_be32(s->crypt_method_header),
        .l1_size                = cpu_to_be32(s->l1_size),
        .l1_table_offset        = cpu_to_be64(s->l1_table_offset),
        .refcount_table_offset  = cpu_to_be64(s->refcount_table_offset),
        .refcount_table_clusters = cpu_to_be32(refcount_table_clusters),
        .nb_snapshots           = cpu_to_be32(s->nb_snapshots),
        .snapshots_offset       = cpu_to_be64(s->snapshots_offset),
K
Kevin Wolf 已提交
1867 1868 1869 1870 1871

        /* Version 3 fields */
        .incompatible_features  = cpu_to_be64(s->incompatible_features),
        .compatible_features    = cpu_to_be64(s->compatible_features),
        .autoclear_features     = cpu_to_be64(s->autoclear_features),
1872
        .refcount_order         = cpu_to_be32(s->refcount_order),
K
Kevin Wolf 已提交
1873
        .header_length          = cpu_to_be32(header_length),
K
Kevin Wolf 已提交
1874
    };
K
Kevin Wolf 已提交
1875

K
Kevin Wolf 已提交
1876 1877 1878 1879 1880 1881 1882 1883 1884
    /* For older versions, write a shorter header */
    switch (s->qcow_version) {
    case 2:
        ret = offsetof(QCowHeader, incompatible_features);
        break;
    case 3:
        ret = sizeof(*header);
        break;
    default:
1885 1886
        ret = -EINVAL;
        goto fail;
K
Kevin Wolf 已提交
1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903
    }

    buf += ret;
    buflen -= ret;
    memset(buf, 0, buflen);

    /* Preserve any unknown field in the header */
    if (s->unknown_header_fields_size) {
        if (buflen < s->unknown_header_fields_size) {
            ret = -ENOSPC;
            goto fail;
        }

        memcpy(buf, s->unknown_header_fields, s->unknown_header_fields_size);
        buf += s->unknown_header_fields_size;
        buflen -= s->unknown_header_fields_size;
    }
K
Kevin Wolf 已提交
1904

K
Kevin Wolf 已提交
1905
    /* Backing file format header extension */
1906
    if (s->image_backing_format) {
K
Kevin Wolf 已提交
1907
        ret = header_ext_add(buf, QCOW2_EXT_MAGIC_BACKING_FORMAT,
1908 1909
                             s->image_backing_format,
                             strlen(s->image_backing_format),
K
Kevin Wolf 已提交
1910 1911 1912
                             buflen);
        if (ret < 0) {
            goto fail;
K
Kevin Wolf 已提交
1913 1914
        }

K
Kevin Wolf 已提交
1915 1916
        buf += ret;
        buflen -= ret;
K
Kevin Wolf 已提交
1917 1918
    }

1919
    /* Feature table */
1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945
    if (s->qcow_version >= 3) {
        Qcow2Feature features[] = {
            {
                .type = QCOW2_FEAT_TYPE_INCOMPATIBLE,
                .bit  = QCOW2_INCOMPAT_DIRTY_BITNR,
                .name = "dirty bit",
            },
            {
                .type = QCOW2_FEAT_TYPE_INCOMPATIBLE,
                .bit  = QCOW2_INCOMPAT_CORRUPT_BITNR,
                .name = "corrupt bit",
            },
            {
                .type = QCOW2_FEAT_TYPE_COMPATIBLE,
                .bit  = QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR,
                .name = "lazy refcounts",
            },
        };

        ret = header_ext_add(buf, QCOW2_EXT_MAGIC_FEATURE_TABLE,
                             features, sizeof(features), buflen);
        if (ret < 0) {
            goto fail;
        }
        buf += ret;
        buflen -= ret;
1946 1947
    }

1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958
    /* Keep unknown header extensions */
    QLIST_FOREACH(uext, &s->unknown_header_ext, next) {
        ret = header_ext_add(buf, uext->magic, uext->data, uext->len, buflen);
        if (ret < 0) {
            goto fail;
        }

        buf += ret;
        buflen -= ret;
    }

K
Kevin Wolf 已提交
1959 1960
    /* End of header extensions */
    ret = header_ext_add(buf, QCOW2_EXT_MAGIC_END, NULL, 0, buflen);
K
Kevin Wolf 已提交
1961 1962 1963 1964
    if (ret < 0) {
        goto fail;
    }

K
Kevin Wolf 已提交
1965 1966
    buf += ret;
    buflen -= ret;
K
Kevin Wolf 已提交
1967

K
Kevin Wolf 已提交
1968
    /* Backing file name */
1969 1970
    if (s->image_backing_file) {
        size_t backing_file_len = strlen(s->image_backing_file);
K
Kevin Wolf 已提交
1971 1972 1973 1974 1975 1976

        if (buflen < backing_file_len) {
            ret = -ENOSPC;
            goto fail;
        }

1977
        /* Using strncpy is ok here, since buf is not NUL-terminated. */
1978
        strncpy(buf, s->image_backing_file, buflen);
K
Kevin Wolf 已提交
1979 1980 1981

        header->backing_file_offset = cpu_to_be64(buf - ((char*) header));
        header->backing_file_size   = cpu_to_be32(backing_file_len);
K
Kevin Wolf 已提交
1982 1983
    }

K
Kevin Wolf 已提交
1984
    /* Write the new header */
1985
    ret = bdrv_pwrite(bs->file, 0, header, s->cluster_size);
K
Kevin Wolf 已提交
1986 1987 1988 1989 1990 1991
    if (ret < 0) {
        goto fail;
    }

    ret = 0;
fail:
K
Kevin Wolf 已提交
1992
    qemu_vfree(header);
K
Kevin Wolf 已提交
1993 1994 1995 1996 1997 1998
    return ret;
}

static int qcow2_change_backing_file(BlockDriverState *bs,
    const char *backing_file, const char *backing_fmt)
{
1999
    BDRVQcow2State *s = bs->opaque;
2000

2001 2002 2003 2004
    if (backing_file && strlen(backing_file) > 1023) {
        return -EINVAL;
    }

K
Kevin Wolf 已提交
2005 2006 2007
    pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: "");
    pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: "");

2008 2009 2010 2011 2012 2013
    g_free(s->image_backing_file);
    g_free(s->image_backing_format);

    s->image_backing_file = backing_file ? g_strdup(bs->backing_file) : NULL;
    s->image_backing_format = backing_fmt ? g_strdup(bs->backing_format) : NULL;

K
Kevin Wolf 已提交
2014
    return qcow2_update_header(bs);
K
Kevin Wolf 已提交
2015 2016
}

K
Kevin Wolf 已提交
2017 2018
static int preallocate(BlockDriverState *bs)
{
K
Kevin Wolf 已提交
2019
    uint64_t bytes;
K
Kevin Wolf 已提交
2020
    uint64_t offset;
K
Kevin Wolf 已提交
2021
    uint64_t host_offset = 0;
K
Kevin Wolf 已提交
2022
    unsigned int cur_bytes;
2023
    int ret;
2024
    QCowL2Meta *meta;
K
Kevin Wolf 已提交
2025

K
Kevin Wolf 已提交
2026
    bytes = bdrv_getlength(bs);
K
Kevin Wolf 已提交
2027 2028
    offset = 0;

K
Kevin Wolf 已提交
2029 2030 2031
    while (bytes) {
        cur_bytes = MIN(bytes, INT_MAX);
        ret = qcow2_alloc_cluster_offset(bs, offset, &cur_bytes,
K
Kevin Wolf 已提交
2032
                                         &host_offset, &meta);
2033
        if (ret < 0) {
2034
            return ret;
K
Kevin Wolf 已提交
2035 2036
        }

2037 2038 2039
        while (meta) {
            QCowL2Meta *next = meta->next;

H
Hu Tao 已提交
2040 2041 2042 2043 2044 2045 2046 2047 2048
            ret = qcow2_alloc_cluster_link_l2(bs, meta);
            if (ret < 0) {
                qcow2_free_any_clusters(bs, meta->alloc_offset,
                                        meta->nb_clusters, QCOW2_DISCARD_NEVER);
                return ret;
            }

            /* There are no dependent requests, but we need to remove our
             * request from the list of in-flight requests */
2049
            QLIST_REMOVE(meta, next_in_flight);
2050 2051 2052

            g_free(meta);
            meta = next;
2053
        }
2054

K
Kevin Wolf 已提交
2055 2056
        /* TODO Preallocate data if requested */

K
Kevin Wolf 已提交
2057 2058
        bytes -= cur_bytes;
        offset += cur_bytes;
K
Kevin Wolf 已提交
2059 2060 2061 2062 2063 2064 2065
    }

    /*
     * It is expected that the image file is large enough to actually contain
     * all of the allocated clusters (otherwise we get failing reads after
     * EOF). Extend the image to the last allocated sector.
     */
K
Kevin Wolf 已提交
2066
    if (host_offset != 0) {
K
Kevin Wolf 已提交
2067
        uint8_t data = 0;
2068
        ret = bdrv_pwrite(bs->file, (host_offset + cur_bytes) - 1,
K
Kevin Wolf 已提交
2069
                          &data, 1);
2070 2071 2072
        if (ret < 0) {
            return ret;
        }
K
Kevin Wolf 已提交
2073 2074 2075 2076 2077
    }

    return 0;
}

2078 2079
static int qcow2_create2(const char *filename, int64_t total_size,
                         const char *backing_file, const char *backing_format,
2080
                         int flags, size_t cluster_size, PreallocMode prealloc,
2081
                         QemuOpts *opts, int version, int refcount_order,
M
Max Reitz 已提交
2082
                         Error **errp)
K
Kevin Wolf 已提交
2083 2084
{
    int cluster_bits;
2085 2086 2087
    QDict *options;

    /* Calculate cluster_bits */
2088
    cluster_bits = ctz32(cluster_size);
K
Kevin Wolf 已提交
2089 2090 2091
    if (cluster_bits < MIN_CLUSTER_BITS || cluster_bits > MAX_CLUSTER_BITS ||
        (1 << cluster_bits) != cluster_size)
    {
M
Max Reitz 已提交
2092 2093
        error_setg(errp, "Cluster size must be a power of two between %d and "
                   "%dk", 1 << MIN_CLUSTER_BITS, 1 << (MAX_CLUSTER_BITS - 10));
K
Kevin Wolf 已提交
2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108
        return -EINVAL;
    }

    /*
     * Open the image file and write a minimal qcow2 header.
     *
     * We keep things simple and start with a zero-sized image. We also
     * do without refcount blocks or a L1 table for now. We'll fix the
     * inconsistency later.
     *
     * We do need a refcount table because growing the refcount table means
     * allocating two new refcount blocks - the seconds of which would be at
     * 2 GB for 64k clusters, and we don't want to have a 2 GB initial file
     * size for any qcow2 image.
     */
2109
    BlockBackend *blk;
2110
    QCowHeader *header;
2111
    uint64_t* refcount_table;
M
Max Reitz 已提交
2112
    Error *local_err = NULL;
K
Kevin Wolf 已提交
2113 2114
    int ret;

2115
    if (prealloc == PREALLOC_MODE_FULL || prealloc == PREALLOC_MODE_FALLOC) {
2116 2117 2118 2119 2120
        /* Note: The following calculation does not need to be exact; if it is a
         * bit off, either some bytes will be "leaked" (which is fine) or we
         * will need to increase the file size by some bytes (which is fine,
         * too, as long as the bulk is allocated here). Therefore, using
         * floating point arithmetic is fine. */
2121 2122 2123
        int64_t meta_size = 0;
        uint64_t nreftablee, nrefblocke, nl1e, nl2e;
        int64_t aligned_total_size = align_offset(total_size, cluster_size);
2124 2125 2126 2127 2128 2129 2130
        int refblock_bits, refblock_size;
        /* refcount entry size in bytes */
        double rces = (1 << refcount_order) / 8.;

        /* see qcow2_open() */
        refblock_bits = cluster_bits - (refcount_order - 3);
        refblock_size = 1 << refblock_bits;
2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154

        /* header: 1 cluster */
        meta_size += cluster_size;

        /* total size of L2 tables */
        nl2e = aligned_total_size / cluster_size;
        nl2e = align_offset(nl2e, cluster_size / sizeof(uint64_t));
        meta_size += nl2e * sizeof(uint64_t);

        /* total size of L1 tables */
        nl1e = nl2e * sizeof(uint64_t) / cluster_size;
        nl1e = align_offset(nl1e, cluster_size / sizeof(uint64_t));
        meta_size += nl1e * sizeof(uint64_t);

        /* total size of refcount blocks
         *
         * note: every host cluster is reference-counted, including metadata
         * (even refcount blocks are recursively included).
         * Let:
         *   a = total_size (this is the guest disk size)
         *   m = meta size not including refcount blocks and refcount tables
         *   c = cluster size
         *   y1 = number of refcount blocks entries
         *   y2 = meta size including everything
2155
         *   rces = refcount entry size in bytes
2156 2157
         * then,
         *   y1 = (y2 + a)/c
2158
         *   y2 = y1 * rces + y1 * rces * sizeof(u64) / c + m
2159
         * we can get y1:
2160
         *   y1 = (a + m) / (c - rces - rces * sizeof(u64) / c)
2161
         */
2162 2163 2164 2165
        nrefblocke = (aligned_total_size + meta_size + cluster_size)
                   / (cluster_size - rces - rces * sizeof(uint64_t)
                                                 / cluster_size);
        meta_size += DIV_ROUND_UP(nrefblocke, refblock_size) * cluster_size;
2166 2167

        /* total size of refcount tables */
2168
        nreftablee = nrefblocke / refblock_size;
2169 2170 2171 2172
        nreftablee = align_offset(nreftablee, cluster_size / sizeof(uint64_t));
        meta_size += nreftablee * sizeof(uint64_t);

        qemu_opt_set_number(opts, BLOCK_OPT_SIZE,
2173
                            aligned_total_size + meta_size, &error_abort);
2174 2175
        qemu_opt_set(opts, BLOCK_OPT_PREALLOC, PreallocMode_lookup[prealloc],
                     &error_abort);
2176 2177
    }

C
Chunyan Liu 已提交
2178
    ret = bdrv_create_file(filename, opts, &local_err);
K
Kevin Wolf 已提交
2179
    if (ret < 0) {
M
Max Reitz 已提交
2180
        error_propagate(errp, local_err);
K
Kevin Wolf 已提交
2181 2182 2183
        return ret;
    }

2184
    blk = blk_new_open(filename, NULL, NULL,
2185
                       BDRV_O_RDWR | BDRV_O_PROTOCOL, &local_err);
2186
    if (blk == NULL) {
M
Max Reitz 已提交
2187
        error_propagate(errp, local_err);
2188
        return -EIO;
K
Kevin Wolf 已提交
2189 2190
    }

2191 2192
    blk_set_allow_write_beyond_eof(blk, true);

K
Kevin Wolf 已提交
2193
    /* Write the header */
2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204
    QEMU_BUILD_BUG_ON((1 << MIN_CLUSTER_BITS) < sizeof(*header));
    header = g_malloc0(cluster_size);
    *header = (QCowHeader) {
        .magic                      = cpu_to_be32(QCOW_MAGIC),
        .version                    = cpu_to_be32(version),
        .cluster_bits               = cpu_to_be32(cluster_bits),
        .size                       = cpu_to_be64(0),
        .l1_table_offset            = cpu_to_be64(0),
        .l1_size                    = cpu_to_be32(0),
        .refcount_table_offset      = cpu_to_be64(cluster_size),
        .refcount_table_clusters    = cpu_to_be32(1),
2205
        .refcount_order             = cpu_to_be32(refcount_order),
2206 2207
        .header_length              = cpu_to_be32(sizeof(*header)),
    };
K
Kevin Wolf 已提交
2208 2209

    if (flags & BLOCK_FLAG_ENCRYPT) {
2210
        header->crypt_method = cpu_to_be32(QCOW_CRYPT_AES);
K
Kevin Wolf 已提交
2211
    } else {
2212
        header->crypt_method = cpu_to_be32(QCOW_CRYPT_NONE);
K
Kevin Wolf 已提交
2213 2214
    }

2215
    if (flags & BLOCK_FLAG_LAZY_REFCOUNTS) {
2216
        header->compatible_features |=
2217 2218 2219
            cpu_to_be64(QCOW2_COMPAT_LAZY_REFCOUNTS);
    }

2220
    ret = blk_pwrite(blk, 0, header, cluster_size, 0);
2221
    g_free(header);
K
Kevin Wolf 已提交
2222
    if (ret < 0) {
M
Max Reitz 已提交
2223
        error_setg_errno(errp, -ret, "Could not write qcow2 header");
K
Kevin Wolf 已提交
2224 2225 2226
        goto out;
    }

2227 2228 2229
    /* Write a refcount table with one refcount block */
    refcount_table = g_malloc0(2 * cluster_size);
    refcount_table[0] = cpu_to_be64(2 * cluster_size);
2230
    ret = blk_pwrite(blk, cluster_size, refcount_table, 2 * cluster_size, 0);
2231
    g_free(refcount_table);
K
Kevin Wolf 已提交
2232 2233

    if (ret < 0) {
M
Max Reitz 已提交
2234
        error_setg_errno(errp, -ret, "Could not write refcount table");
K
Kevin Wolf 已提交
2235 2236 2237
        goto out;
    }

2238 2239
    blk_unref(blk);
    blk = NULL;
K
Kevin Wolf 已提交
2240 2241 2242 2243 2244 2245

    /*
     * And now open the image and make it consistent first (i.e. increase the
     * refcount of the cluster that is occupied by the header and the refcount
     * table)
     */
2246 2247
    options = qdict_new();
    qdict_put(options, "driver", qstring_from_str("qcow2"));
2248
    blk = blk_new_open(filename, NULL, options,
2249
                       BDRV_O_RDWR | BDRV_O_NO_FLUSH, &local_err);
2250
    if (blk == NULL) {
M
Max Reitz 已提交
2251
        error_propagate(errp, local_err);
2252
        ret = -EIO;
K
Kevin Wolf 已提交
2253 2254 2255
        goto out;
    }

2256
    ret = qcow2_alloc_clusters(blk_bs(blk), 3 * cluster_size);
K
Kevin Wolf 已提交
2257
    if (ret < 0) {
M
Max Reitz 已提交
2258 2259
        error_setg_errno(errp, -ret, "Could not allocate clusters for qcow2 "
                         "header and refcount table");
K
Kevin Wolf 已提交
2260 2261 2262 2263 2264 2265 2266
        goto out;

    } else if (ret != 0) {
        error_report("Huh, first cluster in empty image is already in use?");
        abort();
    }

2267
    /* Create a full header (including things like feature table) */
2268
    ret = qcow2_update_header(blk_bs(blk));
2269 2270 2271 2272 2273
    if (ret < 0) {
        error_setg_errno(errp, -ret, "Could not update qcow2 header");
        goto out;
    }

K
Kevin Wolf 已提交
2274
    /* Okay, now that we have a valid image, let's give it the right size */
2275
    ret = blk_truncate(blk, total_size);
K
Kevin Wolf 已提交
2276
    if (ret < 0) {
M
Max Reitz 已提交
2277
        error_setg_errno(errp, -ret, "Could not resize image");
K
Kevin Wolf 已提交
2278 2279 2280 2281 2282
        goto out;
    }

    /* Want a backing file? There you go.*/
    if (backing_file) {
2283
        ret = bdrv_change_backing_file(blk_bs(blk), backing_file, backing_format);
K
Kevin Wolf 已提交
2284
        if (ret < 0) {
M
Max Reitz 已提交
2285 2286
            error_setg_errno(errp, -ret, "Could not assign backing file '%s' "
                             "with format '%s'", backing_file, backing_format);
K
Kevin Wolf 已提交
2287 2288 2289 2290 2291
            goto out;
        }
    }

    /* And if we're supposed to preallocate metadata, do that now */
2292
    if (prealloc != PREALLOC_MODE_OFF) {
2293
        BDRVQcow2State *s = blk_bs(blk)->opaque;
Z
Zhi Yong Wu 已提交
2294
        qemu_co_mutex_lock(&s->lock);
2295
        ret = preallocate(blk_bs(blk));
Z
Zhi Yong Wu 已提交
2296
        qemu_co_mutex_unlock(&s->lock);
K
Kevin Wolf 已提交
2297
        if (ret < 0) {
M
Max Reitz 已提交
2298
            error_setg_errno(errp, -ret, "Could not preallocate metadata");
K
Kevin Wolf 已提交
2299 2300 2301 2302
            goto out;
        }
    }

2303 2304
    blk_unref(blk);
    blk = NULL;
M
Max Reitz 已提交
2305 2306

    /* Reopen the image without BDRV_O_NO_FLUSH to flush it before returning */
2307 2308
    options = qdict_new();
    qdict_put(options, "driver", qstring_from_str("qcow2"));
2309
    blk = blk_new_open(filename, NULL, options,
2310
                       BDRV_O_RDWR | BDRV_O_NO_BACKING, &local_err);
2311
    if (blk == NULL) {
M
Max Reitz 已提交
2312
        error_propagate(errp, local_err);
2313
        ret = -EIO;
M
Max Reitz 已提交
2314 2315 2316
        goto out;
    }

K
Kevin Wolf 已提交
2317 2318
    ret = 0;
out:
2319 2320
    if (blk) {
        blk_unref(blk);
2321
    }
K
Kevin Wolf 已提交
2322 2323
    return ret;
}
K
Kevin Wolf 已提交
2324

2325
static int qcow2_create(const char *filename, QemuOpts *opts, Error **errp)
K
Kevin Wolf 已提交
2326
{
2327 2328 2329
    char *backing_file = NULL;
    char *backing_fmt = NULL;
    char *buf = NULL;
2330
    uint64_t size = 0;
K
Kevin Wolf 已提交
2331
    int flags = 0;
2332
    size_t cluster_size = DEFAULT_CLUSTER_SIZE;
2333
    PreallocMode prealloc;
2334
    int version = 3;
2335 2336
    uint64_t refcount_bits = 16;
    int refcount_order;
M
Max Reitz 已提交
2337 2338
    Error *local_err = NULL;
    int ret;
K
Kevin Wolf 已提交
2339 2340

    /* Read out options */
2341 2342
    size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0),
                    BDRV_SECTOR_SIZE);
2343 2344 2345 2346 2347 2348 2349 2350
    backing_file = qemu_opt_get_del(opts, BLOCK_OPT_BACKING_FILE);
    backing_fmt = qemu_opt_get_del(opts, BLOCK_OPT_BACKING_FMT);
    if (qemu_opt_get_bool_del(opts, BLOCK_OPT_ENCRYPT, false)) {
        flags |= BLOCK_FLAG_ENCRYPT;
    }
    cluster_size = qemu_opt_get_size_del(opts, BLOCK_OPT_CLUSTER_SIZE,
                                         DEFAULT_CLUSTER_SIZE);
    buf = qemu_opt_get_del(opts, BLOCK_OPT_PREALLOC);
2351
    prealloc = qapi_enum_parse(PreallocMode_lookup, buf,
2352
                               PREALLOC_MODE__MAX, PREALLOC_MODE_OFF,
2353 2354 2355
                               &local_err);
    if (local_err) {
        error_propagate(errp, local_err);
2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374
        ret = -EINVAL;
        goto finish;
    }
    g_free(buf);
    buf = qemu_opt_get_del(opts, BLOCK_OPT_COMPAT_LEVEL);
    if (!buf) {
        /* keep the default */
    } else if (!strcmp(buf, "0.10")) {
        version = 2;
    } else if (!strcmp(buf, "1.1")) {
        version = 3;
    } else {
        error_setg(errp, "Invalid compatibility level: '%s'", buf);
        ret = -EINVAL;
        goto finish;
    }

    if (qemu_opt_get_bool_del(opts, BLOCK_OPT_LAZY_REFCOUNTS, false)) {
        flags |= BLOCK_FLAG_LAZY_REFCOUNTS;
K
Kevin Wolf 已提交
2375 2376
    }

2377
    if (backing_file && prealloc != PREALLOC_MODE_OFF) {
M
Max Reitz 已提交
2378 2379
        error_setg(errp, "Backing file and preallocation cannot be used at "
                   "the same time");
2380 2381
        ret = -EINVAL;
        goto finish;
K
Kevin Wolf 已提交
2382 2383
    }

2384
    if (version < 3 && (flags & BLOCK_FLAG_LAZY_REFCOUNTS)) {
M
Max Reitz 已提交
2385 2386
        error_setg(errp, "Lazy refcounts only supported with compatibility "
                   "level 1.1 and above (use compat=1.1 or greater)");
2387 2388
        ret = -EINVAL;
        goto finish;
2389 2390
    }

2391 2392 2393 2394 2395 2396 2397 2398 2399
    refcount_bits = qemu_opt_get_number_del(opts, BLOCK_OPT_REFCOUNT_BITS,
                                            refcount_bits);
    if (refcount_bits > 64 || !is_power_of_2(refcount_bits)) {
        error_setg(errp, "Refcount width must be a power of two and may not "
                   "exceed 64 bits");
        ret = -EINVAL;
        goto finish;
    }

2400 2401 2402 2403 2404 2405 2406 2407
    if (version < 3 && refcount_bits != 16) {
        error_setg(errp, "Different refcount widths than 16 bits require "
                   "compatibility level 1.1 or above (use compat=1.1 or "
                   "greater)");
        ret = -EINVAL;
        goto finish;
    }

2408
    refcount_order = ctz32(refcount_bits);
2409

2410
    ret = qcow2_create2(filename, size, backing_file, backing_fmt, flags,
2411 2412
                        cluster_size, prealloc, opts, version, refcount_order,
                        &local_err);
2413
    error_propagate(errp, local_err);
2414 2415 2416 2417 2418

finish:
    g_free(backing_file);
    g_free(backing_fmt);
    g_free(buf);
M
Max Reitz 已提交
2419
    return ret;
K
Kevin Wolf 已提交
2420 2421
}

2422

2423 2424
static bool is_zero_sectors(BlockDriverState *bs, int64_t start,
                            uint32_t count)
2425 2426 2427
{
    int nr;
    BlockDriverState *file;
2428
    int64_t res;
2429

2430 2431 2432 2433 2434 2435
    if (!count) {
        return true;
    }
    res = bdrv_get_block_status_above(bs, NULL, start, count,
                                      &nr, &file);
    return res >= 0 && (res & BDRV_BLOCK_ZERO) && nr == count;
2436 2437
}

2438 2439
static coroutine_fn int qcow2_co_pwrite_zeroes(BlockDriverState *bs,
    int64_t offset, int count, BdrvRequestFlags flags)
K
Kevin Wolf 已提交
2440 2441
{
    int ret;
2442
    BDRVQcow2State *s = bs->opaque;
K
Kevin Wolf 已提交
2443

2444 2445
    uint32_t head = offset % s->cluster_size;
    uint32_t tail = (offset + count) % s->cluster_size;
2446

2447
    trace_qcow2_pwrite_zeroes_start_req(qemu_coroutine_self(), offset, count);
2448

2449
    if (head || tail) {
2450
        int64_t cl_start = (offset - head) >> BDRV_SECTOR_BITS;
2451
        uint64_t off;
K
Kevin Wolf 已提交
2452
        unsigned int nr;
2453

2454
        assert(head + count <= s->cluster_size);
2455

2456
        /* check whether remainder of cluster already reads as zero */
2457 2458 2459 2460 2461
        if (!(is_zero_sectors(bs, cl_start,
                              DIV_ROUND_UP(head, BDRV_SECTOR_SIZE)) &&
              is_zero_sectors(bs, (offset + count) >> BDRV_SECTOR_BITS,
                              DIV_ROUND_UP(-tail & (s->cluster_size - 1),
                                           BDRV_SECTOR_SIZE)))) {
2462 2463 2464 2465 2466
            return -ENOTSUP;
        }

        qemu_co_mutex_lock(&s->lock);
        /* We can have new write after previous check */
2467 2468
        offset = cl_start << BDRV_SECTOR_BITS;
        count = s->cluster_size;
K
Kevin Wolf 已提交
2469
        nr = s->cluster_size;
2470
        ret = qcow2_get_cluster_offset(bs, offset, &nr, &off);
2471
        if (ret != QCOW2_CLUSTER_UNALLOCATED && ret != QCOW2_CLUSTER_ZERO) {
2472 2473 2474 2475 2476
            qemu_co_mutex_unlock(&s->lock);
            return -ENOTSUP;
        }
    } else {
        qemu_co_mutex_lock(&s->lock);
K
Kevin Wolf 已提交
2477 2478
    }

2479
    trace_qcow2_pwrite_zeroes(qemu_coroutine_self(), offset, count);
2480

K
Kevin Wolf 已提交
2481
    /* Whatever is left can use real zero clusters */
F
Fam Zheng 已提交
2482
    ret = qcow2_zero_clusters(bs, offset, count >> BDRV_SECTOR_BITS, flags);
K
Kevin Wolf 已提交
2483 2484 2485 2486 2487
    qemu_co_mutex_unlock(&s->lock);

    return ret;
}

2488 2489
static coroutine_fn int qcow2_co_pdiscard(BlockDriverState *bs,
                                          int64_t offset, int count)
K
Kevin Wolf 已提交
2490
{
2491
    int ret;
2492
    BDRVQcow2State *s = bs->opaque;
2493 2494

    qemu_co_mutex_lock(&s->lock);
2495 2496
    ret = qcow2_discard_clusters(bs, offset, count >> BDRV_SECTOR_BITS,
                                 QCOW2_DISCARD_REQUEST, false);
2497 2498
    qemu_co_mutex_unlock(&s->lock);
    return ret;
K
Kevin Wolf 已提交
2499 2500
}

2501 2502
static int qcow2_truncate(BlockDriverState *bs, int64_t offset)
{
2503
    BDRVQcow2State *s = bs->opaque;
2504 2505
    int64_t new_l1_size;
    int ret;
2506 2507

    if (offset & 511) {
2508
        error_report("The new size must be a multiple of 512");
2509 2510 2511 2512 2513
        return -EINVAL;
    }

    /* cannot proceed if image has snapshots */
    if (s->nb_snapshots) {
2514
        error_report("Can't resize an image which has snapshots");
2515 2516 2517 2518 2519
        return -ENOTSUP;
    }

    /* shrinking is currently not supported */
    if (offset < bs->total_sectors * 512) {
2520
        error_report("qcow2 doesn't support shrinking images yet");
2521 2522 2523 2524
        return -ENOTSUP;
    }

    new_l1_size = size_to_l1(s, offset);
2525
    ret = qcow2_grow_l1_table(bs, new_l1_size, true);
2526 2527 2528 2529 2530 2531
    if (ret < 0) {
        return ret;
    }

    /* write updated header.size */
    offset = cpu_to_be64(offset);
2532
    ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader, size),
2533
                           &offset, sizeof(uint64_t));
2534 2535 2536 2537 2538 2539 2540 2541
    if (ret < 0) {
        return ret;
    }

    s->l1_vm_state_index = new_l1_size;
    return 0;
}

B
Blue Swirl 已提交
2542 2543
/* XXX: put compressed sectors first, then all the cluster aligned
   tables to avoid losing bytes in alignment */
2544 2545 2546
static coroutine_fn int
qcow2_co_pwritev_compressed(BlockDriverState *bs, uint64_t offset,
                            uint64_t bytes, QEMUIOVector *qiov)
B
Blue Swirl 已提交
2547
{
2548
    BDRVQcow2State *s = bs->opaque;
2549 2550
    QEMUIOVector hd_qiov;
    struct iovec iov;
B
Blue Swirl 已提交
2551 2552
    z_stream strm;
    int ret, out_len;
2553
    uint8_t *buf, *out_buf;
B
Blue Swirl 已提交
2554 2555
    uint64_t cluster_offset;

2556
    if (bytes == 0) {
B
Blue Swirl 已提交
2557 2558
        /* align end of file to a sector boundary to ease reading with
           sector based I/Os */
K
Kevin Wolf 已提交
2559 2560
        cluster_offset = bdrv_getlength(bs->file->bs);
        return bdrv_truncate(bs->file->bs, cluster_offset);
B
Blue Swirl 已提交
2561 2562
    }

2563
    buf = qemu_blockalign(bs, s->cluster_size);
2564
    if (bytes != s->cluster_size) {
2565 2566
        if (bytes > s->cluster_size ||
            offset + bytes != bs->total_sectors << BDRV_SECTOR_BITS)
2567
        {
2568 2569
            qemu_vfree(buf);
            return -EINVAL;
2570
        }
2571 2572
        /* Zero-pad last write if image size is not cluster aligned */
        memset(buf + bytes, 0, s->cluster_size - bytes);
2573
    }
2574
    qemu_iovec_to_buf(qiov, 0, buf, bytes);
B
Blue Swirl 已提交
2575

2576
    out_buf = g_malloc(s->cluster_size);
B
Blue Swirl 已提交
2577 2578 2579 2580 2581 2582 2583

    /* best compression, small window, no zlib header */
    memset(&strm, 0, sizeof(strm));
    ret = deflateInit2(&strm, Z_DEFAULT_COMPRESSION,
                       Z_DEFLATED, -12,
                       9, Z_DEFAULT_STRATEGY);
    if (ret != 0) {
2584 2585
        ret = -EINVAL;
        goto fail;
B
Blue Swirl 已提交
2586 2587 2588 2589 2590 2591 2592 2593 2594 2595
    }

    strm.avail_in = s->cluster_size;
    strm.next_in = (uint8_t *)buf;
    strm.avail_out = s->cluster_size;
    strm.next_out = out_buf;

    ret = deflate(&strm, Z_FINISH);
    if (ret != Z_STREAM_END && ret != Z_OK) {
        deflateEnd(&strm);
2596 2597
        ret = -EINVAL;
        goto fail;
B
Blue Swirl 已提交
2598 2599 2600 2601 2602 2603 2604
    }
    out_len = strm.next_out - out_buf;

    deflateEnd(&strm);

    if (ret != Z_STREAM_END || out_len >= s->cluster_size) {
        /* could not compress: write normal cluster */
2605
        ret = qcow2_co_pwritev(bs, offset, bytes, qiov, 0);
2606 2607 2608
        if (ret < 0) {
            goto fail;
        }
2609 2610
        goto success;
    }
2611

2612 2613 2614 2615 2616 2617 2618 2619 2620
    qemu_co_mutex_lock(&s->lock);
    cluster_offset =
        qcow2_alloc_compressed_cluster_offset(bs, offset, out_len);
    if (!cluster_offset) {
        qemu_co_mutex_unlock(&s->lock);
        ret = -EIO;
        goto fail;
    }
    cluster_offset &= s->cluster_offset_mask;
2621

2622 2623 2624 2625
    ret = qcow2_pre_write_overlap_check(bs, 0, cluster_offset, out_len);
    qemu_co_mutex_unlock(&s->lock);
    if (ret < 0) {
        goto fail;
B
Blue Swirl 已提交
2626 2627
    }

2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639
    iov = (struct iovec) {
        .iov_base   = out_buf,
        .iov_len    = out_len,
    };
    qemu_iovec_init_external(&hd_qiov, &iov, 1);

    BLKDBG_EVENT(bs->file, BLKDBG_WRITE_COMPRESSED);
    ret = bdrv_co_pwritev(bs->file, cluster_offset, out_len, &hd_qiov, 0);
    if (ret < 0) {
        goto fail;
    }
success:
2640 2641
    ret = 0;
fail:
2642
    qemu_vfree(buf);
2643
    g_free(out_buf);
2644
    return ret;
B
Blue Swirl 已提交
2645 2646
}

M
Max Reitz 已提交
2647 2648
static int make_completely_empty(BlockDriverState *bs)
{
2649
    BDRVQcow2State *s = bs->opaque;
M
Max Reitz 已提交
2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683
    int ret, l1_clusters;
    int64_t offset;
    uint64_t *new_reftable = NULL;
    uint64_t rt_entry, l1_size2;
    struct {
        uint64_t l1_offset;
        uint64_t reftable_offset;
        uint32_t reftable_clusters;
    } QEMU_PACKED l1_ofs_rt_ofs_cls;

    ret = qcow2_cache_empty(bs, s->l2_table_cache);
    if (ret < 0) {
        goto fail;
    }

    ret = qcow2_cache_empty(bs, s->refcount_block_cache);
    if (ret < 0) {
        goto fail;
    }

    /* Refcounts will be broken utterly */
    ret = qcow2_mark_dirty(bs);
    if (ret < 0) {
        goto fail;
    }

    BLKDBG_EVENT(bs->file, BLKDBG_L1_UPDATE);

    l1_clusters = DIV_ROUND_UP(s->l1_size, s->cluster_size / sizeof(uint64_t));
    l1_size2 = (uint64_t)s->l1_size * sizeof(uint64_t);

    /* After this call, neither the in-memory nor the on-disk refcount
     * information accurately describe the actual references */

2684
    ret = bdrv_pwrite_zeroes(bs->file, s->l1_table_offset,
2685
                             l1_clusters * s->cluster_size, 0);
M
Max Reitz 已提交
2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697
    if (ret < 0) {
        goto fail_broken_refcounts;
    }
    memset(s->l1_table, 0, l1_size2);

    BLKDBG_EVENT(bs->file, BLKDBG_EMPTY_IMAGE_PREPARE);

    /* Overwrite enough clusters at the beginning of the sectors to place
     * the refcount table, a refcount block and the L1 table in; this may
     * overwrite parts of the existing refcount and L1 table, which is not
     * an issue because the dirty flag is set, complete data loss is in fact
     * desired and partial data loss is consequently fine as well */
2698
    ret = bdrv_pwrite_zeroes(bs->file, s->cluster_size,
2699
                             (2 + l1_clusters) * s->cluster_size, 0);
M
Max Reitz 已提交
2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713
    /* This call (even if it failed overall) may have overwritten on-disk
     * refcount structures; in that case, the in-memory refcount information
     * will probably differ from the on-disk information which makes the BDS
     * unusable */
    if (ret < 0) {
        goto fail_broken_refcounts;
    }

    BLKDBG_EVENT(bs->file, BLKDBG_L1_UPDATE);
    BLKDBG_EVENT(bs->file, BLKDBG_REFTABLE_UPDATE);

    /* "Create" an empty reftable (one cluster) directly after the image
     * header and an empty L1 table three clusters after the image header;
     * the cluster between those two will be used as the first refblock */
2714 2715 2716
    l1_ofs_rt_ofs_cls.l1_offset = cpu_to_be64(3 * s->cluster_size);
    l1_ofs_rt_ofs_cls.reftable_offset = cpu_to_be64(s->cluster_size);
    l1_ofs_rt_ofs_cls.reftable_clusters = cpu_to_be32(1);
2717
    ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader, l1_table_offset),
M
Max Reitz 已提交
2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747
                           &l1_ofs_rt_ofs_cls, sizeof(l1_ofs_rt_ofs_cls));
    if (ret < 0) {
        goto fail_broken_refcounts;
    }

    s->l1_table_offset = 3 * s->cluster_size;

    new_reftable = g_try_new0(uint64_t, s->cluster_size / sizeof(uint64_t));
    if (!new_reftable) {
        ret = -ENOMEM;
        goto fail_broken_refcounts;
    }

    s->refcount_table_offset = s->cluster_size;
    s->refcount_table_size   = s->cluster_size / sizeof(uint64_t);

    g_free(s->refcount_table);
    s->refcount_table = new_reftable;
    new_reftable = NULL;

    /* Now the in-memory refcount information again corresponds to the on-disk
     * information (reftable is empty and no refblocks (the refblock cache is
     * empty)); however, this means some clusters (e.g. the image header) are
     * referenced, but not refcounted, but the normal qcow2 code assumes that
     * the in-memory information is always correct */

    BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC);

    /* Enter the first refblock into the reftable */
    rt_entry = cpu_to_be64(2 * s->cluster_size);
2748
    ret = bdrv_pwrite_sync(bs->file, s->cluster_size,
M
Max Reitz 已提交
2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772
                           &rt_entry, sizeof(rt_entry));
    if (ret < 0) {
        goto fail_broken_refcounts;
    }
    s->refcount_table[0] = 2 * s->cluster_size;

    s->free_cluster_index = 0;
    assert(3 + l1_clusters <= s->refcount_block_size);
    offset = qcow2_alloc_clusters(bs, 3 * s->cluster_size + l1_size2);
    if (offset < 0) {
        ret = offset;
        goto fail_broken_refcounts;
    } else if (offset > 0) {
        error_report("First cluster in emptied image is in use");
        abort();
    }

    /* Now finally the in-memory information corresponds to the on-disk
     * structures and is correct */
    ret = qcow2_mark_clean(bs);
    if (ret < 0) {
        goto fail;
    }

K
Kevin Wolf 已提交
2773
    ret = bdrv_truncate(bs->file->bs, (3 + l1_clusters) * s->cluster_size);
M
Max Reitz 已提交
2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793
    if (ret < 0) {
        goto fail;
    }

    return 0;

fail_broken_refcounts:
    /* The BDS is unusable at this point. If we wanted to make it usable, we
     * would have to call qcow2_refcount_close(), qcow2_refcount_init(),
     * qcow2_check_refcounts(), qcow2_refcount_close() and qcow2_refcount_init()
     * again. However, because the functions which could have caused this error
     * path to be taken are used by those functions as well, it's very likely
     * that that sequence will fail as well. Therefore, just eject the BDS. */
    bs->drv = NULL;

fail:
    g_free(new_reftable);
    return ret;
}

M
Max Reitz 已提交
2794 2795
static int qcow2_make_empty(BlockDriverState *bs)
{
2796
    BDRVQcow2State *s = bs->opaque;
M
Max Reitz 已提交
2797 2798
    uint64_t start_sector;
    int sector_step = INT_MAX / BDRV_SECTOR_SIZE;
M
Max Reitz 已提交
2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811
    int l1_clusters, ret = 0;

    l1_clusters = DIV_ROUND_UP(s->l1_size, s->cluster_size / sizeof(uint64_t));

    if (s->qcow_version >= 3 && !s->snapshots &&
        3 + l1_clusters <= s->refcount_block_size) {
        /* The following function only works for qcow2 v3 images (it requires
         * the dirty flag) and only as long as there are no snapshots (because
         * it completely empties the image). Furthermore, the L1 table and three
         * additional clusters (image header, refcount table, one refcount
         * block) have to fit inside one refcount block. */
        return make_completely_empty(bs);
    }
M
Max Reitz 已提交
2812

M
Max Reitz 已提交
2813 2814
    /* This fallback code simply discards every active cluster; this is slow,
     * but works in all cases */
M
Max Reitz 已提交
2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834
    for (start_sector = 0; start_sector < bs->total_sectors;
         start_sector += sector_step)
    {
        /* As this function is generally used after committing an external
         * snapshot, QCOW2_DISCARD_SNAPSHOT seems appropriate. Also, the
         * default action for this kind of discard is to pass the discard,
         * which will ideally result in an actually smaller image file, as
         * is probably desired. */
        ret = qcow2_discard_clusters(bs, start_sector * BDRV_SECTOR_SIZE,
                                     MIN(sector_step,
                                         bs->total_sectors - start_sector),
                                     QCOW2_DISCARD_SNAPSHOT, true);
        if (ret < 0) {
            break;
        }
    }

    return ret;
}

2835
static coroutine_fn int qcow2_co_flush_to_os(BlockDriverState *bs)
B
Blue Swirl 已提交
2836
{
2837
    BDRVQcow2State *s = bs->opaque;
K
Kevin Wolf 已提交
2838 2839
    int ret;

P
Paolo Bonzini 已提交
2840
    qemu_co_mutex_lock(&s->lock);
2841
    ret = qcow2_cache_write(bs, s->l2_table_cache);
K
Kevin Wolf 已提交
2842
    if (ret < 0) {
D
Dong Xu Wang 已提交
2843
        qemu_co_mutex_unlock(&s->lock);
P
Paolo Bonzini 已提交
2844
        return ret;
K
Kevin Wolf 已提交
2845 2846
    }

2847
    if (qcow2_need_accurate_refcounts(s)) {
2848
        ret = qcow2_cache_write(bs, s->refcount_block_cache);
2849 2850 2851 2852
        if (ret < 0) {
            qemu_co_mutex_unlock(&s->lock);
            return ret;
        }
K
Kevin Wolf 已提交
2853
    }
P
Paolo Bonzini 已提交
2854
    qemu_co_mutex_unlock(&s->lock);
K
Kevin Wolf 已提交
2855

K
Kevin Wolf 已提交
2856 2857 2858
    return 0;
}

2859
static int qcow2_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
B
Blue Swirl 已提交
2860
{
2861
    BDRVQcow2State *s = bs->opaque;
2862 2863
    bdi->unallocated_blocks_are_zero = true;
    bdi->can_write_zeroes_with_unmap = (s->qcow_version >= 3);
B
Blue Swirl 已提交
2864
    bdi->cluster_size = s->cluster_size;
2865
    bdi->vm_state_offset = qcow2_vm_state_offset(s);
B
Blue Swirl 已提交
2866 2867 2868
    return 0;
}

2869 2870
static ImageInfoSpecific *qcow2_get_specific_info(BlockDriverState *bs)
{
2871
    BDRVQcow2State *s = bs->opaque;
2872 2873 2874
    ImageInfoSpecific *spec_info = g_new(ImageInfoSpecific, 1);

    *spec_info = (ImageInfoSpecific){
2875
        .type  = IMAGE_INFO_SPECIFIC_KIND_QCOW2,
2876
        .u.qcow2.data = g_new(ImageInfoSpecificQCow2, 1),
2877 2878
    };
    if (s->qcow_version == 2) {
2879
        *spec_info->u.qcow2.data = (ImageInfoSpecificQCow2){
2880 2881
            .compat             = g_strdup("0.10"),
            .refcount_bits      = s->refcount_bits,
2882 2883
        };
    } else if (s->qcow_version == 3) {
2884
        *spec_info->u.qcow2.data = (ImageInfoSpecificQCow2){
2885 2886 2887 2888
            .compat             = g_strdup("1.1"),
            .lazy_refcounts     = s->compatible_features &
                                  QCOW2_COMPAT_LAZY_REFCOUNTS,
            .has_lazy_refcounts = true,
2889 2890 2891
            .corrupt            = s->incompatible_features &
                                  QCOW2_INCOMPAT_CORRUPT,
            .has_corrupt        = true,
2892
            .refcount_bits      = s->refcount_bits,
2893
        };
2894 2895 2896 2897
    } else {
        /* if this assertion fails, this probably means a new version was
         * added without having it covered here */
        assert(false);
2898 2899 2900 2901 2902
    }

    return spec_info;
}

B
Blue Swirl 已提交
2903 2904 2905
#if 0
static void dump_refcounts(BlockDriverState *bs)
{
2906
    BDRVQcow2State *s = bs->opaque;
B
Blue Swirl 已提交
2907 2908 2909
    int64_t nb_clusters, k, k1, size;
    int refcount;

K
Kevin Wolf 已提交
2910
    size = bdrv_getlength(bs->file->bs);
B
Blue Swirl 已提交
2911 2912 2913 2914 2915 2916 2917
    nb_clusters = size_to_clusters(s, size);
    for(k = 0; k < nb_clusters;) {
        k1 = k;
        refcount = get_refcount(bs, k);
        k++;
        while (k < nb_clusters && get_refcount(bs, k) == refcount)
            k++;
B
Blue Swirl 已提交
2918 2919
        printf("%" PRId64 ": refcount=%d nb=%" PRId64 "\n", k, refcount,
               k - k1);
B
Blue Swirl 已提交
2920 2921 2922 2923
    }
}
#endif

2924 2925
static int qcow2_save_vmstate(BlockDriverState *bs, QEMUIOVector *qiov,
                              int64_t pos)
B
Blue Swirl 已提交
2926
{
2927
    BDRVQcow2State *s = bs->opaque;
B
Blue Swirl 已提交
2928

2929
    BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_SAVE);
2930 2931
    return bs->drv->bdrv_co_pwritev(bs, qcow2_vm_state_offset(s) + pos,
                                    qiov->size, qiov, 0);
B
Blue Swirl 已提交
2932 2933
}

2934 2935
static int qcow2_load_vmstate(BlockDriverState *bs, QEMUIOVector *qiov,
                              int64_t pos)
B
Blue Swirl 已提交
2936
{
2937
    BDRVQcow2State *s = bs->opaque;
B
Blue Swirl 已提交
2938

2939
    BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_LOAD);
2940 2941
    return bs->drv->bdrv_co_preadv(bs, qcow2_vm_state_offset(s) + pos,
                                   qiov->size, qiov, 0);
B
Blue Swirl 已提交
2942 2943
}

M
Max Reitz 已提交
2944 2945 2946 2947
/*
 * Downgrades an image's version. To achieve this, any incompatible features
 * have to be removed.
 */
2948
static int qcow2_downgrade(BlockDriverState *bs, int target_version,
2949
                           BlockDriverAmendStatusCB *status_cb, void *cb_opaque)
M
Max Reitz 已提交
2950
{
2951
    BDRVQcow2State *s = bs->opaque;
M
Max Reitz 已提交
2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963
    int current_version = s->qcow_version;
    int ret;

    if (target_version == current_version) {
        return 0;
    } else if (target_version > current_version) {
        return -EINVAL;
    } else if (target_version != 2) {
        return -EINVAL;
    }

    if (s->refcount_order != 4) {
2964
        error_report("compat=0.10 requires refcount_bits=16");
M
Max Reitz 已提交
2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991
        return -ENOTSUP;
    }

    /* clear incompatible features */
    if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) {
        ret = qcow2_mark_clean(bs);
        if (ret < 0) {
            return ret;
        }
    }

    /* with QCOW2_INCOMPAT_CORRUPT, it is pretty much impossible to get here in
     * the first place; if that happens nonetheless, returning -ENOTSUP is the
     * best thing to do anyway */

    if (s->incompatible_features) {
        return -ENOTSUP;
    }

    /* since we can ignore compatible features, we can set them to 0 as well */
    s->compatible_features = 0;
    /* if lazy refcounts have been used, they have already been fixed through
     * clearing the dirty flag */

    /* clearing autoclear features is trivial */
    s->autoclear_features = 0;

2992
    ret = qcow2_expand_zero_clusters(bs, status_cb, cb_opaque);
M
Max Reitz 已提交
2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005
    if (ret < 0) {
        return ret;
    }

    s->qcow_version = target_version;
    ret = qcow2_update_header(bs);
    if (ret < 0) {
        s->qcow_version = current_version;
        return ret;
    }
    return 0;
}

3006 3007 3008 3009 3010 3011
typedef enum Qcow2AmendOperation {
    /* This is the value Qcow2AmendHelperCBInfo::last_operation will be
     * statically initialized to so that the helper CB can discern the first
     * invocation from an operation change */
    QCOW2_NO_OPERATION = 0,

3012
    QCOW2_CHANGING_REFCOUNT_ORDER,
3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075
    QCOW2_DOWNGRADING,
} Qcow2AmendOperation;

typedef struct Qcow2AmendHelperCBInfo {
    /* The code coordinating the amend operations should only modify
     * these four fields; the rest will be managed by the CB */
    BlockDriverAmendStatusCB *original_status_cb;
    void *original_cb_opaque;

    Qcow2AmendOperation current_operation;

    /* Total number of operations to perform (only set once) */
    int total_operations;

    /* The following fields are managed by the CB */

    /* Number of operations completed */
    int operations_completed;

    /* Cumulative offset of all completed operations */
    int64_t offset_completed;

    Qcow2AmendOperation last_operation;
    int64_t last_work_size;
} Qcow2AmendHelperCBInfo;

static void qcow2_amend_helper_cb(BlockDriverState *bs,
                                  int64_t operation_offset,
                                  int64_t operation_work_size, void *opaque)
{
    Qcow2AmendHelperCBInfo *info = opaque;
    int64_t current_work_size;
    int64_t projected_work_size;

    if (info->current_operation != info->last_operation) {
        if (info->last_operation != QCOW2_NO_OPERATION) {
            info->offset_completed += info->last_work_size;
            info->operations_completed++;
        }

        info->last_operation = info->current_operation;
    }

    assert(info->total_operations > 0);
    assert(info->operations_completed < info->total_operations);

    info->last_work_size = operation_work_size;

    current_work_size = info->offset_completed + operation_work_size;

    /* current_work_size is the total work size for (operations_completed + 1)
     * operations (which includes this one), so multiply it by the number of
     * operations not covered and divide it by the number of operations
     * covered to get a projection for the operations not covered */
    projected_work_size = current_work_size * (info->total_operations -
                                               info->operations_completed - 1)
                                            / (info->operations_completed + 1);

    info->original_status_cb(bs, info->offset_completed + operation_offset,
                             current_work_size + projected_work_size,
                             info->original_cb_opaque);
}

3076
static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
3077 3078
                               BlockDriverAmendStatusCB *status_cb,
                               void *cb_opaque)
M
Max Reitz 已提交
3079
{
3080
    BDRVQcow2State *s = bs->opaque;
M
Max Reitz 已提交
3081 3082 3083 3084
    int old_version = s->qcow_version, new_version = old_version;
    uint64_t new_size = 0;
    const char *backing_file = NULL, *backing_format = NULL;
    bool lazy_refcounts = s->use_lazy_refcounts;
3085 3086 3087
    const char *compat = NULL;
    uint64_t cluster_size = s->cluster_size;
    bool encrypt;
3088
    int refcount_bits = s->refcount_bits;
M
Max Reitz 已提交
3089
    int ret;
3090
    QemuOptDesc *desc = opts->list->desc;
3091
    Qcow2AmendHelperCBInfo helper_cb_info;
M
Max Reitz 已提交
3092

3093 3094
    while (desc && desc->name) {
        if (!qemu_opt_find(opts, desc->name)) {
M
Max Reitz 已提交
3095
            /* only change explicitly defined options */
3096
            desc++;
M
Max Reitz 已提交
3097 3098 3099
            continue;
        }

3100 3101
        if (!strcmp(desc->name, BLOCK_OPT_COMPAT_LEVEL)) {
            compat = qemu_opt_get(opts, BLOCK_OPT_COMPAT_LEVEL);
3102
            if (!compat) {
M
Max Reitz 已提交
3103
                /* preserve default */
3104
            } else if (!strcmp(compat, "0.10")) {
M
Max Reitz 已提交
3105
                new_version = 2;
3106
            } else if (!strcmp(compat, "1.1")) {
M
Max Reitz 已提交
3107 3108
                new_version = 3;
            } else {
3109
                error_report("Unknown compatibility level %s", compat);
M
Max Reitz 已提交
3110 3111
                return -EINVAL;
            }
3112
        } else if (!strcmp(desc->name, BLOCK_OPT_PREALLOC)) {
3113
            error_report("Cannot change preallocation mode");
M
Max Reitz 已提交
3114
            return -ENOTSUP;
3115 3116 3117 3118 3119 3120 3121 3122
        } else if (!strcmp(desc->name, BLOCK_OPT_SIZE)) {
            new_size = qemu_opt_get_size(opts, BLOCK_OPT_SIZE, 0);
        } else if (!strcmp(desc->name, BLOCK_OPT_BACKING_FILE)) {
            backing_file = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
        } else if (!strcmp(desc->name, BLOCK_OPT_BACKING_FMT)) {
            backing_format = qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT);
        } else if (!strcmp(desc->name, BLOCK_OPT_ENCRYPT)) {
            encrypt = qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT,
3123 3124 3125
                                        !!s->cipher);

            if (encrypt != !!s->cipher) {
3126
                error_report("Changing the encryption flag is not supported");
M
Max Reitz 已提交
3127 3128
                return -ENOTSUP;
            }
3129 3130
        } else if (!strcmp(desc->name, BLOCK_OPT_CLUSTER_SIZE)) {
            cluster_size = qemu_opt_get_size(opts, BLOCK_OPT_CLUSTER_SIZE,
3131 3132
                                             cluster_size);
            if (cluster_size != s->cluster_size) {
3133
                error_report("Changing the cluster size is not supported");
M
Max Reitz 已提交
3134 3135
                return -ENOTSUP;
            }
3136 3137
        } else if (!strcmp(desc->name, BLOCK_OPT_LAZY_REFCOUNTS)) {
            lazy_refcounts = qemu_opt_get_bool(opts, BLOCK_OPT_LAZY_REFCOUNTS,
3138
                                               lazy_refcounts);
3139
        } else if (!strcmp(desc->name, BLOCK_OPT_REFCOUNT_BITS)) {
3140 3141 3142 3143 3144 3145 3146 3147 3148 3149
            refcount_bits = qemu_opt_get_number(opts, BLOCK_OPT_REFCOUNT_BITS,
                                                refcount_bits);

            if (refcount_bits <= 0 || refcount_bits > 64 ||
                !is_power_of_2(refcount_bits))
            {
                error_report("Refcount width must be a power of two and may "
                             "not exceed 64 bits");
                return -EINVAL;
            }
M
Max Reitz 已提交
3150
        } else {
3151
            /* if this point is reached, this probably means a new option was
M
Max Reitz 已提交
3152
             * added without having it covered here */
3153
            abort();
M
Max Reitz 已提交
3154
        }
3155 3156

        desc++;
M
Max Reitz 已提交
3157 3158
    }

3159 3160 3161 3162
    helper_cb_info = (Qcow2AmendHelperCBInfo){
        .original_status_cb = status_cb,
        .original_cb_opaque = cb_opaque,
        .total_operations = (new_version < old_version)
3163
                          + (s->refcount_bits != refcount_bits)
3164 3165
    };

3166 3167 3168 3169 3170 3171 3172
    /* Upgrade first (some features may require compat=1.1) */
    if (new_version > old_version) {
        s->qcow_version = new_version;
        ret = qcow2_update_header(bs);
        if (ret < 0) {
            s->qcow_version = old_version;
            return ret;
M
Max Reitz 已提交
3173 3174 3175
        }
    }

3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196
    if (s->refcount_bits != refcount_bits) {
        int refcount_order = ctz32(refcount_bits);
        Error *local_error = NULL;

        if (new_version < 3 && refcount_bits != 16) {
            error_report("Different refcount widths than 16 bits require "
                         "compatibility level 1.1 or above (use compat=1.1 or "
                         "greater)");
            return -EINVAL;
        }

        helper_cb_info.current_operation = QCOW2_CHANGING_REFCOUNT_ORDER;
        ret = qcow2_change_refcount_order(bs, refcount_order,
                                          &qcow2_amend_helper_cb,
                                          &helper_cb_info, &local_error);
        if (ret < 0) {
            error_report_err(local_error);
            return ret;
        }
    }

M
Max Reitz 已提交
3197
    if (backing_file || backing_format) {
3198 3199 3200
        ret = qcow2_change_backing_file(bs,
                    backing_file ?: s->image_backing_file,
                    backing_format ?: s->image_backing_format);
M
Max Reitz 已提交
3201 3202 3203 3204 3205 3206 3207
        if (ret < 0) {
            return ret;
        }
    }

    if (s->use_lazy_refcounts != lazy_refcounts) {
        if (lazy_refcounts) {
3208
            if (new_version < 3) {
3209 3210
                error_report("Lazy refcounts only supported with compatibility "
                             "level 1.1 and above (use compat=1.1 or greater)");
M
Max Reitz 已提交
3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243
                return -EINVAL;
            }
            s->compatible_features |= QCOW2_COMPAT_LAZY_REFCOUNTS;
            ret = qcow2_update_header(bs);
            if (ret < 0) {
                s->compatible_features &= ~QCOW2_COMPAT_LAZY_REFCOUNTS;
                return ret;
            }
            s->use_lazy_refcounts = true;
        } else {
            /* make image clean first */
            ret = qcow2_mark_clean(bs);
            if (ret < 0) {
                return ret;
            }
            /* now disallow lazy refcounts */
            s->compatible_features &= ~QCOW2_COMPAT_LAZY_REFCOUNTS;
            ret = qcow2_update_header(bs);
            if (ret < 0) {
                s->compatible_features |= QCOW2_COMPAT_LAZY_REFCOUNTS;
                return ret;
            }
            s->use_lazy_refcounts = false;
        }
    }

    if (new_size) {
        ret = bdrv_truncate(bs, new_size);
        if (ret < 0) {
            return ret;
        }
    }

3244 3245
    /* Downgrade last (so unsupported features can be removed before) */
    if (new_version < old_version) {
3246 3247 3248
        helper_cb_info.current_operation = QCOW2_DOWNGRADING;
        ret = qcow2_downgrade(bs, new_version, &qcow2_amend_helper_cb,
                              &helper_cb_info);
3249 3250 3251 3252 3253
        if (ret < 0) {
            return ret;
        }
    }

M
Max Reitz 已提交
3254 3255 3256
    return 0;
}

M
Max Reitz 已提交
3257 3258 3259 3260 3261 3262 3263 3264 3265
/*
 * If offset or size are negative, respectively, they will not be included in
 * the BLOCK_IMAGE_CORRUPTED event emitted.
 * fatal will be ignored for read-only BDS; corruptions found there will always
 * be considered non-fatal.
 */
void qcow2_signal_corruption(BlockDriverState *bs, bool fatal, int64_t offset,
                             int64_t size, const char *message_format, ...)
{
3266
    BDRVQcow2State *s = bs->opaque;
3267
    const char *node_name;
M
Max Reitz 已提交
3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290
    char *message;
    va_list ap;

    fatal = fatal && !bs->read_only;

    if (s->signaled_corruption &&
        (!fatal || (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT)))
    {
        return;
    }

    va_start(ap, message_format);
    message = g_strdup_vprintf(message_format, ap);
    va_end(ap);

    if (fatal) {
        fprintf(stderr, "qcow2: Marking image as corrupt: %s; further "
                "corruption events will be suppressed\n", message);
    } else {
        fprintf(stderr, "qcow2: Image is corrupt: %s; further non-fatal "
                "corruption events will be suppressed\n", message);
    }

3291 3292 3293 3294 3295
    node_name = bdrv_get_node_name(bs);
    qapi_event_send_block_image_corrupted(bdrv_get_device_name(bs),
                                          *node_name != '\0', node_name,
                                          message, offset >= 0, offset,
                                          size >= 0, size,
M
Max Reitz 已提交
3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306
                                          fatal, &error_abort);
    g_free(message);

    if (fatal) {
        qcow2_mark_corrupt(bs);
        bs->drv = NULL; /* make BDS unusable */
    }

    s->signaled_corruption = true;
}

3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345
static QemuOptsList qcow2_create_opts = {
    .name = "qcow2-create-opts",
    .head = QTAILQ_HEAD_INITIALIZER(qcow2_create_opts.head),
    .desc = {
        {
            .name = BLOCK_OPT_SIZE,
            .type = QEMU_OPT_SIZE,
            .help = "Virtual disk size"
        },
        {
            .name = BLOCK_OPT_COMPAT_LEVEL,
            .type = QEMU_OPT_STRING,
            .help = "Compatibility level (0.10 or 1.1)"
        },
        {
            .name = BLOCK_OPT_BACKING_FILE,
            .type = QEMU_OPT_STRING,
            .help = "File name of a base image"
        },
        {
            .name = BLOCK_OPT_BACKING_FMT,
            .type = QEMU_OPT_STRING,
            .help = "Image format of the base image"
        },
        {
            .name = BLOCK_OPT_ENCRYPT,
            .type = QEMU_OPT_BOOL,
            .help = "Encrypt the image",
            .def_value_str = "off"
        },
        {
            .name = BLOCK_OPT_CLUSTER_SIZE,
            .type = QEMU_OPT_SIZE,
            .help = "qcow2 cluster size",
            .def_value_str = stringify(DEFAULT_CLUSTER_SIZE)
        },
        {
            .name = BLOCK_OPT_PREALLOC,
            .type = QEMU_OPT_STRING,
3346 3347
            .help = "Preallocation mode (allowed values: off, metadata, "
                    "falloc, full)"
3348 3349 3350 3351 3352 3353 3354
        },
        {
            .name = BLOCK_OPT_LAZY_REFCOUNTS,
            .type = QEMU_OPT_BOOL,
            .help = "Postpone refcount updates",
            .def_value_str = "off"
        },
3355 3356 3357 3358 3359 3360
        {
            .name = BLOCK_OPT_REFCOUNT_BITS,
            .type = QEMU_OPT_NUMBER,
            .help = "Width of a reference count entry in bits",
            .def_value_str = "16"
        },
3361 3362
        { /* end of list */ }
    }
B
Blue Swirl 已提交
3363 3364
};

3365
BlockDriver bdrv_qcow2 = {
3366
    .format_name        = "qcow2",
3367
    .instance_size      = sizeof(BDRVQcow2State),
3368 3369 3370
    .bdrv_probe         = qcow2_probe,
    .bdrv_open          = qcow2_open,
    .bdrv_close         = qcow2_close,
J
Jeff Cody 已提交
3371
    .bdrv_reopen_prepare  = qcow2_reopen_prepare,
3372 3373
    .bdrv_reopen_commit   = qcow2_reopen_commit,
    .bdrv_reopen_abort    = qcow2_reopen_abort,
3374
    .bdrv_join_options    = qcow2_join_options,
C
Chunyan Liu 已提交
3375
    .bdrv_create        = qcow2_create,
3376
    .bdrv_has_zero_init = bdrv_has_zero_init_1,
3377
    .bdrv_co_get_block_status = qcow2_co_get_block_status,
3378 3379
    .bdrv_set_key       = qcow2_set_key,

K
Kevin Wolf 已提交
3380
    .bdrv_co_preadv         = qcow2_co_preadv,
K
Kevin Wolf 已提交
3381
    .bdrv_co_pwritev        = qcow2_co_pwritev,
K
Kevin Wolf 已提交
3382
    .bdrv_co_flush_to_os    = qcow2_co_flush_to_os,
3383

3384
    .bdrv_co_pwrite_zeroes  = qcow2_co_pwrite_zeroes,
3385
    .bdrv_co_pdiscard       = qcow2_co_pdiscard,
3386
    .bdrv_truncate          = qcow2_truncate,
3387
    .bdrv_co_pwritev_compressed = qcow2_co_pwritev_compressed,
M
Max Reitz 已提交
3388
    .bdrv_make_empty        = qcow2_make_empty,
B
Blue Swirl 已提交
3389 3390 3391 3392 3393

    .bdrv_snapshot_create   = qcow2_snapshot_create,
    .bdrv_snapshot_goto     = qcow2_snapshot_goto,
    .bdrv_snapshot_delete   = qcow2_snapshot_delete,
    .bdrv_snapshot_list     = qcow2_snapshot_list,
3394 3395
    .bdrv_snapshot_load_tmp = qcow2_snapshot_load_tmp,
    .bdrv_get_info          = qcow2_get_info,
3396
    .bdrv_get_specific_info = qcow2_get_specific_info,
B
Blue Swirl 已提交
3397

3398 3399
    .bdrv_save_vmstate    = qcow2_save_vmstate,
    .bdrv_load_vmstate    = qcow2_load_vmstate,
B
Blue Swirl 已提交
3400

3401
    .supports_backing           = true,
B
Blue Swirl 已提交
3402 3403
    .bdrv_change_backing_file   = qcow2_change_backing_file,

3404
    .bdrv_refresh_limits        = qcow2_refresh_limits,
3405
    .bdrv_invalidate_cache      = qcow2_invalidate_cache,
K
Kevin Wolf 已提交
3406
    .bdrv_inactivate            = qcow2_inactivate,
3407

3408 3409
    .create_opts         = &qcow2_create_opts,
    .bdrv_check          = qcow2_check,
C
Chunyan Liu 已提交
3410
    .bdrv_amend_options  = qcow2_amend_options,
3411 3412 3413

    .bdrv_detach_aio_context  = qcow2_detach_aio_context,
    .bdrv_attach_aio_context  = qcow2_attach_aio_context,
B
Blue Swirl 已提交
3414 3415
};

3416 3417 3418 3419 3420 3421
static void bdrv_qcow2_init(void)
{
    bdrv_register(&bdrv_qcow2);
}

block_init(bdrv_qcow2_init);