/*
 * Block driver for the QCOW version 2 format
 *
 * Copyright (c) 2004-2006 Fabrice Bellard
 *
 * 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.
 */

#include "qemu/osdep.h"

#include "block/qdict.h"
#include "sysemu/block-backend.h"
#include "qemu/main-loop.h"
#include "qemu/module.h"
#include "qcow2.h"
#include "qemu/error-report.h"
#include "qapi/error.h"
#include "qapi/qapi-events-block-core.h"
#include "qapi/qmp/qdict.h"
#include "qapi/qmp/qstring.h"
#include "trace.h"
#include "qemu/option_int.h"
#include "qemu/cutils.h"
#include "qemu/bswap.h"
#include "qapi/qobject-input-visitor.h"
#include "qapi/qapi-visit-block-core.h"
#include "crypto.h"
#include "block/aio_task.h"

/*
  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.
  - Size of compressed clusters is stored in sectors to reduce bit usage
    in the cluster offsets.
  - Support for storing additional data (such as the VM state) in the
    snapshots.
  - 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.
*/


typedef struct {
    uint32_t magic;
    uint32_t len;
} QEMU_PACKED QCowExtension;

#define  QCOW2_EXT_MAGIC_END 0
#define  QCOW2_EXT_MAGIC_BACKING_FORMAT 0xe2792aca
#define  QCOW2_EXT_MAGIC_FEATURE_TABLE 0x6803f857
#define  QCOW2_EXT_MAGIC_CRYPTO_HEADER 0x0537be77
#define  QCOW2_EXT_MAGIC_BITMAPS 0x23852875
#define  QCOW2_EXT_MAGIC_DATA_FILE 0x44415441

static int coroutine_fn
qcow2_co_preadv_compressed(BlockDriverState *bs,
                           uint64_t cluster_descriptor,
                           uint64_t offset,
                           uint64_t bytes,
                           QEMUIOVector *qiov,
                           size_t qiov_offset);

static int qcow2_probe(const uint8_t *buf, int buf_size, const char *filename)
{
    const QCowHeader *cow_header = (const void *)buf;

    if (buf_size >= sizeof(QCowHeader) &&
        be32_to_cpu(cow_header->magic) == QCOW_MAGIC &&
        be32_to_cpu(cow_header->version) >= 2)
        return 100;
    else
        return 0;
}


static ssize_t qcow2_crypto_hdr_read_func(QCryptoBlock *block, size_t offset,
                                          uint8_t *buf, size_t buflen,
                                          void *opaque, Error **errp)
{
    BlockDriverState *bs = opaque;
    BDRVQcow2State *s = bs->opaque;
    ssize_t ret;

    if ((offset + buflen) > s->crypto_header.length) {
        error_setg(errp, "Request for data outside of extension header");
        return -1;
    }

    ret = bdrv_pread(bs->file,
                     s->crypto_header.offset + offset, buf, buflen);
    if (ret < 0) {
        error_setg_errno(errp, -ret, "Could not read encryption header");
        return -1;
    }
    return ret;
}


static ssize_t qcow2_crypto_hdr_init_func(QCryptoBlock *block, size_t headerlen,
                                          void *opaque, Error **errp)
{
    BlockDriverState *bs = opaque;
    BDRVQcow2State *s = bs->opaque;
    int64_t ret;
    int64_t clusterlen;

    ret = qcow2_alloc_clusters(bs, headerlen);
    if (ret < 0) {
        error_setg_errno(errp, -ret,
                         "Cannot allocate cluster for LUKS header size %zu",
                         headerlen);
        return -1;
    }

    s->crypto_header.length = headerlen;
    s->crypto_header.offset = ret;

    /*
     * Zero fill all space in cluster so it has predictable
     * content, as we may not initialize some regions of the
     * header (eg only 1 out of 8 key slots will be initialized)
     */
    clusterlen = size_to_clusters(s, headerlen) * s->cluster_size;
    assert(qcow2_pre_write_overlap_check(bs, 0, ret, clusterlen, false) == 0);
    ret = bdrv_pwrite_zeroes(bs->file,
                             ret,
                             clusterlen, 0);
    if (ret < 0) {
        error_setg_errno(errp, -ret, "Could not zero fill encryption header");
        return -1;
    }

    return ret;
}


static ssize_t qcow2_crypto_hdr_write_func(QCryptoBlock *block, size_t offset,
                                           const uint8_t *buf, size_t buflen,
                                           void *opaque, Error **errp)
{
    BlockDriverState *bs = opaque;
    BDRVQcow2State *s = bs->opaque;
    ssize_t ret;

    if ((offset + buflen) > s->crypto_header.length) {
        error_setg(errp, "Request for data outside of extension header");
        return -1;
    }

    ret = bdrv_pwrite(bs->file,
                      s->crypto_header.offset + offset, buf, buflen);
    if (ret < 0) {
        error_setg_errno(errp, -ret, "Could not read encryption header");
        return -1;
    }
    return ret;
}

static QDict*
qcow2_extract_crypto_opts(QemuOpts *opts, const char *fmt, Error **errp)
{
    QDict *cryptoopts_qdict;
    QDict *opts_qdict;

    /* Extract "encrypt." options into a qdict */
    opts_qdict = qemu_opts_to_qdict(opts, NULL);
    qdict_extract_subqdict(opts_qdict, &cryptoopts_qdict, "encrypt.");
    qobject_unref(opts_qdict);
    qdict_put_str(cryptoopts_qdict, "format", fmt);
    return cryptoopts_qdict;
}

/*
 * 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
 */
static int qcow2_read_extensions(BlockDriverState *bs, uint64_t start_offset,
                                 uint64_t end_offset, void **p_feature_table,
                                 int flags, bool *need_update_header,
                                 Error **errp)
{
    BDRVQcow2State *s = bs->opaque;
    QCowExtension ext;
    uint64_t offset;
    int ret;
    Qcow2BitmapHeaderExt bitmaps_ext;

    if (need_update_header != NULL) {
        *need_update_header = false;
    }

#ifdef DEBUG_EXT
    printf("qcow2_read_extensions: start=%ld end=%ld\n", start_offset, end_offset);
#endif
    offset = start_offset;
    while (offset < end_offset) {

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

        printf("attempting to read extended header in offset %lu\n", offset);
#endif

        ret = bdrv_pread(bs->file, offset, &ext, sizeof(ext));
        if (ret < 0) {
            error_setg_errno(errp, -ret, "qcow2_read_extension: ERROR: "
                             "pread fail from offset %" PRIu64, offset);
            return 1;
        }
        ext.magic = be32_to_cpu(ext.magic);
        ext.len = be32_to_cpu(ext.len);
        offset += sizeof(ext);
#ifdef DEBUG_EXT
        printf("ext.magic = 0x%x\n", ext.magic);
#endif
        if (offset > end_offset || ext.len > end_offset - offset) {
            error_setg(errp, "Header extension too large");
            return -EINVAL;
        }

        switch (ext.magic) {
        case QCOW2_EXT_MAGIC_END:
            return 0;

        case QCOW2_EXT_MAGIC_BACKING_FORMAT:
            if (ext.len >= sizeof(bs->backing_format)) {
                error_setg(errp, "ERROR: ext_backing_format: len=%" PRIu32
                           " too large (>=%zu)", ext.len,
                           sizeof(bs->backing_format));
                return 2;
            }
            ret = bdrv_pread(bs->file, offset, bs->backing_format, ext.len);
            if (ret < 0) {
                error_setg_errno(errp, -ret, "ERROR: ext_backing_format: "
                                 "Could not read format name");
                return 3;
            }
            bs->backing_format[ext.len] = '\0';
            s->image_backing_format = g_strdup(bs->backing_format);
#ifdef DEBUG_EXT
            printf("Qcow2: Got format extension %s\n", bs->backing_format);
#endif
            break;

        case QCOW2_EXT_MAGIC_FEATURE_TABLE:
            if (p_feature_table != NULL) {
                void *feature_table = g_malloc0(ext.len + 2 * sizeof(Qcow2Feature));
                ret = bdrv_pread(bs->file, offset , feature_table, ext.len);
                if (ret < 0) {
                    error_setg_errno(errp, -ret, "ERROR: ext_feature_table: "
                                     "Could not read table");
                    return ret;
                }

                *p_feature_table = feature_table;
            }
            break;

        case QCOW2_EXT_MAGIC_CRYPTO_HEADER: {
            unsigned int cflags = 0;
            if (s->crypt_method_header != QCOW_CRYPT_LUKS) {
                error_setg(errp, "CRYPTO header extension only "
                           "expected with LUKS encryption method");
                return -EINVAL;
            }
            if (ext.len != sizeof(Qcow2CryptoHeaderExtension)) {
                error_setg(errp, "CRYPTO header extension size %u, "
                           "but expected size %zu", ext.len,
                           sizeof(Qcow2CryptoHeaderExtension));
                return -EINVAL;
            }

            ret = bdrv_pread(bs->file, offset, &s->crypto_header, ext.len);
            if (ret < 0) {
                error_setg_errno(errp, -ret,
                                 "Unable to read CRYPTO header extension");
                return ret;
            }
            s->crypto_header.offset = be64_to_cpu(s->crypto_header.offset);
            s->crypto_header.length = be64_to_cpu(s->crypto_header.length);

            if ((s->crypto_header.offset % s->cluster_size) != 0) {
                error_setg(errp, "Encryption header offset '%" PRIu64 "' is "
                           "not a multiple of cluster size '%u'",
                           s->crypto_header.offset, s->cluster_size);
                return -EINVAL;
            }

            if (flags & BDRV_O_NO_IO) {
                cflags |= QCRYPTO_BLOCK_OPEN_NO_IO;
            }
            s->crypto = qcrypto_block_open(s->crypto_opts, "encrypt.",
                                           qcow2_crypto_hdr_read_func,
                                           bs, cflags, QCOW2_MAX_THREADS, errp);
            if (!s->crypto) {
                return -EINVAL;
            }
        }   break;

        case QCOW2_EXT_MAGIC_BITMAPS:
            if (ext.len != sizeof(bitmaps_ext)) {
                error_setg_errno(errp, -ret, "bitmaps_ext: "
                                 "Invalid extension length");
                return -EINVAL;
            }

            if (!(s->autoclear_features & QCOW2_AUTOCLEAR_BITMAPS)) {
                if (s->qcow_version < 3) {
                    /* Let's be a bit more specific */
                    warn_report("This qcow2 v2 image contains bitmaps, but "
                                "they may have been modified by a program "
                                "without persistent bitmap support; so now "
                                "they must all be considered inconsistent");
                } else {
                    warn_report("a program lacking bitmap support "
                                "modified this file, so all bitmaps are now "
                                "considered inconsistent");
                }
                error_printf("Some clusters may be leaked, "
                             "run 'qemu-img check -r' on the image "
                             "file to fix.");
                if (need_update_header != NULL) {
                    /* Updating is needed to drop invalid bitmap extension. */
                    *need_update_header = true;
                }
                break;
            }

            ret = bdrv_pread(bs->file, offset, &bitmaps_ext, ext.len);
            if (ret < 0) {
                error_setg_errno(errp, -ret, "bitmaps_ext: "
                                 "Could not read ext header");
                return ret;
            }

            if (bitmaps_ext.reserved32 != 0) {
                error_setg_errno(errp, -ret, "bitmaps_ext: "
                                 "Reserved field is not zero");
                return -EINVAL;
            }

            bitmaps_ext.nb_bitmaps = be32_to_cpu(bitmaps_ext.nb_bitmaps);
            bitmaps_ext.bitmap_directory_size =
                be64_to_cpu(bitmaps_ext.bitmap_directory_size);
            bitmaps_ext.bitmap_directory_offset =
                be64_to_cpu(bitmaps_ext.bitmap_directory_offset);

            if (bitmaps_ext.nb_bitmaps > QCOW2_MAX_BITMAPS) {
                error_setg(errp,
                           "bitmaps_ext: Image has %" PRIu32 " bitmaps, "
                           "exceeding the QEMU supported maximum of %d",
                           bitmaps_ext.nb_bitmaps, QCOW2_MAX_BITMAPS);
                return -EINVAL;
            }

            if (bitmaps_ext.nb_bitmaps == 0) {
                error_setg(errp, "found bitmaps extension with zero bitmaps");
                return -EINVAL;
            }

            if (offset_into_cluster(s, bitmaps_ext.bitmap_directory_offset)) {
                error_setg(errp, "bitmaps_ext: "
                                 "invalid bitmap directory offset");
                return -EINVAL;
            }

            if (bitmaps_ext.bitmap_directory_size >
                QCOW2_MAX_BITMAP_DIRECTORY_SIZE) {
                error_setg(errp, "bitmaps_ext: "
                                 "bitmap directory size (%" PRIu64 ") exceeds "
                                 "the maximum supported size (%d)",
                                 bitmaps_ext.bitmap_directory_size,
                                 QCOW2_MAX_BITMAP_DIRECTORY_SIZE);
                return -EINVAL;
            }

            s->nb_bitmaps = bitmaps_ext.nb_bitmaps;
            s->bitmap_directory_offset =
                    bitmaps_ext.bitmap_directory_offset;
            s->bitmap_directory_size =
                    bitmaps_ext.bitmap_directory_size;

#ifdef DEBUG_EXT
            printf("Qcow2: Got bitmaps extension: "
                   "offset=%" PRIu64 " nb_bitmaps=%" PRIu32 "\n",
                   s->bitmap_directory_offset, s->nb_bitmaps);
#endif
            break;

        case QCOW2_EXT_MAGIC_DATA_FILE:
        {
            s->image_data_file = g_malloc0(ext.len + 1);
            ret = bdrv_pread(bs->file, offset, s->image_data_file, ext.len);
            if (ret < 0) {
                error_setg_errno(errp, -ret,
                                 "ERROR: Could not read data file name");
                return ret;
            }
#ifdef DEBUG_EXT
            printf("Qcow2: Got external data file %s\n", s->image_data_file);
#endif
            break;
        }

        default:
            /* unknown magic - save it in case we need to rewrite the header */
            /* If you add a new feature, make sure to also update the fast
             * path of qcow2_make_empty() to deal with it. */
            {
                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);

                ret = bdrv_pread(bs->file, offset , uext->data, uext->len);
                if (ret < 0) {
                    error_setg_errno(errp, -ret, "ERROR: unknown extension: "
                                     "Could not read data");
                    return ret;
                }
            }
            break;
        }

        offset += ((ext.len + 7) & ~7);
    }

    return 0;
}

static void cleanup_unknown_header_ext(BlockDriverState *bs)
{
    BDRVQcow2State *s = bs->opaque;
    Qcow2UnknownHeaderExtension *uext, *next;

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

static void report_unsupported_feature(Error **errp, Qcow2Feature *table,
                                       uint64_t mask)
{
    g_autoptr(GString) features = g_string_sized_new(60);

    while (table && table->name[0] != '\0') {
        if (table->type == QCOW2_FEAT_TYPE_INCOMPATIBLE) {
            if (mask & (1ULL << table->bit)) {
                if (features->len > 0) {
                    g_string_append(features, ", ");
                }
                g_string_append_printf(features, "%.46s", table->name);
                mask &= ~(1ULL << table->bit);
            }
        }
        table++;
    }

    if (mask) {
        if (features->len > 0) {
            g_string_append(features, ", ");
        }
        g_string_append_printf(features,
                               "Unknown incompatible feature: %" PRIx64, mask);
    }

    error_setg(errp, "Unsupported qcow2 feature(s): %s", features->str);
}

/*
 * 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.
 */
int qcow2_mark_dirty(BlockDriverState *bs)
{
    BDRVQcow2State *s = bs->opaque;
    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);
    ret = bdrv_pwrite(bs->file, offsetof(QCowHeader, incompatible_features),
                      &val, sizeof(val));
    if (ret < 0) {
        return ret;
    }
    ret = bdrv_flush(bs->file->bs);
    if (ret < 0) {
        return ret;
    }

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

/*
 * 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)
{
    BDRVQcow2State *s = bs->opaque;

    if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) {
        int ret;

        s->incompatible_features &= ~QCOW2_INCOMPAT_DIRTY;

        ret = qcow2_flush_caches(bs);
        if (ret < 0) {
            return ret;
        }

        return qcow2_update_header(bs);
    }
    return 0;
}

/*
 * Marks the image as corrupt.
 */
int qcow2_mark_corrupt(BlockDriverState *bs)
{
    BDRVQcow2State *s = bs->opaque;

    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)
{
    BDRVQcow2State *s = bs->opaque;

    if (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT) {
        int ret = qcow2_flush_caches(bs);
        if (ret < 0) {
            return ret;
        }

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

static void qcow2_add_check_result(BdrvCheckResult *out,
                                   const BdrvCheckResult *src,
                                   bool set_allocation_info)
{
    out->corruptions += src->corruptions;
    out->leaks += src->leaks;
    out->check_errors += src->check_errors;
    out->corruptions_fixed += src->corruptions_fixed;
    out->leaks_fixed += src->leaks_fixed;

    if (set_allocation_info) {
        out->image_end_offset = src->image_end_offset;
        out->bfi = src->bfi;
    }
}

static int coroutine_fn qcow2_co_check_locked(BlockDriverState *bs,
                                              BdrvCheckResult *result,
                                              BdrvCheckMode fix)
{
    BdrvCheckResult snapshot_res = {};
    BdrvCheckResult refcount_res = {};
    int ret;

    memset(result, 0, sizeof(*result));

    ret = qcow2_check_read_snapshot_table(bs, &snapshot_res, fix);
    if (ret < 0) {
        qcow2_add_check_result(result, &snapshot_res, false);
        return ret;
    }

    ret = qcow2_check_refcounts(bs, &refcount_res, fix);
    qcow2_add_check_result(result, &refcount_res, true);
    if (ret < 0) {
        qcow2_add_check_result(result, &snapshot_res, false);
        return ret;
    }

    ret = qcow2_check_fix_snapshot_table(bs, &snapshot_res, fix);
    qcow2_add_check_result(result, &snapshot_res, false);
    if (ret < 0) {
        return ret;
    }

    if (fix && result->check_errors == 0 && result->corruptions == 0) {
        ret = qcow2_mark_clean(bs);
        if (ret < 0) {
            return ret;
        }
        return qcow2_mark_consistent(bs);
    }
    return ret;
}

static int coroutine_fn qcow2_co_check(BlockDriverState *bs,
                                       BdrvCheckResult *result,
                                       BdrvCheckMode fix)
{
    BDRVQcow2State *s = bs->opaque;
    int ret;

    qemu_co_mutex_lock(&s->lock);
    ret = qcow2_co_check_locked(bs, result, fix);
    qemu_co_mutex_unlock(&s->lock);
    return ret;
}

int qcow2_validate_table(BlockDriverState *bs, uint64_t offset,
                         uint64_t entries, size_t entry_len,
                         int64_t max_size_bytes, const char *table_name,
                         Error **errp)
{
    BDRVQcow2State *s = bs->opaque;

    if (entries > max_size_bytes / entry_len) {
        error_setg(errp, "%s too large", table_name);
        return -EFBIG;
    }

    /* 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 ((INT64_MAX - entries * entry_len < offset) ||
        (offset_into_cluster(s, offset) != 0)) {
        error_setg(errp, "%s offset invalid", table_name);
        return -EINVAL;
    }

    return 0;
}

static const char *const mutable_opts[] = {
    QCOW2_OPT_LAZY_REFCOUNTS,
    QCOW2_OPT_DISCARD_REQUEST,
    QCOW2_OPT_DISCARD_SNAPSHOT,
    QCOW2_OPT_DISCARD_OTHER,
    QCOW2_OPT_OVERLAP,
    QCOW2_OPT_OVERLAP_TEMPLATE,
    QCOW2_OPT_OVERLAP_MAIN_HEADER,
    QCOW2_OPT_OVERLAP_ACTIVE_L1,
    QCOW2_OPT_OVERLAP_ACTIVE_L2,
    QCOW2_OPT_OVERLAP_REFCOUNT_TABLE,
    QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK,
    QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE,
    QCOW2_OPT_OVERLAP_INACTIVE_L1,
    QCOW2_OPT_OVERLAP_INACTIVE_L2,
    QCOW2_OPT_OVERLAP_BITMAP_DIRECTORY,
    QCOW2_OPT_CACHE_SIZE,
    QCOW2_OPT_L2_CACHE_SIZE,
    QCOW2_OPT_L2_CACHE_ENTRY_SIZE,
    QCOW2_OPT_REFCOUNT_CACHE_SIZE,
    QCOW2_OPT_CACHE_CLEAN_INTERVAL,
    NULL
};

static QemuOptsList qcow2_runtime_opts = {
    .name = "qcow2",
    .head = QTAILQ_HEAD_INITIALIZER(qcow2_runtime_opts.head),
    .desc = {
        {
            .name = QCOW2_OPT_LAZY_REFCOUNTS,
            .type = QEMU_OPT_BOOL,
            .help = "Postpone refcount updates",
        },
        {
            .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",
        },
        {
            .name = QCOW2_OPT_OVERLAP,
            .type = QEMU_OPT_STRING,
            .help = "Selects which overlap checks to perform from a range of "
                    "templates (none, constant, cached, all)",
        },
        {
            .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)",
        },
        {
            .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",
        },
        {
            .name = QCOW2_OPT_OVERLAP_BITMAP_DIRECTORY,
            .type = QEMU_OPT_BOOL,
            .help = "Check for unintended writes into the bitmap directory",
        },
        {
            .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_L2_CACHE_ENTRY_SIZE,
            .type = QEMU_OPT_SIZE,
            .help = "Size of each entry in the L2 cache",
        },
        {
            .name = QCOW2_OPT_REFCOUNT_CACHE_SIZE,
            .type = QEMU_OPT_SIZE,
            .help = "Maximum refcount block cache size",
        },
        {
            .name = QCOW2_OPT_CACHE_CLEAN_INTERVAL,
            .type = QEMU_OPT_NUMBER,
            .help = "Clean unused cache entries after this time (in seconds)",
        },
        BLOCK_CRYPTO_OPT_DEF_KEY_SECRET("encrypt.",
            "ID of secret providing qcow2 AES key or LUKS passphrase"),
        { /* end of list */ }
    },
};

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,
    [QCOW2_OL_BITMAP_DIRECTORY_BITNR] = QCOW2_OPT_OVERLAP_BITMAP_DIRECTORY,
};

static void cache_clean_timer_cb(void *opaque)
{
    BlockDriverState *bs = opaque;
    BDRVQcow2State *s = bs->opaque;
    qcow2_cache_clean_unused(s->l2_table_cache);
    qcow2_cache_clean_unused(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)
{
    BDRVQcow2State *s = bs->opaque;
    if (s->cache_clean_interval > 0) {
        s->cache_clean_timer =
            aio_timer_new_with_attrs(context, QEMU_CLOCK_VIRTUAL,
                                     SCALE_MS, QEMU_TIMER_ATTR_EXTERNAL,
                                     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)
{
    BDRVQcow2State *s = bs->opaque;
    if (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);
}

static bool read_cache_sizes(BlockDriverState *bs, QemuOpts *opts,
                             uint64_t *l2_cache_size,
                             uint64_t *l2_cache_entry_size,
                             uint64_t *refcount_cache_size, Error **errp)
{
    BDRVQcow2State *s = bs->opaque;
    uint64_t combined_cache_size, l2_cache_max_setting;
    bool l2_cache_size_set, refcount_cache_size_set, combined_cache_size_set;
    bool l2_cache_entry_size_set;
    int min_refcount_cache = MIN_REFCOUNT_CACHE_SIZE * s->cluster_size;
    uint64_t virtual_disk_size = bs->total_sectors * BDRV_SECTOR_SIZE;
    uint64_t max_l2_entries = DIV_ROUND_UP(virtual_disk_size, s->cluster_size);
    /* An L2 table is always one cluster in size so the max cache size
     * should be a multiple of the cluster size. */
    uint64_t max_l2_cache = ROUND_UP(max_l2_entries * l2_entry_size(s),
                                     s->cluster_size);

    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);
    l2_cache_entry_size_set = qemu_opt_get(opts, QCOW2_OPT_L2_CACHE_ENTRY_SIZE);

    combined_cache_size = qemu_opt_get_size(opts, QCOW2_OPT_CACHE_SIZE, 0);
    l2_cache_max_setting = qemu_opt_get_size(opts, QCOW2_OPT_L2_CACHE_SIZE,
                                             DEFAULT_L2_CACHE_MAX_SIZE);
    *refcount_cache_size = qemu_opt_get_size(opts,
                                             QCOW2_OPT_REFCOUNT_CACHE_SIZE, 0);

    *l2_cache_entry_size = qemu_opt_get_size(
        opts, QCOW2_OPT_L2_CACHE_ENTRY_SIZE, s->cluster_size);

    *l2_cache_size = MIN(max_l2_cache, l2_cache_max_setting);

    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 "
                       "at the same time");
            return false;
        } else if (l2_cache_size_set &&
                   (l2_cache_max_setting > combined_cache_size)) {
            error_setg(errp, QCOW2_OPT_L2_CACHE_SIZE " may not exceed "
                       QCOW2_OPT_CACHE_SIZE);
            return false;
        } else if (*refcount_cache_size > combined_cache_size) {
            error_setg(errp, QCOW2_OPT_REFCOUNT_CACHE_SIZE " may not exceed "
                       QCOW2_OPT_CACHE_SIZE);
            return false;
        }

        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 {
            /* Assign as much memory as possible to the L2 cache, and
             * use the remainder for the refcount cache */
            if (combined_cache_size >= max_l2_cache + min_refcount_cache) {
                *l2_cache_size = max_l2_cache;
                *refcount_cache_size = combined_cache_size - *l2_cache_size;
            } else {
                *refcount_cache_size =
                    MIN(combined_cache_size, min_refcount_cache);
                *l2_cache_size = combined_cache_size - *refcount_cache_size;
            }
        }
    }

    /*
     * If the L2 cache is not enough to cover the whole disk then
     * default to 4KB entries. Smaller entries reduce the cost of
     * loads and evictions and increase I/O performance.
     */
    if (*l2_cache_size < max_l2_cache && !l2_cache_entry_size_set) {
        *l2_cache_entry_size = MIN(s->cluster_size, 4096);
    }

    /* l2_cache_size and refcount_cache_size are ensured to have at least
     * their minimum values in qcow2_update_options_prepare() */

    if (*l2_cache_entry_size < (1 << MIN_CLUSTER_BITS) ||
        *l2_cache_entry_size > s->cluster_size ||
        !is_power_of_2(*l2_cache_entry_size)) {
        error_setg(errp, "L2 cache entry size must be a power of two "
                   "between %d and the cluster size (%d)",
                   1 << MIN_CLUSTER_BITS, s->cluster_size);
        return false;
    }

    return true;
}

typedef struct Qcow2ReopenState {
    Qcow2Cache *l2_table_cache;
    Qcow2Cache *refcount_block_cache;
    int l2_slice_size; /* Number of entries in a slice of the L2 table */
    bool use_lazy_refcounts;
    int overlap_check;
    bool discard_passthrough[QCOW2_DISCARD_MAX];
    uint64_t cache_clean_interval;
    QCryptoBlockOpenOptions *crypto_opts; /* Disk encryption runtime options */
} Qcow2ReopenState;

static int qcow2_update_options_prepare(BlockDriverState *bs,
                                        Qcow2ReopenState *r,
                                        QDict *options, int flags,
                                        Error **errp)
{
    BDRVQcow2State *s = bs->opaque;
    QemuOpts *opts = NULL;
    const char *opt_overlap_check, *opt_overlap_check_template;
    int overlap_check_template = 0;
    uint64_t l2_cache_size, l2_cache_entry_size, refcount_cache_size;
    int i;
    const char *encryptfmt;
    QDict *encryptopts = NULL;
    int ret;

    qdict_extract_subqdict(options, &encryptopts, "encrypt.");
    encryptfmt = qdict_get_try_str(encryptopts, "format");

    opts = qemu_opts_create(&qcow2_runtime_opts, NULL, 0, &error_abort);
    if (!qemu_opts_absorb_qdict(opts, options, errp)) {
        ret = -EINVAL;
        goto fail;
    }

    /* get L2 table/refcount block cache size from command line options */
    if (!read_cache_sizes(bs, opts, &l2_cache_size, &l2_cache_entry_size,

                          &refcount_cache_size, errp)) {
        ret = -EINVAL;
        goto fail;
    }

    l2_cache_size /= l2_cache_entry_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;
    }

    /* 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;
        }
    }

    r->l2_slice_size = l2_cache_entry_size / l2_entry_size(s);
    r->l2_table_cache = qcow2_cache_create(bs, l2_cache_size,
                                           l2_cache_entry_size);
    r->refcount_block_cache = qcow2_cache_create(bs, refcount_cache_size,
                                                 s->cluster_size);
    if (r->l2_table_cache == NULL || r->refcount_block_cache == NULL) {
        error_setg(errp, "Could not allocate metadata caches");
        ret = -ENOMEM;
        goto fail;
    }

    /* New interval for cache cleanup timer */
    r->cache_clean_interval =
        qemu_opt_get_number(opts, QCOW2_OPT_CACHE_CLEAN_INTERVAL,
                            DEFAULT_CACHE_CLEAN_INTERVAL);
#ifndef CONFIG_LINUX
    if (r->cache_clean_interval != 0) {
        error_setg(errp, QCOW2_OPT_CACHE_CLEAN_INTERVAL
                   " not supported on this host");
        ret = -EINVAL;
        goto fail;
    }
#endif
    if (r->cache_clean_interval > UINT_MAX) {
        error_setg(errp, "Cache clean interval too big");
        ret = -EINVAL;
        goto fail;
    }

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

    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;
        }
    }

    /* Overlap check options */
    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;
    }

    r->overlap_check = 0;
    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 */
        r->overlap_check |=
            qemu_opt_get_bool(opts, overlap_bool_option_names[i],
                              overlap_check_template & (1 << i)) << i;
    }

    r->discard_passthrough[QCOW2_DISCARD_NEVER] = false;
    r->discard_passthrough[QCOW2_DISCARD_ALWAYS] = true;
    r->discard_passthrough[QCOW2_DISCARD_REQUEST] =
        qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_REQUEST,
                          flags & BDRV_O_UNMAP);
    r->discard_passthrough[QCOW2_DISCARD_SNAPSHOT] =
        qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_SNAPSHOT, true);
    r->discard_passthrough[QCOW2_DISCARD_OTHER] =
        qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_OTHER, false);

    switch (s->crypt_method_header) {
    case QCOW_CRYPT_NONE:
        if (encryptfmt) {
            error_setg(errp, "No encryption in image header, but options "
                       "specified format '%s'", encryptfmt);
            ret = -EINVAL;
            goto fail;
        }
        break;

    case QCOW_CRYPT_AES:
        if (encryptfmt && !g_str_equal(encryptfmt, "aes")) {
            error_setg(errp,
                       "Header reported 'aes' encryption format but "
                       "options specify '%s'", encryptfmt);
            ret = -EINVAL;
            goto fail;
        }
        qdict_put_str(encryptopts, "format", "qcow");
        r->crypto_opts = block_crypto_open_opts_init(encryptopts, errp);
        if (!r->crypto_opts) {
            ret = -EINVAL;
            goto fail;
        }
        break;

    case QCOW_CRYPT_LUKS:
        if (encryptfmt && !g_str_equal(encryptfmt, "luks")) {
            error_setg(errp,
                       "Header reported 'luks' encryption format but "
                       "options specify '%s'", encryptfmt);
            ret = -EINVAL;
            goto fail;
        }
        qdict_put_str(encryptopts, "format", "luks");
        r->crypto_opts = block_crypto_open_opts_init(encryptopts, errp);
        if (!r->crypto_opts) {
            ret = -EINVAL;
            goto fail;
        }
        break;

    default:
        error_setg(errp, "Unsupported encryption method %d",
                   s->crypt_method_header);
        ret = -EINVAL;
        goto fail;
    }

    ret = 0;
fail:
    qobject_unref(encryptopts);
    qemu_opts_del(opts);
    opts = NULL;
    return ret;
}

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

    if (s->l2_table_cache) {
        qcow2_cache_destroy(s->l2_table_cache);
    }
    if (s->refcount_block_cache) {
        qcow2_cache_destroy(s->refcount_block_cache);
    }
    s->l2_table_cache = r->l2_table_cache;
    s->refcount_block_cache = r->refcount_block_cache;
    s->l2_slice_size = r->l2_slice_size;

    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];
    }

    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));
    }

    qapi_free_QCryptoBlockOpenOptions(s->crypto_opts);
    s->crypto_opts = r->crypto_opts;
}

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

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);
    }

    return ret;
}

static int validate_compression_type(BDRVQcow2State *s, Error **errp)
{
    switch (s->compression_type) {
    case QCOW2_COMPRESSION_TYPE_ZLIB:
#ifdef CONFIG_ZSTD
    case QCOW2_COMPRESSION_TYPE_ZSTD:
#endif
        break;

    default:
        error_setg(errp, "qcow2: unknown compression type: %u",
                   s->compression_type);
        return -ENOTSUP;
    }

    /*
     * if the compression type differs from QCOW2_COMPRESSION_TYPE_ZLIB
     * the incompatible feature flag must be set
     */
    if (s->compression_type == QCOW2_COMPRESSION_TYPE_ZLIB) {
        if (s->incompatible_features & QCOW2_INCOMPAT_COMPRESSION) {
            error_setg(errp, "qcow2: Compression type incompatible feature "
                             "bit must not be set");
            return -EINVAL;
        }
    } else {
        if (!(s->incompatible_features & QCOW2_INCOMPAT_COMPRESSION)) {
            error_setg(errp, "qcow2: Compression type incompatible feature "
                             "bit must be set");
            return -EINVAL;
        }
    }

    return 0;
}

/* Called with s->lock held.  */
static int coroutine_fn qcow2_do_open(BlockDriverState *bs, QDict *options,
                                      int flags, Error **errp)
{
    ERRP_GUARD();
    BDRVQcow2State *s = bs->opaque;
    unsigned int len, i;
    int ret = 0;
    QCowHeader header;
    uint64_t ext_end;
    uint64_t l1_vm_state_index;
    bool update_header = false;

    ret = bdrv_pread(bs->file, 0, &header, sizeof(header));
    if (ret < 0) {
        error_setg_errno(errp, -ret, "Could not read qcow2 header");
        goto fail;
    }
    header.magic = be32_to_cpu(header.magic);
    header.version = be32_to_cpu(header.version);
    header.backing_file_offset = be64_to_cpu(header.backing_file_offset);
    header.backing_file_size = be32_to_cpu(header.backing_file_size);
    header.size = be64_to_cpu(header.size);
    header.cluster_bits = be32_to_cpu(header.cluster_bits);
    header.crypt_method = be32_to_cpu(header.crypt_method);
    header.l1_table_offset = be64_to_cpu(header.l1_table_offset);
    header.l1_size = be32_to_cpu(header.l1_size);
    header.refcount_table_offset = be64_to_cpu(header.refcount_table_offset);
    header.refcount_table_clusters =
        be32_to_cpu(header.refcount_table_clusters);
    header.snapshots_offset = be64_to_cpu(header.snapshots_offset);
    header.nb_snapshots = be32_to_cpu(header.nb_snapshots);

    if (header.magic != QCOW_MAGIC) {
        error_setg(errp, "Image is not in qcow2 format");
        ret = -EINVAL;
        goto fail;
    }
    if (header.version < 2 || header.version > 3) {
        error_setg(errp, "Unsupported qcow2 version %" PRIu32, header.version);
        ret = -ENOTSUP;
        goto fail;
    }

    s->qcow_version = header.version;

    /* Initialise cluster size */
    if (header.cluster_bits < MIN_CLUSTER_BITS ||
        header.cluster_bits > MAX_CLUSTER_BITS) {
        error_setg(errp, "Unsupported cluster size: 2^%" PRIu32,
                   header.cluster_bits);
        ret = -EINVAL;
        goto fail;
    }

    s->cluster_bits = header.cluster_bits;
    s->cluster_size = 1 << s->cluster_bits;

    /* 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 {
        header.incompatible_features =
            be64_to_cpu(header.incompatible_features);
        header.compatible_features = be64_to_cpu(header.compatible_features);
        header.autoclear_features = be64_to_cpu(header.autoclear_features);
        header.refcount_order = be32_to_cpu(header.refcount_order);
        header.header_length = be32_to_cpu(header.header_length);

        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;
    }

    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);
        ret = bdrv_pread(bs->file, sizeof(header), s->unknown_header_fields,
                         s->unknown_header_fields_size);
        if (ret < 0) {
            error_setg_errno(errp, -ret, "Could not read unknown qcow2 header "
                             "fields");
            goto fail;
        }
    }

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

    if (header.backing_file_offset) {
        ext_end = header.backing_file_offset;
    } else {
        ext_end = 1 << header.cluster_bits;
    }

    /* Handle feature bits */
    s->incompatible_features    = header.incompatible_features;
    s->compatible_features      = header.compatible_features;
    s->autoclear_features       = header.autoclear_features;

    /*
     * Handle compression type
     * Older qcow2 images don't contain the compression type header.
     * Distinguish them by the header length and use
     * the only valid (default) compression type in that case
     */
    if (header.header_length > offsetof(QCowHeader, compression_type)) {
        s->compression_type = header.compression_type;
    } else {
        s->compression_type = QCOW2_COMPRESSION_TYPE_ZLIB;
    }

    ret = validate_compression_type(s, errp);
    if (ret) {
        goto fail;
    }

    if (s->incompatible_features & ~QCOW2_INCOMPAT_MASK) {
        void *feature_table = NULL;
        qcow2_read_extensions(bs, header.header_length, ext_end,
                              &feature_table, flags, NULL, NULL);
        report_unsupported_feature(errp, feature_table,
                                   s->incompatible_features &
                                   ~QCOW2_INCOMPAT_MASK);
        ret = -ENOTSUP;
        g_free(feature_table);
        goto fail;
    }

    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)) {
            error_setg(errp, "qcow2: Image is corrupt; cannot be opened "
                       "read/write");
            ret = -EACCES;
            goto fail;
        }
    }

    s->subclusters_per_cluster =
        has_subclusters(s) ? QCOW_EXTL2_SUBCLUSTERS_PER_CLUSTER : 1;
    s->subcluster_size = s->cluster_size / s->subclusters_per_cluster;
    s->subcluster_bits = ctz32(s->subcluster_size);

    if (s->subcluster_size < (1 << MIN_CLUSTER_BITS)) {
        error_setg(errp, "Unsupported subcluster size: %d", s->subcluster_size);
        ret = -EINVAL;
        goto fail;
    }

    /* Check support for various header values */
    if (header.refcount_order > 6) {
        error_setg(errp, "Reference count entry width too large; may not "
                   "exceed 64 bits");
        ret = -EINVAL;
        goto fail;
    }
    s->refcount_order = header.refcount_order;
    s->refcount_bits = 1 << s->refcount_order;
    s->refcount_max = UINT64_C(1) << (s->refcount_bits - 1);
    s->refcount_max += s->refcount_max - 1;

    s->crypt_method_header = header.crypt_method;
    if (s->crypt_method_header) {
        if (bdrv_uses_whitelist() &&
            s->crypt_method_header == QCOW_CRYPT_AES) {
            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;
        }

        if (s->crypt_method_header == QCOW_CRYPT_AES) {
            s->crypt_physical_offset = false;
        } else {
            /* Assuming LUKS and any future crypt methods we
             * add will all use physical offsets, due to the
             * fact that the alternative is insecure...  */
            s->crypt_physical_offset = true;
        }

        bs->encrypted = true;
    }

    s->l2_bits = s->cluster_bits - ctz32(l2_entry_size(s));
    s->l2_size = 1 << s->l2_bits;
    /* 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;
    bs->total_sectors = header.size / BDRV_SECTOR_SIZE;
    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;

    s->refcount_table_offset = header.refcount_table_offset;
    s->refcount_table_size =
        header.refcount_table_clusters << (s->cluster_bits - 3);

    if (header.refcount_table_clusters == 0 && !(flags & BDRV_O_CHECK)) {
        error_setg(errp, "Image does not contain a reference count table");
        ret = -EINVAL;
        goto fail;
    }

    ret = qcow2_validate_table(bs, s->refcount_table_offset,
                               header.refcount_table_clusters,
                               s->cluster_size, QCOW_MAX_REFTABLE_SIZE,
                               "Reference count table", errp);
    if (ret < 0) {
        goto fail;
    }

    if (!(flags & BDRV_O_CHECK)) {
        /*
         * The total size in bytes of the snapshot table is checked in
         * qcow2_read_snapshots() because the size of each snapshot is
         * variable and we don't know it yet.
         * Here we only check the offset and number of snapshots.
         */
        ret = qcow2_validate_table(bs, header.snapshots_offset,
                                   header.nb_snapshots,
                                   sizeof(QCowSnapshotHeader),
                                   sizeof(QCowSnapshotHeader) *
                                       QCOW_MAX_SNAPSHOTS,
                                   "Snapshot table", errp);
        if (ret < 0) {
            goto fail;
        }
    }

    /* read the level 1 table */
    ret = qcow2_validate_table(bs, header.l1_table_offset,
                               header.l1_size, L1E_SIZE,
                               QCOW_MAX_L1_SIZE, "Active L1 table", errp);
    if (ret < 0) {
        goto fail;
    }
    s->l1_size = header.l1_size;
    s->l1_table_offset = header.l1_table_offset;

    l1_vm_state_index = size_to_l1(s, header.size);
    if (l1_vm_state_index > INT_MAX) {
        error_setg(errp, "Image is too big");
        ret = -EFBIG;
        goto fail;
    }
    s->l1_vm_state_index = l1_vm_state_index;

    /* the L1 table must contain at least enough entries to put
       header.size bytes */
    if (s->l1_size < s->l1_vm_state_index) {
        error_setg(errp, "L1 table is too small");
        ret = -EINVAL;
        goto fail;
    }

    if (s->l1_size > 0) {
        s->l1_table = qemu_try_blockalign(bs->file->bs, s->l1_size * L1E_SIZE);
        if (s->l1_table == NULL) {
            error_setg(errp, "Could not allocate L1 table");
            ret = -ENOMEM;
            goto fail;
        }
        ret = bdrv_pread(bs->file, s->l1_table_offset, s->l1_table,
                         s->l1_size * L1E_SIZE);
        if (ret < 0) {
            error_setg_errno(errp, -ret, "Could not read L1 table");
            goto fail;
        }
        for(i = 0;i < s->l1_size; i++) {
            s->l1_table[i] = be64_to_cpu(s->l1_table[i]);
        }
    }

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

    s->flags = flags;

    ret = qcow2_refcount_init(bs);
    if (ret != 0) {
        error_setg_errno(errp, -ret, "Could not initialize refcount handling");
        goto fail;
    }

    QLIST_INIT(&s->cluster_allocs);
    QTAILQ_INIT(&s->discards);

    /* read qcow2 extensions */
    if (qcow2_read_extensions(bs, header.header_length, ext_end, NULL,
                              flags, &update_header, errp)) {
        ret = -EINVAL;
        goto fail;
    }

    /* Open external data file */
    s->data_file = bdrv_open_child(NULL, options, "data-file", bs,
                                   &child_of_bds, BDRV_CHILD_DATA,
                                   true, errp);
    if (*errp) {
        ret = -EINVAL;
        goto fail;
    }

    if (s->incompatible_features & QCOW2_INCOMPAT_DATA_FILE) {
        if (!s->data_file && s->image_data_file) {
            s->data_file = bdrv_open_child(s->image_data_file, options,
                                           "data-file", bs, &child_of_bds,
                                           BDRV_CHILD_DATA, false, errp);
            if (!s->data_file) {
                ret = -EINVAL;
                goto fail;
            }
        }
        if (!s->data_file) {
            error_setg(errp, "'data-file' is required for this image");
            ret = -EINVAL;
            goto fail;
        }

        /* No data here */
        bs->file->role &= ~BDRV_CHILD_DATA;

        /* Must succeed because we have given up permissions if anything */
        bdrv_child_refresh_perms(bs, bs->file, &error_abort);
    } else {
        if (s->data_file) {
            error_setg(errp, "'data-file' can only be set for images with an "
                             "external data file");
            ret = -EINVAL;
            goto fail;
        }

        s->data_file = bs->file;

        if (data_file_is_raw(bs)) {
            error_setg(errp, "data-file-raw requires a data file");
            ret = -EINVAL;
            goto fail;
        }
    }

    /* qcow2_read_extension may have set up the crypto context
     * if the crypt method needs a header region, some methods
     * don't need header extensions, so must check here
     */
    if (s->crypt_method_header && !s->crypto) {
        if (s->crypt_method_header == QCOW_CRYPT_AES) {
            unsigned int cflags = 0;
            if (flags & BDRV_O_NO_IO) {
                cflags |= QCRYPTO_BLOCK_OPEN_NO_IO;
            }
            s->crypto = qcrypto_block_open(s->crypto_opts, "encrypt.",
                                           NULL, NULL, cflags,
                                           QCOW2_MAX_THREADS, errp);
            if (!s->crypto) {
                ret = -EINVAL;
                goto fail;
            }
        } else if (!(flags & BDRV_O_NO_IO)) {
            error_setg(errp, "Missing CRYPTO header for crypt method %d",
                       s->crypt_method_header);
            ret = -EINVAL;
            goto fail;
        }
    }

    /* read the backing file name */
    if (header.backing_file_offset != 0) {
        len = header.backing_file_size;
        if (len > MIN(1023, s->cluster_size - header.backing_file_offset) ||
            len >= sizeof(bs->backing_file)) {
            error_setg(errp, "Backing file name too long");
            ret = -EINVAL;
            goto fail;
        }
        ret = bdrv_pread(bs->file, header.backing_file_offset,
                         bs->auto_backing_file, len);
        if (ret < 0) {
            error_setg_errno(errp, -ret, "Could not read backing file name");
            goto fail;
        }
        bs->auto_backing_file[len] = '\0';
        pstrcpy(bs->backing_file, sizeof(bs->backing_file),
                bs->auto_backing_file);
        s->image_backing_file = g_strdup(bs->auto_backing_file);
    }

    /*
     * Internal snapshots; skip reading them in check mode, because
     * we do not need them then, and we do not want to abort because
     * of a broken table.
     */
    if (!(flags & BDRV_O_CHECK)) {
        s->snapshots_offset = header.snapshots_offset;
        s->nb_snapshots = header.nb_snapshots;

        ret = qcow2_read_snapshots(bs, errp);
        if (ret < 0) {
            goto fail;
        }
    }

    /* Clear unknown autoclear feature bits */
    update_header |= s->autoclear_features & ~QCOW2_AUTOCLEAR_MASK;
    update_header =
        update_header && !bs->read_only && !(flags & BDRV_O_INACTIVE);
    if (update_header) {
        s->autoclear_features &= QCOW2_AUTOCLEAR_MASK;
    }

    /* == Handle persistent dirty bitmaps ==
     *
     * We want load dirty bitmaps in three cases:
     *
     * 1. Normal open of the disk in active mode, not related to invalidation
     *    after migration.
     *
     * 2. Invalidation of the target vm after pre-copy phase of migration, if
     *    bitmaps are _not_ migrating through migration channel, i.e.
     *    'dirty-bitmaps' capability is disabled.
     *
     * 3. Invalidation of source vm after failed or canceled migration.
     *    This is a very interesting case. There are two possible types of
     *    bitmaps:
     *
     *    A. Stored on inactivation and removed. They should be loaded from the
     *       image.
     *
     *    B. Not stored: not-persistent bitmaps and bitmaps, migrated through
     *       the migration channel (with dirty-bitmaps capability).
     *
     *    On the other hand, there are two possible sub-cases:
     *
     *    3.1 disk was changed by somebody else while were inactive. In this
     *        case all in-RAM dirty bitmaps (both persistent and not) are
     *        definitely invalid. And we don't have any method to determine
     *        this.
     *
     *        Simple and safe thing is to just drop all the bitmaps of type B on
     *        inactivation. But in this case we lose bitmaps in valid 4.2 case.
     *
     *        On the other hand, resuming source vm, if disk was already changed
     *        is a bad thing anyway: not only bitmaps, the whole vm state is
     *        out of sync with disk.
     *
     *        This means, that user or management tool, who for some reason
     *        decided to resume source vm, after disk was already changed by
     *        target vm, should at least drop all dirty bitmaps by hand.
     *
     *        So, we can ignore this case for now, but TODO: "generation"
     *        extension for qcow2, to determine, that image was changed after
     *        last inactivation. And if it is changed, we will drop (or at least
     *        mark as 'invalid' all the bitmaps of type B, both persistent
     *        and not).
     *
     *    3.2 disk was _not_ changed while were inactive. Bitmaps may be saved
     *        to disk ('dirty-bitmaps' capability disabled), or not saved
     *        ('dirty-bitmaps' capability enabled), but we don't need to care
     *        of: let's load bitmaps as always: stored bitmaps will be loaded,
     *        and not stored has flag IN_USE=1 in the image and will be skipped
     *        on loading.
     *
     * One remaining possible case when we don't want load bitmaps:
     *
     * 4. Open disk in inactive mode in target vm (bitmaps are migrating or
     *    will be loaded on invalidation, no needs try loading them before)
     */

    if (!(bdrv_get_flags(bs) & BDRV_O_INACTIVE)) {
        /* It's case 1, 2 or 3.2. Or 3.1 which is BUG in management layer. */
        bool header_updated;
        if (!qcow2_load_dirty_bitmaps(bs, &header_updated, errp)) {
            ret = -EINVAL;
            goto fail;
        }

        update_header = update_header && !header_updated;
    }

    if (update_header) {
        ret = qcow2_update_header(bs);
        if (ret < 0) {
            error_setg_errno(errp, -ret, "Could not update qcow2 header");
            goto fail;
        }
    }

    bs->supported_zero_flags = header.version >= 3 ?
                               BDRV_REQ_MAY_UNMAP | BDRV_REQ_NO_FALLBACK : 0;
    bs->supported_truncate_flags = BDRV_REQ_ZERO_WRITE;

    /* Repair image if dirty */
    if (!(flags & (BDRV_O_CHECK | BDRV_O_INACTIVE)) && !bs->read_only &&
        (s->incompatible_features & QCOW2_INCOMPAT_DIRTY)) {
        BdrvCheckResult result = {0};

        ret = qcow2_co_check_locked(bs, &result,
                                    BDRV_FIX_ERRORS | BDRV_FIX_LEAKS);
        if (ret < 0 || result.check_errors) {
            if (ret >= 0) {
                ret = -EIO;
            }
            error_setg_errno(errp, -ret, "Could not repair dirty image");
            goto fail;
        }
    }

#ifdef DEBUG_ALLOC
    {
        BdrvCheckResult result = {0};
        qcow2_check_refcounts(bs, &result, 0);
    }
#endif

    qemu_co_queue_init(&s->thread_task_queue);

    return ret;

 fail:
    g_free(s->image_data_file);
    if (has_data_file(bs)) {
        bdrv_unref_child(bs, s->data_file);
        s->data_file = NULL;
    }
    g_free(s->unknown_header_fields);
    cleanup_unknown_header_ext(bs);
    qcow2_free_snapshots(bs);
    qcow2_refcount_close(bs);
    qemu_vfree(s->l1_table);
    /* else pre-write overlap checks in cache_destroy may crash */
    s->l1_table = NULL;
    cache_clean_timer_del(bs);
    if (s->l2_table_cache) {
        qcow2_cache_destroy(s->l2_table_cache);
    }
    if (s->refcount_block_cache) {
        qcow2_cache_destroy(s->refcount_block_cache);
    }
    qcrypto_block_free(s->crypto);
    qapi_free_QCryptoBlockOpenOptions(s->crypto_opts);
    return ret;
}

typedef struct QCow2OpenCo {
    BlockDriverState *bs;
    QDict *options;
    int flags;
    Error **errp;
    int ret;
} QCow2OpenCo;

static void coroutine_fn qcow2_open_entry(void *opaque)
{
    QCow2OpenCo *qoc = opaque;
    BDRVQcow2State *s = qoc->bs->opaque;

    qemu_co_mutex_lock(&s->lock);
    qoc->ret = qcow2_do_open(qoc->bs, qoc->options, qoc->flags, qoc->errp);
    qemu_co_mutex_unlock(&s->lock);
}

static int qcow2_open(BlockDriverState *bs, QDict *options, int flags,
                      Error **errp)
{
    BDRVQcow2State *s = bs->opaque;
    QCow2OpenCo qoc = {
        .bs = bs,
        .options = options,
        .flags = flags,
        .errp = errp,
        .ret = -EINPROGRESS
    };

    bs->file = bdrv_open_child(NULL, options, "file", bs, &child_of_bds,
                               BDRV_CHILD_IMAGE, false, errp);
    if (!bs->file) {
        return -EINVAL;
    }

    /* Initialise locks */
    qemu_co_mutex_init(&s->lock);

    if (qemu_in_coroutine()) {
        /* From bdrv_co_create.  */
        qcow2_open_entry(&qoc);
    } else {
        assert(qemu_get_current_aio_context() == qemu_get_aio_context());
        qemu_coroutine_enter(qemu_coroutine_create(qcow2_open_entry, &qoc));
        BDRV_POLL_WHILE(bs, qoc.ret == -EINPROGRESS);
    }
    return qoc.ret;
}

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

    if (bs->encrypted) {
        /* Encryption works on a sector granularity */
        bs->bl.request_alignment = qcrypto_block_get_sector_size(s->crypto);
    }
    bs->bl.pwrite_zeroes_alignment = s->subcluster_size;
    bs->bl.pdiscard_alignment = s->cluster_size;
}

static int qcow2_reopen_prepare(BDRVReopenState *state,
                                BlockReopenQueue *queue, Error **errp)
{
    Qcow2ReopenState *r;
    int ret;

    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. */
    if ((state->flags & BDRV_O_RDWR) == 0) {
        ret = qcow2_reopen_bitmaps_ro(state->bs, errp);
        if (ret < 0) {
            goto fail;
        }

        ret = bdrv_flush(state->bs);
        if (ret < 0) {
            goto fail;
        }

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

    return 0;

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_commit_post(BDRVReopenState *state)
{
    if (state->flags & BDRV_O_RDWR) {
        Error *local_err = NULL;

        if (qcow2_reopen_bitmaps_rw(state->bs, &local_err) < 0) {
            /*
             * This is not fatal, bitmaps just left read-only, so all following
             * writes will fail. User can remove read-only bitmaps to unblock
             * writes or retry reopen.
             */
            error_reportf_err(local_err,
                              "%s: Failed to make dirty bitmaps writable: ",
                              bdrv_get_node_name(state->bs));
        }
    }
}

static void qcow2_reopen_abort(BDRVReopenState *state)
{
    qcow2_update_options_abort(state->bs, state->opaque);
    g_free(state->opaque);
}

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);
    }
}

static int coroutine_fn qcow2_co_block_status(BlockDriverState *bs,
                                              bool want_zero,
                                              int64_t offset, int64_t count,
                                              int64_t *pnum, int64_t *map,
                                              BlockDriverState **file)
{
    BDRVQcow2State *s = bs->opaque;
    uint64_t host_offset;
    unsigned int bytes;
    QCow2SubclusterType type;
    int ret, status = 0;

    qemu_co_mutex_lock(&s->lock);

    if (!s->metadata_preallocation_checked) {
        ret = qcow2_detect_metadata_preallocation(bs);
        s->metadata_preallocation = (ret == 1);
        s->metadata_preallocation_checked = true;
    }

    bytes = MIN(INT_MAX, count);
    ret = qcow2_get_host_offset(bs, offset, &bytes, &host_offset, &type);
    qemu_co_mutex_unlock(&s->lock);
    if (ret < 0) {
        return ret;
    }

    *pnum = bytes;

    if ((type == QCOW2_SUBCLUSTER_NORMAL ||
         type == QCOW2_SUBCLUSTER_ZERO_ALLOC ||
         type == QCOW2_SUBCLUSTER_UNALLOCATED_ALLOC) && !s->crypto) {
        *map = host_offset;
        *file = s->data_file->bs;
        status |= BDRV_BLOCK_OFFSET_VALID;
    }
    if (type == QCOW2_SUBCLUSTER_ZERO_PLAIN ||
        type == QCOW2_SUBCLUSTER_ZERO_ALLOC) {
        status |= BDRV_BLOCK_ZERO;
    } else if (type != QCOW2_SUBCLUSTER_UNALLOCATED_PLAIN &&
               type != QCOW2_SUBCLUSTER_UNALLOCATED_ALLOC) {
        status |= BDRV_BLOCK_DATA;
    }
    if (s->metadata_preallocation && (status & BDRV_BLOCK_DATA) &&
        (status & BDRV_BLOCK_OFFSET_VALID))
    {
        status |= BDRV_BLOCK_RECURSE;
    }
    return status;
}

static coroutine_fn int qcow2_handle_l2meta(BlockDriverState *bs,
                                            QCowL2Meta **pl2meta,
                                            bool link_l2)
{
    int ret = 0;
    QCowL2Meta *l2meta = *pl2meta;

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

        if (link_l2) {
            ret = qcow2_alloc_cluster_link_l2(bs, l2meta);
            if (ret) {
                goto out;
            }
        } else {
            qcow2_alloc_cluster_abort(bs, l2meta);
        }

        /* Take the request off the list of running requests */
        QLIST_REMOVE(l2meta, next_in_flight);

        qemu_co_queue_restart_all(&l2meta->dependent_requests);

        next = l2meta->next;
        g_free(l2meta);
        l2meta = next;
    }
out:
    *pl2meta = l2meta;
    return ret;
}

static coroutine_fn int
qcow2_co_preadv_encrypted(BlockDriverState *bs,
                           uint64_t host_offset,
                           uint64_t offset,
                           uint64_t bytes,
                           QEMUIOVector *qiov,
                           uint64_t qiov_offset)
{
    int ret;
    BDRVQcow2State *s = bs->opaque;
    uint8_t *buf;

    assert(bs->encrypted && s->crypto);
    assert(bytes <= QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size);

    /*
     * For encrypted images, read everything into a temporary
     * contiguous buffer on which the AES functions can work.
     * Also, decryption in a separate buffer is better as it
     * prevents the guest from learning information about the
     * encrypted nature of the virtual disk.
     */

    buf = qemu_try_blockalign(s->data_file->bs, bytes);
    if (buf == NULL) {
        return -ENOMEM;
    }

    BLKDBG_EVENT(bs->file, BLKDBG_READ_AIO);
    ret = bdrv_co_pread(s->data_file, host_offset, bytes, buf, 0);
    if (ret < 0) {
        goto fail;
    }

    if (qcow2_co_decrypt(bs, host_offset, offset, buf, bytes) < 0)
    {
        ret = -EIO;
        goto fail;
    }
    qemu_iovec_from_buf(qiov, qiov_offset, buf, bytes);

fail:
    qemu_vfree(buf);

    return ret;
}

typedef struct Qcow2AioTask {
    AioTask task;

    BlockDriverState *bs;
    QCow2SubclusterType subcluster_type; /* only for read */
    uint64_t host_offset; /* or full descriptor in compressed clusters */
    uint64_t offset;
    uint64_t bytes;
    QEMUIOVector *qiov;
    uint64_t qiov_offset;
    QCowL2Meta *l2meta; /* only for write */
} Qcow2AioTask;

static coroutine_fn int qcow2_co_preadv_task_entry(AioTask *task);
static coroutine_fn int qcow2_add_task(BlockDriverState *bs,
                                       AioTaskPool *pool,
                                       AioTaskFunc func,
                                       QCow2SubclusterType subcluster_type,
                                       uint64_t host_offset,
                                       uint64_t offset,
                                       uint64_t bytes,
                                       QEMUIOVector *qiov,
                                       size_t qiov_offset,
                                       QCowL2Meta *l2meta)
{
    Qcow2AioTask local_task;
    Qcow2AioTask *task = pool ? g_new(Qcow2AioTask, 1) : &local_task;

    *task = (Qcow2AioTask) {
        .task.func = func,
        .bs = bs,
        .subcluster_type = subcluster_type,
        .qiov = qiov,
        .host_offset = host_offset,
        .offset = offset,
        .bytes = bytes,
        .qiov_offset = qiov_offset,
        .l2meta = l2meta,
    };

    trace_qcow2_add_task(qemu_coroutine_self(), bs, pool,
                         func == qcow2_co_preadv_task_entry ? "read" : "write",
                         subcluster_type, host_offset, offset, bytes,
                         qiov, qiov_offset);

    if (!pool) {
        return func(&task->task);
    }

    aio_task_pool_start_task(pool, &task->task);

    return 0;
}

static coroutine_fn int qcow2_co_preadv_task(BlockDriverState *bs,
                                             QCow2SubclusterType subc_type,
                                             uint64_t host_offset,
                                             uint64_t offset, uint64_t bytes,
                                             QEMUIOVector *qiov,
                                             size_t qiov_offset)
{
    BDRVQcow2State *s = bs->opaque;

    switch (subc_type) {
    case QCOW2_SUBCLUSTER_ZERO_PLAIN:
    case QCOW2_SUBCLUSTER_ZERO_ALLOC:
        /* Both zero types are handled in qcow2_co_preadv_part */
        g_assert_not_reached();

    case QCOW2_SUBCLUSTER_UNALLOCATED_PLAIN:
    case QCOW2_SUBCLUSTER_UNALLOCATED_ALLOC:
        assert(bs->backing); /* otherwise handled in qcow2_co_preadv_part */

        BLKDBG_EVENT(bs->file, BLKDBG_READ_BACKING_AIO);
        return bdrv_co_preadv_part(bs->backing, offset, bytes,
                                   qiov, qiov_offset, 0);

    case QCOW2_SUBCLUSTER_COMPRESSED:
        return qcow2_co_preadv_compressed(bs, host_offset,
                                          offset, bytes, qiov, qiov_offset);

    case QCOW2_SUBCLUSTER_NORMAL:
        if (bs->encrypted) {
            return qcow2_co_preadv_encrypted(bs, host_offset,
                                             offset, bytes, qiov, qiov_offset);
        }

        BLKDBG_EVENT(bs->file, BLKDBG_READ_AIO);
        return bdrv_co_preadv_part(s->data_file, host_offset,
                                   bytes, qiov, qiov_offset, 0);

    default:
        g_assert_not_reached();
    }

    g_assert_not_reached();
}

static coroutine_fn int qcow2_co_preadv_task_entry(AioTask *task)
{
    Qcow2AioTask *t = container_of(task, Qcow2AioTask, task);

    assert(!t->l2meta);

    return qcow2_co_preadv_task(t->bs, t->subcluster_type,
                                t->host_offset, t->offset, t->bytes,
                                t->qiov, t->qiov_offset);
}

static coroutine_fn int qcow2_co_preadv_part(BlockDriverState *bs,
                                             uint64_t offset, uint64_t bytes,
                                             QEMUIOVector *qiov,
                                             size_t qiov_offset, int flags)
{
    BDRVQcow2State *s = bs->opaque;
    int ret = 0;
    unsigned int cur_bytes; /* number of bytes in current iteration */
    uint64_t host_offset = 0;
    QCow2SubclusterType type;
    AioTaskPool *aio = NULL;

    while (bytes != 0 && aio_task_pool_status(aio) == 0) {
        /* prepare next request */
        cur_bytes = MIN(bytes, INT_MAX);
        if (s->crypto) {
            cur_bytes = MIN(cur_bytes,
                            QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size);
        }

        qemu_co_mutex_lock(&s->lock);
        ret = qcow2_get_host_offset(bs, offset, &cur_bytes,
                                    &host_offset, &type);
        qemu_co_mutex_unlock(&s->lock);
        if (ret < 0) {
            goto out;
        }

        if (type == QCOW2_SUBCLUSTER_ZERO_PLAIN ||
            type == QCOW2_SUBCLUSTER_ZERO_ALLOC ||
            (type == QCOW2_SUBCLUSTER_UNALLOCATED_PLAIN && !bs->backing) ||
            (type == QCOW2_SUBCLUSTER_UNALLOCATED_ALLOC && !bs->backing))
        {
            qemu_iovec_memset(qiov, qiov_offset, 0, cur_bytes);
        } else {
            if (!aio && cur_bytes != bytes) {
                aio = aio_task_pool_new(QCOW2_MAX_WORKERS);
            }
            ret = qcow2_add_task(bs, aio, qcow2_co_preadv_task_entry, type,
                                 host_offset, offset, cur_bytes,
                                 qiov, qiov_offset, NULL);
            if (ret < 0) {
                goto out;
            }
        }

        bytes -= cur_bytes;
        offset += cur_bytes;
        qiov_offset += cur_bytes;
    }

out:
    if (aio) {
        aio_task_pool_wait_all(aio);
        if (ret == 0) {
            ret = aio_task_pool_status(aio);
        }
        g_free(aio);
    }

    return ret;
}

/* Check if it's possible to merge a write request with the writing of
 * the data from the COW regions */
static bool merge_cow(uint64_t offset, unsigned bytes,
                      QEMUIOVector *qiov, size_t qiov_offset,
                      QCowL2Meta *l2meta)
{
    QCowL2Meta *m;

    for (m = l2meta; m != NULL; m = m->next) {
        /* If both COW regions are empty then there's nothing to merge */
        if (m->cow_start.nb_bytes == 0 && m->cow_end.nb_bytes == 0) {
            continue;
        }

        /* If COW regions are handled already, skip this too */
        if (m->skip_cow) {
            continue;
        }

        /*
         * The write request should start immediately after the first
         * COW region. This does not always happen because the area
         * touched by the request can be larger than the one defined
         * by @m (a single request can span an area consisting of a
         * mix of previously unallocated and allocated clusters, that
         * is why @l2meta is a list).
         */
        if (l2meta_cow_start(m) + m->cow_start.nb_bytes != offset) {
            /* In this case the request starts before this region */
            assert(offset < l2meta_cow_start(m));
            assert(m->cow_start.nb_bytes == 0);
            continue;
        }

        /* The write request should end immediately before the second
         * COW region (see above for why it does not always happen) */
        if (m->offset + m->cow_end.offset != offset + bytes) {
            assert(offset + bytes > m->offset + m->cow_end.offset);
            assert(m->cow_end.nb_bytes == 0);
            continue;
        }

        /* Make sure that adding both COW regions to the QEMUIOVector
         * does not exceed IOV_MAX */
        if (qemu_iovec_subvec_niov(qiov, qiov_offset, bytes) > IOV_MAX - 2) {
            continue;
        }

        m->data_qiov = qiov;
        m->data_qiov_offset = qiov_offset;
        return true;
    }

    return false;
}

/*
 * Return 1 if the COW regions read as zeroes, 0 if not, < 0 on error.
 * Note that returning 0 does not guarantee non-zero data.
 */
static int is_zero_cow(BlockDriverState *bs, QCowL2Meta *m)
{
    /*
     * This check is designed for optimization shortcut so it must be
     * efficient.
     * Instead of is_zero(), use bdrv_co_is_zero_fast() as it is
     * faster (but not as accurate and can result in false negatives).
     */
    int ret = bdrv_co_is_zero_fast(bs, m->offset + m->cow_start.offset,
                                   m->cow_start.nb_bytes);
    if (ret <= 0) {
        return ret;
    }

    return bdrv_co_is_zero_fast(bs, m->offset + m->cow_end.offset,
                                m->cow_end.nb_bytes);
}

static int handle_alloc_space(BlockDriverState *bs, QCowL2Meta *l2meta)
{
    BDRVQcow2State *s = bs->opaque;
    QCowL2Meta *m;

    if (!(s->data_file->bs->supported_zero_flags & BDRV_REQ_NO_FALLBACK)) {
        return 0;
    }

    if (bs->encrypted) {
        return 0;
    }

    for (m = l2meta; m != NULL; m = m->next) {
        int ret;
        uint64_t start_offset = m->alloc_offset + m->cow_start.offset;
        unsigned nb_bytes = m->cow_end.offset + m->cow_end.nb_bytes -
            m->cow_start.offset;

        if (!m->cow_start.nb_bytes && !m->cow_end.nb_bytes) {
            continue;
        }

        ret = is_zero_cow(bs, m);
        if (ret < 0) {
            return ret;
        } else if (ret == 0) {
            continue;
        }

        /*
         * instead of writing zero COW buffers,
         * efficiently zero out the whole clusters
         */

        ret = qcow2_pre_write_overlap_check(bs, 0, start_offset, nb_bytes,
                                            true);
        if (ret < 0) {
            return ret;
        }

        BLKDBG_EVENT(bs->file, BLKDBG_CLUSTER_ALLOC_SPACE);
        ret = bdrv_co_pwrite_zeroes(s->data_file, start_offset, nb_bytes,
                                    BDRV_REQ_NO_FALLBACK);
        if (ret < 0) {
            if (ret != -ENOTSUP && ret != -EAGAIN) {
                return ret;
            }
            continue;
        }

        trace_qcow2_skip_cow(qemu_coroutine_self(), m->offset, m->nb_clusters);
        m->skip_cow = true;
    }
    return 0;
}

/*
 * qcow2_co_pwritev_task
 * Called with s->lock unlocked
 * l2meta  - if not NULL, qcow2_co_pwritev_task() will consume it. Caller must
 *           not use it somehow after qcow2_co_pwritev_task() call
 */
static coroutine_fn int qcow2_co_pwritev_task(BlockDriverState *bs,
                                              uint64_t host_offset,
                                              uint64_t offset, uint64_t bytes,
                                              QEMUIOVector *qiov,
                                              uint64_t qiov_offset,
                                              QCowL2Meta *l2meta)
{
    int ret;
    BDRVQcow2State *s = bs->opaque;
    void *crypt_buf = NULL;
    QEMUIOVector encrypted_qiov;

    if (bs->encrypted) {
        assert(s->crypto);
        assert(bytes <= QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size);
        crypt_buf = qemu_try_blockalign(bs->file->bs, bytes);
        if (crypt_buf == NULL) {
            ret = -ENOMEM;
            goto out_unlocked;
        }
        qemu_iovec_to_buf(qiov, qiov_offset, crypt_buf, bytes);

        if (qcow2_co_encrypt(bs, host_offset, offset, crypt_buf, bytes) < 0) {
            ret = -EIO;
            goto out_unlocked;
        }

        qemu_iovec_init_buf(&encrypted_qiov, crypt_buf, bytes);
        qiov = &encrypted_qiov;
        qiov_offset = 0;
    }

    /* Try to efficiently initialize the physical space with zeroes */
    ret = handle_alloc_space(bs, l2meta);
    if (ret < 0) {
        goto out_unlocked;
    }

    /*
     * If we need to do COW, check if it's possible to merge the
     * writing of the guest data together with that of the COW regions.
     * If it's not possible (or not necessary) then write the
     * guest data now.
     */
    if (!merge_cow(offset, bytes, qiov, qiov_offset, l2meta)) {
        BLKDBG_EVENT(bs->file, BLKDBG_WRITE_AIO);
        trace_qcow2_writev_data(qemu_coroutine_self(), host_offset);
        ret = bdrv_co_pwritev_part(s->data_file, host_offset,
                                   bytes, qiov, qiov_offset, 0);
        if (ret < 0) {
            goto out_unlocked;
        }
    }

    qemu_co_mutex_lock(&s->lock);

    ret = qcow2_handle_l2meta(bs, &l2meta, true);
    goto out_locked;

out_unlocked:
    qemu_co_mutex_lock(&s->lock);

out_locked:
    qcow2_handle_l2meta(bs, &l2meta, false);
    qemu_co_mutex_unlock(&s->lock);

    qemu_vfree(crypt_buf);

    return ret;
}

static coroutine_fn int qcow2_co_pwritev_task_entry(AioTask *task)
{
    Qcow2AioTask *t = container_of(task, Qcow2AioTask, task);

    assert(!t->subcluster_type);

    return qcow2_co_pwritev_task(t->bs, t->host_offset,
                                 t->offset, t->bytes, t->qiov, t->qiov_offset,
                                 t->l2meta);
}

static coroutine_fn int qcow2_co_pwritev_part(
        BlockDriverState *bs, uint64_t offset, uint64_t bytes,
        QEMUIOVector *qiov, size_t qiov_offset, int flags)
{
    BDRVQcow2State *s = bs->opaque;
    int offset_in_cluster;
    int ret;
    unsigned int cur_bytes; /* number of sectors in current iteration */
    uint64_t host_offset;
    QCowL2Meta *l2meta = NULL;
    AioTaskPool *aio = NULL;

    trace_qcow2_writev_start_req(qemu_coroutine_self(), offset, bytes);

    while (bytes != 0 && aio_task_pool_status(aio) == 0) {

        l2meta = NULL;

        trace_qcow2_writev_start_part(qemu_coroutine_self());
        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);
        }

        qemu_co_mutex_lock(&s->lock);

        ret = qcow2_alloc_host_offset(bs, offset, &cur_bytes,
                                      &host_offset, &l2meta);
        if (ret < 0) {
            goto out_locked;
        }

        ret = qcow2_pre_write_overlap_check(bs, 0, host_offset,
                                            cur_bytes, true);
        if (ret < 0) {
            goto out_locked;
        }

        qemu_co_mutex_unlock(&s->lock);

        if (!aio && cur_bytes != bytes) {
            aio = aio_task_pool_new(QCOW2_MAX_WORKERS);
        }
        ret = qcow2_add_task(bs, aio, qcow2_co_pwritev_task_entry, 0,
                             host_offset, offset,
                             cur_bytes, qiov, qiov_offset, l2meta);
        l2meta = NULL; /* l2meta is consumed by qcow2_co_pwritev_task() */
        if (ret < 0) {
            goto fail_nometa;
        }

        bytes -= cur_bytes;
        offset += cur_bytes;
        qiov_offset += cur_bytes;
        trace_qcow2_writev_done_part(qemu_coroutine_self(), cur_bytes);
    }
    ret = 0;

    qemu_co_mutex_lock(&s->lock);

out_locked:
    qcow2_handle_l2meta(bs, &l2meta, false);

    qemu_co_mutex_unlock(&s->lock);

fail_nometa:
    if (aio) {
        aio_task_pool_wait_all(aio);
        if (ret == 0) {
            ret = aio_task_pool_status(aio);
        }
        g_free(aio);
    }

    trace_qcow2_writev_done_req(qemu_coroutine_self(), ret);

    return ret;
}

static int qcow2_inactivate(BlockDriverState *bs)
{
    BDRVQcow2State *s = bs->opaque;
    int ret, result = 0;
    Error *local_err = NULL;

    qcow2_store_persistent_dirty_bitmaps(bs, true, &local_err);
    if (local_err != NULL) {
        result = -EINVAL;
        error_reportf_err(local_err, "Lost persistent bitmaps during "
                          "inactivation of node '%s': ",
                          bdrv_get_device_or_node_name(bs));
    }

    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;
}

static void qcow2_close(BlockDriverState *bs)
{
    BDRVQcow2State *s = bs->opaque;
    qemu_vfree(s->l1_table);
    /* else pre-write overlap checks in cache_destroy may crash */
    s->l1_table = NULL;

    if (!(s->flags & BDRV_O_INACTIVE)) {
        qcow2_inactivate(bs);
    }

    cache_clean_timer_del(bs);
    qcow2_cache_destroy(s->l2_table_cache);
    qcow2_cache_destroy(s->refcount_block_cache);

    qcrypto_block_free(s->crypto);
    s->crypto = NULL;
    qapi_free_QCryptoBlockOpenOptions(s->crypto_opts);

    g_free(s->unknown_header_fields);
    cleanup_unknown_header_ext(bs);

    g_free(s->image_data_file);
    g_free(s->image_backing_file);
    g_free(s->image_backing_format);

    if (has_data_file(bs)) {
        bdrv_unref_child(bs, s->data_file);
        s->data_file = NULL;
    }

    qcow2_refcount_close(bs);
    qcow2_free_snapshots(bs);
}

static void coroutine_fn qcow2_co_invalidate_cache(BlockDriverState *bs,
                                                   Error **errp)
{
    ERRP_GUARD();
    BDRVQcow2State *s = bs->opaque;
    int flags = s->flags;
    QCryptoBlock *crypto = NULL;
    QDict *options;
    int ret;

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

    crypto = s->crypto;
    s->crypto = NULL;

    qcow2_close(bs);

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

    flags &= ~BDRV_O_INACTIVE;
    qemu_co_mutex_lock(&s->lock);
    ret = qcow2_do_open(bs, options, flags, errp);
    qemu_co_mutex_unlock(&s->lock);
    qobject_unref(options);
    if (ret < 0) {
        error_prepend(errp, "Could not reopen qcow2 layer: ");
        bs->drv = NULL;
        return;
    }

    s->crypto = crypto;
}

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),
    };

    if (len) {
        memcpy(buf + sizeof(QCowExtension), s, len);
    }

    return ext_len;
}

/*
 * 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.
 *
 * Returns 0 on success, -errno in error cases.
 */
int qcow2_update_header(BlockDriverState *bs)
{
    BDRVQcow2State *s = bs->opaque;
    QCowHeader *header;
    char *buf;
    size_t buflen = s->cluster_size;
    int ret;
    uint64_t total_size;
    uint32_t refcount_table_clusters;
    size_t header_length;
    Qcow2UnknownHeaderExtension *uext;

    buf = qemu_blockalign(bs, buflen);

    /* Header structure */
    header = (QCowHeader*) buf;

    if (buflen < sizeof(*header)) {
        ret = -ENOSPC;
        goto fail;
    }

    header_length = sizeof(*header) + s->unknown_header_fields_size;
    total_size = bs->total_sectors * BDRV_SECTOR_SIZE;
    refcount_table_clusters = s->refcount_table_size >> (s->cluster_bits - 3);

    ret = validate_compression_type(s, NULL);
    if (ret) {
        goto fail;
    }

    *header = (QCowHeader) {
        /* Version 2 fields */
        .magic                  = cpu_to_be32(QCOW_MAGIC),
        .version                = cpu_to_be32(s->qcow_version),
        .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),

        /* 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),
        .refcount_order         = cpu_to_be32(s->refcount_order),
        .header_length          = cpu_to_be32(header_length),
        .compression_type       = s->compression_type,
    };

    /* 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:
        ret = -EINVAL;
        goto fail;
    }

    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;
    }

    /* Backing file format header extension */
    if (s->image_backing_format) {
        ret = header_ext_add(buf, QCOW2_EXT_MAGIC_BACKING_FORMAT,
                             s->image_backing_format,
                             strlen(s->image_backing_format),
                             buflen);
        if (ret < 0) {
            goto fail;
        }

        buf += ret;
        buflen -= ret;
    }

    /* External data file header extension */
    if (has_data_file(bs) && s->image_data_file) {
        ret = header_ext_add(buf, QCOW2_EXT_MAGIC_DATA_FILE,
                             s->image_data_file, strlen(s->image_data_file),
                             buflen);
        if (ret < 0) {
            goto fail;
        }

        buf += ret;
        buflen -= ret;
    }

    /* Full disk encryption header pointer extension */
    if (s->crypto_header.offset != 0) {
        s->crypto_header.offset = cpu_to_be64(s->crypto_header.offset);
        s->crypto_header.length = cpu_to_be64(s->crypto_header.length);
        ret = header_ext_add(buf, QCOW2_EXT_MAGIC_CRYPTO_HEADER,
                             &s->crypto_header, sizeof(s->crypto_header),
                             buflen);
        s->crypto_header.offset = be64_to_cpu(s->crypto_header.offset);
        s->crypto_header.length = be64_to_cpu(s->crypto_header.length);
        if (ret < 0) {
            goto fail;
        }
        buf += ret;
        buflen -= ret;
    }

    /*
     * Feature table.  A mere 8 feature names occupies 392 bytes, and
     * when coupled with the v3 minimum header of 104 bytes plus the
     * 8-byte end-of-extension marker, that would leave only 8 bytes
     * for a backing file name in an image with 512-byte clusters.
     * Thus, we choose to omit this header for cluster sizes 4k and
     * smaller.
     */
    if (s->qcow_version >= 3 && s->cluster_size > 4096) {
        static const 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_INCOMPATIBLE,
                .bit  = QCOW2_INCOMPAT_DATA_FILE_BITNR,
                .name = "external data file",
            },
            {
                .type = QCOW2_FEAT_TYPE_INCOMPATIBLE,
                .bit  = QCOW2_INCOMPAT_COMPRESSION_BITNR,
                .name = "compression type",
            },
            {
                .type = QCOW2_FEAT_TYPE_INCOMPATIBLE,
                .bit  = QCOW2_INCOMPAT_EXTL2_BITNR,
                .name = "extended L2 entries",
            },
            {
                .type = QCOW2_FEAT_TYPE_COMPATIBLE,
                .bit  = QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR,
                .name = "lazy refcounts",
            },
            {
                .type = QCOW2_FEAT_TYPE_AUTOCLEAR,
                .bit  = QCOW2_AUTOCLEAR_BITMAPS_BITNR,
                .name = "bitmaps",
            },
            {
                .type = QCOW2_FEAT_TYPE_AUTOCLEAR,
                .bit  = QCOW2_AUTOCLEAR_DATA_FILE_RAW_BITNR,
                .name = "raw external data",
            },
        };

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

    /* Bitmap extension */
    if (s->nb_bitmaps > 0) {
        Qcow2BitmapHeaderExt bitmaps_header = {
            .nb_bitmaps = cpu_to_be32(s->nb_bitmaps),
            .bitmap_directory_size =
                    cpu_to_be64(s->bitmap_directory_size),
            .bitmap_directory_offset =
                    cpu_to_be64(s->bitmap_directory_offset)
        };
        ret = header_ext_add(buf, QCOW2_EXT_MAGIC_BITMAPS,
                             &bitmaps_header, sizeof(bitmaps_header),
                             buflen);
        if (ret < 0) {
            goto fail;
        }
        buf += ret;
        buflen -= ret;
    }

    /* 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) {