Am 25.07.2012 14:21, schrieb Stefan Hajnoczi:
> Lazy refcounts is a performance optimization for qcow2 that postpones
> refcount metadata updates and instead marks the image dirty. In the
> case of crash or power failure the image will be left in a dirty state
> and repaired next time it is opened.
>
> Reducing metadata I/O is important for cache=writethrough and
> cache=directsync because these modes guarantee that data is on disk
> after each write (hence we cannot take advantage of caching updates in
> RAM). Refcount metadata is not needed for guest->file block address
> translation and therefore does not need to be on-disk at the time of
> write completion - this is the motivation behind the lazy refcount
> optimization.
>
> The lazy refcount optimization must be enabled at image creation time:
>
> qemu-img create -f qcow2 -o compat=1.1,lazy_refcounts=on a.qcow2 10G
> qemu-system-x86_64 -drive if=virtio,file=a.qcow2,cache=writethrough
>
> Update qemu-iotests 031 and 036 since the extension header size changes
> when we add feature bit table entries.
>
> Signed-off-by: Stefan Hajnoczi <stefa...@linux.vnet.ibm.com>
> ---
> block/qcow2-cluster.c | 5 +++-
> block/qcow2.c | 71
> +++++++++++++++++++++++++++++++++++++++++---
> block/qcow2.h | 13 ++++++++
> block_int.h | 26 ++++++++--------
> tests/qemu-iotests/031.out | 12 ++++----
> tests/qemu-iotests/036.out | 2 +-
> 6 files changed, 105 insertions(+), 24 deletions(-)
>
> diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c
> index d7e0e19..e179211 100644
> --- a/block/qcow2-cluster.c
> +++ b/block/qcow2-cluster.c
> @@ -662,7 +662,10 @@ int qcow2_alloc_cluster_link_l2(BlockDriverState *bs,
> QCowL2Meta *m)
> qcow2_cache_depends_on_flush(s->l2_table_cache);
> }
>
> - qcow2_cache_set_dependency(bs, s->l2_table_cache,
> s->refcount_block_cache);
> + if (qcow2_need_accurate_refcounts(s)) {
> + qcow2_cache_set_dependency(bs, s->l2_table_cache,
> + s->refcount_block_cache);
> + }
> ret = get_cluster_table(bs, m->offset, &l2_table, &l2_index);
> if (ret < 0) {
> goto err;
> diff --git a/block/qcow2.c b/block/qcow2.c
> index 7fe1567..d48527f7 100644
> --- a/block/qcow2.c
> +++ b/block/qcow2.c
> @@ -215,6 +215,39 @@ static void report_unsupported_feature(BlockDriverState
> *bs,
> }
>
> /*
> + * 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.
> + */
> +static int qcow2_mark_dirty(BlockDriverState *bs)
> +{
> + BDRVQcowState *s = bs->opaque;
> + uint64_t val;
> + int ret;
> +
> + 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 you respin, I think would be nice to have either an assert(s->version
== 3) before writing to qcow3 header fields, or to use
qcow2_update_header() in the first place.
Kevin
