Am 23.04.2020 um 12:53 hat Max Reitz geschrieben: > On 22.04.20 17:21, Kevin Wolf wrote: > > If BDRV_REQ_ZERO_WRITE is set and we're extending the image, calling > > qcow2_cluster_zeroize() with flags=0 does the right thing: It doesn't > > undo any previous preallocation, but just adds the zero flag to all > > relevant L2 entries. If an external data file is in use, a write_zeroes > > request to the data file is made instead. > > > > Signed-off-by: Kevin Wolf <kw...@redhat.com> > > --- > > block/qcow2.c | 30 ++++++++++++++++++++++++++++++ > > 1 file changed, 30 insertions(+) > > > > diff --git a/block/qcow2.c b/block/qcow2.c > > index 9cfbdfc939..bd632405d1 100644 > > --- a/block/qcow2.c > > +++ b/block/qcow2.c > > [...] > > > @@ -4214,6 +4215,35 @@ static int coroutine_fn > > qcow2_co_truncate(BlockDriverState *bs, int64_t offset, > > g_assert_not_reached(); > > } > > > > + if ((flags & BDRV_REQ_ZERO_WRITE) && offset > old_length) { > > + uint64_t zero_start = QEMU_ALIGN_UP(old_length, s->cluster_size); > > + uint64_t zero_end = QEMU_ALIGN_UP(offset, s->cluster_size); > > + > > + /* Use zero clusters as much as we can */ > > + ret = qcow2_cluster_zeroize(bs, zero_start, zero_end - zero_start, > > 0); > > It’s kind of a pity that this changes the cluster mappings that were > established when using falloc/full preallocation already (i.e., they > become preallocated zero clusters then, so when writing to them, we need > COW again). > > But falloc/full preallocation do not guarantee that the new data is > zero, so I suppose this is the only thing we can reasonably do.
If we really want, I guess we could make full preallocation first try passing BDRV_REQ_ZERO_WRITE to the protocol layer and if that succeeds, we could skip setting the zero cluster flag at the qcow2 level. Feels like a separate patch, though. Kevin
signature.asc
Description: PGP signature