Am 11.11.20 um 16:39 schrieb Maxim Levitsky:
> This helps avoid unneeded writes and discards.
>
> Signed-off-by: Maxim Levitsky <mlevi...@redhat.com>
> ---
> qemu-img.c | 13 ++++++++-----
> 1 file changed, 8 insertions(+), 5 deletions(-)
>
> diff --git a/qemu-img.c b/qemu-img.c
> index c2c56fc797..7e9b0f659f 100644
> --- a/qemu-img.c
> +++ b/qemu-img.c
> @@ -1722,7 +1722,7 @@ static void convert_select_part(ImgConvertState *s,
> int64_t sector_num,
> static int convert_iteration_sectors(ImgConvertState *s, int64_t sector_num)
> {
> int64_t src_cur_offset;
> - int ret, n, src_cur;
> + int ret, n, src_cur, alignment;
> bool post_backing_zero = false;
>
> convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
> @@ -1785,11 +1785,14 @@ static int convert_iteration_sectors(ImgConvertState
> *s, int64_t sector_num)
> n = DIV_ROUND_UP(count, BDRV_SECTOR_SIZE);
>
> /*
> - * Avoid that s->sector_next_status becomes unaligned to the source
> - * request alignment and/or cluster size to avoid unnecessary read
> - * cycles.
> + * Avoid that s->sector_next_status becomes unaligned to the
> + * source/destination request alignment and/or cluster size to avoid
> + * unnecessary read/write cycles.
> */
> - tail = (sector_num - src_cur_offset + n) % s->src_alignment[src_cur];
> + alignment = MAX(s->src_alignment[src_cur], s->alignment);
> + assert(is_power_of_2(alignment));
> +
> + tail = (sector_num - src_cur_offset + n) % alignment;
> if (n > tail) {
> n -= tail;
> }
I was also considering including the s->alignment when adding this chance.
However, you need the least common multiple of both alignments, not the
maximum, otherwise
you might get misaligned to either source or destination.
Why exactly do you need the power of two requirement?
Peter
