Am 27.06.2013 um 15:11 hat Peter Lieven geschrieben:
> if the device supports unmapping and unmapped blocks read as
> zero ensure that the whole device is unmapped and report
> .has_zero_init = 1 in this case to speed up qemu-img convert.
>
> Signed-off-by: Peter Lieven <p...@kamp.de>
> ---
> block/iscsi.c | 72
> ++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
> 1 file changed, 71 insertions(+), 1 deletion(-)
>
> diff --git a/block/iscsi.c b/block/iscsi.c
> index 92e66a6..621ca40 100644
> --- a/block/iscsi.c
> +++ b/block/iscsi.c
> @@ -1436,7 +1436,8 @@ static int iscsi_truncate(BlockDriverState *bs, int64_t
> offset)
>
> static int iscsi_has_zero_init(BlockDriverState *bs)
> {
> - return 0;
> + IscsiLun *iscsilun = bs->opaque;
> + return (iscsilun->lbpu && iscsilun->lbprz) ? 1 : 0;
> }
>
> static int iscsi_create(const char *filename, QEMUOptionParameter *options)
> @@ -1446,6 +1447,7 @@ static int iscsi_create(const char *filename,
> QEMUOptionParameter *options)
> BlockDriverState bs;
> IscsiLun *iscsilun = NULL;
> QDict *bs_options;
> + struct scsi_task *task = NULL;
>
> memset(&bs, 0, sizeof(BlockDriverState));
>
> @@ -1481,7 +1483,75 @@ static int iscsi_create(const char *filename,
> QEMUOptionParameter *options)
> }
>
> ret = 0;
Noticed something independent from your patch, a bit more context:
if (bs.total_sectors < total_size) {
ret = -ENOSPC;
}
ret = 0;
The -ENOSPC case can't ever trigger, there's a 'goto out' missing. Maybe
you can include another patch too fix this in the next version of your
series?
Kevin
