Am 04.03.2015 um 10:04 hat Fam Zheng geschrieben:
> Aio context switch should just work because the requests will be
> drained, so the scheduled timer(s) on the old context will be freed.
>
> Signed-off-by: Fam Zheng <f...@redhat.com>
You should probably also add co_aio_sleep_ns() calls to
null_co_readv/write/flush().
> block/null.c | 34 ++++++++++++++++++++++++++++++++--
> 1 file changed, 32 insertions(+), 2 deletions(-)
>
> diff --git a/block/null.c b/block/null.c
> index ec2bd27..af49525 100644
> --- a/block/null.c
> +++ b/block/null.c
> @@ -12,8 +12,11 @@
>
> #include "block/block_int.h"
>
> +#define NULL_OPT_LATENCY "latency_ns"
> +
> typedef struct {
> int64_t length;
> + int64_t latency_ns;
> } BDRVNullState;
>
> static QemuOptsList runtime_opts = {
> @@ -30,6 +33,12 @@ static QemuOptsList runtime_opts = {
> .type = QEMU_OPT_SIZE,
> .help = "size of the null block",
> },
> + {
> + .name = NULL_OPT_LATENCY,
> + .type = QEMU_OPT_NUMBER,
> + .help = "nanoseconds (approximated) to wait "
> + "before completing request",
> + },
> { /* end of list */ }
> },
> };
> @@ -44,6 +53,8 @@ static int null_file_open(BlockDriverState *bs, QDict
> *options, int flags,
> qemu_opts_absorb_qdict(opts, options, &error_abort);
> s->length =
> qemu_opt_get_size(opts, BLOCK_OPT_SIZE, 1 << 30);
> + s->latency_ns =
> + qemu_opt_get_number(opts, NULL_OPT_LATENCY, 0);
s->latency_ns is signed, so I'd check for negative values.
The rest looks good to me.
Kevin
