[PATCH v2] nvme-cli/fabrics: Add nr_io_queues parameter to connect command
J Freyensee
james_p_freyensee at linux.intel.com
Mon Aug 29 11:52:07 PDT 2016
On Mon, 2016-08-29 at 08:53 -0700, Steve Wise wrote:
> Added 'nr-io-queues' to 'connect' command so users can specify the
> number
> of io queues to allocate.
>
> usage examples:
>
> nvme connect --transport=rdma --nr-io-queues=8 --trsvcid=4420 --
> traddr=10.0.1.14 --nqn=test-nvme
> nvme connect -t rdma -i 8 -s 4420 -a 10.0.1.14 -n test-nvme
>
> Signed-off-by: Steve Wise <swise at opengridcomputing.com>
> ---
>
> Changes since v1:
>
> -Q -> -i
> nr_io_queues -> nr-io-queues
>
> ---
> fabrics.c | 11 +++++++++++
> 1 file changed, 11 insertions(+)
>
> diff --git a/fabrics.c b/fabrics.c
> index 942e987..18e961c 100644
> --- a/fabrics.c
> +++ b/fabrics.c
> @@ -50,6 +50,7 @@ struct config {
> char *traddr;
> char *trsvcid;
> char *hostnqn;
> + char *nr_io_queues;
I just noticed that the 'nr_io_queues' parameter type used in the host
driver code fabrics.c is unsigned int. Would there be a truncation
problem here, like say someone wants to do:
nvme connect --nr-io-queues=256
??
maybe nr_io_queues here also needs to be of type unsigned int.
Other than that, I'm fine with the changes.
> char *raw;
> char *device;
> } cfg = { 0 };
> @@ -481,6 +482,14 @@ static int build_options(char *argstr, int
> max_len)
> max_len -= len;
> }
>
> + if (cfg.nr_io_queues) {
> + len = snprintf(argstr, max_len, ",nr_io_queues=%s",
> cfg.nr_io_queues);
> + if (len < 0)
> + return -EINVAL;
> + argstr += len;
> + max_len -= len;
> + }
> +
> return 0;
> }
>
> @@ -713,6 +722,8 @@ int connect(const char *desc, int argc, char
> **argv)
> "transport service id (e.g. IP port)" },
> {"hostnqn", 'q', "LIST", CFG_STRING, &cfg.hostnqn,
> required_argument,
> "user-defined hostnqn" },
> + {"nr-io-queues", 'i', "LIST", CFG_STRING,
> &cfg.nr_io_queues, required_argument,
> + "number of io queues to use (default is core
> count)" },
> {0},
> };
>
More information about the Linux-nvme
mailing list