On Mon, Nov 16, 2020 at 04:45:03PM -0800, Jakub Kicinski wrote:
> On Sat, 14 Nov 2020 00:16:55 +0100 Antonio Cardace wrote:
> > Add scripts to test ring and coalesce settings
> > of netdevsim.
> >
> > Signed-off-by: Antonio Cardace <acard...@redhat.com>
>
> > @@ -0,0 +1,68 @@
> > +#!/bin/bash
> > +# SPDX-License-Identifier: GPL-2.0-only
> > +
> > +source ethtool-common.sh
> > +
> > +function get_value {
> > + local key=$1
> > +
> > + echo $(ethtool -c $NSIM_NETDEV | \
> > + awk -F':' -v pattern="$key:" '$0 ~ pattern {gsub(/[ \t]/, "", $2);
> > print $2}')
> > +}
> > +
> > +if ! ethtool -h | grep -q coalesce; then
> > + echo "SKIP: No --coalesce support in ethtool"
> > + exit 4
>
> I think the skip exit code for selftests is 2
In the ethtool-pause.sh selftest the exit code is 4 (I copied it from
there), should I change that too?
>
> > +fi
> > +
> > +NSIM_NETDEV=$(make_netdev)
> > +
> > +set -o pipefail
> > +
> > +declare -A SETTINGS_MAP=(
> > + ["rx-frames-low"]="rx-frame-low"
> > + ["tx-frames-low"]="tx-frame-low"
> > + ["rx-frames-high"]="rx-frame-high"
> > + ["tx-frames-high"]="tx-frame-high"
> > + ["rx-usecs"]="rx-usecs"
> > + ["rx-frames"]="rx-frames"
> > + ["rx-usecs-irq"]="rx-usecs-irq"
> > + ["rx-frames-irq"]="rx-frames-irq"
> > + ["tx-usecs"]="tx-usecs"
> > + ["tx-frames"]="tx-frames"
> > + ["tx-usecs-irq"]="tx-usecs-irq"
> > + ["tx-frames-irq"]="tx-frames-irq"
> > + ["stats-block-usecs"]="stats-block-usecs"
> > + ["pkt-rate-low"]="pkt-rate-low"
> > + ["rx-usecs-low"]="rx-usecs-low"
> > + ["tx-usecs-low"]="tx-usecs-low"
> > + ["pkt-rate-high"]="pkt-rate-high"
> > + ["rx-usecs-high"]="rx-usecs-high"
> > + ["tx-usecs-high"]="tx-usecs-high"
> > + ["sample-interval"]="sample-interval"
> > +)
> > +
> > +for key in ${!SETTINGS_MAP[@]}; do
> > + query_key=${SETTINGS_MAP[$key]}
> > + value=$((RANDOM % $((2**32-1))))
> > + ethtool -C $NSIM_NETDEV "$key" "$value"
> > + s=$(get_value "$query_key")
>
> It would be better to validate the entire config, not just the most
> recently set key. This way we would catch the cases where setting
> attr breaks the value of another.
>
Good idea, will do.
Thanks,
Antonio
