GRO tests are timing dependent and can easily flake. This is partially
mitigated in gro.sh by giving each subtest 3 chances to pass. However,
this still flakes on some machines. Reduce the flakiness by:
- Bumping retries to 6.
- Setting napi_defer_hard_irqs to 1 to reduce the chance that GRO is
flushed prematurely. This also lets us reduce the gro_flush_timeout
from 1ms to 100us.
Tested: Ran `gro.sh -t large` 1000 times. There were no failures with
this change. Ran inside strace to increase flakiness.
Signed-off-by: Kevin Krakauer <krakauer@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://patch.msgid.link/20250226192725.621969-4-krakauer@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
"--smac" "${CLIENT_MAC}" "--test" "${test}" "--verbose" )
setup_ns
- # Each test is run 3 times to deflake, because given the receive timing,
+ # Each test is run 6 times to deflake, because given the receive timing,
# not all packets that should coalesce will be considered in the same flow
# on every try.
- for tries in {1..3}; do
+ for tries in {1..6}; do
# Actual test starts here
ip netns exec $server_ns ./gro "${ARGS[@]}" "--rx" "--iface" "server" \
1>>log.txt &
local -r ns_mac="$4"
[[ -e /var/run/netns/"${ns_name}" ]] || ip netns add "${ns_name}"
- echo 1000000 > "/sys/class/net/${ns_dev}/gro_flush_timeout"
+ echo 100000 > "/sys/class/net/${ns_dev}/gro_flush_timeout"
+ echo 1 > "/sys/class/net/${ns_dev}/napi_defer_hard_irqs"
ip link set dev "${ns_dev}" netns "${ns_name}" mtu 65535
ip -netns "${ns_name}" link set dev "${ns_dev}" up