* Re: [PATCH net-next] selftests: drv-net: devmem: set reuseaddr on the sender's source port
2026-09-08 18:34 [PATCH net-next] selftests: drv-net: devmem: set reuseaddr on the sender's source port Jakub Kicinski
@ 2026-09-09 9:19 ` Breno Leitao
2026-09-09 18:35 ` netdev-bot+sashiko
2026-09-09 18:40 ` Bobby Eshleman
2 siblings, 0 replies; 4+ messages in thread
From: Breno Leitao @ 2026-09-09 9:19 UTC (permalink / raw)
To: Jakub Kicinski
Cc: davem, netdev, edumazet, pabeni, andrew+netdev, horms, shuah,
bobbyeshleman, sdf, razor, linux-kselftest
On Tue, Sep 08, 2026 at 11:34:20AM -0700, Jakub Kicinski wrote:
> devmem.py fails on the HW runners with:
>
> CMD[remote]: dd if=/dev/zero bs=512 count=1 2>/dev/null | socat -b 512 \
> -u - TCP6:[fd00:2::1]:50051,bind=[fd00:2::2]:50051,nodelay
> STDERR: socat[41018] W bind(5, {AF=10 [fd00:2::2]:50051}, 28): \
> Address already in use
>
> ncdevmem installs a 5-tuple flow rule which matches the source port, so
> socat has to bind it explicitly. The port comes from rand_port(), which
> checks availability on the DUT - but we bind on the remote...
> That said the failure rate seems to high to be random collisions
> (~2% per sub-test). It's probably TIME_WAIT sockets on the remote,
> run_rx_hds() alone creates 12 of them.
>
> Set reuseaddr so a TIME_WAIT socket does not fail the bind. Collisions
> with a live socket are still possible, we'll see if they are frequent
> enough to care.
>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Breno Leitao <leitao@debian.org>
> --- a/tools/testing/selftests/drivers/net/hw/devmem_lib.py
> +++ b/tools/testing/selftests/drivers/net/hw/devmem_lib.py
> @@ -201,7 +201,7 @@ def ncdevmem_rx(cfg, port, verify=True, fail_on_linear=False, flow_steer=False,
> else:
> addr = cfg.baddr
>
> - suffix = f",bind={cfg.remote_baddr}:{port}"
> + suffix = f",bind={cfg.remote_baddr}:{port},reuseaddr"
socat_listen(), used by run_tx() and run_tx_chunks(), still builds
without reuseaddr. From my reading, I understand we don't need it given
ncdevmem does the active close on those tests. Leaving it alone seems
right.
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: [PATCH net-next] selftests: drv-net: devmem: set reuseaddr on the sender's source port
2026-09-08 18:34 [PATCH net-next] selftests: drv-net: devmem: set reuseaddr on the sender's source port Jakub Kicinski
2026-09-09 9:19 ` Breno Leitao
@ 2026-09-09 18:35 ` netdev-bot+sashiko
2026-09-09 18:40 ` Bobby Eshleman
2 siblings, 0 replies; 4+ messages in thread
From: netdev-bot+sashiko @ 2026-09-09 18:35 UTC (permalink / raw)
To: kuba
Cc: davem, netdev, edumazet, pabeni, andrew+netdev, horms, shuah,
bobbyeshleman, sdf, leitao, razor, linux-kselftest
Thank you for your contribution! Sashiko AI review found 1 potential
issue(s) to consider:
Critical: 0 · High: 0 · Medium: 1 · Low: 0
- [Medium] Incomplete fix: the sibling helper socat_listen() in the same
file still has no reuse option, so the EADDRINUSE/TIME_WAIT flake the…
--
selftests: drv-net: devmem: set reuseaddr on the sender's source port
The devmem selftest's sender has to bind an explicit source port on the
remote host because ncdevmem installs a 5-tuple flow rule. The port is
picked by rand_port() on the DUT, so lingering TIME_WAIT sockets on the
remote can make the bind fail with EADDRINUSE; this adds reuseaddr to the
socat bind option.
> diff --git a/tools/testing/selftests/drivers/net/hw/devmem_lib.py b/tools/testing/selftests/drivers/net/hw/devmem_lib.py
> index 3554954a6691..b6e52fa9dda5 100644
> --- a/tools/testing/selftests/drivers/net/hw/devmem_lib.py
> +++ b/tools/testing/selftests/drivers/net/hw/devmem_lib.py
> @@ -201,7 +201,7 @@ def socat_send(cfg, port, buf_size=0):
> else:
> addr = cfg.baddr
>
> - suffix = f",bind={cfg.remote_baddr}:{port}"
> + suffix = f",bind={cfg.remote_baddr}:{port},reuseaddr"
>
> buf = ""
> if buf_size:
[Severity: Medium]
Should socat_listen() in the same file get the same treatment? It builds
the remote listener for the TX tests without any reuse option:
tools/testing/selftests/drivers/net/hw/devmem_lib.py:
def socat_listen(cfg, port):
"""Socat listen command for TX tests."""
return f"socat -U - TCP{cfg.addr_ipver}-LISTEN:{port}"
That command also runs on cfg.remote, and the port again comes from
rand_port(), which only binds a probe socket on the DUT. The RX sub-tests
are what leave the TIME_WAIT sockets described in the changelog, and they
sit on remote_baddr:<port> in the same ephemeral range that rand_port()
draws from.
If run_tx() or run_tx_chunks() then draws a colliding port, the wildcard
bind of the listener conflicts with the lingering TIME_WAIT socket without
SO_REUSEADDR, socat exits, and:
with bkg(listen_cmd, host=cfg.remote, exit_wait=True) as socat:
wait_port_listen(port, host=cfg.remote)
times out, failing the sub-test. Is that the same failure mode the patch
removes for the sender, just on the listener side?
For reference, the other socat listeners under drivers/net (ping.py, tso.py,
uso.py, xdp.py, xdp_metadata.py, ntuple.py) all pass ",reuseport"
explicitly, so devmem_lib.py's socat_listen() looks like the only bare one.
--
Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20260908183420.1369630-1-kuba%40kernel.org
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: [PATCH net-next] selftests: drv-net: devmem: set reuseaddr on the sender's source port
2026-09-08 18:34 [PATCH net-next] selftests: drv-net: devmem: set reuseaddr on the sender's source port Jakub Kicinski
2026-09-09 9:19 ` Breno Leitao
2026-09-09 18:35 ` netdev-bot+sashiko
@ 2026-09-09 18:40 ` Bobby Eshleman
2 siblings, 0 replies; 4+ messages in thread
From: Bobby Eshleman @ 2026-09-09 18:40 UTC (permalink / raw)
To: Jakub Kicinski
Cc: davem, netdev, edumazet, pabeni, andrew+netdev, horms, shuah,
bobbyeshleman, sdf, leitao, razor, linux-kselftest
On Tue, Sep 08, 2026 at 11:34:20AM -0700, Jakub Kicinski wrote:
> devmem.py fails on the HW runners with:
>
> CMD[remote]: dd if=/dev/zero bs=512 count=1 2>/dev/null | socat -b 512 \
> -u - TCP6:[fd00:2::1]:50051,bind=[fd00:2::2]:50051,nodelay
> STDERR: socat[41018] W bind(5, {AF=10 [fd00:2::2]:50051}, 28): \
> Address already in use
>
> ncdevmem installs a 5-tuple flow rule which matches the source port, so
> socat has to bind it explicitly. The port comes from rand_port(), which
> checks availability on the DUT - but we bind on the remote...
> That said the failure rate seems to high to be random collisions
> (~2% per sub-test). It's probably TIME_WAIT sockets on the remote,
> run_rx_hds() alone creates 12 of them.
>
> Set reuseaddr so a TIME_WAIT socket does not fail the bind. Collisions
> with a live socket are still possible, we'll see if they are frequent
> enough to care.
>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
> ---
> CC: shuah@kernel.org
> CC: bobbyeshleman@meta.com
> CC: sdf@fomichev.me
> CC: leitao@debian.org
> CC: razor@blackwall.org
> CC: linux-kselftest@vger.kernel.org
> ---
> tools/testing/selftests/drivers/net/hw/devmem_lib.py | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/tools/testing/selftests/drivers/net/hw/devmem_lib.py b/tools/testing/selftests/drivers/net/hw/devmem_lib.py
> index 3554954a6691..b6e52fa9dda5 100644
> --- a/tools/testing/selftests/drivers/net/hw/devmem_lib.py
> +++ b/tools/testing/selftests/drivers/net/hw/devmem_lib.py
> @@ -201,7 +201,7 @@ def ncdevmem_rx(cfg, port, verify=True, fail_on_linear=False, flow_steer=False,
> else:
> addr = cfg.baddr
>
> - suffix = f",bind={cfg.remote_baddr}:{port}"
> + suffix = f",bind={cfg.remote_baddr}:{port},reuseaddr"
>
> buf = ""
> if buf_size:
> --
> 2.55.0
>
Reviewed-by: Bobby Eshleman <bobbyeshleman@meta.com>
^ permalink raw reply [flat|nested] 4+ messages in thread