matttbe
commented
7 months ago I don't know if this one is linked to it, but one MPTCP Join is also "regularly" failing:
# 112 userspace pm add & remove address
# syn [ ok ]
# synack [ ok ]
# ack [ ok ]
# add [ ok ]
# echo [ ok ]
# mptcp_info subflows=2:2 [ ok ]
# subflows_total 3:3 [ ok ]
# mptcp_info add_addr_signal=2:2 [ ok ]
# netlink error -3 (No such process)
# ./pm_nl_ctl: bailing out due to netlink error[s]
# rm [ ok ]
# rmsf [fail] got 1 RM_SUBFLOW[s] expected 2
# Server ns stats
# TcpPassiveOpens 3 0.0
# TcpEstabResets 1 0.0
# TcpInSegs 249 0.0
# TcpOutSegs 200 0.0
# TcpExtDelayedACKs 39 0.0
# TcpExtTCPPureAcks 116 0.0
# TcpExtTCPOrigDataSent 130 0.0
# TcpExtTCPDelivered 129 0.0
# MPTcpExtMPCapableSYNRX 1 0.0
# MPTcpExtMPCapableACKRX 1 0.0
# MPTcpExtMPJoinSynRx 2 0.0
# MPTcpExtMPJoinAckRx 2 0.0
# MPTcpExtEchoAdd 2 0.0
# MPTcpExtRmSubflow 1 0.0
# Client ns stats
# TcpActiveOpens 3 0.0
# TcpInSegs 206 0.0
# TcpOutSegs 253 0.0
# TcpOutRsts 1 0.0
# TcpExtTW 2 0.0
# TcpExtDelayedACKs 49 0.0
# TcpExtTCPPureAcks 69 0.0
# TcpExtTCPOrigDataSent 132 0.0
# TcpExtTCPDelivered 135 0.0
# MPTcpExtMPCapableSYNTX 1 0.0
# MPTcpExtMPCapableSYNACKRX 1 0.0
# MPTcpExtMPJoinSynAckRx 2 0.0
# MPTcpExtDuplicateData 1 0.0
# MPTcpExtAddAddr 2 0.0
# MPTcpExtRmAddr 2 0.0
# MPTcpExtRcvWndShared 2 0.0
# Info: invert
# mptcp_info subflows=0:0 [ ok ]
# subflows_total 1:1 [ ok ]The counter TcpEstabResets is also incremented here in case of error, maybe a link?
The pm_nl_ctl failure is linked to this:
mptcp_info add_addr_signal=2:2 [ ok ]
+ ip netns exec ns1-660c230c-HNYZEv ./pm_nl_ctl rem token 420034353 id 10
+ set +x
+ ip netns exec ns1-660c230c-HNYZEv ./pm_nl_ctl dsf lip ::ffff:10.0.2.1 lport 10111 rip ::ffff:10.0.2.2 rport 47945 token 420034353
+ set +x
+ ip netns exec ns1-660c230c-HNYZEv ./pm_nl_ctl rem token 420034353 id 20
+ set +x
+ ip netns exec ns1-660c230c-HNYZEv ./pm_nl_ctl dsf lip 10.0.3.1 lport 10111 rip ::ffff:10.0.3.2 rport 56869 token 420034353
netlink error -3 (No such process)
./pm_nl_ctl: bailing out due to netlink error[s]
+ set +x
rm [ ok ]
rmsf [fail] got 1 RM_SUBFLOW[s] expected 2So only the 4th command on the server side failed, because it didn't find the subflow for this connection (-3 error). Probably because the subflow has been reset?
If we look at the events, the subflow has been removed (MPTCP_EVENT_SUB_CLOSED = 11):
+ ip netns exec ns1-660c34a1-ebiad5 ./pm_nl_ctl dsf lip ::ffff:10.0.2.1 lport 10111 rip ::ffff:10.0.2.2 rport 50643 token 226034464
type:11,token:226034464,family:10,saddr6:::ffff:10.0.2.1,daddr6:::ffff:10.0.2.2,sport:10111,dport:50643,loc_id:10,rem_id:1,backup:0
(...)
+ ip netns exec ns1-660c34a1-ebiad5 ./pm_nl_ctl dsf lip 10.0.3.1 lport 10111 rip ::ffff:10.0.3.2 rport 36365 token 226034464
netlink error -3 (No such process)
./pm_nl_ctl: bailing out due to netlink error[s]
type:11,token:226034464,family:10,saddr6:::ffff:10.0.3.1,daddr6:::ffff:10.0.3.2,sport:10111,dport:36365,loc_id:20,rem_id:2,backup:0
nkambo from LKFT reported that
mptcp_connect.shsubtests are, e.g.From: https://qa-reports.linaro.org/lkft/linux-stable-rc-linux-6.1.y/build/v6.1.83-273-gaa2042702765/testrun/23261389/suite/kselftest-net-mptcp/test/net_mptcp_mptcp_connect_sh_-_mptcp_connect_peek_mode_saveAfterPeek_ns1_MPTCP__ns1_dead_beef_1_1_10049_MPTCP/log
Or
From: https://qa-reports.linaro.org/lkft/linux-stable-rc-linux-6.1.y/build/v6.1.83-273-gaa2042702765/testrun/23259658/suite/kselftest-net-mptcp/test/net_mptcp_mptcp_connect_sh_-_mptcp_connect_full_disconnect_ns1_MPTCP__ns1_dead_beef_1_1_20004_MPTCP/log
Looking at the history, the
mptcp_connect.shtest seems quite unstable: https://qa-reports.linaro.org/lkft/linux-stable-rc-linux-6.1.y/build/v6.1.83-273-gaa2042702765/testrun/23259658/suite/kselftest-net-mptcp/test/net_mptcp_mptcp_connect_sh/history/