MPTCP deadlocks
Subsystems (2)
Jakub Kicinski
maintainerHi!
Not sure if its the new machines or some of the recent work in MPTCP but we hit a deadlock in the tests a couple of times:
stderr of join.sh:
https://netdev-ctrl.bots.linux.dev/logs/vmksft/mptcp-dbg/results/412720/1-mptcp-join-sh/stderr
decoded:
https://netdev-ctrl.bots.linux.dev/logs/vmksft/mptcp-dbg/results/412720/vm-crash-thr0-0
Hi Jakub,
Quoted 2 lines · Jakub Kicinski
Not sure if its the new machines or some of the recent work in MPTCP but we hit a deadlock in the tests a couple of times:
Thank you for the message, I didn't notice that, and I didn't see this issue in our CI. So maybe due to both the new machines and the recent work.
I will check if one of my syzkaller instances didn't find the same bug, hopefully with a reproducer.
Cheers,
Matt
Signature · 1 line
Sponsored by the NGI0 Core fund.
Paolo Abeni
maintainerQuoted 9 lines · Matthieu Baerts
Quoted 2 lines · Jakub Kicinski
Not sure if its the new machines or some of the recent work in MPTCP but we hit a deadlock in the tests a couple of times:
Thank you for the message, I didn't notice that, and I didn't see this issue in our CI. So maybe due to both the new machines and the recent work.
I will check if one of my syzkaller instances didn't find the same bug, hopefully with a reproducer.
caused by:
ommit f8a1d9b18c5efc76784f5a326e905f641f839894
Author: Paolo Abeni <[email protected]> Date: Mon Jul 14 18:41:44 2025 +0200
mptcp: make fallback action and fallback decision atomic
I think it's just quite unlikely to hit the relevant code path with the suitable status to trigger the bug.
@Matttbe: if it's ok I'll send a patch directly to net
/P