On 3/17/25 09:23, Paolo Abeni wrote:
On 3/16/25 11:45 PM, Michal Luczaj wrote:
Racing signal-interrupted connect() and sockmap update may result in an unconnected (and missing vsock transport) socket in a sockmap.
Test spends 2 seconds attempting to reach WARN_ON_ONCE().
connect / state = SS_CONNECTED / sock_map_update_elem if signal_pending state = SS_UNCONNECTED
connect transport = NULL vsock_bpf_recvmsg WARN_ON_ONCE(!vsk->transport)
Signed-off-by: Michal Luczaj mhal@rbox.co
This is apparently causing some bpf self-test failure. (Timeout? the self-test failure output is not clear to me.)
Could you please have a look?
Sending signal to the whole process group probably isn't the best idea. Not sure how the previous version passed though. Sorry, v4 incoming.
Michal