ironcore-dev / dpservice

DPDK based fast Dataplane / L3 router / SDN enabler, installable on compute nodes / SmartNICs
Apache License 2.0
7 stars 1 forks source link

Dpservice crashes due to early telemetry in multiport-eswitch mode #610

Open PlagueCZ opened 6 days ago

PlagueCZ commented 6 days ago

This happened in OSC environment when using multiport-eswitch and dpservice-exporter.

By asking for /ethdev/xstats during initialization, dpservice crashes.

Easily reproducible by running while true; do echo "/ethdev/xstats,0" | dpdk-telemetry.py; sleep 1; done and then starting dpservice.

Stack trace is showing corrupted stack trace on the telemetry thread, so I would think it's some sort of stack overflow. But there are no debug messages from telemetry to know more.

byteocean commented 10 hours ago

Steps to reproduce using standard tools:

  1. configure the machine using dpservice's prepare script with --multiport-eswitch --pf1-proxy;
  2. start testpmd sudo ./dpdk-testpmd -a 3b:00.0,dv_flow_en=2,representor=pf[0-1]vf0 -- -i --rxq=1 --txq=1 --disable-device-start;
  3. running while true; do echo "/ethdev/xstats,0" | dpdk-telemetry.py; sleep 1; done leads to the crashing of testpmd