Hi All, Thanks for joining the meeting this week. Here are some status updates from various folks.
William: Discussed issue he is having with the verifier/llvm complex program becomes stack limited as a workaround, tried using volatile modifier for some stack variables but this causes the verifier to hit the 64k simulated instruction complexity limit reason: no heuristics to prune states involving on-stack variables suggested workaround: use per-cpu array to hold state Brenden: Saw issue with xdp in virtio warn on once in drivers/pci/msi.c:1261 pci_irq_vector+0xd4/0xe0 will send description to netdev list ctf (tracing and types) topics: Suchakra (tracing) - started with babeltrace python bindings - discussed optional compile-time dependency on libbabeltrace - we agreed for now to work on a separate bcc branch for integration testing Alexei: libraries exist as part of pahole (maintained by Arnaldo): payhole is tool for analyzing holes in c datastructures supports dwarf+ctf parser and emitter may need to write emitter from scratch due to licensing, but not too hard bitfield representation may need some work api: extra section at map create time ability to dump as well via syscall new name to avoid ambiguity with tracing-ctf? Andy: arm64 compiler issues inline assembly parser in llvm must-have for tracing to get accurate header definitions open PR already exists for hacking it in BCC to ifdef out inline assembly (ugly) suggestion to respin patches to arm list to cleanup defines Sargun: new landlock patches a month ago, progress? some discussion, seems stalled (unprivileged requirement) alternative: seccomp syscall filtering at global level (with eBPF) avoids criu problem Attendees: William Tu Teng Suchakra Sharma Sargun Dhillon Michael Schubert Mauricio Vasquez Matteo Bertrone Martin Lau Marco Leogrande Lorenzo David Joe Stringer Jan-Erik Jakub Kicinski Gary Berger Flavio Crisciani Ferris Daniel Borkmann Brenden Blanco Brendan Gregg Andy Gospodarek Alexei Starvoitov _______________________________________________ iovisor-dev mailing list iovisor-dev@lists.iovisor.org https://lists.iovisor.org/mailman/listinfo/iovisor-dev
