Open n0toose opened 6 days ago
(I had set the flag to 1
on a previous operating system, forgot to do so again. I assumed that my environment must've broken things somehow - but not in the way that I expected. Perhaps this could explain why the error wasn't caught earlier.)
Possibly related:
If I'm not wrong, gdb/pwndbg "thinks" that a bunch of barrier.wait()
's have piled up and are keeping the program from exiting.
A working (borderline nonsensical) theory is that the kernel tries to inform Uhyve that it should shut down, but it doesn't (why?/how? probably wrong), therefore, it halts indefinitely:
Therefore, the Vcpu never halts, the VM never exits, and Uhyve won't get stopped.
This also applies to the WIP v2 version of uhyve-interface.
Running Uhyve directly (this does not happen with
cargo run
) seems to cause the process to hang indefinitely without a (graceful) exit:However, running...
seems to work just fine. Using
cargo run
also works.Affects 0222ff73dcf46c44351c34d885522c82b7c5c509, have not managed to bisect where this has seemingly stopped happening yet. Host OS is
Ubuntu 24.04.1 LTS x86_64
with6.8.0-44-generic
.