On 09/08/2016 01:38 AM, Alex Bennée wrote: >> + } else if (r == EXCP_ATOMIC) { >> > + /* ??? When we begin running cpus in parallel, we should >> > + stop all cpus, clear parallel_cpus, and interpret a >> > + single insn with cpu_exec_step. In the meantime, >> > + we should never get here. */ >> > + abort(); > Pranith has been hitting this abort in the latest merged tree with MTTCG > but I'm a little unclear how it got here. So is the plan the MTTCG > thread function should do a step_atomic a-la user mode but we'll never > get here in the single threaded case? >
Yes, that's the plan. I guess I could have filled in that blank, but I see that I haven't even done that in the v3 patchset. r~