On Sat, Feb 08, 2025 at 08:31:27PM -1000, Tejun Heo wrote: > Hello, > > On Sat, Feb 08, 2025 at 10:19:38AM +0100, Andrea Righi wrote: > ... > > > This is contingent on scx_builtin_idle_per_node, right? It's confusing for > > > CPU -> node mapping function to return NUMA_NO_NODE depending on an ops > > > flag. Shouldn't this be a generic mapping function? > > > > The idea is that BPF schedulers can use this kfunc to determine the right > > idle cpumask to use, for example a typical usage could be: > > > > int node = scx_bpf_cpu_node(prev_cpu); > > s32 cpu = scx_bpf_pick_idle_cpu_in_node(p->cpus_ptr, node, SCX_PICK_IDLE_IN_NODE); > > > > Or: > > > > int node = scx_bpf_cpu_node(prev_cpu); > > const struct cpumask *idle_cpumask = scx_bpf_get_idle_cpumask_node(node); > > > > When SCX_OPS_BUILTIN_IDLE_PER_NODE is disabled, we need to point to the > > global idle cpumask, that is identified by NUMA_NO_NODE, so this is why we > > can return NUMA_NO_NODE fro scx_bpf_cpu_node(). > > > > Do you think we should make this more clear / document this better. Or do > > you think we should use a different API? > > I think this is too error-prone. It'd be really easy for users to assume > that scx_bpf_cpu_node() always returns the NUMA node for the given CPU which > can lead to really subtle surprises. Why even allow e.g. > scx_bpf_get_idle_cpumask_node() if IDLE_PER_NODE is not enabled? Ok, for the kfuncs I agree that we should just trigger an scx_ops_error() if any of the scx_*_node() are used when SCX_OPS_BUILTIN_IDLE_PER_NODE is disabled (will change this). About scx_cpu_node(), which is used internally, I think it's convenient to return NUMA_NO_NODE when idle-per-node is disabled, just to avoid repeating the same check for scx_builtin_idle_per_node everywhere, and NUMA_NO_NODE internally always means "use the global cpumask". Do you think this is still error-prone? Or should I try to refactor the code to get rid of this NUMA_NO_NODE == global cpumask logic? Thanks, -Andrea
