On Mon, Feb 27, 2023 at 7:04 PM KP Singh <kpsingh@xxxxxxxxxx> wrote: > > > > > > > if (use_trace_rcu) > > > - call_rcu_tasks_trace(&selem->rcu, bpf_selem_free_rcu); > > > + call_rcu_tasks_trace(&selem->rcu, bpf_selem_free_tasks_trace_rcu); > > > else > > > - kfree_rcu(selem, rcu); > > > + call_rcu(&selem->rcu, bpf_selem_free_rcu); > > > > Instead of adding 'bool can_use_smap' to 'struct bpf_local_storage_elem', can it > > be a different rcu call back when smap->map.record is not NULL and only that new > > rcu call back can use smap? > > I have a use on this 8-byte hole when using bpf_mem_alloc in bpf_local_storage. I've decided it to apply it as-is to speeds things up. Kumar, please follow up addressing Kumar's and KP's suggestions.