On Wed, Oct 23, 2024 at 10:37 AM Eduard Zingerman <eddyz87@xxxxxxxxx> wrote: > > On Wed, 2024-10-23 at 10:33 -0700, Andrii Nakryiko wrote: > > [...] > > > Using two u64s to describe stack slot mask is really-really > > inconvenient. > > Yes > > > and increases memory usage by quite a lot. Given we intend to have > > insn_history for each instruction soon, I'd keep stack size at max > > of 512 bytes, even with bpf_fastcall. > > By 8mb for 1M instructions program. > > > Is it possible? > > If we drop this +40 bytes slack space everything else should work as > expected. I'd be OK with that.
