On Fri, Dec 8, 2023 at 6:28 PM Alexei Starovoitov <alexei.starovoitov@xxxxxxxxx> wrote: > > On Fri, Dec 8, 2023 at 6:15 PM Andrii Nakryiko > <andrii.nakryiko@xxxxxxxxx> wrote: > > > > See above, MISC/ZERO is fine as is due to check_stack_read_fixed_off() > > not setting STACK_ACCESS bit, but I can also send a version that > > unconditionally sets INSNS_F_STACK_ACCESS in > > check_stack_write_fixed_off(). > > but it will significantly increase jmp_history and make > backtracking slower? Good point about jmp_history length, so I guess we can leave the patch as is. I don't think it will make backtracking slower, though, the algorithm will process the same amount of instructions regardless of jmp_history length.
