On Tue, Oct 08, 2019 at 12:45:47PM -0700, Andrii Nakryiko wrote:
> Maps that are read-only both from BPF program side and user space side
> have their contents constant, so verifier can track referenced values
> precisely and use that knowledge for dead code elimination, branch
> pruning, etc. This patch teaches BPF verifier how to do this.
>
> Signed-off-by: Andrii Nakryiko <andr...@fb.com>
> ---
> kernel/bpf/verifier.c | 58 +++++++++++++++++++++++++++++++++++++++++--
> 1 file changed, 56 insertions(+), 2 deletions(-)
>
> diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
> index ffc3e53f5300..1e4e4bd64ca5 100644
> --- a/kernel/bpf/verifier.c
> +++ b/kernel/bpf/verifier.c
> @@ -2739,6 +2739,42 @@ static void coerce_reg_to_size(struct bpf_reg_state
> *reg, int size)
> reg->smax_value = reg->umax_value;
> }
>
> +static bool bpf_map_is_rdonly(const struct bpf_map *map)
> +{
> + return (map->map_flags & BPF_F_RDONLY_PROG) &&
> + ((map->map_flags & BPF_F_RDONLY) || map->frozen);
This is definitely buggy. Testing for 'map->map_flags & BPF_F_RDONLY'
to assume it's RO from user space side is not correct as it's just
related to the current fd, but not the map itself. So the second part
definitely /must/ only be: && map->frozen
Thanks,
Daniel
