On 2/25/21 6:27 PM, Cong Wang wrote:
On Wed, Feb 24, 2021 at 11:33 PM Yonghong Song <yhs@xxxxxx> wrote:diff --git a/kernel/bpf/bpf_iter.c b/kernel/bpf/bpf_iter.c index a0d9eade9c80..931870f9cf56 100644 --- a/kernel/bpf/bpf_iter.c +++ b/kernel/bpf/bpf_iter.c @@ -675,3 +675,19 @@ int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx) */ return ret == 0 ? 0 : -EAGAIN; } + +BPF_CALL_4(bpf_for_each_map_elem, struct bpf_map *, map, void *, callback_fn, + void *, callback_ctx, u64, flags) +{ + return map->ops->map_for_each_callback(map, callback_fn, callback_ctx, flags); +}A quick question: is ->map_for_each_callback() now mandatory for every map? I do not see you check for NULL here. Or you check map types somewhere I miss?
At the call site of bpf_for_each_map_elem(), verifier knows the map and ensure check map->ops->map_for_each_callback() is not null. Otherwise,
it will reject the program. So we are fine here.
At least some maps do not support iteration, for example, queue/stack. If you can document supported maps in bpf_for_each_map_elem() doc, it would be very nice.
Yes, will add more info in uapi/linux/bpf.h. Andrii suggested the same.
Thanks for working on this!
You are welcome.
