Re: [PATCH bpf-next 1/2] bpf: Add ifindex to bpf_sk_lookup

Alexei Starovoitov wrote:

On Fri, Oct 15, 2021 at 4:24 AM Mark Pashmfouroush
[off-list ref] wrote:

quoted

diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
index 6fc59d61937a..9bd3e8b8a659 100644
--- a/include/uapi/linux/bpf.h
+++ b/include/uapi/linux/bpf.h

@@ -6262,6 +6262,7 @@ struct bpf_sk_lookup {
        __u32 local_ip4;        /* Network byte order */
        __u32 local_ip6[4];     /* Network byte order */
        __u32 local_port;       /* Host byte order */
+       __u32 ifindex;          /* Maps to skb->dev->ifindex */

Is the comment accurate?
The bpf_sk_lookup_kern ifindex is populated with inet_iif(skb).
Which is skb->skb_iif at this point (I think).
skb->dev->ifindex would typically mean destination or egress ifindex.
In __sk_buff we have 'ifindex' and 'ingress_ifindex' to differentiate them.
If it's really dev->ifindex than keeping 'ifindex' name here would be correct,
but looking at how it's populated in inet/udp_lookup makes me wonder
whether it should be named 'ingress_ifindex' instead and comment clarified.

If/when you resubmit please trim cc list to a minimum.

At least in the tcp cases its coming from inet_iif which is either
the rtable or skb->skb_iif. Agree would be nice to fixup the comment.

Thanks.