Re: [PATCH mptcp-next v3 0/5] add mptcp_subflow bpf_iter

[Geliang Tang <geliang@kernel.org>]

Hi Matt,

On Tue, 2024-09-10 at 09:56 +0200, Matthieu Baerts wrote:
> Hi Geliang,
> 
> On 10/09/2024 08:30, Geliang Tang wrote:
> > On Tue, 2024-09-10 at 06:02 +0000, MPTCP CI wrote:
> > > Hi Geliang,
> > > 
> > > Thank you for your modifications, that's great!
> > > 
> > > But sadly, our CI spotted some issues with it when trying to
> > > build
> > > it.
> > > 
> > > You can find more details there:
> > > 
> > >  
> > > https://github.com/multipath-tcp/mptcp_net-next/actions/runs/10786552828
> > > 
> > > Status: failure
> > 
> > These errors reported by CI should be ignored:
> > 
> > Unable to compile mptcp source code with make C=1 net/mptcp/bpf.o:
> > net/mptcp/bpf.c:215:17: warning: symbol
> > 'bpf_iter_mptcp_subflow_new'
> > was not declared. Should it be static?
> > Unable to compile mptcp source code with make C=1 net/mptcp/bpf.o:
> > net/mptcp/bpf.c:228:42: warning: symbol
> > 'bpf_iter_mptcp_subflow_next'
> > was not declared. Should it be static?
> > Unable to compile mptcp source code with make C=1 net/mptcp/bpf.o:
> > net/mptcp/bpf.c:243:18: warning: symbol
> > 'bpf_iter_mptcp_subflow_destroy' was not declared. Should it be
> > static?
> > Unable to compile mptcp source code with make C=1 net/mptcp/bpf.o:
> > net/mptcp/bpf.c:247:42: warning: symbol
> > 'bpf_mptcp_subflow_ctx_by_pos'
> > was not declared. Should it be static?
> 
> Can we not squash patch 1/5 and 3/5 together?
> 
> It is strange to add the helpers but not using them, then register
> them
> in another patch.

It's a bit strange indeed. It's better to register them in
common_btf_ids in kernel/bpf/helpers.c, not in
bpf_mptcp_sched_kfunc_ids in net/mptcp/bpf.c.

I'll send two squash-to patches to fix this.

Thanks,
-Geliang

> 
> Or did you split them to ease the review on BPF side?
> 
> Cheers,
> Matt