Re: [PATCH mptcp-next v11 10/11] Squash to "selftests/bpf: Add bpf_rr scheduler"

[Geliang Tang <geliang.tang@suse.com>]

On Mon, Jul 03, 2023 at 05:38:39PM +0200, Paolo Abeni wrote:
> On Tue, 2023-06-27 at 09:06 +0800, Geliang Tang wrote:
> > Use sk_storage to store last_snd, instead of using msk->last_snd.
> > 
> > Signed-off-by: Geliang Tang <geliang.tang@suse.com>
> > ---
> >  tools/testing/selftests/bpf/bpf_tcp_helpers.h |  6 ++-
> >  .../selftests/bpf/progs/mptcp_bpf_rr.c        | 44 ++++++++++++++-----
> >  2 files changed, 39 insertions(+), 11 deletions(-)
> > 
> > diff --git a/tools/testing/selftests/bpf/bpf_tcp_helpers.h b/tools/testing/selftests/bpf/bpf_tcp_helpers.h
> > index b4b766c7a68f..945dd46c98c0 100644
> > --- a/tools/testing/selftests/bpf/bpf_tcp_helpers.h
> > +++ b/tools/testing/selftests/bpf/bpf_tcp_helpers.h
> > @@ -259,7 +259,6 @@ struct mptcp_sched_ops {
> >  struct mptcp_sock {
> >  	struct inet_connection_sock	sk;
> >  
> > -	struct sock	*last_snd;
> >  	__u32		token;
> >  	struct sock	*first;
> >  	char		ca_name[TCP_CA_NAME_MAX];
> > @@ -271,5 +270,10 @@ extern void mptcp_sched_data_set_contexts(const struct mptcp_sock *msk,
> >  					  struct mptcp_sched_data *data) __ksym;
> >  extern struct mptcp_subflow_context *
> >  mptcp_subflow_ctx_by_pos(const struct mptcp_sched_data *data, unsigned int pos) __ksym;
> > +static inline struct sock *
> > +mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
> > +{
> > +	return subflow->tcp_sock;
> > +}
> >  
> >  #endif
> > diff --git a/tools/testing/selftests/bpf/progs/mptcp_bpf_rr.c b/tools/testing/selftests/bpf/progs/mptcp_bpf_rr.c
> > index e101428e5906..21144e96ba56 100644
> > --- a/tools/testing/selftests/bpf/progs/mptcp_bpf_rr.c
> > +++ b/tools/testing/selftests/bpf/progs/mptcp_bpf_rr.c
> > @@ -6,33 +6,49 @@
> >  
> >  char _license[] SEC("license") = "GPL";
> >  
> > +struct mptcp_rr_storage {
> > +	struct sock *last_snd;
> > +};
> > +struct sock *last_snd;
> > +
> > +struct {
> > +	__uint(type, BPF_MAP_TYPE_SK_STORAGE);
> > +	__uint(map_flags, BPF_F_NO_PREALLOC);
> > +	__type(key, int);
> > +	__type(value, struct mptcp_rr_storage);
> > +} mptcp_rr_map SEC(".maps");
> > +
> >  SEC("struct_ops/mptcp_sched_rr_init")
> > -void BPF_PROG(mptcp_sched_rr_init, const struct mptcp_sock *msk)
> > +void BPF_PROG(mptcp_sched_rr_init, struct mptcp_sock *msk)
> >  {
> >  }
> >  
> >  SEC("struct_ops/mptcp_sched_rr_release")
> > -void BPF_PROG(mptcp_sched_rr_release, const struct mptcp_sock *msk)
> > +void BPF_PROG(mptcp_sched_rr_release, struct mptcp_sock *msk)
> >  {
> > +	bpf_sk_storage_delete(&mptcp_rr_map, msk);
> >  }
> >  
> > -void BPF_STRUCT_OPS(bpf_rr_data_init, const struct mptcp_sock *msk,
> > +void BPF_STRUCT_OPS(bpf_rr_data_init, struct mptcp_sock *msk,
> >  		    struct mptcp_sched_data *data)
> >  {
> >  	mptcp_sched_data_set_contexts(msk, data);
> >  }
> >  
> > -int BPF_STRUCT_OPS(bpf_rr_get_subflow, const struct mptcp_sock *msk,
> > -		   struct mptcp_sched_data *data)
> > +int BPF_STRUCT_OPS(bpf_rr_get_subflow, struct mptcp_sock *msk,
> > +		   const struct mptcp_sched_data *data)
> >  {
> > +	struct mptcp_subflow_context *subflow;
> > +	struct mptcp_rr_storage *ptr;
> >  	int nr = 0;
> >  
> > -	for (int i = 0; i < MPTCP_SUBFLOWS_MAX; i++) {
> > -		if (!msk->last_snd || !data->contexts[i])
> > +	for (int i = 0; i < data->subflows && i < MPTCP_SUBFLOWS_MAX; i++) {
> > +		subflow = mptcp_subflow_ctx_by_pos(data, i);
> > +		if (!last_snd || !subflow)
> 
> It looks like this is accessing a global variable ('last_snd'), which
> is probably not very safe.

Yes, it's should be a local variable. I just sent a patch to ML to fix
this:

https://patchwork.kernel.org/project/mptcp/patch/0b57bf87dc30e15782ba1fe2e925a7ce6ef0fc90.1688434547.git.geliang.tang@suse.com/

> 
> I think you could instead use inet_csk(msk)->icsk_ca_priv - currently
> unused.
> 
> As this is the only comment I have, no need to send a whole v12. I
> think we can merge the series as-is and then squash the needed change
> to replace the global variable and map used here with inet_csk(msk)-
> >icsk_ca_priv.

One of the purposes of this test is to demonstrate an example of using a
bpf map to store persistent schedulers data, so I prefer to use the bpf
map to implement it here.

Using inet_csk(msk)->icsk_ca_priv is much more complex. We need to add
the special write accesses for icsk_ca_priv.

Thanks,
-Geliang

> 
> Thanks!
> 
> Paolo
>