On Thu, Sep 13, 2018 at 4:57 PM Alexei Starovoitov
<alexei.starovoi...@gmail.com> wrote:
>
> On Thu, Sep 13, 2018 at 04:51:49PM -0400, Willem de Bruijn wrote:
> > On Thu, Sep 13, 2018 at 3:45 PM Alexei Starovoitov
> > <alexei.starovoi...@gmail.com> wrote:
> > >
> > > On Thu, Sep 13, 2018 at 10:45:53AM -0700, Petar Penkov wrote:
> > > > From: Petar Penkov <ppen...@google.com>
> > > >
> > > > Adds a hook for programs of type BPF_PROG_TYPE_FLOW_DISSECTOR and
> > > > attach type BPF_FLOW_DISSECTOR that is executed in the flow dissector
> > > > path. The BPF program is per-network namespace.
> > > >
> > > > Signed-off-by: Petar Penkov <ppen...@google.com>
> > > > Signed-off-by: Willem de Bruijn <will...@google.com>
> > > ...
> > > > @@ -2333,6 +2335,7 @@ struct __sk_buff {
> > > > /* ... here. */
> > > >
> > > > __u32 data_meta;
> > > > + struct bpf_flow_keys *flow_keys;
> > >
> > > the bpf prog form patch 4 looks much better now. Thanks!
> > >
> > > > };
> > > >
> > > > struct bpf_tunnel_key {
> > > > @@ -2778,4 +2781,27 @@ enum bpf_task_fd_type {
> > > > BPF_FD_TYPE_URETPROBE, /* filename + offset */
> > > > };
> > > >
> > > > +struct bpf_flow_keys {
> > > > + __u16 nhoff;
> > > > + __u16 thoff;
> > > > + __u16 addr_proto; /* ETH_P_* of valid addrs
> > > > */
> > > > + __u8 is_frag;
> > > > + __u8 is_first_frag;
> > > > + __u8 is_encap;
> > > > + __be16 n_proto;
> > > > + __u8 ip_proto;
> > > > + union {
> > > > + struct {
> > > > + __be32 ipv4_src;
> > > > + __be32 ipv4_dst;
> > > > + };
> > > > + struct {
> > > > + __u32 ipv6_src[4]; /* in6_addr; network
> > > > order */
> > > > + __u32 ipv6_dst[4]; /* in6_addr; network
> > > > order */
> > > > + };
> > > > + };
> > > > + __be16 sport;
> > > > + __be16 dport;
> > > > +};
> > >
> > > can you please pack it?
> > > struct bpf_flow_keys {
> > > __u16 nhoff; /* 0 2 */
> > > __u16 thoff; /* 2 2 */
> > > __u16 addr_proto; /* 4 2 */
> > > __u8 is_frag; /* 6 1 */
> > > __u8 is_first_frag; /* 7 1 */
> > > __u8 is_encap; /* 8 1 */
> > >
> > > /* XXX 1 byte hole, try to pack */
> > >
> > > __be16 n_proto; /* 10 2 */
> > > __u8 ip_proto; /* 12 1 */
> > >
> > > /* XXX 3 bytes hole, try to pack */
> > >
> > > union {
> > >
> > > also is_frag and other fields are not used by the kernel and
> > > only used by the prog to pass data between tail_calls ?
> >
> > No, these are mapped directly onto fields in struct flow_keys
> > on return from the BPF program in __skb_flow_bpf_to_target.
> > For is_frag, for instance:
> >
> > if (flow_keys->is_frag)
> > key_control->flags |= FLOW_DIS_IS_FRAGMENT;
>
> right. my search-fu failed me. only packing is needed then.
>
> > This is true for all fields in the struct except nhoff.
>
> > > In such case reserve some space in bpf_flow_keys similar to skb->cb
> > > so it can contain any fields and accommodate for inevitable changes
> > > to bpf flow dissector prog in the future.
> >
> > Do you mean a second scratch space akin to cb[], just a few
> > reserved padding bytes?
>
> it looks to me it's possible to rearrange the fields to avoid all holes,
> so no extra padding bytes necessary.
Absolutely. We forgot to run pahole earlier. Updating now.
> > We have given some thought to forward compatibility. The existing
> > fields cannot be changed, but it should be fine if we need to expand
> > the struct later.
>
> let's keep cb-like idea for later. It seems to me we can add it to
> the end of bpf_flow_keys any time later.
