Re: [MPTCP][PATCH v3 mptcp-next 1/8] mptcp: MP_FAIL suboption sending

[Geliang Tang <geliangtang@gmail.com>]

Hi Paolo,

Thanks for your review!

Paolo Abeni <pabeni@redhat.com> 于2021年7月14日周三 下午4:45写道：
>
> On Mon, 2021-06-28 at 12:28 +0800, Geliang Tang wrote:
> > This patch added the MP_FAIL suboption sending support.
> >
> > Add a new flag named send_mp_fail in struct mptcp_subflow_context. If
> > this flag is set, send out MP_FAIL suboption.
> >
> > Add a new member fail_seq in struct mptcp_out_options to save the data
> > sequence number to put into the MP_FAIL suboption.
> >
> > Signed-off-by: Geliang Tang <geliangtang@gmail.com>
> > ---
> >  include/net/mptcp.h  |  1 +
> >  net/mptcp/options.c  | 54 ++++++++++++++++++++++++++++++++++++++++++--
> >  net/mptcp/protocol.h |  4 ++++
> >  3 files changed, 57 insertions(+), 2 deletions(-)
> >
> > diff --git a/include/net/mptcp.h b/include/net/mptcp.h
> > index cb580b06152f..f48d3b5a3fd4 100644
> > --- a/include/net/mptcp.h
> > +++ b/include/net/mptcp.h
> > @@ -72,6 +72,7 @@ struct mptcp_out_options {
> >       u32 nonce;
> >       u64 thmac;
> >       u32 token;
> > +     u64 fail_seq;
>
> Since we can't have both a valid mapping and MP_FAIL in the same
> packet, it would be better to avoid increasing the 'mptcp_out_options'
> size, e.g. re-using some ext_copy field, or unioning with some other
> field.

The RFC says an MP_FAIL option could be included in a RST or on the
subflow-level ACK. So I think we can't re-using the ext_copy field if it's
on the subflow-level ACK.

How about unioning this fail_seq field with ahmac field?

>
> mptcp_out_options has grown quite a bit and we should really look into
> shrinking it.
>
> >       u8 hmac[20];
> >       struct mptcp_ext ext_copy;
> >  #endif
> > diff --git a/net/mptcp/options.c b/net/mptcp/options.c
> > index b5850afea343..b78defe1aed9 100644
> > --- a/net/mptcp/options.c
> > +++ b/net/mptcp/options.c
> > @@ -771,6 +771,28 @@ static noinline void mptcp_established_options_rst(struct sock *sk, struct sk_bu
> >       opts->reset_reason = subflow->reset_reason;
> >  }
> >
> > +static bool mptcp_established_options_mp_fail(struct sock *sk,
> > +                                           unsigned int *size,
> > +                                           unsigned int remaining,
> > +                                           struct mptcp_out_options *opts)
> > +{
> > +     struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
> > +
> > +     if (!subflow->send_mp_fail)
> > +             return false;
> > +
> > +     if (remaining < TCPOLEN_MPTCP_FAIL)
> > +             return false;
> > +
> > +     *size = TCPOLEN_MPTCP_FAIL;
> > +     opts->suboptions |= OPTION_MPTCP_FAIL;
> > +     opts->fail_seq = subflow->fail_seq;
> > +
> > +     pr_debug("MP_FAIL fail_seq=%llu", opts->fail_seq);
> > +
> > +     return true;
> > +}
> > +
> >  bool mptcp_established_options(struct sock *sk, struct sk_buff *skb,
> >                              unsigned int *size, unsigned int remaining,
> >                              struct mptcp_out_options *opts)
> > @@ -787,15 +809,29 @@ bool mptcp_established_options(struct sock *sk, struct sk_buff *skb,
> >               return false;
> >
> >       if (unlikely(skb && TCP_SKB_CB(skb)->tcp_flags & TCPHDR_RST)) {
> > -             mptcp_established_options_rst(sk, skb, size, remaining, opts);
> > +             if (mptcp_established_options_mp_fail(sk, &opt_size, remaining, opts)) {
> > +                     *size += opt_size;
> > +                     remaining -= opt_size;
> > +             }
> > +             mptcp_established_options_rst(sk, skb, &opt_size, remaining, opts);
> > +             *size += opt_size;
> > +             remaining -= opt_size;
> >               return true;
> >       }
> >
> >       snd_data_fin = mptcp_data_fin_enabled(msk);
> >       if (mptcp_established_options_mp(sk, skb, snd_data_fin, &opt_size, remaining, opts))
> >               ret = true;
> > -     else if (mptcp_established_options_dss(sk, skb, snd_data_fin, &opt_size, remaining, opts))
> > +     else if (mptcp_established_options_dss(sk, skb, snd_data_fin, &opt_size, remaining, opts)) {
> >               ret = true;
> > +             if (opts->ext_copy.use_ack) {
>
> Is the extra check on 'opts->ext_copy.use_ack' necessary? can we just
> look for mp_fail?
>

I added this check since the RFC says an MP_FAIL option could be included
on the subflow-level ACK:

'''
In this case, if a receiver identifies a checksum failure when there is
only one path, it will send back an MP_FAIL option on the subflow-level
ACK, referring to the data-level sequence number of the start of the
segment on which the checksum error was detected.
'''

Do I understand right?

Thanks,
-Geliang


>
> > +                     if (mptcp_established_options_mp_fail(sk, &opt_size, remaining, opts)) {
> > +                             *size += opt_size;
> > +                             remaining -= opt_size;
> > +                             ret = true;
> > +                     }
> > +             }
> > +     }
> >
> >       /* we reserved enough space for the above options, and exceeding the
> >        * TCP option space would be fatal
> > @@ -1334,6 +1370,20 @@ void mptcp_write_options(__be32 *ptr, const struct tcp_sock *tp,
> >                                     opts->backup, TCPOPT_NOP);
> >       }
> >
> > +     if (OPTION_MPTCP_FAIL & opts->suboptions) {
> > +             const struct sock *ssk = (const struct sock *)tp;
> > +             struct mptcp_subflow_context *subflow;
> > +
> > +             subflow = mptcp_subflow_ctx(ssk);
> > +             subflow->send_mp_fail = 0;
> > +
> > +             *ptr++ = mptcp_option(MPTCPOPT_MP_FAIL,
> > +                                   TCPOLEN_MPTCP_FAIL,
> > +                                   0, 0);
> > +             put_unaligned_be64(opts->fail_seq, ptr);
> > +             ptr += 2;
> > +     }
> > +
> >       if (OPTION_MPTCP_MPJ_SYN & opts->suboptions) {
> >               *ptr++ = mptcp_option(MPTCPOPT_MP_JOIN,
> >                                     TCPOLEN_MPTCP_MPJ_SYN,
> > diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
> > index 426ed80fe72f..007af5e4ba3d 100644
> > --- a/net/mptcp/protocol.h
> > +++ b/net/mptcp/protocol.h
> > @@ -26,6 +26,7 @@
> >  #define OPTION_MPTCP_FASTCLOSE       BIT(8)
> >  #define OPTION_MPTCP_PRIO    BIT(9)
> >  #define OPTION_MPTCP_RST     BIT(10)
> > +#define OPTION_MPTCP_FAIL    BIT(11)
> >
> >  /* MPTCP option subtypes */
> >  #define MPTCPOPT_MP_CAPABLE  0
> > @@ -67,6 +68,7 @@
> >  #define TCPOLEN_MPTCP_PRIO_ALIGN     4
> >  #define TCPOLEN_MPTCP_FASTCLOSE              12
> >  #define TCPOLEN_MPTCP_RST            4
> > +#define TCPOLEN_MPTCP_FAIL           12
> >
> >  #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM      (TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA)
> >
> > @@ -417,6 +419,7 @@ struct mptcp_subflow_context {
> >               mpc_map : 1,
> >               backup : 1,
> >               send_mp_prio : 1,
> > +             send_mp_fail : 1,
> >               rx_eof : 1,
> >               can_ack : 1,        /* only after processing the remote a key */
> >               disposable : 1;     /* ctx can be free at ulp release time */
> > @@ -431,6 +434,7 @@ struct mptcp_subflow_context {
> >       u8      reset_seen:1;
> >       u8      reset_transient:1;
> >       u8      reset_reason:4;
> > +     u64     fail_seq;
> >
> >       long    delegated_status;
> >       struct  list_head delegated_node;   /* link into delegated_action, protected by local BH */
>