This patch added a new member last_retrans_seq in mptcp_subflow_context
to track the last retransmitting sequence number.

Add a new helper named mptcp_is_data_contiguous() to check whether the
data is contiguous on a subflow.

When a bad checksum is detected and a single contiguous subflow is in use,
don't send RST + MP_FAIL, send data_ack + MP_FAIL instead.

Suggested-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Geliang Tang <geliangtang@gmail.com>
---
 net/mptcp/protocol.c |  6 ++++++
 net/mptcp/protocol.h |  8 ++++++++
 net/mptcp/subflow.c  | 12 ++++++------
 3 files changed, 20 insertions(+), 6 deletions(-)

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index f8ad049d6941..cf8cccfefb51 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -729,6 +729,9 @@ void mptcp_data_ready(struct sock *sk, struct sock *ssk)
 	if (unlikely(subflow->disposable))
 		return;
 
+	if (!subflow->last_retrans_seq || mptcp_is_data_contiguous(ssk))
+		subflow->last_retrans_seq = tcp_sk(ssk)->snd_una - 1;
+
 	ssk_rbuf = READ_ONCE(ssk->sk_rcvbuf);
 	sk_rbuf = READ_ONCE(sk->sk_rcvbuf);
 	if (unlikely(ssk_rbuf > sk_rbuf))
@@ -2415,6 +2418,7 @@ static void mptcp_check_fastclose(struct mptcp_sock *msk)
 static void __mptcp_retrans(struct sock *sk)
 {
 	struct mptcp_sock *msk = mptcp_sk(sk);
+	struct mptcp_subflow_context *subflow;
 	struct mptcp_sendmsg_info info = {};
 	struct mptcp_data_frag *dfrag;
 	size_t copied = 0;
@@ -2464,6 +2468,8 @@ static void __mptcp_retrans(struct sock *sk)
 		dfrag->already_sent = max(dfrag->already_sent, info.sent);
 		tcp_push(ssk, 0, info.mss_now, tcp_sk(ssk)->nonagle,
 			 info.size_goal);
+		subflow = mptcp_subflow_ctx(ssk);
+		subflow->last_retrans_seq = tcp_sk(ssk)->snd_nxt;
 	}
 
 	release_sock(ssk);
diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
index 7379ab580a7e..e090a9244f8b 100644
--- a/net/mptcp/protocol.h
+++ b/net/mptcp/protocol.h
@@ -416,6 +416,7 @@ struct mptcp_subflow_context {
 	u32	map_data_len;
 	__wsum	map_data_csum;
 	u32	map_csum_len;
+	u32	last_retrans_seq;
 	u32	request_mptcp : 1,  /* send MP_CAPABLE */
 		request_join : 1,   /* send MP_JOIN */
 		request_bkup : 1,
@@ -625,6 +626,13 @@ static inline bool mptcp_has_another_subflow(struct sock *ssk)
 	return false;
 }
 
+static inline bool mptcp_is_data_contiguous(struct sock *ssk)
+{
+	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
+
+	return before(subflow->last_retrans_seq, tcp_sk(ssk)->snd_una);
+}
+
 void __init mptcp_proto_init(void);
 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 int __init mptcp_proto_v6_init(void);
diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
index 6172f380dfb7..a8f46a48feb1 100644
--- a/net/mptcp/subflow.c
+++ b/net/mptcp/subflow.c
@@ -1166,15 +1166,15 @@ static bool subflow_check_data_avail(struct sock *ssk)
 fallback:
 	/* RFC 8684 section 3.7. */
 	if (subflow->send_mp_fail) {
-		if (mptcp_has_another_subflow(ssk)) {
+		if (mptcp_has_another_subflow(ssk) || !mptcp_is_data_contiguous(ssk)) {
+			ssk->sk_err = EBADMSG;
+			tcp_set_state(ssk, TCP_CLOSE);
+			subflow->reset_transient = 0;
+			subflow->reset_reason = MPTCP_RST_EMIDDLEBOX;
+			tcp_send_active_reset(ssk, GFP_ATOMIC);
 			while ((skb = skb_peek(&ssk->sk_receive_queue)))
 				sk_eat_skb(ssk, skb);
 		}
-		ssk->sk_err = EBADMSG;
-		tcp_set_state(ssk, TCP_CLOSE);
-		subflow->reset_transient = 0;
-		subflow->reset_reason = MPTCP_RST_EMIDDLEBOX;
-		tcp_send_active_reset(ssk, GFP_ATOMIC);
 		WRITE_ONCE(subflow->data_avail, 0);
 		return true;
 	}
-- 
2.31.1

On Sun, 26 Sep 2021, Geliang Tang wrote:

> This patch added a new member last_retrans_seq in mptcp_subflow_context
> to track the last retransmitting sequence number.
>
> Add a new helper named mptcp_is_data_contiguous() to check whether the
> data is contiguous on a subflow.
>
> When a bad checksum is detected and a single contiguous subflow is in use,
> don't send RST + MP_FAIL, send data_ack + MP_FAIL instead.
>
> Suggested-by: Paolo Abeni <pabeni@redhat.com>
> Signed-off-by: Geliang Tang <geliangtang@gmail.com>

Hi Geliang -

I know there's been a lot of churn in this part of the patch series, 
hopefully we can simplify a little to allow us to merge the first step of 
infinite mapping send/recv.

In v3 and v4 reviews, I suggested tracking the subflow sequence number 
like you do in this patch. After the discussion in the 2021-09-23 meeting, 
proper tracking of contiguous data appeared even more complicated, so 
Paolo had a suggestion to allow fallback in one simple-to-check case: Only 
do infinite mapping fallback if there is a single subflow AND there have 
been no retransmissions AND there have never been any MP_JOINs.

That can be tracked with a "allow_infinite_fallback" bool in mptcp_sock, 
which gets initialized to 'true' when the connection begins and is set to 
'false' on any retransmit or successful MP_JOIN.

Paolo, is that an accurate summary?


It's important that this type of fallback only happens when it is 
guaranteed that the stream is fully in-sequence, otherwise the 
application's data gets corrupted. The above suggestion lets us merge 
working infinite mapping code, and then figure out the more complicated 
"fallback after retransmissions or join" logic in a separate patch series.


Thanks,

Mat



> ---
> net/mptcp/protocol.c |  6 ++++++
> net/mptcp/protocol.h |  8 ++++++++
> net/mptcp/subflow.c  | 12 ++++++------
> 3 files changed, 20 insertions(+), 6 deletions(-)
>
> diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
> index f8ad049d6941..cf8cccfefb51 100644
> --- a/net/mptcp/protocol.c
> +++ b/net/mptcp/protocol.c
> @@ -729,6 +729,9 @@ void mptcp_data_ready(struct sock *sk, struct sock *ssk)
> 	if (unlikely(subflow->disposable))
> 		return;
>
> +	if (!subflow->last_retrans_seq || mptcp_is_data_contiguous(ssk))
> +		subflow->last_retrans_seq = tcp_sk(ssk)->snd_una - 1;
> +
> 	ssk_rbuf = READ_ONCE(ssk->sk_rcvbuf);
> 	sk_rbuf = READ_ONCE(sk->sk_rcvbuf);
> 	if (unlikely(ssk_rbuf > sk_rbuf))
> @@ -2415,6 +2418,7 @@ static void mptcp_check_fastclose(struct mptcp_sock *msk)
> static void __mptcp_retrans(struct sock *sk)
> {
> 	struct mptcp_sock *msk = mptcp_sk(sk);
> +	struct mptcp_subflow_context *subflow;
> 	struct mptcp_sendmsg_info info = {};
> 	struct mptcp_data_frag *dfrag;
> 	size_t copied = 0;
> @@ -2464,6 +2468,8 @@ static void __mptcp_retrans(struct sock *sk)
> 		dfrag->already_sent = max(dfrag->already_sent, info.sent);
> 		tcp_push(ssk, 0, info.mss_now, tcp_sk(ssk)->nonagle,
> 			 info.size_goal);
> +		subflow = mptcp_subflow_ctx(ssk);
> +		subflow->last_retrans_seq = tcp_sk(ssk)->snd_nxt;
> 	}
>
> 	release_sock(ssk);
> diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
> index 7379ab580a7e..e090a9244f8b 100644
> --- a/net/mptcp/protocol.h
> +++ b/net/mptcp/protocol.h
> @@ -416,6 +416,7 @@ struct mptcp_subflow_context {
> 	u32	map_data_len;
> 	__wsum	map_data_csum;
> 	u32	map_csum_len;
> +	u32	last_retrans_seq;
> 	u32	request_mptcp : 1,  /* send MP_CAPABLE */
> 		request_join : 1,   /* send MP_JOIN */
> 		request_bkup : 1,
> @@ -625,6 +626,13 @@ static inline bool mptcp_has_another_subflow(struct sock *ssk)
> 	return false;
> }
>
> +static inline bool mptcp_is_data_contiguous(struct sock *ssk)
> +{
> +	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
> +
> +	return before(subflow->last_retrans_seq, tcp_sk(ssk)->snd_una);
> +}
> +
> void __init mptcp_proto_init(void);
> #if IS_ENABLED(CONFIG_MPTCP_IPV6)
> int __init mptcp_proto_v6_init(void);
> diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
> index 6172f380dfb7..a8f46a48feb1 100644
> --- a/net/mptcp/subflow.c
> +++ b/net/mptcp/subflow.c
> @@ -1166,15 +1166,15 @@ static bool subflow_check_data_avail(struct sock *ssk)
> fallback:
> 	/* RFC 8684 section 3.7. */
> 	if (subflow->send_mp_fail) {
> -		if (mptcp_has_another_subflow(ssk)) {
> +		if (mptcp_has_another_subflow(ssk) || !mptcp_is_data_contiguous(ssk)) {
> +			ssk->sk_err = EBADMSG;
> +			tcp_set_state(ssk, TCP_CLOSE);
> +			subflow->reset_transient = 0;
> +			subflow->reset_reason = MPTCP_RST_EMIDDLEBOX;
> +			tcp_send_active_reset(ssk, GFP_ATOMIC);
> 			while ((skb = skb_peek(&ssk->sk_receive_queue)))
> 				sk_eat_skb(ssk, skb);
> 		}
> -		ssk->sk_err = EBADMSG;
> -		tcp_set_state(ssk, TCP_CLOSE);
> -		subflow->reset_transient = 0;
> -		subflow->reset_reason = MPTCP_RST_EMIDDLEBOX;
> -		tcp_send_active_reset(ssk, GFP_ATOMIC);
> 		WRITE_ONCE(subflow->data_avail, 0);
> 		return true;
> 	}
> -- 
> 2.31.1
>
>
>

--
Mat Martineau
Intel

Mat Martineau <mathew.j.martineau@linux.intel.com> 于2021年9月28日周二 上午8:52写道：
>
> On Sun, 26 Sep 2021, Geliang Tang wrote:
>
> > This patch added a new member last_retrans_seq in mptcp_subflow_context
> > to track the last retransmitting sequence number.
> >
> > Add a new helper named mptcp_is_data_contiguous() to check whether the
> > data is contiguous on a subflow.
> >
> > When a bad checksum is detected and a single contiguous subflow is in use,
> > don't send RST + MP_FAIL, send data_ack + MP_FAIL instead.
> >
> > Suggested-by: Paolo Abeni <pabeni@redhat.com>
> > Signed-off-by: Geliang Tang <geliangtang@gmail.com>
>
> Hi Geliang -
>
> I know there's been a lot of churn in this part of the patch series,
> hopefully we can simplify a little to allow us to merge the first step of
> infinite mapping send/recv.

Hi Mat, Paolo, Matt,

I prefer to continue doing more iterations of this series instead of
merging part of them first. There's still some work to do in it, but it's
nearly finished. I'll send a v6 recently.

And the infinite mapping support is the last issue assigned to me on
GitHub. So I'm thinking about what should I do next. Could you give me
some suggestions?

Thanks,
-Geliang

>
> In v3 and v4 reviews, I suggested tracking the subflow sequence number
> like you do in this patch. After the discussion in the 2021-09-23 meeting,
> proper tracking of contiguous data appeared even more complicated, so
> Paolo had a suggestion to allow fallback in one simple-to-check case: Only
> do infinite mapping fallback if there is a single subflow AND there have
> been no retransmissions AND there have never been any MP_JOINs.
>
> That can be tracked with a "allow_infinite_fallback" bool in mptcp_sock,
> which gets initialized to 'true' when the connection begins and is set to
> 'false' on any retransmit or successful MP_JOIN.
>
> Paolo, is that an accurate summary?
>
>
> It's important that this type of fallback only happens when it is
> guaranteed that the stream is fully in-sequence, otherwise the
> application's data gets corrupted. The above suggestion lets us merge
> working infinite mapping code, and then figure out the more complicated
> "fallback after retransmissions or join" logic in a separate patch series.
>
>
> Thanks,
>
> Mat
>
>
>
> > ---
> > net/mptcp/protocol.c |  6 ++++++
> > net/mptcp/protocol.h |  8 ++++++++
> > net/mptcp/subflow.c  | 12 ++++++------
> > 3 files changed, 20 insertions(+), 6 deletions(-)
> >
> > diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
> > index f8ad049d6941..cf8cccfefb51 100644
> > --- a/net/mptcp/protocol.c
> > +++ b/net/mptcp/protocol.c
> > @@ -729,6 +729,9 @@ void mptcp_data_ready(struct sock *sk, struct sock *ssk)
> >       if (unlikely(subflow->disposable))
> >               return;
> >
> > +     if (!subflow->last_retrans_seq || mptcp_is_data_contiguous(ssk))
> > +             subflow->last_retrans_seq = tcp_sk(ssk)->snd_una - 1;
> > +
> >       ssk_rbuf = READ_ONCE(ssk->sk_rcvbuf);
> >       sk_rbuf = READ_ONCE(sk->sk_rcvbuf);
> >       if (unlikely(ssk_rbuf > sk_rbuf))
> > @@ -2415,6 +2418,7 @@ static void mptcp_check_fastclose(struct mptcp_sock *msk)
> > static void __mptcp_retrans(struct sock *sk)
> > {
> >       struct mptcp_sock *msk = mptcp_sk(sk);
> > +     struct mptcp_subflow_context *subflow;
> >       struct mptcp_sendmsg_info info = {};
> >       struct mptcp_data_frag *dfrag;
> >       size_t copied = 0;
> > @@ -2464,6 +2468,8 @@ static void __mptcp_retrans(struct sock *sk)
> >               dfrag->already_sent = max(dfrag->already_sent, info.sent);
> >               tcp_push(ssk, 0, info.mss_now, tcp_sk(ssk)->nonagle,
> >                        info.size_goal);
> > +             subflow = mptcp_subflow_ctx(ssk);
> > +             subflow->last_retrans_seq = tcp_sk(ssk)->snd_nxt;
> >       }
> >
> >       release_sock(ssk);
> > diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
> > index 7379ab580a7e..e090a9244f8b 100644
> > --- a/net/mptcp/protocol.h
> > +++ b/net/mptcp/protocol.h
> > @@ -416,6 +416,7 @@ struct mptcp_subflow_context {
> >       u32     map_data_len;
> >       __wsum  map_data_csum;
> >       u32     map_csum_len;
> > +     u32     last_retrans_seq;
> >       u32     request_mptcp : 1,  /* send MP_CAPABLE */
> >               request_join : 1,   /* send MP_JOIN */
> >               request_bkup : 1,
> > @@ -625,6 +626,13 @@ static inline bool mptcp_has_another_subflow(struct sock *ssk)
> >       return false;
> > }
> >
> > +static inline bool mptcp_is_data_contiguous(struct sock *ssk)
> > +{
> > +     struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
> > +
> > +     return before(subflow->last_retrans_seq, tcp_sk(ssk)->snd_una);
> > +}
> > +
> > void __init mptcp_proto_init(void);
> > #if IS_ENABLED(CONFIG_MPTCP_IPV6)
> > int __init mptcp_proto_v6_init(void);
> > diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
> > index 6172f380dfb7..a8f46a48feb1 100644
> > --- a/net/mptcp/subflow.c
> > +++ b/net/mptcp/subflow.c
> > @@ -1166,15 +1166,15 @@ static bool subflow_check_data_avail(struct sock *ssk)
> > fallback:
> >       /* RFC 8684 section 3.7. */
> >       if (subflow->send_mp_fail) {
> > -             if (mptcp_has_another_subflow(ssk)) {
> > +             if (mptcp_has_another_subflow(ssk) || !mptcp_is_data_contiguous(ssk)) {
> > +                     ssk->sk_err = EBADMSG;
> > +                     tcp_set_state(ssk, TCP_CLOSE);
> > +                     subflow->reset_transient = 0;
> > +                     subflow->reset_reason = MPTCP_RST_EMIDDLEBOX;
> > +                     tcp_send_active_reset(ssk, GFP_ATOMIC);
> >                       while ((skb = skb_peek(&ssk->sk_receive_queue)))
> >                               sk_eat_skb(ssk, skb);
> >               }
> > -             ssk->sk_err = EBADMSG;
> > -             tcp_set_state(ssk, TCP_CLOSE);
> > -             subflow->reset_transient = 0;
> > -             subflow->reset_reason = MPTCP_RST_EMIDDLEBOX;
> > -             tcp_send_active_reset(ssk, GFP_ATOMIC);
> >               WRITE_ONCE(subflow->data_avail, 0);
> >               return true;
> >       }
> > --
> > 2.31.1
> >
> >
> >
>
> --
> Mat Martineau
> Intel
>