* [PATCH net] mptcp: always include dack if possible.
@ 2020-03-03 17:22 Paolo Abeni
2020-03-03 18:58 ` Mat Martineau
0 siblings, 1 reply; 3+ messages in thread
From: Paolo Abeni @ 2020-03-03 17:22 UTC (permalink / raw
To: netdev; +Cc: Mat Martineau, Matthieu Baerts, David S. Miller, Christoph Paasch
Currently passive MPTCP socket can skip including the DACK
option - if the peer sends data before accept() completes.
The above happens because the msk 'can_ack' flag is set
only after the accept() call.
Such missing DACK option may cause - as per RFC spec -
unwanted fallback to TCP.
This change addresses the issue using the key material
available in the current subflow, if any, to create a suitable
dack option when msk ack seq is not yet available.
Fixes: d22f4988ffec ("mptcp: process MP_CAPABLE data option")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
---
net/mptcp/options.c | 17 +++++++++++++++--
1 file changed, 15 insertions(+), 2 deletions(-)
diff --git a/net/mptcp/options.c b/net/mptcp/options.c
index 45acd877bef3..9eb84115dc35 100644
--- a/net/mptcp/options.c
+++ b/net/mptcp/options.c
@@ -334,6 +334,8 @@ static bool mptcp_established_options_dss(struct sock *sk, struct sk_buff *skb,
struct mptcp_sock *msk;
unsigned int ack_size;
bool ret = false;
+ bool can_ack;
+ u64 ack_seq;
u8 tcp_fin;
if (skb) {
@@ -360,9 +362,20 @@ static bool mptcp_established_options_dss(struct sock *sk, struct sk_buff *skb,
ret = true;
}
+ /* passive sockets msk will set the 'can_ack' after accept(), even
+ * if the first subflow may have the already the remote key handy
+ */
+ can_ack = true;
opts->ext_copy.use_ack = 0;
msk = mptcp_sk(subflow->conn);
- if (!msk || !READ_ONCE(msk->can_ack)) {
+ if (likely(msk && READ_ONCE(msk->can_ack)))
+ ack_seq = msk->ack_seq;
+ else if (subflow->can_ack)
+ mptcp_crypto_key_sha(subflow->remote_key, NULL, &ack_seq);
+ else
+ can_ack = false;
+
+ if (unlikely(!can_ack)) {
*size = ALIGN(dss_size, 4);
return ret;
}
@@ -375,7 +388,7 @@ static bool mptcp_established_options_dss(struct sock *sk, struct sk_buff *skb,
dss_size += ack_size;
- opts->ext_copy.data_ack = msk->ack_seq;
+ opts->ext_copy.data_ack = ack_seq;
opts->ext_copy.ack64 = 1;
opts->ext_copy.use_ack = 1;
--
2.21.1
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH net] mptcp: always include dack if possible.
2020-03-03 17:22 [PATCH net] mptcp: always include dack if possible Paolo Abeni
@ 2020-03-03 18:58 ` Mat Martineau
2020-03-03 21:14 ` Paolo Abeni
0 siblings, 1 reply; 3+ messages in thread
From: Mat Martineau @ 2020-03-03 18:58 UTC (permalink / raw
To: Paolo Abeni; +Cc: netdev, Matthieu Baerts, David S. Miller, Christoph Paasch
Paolo -
On Tue, 3 Mar 2020, Paolo Abeni wrote:
> Currently passive MPTCP socket can skip including the DACK
> option - if the peer sends data before accept() completes.
>
> The above happens because the msk 'can_ack' flag is set
> only after the accept() call.
>
> Such missing DACK option may cause - as per RFC spec -
> unwanted fallback to TCP.
>
> This change addresses the issue using the key material
> available in the current subflow, if any, to create a suitable
> dack option when msk ack seq is not yet available.
>
> Fixes: d22f4988ffec ("mptcp: process MP_CAPABLE data option")
> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
> ---
> net/mptcp/options.c | 17 +++++++++++++++--
> 1 file changed, 15 insertions(+), 2 deletions(-)
>
> diff --git a/net/mptcp/options.c b/net/mptcp/options.c
> index 45acd877bef3..9eb84115dc35 100644
> --- a/net/mptcp/options.c
> +++ b/net/mptcp/options.c
> @@ -334,6 +334,8 @@ static bool mptcp_established_options_dss(struct sock *sk, struct sk_buff *skb,
> struct mptcp_sock *msk;
> unsigned int ack_size;
> bool ret = false;
> + bool can_ack;
> + u64 ack_seq;
> u8 tcp_fin;
>
> if (skb) {
> @@ -360,9 +362,20 @@ static bool mptcp_established_options_dss(struct sock *sk, struct sk_buff *skb,
> ret = true;
> }
>
> + /* passive sockets msk will set the 'can_ack' after accept(), even
> + * if the first subflow may have the already the remote key handy
> + */
> + can_ack = true;
> opts->ext_copy.use_ack = 0;
> msk = mptcp_sk(subflow->conn);
> - if (!msk || !READ_ONCE(msk->can_ack)) {
> + if (likely(msk && READ_ONCE(msk->can_ack)))
> + ack_seq = msk->ack_seq;
> + else if (subflow->can_ack)
> + mptcp_crypto_key_sha(subflow->remote_key, NULL, &ack_seq);
The other code paths that set the initial sequence number all increment it
before sending (to ack SYN+MP_CAPABLE). It looks like the spec allows the
value calculated here, but we might as well be consistent about the
initial value we send over the wire.
> + else
> + can_ack = false;
> +
> + if (unlikely(!can_ack)) {
> *size = ALIGN(dss_size, 4);
> return ret;
> }
> @@ -375,7 +388,7 @@ static bool mptcp_established_options_dss(struct sock *sk, struct sk_buff *skb,
>
> dss_size += ack_size;
>
> - opts->ext_copy.data_ack = msk->ack_seq;
> + opts->ext_copy.data_ack = ack_seq;
> opts->ext_copy.ack64 = 1;
> opts->ext_copy.use_ack = 1;
>
> --
> 2.21.1
>
>
--
Mat Martineau
Intel
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH net] mptcp: always include dack if possible.
2020-03-03 18:58 ` Mat Martineau
@ 2020-03-03 21:14 ` Paolo Abeni
0 siblings, 0 replies; 3+ messages in thread
From: Paolo Abeni @ 2020-03-03 21:14 UTC (permalink / raw
To: Mat Martineau; +Cc: netdev, Matthieu Baerts, David S. Miller, Christoph Paasch
On Tue, 2020-03-03 at 10:58 -0800, Mat Martineau wrote:
> On Tue, 3 Mar 2020, Paolo Abeni wrote:
>
> > Currently passive MPTCP socket can skip including the DACK
> > option - if the peer sends data before accept() completes.
> >
> > The above happens because the msk 'can_ack' flag is set
> > only after the accept() call.
> >
> > Such missing DACK option may cause - as per RFC spec -
> > unwanted fallback to TCP.
> >
> > This change addresses the issue using the key material
> > available in the current subflow, if any, to create a suitable
> > dack option when msk ack seq is not yet available.
> >
> > Fixes: d22f4988ffec ("mptcp: process MP_CAPABLE data option")
> > Signed-off-by: Paolo Abeni <pabeni@redhat.com>
> > ---
> > net/mptcp/options.c | 17 +++++++++++++++--
> > 1 file changed, 15 insertions(+), 2 deletions(-)
> >
> > diff --git a/net/mptcp/options.c b/net/mptcp/options.c
> > index 45acd877bef3..9eb84115dc35 100644
> > --- a/net/mptcp/options.c
> > +++ b/net/mptcp/options.c
> > @@ -334,6 +334,8 @@ static bool mptcp_established_options_dss(struct sock *sk, struct sk_buff *skb,
> > struct mptcp_sock *msk;
> > unsigned int ack_size;
> > bool ret = false;
> > + bool can_ack;
> > + u64 ack_seq;
> > u8 tcp_fin;
> >
> > if (skb) {
> > @@ -360,9 +362,20 @@ static bool mptcp_established_options_dss(struct sock *sk, struct sk_buff *skb,
> > ret = true;
> > }
> >
> > + /* passive sockets msk will set the 'can_ack' after accept(), even
> > + * if the first subflow may have the already the remote key handy
> > + */
> > + can_ack = true;
> > opts->ext_copy.use_ack = 0;
> > msk = mptcp_sk(subflow->conn);
> > - if (!msk || !READ_ONCE(msk->can_ack)) {
> > + if (likely(msk && READ_ONCE(msk->can_ack)))
> > + ack_seq = msk->ack_seq;
> > + else if (subflow->can_ack)
> > + mptcp_crypto_key_sha(subflow->remote_key, NULL, &ack_seq);
>
> The other code paths that set the initial sequence number all increment it
> before sending (to ack SYN+MP_CAPABLE). It looks like the spec allows the
> value calculated here, but we might as well be consistent about the
> initial value we send over the wire.
Thanks for the feedback! Agreed. I'll send a v2 tomorrow.
Cheers,
Paolo
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2020-03-03 21:14 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2020-03-03 17:22 [PATCH net] mptcp: always include dack if possible Paolo Abeni
2020-03-03 18:58 ` Mat Martineau
2020-03-03 21:14 ` Paolo Abeni
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.