[RFC v2 mptcp-next] mptcp: support MSG_EOR in mptcp_sendmsg

[RFC v2 mptcp-next] mptcp: support MSG_EOR in mptcp_sendmsg

[Gang Yan]

From: Gang Yan <yangang@kylinos.cn>

This patch adds support for the MSG_EOR flag in MPTCP's sendmsg path,
ensuring that data fragments marked with MSG_EOR are properly handled
to prevent coalescing with subsequent data.

Key changes:
1. Added an 'eor' field to struct mptcp_data_frag to track MSG_EOR marking
2. Initialize the eor field to 0 in mptcp_carve_data_frag()
3. In mptcp_sendmsg_frag(), when sending the last chunk of a data fragment
   that has MSG_EOR set, mark the corresponding skb with
   'TCP_SKB_CB(skb)->eor=1' to prevent coalescing with subsequent data
4. Modified mptcp_sendmsg() to:
   - Preserve MSG_EOR flag in msg_flags filtering
   - Mark the last pending data fragment with eor = 1 when MSG_EOR is set
     in the message flags

This ensures that applications using MSG_EOR to indicate record boundaries
have their intent preserved across MPTCP subflows, maintaining proper
message segmentation semantics.

Signed-off-by: Gang Yan <yangang@kylinos.cn>
---

Notes:
    changelog:
      v2:
        - Fix code-style issues.
        - Submitted a pull request to the mptcp_packetdrill repository:
          https://github.com/multipath-tcp/packetdrill/pull/189
    
    Hi Matt:
    
    Thank you for your feedback on v1. I've addressed the suggestions in this
    updated version. Please take a look when you have a moment.
    
    Thanks
    Gang

 net/mptcp/protocol.c | 22 +++++++++++++++++++---
 net/mptcp/protocol.h |  1 +
 2 files changed, 20 insertions(+), 3 deletions(-)

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index c88882062c40..b8200765506f 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -1174,6 +1174,7 @@ mptcp_carve_data_frag(const struct mptcp_sock *msk, struct page_frag *pfrag,
 	dfrag->offset = offset + sizeof(struct mptcp_data_frag);
 	dfrag->already_sent = 0;
 	dfrag->page = pfrag->page;
+	dfrag->eor = 0;
 
 	return dfrag;
 }
@@ -1434,6 +1435,13 @@ static int mptcp_sendmsg_frag(struct sock *sk, struct sock *ssk,
 		mptcp_update_infinite_map(msk, ssk, mpext);
 	trace_mptcp_sendmsg_frag(mpext);
 	mptcp_subflow_ctx(ssk)->rel_write_seq += copy;
+
+	 /* If this is the last chunk of a dfrag with MSG_EOR set
+	  * mark the skb to prevent coalescing with subsequent data
+	  */
+	if (dfrag->eor && info->sent + copy >= dfrag->data_len)
+		TCP_SKB_CB(skb)->eor = 1;
+
 	return copy;
 }
 
@@ -1894,7 +1902,8 @@ static int mptcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 	long timeo;
 
 	/* silently ignore everything else */
-	msg->msg_flags &= MSG_MORE | MSG_DONTWAIT | MSG_NOSIGNAL | MSG_FASTOPEN;
+	msg->msg_flags &= MSG_MORE | MSG_DONTWAIT | MSG_NOSIGNAL |
+			  MSG_FASTOPEN | MSG_EOR;
 
 	lock_sock(sk);
 
@@ -2001,9 +2010,16 @@ static int mptcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 			goto do_error;
 	}
 
-	if (copied)
-		__mptcp_push_pending(sk, msg->msg_flags);
+	if (copied) {
+		/* Mark the last dfrag with EOR if MSG_EOR was set */
+		if (msg->msg_flags & MSG_EOR) {
+			struct mptcp_data_frag *dfrag = mptcp_pending_tail(sk);
 
+			if (dfrag)
+				dfrag->eor = 1;
+		}
+		__mptcp_push_pending(sk, msg->msg_flags);
+	}
 out:
 	release_sock(sk);
 	return copied;
diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
index cd5266099993..8b243062009c 100644
--- a/net/mptcp/protocol.h
+++ b/net/mptcp/protocol.h
@@ -267,6 +267,7 @@ struct mptcp_data_frag {
 	u16 overhead;
 	u16 already_sent;
 	struct page *page;
+	u8 eor;  /* Is MSG_EOR marked? Prevents coalescing with next frag */
 };
 
 /* Arbitrary compromise between as low as possible to react timely to subflow
-- 
2.43.0

Re: [RFC v2 mptcp-next] mptcp: support MSG_EOR in mptcp_sendmsg

[MPTCP CI]

Hi Gang,

Thank you for your modifications, that's great!

Our CI did some validations and here is its report:

- KVM Validation: normal (except selftest_mptcp_join): Success! ✅
- KVM Validation: normal (only selftest_mptcp_join): Success! ✅
- KVM Validation: debug (except selftest_mptcp_join): Unstable: 2 failed test(s): packetdrill_dss packetdrill_fastopen 🔴
- KVM Validation: debug (only selftest_mptcp_join): Success! ✅
- KVM Validation: btf-normal (only bpftest_all): Success! ✅
- KVM Validation: btf-debug (only bpftest_all): Success! ✅
- Task: https://github.com/multipath-tcp/mptcp_net-next/actions/runs/21614981304

Initiator: Patchew Applier
Commits: https://github.com/multipath-tcp/mptcp_net-next/commits/68562b159b7f
Patchwork: https://patchwork.kernel.org/project/mptcp/list/?series=1050034


If there are some issues, you can reproduce them using the same environment as
the one used by the CI thanks to a docker image, e.g.:

    $ cd [kernel source code]
    $ docker run -v "${PWD}:${PWD}:rw" -w "${PWD}" --privileged --rm -it \
        --pull always mptcp/mptcp-upstream-virtme-docker:latest \
        auto-normal

For more details:

    https://github.com/multipath-tcp/mptcp-upstream-virtme-docker


Please note that despite all the efforts that have been already done to have a
stable tests suite when executed on a public CI like here, it is possible some
reported issues are not due to your modifications. Still, do not hesitate to
help us improve that ;-)

Cheers,
MPTCP GH Action bot
Bot operated by Matthieu Baerts (NGI0 Core)

Re: [RFC v2 mptcp-next] mptcp: support MSG_EOR in mptcp_sendmsg

[Paolo Abeni]

On 2/3/26 3:30 AM, Gang Yan wrote:
> diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
> index cd5266099993..8b243062009c 100644
> --- a/net/mptcp/protocol.h
> +++ b/net/mptcp/protocol.h
> @@ -267,6 +267,7 @@ struct mptcp_data_frag {
>  	u16 overhead;
>  	u16 already_sent;
>  	struct page *page;
> +	u8 eor;  /* Is MSG_EOR marked? Prevents coalescing with next frag */

I'm sorry for the late feedback, but the above will waste quite a bit of
memory due to alignment. I think you can change the 'overhead' size to
u8 and place 'eor' after such field.

Such change could be paired with build time check vs 'overhead'
overflow. i.e.

	BUILD_BUG_ON(ALIGN(1, sizeof(long)) + sizeof(struct mptcp_data_frag) >
U8_MAX);

/P

Re: [RFC v2 mptcp-next] mptcp: support MSG_EOR in mptcp_sendmsg

[Geliang Tang]

Hi Gang,

On Thu, 2026-02-12 at 10:04 +0100, Paolo Abeni wrote:
> On 2/3/26 3:30 AM, Gang Yan wrote:
> > diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
> > index cd5266099993..8b243062009c 100644
> > --- a/net/mptcp/protocol.h
> > +++ b/net/mptcp/protocol.h
> > @@ -267,6 +267,7 @@ struct mptcp_data_frag {
> >  	u16 overhead;
> >  	u16 already_sent;
> >  	struct page *page;
> > +	u8 eor;  /* Is MSG_EOR marked? Prevents coalescing with
> > next frag */
> 
> I'm sorry for the late feedback, but the above will waste quite a bit
> of
> memory due to alignment. I think you can change the 'overhead' size
> to
> u8 and place 'eor' after such field.

I completely agree.
I think it's better not to let 'eor' use all 8 bits here; just one bit
is
sufficient, leaving the other 7 bits for future use, similar to:

	u8 eor:1,
	   __unused:7;

I changed the state of this patch as "Changes Requested".

Thanks,
-Geliang

> 
> Such change could be paired with build time check vs 'overhead'
> overflow. i.e.
> 
> 	BUILD_BUG_ON(ALIGN(1, sizeof(long)) + sizeof(struct
> mptcp_data_frag) >
> U8_MAX);
> 
> /P
> 

Re: [RFC v2 mptcp-next] mptcp: support MSG_EOR in mptcp_sendmsg

[Paolo Abeni]

On 2/28/26 10:12 AM, Geliang Tang wrote:
> On Thu, 2026-02-12 at 10:04 +0100, Paolo Abeni wrote:
>> On 2/3/26 3:30 AM, Gang Yan wrote:
>>> diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
>>> index cd5266099993..8b243062009c 100644
>>> --- a/net/mptcp/protocol.h
>>> +++ b/net/mptcp/protocol.h
>>> @@ -267,6 +267,7 @@ struct mptcp_data_frag {
>>>  	u16 overhead;
>>>  	u16 already_sent;
>>>  	struct page *page;
>>> +	u8 eor;  /* Is MSG_EOR marked? Prevents coalescing with
>>> next frag */
>>
>> I'm sorry for the late feedback, but the above will waste quite a bit
>> of
>> memory due to alignment. I think you can change the 'overhead' size
>> to
>> u8 and place 'eor' after such field.
> 
> I completely agree.
> I think it's better not to let 'eor' use all 8 bits here; just one bit
> is
> sufficient, leaving the other 7 bits for future use, similar to:
> 
> 	u8 eor:1,
> 	   __unused:7;
> 
> I changed the state of this patch as "Changes Requested".

Note that adding a trailing filed will still increases the struct
mptcp_data_frag effective size by 8 bytes.

The maximum value that can land into the 'overhead' field is
'sizeof(struct mptcp_data_frag)' + 7; we can shrink such field to u8 and
move the new one in the created hole. This will prevent struct
mptcp_data_frag size increase.

An eventual build time check on the max 'overhead' value could be a
plus, but it's probably too conservative, given that max is currently 47
and should not increase.

/P