Skip to contentSkip to conversation

implement .splice_eof

PATCH Superseded Geliang Tang 3 participants 10 messages Feb 2 – Feb 3

Subsystems (4)

v1 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Geliang Tang

author reviewer
#0 lore raw

cover implement .splice_eof

From: Geliang Tang <[email protected]>

This set implements .splice_eof for MPTCP and tests it.

Geliang Tang (3):
tcp: export do_tcp_splice_eof
mptcp: implement .splice_eof
selftests: mptcp: connect: trigger splice_eof

 include/net/tcp.h                                |  1 +
 net/ipv4/tcp.c                                   |  8 ++++++--
 net/mptcp/protocol.c                             | 16 ++++++++++++++++
 .../testing/selftests/net/mptcp/mptcp_connect.c  |  2 +-
 4 files changed, 24 insertions(+), 3 deletions(-)
Signature · 1 line
2.51.0
author reviewer
#1 lore raw

PATCH 1/3 tcp: export do_tcp_splice_eof

From: Geliang Tang <[email protected]>

Extract a do_tcp_splice_eof() helper from tcp_splice_eof() and export it to net/tcp.h, so that it can be used in MPTCP.

---
 include/net/tcp.h | 1 +
 net/ipv4/tcp.c    | 8 ++++++--
 2 files changed, 7 insertions(+), 2 deletions(-)
include/net/tcp.h +1 -0
index cecec1a92d5e..9564fd7b73c7 100644
@@ -370,6 +370,7 @@ int tcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t size);
 int tcp_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t size);
 int tcp_sendmsg_fastopen(struct sock *sk, struct msghdr *msg, int *copied,
 			 size_t size, struct ubuf_info *uarg);
+void do_tcp_splice_eof(struct sock *sk);
 void tcp_splice_eof(struct socket *sock);
 int tcp_send_mss(struct sock *sk, int *size_goal, int flags);
 int tcp_wmem_schedule(struct sock *sk, int copy);
net/ipv4/tcp.c +6 -2
index ae0000215ff4..80d374fe6543 100644
@@ -1468,9 +1468,8 @@ int tcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t size)
 }
 EXPORT_SYMBOL(tcp_sendmsg);
 
-void tcp_splice_eof(struct socket *sock)
+void do_tcp_splice_eof(struct sock *sk)
 {
-	struct sock *sk = sock->sk;
 	struct tcp_sock *tp = tcp_sk(sk);
 	int mss_now, size_goal;
 
@@ -1482,6 +1481,11 @@ void tcp_splice_eof(struct socket *sock)
 	tcp_push(sk, 0, mss_now, tp->nonagle, size_goal);
 	release_sock(sk);
 }
+
+void tcp_splice_eof(struct socket *sock)
+{
+	do_tcp_splice_eof(sock->sk);
+}
 EXPORT_IPV6_MOD_GPL(tcp_splice_eof);
 
 /*
Signature · 1 line
2.51.0
author reviewer
#2 lore raw

PATCH 2/3 mptcp: implement .splice_eof

In reply to Geliang Tang

From: Geliang Tang <[email protected]>

This patch implements the .splice_eof interface for MPTCP, namely mptcp_splice_eof(), which sequentially calls do_tcp_splice_eof() for each subflow.

---
 net/mptcp/protocol.c | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
net/mptcp/protocol.c +16 -0
index c88882062c40..5635d196cb9f 100644
@@ -4018,6 +4018,20 @@ static int mptcp_connect(struct sock *sk, struct sockaddr_unsized *uaddr,
 	return 0;
 }
 
+static void mptcp_splice_eof(struct socket *sock)
+{
+	struct mptcp_subflow_context *subflow;
+	struct sock *sk = sock->sk, *ssk;
+
+	lock_sock(sk);
+	mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
+		ssk = mptcp_subflow_tcp_sock(subflow);
+
+		do_tcp_splice_eof(ssk);
+	}
+	release_sock(sk);
+}
+
 static struct proto mptcp_prot = {
 	.name		= "MPTCP",
 	.owner		= THIS_MODULE,
@@ -4049,6 +4063,7 @@ static struct proto mptcp_prot = {
 	.obj_size	= sizeof(struct mptcp_sock),
 	.slab_flags	= SLAB_TYPESAFE_BY_RCU,
 	.no_autobind	= true,
+	.splice_eof	= mptcp_splice_eof,
 };
 
 static int mptcp_bind(struct socket *sock, struct sockaddr_unsized *uaddr, int addr_len)
@@ -4540,6 +4555,7 @@ static const struct proto_ops mptcp_stream_ops = {
 	.set_rcvlowat	   = mptcp_set_rcvlowat,
 	.read_sock	   = mptcp_read_sock,
 	.splice_read	   = mptcp_splice_read,
+	.splice_eof	   = inet_splice_eof,
 };
 
 static struct inet_protosw mptcp_protosw = {
Signature · 1 line
2.51.0
maintainer
#3 lore raw

Re: [PATCH mptcp-next 2/3] mptcp: implement .splice_eof

Hi Geliang,

Thank you for looking at that!

Quoted 5 lines · Geliang Tang

From: Geliang Tang <[email protected]>

This patch implements the .splice_eof interface for MPTCP, namely mptcp_splice_eof(), which sequentially calls do_tcp_splice_eof() for each subflow.

Can you please explain what this hook is supposed to do / used for please? And also why the solution is to call do_tcp_splice_eof() on each subflow?

Also, I'm a bit confused: why is this needed? Does it fix something or is it a new feature or an optimisation?

Quoted 14 lines
---
 net/mptcp/protocol.c | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
net/mptcp/protocol.c +0 -0
index c88882062c40..5635d196cb9f 100644
@@ -4018,6 +4018,20 @@ static int mptcp_connect(struct sock *sk, struct sockaddr_unsized *uaddr,
 	return 0;
 }
 

net/mptcp/protocol.c static int mptcp_connect(struct sock *sk, struct sockaddr_unsized *uaddr, · line 4030

Quoted 10 lines
net/mptcp/protocol.c +10 -0
+static void mptcp_splice_eof(struct socket *sock)
+{
+	struct mptcp_subflow_context *subflow;
+	struct sock *sk = sock->sk, *ssk;
+
+	lock_sock(sk);
+	mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
+		ssk = mptcp_subflow_tcp_sock(subflow);
+
+		do_tcp_splice_eof(ssk);

Is it fine to call this on closed subflows? e.g. if the initial subflow has been closed. (I didn't check, maybe that's OK)

Quoted 16 lines
net/mptcp/protocol.c +5 -0
+	}
+	release_sock(sk);
+}
+
 static struct proto mptcp_prot = {
 	.name		= "MPTCP",
 	.owner		= THIS_MODULE,
@@ -4049,6 +4063,7 @@ static struct proto mptcp_prot = {
 	.obj_size	= sizeof(struct mptcp_sock),
 	.slab_flags	= SLAB_TYPESAFE_BY_RCU,
 	.no_autobind	= true,
+	.splice_eof	= mptcp_splice_eof,
 };
 
 static int mptcp_bind(struct socket *sock, struct sockaddr_unsized *uaddr, int addr_len)

net/mptcp/protocol.c static const struct proto_ops mptcp_stream_ops = { · line 4558

Quoted 4 lines
net/mptcp/protocol.c +1 -0
@@ -4540,6 +4555,7 @@ static const struct proto_ops mptcp_stream_ops = {
 	.set_rcvlowat	   = mptcp_set_rcvlowat,
 	.read_sock	   = mptcp_read_sock,
 	.splice_read	   = mptcp_splice_read,
+	.splice_eof	   = inet_splice_eof,

Is this line required? Will it not call inet_splice_eof() by default? (I didn't check)

net/mptcp/protocol.c static const struct proto_ops mptcp_stream_ops = { · line 4561

Quoted 3 lines
net/mptcp/protocol.c +0 -0
 };
 
 static struct inet_protosw mptcp_protosw = {

Cheers,
Matt

Signature · 1 line
Sponsored by the NGI0 Core fund.
author reviewer
#4 lore raw

Re: [PATCH mptcp-next 2/3] mptcp: implement .splice_eof

Hi Matt,

Quoted 7 lines · Matthieu Baerts

Hi Geliang,

Thank you for looking at that!

Quoted 2 lines · Geliang Tang

From: Geliang Tang <[email protected]>

Quoted 7 lines
Quoted 4 lines

This patch implements the .splice_eof interface for MPTCP, namely mptcp_splice_eof(), which sequentially calls do_tcp_splice_eof() for
each subflow.

Can you please explain what this hook is supposed to do / used for please?

do_tcp_splice_eof() ensures that any remaining data in the TCP send queue is flushed immediately when a sendfile() operation reaches end- of-file (EOF).

Quoted 2 lines

And also why the solution is to call do_tcp_splice_eof() on each subflow?

MPTCP operates over multiple TCP subflows. When splicing data through an MPTCP socket, each subflow may have pending data in its send buffer that needs to be properly finalized. So here calls do_tcp_splice_eof() on each subflow.

Quoted 4 lines

Also, I'm a bit confused: why is this needed? Does it fix something or
is it a new feature or an optimisation?

It is not a fix, but a new feature, to keep consistent with TCP. Since TCP handles the splice EOF notification but MPTCP didn't.

Quoted 22 lines
Quoted 22 lines

---
 net/mptcp/protocol.c | 16 ++++++++++++++++

 1 file changed, 16 insertions(+)
net/mptcp/protocol.c +0 -0
index c88882062c40..5635d196cb9f 100644
@@ -4018,6 +4018,20 @@ static int mptcp_connect(struct sock *sk,
 struct sockaddr_unsized *uaddr,
  	return 0;
  }

net/mptcp/protocol.c +6 -0
+static void mptcp_splice_eof(struct socket *sock)
+{
+	struct mptcp_subflow_context *subflow;
+	struct sock *sk = sock->sk, *ssk;
+
+	lock_sock(sk);
Quoted 8 lines

net/mptcp/protocol.c static int mptcp_connect(struct sock *sk, struct sockaddr_unsized *uaddr, · line 4030

Quoted 4 lines
net/mptcp/protocol.c +4 -0
+	mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
+		ssk = mptcp_subflow_tcp_sock(subflow);
+
+		do_tcp_splice_eof(ssk);

Is it fine to call this on closed subflows? e.g. if the initial subflow
has been closed. (I didn't check, maybe that's OK)

Good point. I will add this check in v2:

       if (ssk->sk_state == TCP_CLOSE)
               continue;
Quoted 19 lines
Quoted 19 lines
net/mptcp/protocol.c +4 -0
+	}
+	release_sock(sk);
+}
+

 static struct proto mptcp_prot = {
  .name = "MPTCP",
  .owner = THIS_MODULE,

net/mptcp/protocol.c +1 -0
@@ -4049,6 +4063,7 @@ static struct proto mptcp_prot = {
  	.obj_size	= sizeof(struct mptcp_sock),
  	.slab_flags	= SLAB_TYPESAFE_BY_RCU,
  	.no_autobind	= true,
+	.splice_eof	= mptcp_splice_eof,
  };


 static int mptcp_bind(struct socket *sock, struct sockaddr_unsized *uaddr, int addr_len)

net/mptcp/protocol.c +0 -0
@@ -4540,6 +4555,7 @@ static const struct proto_ops
 mptcp_stream_ops = {
Quoted 8 lines

net/mptcp/protocol.c static const struct proto_ops mptcp_stream_ops = { · line 4558

Quoted 4 lines
net/mptcp/protocol.c +1 -0
  	.set_rcvlowat	   = mptcp_set_rcvlowat,
  	.read_sock	   = mptcp_read_sock,
  	.splice_read	   = mptcp_splice_read,
+	.splice_eof	   = inet_splice_eof,

Is this line required? Will it not call inet_splice_eof() by default? (I
didn't check)

Yes, this is required.

sock_splice_eof() needs to call the .splice_eof interface from struct proto_ops. To maintain consistency with regular TCP behavior, the .splice_eof interface of mptcp_stream_ops is set to inet_splice_eof too. inet_splice_eof() will switch to the protocol-specific implementation (sk->sk_prot->splice_eof), which for MPTCP is mptcp_splice_eof().

Thanks,
-Geliang

Quoted 7 lines
Quoted 4 lines
net/mptcp/protocol.c +0 -0
  };


 static struct inet_protosw mptcp_protosw = {

Cheers,
Matt

maintainer
#5 lore raw

Re: [PATCH mptcp-next 2/3] mptcp: implement .splice_eof

Hi Geliang,

Thank you for your reply!

Quoted 15 lines · Geliang Tang

Hi Matt,

Quoted 12 lines · Matthieu Baerts

Hi Geliang,

Thank you for looking at that!

Quoted 6 lines · Geliang Tang

From: Geliang Tang <[email protected]>

This patch implements the .splice_eof interface for MPTCP, namely mptcp_splice_eof(), which sequentially calls do_tcp_splice_eof() for
each subflow.

Quoted 6 lines
Quoted 2 lines

Can you please explain what this hook is supposed to do / used for please?

do_tcp_splice_eof() ensures that any remaining data in the TCP send queue is flushed immediately when a sendfile() operation reaches end- of-file (EOF).

OK, so if I understand correctly, it means that without .splice_eof() support, the queue is not flushed immediately when a sendfile() operation reaches end-of-file (EOF). But that's OK, it will be flushed, eventually with a small delay but the most important is that all data will be transferred. Is that correct?

If it is, can you please reflect that in the commit message? I think it is essential to mention it is not linked to the 'splice()' syscall, it is an improvement, and nothing was broken before.

Quoted 7 lines
Quoted 2 lines

And also why the solution is to call do_tcp_splice_eof() on each subflow?

MPTCP operates over multiple TCP subflows. When splicing data through an MPTCP socket, each subflow may have pending data in its send buffer that needs to be properly finalized. So here calls do_tcp_splice_eof() on each subflow.

Can you also please add a note about that in the commit message?

Quoted 71 lines
Quoted 3 lines

Also, I'm a bit confused: why is this needed? Does it fix something or
is it a new feature or an optimisation?

It is not a fix, but a new feature, to keep consistent with TCP. Since TCP handles the splice EOF notification but MPTCP didn't.

Quoted 22 lines
Quoted 22 lines

---
 net/mptcp/protocol.c | 16 ++++++++++++++++

 1 file changed, 16 insertions(+)
net/mptcp/protocol.c +0 -0
index c88882062c40..5635d196cb9f 100644
@@ -4018,6 +4018,20 @@ static int mptcp_connect(struct sock *sk,
 struct sockaddr_unsized *uaddr,
  	return 0;
  }

net/mptcp/protocol.c +6 -0
+static void mptcp_splice_eof(struct socket *sock)
+{
+	struct mptcp_subflow_context *subflow;
+	struct sock *sk = sock->sk, *ssk;
+
+	lock_sock(sk);
Quoted 8 lines

net/mptcp/protocol.c static int mptcp_connect(struct sock *sk, struct sockaddr_unsized *uaddr, · line 4030

Quoted 4 lines
net/mptcp/protocol.c +4 -0
+	mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
+		ssk = mptcp_subflow_tcp_sock(subflow);
+
+		do_tcp_splice_eof(ssk);

Is it fine to call this on closed subflows? e.g. if the initial subflow
has been closed. (I didn't check, maybe that's OK)

Good point. I will add this check in v2:

       if (ssk->sk_state == TCP_CLOSE)
               continue;
Quoted 19 lines
Quoted 19 lines
net/mptcp/protocol.c +4 -0
+	}
+	release_sock(sk);
+}
+

 static struct proto mptcp_prot = {
  .name = "MPTCP",
  .owner = THIS_MODULE,

net/mptcp/protocol.c +1 -0
@@ -4049,6 +4063,7 @@ static struct proto mptcp_prot = {
  	.obj_size	= sizeof(struct mptcp_sock),
  	.slab_flags	= SLAB_TYPESAFE_BY_RCU,
  	.no_autobind	= true,
+	.splice_eof	= mptcp_splice_eof,
  };


 static int mptcp_bind(struct socket *sock, struct sockaddr_unsized *uaddr, int addr_len)

net/mptcp/protocol.c +0 -0
@@ -4540,6 +4555,7 @@ static const struct proto_ops
 mptcp_stream_ops = {
Quoted 8 lines

net/mptcp/protocol.c static const struct proto_ops mptcp_stream_ops = { · line 4558

Quoted 4 lines
net/mptcp/protocol.c +1 -0
  	.set_rcvlowat	   = mptcp_set_rcvlowat,
  	.read_sock	   = mptcp_read_sock,
  	.splice_read	   = mptcp_splice_read,
+	.splice_eof	   = inet_splice_eof,

Is this line required? Will it not call inet_splice_eof() by default? (I
didn't check)

Quoted 8 lines

Yes, this is required.

sock_splice_eof() needs to call the .splice_eof interface from struct proto_ops. To maintain consistency with regular TCP behavior, the .splice_eof interface of mptcp_stream_ops is set to inet_splice_eof too. inet_splice_eof() will switch to the protocol-specific implementation (sk->sk_prot->splice_eof), which for MPTCP is mptcp_splice_eof().

OK. Then be careful that inet_splice_eof() will call inet_send_prepare() which will call sock_rps_record_flow(sk). mptcp_rps_record_subflows() should be called on each subflow, probably from mptcp_splice_eof().

Cheers,
Matt

Signature · 1 line
Sponsored by the NGI0 Core fund.
author reviewer
#6 lore raw

PATCH 3/3 selftests: mptcp: connect: trigger splice_eof

In reply to Geliang Tang

From: Geliang Tang <[email protected]>

Increase the sendfile count by one to ensure the transmission size exceeds the actual data length. This triggers the splice_eof path in the kernel, allowing the newly implemented MPTCP splice_eof interface to be exercised during testing.

The change from 'count' to 'count + 1' forces the sendfile operation to attempt sending one more byte than available, which activates the end-of-file handling in the splicing logic and ensures coverage of the related MPTCP code paths.

---
 tools/testing/selftests/net/mptcp/mptcp_connect.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/testing/selftests/net/mptcp/mptcp_connect.c +1 -1
index cbe573c4ab3a..2aaf3ed11315 100644
@@ -870,7 +870,7 @@ static int do_sendfile(int infd, int outfd, unsigned int count,
 	while (count > 0) {
 		ssize_t r;
 
-		r = sendfile(outfd, infd, NULL, count);
+		r = sendfile(outfd, infd, NULL, count + 1);
 		if (r < 0) {
 			perror("sendfile");
 			return 3;
Signature · 1 line
2.51.0
maintainer
#7 lore raw

Re: [PATCH mptcp-next 3/3] selftests: mptcp: connect: trigger splice_eof

Hi Geliang,

Quoted 11 lines · Geliang Tang

From: Geliang Tang <[email protected]>

Increase the sendfile count by one to ensure the transmission size exceeds the actual data length. This triggers the splice_eof path in the kernel, allowing the newly implemented MPTCP splice_eof interface to be exercised during testing.

The change from 'count' to 'count + 1' forces the sendfile operation to attempt sending one more byte than available, which activates the end-of-file handling in the splicing logic and ensures coverage of the related MPTCP code paths.

I'm a bit confused: is this splice_eof interface not linked to "splice"? Why is it used with sendfile()?

Also, what's the behaviour without the implementation of "splice_eof()"? Was it a wrong behaviour or is it the same? What's the differences between the situation before and after this series?

Cheers,
Matt

Signature · 1 line
Sponsored by the NGI0 Core fund.
author reviewer
#8 lore raw

Re: [PATCH mptcp-next 3/3] selftests: mptcp: connect: trigger splice_eof

Hi Matt,

Quoted 17 lines · Matthieu Baerts

Hi Geliang,

Quoted 7 lines · Geliang Tang

From: Geliang Tang <[email protected]>

Increase the sendfile count by one to ensure the transmission size exceeds the actual data length. This triggers the splice_eof path in the kernel, allowing the newly implemented MPTCP splice_eof interface to be exercised during testing.

Quoted 6 lines

The change from 'count' to 'count + 1' forces the sendfile operation
to attempt sending one more byte than available, which activates the
end-of-file handling in the splicing logic and ensures coverage of the related MPTCP code paths.

Quoted 3 lines

I'm a bit confused: is this splice_eof interface not linked to "splice"?
Why is it used with sendfile()?

.splice_eof hook is indeed triggered by sendfile(), not by splice().

sendfile -> do_sendfile -> do_splice_direct -> do_splice_direct_actor -> splice_direct_to_actor -> do_splice_eof.

Quoted 5 lines

Also, what's the behaviour without the implementation of "splice_eof()"?
Was it a wrong behaviour or is it the same? What's the differences between the situation before and after this series?

Without 'splice_eof', it was no a wrong behaviour, but MPTCP didn't previously handle the splice EOF notification. TCP handled it.

Quoted 3 lines

Cheers,
Matt

#9 lore raw
In reply to Geliang Tang

Hi Geliang,

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: 1 failed test(s): packetdrill_dss 🔴
- 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/21584982091

Initiator: Patchew Applier
Commits: https://github.com/multipath-tcp/mptcp_net-next/commits/b605c7ce18aa
Patchwork: https://patchwork.kernel.org/project/mptcp/list/?series=1049653

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)