This patch implements the round-robin BPF MPTCP scheduler, named bpf_rr,
which always picks the next available subflow to send data. If no such
next subflow available, picks the first one.

Signed-off-by: Geliang Tang <geliang.tang@suse.com>
---
 .../testing/selftests/bpf/bpf_mptcp_helpers.h |  5 ++
 .../selftests/bpf/progs/mptcp_bpf_rr.c        | 49 +++++++++++++++++++
 2 files changed, 54 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/progs/mptcp_bpf_rr.c

diff --git a/tools/testing/selftests/bpf/bpf_mptcp_helpers.h b/tools/testing/selftests/bpf/bpf_mptcp_helpers.h
index 1fe3d0a97429..8ce5b1603962 100644
--- a/tools/testing/selftests/bpf/bpf_mptcp_helpers.h
+++ b/tools/testing/selftests/bpf/bpf_mptcp_helpers.h
@@ -37,4 +37,9 @@ struct mptcp_sock {
 	char		ca_name[TCP_CA_NAME_MAX];
 } __attribute__((preserve_access_index));
 
+struct mptcp_subflow_context {
+	__u32	token;
+	struct	sock *tcp_sock;	    /* tcp sk backpointer */
+} __attribute__((preserve_access_index));
+
 #endif
diff --git a/tools/testing/selftests/bpf/progs/mptcp_bpf_rr.c b/tools/testing/selftests/bpf/progs/mptcp_bpf_rr.c
new file mode 100644
index 000000000000..2f31de4cfc84
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/mptcp_bpf_rr.c
@@ -0,0 +1,49 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2022, SUSE. */
+
+#include <linux/bpf.h>
+#include <linux/stddef.h>
+#include <linux/tcp.h>
+#include "bpf_mptcp_helpers.h"
+
+char _license[] SEC("license") = "GPL";
+
+SEC("struct_ops/mptcp_sched_rr_init")
+void BPF_PROG(mptcp_sched_rr_init, struct mptcp_sock *msk)
+{
+}
+
+SEC("struct_ops/mptcp_sched_rr_release")
+void BPF_PROG(mptcp_sched_rr_release, struct mptcp_sock *msk)
+{
+}
+
+void BPF_STRUCT_OPS(bpf_rr_get_subflow, struct mptcp_sock *msk,
+		    bool reinject, struct mptcp_sched_data *data)
+{
+	struct mptcp_subflow_context *subflow;
+	struct sock *ssk = msk->first;
+
+	for (int i = 0; i < MPTCP_SUBFLOWS_MAX; i++) {
+		if (i >= data->subflows)
+			break;
+
+		subflow = data->array[i];
+		if (subflow->tcp_sock != msk->last_snd) {
+			ssk = subflow->tcp_sock;
+			break;
+		}
+	}
+
+	msk->last_snd = ssk;
+	data->sock = ssk;
+	data->call_again = 0;
+}
+
+SEC(".struct_ops")
+struct mptcp_sched_ops rr = {
+	.init		= (void *)mptcp_sched_rr_init,
+	.release	= (void *)mptcp_sched_rr_release,
+	.get_subflow	= (void *)bpf_rr_get_subflow,
+	.name		= "bpf_rr",
+};
-- 
2.34.1

On Fri, 29 Apr 2022, Geliang Tang wrote:

> This patch implements the round-robin BPF MPTCP scheduler, named bpf_rr,
> which always picks the next available subflow to send data. If no such
> next subflow available, picks the first one.
>
> Signed-off-by: Geliang Tang <geliang.tang@suse.com>
> ---
> .../testing/selftests/bpf/bpf_mptcp_helpers.h |  5 ++
> .../selftests/bpf/progs/mptcp_bpf_rr.c        | 49 +++++++++++++++++++
> 2 files changed, 54 insertions(+)
> create mode 100644 tools/testing/selftests/bpf/progs/mptcp_bpf_rr.c
>
> diff --git a/tools/testing/selftests/bpf/bpf_mptcp_helpers.h b/tools/testing/selftests/bpf/bpf_mptcp_helpers.h
> index 1fe3d0a97429..8ce5b1603962 100644
> --- a/tools/testing/selftests/bpf/bpf_mptcp_helpers.h
> +++ b/tools/testing/selftests/bpf/bpf_mptcp_helpers.h
> @@ -37,4 +37,9 @@ struct mptcp_sock {
> 	char		ca_name[TCP_CA_NAME_MAX];
> } __attribute__((preserve_access_index));
>
> +struct mptcp_subflow_context {
> +	__u32	token;
> +	struct	sock *tcp_sock;	    /* tcp sk backpointer */
> +} __attribute__((preserve_access_index));
> +
> #endif
> diff --git a/tools/testing/selftests/bpf/progs/mptcp_bpf_rr.c b/tools/testing/selftests/bpf/progs/mptcp_bpf_rr.c
> new file mode 100644
> index 000000000000..2f31de4cfc84
> --- /dev/null
> +++ b/tools/testing/selftests/bpf/progs/mptcp_bpf_rr.c
> @@ -0,0 +1,49 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/* Copyright (c) 2022, SUSE. */
> +
> +#include <linux/bpf.h>
> +#include <linux/stddef.h>
> +#include <linux/tcp.h>
> +#include "bpf_mptcp_helpers.h"
> +
> +char _license[] SEC("license") = "GPL";
> +
> +SEC("struct_ops/mptcp_sched_rr_init")
> +void BPF_PROG(mptcp_sched_rr_init, struct mptcp_sock *msk)
> +{
> +}
> +
> +SEC("struct_ops/mptcp_sched_rr_release")
> +void BPF_PROG(mptcp_sched_rr_release, struct mptcp_sock *msk)
> +{
> +}
> +
> +void BPF_STRUCT_OPS(bpf_rr_get_subflow, struct mptcp_sock *msk,
> +		    bool reinject, struct mptcp_sched_data *data)
> +{
> +	struct mptcp_subflow_context *subflow;
> +	struct sock *ssk = msk->first;
> +
> +	for (int i = 0; i < MPTCP_SUBFLOWS_MAX; i++) {
> +		if (i >= data->subflows)
> +			break;

Is the extra 'if' statement required by the BPF verifier? If this is not 
allowed:

for (int i = 0; i < MPTCP_SUBFLOWS_MAX; i++)

is this?

for (int i = 0; i < MPTCP_SUBFLOWS_MAX && i < data->subflows; i++)

> +
> +		subflow = data->array[i];
> +		if (subflow->tcp_sock != msk->last_snd) {
> +			ssk = subflow->tcp_sock;
> +			break;
> +		}
> +	}
> +
> +	msk->last_snd = ssk;
> +	data->sock = ssk;
> +	data->call_again = 0;
> +}
> +
> +SEC(".struct_ops")
> +struct mptcp_sched_ops rr = {
> +	.init		= (void *)mptcp_sched_rr_init,
> +	.release	= (void *)mptcp_sched_rr_release,
> +	.get_subflow	= (void *)bpf_rr_get_subflow,
> +	.name		= "bpf_rr",
> +};
> -- 
> 2.34.1
>
>
>

--
Mat Martineau
Intel