Skip to contentSkip to conversation

BPF packet scheduler

PATCH RESEND Superseded Geliang Tang 3 participants 20 messages Mar 24, 2022 – Mar 25, 2022

Subsystems (6)

v5 of 18 Superseded Latest is v18 →

lore
Share & tools
lore
#0 lore raw

v5 cover BPF packet scheduler

RESEND:
- rebased to export/20220324T054815.
- depends on: "add skc_to_mptcp_sock" v12.

v5:
- patch 1: define per-namespace sched_list (but only used init_net namespace. It is difficult to get 'net' in bpf_mptcp_sched_reg and bpf_mptcp_sched_unreg. I need some suggestions here.)
- patch 2: skip mptcp_sched_default in mptcp_unregister_scheduler.
- patch 8: add tests into mptcp.c, instead of bpf_tcp_ca.c.

v4:
- set msk->sched to &mptcp_sched_default when the sched argument is NULL
in mptcp_init_sched().

v3:
- add mptcp_release_sched helper in patch 4.
- rename mptcp_set_sched to mptcp_init_sched in patch 4.
- add mptcp_sched_first_release in patch 7.
- do some cleanups.

v2:
- split into more small patches.
- change all parameters of mptcp_sched_ops from sk to msk:

       void (*init)(struct mptcp_sock *msk);
       void (*release)(struct mptcp_sock *msk);
       struct sock *   (*get_subflow)(struct mptcp_sock *msk);

- add tests in bpf_tcp_ca.c, instead of adding a new one.

v1:
- Addressed to the commends in the RFC version.

Geliang Tang (8):
mptcp: add struct mptcp_sched_ops
mptcp: register default scheduler
mptcp: add a new sysctl scheduler
mptcp: add sched in mptcp_sock
mptcp: add get_subflow wrapper
mptcp: add bpf_mptcp_sched_ops
selftests: bpf: add bpf_first scheduler
selftests: bpf: add bpf_first test

 Documentation/networking/mptcp-sysctl.rst     |   8 +
 include/net/mptcp.h                           |  13 ++
 kernel/bpf/bpf_struct_ops_types.h             |   4 +
 net/mptcp/Makefile                            |   2 +-
 net/mptcp/bpf.c                               | 102 ++++++++++++
 net/mptcp/ctrl.c                              |  14 ++
 net/mptcp/protocol.c                          |  13 +-
 net/mptcp/protocol.h                          |  14 ++
 net/mptcp/sched.c                             | 152 ++++++++++++++++++
 tools/testing/selftests/bpf/bpf_tcp_helpers.h |  12 ++
 .../testing/selftests/bpf/prog_tests/mptcp.c  |  42 +++++
 tools/testing/selftests/bpf/progs/bpf_first.c |  30 ++++
 12 files changed, 401 insertions(+), 5 deletions(-)
 create mode 100644 net/mptcp/sched.c
 create mode 100644 tools/testing/selftests/bpf/progs/bpf_first.c
Signature · 1 line
2.34.1
#1 lore raw

PATCH RESEND v5 1/8 mptcp: add struct mptcp_sched_ops

This patch added struct mptcp_sched_ops. And define the scheduler register, unregister and find functions.

---
 include/net/mptcp.h  |  13 +++++
 net/mptcp/Makefile   |   2 +-
 net/mptcp/protocol.c |   1 +
 net/mptcp/protocol.h |   7 +++
 net/mptcp/sched.c    | 114 +++++++++++++++++++++++++++++++++++++++++++
 5 files changed, 136 insertions(+), 1 deletion(-)
 create mode 100644 net/mptcp/sched.c
include/net/mptcp.h +13 -0
index 8b1afd6f5cc4..e3a0baa8dbd7 100644
@@ -95,6 +95,19 @@ struct mptcp_out_options {
 #endif
 };
 
+#define MPTCP_SCHED_NAME_MAX 16
+
+struct mptcp_sched_ops {
+	struct sock *	(*get_subflow)(struct mptcp_sock *msk);
+
+	char			name[MPTCP_SCHED_NAME_MAX];
+	struct module		*owner;
+	struct list_head	list;
+
+	void (*init)(struct mptcp_sock *msk);
+	void (*release)(struct mptcp_sock *msk);
+} ____cacheline_aligned_in_smp;
+
 #ifdef CONFIG_MPTCP
 extern struct request_sock_ops mptcp_subflow_request_sock_ops;
 
net/mptcp/Makefile +1 -1
index 0a0608b6b4b4..aa5c10d1b80a 100644
@@ -3,7 +3,7 @@ obj-$(CONFIG_MPTCP) += mptcp.o
 ccflags-y += -DDEBUG
 
 mptcp-y := protocol.o subflow.o options.o token.o crypto.o ctrl.o pm.o diag.o \
-	   mib.o pm_netlink.o sockopt.o
+	   mib.o pm_netlink.o sockopt.o sched.o
 
 obj-$(CONFIG_SYN_COOKIES) += syncookies.o
 obj-$(CONFIG_INET_MPTCP_DIAG) += mptcp_diag.o
net/mptcp/protocol.c +1 -0
index d3887f628b54..b1d7c8b0c112 100644
@@ -3807,6 +3807,7 @@ void __init mptcp_proto_init(void)
 
 	mptcp_subflow_init();
 	mptcp_pm_init();
+	mptcp_sched_init();
 	mptcp_token_init();
 
 	if (proto_register(&mptcp_prot, MPTCP_USE_SLAB) != 0)
net/mptcp/protocol.h +7 -0
index fd82fd113113..3258b740c8ee 100644
@@ -608,6 +608,13 @@ int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock);
 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
 			 struct sockaddr_storage *addr,
 			 unsigned short family);
+struct mptcp_sched_ops *mptcp_sched_find(const struct net *net,
+					 const char *name);
+int mptcp_register_scheduler(const struct net *net,
+			     struct mptcp_sched_ops *sched);
+void mptcp_unregister_scheduler(const struct net *net,
+				struct mptcp_sched_ops *sched);
+void mptcp_sched_init(void);
 
 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
 {
net/mptcp/sched.c new file +114 -0

114 lines not on this page · raw message

Signature · 1 line
2.34.1
#11 lore raw

PATCH RESEND v5 2/8 mptcp: register default scheduler

In reply to Geliang Tang

This patch defined the default packet scheduler mptcp_sched_default, registered it in mptcp_sched_init().

---
 net/mptcp/protocol.c |  2 +-
 net/mptcp/protocol.h |  1 +
 net/mptcp/sched.c    | 10 ++++++++++
 3 files changed, 12 insertions(+), 1 deletion(-)
net/mptcp/protocol.c +1 -1
index b1d7c8b0c112..81b77ade9511 100644
@@ -1424,7 +1424,7 @@ bool mptcp_subflow_active(struct mptcp_subflow_context *subflow)
  * returns the subflow that will transmit the next DSS
  * additionally updates the rtx timeout
  */
-static struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk)
+struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk)
 {
 	struct subflow_send_info send_info[SSK_MODE_MAX];
 	struct mptcp_subflow_context *subflow;
net/mptcp/protocol.h +1 -0
index 3258b740c8ee..46ac5b4528d5 100644
@@ -614,6 +614,7 @@ int mptcp_register_scheduler(const struct net *net,
 			     struct mptcp_sched_ops *sched);
 void mptcp_unregister_scheduler(const struct net *net,
 				struct mptcp_sched_ops *sched);
+struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
 void mptcp_sched_init(void);
 
 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
net/mptcp/sched.c +10 -0
index 3798a5cefeb6..f6ace8cfe865 100644
@@ -23,6 +23,12 @@ struct sched_pernet {
 	struct list_head	sched_list;
 };
 
+static struct mptcp_sched_ops mptcp_sched_default = {
+	.get_subflow    = mptcp_subflow_get_send,
+	.name           = "default",
+	.owner          = THIS_MODULE,
+};
+
 static struct sched_pernet *sched_get_pernet(const struct net *net)
 {
 	return net_generic(net, sched_pernet_id);
@@ -70,6 +76,9 @@ void mptcp_unregister_scheduler(const struct net *net,
 {
 	struct sched_pernet *pernet = sched_get_pernet(net);
 
+	if (sched == &mptcp_sched_default)
+		return;
+
 	spin_lock(&pernet->lock);
 	list_del_rcu(&sched->list);
 	spin_unlock(&pernet->lock);
@@ -111,4 +120,5 @@ void mptcp_sched_init(void)
 {
 	if (register_pernet_subsys(&mptcp_sched_pernet_ops) < 0)
 		panic("Failed to register MPTCP sched pernet subsystem.\n");
+	mptcp_register_scheduler(&init_net, &mptcp_sched_default);
 }
Signature · 1 line
2.34.1
#12 lore raw

PATCH RESEND v5 3/8 mptcp: add a new sysctl scheduler

In reply to Geliang Tang

This patch added a new sysctl, named scheduler, to support for selection of different schedulers.

---
 Documentation/networking/mptcp-sysctl.rst |  8 ++++++++
 net/mptcp/ctrl.c                          | 14 ++++++++++++++
 net/mptcp/protocol.h                      |  1 +
 3 files changed, 23 insertions(+)
Documentation/networking/mptcp-sysctl.rst +8 -0
index e263dfcc4b40..d9e69fdc7ea3 100644
@@ -75,3 +75,11 @@ stale_loss_cnt - INTEGER
 	This is a per-namespace sysctl.
 
 	Default: 4
+
+scheduler - STRING
+	Select the scheduler of your choice.
+
+	Support for selection of different schedulers. This is a per-namespace
+	sysctl.
+
+	Default: "default"
net/mptcp/ctrl.c +14 -0
index ae20b7d92e28..c46c22a84d23 100644
@@ -32,6 +32,7 @@ struct mptcp_pernet {
 	u8 checksum_enabled;
 	u8 allow_join_initial_addr_port;
 	u8 pm_type;
+	char scheduler[MPTCP_SCHED_NAME_MAX];
 };
 
 static struct mptcp_pernet *mptcp_get_pernet(const struct net *net)
@@ -69,6 +70,11 @@ int mptcp_get_pm_type(const struct net *net)
 	return mptcp_get_pernet(net)->pm_type;
 }
 
+const char *mptcp_get_scheduler(const struct net *net)
+{
+	return mptcp_get_pernet(net)->scheduler;
+}
+
 static void mptcp_pernet_set_defaults(struct mptcp_pernet *pernet)
 {
 	pernet->mptcp_enabled = 1;
@@ -77,6 +83,7 @@ static void mptcp_pernet_set_defaults(struct mptcp_pernet *pernet)
 	pernet->allow_join_initial_addr_port = 1;
 	pernet->stale_loss_cnt = 4;
 	pernet->pm_type = MPTCP_PM_TYPE_KERNEL;
+	strcpy(pernet->scheduler, "default");
 }
 
 #ifdef CONFIG_SYSCTL
@@ -128,6 +135,12 @@ static struct ctl_table mptcp_sysctl_table[] = {
 		.extra1       = SYSCTL_ZERO,
 		.extra2       = &mptcp_pm_type_max
 	},
+	{
+		.procname = "scheduler",
+		.maxlen	= MPTCP_SCHED_NAME_MAX,
+		.mode = 0644,
+		.proc_handler = proc_dostring,
+	},
 	{}
 };
 
@@ -149,6 +162,7 @@ static int mptcp_pernet_new_table(struct net *net, struct mptcp_pernet *pernet)
 	table[3].data = &pernet->allow_join_initial_addr_port;
 	table[4].data = &pernet->stale_loss_cnt;
 	table[5].data = &pernet->pm_type;
+	table[6].data = &pernet->scheduler;
 
 	hdr = register_net_sysctl(net, MPTCP_SYSCTL_PATH, table);
 	if (!hdr)
net/mptcp/protocol.h +1 -0
index 46ac5b4528d5..3e69c1e5600a 100644
@@ -586,6 +586,7 @@ int mptcp_is_checksum_enabled(const struct net *net);
 int mptcp_allow_join_id0(const struct net *net);
 unsigned int mptcp_stale_loss_cnt(const struct net *net);
 int mptcp_get_pm_type(const struct net *net);
+const char *mptcp_get_scheduler(const struct net *net);
 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow,
 				     struct mptcp_options_received *mp_opt);
 bool __mptcp_retransmit_pending_data(struct sock *sk);
Signature · 1 line
2.34.1
#13 lore raw

PATCH RESEND v5 4/8 mptcp: add sched in mptcp_sock

In reply to Geliang Tang

This patch added a new struct member sched in struct mptcp_sock. And two helpers mptcp_init_sched() and mptcp_release_sched() to init and release it.

Init it with the sysctl scheduler in mptcp_init_sock(), copy the scheduler from the parent in mptcp_sk_clone(), and release it in __mptcp_destroy_sock().

---
 net/mptcp/protocol.c |  4 ++++
 net/mptcp/protocol.h |  4 ++++
 net/mptcp/sched.c    | 21 +++++++++++++++++++++
 3 files changed, 29 insertions(+)
net/mptcp/protocol.c +4 -0
index 81b77ade9511..c7969b9b567e 100644
@@ -2661,6 +2661,8 @@ static int mptcp_init_sock(struct sock *sk)
 	 * propagate the correct value
 	 */
 	mptcp_ca_reset(sk);
+	mptcp_init_sched(mptcp_sk(sk),
+			 mptcp_sched_find(net, mptcp_get_scheduler(net)));
 
 	sk_sockets_allocated_inc(sk);
 	sk->sk_rcvbuf = sock_net(sk)->ipv4.sysctl_tcp_rmem[1];
@@ -2816,6 +2818,7 @@ static void __mptcp_destroy_sock(struct sock *sk)
 	sk_stop_timer(sk, &sk->sk_timer);
 	mptcp_data_unlock(sk);
 	msk->pm.status = 0;
+	mptcp_release_sched(msk);
 
 	/* clears msk->subflow, allowing the following loop to close
 	 * even the initial subflow
@@ -2993,6 +2996,7 @@ struct sock *mptcp_sk_clone(const struct sock *sk,
 	msk->snd_una = msk->write_seq;
 	msk->wnd_end = msk->snd_nxt + req->rsk_rcv_wnd;
 	msk->setsockopt_seq = mptcp_sk(sk)->setsockopt_seq;
+	mptcp_init_sched(msk, mptcp_sk(sk)->sched);
 
 	if (mp_opt->suboptions & OPTIONS_MPTCP_MPC) {
 		msk->can_ack = true;
net/mptcp/protocol.h +4 -0
index 3e69c1e5600a..ad45281e8562 100644
@@ -288,6 +288,7 @@ struct mptcp_sock {
 	struct socket	*subflow; /* outgoing connect/listener/!mp_capable */
 	struct sock	*first;
 	struct mptcp_pm_data	pm;
+	struct mptcp_sched_ops	*sched;
 	struct {
 		u32	space;	/* bytes copied in last measurement window */
 		u32	copied; /* bytes copied in this measurement window */
@@ -617,6 +618,9 @@ void mptcp_unregister_scheduler(const struct net *net,
 				struct mptcp_sched_ops *sched);
 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
 void mptcp_sched_init(void);
+void mptcp_init_sched(struct mptcp_sock *msk,
+		      struct mptcp_sched_ops *sched);
+void mptcp_release_sched(struct mptcp_sock *msk);
 
 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
 {
net/mptcp/sched.c +21 -0
index f6ace8cfe865..e431b1eec5e1 100644
@@ -122,3 +122,24 @@ void mptcp_sched_init(void)
 		panic("Failed to register MPTCP sched pernet subsystem.\n");
 	mptcp_register_scheduler(&init_net, &mptcp_sched_default);
 }
+
+void mptcp_init_sched(struct mptcp_sock *msk,
+		      struct mptcp_sched_ops *sched)
+{
+	if (!sched)
+		msk->sched = &mptcp_sched_default;
+	else
+		msk->sched = sched;
+
+	if (msk->sched->init)
+		msk->sched->init(msk);
+
+	pr_debug("sched=%s", msk->sched->name);
+}
+
+void mptcp_release_sched(struct mptcp_sock *msk)
+{
+	if (msk->sched && msk->sched->release)
+		msk->sched->release(msk);
+	msk->sched = NULL;
+}
Signature · 1 line
2.34.1
#14 lore raw

PATCH RESEND v5 5/8 mptcp: add get_subflow wrapper

In reply to Geliang Tang

This patch defined a new wrapper mptcp_sched_get_subflow(), invoke get_subflow of msk->sched in it. Use the wrapper instead of using mptcp_subflow_get_send() directly.

---
 net/mptcp/protocol.c | 6 +++---
 net/mptcp/protocol.h | 1 +
 net/mptcp/sched.c    | 7 +++++++
 3 files changed, 11 insertions(+), 3 deletions(-)
net/mptcp/protocol.c +3 -3
index c7969b9b567e..241cfda822c0 100644
@@ -1567,7 +1567,7 @@ void __mptcp_push_pending(struct sock *sk, unsigned int flags)
 			int ret = 0;
 
 			prev_ssk = ssk;
-			ssk = mptcp_subflow_get_send(msk);
+			ssk = mptcp_sched_get_subflow(msk);
 
 			/* First check. If the ssk has changed since
 			 * the last round, release prev_ssk
@@ -1636,7 +1636,7 @@ static void __mptcp_subflow_push_pending(struct sock *sk, struct sock *ssk)
 			 * check for a different subflow usage only after
 			 * spooling the first chunk of data
 			 */
-			xmit_ssk = first ? ssk : mptcp_subflow_get_send(mptcp_sk(sk));
+			xmit_ssk = first ? ssk : mptcp_sched_get_subflow(mptcp_sk(sk));
 			if (!xmit_ssk)
 				goto out;
 			if (xmit_ssk != ssk) {
@@ -3134,7 +3134,7 @@ void __mptcp_check_push(struct sock *sk, struct sock *ssk)
 		return;
 
 	if (!sock_owned_by_user(sk)) {
-		struct sock *xmit_ssk = mptcp_subflow_get_send(mptcp_sk(sk));
+		struct sock *xmit_ssk = mptcp_sched_get_subflow(mptcp_sk(sk));
 
 		if (xmit_ssk == ssk)
 			__mptcp_subflow_push_pending(sk, ssk);
net/mptcp/protocol.h +1 -0
index ad45281e8562..7781eb527b9a 100644
@@ -621,6 +621,7 @@ void mptcp_sched_init(void);
 void mptcp_init_sched(struct mptcp_sock *msk,
 		      struct mptcp_sched_ops *sched);
 void mptcp_release_sched(struct mptcp_sock *msk);
+struct sock *mptcp_sched_get_subflow(struct mptcp_sock *msk);
 
 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
 {
net/mptcp/sched.c +7 -0
index e431b1eec5e1..2fd2a034ba49 100644
@@ -11,6 +11,7 @@
 #include <linux/list.h>
 #include <linux/rculist.h>
 #include <linux/spinlock.h>
+#include <linux/indirect_call_wrapper.h>
 #include <net/tcp.h>
 #include <net/netns/generic.h>
 #include "protocol.h"
@@ -143,3 +144,9 @@ void mptcp_release_sched(struct mptcp_sock *msk)
 		msk->sched->release(msk);
 	msk->sched = NULL;
 }
+
+struct sock *mptcp_sched_get_subflow(struct mptcp_sock *msk)
+{
+	return INDIRECT_CALL_INET_1(msk->sched->get_subflow,
+				    mptcp_subflow_get_send, msk);
+}
Signature · 1 line
2.34.1
#16 lore raw

PATCH RESEND v5 6/8 mptcp: add bpf_mptcp_sched_ops

In reply to Geliang Tang

This patch implemented a new struct bpf_struct_ops, bpf_mptcp_sched_ops. Register and unregister the mptcp scheduler in .reg and .unreg.

---
 kernel/bpf/bpf_struct_ops_types.h |   4 ++
 net/mptcp/bpf.c                   | 102 ++++++++++++++++++++++++++++++
 2 files changed, 106 insertions(+)
kernel/bpf/bpf_struct_ops_types.h +4 -0
index 5678a9ddf817..5a6b0c0d8d3d 100644
@@ -8,5 +8,9 @@ BPF_STRUCT_OPS_TYPE(bpf_dummy_ops)
 #ifdef CONFIG_INET
 #include <net/tcp.h>
 BPF_STRUCT_OPS_TYPE(tcp_congestion_ops)
+#ifdef CONFIG_MPTCP
+#include <net/mptcp.h>
+BPF_STRUCT_OPS_TYPE(mptcp_sched_ops)
+#endif
 #endif
 #endif
net/mptcp/bpf.c +102 -0

110 lines not on this page · raw message

Signature · 1 line
2.34.1
#17 lore raw

PATCH RESEND v5 7/8 selftests: bpf: add bpf_first scheduler

In reply to Geliang Tang

This patch implemented the simplest MPTCP scheduler, named bpf_first, which always picks the first subflow to send data.

---
 tools/testing/selftests/bpf/bpf_tcp_helpers.h | 12 ++++++++
 tools/testing/selftests/bpf/progs/bpf_first.c | 30 +++++++++++++++++++
 2 files changed, 42 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/progs/bpf_first.c
tools/testing/selftests/bpf/bpf_tcp_helpers.h +12 -0
index f92357597e63..5bb21b24231f 100644
@@ -197,6 +197,18 @@ struct tcp_congestion_ops {
 	void *owner;
 };
 
+#define MPTCP_SCHED_NAME_MAX 16
+
+struct mptcp_sched_ops {
+	char name[MPTCP_SCHED_NAME_MAX];
+
+	void (*init)(struct mptcp_sock *msk);
+	void (*release)(struct mptcp_sock *msk);
+
+	struct sock *	(*get_subflow)(struct mptcp_sock *msk);
+	void *owner;
+};
+
 #define min(a, b) ((a) < (b) ? (a) : (b))
 #define max(a, b) ((a) > (b) ? (a) : (b))
 #define min_not_zero(x, y) ({			\
tools/testing/selftests/bpf/progs/bpf_first.c new file +30 -0
new file mode 100644
index 000000000000..3ae5c1eab5f5
@@ -0,0 +1,30 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* Copyright (c) 2022, SUSE. */
+
+#include <linux/bpf.h>
+#include "bpf_tcp_helpers.h"
+
+char _license[] SEC("license") = "GPL";
+
+SEC("struct_ops/mptcp_sched_first_init")
+void BPF_PROG(mptcp_sched_first_init, struct mptcp_sock *msk)
+{
+}
+
+SEC("struct_ops/mptcp_sched_first_release")
+void BPF_PROG(mptcp_sched_first_release, struct mptcp_sock *msk)
+{
+}
+
+struct sock *BPF_STRUCT_OPS(bpf_first_get_subflow, struct mptcp_sock *msk)
+{
+	return msk->first;
+}
+
+SEC(".struct_ops")
+struct mptcp_sched_ops first = {
+	.init		= (void *)mptcp_sched_first_init,
+	.release	= (void *)mptcp_sched_first_release,
+	.get_subflow	= (void *)bpf_first_get_subflow,
+	.name		= "bpf_first",
+};
Signature · 1 line
2.34.1
#18 lore raw

PATCH RESEND v5 8/8 selftests: bpf: add bpf_first test

In reply to Geliang Tang

This patch expended the MPTCP test base to support MPTCP packet scheduler tests. Add the bpf_first MPTCP sched test in it. Use sysctl to set net.mptcp.scheduler to use this sched.

---
 .../testing/selftests/bpf/prog_tests/mptcp.c  | 42 +++++++++++++++++++
 1 file changed, 42 insertions(+)
tools/testing/selftests/bpf/prog_tests/mptcp.c +42 -0
index b7c230c4efef..151350202fa6 100644
@@ -4,6 +4,7 @@
 #include <test_progs.h>
 #include "cgroup_helpers.h"
 #include "network_helpers.h"
+#include "bpf_first.skel.h"
 
 #ifndef TCP_CA_NAME_MAX
 #define TCP_CA_NAME_MAX	16
@@ -19,6 +20,15 @@ struct mptcp_storage {
 };
 
 static char monitor_log_path[64];
+static int duration;
+
+static void send_byte(int fd)
+{
+	char b = 0x55;
+
+	if (CHECK_FAIL(write(fd, &b, sizeof(b)) != 1))
+		perror("Failed to send single byte");
+}
 
 static int verify_tsk(int map_fd, int client_fd)
 {
@@ -202,6 +212,8 @@ static int run_test(int cgroup_fd, int server_fd, bool is_mptcp)
 		goto close_client_fd;
 	}
 
+	send_byte(client_fd);
+
 	err += is_mptcp ? verify_msk(map_fd, client_fd) :
 			  verify_tsk(map_fd, client_fd);
 
@@ -254,8 +266,38 @@ void test_base(void)
 	close(cgroup_fd);
 }
 
+static void test_first(void)
+{
+	struct bpf_first *first_skel;
+	int server_fd, client_fd;
+	struct bpf_link *link;
+
+	first_skel = bpf_first__open_and_load();
+	if (CHECK(!first_skel, "bpf_first__open_and_load", "failed\n"))
+		return;
+
+	link = bpf_map__attach_struct_ops(first_skel->maps.first);
+	if (!ASSERT_OK_PTR(link, "bpf_map__attach_struct_ops")) {
+		bpf_first__destroy(first_skel);
+		return;
+	}
+
+	system("sysctl -q net.mptcp.scheduler=bpf_first");
+	server_fd = start_mptcp_server(AF_INET, NULL, 0, 0);
+	client_fd = connect_to_mptcp_fd(server_fd, 0);
+
+	send_byte(client_fd);
+
+	close(client_fd);
+	close(server_fd);
+	bpf_link__destroy(link);
+	bpf_first__destroy(first_skel);
+}
+
 void test_mptcp(void)
 {
 	if (test__start_subtest("base"))
 		test_base();
+	if (test__start_subtest("first"))
+		test_first();
 }
Signature · 1 line
2.34.1
#2 lore raw

Re: [PATCH RESEND mptcp-next v5 1/8] mptcp: add struct mptcp_sched_ops

In reply to Geliang Tang
Quoted 2 lines · Geliang Tang

This patch added struct mptcp_sched_ops. And define the scheduler register, unregister and find functions.

... but why are they pernet? Makes no sense to me, so an explanation would help. Or, remove the pernet ops.

All callers pass &init_net, so I don't think there is any reason for pernet data structures here.

Quoted 5 lines
net/mptcp/sched.c new file +5 -0
+struct mptcp_sched_ops *mptcp_sched_find(const struct net *net,
+					 const char *name)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+	struct mptcp_sched_ops *sched, *ret = NULL;

net/mptcp/sched.c line 45

Quoted 10 lines
net/mptcp/sched.c new file +10 -0
+
+	spin_lock(&pernet->lock);
+	list_for_each_entry_rcu(sched, &pernet->sched_list, list) {
+		if (!strcmp(sched->name, name)) {
+			ret = sched;
+			break;
+		}
+	}
+	spin_unlock(&pernet->lock);
+

Locking isn't needed here if caller holds rcu read lock.

Quoted 4 lines
net/mptcp/sched.c new file +4 -0
+int mptcp_register_scheduler(const struct net *net,
+			     struct mptcp_sched_ops *sched)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);

net/mptcp/sched.c line 62

Quoted 10 lines
net/mptcp/sched.c new file +10 -0
+
+	if (!sched->get_subflow)
+		return -EINVAL;
+
+	if (mptcp_sched_find(net, sched->name))
+		return -EEXIST;
+
+	spin_lock(&pernet->lock);
+	list_add_tail_rcu(&sched->list, &pernet->sched_list);
+	spin_unlock(&pernet->lock);

I would use global list/lock, not pernet registration. I don't see a reason why that is needed, its just duplicate info.

net/mptcp/sched.c line 77

Quoted 10 lines
net/mptcp/sched.c new file +10 -0
+void mptcp_unregister_scheduler(const struct net *net,
+				struct mptcp_sched_ops *sched)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+
+	spin_lock(&pernet->lock);
+	list_del_rcu(&sched->list);
+	spin_unlock(&pernet->lock);
+
+	synchronize_rcu();

Why synchronize_rcu()? Comment please why that is needed, no freeing happens?

net/mptcp/sched.c line 100

Quoted 9 lines
net/mptcp/sched.c new file +9 -0
+	struct sched_pernet *pernet = sched_get_pernet(net);
+	struct mptcp_sched_ops *sched;
+
+	spin_lock(&pernet->lock);
+	list_for_each_entry_rcu(sched, &pernet->sched_list, list)
+		list_del_rcu(&sched->list);
+	spin_unlock(&pernet->lock);
+
+	synchronize_rcu();

Same, why is it needed?

#3 lore raw

Re: [PATCH RESEND mptcp-next v5 1/8] mptcp: add struct mptcp_sched_ops

Quoted 9 lines · Florian Westphal
Quoted 2 lines · Geliang Tang

This patch added struct mptcp_sched_ops. And define the scheduler register, unregister and find functions.

... but why are they pernet? Makes no sense to me, so an explanation would help. Or, remove the pernet ops.

All callers pass &init_net, so I don't think there is any reason for pernet data structures here.

v4 used global list instead of pernet list:

https://patchwork.kernel.org/project/mptcp/patch/02c1d2d2eee134713737fe5c5f73127e8c741589.1647942374.git.geliang.tang@suse.com/

Should I go back to v4?

Quoted 63 lines
Quoted 6 lines
net/mptcp/sched.c new file +5 -0
+struct mptcp_sched_ops *mptcp_sched_find(const struct net *net,
+					 const char *name)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+	struct mptcp_sched_ops *sched, *ret = NULL;

net/mptcp/sched.c line 45

Quoted 10 lines
net/mptcp/sched.c new file +10 -0
+
+	spin_lock(&pernet->lock);
+	list_for_each_entry_rcu(sched, &pernet->sched_list, list) {
+		if (!strcmp(sched->name, name)) {
+			ret = sched;
+			break;
+		}
+	}
+	spin_unlock(&pernet->lock);
+

Locking isn't needed here if caller holds rcu read lock.

Quoted 4 lines
net/mptcp/sched.c new file +4 -0
+int mptcp_register_scheduler(const struct net *net,
+			     struct mptcp_sched_ops *sched)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);

net/mptcp/sched.c line 62

Quoted 10 lines
net/mptcp/sched.c new file +10 -0
+
+	if (!sched->get_subflow)
+		return -EINVAL;
+
+	if (mptcp_sched_find(net, sched->name))
+		return -EEXIST;
+
+	spin_lock(&pernet->lock);
+	list_add_tail_rcu(&sched->list, &pernet->sched_list);
+	spin_unlock(&pernet->lock);

I would use global list/lock, not pernet registration. I don't see a reason why that is needed, its just duplicate info.

net/mptcp/sched.c line 77

Quoted 10 lines
net/mptcp/sched.c new file +10 -0
+void mptcp_unregister_scheduler(const struct net *net,
+				struct mptcp_sched_ops *sched)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+
+	spin_lock(&pernet->lock);
+	list_del_rcu(&sched->list);
+	spin_unlock(&pernet->lock);
+
+	synchronize_rcu();

Why synchronize_rcu()? Comment please why that is needed, no freeing happens?

net/mptcp/sched.c line 100

Quoted 9 lines
net/mptcp/sched.c new file +9 -0
+	struct sched_pernet *pernet = sched_get_pernet(net);
+	struct mptcp_sched_ops *sched;
+
+	spin_lock(&pernet->lock);
+	list_for_each_entry_rcu(sched, &pernet->sched_list, list)
+		list_del_rcu(&sched->list);
+	spin_unlock(&pernet->lock);
+
+	synchronize_rcu();

Same, why is it needed?

#4 lore raw

Re: [PATCH RESEND mptcp-next v5 1/8] mptcp: add struct mptcp_sched_ops

Quoted 8 lines · Geliang Tang
Quoted 7 lines · Florian Westphal
Quoted 2 lines · Geliang Tang

This patch added struct mptcp_sched_ops. And define the scheduler register, unregister and find functions.

... but why are they pernet? Makes no sense to me, so an explanation would help. Or, remove the pernet ops.

Quoted 8 lines
Quoted 2 lines

All callers pass &init_net, so I don't think there is any reason for pernet data structures here.

v4 used global list instead of pernet list:

https://patchwork.kernel.org/project/mptcp/patch/02c1d2d2eee134713737fe5c5f73127e8c741589.1647942374.git.geliang.tang@suse.com/

Should I go back to v4?

Mhh. Mat, why the pernet thing? AFAICS all net namespaces would contain the same (albeit we'd have to kmemdup first to get unique list_heads....) info, so I don't understand why thats a good idea?

A single/global lock is fine if its not taken in the normal (per packet) case.

#5 lore raw

Re: [PATCH RESEND mptcp-next v5 1/8] mptcp: add struct mptcp_sched_ops

Quoted 9 lines · Florian Westphal
Quoted 8 lines · Geliang Tang
Quoted 7 lines · Florian Westphal
Quoted 2 lines · Geliang Tang

This patch added struct mptcp_sched_ops. And define the scheduler register, unregister and find functions.

... but why are they pernet? Makes no sense to me, so an explanation would help. Or, remove the pernet ops.

Quoted 6 lines
Quoted 6 lines
Quoted 2 lines

All callers pass &init_net, so I don't think there is any reason for pernet data structures here.

v4 used global list instead of pernet list:

https://patchwork.kernel.org/project/mptcp/patch/02c1d2d2eee134713737fe5c5f73127e8c741589.1647942374.git.geliang.tang@suse.com/

Florian, should I remove synchronize_rcu() in v4 too?

Quoted 9 lines
Quoted 2 lines

Should I go back to v4?

Mhh. Mat, why the pernet thing? AFAICS all net namespaces would contain the same (albeit we'd have to kmemdup first to get unique list_heads....) info, so I don't understand why thats a good idea?

A single/global lock is fine if its not taken in the normal (per packet) case.

#6 lore raw

Re: [PATCH RESEND mptcp-next v5 1/8] mptcp: add struct mptcp_sched_ops

Quoted 10 lines · Geliang Tang
Quoted 9 lines · Florian Westphal
Quoted 8 lines · Geliang Tang
Quoted 7 lines · Florian Westphal
Quoted 2 lines · Geliang Tang

This patch added struct mptcp_sched_ops. And define the scheduler register, unregister and find functions.

... but why are they pernet? Makes no sense to me, so an explanation would help. Or, remove the pernet ops.

Quoted 8 lines
Quoted 6 lines
Quoted 6 lines
Quoted 2 lines

All callers pass &init_net, so I don't think there is any reason for pernet data structures here.

v4 used global list instead of pernet list:

https://patchwork.kernel.org/project/mptcp/patch/02c1d2d2eee134713737fe5c5f73127e8c741589.1647942374.git.geliang.tang@suse.com/

Florian, should I remove synchronize_rcu() in v4 too?

I think so. Else you need to add a comment as to why its needed.

I suspect full 'rmmod mptcp_sched_example' sequence would be:

1. mptcp_unregister_scheduler(&example_sched);
2. walk all active mptcp sockets to re-set their scheduler in case something still uses 'example_sched'
3. complete the rmmod

In that case, we might need a synchronize_rcu after 2), but not 1.

#7 lore raw

Re: [PATCH RESEND mptcp-next v5 1/8] mptcp: add struct mptcp_sched_ops

In reply to Geliang Tang
Quoted 42 lines · Geliang Tang

This patch added struct mptcp_sched_ops. And define the scheduler register, unregister and find functions.

---

include/net/mptcp.h  |  13 +++++
net/mptcp/Makefile   |   2 +-
net/mptcp/protocol.c |   1 +
net/mptcp/protocol.h |   7 +++
net/mptcp/sched.c    | 114 +++++++++++++++++++++++++++++++++++++++++++
5 files changed, 136 insertions(+), 1 deletion(-)

create mode 100644 net/mptcp/sched.c

include/net/mptcp.h +13 -0
index 8b1afd6f5cc4..e3a0baa8dbd7 100644
@@ -95,6 +95,19 @@ struct mptcp_out_options {
 #endif
 };
 
+#define MPTCP_SCHED_NAME_MAX 16
+
+struct mptcp_sched_ops {
+	struct sock *	(*get_subflow)(struct mptcp_sock *msk);
+
+	char			name[MPTCP_SCHED_NAME_MAX];
+	struct module		*owner;
+	struct list_head	list;
+
+	void (*init)(struct mptcp_sock *msk);
+	void (*release)(struct mptcp_sock *msk);
+} ____cacheline_aligned_in_smp;
+
 #ifdef CONFIG_MPTCP
 extern struct request_sock_ops mptcp_subflow_request_sock_ops;
 
net/mptcp/Makefile +0 -0
index 0a0608b6b4b4..aa5c10d1b80a 100644
@@ -3,7 +3,7 @@ obj-$(CONFIG_MPTCP) += mptcp.o
Quoted 1 line

ccflags-y += -DDEBUG

This line of context keeps making 'git am' fail - looks like you have a debug-enabling commit in your private tree?

This is probably why patchew is having trouble with this series too.

- Mat

Quoted 162 lines

mptcp-y := protocol.o subflow.o options.o token.o crypto.o ctrl.o pm.o diag.o \

net/mptcp/Makefile +1 -1
-	   mib.o pm_netlink.o sockopt.o
+	   mib.o pm_netlink.o sockopt.o sched.o

obj-$(CONFIG_SYN_COOKIES) += syncookies.o
obj-$(CONFIG_INET_MPTCP_DIAG) += mptcp_diag.o

net/mptcp/protocol.c +1 -0
index d3887f628b54..b1d7c8b0c112 100644
@@ -3807,6 +3807,7 @@ void __init mptcp_proto_init(void)
 
 	mptcp_subflow_init();
 	mptcp_pm_init();
+	mptcp_sched_init();
 	mptcp_token_init();
 
 	if (proto_register(&mptcp_prot, MPTCP_USE_SLAB) != 0)
net/mptcp/protocol.h +7 -0
index fd82fd113113..3258b740c8ee 100644
@@ -608,6 +608,13 @@ int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock);
 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
 			 struct sockaddr_storage *addr,
 			 unsigned short family);
+struct mptcp_sched_ops *mptcp_sched_find(const struct net *net,
+					 const char *name);
+int mptcp_register_scheduler(const struct net *net,
+			     struct mptcp_sched_ops *sched);
+void mptcp_unregister_scheduler(const struct net *net,
+				struct mptcp_sched_ops *sched);
+void mptcp_sched_init(void);
 
 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
 {
net/mptcp/sched.c new file +114 -0
new file mode 100644
index 000000000000..3798a5cefeb6
@@ -0,0 +1,114 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Multipath TCP
+ *
+ * Copyright (c) 2022, SUSE.
+ */
+
+#define pr_fmt(fmt) "MPTCP: " fmt
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/list.h>
+#include <linux/rculist.h>
+#include <linux/spinlock.h>
+#include <net/tcp.h>
+#include <net/netns/generic.h>
+#include "protocol.h"
+
+static int sched_pernet_id;
+
+struct sched_pernet {
+	/* protects pernet updates */
+	spinlock_t		lock;
+	struct list_head	sched_list;
+};
+
+static struct sched_pernet *sched_get_pernet(const struct net *net)
+{
+	return net_generic(net, sched_pernet_id);
+}
+
+struct mptcp_sched_ops *mptcp_sched_find(const struct net *net,
+					 const char *name)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+	struct mptcp_sched_ops *sched, *ret = NULL;
+
+	spin_lock(&pernet->lock);
+	list_for_each_entry_rcu(sched, &pernet->sched_list, list) {
+		if (!strcmp(sched->name, name)) {
+			ret = sched;
+			break;
+		}
+	}
+	spin_unlock(&pernet->lock);
+
+	return ret;
+}
+
+int mptcp_register_scheduler(const struct net *net,
+			     struct mptcp_sched_ops *sched)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+
+	if (!sched->get_subflow)
+		return -EINVAL;
+
+	if (mptcp_sched_find(net, sched->name))
+		return -EEXIST;
+
+	spin_lock(&pernet->lock);
+	list_add_tail_rcu(&sched->list, &pernet->sched_list);
+	spin_unlock(&pernet->lock);
+
+	pr_debug("%s registered", sched->name);
+	return 0;
+}
+
+void mptcp_unregister_scheduler(const struct net *net,
+				struct mptcp_sched_ops *sched)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+
+	spin_lock(&pernet->lock);
+	list_del_rcu(&sched->list);
+	spin_unlock(&pernet->lock);
+
+	synchronize_rcu();
+}
+
+static int __net_init sched_init_net(struct net *net)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+
+	INIT_LIST_HEAD_RCU(&pernet->sched_list);
+	spin_lock_init(&pernet->lock);
+
+	return 0;
+}
+
+static void __net_exit sched_exit_net(struct net *net)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+	struct mptcp_sched_ops *sched;
+
+	spin_lock(&pernet->lock);
+	list_for_each_entry_rcu(sched, &pernet->sched_list, list)
+		list_del_rcu(&sched->list);
+	spin_unlock(&pernet->lock);
+
+	synchronize_rcu();
+}
+
+static struct pernet_operations mptcp_sched_pernet_ops = {
+	.init = sched_init_net,
+	.exit = sched_exit_net,
+	.id = &sched_pernet_id,
+	.size = sizeof(struct sched_pernet),
+};
+
+void mptcp_sched_init(void)
+{
+	if (register_pernet_subsys(&mptcp_sched_pernet_ops) < 0)
+		panic("Failed to register MPTCP sched pernet subsystem.\n");
+}
Signature · 1 line
2.34.1

--
Mat Martineau
Intel

#19 lore raw
In reply to Geliang Tang
Quoted 11 lines · Geliang Tang

RESEND:
- rebased to export/20220324T054815.
- depends on: "add skc_to_mptcp_sock" v12.

v5:
- patch 1: define per-namespace sched_list (but only used init_net namespace. It is difficult to get 'net' in bpf_mptcp_sched_reg and bpf_mptcp_sched_unreg. I need some suggestions here.)
- patch 2: skip mptcp_sched_default in mptcp_unregister_scheduler.
- patch 8: add tests into mptcp.c, instead of bpf_tcp_ca.c.

I ran in to a build problem because CONFIG_BPF_SYSCALL was not set in my .config and net/mptcp/Kconfig did not capture any BPF dependencies.

I think it would be best to allow MPTCP use without BPF enabled, so a new CONFIG_MPTCP_BPF_SCHEDULER config option (with dependencies on MPTCP and BPF_SYSCALL) would be good.

--
Mat Martineau
Intel

#15 lore raw

Re: [PATCH RESEND mptcp-next v5 5/8] mptcp: add get_subflow wrapper

In reply to Geliang Tang
Quoted 67 lines · Geliang Tang

This patch defined a new wrapper mptcp_sched_get_subflow(), invoke get_subflow of msk->sched in it. Use the wrapper instead of using mptcp_subflow_get_send() directly.

---

net/mptcp/protocol.c | 6 +++---
net/mptcp/protocol.h | 1 +
net/mptcp/sched.c    | 7 +++++++
3 files changed, 11 insertions(+), 3 deletions(-)
net/mptcp/protocol.c +3 -3
index c7969b9b567e..241cfda822c0 100644
@@ -1567,7 +1567,7 @@ void __mptcp_push_pending(struct sock *sk, unsigned int flags)
 			int ret = 0;
 
 			prev_ssk = ssk;
-			ssk = mptcp_subflow_get_send(msk);
+			ssk = mptcp_sched_get_subflow(msk);
 
 			/* First check. If the ssk has changed since
 			 * the last round, release prev_ssk
@@ -1636,7 +1636,7 @@ static void __mptcp_subflow_push_pending(struct sock *sk, struct sock *ssk)
 			 * check for a different subflow usage only after
 			 * spooling the first chunk of data
 			 */
-			xmit_ssk = first ? ssk : mptcp_subflow_get_send(mptcp_sk(sk));
+			xmit_ssk = first ? ssk : mptcp_sched_get_subflow(mptcp_sk(sk));
 			if (!xmit_ssk)
 				goto out;
 			if (xmit_ssk != ssk) {
@@ -3134,7 +3134,7 @@ void __mptcp_check_push(struct sock *sk, struct sock *ssk)
 		return;
 
 	if (!sock_owned_by_user(sk)) {
-		struct sock *xmit_ssk = mptcp_subflow_get_send(mptcp_sk(sk));
+		struct sock *xmit_ssk = mptcp_sched_get_subflow(mptcp_sk(sk));
 
 		if (xmit_ssk == ssk)
 			__mptcp_subflow_push_pending(sk, ssk);
net/mptcp/protocol.h +1 -0
index ad45281e8562..7781eb527b9a 100644
@@ -621,6 +621,7 @@ void mptcp_sched_init(void);
 void mptcp_init_sched(struct mptcp_sock *msk,
 		      struct mptcp_sched_ops *sched);
 void mptcp_release_sched(struct mptcp_sock *msk);
+struct sock *mptcp_sched_get_subflow(struct mptcp_sock *msk);
 
 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
 {
net/mptcp/sched.c +1 -0
index e431b1eec5e1..2fd2a034ba49 100644
@@ -11,6 +11,7 @@
 #include <linux/list.h>
 #include <linux/rculist.h>
 #include <linux/spinlock.h>
+#include <linux/indirect_call_wrapper.h>
 #include <net/tcp.h>
 #include <net/netns/generic.h>
 #include "protocol.h"

net/mptcp/sched.c void mptcp_release_sched(struct mptcp_sock *msk) · line 152

Quoted 9 lines
net/mptcp/sched.c +6 -0
@@ -143,3 +144,9 @@ void mptcp_release_sched(struct mptcp_sock *msk)
 		msk->sched->release(msk);
 	msk->sched = NULL;
 }
+
+struct sock *mptcp_sched_get_subflow(struct mptcp_sock *msk)
+{
+	return INDIRECT_CALL_INET_1(msk->sched->get_subflow,
+				    mptcp_subflow_get_send, msk);
+}

Since the purpose of this is to limit function call overhead, this could be implemented as a static inline function in protocol.h

--
Mat Martineau
Intel

#8 lore raw

Re: [PATCH RESEND mptcp-next v5 1/8] mptcp: add struct mptcp_sched_ops

In reply to Mat Martineau
Quoted 143 lines · Mat Martineau
Quoted 42 lines · Geliang Tang

This patch added struct mptcp_sched_ops. And define the scheduler register, unregister and find functions.

---

include/net/mptcp.h  |  13 +++++
net/mptcp/Makefile   |   2 +-
net/mptcp/protocol.c |   1 +
net/mptcp/protocol.h |   7 +++
net/mptcp/sched.c    | 114 +++++++++++++++++++++++++++++++++++++++++++
5 files changed, 136 insertions(+), 1 deletion(-)

create mode 100644 net/mptcp/sched.c

include/net/mptcp.h +13 -0
index 8b1afd6f5cc4..e3a0baa8dbd7 100644
@@ -95,6 +95,19 @@ struct mptcp_out_options {
 #endif
 };
 
+#define MPTCP_SCHED_NAME_MAX 16
+
+struct mptcp_sched_ops {
+	struct sock *	(*get_subflow)(struct mptcp_sock *msk);
+
+	char			name[MPTCP_SCHED_NAME_MAX];
+	struct module		*owner;
+	struct list_head	list;
+
+	void (*init)(struct mptcp_sock *msk);
+	void (*release)(struct mptcp_sock *msk);
+} ____cacheline_aligned_in_smp;
+
 #ifdef CONFIG_MPTCP
 extern struct request_sock_ops mptcp_subflow_request_sock_ops;
 
net/mptcp/Makefile +0 -0
index 0a0608b6b4b4..aa5c10d1b80a 100644
@@ -3,7 +3,7 @@ obj-$(CONFIG_MPTCP) += mptcp.o
Quoted 1 line

ccflags-y += -DDEBUG

This line of context keeps making 'git am' fail - looks like you have a debug-enabling commit in your private tree?

This is probably why patchew is having trouble with this series too.

- Mat

Quoted 91 lines

mptcp-y := protocol.o subflow.o options.o token.o crypto.o ctrl.o pm.o diag.o \

net/mptcp/Makefile +1 -1
-	   mib.o pm_netlink.o sockopt.o
+	   mib.o pm_netlink.o sockopt.o sched.o

obj-$(CONFIG_SYN_COOKIES) += syncookies.o
obj-$(CONFIG_INET_MPTCP_DIAG) += mptcp_diag.o

net/mptcp/protocol.c +1 -0
index d3887f628b54..b1d7c8b0c112 100644
@@ -3807,6 +3807,7 @@ void __init mptcp_proto_init(void)
 
 	mptcp_subflow_init();
 	mptcp_pm_init();
+	mptcp_sched_init();
 	mptcp_token_init();
 
 	if (proto_register(&mptcp_prot, MPTCP_USE_SLAB) != 0)
net/mptcp/protocol.h +7 -0
index fd82fd113113..3258b740c8ee 100644
@@ -608,6 +608,13 @@ int mptcp_subflow_create_socket(struct sock *sk, 
 struct socket **new_sock);
 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
 			 struct sockaddr_storage *addr,
 			 unsigned short family);
+struct mptcp_sched_ops *mptcp_sched_find(const struct net *net,
+					 const char *name);
+int mptcp_register_scheduler(const struct net *net,
+			     struct mptcp_sched_ops *sched);
+void mptcp_unregister_scheduler(const struct net *net,
+				struct mptcp_sched_ops *sched);
+void mptcp_sched_init(void);
 
 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)

{

net/mptcp/sched.c new file +47 -0
new file mode 100644
index 000000000000..3798a5cefeb6
@@ -0,0 +1,114 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Multipath TCP
+ *
+ * Copyright (c) 2022, SUSE.
+ */
+
+#define pr_fmt(fmt) "MPTCP: " fmt
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/list.h>
+#include <linux/rculist.h>
+#include <linux/spinlock.h>
+#include <net/tcp.h>
+#include <net/netns/generic.h>
+#include "protocol.h"
+
+static int sched_pernet_id;
+
+struct sched_pernet {
+	/* protects pernet updates */
+	spinlock_t		lock;
+	struct list_head	sched_list;
+};
+
+static struct sched_pernet *sched_get_pernet(const struct net *net)
+{
+	return net_generic(net, sched_pernet_id);
+}
+
+struct mptcp_sched_ops *mptcp_sched_find(const struct net *net,
+					 const char *name)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+	struct mptcp_sched_ops *sched, *ret = NULL;
+
+	spin_lock(&pernet->lock);
+	list_for_each_entry_rcu(sched, &pernet->sched_list, list) {
+		if (!strcmp(sched->name, name)) {
+			ret = sched;
+			break;
+		}
+	}
+	spin_unlock(&pernet->lock);
+
+	return ret;
+}
Quoted 8 lines

net/mptcp/sched.c line 55

Quoted 8 lines
net/mptcp/sched.c new file +8 -0
+
+int mptcp_register_scheduler(const struct net *net,
+			     struct mptcp_sched_ops *sched)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+
+	if (!sched->get_subflow)
+		return -EINVAL;

Should also validate sched->name, don't want to allow a NULL pointer there.

Quoted 51 lines
Quoted 51 lines
net/mptcp/sched.c new file +51 -0
+
+	if (mptcp_sched_find(net, sched->name))
+		return -EEXIST;
+
+	spin_lock(&pernet->lock);
+	list_add_tail_rcu(&sched->list, &pernet->sched_list);
+	spin_unlock(&pernet->lock);
+
+	pr_debug("%s registered", sched->name);
+	return 0;
+}
+
+void mptcp_unregister_scheduler(const struct net *net,
+				struct mptcp_sched_ops *sched)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+
+	spin_lock(&pernet->lock);
+	list_del_rcu(&sched->list);
+	spin_unlock(&pernet->lock);
+
+	synchronize_rcu();
+}
+
+static int __net_init sched_init_net(struct net *net)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+
+	INIT_LIST_HEAD_RCU(&pernet->sched_list);
+	spin_lock_init(&pernet->lock);
+
+	return 0;
+}
+
+static void __net_exit sched_exit_net(struct net *net)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+	struct mptcp_sched_ops *sched;
+
+	spin_lock(&pernet->lock);
+	list_for_each_entry_rcu(sched, &pernet->sched_list, list)
+		list_del_rcu(&sched->list);
+	spin_unlock(&pernet->lock);
+
+	synchronize_rcu();
+}
+
+static struct pernet_operations mptcp_sched_pernet_ops = {
+	.init = sched_init_net,
+	.exit = sched_exit_net,
+	.id = &sched_pernet_id,
Quoted 8 lines

net/mptcp/sched.c line 114

Quoted 8 lines
net/mptcp/sched.c new file +8 -0
+	.size = sizeof(struct sched_pernet),
+};
+
+void mptcp_sched_init(void)
+{
+	if (register_pernet_subsys(&mptcp_sched_pernet_ops) < 0)
+		panic("Failed to register MPTCP sched pernet subsystem.\n");
+}

--
Mat Martineau
Intel

#9 lore raw

Re: [PATCH RESEND mptcp-next v5 1/8] mptcp: add struct mptcp_sched_ops

Quoted 147 lines · Mat Martineau
Quoted 145 lines · Mat Martineau
Quoted 42 lines · Geliang Tang

This patch added struct mptcp_sched_ops. And define the scheduler register, unregister and find functions.

---

include/net/mptcp.h  |  13 +++++
net/mptcp/Makefile   |   2 +-
net/mptcp/protocol.c |   1 +
net/mptcp/protocol.h |   7 +++
net/mptcp/sched.c    | 114 +++++++++++++++++++++++++++++++++++++++++++
5 files changed, 136 insertions(+), 1 deletion(-)

create mode 100644 net/mptcp/sched.c

include/net/mptcp.h +13 -0
index 8b1afd6f5cc4..e3a0baa8dbd7 100644
@@ -95,6 +95,19 @@ struct mptcp_out_options {
 #endif
 };
 
+#define MPTCP_SCHED_NAME_MAX 16
+
+struct mptcp_sched_ops {
+	struct sock *	(*get_subflow)(struct mptcp_sock *msk);
+
+	char			name[MPTCP_SCHED_NAME_MAX];
+	struct module		*owner;
+	struct list_head	list;
+
+	void (*init)(struct mptcp_sock *msk);
+	void (*release)(struct mptcp_sock *msk);
+} ____cacheline_aligned_in_smp;
+
 #ifdef CONFIG_MPTCP
 extern struct request_sock_ops mptcp_subflow_request_sock_ops;
 
net/mptcp/Makefile +0 -0
index 0a0608b6b4b4..aa5c10d1b80a 100644
@@ -3,7 +3,7 @@ obj-$(CONFIG_MPTCP) += mptcp.o
Quoted 1 line

ccflags-y += -DDEBUG

This line of context keeps making 'git am' fail - looks like you have a debug-enabling commit in your private tree?

This is probably why patchew is having trouble with this series too.

- Mat

Quoted 93 lines

mptcp-y := protocol.o subflow.o options.o token.o crypto.o ctrl.o pm.o diag.o \

net/mptcp/Makefile +1 -1
-	   mib.o pm_netlink.o sockopt.o
+	   mib.o pm_netlink.o sockopt.o sched.o

obj-$(CONFIG_SYN_COOKIES) += syncookies.o
obj-$(CONFIG_INET_MPTCP_DIAG) += mptcp_diag.o

net/mptcp/protocol.c +1 -0
index d3887f628b54..b1d7c8b0c112 100644
@@ -3807,6 +3807,7 @@ void __init mptcp_proto_init(void)
 
 	mptcp_subflow_init();
 	mptcp_pm_init();
+	mptcp_sched_init();
 	mptcp_token_init();
 
 	if (proto_register(&mptcp_prot, MPTCP_USE_SLAB) != 0)
net/mptcp/protocol.h +7 -0
index fd82fd113113..3258b740c8ee 100644
@@ -608,6 +608,13 @@ int mptcp_subflow_create_socket(struct sock *sk, 
 struct socket **new_sock);
 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
 			 struct sockaddr_storage *addr,
 			 unsigned short family);
+struct mptcp_sched_ops *mptcp_sched_find(const struct net *net,
+					 const char *name);
+int mptcp_register_scheduler(const struct net *net,
+			     struct mptcp_sched_ops *sched);
+void mptcp_unregister_scheduler(const struct net *net,
+				struct mptcp_sched_ops *sched);
+void mptcp_sched_init(void);
 
 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)

{

net/mptcp/sched.c new file +49 -0
new file mode 100644
index 000000000000..3798a5cefeb6
@@ -0,0 +1,114 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Multipath TCP
+ *
+ * Copyright (c) 2022, SUSE.
+ */
+
+#define pr_fmt(fmt) "MPTCP: " fmt
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/list.h>
+#include <linux/rculist.h>
+#include <linux/spinlock.h>
+#include <net/tcp.h>
+#include <net/netns/generic.h>
+#include "protocol.h"
+
+static int sched_pernet_id;
+
+struct sched_pernet {
+	/* protects pernet updates */
+	spinlock_t		lock;
+	struct list_head	sched_list;
+};
+
+static struct sched_pernet *sched_get_pernet(const struct net *net)
+{
+	return net_generic(net, sched_pernet_id);
+}
+
+struct mptcp_sched_ops *mptcp_sched_find(const struct net *net,
+					 const char *name)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+	struct mptcp_sched_ops *sched, *ret = NULL;
+
+	spin_lock(&pernet->lock);
+	list_for_each_entry_rcu(sched, &pernet->sched_list, list) {
+		if (!strcmp(sched->name, name)) {
+			ret = sched;
+			break;
+		}
+	}
+	spin_unlock(&pernet->lock);
+
+	return ret;
+}
+
+int mptcp_register_scheduler(const struct net *net,
Quoted 8 lines
Quoted 6 lines

net/mptcp/sched.c line 55

Quoted 6 lines
net/mptcp/sched.c new file +6 -0
+			     struct mptcp_sched_ops *sched)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+
+	if (!sched->get_subflow)
+		return -EINVAL;

Should also validate sched->name, don't want to allow a NULL pointer there.

Oops, never mind - name is an array, not a pointer!

Quoted 55 lines
Quoted 55 lines
Quoted 55 lines
net/mptcp/sched.c new file +55 -0
+
+	if (mptcp_sched_find(net, sched->name))
+		return -EEXIST;
+
+	spin_lock(&pernet->lock);
+	list_add_tail_rcu(&sched->list, &pernet->sched_list);
+	spin_unlock(&pernet->lock);
+
+	pr_debug("%s registered", sched->name);
+	return 0;
+}
+
+void mptcp_unregister_scheduler(const struct net *net,
+				struct mptcp_sched_ops *sched)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+
+	spin_lock(&pernet->lock);
+	list_del_rcu(&sched->list);
+	spin_unlock(&pernet->lock);
+
+	synchronize_rcu();
+}
+
+static int __net_init sched_init_net(struct net *net)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+
+	INIT_LIST_HEAD_RCU(&pernet->sched_list);
+	spin_lock_init(&pernet->lock);
+
+	return 0;
+}
+
+static void __net_exit sched_exit_net(struct net *net)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+	struct mptcp_sched_ops *sched;
+
+	spin_lock(&pernet->lock);
+	list_for_each_entry_rcu(sched, &pernet->sched_list, list)
+		list_del_rcu(&sched->list);
+	spin_unlock(&pernet->lock);
+
+	synchronize_rcu();
+}
+
+static struct pernet_operations mptcp_sched_pernet_ops = {
+	.init = sched_init_net,
+	.exit = sched_exit_net,
+	.id = &sched_pernet_id,
+	.size = sizeof(struct sched_pernet),
+};
+
+void mptcp_sched_init(void)
Quoted 10 lines
Quoted 4 lines

net/mptcp/sched.c line 114

Quoted 4 lines
net/mptcp/sched.c new file +4 -0
+{
+	if (register_pernet_subsys(&mptcp_sched_pernet_ops) < 0)
+		panic("Failed to register MPTCP sched pernet subsystem.\n");
+}

--
Mat Martineau
Intel

--
Mat Martineau
Intel

#10 lore raw

Re: [PATCH RESEND mptcp-next v5 1/8] mptcp: add struct mptcp_sched_ops

Sorry Mat, I gave you the wrong feedback at yesterday's weekly meeting. BPF does work in different namespaces, see kernel/bpf/net_namespace.c and include/linux/bpf-netns.h. I think we should continue to use the pernet sched_list. v6 just sent out.

Thanks,
-Geliang

Quoted 233 lines · Mat Martineau
Quoted 150 lines · Mat Martineau
Quoted 148 lines · Mat Martineau
Quoted 42 lines · Geliang Tang

This patch added struct mptcp_sched_ops. And define the scheduler register, unregister and find functions.

---

include/net/mptcp.h  |  13 +++++
net/mptcp/Makefile   |   2 +-
net/mptcp/protocol.c |   1 +
net/mptcp/protocol.h |   7 +++
net/mptcp/sched.c    | 114 +++++++++++++++++++++++++++++++++++++++++++
5 files changed, 136 insertions(+), 1 deletion(-)

create mode 100644 net/mptcp/sched.c

include/net/mptcp.h +13 -0
index 8b1afd6f5cc4..e3a0baa8dbd7 100644
@@ -95,6 +95,19 @@ struct mptcp_out_options {
 #endif
 };
 
+#define MPTCP_SCHED_NAME_MAX 16
+
+struct mptcp_sched_ops {
+	struct sock *	(*get_subflow)(struct mptcp_sock *msk);
+
+	char			name[MPTCP_SCHED_NAME_MAX];
+	struct module		*owner;
+	struct list_head	list;
+
+	void (*init)(struct mptcp_sock *msk);
+	void (*release)(struct mptcp_sock *msk);
+} ____cacheline_aligned_in_smp;
+
 #ifdef CONFIG_MPTCP
 extern struct request_sock_ops mptcp_subflow_request_sock_ops;
 
net/mptcp/Makefile +0 -0
index 0a0608b6b4b4..aa5c10d1b80a 100644
@@ -3,7 +3,7 @@ obj-$(CONFIG_MPTCP) += mptcp.o
Quoted 1 line

ccflags-y += -DDEBUG

This line of context keeps making 'git am' fail - looks like you have a debug-enabling commit in your private tree?

This is probably why patchew is having trouble with this series too.

- Mat

Quoted 95 lines

mptcp-y := protocol.o subflow.o options.o token.o crypto.o ctrl.o pm.o diag.o \

net/mptcp/Makefile +1 -1
-	   mib.o pm_netlink.o sockopt.o
+	   mib.o pm_netlink.o sockopt.o sched.o

obj-$(CONFIG_SYN_COOKIES) += syncookies.o
obj-$(CONFIG_INET_MPTCP_DIAG) += mptcp_diag.o

net/mptcp/protocol.c +1 -0
index d3887f628b54..b1d7c8b0c112 100644
@@ -3807,6 +3807,7 @@ void __init mptcp_proto_init(void)
 
 	mptcp_subflow_init();
 	mptcp_pm_init();
+	mptcp_sched_init();
 	mptcp_token_init();
 
 	if (proto_register(&mptcp_prot, MPTCP_USE_SLAB) != 0)
net/mptcp/protocol.h +7 -0
index fd82fd113113..3258b740c8ee 100644
@@ -608,6 +608,13 @@ int mptcp_subflow_create_socket(struct sock
 *sk, struct socket **new_sock);
 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
 			 struct sockaddr_storage *addr,
 			 unsigned short family);
+struct mptcp_sched_ops *mptcp_sched_find(const struct net *net,
+					 const char *name);
+int mptcp_register_scheduler(const struct net *net,
+			     struct mptcp_sched_ops *sched);
+void mptcp_unregister_scheduler(const struct net *net,
+				struct mptcp_sched_ops *sched);
+void mptcp_sched_init(void);
 

static inline bool __mptcp_subflow_active(struct
mptcp_subflow_context *subflow)
{

net/mptcp/sched.c new file +49 -0
new file mode 100644
index 000000000000..3798a5cefeb6
@@ -0,0 +1,114 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Multipath TCP
+ *
+ * Copyright (c) 2022, SUSE.
+ */
+
+#define pr_fmt(fmt) "MPTCP: " fmt
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/list.h>
+#include <linux/rculist.h>
+#include <linux/spinlock.h>
+#include <net/tcp.h>
+#include <net/netns/generic.h>
+#include "protocol.h"
+
+static int sched_pernet_id;
+
+struct sched_pernet {
+	/* protects pernet updates */
+	spinlock_t		lock;
+	struct list_head	sched_list;
+};
+
+static struct sched_pernet *sched_get_pernet(const struct net *net)
+{
+	return net_generic(net, sched_pernet_id);
+}
+
+struct mptcp_sched_ops *mptcp_sched_find(const struct net *net,
+					 const char *name)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+	struct mptcp_sched_ops *sched, *ret = NULL;
+
+	spin_lock(&pernet->lock);
+	list_for_each_entry_rcu(sched, &pernet->sched_list, list) {
+		if (!strcmp(sched->name, name)) {
+			ret = sched;
+			break;
+		}
+	}
+	spin_unlock(&pernet->lock);
+
+	return ret;
+}
+
+int mptcp_register_scheduler(const struct net *net,
Quoted 8 lines
Quoted 6 lines

net/mptcp/sched.c line 55

Quoted 6 lines
net/mptcp/sched.c new file +6 -0
+			     struct mptcp_sched_ops *sched)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+
+	if (!sched->get_subflow)
+		return -EINVAL;

Should also validate sched->name, don't want to allow a NULL pointer there.

Oops, never mind - name is an array, not a pointer!

Quoted 55 lines
Quoted 55 lines
Quoted 55 lines
net/mptcp/sched.c new file +55 -0
+
+	if (mptcp_sched_find(net, sched->name))
+		return -EEXIST;
+
+	spin_lock(&pernet->lock);
+	list_add_tail_rcu(&sched->list, &pernet->sched_list);
+	spin_unlock(&pernet->lock);
+
+	pr_debug("%s registered", sched->name);
+	return 0;
+}
+
+void mptcp_unregister_scheduler(const struct net *net,
+				struct mptcp_sched_ops *sched)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+
+	spin_lock(&pernet->lock);
+	list_del_rcu(&sched->list);
+	spin_unlock(&pernet->lock);
+
+	synchronize_rcu();
+}
+
+static int __net_init sched_init_net(struct net *net)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+
+	INIT_LIST_HEAD_RCU(&pernet->sched_list);
+	spin_lock_init(&pernet->lock);
+
+	return 0;
+}
+
+static void __net_exit sched_exit_net(struct net *net)
+{
+	struct sched_pernet *pernet = sched_get_pernet(net);
+	struct mptcp_sched_ops *sched;
+
+	spin_lock(&pernet->lock);
+	list_for_each_entry_rcu(sched, &pernet->sched_list, list)
+		list_del_rcu(&sched->list);
+	spin_unlock(&pernet->lock);
+
+	synchronize_rcu();
+}
+
+static struct pernet_operations mptcp_sched_pernet_ops = {
+	.init = sched_init_net,
+	.exit = sched_exit_net,
+	.id = &sched_pernet_id,
+	.size = sizeof(struct sched_pernet),
+};
+
+void mptcp_sched_init(void)
Quoted 10 lines
Quoted 4 lines

net/mptcp/sched.c line 114

Quoted 4 lines
net/mptcp/sched.c new file +4 -0
+{
+	if (register_pernet_subsys(&mptcp_sched_pernet_ops) < 0)
+		panic("Failed to register MPTCP sched pernet subsystem.\n");
+}

--
Mat Martineau
Intel

--
Mat Martineau
Intel