Skip to contentSkip to conversation

BPF packet scheduler

PATCH Superseded Geliang Tang 6 participants 15 messages Apr 27, 2022

Subsystems (6)

v16 of 18 Superseded Latest is v18 →

lore
Share & tools
lore
#0 lore raw

v16 cover BPF packet scheduler

v16:
- add CONFIG_BPF_JIT check around the scheduler code inside bpf.c
- depends on export/20220426T102605 +
Squash to "bpf: add bpf_skc_to_mptcp_sock_proto" + Squash to "selftests: bpf: test bpf_skc_to_mptcp_sock" Squash to "selftests: bpf: add MPTCP test base"

v15:
- set mptcp_sched_id and mptcp_sched_type in bpf_mptcp_sched_init().
- add CONFIG_BPF_JIT in patch 6.
- drop '#include <sys/param.h>' in patch 8.
- depends on export/20220426T102605 +
Squash to "bpf: add bpf_skc_to_mptcp_sock_proto" + Squash to "selftests: bpf: test bpf_skc_to_mptcp_sock"

v14:
- add struct mptcp_sched_data.

base-commit: export/20220422T060337

v13:
- rename retrans to reinject
- drop "add last_snd write access" patch
- change %lu to %zu to fix the build break on i386.

base-commit: export/20220420T152103

v12:
- add call_me_again flag.

base-commit: export/20220419T055400

v11:
- add retrans argment for get_subflow()

base-commit: export/20220408T100826

v10:
- patch 5: keep msk->last_snd setting in get_subflow().
- patch 6: add bpf_mptcp_sched_btf_struct_access().
- patch 8: use MIN() in sys/param.h, instead of defining a new one.
- update commit logs.

base-commit: export/20220406T054706

v9:
- patch 2: add the missing mptcp_sched_init() invoking in mptcp_proto_init().
- patch 5: set last_snd after invoking get_subflow().
- patch 7: merge the squash-to patch.

v8:
- use global sched_list instead of pernet sched_list.
- drop synchronize_rcu() in mptcp_unregister_scheduler().
- update mptcp_init_sched and mptcp_release_sched as Mat and Florian suggested.
- fix the build break in patch 8.
- depends on: "add skc_to_mptcp_sock" v14.
- export/20220325T055307

v7:
- add bpf_try_module_get in mptcp_init_sched.
- add bpf_module_put in mptcp_release_sched.
- rename bpf_first to mptcp_bpf_first.
- update commit logs.

v6:
- still use pernet sched_list, use current->nsproxy->net_ns in BPF context instead of using init_net.
- patch 1:
- use rcu_read_lock instead of spin_lock in mptcp_sched_find as Florian suggested.
- drop synchronize_rcu in sched_exit_net as Florian suggested.
- keep synchronize_rcu in mptcp_unregister_scheduler, otherwise, got a workqueue lockup in my test.
- update Makefile as Mat suggested.
- patch 2:
- add mptcp_sched_data_init to register default sched, instead of registering it in init_net.
- patch 5:
- move mptcp_sched_get_subflow to protocol.h as Mat suggested.
- patch 6:
- use current->nsproxy->net_ns instead of init_net.
- patch 8:
- add send_data to send more data, instead of send_byte.

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                           |  19 +++
 kernel/bpf/bpf_struct_ops_types.h             |   4 +
 net/mptcp/Makefile                            |   2 +-
 net/mptcp/bpf.c                               | 151 ++++++++++++++++++
 net/mptcp/ctrl.c                              |  14 ++
 net/mptcp/protocol.c                          |  25 ++-
 net/mptcp/protocol.h                          |  27 ++++
 net/mptcp/sched.c                             | 104 ++++++++++++
 .../testing/selftests/bpf/bpf_mptcp_helpers.h |  19 +++
 .../testing/selftests/bpf/prog_tests/mptcp.c  | 112 +++++++++++++
 .../selftests/bpf/progs/mptcp_bpf_first.c     |  32 ++++
 12 files changed, 511 insertions(+), 6 deletions(-)
 create mode 100644 net/mptcp/sched.c
 create mode 100644 tools/testing/selftests/bpf/progs/mptcp_bpf_first.c
Signature · 1 line
2.34.1
#1 lore raw

PATCH v16 1/8 mptcp: add struct mptcp_sched_ops

This patch defines struct mptcp_sched_ops, which has three struct members, name, owner and list, and three function pointers, init, release and get_subflow.

Add the scheduler registering, unregistering and finding functions to add, delete and find a packet scheduler on the global list mptcp_sched_list.

For supporting a "redundant" packet scheduler in the future, this patch adds a flag in struct mptcp_sched_data named call_again to indicate that get_subflow() function needs to be called again.

---
 include/net/mptcp.h                           | 19 +++++++
 net/mptcp/Makefile                            |  2 +-
 net/mptcp/protocol.h                          |  3 +
 net/mptcp/sched.c                             | 56 +++++++++++++++++++
 .../testing/selftests/bpf/bpf_mptcp_helpers.h | 18 ++++++
 5 files changed, 97 insertions(+), 1 deletion(-)
 create mode 100644 net/mptcp/sched.c
include/net/mptcp.h +19 -0
index 4d761ad530c9..dd4ee7a77567 100644
@@ -95,6 +95,25 @@ struct mptcp_out_options {
 #endif
 };
 
+#define MPTCP_SCHED_NAME_MAX 16
+
+struct mptcp_sched_data {
+	struct sock	*sock;
+	bool		call_again;
+};
+
+struct mptcp_sched_ops {
+	void (*get_subflow)(struct mptcp_sock *msk, bool reinject,
+			    struct mptcp_sched_data *data);
+
+	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 6e7df47c9584..8a7f68efa35f 100644
@@ -2,7 +2,7 @@
 obj-$(CONFIG_MPTCP) += mptcp.o
 
 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 pm_userspace.o
+	   mib.o pm_netlink.o sockopt.o pm_userspace.o sched.o
 
 obj-$(CONFIG_SYN_COOKIES) += syncookies.o
 obj-$(CONFIG_INET_MPTCP_DIAG) += mptcp_diag.o
net/mptcp/protocol.h +3 -0
index 4672901d0dfe..221bb6b9860e 100644
@@ -621,6 +621,9 @@ 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 char *name);
+int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
+void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
 
 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
 {
net/mptcp/sched.c new file +56 -0

56 lines not on this page · raw message

tools/testing/selftests/bpf/bpf_mptcp_helpers.h +18 -0
index b5a43b108982..de20fbdb6d98 100644
@@ -6,6 +6,24 @@
 
 #include "bpf_tcp_helpers.h"
 
+#define MPTCP_SCHED_NAME_MAX 16
+
+struct mptcp_sched_data {
+	struct sock	*sock;
+	bool		call_again;
+};
+
+struct mptcp_sched_ops {
+	char name[MPTCP_SCHED_NAME_MAX];
+
+	void (*init)(struct mptcp_sock *msk);
+	void (*release)(struct mptcp_sock *msk);
+
+	void (*get_subflow)(struct mptcp_sock *msk, bool reinject,
+			    struct mptcp_sched_data *data);
+	void *owner;
+};
+
 struct mptcp_sock {
 	struct inet_connection_sock	sk;
 
Signature · 1 line
2.34.1
#2 lore raw

PATCH v16 2/8 mptcp: register default scheduler

In reply to Geliang Tang

This patch defines the default packet scheduler mptcp_sched_default, register it in mptcp_sched_init(), which is invoked in mptcp_proto_init(). Skip deleting this default scheduler in mptcp_unregister_scheduler().

---
 net/mptcp/protocol.c |  9 +++++++++
 net/mptcp/protocol.h |  3 +++
 net/mptcp/sched.c    | 14 ++++++++++++++
 3 files changed, 26 insertions(+)
net/mptcp/protocol.c +9 -0
index 653757ea0aca..bf9d5bd8dfa6 100644
@@ -2259,6 +2259,14 @@ static struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk)
 	return min_stale_count > 1 ? backup : NULL;
 }
 
+void mptcp_get_subflow_default(struct mptcp_sock *msk, bool reinject,
+			       struct mptcp_sched_data *data)
+{
+	data->sock = reinject ? mptcp_subflow_get_retrans(msk) :
+				mptcp_subflow_get_send(msk);
+	data->call_again = 0;
+}
+
 static void mptcp_dispose_initial_subflow(struct mptcp_sock *msk)
 {
 	if (msk->subflow) {
@@ -3813,6 +3821,7 @@ 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 +3 -0
index 221bb6b9860e..65a765c8d2be 100644
@@ -624,6 +624,9 @@ void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
 struct mptcp_sched_ops *mptcp_sched_find(const char *name);
 int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
+void mptcp_get_subflow_default(struct mptcp_sock *msk, bool reinject,
+			       struct mptcp_sched_data *data);
+void mptcp_sched_init(void);
 
 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
 {
net/mptcp/sched.c +14 -0
index c5d3bbafba71..bd0beff8cac8 100644
@@ -13,6 +13,12 @@
 #include <linux/spinlock.h>
 #include "protocol.h"
 
+static struct mptcp_sched_ops mptcp_sched_default = {
+	.get_subflow    = mptcp_get_subflow_default,
+	.name           = "default",
+	.owner          = THIS_MODULE,
+};
+
 static DEFINE_SPINLOCK(mptcp_sched_list_lock);
 static LIST_HEAD(mptcp_sched_list);
 
@@ -50,7 +56,15 @@ int mptcp_register_scheduler(struct mptcp_sched_ops *sched)
 
 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched)
 {
+	if (sched == &mptcp_sched_default)
+		return;
+
 	spin_lock(&mptcp_sched_list_lock);
 	list_del_rcu(&sched->list);
 	spin_unlock(&mptcp_sched_list_lock);
 }
+
+void mptcp_sched_init(void)
+{
+	mptcp_register_scheduler(&mptcp_sched_default);
+}
Signature · 1 line
2.34.1
#3 lore raw

PATCH v16 3/8 mptcp: add a new sysctl scheduler

In reply to Geliang Tang

This patch adds a new sysctl, named scheduler, to support for selection of different schedulers. Export mptcp_get_scheduler helper to get this sysctl.

---
 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 65a765c8d2be..4842a28f34c2 100644
@@ -596,6 +596,7 @@ 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
#4 lore raw

PATCH v16 4/8 mptcp: add sched in mptcp_sock

In reply to Geliang Tang

This patch adds 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                          |  7 ++++
 net/mptcp/protocol.h                          |  4 +++
 net/mptcp/sched.c                             | 34 +++++++++++++++++++
 .../testing/selftests/bpf/bpf_mptcp_helpers.h |  1 +
 4 files changed, 46 insertions(+)
net/mptcp/protocol.c +7 -0
index bf9d5bd8dfa6..7590e2d29f39 100644
@@ -2673,6 +2673,11 @@ static int mptcp_init_sock(struct sock *sk)
 	if (ret)
 		return ret;
 
+	ret = mptcp_init_sched(mptcp_sk(sk),
+			       mptcp_sched_find(mptcp_get_scheduler(net)));
+	if (ret)
+		return ret;
+
 	/* fetch the ca name; do it outside __mptcp_init_sock(), so that clone will
 	 * propagate the correct value
 	 */
@@ -2832,6 +2837,7 @@ 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
@@ -3009,6 +3015,7 @@ 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 4842a28f34c2..22f3f41e1e32 100644
@@ -298,6 +298,7 @@ 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 */
@@ -628,6 +629,9 @@ void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
 void mptcp_get_subflow_default(struct mptcp_sock *msk, bool reinject,
 			       struct mptcp_sched_data *data);
 void mptcp_sched_init(void);
+int 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 +34 -0
index bd0beff8cac8..8025dc51fbe9 100644
@@ -68,3 +68,37 @@ void mptcp_sched_init(void)
 {
 	mptcp_register_scheduler(&mptcp_sched_default);
 }
+
+int mptcp_init_sched(struct mptcp_sock *msk,
+		     struct mptcp_sched_ops *sched)
+{
+	struct mptcp_sched_ops *sched_init = &mptcp_sched_default;
+
+	if (sched)
+		sched_init = sched;
+
+	if (!bpf_try_module_get(sched_init, sched_init->owner))
+		return -EBUSY;
+
+	msk->sched = sched_init;
+	if (msk->sched->init)
+		msk->sched->init(msk);
+
+	pr_debug("sched=%s", msk->sched->name);
+
+	return 0;
+}
+
+void mptcp_release_sched(struct mptcp_sock *msk)
+{
+	struct mptcp_sched_ops *sched = msk->sched;
+
+	if (!sched)
+		return;
+
+	msk->sched = NULL;
+	if (sched->release)
+		sched->release(msk);
+
+	bpf_module_put(sched, sched->owner);
+}
tools/testing/selftests/bpf/bpf_mptcp_helpers.h +1 -0
index de20fbdb6d98..a0b83fbe8133 100644
@@ -29,6 +29,7 @@ struct mptcp_sock {
 
 	__u32		token;
 	struct sock	*first;
+	struct mptcp_sched_ops *sched;
 	char		ca_name[TCP_CA_NAME_MAX];
 } __attribute__((preserve_access_index));
 
Signature · 1 line
2.34.1
#5 lore raw

PATCH v16 5/8 mptcp: add get_subflow wrapper

In reply to Geliang Tang

This patch defines 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 |  9 ++++-----
 net/mptcp/protocol.h | 16 ++++++++++++++++
 2 files changed, 20 insertions(+), 5 deletions(-)
net/mptcp/protocol.c +4 -5
index 7590e2d29f39..c6e963848b18 100644
@@ -1515,7 +1515,6 @@ static struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk)
 	subflow->avg_pacing_rate = div_u64((u64)subflow->avg_pacing_rate * wmem +
 					   READ_ONCE(ssk->sk_pacing_rate) * burst,
 					   burst + wmem);
-	msk->last_snd = ssk;
 	msk->snd_burst = burst;
 	return ssk;
 }
@@ -1575,7 +1574,7 @@ 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, false);
 
 			/* First check. If the ssk has changed since
 			 * the last round, release prev_ssk
@@ -1644,7 +1643,7 @@ 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), false);
 			if (!xmit_ssk)
 				goto out;
 			if (xmit_ssk != ssk) {
@@ -2489,7 +2488,7 @@ static void __mptcp_retrans(struct sock *sk)
 	mptcp_clean_una_wakeup(sk);
 
 	/* first check ssk: need to kick "stale" logic */
-	ssk = mptcp_subflow_get_retrans(msk);
+	ssk = mptcp_sched_get_subflow(msk, true);
 	dfrag = mptcp_rtx_head(sk);
 	if (!dfrag) {
 		if (mptcp_data_fin_enabled(msk)) {
@@ -3154,7 +3153,7 @@ 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), false);
 
 		if (xmit_ssk == ssk)
 			__mptcp_subflow_push_pending(sk, ssk);
net/mptcp/protocol.h +16 -0
index 22f3f41e1e32..91512fc25128 100644
@@ -633,6 +633,22 @@ int mptcp_init_sched(struct mptcp_sock *msk,
 		     struct mptcp_sched_ops *sched);
 void mptcp_release_sched(struct mptcp_sock *msk);
 
+static inline struct sock *mptcp_sched_get_subflow(struct mptcp_sock *msk, bool reinject)
+{
+	struct mptcp_sched_data data = {
+		.sock		= msk->first,
+		.call_again	= 0,
+	};
+
+	msk->sched ? INDIRECT_CALL_INET_1(msk->sched->get_subflow,
+					  mptcp_get_subflow_default,
+					  msk, reinject, &data) :
+		     mptcp_get_subflow_default(msk, reinject, &data);
+
+	msk->last_snd = data.sock;
+	return data.sock;
+}
+
 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
 {
 	struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
Signature · 1 line
2.34.1
#10 lore raw

PATCH v16 6/8 mptcp: add bpf_mptcp_sched_ops

In reply to Geliang Tang

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

This MPTCP BPF scheduler implementation is similar to BPF TCP CC. And net/ipv4/bpf_tcp_ca.c is a frame of reference for this patch.

---
 kernel/bpf/bpf_struct_ops_types.h |   4 +
 net/mptcp/bpf.c                   | 151 ++++++++++++++++++++++++++++++
 2 files changed, 155 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 +151 -0

159 lines not on this page · raw message

Signature · 1 line
2.34.1
#12 lore raw

PATCH v16 7/8 selftests: bpf: add bpf_first scheduler

In reply to Geliang Tang

This patch implements the simplest MPTCP scheduler, named bpf_first, which always picks the first subflow to send data. It's a sample of MPTCP BPF scheduler implementations.

---
 .../selftests/bpf/progs/mptcp_bpf_first.c     | 32 +++++++++++++++++++
 1 file changed, 32 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/progs/mptcp_bpf_first.c
tools/testing/selftests/bpf/progs/mptcp_bpf_first.c new file +32 -0
new file mode 100644
index 000000000000..0ca9754c078d
@@ -0,0 +1,32 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2022, SUSE. */
+
+#include <linux/bpf.h>
+#include "bpf_mptcp_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)
+{
+}
+
+void BPF_STRUCT_OPS(bpf_first_get_subflow, struct mptcp_sock *msk,
+		    bool reinject, struct mptcp_sched_data *data)
+{
+	data->sock = msk->first;
+	data->call_again = 0;
+}
+
+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
#13 lore raw

PATCH v16 8/8 selftests: bpf: add bpf_first test

In reply to Geliang Tang

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

Some code in send_data() is from prog_tests/bpf_tcp_ca.c.

---
 .../testing/selftests/bpf/prog_tests/mptcp.c  | 112 ++++++++++++++++++
 1 file changed, 112 insertions(+)
tools/testing/selftests/bpf/prog_tests/mptcp.c +112 -0
index 7e704f5aab05..44484a63e62a 100644
@@ -4,6 +4,7 @@
 #include <test_progs.h>
 #include "cgroup_helpers.h"
 #include "network_helpers.h"
+#include "mptcp_bpf_first.skel.h"
 
 #ifndef TCP_CA_NAME_MAX
 #define TCP_CA_NAME_MAX	16
@@ -19,6 +20,8 @@ struct mptcp_storage {
 };
 
 static char monitor_log_path[64];
+static const unsigned int total_bytes = 10 * 1024 * 1024;
+static int stop, duration;
 
 static int verify_tsk(int map_fd, int client_fd)
 {
@@ -251,8 +254,117 @@ void test_base(void)
 	close(cgroup_fd);
 }
 
+static void *server(void *arg)
+{
+	int lfd = (int)(long)arg, err = 0, fd;
+	ssize_t nr_sent = 0, bytes = 0;
+	char batch[1500];
+
+	fd = accept(lfd, NULL, NULL);
+	while (fd == -1) {
+		if (errno == EINTR)
+			continue;
+		err = -errno;
+		goto done;
+	}
+
+	if (settimeo(fd, 0)) {
+		err = -errno;
+		goto done;
+	}
+
+	while (bytes < total_bytes && !READ_ONCE(stop)) {
+		nr_sent = send(fd, &batch,
+			       MIN(total_bytes - bytes, sizeof(batch)), 0);
+		if (nr_sent == -1 && errno == EINTR)
+			continue;
+		if (nr_sent == -1) {
+			err = -errno;
+			break;
+		}
+		bytes += nr_sent;
+	}
+
+	CHECK(bytes != total_bytes, "send", "%zd != %u nr_sent:%zd errno:%d\n",
+	      bytes, total_bytes, nr_sent, errno);
+
+done:
+	if (fd >= 0)
+		close(fd);
+	if (err) {
+		WRITE_ONCE(stop, 1);
+		return ERR_PTR(err);
+	}
+	return NULL;
+}
+
+static void send_data(int lfd, int fd)
+{
+	ssize_t nr_recv = 0, bytes = 0;
+	pthread_t srv_thread;
+	void *thread_ret;
+	char batch[1500];
+	int err;
+
+	WRITE_ONCE(stop, 0);
+
+	err = pthread_create(&srv_thread, NULL, server, (void *)(long)lfd);
+	if (CHECK(err != 0, "pthread_create", "err:%d errno:%d\n", err, errno))
+		return;
+
+	/* recv total_bytes */
+	while (bytes < total_bytes && !READ_ONCE(stop)) {
+		nr_recv = recv(fd, &batch,
+			       MIN(total_bytes - bytes, sizeof(batch)), 0);
+		if (nr_recv == -1 && errno == EINTR)
+			continue;
+		if (nr_recv == -1)
+			break;
+		bytes += nr_recv;
+	}
+
+	CHECK(bytes != total_bytes, "recv", "%zd != %u nr_recv:%zd errno:%d\n",
+	      bytes, total_bytes, nr_recv, errno);
+
+	WRITE_ONCE(stop, 1);
+
+	pthread_join(srv_thread, &thread_ret);
+	CHECK(IS_ERR(thread_ret), "pthread_join", "thread_ret:%ld",
+	      PTR_ERR(thread_ret));
+}
+
+static void test_first(void)
+{
+	struct mptcp_bpf_first *first_skel;
+	int server_fd, client_fd;
+	struct bpf_link *link;
+
+	first_skel = mptcp_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")) {
+		mptcp_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_data(server_fd, client_fd);
+
+	close(client_fd);
+	close(server_fd);
+	bpf_link__destroy(link);
+	mptcp_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
#14 lore raw

Re: selftests: bpf: add bpf_first test: Tests Results

Hi Geliang,

Thank you for your modifications, that's great!

Our CI did some validations and here is its report:

- KVM Validation: normal:
- Success! ✅:
- Task: https://cirrus-ci.com/task/5378906021691392
- Summary: https://api.cirrus-ci.com/v1/artifact/task/5378906021691392/summary/summary.txt

- KVM Validation: debug:
- Unstable: 2 failed test(s): selftest_diag selftest_mptcp_join 🔴:
- Task: https://cirrus-ci.com/task/6504805928534016
- Summary: https://api.cirrus-ci.com/v1/artifact/task/6504805928534016/summary/summary.txt

Initiator: Patchew Applier
Commits: https://github.com/multipath-tcp/mptcp_net-next/commits/ffe30335bc16

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-debug

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 (Tessares)

maintainer
#6 lore raw

Re: [PATCH mptcp-next v16 5/8] mptcp: add get_subflow wrapper

In reply to Geliang Tang

Hello,

First of all I'm sorry for the very late feedback: I had some difficulties to allocate time to follow this development.

Quoted 15 lines · Geliang Tang

This patch defines 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 |  9 ++++-----
 net/mptcp/protocol.h | 16 ++++++++++++++++
 2 files changed, 20 insertions(+), 5 deletions(-)
net/mptcp/protocol.c +0 -0
index 7590e2d29f39..c6e963848b18 100644

net/mptcp/protocol.c static struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk) · line 1520

Quoted 7 lines
net/mptcp/protocol.c +0 -1
@@ -1515,7 +1515,6 @@ static struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk)
 	subflow->avg_pacing_rate = div_u64((u64)subflow->avg_pacing_rate * wmem +
 					   READ_ONCE(ssk->sk_pacing_rate) * burst,
 					   burst + wmem);
-	msk->last_snd = ssk;
 	msk->snd_burst = burst;
 	return ssk;
 }

This breaks "mptcp: really share subflow snd_wnd": When the MPTCP-level cwin is 0, we want to do 0-window probe (so mptcp_subflow_get_send() return a 'valid' subflow) but we don't want to start a new burst for such 0 win probe (so mptcp_subflow_get_send() clears last_snd).

With this change, when MPTCP-level cwin is 0, 'last_send' is not cleared anymore.

A simple sulution would be:

---

net/mptcp/protocol.c +1 -2
index 653757ea0aca..c20f5fd04cad 100644
@@ -1507,7 +1507,7 @@ static struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk)
        burst = min_t(int, MPTCP_SEND_BURST_SIZE, mptcp_wnd_end(msk) - msk->snd_nxt);
        wmem = READ_ONCE(ssk->sk_wmem_queued);
        if (!burst) {
-               msk->last_snd = NULL;
+               msk->snd_burst = 0;
                return ssk;
        }
---

Probably a better solution would be adding 'snd_burst' to struct mptcp_sched_data, and let mptcp_sched_get_subflow() update msk-

Quoted 1 line

snd_burst as specified by the scheduler. With this latter change the

'msk' argument in mptcp_sched_ops->schedule() could be const.

Quoted 41 lines
net/mptcp/protocol.c +4 -4
@@ -1575,7 +1574,7 @@ 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, false);
 
 			/* First check. If the ssk has changed since
 			 * the last round, release prev_ssk
@@ -1644,7 +1643,7 @@ 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), false);
 			if (!xmit_ssk)
 				goto out;
 			if (xmit_ssk != ssk) {
@@ -2489,7 +2488,7 @@ static void __mptcp_retrans(struct sock *sk)
 	mptcp_clean_una_wakeup(sk);
 
 	/* first check ssk: need to kick "stale" logic */
-	ssk = mptcp_subflow_get_retrans(msk);
+	ssk = mptcp_sched_get_subflow(msk, true);
 	dfrag = mptcp_rtx_head(sk);
 	if (!dfrag) {
 		if (mptcp_data_fin_enabled(msk)) {
@@ -3154,7 +3153,7 @@ 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), false);
 
 		if (xmit_ssk == ssk)
 			__mptcp_subflow_push_pending(sk, ssk);
net/mptcp/protocol.h +0 -0
index 22f3f41e1e32..91512fc25128 100644

net/mptcp/protocol.h int mptcp_init_sched(struct mptcp_sock *msk, · line 641

Quoted 9 lines
net/mptcp/protocol.h +6 -0
@@ -633,6 +633,22 @@ int mptcp_init_sched(struct mptcp_sock *msk,
 		     struct mptcp_sched_ops *sched);
 void mptcp_release_sched(struct mptcp_sock *msk);
 
+static inline struct sock *mptcp_sched_get_subflow(struct mptcp_sock *msk, bool reinject)
+{
+	struct mptcp_sched_data data = {
+		.sock		= msk->first,
+		.call_again	= 0,
+	};

It's not clear to me:
- why we need the 'sock' argument, since the scheduler already get 'msk' as argument?
- why we need to initialize it (looks like an 'output' argument ?!?)
- what is the goal/role of the 'call_again' argument. It looks like we just ignore it?!?

Side note: perhaps we should move the following chunk:

---

       sock_owned_by_me(sk);

        if (__mptcp_check_fallback(msk)) {
                if (!msk->first)
                        return NULL;
                return sk_stream_memory_free(msk->first) ? msk->first : NULL;
        }

---

out of mptcp_subflow_get_send() into mptcp_sched_get_subflow(), so
that every scheduler will not have to deal with fallback sockets.

net/mptcp/protocol.h int mptcp_init_sched(struct mptcp_sock *msk, · line 646

Quoted 5 lines
net/mptcp/protocol.h +5 -0
+
+	msk->sched ? INDIRECT_CALL_INET_1(msk->sched->get_subflow,
+					  mptcp_get_subflow_default,
+					  msk, reinject, &data) :
+		     mptcp_get_subflow_default(msk, reinject, &data);

With this we have quite a lot of conditionals for the default scheduler. I think we can drop the INDIRECT_CALL_INET_1() wrapper, and rework the code so that msk->sched is always NULL with the default scheduler.

And sorry to bring the next topic so late, but why a 'reinject' argument instead of an additional mptcp_sched_ops op? the latter option will avoid another branch in fast-path.

Thanks!

Paolo

maintainer
#7 lore raw

Re: [PATCH mptcp-next v16 5/8] mptcp: add get_subflow wrapper

Quoted 10 lines · Paolo Abeni

Review

Quoted 5 lines
(quoted diff) +5 -0
+
+	msk->sched ? INDIRECT_CALL_INET_1(msk->sched->get_subflow,
+					  mptcp_get_subflow_default,
+					  msk, reinject, &data) :
+		     mptcp_get_subflow_default(msk, reinject, &data);

With this we have quite a lot of conditionals for the default scheduler. I think we can drop the INDIRECT_CALL_INET_1() wrapper, and rework the code so that msk->sched is always NULL with the default scheduler.

Even better, we could drop the mptcp_get_subflow_default() wrapper and call directly the current scheduler when msk->sched is NULL, even before filling the 'data' struct. That should fit nicely expecially if we use 2 separate ops for plain xmit and retrans.

Overall that would lead to something alike:

---
static inline struct sock *mptcp_sched_get_send(struct mptcp_sock msk)

{
	struct mptcp_sched_data data;

        sock_owned_by_me(sk);

	/* the following check is moved out of mptcp_subflow_get_send */
        if (__mptcp_check_fallback(msk)) {
                if (!msk->first)
                        return NULL;
                return sk_stream_memory_free(msk->first) ? msk->first : NULL;
        }

	if (!msk->sched)
		return mptcp_subflow_get_send(msk);

	data.sock = msk->first; /* if needed */
	data.call_again = 0;	/* if needed */
	msk->sched->get_send(msk, &data);
	msk->last_snd = data.sock;
	return data.sock;
}

---
plus something similar for retrans.

Thanks!

Paolo

#11 lore raw

Re: [PATCH mptcp-next v16 6/8] mptcp: add bpf_mptcp_sched_ops

In reply to Geliang Tang
Thank you for the patch! Perhaps something to improve:

Hi Geliang,

Thank you for the patch! Perhaps something to improve:

[auto build test WARNING on mptcp/export]
[cannot apply to bpf-next/master bpf/master linus/master v5.18-rc4 next-20220427] [If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]

url:    https://github.com/intel-lab-lkp/linux/commits/Geliang-Tang/BPF-packet-scheduler/20220427-095943
base:   https://github.com/multipath-tcp/mptcp_net-next.git export
config: i386-randconfig-m021 (https://download.01.org/0day-ci/archive/20220427/[email protected]/config)
compiler: gcc-11 (Debian 11.2.0-20) 11.2.0

If you fix the issue, kindly add following tag as appropriate

smatch warnings:
net/mptcp/bpf.c:148 bpf_mptcp_sched_init() warn: unsigned 'mptcp_sched_id' is never less than zero.

vim +/mptcp_sched_id +148 net/mptcp/bpf.c

143
144 static int bpf_mptcp_sched_init(struct btf *btf)
145 {
146 mptcp_sched_id = btf_find_by_name_kind(btf, "mptcp_sched_data",
147 BTF_KIND_STRUCT);

 > 148		if (mptcp_sched_id < 0)

149 return -EINVAL;
150 mptcp_sched_type = btf_type_by_id(btf, mptcp_sched_id);
151
152 return 0;
153 }
154

Signature · 2 lines
0-DAY CI Kernel Test Service
https://01.org/lkp
#8 lore raw

Re: [PATCH mptcp-next v16 5/8] mptcp: add get_subflow wrapper

In reply to Paolo Abeni

Hi Paolo,

Thanks for your review!

Quoted 46 lines · Paolo Abeni <[email protected]> 于2022年4月27日周三 16:27写道:

Hello,

First of all I'm sorry for the very late feedback: I had some difficulties to allocate time to follow this development.

Quoted 15 lines · Geliang Tang

This patch defines 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 |  9 ++++-----
 net/mptcp/protocol.h | 16 ++++++++++++++++
 2 files changed, 20 insertions(+), 5 deletions(-)
net/mptcp/protocol.c +0 -0
index 7590e2d29f39..c6e963848b18 100644

net/mptcp/protocol.c static struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk) · line 1520

Quoted 7 lines
net/mptcp/protocol.c +0 -1
@@ -1515,7 +1515,6 @@ static struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk)
      subflow->avg_pacing_rate = div_u64((u64)subflow->avg_pacing_rate * wmem +
                                         READ_ONCE(ssk->sk_pacing_rate) * burst,
                                         burst + wmem);
-     msk->last_snd = ssk;
      msk->snd_burst = burst;
      return ssk;
 }

This breaks "mptcp: really share subflow snd_wnd": When the MPTCP-level cwin is 0, we want to do 0-window probe (so mptcp_subflow_get_send() return a 'valid' subflow) but we don't want to start a new burst for such 0 win probe (so mptcp_subflow_get_send() clears last_snd).

With this change, when MPTCP-level cwin is 0, 'last_send' is not cleared anymore.

A simple sulution would be:

---

net/mptcp/protocol.c +0 -0
index 653757ea0aca..c20f5fd04cad 100644

net/mptcp/protocol.c static struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk) · line 1513

Quoted 8 lines
net/mptcp/protocol.c +1 -2
@@ -1507,7 +1507,7 @@ static struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk)
        burst = min_t(int, MPTCP_SEND_BURST_SIZE, mptcp_wnd_end(msk) - msk->snd_nxt);
        wmem = READ_ONCE(ssk->sk_wmem_queued);
        if (!burst) {
-               msk->last_snd = NULL;
+               msk->snd_burst = 0;
                return ssk;
        }
---

I think it's better to add back "msk->last_snd = ssk;" in mptcp_subflow_get_send(), keep the original mptcp_subflow_get_send() unchanged. And set msk->last_snd in the get_subflow() of the scheduler, the same as v12:

https://patchwork.kernel.org/project/mptcp/patch/396f8edcbf8e4f65a1d76c228dc3d12aa4dd2f11.1650430389.git.geliang.tang@suse.com/

Quoted 50 lines

Probably a better solution would be adding 'snd_burst' to struct mptcp_sched_data, and let mptcp_sched_get_subflow() update msk-

Quoted 1 line

snd_burst as specified by the scheduler. With this latter change the

'msk' argument in mptcp_sched_ops->schedule() could be const.

Quoted 44 lines
net/mptcp/protocol.c +4 -4
@@ -1575,7 +1574,7 @@ 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, false);
 
                      /* First check. If the ssk has changed since
                       * the last round, release prev_ssk
@@ -1644,7 +1643,7 @@ 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), false);
                      if (!xmit_ssk)
                              goto out;
                      if (xmit_ssk != ssk) {
@@ -2489,7 +2488,7 @@ static void __mptcp_retrans(struct sock *sk)
      mptcp_clean_una_wakeup(sk);
 
      /* first check ssk: need to kick "stale" logic */
-     ssk = mptcp_subflow_get_retrans(msk);
+     ssk = mptcp_sched_get_subflow(msk, true);
      dfrag = mptcp_rtx_head(sk);
      if (!dfrag) {
              if (mptcp_data_fin_enabled(msk)) {
@@ -3154,7 +3153,7 @@ 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), false);
 
              if (xmit_ssk == ssk)
                      __mptcp_subflow_push_pending(sk, ssk);
net/mptcp/protocol.h +0 -0
index 22f3f41e1e32..91512fc25128 100644
@@ -633,6 +633,22 @@ int mptcp_init_sched(struct mptcp_sock *msk,
                   struct mptcp_sched_ops *sched);
 void mptcp_release_sched(struct mptcp_sock *msk);
 
Quoted 10 lines

net/mptcp/protocol.h int mptcp_init_sched(struct mptcp_sock *msk, · line 641

Quoted 6 lines
net/mptcp/protocol.h +6 -0
+static inline struct sock *mptcp_sched_get_subflow(struct mptcp_sock *msk, bool reinject)
+{
+     struct mptcp_sched_data data = {
+             .sock           = msk->first,
+             .call_again     = 0,
+     };

It's not clear to me:
- why we need the 'sock' argument, since the scheduler already get 'msk' as argument?

sock is the selected subflow to return in the scheduer.

Quoted 1 line

- why we need to initialize it (looks like an 'output' argument ?!?)

Yes, it's an output argument. No need to init it.

Quoted 2 lines

- what is the goal/role of the 'call_again' argument. It looks like we just ignore it?!?

call_again is for supporting a "redundant" packet scheduler in the future indicate that get_subflow() function needs to be called again.

Quoted 22 lines

Side note: perhaps we should move the following chunk:

---

       sock_owned_by_me(sk);

        if (__mptcp_check_fallback(msk)) {
                if (!msk->first)
                        return NULL;
                return sk_stream_memory_free(msk->first) ? msk->first : NULL;
        }

---

out of mptcp_subflow_get_send() into mptcp_sched_get_subflow(), so
that every scheduler will not have to deal with fallback sockets.

net/mptcp/protocol.h int mptcp_init_sched(struct mptcp_sock *msk, · line 646

Quoted 5 lines
net/mptcp/protocol.h +5 -0
+
+     msk->sched ? INDIRECT_CALL_INET_1(msk->sched->get_subflow,
+                                       mptcp_get_subflow_default,
+                                       msk, reinject, &data) :
+                  mptcp_get_subflow_default(msk, reinject, &data);
Quoted 4 lines

With this we have quite a lot of conditionals for the default scheduler. I think we can drop the INDIRECT_CALL_INET_1() wrapper, and rework the code so that msk->sched is always NULL with the default scheduler.

So I need to drop patch 2 (2/8 mptcp: register default scheduler) in this series, right?

Quoted 4 lines

And sorry to bring the next topic so late, but why a 'reinject' argument instead of an additional mptcp_sched_ops op? the latter option will avoid another branch in fast-path.

How about keeping th 'reinject' argument like this:

----
static inline struct sock *mptcp_sched_get_send(struct mptcp_sock *msk)

{
        struct mptcp_sched_data data;

        sock_owned_by_me((struct sock *)msk);

        /* the following check is moved out of mptcp_subflow_get_send */
        if (__mptcp_check_fallback(msk)) {
                if (!msk->first)
                        return NULL;
                return sk_stream_memory_free(msk->first) ? msk->first : NULL;
        }

        if (!msk->sched)
                return mptcp_subflow_get_send(msk);

        msk->sched->get_subflow(msk, false, &data);

        return data.sock;
}

static inline struct sock *mptcp_sched_get_retrans(struct mptcp_sock *msk)

{
        struct mptcp_sched_data data;

        sock_owned_by_me((const struct sock *)msk);

        if (__mptcp_check_fallback(msk))
                return NULL;

        if (!msk->sched)
                return mptcp_subflow_get_retrans(msk);

        msk->sched->get_subflow(msk, true, &data);

        return data.sock;
}

-----

Thanks,
-Geliang

Quoted 6 lines

Thanks!

Paolo

#9 lore raw

Re: [PATCH mptcp-next v16 5/8] mptcp: add get_subflow wrapper

Hi Geliang, Paolo,

Quoted 65 lines · Geliang Tang

Hi Paolo,

Thanks for your review!

Quoted 46 lines · Paolo Abeni <[email protected]> 于2022年4月27日周三 16:27写道:

Hello,

First of all I'm sorry for the very late feedback: I had some difficulties to allocate time to follow this development.

Quoted 15 lines · Geliang Tang

This patch defines 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 |  9 ++++-----
 net/mptcp/protocol.h | 16 ++++++++++++++++
 2 files changed, 20 insertions(+), 5 deletions(-)
net/mptcp/protocol.c +0 -0
index 7590e2d29f39..c6e963848b18 100644

net/mptcp/protocol.c static struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk) · line 1520

Quoted 7 lines
net/mptcp/protocol.c +0 -1
@@ -1515,7 +1515,6 @@ static struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk)
      subflow->avg_pacing_rate = div_u64((u64)subflow->avg_pacing_rate * wmem +
                                         READ_ONCE(ssk->sk_pacing_rate) * burst,
                                         burst + wmem);
-     msk->last_snd = ssk;
      msk->snd_burst = burst;
      return ssk;
 }

This breaks "mptcp: really share subflow snd_wnd": When the MPTCP-level cwin is 0, we want to do 0-window probe (so mptcp_subflow_get_send() return a 'valid' subflow) but we don't want to start a new burst for such 0 win probe (so mptcp_subflow_get_send() clears last_snd).

With this change, when MPTCP-level cwin is 0, 'last_send' is not cleared anymore.

A simple sulution would be:

---

net/mptcp/protocol.c +0 -0
index 653757ea0aca..c20f5fd04cad 100644

net/mptcp/protocol.c static struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk) · line 1513

Quoted 8 lines
net/mptcp/protocol.c +1 -2
@@ -1507,7 +1507,7 @@ static struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk)
        burst = min_t(int, MPTCP_SEND_BURST_SIZE, mptcp_wnd_end(msk) - msk->snd_nxt);
        wmem = READ_ONCE(ssk->sk_wmem_queued);
        if (!burst) {
-               msk->last_snd = NULL;
+               msk->snd_burst = 0;
                return ssk;
        }
---

I think it's better to add back "msk->last_snd = ssk;" in mptcp_subflow_get_send(), keep the original mptcp_subflow_get_send() unchanged. And set msk->last_snd in the get_subflow() of the scheduler, the same as v12:

Quoted 1 line

https://patchwork.kernel.org/project/mptcp/patch/396f8edcbf8e4f65a1d76c228dc3d12aa4dd2f11.1650430389.git.geliang.tang@suse.com/

It might be interesting to discuss about the scheduler API at the meeting tomorrow (sadly I will not be there).

Maybe it is interesting to separate the work in two steps like it is done in mptcp.org: select data, then select subflow.

If the scheduler needs to know if it is is selecting a subflow for a zero window test or not: we are getting closer to the mptcp.org API :)

(or maybe just enough to reset msk->snd_burst instead of asking all schedulers to set 'last_snd' + give access to that for BPF)

Cheers,
Matt

Quoted 161 lines
Quoted 51 lines

Probably a better solution would be adding 'snd_burst' to struct mptcp_sched_data, and let mptcp_sched_get_subflow() update msk-

Quoted 1 line

snd_burst as specified by the scheduler. With this latter change the

'msk' argument in mptcp_sched_ops->schedule() could be const.

Quoted 44 lines
net/mptcp/protocol.c +4 -4
@@ -1575,7 +1574,7 @@ 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, false);
 
                      /* First check. If the ssk has changed since
                       * the last round, release prev_ssk
@@ -1644,7 +1643,7 @@ 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), false);
                      if (!xmit_ssk)
                              goto out;
                      if (xmit_ssk != ssk) {
@@ -2489,7 +2488,7 @@ static void __mptcp_retrans(struct sock *sk)
      mptcp_clean_una_wakeup(sk);
 
      /* first check ssk: need to kick "stale" logic */
-     ssk = mptcp_subflow_get_retrans(msk);
+     ssk = mptcp_sched_get_subflow(msk, true);
      dfrag = mptcp_rtx_head(sk);
      if (!dfrag) {
              if (mptcp_data_fin_enabled(msk)) {
@@ -3154,7 +3153,7 @@ 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), false);
 
              if (xmit_ssk == ssk)
                      __mptcp_subflow_push_pending(sk, ssk);
net/mptcp/protocol.h +0 -0
index 22f3f41e1e32..91512fc25128 100644
@@ -633,6 +633,22 @@ int mptcp_init_sched(struct mptcp_sock *msk,
                   struct mptcp_sched_ops *sched);
 void mptcp_release_sched(struct mptcp_sock *msk);
 
Quoted 10 lines

net/mptcp/protocol.h int mptcp_init_sched(struct mptcp_sock *msk, · line 641

Quoted 6 lines
net/mptcp/protocol.h +6 -0
+static inline struct sock *mptcp_sched_get_subflow(struct mptcp_sock *msk, bool reinject)
+{
+     struct mptcp_sched_data data = {
+             .sock           = msk->first,
+             .call_again     = 0,
+     };

It's not clear to me:
- why we need the 'sock' argument, since the scheduler already get 'msk' as argument?

sock is the selected subflow to return in the scheduer.

Quoted 1 line

- why we need to initialize it (looks like an 'output' argument ?!?)

Yes, it's an output argument. No need to init it.

Quoted 2 lines

- what is the goal/role of the 'call_again' argument. It looks like we just ignore it?!?

call_again is for supporting a "redundant" packet scheduler in the future indicate that get_subflow() function needs to be called again.

Quoted 22 lines

Side note: perhaps we should move the following chunk:

---

       sock_owned_by_me(sk);

        if (__mptcp_check_fallback(msk)) {
                if (!msk->first)
                        return NULL;
                return sk_stream_memory_free(msk->first) ? msk->first : NULL;
        }

---

out of mptcp_subflow_get_send() into mptcp_sched_get_subflow(), so
that every scheduler will not have to deal with fallback sockets.

net/mptcp/protocol.h int mptcp_init_sched(struct mptcp_sock *msk, · line 646

Quoted 5 lines
net/mptcp/protocol.h +5 -0
+
+     msk->sched ? INDIRECT_CALL_INET_1(msk->sched->get_subflow,
+                                       mptcp_get_subflow_default,
+                                       msk, reinject, &data) :
+                  mptcp_get_subflow_default(msk, reinject, &data);
Quoted 4 lines

With this we have quite a lot of conditionals for the default scheduler. I think we can drop the INDIRECT_CALL_INET_1() wrapper, and rework the code so that msk->sched is always NULL with the default scheduler.

So I need to drop patch 2 (2/8 mptcp: register default scheduler) in this series, right?

Quoted 4 lines

And sorry to bring the next topic so late, but why a 'reinject' argument instead of an additional mptcp_sched_ops op? the latter option will avoid another branch in fast-path.

How about keeping th 'reinject' argument like this:

----
static inline struct sock *mptcp_sched_get_send(struct mptcp_sock *msk)

{
        struct mptcp_sched_data data;

        sock_owned_by_me((struct sock *)msk);

        /* the following check is moved out of mptcp_subflow_get_send */
        if (__mptcp_check_fallback(msk)) {
                if (!msk->first)
                        return NULL;
                return sk_stream_memory_free(msk->first) ? msk->first : NULL;
        }

        if (!msk->sched)
                return mptcp_subflow_get_send(msk);

        msk->sched->get_subflow(msk, false, &data);

        return data.sock;
}

static inline struct sock *mptcp_sched_get_retrans(struct mptcp_sock *msk)

{
        struct mptcp_sched_data data;

        sock_owned_by_me((const struct sock *)msk);

        if (__mptcp_check_fallback(msk))
                return NULL;

        if (!msk->sched)
                return mptcp_subflow_get_retrans(msk);

        msk->sched->get_subflow(msk, true, &data);

        return data.sock;
}

-----

Thanks,
-Geliang

Quoted 6 lines

Thanks!

Paolo

Signature · 2 lines
Tessares | Belgium | Hybrid Access Solutions
www.tessares.net