Series comparison

From: Geliang Tang <tanggeliang@kylinos.cn>

I got the following error when running bpf burst sched selftests in auto-btf-debug mode:

[  167.258293][   T49] BUG: sleeping function called from invalid context at net/core/sock.c:3652
[  167.258981][   T49] in_atomic(): 0, irqs_disabled(): 0, non_block: 0, pid: 49, name: kworker/0:2
[  167.259547][   T49] preempt_count: 0, expected: 0
[  167.259826][   T49] RCU nest depth: 1, expected: 0
[  167.259975][   T49] 5 locks held by kworker/0:2/49:
[  167.260137][   T49]  #0: ffff888001130948 ((wq_completion)events){+.+.}-{0:0}, at: process_one_work+0x7e4/0x16b0
[  167.260741][   T49]  #1: ffffc90000347d90 ((work_completion)(&msk->work)){+.+.}-{0:0}, at: process_one_work+0xdf9/0x16b0
[  167.261206][   T49]  #2: ffff88800c620e58 (sk_lock-AF_INET){+.+.}-{0:0}, at: mptcp_worker+0x7b/0xad0
[  167.261476][   T49]  #3: ffffffff8977e620 (rcu_read_lock){....}-{1:2}, at: __bpf_prog_enter+0x1f/0x170
[  167.261742][   T49]  #4: ffff88800b6f8258 (k-sk_lock-AF_INET#2){+.+.}-{0:0}, at: bpf_prog_87e3c1bd65224f1c_bpf_burst_get_subflow+0x12c/0x58b
[  167.262309][   T49] CPU: 0 UID: 0 PID: 49 Comm: kworker/0:2 Tainted: G           OE      6.12.0-rc6+ #9
[  167.262591][   T49] Tainted: [O]=OOT_MODULE, [E]=UNSIGNED_MODULE
[  167.262774][   T49] Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011
[  167.262950][   T49] Workqueue: events mptcp_worker
[  167.263097][   T49] Call Trace:
[  167.263211][   T49]  <TASK>
[  167.263301][   T49]  dump_stack_lvl+0x9e/0xe0
[  167.263453][   T49]  __might_resched+0x35d/0x590
[  167.263599][   T49]  ? __pfx___might_resched+0x10/0x10
[  167.263829][   T49]  __lock_sock_fast+0x2f/0xd0
[  167.264035][   T49]  mptcp_pm_nl_subflow_chk_stale+0x1e2/0x3c0
[  167.264304][   T49]  ? bpf_prog_87e3c1bd65224f1c_bpf_burst_get_subflow+0x12c/0x58b
[  167.264604][   T49]  bpf_prog_87e3c1bd65224f1c_bpf_burst_get_subflow+0x12c/0x58b
[  167.264995][   T49]  ? mptcp_sched_get_retrans+0x1f8/0x330
[  167.265215][   T49]  ? __pfx_mptcp_sched_get_retrans+0x10/0x10
[  167.265432][   T49]  ? mark_lock+0x371/0x3c0
[  167.265678][   T49]  ? __mptcp_retrans+0xf4/0x9a0
[  167.265973][   T49]  ? mptcp_worker+0x7b/0xad0

mptcp_pm_subflow_chk_stale is a sleeping function, it shouldn't be
invoked under BPF rcu_read_lock.

One solution is to set mptcp_pm_subflow_chk_stale with KF_SLEEPABLE
flag:

BTF_ID_FLAGS(func, mptcp_pm_subflow_chk_stale, KF_SLEEPABLE)

Then set bpf_burst_get_subflow with BPF_F_SLEEPABLE:

err = bpf_program__set_flags(skel->progs.bpf_burst_get_subflow,
                                   BPF_F_SLEEPABLE);

But another error ocurrs:

[  101.732098][    C0] BUG: sleeping function called from invalid context at kernel/bpf/trampoline.c:968
[  101.732781][    C0] in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 1640, name: test_progs-no_a
[  101.733477][    C0] preempt_count: 303, expected: 0
[  101.733793][    C0] RCU nest depth: 2, expected: 0
[  101.734043][    C0] 6 locks held by test_progs-no_a/1640:
[  101.734187][    C0] #0: ffff888007526e58 (sk_lock-AF_INET){+.+.}-{0:0}, at: sk_wait_data (net/core/sock.c:3131 (discriminator 2)) 
[  101.734440][    C0] #1: ffffffff95b7e620 (rcu_read_lock){....}-{1:2}, at: process_backlog (include/linux/local_lock_internal.h:38 (discriminator 1) net/core/dev.c:6115 (discriminator 1)) 
[  101.734728][    C0] #2: ffffffff95b7e620 (rcu_read_lock){....}-{1:2}, at: ip_local_deliver_finish (include/linux/rcupdate.h:337 (discriminator 1) include/linux/rcupdate.h:849 (discriminator 1) net/ipv4/ip_input.c:232 (discriminator 1)) 
[  101.735103][    C0] #3: ffff888008eb8e58 (k-slock-AF_INET/1){+.-.}-{2:2}, at: tcp_v4_rcv (include/linux/skbuff.h:1670 include/net/tcp.h:2530 net/ipv4/tcp_ipv4.c:2348) 
[  101.735560][    C0] #4: ffff8880075261d8 (slock-AF_INET){+.-.}-{2:2}, at: mptcp_incoming_options (net/mptcp/options.c:1053 net/mptcp/options.c:1203) 
[  101.736023][    C0] #5: ffffffff95b7d980 (rcu_read_lock_trace){....}-{0:0}, at: __bpf_prog_enter_sleepable (include/linux/rcupdate.h:337 (discriminator 1) include/linux/rcupdate_trace.h:57 (discriminator 1) kernel/bpf/trampoline.c:966 (discriminator 1)) 
[  101.736325][    C0] CPU: 0 UID: 0 PID: 1640 Comm: test_progs-no_a Tainted: G           OE      6.12.0-rc7+ #5
[  101.736676][    C0] Tainted: [O]=OOT_MODULE, [E]=UNSIGNED_MODULE
[  101.736940][    C0] Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011
[  101.737146][    C0] Call Trace:
[  101.737259][    C0]  <IRQ>
[  101.737331][    C0] dump_stack_lvl (lib/dump_stack.c:123) 
[  101.737489][    C0] __might_resched (kernel/sched/core.c:8657) 
[  101.737625][    C0] ? __lock_acquire (kernel/locking/lockdep.c:5202 (discriminator 1)) 
[  101.737772][    C0] ? __pfx___might_resched (kernel/sched/core.c:8611) 
[  101.737919][    C0] __might_fault (mm/memory.c:6715 (discriminator 1)) 
[  101.738051][    C0] __bpf_prog_enter_sleepable (include/linux/bpf.h:2089 (discriminator 1) kernel/bpf/trampoline.c:970 (discriminator 1)) 
[  101.738195][    C0] ? __bpf_prog_enter_sleepable (include/linux/rcupdate.h:337 (discriminator 1) include/linux/rcupdate_trace.h:57 (discriminator 1) kernel/bpf/trampoline.c:966 (discriminator 1)) 
[  101.738357][    C0] ? mptcp_sched_get_send (net/mptcp/sched.c:183) 
[  101.738511][    C0] ? __pfx_mptcp_sched_get_send (net/mptcp/sched.c:158) 
[  101.738641][    C0] ? __bpf_tramp_enter (include/linux/rcupdate.h:347 (discriminator 1) include/linux/rcupdate.h:880 (discriminator 1) include/linux/percpu-refcount.h:209 (discriminator 1) include/linux/percpu-refcount.h:222 (discriminator 1) kernel/bpf/trampoline.c:1010 (discriminator 1)) 
[  101.738774][    C0] ? bpf_trampoline_6442561300+0x31/0xd7 
[  101.738911][    C0] ? mptcp_sched_get_send (net/mptcp/sched.c:158) 
[  101.739067][    C0] ? __mptcp_subflow_push_pending (net/mptcp/protocol.c:1682 (discriminator 1)) 
[  101.739231][    C0] ? __pfx___mptcp_subflow_push_pending (net/mptcp/protocol.c:1655) 
[  101.739378][    C0] ? do_raw_spin_lock (arch/x86/include/asm/atomic.h:107 (discriminator 1) include/linux/atomic/atomic-arch-fallback.h:2170 (discriminator 1) include/linux/atomic/atomic-instrumented.h:1302 (discriminator 1) include/asm-generic/qspinlock.h:111 (discriminator 1) kernel/locking/spinlock_debug.c:116 (discriminator 1)) 
[  101.739556][    C0] ? __pfx_do_raw_spin_lock (kernel/locking/spinlock_debug.c:114) 
[  101.739743][    C0] ? mptcp_incoming_options (net/mptcp/options.c:1067 net/mptcp/options.c:1203)

This means get_send() interface of this scheduler can't be set with
BPF_F_SLEEPABLE flag since it's invoked in ack_update_msk() under mptcp
data lock.

So this patch has to split get_subflow() interface of packet scheduer into
two interfaces: get_send() and get_retrans(). Then we can set get_retrans()
interface alone with BPF_F_SLEEPABLE flag.

Thanks to Barry for helping me solve this issue.
Cc: Barry Song <baohua@kernel.org>

Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/529

Geliang Tang (8):
  mptcp: split get_subflow interface into two
  Squash to "selftests/bpf: Add bpf_first scheduler & test"
  Squash to "selftests/bpf: Add bpf_bkup scheduler & test"
  Squash to "selftests/bpf: Add bpf_rr scheduler & test"
  Squash to "selftests/bpf: Add bpf_red scheduler & test"
  Squash to "selftests/bpf: Add bpf_burst scheduler & test"
  Squash to "bpf: Export mptcp packet scheduler helpers"
  Squash to "selftests/bpf: Add bpf_burst scheduler & test"

include/net/mptcp.h                           |  5 +--
 net/mptcp/bpf.c                               | 13 ++++++--
 net/mptcp/sched.c                             | 33 ++++++++++++-------
 .../testing/selftests/bpf/prog_tests/mptcp.c  | 14 ++++++--
 .../selftests/bpf/progs/mptcp_bpf_bkup.c      | 24 +++++++++++---
 .../selftests/bpf/progs/mptcp_bpf_burst.c     | 32 ++++++++----------
 .../selftests/bpf/progs/mptcp_bpf_first.c     | 24 +++++++++++---
 .../selftests/bpf/progs/mptcp_bpf_red.c       | 24 +++++++++++---
 .../selftests/bpf/progs/mptcp_bpf_rr.c        | 24 +++++++++++---
 9 files changed, 136 insertions(+), 57 deletions(-)

-- 
2.43.0

From: Geliang Tang <tanggeliang@kylinos.cn>

get_retrans() interface of the burst packet scheduler invokes a sleeping
function mptcp_pm_subflow_chk_stale(), which calls __lock_sock_fast().
So get_retrans() interface should be set with BPF_F_SLEEPABLE flag in
BPF. But get_send() interface of this scheduler can't be set with
BPF_F_SLEEPABLE flag since it's invoked in ack_update_msk() under mptcp
data lock.

So this patch has to split get_subflow() interface of packet scheduer into
two interfaces: get_send() and get_retrans(). Then we can set get_retrans()
interface alone with BPF_F_SLEEPABLE flag.

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 include/net/mptcp.h |  5 +++--
 net/mptcp/bpf.c     | 11 +++++++++--
 net/mptcp/sched.c   | 33 ++++++++++++++++++++++-----------
 3 files changed, 34 insertions(+), 15 deletions(-)

diff --git a/include/net/mptcp.h b/include/net/mptcp.h
index XXXXXXX..XXXXXXX 100644
--- a/include/net/mptcp.h
+++ b/include/net/mptcp.h
@@ -XXX,XX +XXX,XX @@ struct mptcp_out_options {
 #define MPTCP_SUBFLOWS_MAX	8
 
 struct mptcp_sched_data {
-	bool	reinject;
 	u8	subflows;
 	struct mptcp_subflow_context *contexts[MPTCP_SUBFLOWS_MAX];
 };
 
 struct mptcp_sched_ops {
-	int (*get_subflow)(struct mptcp_sock *msk,
+	int (*get_send)(struct mptcp_sock *msk,
+			struct mptcp_sched_data *data);
+	int (*get_retrans)(struct mptcp_sock *msk,
 			   struct mptcp_sched_data *data);
 
 	char			name[MPTCP_SCHED_NAME_MAX];
diff --git a/net/mptcp/bpf.c b/net/mptcp/bpf.c
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/bpf.c
+++ b/net/mptcp/bpf.c
@@ -XXX,XX +XXX,XX @@ static int bpf_mptcp_sched_init(struct btf *btf)
 	return 0;
 }
 
-static int __bpf_mptcp_sched_get_subflow(struct mptcp_sock *msk,
+static int __bpf_mptcp_sched_get_send(struct mptcp_sock *msk,
+				      struct mptcp_sched_data *data)
+{
+	return 0;
+}
+
+static int __bpf_mptcp_sched_get_retrans(struct mptcp_sock *msk,
 					 struct mptcp_sched_data *data)
 {
 	return 0;
@@ -XXX,XX +XXX,XX @@ static void __bpf_mptcp_sched_release(struct mptcp_sock *msk)
 }
 
 static struct mptcp_sched_ops __bpf_mptcp_sched_ops = {
-	.get_subflow	= __bpf_mptcp_sched_get_subflow,
+	.get_send	= __bpf_mptcp_sched_get_send,
+	.get_retrans	= __bpf_mptcp_sched_get_retrans,
 	.init		= __bpf_mptcp_sched_init,
 	.release	= __bpf_mptcp_sched_release,
 };
diff --git a/net/mptcp/sched.c b/net/mptcp/sched.c
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/sched.c
+++ b/net/mptcp/sched.c
@@ -XXX,XX +XXX,XX @@
 static DEFINE_SPINLOCK(mptcp_sched_list_lock);
 static LIST_HEAD(mptcp_sched_list);
 
-static int mptcp_sched_default_get_subflow(struct mptcp_sock *msk,
+static int mptcp_sched_default_get_send(struct mptcp_sock *msk,
+					struct mptcp_sched_data *data)
+{
+	struct sock *ssk;
+
+	ssk = mptcp_subflow_get_send(msk);
+	if (!ssk)
+		return -EINVAL;
+
+	mptcp_subflow_set_scheduled(mptcp_subflow_ctx(ssk), true);
+	return 0;
+}
+
+static int mptcp_sched_default_get_retrans(struct mptcp_sock *msk,
 					   struct mptcp_sched_data *data)
 {
 	struct sock *ssk;
 
-	ssk = data->reinject ? mptcp_subflow_get_retrans(msk) :
-			       mptcp_subflow_get_send(msk);
+	ssk = mptcp_subflow_get_retrans(msk);
 	if (!ssk)
 		return -EINVAL;
 
@@ -XXX,XX +XXX,XX @@ static int mptcp_sched_default_get_subflow(struct mptcp_sock *msk,
 }
 
 static struct mptcp_sched_ops mptcp_sched_default = {
-	.get_subflow	= mptcp_sched_default_get_subflow,
+	.get_send	= mptcp_sched_default_get_send,
+	.get_retrans	= mptcp_sched_default_get_retrans,
 	.name		= "default",
 	.owner		= THIS_MODULE,
 };
@@ -XXX,XX +XXX,XX @@ void mptcp_get_available_schedulers(char *buf, size_t maxlen)
 
 int mptcp_register_scheduler(struct mptcp_sched_ops *sched)
 {
-	if (!sched->get_subflow)
+	if (!sched->get_send || !sched->get_retrans)
 		return -EINVAL;
 
 	spin_lock(&mptcp_sched_list_lock);
@@ -XXX,XX +XXX,XX @@ int mptcp_sched_get_send(struct mptcp_sock *msk)
 			return 0;
 	}
 
-	data.reinject = false;
 	if (msk->sched == &mptcp_sched_default || !msk->sched)
-		return mptcp_sched_default_get_subflow(msk, &data);
+		return mptcp_sched_default_get_send(msk, &data);
 	mptcp_sched_data_set_contexts(msk, &data);
-	return msk->sched->get_subflow(msk, &data);
+	return msk->sched->get_send(msk, &data);
 }
 
 int mptcp_sched_get_retrans(struct mptcp_sock *msk)
@@ -XXX,XX +XXX,XX @@ int mptcp_sched_get_retrans(struct mptcp_sock *msk)
 			return 0;
 	}
 
-	data.reinject = true;
 	if (msk->sched == &mptcp_sched_default || !msk->sched)
-		return mptcp_sched_default_get_subflow(msk, &data);
+		return mptcp_sched_default_get_retrans(msk, &data);
 	mptcp_sched_data_set_contexts(msk, &data);
-	return msk->sched->get_subflow(msk, &data);
+	return msk->sched->get_retrans(msk, &data);
 }
-- 
2.43.0

From: Geliang Tang <tanggeliang@kylinos.cn>

Split get_subflow() interface into two: get_send() and get_retrans().

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 .../selftests/bpf/progs/mptcp_bpf_first.c     | 24 +++++++++++++++----
 1 file changed, 19 insertions(+), 5 deletions(-)

diff --git a/tools/testing/selftests/bpf/progs/mptcp_bpf_first.c b/tools/testing/selftests/bpf/progs/mptcp_bpf_first.c
index XXXXXXX..XXXXXXX 100644
--- a/tools/testing/selftests/bpf/progs/mptcp_bpf_first.c
+++ b/tools/testing/selftests/bpf/progs/mptcp_bpf_first.c
@@ -XXX,XX +XXX,XX @@ void BPF_PROG(mptcp_sched_first_release, struct mptcp_sock *msk)
 {
 }
 
-SEC("struct_ops")
-int BPF_PROG(bpf_first_get_subflow, struct mptcp_sock *msk,
-	     struct mptcp_sched_data *data)
+static int bpf_first_get_subflow(struct mptcp_sock *msk,
+				 struct mptcp_sched_data *data)
 {
 	mptcp_subflow_set_scheduled(bpf_mptcp_subflow_ctx_by_pos(data, 0), true);
 	return 0;
 }
 
-SEC(".struct_ops")
+SEC("struct_ops")
+int BPF_PROG(bpf_first_get_send, struct mptcp_sock *msk,
+	     struct mptcp_sched_data *data)
+{
+	return bpf_first_get_subflow(msk, data);
+}
+
+SEC("struct_ops")
+int BPF_PROG(bpf_first_get_retrans, struct mptcp_sock *msk,
+	     struct mptcp_sched_data *data)
+{
+	return bpf_first_get_subflow(msk, data);
+}
+
+SEC(".struct_ops.link")
 struct mptcp_sched_ops first = {
 	.init		= (void *)mptcp_sched_first_init,
 	.release	= (void *)mptcp_sched_first_release,
-	.get_subflow	= (void *)bpf_first_get_subflow,
+	.get_send	= (void *)bpf_first_get_send,
+	.get_retrans	= (void *)bpf_first_get_retrans,
 	.name		= "bpf_first",
 };
-- 
2.43.0

From: Geliang Tang <tanggeliang@kylinos.cn>

Split get_subflow() interface into two: get_send() and get_retrans().

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 .../selftests/bpf/progs/mptcp_bpf_bkup.c      | 24 +++++++++++++++----
 1 file changed, 19 insertions(+), 5 deletions(-)

diff --git a/tools/testing/selftests/bpf/progs/mptcp_bpf_bkup.c b/tools/testing/selftests/bpf/progs/mptcp_bpf_bkup.c
index XXXXXXX..XXXXXXX 100644
--- a/tools/testing/selftests/bpf/progs/mptcp_bpf_bkup.c
+++ b/tools/testing/selftests/bpf/progs/mptcp_bpf_bkup.c
@@ -XXX,XX +XXX,XX @@ void BPF_PROG(mptcp_sched_bkup_release, struct mptcp_sock *msk)
 {
 }
 
-SEC("struct_ops")
-int BPF_PROG(bpf_bkup_get_subflow, struct mptcp_sock *msk,
-	     struct mptcp_sched_data *data)
+static int bpf_bkup_get_subflow(struct mptcp_sock *msk,
+				struct mptcp_sched_data *data)
 {
 	int nr = -1;
 
@@ -XXX,XX +XXX,XX @@ int BPF_PROG(bpf_bkup_get_subflow, struct mptcp_sock *msk,
 	return 0;
 }
 
-SEC(".struct_ops")
+SEC("struct_ops")
+int BPF_PROG(bpf_bkup_get_send, struct mptcp_sock *msk,
+	     struct mptcp_sched_data *data)
+{
+	return bpf_bkup_get_subflow(msk, data);
+}
+
+SEC("struct_ops")
+int BPF_PROG(bpf_bkup_get_retrans, struct mptcp_sock *msk,
+	     struct mptcp_sched_data *data)
+{
+	return bpf_bkup_get_subflow(msk, data);
+}
+
+SEC(".struct_ops.link")
 struct mptcp_sched_ops bkup = {
 	.init		= (void *)mptcp_sched_bkup_init,
 	.release	= (void *)mptcp_sched_bkup_release,
-	.get_subflow	= (void *)bpf_bkup_get_subflow,
+	.get_send	= (void *)bpf_bkup_get_send,
+	.get_retrans	= (void *)bpf_bkup_get_retrans,
 	.name		= "bpf_bkup",
 };
-- 
2.43.0

From: Geliang Tang <tanggeliang@kylinos.cn>

Split get_subflow() interface into two: get_send() and get_retrans().

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 .../selftests/bpf/progs/mptcp_bpf_rr.c        | 24 +++++++++++++++----
 1 file changed, 19 insertions(+), 5 deletions(-)

diff --git a/tools/testing/selftests/bpf/progs/mptcp_bpf_rr.c b/tools/testing/selftests/bpf/progs/mptcp_bpf_rr.c
index XXXXXXX..XXXXXXX 100644
--- a/tools/testing/selftests/bpf/progs/mptcp_bpf_rr.c
+++ b/tools/testing/selftests/bpf/progs/mptcp_bpf_rr.c
@@ -XXX,XX +XXX,XX @@ void BPF_PROG(mptcp_sched_rr_release, struct mptcp_sock *msk)
 	bpf_sk_storage_delete(&mptcp_rr_map, msk);
 }
 
-SEC("struct_ops")
-int BPF_PROG(bpf_rr_get_subflow, struct mptcp_sock *msk,
-	     struct mptcp_sched_data *data)
+static int bpf_rr_get_subflow(struct mptcp_sock *msk,
+			      struct mptcp_sched_data *data)
 {
 	struct mptcp_subflow_context *subflow;
 	struct mptcp_rr_storage *ptr;
@@ -XXX,XX +XXX,XX @@ int BPF_PROG(bpf_rr_get_subflow, struct mptcp_sock *msk,
 	return 0;
 }
 
-SEC(".struct_ops")
+SEC("struct_ops")
+int BPF_PROG(bpf_rr_get_send, struct mptcp_sock *msk,
+	     struct mptcp_sched_data *data)
+{
+	return bpf_rr_get_subflow(msk, data);
+}
+
+SEC("struct_ops")
+int BPF_PROG(bpf_rr_get_retrans, struct mptcp_sock *msk,
+	     struct mptcp_sched_data *data)
+{
+	return bpf_rr_get_subflow(msk, data);
+}
+
+SEC(".struct_ops.link")
 struct mptcp_sched_ops rr = {
 	.init		= (void *)mptcp_sched_rr_init,
 	.release	= (void *)mptcp_sched_rr_release,
-	.get_subflow	= (void *)bpf_rr_get_subflow,
+	.get_send	= (void *)bpf_rr_get_send,
+	.get_retrans	= (void *)bpf_rr_get_retrans,
 	.name		= "bpf_rr",
 };
-- 
2.43.0

From: Geliang Tang <tanggeliang@kylinos.cn>

Split get_subflow() interface into two: get_send() and get_retrans().

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 .../selftests/bpf/progs/mptcp_bpf_red.c       | 24 +++++++++++++++----
 1 file changed, 19 insertions(+), 5 deletions(-)

diff --git a/tools/testing/selftests/bpf/progs/mptcp_bpf_red.c b/tools/testing/selftests/bpf/progs/mptcp_bpf_red.c
index XXXXXXX..XXXXXXX 100644
--- a/tools/testing/selftests/bpf/progs/mptcp_bpf_red.c
+++ b/tools/testing/selftests/bpf/progs/mptcp_bpf_red.c
@@ -XXX,XX +XXX,XX @@ void BPF_PROG(mptcp_sched_red_release, struct mptcp_sock *msk)
 {
 }
 
-SEC("struct_ops")
-int BPF_PROG(bpf_red_get_subflow, struct mptcp_sock *msk,
-	     struct mptcp_sched_data *data)
+static int bpf_red_get_subflow(struct mptcp_sock *msk,
+			       struct mptcp_sched_data *data)
 {
 	for (int i = 0; i < data->subflows && i < MPTCP_SUBFLOWS_MAX; i++) {
 		if (!bpf_mptcp_subflow_ctx_by_pos(data, i))
@@ -XXX,XX +XXX,XX @@ int BPF_PROG(bpf_red_get_subflow, struct mptcp_sock *msk,
 	return 0;
 }
 
-SEC(".struct_ops")
+SEC("struct_ops")
+int BPF_PROG(bpf_red_get_send, struct mptcp_sock *msk,
+	     struct mptcp_sched_data *data)
+{
+	return bpf_red_get_subflow(msk, data);
+}
+
+SEC("struct_ops")
+int BPF_PROG(bpf_red_get_retrans, struct mptcp_sock *msk,
+	     struct mptcp_sched_data *data)
+{
+	return bpf_red_get_subflow(msk, data);
+}
+
+SEC(".struct_ops.link")
 struct mptcp_sched_ops red = {
 	.init		= (void *)mptcp_sched_red_init,
 	.release	= (void *)mptcp_sched_red_release,
-	.get_subflow	= (void *)bpf_red_get_subflow,
+	.get_send	= (void *)bpf_red_get_send,
+	.get_retrans	= (void *)bpf_red_get_retrans,
 	.name		= "bpf_red",
 };
-- 
2.43.0

From: Geliang Tang <tanggeliang@kylinos.cn>

Split get_subflow() interface into two: get_send() and get_retrans().

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 .../selftests/bpf/progs/mptcp_bpf_burst.c     | 32 ++++++++-----------
 1 file changed, 13 insertions(+), 19 deletions(-)

diff --git a/tools/testing/selftests/bpf/progs/mptcp_bpf_burst.c b/tools/testing/selftests/bpf/progs/mptcp_bpf_burst.c
index XXXXXXX..XXXXXXX 100644
--- a/tools/testing/selftests/bpf/progs/mptcp_bpf_burst.c
+++ b/tools/testing/selftests/bpf/progs/mptcp_bpf_burst.c
@@ -XXX,XX +XXX,XX @@ char _license[] SEC("license") = "GPL";
 
 #define MPTCP_SEND_BURST_SIZE	65428
 
+#define SSK_MODE_ACTIVE	0
+#define SSK_MODE_BACKUP	1
+#define SSK_MODE_MAX	2
+
 #define min(a, b) ((a) < (b) ? (a) : (b))
 
 struct bpf_subflow_send_info {
@@ -XXX,XX +XXX,XX @@ extern bool tcp_stream_memory_free(const struct sock *sk, int wake) __ksym;
 extern bool bpf_mptcp_subflow_queues_empty(struct sock *sk) __ksym;
 extern void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk) __ksym;
 
-#define SSK_MODE_ACTIVE	0
-#define SSK_MODE_BACKUP	1
-#define SSK_MODE_MAX	2
-
 static __always_inline __u64 div_u64(__u64 dividend, __u32 divisor)
 {
 	return dividend / divisor;
@@ -XXX,XX +XXX,XX @@ void BPF_PROG(mptcp_sched_burst_release, struct mptcp_sock *msk)
 {
 }
 
-static int bpf_burst_get_send(struct mptcp_sock *msk,
-			      struct mptcp_sched_data *data)
+SEC("struct_ops")
+int BPF_PROG(bpf_burst_get_send, struct mptcp_sock *msk,
+	     struct mptcp_sched_data *data)
 {
 	struct bpf_subflow_send_info send_info[SSK_MODE_MAX];
 	struct mptcp_subflow_context *subflow;
@@ -XXX,XX +XXX,XX @@ static int bpf_burst_get_send(struct mptcp_sock *msk,
 	return 0;
 }
 
-static int bpf_burst_get_retrans(struct mptcp_sock *msk,
-				 struct mptcp_sched_data *data)
+SEC("struct_ops")
+int BPF_PROG(bpf_burst_get_retrans, struct mptcp_sock *msk,
+	     struct mptcp_sched_data *data)
 {
 	int backup = MPTCP_SUBFLOWS_MAX, pick = MPTCP_SUBFLOWS_MAX, subflow_id;
 	struct mptcp_subflow_context *subflow;
@@ -XXX,XX +XXX,XX @@ static int bpf_burst_get_retrans(struct mptcp_sock *msk,
 	return 0;
 }
 
-SEC("struct_ops")
-int BPF_PROG(bpf_burst_get_subflow, struct mptcp_sock *msk,
-	     struct mptcp_sched_data *data)
-{
-	if (data->reinject)
-		return bpf_burst_get_retrans(msk, data);
-	return bpf_burst_get_send(msk, data);
-}
-
-SEC(".struct_ops")
+SEC(".struct_ops.link")
 struct mptcp_sched_ops burst = {
 	.init		= (void *)mptcp_sched_burst_init,
 	.release	= (void *)mptcp_sched_burst_release,
-	.get_subflow	= (void *)bpf_burst_get_subflow,
+	.get_send	= (void *)bpf_burst_get_send,
+	.get_retrans	= (void *)bpf_burst_get_retrans,
 	.name		= "bpf_burst",
 };
-- 
2.43.0

From: Geliang Tang <tanggeliang@kylinos.cn>

Set BPF_F_SLEEPABLE flags for bpf_burst_get_retrans.

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 tools/testing/selftests/bpf/prog_tests/mptcp.c | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)

diff --git a/tools/testing/selftests/bpf/prog_tests/mptcp.c b/tools/testing/selftests/bpf/prog_tests/mptcp.c
index XXXXXXX..XXXXXXX 100644
--- a/tools/testing/selftests/bpf/prog_tests/mptcp.c
+++ b/tools/testing/selftests/bpf/prog_tests/mptcp.c
@@ -XXX,XX +XXX,XX @@ static void test_red(void)
 static void test_burst(void)
 {
 	struct mptcp_bpf_burst *skel;
+	int err;
 
-	skel = mptcp_bpf_burst__open_and_load();
-	if (!ASSERT_OK_PTR(skel, "open_and_load: burst"))
+	skel = mptcp_bpf_burst__open();
+	if (!ASSERT_OK_PTR(skel, "open: burst"))
 		return;
 
+	err = bpf_program__set_flags(skel->progs.bpf_burst_get_retrans,
+				     BPF_F_SLEEPABLE);
+	if (!ASSERT_OK(err, "set sleepable flags"))
+		goto skel_destroy;
+
+	if (!ASSERT_OK(mptcp_bpf_burst__load(skel), "load: burst"))
+		goto skel_destroy;
+
 	test_bpf_sched(skel->obj, "burst", WITH_DATA, WITH_DATA);
+skel_destroy:
 	mptcp_bpf_burst__destroy(skel);
 }
 
-- 
2.43.0

From: Geliang Tang <tanggeliang@kylinos.cn>

v2:
 - use get_send for retransmit scheduling if get_retrans is NULL as Mat
   suggested (thanks)

I got the following error when running bpf burst sched selftests in auto-btf-debug mode:

mptcp_pm_subflow_chk_stale is a sleeping function, it shouldn't be
invoked under BPF rcu_read_lock.

One solution is to set mptcp_pm_subflow_chk_stale with KF_SLEEPABLE
flag:

BTF_ID_FLAGS(func, mptcp_pm_subflow_chk_stale, KF_SLEEPABLE)

Then set bpf_burst_get_subflow with BPF_F_SLEEPABLE:

err = bpf_program__set_flags(skel->progs.bpf_burst_get_subflow,
                                   BPF_F_SLEEPABLE);

But another error ocurrs:

This means get_send() interface of this scheduler can't be set with
BPF_F_SLEEPABLE flag since it's invoked in ack_update_msk() under mptcp
data lock.

So this patch has to split get_subflow() interface of packet scheduer into
two interfaces: get_send() and get_retrans(). Then we can set get_retrans()
interface alone with BPF_F_SLEEPABLE flag.

Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/529

Geliang Tang (9):
  mptcp: split get_subflow interface into two
  Squash to "bpf: Add bpf_mptcp_sched_ops"
  Squash to "selftests/bpf: Add bpf_first scheduler & test"
  Squash to "selftests/bpf: Add bpf_bkup scheduler & test"
  Squash to "selftests/bpf: Add bpf_rr scheduler & test"
  Squash to "selftests/bpf: Add bpf_red scheduler & test"
  Squash to "selftests/bpf: Add bpf_burst scheduler & test"
  Squash to "bpf: Export mptcp packet scheduler helpers"
  Squash to "selftests/bpf: Add bpf_burst scheduler & test"

include/net/mptcp.h                           |  5 +--
 net/mptcp/bpf.c                               | 13 +++++--
 net/mptcp/sched.c                             | 35 +++++++++++++------
 .../testing/selftests/bpf/prog_tests/mptcp.c  | 14 ++++++--
 .../selftests/bpf/progs/mptcp_bpf_bkup.c      |  4 +--
 .../selftests/bpf/progs/mptcp_bpf_burst.c     | 22 +++++-------
 .../selftests/bpf/progs/mptcp_bpf_first.c     |  4 +--
 .../selftests/bpf/progs/mptcp_bpf_red.c       |  4 +--
 .../selftests/bpf/progs/mptcp_bpf_rr.c        |  4 +--
 9 files changed, 65 insertions(+), 40 deletions(-)

-- 
2.45.2

From: Geliang Tang <tanggeliang@kylinos.cn>

So this patch has to split get_subflow() interface of packet scheduer into
two interfaces: get_send() and get_retrans(). Then we can set get_retrans()
interface alone with BPF_F_SLEEPABLE flag.

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 include/net/mptcp.h |  5 +++--
 net/mptcp/sched.c   | 35 ++++++++++++++++++++++++-----------
 2 files changed, 27 insertions(+), 13 deletions(-)

diff --git a/include/net/mptcp.h b/include/net/mptcp.h
index XXXXXXX..XXXXXXX 100644
--- a/include/net/mptcp.h
+++ b/include/net/mptcp.h
@@ -XXX,XX +XXX,XX @@ struct mptcp_out_options {
 #define MPTCP_SUBFLOWS_MAX	8
 
 struct mptcp_sched_data {
-	bool	reinject;
 	u8	subflows;
 	struct mptcp_subflow_context *contexts[MPTCP_SUBFLOWS_MAX];
 };
 
 struct mptcp_sched_ops {
-	int (*get_subflow)(struct mptcp_sock *msk,
+	int (*get_send)(struct mptcp_sock *msk,
+			struct mptcp_sched_data *data);
+	int (*get_retrans)(struct mptcp_sock *msk,
 			   struct mptcp_sched_data *data);
 
 	char			name[MPTCP_SCHED_NAME_MAX];
diff --git a/net/mptcp/sched.c b/net/mptcp/sched.c
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/sched.c
+++ b/net/mptcp/sched.c
@@ -XXX,XX +XXX,XX @@
 static DEFINE_SPINLOCK(mptcp_sched_list_lock);
 static LIST_HEAD(mptcp_sched_list);
 
-static int mptcp_sched_default_get_subflow(struct mptcp_sock *msk,
+static int mptcp_sched_default_get_send(struct mptcp_sock *msk,
+					struct mptcp_sched_data *data)
+{
+	struct sock *ssk;
+
+	ssk = mptcp_subflow_get_send(msk);
+	if (!ssk)
+		return -EINVAL;
+
+	mptcp_subflow_set_scheduled(mptcp_subflow_ctx(ssk), true);
+	return 0;
+}
+
+static int mptcp_sched_default_get_retrans(struct mptcp_sock *msk,
 					   struct mptcp_sched_data *data)
 {
 	struct sock *ssk;
 
-	ssk = data->reinject ? mptcp_subflow_get_retrans(msk) :
-			       mptcp_subflow_get_send(msk);
+	ssk = mptcp_subflow_get_retrans(msk);
 	if (!ssk)
 		return -EINVAL;
 
@@ -XXX,XX +XXX,XX @@ static int mptcp_sched_default_get_subflow(struct mptcp_sock *msk,
 }
 
 static struct mptcp_sched_ops mptcp_sched_default = {
-	.get_subflow	= mptcp_sched_default_get_subflow,
+	.get_send	= mptcp_sched_default_get_send,
+	.get_retrans	= mptcp_sched_default_get_retrans,
 	.name		= "default",
 	.owner		= THIS_MODULE,
 };
@@ -XXX,XX +XXX,XX @@ void mptcp_get_available_schedulers(char *buf, size_t maxlen)
 
 int mptcp_register_scheduler(struct mptcp_sched_ops *sched)
 {
-	if (!sched->get_subflow)
+	if (!sched->get_send)
 		return -EINVAL;
 
 	spin_lock(&mptcp_sched_list_lock);
@@ -XXX,XX +XXX,XX @@ int mptcp_sched_get_send(struct mptcp_sock *msk)
 			return 0;
 	}
 
-	data.reinject = false;
 	if (msk->sched == &mptcp_sched_default || !msk->sched)
-		return mptcp_sched_default_get_subflow(msk, &data);
+		return mptcp_sched_default_get_send(msk, &data);
 	mptcp_sched_data_set_contexts(msk, &data);
-	return msk->sched->get_subflow(msk, &data);
+	return msk->sched->get_send(msk, &data);
 }
 
 int mptcp_sched_get_retrans(struct mptcp_sock *msk)
@@ -XXX,XX +XXX,XX @@ int mptcp_sched_get_retrans(struct mptcp_sock *msk)
 			return 0;
 	}
 
-	data.reinject = true;
 	if (msk->sched == &mptcp_sched_default || !msk->sched)
-		return mptcp_sched_default_get_subflow(msk, &data);
+		return mptcp_sched_default_get_retrans(msk, &data);
 	mptcp_sched_data_set_contexts(msk, &data);
-	return msk->sched->get_subflow(msk, &data);
+	if (!msk->sched->get_retrans)
+		return msk->sched->get_send(msk, &data);
+	return msk->sched->get_retrans(msk, &data);
 }
-- 
2.45.2

From: Geliang Tang <tanggeliang@kylinos.cn>

Please update the subject to

bpf: Add mptcp packet scheduler struct_ops

Split get_subflow interface into two.

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 net/mptcp/bpf.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/net/mptcp/bpf.c b/net/mptcp/bpf.c
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/bpf.c
+++ b/net/mptcp/bpf.c
@@ -XXX,XX +XXX,XX @@ static int bpf_mptcp_sched_init(struct btf *btf)
 	return 0;
 }
 
-static int __bpf_mptcp_sched_get_subflow(struct mptcp_sock *msk,
+static int __bpf_mptcp_sched_get_send(struct mptcp_sock *msk,
+				      struct mptcp_sched_data *data)
+{
+	return 0;
+}
+
+static int __bpf_mptcp_sched_get_retrans(struct mptcp_sock *msk,
 					 struct mptcp_sched_data *data)
 {
 	return 0;
@@ -XXX,XX +XXX,XX @@ static void __bpf_mptcp_sched_release(struct mptcp_sock *msk)
 }
 
 static struct mptcp_sched_ops __bpf_mptcp_sched_ops = {
-	.get_subflow	= __bpf_mptcp_sched_get_subflow,
+	.get_send	= __bpf_mptcp_sched_get_send,
+	.get_retrans	= __bpf_mptcp_sched_get_retrans,
 	.init		= __bpf_mptcp_sched_init,
 	.release	= __bpf_mptcp_sched_release,
 };
-- 
2.45.2

From: Geliang Tang <tanggeliang@kylinos.cn>

Use new interface get_send() instead of get_subflow().

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 tools/testing/selftests/bpf/progs/mptcp_bpf_first.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

From: Geliang Tang <tanggeliang@kylinos.cn>

Use new interface get_send() instead of get_subflow().

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 tools/testing/selftests/bpf/progs/mptcp_bpf_bkup.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

From: Geliang Tang <tanggeliang@kylinos.cn>

Use new interface get_send() instead of get_subflow().

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 tools/testing/selftests/bpf/progs/mptcp_bpf_rr.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

From: Geliang Tang <tanggeliang@kylinos.cn>

Use new interface get_send() instead of get_subflow().

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 tools/testing/selftests/bpf/progs/mptcp_bpf_red.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

From: Geliang Tang <tanggeliang@kylinos.cn>

Split get_subflow() interface into two: get_send() and get_retrans().

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 .../selftests/bpf/progs/mptcp_bpf_burst.c     | 22 +++++++------------
 1 file changed, 8 insertions(+), 14 deletions(-)

diff --git a/tools/testing/selftests/bpf/progs/mptcp_bpf_burst.c b/tools/testing/selftests/bpf/progs/mptcp_bpf_burst.c
index XXXXXXX..XXXXXXX 100644
--- a/tools/testing/selftests/bpf/progs/mptcp_bpf_burst.c
+++ b/tools/testing/selftests/bpf/progs/mptcp_bpf_burst.c
@@ -XXX,XX +XXX,XX @@ void BPF_PROG(mptcp_sched_burst_release, struct mptcp_sock *msk)
 {
 }
 
-static int bpf_burst_get_send(struct mptcp_sock *msk,
-			      struct mptcp_sched_data *data)
+SEC("struct_ops")
+int BPF_PROG(bpf_burst_get_send, struct mptcp_sock *msk,
+	     struct mptcp_sched_data *data)
 {
 	struct bpf_subflow_send_info send_info[SSK_MODE_MAX];
 	struct mptcp_subflow_context *subflow;
@@ -XXX,XX +XXX,XX @@ static int bpf_burst_get_send(struct mptcp_sock *msk,
 	return 0;
 }
 
-static int bpf_burst_get_retrans(struct mptcp_sock *msk,
-				 struct mptcp_sched_data *data)
+SEC("struct_ops")
+int BPF_PROG(bpf_burst_get_retrans, struct mptcp_sock *msk,
+	     struct mptcp_sched_data *data)
 {
 	int backup = MPTCP_SUBFLOWS_MAX, pick = MPTCP_SUBFLOWS_MAX, subflow_id;
 	struct mptcp_subflow_context *subflow;
@@ -XXX,XX +XXX,XX @@ static int bpf_burst_get_retrans(struct mptcp_sock *msk,
 	return 0;
 }
 
-SEC("struct_ops")
-int BPF_PROG(bpf_burst_get_subflow, struct mptcp_sock *msk,
-	     struct mptcp_sched_data *data)
-{
-	if (data->reinject)
-		return bpf_burst_get_retrans(msk, data);
-	return bpf_burst_get_send(msk, data);
-}
-
 SEC(".struct_ops")
 struct mptcp_sched_ops burst = {
 	.init		= (void *)mptcp_sched_burst_init,
 	.release	= (void *)mptcp_sched_burst_release,
-	.get_subflow	= (void *)bpf_burst_get_subflow,
+	.get_send	= (void *)bpf_burst_get_send,
+	.get_retrans	= (void *)bpf_burst_get_retrans,
 	.name		= "bpf_burst",
 };
-- 
2.45.2

From: Geliang Tang <tanggeliang@kylinos.cn>

Set BPF_F_SLEEPABLE flags for bpf_burst_get_retrans.

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 tools/testing/selftests/bpf/prog_tests/mptcp.c | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)