Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   Yafang Shao <laoar.shao@gmail.com>
To:     davem@davemloft.net, satoru.moriya@hds.com
Cc:     netdev@vger.kernel.org, linux-kernel@vger.kernel.org,
        Yafang Shao <laoar.shao@gmail.com>
Subject: [PATCH] net: expose sk wmem in sock_exceed_buf_limit tracepoint
Date:   Sun,  1 Jul 2018 23:31:30 +0800
Message-Id: <1530459090-11037-1-git-send-email-laoar.shao@gmail.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Currently trace_sock_exceed_buf_limit() only show rmem info,
but wmem limit may also be hit.
So expose wmem info in this tracepoint as well.

Regarding memcg, I think it is better to introduce a new tracepoint(if
that is needed), i.e. trace_memcg_limit_hit other than show memcg info in
trace_sock_exceed_buf_limit.

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
---
 include/trace/events/sock.h | 30 +++++++++++++++++++++++++-----
 net/core/sock.c             |  6 ++++--
 2 files changed, 29 insertions(+), 7 deletions(-)

diff --git a/include/trace/events/sock.h b/include/trace/events/sock.h
index 3176a39..a0c4b8a 100644
--- a/include/trace/events/sock.h
+++ b/include/trace/events/sock.h
@@ -35,6 +35,10 @@
 		EM(TCP_CLOSING)			\
 		EMe(TCP_NEW_SYN_RECV)
 
+#define skmem_kind_names			\
+		EM(SK_MEM_SEND)			\
+		EMe(SK_MEM_RECV)
+
 /* enums need to be exported to user space */
 #undef EM
 #undef EMe
@@ -44,6 +48,7 @@
 family_names
 inet_protocol_names
 tcp_state_names
+skmem_kind_names
 
 #undef EM
 #undef EMe
@@ -59,6 +64,9 @@
 #define show_tcp_state_name(val)        \
 	__print_symbolic(val, tcp_state_names)
 
+#define show_skmem_kind_names(val)	\
+	__print_symbolic(val, skmem_kind_names)
+
 TRACE_EVENT(sock_rcvqueue_full,
 
 	TP_PROTO(struct sock *sk, struct sk_buff *skb),
@@ -83,9 +91,9 @@
 
 TRACE_EVENT(sock_exceed_buf_limit,
 
-	TP_PROTO(struct sock *sk, struct proto *prot, long allocated),
+	TP_PROTO(struct sock *sk, struct proto *prot, long allocated, int kind),
 
-	TP_ARGS(sk, prot, allocated),
+	TP_ARGS(sk, prot, allocated, kind),
 
 	TP_STRUCT__entry(
 		__array(char, name, 32)
@@ -93,6 +101,10 @@
 		__field(long, allocated)
 		__field(int, sysctl_rmem)
 		__field(int, rmem_alloc)
+		__field(int, sysctl_wmem)
+		__field(int, wmem_alloc)
+		__field(int, wmem_queued)
+		__field(int, kind)
 	),
 
 	TP_fast_assign(
@@ -101,17 +113,25 @@
 		__entry->allocated = allocated;
 		__entry->sysctl_rmem = sk_get_rmem0(sk, prot);
 		__entry->rmem_alloc = atomic_read(&sk->sk_rmem_alloc);
+		__entry->sysctl_wmem = sk_get_wmem0(sk, prot);
+		__entry->wmem_alloc = refcount_read(&sk->sk_wmem_alloc);
+		__entry->wmem_queued = sk->sk_wmem_queued;
+		__entry->kind = kind;
 	),
 
-	TP_printk("proto:%s sysctl_mem=%ld,%ld,%ld allocated=%ld "
-		"sysctl_rmem=%d rmem_alloc=%d",
+	TP_printk("proto:%s sysctl_mem=%ld,%ld,%ld allocated=%ld sysctl_rmem=%d rmem_alloc=%d sysctl_wmem=%d wmem_alloc=%d wmem_queued=%d kind=%s",
 		__entry->name,
 		__entry->sysctl_mem[0],
 		__entry->sysctl_mem[1],
 		__entry->sysctl_mem[2],
 		__entry->allocated,
 		__entry->sysctl_rmem,
-		__entry->rmem_alloc)
+		__entry->rmem_alloc,
+		__entry->sysctl_wmem,
+		__entry->wmem_alloc,
+		__entry->wmem_queued,
+		show_skmem_kind_names(__entry->kind)
+	)
 );
 
 TRACE_EVENT(inet_sock_set_state,
diff --git a/net/core/sock.c b/net/core/sock.c
index bcc4182..65350ed 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -2401,9 +2401,10 @@ int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind)
 {
 	struct proto *prot = sk->sk_prot;
 	long allocated = sk_memory_allocated_add(sk, amt);
+	bool charged = true;
 
 	if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
-	    !mem_cgroup_charge_skmem(sk->sk_memcg, amt))
+	    !(charged = mem_cgroup_charge_skmem(sk->sk_memcg, amt)))
 		goto suppress_allocation;
 
 	/* Under limit. */
@@ -2461,7 +2462,8 @@ int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind)
 			return 1;
 	}
 
-	trace_sock_exceed_buf_limit(sk, prot, allocated);
+	if (kind == SK_MEM_SEND || (kind == SK_MEM_RECV && charged))
+		trace_sock_exceed_buf_limit(sk, prot, allocated, kind);
 
 	sk_memory_allocated_sub(sk, amt);
 
-- 
1.8.3.1