[PATCH net-next 0/2] net: skb: introduce kfree_skb_with_reason()

[PATCH net-next 0/2] net: skb: introduce kfree_skb_with_reason()

[menglong8.dong]

From: Menglong Dong <imagedong@tencent.com>

In this series patch, the interface kfree_skb_with_reason() is
introduced(), which is used to collect skb drop reason, and pass
it to 'kfree_skb' tracepoint. Therefor, 'drop_monitor' or eBPF is
able to monitor abnormal skb with detail reason.

In fact, this series patches are out of the intelligence of David
and Steve, I'm just a truck man :/

Previous discussion is here:

https://lore.kernel.org/netdev/20211118105752.1d46e990@gandalf.local.home/
https://lore.kernel.org/netdev/67b36bd8-2477-88ac-83a0-35a1eeaf40c9@gmail.com/

In the first patch, kfree_skb_with_reason() is introduced and
the 'reason' field is added to 'kfree_skb' tracepoint. In the
second patch, 'kfree_skb()' in replaced with 'kfree_skb_with_reason()'
in tcp_v4_rcv().


Menglong Dong (2):
  net: skb: introduce kfree_skb_with_reason()
  net: skb: use kfree_skb_with_reason() in tcp_v4_rcv()

 include/linux/skbuff.h     | 16 ++++++++++++++++
 include/trace/events/skb.h | 39 +++++++++++++++++++++++++++++++-------
 net/core/dev.c             |  3 ++-
 net/core/drop_monitor.c    | 10 +++++++---
 net/core/skbuff.c          | 22 ++++++++++++++++++++-
 net/ipv4/tcp_ipv4.c        | 10 ++++++++--
 6 files changed, 86 insertions(+), 14 deletions(-)

-- 
2.27.0

[PATCH net-next 1/2] net: skb: introduce kfree_skb_with_reason()

[menglong8.dong]

From: Menglong Dong <imagedong@tencent.com>

Introduce the interface kfree_skb_with_reason(), which is used to pass
the reason why the skb is dropped to 'kfree_skb' tracepoint.

Add the 'reason' field to 'trace_kfree_skb', therefor user can get
more detail information about abnormal skb with 'drop_monitor' or
eBPF.

Signed-off-by: Menglong Dong <imagedong@tencent.com>
---
 include/linux/skbuff.h     | 13 +++++++++++++
 include/trace/events/skb.h | 36 +++++++++++++++++++++++++++++-------
 net/core/dev.c             |  3 ++-
 net/core/drop_monitor.c    | 10 +++++++---
 net/core/skbuff.c          | 22 +++++++++++++++++++++-
 5 files changed, 72 insertions(+), 12 deletions(-)

diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index aa9d42724e20..3620b3ff2154 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -305,6 +305,17 @@ struct sk_buff_head {
 
 struct sk_buff;
 
+/* The reason of skb drop, which is used in kfree_skb_with_reason().
+ * en...maybe they should be splited by group?
+ *
+ * Each item here should also be in 'TRACE_SKB_DROP_REASON', which is
+ * used to translate the reason to string.
+ */
+enum skb_drop_reason {
+	SKB_DROP_REASON_NOT_SPECIFIED,
+	SKB_DROP_REASON_MAX,
+};
+
 /* To allow 64K frame to be packed as single skb without frag_list we
  * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
  * buffers which do not start on a page boundary.
@@ -1087,6 +1098,8 @@ static inline bool skb_unref(struct sk_buff *skb)
 
 void skb_release_head_state(struct sk_buff *skb);
 void kfree_skb(struct sk_buff *skb);
+void kfree_skb_with_reason(struct sk_buff *skb,
+			   enum skb_drop_reason reason);
 void kfree_skb_list(struct sk_buff *segs);
 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
 void skb_tx_error(struct sk_buff *skb);
diff --git a/include/trace/events/skb.h b/include/trace/events/skb.h
index 9e92f22eb086..cab1c08a30cd 100644
--- a/include/trace/events/skb.h
+++ b/include/trace/events/skb.h
@@ -9,29 +9,51 @@
 #include <linux/netdevice.h>
 #include <linux/tracepoint.h>
 
+#define TRACE_SKB_DROP_REASON					\
+	EM(SKB_DROP_REASON_NOT_SPECIFIED, NOT_SPECIFIED)	\
+	EMe(SKB_DROP_REASON_MAX, HAHA_MAX)
+
+#undef EM
+#undef EMe
+
+#define EM(a, b)	TRACE_DEFINE_ENUM(a);
+#define EMe(a, b)	TRACE_DEFINE_ENUM(a);
+
+TRACE_SKB_DROP_REASON
+
+#undef EM
+#undef EMe
+#define EM(a, b)	{ a, #b },
+#define EMe(a, b)	{ a, #b }
+
+
 /*
  * Tracepoint for free an sk_buff:
  */
 TRACE_EVENT(kfree_skb,
 
-	TP_PROTO(struct sk_buff *skb, void *location),
+	TP_PROTO(struct sk_buff *skb, void *location,
+		 enum skb_drop_reason reason),
 
-	TP_ARGS(skb, location),
+	TP_ARGS(skb, location, reason),
 
 	TP_STRUCT__entry(
-		__field(	void *,		skbaddr		)
-		__field(	void *,		location	)
-		__field(	unsigned short,	protocol	)
+		__field(void *,		skbaddr)
+		__field(void *,		location)
+		__field(unsigned short,	protocol)
+		__field(enum skb_drop_reason,	reason)
 	),
 
 	TP_fast_assign(
 		__entry->skbaddr = skb;
 		__entry->location = location;
 		__entry->protocol = ntohs(skb->protocol);
+		__entry->reason = reason;
 	),
 
-	TP_printk("skbaddr=%p protocol=%u location=%p",
-		__entry->skbaddr, __entry->protocol, __entry->location)
+	TP_printk("skbaddr=%p protocol=%u location=%p reason: %s",
+		__entry->skbaddr, __entry->protocol, __entry->location,
+		__print_symbolic(__entry->reason, TRACE_SKB_DROP_REASON))
 );
 
 TRACE_EVENT(consume_skb,
diff --git a/net/core/dev.c b/net/core/dev.c
index 644b9c8be3a8..9464dbf9e3d6 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -4899,7 +4899,8 @@ static __latent_entropy void net_tx_action(struct softirq_action *h)
 			if (likely(get_kfree_skb_cb(skb)->reason == SKB_REASON_CONSUMED))
 				trace_consume_skb(skb);
 			else
-				trace_kfree_skb(skb, net_tx_action);
+				trace_kfree_skb(skb, net_tx_action,
+						SKB_DROP_REASON_NOT_SPECIFIED);
 
 			if (skb->fclone != SKB_FCLONE_UNAVAILABLE)
 				__kfree_skb(skb);
diff --git a/net/core/drop_monitor.c b/net/core/drop_monitor.c
index 3d0ab2eec916..7b288a121a41 100644
--- a/net/core/drop_monitor.c
+++ b/net/core/drop_monitor.c
@@ -110,7 +110,8 @@ static u32 net_dm_queue_len = 1000;
 
 struct net_dm_alert_ops {
 	void (*kfree_skb_probe)(void *ignore, struct sk_buff *skb,
-				void *location);
+				void *location,
+				enum skb_drop_reason reason);
 	void (*napi_poll_probe)(void *ignore, struct napi_struct *napi,
 				int work, int budget);
 	void (*work_item_func)(struct work_struct *work);
@@ -262,7 +263,9 @@ static void trace_drop_common(struct sk_buff *skb, void *location)
 	spin_unlock_irqrestore(&data->lock, flags);
 }
 
-static void trace_kfree_skb_hit(void *ignore, struct sk_buff *skb, void *location)
+static void trace_kfree_skb_hit(void *ignore, struct sk_buff *skb,
+				void *location,
+				enum skb_drop_reason reason)
 {
 	trace_drop_common(skb, location);
 }
@@ -490,7 +493,8 @@ static const struct net_dm_alert_ops net_dm_alert_summary_ops = {
 
 static void net_dm_packet_trace_kfree_skb_hit(void *ignore,
 					      struct sk_buff *skb,
-					      void *location)
+					      void *location,
+					      enum skb_drop_reason reason)
 {
 	ktime_t tstamp = ktime_get_real();
 	struct per_cpu_dm_data *data;
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index 275f7b8416fe..570dc022a8a1 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -770,11 +770,31 @@ void kfree_skb(struct sk_buff *skb)
 	if (!skb_unref(skb))
 		return;
 
-	trace_kfree_skb(skb, __builtin_return_address(0));
+	trace_kfree_skb(skb, __builtin_return_address(0),
+			SKB_DROP_REASON_NOT_SPECIFIED);
 	__kfree_skb(skb);
 }
 EXPORT_SYMBOL(kfree_skb);
 
+/**
+ *	kfree_skb_with_reason - free an sk_buff with reason
+ *	@skb: buffer to free
+ *	@reason: reason why this skb is dropped
+ *
+ *	The same as kfree_skb() except that this function will pass
+ *	the drop reason to 'kfree_skb' tracepoint.
+ */
+void kfree_skb_with_reason(struct sk_buff *skb,
+			   enum skb_drop_reason reason)
+{
+	if (!skb_unref(skb))
+		return;
+
+	trace_kfree_skb(skb, __builtin_return_address(0), reason);
+	__kfree_skb(skb);
+}
+EXPORT_SYMBOL(kfree_skb_with_reason);
+
 void kfree_skb_list(struct sk_buff *segs)
 {
 	while (segs) {
-- 
2.30.2

[PATCH net-next 2/2] net: skb: use kfree_skb_with_reason() in tcp_v4_rcv()

[menglong8.dong]

From: Menglong Dong <imagedong@tencent.com>

Replace kfree_skb() with kfree_skb_with_reason() in tcp_v4_rcv().
Following drop reason are added:

SKB_DROP_REASON_NO_SOCK
SKB_DROP_REASON_BAD_PACKET
SKB_DROP_REASON_TCP_CSUM

After this patch, 'kfree_skb' event will print message like this:

$           TASK-PID     CPU#  |||||  TIMESTAMP  FUNCTION
$              | |         |   |||||     |         |
          <idle>-0       [000] ..s1.    36.113438: kfree_skb: skbaddr=(____ptrval____) protocol=2048 location=(____ptrval____) reason: NO_SOCK

The reason of skb drop is printed too.

Signed-off-by: Menglong Dong <imagedong@tencent.com>
---
 include/linux/skbuff.h     |  3 +++
 include/trace/events/skb.h |  3 +++
 net/ipv4/tcp_ipv4.c        | 10 ++++++++--
 3 files changed, 14 insertions(+), 2 deletions(-)

diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index 3620b3ff2154..f85db6c035d1 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -313,6 +313,9 @@ struct sk_buff;
  */
 enum skb_drop_reason {
 	SKB_DROP_REASON_NOT_SPECIFIED,
+	SKB_DROP_REASON_NO_SOCK,
+	SKB_DROP_REASON_BAD_PACKET,
+	SKB_DROP_REASON_TCP_CSUM,
 	SKB_DROP_REASON_MAX,
 };
 
diff --git a/include/trace/events/skb.h b/include/trace/events/skb.h
index cab1c08a30cd..b9ea6b4ed7ec 100644
--- a/include/trace/events/skb.h
+++ b/include/trace/events/skb.h
@@ -11,6 +11,9 @@
 
 #define TRACE_SKB_DROP_REASON					\
 	EM(SKB_DROP_REASON_NOT_SPECIFIED, NOT_SPECIFIED)	\
+	EM(SKB_DROP_REASON_NO_SOCK, NO_SOCK)			\
+	EM(SKB_DROP_REASON_BAD_PACKET, BAD_PACKET)		\
+	EM(SKB_DROP_REASON_TCP_CSUM, TCP_CSUM)			\
 	EMe(SKB_DROP_REASON_MAX, HAHA_MAX)
 
 #undef EM
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index ac10e4cdd8d0..03dc4c79b84b 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -1971,8 +1971,10 @@ int tcp_v4_rcv(struct sk_buff *skb)
 	const struct tcphdr *th;
 	bool refcounted;
 	struct sock *sk;
+	int drop_reason;
 	int ret;
 
+	drop_reason = 0;
 	if (skb->pkt_type != PACKET_HOST)
 		goto discard_it;
 
@@ -1984,8 +1986,10 @@ int tcp_v4_rcv(struct sk_buff *skb)
 
 	th = (const struct tcphdr *)skb->data;
 
-	if (unlikely(th->doff < sizeof(struct tcphdr) / 4))
+	if (unlikely(th->doff < sizeof(struct tcphdr) / 4)) {
+		drop_reason = SKB_DROP_REASON_BAD_PACKET;
 		goto bad_packet;
+	}
 	if (!pskb_may_pull(skb, th->doff * 4))
 		goto discard_it;
 
@@ -2124,6 +2128,7 @@ int tcp_v4_rcv(struct sk_buff *skb)
 	return ret;
 
 no_tcp_socket:
+	drop_reason = SKB_DROP_REASON_NO_SOCK;
 	if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 		goto discard_it;
 
@@ -2131,6 +2136,7 @@ int tcp_v4_rcv(struct sk_buff *skb)
 
 	if (tcp_checksum_complete(skb)) {
 csum_error:
+		drop_reason = SKB_DROP_REASON_TCP_CSUM;
 		trace_tcp_bad_csum(skb);
 		__TCP_INC_STATS(net, TCP_MIB_CSUMERRORS);
 bad_packet:
@@ -2141,7 +2147,7 @@ int tcp_v4_rcv(struct sk_buff *skb)
 
 discard_it:
 	/* Discard frame. */
-	kfree_skb(skb);
+	kfree_skb_with_reason(skb, drop_reason);
 	return 0;
 
 discard_and_relse:
-- 
2.30.2

Re: [PATCH net-next 2/2] net: skb: use kfree_skb_with_reason() in tcp_v4_rcv()

[David Ahern]

On 12/29/21 7:32 AM, menglong8.dong@gmail.com wrote:
> From: Menglong Dong <imagedong@tencent.com>
> 
> Replace kfree_skb() with kfree_skb_with_reason() in tcp_v4_rcv().
> Following drop reason are added:
> 
> SKB_DROP_REASON_NO_SOCK
> SKB_DROP_REASON_BAD_PACKET
> SKB_DROP_REASON_TCP_CSUM
> 
> After this patch, 'kfree_skb' event will print message like this:
> 
> $           TASK-PID     CPU#  |||||  TIMESTAMP  FUNCTION
> $              | |         |   |||||     |         |
>           <idle>-0       [000] ..s1.    36.113438: kfree_skb: skbaddr=(____ptrval____) protocol=2048 location=(____ptrval____) reason: NO_SOCK
> 
> The reason of skb drop is printed too.
> 
> Signed-off-by: Menglong Dong <imagedong@tencent.com>
> ---
>  include/linux/skbuff.h     |  3 +++
>  include/trace/events/skb.h |  3 +++
>  net/ipv4/tcp_ipv4.c        | 10 ++++++++--

your first patch set was targeting UDP and now you are starting with tcp?


>  3 files changed, 14 insertions(+), 2 deletions(-)
> 
> diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
> index 3620b3ff2154..f85db6c035d1 100644
> --- a/include/linux/skbuff.h
> +++ b/include/linux/skbuff.h
> @@ -313,6 +313,9 @@ struct sk_buff;
>   */
>  enum skb_drop_reason {
>  	SKB_DROP_REASON_NOT_SPECIFIED,
> +	SKB_DROP_REASON_NO_SOCK,

SKB_DROP_REASON_NO_SOCKET

> +	SKB_DROP_REASON_BAD_PACKET,

SKB_DROP_REASON_PKT_TOO_SMALL

User oriented messages, not code based.


> +	SKB_DROP_REASON_TCP_CSUM,
>  	SKB_DROP_REASON_MAX,
>  };
>  
> diff --git a/include/trace/events/skb.h b/include/trace/events/skb.h
> index cab1c08a30cd..b9ea6b4ed7ec 100644
> --- a/include/trace/events/skb.h
> +++ b/include/trace/events/skb.h
> @@ -11,6 +11,9 @@
>  
>  #define TRACE_SKB_DROP_REASON					\
>  	EM(SKB_DROP_REASON_NOT_SPECIFIED, NOT_SPECIFIED)	\
> +	EM(SKB_DROP_REASON_NO_SOCK, NO_SOCK)			\
> +	EM(SKB_DROP_REASON_BAD_PACKET, BAD_PACKET)		\
> +	EM(SKB_DROP_REASON_TCP_CSUM, TCP_CSUM)			\
>  	EMe(SKB_DROP_REASON_MAX, HAHA_MAX)
>  
>  #undef EM
> diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
> index ac10e4cdd8d0..03dc4c79b84b 100644
> --- a/net/ipv4/tcp_ipv4.c
> +++ b/net/ipv4/tcp_ipv4.c
> @@ -1971,8 +1971,10 @@ int tcp_v4_rcv(struct sk_buff *skb)
>  	const struct tcphdr *th;
>  	bool refcounted;
>  	struct sock *sk;
> +	int drop_reason;
>  	int ret;
>  
> +	drop_reason = 0;

	drop_reason = SKB_DROP_REASON_NOT_SPECIFIED;

Re: [PATCH net-next 2/2] net: skb: use kfree_skb_with_reason() in tcp_v4_rcv()

[Menglong Dong]

On Thu, Dec 30, 2021 at 4:18 AM David Ahern <dsahern@gmail.com> wrote:
>
> On 12/29/21 7:32 AM, menglong8.dong@gmail.com wrote:
> > From: Menglong Dong <imagedong@tencent.com>
> >
> > Replace kfree_skb() with kfree_skb_with_reason() in tcp_v4_rcv().
> > Following drop reason are added:
> >
> > SKB_DROP_REASON_NO_SOCK
> > SKB_DROP_REASON_BAD_PACKET
> > SKB_DROP_REASON_TCP_CSUM
> >
> > After this patch, 'kfree_skb' event will print message like this:
> >
> > $           TASK-PID     CPU#  |||||  TIMESTAMP  FUNCTION
> > $              | |         |   |||||     |         |
> >           <idle>-0       [000] ..s1.    36.113438: kfree_skb: skbaddr=(____ptrval____) protocol=2048 location=(____ptrval____) reason: NO_SOCK
> >
> > The reason of skb drop is printed too.
> >
> > Signed-off-by: Menglong Dong <imagedong@tencent.com>
> > ---
> >  include/linux/skbuff.h     |  3 +++
> >  include/trace/events/skb.h |  3 +++
> >  net/ipv4/tcp_ipv4.c        | 10 ++++++++--
>
> your first patch set was targeting UDP and now you are starting with tcp?

Yeah, I think TCP is used more, which can be a good starting point. After
all, general protocols are all my target.

>
>
> >  3 files changed, 14 insertions(+), 2 deletions(-)
> >
> > diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
> > index 3620b3ff2154..f85db6c035d1 100644
> > --- a/include/linux/skbuff.h
> > +++ b/include/linux/skbuff.h
> > @@ -313,6 +313,9 @@ struct sk_buff;
> >   */
> >  enum skb_drop_reason {
> >       SKB_DROP_REASON_NOT_SPECIFIED,
> > +     SKB_DROP_REASON_NO_SOCK,
>
> SKB_DROP_REASON_NO_SOCKET
>
> > +     SKB_DROP_REASON_BAD_PACKET,
>
> SKB_DROP_REASON_PKT_TOO_SMALL
>
> User oriented messages, not code based.
>

Ok, get it!

Thanks!
Menglong Dong

>
> > +     SKB_DROP_REASON_TCP_CSUM,
> >       SKB_DROP_REASON_MAX,
> >  };
> >
> > diff --git a/include/trace/events/skb.h b/include/trace/events/skb.h
> > index cab1c08a30cd..b9ea6b4ed7ec 100644
> > --- a/include/trace/events/skb.h
> > +++ b/include/trace/events/skb.h
> > @@ -11,6 +11,9 @@
> >
> >  #define TRACE_SKB_DROP_REASON                                        \
> >       EM(SKB_DROP_REASON_NOT_SPECIFIED, NOT_SPECIFIED)        \
> > +     EM(SKB_DROP_REASON_NO_SOCK, NO_SOCK)                    \
> > +     EM(SKB_DROP_REASON_BAD_PACKET, BAD_PACKET)              \
> > +     EM(SKB_DROP_REASON_TCP_CSUM, TCP_CSUM)                  \
> >       EMe(SKB_DROP_REASON_MAX, HAHA_MAX)
> >
> >  #undef EM
> > diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
> > index ac10e4cdd8d0..03dc4c79b84b 100644
> > --- a/net/ipv4/tcp_ipv4.c
> > +++ b/net/ipv4/tcp_ipv4.c
> > @@ -1971,8 +1971,10 @@ int tcp_v4_rcv(struct sk_buff *skb)
> >       const struct tcphdr *th;
> >       bool refcounted;
> >       struct sock *sk;
> > +     int drop_reason;
> >       int ret;
> >
> > +     drop_reason = 0;
>
>         drop_reason = SKB_DROP_REASON_NOT_SPECIFIED;
>