[PATCH net-next 0/7] XDP redirect tracepoints

[PATCH net-next 0/7] XDP redirect tracepoints

[Jesper Dangaard Brouer]

I feel this is as far as I can take the tracepoint infrastructure to
assist XDP monitoring.

Tracepoints comes with a base overhead of 25 nanosec for an attached
bpf_prog, and 48 nanosec for using a full perf record. This is
problematic for the XDP use-case, but it is very convenient to use the
existing perf infrastructure.

>From a performance perspective, the real solution would be to attach
another bpf_prog (that understand xdp_buff), but I'm not sure we want
to introduce yet another bpf attach API for this.

One thing left is to standardize the possible err return codes, to a
limited set, to allow easier (and faster) mapping into a bpf map.

---

Jesper Dangaard Brouer (7):
      xdp: remove redundant argument to trace_xdp_redirect
      xdp: tracepoint xdp_redirect also need a map argument
      xdp: make xdp tracepoints report bpf prog id instead of prog_tag
      xdp: separate xdp_redirect tracepoint in error case
      xdp: separate xdp_redirect tracepoint in map case
      samples/bpf: xdp_redirect load XDP dummy prog on TX device
      samples/bpf: xdp_monitor tool based on tracepoints


 include/trace/events/xdp.h          |  100 ++++++++++--
 net/core/filter.c                   |   37 +++-
 samples/bpf/Makefile                |    4 
 samples/bpf/xdp_monitor_kern.c      |   88 ++++++++++
 samples/bpf/xdp_monitor_user.c      |  295 +++++++++++++++++++++++++++++++++++
 samples/bpf/xdp_redirect_kern.c     |   11 +
 samples/bpf/xdp_redirect_map_kern.c |   11 +
 samples/bpf/xdp_redirect_map_user.c |   22 ++-
 samples/bpf/xdp_redirect_user.c     |   21 ++
 9 files changed, 543 insertions(+), 46 deletions(-)
 create mode 100644 samples/bpf/xdp_monitor_kern.c
 create mode 100644 samples/bpf/xdp_monitor_user.c

[PATCH net-next 1/7] xdp: remove redundant argument to trace_xdp_redirect

[Jesper Dangaard Brouer]

Supplying the action argument XDP_REDIRECT to the tracepoint xdp_redirect
is redundant as it is only called in-case this action was specified.

Remove the argument, but keep "act" member of the tracepoint struct and
populate it with XDP_REDIRECT.  This makes it easier to write a common bpf_prog
processing events.

Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
---
 include/trace/events/xdp.h |    6 +++---
 net/core/filter.c          |    6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/include/trace/events/xdp.h b/include/trace/events/xdp.h
index 27cf2ef35f5f..f684f3b36bca 100644
--- a/include/trace/events/xdp.h
+++ b/include/trace/events/xdp.h
@@ -52,10 +52,10 @@ TRACE_EVENT(xdp_exception,
 TRACE_EVENT(xdp_redirect,
 
 	TP_PROTO(const struct net_device *dev,
-		 const struct bpf_prog *xdp, u32 act,
+		 const struct bpf_prog *xdp,
 		 int to_index, int err),
 
-	TP_ARGS(dev, xdp, act, to_index, err),
+	TP_ARGS(dev, xdp, to_index, err),
 
 	TP_STRUCT__entry(
 		__array(u8, prog_tag, 8)
@@ -68,7 +68,7 @@ TRACE_EVENT(xdp_redirect,
 	TP_fast_assign(
 		BUILD_BUG_ON(sizeof(__entry->prog_tag) != sizeof(xdp->tag));
 		memcpy(__entry->prog_tag, xdp->tag, sizeof(xdp->tag));
-		__entry->act		= act;
+		__entry->act		= XDP_REDIRECT;
 		__entry->ifindex	= dev->ifindex;
 		__entry->to_index	= to_index;
 		__entry->err		= err;
diff --git a/net/core/filter.c b/net/core/filter.c
index 4bcd6baa80c9..de31fb684ad4 100644
--- a/net/core/filter.c
+++ b/net/core/filter.c
@@ -2524,7 +2524,7 @@ static int xdp_do_redirect_map(struct net_device *dev, struct xdp_buff *xdp,
 	if (likely(!err))
 		ri->map_to_flush = map;
 out:
-	trace_xdp_redirect(dev, xdp_prog, XDP_REDIRECT, index, err);
+	trace_xdp_redirect(dev, xdp_prog, index, err);
 	return err;
 }
 
@@ -2548,7 +2548,7 @@ int xdp_do_redirect(struct net_device *dev, struct xdp_buff *xdp,
 
 	err = __bpf_tx_xdp(fwd, NULL, xdp, 0);
 out:
-	trace_xdp_redirect(dev, xdp_prog, XDP_REDIRECT, index, err);
+	trace_xdp_redirect(dev, xdp_prog, index, err);
 	return err;
 }
 EXPORT_SYMBOL_GPL(xdp_do_redirect);
@@ -2582,7 +2582,7 @@ int xdp_do_generic_redirect(struct net_device *dev, struct sk_buff *skb,
 
 	skb->dev = fwd;
 out:
-	trace_xdp_redirect(dev, xdp_prog, XDP_REDIRECT, index, err);
+	trace_xdp_redirect(dev, xdp_prog, index, err);
 	return err;
 }
 EXPORT_SYMBOL_GPL(xdp_do_generic_redirect);

[PATCH net-next 2/7] xdp: tracepoint xdp_redirect also need a map argument

[Jesper Dangaard Brouer]

To make sense of the map index, the tracepoint user also need to know
that map we are talking about.  Supply the map pointer but only expose
the map->id.

The 'to_index' is renamed 'to_ifindex'.  In the xdp_redirect_map case,
this is the result of the devmap lookup. The map lookup key is exposed
as map_index, which is needed to troubleshoot in case the lookup failed.
The 'to_ifindex' is placed after 'err' to keep TP_STRUCT as common as
possible.

This also keeps the TP_STRUCT similar enough, that userspace can write
a monitor program, that doesn't need to care about whether
bpf_redirect or bpf_redirect_map were used.

Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
---
 include/trace/events/xdp.h |   38 ++++++++++++++++++++++++++++++--------
 net/core/filter.c          |    6 +++---
 2 files changed, 33 insertions(+), 11 deletions(-)

diff --git a/include/trace/events/xdp.h b/include/trace/events/xdp.h
index f684f3b36bca..573dcfa1aeaa 100644
--- a/include/trace/events/xdp.h
+++ b/include/trace/events/xdp.h
@@ -49,20 +49,23 @@ TRACE_EVENT(xdp_exception,
 		  __entry->ifindex)
 );
 
-TRACE_EVENT(xdp_redirect,
+DECLARE_EVENT_CLASS(xdp_redirect_template,
 
 	TP_PROTO(const struct net_device *dev,
 		 const struct bpf_prog *xdp,
-		 int to_index, int err),
+		 int to_ifindex, int err,
+		 const struct bpf_map *map, u32 map_index),
 
-	TP_ARGS(dev, xdp, to_index, err),
+	TP_ARGS(dev, xdp, to_ifindex, err, map, map_index),
 
 	TP_STRUCT__entry(
 		__array(u8, prog_tag, 8)
 		__field(u32, act)
 		__field(int, ifindex)
-		__field(int, to_index)
 		__field(int, err)
+		__field(int, to_ifindex)
+		__field(u32, map_id)
+		__field(int, map_index)
 	),
 
 	TP_fast_assign(
@@ -70,16 +73,35 @@ TRACE_EVENT(xdp_redirect,
 		memcpy(__entry->prog_tag, xdp->tag, sizeof(xdp->tag));
 		__entry->act		= XDP_REDIRECT;
 		__entry->ifindex	= dev->ifindex;
-		__entry->to_index	= to_index;
 		__entry->err		= err;
+		__entry->to_ifindex	= to_ifindex;
+		__entry->map_id		= map ? map->id : 0;
+		__entry->map_index	= map_index;
 	),
 
-	TP_printk("prog=%s action=%s ifindex=%d to_index=%d err=%d",
+	TP_printk("prog=%s action=%s ifindex=%d to_ifindex=%d err=%d"
+		  " map_id=%d map_index=%d",
 		  __print_hex_str(__entry->prog_tag, 8),
 		  __print_symbolic(__entry->act, __XDP_ACT_SYM_TAB),
-		  __entry->ifindex, __entry->to_index,
-		  __entry->err)
+		  __entry->ifindex, __entry->to_ifindex,
+		  __entry->err,
+		  __entry->map_id, __entry->map_index)
 );
+
+DEFINE_EVENT(xdp_redirect_template, xdp_redirect,
+	TP_PROTO(const struct net_device *dev,
+		 const struct bpf_prog *xdp,
+		 int to_ifindex, int err,
+		 const struct bpf_map *map, u32 map_index),
+	TP_ARGS(dev, xdp, to_ifindex, err, map, map_index)
+);
+
+#define _trace_xdp_redirect(dev, xdp, to, err)	\
+	 trace_xdp_redirect(dev, xdp, to, err, NULL, 0);
+
+#define trace_xdp_redirect_map(dev, xdp, fwd, err, map, idx)	\
+	trace_xdp_redirect(dev, xdp, fwd ? fwd->ifindex : 0, err, map, idx);
+
 #endif /* _TRACE_XDP_H */
 
 #include <trace/define_trace.h>
diff --git a/net/core/filter.c b/net/core/filter.c
index de31fb684ad4..31eab77cc842 100644
--- a/net/core/filter.c
+++ b/net/core/filter.c
@@ -2524,7 +2524,7 @@ static int xdp_do_redirect_map(struct net_device *dev, struct xdp_buff *xdp,
 	if (likely(!err))
 		ri->map_to_flush = map;
 out:
-	trace_xdp_redirect(dev, xdp_prog, index, err);
+	trace_xdp_redirect_map(dev, xdp_prog, fwd, err, map, index);
 	return err;
 }
 
@@ -2548,7 +2548,7 @@ int xdp_do_redirect(struct net_device *dev, struct xdp_buff *xdp,
 
 	err = __bpf_tx_xdp(fwd, NULL, xdp, 0);
 out:
-	trace_xdp_redirect(dev, xdp_prog, index, err);
+	_trace_xdp_redirect(dev, xdp_prog, index, err);
 	return err;
 }
 EXPORT_SYMBOL_GPL(xdp_do_redirect);
@@ -2582,7 +2582,7 @@ int xdp_do_generic_redirect(struct net_device *dev, struct sk_buff *skb,
 
 	skb->dev = fwd;
 out:
-	trace_xdp_redirect(dev, xdp_prog, index, err);
+	_trace_xdp_redirect(dev, xdp_prog, index, err);
 	return err;
 }
 EXPORT_SYMBOL_GPL(xdp_do_generic_redirect);

[PATCH net-next 3/7] xdp: make xdp tracepoints report bpf prog id instead of prog_tag

[Jesper Dangaard Brouer]

Given previous patch expose the map_id, it seems natural to also
report the bpf prog id.

Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
---
 include/trace/events/xdp.h |   18 ++++++++----------
 1 file changed, 8 insertions(+), 10 deletions(-)

diff --git a/include/trace/events/xdp.h b/include/trace/events/xdp.h
index 573dcfa1aeaa..89eba564e199 100644
--- a/include/trace/events/xdp.h
+++ b/include/trace/events/xdp.h
@@ -31,20 +31,19 @@ TRACE_EVENT(xdp_exception,
 	TP_ARGS(dev, xdp, act),
 
 	TP_STRUCT__entry(
-		__array(u8, prog_tag, 8)
+		__field(int, prog_id)
 		__field(u32, act)
 		__field(int, ifindex)
 	),
 
 	TP_fast_assign(
-		BUILD_BUG_ON(sizeof(__entry->prog_tag) != sizeof(xdp->tag));
-		memcpy(__entry->prog_tag, xdp->tag, sizeof(xdp->tag));
+		__entry->prog_id	= xdp->aux->id;
 		__entry->act		= act;
 		__entry->ifindex	= dev->ifindex;
 	),
 
-	TP_printk("prog=%s action=%s ifindex=%d",
-		  __print_hex_str(__entry->prog_tag, 8),
+	TP_printk("prog_id=%d action=%s ifindex=%d",
+		  __entry->prog_id,
 		  __print_symbolic(__entry->act, __XDP_ACT_SYM_TAB),
 		  __entry->ifindex)
 );
@@ -59,7 +58,7 @@ DECLARE_EVENT_CLASS(xdp_redirect_template,
 	TP_ARGS(dev, xdp, to_ifindex, err, map, map_index),
 
 	TP_STRUCT__entry(
-		__array(u8, prog_tag, 8)
+		__field(int, prog_id)
 		__field(u32, act)
 		__field(int, ifindex)
 		__field(int, err)
@@ -69,8 +68,7 @@ DECLARE_EVENT_CLASS(xdp_redirect_template,
 	),
 
 	TP_fast_assign(
-		BUILD_BUG_ON(sizeof(__entry->prog_tag) != sizeof(xdp->tag));
-		memcpy(__entry->prog_tag, xdp->tag, sizeof(xdp->tag));
+		__entry->prog_id	= xdp->aux->id;
 		__entry->act		= XDP_REDIRECT;
 		__entry->ifindex	= dev->ifindex;
 		__entry->err		= err;
@@ -79,9 +77,9 @@ DECLARE_EVENT_CLASS(xdp_redirect_template,
 		__entry->map_index	= map_index;
 	),
 
-	TP_printk("prog=%s action=%s ifindex=%d to_ifindex=%d err=%d"
+	TP_printk("prog_id=%d action=%s ifindex=%d to_ifindex=%d err=%d"
 		  " map_id=%d map_index=%d",
-		  __print_hex_str(__entry->prog_tag, 8),
+		  __entry->prog_id,
 		  __print_symbolic(__entry->act, __XDP_ACT_SYM_TAB),
 		  __entry->ifindex, __entry->to_ifindex,
 		  __entry->err,

[PATCH net-next 4/7] xdp: separate xdp_redirect tracepoint in error case

[Jesper Dangaard Brouer]

There is a need to separate the xdp_redirect tracepoint into two
tracepoints, for separating the error case from the normal forward
case.

Due to the extreme speeds XDP is operating at, loading a tracepoint
have a measurable impact.  Single core XDP REDIRECT (ethtool tuned
rx-usecs 25) can do 13.7 Mpps forwarding, but loading a simple
bpf_prog at the tracepoint (with a return 0) reduce perf to 10.2 Mpps
(CPU E5-1650 v4 @ 3.60GHz, driver: ixgbe)

The overhead of loading a bpf-based tracepoint can be calculated to
cost 25 nanosec ((1/13782002-1/10267937)*10^9 = -24.83 ns).

Using perf record on the tracepoint event, with a non-matching --filter
expression, the overhead is much larger. Performance drops to 8.3 Mpps,
cost 48 nanosec ((1/13782002-1/8312497)*10^9 = -47.74))

Having a separate tracepoint for err cases, which should be less
frequent, allow running a continuous monitor for errors while not
affecting the redirect forward performance (this have also been
verified by measurements).

Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
---
 include/trace/events/xdp.h |   22 ++++++++++++++++++----
 net/core/filter.c          |   37 ++++++++++++++++++++++++-------------
 2 files changed, 42 insertions(+), 17 deletions(-)

diff --git a/include/trace/events/xdp.h b/include/trace/events/xdp.h
index 89eba564e199..1eebad55ebd4 100644
--- a/include/trace/events/xdp.h
+++ b/include/trace/events/xdp.h
@@ -94,11 +94,25 @@ DEFINE_EVENT(xdp_redirect_template, xdp_redirect,
 	TP_ARGS(dev, xdp, to_ifindex, err, map, map_index)
 );
 
-#define _trace_xdp_redirect(dev, xdp, to, err)	\
-	 trace_xdp_redirect(dev, xdp, to, err, NULL, 0);
+DEFINE_EVENT(xdp_redirect_template, xdp_redirect_err,
+	TP_PROTO(const struct net_device *dev,
+		 const struct bpf_prog *xdp,
+		 int to_ifindex, int err,
+		 const struct bpf_map *map, u32 map_index),
+	TP_ARGS(dev, xdp, to_ifindex, err, map, map_index)
+);
+
+#define _trace_xdp_redirect(dev, xdp, to)		\
+	 trace_xdp_redirect(dev, xdp, to, 0, NULL, 0);
+
+#define _trace_xdp_redirect_err(dev, xdp, to, err)	\
+	 trace_xdp_redirect_err(dev, xdp, to, err, NULL, 0);
+
+#define trace_xdp_redirect_map(dev, xdp, fwd, map, idx)	\
+	trace_xdp_redirect(dev, xdp, fwd ? fwd->ifindex : 0, 0, map, idx);
 
-#define trace_xdp_redirect_map(dev, xdp, fwd, err, map, idx)	\
-	trace_xdp_redirect(dev, xdp, fwd ? fwd->ifindex : 0, err, map, idx);
+#define trace_xdp_redirect_map_err(dev, xdp, fwd, map, idx, err)	\
+	trace_xdp_redirect_err(dev, xdp, fwd ? fwd->ifindex : 0, err, map, idx);
 
 #endif /* _TRACE_XDP_H */
 
diff --git a/net/core/filter.c b/net/core/filter.c
index 31eab77cc842..096e78de0b97 100644
--- a/net/core/filter.c
+++ b/net/core/filter.c
@@ -2515,16 +2515,20 @@ static int xdp_do_redirect_map(struct net_device *dev, struct xdp_buff *xdp,
 	fwd = __dev_map_lookup_elem(map, index);
 	if (!fwd) {
 		err = -EINVAL;
-		goto out;
+		goto err;
 	}
 	if (ri->map_to_flush && ri->map_to_flush != map)
 		xdp_do_flush_map();
 
 	err = __bpf_tx_xdp(fwd, map, xdp, index);
-	if (likely(!err))
-		ri->map_to_flush = map;
-out:
-	trace_xdp_redirect_map(dev, xdp_prog, fwd, err, map, index);
+	if (unlikely(err))
+		goto err;
+
+	ri->map_to_flush = map;
+	trace_xdp_redirect_map(dev, xdp_prog, fwd, map, index);
+	return 0;
+err:
+	trace_xdp_redirect_map_err(dev, xdp_prog, fwd, map, index, err);
 	return err;
 }
 
@@ -2543,12 +2547,17 @@ int xdp_do_redirect(struct net_device *dev, struct xdp_buff *xdp,
 	ri->ifindex = 0;
 	if (unlikely(!fwd)) {
 		err = -EINVAL;
-		goto out;
+		goto err;
 	}
 
 	err = __bpf_tx_xdp(fwd, NULL, xdp, 0);
-out:
-	_trace_xdp_redirect(dev, xdp_prog, index, err);
+	if (unlikely(err))
+		goto err;
+
+	_trace_xdp_redirect(dev, xdp_prog, index);
+	return 0;
+err:
+	_trace_xdp_redirect_err(dev, xdp_prog, index, err);
 	return err;
 }
 EXPORT_SYMBOL_GPL(xdp_do_redirect);
@@ -2566,23 +2575,25 @@ int xdp_do_generic_redirect(struct net_device *dev, struct sk_buff *skb,
 	ri->ifindex = 0;
 	if (unlikely(!fwd)) {
 		err = -EINVAL;
-		goto out;
+		goto err;
 	}
 
 	if (unlikely(!(fwd->flags & IFF_UP))) {
 		err = -ENETDOWN;
-		goto out;
+		goto err;
 	}
 
 	len = fwd->mtu + fwd->hard_header_len + VLAN_HLEN;
 	if (skb->len > len) {
 		err = -EMSGSIZE;
-		goto out;
+		goto err;
 	}
 
 	skb->dev = fwd;
-out:
-	_trace_xdp_redirect(dev, xdp_prog, index, err);
+	_trace_xdp_redirect(dev, xdp_prog, index);
+	return 0;
+err:
+	_trace_xdp_redirect_err(dev, xdp_prog, index, err);
 	return err;
 }
 EXPORT_SYMBOL_GPL(xdp_do_generic_redirect);

Re: [PATCH net-next 4/7] xdp: separate xdp_redirect tracepoint in error case

[Alexei Starovoitov]

On Tue, Aug 29, 2017 at 04:37:56PM +0200, Jesper Dangaard Brouer wrote:
> There is a need to separate the xdp_redirect tracepoint into two
> tracepoints, for separating the error case from the normal forward
> case.
> 
> Due to the extreme speeds XDP is operating at, loading a tracepoint
> have a measurable impact.  Single core XDP REDIRECT (ethtool tuned
> rx-usecs 25) can do 13.7 Mpps forwarding, but loading a simple
> bpf_prog at the tracepoint (with a return 0) reduce perf to 10.2 Mpps
> (CPU E5-1650 v4 @ 3.60GHz, driver: ixgbe)
> 
> The overhead of loading a bpf-based tracepoint can be calculated to
> cost 25 nanosec ((1/13782002-1/10267937)*10^9 = -24.83 ns).
> 
> Using perf record on the tracepoint event, with a non-matching --filter
> expression, the overhead is much larger. Performance drops to 8.3 Mpps,
> cost 48 nanosec ((1/13782002-1/8312497)*10^9 = -47.74))
> 
> Having a separate tracepoint for err cases, which should be less
> frequent, allow running a continuous monitor for errors while not
> affecting the redirect forward performance (this have also been
> verified by measurements).
> 
> Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>

thanks for detailed analysis of performance impact of the changes.
looks great to me.
Acked-by: Alexei Starovoitov <ast@kernel.org>

[PATCH net-next 5/7] xdp: separate xdp_redirect tracepoint in map case

[Jesper Dangaard Brouer]

Creating as specific xdp_redirect_map variant of the xdp tracepoints
allow users to write simpler/faster BPF progs that get attached to
these tracepoints.

Goal is to still keep the tracepoints in xdp_redirect and xdp_redirect_map
similar enough, that a tool can read the top part of the TP_STRUCT and
produce similar monitor statistics.

Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
---
 include/trace/events/xdp.h |   46 ++++++++++++++++++++++++++++++++++++--------
 net/core/filter.c          |    4 ++--
 2 files changed, 40 insertions(+), 10 deletions(-)

diff --git a/include/trace/events/xdp.h b/include/trace/events/xdp.h
index 1eebad55ebd4..862575ac8da9 100644
--- a/include/trace/events/xdp.h
+++ b/include/trace/events/xdp.h
@@ -77,13 +77,11 @@ DECLARE_EVENT_CLASS(xdp_redirect_template,
 		__entry->map_index	= map_index;
 	),
 
-	TP_printk("prog_id=%d action=%s ifindex=%d to_ifindex=%d err=%d"
-		  " map_id=%d map_index=%d",
+	TP_printk("prog_id=%d action=%s ifindex=%d to_ifindex=%d err=%d",
 		  __entry->prog_id,
 		  __print_symbolic(__entry->act, __XDP_ACT_SYM_TAB),
 		  __entry->ifindex, __entry->to_ifindex,
-		  __entry->err,
-		  __entry->map_id, __entry->map_index)
+		  __entry->err)
 );
 
 DEFINE_EVENT(xdp_redirect_template, xdp_redirect,
@@ -108,11 +106,43 @@ DEFINE_EVENT(xdp_redirect_template, xdp_redirect_err,
 #define _trace_xdp_redirect_err(dev, xdp, to, err)	\
 	 trace_xdp_redirect_err(dev, xdp, to, err, NULL, 0);
 
-#define trace_xdp_redirect_map(dev, xdp, fwd, map, idx)	\
-	trace_xdp_redirect(dev, xdp, fwd ? fwd->ifindex : 0, 0, map, idx);
+DEFINE_EVENT_PRINT(xdp_redirect_template, xdp_redirect_map,
+	TP_PROTO(const struct net_device *dev,
+		 const struct bpf_prog *xdp,
+		 int to_ifindex, int err,
+		 const struct bpf_map *map, u32 map_index),
+	TP_ARGS(dev, xdp, to_ifindex, err, map, map_index),
+	TP_printk("prog_id=%d action=%s ifindex=%d to_ifindex=%d err=%d"
+		  " map_id=%d map_index=%d",
+		  __entry->prog_id,
+		  __print_symbolic(__entry->act, __XDP_ACT_SYM_TAB),
+		  __entry->ifindex, __entry->to_ifindex,
+		  __entry->err,
+		  __entry->map_id, __entry->map_index)
+);
+
+DEFINE_EVENT_PRINT(xdp_redirect_template, xdp_redirect_map_err,
+	TP_PROTO(const struct net_device *dev,
+		 const struct bpf_prog *xdp,
+		 int to_ifindex, int err,
+		 const struct bpf_map *map, u32 map_index),
+	TP_ARGS(dev, xdp, to_ifindex, err, map, map_index),
+	TP_printk("prog_id=%d action=%s ifindex=%d to_ifindex=%d err=%d"
+		  " map_id=%d map_index=%d",
+		  __entry->prog_id,
+		  __print_symbolic(__entry->act, __XDP_ACT_SYM_TAB),
+		  __entry->ifindex, __entry->to_ifindex,
+		  __entry->err,
+		  __entry->map_id, __entry->map_index)
+);
+
+#define _trace_xdp_redirect_map(dev, xdp, fwd, map, idx)		\
+	 trace_xdp_redirect_map(dev, xdp, fwd ? fwd->ifindex : 0,	\
+				0, map, idx);
 
-#define trace_xdp_redirect_map_err(dev, xdp, fwd, map, idx, err)	\
-	trace_xdp_redirect_err(dev, xdp, fwd ? fwd->ifindex : 0, err, map, idx);
+#define _trace_xdp_redirect_map_err(dev, xdp, fwd, map, idx, err)	\
+	 trace_xdp_redirect_map_err(dev, xdp, fwd ? fwd->ifindex : 0,	\
+				    err, map, idx);
 
 #endif /* _TRACE_XDP_H */
 
diff --git a/net/core/filter.c b/net/core/filter.c
index 096e78de0b97..c6a37fe0285b 100644
--- a/net/core/filter.c
+++ b/net/core/filter.c
@@ -2525,10 +2525,10 @@ static int xdp_do_redirect_map(struct net_device *dev, struct xdp_buff *xdp,
 		goto err;
 
 	ri->map_to_flush = map;
-	trace_xdp_redirect_map(dev, xdp_prog, fwd, map, index);
+	_trace_xdp_redirect_map(dev, xdp_prog, fwd, map, index);
 	return 0;
 err:
-	trace_xdp_redirect_map_err(dev, xdp_prog, fwd, map, index, err);
+	_trace_xdp_redirect_map_err(dev, xdp_prog, fwd, map, index, err);
 	return err;
 }
 

[PATCH net-next 6/7] samples/bpf: xdp_redirect load XDP dummy prog on TX device

[Jesper Dangaard Brouer]

For supporting XDP_REDIRECT, a device driver must (obviously)
implement the "TX" function ndo_xdp_xmit().  An additional requirement
is you cannot TX out a device, unless it also have a xdp bpf program
attached. This dependency is caused by the driver code need to setup
XDP resources before it can ndo_xdp_xmit.

Update bpf samples xdp_redirect and xdp_redirect_map to automatically
attach a dummy XDP program to the configured ifindex_out device.  Use
the XDP flag XDP_FLAGS_UPDATE_IF_NOEXIST on the dummy load, to avoid
overriding an existing XDP prog on the device.

Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
---
 samples/bpf/xdp_redirect_kern.c     |   11 ++++++++++-
 samples/bpf/xdp_redirect_map_kern.c |   11 ++++++++++-
 samples/bpf/xdp_redirect_map_user.c |   22 +++++++++++++++-------
 samples/bpf/xdp_redirect_user.c     |   21 +++++++++++++++------
 4 files changed, 50 insertions(+), 15 deletions(-)

diff --git a/samples/bpf/xdp_redirect_kern.c b/samples/bpf/xdp_redirect_kern.c
index a34ad457a684..1c90288d0203 100644
--- a/samples/bpf/xdp_redirect_kern.c
+++ b/samples/bpf/xdp_redirect_kern.c
@@ -26,6 +26,9 @@ struct bpf_map_def SEC("maps") tx_port = {
 	.max_entries = 1,
 };
 
+/* Count RX packets, as XDP bpf_prog doesn't get direct TX-success
+ * feedback.  Redirect TX errors can be caught via a tracepoint.
+ */
 struct bpf_map_def SEC("maps") rxcnt = {
 	.type = BPF_MAP_TYPE_PERCPU_ARRAY,
 	.key_size = sizeof(u32),
@@ -33,7 +36,6 @@ struct bpf_map_def SEC("maps") rxcnt = {
 	.max_entries = 1,
 };
 
-
 static void swap_src_dst_mac(void *data)
 {
 	unsigned short *p = data;
@@ -78,4 +80,11 @@ int xdp_redirect_prog(struct xdp_md *ctx)
 	return bpf_redirect(*ifindex, 0);
 }
 
+/* Redirect require an XDP bpf_prog loaded on the TX device */
+SEC("xdp_redirect_dummy")
+int xdp_redirect_dummy(struct xdp_md *ctx)
+{
+	return XDP_PASS;
+}
+
 char _license[] SEC("license") = "GPL";
diff --git a/samples/bpf/xdp_redirect_map_kern.c b/samples/bpf/xdp_redirect_map_kern.c
index 2faf196e17ea..79795d41ad0d 100644
--- a/samples/bpf/xdp_redirect_map_kern.c
+++ b/samples/bpf/xdp_redirect_map_kern.c
@@ -26,6 +26,9 @@ struct bpf_map_def SEC("maps") tx_port = {
 	.max_entries = 100,
 };
 
+/* Count RX packets, as XDP bpf_prog doesn't get direct TX-success
+ * feedback.  Redirect TX errors can be caught via a tracepoint.
+ */
 struct bpf_map_def SEC("maps") rxcnt = {
 	.type = BPF_MAP_TYPE_PERCPU_ARRAY,
 	.key_size = sizeof(u32),
@@ -33,7 +36,6 @@ struct bpf_map_def SEC("maps") rxcnt = {
 	.max_entries = 1,
 };
 
-
 static void swap_src_dst_mac(void *data)
 {
 	unsigned short *p = data;
@@ -80,4 +82,11 @@ int xdp_redirect_map_prog(struct xdp_md *ctx)
 	return bpf_redirect_map(&tx_port, vport, 0);
 }
 
+/* Redirect require an XDP bpf_prog loaded on the TX device */
+SEC("xdp_redirect_dummy")
+int xdp_redirect_dummy(struct xdp_md *ctx)
+{
+	return XDP_PASS;
+}
+
 char _license[] SEC("license") = "GPL";
diff --git a/samples/bpf/xdp_redirect_map_user.c b/samples/bpf/xdp_redirect_map_user.c
index a1ad00fdaa8a..d4d86a273fba 100644
--- a/samples/bpf/xdp_redirect_map_user.c
+++ b/samples/bpf/xdp_redirect_map_user.c
@@ -16,6 +16,7 @@
 #include <signal.h>
 #include <stdio.h>
 #include <stdlib.h>
+#include <stdbool.h>
 #include <string.h>
 #include <unistd.h>
 #include <libgen.h>
@@ -26,17 +27,18 @@
 
 static int ifindex_in;
 static int ifindex_out;
+static bool ifindex_out_xdp_dummy_attached = true;
 
 static __u32 xdp_flags;
 
 static void int_exit(int sig)
 {
 	set_link_xdp_fd(ifindex_in, -1, xdp_flags);
+	if (ifindex_out_xdp_dummy_attached)
+		set_link_xdp_fd(ifindex_out, -1, xdp_flags);
 	exit(0);
 }
 
-/* simple per-protocol drop counter
- */
 static void poll_stats(int interval, int ifindex)
 {
 	unsigned int nr_cpus = bpf_num_possible_cpus();
@@ -70,7 +72,6 @@ static void usage(const char *prog)
 		prog);
 }
 
-
 int main(int argc, char **argv)
 {
 	const char *optstr = "SN";
@@ -112,14 +113,21 @@ int main(int argc, char **argv)
 		return 1;
 	}
 
-	signal(SIGINT, int_exit);
-	signal(SIGTERM, int_exit);
-
 	if (set_link_xdp_fd(ifindex_in, prog_fd[0], xdp_flags) < 0) {
-		printf("link set xdp fd failed\n");
+		printf("ERROR: link set xdp fd failed on %d\n", ifindex_in);
 		return 1;
 	}
 
+	/* Loading dummy XDP prog on out-device */
+	if (set_link_xdp_fd(ifindex_out, prog_fd[1],
+			    (xdp_flags | XDP_FLAGS_UPDATE_IF_NOEXIST)) < 0) {
+		printf("WARN: link set xdp fd failed on %d\n", ifindex_out);
+		ifindex_out_xdp_dummy_attached = false;
+	}
+
+	signal(SIGINT, int_exit);
+	signal(SIGTERM, int_exit);
+
 	printf("map[0] (vports) = %i, map[1] (map) = %i, map[2] (count) = %i\n",
 		map_fd[0], map_fd[1], map_fd[2]);
 
diff --git a/samples/bpf/xdp_redirect_user.c b/samples/bpf/xdp_redirect_user.c
index f705a1905d2d..4475d837bf2c 100644
--- a/samples/bpf/xdp_redirect_user.c
+++ b/samples/bpf/xdp_redirect_user.c
@@ -16,6 +16,7 @@
 #include <signal.h>
 #include <stdio.h>
 #include <stdlib.h>
+#include <stdbool.h>
 #include <string.h>
 #include <unistd.h>
 #include <libgen.h>
@@ -26,17 +27,18 @@
 
 static int ifindex_in;
 static int ifindex_out;
+static bool ifindex_out_xdp_dummy_attached = true;
 
 static __u32 xdp_flags;
 
 static void int_exit(int sig)
 {
 	set_link_xdp_fd(ifindex_in, -1, xdp_flags);
+	if (ifindex_out_xdp_dummy_attached)
+		set_link_xdp_fd(ifindex_out, -1, xdp_flags);
 	exit(0);
 }
 
-/* simple per-protocol drop counter
- */
 static void poll_stats(int interval, int ifindex)
 {
 	unsigned int nr_cpus = bpf_num_possible_cpus();
@@ -112,14 +114,21 @@ int main(int argc, char **argv)
 		return 1;
 	}
 
-	signal(SIGINT, int_exit);
-	signal(SIGTERM, int_exit);
-
 	if (set_link_xdp_fd(ifindex_in, prog_fd[0], xdp_flags) < 0) {
-		printf("link set xdp fd failed\n");
+		printf("ERROR: link set xdp fd failed on %d\n", ifindex_in);
 		return 1;
 	}
 
+	/* Loading dummy XDP prog on out-device */
+	if (set_link_xdp_fd(ifindex_out, prog_fd[1],
+			    (xdp_flags | XDP_FLAGS_UPDATE_IF_NOEXIST)) < 0) {
+		printf("WARN: link set xdp fd failed on %d\n", ifindex_out);
+		ifindex_out_xdp_dummy_attached = false;
+	}
+
+	signal(SIGINT, int_exit);
+	signal(SIGTERM, int_exit);
+
 	/* bpf redirect port */
 	ret = bpf_map_update_elem(map_fd[0], &key, &ifindex_out, 0);
 	if (ret) {

Re: [PATCH net-next 6/7] samples/bpf: xdp_redirect load XDP dummy prog on TX device

[Tariq Toukan]

Hi Jesper,

On 29/08/2017 5:38 PM, Jesper Dangaard Brouer wrote:
>   
> +/* Redirect require an XDP bpf_prog loaded on the TX device */
> +SEC("xdp_redirect_dummy")
> +int xdp_redirect_dummy(struct xdp_md *ctx)
> +{
> +	return XDP_PASS;
> +}
> +

I get a compilation error related to this:

$ make samples/bpf/

...
LLVM ERROR: 'xdp_redirect_dummy' label emitted multiple times to 
assembly file
make[1]: *** [samples/bpf/xdp_redirect_kern.o] Error 1
make: *** [samples/bpf/] Error 2


It can be fixed by the following patch.
I can submit it in a separate mail if you want to.


diff --git a/samples/bpf/xdp_redirect_kern.c 
b/samples/bpf/xdp_redirect_kern.c
index 1c90288d0203..8abb151e385f 100644
--- a/samples/bpf/xdp_redirect_kern.c
+++ b/samples/bpf/xdp_redirect_kern.c
@@ -82,7 +82,7 @@ int xdp_redirect_prog(struct xdp_md *ctx)

  /* Redirect require an XDP bpf_prog loaded on the TX device */
  SEC("xdp_redirect_dummy")
-int xdp_redirect_dummy(struct xdp_md *ctx)
+int xdp_redirect_dummy_prog(struct xdp_md *ctx)
  {
         return XDP_PASS;
  }
diff --git a/samples/bpf/xdp_redirect_map_kern.c 
b/samples/bpf/xdp_redirect_map_kern.c
index 79795d41ad0d..740a529ba84f 100644
--- a/samples/bpf/xdp_redirect_map_kern.c
+++ b/samples/bpf/xdp_redirect_map_kern.c
@@ -84,7 +84,7 @@ int xdp_redirect_map_prog(struct xdp_md *ctx)

  /* Redirect require an XDP bpf_prog loaded on the TX device */
  SEC("xdp_redirect_dummy")
-int xdp_redirect_dummy(struct xdp_md *ctx)
+int xdp_redirect_dummy_prog(struct xdp_md *ctx)
  {
         return XDP_PASS;
  }


Regards,
Tariq

Re: [PATCH net-next 6/7] samples/bpf: xdp_redirect load XDP dummy prog on TX device

[Jesper Dangaard Brouer]

On Thu, 31 Aug 2017 13:08:22 +0300
Tariq Toukan <tariqt@mellanox.com> wrote:

> Hi Jesper,
> 
> On 29/08/2017 5:38 PM, Jesper Dangaard Brouer wrote:
> >   
> > +/* Redirect require an XDP bpf_prog loaded on the TX device */
> > +SEC("xdp_redirect_dummy")
> > +int xdp_redirect_dummy(struct xdp_md *ctx)
> > +{
> > +	return XDP_PASS;
> > +}
> > +  
> 
> I get a compilation error related to this:
> 
> $ make samples/bpf/
> 
> ...
> LLVM ERROR: 'xdp_redirect_dummy' label emitted multiple times to 
> assembly file
> make[1]: *** [samples/bpf/xdp_redirect_kern.o] Error 1
> make: *** [samples/bpf/] Error 2
> 
> 
> It can be fixed by the following patch.
> I can submit it in a separate mail if you want to.

Ups! - yes please submit an official fix-patch for this!

Below fix looks good to me...

> diff --git a/samples/bpf/xdp_redirect_kern.c 
> b/samples/bpf/xdp_redirect_kern.c
> index 1c90288d0203..8abb151e385f 100644
> --- a/samples/bpf/xdp_redirect_kern.c
> +++ b/samples/bpf/xdp_redirect_kern.c
> @@ -82,7 +82,7 @@ int xdp_redirect_prog(struct xdp_md *ctx)
> 
>   /* Redirect require an XDP bpf_prog loaded on the TX device */
>   SEC("xdp_redirect_dummy")
> -int xdp_redirect_dummy(struct xdp_md *ctx)
> +int xdp_redirect_dummy_prog(struct xdp_md *ctx)
>   {
>          return XDP_PASS;
>   }
> diff --git a/samples/bpf/xdp_redirect_map_kern.c 
> b/samples/bpf/xdp_redirect_map_kern.c
> index 79795d41ad0d..740a529ba84f 100644
> --- a/samples/bpf/xdp_redirect_map_kern.c
> +++ b/samples/bpf/xdp_redirect_map_kern.c
> @@ -84,7 +84,7 @@ int xdp_redirect_map_prog(struct xdp_md *ctx)
> 
>   /* Redirect require an XDP bpf_prog loaded on the TX device */
>   SEC("xdp_redirect_dummy")
> -int xdp_redirect_dummy(struct xdp_md *ctx)
> +int xdp_redirect_dummy_prog(struct xdp_md *ctx)
>   {
>          return XDP_PASS;
>   }

Best regards,
  Jesper Dangaard Brouer
  MSc.CS, Principal Kernel Engineer at Red Hat
  LinkedIn: http://www.linkedin.com/in/brouer

[PATCH net-next 7/7] samples/bpf: xdp_monitor tool based on tracepoints

[Jesper Dangaard Brouer]

This tool xdp_monitor demonstrate how to use the different xdp_redirect
tracepoints xdp_redirect{,_map}{,_err} from a BPF program.

The default mode is to only monitor the error counters, to avoid
affecting the per packet performance. Tracepoints comes with a base
overhead of 25 nanosec for an attached bpf_prog, and 48 nanosec for
using a full perf record (with non-matching filter).  Thus, default
loading the --stats mode could affect the maximum performance.

This version of the tool is very simple and count all types of errors
as one.  It will be natural to extend this later with the different
types of errors that can occur, which should help users quickly
identify common mistakes.

Because the TP_STRUCT was kept in sync all the tracepoints loads the
same BPF code.  It would also be natural to extend the map version to
demonstrate how the map information could be used.

Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
---
 samples/bpf/Makefile           |    4 +
 samples/bpf/xdp_monitor_kern.c |   88 ++++++++++++
 samples/bpf/xdp_monitor_user.c |  295 ++++++++++++++++++++++++++++++++++++++++
 3 files changed, 387 insertions(+)
 create mode 100644 samples/bpf/xdp_monitor_kern.c
 create mode 100644 samples/bpf/xdp_monitor_user.c

diff --git a/samples/bpf/Makefile b/samples/bpf/Makefile
index f1010fe759fe..cf17c7932a6e 100644
--- a/samples/bpf/Makefile
+++ b/samples/bpf/Makefile
@@ -39,6 +39,7 @@ hostprogs-y += per_socket_stats_example
 hostprogs-y += load_sock_ops
 hostprogs-y += xdp_redirect
 hostprogs-y += xdp_redirect_map
+hostprogs-y += xdp_monitor
 hostprogs-y += syscall_tp
 
 # Libbpf dependencies
@@ -83,6 +84,7 @@ test_map_in_map-objs := bpf_load.o $(LIBBPF) test_map_in_map_user.o
 per_socket_stats_example-objs := $(LIBBPF) cookie_uid_helper_example.o
 xdp_redirect-objs := bpf_load.o $(LIBBPF) xdp_redirect_user.o
 xdp_redirect_map-objs := bpf_load.o $(LIBBPF) xdp_redirect_map_user.o
+xdp_monitor-objs := bpf_load.o $(LIBBPF) xdp_monitor_user.o
 syscall_tp-objs := bpf_load.o $(LIBBPF) syscall_tp_user.o
 
 # Tell kbuild to always build the programs
@@ -127,6 +129,7 @@ always += tcp_iw_kern.o
 always += tcp_clamp_kern.o
 always += xdp_redirect_kern.o
 always += xdp_redirect_map_kern.o
+always += xdp_monitor_kern.o
 always += syscall_tp_kern.o
 
 HOSTCFLAGS += -I$(objtree)/usr/include
@@ -166,6 +169,7 @@ HOSTLOADLIBES_xdp_tx_iptunnel += -lelf
 HOSTLOADLIBES_test_map_in_map += -lelf
 HOSTLOADLIBES_xdp_redirect += -lelf
 HOSTLOADLIBES_xdp_redirect_map += -lelf
+HOSTLOADLIBES_xdp_monitor += -lelf
 HOSTLOADLIBES_syscall_tp += -lelf
 
 # Allows pointing LLC/CLANG to a LLVM backend with bpf support, redefine on cmdline:
diff --git a/samples/bpf/xdp_monitor_kern.c b/samples/bpf/xdp_monitor_kern.c
new file mode 100644
index 000000000000..74f3fd8ed729
--- /dev/null
+++ b/samples/bpf/xdp_monitor_kern.c
@@ -0,0 +1,88 @@
+/* XDP monitor tool, based on tracepoints
+ *
+ *  Copyright(c) 2017 Jesper Dangaard Brouer, Red Hat Inc.
+ */
+#include <uapi/linux/bpf.h>
+#include "bpf_helpers.h"
+
+struct bpf_map_def SEC("maps") redirect_err_cnt = {
+	.type = BPF_MAP_TYPE_PERCPU_ARRAY,
+	.key_size = sizeof(u32),
+	.value_size = sizeof(u64),
+	.max_entries = 2,
+	/* TODO: have entries for all possible errno's */
+};
+
+/* Tracepoint format: /sys/kernel/debug/tracing/events/xdp/xdp_redirect/format
+ * Code in:                kernel/include/trace/events/xdp.h
+ */
+struct xdp_redirect_ctx {
+	unsigned short common_type;	//	offset:0;  size:2; signed:0;
+	unsigned char common_flags;	//	offset:2;  size:1; signed:0;
+	unsigned char common_preempt_count;//	offset:3;  size:1; signed:0;
+	int common_pid;			//	offset:4;  size:4; signed:1;
+
+	int prog_id;			//	offset:8;  size:4; signed:1;
+	u32 act;			//	offset:12  size:4; signed:0;
+	int ifindex;			//	offset:16  size:4; signed:1;
+	int err;			//	offset:20  size:4; signed:1;
+	int to_ifindex;			//	offset:24  size:4; signed:1;
+	u32 map_id;			//	offset:28  size:4; signed:0;
+	int map_index;			//	offset:32  size:4; signed:1;
+};					//	offset:36
+
+enum {
+	XDP_REDIRECT_SUCCESS = 0,
+	XDP_REDIRECT_ERROR = 1
+};
+
+static __always_inline
+int xdp_redirect_collect_stat(struct xdp_redirect_ctx *ctx)
+{
+	u32 key = XDP_REDIRECT_ERROR;
+	int err = ctx->err;
+	u64 *cnt;
+
+	if (!err)
+		key = XDP_REDIRECT_SUCCESS;
+
+	cnt  = bpf_map_lookup_elem(&redirect_err_cnt, &key);
+	if (!cnt)
+		return 0;
+	*cnt += 1;
+
+	return 0; /* Indicate event was filtered (no further processing)*/
+	/*
+	 * Returning 1 here would allow e.g. a perf-record tracepoint
+	 * to see and record these events, but it doesn't work well
+	 * in-practice as stopping perf-record also unload this
+	 * bpf_prog.  Plus, there is additional overhead of doing so.
+	 */
+}
+
+SEC("tracepoint/xdp/xdp_redirect_err")
+int trace_xdp_redirect_err(struct xdp_redirect_ctx *ctx)
+{
+	return xdp_redirect_collect_stat(ctx);
+}
+
+
+SEC("tracepoint/xdp/xdp_redirect_map_err")
+int trace_xdp_redirect_map_err(struct xdp_redirect_ctx *ctx)
+{
+	return xdp_redirect_collect_stat(ctx);
+}
+
+/* Likely unloaded when prog starts */
+SEC("tracepoint/xdp/xdp_redirect")
+int trace_xdp_redirect(struct xdp_redirect_ctx *ctx)
+{
+	return xdp_redirect_collect_stat(ctx);
+}
+
+/* Likely unloaded when prog starts */
+SEC("tracepoint/xdp/xdp_redirect_map")
+int trace_xdp_redirect_map(struct xdp_redirect_ctx *ctx)
+{
+	return xdp_redirect_collect_stat(ctx);
+}
diff --git a/samples/bpf/xdp_monitor_user.c b/samples/bpf/xdp_monitor_user.c
new file mode 100644
index 000000000000..b51b4f5e3257
--- /dev/null
+++ b/samples/bpf/xdp_monitor_user.c
@@ -0,0 +1,295 @@
+/* Copyright(c) 2017 Jesper Dangaard Brouer, Red Hat, Inc.
+ */
+static const char *__doc__=
+ "XDP monitor tool, based on tracepoints\n"
+;
+
+static const char *__doc_err_only__=
+ " NOTICE: Only tracking XDP redirect errors\n"
+ "         Enable TX success stats via '--stats'\n"
+ "         (which comes with a per packet processing overhead)\n"
+;
+
+#include <errno.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdbool.h>
+#include <stdint.h>
+#include <string.h>
+#include <ctype.h>
+#include <unistd.h>
+#include <locale.h>
+
+#include <getopt.h>
+#include <net/if.h>
+#include <time.h>
+
+#include "libbpf.h"
+#include "bpf_load.h"
+#include "bpf_util.h"
+
+static int verbose = 1;
+static bool debug = false;
+
+static const struct option long_options[] = {
+	{"help",	no_argument,		NULL, 'h' },
+	{"debug",	no_argument,		NULL, 'D' },
+	{"stats",	no_argument,		NULL, 'S' },
+	{"sec", 	required_argument,	NULL, 's' },
+	{0, 0, NULL,  0 }
+};
+
+static void usage(char *argv[])
+{
+	int i;
+	printf("\nDOCUMENTATION:\n%s\n", __doc__);
+	printf("\n");
+	printf(" Usage: %s (options-see-below)\n",
+	       argv[0]);
+	printf(" Listing options:\n");
+	for (i = 0; long_options[i].name != 0; i++) {
+		printf(" --%-15s", long_options[i].name);
+		if (long_options[i].flag != NULL)
+			printf(" flag (internal value:%d)",
+			       *long_options[i].flag);
+		else
+			printf("(internal short-option: -%c)",
+			       long_options[i].val);
+		printf("\n");
+	}
+	printf("\n");
+}
+
+#define NANOSEC_PER_SEC 1000000000 /* 10^9 */
+__u64 gettime(void)
+{
+	struct timespec t;
+	int res;
+
+	res = clock_gettime(CLOCK_MONOTONIC, &t);
+	if (res < 0) {
+		fprintf(stderr, "Error with gettimeofday! (%i)\n", res);
+		exit(EXIT_FAILURE);
+	}
+	return (__u64) t.tv_sec * NANOSEC_PER_SEC + t.tv_nsec;
+}
+
+enum {
+	REDIR_SUCCESS = 0,
+	REDIR_ERROR = 1,
+};
+#define REDIR_RES_MAX 2
+static const char *redir_names[REDIR_RES_MAX] = {
+	[REDIR_SUCCESS]	= "Success",
+	[REDIR_ERROR]	= "Error",
+};
+static const char *err2str(int err)
+{
+	if (err < REDIR_RES_MAX)
+		return redir_names[err];
+	return NULL;
+}
+
+struct record {
+	__u64 counter;
+	__u64 timestamp;
+};
+
+struct stats_record {
+	struct record xdp_redir[REDIR_RES_MAX];
+};
+
+static void stats_print_headers(bool err_only)
+{
+	if (err_only)
+		printf("\n%s\n", __doc_err_only__);
+
+	printf("%-14s %-10s %-18s %-9s\n",
+	       "XDP_REDIRECT", "pps ", "pps-human-readable", "measure-period");
+}
+
+static void stats_print(struct stats_record *rec,
+			struct stats_record *prev,
+			bool err_only)
+{
+	int i = 0;
+
+	if (err_only)
+		i = REDIR_ERROR;
+
+	for (; i < REDIR_RES_MAX; i++) {
+		struct record *r = &rec->xdp_redir[i];
+		struct record *p = &prev->xdp_redir[i];
+		__u64 period  = 0;
+		__u64 packets = 0;
+		double pps = 0;
+		double period_ = 0;
+
+		if (p->timestamp) {
+			packets = r->counter - p->counter;
+			period  = r->timestamp - p->timestamp;
+			if (period > 0) {
+				period_ = ((double) period / NANOSEC_PER_SEC);
+				pps = packets / period_;
+			}
+		}
+
+		printf("%-14s %-10.0f %'-18.0f %f\n",
+		       err2str(i), pps, pps, period_);
+	}
+}
+
+static __u64 get_key32_value64_percpu(int fd, __u32 key)
+{
+	/* For percpu maps, userspace gets a value per possible CPU */
+	unsigned int nr_cpus = bpf_num_possible_cpus();
+	__u64 values[nr_cpus];
+	__u64 sum = 0;
+	int i;
+
+	if ((bpf_map_lookup_elem(fd, &key, values)) != 0) {
+		fprintf(stderr,
+			"ERR: bpf_map_lookup_elem failed key:0x%X\n", key);
+		return 0;
+	}
+
+	/* Sum values from each CPU */
+	for (i = 0; i < nr_cpus; i++) {
+		sum += values[i];
+	}
+	return sum;
+}
+
+static bool stats_collect(int fd, struct stats_record *rec)
+{
+	int i;
+
+	/* TODO: Detect if someone unloaded the perf event_fd's, as
+	 * this can happen by someone running perf-record -e
+	 */
+
+	for (i = 0; i < REDIR_RES_MAX; i++) {
+		rec->xdp_redir[i].timestamp = gettime();
+		rec->xdp_redir[i].counter = get_key32_value64_percpu(fd, i);
+	}
+	return true;
+}
+
+static void stats_poll(int interval, bool err_only)
+{
+	struct stats_record rec, prev;
+	int map_fd;
+
+	memset(&rec, 0, sizeof(rec));
+
+	/* Trick to pretty printf with thousands separators use %' */
+	setlocale(LC_NUMERIC, "en_US");
+
+	/* Header */
+	if (verbose)
+		printf("\n%s", __doc__);
+
+	/* TODO Need more advanced stats on error types */
+	if (verbose)
+		printf(" - Stats map: %s\n", map_data[0].name);
+	map_fd = map_data[0].fd;
+
+	stats_print_headers(err_only);
+	fflush(stdout);
+
+	while (1) {
+		memcpy(&prev, &rec, sizeof(rec));
+		stats_collect(map_fd, &rec);
+		stats_print(&rec, &prev, err_only);
+		fflush(stdout);
+		sleep(interval);
+	}
+}
+
+void print_bpf_prog_info(void)
+{
+	int i;
+
+	/* Prog info */
+	printf("Loaded BPF prog have %d bpf program(s)\n", prog_cnt);
+	for (i = 0; i < prog_cnt; i++) {
+		printf(" - prog_fd[%d] = fd(%d)\n", i, prog_fd[i]);
+	}
+
+	/* Maps info */
+	printf("Loaded BPF prog have %d map(s)\n", map_data_count);
+	for (i = 0; i < map_data_count; i++) {
+		char *name = map_data[i].name;
+		int fd     = map_data[i].fd;
+
+		printf(" - map_data[%d] = fd(%d) name:%s\n", i, fd, name);
+	}
+
+	/* Event info */
+	printf("Searching for (max:%d) event file descriptor(s)\n", prog_cnt);
+	for (i = 0; i < prog_cnt; i++) {
+		if (event_fd[i] != -1)
+			printf(" - event_fd[%d] = fd(%d)\n", i, event_fd[i]);
+	}
+}
+
+int main(int argc, char **argv)
+{
+	int longindex = 0, opt;
+	int ret = EXIT_SUCCESS;
+	char bpf_obj_file[256];
+
+	/* Default settings: */
+	bool errors_only = true;
+	int interval = 2;
+
+	snprintf(bpf_obj_file, sizeof(bpf_obj_file), "%s_kern.o", argv[0]);
+
+	/* Parse commands line args */
+	while ((opt = getopt_long(argc, argv, "h",
+				  long_options, &longindex)) != -1) {
+		switch (opt) {
+		case 'D':
+			debug = true;
+			break;
+		case 'S':
+			errors_only = false;
+			break;
+		case 's':
+			interval = atoi(optarg);
+			break;
+		case 'h':
+		default:
+			usage(argv);
+			return EXIT_FAILURE;
+		}
+	}
+
+	if (load_bpf_file(bpf_obj_file)) {
+		printf("ERROR - bpf_log_buf: %s", bpf_log_buf);
+		return 1;
+	}
+	if (!prog_fd[0]) {
+		printf("ERROR - load_bpf_file: %s\n", strerror(errno));
+		return 1;
+	}
+
+	if (debug) {
+		print_bpf_prog_info();
+	}
+
+	/* Unload/stop tracepoint event by closing fd's */
+	if (errors_only) {
+		/* The prog_fd[i] and event_fd[i] depend on the
+		 * order the functions was defined in _kern.c
+		 */
+		close(event_fd[2]); /* tracepoint/xdp/xdp_redirect */
+		close(prog_fd[2]);  /* func: trace_xdp_redirect */
+		close(event_fd[3]); /* tracepoint/xdp/xdp_redirect_map */
+		close(prog_fd[3]);  /* func: trace_xdp_redirect_map */
+	}
+
+	stats_poll(interval, errors_only);
+
+	return ret;
+}

Re: [PATCH net-next 7/7] samples/bpf: xdp_monitor tool based on tracepoints

[Alexei Starovoitov]

On Tue, Aug 29, 2017 at 04:38:11PM +0200, Jesper Dangaard Brouer wrote:
> This tool xdp_monitor demonstrate how to use the different xdp_redirect
> tracepoints xdp_redirect{,_map}{,_err} from a BPF program.
> 
> The default mode is to only monitor the error counters, to avoid
> affecting the per packet performance. Tracepoints comes with a base
> overhead of 25 nanosec for an attached bpf_prog, and 48 nanosec for
> using a full perf record (with non-matching filter).  Thus, default
> loading the --stats mode could affect the maximum performance.
> 
> This version of the tool is very simple and count all types of errors
> as one.  It will be natural to extend this later with the different
> types of errors that can occur, which should help users quickly
> identify common mistakes.
> 
> Because the TP_STRUCT was kept in sync all the tracepoints loads the
> same BPF code.  It would also be natural to extend the map version to
> demonstrate how the map information could be used.
> 
> Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>

Nice. Did you consider using libbbpf (instead of old bpf_load.c hack)
and make full standalone tool out of it? Looks very useful.
Acked-by: Alexei Starovoitov <ast@kernel.org>

Re: [PATCH net-next 7/7] samples/bpf: xdp_monitor tool based on tracepoints

[Daniel Borkmann]

On 08/29/2017 07:05 PM, Alexei Starovoitov wrote:
> On Tue, Aug 29, 2017 at 04:38:11PM +0200, Jesper Dangaard Brouer wrote:
>> This tool xdp_monitor demonstrate how to use the different xdp_redirect
>> tracepoints xdp_redirect{,_map}{,_err} from a BPF program.
>>
>> The default mode is to only monitor the error counters, to avoid
>> affecting the per packet performance. Tracepoints comes with a base
>> overhead of 25 nanosec for an attached bpf_prog, and 48 nanosec for
>> using a full perf record (with non-matching filter).  Thus, default
>> loading the --stats mode could affect the maximum performance.
>>
>> This version of the tool is very simple and count all types of errors
>> as one.  It will be natural to extend this later with the different
>> types of errors that can occur, which should help users quickly
>> identify common mistakes.
>>
>> Because the TP_STRUCT was kept in sync all the tracepoints loads the
>> same BPF code.  It would also be natural to extend the map version to
>> demonstrate how the map information could be used.
>>
>> Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
>
> Nice. Did you consider using libbbpf (instead of old bpf_load.c hack)
> and make full standalone tool out of it? Looks very useful.

+1 also my suggestion. ;)

> Acked-by: Alexei Starovoitov <ast@kernel.org>

Re: [PATCH net-next 0/7] XDP redirect tracepoints

[David Miller]

From: Jesper Dangaard Brouer <brouer@redhat.com>
Date: Tue, 29 Aug 2017 16:37:35 +0200

> I feel this is as far as I can take the tracepoint infrastructure to
> assist XDP monitoring.
> 
> Tracepoints comes with a base overhead of 25 nanosec for an attached
> bpf_prog, and 48 nanosec for using a full perf record. This is
> problematic for the XDP use-case, but it is very convenient to use the
> existing perf infrastructure.
> 
>>From a performance perspective, the real solution would be to attach
> another bpf_prog (that understand xdp_buff), but I'm not sure we want
> to introduce yet another bpf attach API for this.
> 
> One thing left is to standardize the possible err return codes, to a
> limited set, to allow easier (and faster) mapping into a bpf map.

Series applied, thanks Jesper.