src/patches/suse-2.6.27.25/patches.trace/lttng-instrumentation-net.patch

   1 From: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
   2 Subject: LTTng instrumentation - net
   3
   4 Original patch header:
   5   LTTng instrumentation - net
   6
   7   Network device activity instrumentation (xmit/receive). Allows to detect when a
   8   packet had arrived on the network card or when it is going to be sent. This is
   9   the instrumentation point outside of the drivers that is the closest to the
  10   hardware. It allows to detect the amount of time taken by a packet to go through
  11   the kernel between the system call and the actual delivery to the network card
  12   (given that system calls are instrumented).
  13
  14   Those tracepoints are used by LTTng.
  15
  16   About the performance impact of tracepoints (which is comparable to markers),
  17   even without immediate values optimizations, tests done by Hideo Aoki on ia64
  18   show no regression. His test case was using hackbench on a kernel where
  19   scheduler instrumentation (about 5 events in code scheduler code) was added.
  20   See the "Tracepoints" patch header for performance result detail.
  21
  22   Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
  23   CC: Andrew Morton <akpm@linux-foundation.org>
  24   CC: netdev@vger.kernel.org
  25   CC: Jeff Garzik <jgarzik@pobox.com>
  26   CC: Masami Hiramatsu <mhiramat@redhat.com>
  27   CC: 'Peter Zijlstra' <peterz@infradead.org>
  28   CC: "Frank Ch. Eigler" <fche@redhat.com>
  29   CC: 'Ingo Molnar' <mingo@elte.hu>
  30   CC: 'Hideo AOKI' <haoki@redhat.com>
  31   CC: Takashi Nishiie <t-nishiie@np.css.fujitsu.com>
  32   CC: 'Steven Rostedt' <rostedt@goodmis.org>
  33   CC: Eduard - Gabriel Munteanu <eduard.munteanu@linux360.ro>
  34
  35 Acked-by: Jan Blunck <jblunck@suse.de>
  36 ---
  37 ---
  38  include/trace/net.h |   14 ++++++++++++++
  39  net/core/dev.c      |    3 +++
  40  2 files changed, 17 insertions(+)
  41
  42 --- /dev/null
  43 +++ b/include/trace/net.h
  44 @@ -0,0 +1,14 @@
  45 +#ifndef _TRACE_NET_H
  46 +#define _TRACE_NET_H
  47 +
  48 +#include <net/sock.h>
  49 +#include <linux/tracepoint.h>
  50 +
  51 +DEFINE_TRACE(net_dev_xmit,
  52 +       TPPROTO(struct sk_buff *skb),
  53 +       TPARGS(skb));
  54 +DEFINE_TRACE(net_dev_receive,
  55 +       TPPROTO(struct sk_buff *skb),
  56 +       TPARGS(skb));
  57 +
  58 +#endif
  59 --- a/net/core/dev.c
  60 +++ b/net/core/dev.c
  61 @@ -127,6 +127,7 @@
  62  #include <linux/in.h>
  63  #include <linux/jhash.h>
  64  #include <linux/random.h>
  65 +#include <trace/net.h>
  66
  67  #include "net-sysfs.h"
  68
  69 @@ -1794,6 +1795,7 @@ int dev_queue_xmit(struct sk_buff *skb)
  70         }
  71
  72  gso:
  73 +       trace_net_dev_xmit(skb);
  74         /* Disable soft irqs for various locks below. Also
  75          * stops preemption for RCU.
  76          */
  77 @@ -2215,6 +2217,7 @@ int netif_receive_skb(struct sk_buff *sk
  78
  79         __get_cpu_var(netdev_rx_stat).total++;
  80
  81 +       trace_net_dev_receive(skb);
  82         skb_reset_network_header(skb);
  83         skb_reset_transport_header(skb);
  84         skb->mac_len = skb->network_header - skb->mac_header;