tracing: pass around ring buffer instead of tracer
authorSteven Rostedt <srostedt@redhat.com>
Wed, 2 Sep 2009 18:17:06 +0000 (14:17 -0400)
committerSteven Rostedt <rostedt@goodmis.org>
Fri, 4 Sep 2009 22:59:39 +0000 (18:59 -0400)
commite77405ad80f53966524b5c31244e13fbbbecbd84
tree65c05f9e1573e9958e52bb72655e00c8592aacd2
parentf633903af2ceb0cec07d45e499a072b6593d0ed1
tracing: pass around ring buffer instead of tracer

The latency tracers (irqsoff and wakeup) can swap trace buffers
on the fly. If an event is happening and has reserved data on one of
the buffers, and the latency tracer swaps the global buffer with the
max buffer, the result is that the event may commit the data to the
wrong buffer.

This patch changes the API to the trace recording to be recieve the
buffer that was used to reserve a commit. Then this buffer can be passed
in to the commit.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
12 files changed:
include/linux/ftrace_event.h
include/trace/ftrace.h
kernel/trace/blktrace.c
kernel/trace/trace.c
kernel/trace/trace.h
kernel/trace/trace_boot.c
kernel/trace/trace_events.c
kernel/trace/trace_functions_graph.c
kernel/trace/trace_mmiotrace.c
kernel/trace/trace_power.c
kernel/trace/trace_sched_switch.c
kernel/trace/trace_syscalls.c