Ftrace - Revision history

Tim Bird: /* find deepest kernel stack */

2013-03-21T18:28:42Z

‎find deepest kernel stack

Tim Bird: /* find deepest kernel stack */

2013-03-21T18:24:45Z

‎find deepest kernel stack

Tim Bird: /* find deepest kernel stack */ refine instructions

2013-03-21T18:22:08Z

‎find deepest kernel stack: refine instructions

Tim Bird at 00:11, 10 May 2012

2012-05-10T00:11:20Z

Tim Bird: add output to stack tracer tip

2012-05-10T00:07:43Z

add output to stack tracer tip

Tim Bird at 00:05, 10 May 2012

2012-05-10T00:05:12Z

Tim Bird: add some interesting ftrace tips

2012-05-10T00:00:24Z

add some interesting ftrace tips

Show changes

Jsujjava: Add Category Linux tracing

2011-05-12T18:05:49Z

Add Category Linux tracing

Rostedt at 01:50, 13 April 2011

2011-04-13T01:50:12Z

Tim Bird: Created page with 'This page has information about ftrace, that I couldn't find elsewhere. == Tips == === Tracing a specific process === (Adapted from email by Steven Rostedt) To trace just the ke…'

2010-01-12T19:16:06Z

Created page with 'This page has information about ftrace, that I couldn't find elsewhere. == Tips == === Tracing a specific process === (Adapted from email by Steven Rostedt) To trace just the ke…'

New page

This page has information about ftrace, that I couldn't find elsewhere.

== Tips ==
=== Tracing a specific process ===
(Adapted from email by Steven Rostedt)
To trace just the kernel functions executed in the context of a particular function,
set the pseudo-variable 'set-ftrace-pid', to the process id (pid) of the process.

If the process is not already running, you can use a wrapper shell script and the
'exec' command, to execute a command as a known pid.

Like so:
<pre>
#!/bin/sh
echo $$ > /debug/tracing/set_ftrace_pid
# can set other filtering here
echo function > /debug/tracing/current_tracer
exec $*
</pre>

In this example, '$$' is the pid of the currently executing process (the shell
script. This is set into the 'set_ftrace_pid' variable, then the 'function'
tracer is enabled. Then this script exec's the command (specified by the first
argument to the script).

Example usage (assuming script is called 'trace_cmd'):
trace_cmd ls

@@ Line 246: / Line 246: @@
-=== find deepest kernel stack ===
+=== Find deepest kernel stack ===
 This is useful to find the routine with the deepest kernel stack
 The system continually monitors the stack depth of all processes, and

@@ Line 269: / Line 269: @@
 *** echo 0 >/proc/sys/kernel/stack_tracer_enabled
-Here is what the output looks like:
+=== additional resources ===
 <pre>
 /debug/tracing # cat stack_trace

@@ Line 248: / Line 248: @@
 === find deepest kernel stack ===
 This is useful to find the routine with the deepest kernel stack
-* set the following kernel configuration options:
+(The following instructions work for a v3.0 Linux kernel)
 ** CONFIG_FTRACE: "Tracers"
 ** CONFIG_FUNCTION_TRACER: "Kernel Function Tracer"
 ** CONFIG_STACK_TRACER: "Trace max stack"
-* use the following on the kernel command line:
+* Turning it on: You can turn it on at boot time or at runtime.
-** stacktrace
+** At boot time, use the following on the kernel command line:
-* to get the data:
+*** stacktrace
 ** $ mount -t debugfs debugfs /debug
 ** $ cat /debug/tracing/stack_trace
 * scope of operation
 ** the stack tracer will continue operating until you turn it off, which can be done with:
-*** echo 0 >/debug/tracing/stacktrace
+*** echo 0 >/proc/sys/kernel/stack_tracer_enabled
 Here is what the output looks like:

@@ Line 5: / Line 5: @@
 The interface for Ftrace resides in the debugfs file system in the tracing directory.
-Documentation for this can be found in the Linux kernel Documentation directory in
+Documentation for this can be found in the Linux kernel source tree at
-trace/ftrace.txt.
+[http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;f=Documentation/trace/ftrace.txt; Documentation/trace/ftrace.txt].
 == trace-cmd ==

@@ Line 89: / Line 89: @@
   # trace-cmd record -p function -F ls
-=== Capturing an oops to the serial console ===
+=== Capturing an oops (from startup) to the serial console ===
 You can capture the function calls leading up to a panic
 by placing the following on the kernel command line:
   ftrace=function ftrace_dump_on_oops
@@ Line 224: / Line 225: @@
 </pre>
-===== Find latencies on kernel startup ==
+=== Find latencies on kernel startup ===
 It is possible to use ftrace to record functions that exceed a certain amount of time, using
 the 'tracing_thresh' option.
@@ Line 256: / Line 257: @@
 * to get the data:
 ** $ mount -t debugfs debugfs /debug
-** $ cat /debug/tracing/trace
+** $ cat /debug/tracing/stack_trace
 * scope of operation
 ** the stack tracer will continue operating until you turn it off, which can be done with:
 *** echo 0 >/debug/tracing/stacktrace
 [[Category:Linux tracing]]

← Older revision		Revision as of 18:05, 12 May 2011
Line 88:		Line 88:

	# trace-cmd record -p function -F ls		# trace-cmd record -p function -F ls
		+
		+	[[Category:Linux tracing]]

@@ Line 1: / Line 1: @@
-This page has information about ftrace, that I couldn't find elsewhere.
+Ftrace is the Linux kernel internal tracer that was included in the Linux kernel in 2.6.27.
 == Tips ==
-=== Tracing a specific process ===
+=== Tracing a specific process with the Ftrace interface ===
 (Adapted from email by Steven Rostedt)
 To trace just the kernel functions executed in the context of a particular function,
@@ Line 24: / Line 82: @@
 argument to the script).
-Example usage (assuming script is called 'trace_cmd'):
+Example usage (assuming script is called 'trace_command'):
-  trace_cmd ls
+  trace_command ls