The display of absolute time and duration fields is based on the
latency field. This was added during the irqsoff/wakeup tracers
graph support changes.

It's causing confusion in what fields will be displayed for the
function_graph tracer itself. So I'm removing this depency, and
adding absolute time and duration fields to the preemptirqsoff
preemptoff irqsoff wakeup tracers.

With following commands:
	# echo function_graph > ./current_tracer
	# cat trace

This is what it looked like before:
# tracer: function_graph
#
#     TIME        CPU  DURATION                  FUNCTION CALLS
#      |          |     |   |                     |   |   |   |
 0)   0.068 us    |          } /* page_add_file_rmap */
 0)               |          _raw_spin_unlock() {
...

This is what it looks like now:
# tracer: function_graph
#
# CPU  DURATION                  FUNCTION CALLS
# |     |   |                     |   |   |   |
 0)   0.068 us    |                } /* add_preempt_count */
 0)   0.993 us    |              } /* vfsmount_lock_local_lock */
...

For preemptirqsoff preemptoff irqsoff wakeup tracers,
this is what it looked like before:
SNIP
#                       _-----=> irqs-off
#                      / _----=> need-resched
#                     | / _---=> hardirq/softirq
#                     || / _--=> preempt-depth
#                     ||| / _-=> lock-depth
#                     |||| /
# CPU  TASK/PID       |||||  DURATION                  FUNCTION CALLS
# |     |    |        |||||   |   |                     |   |   |   |
 1)    <idle>-0    |  d..1  0.000 us    |  acpi_idle_enter_simple();
...

This is what it looks like now:
SNIP
#
#                                       _-----=> irqs-off
#                                      / _----=> need-resched
#                                     | / _---=> hardirq/softirq
#                                     || / _--=> preempt-depth
#                                     ||| /
#     TIME        CPU  TASK/PID       ||||  DURATION                  FUNCTION CALLS
#      |          |     |    |        ||||   |   |                     |   |   |   |
   19.847735 |   1)    <idle>-0    |  d..1  0.000 us    |  acpi_idle_enter_simple();
...
Signed-off-by: NJiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1307113131-10045-2-git-send-email-jolsa@redhat.comSigned-off-by: NSteven Rostedt <rostedt@goodmis.org>

ftrace_ops that are registered to trace functions can now be
agnostic to each other in respect to what functions they trace.
Each ops has their own hash of the functions they want to trace
and a hash to what they do not want to trace. A empty hash for
the functions they want to trace denotes all functions should
be traced that are not in the notrace hash.

Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

Fixes generated by 'codespell' and manually reviewed.
Signed-off-by: NLucas De Marchi <lucas.demarchi@profusion.mobi>

During early boot, local IRQ is disabled until IRQ subsystem is
properly initialized.  During this time, no one should enable
local IRQ and some operations which usually are not allowed with
IRQ disabled, e.g. operations which might sleep or require
communications with other processors, are allowed.

lockdep tracked this with early_boot_irqs_off/on() callbacks.
As other subsystems need this information too, move it to
init/main.c and make it generally available.  While at it,
toggle the boolean to early_boot_irqs_disabled instead of
enabled so that it can be initialized with %false and %true
indicates the exceptional condition.
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: NPekka Enberg <penberg@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
LKML-Reference: <20110120110635.GB6036@htj.dyndns.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

The preempt and irqsoff tracers have three types of function tracers.
Normal function tracer, function graph entry, and function graph return.
Each of these use a complex dance to prevent recursion and whether
to trace the data or not (depending if interrupts are enabled or not).

This patch moves the duplicate code into a single routine, to
prevent future mistakes with modifying duplicate complex code.

Cc: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

Move trace_graph_function() and print_graph_headers_flags() functions
to the trace_function_graph.c to be globaly available.
Signed-off-by: NJiri Olsa <jolsa@redhat.com>
LKML-Reference: <1285243253-7372-3-git-send-email-jolsa@redhat.com>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

Documentation/trace/ftrace.txt says

  buffer_size_kb:

        This sets or displays the number of kilobytes each CPU
        buffer can hold. The tracer buffers are the same size
        for each CPU. The displayed number is the size of the
        CPU buffer and not total size of all buffers. The
        trace buffers are allocated in pages (blocks of memory
        that the kernel uses for allocation, usually 4 KB in size).
        If the last page allocated has room for more bytes
        than requested, the rest of the page will be used,
        making the actual allocation bigger than requested.
        ( Note, the size may not be a multiple of the page size
          due to buffer management overhead. )

        This can only be updated when the current_tracer
        is set to "nop".

But it's incorrect. currently total memory consumption is
'buffer_size_kb x CPUs x 2'.

Why two times difference is there? because ftrace implicitly allocate
the buffer for max latency too.

That makes sad result when admin want to use large buffer. (If admin
want full logging and makes detail analysis). example, If admin
have 24 CPUs machine and write 200MB to buffer_size_kb, the system
consume ~10GB memory (200MB x 24 x 2). umm.. 5GB memory waste is
usually unacceptable.

Fortunatelly, almost all users don't use max latency feature.
The max latency buffer can be disabled easily.

This patch shrink buffer size of the max latency buffer if
unnecessary.
Signed-off-by: NKOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
LKML-Reference: <20100701104554.DA2D.A69D9226@jp.fujitsu.com>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

Add function graph output to irqsoff tracer.

The graph output is enabled by setting new 'display-graph' trace option.
Signed-off-by: NJiri Olsa <jolsa@redhat.com>
LKML-Reference: <1270227683-14631-4-git-send-email-jolsa@redhat.com>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

The irqsoff and friends tracers help in finding causes of latency in the
kernel. The also work with the function tracer to show what was happening
when interrupts or preemption are disabled. But the function tracer has
a bit of an overhead and can cause exagerated readings.

Currently, when tracing with /proc/sys/kernel/ftrace_enabled = 0, where the
function tracer is disabled, the information that is provided can end up
being useless. For example, a 2 and a half millisecond latency only showed:

 # tracer: preemptirqsoff
 #
 # preemptirqsoff latency trace v1.1.5 on 2.6.32
 # --------------------------------------------------------------------
 # latency: 2463 us, #4/4, CPU#2 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
 #    -----------------
 #    | task: -4242 (uid:0 nice:0 policy:0 rt_prio:0)
 #    -----------------
 #  => started at: _spin_lock_irqsave
 #  => ended at:   remove_wait_queue
 #
 #
 #                  _------=> CPU#
 #                 / _-----=> irqs-off
 #                | / _----=> need-resched
 #                || / _---=> hardirq/softirq
 #                ||| / _--=> preempt-depth
 #                |||| /_--=> lock-depth
 #                |||||/     delay
 #  cmd     pid   |||||| time  |   caller
 #     \   /      ||||||   \   |   /
 hackbenc-4242    2d....    0us!: trace_hardirqs_off <-_spin_lock_irqsave
 hackbenc-4242    2...1. 2463us+: _spin_unlock_irqrestore <-remove_wait_queue
 hackbenc-4242    2...1. 2466us : trace_preempt_on <-remove_wait_queue

The above lets us know that hackbench with pid 2463 grabbed a spin lock
somewhere and enabled preemption at remove_wait_queue. This helps a little
but where this actually happened is not informative.

This patch adds the stack dump to the end of the irqsoff tracer. This provides
the following output:

 hackbenc-4242    2d....    0us!: trace_hardirqs_off <-_spin_lock_irqsave
 hackbenc-4242    2...1. 2463us+: _spin_unlock_irqrestore <-remove_wait_queue
 hackbenc-4242    2...1. 2466us : trace_preempt_on <-remove_wait_queue
 hackbenc-4242    2...1. 2467us : <stack trace>
  => sub_preempt_count
  => _spin_unlock_irqrestore
  => remove_wait_queue
  => free_poll_entry
  => poll_freewait
  => do_sys_poll
  => sys_poll
  => system_call_fastpath

Now we see that the culprit of this latency was the free_poll_entry code.
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

The state of the function pair tracing_stop()/tracing_start() is
correctly considered when tracer data are updated. However, the global
and externally accessible variable tracing_max_latency is always updated
- even when tracing is stopped.

The update should only occur, if tracing was not stopped.
Signed-off-by: NCarsten Emde <C.Emde@osadl.org>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

When the nsecs_to_usecs() conversion in probe_wakeup_sched_switch() and
check_critical_timing() was moved to a later stage in order to avoid
unnecessary computing, it was overlooked to remove the original
variables, assignments and comments..
Signed-off-by: NCarsten Emde <C.Emde@osadl.org>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

Currently the latency tracers reset the ring buffer. Unfortunately
if a commit is in process (due to a trace event), this can corrupt
the ring buffer. When this happens, the ring buffer will detect
the corruption and then permanently disable the ring buffer.

The bug does not crash the system, but it does prevent further tracing
after the bug is hit.

Instead of reseting the trace buffers, the timestamp of the start of
the trace is used instead. The buffers will still contain the previous
data, but the output will not count any data that is before the
timestamp of the trace.

Note, this only affects the static trace output (trace) and not the
runtime trace output (trace_pipe). The runtime trace output does not
make sense for the latency tracers anyway.
Reported-by: NArnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

The latency tracers (irqsoff, preemptoff, preemptirqsoff, and wakeup)
are pretty useless with the default output format. This patch makes them
automatically enable the latency format when they are selected. They
also record the state of the latency option, and if it was not enabled
when selected, they disable it on reset.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>

Impact: cleanup

Fix incorrect hint message in code and typos in comments.
Signed-off-by: NWenji Huang <wenji.huang@oracle.com>
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>

Impact: cleanup
Signed-off-by: NArnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Impact: cleanup of duplicate features

The trace output disables the ring buffer and prevents tracing to
occur. The code in irqsoff to do the same thing is no longer needed.
This patch removes it.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Impact: trace max latencies on start of latency tracing

This patch sets the max latency to zero whenever one of the
irq variant tracers or the wakeup tracer is set to current tracer.

Most developers expect to see output when starting up a latency
tracer. But since the max_latency is already set to max, and
it takes a latency greater than max_latency to be recorded, there
is no trace. This is not the expected behavior and has even confused
myself.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Impact: trace max latencies on start of latency tracing

This patch sets the max latency to zero whenever one of the
irq variant tracers or the wakeup tracer is set to current tracer.

Most developers expect to see output when starting up a latency
tracer. But since the max_latency is already set to max, and
it takes a latency greater than max_latency to be recorded, there
is no trace. This is not the expected behavior and has even confused
myself.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Impact: extend the ->init() method with the ability to fail

This bring a way to know if the initialization of a tracer successed.
A tracer must return 0 on success and a traditional error (ie:
-ENOMEM) if it fails.

If a tracer fails to init, it is free to print a detailed warn. The
tracing api will not and switch to a new tracer will just return the
error from the init callback.

Note: this will be used for the return tracer.
Signed-off-by: NFrederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Impact: remove obsolete variable in trace_array structure

With the new start / stop method of ftrace, the ctrl variable
in the trace_array structure is now obsolete. Remove it.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Impact: Remove the ctrl_update tracer method

With the new quick start/stop method of tracing, the ctrl_update
method is out of date.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Impact: fix to irqsoff tracer output

In converting to the new start / stop ftrace handling, the irqsoff
tracer start called the irqsoff reset function. irqsoff tracer is
not the same as the other traces, and it resets the buffers while
searching for the longest latency.

The reset that the irqsoff stop method calls disables the function
tracing. That means that, by starting the tracer, the function
tracer is disabled incorrectly.

This patch simply removes the call to reset which keeps the function
tracing enabled. Reset is not needed for the irqsoff stop method.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Impact: change where tracing is started up and stopped

Currently, when a new tracer is selected via echo'ing a tracer name into
the current_tracer file, the startup is only done if tracing_enabled is
set to one. If tracing_enabled is changed to zero (by echo'ing 0 into
the tracing_enabled file) a full shutdown is performed.

The full startup and shutdown of a tracer can be expensive and the
user can lose out traces when echo'ing in 0 to the tracing_enabled file,
because the process takes too long. There can also be places that
the user would like to start and stop the tracer several times and
doing the full startup and shutdown of a tracer might be too expensive.

This patch performs the full startup and shutdown when a tracer is
selected. It also adds a way to do a quick start or stop of a tracer.
The quick version is just a flag that prevents the tracing from
taking place, but the overhead of the code is still there.

For example, the startup of a tracer may enable tracepoints, or enable
the function tracer.  The stop and start will just set a flag to
have the tracer ignore the calls when the tracepoint or function trace
is called.  The overhead of the tracer may still be present when
the tracer is stopped, but no tracing will occur. Setting the tracer
to the 'nop' tracer (or any other tracer) will perform the shutdown
of the tracer which will disable the tracepoint or disable the
function tracer.

The tracing_enabled file will simply start or stop tracing.

This change is all internal. The end result for the user should be the same
as before. If tracing_enabled is not set, no trace will happen.
If tracing_enabled is set, then the trace will happen. The tracing_enabled
variable is static between tracers. Enabling  tracing_enabled and
going to another tracer will keep tracing_enabled enabled. Same
is true with disabling tracing_enabled.

This patch will now provide a fast start/stop method to the users
for enabling or disabling tracing.

Note: There were two methods to the struct tracer that were never
 used: The methods start and stop. These were to be used as a hook
 to the reading of the trace output, but ended up not being
 necessary. These two methods are now used to enable the start
 and stop of each tracer, in case the tracer needs to do more than
 just not write into the buffer. For example, the irqsoff tracer
 must stop recording max latencies when tracing is stopped.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Due to confusion between the ftrace infrastructure and the gcc profiling
tracer "ftrace", this patch renames the config options from FTRACE to
FUNCTION_TRACER.  The other two names that are offspring from FTRACE
DYNAMIC_FTRACE and FTRACE_MCOUNT_RECORD will stay the same.

This patch was generated mostly by script, and partially by hand.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

The assigning of the pc counter is in the wrong spot in the
check_critical_timing function. The pc variable is used in the
out jump.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

With the new ring buffer infrastructure in ftrace, I'm trying to make
ftrace a little more light weight.

This patch converts a lot of the local_irq_save/restore into
preempt_disable/enable.  The original preempt count in a lot of cases
has to be sent in as a parameter so that it can be recorded correctly.
Some places were recording it incorrectly before anyway.

This is also laying the ground work to make ftrace a little bit
more reentrant, and remove all locking. The function tracers must
still protect from reentrancy.

Note: All the function tracers must be careful when using preempt_disable.
  It must do the following:

  resched = need_resched();
  preempt_disable_notrace();
  [...]
  if (resched)
	preempt_enable_no_resched_notrace();
  else
	preempt_enable_notrace();

The reason is that if this function traces schedule() itself, the
preempt_enable_notrace() will cause a schedule, which will lead
us into a recursive failure.

If we needed to reschedule before calling preempt_disable, we
should have already scheduled. Since we did not, this is most
likely that we should not and are probably inside a schedule
function.

If resched was not set, we still need to catch the need resched
flag being set when preemption was off and the if case at the
end will catch that for us.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

This patch ports ftrace over to the new ring buffer.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

fix:

 ERROR: "start_critical_timings" [drivers/acpi/processor.ko] undefined!
 ERROR: "stop_critical_timings" [drivers/acpi/processor.ko] undefined!
Signed-off-by: NIngo Molnar <mingo@elte.hu>

When PREEMPT_TRACER and IRQSOFF_TRACER are both configured and irqsoff
tracer is running, the preempt_off sections might also be traced.

Thanks to Andrew Morton for pointing out my mistake of spin_lock disabling
interrupts while he was reviewing ftrace.txt. Seems that my example I used
actually hit this bug.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Printing out new max latencies was fine for the old RT tracer. But for
mainline it is a bit messy. We also need to test if the run queue
is locked before we can do the print. This means that we may not be
printing out latencies if the run queue is locked on another CPU.
This produces inconsistencies in the output.

This patch simply removes the print altogether.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Cc: pq@iki.fi
Cc: proski@gnu.org
Cc: sandmann@redhat.com
Cc: a.p.zijlstra@chello.nl
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>

This patch changes the use of __get_cpu_var to explicitly calling
raw_smp_processor_id and using the per_cpu() macro. On some debug
configurations, the use of __get_cpu_var may cause ftrace to trigger
and this can cause problems with the irqsoff tracing.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>

printk called from wakeup critical timings and irqs off can
cause deadlocks since printk might do a wakeup itself. If the
call to printk happens with the runqueue lock held, it can
deadlock.

This patch protects the printk from being called in trace irqs off
with a test to see if the runqueue for the current CPU is locked.
If it is locked, the printk is skipped.

The wakeup always holds the runqueue lock, so the printk is
simply removed.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>

now that we have a kbuild method for notrace, no need to pollute the
C code with the annotations.
Signed-off-by: NIngo Molnar <mingo@elte.hu>
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>

clean up recent code.
Signed-off-by: NIngo Molnar <mingo@elte.hu>
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>

A new check was added in the ftrace function that wont trace if the CPU
trace buffer is disabled.  Unfortunately, other tracers used ftrace() to
write to the buffer after they disabled it. The new disable check makes
these calls into a nop.

This patch changes the __ftrace that is called without the check into a
new api for the other tracers to use, called "trace_function". The other
tracers use this interface instead when the trace CPU buffer is already
disabled.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>

rename and uninline now() to ftrace_now().
Signed-off-by: NIngo Molnar <mingo@elte.hu>
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>

factor out code and clean it up.
Signed-off-by: NIngo Molnar <mingo@elte.hu>
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>

Signed-off-by: NIngo Molnar <mingo@elte.hu>
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>

This patch fixes some bugs to the updating of the max trace that
was caused by implementing the new buffering.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>

The irqsoff function tracer did a __get_cpu_var to determine
if it should trace the function or not. The problem is that
__get_cpu_var can preempt between getting the CPU and reading
the cpu variable. This means that the cpu variable that is
being read is not from the cpu being run on.

At worst, this can give a false positive, where we trace the
function when we should not.  It will never give a false negative
since we only want to trace when interrupts are disabled
and we never preempt when they are.

This fix adds a check after reading the irq flags to only
trace if the interrupts are actually disabled. It also changes
the reading of the cpu variable to use a raw_smp_processor_id
since we now don't care if we preempt. We still catch that fact.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>