Re: [PATCH v2 24/32] selftests/ftrace: Add max stack tracer testcase

[Steven Rostedt <rostedt@goodmis.org>]

On Sat, 25 Aug 2018 10:45:09 +0900
Masami Hiramatsu <mhiramat@kernel.org> wrote:

> On Fri, 24 Aug 2018 19:23:52 -0400
> Steven Rostedt <rostedt@goodmis.org> wrote:
> 
> > On Fri, 17 Aug 2018 01:40:30 +0900
> > Masami Hiramatsu <mhiramat@kernel.org> wrote:
> >   
> > > Add a testcase for max stack tracer, which checks basic
> > > max stack usage tracing and its filter feature.
> > > 
> > > Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
> > > ---
> > >  .../ftrace/test.d/ftrace/func_stack_tracer.tc      |   39 ++++++++++++++++++++
> > >  1 file changed, 39 insertions(+)
> > >  create mode 100644 tools/testing/selftests/ftrace/test.d/ftrace/func_stack_tracer.tc
> > > 
> > > diff --git a/tools/testing/selftests/ftrace/test.d/ftrace/func_stack_tracer.tc b/tools/testing/selftests/ftrace/test.d/ftrace/func_stack_tracer.tc
> > > new file mode 100644
> > > index 000000000000..b414f0e3c646
> > > --- /dev/null
> > > +++ b/tools/testing/selftests/ftrace/test.d/ftrace/func_stack_tracer.tc
> > > @@ -0,0 +1,39 @@
> > > +#!/bin/sh
> > > +# SPDX-License-Identifier: GPL-2.0
> > > +# description: ftrace - Max stack tracer
> > > +# Test the basic function of max-stack usage tracing
> > > +
> > > +if [ ! -f stack_trace ]; then
> > > +  echo "Max stack tracer is not supported - please make CONFIG_STACK_TRACER=y"
> > > +  exit_unsupported
> > > +fi
> > > +
> > > +echo > stack_trace_filter
> > > +echo 0 > stack_max_size
> > > +echo 1 > /proc/sys/kernel/stack_tracer_enabled
> > > +
> > > +: "Fork and wait for the first entry become !lock"
> > > +timeout=10
> > > +while [ $timeout -ne 0 ]; do
> > > +  ( echo "forked" )
> > > +  FL=`grep " 0)" stack_trace`
> > > +  echo $FL | grep -q "lock" || break;
> > > +  timeout=$((timeout - 1))
> > > +done
> > > +echo 0 > /proc/sys/kernel/stack_tracer_enabled
> > > +
> > > +echo '*lock*' > stack_trace_filter
> > > +test `cat stack_trace_filter | wc -l` -eq `grep lock stack_trace_filter | wc -l`
> > > +
> > > +echo 0 > stack_max_size
> > > +echo 1 > /proc/sys/kernel/stack_tracer_enabled  
> > 
> > This test fails quite a bit because it doesn't guarantee that another
> > stack trace gets in that is bigger, and we don't find the "lock".  
> 
> Hmm, maybe I misunderstand what is the stack_trace_filter.
> I thought it filtered the function that triggered stack trace,
> so the top of stack trace was always filtered function.
> If not, what does this filter?

Ah, I missed the filter part, no its correct, and actually found a
bug ;-) By adding this patch, it works again:

diff --git a/kernel/trace/trace_stack.c b/kernel/trace/trace_stack.c
index 4237eba4ef20..2b0d1ee3241c 100644
--- a/kernel/trace/trace_stack.c
+++ b/kernel/trace/trace_stack.c
@@ -111,7 +111,7 @@ check_stack(unsigned long ip, unsigned long *stack)
 	stack_trace_max_size = this_size;
 
 	stack_trace_max.nr_entries = 0;
-	stack_trace_max.skip = 3;
+	stack_trace_max.skip = 0;
 
 	save_stack_trace(&stack_trace_max);
 


-- Steve

> >   
> > > +
> > > +: "Fork and always the first entry including lock"
> > > +timeout=10
> > > +while [ $timeout -ne 0 ]; do
> > > +  ( echo "forked" )
> > > +  FL=`grep " 0)" stack_trace`
> > > +  echo $FL | grep -q "lock"
> > > +  timeout=$((timeout - 1))
> > > +done
> > > +echo 0 > /proc/sys/kernel/stack_tracer_enabled  
> >   
> 
>