Re: [PATCH 3/3] powerpc/ftrace: simplify prepare_ftrace_return

On Wed, 17 Sep 2014 17:07:04 +1000
Anton Blanchard [off-list ref] wrote:

Instead of passing in the stack address of the link register
to be modified, just pass in the old value and return the
new value and rely on ftrace_graph_caller to do the
modification.

This removes the exception handling around the stack update -
it isn't needed and we weren't consistent about it. Later on
we would do an unprotected modification:

       if (!ftrace_graph_entry(&trace)) {
               *parent = old;

First I'll say this is something I've been wanting to do with x86 for
some time. That said...

With this patch, things move much further in my tests. The stress test
passes again. But then it fails on my stack trace test. Which is
because this is what I have in the stack traces:

           sleep-3557  [000] d...   100.206808: <stack trace>
 => 0
 => 0
 => 0
 => 0
 => 0
 => 0
 => 0
 => 0


Where without the patches I have something like this:

          sleep-3641  [001] d...   304.023550: <stack trace>
 => .ftrace_raw_event_sched_switch
 => .__schedule
 => .schedule
 => .do_nanosleep
 => .hrtimer_nanosleep
 => .compat_SyS_nanosleep
 => syscall_exit
 => 0

This could be broken from the earlier patches, I haven't run just this
test. I probably should on them.

I've attached the test.

-- Steve