Re: [PATCH] PM: trace events for suspend/resume

[Todd E Brandt <todd.e.brandt@linux.intel.com>]

Hi guys, I'm not hearing any objections to revamping the machine_suspend trace
event, so I'll resubmit a patch with it removed and supplanted with the
new suspend_resume event. I've also discovered that this first patch doesn't
completely cover freeze, so I'm moving a couple of the calls around.

On Fri, May 23, 2014 at 01:06:32AM +0200, Rafael J. Wysocki wrote:
> On Thursday, May 22, 2014 06:41:45 PM Steven Rostedt wrote:
> > On Thu, 2014-05-22 at 10:07 -0700, Todd E Brandt wrote:
> > >  
> > > > >  #include <asm/io.h>
> > > > >  #include <asm/uaccess.h>
> > > > > @@ -835,7 +836,9 @@ acpi_status acpi_os_remove_interrupt_handler(u32 irq, acpi_osd_handler handler)
> > > > >  
> > > > >  void acpi_os_sleep(u64 ms)
> > > > >  {
> > > > > +	trace_suspend_resume("acpi_os_sleep", true);
> > > > >  	msleep(ms);
> > > > > +	trace_suspend_resume("acpi_os_sleep", false);
> > > > >  }
> > > > 
> > > > This function doesn't have anything to do with system suspend/resume.
> > > > It is for waiting. :-)
> > > 
> > > Yea but it really helps you diagnose bad firmware during suspend/resume
> > > (for instance when there's just an arbitrary sleep 100ms in the ACPI code).
> > > But it doesn't matter for general performance testing, I'll remove it on v2
> > > submit.
> > > 
> > > > 
> > > > The rest of the patch looks OK to me.
> > > > 
> > > > Steven, what about the tracing angle?
> > 
> > It should be under an ACPI system (if there is one). Looks rather silly
> > to call it for suspend.
> 
> I meant the rest of the patch, not this particular hunk. :-)
> 
> Rafael
>