[Intel-gfx] [PATCH igt] lib: Always enable ftrace-dump-on-oops

[Petri Latvala]

On Wed, Nov 08, 2017 at 12:36:32PM +0000, Chris Wilson wrote:
> Quoting Petri Latvala (2017-11-08 12:32:09)
> > On Wed, Nov 08, 2017 at 12:18:40PM +0000, Chris Wilson wrote:
> > > Some debugging information is too voluminous to enable by default, yet
> > > may be invaluable when it comes to post-mortem debugging. trace_printk()
> > > provides the facility for the trace ringbuffer to be dumped on oops,
> > > this way we can cheaply spam the debug log and only present it in case
> > > of emergency.
> > > 
> > > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > > ---
> > >  lib/igt_core.c | 23 +++++++++++++++++++++++
> > >  1 file changed, 23 insertions(+)
> > > 
> > > diff --git a/lib/igt_core.c b/lib/igt_core.c
> > > index 538a4472..ba4622d6 100644
> > > --- a/lib/igt_core.c
> > > +++ b/lib/igt_core.c
> > > @@ -561,6 +561,28 @@ static void low_mem_killer_disable(bool disable)
> > >       chmod(adj_fname, buf.st_mode);
> > >  }
> > >  
> > > +/*
> > > + * If the test takes out the machine, in addition to the usual dmesg
> > > + * spam, the kernel may also emit a "death rattle" -- extra debug
> > > + * information that is overkill for normal successful tests, but
> > > + * vital for post-mortem analysis.
> > > + */
> > > +static void ftrace_dump_on_oops(bool enable)
> > > +{
> > > +     int fd;
> > > +
> > > +     fd = open("/proc/sys/kernel/ftrace_dump_on_oops", O_WRONLY);
> > > +     if (fd < 0)
> > > +             return;
> > > +
> > > +     /*
> > > +      * If we fail, we do not get the death rattle we wish, but we
> > > +      * still want to run the tests anyway.
> > > +      */
> > > +     igt_ignore_warn(write(fd, enable ? "1\n" : "0\n", 2));
> > > +     close(fd);
> > > +}
> > > +
> > >  bool igt_exit_called;
> > >  static void common_exit_handler(int sig)
> > >  {
> > > @@ -858,6 +880,7 @@ out:
> > >               sync();
> > >               oom_adjust_for_doom();
> > >               low_mem_killer_disable(true);
> > > +             ftrace_dump_on_oops(true);
> > 
> > 
> > Is there a call to disable it too? Or rather, restore the value that
> > was there before executing a test.
> 
> No, because I didn't care. If the machine dies after the process
> excited, that death rattle may still be important.

Fair enough, there's indeed no good place where disabling again makes
sense.



Reviewed-by: Petri Latvala <petri.latvala at intel.com>