[Intel-gfx] [PATCH i-g-t] lib/pm_rpm: Reload the module with full mmio debugging
Chris Wilson
chris at chris-wilson.co.uk
Tue Sep 4 14:21:09 UTC 2018
Quoting Imre Deak (2018-09-04 15:14:06)
> On Tue, Sep 04, 2018 at 11:20:04AM +0100, Chris Wilson wrote:
> > Our unclaimed mmio access debugging is lazy, doing cheap checks
> > periodically and only if they fail do a full check around every mmio
> > access. When testing for runtime pm, enable the full mmio debugging from
> > the initial load.
> >
> > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > Cc: Imre Deak <imre.deak at intel.com>
> > ---
> > tests/pm_rpm.c | 7 +++++--
> > 1 file changed, 5 insertions(+), 2 deletions(-)
> >
> > diff --git a/tests/pm_rpm.c b/tests/pm_rpm.c
> > index e3bb6227a..91aef0889 100644
> > --- a/tests/pm_rpm.c
> > +++ b/tests/pm_rpm.c
> > @@ -2058,7 +2058,7 @@ int main(int argc, char *argv[])
> > igt_subtest("module-reload") {
> > igt_debug("Reload w/o display\n");
> > igt_i915_driver_unload();
> > - igt_assert_eq(igt_i915_driver_load("disable_display=1"), 0);
> > + igt_assert_eq(igt_i915_driver_load("disable_display=1 mmio_debug=2147483647"), 0);
> >
> > igt_assert(setup_environment());
> > igt_assert(igt_wait(device_in_pci_d3(), 2000, 100));
> > @@ -2066,13 +2066,16 @@ int main(int argc, char *argv[])
> >
> > igt_debug("Reload as normal\n");
> > igt_i915_driver_unload();
> > - igt_assert_eq(igt_i915_driver_load(NULL), 0);
> > + igt_assert_eq(igt_i915_driver_load("mmio_debug=2147483647"), 0);
>
> Could've been in hex.
-1 should work as well as INT_MAX, I was just a bit dubious about that
claim at the time. But the kernel looks to be happy enough to start
counting down from -1.
> > igt_assert(setup_environment());
> > igt_assert(igt_wait(device_in_pci_d3(), 2000, 100));
> > if (enable_one_screen_with_type(&ms_data, SCREEN_TYPE_ANY))
> > drm_resources_equal_subtest();
> > teardown_environment();
> > +
> > + /* Remove our mmio_debugging module */
> > + igt_i915_driver_unload();
>
> What loads it again?
We automatically load it upon a failed drm_driver_open(). We have a list
of modules to try and modprobe if we find no matching /dev/dri/* fd.
-Chris
More information about the Intel-gfx
mailing list