On Tue, 2015-11-17 at 23:30 +0200, Ville Syrjälä wrote:
> On Tue, Nov 17, 2015 at 10:18:41PM +0100, Daniel Vetter wrote:
> > If we can't acquire dev->struct_mutex we need to fail runtime suspend,
> > at least with the current design. Currently we do that using -EAGAIN,
> > but that upsets the pm core, resulting in the occasional fail testcase
> > in our CI with the following dmesg dirt:
> >
> > pci_pm_runtime_suspend(): intel_runtime_suspend+0x0/0x240 [i915] returns -11
> >
> > Chris has some ideas to improve this, but for now just shut up the
> > error.
> >
> > Cc: Paulo Zanoni <przanoni@xxxxxxxxx>
> > Cc: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
> > Signed-off-by: Daniel Vetter <daniel.vetter@xxxxxxxxx>
> > ---
> > drivers/gpu/drm/i915/i915_drv.c | 6 +++---
> > 1 file changed, 3 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
> > index 5a70aca71d6b..ab8ffbc48e2d 100644
> > --- a/drivers/gpu/drm/i915/i915_drv.c
> > +++ b/drivers/gpu/drm/i915/i915_drv.c
> > @@ -1497,8 +1497,7 @@ static int intel_runtime_suspend(struct device *device)
> > * We could deadlock here in case another thread holding struct_mutex
> > * calls RPM suspend concurrently, since the RPM suspend will wait
> > * first for this RPM suspend to finish. In this case the concurrent
> > - * RPM resume will be followed by its RPM suspend counterpart. Still
> > - * for consistency return -EAGAIN, which will reschedule this suspend.
> > + * RPM resume will be followed by its RPM suspend counterpart.
> > */
> > if (!mutex_trylock(&dev->struct_mutex)) {
> > DRM_DEBUG_KMS("device lock contention, deffering suspend\n");
> > @@ -1508,7 +1507,8 @@ static int intel_runtime_suspend(struct device *device)
> > */
> > pm_runtime_mark_last_busy(device);
> >
> > - return -EAGAIN;
> > + /* Fail silently to avoid upsetting the pm core. */
> > + return 0;
>
> So the core will assume we're now suspended and then resume gets called
> while we're still powered on. Sounds like a bad plan to me. I'm
> especially worried about VLV here with its GT no wake dance and manual
> save/restore.
Also the PCI core will put the device into D3 if we report success.
>
> > }
> > /*
> > * We are safe here against re-faults, since the fault handler takes
> > --
> > 2.5.1
> >
> > _______________________________________________
> > Intel-gfx mailing list
> > Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
> > http://lists.freedesktop.org/mailman/listinfo/intel-gfx
>
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
