[Intel-gfx] [PATCH v2] drm/i915: Apply missed interrupt after reset w/a to all ringbuffer gen

[Chris Wilson]

Quoting Ville Syrjälä (2018-12-13 12:45:00)
> On Thu, Dec 13, 2018 at 12:34:02PM +0000, Chris Wilson wrote:
> > Quoting Ville Syrjälä (2018-12-13 12:29:15)
> > > On Thu, Dec 13, 2018 at 12:07:35PM +0000, Chris Wilson wrote:
> > > > Quoting Ville Syrjälä (2018-12-13 11:59:28)
> > > > > On Thu, Dec 13, 2018 at 11:01:05AM +0000, Chris Wilson wrote:
> > > > > > Having completed a test run of gem_eio across all machines in CI we also
> > > > > > observe the phenomenon (of lost interrupts after resetting the GPU) on
> > > > > > gen3 machines as well as the previously sighted gen6/gen7. Let's apply
> > > > > > the same HWSTAM workaround that was effective for gen6+ for all, as
> > > > > > although we haven't seen the same failure on gen4/5 it seems prudent to
> > > > > > keep the code the same.
> > > > > > 
> > > > > > As a consequence we can remove the extra setting of HWSTAM and apply the
> > > > > > register from a single site.
> > > > > > 
> > > > > > v2: Delazy and move the HWSTAM into its own function
> > > > > > 
> > > > > > References: https://bugs.freedesktop.org/show_bug.cgi?id=108735
> > > > > > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > > > > > Cc: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
> > > > > > ---
> > > > > >  drivers/gpu/drm/i915/i915_irq.c         |  9 ------
> > > > > >  drivers/gpu/drm/i915/intel_ringbuffer.c | 41 ++++++++++++++++---------
> > > > > >  2 files changed, 27 insertions(+), 23 deletions(-)
> > > > > > 
> > > > > > diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
> > > > > > index e2dac9b5f4ce..0c7fc9890891 100644
> > > > > > --- a/drivers/gpu/drm/i915/i915_irq.c
> > > > > > +++ b/drivers/gpu/drm/i915/i915_irq.c
> > > > > > @@ -3586,9 +3586,6 @@ static void ironlake_irq_reset(struct drm_device *dev)
> > > > > >  {
> > > > > >       struct drm_i915_private *dev_priv = to_i915(dev);
> > > > > >  
> > > > > > -     if (IS_GEN(dev_priv, 5))
> > > > > > -             I915_WRITE(HWSTAM, 0xffffffff);
> > > > > > -
> > > > > >       GEN3_IRQ_RESET(DE);
> > > > > >       if (IS_GEN(dev_priv, 7))
> > > > > >               I915_WRITE(GEN7_ERR_INT, 0xffffffff);
> > > > > > @@ -4368,8 +4365,6 @@ static void i8xx_irq_reset(struct drm_device *dev)
> > > > > >  
> > > > > >       i9xx_pipestat_irq_reset(dev_priv);
> > > > > >  
> > > > > > -     I915_WRITE16(HWSTAM, 0xffff);
> > > > > > -
> > > > > >       GEN2_IRQ_RESET();
> > > > > >  }
> > > > > >  
> > > > > > @@ -4537,8 +4532,6 @@ static void i915_irq_reset(struct drm_device *dev)
> > > > > >  
> > > > > >       i9xx_pipestat_irq_reset(dev_priv);
> > > > > >  
> > > > > > -     I915_WRITE(HWSTAM, 0xffffffff);
> > > > > > -
> > > > > >       GEN3_IRQ_RESET();
> > > > > >  }
> > > > > >  
> > > > > > @@ -4648,8 +4641,6 @@ static void i965_irq_reset(struct drm_device *dev)
> > > > > >  
> > > > > >       i9xx_pipestat_irq_reset(dev_priv);
> > > > > >  
> > > > > > -     I915_WRITE(HWSTAM, 0xffffffff);
> > > > > > -
> > > > > >       GEN3_IRQ_RESET();
> > > > > >  }
> > > > > 
> > > > > So we're not worried about enabling interrupts and having
> > > > > something unmasked in HWSTAM by accident before we have the
> > > > > status page set up?
> > > > 
> > > > Sanitization of the HWSP setup would be off during early engine setup.
> > > > We do the irq install & reset during i915_load_modeset_init after we do
> > > > the status page setup. Unless I'm mistaken, moving the HWSTAM alongside
> > > > HWSP moves the sanitisation earlier.
> > > 
> > > To me it looks like the hwsp setup happens via i915_gem_init(), which
> > > gets called after irq install. But maybe I'm just hopelessly lost
> > > in the maze.
> > 
> > i915_driver_init_hw -> i915_gem_init_hw -> init_ringbuffer...
> 
> i915_gem_init_hw() doesn't seem to be called from there, unless I'm
> looking at a stale tree already. I did checkout ~4 hours ago so
> could very well be outdated already :)

Indeed not. So moved the sanitization to
	i915_driver_init_mmio -> intel_engines_init_mmio
which is unarguably as early as we can do it :)
-Chris