On Fri, Jun 21, 2013 at 03:40:04PM +0100, Chris Wilson wrote: > Do not trust our bookkeeping when reporting errors, and instead dump the > register contents. In particular, this solves one particular issue when > an error is reported before we finish setting up the outputs and have a > complete mapping (i.e. during initialisation we set garbage state). If > an error occurs at that early stage, it is vital that we get an accurate > report of the hardware state and not conflated with our own inaccurate > opinions. > > This fixes a panic for a large number of pre-Haswell machines that > currently trigger an error during KMS takeover. > > Reported-by: Dustin King <daking at rescomp.stanford.edu> > Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=60021 > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk> > Cc: stable at vger.kernel.org Shouldn't we just dump all transcoder registers on Haswell instead of potentially fragile dances trying to reconstruct state? -Daniel > --- > drivers/gpu/drm/i915/intel_display.c | 37 ++++++++++++++++++++++++++++++++++-- > 1 file changed, 35 insertions(+), 2 deletions(-) > > diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c > index 5988bda..7ce4588 100644 > --- a/drivers/gpu/drm/i915/intel_display.c > +++ b/drivers/gpu/drm/i915/intel_display.c > @@ -10273,12 +10273,43 @@ struct intel_display_error_state { > } plane[I915_MAX_PIPES]; > }; > > +static enum transcoder > +read_cpu_transcoder(struct drm_i915_private *dev_priv, int pipe) > +{ > + if (HAS_DDI(dev_priv->dev)) { > + int edp_pipe; > + > + switch (I915_READ(TRANS_DDI_FUNC_CTL(TRANSCODER_EDP)) & (TRANS_DDI_FUNC_ENABLE | TRANS_DDI_EDP_INPUT_MASK)) { > + default: > + edp_pipe = -1; > + break; > + > + case TRANS_DDI_EDP_INPUT_A_ONOFF | TRANS_DDI_FUNC_ENABLE: > + case TRANS_DDI_EDP_INPUT_A_ON | TRANS_DDI_FUNC_ENABLE: > + edp_pipe = PIPE_A; > + break; > + > + case TRANS_DDI_EDP_INPUT_B_ONOFF | TRANS_DDI_FUNC_ENABLE: > + edp_pipe = PIPE_B; > + break; > + > + case TRANS_DDI_EDP_INPUT_C_ONOFF | TRANS_DDI_FUNC_ENABLE: > + edp_pipe = PIPE_C; > + break; > + } > + > + if (edp_pipe == pipe) > + pipe = TRANSCODER_EDP; > + } > + > + return pipe; > +} > + > struct intel_display_error_state * > intel_display_capture_error_state(struct drm_device *dev) > { > drm_i915_private_t *dev_priv = dev->dev_private; > struct intel_display_error_state *error; > - enum transcoder cpu_transcoder; > int i; > > error = kmalloc(sizeof(*error), GFP_ATOMIC); > @@ -10289,7 +10320,9 @@ intel_display_capture_error_state(struct drm_device *dev) > error->power_well_driver = I915_READ(HSW_PWR_WELL_DRIVER); > > for_each_pipe(i) { > - cpu_transcoder = intel_pipe_to_cpu_transcoder(dev_priv, i); > + enum transcoder cpu_transcoder; > + > + cpu_transcoder = read_cpu_transcoder(dev_priv, i); > error->pipe[i].cpu_transcoder = cpu_transcoder; > > if (INTEL_INFO(dev)->gen <= 6 || IS_VALLEYVIEW(dev)) { > -- > 1.8.3.1 > > _______________________________________________ > Intel-gfx mailing list > Intel-gfx at lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/intel-gfx -- Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
