If runtime PM is enabled and we unset all modes, we will runtime
suspend after __intel_set_mode() , then function
intel_modeset_check_state() will try to read the HW state while it is
suspended and trigger lots of WARNs because it shouldn't be reading
registers.
So on this patch we make intel_ddi_connector_get_hw_state() return
false in case the power domain is disabled, and we also make
intel_display_power_enabled() return false in case the device is
suspended. Notice that we can't just use
intel_display_power_enabled_sw() because while the driver is being
initialized the power domain refcounts are not reflecting the real
state of the hardware.
Just for reference, I have previously published an alternate patch for
this problem, called "drm/i915: get runtime PM at intel_set_mode".
Testcase: igt/pm_pc8
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
enum port port = intel_ddi_get_encoder_port(intel_encoder);
enum pipe pipe = 0;
enum transcoder cpu_transcoder;
+ enum intel_display_power_domain power_domain;
uint32_t tmp;
+ power_domain = intel_display_port_power_domain(intel_encoder);
+ if (!intel_display_power_enabled(dev_priv, power_domain))
+ return false;
+
if (!intel_encoder->get_hw_state(intel_encoder, &pipe))
return false;
bool is_enabled;
int i;
+ if (dev_priv->pm.suspended)
+ return false;
+
power_domains = &dev_priv->power_domains;
is_enabled = true;