[Intel-gfx] [PATCH] drm/i915: Add documentation to gen9_set_dc_state()

[Dhinakaran Pandiyan]

On Wed, 2018-04-25 at 10:45 -0700, Rodrigo Vivi wrote:
> On Wed, Apr 25, 2018 at 02:09:14PM +0300, Imre Deak wrote:
> > On Wed, Apr 25, 2018 at 12:50:06PM +0300, Jani Nikula wrote:
> > > 
> > > Argh, now with Ville's correct address.
> > > 
> > > On Wed, 25 Apr 2018, Jani Nikula <jani.nikula at intel.com> wrote:
> > > > Cc: Rodrigo, DK, Ville
> > > >
> > > > On Tue, 17 Apr 2018, Imre Deak <imre.deak at intel.com> wrote:
> > > >> Add documentation to gen9_set_dc_state() on what enabling a given DC
> > > >> state means and at what point HW/DMC actually enters/exits these states.
> > > >>
> > > >> Cc: Jani Nikula <jani.nikula at intel.com>
> > > >> Cc: Daniel Vetter <daniel.vetter at ffwll.ch>
> > > >> Signed-off-by: Imre Deak <imre.deak at intel.com>
> > > >> ---
> > > >>  drivers/gpu/drm/i915/intel_runtime_pm.c | 23 +++++++++++++++++++++++
> > > >>  1 file changed, 23 insertions(+)
> > > >>
> > > >> [ On IRC I stated that PSR entry would be prevented in a given DC state,
> > > >>   but looking more at it I haven't found any proof for this. So as I
> > > >>   understand the only connection between PSR and DC states is that if
> > > >>   DC5/6 is disabled power saving will be blocked, which would otherwise
> > > >>   be possible when PSR is active and the display pipe is off. ]
> > > >
> > > > I think I'm still missing a definitive answer to the question, who
> > > > disables PSR before DP AUX transactions?
> 
> no one. :(
> 
> This is a gap that we are aware. I believe Jose is working to address that.
> 
> cc: Jose.
> 
> > > >
> > > > Does intel_display_power_get(dev_priv, intel_dp->aux_power_domain) in
> > > > pps_lock() cause PSR exit, through the DC state change? If yes, this
> > > > needs to be properly documented in code. Maybe with a WARN_ON(psr
> > > > active) on top.
> > 
> > No, that was only my misunderstanding earlier on IRC. Disabling DC
> > states doesn't prevent PSR entry. So the PSR active state is independent
> > of DC states; if PSR is enabled with DC5/6 disabled it just means there
> > will be less power saving during PSR active periods than it would be
> > possible with DC5/6 enabled.
> 
> Yeap, they are independent.
> 
> > 
> > Disabling PSR then has to happen by some other means while the driver
> > performs AUX transfers, either disabling it manually from the driver, or
> > by using the AUX HW mutex.
> 
> AUX HW mutex is not an option. It got discarded again. Guidance from HW eng
> is to avoid it and disable PSR manually before aux transactions.
> 
> > 
> > > >
> > > > Quoting bspec 7530, "DDI A AUX channel transactions must not be sent
> > > > while SRD is enabled. SRD must be completely disabled before a DDI A AUX
> > > > channel transaction can be sent."
> > > >
> > > > I'm also confused how sink CRC would ever work with PSR enabled.

The bspec quote is specifically for BDW and HSW. The reason, presumably,
is HW sends aux transactions for PSR exit (this, we know for sure) and
the transactions can interfere with driver initiated aux transactions.

afaict, there is nothing in the DP spec against aux reads (for sink-crc)
when PSR is enabled.

So, we have to be careful about reading sink-crc and not start the reads
when PSR exit events happen. The current sink-crc code does the exact
opposite by reading sink-crc after triggering PSR exit events, namely
vbi enable.

-DK

> 
> Our driver does aux transactions on modesets and hw does them on psr sync.
> 
> Sink CRC is used on the tests after modeset and hopefully when psr is
> stable. But this actually could explain most of sink CRC issues we had
> so far, indeed. Too much assumption and so less checks and protections :(
> 
> my bad...
> 
> > > >
> > > > BR,
> > > > Jani.
> > > >
> > > >
> > > >>
> > > >> diff --git a/drivers/gpu/drm/i915/intel_runtime_pm.c b/drivers/gpu/drm/i915/intel_runtime_pm.c
> > > >> index 53ea564f971e..40a7955886d4 100644
> > > >> --- a/drivers/gpu/drm/i915/intel_runtime_pm.c
> > > >> +++ b/drivers/gpu/drm/i915/intel_runtime_pm.c
> > > >> @@ -542,6 +542,29 @@ void gen9_sanitize_dc_state(struct drm_i915_private *dev_priv)
> > > >>  	dev_priv->csr.dc_state = val;
> > > >>  }
> > > >>  
> > > >> +/**
> > > >> + * gen9_set_dc_state - set target display C power state
> > > >> + * @dev_priv: i915 device instance
> > > >> + * @state: target DC power state
> > > >> + * - DC_STATE_DISABLE
> > > >> + * - DC_STATE_EN_UPTO_DC5
> > > >> + * - DC_STATE_EN_UPTO_DC6
> > > >> + * - DC_STATE_EN_DC9
> > > >> + *
> > > >> + * Signal to DMC firmware/HW the target DC power state passed in @state.
> > > >> + * DMC/HW can turn off individual display clocks and power rails when entering
> > > >> + * a deeper DC power state (higher in number) and turns these back when exiting
> > > >> + * that state to a shallower power state (lower in number). The HW will decide
> > > >> + * when to actually enter a given state on an on-demand basis, for instance
> > > >> + * depending on the active state of display pipes. The state of display
> > > >> + * registers backed by affected power rails are saved/restored as needed.
> > > >> + *
> > > >> + * Based on the above enabling a deeper DC power state is asynchronous wrt.
> > > >> + * enabling it. Disabling a deeper power state is synchronous: for instance
> > > >> + * setting %DC_STATE_DISABLE won't complete until all HW resources are turned
> > > >> + * back on and register state is restored. This is guaranteed by the MMIO write
> > > >> + * to DC_STATE_EN blocking until the state is restored.
> > > >> + */
> > > >>  static void gen9_set_dc_state(struct drm_i915_private *dev_priv, uint32_t state)
> > > >>  {
> > > >>  	uint32_t val;
> > > 
> > > -- 
> > > Jani Nikula, Intel Open Source Technology Center