The idea that ctg uses different HPD live state bits is
total nonsense, at least on my machine (Dell Latitude
E5400).
The only reason DP-B even works on my ctg is that DP-D
live state is stuck high, even though there is no physical
DP-D port. So when the detect checks DP-B live state it
sees the stuck live state of DP-D instead. If I hack
the driver to not register DP-D at all, and thus we never
enabe DP-D HPD, DP-B stops working as well.
Just to put some conclusive evidence into this mess,
here are the actual hotplug register values for each port:
Everything disconnected:
PORT_HOTPLUG_EN (0x00061110): 0x00000000
PORT_HOTPLUG_STAT (0x00061114): 0x00000000
PORT_HOTPLUG_EN (0x00061110): 0x08000000
PORT_HOTPLUG_STAT (0x00061114): 0x08000000
PORT_HOTPLUG_EN (0x00061110): 0x10000000
PORT_HOTPLUG_STAT (0x00061114): 0x00000000
PORT_HOTPLUG_EN (0x00061110): 0x20000000
PORT_HOTPLUG_STAT (0x00061114): 0x00000000
Only port B connected:
PORT_HOTPLUG_EN (0x00061110): 0x00000000
PORT_HOTPLUG_STAT (0x00061114): 0x00000000
PORT_HOTPLUG_EN (0x00061110): 0x08000000
PORT_HOTPLUG_STAT (0x00061114): 0x08000000
PORT_HOTPLUG_EN (0x00061110): 0x10000000
PORT_HOTPLUG_STAT (0x00061114): 0x00000000
PORT_HOTPLUG_EN (0x00061110): 0x20000000
PORT_HOTPLUG_STAT (0x00061114): 0x20000000
Only port C connected:
PORT_HOTPLUG_EN (0x00061110): 0x00000000
PORT_HOTPLUG_STAT (0x00061114): 0x00000000
PORT_HOTPLUG_EN (0x00061110): 0x08000000
PORT_HOTPLUG_STAT (0x00061114): 0x08000000
PORT_HOTPLUG_EN (0x00061110): 0x10000000
PORT_HOTPLUG_STAT (0x00061114): 0x10000000
PORT_HOTPLUG_EN (0x00061110): 0x20000000
PORT_HOTPLUG_STAT (0x00061114): 0x00000000
So the enable bit and live state bit always match 1:1.
Acked-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230302161013.29213-4-ville.syrjala@linux.intel.com
sizeof(struct intel_dmc) > 1024 bytes, allocated on all platforms as
part of struct drm_i915_private, whether they have DMC or not.
Allocate struct intel_dmc dynamically, and hide all the dmc details
behind an opaque pointer in intel_dmc.c.
Care must be taken to take into account all cases: DMC not supported on
the platform, DMC supported but not initialized, and DMC initialized but
not loaded. For the second case, we need to move the wakeref out of
struct intel_dmc.
v2:
- Rebase to kzalloc dmc after runtime pm get (Imre)
Cc: Imre Deak <imre.deak@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230301122944.1298929-4-jani.nikula@intel.com
There's only one reference to the struct intel_dmc members dc_state,
target_dc_state, and allowed_dc_mask within intel_dmc.c, begging the
question why they are under struct intel_dmc to begin with.
Moreover, the only references to i915->display.dmc outside of
intel_dmc.c are to these members.
They don't belong. Move them from struct intel_dmc to struct
i915_power_domains, which seems like a more suitable place.
Cc: Imre Deak <imre.deak@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230301122944.1298929-1-jani.nikula@intel.com
All intel_suspend_hw() does is clear PCH_LP_PARTITION_LEVEL_DISABLE bit
in SOUTH_DSPCLK_GATE_D for LPT LP. intel_suspend_hw() gets called from
i915_drm_suspend().
However, i915_drm_suspend_late() calls
intel_display_power_suspend_late(), which in turn calls hsw_enable_pc8()
on HSW and BDW. The first thing that does is clear
PCH_LP_PARTITION_LEVEL_DISABLE bit in SOUTH_DSPCLK_GATE_D.
Remove the duplicated clearing of the bit, effectively delaying it from
i915_drm_suspend() to i915_drm_suspend_late(), and remove the
unnecessary intel_suspend_hw() function altogether.
Cc: Imre Deak <imre.deak@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/f732a7922c2450b41169c9b79a80fba97ab00592.1677678803.git.jani.nikula@intel.com
Grab the HDR DPCD refresh timeout (time we need to wait after
writing the sourc OUI before the HDR DPCD registers are ready)
from the VBT.
Windows doesn't even seem to have any default value for this,
which is perhaps a bit weird since the VBT value is documented
as TGL+ and I thought the HDR backlight stuff might already be
used on earlier platforms. To play it safe I left the old
hardcoded 30ms default in place. Digging through some internal
stuff that seems to have been a number given by the vendor for
one particularly slow TCON. Although I did see 50ms mentioned
somewhere as well.
Let's also include the value in the debug print to ease
debugging, and toss in the customary connector id+name as well.
The TGL Thinkpad T14 I have sets this to 0 btw. So the delay
is now gone on this machine:
[CONNECTOR:308:eDP-1] Detected Intel HDR backlight interface version 1
[CONNECTOR:308:eDP-1] Using Intel proprietary eDP backlight controls
[CONNECTOR:308:eDP-1] SDR backlight is controlled through PWM
[CONNECTOR:308:eDP-1] Using native PCH PWM for backlight control (controller=0)
[CONNECTOR:308:eDP-1] Using AUX HDR interface for backlight control (range 0..496)
[CONNECTOR:308:eDP-1] Performing OUI wait (0 ms)
Cc: Lyude Paul <lyude@redhat.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230220164718.23117-1-ville.syrjala@linux.intel.com
Reviewed-by: Jouni Högander <jouni.hogander@intel.com>
FBC on gen2/3 seems to trigger page table errors. No visual
artifacts are visible, and essentially the same FBC
code works on gen4 so these seem entirely spurious. There
are also hints in gen3 bspec indicating that certain bits
in PGTBL_ER are just not wired up correctly in the
hardware.
Ideally we'd want to mask out only the bogus bits, but
sadly there is no mask for PGTBL_ER, and instead we are
forced to mask out all page table errors via EMR :(
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230125185234.21599-6-ville.syrjala@linux.intel.com
Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
At least on some platforms (tested on ctg) the way
vgacon does screen blanking seems to flag constant
FIFO underruns, which means we have to be prepared
for them while the driver is loading. Currently
there is a time window between drm_crtc_init() and
intel_sanitize_fifo_underrun_reporting() during
which FIFO underrun reporting is in fact marked as
enabled. Thus we may end up mistakenly detecting
these bogus underruns during driver init.
Close the race by marking FIFO underrun reporting
as disabled prior to even registering the crtc.
intel_sanitize_fifo_underrun_reporting()/etc. will
re-enable it later if needed.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230125185234.21599-2-ville.syrjala@linux.intel.com
Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
The audio logic lives in the transcoder rather than the pipe,
so start tracking it like that.
This is only really important for bigjoiner cases where tracking
by pipe doesn't work at all since intel_audio_codec_{enable,disable}()
won't even be called for the slave pipe. This means the state
checker won't find the ELD for the slave pipe and gets upset.
The PD->has_audio readout does currently work since that gets
read out from the same transcoder for both pipes.
For other cases this doesn't actually matter since it's only
the normal pipe transcoders that are audio capable, whereas
the more special transcoders (EDP/DSI) are not.
v2: Fix kernel docs
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230222151454.24888-1-ville.syrjala@linux.intel.com
Reviewed-by: Uma Shankar <uma.shankar@intel.com>
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/8222
The DSI code has some local hacks to program TRANS_VBLANK on
TGL+ (ICL DSI transcoders didn't have this register). That
will not work when we need to start using the delayed vblank
(for DSB purposes). Too lazy to figure out what the is going
on there, so just sprinkle FIXMEs in the hopes someone else
will spot them eventually.
v2: Only TRANS_{HBLANK,SET_CONTEXT_LATENCY} still no not
exist for DSI transcoders, only TRANS_VBLANK
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230213225258.2127-12-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
On TGL VBLANK.VBLANK_START was the mechanism by which we can
delay the pipe's internal vblank in relation to the transcoder's
vblank. On ADL+ that no longer does anything. Instead we must
now use the new TRANS_SET_CONTEXT_LATENCY register. Program it
accordingly.
And since VBLANK.VBLANK_START is no longer used by the hardware
on ADL+ let's just zero it out to make it stand out in register
dumps. Seeing the zeroed value should hopefully remind people
to check the other register instead.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230213225258.2127-11-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Define the contents of the transcoder timing registers using
REG_GENMASK() & co. For ease of maintenance let's just define
the bitmasks with the full 16bit width (also used by the
current hand rolled stuff) even though not all bits are actually
used. None of the unsued bits have ever contained anything.
Jani spotted that the CRT load detection code did use narrower
bitmasks, so that is now going to change. But that is fine
since any garbage in the high bits would have been caught by
the state checker that always used the full 16bit masks.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230213225258.2127-10-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
