Currently the KMD is using enum i915_cache_level to set caching policy for
buffer objects. This is flaky because the PAT index which really controls
the caching behavior in PTE has far more levels than what's defined in the
enum. In addition, the PAT index is platform dependent, having to translate
between i915_cache_level and PAT index is not reliable, and makes the code
more complicated.
From UMD's perspective there is also a necessity to set caching policy for
performance fine tuning. It's much easier for the UMD to directly use PAT
index because the behavior of each PAT index is clearly defined in Bspec.
Having the abstracted i915_cache_level sitting in between would only cause
more ambiguity. PAT is expected to work much like MOCS already works today,
and by design userspace is expected to select the index that exactly
matches the desired behavior described in the hardware specification.
For these reasons this patch replaces i915_cache_level with PAT index. Also
note, the cache_level is not completely removed yet, because the KMD still
has the need of creating buffer objects with simple cache settings such as
cached, uncached, or writethrough. For kernel objects, cache_level is used
for simplicity and backward compatibility. For Pre-gen12 platforms PAT can
have 1:1 mapping to i915_cache_level, so these two are interchangeable. see
the use of LEGACY_CACHELEVEL.
One consequence of this change is that gen8_pte_encode is no longer working
for gen12 platforms due to the fact that gen12 platforms has different PAT
definitions. In the meantime the mtl_pte_encode introduced specfically for
MTL becomes generic for all gen12 platforms. This patch renames the MTL
PTE encode function into gen12_pte_encode and apply it to all gen12. Even
though this change looks unrelated, but separating them would temporarily
break gen12 PTE encoding, thus squash them in one patch.
Special note: this patch changes the way caching behavior is controlled in
the sense that some objects are left to be managed by userspace. For such
objects we need to be careful not to change the userspace settings.There
are kerneldoc and comments added around obj->cache_coherent, cache_dirty,
and how to bypass the checkings by i915_gem_object_has_cache_level. For
full understanding, these changes need to be looked at together with the
two follow-up patches, one disables the {set|get}_caching ioctl's and the
other adds set_pat extension to the GEM_CREATE uAPI.
Bspec: 63019
Cc: Chris Wilson <chris.p.wilson@linux.intel.com>
Signed-off-by: Fei Yang <fei.yang@intel.com>
Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Andi Shyti <andi.shyti@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230509165200.1740-3-fei.yang@intel.com
The big switch+if statement mess in map_aux_ch() is
illegible. Split up into cleaner per-platform arrays
like we already have for the gmbus pins.
We use enum aux_ch as the index and the VBT thing as
the value. Slightly non-intuitive perhaps but if we
did it the other way around we'd have problems with
AUX_CH_A being zero, and thus any non-populated
element would look like AUX_CH_A.
v2: flip the index vs. value around
TODO: Didn't bother with the platform variants beyond the
ones that really need remapping, which means if the
VBT is bogus we end up with a nonexistent aux ch.
Might be nice to check this a bit better.
Yet another bitmask in device info?
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230509160206.25971-1-ville.syrjala@linux.intel.com
Reviewed-by: Gustavo Sousa <gustavo.sousa@intel.com>
drm_dp_dsc_sink_max_slice_count() may return 0 if something goes
wrong on the part of the DSC sink and its DPCD register. This null
value may be later used as a divisor in intel_dsc_compute_params(),
which will lead to an error.
In the unlikely event that this issue occurs, fix it by testing the
return value of drm_dp_dsc_sink_max_slice_count() against zero.
Found by Linux Verification Center (linuxtesting.org) with static
analysis tool SVACE.
Fixes: a4a157777c ("drm/i915/dp: Compute DSC pipe config in atomic check")
Signed-off-by: Nikita Zhandarovich <n.zhandarovich@fintech.ru>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230418140430.69902-1-n.zhandarovich@fintech.ru
(cherry picked from commit 51f7008239)
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
On g4x the hardware has only one audio/video DIP block. Thus
only one HDMI port can transmit audio/infoframes at a time.
Currently we pretend that multiple ports can do it at the same
time, but that doesn't actually work for two reasons:
- the state of the single hw block will get clobbered by
the multiple ports, leading to state checker failures
- the hardware will automagically disable audio/infoframe
transmission when enabled on multiple ports
To fix this let's allow only one of the ports to transmit
audio/infoframes at a time. We'll just go over all the HDMI
ports and pick the first one that is otherwise capable of
audio/infoframes. Further HDMI ports will be treated as if
they had a DVI sink connected.
In order to compute this consistently we'll also need to
always add all HDMI ports to the atomic state.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20221107194604.15227-6-ville.syrjala@linux.intel.com
Reviewed-by: Uma Shankar <uma.shankar@intel.com>
The decision to use DFP output format conversion capabilities should be
during compute_config phase.
This patch uses the members of intel_dp->dfp to only store the
format conversion capabilities of the DP device and uses the crtc_state
sink_format member, to program the protocol-converter for
colorspace/format conversion.
v2: Use sink_format to determine the color conversion config for the
pcon (Ville).
v3: Fix typo: missing 'break' in switch case (lkp kernel test robot).
v4: Add helper to check if DP supports YCBCR420.
v5: Simplify logic for computing output_format, based on the given
sink_format. (Ville).
Added scaler constraint for YCbCr420 output.
v6: Split the patch for Scaler constraint for Ycbcr420.
v7: Simplify the policy for selecting output_format:
Always try for RGB first, followed by YCBCR444, and finally by YCBCR420.
v8: Removed redundant comments, minor refactoring. (Ville)
v9: Added member for ycbcr420 passthrough cap, fixed minor issues. (Ville)
Signed-off-by: Ankit Nautiyal <ankit.k.nautiyal@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230427125605.487769-3-ankit.k.nautiyal@intel.com
The decision to use DFP output format conversion capabilities should be
during compute_config phase.
This patch adds new member to crtc_state to represent the final
output_format to the sink. In case of a DFP this can be different than
the output_format, as per the format conversion done via the PCON.
This will help to store only the format conversion capabilities of the
DP device in intel_dp->dfp, and use crtc_state to compute and store the
configuration for color/format conversion for a given mode.
v2: modified the new member to crtc_state to represent the final
output_format that eaches the sink, after possible conversion by
PCON kind of devices. (Ville)
v3: Addressed comments from Ville:
-Added comments to clarify difference between sink_format and
output_format.
-Corrected the order of setting sink_format and output_format.
-Added readout for sink_format in get_pipe_config hooks.
v4: Set sink_format for intel_sdvo too. (Ville)
v5: Rebased.
v6: Fixed condition to go for YCbCr420 format for dp and hdmi. (Ville)
v7: Fix the condition to set sink_format for HDMI.
Set hdmi output_format simply as sink_format. (Ville)
Signed-off-by: Ankit Nautiyal <ankit.k.nautiyal@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> (v3)
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230427125605.487769-2-ankit.k.nautiyal@intel.com
drm_dp_dsc_sink_max_slice_count() may return 0 if something goes
wrong on the part of the DSC sink and its DPCD register. This null
value may be later used as a divisor in intel_dsc_compute_params(),
which will lead to an error.
In the unlikely event that this issue occurs, fix it by testing the
return value of drm_dp_dsc_sink_max_slice_count() against zero.
Found by Linux Verification Center (linuxtesting.org) with static
analysis tool SVACE.
Fixes: a4a157777c ("drm/i915/dp: Compute DSC pipe config in atomic check")
Signed-off-by: Nikita Zhandarovich <n.zhandarovich@fintech.ru>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230418140430.69902-1-n.zhandarovich@fintech.ru
intel_atomic_get_new_crtc_state can return NULL, unless crtc state wasn't
obtained previously with intel_atomic_get_crtc_state, so we must check it
for NULLness here, just as in many other places, where we can't guarantee
that intel_atomic_get_crtc_state was called.
We are currently getting NULL ptr deref because of that, so this fix was
confirmed to help.
Fixes: 74a75dc908 ("drm/i915/display: move plane prepare/cleanup to intel_atomic_plane.c")
Signed-off-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230505082212.27089-1-stanislav.lisovskiy@intel.com
Display to communicate display pipe count/CDCLK/voltage configuration
to Pcode for more accurate power accounting for DG2.
Existing sequence is only sending the voltage value to the Pcode.
Adding new sequence with current cdclk associate with voltage value masking.
Adding pcode request when any pipe power well will disable or enable.
v2: - Make intel_cdclk_need_serialize static to make CI compiler happy.
v3: - Removed redundant return(Jani Nikula)
- Changed intel_cdclk_power_usage_to_pcode_(pre|post)_notification to be
static and also naming to intel_cdclk_pcode_(pre|post)_notify(Jani Nikula)
- Changed u8 to be u16 for cdclk parameter in intel_pcode_notify function,
as according to BSpec it requires 10 bits(Jani Nikula)
- Replaced dev_priv's with i915's(Jani Nikula)
- Simplified expression in intel_cdclk_need_serialize(Jani Nikula)
- Removed redundant kernel-doc and indentation(Jani Nikula)
v4: - Fixed some checkpatch warnings
v5: - According to HW team comments that change should affect only DG2,
fix correspodent platform check to account this.
v6: - Added one more missing IS_DG2 check(Vinod Govindapillai)
Signed-off-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230504093959.12085-1-stanislav.lisovskiy@intel.com
Define and use the bitmasks for the x/y components
of the skl+ scaler window pos/size registers.
We stick to the full 16 bits mask here even though the
hardware limits are actually lower. The current (ADL)
hardware maximums are in fact: 14 bits for X size, 13 bits
for X pos, 13 bits for Y size/pos. Yes, that is correct,
X pos has less bits than the X size for some reason. But
that doesn't actually matter for now as we don't currently
even support such wide displays without the use of bigjoiner
(due to max plane width limit).
v2: Switch back to full 16bit masks since that's what
we use transcoder timign regs and PIPESRC as well
Cc: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230426135019.7603-6-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Define and use the bitmasks for the x/y components
of the ilk+ panel filter window pos/size registers.
Note that we stick to the full 16 bit mask even though
the actual hardware limits are lower (and somewhat
platform dependent). BDW is actually limited to
13 bits horizontal and 12 bits vertical, with the high
bits being hardwired to zero. HSW should have the same
limits as BDW. And pre-HSW should be limited to 12bits
in both directions as that's already the limit of the
transcoder timing registers. Curiously on HSW and earlier
platforms all 16 bits can actually be set, but presumably
the hardware ignores the high bits.
v2: Switch back to full 16bit masks since that's what
we use transcoder timign regs and PIPESRC as well
Cc: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230426135019.7603-2-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
The intel_dsi_msleep() helper skips sleeping if the MIPI-sequences have
a version of 3 or newer and the panel is in vid-mode.
This is based on the big comment around line 730 which starts with
"Panel enable/disable sequences from the VBT spec.", where
the "v3 video mode seq" column does not have any wait t# entries.
Checking the Windows driver shows that it does always honor
the VBT delays independent of the version of the VBT sequences.
Commit 6fdb335f1c ("drm/i915/dsi: Use unconditional msleep for
the panel_on_delay when there is no reset-deassert MIPI-sequence")
switched to a direct msleep() instead of intel_dsi_msleep()
when there is no MIPI_SEQ_DEASSERT_RESET sequence, to fix
the panel on an Acer Aspire Switch 10 E SW3-016 not turning on.
And now testing on a Nextbook Ares 8A shows that panel_on_delay
must always be honored otherwise the panel will not turn on.
Instead of only always using regular msleep() for panel_on_delay
do as Windows does and always use regular msleep() everywhere
were intel_dsi_msleep() is used and drop the intel_dsi_msleep()
helper.
Changes in v2:
- Replace all intel_dsi_msleep() calls instead of just
the intel_dsi_msleep(panel_on_delay) call
Cc: stable@vger.kernel.org
Fixes: 6fdb335f1c ("drm/i915/dsi: Use unconditional msleep for the panel_on_delay when there is no reset-deassert MIPI-sequence")
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230425194441.68086-1-hdegoede@redhat.com
(cherry picked from commit fa83c12132)
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
The intel_dsi_msleep() helper skips sleeping if the MIPI-sequences have
a version of 3 or newer and the panel is in vid-mode.
This is based on the big comment around line 730 which starts with
"Panel enable/disable sequences from the VBT spec.", where
the "v3 video mode seq" column does not have any wait t# entries.
Checking the Windows driver shows that it does always honor
the VBT delays independent of the version of the VBT sequences.
Commit 6fdb335f1c ("drm/i915/dsi: Use unconditional msleep for
the panel_on_delay when there is no reset-deassert MIPI-sequence")
switched to a direct msleep() instead of intel_dsi_msleep()
when there is no MIPI_SEQ_DEASSERT_RESET sequence, to fix
the panel on an Acer Aspire Switch 10 E SW3-016 not turning on.
And now testing on a Nextbook Ares 8A shows that panel_on_delay
must always be honored otherwise the panel will not turn on.
Instead of only always using regular msleep() for panel_on_delay
do as Windows does and always use regular msleep() everywhere
were intel_dsi_msleep() is used and drop the intel_dsi_msleep()
helper.
Changes in v2:
- Replace all intel_dsi_msleep() calls instead of just
the intel_dsi_msleep(panel_on_delay) call
Cc: stable@vger.kernel.org
Fixes: 6fdb335f1c ("drm/i915/dsi: Use unconditional msleep for the panel_on_delay when there is no reset-deassert MIPI-sequence")
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230425194441.68086-1-hdegoede@redhat.com
Type-C PHYs are taking longer than expected for Aux IO Power Enabling.
Workaround: Increase the timeout.
---v2
-change style on how we mention WA [Ankit]
-fix bat error by creating new func that is only called for aux power
well scenarios so we can avoid null pointer error as it is called
everywhere.
--v3
-Add non-default enable_timeout to power well descriptor which avoids
adding more platform checks [Imre]
--v4
-Remove Bspec link from top to bottom remove WA link from commit put it
on comment [Jani]
-enable_timeout in ms and add .fixed_enable_delay too [Imre]
--v5
-move power_wells instead of duplicating them [Imre]
Bspec: 55480
Cc: Jani Nikula <jani.nikula@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Signed-off-by: Suraj Kandpal <suraj.kandpal@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Animesh Manna <animesh.manna@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230418131425.1285088-1-suraj.kandpal@intel.com
On ADLP+ Bspec allows DC5/6 to be enabled while power well 2 is enabled.
Since the AUX and DDI power wells (except for port A/B) are also backed
by power well 2, this would suggest that DC5/6 can be enabled while any
of these AUX or DDI port functionalities are used. As opposed to this
AUX transfers will time out on ADLP TypeC ports while DC6 is enabled.
Until the restriction for DC5/6 is clarified in Bspec let's assume that
the intention is to allow for using these power states while pipe A/B is
enabled, but only for combo ports which can be used with eDP outputs.
Similarly assume that AUX transaction initiated by the driver on any port
requires DC states to be disabled.
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Radhakrishna Sripada <radhakrishna.sripada@intel.com>
Fixes: 88c4879384 ("drm/i915: Use separate "DC off" power well for ADL-P and DG2")
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230424200205.1732941-1-imre.deak@intel.com
