Atm, the driver programs explicitly the default transparent link
training mode (0x55) to DP_PHY_REPEATER_MODE even if no LTTPRs are
detected.
This conforms to the spec (3.6.6.1):
"DP upstream devices that do not enable the Non-transparent mode of
LTTPRs shall program the PHY_REPEATER_MODE register (DPCD Address
F0003h) to 55h (default) prior to link training"
however writing the default value to this DPCD register seems to cause
occasional link training errors at least for a DELL WD19TB TBT dock, when
no LTTPRs are detected.
Writing to DP_PHY_REPEATER_MODE will also cause an unnecessary timeout
on systems without any LTTPR.
To fix the above two issues let's assume that setting the default mode
is redundant when no LTTPRs are detected. Keep the existing behavior and
program the default mode if more than 8 LTTPRs are detected or in case
the read from DP_PHY_REPEATER_CNT returns an invalid value.
References: https://gitlab.freedesktop.org/drm/intel/-/issues/2801
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Khaled Almahallawy <khaled.almahallawy@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210118183143.1145707-1-imre.deak@intel.com
Limit pre-skl plane stride to below 4k or 8k pixels (depending on
the platform). We do this in order guarantee that TILEOFF/OFFSET.x
does not get too big.
Currently this is not a problem as we align SURF to 4k, and so
TILEOFF/OFFSET only have to deal with a single tile's worth of
pixels. But for async flips we're going to have to bump SURF
alignment to 256k, and thus we can no longer guarantee
TILEOFF/OFFSET.x will stay within acceptable bounds. We can avoid
this by borrowing a trick from the skl+ code and limit the max
plane stride to whatever value we can fit into TILEOFF/OFFSET.x.
The slight downside is that we may end up doing GTT remapping in
a few more cases where previously we did not have to. But since
that will only happen with huge buffers I'm not really concerned
about it.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210111163711.12913-3-ville.syrjala@linux.intel.com
Reviewed-by: Karthik B S <karthik.b.s@intel.com>
Up to now we were reading some DRAM information from MCHBAR register
and from pcode what is already not good but some GEN12(TGL-H and ADL-S)
platforms have MCHBAR DRAM information in different offsets.
This was notified to HW team that decided that the best alternative is
always apply the 16gb_dimm watermark adjustment for GEN12+ platforms
and read the remaning DRAM information needed to other display
programming from pcode.
So here moving the DRAM pcode function to intel_dram.c, removing
the duplicated fields from intel_qgv_info, setting and using
information from dram_info.
v2:
- bring back num_points to intel_qgv_info as num_qgv_point can be
overwritten in icl_get_qgv_points()
- add gen12_get_dram_info() and simplify gen11_get_dram_info()
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210128164312.91160-2-jose.souza@intel.com
Atm the driver will calculate a wrong MST timeslots/MTP (aka time unit)
value for MST streams if the link parameters (link rate or lane count)
are limited in a way independent of the sink capabilities (reported by
DPCD).
One example of such a limitation is when a MUX between the sink and
source connects only a limited number of lanes to the display and
connects the rest of the lanes to other peripherals (USB).
Another issue is that atm MST core calculates the divider based on the
backwards compatible DPCD (at address 0x0000) vs. the extended
capability info (at address 0x2200). This can result in leaving some
part of the MST BW unused (For instance in case of the WD19TB dock).
Fix the above two issues by calculating the PBN divider value based on
the rate and lane count link parameters that the driver uses for all
other computation.
Bugzilla: https://gitlab.freedesktop.org/drm/intel/-/issues/2977
Cc: Lyude Paul <lyude@redhat.com>
Cc: Ville Syrjala <ville.syrjala@intel.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ville Syrjala <ville.syrjala@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210125173636.1733812-2-imre.deak@intel.com
Prevent the ICL HDR plane pipeline from performing YUV color range
correction twice when the input is in limited range. This is done by
removing the limited-range code from icl_program_input_csc().
Before this patch the following could happen: user space gives us a YUV
buffer in limited range; per the pipeline in [1], the plane would first
go through a "YUV Range correct" stage that expands the range; the plane
would then go through the "Input CSC" stage which would also expand the
range because icl_program_input_csc() would use a matrix and an offset
that assume limited-range input; this would ultimately cause dark and
light colors to appear darker and lighter than they should respectively.
This is an issue because if a buffer switches between being scanned out
and being composited with the GPU, the user will see a color difference.
If this switching happens quickly and frequently, the user will perceive
this as a flickering.
[1] https://01.org/sites/default/files/documentation/intel-gfx-prm-osrc-icllp-vol12-displayengine_0.pdf#page=281
Cc: stable@vger.kernel.org
Signed-off-by: Andres Calderon Jaramillo <andrescj@chromium.org>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201215224219.3896256-1-andrescj@google.com
In order to make the dbuf state computation less fragile
let's make it stand on its own feet by not requiring someone
to peek into a crystall ball ahead of time to figure out
which pipes need to be added to the state under which potential
future conditions. Instead we compute each piece of the state
as we go along, and if any fallout occurs that affects more than
the current set of pipes we add the affected pipes to the state
naturally.
That requires that we track a few extra thigns in the global
dbuf state: dbuf slices for each pipe, and the weight each
pipe has when distributing the same set of slice(s) between
multiple pipes. Easy enough.
We do need to follow a somewhat careful sequence of computations
though as there are several steps involved in cooking up the dbuf
state. Thoguh we could avoid some of that by computing more things
on demand instead of relying on earlier step of the algorithm to
have filled it out. I think the end result is still reasonable
as the entire sequence is pretty much consolidated into a single
function instead of being spread around all over.
The rough sequence is this:
1. calculate active_pipes
2. calculate dbuf slices for every pipe
3. calculate total enabled slices
4. calculate new dbuf weights for any crtc in the state
5. calculate new ddb entry for every pipe based on the sets of
slices and weights, and add any affected crtc to the state
6. calculate new plane ddb entries for all crtcs in the state,
and add any affected plane to the state so that we'll perform
the requisite hw reprogramming
And as a nice bonus we get to throw dev_priv->wm.distrust_bios_wm
out the window.
v2: Keep crtc_state->wm.skl.ddb
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210122205633.18492-8-ville.syrjala@linux.intel.com
The dbuf state will be where we collect all the inter-pipe dbuf
allocation stuff. Start by adding the actual per-pipe ddb entries
there.
Originally the plan was to move them there outright, but that no longer
works as we're no longer guaranteed to have a dbuf state when it comes
time to sanity check the ddb overlaps in skl_commit_modeset_enables().
I think when I wrote this originally we did the watermark/ddb
calculation last, and so we couldn't have any crtcs in the state w/o
also having the dbuf state. But that has since changed and we do the
watermark/ddb calculation much earlier, and thus it is now possible
to commit crtcs w/o a dbuf state. So we keep another copy of the
information in the crtc state.
v2: Rebase
v3: Duplicate the entries instead of moving
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210122205633.18492-6-ville.syrjala@linux.intel.com
To get sensible vblank timestamping behaviour we need to feed
the vmax based timings to the vblank code, otherwise it'll chop
off the scanline counter when it exceeds the minumum vtotal.
Additionally with VRR we have three cases to consider when we
generate the vblank timestamp:
1) we are in vertical active
-> nothing special needs to be done, just return the current
scanout position and the core will calculate the timestamp
corresponding to the past time when the current vertical
active started
2) we are in vertical blank and no push has been sent
-> the hardware will keep extending the vblank presumably
to its maximum length, so we make the timestmap match the
expected time when the max length vblank will end. Since
the timings used for this are now based on vmax nothing
special actually needs to be done
3) we are in vblank and a push has been sent so the vblank is
about to terminate
-> presumably we want the timestmap to accurately reflect
when the vblank will terminate, so we use the sampled
frame timestamp vs. current timestamp to guesstimate
how far along the vblank exit we are, and then we
adjust the reported scanout position accordingly so
that the core will see that the vblank is close to
ending.
v2:
* Fix the else if (use_scanline_Counter) (Manasi)
Signed-off-by: Manasi Navare <manasi.d.navare@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Manasi Navare <manasi.d.navare@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210122232647.22688-17-manasi.d.navare@intel.com
VRR achieves vblank stretching using the HW PUSH functionality.
So once the VRR is enabled during modeset then for each flip
request from userspace, in the atomic tail pipe_update_end()
we need to set the VRR push bit in HW for it to terminate
the vblank at configured flipline or anytime after flipline
or latest at the Vmax.
The HW clears the PUSH bit after the double buffer updates
are completed.
v2:
* Move send push to after irq en (Manasi)
* Call send push unconditionally (Jani N)
v3:
* Stall w.r.t Vrr vmax (Manasi, Gary Smith)
v4:
* Remove the rmw (Ville)
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Gary Smith <gary.k.smith@intel.com>
Signed-off-by: Manasi Navare <manasi.d.navare@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210122232647.22688-11-manasi.d.navare@intel.com
This forces a complete modeset if vrr drm crtc state goes
from enabled to disabled and vice versa.
This patch also computes vrr state variables from the mode timings
and based on the vrr property set by userspace as well as hardware's
vrr capability.
v2:
*Rebase
v3:
* Vmin = max (vtotal, vmin) (Manasi)
v4:
* set crtc_state->vrr.enable = 0 for disable request
v5:
* drm_dbg_kms, squash crtc states def patch (Jani N)
v6:
* Move vrr modeset check to separate function (Jani N)
v7:
* Ville's fixes - vmin, vmax rename, fix rounding dir
* Add pipeline full, flipline to crtc state
* Pass conn state to vrr_compute_config (Ville)
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Signed-off-by: Manasi Navare <manasi.d.navare@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210122232647.22688-6-manasi.d.navare@intel.com
This function sets the VRR property for connector based
on the platform support, EDID monitor range and DP sink
DPCD capability of outputing video without msa
timing information.
v8:
* Use HAS_VRR, remove drm_conn declaration (Jani N)
* Fix typos in Comment (Jani N)
v7:
* Move the helper to separate file (Manasi)
v6:
* Remove unset of prop
v5:
* Fix the vrr prop not being set in kernel (Manasi)
* Unset the prop on connector disconnect (Manasi)
v4:
* Rebase (Mansi)
v3:
* intel_dp_is_vrr_capable can be used for debugfs, make it
non static (Manasi)
v2:
* Just set this in intel_dp_get_modes instead of new hook (Jani)
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Signed-off-by: Aditya Swarup <aditya.swarup@intel.com>
Signed-off-by: Manasi Navare <manasi.d.navare@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210122232647.22688-2-manasi.d.navare@intel.com
Render Decompression is supported with Y-Tiled main surface. The CCS is
linear and has 4 bits of data for each main surface cache line pair, a
ratio of 1:256. Additional Clear Color information is passed from the
user-space through an offset in the GEM BO. Add a new modifier to identify
and parse new Clear Color information and extend Gen12 render decompression
functionality to the newly added modifier.
v2: Fix has_alpha flag for modifiers, omit CC modifier during initial
plane config(Matt). Fix Lookup error.
v3: Fix the panic while running kms_cube
v4: Add alignment check and reuse the comments for ge12_ccs_formats(Matt)
v5: Fix typos and wrap comments(Matt)
v6:
- Use format block descriptors to get the subsampling calculations for
the CCS surface right.
- Use helpers to convert between main and CCS surfaces.
- Prevent coordinate checks for the CC surface.
- Simplify reading CC value from surface map, add description of CC val
layout.
- Remove redundant ccval variable from skl_program_plane().
v7:
- Move the CC value readout after syncing against any GPU write on the
FB obj (Nanley, Chris)
- Make sure the CC value readout works on platforms w/o struct pages
(dGFX) and other non-coherent platforms wrt. CPU reads (none atm).
(Chris)
v8:
- Rebase on the function param order change of
i915_gem_object_read_from_page().
- Clarify code comment on the clear color value format and the required
FB obj pinning/syncing by the caller.
- Remove redundant variables in
intel_atomic_prepare_plane_clear_colors().
v9:
- Fix s/sizeof(&ccval)/sizeof(ccval)/ typo.
Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Cc: Ville Syrjala <ville.syrjala@intel.com>
Cc: Shashank Sharma <shashank.sharma@intel.com>
Cc: Rafael Antognolli <rafael.antognolli@intel.com>
Cc: Nanley G Chery <nanley.g.chery@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Radhakrishna Sripada <radhakrishna.sripada@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210115213952.1040398-1-imre.deak@intel.com
Add a simple helper to read data with the CPU from the page of a GEM
object. Do the read either via a kmap if the object has struct pages
or an iomap otherwise. This is needed by the next patch, reading a u64
value from the object (w/o requiring the obj to be mapped to the GPU).
Suggested by Chris.
v2 (Chris):
- Sanitize the type and order of func params.
- Avoid consts requiring too many casts.
- Use BUG_ON instead of WARN_ON, simplify the conditions.
- Fix __iomem sparse errors.
- Leave locking/syncing/pinning up to the caller, require only that the
caller has pinned the object pages.
- Check for iomem backing store before reading via an iomap.
v3:
- Fix offset passed to io_mapping_map_wc() missing a mem.region.start
delta. (Chris, Matthew)
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20210120213834.1435710-1-imre.deak@intel.com
