Instead of direct calls of the link training functions, use a modeset
commit to retrain a DP link in SST mode, similarly to how this is done
in DP-MST mode. Originally the current way was chosen presumedly,
because there wasn't a well-established way in place for the driver to
do an internal (vs. userspace/kernel client) commit. Since then such
internal commits became a common place (initial-, HDMI/TC link reset
commit), so there is no reason to handle the DP-SST link-retraining case
differently.
At the end of the current sequence the HW reported a FIFO underrun -
without other issues visible to users - because during retraining the
link's encoder/port was disabled/re-enabled without also
disabling/re-enabling the corresponding pipe/transcoder (as required by
the spec); the corresponding underrun error message was suppressed as a
known issue. Based on Ankit's test on DG2 the underrun error was still
reported as it got detected with some (vblank) delay wrt. other
platforms. Switching to a modeset commit resolves these underrun related
issues.
Cc: Ankit Nautiyal <ankit.k.nautiyal@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Suraj Kandpal <suraj.kandpal@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240712135724.660399-1-imre.deak@intel.com
Calculate the vblank delay in the vblank evasion code correctly
for interlaced modes.
The current code assumes that we won't be using an interlaced mode.
That assumption is actually valid since we've defeatured interlaced
scanout in commit f71c9b7bc3 ("drm/i915/display: Prune Interlace
modes for Display >=12") for DSB capable platforms. However the
feature is still present in the hardware, and if we ever find the
need to re-enable it seems better to calculate the vblank delay
correctly.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240624191032.27333-2-ville.syrjala@linux.intel.com
Reviewed-by: Animesh Manna <animesh.manna@intel.com>
Starting with Xe_LPD+, although FIA is still used to readout Type-C pin
assignment, part of Type-C support is moved to PICA and programming
PORT_TX_DFLEXDPMLE1(*) registers is not applicable anymore like it was
for previous display IPs (e.g. see BSpec 49190).
v2:
- Mention Bspec 49190 as a reference of instructions for previous
IPs. (Shekhar Chauhan)
- s/Xe_LPDP/Xe_LPD+/ in the commit message. (Matt Roper)
- Update commit message to be more accurate to the changes in the IP.
(Imre Deak)
Bspec: 65750, 65448
Reviewed-by: Shekhar Chauhan <shekhar.chauhan@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240625202652.315936-1-gustavo.sousa@intel.com
Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
DSB_STATUS bit 16 is supposed to be a sticky bit informing us whether
the DSB was idle or not when the pipe's delayed vblank (when double
buffered registers latch) occurred. Unfortunately it turns out this
is a level triggred signal, ie. the bit will be set whenever the
DSB is busy during the scanline window between start of delayed
vblank and vtotal. Try to document that fact by renaming the bit.
Sadly this also thwarts my plan to use this bit to sanity check
that the (to be introduced) DSB based vblank evasion did its job
correctly. That would require an edge triggered signal instead.
So looks like we'll have to rely mostly on luck instead :(
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240611133344.30673-12-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
eDP1.5 support ALPM with Panel Replay as well. We need to check ALPM
related things for Panel Replay as well.
Bspec: 68920
v5:
- instead of using crtc_state->has_panel_replay provide aux_less flag
v4:
- do not move _compute_psr2_sdp_prior_scanline_indication
- move ALPM related checks to separate function
v3: move vblank check as well
v2: do not move Vblank >= PSR2_CTL Block Count Number maximum line count
check
Signed-off-by: Jouni Högander <jouni.hogander@intel.com>
Reviewed-by: Animesh Manna <animesh.manna@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240619062131.4021196-9-jouni.hogander@intel.com
In case of legacy cursor update, the cursor VMA needs to be unpinned
only after vblank. This exceeds the lifetime of the whole atomic commit.
Any trick I attempted to keep the atomic commit alive didn't work, as
drm_atomic_helper_setup_commit() force throttles on any old commit that
wasn't cleaned up.
The only option remaining is to remove the plane from the atomic commit,
and use the same path as the legacy cursor update to clean the state
after vblank.
Changes since previous version:
- Call the memset for plane state immediately when scheduling vblank,
this prevents a use-after-free in cursor cleanup.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Uma Shankar <uma.shankar@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240522053341.137592-4-maarten.lankhorst@linux.intel.com
The cursor hardware only does sync updates, and thus the hardware
will be scanning out from the old fb until the next start of vblank.
So in order to make the legacy cursor fastpath actually safe we
should not unpin the old fb until we're sure the hardware has
ceased accessing it. The simplest approach is to just use a vblank
work here to do the delayed unpin.
Not 100% sure it's a good idea to put this onto the same high
priority vblank worker as eg. our timing critical gamma updates.
But let's keep it simple for now, and it we later discover that
this is causing problems we can think about adding a lower
priority worker for such things.
This patch is slightly reworked by Maarten
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Uma Shankar <uma.shankar@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240522053341.137592-3-maarten.lankhorst@linux.intel.com
I don't think the display hardware really has such chroma
plane tile row alignment requirements as outlined in
commit d156135e6a ("drm/i915/tgl: Make sure a semiplanar
UV plane is tile row size aligned")
Bspec had the same exact thing to say about earlier hardware
as well, but we never cared and things work just fine.
The one thing mentioned in that commit that is definitely
true however is the fence alignment issue. But we don't
deal with that on earlier hardware either. We do have code
to deal with that issue for the first color plane, but not
the chroma planes. So I think if we did want to check this
more extensively we should do it in the same places where
we already check the first color plane (namely
convert_plane_offset_to_xy() and intel_fb_bo_framebuffer_init()).
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240612204712.31404-10-ville.syrjala@linux.intel.com
Different planes could have different alignment requirements
even for the same format/modifier. Collect the alignment
requirements across all planes capable of scanning out the
fb such that the alignment is satisfactory to all those
planes.
So far this was sort of handled by making sure intel_surf_alignment()
declares the superset of all planes' alignment requirements,
but maintaining that manually is annoying. So we're going to move
towards each plane declaring only its own requirements, and thus
we need code to generate the superset.
v2: Drop the borked per-plane vma optimization (Imre)
Assert that the plane's declared alignment is POT (Imre)
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240612204712.31404-5-ville.syrjala@linux.intel.com
Reviewed-by: Imre Deak <imre.deak@intel.com>
Different hardware generations have different scanout alignment
requirements. Introduce a new vfunc that will allow us to
make that distinction without horrible if-ladders.
For now we directly plug in the existing intel_surf_alignment()
and intel_cursor_alignment() functions.
For fbdev we (temporarily) introduce intel_fbdev_min_alignment()
that simply queries the alignment from the primary plane of
the first crtc.
TODO: someone will need to fix xe's alignment handling
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240612204712.31404-4-ville.syrjala@linux.intel.com
Acked-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
