If an MST stream is modeset, its state must be checked along all the
other streams on the same MST link, for instance to resolve a BW
overallocation of a non-sink MST port or to make sure that the FEC is
enabled/disabled the same way for all these streams.
To prepare for that this patch adds all the stream CRTCs to the atomic
state and marks them for modeset similarly to tgl+ platforms. (If the
state computation doesn't change the state the CRTC is switched back to
fastset mode.)
So far on tgl+ this was required because all streams in the topology
shared the master transcoder. For older platforms this didn't apply but
adding all the state is required now on all platforms based on the
above.
v2:
- Add code and commit log comment clarifying the requirements on old/new
platforms. (Stan)
- Rename the function based on the new semantic. (Ville)
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231030155843.2251023-15-imre.deak@intel.com
Atm, the BW allocated for an MST stream doesn't take into account the
DSC control symbol (EOC) and data alignment overhead on the local (first
downstream) MST link (reflected by the data M/N/TU values) and - besides
the above overheads - the FEC symbol overhead on 8b/10b remote
(after a downstream branch device) MST links.
In addition the FEC overhead used on the local link is a fixed amount,
which only applies to certain modes, but not enough for all modes; add a
code comment clarifying this.
Fix the above by calculating the data M/N values with the total BW
overhead (not including the SSC overhead, since this isn't enabled by
the source device) and using this the PBN and TU values for the local
link and PBN for remote links (including SSC, since this is mandatory
for links after downstream branch devices).
For now keep the current fixed FEC overhead as a minimum, since this is
what bspec requires for audio functionality.
Calculate the effective link BW in a clearer way, applying the channel
coding efficiency based on the coding type. The calculation was correct
for 8b/10b, but not for 128b/132b links; this patch leaves the behavior
for this unchanged, leaving the fix for a follow-up.
v2:
- Fix TU size programmed to the HW, making it match the payload size
programmed to the payload table.
v3:
- Add code comment about the connection between the payload's size in
the payload table and the corresponding PBN value. (Ville)
- Add WARN_ON(remote_m_n.tu < dp_m_n.tu). (Ville)
- Add code comment about factors not accounted for by the BW
calculation in intel_dp_mst_mode_valid_ctx() (and compute config).
(Ville)
- Simplify calculation of PBN to remote_m_n.tu * mst_state->pbn_div.
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com> (v1)
Signed-off-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231107001505.3370108-2-imre.deak@intel.com
Enable FEC in crtc_state, as soon as it's known it will be needed by
DSC. This fixes the calculation of BW allocation overhead, in case DSC
is enabled by falling back to it during the encoder compute config
phase (vs. enabling FEC due to DSC being enabled on other streams).
v2:
- Enable FEC only in intel_dp_mst_find_vcpi_slots_for_bpp(), since
only by that will crtc_state->port_clock be set, which in turn is
needed by intel_dp_is_uhbr().
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com> (v1)
Signed-off-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231030155843.2251023-11-imre.deak@intel.com
After drm_connector_init() is called the connector is visible to the
rest of the kernel via the drm_mode_config::connector_list. Make
sure that the DSC AUX device and capabilities are setup by that time.
Another race condition is adding the connector to the connector list
before drm_connector_helper_add() sets the connector helper functions.
That's an unrelated issue, for which the fix is for a follow-up. One
solution would be adding the connector to the connector list only
during its registration in drm_connector_register().
Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Fixes: 808b43fa7e ("drm/i915/dp_mst: Set connector DSC capabilities and decompression AUX")
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231030155843.2251023-2-imre.deak@intel.com
Display driver shall read DPCD 00071h[3:1] during configuration
to get PSR setup time. This register provides the setup time
requirement on the VSC SDP entry packet. If setup time cannot be
met with the current timings
(e.g., PSR setup time + other blanking requirements > blanking time),
driver should enable sending VSC SDP one frame earlier before sending
the capture frame.
BSpec: 69895 (PSR Entry Setup Frames 17:16)
v2: Write frames before su entry to correct register (Ville, Jouni)
Move frames before su entry calculation to it's
own function (Ville, Jouni)
Rename PSR Entry Setup Frames register to indicate
Lunarlake specificity (Jouni)
v3: Modify setup entry frames calculation function to
return the actual frames (Ville)
Match comment with actual implementation (Jouni)
v4: Drop "set" from function naming (Jouni, Ville)
Use i915 instead of dev_priv (Jouni)
Signed-off-by: Mika Kahola <mika.kahola@intel.com>
Reviewed-by: Jouni Högander <jouni.hogander@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231106114228.146574-1-mika.kahola@intel.com
Fix below compiler warning:
intel_tc.c:1879:11: error: ‘%d’ directive output may be truncated
writing between 1 and 11 bytes into a region of size 3
[-Werror=format-truncation=]
"%c/TC#%d", port_name(port), tc_port + 1);
^~
intel_tc.c:1878:2: note: ‘snprintf’ output between 7 and 17 bytes
into a destination of size 8
snprintf(tc->port_name, sizeof(tc->port_name),
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
"%c/TC#%d", port_name(port), tc_port + 1);
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
v2: use kasprintf(Imre)
v3: use const for port_name, and fix tc mem leak(Imre)
Fixes: 3eafcddf76 ("drm/i915/tc: Move TC port fields to a new intel_tc_port struct")
Cc: Mika Kahola <mika.kahola@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Nirmoy Das <nirmoy.das@intel.com>
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231026125636.5080-1-nirmoy.das@intel.com
On GLK CDCLK frequency needs to be at least 2*96 MHz when accessing
the audio hardware. Currently we bump the CDCLK frequency up
temporarily (if not high enough already) whenever audio hardware
is being accessed, and drop it back down afterwards.
With a single active pipe this works just fine as we can switch
between all the valid CDCLK frequencies by changing the cd2x
divider, which doesn't require a full modeset. However with
multiple active pipes the cd2x divider trick no longer works,
and thus we end up blinking all displays off and back on.
To avoid this let's just bump the CDCLK frequency to >=2*96MHz
whenever multiple pipes are active. The downside is slightly
higher power consumption, but that seems like an acceptable
tradeoff. With a single active pipe we can stick to the current
more optiomal (from power comsumption POV) behaviour.
Cc: stable@vger.kernel.org
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/9599
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231031160800.18371-1-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
We are preparing for Xe driver. Xe driver doesn't have i915_sw_fence
implementation. Lets drop i915_sw_fence usage from display code and
use dma_fence interfaces directly.
For this purpose stack dma fences from related objects into new plane
state. Drm_gem_plane_helper_prepare_fb can be used for fences in new
fb. Separate local implementation is used for Stacking fences from old fb
into new plane state. Then wait for these stacked fences during atomic
commit. There is no be need for separate GPU reset handling in
intel_atomic_commit_fence_wait as the fences are signaled when GPU hang is
detected and GPU is being reset.
v4:
- Drop to_new_plane_state suffix from add_dma_resv_fences
- Use dma_resv_usage_rw(false) (DMA_RESV_USAGE_WRITE)
v3:
- Rename add_fences and it's parameters
- Remove signaled check
- Remove waiting old_plane_state fences
v2:
- Add fences from old fb into new_plane_state->uapi.fence rather than
into old_plane_state->uapi.fence
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: José Roberto de Souza <jose.souza@intel.com>
Signed-off-by: Jouni Högander <jouni.hogander@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231031084557.1181630-1-jouni.hogander@intel.com
The extra DPLL power domain is currently handled in three places:
- combo_pll_enable()
- combo_pll_disable()
- readout_dpll_hw_state()
First two of those are low level PLL funcs, but the third is a higher
level thing. So the current situation is rather inconsistent. Unify
this by moving the PLL enable/disable up one level. This also means
the extra power domain could be trivially be used by other platforms
as well.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231012123522.26045-4-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Just include the JSL/EHL DPLL4 extra power domain in the dpll_info
struct. This way the same approach could be used by other platforms
as well (should the need arise), and we don't have to sprinkle
platform checks all over the place.
Note that I'm perhaps slightly abusing things here as
power_domain==0 (which is actually POWER_DOMAIN_DISPLAY_CORE) now
indicates that no extra power domain is needed. I suppose using
POWER_DOMAIN_INVALID would be more correct, but then we'd have to
sprinkle that to all the other DPLLs.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231012123522.26045-3-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
With MTL adding PICA between the port and the real phy, the path
add for DG2 stopped being followed and newer platforms are simply using
the older path for TC phys. LNL is no different than MTL in this aspect,
so just add it to the mess. In future the phy and port designation and
deciding if it's TC should better be cleaned up.
To make it just a bit better, also change intel_phy_is_snps() to show
this is DG2-only.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Gustavo Sousa <gustavo.sousa@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231026184045.1015655-3-lucas.demarchi@intel.com
For Lunar Lake, DDI-A is connected to C10 PHY, while TC1-TC3 are connected
to C20 phy, like in Meteor Lake. Update the check in intel_is_c10phy()
accordingly.
This reverts the change in commit e388ae97e2 ("drm/i915/display:
Eliminate IS_METEORLAKE checks") that turned that into a display engine
version check. The phy <-> port connection is very SoC-specific and not
related to that version.
IS_LUNARLAKE() is defined to 0 in i915 as it's expected that the
(upcoming) xe driver is the one defining the platform, with i915 only
driving the display side.
Bspec: 70818
Reviewed-by: Gustavo Sousa <gustavo.sousa@intel.com>
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231026184045.1015655-2-lucas.demarchi@intel.com
Every know and then we receive the following error when running
for example IGT test kms_flip.
[drm] *ERROR* PHY G Read 0d80 failed after 3 retries.
[drm] *ERROR* PHY G Write 0d81 failed after 3 retries.
Since the error is sporadic in nature, the patch proposes
to reset the message bus after every successful or unsuccessful
read or write operation.
v2: Add FIXME's to indicate the experimental nature of
this workaround (Rodrigo)
v3: Dropping the additional delay as moving reset to *_read_once()
and *_write_once() functions seem unnecessary delay
Signed-off-by: Mika Kahola <mika.kahola@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231016125544.719963-1-mika.kahola@intel.com
