[Why & How]
Consolidate dmub access to a single interface. This makes it easier to
add code in the future that needs to run every time a dmub command is
requested (e.g. instrumentation, locking etc).
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Qingqing Zhuo <qingqing.zhuo@amd.com>
Signed-off-by: Josip Pavic <Josip.Pavic@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Description]
- Having excessively large margin causes failure in the static
schedulability check in some cases for SubVP + DRR
- 100us of DRR margin is sufficient based on a weeks worth of
stress testing on different display configs
Reviewed-by: Michael Strauss <Michael.Strauss@amd.com>
Acked-by: Qingqing Zhuo <qingqing.zhuo@amd.com>
Signed-off-by: Alvin Lee <Alvin.Lee2@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Description]
- Previous bug fix for audio issue included dtbclk and p-state
on the optimized boot path which is incorarect
- We only care about DISPCLK in the optimized vs. non-optimized
boot path to avoid audio issues
Reviewed-by: Saaem Rizvi <syedsaaem.rizvi@amd.com>
Acked-by: Qingqing Zhuo <qingqing.zhuo@amd.com>
Signed-off-by: Alvin Lee <Alvin.Lee2@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
when amdgpu_dm_update_connector_after_detect is called
two times successively with valid sink, memory allocated of
aconnector->timing_requested for the first call is not free.
this causes memeleak.
[How]
allocate memory only when aconnector->timing_requested
is null.
Reviewed-by: Qingqing Zhuo <Qingqing.Zhuo@amd.com>
Acked-by: Qingqing Zhuo <qingqing.zhuo@amd.com>
Signed-off-by: Hersen Wu <hersenxs.wu@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why&How]
When skipping full modeset since the only state change was a front porch
change, the DC commit sequence requires extra checks to handle non
existant plane states being asked to be removed from context.
Reviewed-by: Alvin Lee <Alvin.Lee2@amd.com>
Acked-by: Qingqing Zhuo <qingqing.zhuo@amd.com>
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
Currently we don't check if secure display is enabled before we send
command to disable secure display in dmub. It will accidentally cause
some other igt tests to fail, eg, crtc-linear-degamma.
[How]
Refactor the code we reset the secure display state to check secure
display was enabled or not before stopping it in dmub.
Reviewed-by: Wayne Lin <Wayne.Lin@amd.com>
Acked-by: Qingqing Zhuo <qingqing.zhuo@amd.com>
Signed-off-by: Alan Liu <HaoPing.Liu@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
smatch reports
drivers/gpu/drm/amd/amdgpu/../display/dc/dcn314/dcn314_dccg.c:277:6: warning: symbol
'dccg314_init' was not declared. Should it be static?
This variable is only used in one file so should be static.
Signed-off-by: Tom Rix <trix@redhat.com>
Signed-off-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
When IGT's kms_hdmi_inject forces EDID for HDMI audio, dc rejects the
request because virtual signal is not in dc_is_audio_capable_signal
function.
[How]
Includes SIGNAL_TYPE_VIRTUAL as audio capable.
Reviewed-by: Chao-kai Wang <Stylon.Wang@amd.com>
Acked-by: Qingqing Zhuo <qingqing.zhuo@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Wenchieh Chien <wenchieh.chien@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When the commit fff7eb56b3 ("drm/amd/display: Don't set dram clock
change requirement for SubVP") was merged, we missed some parts
associated with the MCLK switch. This commit adds all the missing parts.
Fixes: fff7eb56b3 ("drm/amd/display: Don't set dram clock change requirement for SubVP")
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[WHY]
It is possible to commit state multiple times in rapid succession with
FAMS enabled; if each of these commits were to set optimized_required,
then the user may see latency.
[HOW]
fw_based_mclk_switching is currently not used in dc->clk_mgr; use it
to track whether the current state has FAMS enabled;
if it has, then do not disable FAMS in prepare_bandwidth, and do not set
optimized_required.
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Wesley Chalmers <Wesley.Chalmers@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[WHY]
Writing to DRR registers such as OTG_V_TOTAL_MIN on the same frame as a
pipe commit can cause underflow.
[HOW]
Move DMUB p-state delegate into optimze_bandwidth; enabling FAMS sets
optimized_required.
This change expects that Freesync requests are blocked when
optimized_required is true.
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Wesley Chalmers <Wesley.Chalmers@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
set ABM pipe/backlight gets some issues when abm callback func pointers
are NULL. For some usecase, driver would like to control PWM level before
ABM resource is ready. However, recent flow refactor of ABM didn't
consider that use case.
[How]
Rollback flow that sending inbox command to dmub directly when ABM
function pointers aren't ready.
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Leon Huang <Leon.Huang1@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The previous 'commit ca9beb8aac ("drm/amd/display: Add logging when
setting DP sink power state fails")', it is better to unconditionally
print "failed to power up sink", because we are returning
DC_ERROR_UNEXPECTED.
Fixes: ca9beb8aac ("drm/amd/display: Add logging when setting DP sink power state fails")
Cc: Aurabindo Pillai <aurabindo.pillai@amd.com>
Cc: Fangzhi Zuo <Jerry.Zuo@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When using FPO, there is some misconfiguration that happens for the lack
of configuration of the MCLK switch in some circumstances. This commit
adds the required field update when using the MCLK switch.
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
All clock values came from firmware, but bounding box values can be
helpful in some debug situations. This commit updates some of the values
associated with clock speed and memory channels.
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[WHY]
Function "amdgpu_irq_update()" called by "amdgpu_device_ip_late_init()" is an atomic context.
We shouldn't access registers through KIQ since "msleep()" may be called in "amdgpu_kiq_rreg()".
[HOW]
Move function "amdgpu_virt_release_full_gpu()" after function "amdgpu_device_ip_late_init()",
to ensure that registers be accessed through RLCG instead of KIQ.
Call Trace:
<TASK>
show_stack+0x52/0x69
dump_stack_lvl+0x49/0x6d
dump_stack+0x10/0x18
__schedule_bug.cold+0x4f/0x6b
__schedule+0x473/0x5d0
? __wake_up_klogd.part.0+0x40/0x70
? vprintk_emit+0xbe/0x1f0
schedule+0x68/0x110
schedule_timeout+0x87/0x160
? timer_migration_handler+0xa0/0xa0
msleep+0x2d/0x50
amdgpu_kiq_rreg+0x18d/0x1f0 [amdgpu]
amdgpu_device_rreg.part.0+0x59/0xd0 [amdgpu]
amdgpu_device_rreg+0x3a/0x50 [amdgpu]
amdgpu_sriov_rreg+0x3c/0xb0 [amdgpu]
gfx_v10_0_set_gfx_eop_interrupt_state.constprop.0+0x16c/0x190 [amdgpu]
gfx_v10_0_set_eop_interrupt_state+0xa5/0xb0 [amdgpu]
amdgpu_irq_update+0x53/0x80 [amdgpu]
amdgpu_irq_get+0x7c/0xb0 [amdgpu]
amdgpu_fence_driver_hw_init+0x58/0x90 [amdgpu]
amdgpu_device_init.cold+0x16b7/0x2022 [amdgpu]
Signed-off-by: Chong Li <chongli2@amd.com>
Reviewed-by: JingWen.Chen2@amd.com
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
gcc with W=1 reports
In file included from drivers/gpu/drm/amd/amdgpu/../pm/swsmu/smu13/smu_v13_0.c:36:
./drivers/gpu/drm/amd/amdgpu/../pm/swsmu/inc/smu_v13_0.h:66:18: error:
‘pmfw_decoded_link_width’ defined but not used [-Werror=unused-const-variable=]
66 | static const int pmfw_decoded_link_width[7] = {0, 1, 2, 4, 8, 12, 16};
| ^~~~~~~~~~~~~~~~~~~~~~~
./drivers/gpu/drm/amd/amdgpu/../pm/swsmu/inc/smu_v13_0.h:65:18: error:
‘pmfw_decoded_link_speed’ defined but not used [-Werror=unused-const-variable=]
65 | static const int pmfw_decoded_link_speed[5] = {1, 2, 3, 4, 5};
| ^~~~~~~~~~~~~~~~~~~~~~~
These variables are defined and used in smu_v13_0_7_ppt.c and smu_v13_0_0_ppt.c.
There should be only one definition. So define the variables as globals
in smu_v13_0.c
Signed-off-by: Tom Rix <trix@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We need to untangle the mess where some SKL machines (at least)
declare both DDI A and DDI E to be present in their VBT, and
both using AUX A. DDI A is a ghost eDP, wheres DDI E may be a
real DP->VGA converter.
Currently that is handled by checking the VBT child devices
for conflicts before output probing. But that kind of solution
will not work for the ADL phantom dual eDP VBTs. I think on
those we just have to probe the eDP first. And would be nice
to use the same probe scheme for everything.
On these SKL systems if we probe DDI A first (which is only
natural given it's declared by VBT first) we will get an answer
via AUX, but it came from the DP->VGA converter hooked to the
DDI E, not DDI A. Thus we mistakenly register eDP on DDI A
and screw up the real DP device in DDI E.
To fix this let's check the HPD live state during the eDP probe.
If we got an answer via DPCD but HPD is still down let's assume
we got the answer from someone else.
Smoke tested on all my eDP machines (ilk,hsw-ult,tgl,adl) and
I also tested turning off all HPD hardware prior to loading
i915 to make sure it all comes up properly. And I simulated
the failure path too by not turning on HPD sense and that
correctly gave up on eDP.
I *think* Windows might just fully depend on HPD here. I
couldn't really find any other way they probe displays. And
I did find code where they also check the live state prior
to AUX transfers (something Imre and I have also talked
about perhaps doing). That would also solve this as we'd
not succeed in the eDP probe DPCD reads.
Other solutions I've considered:
- Reintrduce DDI strap checks on SKL. Unfortunately we just
don't have any idea how reliable they are on real production
hardware, and commit 5a2376d136 ("drm/i915/skl: WaIgnoreDDIAStrap
is forever, always init DDI A") does suggest that not very.
Sadly that commit is very poor in details :/
Also the systems (Asrock B250M-HDV at least) fixed by commit
41e35ffb38 ("drm/i915: Favor last VBT child device with
conflicting AUX ch/DDC pin") might still not work since we
don't know what their straps indicate. Stupid me for not
asking the reporter to check those at the time :(
We have currently two CI machines (fi-cfl-guc,fi-cfl-8700k
both MS-7B54/Z370M) that also declare both DDI A and DDI E
in VBT to use AUX A, and on these the DDI A strap is also
set. There doesn't seem to be anything hooked up to either
DDI however. But given the DDI A strap is wrong on these it
might well be wrong on the Asrock too.
Most other CI machines seem to have straps that generally
match the VBT. fi-kbl-soraka is an exception though as DDI D
strap is not set, but it is declared in VBT as a DP++ port.
No idea if there's a real physical port to go with it or not.
- Some kind of quirk just for the cases where both DDI A and DDI E
are present in VBT. Might be feasible given we've ignored
DDI A in these cases up to now successfully. But feels rather
unsatisfactory, and not very future proof against funny VBTs.
References: https://bugs.freedesktop.org/show_bug.cgi?id=111966
Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230417131728.7705-4-ville.syrjala@linux.intel.com
Add a mechanism by which we can enable the HPD sense for
individual encoders.
This will be used during eDP probing to figure out if
anything is actually connected. The normal intel_hpd_irq_setup()
thing doesn't work since we only do that after probing the
outputs, and we only enable HPD sense for encoders that were
successfully probed.
The other idea that crossed my minds was to just turn on
HPD sense for all pins before output probing and let hpd_irq_setup()
clean it up afterwards. But that doesn't work for BXT/GLK where
the HPD invert information comes from the VBT child device.
So looks like this really needs to be per-encoder.
v2: Give it a better name (Jani)
v3: Deal with mtl
Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com> #v2
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230417131728.7705-3-ville.syrjala@linux.intel.com
The address of a data structure member was determined before
a corresponding null pointer check in the implementation of
the functions “nvkm_fanpwm_create” and “nvkm_fantog_create”.
Thus avoid the risk for undefined behaviour by moving the assignment
for the data structure member “fan” behind two null pointer checks.
This issue was detected by using the Coccinelle software.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/3d0215dc-74d4-2c42-2eee-7a5fcf62b9f1@web.de
During a suspend/resume cycle the VO power domain will be disabled and
the VOP2 registers will reset to their default values. After that the
cached register values will be out of sync and the read/modify/write
operations we do on the window registers will result in bogus values
written. Fix this by re-initializing the register cache each time we
enable the VOP2. With this the VOP2 will show a picture after a
suspend/resume cycle whereas without this the screen stays dark.
Fixes: 604be85547 ("drm/rockchip: Add VOP2 driver")
Cc: stable@vger.kernel.org
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Tested-by: Chris Morgan <macromorgan@hotmail.com>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20230413144347.3506023-1-s.hauer@pengutronix.de
