Nicolai Hähnle [Tue, 20 Aug 2019 13:39:53 +0000 (15:39 +0200)]
drm/amdgpu: prevent memory leaks in AMDGPU_CS ioctl
Error out if the AMDGPU_CS ioctl is called with multiple SYNCOBJ_OUT and/or
TIMELINE_SIGNAL chunks, since otherwise the last chunk wins while the
allocated array as well as the reference counts of sync objects are leaked.
drm/amd/display: Lock the CRTC when setting CRC source
[Why]
We need to ensure that we're holding the lock on the CRTC when setting
the CRC source since we're modifying the CRTC state directly.
We also need to wait for any outstanding non-blocking commits to finish
so they aren't reading state that's potentially being modified -
non-blocking commits don't hold the CRTC lock while doing commit tail
work.
[How]
Lock the CRTC using its mutex. While holding the lock check if there's
any commit active on the CRTC - if there is, it's non-blocking and
we should wait until it's finished by waiting for hw_done to be
signaled since that's the last point where we touch CRTC state.
drm/amd/display: Split out DC programming for CRC capture
[Why]
Calling amdgpu_dm_crtc_set_crc_source in amdgpu_dm directly has the
consequence of adding additional vblank references or starting DPRX
CRC capture more than once without calling stop first.
Vblank references for CRC capture should be managed entirely by opening
and closing the CRC file from userspace.
Stream state also shouldn't be required on the CRC so we can close the
file after the CRTC has been disabled.
[How]
Do DC programming required for configuring CRC capture separately from
setting the source. Whenever we re-enable or reset a CRC this
programming should be reapplied.
CRC vblank reference handling in amdgpu_dm can be entirely dropped after
this.
Stream state also no longer needs to be required since we can just defer
the programming to when the stream is actually enabled.
drm/amd/display: Use connector list for finding DPRX CRC aux
[Why]
This change is a refactor in preparation for adding locking and removing
the requirement for a stream state on the CRTC for enabling CRC capture
to fix igt@kms_plane_multiple@* warnings.
[How]
We can get the aux by finding the matching connector for the CRTC
with the assumption that we're not doing cloning.
drm/amd/display: Check return code for CRC drm_crtc_vblank_get
[Why]
The call to drm_crtc_vblank_get can fail if vblank is disabled and
we try to increment the reference.
Since drm_crtc_vblank_get internally drops the reference when it fails
it means the subsequent drm_crtc_vblank_put(...) when closing the file
drops a zero reference.
This was found via igt@kms_plane@pixel-format-pipe-A-planes.
[How]
Check the return code and return it on failure.
We wouldn't have been able to enable CRC reading anyway since vblank
wasn't enabled.
Evan Quan [Fri, 16 Aug 2019 09:11:46 +0000 (17:11 +0800)]
drm/amd/powerplay: correct SW smu11 thermal range settings
Problems with current settings:
1. The min value was overrided to 0 on Vega20 & Navi10. While
the expected should be -273.15 C.
2. The thermal min/max threshold was output in wrong unit on
Navi10 & Arcturus. As TEMP_RANGE_MIN/MAX is already in
millicelsius. And "*1000" in smu_v11_0_start_thermal_control
makes the output wrongly.
Kevin Wang [Mon, 19 Aug 2019 15:38:02 +0000 (23:38 +0800)]
drm/amd/powerplay: fix variable type errors in smu_v11_0_setup_pptable
fix size type errors, from uint32_t to uint16_t.
it will cause only initializes the highest 16 bits in
smu_get_atom_data_table function.
bug report:
This fixes the following static checker warning.
drivers/gpu/drm/amd/amdgpu/../powerplay/smu_v11_0.c:390 smu_v11_0_setup_pptable()
warn: passing casted pointer '&size' to 'smu_get_atom_data_table()' 32 vs 16.
Kevin Wang [Fri, 16 Aug 2019 07:17:42 +0000 (15:17 +0800)]
drm/amd/powerplay: add smu_smc_read_sensor support for arcturus
the bellow patch refine the sensor read sequence,
but missed to add arcuturs support. (arcuturs_ppt.c)
it will cause some sensor is not supported in arcturus.
drm/amd/powerplay: change smu_read_sensor sequence in smu
Yong Zhao [Fri, 2 Aug 2019 02:55:50 +0000 (22:55 -0400)]
drm/amdkfd: Fill the name field in node topology with asic name v2
The name field in node topology has not been used. We re-purpose it to
hold the asic name, which can be queried by user space applications
through sysfs.
drm/amd/display: Calculate bpc based on max_requested_bpc
[Why]
The only place where state->max_bpc is updated on the connector is
at the start of atomic check during drm_atomic_connector_check. It
isn't updated when adding the connectors to the atomic state after
the fact. It also doesn't necessarily reflect the right value when
called in amdgpu during mode validation outside of atomic check.
This can cause the wrong bpc to be used even if the max_requested_bpc
is the correct value.
[How]
Don't rely on state->max_bpc reflecting the real bpc value and just
do the min(...) based on display info bpc and max_requested_bpc.
Fixes: 01933ba42d3d ("drm/amd/display: Use current connector state if NULL when checking bpc") Signed-off-by: Nicholas Kazlauskas <[email protected]> Reviewed-by: Leo Li <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
Alex Deucher [Thu, 15 Aug 2019 13:27:09 +0000 (08:27 -0500)]
drm/amdgpu/gfx9: update pg_flags after determining if gfx off is possible
We need to set certain power gating flags after we determine
if the firmware version is sufficient to support gfxoff.
Previously we set the pg flags in early init, but we later
we might have disabled gfxoff if the firmware versions didn't
support it. Move adding the additional pg flags after we
determine whether or not to support gfxoff.
Yong Zhao [Tue, 13 Aug 2019 18:38:03 +0000 (14:38 -0400)]
drm/amdgpu: Set VM_L2_CNTL.PDE_FAULT_CLASSIFICATION to 0 for GFX10
We have done this for pre-GFX10 asics, but GFX10 did not pick up the
new change. The below is the commit message for that change.
This is recommended by HW designers. Previously when it was set to 1,
the PDE walk error in VM fault will be treated as
PERMISSION_OR_INVALID_PAGE_FAULT rather than usually expected OTHER_FAULT.
As a result, the retry control in VM_CONTEXT*_CNTL will change accordingly.
The above behavior is kind of abnormal. Furthermore, the
PDE_FAULT_CLASSIFICATION == 1 feature was targeted for very old ASICs
and it never made it way to production. Therefore, we should set it to 0.
Oak Zeng [Thu, 1 Aug 2019 19:55:45 +0000 (14:55 -0500)]
drm/amdgpu: Export function to flush TLB of specific vm hub
This is for kfd to reuse amdgpu TLB invalidation function.
On gfx10, kfd only needs to flush TLB on gfx hub but not
on mm hub. So export a function for KFD flush TLB only on
specific hub.
Leo Li [Tue, 6 Aug 2019 17:50:28 +0000 (13:50 -0400)]
drm/amd/display: Load NV12 SOC BB from firmware
[Why]
Previous SOC bounding box firmware loading logic was for NV10, when we
still had it in firmware. Now that it's brought into driver code, and
NV12 BB is in firmware, this logic needs to be repurposed for NV12.
[How]
Set SOC_BOUNDING_BOX_VALID to false, and add the dcn_2_0_nv12_soc BB
struct. In init_soc_bounding_box, load firmware to nv12 BB instead.
In addition, conditionally update and patch the BB, depending on ASIC
REV.
Michael Strauss [Fri, 26 Jul 2019 16:04:12 +0000 (12:04 -0400)]
drm/amd/display: Enable MPO with pre-blend color processing (RGB)
[Why]
DCN10 performs color processing before MPC combination, causes color
shift in RGB colorspaces when positive brightness offset is applied
However, YCbCr is still unfixed and remains disabled
[How]
Add layerIndex to dc_plane_state and dc_plane_info structs
Re-enable MPO when brightness is adjusted and colorspace is not YCbCr
Set rear plane's brightness offset to 0 when front plane visible
yanyan kang [Mon, 29 Jul 2019 09:46:44 +0000 (17:46 +0800)]
drm/amd/display: audio cannot switch to internal when display turns off
[why]
disable_az_endpoint has been skipped because
dc->debug.az_endpoint_mute_only = true.
[how]
set dc->debug.az_endpoint_mute_only false when PPLIB’s PME notification function
pointer is not NULL at the dcn10_resource construct function,because right now
SMU/PPLIB and DAL all have the AZ D3 force PME notification implemented. AZ D3 should work.
Wenjing Liu [Fri, 26 Jul 2019 18:53:20 +0000 (14:53 -0400)]
drm/amd/display: check hpd before retry verify link cap
[why]
During detection link training if a display is disconnected,
the current code will retry 3 times of link training
on disconnected link before giving up.
[how]
Before each retry check for HPD status, only retry
verify link cap when HPD is still high.
Also put a 10ms delay between each retry to improve
the chance of success.
drm/amd/display: fix dcn20 global sync dml param extraction
Currently the paremeters are extracted as if dml is calculating
using pipes as we pass them in. in reality, dml internally merges
pipes into planes if pipe split is detected.
This change adds reverse logic to dcn20_calculate_dlg_params so
that the global sync parameters can be correctly extracted for
all the pipes when pipe split is enabled.
drm/amd/display: Register VUPDATE_NO_LOCK interrupts for DCN2
[Why]
These are needed to send back DRM vblank events in the case where VRR
is on. Without the interrupt enabled we're deferring the events into the
vblank queue and userspace is left waiting forever to get back the
events they need.
Found using igt@kms_vrr - the test fails immediately due to vblank
timeout.
[How]
Register them the same way we're handling it for DCN1.
Charlene Liu [Wed, 24 Jul 2019 22:14:46 +0000 (18:14 -0400)]
drm/amd/display: wake up ogam mem pwr before programming ocsc
[Description]
OGAM_MEM_PWR could stay in light up when driver woke up to update gamma.
either disable MEM_LOW power feature or set to OGAM_bypass could make artificial color distortion goes away.
Easy reproduce after LOW_MEM Power feature enables and resume from S3.
Nikola Cornij [Tue, 23 Jul 2019 20:38:02 +0000 (16:38 -0400)]
drm/amd/display: Remove 4:2:2 DSC support
[why]
If DSC is available, a higher picture quality is achieved by using
DSC with 4:4:4 format. Using 4:2:2 instead does not offer any benefit
and would only introduce loss of quality. Removing it reduces
maintenance and testing effort.
Lewis Huang [Mon, 22 Jul 2019 19:23:32 +0000 (15:23 -0400)]
drm/amd/display: reprogram VM config when system resume
[Why]
The vm config will be clear to 0 when system enter S4. It will
cause hubbub didn't know how to fetch data when system resume.
The flip always pending because earliest_inuse_address and
request_address are different.
Eric Yang [Thu, 11 Jul 2019 02:41:51 +0000 (22:41 -0400)]
drm/amd/display: Enable type C hotplug
[Why and How]
We want to change where timing is done for alt mode.
Some of the commented out #ifs are needed for DCN20
so we enable them for that case.
drm/amd/display: Remove duplicate interface for programming FB
[Why]
There are currently two interfaces for exactly the same thing:
hupb_update_dchub in hupb and update_dchub in hubbub. The hubbub
version is currently unused past dcn10, largely because the call
from the dcn10 hardware sequencer does not call through the
interface, so the hupb interface was used instead. This is
confusing because of the duplicate code, the unused functions,
and the fact that more that one block currently owns this set
of registers.
[How]
Remove the hubp interface entirely, as well as the register
declarations that are not longer needed because of this. Change
the call site to always call the hubbub version through the
interface. Fix the update_dchub function in dcn20_hubbub.c to
program the correct registers for dcn20.
Jun Lei [Mon, 22 Jul 2019 13:45:20 +0000 (09:45 -0400)]
drm/amd/display: fix pipe selection logic in validate
[why]
Resource mapping done in dcn20_validate_bandwidth has a flaw: When a full
update is performed, the HWSS will only update the MPCC tree for the stream
that is updated as opposed to all streams. This means that when mapping pipes
in validation, care must be taken to not change any existing mapping, otherwise it
leads to partial hw programming
[how]
it's not strictly necessary to track which stream/mpcc tree is being updated, but
rather it's sufficient to compare current and new state and just keep pipes that were
previously already mapped unchanged.
Nikola Cornij [Wed, 17 Jul 2019 23:02:14 +0000 (19:02 -0400)]
drm/amd/display: Set DSC before DIG front-end is connected to its back-end
[why]
At the time DIG FE is connected to its BE, the clocks in OTG are enabled and
PHY will also be set up. When DSC has to be used to fit the stream into the
available bandwidth, without DSC being set DIG could get exposed to the
higer bandwidth it (or link) could handle. This causes the HW to "reject"
video enable setup (the register shows that video enable was attempted, but
the status bit shows it as disabled).
[how]
- Separate DSC setup into DSC register config and DSC PPS SDP setup
- Move most of the DSC setup (register config) to before
dcn10_link_encoder_connect_dig_be_to_fe() is called
- Set up DSC PPS SDP after DIG FE is connected to its BE. This is because
setting DSC PPS SDP before that has no effect.
Jun Lei [Thu, 18 Jul 2019 14:02:40 +0000 (10:02 -0400)]
drm/amd/display: clean up DML for DCN2x
[why]
Previous "less risky" implemenation of 3 tiered fallback is no longer necessary since
DMLv2 has gone through proper validation. v2 can now be used as the default and 1
level of fallback can be removed
[Why]
underflow seen on certain monitor setups caused by making dcnxx_init_hw
generic
[How]
by moving dcn20_init_hw into dcn10, we added a dcn-specific clk_mgr
init (dc->clk_mgr->funcs->init_clocks()). Thus, put old clk_mgr
memset in an else statement so both memsets don't get set
Martin Leung [Tue, 9 Jul 2019 19:15:17 +0000 (15:15 -0400)]
drm/amd/display: Make init_hw and init_pipes generic for seamless boot
[Why]
For seamless boot the init_hw sequence must be split into
actual hardware vs pipes, in order to defer pipe initialization to set mode
and skip of pipe-destructive sequences
[How]
made dcn10_init_hw and dcn10_init_pipes generic for future dcns to inherit
deleted dcn20 specific versions. This is part 1 of a 2 partimplementation
of seamless boot
Jun Lei [Tue, 16 Jul 2019 17:02:41 +0000 (13:02 -0400)]
drm/amd/display: wait for pending complete when enabling a plane
[why]
When planes are enabled, they must be enabled using VSYNC update (not immediate).
However, before the VUPDATE occurs, DM may call with an "immediate" flip which is address
only. This operation would normally be okay, but if the locking for immediate flip happens
to occur before the VUPDATE associated with the initial plane enablement, it will cause HW
to hang.
[how]
HWSS should enforce plane enable in HW to be synchronous with the call that enables the plane.
Jun Lei [Mon, 15 Jul 2019 14:41:47 +0000 (10:41 -0400)]
drm/amd/display: fixup DPP programming sequence
[why]
DC does not correct account for the fact that DPP DTO is double buffered while DPP ref is not.
This means that when DPP ref clock is lowered when it's "safe to lower", the DPP blocks that need
an increased divider will temporarily have actual DPP clock drop below minimum while DTO
double buffering takes effect. This results in temporary underflow.
[how]
To fix this, DPP clock cannot be programmed atomically, but rather be broken up into the DTO and the
ref. Each has a separate "safe to lower" logic. When doing "prepare" the ref and dividers may only increase.
When doing "optimize", both may decrease. It is guaranteed that we won't exceed max DPP clock because
we do not use dividers larger than 1.
drm/amd/display: make firmware info only load once during dc_bios create
Currently every time DC wants to access firmware info we make a call
into VBIOS. This makes no sense as there is nothing that can change
runtime inside fw info and can cause issues when calling unstable
bios during bringup.
This change eliminate this behavior by only calling bios once for fw
info and keeping it stored as part of dc_bios.
David Galiffi [Mon, 24 Jun 2019 14:34:13 +0000 (10:34 -0400)]
drm/amd/display: Synchronous DisplayPort Link Training
[WHY]
We require a method to perform synchronous link training.
[HOW]
Sync LT is broken into 3 basic steps.
"Begin" starts the state machine, and resets "preferred" link settings.
"Attempt" will attempt to train the link with a given set of training
parameters.
"End" stops the state machine, and will optionally disable the link phy.
Between "Begin" and "End" DPCD:600h must not be set to "2"
(D3:Powered Down).
Between "Begin" and "End", there may be multiple "Attempts" with different
training parameters.
Bayan Zabihiyan [Wed, 10 Jul 2019 20:00:53 +0000 (16:00 -0400)]
drm/amd/display: Fix frames_to_insert math
[Why]
The math on deciding on how many
"frames to insert" sometimes sent us over the max refresh rate.
Also integer overflow can occur if we have high refresh rates.
[How]
Instead of clipping the frame duration such that it doesn’t go below the min,
just remove a frame from the number of frames to insert. +
Use unsigned long long for intermediate calculations to prevent
integer overflow.
Anthony Koo [Fri, 12 Jul 2019 14:52:54 +0000 (10:52 -0400)]
drm/amd/display: fix issue where 252-255 values are clipped
[Why]
When endpoint is at the boundary of a region, such as at 2^0=1
we find that the last segment has a sharp slope and some points
are clipped at the top.
[How]
If end point is 1, which is exactly at the 2^0 region boundary, we
need to program an additional region beyond this point.
Kevin Wang [Tue, 13 Aug 2019 02:25:25 +0000 (10:25 +0800)]
drm/amd/powerplay: fix message of SetHardMinByFreq failed when feature is disabled
the direct send message to smc to set hard clokc will failed
when smc clock dpm feature is disabled.
so use function of smu_set_hard_freq_range to replace it.
the function will check feature enablement.
eg: when uclk (mclk) dpm feature is disabled on navi10
[ 300.675901] amdgpu: [powerplay] failed send message: SetHardMinByFreq(28)
param: 0x00020064 response 0xfffffffb
drm/amdgpu: simplify and cleanup setting the dma mask
Use dma_set_mask_and_coherent to set both masks in one go, and remove
the no longer required fallback, as the kernel now always accepts
larger than required DMA masks. Fail the driver probe if we can't
set the DMA mask, as that means the system can only support a larger
mask.
drm/radeon: simplify and cleanup setting the dma mask
Use dma_set_mask_and_coherent to set both masks in one go, and remove
the no longer required fallback, as the kernel now always accepts
larger than required DMA masks. Fail the driver probe if we can't
set the DMA mask, as that means the system can only support a larger
mask.
drm/amdgpu: handle PCIe root ports with addressing limitations
amdgpu uses a need_dma32 flag to indicate to the drm core that some
allocations need to be done using GFP_DMA32, but it only checks the
device addressing capabilities to make that decision. Unfortunately
PCIe root ports that have limited addressing exist as well. Use the
dma_addressing_limited instead to also take those into account.
projects / linux.git / log