- 07 Aug, 2023 40 commits
-
-
Martin Leung authored
This version brings along following fixes: - Fix a regression in blank pixel data caused by coding mistake - Fix a bug when searching for insert_above_mpcc - Update add plane to context logic with a new algorithm - Adjust visual confirm dpp height offset to be 1/240 of v addressable - Add interface to enable DPIA trace - Support plane clip with zero recout size - Blocking invalid 420 modes on HDMI TMDS for DCN31 and DCN314 - Make mpc32 functions available to future DCNs - Change HDCP update sequence for DM Acked-by:
Tom Chung <chiahsuan.chung@amd.com> Signed-off-by:
Martin Leung <martin.leung@amd.com> Tested-by:
Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by:
Alex Deucher <alexander.deucher@amd.com>
-
Bhawanpreet Lakha authored
Refactor the sequence in hdcp_update_display() to use mod_hdcp_update_display(). Previous sequence: - remove()->add() This Sequence was used to update the display, (mod_hdcp_update_display didn't exist at the time). This meant for any hdcp updates (type changes, enable/disable) we would remove, reconstruct, and add. This leads to unnecessary calls to psp eventually New Sequence using mod_hdcp_update_display(): - add() once when stream is enabled - use update() for all updates The update function checks for prev == new states and will not unnecessarily end up calling psp via add/remove. Reviewed-by:
Qingqing Zhuo <qingqing.zhuo@amd.com> Acked-by:
Bhawanpreet Lakha <bhawanpreet.lakha@amd.com> Tested-by:
-
Wenjing Liu authored
[why] Secondary DPP pipes are used for rendering secondary layers of planes. The name "for layer" doesn't make it obvious. The function is acquiring a free pipe as secondary dpp pipe only. We rename it so it is more obvious. In a future follow up change, we want to add functions to acquire free pipe as opp head pipe or otg master pipe as well. They will have their separate allocation priority. Reviewed-by:
Jun Lei <jun.lei@amd.com> Acked-by:
Wenjing Liu <wenjing.liu@amd.com> Tested-by:
-
Wenjing Liu authored
[why] generic dc resource file should not know what an optimal idle pipe is because this is dcn hardware dependent. [how] We move the optimial pipe searching logic in dcn specific layer. Reviewed-by:
-
Alvin Lee authored
[Description] In overclocking scenarios the max memclk could be higher than the DC mode limit. However, for configs that don't support MCLK switching we need to set the max memclk to the overclocked max instead of the DC mode max or we could result in underflow. Reviewed-by:
Samson Tam <samson.tam@amd.com> Acked-by:
Alvin Lee <alvin.lee2@amd.com> Tested-by:
-
Mike Hsieh authored
Add a new vendor ID code for reference Reviewed-by:
Mike Hsieh <mike.hsieh@amd.com> Tested-by:
-
Leo Chen authored
[Why & How] HDMI TMDS does not have ODM support. Filtering 420 modes that exceed the 4096 FMT limitation on DCN314 will resolve intermittent corruptions issues. Reviewed-by:
Nicholas Kazlauskas <nicholas.kazlauskas@amd.com> Acked-by:
Leo Chen <sancchen@amd.com> Tested-by:
-
Leo Chen authored
[Why & How] HDMI TMDS does not have ODM support. Filtering 420 modes that exceed the 4096 FMT limitation on DCN31 will resolve intermittent corruptions issues. Reviewed-by:
-
Leo Chen authored
[Why & How] Revert commit 6917b0b7 ("drm/amd/display: Read down-spread percentage from lut to adjust dprefclk.") This change was causing 240hz display to not light up after s0i3. Reviewed-by:
-
Wenjing Liu authored
[why] With previous ODM + MPO policy we will only allocate pipes when MPO plane has intersection with current ODM slice. With this policy we have to indicate scaling change when plane clip is updated in case plane clip change causes MPO plane to go in or out current ODM slice. With new ODM + MPO policy we allocate pipe independent from plane clip size so we no longer treat it as a scaling change. There no need to do a full update due to clip size change anymore. Reviewed-by:
-
Wenjing Liu authored
[why] For timing with large v addressable visual confirm is just too small. It is difficult to tell visually which DPP we are using. On the other hand with timing with small v addressable visual confirm is too large and covers the UI area. [how] We calculate visual confirm dpp height offset based on v addressable so it stays relatively the same height i.e. 1/240 verticle portion of the screen. Reviewed-by:
-
Wenjing Liu authored
[why] Preivous algorithm for finding an optimal idle pipe for a new plane was implemented to handle dynamic pipe allocation when MPO plane moves from one ODM slice to the other. Now pipe allocation is more static so it no longer depends on the MPO plane's position. We are simplifying our logic and remove unnecessary handling in our code. [how] Apply a new simplified version of pipe resource allocation logic to reduce unnecessary flip delay caused by swapping secondary dpp pipe to other MPC blending tree. Reviewed-by:
-
Wenjing Liu authored
[why] We only acquire MPO pipe for blending tree where the plane clip will be rendered. If an MPO plane is outside current ODM slice rect, we will skip pipe allocation. With new programming policy we want to allocate pipes for every ODM slice blending tree even for those whose ODM slice rect doesn't intersect with plane clip. This is aligned with DML validation so the pipe topology is programmed independently from the plane's position and dst plane size. [how] - Remove the logic to allocate pipe only when the MPO plane intersects with ODM slice and replace with the new logic to always allocate pipes. - Remove the logic to tear down ODM configuration in favor for supporting secondary MPO planes. - Remove the logic to use full update when MPO goes accross ODM slice boundary. Reviewed-by:
-
Wenjing Liu authored
[why] If plane clip is outside current pipe's ODM slice rect, our logic will fail validation because we assume that when a pipe is acquired to support a plane clip, it must blend some portion of the plane on the screen. This assumption needs to be changed. When a pipe is acquired to render the plane, we are now allowing it to support a case where it can take minimum viewport and draw with zero recout size when the plane clip is outside current ODM slice rect. The reason is that we want to allocate and get the pipe pre-programmed so it is ready to be rendered when user moves the plane over to the current ODM slice with a fast update. Whereas with the existing solution when user moves the plane over, we will need to allocate a pipe as needed and power it up and program it through a full update. This not only impacts the user experience with unnecessary delay of a frame but also doesn't generate any benefit to the user because DML doesn't support it. DML will invalidate based on worst case scenario and it doesn't depend on the plane location. So having our programming sequence support such dynamic pipe allocation is not meaningful anyway. [how] In build scaler params allow recout to be zero size and if viewport is smaller than minimum, set minimum viewport size. Reviewed-by:
-
Austin Zheng authored
[Why] Flash of corruption observed when UCLK switching after transitioning from DTBCLK to DPREFCLK on subVP(DP) + subVP(HDMI) config Scenario where DPREFCLK is required instead of DTBCLK is not expected [How] Always set the DTBCLK source as DTBCLK0 Reviewed-by:
Austin Zheng <austin.zheng@amd.com> Tested-by:
-
Austin Zheng authored
[Why] Front-end would be programmed using the stream update flags set from the previous update and the full update should be triggered whenever commit_state_no_check gets called. [How] Set all stream update flags before programming the front-end Clear all flags that got set to avoid redundant programming Reviewed-by:
-
Wesley Chalmers authored
Make the mpc32 functions available for future DCNs to use Reviewed-by:
Wesley Chalmers <wesley.chalmers@amd.com> Tested-by:
-
Wesley Chalmers authored
[WHY] Currently, when insert_plane is called with insert_above_mpcc parameter that is equal to tree->opp_list, the function returns NULL. [HOW] Instead, the function should insert the plane at the top of the tree. Cc: Mario Limonciello <mario.limonciello@amd.com> Cc: Alex Deucher <alexander.deucher@amd.com> Cc: stable@vger.kernel.org Reviewed-by:
-
Stylon Wang authored
[Why] DPIA traces from DMUB is not enabled by default, which is less convenient to debug DPIA related issues because we have to resort to other means to enable DPIA trace. [How] Reuse existing DC debug mask to enable DPIA trace log from kernel command line. This makes debugging DPIA issues easier especially when system power state (suspend/reboot) is involved. To turn on DPIA trace, simply add "amdgpu.dcdebugmask=0x80" to kernel command line. Reviewed-by:
Aurabindo Pillai <aurabindo.pillai@amd.com> Acked-by:
Stylon Wang <stylon.wang@amd.com> Tested-by:
-
Stylon Wang authored
[Why] DPIA traces from DMUB is not enabled by default, which is less convenient to debug DPIA related issues because we have to resort to other debug tools to enable DPIA trace. [How] Exposes interfaces to update trace mask from the DMUB GPINT commands. Also provides DC implementations to enable DPIA trace. Reviewed-by:
-
Wenjing Liu authored
[why] There was unfortunately a coding mistake. It gets caught with an ultrawide monitor that requires ODM 4:1 combine. We are blanking or unblanking pixel data we are supposed to enumerate through all ODM pipes and program DPG for each of those pipes. However the coding mistake causes us to program only the first and last ODM pipes. Cc: Mario Limonciello <mario.limonciello@amd.com> Cc: Alex Deucher <alexander.deucher@amd.com> Cc: stable@vger.kernel.org Reviewed-by:
-
Shashank Sharma authored
MES allocates process level doorbells, but there is no userspace client to consume it. It was only being used for the MES ring tests (in kernel), and was written by kernel doorbell write. The previous patch of this series has changed the MES ring test code to use kernel level MES doorbells. This patch now cleans up the process level doorbell allocation code which is not required. Cc: Alex Deucher <alexander.deucher@amd.com> Cc: Christian Koenig <christian.koenig@amd.com> Reviewed-by:
Christian König <christian.koenig@amd.com> Signed-off-by:
Shashank Sharma <shashank.sharma@amd.com> Signed-off-by:
Arvind Yadav <arvind.yadav@amd.com> Signed-off-by:
-
Shashank Sharma authored
This patch: - Removes the existing doorbell management code, and its variables from the doorbell_init function, it will be done in doorbell manager now. - uses the doorbell page created for MES kernel level needs (doorbells for MES self tests) - current MES code was allocating MES doorbells in MES process context, but those were getting written using kernel doorbell calls. This patch instead allocates a MES kernel doorbell for this (in add_hw_queue). V2: Create an extra page of doorbells for MES during kernel doorbell creation (Alex) V4: Move MES doorbell size and page offset objects in this patch from patch 6. Cc: Alex Deucher <alexander.deucher@amd.com> Cc: Christian Koenig <christian.koenig@amd.com> Reviewed-by: -
Shashank Sharma authored
This patch removes some variables and functions from KFD doorbell handling code, which are no more required since doorbell manager is handling doorbell calculations. Cc: Alex Deucher <alexander.deucher@amd.com> Cc: Christian Koenig <christian.koenig@amd.com> Cc: Felix Kuehling <Felix.Kuehling@amd.com> Reviewed-by:
Felix Kuehling <Felix.Kuehling@amd.com> Signed-off-by:
-
Hamza Mahfooz authored
We should be checking to see if async flips are supported in amdgpu_dm_atomic_check() (i.e. not dm_crtc_helper_atomic_check()). Also, async flipping isn't supported if a plane's framebuffer changes memory domains during an atomic commit. So, move the check from dm_crtc_helper_atomic_check() to amdgpu_dm_atomic_check() and check if the memory domain has changed in amdgpu_dm_atomic_check(). Cc: stable@vger.kernel.org Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2733 Fixes: c1e18c44 ("drm/amd/display: only accept async flips for fast updates") Reviewed-by:
Harry Wentland <harry.wentland@amd.com> Signed-off-by:
Hamza Mahfooz <hamza.mahfooz@amd.com> Signed-off-by:
-
Shashank Sharma authored
This patch: - adds a doorbell object in kfd pdd structure. - allocates doorbells for a process while creating its queue. - frees the doorbells with pdd destroy. - moves doorbell bitmap init function to kfd_doorbell.c PS: This patch ensures that we don't break the existing KFD functionality, but now KFD userspace library should also create doorbell pages as AMDGPU GEM objects using libdrm functions in userspace. The reference code for the same is available with AMDGPU Usermode queue libdrm MR. Once this is done, we will not need to create process doorbells in kernel. V2: - Do not use doorbell wrapper API, use amdgpu_bo_create_kernel instead (Alex). - Do not use custom doorbell structure, instead use separate variables for bo and doorbell_bitmap (Alex) V3: - Do not allocate doorbell page with PDD, delay doorbell process page allocation until really needed (Felix) Cc: Alex Deucher <alexander.deucher@amd.com> Cc: Christian Koenig <christian.koenig@amd.com> Cc: Felix Kuehling <Felix.Kuehling@amd.com> Acked-by:Felix Kuehling <Felilx.Kuehling@amd.com> Signed-off-by:
-
Shashank Sharma authored
This patch: - adds a doorbell bo in kfd device structure. - creates doorbell page for kfd kernel usages. - updates the get_kernel_doorbell and free_kernel_doorbell functions accordingly V2: Do not use wrapper API, use direct amdgpu_create_kernel(Alex) V3: - Move single variable declaration below (Christian) - Add a to-do item to reuse the KGD kernel level doorbells for KFD for non-MES cases, instead of reserving one page (Felix) Cc: Alex Deucher <alexander.deucher@amd.com> Cc: Christian Koenig <christian.koenig@amd.com> Cc: Felix Kuehling <Felix.Kuehling@amd.com> Reviewed-by:
-
Bhawanpreet Lakha authored
Read DP_SINK_PR_PIXEL_DEVIATION_PER_LINE and DP_SINK_PR_MAX_NUMBER_OF_DEVIATION_LINE Signed-off-by:
-
Bhawanpreet Lakha authored
We need certain conditions for replay to be enabled, so create an interface in DM to enable/disable replay. Signed-off-by:
-
Yang Wang authored
v1: Enable following message in vf mode. - PPSMC_MSG_GetMinGfxclkFreqquency - PPSMC_MSG_GetMaxGfxclkFreqquency - PPSMC_MSG_GetMinDpmFreq - PPSMC_MSG_GetMaxDpmFreq these message will cause pp_dpm_* device node not work properly. v2: the following message is disabled in VF mode. (since pmfw 85.69.0) - PPSMC_MSG_EnableAllSmuFeatures Signed-off-by:
Yang Wang <kevinyang.wang@amd.com> Acked-by:
Hawking Zhang <Hawking.Zhang@amd.com> Reviewed-by:
Lijo Lazar <lijo.lazar@amd.com> Signed-off-by:
-
Ori Messinger authored
Added missing MES firmware versions to the 'fw_version' sysfs directory, they should now exist as a files named "mes_fw_version" and "mes_kiq_fw_version" found at: /sys/class/drm/cardX/device/fw_version/mes_fw_version /sys/class/drm/cardX/device/fw_version/mes_kiq_fw_version Where X is the card number, and the version is displayed in hexadecimal. Signed-off-by:
Ori Messinger <ori.messinger@amd.com> Reviewed-by:
Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com> Signed-off-by:
-
Shashank Sharma authored
This patch adds a helper function which converts a doorbell's relative index in a BO to an absolute doorbell offset in the doorbell BAR. V2: No space between the variable name doc (Luben) Cc: Alex Deucher <alexander.deucher@amd.com> Cc: Christian Koenig <christian.koenig@amd.com> Acked-by:
-
Lijo Lazar authored
Energy counter should be reported in units of 15.259 uJ. Don't apply any conversion. Signed-off-by:
-
Shashank Sharma authored
This patch: - creates a doorbell page for graphics driver usages. - adds a few new varlables in adev->doorbell structure to keep track of kernel's doorbell-bo. - removes the adev->doorbell.ptr variable, replaces it with kernel-doorbell-bo's cpu address. V2: - Create doorbell BO directly, no wrappe functions (Alex) - no additional doorbell structure (Alex, Christian) - Use doorbell_cpu_ptr, remove ioremap (Christian, Alex) - Allocate one extra page of doorbells for MES (Alex) V4: Move MES doorbell base init into MES related patch (Christian) Cc: Alex Deucher <alexander.deucher@amd.com> Cc: Christian Koenig <christian.koenig@amd.com> Reviewed-by: -
Lijo Lazar authored
Make the new ascis to follow nbio callback method to get pcie replay count. Signed-off-by:
-
Yang Wang authored
- fix variable ('attr') dereferenced issue. - using condition check instead of BUG_ON(). Fixes: 4e01847c ("drm/amdgpu: optimize amdgpu device attribute code") Cc: Dan Carpenter <dan.carpenter@linaro.org> Signed-off-by:Kenneth Feng <kenneth.feng@amd.com> Signed-off-by:
-
Lijo Lazar authored
Add implementation to get pcie replay count for nbio v7.9. Signed-off-by:
-
Shashank Sharma authored
This patch initialzes the ttm resource manager for doorbells. V2: Do not round up doorbell size (Alex) Cc: Alex Deucher <alexander.deucher@amd.com> Cc: Christian Koenig <christian.koenig@amd.com> Reviewed-by:
-
Alex Deucher authored
This patch adds changes: - to accommodate the new GEM domain for DOORBELLs - to accommodate the new TTM PL for DOORBELLs in order to manage doorbell pages as GEM object. V2: Addressed reviwe comments from Christian - drop the doorbell changes for pinning/unpinning - drop the doorbell changes for dma-buf map - drop the doorbell changes for sgt - no need to handle TTM_PL_FLAG_CONTIGUOUS for doorbell - add caching type for doorbell V3: - Removed unrelated empty line (Christian) - Add PL_DOORBELL in mem_type_to_domain() as well (Alex) Cc: Alex Deucher <alexander.deucher@amd.com> Cc: Christian Koenig <christian.koenig@amd.com> Reviewed-by: -
Jay Cornwall authored
SMEM instructions can reach addresses above 47 bits but require bit 47 to be sign-extended through bits [63:48]. This allows the TMA to be relocated in a following patch. Signed-off-by:
Jay Cornwall <jay.cornwall@amd.com> Reviewed-by:
-
