wireguard-linux/drivers/gpu/drm/amd/display/amdgpu_dm, branch stable

drm/amd/display: make DC_RUN_WITH_PREEMPTION_ENABLED misuse a build error

2026-08-12T13:45:03Z

Inside an FPU compilation unit DC_FP_START() and DC_FP_END() are defined as BUILD_BUG(), so using them there fails the build. That was done on purpose by commit a574f53ed52e ("drm/amd/display: Permit DC_FP_START/END only in non-FP compilation units"). DC_RUN_WITH_PREEMPTION_ENABLED() was added later by commit 3539437f354b ("drm/amd/display: Move FPU Guards From DML To DC - Part 1") and defined as a plain pass-through in that same branch instead. A wrap placed inside an FPU compilation unit therefore compiles cleanly, reads as correct during review, and does nothing at all. This is not hypothetical. While chasing a "scheduling while atomic" splat in dc_create_plane_state() on PREEMPT_RT, an attempt to place the guard further up the call chain, in dml21_add_phantom_plane() in dc/dml2_0/dml21/dml21_utils.c, had no effect for exactly this reason: dc/dml2_0/Makefile applies CC_FLAGS_FPU to every object under that directory, and the top level Makefile adds -D_LINUX_FPU_COMPILATION_UNIT to CC_FLAGS_FPU. Define the macro as BUILD_BUG() there as well, so that the mistake is a compile error rather than a guard that silently does nothing. The code argument is kept in the expansion so the BUILD_BUG() failure is not accompanied by set-but-unused diagnostics for variables assigned inside it. No current user is affected. dc/core/dc_stream.c and dc/resource/dcn32/dcn32_resource.c are outside the dml directories, and dc/dml2_0/dml2_wrapper.c and dc/dml2_0/dml21/dml21_wrapper.c are built without the FPU flags because dc/dml2_0/Makefile replaces their CFLAGS with CC_FLAGS_NO_FPU and removes CC_FLAGS_FPU. Link: https://lore.kernel.org/all/1ead313022bc62dce1f42af9f855727eb9074443.camel@web.de/ Signed-off-by: Mikhail Gavrilov Reviewed-by: Tom Chung Signed-off-by: Alex Deucher

drm/amd/display: Fix NULL pointer dereference in amdgpu_dm_crtc_set_vblank()

2026-08-12T13:38:11Z

amdgpu_dm_crtc_set_vblank() dereferences acrtc_state->stream when vblank is enabled/queried from DRM_IOCTL_MODE_CRTC_GET_SEQUENCE before a stream is attached to it. BUG: kernel NULL pointer dereference, address: 0000000000000008 RIP: amdgpu_dm_crtc_set_vblank+0x6b/0x4d0 [amdgpu] Call Trace: drm_vblank_enable drm_vblank_get drm_crtc_get_sequence_ioctl drm_ioctl_kernel drm_ioctl Reproduced by running VKCTS with WSI tests enabled on RADV. Guard the enable path on acrtc_state->stream being non-NULL, matching the existing checks in this function. Fixes: 34d66bc7ff10 ("drm/amd/display: Fix Xorg desktop unresponsive on Replay panel") Reviewed-by: Melissa Wen Signed-off-by: Samuel Pitoiset Signed-off-by: Alex Deucher

drm/amd/display: Set native cursor mode for disabled CRTCs again

2026-08-06T17:45:44Z

It was regressed by a refactor. Let's fix it again. Always set native cursor mode when the CRTC is disabled, to make sure it doesn't cause atomic commits to fail when they are trying to disable the CRTC. Fixes: 87a6dab12fc0 ("drm/amd/display: add cursor module") Cc: Alex Hung Reviewed-by: Ivan Lipski Signed-off-by: Timur Kristóf Signed-off-by: Alex Deucher

drm/amd/display: allow self-refresh exit while entry is blocked

2026-08-06T16:48:00Z

amdgpu_dm_crtc_set_static_screen_optimze() maps sso_enable to the Replay and PSR1 vsync events. allow_sr_entry is an entry gate, but the helper currently applies it to both directions. A non-fast update clears allow_sr_entry. During a modeset, a separate hardware-programming event keeps self-refresh blocked while the stream is reprogrammed. If vblank is enabled before the entry delay expires, the ISM calls the helper with sso_enable false. The early return drops the disable request, so the vsync events are not set. After enough fast commits, allow_sr_entry becomes true and the hardware-programming event can be cleared. Since the vblank reference remains held, there is no further zero-to-one vblank transition to restore the missing vsync events. Replay or PSR1 can then become active while vblank is still enabled. Gate only requests that enable static-screen optimization. Always process disable requests so a vblank requestor keeps Replay and PSR1 blocked. On a Phoenix system, repeated SDDM-to-VT handoffs produced stuck flips followed by flip_done and commit-wait timeouts. The timeout was not observed with this change applied. Fixes: 3c108046e1d6 ("drm/amd/display: Add power module on Linux") Assisted-by: Codex:gpt-5.6-sol Assisted-by: Claude:opus-5 Signed-off-by: David Weber Reviewed-by: Leo Li Signed-off-by: Alex Deucher

drm/amd: Disable DP audio spread spectrum for Cyan Skillfish

2026-08-06T16:47:56Z

The VBIOS for Cyan Skillfish devices (DCN201) indicates there is DisplayPort ref clock spread spectrum downspread, so the audio clock is corrected for it. However, the clock source in this hardware does not seem to actually be running with a clock downspread, so DisplayPort audio desyncs with video after several minutes. Ignore dprefclk SS downspread on CYAN_SKILLFISH2 asic. Closes: https://gitlab.freedesktop.org/drm/amd/-/work_items/5429 Signed-off-by: Travis K. Bangs Signed-off-by: Alex Deucher

drm/amd/display: Fix more KUnit connector use-after-free bugs

2026-08-06T16:44:10Z

drmm_connector_init() and drmm_encoder_init() register their cleanup (drm_connector_cleanup() / drm_encoder_cleanup()) as DRM-managed actions tied to the drm_device lifetime. When the object memory is owned by KUnit, it is freed before that action runs, so the cleanup touches freed memory. Allocate these objects with drmm_kzalloc() so their lifetime matches the cleanup action. Signed-off-by: Alex Hung Assisted-by: Copilot:Claude-Opus-4.8 Acked-by: Roman Li Tested-by: Dan Wheeler Signed-off-by: Alex Deucher

drm/amd/display: Configure all CRC engines in pipe CRC source path

2026-08-06T16:44:05Z

Iterate every CRC engine instance when enabling or disabling the pipe CRC source so DCN31x OTG_CRC1_EN is handled alongside OTG_CRC_EN, matching the legacy shared enable behavior used by DisplayCRC multi-ROI. Reviewed-by: ChiaHsuan (Tom) Chung Signed-off-by: Wayne Lin Signed-off-by: Roman Li Tested-by: Dan Wheeler Signed-off-by: Alex Deucher

drm/amd/display: Cover crtc destroy_state stream release

2026-08-06T16:43:28Z

Add dm_test_crtc_destroy_state_releases_stream to cover the cur->stream branch of amdgpu_dm_crtc_destroy_state(), complementing the existing no-stream test. The test attaches a DC stream to the CRTC state and takes an extra stream reference so the destroy path drops back to the KUnit-managed reference instead of freeing the stream, then verifies exactly one reference was released. Assisted-by: Copilot:Claude-Opus-4.8 Reviewed-by: Alex Hung Signed-off-by: Bhawanpreet Lakha Signed-off-by: Roman Li Tested-by: Dan Wheeler Signed-off-by: Alex Deucher

drm/amd/display: Cover crtc vblank restore replay-supported path

2026-08-06T16:43:24Z

Add dm_test_crtc_enable_vblank_ips_restore_replay to cover the pr->config.replay_supported side of the sr_supported OR in amdgpu_dm_crtc_set_vblank(). The existing IPS restore test establishes self-refresh support via the PSR version. This test instead marks the PSR version unsupported and sets replay_supported, forcing the sr_supported computation to fall through to the replay branch while still calling drm_crtc_vblank_restore(). Assisted-by: Copilot:Claude-Opus-4.8 Reviewed-by: Alex Hung Signed-off-by: Bhawanpreet Lakha Signed-off-by: Roman Li Tested-by: Dan Wheeler Signed-off-by: Alex Deucher

drm/amd/display: Cover crtc vblank IPS self-refresh restore

2026-08-06T16:43:14Z

Add dm_test_crtc_enable_vblank_ips_restore to cover the IPS/self-refresh branch of amdgpu_dm_crtc_set_vblank() that calls drm_crtc_vblank_restore(). The test primes the DC with ips_support set and IPS not fully disabled, a supported PSR version (self-refresh supported) and an immediate-disable vblank config, so all four conditions gating the restore hold. A stub get_vblank_timestamp hook is installed on the CRTC so the restore helper passes its sanity check, and the enable path then runs to completion. Assisted-by: Copilot:Claude-Opus-4.8 Reviewed-by: Alex Hung Signed-off-by: Bhawanpreet Lakha Signed-off-by: Roman Li Tested-by: Dan Wheeler Signed-off-by: Alex Deucher