[PATCH] drm/amd/display: Wrap dcn32_override_min_req_memclk() in DC_FP_{START, END}

[PATCH] drm/amd/display: Wrap dcn32_override_min_req_memclk() in DC_FP_{START, END}

[Xi Ruoyao]

[Why]
The dcn32_override_min_req_memclk function is in dcn32_fpu.c, which is
compiled with CC_FLAGS_FPU into FP instructions.  So when we call it we
must use DC_FP_{START,END} to save and restore the FP context, and
prepare the FP unit on architectures like LoongArch where the FP unit
isn't always on.

Reported-by: LiarOnce <liaronce@hotmail.com>
Fixes: ee7be8f3de1c ("drm/amd/display: Limit DCN32 8 channel or less parts to DPM1 for FPO")
Cc: stable@vger.kernel.org
Signed-off-by: Xi Ruoyao <xry111@xry111.site>
---
 drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c
index 7ebb7d1193af..c7fd604024d6 100644
--- a/drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c
+++ b/drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c
@@ -1785,7 +1785,10 @@ static bool dml1_validate(struct dc *dc, struct dc_state *context, enum dc_valid
 
 	dc->res_pool->funcs->calculate_wm_and_dlg(dc, context, pipes, pipe_cnt, vlevel);
 
+	DC_FP_START();
 	dcn32_override_min_req_memclk(dc, context);
+	DC_FP_END();
+
 	dcn32_override_min_req_dcfclk(dc, context);
 
 	BW_VAL_TRACE_END_WATERMARKS();
-- 
2.53.0

Claude review: drm/amd/display: Wrap dcn32_override_min_req_memclk() in DC_FP_{START, END}

[Claude Code Review Bot]

Overall Series Review

Subject: drm/amd/display: Wrap dcn32_override_min_req_memclk() in DC_FP_{START, END}
Author: Xi Ruoyao <xry111@xry111.site>
Patches: 1
Reviewed: 2026-03-09T09:08:52.939278

---

This is a single-patch series that fixes a missing FP context save/restore around a call to `dcn32_override_min_req_memclk()`. The function lives in `dcn32_fpu.c`, which is compiled with `CC_FLAGS_FPU` (enabling FP instructions). On architectures like LoongArch where the FPU isn't always enabled, calling into FPU-compiled code without the proper `DC_FP_START()`/`DC_FP_END()` guards will cause faults or incorrect behavior.

The fix is correct, minimal, and follows the established pattern already used throughout the same file (e.g., the `dcn32_internal_validate_bw()` call at lines 1770-1771 uses the same `DC_FP_START/END` wrapping). The `Fixes:` tag and `Cc: stable` are appropriate since this is a real bug that affects users on LoongArch.

**Verdict: Good to merge.**

---
Generated by Claude Code Patch Reviewer

Claude review: drm/amd/display: Wrap dcn32_override_min_req_memclk() in DC_FP_{START, END}

[Claude Code Review Bot]

Patch Review

**Correctness: Correct.**

The added guards:
```c
DC_FP_START();
dcn32_override_min_req_memclk(dc, context);
DC_FP_END();
```

are necessary because `dcn32_override_min_req_memclk()` is defined in `dcn32_fpu.c` (confirmed at `dcn32_fpu.c:3600`), which is compiled with `CC_FLAGS_FPU` per `dc/dml/Makefile:28,65`. The sibling function `dcn32_override_min_req_dcfclk()` at line 1792 does **not** need wrapping because it lives in `dcn32_resource_helpers.c` (a non-FPU file), which is correct.

**Observations:**

1. The commit message is clear and well-structured with the `[Why]` format used by the amdgpu display team.
2. The `Fixes:` tag correctly references `ee7be8f3de1c` which introduced the unguarded call.
3. The `Cc: stable@vger.kernel.org` is appropriate for a bug fix that causes real failures on LoongArch hardware.
4. The pattern is consistent with how `dcn32_internal_validate_bw()` is already wrapped at lines 1770-1771 in the same function.

**No issues found.**

---
Generated by Claude Code Patch Reviewer