Le 19/10/2025 à 11:16, Federico Amedeo Izzo via B4 Relay a écrit :
From: Federico Amedeo Izzo <federico-3FKkmSgw/[email protected]>
Add support for DSPP GC block in DPU driver for Qualcomm SoCs.
Expose the GAMMA_LUT DRM property, which is needed to enable
night light and basic screen color calibration.
I used LineageOS downstream kernel as a reference and found the LUT
format by trial-and-error on OnePlus 6.
Tested on oneplus-enchilada (sdm845-mainline 6.16-dev) and xiaomi-tissot
(msm8953-mainline 6.12/main).
Tested-by: David Heidelberg <[email protected]> # Pixel 3
(next-20251018)
Tested-by: Guido Günther <[email protected]> # on
sdm845-shift-axolotl
Signed-off-by: Federico Amedeo Izzo <federico-3FKkmSgw/[email protected]>
---
DRM GAMMA_LUT support was missing on sdm845 and other Qualcomm SoCs using
DPU for CRTC. This is needed in userspace to enable features like Night
Light or basic color calibration.
I wrote this driver to enable Night Light on OnePlus 6, and after the
driver was working I found out it applies to the 29 different Qualcomm SoCs
that use the DPU display engine, including X1E for laptops.
I used the LineageOS downstream kernel as reference and found the correct
LUT format by trial-and-error on OnePlus 6.
This was my first Linux driver and it's been a great learning
experience.
The patch was reviewed by postmarketOS contributors here:
https://gitlab.com/sdm845-mainline/linux/-/merge_requests/137
During review the patch was tested successfully on hamoa (X1E).
---
Hi,
...
@@ -830,19 +862,40 @@ static void _dpu_crtc_setup_cp_blocks(struct drm_crtc
*crtc)
ctl = mixer[i].lm_ctl;
dspp = mixer[i].hw_dspp;
- if (!dspp || !dspp->ops.setup_pcc)
+ if (!dspp)
continue;
- if (!state->ctm) {
- dspp->ops.setup_pcc(dspp, NULL);
- } else {
- _dpu_crtc_get_pcc_coeff(state, &cfg);
- dspp->ops.setup_pcc(dspp, &cfg);
+ if (dspp->ops.setup_pcc) {
+ if (!state->ctm) {
+ dspp->ops.setup_pcc(dspp, NULL);
+ } else {
+ _dpu_crtc_get_pcc_coeff(state, &cfg);
+ dspp->ops.setup_pcc(dspp, &cfg);
+ }
+
+ /* stage config flush mask */
+ ctl->ops.update_pending_flush_dspp(ctl,
+ mixer[i].hw_dspp->idx, DPU_DSPP_PCC);
}
- /* stage config flush mask */
- ctl->ops.update_pending_flush_dspp(ctl,
- mixer[i].hw_dspp->idx, DPU_DSPP_PCC);
+ if (dspp->ops.setup_gc) {
+ if (!state->gamma_lut) {
+ dspp->ops.setup_gc(dspp, NULL);
+ } else {
+ gc_lut = kzalloc(sizeof(*gc_lut), GFP_KERNEL);
The memory is already cleared in _dpu_crtc_get_gc_lut().
Eiher this could be changed into kmalloc(), or the memset in
_dpu_crtc_get_gc_lut() could be removed.
+ if (!gc_lut) {
+ DRM_ERROR("failed to allocate
gc_lut\n");
usually,message related to memory allocation errors are not needed,
because things are already verbose in such a case.
+ continue;
+ }
+ _dpu_crtc_get_gc_lut(state, gc_lut);
+ dspp->ops.setup_gc(dspp, gc_lut);
+ kfree(gc_lut);
+ }
+
+ /* stage config flush mask */
+ ctl->ops.update_pending_flush_dspp(ctl,
+ mixer[i].hw_dspp->idx, DPU_DSPP_GC);
+ }
}
}
...
diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_hw_dspp.c
b/drivers/gpu/drm/msm/disp/dpu1/dpu_hw_dspp.c
index 54b20faa0b69..7ebe7d8a5382 100644
--- a/drivers/gpu/drm/msm/disp/dpu1/dpu_hw_dspp.c
+++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_hw_dspp.c
...
@@ -63,6 +75,47 @@ static void dpu_setup_dspp_pcc(struct dpu_hw_dspp *ctx,
DPU_REG_WRITE(&ctx->hw, base, PCC_EN);
}
+static void dpu_setup_dspp_gc(struct dpu_hw_dspp *ctx,
+ struct dpu_hw_gc_lut *gc_lut)
+{
+ int i = 0;
+ u32 base, reg;
+
+ if (!ctx) {
+ DRM_ERROR("invalid ctx %pK\n", ctx);
ctx is known to be NULL here. So the message can be simplified.
+ return;
+ }
+
+ base = ctx->cap->sblk->gc.base;
+
+ if (!base) {
+ DRM_ERROR("invalid ctx %pK gc base 0x%x\n", ctx, base);
base is known to be NULL here. So the message can be simplified.
+ return;
+ }
+
+ if (!gc_lut) {
+ DRM_DEBUG_DRIVER("disable gc feature\n");
+ DPU_REG_WRITE(&ctx->hw, base, GC_DIS);
+ return;
+ }
+
+ reg = 0;
+ DPU_REG_WRITE(&ctx->hw, base + GC_C0_INDEX_OFF, reg);
+ DPU_REG_WRITE(&ctx->hw, base + GC_C1_INDEX_OFF, reg);
+ DPU_REG_WRITE(&ctx->hw, base + GC_C2_INDEX_OFF, reg);
Why not using 0 explicitly, instead of using reg here?
+
+ for (i = 0; i < PGC_TBL_LEN; i++) {
+ DPU_REG_WRITE(&ctx->hw, base + GC_C0_OFF, gc_lut->c0[i]);
+ DPU_REG_WRITE(&ctx->hw, base + GC_C1_OFF, gc_lut->c1[i]);
+ DPU_REG_WRITE(&ctx->hw, base + GC_C2_OFF, gc_lut->c2[i]);
+ }
+
+ DPU_REG_WRITE(&ctx->hw, base + GC_LUT_SWAP_OFF, BIT(0));
+
+ reg = GC_EN | ((gc_lut->flags & PGC_8B_ROUND) ? GC_8B_ROUND_EN : 0);
+ DPU_REG_WRITE(&ctx->hw, base, reg);
+}
...
CJ
