[PATCH v3] drm/msm/a6xx: Add support for Adreno 612

[PATCH v3] drm/msm/a6xx: Add support for Adreno 612

[Akhil P Oommen]

From: Jie Zhang <quic_jiezh@quicinc.com>

Add support for Adreno 612 GPU found in SM6150/QCS615 chipsets.
A612 falls under ADRENO_6XX_GEN1 family and is a cut down version
of A615 GPU.

A612 has a new IP called Reduced Graphics Management Unit or RGMU
which is a small state machine which helps to toggle GX GDSC
(connected to CX rail) to implement IFPC feature. It doesn't support
any other features of a full fledged GMU like clock control, resource
voting to rpmh etc. So we need linux clock driver support like other
gmu-wrapper implementations to control gpu core clock and gpu GX gdsc.
Since there is no benefit with enabling RGMU at the moment, RGMU is
entirely skipped in this patch.

Signed-off-by: Jie Zhang <quic_jiezh@quicinc.com>
Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
Reviewed-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
---
Mesa support is already available for A612. Verified Glmark2 with
weston.

Some dependencies for the devicetree change are not yet available
in the mailing lists. I will send it out as a separate patch later.
---
Changes in v3:
- Drop the NO_SYSCACHE quirk patch (Konrad/Rob)
- Use the new "qcom,adreno-rgmu" compatible string (Konrad)
- Link to v2: https://lore.kernel.org/r/20241125-a612-gpu-support-v2-0-b7cc38e60191@quicinc.com

Changes in v2:
- Added a new quirk to check LLC support (new patch). This helps to
correct LLC handling in A612's patch.
- Rebased on msm-next tip
- Captured R-b from Konrad
- Link to v1: https://lore.kernel.org/r/20241101-a612-gpu-support-v1-1-bdfe8f6d9306@quicinc.com
---
 drivers/gpu/drm/msm/adreno/a6xx_catalog.c | 15 ++++++++++
 drivers/gpu/drm/msm/adreno/a6xx_gpu.c     | 47 ++++++++++++++++++++++---------
 drivers/gpu/drm/msm/adreno/adreno_gpu.h   | 11 ++++++--
 3 files changed, 57 insertions(+), 16 deletions(-)

diff --git a/drivers/gpu/drm/msm/adreno/a6xx_catalog.c b/drivers/gpu/drm/msm/adreno/a6xx_catalog.c
index 0c560e84ad5a..234083b69844 100644
--- a/drivers/gpu/drm/msm/adreno/a6xx_catalog.c
+++ b/drivers/gpu/drm/msm/adreno/a6xx_catalog.c
@@ -704,6 +704,21 @@ static const struct adreno_info a6xx_gpus[] = {
 			{ 157, 3 },
 			{ 127, 4 },
 		),
+	}, {
+		.chip_ids = ADRENO_CHIP_IDS(0x06010200),
+		.family = ADRENO_6XX_GEN1,
+		.fw = {
+			[ADRENO_FW_SQE] = "a630_sqe.fw",
+		},
+		.gmem = (SZ_128K + SZ_4K),
+		.inactive_period = DRM_MSM_INACTIVE_PERIOD,
+		.init = a6xx_gpu_init,
+		.a6xx = &(const struct a6xx_info) {
+			.hwcg = a612_hwcg,
+			.protect = &a630_protect,
+			.gmu_cgc_mode = 0x00000022,
+			.prim_fifo_threshold = 0x00080000,
+		},
 	}, {
 		.chip_ids = ADRENO_CHIP_IDS(0x06010500),
 		.family = ADRENO_6XX_GEN1,
diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
index 019610341df1..630932693951 100644
--- a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
@@ -504,15 +504,26 @@ static void a6xx_set_hwcg(struct msm_gpu *gpu, bool state)
 
 	if (adreno_is_a630(adreno_gpu))
 		clock_cntl_on = 0x8aa8aa02;
-	else if (adreno_is_a610(adreno_gpu))
+	else if (adreno_is_a610(adreno_gpu) || adreno_is_a612(adreno_gpu))
 		clock_cntl_on = 0xaaa8aa82;
 	else if (adreno_is_a702(adreno_gpu))
 		clock_cntl_on = 0xaaaaaa82;
 	else
 		clock_cntl_on = 0x8aa8aa82;
 
-	cgc_delay = adreno_is_a615_family(adreno_gpu) ? 0x111 : 0x10111;
-	cgc_hyst = adreno_is_a615_family(adreno_gpu) ? 0x555 : 0x5555;
+	if (adreno_is_a612(adreno_gpu))
+		cgc_delay = 0x11;
+	else if (adreno_is_a615_family(adreno_gpu))
+		cgc_delay = 0x111;
+	else
+		cgc_delay = 0x10111;
+
+	if (adreno_is_a612(adreno_gpu))
+		cgc_hyst = 0x55;
+	else if (adreno_is_a615_family(adreno_gpu))
+		cgc_delay = 0x555;
+	else
+		cgc_delay = 0x5555;
 
 	gmu_write(&a6xx_gpu->gmu, REG_A6XX_GPU_GMU_AO_GMU_CGC_MODE_CNTL,
 			state ? adreno_gpu->info->a6xx->gmu_cgc_mode : 0);
@@ -600,6 +611,9 @@ static void a6xx_calc_ubwc_config(struct adreno_gpu *gpu)
 		gpu->ubwc_config.ubwc_swizzle = 0x7;
 	}
 
+	if (adreno_is_a612(gpu))
+		gpu->ubwc_config.highest_bank_bit = 13;
+
 	if (adreno_is_a618(gpu))
 		gpu->ubwc_config.highest_bank_bit = 14;
 
@@ -1165,7 +1179,7 @@ static int hw_init(struct msm_gpu *gpu)
 		gpu_write(gpu, REG_A6XX_CP_LPAC_PROG_FIFO_SIZE, 0x00000020);
 
 	/* Setting the mem pool size */
-	if (adreno_is_a610(adreno_gpu)) {
+	if (adreno_is_a610(adreno_gpu) || adreno_is_a612(adreno_gpu)) {
 		gpu_write(gpu, REG_A6XX_CP_MEM_POOL_SIZE, 48);
 		gpu_write(gpu, REG_A6XX_CP_MEM_POOL_DBG_ADDR, 47);
 	} else if (adreno_is_a702(adreno_gpu)) {
@@ -1199,7 +1213,7 @@ static int hw_init(struct msm_gpu *gpu)
 
 	/* Enable fault detection */
 	if (adreno_is_a730(adreno_gpu) ||
-	    adreno_is_a740_family(adreno_gpu))
+	    adreno_is_a740_family(adreno_gpu) || adreno_is_a612(adreno_gpu))
 		gpu_write(gpu, REG_A6XX_RBBM_INTERFACE_HANG_INT_CNTL, (1 << 30) | 0xcfffff);
 	else if (adreno_is_a690(adreno_gpu))
 		gpu_write(gpu, REG_A6XX_RBBM_INTERFACE_HANG_INT_CNTL, (1 << 30) | 0x4fffff);
@@ -1863,8 +1877,8 @@ static void a7xx_llc_activate(struct a6xx_gpu *a6xx_gpu)
 
 static void a6xx_llc_slices_destroy(struct a6xx_gpu *a6xx_gpu)
 {
-	/* No LLCC on non-RPMh (and by extension, non-GMU) SoCs */
-	if (adreno_has_gmu_wrapper(&a6xx_gpu->base))
+	/* A612 is actually not a gmu-wrapper and has LLCC */
+	if (adreno_has_gmu_wrapper(&a6xx_gpu->base) && !adreno_is_a612(&a6xx_gpu->base))
 		return;
 
 	llcc_slice_putd(a6xx_gpu->llc_slice);
@@ -1876,8 +1890,8 @@ static void a6xx_llc_slices_init(struct platform_device *pdev,
 {
 	struct device_node *phandle;
 
-	/* No LLCC on non-RPMh (and by extension, non-GMU) SoCs */
-	if (adreno_has_gmu_wrapper(&a6xx_gpu->base))
+	/* A612 is actually not a gmu-wrapper and has LLCC */
+	if (adreno_has_gmu_wrapper(&a6xx_gpu->base) && !adreno_is_a612(&a6xx_gpu->base))
 		return;
 
 	/*
@@ -2081,6 +2095,9 @@ static int a6xx_pm_resume(struct msm_gpu *gpu)
 	if (!ret)
 		msm_devfreq_resume(gpu);
 
+	if (adreno_is_a612(&a6xx_gpu->base))
+		a6xx_llc_activate(a6xx_gpu);
+
 	return ret;
 }
 
@@ -2120,6 +2137,9 @@ static int a6xx_pm_suspend(struct msm_gpu *gpu)
 
 	trace_msm_gpu_suspend(0);
 
+	if (adreno_is_a612(&a6xx_gpu->base))
+		a6xx_llc_deactivate(a6xx_gpu);
+
 	msm_devfreq_suspend(gpu);
 
 	mutex_lock(&a6xx_gpu->gmu.lock);
@@ -2475,7 +2495,9 @@ struct msm_gpu *a6xx_gpu_init(struct drm_device *dev)
 	/* FIXME: How do we gracefully handle this? */
 	BUG_ON(!node);
 
-	adreno_gpu->gmu_is_wrapper = of_device_is_compatible(node, "qcom,adreno-gmu-wrapper");
+	/* We do not support RGMU at the moment, so assume it is a gmu wrapper for now */
+	adreno_gpu->gmu_is_wrapper = of_device_is_compatible(node, "qcom,adreno-gmu-wrapper") ||
+		of_device_is_compatible(node, "qcom,adreno-rgmu");
 
 	adreno_gpu->base.hw_apriv =
 		!!(config->info->quirks & ADRENO_QUIRK_HAS_HW_APRIV);
@@ -2485,11 +2507,8 @@ struct msm_gpu *a6xx_gpu_init(struct drm_device *dev)
 		  config->info->family == ADRENO_7XX_GEN2 ||
 		  config->info->family == ADRENO_7XX_GEN3;
 
-	a6xx_llc_slices_init(pdev, a6xx_gpu, is_a7xx);
-
 	ret = a6xx_set_supported_hw(&pdev->dev, config->info);
 	if (ret) {
-		a6xx_llc_slices_destroy(a6xx_gpu);
 		kfree(a6xx_gpu);
 		return ERR_PTR(ret);
 	}
@@ -2508,6 +2527,8 @@ struct msm_gpu *a6xx_gpu_init(struct drm_device *dev)
 		return ERR_PTR(ret);
 	}
 
+	a6xx_llc_slices_init(pdev, a6xx_gpu, is_a7xx);
+
 	/*
 	 * For now only clamp to idle freq for devices where this is known not
 	 * to cause power supply issues:
diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.h b/drivers/gpu/drm/msm/adreno/adreno_gpu.h
index e71f420f8b3a..61b9141a63ab 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_gpu.h
+++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.h
@@ -420,6 +420,11 @@ static inline int adreno_is_a610(const struct adreno_gpu *gpu)
 	return adreno_is_revn(gpu, 610);
 }
 
+static inline int adreno_is_a612(const struct adreno_gpu *gpu)
+{
+	return gpu->info->chip_ids[0] == 0x06010200;
+}
+
 static inline int adreno_is_a618(const struct adreno_gpu *gpu)
 {
 	return adreno_is_revn(gpu, 618);
@@ -489,9 +494,9 @@ static inline int adreno_is_a610_family(const struct adreno_gpu *gpu)
 {
 	if (WARN_ON_ONCE(!gpu->info))
 		return false;
-
-	/* TODO: A612 */
-	return adreno_is_a610(gpu) || adreno_is_a702(gpu);
+	return adreno_is_a610(gpu) ||
+	       adreno_is_a612(gpu) ||
+	       adreno_is_a702(gpu);
 }
 
 /* TODO: 615/616 */

---
base-commit: f4a867a46862c1743501bbe8c813238456ec8699
change-id: 20241031-a612-gpu-support-d6330f17d01f

Best regards,
-- 
Akhil P Oommen <quic_akhilpo@quicinc.com>

Re: [PATCH v3] drm/msm/a6xx: Add support for Adreno 612

[Konrad Dybcio]

On 13.12.2024 12:46 PM, Akhil P Oommen wrote:
> From: Jie Zhang <quic_jiezh@quicinc.com>
> 
> Add support for Adreno 612 GPU found in SM6150/QCS615 chipsets.
> A612 falls under ADRENO_6XX_GEN1 family and is a cut down version
> of A615 GPU.
> 
> A612 has a new IP called Reduced Graphics Management Unit or RGMU
> which is a small state machine which helps to toggle GX GDSC
> (connected to CX rail) to implement IFPC feature. It doesn't support
> any other features of a full fledged GMU like clock control, resource
> voting to rpmh etc. So we need linux clock driver support like other
> gmu-wrapper implementations to control gpu core clock and gpu GX gdsc.
> Since there is no benefit with enabling RGMU at the moment, RGMU is
> entirely skipped in this patch.
> 
> Signed-off-by: Jie Zhang <quic_jiezh@quicinc.com>
> Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
> Reviewed-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
> ---

So we talked offline a bit, and the RGMU requires a piece of firmware.

We concluded it's best to describe that from the get-go, so that the
user doesn't get surprised when a new kernel update brings new firmware
requirements for previously-working hardware.

Please wait for the new revision.

Konrad

Re: [PATCH v3] drm/msm/a6xx: Add support for Adreno 612

[Dmitry Baryshkov]

On Fri, Dec 20, 2024 at 08:56:31PM +0100, Konrad Dybcio wrote:
> On 13.12.2024 12:46 PM, Akhil P Oommen wrote:
> > From: Jie Zhang <quic_jiezh@quicinc.com>
> > 
> > Add support for Adreno 612 GPU found in SM6150/QCS615 chipsets.
> > A612 falls under ADRENO_6XX_GEN1 family and is a cut down version
> > of A615 GPU.
> > 
> > A612 has a new IP called Reduced Graphics Management Unit or RGMU
> > which is a small state machine which helps to toggle GX GDSC
> > (connected to CX rail) to implement IFPC feature. It doesn't support
> > any other features of a full fledged GMU like clock control, resource
> > voting to rpmh etc. So we need linux clock driver support like other
> > gmu-wrapper implementations to control gpu core clock and gpu GX gdsc.
> > Since there is no benefit with enabling RGMU at the moment, RGMU is
> > entirely skipped in this patch.
> > 
> > Signed-off-by: Jie Zhang <quic_jiezh@quicinc.com>
> > Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
> > Reviewed-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
> > ---
> 
> So we talked offline a bit, and the RGMU requires a piece of firmware.
> 
> We concluded it's best to describe that from the get-go, so that the
> user doesn't get surprised when a new kernel update brings new firmware
> requirements for previously-working hardware.

I'd say, please make sure that the RGMU firmware is also a part of the
linux-firmware from the beginning.

> 
> Please wait for the new revision.
> 
> Konrad

-- 
With best wishes
Dmitry

Re: [PATCH v3] drm/msm/a6xx: Add support for Adreno 612

[Akhil P Oommen]

On 12/21/2024 2:28 AM, Dmitry Baryshkov wrote:
> On Fri, Dec 20, 2024 at 08:56:31PM +0100, Konrad Dybcio wrote:
>> On 13.12.2024 12:46 PM, Akhil P Oommen wrote:
>>> From: Jie Zhang <quic_jiezh@quicinc.com>
>>>
>>> Add support for Adreno 612 GPU found in SM6150/QCS615 chipsets.
>>> A612 falls under ADRENO_6XX_GEN1 family and is a cut down version
>>> of A615 GPU.
>>>
>>> A612 has a new IP called Reduced Graphics Management Unit or RGMU
>>> which is a small state machine which helps to toggle GX GDSC
>>> (connected to CX rail) to implement IFPC feature. It doesn't support
>>> any other features of a full fledged GMU like clock control, resource
>>> voting to rpmh etc. So we need linux clock driver support like other
>>> gmu-wrapper implementations to control gpu core clock and gpu GX gdsc.
>>> Since there is no benefit with enabling RGMU at the moment, RGMU is
>>> entirely skipped in this patch.
>>>
>>> Signed-off-by: Jie Zhang <quic_jiezh@quicinc.com>
>>> Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
>>> Reviewed-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
>>> ---
>>
>> So we talked offline a bit, and the RGMU requires a piece of firmware.
>>
>> We concluded it's best to describe that from the get-go, so that the
>> user doesn't get surprised when a new kernel update brings new firmware
>> requirements for previously-working hardware.
> 
> I'd say, please make sure that the RGMU firmware is also a part of the
> linux-firmware from the beginning.

Yes, I am working on that part. I will send another revision of this patch.

-Akhil

> 
>>
>> Please wait for the new revision.
>>
>> Konrad
>