[PATCH 1/2] drm/amdgpu/gfx10: add wraparound gpu counter check for APUs as well

[PATCH 1/2] drm/amdgpu/gfx10: add wraparound gpu counter check for APUs as well

[Alex Deucher]

Apply the same check we do for dGPUs for APUs as well.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 15 +++++++++++++--
 1 file changed, 13 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index e7dfeb466a0e..dbe7442fb25c 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
@@ -7707,8 +7707,19 @@ static uint64_t gfx_v10_0_get_gpu_clock_counter(struct amdgpu_device *adev)
 	switch (adev->ip_versions[GC_HWIP][0]) {
 	case IP_VERSION(10, 3, 1):
 	case IP_VERSION(10, 3, 3):
-		clock = (uint64_t)RREG32_SOC15(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Vangogh) |
-			((uint64_t)RREG32_SOC15(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Vangogh) << 32ULL);
+		preempt_disable();
+		clock_hi = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Vangogh);
+		clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Vangogh);
+		hi_check = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Vangogh);
+		/* The SMUIO TSC clock frequency is 100MHz, which sets 32-bit carry over
+		 * roughly every 42 seconds.
+		 */
+		if (hi_check != clock_hi) {
+			clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Vangogh);
+			clock_hi = hi_check;
+		}
+		preempt_enable();
+		clock = clock_lo | (clock_hi << 32ULL);
 		break;
 	default:
 		preempt_disable();
-- 
2.31.1

[PATCH 2/2] drm/amdgpu/gfx9: switch to golden tsc registers for renoir+

[Alex Deucher]

Renoir and newer gfx9 APUs have new TSC register that is
not part of the gfxoff tile, so it can be read without
needing to disable gfx off.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 46 ++++++++++++++++++++-------
 1 file changed, 35 insertions(+), 11 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index 2d0bc1c91426..154fa6facf19 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
@@ -140,6 +140,11 @@ MODULE_FIRMWARE("amdgpu/aldebaran_rlc.bin");
 #define mmTCP_CHAN_STEER_5_ARCT								0x0b0c
 #define mmTCP_CHAN_STEER_5_ARCT_BASE_IDX							0
 
+#define mmGOLDEN_TSC_COUNT_UPPER_Renoir                0x0025
+#define mmGOLDEN_TSC_COUNT_UPPER_Renoir_BASE_IDX       1
+#define mmGOLDEN_TSC_COUNT_LOWER_Renoir                0x0026
+#define mmGOLDEN_TSC_COUNT_LOWER_Renoir_BASE_IDX       1
+
 enum ta_ras_gfx_subblock {
 	/*CPC*/
 	TA_RAS_BLOCK__GFX_CPC_INDEX_START = 0,
@@ -4240,19 +4245,38 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)
 
 static uint64_t gfx_v9_0_get_gpu_clock_counter(struct amdgpu_device *adev)
 {
-	uint64_t clock;
+	uint64_t clock, clock_lo, clock_hi, hi_check;
 
-	amdgpu_gfx_off_ctrl(adev, false);
-	mutex_lock(&adev->gfx.gpu_clock_mutex);
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 0, 1) && amdgpu_sriov_runtime(adev)) {
-		clock = gfx_v9_0_kiq_read_clock(adev);
-	} else {
-		WREG32_SOC15(GC, 0, mmRLC_CAPTURE_GPU_CLOCK_COUNT, 1);
-		clock = (uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_LSB) |
-			((uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_MSB) << 32ULL);
+	switch (adev->ip_versions[GC_HWIP][0]) {
+	case IP_VERSION(9, 3, 0):
+		preempt_disable();
+		clock_hi = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Renoir);
+		clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Renoir);
+		hi_check = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Renoir);
+		/* The SMUIO TSC clock frequency is 100MHz, which sets 32-bit carry over
+		 * roughly every 42 seconds.
+		 */
+		if (hi_check != clock_hi) {
+			clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Renoir);
+			clock_hi = hi_check;
+		}
+		preempt_enable();
+		clock = clock_lo | (clock_hi << 32ULL);
+		break;
+	default:
+		amdgpu_gfx_off_ctrl(adev, false);
+		mutex_lock(&adev->gfx.gpu_clock_mutex);
+		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 0, 1) && amdgpu_sriov_runtime(adev)) {
+			clock = gfx_v9_0_kiq_read_clock(adev);
+		} else {
+			WREG32_SOC15(GC, 0, mmRLC_CAPTURE_GPU_CLOCK_COUNT, 1);
+			clock = (uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_LSB) |
+				((uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_MSB) << 32ULL);
+		}
+		mutex_unlock(&adev->gfx.gpu_clock_mutex);
+		amdgpu_gfx_off_ctrl(adev, true);
+		break;
 	}
-	mutex_unlock(&adev->gfx.gpu_clock_mutex);
-	amdgpu_gfx_off_ctrl(adev, true);
 	return clock;
 }
 
-- 
2.31.1

Re: [PATCH 1/2] drm/amdgpu/gfx10: add wraparound gpu counter check for APUs as well

[Luben Tuikov]

Seems reasonable.

Series is Acked-by: Luben Tuikov <luben.tuikov@amd.com>

Regards,
Luben

On 2021-11-18 14:54, Alex Deucher wrote:
> Apply the same check we do for dGPUs for APUs as well.
>
> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 15 +++++++++++++--
>  1 file changed, 13 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
> index e7dfeb466a0e..dbe7442fb25c 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
> @@ -7707,8 +7707,19 @@ static uint64_t gfx_v10_0_get_gpu_clock_counter(struct amdgpu_device *adev)
>  	switch (adev->ip_versions[GC_HWIP][0]) {
>  	case IP_VERSION(10, 3, 1):
>  	case IP_VERSION(10, 3, 3):
> -		clock = (uint64_t)RREG32_SOC15(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Vangogh) |
> -			((uint64_t)RREG32_SOC15(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Vangogh) << 32ULL);
> +		preempt_disable();
> +		clock_hi = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Vangogh);
> +		clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Vangogh);
> +		hi_check = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Vangogh);
> +		/* The SMUIO TSC clock frequency is 100MHz, which sets 32-bit carry over
> +		 * roughly every 42 seconds.
> +		 */
> +		if (hi_check != clock_hi) {
> +			clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Vangogh);
> +			clock_hi = hi_check;
> +		}
> +		preempt_enable();
> +		clock = clock_lo | (clock_hi << 32ULL);
>  		break;
>  	default:
>  		preempt_disable();