[PATCH 142/207] drm/amdgpu: add start DPG mode for VCN3.0
Alex Deucher
alexdeucher at gmail.com
Mon Jun 1 18:19:49 UTC 2020
From: Boyuan Zhang <boyuan.zhang at amd.com>
Add vcn_v3_0_start_dpg_mode to setup and start VCN block in DPG mode for VCN3.0
V2: Separate from previous patch-0002, and update description.
Signed-off-by: Boyuan Zhang <boyuan.zhang at amd.com>
Reviewed-by: James Zhu <james.zhu at amd.com>
Reviewed-by: Alex Deucher <alexander.deucher at amd.com>
Signed-off-by: Alex Deucher <alexander.deucher at amd.com>
---
drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c | 141 ++++++++++++++++++++++++++
1 file changed, 141 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
index 86db36537371..5578af5cb399 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
@@ -818,6 +818,142 @@ static void vcn_v3_0_enable_clock_gating(struct amdgpu_device *adev, int inst)
WREG32_SOC15(VCN, inst, mmUVD_SUVD_CGC_CTRL, data);
}
+static int vcn_v3_0_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, bool indirect)
+{
+ struct amdgpu_ring *ring;
+ uint32_t rb_bufsz, tmp;
+
+ /* disable register anti-hang mechanism */
+ WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, mmUVD_POWER_STATUS), 1,
+ ~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
+ /* enable dynamic power gating mode */
+ tmp = RREG32_SOC15(VCN, inst_idx, mmUVD_POWER_STATUS);
+ tmp |= UVD_POWER_STATUS__UVD_PG_MODE_MASK;
+ tmp |= UVD_POWER_STATUS__UVD_PG_EN_MASK;
+ WREG32_SOC15(VCN, inst_idx, mmUVD_POWER_STATUS, tmp);
+
+ if (indirect)
+ adev->vcn.inst[inst_idx].dpg_sram_curr_addr = (uint32_t*)adev->vcn.inst[inst_idx].dpg_sram_cpu_addr;
+
+ /* enable clock gating */
+ vcn_v3_0_clock_gating_dpg_mode(adev, 0, inst_idx, indirect);
+
+ /* enable VCPU clock */
+ tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
+ tmp |= UVD_VCPU_CNTL__CLK_EN_MASK;
+ tmp |= UVD_VCPU_CNTL__BLK_RST_MASK;
+ WREG32_SOC15_DPG_MODE_2_0(inst_idx, SOC15_DPG_MODE_OFFSET_2_0(
+ VCN, inst_idx, mmUVD_VCPU_CNTL), tmp, 0, indirect);
+
+ /* disable master interupt */
+ WREG32_SOC15_DPG_MODE_2_0(inst_idx, SOC15_DPG_MODE_OFFSET_2_0(
+ VCN, inst_idx, mmUVD_MASTINT_EN), 0, 0, indirect);
+
+ /* setup mmUVD_LMI_CTRL */
+ tmp = (0x8 | UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
+ UVD_LMI_CTRL__REQ_MODE_MASK |
+ UVD_LMI_CTRL__CRC_RESET_MASK |
+ UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
+ UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
+ UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK |
+ (8 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) |
+ 0x00100000L);
+ WREG32_SOC15_DPG_MODE_2_0(inst_idx, SOC15_DPG_MODE_OFFSET_2_0(
+ VCN, inst_idx, mmUVD_LMI_CTRL), tmp, 0, indirect);
+
+ WREG32_SOC15_DPG_MODE_2_0(inst_idx, SOC15_DPG_MODE_OFFSET_2_0(
+ VCN, inst_idx, mmUVD_MPC_CNTL),
+ 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT, 0, indirect);
+
+ WREG32_SOC15_DPG_MODE_2_0(inst_idx, SOC15_DPG_MODE_OFFSET_2_0(
+ VCN, inst_idx, mmUVD_MPC_SET_MUXA0),
+ ((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) |
+ (0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) |
+ (0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) |
+ (0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)), 0, indirect);
+
+ WREG32_SOC15_DPG_MODE_2_0(inst_idx, SOC15_DPG_MODE_OFFSET_2_0(
+ VCN, inst_idx, mmUVD_MPC_SET_MUXB0),
+ ((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) |
+ (0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) |
+ (0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) |
+ (0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)), 0, indirect);
+
+ WREG32_SOC15_DPG_MODE_2_0(inst_idx, SOC15_DPG_MODE_OFFSET_2_0(
+ VCN, inst_idx, mmUVD_MPC_SET_MUX),
+ ((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) |
+ (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
+ (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)), 0, indirect);
+
+ vcn_v3_0_mc_resume_dpg_mode(adev, inst_idx, indirect);
+
+ WREG32_SOC15_DPG_MODE_2_0(inst_idx, SOC15_DPG_MODE_OFFSET_2_0(
+ VCN, inst_idx, mmUVD_REG_XX_MASK), 0x10, 0, indirect);
+ WREG32_SOC15_DPG_MODE_2_0(inst_idx, SOC15_DPG_MODE_OFFSET_2_0(
+ VCN, inst_idx, mmUVD_RBC_XX_IB_REG_CHECK), 0x3, 0, indirect);
+
+ /* enable LMI MC and UMC channels */
+ WREG32_SOC15_DPG_MODE_2_0(inst_idx, SOC15_DPG_MODE_OFFSET_2_0(
+ VCN, inst_idx, mmUVD_LMI_CTRL2), 0, 0, indirect);
+
+ /* unblock VCPU register access */
+ WREG32_SOC15_DPG_MODE_2_0(inst_idx, SOC15_DPG_MODE_OFFSET_2_0(
+ VCN, inst_idx, mmUVD_RB_ARB_CTRL), 0, 0, indirect);
+
+ tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
+ tmp |= UVD_VCPU_CNTL__CLK_EN_MASK;
+ WREG32_SOC15_DPG_MODE_2_0(inst_idx, SOC15_DPG_MODE_OFFSET_2_0(
+ VCN, inst_idx, mmUVD_VCPU_CNTL), tmp, 0, indirect);
+
+ /* enable master interrupt */
+ WREG32_SOC15_DPG_MODE_2_0(inst_idx, SOC15_DPG_MODE_OFFSET_2_0(
+ VCN, inst_idx, mmUVD_MASTINT_EN),
+ UVD_MASTINT_EN__VCPU_EN_MASK, 0, indirect);
+
+ /* add nop to workaround PSP size check */
+ WREG32_SOC15_DPG_MODE_2_0(inst_idx, SOC15_DPG_MODE_OFFSET_2_0(
+ VCN, inst_idx, mmUVD_VCPU_CNTL), tmp, 0, indirect);
+
+ if (indirect)
+ psp_update_vcn_sram(adev, inst_idx, adev->vcn.inst[inst_idx].dpg_sram_gpu_addr,
+ (uint32_t)((uintptr_t)adev->vcn.inst[inst_idx].dpg_sram_curr_addr -
+ (uintptr_t)adev->vcn.inst[inst_idx].dpg_sram_cpu_addr));
+
+ ring = &adev->vcn.inst[inst_idx].ring_dec;
+ /* force RBC into idle state */
+ rb_bufsz = order_base_2(ring->ring_size);
+ tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz);
+ tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1);
+ tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1);
+ tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1);
+ tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1);
+ WREG32_SOC15(VCN, inst_idx, mmUVD_RBC_RB_CNTL, tmp);
+
+ /* set the write pointer delay */
+ WREG32_SOC15(VCN, inst_idx, mmUVD_RBC_RB_WPTR_CNTL, 0);
+
+ /* set the wb address */
+ WREG32_SOC15(VCN, inst_idx, mmUVD_RBC_RB_RPTR_ADDR,
+ (upper_32_bits(ring->gpu_addr) >> 2));
+
+ /* programm the RB_BASE for ring buffer */
+ WREG32_SOC15(VCN, inst_idx, mmUVD_LMI_RBC_RB_64BIT_BAR_LOW,
+ lower_32_bits(ring->gpu_addr));
+ WREG32_SOC15(VCN, inst_idx, mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH,
+ upper_32_bits(ring->gpu_addr));
+
+ /* Initialize the ring buffer's read and write pointers */
+ WREG32_SOC15(VCN, inst_idx, mmUVD_RBC_RB_RPTR, 0);
+
+ WREG32_SOC15(VCN, inst_idx, mmUVD_SCRATCH2, 0);
+
+ ring->wptr = RREG32_SOC15(VCN, inst_idx, mmUVD_RBC_RB_RPTR);
+ WREG32_SOC15(VCN, inst_idx, mmUVD_RBC_RB_WPTR,
+ lower_32_bits(ring->wptr));
+
+ return 0;
+}
+
static int vcn_v3_0_start(struct amdgpu_device *adev)
{
struct amdgpu_ring *ring;
@@ -831,6 +967,11 @@ static int vcn_v3_0_start(struct amdgpu_device *adev)
if (adev->vcn.harvest_config & (1 << i))
continue;
+ if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG){
+ r = vcn_v3_0_start_dpg_mode(adev, i, adev->vcn.indirect_sram);
+ continue;
+ }
+
/* disable VCN power gating */
vcn_v3_0_disable_static_power_gating(adev, i);
--
2.25.4
More information about the amd-gfx
mailing list