Skip to content

L
linux
Commit fd91d38b authored by Lijo Lazar's avatar Lijo Lazar Committed by Alex Deucher
Browse files
Options

drm/amdgpu: Use logical ids for VCN/JPEG v4.0.3 

Address VCN/JPEG instances using logical ids. Whenever register access is
required, get the physical instance using GET_INST.
Signed-off-by: default avatarLijo Lazar <lijo.lazar@amd.com>
Acked-by: default avatarLeo Liu <leo.liu@amd.com>
Tested-by: default avatarJames Zhu <James.Zhu@amd.com>
Signed-off-by: default avatarAlex Deucher <alexander.deucher@amd.com>
parent 07bc0ac8
Show whitespace changes
Inline Side-by-side
Showing with 408 additions and 317 deletions
drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
View file @ fd91d38b
...@@ -144,14 +144,19 @@ ...@@ -144,14 +144,19 @@
#define WREG32_SOC15_DPG_MODE(inst_idx, offset, value, mask_en, indirect) \ #define WREG32_SOC15_DPG_MODE(inst_idx, offset, value, mask_en, indirect) \
do { \ do { \
if (!indirect) { \ if (!indirect) { \
WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_DATA, value); \ WREG32_SOC15(VCN, GET_INST(VCN, inst_idx), \
WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_CTL, \ mmUVD_DPG_LMA_DATA, value); \
WREG32_SOC15( \
VCN, GET_INST(VCN, inst_idx), \
mmUVD_DPG_LMA_CTL, \
(0x1 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT | \ (0x1 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT | \
mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT | \ mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT | \
offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT)); \ offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT)); \
} else { \ } else { \
*adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ = offset; \ *adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ = \
*adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ = value; \ offset; \
*adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ = \
value; \
} \ } \
} while (0) } while (0)
......
drivers/gpu/drm/amd/amdgpu/aqua_vanjaram_reg_init.c
View file @ fd91d38b
...@@ -70,6 +70,8 @@ static int8_t aqua_vanjaram_logical_to_dev_inst(struct amdgpu_device *adev, ...@@ -70,6 +70,8 @@ static int8_t aqua_vanjaram_logical_to_dev_inst(struct amdgpu_device *adev,
switch (block) { switch (block) {
case GC_HWIP: case GC_HWIP:
case SDMA0_HWIP: case SDMA0_HWIP:
/* Both JPEG and VCN as JPEG is only alias of VCN */
case VCN_HWIP:
dev_inst = adev->ip_map.dev_inst[block][inst]; dev_inst = adev->ip_map.dev_inst[block][inst];
break; break;
default: default:
...@@ -379,7 +381,7 @@ static int aqua_vanjaram_xcp_mgr_init(struct amdgpu_device *adev) ...@@ -379,7 +381,7 @@ static int aqua_vanjaram_xcp_mgr_init(struct amdgpu_device *adev)
int aqua_vanjaram_init_soc_config(struct amdgpu_device *adev) int aqua_vanjaram_init_soc_config(struct amdgpu_device *adev)
{ {
u32 mask, inst_mask = adev->sdma.sdma_mask; u32 mask, inst_mask = adev->sdma.sdma_mask;
int ret, i, num_inst; int ret, i;
/* generally 1 AID supports 4 instances */ /* generally 1 AID supports 4 instances */
adev->sdma.num_inst_per_aid = 4; adev->sdma.num_inst_per_aid = 4;
...@@ -394,11 +396,15 @@ int aqua_vanjaram_init_soc_config(struct amdgpu_device *adev) ...@@ -394,11 +396,15 @@ int aqua_vanjaram_init_soc_config(struct amdgpu_device *adev)
adev->aid_mask |= (1 << i); adev->aid_mask |= (1 << i);
} }
num_inst = hweight32(adev->aid_mask); /* Harvest config is not used for aqua vanjaram. VCN and JPEGs will be
* addressed based on logical instance ids.
*/
adev->vcn.harvest_config = 0;
adev->vcn.num_inst_per_aid = 1; adev->vcn.num_inst_per_aid = 1;
adev->vcn.num_vcn_inst = adev->vcn.num_inst_per_aid * num_inst; adev->vcn.num_vcn_inst = hweight32(adev->vcn.inst_mask);
adev->jpeg.harvest_config = 0;
adev->jpeg.num_inst_per_aid = 1; adev->jpeg.num_inst_per_aid = 1;
adev->jpeg.num_jpeg_inst = adev->jpeg.num_inst_per_aid * num_inst; adev->jpeg.num_jpeg_inst = hweight32(adev->jpeg.inst_mask);
ret = aqua_vanjaram_xcp_mgr_init(adev); ret = aqua_vanjaram_xcp_mgr_init(adev);
if (ret) if (ret)
......
drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c
View file @ fd91d38b
...@@ -82,7 +82,7 @@ static int jpeg_v4_0_3_sw_init(void *handle) ...@@ -82,7 +82,7 @@ static int jpeg_v4_0_3_sw_init(void *handle)
{ {
struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct amdgpu_device *adev = (struct amdgpu_device *)handle;
struct amdgpu_ring *ring; struct amdgpu_ring *ring;
int i, j, r; int i, j, r, jpeg_inst;
for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) { for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
/* JPEG TRAP */ /* JPEG TRAP */
...@@ -101,14 +101,15 @@ static int jpeg_v4_0_3_sw_init(void *handle) ...@@ -101,14 +101,15 @@ static int jpeg_v4_0_3_sw_init(void *handle)
return r; return r;
for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) { for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
if (adev->jpeg.harvest_config & (1 << i)) jpeg_inst = GET_INST(JPEG, i);
continue;
for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) { for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
ring = &adev->jpeg.inst[i].ring_dec[j]; ring = &adev->jpeg.inst[i].ring_dec[j];
ring->use_doorbell = true; ring->use_doorbell = true;
ring->vm_hub = AMDGPU_MMHUB0(adev->jpeg.inst[i].aid_id); ring->vm_hub = AMDGPU_MMHUB0(adev->jpeg.inst[i].aid_id);
ring->doorbell_index = ring->doorbell_index =
(adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1 + j + 9 * i; (adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
1 + j + 9 * jpeg_inst;
sprintf(ring->name, "jpeg_dec_%d.%d", i, j); sprintf(ring->name, "jpeg_dec_%d.%d", i, j);
r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq, 0, r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq, 0,
AMDGPU_RING_PRIO_DEFAULT, NULL); AMDGPU_RING_PRIO_DEFAULT, NULL);
...@@ -118,8 +119,10 @@ static int jpeg_v4_0_3_sw_init(void *handle) ...@@ -118,8 +119,10 @@ static int jpeg_v4_0_3_sw_init(void *handle)
adev->jpeg.internal.jpeg_pitch[j] = adev->jpeg.internal.jpeg_pitch[j] =
regUVD_JRBC0_UVD_JRBC_SCRATCH0_INTERNAL_OFFSET; regUVD_JRBC0_UVD_JRBC_SCRATCH0_INTERNAL_OFFSET;
adev->jpeg.inst[i].external.jpeg_pitch[j] = adev->jpeg.inst[i].external.jpeg_pitch[j] =
SOC15_REG_OFFSET1(JPEG, i, regUVD_JRBC0_UVD_JRBC_SCRATCH0, SOC15_REG_OFFSET1(
(j?(0x40 * j - 0xc80):0)); JPEG, jpeg_inst,
regUVD_JRBC0_UVD_JRBC_SCRATCH0,
(j ? (0x40 * j - 0xc80) : 0));
} }
} }
...@@ -157,24 +160,29 @@ static int jpeg_v4_0_3_hw_init(void *handle) ...@@ -157,24 +160,29 @@ static int jpeg_v4_0_3_hw_init(void *handle)
{ {
struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct amdgpu_device *adev = (struct amdgpu_device *)handle;
struct amdgpu_ring *ring; struct amdgpu_ring *ring;
int i, j, r; int i, j, r, jpeg_inst;
for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) { for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
if (adev->jpeg.harvest_config & (1 << i)) jpeg_inst = GET_INST(JPEG, i);
continue;
ring = adev->jpeg.inst[i].ring_dec; ring = adev->jpeg.inst[i].ring_dec;
if (ring->use_doorbell) if (ring->use_doorbell)
adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell, adev->nbio.funcs->vcn_doorbell_range(
(adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 9 * i, adev, ring->use_doorbell,
(adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
9 * jpeg_inst,
adev->jpeg.inst[i].aid_id); adev->jpeg.inst[i].aid_id);
for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) { for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
ring = &adev->jpeg.inst[i].ring_dec[j]; ring = &adev->jpeg.inst[i].ring_dec[j];
if (ring->use_doorbell) if (ring->use_doorbell)
WREG32_SOC15_OFFSET(VCN, i, regVCN_JPEG_DB_CTRL, WREG32_SOC15_OFFSET(
(ring->pipe?(ring->pipe - 0x15):0), VCN, GET_INST(VCN, i),
ring->doorbell_index << VCN_JPEG_DB_CTRL__OFFSET__SHIFT | regVCN_JPEG_DB_CTRL,
(ring->pipe ? (ring->pipe - 0x15) : 0),
ring->doorbell_index
<< VCN_JPEG_DB_CTRL__OFFSET__SHIFT |
VCN_JPEG_DB_CTRL__EN_MASK); VCN_JPEG_DB_CTRL__EN_MASK);
r = amdgpu_ring_test_helper(ring); r = amdgpu_ring_test_helper(ring);
if (r) if (r)
...@@ -250,10 +258,11 @@ static int jpeg_v4_0_3_resume(void *handle) ...@@ -250,10 +258,11 @@ static int jpeg_v4_0_3_resume(void *handle)
static void jpeg_v4_0_3_disable_clock_gating(struct amdgpu_device *adev, int inst_idx) static void jpeg_v4_0_3_disable_clock_gating(struct amdgpu_device *adev, int inst_idx)
{ {
int i, jpeg_inst;
uint32_t data; uint32_t data;
int i;
data = RREG32_SOC15(JPEG, inst_idx, regJPEG_CGC_CTRL); jpeg_inst = GET_INST(JPEG, inst_idx);
data = RREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_CTRL);
if (adev->cg_flags & AMD_CG_SUPPORT_JPEG_MGCG) { if (adev->cg_flags & AMD_CG_SUPPORT_JPEG_MGCG) {
data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
data &= (~(JPEG_CGC_CTRL__JPEG0_DEC_MODE_MASK << 1)); data &= (~(JPEG_CGC_CTRL__JPEG0_DEC_MODE_MASK << 1));
...@@ -263,21 +272,22 @@ static void jpeg_v4_0_3_disable_clock_gating(struct amdgpu_device *adev, int ins ...@@ -263,21 +272,22 @@ static void jpeg_v4_0_3_disable_clock_gating(struct amdgpu_device *adev, int ins
data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT; data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT; data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
WREG32_SOC15(JPEG, inst_idx, regJPEG_CGC_CTRL, data); WREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_CTRL, data);
data = RREG32_SOC15(JPEG, inst_idx, regJPEG_CGC_GATE); data = RREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_GATE);
data &= ~(JPEG_CGC_GATE__JMCIF_MASK | JPEG_CGC_GATE__JRBBM_MASK); data &= ~(JPEG_CGC_GATE__JMCIF_MASK | JPEG_CGC_GATE__JRBBM_MASK);
for (i = 0; i < adev->jpeg.num_jpeg_rings; ++i) for (i = 0; i < adev->jpeg.num_jpeg_rings; ++i)
data &= ~(JPEG_CGC_GATE__JPEG0_DEC_MASK << i); data &= ~(JPEG_CGC_GATE__JPEG0_DEC_MASK << i);
WREG32_SOC15(JPEG, inst_idx, regJPEG_CGC_GATE, data); WREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_GATE, data);
} }
static void jpeg_v4_0_3_enable_clock_gating(struct amdgpu_device *adev, int inst_idx) static void jpeg_v4_0_3_enable_clock_gating(struct amdgpu_device *adev, int inst_idx)
{ {
int i, jpeg_inst;
uint32_t data; uint32_t data;
int i;
data = RREG32_SOC15(JPEG, inst_idx, regJPEG_CGC_CTRL); jpeg_inst = GET_INST(JPEG, inst_idx);
data = RREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_CTRL);
if (adev->cg_flags & AMD_CG_SUPPORT_JPEG_MGCG) { if (adev->cg_flags & AMD_CG_SUPPORT_JPEG_MGCG) {
data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
data |= (JPEG_CGC_CTRL__JPEG0_DEC_MODE_MASK << 1); data |= (JPEG_CGC_CTRL__JPEG0_DEC_MODE_MASK << 1);
...@@ -287,13 +297,13 @@ static void jpeg_v4_0_3_enable_clock_gating(struct amdgpu_device *adev, int inst ...@@ -287,13 +297,13 @@ static void jpeg_v4_0_3_enable_clock_gating(struct amdgpu_device *adev, int inst
data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT; data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT; data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
WREG32_SOC15(JPEG, inst_idx, regJPEG_CGC_CTRL, data); WREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_CTRL, data);
data = RREG32_SOC15(JPEG, inst_idx, regJPEG_CGC_GATE); data = RREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_GATE);
data |= (JPEG_CGC_GATE__JMCIF_MASK | JPEG_CGC_GATE__JRBBM_MASK); data |= (JPEG_CGC_GATE__JMCIF_MASK | JPEG_CGC_GATE__JRBBM_MASK);
for (i = 0; i < adev->jpeg.num_jpeg_rings; ++i) for (i = 0; i < adev->jpeg.num_jpeg_rings; ++i)
data |= (JPEG_CGC_GATE__JPEG0_DEC_MASK << i); data |= (JPEG_CGC_GATE__JPEG0_DEC_MASK << i);
WREG32_SOC15(JPEG, inst_idx, regJPEG_CGC_GATE, data); WREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_GATE, data);
} }
/** /**
...@@ -306,33 +316,35 @@ static void jpeg_v4_0_3_enable_clock_gating(struct amdgpu_device *adev, int inst ...@@ -306,33 +316,35 @@ static void jpeg_v4_0_3_enable_clock_gating(struct amdgpu_device *adev, int inst
static int jpeg_v4_0_3_start(struct amdgpu_device *adev) static int jpeg_v4_0_3_start(struct amdgpu_device *adev)
{ {
struct amdgpu_ring *ring; struct amdgpu_ring *ring;
int i, j; int i, j, jpeg_inst;
for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) { for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
if (adev->jpeg.harvest_config & (1 << i)) jpeg_inst = GET_INST(JPEG, i);
continue;
WREG32_SOC15(JPEG, i, regUVD_PGFSM_CONFIG, WREG32_SOC15(JPEG, jpeg_inst, regUVD_PGFSM_CONFIG,
1 << UVD_PGFSM_CONFIG__UVDJ_PWR_CONFIG__SHIFT); 1 << UVD_PGFSM_CONFIG__UVDJ_PWR_CONFIG__SHIFT);
SOC15_WAIT_ON_RREG(JPEG, i, regUVD_PGFSM_STATUS, SOC15_WAIT_ON_RREG(
UVD_PGFSM_STATUS__UVDJ_PWR_ON << JPEG, jpeg_inst, regUVD_PGFSM_STATUS,
UVD_PGFSM_STATUS__UVDJ_PWR_STATUS__SHIFT, UVD_PGFSM_STATUS__UVDJ_PWR_ON
<< UVD_PGFSM_STATUS__UVDJ_PWR_STATUS__SHIFT,
UVD_PGFSM_STATUS__UVDJ_PWR_STATUS_MASK); UVD_PGFSM_STATUS__UVDJ_PWR_STATUS_MASK);
/* disable anti hang mechanism */ /* disable anti hang mechanism */
WREG32_P(SOC15_REG_OFFSET(JPEG, i, regUVD_JPEG_POWER_STATUS), 0, WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst,
~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK); regUVD_JPEG_POWER_STATUS),
0, ~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK);
/* JPEG disable CGC */ /* JPEG disable CGC */
jpeg_v4_0_3_disable_clock_gating(adev, i); jpeg_v4_0_3_disable_clock_gating(adev, i);
/* MJPEG global tiling registers */ /* MJPEG global tiling registers */
WREG32_SOC15(JPEG, i, regJPEG_DEC_GFX8_ADDR_CONFIG, WREG32_SOC15(JPEG, jpeg_inst, regJPEG_DEC_GFX8_ADDR_CONFIG,
adev->gfx.config.gb_addr_config); adev->gfx.config.gb_addr_config);
WREG32_SOC15(JPEG, i, regJPEG_DEC_GFX10_ADDR_CONFIG, WREG32_SOC15(JPEG, jpeg_inst, regJPEG_DEC_GFX10_ADDR_CONFIG,
adev->gfx.config.gb_addr_config); adev->gfx.config.gb_addr_config);
/* enable JMI channel */ /* enable JMI channel */
WREG32_P(SOC15_REG_OFFSET(JPEG, i, regUVD_JMI_CNTL), 0, WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst, regUVD_JMI_CNTL), 0,
~UVD_JMI_CNTL__SOFT_RESET_MASK); ~UVD_JMI_CNTL__SOFT_RESET_MASK);
for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) { for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
...@@ -341,25 +353,40 @@ static int jpeg_v4_0_3_start(struct amdgpu_device *adev) ...@@ -341,25 +353,40 @@ static int jpeg_v4_0_3_start(struct amdgpu_device *adev)
ring = &adev->jpeg.inst[i].ring_dec[j]; ring = &adev->jpeg.inst[i].ring_dec[j];
/* enable System Interrupt for JRBC */ /* enable System Interrupt for JRBC */
WREG32_P(SOC15_REG_OFFSET(JPEG, i, regJPEG_SYS_INT_EN), WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst,
regJPEG_SYS_INT_EN),
JPEG_SYS_INT_EN__DJRBC0_MASK << j, JPEG_SYS_INT_EN__DJRBC0_MASK << j,
~(JPEG_SYS_INT_EN__DJRBC0_MASK << j)); ~(JPEG_SYS_INT_EN__DJRBC0_MASK << j));
WREG32_SOC15_OFFSET(JPEG, i, WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
regUVD_JMI0_UVD_LMI_JRBC_RB_VMID, reg_offset, 0); regUVD_JMI0_UVD_LMI_JRBC_RB_VMID,
WREG32_SOC15_OFFSET(JPEG, i, regUVD_JRBC0_UVD_JRBC_RB_CNTL, reg_offset, reg_offset, 0);
WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
regUVD_JRBC0_UVD_JRBC_RB_CNTL,
reg_offset,
(0x00000001L | 0x00000002L)); (0x00000001L | 0x00000002L));
WREG32_SOC15_OFFSET(JPEG, i, regUVD_JMI0_UVD_LMI_JRBC_RB_64BIT_BAR_LOW, WREG32_SOC15_OFFSET(
JPEG, jpeg_inst,
regUVD_JMI0_UVD_LMI_JRBC_RB_64BIT_BAR_LOW,
reg_offset, lower_32_bits(ring->gpu_addr)); reg_offset, lower_32_bits(ring->gpu_addr));
WREG32_SOC15_OFFSET(JPEG, i, regUVD_JMI0_UVD_LMI_JRBC_RB_64BIT_BAR_HIGH, WREG32_SOC15_OFFSET(
JPEG, jpeg_inst,
regUVD_JMI0_UVD_LMI_JRBC_RB_64BIT_BAR_HIGH,
reg_offset, upper_32_bits(ring->gpu_addr)); reg_offset, upper_32_bits(ring->gpu_addr));
WREG32_SOC15_OFFSET(JPEG, i, regUVD_JRBC0_UVD_JRBC_RB_RPTR, reg_offset, 0); WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
WREG32_SOC15_OFFSET(JPEG, i, regUVD_JRBC0_UVD_JRBC_RB_WPTR, reg_offset, 0); regUVD_JRBC0_UVD_JRBC_RB_RPTR,
WREG32_SOC15_OFFSET(JPEG, i, regUVD_JRBC0_UVD_JRBC_RB_CNTL, reg_offset, reg_offset, 0);
0x00000002L); WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
WREG32_SOC15_OFFSET(JPEG, i, regUVD_JRBC0_UVD_JRBC_RB_SIZE, reg_offset, regUVD_JRBC0_UVD_JRBC_RB_WPTR,
ring->ring_size / 4); reg_offset, 0);
ring->wptr = RREG32_SOC15_OFFSET(JPEG, i, regUVD_JRBC0_UVD_JRBC_RB_WPTR, WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
regUVD_JRBC0_UVD_JRBC_RB_CNTL,
reg_offset, 0x00000002L);
WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
regUVD_JRBC0_UVD_JRBC_RB_SIZE,
reg_offset, ring->ring_size / 4);
ring->wptr = RREG32_SOC15_OFFSET(
JPEG, jpeg_inst, regUVD_JRBC0_UVD_JRBC_RB_WPTR,
reg_offset); reg_offset);
} }
} }
...@@ -376,29 +403,29 @@ static int jpeg_v4_0_3_start(struct amdgpu_device *adev) ...@@ -376,29 +403,29 @@ static int jpeg_v4_0_3_start(struct amdgpu_device *adev)
*/ */
static int jpeg_v4_0_3_stop(struct amdgpu_device *adev) static int jpeg_v4_0_3_stop(struct amdgpu_device *adev)
{ {
int i; int i, jpeg_inst;
for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) { for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
if (adev->jpeg.harvest_config & (1 << i)) jpeg_inst = GET_INST(JPEG, i);
continue;
/* reset JMI */ /* reset JMI */
WREG32_P(SOC15_REG_OFFSET(JPEG, i, regUVD_JMI_CNTL), WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst, regUVD_JMI_CNTL),
UVD_JMI_CNTL__SOFT_RESET_MASK, UVD_JMI_CNTL__SOFT_RESET_MASK,
~UVD_JMI_CNTL__SOFT_RESET_MASK); ~UVD_JMI_CNTL__SOFT_RESET_MASK);
jpeg_v4_0_3_enable_clock_gating(adev, i); jpeg_v4_0_3_enable_clock_gating(adev, i);
/* enable anti hang mechanism */ /* enable anti hang mechanism */
WREG32_P(SOC15_REG_OFFSET(JPEG, i, regUVD_JPEG_POWER_STATUS), WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst,
regUVD_JPEG_POWER_STATUS),
UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK, UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK,
~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK); ~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK);
WREG32_SOC15(JPEG, i, regUVD_PGFSM_CONFIG, WREG32_SOC15(JPEG, jpeg_inst, regUVD_PGFSM_CONFIG,
2 << UVD_PGFSM_CONFIG__UVDJ_PWR_CONFIG__SHIFT); 2 << UVD_PGFSM_CONFIG__UVDJ_PWR_CONFIG__SHIFT);
SOC15_WAIT_ON_RREG(JPEG, i, regUVD_PGFSM_STATUS, SOC15_WAIT_ON_RREG(
UVD_PGFSM_STATUS__UVDJ_PWR_OFF << JPEG, jpeg_inst, regUVD_PGFSM_STATUS,
UVD_PGFSM_STATUS__UVDJ_PWR_STATUS__SHIFT, UVD_PGFSM_STATUS__UVDJ_PWR_OFF
<< UVD_PGFSM_STATUS__UVDJ_PWR_STATUS__SHIFT,
UVD_PGFSM_STATUS__UVDJ_PWR_STATUS_MASK); UVD_PGFSM_STATUS__UVDJ_PWR_STATUS_MASK);
} }
...@@ -416,8 +443,9 @@ static uint64_t jpeg_v4_0_3_dec_ring_get_rptr(struct amdgpu_ring *ring) ...@@ -416,8 +443,9 @@ static uint64_t jpeg_v4_0_3_dec_ring_get_rptr(struct amdgpu_ring *ring)
{ {
struct amdgpu_device *adev = ring->adev; struct amdgpu_device *adev = ring->adev;
return RREG32_SOC15_OFFSET(JPEG, ring->me, regUVD_JRBC0_UVD_JRBC_RB_RPTR, return RREG32_SOC15_OFFSET(
ring->pipe?(0x40 * ring->pipe - 0xc80):0); JPEG, GET_INST(JPEG, ring->me), regUVD_JRBC0_UVD_JRBC_RB_RPTR,
ring->pipe ? (0x40 * ring->pipe - 0xc80) : 0);
} }
/** /**
...@@ -434,8 +462,10 @@ static uint64_t jpeg_v4_0_3_dec_ring_get_wptr(struct amdgpu_ring *ring) ...@@ -434,8 +462,10 @@ static uint64_t jpeg_v4_0_3_dec_ring_get_wptr(struct amdgpu_ring *ring)
if (ring->use_doorbell) if (ring->use_doorbell)
return adev->wb.wb[ring->wptr_offs]; return adev->wb.wb[ring->wptr_offs];
else else
return RREG32_SOC15_OFFSET(JPEG, ring->me, regUVD_JRBC0_UVD_JRBC_RB_WPTR, return RREG32_SOC15_OFFSET(
ring->pipe?(0x40 * ring->pipe - 0xc80):0); JPEG, GET_INST(JPEG, ring->me),
regUVD_JRBC0_UVD_JRBC_RB_WPTR,
ring->pipe ? (0x40 * ring->pipe - 0xc80) : 0);
} }
/** /**
...@@ -453,8 +483,11 @@ static void jpeg_v4_0_3_dec_ring_set_wptr(struct amdgpu_ring *ring) ...@@ -453,8 +483,11 @@ static void jpeg_v4_0_3_dec_ring_set_wptr(struct amdgpu_ring *ring)
adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr);
WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr));
} else { } else {
WREG32_SOC15_OFFSET(JPEG, ring->me, regUVD_JRBC0_UVD_JRBC_RB_WPTR, WREG32_SOC15_OFFSET(JPEG, GET_INST(JPEG, ring->me),
(ring->pipe?(0x40 * ring->pipe - 0xc80):0), lower_32_bits(ring->wptr)); regUVD_JRBC0_UVD_JRBC_RB_WPTR,
(ring->pipe ? (0x40 * ring->pipe - 0xc80) :
0),
lower_32_bits(ring->wptr));
} }
} }
...@@ -703,13 +736,13 @@ static bool jpeg_v4_0_3_is_idle(void *handle) ...@@ -703,13 +736,13 @@ static bool jpeg_v4_0_3_is_idle(void *handle)
int i, j; int i, j;
for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) { for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
if (adev->jpeg.harvest_config & (1 << i))
continue;
for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) { for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
unsigned int reg_offset = (j?(0x40 * j - 0xc80):0); unsigned int reg_offset = (j?(0x40 * j - 0xc80):0);
ret &= ((RREG32_SOC15_OFFSET(JPEG, i, ret &= ((RREG32_SOC15_OFFSET(
regUVD_JRBC0_UVD_JRBC_STATUS, reg_offset) & JPEG, GET_INST(JPEG, i),
regUVD_JRBC0_UVD_JRBC_STATUS,
reg_offset) &
UVD_JRBC0_UVD_JRBC_STATUS__RB_JOB_DONE_MASK) == UVD_JRBC0_UVD_JRBC_STATUS__RB_JOB_DONE_MASK) ==
UVD_JRBC0_UVD_JRBC_STATUS__RB_JOB_DONE_MASK); UVD_JRBC0_UVD_JRBC_STATUS__RB_JOB_DONE_MASK);
} }
...@@ -725,12 +758,11 @@ static int jpeg_v4_0_3_wait_for_idle(void *handle) ...@@ -725,12 +758,11 @@ static int jpeg_v4_0_3_wait_for_idle(void *handle)
int i, j; int i, j;
for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) { for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
if (adev->jpeg.harvest_config & (1 << i))
continue;
for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) { for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
unsigned int reg_offset = (j?(0x40 * j - 0xc80):0); unsigned int reg_offset = (j?(0x40 * j - 0xc80):0);
ret &= SOC15_WAIT_ON_RREG_OFFSET(JPEG, i, ret &= SOC15_WAIT_ON_RREG_OFFSET(
JPEG, GET_INST(JPEG, i),
regUVD_JRBC0_UVD_JRBC_STATUS, reg_offset, regUVD_JRBC0_UVD_JRBC_STATUS, reg_offset,
UVD_JRBC0_UVD_JRBC_STATUS__RB_JOB_DONE_MASK, UVD_JRBC0_UVD_JRBC_STATUS__RB_JOB_DONE_MASK,
UVD_JRBC0_UVD_JRBC_STATUS__RB_JOB_DONE_MASK); UVD_JRBC0_UVD_JRBC_STATUS__RB_JOB_DONE_MASK);
...@@ -747,8 +779,6 @@ static int jpeg_v4_0_3_set_clockgating_state(void *handle, ...@@ -747,8 +779,6 @@ static int jpeg_v4_0_3_set_clockgating_state(void *handle,
int i; int i;
for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) { for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
if (adev->jpeg.harvest_config & (1 << i))
continue;
if (enable) { if (enable) {
if (!jpeg_v4_0_3_is_idle(handle)) if (!jpeg_v4_0_3_is_idle(handle))
return -EBUSY; return -EBUSY;
...@@ -792,35 +822,46 @@ static int jpeg_v4_0_3_process_interrupt(struct amdgpu_device *adev, ...@@ -792,35 +822,46 @@ static int jpeg_v4_0_3_process_interrupt(struct amdgpu_device *adev,
struct amdgpu_irq_src *source, struct amdgpu_irq_src *source,
struct amdgpu_iv_entry *entry) struct amdgpu_iv_entry *entry)
{ {
uint32_t i; uint32_t i, inst;
i = node_id_to_phys_map[entry->node_id]; i = node_id_to_phys_map[entry->node_id];
DRM_DEV_DEBUG(adev->dev, "IH: JPEG TRAP\n"); DRM_DEV_DEBUG(adev->dev, "IH: JPEG TRAP\n");
for (inst = 0; inst < adev->jpeg.num_jpeg_inst; ++inst)
if (adev->jpeg.inst[inst].aid_id == i)
break;
if (inst >= adev->jpeg.num_jpeg_inst) {
dev_WARN_ONCE(adev->dev, 1,
"Interrupt received for unknown JPEG instance %d",
entry->node_id);
return 0;
}
switch (entry->src_id) { switch (entry->src_id) {
case VCN_4_0__SRCID__JPEG_DECODE: case VCN_4_0__SRCID__JPEG_DECODE:
amdgpu_fence_process(&adev->jpeg.inst[i].ring_dec[0]); amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[0]);
break; break;
case VCN_4_0__SRCID__JPEG1_DECODE: case VCN_4_0__SRCID__JPEG1_DECODE:
amdgpu_fence_process(&adev->jpeg.inst[i].ring_dec[1]); amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[1]);
break; break;
case VCN_4_0__SRCID__JPEG2_DECODE: case VCN_4_0__SRCID__JPEG2_DECODE:
amdgpu_fence_process(&adev->jpeg.inst[i].ring_dec[2]); amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[2]);
break; break;
case VCN_4_0__SRCID__JPEG3_DECODE: case VCN_4_0__SRCID__JPEG3_DECODE:
amdgpu_fence_process(&adev->jpeg.inst[i].ring_dec[3]); amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[3]);
break; break;
case VCN_4_0__SRCID__JPEG4_DECODE: case VCN_4_0__SRCID__JPEG4_DECODE:
amdgpu_fence_process(&adev->jpeg.inst[i].ring_dec[4]); amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[4]);
break; break;
case VCN_4_0__SRCID__JPEG5_DECODE: case VCN_4_0__SRCID__JPEG5_DECODE:
amdgpu_fence_process(&adev->jpeg.inst[i].ring_dec[5]); amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[5]);
break; break;
case VCN_4_0__SRCID__JPEG6_DECODE: case VCN_4_0__SRCID__JPEG6_DECODE:
amdgpu_fence_process(&adev->jpeg.inst[i].ring_dec[6]); amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[6]);
break; break;
case VCN_4_0__SRCID__JPEG7_DECODE: case VCN_4_0__SRCID__JPEG7_DECODE:
amdgpu_fence_process(&adev->jpeg.inst[i].ring_dec[7]); amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[7]);
break; break;
default: default:
DRM_DEV_ERROR(adev->dev, "Unhandled interrupt: %d %d\n", DRM_DEV_ERROR(adev->dev, "Unhandled interrupt: %d %d\n",
...@@ -882,17 +923,17 @@ static const struct amdgpu_ring_funcs jpeg_v4_0_3_dec_ring_vm_funcs = { ...@@ -882,17 +923,17 @@ static const struct amdgpu_ring_funcs jpeg_v4_0_3_dec_ring_vm_funcs = {
static void jpeg_v4_0_3_set_dec_ring_funcs(struct amdgpu_device *adev) static void jpeg_v4_0_3_set_dec_ring_funcs(struct amdgpu_device *adev)
{ {
int i, j; int i, j, jpeg_inst;
for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) { for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
if (adev->jpeg.harvest_config & (1 << i))
continue;
for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) { for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
adev->jpeg.inst[i].ring_dec[j].funcs = &jpeg_v4_0_3_dec_ring_vm_funcs; adev->jpeg.inst[i].ring_dec[j].funcs = &jpeg_v4_0_3_dec_ring_vm_funcs;
adev->jpeg.inst[i].ring_dec[j].me = i; adev->jpeg.inst[i].ring_dec[j].me = i;
adev->jpeg.inst[i].ring_dec[j].pipe = j; adev->jpeg.inst[i].ring_dec[j].pipe = j;
} }
adev->jpeg.inst[i].aid_id = i / adev->jpeg.num_inst_per_aid; jpeg_inst = GET_INST(JPEG, i);
adev->jpeg.inst[i].aid_id =
jpeg_inst / adev->jpeg.num_inst_per_aid;
} }
DRM_DEV_INFO(adev->dev, "JPEG decode is enabled in VM mode\n"); DRM_DEV_INFO(adev->dev, "JPEG decode is enabled in VM mode\n");
} }
...@@ -907,8 +948,6 @@ static void jpeg_v4_0_3_set_irq_funcs(struct amdgpu_device *adev) ...@@ -907,8 +948,6 @@ static void jpeg_v4_0_3_set_irq_funcs(struct amdgpu_device *adev)
int i; int i;
for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) { for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
if (adev->jpeg.harvest_config & (1 << i))
continue;
adev->jpeg.inst->irq.num_types += adev->jpeg.num_jpeg_rings; adev->jpeg.inst->irq.num_types += adev->jpeg.num_jpeg_rings;
} }
adev->jpeg.inst->irq.funcs = &jpeg_v4_0_3_irq_funcs; adev->jpeg.inst->irq.funcs = &jpeg_v4_0_3_irq_funcs;
......
drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c
View file @ fd91d38b
...@@ -83,7 +83,7 @@ static int vcn_v4_0_3_sw_init(void *handle) ...@@ -83,7 +83,7 @@ static int vcn_v4_0_3_sw_init(void *handle)
{ {
struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct amdgpu_device *adev = (struct amdgpu_device *)handle;
struct amdgpu_ring *ring; struct amdgpu_ring *ring;
int i, r; int i, r, vcn_inst;
r = amdgpu_vcn_sw_init(adev); r = amdgpu_vcn_sw_init(adev);
if (r) if (r)
...@@ -104,12 +104,13 @@ static int vcn_v4_0_3_sw_init(void *handle) ...@@ -104,12 +104,13 @@ static int vcn_v4_0_3_sw_init(void *handle)
for (i = 0; i < adev->vcn.num_vcn_inst; i++) { for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
volatile struct amdgpu_vcn4_fw_shared *fw_shared; volatile struct amdgpu_vcn4_fw_shared *fw_shared;
if (adev->vcn.harvest_config & (1 << i)) vcn_inst = GET_INST(VCN, i);
continue;
ring = &adev->vcn.inst[i].ring_enc[0]; ring = &adev->vcn.inst[i].ring_enc[0];
ring->use_doorbell = true; ring->use_doorbell = true;
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 9 * i; ring->doorbell_index =
(adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
9 * vcn_inst;
ring->vm_hub = AMDGPU_MMHUB0(adev->vcn.inst[i].aid_id); ring->vm_hub = AMDGPU_MMHUB0(adev->vcn.inst[i].aid_id);
sprintf(ring->name, "vcn_unified_%d", i); sprintf(ring->name, "vcn_unified_%d", i);
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0, r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0,
...@@ -148,8 +149,6 @@ static int vcn_v4_0_3_sw_fini(void *handle) ...@@ -148,8 +149,6 @@ static int vcn_v4_0_3_sw_fini(void *handle)
for (i = 0; i < adev->vcn.num_vcn_inst; i++) { for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
volatile struct amdgpu_vcn4_fw_shared *fw_shared; volatile struct amdgpu_vcn4_fw_shared *fw_shared;
if (adev->vcn.harvest_config & (1 << i))
continue;
fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr; fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
fw_shared->present_flag_0 = 0; fw_shared->present_flag_0 = 0;
fw_shared->sq.is_enabled = cpu_to_le32(false); fw_shared->sq.is_enabled = cpu_to_le32(false);
...@@ -177,20 +176,24 @@ static int vcn_v4_0_3_hw_init(void *handle) ...@@ -177,20 +176,24 @@ static int vcn_v4_0_3_hw_init(void *handle)
{ {
struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct amdgpu_device *adev = (struct amdgpu_device *)handle;
struct amdgpu_ring *ring; struct amdgpu_ring *ring;
int i, r; int i, r, vcn_inst;
for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
if (adev->vcn.harvest_config & (1 << i)) vcn_inst = GET_INST(VCN, i);
continue;
ring = &adev->vcn.inst[i].ring_enc[0]; ring = &adev->vcn.inst[i].ring_enc[0];
if (ring->use_doorbell) { if (ring->use_doorbell) {
adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell, adev->nbio.funcs->vcn_doorbell_range(
(adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 9 * i, adev, ring->use_doorbell,
(adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
9 * vcn_inst,
adev->vcn.inst[i].aid_id); adev->vcn.inst[i].aid_id);
WREG32_SOC15(VCN, ring->me, regVCN_RB1_DB_CTRL, WREG32_SOC15(
ring->doorbell_index << VCN_RB1_DB_CTRL__OFFSET__SHIFT | VCN, GET_INST(VCN, ring->me),
regVCN_RB1_DB_CTRL,
ring->doorbell_index
<< VCN_RB1_DB_CTRL__OFFSET__SHIFT |
VCN_RB1_DB_CTRL__EN_MASK); VCN_RB1_DB_CTRL__EN_MASK);
} }
...@@ -278,54 +281,67 @@ static int vcn_v4_0_3_resume(void *handle) ...@@ -278,54 +281,67 @@ static int vcn_v4_0_3_resume(void *handle)
*/ */
static void vcn_v4_0_3_mc_resume(struct amdgpu_device *adev, int inst_idx) static void vcn_v4_0_3_mc_resume(struct amdgpu_device *adev, int inst_idx)
{ {
uint32_t offset, size; uint32_t offset, size, vcn_inst;
const struct common_firmware_header *hdr; const struct common_firmware_header *hdr;
hdr = (const struct common_firmware_header *)adev->vcn.fw->data; hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8); size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
vcn_inst = GET_INST(VCN, inst_idx);
/* cache window 0: fw */ /* cache window 0: fw */
if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
WREG32_SOC15(VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, WREG32_SOC15(
(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst_idx].tmr_mc_addr_lo)); VCN, vcn_inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
WREG32_SOC15(VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst_idx]
(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst_idx].tmr_mc_addr_hi)); .tmr_mc_addr_lo));
WREG32_SOC15(VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET0, 0); WREG32_SOC15(
VCN, vcn_inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst_idx]
.tmr_mc_addr_hi));
WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_OFFSET0, 0);
offset = 0; offset = 0;
} else { } else {
WREG32_SOC15(VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr)); lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr));
WREG32_SOC15(VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, WREG32_SOC15(VCN, vcn_inst,
regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr)); upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr));
offset = size; offset = size;
WREG32_SOC15(VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET0, WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_OFFSET0,
AMDGPU_UVD_FIRMWARE_OFFSET >> 3); AMDGPU_UVD_FIRMWARE_OFFSET >> 3);
} }
WREG32_SOC15(VCN, inst_idx, regUVD_VCPU_CACHE_SIZE0, size); WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_SIZE0, size);
/* cache window 1: stack */ /* cache window 1: stack */
WREG32_SOC15(VCN, inst_idx, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW, WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW,
lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset)); lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset));
WREG32_SOC15(VCN, inst_idx, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH, WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH,
upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset)); upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset));
WREG32_SOC15(VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET1, 0); WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_OFFSET1, 0);
WREG32_SOC15(VCN, inst_idx, regUVD_VCPU_CACHE_SIZE1, AMDGPU_VCN_STACK_SIZE); WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_SIZE1,
AMDGPU_VCN_STACK_SIZE);
/* cache window 2: context */ /* cache window 2: context */
WREG32_SOC15(VCN, inst_idx, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW, WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW,
lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE)); lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset +
WREG32_SOC15(VCN, inst_idx, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH, AMDGPU_VCN_STACK_SIZE));
upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE)); WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH,
WREG32_SOC15(VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET2, 0); upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset +
WREG32_SOC15(VCN, inst_idx, regUVD_VCPU_CACHE_SIZE2, AMDGPU_VCN_CONTEXT_SIZE); AMDGPU_VCN_STACK_SIZE));
WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_OFFSET2, 0);
WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_SIZE2,
AMDGPU_VCN_CONTEXT_SIZE);
/* non-cache window */ /* non-cache window */
WREG32_SOC15(VCN, inst_idx, regUVD_LMI_VCPU_NC0_64BIT_BAR_LOW, WREG32_SOC15(
VCN, vcn_inst, regUVD_LMI_VCPU_NC0_64BIT_BAR_LOW,
lower_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr)); lower_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr));
WREG32_SOC15(VCN, inst_idx, regUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH, WREG32_SOC15(
VCN, vcn_inst, regUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH,
upper_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr)); upper_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr));
WREG32_SOC15(VCN, inst_idx, regUVD_VCPU_NONCACHE_OFFSET0, 0); WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_NONCACHE_OFFSET0, 0);
WREG32_SOC15(VCN, inst_idx, regUVD_VCPU_NONCACHE_SIZE0, WREG32_SOC15(
VCN, vcn_inst, regUVD_VCPU_NONCACHE_SIZE0,
AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn4_fw_shared))); AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn4_fw_shared)));
} }
...@@ -454,18 +470,21 @@ static void vcn_v4_0_3_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_i ...@@ -454,18 +470,21 @@ static void vcn_v4_0_3_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_i
static void vcn_v4_0_3_disable_clock_gating(struct amdgpu_device *adev, int inst_idx) static void vcn_v4_0_3_disable_clock_gating(struct amdgpu_device *adev, int inst_idx)
{ {
uint32_t data; uint32_t data;
int vcn_inst;
if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG) if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
return; return;
vcn_inst = GET_INST(VCN, inst_idx);
/* VCN disable CGC */ /* VCN disable CGC */
data = RREG32_SOC15(VCN, inst_idx, regUVD_CGC_CTRL); data = RREG32_SOC15(VCN, vcn_inst, regUVD_CGC_CTRL);
data &= ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK; data &= ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK;
data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT; data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT; data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
WREG32_SOC15(VCN, inst_idx, regUVD_CGC_CTRL, data); WREG32_SOC15(VCN, vcn_inst, regUVD_CGC_CTRL, data);
data = RREG32_SOC15(VCN, inst_idx, regUVD_CGC_GATE); data = RREG32_SOC15(VCN, vcn_inst, regUVD_CGC_GATE);
data &= ~(UVD_CGC_GATE__SYS_MASK data &= ~(UVD_CGC_GATE__SYS_MASK
| UVD_CGC_GATE__MPEG2_MASK | UVD_CGC_GATE__MPEG2_MASK
| UVD_CGC_GATE__REGS_MASK | UVD_CGC_GATE__REGS_MASK
...@@ -479,10 +498,10 @@ static void vcn_v4_0_3_disable_clock_gating(struct amdgpu_device *adev, int inst ...@@ -479,10 +498,10 @@ static void vcn_v4_0_3_disable_clock_gating(struct amdgpu_device *adev, int inst
| UVD_CGC_GATE__VCPU_MASK | UVD_CGC_GATE__VCPU_MASK
| UVD_CGC_GATE__MMSCH_MASK); | UVD_CGC_GATE__MMSCH_MASK);
WREG32_SOC15(VCN, inst_idx, regUVD_CGC_GATE, data); WREG32_SOC15(VCN, vcn_inst, regUVD_CGC_GATE, data);
SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_CGC_GATE, 0, 0xFFFFFFFF); SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_CGC_GATE, 0, 0xFFFFFFFF);
data = RREG32_SOC15(VCN, inst_idx, regUVD_CGC_CTRL); data = RREG32_SOC15(VCN, vcn_inst, regUVD_CGC_CTRL);
data &= ~(UVD_CGC_CTRL__SYS_MODE_MASK data &= ~(UVD_CGC_CTRL__SYS_MODE_MASK
| UVD_CGC_CTRL__MPEG2_MODE_MASK | UVD_CGC_CTRL__MPEG2_MODE_MASK
| UVD_CGC_CTRL__REGS_MODE_MASK | UVD_CGC_CTRL__REGS_MODE_MASK
...@@ -495,9 +514,9 @@ static void vcn_v4_0_3_disable_clock_gating(struct amdgpu_device *adev, int inst ...@@ -495,9 +514,9 @@ static void vcn_v4_0_3_disable_clock_gating(struct amdgpu_device *adev, int inst
| UVD_CGC_CTRL__WCB_MODE_MASK | UVD_CGC_CTRL__WCB_MODE_MASK
| UVD_CGC_CTRL__VCPU_MODE_MASK | UVD_CGC_CTRL__VCPU_MODE_MASK
| UVD_CGC_CTRL__MMSCH_MODE_MASK); | UVD_CGC_CTRL__MMSCH_MODE_MASK);
WREG32_SOC15(VCN, inst_idx, regUVD_CGC_CTRL, data); WREG32_SOC15(VCN, vcn_inst, regUVD_CGC_CTRL, data);
data = RREG32_SOC15(VCN, inst_idx, regUVD_SUVD_CGC_GATE); data = RREG32_SOC15(VCN, vcn_inst, regUVD_SUVD_CGC_GATE);
data |= (UVD_SUVD_CGC_GATE__SRE_MASK data |= (UVD_SUVD_CGC_GATE__SRE_MASK
| UVD_SUVD_CGC_GATE__SIT_MASK | UVD_SUVD_CGC_GATE__SIT_MASK
| UVD_SUVD_CGC_GATE__SMP_MASK | UVD_SUVD_CGC_GATE__SMP_MASK
...@@ -519,9 +538,9 @@ static void vcn_v4_0_3_disable_clock_gating(struct amdgpu_device *adev, int inst ...@@ -519,9 +538,9 @@ static void vcn_v4_0_3_disable_clock_gating(struct amdgpu_device *adev, int inst
| UVD_SUVD_CGC_GATE__SIT_VP9_DEC_MASK | UVD_SUVD_CGC_GATE__SIT_VP9_DEC_MASK
| UVD_SUVD_CGC_GATE__SDB_VP9_MASK | UVD_SUVD_CGC_GATE__SDB_VP9_MASK
| UVD_SUVD_CGC_GATE__IME_HEVC_MASK); | UVD_SUVD_CGC_GATE__IME_HEVC_MASK);
WREG32_SOC15(VCN, inst_idx, regUVD_SUVD_CGC_GATE, data); WREG32_SOC15(VCN, vcn_inst, regUVD_SUVD_CGC_GATE, data);
data = RREG32_SOC15(VCN, inst_idx, regUVD_SUVD_CGC_CTRL); data = RREG32_SOC15(VCN, vcn_inst, regUVD_SUVD_CGC_CTRL);
data &= ~(UVD_SUVD_CGC_CTRL__SRE_MODE_MASK data &= ~(UVD_SUVD_CGC_CTRL__SRE_MODE_MASK
| UVD_SUVD_CGC_CTRL__SIT_MODE_MASK | UVD_SUVD_CGC_CTRL__SIT_MODE_MASK
| UVD_SUVD_CGC_CTRL__SMP_MODE_MASK | UVD_SUVD_CGC_CTRL__SMP_MODE_MASK
...@@ -530,7 +549,7 @@ static void vcn_v4_0_3_disable_clock_gating(struct amdgpu_device *adev, int inst ...@@ -530,7 +549,7 @@ static void vcn_v4_0_3_disable_clock_gating(struct amdgpu_device *adev, int inst
| UVD_SUVD_CGC_CTRL__ENT_MODE_MASK | UVD_SUVD_CGC_CTRL__ENT_MODE_MASK
| UVD_SUVD_CGC_CTRL__IME_MODE_MASK | UVD_SUVD_CGC_CTRL__IME_MODE_MASK
| UVD_SUVD_CGC_CTRL__SITE_MODE_MASK); | UVD_SUVD_CGC_CTRL__SITE_MODE_MASK);
WREG32_SOC15(VCN, inst_idx, regUVD_SUVD_CGC_CTRL, data); WREG32_SOC15(VCN, vcn_inst, regUVD_SUVD_CGC_CTRL, data);
} }
/** /**
...@@ -595,18 +614,21 @@ static void vcn_v4_0_3_disable_clock_gating_dpg_mode(struct amdgpu_device *adev, ...@@ -595,18 +614,21 @@ static void vcn_v4_0_3_disable_clock_gating_dpg_mode(struct amdgpu_device *adev,
static void vcn_v4_0_3_enable_clock_gating(struct amdgpu_device *adev, int inst_idx) static void vcn_v4_0_3_enable_clock_gating(struct amdgpu_device *adev, int inst_idx)
{ {
uint32_t data; uint32_t data;
int vcn_inst;
if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG) if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
return; return;
vcn_inst = GET_INST(VCN, inst_idx);
/* enable VCN CGC */ /* enable VCN CGC */
data = RREG32_SOC15(VCN, inst_idx, regUVD_CGC_CTRL); data = RREG32_SOC15(VCN, vcn_inst, regUVD_CGC_CTRL);
data |= 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; data |= 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT; data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT; data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
WREG32_SOC15(VCN, inst_idx, regUVD_CGC_CTRL, data); WREG32_SOC15(VCN, vcn_inst, regUVD_CGC_CTRL, data);
data = RREG32_SOC15(VCN, inst_idx, regUVD_CGC_CTRL); data = RREG32_SOC15(VCN, vcn_inst, regUVD_CGC_CTRL);
data |= (UVD_CGC_CTRL__SYS_MODE_MASK data |= (UVD_CGC_CTRL__SYS_MODE_MASK
| UVD_CGC_CTRL__MPEG2_MODE_MASK | UVD_CGC_CTRL__MPEG2_MODE_MASK
| UVD_CGC_CTRL__REGS_MODE_MASK | UVD_CGC_CTRL__REGS_MODE_MASK
...@@ -618,9 +640,9 @@ static void vcn_v4_0_3_enable_clock_gating(struct amdgpu_device *adev, int inst_ ...@@ -618,9 +640,9 @@ static void vcn_v4_0_3_enable_clock_gating(struct amdgpu_device *adev, int inst_
| UVD_CGC_CTRL__LRBBM_MODE_MASK | UVD_CGC_CTRL__LRBBM_MODE_MASK
| UVD_CGC_CTRL__WCB_MODE_MASK | UVD_CGC_CTRL__WCB_MODE_MASK
| UVD_CGC_CTRL__VCPU_MODE_MASK); | UVD_CGC_CTRL__VCPU_MODE_MASK);
WREG32_SOC15(VCN, inst_idx, regUVD_CGC_CTRL, data); WREG32_SOC15(VCN, vcn_inst, regUVD_CGC_CTRL, data);
data = RREG32_SOC15(VCN, inst_idx, regUVD_SUVD_CGC_CTRL); data = RREG32_SOC15(VCN, vcn_inst, regUVD_SUVD_CGC_CTRL);
data |= (UVD_SUVD_CGC_CTRL__SRE_MODE_MASK data |= (UVD_SUVD_CGC_CTRL__SRE_MODE_MASK
| UVD_SUVD_CGC_CTRL__SIT_MODE_MASK | UVD_SUVD_CGC_CTRL__SIT_MODE_MASK
| UVD_SUVD_CGC_CTRL__SMP_MODE_MASK | UVD_SUVD_CGC_CTRL__SMP_MODE_MASK
...@@ -629,7 +651,7 @@ static void vcn_v4_0_3_enable_clock_gating(struct amdgpu_device *adev, int inst_ ...@@ -629,7 +651,7 @@ static void vcn_v4_0_3_enable_clock_gating(struct amdgpu_device *adev, int inst_
| UVD_SUVD_CGC_CTRL__ENT_MODE_MASK | UVD_SUVD_CGC_CTRL__ENT_MODE_MASK
| UVD_SUVD_CGC_CTRL__IME_MODE_MASK | UVD_SUVD_CGC_CTRL__IME_MODE_MASK
| UVD_SUVD_CGC_CTRL__SITE_MODE_MASK); | UVD_SUVD_CGC_CTRL__SITE_MODE_MASK);
WREG32_SOC15(VCN, inst_idx, regUVD_SUVD_CGC_CTRL, data); WREG32_SOC15(VCN, vcn_inst, regUVD_SUVD_CGC_CTRL, data);
} }
/** /**
...@@ -646,16 +668,18 @@ static int vcn_v4_0_3_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, b ...@@ -646,16 +668,18 @@ static int vcn_v4_0_3_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, b
volatile struct amdgpu_vcn4_fw_shared *fw_shared = volatile struct amdgpu_vcn4_fw_shared *fw_shared =
adev->vcn.inst[inst_idx].fw_shared.cpu_addr; adev->vcn.inst[inst_idx].fw_shared.cpu_addr;
struct amdgpu_ring *ring; struct amdgpu_ring *ring;
int vcn_inst;
uint32_t tmp; uint32_t tmp;
vcn_inst = GET_INST(VCN, inst_idx);
/* disable register anti-hang mechanism */ /* disable register anti-hang mechanism */
WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, regUVD_POWER_STATUS), 1, WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_POWER_STATUS), 1,
~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK); ~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
/* enable dynamic power gating mode */ /* enable dynamic power gating mode */
tmp = RREG32_SOC15(VCN, inst_idx, regUVD_POWER_STATUS); tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_POWER_STATUS);
tmp |= UVD_POWER_STATUS__UVD_PG_MODE_MASK; tmp |= UVD_POWER_STATUS__UVD_PG_MODE_MASK;
tmp |= UVD_POWER_STATUS__UVD_PG_EN_MASK; tmp |= UVD_POWER_STATUS__UVD_PG_EN_MASK;
WREG32_SOC15(VCN, inst_idx, regUVD_POWER_STATUS, tmp); WREG32_SOC15(VCN, vcn_inst, regUVD_POWER_STATUS, tmp);
if (indirect) if (indirect)
adev->vcn.inst[inst_idx].dpg_sram_curr_addr = adev->vcn.inst[inst_idx].dpg_sram_curr_addr =
...@@ -737,27 +761,28 @@ static int vcn_v4_0_3_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, b ...@@ -737,27 +761,28 @@ static int vcn_v4_0_3_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, b
ring = &adev->vcn.inst[inst_idx].ring_enc[0]; ring = &adev->vcn.inst[inst_idx].ring_enc[0];
/* program the RB_BASE for ring buffer */ /* program the RB_BASE for ring buffer */
WREG32_SOC15(VCN, inst_idx, regUVD_RB_BASE_LO, WREG32_SOC15(VCN, vcn_inst, regUVD_RB_BASE_LO,
lower_32_bits(ring->gpu_addr)); lower_32_bits(ring->gpu_addr));
WREG32_SOC15(VCN, inst_idx, regUVD_RB_BASE_HI, WREG32_SOC15(VCN, vcn_inst, regUVD_RB_BASE_HI,
upper_32_bits(ring->gpu_addr)); upper_32_bits(ring->gpu_addr));
WREG32_SOC15(VCN, inst_idx, regUVD_RB_SIZE, ring->ring_size / sizeof(uint32_t)); WREG32_SOC15(VCN, vcn_inst, regUVD_RB_SIZE,
ring->ring_size / sizeof(uint32_t));
/* resetting ring, fw should not check RB ring */ /* resetting ring, fw should not check RB ring */
tmp = RREG32_SOC15(VCN, inst_idx, regVCN_RB_ENABLE); tmp = RREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE);
tmp &= ~(VCN_RB_ENABLE__RB_EN_MASK); tmp &= ~(VCN_RB_ENABLE__RB_EN_MASK);
WREG32_SOC15(VCN, inst_idx, regVCN_RB_ENABLE, tmp); WREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE, tmp);
fw_shared->sq.queue_mode |= FW_QUEUE_RING_RESET; fw_shared->sq.queue_mode |= FW_QUEUE_RING_RESET;
/* Initialize the ring buffer's read and write pointers */ /* Initialize the ring buffer's read and write pointers */
WREG32_SOC15(VCN, inst_idx, regUVD_RB_RPTR, 0); WREG32_SOC15(VCN, vcn_inst, regUVD_RB_RPTR, 0);
WREG32_SOC15(VCN, inst_idx, regUVD_RB_WPTR, 0); WREG32_SOC15(VCN, vcn_inst, regUVD_RB_WPTR, 0);
ring->wptr = RREG32_SOC15(VCN, inst_idx, regUVD_RB_WPTR); ring->wptr = RREG32_SOC15(VCN, vcn_inst, regUVD_RB_WPTR);
tmp = RREG32_SOC15(VCN, inst_idx, regVCN_RB_ENABLE); tmp = RREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE);
tmp |= VCN_RB_ENABLE__RB_EN_MASK; tmp |= VCN_RB_ENABLE__RB_EN_MASK;
WREG32_SOC15(VCN, inst_idx, regVCN_RB_ENABLE, tmp); WREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE, tmp);
fw_shared->sq.queue_mode &= ~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF); fw_shared->sq.queue_mode &= ~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF);
/*resetting done, fw can check RB ring */ /*resetting done, fw can check RB ring */
...@@ -777,74 +802,75 @@ static int vcn_v4_0_3_start(struct amdgpu_device *adev) ...@@ -777,74 +802,75 @@ static int vcn_v4_0_3_start(struct amdgpu_device *adev)
{ {
volatile struct amdgpu_vcn4_fw_shared *fw_shared; volatile struct amdgpu_vcn4_fw_shared *fw_shared;
struct amdgpu_ring *ring; struct amdgpu_ring *ring;
int i, j, k, r, vcn_inst;
uint32_t tmp; uint32_t tmp;
int i, j, k, r;
if (adev->pm.dpm_enabled) if (adev->pm.dpm_enabled)
amdgpu_dpm_enable_uvd(adev, true); amdgpu_dpm_enable_uvd(adev, true);
for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
if (adev->vcn.harvest_config & (1 << i))
continue;
if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) { if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
r = vcn_v4_0_3_start_dpg_mode(adev, i, adev->vcn.indirect_sram); r = vcn_v4_0_3_start_dpg_mode(adev, i, adev->vcn.indirect_sram);
continue; continue;
} }
vcn_inst = GET_INST(VCN, i);
/* set VCN status busy */ /* set VCN status busy */
tmp = RREG32_SOC15(VCN, i, regUVD_STATUS) | UVD_STATUS__UVD_BUSY; tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_STATUS) |
WREG32_SOC15(VCN, i, regUVD_STATUS, tmp); UVD_STATUS__UVD_BUSY;
WREG32_SOC15(VCN, vcn_inst, regUVD_STATUS, tmp);
/*SW clock gating */ /*SW clock gating */
vcn_v4_0_3_disable_clock_gating(adev, i); vcn_v4_0_3_disable_clock_gating(adev, i);
/* enable VCPU clock */ /* enable VCPU clock */
WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_VCPU_CNTL),
UVD_VCPU_CNTL__CLK_EN_MASK, ~UVD_VCPU_CNTL__CLK_EN_MASK); UVD_VCPU_CNTL__CLK_EN_MASK,
~UVD_VCPU_CNTL__CLK_EN_MASK);
/* disable master interrupt */ /* disable master interrupt */
WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_MASTINT_EN), 0, WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_MASTINT_EN), 0,
~UVD_MASTINT_EN__VCPU_EN_MASK); ~UVD_MASTINT_EN__VCPU_EN_MASK);
/* enable LMI MC and UMC channels */ /* enable LMI MC and UMC channels */
WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_LMI_CTRL2), 0, WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_LMI_CTRL2), 0,
~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
tmp = RREG32_SOC15(VCN, i, regUVD_SOFT_RESET); tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_SOFT_RESET);
tmp &= ~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK; tmp &= ~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
tmp &= ~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK; tmp &= ~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
WREG32_SOC15(VCN, i, regUVD_SOFT_RESET, tmp); WREG32_SOC15(VCN, vcn_inst, regUVD_SOFT_RESET, tmp);
/* setup regUVD_LMI_CTRL */ /* setup regUVD_LMI_CTRL */
tmp = RREG32_SOC15(VCN, i, regUVD_LMI_CTRL); tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_LMI_CTRL);
WREG32_SOC15(VCN, i, regUVD_LMI_CTRL, tmp | WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_CTRL,
UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK | tmp | UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
UVD_LMI_CTRL__MASK_MC_URGENT_MASK | UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK | UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK); UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK);
/* setup regUVD_MPC_CNTL */ /* setup regUVD_MPC_CNTL */
tmp = RREG32_SOC15(VCN, i, regUVD_MPC_CNTL); tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_MPC_CNTL);
tmp &= ~UVD_MPC_CNTL__REPLACEMENT_MODE_MASK; tmp &= ~UVD_MPC_CNTL__REPLACEMENT_MODE_MASK;
tmp |= 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT; tmp |= 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT;
WREG32_SOC15(VCN, i, regUVD_MPC_CNTL, tmp); WREG32_SOC15(VCN, vcn_inst, regUVD_MPC_CNTL, tmp);
/* setup UVD_MPC_SET_MUXA0 */ /* setup UVD_MPC_SET_MUXA0 */
WREG32_SOC15(VCN, i, regUVD_MPC_SET_MUXA0, WREG32_SOC15(VCN, vcn_inst, regUVD_MPC_SET_MUXA0,
((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) | ((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) |
(0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) | (0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) |
(0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) | (0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) |
(0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT))); (0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)));
/* setup UVD_MPC_SET_MUXB0 */ /* setup UVD_MPC_SET_MUXB0 */
WREG32_SOC15(VCN, i, regUVD_MPC_SET_MUXB0, WREG32_SOC15(VCN, vcn_inst, regUVD_MPC_SET_MUXB0,
((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) | ((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) |
(0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) | (0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) |
(0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) | (0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) |
(0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT))); (0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)));
/* setup UVD_MPC_SET_MUX */ /* setup UVD_MPC_SET_MUX */
WREG32_SOC15(VCN, i, regUVD_MPC_SET_MUX, WREG32_SOC15(VCN, vcn_inst, regUVD_MPC_SET_MUX,
((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) | ((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) |
(0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) | (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
(0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT))); (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)));
...@@ -852,24 +878,25 @@ static int vcn_v4_0_3_start(struct amdgpu_device *adev) ...@@ -852,24 +878,25 @@ static int vcn_v4_0_3_start(struct amdgpu_device *adev)
vcn_v4_0_3_mc_resume(adev, i); vcn_v4_0_3_mc_resume(adev, i);
/* VCN global tiling registers */ /* VCN global tiling registers */
WREG32_SOC15(VCN, i, regUVD_GFX8_ADDR_CONFIG, WREG32_SOC15(VCN, vcn_inst, regUVD_GFX8_ADDR_CONFIG,
adev->gfx.config.gb_addr_config); adev->gfx.config.gb_addr_config);
WREG32_SOC15(VCN, i, regUVD_GFX10_ADDR_CONFIG, WREG32_SOC15(VCN, vcn_inst, regUVD_GFX10_ADDR_CONFIG,
adev->gfx.config.gb_addr_config); adev->gfx.config.gb_addr_config);
/* unblock VCPU register access */ /* unblock VCPU register access */
WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_RB_ARB_CTRL), 0, WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_RB_ARB_CTRL), 0,
~UVD_RB_ARB_CTRL__VCPU_DIS_MASK); ~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
/* release VCPU reset to boot */ /* release VCPU reset to boot */
WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), 0, WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_VCPU_CNTL), 0,
~UVD_VCPU_CNTL__BLK_RST_MASK); ~UVD_VCPU_CNTL__BLK_RST_MASK);
for (j = 0; j < 10; ++j) { for (j = 0; j < 10; ++j) {
uint32_t status; uint32_t status;
for (k = 0; k < 100; ++k) { for (k = 0; k < 100; ++k) {
status = RREG32_SOC15(VCN, i, regUVD_STATUS); status = RREG32_SOC15(VCN, vcn_inst,
regUVD_STATUS);
if (status & 2) if (status & 2)
break; break;
mdelay(10); mdelay(10);
...@@ -880,12 +907,14 @@ static int vcn_v4_0_3_start(struct amdgpu_device *adev) ...@@ -880,12 +907,14 @@ static int vcn_v4_0_3_start(struct amdgpu_device *adev)
DRM_DEV_ERROR(adev->dev, DRM_DEV_ERROR(adev->dev,
"VCN decode not responding, trying to reset the VCPU!!!\n"); "VCN decode not responding, trying to reset the VCPU!!!\n");
WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst,
regUVD_VCPU_CNTL),
UVD_VCPU_CNTL__BLK_RST_MASK, UVD_VCPU_CNTL__BLK_RST_MASK,
~UVD_VCPU_CNTL__BLK_RST_MASK); ~UVD_VCPU_CNTL__BLK_RST_MASK);
mdelay(10); mdelay(10);
WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), 0, WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst,
~UVD_VCPU_CNTL__BLK_RST_MASK); regUVD_VCPU_CNTL),
0, ~UVD_VCPU_CNTL__BLK_RST_MASK);
mdelay(10); mdelay(10);
r = -1; r = -1;
...@@ -897,39 +926,40 @@ static int vcn_v4_0_3_start(struct amdgpu_device *adev) ...@@ -897,39 +926,40 @@ static int vcn_v4_0_3_start(struct amdgpu_device *adev)
} }
/* enable master interrupt */ /* enable master interrupt */
WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_MASTINT_EN), WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_MASTINT_EN),
UVD_MASTINT_EN__VCPU_EN_MASK, UVD_MASTINT_EN__VCPU_EN_MASK,
~UVD_MASTINT_EN__VCPU_EN_MASK); ~UVD_MASTINT_EN__VCPU_EN_MASK);
/* clear the busy bit of VCN_STATUS */ /* clear the busy bit of VCN_STATUS */
WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_STATUS), 0, WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_STATUS), 0,
~(2 << UVD_STATUS__VCPU_REPORT__SHIFT)); ~(2 << UVD_STATUS__VCPU_REPORT__SHIFT));
ring = &adev->vcn.inst[i].ring_enc[0]; ring = &adev->vcn.inst[i].ring_enc[0];
fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr; fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
/* program the RB_BASE for ring buffer */ /* program the RB_BASE for ring buffer */
WREG32_SOC15(VCN, i, regUVD_RB_BASE_LO, WREG32_SOC15(VCN, vcn_inst, regUVD_RB_BASE_LO,
lower_32_bits(ring->gpu_addr)); lower_32_bits(ring->gpu_addr));
WREG32_SOC15(VCN, i, regUVD_RB_BASE_HI, WREG32_SOC15(VCN, vcn_inst, regUVD_RB_BASE_HI,
upper_32_bits(ring->gpu_addr)); upper_32_bits(ring->gpu_addr));
WREG32_SOC15(VCN, i, regUVD_RB_SIZE, ring->ring_size / sizeof(uint32_t)); WREG32_SOC15(VCN, vcn_inst, regUVD_RB_SIZE,
ring->ring_size / sizeof(uint32_t));
/* resetting ring, fw should not check RB ring */ /* resetting ring, fw should not check RB ring */
tmp = RREG32_SOC15(VCN, i, regVCN_RB_ENABLE); tmp = RREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE);
tmp &= ~(VCN_RB_ENABLE__RB_EN_MASK); tmp &= ~(VCN_RB_ENABLE__RB_EN_MASK);
WREG32_SOC15(VCN, i, regVCN_RB_ENABLE, tmp); WREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE, tmp);
/* Initialize the ring buffer's read and write pointers */ /* Initialize the ring buffer's read and write pointers */
WREG32_SOC15(VCN, i, regUVD_RB_RPTR, 0); WREG32_SOC15(VCN, vcn_inst, regUVD_RB_RPTR, 0);
WREG32_SOC15(VCN, i, regUVD_RB_WPTR, 0); WREG32_SOC15(VCN, vcn_inst, regUVD_RB_WPTR, 0);
tmp = RREG32_SOC15(VCN, i, regVCN_RB_ENABLE); tmp = RREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE);
tmp |= VCN_RB_ENABLE__RB_EN_MASK; tmp |= VCN_RB_ENABLE__RB_EN_MASK;
WREG32_SOC15(VCN, i, regVCN_RB_ENABLE, tmp); WREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE, tmp);
ring->wptr = RREG32_SOC15(VCN, i, regUVD_RB_WPTR); ring->wptr = RREG32_SOC15(VCN, vcn_inst, regUVD_RB_WPTR);
fw_shared->sq.queue_mode &= fw_shared->sq.queue_mode &=
cpu_to_le32(~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF)); cpu_to_le32(~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF));
...@@ -948,20 +978,23 @@ static int vcn_v4_0_3_start(struct amdgpu_device *adev) ...@@ -948,20 +978,23 @@ static int vcn_v4_0_3_start(struct amdgpu_device *adev)
static int vcn_v4_0_3_stop_dpg_mode(struct amdgpu_device *adev, int inst_idx) static int vcn_v4_0_3_stop_dpg_mode(struct amdgpu_device *adev, int inst_idx)
{ {
uint32_t tmp; uint32_t tmp;
int vcn_inst;
vcn_inst = GET_INST(VCN, inst_idx);
/* Wait for power status to be 1 */ /* Wait for power status to be 1 */
SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_POWER_STATUS, 1, SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_POWER_STATUS, 1,
UVD_POWER_STATUS__UVD_POWER_STATUS_MASK); UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
/* wait for read ptr to be equal to write ptr */ /* wait for read ptr to be equal to write ptr */
tmp = RREG32_SOC15(VCN, inst_idx, regUVD_RB_WPTR); tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_RB_WPTR);
SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_RB_RPTR, tmp, 0xFFFFFFFF); SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_RB_RPTR, tmp, 0xFFFFFFFF);
SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_POWER_STATUS, 1, SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_POWER_STATUS, 1,
UVD_POWER_STATUS__UVD_POWER_STATUS_MASK); UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
/* disable dynamic power gating mode */ /* disable dynamic power gating mode */
WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, regUVD_POWER_STATUS), 0, WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_POWER_STATUS), 0,
~UVD_POWER_STATUS__UVD_PG_MODE_MASK); ~UVD_POWER_STATUS__UVD_PG_MODE_MASK);
return 0; return 0;
} }
...@@ -976,12 +1009,11 @@ static int vcn_v4_0_3_stop_dpg_mode(struct amdgpu_device *adev, int inst_idx) ...@@ -976,12 +1009,11 @@ static int vcn_v4_0_3_stop_dpg_mode(struct amdgpu_device *adev, int inst_idx)
static int vcn_v4_0_3_stop(struct amdgpu_device *adev) static int vcn_v4_0_3_stop(struct amdgpu_device *adev)
{ {
volatile struct amdgpu_vcn4_fw_shared *fw_shared; volatile struct amdgpu_vcn4_fw_shared *fw_shared;
int i, r = 0, vcn_inst;
uint32_t tmp; uint32_t tmp;
int i, r = 0;
for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
if (adev->vcn.harvest_config & (1 << i)) vcn_inst = GET_INST(VCN, i);
continue;
fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr; fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
fw_shared->sq.queue_mode |= FW_QUEUE_DPG_HOLD_OFF; fw_shared->sq.queue_mode |= FW_QUEUE_DPG_HOLD_OFF;
...@@ -992,7 +1024,8 @@ static int vcn_v4_0_3_stop(struct amdgpu_device *adev) ...@@ -992,7 +1024,8 @@ static int vcn_v4_0_3_stop(struct amdgpu_device *adev)
} }
/* wait for vcn idle */ /* wait for vcn idle */
r = SOC15_WAIT_ON_RREG(VCN, i, regUVD_STATUS, UVD_STATUS__IDLE, 0x7); r = SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_STATUS,
UVD_STATUS__IDLE, 0x7);
if (r) if (r)
goto Done; goto Done;
...@@ -1000,45 +1033,47 @@ static int vcn_v4_0_3_stop(struct amdgpu_device *adev) ...@@ -1000,45 +1033,47 @@ static int vcn_v4_0_3_stop(struct amdgpu_device *adev)
UVD_LMI_STATUS__READ_CLEAN_MASK | UVD_LMI_STATUS__READ_CLEAN_MASK |
UVD_LMI_STATUS__WRITE_CLEAN_MASK | UVD_LMI_STATUS__WRITE_CLEAN_MASK |
UVD_LMI_STATUS__WRITE_CLEAN_RAW_MASK; UVD_LMI_STATUS__WRITE_CLEAN_RAW_MASK;
r = SOC15_WAIT_ON_RREG(VCN, i, regUVD_LMI_STATUS, tmp, tmp); r = SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_LMI_STATUS, tmp,
tmp);
if (r) if (r)
goto Done; goto Done;
/* stall UMC channel */ /* stall UMC channel */
tmp = RREG32_SOC15(VCN, i, regUVD_LMI_CTRL2); tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_LMI_CTRL2);
tmp |= UVD_LMI_CTRL2__STALL_ARB_UMC_MASK; tmp |= UVD_LMI_CTRL2__STALL_ARB_UMC_MASK;
WREG32_SOC15(VCN, i, regUVD_LMI_CTRL2, tmp); WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_CTRL2, tmp);
tmp = UVD_LMI_STATUS__UMC_READ_CLEAN_RAW_MASK | tmp = UVD_LMI_STATUS__UMC_READ_CLEAN_RAW_MASK |
UVD_LMI_STATUS__UMC_WRITE_CLEAN_RAW_MASK; UVD_LMI_STATUS__UMC_WRITE_CLEAN_RAW_MASK;
r = SOC15_WAIT_ON_RREG(VCN, i, regUVD_LMI_STATUS, tmp, tmp); r = SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_LMI_STATUS, tmp,
tmp);
if (r) if (r)
goto Done; goto Done;
/* Unblock VCPU Register access */ /* Unblock VCPU Register access */
WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_RB_ARB_CTRL), WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_RB_ARB_CTRL),
UVD_RB_ARB_CTRL__VCPU_DIS_MASK, UVD_RB_ARB_CTRL__VCPU_DIS_MASK,
~UVD_RB_ARB_CTRL__VCPU_DIS_MASK); ~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
/* release VCPU reset to boot */ /* release VCPU reset to boot */
WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_VCPU_CNTL),
UVD_VCPU_CNTL__BLK_RST_MASK, UVD_VCPU_CNTL__BLK_RST_MASK,
~UVD_VCPU_CNTL__BLK_RST_MASK); ~UVD_VCPU_CNTL__BLK_RST_MASK);
/* disable VCPU clock */ /* disable VCPU clock */
WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), 0, WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_VCPU_CNTL), 0,
~(UVD_VCPU_CNTL__CLK_EN_MASK)); ~(UVD_VCPU_CNTL__CLK_EN_MASK));
/* reset LMI UMC/LMI/VCPU */ /* reset LMI UMC/LMI/VCPU */
tmp = RREG32_SOC15(VCN, i, regUVD_SOFT_RESET); tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_SOFT_RESET);
tmp |= UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK; tmp |= UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
WREG32_SOC15(VCN, i, regUVD_SOFT_RESET, tmp); WREG32_SOC15(VCN, vcn_inst, regUVD_SOFT_RESET, tmp);
tmp = RREG32_SOC15(VCN, i, regUVD_SOFT_RESET); tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_SOFT_RESET);
tmp |= UVD_SOFT_RESET__LMI_SOFT_RESET_MASK; tmp |= UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
WREG32_SOC15(VCN, i, regUVD_SOFT_RESET, tmp); WREG32_SOC15(VCN, vcn_inst, regUVD_SOFT_RESET, tmp);
/* clear VCN status */ /* clear VCN status */
WREG32_SOC15(VCN, i, regUVD_STATUS, 0); WREG32_SOC15(VCN, vcn_inst, regUVD_STATUS, 0);
/* apply HW clock gating */ /* apply HW clock gating */
vcn_v4_0_3_enable_clock_gating(adev, i); vcn_v4_0_3_enable_clock_gating(adev, i);
...@@ -1080,7 +1115,7 @@ static uint64_t vcn_v4_0_3_unified_ring_get_rptr(struct amdgpu_ring *ring) ...@@ -1080,7 +1115,7 @@ static uint64_t vcn_v4_0_3_unified_ring_get_rptr(struct amdgpu_ring *ring)
if (ring != &adev->vcn.inst[ring->me].ring_enc[0]) if (ring != &adev->vcn.inst[ring->me].ring_enc[0])
DRM_ERROR("wrong ring id is identified in %s", __func__); DRM_ERROR("wrong ring id is identified in %s", __func__);
return RREG32_SOC15(VCN, ring->me, regUVD_RB_RPTR); return RREG32_SOC15(VCN, GET_INST(VCN, ring->me), regUVD_RB_RPTR);
} }
/** /**
...@@ -1100,7 +1135,8 @@ static uint64_t vcn_v4_0_3_unified_ring_get_wptr(struct amdgpu_ring *ring) ...@@ -1100,7 +1135,8 @@ static uint64_t vcn_v4_0_3_unified_ring_get_wptr(struct amdgpu_ring *ring)
if (ring->use_doorbell) if (ring->use_doorbell)
return *ring->wptr_cpu_addr; return *ring->wptr_cpu_addr;
else else
return RREG32_SOC15(VCN, ring->me, regUVD_RB_WPTR); return RREG32_SOC15(VCN, GET_INST(VCN, ring->me),
regUVD_RB_WPTR);
} }
/** /**
...@@ -1121,7 +1157,8 @@ static void vcn_v4_0_3_unified_ring_set_wptr(struct amdgpu_ring *ring) ...@@ -1121,7 +1157,8 @@ static void vcn_v4_0_3_unified_ring_set_wptr(struct amdgpu_ring *ring)
*ring->wptr_cpu_addr = lower_32_bits(ring->wptr); *ring->wptr_cpu_addr = lower_32_bits(ring->wptr);
WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr));
} else { } else {
WREG32_SOC15(VCN, ring->me, regUVD_RB_WPTR, lower_32_bits(ring->wptr)); WREG32_SOC15(VCN, GET_INST(VCN, ring->me), regUVD_RB_WPTR,
lower_32_bits(ring->wptr));
} }
} }
...@@ -1163,14 +1200,14 @@ static const struct amdgpu_ring_funcs vcn_v4_0_3_unified_ring_vm_funcs = { ...@@ -1163,14 +1200,14 @@ static const struct amdgpu_ring_funcs vcn_v4_0_3_unified_ring_vm_funcs = {
*/ */
static void vcn_v4_0_3_set_unified_ring_funcs(struct amdgpu_device *adev) static void vcn_v4_0_3_set_unified_ring_funcs(struct amdgpu_device *adev)
{ {
int i; int i, vcn_inst;
for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
if (adev->vcn.harvest_config & (1 << i))
continue;
adev->vcn.inst[i].ring_enc[0].funcs = &vcn_v4_0_3_unified_ring_vm_funcs; adev->vcn.inst[i].ring_enc[0].funcs = &vcn_v4_0_3_unified_ring_vm_funcs;
adev->vcn.inst[i].ring_enc[0].me = i; adev->vcn.inst[i].ring_enc[0].me = i;
adev->vcn.inst[i].aid_id = i / adev->vcn.num_inst_per_aid; vcn_inst = GET_INST(VCN, i);
adev->vcn.inst[i].aid_id =
vcn_inst / adev->vcn.num_inst_per_aid;
} }
DRM_DEV_INFO(adev->dev, "VCN decode is enabled in VM mode\n"); DRM_DEV_INFO(adev->dev, "VCN decode is enabled in VM mode\n");
} }
...@@ -1188,9 +1225,8 @@ static bool vcn_v4_0_3_is_idle(void *handle) ...@@ -1188,9 +1225,8 @@ static bool vcn_v4_0_3_is_idle(void *handle)
int i, ret = 1; int i, ret = 1;
for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
if (adev->vcn.harvest_config & (1 << i)) ret &= (RREG32_SOC15(VCN, GET_INST(VCN, i), regUVD_STATUS) ==
continue; UVD_STATUS__IDLE);
ret &= (RREG32_SOC15(VCN, i, regUVD_STATUS) == UVD_STATUS__IDLE);
} }
return ret; return ret;
...@@ -1209,10 +1245,8 @@ static int vcn_v4_0_3_wait_for_idle(void *handle) ...@@ -1209,10 +1245,8 @@ static int vcn_v4_0_3_wait_for_idle(void *handle)
int i, ret = 0; int i, ret = 0;
for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
if (adev->vcn.harvest_config & (1 << i)) ret = SOC15_WAIT_ON_RREG(VCN, GET_INST(VCN, i), regUVD_STATUS,
continue; UVD_STATUS__IDLE, UVD_STATUS__IDLE);
ret = SOC15_WAIT_ON_RREG(VCN, i, regUVD_STATUS, UVD_STATUS__IDLE,
UVD_STATUS__IDLE);
if (ret) if (ret)
return ret; return ret;
} }
...@@ -1235,10 +1269,9 @@ static int vcn_v4_0_3_set_clockgating_state(void *handle, ...@@ -1235,10 +1269,9 @@ static int vcn_v4_0_3_set_clockgating_state(void *handle,
int i; int i;
for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
if (adev->vcn.harvest_config & (1 << i))
continue;
if (enable) { if (enable) {
if (RREG32_SOC15(VCN, i, regUVD_STATUS) != UVD_STATUS__IDLE) if (RREG32_SOC15(VCN, GET_INST(VCN, i),
regUVD_STATUS) != UVD_STATUS__IDLE)
return -EBUSY; return -EBUSY;
vcn_v4_0_3_enable_clock_gating(adev, i); vcn_v4_0_3_enable_clock_gating(adev, i);
} else { } else {
...@@ -1307,15 +1340,26 @@ static int vcn_v4_0_3_process_interrupt(struct amdgpu_device *adev, ...@@ -1307,15 +1340,26 @@ static int vcn_v4_0_3_process_interrupt(struct amdgpu_device *adev,
struct amdgpu_irq_src *source, struct amdgpu_irq_src *source,
struct amdgpu_iv_entry *entry) struct amdgpu_iv_entry *entry)
{ {
uint32_t i; uint32_t i, inst;
i = node_id_to_phys_map[entry->node_id]; i = node_id_to_phys_map[entry->node_id];
DRM_DEV_DEBUG(adev->dev, "IH: VCN TRAP\n"); DRM_DEV_DEBUG(adev->dev, "IH: VCN TRAP\n");
for (inst = 0; inst < adev->vcn.num_vcn_inst; ++inst)
if (adev->vcn.inst[inst].aid_id == i)
break;
if (inst >= adev->vcn.num_vcn_inst) {
dev_WARN_ONCE(adev->dev, 1,
"Interrupt received for unknown VCN instance %d",
entry->node_id);
return 0;
}
switch (entry->src_id) { switch (entry->src_id) {
case VCN_4_0__SRCID__UVD_ENC_GENERAL_PURPOSE: case VCN_4_0__SRCID__UVD_ENC_GENERAL_PURPOSE:
amdgpu_fence_process(&adev->vcn.inst[i].ring_enc[0]); amdgpu_fence_process(&adev->vcn.inst[inst].ring_enc[0]);
break; break;
default: default:
DRM_DEV_ERROR(adev->dev, "Unhandled interrupt: %d %d\n", DRM_DEV_ERROR(adev->dev, "Unhandled interrupt: %d %d\n",
...@@ -1343,9 +1387,6 @@ static void vcn_v4_0_3_set_irq_funcs(struct amdgpu_device *adev) ...@@ -1343,9 +1387,6 @@ static void vcn_v4_0_3_set_irq_funcs(struct amdgpu_device *adev)
int i; int i;
for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
if (adev->vcn.harvest_config & (1 << i))
continue;
adev->vcn.inst->irq.num_types++; adev->vcn.inst->irq.num_types++;
} }
adev->vcn.inst->irq.funcs = &vcn_v4_0_3_irq_funcs; adev->vcn.inst->irq.funcs = &vcn_v4_0_3_irq_funcs;
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7