WREG32/RREG32_RLC should specify the instance so the correct XCC's RLCG
interface can be used.
Signed-off-by: Victor Lu <victorchengchi.lu@xxxxxxx>
---
.../drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c | 4 +--
.../drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c | 25 +++++++--------
.../drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c | 4 +--
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c | 32 +++++++++----------
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 6 ++--
drivers/gpu/drm/amd/amdgpu/soc15.c | 2 +-
drivers/gpu/drm/amd/amdgpu/soc15_common.h | 8 ++---
7 files changed, 40 insertions(+), 41 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
index e2fed6edbdd0..4b6007de24ee 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
@@ -154,11 +154,11 @@ static uint32_t kgd_gfx_aldebaran_set_address_watch(
WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_H) +
(watch_id * TCP_WATCH_STRIDE)),
- watch_address_high);
+ watch_address_high, 0);
WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_L) +
(watch_id * TCP_WATCH_STRIDE)),
- watch_address_low);
+ watch_address_low, 0);
return watch_address_cntl;
}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c
index 490c8f5ddb60..218226c10b93 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c
@@ -300,14 +300,14 @@ static int kgd_gfx_v9_4_3_hqd_load(struct amdgpu_device *adev, void *mqd,
hqd_end = SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_HQD_AQL_DISPATCH_ID_HI);
for (reg = hqd_base; reg <= hqd_end; reg++)
- WREG32_RLC(reg, mqd_hqd[reg - hqd_base]);
+ WREG32_RLC(reg, mqd_hqd[reg - hqd_base], inst);
/* Activate doorbell logic before triggering WPTR poll. */
data = REG_SET_FIELD(m->cp_hqd_pq_doorbell_control,
CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1);
WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_HQD_PQ_DOORBELL_CONTROL),
- data);
+ data, inst);
if (wptr) {
/* Don't read wptr with get_user because the user
@@ -337,26 +337,25 @@ static int kgd_gfx_v9_4_3_hqd_load(struct amdgpu_device *adev, void *mqd,
guessed_wptr += (uint64_t)m->cp_hqd_pq_wptr_hi << 32;
WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_HQD_PQ_WPTR_LO),
- lower_32_bits(guessed_wptr));
+ lower_32_bits(guessed_wptr), inst);
WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_HQD_PQ_WPTR_HI),
- upper_32_bits(guessed_wptr));
+ upper_32_bits(guessed_wptr), inst);
WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_HQD_PQ_WPTR_POLL_ADDR),
- lower_32_bits((uintptr_t)wptr));
+ lower_32_bits((uintptr_t)wptr), inst);
WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst),
regCP_HQD_PQ_WPTR_POLL_ADDR_HI),
- upper_32_bits((uintptr_t)wptr));
- WREG32(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_PQ_WPTR_POLL_CNTL1),
- (uint32_t)kgd_gfx_v9_get_queue_mask(adev, pipe_id,
- queue_id));
+ upper_32_bits((uintptr_t)wptr), inst);
+ WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_PQ_WPTR_POLL_CNTL1),
+ (uint32_t)kgd_gfx_v9_get_queue_mask(adev, pipe_id, queue_id), inst);
}
/* Start the EOP fetcher */
WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_HQD_EOP_RPTR),
REG_SET_FIELD(m->cp_hqd_eop_rptr,
- CP_HQD_EOP_RPTR, INIT_FETCHER, 1));
+ CP_HQD_EOP_RPTR, INIT_FETCHER, 1), inst);
data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);
- WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_HQD_ACTIVE), data);
+ WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_HQD_ACTIVE), data, inst);
kgd_gfx_v9_release_queue(adev, inst);
@@ -497,12 +496,12 @@ static uint32_t kgd_gfx_v9_4_3_set_address_watch(
WREG32_RLC((SOC15_REG_OFFSET(GC, GET_INST(GC, inst),
regTCP_WATCH0_ADDR_H) +
(watch_id * TCP_WATCH_STRIDE)),
- watch_address_high);
+ watch_address_high, inst);
WREG32_RLC((SOC15_REG_OFFSET(GC, GET_INST(GC, inst),
regTCP_WATCH0_ADDR_L) +
(watch_id * TCP_WATCH_STRIDE)),
- watch_address_low);
+ watch_address_low, inst);
return watch_address_cntl;
}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c
index d67d003bada2..a89902d6dc35 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c
@@ -771,11 +771,11 @@ static uint32_t kgd_gfx_v11_set_address_watch(struct amdgpu_device *adev,
WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_H) +
(watch_id * TCP_WATCH_STRIDE)),
- watch_address_high);
+ watch_address_high, 0);
WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_L) +
(watch_id * TCP_WATCH_STRIDE)),
- watch_address_low);
+ watch_address_low, 0);
return watch_address_cntl;
}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
index 28963726bc97..93d6719e89dd 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -91,8 +91,8 @@ void kgd_gfx_v9_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmi
{
kgd_gfx_v9_lock_srbm(adev, 0, 0, 0, vmid, inst);
- WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmSH_MEM_CONFIG), sh_mem_config);
- WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmSH_MEM_BASES), sh_mem_bases);
+ WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmSH_MEM_CONFIG), sh_mem_config, inst);
+ WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmSH_MEM_BASES), sh_mem_bases, inst);
/* APE1 no longer exists on GFX9 */
kgd_gfx_v9_unlock_srbm(adev, inst);
@@ -239,14 +239,14 @@ int kgd_gfx_v9_hqd_load(struct amdgpu_device *adev, void *mqd,
for (reg = hqd_base;
reg <= SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_HQD_PQ_WPTR_HI); reg++)
- WREG32_RLC(reg, mqd_hqd[reg - hqd_base]);
+ WREG32_RLC(reg, mqd_hqd[reg - hqd_base], inst);
/* Activate doorbell logic before triggering WPTR poll. */
data = REG_SET_FIELD(m->cp_hqd_pq_doorbell_control,
CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1);
WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_HQD_PQ_DOORBELL_CONTROL),
- data);
+ data, inst);
if (wptr) {
/* Don't read wptr with get_user because the user
@@ -276,13 +276,13 @@ int kgd_gfx_v9_hqd_load(struct amdgpu_device *adev, void *mqd,
guessed_wptr += (uint64_t)m->cp_hqd_pq_wptr_hi << 32;
WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_HQD_PQ_WPTR_LO),
- lower_32_bits(guessed_wptr));
+ lower_32_bits(guessed_wptr), inst);
WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_HQD_PQ_WPTR_HI),
- upper_32_bits(guessed_wptr));
+ upper_32_bits(guessed_wptr), inst);
WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_HQD_PQ_WPTR_POLL_ADDR),
- lower_32_bits((uintptr_t)wptr));
+ lower_32_bits((uintptr_t)wptr), inst);
WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_HQD_PQ_WPTR_POLL_ADDR_HI),
- upper_32_bits((uintptr_t)wptr));
+ upper_32_bits((uintptr_t)wptr), inst);
WREG32_SOC15(GC, GET_INST(GC, inst), mmCP_PQ_WPTR_POLL_CNTL1,
(uint32_t)kgd_gfx_v9_get_queue_mask(adev, pipe_id, queue_id));
}
@@ -290,10 +290,10 @@ int kgd_gfx_v9_hqd_load(struct amdgpu_device *adev, void *mqd,
/* Start the EOP fetcher */
WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_HQD_EOP_RPTR),
REG_SET_FIELD(m->cp_hqd_eop_rptr,
- CP_HQD_EOP_RPTR, INIT_FETCHER, 1));
+ CP_HQD_EOP_RPTR, INIT_FETCHER, 1), inst);
data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);
- WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_HQD_ACTIVE), data);
+ WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_HQD_ACTIVE), data, inst);
kgd_gfx_v9_release_queue(adev, inst);
@@ -556,7 +556,7 @@ int kgd_gfx_v9_hqd_destroy(struct amdgpu_device *adev, void *mqd,
break;
}
- WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_HQD_DEQUEUE_REQUEST), type);
+ WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_HQD_DEQUEUE_REQUEST), type, inst);
end_jiffies = (utimeout * HZ / 1000) + jiffies;
while (true) {
@@ -855,15 +855,15 @@ uint32_t kgd_gfx_v9_set_address_watch(struct amdgpu_device *adev,
WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
(watch_id * TCP_WATCH_STRIDE)),
- watch_address_cntl);
+ watch_address_cntl, 0);
WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_ADDR_H) +
(watch_id * TCP_WATCH_STRIDE)),
- watch_address_high);
+ watch_address_high, 0);
WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_ADDR_L) +
(watch_id * TCP_WATCH_STRIDE)),
- watch_address_low);
+ watch_address_low, 0);
/* Enable the watch point */
watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
@@ -873,7 +873,7 @@ uint32_t kgd_gfx_v9_set_address_watch(struct amdgpu_device *adev,
WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
(watch_id * TCP_WATCH_STRIDE)),
- watch_address_cntl);
+ watch_address_cntl, 0);
return 0;
}
@@ -887,7 +887,7 @@ uint32_t kgd_gfx_v9_clear_address_watch(struct amdgpu_device *adev,
WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
(watch_id * TCP_WATCH_STRIDE)),
- watch_address_cntl);
+ watch_address_cntl, 0);
return 0;
}
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index 458faf657042..5ea1497b8300 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
@@ -2485,11 +2485,11 @@ static void gfx_v9_0_init_csb(struct amdgpu_device *adev)
adev->gfx.rlc.funcs->get_csb_buffer(adev, adev->gfx.rlc.cs_ptr);
/* csib */
WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmRLC_CSIB_ADDR_HI),
- adev->gfx.rlc.clear_state_gpu_addr >> 32);
+ adev->gfx.rlc.clear_state_gpu_addr >> 32, 0);
WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmRLC_CSIB_ADDR_LO),
- adev->gfx.rlc.clear_state_gpu_addr & 0xfffffffc);
+ adev->gfx.rlc.clear_state_gpu_addr & 0xfffffffc, 0);
WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmRLC_CSIB_LENGTH),
- adev->gfx.rlc.clear_state_size);
+ adev->gfx.rlc.clear_state_size, 0);
}
static void gfx_v9_1_parse_ind_reg_list(int *register_list_format,
diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c
index 158e611474ca..cca7bbc9fae9 100644
--- a/drivers/gpu/drm/amd/amdgpu/soc15.c
+++ b/drivers/gpu/drm/amd/amdgpu/soc15.c
@@ -469,7 +469,7 @@ void soc15_program_register_sequence(struct amdgpu_device *adev,
reg == SOC15_REG_OFFSET(GC, 0, mmPA_SC_ENHANCE) ||
reg == SOC15_REG_OFFSET(GC, 0, mmPA_SC_ENHANCE_1) ||
reg == SOC15_REG_OFFSET(GC, 0, mmSH_MEM_CONFIG))
- WREG32_RLC(reg, tmp);
+ WREG32_RLC(reg, tmp, 0);
else
(entry->hwip == GC_HWIP) ?
WREG32_SOC15_IP(GC, reg, tmp) : WREG32(reg, tmp);
diff --git a/drivers/gpu/drm/amd/amdgpu/soc15_common.h b/drivers/gpu/drm/amd/amdgpu/soc15_common.h
index c75e9cd5c98b..148eb2603b9a 100644
--- a/drivers/gpu/drm/amd/amdgpu/soc15_common.h
+++ b/drivers/gpu/drm/amd/amdgpu/soc15_common.h
@@ -107,8 +107,8 @@
(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + (reg) + (offset)), \
#reg, expected_value, mask)
-#define WREG32_RLC(reg, value) \
- __WREG32_SOC15_RLC__(reg, value, AMDGPU_REGS_RLC, GC_HWIP, 0)
+#define WREG32_RLC(reg, value, inst) \
+ __WREG32_SOC15_RLC__(reg, value, AMDGPU_REGS_RLC, GC_HWIP, inst)
#define WREG32_RLC_EX(prefix, reg, value, inst) \
do { \
@@ -139,8 +139,8 @@
__WREG32_SOC15_RLC__((adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg), value, AMDGPU_REGS_RLC, GC_HWIP, inst)
/* for GC only */
-#define RREG32_RLC(reg) \
- __RREG32_SOC15_RLC__(reg, AMDGPU_REGS_RLC, GC_HWIP)
+#define RREG32_RLC(reg, inst) \
+ __RREG32_SOC15_RLC__(reg, AMDGPU_REGS_RLC, GC_HWIP, inst)
#define WREG32_RLC_NO_KIQ(reg, value, hwip) \
__WREG32_SOC15_RLC__(reg, value, AMDGPU_REGS_NO_KIQ | AMDGPU_REGS_RLC, hwip, 0)
--
2.34.1
