summaryrefslogtreecommitdiffstats
path: root/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c')
-rw-r--r--drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c83
1 files changed, 41 insertions, 42 deletions
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 ef3d93b995b2..d5af41143d12 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -225,8 +225,8 @@ static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
lock_srbm(kgd, 0, 0, 0, vmid);
- WREG32(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_CONFIG), sh_mem_config);
- WREG32(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_BASES), sh_mem_bases);
+ WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_CONFIG), sh_mem_config);
+ WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_BASES), sh_mem_bases);
/* APE1 no longer exists on GFX9 */
unlock_srbm(kgd);
@@ -369,7 +369,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
value = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CP_SCHEDULERS));
value = REG_SET_FIELD(value, RLC_CP_SCHEDULERS, scheduler1,
((mec << 5) | (pipe << 3) | queue_id | 0x80));
- WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CP_SCHEDULERS), value);
+ WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmRLC_CP_SCHEDULERS), value);
}
/* HQD registers extend from CP_MQD_BASE_ADDR to CP_HQD_EOP_WPTR_MEM. */
@@ -378,13 +378,13 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
for (reg = hqd_base;
reg <= SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI); reg++)
- WREG32(reg, mqd_hqd[reg - hqd_base]);
+ WREG32_RLC(reg, mqd_hqd[reg - hqd_base]);
/* 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(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL), data);
+ WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL), data);
if (wptr) {
/* Don't read wptr with get_user because the user
@@ -413,25 +413,25 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
guessed_wptr += m->cp_hqd_pq_wptr_lo & ~(queue_size - 1);
guessed_wptr += (uint64_t)m->cp_hqd_pq_wptr_hi << 32;
- WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_LO),
+ WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_LO),
lower_32_bits(guessed_wptr));
- WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI),
+ WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI),
upper_32_bits(guessed_wptr));
- WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR),
+ WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR),
lower_32_bits((uintptr_t)wptr));
- WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR_HI),
+ WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR_HI),
upper_32_bits((uintptr_t)wptr));
WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_PQ_WPTR_POLL_CNTL1),
get_queue_mask(adev, pipe_id, queue_id));
}
/* Start the EOP fetcher */
- WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_EOP_RPTR),
+ WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_EOP_RPTR),
REG_SET_FIELD(m->cp_hqd_eop_rptr,
CP_HQD_EOP_RPTR, INIT_FETCHER, 1));
data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);
- WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_ACTIVE), data);
+ WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_ACTIVE), data);
release_queue(kgd);
@@ -633,7 +633,7 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd,
acquire_queue(kgd, pipe_id, queue_id);
if (m->cp_hqd_vmid == 0)
- WREG32_FIELD15(GC, 0, RLC_CP_SCHEDULERS, scheduler1, 0);
+ WREG32_FIELD15_RLC(GC, 0, RLC_CP_SCHEDULERS, scheduler1, 0);
switch (reset_type) {
case KFD_PREEMPT_TYPE_WAVEFRONT_DRAIN:
@@ -647,7 +647,7 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd,
break;
}
- WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_DEQUEUE_REQUEST), type);
+ WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_DEQUEUE_REQUEST), type);
end_jiffies = (utimeout * HZ / 1000) + jiffies;
while (true) {
@@ -726,29 +726,8 @@ static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd,
return reg & ATC_VMID0_PASID_MAPPING__PASID_MASK;
}
-static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t vmid)
-{
- struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
-
- /* Use legacy mode tlb invalidation.
- *
- * Currently on Raven the code below is broken for anything but
- * legacy mode due to a MMHUB power gating problem. A workaround
- * is for MMHUB to wait until the condition PER_VMID_INVALIDATE_REQ
- * == PER_VMID_INVALIDATE_ACK instead of simply waiting for the ack
- * bit.
- *
- * TODO 1: agree on the right set of invalidation registers for
- * KFD use. Use the last one for now. Invalidate both GC and
- * MMHUB.
- *
- * TODO 2: support range-based invalidation, requires kfg2kgd
- * interface change
- */
- amdgpu_gmc_flush_gpu_tlb(adev, vmid, 0);
-}
-
-static int invalidate_tlbs_with_kiq(struct amdgpu_device *adev, uint16_t pasid)
+static int invalidate_tlbs_with_kiq(struct amdgpu_device *adev, uint16_t pasid,
+ uint32_t flush_type)
{
signed long r;
uint32_t seq;
@@ -761,7 +740,7 @@ static int invalidate_tlbs_with_kiq(struct amdgpu_device *adev, uint16_t pasid)
PACKET3_INVALIDATE_TLBS_DST_SEL(1) |
PACKET3_INVALIDATE_TLBS_ALL_HUB(1) |
PACKET3_INVALIDATE_TLBS_PASID(pasid) |
- PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(0)); /* legacy */
+ PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(flush_type));
amdgpu_fence_emit_polling(ring, &seq);
amdgpu_ring_commit(ring);
spin_unlock(&adev->gfx.kiq.ring_lock);
@@ -780,12 +759,16 @@ static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid)
struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
int vmid;
struct amdgpu_ring *ring = &adev->gfx.kiq.ring;
+ uint32_t flush_type = 0;
if (adev->in_gpu_reset)
return -EIO;
+ if (adev->gmc.xgmi.num_physical_nodes &&
+ adev->asic_type == CHIP_VEGA20)
+ flush_type = 2;
if (ring->sched.ready)
- return invalidate_tlbs_with_kiq(adev, pasid);
+ return invalidate_tlbs_with_kiq(adev, pasid, flush_type);
for (vmid = 0; vmid < 16; vmid++) {
if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid))
@@ -793,7 +776,8 @@ static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid)
if (get_atc_vmid_pasid_mapping_valid(kgd, vmid)) {
if (get_atc_vmid_pasid_mapping_pasid(kgd, vmid)
== pasid) {
- write_vmid_invalidate_request(kgd, vmid);
+ amdgpu_gmc_flush_gpu_tlb(adev, vmid,
+ flush_type);
break;
}
}
@@ -811,7 +795,22 @@ static int invalidate_tlbs_vmid(struct kgd_dev *kgd, uint16_t vmid)
return 0;
}
- write_vmid_invalidate_request(kgd, vmid);
+ /* Use legacy mode tlb invalidation.
+ *
+ * Currently on Raven the code below is broken for anything but
+ * legacy mode due to a MMHUB power gating problem. A workaround
+ * is for MMHUB to wait until the condition PER_VMID_INVALIDATE_REQ
+ * == PER_VMID_INVALIDATE_ACK instead of simply waiting for the ack
+ * bit.
+ *
+ * TODO 1: agree on the right set of invalidation registers for
+ * KFD use. Use the last one for now. Invalidate both GC and
+ * MMHUB.
+ *
+ * TODO 2: support range-based invalidation, requires kfg2kgd
+ * interface change
+ */
+ amdgpu_gmc_flush_gpu_tlb(adev, vmid, 0);
return 0;
}
@@ -838,7 +837,7 @@ static int kgd_wave_control_execute(struct kgd_dev *kgd,
mutex_lock(&adev->grbm_idx_mutex);
- WREG32(SOC15_REG_OFFSET(GC, 0, mmGRBM_GFX_INDEX), gfx_index_val);
+ WREG32_SOC15_RLC_SHADOW(GC, 0, mmGRBM_GFX_INDEX, gfx_index_val);
WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_CMD), sq_cmd);
data = REG_SET_FIELD(data, GRBM_GFX_INDEX,
@@ -848,7 +847,7 @@ static int kgd_wave_control_execute(struct kgd_dev *kgd,
data = REG_SET_FIELD(data, GRBM_GFX_INDEX,
SE_BROADCAST_WRITES, 1);
- WREG32(SOC15_REG_OFFSET(GC, 0, mmGRBM_GFX_INDEX), data);
+ WREG32_SOC15_RLC_SHADOW(GC, 0, mmGRBM_GFX_INDEX, data);
mutex_unlock(&adev->grbm_idx_mutex);
return 0;
generated by cgit v1.2.3-55-g7522 (git 2.39.0) at 2025-05-01 14:36:52 +0200