drm/amdgpu/mes12: configure two pipes hardware resources

Configure two pipes with different hardware resources.

Signed-off-by: Jack Xiao <Jack.Xiao@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This commit is contained in:
Jack Xiao 2024-08-07 14:49:30 +08:00 committed by Alex Deucher
parent aa539da8af
commit ea5d6db17a
4 changed files with 81 additions and 64 deletions

View File

@ -165,36 +165,38 @@ int amdgpu_mes_init(struct amdgpu_device *adev)
adev->mes.sdma_hqd_mask[i] = 0xfc;
}
r = amdgpu_device_wb_get(adev, &adev->mes.sch_ctx_offs);
if (r) {
dev_err(adev->dev,
"(%d) ring trail_fence_offs wb alloc failed\n", r);
goto error_ids;
}
adev->mes.sch_ctx_gpu_addr =
adev->wb.gpu_addr + (adev->mes.sch_ctx_offs * 4);
adev->mes.sch_ctx_ptr =
(uint64_t *)&adev->wb.wb[adev->mes.sch_ctx_offs];
for (i = 0; i < AMDGPU_MAX_MES_PIPES; i++) {
r = amdgpu_device_wb_get(adev, &adev->mes.sch_ctx_offs[i]);
if (r) {
dev_err(adev->dev,
"(%d) ring trail_fence_offs wb alloc failed\n",
r);
goto error;
}
adev->mes.sch_ctx_gpu_addr[i] =
adev->wb.gpu_addr + (adev->mes.sch_ctx_offs[i] * 4);
adev->mes.sch_ctx_ptr[i] =
(uint64_t *)&adev->wb.wb[adev->mes.sch_ctx_offs[i]];
r = amdgpu_device_wb_get(adev, &adev->mes.query_status_fence_offs);
if (r) {
amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs);
dev_err(adev->dev,
"(%d) query_status_fence_offs wb alloc failed\n", r);
goto error_ids;
r = amdgpu_device_wb_get(adev,
&adev->mes.query_status_fence_offs[i]);
if (r) {
dev_err(adev->dev,
"(%d) query_status_fence_offs wb alloc failed\n",
r);
goto error;
}
adev->mes.query_status_fence_gpu_addr[i] = adev->wb.gpu_addr +
(adev->mes.query_status_fence_offs[i] * 4);
adev->mes.query_status_fence_ptr[i] =
(uint64_t *)&adev->wb.wb[adev->mes.query_status_fence_offs[i]];
}
adev->mes.query_status_fence_gpu_addr =
adev->wb.gpu_addr + (adev->mes.query_status_fence_offs * 4);
adev->mes.query_status_fence_ptr =
(uint64_t *)&adev->wb.wb[adev->mes.query_status_fence_offs];
r = amdgpu_device_wb_get(adev, &adev->mes.read_val_offs);
if (r) {
amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs);
amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs);
dev_err(adev->dev,
"(%d) read_val_offs alloc failed\n", r);
goto error_ids;
goto error;
}
adev->mes.read_val_gpu_addr =
adev->wb.gpu_addr + (adev->mes.read_val_offs * 4);
@ -214,10 +216,16 @@ int amdgpu_mes_init(struct amdgpu_device *adev)
error_doorbell:
amdgpu_mes_doorbell_free(adev);
error:
amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs);
amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs);
amdgpu_device_wb_free(adev, adev->mes.read_val_offs);
error_ids:
for (i = 0; i < AMDGPU_MAX_MES_PIPES; i++) {
if (adev->mes.sch_ctx_ptr[i])
amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs[i]);
if (adev->mes.query_status_fence_ptr[i])
amdgpu_device_wb_free(adev,
adev->mes.query_status_fence_offs[i]);
}
if (adev->mes.read_val_ptr)
amdgpu_device_wb_free(adev, adev->mes.read_val_offs);
idr_destroy(&adev->mes.pasid_idr);
idr_destroy(&adev->mes.gang_id_idr);
idr_destroy(&adev->mes.queue_id_idr);
@ -228,13 +236,22 @@ error_ids:
void amdgpu_mes_fini(struct amdgpu_device *adev)
{
int i;
amdgpu_bo_free_kernel(&adev->mes.event_log_gpu_obj,
&adev->mes.event_log_gpu_addr,
&adev->mes.event_log_cpu_addr);
amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs);
amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs);
amdgpu_device_wb_free(adev, adev->mes.read_val_offs);
for (i = 0; i < AMDGPU_MAX_MES_PIPES; i++) {
if (adev->mes.sch_ctx_ptr[i])
amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs[i]);
if (adev->mes.query_status_fence_ptr[i])
amdgpu_device_wb_free(adev,
adev->mes.query_status_fence_offs[i]);
}
if (adev->mes.read_val_ptr)
amdgpu_device_wb_free(adev, adev->mes.read_val_offs);
amdgpu_mes_doorbell_free(adev);
idr_destroy(&adev->mes.pasid_idr);

View File

@ -112,12 +112,12 @@ struct amdgpu_mes {
uint32_t gfx_hqd_mask[AMDGPU_MES_MAX_GFX_PIPES];
uint32_t sdma_hqd_mask[AMDGPU_MES_MAX_SDMA_PIPES];
uint32_t aggregated_doorbells[AMDGPU_MES_PRIORITY_NUM_LEVELS];
uint32_t sch_ctx_offs;
uint64_t sch_ctx_gpu_addr;
uint64_t *sch_ctx_ptr;
uint32_t query_status_fence_offs;
uint64_t query_status_fence_gpu_addr;
uint64_t *query_status_fence_ptr;
uint32_t sch_ctx_offs[AMDGPU_MAX_MES_PIPES];
uint64_t sch_ctx_gpu_addr[AMDGPU_MAX_MES_PIPES];
uint64_t *sch_ctx_ptr[AMDGPU_MAX_MES_PIPES];
uint32_t query_status_fence_offs[AMDGPU_MAX_MES_PIPES];
uint64_t query_status_fence_gpu_addr[AMDGPU_MAX_MES_PIPES];
uint64_t *query_status_fence_ptr[AMDGPU_MAX_MES_PIPES];
uint32_t read_val_offs;
uint64_t read_val_gpu_addr;
uint32_t *read_val_ptr;

View File

@ -522,9 +522,9 @@ static int mes_v11_0_set_hw_resources(struct amdgpu_mes *mes)
mes_set_hw_res_pkt.vmid_mask_gfxhub = mes->vmid_mask_gfxhub;
mes_set_hw_res_pkt.gds_size = adev->gds.gds_size;
mes_set_hw_res_pkt.paging_vmid = 0;
mes_set_hw_res_pkt.g_sch_ctx_gpu_mc_ptr = mes->sch_ctx_gpu_addr;
mes_set_hw_res_pkt.g_sch_ctx_gpu_mc_ptr = mes->sch_ctx_gpu_addr[0];
mes_set_hw_res_pkt.query_status_fence_gpu_mc_ptr =
mes->query_status_fence_gpu_addr;
mes->query_status_fence_gpu_addr[0];
for (i = 0; i < MAX_COMPUTE_PIPES; i++)
mes_set_hw_res_pkt.compute_hqd_mask[i] =
@ -1243,9 +1243,6 @@ static int mes_v11_0_sw_fini(void *handle)
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
int pipe;
amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs);
amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs);
for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) {
kfree(adev->mes.mqd_backup[pipe]);

View File

@ -542,27 +542,33 @@ static int mes_v12_0_set_hw_resources(struct amdgpu_mes *mes, int pipe)
mes_set_hw_res_pkt.header.opcode = MES_SCH_API_SET_HW_RSRC;
mes_set_hw_res_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
mes_set_hw_res_pkt.vmid_mask_mmhub = mes->vmid_mask_mmhub;
mes_set_hw_res_pkt.vmid_mask_gfxhub = mes->vmid_mask_gfxhub;
mes_set_hw_res_pkt.gds_size = adev->gds.gds_size;
mes_set_hw_res_pkt.paging_vmid = 0;
mes_set_hw_res_pkt.g_sch_ctx_gpu_mc_ptr = mes->sch_ctx_gpu_addr;
if (pipe == AMDGPU_MES_SCHED_PIPE) {
mes_set_hw_res_pkt.vmid_mask_mmhub = mes->vmid_mask_mmhub;
mes_set_hw_res_pkt.vmid_mask_gfxhub = mes->vmid_mask_gfxhub;
mes_set_hw_res_pkt.gds_size = adev->gds.gds_size;
mes_set_hw_res_pkt.paging_vmid = 0;
for (i = 0; i < MAX_COMPUTE_PIPES; i++)
mes_set_hw_res_pkt.compute_hqd_mask[i] =
mes->compute_hqd_mask[i];
for (i = 0; i < MAX_GFX_PIPES; i++)
mes_set_hw_res_pkt.gfx_hqd_mask[i] =
mes->gfx_hqd_mask[i];
for (i = 0; i < MAX_SDMA_PIPES; i++)
mes_set_hw_res_pkt.sdma_hqd_mask[i] =
mes->sdma_hqd_mask[i];
for (i = 0; i < AMD_PRIORITY_NUM_LEVELS; i++)
mes_set_hw_res_pkt.aggregated_doorbells[i] =
mes->aggregated_doorbells[i];
}
mes_set_hw_res_pkt.g_sch_ctx_gpu_mc_ptr =
mes->sch_ctx_gpu_addr[pipe];
mes_set_hw_res_pkt.query_status_fence_gpu_mc_ptr =
mes->query_status_fence_gpu_addr;
for (i = 0; i < MAX_COMPUTE_PIPES; i++)
mes_set_hw_res_pkt.compute_hqd_mask[i] =
mes->compute_hqd_mask[i];
for (i = 0; i < MAX_GFX_PIPES; i++)
mes_set_hw_res_pkt.gfx_hqd_mask[i] = mes->gfx_hqd_mask[i];
for (i = 0; i < MAX_SDMA_PIPES; i++)
mes_set_hw_res_pkt.sdma_hqd_mask[i] = mes->sdma_hqd_mask[i];
for (i = 0; i < AMD_PRIORITY_NUM_LEVELS; i++)
mes_set_hw_res_pkt.aggregated_doorbells[i] =
mes->aggregated_doorbells[i];
mes->query_status_fence_gpu_addr[pipe];
for (i = 0; i < 5; i++) {
mes_set_hw_res_pkt.gc_base[i] = adev->reg_offset[GC_HWIP][0][i];
@ -1331,9 +1337,6 @@ static int mes_v12_0_sw_fini(void *handle)
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
int pipe;
amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs);
amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs);
for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) {
kfree(adev->mes.mqd_backup[pipe]);