Before fire a new host trap, check the host trap status.
Signed-off-by: James Zhu <James.Zhu@xxxxxxx>
---
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c | 35 +++++++++++++++++++
.../amd/include/asic_reg/gc/gc_9_0_offset.h | 2 ++
.../amd/include/asic_reg/gc/gc_9_0_sh_mask.h | 5 +++
3 files changed, 42 insertions(+)
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 adfe5e5585e5..43edd62df5fe 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -1144,6 +1144,35 @@ void kgd_gfx_v9_program_trap_handler_settings(struct amdgpu_device *adev,
kgd_gfx_v9_unlock_srbm(adev, inst);
}
+static uint32_t kgd_aldebaran_get_hosttrap_status(struct amdgpu_device *adev)
+{
+ uint32_t sq_hosttrap_status = 0x0;
+ int i, j;
+
+ mutex_lock(&adev->grbm_idx_mutex);
+ for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
+ for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
+ amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff, 0);
+ sq_hosttrap_status = RREG32_SOC15(GC, 0, mmSQ_HOSTTRAP_STATUS);
+
+ if (sq_hosttrap_status & SQ_HOSTTRAP_STATUS__HTPENDING_OVERRIDE_MASK) {
+ WREG32_SOC15(GC, 0, mmSQ_HOSTTRAP_STATUS,
+ SQ_HOSTTRAP_STATUS__HTPENDING_OVERRIDE_MASK);
+ sq_hosttrap_status = 0x0;
+ continue;
+ }
+ if (sq_hosttrap_status)
+ goto out;
+ }
+ }
+
+out:
+ amdgpu_gfx_select_se_sh(adev, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0);
+ mutex_unlock(&adev->grbm_idx_mutex);
+
+ return sq_hosttrap_status;
+}
+
uint32_t kgd_gfx_v9_trigger_pc_sample_trap(struct amdgpu_device *adev,
uint32_t vmid,
uint32_t max_wave_slot,
@@ -1154,6 +1183,12 @@ uint32_t kgd_gfx_v9_trigger_pc_sample_trap(struct amdgpu_device *adev,
{
if (method == KFD_IOCTL_PCS_METHOD_HOSTTRAP) {
uint32_t value = 0;
+ uint32_t sq_hosttrap_status = 0x0;
+
+ sq_hosttrap_status = kgd_aldebaran_get_hosttrap_status(adev);
+ /* skip when last host trap request is still pending to complete */
+ if (sq_hosttrap_status)
+ return 0;
value = REG_SET_FIELD(value, SQ_CMD, CMD, SQ_IND_CMD_CMD_TRAP);
value = REG_SET_FIELD(value, SQ_CMD, MODE, SQ_IND_CMD_MODE_SINGLE);
diff --git a/drivers/gpu/drm/amd/include/asic_reg/gc/gc_9_0_offset.h b/drivers/gpu/drm/amd/include/asic_reg/gc/gc_9_0_offset.h
index 12d451e5475b..5b17d9066452 100644
--- a/drivers/gpu/drm/amd/include/asic_reg/gc/gc_9_0_offset.h
+++ b/drivers/gpu/drm/amd/include/asic_reg/gc/gc_9_0_offset.h
@@ -462,6 +462,8 @@
#define mmSQ_IND_DATA_BASE_IDX 0
#define mmSQ_CMD 0x037b
#define mmSQ_CMD_BASE_IDX 0
+#define mmSQ_HOSTTRAP_STATUS 0x0376
+#define mmSQ_HOSTTRAP_STATUS_BASE_IDX 0
#define mmSQ_TIME_HI 0x037c
#define mmSQ_TIME_HI_BASE_IDX 0
#define mmSQ_TIME_LO 0x037d
diff --git a/drivers/gpu/drm/amd/include/asic_reg/gc/gc_9_0_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/gc/gc_9_0_sh_mask.h
index efc16ddf274a..3dfe4ab31421 100644
--- a/drivers/gpu/drm/amd/include/asic_reg/gc/gc_9_0_sh_mask.h
+++ b/drivers/gpu/drm/amd/include/asic_reg/gc/gc_9_0_sh_mask.h
@@ -2616,6 +2616,11 @@
//SQ_CMD_TIMESTAMP
#define SQ_CMD_TIMESTAMP__TIMESTAMP__SHIFT 0x0
#define SQ_CMD_TIMESTAMP__TIMESTAMP_MASK 0x000000FFL
+//SQ_HOSTTRAP_STATUS
+#define SQ_HOSTTRAP_STATUS__HTPENDINGCOUNT__SHIFT 0x0
+#define SQ_HOSTTRAP_STATUS__HTPENDING_OVERRIDE__SHIFT 0x8
+#define SQ_HOSTTRAP_STATUS__HTPENDINGCOUNT_MASK 0x000000FFL
+#define SQ_HOSTTRAP_STATUS__HTPENDING_OVERRIDE_MASK 0x00000100L
//SQ_IND_INDEX
#define SQ_IND_INDEX__WAVE_ID__SHIFT 0x0
#define SQ_IND_INDEX__SIMD_ID__SHIFT 0x4
--
2.25.1
