On vega20, rocm-smi gets the wrong gfx voltage after baco reset.
This can be reproduced as below.
:~$ rocm-smi --showvoltage
GPU[0] : Voltage (mV): 737
:~$ rocm-smi -d0 --gpureset
GPU[0] : GPU reset was successful
:~$ rocm-smi --showvoltage
GPU[0] : Voltage (mV): 1550
Root cause: telemetry is disabled in the asic_init after baco exit.
This fix targets to re-enable telemetry then all the power and voltage
info can be fetched correctly, mp1 firmware also depends on this setting
for dpm arbitration.
Signed-off-by: Kenneth Feng <kenneth.feng@xxxxxxx>
---
.../drm/amd/pm/powerplay/hwmgr/vega20_baco.c | 17 +++++++++++++++++
.../drm/amd/pm/powerplay/hwmgr/vega20_baco.h | 2 +-
.../drm/amd/pm/powerplay/hwmgr/vega20_hwmgr.c | 1 +
3 files changed, 19 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_baco.c b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_baco.c
index 2a28c9df15a0..bb58097a925c 100644
--- a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_baco.c
+++ b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_baco.c
@@ -28,9 +28,16 @@
#include "vega20_ppsmc.h"
#include "vega20_baco.h"
#include "vega20_smumgr.h"
+#include "smuio/smuio_9_0_offset.h"
+#include "smuio/smuio_9_0_sh_mask.h"
#include "amdgpu_ras.h"
+#define mmSMUSVI0_TFN 0x2
+#define SMUSVI0_TFN___PLANE0_MASK 0x1
+#define SMUSVI0_TFN___PLANE1_MASK 0x2
+#define mmSMUSVI0_TFN_BASE_IDX 0
+
static const struct soc15_baco_cmd_entry clean_baco_tbl[] =
{
{CMD_WRITE, SOC15_REG_ENTRY(NBIF, 0, mmBIOS_SCRATCH_6), 0, 0, 0, 0},
@@ -120,3 +127,13 @@ int vega20_baco_apply_vdci_flush_workaround(struct pp_hwmgr *hwmgr)
return smum_send_msg_to_smc(hwmgr, PPSMC_MSG_BacoWorkAroundFlushVDCI, NULL);
}
+
+void vega20_baco_override_telemetry_parameters(struct pp_hwmgr *hwmgr)
+{
+ struct amdgpu_device *adev = (struct amdgpu_device *)(hwmgr->adev);
+ uint32_t data = RREG32_SOC15(SMUIO, 0, mmSMUSVI0_TFN);
+
+ data &= (~SMUSVI0_TFN___PLANE0_MASK);
+ data |= SMUSVI0_TFN___PLANE1_MASK;
+ WREG32_SOC15(SMUIO, 0, mmSMUSVI0_TFN, data);
+}
diff --git a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_baco.h b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_baco.h
index f06471e712dc..9ca39569ba0e 100644
--- a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_baco.h
+++ b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_baco.h
@@ -29,5 +29,5 @@ extern int vega20_baco_get_capability(struct pp_hwmgr *hwmgr, bool *cap);
extern int vega20_baco_get_state(struct pp_hwmgr *hwmgr, enum BACO_STATE *state);
extern int vega20_baco_set_state(struct pp_hwmgr *hwmgr, enum BACO_STATE state);
extern int vega20_baco_apply_vdci_flush_workaround(struct pp_hwmgr *hwmgr);
-
+extern void vega20_baco_override_telemetry_parameters(struct pp_hwmgr *hwmgr);
#endif
diff --git a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_hwmgr.c b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_hwmgr.c
index 213c9c6b4462..12830a8dd923 100644
--- a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_hwmgr.c
+++ b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_hwmgr.c
@@ -503,6 +503,7 @@ static int vega20_setup_asic_task(struct pp_hwmgr *hwmgr)
ret = vega20_baco_apply_vdci_flush_workaround(hwmgr);
if (ret)
pr_err("Failed to apply vega20 baco workaround!\n");
+ vega20_baco_override_telemetry_parameters(hwmgr);
}
return ret;
--
2.17.1
_______________________________________________
amd-gfx mailing list
amd-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
