RE: [PATCH] drm/amd/pm: bug fix for baco reset

[Quan, Evan]

[AMD Public Use]

Reviewed-by: Evan Quan <evan.q...@amd.com>

-----Original Message-----
From: amd-gfx <amd-gfx-boun...@lists.freedesktop.org> On Behalf Of Kenneth Feng
Sent: Thursday, March 18, 2021 12:07 PM
To: amd-gfx@lists.freedesktop.org
Cc: Feng, Kenneth <kenneth.f...@amd.com>
Subject: [PATCH] drm/amd/pm: bug fix for baco reset

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.f...@amd.com>
---
 .../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@lists.freedesktop.org
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&amp;data=04%7C01%7Cevan.quan%40amd.com%7C7fee5a2166be4f31286408d8e9c33f55%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637516372084403923%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&amp;sdata=KEFE61ILud7i%2F28piougWFNBBSYz%2B5mi3dBfHSghllw%3D&amp;reserved=0
_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx