Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
drm/amd/pm: correct mem_busy_percent display due to calculation errors
PMFW may return invalid values due to internal calculation errors.
so, the kmd driver must validate and sanitize the returned values to
prevent issues caused by firmware calculation errors.
For example, values 0xfffe (-2) and 0xffff (-1) are treated
as invalid and clamped to 0.
this applies to devices with CAB (Cache As Buffer) functionality.
Closes: https://gitlab.freedesktop.org/drm/amd/-/work_items/4905
Signed-off-by: Yang Wang <kevinyang.wang@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
authored by
Yang Wang
and committed by
Alex Deucher
592713a8
95a599c8
+32
-15
+17
drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h
+17
drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h
···
2164
2164
smu_feature_list_clear_all(smu, SMU_FEATURE_LIST_ALLOWED);
2165
2165
}
2166
2166
2167
+
/*
2168
+
* smu_safe_u16_nn - Make u16 safe by filtering negative overflow errors
2169
+
* @val: Input u16 value, may contain invalid negative overflows
2170
+
*
2171
+
* Convert u16 to non-negative value. Cast to s16 to detect negative values
2172
+
* caused by calculation errors. Return 0 for negative errors, return
2173
+
* original value if valid.
2174
+
*
2175
+
* Return: Valid u16 value or 0
2176
+
*/
2177
+
static inline u16 smu_safe_u16_nn(u16 val)
2178
+
{
2179
+
s16 tmp = (s16)val;
2180
+
2181
+
return tmp < 0 ? 0 : val;
2182
+
}
2183
+
2167
2184
#endif
+5
-5
drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c
+5
-5
drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c
···
773
773
*value = metrics->AverageGfxclkFrequencyPreDs;
774
774
break;
775
775
case METRICS_AVERAGE_FCLK:
776
-
if (metrics->AverageUclkActivity <= SMU_13_0_0_BUSY_THRESHOLD)
776
+
if (smu_safe_u16_nn(metrics->AverageUclkActivity) <= SMU_13_0_0_BUSY_THRESHOLD)
777
777
*value = metrics->AverageFclkFrequencyPostDs;
778
778
else
779
779
*value = metrics->AverageFclkFrequencyPreDs;
780
780
break;
781
781
case METRICS_AVERAGE_UCLK:
782
-
if (metrics->AverageUclkActivity <= SMU_13_0_0_BUSY_THRESHOLD)
782
+
if (smu_safe_u16_nn(metrics->AverageUclkActivity) <= SMU_13_0_0_BUSY_THRESHOLD)
783
783
*value = metrics->AverageMemclkFrequencyPostDs;
784
784
else
785
785
*value = metrics->AverageMemclkFrequencyPreDs;
···
800
800
*value = metrics->AverageGfxActivity;
801
801
break;
802
802
case METRICS_AVERAGE_MEMACTIVITY:
803
-
*value = metrics->AverageUclkActivity;
803
+
*value = smu_safe_u16_nn(metrics->AverageUclkActivity);
804
804
break;
805
805
case METRICS_AVERAGE_VCNACTIVITY:
806
806
*value = max(metrics->Vcn0ActivityPercentage,
···
2085
2085
metrics->AvgTemperature[TEMP_VR_MEM1]);
2086
2086
2087
2087
gpu_metrics->average_gfx_activity = metrics->AverageGfxActivity;
2088
-
gpu_metrics->average_umc_activity = metrics->AverageUclkActivity;
2088
+
gpu_metrics->average_umc_activity = smu_safe_u16_nn(metrics->AverageUclkActivity);
2089
2089
gpu_metrics->average_mm_activity = max(metrics->Vcn0ActivityPercentage,
2090
2090
metrics->Vcn1ActivityPercentage);
2091
2091
···
2102
2102
else
2103
2103
gpu_metrics->average_gfxclk_frequency = metrics->AverageGfxclkFrequencyPreDs;
2104
2104
2105
-
if (metrics->AverageUclkActivity <= SMU_13_0_0_BUSY_THRESHOLD)
2105
+
if (smu_safe_u16_nn(metrics->AverageUclkActivity) <= SMU_13_0_0_BUSY_THRESHOLD)
2106
2106
gpu_metrics->average_uclk_frequency = metrics->AverageMemclkFrequencyPostDs;
2107
2107
else
2108
2108
gpu_metrics->average_uclk_frequency = metrics->AverageMemclkFrequencyPreDs;
+5
-5
drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c
+5
-5
drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c
···
783
783
*value = metrics->AverageGfxclkFrequencyPreDs;
784
784
break;
785
785
case METRICS_AVERAGE_FCLK:
786
-
if (metrics->AverageUclkActivity <= SMU_13_0_7_BUSY_THRESHOLD)
786
+
if (smu_safe_u16_nn(metrics->AverageUclkActivity) <= SMU_13_0_7_BUSY_THRESHOLD)
787
787
*value = metrics->AverageFclkFrequencyPostDs;
788
788
else
789
789
*value = metrics->AverageFclkFrequencyPreDs;
790
790
break;
791
791
case METRICS_AVERAGE_UCLK:
792
-
if (metrics->AverageUclkActivity <= SMU_13_0_7_BUSY_THRESHOLD)
792
+
if (smu_safe_u16_nn(metrics->AverageUclkActivity) <= SMU_13_0_7_BUSY_THRESHOLD)
793
793
*value = metrics->AverageMemclkFrequencyPostDs;
794
794
else
795
795
*value = metrics->AverageMemclkFrequencyPreDs;
···
814
814
*value = metrics->AverageGfxActivity;
815
815
break;
816
816
case METRICS_AVERAGE_MEMACTIVITY:
817
-
*value = metrics->AverageUclkActivity;
817
+
*value = smu_safe_u16_nn(metrics->AverageUclkActivity);
818
818
break;
819
819
case METRICS_AVERAGE_SOCKETPOWER:
820
820
*value = metrics->AverageSocketPower << 8;
···
2091
2091
metrics->AvgTemperature[TEMP_VR_MEM1]);
2092
2092
2093
2093
gpu_metrics->average_gfx_activity = metrics->AverageGfxActivity;
2094
-
gpu_metrics->average_umc_activity = metrics->AverageUclkActivity;
2094
+
gpu_metrics->average_umc_activity = smu_safe_u16_nn(metrics->AverageUclkActivity);
2095
2095
gpu_metrics->average_mm_activity = max(metrics->Vcn0ActivityPercentage,
2096
2096
metrics->Vcn1ActivityPercentage);
2097
2097
···
2104
2104
else
2105
2105
gpu_metrics->average_gfxclk_frequency = metrics->AverageGfxclkFrequencyPreDs;
2106
2106
2107
-
if (metrics->AverageUclkActivity <= SMU_13_0_7_BUSY_THRESHOLD)
2107
+
if (smu_safe_u16_nn(metrics->AverageUclkActivity) <= SMU_13_0_7_BUSY_THRESHOLD)
2108
2108
gpu_metrics->average_uclk_frequency = metrics->AverageMemclkFrequencyPostDs;
2109
2109
else
2110
2110
gpu_metrics->average_uclk_frequency = metrics->AverageMemclkFrequencyPreDs;
+5
-5
drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c
+5
-5
drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c
···
661
661
*value = metrics->AverageGfxclkFrequencyPreDs;
662
662
break;
663
663
case METRICS_AVERAGE_FCLK:
664
-
if (metrics->AverageUclkActivity <= SMU_14_0_2_BUSY_THRESHOLD)
664
+
if (smu_safe_u16_nn(metrics->AverageUclkActivity) <= SMU_14_0_2_BUSY_THRESHOLD)
665
665
*value = metrics->AverageFclkFrequencyPostDs;
666
666
else
667
667
*value = metrics->AverageFclkFrequencyPreDs;
668
668
break;
669
669
case METRICS_AVERAGE_UCLK:
670
-
if (metrics->AverageUclkActivity <= SMU_14_0_2_BUSY_THRESHOLD)
670
+
if (smu_safe_u16_nn(metrics->AverageUclkActivity) <= SMU_14_0_2_BUSY_THRESHOLD)
671
671
*value = metrics->AverageMemclkFrequencyPostDs;
672
672
else
673
673
*value = metrics->AverageMemclkFrequencyPreDs;
···
688
688
*value = metrics->AverageGfxActivity;
689
689
break;
690
690
case METRICS_AVERAGE_MEMACTIVITY:
691
-
*value = metrics->AverageUclkActivity;
691
+
*value = smu_safe_u16_nn(metrics->AverageUclkActivity);
692
692
break;
693
693
case METRICS_AVERAGE_VCNACTIVITY:
694
694
*value = max(metrics->AverageVcn0ActivityPercentage,
···
2147
2147
metrics->AvgTemperature[TEMP_VR_MEM1]);
2148
2148
2149
2149
gpu_metrics->average_gfx_activity = metrics->AverageGfxActivity;
2150
-
gpu_metrics->average_umc_activity = metrics->AverageUclkActivity;
2150
+
gpu_metrics->average_umc_activity = smu_safe_u16_nn(metrics->AverageUclkActivity);
2151
2151
gpu_metrics->average_mm_activity = max(metrics->AverageVcn0ActivityPercentage,
2152
2152
metrics->Vcn1ActivityPercentage);
2153
2153
···
2159
2159
else
2160
2160
gpu_metrics->average_gfxclk_frequency = metrics->AverageGfxclkFrequencyPreDs;
2161
2161
2162
-
if (metrics->AverageUclkActivity <= SMU_14_0_2_BUSY_THRESHOLD)
2162
+
if (smu_safe_u16_nn(metrics->AverageUclkActivity) <= SMU_14_0_2_BUSY_THRESHOLD)
2163
2163
gpu_metrics->average_uclk_frequency = metrics->AverageMemclkFrequencyPostDs;
2164
2164
else
2165
2165
gpu_metrics->average_uclk_frequency = metrics->AverageMemclkFrequencyPreDs;