Enable RAS feature for the gc of Arcturus
Subsystems (2)
cover Enable RAS feature for the gc of Arcturus
Refactor the ras related codes of vega20:
1. refine the security check for RAS functions.
2. abstract clearing edc counters to a separated function.
3. add ip prefix to ip related codes.
Implementation of RAS feature for Arcturus gfx:
1. add new register head files for gfx v9.4.1.
2. add codes to support querying of EDC counter and error injection.
Dennis Li (4):
drm/amdgpu: refine the security check for RAS functions drm/amdgpu: abstract EDC counter clear to a separated function drm/amdgpu: add EDC counter registers of gc for Arcturus drm/amdgpu: add RAS support for the gfx block of Arcturus
drivers/gpu/drm/amd/amdgpu/Makefile | 1 + drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 138 ++- drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c | 978 ++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h | 35 + .../amd/include/asic_reg/gc/gc_9_4_1_offset.h | 264 +++++ .../include/asic_reg/gc/gc_9_4_1_sh_mask.h | 748 ++++++++++++++ 6 files changed, 2128 insertions(+), 36 deletions(-) create mode 100644 drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c create mode 100644 drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h create mode 100644 drivers/gpu/drm/amd/include/asic_reg/gc/gc_9_4_1_offset.h create mode 100644 drivers/gpu/drm/amd/include/asic_reg/gc/gc_9_4_1_sh_mask.h
Signature · 6 lines
2.17.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 1/4 drm/amdgpu: refine the security check for RAS functions
To avoid calling RAS related functions when RAS feature isn't supported in hardware. Change to check supported features, instead of checking asic type.
v2: reuse amdgpu_ras_is_supported function, instead of introducing a new flag for hardware ras feature.
--- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +2 -2
@@ -5967,7 +5967,7 @@ static int gfx_v9_0_ras_error_inject(struct amdgpu_device *adev,int ret;struct ta_ras_trigger_error_input block_info = {0 }; -if (adev->asic_type !=CHIP_VEGA20 ) +if (!amdgpu_ras_is_supported (adev,AMDGPU_RAS_BLOCK__GFX ))return -EINVAL ;if (info->head.sub_block_index >=ARRAY_SIZE (ras_gfx_subblocks))
@@ -6218,7 +6218,7 @@ static int gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev,uint32_t i, j, k;uint32_t reg_value; -if (adev->asic_type !=CHIP_VEGA20 ) +if (!amdgpu_ras_is_supported (adev,AMDGPU_RAS_BLOCK__GFX ))return -EINVAL ; err_data->ue_count =0 ;
Signature · 6 lines
2.17.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 2/4 drm/amdgpu: abstract EDC counter clear to a separated function
In reply to Dennis Li1. Add IP prefix for the IP related codes.
2. Refactor the code to clear EDC counter.
--- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 112 ++++++++++++++++++-------- 1 file changed, 77 insertions(+), 35 deletions(-)
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +77 -35
@@ -736,6 +736,7 @@ static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring);static u64 gfx_v9_0_ring_get_rptr_compute (struct amdgpu_ring *ring);static int gfx_v9_0_query_ras_error_count (struct amdgpu_device *adev,void *ras_error_status); +static void gfx_v9_0_clear_ras_edc_counter (struct amdgpu_device *adev);static int gfx_v9_0_ras_error_inject (struct amdgpu_device *adev,void *inject_if);
@@ -3996,7 +3997,7 @@ static const struct soc15_reg_entry sgpr2_init_regs[] = { { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE7), 0x0000ff00 }, }; -static const struct soc15_reg_entry sec_ded_counter_registers[] = { +static const struct soc15_reg_entry gfx_v9_0_edc_counter_regs[] = { {SOC15_REG_ENTRY (GC ,0 , mmCPC_EDC_SCRATCH_CNT),0 ,1 ,1 }, {SOC15_REG_ENTRY (GC ,0 , mmCPC_EDC_UCODE_CNT),0 ,1 ,1 }, { SOC15_REG_ENTRY(GC, 0, mmCPF_EDC_ROQ_CNT), 0, 1, 1},
@@ -4085,7 +4086,7 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev)struct amdgpu_ring *ring = &adev->gfx.compute_ring[0 ];struct amdgpu_ib ib;struct dma_fence *f =NULL ; -int r, i, j, k; +int r, i;unsigned total_size, vgpr_offset, sgpr_offset;u64 gpu_addr;
@@ -4231,18 +4232,7 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) goto fail; } -/* read back registers to clear the counters */ -mutex_lock (&adev->grbm_idx_mutex); -for (i =0 ; i <ARRAY_SIZE (sec_ded_counter_registers); i++) { -for (j =0 ; j < sec_ded_counter_registers[i].se_num; j++) { -for (k =0 ; k < sec_ded_counter_registers[i].instance; k++) { -gfx_v9_0_select_se_sh (adev, j,0x0 , k); -RREG32 (SOC15_REG_ENTRY_OFFSET (sec_ded_counter_registers[i])); - } - } - } -WREG32_SOC15 (GC ,0 , mmGRBM_GFX_INDEX,0xe0000000 ); -mutex_unlock (&adev->grbm_idx_mutex); +gfx_v9_0_clear_ras_edc_counter (adev); fail:amdgpu_ib_free (adev, &ib,NULL );
@@ -5519,7 +5509,7 @@ static int gfx_v9_0_priv_inst_irq(struct amdgpu_device *adev, } -static const struct soc15_ras_field_entry gc_ras_fields_vg20[] = { +static const struct soc15_ras_field_entry gfx_v9_0_ras_fields[] = { {"CPC_SCRATCH" ,SOC15_REG_ENTRY (GC ,0 , mmCPC_EDC_SCRATCH_CNT),SOC15_REG_FIELD (CPC_EDC_SCRATCH_CNT ,SEC_COUNT ),SOC15_REG_FIELD (CPC_EDC_SCRATCH_CNT ,DED_COUNT )
@@ -6092,7 +6082,7 @@ static int gfx_v9_0_query_utc_edc_status(struct amdgpu_device *adev,WREG32_SOC15 (GC ,0 , mmATC_L2_CACHE_4K_EDC_INDEX,255 );WREG32_SOC15 (GC ,0 , mmATC_L2_CACHE_4K_EDC_CNT,0 ); -for (i =0 ; i <16 ; i++) { +for (i =0 ; i <ARRAY_SIZE (vml2_mems); i++) {WREG32_SOC15 (GC ,0 , mmVM_L2_MEM_ECC_INDEX, i); data =RREG32_SOC15 (GC ,0 , mmVM_L2_MEM_ECC_CNT);
@@ -6111,7 +6101,7 @@ static int gfx_v9_0_query_utc_edc_status(struct amdgpu_device *adev, } } -for (i =0 ; i <7 ; i++) { +for (i =0 ; i <ARRAY_SIZE (vml2_walker_mems); i++) {WREG32_SOC15 (GC ,0 , mmVM_L2_WALKER_MEM_ECC_INDEX, i); data =RREG32_SOC15 (GC ,0 , mmVM_L2_WALKER_MEM_ECC_CNT);
@@ -6132,7 +6122,7 @@ static int gfx_v9_0_query_utc_edc_status(struct amdgpu_device *adev, } } -for (i =0 ; i <4 ; i++) { +for (i =0 ; i <ARRAY_SIZE (atc_l2_cache_2m_mems); i++) {WREG32_SOC15 (GC ,0 , mmATC_L2_CACHE_2M_EDC_INDEX, i); data =RREG32_SOC15 (GC ,0 , mmATC_L2_CACHE_2M_EDC_CNT);
@@ -6144,7 +6134,7 @@ static int gfx_v9_0_query_utc_edc_status(struct amdgpu_device *adev, } } -for (i =0 ; i <32 ; i++) { +for (i =0 ; i <ARRAY_SIZE (atc_l2_cache_4k_mems); i++) {WREG32_SOC15 (GC ,0 , mmATC_L2_CACHE_4K_EDC_INDEX, i); data =RREG32_SOC15 (GC ,0 , mmATC_L2_CACHE_4K_EDC_CNT);
@@ -6171,36 +6161,36 @@ static int gfx_v9_0_query_utc_edc_status(struct amdgpu_device *adev,return 0 ; } -static int __get_ras_error_count (const struct soc15_reg_entry *reg, +static int gfx_v9_0_ras_error_count (const struct soc15_reg_entry *reg,uint32_t se_id,uint32_t inst_id,uint32_t value,uint32_t *sec_count,uint32_t *ded_count) {uint32_t i;uint32_t sec_cnt, ded_cnt; -for (i =0 ; i <ARRAY_SIZE (gc_ras_fields_vg20); i++) { -if (gc_ras_fields_vg20[i].reg_offset != reg->reg_offset || - gc_ras_fields_vg20[i].seg != reg->seg || - gc_ras_fields_vg20[i].inst != reg->inst) +for (i =0 ; i <ARRAY_SIZE (gfx_v9_0_ras_fields); i++) { +if (gfx_v9_0_ras_fields[i].reg_offset != reg->reg_offset || + gfx_v9_0_ras_fields[i].seg != reg->seg || + gfx_v9_0_ras_fields[i].inst != reg->inst)continue ; sec_cnt = (value & - gc_ras_fields_vg20[i].sec_count_mask) >> - gc_ras_fields_vg20[i].sec_count_shift; + gfx_v9_0_ras_fields[i].sec_count_mask) >> + gfx_v9_0_ras_fields[i].sec_count_shift;if (sec_cnt) {DRM_INFO ("GFX SubBlock %s, Instance[%d][%d], SEC %d\n" , - gc_ras_fields_vg20[i].name, + gfx_v9_0_ras_fields[i].name, se_id, inst_id, sec_cnt); *sec_count += sec_cnt; } ded_cnt = (value & - gc_ras_fields_vg20[i].ded_count_mask) >> - gc_ras_fields_vg20[i].ded_count_shift; + gfx_v9_0_ras_fields[i].ded_count_mask) >> + gfx_v9_0_ras_fields[i].ded_count_shift;if (ded_cnt) {DRM_INFO ("GFX SubBlock %s, Instance[%d][%d], DED %d\n" , - gc_ras_fields_vg20[i].name, + gfx_v9_0_ras_fields[i].name, se_id, inst_id, ded_cnt); *ded_count += ded_cnt;
@@ -6210,6 +6200,58 @@ static int __get_ras_error_count(const struct soc15_reg_entry *reg,return 0 ; } +static void gfx_v9_0_clear_ras_edc_counter (struct amdgpu_device *adev) +{ +int i, j, k; + +/* read back registers to clear the counters */ +mutex_lock (&adev->grbm_idx_mutex); +for (i =0 ; i <ARRAY_SIZE (gfx_v9_0_edc_counter_regs); i++) { +for (j =0 ; j < gfx_v9_0_edc_counter_regs[i].se_num; j++) { +for (k =0 ; k < gfx_v9_0_edc_counter_regs[i].instance; k++) { +gfx_v9_0_select_se_sh (adev, j,0x0 , k); +RREG32 (SOC15_REG_ENTRY_OFFSET (gfx_v9_0_edc_counter_regs[i])); + } + } + } +WREG32_SOC15 (GC ,0 , mmGRBM_GFX_INDEX,0xe0000000 ); +mutex_unlock (&adev->grbm_idx_mutex); + +WREG32_SOC15 (GC ,0 , mmVM_L2_MEM_ECC_INDEX,255 ); +WREG32_SOC15 (GC ,0 , mmVM_L2_MEM_ECC_CNT,0 ); +WREG32_SOC15 (GC ,0 , mmVM_L2_WALKER_MEM_ECC_INDEX,255 ); +WREG32_SOC15 (GC ,0 , mmVM_L2_WALKER_MEM_ECC_CNT,0 ); +WREG32_SOC15 (GC ,0 , mmATC_L2_CACHE_2M_EDC_INDEX,255 ); +WREG32_SOC15 (GC ,0 , mmATC_L2_CACHE_2M_EDC_CNT,0 ); +WREG32_SOC15 (GC ,0 , mmATC_L2_CACHE_4K_EDC_INDEX,255 ); +WREG32_SOC15 (GC ,0 , mmATC_L2_CACHE_4K_EDC_CNT,0 ); + +for (i =0 ; i <ARRAY_SIZE (vml2_mems); i++) { +WREG32_SOC15 (GC ,0 , mmVM_L2_MEM_ECC_INDEX, i); +RREG32_SOC15 (GC ,0 , mmVM_L2_MEM_ECC_CNT); + } + +for (i =0 ; i <ARRAY_SIZE (vml2_walker_mems); i++) { +WREG32_SOC15 (GC ,0 , mmVM_L2_WALKER_MEM_ECC_INDEX, i); +RREG32_SOC15 (GC ,0 , mmVM_L2_WALKER_MEM_ECC_CNT); + } + +for (i =0 ; i <ARRAY_SIZE (atc_l2_cache_2m_mems); i++) { +WREG32_SOC15 (GC ,0 , mmATC_L2_CACHE_2M_EDC_INDEX, i); +RREG32_SOC15 (GC ,0 , mmATC_L2_CACHE_2M_EDC_CNT); + } + +for (i =0 ; i <ARRAY_SIZE (atc_l2_cache_4k_mems); i++) { +WREG32_SOC15 (GC ,0 , mmATC_L2_CACHE_4K_EDC_INDEX, i); +RREG32_SOC15 (GC ,0 , mmATC_L2_CACHE_4K_EDC_CNT); + } + +WREG32_SOC15 (GC ,0 , mmVM_L2_MEM_ECC_INDEX,255 ); +WREG32_SOC15 (GC ,0 , mmVM_L2_WALKER_MEM_ECC_INDEX,255 ); +WREG32_SOC15 (GC ,0 , mmATC_L2_CACHE_2M_EDC_INDEX,255 ); +WREG32_SOC15 (GC ,0 , mmATC_L2_CACHE_4K_EDC_INDEX,255 ); +} +static int gfx_v9_0_query_ras_error_count (struct amdgpu_device *adev,void *ras_error_status) {
@@ -6226,14 +6268,14 @@ static int gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev,mutex_lock (&adev->grbm_idx_mutex); -for (i =0 ; i <ARRAY_SIZE (sec_ded_counter_registers); i++) { -for (j =0 ; j < sec_ded_counter_registers[i].se_num; j++) { -for (k =0 ; k < sec_ded_counter_registers[i].instance; k++) { +for (i =0 ; i <ARRAY_SIZE (gfx_v9_0_edc_counter_regs); i++) { +for (j =0 ; j < gfx_v9_0_edc_counter_regs[i].se_num; j++) { +for (k =0 ; k < gfx_v9_0_edc_counter_regs[i].instance; k++) {gfx_v9_0_select_se_sh (adev, j,0 , k); reg_value = -RREG32 (SOC15_REG_ENTRY_OFFSET (sec_ded_counter_registers[i])); +RREG32 (SOC15_REG_ENTRY_OFFSET (gfx_v9_0_edc_counter_regs[i]));if (reg_value) -__get_ras_error_count (&sec_ded_counter_registers[i], +gfx_v9_0_ras_error_count (&gfx_v9_0_edc_counter_regs[i], j, k, reg_value, &sec_count, &ded_count); }
Signature · 6 lines
2.17.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 3/4 drm/amdgpu: add EDC counter registers of gc for Arcturus
In reply to Dennis Liadd reg headers to gc includes
v2: remove unused registers and fields in this patch set
--- .../amd/include/asic_reg/gc/gc_9_4_1_offset.h | 264 +++++++ .../include/asic_reg/gc/gc_9_4_1_sh_mask.h | 748 ++++++++++++++++++ 2 files changed, 1012 insertions(+) create mode 100644 drivers/gpu/drm/amd/include/asic_reg/gc/gc_9_4_1_offset.h create mode 100644 drivers/gpu/drm/amd/include/asic_reg/gc/gc_9_4_1_sh_mask.h
drivers/gpu/drm/amd/include/asic_reg/gc/gc_9_4_1_offset.h new file +264 -0
@@ -0,0 +1,264 @@ +/* +* Copyright (C) 2020 Advanced Micro Devices, Inc. +* +* Permission is hereby granted, free of charge, to any person obtaining a +* copy of this software and associated documentation files (the "Software"), +* to deal in the Software without restriction, including without limitation +* the rights to use, copy, modify, merge, publish, distribute, sublicense, +* and/or sell copies of the Software, and to permit persons to whom the +* Software is furnished to do so, subject to the following conditions: +* +* The above copyright notice and this permission notice shall be included +* in all copies or substantial portions of the Software. +* +* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS +* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL +* THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN +* AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN +* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +*/ +#ifndef _gc_9_4_1_OFFSET_HEADER +#define _gc_9_4_1_OFFSET_HEADER + +// addressBlock: gc_grbmdec +// base address: 0x8000 +#define mmGRBM_CNTL 0x0000 +#define mmGRBM_CNTL_BASE_IDX 0 +#define mmGRBM_SKEW_CNTL 0x0001 +#define mmGRBM_SKEW_CNTL_BASE_IDX 0 +#define mmGRBM_STATUS2 0x0002 +#define mmGRBM_STATUS2_BASE_IDX 0 +#define mmGRBM_PWR_CNTL 0x0003 +#define mmGRBM_PWR_CNTL_BASE_IDX 0 +#define mmGRBM_STATUS 0x0004 +#define mmGRBM_STATUS_BASE_IDX 0 +#define mmGRBM_STATUS_SE0 0x0005 +#define mmGRBM_STATUS_SE0_BASE_IDX 0 +#define mmGRBM_STATUS_SE1 0x0006 +#define mmGRBM_STATUS_SE1_BASE_IDX 0 +#define mmGRBM_SOFT_RESET 0x0008 +#define mmGRBM_SOFT_RESET_BASE_IDX 0 +#define mmGRBM_GFX_CLKEN_CNTL 0x000c +#define mmGRBM_GFX_CLKEN_CNTL_BASE_IDX 0 +#define mmGRBM_WAIT_IDLE_CLOCKS 0x000d +#define mmGRBM_WAIT_IDLE_CLOCKS_BASE_IDX 0 +#define mmGRBM_STATUS_SE2 0x000e +#define mmGRBM_STATUS_SE2_BASE_IDX 0 +#define mmGRBM_STATUS_SE3 0x000f +#define mmGRBM_STATUS_SE3_BASE_IDX 0 +#define mmGRBM_READ_ERROR 0x0016 +#define mmGRBM_READ_ERROR_BASE_IDX 0 +#define mmGRBM_READ_ERROR2 0x0017 +#define mmGRBM_READ_ERROR2_BASE_IDX 0 +#define mmGRBM_INT_CNTL 0x0018 +#define mmGRBM_INT_CNTL_BASE_IDX 0 +#define mmGRBM_TRAP_OP 0x0019 +#define mmGRBM_TRAP_OP_BASE_IDX 0 +#define mmGRBM_TRAP_ADDR 0x001a +#define mmGRBM_TRAP_ADDR_BASE_IDX 0 +#define mmGRBM_TRAP_ADDR_MSK 0x001b +#define mmGRBM_TRAP_ADDR_MSK_BASE_IDX 0 +#define mmGRBM_TRAP_WD 0x001c +#define mmGRBM_TRAP_WD_BASE_IDX 0 +#define mmGRBM_TRAP_WD_MSK 0x001d +#define mmGRBM_TRAP_WD_MSK_BASE_IDX 0 +#define mmGRBM_DSM_BYPASS 0x001e +#define mmGRBM_DSM_BYPASS_BASE_IDX 0 +#define mmGRBM_WRITE_ERROR 0x001f +#define mmGRBM_WRITE_ERROR_BASE_IDX 0 +#define mmGRBM_IOV_ERROR 0x0020 +#define mmGRBM_IOV_ERROR_BASE_IDX 0 +#define mmGRBM_CHIP_REVISION 0x0021 +#define mmGRBM_CHIP_REVISION_BASE_IDX 0 +#define mmGRBM_GFX_CNTL 0x0022 +#define mmGRBM_GFX_CNTL_BASE_IDX 0 +#define mmGRBM_RSMU_CFG 0x0023 +#define mmGRBM_RSMU_CFG_BASE_IDX 0 +#define mmGRBM_IH_CREDIT 0x0024 +#define mmGRBM_IH_CREDIT_BASE_IDX 0 +#define mmGRBM_PWR_CNTL2 0x0025 +#define mmGRBM_PWR_CNTL2_BASE_IDX 0 +#define mmGRBM_UTCL2_INVAL_RANGE_START 0x0026 +#define mmGRBM_UTCL2_INVAL_RANGE_START_BASE_IDX 0 +#define mmGRBM_UTCL2_INVAL_RANGE_END 0x0027 +#define mmGRBM_UTCL2_INVAL_RANGE_END_BASE_IDX 0 +#define mmGRBM_RSMU_READ_ERROR 0x0028 +#define mmGRBM_RSMU_READ_ERROR_BASE_IDX 0 +#define mmGRBM_CHICKEN_BITS 0x0029 +#define mmGRBM_CHICKEN_BITS_BASE_IDX 0 +#define mmGRBM_FENCE_RANGE0 0x002a +#define mmGRBM_FENCE_RANGE0_BASE_IDX 0 +#define mmGRBM_FENCE_RANGE1 0x002b +#define mmGRBM_FENCE_RANGE1_BASE_IDX 0 +#define mmGRBM_NOWHERE 0x003f +#define mmGRBM_NOWHERE_BASE_IDX 0 +#define mmGRBM_SCRATCH_REG0 0x0040 +#define mmGRBM_SCRATCH_REG0_BASE_IDX 0 +#define mmGRBM_SCRATCH_REG1 0x0041 +#define mmGRBM_SCRATCH_REG1_BASE_IDX 0 +#define mmGRBM_SCRATCH_REG2 0x0042 +#define mmGRBM_SCRATCH_REG2_BASE_IDX 0 +#define mmGRBM_SCRATCH_REG3 0x0043 +#define mmGRBM_SCRATCH_REG3_BASE_IDX 0 +#define mmGRBM_SCRATCH_REG4 0x0044 +#define mmGRBM_SCRATCH_REG4_BASE_IDX 0 +#define mmGRBM_SCRATCH_REG5 0x0045 +#define mmGRBM_SCRATCH_REG5_BASE_IDX 0 +#define mmGRBM_SCRATCH_REG6 0x0046 +#define mmGRBM_SCRATCH_REG6_BASE_IDX 0 +#define mmGRBM_SCRATCH_REG7 0x0047 +#define mmGRBM_SCRATCH_REG7_BASE_IDX 0 + +// addressBlock: gc_cppdec2 +// base address: 0xc600 +#define mmCPF_EDC_TAG_CNT 0x1189 +#define mmCPF_EDC_TAG_CNT_BASE_IDX 0 +#define mmCPF_EDC_ROQ_CNT 0x118a +#define mmCPF_EDC_ROQ_CNT_BASE_IDX 0 +#define mmCPG_EDC_TAG_CNT 0x118b +#define mmCPG_EDC_TAG_CNT_BASE_IDX 0 +#define mmCPG_EDC_DMA_CNT 0x118d +#define mmCPG_EDC_DMA_CNT_BASE_IDX 0 +#define mmCPC_EDC_SCRATCH_CNT 0x118e +#define mmCPC_EDC_SCRATCH_CNT_BASE_IDX 0 +#define mmCPC_EDC_UCODE_CNT 0x118f +#define mmCPC_EDC_UCODE_CNT_BASE_IDX 0 +#define mmDC_EDC_STATE_CNT 0x1191 +#define mmDC_EDC_STATE_CNT_BASE_IDX 0 +#define mmDC_EDC_CSINVOC_CNT 0x1192 +#define mmDC_EDC_CSINVOC_CNT_BASE_IDX 0 +#define mmDC_EDC_RESTORE_CNT 0x1193 +#define mmDC_EDC_RESTORE_CNT_BASE_IDX 0 + +// addressBlock: gc_gdsdec +// base address: 0x9700 +#define mmGDS_EDC_CNT 0x05c5 +#define mmGDS_EDC_CNT_BASE_IDX 0 +#define mmGDS_EDC_GRBM_CNT 0x05c6 +#define mmGDS_EDC_GRBM_CNT_BASE_IDX 0 +#define mmGDS_EDC_OA_DED 0x05c7 +#define mmGDS_EDC_OA_DED_BASE_IDX 0 +#define mmGDS_EDC_OA_PHY_CNT 0x05cb +#define mmGDS_EDC_OA_PHY_CNT_BASE_IDX 0 +#define mmGDS_EDC_OA_PIPE_CNT 0x05cc +#define mmGDS_EDC_OA_PIPE_CNT_BASE_IDX 0 + +// addressBlock: gc_shsdec +// base address: 0x9000 +#define mmSPI_EDC_CNT 0x0445 +#define mmSPI_EDC_CNT_BASE_IDX 0 + +// addressBlock: gc_sqdec +// base address: 0x8c00 +#define mmSQC_EDC_CNT2 0x032c +#define mmSQC_EDC_CNT2_BASE_IDX 0 +#define mmSQC_EDC_CNT3 0x032d +#define mmSQC_EDC_CNT3_BASE_IDX 0 +#define mmSQC_EDC_PARITY_CNT3 0x032e +#define mmSQC_EDC_PARITY_CNT3_BASE_IDX 0 +#define mmSQC_EDC_CNT 0x03a2 +#define mmSQC_EDC_CNT_BASE_IDX 0 +#define mmSQ_EDC_SEC_CNT 0x03a3 +#define mmSQ_EDC_SEC_CNT_BASE_IDX 0 +#define mmSQ_EDC_DED_CNT 0x03a4 +#define mmSQ_EDC_DED_CNT_BASE_IDX 0 +#define mmSQ_EDC_INFO 0x03a5 +#define mmSQ_EDC_INFO_BASE_IDX 0 +#define mmSQ_EDC_CNT 0x03a6 +#define mmSQ_EDC_CNT_BASE_IDX 0 + +// addressBlock: gc_tpdec +// base address: 0x9400 +#define mmTA_EDC_CNT 0x0586 +#define mmTA_EDC_CNT_BASE_IDX 0 + +// addressBlock: gc_tcdec +// base address: 0xac00 +#define mmTCP_EDC_CNT 0x0b17 +#define mmTCP_EDC_CNT_BASE_IDX 0 +#define mmTCP_EDC_CNT_NEW 0x0b18 +#define mmTCP_EDC_CNT_NEW_BASE_IDX 0 +#define mmTCP_ATC_EDC_GATCL1_CNT 0x12b1 +#define mmTCP_ATC_EDC_GATCL1_CNT_BASE_IDX 0 +#define mmTCI_EDC_CNT 0x0b60 +#define mmTCI_EDC_CNT_BASE_IDX 0 +#define mmTCC_EDC_CNT 0x0b82 +#define mmTCC_EDC_CNT_BASE_IDX 0 +#define mmTCC_EDC_CNT2 0x0b83 +#define mmTCC_EDC_CNT2_BASE_IDX 0 +#define mmTCA_EDC_CNT 0x0bc5 +#define mmTCA_EDC_CNT_BASE_IDX 0 + +// addressBlock: gc_tpdec +// base address: 0x9400 +#define mmTD_EDC_CNT 0x052e +#define mmTD_EDC_CNT_BASE_IDX 0 +#define mmTA_EDC_CNT 0x0586 +#define mmTA_EDC_CNT_BASE_IDX 0 + +// addressBlock: gc_ea_gceadec2 +// base address: 0x9c00 +#define mmGCEA_EDC_CNT 0x0706 +#define mmGCEA_EDC_CNT_BASE_IDX 0 +#define mmGCEA_EDC_CNT2 0x0707 +#define mmGCEA_EDC_CNT2_BASE_IDX 0 +#define mmGCEA_EDC_CNT3 0x071b +#define mmGCEA_EDC_CNT3_BASE_IDX 0 + +// addressBlock: gc_gfxudec +// base address: 0x30000 +#define mmSCRATCH_REG0 0x2040 +#define mmSCRATCH_REG0_BASE_IDX 1 +#define mmSCRATCH_REG1 0x2041 +#define mmSCRATCH_REG1_BASE_IDX 1 +#define mmSCRATCH_REG2 0x2042 +#define mmSCRATCH_REG2_BASE_IDX 1 +#define mmSCRATCH_REG3 0x2043 +#define mmSCRATCH_REG3_BASE_IDX 1 +#define mmSCRATCH_REG4 0x2044 +#define mmSCRATCH_REG4_BASE_IDX 1 +#define mmSCRATCH_REG5 0x2045 +#define mmSCRATCH_REG5_BASE_IDX 1 +#define mmSCRATCH_REG6 0x2046 +#define mmSCRATCH_REG6_BASE_IDX 1 +#define mmSCRATCH_REG7 0x2047 +#define mmSCRATCH_REG7_BASE_IDX 1 +#define mmGRBM_GFX_INDEX 0x2200 +#define mmGRBM_GFX_INDEX_BASE_IDX 1 + +// addressBlock: gc_utcl2_atcl2dec +// base address: 0xa000 +#define mmATC_L2_CACHE_4K_DSM_INDEX 0x080e +#define mmATC_L2_CACHE_4K_DSM_INDEX_BASE_IDX 0 +#define mmATC_L2_CACHE_2M_DSM_INDEX 0x080f +#define mmATC_L2_CACHE_2M_DSM_INDEX_BASE_IDX 0 +#define mmATC_L2_CACHE_4K_DSM_CNTL 0x0810 +#define mmATC_L2_CACHE_4K_DSM_CNTL_BASE_IDX 0 +#define mmATC_L2_CACHE_2M_DSM_CNTL 0x0811 +#define mmATC_L2_CACHE_2M_DSM_CNTL_BASE_IDX 0 + +// addressBlock: gc_utcl2_vml2pfdec +// base address: 0xa100 +#define mmVML2_MEM_ECC_INDEX 0x0860 +#define mmVML2_MEM_ECC_INDEX_BASE_IDX 0 +#define mmVML2_WALKER_MEM_ECC_INDEX 0x0861 +#define mmVML2_WALKER_MEM_ECC_INDEX_BASE_IDX 0 +#define mmUTCL2_MEM_ECC_INDEX 0x0862 +#define mmUTCL2_MEM_ECC_INDEX_BASE_IDX 0 + +#define mmVML2_MEM_ECC_CNTL 0x0863 +#define mmVML2_MEM_ECC_CNTL_BASE_IDX 0 +#define mmVML2_WALKER_MEM_ECC_CNTL 0x0864 +#define mmVML2_WALKER_MEM_ECC_CNTL_BASE_IDX 0 +#define mmUTCL2_MEM_ECC_CNTL 0x0865 +#define mmUTCL2_MEM_ECC_CNTL_BASE_IDX 0 + +// addressBlock: gc_rlcpdec +// base address: 0x3b000 +#define mmRLC_EDC_CNT 0x4d40 +#define mmRLC_EDC_CNT_BASE_IDX 1 +#define mmRLC_EDC_CNT2 0x4d41 +#define mmRLC_EDC_CNT2_BASE_IDX 1 + +#endif
drivers/gpu/drm/amd/include/asic_reg/gc/gc_9_4_1_sh_mask.h new file +748 -0
749 lines not on this page · raw message
Signature · 6 lines
2.17.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 4/4 drm/amdgpu: add RAS support for the gfx block of Arcturus
In reply to Dennis LiImplement functions to do the RAS error injection and query EDC counter.
--- drivers/gpu/drm/amd/amdgpu/Makefile | 1 + drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 26 +- drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c | 978 ++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h | 35 + 4 files changed, 1039 insertions(+), 1 deletion(-) create mode 100644 drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c create mode 100644 drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h
drivers/gpu/drm/amd/amdgpu/Makefile +1 -0
@@ -120,6 +120,7 @@ amdgpu-y += \ amdgpu_rlc.o \ gfx_v8_0.o \ gfx_v9_0.o \ + gfx_v9_4.o \ gfx_v10_0.o # add async DMA block
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +25 -1
@@ -48,6 +48,8 @@#include "amdgpu_ras.h" +#include "gfx_v9_4.h" +#define GFX9_NUM_GFX_RINGS 1#define GFX9_MEC_HPD_SIZE 4096#define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L
@@ -1822,6 +1824,17 @@ static const struct amdgpu_gfx_funcs gfx_v9_0_gfx_funcs = { .query_ras_error_count = &gfx_v9_0_query_ras_error_count }; +static const struct amdgpu_gfx_funcs gfx_v9_4_gfx_funcs = { + .get_gpu_clock_counter = &gfx_v9_0_get_gpu_clock_counter, + .select_se_sh = &gfx_v9_0_select_se_sh, + .read_wave_data = &gfx_v9_0_read_wave_data, + .read_wave_sgprs = &gfx_v9_0_read_wave_sgprs, + .read_wave_vgprs = &gfx_v9_0_read_wave_vgprs, + .select_me_pipe_q = &gfx_v9_0_select_me_pipe_q, + .ras_error_inject = &gfx_v9_4_ras_error_inject, + .query_ras_error_count = &gfx_v9_4_query_ras_error_count +}; +static int gfx_v9_0_gpu_early_init (struct amdgpu_device *adev) {u32 gb_addr_config;
@@ -1873,6 +1886,7 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) gb_addr_config =RAVEN_GB_ADDR_CONFIG_GOLDEN ;break ;case CHIP_ARCTURUS : + adev->gfx.funcs = &gfx_v9_4_gfx_funcs; adev->gfx.config.max_hw_contexts =8 ; adev->gfx.config.sc_prim_fifo_size_frontend =0x20 ; adev->gfx.config.sc_prim_fifo_size_backend =0x100 ;
@@ -4232,7 +4246,17 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) goto fail; } -gfx_v9_0_clear_ras_edc_counter (adev); +switch (adev->asic_type) + { +case CHIP_VEGA20 : +gfx_v9_0_clear_ras_edc_counter (adev); +break ; +case CHIP_ARCTURUS : +gfx_v9_4_clear_ras_edc_counter (adev); +break ; +default : +break ; + } fail:amdgpu_ib_free (adev, &ib,NULL );
drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c new file +978 -0
978 lines not on this page · raw message
drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h new file +35 -0
@@ -0,0 +1,35 @@ +/* +* Copyright 2020 Advanced Micro Devices, Inc. +* +* Permission is hereby granted, free of charge, to any person obtaining a +* copy of this software and associated documentation files (the "Software"), +* to deal in the Software without restriction, including without limitation +* the rights to use, copy, modify, merge, publish, distribute, sublicense, +* and/or sell copies of the Software, and to permit persons to whom the +* Software is furnished to do so, subject to the following conditions: +* +* The above copyright notice and this permission notice shall be included in +* all copies or substantial portions of the Software. +* +* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL +* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR +* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, +* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR +* OTHER DEALINGS IN THE SOFTWARE. +* +*/ + +#ifndef __GFX_V9_4_H__ +#define __GFX_V9_4_H__ + +void gfx_v9_4_clear_ras_edc_counter (struct amdgpu_device *adev); + +int gfx_v9_4_query_ras_error_count (struct amdgpu_device *adev, +void *ras_error_status); + +int gfx_v9_4_ras_error_inject (struct amdgpu_device *adev, +void *inject_if); + +#endif /* __GFX_V9_4_H__ */
Signature · 6 lines
2.17.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
RE: [PATCH 4/4] drm/amdgpu: add RAS support for the gfx block of Arcturus
[AMD Public Use]
+ switch (adev->asic_type)
+ {
Please correct the coding style. '{' should stay at the same line of switch.
+ mutex_unlock(&adev->grbm_idx_mutex);
+
+ gfx_v9_4_query_utc_edc_status(adev, err_data);
Is it necessary to move gfx_v9_4_query_utc_edc_status calling above to be protected by lock? This can possibly avoid the problem of concurrent multi query?
With above two fixed, the series is: Reviewed-by: Guchun Chen <[email protected]>
Regards,
Guchun
-----Original Message-----
From: Dennis Li <[email protected]>
Sent: Sunday, January 19, 2020 10:46 AM
To: [email protected]; Deucher, Alexander <[email protected]>; Zhou1, Tao <[email protected]>; Zhang, Hawking <[email protected]>; Chen, Guchun <[email protected]>
Cc: Li, Dennis <[email protected]>
Subject: [PATCH 4/4] drm/amdgpu: add RAS support for the gfx block of Arcturus
Implement functions to do the RAS error injection and query EDC counter.
--- drivers/gpu/drm/amd/amdgpu/Makefile | 1 + drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 26 +- drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c | 978 ++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h | 35 + 4 files changed, 1039 insertions(+), 1 deletion(-) create mode 100644 drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c create mode 100644 drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h
drivers/gpu/drm/amd/amdgpu/Makefile +1 -0
@@ -120,6 +120,7 @@ amdgpu-y += \ amdgpu_rlc.o \ gfx_v8_0.o \ gfx_v9_0.o \ + gfx_v9_4.o \ gfx_v10_0.o # add async DMA block
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +2 -0
@@ -48,6 +48,8 @@#include "amdgpu_ras.h" +#include "gfx_v9_4.h" +#define GFX9_NUM_GFX_RINGS 1#define GFX9_MEC_HPD_SIZE 4096#define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L @@ -1822,6 +1824,17 @@ static const struct amdgpu_gfx_funcs gfx_v9_0_gfx_funcs = {
.query_ras_error_count = &gfx_v9_0_query_ras_error_count };
+static const struct amdgpu_gfx_funcs gfx_v9_4_gfx_funcs = {
+ .get_gpu_clock_counter = &gfx_v9_0_get_gpu_clock_counter,
+ .select_se_sh = &gfx_v9_0_select_se_sh,
+ .read_wave_data = &gfx_v9_0_read_wave_data,
+ .read_wave_sgprs = &gfx_v9_0_read_wave_sgprs,
+ .read_wave_vgprs = &gfx_v9_0_read_wave_vgprs,
+ .select_me_pipe_q = &gfx_v9_0_select_me_pipe_q,
+ .ras_error_inject = &gfx_v9_4_ras_error_inject,
+ .query_ras_error_count = &gfx_v9_4_query_ras_error_count };
+
static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) {
u32 gb_addr_config;
@@ -1873,6 +1886,7 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev)
gb_addr_config = RAVEN_GB_ADDR_CONFIG_GOLDEN;
break;
case CHIP_ARCTURUS:
+ adev->gfx.funcs = &gfx_v9_4_gfx_funcs;
adev->gfx.config.max_hw_contexts = 8;
adev->gfx.config.sc_prim_fifo_size_frontend = 0x20;
adev->gfx.config.sc_prim_fifo_size_backend = 0x100; @@ -4232,7 +4246,17 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) goto fail;
}
- gfx_v9_0_clear_ras_edc_counter(adev);
+ switch (adev->asic_type)
+ {
+ case CHIP_VEGA20:
+ gfx_v9_0_clear_ras_edc_counter(adev);
+ break;
+ case CHIP_ARCTURUS:
+ gfx_v9_4_clear_ras_edc_counter(adev);
+ break;
+ default:
+ break;
+ }
fail:
amdgpu_ib_free(adev, &ib, NULL);
drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c new file +978 -0
978 lines not on this page · raw message
+ block_info.value = info->value;
+
+ mutex_lock(&adev->grbm_idx_mutex);
+ ret = psp_ras_trigger_error(&adev->psp, &block_info);
+ mutex_unlock(&adev->grbm_idx_mutex);
+
+ return ret;
+}
drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h new file +35 -0
@@ -0,0 +1,35 @@ +/* +* Copyright 2020 Advanced Micro Devices, Inc. +* +* Permission is hereby granted, free of charge, to any person +obtaining a +* copy of this software and associated documentation files (the +"Software"), +* to deal in the Software without restriction, including without +limitation +* the rights to use, copy, modify, merge, publish, distribute, +sublicense, +* and/or sell copies of the Software, and to permit persons to whom +the +* Software is furnished to do so, subject to the following conditions: +* +* The above copyright notice and this permission notice shall be +included in +* all copies or substantial portions of the Software. +* +* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR +* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, +* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT +SHALL +* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, +DAMAGES OR +* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR +OTHERWISE, +* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE +OR +* OTHER DEALINGS IN THE SOFTWARE. +* +*/ +
+#ifndef __GFX_V9_4_H__
+#define __GFX_V9_4_H__
+
+void gfx_v9_4_clear_ras_edc_counter(struct amdgpu_device *adev); +
+int gfx_v9_4_query_ras_error_count(struct amdgpu_device *adev,
+ void *ras_error_status);
+
+int gfx_v9_4_ras_error_inject(struct amdgpu_device *adev,
+ void *inject_if);
+
+#endif /* __GFX_V9_4_H__ */
--
2.17.1
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
RE: [PATCH 4/4] drm/amdgpu: add RAS support for the gfx block of Arcturus
[AMD Public Use]
Hi, Guchun,
adev->grbm_idx_mutex is only used to protect the access to registers whose instance switch is indexed by grbm_index.
Best Regards
Dennis Li
-----Original Message-----
From: Chen, Guchun <[email protected]>
Sent: Sunday, January 19, 2020 11:40 AM
To: Li, Dennis <[email protected]>; [email protected]; Deucher, Alexander <[email protected]>; Zhou1, Tao <[email protected]>; Zhang, Hawking <[email protected]>
Cc: Li, Dennis <[email protected]>
Subject: RE: [PATCH 4/4] drm/amdgpu: add RAS support for the gfx block of Arcturus
[AMD Public Use]
+ switch (adev->asic_type)
+ {
Please correct the coding style. '{' should stay at the same line of switch.
+ mutex_unlock(&adev->grbm_idx_mutex);
+
+ gfx_v9_4_query_utc_edc_status(adev, err_data);
Is it necessary to move gfx_v9_4_query_utc_edc_status calling above to be protected by lock? This can possibly avoid the problem of concurrent multi query?
With above two fixed, the series is: Reviewed-by: Guchun Chen <[email protected]>
Regards,
Guchun
-----Original Message-----
From: Dennis Li <[email protected]>
Sent: Sunday, January 19, 2020 10:46 AM
To: [email protected]; Deucher, Alexander <[email protected]>; Zhou1, Tao <[email protected]>; Zhang, Hawking <[email protected]>; Chen, Guchun <[email protected]>
Cc: Li, Dennis <[email protected]>
Subject: [PATCH 4/4] drm/amdgpu: add RAS support for the gfx block of Arcturus
Implement functions to do the RAS error injection and query EDC counter.
--- drivers/gpu/drm/amd/amdgpu/Makefile | 1 + drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 26 +- drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c | 978 ++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h | 35 + 4 files changed, 1039 insertions(+), 1 deletion(-) create mode 100644 drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c create mode 100644 drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h
drivers/gpu/drm/amd/amdgpu/Makefile +1 -0
@@ -120,6 +120,7 @@ amdgpu-y += \ amdgpu_rlc.o \ gfx_v8_0.o \ gfx_v9_0.o \ + gfx_v9_4.o \ gfx_v10_0.o # add async DMA block
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +2 -0
@@ -48,6 +48,8 @@#include "amdgpu_ras.h" +#include "gfx_v9_4.h" +#define GFX9_NUM_GFX_RINGS 1#define GFX9_MEC_HPD_SIZE 4096#define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L @@ -1822,6 +1824,17 @@ static const struct amdgpu_gfx_funcs gfx_v9_0_gfx_funcs = {
.query_ras_error_count = &gfx_v9_0_query_ras_error_count };
+static const struct amdgpu_gfx_funcs gfx_v9_4_gfx_funcs = {
+ .get_gpu_clock_counter = &gfx_v9_0_get_gpu_clock_counter,
+ .select_se_sh = &gfx_v9_0_select_se_sh,
+ .read_wave_data = &gfx_v9_0_read_wave_data,
+ .read_wave_sgprs = &gfx_v9_0_read_wave_sgprs,
+ .read_wave_vgprs = &gfx_v9_0_read_wave_vgprs,
+ .select_me_pipe_q = &gfx_v9_0_select_me_pipe_q,
+ .ras_error_inject = &gfx_v9_4_ras_error_inject,
+ .query_ras_error_count = &gfx_v9_4_query_ras_error_count };
+
static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) {
u32 gb_addr_config;
@@ -1873,6 +1886,7 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev)
gb_addr_config = RAVEN_GB_ADDR_CONFIG_GOLDEN;
break;
case CHIP_ARCTURUS:
+ adev->gfx.funcs = &gfx_v9_4_gfx_funcs;
adev->gfx.config.max_hw_contexts = 8;
adev->gfx.config.sc_prim_fifo_size_frontend = 0x20;
adev->gfx.config.sc_prim_fifo_size_backend = 0x100; @@ -4232,7 +4246,17 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) goto fail;
}
- gfx_v9_0_clear_ras_edc_counter(adev);
+ switch (adev->asic_type)
+ {
+ case CHIP_VEGA20:
+ gfx_v9_0_clear_ras_edc_counter(adev);
+ break;
+ case CHIP_ARCTURUS:
+ gfx_v9_4_clear_ras_edc_counter(adev);
+ break;
+ default:
+ break;
+ }
fail:
amdgpu_ib_free(adev, &ib, NULL);
drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c new file +978 -0
978 lines not on this page · raw message
+ block_info.value = info->value;
+
+ mutex_lock(&adev->grbm_idx_mutex);
+ ret = psp_ras_trigger_error(&adev->psp, &block_info);
+ mutex_unlock(&adev->grbm_idx_mutex);
+
+ return ret;
+}
drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h new file +35 -0
@@ -0,0 +1,35 @@ +/* +* Copyright 2020 Advanced Micro Devices, Inc. +* +* Permission is hereby granted, free of charge, to any person +obtaining a +* copy of this software and associated documentation files (the +"Software"), +* to deal in the Software without restriction, including without +limitation +* the rights to use, copy, modify, merge, publish, distribute, +sublicense, +* and/or sell copies of the Software, and to permit persons to whom +the +* Software is furnished to do so, subject to the following conditions: +* +* The above copyright notice and this permission notice shall be +included in +* all copies or substantial portions of the Software. +* +* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR +* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, +* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT +SHALL +* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, +DAMAGES OR +* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR +OTHERWISE, +* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE +OR +* OTHER DEALINGS IN THE SOFTWARE. +* +*/ +
+#ifndef __GFX_V9_4_H__
+#define __GFX_V9_4_H__
+
+void gfx_v9_4_clear_ras_edc_counter(struct amdgpu_device *adev); +
+int gfx_v9_4_query_ras_error_count(struct amdgpu_device *adev,
+ void *ras_error_status);
+
+int gfx_v9_4_ras_error_inject(struct amdgpu_device *adev,
+ void *inject_if);
+
+#endif /* __GFX_V9_4_H__ */
--
2.17.1
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx