diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c')
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 1791 |
1 files changed, 1243 insertions, 548 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 97105a5bb246..7f944bb11298 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -38,7 +38,6 @@ #include "gc/gc_9_0_sh_mask.h" #include "vega10_enum.h" -#include "hdp/hdp_4_0_offset.h" #include "soc15_common.h" #include "clearstate_gfx9.h" @@ -48,18 +47,19 @@ #include "amdgpu_ras.h" +#include "gfx_v9_4.h" +#include "gfx_v9_0.h" +#include "gfx_v9_4_2.h" + +#include "asic_reg/pwr/pwr_10_0_offset.h" +#include "asic_reg/pwr/pwr_10_0_sh_mask.h" +#include "asic_reg/gc/gc_9_0_default.h" + #define GFX9_NUM_GFX_RINGS 1 #define GFX9_MEC_HPD_SIZE 4096 #define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L #define RLC_SAVE_RESTORE_ADDR_STARTING_OFFSET 0x00000000L -#define mmPWR_MISC_CNTL_STATUS 0x0183 -#define mmPWR_MISC_CNTL_STATUS_BASE_IDX 0 -#define PWR_MISC_CNTL_STATUS__PWR_GFX_RLC_CGPG_EN__SHIFT 0x0 -#define PWR_MISC_CNTL_STATUS__PWR_GFXOFF_STATUS__SHIFT 0x1 -#define PWR_MISC_CNTL_STATUS__PWR_GFX_RLC_CGPG_EN_MASK 0x00000001L -#define PWR_MISC_CNTL_STATUS__PWR_GFXOFF_STATUS_MASK 0x00000006L - #define mmGCEA_PROBE_MAP 0x070c #define mmGCEA_PROBE_MAP_BASE_IDX 0 @@ -108,16 +108,25 @@ MODULE_FIRMWARE("amdgpu/raven2_rlc.bin"); MODULE_FIRMWARE("amdgpu/raven_kicker_rlc.bin"); MODULE_FIRMWARE("amdgpu/arcturus_mec.bin"); -MODULE_FIRMWARE("amdgpu/arcturus_mec2.bin"); MODULE_FIRMWARE("amdgpu/arcturus_rlc.bin"); MODULE_FIRMWARE("amdgpu/renoir_ce.bin"); MODULE_FIRMWARE("amdgpu/renoir_pfp.bin"); MODULE_FIRMWARE("amdgpu/renoir_me.bin"); MODULE_FIRMWARE("amdgpu/renoir_mec.bin"); -MODULE_FIRMWARE("amdgpu/renoir_mec2.bin"); MODULE_FIRMWARE("amdgpu/renoir_rlc.bin"); +MODULE_FIRMWARE("amdgpu/green_sardine_ce.bin"); +MODULE_FIRMWARE("amdgpu/green_sardine_pfp.bin"); +MODULE_FIRMWARE("amdgpu/green_sardine_me.bin"); +MODULE_FIRMWARE("amdgpu/green_sardine_mec.bin"); +MODULE_FIRMWARE("amdgpu/green_sardine_mec2.bin"); +MODULE_FIRMWARE("amdgpu/green_sardine_rlc.bin"); + +MODULE_FIRMWARE("amdgpu/aldebaran_mec.bin"); +MODULE_FIRMWARE("amdgpu/aldebaran_mec2.bin"); +MODULE_FIRMWARE("amdgpu/aldebaran_rlc.bin"); + #define mmTCP_CHAN_STEER_0_ARCT 0x0b03 #define mmTCP_CHAN_STEER_0_ARCT_BASE_IDX 0 #define mmTCP_CHAN_STEER_1_ARCT 0x0b04 @@ -131,18 +140,6 @@ MODULE_FIRMWARE("amdgpu/renoir_rlc.bin"); #define mmTCP_CHAN_STEER_5_ARCT 0x0b0c #define mmTCP_CHAN_STEER_5_ARCT_BASE_IDX 0 -struct ras_gfx_subblock_reg { - const char *name; - uint32_t hwip; - uint32_t inst; - uint32_t seg; - uint32_t reg_offset; - uint32_t sec_count_mask; - uint32_t sec_count_shift; - uint32_t ded_count_mask; - uint32_t ded_count_shift; -}; - enum ta_ras_gfx_subblock { /*CPC*/ TA_RAS_BLOCK__GFX_CPC_INDEX_START = 0, @@ -521,8 +518,8 @@ static const struct soc15_reg_golden golden_settings_gc_9_0[] = SOC15_REG_GOLDEN_VALUE(GC, 0, mmSH_MEM_CONFIG, 0x00001000, 0x00001000), SOC15_REG_GOLDEN_VALUE(GC, 0, mmSPI_RESOURCE_RESERVE_CU_0, 0x0007ffff, 0x00000800), SOC15_REG_GOLDEN_VALUE(GC, 0, mmSPI_RESOURCE_RESERVE_CU_1, 0x0007ffff, 0x00000800), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmSPI_RESOURCE_RESERVE_EN_CU_0, 0x01ffffff, 0x0000ff87), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmSPI_RESOURCE_RESERVE_EN_CU_1, 0x01ffffff, 0x0000ff8f), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSPI_RESOURCE_RESERVE_EN_CU_0, 0x01ffffff, 0x00ffff87), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSPI_RESOURCE_RESERVE_EN_CU_1, 0x01ffffff, 0x00ffff8f), SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQC_CONFIG, 0x03000000, 0x020a2000), SOC15_REG_GOLDEN_VALUE(GC, 0, mmTA_CNTL_AUX, 0xfffffeef, 0x010b0000), SOC15_REG_GOLDEN_VALUE(GC, 0, mmTCP_CHAN_STEER_HI, 0xffffffff, 0x4a2c0e68), @@ -705,6 +702,12 @@ static const struct soc15_reg_golden golden_settings_gc_9_4_1_arct[] = SOC15_REG_GOLDEN_VALUE(GC, 0, mmTCP_CHAN_STEER_5_ARCT, 0x3ff, 0x135), SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_CONFIG, 0xffffffff, 0x011A0000), SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_FIFO_SIZES, 0xffffffff, 0x00000f00), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmTCP_UTCL1_CNTL1, 0x30000000, 0x30000000) +}; + +static const struct soc15_reg_rlcg rlcg_access_gc_9_0[] = { + {SOC15_REG_ENTRY(GC, 0, mmGRBM_GFX_INDEX)}, + {SOC15_REG_ENTRY(GC, 0, mmSQ_IND_INDEX)}, }; static const u32 GFX_RLC_SRM_INDEX_CNTL_ADDR_OFFSETS[] = @@ -731,6 +734,75 @@ static const u32 GFX_RLC_SRM_INDEX_CNTL_DATA_OFFSETS[] = mmRLC_SRM_INDEX_CNTL_DATA_7 - mmRLC_SRM_INDEX_CNTL_DATA_0, }; +static void gfx_v9_0_rlcg_w(struct amdgpu_device *adev, u32 offset, u32 v, u32 flag) +{ + static void *scratch_reg0; + static void *scratch_reg1; + static void *scratch_reg2; + static void *scratch_reg3; + static void *spare_int; + static uint32_t grbm_cntl; + static uint32_t grbm_idx; + + scratch_reg0 = adev->rmmio + (adev->reg_offset[GC_HWIP][0][mmSCRATCH_REG0_BASE_IDX] + mmSCRATCH_REG0)*4; + scratch_reg1 = adev->rmmio + (adev->reg_offset[GC_HWIP][0][mmSCRATCH_REG1_BASE_IDX] + mmSCRATCH_REG1)*4; + scratch_reg2 = adev->rmmio + (adev->reg_offset[GC_HWIP][0][mmSCRATCH_REG1_BASE_IDX] + mmSCRATCH_REG2)*4; + scratch_reg3 = adev->rmmio + (adev->reg_offset[GC_HWIP][0][mmSCRATCH_REG1_BASE_IDX] + mmSCRATCH_REG3)*4; + spare_int = adev->rmmio + (adev->reg_offset[GC_HWIP][0][mmRLC_SPARE_INT_BASE_IDX] + mmRLC_SPARE_INT)*4; + + grbm_cntl = adev->reg_offset[GC_HWIP][0][mmGRBM_GFX_CNTL_BASE_IDX] + mmGRBM_GFX_CNTL; + grbm_idx = adev->reg_offset[GC_HWIP][0][mmGRBM_GFX_INDEX_BASE_IDX] + mmGRBM_GFX_INDEX; + + if (amdgpu_sriov_runtime(adev)) { + pr_err("shouldn't call rlcg write register during runtime\n"); + return; + } + + if (offset == grbm_cntl || offset == grbm_idx) { + if (offset == grbm_cntl) + writel(v, scratch_reg2); + else if (offset == grbm_idx) + writel(v, scratch_reg3); + + writel(v, ((void __iomem *)adev->rmmio) + (offset * 4)); + } else { + uint32_t i = 0; + uint32_t retries = 50000; + + writel(v, scratch_reg0); + writel(offset | 0x80000000, scratch_reg1); + writel(1, spare_int); + for (i = 0; i < retries; i++) { + u32 tmp; + + tmp = readl(scratch_reg1); + if (!(tmp & 0x80000000)) + break; + + udelay(10); + } + if (i >= retries) + pr_err("timeout: rlcg program reg:0x%05x failed !\n", offset); + } + +} + +static void gfx_v9_0_sriov_wreg(struct amdgpu_device *adev, u32 offset, + u32 v, u32 acc_flags, u32 hwip) +{ + if ((acc_flags & AMDGPU_REGS_RLC) && + amdgpu_sriov_fullaccess(adev)) { + gfx_v9_0_rlcg_w(adev, offset, v, acc_flags); + + return; + } + + if (acc_flags & AMDGPU_REGS_NO_KIQ) + WREG32_NO_KIQ(offset, v); + else + WREG32(offset, v); +} + #define VEGA10_GB_ADDR_CONFIG_GOLDEN 0x2a114042 #define VEGA12_GB_ADDR_CONFIG_GOLDEN 0x24104041 #define RAVEN_GB_ADDR_CONFIG_GOLDEN 0x24000042 @@ -741,20 +813,148 @@ static void gfx_v9_0_set_irq_funcs(struct amdgpu_device *adev); static void gfx_v9_0_set_gds_init(struct amdgpu_device *adev); static void gfx_v9_0_set_rlc_funcs(struct amdgpu_device *adev); static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev, - struct amdgpu_cu_info *cu_info); + struct amdgpu_cu_info *cu_info); static uint64_t gfx_v9_0_get_gpu_clock_counter(struct amdgpu_device *adev); -static void gfx_v9_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num, u32 instance); 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 int gfx_v9_0_ras_error_inject(struct amdgpu_device *adev, void *inject_if); +static void gfx_v9_0_reset_ras_error_count(struct amdgpu_device *adev); + +static void gfx_v9_0_kiq_set_resources(struct amdgpu_ring *kiq_ring, + uint64_t queue_mask) +{ + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_RESOURCES, 6)); + amdgpu_ring_write(kiq_ring, + PACKET3_SET_RESOURCES_VMID_MASK(0) | + /* vmid_mask:0* queue_type:0 (KIQ) */ + PACKET3_SET_RESOURCES_QUEUE_TYPE(0)); + amdgpu_ring_write(kiq_ring, + lower_32_bits(queue_mask)); /* queue mask lo */ + amdgpu_ring_write(kiq_ring, + upper_32_bits(queue_mask)); /* queue mask hi */ + amdgpu_ring_write(kiq_ring, 0); /* gws mask lo */ + amdgpu_ring_write(kiq_ring, 0); /* gws mask hi */ + amdgpu_ring_write(kiq_ring, 0); /* oac mask */ + amdgpu_ring_write(kiq_ring, 0); /* gds heap base:0, gds heap size:0 */ +} + +static void gfx_v9_0_kiq_map_queues(struct amdgpu_ring *kiq_ring, + struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = kiq_ring->adev; + uint64_t mqd_addr = amdgpu_bo_gpu_offset(ring->mqd_obj); + uint64_t wptr_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0; + + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_MAP_QUEUES, 5)); + /* Q_sel:0, vmid:0, vidmem: 1, engine:0, num_Q:1*/ + amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */ + PACKET3_MAP_QUEUES_QUEUE_SEL(0) | /* Queue_Sel */ + PACKET3_MAP_QUEUES_VMID(0) | /* VMID */ + PACKET3_MAP_QUEUES_QUEUE(ring->queue) | + PACKET3_MAP_QUEUES_PIPE(ring->pipe) | + PACKET3_MAP_QUEUES_ME((ring->me == 1 ? 0 : 1)) | + /*queue_type: normal compute queue */ + PACKET3_MAP_QUEUES_QUEUE_TYPE(0) | + /* alloc format: all_on_one_pipe */ + PACKET3_MAP_QUEUES_ALLOC_FORMAT(0) | + PACKET3_MAP_QUEUES_ENGINE_SEL(eng_sel) | + /* num_queues: must be 1 */ + PACKET3_MAP_QUEUES_NUM_QUEUES(1)); + amdgpu_ring_write(kiq_ring, + PACKET3_MAP_QUEUES_DOORBELL_OFFSET(ring->doorbell_index)); + amdgpu_ring_write(kiq_ring, lower_32_bits(mqd_addr)); + amdgpu_ring_write(kiq_ring, upper_32_bits(mqd_addr)); + amdgpu_ring_write(kiq_ring, lower_32_bits(wptr_addr)); + amdgpu_ring_write(kiq_ring, upper_32_bits(wptr_addr)); +} + +static void gfx_v9_0_kiq_unmap_queues(struct amdgpu_ring *kiq_ring, + struct amdgpu_ring *ring, + enum amdgpu_unmap_queues_action action, + u64 gpu_addr, u64 seq) +{ + uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0; + + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4)); + amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */ + PACKET3_UNMAP_QUEUES_ACTION(action) | + PACKET3_UNMAP_QUEUES_QUEUE_SEL(0) | + PACKET3_UNMAP_QUEUES_ENGINE_SEL(eng_sel) | + PACKET3_UNMAP_QUEUES_NUM_QUEUES(1)); + amdgpu_ring_write(kiq_ring, + PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index)); + + if (action == PREEMPT_QUEUES_NO_UNMAP) { + amdgpu_ring_write(kiq_ring, lower_32_bits(gpu_addr)); + amdgpu_ring_write(kiq_ring, upper_32_bits(gpu_addr)); + amdgpu_ring_write(kiq_ring, seq); + } else { + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); + } +} + +static void gfx_v9_0_kiq_query_status(struct amdgpu_ring *kiq_ring, + struct amdgpu_ring *ring, + u64 addr, + u64 seq) +{ + uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0; + + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_QUERY_STATUS, 5)); + amdgpu_ring_write(kiq_ring, + PACKET3_QUERY_STATUS_CONTEXT_ID(0) | + PACKET3_QUERY_STATUS_INTERRUPT_SEL(0) | + PACKET3_QUERY_STATUS_COMMAND(2)); + /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */ + amdgpu_ring_write(kiq_ring, + PACKET3_QUERY_STATUS_DOORBELL_OFFSET(ring->doorbell_index) | + PACKET3_QUERY_STATUS_ENG_SEL(eng_sel)); + amdgpu_ring_write(kiq_ring, lower_32_bits(addr)); + amdgpu_ring_write(kiq_ring, upper_32_bits(addr)); + amdgpu_ring_write(kiq_ring, lower_32_bits(seq)); + amdgpu_ring_write(kiq_ring, upper_32_bits(seq)); +} + +static void gfx_v9_0_kiq_invalidate_tlbs(struct amdgpu_ring *kiq_ring, + uint16_t pasid, uint32_t flush_type, + bool all_hub) +{ + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_INVALIDATE_TLBS, 0)); + amdgpu_ring_write(kiq_ring, + PACKET3_INVALIDATE_TLBS_DST_SEL(1) | + PACKET3_INVALIDATE_TLBS_ALL_HUB(all_hub) | + PACKET3_INVALIDATE_TLBS_PASID(pasid) | + PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(flush_type)); +} + +static const struct kiq_pm4_funcs gfx_v9_0_kiq_pm4_funcs = { + .kiq_set_resources = gfx_v9_0_kiq_set_resources, + .kiq_map_queues = gfx_v9_0_kiq_map_queues, + .kiq_unmap_queues = gfx_v9_0_kiq_unmap_queues, + .kiq_query_status = gfx_v9_0_kiq_query_status, + .kiq_invalidate_tlbs = gfx_v9_0_kiq_invalidate_tlbs, + .set_resources_size = 8, + .map_queues_size = 7, + .unmap_queues_size = 6, + .query_status_size = 7, + .invalidate_tlbs_size = 2, +}; + +static void gfx_v9_0_set_kiq_pm4_funcs(struct amdgpu_device *adev) +{ + adev->gfx.kiq.pmf = &gfx_v9_0_kiq_pm4_funcs; +} static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) { - switch (adev->asic_type) { - case CHIP_VEGA10: + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(9, 0, 1): soc15_program_register_sequence(adev, golden_settings_gc_9_0, ARRAY_SIZE(golden_settings_gc_9_0)); @@ -762,7 +962,7 @@ static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) golden_settings_gc_9_0_vg10, ARRAY_SIZE(golden_settings_gc_9_0_vg10)); break; - case CHIP_VEGA12: + case IP_VERSION(9, 2, 1): soc15_program_register_sequence(adev, golden_settings_gc_9_2_1, ARRAY_SIZE(golden_settings_gc_9_2_1)); @@ -770,7 +970,7 @@ static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) golden_settings_gc_9_2_1_vg12, ARRAY_SIZE(golden_settings_gc_9_2_1_vg12)); break; - case CHIP_VEGA20: + case IP_VERSION(9, 4, 0): soc15_program_register_sequence(adev, golden_settings_gc_9_0, ARRAY_SIZE(golden_settings_gc_9_0)); @@ -778,15 +978,16 @@ static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) golden_settings_gc_9_0_vg20, ARRAY_SIZE(golden_settings_gc_9_0_vg20)); break; - case CHIP_ARCTURUS: + case IP_VERSION(9, 4, 1): soc15_program_register_sequence(adev, golden_settings_gc_9_4_1_arct, ARRAY_SIZE(golden_settings_gc_9_4_1_arct)); break; - case CHIP_RAVEN: + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): soc15_program_register_sequence(adev, golden_settings_gc_9_1, ARRAY_SIZE(golden_settings_gc_9_1)); - if (adev->rev_id >= 8) + if (adev->apu_flags & AMD_APU_IS_RAVEN2) soc15_program_register_sequence(adev, golden_settings_gc_9_1_rv2, ARRAY_SIZE(golden_settings_gc_9_1_rv2)); @@ -795,16 +996,21 @@ static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) golden_settings_gc_9_1_rv1, ARRAY_SIZE(golden_settings_gc_9_1_rv1)); break; - case CHIP_RENOIR: + case IP_VERSION(9, 3, 0): soc15_program_register_sequence(adev, golden_settings_gc_9_1_rn, ARRAY_SIZE(golden_settings_gc_9_1_rn)); return; /* for renoir, don't need common goldensetting */ + case IP_VERSION(9, 4, 2): + gfx_v9_4_2_init_golden_registers(adev, + adev->smuio.funcs->get_die_id(adev)); + break; default: break; } - if (adev->asic_type != CHIP_ARCTURUS) + if ((adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 1)) && + (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 2))) soc15_program_register_sequence(adev, golden_settings_gc_9_x_common, (const u32)ARRAY_SIZE(golden_settings_gc_9_x_common)); } @@ -905,7 +1111,8 @@ static int gfx_v9_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) gpu_addr = adev->wb.gpu_addr + (index * 4); adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD); memset(&ib, 0, sizeof(ib)); - r = amdgpu_ib_get(adev, NULL, 16, &ib); + r = amdgpu_ib_get(adev, NULL, 16, + AMDGPU_IB_POOL_DIRECT, &ib); if (r) goto err1; @@ -987,15 +1194,15 @@ static void gfx_v9_0_check_fw_write_wait(struct amdgpu_device *adev) adev->gfx.me_fw_write_wait = false; adev->gfx.mec_fw_write_wait = false; - if ((adev->gfx.mec_fw_version < 0x000001a5) || + if ((adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 1)) && + ((adev->gfx.mec_fw_version < 0x000001a5) || (adev->gfx.mec_feature_version < 46) || (adev->gfx.pfp_fw_version < 0x000000b7) || - (adev->gfx.pfp_feature_version < 46)) - DRM_WARN_ONCE("Warning: check cp_fw_version and update it to realize \ - GRBM requires 1-cycle delay in cp firmware\n"); + (adev->gfx.pfp_feature_version < 46))) + DRM_WARN_ONCE("CP firmware version too old, please update!"); - switch (adev->asic_type) { - case CHIP_VEGA10: + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(9, 0, 1): if ((adev->gfx.me_fw_version >= 0x0000009c) && (adev->gfx.me_feature_version >= 42) && (adev->gfx.pfp_fw_version >= 0x000000b1) && @@ -1006,7 +1213,7 @@ static void gfx_v9_0_check_fw_write_wait(struct amdgpu_device *adev) (adev->gfx.mec_feature_version >= 42)) adev->gfx.mec_fw_write_wait = true; break; - case CHIP_VEGA12: + case IP_VERSION(9, 2, 1): if ((adev->gfx.me_fw_version >= 0x0000009c) && (adev->gfx.me_feature_version >= 44) && (adev->gfx.pfp_fw_version >= 0x000000b2) && @@ -1017,7 +1224,7 @@ static void gfx_v9_0_check_fw_write_wait(struct amdgpu_device *adev) (adev->gfx.mec_feature_version >= 44)) adev->gfx.mec_fw_write_wait = true; break; - case CHIP_VEGA20: + case IP_VERSION(9, 4, 0): if ((adev->gfx.me_fw_version >= 0x0000009c) && (adev->gfx.me_feature_version >= 44) && (adev->gfx.pfp_fw_version >= 0x000000b2) && @@ -1028,7 +1235,8 @@ static void gfx_v9_0_check_fw_write_wait(struct amdgpu_device *adev) (adev->gfx.mec_feature_version >= 44)) adev->gfx.mec_fw_write_wait = true; break; - case CHIP_RAVEN: + case IP_VERSION(9, 1, 0): + case IP_VERSION(9, 2, 2): if ((adev->gfx.me_fw_version >= 0x0000009c) && (adev->gfx.me_feature_version >= 42) && (adev->gfx.pfp_fw_version >= 0x000000b1) && @@ -1040,22 +1248,83 @@ static void gfx_v9_0_check_fw_write_wait(struct amdgpu_device *adev) adev->gfx.mec_fw_write_wait = true; break; default: + adev->gfx.me_fw_write_wait = true; + adev->gfx.mec_fw_write_wait = true; break; } } +struct amdgpu_gfxoff_quirk { + u16 chip_vendor; + u16 chip_device; + u16 subsys_vendor; + u16 subsys_device; + u8 revision; +}; + +static const struct amdgpu_gfxoff_quirk amdgpu_gfxoff_quirk_list[] = { + /* https://bugzilla.kernel.org/show_bug.cgi?id=204689 */ + { 0x1002, 0x15dd, 0x1002, 0x15dd, 0xc8 }, + /* https://bugzilla.kernel.org/show_bug.cgi?id=207171 */ + { 0x1002, 0x15dd, 0x103c, 0x83e7, 0xd3 }, + /* GFXOFF is unstable on C6 parts with a VBIOS 113-RAVEN-114 */ + { 0x1002, 0x15dd, 0x1002, 0x15dd, 0xc6 }, + { 0, 0, 0, 0, 0 }, +}; + +static bool gfx_v9_0_should_disable_gfxoff(struct pci_dev *pdev) +{ + const struct amdgpu_gfxoff_quirk *p = amdgpu_gfxoff_quirk_list; + + while (p && p->chip_device != 0) { + if (pdev->vendor == p->chip_vendor && + pdev->device == p->chip_device && + pdev->subsystem_vendor == p->subsys_vendor && + pdev->subsystem_device == p->subsys_device && + pdev->revision == p->revision) { + return true; + } + ++p; + } + return false; +} + +static bool is_raven_kicker(struct amdgpu_device *adev) +{ + if (adev->pm.fw_version >= 0x41e2b) + return true; + else + return false; +} + +static bool check_if_enlarge_doorbell_range(struct amdgpu_device *adev) +{ + if ((adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 3, 0)) && + (adev->gfx.me_fw_version >= 0x000000a5) && + (adev->gfx.me_feature_version >= 52)) + return true; + else + return false; +} + static void gfx_v9_0_check_if_need_gfxoff(struct amdgpu_device *adev) { - switch (adev->asic_type) { - case CHIP_VEGA10: - case CHIP_VEGA12: - case CHIP_VEGA20: + if (gfx_v9_0_should_disable_gfxoff(adev->pdev)) + adev->pm.pp_feature &= ~PP_GFXOFF_MASK; + + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(9, 0, 1): + case IP_VERSION(9, 2, 1): + case IP_VERSION(9, 4, 0): break; - case CHIP_RAVEN: - if (!(adev->rev_id >= 0x8 || - adev->pdev->device == 0x15d8) && - (adev->pm.fw_version < 0x41e2b || /* not raven1 fresh */ - !adev->gfx.rlc.is_rlc_v2_1)) /* without rlc save restore ucodes */ + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): + if (!((adev->apu_flags & AMD_APU_IS_RAVEN2) || + (adev->apu_flags & AMD_APU_IS_PICASSO)) && + ((!is_raven_kicker(adev) && + adev->gfx.rlc_fw_version < 531) || + (adev->gfx.rlc_feature_version < 1) || + !adev->gfx.rlc.is_rlc_v2_1)) adev->pm.pp_feature &= ~PP_GFXOFF_MASK; if (adev->pm.pp_feature & PP_GFXOFF_MASK) @@ -1063,7 +1332,7 @@ static void gfx_v9_0_check_if_need_gfxoff(struct amdgpu_device *adev) AMD_PG_SUPPORT_CP | AMD_PG_SUPPORT_RLC_SMU_HS; break; - case CHIP_RENOIR: + case IP_VERSION(9, 3, 0): if (adev->pm.pp_feature & PP_GFXOFF_MASK) adev->pg_flags |= AMD_PG_SUPPORT_GFX_PG | AMD_PG_SUPPORT_CP | @@ -1285,6 +1554,16 @@ out: return err; } +static bool gfx_v9_0_load_mec2_fw_bin_support(struct amdgpu_device *adev) +{ + if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2) || + adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1) || + adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 3, 0)) + return false; + + return true; +} + static int gfx_v9_0_init_cp_compute_microcode(struct amdgpu_device *adev, const char *chip_name) { @@ -1306,21 +1585,26 @@ static int gfx_v9_0_init_cp_compute_microcode(struct amdgpu_device *adev, adev->gfx.mec_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name); - err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev); - if (!err) { - err = amdgpu_ucode_validate(adev->gfx.mec2_fw); - if (err) - goto out; - cp_hdr = (const struct gfx_firmware_header_v1_0 *) - adev->gfx.mec2_fw->data; - adev->gfx.mec2_fw_version = - le32_to_cpu(cp_hdr->header.ucode_version); - adev->gfx.mec2_feature_version = - le32_to_cpu(cp_hdr->ucode_feature_version); + if (gfx_v9_0_load_mec2_fw_bin_support(adev)) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name); + err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev); + if (!err) { + err = amdgpu_ucode_validate(adev->gfx.mec2_fw); + if (err) + goto out; + cp_hdr = (const struct gfx_firmware_header_v1_0 *) + adev->gfx.mec2_fw->data; + adev->gfx.mec2_fw_version = + le32_to_cpu(cp_hdr->header.ucode_version); + adev->gfx.mec2_feature_version = + le32_to_cpu(cp_hdr->ucode_feature_version); + } else { + err = 0; + adev->gfx.mec2_fw = NULL; + } } else { - err = 0; - adev->gfx.mec2_fw = NULL; + adev->gfx.mec2_fw_version = adev->gfx.mec_fw_version; + adev->gfx.mec2_feature_version = adev->gfx.mec_feature_version; } if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { @@ -1349,8 +1633,7 @@ static int gfx_v9_0_init_cp_compute_microcode(struct amdgpu_device *adev, /* TODO: Determine if MEC2 JT FW loading can be removed for all GFX V9 asic and above */ - if (adev->asic_type != CHIP_ARCTURUS && - adev->asic_type != CHIP_RENOIR) { + if (gfx_v9_0_load_mec2_fw_bin_support(adev)) { info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC2_JT]; info->ucode_id = AMDGPU_UCODE_ID_CP_MEC2_JT; info->fw = adev->gfx.mec2_fw; @@ -1383,36 +1666,43 @@ static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) DRM_DEBUG("\n"); - switch (adev->asic_type) { - case CHIP_VEGA10: + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(9, 0, 1): chip_name = "vega10"; break; - case CHIP_VEGA12: + case IP_VERSION(9, 2, 1): chip_name = "vega12"; break; - case CHIP_VEGA20: + case IP_VERSION(9, 4, 0): chip_name = "vega20"; break; - case CHIP_RAVEN: - if (adev->rev_id >= 8) + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): + if (adev->apu_flags & AMD_APU_IS_RAVEN2) chip_name = "raven2"; - else if (adev->pdev->device == 0x15d8) + else if (adev->apu_flags & AMD_APU_IS_PICASSO) chip_name = "picasso"; else chip_name = "raven"; break; - case CHIP_ARCTURUS: + case IP_VERSION(9, 4, 1): chip_name = "arcturus"; break; - case CHIP_RENOIR: - chip_name = "renoir"; + case IP_VERSION(9, 3, 0): + if (adev->apu_flags & AMD_APU_IS_RENOIR) + chip_name = "renoir"; + else + chip_name = "green_sardine"; + break; + case IP_VERSION(9, 4, 2): + chip_name = "aldebaran"; break; default: BUG(); } /* No CPG in Arcturus */ - if (adev->asic_type != CHIP_ARCTURUS) { + if (adev->gfx.num_gfx_rings) { r = gfx_v9_0_init_cp_gfx_microcode(adev, chip_name); if (r) return r; @@ -1508,7 +1798,7 @@ static void gfx_v9_0_init_always_on_cu_mask(struct amdgpu_device *adev) if (adev->flags & AMD_IS_APU) always_on_cu_num = 4; - else if (adev->asic_type == CHIP_VEGA12) + else if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 2, 1)) always_on_cu_num = 8; else always_on_cu_num = 12; @@ -1647,7 +1937,10 @@ static void gfx_v9_0_enable_lbpw(struct amdgpu_device *adev, bool enable) static int gfx_v9_0_cp_jump_table_num(struct amdgpu_device *adev) { - return 5; + if (gfx_v9_0_load_mec2_fw_bin_support(adev)) + return 5; + else + return 4; } static int gfx_v9_0_rlc_init(struct amdgpu_device *adev) @@ -1666,7 +1959,7 @@ static int gfx_v9_0_rlc_init(struct amdgpu_device *adev) return r; } - if (adev->asic_type == CHIP_RAVEN || adev->asic_type == CHIP_RENOIR) { + if (adev->flags & AMD_IS_APU) { /* TODO: double check the cp_table_size for RV */ adev->gfx.rlc.cp_table_size = ALIGN(96 * 5 * 4, 2048) + (64 * 1024); /* JT + GDS */ r = amdgpu_gfx_rlc_init_cpt(adev); @@ -1674,17 +1967,22 @@ static int gfx_v9_0_rlc_init(struct amdgpu_device *adev) return r; } - switch (adev->asic_type) { - case CHIP_RAVEN: + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): gfx_v9_0_init_lbpw(adev); break; - case CHIP_VEGA20: + case IP_VERSION(9, 4, 0): gfx_v9_4_init_lbpw(adev); break; default: break; } + /* init spm vmid with 0xf */ + if (adev->gfx.rlc.funcs->update_spm_vmid) + adev->gfx.rlc.funcs->update_spm_vmid(adev, 0xf); + return 0; } @@ -1710,29 +2008,30 @@ static int gfx_v9_0_mec_init(struct amdgpu_device *adev) /* take ownership of the relevant compute queues */ amdgpu_gfx_compute_queue_acquire(adev); mec_hpd_size = adev->gfx.num_compute_rings * GFX9_MEC_HPD_SIZE; + if (mec_hpd_size) { + r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.mec.hpd_eop_obj, + &adev->gfx.mec.hpd_eop_gpu_addr, + (void **)&hpd); + if (r) { + dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r); + gfx_v9_0_mec_fini(adev); + return r; + } - r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.mec.hpd_eop_obj, - &adev->gfx.mec.hpd_eop_gpu_addr, - (void **)&hpd); - if (r) { - dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r); - gfx_v9_0_mec_fini(adev); - return r; - } - - memset(hpd, 0, adev->gfx.mec.hpd_eop_obj->tbo.mem.size); + memset(hpd, 0, mec_hpd_size); - amdgpu_bo_kunmap(adev->gfx.mec.hpd_eop_obj); - amdgpu_bo_unreserve(adev->gfx.mec.hpd_eop_obj); + amdgpu_bo_kunmap(adev->gfx.mec.hpd_eop_obj); + amdgpu_bo_unreserve(adev->gfx.mec.hpd_eop_obj); + } mec_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data; fw_data = (const __le32 *) (adev->gfx.mec_fw->data + le32_to_cpu(mec_hdr->header.ucode_array_offset_bytes)); - fw_size = le32_to_cpu(mec_hdr->header.ucode_size_bytes) / 4; + fw_size = le32_to_cpu(mec_hdr->header.ucode_size_bytes); r = amdgpu_bo_create_reserved(adev, mec_hdr->header.ucode_size_bytes, PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, @@ -1755,7 +2054,7 @@ static int gfx_v9_0_mec_init(struct amdgpu_device *adev) static uint32_t wave_read_ind(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t address) { - WREG32_SOC15(GC, 0, mmSQ_IND_INDEX, + WREG32_SOC15_RLC(GC, 0, mmSQ_IND_INDEX, (wave << SQ_IND_INDEX__WAVE_ID__SHIFT) | (simd << SQ_IND_INDEX__SIMD_ID__SHIFT) | (address << SQ_IND_INDEX__INDEX__SHIFT) | @@ -1767,7 +2066,7 @@ static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t thread, uint32_t regno, uint32_t num, uint32_t *out) { - WREG32_SOC15(GC, 0, mmSQ_IND_INDEX, + WREG32_SOC15_RLC(GC, 0, mmSQ_IND_INDEX, (wave << SQ_IND_INDEX__WAVE_ID__SHIFT) | (simd << SQ_IND_INDEX__SIMD_ID__SHIFT) | (regno << SQ_IND_INDEX__INDEX__SHIFT) | @@ -1796,6 +2095,7 @@ static void gfx_v9_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, u dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_IB_STS); dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_IB_DBG0); dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_M0); + dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_MODE); } static void gfx_v9_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd, @@ -1824,14 +2124,20 @@ static void gfx_v9_0_select_me_pipe_q(struct amdgpu_device *adev, } static const struct amdgpu_gfx_funcs gfx_v9_0_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, + .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, +}; + +static const struct amdgpu_gfx_ras_funcs gfx_v9_0_ras_funcs = { + .ras_late_init = amdgpu_gfx_ras_late_init, + .ras_fini = amdgpu_gfx_ras_fini, .ras_error_inject = &gfx_v9_0_ras_error_inject, - .query_ras_error_count = &gfx_v9_0_query_ras_error_count + .query_ras_error_count = &gfx_v9_0_query_ras_error_count, + .reset_ras_error_count = &gfx_v9_0_reset_ras_error_count, }; static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) @@ -1841,8 +2147,8 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) adev->gfx.funcs = &gfx_v9_0_gfx_funcs; - switch (adev->asic_type) { - case CHIP_VEGA10: + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(9, 0, 1): 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; @@ -1850,7 +2156,7 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) adev->gfx.config.sc_earlyz_tile_fifo_size = 0x4C0; gb_addr_config = VEGA10_GB_ADDR_CONFIG_GOLDEN; break; - case CHIP_VEGA12: + case IP_VERSION(9, 2, 1): 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; @@ -1859,7 +2165,8 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) gb_addr_config = VEGA12_GB_ADDR_CONFIG_GOLDEN; DRM_INFO("fix gfx.config for vega12\n"); break; - case CHIP_VEGA20: + case IP_VERSION(9, 4, 0): + adev->gfx.ras_funcs = &gfx_v9_0_ras_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; @@ -1873,18 +2180,20 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) if (err) return err; break; - case CHIP_RAVEN: + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): 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; adev->gfx.config.sc_hiz_tile_fifo_size = 0x30; adev->gfx.config.sc_earlyz_tile_fifo_size = 0x4C0; - if (adev->rev_id >= 8) + if (adev->apu_flags & AMD_APU_IS_RAVEN2) gb_addr_config = RAVEN2_GB_ADDR_CONFIG_GOLDEN; else gb_addr_config = RAVEN_GB_ADDR_CONFIG_GOLDEN; break; - case CHIP_ARCTURUS: + case IP_VERSION(9, 4, 1): + adev->gfx.ras_funcs = &gfx_v9_4_ras_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; @@ -1894,7 +2203,7 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) gb_addr_config &= ~0xf3e777ff; gb_addr_config |= 0x22014042; break; - case CHIP_RENOIR: + case IP_VERSION(9, 3, 0): 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; @@ -1904,6 +2213,21 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) gb_addr_config &= ~0xf3e777ff; gb_addr_config |= 0x22010042; break; + case IP_VERSION(9, 4, 2): + adev->gfx.ras_funcs = &gfx_v9_4_2_ras_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; + adev->gfx.config.sc_hiz_tile_fifo_size = 0x30; + adev->gfx.config.sc_earlyz_tile_fifo_size = 0x4C0; + gb_addr_config = RREG32_SOC15(GC, 0, mmGB_ADDR_CONFIG); + gb_addr_config &= ~0xf3e777ff; + gb_addr_config |= 0x22014042; + /* check vbios table if gpu info is not available */ + err = amdgpu_atomfirmware_get_gfx_info(adev); + if (err) + return err; + break; default: BUG(); break; @@ -1952,9 +2276,9 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) static int gfx_v9_0_compute_ring_init(struct amdgpu_device *adev, int ring_id, int mec, int pipe, int queue) { - int r; unsigned irq_type; struct amdgpu_ring *ring = &adev->gfx.compute_ring[ring_id]; + unsigned int hw_prio; ring = &adev->gfx.compute_ring[ring_id]; @@ -1973,15 +2297,11 @@ static int gfx_v9_0_compute_ring_init(struct amdgpu_device *adev, int ring_id, irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP + ((ring->me - 1) * adev->gfx.mec.num_pipe_per_mec) + ring->pipe; - + hw_prio = amdgpu_gfx_is_high_priority_compute_queue(adev, ring) ? + AMDGPU_GFX_PIPE_PRIO_HIGH : AMDGPU_GFX_PIPE_PRIO_NORMAL; /* type-2 packets are deprecated on MEC, use type-3 instead */ - r = amdgpu_ring_init(adev, ring, 1024, - &adev->gfx.eop_irq, irq_type); - if (r) - return r; - - - return 0; + return amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type, + hw_prio, NULL); } static int gfx_v9_0_sw_init(void *handle) @@ -1991,13 +2311,15 @@ static int gfx_v9_0_sw_init(void *handle) struct amdgpu_kiq *kiq; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - switch (adev->asic_type) { - case CHIP_VEGA10: - case CHIP_VEGA12: - case CHIP_VEGA20: - case CHIP_RAVEN: - case CHIP_ARCTURUS: - case CHIP_RENOIR: + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(9, 0, 1): + case IP_VERSION(9, 2, 1): + case IP_VERSION(9, 4, 0): + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): + case IP_VERSION(9, 4, 1): + case IP_VERSION(9, 3, 0): + case IP_VERSION(9, 4, 2): adev->gfx.mec.num_mec = 2; break; default: @@ -2069,8 +2391,9 @@ static int gfx_v9_0_sw_init(void *handle) sprintf(ring->name, "gfx_%d", i); ring->use_doorbell = true; ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1; - r = amdgpu_ring_init(adev, ring, 1024, - &adev->gfx.eop_irq, AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP); + r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, + AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP, + AMDGPU_RING_PRIO_DEFAULT, NULL); if (r) return r; } @@ -2125,7 +2448,9 @@ static int gfx_v9_0_sw_fini(void *handle) int i; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - amdgpu_gfx_ras_fini(adev); + if (adev->gfx.ras_funcs && + adev->gfx.ras_funcs->ras_fini) + adev->gfx.ras_funcs->ras_fini(adev); for (i = 0; i < adev->gfx.num_gfx_rings; i++) amdgpu_ring_fini(&adev->gfx.gfx_ring[i]); @@ -2138,7 +2463,7 @@ static int gfx_v9_0_sw_fini(void *handle) gfx_v9_0_mec_fini(adev); amdgpu_bo_unref(&adev->gfx.rlc.clear_state_obj); - if (adev->asic_type == CHIP_RAVEN || adev->asic_type == CHIP_RENOIR) { + if (adev->flags & AMD_IS_APU) { amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj, &adev->gfx.rlc.cp_table_gpu_addr, (void **)&adev->gfx.rlc.cp_table_ptr); @@ -2154,7 +2479,8 @@ static void gfx_v9_0_tiling_mode_table_init(struct amdgpu_device *adev) /* TODO */ } -static void gfx_v9_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num, u32 instance) +void gfx_v9_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num, + u32 instance) { u32 data; @@ -2217,8 +2543,6 @@ static void gfx_v9_0_setup_rb(struct amdgpu_device *adev) } #define DEFAULT_SH_MEM_BASES (0x6000) -#define FIRST_COMPUTE_VMID (8) -#define LAST_COMPUTE_VMID (16) static void gfx_v9_0_init_compute_vmid(struct amdgpu_device *adev) { int i; @@ -2238,7 +2562,7 @@ static void gfx_v9_0_init_compute_vmid(struct amdgpu_device *adev) SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT; mutex_lock(&adev->srbm_mutex); - for (i = FIRST_COMPUTE_VMID; i < LAST_COMPUTE_VMID; i++) { + for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) { soc15_grbm_select(adev, 0, 0, 0, i); /* CP and shaders */ WREG32_SOC15_RLC(GC, 0, mmSH_MEM_CONFIG, sh_mem_config); @@ -2249,7 +2573,7 @@ static void gfx_v9_0_init_compute_vmid(struct amdgpu_device *adev) /* Initialize all compute VMIDs to have no GDS, GWS, or OA acccess. These should be enabled by FW for target VMIDs. */ - for (i = FIRST_COMPUTE_VMID; i < LAST_COMPUTE_VMID; i++) { + for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) { WREG32_SOC15_OFFSET(GC, 0, mmGDS_VMID0_BASE, 2 * i, 0); WREG32_SOC15_OFFSET(GC, 0, mmGDS_VMID0_SIZE, 2 * i, 0); WREG32_SOC15_OFFSET(GC, 0, mmGDS_GWS_VMID0, i, 0); @@ -2267,7 +2591,7 @@ static void gfx_v9_0_init_gds_vmid(struct amdgpu_device *adev) * the driver can enable them for graphics. VMID0 should maintain * access so that HWS firmware can save/restore entries. */ - for (vmid = 1; vmid < 16; vmid++) { + for (vmid = 1; vmid < AMDGPU_NUM_VMID; vmid++) { WREG32_SOC15_OFFSET(GC, 0, mmGDS_VMID0_BASE, 2 * vmid, 0); WREG32_SOC15_OFFSET(GC, 0, mmGDS_VMID0_SIZE, 2 * vmid, 0); WREG32_SOC15_OFFSET(GC, 0, mmGDS_GWS_VMID0, vmid, 0); @@ -2275,6 +2599,22 @@ static void gfx_v9_0_init_gds_vmid(struct amdgpu_device *adev) } } +static void gfx_v9_0_init_sq_config(struct amdgpu_device *adev) +{ + uint32_t tmp; + + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(9, 4, 1): + tmp = RREG32_SOC15(GC, 0, mmSQ_CONFIG); + tmp = REG_SET_FIELD(tmp, SQ_CONFIG, + DISABLE_BARRIER_WAITCNT, 1); + WREG32_SOC15(GC, 0, mmSQ_CONFIG, tmp); + break; + default: + break; + } +} + static void gfx_v9_0_constants_init(struct amdgpu_device *adev) { u32 tmp; @@ -2298,14 +2638,14 @@ static void gfx_v9_0_constants_init(struct amdgpu_device *adev) tmp = REG_SET_FIELD(0, SH_MEM_CONFIG, ALIGNMENT_MODE, SH_MEM_ALIGNMENT_MODE_UNALIGNED); tmp = REG_SET_FIELD(tmp, SH_MEM_CONFIG, RETRY_DISABLE, - !!amdgpu_noretry); + !!adev->gmc.noretry); WREG32_SOC15_RLC(GC, 0, mmSH_MEM_CONFIG, tmp); WREG32_SOC15_RLC(GC, 0, mmSH_MEM_BASES, 0); } else { tmp = REG_SET_FIELD(0, SH_MEM_CONFIG, ALIGNMENT_MODE, SH_MEM_ALIGNMENT_MODE_UNALIGNED); tmp = REG_SET_FIELD(tmp, SH_MEM_CONFIG, RETRY_DISABLE, - !!amdgpu_noretry); + !!adev->gmc.noretry); WREG32_SOC15_RLC(GC, 0, mmSH_MEM_CONFIG, tmp); tmp = REG_SET_FIELD(0, SH_MEM_BASES, PRIVATE_BASE, (adev->gmc.private_aperture_start >> 48)); @@ -2320,6 +2660,7 @@ static void gfx_v9_0_constants_init(struct amdgpu_device *adev) gfx_v9_0_init_compute_vmid(adev); gfx_v9_0_init_gds_vmid(adev); + gfx_v9_0_init_sq_config(adev); } static void gfx_v9_0_wait_for_rlc_serdes(struct amdgpu_device *adev) @@ -2363,12 +2704,17 @@ static void gfx_v9_0_wait_for_rlc_serdes(struct amdgpu_device *adev) static void gfx_v9_0_enable_gui_idle_interrupt(struct amdgpu_device *adev, bool enable) { - u32 tmp = RREG32_SOC15(GC, 0, mmCP_INT_CNTL_RING0); + u32 tmp; + + /* These interrupts should be enabled to drive DS clock */ + + tmp= RREG32_SOC15(GC, 0, mmCP_INT_CNTL_RING0); tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE, enable ? 1 : 0); tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_EMPTY_INT_ENABLE, enable ? 1 : 0); tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE, enable ? 1 : 0); - tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, GFX_IDLE_INT_ENABLE, enable ? 1 : 0); + if(adev->gfx.num_gfx_rings) + tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, GFX_IDLE_INT_ENABLE, enable ? 1 : 0); WREG32_SOC15(GC, 0, mmCP_INT_CNTL_RING0, tmp); } @@ -2537,7 +2883,7 @@ static void pwr_10_0_gfxip_control_over_cgpg(struct amdgpu_device *adev, uint32_t default_data = 0; default_data = data = RREG32(SOC15_REG_OFFSET(PWR, 0, mmPWR_MISC_CNTL_STATUS)); - if (enable == true) { + if (enable) { /* enable GFXIP control over CGPG */ data |= PWR_MISC_CNTL_STATUS__PWR_GFX_RLC_CGPG_EN_MASK; if(default_data != data) @@ -2593,8 +2939,8 @@ static void gfx_v9_0_init_gfx_power_gating(struct amdgpu_device *adev) /* program GRBM_REG_SAVE_GFX_IDLE_THRESHOLD to 0x55f0 */ data |= (0x55f0 << RLC_AUTO_PG_CTRL__GRBM_REG_SAVE_GFX_IDLE_THRESHOLD__SHIFT); WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_AUTO_PG_CTRL), data); - - pwr_10_0_gfxip_control_over_cgpg(adev, true); + if (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 3, 0)) + pwr_10_0_gfxip_control_over_cgpg(adev, true); } } @@ -2705,9 +3051,8 @@ static void gfx_v9_0_init_pg(struct amdgpu_device *adev) * And it's needed by gfxoff feature. */ if (adev->gfx.rlc.is_rlc_v2_1) { - if (adev->asic_type == CHIP_VEGA12 || - (adev->asic_type == CHIP_RAVEN && - adev->rev_id >= 8)) + if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 2, 1) || + (adev->apu_flags & AMD_APU_IS_RAVEN2)) gfx_v9_1_init_rlc_save_restore_list(adev); gfx_v9_0_enable_save_restore_machine(adev); } @@ -2724,7 +3069,7 @@ static void gfx_v9_0_init_pg(struct amdgpu_device *adev) } } -void gfx_v9_0_rlc_stop(struct amdgpu_device *adev) +static void gfx_v9_0_rlc_stop(struct amdgpu_device *adev) { WREG32_FIELD15(GC, 0, RLC_CNTL, RLC_ENABLE_F32, 0); gfx_v9_0_enable_gui_idle_interrupt(adev, false); @@ -2819,14 +3164,15 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev) return r; } - switch (adev->asic_type) { - case CHIP_RAVEN: + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): if (amdgpu_lbpw == 0) gfx_v9_0_enable_lbpw(adev, false); else gfx_v9_0_enable_lbpw(adev, true); break; - case CHIP_VEGA20: + case IP_VERSION(9, 4, 0): if (amdgpu_lbpw > 0) gfx_v9_0_enable_lbpw(adev, true); else @@ -2843,16 +3189,11 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev) static void gfx_v9_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable) { - int i; u32 tmp = RREG32_SOC15(GC, 0, mmCP_ME_CNTL); tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_HALT, enable ? 0 : 1); tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_HALT, enable ? 0 : 1); tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_HALT, enable ? 0 : 1); - if (!enable) { - for (i = 0; i < adev->gfx.num_gfx_rings; i++) - adev->gfx.gfx_ring[i].sched.ready = false; - } WREG32_SOC15_RLC(GC, 0, mmCP_ME_CNTL, tmp); udelay(50); } @@ -3048,15 +3389,11 @@ static int gfx_v9_0_cp_gfx_resume(struct amdgpu_device *adev) static void gfx_v9_0_cp_compute_enable(struct amdgpu_device *adev, bool enable) { - int i; - if (enable) { WREG32_SOC15_RLC(GC, 0, mmCP_MEC_CNTL, 0); } else { WREG32_SOC15_RLC(GC, 0, mmCP_MEC_CNTL, (CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK)); - for (i = 0; i < adev->gfx.num_compute_rings; i++) - adev->gfx.compute_ring[i].sched.ready = false; adev->gfx.kiq.ring.sched.ready = false; } udelay(50); @@ -3119,72 +3456,17 @@ static void gfx_v9_0_kiq_setting(struct amdgpu_ring *ring) WREG32_SOC15_RLC(GC, 0, mmRLC_CP_SCHEDULERS, tmp); } -static int gfx_v9_0_kiq_kcq_enable(struct amdgpu_device *adev) +static void gfx_v9_0_mqd_set_priority(struct amdgpu_ring *ring, struct v9_mqd *mqd) { - struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring; - uint64_t queue_mask = 0; - int r, i; - - for (i = 0; i < AMDGPU_MAX_COMPUTE_QUEUES; ++i) { - if (!test_bit(i, adev->gfx.mec.queue_bitmap)) - continue; + struct amdgpu_device *adev = ring->adev; - /* This situation may be hit in the future if a new HW - * generation exposes more than 64 queues. If so, the - * definition of queue_mask needs updating */ - if (WARN_ON(i >= (sizeof(queue_mask)*8))) { - DRM_ERROR("Invalid KCQ enabled: %d\n", i); - break; + if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE) { + if (amdgpu_gfx_is_high_priority_compute_queue(adev, ring)) { + mqd->cp_hqd_pipe_priority = AMDGPU_GFX_PIPE_PRIO_HIGH; + mqd->cp_hqd_queue_priority = + AMDGPU_GFX_QUEUE_PRIORITY_MAXIMUM; } - - queue_mask |= (1ull << i); } - - r = amdgpu_ring_alloc(kiq_ring, (7 * adev->gfx.num_compute_rings) + 8); - if (r) { - DRM_ERROR("Failed to lock KIQ (%d).\n", r); - return r; - } - - /* set resources */ - amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_RESOURCES, 6)); - amdgpu_ring_write(kiq_ring, PACKET3_SET_RESOURCES_VMID_MASK(0) | - PACKET3_SET_RESOURCES_QUEUE_TYPE(0)); /* vmid_mask:0 queue_type:0 (KIQ) */ - amdgpu_ring_write(kiq_ring, lower_32_bits(queue_mask)); /* queue mask lo */ - amdgpu_ring_write(kiq_ring, upper_32_bits(queue_mask)); /* queue mask hi */ - amdgpu_ring_write(kiq_ring, 0); /* gws mask lo */ - amdgpu_ring_write(kiq_ring, 0); /* gws mask hi */ - amdgpu_ring_write(kiq_ring, 0); /* oac mask */ - amdgpu_ring_write(kiq_ring, 0); /* gds heap base:0, gds heap size:0 */ - for (i = 0; i < adev->gfx.num_compute_rings; i++) { - struct amdgpu_ring *ring = &adev->gfx.compute_ring[i]; - uint64_t mqd_addr = amdgpu_bo_gpu_offset(ring->mqd_obj); - uint64_t wptr_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); - - amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_MAP_QUEUES, 5)); - /* Q_sel:0, vmid:0, vidmem: 1, engine:0, num_Q:1*/ - amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */ - PACKET3_MAP_QUEUES_QUEUE_SEL(0) | /* Queue_Sel */ - PACKET3_MAP_QUEUES_VMID(0) | /* VMID */ - PACKET3_MAP_QUEUES_QUEUE(ring->queue) | - PACKET3_MAP_QUEUES_PIPE(ring->pipe) | - PACKET3_MAP_QUEUES_ME((ring->me == 1 ? 0 : 1)) | - PACKET3_MAP_QUEUES_QUEUE_TYPE(0) | /*queue_type: normal compute queue */ - PACKET3_MAP_QUEUES_ALLOC_FORMAT(0) | /* alloc format: all_on_one_pipe */ - PACKET3_MAP_QUEUES_ENGINE_SEL(0) | /* engine_sel: compute */ - PACKET3_MAP_QUEUES_NUM_QUEUES(1)); /* num_queues: must be 1 */ - amdgpu_ring_write(kiq_ring, PACKET3_MAP_QUEUES_DOORBELL_OFFSET(ring->doorbell_index)); - amdgpu_ring_write(kiq_ring, lower_32_bits(mqd_addr)); - amdgpu_ring_write(kiq_ring, upper_32_bits(mqd_addr)); - amdgpu_ring_write(kiq_ring, lower_32_bits(wptr_addr)); - amdgpu_ring_write(kiq_ring, upper_32_bits(wptr_addr)); - } - - r = amdgpu_ring_test_helper(kiq_ring); - if (r) - DRM_ERROR("KCQ enable failed\n"); - - return r; } static int gfx_v9_0_mqd_init(struct amdgpu_ring *ring) @@ -3323,8 +3605,15 @@ static int gfx_v9_0_mqd_init(struct amdgpu_ring *ring) tmp = REG_SET_FIELD(tmp, CP_HQD_IB_CONTROL, MIN_IB_AVAIL_SIZE, 3); mqd->cp_hqd_ib_control = tmp; - /* activate the queue */ - mqd->cp_hqd_active = 1; + /* set static priority for a queue/ring */ + gfx_v9_0_mqd_set_priority(ring, mqd); + mqd->cp_hqd_quantum = RREG32_SOC15(GC, 0, mmCP_HQD_QUANTUM); + + /* map_queues packet doesn't need activate the queue, + * so only kiq need set this field. + */ + if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ) + mqd->cp_hqd_active = 1; return 0; } @@ -3405,7 +3694,16 @@ static int gfx_v9_0_kiq_init_register(struct amdgpu_ring *ring) if (ring->use_doorbell) { WREG32_SOC15(GC, 0, mmCP_MEC_DOORBELL_RANGE_LOWER, (adev->doorbell_index.kiq * 2) << 2); - WREG32_SOC15(GC, 0, mmCP_MEC_DOORBELL_RANGE_UPPER, + /* If GC has entered CGPG, ringing doorbell > first page + * doesn't wakeup GC. Enlarge CP_MEC_DOORBELL_RANGE_UPPER to + * workaround this issue. And this change has to align with firmware + * update. + */ + if (check_if_enlarge_doorbell_range(adev)) + WREG32_SOC15(GC, 0, mmCP_MEC_DOORBELL_RANGE_UPPER, + (adev->doorbell.size - 4)); + else + WREG32_SOC15(GC, 0, mmCP_MEC_DOORBELL_RANGE_UPPER, (adev->doorbell_index.userqueue_end * 2) << 2); } @@ -3478,11 +3776,18 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; struct v9_mqd *mqd = ring->mqd_ptr; int mqd_idx = AMDGPU_MAX_COMPUTE_RINGS; + struct v9_mqd *tmp_mqd; gfx_v9_0_kiq_setting(ring); - if (adev->in_gpu_reset) { /* for GPU_RESET case */ - /* reset MQD to a clean status */ + /* GPU could be in bad state during probe, driver trigger the reset + * after load the SMU, in this case , the mqd is not be initialized. + * driver need to re-init the mqd. + * check mqd->cp_hqd_pq_control since this value should not be 0 + */ + tmp_mqd = (struct v9_mqd *)adev->gfx.mec.mqd_backup[mqd_idx]; + if (amdgpu_in_reset(adev) && tmp_mqd->cp_hqd_pq_control){ + /* for GPU_RESET case , reset MQD to a clean status */ if (adev->gfx.mec.mqd_backup[mqd_idx]) memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation)); @@ -3518,8 +3823,15 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; struct v9_mqd *mqd = ring->mqd_ptr; int mqd_idx = ring - &adev->gfx.compute_ring[0]; + struct v9_mqd *tmp_mqd; + + /* Same as above kiq init, driver need to re-init the mqd if mqd->cp_hqd_pq_control + * is not be initialized before + */ + tmp_mqd = (struct v9_mqd *)adev->gfx.mec.mqd_backup[mqd_idx]; - if (!adev->in_gpu_reset && !adev->in_suspend) { + if (!tmp_mqd->cp_hqd_pq_control || + (!amdgpu_in_reset(adev) && !adev->in_suspend)) { memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation)); ((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; ((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; @@ -3531,13 +3843,14 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring) if (adev->gfx.mec.mqd_backup[mqd_idx]) memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation)); - } else if (adev->in_gpu_reset) { /* for GPU_RESET case */ + } else if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */ /* reset MQD to a clean status */ if (adev->gfx.mec.mqd_backup[mqd_idx]) memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation)); /* reset ring buffer */ ring->wptr = 0; + atomic64_set((atomic64_t *)&adev->wb.wb[ring->wptr_offs], 0); amdgpu_ring_clear_ring(ring); } else { amdgpu_ring_clear_ring(ring); @@ -3593,7 +3906,7 @@ static int gfx_v9_0_kcq_resume(struct amdgpu_device *adev) goto done; } - r = gfx_v9_0_kiq_kcq_enable(adev); + r = amdgpu_gfx_enable_kcq(adev); done: return r; } @@ -3607,7 +3920,7 @@ static int gfx_v9_0_cp_resume(struct amdgpu_device *adev) gfx_v9_0_enable_gui_idle_interrupt(adev, false); if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) { - if (adev->asic_type != CHIP_ARCTURUS) { + if (adev->gfx.num_gfx_rings) { /* legacy firmware loading */ r = gfx_v9_0_cp_gfx_load_microcode(adev); if (r) @@ -3623,7 +3936,7 @@ static int gfx_v9_0_cp_resume(struct amdgpu_device *adev) if (r) return r; - if (adev->asic_type != CHIP_ARCTURUS) { + if (adev->gfx.num_gfx_rings) { r = gfx_v9_0_cp_gfx_resume(adev); if (r) return r; @@ -3633,7 +3946,7 @@ static int gfx_v9_0_cp_resume(struct amdgpu_device *adev) if (r) return r; - if (adev->asic_type != CHIP_ARCTURUS) { + if (adev->gfx.num_gfx_rings) { ring = &adev->gfx.gfx_ring[0]; r = amdgpu_ring_test_helper(ring); if (r) @@ -3650,9 +3963,27 @@ static int gfx_v9_0_cp_resume(struct amdgpu_device *adev) return 0; } +static void gfx_v9_0_init_tcp_config(struct amdgpu_device *adev) +{ + u32 tmp; + + if (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 1) && + adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 2)) + return; + + tmp = RREG32_SOC15(GC, 0, mmTCP_ADDR_CONFIG); + tmp = REG_SET_FIELD(tmp, TCP_ADDR_CONFIG, ENABLE64KHASH, + adev->df.hash_status.hash_64k); + tmp = REG_SET_FIELD(tmp, TCP_ADDR_CONFIG, ENABLE2MHASH, + adev->df.hash_status.hash_2m); + tmp = REG_SET_FIELD(tmp, TCP_ADDR_CONFIG, ENABLE1GHASH, + adev->df.hash_status.hash_1g); + WREG32_SOC15(GC, 0, mmTCP_ADDR_CONFIG, tmp); +} + static void gfx_v9_0_cp_enable(struct amdgpu_device *adev, bool enable) { - if (adev->asic_type != CHIP_ARCTURUS) + if (adev->gfx.num_gfx_rings) gfx_v9_0_cp_gfx_enable(adev, enable); gfx_v9_0_cp_compute_enable(adev, enable); } @@ -3667,6 +3998,8 @@ static int gfx_v9_0_hw_init(void *handle) gfx_v9_0_constants_init(adev); + gfx_v9_0_init_tcp_config(adev); + r = adev->gfx.rlc.funcs->resume(adev); if (r) return r; @@ -3675,35 +4008,8 @@ static int gfx_v9_0_hw_init(void *handle) if (r) return r; - return r; -} - -static int gfx_v9_0_kcq_disable(struct amdgpu_device *adev) -{ - int r, i; - struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring; - - r = amdgpu_ring_alloc(kiq_ring, 6 * adev->gfx.num_compute_rings); - if (r) - DRM_ERROR("Failed to lock KIQ (%d).\n", r); - - for (i = 0; i < adev->gfx.num_compute_rings; i++) { - struct amdgpu_ring *ring = &adev->gfx.compute_ring[i]; - - amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4)); - amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */ - PACKET3_UNMAP_QUEUES_ACTION(1) | /* RESET_QUEUES */ - PACKET3_UNMAP_QUEUES_QUEUE_SEL(0) | - PACKET3_UNMAP_QUEUES_ENGINE_SEL(0) | - PACKET3_UNMAP_QUEUES_NUM_QUEUES(1)); - amdgpu_ring_write(kiq_ring, PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index)); - amdgpu_ring_write(kiq_ring, 0); - amdgpu_ring_write(kiq_ring, 0); - amdgpu_ring_write(kiq_ring, 0); - } - r = amdgpu_ring_test_helper(kiq_ring); - if (r) - DRM_ERROR("KCQ disable failed\n"); + if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2)) + gfx_v9_4_2_set_power_brake_sequence(adev); return r; } @@ -3719,7 +4025,7 @@ static int gfx_v9_0_hw_fini(void *handle) /* DF freeze and kcq disable will fail */ if (!amdgpu_ras_intr_triggered()) /* disable KCQ to avoid CPC touch memory not valid anymore */ - gfx_v9_0_kcq_disable(adev); + amdgpu_gfx_disable_kcq(adev); if (amdgpu_sriov_vf(adev)) { gfx_v9_0_cp_gfx_enable(adev, false); @@ -3735,7 +4041,7 @@ static int gfx_v9_0_hw_fini(void *handle) /* Use deinitialize sequence from CAIL when unbinding device from driver, * otherwise KIQ is hanging when binding back */ - if (!adev->in_gpu_reset && !adev->in_suspend) { + if (!amdgpu_in_reset(adev) && !adev->in_suspend) { mutex_lock(&adev->srbm_mutex); soc15_grbm_select(adev, adev->gfx.kiq.ring.me, adev->gfx.kiq.ring.pipe, @@ -3746,8 +4052,14 @@ static int gfx_v9_0_hw_fini(void *handle) } gfx_v9_0_cp_enable(adev, false); - adev->gfx.rlc.funcs->stop(adev); + /* Skip suspend with A+A reset */ + if (adev->gmc.xgmi.connected_to_cpu && amdgpu_in_reset(adev)) { + dev_dbg(adev->dev, "Device in reset. Skipping RLC halt\n"); + return 0; + } + + adev->gfx.rlc.funcs->stop(adev); return 0; } @@ -3821,7 +4133,7 @@ static int gfx_v9_0_soft_reset(void *handle) /* stop the rlc */ adev->gfx.rlc.funcs->stop(adev); - if (adev->asic_type != CHIP_ARCTURUS) + if (adev->gfx.num_gfx_rings) /* Disable GFX parsing/prefetching */ gfx_v9_0_cp_gfx_enable(adev, false); @@ -3848,28 +4160,95 @@ static int gfx_v9_0_soft_reset(void *handle) return 0; } +static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev) +{ + signed long r, cnt = 0; + unsigned long flags; + uint32_t seq, reg_val_offs = 0; + uint64_t value = 0; + struct amdgpu_kiq *kiq = &adev->gfx.kiq; + struct amdgpu_ring *ring = &kiq->ring; + + BUG_ON(!ring->funcs->emit_rreg); + + spin_lock_irqsave(&kiq->ring_lock, flags); + if (amdgpu_device_wb_get(adev, ®_val_offs)) { + pr_err("critical bug! too many kiq readers\n"); + goto failed_unlock; + } + amdgpu_ring_alloc(ring, 32); + amdgpu_ring_write(ring, PACKET3(PACKET3_COPY_DATA, 4)); + amdgpu_ring_write(ring, 9 | /* src: register*/ + (5 << 8) | /* dst: memory */ + (1 << 16) | /* count sel */ + (1 << 20)); /* write confirm */ + amdgpu_ring_write(ring, 0); + amdgpu_ring_write(ring, 0); + amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr + + reg_val_offs * 4)); + amdgpu_ring_write(ring, upper_32_bits(adev->wb.gpu_addr + + reg_val_offs * 4)); + r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT); + if (r) + goto failed_undo; + + amdgpu_ring_commit(ring); + spin_unlock_irqrestore(&kiq->ring_lock, flags); + + r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT); + + /* don't wait anymore for gpu reset case because this way may + * block gpu_recover() routine forever, e.g. this virt_kiq_rreg + * is triggered in TTM and ttm_bo_lock_delayed_workqueue() will + * never return if we keep waiting in virt_kiq_rreg, which cause + * gpu_recover() hang there. + * + * also don't wait anymore for IRQ context + * */ + if (r < 1 && (amdgpu_in_reset(adev))) + goto failed_kiq_read; + + might_sleep(); + while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) { + msleep(MAX_KIQ_REG_BAILOUT_INTERVAL); + r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT); + } + + if (cnt > MAX_KIQ_REG_TRY) + goto failed_kiq_read; + + mb(); + value = (uint64_t)adev->wb.wb[reg_val_offs] | + (uint64_t)adev->wb.wb[reg_val_offs + 1 ] << 32ULL; + amdgpu_device_wb_free(adev, reg_val_offs); + return value; + +failed_undo: + amdgpu_ring_undo(ring); +failed_unlock: + spin_unlock_irqrestore(&kiq->ring_lock, flags); +failed_kiq_read: + if (reg_val_offs) + amdgpu_device_wb_free(adev, reg_val_offs); + pr_err("failed to read gpu clock\n"); + return ~0; +} + static uint64_t gfx_v9_0_get_gpu_clock_counter(struct amdgpu_device *adev) { uint64_t clock; + amdgpu_gfx_off_ctrl(adev, false); mutex_lock(&adev->gfx.gpu_clock_mutex); - if (adev->asic_type == CHIP_VEGA10 && amdgpu_sriov_runtime(adev)) { - uint32_t tmp, lsb, msb, i = 0; - do { - if (i != 0) - udelay(1); - tmp = RREG32_SOC15(GC, 0, mmRLC_REFCLOCK_TIMESTAMP_MSB); - lsb = RREG32_SOC15(GC, 0, mmRLC_REFCLOCK_TIMESTAMP_LSB); - msb = RREG32_SOC15(GC, 0, mmRLC_REFCLOCK_TIMESTAMP_MSB); - i++; - } while (unlikely(tmp != msb) && (i < adev->usec_timeout)); - clock = (uint64_t)lsb | ((uint64_t)msb << 32ULL); + if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 0, 1) && amdgpu_sriov_runtime(adev)) { + clock = gfx_v9_0_kiq_read_clock(adev); } else { WREG32_SOC15(GC, 0, mmRLC_CAPTURE_GPU_CLOCK_COUNT, 1); clock = (uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_LSB) | ((uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_MSB) << 32ULL); } mutex_unlock(&adev->gfx.gpu_clock_mutex); + amdgpu_gfx_off_ctrl(adev, true); return clock; } @@ -3936,33 +4315,173 @@ static const u32 sgpr_init_compute_shader[] = 0xbe800080, 0xbf810000, }; +static const u32 vgpr_init_compute_shader_arcturus[] = { + 0xd3d94000, 0x18000080, 0xd3d94001, 0x18000080, 0xd3d94002, 0x18000080, + 0xd3d94003, 0x18000080, 0xd3d94004, 0x18000080, 0xd3d94005, 0x18000080, + 0xd3d94006, 0x18000080, 0xd3d94007, 0x18000080, 0xd3d94008, 0x18000080, + 0xd3d94009, 0x18000080, 0xd3d9400a, 0x18000080, 0xd3d9400b, 0x18000080, + 0xd3d9400c, 0x18000080, 0xd3d9400d, 0x18000080, 0xd3d9400e, 0x18000080, + 0xd3d9400f, 0x18000080, 0xd3d94010, 0x18000080, 0xd3d94011, 0x18000080, + 0xd3d94012, 0x18000080, 0xd3d94013, 0x18000080, 0xd3d94014, 0x18000080, + 0xd3d94015, 0x18000080, 0xd3d94016, 0x18000080, 0xd3d94017, 0x18000080, + 0xd3d94018, 0x18000080, 0xd3d94019, 0x18000080, 0xd3d9401a, 0x18000080, + 0xd3d9401b, 0x18000080, 0xd3d9401c, 0x18000080, 0xd3d9401d, 0x18000080, + 0xd3d9401e, 0x18000080, 0xd3d9401f, 0x18000080, 0xd3d94020, 0x18000080, + 0xd3d94021, 0x18000080, 0xd3d94022, 0x18000080, 0xd3d94023, 0x18000080, + 0xd3d94024, 0x18000080, 0xd3d94025, 0x18000080, 0xd3d94026, 0x18000080, + 0xd3d94027, 0x18000080, 0xd3d94028, 0x18000080, 0xd3d94029, 0x18000080, + 0xd3d9402a, 0x18000080, 0xd3d9402b, 0x18000080, 0xd3d9402c, 0x18000080, + 0xd3d9402d, 0x18000080, 0xd3d9402e, 0x18000080, 0xd3d9402f, 0x18000080, + 0xd3d94030, 0x18000080, 0xd3d94031, 0x18000080, 0xd3d94032, 0x18000080, + 0xd3d94033, 0x18000080, 0xd3d94034, 0x18000080, 0xd3d94035, 0x18000080, + 0xd3d94036, 0x18000080, 0xd3d94037, 0x18000080, 0xd3d94038, 0x18000080, + 0xd3d94039, 0x18000080, 0xd3d9403a, 0x18000080, 0xd3d9403b, 0x18000080, + 0xd3d9403c, 0x18000080, 0xd3d9403d, 0x18000080, 0xd3d9403e, 0x18000080, + 0xd3d9403f, 0x18000080, 0xd3d94040, 0x18000080, 0xd3d94041, 0x18000080, + 0xd3d94042, 0x18000080, 0xd3d94043, 0x18000080, 0xd3d94044, 0x18000080, + 0xd3d94045, 0x18000080, 0xd3d94046, 0x18000080, 0xd3d94047, 0x18000080, + 0xd3d94048, 0x18000080, 0xd3d94049, 0x18000080, 0xd3d9404a, 0x18000080, + 0xd3d9404b, 0x18000080, 0xd3d9404c, 0x18000080, 0xd3d9404d, 0x18000080, + 0xd3d9404e, 0x18000080, 0xd3d9404f, 0x18000080, 0xd3d94050, 0x18000080, + 0xd3d94051, 0x18000080, 0xd3d94052, 0x18000080, 0xd3d94053, 0x18000080, + 0xd3d94054, 0x18000080, 0xd3d94055, 0x18000080, 0xd3d94056, 0x18000080, + 0xd3d94057, 0x18000080, 0xd3d94058, 0x18000080, 0xd3d94059, 0x18000080, + 0xd3d9405a, 0x18000080, 0xd3d9405b, 0x18000080, 0xd3d9405c, 0x18000080, + 0xd3d9405d, 0x18000080, 0xd3d9405e, 0x18000080, 0xd3d9405f, 0x18000080, + 0xd3d94060, 0x18000080, 0xd3d94061, 0x18000080, 0xd3d94062, 0x18000080, + 0xd3d94063, 0x18000080, 0xd3d94064, 0x18000080, 0xd3d94065, 0x18000080, + 0xd3d94066, 0x18000080, 0xd3d94067, 0x18000080, 0xd3d94068, 0x18000080, + 0xd3d94069, 0x18000080, 0xd3d9406a, 0x18000080, 0xd3d9406b, 0x18000080, + 0xd3d9406c, 0x18000080, 0xd3d9406d, 0x18000080, 0xd3d9406e, 0x18000080, + 0xd3d9406f, 0x18000080, 0xd3d94070, 0x18000080, 0xd3d94071, 0x18000080, + 0xd3d94072, 0x18000080, 0xd3d94073, 0x18000080, 0xd3d94074, 0x18000080, + 0xd3d94075, 0x18000080, 0xd3d94076, 0x18000080, 0xd3d94077, 0x18000080, + 0xd3d94078, 0x18000080, 0xd3d94079, 0x18000080, 0xd3d9407a, 0x18000080, + 0xd3d9407b, 0x18000080, 0xd3d9407c, 0x18000080, 0xd3d9407d, 0x18000080, + 0xd3d9407e, 0x18000080, 0xd3d9407f, 0x18000080, 0xd3d94080, 0x18000080, + 0xd3d94081, 0x18000080, 0xd3d94082, 0x18000080, 0xd3d94083, 0x18000080, + 0xd3d94084, 0x18000080, 0xd3d94085, 0x18000080, 0xd3d94086, 0x18000080, + 0xd3d94087, 0x18000080, 0xd3d94088, 0x18000080, 0xd3d94089, 0x18000080, + 0xd3d9408a, 0x18000080, 0xd3d9408b, 0x18000080, 0xd3d9408c, 0x18000080, + 0xd3d9408d, 0x18000080, 0xd3d9408e, 0x18000080, 0xd3d9408f, 0x18000080, + 0xd3d94090, 0x18000080, 0xd3d94091, 0x18000080, 0xd3d94092, 0x18000080, + 0xd3d94093, 0x18000080, 0xd3d94094, 0x18000080, 0xd3d94095, 0x18000080, + 0xd3d94096, 0x18000080, 0xd3d94097, 0x18000080, 0xd3d94098, 0x18000080, + 0xd3d94099, 0x18000080, 0xd3d9409a, 0x18000080, 0xd3d9409b, 0x18000080, + 0xd3d9409c, 0x18000080, 0xd3d9409d, 0x18000080, 0xd3d9409e, 0x18000080, + 0xd3d9409f, 0x18000080, 0xd3d940a0, 0x18000080, 0xd3d940a1, 0x18000080, + 0xd3d940a2, 0x18000080, 0xd3d940a3, 0x18000080, 0xd3d940a4, 0x18000080, + 0xd3d940a5, 0x18000080, 0xd3d940a6, 0x18000080, 0xd3d940a7, 0x18000080, + 0xd3d940a8, 0x18000080, 0xd3d940a9, 0x18000080, 0xd3d940aa, 0x18000080, + 0xd3d940ab, 0x18000080, 0xd3d940ac, 0x18000080, 0xd3d940ad, 0x18000080, + 0xd3d940ae, 0x18000080, 0xd3d940af, 0x18000080, 0xd3d940b0, 0x18000080, + 0xd3d940b1, 0x18000080, 0xd3d940b2, 0x18000080, 0xd3d940b3, 0x18000080, + 0xd3d940b4, 0x18000080, 0xd3d940b5, 0x18000080, 0xd3d940b6, 0x18000080, + 0xd3d940b7, 0x18000080, 0xd3d940b8, 0x18000080, 0xd3d940b9, 0x18000080, + 0xd3d940ba, 0x18000080, 0xd3d940bb, 0x18000080, 0xd3d940bc, 0x18000080, + 0xd3d940bd, 0x18000080, 0xd3d940be, 0x18000080, 0xd3d940bf, 0x18000080, + 0xd3d940c0, 0x18000080, 0xd3d940c1, 0x18000080, 0xd3d940c2, 0x18000080, + 0xd3d940c3, 0x18000080, 0xd3d940c4, 0x18000080, 0xd3d940c5, 0x18000080, + 0xd3d940c6, 0x18000080, 0xd3d940c7, 0x18000080, 0xd3d940c8, 0x18000080, + 0xd3d940c9, 0x18000080, 0xd3d940ca, 0x18000080, 0xd3d940cb, 0x18000080, + 0xd3d940cc, 0x18000080, 0xd3d940cd, 0x18000080, 0xd3d940ce, 0x18000080, + 0xd3d940cf, 0x18000080, 0xd3d940d0, 0x18000080, 0xd3d940d1, 0x18000080, + 0xd3d940d2, 0x18000080, 0xd3d940d3, 0x18000080, 0xd3d940d4, 0x18000080, + 0xd3d940d5, 0x18000080, 0xd3d940d6, 0x18000080, 0xd3d940d7, 0x18000080, + 0xd3d940d8, 0x18000080, 0xd3d940d9, 0x18000080, 0xd3d940da, 0x18000080, + 0xd3d940db, 0x18000080, 0xd3d940dc, 0x18000080, 0xd3d940dd, 0x18000080, + 0xd3d940de, 0x18000080, 0xd3d940df, 0x18000080, 0xd3d940e0, 0x18000080, + 0xd3d940e1, 0x18000080, 0xd3d940e2, 0x18000080, 0xd3d940e3, 0x18000080, + 0xd3d940e4, 0x18000080, 0xd3d940e5, 0x18000080, 0xd3d940e6, 0x18000080, + 0xd3d940e7, 0x18000080, 0xd3d940e8, 0x18000080, 0xd3d940e9, 0x18000080, + 0xd3d940ea, 0x18000080, 0xd3d940eb, 0x18000080, 0xd3d940ec, 0x18000080, + 0xd3d940ed, 0x18000080, 0xd3d940ee, 0x18000080, 0xd3d940ef, 0x18000080, + 0xd3d940f0, 0x18000080, 0xd3d940f1, 0x18000080, 0xd3d940f2, 0x18000080, + 0xd3d940f3, 0x18000080, 0xd3d940f4, 0x18000080, 0xd3d940f5, 0x18000080, + 0xd3d940f6, 0x18000080, 0xd3d940f7, 0x18000080, 0xd3d940f8, 0x18000080, + 0xd3d940f9, 0x18000080, 0xd3d940fa, 0x18000080, 0xd3d940fb, 0x18000080, + 0xd3d940fc, 0x18000080, 0xd3d940fd, 0x18000080, 0xd3d940fe, 0x18000080, + 0xd3d940ff, 0x18000080, 0xb07c0000, 0xbe8a00ff, 0x000000f8, 0xbf11080a, + 0x7e000280, 0x7e020280, 0x7e040280, 0x7e060280, 0x7e080280, 0x7e0a0280, + 0x7e0c0280, 0x7e0e0280, 0x808a880a, 0xbe80320a, 0xbf84fff5, 0xbf9c0000, + 0xd28c0001, 0x0001007f, 0xd28d0001, 0x0002027e, 0x10020288, 0xb88b0904, + 0xb78b4000, 0xd1196a01, 0x00001701, 0xbe8a0087, 0xbefc00c1, 0xd89c4000, + 0x00020201, 0xd89cc080, 0x00040401, 0x320202ff, 0x00000800, 0x808a810a, + 0xbf84fff8, 0xbf810000, +}; + +/* When below register arrays changed, please update gpr_reg_size, + and sec_ded_counter_reg_size in function gfx_v9_0_do_edc_gpr_workarounds, + to cover all gfx9 ASICs */ static const struct soc15_reg_entry vgpr_init_regs[] = { + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_RESOURCE_LIMITS), 0x0000000 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_X), 0x40 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Y), 4 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Z), 1 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC1), 0x3f }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC2), 0x400000 }, /* 64KB LDS */ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE0), 0xffffffff }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE1), 0xffffffff }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE2), 0xffffffff }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE3), 0xffffffff }, - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_RESOURCE_LIMITS), 0x1000000 }, /* CU_GROUP_COUNT=1 */ - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_X), 256*2 }, - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Y), 1 }, - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Z), 1 }, - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC1), 0x100007f }, /* VGPRS=15 (256 logical VGPRs, SGPRS=1 (16 SGPRs, BULKY=1 */ - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC2), 0x400000 }, /* 64KB LDS */ + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE4), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE5), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE6), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE7), 0xffffffff }, }; -static const struct soc15_reg_entry sgpr_init_regs[] = { +static const struct soc15_reg_entry vgpr_init_regs_arcturus[] = { + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_RESOURCE_LIMITS), 0x0000000 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_X), 0x40 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Y), 4 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Z), 1 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC1), 0xbf }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC2), 0x400000 }, /* 64KB LDS */ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE0), 0xffffffff }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE1), 0xffffffff }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE2), 0xffffffff }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE3), 0xffffffff }, - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_RESOURCE_LIMITS), 0x1000000 }, /* CU_GROUP_COUNT=1 */ - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_X), 256*2 }, - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Y), 1 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE4), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE5), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE6), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE7), 0xffffffff }, +}; + +static const struct soc15_reg_entry sgpr1_init_regs[] = { + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_RESOURCE_LIMITS), 0x0000000 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_X), 0x40 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Y), 8 }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Z), 1 }, - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC1), 0x340 }, /* SGPRS=13 (112 GPRS) */ + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC1), 0x240 }, /* (80 GPRS) */ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC2), 0x0 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE0), 0x000000ff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE1), 0x000000ff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE2), 0x000000ff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE3), 0x000000ff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE4), 0x000000ff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE5), 0x000000ff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE6), 0x000000ff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE7), 0x000000ff }, }; -static const struct soc15_reg_entry sec_ded_counter_registers[] = { +static const struct soc15_reg_entry sgpr2_init_regs[] = { + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_RESOURCE_LIMITS), 0x0000000 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_X), 0x40 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Y), 8 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Z), 1 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC1), 0x240 }, /* (80 GPRS) */ + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC2), 0x0 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE0), 0x0000ff00 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE1), 0x0000ff00 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE2), 0x0000ff00 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE3), 0x0000ff00 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE4), 0x0000ff00 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE5), 0x0000ff00 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE6), 0x0000ff00 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE7), 0x0000ff00 }, +}; + +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}, @@ -4050,10 +4569,19 @@ 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; + int compute_dim_x = adev->gfx.config.max_shader_engines * + adev->gfx.config.max_cu_per_sh * + adev->gfx.config.max_sh_per_se; + int sgpr_work_group_size = 5; + int gpr_reg_size = adev->gfx.config.max_shader_engines + 6; + int vgpr_init_shader_size; + const u32 *vgpr_init_shader_ptr; + const struct soc15_reg_entry *vgpr_init_regs_ptr; + /* only support when RAS is enabled */ if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX)) return 0; @@ -4062,27 +4590,40 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) if (!ring->sched.ready) return 0; + if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1)) { + vgpr_init_shader_ptr = vgpr_init_compute_shader_arcturus; + vgpr_init_shader_size = sizeof(vgpr_init_compute_shader_arcturus); + vgpr_init_regs_ptr = vgpr_init_regs_arcturus; + } else { + vgpr_init_shader_ptr = vgpr_init_compute_shader; + vgpr_init_shader_size = sizeof(vgpr_init_compute_shader); + vgpr_init_regs_ptr = vgpr_init_regs; + } + total_size = - ((ARRAY_SIZE(vgpr_init_regs) * 3) + 4 + 5 + 2) * 4; + (gpr_reg_size * 3 + 4 + 5 + 2) * 4; /* VGPRS */ total_size += - ((ARRAY_SIZE(sgpr_init_regs) * 3) + 4 + 5 + 2) * 4; + (gpr_reg_size * 3 + 4 + 5 + 2) * 4; /* SGPRS1 */ + total_size += + (gpr_reg_size * 3 + 4 + 5 + 2) * 4; /* SGPRS2 */ total_size = ALIGN(total_size, 256); vgpr_offset = total_size; - total_size += ALIGN(sizeof(vgpr_init_compute_shader), 256); + total_size += ALIGN(vgpr_init_shader_size, 256); sgpr_offset = total_size; total_size += sizeof(sgpr_init_compute_shader); /* allocate an indirect buffer to put the commands in */ memset(&ib, 0, sizeof(ib)); - r = amdgpu_ib_get(adev, NULL, total_size, &ib); + r = amdgpu_ib_get(adev, NULL, total_size, + AMDGPU_IB_POOL_DIRECT, &ib); if (r) { DRM_ERROR("amdgpu: failed to get ib (%d).\n", r); return r; } /* load the compute shaders */ - for (i = 0; i < ARRAY_SIZE(vgpr_init_compute_shader); i++) - ib.ptr[i + (vgpr_offset / 4)] = vgpr_init_compute_shader[i]; + for (i = 0; i < vgpr_init_shader_size/sizeof(u32); i++) + ib.ptr[i + (vgpr_offset / 4)] = vgpr_init_shader_ptr[i]; for (i = 0; i < ARRAY_SIZE(sgpr_init_compute_shader); i++) ib.ptr[i + (sgpr_offset / 4)] = sgpr_init_compute_shader[i]; @@ -4092,11 +4633,11 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) /* VGPR */ /* write the register state for the compute dispatch */ - for (i = 0; i < ARRAY_SIZE(vgpr_init_regs); i++) { + for (i = 0; i < gpr_reg_size; i++) { ib.ptr[ib.length_dw++] = PACKET3(PACKET3_SET_SH_REG, 1); - ib.ptr[ib.length_dw++] = SOC15_REG_ENTRY_OFFSET(vgpr_init_regs[i]) + ib.ptr[ib.length_dw++] = SOC15_REG_ENTRY_OFFSET(vgpr_init_regs_ptr[i]) - PACKET3_SET_SH_REG_START; - ib.ptr[ib.length_dw++] = vgpr_init_regs[i].reg_value; + ib.ptr[ib.length_dw++] = vgpr_init_regs_ptr[i].reg_value; } /* write the shader start address: mmCOMPUTE_PGM_LO, mmCOMPUTE_PGM_HI */ gpu_addr = (ib.gpu_addr + (u64)vgpr_offset) >> 8; @@ -4108,7 +4649,35 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) /* write dispatch packet */ ib.ptr[ib.length_dw++] = PACKET3(PACKET3_DISPATCH_DIRECT, 3); - ib.ptr[ib.length_dw++] = 128; /* x */ + ib.ptr[ib.length_dw++] = compute_dim_x * 2; /* x */ + ib.ptr[ib.length_dw++] = 1; /* y */ + ib.ptr[ib.length_dw++] = 1; /* z */ + ib.ptr[ib.length_dw++] = + REG_SET_FIELD(0, COMPUTE_DISPATCH_INITIATOR, COMPUTE_SHADER_EN, 1); + + /* write CS partial flush packet */ + ib.ptr[ib.length_dw++] = PACKET3(PACKET3_EVENT_WRITE, 0); + ib.ptr[ib.length_dw++] = EVENT_TYPE(7) | EVENT_INDEX(4); + + /* SGPR1 */ + /* write the register state for the compute dispatch */ + for (i = 0; i < gpr_reg_size; i++) { + ib.ptr[ib.length_dw++] = PACKET3(PACKET3_SET_SH_REG, 1); + ib.ptr[ib.length_dw++] = SOC15_REG_ENTRY_OFFSET(sgpr1_init_regs[i]) + - PACKET3_SET_SH_REG_START; + ib.ptr[ib.length_dw++] = sgpr1_init_regs[i].reg_value; + } + /* write the shader start address: mmCOMPUTE_PGM_LO, mmCOMPUTE_PGM_HI */ + gpu_addr = (ib.gpu_addr + (u64)sgpr_offset) >> 8; + ib.ptr[ib.length_dw++] = PACKET3(PACKET3_SET_SH_REG, 2); + ib.ptr[ib.length_dw++] = SOC15_REG_OFFSET(GC, 0, mmCOMPUTE_PGM_LO) + - PACKET3_SET_SH_REG_START; + ib.ptr[ib.length_dw++] = lower_32_bits(gpu_addr); + ib.ptr[ib.length_dw++] = upper_32_bits(gpu_addr); + + /* write dispatch packet */ + ib.ptr[ib.length_dw++] = PACKET3(PACKET3_DISPATCH_DIRECT, 3); + ib.ptr[ib.length_dw++] = compute_dim_x / 2 * sgpr_work_group_size; /* x */ ib.ptr[ib.length_dw++] = 1; /* y */ ib.ptr[ib.length_dw++] = 1; /* z */ ib.ptr[ib.length_dw++] = @@ -4118,13 +4687,13 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) ib.ptr[ib.length_dw++] = PACKET3(PACKET3_EVENT_WRITE, 0); ib.ptr[ib.length_dw++] = EVENT_TYPE(7) | EVENT_INDEX(4); - /* SGPR */ + /* SGPR2 */ /* write the register state for the compute dispatch */ - for (i = 0; i < ARRAY_SIZE(sgpr_init_regs); i++) { + for (i = 0; i < gpr_reg_size; i++) { ib.ptr[ib.length_dw++] = PACKET3(PACKET3_SET_SH_REG, 1); - ib.ptr[ib.length_dw++] = SOC15_REG_ENTRY_OFFSET(sgpr_init_regs[i]) + ib.ptr[ib.length_dw++] = SOC15_REG_ENTRY_OFFSET(sgpr2_init_regs[i]) - PACKET3_SET_SH_REG_START; - ib.ptr[ib.length_dw++] = sgpr_init_regs[i].reg_value; + ib.ptr[ib.length_dw++] = sgpr2_init_regs[i].reg_value; } /* write the shader start address: mmCOMPUTE_PGM_LO, mmCOMPUTE_PGM_HI */ gpu_addr = (ib.gpu_addr + (u64)sgpr_offset) >> 8; @@ -4136,7 +4705,7 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) /* write dispatch packet */ ib.ptr[ib.length_dw++] = PACKET3(PACKET3_DISPATCH_DIRECT, 3); - ib.ptr[ib.length_dw++] = 128; /* x */ + ib.ptr[ib.length_dw++] = compute_dim_x / 2 * sgpr_work_group_size; /* x */ ib.ptr[ib.length_dw++] = 1; /* y */ ib.ptr[ib.length_dw++] = 1; /* z */ ib.ptr[ib.length_dw++] = @@ -4160,19 +4729,6 @@ 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); - fail: amdgpu_ib_free(adev, &ib, NULL); dma_fence_put(f); @@ -4184,11 +4740,14 @@ static int gfx_v9_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (adev->asic_type == CHIP_ARCTURUS) + if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1) || + adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2)) adev->gfx.num_gfx_rings = 0; else adev->gfx.num_gfx_rings = GFX9_NUM_GFX_RINGS; - adev->gfx.num_compute_rings = AMDGPU_MAX_COMPUTE_RINGS; + adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev), + AMDGPU_MAX_COMPUTE_RINGS); + gfx_v9_0_set_kiq_pm4_funcs(adev); gfx_v9_0_set_ring_funcs(adev); gfx_v9_0_set_irq_funcs(adev); gfx_v9_0_set_gds_init(adev); @@ -4202,19 +4761,39 @@ static int gfx_v9_0_ecc_late_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; int r; - r = amdgpu_gfx_ras_late_init(adev); - if (r) - return r; - - r = gfx_v9_0_do_edc_gds_workarounds(adev); - if (r) - return r; + /* + * Temp workaround to fix the issue that CP firmware fails to + * update read pointer when CPDMA is writing clearing operation + * to GDS in suspend/resume sequence on several cards. So just + * limit this operation in cold boot sequence. + */ + if ((!adev->in_suspend) && + (adev->gds.gds_size)) { + r = gfx_v9_0_do_edc_gds_workarounds(adev); + if (r) + return r; + } /* requires IBs so do in late init after IB pool is initialized */ - r = gfx_v9_0_do_edc_gpr_workarounds(adev); + if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2)) + r = gfx_v9_4_2_do_edc_gpr_workarounds(adev); + else + r = gfx_v9_0_do_edc_gpr_workarounds(adev); + if (r) return r; + if (adev->gfx.ras_funcs && + adev->gfx.ras_funcs->ras_late_init) { + r = adev->gfx.ras_funcs->ras_late_init(adev); + if (r) + return r; + } + + if (adev->gfx.ras_funcs && + adev->gfx.ras_funcs->enable_watchdog_timer) + adev->gfx.ras_funcs->enable_watchdog_timer(adev); + return 0; } @@ -4324,7 +4903,7 @@ static void gfx_v9_0_update_medium_grain_clock_gating(struct amdgpu_device *adev /* 1 - RLC_CGTT_MGCG_OVERRIDE */ def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE); - if (adev->asic_type != CHIP_VEGA12) + if (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 2, 1)) data &= ~RLC_CGTT_MGCG_OVERRIDE__CPF_CGTT_SCLK_OVERRIDE_MASK; data &= ~(RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK | @@ -4358,7 +4937,7 @@ static void gfx_v9_0_update_medium_grain_clock_gating(struct amdgpu_device *adev /* 1 - MGCG_OVERRIDE */ def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE); - if (adev->asic_type != CHIP_VEGA12) + if (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 2, 1)) data |= RLC_CGTT_MGCG_OVERRIDE__CPF_CGTT_SCLK_OVERRIDE_MASK; data |= (RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK | @@ -4392,13 +4971,13 @@ static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev, { uint32_t data, def; - if (adev->asic_type == CHIP_ARCTURUS) + if (!adev->gfx.num_gfx_rings) return; amdgpu_gfx_rlc_enter_safe_mode(adev); /* Enable 3D CGCG/CGLS */ - if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGCG)) { + if (enable) { /* write cmd to clear cgcg/cgls ov */ def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE); /* unset CGCG override */ @@ -4410,8 +4989,12 @@ static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev, /* enable 3Dcgcg FSM(0x0000363f) */ def = RREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D); - data = (0x36 << RLC_CGCG_CGLS_CTRL_3D__CGCG_GFX_IDLE_THRESHOLD__SHIFT) | - RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK; + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGCG) + data = (0x36 << RLC_CGCG_CGLS_CTRL_3D__CGCG_GFX_IDLE_THRESHOLD__SHIFT) | + RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK; + else + data = 0x0 << RLC_CGCG_CGLS_CTRL_3D__CGCG_GFX_IDLE_THRESHOLD__SHIFT; + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGLS) data |= (0x000F << RLC_CGCG_CGLS_CTRL_3D__CGLS_REP_COMPANSAT_DELAY__SHIFT) | RLC_CGCG_CGLS_CTRL_3D__CGLS_EN_MASK; @@ -4460,7 +5043,7 @@ static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev /* enable cgcg FSM(0x0000363F) */ def = RREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL); - if (adev->asic_type == CHIP_ARCTURUS) + if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1)) data = (0x2000 << RLC_CGCG_CGLS_CTRL__CGCG_GFX_IDLE_THRESHOLD__SHIFT) | RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK; else @@ -4515,6 +5098,54 @@ static int gfx_v9_0_update_gfx_clock_gating(struct amdgpu_device *adev, return 0; } +static void gfx_v9_0_update_spm_vmid(struct amdgpu_device *adev, unsigned vmid) +{ + u32 reg, data; + + reg = SOC15_REG_OFFSET(GC, 0, mmRLC_SPM_MC_CNTL); + if (amdgpu_sriov_is_pp_one_vf(adev)) + data = RREG32_NO_KIQ(reg); + else + data = RREG32(reg); + + data &= ~RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK; + data |= (vmid & RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK) << RLC_SPM_MC_CNTL__RLC_SPM_VMID__SHIFT; + + if (amdgpu_sriov_is_pp_one_vf(adev)) + WREG32_SOC15_NO_KIQ(GC, 0, mmRLC_SPM_MC_CNTL, data); + else + WREG32_SOC15(GC, 0, mmRLC_SPM_MC_CNTL, data); +} + +static bool gfx_v9_0_check_rlcg_range(struct amdgpu_device *adev, + uint32_t offset, + struct soc15_reg_rlcg *entries, int arr_size) +{ + int i; + uint32_t reg; + + if (!entries) + return false; + + for (i = 0; i < arr_size; i++) { + const struct soc15_reg_rlcg *entry; + + entry = &entries[i]; + reg = adev->reg_offset[entry->hwip][entry->instance][entry->segment] + entry->reg; + if (offset == reg) + return true; + } + + return false; +} + +static bool gfx_v9_0_is_rlcg_access_range(struct amdgpu_device *adev, u32 offset) +{ + return gfx_v9_0_check_rlcg_range(adev, offset, + (void *)rlcg_access_gc_9_0, + ARRAY_SIZE(rlcg_access_gc_9_0)); +} + static const struct amdgpu_rlc_funcs gfx_v9_0_rlc_funcs = { .is_rlc_enabled = gfx_v9_0_is_rlc_enabled, .set_safe_mode = gfx_v9_0_set_safe_mode, @@ -4526,22 +5157,25 @@ static const struct amdgpu_rlc_funcs gfx_v9_0_rlc_funcs = { .resume = gfx_v9_0_rlc_resume, .stop = gfx_v9_0_rlc_stop, .reset = gfx_v9_0_rlc_reset, - .start = gfx_v9_0_rlc_start + .start = gfx_v9_0_rlc_start, + .update_spm_vmid = gfx_v9_0_update_spm_vmid, + .sriov_wreg = gfx_v9_0_sriov_wreg, + .is_rlcg_access_range = gfx_v9_0_is_rlcg_access_range, }; static int gfx_v9_0_set_powergating_state(void *handle, enum amd_powergating_state state) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - bool enable = (state == AMD_PG_STATE_GATE) ? true : false; + bool enable = (state == AMD_PG_STATE_GATE); - switch (adev->asic_type) { - case CHIP_RAVEN: - case CHIP_RENOIR: - if (!enable) { + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): + case IP_VERSION(9, 3, 0): + if (!enable) amdgpu_gfx_off_ctrl(adev, false); - cancel_delayed_work_sync(&adev->gfx.gfx_off_delay_work); - } + if (adev->pg_flags & AMD_PG_SUPPORT_RLC_SMU_HS) { gfx_v9_0_enable_sck_slow_down_on_power_up(adev, true); gfx_v9_0_enable_sck_slow_down_on_power_down(adev, true); @@ -4564,13 +5198,8 @@ static int gfx_v9_0_set_powergating_state(void *handle, if (enable) amdgpu_gfx_off_ctrl(adev, true); break; - case CHIP_VEGA12: - if (!enable) { - amdgpu_gfx_off_ctrl(adev, false); - cancel_delayed_work_sync(&adev->gfx.gfx_off_delay_work); - } else { - amdgpu_gfx_off_ctrl(adev, true); - } + case IP_VERSION(9, 2, 1): + amdgpu_gfx_off_ctrl(adev, enable); break; default: break; @@ -4587,15 +5216,17 @@ static int gfx_v9_0_set_clockgating_state(void *handle, if (amdgpu_sriov_vf(adev)) return 0; - switch (adev->asic_type) { - case CHIP_VEGA10: - case CHIP_VEGA12: - case CHIP_VEGA20: - case CHIP_RAVEN: - case CHIP_ARCTURUS: - case CHIP_RENOIR: + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(9, 0, 1): + case IP_VERSION(9, 2, 1): + case IP_VERSION(9, 4, 0): + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): + case IP_VERSION(9, 4, 1): + case IP_VERSION(9, 3, 0): + case IP_VERSION(9, 4, 2): gfx_v9_0_update_gfx_clock_gating(adev, - state == AMD_CG_STATE_GATE ? true : false); + state == AMD_CG_STATE_GATE); break; default: break; @@ -4612,12 +5243,12 @@ static void gfx_v9_0_get_clockgating_state(void *handle, u32 *flags) *flags = 0; /* AMD_CG_SUPPORT_GFX_MGCG */ - data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE); + data = RREG32_KIQ(SOC15_REG_OFFSET(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE)); if (!(data & RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK)) *flags |= AMD_CG_SUPPORT_GFX_MGCG; /* AMD_CG_SUPPORT_GFX_CGCG */ - data = RREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL); + data = RREG32_KIQ(SOC15_REG_OFFSET(GC, 0, mmRLC_CGCG_CGLS_CTRL)); if (data & RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK) *flags |= AMD_CG_SUPPORT_GFX_CGCG; @@ -4626,18 +5257,18 @@ static void gfx_v9_0_get_clockgating_state(void *handle, u32 *flags) *flags |= AMD_CG_SUPPORT_GFX_CGLS; /* AMD_CG_SUPPORT_GFX_RLC_LS */ - data = RREG32_SOC15(GC, 0, mmRLC_MEM_SLP_CNTL); + data = RREG32_KIQ(SOC15_REG_OFFSET(GC, 0, mmRLC_MEM_SLP_CNTL)); if (data & RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK) *flags |= AMD_CG_SUPPORT_GFX_RLC_LS | AMD_CG_SUPPORT_GFX_MGLS; /* AMD_CG_SUPPORT_GFX_CP_LS */ - data = RREG32_SOC15(GC, 0, mmCP_MEM_SLP_CNTL); + data = RREG32_KIQ(SOC15_REG_OFFSET(GC, 0, mmCP_MEM_SLP_CNTL)); if (data & CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK) *flags |= AMD_CG_SUPPORT_GFX_CP_LS | AMD_CG_SUPPORT_GFX_MGLS; - if (adev->asic_type != CHIP_ARCTURUS) { + if (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 1)) { /* AMD_CG_SUPPORT_GFX_3D_CGCG */ - data = RREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D); + data = RREG32_KIQ(SOC15_REG_OFFSET(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D)); if (data & RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK) *flags |= AMD_CG_SUPPORT_GFX_3D_CGCG; @@ -4674,7 +5305,7 @@ static void gfx_v9_0_ring_set_wptr_gfx(struct amdgpu_ring *ring) if (ring->use_doorbell) { /* XXX check if swapping is necessary on BE */ - atomic64_set((atomic64_t*)&adev->wb.wb[ring->wptr_offs], ring->wptr); + atomic64_set((atomic64_t *)&adev->wb.wb[ring->wptr_offs], ring->wptr); WDOORBELL64(ring->doorbell_index, ring->wptr); } else { WREG32_SOC15(GC, 0, mmCP_RB0_WPTR, lower_32_bits(ring->wptr)); @@ -4729,7 +5360,7 @@ static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, if (amdgpu_sriov_vf(ring->adev) && (ib->flags & AMDGPU_IB_FLAG_PREEMPT)) { control |= INDIRECT_BUFFER_PRE_ENB(1); - if (!(ib->flags & AMDGPU_IB_FLAG_CE)) + if (!(ib->flags & AMDGPU_IB_FLAG_CE) && vmid) gfx_v9_0_ring_emit_de_meta(ring); } @@ -4854,112 +5485,13 @@ static u64 gfx_v9_0_ring_get_wptr_compute(struct amdgpu_ring *ring) return wptr; } -static void gfx_v9_0_ring_set_pipe_percent(struct amdgpu_ring *ring, - bool acquire) -{ - struct amdgpu_device *adev = ring->adev; - int pipe_num, tmp, reg; - int pipe_percent = acquire ? SPI_WCL_PIPE_PERCENT_GFX__VALUE_MASK : 0x1; - - pipe_num = ring->me * adev->gfx.mec.num_pipe_per_mec + ring->pipe; - - /* first me only has 2 entries, GFX and HP3D */ - if (ring->me > 0) - pipe_num -= 2; - - reg = SOC15_REG_OFFSET(GC, 0, mmSPI_WCL_PIPE_PERCENT_GFX) + pipe_num; - tmp = RREG32(reg); - tmp = REG_SET_FIELD(tmp, SPI_WCL_PIPE_PERCENT_GFX, VALUE, pipe_percent); - WREG32(reg, tmp); -} - -static void gfx_v9_0_pipe_reserve_resources(struct amdgpu_device *adev, - struct amdgpu_ring *ring, - bool acquire) -{ - int i, pipe; - bool reserve; - struct amdgpu_ring *iring; - - mutex_lock(&adev->gfx.pipe_reserve_mutex); - pipe = amdgpu_gfx_mec_queue_to_bit(adev, ring->me, ring->pipe, 0); - if (acquire) - set_bit(pipe, adev->gfx.pipe_reserve_bitmap); - else - clear_bit(pipe, adev->gfx.pipe_reserve_bitmap); - - if (!bitmap_weight(adev->gfx.pipe_reserve_bitmap, AMDGPU_MAX_COMPUTE_QUEUES)) { - /* Clear all reservations - everyone reacquires all resources */ - for (i = 0; i < adev->gfx.num_gfx_rings; ++i) - gfx_v9_0_ring_set_pipe_percent(&adev->gfx.gfx_ring[i], - true); - - for (i = 0; i < adev->gfx.num_compute_rings; ++i) - gfx_v9_0_ring_set_pipe_percent(&adev->gfx.compute_ring[i], - true); - } else { - /* Lower all pipes without a current reservation */ - for (i = 0; i < adev->gfx.num_gfx_rings; ++i) { - iring = &adev->gfx.gfx_ring[i]; - pipe = amdgpu_gfx_mec_queue_to_bit(adev, - iring->me, - iring->pipe, - 0); - reserve = test_bit(pipe, adev->gfx.pipe_reserve_bitmap); - gfx_v9_0_ring_set_pipe_percent(iring, reserve); - } - - for (i = 0; i < adev->gfx.num_compute_rings; ++i) { - iring = &adev->gfx.compute_ring[i]; - pipe = amdgpu_gfx_mec_queue_to_bit(adev, - iring->me, - iring->pipe, - 0); - reserve = test_bit(pipe, adev->gfx.pipe_reserve_bitmap); - gfx_v9_0_ring_set_pipe_percent(iring, reserve); - } - } - - mutex_unlock(&adev->gfx.pipe_reserve_mutex); -} - -static void gfx_v9_0_hqd_set_priority(struct amdgpu_device *adev, - struct amdgpu_ring *ring, - bool acquire) -{ - uint32_t pipe_priority = acquire ? 0x2 : 0x0; - uint32_t queue_priority = acquire ? 0xf : 0x0; - - mutex_lock(&adev->srbm_mutex); - soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); - - WREG32_SOC15_RLC(GC, 0, mmCP_HQD_PIPE_PRIORITY, pipe_priority); - WREG32_SOC15_RLC(GC, 0, mmCP_HQD_QUEUE_PRIORITY, queue_priority); - - soc15_grbm_select(adev, 0, 0, 0, 0); - mutex_unlock(&adev->srbm_mutex); -} - -static void gfx_v9_0_ring_set_priority_compute(struct amdgpu_ring *ring, - enum drm_sched_priority priority) -{ - struct amdgpu_device *adev = ring->adev; - bool acquire = priority == DRM_SCHED_PRIORITY_HIGH_HW; - - if (ring->funcs->type != AMDGPU_RING_TYPE_COMPUTE) - return; - - gfx_v9_0_hqd_set_priority(adev, ring, acquire); - gfx_v9_0_pipe_reserve_resources(adev, ring, acquire); -} - static void gfx_v9_0_ring_set_wptr_compute(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; /* XXX check if swapping is necessary on BE */ if (ring->use_doorbell) { - atomic64_set((atomic64_t*)&adev->wb.wb[ring->wptr_offs], ring->wptr); + atomic64_set((atomic64_t *)&adev->wb.wb[ring->wptr_offs], ring->wptr); WDOORBELL64(ring->doorbell_index, ring->wptr); } else{ BUG(); /* only DOORBELL method supported on gfx9 now */ @@ -5040,10 +5572,13 @@ static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring) amdgpu_ring_write_multiple(ring, (void *)&de_payload, sizeof(de_payload) >> 2); } -static void gfx_v9_0_ring_emit_tmz(struct amdgpu_ring *ring, bool start) +static void gfx_v9_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start, + bool secure) { + uint32_t v = secure ? FRAME_TMZ : 0; + amdgpu_ring_write(ring, PACKET3(PACKET3_FRAME_CONTROL, 0)); - amdgpu_ring_write(ring, FRAME_CMD(start ? 0 : 1)); /* frame_end */ + amdgpu_ring_write(ring, v | FRAME_CMD(start ? 0 : 1)); } static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags) @@ -5053,8 +5588,6 @@ static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags) if (amdgpu_sriov_vf(ring->adev)) gfx_v9_0_ring_emit_ce_meta(ring); - gfx_v9_0_ring_emit_tmz(ring, true); - dw2 |= 0x80000000; /* set load_enable otherwise this package is just NOPs */ if (flags & AMDGPU_HAVE_CTX_SWITCH) { /* set load_global_config & load_global_uconfig */ @@ -5105,7 +5638,8 @@ static void gfx_v9_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne ring->ring[offset] = (ring->ring_size>>2) - offset + cur; } -static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg, + uint32_t reg_val_offs) { struct amdgpu_device *adev = ring->adev; @@ -5116,9 +5650,9 @@ static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr + - adev->virt.reg_val_offs * 4)); + reg_val_offs * 4)); amdgpu_ring_write(ring, upper_32_bits(adev->wb.gpu_addr + - adev->virt.reg_val_offs * 4)); + reg_val_offs * 4)); } static void gfx_v9_0_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg, @@ -5277,6 +5811,7 @@ static int gfx_v9_0_set_priv_inst_fault_state(struct amdgpu_device *adev, WREG32_FIELD15(GC, 0, CP_INT_CNTL_RING0, PRIV_INSTR_INT_ENABLE, state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0); + break; default: break; } @@ -5440,7 +5975,7 @@ static int gfx_v9_0_priv_inst_irq(struct amdgpu_device *adev, } -static const struct ras_gfx_subblock_reg ras_subblock_regs[] = { +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) @@ -5888,7 +6423,7 @@ 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)) @@ -6013,73 +6548,76 @@ 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); sec_count = REG_GET_FIELD(data, VM_L2_MEM_ECC_CNT, SEC_COUNT); if (sec_count) { - DRM_INFO("Instance[%d]: SubBlock %s, SEC %d\n", i, - vml2_mems[i], sec_count); + dev_info(adev->dev, "Instance[%d]: SubBlock %s, " + "SEC %d\n", i, vml2_mems[i], sec_count); err_data->ce_count += sec_count; } ded_count = REG_GET_FIELD(data, VM_L2_MEM_ECC_CNT, DED_COUNT); if (ded_count) { - DRM_INFO("Instance[%d]: SubBlock %s, DED %d\n", i, - vml2_mems[i], ded_count); + dev_info(adev->dev, "Instance[%d]: SubBlock %s, " + "DED %d\n", i, vml2_mems[i], ded_count); err_data->ue_count += ded_count; } } - 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); sec_count = REG_GET_FIELD(data, VM_L2_WALKER_MEM_ECC_CNT, SEC_COUNT); if (sec_count) { - DRM_INFO("Instance[%d]: SubBlock %s, SEC %d\n", i, - vml2_walker_mems[i], sec_count); + dev_info(adev->dev, "Instance[%d]: SubBlock %s, " + "SEC %d\n", i, vml2_walker_mems[i], sec_count); err_data->ce_count += sec_count; } ded_count = REG_GET_FIELD(data, VM_L2_WALKER_MEM_ECC_CNT, DED_COUNT); if (ded_count) { - DRM_INFO("Instance[%d]: SubBlock %s, DED %d\n", i, - vml2_walker_mems[i], ded_count); + dev_info(adev->dev, "Instance[%d]: SubBlock %s, " + "DED %d\n", i, vml2_walker_mems[i], ded_count); err_data->ue_count += ded_count; } } - 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); sec_count = (data & 0x00006000L) >> 0xd; if (sec_count) { - DRM_INFO("Instance[%d]: SubBlock %s, SEC %d\n", i, - atc_l2_cache_2m_mems[i], sec_count); + dev_info(adev->dev, "Instance[%d]: SubBlock %s, " + "SEC %d\n", i, atc_l2_cache_2m_mems[i], + sec_count); err_data->ce_count += sec_count; } } - 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); sec_count = (data & 0x00006000L) >> 0xd; if (sec_count) { - DRM_INFO("Instance[%d]: SubBlock %s, SEC %d\n", i, - atc_l2_cache_4k_mems[i], sec_count); + dev_info(adev->dev, "Instance[%d]: SubBlock %s, " + "SEC %d\n", i, atc_l2_cache_4k_mems[i], + sec_count); err_data->ce_count += sec_count; } ded_count = (data & 0x00018000L) >> 0xf; if (ded_count) { - DRM_INFO("Instance[%d]: SubBlock %s, DED %d\n", i, - atc_l2_cache_4k_mems[i], ded_count); + dev_info(adev->dev, "Instance[%d]: SubBlock %s, " + "DED %d\n", i, atc_l2_cache_4k_mems[i], + ded_count); err_data->ue_count += ded_count; } } @@ -6092,36 +6630,39 @@ 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(struct amdgpu_device *adev, + 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(ras_subblock_regs); i++) { - if(ras_subblock_regs[i].reg_offset != reg->reg_offset || - ras_subblock_regs[i].seg != reg->seg || - ras_subblock_regs[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 & - ras_subblock_regs[i].sec_count_mask) >> - ras_subblock_regs[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", - ras_subblock_regs[i].name, + dev_info(adev->dev, "GFX SubBlock %s, " + "Instance[%d][%d], SEC %d\n", + gfx_v9_0_ras_fields[i].name, se_id, inst_id, sec_cnt); *sec_count += sec_cnt; } ded_cnt = (value & - ras_subblock_regs[i].ded_count_mask) >> - ras_subblock_regs[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", - ras_subblock_regs[i].name, + dev_info(adev->dev, "GFX SubBlock %s, " + "Instance[%d][%d], DED %d\n", + gfx_v9_0_ras_fields[i].name, se_id, inst_id, ded_cnt); *ded_count += ded_cnt; @@ -6131,6 +6672,61 @@ static int __get_ras_error_count(const struct soc15_reg_entry *reg, return 0; } +static void gfx_v9_0_reset_ras_error_count(struct amdgpu_device *adev) +{ + int i, j, k; + + if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX)) + return; + + /* 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) { @@ -6139,7 +6735,7 @@ 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; @@ -6147,16 +6743,17 @@ 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], - j, k, reg_value, - &sec_count, &ded_count); + gfx_v9_0_ras_error_count(adev, + &gfx_v9_0_edc_counter_regs[i], + j, k, reg_value, + &sec_count, &ded_count); } } } @@ -6172,6 +6769,84 @@ static int gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, return 0; } +static void gfx_v9_0_emit_mem_sync(struct amdgpu_ring *ring) +{ + const unsigned int cp_coher_cntl = + PACKET3_ACQUIRE_MEM_CP_COHER_CNTL_SH_ICACHE_ACTION_ENA(1) | + PACKET3_ACQUIRE_MEM_CP_COHER_CNTL_SH_KCACHE_ACTION_ENA(1) | + PACKET3_ACQUIRE_MEM_CP_COHER_CNTL_TC_ACTION_ENA(1) | + PACKET3_ACQUIRE_MEM_CP_COHER_CNTL_TCL1_ACTION_ENA(1) | + PACKET3_ACQUIRE_MEM_CP_COHER_CNTL_TC_WB_ACTION_ENA(1); + + /* ACQUIRE_MEM -make one or more surfaces valid for use by the subsequent operations */ + amdgpu_ring_write(ring, PACKET3(PACKET3_ACQUIRE_MEM, 5)); + amdgpu_ring_write(ring, cp_coher_cntl); /* CP_COHER_CNTL */ + amdgpu_ring_write(ring, 0xffffffff); /* CP_COHER_SIZE */ + amdgpu_ring_write(ring, 0xffffff); /* CP_COHER_SIZE_HI */ + amdgpu_ring_write(ring, 0); /* CP_COHER_BASE */ + amdgpu_ring_write(ring, 0); /* CP_COHER_BASE_HI */ + amdgpu_ring_write(ring, 0x0000000A); /* POLL_INTERVAL */ +} + +static void gfx_v9_0_emit_wave_limit_cs(struct amdgpu_ring *ring, + uint32_t pipe, bool enable) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t val; + uint32_t wcl_cs_reg; + + /* mmSPI_WCL_PIPE_PERCENT_CS[0-7]_DEFAULT values are same */ + val = enable ? 0x1 : mmSPI_WCL_PIPE_PERCENT_CS0_DEFAULT; + + switch (pipe) { + case 0: + wcl_cs_reg = SOC15_REG_OFFSET(GC, 0, mmSPI_WCL_PIPE_PERCENT_CS0); + break; + case 1: + wcl_cs_reg = SOC15_REG_OFFSET(GC, 0, mmSPI_WCL_PIPE_PERCENT_CS1); + break; + case 2: + wcl_cs_reg = SOC15_REG_OFFSET(GC, 0, mmSPI_WCL_PIPE_PERCENT_CS2); + break; + case 3: + wcl_cs_reg = SOC15_REG_OFFSET(GC, 0, mmSPI_WCL_PIPE_PERCENT_CS3); + break; + default: + DRM_DEBUG("invalid pipe %d\n", pipe); + return; + } + + amdgpu_ring_emit_wreg(ring, wcl_cs_reg, val); + +} +static void gfx_v9_0_emit_wave_limit(struct amdgpu_ring *ring, bool enable) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t val; + int i; + + + /* mmSPI_WCL_PIPE_PERCENT_GFX is 7 bit multiplier register to limit + * number of gfx waves. Setting 5 bit will make sure gfx only gets + * around 25% of gpu resources. + */ + val = enable ? 0x1f : mmSPI_WCL_PIPE_PERCENT_GFX_DEFAULT; + amdgpu_ring_emit_wreg(ring, + SOC15_REG_OFFSET(GC, 0, mmSPI_WCL_PIPE_PERCENT_GFX), + val); + + /* Restrict waves for normal/low priority compute queues as well + * to get best QoS for high priority compute jobs. + * + * amdgpu controls only 1st ME(0-3 CS pipes). + */ + for (i = 0; i < adev->gfx.mec.num_pipe_per_mec; i++) { + if (i != ring->pipe) + gfx_v9_0_emit_wave_limit_cs(ring, i, enable); + + } +} + static const struct amd_ip_funcs gfx_v9_0_ip_funcs = { .name = "gfx_v9_0", .early_init = gfx_v9_0_early_init, @@ -6218,7 +6893,8 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = { 3 + /* CNTX_CTRL */ 5 + /* HDP_INVL */ 8 + 8 + /* FENCE x2 */ - 2, /* SWITCH_BUFFER */ + 2 + /* SWITCH_BUFFER */ + 7, /* gfx_v9_0_emit_mem_sync */ .emit_ib_size = 4, /* gfx_v9_0_ring_emit_ib_gfx */ .emit_ib = gfx_v9_0_ring_emit_ib_gfx, .emit_fence = gfx_v9_0_ring_emit_fence, @@ -6234,11 +6910,12 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = { .emit_cntxcntl = gfx_v9_ring_emit_cntxcntl, .init_cond_exec = gfx_v9_0_ring_emit_init_cond_exec, .patch_cond_exec = gfx_v9_0_ring_emit_patch_cond_exec, - .emit_tmz = gfx_v9_0_ring_emit_tmz, + .emit_frame_cntl = gfx_v9_0_ring_emit_frame_cntl, .emit_wreg = gfx_v9_0_ring_emit_wreg, .emit_reg_wait = gfx_v9_0_ring_emit_reg_wait, .emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait, .soft_recovery = gfx_v9_0_ring_soft_recovery, + .emit_mem_sync = gfx_v9_0_emit_mem_sync, }; static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = { @@ -6258,7 +6935,10 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = { SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 + 2 + /* gfx_v9_0_ring_emit_vm_flush */ - 8 + 8 + 8, /* gfx_v9_0_ring_emit_fence x3 for user fence, vm fence */ + 8 + 8 + 8 + /* gfx_v9_0_ring_emit_fence x3 for user fence, vm fence */ + 7 + /* gfx_v9_0_emit_mem_sync */ + 5 + /* gfx_v9_0_emit_wave_limit for updating mmSPI_WCL_PIPE_PERCENT_GFX register */ + 15, /* for updating 3 mmSPI_WCL_PIPE_PERCENT_CS registers */ .emit_ib_size = 7, /* gfx_v9_0_ring_emit_ib_compute */ .emit_ib = gfx_v9_0_ring_emit_ib_compute, .emit_fence = gfx_v9_0_ring_emit_fence, @@ -6270,10 +6950,11 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = { .test_ib = gfx_v9_0_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, - .set_priority = gfx_v9_0_ring_set_priority_compute, .emit_wreg = gfx_v9_0_ring_emit_wreg, .emit_reg_wait = gfx_v9_0_ring_emit_reg_wait, .emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait, + .emit_mem_sync = gfx_v9_0_emit_mem_sync, + .emit_wave_limit = gfx_v9_0_emit_wave_limit, }; static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = { @@ -6356,13 +7037,15 @@ static void gfx_v9_0_set_irq_funcs(struct amdgpu_device *adev) static void gfx_v9_0_set_rlc_funcs(struct amdgpu_device *adev) { - switch (adev->asic_type) { - case CHIP_VEGA10: - case CHIP_VEGA12: - case CHIP_VEGA20: - case CHIP_RAVEN: - case CHIP_ARCTURUS: - case CHIP_RENOIR: + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(9, 0, 1): + case IP_VERSION(9, 2, 1): + case IP_VERSION(9, 4, 0): + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): + case IP_VERSION(9, 4, 1): + case IP_VERSION(9, 3, 0): + case IP_VERSION(9, 4, 2): adev->gfx.rlc.funcs = &gfx_v9_0_rlc_funcs; break; default: @@ -6373,38 +7056,50 @@ static void gfx_v9_0_set_rlc_funcs(struct amdgpu_device *adev) static void gfx_v9_0_set_gds_init(struct amdgpu_device *adev) { /* init asci gds info */ - switch (adev->asic_type) { - case CHIP_VEGA10: - case CHIP_VEGA12: - case CHIP_VEGA20: + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(9, 0, 1): + case IP_VERSION(9, 2, 1): + case IP_VERSION(9, 4, 0): adev->gds.gds_size = 0x10000; break; - case CHIP_RAVEN: - case CHIP_ARCTURUS: + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): + case IP_VERSION(9, 4, 1): adev->gds.gds_size = 0x1000; break; + case IP_VERSION(9, 4, 2): + /* aldebaran removed all the GDS internal memory, + * only support GWS opcode in kernel, like barrier + * semaphore.etc */ + adev->gds.gds_size = 0; + break; default: adev->gds.gds_size = 0x10000; break; } - switch (adev->asic_type) { - case CHIP_VEGA10: - case CHIP_VEGA20: + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(9, 0, 1): + case IP_VERSION(9, 4, 0): adev->gds.gds_compute_max_wave_id = 0x7ff; break; - case CHIP_VEGA12: + case IP_VERSION(9, 2, 1): adev->gds.gds_compute_max_wave_id = 0x27f; break; - case CHIP_RAVEN: - if (adev->rev_id >= 0x8) + case IP_VERSION(9, 2, 2): + case IP_VERSION(9, 1, 0): + if (adev->apu_flags & AMD_APU_IS_RAVEN2) adev->gds.gds_compute_max_wave_id = 0x77; /* raven2 */ else adev->gds.gds_compute_max_wave_id = 0x15f; /* raven1 */ break; - case CHIP_ARCTURUS: + case IP_VERSION(9, 4, 1): adev->gds.gds_compute_max_wave_id = 0xfff; break; + case IP_VERSION(9, 4, 2): + /* deprecated for Aldebaran, no usage at all */ + adev->gds.gds_compute_max_wave_id = 0; + break; default: /* this really depends on the chip */ adev->gds.gds_compute_max_wave_id = 0x7ff; |