From 2096458a304a38ca3f983174fba7a82946c0e5dd Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 24 Nov 2015 17:43:42 -0500 Subject: [PATCH 0011/1110] drm/amdgpu: add EDC support for CZ (v3) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This adds EDC support for CZ. EDC = Error Correction and Detection This code properly initializes the EDC hardware and resets the error counts. This is done in late_init since it requires the IB pool which is not initialized during hw_init. v2: fix the IB size as noted by Felix, fix shader pgm register programming v3: use the IB for the shaders as suggested by Christian Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 331 +++++++++++++++++++++++++++++++++- 1 file changed, 330 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index d105403..bc72883 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -964,6 +964,322 @@ static int gfx_v8_0_mec_init(struct amdgpu_device *adev) return 0; } +static const u32 vgpr_init_compute_shader[] = +{ + 0x7e000209, 0x7e020208, + 0x7e040207, 0x7e060206, + 0x7e080205, 0x7e0a0204, + 0x7e0c0203, 0x7e0e0202, + 0x7e100201, 0x7e120200, + 0x7e140209, 0x7e160208, + 0x7e180207, 0x7e1a0206, + 0x7e1c0205, 0x7e1e0204, + 0x7e200203, 0x7e220202, + 0x7e240201, 0x7e260200, + 0x7e280209, 0x7e2a0208, + 0x7e2c0207, 0x7e2e0206, + 0x7e300205, 0x7e320204, + 0x7e340203, 0x7e360202, + 0x7e380201, 0x7e3a0200, + 0x7e3c0209, 0x7e3e0208, + 0x7e400207, 0x7e420206, + 0x7e440205, 0x7e460204, + 0x7e480203, 0x7e4a0202, + 0x7e4c0201, 0x7e4e0200, + 0x7e500209, 0x7e520208, + 0x7e540207, 0x7e560206, + 0x7e580205, 0x7e5a0204, + 0x7e5c0203, 0x7e5e0202, + 0x7e600201, 0x7e620200, + 0x7e640209, 0x7e660208, + 0x7e680207, 0x7e6a0206, + 0x7e6c0205, 0x7e6e0204, + 0x7e700203, 0x7e720202, + 0x7e740201, 0x7e760200, + 0x7e780209, 0x7e7a0208, + 0x7e7c0207, 0x7e7e0206, + 0xbf8a0000, 0xbf810000, +}; + +static const u32 sgpr_init_compute_shader[] = +{ + 0xbe8a0100, 0xbe8c0102, + 0xbe8e0104, 0xbe900106, + 0xbe920108, 0xbe940100, + 0xbe960102, 0xbe980104, + 0xbe9a0106, 0xbe9c0108, + 0xbe9e0100, 0xbea00102, + 0xbea20104, 0xbea40106, + 0xbea60108, 0xbea80100, + 0xbeaa0102, 0xbeac0104, + 0xbeae0106, 0xbeb00108, + 0xbeb20100, 0xbeb40102, + 0xbeb60104, 0xbeb80106, + 0xbeba0108, 0xbebc0100, + 0xbebe0102, 0xbec00104, + 0xbec20106, 0xbec40108, + 0xbec60100, 0xbec80102, + 0xbee60004, 0xbee70005, + 0xbeea0006, 0xbeeb0007, + 0xbee80008, 0xbee90009, + 0xbefc0000, 0xbf8a0000, + 0xbf810000, 0x00000000, +}; + +static const u32 vgpr_init_regs[] = +{ + mmCOMPUTE_STATIC_THREAD_MGMT_SE0, 0xffffffff, + mmCOMPUTE_RESOURCE_LIMITS, 0, + mmCOMPUTE_NUM_THREAD_X, 256*4, + mmCOMPUTE_NUM_THREAD_Y, 1, + mmCOMPUTE_NUM_THREAD_Z, 1, + mmCOMPUTE_PGM_RSRC2, 20, + mmCOMPUTE_USER_DATA_0, 0xedcedc00, + mmCOMPUTE_USER_DATA_1, 0xedcedc01, + mmCOMPUTE_USER_DATA_2, 0xedcedc02, + mmCOMPUTE_USER_DATA_3, 0xedcedc03, + mmCOMPUTE_USER_DATA_4, 0xedcedc04, + mmCOMPUTE_USER_DATA_5, 0xedcedc05, + mmCOMPUTE_USER_DATA_6, 0xedcedc06, + mmCOMPUTE_USER_DATA_7, 0xedcedc07, + mmCOMPUTE_USER_DATA_8, 0xedcedc08, + mmCOMPUTE_USER_DATA_9, 0xedcedc09, +}; + +static const u32 sgpr1_init_regs[] = +{ + mmCOMPUTE_STATIC_THREAD_MGMT_SE0, 0x0f, + mmCOMPUTE_RESOURCE_LIMITS, 0x1000000, + mmCOMPUTE_NUM_THREAD_X, 256*5, + mmCOMPUTE_NUM_THREAD_Y, 1, + mmCOMPUTE_NUM_THREAD_Z, 1, + mmCOMPUTE_PGM_RSRC2, 20, + mmCOMPUTE_USER_DATA_0, 0xedcedc00, + mmCOMPUTE_USER_DATA_1, 0xedcedc01, + mmCOMPUTE_USER_DATA_2, 0xedcedc02, + mmCOMPUTE_USER_DATA_3, 0xedcedc03, + mmCOMPUTE_USER_DATA_4, 0xedcedc04, + mmCOMPUTE_USER_DATA_5, 0xedcedc05, + mmCOMPUTE_USER_DATA_6, 0xedcedc06, + mmCOMPUTE_USER_DATA_7, 0xedcedc07, + mmCOMPUTE_USER_DATA_8, 0xedcedc08, + mmCOMPUTE_USER_DATA_9, 0xedcedc09, +}; + +static const u32 sgpr2_init_regs[] = +{ + mmCOMPUTE_STATIC_THREAD_MGMT_SE0, 0xf0, + mmCOMPUTE_RESOURCE_LIMITS, 0x1000000, + mmCOMPUTE_NUM_THREAD_X, 256*5, + mmCOMPUTE_NUM_THREAD_Y, 1, + mmCOMPUTE_NUM_THREAD_Z, 1, + mmCOMPUTE_PGM_RSRC2, 20, + mmCOMPUTE_USER_DATA_0, 0xedcedc00, + mmCOMPUTE_USER_DATA_1, 0xedcedc01, + mmCOMPUTE_USER_DATA_2, 0xedcedc02, + mmCOMPUTE_USER_DATA_3, 0xedcedc03, + mmCOMPUTE_USER_DATA_4, 0xedcedc04, + mmCOMPUTE_USER_DATA_5, 0xedcedc05, + mmCOMPUTE_USER_DATA_6, 0xedcedc06, + mmCOMPUTE_USER_DATA_7, 0xedcedc07, + mmCOMPUTE_USER_DATA_8, 0xedcedc08, + mmCOMPUTE_USER_DATA_9, 0xedcedc09, +}; + +static const u32 sec_ded_counter_registers[] = +{ + mmCPC_EDC_ATC_CNT, + mmCPC_EDC_SCRATCH_CNT, + mmCPC_EDC_UCODE_CNT, + mmCPF_EDC_ATC_CNT, + mmCPF_EDC_ROQ_CNT, + mmCPF_EDC_TAG_CNT, + mmCPG_EDC_ATC_CNT, + mmCPG_EDC_DMA_CNT, + mmCPG_EDC_TAG_CNT, + mmDC_EDC_CSINVOC_CNT, + mmDC_EDC_RESTORE_CNT, + mmDC_EDC_STATE_CNT, + mmGDS_EDC_CNT, + mmGDS_EDC_GRBM_CNT, + mmGDS_EDC_OA_DED, + mmSPI_EDC_CNT, + mmSQC_ATC_EDC_GATCL1_CNT, + mmSQC_EDC_CNT, + mmSQ_EDC_DED_CNT, + mmSQ_EDC_INFO, + mmSQ_EDC_SEC_CNT, + mmTCC_EDC_CNT, + mmTCP_ATC_EDC_GATCL1_CNT, + mmTCP_EDC_CNT, + mmTD_EDC_CNT +}; + +static int gfx_v8_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring = &adev->gfx.compute_ring[0]; + struct amdgpu_ib ib; + struct fence *f = NULL; + int r, i; + u32 tmp; + unsigned total_size, vgpr_offset, sgpr_offset; + u64 gpu_addr; + + /* only supported on CZ */ + if (adev->asic_type != CHIP_CARRIZO) + return 0; + + /* bail if the compute ring is not ready */ + if (!ring->ready) + return 0; + + tmp = RREG32(mmGB_EDC_MODE); + WREG32(mmGB_EDC_MODE, 0); + + total_size = + (((ARRAY_SIZE(vgpr_init_regs) / 2) * 3) + 4 + 5 + 2) * 4; + total_size += + (((ARRAY_SIZE(sgpr1_init_regs) / 2) * 3) + 4 + 5 + 2) * 4; + total_size += + (((ARRAY_SIZE(sgpr2_init_regs) / 2) * 3) + 4 + 5 + 2) * 4; + total_size = ALIGN(total_size, 256); + vgpr_offset = total_size; + total_size += ALIGN(sizeof(vgpr_init_compute_shader), 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(ring, NULL, total_size, &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 < ARRAY_SIZE(sgpr_init_compute_shader); i++) + ib.ptr[i + (sgpr_offset / 4)] = sgpr_init_compute_shader[i]; + + /* init the ib length to 0 */ + ib.length_dw = 0; + + /* VGPR */ + /* write the register state for the compute dispatch */ + for (i = 0; i < ARRAY_SIZE(vgpr_init_regs); i += 2) { + ib.ptr[ib.length_dw++] = PACKET3(PACKET3_SET_SH_REG, 1); + ib.ptr[ib.length_dw++] = vgpr_init_regs[i] - PACKET3_SET_SH_REG_START; + ib.ptr[ib.length_dw++] = vgpr_init_regs[i + 1]; + } + /* write the shader start address: mmCOMPUTE_PGM_LO, mmCOMPUTE_PGM_HI */ + gpu_addr = (ib.gpu_addr + (u64)vgpr_offset) >> 8; + ib.ptr[ib.length_dw++] = PACKET3(PACKET3_SET_SH_REG, 2); + ib.ptr[ib.length_dw++] = 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++] = 8; /* 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 < ARRAY_SIZE(sgpr1_init_regs); i += 2) { + ib.ptr[ib.length_dw++] = PACKET3(PACKET3_SET_SH_REG, 1); + ib.ptr[ib.length_dw++] = sgpr1_init_regs[i] - PACKET3_SET_SH_REG_START; + ib.ptr[ib.length_dw++] = sgpr1_init_regs[i + 1]; + } + /* 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++] = 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++] = 8; /* 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); + + /* SGPR2 */ + /* write the register state for the compute dispatch */ + for (i = 0; i < ARRAY_SIZE(sgpr2_init_regs); i += 2) { + ib.ptr[ib.length_dw++] = PACKET3(PACKET3_SET_SH_REG, 1); + ib.ptr[ib.length_dw++] = sgpr2_init_regs[i] - PACKET3_SET_SH_REG_START; + ib.ptr[ib.length_dw++] = sgpr2_init_regs[i + 1]; + } + /* 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++] = 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++] = 8; /* 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); + + /* shedule the ib on the ring */ + r = amdgpu_sched_ib_submit_kernel_helper(adev, ring, &ib, 1, NULL, + AMDGPU_FENCE_OWNER_UNDEFINED, + &f); + if (r) { + DRM_ERROR("amdgpu: ib submit failed (%d).\n", r); + goto fail; + } + + /* wait for the GPU to finish processing the IB */ + r = fence_wait(f, false); + if (r) { + DRM_ERROR("amdgpu: fence wait failed (%d).\n", r); + goto fail; + } + + tmp = REG_SET_FIELD(tmp, GB_EDC_MODE, DED_MODE, 2); + tmp = REG_SET_FIELD(tmp, GB_EDC_MODE, PROP_FED, 1); + WREG32(mmGB_EDC_MODE, tmp); + + tmp = RREG32(mmCC_GC_EDC_CONFIG); + tmp = REG_SET_FIELD(tmp, CC_GC_EDC_CONFIG, DIS_EDC, 0) | 1; + WREG32(mmCC_GC_EDC_CONFIG, tmp); + + + /* read back registers to clear the counters */ + for (i = 0; i < ARRAY_SIZE(sec_ded_counter_registers); i++) + RREG32(sec_ded_counter_registers[i]); + +fail: + fence_put(f); + amdgpu_ib_free(adev, &ib); + + return r; +} + static void gfx_v8_0_gpu_early_init(struct amdgpu_device *adev) { u32 gb_addr_config; @@ -4458,6 +4774,19 @@ static int gfx_v8_0_early_init(void *handle) return 0; } +static int gfx_v8_0_late_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int r; + + /* requires IBs so do in late init after IB pool is initialized */ + r = gfx_v8_0_do_edc_gpr_workarounds(adev); + if (r) + return r; + + return 0; +} + static int gfx_v8_0_set_powergating_state(void *handle, enum amd_powergating_state state) { @@ -4996,7 +5325,7 @@ static int gfx_v8_0_priv_inst_irq(struct amdgpu_device *adev, const struct amd_ip_funcs gfx_v8_0_ip_funcs = { .early_init = gfx_v8_0_early_init, - .late_init = NULL, + .late_init = gfx_v8_0_late_init, .sw_init = gfx_v8_0_sw_init, .sw_fini = gfx_v8_0_sw_fini, .hw_init = gfx_v8_0_hw_init, -- 2.7.4