diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c')
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 2575 |
1 files changed, 1321 insertions, 1254 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index aa5a50f5eac8..1c87375e1dd5 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -20,10 +20,16 @@ * OTHER DEALINGS IN THE SOFTWARE. * */ + +#include <linux/delay.h> +#include <linux/kernel.h> #include <linux/firmware.h> -#include <drm/drmP.h> +#include <linux/module.h> +#include <linux/pci.h> + #include "amdgpu.h" #include "amdgpu_gfx.h" +#include "amdgpu_ring.h" #include "vi.h" #include "vi_structs.h" #include "vid.h" @@ -43,15 +49,16 @@ #include "gca/gfx_8_0_d.h" #include "gca/gfx_8_0_enum.h" #include "gca/gfx_8_0_sh_mask.h" -#include "gca/gfx_8_0_enum.h" #include "dce/dce_10_0_d.h" #include "dce/dce_10_0_sh_mask.h" #include "smu/smu_7_1_3_d.h" +#include "ivsrcid/ivsrcid_vislands30.h" + #define GFX8_NUM_GFX_RINGS 1 -#define GFX8_MEC_HPD_SIZE 2048 +#define GFX8_MEC_HPD_SIZE 4096 #define TOPAZ_GB_ADDR_CONFIG_GOLDEN 0x22010001 #define CARRIZO_GB_ADDR_CONFIG_GOLDEN 0x22010001 @@ -124,27 +131,49 @@ MODULE_FIRMWARE("amdgpu/fiji_mec.bin"); MODULE_FIRMWARE("amdgpu/fiji_mec2.bin"); MODULE_FIRMWARE("amdgpu/fiji_rlc.bin"); -MODULE_FIRMWARE("amdgpu/polaris11_ce.bin"); -MODULE_FIRMWARE("amdgpu/polaris11_pfp.bin"); -MODULE_FIRMWARE("amdgpu/polaris11_me.bin"); -MODULE_FIRMWARE("amdgpu/polaris11_mec.bin"); -MODULE_FIRMWARE("amdgpu/polaris11_mec2.bin"); -MODULE_FIRMWARE("amdgpu/polaris11_rlc.bin"); - MODULE_FIRMWARE("amdgpu/polaris10_ce.bin"); +MODULE_FIRMWARE("amdgpu/polaris10_ce_2.bin"); MODULE_FIRMWARE("amdgpu/polaris10_pfp.bin"); +MODULE_FIRMWARE("amdgpu/polaris10_pfp_2.bin"); MODULE_FIRMWARE("amdgpu/polaris10_me.bin"); +MODULE_FIRMWARE("amdgpu/polaris10_me_2.bin"); MODULE_FIRMWARE("amdgpu/polaris10_mec.bin"); +MODULE_FIRMWARE("amdgpu/polaris10_mec_2.bin"); MODULE_FIRMWARE("amdgpu/polaris10_mec2.bin"); +MODULE_FIRMWARE("amdgpu/polaris10_mec2_2.bin"); MODULE_FIRMWARE("amdgpu/polaris10_rlc.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_ce.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_ce_2.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_pfp.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_pfp_2.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_me.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_me_2.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_mec.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_mec_2.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_mec2.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_mec2_2.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_rlc.bin"); + MODULE_FIRMWARE("amdgpu/polaris12_ce.bin"); +MODULE_FIRMWARE("amdgpu/polaris12_ce_2.bin"); MODULE_FIRMWARE("amdgpu/polaris12_pfp.bin"); +MODULE_FIRMWARE("amdgpu/polaris12_pfp_2.bin"); MODULE_FIRMWARE("amdgpu/polaris12_me.bin"); +MODULE_FIRMWARE("amdgpu/polaris12_me_2.bin"); MODULE_FIRMWARE("amdgpu/polaris12_mec.bin"); +MODULE_FIRMWARE("amdgpu/polaris12_mec_2.bin"); MODULE_FIRMWARE("amdgpu/polaris12_mec2.bin"); +MODULE_FIRMWARE("amdgpu/polaris12_mec2_2.bin"); MODULE_FIRMWARE("amdgpu/polaris12_rlc.bin"); +MODULE_FIRMWARE("amdgpu/vegam_ce.bin"); +MODULE_FIRMWARE("amdgpu/vegam_pfp.bin"); +MODULE_FIRMWARE("amdgpu/vegam_me.bin"); +MODULE_FIRMWARE("amdgpu/vegam_mec.bin"); +MODULE_FIRMWARE("amdgpu/vegam_mec2.bin"); +MODULE_FIRMWARE("amdgpu/vegam_rlc.bin"); + static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] = { {mmGDS_VMID0_BASE, mmGDS_VMID0_SIZE, mmGDS_GWS_VMID0, mmGDS_OA_VMID0}, @@ -193,8 +222,8 @@ static const u32 tonga_golden_common_all[] = mmGB_ADDR_CONFIG, 0xffffffff, 0x22011003, mmSPI_RESOURCE_RESERVE_CU_0, 0xffffffff, 0x00000800, mmSPI_RESOURCE_RESERVE_CU_1, 0xffffffff, 0x00000800, - mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00007FBF, - mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00007FAF + mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00FF7FBF, + mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00FF7FAF }; static const u32 tonga_mgcg_cgcg_init[] = @@ -276,6 +305,37 @@ static const u32 tonga_mgcg_cgcg_init[] = mmCP_MEM_SLP_CNTL, 0x00000001, 0x00000001, }; +static const u32 golden_settings_vegam_a11[] = +{ + mmCB_HW_CONTROL, 0x0001f3cf, 0x00007208, + mmCB_HW_CONTROL_2, 0x0f000000, 0x0d000000, + mmCB_HW_CONTROL_3, 0x000001ff, 0x00000040, + mmDB_DEBUG2, 0xf00fffff, 0x00000400, + mmPA_SC_ENHANCE, 0xffffffff, 0x20000001, + mmPA_SC_LINE_STIPPLE_STATE, 0x0000ff0f, 0x00000000, + mmPA_SC_RASTER_CONFIG, 0x3f3fffff, 0x3a00161a, + mmPA_SC_RASTER_CONFIG_1, 0x0000003f, 0x0000002e, + mmRLC_CGCG_CGLS_CTRL, 0x00000003, 0x0001003c, + mmRLC_CGCG_CGLS_CTRL_3D, 0xffffffff, 0x0001003c, + mmSQ_CONFIG, 0x07f80000, 0x01180000, + mmTA_CNTL_AUX, 0x000f000f, 0x000b0000, + mmTCC_CTRL, 0x00100000, 0xf31fff7f, + mmTCP_ADDR_CONFIG, 0x000003ff, 0x000000f7, + mmTCP_CHAN_STEER_HI, 0xffffffff, 0x00000000, + mmTCP_CHAN_STEER_LO, 0xffffffff, 0x32761054, + mmVGT_RESET_DEBUG, 0x00000004, 0x00000004, +}; + +static const u32 vegam_golden_common_all[] = +{ + mmGRBM_GFX_INDEX, 0xffffffff, 0xe0000000, + mmGB_ADDR_CONFIG, 0xffffffff, 0x22011003, + mmSPI_RESOURCE_RESERVE_CU_0, 0xffffffff, 0x00000800, + mmSPI_RESOURCE_RESERVE_CU_1, 0xffffffff, 0x00000800, + mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00FF7FBF, + mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00FF7FAF, +}; + static const u32 golden_settings_polaris11_a11[] = { mmCB_HW_CONTROL, 0x0000f3cf, 0x00007208, @@ -303,8 +363,8 @@ static const u32 polaris11_golden_common_all[] = mmGB_ADDR_CONFIG, 0xffffffff, 0x22011002, mmSPI_RESOURCE_RESERVE_CU_0, 0xffffffff, 0x00000800, mmSPI_RESOURCE_RESERVE_CU_1, 0xffffffff, 0x00000800, - mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00007FBF, - mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00007FAF, + mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00FF7FBF, + mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00FF7FAF, }; static const u32 golden_settings_polaris10_a11[] = @@ -336,8 +396,8 @@ static const u32 polaris10_golden_common_all[] = mmGB_ADDR_CONFIG, 0xffffffff, 0x22011003, mmSPI_RESOURCE_RESERVE_CU_0, 0xffffffff, 0x00000800, mmSPI_RESOURCE_RESERVE_CU_1, 0xffffffff, 0x00000800, - mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00007FBF, - mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00007FAF, + mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00FF7FBF, + mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00FF7FAF, }; static const u32 fiji_golden_common_all[] = @@ -348,8 +408,8 @@ static const u32 fiji_golden_common_all[] = mmGB_ADDR_CONFIG, 0xffffffff, 0x22011003, mmSPI_RESOURCE_RESERVE_CU_0, 0xffffffff, 0x00000800, mmSPI_RESOURCE_RESERVE_CU_1, 0xffffffff, 0x00000800, - mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00007FBF, - mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00007FAF, + mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00FF7FBF, + mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00FF7FAF, mmGRBM_GFX_INDEX, 0xffffffff, 0xe0000000, mmSPI_CONFIG_CNTL_1, 0x0000000f, 0x00000009, }; @@ -436,8 +496,8 @@ static const u32 iceland_golden_common_all[] = mmGB_ADDR_CONFIG, 0xffffffff, 0x22010001, mmSPI_RESOURCE_RESERVE_CU_0, 0xffffffff, 0x00000800, mmSPI_RESOURCE_RESERVE_CU_1, 0xffffffff, 0x00000800, - mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00007FBF, - mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00007FAF + mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00FF7FBF, + mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00FF7FAF }; static const u32 iceland_mgcg_cgcg_init[] = @@ -532,8 +592,8 @@ static const u32 cz_golden_common_all[] = mmGB_ADDR_CONFIG, 0xffffffff, 0x22010001, mmSPI_RESOURCE_RESERVE_CU_0, 0xffffffff, 0x00000800, mmSPI_RESOURCE_RESERVE_CU_1, 0xffffffff, 0x00000800, - mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00007FBF, - mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00007FAF + mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00FF7FBF, + mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00FF7FAF }; static const u32 cz_mgcg_cgcg_init[] = @@ -637,8 +697,8 @@ static const u32 stoney_golden_common_all[] = mmGB_ADDR_CONFIG, 0xffffffff, 0x12010001, mmSPI_RESOURCE_RESERVE_CU_0, 0xffffffff, 0x00000800, mmSPI_RESOURCE_RESERVE_CU_1, 0xffffffff, 0x00000800, - mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00007FBF, - mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00007FAF, + mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00FF7FBF, + mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00FF7FAF, }; static const u32 stoney_mgcg_cgcg_init[] = @@ -650,6 +710,17 @@ static const u32 stoney_mgcg_cgcg_init[] = mmCGTS_SM_CTRL_REG, 0xffffffff, 0x96940200, }; + +static const char * const sq_edc_source_names[] = { + "SQ_EDC_INFO_SOURCE_INVALID: No EDC error has occurred", + "SQ_EDC_INFO_SOURCE_INST: EDC source is Instruction Fetch", + "SQ_EDC_INFO_SOURCE_SGPR: EDC source is SGPR or SQC data return", + "SQ_EDC_INFO_SOURCE_VGPR: EDC source is VGPR", + "SQ_EDC_INFO_SOURCE_LDS: EDC source is LDS", + "SQ_EDC_INFO_SOURCE_GDS: EDC source is GDS", + "SQ_EDC_INFO_SOURCE_TA: EDC source is TA", +}; + static void gfx_v8_0_set_ring_funcs(struct amdgpu_device *adev); static void gfx_v8_0_set_irq_funcs(struct amdgpu_device *adev); static void gfx_v8_0_set_gds_init(struct amdgpu_device *adev); @@ -659,143 +730,138 @@ static void gfx_v8_0_get_cu_info(struct amdgpu_device *adev); static void gfx_v8_0_ring_emit_ce_meta(struct amdgpu_ring *ring); static void gfx_v8_0_ring_emit_de_meta(struct amdgpu_ring *ring); +#define CG_ACLK_CNTL__ACLK_DIVIDER_MASK 0x0000007fL +#define CG_ACLK_CNTL__ACLK_DIVIDER__SHIFT 0x00000000L + static void gfx_v8_0_init_golden_registers(struct amdgpu_device *adev) { + uint32_t data; + switch (adev->asic_type) { case CHIP_TOPAZ: - amdgpu_program_register_sequence(adev, - iceland_mgcg_cgcg_init, - (const u32)ARRAY_SIZE(iceland_mgcg_cgcg_init)); - amdgpu_program_register_sequence(adev, - golden_settings_iceland_a11, - (const u32)ARRAY_SIZE(golden_settings_iceland_a11)); - amdgpu_program_register_sequence(adev, - iceland_golden_common_all, - (const u32)ARRAY_SIZE(iceland_golden_common_all)); + amdgpu_device_program_register_sequence(adev, + iceland_mgcg_cgcg_init, + ARRAY_SIZE(iceland_mgcg_cgcg_init)); + amdgpu_device_program_register_sequence(adev, + golden_settings_iceland_a11, + ARRAY_SIZE(golden_settings_iceland_a11)); + amdgpu_device_program_register_sequence(adev, + iceland_golden_common_all, + ARRAY_SIZE(iceland_golden_common_all)); break; case CHIP_FIJI: - amdgpu_program_register_sequence(adev, - fiji_mgcg_cgcg_init, - (const u32)ARRAY_SIZE(fiji_mgcg_cgcg_init)); - amdgpu_program_register_sequence(adev, - golden_settings_fiji_a10, - (const u32)ARRAY_SIZE(golden_settings_fiji_a10)); - amdgpu_program_register_sequence(adev, - fiji_golden_common_all, - (const u32)ARRAY_SIZE(fiji_golden_common_all)); + amdgpu_device_program_register_sequence(adev, + fiji_mgcg_cgcg_init, + ARRAY_SIZE(fiji_mgcg_cgcg_init)); + amdgpu_device_program_register_sequence(adev, + golden_settings_fiji_a10, + ARRAY_SIZE(golden_settings_fiji_a10)); + amdgpu_device_program_register_sequence(adev, + fiji_golden_common_all, + ARRAY_SIZE(fiji_golden_common_all)); break; case CHIP_TONGA: - amdgpu_program_register_sequence(adev, - tonga_mgcg_cgcg_init, - (const u32)ARRAY_SIZE(tonga_mgcg_cgcg_init)); - amdgpu_program_register_sequence(adev, - golden_settings_tonga_a11, - (const u32)ARRAY_SIZE(golden_settings_tonga_a11)); - amdgpu_program_register_sequence(adev, - tonga_golden_common_all, - (const u32)ARRAY_SIZE(tonga_golden_common_all)); + amdgpu_device_program_register_sequence(adev, + tonga_mgcg_cgcg_init, + ARRAY_SIZE(tonga_mgcg_cgcg_init)); + amdgpu_device_program_register_sequence(adev, + golden_settings_tonga_a11, + ARRAY_SIZE(golden_settings_tonga_a11)); + amdgpu_device_program_register_sequence(adev, + tonga_golden_common_all, + ARRAY_SIZE(tonga_golden_common_all)); + break; + case CHIP_VEGAM: + amdgpu_device_program_register_sequence(adev, + golden_settings_vegam_a11, + ARRAY_SIZE(golden_settings_vegam_a11)); + amdgpu_device_program_register_sequence(adev, + vegam_golden_common_all, + ARRAY_SIZE(vegam_golden_common_all)); break; case CHIP_POLARIS11: case CHIP_POLARIS12: - amdgpu_program_register_sequence(adev, - golden_settings_polaris11_a11, - (const u32)ARRAY_SIZE(golden_settings_polaris11_a11)); - amdgpu_program_register_sequence(adev, - polaris11_golden_common_all, - (const u32)ARRAY_SIZE(polaris11_golden_common_all)); + amdgpu_device_program_register_sequence(adev, + golden_settings_polaris11_a11, + ARRAY_SIZE(golden_settings_polaris11_a11)); + amdgpu_device_program_register_sequence(adev, + polaris11_golden_common_all, + ARRAY_SIZE(polaris11_golden_common_all)); break; case CHIP_POLARIS10: - amdgpu_program_register_sequence(adev, - golden_settings_polaris10_a11, - (const u32)ARRAY_SIZE(golden_settings_polaris10_a11)); - amdgpu_program_register_sequence(adev, - polaris10_golden_common_all, - (const u32)ARRAY_SIZE(polaris10_golden_common_all)); - WREG32_SMC(ixCG_ACLK_CNTL, 0x0000001C); - if (adev->pdev->revision == 0xc7 && + amdgpu_device_program_register_sequence(adev, + golden_settings_polaris10_a11, + ARRAY_SIZE(golden_settings_polaris10_a11)); + amdgpu_device_program_register_sequence(adev, + polaris10_golden_common_all, + ARRAY_SIZE(polaris10_golden_common_all)); + data = RREG32_SMC(ixCG_ACLK_CNTL); + data &= ~CG_ACLK_CNTL__ACLK_DIVIDER_MASK; + data |= 0x18 << CG_ACLK_CNTL__ACLK_DIVIDER__SHIFT; + WREG32_SMC(ixCG_ACLK_CNTL, data); + if ((adev->pdev->device == 0x67DF) && (adev->pdev->revision == 0xc7) && ((adev->pdev->subsystem_device == 0xb37 && adev->pdev->subsystem_vendor == 0x1002) || (adev->pdev->subsystem_device == 0x4a8 && adev->pdev->subsystem_vendor == 0x1043) || - (adev->pdev->subsystem_device == 0x9480 && adev->pdev->subsystem_vendor == 0x1682))) { + (adev->pdev->subsystem_device == 0x9480 && adev->pdev->subsystem_vendor == 0x1680))) { amdgpu_atombios_i2c_channel_trans(adev, 0x10, 0x96, 0x1E, 0xDD); amdgpu_atombios_i2c_channel_trans(adev, 0x10, 0x96, 0x1F, 0xD0); } break; case CHIP_CARRIZO: - amdgpu_program_register_sequence(adev, - cz_mgcg_cgcg_init, - (const u32)ARRAY_SIZE(cz_mgcg_cgcg_init)); - amdgpu_program_register_sequence(adev, - cz_golden_settings_a11, - (const u32)ARRAY_SIZE(cz_golden_settings_a11)); - amdgpu_program_register_sequence(adev, - cz_golden_common_all, - (const u32)ARRAY_SIZE(cz_golden_common_all)); + amdgpu_device_program_register_sequence(adev, + cz_mgcg_cgcg_init, + ARRAY_SIZE(cz_mgcg_cgcg_init)); + amdgpu_device_program_register_sequence(adev, + cz_golden_settings_a11, + ARRAY_SIZE(cz_golden_settings_a11)); + amdgpu_device_program_register_sequence(adev, + cz_golden_common_all, + ARRAY_SIZE(cz_golden_common_all)); break; case CHIP_STONEY: - amdgpu_program_register_sequence(adev, - stoney_mgcg_cgcg_init, - (const u32)ARRAY_SIZE(stoney_mgcg_cgcg_init)); - amdgpu_program_register_sequence(adev, - stoney_golden_settings_a11, - (const u32)ARRAY_SIZE(stoney_golden_settings_a11)); - amdgpu_program_register_sequence(adev, - stoney_golden_common_all, - (const u32)ARRAY_SIZE(stoney_golden_common_all)); + amdgpu_device_program_register_sequence(adev, + stoney_mgcg_cgcg_init, + ARRAY_SIZE(stoney_mgcg_cgcg_init)); + amdgpu_device_program_register_sequence(adev, + stoney_golden_settings_a11, + ARRAY_SIZE(stoney_golden_settings_a11)); + amdgpu_device_program_register_sequence(adev, + stoney_golden_common_all, + ARRAY_SIZE(stoney_golden_common_all)); break; default: break; } } -static void gfx_v8_0_scratch_init(struct amdgpu_device *adev) -{ - adev->gfx.scratch.num_reg = 7; - adev->gfx.scratch.reg_base = mmSCRATCH_REG0; - adev->gfx.scratch.free_mask = (1u << adev->gfx.scratch.num_reg) - 1; -} - static int gfx_v8_0_ring_test_ring(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - uint32_t scratch; uint32_t tmp = 0; unsigned i; int r; - r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) { - DRM_ERROR("amdgpu: cp failed to get scratch reg (%d).\n", r); - return r; - } - WREG32(scratch, 0xCAFEDEAD); + WREG32(mmSCRATCH_REG0, 0xCAFEDEAD); r = amdgpu_ring_alloc(ring, 3); - if (r) { - DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n", - ring->idx, r); - amdgpu_gfx_scratch_free(adev, scratch); + if (r) return r; - } + amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); - amdgpu_ring_write(ring, (scratch - PACKET3_SET_UCONFIG_REG_START)); + amdgpu_ring_write(ring, mmSCRATCH_REG0 - PACKET3_SET_UCONFIG_REG_START); amdgpu_ring_write(ring, 0xDEADBEEF); amdgpu_ring_commit(ring); for (i = 0; i < adev->usec_timeout; i++) { - tmp = RREG32(scratch); + tmp = RREG32(mmSCRATCH_REG0); if (tmp == 0xDEADBEEF) break; - DRM_UDELAY(1); - } - if (i < adev->usec_timeout) { - DRM_INFO("ring test on %d succeeded in %d usecs\n", - ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed (scratch(0x%04X)=0x%08X)\n", - ring->idx, scratch, tmp); - r = -EINVAL; + udelay(1); } - amdgpu_gfx_scratch_free(adev, scratch); + + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; + return r; } @@ -804,26 +870,30 @@ static int gfx_v8_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) struct amdgpu_device *adev = ring->adev; struct amdgpu_ib ib; struct dma_fence *f = NULL; - uint32_t scratch; - uint32_t tmp = 0; + + unsigned int index; + uint64_t gpu_addr; + uint32_t tmp; long r; - r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) { - DRM_ERROR("amdgpu: failed to get scratch reg (%ld).\n", r); + r = amdgpu_device_wb_get(adev, &index); + if (r) return r; - } - WREG32(scratch, 0xCAFEDEAD); + + 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, 256, &ib); - if (r) { - DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); + + r = amdgpu_ib_get(adev, NULL, 20, AMDGPU_IB_POOL_DIRECT, &ib); + if (r) goto err1; - } - ib.ptr[0] = PACKET3(PACKET3_SET_UCONFIG_REG, 1); - ib.ptr[1] = ((scratch - PACKET3_SET_UCONFIG_REG_START)); - ib.ptr[2] = 0xDEADBEEF; - ib.length_dw = 3; + + ib.ptr[0] = PACKET3(PACKET3_WRITE_DATA, 3); + ib.ptr[1] = WRITE_DATA_DST_SEL(5) | WR_CONFIRM; + ib.ptr[2] = lower_32_bits(gpu_addr); + ib.ptr[3] = upper_32_bits(gpu_addr); + ib.ptr[4] = 0xDEADBEEF; + ib.length_dw = 5; r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); if (r) @@ -831,47 +901,37 @@ static int gfx_v8_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) r = dma_fence_wait_timeout(f, false, timeout); if (r == 0) { - DRM_ERROR("amdgpu: IB test timed out.\n"); r = -ETIMEDOUT; goto err2; } else if (r < 0) { - DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); goto err2; } - tmp = RREG32(scratch); - if (tmp == 0xDEADBEEF) { - DRM_INFO("ib test on ring %d succeeded\n", ring->idx); + + tmp = adev->wb.wb[index]; + if (tmp == 0xDEADBEEF) r = 0; - } else { - DRM_ERROR("amdgpu: ib test failed (scratch(0x%04X)=0x%08X)\n", - scratch, tmp); + else r = -EINVAL; - } + err2: - amdgpu_ib_free(adev, &ib, NULL); + amdgpu_ib_free(&ib, NULL); dma_fence_put(f); err1: - amdgpu_gfx_scratch_free(adev, scratch); + amdgpu_device_wb_free(adev, index); return r; } static void gfx_v8_0_free_microcode(struct amdgpu_device *adev) { - release_firmware(adev->gfx.pfp_fw); - adev->gfx.pfp_fw = NULL; - release_firmware(adev->gfx.me_fw); - adev->gfx.me_fw = NULL; - release_firmware(adev->gfx.ce_fw); - adev->gfx.ce_fw = NULL; - release_firmware(adev->gfx.rlc_fw); - adev->gfx.rlc_fw = NULL; - release_firmware(adev->gfx.mec_fw); - adev->gfx.mec_fw = NULL; + amdgpu_ucode_release(&adev->gfx.pfp_fw); + amdgpu_ucode_release(&adev->gfx.me_fw); + amdgpu_ucode_release(&adev->gfx.ce_fw); + amdgpu_ucode_release(&adev->gfx.rlc_fw); + amdgpu_ucode_release(&adev->gfx.mec_fw); if ((adev->asic_type != CHIP_STONEY) && (adev->asic_type != CHIP_TOPAZ)) - release_firmware(adev->gfx.mec2_fw); - adev->gfx.mec2_fw = NULL; + amdgpu_ucode_release(&adev->gfx.mec2_fw); kfree(adev->gfx.rlc.register_list_format); } @@ -879,7 +939,6 @@ static void gfx_v8_0_free_microcode(struct amdgpu_device *adev) static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) { const char *chip_name; - char fw_name[30]; int err; struct amdgpu_firmware_info *info = NULL; const struct common_firmware_header *header = NULL; @@ -902,38 +961,59 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) case CHIP_FIJI: chip_name = "fiji"; break; - case CHIP_POLARIS11: - chip_name = "polaris11"; + case CHIP_STONEY: + chip_name = "stoney"; break; case CHIP_POLARIS10: chip_name = "polaris10"; break; + case CHIP_POLARIS11: + chip_name = "polaris11"; + break; case CHIP_POLARIS12: chip_name = "polaris12"; break; - case CHIP_STONEY: - chip_name = "stoney"; + case CHIP_VEGAM: + chip_name = "vegam"; break; default: BUG(); } - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name); - err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev); - if (err) - goto out; - err = amdgpu_ucode_validate(adev->gfx.pfp_fw); + if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) { + err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, + AMDGPU_UCODE_OPTIONAL, + "amdgpu/%s_pfp_2.bin", chip_name); + if (err == -ENODEV) { + err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_pfp.bin", chip_name); + } + } else { + err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_pfp.bin", chip_name); + } if (err) goto out; cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.pfp_fw->data; adev->gfx.pfp_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); adev->gfx.pfp_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name); - err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev); - if (err) - goto out; - err = amdgpu_ucode_validate(adev->gfx.me_fw); + if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) { + err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, + AMDGPU_UCODE_OPTIONAL, + "amdgpu/%s_me_2.bin", chip_name); + if (err == -ENODEV) { + err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_me.bin", chip_name); + } + } else { + err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_me.bin", chip_name); + } if (err) goto out; cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.me_fw->data; @@ -941,11 +1021,20 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) adev->gfx.me_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name); - err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev); - if (err) - goto out; - err = amdgpu_ucode_validate(adev->gfx.ce_fw); + if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) { + err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, + AMDGPU_UCODE_OPTIONAL, + "amdgpu/%s_ce_2.bin", chip_name); + if (err == -ENODEV) { + err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_ce.bin", chip_name); + } + } else { + err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_ce.bin", chip_name); + } if (err) goto out; cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data; @@ -963,11 +1052,11 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) } else adev->virt.chained_ib_support = false; - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name); - err = request_firmware(&adev->gfx.rlc_fw, fw_name, adev->dev); + err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_rlc.bin", chip_name); if (err) goto out; - err = amdgpu_ucode_validate(adev->gfx.rlc_fw); rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data; adev->gfx.rlc_fw_version = le32_to_cpu(rlc_hdr->header.ucode_version); adev->gfx.rlc_feature_version = le32_to_cpu(rlc_hdr->ucode_feature_version); @@ -1002,21 +1091,30 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) tmp = (unsigned int *)((uintptr_t)rlc_hdr + le32_to_cpu(rlc_hdr->reg_list_format_array_offset_bytes)); - for (i = 0 ; i < (rlc_hdr->reg_list_format_size_bytes >> 2); i++) + for (i = 0 ; i < (adev->gfx.rlc.reg_list_format_size_bytes >> 2); i++) adev->gfx.rlc.register_list_format[i] = le32_to_cpu(tmp[i]); adev->gfx.rlc.register_restore = adev->gfx.rlc.register_list_format + i; tmp = (unsigned int *)((uintptr_t)rlc_hdr + le32_to_cpu(rlc_hdr->reg_list_array_offset_bytes)); - for (i = 0 ; i < (rlc_hdr->reg_list_size_bytes >> 2); i++) + for (i = 0 ; i < (adev->gfx.rlc.reg_list_size_bytes >> 2); i++) adev->gfx.rlc.register_restore[i] = le32_to_cpu(tmp[i]); - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name); - err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev); - if (err) - goto out; - err = amdgpu_ucode_validate(adev->gfx.mec_fw); + if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) { + err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, + AMDGPU_UCODE_OPTIONAL, + "amdgpu/%s_mec_2.bin", chip_name); + if (err == -ENODEV) { + err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_mec.bin", chip_name); + } + } else { + err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_mec.bin", chip_name); + } if (err) goto out; cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data; @@ -1025,12 +1123,21 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) if ((adev->asic_type != CHIP_STONEY) && (adev->asic_type != CHIP_TOPAZ)) { - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name); - err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev); + if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) { + err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, + AMDGPU_UCODE_OPTIONAL, + "amdgpu/%s_mec2_2.bin", chip_name); + if (err == -ENODEV) { + err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_mec2.bin", chip_name); + } + } else { + err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_mec2.bin", chip_name); + } 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 = @@ -1043,230 +1150,106 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) } } - if (adev->firmware.load_type == AMDGPU_FW_LOAD_SMU) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_PFP]; - info->ucode_id = AMDGPU_UCODE_ID_CP_PFP; - info->fw = adev->gfx.pfp_fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_ME]; - info->ucode_id = AMDGPU_UCODE_ID_CP_ME; - info->fw = adev->gfx.me_fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_CE]; - info->ucode_id = AMDGPU_UCODE_ID_CP_CE; - info->fw = adev->gfx.ce_fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_PFP]; + info->ucode_id = AMDGPU_UCODE_ID_CP_PFP; + info->fw = adev->gfx.pfp_fw; + header = (const struct common_firmware_header *)info->fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_ME]; + info->ucode_id = AMDGPU_UCODE_ID_CP_ME; + info->fw = adev->gfx.me_fw; + header = (const struct common_firmware_header *)info->fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_CE]; + info->ucode_id = AMDGPU_UCODE_ID_CP_CE; + info->fw = adev->gfx.ce_fw; + header = (const struct common_firmware_header *)info->fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_G]; + info->ucode_id = AMDGPU_UCODE_ID_RLC_G; + info->fw = adev->gfx.rlc_fw; + header = (const struct common_firmware_header *)info->fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC1]; + info->ucode_id = AMDGPU_UCODE_ID_CP_MEC1; + info->fw = adev->gfx.mec_fw; + header = (const struct common_firmware_header *)info->fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + + /* we need account JT in */ + cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(cp_hdr->jt_size) << 2, PAGE_SIZE); - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_G]; - info->ucode_id = AMDGPU_UCODE_ID_RLC_G; - info->fw = adev->gfx.rlc_fw; - header = (const struct common_firmware_header *)info->fw->data; + if (amdgpu_sriov_vf(adev)) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_STORAGE]; + info->ucode_id = AMDGPU_UCODE_ID_STORAGE; + info->fw = adev->gfx.mec_fw; adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + ALIGN(le32_to_cpu(64 * PAGE_SIZE), PAGE_SIZE); + } - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC1]; - info->ucode_id = AMDGPU_UCODE_ID_CP_MEC1; - info->fw = adev->gfx.mec_fw; + if (adev->gfx.mec2_fw) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC2]; + info->ucode_id = AMDGPU_UCODE_ID_CP_MEC2; + info->fw = adev->gfx.mec2_fw; header = (const struct common_firmware_header *)info->fw->data; adev->firmware.fw_size += ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - - /* we need account JT in */ - cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(cp_hdr->jt_size) << 2, PAGE_SIZE); - - if (amdgpu_sriov_vf(adev)) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_STORAGE]; - info->ucode_id = AMDGPU_UCODE_ID_STORAGE; - info->fw = adev->gfx.mec_fw; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(64 * PAGE_SIZE), PAGE_SIZE); - } - - if (adev->gfx.mec2_fw) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC2]; - info->ucode_id = AMDGPU_UCODE_ID_CP_MEC2; - info->fw = adev->gfx.mec2_fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - } - } out: if (err) { - dev_err(adev->dev, - "gfx8: Failed to load firmware \"%s\"\n", - fw_name); - release_firmware(adev->gfx.pfp_fw); - adev->gfx.pfp_fw = NULL; - release_firmware(adev->gfx.me_fw); - adev->gfx.me_fw = NULL; - release_firmware(adev->gfx.ce_fw); - adev->gfx.ce_fw = NULL; - release_firmware(adev->gfx.rlc_fw); - adev->gfx.rlc_fw = NULL; - release_firmware(adev->gfx.mec_fw); - adev->gfx.mec_fw = NULL; - release_firmware(adev->gfx.mec2_fw); - adev->gfx.mec2_fw = NULL; + dev_err(adev->dev, "gfx8: Failed to load firmware %s gfx firmware\n", chip_name); + amdgpu_ucode_release(&adev->gfx.pfp_fw); + amdgpu_ucode_release(&adev->gfx.me_fw); + amdgpu_ucode_release(&adev->gfx.ce_fw); + amdgpu_ucode_release(&adev->gfx.rlc_fw); + amdgpu_ucode_release(&adev->gfx.mec_fw); + amdgpu_ucode_release(&adev->gfx.mec2_fw); } return err; } -static void gfx_v8_0_get_csb_buffer(struct amdgpu_device *adev, - volatile u32 *buffer) +static void gfx_v8_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer) { - u32 count = 0, i; - const struct cs_section_def *sect = NULL; - const struct cs_extent_def *ext = NULL; + u32 count = 0; if (adev->gfx.rlc.cs_data == NULL) return; if (buffer == NULL) return; - buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0)); - buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE); - - buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1)); - buffer[count++] = cpu_to_le32(0x80000000); - buffer[count++] = cpu_to_le32(0x80000000); - - for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) { - for (ext = sect->section; ext->extent != NULL; ++ext) { - if (sect->id == SECT_CONTEXT) { - buffer[count++] = - cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count)); - buffer[count++] = cpu_to_le32(ext->reg_index - - PACKET3_SET_CONTEXT_REG_START); - for (i = 0; i < ext->reg_count; i++) - buffer[count++] = cpu_to_le32(ext->extent[i]); - } else { - return; - } - } - } + count = amdgpu_gfx_csb_preamble_start(buffer); + count = amdgpu_gfx_csb_data_parser(adev, buffer, count); buffer[count++] = cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, 2)); - buffer[count++] = cpu_to_le32(mmPA_SC_RASTER_CONFIG - - PACKET3_SET_CONTEXT_REG_START); + buffer[count++] = cpu_to_le32(mmPA_SC_RASTER_CONFIG - PACKET3_SET_CONTEXT_REG_START); buffer[count++] = cpu_to_le32(adev->gfx.config.rb_config[0][0].raster_config); buffer[count++] = cpu_to_le32(adev->gfx.config.rb_config[0][0].raster_config_1); - buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0)); - buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE); - - buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0)); - buffer[count++] = cpu_to_le32(0); + amdgpu_gfx_csb_preamble_end(buffer, count); } -static void cz_init_cp_jump_table(struct amdgpu_device *adev) +static int gfx_v8_0_cp_jump_table_num(struct amdgpu_device *adev) { - const __le32 *fw_data; - volatile u32 *dst_ptr; - int me, i, max_me = 4; - u32 bo_offset = 0; - u32 table_offset, table_size; - if (adev->asic_type == CHIP_CARRIZO) - max_me = 5; - - /* write the cp table buffer */ - dst_ptr = adev->gfx.rlc.cp_table_ptr; - for (me = 0; me < max_me; me++) { - if (me == 0) { - const struct gfx_firmware_header_v1_0 *hdr = - (const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data; - fw_data = (const __le32 *) - (adev->gfx.ce_fw->data + - le32_to_cpu(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } else if (me == 1) { - const struct gfx_firmware_header_v1_0 *hdr = - (const struct gfx_firmware_header_v1_0 *)adev->gfx.pfp_fw->data; - fw_data = (const __le32 *) - (adev->gfx.pfp_fw->data + - le32_to_cpu(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } else if (me == 2) { - const struct gfx_firmware_header_v1_0 *hdr = - (const struct gfx_firmware_header_v1_0 *)adev->gfx.me_fw->data; - fw_data = (const __le32 *) - (adev->gfx.me_fw->data + - le32_to_cpu(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } else if (me == 3) { - const struct gfx_firmware_header_v1_0 *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(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } else if (me == 4) { - const struct gfx_firmware_header_v1_0 *hdr = - (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec2_fw->data; - fw_data = (const __le32 *) - (adev->gfx.mec2_fw->data + - le32_to_cpu(hdr->header.ucode_array_offset_bytes)); - table_offset = le32_to_cpu(hdr->jt_offset); - table_size = le32_to_cpu(hdr->jt_size); - } - - for (i = 0; i < table_size; i ++) { - dst_ptr[bo_offset + i] = - cpu_to_le32(le32_to_cpu(fw_data[table_offset + i])); - } - - bo_offset += table_size; - } -} - -static void gfx_v8_0_rlc_fini(struct amdgpu_device *adev) -{ - int r; - - /* clear state block */ - if (adev->gfx.rlc.clear_state_obj) { - r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, true); - if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve RLC cbs bo failed\n", r); - amdgpu_bo_unpin(adev->gfx.rlc.clear_state_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); - amdgpu_bo_unref(&adev->gfx.rlc.clear_state_obj); - adev->gfx.rlc.clear_state_obj = NULL; - } - - /* jump table block */ - if (adev->gfx.rlc.cp_table_obj) { - r = amdgpu_bo_reserve(adev->gfx.rlc.cp_table_obj, true); - if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve RLC cp table bo failed\n", r); - amdgpu_bo_unpin(adev->gfx.rlc.cp_table_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj); - amdgpu_bo_unref(&adev->gfx.rlc.cp_table_obj); - adev->gfx.rlc.cp_table_obj = NULL; - } + return 5; + else + return 4; } static int gfx_v8_0_rlc_init(struct amdgpu_device *adev) { - volatile u32 *dst_ptr; - u32 dws; const struct cs_section_def *cs_data; int r; @@ -1275,105 +1258,30 @@ static int gfx_v8_0_rlc_init(struct amdgpu_device *adev) cs_data = adev->gfx.rlc.cs_data; if (cs_data) { - /* clear state block */ - adev->gfx.rlc.clear_state_size = dws = gfx_v8_0_get_csb_size(adev); - - if (adev->gfx.rlc.clear_state_obj == NULL) { - r = amdgpu_bo_create(adev, dws * 4, PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, - &adev->gfx.rlc.clear_state_obj); - if (r) { - dev_warn(adev->dev, "(%d) create RLC c bo failed\n", r); - gfx_v8_0_rlc_fini(adev); - return r; - } - } - r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, false); - if (unlikely(r != 0)) { - gfx_v8_0_rlc_fini(adev); - return r; - } - r = amdgpu_bo_pin(adev->gfx.rlc.clear_state_obj, AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.clear_state_gpu_addr); - if (r) { - amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); - dev_warn(adev->dev, "(%d) pin RLC cbs bo failed\n", r); - gfx_v8_0_rlc_fini(adev); - return r; - } - - r = amdgpu_bo_kmap(adev->gfx.rlc.clear_state_obj, (void **)&adev->gfx.rlc.cs_ptr); - if (r) { - dev_warn(adev->dev, "(%d) map RLC cbs bo failed\n", r); - gfx_v8_0_rlc_fini(adev); + /* init clear state block */ + r = amdgpu_gfx_rlc_init_csb(adev); + if (r) return r; - } - /* set up the cs buffer */ - dst_ptr = adev->gfx.rlc.cs_ptr; - gfx_v8_0_get_csb_buffer(adev, dst_ptr); - amdgpu_bo_kunmap(adev->gfx.rlc.clear_state_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); } if ((adev->asic_type == CHIP_CARRIZO) || (adev->asic_type == CHIP_STONEY)) { adev->gfx.rlc.cp_table_size = ALIGN(96 * 5 * 4, 2048) + (64 * 1024); /* JT + GDS */ - if (adev->gfx.rlc.cp_table_obj == NULL) { - r = amdgpu_bo_create(adev, adev->gfx.rlc.cp_table_size, PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, - &adev->gfx.rlc.cp_table_obj); - if (r) { - dev_warn(adev->dev, "(%d) create RLC cp table bo failed\n", r); - return r; - } - } - - r = amdgpu_bo_reserve(adev->gfx.rlc.cp_table_obj, false); - if (unlikely(r != 0)) { - dev_warn(adev->dev, "(%d) reserve RLC cp table bo failed\n", r); - return r; - } - r = amdgpu_bo_pin(adev->gfx.rlc.cp_table_obj, AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.cp_table_gpu_addr); - if (r) { - amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj); - dev_warn(adev->dev, "(%d) pin RLC cp table bo failed\n", r); - return r; - } - r = amdgpu_bo_kmap(adev->gfx.rlc.cp_table_obj, (void **)&adev->gfx.rlc.cp_table_ptr); - if (r) { - dev_warn(adev->dev, "(%d) map RLC cp table bo failed\n", r); + r = amdgpu_gfx_rlc_init_cpt(adev); + if (r) return r; - } - - cz_init_cp_jump_table(adev); - - amdgpu_bo_kunmap(adev->gfx.rlc.cp_table_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj); } + /* init spm vmid with 0xf */ + if (adev->gfx.rlc.funcs->update_spm_vmid) + adev->gfx.rlc.funcs->update_spm_vmid(adev, NULL, 0xf); + return 0; } static void gfx_v8_0_mec_fini(struct amdgpu_device *adev) { - int r; - - if (adev->gfx.mec.hpd_eop_obj) { - r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, true); - if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve HPD EOP bo failed\n", r); - amdgpu_bo_unpin(adev->gfx.mec.hpd_eop_obj); - amdgpu_bo_unreserve(adev->gfx.mec.hpd_eop_obj); - amdgpu_bo_unref(&adev->gfx.mec.hpd_eop_obj); - adev->gfx.mec.hpd_eop_obj = NULL; - } + amdgpu_bo_free_kernel(&adev->gfx.mec.hpd_eop_obj, NULL, NULL); } static int gfx_v8_0_mec_init(struct amdgpu_device *adev) @@ -1382,48 +1290,29 @@ static int gfx_v8_0_mec_init(struct amdgpu_device *adev) u32 *hpd; size_t mec_hpd_size; - bitmap_zero(adev->gfx.mec.queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES); + bitmap_zero(adev->gfx.mec_bitmap[0].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES); /* take ownership of the relevant compute queues */ amdgpu_gfx_compute_queue_acquire(adev); mec_hpd_size = adev->gfx.num_compute_rings * GFX8_MEC_HPD_SIZE; - - if (adev->gfx.mec.hpd_eop_obj == NULL) { - r = amdgpu_bo_create(adev, - mec_hpd_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL, - &adev->gfx.mec.hpd_eop_obj); + if (mec_hpd_size) { + r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM | + AMDGPU_GEM_DOMAIN_GTT, + &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); return r; } - } - - r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, false); - if (unlikely(r != 0)) { - gfx_v8_0_mec_fini(adev); - return r; - } - r = amdgpu_bo_pin(adev->gfx.mec.hpd_eop_obj, AMDGPU_GEM_DOMAIN_GTT, - &adev->gfx.mec.hpd_eop_gpu_addr); - if (r) { - dev_warn(adev->dev, "(%d) pin HDP EOP bo failed\n", r); - gfx_v8_0_mec_fini(adev); - return r; - } - r = amdgpu_bo_kmap(adev->gfx.mec.hpd_eop_obj, (void **)&hpd); - if (r) { - dev_warn(adev->dev, "(%d) map HDP EOP bo failed\n", r); - gfx_v8_0_mec_fini(adev); - return r; - } - memset(hpd, 0, mec_hpd_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); + } return 0; } @@ -1493,10 +1382,11 @@ static const u32 sgpr_init_compute_shader[] = static const u32 vgpr_init_regs[] = { mmCOMPUTE_STATIC_THREAD_MGMT_SE0, 0xffffffff, - mmCOMPUTE_RESOURCE_LIMITS, 0, + mmCOMPUTE_RESOURCE_LIMITS, 0x1000000, /* CU_GROUP_COUNT=1 */ mmCOMPUTE_NUM_THREAD_X, 256*4, mmCOMPUTE_NUM_THREAD_Y, 1, mmCOMPUTE_NUM_THREAD_Z, 1, + mmCOMPUTE_PGM_RSRC1, 0x100004f, /* VGPRS=15 (64 logical VGPRs), SGPRS=1 (16 SGPRs), BULKY=1 */ mmCOMPUTE_PGM_RSRC2, 20, mmCOMPUTE_USER_DATA_0, 0xedcedc00, mmCOMPUTE_USER_DATA_1, 0xedcedc01, @@ -1513,10 +1403,11 @@ static const u32 vgpr_init_regs[] = static const u32 sgpr1_init_regs[] = { mmCOMPUTE_STATIC_THREAD_MGMT_SE0, 0x0f, - mmCOMPUTE_RESOURCE_LIMITS, 0x1000000, + mmCOMPUTE_RESOURCE_LIMITS, 0x1000000, /* CU_GROUP_COUNT=1 */ mmCOMPUTE_NUM_THREAD_X, 256*5, mmCOMPUTE_NUM_THREAD_Y, 1, mmCOMPUTE_NUM_THREAD_Z, 1, + mmCOMPUTE_PGM_RSRC1, 0x240, /* SGPRS=9 (80 GPRS) */ mmCOMPUTE_PGM_RSRC2, 20, mmCOMPUTE_USER_DATA_0, 0xedcedc00, mmCOMPUTE_USER_DATA_1, 0xedcedc01, @@ -1537,6 +1428,7 @@ static const u32 sgpr2_init_regs[] = mmCOMPUTE_NUM_THREAD_X, 256*5, mmCOMPUTE_NUM_THREAD_Y, 1, mmCOMPUTE_NUM_THREAD_Z, 1, + mmCOMPUTE_PGM_RSRC1, 0x240, /* SGPRS=9 (80 GPRS) */ mmCOMPUTE_PGM_RSRC2, 20, mmCOMPUTE_USER_DATA_0, 0xedcedc00, mmCOMPUTE_USER_DATA_1, 0xedcedc01, @@ -1594,7 +1486,7 @@ static int gfx_v8_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) return 0; /* bail if the compute ring is not ready */ - if (!ring->ready) + if (!ring->sched.ready) return 0; tmp = RREG32(mmGB_EDC_MODE); @@ -1614,7 +1506,8 @@ static int gfx_v8_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) /* 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; @@ -1736,7 +1629,7 @@ static int gfx_v8_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) RREG32(sec_ded_counter_registers[i]); fail: - amdgpu_ib_free(adev, &ib, NULL); + amdgpu_ib_free(&ib, NULL); dma_fence_put(f); return r; @@ -1745,7 +1638,7 @@ fail: static int gfx_v8_0_gpu_early_init(struct amdgpu_device *adev) { u32 gb_addr_config; - u32 mc_shared_chmap, mc_arb_ramcfg; + u32 mc_arb_ramcfg; u32 dimm00_addr_map, dimm01_addr_map, dimm10_addr_map, dimm11_addr_map; u32 tmp; int ret; @@ -1801,6 +1694,7 @@ static int gfx_v8_0_gpu_early_init(struct amdgpu_device *adev) gb_addr_config = POLARIS11_GB_ADDR_CONFIG_GOLDEN; break; case CHIP_POLARIS10: + case CHIP_VEGAM: ret = amdgpu_atombios_get_gfx_info(adev); if (ret) return ret; @@ -1884,10 +1778,14 @@ static int gfx_v8_0_gpu_early_init(struct amdgpu_device *adev) break; } - mc_shared_chmap = RREG32(mmMC_SHARED_CHMAP); adev->gfx.config.mc_arb_ramcfg = RREG32(mmMC_ARB_RAMCFG); mc_arb_ramcfg = adev->gfx.config.mc_arb_ramcfg; + adev->gfx.config.num_banks = REG_GET_FIELD(mc_arb_ramcfg, + MC_ARB_RAMCFG, NOOFBANK); + adev->gfx.config.num_ranks = REG_GET_FIELD(mc_arb_ramcfg, + MC_ARB_RAMCFG, NOOFRANKS); + adev->gfx.config.num_tile_pipes = adev->gfx.config.max_tile_pipes; adev->gfx.config.mem_max_burst_length_bytes = 256; if (adev->flags & AMD_IS_APU) { @@ -1951,6 +1849,7 @@ static int gfx_v8_0_compute_ring_init(struct amdgpu_device *adev, int ring_id, 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]; @@ -1961,7 +1860,7 @@ static int gfx_v8_0_compute_ring_init(struct amdgpu_device *adev, int ring_id, ring->ring_obj = NULL; ring->use_doorbell = true; - ring->doorbell_index = AMDGPU_DOORBELL_MEC_RING0 + ring_id; + ring->doorbell_index = adev->doorbell_index.mec_ring0 + ring_id; ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr + (ring_id * GFX8_MEC_HPD_SIZE); sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue); @@ -1970,9 +1869,11 @@ static int gfx_v8_0_compute_ring_init(struct amdgpu_device *adev, int ring_id, + ((ring->me - 1) * adev->gfx.mec.num_pipe_per_mec) + ring->pipe; + hw_prio = amdgpu_gfx_is_high_priority_compute_queue(adev, ring) ? + AMDGPU_RING_PRIO_2 : AMDGPU_RING_PRIO_DEFAULT; /* type-2 packets are deprecated on MEC, use type-3 instead */ - r = amdgpu_ring_init(adev, ring, 1024, - &adev->gfx.eop_irq, irq_type); + r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type, + hw_prio, NULL); if (r) return r; @@ -1980,20 +1881,23 @@ static int gfx_v8_0_compute_ring_init(struct amdgpu_device *adev, int ring_id, return 0; } -static int gfx_v8_0_sw_init(void *handle) +static void gfx_v8_0_sq_irq_work_func(struct work_struct *work); + +static int gfx_v8_0_sw_init(struct amdgpu_ip_block *ip_block) { int i, j, k, r, ring_id; + int xcc_id = 0; struct amdgpu_ring *ring; - struct amdgpu_kiq *kiq; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; switch (adev->asic_type) { - case CHIP_FIJI: case CHIP_TONGA: + case CHIP_CARRIZO: + case CHIP_FIJI: + case CHIP_POLARIS10: case CHIP_POLARIS11: case CHIP_POLARIS12: - case CHIP_POLARIS10: - case CHIP_CARRIZO: + case CHIP_VEGAM: adev->gfx.mec.num_mec = 2; break; case CHIP_TOPAZ: @@ -2006,31 +1910,40 @@ static int gfx_v8_0_sw_init(void *handle) adev->gfx.mec.num_pipe_per_mec = 4; adev->gfx.mec.num_queue_per_pipe = 8; - /* KIQ event */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 178, &adev->gfx.kiq.irq); - if (r) - return r; - /* EOP Event */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 181, &adev->gfx.eop_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_CP_END_OF_PIPE, &adev->gfx.eop_irq); if (r) return r; /* Privileged reg */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 184, + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_CP_PRIV_REG_FAULT, &adev->gfx.priv_reg_irq); if (r) return r; /* Privileged inst */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 185, + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_CP_PRIV_INSTR_FAULT, &adev->gfx.priv_inst_irq); if (r) return r; - adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE; + /* Add CP EDC/ECC irq */ + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_CP_ECC_ERROR, + &adev->gfx.cp_ecc_error_irq); + if (r) + return r; - gfx_v8_0_scratch_init(adev); + /* SQ interrupts. */ + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SQ_INTERRUPT_MSG, + &adev->gfx.sq_irq); + if (r) { + DRM_ERROR("amdgpu_irq_add() for SQ failed: %d\n", r); + return r; + } + + INIT_WORK(&adev->gfx.sq_work.work, gfx_v8_0_sq_irq_work_func); + + adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE; r = gfx_v8_0_init_microcode(adev); if (r) { @@ -2038,7 +1951,7 @@ static int gfx_v8_0_sw_init(void *handle) return r; } - r = gfx_v8_0_rlc_init(adev); + r = adev->gfx.rlc.funcs->init(adev); if (r) { DRM_ERROR("Failed to init rlc BOs!\n"); return r; @@ -2058,11 +1971,12 @@ static int gfx_v8_0_sw_init(void *handle) /* no gfx doorbells on iceland */ if (adev->asic_type != CHIP_TOPAZ) { ring->use_doorbell = true; - ring->doorbell_index = AMDGPU_DOORBELL_GFX_RING0; + ring->doorbell_index = adev->doorbell_index.gfx_ring0; } r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, - AMDGPU_CP_IRQ_GFX_EOP); + AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP, + AMDGPU_RING_PRIO_DEFAULT, NULL); if (r) return r; } @@ -2073,7 +1987,8 @@ static int gfx_v8_0_sw_init(void *handle) for (i = 0; i < adev->gfx.mec.num_mec; ++i) { for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) { for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) { - if (!amdgpu_gfx_is_mec_queue_enabled(adev, i, k, j)) + if (!amdgpu_gfx_is_mec_queue_enabled(adev, 0, i, + k, j)) continue; r = gfx_v8_0_compute_ring_init(adev, @@ -2087,38 +2002,18 @@ static int gfx_v8_0_sw_init(void *handle) } } - r = amdgpu_gfx_kiq_init(adev, GFX8_MEC_HPD_SIZE); + r = amdgpu_gfx_kiq_init(adev, GFX8_MEC_HPD_SIZE, 0); if (r) { DRM_ERROR("Failed to init KIQ BOs!\n"); return r; } - kiq = &adev->gfx.kiq; - r = amdgpu_gfx_kiq_init_ring(adev, &kiq->ring, &kiq->irq); + r = amdgpu_gfx_kiq_init_ring(adev, xcc_id); if (r) return r; /* create MQD for all compute queues as well as KIQ for SRIOV case */ - r = amdgpu_gfx_compute_mqd_sw_init(adev, sizeof(struct vi_mqd_allocation)); - if (r) - return r; - - /* reserve GDS, GWS and OA resource for gfx */ - r = amdgpu_bo_create_kernel(adev, adev->gds.mem.gfx_partition_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_GDS, - &adev->gds.gds_gfx_bo, NULL, NULL); - if (r) - return r; - - r = amdgpu_bo_create_kernel(adev, adev->gds.gws.gfx_partition_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_GWS, - &adev->gds.gws_gfx_bo, NULL, NULL); - if (r) - return r; - - r = amdgpu_bo_create_kernel(adev, adev->gds.oa.gfx_partition_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_OA, - &adev->gds.oa_gfx_bo, NULL, NULL); + r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct vi_mqd_allocation), 0); if (r) return r; @@ -2128,29 +2023,39 @@ static int gfx_v8_0_sw_init(void *handle) if (r) return r; + adev->gfx.gfx_supported_reset = + amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]); + adev->gfx.compute_supported_reset = + amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]); + return 0; } -static int gfx_v8_0_sw_fini(void *handle) +static int gfx_v8_0_sw_fini(struct amdgpu_ip_block *ip_block) { + struct amdgpu_device *adev = ip_block->adev; int i; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - amdgpu_bo_free_kernel(&adev->gds.oa_gfx_bo, NULL, NULL); - amdgpu_bo_free_kernel(&adev->gds.gws_gfx_bo, NULL, NULL); - amdgpu_bo_free_kernel(&adev->gds.gds_gfx_bo, NULL, NULL); for (i = 0; i < adev->gfx.num_gfx_rings; i++) amdgpu_ring_fini(&adev->gfx.gfx_ring[i]); for (i = 0; i < adev->gfx.num_compute_rings; i++) amdgpu_ring_fini(&adev->gfx.compute_ring[i]); - amdgpu_gfx_compute_mqd_sw_fini(adev); - amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring, &adev->gfx.kiq.irq); - amdgpu_gfx_kiq_fini(adev); + amdgpu_gfx_mqd_sw_fini(adev, 0); + amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq[0].ring); + amdgpu_gfx_kiq_fini(adev, 0); gfx_v8_0_mec_fini(adev); - gfx_v8_0_rlc_fini(adev); + amdgpu_gfx_rlc_fini(adev); + amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj, + &adev->gfx.rlc.clear_state_gpu_addr, + (void **)&adev->gfx.rlc.cs_ptr); + if ((adev->asic_type == CHIP_CARRIZO) || + (adev->asic_type == CHIP_STONEY)) { + amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj, + &adev->gfx.rlc.cp_table_gpu_addr, + (void **)&adev->gfx.rlc.cp_table_ptr); + } gfx_v8_0_free_microcode(adev); return 0; @@ -2345,6 +2250,7 @@ static void gfx_v8_0_tiling_mode_table_init(struct amdgpu_device *adev) break; case CHIP_FIJI: + case CHIP_VEGAM: modearray[0] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) | TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) | @@ -3302,6 +3208,7 @@ static void gfx_v8_0_tiling_mode_table_init(struct amdgpu_device *adev) dev_warn(adev->dev, "Unknown chip type (%d) in function gfx_v8_0_tiling_mode_table_init() falling through to CHIP_CARRIZO\n", adev->asic_type); + fallthrough; case CHIP_CARRIZO: modearray[0] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | @@ -3478,7 +3385,8 @@ static void gfx_v8_0_tiling_mode_table_init(struct amdgpu_device *adev) } static void gfx_v8_0_select_se_sh(struct amdgpu_device *adev, - u32 se_num, u32 sh_num, u32 instance) + u32 se_num, u32 sh_num, u32 instance, + int xcc_id) { u32 data; @@ -3500,6 +3408,12 @@ static void gfx_v8_0_select_se_sh(struct amdgpu_device *adev, WREG32(mmGRBM_GFX_INDEX, data); } +static void gfx_v8_0_select_me_pipe_q(struct amdgpu_device *adev, + u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id) +{ + vi_srbm_select(adev, me, pipe, q, vm); +} + static u32 gfx_v8_0_get_rb_active_bitmap(struct amdgpu_device *adev) { u32 data, mask; @@ -3520,6 +3434,7 @@ gfx_v8_0_raster_config(struct amdgpu_device *adev, u32 *rconf, u32 *rconf1) { switch (adev->asic_type) { case CHIP_FIJI: + case CHIP_VEGAM: *rconf |= RB_MAP_PKR0(2) | RB_MAP_PKR1(2) | RB_XSEL2(1) | PKR_MAP(2) | PKR_XSEL(1) | PKR_YSEL(1) | @@ -3655,13 +3570,13 @@ gfx_v8_0_write_harvested_raster_configs(struct amdgpu_device *adev, } /* GRBM_GFX_INDEX has a different offset on VI */ - gfx_v8_0_select_se_sh(adev, se, 0xffffffff, 0xffffffff); + gfx_v8_0_select_se_sh(adev, se, 0xffffffff, 0xffffffff, 0); WREG32(mmPA_SC_RASTER_CONFIG, raster_config_se); WREG32(mmPA_SC_RASTER_CONFIG_1, raster_config_1); } /* GRBM_GFX_INDEX has a different offset on VI */ - gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0); } static void gfx_v8_0_setup_rb(struct amdgpu_device *adev) @@ -3677,13 +3592,13 @@ static void gfx_v8_0_setup_rb(struct amdgpu_device *adev) mutex_lock(&adev->grbm_idx_mutex); for (i = 0; i < adev->gfx.config.max_shader_engines; i++) { for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) { - gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff); + gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff, 0); data = gfx_v8_0_get_rb_active_bitmap(adev); active_rbs |= data << ((i * adev->gfx.config.max_sh_per_se + j) * rb_bitmap_width_per_sh); } } - gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0); adev->gfx.config.backend_enable_mask = active_rbs; adev->gfx.config.num_rbs = hweight32(active_rbs); @@ -3706,7 +3621,7 @@ static void gfx_v8_0_setup_rb(struct amdgpu_device *adev) /* cache the values for userspace */ for (i = 0; i < adev->gfx.config.max_shader_engines; i++) { for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) { - gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff); + gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff, 0); adev->gfx.config.rb_config[i][j].rb_backend_disable = RREG32(mmCC_RB_BACKEND_DISABLE); adev->gfx.config.rb_config[i][j].user_rb_backend_disable = @@ -3717,10 +3632,11 @@ static void gfx_v8_0_setup_rb(struct amdgpu_device *adev) RREG32(mmPA_SC_RASTER_CONFIG_1); } } - gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0); mutex_unlock(&adev->grbm_idx_mutex); } +#define DEFAULT_SH_MEM_BASES (0x6000) /** * gfx_v8_0_init_compute_vmid - gart enable * @@ -3729,9 +3645,6 @@ static void gfx_v8_0_setup_rb(struct amdgpu_device *adev) * Initialize compute vmid sh_mem registers * */ -#define DEFAULT_SH_MEM_BASES (0x6000) -#define FIRST_COMPUTE_VMID (8) -#define LAST_COMPUTE_VMID (16) static void gfx_v8_0_init_compute_vmid(struct amdgpu_device *adev) { int i; @@ -3754,7 +3667,7 @@ static void gfx_v8_0_init_compute_vmid(struct amdgpu_device *adev) SH_MEM_CONFIG__PRIVATE_ATC_MASK; 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++) { vi_srbm_select(adev, 0, 0, 0, i); /* CP and shaders */ WREG32(mmSH_MEM_CONFIG, sh_mem_config); @@ -3764,6 +3677,33 @@ static void gfx_v8_0_init_compute_vmid(struct amdgpu_device *adev) } vi_srbm_select(adev, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); + + /* Initialize all compute VMIDs to have no GDS, GWS, or OA + access. These should be enabled by FW for target VMIDs. */ + for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) { + WREG32(amdgpu_gds_reg_offset[i].mem_base, 0); + WREG32(amdgpu_gds_reg_offset[i].mem_size, 0); + WREG32(amdgpu_gds_reg_offset[i].gws, 0); + WREG32(amdgpu_gds_reg_offset[i].oa, 0); + } +} + +static void gfx_v8_0_init_gds_vmid(struct amdgpu_device *adev) +{ + int vmid; + + /* + * Initialize all compute and user-gfx VMIDs to have no GDS, GWS, or OA + * access. Compute VMIDs should be enabled by FW for target VMIDs, + * the driver can enable them for graphics. VMID0 should maintain + * access so that HWS firmware can save/restore entries. + */ + for (vmid = 1; vmid < AMDGPU_NUM_VMID; vmid++) { + WREG32(amdgpu_gds_reg_offset[vmid].mem_base, 0); + WREG32(amdgpu_gds_reg_offset[vmid].mem_size, 0); + WREG32(amdgpu_gds_reg_offset[vmid].gws, 0); + WREG32(amdgpu_gds_reg_offset[vmid].oa, 0); + } } static void gfx_v8_0_config_init(struct amdgpu_device *adev) @@ -3779,7 +3719,7 @@ static void gfx_v8_0_config_init(struct amdgpu_device *adev) } } -static void gfx_v8_0_gpu_init(struct amdgpu_device *adev) +static void gfx_v8_0_constants_init(struct amdgpu_device *adev) { u32 tmp, sh_static_mem_cfg; int i; @@ -3802,6 +3742,8 @@ static void gfx_v8_0_gpu_init(struct amdgpu_device *adev) ELEMENT_SIZE, 1); sh_static_mem_cfg = REG_SET_FIELD(sh_static_mem_cfg, SH_STATIC_MEM_CONFIG, INDEX_STRIDE, 3); + WREG32(mmSH_STATIC_MEM_CONFIG, sh_static_mem_cfg); + mutex_lock(&adev->srbm_mutex); for (i = 0; i < adev->vm_manager.id_mgr[0].num_ids; i++) { vi_srbm_select(adev, 0, 0, 0, i); @@ -3819,25 +3761,25 @@ static void gfx_v8_0_gpu_init(struct amdgpu_device *adev) tmp = REG_SET_FIELD(tmp, SH_MEM_CONFIG, ALIGNMENT_MODE, SH_MEM_ALIGNMENT_MODE_UNALIGNED); WREG32(mmSH_MEM_CONFIG, tmp); - tmp = adev->mc.shared_aperture_start >> 48; + tmp = adev->gmc.shared_aperture_start >> 48; WREG32(mmSH_MEM_BASES, tmp); } WREG32(mmSH_MEM_APE1_BASE, 1); WREG32(mmSH_MEM_APE1_LIMIT, 0); - WREG32(mmSH_STATIC_MEM_CONFIG, sh_static_mem_cfg); } vi_srbm_select(adev, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); gfx_v8_0_init_compute_vmid(adev); + gfx_v8_0_init_gds_vmid(adev); mutex_lock(&adev->grbm_idx_mutex); /* * making sure that the following register writes will be broadcasted * to all the shaders */ - gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0); WREG32(mmPA_SC_FIFO_SIZE, (adev->gfx.config.sc_prim_fifo_size_frontend << @@ -3868,15 +3810,23 @@ static void gfx_v8_0_wait_for_rlc_serdes(struct amdgpu_device *adev) mutex_lock(&adev->grbm_idx_mutex); for (i = 0; i < adev->gfx.config.max_shader_engines; i++) { for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) { - gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff); + gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff, 0); for (k = 0; k < adev->usec_timeout; k++) { if (RREG32(mmRLC_SERDES_CU_MASTER_BUSY) == 0) break; udelay(1); } + if (k == adev->usec_timeout) { + gfx_v8_0_select_se_sh(adev, 0xffffffff, + 0xffffffff, 0xffffffff, 0); + mutex_unlock(&adev->grbm_idx_mutex); + DRM_INFO("Timeout wait for RLC serdes %u,%u\n", + i, j); + return; + } } } - gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0); mutex_unlock(&adev->grbm_idx_mutex); mask = RLC_SERDES_NONCU_MASTER_BUSY__SE_MASTER_BUSY_MASK | @@ -3905,6 +3855,7 @@ static void gfx_v8_0_enable_gui_idle_interrupt(struct amdgpu_device *adev, static void gfx_v8_0_init_csb(struct amdgpu_device *adev) { + adev->gfx.rlc.funcs->get_csb_buffer(adev, adev->gfx.rlc.cs_ptr); /* csib */ WREG32(mmRLC_CSIB_ADDR_HI, adev->gfx.rlc.clear_state_gpu_addr >> 32); @@ -3974,21 +3925,20 @@ static int gfx_v8_0_init_save_restore_list(struct amdgpu_device *adev) int list_size; unsigned int *register_list_format = - kmalloc(adev->gfx.rlc.reg_list_format_size_bytes, GFP_KERNEL); + kmemdup(adev->gfx.rlc.register_list_format, + adev->gfx.rlc.reg_list_format_size_bytes, GFP_KERNEL); if (!register_list_format) return -ENOMEM; - memcpy(register_list_format, adev->gfx.rlc.register_list_format, - adev->gfx.rlc.reg_list_format_size_bytes); gfx_v8_0_parse_ind_reg_list(register_list_format, RLC_FormatDirectRegListLength, adev->gfx.rlc.reg_list_format_size_bytes >> 2, unique_indices, &indices_count, - sizeof(unique_indices) / sizeof(int), + ARRAY_SIZE(unique_indices), indirect_start_offsets, &offset_count, - sizeof(indirect_start_offsets)/sizeof(int)); + ARRAY_SIZE(indirect_start_offsets)); /* save and restore list */ WREG32_FIELD(RLC_SRM_CNTL, AUTO_INCR_ADDR, 1); @@ -4010,14 +3960,14 @@ static int gfx_v8_0_init_save_restore_list(struct amdgpu_device *adev) /* starting offsets starts */ WREG32(mmRLC_GPM_SCRATCH_ADDR, adev->gfx.rlc.starting_offsets_start); - for (i = 0; i < sizeof(indirect_start_offsets)/sizeof(int); i++) + for (i = 0; i < ARRAY_SIZE(indirect_start_offsets); i++) WREG32(mmRLC_GPM_SCRATCH_DATA, indirect_start_offsets[i]); /* unique indices */ temp = mmRLC_SRM_INDEX_CNTL_ADDR_0; data = mmRLC_SRM_INDEX_CNTL_DATA_0; - for (i = 0; i < sizeof(unique_indices) / sizeof(int); i++) { + for (i = 0; i < ARRAY_SIZE(unique_indices); i++) { if (unique_indices[i] != 0) { WREG32(temp + i, unique_indices[i] & 0x3FFFF); WREG32(data + i, unique_indices[i] >> 20); @@ -4078,7 +4028,8 @@ static void gfx_v8_0_init_pg(struct amdgpu_device *adev) gfx_v8_0_init_power_gating(adev); WREG32(mmRLC_PG_ALWAYS_ON_CU_MASK, adev->gfx.cu_info.ao_cu_mask); } else if ((adev->asic_type == CHIP_POLARIS11) || - (adev->asic_type == CHIP_POLARIS12)) { + (adev->asic_type == CHIP_POLARIS12) || + (adev->asic_type == CHIP_VEGAM)) { gfx_v8_0_init_csb(adev); gfx_v8_0_init_save_restore_list(adev); gfx_v8_0_enable_save_restore_machine(adev); @@ -4115,78 +4066,23 @@ static void gfx_v8_0_rlc_start(struct amdgpu_device *adev) udelay(50); } -static int gfx_v8_0_rlc_load_microcode(struct amdgpu_device *adev) -{ - const struct rlc_firmware_header_v2_0 *hdr; - const __le32 *fw_data; - unsigned i, fw_size; - - if (!adev->gfx.rlc_fw) - return -EINVAL; - - hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data; - amdgpu_ucode_print_rlc_hdr(&hdr->header); - - fw_data = (const __le32 *)(adev->gfx.rlc_fw->data + - le32_to_cpu(hdr->header.ucode_array_offset_bytes)); - fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4; - - WREG32(mmRLC_GPM_UCODE_ADDR, 0); - for (i = 0; i < fw_size; i++) - WREG32(mmRLC_GPM_UCODE_DATA, le32_to_cpup(fw_data++)); - WREG32(mmRLC_GPM_UCODE_ADDR, adev->gfx.rlc_fw_version); - - return 0; -} - static int gfx_v8_0_rlc_resume(struct amdgpu_device *adev) { - int r; - u32 tmp; - - gfx_v8_0_rlc_stop(adev); - - /* disable CG */ - tmp = RREG32(mmRLC_CGCG_CGLS_CTRL); - tmp &= ~(RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK | - RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK); - WREG32(mmRLC_CGCG_CGLS_CTRL, tmp); - if (adev->asic_type == CHIP_POLARIS11 || - adev->asic_type == CHIP_POLARIS10 || - adev->asic_type == CHIP_POLARIS12) { - tmp = RREG32(mmRLC_CGCG_CGLS_CTRL_3D); - tmp &= ~0x3; - WREG32(mmRLC_CGCG_CGLS_CTRL_3D, tmp); + if (amdgpu_sriov_vf(adev)) { + gfx_v8_0_init_csb(adev); + return 0; } - /* disable PG */ - WREG32(mmRLC_PG_CNTL, 0); - - gfx_v8_0_rlc_reset(adev); + adev->gfx.rlc.funcs->stop(adev); + adev->gfx.rlc.funcs->reset(adev); gfx_v8_0_init_pg(adev); - - if (!adev->pp_enabled) { - if (adev->firmware.load_type != AMDGPU_FW_LOAD_SMU) { - /* legacy rlc firmware loading */ - r = gfx_v8_0_rlc_load_microcode(adev); - if (r) - return r; - } else { - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_RLC_G); - if (r) - return -EINVAL; - } - } - - gfx_v8_0_rlc_start(adev); + adev->gfx.rlc.funcs->start(adev); return 0; } static void gfx_v8_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable) { - int i; u32 tmp = RREG32(mmCP_ME_CNTL); if (enable) { @@ -4197,70 +4093,11 @@ static void gfx_v8_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable) tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_HALT, 1); tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_HALT, 1); tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_HALT, 1); - for (i = 0; i < adev->gfx.num_gfx_rings; i++) - adev->gfx.gfx_ring[i].ready = false; } WREG32(mmCP_ME_CNTL, tmp); udelay(50); } -static int gfx_v8_0_cp_gfx_load_microcode(struct amdgpu_device *adev) -{ - const struct gfx_firmware_header_v1_0 *pfp_hdr; - const struct gfx_firmware_header_v1_0 *ce_hdr; - const struct gfx_firmware_header_v1_0 *me_hdr; - const __le32 *fw_data; - unsigned i, fw_size; - - if (!adev->gfx.me_fw || !adev->gfx.pfp_fw || !adev->gfx.ce_fw) - return -EINVAL; - - pfp_hdr = (const struct gfx_firmware_header_v1_0 *) - adev->gfx.pfp_fw->data; - ce_hdr = (const struct gfx_firmware_header_v1_0 *) - adev->gfx.ce_fw->data; - me_hdr = (const struct gfx_firmware_header_v1_0 *) - adev->gfx.me_fw->data; - - amdgpu_ucode_print_gfx_hdr(&pfp_hdr->header); - amdgpu_ucode_print_gfx_hdr(&ce_hdr->header); - amdgpu_ucode_print_gfx_hdr(&me_hdr->header); - - gfx_v8_0_cp_gfx_enable(adev, false); - - /* PFP */ - fw_data = (const __le32 *) - (adev->gfx.pfp_fw->data + - le32_to_cpu(pfp_hdr->header.ucode_array_offset_bytes)); - fw_size = le32_to_cpu(pfp_hdr->header.ucode_size_bytes) / 4; - WREG32(mmCP_PFP_UCODE_ADDR, 0); - for (i = 0; i < fw_size; i++) - WREG32(mmCP_PFP_UCODE_DATA, le32_to_cpup(fw_data++)); - WREG32(mmCP_PFP_UCODE_ADDR, adev->gfx.pfp_fw_version); - - /* CE */ - fw_data = (const __le32 *) - (adev->gfx.ce_fw->data + - le32_to_cpu(ce_hdr->header.ucode_array_offset_bytes)); - fw_size = le32_to_cpu(ce_hdr->header.ucode_size_bytes) / 4; - WREG32(mmCP_CE_UCODE_ADDR, 0); - for (i = 0; i < fw_size; i++) - WREG32(mmCP_CE_UCODE_DATA, le32_to_cpup(fw_data++)); - WREG32(mmCP_CE_UCODE_ADDR, adev->gfx.ce_fw_version); - - /* ME */ - fw_data = (const __le32 *) - (adev->gfx.me_fw->data + - le32_to_cpu(me_hdr->header.ucode_array_offset_bytes)); - fw_size = le32_to_cpu(me_hdr->header.ucode_size_bytes) / 4; - WREG32(mmCP_ME_RAM_WADDR, 0); - for (i = 0; i < fw_size; i++) - WREG32(mmCP_ME_RAM_DATA, le32_to_cpup(fw_data++)); - WREG32(mmCP_ME_RAM_WADDR, adev->gfx.me_fw_version); - - return 0; -} - static u32 gfx_v8_0_get_csb_size(struct amdgpu_device *adev) { u32 count = 0; @@ -4334,37 +4171,8 @@ static int gfx_v8_0_cp_gfx_start(struct amdgpu_device *adev) amdgpu_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 2)); amdgpu_ring_write(ring, mmPA_SC_RASTER_CONFIG - PACKET3_SET_CONTEXT_REG_START); - switch (adev->asic_type) { - case CHIP_TONGA: - case CHIP_POLARIS10: - amdgpu_ring_write(ring, 0x16000012); - amdgpu_ring_write(ring, 0x0000002A); - break; - case CHIP_POLARIS11: - case CHIP_POLARIS12: - amdgpu_ring_write(ring, 0x16000012); - amdgpu_ring_write(ring, 0x00000000); - break; - case CHIP_FIJI: - amdgpu_ring_write(ring, 0x3a00161a); - amdgpu_ring_write(ring, 0x0000002e); - break; - case CHIP_CARRIZO: - amdgpu_ring_write(ring, 0x00000002); - amdgpu_ring_write(ring, 0x00000000); - break; - case CHIP_TOPAZ: - amdgpu_ring_write(ring, adev->gfx.config.num_rbs == 1 ? - 0x00000000 : 0x00000002); - amdgpu_ring_write(ring, 0x00000000); - break; - case CHIP_STONEY: - amdgpu_ring_write(ring, 0x00000000); - amdgpu_ring_write(ring, 0x00000000); - break; - default: - BUG(); - } + amdgpu_ring_write(ring, adev->gfx.config.rb_config[0][0].raster_config); + amdgpu_ring_write(ring, adev->gfx.config.rb_config[0][0].raster_config_1); amdgpu_ring_write(ring, PACKET3(PACKET3_PREAMBLE_CNTL, 0)); amdgpu_ring_write(ring, PACKET3_PREAMBLE_END_CLEAR_STATE); @@ -4409,7 +4217,7 @@ static void gfx_v8_0_set_cpg_door_bell(struct amdgpu_device *adev, struct amdgpu tmp = REG_SET_FIELD(0, CP_RB_DOORBELL_RANGE_LOWER, DOORBELL_RANGE_LOWER, - AMDGPU_DOORBELL_GFX_RING0); + adev->doorbell_index.gfx_ring0); WREG32(mmCP_RB_DOORBELL_RANGE_LOWER, tmp); WREG32(mmCP_RB_DOORBELL_RANGE_UPPER, @@ -4422,7 +4230,6 @@ static int gfx_v8_0_cp_gfx_resume(struct amdgpu_device *adev) u32 tmp; u32 rb_bufsz; u64 rb_addr, rptr_addr, wptr_gpu_addr; - int r; /* Set the write pointer delay */ WREG32(mmCP_RB_WPTR_DELAY, 0); @@ -4447,12 +4254,12 @@ static int gfx_v8_0_cp_gfx_resume(struct amdgpu_device *adev) ring->wptr = 0; WREG32(mmCP_RB0_WPTR, lower_32_bits(ring->wptr)); - /* set the wb address wether it's enabled or not */ - rptr_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); + /* set the wb address whether it's enabled or not */ + rptr_addr = ring->rptr_gpu_addr; WREG32(mmCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr)); WREG32(mmCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & 0xFF); - wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + wptr_gpu_addr = ring->wptr_gpu_addr; WREG32(mmCP_RB_WPTR_POLL_ADDR_LO, lower_32_bits(wptr_gpu_addr)); WREG32(mmCP_RB_WPTR_POLL_ADDR_HI, upper_32_bits(wptr_gpu_addr)); mdelay(1); @@ -4466,75 +4273,21 @@ static int gfx_v8_0_cp_gfx_resume(struct amdgpu_device *adev) /* start the ring */ amdgpu_ring_clear_ring(ring); gfx_v8_0_cp_gfx_start(adev); - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) - ring->ready = false; - return r; + return 0; } static void gfx_v8_0_cp_compute_enable(struct amdgpu_device *adev, bool enable) { - int i; - if (enable) { WREG32(mmCP_MEC_CNTL, 0); } else { WREG32(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].ready = false; - adev->gfx.kiq.ring.ready = false; + adev->gfx.kiq[0].ring.sched.ready = false; } udelay(50); } -static int gfx_v8_0_cp_compute_load_microcode(struct amdgpu_device *adev) -{ - const struct gfx_firmware_header_v1_0 *mec_hdr; - const __le32 *fw_data; - unsigned i, fw_size; - - if (!adev->gfx.mec_fw) - return -EINVAL; - - gfx_v8_0_cp_compute_enable(adev, false); - - mec_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data; - amdgpu_ucode_print_gfx_hdr(&mec_hdr->header); - - 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; - - /* MEC1 */ - WREG32(mmCP_MEC_ME1_UCODE_ADDR, 0); - for (i = 0; i < fw_size; i++) - WREG32(mmCP_MEC_ME1_UCODE_DATA, le32_to_cpup(fw_data+i)); - WREG32(mmCP_MEC_ME1_UCODE_ADDR, adev->gfx.mec_fw_version); - - /* Loading MEC2 firmware is only necessary if MEC2 should run different microcode than MEC1. */ - if (adev->gfx.mec2_fw) { - const struct gfx_firmware_header_v1_0 *mec2_hdr; - - mec2_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec2_fw->data; - amdgpu_ucode_print_gfx_hdr(&mec2_hdr->header); - - fw_data = (const __le32 *) - (adev->gfx.mec2_fw->data + - le32_to_cpu(mec2_hdr->header.ucode_array_offset_bytes)); - fw_size = le32_to_cpu(mec2_hdr->header.ucode_size_bytes) / 4; - - WREG32(mmCP_MEC_ME2_UCODE_ADDR, 0); - for (i = 0; i < fw_size; i++) - WREG32(mmCP_MEC_ME2_UCODE_DATA, le32_to_cpup(fw_data+i)); - WREG32(mmCP_MEC_ME2_UCODE_ADDR, adev->gfx.mec2_fw_version); - } - - return 0; -} - /* KIQ functions */ static void gfx_v8_0_kiq_setting(struct amdgpu_ring *ring) { @@ -4545,26 +4298,23 @@ static void gfx_v8_0_kiq_setting(struct amdgpu_ring *ring) tmp = RREG32(mmRLC_CP_SCHEDULERS); tmp &= 0xffffff00; tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue); - WREG32(mmRLC_CP_SCHEDULERS, tmp); - tmp |= 0x80; - WREG32(mmRLC_CP_SCHEDULERS, tmp); + WREG32(mmRLC_CP_SCHEDULERS, tmp | 0x80); } static int gfx_v8_0_kiq_kcq_enable(struct amdgpu_device *adev) { - struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring; - uint32_t scratch, tmp = 0; + struct amdgpu_ring *kiq_ring = &adev->gfx.kiq[0].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)) + if (!test_bit(i, adev->gfx.mec_bitmap[0].queue_bitmap)) continue; /* 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))) { + if (WARN_ON(i >= (sizeof(queue_mask)*8))) { DRM_ERROR("Invalid KCQ enabled: %d\n", i); break; } @@ -4572,17 +4322,9 @@ static int gfx_v8_0_kiq_kcq_enable(struct amdgpu_device *adev) queue_mask |= (1ull << i); } - r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) { - DRM_ERROR("Failed to get scratch reg (%d).\n", r); - return r; - } - WREG32(scratch, 0xCAFEDEAD); - - r = amdgpu_ring_alloc(kiq_ring, (8 * adev->gfx.num_compute_rings) + 11); + r = amdgpu_ring_alloc(kiq_ring, (8 * adev->gfx.num_compute_rings) + 8); if (r) { DRM_ERROR("Failed to lock KIQ (%d).\n", r); - amdgpu_gfx_scratch_free(adev, scratch); return r; } /* set resources */ @@ -4597,7 +4339,7 @@ static int gfx_v8_0_kiq_kcq_enable(struct amdgpu_device *adev) 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); + uint64_t wptr_addr = ring->wptr_gpu_addr; /* map queues */ amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_MAP_QUEUES, 5)); @@ -4614,26 +4356,10 @@ static int gfx_v8_0_kiq_kcq_enable(struct amdgpu_device *adev) amdgpu_ring_write(kiq_ring, lower_32_bits(wptr_addr)); amdgpu_ring_write(kiq_ring, upper_32_bits(wptr_addr)); } - /* write to scratch for completion */ - amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); - amdgpu_ring_write(kiq_ring, (scratch - PACKET3_SET_UCONFIG_REG_START)); - amdgpu_ring_write(kiq_ring, 0xDEADBEEF); - amdgpu_ring_commit(kiq_ring); - for (i = 0; i < adev->usec_timeout; i++) { - tmp = RREG32(scratch); - if (tmp == 0xDEADBEEF) - break; - DRM_UDELAY(1); - } - if (i >= adev->usec_timeout) { - DRM_ERROR("KCQ enable failed (scratch(0x%04X)=0x%08X)\n", - scratch, tmp); - r = -EINVAL; - } - amdgpu_gfx_scratch_free(adev, scratch); + amdgpu_ring_commit(kiq_ring); - return r; + return 0; } static int gfx_v8_0_deactivate_hqd(struct amdgpu_device *adev, u32 req) @@ -4657,6 +4383,19 @@ static int gfx_v8_0_deactivate_hqd(struct amdgpu_device *adev, u32 req) return r; } +static void gfx_v8_0_mqd_set_priority(struct amdgpu_ring *ring, struct vi_mqd *mqd) +{ + struct amdgpu_device *adev = ring->adev; + + 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; + } + } +} + static int gfx_v8_0_mqd_init(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; @@ -4671,12 +4410,10 @@ static int gfx_v8_0_mqd_init(struct amdgpu_ring *ring) mqd->compute_static_thread_mgmt_se2 = 0xffffffff; mqd->compute_static_thread_mgmt_se3 = 0xffffffff; mqd->compute_misc_reserved = 0x00000003; - if (!(adev->flags & AMD_IS_APU)) { - mqd->dynamic_cu_mask_addr_lo = lower_32_bits(ring->mqd_gpu_addr - + offsetof(struct vi_mqd_allocation, dyamic_cu_mask)); - mqd->dynamic_cu_mask_addr_hi = upper_32_bits(ring->mqd_gpu_addr - + offsetof(struct vi_mqd_allocation, dyamic_cu_mask)); - } + mqd->dynamic_cu_mask_addr_lo = lower_32_bits(ring->mqd_gpu_addr + + offsetof(struct vi_mqd_allocation, dynamic_cu_mask)); + mqd->dynamic_cu_mask_addr_hi = upper_32_bits(ring->mqd_gpu_addr + + offsetof(struct vi_mqd_allocation, dynamic_cu_mask)); eop_base_addr = ring->eop_gpu_addr >> 8; mqd->cp_hqd_eop_base_addr_lo = eop_base_addr; mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr); @@ -4715,7 +4452,7 @@ static int gfx_v8_0_mqd_init(struct amdgpu_ring *ring) tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE, (order_base_2(ring->ring_size / 4) - 1)); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE, - ((order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1) << 8)); + (order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1)); #ifdef __BIG_ENDIAN tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, ENDIAN_SWAP, 1); #endif @@ -4726,13 +4463,13 @@ static int gfx_v8_0_mqd_init(struct amdgpu_ring *ring) mqd->cp_hqd_pq_control = tmp; /* set the wb address whether it's enabled or not */ - wb_gpu_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4); + wb_gpu_addr = ring->rptr_gpu_addr; mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc; mqd->cp_hqd_pq_rptr_report_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; /* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */ - wb_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + wb_gpu_addr = ring->wptr_gpu_addr; mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffffc; mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; @@ -4782,9 +4519,6 @@ static int gfx_v8_0_mqd_init(struct amdgpu_ring *ring) /* defaults */ mqd->cp_hqd_eop_rptr = RREG32(mmCP_HQD_EOP_RPTR); mqd->cp_hqd_eop_wptr = RREG32(mmCP_HQD_EOP_WPTR); - mqd->cp_hqd_pipe_priority = RREG32(mmCP_HQD_PIPE_PRIORITY); - mqd->cp_hqd_queue_priority = RREG32(mmCP_HQD_QUEUE_PRIORITY); - mqd->cp_hqd_quantum = RREG32(mmCP_HQD_QUANTUM); mqd->cp_hqd_ctx_save_base_addr_lo = RREG32(mmCP_HQD_CTX_SAVE_BASE_ADDR_LO); mqd->cp_hqd_ctx_save_base_addr_hi = RREG32(mmCP_HQD_CTX_SAVE_BASE_ADDR_HI); mqd->cp_hqd_cntl_stack_offset = RREG32(mmCP_HQD_CNTL_STACK_OFFSET); @@ -4796,13 +4530,20 @@ static int gfx_v8_0_mqd_init(struct amdgpu_ring *ring) mqd->cp_hqd_eop_wptr_mem = RREG32(mmCP_HQD_EOP_WPTR_MEM); mqd->cp_hqd_eop_dones = RREG32(mmCP_HQD_EOP_DONES); - /* activate the queue */ - mqd->cp_hqd_active = 1; + /* set static priority for a queue/ring */ + gfx_v8_0_mqd_set_priority(ring, mqd); + mqd->cp_hqd_quantum = RREG32(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; } -int gfx_v8_0_mqd_commit(struct amdgpu_device *adev, +static int gfx_v8_0_mqd_commit(struct amdgpu_device *adev, struct vi_mqd *mqd) { uint32_t mqd_reg; @@ -4843,14 +4584,13 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; struct vi_mqd *mqd = ring->mqd_ptr; - int mqd_idx = AMDGPU_MAX_COMPUTE_RINGS; gfx_v8_0_kiq_setting(ring); - if (adev->gfx.in_reset) { /* for GPU_RESET case */ + 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 vi_mqd_allocation)); + if (adev->gfx.kiq[0].mqd_backup) + memcpy(mqd, adev->gfx.kiq[0].mqd_backup, sizeof(struct vi_mqd_allocation)); /* reset ring buffer */ ring->wptr = 0; @@ -4862,8 +4602,10 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring) mutex_unlock(&adev->srbm_mutex); } else { memset((void *)mqd, 0, sizeof(struct vi_mqd_allocation)); - ((struct vi_mqd_allocation *)mqd)->dyamic_cu_mask = 0xFFFFFFFF; - ((struct vi_mqd_allocation *)mqd)->dyamic_rb_mask = 0xFFFFFFFF; + ((struct vi_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; + ((struct vi_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; + if (amdgpu_sriov_vf(adev) && adev->in_suspend) + amdgpu_ring_clear_ring(ring); mutex_lock(&adev->srbm_mutex); vi_srbm_select(adev, ring->me, ring->pipe, ring->queue, 0); gfx_v8_0_mqd_init(ring); @@ -4871,8 +4613,8 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring) vi_srbm_select(adev, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); - if (adev->gfx.mec.mqd_backup[mqd_idx]) - memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct vi_mqd_allocation)); + if (adev->gfx.kiq[0].mqd_backup) + memcpy(adev->gfx.kiq[0].mqd_backup, mqd, sizeof(struct vi_mqd_allocation)); } return 0; @@ -4884,10 +4626,10 @@ static int gfx_v8_0_kcq_init_queue(struct amdgpu_ring *ring) struct vi_mqd *mqd = ring->mqd_ptr; int mqd_idx = ring - &adev->gfx.compute_ring[0]; - if (!adev->gfx.in_reset && !adev->gfx.in_suspend) { + if (!amdgpu_in_reset(adev) && !adev->in_suspend) { memset((void *)mqd, 0, sizeof(struct vi_mqd_allocation)); - ((struct vi_mqd_allocation *)mqd)->dyamic_cu_mask = 0xFFFFFFFF; - ((struct vi_mqd_allocation *)mqd)->dyamic_rb_mask = 0xFFFFFFFF; + ((struct vi_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; + ((struct vi_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; mutex_lock(&adev->srbm_mutex); vi_srbm_select(adev, ring->me, ring->pipe, ring->queue, 0); gfx_v8_0_mqd_init(ring); @@ -4896,14 +4638,13 @@ static int gfx_v8_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 vi_mqd_allocation)); - } else if (adev->gfx.in_reset) { /* for GPU_RESET case */ - /* reset MQD to a clean status */ + } else { + /* restore MQD to a clean status */ if (adev->gfx.mec.mqd_backup[mqd_idx]) memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct vi_mqd_allocation)); /* reset ring buffer */ ring->wptr = 0; - amdgpu_ring_clear_ring(ring); - } else { + atomic64_set((atomic64_t *)ring->wptr_cpu_addr, 0); amdgpu_ring_clear_ring(ring); } return 0; @@ -4912,8 +4653,8 @@ static int gfx_v8_0_kcq_init_queue(struct amdgpu_ring *ring) static void gfx_v8_0_set_mec_doorbell_range(struct amdgpu_device *adev) { if (adev->asic_type > CHIP_TONGA) { - WREG32(mmCP_MEC_DOORBELL_RANGE_LOWER, AMDGPU_DOORBELL_KIQ << 2); - WREG32(mmCP_MEC_DOORBELL_RANGE_UPPER, AMDGPU_DOORBELL_MEC_RING7 << 2); + WREG32(mmCP_MEC_DOORBELL_RANGE_LOWER, adev->doorbell_index.kiq << 2); + WREG32(mmCP_MEC_DOORBELL_RANGE_UPPER, adev->doorbell_index.mec_ring7 << 2); } /* enable doorbells */ WREG32_FIELD(CP_PQ_STATUS, DOORBELL_ENABLE, 1); @@ -4921,70 +4662,49 @@ static void gfx_v8_0_set_mec_doorbell_range(struct amdgpu_device *adev) static int gfx_v8_0_kiq_resume(struct amdgpu_device *adev) { - struct amdgpu_ring *ring = NULL; - int r = 0, i; - - gfx_v8_0_cp_compute_enable(adev, true); - - ring = &adev->gfx.kiq.ring; + gfx_v8_0_kiq_init_queue(&adev->gfx.kiq[0].ring); + return 0; +} - r = amdgpu_bo_reserve(ring->mqd_obj, false); - if (unlikely(r != 0)) - goto done; +static int gfx_v8_0_kcq_resume(struct amdgpu_device *adev) +{ + int i, r; - r = amdgpu_bo_kmap(ring->mqd_obj, &ring->mqd_ptr); - if (!r) { - r = gfx_v8_0_kiq_init_queue(ring); - amdgpu_bo_kunmap(ring->mqd_obj); - ring->mqd_ptr = NULL; - } - amdgpu_bo_unreserve(ring->mqd_obj); - if (r) - goto done; + gfx_v8_0_cp_compute_enable(adev, true); for (i = 0; i < adev->gfx.num_compute_rings; i++) { - ring = &adev->gfx.compute_ring[i]; - - r = amdgpu_bo_reserve(ring->mqd_obj, false); - if (unlikely(r != 0)) - goto done; - r = amdgpu_bo_kmap(ring->mqd_obj, &ring->mqd_ptr); - if (!r) { - r = gfx_v8_0_kcq_init_queue(ring); - amdgpu_bo_kunmap(ring->mqd_obj); - ring->mqd_ptr = NULL; - } - amdgpu_bo_unreserve(ring->mqd_obj); + r = gfx_v8_0_kcq_init_queue(&adev->gfx.compute_ring[i]); if (r) - goto done; + return r; } gfx_v8_0_set_mec_doorbell_range(adev); - r = gfx_v8_0_kiq_kcq_enable(adev); + return gfx_v8_0_kiq_kcq_enable(adev); +} + +static int gfx_v8_0_cp_test_all_rings(struct amdgpu_device *adev) +{ + int r, i; + struct amdgpu_ring *ring; + + /* collect all the ring_tests here, gfx, kiq, compute */ + ring = &adev->gfx.gfx_ring[0]; + r = amdgpu_ring_test_helper(ring); if (r) - goto done; + return r; - /* Test KIQ */ - ring = &adev->gfx.kiq.ring; - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; - goto done; - } + ring = &adev->gfx.kiq[0].ring; + r = amdgpu_ring_test_helper(ring); + if (r) + return r; - /* Test KCQs */ for (i = 0; i < adev->gfx.num_compute_rings; i++) { ring = &adev->gfx.compute_ring[i]; - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) - ring->ready = false; + amdgpu_ring_test_helper(ring); } -done: - return r; + return 0; } static int gfx_v8_0_cp_resume(struct amdgpu_device *adev) @@ -4994,50 +4714,19 @@ static int gfx_v8_0_cp_resume(struct amdgpu_device *adev) if (!(adev->flags & AMD_IS_APU)) gfx_v8_0_enable_gui_idle_interrupt(adev, false); - if (!adev->pp_enabled) { - if (adev->firmware.load_type != AMDGPU_FW_LOAD_SMU) { - /* legacy firmware loading */ - r = gfx_v8_0_cp_gfx_load_microcode(adev); - if (r) - return r; - - r = gfx_v8_0_cp_compute_load_microcode(adev); - if (r) - return r; - } else { - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_CP_CE); - if (r) - return -EINVAL; - - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_CP_PFP); - if (r) - return -EINVAL; - - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_CP_ME); - if (r) - return -EINVAL; - - if (adev->asic_type == CHIP_TOPAZ) { - r = gfx_v8_0_cp_compute_load_microcode(adev); - if (r) - return r; - } else { - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_CP_MEC1); - if (r) - return -EINVAL; - } - } - } + r = gfx_v8_0_kiq_resume(adev); + if (r) + return r; r = gfx_v8_0_cp_gfx_resume(adev); if (r) return r; - r = gfx_v8_0_kiq_resume(adev); + r = gfx_v8_0_kcq_resume(adev); + if (r) + return r; + + r = gfx_v8_0_cp_test_all_rings(adev); if (r) return r; @@ -5052,15 +4741,15 @@ static void gfx_v8_0_cp_enable(struct amdgpu_device *adev, bool enable) gfx_v8_0_cp_compute_enable(adev, enable); } -static int gfx_v8_0_hw_init(void *handle) +static int gfx_v8_0_hw_init(struct amdgpu_ip_block *ip_block) { int r; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; gfx_v8_0_init_golden_registers(adev); - gfx_v8_0_gpu_init(adev); + gfx_v8_0_constants_init(adev); - r = gfx_v8_0_rlc_resume(adev); + r = adev->gfx.rlc.funcs->resume(adev); if (r) return r; @@ -5069,59 +4758,71 @@ static int gfx_v8_0_hw_init(void *handle) return r; } -static int gfx_v8_0_hw_fini(void *handle) +static int gfx_v8_0_kcq_disable(struct amdgpu_device *adev) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int r, i; + struct amdgpu_ring *kiq_ring = &adev->gfx.kiq[0].ring; - amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0); - amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0); - if (amdgpu_sriov_vf(adev)) { - pr_debug("For SRIOV client, shouldn't do anything.\n"); - return 0; - } - gfx_v8_0_cp_enable(adev, false); - gfx_v8_0_rlc_stop(adev); + r = amdgpu_ring_alloc(kiq_ring, 6 * adev->gfx.num_compute_rings); + if (r) + DRM_ERROR("Failed to lock KIQ (%d).\n", r); - amdgpu_set_powergating_state(adev, - AMD_IP_BLOCK_TYPE_GFX, AMD_PG_STATE_UNGATE); + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + struct amdgpu_ring *ring = &adev->gfx.compute_ring[i]; - return 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(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); + } + /* Submit unmap queue packet */ + amdgpu_ring_commit(kiq_ring); + /* + * Ring test will do a basic scratch register change check. Just run + * this to ensure that unmap queues that is submitted before got + * processed successfully before returning. + */ + r = amdgpu_ring_test_helper(kiq_ring); + if (r) + DRM_ERROR("KCQ disable failed\n"); -static int gfx_v8_0_suspend(void *handle) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - adev->gfx.in_suspend = true; - return gfx_v8_0_hw_fini(adev); + return r; } -static int gfx_v8_0_resume(void *handle) +static bool gfx_v8_0_is_idle(struct amdgpu_ip_block *ip_block) { - int r; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; - r = gfx_v8_0_hw_init(adev); - adev->gfx.in_suspend = false; - return r; + if (REG_GET_FIELD(RREG32(mmGRBM_STATUS), GRBM_STATUS, GUI_ACTIVE) + || RREG32(mmGRBM_STATUS2) != 0x8) + return false; + else + return true; } -static bool gfx_v8_0_is_idle(void *handle) +static bool gfx_v8_0_rlc_is_idle(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (REG_GET_FIELD(RREG32(mmGRBM_STATUS), GRBM_STATUS, GUI_ACTIVE)) + if (RREG32(mmGRBM_STATUS2) != 0x8) return false; else return true; } -static int gfx_v8_0_wait_for_idle(void *handle) +static int gfx_v8_0_wait_for_rlc_idle(void *handle) { - unsigned i; + unsigned int i; struct amdgpu_device *adev = (struct amdgpu_device *)handle; for (i = 0; i < adev->usec_timeout; i++) { - if (gfx_v8_0_is_idle(handle)) + if (gfx_v8_0_rlc_is_idle(handle)) return 0; udelay(1); @@ -5129,9 +4830,66 @@ static int gfx_v8_0_wait_for_idle(void *handle) return -ETIMEDOUT; } -static bool gfx_v8_0_check_soft_reset(void *handle) +static int gfx_v8_0_wait_for_idle(struct amdgpu_ip_block *ip_block) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + unsigned int i; + struct amdgpu_device *adev = ip_block->adev; + + for (i = 0; i < adev->usec_timeout; i++) { + if (gfx_v8_0_is_idle(ip_block)) + return 0; + + udelay(1); + } + return -ETIMEDOUT; +} + +static int gfx_v8_0_hw_fini(struct amdgpu_ip_block *ip_block) +{ + struct amdgpu_device *adev = ip_block->adev; + + amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0); + amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0); + + amdgpu_irq_put(adev, &adev->gfx.cp_ecc_error_irq, 0); + + amdgpu_irq_put(adev, &adev->gfx.sq_irq, 0); + + /* disable KCQ to avoid CPC touch memory not valid anymore */ + gfx_v8_0_kcq_disable(adev); + + if (amdgpu_sriov_vf(adev)) { + pr_debug("For SRIOV client, shouldn't do anything.\n"); + return 0; + } + + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); + if (!gfx_v8_0_wait_for_idle(ip_block)) + gfx_v8_0_cp_enable(adev, false); + else + pr_err("cp is busy, skip halt cp\n"); + if (!gfx_v8_0_wait_for_rlc_idle(adev)) + adev->gfx.rlc.funcs->stop(adev); + else + pr_err("rlc is busy, skip halt rlc\n"); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); + + return 0; +} + +static int gfx_v8_0_suspend(struct amdgpu_ip_block *ip_block) +{ + return gfx_v8_0_hw_fini(ip_block); +} + +static int gfx_v8_0_resume(struct amdgpu_ip_block *ip_block) +{ + return gfx_v8_0_hw_init(ip_block); +} + +static bool gfx_v8_0_check_soft_reset(struct amdgpu_ip_block *ip_block) +{ + struct amdgpu_device *adev = ip_block->adev; u32 grbm_soft_reset = 0, srbm_soft_reset = 0; u32 tmp; @@ -5191,20 +4949,19 @@ static bool gfx_v8_0_check_soft_reset(void *handle) } } -static int gfx_v8_0_pre_soft_reset(void *handle) +static int gfx_v8_0_pre_soft_reset(struct amdgpu_ip_block *ip_block) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - u32 grbm_soft_reset = 0, srbm_soft_reset = 0; + struct amdgpu_device *adev = ip_block->adev; + u32 grbm_soft_reset = 0; if ((!adev->gfx.grbm_soft_reset) && (!adev->gfx.srbm_soft_reset)) return 0; grbm_soft_reset = adev->gfx.grbm_soft_reset; - srbm_soft_reset = adev->gfx.srbm_soft_reset; /* stop the rlc */ - gfx_v8_0_rlc_stop(adev); + adev->gfx.rlc.funcs->stop(adev); if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) || REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_GFX)) @@ -5230,12 +4987,12 @@ static int gfx_v8_0_pre_soft_reset(void *handle) gfx_v8_0_cp_compute_enable(adev, false); } - return 0; + return 0; } -static int gfx_v8_0_soft_reset(void *handle) +static int gfx_v8_0_soft_reset(struct amdgpu_ip_block *ip_block) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; u32 grbm_soft_reset = 0, srbm_soft_reset = 0; u32 tmp; @@ -5295,21 +5052,16 @@ static int gfx_v8_0_soft_reset(void *handle) return 0; } -static int gfx_v8_0_post_soft_reset(void *handle) +static int gfx_v8_0_post_soft_reset(struct amdgpu_ip_block *ip_block) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - u32 grbm_soft_reset = 0, srbm_soft_reset = 0; + struct amdgpu_device *adev = ip_block->adev; + u32 grbm_soft_reset = 0; if ((!adev->gfx.grbm_soft_reset) && (!adev->gfx.srbm_soft_reset)) return 0; grbm_soft_reset = adev->gfx.grbm_soft_reset; - srbm_soft_reset = adev->gfx.srbm_soft_reset; - - if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) || - REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_GFX)) - gfx_v8_0_cp_gfx_resume(adev); if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) || REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CPF) || @@ -5327,8 +5079,16 @@ static int gfx_v8_0_post_soft_reset(void *handle) mutex_unlock(&adev->srbm_mutex); } gfx_v8_0_kiq_resume(adev); + gfx_v8_0_kcq_resume(adev); } - gfx_v8_0_rlc_start(adev); + + if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) || + REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_GFX)) + gfx_v8_0_cp_gfx_resume(adev); + + gfx_v8_0_cp_test_all_rings(adev); + + adev->gfx.rlc.funcs->start(adev); return 0; } @@ -5359,15 +5119,6 @@ static void gfx_v8_0_ring_emit_gds_switch(struct amdgpu_ring *ring, uint32_t gws_base, uint32_t gws_size, uint32_t oa_base, uint32_t oa_size) { - gds_base = gds_base >> AMDGPU_GDS_SHIFT; - gds_size = gds_size >> AMDGPU_GDS_SHIFT; - - gws_base = gws_base >> AMDGPU_GWS_SHIFT; - gws_size = gws_size >> AMDGPU_GWS_SHIFT; - - oa_base = oa_base >> AMDGPU_OA_SHIFT; - oa_size = oa_size >> AMDGPU_OA_SHIFT; - /* GDS Base */ amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | @@ -5426,7 +5177,7 @@ static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd, *(out++) = RREG32(mmSQ_IND_DATA); } -static void gfx_v8_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields) +static void gfx_v8_0_read_wave_data(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields) { /* type 0 wave data */ dst[(*no_fields)++] = 0; @@ -5448,9 +5199,10 @@ static void gfx_v8_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, u dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_TMA_HI); 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_v8_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd, +static void gfx_v8_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t start, uint32_t size, uint32_t *dst) { @@ -5465,14 +5217,17 @@ static const struct amdgpu_gfx_funcs gfx_v8_0_gfx_funcs = { .select_se_sh = &gfx_v8_0_select_se_sh, .read_wave_data = &gfx_v8_0_read_wave_data, .read_wave_sgprs = &gfx_v8_0_read_wave_sgprs, + .select_me_pipe_q = &gfx_v8_0_select_me_pipe_q }; -static int gfx_v8_0_early_init(void *handle) +static int gfx_v8_0_early_init(struct amdgpu_ip_block *ip_block) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; + adev->gfx.xcc_mask = 1; adev->gfx.num_gfx_rings = GFX8_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); adev->gfx.funcs = &gfx_v8_0_gfx_funcs; gfx_v8_0_set_ring_funcs(adev); gfx_v8_0_set_irq_funcs(adev); @@ -5482,9 +5237,9 @@ static int gfx_v8_0_early_init(void *handle) return 0; } -static int gfx_v8_0_late_init(void *handle) +static int gfx_v8_0_late_init(struct amdgpu_ip_block *ip_block) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; int r; r = amdgpu_irq_get(adev, &adev->gfx.priv_reg_irq, 0); @@ -5500,8 +5255,19 @@ static int gfx_v8_0_late_init(void *handle) if (r) return r; - amdgpu_set_powergating_state(adev, - AMD_IP_BLOCK_TYPE_GFX, AMD_PG_STATE_GATE); + r = amdgpu_irq_get(adev, &adev->gfx.cp_ecc_error_irq, 0); + if (r) { + DRM_ERROR("amdgpu_irq_get() failed to get IRQ for EDC, r: %d.\n", r); + return r; + } + + r = amdgpu_irq_get(adev, &adev->gfx.sq_irq, 0); + if (r) { + DRM_ERROR( + "amdgpu_irq_get() failed to get IRQ for SQ, r: %d.\n", + r); + return r; + } return 0; } @@ -5510,12 +5276,10 @@ static void gfx_v8_0_enable_gfx_static_mg_power_gating(struct amdgpu_device *ade bool enable) { if ((adev->asic_type == CHIP_POLARIS11) || - (adev->asic_type == CHIP_POLARIS12)) + (adev->asic_type == CHIP_POLARIS12) || + (adev->asic_type == CHIP_VEGAM)) /* Send msg to SMU via Powerplay */ - amdgpu_set_powergating_state(adev, - AMD_IP_BLOCK_TYPE_SMC, - enable ? - AMD_PG_STATE_GATE : AMD_PG_STATE_UNGATE); + amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, enable, 0); WREG32_FIELD(RLC_PG_CNTL, STATIC_PER_CU_PG_ENABLE, enable ? 1 : 0); } @@ -5561,15 +5325,20 @@ static void cz_update_gfx_cg_power_gating(struct amdgpu_device *adev, } } -static int gfx_v8_0_set_powergating_state(void *handle, +static int gfx_v8_0_set_powergating_state(struct amdgpu_ip_block *ip_block, enum amd_powergating_state state) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; bool enable = (state == AMD_PG_STATE_GATE); if (amdgpu_sriov_vf(adev)) return 0; + if (adev->pg_flags & (AMD_PG_SUPPORT_GFX_SMG | + AMD_PG_SUPPORT_RLC_SMU_HS | + AMD_PG_SUPPORT_CP | + AMD_PG_SUPPORT_GFX_DMG)) + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); switch (adev->asic_type) { case CHIP_CARRIZO: case CHIP_STONEY: @@ -5600,6 +5369,7 @@ static int gfx_v8_0_set_powergating_state(void *handle, break; case CHIP_POLARIS11: case CHIP_POLARIS12: + case CHIP_VEGAM: if ((adev->pg_flags & AMD_PG_SUPPORT_GFX_SMG) && enable) gfx_v8_0_enable_gfx_static_mg_power_gating(adev, true); else @@ -5618,13 +5388,17 @@ static int gfx_v8_0_set_powergating_state(void *handle, default: break; } - + if (adev->pg_flags & (AMD_PG_SUPPORT_GFX_SMG | + AMD_PG_SUPPORT_RLC_SMU_HS | + AMD_PG_SUPPORT_CP | + AMD_PG_SUPPORT_GFX_DMG)) + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); return 0; } -static void gfx_v8_0_get_clockgating_state(void *handle, u32 *flags) +static void gfx_v8_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; int data; if (amdgpu_sriov_vf(adev)) @@ -5669,7 +5443,7 @@ static void gfx_v8_0_send_serdes_cmd(struct amdgpu_device *adev, { uint32_t data; - gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0); WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff); WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff); @@ -5712,57 +5486,53 @@ static void gfx_v8_0_send_serdes_cmd(struct amdgpu_device *adev, #define RLC_GPR_REG2__MESSAGE__SHIFT 0x00000001 #define RLC_GPR_REG2__MESSAGE_MASK 0x0000001e -static void iceland_enter_rlc_safe_mode(struct amdgpu_device *adev) +static bool gfx_v8_0_is_rlc_enabled(struct amdgpu_device *adev) { - u32 data; - unsigned i; + uint32_t rlc_setting; - data = RREG32(mmRLC_CNTL); - if (!(data & RLC_CNTL__RLC_ENABLE_F32_MASK)) - return; + rlc_setting = RREG32(mmRLC_CNTL); + if (!(rlc_setting & RLC_CNTL__RLC_ENABLE_F32_MASK)) + return false; - if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_MGCG)) { - data |= RLC_SAFE_MODE__CMD_MASK; - data &= ~RLC_SAFE_MODE__MESSAGE_MASK; - data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT); - WREG32(mmRLC_SAFE_MODE, data); + return true; +} - for (i = 0; i < adev->usec_timeout; i++) { - if ((RREG32(mmRLC_GPM_STAT) & - (RLC_GPM_STAT__GFX_CLOCK_STATUS_MASK | - RLC_GPM_STAT__GFX_POWER_STATUS_MASK)) == - (RLC_GPM_STAT__GFX_CLOCK_STATUS_MASK | - RLC_GPM_STAT__GFX_POWER_STATUS_MASK)) - break; - udelay(1); - } +static void gfx_v8_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id) +{ + uint32_t data; + unsigned i; + data = RREG32(mmRLC_CNTL); + data |= RLC_SAFE_MODE__CMD_MASK; + data &= ~RLC_SAFE_MODE__MESSAGE_MASK; + data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT); + WREG32(mmRLC_SAFE_MODE, data); - for (i = 0; i < adev->usec_timeout; i++) { - if (!REG_GET_FIELD(RREG32(mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD)) - break; - udelay(1); - } - adev->gfx.rlc.in_safe_mode = true; + /* wait for RLC_SAFE_MODE */ + for (i = 0; i < adev->usec_timeout; i++) { + if ((RREG32(mmRLC_GPM_STAT) & + (RLC_GPM_STAT__GFX_CLOCK_STATUS_MASK | + RLC_GPM_STAT__GFX_POWER_STATUS_MASK)) == + (RLC_GPM_STAT__GFX_CLOCK_STATUS_MASK | + RLC_GPM_STAT__GFX_POWER_STATUS_MASK)) + break; + udelay(1); + } + for (i = 0; i < adev->usec_timeout; i++) { + if (!REG_GET_FIELD(RREG32(mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD)) + break; + udelay(1); } } -static void iceland_exit_rlc_safe_mode(struct amdgpu_device *adev) +static void gfx_v8_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id) { - u32 data = 0; + uint32_t data; unsigned i; data = RREG32(mmRLC_CNTL); - if (!(data & RLC_CNTL__RLC_ENABLE_F32_MASK)) - return; - - if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_MGCG)) { - if (adev->gfx.rlc.in_safe_mode) { - data |= RLC_SAFE_MODE__CMD_MASK; - data &= ~RLC_SAFE_MODE__MESSAGE_MASK; - WREG32(mmRLC_SAFE_MODE, data); - adev->gfx.rlc.in_safe_mode = false; - } - } + data |= RLC_SAFE_MODE__CMD_MASK; + data &= ~RLC_SAFE_MODE__MESSAGE_MASK; + WREG32(mmRLC_SAFE_MODE, data); for (i = 0; i < adev->usec_timeout; i++) { if (!REG_GET_FIELD(RREG32(mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD)) @@ -5771,9 +5541,41 @@ static void iceland_exit_rlc_safe_mode(struct amdgpu_device *adev) } } +static void gfx_v8_0_update_spm_vmid(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned vmid) +{ + u32 data; + + amdgpu_gfx_off_ctrl(adev, false); + + if (amdgpu_sriov_is_pp_one_vf(adev)) + data = RREG32_NO_KIQ(mmRLC_SPM_VMID); + else + data = RREG32(mmRLC_SPM_VMID); + + data &= ~RLC_SPM_VMID__RLC_SPM_VMID_MASK; + data |= (vmid & RLC_SPM_VMID__RLC_SPM_VMID_MASK) << RLC_SPM_VMID__RLC_SPM_VMID__SHIFT; + + if (amdgpu_sriov_is_pp_one_vf(adev)) + WREG32_NO_KIQ(mmRLC_SPM_VMID, data); + else + WREG32(mmRLC_SPM_VMID, data); + + amdgpu_gfx_off_ctrl(adev, true); +} + static const struct amdgpu_rlc_funcs iceland_rlc_funcs = { - .enter_safe_mode = iceland_enter_rlc_safe_mode, - .exit_safe_mode = iceland_exit_rlc_safe_mode + .is_rlc_enabled = gfx_v8_0_is_rlc_enabled, + .set_safe_mode = gfx_v8_0_set_safe_mode, + .unset_safe_mode = gfx_v8_0_unset_safe_mode, + .init = gfx_v8_0_rlc_init, + .get_csb_size = gfx_v8_0_get_csb_size, + .get_csb_buffer = gfx_v8_0_get_csb_buffer, + .get_cp_table_num = gfx_v8_0_cp_jump_table_num, + .resume = gfx_v8_0_rlc_resume, + .stop = gfx_v8_0_rlc_stop, + .reset = gfx_v8_0_rlc_reset, + .start = gfx_v8_0_rlc_start, + .update_spm_vmid = gfx_v8_0_update_spm_vmid }; static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev, @@ -5781,8 +5583,6 @@ static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev { uint32_t temp, data; - adev->gfx.rlc.funcs->enter_safe_mode(adev); - /* It is disabled by HW by default */ if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) { if (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGLS) { @@ -5876,8 +5676,6 @@ static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev /* 7- wait for RLC_SERDES_CU_MASTER & RLC_SERDES_NONCU_MASTER idle */ gfx_v8_0_wait_for_rlc_serdes(adev); } - - adev->gfx.rlc.funcs->exit_safe_mode(adev); } static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev, @@ -5887,8 +5685,6 @@ static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev temp = data = RREG32(mmRLC_CGCG_CGLS_CTRL); - adev->gfx.rlc.funcs->enter_safe_mode(adev); - if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)) { temp1 = data1 = RREG32(mmRLC_CGTT_MGCG_OVERRIDE); data1 &= ~RLC_CGTT_MGCG_OVERRIDE__CGCG_MASK; @@ -5950,7 +5746,7 @@ static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev /* wait for RLC_SERDES_CU_MASTER & RLC_SERDES_NONCU_MASTER idle */ gfx_v8_0_wait_for_rlc_serdes(adev); - /* write cmd to Set CGCG Overrride */ + /* write cmd to Set CGCG Override */ gfx_v8_0_send_serdes_cmd(adev, BPM_REG_CGCG_OVERRIDE, SET_BPM_SERDES_CMD); /* wait for RLC_SERDES_CU_MASTER & RLC_SERDES_NONCU_MASTER idle */ @@ -5969,12 +5765,12 @@ static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev } gfx_v8_0_wait_for_rlc_serdes(adev); - - adev->gfx.rlc.funcs->exit_safe_mode(adev); } static int gfx_v8_0_update_gfx_clock_gating(struct amdgpu_device *adev, bool enable) { + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); + if (enable) { /* CGCG/CGLS should be enabled after MGCG/MGLS/TS(CG/LS) * === MGCG + MGLS + TS(CG/LS) === @@ -5988,6 +5784,8 @@ static int gfx_v8_0_update_gfx_clock_gating(struct amdgpu_device *adev, gfx_v8_0_update_coarse_grain_clock_gating(adev, enable); gfx_v8_0_update_medium_grain_clock_gating(adev, enable); } + + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); return 0; } @@ -5996,7 +5794,6 @@ static int gfx_v8_0_tonga_update_gfx_clock_gating(struct amdgpu_device *adev, { uint32_t msg_id, pp_state = 0; uint32_t pp_support_state = 0; - void *pp_handle = adev->powerplay.pp_handle; if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_CGLS)) { if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS) { @@ -6014,7 +5811,7 @@ static int gfx_v8_0_tonga_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_CG, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_MGLS)) { @@ -6035,7 +5832,7 @@ static int gfx_v8_0_tonga_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_MG, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } return 0; @@ -6047,7 +5844,6 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev, uint32_t msg_id, pp_state = 0; uint32_t pp_support_state = 0; - void *pp_handle = adev->powerplay.pp_handle; if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_CGLS)) { if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS) { @@ -6065,7 +5861,7 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_CG, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_3D_CGCG | AMD_CG_SUPPORT_GFX_3D_CGLS)) { @@ -6084,7 +5880,7 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_3D, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_MGLS)) { @@ -6105,7 +5901,7 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_MG, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & AMD_CG_SUPPORT_GFX_RLC_LS) { @@ -6120,7 +5916,7 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_RLC, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CP_LS) { @@ -6134,16 +5930,16 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_CP, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } return 0; } -static int gfx_v8_0_set_clockgating_state(void *handle, +static int gfx_v8_0_set_clockgating_state(struct amdgpu_ip_block *ip_block, enum amd_clockgating_state state) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; if (amdgpu_sriov_vf(adev)) return 0; @@ -6161,6 +5957,7 @@ static int gfx_v8_0_set_clockgating_state(void *handle, case CHIP_POLARIS10: case CHIP_POLARIS11: case CHIP_POLARIS12: + case CHIP_VEGAM: gfx_v8_0_polaris_update_gfx_clock_gating(adev, state); break; default: @@ -6171,7 +5968,7 @@ static int gfx_v8_0_set_clockgating_state(void *handle, static u64 gfx_v8_0_ring_get_rptr(struct amdgpu_ring *ring) { - return ring->adev->wb.wb[ring->rptr_offs]; + return *ring->rptr_cpu_addr; } static u64 gfx_v8_0_ring_get_wptr_gfx(struct amdgpu_ring *ring) @@ -6180,7 +5977,7 @@ static u64 gfx_v8_0_ring_get_wptr_gfx(struct amdgpu_ring *ring) if (ring->use_doorbell) /* XXX check if swapping is necessary on BE */ - return ring->adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; else return RREG32(mmCP_RB0_WPTR); } @@ -6191,7 +5988,7 @@ static void gfx_v8_0_ring_set_wptr_gfx(struct amdgpu_ring *ring) if (ring->use_doorbell) { /* XXX check if swapping is necessary on BE */ - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); } else { WREG32(mmCP_RB0_WPTR, lower_32_bits(ring->wptr)); @@ -6243,23 +6040,12 @@ static void gfx_v8_0_ring_emit_vgt_flush(struct amdgpu_ring *ring) EVENT_INDEX(0)); } - -static void gfx_v8_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) -{ - amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); - amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | - WRITE_DATA_DST_SEL(0) | - WR_CONFIRM)); - amdgpu_ring_write(ring, mmHDP_DEBUG0); - amdgpu_ring_write(ring, 0); - amdgpu_ring_write(ring, 1); - -} - static void gfx_v8_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, - struct amdgpu_ib *ib, - unsigned vm_id, bool ctx_switch) + struct amdgpu_job *job, + struct amdgpu_ib *ib, + uint32_t flags) { + unsigned vmid = AMDGPU_JOB_GET_VMID(job); u32 header, control = 0; if (ib->flags & AMDGPU_IB_FLAG_CE) @@ -6267,12 +6053,12 @@ static void gfx_v8_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, else header = PACKET3(PACKET3_INDIRECT_BUFFER, 2); - control |= ib->length_dw | (vm_id << 24); + control |= ib->length_dw | (vmid << 24); 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_v8_0_ring_emit_de_meta(ring); } @@ -6287,10 +6073,28 @@ static void gfx_v8_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, } static void gfx_v8_0_ring_emit_ib_compute(struct amdgpu_ring *ring, + struct amdgpu_job *job, struct amdgpu_ib *ib, - unsigned vm_id, bool ctx_switch) -{ - u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vm_id << 24); + uint32_t flags) +{ + unsigned vmid = AMDGPU_JOB_GET_VMID(job); + u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vmid << 24); + + /* Currently, there is a high possibility to get wave ID mismatch + * between ME and GDS, leading to a hw deadlock, because ME generates + * different wave IDs than the GDS expects. This situation happens + * randomly when at least 5 compute pipes use GDS ordered append. + * The wave IDs generated by ME are also wrong after suspend/resume. + * Those are probably bugs somewhere else in the kernel driver. + * + * Writing GDS_COMPUTE_MAX_WAVE_ID resets wave ID counters in ME and + * GDS to 0 for this ring (me/pipe). + */ + if (ib->flags & AMDGPU_IB_FLAG_RESET_GDS_MAX_WAVE_ID) { + amdgpu_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)); + amdgpu_ring_write(ring, mmGDS_COMPUTE_MAX_WAVE_ID - PACKET3_SET_CONFIG_REG_START); + amdgpu_ring_write(ring, ring->adev->gds.gds_compute_max_wave_id); + } amdgpu_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2)); amdgpu_ring_write(ring, @@ -6307,8 +6111,11 @@ static void gfx_v8_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr, { bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT; bool int_sel = flags & AMDGPU_FENCE_FLAG_INT; + bool exec = flags & AMDGPU_FENCE_FLAG_EXEC; - /* EVENT_WRITE_EOP - flush caches, send int */ + /* Workaround for cache flush problems. First send a dummy EOP + * event down the pipe with seq one below. + */ amdgpu_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE_EOP, 4)); amdgpu_ring_write(ring, (EOP_TCL1_ACTION_EN | EOP_TC_ACTION_EN | @@ -6317,6 +6124,21 @@ static void gfx_v8_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr, EVENT_INDEX(5))); amdgpu_ring_write(ring, addr & 0xfffffffc); amdgpu_ring_write(ring, (upper_32_bits(addr) & 0xffff) | + DATA_SEL(1) | INT_SEL(0)); + amdgpu_ring_write(ring, lower_32_bits(seq - 1)); + amdgpu_ring_write(ring, upper_32_bits(seq - 1)); + + /* Then send the real EOP event down the pipe: + * EVENT_WRITE_EOP - flush caches, send int */ + amdgpu_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE_EOP, 4)); + amdgpu_ring_write(ring, (EOP_TCL1_ACTION_EN | + EOP_TC_ACTION_EN | + EOP_TC_WB_ACTION_EN | + EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) | + EVENT_INDEX(5) | + (exec ? EOP_EXEC : 0))); + amdgpu_ring_write(ring, addr & 0xfffffffc); + amdgpu_ring_write(ring, (upper_32_bits(addr) & 0xffff) | DATA_SEL(write64bit ? 2 : 1) | INT_SEL(int_sel ? 2 : 0)); amdgpu_ring_write(ring, lower_32_bits(seq)); amdgpu_ring_write(ring, upper_32_bits(seq)); @@ -6341,32 +6163,11 @@ static void gfx_v8_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring) } static void gfx_v8_0_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned vm_id, uint64_t pd_addr) + unsigned vmid, uint64_t pd_addr) { int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX); - amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); - amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(usepfp) | - WRITE_DATA_DST_SEL(0)) | - WR_CONFIRM); - if (vm_id < 8) { - amdgpu_ring_write(ring, - (mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR + vm_id)); - } else { - amdgpu_ring_write(ring, - (mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vm_id - 8)); - } - amdgpu_ring_write(ring, 0); - amdgpu_ring_write(ring, pd_addr >> 12); - - /* bits 0-15 are the VM contexts0-15 */ - /* invalidate the cache */ - amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); - amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | - WRITE_DATA_DST_SEL(0))); - amdgpu_ring_write(ring, mmVM_INVALIDATE_REQUEST); - amdgpu_ring_write(ring, 0); - amdgpu_ring_write(ring, 1 << vm_id); + amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); /* wait for the invalidate to complete */ amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5)); @@ -6389,7 +6190,7 @@ static void gfx_v8_0_ring_emit_vm_flush(struct amdgpu_ring *ring, static u64 gfx_v8_0_ring_get_wptr_compute(struct amdgpu_ring *ring) { - return ring->adev->wb.wb[ring->wptr_offs]; + return *ring->wptr_cpu_addr; } static void gfx_v8_0_ring_set_wptr_compute(struct amdgpu_ring *ring) @@ -6397,7 +6198,7 @@ static void gfx_v8_0_ring_set_wptr_compute(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; /* XXX check if swapping is necessary on BE */ - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr); + *ring->wptr_cpu_addr = lower_32_bits(ring->wptr); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); } @@ -6486,34 +6287,24 @@ static void gfx_v8_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags) amdgpu_ring_write(ring, 0); } -static unsigned gfx_v8_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring) +static unsigned gfx_v8_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring, + uint64_t addr) { unsigned ret; amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3)); - amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr)); - amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr)); - amdgpu_ring_write(ring, 0); /* discard following DWs if *cond_exec_gpu_addr==0 */ + amdgpu_ring_write(ring, lower_32_bits(addr)); + amdgpu_ring_write(ring, upper_32_bits(addr)); + /* discard following DWs if *cond_exec_gpu_addr==0 */ + amdgpu_ring_write(ring, 0); ret = ring->wptr & ring->buf_mask; - amdgpu_ring_write(ring, 0x55aa55aa); /* patch dummy value later */ + /* patch dummy value later */ + amdgpu_ring_write(ring, 0); return ret; } -static void gfx_v8_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigned offset) -{ - unsigned cur; - - BUG_ON(offset > ring->buf_mask); - BUG_ON(ring->ring[offset] != 0x55aa55aa); - - cur = (ring->wptr & ring->buf_mask) - 1; - if (likely(cur > offset)) - ring->ring[offset] = cur - offset; - else - ring->ring[offset] = (ring->ring_size >> 2) - offset + cur; -} - -static void gfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg, + uint32_t reg_val_offs) { struct amdgpu_device *adev = ring->adev; @@ -6524,21 +6315,47 @@ static void gfx_v8_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_v8_0_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg, uint32_t val) { + uint32_t cmd; + + switch (ring->funcs->type) { + case AMDGPU_RING_TYPE_GFX: + cmd = WRITE_DATA_ENGINE_SEL(1) | WR_CONFIRM; + break; + case AMDGPU_RING_TYPE_KIQ: + cmd = 1 << 16; /* no inc addr */ + break; + default: + cmd = WR_CONFIRM; + break; + } + amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); - amdgpu_ring_write(ring, (1 << 16)); /* no inc addr */ + amdgpu_ring_write(ring, cmd); amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, val); } +static void gfx_v8_0_ring_soft_recovery(struct amdgpu_ring *ring, unsigned vmid) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t value = 0; + + value = REG_SET_FIELD(value, SQ_CMD, CMD, 0x03); + value = REG_SET_FIELD(value, SQ_CMD, MODE, 0x01); + value = REG_SET_FIELD(value, SQ_CMD, CHECK_VMID, 1); + value = REG_SET_FIELD(value, SQ_CMD, VM_ID, vmid); + WREG32(mmSQ_CMD, value); +} + static void gfx_v8_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev, enum amdgpu_interrupt_state state) { @@ -6625,7 +6442,7 @@ static int gfx_v8_0_set_eop_interrupt_state(struct amdgpu_device *adev, enum amdgpu_interrupt_state state) { switch (type) { - case AMDGPU_CP_IRQ_GFX_EOP: + case AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP: gfx_v8_0_set_gfx_eop_interrupt_state(adev, state); break; case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP: @@ -6658,6 +6475,77 @@ static int gfx_v8_0_set_eop_interrupt_state(struct amdgpu_device *adev, return 0; } +static int gfx_v8_0_set_cp_ecc_int_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned int type, + enum amdgpu_interrupt_state state) +{ + int enable_flag; + + switch (state) { + case AMDGPU_IRQ_STATE_DISABLE: + enable_flag = 0; + break; + + case AMDGPU_IRQ_STATE_ENABLE: + enable_flag = 1; + break; + + default: + return -EINVAL; + } + + WREG32_FIELD(CP_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, enable_flag); + WREG32_FIELD(CP_INT_CNTL_RING0, CP_ECC_ERROR_INT_ENABLE, enable_flag); + WREG32_FIELD(CP_INT_CNTL_RING1, CP_ECC_ERROR_INT_ENABLE, enable_flag); + WREG32_FIELD(CP_INT_CNTL_RING2, CP_ECC_ERROR_INT_ENABLE, enable_flag); + WREG32_FIELD(CPC_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, enable_flag); + WREG32_FIELD(CP_ME1_PIPE0_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, + enable_flag); + WREG32_FIELD(CP_ME1_PIPE1_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, + enable_flag); + WREG32_FIELD(CP_ME1_PIPE2_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, + enable_flag); + WREG32_FIELD(CP_ME1_PIPE3_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, + enable_flag); + WREG32_FIELD(CP_ME2_PIPE0_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, + enable_flag); + WREG32_FIELD(CP_ME2_PIPE1_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, + enable_flag); + WREG32_FIELD(CP_ME2_PIPE2_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, + enable_flag); + WREG32_FIELD(CP_ME2_PIPE3_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, + enable_flag); + + return 0; +} + +static int gfx_v8_0_set_sq_int_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned int type, + enum amdgpu_interrupt_state state) +{ + int enable_flag; + + switch (state) { + case AMDGPU_IRQ_STATE_DISABLE: + enable_flag = 1; + break; + + case AMDGPU_IRQ_STATE_ENABLE: + enable_flag = 0; + break; + + default: + return -EINVAL; + } + + WREG32_FIELD(SQ_INTERRUPT_MSG_CTRL, STALL, + enable_flag); + + return 0; +} + static int gfx_v8_0_eop_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) @@ -6690,12 +6578,39 @@ static int gfx_v8_0_eop_irq(struct amdgpu_device *adev, return 0; } +static void gfx_v8_0_fault(struct amdgpu_device *adev, + struct amdgpu_iv_entry *entry) +{ + u8 me_id, pipe_id, queue_id; + struct amdgpu_ring *ring; + int i; + + me_id = (entry->ring_id & 0x0c) >> 2; + pipe_id = (entry->ring_id & 0x03) >> 0; + queue_id = (entry->ring_id & 0x70) >> 4; + + switch (me_id) { + case 0: + drm_sched_fault(&adev->gfx.gfx_ring[0].sched); + break; + case 1: + case 2: + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + ring = &adev->gfx.compute_ring[i]; + if (ring->me == me_id && ring->pipe == pipe_id && + ring->queue == queue_id) + drm_sched_fault(&ring->sched); + } + break; + } +} + static int gfx_v8_0_priv_reg_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { DRM_ERROR("Illegal register access in command stream\n"); - schedule_work(&adev->reset_work); + gfx_v8_0_fault(adev, entry); return 0; } @@ -6704,54 +6619,205 @@ static int gfx_v8_0_priv_inst_irq(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry) { DRM_ERROR("Illegal instruction in command stream\n"); - schedule_work(&adev->reset_work); + gfx_v8_0_fault(adev, entry); return 0; } -static int gfx_v8_0_kiq_set_interrupt_state(struct amdgpu_device *adev, - struct amdgpu_irq_src *src, - unsigned int type, - enum amdgpu_interrupt_state state) +static int gfx_v8_0_cp_ecc_error_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) { - struct amdgpu_ring *ring = &(adev->gfx.kiq.ring); + DRM_ERROR("CP EDC/ECC error detected."); + return 0; +} - switch (type) { - case AMDGPU_CP_KIQ_IRQ_DRIVER0: - WREG32_FIELD(CPC_INT_CNTL, GENERIC2_INT_ENABLE, - state == AMDGPU_IRQ_STATE_DISABLE ? 0 : 1); - if (ring->me == 1) - WREG32_FIELD_OFFSET(CP_ME1_PIPE0_INT_CNTL, - ring->pipe, - GENERIC2_INT_ENABLE, - state == AMDGPU_IRQ_STATE_DISABLE ? 0 : 1); - else - WREG32_FIELD_OFFSET(CP_ME2_PIPE0_INT_CNTL, - ring->pipe, - GENERIC2_INT_ENABLE, - state == AMDGPU_IRQ_STATE_DISABLE ? 0 : 1); +static void gfx_v8_0_parse_sq_irq(struct amdgpu_device *adev, unsigned ih_data, + bool from_wq) +{ + u32 enc, se_id, sh_id, cu_id; + char type[20]; + int sq_edc_source = -1; + + enc = REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_CMN, ENCODING); + se_id = REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_CMN, SE_ID); + + switch (enc) { + case 0: + DRM_INFO("SQ general purpose intr detected:" + "se_id %d, immed_overflow %d, host_reg_overflow %d," + "host_cmd_overflow %d, cmd_timestamp %d," + "reg_timestamp %d, thread_trace_buff_full %d," + "wlt %d, thread_trace %d.\n", + se_id, + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_AUTO, IMMED_OVERFLOW), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_AUTO, HOST_REG_OVERFLOW), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_AUTO, HOST_CMD_OVERFLOW), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_AUTO, CMD_TIMESTAMP), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_AUTO, REG_TIMESTAMP), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_AUTO, THREAD_TRACE_BUF_FULL), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_AUTO, WLT), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_AUTO, THREAD_TRACE) + ); + break; + case 1: + case 2: + + cu_id = REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, CU_ID); + sh_id = REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, SH_ID); + + /* + * This function can be called either directly from ISR + * or from BH in which case we can access SQ_EDC_INFO + * instance + */ + if (from_wq) { + mutex_lock(&adev->grbm_idx_mutex); + gfx_v8_0_select_se_sh(adev, se_id, sh_id, cu_id, 0); + + sq_edc_source = REG_GET_FIELD(RREG32(mmSQ_EDC_INFO), SQ_EDC_INFO, SOURCE); + + gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0); + mutex_unlock(&adev->grbm_idx_mutex); + } + + if (enc == 1) + sprintf(type, "instruction intr"); + else + sprintf(type, "EDC/ECC error"); + + DRM_INFO( + "SQ %s detected: " + "se_id %d, sh_id %d, cu_id %d, simd_id %d, wave_id %d, vm_id %d " + "trap %s, sq_ed_info.source %s.\n", + type, se_id, sh_id, cu_id, + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, SIMD_ID), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, WAVE_ID), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, VM_ID), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, PRIV) ? "true" : "false", + (sq_edc_source != -1) ? sq_edc_source_names[sq_edc_source] : "unavailable" + ); + break; + default: + DRM_ERROR("SQ invalid encoding type\n."); + } +} + +static void gfx_v8_0_sq_irq_work_func(struct work_struct *work) +{ + + struct amdgpu_device *adev = container_of(work, struct amdgpu_device, gfx.sq_work.work); + struct sq_work *sq_work = container_of(work, struct sq_work, work); + + gfx_v8_0_parse_sq_irq(adev, sq_work->ih_data, true); +} + +static int gfx_v8_0_sq_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + unsigned ih_data = entry->src_data[0]; + + /* + * Try to submit work so SQ_EDC_INFO can be accessed from + * BH. If previous work submission hasn't finished yet + * just print whatever info is possible directly from the ISR. + */ + if (work_pending(&adev->gfx.sq_work.work)) { + gfx_v8_0_parse_sq_irq(adev, ih_data, false); + } else { + adev->gfx.sq_work.ih_data = ih_data; + schedule_work(&adev->gfx.sq_work.work); + } + + return 0; +} + +static void gfx_v8_0_emit_mem_sync(struct amdgpu_ring *ring) +{ + amdgpu_ring_write(ring, PACKET3(PACKET3_SURFACE_SYNC, 3)); + amdgpu_ring_write(ring, PACKET3_TCL1_ACTION_ENA | + PACKET3_TC_ACTION_ENA | + PACKET3_SH_KCACHE_ACTION_ENA | + PACKET3_SH_ICACHE_ACTION_ENA | + PACKET3_TC_WB_ACTION_ENA); /* CP_COHER_CNTL */ + amdgpu_ring_write(ring, 0xffffffff); /* CP_COHER_SIZE */ + amdgpu_ring_write(ring, 0); /* CP_COHER_BASE */ + amdgpu_ring_write(ring, 0x0000000A); /* poll interval */ +} + +static void gfx_v8_0_emit_mem_sync_compute(struct amdgpu_ring *ring) +{ + amdgpu_ring_write(ring, PACKET3(PACKET3_ACQUIRE_MEM, 5)); + amdgpu_ring_write(ring, PACKET3_TCL1_ACTION_ENA | + PACKET3_TC_ACTION_ENA | + PACKET3_SH_KCACHE_ACTION_ENA | + PACKET3_SH_ICACHE_ACTION_ENA | + PACKET3_TC_WB_ACTION_ENA); /* CP_COHER_CNTL */ + amdgpu_ring_write(ring, 0xffffffff); /* CP_COHER_SIZE */ + amdgpu_ring_write(ring, 0xff); /* 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 */ +} + + +/* mmSPI_WCL_PIPE_PERCENT_CS[0-7]_DEFAULT values are same */ +#define mmSPI_WCL_PIPE_PERCENT_CS_DEFAULT 0x0000007f +static void gfx_v8_0_emit_wave_limit_cs(struct amdgpu_ring *ring, + uint32_t pipe, bool enable) +{ + uint32_t val; + uint32_t wcl_cs_reg; + + val = enable ? 0x1 : mmSPI_WCL_PIPE_PERCENT_CS_DEFAULT; + + switch (pipe) { + case 0: + wcl_cs_reg = mmSPI_WCL_PIPE_PERCENT_CS0; break; - default: - BUG(); /* kiq only support GENERIC2_INT now */ + case 1: + wcl_cs_reg = mmSPI_WCL_PIPE_PERCENT_CS1; + break; + case 2: + wcl_cs_reg = mmSPI_WCL_PIPE_PERCENT_CS2; + break; + case 3: + wcl_cs_reg = mmSPI_WCL_PIPE_PERCENT_CS3; break; + default: + DRM_DEBUG("invalid pipe %d\n", pipe); + return; } - return 0; + + amdgpu_ring_emit_wreg(ring, wcl_cs_reg, val); + } -static int gfx_v8_0_kiq_irq(struct amdgpu_device *adev, - struct amdgpu_irq_src *source, - struct amdgpu_iv_entry *entry) +#define mmSPI_WCL_PIPE_PERCENT_GFX_DEFAULT 0x07ffffff +static void gfx_v8_0_emit_wave_limit(struct amdgpu_ring *ring, bool enable) { - u8 me_id, pipe_id, queue_id; - struct amdgpu_ring *ring = &(adev->gfx.kiq.ring); + struct amdgpu_device *adev = ring->adev; + uint32_t val; + int i; - me_id = (entry->ring_id & 0x0c) >> 2; - pipe_id = (entry->ring_id & 0x03) >> 0; - queue_id = (entry->ring_id & 0x70) >> 4; - DRM_DEBUG("IH: CPC GENERIC2_INT, me:%d, pipe:%d, queue:%d\n", - me_id, pipe_id, queue_id); + /* 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, 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_v8_0_emit_wave_limit_cs(ring, i, enable); + + } - amdgpu_fence_process(ring); - return 0; } static const struct amd_ip_funcs gfx_v8_0_ip_funcs = { @@ -6786,8 +6852,8 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = { .emit_frame_size = /* maximum 215dw if count 16 IBs in */ 5 + /* COND_EXEC */ 7 + /* PIPELINE_SYNC */ - 19 + /* VM_FLUSH */ - 8 + /* FENCE for VM_FLUSH */ + VI_FLUSH_GPU_TLB_NUM_WREG * 5 + 9 + /* VM_FLUSH */ + 12 + /* FENCE for VM_FLUSH */ 20 + /* GDS switch */ 4 + /* double SWITCH_BUFFER, the first COND_EXEC jump to the place just @@ -6799,8 +6865,9 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = { 31 + /* DE_META */ 3 + /* CNTX_CTRL */ 5 + /* HDP_INVL */ - 8 + 8 + /* FENCE x2 */ - 2, /* SWITCH_BUFFER */ + 12 + 12 + /* FENCE x2 */ + 2 + /* SWITCH_BUFFER */ + 5, /* SURFACE_SYNC */ .emit_ib_size = 4, /* gfx_v8_0_ring_emit_ib_gfx */ .emit_ib = gfx_v8_0_ring_emit_ib_gfx, .emit_fence = gfx_v8_0_ring_emit_fence_gfx, @@ -6808,7 +6875,6 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = { .emit_vm_flush = gfx_v8_0_ring_emit_vm_flush, .emit_gds_switch = gfx_v8_0_ring_emit_gds_switch, .emit_hdp_flush = gfx_v8_0_ring_emit_hdp_flush, - .emit_hdp_invalidate = gfx_v8_0_ring_emit_hdp_invalidate, .test_ring = gfx_v8_0_ring_test_ring, .test_ib = gfx_v8_0_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, @@ -6816,7 +6882,9 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = { .emit_switch_buffer = gfx_v8_ring_emit_sb, .emit_cntxcntl = gfx_v8_ring_emit_cntxcntl, .init_cond_exec = gfx_v8_0_ring_emit_init_cond_exec, - .patch_cond_exec = gfx_v8_0_ring_emit_patch_cond_exec, + .emit_wreg = gfx_v8_0_ring_emit_wreg, + .soft_recovery = gfx_v8_0_ring_soft_recovery, + .emit_mem_sync = gfx_v8_0_emit_mem_sync, }; static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = { @@ -6830,22 +6898,28 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = { .emit_frame_size = 20 + /* gfx_v8_0_ring_emit_gds_switch */ 7 + /* gfx_v8_0_ring_emit_hdp_flush */ - 5 + /* gfx_v8_0_ring_emit_hdp_invalidate */ + 5 + /* hdp_invalidate */ 7 + /* gfx_v8_0_ring_emit_pipeline_sync */ - 17 + /* gfx_v8_0_ring_emit_vm_flush */ - 7 + 7 + 7, /* gfx_v8_0_ring_emit_fence_compute x3 for user fence, vm fence */ - .emit_ib_size = 4, /* gfx_v8_0_ring_emit_ib_compute */ + VI_FLUSH_GPU_TLB_NUM_WREG * 5 + 7 + /* gfx_v8_0_ring_emit_vm_flush */ + 7 + 7 + 7 + /* gfx_v8_0_ring_emit_fence_compute x3 for user fence, vm fence */ + 7 + /* gfx_v8_0_emit_mem_sync_compute */ + 5 + /* gfx_v8_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_v8_0_ring_emit_ib_compute */ .emit_ib = gfx_v8_0_ring_emit_ib_compute, .emit_fence = gfx_v8_0_ring_emit_fence_compute, .emit_pipeline_sync = gfx_v8_0_ring_emit_pipeline_sync, .emit_vm_flush = gfx_v8_0_ring_emit_vm_flush, .emit_gds_switch = gfx_v8_0_ring_emit_gds_switch, .emit_hdp_flush = gfx_v8_0_ring_emit_hdp_flush, - .emit_hdp_invalidate = gfx_v8_0_ring_emit_hdp_invalidate, .test_ring = gfx_v8_0_ring_test_ring, .test_ib = gfx_v8_0_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, + .emit_wreg = gfx_v8_0_ring_emit_wreg, + .soft_recovery = gfx_v8_0_ring_soft_recovery, + .emit_mem_sync = gfx_v8_0_emit_mem_sync_compute, + .emit_wave_limit = gfx_v8_0_emit_wave_limit, }; static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_kiq = { @@ -6859,26 +6933,25 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_kiq = { .emit_frame_size = 20 + /* gfx_v8_0_ring_emit_gds_switch */ 7 + /* gfx_v8_0_ring_emit_hdp_flush */ - 5 + /* gfx_v8_0_ring_emit_hdp_invalidate */ + 5 + /* hdp_invalidate */ 7 + /* gfx_v8_0_ring_emit_pipeline_sync */ 17 + /* gfx_v8_0_ring_emit_vm_flush */ 7 + 7 + 7, /* gfx_v8_0_ring_emit_fence_kiq x3 for user fence, vm fence */ - .emit_ib_size = 4, /* gfx_v8_0_ring_emit_ib_compute */ - .emit_ib = gfx_v8_0_ring_emit_ib_compute, + .emit_ib_size = 7, /* gfx_v8_0_ring_emit_ib_compute */ .emit_fence = gfx_v8_0_ring_emit_fence_kiq, .test_ring = gfx_v8_0_ring_test_ring, - .test_ib = gfx_v8_0_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, .emit_rreg = gfx_v8_0_ring_emit_rreg, .emit_wreg = gfx_v8_0_ring_emit_wreg, + .emit_hdp_flush = gfx_v8_0_ring_emit_hdp_flush, }; static void gfx_v8_0_set_ring_funcs(struct amdgpu_device *adev) { int i; - adev->gfx.kiq.ring.funcs = &gfx_v8_0_ring_funcs_kiq; + adev->gfx.kiq[0].ring.funcs = &gfx_v8_0_ring_funcs_kiq; for (i = 0; i < adev->gfx.num_gfx_rings; i++) adev->gfx.gfx_ring[i].funcs = &gfx_v8_0_ring_funcs_gfx; @@ -6902,9 +6975,14 @@ static const struct amdgpu_irq_src_funcs gfx_v8_0_priv_inst_irq_funcs = { .process = gfx_v8_0_priv_inst_irq, }; -static const struct amdgpu_irq_src_funcs gfx_v8_0_kiq_irq_funcs = { - .set = gfx_v8_0_kiq_set_interrupt_state, - .process = gfx_v8_0_kiq_irq, +static const struct amdgpu_irq_src_funcs gfx_v8_0_cp_ecc_error_irq_funcs = { + .set = gfx_v8_0_set_cp_ecc_int_state, + .process = gfx_v8_0_cp_ecc_error_irq, +}; + +static const struct amdgpu_irq_src_funcs gfx_v8_0_sq_irq_funcs = { + .set = gfx_v8_0_set_sq_int_state, + .process = gfx_v8_0_sq_irq, }; static void gfx_v8_0_set_irq_funcs(struct amdgpu_device *adev) @@ -6918,8 +6996,11 @@ static void gfx_v8_0_set_irq_funcs(struct amdgpu_device *adev) adev->gfx.priv_inst_irq.num_types = 1; adev->gfx.priv_inst_irq.funcs = &gfx_v8_0_priv_inst_irq_funcs; - adev->gfx.kiq.irq.num_types = AMDGPU_CP_KIQ_IRQ_LAST; - adev->gfx.kiq.irq.funcs = &gfx_v8_0_kiq_irq_funcs; + adev->gfx.cp_ecc_error_irq.num_types = 1; + adev->gfx.cp_ecc_error_irq.funcs = &gfx_v8_0_cp_ecc_error_irq_funcs; + + adev->gfx.sq_irq.num_types = 1; + adev->gfx.sq_irq.funcs = &gfx_v8_0_sq_irq_funcs; } static void gfx_v8_0_set_rlc_funcs(struct amdgpu_device *adev) @@ -6930,29 +7011,10 @@ static void gfx_v8_0_set_rlc_funcs(struct amdgpu_device *adev) static void gfx_v8_0_set_gds_init(struct amdgpu_device *adev) { /* init asci gds info */ - adev->gds.mem.total_size = RREG32(mmGDS_VMID0_SIZE); - adev->gds.gws.total_size = 64; - adev->gds.oa.total_size = 16; - - if (adev->gds.mem.total_size == 64 * 1024) { - adev->gds.mem.gfx_partition_size = 4096; - adev->gds.mem.cs_partition_size = 4096; - - adev->gds.gws.gfx_partition_size = 4; - adev->gds.gws.cs_partition_size = 4; - - adev->gds.oa.gfx_partition_size = 4; - adev->gds.oa.cs_partition_size = 1; - } else { - adev->gds.mem.gfx_partition_size = 1024; - adev->gds.mem.cs_partition_size = 1024; - - adev->gds.gws.gfx_partition_size = 16; - adev->gds.gws.cs_partition_size = 16; - - adev->gds.oa.gfx_partition_size = 4; - adev->gds.oa.cs_partition_size = 4; - } + adev->gds.gds_size = RREG32(mmGDS_VMID0_SIZE); + adev->gds.gws_size = 64; + adev->gds.oa_size = 16; + adev->gds.gds_compute_max_wave_id = RREG32(mmGDS_COMPUTE_MAX_WAVE_ID); } static void gfx_v8_0_set_user_cu_inactive_bitmap(struct amdgpu_device *adev, @@ -7004,12 +7066,12 @@ static void gfx_v8_0_get_cu_info(struct amdgpu_device *adev) mask = 1; ao_bitmap = 0; counter = 0; - gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff); + gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff, 0); if (i < 4 && j < 2) gfx_v8_0_set_user_cu_inactive_bitmap( adev, disable_masks[i * 2 + j]); bitmap = gfx_v8_0_get_cu_active_bitmap(adev); - cu_info->bitmap[i][j] = bitmap; + cu_info->bitmap[0][i][j] = bitmap; for (k = 0; k < adev->gfx.config.max_cu_per_sh; k ++) { if (bitmap & mask) { @@ -7025,11 +7087,16 @@ static void gfx_v8_0_get_cu_info(struct amdgpu_device *adev) cu_info->ao_cu_bitmap[i][j] = ao_bitmap; } } - gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0); mutex_unlock(&adev->grbm_idx_mutex); cu_info->number = active_cu_number; cu_info->ao_cu_mask = ao_cu_mask; + cu_info->simd_per_cu = NUM_SIMD_PER_CU; + cu_info->max_waves_per_simd = 10; + cu_info->max_scratch_slots_per_cu = 32; + cu_info->wave_front_size = 64; + cu_info->lds_size = 64; } const struct amdgpu_ip_block_version gfx_v8_0_ip_block = @@ -7054,18 +7121,18 @@ static void gfx_v8_0_ring_emit_ce_meta(struct amdgpu_ring *ring) { uint64_t ce_payload_addr; int cnt_ce; - static union { + union { struct vi_ce_ib_state regular; struct vi_ce_ib_state_chained_ib chained; } ce_payload = {}; if (ring->adev->virt.chained_ib_support) { - ce_payload_addr = AMDGPU_VA_RESERVED_SIZE - 2 * 4096 + - offsetof(struct vi_gfx_meta_data_chained_ib, ce_payload); + ce_payload_addr = amdgpu_csa_vaddr(ring->adev) + + offsetof(struct vi_gfx_meta_data_chained_ib, ce_payload); cnt_ce = (sizeof(ce_payload.chained) >> 2) + 4 - 2; } else { - ce_payload_addr = AMDGPU_VA_RESERVED_SIZE - 2 * 4096 + - offsetof(struct vi_gfx_meta_data, ce_payload); + ce_payload_addr = amdgpu_csa_vaddr(ring->adev) + + offsetof(struct vi_gfx_meta_data, ce_payload); cnt_ce = (sizeof(ce_payload.regular) >> 2) + 4 - 2; } @@ -7083,12 +7150,12 @@ static void gfx_v8_0_ring_emit_de_meta(struct amdgpu_ring *ring) { uint64_t de_payload_addr, gds_addr, csa_addr; int cnt_de; - static union { + union { struct vi_de_ib_state regular; struct vi_de_ib_state_chained_ib chained; } de_payload = {}; - csa_addr = AMDGPU_VA_RESERVED_SIZE - 2 * 4096; + csa_addr = amdgpu_csa_vaddr(ring->adev); gds_addr = csa_addr + 4096; if (ring->adev->virt.chained_ib_support) { de_payload.chained.gds_backup_addrlo = lower_32_bits(gds_addr); |