UMSCH 4.0
Subsystems (2)
Alex Deucher
maintainercover UMSCH 4.0
Add UMSCH 4.0 support. UMSCH handles scheduling
of multimedia queues similar to MES for GFX and
SDMA.
Lang Yu (14):
drm/amdgpu: add UMSCH 4.0 register headers drm/amdgpu: add UMSCH IP BLOCK TYPE definition drm/amdgpu: add UMSCH RING TYPE definition drm/amdgpu: add UMSCH firmware header definition drm/amdgpu: add UMSCH 4.0 api definition drm/amdgpu: add initial support for UMSCH drm/amdgpu: enable UMSCH scheduling for VPE drm/amdgpu: add selftest framework for UMSCH drm/amdgpu: add VPE queue submission test drm/amdgpu: reserve mmhub engine 3 for UMSCH FW drm/amdgpu: add PSP loading support for UMSCH drm/amdgpu/discovery: enable UMSCH 4.0 in IP discovery drm/amdgpu: add amdgpu_umsch_mm module parameter drm/amdgpu: update SET_HW_RESOURCES definition for UMSCH
drivers/gpu/drm/amd/amdgpu/Makefile | 5 + drivers/gpu/drm/amd/amdgpu/amdgpu.h | 6 + drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 20 + drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 10 + drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 6 +- drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 9 + drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 3 +- drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c | 21 + drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h | 20 + drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c | 875 +++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.h | 218 +++++ drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c | 408 ++++++++ drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.h | 30 + drivers/gpu/drm/amd/include/amd_shared.h | 2 + .../include/asic_reg/vcn/vcn_4_0_0_offset.h | 422 +++++++++ .../include/asic_reg/vcn/vcn_4_0_0_sh_mask.h | 882 ++++++++++++++++++ .../drm/amd/include/umsch_mm_4_0_api_def.h | 437 +++++++++ 17 files changed, 3372 insertions(+), 2 deletions(-) create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.h create mode 100644 drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c create mode 100644 drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.h create mode 100644 drivers/gpu/drm/amd/include/umsch_mm_4_0_api_def.h
Signature · 1 line
2.41.0
Alex Deucher
maintainerPATCH 1/14 drm/amdgpu: add UMSCH 4.0 register headers
From: Lang Yu <[email protected]>
Add headers for UMSCH 4.0.
v2: updates (Alex)
--- .../include/asic_reg/vcn/vcn_4_0_0_offset.h | 422 +++++++++ .../include/asic_reg/vcn/vcn_4_0_0_sh_mask.h | 882 ++++++++++++++++++ 2 files changed, 1304 insertions(+)
drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_4_0_0_offset.h +422 -0
@@ -1305,6 +1305,32 @@#define regVCN_RB3_DB_CTRL_BASE_IDX 1#define regVCN_RB4_DB_CTRL 0x0075#define regVCN_RB4_DB_CTRL_BASE_IDX 1 +#define regVCN_UMSCH_RB_DB_CTRL 0x0076 +#define regVCN_UMSCH_RB_DB_CTRL_BASE_IDX 1 +#define regVCN_AGDB_CTRL0 0x0079 +#define regVCN_AGDB_CTRL0_BASE_IDX 1 +#define regVCN_AGDB_CTRL1 0x007a +#define regVCN_AGDB_CTRL1_BASE_IDX 1 +#define regVCN_AGDB_CTRL2 0x007b +#define regVCN_AGDB_CTRL2_BASE_IDX 1 +#define regVCN_AGDB_CTRL3 0x007c +#define regVCN_AGDB_CTRL3_BASE_IDX 1 +#define regVCN_AGDB_CTRL4 0x007d +#define regVCN_AGDB_CTRL4_BASE_IDX 1 +#define regVCN_AGDB_CTRL5 0x007e +#define regVCN_AGDB_CTRL5_BASE_IDX 1 +#define regVCN_AGDB_MASK0 0x007f +#define regVCN_AGDB_MASK0_BASE_IDX 1 +#define regVCN_AGDB_MASK1 0x0080 +#define regVCN_AGDB_MASK1_BASE_IDX 1 +#define regVCN_AGDB_MASK2 0x0081 +#define regVCN_AGDB_MASK2_BASE_IDX 1 +#define regVCN_AGDB_MASK3 0x0082 +#define regVCN_AGDB_MASK3_BASE_IDX 1 +#define regVCN_AGDB_MASK4 0x0083 +#define regVCN_AGDB_MASK4_BASE_IDX 1 +#define regVCN_AGDB_MASK5 0x0084 +#define regVCN_AGDB_MASK5_BASE_IDX 1#define regVCN_RB_ENABLE 0x0085#define regVCN_RB_ENABLE_BASE_IDX 1#define regVCN_RB_WPTR_CTRL 0x0086
@@ -1556,6 +1582,402 @@#define regVCN_RAS_CNTL_MMSCH 0x0914#define regVCN_RAS_CNTL_MMSCH_BASE_IDX 1 +#define regVCN_UMSCH_MES_UTCL1_CNTL 0x0759 +#define regVCN_UMSCH_MES_UTCL1_CNTL_BASE_IDX 1 +#define regVCN_UMSCH_MES_BUSY 0x075a +#define regVCN_UMSCH_MES_BUSY_BASE_IDX 1 +#define regVCN_UMSCH_RB_BASE_LO 0x075b +#define regVCN_UMSCH_RB_BASE_LO_BASE_IDX 1 +#define regVCN_UMSCH_RB_BASE_HI 0x075c +#define regVCN_UMSCH_RB_BASE_HI_BASE_IDX 1 +#define regVCN_UMSCH_RB_SIZE 0x075d +#define regVCN_UMSCH_RB_SIZE_BASE_IDX 1 +#define regVCN_UMSCH_RB_RPTR 0x075e +#define regVCN_UMSCH_RB_RPTR_BASE_IDX 1 +#define regVCN_UMSCH_RB_WPTR 0x075f +#define regVCN_UMSCH_RB_WPTR_BASE_IDX 1 +#define regVCN_UMSCH_MASTINT_EN 0x0760 +#define regVCN_UMSCH_MASTINT_EN_BASE_IDX 1 +#define regVCN_UMSCH_IH_CTRL 0x0761 +#define regVCN_UMSCH_IH_CTRL_BASE_IDX 1 +#define regVCN_UMSCH_SYS_INT_EN 0x0762 +#define regVCN_UMSCH_SYS_INT_EN_BASE_IDX 1 +#define regVCN_UMSCH_SYS_INT_STATUS 0x0763 +#define regVCN_UMSCH_SYS_INT_STATUS_BASE_IDX 1 +#define regVCN_UMSCH_SYS_INT_ACK 0x0764 +#define regVCN_UMSCH_SYS_INT_ACK_BASE_IDX 1 +#define regVCN_UMSCH_SYS_INT_SRC 0x0765 +#define regVCN_UMSCH_SYS_INT_SRC_BASE_IDX 1 +#define regVCN_UMSCH_IH_CTX_CTRL 0x0766 +#define regVCN_UMSCH_IH_CTX_CTRL_BASE_IDX 1 +#define regVCN_UMSCH_CGC_CTRL 0x0767 +#define regVCN_UMSCH_CGC_CTRL_BASE_IDX 1 +#define regVCN_UMSCH_CGC_STATUS 0x0768 +#define regVCN_UMSCH_CGC_STATUS_BASE_IDX 1 +#define regVCN_UMSCH_CGC_MEM_CTRL 0x0769 +#define regVCN_UMSCH_CGC_MEM_CTRL_BASE_IDX 1 +#define regUVD_INTERNAL_REG_VIOLATION_8 0x076a +#define regUVD_INTERNAL_REG_VIOLATION_8_BASE_IDX 1 +#define regUVD_UMSCH_FORCE 0x076b +#define regUVD_UMSCH_FORCE_BASE_IDX 1 +#define regUVD_UMSCH_DEBUG_INDEX 0x076c +#define regUVD_UMSCH_DEBUG_INDEX_BASE_IDX 1 +#define regUVD_UMSCH_DEBUG_DATA_LO 0x076d +#define regUVD_UMSCH_DEBUG_DATA_LO_BASE_IDX 1 +#define regUVD_UMSCH_DEBUG_DATA_HI 0x076e +#define regUVD_UMSCH_DEBUG_DATA_HI_BASE_IDX 1 +#define regUVD_UMSCH_DEBUG_UTCL2_TCIU_IF 0x076f +#define regUVD_UMSCH_DEBUG_UTCL2_TCIU_IF_BASE_IDX 1 +#define regUMSCH_MES_RESET_CTRL 0x0770 +#define regUMSCH_MES_RESET_CTRL_BASE_IDX 1 + +#define regVCN_MES_PRGRM_CNTR_START 0x0780 +#define regVCN_MES_PRGRM_CNTR_START_BASE_IDX 1 +#define regVCN_MES_INTR_ROUTINE_START 0x0781 +#define regVCN_MES_INTR_ROUTINE_START_BASE_IDX 1 +#define regVCN_MES_MTVEC_LO 0x0781 +#define regVCN_MES_MTVEC_LO_BASE_IDX 1 +#define regVCN_MES_INTR_ROUTINE_START_HI 0x0782 +#define regVCN_MES_INTR_ROUTINE_START_HI_BASE_IDX 1 +#define regVCN_MES_MTVEC_HI 0x0782 +#define regVCN_MES_MTVEC_HI_BASE_IDX 1 +#define regVCN_MES_CNTL 0x0787 +#define regVCN_MES_CNTL_BASE_IDX 1 +#define regVCN_MES_PIPE_PRIORITY_CNTS 0x0788 +#define regVCN_MES_PIPE_PRIORITY_CNTS_BASE_IDX 1 +#define regVCN_MES_PIPE0_PRIORITY 0x0789 +#define regVCN_MES_PIPE0_PRIORITY_BASE_IDX 1 +#define regVCN_MES_PIPE1_PRIORITY 0x078a +#define regVCN_MES_PIPE1_PRIORITY_BASE_IDX 1 +#define regVCN_MES_PIPE2_PRIORITY 0x078b +#define regVCN_MES_PIPE2_PRIORITY_BASE_IDX 1 +#define regVCN_MES_PIPE3_PRIORITY 0x078c +#define regVCN_MES_PIPE3_PRIORITY_BASE_IDX 1 +#define regVCN_MES_HEADER_DUMP 0x078d +#define regVCN_MES_HEADER_DUMP_BASE_IDX 1 +#define regVCN_MES_MIE_LO 0x078e +#define regVCN_MES_MIE_LO_BASE_IDX 1 +#define regVCN_MES_MIE_HI 0x078f +#define regVCN_MES_MIE_HI_BASE_IDX 1 +#define regVCN_MES_INTERRUPT 0x0790 +#define regVCN_MES_INTERRUPT_BASE_IDX 1 +#define regVCN_MES_SCRATCH_INDEX 0x0791 +#define regVCN_MES_SCRATCH_INDEX_BASE_IDX 1 +#define regVCN_MES_SCRATCH_DATA 0x0792 +#define regVCN_MES_SCRATCH_DATA_BASE_IDX 1 +#define regVCN_MES_INSTR_PNTR 0x0793 +#define regVCN_MES_INSTR_PNTR_BASE_IDX 1 +#define regVCN_MES_MSCRATCH_HI 0x0794 +#define regVCN_MES_MSCRATCH_HI_BASE_IDX 1 +#define regVCN_MES_MSCRATCH_LO 0x0795 +#define regVCN_MES_MSCRATCH_LO_BASE_IDX 1 +#define regVCN_MES_MSTATUS_LO 0x0796 +#define regVCN_MES_MSTATUS_LO_BASE_IDX 1 +#define regVCN_MES_MSTATUS_HI 0x0797 +#define regVCN_MES_MSTATUS_HI_BASE_IDX 1 +#define regVCN_MES_MEPC_LO 0x0798 +#define regVCN_MES_MEPC_LO_BASE_IDX 1 +#define regVCN_MES_MEPC_HI 0x0799 +#define regVCN_MES_MEPC_HI_BASE_IDX 1 +#define regVCN_MES_MCAUSE_LO 0x079a +#define regVCN_MES_MCAUSE_LO_BASE_IDX 1 +#define regVCN_MES_MCAUSE_HI 0x079b +#define regVCN_MES_MCAUSE_HI_BASE_IDX 1 +#define regVCN_MES_MBADADDR_LO 0x079c +#define regVCN_MES_MBADADDR_LO_BASE_IDX 1 +#define regVCN_MES_MBADADDR_HI 0x079d +#define regVCN_MES_MBADADDR_HI_BASE_IDX 1 +#define regVCN_MES_MIP_LO 0x079e +#define regVCN_MES_MIP_LO_BASE_IDX 1 +#define regVCN_MES_MIP_HI 0x079f +#define regVCN_MES_MIP_HI_BASE_IDX 1 +#define regVCN_MES_IC_OP_CNTL 0x07a0 +#define regVCN_MES_IC_OP_CNTL_BASE_IDX 1 +#define regVCN_MES_MCYCLE_LO 0x07a6 +#define regVCN_MES_MCYCLE_LO_BASE_IDX 1 +#define regVCN_MES_MCYCLE_HI 0x07a7 +#define regVCN_MES_MCYCLE_HI_BASE_IDX 1 +#define regVCN_MES_MTIME_LO 0x07a8 +#define regVCN_MES_MTIME_LO_BASE_IDX 1 +#define regVCN_MES_MTIME_HI 0x07a9 +#define regVCN_MES_MTIME_HI_BASE_IDX 1 +#define regVCN_MES_MINSTRET_LO 0x07aa +#define regVCN_MES_MINSTRET_LO_BASE_IDX 1 +#define regVCN_MES_MINSTRET_HI 0x07ab +#define regVCN_MES_MINSTRET_HI_BASE_IDX 1 +#define regVCN_MES_MISA_LO 0x07ac +#define regVCN_MES_MISA_LO_BASE_IDX 1 +#define regVCN_MES_MISA_HI 0x07ad +#define regVCN_MES_MISA_HI_BASE_IDX 1 +#define regVCN_MES_MVENDORID_LO 0x07ae +#define regVCN_MES_MVENDORID_LO_BASE_IDX 1 +#define regVCN_MES_MVENDORID_HI 0x07af +#define regVCN_MES_MVENDORID_HI_BASE_IDX 1 +#define regVCN_MES_MARCHID_LO 0x07b0 +#define regVCN_MES_MARCHID_LO_BASE_IDX 1 +#define regVCN_MES_MARCHID_HI 0x07b1 +#define regVCN_MES_MARCHID_HI_BASE_IDX 1 +#define regVCN_MES_MIMPID_LO 0x07b2 +#define regVCN_MES_MIMPID_LO_BASE_IDX 1 +#define regVCN_MES_MIMPID_HI 0x07b3 +#define regVCN_MES_MIMPID_HI_BASE_IDX 1 +#define regVCN_MES_MHARTID_LO 0x07b4 +#define regVCN_MES_MHARTID_LO_BASE_IDX 1 +#define regVCN_MES_MHARTID_HI 0x07b5 +#define regVCN_MES_MHARTID_HI_BASE_IDX 1 +#define regVCN_MES_DC_BASE_CNTL 0x07b6 +#define regVCN_MES_DC_BASE_CNTL_BASE_IDX 1 +#define regVCN_MES_DC_OP_CNTL 0x07b7 +#define regVCN_MES_DC_OP_CNTL_BASE_IDX 1 +#define regVCN_MES_MTIMECMP_LO 0x07b8 +#define regVCN_MES_MTIMECMP_LO_BASE_IDX 1 +#define regVCN_MES_MTIMECMP_HI 0x07b9 +#define regVCN_MES_MTIMECMP_HI_BASE_IDX 1 +#define regVCN_MES_DEBUG_INTERRUPT_INSTR_PNTR 0x07c2 +#define regVCN_MES_DEBUG_INTERRUPT_INSTR_PNTR_BASE_IDX 1 +#define regVCN_MES_GP0_LO 0x07c3 +#define regVCN_MES_GP0_LO_BASE_IDX 1 +#define regVCN_MES_GP0_HI 0x07c4 +#define regVCN_MES_GP0_HI_BASE_IDX 1 +#define regVCN_MES_GP1_LO 0x07c5 +#define regVCN_MES_GP1_LO_BASE_IDX 1 +#define regVCN_MES_GP1_HI 0x07c6 +#define regVCN_MES_GP1_HI_BASE_IDX 1 +#define regVCN_MES_GP2_LO 0x07c7 +#define regVCN_MES_GP2_LO_BASE_IDX 1 +#define regVCN_MES_GP2_HI 0x07c8 +#define regVCN_MES_GP2_HI_BASE_IDX 1 +#define regVCN_MES_GP3_LO 0x07c9 +#define regVCN_MES_GP3_LO_BASE_IDX 1 +#define regVCN_MES_GP3_HI 0x07ca +#define regVCN_MES_GP3_HI_BASE_IDX 1 +#define regVCN_MES_GP4_LO 0x07cb +#define regVCN_MES_GP4_LO_BASE_IDX 1 +#define regVCN_MES_GP4_HI 0x07cc +#define regVCN_MES_GP4_HI_BASE_IDX 1 +#define regVCN_MES_GP5_LO 0x07cd +#define regVCN_MES_GP5_LO_BASE_IDX 1 +#define regVCN_MES_GP5_HI 0x07ce +#define regVCN_MES_GP5_HI_BASE_IDX 1 +#define regVCN_MES_GP6_LO 0x07cf +#define regVCN_MES_GP6_LO_BASE_IDX 1 +#define regVCN_MES_GP6_HI 0x07d0 +#define regVCN_MES_GP6_HI_BASE_IDX 1 +#define regVCN_MES_GP7_LO 0x07d1 +#define regVCN_MES_GP7_LO_BASE_IDX 1 +#define regVCN_MES_GP7_HI 0x07d2 +#define regVCN_MES_GP7_HI_BASE_IDX 1 +#define regVCN_MES_GP8_LO 0x07d3 +#define regVCN_MES_GP8_LO_BASE_IDX 1 +#define regVCN_MES_GP8_HI 0x07d4 +#define regVCN_MES_GP8_HI_BASE_IDX 1 +#define regVCN_MES_GP9_LO 0x07d5 +#define regVCN_MES_GP9_LO_BASE_IDX 1 +#define regVCN_MES_GP9_HI 0x07d6 +#define regVCN_MES_GP9_HI_BASE_IDX 1 +#define regVCN_MES_DM_INDEX_ADDR 0x0800 +#define regVCN_MES_DM_INDEX_ADDR_BASE_IDX 1 +#define regVCN_MES_DM_INDEX_DATA 0x0801 +#define regVCN_MES_DM_INDEX_DATA_BASE_IDX 1 +#define regVCN_MES_DBG_FROM_RST 0x0802 +#define regVCN_MES_DBG_FROM_RST_BASE_IDX 1 +#define regVCN_MES_LOCAL_BASE0_LO 0x0803 +#define regVCN_MES_LOCAL_BASE0_LO_BASE_IDX 1 +#define regVCN_MES_LOCAL_BASE0_HI 0x0804 +#define regVCN_MES_LOCAL_BASE0_HI_BASE_IDX 1 +#define regVCN_MES_LOCAL_MASK0_LO 0x0805 +#define regVCN_MES_LOCAL_MASK0_LO_BASE_IDX 1 +#define regVCN_MES_LOCAL_MASK0_HI 0x0806 +#define regVCN_MES_LOCAL_MASK0_HI_BASE_IDX 1 +#define regVCN_MES_LOCAL_APERTURE 0x0807 +#define regVCN_MES_LOCAL_APERTURE_BASE_IDX 1 +#define regVCN_MES_LOCAL_INSTR_BASE_LO 0x0808 +#define regVCN_MES_LOCAL_INSTR_BASE_LO_BASE_IDX 1 +#define regVCN_MES_LOCAL_INSTR_BASE_HI 0x0809 +#define regVCN_MES_LOCAL_INSTR_BASE_HI_BASE_IDX 1 +#define regVCN_MES_LOCAL_INSTR_MASK_LO 0x080a +#define regVCN_MES_LOCAL_INSTR_MASK_LO_BASE_IDX 1 +#define regVCN_MES_LOCAL_INSTR_MASK_HI 0x080b +#define regVCN_MES_LOCAL_INSTR_MASK_HI_BASE_IDX 1 +#define regVCN_MES_LOCAL_INSTR_APERTURE 0x080c +#define regVCN_MES_LOCAL_INSTR_APERTURE_BASE_IDX 1 +#define regVCN_MES_LOCAL_SCRATCH_APERTURE 0x080d +#define regVCN_MES_LOCAL_SCRATCH_APERTURE_BASE_IDX 1 +#define regVCN_MES_LOCAL_SCRATCH_BASE_LO 0x080e +#define regVCN_MES_LOCAL_SCRATCH_BASE_LO_BASE_IDX 1 +#define regVCN_MES_LOCAL_SCRATCH_BASE_HI 0x080f +#define regVCN_MES_LOCAL_SCRATCH_BASE_HI_BASE_IDX 1 +#define regVCN_MES_PERFCOUNT_CNTL 0x0819 +#define regVCN_MES_PERFCOUNT_CNTL_BASE_IDX 1 +#define regVCN_MES_PENDING_INTERRUPT 0x081a +#define regVCN_MES_PENDING_INTERRUPT_BASE_IDX 1 +#define regVCN_MES_PRIV_LEVEL 0x081b +#define regVCN_MES_PRIV_LEVEL_BASE_IDX 1 +#define regVCN_MES_PRIV_LEVEL_VIOLATION_STATUS 0x081c +#define regVCN_MES_PRIV_LEVEL_VIOLATION_STATUS_BASE_IDX 1 +#define regVCN_MES_PRGRM_CNTR_START_HI 0x081d +#define regVCN_MES_PRGRM_CNTR_START_HI_BASE_IDX 1 +#define regVCN_MES_DEBUG_INTERRUPT_INSTR_PNTR_HI 0x081e +#define regVCN_MES_DEBUG_INTERRUPT_INSTR_PNTR_HI_BASE_IDX 1 +#define regVCN_MES_INTERRUPT_DATA_16 0x081f +#define regVCN_MES_INTERRUPT_DATA_16_BASE_IDX 1 +#define regVCN_MES_INTERRUPT_DATA_17 0x0820 +#define regVCN_MES_INTERRUPT_DATA_17_BASE_IDX 1 +#define regVCN_MES_INTERRUPT_DATA_18 0x0821 +#define regVCN_MES_INTERRUPT_DATA_18_BASE_IDX 1 +#define regVCN_MES_INTERRUPT_DATA_19 0x0822 +#define regVCN_MES_INTERRUPT_DATA_19_BASE_IDX 1 +#define regVCN_MES_INTERRUPT_DATA_20 0x0823 +#define regVCN_MES_INTERRUPT_DATA_20_BASE_IDX 1 +#define regVCN_MES_INTERRUPT_DATA_21 0x0824 +#define regVCN_MES_INTERRUPT_DATA_21_BASE_IDX 1 +#define regVCN_MES_INTERRUPT_DATA_22 0x0825 +#define regVCN_MES_INTERRUPT_DATA_22_BASE_IDX 1 +#define regVCN_MES_INTERRUPT_DATA_23 0x0826 +#define regVCN_MES_INTERRUPT_DATA_23_BASE_IDX 1 +#define regVCN_MES_INTERRUPT_DATA_24 0x0827 +#define regVCN_MES_INTERRUPT_DATA_24_BASE_IDX 1 +#define regVCN_MES_INTERRUPT_DATA_25 0x0828 +#define regVCN_MES_INTERRUPT_DATA_25_BASE_IDX 1 +#define regVCN_MES_INTERRUPT_DATA_26 0x0829 +#define regVCN_MES_INTERRUPT_DATA_26_BASE_IDX 1 +#define regVCN_MES_INTERRUPT_DATA_27 0x082a +#define regVCN_MES_INTERRUPT_DATA_27_BASE_IDX 1 +#define regVCN_MES_INTERRUPT_DATA_28 0x082b +#define regVCN_MES_INTERRUPT_DATA_28_BASE_IDX 1 +#define regVCN_MES_INTERRUPT_DATA_29 0x082c +#define regVCN_MES_INTERRUPT_DATA_29_BASE_IDX 1 +#define regVCN_MES_INTERRUPT_DATA_30 0x082d +#define regVCN_MES_INTERRUPT_DATA_30_BASE_IDX 1 +#define regVCN_MES_INTERRUPT_DATA_31 0x082e +#define regVCN_MES_INTERRUPT_DATA_31_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE0_BASE 0x082f +#define regVCN_MES_DC_APERTURE0_BASE_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE0_MASK 0x0830 +#define regVCN_MES_DC_APERTURE0_MASK_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE0_CNTL 0x0831 +#define regVCN_MES_DC_APERTURE0_CNTL_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE1_BASE 0x0832 +#define regVCN_MES_DC_APERTURE1_BASE_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE1_MASK 0x0833 +#define regVCN_MES_DC_APERTURE1_MASK_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE1_CNTL 0x0834 +#define regVCN_MES_DC_APERTURE1_CNTL_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE2_BASE 0x0835 +#define regVCN_MES_DC_APERTURE2_BASE_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE2_MASK 0x0836 +#define regVCN_MES_DC_APERTURE2_MASK_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE2_CNTL 0x0837 +#define regVCN_MES_DC_APERTURE2_CNTL_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE3_BASE 0x0838 +#define regVCN_MES_DC_APERTURE3_BASE_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE3_MASK 0x0839 +#define regVCN_MES_DC_APERTURE3_MASK_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE3_CNTL 0x083a +#define regVCN_MES_DC_APERTURE3_CNTL_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE4_BASE 0x083b +#define regVCN_MES_DC_APERTURE4_BASE_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE4_MASK 0x083c +#define regVCN_MES_DC_APERTURE4_MASK_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE4_CNTL 0x083d +#define regVCN_MES_DC_APERTURE4_CNTL_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE5_BASE 0x083e +#define regVCN_MES_DC_APERTURE5_BASE_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE5_MASK 0x083f +#define regVCN_MES_DC_APERTURE5_MASK_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE5_CNTL 0x0840 +#define regVCN_MES_DC_APERTURE5_CNTL_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE6_BASE 0x0841 +#define regVCN_MES_DC_APERTURE6_BASE_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE6_MASK 0x0842 +#define regVCN_MES_DC_APERTURE6_MASK_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE6_CNTL 0x0843 +#define regVCN_MES_DC_APERTURE6_CNTL_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE7_BASE 0x0844 +#define regVCN_MES_DC_APERTURE7_BASE_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE7_MASK 0x0845 +#define regVCN_MES_DC_APERTURE7_MASK_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE7_CNTL 0x0846 +#define regVCN_MES_DC_APERTURE7_CNTL_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE8_BASE 0x0847 +#define regVCN_MES_DC_APERTURE8_BASE_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE8_MASK 0x0848 +#define regVCN_MES_DC_APERTURE8_MASK_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE8_CNTL 0x0849 +#define regVCN_MES_DC_APERTURE8_CNTL_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE9_BASE 0x084a +#define regVCN_MES_DC_APERTURE9_BASE_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE9_MASK 0x084b +#define regVCN_MES_DC_APERTURE9_MASK_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE9_CNTL 0x084c +#define regVCN_MES_DC_APERTURE9_CNTL_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE10_BASE 0x084d +#define regVCN_MES_DC_APERTURE10_BASE_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE10_MASK 0x084e +#define regVCN_MES_DC_APERTURE10_MASK_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE10_CNTL 0x084f +#define regVCN_MES_DC_APERTURE10_CNTL_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE11_BASE 0x0850 +#define regVCN_MES_DC_APERTURE11_BASE_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE11_MASK 0x0851 +#define regVCN_MES_DC_APERTURE11_MASK_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE11_CNTL 0x0852 +#define regVCN_MES_DC_APERTURE11_CNTL_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE12_BASE 0x0853 +#define regVCN_MES_DC_APERTURE12_BASE_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE12_MASK 0x0854 +#define regVCN_MES_DC_APERTURE12_MASK_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE12_CNTL 0x0855 +#define regVCN_MES_DC_APERTURE12_CNTL_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE13_BASE 0x0856 +#define regVCN_MES_DC_APERTURE13_BASE_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE13_MASK 0x0857 +#define regVCN_MES_DC_APERTURE13_MASK_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE13_CNTL 0x0858 +#define regVCN_MES_DC_APERTURE13_CNTL_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE14_BASE 0x0859 +#define regVCN_MES_DC_APERTURE14_BASE_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE14_MASK 0x085a +#define regVCN_MES_DC_APERTURE14_MASK_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE14_CNTL 0x085b +#define regVCN_MES_DC_APERTURE14_CNTL_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE15_BASE 0x085c +#define regVCN_MES_DC_APERTURE15_BASE_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE15_MASK 0x085d +#define regVCN_MES_DC_APERTURE15_MASK_BASE_IDX 1 +#define regVCN_MES_DC_APERTURE15_CNTL 0x085e +#define regVCN_MES_DC_APERTURE15_CNTL_BASE_IDX 1 + +#define regVCN_HYP_ME1_PIPE0_VMID_CNTL 0x0890 +#define regVCN_HYP_ME1_PIPE0_VMID_CNTL_BASE_IDX 1 +#define regVCN_HYP_ME1_PIPE1_VMID_CNTL 0x0891 +#define regVCN_HYP_ME1_PIPE1_VMID_CNTL_BASE_IDX 1 +#define regVCN_MES_IC_BASE_LO 0x08d0 +#define regVCN_MES_IC_BASE_LO_BASE_IDX 1 +#define regVCN_MES_MIBASE_LO 0x08d0 +#define regVCN_MES_MIBASE_LO_BASE_IDX 1 +#define regVCN_MES_IC_BASE_HI 0x08d1 +#define regVCN_MES_IC_BASE_HI_BASE_IDX 1 +#define regVCN_MES_MIBASE_HI 0x08d1 +#define regVCN_MES_MIBASE_HI_BASE_IDX 1 +#define regVCN_MES_IC_BASE_CNTL 0x08d2 +#define regVCN_MES_IC_BASE_CNTL_BASE_IDX 1 +#define regVCN_MES_DC_BASE_LO 0x08d4 +#define regVCN_MES_DC_BASE_LO_BASE_IDX 1 +#define regVCN_MES_MDBASE_LO 0x08d4 +#define regVCN_MES_MDBASE_LO_BASE_IDX 1 +#define regVCN_MES_DC_BASE_HI 0x08d5 +#define regVCN_MES_DC_BASE_HI_BASE_IDX 1 +#define regVCN_MES_MDBASE_HI 0x08d5 +#define regVCN_MES_MDBASE_HI_BASE_IDX 1 +#define regVCN_MES_MIBOUND_LO 0x08db +#define regVCN_MES_MIBOUND_LO_BASE_IDX 1 +#define regVCN_MES_MIBOUND_HI 0x08dc +#define regVCN_MES_MIBOUND_HI_BASE_IDX 1 +#define regVCN_MES_MDBOUND_LO 0x08dd +#define regVCN_MES_MDBOUND_LO_BASE_IDX 1 +#define regVCN_MES_MDBOUND_HI 0x08de +#define regVCN_MES_MDBOUND_HI_BASE_IDX 1// addressBlock: uvdctxind // base address: 0x0
drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_4_0_0_sh_mask.h +882 -0
893 lines not on this page · raw message
Signature · 1 line
2.41.0
Alex Deucher
maintainerPATCH 2/14 drm/amdgpu: add UMSCH IP BLOCK TYPE definition
In reply to Alex DeucherFrom: Lang Yu <[email protected]>
Add IP BLOCK TYPE definition for Multimedia User Mode Scheduler which is a hardware scheduler for VCN and VPE workload.
--- drivers/gpu/drm/amd/include/amd_shared.h | 2 ++ 1 file changed, 2 insertions(+)
drivers/gpu/drm/amd/include/amd_shared.h +2 -0
@@ -83,6 +83,7 @@ enum amd_apu_flags {* @AMD_IP_BLOCK_TYPE_VCN: Video Core/Codec Next * @AMD_IP_BLOCK_TYPE_MES: Micro-Engine Scheduler * @AMD_IP_BLOCK_TYPE_JPEG: JPEG Engine +* @AMD_IP_BLOCK_TYPE_UMSCH_MM: User Mode Schduler for Multimedia * @AMD_IP_BLOCK_TYPE_NUM: Total number of IP block types */ enum amd_ip_block_type {
@@ -101,6 +102,7 @@ enum amd_ip_block_type {AMD_IP_BLOCK_TYPE_MES ,AMD_IP_BLOCK_TYPE_JPEG ,AMD_IP_BLOCK_TYPE_VPE , +AMD_IP_BLOCK_TYPE_UMSCH_MM , AMD_IP_BLOCK_TYPE_NUM, };
Signature · 1 line
2.41.0
Alex Deucher
maintainerPATCH 3/14 drm/amdgpu: add UMSCH RING TYPE definition
In reply to Alex DeucherFrom: Lang Yu <[email protected]>
Add RING TYPE definition for Multi Mdeia User Mode Scheduler.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2 -1
@@ -80,7 +80,8 @@ enum amdgpu_ring_type {AMDGPU_RING_TYPE_VCN_JPEG =AMDGPU_HW_IP_VCN_JPEG ,AMDGPU_RING_TYPE_VPE =AMDGPU_HW_IP_VPE ,AMDGPU_RING_TYPE_KIQ , -AMDGPU_RING_TYPE_MES +AMDGPU_RING_TYPE_MES , + AMDGPU_RING_TYPE_UMSCH_MM, }; enum amdgpu_ib_pool_type {
Signature · 1 line
2.41.0
Alex Deucher
maintainerPATCH 4/14 drm/amdgpu: add UMSCH firmware header definition
In reply to Alex DeucherFrom: Lang Yu <[email protected]>
Add firmware header definition for UMSCH.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h | 17 +++++++++++++++++ 1 file changed, 17 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h +17 -0
@@ -328,6 +328,23 @@ struct vpe_firmware_header_v1_0 {uint32_t ctl_jt_size;/* control thread size of jt */ }; +/* version_major=1, version_minor=0 */ +struct umsch_mm_firmware_header_v1_0 { +struct common_firmware_header header; +uint32_t umsch_mm_ucode_version; +uint32_t umsch_mm_ucode_size_bytes; +uint32_t umsch_mm_ucode_offset_bytes; +uint32_t umsch_mm_ucode_data_version; +uint32_t umsch_mm_ucode_data_size_bytes; +uint32_t umsch_mm_ucode_data_offset_bytes; +uint32_t umsch_mm_irq_start_addr_lo; +uint32_t umsch_mm_irq_start_addr_hi; +uint32_t umsch_mm_uc_start_addr_lo; +uint32_t umsch_mm_uc_start_addr_hi; +uint32_t umsch_mm_data_start_addr_lo; +uint32_t umsch_mm_data_start_addr_hi; +}; +/* gpu info payload */ struct gpu_info_firmware_v1_0 {uint32_t gc_num_se;
Signature · 1 line
2.41.0
Alex Deucher
maintainerPATCH 5/14 drm/amdgpu: add UMSCH 4.0 api definition
In reply to Alex DeucherFrom: Lang Yu <[email protected]>
Add api definition for UMSCH 4.0.
v2: adjust coding style.
--- .../drm/amd/include/umsch_mm_4_0_api_def.h | 434 ++++++++++++++++++ 1 file changed, 434 insertions(+) create mode 100644 drivers/gpu/drm/amd/include/umsch_mm_4_0_api_def.h
drivers/gpu/drm/amd/include/umsch_mm_4_0_api_def.h new file +434 -0
@@ -0,0 +1,434 @@ +/* SPDX-License-Identifier: MIT */ +/* +* Copyright 2023 Advanced Micro Devices, Inc. +* +* Permission is hereby granted, free of charge, to any person obtaining a +* copy of this software and associated documentation files (the "Software"), +* to deal in the Software without restriction, including without limitation +* the rights to use, copy, modify, merge, publish, distribute, sublicense, +* and/or sell copies of the Software, and to permit persons to whom the +* Software is furnished to do so, subject to the following conditions: +* +* The above copyright notice and this permission notice shall be included in +* all copies or substantial portions of the Software. +* +* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL +* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR +* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, +* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR +* OTHER DEALINGS IN THE SOFTWARE. +* +*/ + +#ifndef __UMSCH_MM_API_DEF_H__ +#define __UMSCH_MM_API_DEF_H__ + +#pragma once + +#pragma pack(push, 4) + +#define UMSCH_API_VERSION 1 + +/* +* Driver submits one API(cmd) as a single Frame and this command size is same for all API +* to ease the debugging and parsing of ring buffer. +*/ +enum {API_FRAME_SIZE_IN_DWORDS =64 }; + +/* +* To avoid command in scheduler context to be overwritten whenever multiple interrupts come in, +* this creates another queue. +*/ +enum {API_NUMBER_OF_COMMAND_MAX =32 }; + +enum {UMSCH_INSTANCE_DB_OFFSET_MAX =16 }; + +enum UMSCH_API_TYPE { +UMSCH_API_TYPE_SCHEDULER =1 , +UMSCH_API_TYPE_MAX +}; + +enum UMSCH_MS_LOG_CONTEXT_STATE { +UMSCH_LOG_CONTEXT_STATE_IDLE =0 , +UMSCH_LOG_CONTEXT_STATE_RUNNING =1 , +UMSCH_LOG_CONTEXT_STATE_READY =2 , +UMSCH_LOG_CONTEXT_STATE_READY_STANDBY =3 , +UMSCH_LOG_CONTEXT_STATE_INVALID =0xF , +}; + +enum UMSCH_MS_LOG_OPERATION { +UMSCH_LOG_OPERATION_CONTEXT_STATE_CHANGE =0 , +UMSCH_LOG_OPERATION_QUEUE_NEW_WORK =1 , +UMSCH_LOG_OPERATION_QUEUE_UNWAIT_SYNC_OBJECT =2 , +UMSCH_LOG_OPERATION_QUEUE_NO_MORE_WORK =3 , +UMSCH_LOG_OPERATION_QUEUE_WAIT_SYNC_OBJECT =4 , +UMSCH_LOG_OPERATION_QUEUE_INVALID =0xF , +}; + +struct UMSCH_INSTANCE_DB_OFFSET { +uint32_t instance_index; +uint32_t doorbell_offset; +}; + +struct UMSCH_LOG_CONTEXT_STATE_CHANGE { +uint64_t h_context; +enum UMSCH_MS_LOG_CONTEXT_STATE new_context_state; +}; + +struct UMSCH_LOG_QUEUE_NEW_WORK { +uint64_t h_queue; +uint64_t reserved; +}; + +struct UMSCH_LOG_QUEUE_UNWAIT_SYNC_OBJECT { +uint64_t h_queue; +uint64_t h_sync_object; +}; + +struct UMSCH_LOG_QUEUE_NO_MORE_WORK { +uint64_t h_queue; +uint64_t reserved; +}; + +struct UMSCH_LOG_QUEUE_WAIT_SYNC_OBJECT { +uint64_t h_queue; +uint64_t h_sync_object; +}; + +struct UMSCH_LOG_ENTRY_HEADER { +uint32_t first_free_entry_index; +uint32_t wraparound_count; +uint64_t number_of_entries; +uint64_t reserved[2 ]; +}; + +struct UMSCH_LOG_ENTRY_DATA { +uint64_t gpu_time_stamp; +uint32_t operation_type;/* operation_type is of UMSCH_LOG_OPERATION type */ +uint32_t reserved_operation_type_bits; +union { +struct UMSCH_LOG_CONTEXT_STATE_CHANGE context_state_change; +struct UMSCH_LOG_QUEUE_NEW_WORK queue_new_work; +struct UMSCH_LOG_QUEUE_UNWAIT_SYNC_OBJECT queue_unwait_sync_object; +struct UMSCH_LOG_QUEUE_NO_MORE_WORK queue_no_more_work; +struct UMSCH_LOG_QUEUE_WAIT_SYNC_OBJECT queue_wait_sync_object; +uint64_t all[2 ]; + }; +}; + +struct UMSCH_LOG_BUFFER { +struct UMSCH_LOG_ENTRY_HEADER header; +struct UMSCH_LOG_ENTRY_DATA entries[1 ]; +}; + +enum UMSCH_API_OPCODE { +UMSCH_API_SET_HW_RSRC =0x00 , +UMSCH_API_SET_SCHEDULING_CONFIG =0x1 , +UMSCH_API_ADD_QUEUE =0x2 , +UMSCH_API_REMOVE_QUEUE =0x3 , +UMSCH_API_PERFORM_YIELD =0x4 , +UMSCH_API_SUSPEND =0x5 , +UMSCH_API_RESUME =0x6 , +UMSCH_API_RESET =0x7 , +UMSCH_API_SET_LOG_BUFFER =0x8 , +UMSCH_API_CHANGE_CONTEXT_PRIORITY =0x9 , +UMSCH_API_QUERY_SCHEDULER_STATUS =0xA , +UMSCH_API_UPDATE_AFFINITY =0xB , +UMSCH_API_MAX =0xFF +}; + +union UMSCH_API_HEADER { +struct { +uint32_t type :4 ;/* 0 - Invalid; 1 - Scheduling; 2 - TBD */ +uint32_t opcode :8 ; +uint32_t dwsize :8 ; +uint32_t reserved :12 ; + }; + +uint32_t u32All; +}; + +enum UMSCH_AMD_PRIORITY_LEVEL { +AMD_PRIORITY_LEVEL_IDLE =0 , +AMD_PRIORITY_LEVEL_NORMAL =1 , +AMD_PRIORITY_LEVEL_FOCUS =2 , +AMD_PRIORITY_LEVEL_REALTIME =3 , +AMD_PRIORITY_NUM_LEVELS +}; + +enum UMSCH_ENGINE_TYPE { +UMSCH_ENGINE_TYPE_VCN0 =0 , +UMSCH_ENGINE_TYPE_VCN1 =1 , +UMSCH_ENGINE_TYPE_VCN =2 , +UMSCH_ENGINE_TYPE_VPE =3 , +UMSCH_ENGINE_TYPE_MAX +}; + +#define AFFINITY_DISABLE 0 +#define AFFINITY_ENABLE 1 +#define AFFINITY_MAX 2 + +union UMSCH_AFFINITY { +struct { +unsigned int vcn0Affinity :2 ;/* enable 1 disable 0 */ +unsigned int vcn1Affinity :2 ; +unsigned int reserved :28 ; + }; +unsigned int u32All; +}; + +struct UMSCH_API_STATUS { +uint64_t api_completion_fence_addr; +uint32_t api_completion_fence_value; +}; + +enum {MAX_VCN0_INSTANCES =1 }; +enum {MAX_VCN1_INSTANCES =1 }; +enum {MAX_VCN_INSTANCES =2 }; + +enum {MAX_VPE_INSTANCES =1 }; + +enum {MAX_VCN_QUEUES =4 }; +enum {MAX_VPE_QUEUES =8 }; + +enum {MAX_QUEUES_IN_A_CONTEXT =1 }; + +enum {UMSCH_MAX_HWIP_SEGMENT =8 }; + +enum VM_HUB_TYPE { +VM_HUB_TYPE_GC =0 , +VM_HUB_TYPE_MM =1 , +VM_HUB_TYPE_MAX , +}; + +enum {VMID_INVALID =0xffff }; + +enum {MAX_VMID_MMHUB =16 }; + +union UMSCHAPI__SET_HW_RESOURCES { +struct { +union UMSCH_API_HEADER header; +uint32_t vmid_mask_mm_vcn; +uint32_t vmid_mask_mm_vpe; +uint32_t engine_mask; +uint32_t logging_vmid; +uint32_t vcn0_hqd_mask[MAX_VCN0_INSTANCES ]; +uint32_t vcn1_hqd_mask[MAX_VCN1_INSTANCES ]; +uint32_t vcn_hqd_mask[MAX_VCN_INSTANCES ]; +uint32_t vpe_hqd_mask[MAX_VPE_INSTANCES ]; +uint64_t g_sch_ctx_gpu_mc_ptr; +uint32_t mmhub_base[UMSCH_MAX_HWIP_SEGMENT ]; +uint32_t mmhub_version; +uint32_t osssys_base[UMSCH_MAX_HWIP_SEGMENT ]; +uint32_t osssys_version; +struct UMSCH_API_STATUS api_status; +union { +struct { +uint32_t disable_reset :1 ; +uint32_t disable_umsch_log :1 ; +uint32_t enable_level_process_quantum_check :1 ; +uint32_t is_vcn0_enabled :1 ; +uint32_t is_vcn1_enabled :1 ; +uint32_t reserved :27 ; + }; +uint32_t uint32_all; + }; + }; + +uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS ]; +}; +static_assert (sizeof (union UMSCHAPI__SET_HW_RESOURCES ) <=API_FRAME_SIZE_IN_DWORDS *sizeof (uint32_t ), +"size of UMSCHAPI__SET_HW_RESOURCES must be less than 256 bytes" ); + +union UMSCHAPI__SET_SCHEDULING_CONFIG { +struct { +union UMSCH_API_HEADER header; +/* +* Grace period when preempting another priority band for this priority band. +* The value for idle priority band is ignored, as it never preempts other bands. +*/ +uint64_t grace_period_other_levels[AMD_PRIORITY_NUM_LEVELS ]; + +/* Default quantum for scheduling across processes within a priority band. */ +uint64_t process_quantum_for_level[AMD_PRIORITY_NUM_LEVELS ]; + +/* Default grace period for processes that preempt each other within a priority band. */ +uint64_t process_grace_period_same_level[AMD_PRIORITY_NUM_LEVELS ]; + +/* +* For normal level this field specifies the target GPU percentage in situations +* when it's starved by the high level. Valid values are between 0 and 50, +* with the default being 10. +*/ +uint32_t normal_yield_percent; + +struct UMSCH_API_STATUS api_status; + }; + +uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS ]; +}; + +union UMSCHAPI__ADD_QUEUE { +struct { +union UMSCH_API_HEADER header; +uint32_t process_id; +uint64_t page_table_base_addr; +uint64_t process_va_start; +uint64_t process_va_end; +uint64_t process_quantum; +uint64_t process_csa_addr; +uint64_t context_quantum; +uint64_t context_csa_addr; +uint32_t inprocess_context_priority; +enum UMSCH_AMD_PRIORITY_LEVEL context_global_priority_level; +uint32_t doorbell_offset_0; +uint32_t doorbell_offset_1; +union UMSCH_AFFINITY affinity; +uint64_t mqd_addr; +uint64_t h_context; +uint64_t h_queue; +enum UMSCH_ENGINE_TYPE engine_type; +uint32_t vm_context_cntl; + +struct { +uint32_t is_context_suspended :1 ; +uint32_t reserved :31 ; + }; +struct UMSCH_API_STATUS api_status; + }; + +uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS ]; +}; + + +union UMSCHAPI__REMOVE_QUEUE { +struct { +union UMSCH_API_HEADER header; +uint32_t doorbell_offset_0; +uint32_t doorbell_offset_1; +uint64_t context_csa_addr; + +struct UMSCH_API_STATUS api_status; + }; + +uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS ]; +}; + +union UMSCHAPI__PERFORM_YIELD { +struct { +union UMSCH_API_HEADER header; +uint32_t dummy; +struct UMSCH_API_STATUS api_status; + }; + +uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS ]; +}; + +union UMSCHAPI__SUSPEND { +struct { +union UMSCH_API_HEADER header; +uint64_t context_csa_addr; +uint64_t suspend_fence_addr; +uint32_t suspend_fence_value; + +struct UMSCH_API_STATUS api_status; + }; + +uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS ]; +}; + +enum UMSCH_RESUME_OPTION { +CONTEXT_RESUME =0 , +ENGINE_SCHEDULE_RESUME =1 , +}; + +union UMSCHAPI__RESUME { +struct { +union UMSCH_API_HEADER header; + +enum UMSCH_RESUME_OPTION resume_option; +uint64_t context_csa_addr;/* valid only for UMSCH_SWIP_CONTEXT_RESUME */ +enum UMSCH_ENGINE_TYPE engine_type; + +struct UMSCH_API_STATUS api_status; + }; + +uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS ]; +}; + +enum UMSCH_RESET_OPTION { +HANG_DETECT_AND_RESET =0 , +HANG_DETECT_ONLY =1 , +}; + +union UMSCHAPI__RESET { +struct { +union UMSCH_API_HEADER header; + +enum UMSCH_RESET_OPTION reset_option; +uint64_t doorbell_offset_addr; +enum UMSCH_ENGINE_TYPE engine_type; + +struct UMSCH_API_STATUS api_status; + }; + +uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS ]; +}; + +union UMSCHAPI__SET_LOGGING_BUFFER { +struct { +union UMSCH_API_HEADER header; +/* There are separate log buffers for each queue type */ +enum UMSCH_ENGINE_TYPE log_type; +/* Log buffer GPU Address */ +uint64_t logging_buffer_addr; +/* Number of entries in the log buffer */ +uint32_t number_of_entries; +/* Entry index at which CPU interrupt needs to be signalled */ +uint32_t interrupt_entry; + +struct UMSCH_API_STATUS api_status; + }; + +uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS ]; +}; + +union UMSCHAPI__UPDATE_AFFINITY { +struct { +union UMSCH_API_HEADER header; +union UMSCH_AFFINITY affinity; +uint64_t context_csa_addr; +struct UMSCH_API_STATUS api_status; + }; + +uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS ]; +}; + +union UMSCHAPI__CHANGE_CONTEXT_PRIORITY_LEVEL { +struct { +union UMSCH_API_HEADER header; +uint32_t inprocess_context_priority; +enum UMSCH_AMD_PRIORITY_LEVEL context_global_priority_level; +uint64_t context_quantum; +uint64_t context_csa_addr; +struct UMSCH_API_STATUS api_status; + }; + +uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS ]; +}; + +union UMSCHAPI__QUERY_UMSCH_STATUS { +struct { +union UMSCH_API_HEADER header; +bool umsch_mm_healthy;/* 0 - not healthy, 1 - healthy */ +struct UMSCH_API_STATUS api_status; + }; + +uint32_t max_dwords_in_api[API_FRAME_SIZE_IN_DWORDS ]; +}; + +#pragma pack(pop) + +#endif
Signature · 1 line
2.41.0
Alex Deucher
maintainerPATCH 6/14 drm/amdgpu: add initial support for UMSCH
In reply to Alex DeucherFrom: Lang Yu <[email protected]>
Add basic data structure, dummy ring functions
and ip functions for UMSCH.
Implement sw_init(ring_init and init_microcodede) and hw_init(load_microcode), UMSCH can boot up now.
Implement hw_init(ring_start) and hw_fini(ring_stop), UMSCH is ready for command submission now.
Implement set_hw_resources and add/remove_queue,
UMSCH is ready for scheduling now.
Aggregated doorbell is used to notify UMSCH FW that there is unmapped queue with corresponding priority level (e.g., AGDB[0] for Real time band, etc.) is updating its job.
v2: squash together initial patches to avoid breaking the
build (Alex)
--- drivers/gpu/drm/amd/amdgpu/Makefile | 5 + drivers/gpu/drm/amd/amdgpu/amdgpu.h | 5 + drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c | 361 +++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.h | 210 ++++++++++ drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c | 385 +++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.h | 30 ++ 6 files changed, 996 insertions(+) create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.h create mode 100644 drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c create mode 100644 drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.h
drivers/gpu/drm/amd/amdgpu/Makefile +5 -0
@@ -221,6 +221,11 @@ amdgpu-y += \ amdgpu_vpe.o \ vpe_v6_1.o +# add UMSCH block +amdgpu-y += \ + amdgpu_umsch_mm.o \ + umsch_mm_v4_0.o + # # add ATHUB block amdgpu-y += \
drivers/gpu/drm/amd/amdgpu/amdgpu.h +5 -0
@@ -81,6 +81,7 @@#include "amdgpu_vcn.h" #include "amdgpu_jpeg.h" #include "amdgpu_vpe.h" +#include "amdgpu_umsch_mm.h" #include "amdgpu_gmc.h" #include "amdgpu_gfx.h" #include "amdgpu_sdma.h"
@@ -952,6 +953,10 @@ struct amdgpu_device {/* vpe */ struct amdgpu_vpe vpe; +/* umsch */ +struct amdgpu_umsch_mm umsch_mm; +bool enable_umsch_mm; +/* firmwares */ struct amdgpu_firmware firmware;
drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c new file +361 -0
361 lines not on this page · raw message
drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.h new file +210 -0
210 lines not on this page · raw message
drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c new file +385 -0
385 lines not on this page · raw message
drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.h new file +30 -0
@@ -0,0 +1,30 @@ +/* SPDX-License-Identifier: MIT */ +/* +* Copyright 2023 Advanced Micro Devices, Inc. +* +* Permission is hereby granted, free of charge, to any person obtaining a +* copy of this software and associated documentation files (the "Software"), +* to deal in the Software without restriction, including without limitation +* the rights to use, copy, modify, merge, publish, distribute, sublicense, +* and/or sell copies of the Software, and to permit persons to whom the +* Software is furnished to do so, subject to the following conditions: +* +* The above copyright notice and this permission notice shall be included in +* all copies or substantial portions of the Software. +* +* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL +* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR +* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, +* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR +* OTHER DEALINGS IN THE SOFTWARE. +* +*/ + +#ifndef __UMSCH_MM_V4_0_H__ +#define __UMSCH_MM_V4_0_H__ + +void umsch_mm_v4_0_set_funcs (struct amdgpu_umsch_mm *umsch); + +#endif
Signature · 1 line
2.41.0
Alex Deucher
maintainerPATCH 7/14 drm/amdgpu: enable UMSCH scheduling for VPE
In reply to Alex DeucherFrom: Lang Yu <[email protected]>
Add VPE into UMSCH hw resourses,
set vmid mask to 0xf00,
set hqd mask to 0xfe,
then UMSCH can schedule VPE queues.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c | 4 ++++ 1 file changed, 4 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c +4 -0
@@ -231,6 +231,10 @@ static int umsch_mm_init(struct amdgpu_device *adev) {int r; + adev->umsch_mm.vmid_mask_mm_vpe =0xf00 ; + adev->umsch_mm.engine_mask = (1 <<UMSCH_SWIP_ENGINE_TYPE_VPE ); + adev->umsch_mm.vpe_hqd_mask =0xfe ; + r =amdgpu_device_wb_get (adev, &adev->umsch_mm.wb_index);if (r) {dev_err (adev->dev,"failed to alloc wb for umsch: %d\n" , r);
Signature · 1 line
2.41.0
Alex Deucher
maintainerPATCH 8/14 drm/amdgpu: add selftest framework for UMSCH
In reply to Alex DeucherFrom: Lang Yu <[email protected]>
Prepare for VPE and VCN queue submission test.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c | 406 ++++++++++++++++++- 1 file changed, 405 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c +405 -1
@@ -28,6 +28,408 @@#include "amdgpu_umsch_mm.h" #include "umsch_mm_v4_0.h" +struct umsch_mm_test_ctx_data { +uint8_t process_csa[PAGE_SIZE ]; +uint8_t vpe_ctx_csa[PAGE_SIZE ]; +uint8_t vcn_ctx_csa[PAGE_SIZE ]; +}; + +struct umsch_mm_test_mqd_data { +uint8_t vpe_mqd[PAGE_SIZE ]; +uint8_t vcn_mqd[PAGE_SIZE ]; +}; + +struct umsch_mm_test_ring_data { +uint8_t vpe_ring[PAGE_SIZE ]; +uint8_t vpe_ib[PAGE_SIZE ]; +uint8_t vcn_ring[PAGE_SIZE ]; +uint8_t vcn_ib[PAGE_SIZE ]; +}; + +struct umsch_mm_test_queue_info { +uint64_t mqd_addr; +uint64_t csa_addr; +uint32_t doorbell_offset_0; +uint32_t doorbell_offset_1; +enum UMSCH_SWIP_ENGINE_TYPE engine; +}; + +struct umsch_mm_test { +struct amdgpu_bo *ctx_data_obj; +uint64_t ctx_data_gpu_addr; +uint32_t *ctx_data_cpu_addr; + +struct amdgpu_bo *mqd_data_obj; +uint64_t mqd_data_gpu_addr; +uint32_t *mqd_data_cpu_addr; + +struct amdgpu_bo *ring_data_obj; +uint64_t ring_data_gpu_addr; +uint32_t *ring_data_cpu_addr; + + +struct amdgpu_vm *vm; +struct amdgpu_bo_va *bo_va; +uint32_t pasid; +uint32_t vm_cntx_cntl; +uint32_t num_queues; +}; + +static int map_ring_data (struct amdgpu_device *adev,struct amdgpu_vm *vm, +struct amdgpu_bo *bo,struct amdgpu_bo_va **bo_va, +uint64_t addr,uint32_t size) +{ +struct ww_acquire_ctx ticket; +struct list_head list; +struct amdgpu_bo_list_entry pd; +struct ttm_validate_buffer csa_tv; +struct amdgpu_sync sync; +int r; + +INIT_LIST_HEAD (&list); +INIT_LIST_HEAD (&csa_tv.head); + csa_tv.bo = &bo->tbo; + csa_tv.num_shared =1 ; + +list_add (&csa_tv.head, &list); +amdgpu_vm_get_pd_bo (vm, &list, &pd); + +amdgpu_sync_create (&sync); + + r =ttm_eu_reserve_buffers (&ticket, &list, true,NULL ); +if (r) +return r; + + *bo_va =amdgpu_vm_bo_add (adev, vm, bo); +if (!*bo_va) { +ttm_eu_backoff_reservation (&ticket, &list); +return -ENOMEM ; + } + + r =amdgpu_vm_bo_map (adev, *bo_va, addr,0 , size, +AMDGPU_PTE_READABLE |AMDGPU_PTE_WRITEABLE | +AMDGPU_PTE_EXECUTABLE ); + +if (r) { +amdgpu_vm_bo_del (adev, *bo_va); +ttm_eu_backoff_reservation (&ticket, &list); +return r; + } + + + r =amdgpu_vm_bo_update (adev, *bo_va, false); +if (r) + goto error; + +amdgpu_sync_fence (&sync, (*bo_va)->last_pt_update); + + r =amdgpu_vm_update_pdes (adev, vm, false); +if (r) + goto error; + +amdgpu_sync_fence (&sync, vm->last_update); + +amdgpu_sync_wait (&sync, false); + +ttm_eu_backoff_reservation (&ticket, &list); +amdgpu_sync_free (&sync); + +return 0 ; +error: +ttm_eu_backoff_reservation (&ticket, &list); +amdgpu_sync_free (&sync); +return r; +} + +static int unmap_ring_data (struct amdgpu_device *adev,struct amdgpu_vm *vm, +struct amdgpu_bo *bo,struct amdgpu_bo_va *bo_va, +uint64_t addr) +{ +struct ww_acquire_ctx ticket; +struct list_head list; +struct amdgpu_bo_list_entry pd; +struct ttm_validate_buffer csa_tv; +int r; + +INIT_LIST_HEAD (&list); +INIT_LIST_HEAD (&csa_tv.head); + csa_tv.bo = &bo->tbo; + csa_tv.num_shared =1 ; + +list_add (&csa_tv.head, &list); +amdgpu_vm_get_pd_bo (vm, &list, &pd); + + r =ttm_eu_reserve_buffers (&ticket, &list, true,NULL ); +if (r) +return r; + + r =amdgpu_vm_bo_unmap (adev, bo_va, addr); +if (r) { +ttm_eu_backoff_reservation (&ticket, &list); +return r; + } + +amdgpu_vm_bo_del (adev, bo_va); + +ttm_eu_backoff_reservation (&ticket, &list); + +return 0 ; +} + +static void setup_vpe_queue (struct amdgpu_device *adev, +struct umsch_mm_test *test, +struct umsch_mm_test_queue_info *qinfo) +{ +} + +static void setup_vcn_queue (struct amdgpu_device *adev, +struct umsch_mm_test *test, +struct umsch_mm_test_queue_info *qinfo) +{ +} + +static int add_test_queue (struct amdgpu_device *adev, +struct umsch_mm_test *test, +struct umsch_mm_test_queue_info *qinfo) +{ +struct umsch_mm_add_queue_input queue_input = {}; +int r; + + queue_input.process_id = test->pasid; + queue_input.page_table_base_addr =amdgpu_gmc_pd_addr (test->vm->root.bo); + + queue_input.process_va_start =0 ; + queue_input.process_va_end = (adev->vm_manager.max_pfn -1 ) <<AMDGPU_GPU_PAGE_SHIFT ; + + queue_input.process_quantum =100000 ;/* 10ms */ + queue_input.process_csa_addr = test->ctx_data_gpu_addr + + offsetof(struct umsch_mm_test_ctx_data , process_csa); + + queue_input.context_quantum =10000 ;/* 1ms */ + queue_input.context_csa_addr = qinfo->csa_addr; + + queue_input.inprocess_context_priority =CONTEXT_PRIORITY_LEVEL_NORMAL ; + queue_input.context_global_priority_level =CONTEXT_PRIORITY_LEVEL_NORMAL ; + queue_input.doorbell_offset_0 = qinfo->doorbell_offset_0; + queue_input.doorbell_offset_1 = qinfo->doorbell_offset_1; + + queue_input.engine_type = qinfo->engine; + queue_input.mqd_addr = qinfo->mqd_addr; + queue_input.vm_context_cntl = test->vm_cntx_cntl; + +amdgpu_umsch_mm_lock (&adev->umsch_mm); + r = adev->umsch_mm.funcs->add_queue (&adev->umsch_mm, &queue_input); +amdgpu_umsch_mm_unlock (&adev->umsch_mm); +if (r) +return r; + +return 0 ; +} + +static int remove_test_queue (struct amdgpu_device *adev, +struct umsch_mm_test *test, +struct umsch_mm_test_queue_info *qinfo) +{ +struct umsch_mm_remove_queue_input queue_input = {}; +int r; + + queue_input.doorbell_offset_0 = qinfo->doorbell_offset_0; + queue_input.doorbell_offset_1 = qinfo->doorbell_offset_1; + queue_input.context_csa_addr = qinfo->csa_addr; + +amdgpu_umsch_mm_lock (&adev->umsch_mm); + r = adev->umsch_mm.funcs->remove_queue (&adev->umsch_mm, &queue_input); +amdgpu_umsch_mm_unlock (&adev->umsch_mm); +if (r) +return r; + +return 0 ; +} + +static int submit_vpe_queue (struct amdgpu_device *adev,struct umsch_mm_test *test) +{ +return 0 ; +} + +static int submit_vcn_queue (struct amdgpu_device *adev,struct umsch_mm_test *test) +{ +return 0 ; +} + +static int setup_umsch_mm_test (struct amdgpu_device *adev, +struct umsch_mm_test *test) +{ +struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0 (0 )]; +int r; + + test->vm_cntx_cntl = hub->vm_cntx_cntl; + + test->vm =kzalloc (sizeof (*test->vm),GFP_KERNEL ); +if (!test->vm) { + r = -ENOMEM ; +return r; + } + + r =amdgpu_vm_init (adev, test->vm,-1 ); +if (r) + goto error_free_vm; + + test->pasid =amdgpu_pasid_alloc (16 ); +if (test->pasid <0 ) { + r = test->pasid; + goto error_fini_vm; + } + + r =amdgpu_bo_create_kernel (adev,sizeof (struct umsch_mm_test_ctx_data ), +PAGE_SIZE ,AMDGPU_GEM_DOMAIN_GTT , + &test->ctx_data_obj, + &test->ctx_data_gpu_addr, + (void **)&test->ctx_data_cpu_addr); +if (r) + goto error_free_pasid; + +memset (test->ctx_data_cpu_addr,0 ,sizeof (struct umsch_mm_test_ctx_data )); + + r =amdgpu_bo_create_kernel (adev,PAGE_SIZE , +PAGE_SIZE ,AMDGPU_GEM_DOMAIN_GTT , + &test->mqd_data_obj, + &test->mqd_data_gpu_addr, + (void **)&test->mqd_data_cpu_addr); +if (r) + goto error_free_ctx_data_obj; + +memset (test->mqd_data_cpu_addr,0 ,PAGE_SIZE ); + + r =amdgpu_bo_create_kernel (adev,sizeof (struct umsch_mm_test_ring_data ), +PAGE_SIZE ,AMDGPU_GEM_DOMAIN_GTT , + &test->ring_data_obj, +NULL , + (void **)&test->ring_data_cpu_addr); +if (r) + goto error_free_mqd_data_obj; + +memset (test->ring_data_cpu_addr,0 ,sizeof (struct umsch_mm_test_ring_data )); + + test->ring_data_gpu_addr =AMDGPU_VA_RESERVED_SIZE ; + r =map_ring_data (adev, test->vm, test->ring_data_obj, &test->bo_va, + test->ring_data_gpu_addr,sizeof (struct umsch_mm_test_ring_data )); +if (r) + goto error_free_ring_data_obj; + +return 0 ; + +error_free_ring_data_obj: +amdgpu_bo_free_kernel (&test->ring_data_obj,NULL , + (void **)&test->ring_data_cpu_addr); +error_free_mqd_data_obj: +amdgpu_bo_free_kernel (&test->mqd_data_obj, &test->mqd_data_gpu_addr, + (void **)&test->mqd_data_cpu_addr); +error_free_ctx_data_obj: +amdgpu_bo_free_kernel (&test->ctx_data_obj, &test->ctx_data_gpu_addr, + (void **)&test->ctx_data_cpu_addr); +error_free_pasid: +amdgpu_pasid_free (test->pasid); +error_fini_vm: +amdgpu_vm_fini (adev, test->vm); +error_free_vm: +kfree (test->vm); + +return r; +} + +static void cleanup_umsch_mm_test (struct amdgpu_device *adev, +struct umsch_mm_test *test) +{ +unmap_ring_data (adev, test->vm, test->ring_data_obj, + test->bo_va, test->ring_data_gpu_addr); +amdgpu_bo_free_kernel (&test->mqd_data_obj, &test->mqd_data_gpu_addr, + (void **)&test->mqd_data_cpu_addr); +amdgpu_bo_free_kernel (&test->ring_data_obj,NULL , + (void **)&test->ring_data_cpu_addr); +amdgpu_bo_free_kernel (&test->ctx_data_obj, &test->ctx_data_gpu_addr, + (void **)&test->ctx_data_cpu_addr); +amdgpu_pasid_free (test->pasid); +amdgpu_vm_fini (adev, test->vm); +kfree (test->vm); +} + +static int setup_test_queues (struct amdgpu_device *adev, +struct umsch_mm_test *test, +struct umsch_mm_test_queue_info *qinfo) +{ +int i, r; + +for (i =0 ; i < test->num_queues; i++) { +if (qinfo[i].engine ==UMSCH_SWIP_ENGINE_TYPE_VPE ) +setup_vpe_queue (adev, test, &qinfo[i]); +else +setup_vcn_queue (adev, test, &qinfo[i]); + + r =add_test_queue (adev, test, &qinfo[i]); +if (r) +return r; + } + +return 0 ; +} + +static int submit_test_queues (struct amdgpu_device *adev, +struct umsch_mm_test *test, +struct umsch_mm_test_queue_info *qinfo) +{ +int i, r; + +for (i =0 ; i < test->num_queues; i++) { +if (qinfo[i].engine ==UMSCH_SWIP_ENGINE_TYPE_VPE ) + r =submit_vpe_queue (adev, test); +else + r =submit_vcn_queue (adev, test); +if (r) +return r; + } + +return 0 ; +} + +static void cleanup_test_queues (struct amdgpu_device *adev, +struct umsch_mm_test *test, +struct umsch_mm_test_queue_info *qinfo) +{ +int i; + +for (i =0 ; i < test->num_queues; i++) +remove_test_queue (adev, test, &qinfo[i]); +} + +static int umsch_mm_test (struct amdgpu_device *adev) +{ +struct umsch_mm_test_queue_info qinfo[] = {}; +struct umsch_mm_test test = { .num_queues =ARRAY_SIZE (qinfo) }; +int r; + + r =setup_umsch_mm_test (adev, &test); +if (r) +return r; + + r =setup_test_queues (adev, &test, qinfo); +if (r) + goto cleanup; + + r =submit_test_queues (adev, &test, qinfo); +if (r) + goto cleanup; + +cleanup_test_queues (adev, &test, qinfo); +cleanup_umsch_mm_test (adev, &test); + +return 0 ; + +cleanup: +cleanup_test_queues (adev, &test, qinfo); +cleanup_umsch_mm_test (adev, &test); +return r; +} +int amdgpu_umsch_mm_submit_pkt (struct amdgpu_umsch_mm *umsch,void *pkt,int ndws) {struct amdgpu_ring *ring = &umsch->ring;
@@ -272,7 +674,9 @@ static int umsch_mm_early_init(void *handle)static int umsch_mm_late_init (void *handle) { -return 0 ; +struct amdgpu_device *adev = (struct amdgpu_device *)handle; + +return umsch_mm_test (adev); }static int umsch_mm_sw_init (void *handle)
Signature · 1 line
2.41.0
Alex Deucher
maintainerPATCH 9/14 drm/amdgpu: add VPE queue submission test
In reply to Alex DeucherFrom: Lang Yu <[email protected]>
Submit a fence command through indirect buffer.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c | 56 +++++++++++++++++++- 1 file changed, 54 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c +54 -2
@@ -180,6 +180,21 @@ static void setup_vpe_queue(struct amdgpu_device *adev,struct umsch_mm_test *test, struct umsch_mm_test_queue_info *qinfo) { + struct MQD_INFO *mqd = (struct MQD_INFO *)test->mqd_data_cpu_addr; +uint64_t ring_gpu_addr = test->ring_data_gpu_addr; + + mqd->rb_base_lo = (ring_gpu_addr >>8 ); + mqd->rb_base_hi = (ring_gpu_addr >>40 ); + mqd->rb_size =PAGE_SIZE /4 ; + mqd->wptr_val =0 ; + mqd->rptr_val =0 ; + mqd->unmapped =1 ; + + qinfo->mqd_addr = test->mqd_data_gpu_addr; + qinfo->csa_addr = test->ctx_data_gpu_addr + + offsetof(struct umsch_mm_test_ctx_data , vpe_ctx_csa); + qinfo->doorbell_offset_0 = (adev->doorbell_index.vpe_ring +1 ) <<1 ; + qinfo->doorbell_offset_1 =0 ; } static void setup_vcn_queue(struct amdgpu_device *adev,
@@ -248,7 +263,42 @@ static int remove_test_queue(struct amdgpu_device *adev,static int submit_vpe_queue (struct amdgpu_device *adev,struct umsch_mm_test *test) { -return 0 ; +struct MQD_INFO *mqd = (struct MQD_INFO *)test->mqd_data_cpu_addr; +uint32_t *ring = test->ring_data_cpu_addr + + offsetof(struct umsch_mm_test_ring_data , vpe_ring) /4 ; +uint32_t *ib = test->ring_data_cpu_addr + + offsetof(struct umsch_mm_test_ring_data , vpe_ib) /4 ; +uint64_t ib_gpu_addr = test->ring_data_gpu_addr + + offsetof(struct umsch_mm_test_ring_data , vpe_ib); +uint32_t *fence = ib +2048 /4 ; +uint64_t fence_gpu_addr = ib_gpu_addr +2048 ; +const uint32_t test_pattern =0xdeadbeef ; +int i; + + ib[0 ] =VPE_CMD_HEADER (VPE_CMD_OPCODE_FENCE ,0 ); + ib[1 ] =lower_32_bits (fence_gpu_addr); + ib[2 ] =upper_32_bits (fence_gpu_addr); + ib[3 ] = test_pattern; + + ring[0 ] =VPE_CMD_HEADER (VPE_CMD_OPCODE_INDIRECT ,0 ); + ring[1 ] = (ib_gpu_addr &0xffffffe0 ); + ring[2 ] =upper_32_bits (ib_gpu_addr); + ring[3 ] =4 ; + ring[4 ] =0 ; + ring[5 ] =0 ; + + mqd->wptr_val = (6 <<2 ); +// WDOORBELL32(adev->umsch_mm.agdb_index[CONTEXT_PRIORITY_LEVEL_NORMAL], mqd->wptr_val); + +for (i =0 ; i < adev->usec_timeout; i++) { +if (*fence == test_pattern) +return 0 ; +udelay (1 ); + } + +dev_err (adev->dev,"vpe queue submission timeout\n" ); + +return -ETIMEDOUT ; }static int submit_vcn_queue (struct amdgpu_device *adev,struct umsch_mm_test *test)
@@ -403,7 +453,9 @@ static void cleanup_test_queues(struct amdgpu_device *adev,static int umsch_mm_test (struct amdgpu_device *adev) { -struct umsch_mm_test_queue_info qinfo[] = {}; +struct umsch_mm_test_queue_info qinfo[] = { + { .engine =UMSCH_SWIP_ENGINE_TYPE_VPE }, + };struct umsch_mm_test test = { .num_queues =ARRAY_SIZE (qinfo) };int r;
Signature · 1 line
2.41.0
Alex Deucher
maintainerPATCH 10/14 drm/amdgpu: reserve mmhub engine 3 for UMSCH FW
In reply to Alex DeucherFrom: Lang Yu <[email protected]>
UMSCH FW uses mmhub engine 3 for invalidation.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +5 -1
@@ -549,13 +549,17 @@ int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev)/* reserve engine 5 for firmware */ if (adev->enable_mes) vm_inv_engs[i] &= ~(1 <<5 ); +/* reserve mmhub engine 3 for firmware */ +if (adev->enable_umsch_mm) + vm_inv_engs[i] &= ~(1 <<3 ); }for (i =0 ; i < adev->num_rings; ++i) { ring = adev->rings[i]; vmhub = ring->vm_hub; -if (ring == &adev->mes.ring) +if (ring == &adev->mes.ring || + ring == &adev->umsch_mm.ring)continue ; inv_eng =ffs (vm_inv_engs[vmhub]);
Signature · 1 line
2.41.0
Alex Deucher
maintainerPATCH 11/14 drm/amdgpu: add PSP loading support for UMSCH
In reply to Alex DeucherFrom: Lang Yu <[email protected]>
Add front door loading support.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 9 ++ drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c | 21 ++++ drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h | 3 + drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c | 54 ++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.h | 8 ++ drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c | 102 +++++++++++-------- 6 files changed, 156 insertions(+), 41 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +9 -0
@@ -2399,6 +2399,15 @@ static int psp_get_fw_type(struct amdgpu_firmware_info *ucode,case AMDGPU_UCODE_ID_VPE : *type =GFX_FW_TYPE_VPE ;break ; +case AMDGPU_UCODE_ID_UMSCH_MM_UCODE : + *type =GFX_FW_TYPE_UMSCH_UCODE ; +break ; +case AMDGPU_UCODE_ID_UMSCH_MM_DATA : + *type =GFX_FW_TYPE_UMSCH_DATA ; +break ; +case AMDGPU_UCODE_ID_UMSCH_MM_CMD_BUFFER : + *type =GFX_FW_TYPE_UMSCH_CMD_BUFFER ; +break ;case AMDGPU_UCODE_ID_MAXIMUM :default :return -EINVAL ;
drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c +21 -0
@@ -664,6 +664,16 @@ const char *amdgpu_ucode_name(enum AMDGPU_UCODE_ID ucode_id)return "DMCUB" ;case AMDGPU_UCODE_ID_CAP :return "CAP" ; +case AMDGPU_UCODE_ID_VPE_CTX : +return "VPE_CTX" ; +case AMDGPU_UCODE_ID_VPE_CTL : +return "VPE_CTL" ; +case AMDGPU_UCODE_ID_VPE : +return "VPE" ; +case AMDGPU_UCODE_ID_UMSCH_MM_UCODE : +return "UMSCH_MM_UCODE" ; +case AMDGPU_UCODE_ID_UMSCH_MM_DATA : +return "UMSCH_MM_DATA" ;default :return "UNKNOWN UCODE" ; }
@@ -750,6 +760,7 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev,const struct sdma_firmware_header_v2_0 *sdma_hdr =NULL ;const struct imu_firmware_header_v1_0 *imu_hdr =NULL ;const struct vpe_firmware_header_v1_0 *vpe_hdr =NULL ; +const struct umsch_mm_firmware_header_v1_0 *umsch_mm_hdr =NULL ;u8 *ucode_addr;if (!ucode->fw)
@@ -962,6 +973,16 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev, ucode_addr = (u8 *)ucode->fw->data +le32_to_cpu (vpe_hdr->ctl_ucode_offset);break ; +case AMDGPU_UCODE_ID_UMSCH_MM_UCODE : + ucode->ucode_size =le32_to_cpu (umsch_mm_hdr->umsch_mm_ucode_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + +le32_to_cpu (umsch_mm_hdr->header.ucode_array_offset_bytes); +break ; +case AMDGPU_UCODE_ID_UMSCH_MM_DATA : + ucode->ucode_size =le32_to_cpu (umsch_mm_hdr->umsch_mm_ucode_data_size_bytes); + ucode_addr = (u8 *)ucode->fw->data + +le32_to_cpu (umsch_mm_hdr->umsch_mm_ucode_data_offset_bytes); +break ;default : ucode->ucode_size =le32_to_cpu (header->ucode_size_bytes); ucode_addr = (u8 *)ucode->fw->data +
drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h +3 -0
@@ -507,6 +507,9 @@ enum AMDGPU_UCODE_ID {AMDGPU_UCODE_ID_VPE_CTX ,AMDGPU_UCODE_ID_VPE_CTL ,AMDGPU_UCODE_ID_VPE , +AMDGPU_UCODE_ID_UMSCH_MM_UCODE , +AMDGPU_UCODE_ID_UMSCH_MM_DATA , +AMDGPU_UCODE_ID_UMSCH_MM_CMD_BUFFER , AMDGPU_UCODE_ID_MAXIMUM, };
drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c +54 -0
84 lines not on this page · raw message
drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.h +8 -0
@@ -147,6 +147,10 @@ struct amdgpu_umsch_mm {uint64_t data_start_addr;uint32_t data_size; +struct amdgpu_bo *cmd_buf_obj; +uint64_t cmd_buf_gpu_addr; +uint32_t *cmd_buf_ptr; +uint32_t wb_index;uint64_t sch_ctx_gpu_addr;uint32_t *sch_ctx_cpu_addr;
@@ -163,12 +167,16 @@ struct amdgpu_umsch_mm {struct mutex mutex_hidden; }; +int umsch_mm_psp_update_sram (struct amdgpu_device *adev,u32 ucode_size); +int amdgpu_umsch_mm_submit_pkt (struct amdgpu_umsch_mm *umsch,void *pkt,int ndws);int amdgpu_umsch_mm_query_fence (struct amdgpu_umsch_mm *umsch);int amdgpu_umsch_mm_init_microcode (struct amdgpu_umsch_mm *umsch);int amdgpu_umsch_mm_allocate_ucode_buffer (struct amdgpu_umsch_mm *umsch);int amdgpu_umsch_mm_allocate_ucode_data_buffer (struct amdgpu_umsch_mm *umsch); +void *amdgpu_umsch_mm_add_cmd (struct amdgpu_umsch_mm *umsch, +void * cmd_ptr,uint32_t reg_offset,uint32_t reg_data);int amdgpu_umsch_mm_ring_init (struct amdgpu_umsch_mm *umsch);
drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c +61 -41
158 lines not on this page · raw message
Signature · 1 line
2.41.0
Alex Deucher
maintainerPATCH 12/14 drm/amdgpu/discovery: enable UMSCH 4.0 in IP discovery
In reply to Alex DeucherFrom: Lang Yu <[email protected]>
Enable UMSCH to support VPE and VCN user queues.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +18 -0
@@ -2159,6 +2159,20 @@ static int amdgpu_discovery_set_vpe_ip_blocks(struct amdgpu_device *adev)return 0 ; } +static int amdgpu_discovery_set_umsch_mm_ip_blocks (struct amdgpu_device *adev) +{ +switch (adev->ip_versions[VCN_HWIP ][0 ]) { +case IP_VERSION (4 ,0 ,5 ): +amdgpu_device_ip_block_add (adev, &umsch_mm_v4_0_ip_block); + adev->enable_umsch_mm = true; +break ; +default : +break ; + } + +return 0 ; +} +int amdgpu_discovery_set_ip_blocks (struct amdgpu_device *adev) {int r;
@@ -2650,6 +2664,10 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)if (r)return r; + r =amdgpu_discovery_set_umsch_mm_ip_blocks (adev); +if (r) +return r; +return 0 ; }
Signature · 1 line
2.41.0
Alex Deucher
maintainerPATCH 13/14 drm/amdgpu: add amdgpu_umsch_mm module parameter
In reply to Alex DeucherFrom: Lang Yu <[email protected]>
Enable Multi Media User Mode Scheduler
(0 = disabled (default), 1 = enabled).
--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 6 ++++-- drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 10 ++++++++++ 3 files changed, 15 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu.h +1 -0
@@ -245,6 +245,7 @@ extern int amdgpu_num_kcq;#define AMDGPU_VCNFW_LOG_SIZE (32 * 1024)extern int amdgpu_vcnfw_log;extern int amdgpu_sg_display; +extern int amdgpu_umsch_mm;extern int amdgpu_user_partt_mode;
drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +4 -2
@@ -2163,8 +2163,10 @@ static int amdgpu_discovery_set_umsch_mm_ip_blocks(struct amdgpu_device *adev) {switch (adev->ip_versions[VCN_HWIP ][0 ]) {case IP_VERSION (4 ,0 ,5 ): -amdgpu_device_ip_block_add (adev, &umsch_mm_v4_0_ip_block); - adev->enable_umsch_mm = true; +if (amdgpu_umsch_mm &0x1 ) { +amdgpu_device_ip_block_add (adev, &umsch_mm_v4_0_ip_block); + adev->enable_umsch_mm = true; + }break ;default :break ;
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +10 -0
@@ -194,6 +194,7 @@ int amdgpu_use_xgmi_p2p = 1;int amdgpu_vcnfw_log;int amdgpu_sg_display =-1 ;/* auto */ int amdgpu_user_partt_mode =AMDGPU_AUTO_COMPUTE_PARTITION_MODE ; +int amdgpu_umsch_mm;static void amdgpu_drv_delayed_reset_work_handler (struct work_struct *work);
@@ -907,6 +908,15 @@ module_param_named(vcnfw_log, amdgpu_vcnfw_log, int, 0444);MODULE_PARM_DESC (sg_display,"S/G Display (-1 = auto (default), 0 = disable)" );module_param_named (sg_display, amdgpu_sg_display, int,0444 ); +/** +* DOC: umsch_mm (int) +* Enable Multi Media User Mode Scheduler. This is a HW scheduling engine for VCN and VPE. +* (0 = disabled (default), 1 = enabled) +*/ +MODULE_PARM_DESC (umsch_mm, +"Enable Multi Media User Mode Scheduler (0 = disabled (default), 1 = enabled)" ); +module_param_named (umsch_mm, amdgpu_umsch_mm, int,0444 ); +/** * DOC: smu_pptable_id (int) * Used to override pptable id. id = 0 use VBIOS pptable.
Signature · 1 line
2.41.0
Alex Deucher
maintainerPATCH 14/14 drm/amdgpu: update SET_HW_RESOURCES definition for UMSCH
In reply to Alex DeucherFrom: Lang Yu <[email protected]>
Align with FW changes.
--- drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c | 3 +++ drivers/gpu/drm/amd/include/umsch_mm_4_0_api_def.h | 3 +++ 2 files changed, 6 insertions(+)
drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c +3 -0
@@ -279,6 +279,9 @@ static int umsch_mm_v4_0_set_hw_resources(struct amdgpu_umsch_mm *umsch) sizeof(uint32_t) * 5); set_hw_resources.osssys_version = adev->ip_versions[OSSSYS_HWIP ][0 ]; + set_hw_resources.vcn_version = adev->ip_versions[VCN_HWIP ][0 ]; + set_hw_resources.vpe_version = adev->ip_versions[VPE_HWIP ][0 ]; + set_hw_resources.api_status.api_completion_fence_addr = umsch->ring.fence_drv.gpu_addr; set_hw_resources.api_status.api_completion_fence_value = ++umsch->ring.fence_drv.sync_seq;
drivers/gpu/drm/amd/include/umsch_mm_4_0_api_def.h +3 -0
@@ -212,6 +212,7 @@ union UMSCHAPI__SET_HW_RESOURCES {union UMSCH_API_HEADER header;uint32_t vmid_mask_mm_vcn;uint32_t vmid_mask_mm_vpe; +uint32_t collaboration_mask_vpe;uint32_t engine_mask;uint32_t logging_vmid;uint32_t vcn0_hqd_mask[MAX_VCN0_INSTANCES ];
@@ -223,6 +224,8 @@ union UMSCHAPI__SET_HW_RESOURCES {uint32_t mmhub_version;uint32_t osssys_base[UMSCH_MAX_HWIP_SEGMENT ];uint32_t osssys_version; +uint32_t vcn_version; +uint32_t vpe_version;struct UMSCH_API_STATUS api_status; union { struct {
Signature · 1 line
2.41.0