From e60f8db5e4c85d7f8868a06b02f06c0ae0e4c332 Mon Sep 17 00:00:00 2001 From: Alex Xie Date: Thu, 9 Mar 2017 11:36:26 -0500 Subject: [PATCH] drm/amdgpu: Add GMC 9.0 support (v2) On SOC-15 parts, the GMC (Graphics Memory Controller) consists of two hubs: GFX (graphics and compute) and MM (sdma, uvd, vce). v2: drop sdma from Makefile, fix duplicate return statement. Signed-off-by: Alex Xie Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/Makefile | 3 +- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 30 ++ drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 28 +- drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c | 447 +++++++++++++++++ drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.h | 35 ++ drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 825 +++++++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/gmc_v9_0.h | 30 ++ drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c | 585 ++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h | 35 ++ drivers/gpu/drm/amd/include/amd_shared.h | 2 + 10 files changed, 2013 insertions(+), 7 deletions(-) create mode 100644 drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c create mode 100644 drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.h create mode 100644 drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c create mode 100644 drivers/gpu/drm/amd/amdgpu/gmc_v9_0.h create mode 100644 drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c create mode 100644 drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile index 4ae5ddc..ccbc4c0 100644 --- a/drivers/gpu/drm/amd/amdgpu/Makefile +++ b/drivers/gpu/drm/amd/amdgpu/Makefile @@ -39,7 +39,8 @@ amdgpu-y += \ # add GMC block amdgpu-y += \ gmc_v7_0.o \ - gmc_v8_0.o + gmc_v8_0.o \ + gfxhub_v1_0.o mmhub_v1_0.o gmc_v9_0.o # add IH block amdgpu-y += \ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index f990eec..13ea68a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -121,6 +121,11 @@ extern int amdgpu_param_buf_per_se; /* max number of IP instances */ #define AMDGPU_MAX_SDMA_INSTANCES 2 +/* max number of VMHUB */ +#define AMDGPU_MAX_VMHUBS 2 +#define AMDGPU_MMHUB 0 +#define AMDGPU_GFXHUB 1 + /* hardcode that limit for now */ #define AMDGPU_VA_RESERVED_SIZE (8 << 20) @@ -308,6 +313,12 @@ struct amdgpu_gart_funcs { uint32_t flags); }; +/* provided by the mc block */ +struct amdgpu_mc_funcs { + /* adjust mc addr in fb for APU case */ + u64 (*adjust_mc_addr)(struct amdgpu_device *adev, u64 addr); +}; + /* provided by the ih block */ struct amdgpu_ih_funcs { /* ring read/write ptr handling, called from interrupt context */ @@ -557,6 +568,21 @@ int amdgpu_gart_bind(struct amdgpu_device *adev, uint64_t offset, int amdgpu_ttm_recover_gart(struct amdgpu_device *adev); /* + * VMHUB structures, functions & helpers + */ +struct amdgpu_vmhub { + uint32_t ctx0_ptb_addr_lo32; + uint32_t ctx0_ptb_addr_hi32; + uint32_t vm_inv_eng0_req; + uint32_t vm_inv_eng0_ack; + uint32_t vm_context0_cntl; + uint32_t vm_l2_pro_fault_status; + uint32_t vm_l2_pro_fault_cntl; + uint32_t (*get_invalidate_req)(unsigned int vm_id); + uint32_t (*get_vm_protection_bits)(void); +}; + +/* * GPU MC structures, functions & helpers */ struct amdgpu_mc { @@ -589,6 +615,9 @@ struct amdgpu_mc { u64 shared_aperture_end; u64 private_aperture_start; u64 private_aperture_end; + /* protects concurrent invalidation */ + spinlock_t invalidate_lock; + const struct amdgpu_mc_funcs *mc_funcs; }; /* @@ -1473,6 +1502,7 @@ struct amdgpu_device { struct amdgpu_gart gart; struct amdgpu_dummy_page dummy_page; struct amdgpu_vm_manager vm_manager; + struct amdgpu_vmhub vmhub[AMDGPU_MAX_VMHUBS]; /* memory management */ struct amdgpu_mman mman; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 2d9ec9c..3d6b9a9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -447,6 +447,16 @@ static bool amdgpu_vm_ring_has_compute_vm_bug(struct amdgpu_ring *ring) return false; } +static u64 amdgpu_vm_adjust_mc_addr(struct amdgpu_device *adev, u64 mc_addr) +{ + u64 addr = mc_addr; + + if (adev->mc.mc_funcs && adev->mc.mc_funcs->adjust_mc_addr) + addr = adev->mc.mc_funcs->adjust_mc_addr(adev, addr); + + return addr; +} + /** * amdgpu_vm_flush - hardware flush the vm * @@ -477,9 +487,10 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job) if (ring->funcs->emit_vm_flush && (job->vm_needs_flush || amdgpu_vm_is_gpu_reset(adev, id))) { struct dma_fence *fence; + u64 pd_addr = amdgpu_vm_adjust_mc_addr(adev, job->vm_pd_addr); - trace_amdgpu_vm_flush(job->vm_pd_addr, ring->idx, job->vm_id); - amdgpu_ring_emit_vm_flush(ring, job->vm_id, job->vm_pd_addr); + trace_amdgpu_vm_flush(pd_addr, ring->idx, job->vm_id); + amdgpu_ring_emit_vm_flush(ring, job->vm_id, pd_addr); r = amdgpu_fence_emit(ring, &fence); if (r) @@ -715,15 +726,18 @@ int amdgpu_vm_update_page_directory(struct amdgpu_device *adev, (count == AMDGPU_VM_MAX_UPDATE_SIZE)) { if (count) { + uint64_t pt_addr = + amdgpu_vm_adjust_mc_addr(adev, last_pt); + if (shadow) amdgpu_vm_do_set_ptes(¶ms, last_shadow, - last_pt, count, + pt_addr, count, incr, AMDGPU_PTE_VALID); amdgpu_vm_do_set_ptes(¶ms, last_pde, - last_pt, count, incr, + pt_addr, count, incr, AMDGPU_PTE_VALID); } @@ -737,11 +751,13 @@ int amdgpu_vm_update_page_directory(struct amdgpu_device *adev, } if (count) { + uint64_t pt_addr = amdgpu_vm_adjust_mc_addr(adev, last_pt); + if (vm->page_directory->shadow) - amdgpu_vm_do_set_ptes(¶ms, last_shadow, last_pt, + amdgpu_vm_do_set_ptes(¶ms, last_shadow, pt_addr, count, incr, AMDGPU_PTE_VALID); - amdgpu_vm_do_set_ptes(¶ms, last_pde, last_pt, + amdgpu_vm_do_set_ptes(¶ms, last_pde, pt_addr, count, incr, AMDGPU_PTE_VALID); } diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c new file mode 100644 index 0000000..1ff019c --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c @@ -0,0 +1,447 @@ +/* + * Copyright 2016 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. + * + */ +#include "amdgpu.h" +#include "gfxhub_v1_0.h" + +#include "vega10/soc15ip.h" +#include "vega10/GC/gc_9_0_offset.h" +#include "vega10/GC/gc_9_0_sh_mask.h" +#include "vega10/GC/gc_9_0_default.h" +#include "vega10/vega10_enum.h" + +#include "soc15_common.h" + +int gfxhub_v1_0_gart_enable(struct amdgpu_device *adev) +{ + u32 tmp; + u64 value; + u32 i; + + /* Program MC. */ + /* Update configuration */ + WREG32(SOC15_REG_OFFSET(GC, 0, mmMC_VM_SYSTEM_APERTURE_LOW_ADDR), + adev->mc.vram_start >> 18); + WREG32(SOC15_REG_OFFSET(GC, 0, mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR), + adev->mc.vram_end >> 18); + + value = adev->vram_scratch.gpu_addr - adev->mc.vram_start + + adev->vm_manager.vram_base_offset; + WREG32(SOC15_REG_OFFSET(GC, 0, + mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB), + (u32)(value >> 12)); + WREG32(SOC15_REG_OFFSET(GC, 0, + mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB), + (u32)(value >> 44)); + + /* Disable AGP. */ + WREG32(SOC15_REG_OFFSET(GC, 0, mmMC_VM_AGP_BASE), 0); + WREG32(SOC15_REG_OFFSET(GC, 0, mmMC_VM_AGP_TOP), 0); + WREG32(SOC15_REG_OFFSET(GC, 0, mmMC_VM_AGP_BOT), 0xFFFFFFFF); + + /* GART Enable. */ + + /* Setup TLB control */ + tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmMC_VM_MX_L1_TLB_CNTL)); + tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1); + tmp = REG_SET_FIELD(tmp, + MC_VM_MX_L1_TLB_CNTL, + SYSTEM_ACCESS_MODE, + 3); + tmp = REG_SET_FIELD(tmp, + MC_VM_MX_L1_TLB_CNTL, + ENABLE_ADVANCED_DRIVER_MODEL, + 1); + tmp = REG_SET_FIELD(tmp, + MC_VM_MX_L1_TLB_CNTL, + SYSTEM_APERTURE_UNMAPPED_ACCESS, + 0); + tmp = REG_SET_FIELD(tmp, + MC_VM_MX_L1_TLB_CNTL, + ECO_BITS, + 0); + tmp = REG_SET_FIELD(tmp, + MC_VM_MX_L1_TLB_CNTL, + MTYPE, + MTYPE_UC);/* XXX for emulation. */ + tmp = REG_SET_FIELD(tmp, + MC_VM_MX_L1_TLB_CNTL, + ATC_EN, + 1); + WREG32(SOC15_REG_OFFSET(GC, 0, mmMC_VM_MX_L1_TLB_CNTL), tmp); + + /* Setup L2 cache */ + tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmVM_L2_CNTL)); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 1); + tmp = REG_SET_FIELD(tmp, + VM_L2_CNTL, + ENABLE_L2_FRAGMENT_PROCESSING, + 0); + tmp = REG_SET_FIELD(tmp, + VM_L2_CNTL, + L2_PDE0_CACHE_TAG_GENERATION_MODE, + 0);/* XXX for emulation, Refer to closed source code.*/ + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 1); + tmp = REG_SET_FIELD(tmp, + VM_L2_CNTL, + CONTEXT1_IDENTITY_ACCESS_MODE, + 1); + tmp = REG_SET_FIELD(tmp, + VM_L2_CNTL, + IDENTITY_MODE_FRAGMENT_SIZE, + 0); + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_L2_CNTL), tmp); + + tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmVM_L2_CNTL2)); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_L2_CACHE, 1); + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_L2_CNTL2), tmp); + + tmp = mmVM_L2_CNTL3_DEFAULT; + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_L2_CNTL3), tmp); + + tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmVM_L2_CNTL4)); + tmp = REG_SET_FIELD(tmp, + VM_L2_CNTL4, + VMC_TAP_PDE_REQUEST_PHYSICAL, + 0); + tmp = REG_SET_FIELD(tmp, + VM_L2_CNTL4, + VMC_TAP_PTE_REQUEST_PHYSICAL, + 0); + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_L2_CNTL4), tmp); + + /* setup context0 */ + WREG32(SOC15_REG_OFFSET(GC, 0, + mmVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32), + (u32)(adev->mc.gtt_start >> 12)); + WREG32(SOC15_REG_OFFSET(GC, 0, + mmVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32), + (u32)(adev->mc.gtt_start >> 44)); + + WREG32(SOC15_REG_OFFSET(GC, 0, + mmVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32), + (u32)(adev->mc.gtt_end >> 12)); + WREG32(SOC15_REG_OFFSET(GC, 0, + mmVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32), + (u32)(adev->mc.gtt_end >> 44)); + + BUG_ON(adev->gart.table_addr & (~0x0000FFFFFFFFF000ULL)); + value = adev->gart.table_addr - adev->mc.vram_start + + adev->vm_manager.vram_base_offset; + value &= 0x0000FFFFFFFFF000ULL; + value |= 0x1; /*valid bit*/ + + WREG32(SOC15_REG_OFFSET(GC, 0, + mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32), + (u32)value); + WREG32(SOC15_REG_OFFSET(GC, 0, + mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32), + (u32)(value >> 32)); + + WREG32(SOC15_REG_OFFSET(GC, 0, + mmVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32), + (u32)(adev->dummy_page.addr >> 12)); + WREG32(SOC15_REG_OFFSET(GC, 0, + mmVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32), + (u32)(adev->dummy_page.addr >> 44)); + + tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmVM_L2_PROTECTION_FAULT_CNTL2)); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL2, + ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, + 1); + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_L2_PROTECTION_FAULT_CNTL2), tmp); + + tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_CNTL)); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH, 0); + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_CNTL), tmp); + + /* Disable identity aperture.*/ + WREG32(SOC15_REG_OFFSET(GC, 0, + mmVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32), 0XFFFFFFFF); + WREG32(SOC15_REG_OFFSET(GC, 0, + mmVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32), 0x0000000F); + + WREG32(SOC15_REG_OFFSET(GC, 0, + mmVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32), 0); + WREG32(SOC15_REG_OFFSET(GC, 0, + mmVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32), 0); + + WREG32(SOC15_REG_OFFSET(GC, 0, + mmVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32), 0); + WREG32(SOC15_REG_OFFSET(GC, 0, + mmVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32), 0); + + for (i = 0; i <= 14; i++) { + tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT1_CNTL) + i); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + PAGE_TABLE_BLOCK_SIZE, + amdgpu_vm_block_size - 9); + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT1_CNTL) + i, tmp); + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32) + i*2, 0); + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32) + i*2, 0); + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32) + i*2, + adev->vm_manager.max_pfn - 1); + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32) + i*2, 0); + } + + + return 0; +} + +void gfxhub_v1_0_gart_disable(struct amdgpu_device *adev) +{ + u32 tmp; + u32 i; + + /* Disable all tables */ + for (i = 0; i < 16; i++) + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_CNTL) + i, 0); + + /* Setup TLB control */ + tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmMC_VM_MX_L1_TLB_CNTL)); + tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0); + tmp = REG_SET_FIELD(tmp, + MC_VM_MX_L1_TLB_CNTL, + ENABLE_ADVANCED_DRIVER_MODEL, + 0); + WREG32(SOC15_REG_OFFSET(GC, 0, mmMC_VM_MX_L1_TLB_CNTL), tmp); + + /* Setup L2 cache */ + tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmVM_L2_CNTL)); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 0); + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_L2_CNTL), tmp); + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_L2_CNTL3), 0); +} + +/** + * gfxhub_v1_0_set_fault_enable_default - update GART/VM fault handling + * + * @adev: amdgpu_device pointer + * @value: true redirects VM faults to the default page + */ +void gfxhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev, + bool value) +{ + u32 tmp; + tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmVM_L2_PROTECTION_FAULT_CNTL)); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, + VM_L2_PROTECTION_FAULT_CNTL, + TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT, + value); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + READ_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_L2_PROTECTION_FAULT_CNTL), tmp); +} + +static uint32_t gfxhub_v1_0_get_invalidate_req(unsigned int vm_id) +{ + u32 req = 0; + + /* invalidate using legacy mode on vm_id*/ + req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, + PER_VMID_INVALIDATE_REQ, 1 << vm_id); + req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, 0); + req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PTES, 1); + req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE0, 1); + req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE1, 1); + req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE2, 1); + req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, INVALIDATE_L1_PTES, 1); + req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, + CLEAR_PROTECTION_FAULT_STATUS_ADDR, 0); + + return req; +} + +static uint32_t gfxhub_v1_0_get_vm_protection_bits(void) +{ + return (VM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + VM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + VM_CONTEXT1_CNTL__PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + VM_CONTEXT1_CNTL__VALID_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + VM_CONTEXT1_CNTL__READ_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + VM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + VM_CONTEXT1_CNTL__EXECUTE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK); +} + +static int gfxhub_v1_0_early_init(void *handle) +{ + return 0; +} + +static int gfxhub_v1_0_late_init(void *handle) +{ + return 0; +} + +static int gfxhub_v1_0_sw_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB]; + + hub->ctx0_ptb_addr_lo32 = + SOC15_REG_OFFSET(GC, 0, + mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32); + hub->ctx0_ptb_addr_hi32 = + SOC15_REG_OFFSET(GC, 0, + mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32); + hub->vm_inv_eng0_req = + SOC15_REG_OFFSET(GC, 0, mmVM_INVALIDATE_ENG0_REQ); + hub->vm_inv_eng0_ack = + SOC15_REG_OFFSET(GC, 0, mmVM_INVALIDATE_ENG0_ACK); + hub->vm_context0_cntl = + SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_CNTL); + hub->vm_l2_pro_fault_status = + SOC15_REG_OFFSET(GC, 0, mmVM_L2_PROTECTION_FAULT_STATUS); + hub->vm_l2_pro_fault_cntl = + SOC15_REG_OFFSET(GC, 0, mmVM_L2_PROTECTION_FAULT_CNTL); + + hub->get_invalidate_req = gfxhub_v1_0_get_invalidate_req; + hub->get_vm_protection_bits = gfxhub_v1_0_get_vm_protection_bits; + + return 0; +} + +static int gfxhub_v1_0_sw_fini(void *handle) +{ + return 0; +} + +static int gfxhub_v1_0_hw_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + unsigned i; + + for (i = 0 ; i < 18; ++i) { + WREG32(SOC15_REG_OFFSET(GC, 0, + mmVM_INVALIDATE_ENG0_ADDR_RANGE_LO32) + + 2 * i, 0xffffffff); + WREG32(SOC15_REG_OFFSET(GC, 0, + mmVM_INVALIDATE_ENG0_ADDR_RANGE_HI32) + + 2 * i, 0x1f); + } + + return 0; +} + +static int gfxhub_v1_0_hw_fini(void *handle) +{ + return 0; +} + +static int gfxhub_v1_0_suspend(void *handle) +{ + return 0; +} + +static int gfxhub_v1_0_resume(void *handle) +{ + return 0; +} + +static bool gfxhub_v1_0_is_idle(void *handle) +{ + return true; +} + +static int gfxhub_v1_0_wait_for_idle(void *handle) +{ + return 0; +} + +static int gfxhub_v1_0_soft_reset(void *handle) +{ + return 0; +} + +static int gfxhub_v1_0_set_clockgating_state(void *handle, + enum amd_clockgating_state state) +{ + return 0; +} + +static int gfxhub_v1_0_set_powergating_state(void *handle, + enum amd_powergating_state state) +{ + return 0; +} + +const struct amd_ip_funcs gfxhub_v1_0_ip_funcs = { + .name = "gfxhub_v1_0", + .early_init = gfxhub_v1_0_early_init, + .late_init = gfxhub_v1_0_late_init, + .sw_init = gfxhub_v1_0_sw_init, + .sw_fini = gfxhub_v1_0_sw_fini, + .hw_init = gfxhub_v1_0_hw_init, + .hw_fini = gfxhub_v1_0_hw_fini, + .suspend = gfxhub_v1_0_suspend, + .resume = gfxhub_v1_0_resume, + .is_idle = gfxhub_v1_0_is_idle, + .wait_for_idle = gfxhub_v1_0_wait_for_idle, + .soft_reset = gfxhub_v1_0_soft_reset, + .set_clockgating_state = gfxhub_v1_0_set_clockgating_state, + .set_powergating_state = gfxhub_v1_0_set_powergating_state, +}; + +const struct amdgpu_ip_block_version gfxhub_v1_0_ip_block = +{ + .type = AMD_IP_BLOCK_TYPE_GFXHUB, + .major = 1, + .minor = 0, + .rev = 0, + .funcs = &gfxhub_v1_0_ip_funcs, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.h b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.h new file mode 100644 index 0000000..5129a8f --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.h @@ -0,0 +1,35 @@ +/* + * Copyright 2016 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 __GFXHUB_V1_0_H__ +#define __GFXHUB_V1_0_H__ + +int gfxhub_v1_0_gart_enable(struct amdgpu_device *adev); +void gfxhub_v1_0_gart_disable(struct amdgpu_device *adev); +void gfxhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev, + bool value); + +extern const struct amd_ip_funcs gfxhub_v1_0_ip_funcs; +extern const struct amdgpu_ip_block_version gfxhub_v1_0_ip_block; + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c new file mode 100644 index 0000000..c069db7 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -0,0 +1,825 @@ +/* + * Copyright 2016 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. + * + */ +#include +#include "amdgpu.h" +#include "gmc_v9_0.h" + +#include "vega10/soc15ip.h" +#include "vega10/HDP/hdp_4_0_offset.h" +#include "vega10/HDP/hdp_4_0_sh_mask.h" +#include "vega10/GC/gc_9_0_sh_mask.h" +#include "vega10/vega10_enum.h" + +#include "soc15_common.h" + +#include "nbio_v6_1.h" +#include "gfxhub_v1_0.h" +#include "mmhub_v1_0.h" + +#define mmDF_CS_AON0_DramBaseAddress0 0x0044 +#define mmDF_CS_AON0_DramBaseAddress0_BASE_IDX 0 +//DF_CS_AON0_DramBaseAddress0 +#define DF_CS_AON0_DramBaseAddress0__AddrRngVal__SHIFT 0x0 +#define DF_CS_AON0_DramBaseAddress0__LgcyMmioHoleEn__SHIFT 0x1 +#define DF_CS_AON0_DramBaseAddress0__IntLvNumChan__SHIFT 0x4 +#define DF_CS_AON0_DramBaseAddress0__IntLvAddrSel__SHIFT 0x8 +#define DF_CS_AON0_DramBaseAddress0__DramBaseAddr__SHIFT 0xc +#define DF_CS_AON0_DramBaseAddress0__AddrRngVal_MASK 0x00000001L +#define DF_CS_AON0_DramBaseAddress0__LgcyMmioHoleEn_MASK 0x00000002L +#define DF_CS_AON0_DramBaseAddress0__IntLvNumChan_MASK 0x000000F0L +#define DF_CS_AON0_DramBaseAddress0__IntLvAddrSel_MASK 0x00000700L +#define DF_CS_AON0_DramBaseAddress0__DramBaseAddr_MASK 0xFFFFF000L + +/* XXX Move this macro to VEGA10 header file, which is like vid.h for VI.*/ +#define AMDGPU_NUM_OF_VMIDS 8 + +static const u32 golden_settings_vega10_hdp[] = +{ + 0xf64, 0x0fffffff, 0x00000000, + 0xf65, 0x0fffffff, 0x00000000, + 0xf66, 0x0fffffff, 0x00000000, + 0xf67, 0x0fffffff, 0x00000000, + 0xf68, 0x0fffffff, 0x00000000, + 0xf6a, 0x0fffffff, 0x00000000, + 0xf6b, 0x0fffffff, 0x00000000, + 0xf6c, 0x0fffffff, 0x00000000, + 0xf6d, 0x0fffffff, 0x00000000, + 0xf6e, 0x0fffffff, 0x00000000, +}; + +static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *src, + unsigned type, + enum amdgpu_interrupt_state state) +{ + struct amdgpu_vmhub *hub; + u32 tmp, reg, bits, i; + + switch (state) { + case AMDGPU_IRQ_STATE_DISABLE: + /* MM HUB */ + hub = &adev->vmhub[AMDGPU_MMHUB]; + bits = hub->get_vm_protection_bits(); + for (i = 0; i< 16; i++) { + reg = hub->vm_context0_cntl + i; + tmp = RREG32(reg); + tmp &= ~bits; + WREG32(reg, tmp); + } + + /* GFX HUB */ + hub = &adev->vmhub[AMDGPU_GFXHUB]; + bits = hub->get_vm_protection_bits(); + for (i = 0; i < 16; i++) { + reg = hub->vm_context0_cntl + i; + tmp = RREG32(reg); + tmp &= ~bits; + WREG32(reg, tmp); + } + break; + case AMDGPU_IRQ_STATE_ENABLE: + /* MM HUB */ + hub = &adev->vmhub[AMDGPU_MMHUB]; + bits = hub->get_vm_protection_bits(); + for (i = 0; i< 16; i++) { + reg = hub->vm_context0_cntl + i; + tmp = RREG32(reg); + tmp |= bits; + WREG32(reg, tmp); + } + + /* GFX HUB */ + hub = &adev->vmhub[AMDGPU_GFXHUB]; + bits = hub->get_vm_protection_bits(); + for (i = 0; i < 16; i++) { + reg = hub->vm_context0_cntl + i; + tmp = RREG32(reg); + tmp |= bits; + WREG32(reg, tmp); + } + break; + default: + break; + } + + return 0; +} + +static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + struct amdgpu_vmhub *gfxhub = &adev->vmhub[AMDGPU_GFXHUB]; + struct amdgpu_vmhub *mmhub = &adev->vmhub[AMDGPU_MMHUB]; + uint32_t status; + u64 addr; + + addr = (u64)entry->src_data[0] << 12; + addr |= ((u64)entry->src_data[1] & 0xf) << 44; + + if (entry->vm_id_src) { + status = RREG32(mmhub->vm_l2_pro_fault_status); + WREG32_P(mmhub->vm_l2_pro_fault_cntl, 1, ~1); + } else { + status = RREG32(gfxhub->vm_l2_pro_fault_status); + WREG32_P(gfxhub->vm_l2_pro_fault_cntl, 1, ~1); + } + + DRM_ERROR("[%s]VMC page fault (src_id:%u ring:%u vm_id:%u pas_id:%u) " + "at page 0x%016llx from %d\n" + "VM_L2_PROTECTION_FAULT_STATUS:0x%08X\n", + entry->vm_id_src ? "mmhub" : "gfxhub", + entry->src_id, entry->ring_id, entry->vm_id, entry->pas_id, + addr, entry->client_id, status); + + return 0; +} + +static const struct amdgpu_irq_src_funcs gmc_v9_0_irq_funcs = { + .set = gmc_v9_0_vm_fault_interrupt_state, + .process = gmc_v9_0_process_interrupt, +}; + +static void gmc_v9_0_set_irq_funcs(struct amdgpu_device *adev) +{ + adev->mc.vm_fault.num_types = 1; + adev->mc.vm_fault.funcs = &gmc_v9_0_irq_funcs; +} + +/* + * GART + * VMID 0 is the physical GPU addresses as used by the kernel. + * VMIDs 1-15 are used for userspace clients and are handled + * by the amdgpu vm/hsa code. + */ + +/** + * gmc_v9_0_gart_flush_gpu_tlb - gart tlb flush callback + * + * @adev: amdgpu_device pointer + * @vmid: vm instance to flush + * + * Flush the TLB for the requested page table. + */ +static void gmc_v9_0_gart_flush_gpu_tlb(struct amdgpu_device *adev, + uint32_t vmid) +{ + /* Use register 17 for GART */ + const unsigned eng = 17; + unsigned i, j; + + /* flush hdp cache */ + nbio_v6_1_hdp_flush(adev); + + spin_lock(&adev->mc.invalidate_lock); + + for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) { + struct amdgpu_vmhub *hub = &adev->vmhub[i]; + u32 tmp = hub->get_invalidate_req(vmid); + + WREG32(hub->vm_inv_eng0_req + eng, tmp); + + /* Busy wait for ACK.*/ + for (j = 0; j < 100; j++) { + tmp = RREG32(hub->vm_inv_eng0_ack + eng); + tmp &= 1 << vmid; + if (tmp) + break; + cpu_relax(); + } + if (j < 100) + continue; + + /* Wait for ACK with a delay.*/ + for (j = 0; j < adev->usec_timeout; j++) { + tmp = RREG32(hub->vm_inv_eng0_ack + eng); + tmp &= 1 << vmid; + if (tmp) + break; + udelay(1); + } + if (j < adev->usec_timeout) + continue; + + DRM_ERROR("Timeout waiting for VM flush ACK!\n"); + } + + spin_unlock(&adev->mc.invalidate_lock); +} + +/** + * gmc_v9_0_gart_set_pte_pde - update the page tables using MMIO + * + * @adev: amdgpu_device pointer + * @cpu_pt_addr: cpu address of the page table + * @gpu_page_idx: entry in the page table to update + * @addr: dst addr to write into pte/pde + * @flags: access flags + * + * Update the page tables using the CPU. + */ +static int gmc_v9_0_gart_set_pte_pde(struct amdgpu_device *adev, + void *cpu_pt_addr, + uint32_t gpu_page_idx, + uint64_t addr, + uint64_t flags) +{ + void __iomem *ptr = (void *)cpu_pt_addr; + uint64_t value; + + /* + * PTE format on VEGA 10: + * 63:59 reserved + * 58:57 mtype + * 56 F + * 55 L + * 54 P + * 53 SW + * 52 T + * 50:48 reserved + * 47:12 4k physical page base address + * 11:7 fragment + * 6 write + * 5 read + * 4 exe + * 3 Z + * 2 snooped + * 1 system + * 0 valid + * + * PDE format on VEGA 10: + * 63:59 block fragment size + * 58:55 reserved + * 54 P + * 53:48 reserved + * 47:6 physical base address of PD or PTE + * 5:3 reserved + * 2 C + * 1 system + * 0 valid + */ + + /* + * The following is for PTE only. GART does not have PDEs. + */ + value = addr & 0x0000FFFFFFFFF000ULL; + value |= flags; + writeq(value, ptr + (gpu_page_idx * 8)); + return 0; +} + +static uint64_t gmc_v9_0_get_vm_pte_flags(struct amdgpu_device *adev, + uint32_t flags) + +{ + uint64_t pte_flag = 0; + + if (flags & AMDGPU_VM_PAGE_EXECUTABLE) + pte_flag |= AMDGPU_PTE_EXECUTABLE; + if (flags & AMDGPU_VM_PAGE_READABLE) + pte_flag |= AMDGPU_PTE_READABLE; + if (flags & AMDGPU_VM_PAGE_WRITEABLE) + pte_flag |= AMDGPU_PTE_WRITEABLE; + + switch (flags & AMDGPU_VM_MTYPE_MASK) { + case AMDGPU_VM_MTYPE_DEFAULT: + pte_flag |= AMDGPU_PTE_MTYPE(MTYPE_NC); + break; + case AMDGPU_VM_MTYPE_NC: + pte_flag |= AMDGPU_PTE_MTYPE(MTYPE_NC); + break; + case AMDGPU_VM_MTYPE_WC: + pte_flag |= AMDGPU_PTE_MTYPE(MTYPE_WC); + break; + case AMDGPU_VM_MTYPE_CC: + pte_flag |= AMDGPU_PTE_MTYPE(MTYPE_CC); + break; + case AMDGPU_VM_MTYPE_UC: + pte_flag |= AMDGPU_PTE_MTYPE(MTYPE_UC); + break; + default: + pte_flag |= AMDGPU_PTE_MTYPE(MTYPE_NC); + break; + } + + if (flags & AMDGPU_VM_PAGE_PRT) + pte_flag |= AMDGPU_PTE_PRT; + + return pte_flag; +} + +static const struct amdgpu_gart_funcs gmc_v9_0_gart_funcs = { + .flush_gpu_tlb = gmc_v9_0_gart_flush_gpu_tlb, + .set_pte_pde = gmc_v9_0_gart_set_pte_pde, + .get_vm_pte_flags = gmc_v9_0_get_vm_pte_flags +}; + +static void gmc_v9_0_set_gart_funcs(struct amdgpu_device *adev) +{ + if (adev->gart.gart_funcs == NULL) + adev->gart.gart_funcs = &gmc_v9_0_gart_funcs; +} + +static u64 gmc_v9_0_adjust_mc_addr(struct amdgpu_device *adev, u64 mc_addr) +{ + return adev->vm_manager.vram_base_offset + mc_addr - adev->mc.vram_start; +} + +static const struct amdgpu_mc_funcs gmc_v9_0_mc_funcs = { + .adjust_mc_addr = gmc_v9_0_adjust_mc_addr, +}; + +static void gmc_v9_0_set_mc_funcs(struct amdgpu_device *adev) +{ + adev->mc.mc_funcs = &gmc_v9_0_mc_funcs; +} + +static int gmc_v9_0_early_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + gmc_v9_0_set_gart_funcs(adev); + gmc_v9_0_set_mc_funcs(adev); + gmc_v9_0_set_irq_funcs(adev); + + return 0; +} + +static int gmc_v9_0_late_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + return amdgpu_irq_get(adev, &adev->mc.vm_fault, 0); +} + +static void gmc_v9_0_vram_gtt_location(struct amdgpu_device *adev, + struct amdgpu_mc *mc) +{ + u64 base = mmhub_v1_0_get_fb_location(adev); + amdgpu_vram_location(adev, &adev->mc, base); + adev->mc.gtt_base_align = 0; + amdgpu_gtt_location(adev, mc); +} + +/** + * gmc_v9_0_mc_init - initialize the memory controller driver params + * + * @adev: amdgpu_device pointer + * + * Look up the amount of vram, vram width, and decide how to place + * vram and gart within the GPU's physical address space. + * Returns 0 for success. + */ +static int gmc_v9_0_mc_init(struct amdgpu_device *adev) +{ + u32 tmp; + int chansize, numchan; + + /* hbm memory channel size */ + chansize = 128; + + tmp = RREG32(SOC15_REG_OFFSET(DF, 0, mmDF_CS_AON0_DramBaseAddress0)); + tmp &= DF_CS_AON0_DramBaseAddress0__IntLvNumChan_MASK; + tmp >>= DF_CS_AON0_DramBaseAddress0__IntLvNumChan__SHIFT; + switch (tmp) { + case 0: + default: + numchan = 1; + break; + case 1: + numchan = 2; + break; + case 2: + numchan = 0; + break; + case 3: + numchan = 4; + break; + case 4: + numchan = 0; + break; + case 5: + numchan = 8; + break; + case 6: + numchan = 0; + break; + case 7: + numchan = 16; + break; + case 8: + numchan = 2; + break; + } + adev->mc.vram_width = numchan * chansize; + + /* Could aper size report 0 ? */ + adev->mc.aper_base = pci_resource_start(adev->pdev, 0); + adev->mc.aper_size = pci_resource_len(adev->pdev, 0); + /* size in MB on si */ + adev->mc.mc_vram_size = + nbio_v6_1_get_memsize(adev) * 1024ULL * 1024ULL; + adev->mc.real_vram_size = adev->mc.mc_vram_size; + adev->mc.visible_vram_size = adev->mc.aper_size; + + /* In case the PCI BAR is larger than the actual amount of vram */ + if (adev->mc.visible_vram_size > adev->mc.real_vram_size) + adev->mc.visible_vram_size = adev->mc.real_vram_size; + + /* unless the user had overridden it, set the gart + * size equal to the 1024 or vram, whichever is larger. + */ + if (amdgpu_gart_size == -1) + adev->mc.gtt_size = max((1024ULL << 20), adev->mc.mc_vram_size); + else + adev->mc.gtt_size = (uint64_t)amdgpu_gart_size << 20; + + gmc_v9_0_vram_gtt_location(adev, &adev->mc); + + return 0; +} + +static int gmc_v9_0_gart_init(struct amdgpu_device *adev) +{ + int r; + + if (adev->gart.robj) { + WARN(1, "VEGA10 PCIE GART already initialized\n"); + return 0; + } + /* Initialize common gart structure */ + r = amdgpu_gart_init(adev); + if (r) + return r; + adev->gart.table_size = adev->gart.num_gpu_pages * 8; + adev->gart.gart_pte_flags = AMDGPU_PTE_MTYPE(MTYPE_UC) | + AMDGPU_PTE_EXECUTABLE; + return amdgpu_gart_table_vram_alloc(adev); +} + +/* + * vm + * VMID 0 is the physical GPU addresses as used by the kernel. + * VMIDs 1-15 are used for userspace clients and are handled + * by the amdgpu vm/hsa code. + */ +/** + * gmc_v9_0_vm_init - vm init callback + * + * @adev: amdgpu_device pointer + * + * Inits vega10 specific vm parameters (number of VMs, base of vram for + * VMIDs 1-15) (vega10). + * Returns 0 for success. + */ +static int gmc_v9_0_vm_init(struct amdgpu_device *adev) +{ + /* + * number of VMs + * VMID 0 is reserved for System + * amdgpu graphics/compute will use VMIDs 1-7 + * amdkfd will use VMIDs 8-15 + */ + adev->vm_manager.num_ids = AMDGPU_NUM_OF_VMIDS; + amdgpu_vm_manager_init(adev); + + /* base offset of vram pages */ + /*XXX This value is not zero for APU*/ + adev->vm_manager.vram_base_offset = 0; + + return 0; +} + +/** + * gmc_v9_0_vm_fini - vm fini callback + * + * @adev: amdgpu_device pointer + * + * Tear down any asic specific VM setup. + */ +static void gmc_v9_0_vm_fini(struct amdgpu_device *adev) +{ + return; +} + +static int gmc_v9_0_sw_init(void *handle) +{ + int r; + int dma_bits; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + spin_lock_init(&adev->mc.invalidate_lock); + + if (adev->flags & AMD_IS_APU) { + adev->mc.vram_type = AMDGPU_VRAM_TYPE_UNKNOWN; + } else { + /* XXX Don't know how to get VRAM type yet. */ + adev->mc.vram_type = AMDGPU_VRAM_TYPE_HBM; + } + + /* This interrupt is VMC page fault.*/ + r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_VMC, 0, + &adev->mc.vm_fault); + + if (r) + return r; + + /* Adjust VM size here. + * Currently default to 64GB ((16 << 20) 4k pages). + * Max GPUVM size is 48 bits. + */ + adev->vm_manager.max_pfn = amdgpu_vm_size << 18; + + /* Set the internal MC address mask + * This is the max address of the GPU's + * internal address space. + */ + adev->mc.mc_mask = 0xffffffffffffULL; /* 48 bit MC */ + + /* set DMA mask + need_dma32 flags. + * PCIE - can handle 44-bits. + * IGP - can handle 44-bits + * PCI - dma32 for legacy pci gart, 44 bits on vega10 + */ + adev->need_dma32 = false; + dma_bits = adev->need_dma32 ? 32 : 44; + r = pci_set_dma_mask(adev->pdev, DMA_BIT_MASK(dma_bits)); + if (r) { + adev->need_dma32 = true; + dma_bits = 32; + printk(KERN_WARNING "amdgpu: No suitable DMA available.\n"); + } + r = pci_set_consistent_dma_mask(adev->pdev, DMA_BIT_MASK(dma_bits)); + if (r) { + pci_set_consistent_dma_mask(adev->pdev, DMA_BIT_MASK(32)); + printk(KERN_WARNING "amdgpu: No coherent DMA available.\n"); + } + + r = gmc_v9_0_mc_init(adev); + if (r) + return r; + + /* Memory manager */ + r = amdgpu_bo_init(adev); + if (r) + return r; + + r = gmc_v9_0_gart_init(adev); + if (r) + return r; + + if (!adev->vm_manager.enabled) { + r = gmc_v9_0_vm_init(adev); + if (r) { + dev_err(adev->dev, "vm manager initialization failed (%d).\n", r); + return r; + } + adev->vm_manager.enabled = true; + } + return r; +} + +/** + * gmc_v8_0_gart_fini - vm fini callback + * + * @adev: amdgpu_device pointer + * + * Tears down the driver GART/VM setup (CIK). + */ +static void gmc_v9_0_gart_fini(struct amdgpu_device *adev) +{ + amdgpu_gart_table_vram_free(adev); + amdgpu_gart_fini(adev); +} + +static int gmc_v9_0_sw_fini(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (adev->vm_manager.enabled) { + amdgpu_vm_manager_fini(adev); + gmc_v9_0_vm_fini(adev); + adev->vm_manager.enabled = false; + } + gmc_v9_0_gart_fini(adev); + amdgpu_gem_force_release(adev); + amdgpu_bo_fini(adev); + + return 0; +} + +static void gmc_v9_0_init_golden_registers(struct amdgpu_device *adev) +{ + switch (adev->asic_type) { + case CHIP_VEGA10: + break; + default: + break; + } +} + +/** + * gmc_v9_0_gart_enable - gart enable + * + * @adev: amdgpu_device pointer + */ +static int gmc_v9_0_gart_enable(struct amdgpu_device *adev) +{ + int r; + bool value; + u32 tmp; + + amdgpu_program_register_sequence(adev, + golden_settings_vega10_hdp, + (const u32)ARRAY_SIZE(golden_settings_vega10_hdp)); + + if (adev->gart.robj == NULL) { + dev_err(adev->dev, "No VRAM object for PCIE GART.\n"); + return -EINVAL; + } + r = amdgpu_gart_table_vram_pin(adev); + if (r) + return r; + + /* After HDP is initialized, flush HDP.*/ + nbio_v6_1_hdp_flush(adev); + + r = gfxhub_v1_0_gart_enable(adev); + if (r) + return r; + + r = mmhub_v1_0_gart_enable(adev); + if (r) + return r; + + tmp = RREG32(SOC15_REG_OFFSET(HDP, 0, mmHDP_MISC_CNTL)); + tmp |= HDP_MISC_CNTL__FLUSH_INVALIDATE_CACHE_MASK; + WREG32(SOC15_REG_OFFSET(HDP, 0, mmHDP_MISC_CNTL), tmp); + + tmp = RREG32(SOC15_REG_OFFSET(HDP, 0, mmHDP_HOST_PATH_CNTL)); + WREG32(SOC15_REG_OFFSET(HDP, 0, mmHDP_HOST_PATH_CNTL), tmp); + + + if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS) + value = false; + else + value = true; + + gfxhub_v1_0_set_fault_enable_default(adev, value); + mmhub_v1_0_set_fault_enable_default(adev, value); + + gmc_v9_0_gart_flush_gpu_tlb(adev, 0); + + DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n", + (unsigned)(adev->mc.gtt_size >> 20), + (unsigned long long)adev->gart.table_addr); + adev->gart.ready = true; + return 0; +} + +static int gmc_v9_0_hw_init(void *handle) +{ + int r; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + /* The sequence of these two function calls matters.*/ + gmc_v9_0_init_golden_registers(adev); + + r = gmc_v9_0_gart_enable(adev); + + return r; +} + +/** + * gmc_v9_0_gart_disable - gart disable + * + * @adev: amdgpu_device pointer + * + * This disables all VM page table. + */ +static void gmc_v9_0_gart_disable(struct amdgpu_device *adev) +{ + gfxhub_v1_0_gart_disable(adev); + mmhub_v1_0_gart_disable(adev); + amdgpu_gart_table_vram_unpin(adev); +} + +static int gmc_v9_0_hw_fini(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + amdgpu_irq_put(adev, &adev->mc.vm_fault, 0); + gmc_v9_0_gart_disable(adev); + + return 0; +} + +static int gmc_v9_0_suspend(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (adev->vm_manager.enabled) { + gmc_v9_0_vm_fini(adev); + adev->vm_manager.enabled = false; + } + gmc_v9_0_hw_fini(adev); + + return 0; +} + +static int gmc_v9_0_resume(void *handle) +{ + int r; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + r = gmc_v9_0_hw_init(adev); + if (r) + return r; + + if (!adev->vm_manager.enabled) { + r = gmc_v9_0_vm_init(adev); + if (r) { + dev_err(adev->dev, + "vm manager initialization failed (%d).\n", r); + return r; + } + adev->vm_manager.enabled = true; + } + + return r; +} + +static bool gmc_v9_0_is_idle(void *handle) +{ + /* MC is always ready in GMC v9.*/ + return true; +} + +static int gmc_v9_0_wait_for_idle(void *handle) +{ + /* There is no need to wait for MC idle in GMC v9.*/ + return 0; +} + +static int gmc_v9_0_soft_reset(void *handle) +{ + /* XXX for emulation.*/ + return 0; +} + +static int gmc_v9_0_set_clockgating_state(void *handle, + enum amd_clockgating_state state) +{ + return 0; +} + +static int gmc_v9_0_set_powergating_state(void *handle, + enum amd_powergating_state state) +{ + return 0; +} + +const struct amd_ip_funcs gmc_v9_0_ip_funcs = { + .name = "gmc_v9_0", + .early_init = gmc_v9_0_early_init, + .late_init = gmc_v9_0_late_init, + .sw_init = gmc_v9_0_sw_init, + .sw_fini = gmc_v9_0_sw_fini, + .hw_init = gmc_v9_0_hw_init, + .hw_fini = gmc_v9_0_hw_fini, + .suspend = gmc_v9_0_suspend, + .resume = gmc_v9_0_resume, + .is_idle = gmc_v9_0_is_idle, + .wait_for_idle = gmc_v9_0_wait_for_idle, + .soft_reset = gmc_v9_0_soft_reset, + .set_clockgating_state = gmc_v9_0_set_clockgating_state, + .set_powergating_state = gmc_v9_0_set_powergating_state, +}; + +const struct amdgpu_ip_block_version gmc_v9_0_ip_block = +{ + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 9, + .minor = 0, + .rev = 0, + .funcs = &gmc_v9_0_ip_funcs, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.h b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.h new file mode 100644 index 0000000..b030ca5 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.h @@ -0,0 +1,30 @@ +/* + * Copyright 2016 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 __GMC_V9_0_H__ +#define __GMC_V9_0_H__ + +extern const struct amd_ip_funcs gmc_v9_0_ip_funcs; +extern const struct amdgpu_ip_block_version gmc_v9_0_ip_block; + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c new file mode 100644 index 0000000..b1e0e6b --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c @@ -0,0 +1,585 @@ +/* + * Copyright 2016 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. + * + */ +#include "amdgpu.h" +#include "mmhub_v1_0.h" + +#include "vega10/soc15ip.h" +#include "vega10/MMHUB/mmhub_1_0_offset.h" +#include "vega10/MMHUB/mmhub_1_0_sh_mask.h" +#include "vega10/MMHUB/mmhub_1_0_default.h" +#include "vega10/ATHUB/athub_1_0_offset.h" +#include "vega10/ATHUB/athub_1_0_sh_mask.h" +#include "vega10/ATHUB/athub_1_0_default.h" +#include "vega10/vega10_enum.h" + +#include "soc15_common.h" + +u64 mmhub_v1_0_get_fb_location(struct amdgpu_device *adev) +{ + u64 base = RREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMC_VM_FB_LOCATION_BASE)); + + base &= MC_VM_FB_LOCATION_BASE__FB_BASE_MASK; + base <<= 24; + + return base; +} + +int mmhub_v1_0_gart_enable(struct amdgpu_device *adev) +{ + u32 tmp; + u64 value; + uint64_t addr; + u32 i; + + /* Program MC. */ + /* Update configuration */ + DRM_INFO("%s -- in\n", __func__); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMC_VM_SYSTEM_APERTURE_LOW_ADDR), + adev->mc.vram_start >> 18); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR), + adev->mc.vram_end >> 18); + value = adev->vram_scratch.gpu_addr - adev->mc.vram_start + + adev->vm_manager.vram_base_offset; + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB), + (u32)(value >> 12)); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB), + (u32)(value >> 44)); + + /* Disable AGP. */ + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMC_VM_AGP_BASE), 0); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMC_VM_AGP_TOP), 0); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMC_VM_AGP_BOT), 0x00FFFFFF); + + /* GART Enable. */ + + /* Setup TLB control */ + tmp = RREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMC_VM_MX_L1_TLB_CNTL)); + tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1); + tmp = REG_SET_FIELD(tmp, + MC_VM_MX_L1_TLB_CNTL, + SYSTEM_ACCESS_MODE, + 3); + tmp = REG_SET_FIELD(tmp, + MC_VM_MX_L1_TLB_CNTL, + ENABLE_ADVANCED_DRIVER_MODEL, + 1); + tmp = REG_SET_FIELD(tmp, + MC_VM_MX_L1_TLB_CNTL, + SYSTEM_APERTURE_UNMAPPED_ACCESS, + 0); + tmp = REG_SET_FIELD(tmp, + MC_VM_MX_L1_TLB_CNTL, + ECO_BITS, + 0); + tmp = REG_SET_FIELD(tmp, + MC_VM_MX_L1_TLB_CNTL, + MTYPE, + MTYPE_UC);/* XXX for emulation. */ + tmp = REG_SET_FIELD(tmp, + MC_VM_MX_L1_TLB_CNTL, + ATC_EN, + 1); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMC_VM_MX_L1_TLB_CNTL), tmp); + + /* Setup L2 cache */ + tmp = RREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_L2_CNTL)); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 1); + tmp = REG_SET_FIELD(tmp, + VM_L2_CNTL, + ENABLE_L2_FRAGMENT_PROCESSING, + 0); + tmp = REG_SET_FIELD(tmp, + VM_L2_CNTL, + L2_PDE0_CACHE_TAG_GENERATION_MODE, + 0);/* XXX for emulation, Refer to closed source code.*/ + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 1); + tmp = REG_SET_FIELD(tmp, + VM_L2_CNTL, + CONTEXT1_IDENTITY_ACCESS_MODE, + 1); + tmp = REG_SET_FIELD(tmp, + VM_L2_CNTL, + IDENTITY_MODE_FRAGMENT_SIZE, + 0); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_L2_CNTL), tmp); + + tmp = RREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_L2_CNTL2)); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_L2_CACHE, 1); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_L2_CNTL2), tmp); + + tmp = mmVM_L2_CNTL3_DEFAULT; + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_L2_CNTL3), tmp); + + tmp = RREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_L2_CNTL4)); + tmp = REG_SET_FIELD(tmp, + VM_L2_CNTL4, + VMC_TAP_PDE_REQUEST_PHYSICAL, + 0); + tmp = REG_SET_FIELD(tmp, + VM_L2_CNTL4, + VMC_TAP_PTE_REQUEST_PHYSICAL, + 0); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_L2_CNTL4), tmp); + + /* setup context0 */ + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32), + (u32)(adev->mc.gtt_start >> 12)); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32), + (u32)(adev->mc.gtt_start >> 44)); + + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32), + (u32)(adev->mc.gtt_end >> 12)); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32), + (u32)(adev->mc.gtt_end >> 44)); + + BUG_ON(adev->gart.table_addr & (~0x0000FFFFFFFFF000ULL)); + value = adev->gart.table_addr - adev->mc.vram_start + + adev->vm_manager.vram_base_offset; + value &= 0x0000FFFFFFFFF000ULL; + value |= 0x1; /* valid bit */ + + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32), + (u32)value); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32), + (u32)(value >> 32)); + + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32), + (u32)(adev->dummy_page.addr >> 12)); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32), + (u32)(adev->dummy_page.addr >> 44)); + + tmp = RREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_L2_PROTECTION_FAULT_CNTL2)); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL2, + ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, + 1); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_L2_PROTECTION_FAULT_CNTL2), tmp); + + addr = SOC15_REG_OFFSET(MMHUB, 0, mmVM_CONTEXT0_CNTL); + tmp = RREG32(addr); + + tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH, 0); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_CONTEXT0_CNTL), tmp); + + tmp = RREG32(addr); + + /* Disable identity aperture.*/ + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32), 0XFFFFFFFF); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32), 0x0000000F); + + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32), 0); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32), 0); + + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32), 0); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32), 0); + + for (i = 0; i <= 14; i++) { + tmp = RREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_CONTEXT1_CNTL) + + i); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + ENABLE_CONTEXT, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + PAGE_TABLE_DEPTH, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + PAGE_TABLE_BLOCK_SIZE, + amdgpu_vm_block_size - 9); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_CONTEXT1_CNTL) + i, tmp); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32) + i*2, 0); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32) + i*2, 0); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32) + i*2, + adev->vm_manager.max_pfn - 1); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32) + i*2, 0); + } + + return 0; +} + +void mmhub_v1_0_gart_disable(struct amdgpu_device *adev) +{ + u32 tmp; + u32 i; + + /* Disable all tables */ + for (i = 0; i < 16; i++) + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_CONTEXT0_CNTL) + i, 0); + + /* Setup TLB control */ + tmp = RREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMC_VM_MX_L1_TLB_CNTL)); + tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0); + tmp = REG_SET_FIELD(tmp, + MC_VM_MX_L1_TLB_CNTL, + ENABLE_ADVANCED_DRIVER_MODEL, + 0); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMC_VM_MX_L1_TLB_CNTL), tmp); + + /* Setup L2 cache */ + tmp = RREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_L2_CNTL)); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 0); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_L2_CNTL), tmp); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_L2_CNTL3), 0); +} + +/** + * mmhub_v1_0_set_fault_enable_default - update GART/VM fault handling + * + * @adev: amdgpu_device pointer + * @value: true redirects VM faults to the default page + */ +void mmhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev, bool value) +{ + u32 tmp; + tmp = RREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_L2_PROTECTION_FAULT_CNTL)); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, + VM_L2_PROTECTION_FAULT_CNTL, + TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT, + value); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + READ_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmVM_L2_PROTECTION_FAULT_CNTL), tmp); +} + +static uint32_t mmhub_v1_0_get_invalidate_req(unsigned int vm_id) +{ + u32 req = 0; + + /* invalidate using legacy mode on vm_id*/ + req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, + PER_VMID_INVALIDATE_REQ, 1 << vm_id); + req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, 0); + req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PTES, 1); + req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE0, 1); + req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE1, 1); + req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE2, 1); + req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, INVALIDATE_L1_PTES, 1); + req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, + CLEAR_PROTECTION_FAULT_STATUS_ADDR, 0); + + return req; +} + +static uint32_t mmhub_v1_0_get_vm_protection_bits(void) +{ + return (VM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + VM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + VM_CONTEXT1_CNTL__PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + VM_CONTEXT1_CNTL__VALID_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + VM_CONTEXT1_CNTL__READ_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + VM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | + VM_CONTEXT1_CNTL__EXECUTE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK); +} + +static int mmhub_v1_0_early_init(void *handle) +{ + return 0; +} + +static int mmhub_v1_0_late_init(void *handle) +{ + return 0; +} + +static int mmhub_v1_0_sw_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB]; + + hub->ctx0_ptb_addr_lo32 = + SOC15_REG_OFFSET(MMHUB, 0, + mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32); + hub->ctx0_ptb_addr_hi32 = + SOC15_REG_OFFSET(MMHUB, 0, + mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32); + hub->vm_inv_eng0_req = + SOC15_REG_OFFSET(MMHUB, 0, mmVM_INVALIDATE_ENG0_REQ); + hub->vm_inv_eng0_ack = + SOC15_REG_OFFSET(MMHUB, 0, mmVM_INVALIDATE_ENG0_ACK); + hub->vm_context0_cntl = + SOC15_REG_OFFSET(MMHUB, 0, mmVM_CONTEXT0_CNTL); + hub->vm_l2_pro_fault_status = + SOC15_REG_OFFSET(MMHUB, 0, mmVM_L2_PROTECTION_FAULT_STATUS); + hub->vm_l2_pro_fault_cntl = + SOC15_REG_OFFSET(MMHUB, 0, mmVM_L2_PROTECTION_FAULT_CNTL); + + hub->get_invalidate_req = mmhub_v1_0_get_invalidate_req; + hub->get_vm_protection_bits = mmhub_v1_0_get_vm_protection_bits; + + return 0; +} + +static int mmhub_v1_0_sw_fini(void *handle) +{ + return 0; +} + +static int mmhub_v1_0_hw_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + unsigned i; + + for (i = 0; i < 18; ++i) { + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmVM_INVALIDATE_ENG0_ADDR_RANGE_LO32) + + 2 * i, 0xffffffff); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmVM_INVALIDATE_ENG0_ADDR_RANGE_HI32) + + 2 * i, 0x1f); + } + + return 0; +} + +static int mmhub_v1_0_hw_fini(void *handle) +{ + return 0; +} + +static int mmhub_v1_0_suspend(void *handle) +{ + return 0; +} + +static int mmhub_v1_0_resume(void *handle) +{ + return 0; +} + +static bool mmhub_v1_0_is_idle(void *handle) +{ + return true; +} + +static int mmhub_v1_0_wait_for_idle(void *handle) +{ + return 0; +} + +static int mmhub_v1_0_soft_reset(void *handle) +{ + return 0; +} + +static void mmhub_v1_0_update_medium_grain_clock_gating(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data, def1, data1, def2, data2; + + def = data = RREG32(SOC15_REG_OFFSET(MMHUB, 0, mmATC_L2_MISC_CG)); + def1 = data1 = RREG32(SOC15_REG_OFFSET(MMHUB, 0, mmDAGB0_CNTL_MISC2)); + def2 = data2 = RREG32(SOC15_REG_OFFSET(MMHUB, 0, mmDAGB1_CNTL_MISC2)); + + if (enable && (adev->cg_flags & AMD_CG_SUPPORT_MC_MGCG)) { + data |= ATC_L2_MISC_CG__ENABLE_MASK; + + data1 &= ~(DAGB0_CNTL_MISC2__DISABLE_WRREQ_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_WRRET_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_RDREQ_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_RDRET_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_TLBWR_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_TLBRD_CG_MASK); + + data2 &= ~(DAGB1_CNTL_MISC2__DISABLE_WRREQ_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_WRRET_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_RDREQ_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_RDRET_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_TLBWR_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_TLBRD_CG_MASK); + } else { + data &= ~ATC_L2_MISC_CG__ENABLE_MASK; + + data1 |= (DAGB0_CNTL_MISC2__DISABLE_WRREQ_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_WRRET_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_RDREQ_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_RDRET_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_TLBWR_CG_MASK | + DAGB0_CNTL_MISC2__DISABLE_TLBRD_CG_MASK); + + data2 |= (DAGB1_CNTL_MISC2__DISABLE_WRREQ_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_WRRET_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_RDREQ_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_RDRET_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_TLBWR_CG_MASK | + DAGB1_CNTL_MISC2__DISABLE_TLBRD_CG_MASK); + } + + if (def != data) + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmATC_L2_MISC_CG), data); + + if (def1 != data1) + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmDAGB0_CNTL_MISC2), data1); + + if (def2 != data2) + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmDAGB1_CNTL_MISC2), data2); +} + +static void athub_update_medium_grain_clock_gating(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data; + + def = data = RREG32(SOC15_REG_OFFSET(ATHUB, 0, mmATHUB_MISC_CNTL)); + + if (enable && (adev->cg_flags & AMD_CG_SUPPORT_MC_MGCG)) + data |= ATHUB_MISC_CNTL__CG_ENABLE_MASK; + else + data &= ~ATHUB_MISC_CNTL__CG_ENABLE_MASK; + + if (def != data) + WREG32(SOC15_REG_OFFSET(ATHUB, 0, mmATHUB_MISC_CNTL), data); +} + +static void mmhub_v1_0_update_medium_grain_light_sleep(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data; + + def = data = RREG32(SOC15_REG_OFFSET(MMHUB, 0, mmATC_L2_MISC_CG)); + + if (enable && (adev->cg_flags & AMD_CG_SUPPORT_MC_LS)) + data |= ATC_L2_MISC_CG__MEM_LS_ENABLE_MASK; + else + data &= ~ATC_L2_MISC_CG__MEM_LS_ENABLE_MASK; + + if (def != data) + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmATC_L2_MISC_CG), data); +} + +static void athub_update_medium_grain_light_sleep(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data; + + def = data = RREG32(SOC15_REG_OFFSET(ATHUB, 0, mmATHUB_MISC_CNTL)); + + if (enable && (adev->cg_flags & AMD_CG_SUPPORT_MC_LS) && + (adev->cg_flags & AMD_CG_SUPPORT_HDP_LS)) + data |= ATHUB_MISC_CNTL__CG_MEM_LS_ENABLE_MASK; + else + data &= ~ATHUB_MISC_CNTL__CG_MEM_LS_ENABLE_MASK; + + if(def != data) + WREG32(SOC15_REG_OFFSET(ATHUB, 0, mmATHUB_MISC_CNTL), data); +} + +static int mmhub_v1_0_set_clockgating_state(void *handle, + enum amd_clockgating_state state) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + switch (adev->asic_type) { + case CHIP_VEGA10: + mmhub_v1_0_update_medium_grain_clock_gating(adev, + state == AMD_CG_STATE_GATE ? true : false); + athub_update_medium_grain_clock_gating(adev, + state == AMD_CG_STATE_GATE ? true : false); + mmhub_v1_0_update_medium_grain_light_sleep(adev, + state == AMD_CG_STATE_GATE ? true : false); + athub_update_medium_grain_light_sleep(adev, + state == AMD_CG_STATE_GATE ? true : false); + break; + default: + break; + } + + return 0; +} + +static int mmhub_v1_0_set_powergating_state(void *handle, + enum amd_powergating_state state) +{ + return 0; +} + +const struct amd_ip_funcs mmhub_v1_0_ip_funcs = { + .name = "mmhub_v1_0", + .early_init = mmhub_v1_0_early_init, + .late_init = mmhub_v1_0_late_init, + .sw_init = mmhub_v1_0_sw_init, + .sw_fini = mmhub_v1_0_sw_fini, + .hw_init = mmhub_v1_0_hw_init, + .hw_fini = mmhub_v1_0_hw_fini, + .suspend = mmhub_v1_0_suspend, + .resume = mmhub_v1_0_resume, + .is_idle = mmhub_v1_0_is_idle, + .wait_for_idle = mmhub_v1_0_wait_for_idle, + .soft_reset = mmhub_v1_0_soft_reset, + .set_clockgating_state = mmhub_v1_0_set_clockgating_state, + .set_powergating_state = mmhub_v1_0_set_powergating_state, +}; + +const struct amdgpu_ip_block_version mmhub_v1_0_ip_block = +{ + .type = AMD_IP_BLOCK_TYPE_MMHUB, + .major = 1, + .minor = 0, + .rev = 0, + .funcs = &mmhub_v1_0_ip_funcs, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h new file mode 100644 index 0000000..aadedf9 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h @@ -0,0 +1,35 @@ +/* + * Copyright 2016 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 __MMHUB_V1_0_H__ +#define __MMHUB_V1_0_H__ + +u64 mmhub_v1_0_get_fb_location(struct amdgpu_device *adev); +int mmhub_v1_0_gart_enable(struct amdgpu_device *adev); +void mmhub_v1_0_gart_disable(struct amdgpu_device *adev); +void mmhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev, + bool value); + +extern const struct amd_ip_funcs mmhub_v1_0_ip_funcs; +extern const struct amdgpu_ip_block_version mmhub_v1_0_ip_block; + +#endif diff --git a/drivers/gpu/drm/amd/include/amd_shared.h b/drivers/gpu/drm/amd/include/amd_shared.h index 717d6be..a94420d 100644 --- a/drivers/gpu/drm/amd/include/amd_shared.h +++ b/drivers/gpu/drm/amd/include/amd_shared.h @@ -74,6 +74,8 @@ enum amd_ip_block_type { AMD_IP_BLOCK_TYPE_UVD, AMD_IP_BLOCK_TYPE_VCE, AMD_IP_BLOCK_TYPE_ACP, + AMD_IP_BLOCK_TYPE_GFXHUB, + AMD_IP_BLOCK_TYPE_MMHUB }; enum amd_clockgating_state { -- 2.7.4