From: Zhang, Jerry Date: Wed, 29 Mar 2017 08:08:32 +0000 (+0800) Subject: drm/amdgpu: fix vm size and block size for VMPT (v5) X-Git-Tag: v4.12-rc1~116^2~15^2~24 X-Git-Url: https://asedeno.scripts.mit.edu/gitweb/?a=commitdiff_plain;h=36b32a682bc32693e681cb984aac9c291a09c519;p=linux.git drm/amdgpu: fix vm size and block size for VMPT (v5) Set reasonable defaults per family. v2: set both of them in gmc v3: move vm size and block size in vm manager v4: squash in warning fix from Alex Xie v5: squash in min() warning fix Signed-off-by: Junwei Zhang Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index d78c523c93fe..831c2bfd2072 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -1042,14 +1042,6 @@ static bool amdgpu_check_pot_argument(int arg) static void amdgpu_get_block_size(struct amdgpu_device *adev) { - /* from AI, asic starts to support multiple level VMPT */ - if (adev->asic_type >= CHIP_VEGA10) { - if (amdgpu_vm_block_size != 9) - dev_warn(adev->dev, - "Multi-VMPT limits block size to one page!\n"); - amdgpu_vm_block_size = 9; - return; - } /* defines number of bits in page table versus page directory, * a page is 4KB so we have 12 bits offset, minimum 9 bits in the * page table and the remaining bits are in the page directory */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 14f0889f7563..2895d9d86f29 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -100,13 +100,14 @@ static unsigned amdgpu_vm_num_entries(struct amdgpu_device *adev, if (level == 0) /* For the root directory */ return adev->vm_manager.max_pfn >> - (amdgpu_vm_block_size * adev->vm_manager.num_level); + (adev->vm_manager.block_size * + adev->vm_manager.num_level); else if (level == adev->vm_manager.num_level) /* For the page tables on the leaves */ - return AMDGPU_VM_PTE_COUNT; + return AMDGPU_VM_PTE_COUNT(adev); else /* Everything in between */ - return 1 << amdgpu_vm_block_size; + return 1 << adev->vm_manager.block_size; } /** @@ -271,7 +272,7 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev, unsigned level) { unsigned shift = (adev->vm_manager.num_level - level) * - amdgpu_vm_block_size; + adev->vm_manager.block_size; unsigned pt_idx, from, to; int r; @@ -976,7 +977,7 @@ static struct amdgpu_bo *amdgpu_vm_get_pt(struct amdgpu_pte_update_params *p, unsigned idx, level = p->adev->vm_manager.num_level; while (entry->entries) { - idx = addr >> (amdgpu_vm_block_size * level--); + idx = addr >> (p->adev->vm_manager.block_size * level--); idx %= amdgpu_bo_size(entry->bo) / 8; entry = &entry->entries[idx]; } @@ -1003,7 +1004,8 @@ static void amdgpu_vm_update_ptes(struct amdgpu_pte_update_params *params, uint64_t start, uint64_t end, uint64_t dst, uint64_t flags) { - const uint64_t mask = AMDGPU_VM_PTE_COUNT - 1; + struct amdgpu_device *adev = params->adev; + const uint64_t mask = AMDGPU_VM_PTE_COUNT(adev) - 1; uint64_t cur_pe_start, cur_nptes, cur_dst; uint64_t addr; /* next GPU address to be updated */ @@ -1027,7 +1029,7 @@ static void amdgpu_vm_update_ptes(struct amdgpu_pte_update_params *params, if ((addr & ~mask) == (end & ~mask)) nptes = end - addr; else - nptes = AMDGPU_VM_PTE_COUNT - (addr & mask); + nptes = AMDGPU_VM_PTE_COUNT(adev) - (addr & mask); cur_pe_start = amdgpu_bo_gpu_offset(pt); cur_pe_start += (addr & mask) * 8; @@ -1055,7 +1057,7 @@ static void amdgpu_vm_update_ptes(struct amdgpu_pte_update_params *params, if ((addr & ~mask) == (end & ~mask)) nptes = end - addr; else - nptes = AMDGPU_VM_PTE_COUNT - (addr & mask); + nptes = AMDGPU_VM_PTE_COUNT(adev) - (addr & mask); next_pe_start = amdgpu_bo_gpu_offset(pt); next_pe_start += (addr & mask) * 8; @@ -1202,7 +1204,7 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, * reserve space for one command every (1 << BLOCK_SIZE) * entries or 2k dwords (whatever is smaller) */ - ncmds = (nptes >> min(amdgpu_vm_block_size, 11)) + 1; + ncmds = (nptes >> min(adev->vm_manager.block_size, 11u)) + 1; /* padding, etc. */ ndw = 64; @@ -2073,7 +2075,7 @@ void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev, int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm) { const unsigned align = min(AMDGPU_VM_PTB_ALIGN_SIZE, - AMDGPU_VM_PTE_COUNT * 8); + AMDGPU_VM_PTE_COUNT(adev) * 8); unsigned ring_instance; struct amdgpu_ring *ring; struct amd_sched_rq *rq; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index ed2467881b74..02b0dd3b135f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -45,7 +45,7 @@ struct amdgpu_bo_list_entry; #define AMDGPU_VM_MAX_UPDATE_SIZE 0x3FFFF /* number of entries in page table */ -#define AMDGPU_VM_PTE_COUNT (1 << amdgpu_vm_block_size) +#define AMDGPU_VM_PTE_COUNT(adev) (1 << (adev)->vm_manager.block_size) /* PTBs (Page Table Blocks) need to be aligned to 32K */ #define AMDGPU_VM_PTB_ALIGN_SIZE 32768 @@ -162,6 +162,8 @@ struct amdgpu_vm_manager { uint64_t max_pfn; uint32_t num_level; + uint64_t vm_size; + uint32_t block_size; /* vram base address for page table entry */ u64 vram_base_offset; /* is vm enabled? */ diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c index b808d4ce86d6..70c21f9b904b 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c @@ -222,7 +222,7 @@ int gfxhub_v1_0_gart_enable(struct amdgpu_device *adev) EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_BLOCK_SIZE, - amdgpu_vm_block_size - 9); + adev->vm_manager.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); diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c index d9586601a437..8f18d14f8eda 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c @@ -543,7 +543,8 @@ static int gmc_v6_0_gart_enable(struct amdgpu_device *adev) WREG32(mmVM_CONTEXT1_CNTL, VM_CONTEXT1_CNTL__ENABLE_CONTEXT_MASK | (1UL << VM_CONTEXT1_CNTL__PAGE_TABLE_DEPTH__SHIFT) | - ((amdgpu_vm_block_size - 9) << VM_CONTEXT1_CNTL__PAGE_TABLE_BLOCK_SIZE__SHIFT)); + ((adev->vm_manager.block_size - 9) + << VM_CONTEXT1_CNTL__PAGE_TABLE_BLOCK_SIZE__SHIFT)); if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS) gmc_v6_0_set_fault_enable_default(adev, false); else @@ -848,7 +849,12 @@ static int gmc_v6_0_sw_init(void *handle) if (r) return r; - adev->vm_manager.max_pfn = amdgpu_vm_size << 18; + adev->vm_manager.vm_size = amdgpu_vm_size; + adev->vm_manager.block_size = amdgpu_vm_block_size; + adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18; + + DRM_INFO("vm size is %llu GB, block size is %d-bit\n", + adev->vm_manager.vm_size, adev->vm_manager.block_size); adev->mc.mc_mask = 0xffffffffffULL; diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c index 78643a1baa5c..b86b454197f8 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c @@ -644,7 +644,7 @@ static int gmc_v7_0_gart_enable(struct amdgpu_device *adev) 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, PAGE_TABLE_BLOCK_SIZE, - amdgpu_vm_block_size - 9); + adev->vm_manager.block_size - 9); WREG32(mmVM_CONTEXT1_CNTL, tmp); if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS) gmc_v7_0_set_fault_enable_default(adev, false); @@ -1003,7 +1003,12 @@ static int gmc_v7_0_sw_init(void *handle) * Currently set to 4GB ((1 << 20) 4k pages). * Max GPUVM size for cayman and SI is 40 bits. */ - adev->vm_manager.max_pfn = amdgpu_vm_size << 18; + adev->vm_manager.vm_size = amdgpu_vm_size; + adev->vm_manager.block_size = amdgpu_vm_block_size; + adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18; + + DRM_INFO("vm size is %llu GB, block size is %d-bit\n", + adev->vm_manager.vm_size, adev->vm_manager.block_size); /* Set the internal MC address mask * This is the max address of the GPU's diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c index 42b2f357a799..108a20e832cf 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c @@ -853,7 +853,7 @@ static int gmc_v8_0_gart_enable(struct amdgpu_device *adev) 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); + adev->vm_manager.block_size - 9); WREG32(mmVM_CONTEXT1_CNTL, tmp); if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS) gmc_v8_0_set_fault_enable_default(adev, false); @@ -1087,7 +1087,12 @@ static int gmc_v8_0_sw_init(void *handle) * Currently set to 4GB ((1 << 20) 4k pages). * Max GPUVM size for cayman and SI is 40 bits. */ - adev->vm_manager.max_pfn = amdgpu_vm_size << 18; + adev->vm_manager.vm_size = amdgpu_vm_size; + adev->vm_manager.block_size = amdgpu_vm_block_size; + adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18; + + DRM_INFO("vm size is %llu GB, block size is %d-bit\n", + adev->vm_manager.vm_size, adev->vm_manager.block_size); /* Set the internal MC address mask * This is the max address of the GPU's diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index d81372357ae6..7632f4ad7aa9 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -532,11 +532,23 @@ static int gmc_v9_0_sw_init(void *handle) if (adev->flags & AMD_IS_APU) { adev->mc.vram_type = AMDGPU_VRAM_TYPE_UNKNOWN; + adev->vm_manager.vm_size = amdgpu_vm_size; + adev->vm_manager.block_size = amdgpu_vm_block_size; } else { /* XXX Don't know how to get VRAM type yet. */ adev->mc.vram_type = AMDGPU_VRAM_TYPE_HBM; + /* + * To fulfill 4-level page support, + * vm size is 256TB (48bit), maximum size of Vega10, + * block size 512 (9bit) + */ + adev->vm_manager.vm_size = 1U << 18; + adev->vm_manager.block_size = 9; } + DRM_INFO("vm size is %llu GB, block size is %d-bit\n", + adev->vm_manager.vm_size, adev->vm_manager.block_size); + /* This interrupt is VMC page fault.*/ r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_VMC, 0, &adev->mc.vm_fault); @@ -546,14 +558,7 @@ static int gmc_v9_0_sw_init(void *handle) if (r) return r; - /* Because of four level VMPTs, vm size is at least 512GB. - * The maximum size is 256TB (48bit). - */ - if (amdgpu_vm_size < 512) { - DRM_WARN("VM size is at least 512GB!\n"); - amdgpu_vm_size = 512; - } - adev->vm_manager.max_pfn = (uint64_t)amdgpu_vm_size << 18; + adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18; /* Set the internal MC address mask * This is the max address of the GPU's diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c index a065b4394ea7..3c9e27effc6a 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c @@ -242,7 +242,7 @@ int mmhub_v1_0_gart_enable(struct amdgpu_device *adev) EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1); tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_BLOCK_SIZE, - amdgpu_vm_block_size - 9); + adev->vm_manager.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);