drm/amdgpu: allow get_vm_pde to change flags as well

And also provide the level for which we need a PDE.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/amdgpu/amdgpu.h

@@ -346,7 +346,8 @@ struct amdgpu_gart_funcs {
 	uint64_t (*get_vm_pte_flags)(struct amdgpu_device *adev,
 				     uint32_t flags);
 	/* get the pde for a given mc addr */
-	u64 (*get_vm_pde)(struct amdgpu_device *adev, u64 addr);
+	void (*get_vm_pde)(struct amdgpu_device *adev, int level,
+			   u64 *dst, u64 *flags);
 	uint32_t (*get_invalidate_req)(unsigned int vm_id);
 };
 
@@ -1826,7 +1827,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring)
 #define amdgpu_asic_get_config_memsize(adev) (adev)->asic_funcs->get_config_memsize((adev))
 #define amdgpu_gart_flush_gpu_tlb(adev, vmid) (adev)->gart.gart_funcs->flush_gpu_tlb((adev), (vmid))
 #define amdgpu_gart_set_pte_pde(adev, pt, idx, addr, flags) (adev)->gart.gart_funcs->set_pte_pde((adev), (pt), (idx), (addr), (flags))
-#define amdgpu_gart_get_vm_pde(adev, addr) (adev)->gart.gart_funcs->get_vm_pde((adev), (addr))
+#define amdgpu_gart_get_vm_pde(adev, level, dst, flags) (adev)->gart.gart_funcs->get_vm_pde((adev), (level), (dst), (flags))
 #define amdgpu_vm_copy_pte(adev, ib, pe, src, count) ((adev)->vm_manager.vm_pte_funcs->copy_pte((ib), (pe), (src), (count)))
 #define amdgpu_vm_write_pte(adev, ib, pe, value, count, incr) ((adev)->vm_manager.vm_pte_funcs->write_pte((ib), (pe), (value), (count), (incr)))
 #define amdgpu_vm_set_pte_pde(adev, ib, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->set_pte_pde((ib), (pe), (addr), (count), (incr), (flags)))

drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c

@@ -1070,9 +1070,10 @@ static void amdgpu_vm_update_pde(struct amdgpu_pte_update_params *params,
 				 struct amdgpu_vm_pt *parent,
 				 struct amdgpu_vm_pt *entry)
 {
-	struct amdgpu_bo *bo = entry->base.bo, *shadow = NULL;
+	struct amdgpu_bo *bo = entry->base.bo, *shadow = NULL, *pbo;
 	uint64_t pd_addr, shadow_addr = 0;
-	uint64_t pde, pt;
+	uint64_t pde, pt, flags;
+	unsigned level;
 
 	/* Don't update huge pages here */
 	if (entry->huge)
@@ -1087,15 +1088,19 @@ static void amdgpu_vm_update_pde(struct amdgpu_pte_update_params *params,
 			shadow_addr = amdgpu_bo_gpu_offset(shadow);
 	}
 
+	for (level = 0, pbo = parent->base.bo->parent; pbo; ++level)
+		pbo = pbo->parent;
+
 	pt = amdgpu_bo_gpu_offset(bo);
-	pt = amdgpu_gart_get_vm_pde(params->adev, pt);
+	flags = AMDGPU_PTE_VALID;
+	amdgpu_gart_get_vm_pde(params->adev, level, &pt, &flags);
 	if (shadow) {
 		pde = shadow_addr + (entry - parent->entries) * 8;
-		params->func(params, pde, pt, 1, 0, AMDGPU_PTE_VALID);
+		params->func(params, pde, pt, 1, 0, flags);
 	}
 
 	pde = pd_addr + (entry - parent->entries) * 8;
-	params->func(params, pde, pt, 1, 0, AMDGPU_PTE_VALID);
+	params->func(params, pde, pt, 1, 0, flags);
 }
 
 /*
@@ -1305,7 +1310,6 @@ static void amdgpu_vm_handle_huge_pages(struct amdgpu_pte_update_params *p,
 	    !(flags & AMDGPU_PTE_VALID)) {
 
 		dst = amdgpu_bo_gpu_offset(entry->base.bo);
-		dst = amdgpu_gart_get_vm_pde(p->adev, dst);
 		flags = AMDGPU_PTE_VALID;
 	} else {
 		/* Set the huge page flag to stop scanning at this PDE */
@@ -1314,9 +1318,11 @@ static void amdgpu_vm_handle_huge_pages(struct amdgpu_pte_update_params *p,
 
 	if (!entry->huge && !(flags & AMDGPU_PDE_PTE))
 		return;
-
 	entry->huge = !!(flags & AMDGPU_PDE_PTE);
 
+	amdgpu_gart_get_vm_pde(p->adev, p->adev->vm_manager.num_level - 1,
+			       &dst, &flags);
+
 	if (use_cpu_update) {
 		/* In case a huge page is replaced with a system
 		 * memory mapping, p->pages_addr != NULL and

drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c

@@ -3686,10 +3686,11 @@ static void gfx_v9_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX);
 	uint32_t req = ring->adev->gart.gart_funcs->get_invalidate_req(vm_id);
+	uint64_t flags = AMDGPU_PTE_VALID;
 	unsigned eng = ring->vm_inv_eng;
 
-	pd_addr = amdgpu_gart_get_vm_pde(ring->adev, pd_addr);
-	pd_addr |= AMDGPU_PTE_VALID;
+	amdgpu_gart_get_vm_pde(ring->adev, -1, &pd_addr, &flags);
+	pd_addr |= flags;
 
 	gfx_v9_0_write_data_to_reg(ring, usepfp, true,
 				   hub->ctx0_ptb_addr_lo32 + (2 * vm_id),

drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c

@@ -395,10 +395,10 @@ static uint64_t gmc_v6_0_get_vm_pte_flags(struct amdgpu_device *adev,
 	return pte_flag;
 }
 
-static uint64_t gmc_v6_0_get_vm_pde(struct amdgpu_device *adev, uint64_t addr)
+static void gmc_v6_0_get_vm_pde(struct amdgpu_device *adev, int level,
+				uint64_t *addr, uint64_t *flags)
 {
-	BUG_ON(addr & 0xFFFFFF0000000FFFULL);
-	return addr;
+	BUG_ON(*addr & 0xFFFFFF0000000FFFULL);
 }
 
 static void gmc_v6_0_set_fault_enable_default(struct amdgpu_device *adev,

drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c

@@ -480,10 +480,10 @@ static uint64_t gmc_v7_0_get_vm_pte_flags(struct amdgpu_device *adev,
 	return pte_flag;
 }
 
-static uint64_t gmc_v7_0_get_vm_pde(struct amdgpu_device *adev, uint64_t addr)
+static void gmc_v7_0_get_vm_pde(struct amdgpu_device *adev, int level,
+				uint64_t *addr, uint64_t *flags)
 {
-	BUG_ON(addr & 0xFFFFFF0000000FFFULL);
-	return addr;
+	BUG_ON(*addr & 0xFFFFFF0000000FFFULL);
 }
 
 /**

drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c

@@ -677,10 +677,10 @@ static uint64_t gmc_v8_0_get_vm_pte_flags(struct amdgpu_device *adev,
 	return pte_flag;
 }
 
-static uint64_t gmc_v8_0_get_vm_pde(struct amdgpu_device *adev, uint64_t addr)
+static void gmc_v8_0_get_vm_pde(struct amdgpu_device *adev, int level,
+				uint64_t *addr, uint64_t *flags)
 {
-	BUG_ON(addr & 0xFFFFFF0000000FFFULL);
-	return addr;
+	BUG_ON(*addr & 0xFFFFFF0000000FFFULL);
 }
 
 /**

drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c

@@ -474,11 +474,13 @@ static uint64_t gmc_v9_0_get_vm_pte_flags(struct amdgpu_device *adev,
 	return pte_flag;
 }
 
-static u64 gmc_v9_0_get_vm_pde(struct amdgpu_device *adev, u64 addr)
+static void gmc_v9_0_get_vm_pde(struct amdgpu_device *adev, int level,
+				uint64_t *addr, uint64_t *flags)
 {
-	addr = adev->vm_manager.vram_base_offset + addr - adev->mc.vram_start;
-	BUG_ON(addr & 0xFFFF00000000003FULL);
-	return addr;
+	if (!(*flags & AMDGPU_PDE_PTE))
+		*addr = adev->vm_manager.vram_base_offset + *addr -
+			adev->mc.vram_start;
+	BUG_ON(*addr & 0xFFFF00000000003FULL);
 }
 
 static const struct amdgpu_gart_funcs gmc_v9_0_gart_funcs = {

drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c

@@ -1146,10 +1146,11 @@ static void sdma_v4_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
 {
 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	uint32_t req = ring->adev->gart.gart_funcs->get_invalidate_req(vm_id);
+	uint64_t flags = AMDGPU_PTE_VALID;
 	unsigned eng = ring->vm_inv_eng;
 
-	pd_addr = amdgpu_gart_get_vm_pde(ring->adev, pd_addr);
-	pd_addr |= AMDGPU_PTE_VALID;
+	amdgpu_gart_get_vm_pde(ring->adev, -1, &pd_addr, &flags);
+	pd_addr |= flags;
 
 	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) |
 			  SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf));

drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c

@@ -1295,11 +1295,12 @@ static void uvd_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
 {
 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	uint32_t req = ring->adev->gart.gart_funcs->get_invalidate_req(vm_id);
-	uint32_t data0, data1, mask;
+	uint64_t flags = AMDGPU_PTE_VALID;
 	unsigned eng = ring->vm_inv_eng;
+	uint32_t data0, data1, mask;
 
-	pd_addr = amdgpu_gart_get_vm_pde(ring->adev, pd_addr);
-	pd_addr |= AMDGPU_PTE_VALID;
+	amdgpu_gart_get_vm_pde(ring->adev, -1, &pd_addr, &flags);
+	pd_addr |= flags;
 
 	data0 = (hub->ctx0_ptb_addr_hi32 + vm_id * 2) << 2;
 	data1 = upper_32_bits(pd_addr);
@@ -1346,10 +1347,11 @@ static void uvd_v7_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring,
 {
 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	uint32_t req = ring->adev->gart.gart_funcs->get_invalidate_req(vm_id);
+	uint64_t flags = AMDGPU_PTE_VALID;
 	unsigned eng = ring->vm_inv_eng;
 
-	pd_addr = amdgpu_gart_get_vm_pde(ring->adev, pd_addr);
-	pd_addr |= AMDGPU_PTE_VALID;
+	amdgpu_gart_get_vm_pde(ring->adev, -1, &pd_addr, &flags);
+	pd_addr |= flags;
 
 	amdgpu_ring_write(ring, HEVC_ENC_CMD_REG_WRITE);
 	amdgpu_ring_write(ring,	(hub->ctx0_ptb_addr_hi32 + vm_id * 2) << 2);

drivers/gpu/drm/amd/amdgpu/vce_v4_0.c

@@ -969,10 +969,11 @@ static void vce_v4_0_emit_vm_flush(struct amdgpu_ring *ring,
 {
 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	uint32_t req = ring->adev->gart.gart_funcs->get_invalidate_req(vm_id);
+	uint64_t flags = AMDGPU_PTE_VALID;
 	unsigned eng = ring->vm_inv_eng;
 
-	pd_addr = amdgpu_gart_get_vm_pde(ring->adev, pd_addr);
-	pd_addr |= AMDGPU_PTE_VALID;
+	amdgpu_gart_get_vm_pde(ring->adev, -1, &pd_addr, &flags);
+	pd_addr |= flags;
 
 	amdgpu_ring_write(ring, VCE_CMD_REG_WRITE);
 	amdgpu_ring_write(ring,	(hub->ctx0_ptb_addr_hi32 + vm_id * 2) << 2);

drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c

@@ -892,11 +892,12 @@ static void vcn_v1_0_dec_ring_emit_vm_flush(struct amdgpu_ring *ring,
 {
 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	uint32_t req = ring->adev->gart.gart_funcs->get_invalidate_req(vm_id);
-	uint32_t data0, data1, mask;
+	uint64_t flags = AMDGPU_PTE_VALID;
 	unsigned eng = ring->vm_inv_eng;
+	uint32_t data0, data1, mask;
 
-	pd_addr = amdgpu_gart_get_vm_pde(ring->adev, pd_addr);
-	pd_addr |= AMDGPU_PTE_VALID;
+	amdgpu_gart_get_vm_pde(ring->adev, -1, &pd_addr, &flags);
+	pd_addr |= flags;
 
 	data0 = (hub->ctx0_ptb_addr_hi32 + vm_id * 2) << 2;
 	data1 = upper_32_bits(pd_addr);
@@ -1024,10 +1025,11 @@ static void vcn_v1_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring,
 {
 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	uint32_t req = ring->adev->gart.gart_funcs->get_invalidate_req(vm_id);
+	uint64_t flags = AMDGPU_PTE_VALID;
 	unsigned eng = ring->vm_inv_eng;
 
-	pd_addr = amdgpu_gart_get_vm_pde(ring->adev, pd_addr);
-	pd_addr |= AMDGPU_PTE_VALID;
+	amdgpu_gart_get_vm_pde(ring->adev, -1, &pd_addr, &flags);
+	pd_addr |= flags;
 
 	amdgpu_ring_write(ring, VCN_ENC_CMD_REG_WRITE);
 	amdgpu_ring_write(ring,