drm/amdgpu/gmc9: fix 64 bit division in partition code

Rework logic or use do_div() to avoid problems on 32 bit.

v2: add a missing case for XCP macro
v3: fix out of bounds array access
v4: fix xcp handling harder

Acked-by: Guchun Chen <guchun.chen@amd.com> (v1)
Reviewed-by: Mukul Joshi <mukul.joshi@amd.com> (v3)
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

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

@@ -794,3 +794,18 @@ void amdgpu_amdkfd_unlock_kfd(struct amdgpu_device *adev)
 {
 	kgd2kfd_unlock_kfd();
 }
+
+
+u64 amdgpu_amdkfd_xcp_memory_size(struct amdgpu_device *adev, int xcp_id)
+{
+	u64 tmp;
+	s8 mem_id = KFD_XCP_MEM_ID(adev, xcp_id);
+
+	if (adev->gmc.num_mem_partitions && xcp_id >= 0 && mem_id >= 0) {
+		tmp = adev->gmc.mem_partitions[mem_id].size;
+		do_div(tmp, adev->xcp_mgr->num_xcp_per_mem_partition);
+		return tmp;
+	} else {
+		return adev->gmc.real_vram_size;
+	}
+}

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

@@ -333,15 +333,14 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
 void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
 		uint64_t size, u32 alloc_flag, int8_t xcp_id);
 
+u64 amdgpu_amdkfd_xcp_memory_size(struct amdgpu_device *adev, int xcp_id);
+
 #define KFD_XCP_MEM_ID(adev, xcp_id) \
 		((adev)->xcp_mgr && (xcp_id) >= 0 ?\
 		(adev)->xcp_mgr->xcp[(xcp_id)].mem_id : -1)
 
-#define KFD_XCP_MEMORY_SIZE(adev, xcp_id)\
-		((adev)->gmc.num_mem_partitions && (xcp_id) >= 0 ?\
-		(adev)->gmc.mem_partitions[KFD_XCP_MEM_ID((adev), (xcp_id))].size /\
-		(adev)->xcp_mgr->num_xcp_per_mem_partition :\
-		(adev)->gmc.real_vram_size)
+#define KFD_XCP_MEMORY_SIZE(adev, xcp_id) amdgpu_amdkfd_xcp_memory_size((adev), (xcp_id))
+
 
 #if IS_ENABLED(CONFIG_HSA_AMD)
 void amdgpu_amdkfd_gpuvm_init_mem_limits(void);

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

@@ -814,11 +814,14 @@ static void amdgpu_ttm_gart_bind_gfx9_mqd(struct amdgpu_device *adev,
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
 	uint64_t total_pages = ttm->num_pages;
 	int num_xcc = max(1U, adev->gfx.num_xcc_per_xcp);
-	uint64_t page_idx, pages_per_xcc = total_pages / num_xcc;
+	uint64_t page_idx, pages_per_xcc;
 	int i;
 	uint64_t ctrl_flags = (flags & ~AMDGPU_PTE_MTYPE_VG10_MASK) |
 			AMDGPU_PTE_MTYPE_VG10(AMDGPU_MTYPE_NC);
 
+	pages_per_xcc = total_pages;
+	do_div(pages_per_xcc, num_xcc);
+
 	for (i = 0, page_idx = 0; i < num_xcc; i++, page_idx += pages_per_xcc) {
 		/* MQD page: use default flags */
 		amdgpu_gart_bind(adev,

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

@@ -1914,9 +1914,10 @@ gmc_v9_0_init_acpi_mem_ranges(struct amdgpu_device *adev,
 	adev->gmc.num_mem_partitions = num_ranges;
 
 	/* If there is only partition, don't use entire size */
-	if (adev->gmc.num_mem_partitions == 1)
-		mem_ranges[0].size =
-			(mem_ranges[0].size * (mem_groups - 1) / mem_groups);
+	if (adev->gmc.num_mem_partitions == 1) {
+		mem_ranges[0].size = mem_ranges[0].size * (mem_groups - 1);
+		do_div(mem_ranges[0].size, mem_groups);
+	}
 }
 
 static void
@@ -1948,8 +1949,8 @@ gmc_v9_0_init_sw_mem_ranges(struct amdgpu_device *adev,
 		break;
 	}
 
-	size = (adev->gmc.real_vram_size >> AMDGPU_GPU_PAGE_SHIFT) /
-	       adev->gmc.num_mem_partitions;
+	size = adev->gmc.real_vram_size >> AMDGPU_GPU_PAGE_SHIFT;
+	size /= adev->gmc.num_mem_partitions;
 
 	for (i = 0; i < adev->gmc.num_mem_partitions; ++i) {
 		mem_ranges[i].range.fpfn = start_addr;

drivers/gpu/drm/amd/amdkfd/kfd_svm.c

@@ -1939,10 +1939,14 @@ void svm_range_set_max_pages(struct amdgpu_device *adev)
 	uint64_t max_pages;
 	uint64_t pages, _pages;
 	uint64_t min_pages = 0;
-	int i;
+	int i, id;
 
 	for (i = 0; i < adev->kfd.dev->num_nodes; i++) {
-		pages = KFD_XCP_MEMORY_SIZE(adev, adev->kfd.dev->nodes[i]->xcp->id) >> 17;
+		if (adev->kfd.dev->nodes[i]->xcp)
+			id = adev->kfd.dev->nodes[i]->xcp->id;
+		else
+			id = -1;
+		pages = KFD_XCP_MEMORY_SIZE(adev, id) >> 17;
 		pages = clamp(pages, 1ULL << 9, 1ULL << 18);
 		pages = rounddown_pow_of_two(pages);
 		min_pages = min_not_zero(min_pages, pages);