diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdkfd/kfd_topology.c')
| -rw-r--r-- | drivers/gpu/drm/amd/amdkfd/kfd_topology.c | 78 |
1 files changed, 38 insertions, 40 deletions
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c index 9476e30d6baa..e477d7509646 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c @@ -108,24 +108,6 @@ struct kfd_node *kfd_device_by_id(uint32_t gpu_id) return top_dev->gpu; } -struct kfd_node *kfd_device_by_pci_dev(const struct pci_dev *pdev) -{ - struct kfd_topology_device *top_dev; - struct kfd_node *device = NULL; - - down_read(&topology_lock); - - list_for_each_entry(top_dev, &topology_device_list, list) - if (top_dev->gpu && top_dev->gpu->adev->pdev == pdev) { - device = top_dev->gpu; - break; - } - - up_read(&topology_lock); - - return device; -} - /* Called with write topology_lock acquired */ static void kfd_release_topology_device(struct kfd_topology_device *dev) { @@ -537,6 +519,8 @@ static ssize_t node_show(struct kobject *kobj, struct attribute *attr, dev->gpu->kfd->mec_fw_version); sysfs_show_32bit_prop(buffer, offs, "capability", dev->node_props.capability); + sysfs_show_32bit_prop(buffer, offs, "capability2", + dev->node_props.capability2); sysfs_show_64bit_prop(buffer, offs, "debug_prop", dev->node_props.debug_prop); sysfs_show_32bit_prop(buffer, offs, "sdma_fw_version", @@ -968,24 +952,23 @@ static void kfd_update_system_properties(void) up_read(&topology_lock); } -static void find_system_memory(const struct dmi_header *dm, - void *private) +static void find_system_memory(const struct dmi_header *dm, void *private) { + struct dmi_mem_device *memdev = container_of(dm, struct dmi_mem_device, header); struct kfd_mem_properties *mem; - u16 mem_width, mem_clock; struct kfd_topology_device *kdev = (struct kfd_topology_device *)private; - const u8 *dmi_data = (const u8 *)(dm + 1); - - if (dm->type == DMI_ENTRY_MEM_DEVICE && dm->length >= 0x15) { - mem_width = (u16)(*(const u16 *)(dmi_data + 0x6)); - mem_clock = (u16)(*(const u16 *)(dmi_data + 0x11)); - list_for_each_entry(mem, &kdev->mem_props, list) { - if (mem_width != 0xFFFF && mem_width != 0) - mem->width = mem_width; - if (mem_clock != 0) - mem->mem_clk_max = mem_clock; - } + + if (memdev->header.type != DMI_ENTRY_MEM_DEVICE) + return; + if (memdev->header.length < sizeof(struct dmi_mem_device)) + return; + + list_for_each_entry(mem, &kdev->mem_props, list) { + if (memdev->total_width != 0xFFFF && memdev->total_width != 0) + mem->width = memdev->total_width; + if (memdev->speed != 0) + mem->mem_clk_max = memdev->speed; } } @@ -1683,17 +1666,32 @@ static int fill_in_l2_l3_pcache(struct kfd_cache_properties **props_ext, int cache_type, unsigned int cu_processor_id, struct kfd_node *knode) { - unsigned int cu_sibling_map_mask; + unsigned int cu_sibling_map_mask = 0; int first_active_cu; int i, j, k, xcc, start, end; int num_xcc = NUM_XCC(knode->xcc_mask); struct kfd_cache_properties *pcache = NULL; enum amdgpu_memory_partition mode; struct amdgpu_device *adev = knode->adev; + bool found = false; start = ffs(knode->xcc_mask) - 1; end = start + num_xcc; - cu_sibling_map_mask = cu_info->bitmap[start][0][0]; + + /* To find the bitmap in the first active cu in the first + * xcc, it is based on the assumption that evrey xcc must + * have at least one active cu. + */ + for (i = 0; i < gfx_info->max_shader_engines && !found; i++) { + for (j = 0; j < gfx_info->max_sh_per_se && !found; j++) { + if (cu_info->bitmap[start][i % 4][j % 4]) { + cu_sibling_map_mask = + cu_info->bitmap[start][i % 4][j % 4]; + found = true; + } + } + } + cu_sibling_map_mask &= ((1 << pcache_info[cache_type].num_cu_shared) - 1); first_active_cu = ffs(cu_sibling_map_mask); @@ -1714,7 +1712,8 @@ static int fill_in_l2_l3_pcache(struct kfd_cache_properties **props_ext, pcache->cacheline_size = pcache_info[cache_type].cache_line_size; if (KFD_GC_VERSION(knode) == IP_VERSION(9, 4, 3) || - KFD_GC_VERSION(knode) == IP_VERSION(9, 4, 4)) + KFD_GC_VERSION(knode) == IP_VERSION(9, 4, 4) || + KFD_GC_VERSION(knode) == IP_VERSION(9, 5, 0)) mode = adev->gmc.gmc_funcs->query_mem_partition_mode(adev); else mode = UNKNOWN_MEMORY_PARTITION_MODE; @@ -1776,7 +1775,7 @@ static void kfd_fill_cache_non_crat_info(struct kfd_topology_device *dev, struct struct amdgpu_cu_info *cu_info = &kdev->adev->gfx.cu_info; struct amdgpu_gfx_config *gfx_info = &kdev->adev->gfx.config; int gpu_processor_id; - struct kfd_cache_properties *props_ext; + struct kfd_cache_properties *props_ext = NULL; int num_of_entries = 0; int num_of_cache_types = 0; struct kfd_gpu_cache_info cache_info[KFD_MAX_CACHE_TYPES]; @@ -1984,6 +1983,9 @@ static void kfd_topology_set_capabilities(struct kfd_topology_device *dev) if (kfd_dbg_has_ttmps_always_setup(dev->gpu)) dev->node_props.debug_prop |= HSA_DBG_DISPATCH_INFO_ALWAYS_VALID; + if (dev->gpu->adev->sdma.supported_reset & AMDGPU_RESET_TYPE_PER_QUEUE) + dev->node_props.capability2 |= HSA_CAP2_PER_SDMA_QUEUE_RESET_SUPPORTED; + if (KFD_GC_VERSION(dev->gpu) < IP_VERSION(10, 0, 0)) { if (KFD_GC_VERSION(dev->gpu) == IP_VERSION(9, 4, 3) || KFD_GC_VERSION(dev->gpu) == IP_VERSION(9, 4, 4)) @@ -2004,10 +2006,6 @@ static void kfd_topology_set_capabilities(struct kfd_topology_device *dev) dev->node_props.debug_prop |= HSA_DBG_WATCH_ADDR_MASK_LO_BIT_GFX10 | HSA_DBG_WATCH_ADDR_MASK_HI_BIT; - if (KFD_GC_VERSION(dev->gpu) >= IP_VERSION(11, 0, 0)) - dev->node_props.capability |= - HSA_CAP_TRAP_DEBUG_PRECISE_MEMORY_OPERATIONS_SUPPORTED; - if (KFD_GC_VERSION(dev->gpu) >= IP_VERSION(12, 0, 0)) dev->node_props.capability |= HSA_CAP_TRAP_DEBUG_PRECISE_ALU_OPERATIONS_SUPPORTED; |