[SWDEV-453922] AMD SMI to provide mapping feature of other enumeration methods (#51)
Added enumeration mapping for - drm render - drm card - hsa id - hip id - hip uuid (rocminfo uuid) Signed-off-by: AL Musaffar, Yazen <Yazen.ALMusaffar@amd.com> Signed-off-by: Maisam Arif <Maisam.Arif@amd.com> Co-authored-by: Maisam Arif <Maisam.Arif@amd.com>
Este cometimento está contido em:
AL Musaffar, Yazen
cometido por
GitHub
GitHub
ascendente
d5b2763aba
cometimento
2936e00fed
+101
-34
@@ -54,6 +54,7 @@
|
||||
#include "rocm_smi/rocm_smi_logger.h"
|
||||
#include "rocm_smi/rocm_smi_utils.h"
|
||||
#include "rocm_smi/rocm_smi.h"
|
||||
#include "rocm_smi/rocm_smi_kfd.h"
|
||||
|
||||
// a global instance of std::mutex to protect data passed during threads
|
||||
std::mutex myMutex;
|
||||
@@ -124,6 +125,7 @@ static amdsmi_status_t get_gpu_device_from_handle(amdsmi_processor_handle proces
|
||||
return AMDSMI_STATUS_NOT_SUPPORTED;
|
||||
}
|
||||
|
||||
|
||||
template <typename F, typename ...Args>
|
||||
amdsmi_status_t rsmi_wrapper(F && f,
|
||||
amdsmi_processor_handle processor_handle, uint32_t increment_gpu_id = 0, Args &&... args) {
|
||||
@@ -510,8 +512,8 @@ amdsmi_status_t amdsmi_get_processor_handles_by_type(amdsmi_socket_handle socket
|
||||
return AMDSMI_STATUS_SUCCESS;
|
||||
}
|
||||
|
||||
|
||||
#endif
|
||||
|
||||
amdsmi_status_t amdsmi_get_processor_type(amdsmi_processor_handle processor_handle ,
|
||||
processor_type_t* processor_type) {
|
||||
|
||||
@@ -549,7 +551,103 @@ amdsmi_get_gpu_device_bdf(amdsmi_processor_handle processor_handle, amdsmi_bdf_t
|
||||
return AMDSMI_STATUS_SUCCESS;
|
||||
}
|
||||
|
||||
amdsmi_status_t amdsmi_get_gpu_board_info(amdsmi_processor_handle processor_handle, amdsmi_board_info_t *board_info) {
|
||||
amdsmi_status_t
|
||||
amdsmi_get_gpu_device_uuid(amdsmi_processor_handle processor_handle,
|
||||
unsigned int *uuid_length,
|
||||
char *uuid) {
|
||||
AMDSMI_CHECK_INIT();
|
||||
|
||||
if (uuid_length == nullptr || uuid == nullptr || uuid_length == nullptr || *uuid_length < AMDSMI_GPU_UUID_SIZE) {
|
||||
return AMDSMI_STATUS_INVAL;
|
||||
}
|
||||
|
||||
amd::smi::AMDSmiGPUDevice* gpu_device = nullptr;
|
||||
amdsmi_status_t r = get_gpu_device_from_handle(processor_handle, &gpu_device);
|
||||
if (r != AMDSMI_STATUS_SUCCESS)
|
||||
return r;
|
||||
|
||||
amdsmi_status_t status = AMDSMI_STATUS_SUCCESS;
|
||||
SMIGPUDEVICE_MUTEX(gpu_device->get_mutex())
|
||||
|
||||
amdsmi_asic_info_t asic_info = {};
|
||||
const uint8_t fcn = 0xff;
|
||||
|
||||
status = amdsmi_get_gpu_asic_info(processor_handle, &asic_info);
|
||||
if (status != AMDSMI_STATUS_SUCCESS) {
|
||||
printf("Getting asic info failed. Return code: %d", status);
|
||||
return status;
|
||||
}
|
||||
|
||||
/* generate random UUID */
|
||||
status = amdsmi_uuid_gen(uuid,
|
||||
strtoull(asic_info.asic_serial, nullptr, 16),
|
||||
(uint16_t)asic_info.device_id, fcn);
|
||||
return status;
|
||||
}
|
||||
|
||||
amdsmi_status_t
|
||||
amdsmi_get_gpu_enumeration_info(amdsmi_processor_handle processor_handle,
|
||||
amdsmi_enumeration_info_t *info){
|
||||
|
||||
// Ensure library initialization
|
||||
AMDSMI_CHECK_INIT();
|
||||
|
||||
if (info == nullptr) {
|
||||
return AMDSMI_STATUS_INVAL;
|
||||
}
|
||||
|
||||
amdsmi_status_t status;
|
||||
|
||||
// Retrieve GPU device from the processor handle
|
||||
amd::smi::AMDSmiGPUDevice* gpu_device = nullptr;
|
||||
status = get_gpu_device_from_handle(processor_handle, &gpu_device);
|
||||
if (status != AMDSMI_STATUS_SUCCESS) {
|
||||
return status;
|
||||
}
|
||||
|
||||
// Retrieve DRM Card ID
|
||||
info->drm_card = gpu_device->get_card_from_bdf();
|
||||
|
||||
// Retrieve DRM Render ID
|
||||
info->drm_render = gpu_device->get_render_id();
|
||||
|
||||
// Retrieve HIP ID (difference from the smallest node ID) and HSA ID
|
||||
std::map<uint64_t, std::shared_ptr<amd::smi::KFDNode>> nodes;
|
||||
if (amd::smi::DiscoverKFDNodes(&nodes) == 0) {
|
||||
uint32_t smallest_node_id = std::numeric_limits<uint32_t>::max();
|
||||
for (const auto& node_pair : nodes) {
|
||||
uint32_t node_id = 0;
|
||||
if (node_pair.second->get_node_id(&node_id) == 0) {
|
||||
smallest_node_id = std::min(smallest_node_id, node_id);
|
||||
}
|
||||
}
|
||||
|
||||
// Default to 0xffffffff as not supported
|
||||
info->hsa_id = std::numeric_limits<uint32_t>::max();
|
||||
info->hip_id = std::numeric_limits<uint32_t>::max();
|
||||
amdsmi_kfd_info_t kfd_info;
|
||||
status = amdsmi_get_gpu_kfd_info(processor_handle, &kfd_info);
|
||||
if (status == AMDSMI_STATUS_SUCCESS) {
|
||||
info->hsa_id = kfd_info.node_id;
|
||||
info->hip_id = kfd_info.node_id - smallest_node_id;
|
||||
}
|
||||
}
|
||||
|
||||
// Retrieve HIP UUID
|
||||
std::string hip_uuid_str = "GPU-";
|
||||
amdsmi_asic_info_t asic_info = {};
|
||||
status = amdsmi_get_gpu_asic_info(processor_handle, &asic_info);
|
||||
if (status == AMDSMI_STATUS_SUCCESS) {
|
||||
hip_uuid_str += std::string(asic_info.asic_serial).substr(0, sizeof(info->hip_uuid) - hip_uuid_str.size() - 1);
|
||||
std::strncpy(info->hip_uuid, hip_uuid_str.c_str(), sizeof(info->hip_uuid) - 1);
|
||||
info->hip_uuid[sizeof(info->hip_uuid) - 1] = '\0'; // Ensure null termination
|
||||
}
|
||||
|
||||
return AMDSMI_STATUS_SUCCESS;
|
||||
}
|
||||
|
||||
amdsmi_status_t
|
||||
amdsmi_get_gpu_board_info(amdsmi_processor_handle processor_handle, amdsmi_board_info_t *board_info) {
|
||||
AMDSMI_CHECK_INIT();
|
||||
|
||||
if (board_info == nullptr) {
|
||||
@@ -1277,6 +1375,7 @@ amdsmi_get_gpu_asic_info(amdsmi_processor_handle processor_handle, amdsmi_asic_i
|
||||
return AMDSMI_STATUS_SUCCESS;
|
||||
}
|
||||
|
||||
|
||||
amdsmi_status_t
|
||||
amdsmi_get_gpu_xgmi_link_status(amdsmi_processor_handle processor_handle,
|
||||
amdsmi_xgmi_link_status_t *link_status) {
|
||||
@@ -3399,38 +3498,6 @@ amdsmi_status_t amdsmi_get_gpu_driver_info(amdsmi_processor_handle processor_han
|
||||
}
|
||||
|
||||
|
||||
amdsmi_status_t
|
||||
amdsmi_get_gpu_device_uuid(amdsmi_processor_handle processor_handle, unsigned int *uuid_length, char *uuid) {
|
||||
AMDSMI_CHECK_INIT();
|
||||
|
||||
if (uuid_length == nullptr || uuid == nullptr || uuid_length == nullptr || *uuid_length < AMDSMI_GPU_UUID_SIZE) {
|
||||
return AMDSMI_STATUS_INVAL;
|
||||
}
|
||||
|
||||
amd::smi::AMDSmiGPUDevice* gpu_device = nullptr;
|
||||
amdsmi_status_t r = get_gpu_device_from_handle(processor_handle, &gpu_device);
|
||||
if (r != AMDSMI_STATUS_SUCCESS)
|
||||
return r;
|
||||
|
||||
amdsmi_status_t status = AMDSMI_STATUS_SUCCESS;
|
||||
SMIGPUDEVICE_MUTEX(gpu_device->get_mutex())
|
||||
|
||||
amdsmi_asic_info_t asic_info = {};
|
||||
const uint8_t fcn = 0xff;
|
||||
|
||||
status = amdsmi_get_gpu_asic_info(processor_handle, &asic_info);
|
||||
if (status != AMDSMI_STATUS_SUCCESS) {
|
||||
printf("Getting asic info failed. Return code: %d", status);
|
||||
return status;
|
||||
}
|
||||
|
||||
/* generate random UUID */
|
||||
status = amdsmi_uuid_gen(uuid,
|
||||
strtoull(asic_info.asic_serial, nullptr, 16),
|
||||
(uint16_t)asic_info.device_id, fcn);
|
||||
return status;
|
||||
}
|
||||
|
||||
amdsmi_status_t amdsmi_get_pcie_info(amdsmi_processor_handle processor_handle, amdsmi_pcie_info_t *info) {
|
||||
AMDSMI_CHECK_INIT();
|
||||
|
||||
|
||||
@@ -30,6 +30,7 @@
|
||||
#include <map>
|
||||
#include <memory>
|
||||
#include <unordered_set>
|
||||
#include <filesystem>
|
||||
|
||||
namespace amd {
|
||||
namespace smi {
|
||||
@@ -285,6 +286,77 @@ const GPUComputeProcessList_t& AMDSmiGPUDevice::amdgpu_get_compute_process_list(
|
||||
return compute_process_list_;
|
||||
}
|
||||
|
||||
// Convert `amdsmi_bdf_t` to a PCI BDF string
|
||||
std::string AMDSmiGPUDevice::bdf_to_string() const {
|
||||
std::ostringstream oss;
|
||||
oss << std::setfill('0') << std::hex // Use hexadecimal formatting
|
||||
<< std::setw(4) << bdf_.domain_number << ":" // Domain (4 digits)
|
||||
<< std::setw(2) << static_cast<int>(bdf_.bus_number) << ":" // Bus (2 digits)
|
||||
<< std::setw(2) << static_cast<int>(bdf_.device_number) << "." // Device (2 digits)
|
||||
<< static_cast<int>(bdf_.function_number); // Function (1 digit)
|
||||
return oss.str();
|
||||
}
|
||||
|
||||
uint32_t AMDSmiGPUDevice::get_card_from_bdf() const {
|
||||
const std::string drm_path = "/sys/class/drm/";
|
||||
|
||||
// Iterate over the contents of /sys/class/drm/
|
||||
for (const auto& entry : std::filesystem::directory_iterator(drm_path)) {
|
||||
const std::string device_name = entry.path().filename();
|
||||
|
||||
// Check if the entry starts with "card"
|
||||
if (device_name.find("card") == 0) {
|
||||
const std::string card_path = drm_path + device_name + "/device";
|
||||
|
||||
// Open the uevent file for the device
|
||||
std::ifstream uevent_file(card_path + "/uevent");
|
||||
if (!uevent_file) {
|
||||
continue; // Skip if the file is not found
|
||||
}
|
||||
|
||||
std::string line;
|
||||
while (std::getline(uevent_file, line)) {
|
||||
// Check for the PCI_SLOT_NAME and if it contains the BDF
|
||||
if (line.rfind("PCI_SLOT_NAME", 0) == 0 && line.find(bdf_to_string()) != std::string::npos) {
|
||||
return std::stoi(device_name.substr(4)); // Convert extracted number to int
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
return std::numeric_limits<uint32_t>::max(); // Return -1 if no matching card is found
|
||||
}
|
||||
|
||||
uint32_t AMDSmiGPUDevice::get_render_id() const {
|
||||
const std::string drm_path = "/sys/class/drm/";
|
||||
|
||||
// Iterate over the contents of /sys/class/drm/
|
||||
for (const auto& entry : std::filesystem::directory_iterator(drm_path)) {
|
||||
const std::string device_name = entry.path().filename();
|
||||
|
||||
// Check if the entry starts with "renderD"
|
||||
if (device_name.find("renderD") == 0) {
|
||||
const std::string render_path = drm_path + device_name + "/device";
|
||||
|
||||
// Open the uevent file for the device
|
||||
std::ifstream uevent_file(render_path + "/uevent");
|
||||
if (!uevent_file) {
|
||||
continue; // Skip if the file is not found
|
||||
}
|
||||
|
||||
std::string line;
|
||||
while (std::getline(uevent_file, line)) {
|
||||
// Check for the PCI_SLOT_NAME and if it contains the BDF
|
||||
if (line.rfind("PCI_SLOT_NAME", 0) == 0 && line.find(bdf_to_string()) != std::string::npos) {
|
||||
return std::stoi(device_name.substr(7)); // Extract only the number after "renderD"
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
return std::numeric_limits<uint32_t>::max(); // Return -1 if no matching render ID is found
|
||||
}
|
||||
|
||||
|
||||
} // namespace smi
|
||||
} // namespace amd
|
||||
|
||||
Criar uma nova questão referindo esta
Bloquear um utilizador