gpu : introduce notion of a device
The concept of a device is being introduced for the GPU, which enables a
more efficient and scalable approach to managing multiple sensors for a
single endpoint.
Refactoring of Entity-Manager Configuration and Endpoint Discovery To
support multiple sensors for an endpoint, the following changes are
being made:
Entity-Manager Configuration Discovery: This task will be moved out of
the GPU Temperature Sensor implementation.
Endpoint Discovery: Similarly, endpoint discovery will also be performed
outside of the GPU Temperature Sensor implementation.
Frequency of Task Execution: Both entity-manager configuration discovery
and endpoint discovery will be performed only once per endpoint, rather
than repeatedly for each sensor. This optimization will improve
performance and reduce redundancy.
Tested.
```
$ curl -k -u 'root:0penBmc' https://10.137.203.137/redfish/v1/Chassis/NVIDIA_GB200_1/Sensors/temperature_NVIDIA_GB200_GPU_TEMP_0
{
"@odata.id": "/redfish/v1/Chassis/NVIDIA_GB200_1/Sensors/temperature_NVIDIA_GB200_GPU_TEMP_0",
"@odata.type": "#Sensor.v1_2_0.Sensor",
"Id": "temperature_NVIDIA_GB200_GPU_TEMP_0",
"Name": "NVIDIA GB200 GPU TEMP 0",
"Reading": 35.96875,
"ReadingRangeMax": 127.0,
"ReadingRangeMin": -128.0,
"ReadingType": "Temperature",
"ReadingUnits": "Cel",
"Status": {
"Health": "OK",
"State": "Enabled"
}
}%
```
Change-Id: Ie3dcd43caa031b4aaa61d8be3f5d71aefd53bc9a
Signed-off-by: Harshit Aghera <haghera@nvidia.com>
diff --git a/src/gpu/GpuSensor.hpp b/src/gpu/GpuSensor.hpp
index 7c70d55..2961404 100644
--- a/src/gpu/GpuSensor.hpp
+++ b/src/gpu/GpuSensor.hpp
@@ -7,34 +7,16 @@
#include "MctpRequester.hpp"
#include "Thresholds.hpp"
-#include "sensor.hpp"
+#include "UpdatableSensor.hpp"
-#include <boost/asio/io_context.hpp>
-#include <boost/asio/steady_timer.hpp>
-#include <boost/container/flat_map.hpp>
#include <sdbusplus/asio/connection.hpp>
#include <sdbusplus/asio/object_server.hpp>
-#include <sdbusplus/message.hpp>
-#include <chrono>
#include <cstdint>
-#include <map>
#include <memory>
#include <string>
-#include <utility>
-#include <variant>
#include <vector>
-constexpr const char* sensorPathPrefix = "/xyz/openbmc_project/sensors/";
-constexpr const char* sensorType = "NvidiaMctpVdm";
-
-using getSubTreeRet = std::vector<
- std::pair<std::string,
- std::vector<std::pair<std::string, std::vector<std::string>>>>>;
-using GpuSensorConfigMap =
- std::map<std::string, std::variant<std::string, bool, uint32_t, uint8_t,
- int64_t, std::vector<uint8_t>>>;
-
/**
* @struct DeviceInfo
* @brief Contains information about a device
@@ -52,30 +34,27 @@
* management via std::enable_shared_from_this
*/
struct GpuTempSensor :
- public Sensor,
+ public GpuSensor,
public std::enable_shared_from_this<GpuTempSensor>
{
public:
/**
* @brief Constructor for GpuTempSensor
- * @param conn D-Bus connection
- * @param io Boost ASIO I/O context for asynchronous operations
+ * @param conn D-Bus connection for system communication
* @param mctpRequester MCTP protocol requester for GPU communication
- * @param name Name of the sensor
- * @param sensorConfiguration Configuration string for the sensor
- * @param objectServer D-Bus object server
- * @param thresholdData Vector of threshold configurations
- * @param pollRate How often to poll for new readings
- * @param deviceInfo Information about the GPU device
- * @param verbose Whether to enable verbose logging
+ * @param name Name of the sensor for identification in the system
+ * @param sensorConfiguration Configuration string for the sensor containing
+ * setup parameters
+ * @param eid EID of the device endpoint
+ * @param objectServer D-Bus object server for exposing sensor interfaces
+ * @param thresholdData Vector of threshold configurations for temperature
+ * monitoring
*/
GpuTempSensor(std::shared_ptr<sdbusplus::asio::connection>& conn,
- boost::asio::io_context& io,
mctp::MctpRequester& mctpRequester, const std::string& name,
- const std::string& sensorConfiguration,
+ const std::string& sensorConfiguration, uint8_t eid,
sdbusplus::asio::object_server& objectServer,
- std::vector<thresholds::Threshold>&& thresholdData,
- std::chrono::milliseconds pollRate);
+ std::vector<thresholds::Threshold>&& thresholdData);
/**
* @brief Destructor
@@ -91,47 +70,9 @@
private:
/**
- * @brief Read the current temperature value from the GPU
- */
- void read();
-
- /**
- * @brief Initialize the sensor
- */
- void init();
-
- /**
* @brief Update the sensor reading
*/
- void update();
-
- /**
- * @brief Discover available GPUs on the system
- */
- void discoverGpus();
-
- /**
- * @brief Process MCTP endpoints discovered on the system
- *
- * @param[in] ec Error code from the D-Bus method call
- * @param[in] ret Object tree results containing MCTP endpoint information
- */
- void processMctpEndpoints(const boost::system::error_code& ec,
- const getSubTreeRet& ret);
-
- /**
- * @brief Process configuration properties for MCTP endpoints
- *
- * @param[in] ec Error code from the D-Bus properties method call
- * @param[in] configs Map of configuration properties for the endpoint
- */
- void processEndpointConfigs(const boost::system::error_code& ec,
- const GpuSensorConfigMap& configs);
- /**
- * @brief Process a discovered GPU endpoint
- * @param eid The endpoint ID of the discovered GPU
- */
- void processGpuEndpoint(uint8_t eid);
+ void update() final;
/**
* @brief MCTP endpoint ID
@@ -144,52 +85,12 @@
uint8_t sensorId;
/**
- * @brief How often to poll the sensor in milliseconds
- */
- std::chrono::milliseconds sensorPollMs;
-
- /**
- * @brief Timer for scheduling sensor reads
- */
- boost::asio::steady_timer waitTimer;
-
- /**
* @brief Reference to the MCTP requester for communication
*/
mctp::MctpRequester& mctpRequester;
/**
- * @brief D-Bus connection
- */
- std::shared_ptr<sdbusplus::asio::connection> conn;
-
- /**
* @brief D-Bus object server
*/
sdbusplus::asio::object_server& objectServer;
};
-
-/**
- * @brief Create GPU temperature sensors
- * @param io Boost ASIO I/O context
- * @param objectServer D-Bus object server
- * @param sensors Map to store created sensors
- * @param dbusConnection D-Bus connection
- * @param mctpRequester MCTP requester for GPU communication
- */
-void createSensors(
- boost::asio::io_context& io, sdbusplus::asio::object_server& objectServer,
- boost::container::flat_map<std::string, std::shared_ptr<GpuTempSensor>>&
- sensors,
- std::shared_ptr<sdbusplus::asio::connection>& dbusConnection,
- mctp::MctpRequester& mctpRequester);
-
-/**
- * @brief Handle D-Bus interface removal events
- * @param message D-Bus message containing interface removal information
- * @param sensors Map of GPU temperature sensors to check for removal
- */
-void interfaceRemoved(
- sdbusplus::message_t& message,
- boost::container::flat_map<std::string, std::shared_ptr<GpuTempSensor>>&
- sensors);