nvidia-gpu: introduce notion of a device
Perform device discovery tasks only once per device to prepare for
introducing additional gpu sensors.
In the current implementation, sensor updates and device discovery via
MCTP are managed within a single class for simplicity. However, since a
GPU device typically includes multiple sensors, performing device
discovery for each individual sensor is inefficient. Instead, it would
be more effective to execute device discovery once per device.
Tested: Build an image for gb200nvl-obmc machine with the following
patches cherry picked. This patches are needed to enable the mctp stack.
https://gerrit.openbmc.org/c/openbmc/openbmc/+/79422
```
$ curl -k -u 'root:0penBmc' https://10.137.203.137/redfish/v1/Chassis/NVIDIA_GB200_1/Sensors/temperature_NVIDIA_GB200_GPU_0_TEMP_0
{
"@odata.id": "/redfish/v1/Chassis/NVIDIA_GB200_1/Sensors/temperature_NVIDIA_GB200_GPU_0_TEMP_0",
"@odata.type": "#Sensor.v1_2_0.Sensor",
"Id": "temperature_NVIDIA_GB200_GPU_0_TEMP_0",
"Name": "NVIDIA GB200 GPU 0 TEMP 0",
"Reading": 37.6875,
"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/nvidia-gpu/NvidiaGpuSensor.cpp b/src/nvidia-gpu/NvidiaGpuSensor.cpp
index 86b356b..1626545 100644
--- a/src/nvidia-gpu/NvidiaGpuSensor.cpp
+++ b/src/nvidia-gpu/NvidiaGpuSensor.cpp
@@ -14,25 +14,19 @@
#include <bits/basic_string.h>
#include <MctpRequester.hpp>
+#include <NvidiaDeviceDiscovery.hpp>
#include <NvidiaGpuMctpVdm.hpp>
#include <OcpMctpVdm.hpp>
-#include <boost/asio/io_context.hpp>
-#include <boost/container/flat_map.hpp>
#include <phosphor-logging/lg2.hpp>
#include <sdbusplus/asio/connection.hpp>
#include <sdbusplus/asio/object_server.hpp>
-#include <sdbusplus/message.hpp>
-#include <sdbusplus/message/native_types.hpp>
-#include <algorithm>
-#include <chrono>
#include <cstddef>
#include <cstdint>
#include <functional>
#include <memory>
#include <string>
#include <utility>
-#include <variant>
#include <vector>
using namespace std::literals;
@@ -41,19 +35,17 @@
static constexpr double gpuTempSensorMaxReading = 127;
static constexpr double gpuTempSensorMinReading = -128;
-GpuTempSensor::GpuTempSensor(
+NvidiaGpuTempSensor::NvidiaGpuTempSensor(
std::shared_ptr<sdbusplus::asio::connection>& conn,
- boost::asio::io_context& io, mctp::MctpRequester& mctpRequester,
- const std::string& name, const std::string& sensorConfiguration,
+ mctp::MctpRequester& mctpRequester, const std::string& name,
+ const std::string& sensorConfiguration, const uint8_t eid,
sdbusplus::asio::object_server& objectServer,
- std::vector<thresholds::Threshold>&& thresholdData,
- std::chrono::milliseconds pollRate) :
+ std::vector<thresholds::Threshold>&& thresholdData) :
Sensor(escapeName(name), std::move(thresholdData), sensorConfiguration,
"temperature", false, true, gpuTempSensorMaxReading,
gpuTempSensorMinReading, conn),
- sensorId{gpuTempSensorId}, sensorPollMs(pollRate),
- waitTimer(io, std::chrono::steady_clock::duration(0)),
- mctpRequester(mctpRequester), conn(conn), objectServer(objectServer)
+ eid(eid), sensorId{gpuTempSensorId}, mctpRequester(mctpRequester),
+ objectServer(objectServer)
{
std::string dbusPath =
sensorPathPrefix + "temperature/"s + escapeName(name);
@@ -70,12 +62,11 @@
association = objectServer.add_interface(dbusPath, association::interface);
- discoverGpus();
+ setInitialProperties(sensor_paths::unitDegreesC);
}
-GpuTempSensor::~GpuTempSensor()
+NvidiaGpuTempSensor::~NvidiaGpuTempSensor()
{
- waitTimer.cancel();
for (const auto& iface : thresholdInterfaces)
{
objectServer.remove_interface(iface);
@@ -84,72 +75,18 @@
objectServer.remove_interface(sensorInterface);
}
-void GpuTempSensor::checkThresholds()
+void NvidiaGpuTempSensor::checkThresholds()
{
thresholds::checkThresholds(this);
}
-void GpuTempSensor::queryEndpoints(const boost::system::error_code& ec,
- const GetSubTreeType& ret)
-{
- if (ec)
- {
- lg2::error("Error querying endoints :{ERROR}", "ERROR", ec.message());
- return;
- }
-
- if (ret.empty())
- {
- return;
- }
-
- for (const auto& [objPath, services] : ret)
- {
- for (const auto& [service, ifaces] : services)
- {
- for (const auto& iface : ifaces)
- {
- if (iface == "xyz.openbmc_project.MCTP.Endpoint")
- {
- conn->async_method_call(
- [this](const boost::system::error_code& ec,
- const SensorBaseConfigMap& configs) {
- this->processEndpoint(ec, configs);
- },
- service, objPath, "org.freedesktop.DBus.Properties",
- "GetAll", iface);
- }
- }
- }
- }
-}
-
-void GpuTempSensor::read()
-{
- update();
-
- waitTimer.expires_after(std::chrono::milliseconds(sensorPollMs));
- waitTimer.async_wait(
- [weakPtrToThis = std::weak_ptr<GpuTempSensor>{shared_from_this()}](
- const boost::system::error_code& ec) {
- if (ec)
- {
- return;
- }
- if (auto ptr = weakPtrToThis.lock())
- {
- ptr->read();
- }
- });
-}
-
-void GpuTempSensor::processResponse(int sendRecvMsgResult)
+void NvidiaGpuTempSensor::processResponse(int sendRecvMsgResult)
{
if (sendRecvMsgResult != 0)
{
lg2::error(
- "Error updating Temperature Sensor: sending message over MCTP failed, rc={RC}",
- "RC", sendRecvMsgResult);
+ "Error updating Temperature Sensor for eid {EID} and sensor id {SID} : sending message over MCTP failed, rc={RC}",
+ "EID", eid, "SID", sensorId, "RC", sendRecvMsgResult);
return;
}
@@ -163,264 +100,29 @@
if (rc != 0 || cc != ocp::accelerator_management::CompletionCode::SUCCESS)
{
lg2::error(
- "Error updating Temperature Sensor: decode failed, rc={RC}, cc={CC}, reasonCode={RESC}",
- "RC", rc, "CC", cc, "RESC", reasonCode);
+ "Error updating Temperature Sensor for eid {EID} and sensor id {SID} : decode failed. "
+ "rc={RC}, cc={CC}, reasonCode={RESC}",
+ "EID", eid, "SID", sensorId, "RC", rc, "CC", cc, "RESC",
+ reasonCode);
return;
}
updateValue(tempValue);
}
-void GpuTempSensor::update()
+void NvidiaGpuTempSensor::update()
{
auto rc = gpu::encodeGetTemperatureReadingRequest(
0, sensorId, getTemperatureReadingRequest);
+
if (rc != 0)
{
- lg2::error("Error updating Temperature Sensor: encode failed, rc={RC}",
- "RC", rc);
- return;
+ lg2::error(
+ "Error updating Temperature Sensor for eid {EID} and sensor id {SID} : encode failed, rc={RC}",
+ "EID", eid, "SID", sensorId, "RC", rc);
}
mctpRequester.sendRecvMsg(
eid, getTemperatureReadingRequest, getTemperatureReadingResponse,
[this](int sendRecvMsgResult) { processResponse(sendRecvMsgResult); });
}
-
-void GpuTempSensor::processQueryDeviceIdResponse(uint8_t eid,
- int sendRecvMsgResult)
-{
- if (sendRecvMsgResult != 0)
- {
- lg2::error(
- "Error processing GPU endpoint: sending message over MCTP failed, rc={RC}",
- "RC", sendRecvMsgResult);
- return;
- }
-
- ocp::accelerator_management::CompletionCode cc{};
- uint16_t reasonCode = 0;
- uint8_t responseDeviceType = 0;
- uint8_t responseInstanceId = 0;
-
- auto rc = gpu::decodeQueryDeviceIdentificationResponse(
- queryDeviceIdentificationResponse, cc, reasonCode, responseDeviceType,
- responseInstanceId);
-
- if (rc != 0 || cc != ocp::accelerator_management::CompletionCode::SUCCESS)
- {
- lg2::error(
- "Error processing GPU endpoint: decode failed, rc={RC}, cc={CC}, reasonCode={RESC}",
- "RC", rc, "CC", cc, "RESC", reasonCode);
- return;
- }
-
- if (responseDeviceType ==
- static_cast<uint8_t>(gpu::DeviceIdentification::DEVICE_GPU))
- {
- lg2::info(
- "Found the GPU with EID {EID}, DeviceType {DEVTYPE}, InstanceId {IID}.",
- "EID", eid, "DEVTYPE", responseDeviceType, "IID",
- responseInstanceId);
-
- this->eid = eid;
- setInitialProperties(sensor_paths::unitDegreesC);
- read();
- }
-}
-
-void GpuTempSensor::processGpuEndpoint(uint8_t eid)
-{
- auto rc = gpu::encodeQueryDeviceIdentificationRequest(
- 0, queryDeviceIdentificationRequest);
- if (rc != 0)
- {
- lg2::error("Error processing GPU endpoint: encode failed, rc={RC}",
- "RC", rc);
- return;
- }
-
- mctpRequester.sendRecvMsg(
- eid, queryDeviceIdentificationRequest,
- queryDeviceIdentificationResponse, [this, eid](int sendRecvMsgResult) {
- processQueryDeviceIdResponse(eid, sendRecvMsgResult);
- });
-}
-
-void GpuTempSensor::processEndpoint(const boost::system::error_code& ec,
- const SensorBaseConfigMap& endpoint)
-{
- if (ec)
- {
- lg2::error("Error processing MCTP endpoint: {ERROR}", "ERROR",
- ec.message());
- return;
- }
-
- uint8_t eid{};
- std::vector<uint8_t> mctpTypes{};
-
- auto hasEid = endpoint.find("EID");
- if (hasEid != endpoint.end())
- {
- const auto* eidPtr = std::get_if<uint8_t>(&hasEid->second);
- if (eidPtr != nullptr)
- {
- eid = *eidPtr;
- }
- else
- {
- lg2::error(
- "Error processing MCTP endpoint: Property EID does not have valid type.");
- return;
- }
- }
- else
- {
- lg2::error(
- "Error processing MCTP endpoint: Property EID not found in the configuration.");
- return;
- }
-
- auto hasMctpTypes = endpoint.find("SupportedMessageTypes");
- if (hasMctpTypes != endpoint.end())
- {
- const auto* mctpTypePtr =
- std::get_if<std::vector<uint8_t>>(&hasMctpTypes->second);
- if (mctpTypePtr != nullptr)
- {
- mctpTypes = *mctpTypePtr;
- }
- else
- {
- lg2::error(
- "Error processing MCTP endpoint: Property SupportedMessageTypes does not have valid type.");
- return;
- }
- }
- else
- {
- lg2::error(
- "Error processing MCTP endpoint: Property SupportedMessageTypes not found in the configuration.");
- return;
- }
-
- if (std::find(mctpTypes.begin(), mctpTypes.end(),
- ocp::accelerator_management::messageType) != mctpTypes.end())
- {
- lg2::info(
- "GpuTempSensor::discoverGpus(): Found OCP MCTP VDM Endpoint with ID {EID}",
- "EID", eid);
- this->processGpuEndpoint(eid);
- }
-}
-
-void GpuTempSensor::discoverGpus()
-{
- std::string searchPath{"/au/com/codeconstruct/"};
- std::vector<std::string> ifaceList{{"xyz.openbmc_project.MCTP.Endpoint"}};
-
- conn->async_method_call(
- [this](const boost::system::error_code& ec, const GetSubTreeType& ret) {
- queryEndpoints(ec, ret);
- },
- "xyz.openbmc_project.ObjectMapper",
- "/xyz/openbmc_project/object_mapper",
- "xyz.openbmc_project.ObjectMapper", "GetSubTree", searchPath, 0,
- ifaceList);
-}
-
-void processSensorConfigs(
- 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, const ManagedObjectType& resp)
-{
- for (const auto& [path, interfaces] : resp)
- {
- for (const auto& [intf, cfg] : interfaces)
- {
- if (intf != configInterfaceName(sensorType))
- {
- continue;
- }
-
- std::string name = loadVariant<std::string>(cfg, "Name");
-
- uint64_t pollRate = loadVariant<uint64_t>(cfg, "PollRate");
-
- sensors[name] = std::make_shared<GpuTempSensor>(
- dbusConnection, io, mctpRequester, name, path, objectServer,
- std::vector<thresholds::Threshold>{},
- std::chrono::milliseconds{pollRate});
-
- lg2::info(
- "Added GPU Temperature Sensor {NAME} with chassis path: {PATH}.",
- "NAME", name, "PATH", path);
- }
- }
-}
-
-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)
-{
- if (!dbusConnection)
- {
- lg2::error("Connection not created");
- return;
- }
- dbusConnection->async_method_call(
- [&sensors, &mctpRequester, &dbusConnection, &io,
- &objectServer](const boost::system::error_code& ec,
- const ManagedObjectType& resp) {
- if (ec)
- {
- lg2::error("Error contacting entity manager");
- return;
- }
-
- processSensorConfigs(io, objectServer, sensors, dbusConnection,
- mctpRequester, resp);
- },
- entityManagerName, "/xyz/openbmc_project/inventory",
- "org.freedesktop.DBus.ObjectManager", "GetManagedObjects");
-}
-
-void interfaceRemoved(
- sdbusplus::message_t& message,
- boost::container::flat_map<std::string, std::shared_ptr<GpuTempSensor>>&
- sensors)
-{
- if (message.is_method_error())
- {
- lg2::error("interfacesRemoved callback method error");
- return;
- }
-
- sdbusplus::message::object_path removedPath;
- std::vector<std::string> interfaces;
-
- message.read(removedPath, interfaces);
-
- // If the xyz.openbmc_project.Confguration.X interface was removed
- // for one or more sensors, delete those sensor objects.
- auto sensorIt = sensors.begin();
- while (sensorIt != sensors.end())
- {
- if ((sensorIt->second->configurationPath == removedPath) &&
- (std::find(interfaces.begin(), interfaces.end(),
- configInterfaceName(sensorType)) != interfaces.end()))
- {
- sensorIt = sensors.erase(sensorIt);
- }
- else
- {
- sensorIt++;
- }
- }
-}