nvidia-gpu: add BoostClockFrequency property

Implement BoostClockFrequency property for NVIDIA GPU
inventory to expose the default boost clock frequency of GPU
accelerators.

The property is added t

nvidia-gpu: add BoostClockFrequency property

Implement BoostClockFrequency property for NVIDIA GPU
inventory to expose the default boost clock frequency of GPU
accelerators.

The property is added to xyz.openbmc_project.Inventory.Item
.Accelerator interface, utilizing the existing MCTP VDM Property
ID 21 (DEFAULT_BOOST_CLOCKS) to query the GPU hardware over MCTP
and populate the property value.

Changes:
- src/nvidia-gpu/NvidiaGpuMctpVdm.hpp: Add uint64_t to
InventoryValue variant to support numeric clock speed values.
- src/nvidia-gpu/NvidiaGpuMctpVdm.cpp: Add DEFAULT_BOOST_CLOCKS
case to decodeInventoryData to parse uint64_t clock speed from
MCTP response payload.
- src/nvidia-gpu/Inventory.cpp: Register BoostClockFrequency
property on Accelerator interface, add DEFAULT_BOOST_CLOCKS to
properties query map, and handle uint64_t response in
handleInventoryPropertyResponse.

Tested: Build an image for nvl32-obmc machine with the following patch
cherry picked.

https://gerrit.openbmc.org/c/openbmc/openbmc/+/85763
https://gerrit.openbmc.org/c/openbmc/dbus-sensors/+/85080

Verified via busctl that BoostClockFrequency property
appears under xyz.openbmc_project.Inventory.Item.Accelerator
interface for GPU devices and contains the correct boost clock
value (e.g., 2430 MHz). Confirmed successful MCTP query and
property update through nvidiagpusensor service logs.

Change-Id: I3d7410e1b1a455a81263c89f63ac1c6338eeefe1
Signed-off-by: Eric Liu <liuer@nvidia.com>

show more ...

nvidia-gpu: implement PhysicalContext interface

This change implements the xyz.openbmc_project.Common.PhysicalContext
interface for NVIDIA GPU sensors (Energy, Power, Temperature, and
Voltage). This

nvidia-gpu: implement PhysicalContext interface

This change implements the xyz.openbmc_project.Common.PhysicalContext
interface for NVIDIA GPU sensors (Energy, Power, Temperature, and
Voltage). This allows sensors to expose their hardware context
information to external interfaces like Redfish.

Instead of hardcoding the PhysicalContext value, this implementation
uses a device type enum (gpu::DeviceIdentification) to determine the
appropriate PhysicalContext. A helper function maps device types to
their corresponding D-Bus PhysicalContext values, ensuring proper
separation of concerns.

For GPU devices, the Type property is set to 'GPU'. For other device
types (SMA, PCIe), no PhysicalContext interface is created, keeping
their D-Bus representation clean.

This implementation follows the interface definition introduced in:
https://gerrit.openbmc.org/c/openbmc/phosphor-dbus-interfaces/+/86504

Changes:
- src/nvidia-gpu/NvidiaSensorUtils.hpp: New helper file containing
deviceTypeToPhysicalContext() function that maps DeviceIdentification
enum to D-Bus PhysicalContext paths.
- src/nvidia-gpu/*.hpp: Update sensor constructors to accept
gpu::DeviceIdentification deviceType parameter (no default value).
- src/nvidia-gpu/*Sensor.cpp: Use helper function to determine
PhysicalContext. Conditionally create interface only when a valid
context is returned. Register 'Type' property with the mapped value.
- src/nvidia-gpu/NvidiaGpuDevice.cpp: Pass DEVICE_GPU enum for all GPU
sensors. Fix io member initialization in constructor.
- src/nvidia-gpu/NvidiaSmaDevice.cpp: Pass DEVICE_SMA enum for SMA
sensors (no PhysicalContext interface created).

Design rationale:
- GpuDevice class doesn't need to know D-Bus implementation details
- Centralized mapping function makes maintenance easier
- Type-safe enum prevents typos and provides compile-time checking
- Automatic handling for different device types without conditional
logic in device classes

Tested: Build an image for nvl32-obmc machine with the following patch
cherry picked.

https://gerrit.openbmc.org/c/openbmc/openbmc/+/85490

Verified via busctl that the Type property appears under
xyz.openbmc_project.Common.PhysicalContext interface for GPU sensors
only and contains the correct value 'GPU'. SMA device sensors do not
have this interface, as expected.

Depends-On: I83dcbe4810139fb92fddf6b099f5a1a057e7e05e
Change-Id: I1d5abfa5d4416af3565bf315e0f28cb6af56f14c
Signed-off-by: Ender Hsieh <andhsieh@nvidia.com>

show more ...

nvidia-gpu: add ConnectX Ethernet Port Metrics

Add xyz.openbmc_project.Metric.Value interface for each of the following
Ethernet port metrics of a ConnectX device.

- TXBytes
- RXBytes
- RXMulticast

nvidia-gpu: add ConnectX Ethernet Port Metrics

Add xyz.openbmc_project.Metric.Value interface for each of the following
Ethernet port metrics of a ConnectX device.

- TXBytes
- RXBytes
- RXMulticastFrames
- TXMulticastFrames
- RXUnicastFrames
- TXUnicastFrames
- RXBroadcastFrames
- TXBroadcastFrames
- RXFCSErrors
- RXFrameAlignmentErrors
- RXFalseCarrierErrors
- RXUndersizeFrames
- RXOversizeFrames
- RXPauseXONFrames
- RXPauseXOFFFrames
- TXPauseXONFrames
- TXPauseXOFFFrames
- TXSingleCollisions
- TXMultipleCollisions
- TXLateCollisions
- TXExcessiveCollisions

PDI Patch -
https://gerrit.openbmc.org/c/openbmc/phosphor-dbus-interfaces/+/84847

Tested: Build an image for nvl32-obmc machine with the following patch
cherry picked.

https://gerrit.openbmc.org/c/openbmc/entity-manager/+/84257
https://gerrit.openbmc.org/c/openbmc/openbmc/+/85490

The openbmc patch cherry-picks the following patches that are currently
under review.

```
1. device tree
https://lore.kernel.org/all/aRbLqH8pLWCQryhu@molberding.nvidia.com/
2. mctpd patches
https://github.com/CodeConstruct/mctp/pull/85
3. u-boot changes
https://lore.kernel.org/openbmc/20251121-msx4-v1-0-fc0118b666c1@nvidia.com/T/#t
4. kernel changes as specified in the openbmc patch (for espi)
5. entity-manager changes
https://gerrit.openbmc.org/c/openbmc/entity-manager/+/85455
6. platform-init changes
https://gerrit.openbmc.org/c/openbmc/platform-init/+/85456
7. spi changes
https://lore.kernel.org/all/20251121-w25q01jv_fixup-v1-1-3d175050db73@nvidia.com/
```

```
root@nvl32-obmc:~# busctl tree xyz.openbmc_project.GpuSensor
`- /xyz
`- /xyz/openbmc_project
|- /xyz/openbmc_project/inventory
| |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_0_NIC
| | |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_0_NIC/Port_1
| | `- /xyz/openbmc_project/inventory/Nvidia_ConnectX_0_NIC/Port_2
| |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_0_PCIe
| | |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_0_PCIe/DOWN_0
| | |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_0_PCIe/DOWN_1
| | `- /xyz/openbmc_project/inventory/Nvidia_ConnectX_0_PCIe/UP_0
| |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_2_NIC
| | |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_2_NIC/Port_1
| | `- /xyz/openbmc_project/inventory/Nvidia_ConnectX_2_NIC/Port_2
| |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_2_PCIe
| | |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_2_PCIe/DOWN_0
| | |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_2_PCIe/DOWN_1
| | `- /xyz/openbmc_project/inventory/Nvidia_ConnectX_2_PCIe/UP_0
| |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_3_NIC
| | |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_3_NIC/Port_1
| | `- /xyz/openbmc_project/inventory/Nvidia_ConnectX_3_NIC/Port_2
| `- /xyz/openbmc_project/inventory/Nvidia_ConnectX_3_PCIe
| |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_3_PCIe/DOWN_0
| |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_3_PCIe/DOWN_1
| `- /xyz/openbmc_project/inventory/Nvidia_ConnectX_3_PCIe/UP_0
|- /xyz/openbmc_project/metric
| |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1
| | `- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/rx_broadcast_frames
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/rx_bytes
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/rx_false_carrier_errors
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/rx_fcs_errors
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/rx_frame_alignment_errors
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/rx_multicast_frames
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/rx_oversize_frames
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/rx_pause_xoff_frames
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/rx_pause_xon_frames
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/rx_undersize_frames
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/rx_unicast_frames
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/tx_broadcast_frames
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/tx_bytes
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/tx_excessive_collisions
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/tx_late_collisions
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/tx_multicast_frames
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/tx_multiple_collisions
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/tx_pause_xoff_frames
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/tx_pause_xon_frames
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/tx_single_collisions
| | `- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_NIC_Port_1/nic/tx_unicast_frames

root@nvl32-obmc:~# busctl introspect xyz.openbmc_project.GpuSensor /xyz/openbmc_project/metric/port_Nvidia_ConnectX_3_NIC_Port_2/nic/rx_bytes
NAME TYPE SIGNATURE RESULT/VALUE FLAGS
org.freedesktop.DBus.Introspectable interface - - -
.Introspect method - s -
org.freedesktop.DBus.Peer interface - - -
.GetMachineId method - s -
.Ping method - - -
org.freedesktop.DBus.Properties interface - - -
.Get method ss v -
.GetAll method s a{sv} -
.Set method ssv - -
.PropertiesChanged signal sa{sv}as - -
xyz.openbmc_project.Association.Definitions interface - - -
.Associations property a(sss) 1 "measuring" "measured_by" "/xyz/ope... emits-change
xyz.openbmc_project.Metric.Value interface - - -
.Unit property s "xyz.openbmc_project.Metric.Value.Uni... emits-change
.Value property d 0 emits-change
```

Change-Id: I30123e35b759182039cb6f25526fafe733c0f354
Signed-off-by: Harshit Aghera <haghera@nvidia.com>

show more ...

nvidia-gpu: add support for PCIe port metrics

Add xyz.openbmc_project.Metric.Value interface for each of the following
PCIe port metric of a ConnectX device.

PCIeErrors.CorrectableErrorCount
PCIeEr

nvidia-gpu: add support for PCIe port metrics

Add xyz.openbmc_project.Metric.Value interface for each of the following
PCIe port metric of a ConnectX device.

PCIeErrors.CorrectableErrorCount
PCIeErrors.NonFatalErrorCount
PCIeErrors.FatalErrorCount
PCIeErrors.L0ToRecoveryCount
PCIeErrors.ReplayCount
PCIeErrors.ReplayRolloverCount
PCIeErrors.NAKSentCount
PCIeErrors.NAKReceivedCount
PCIeErrors.UnsupportedRequestCount

PDI Patch -
https://gerrit.openbmc.org/c/openbmc/phosphor-dbus-interfaces/+/84839

Tested: Build an image for nvl32-obmc machine with the following patch
cherry picked.

https://gerrit.openbmc.org/c/openbmc/openbmc/+/85490

The patch cherry-picks the following patches that are currently under
review.

```
1. device tree
https://lore.kernel.org/all/aRbLqH8pLWCQryhu@molberding.nvidia.com/
2. mctpd patches
https://github.com/CodeConstruct/mctp/pull/85
3. u-boot changes
https://lore.kernel.org/openbmc/20251121-msx4-v1-0-fc0118b666c1@nvidia.com/T/#t
4. kernel changes as specified in the openbmc patch (for espi)
5. entity-manager changes
https://gerrit.openbmc.org/c/openbmc/entity-manager/+/85455
6. platform-init changes
https://gerrit.openbmc.org/c/openbmc/platform-init/+/85456
7. spi changes
https://lore.kernel.org/all/20251121-w25q01jv_fixup-v1-1-3d175050db73@nvidia.com/
```

```
root@nvl32-obmc:~# busctl tree xyz.openbmc_project.GpuSensor
`- /xyz
`- /xyz/openbmc_project
|- /xyz/openbmc_project/inventory
| |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_0_PCIe
| | |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_0_PCIe/DOWN_0
| | |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_0_PCIe/DOWN_1
| | `- /xyz/openbmc_project/inventory/Nvidia_ConnectX_0_PCIe/UP_0
| |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_2_PCIe
| | |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_2_PCIe/DOWN_0
| | |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_2_PCIe/DOWN_1
| | `- /xyz/openbmc_project/inventory/Nvidia_ConnectX_2_PCIe/UP_0
| `- /xyz/openbmc_project/inventory/Nvidia_ConnectX_3_PCIe
| |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_3_PCIe/DOWN_0
| |- /xyz/openbmc_project/inventory/Nvidia_ConnectX_3_PCIe/DOWN_1
| `- /xyz/openbmc_project/inventory/Nvidia_ConnectX_3_PCIe/UP_0
|- /xyz/openbmc_project/metric
| |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_PCIe_DOWN_0
| | `- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_PCIe_DOWN_0/pcie
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_PCIe_DOWN_0/pcie/correctable_error_count
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_PCIe_DOWN_0/pcie/fatal_error_count
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_PCIe_DOWN_0/pcie/l0_to_recovery_count
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_PCIe_DOWN_0/pcie/nak_received_count
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_PCIe_DOWN_0/pcie/nak_sent_count
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_PCIe_DOWN_0/pcie/non_fatal_error_count
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_PCIe_DOWN_0/pcie/replay_count
| | |- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_PCIe_DOWN_0/pcie/replay_rollover_count
| | `- /xyz/openbmc_project/metric/port_Nvidia_ConnectX_0_PCIe_DOWN_0/pcie/unsupported_request_count

root@nvl32-obmc:~# busctl introspect xyz.openbmc_project.GpuSensor /xyz/openbmc_project/metric/port_Nvidia_ConnectX_3_PCIe_DOWN_1/pcie/l0_to_recovery_count
NAME TYPE SIGNATURE RESULT/VALUE FLAGS
org.freedesktop.DBus.Introspectable interface - - -
.Introspect method - s -
org.freedesktop.DBus.Peer interface - - -
.GetMachineId method - s -
.Ping method - - -
org.freedesktop.DBus.Properties interface - - -
.Get method ss v -
.GetAll method s a{sv} -
.Set method ssv - -
.PropertiesChanged signal sa{sv}as - -
xyz.openbmc_project.Association.Definitions interface - - -
.Associations property a(sss) 1 "measuring" "measured_by" "/xyz/ope... emits-change
xyz.openbmc_project.Metric.Value interface - - -
.Unit property s "xyz.openbmc_project.Metric.Value.Uni... emits-change
.Value property d 1 emits-change
```

Change-Id: I3379c09346653d6a6bf2921bf765f0adf5a22098
Signed-off-by: Harshit Aghera <haghera@nvidia.com>

show more ...

nvidia-gpu: enable gpu software inventory

The patch uses the MCTP VDM command to retrieve the GPU driver version
and updates the DBus interface xyz.openbmc_project.Software.Version with
this informa

nvidia-gpu: enable gpu software inventory

The patch uses the MCTP VDM command to retrieve the GPU driver version
and updates the DBus interface xyz.openbmc_project.Software.Version with
this information at DBus object path /xyz/openbmc_project/software/. The
patch also associates software inventory to the chassis inventory
item. The GPU driver version is made available in Redfish at the URI
/redfish/v1/UpdateService/FirmwareInventory/.

Tested: Build an image for nvl32-obmc machine with the following patches
cherry picked.

https://gerrit.openbmc.org/c/openbmc/openbmc/+/85490

The patch cherry-picks the following patches that are currently under
review.

```
1. device tree
https://lore.kernel.org/all/aRbLqH8pLWCQryhu@molberding.nvidia.com/
2. mctpd patches
https://github.com/CodeConstruct/mctp/pull/85
3. u-boot changes
https://lore.kernel.org/openbmc/20251121-msx4-v1-0-fc0118b666c1@nvidia.com/T/#t
4. kernel changes as specified in the openbmc patch (for espi)
5. entity-manager changes
https://gerrit.openbmc.org/c/openbmc/entity-manager/+/85455
6. platform-init changes
https://gerrit.openbmc.org/c/openbmc/platform-init/+/85456
7. spi changes
https://lore.kernel.org/all/20251121-w25q01jv_fixup-v1-1-3d175050db73@nvidia.com/
```

The GPU driver version shows up on the DBus.

Change-Id: I712fe0952a02f36e386d3f37a5d4a8192ba641de
Signed-off-by: Harshit Aghera <haghera@nvidia.com>

show more ...

nvidia-gpu: add support for PCIe port telemetry

Add xyz.openbmc_project.Inventory.Connector.Port Interface for each PCIe
port of a ConnectX device.

PDI patches to extend the xyz.openbmc_project.Inv

nvidia-gpu: add support for PCIe port telemetry

Add xyz.openbmc_project.Inventory.Connector.Port Interface for each PCIe
port of a ConnectX device.

PDI patches to extend the xyz.openbmc_project.Inventory.Connector.Port
Interface -
https://gerrit.openbmc.org/c/openbmc/phosphor-dbus-interfaces/+/84653
https://gerrit.openbmc.org/c/openbmc/phosphor-dbus-interfaces/+/84652

Tested: Build an image for nvl32-obmc machine with the following patch
cherry picked.

https://gerrit.openbmc.org/c/openbmc/openbmc/+/85490

The patch cherry-picks the following patches that are currently under
review.

```
1. device tree
https://lore.kernel.org/all/aRbLqH8pLWCQryhu@molberding.nvidia.com/
2. mctpd patches
https://github.com/CodeConstruct/mctp/pull/85
3. u-boot changes
https://lore.kernel.org/openbmc/20251121-msx4-v1-0-fc0118b666c1@nvidia.com/T/#t
4. kernel changes as specified in the openbmc patch (for espi)
5. entity-manager changes
https://gerrit.openbmc.org/c/openbmc/entity-manager/+/85455
6. platform-init changes
https://gerrit.openbmc.org/c/openbmc/platform-init/+/85456
7. spi changes
https://lore.kernel.org/all/20251121-w25q01jv_fixup-v1-1-3d175050db73@nvidia.com/
```

```
root@nvl32-obmc:~# busctl tree xyz.openbmc_project.GpuSensor
`- /xyz
`- /xyz/openbmc_project
|- /xyz/openbmc_project/inventory
| `- /xyz/openbmc_project/inventory/pcie_devices
| |- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_0
| | |- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_0/DOWN_0
| | |- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_0/DOWN_1
| | `- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_0/UP_0
| |- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_1
| | |- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_1/DOWN_0
| | |- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_1/DOWN_1
| | `- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_1/UP_0
| |- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_2
| | |- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_2/DOWN_0
| | |- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_2/DOWN_1
| | `- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_2/UP_0
| `- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_3
| |- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_3/DOWN_0
| |- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_3/DOWN_1
| `- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_3/UP_0
`- /xyz/openbmc_project/sensors

root@nvl32-obmc:~# busctl -l introspect xyz.openbmc_project.GpuSensor /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_1/DOWN_0
NAME TYPE SIGNATURE RESULT/VALUE FLAGS
org.freedesktop.DBus.Introspectable interface - - -
.Introspect method - s -
org.freedesktop.DBus.Peer interface - - -
.GetMachineId method - s -
.Ping method - - -
org.freedesktop.DBus.Properties interface - - -
.Get method ss v -
.GetAll method s a{sv} -
.Set method ssv - -
.PropertiesChanged signal sa{sv}as - -
xyz.openbmc_project.Association.Definitions interface - - -
.Associations property a(sss) 1 "connected_to" "connecting" "/xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_1" emits-change
xyz.openbmc_project.Inventory.Connector.Port interface - - -
.PortProtocol property s "xyz.openbmc_project.Inventory.Connector.Port.PortProtocol.PCIe" emits-change
.PortType property s "xyz.openbmc_project.Inventory.Connector.Port.PortType.DownstreamPort" emits-change
.Speed property t 34359738368 emits-change
.Width property u 16 emits-change
```

Change-Id: I2845f090ac92c8ff6a742ec83c23073e6ea4e1b6
Signed-off-by: Harshit Aghera <haghera@nvidia.com>

show more ...

nvidia-gpu: Use common class for mctp endpoints

The common endpoint class should be used for send and receive.

Tested: Used by MctpRequester, tested on nvl32-obmc

Change-Id: I0060a66a5bcb4decfbe66

nvidia-gpu: Use common class for mctp endpoints

The common endpoint class should be used for send and receive.

Tested: Used by MctpRequester, tested on nvl32-obmc

Change-Id: I0060a66a5bcb4decfbe663d46ba88529e01e2209
Signed-off-by: Ed Tanous <etanous@nvidia.com>

show more ...

Remove redundant is_method_error() checks

The handlers registered through sdbusplus::bus::match_t only receive
D-Bus signals. Signal messages are never sent as method-error
replies, and therefore me

Remove redundant is_method_error() checks

The handlers registered through sdbusplus::bus::match_t only receive
D-Bus signals. Signal messages are never sent as method-error
replies, and therefore message.is_method_error() can never be true in
these callbacks.

This change removes all unnecessary is_method_error() checks from
signal handlers to simplify the code and avoid confusion.

Change-Id: I43e4a564c1bf401a5da9819dd201464e4a59c871
Signed-off-by: George Liu <liuxiwei@ieisystem.com>

show more ...

request maintainer role for nvidia-gpu

I have actively contributed to and reviewed patches for nvidia-gpu
application since its inception in May 2025. Additionally, I have
contributed and reviewed p

request maintainer role for nvidia-gpu

I have actively contributed to and reviewed patches for nvidia-gpu
application since its inception in May 2025. Additionally, I have
contributed and reviewed phosphor-dbus-interfaces and bmcweb patches
related to nvidia-gpu application.

Change-Id: I8eca227699b09c5cdb49495d5237a545c8609e86
Signed-off-by: Harshit Aghera <haghera@nvidia.com>

show more ...

Fix test build if nvidia-gpu is disabled

When nvidia-gpu is disabled, unit tests don't build because of the
shared gpusensor_sources variable. Make a quick fix to fix the build.
Going forward we ma

Fix test build if nvidia-gpu is disabled

When nvidia-gpu is disabled, unit tests don't build because of the
shared gpusensor_sources variable. Make a quick fix to fix the build.
Going forward we may need the option checking to be put into the sub
meson file rather than the top level, so that unit test deps can build
separately.

Change-Id: Ib87487fe15e80df44afbd9c3421163c6fbc16f74
Signed-off-by: Ed Tanous <etanous@nvidia.com>

show more ...

nvidia-gpu: fix GPU power PeakReading PDI usage

The GPU power peak reading, which uses the Telemetry.Report PDI, was
relying on a string ("PeakReading") to expose the reading. This string
is Redfish

nvidia-gpu: fix GPU power PeakReading PDI usage

The GPU power peak reading, which uses the Telemetry.Report PDI, was
relying on a string ("PeakReading") to expose the reading. This string
is Redfish specific. Instead, use the OperationType.Maximum enum defined
in the PDI. Bmcweb code can map this to PeakReading.

Tested: Build an image for nvl32-obmc machine with the following patches
cherry picked.

https://gerrit.openbmc.org/c/openbmc/openbmc/+/85490
https://gerrit.openbmc.org/c/openbmc/bmcweb/+/82449.

The patch cherry-picks the following patches that are currently under
review.

```
1. device tree
https://lore.kernel.org/all/aRbLqH8pLWCQryhu@molberding.nvidia.com/
2. mctpd patches
https://github.com/CodeConstruct/mctp/pull/85
3. u-boot changes
https://lore.kernel.org/openbmc/20251121-msx4-v1-0-fc0118b666c1@nvidia.com/T/#t
4. kernel changes as specified in the openbmc patch (for espi)
5. entity-manager changes
https://gerrit.openbmc.org/c/openbmc/entity-manager/+/85455
6. platform-init changes
https://gerrit.openbmc.org/c/openbmc/platform-init/+/85456
7. spi changes
https://lore.kernel.org/all/20251121-w25q01jv_fixup-v1-1-3d175050db73@nvidia.com/
```

The GPU Power PeakReading is correctly reported on DBus and on redfish.

Change-Id: I39b2b4987d845f878ffdedcfdb02cdfdc02a4499
Signed-off-by: Deepak Kodihalli <deepak.kodihalli.83@gmail.com>
Signed-off-by: Harshit Aghera <haghera@nvidia.com>

show more ...

nvidia-gpu: add support for ConnectX device

Add support to discover ConnectX devices and to populate PCIe interface
properties using Phosphor DBus Interface
xyz.openbmc_project.Inventory.Item.PCIeDe

nvidia-gpu: add support for ConnectX device

Add support to discover ConnectX devices and to populate PCIe interface
properties using Phosphor DBus Interface
xyz.openbmc_project.Inventory.Item.PCIeDevice.

ConnectX device has an integrated PCIe Switch. The patch uses
xyz.openbmc_project.Inventory.Item.PCIeSwitch PDI to define the PCIe
Switch resource.

Tested: Build an image for nvl32-obmc machine with the following patch
cherry picked.

https://gerrit.openbmc.org/c/openbmc/openbmc/+/85490

The patch cherry-picks the following patches that are currently under
review.

```
1. device tree
https://lore.kernel.org/all/aRbLqH8pLWCQryhu@molberding.nvidia.com/
2. mctpd patches
https://github.com/CodeConstruct/mctp/pull/85
3. u-boot changes
https://lore.kernel.org/openbmc/20251121-msx4-v1-0-fc0118b666c1@nvidia.com/T/#t
4. kernel changes as specified in the openbmc patch (for espi)
5. entity-manager changes
https://gerrit.openbmc.org/c/openbmc/entity-manager/+/85455
6. platform-init changes
https://gerrit.openbmc.org/c/openbmc/platform-init/+/85456
7. spi changes
https://lore.kernel.org/all/20251121-w25q01jv_fixup-v1-1-3d175050db73@nvidia.com/
```

```
root@nvl32-bmc:~# busctl tree xyz.openbmc_project.GpuSensor
`- /xyz
`- /xyz/openbmc_project
|- /xyz/openbmc_project/inventory
| `- /xyz/openbmc_project/inventory/pcie_devices
| |- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_0
| |- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_1
| |- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_2
| `- /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_3

root@nvl32-obmc:~# busctl introspect xyz.openbmc_project.GpuSensor /xyz/openbmc_project/inventory/pcie_devices/Nvidia_ConnectX_0
NAME TYPE SIGNATURE RESULT/VALUE FLAGS
org.freedesktop.DBus.Introspectable interface - - -
.Introspect method - s -
org.freedesktop.DBus.Peer interface - - -
.GetMachineId method - s -
.Ping method - - -
org.freedesktop.DBus.Properties interface - - -
.Get method ss v -
.GetAll method s a{sv} -
.Set method ssv - -
.PropertiesChanged signal sa{sv}as - -
xyz.openbmc_project.Inventory.Item.PCIeDevice interface - - -
.GenerationInUse property s "xyz.openbmc_project.Inventory.Item.P... emits-change
.GenerationSupported property s "xyz.openbmc_project.Inventory.Item.P... emits-change
.LanesInUse property u 8 emits-change
.MaxLanes property u 16 emits-change
xyz.openbmc_project.Inventory.Item.PCIeSwitch interface - - -

$ curl -s -k -u 'root:0penBmc' https://${bmc_ip}/redfish/v1/Systems/system/PCIeDevices/Nvidia_ConnectX_0
{
"@odata.id": "/redfish/v1/Systems/system/PCIeDevices/Nvidia_ConnectX_0",
"@odata.type": "#PCIeDevice.v1_19_0.PCIeDevice",
"Id": "Nvidia_ConnectX_0",
"Name": "PCIe Device",
"PCIeFunctions": {
"@odata.id": "/redfish/v1/Systems/system/PCIeDevices/Nvidia_ConnectX_0/PCIeFunctions"
},
"PCIeInterface": {
"LanesInUse": 8,
"MaxLanes": 16,
"MaxPCIeType": "Gen5",
"PCIeType": "Gen5"
},
"Status": {
"Health": "OK",
"State": "Enabled"
}
}%
```

Change-Id: Id89ce8a298ebb16934e94efcb9ca4679f91a7b26
Signed-off-by: Harshit Aghera <haghera@nvidia.com>

show more ...

nvidia-gpu: move unused member

Move MaxMessageSize to where it's used

Change-Id: I6c45157e6e3e52672cab86c82af1ea45a3628d19
Signed-off-by: Ed Tanous <etanous@nvidia.com>

nvidia-gpu: Declare send endpoint on stack

There's no reason to store this small class in between transactions.
Just construct on stack as part of the send.

Tested: On last patchset in series

Chan

nvidia-gpu: Declare send endpoint on stack

There's no reason to store this small class in between transactions.
Just construct on stack as part of the send.

Tested: On last patchset in series

Change-Id: I00090942665f022bfa2552b9c31c7c3da000646b
Signed-off-by: Ed Tanous <etanous@nvidia.com>

show more ...

Change copyright to match linux foundation

We should use SPDX identifiers wherever possible for simplification.

Change-Id: If3a7bfe506d7fded64a3ac929cc643834b16303e
Signed-off-by: Ed Tanous <etanou

Change copyright to match linux foundation

We should use SPDX identifiers wherever possible for simplification.

Change-Id: If3a7bfe506d7fded64a3ac929cc643834b16303e
Signed-off-by: Ed Tanous <etanous@nvidia.com>

show more ...

nvidia-gpu: add TLimit sensor properties

Add support for DMTF Redfish properties ReadingBasis and Implementation
for GPU TLimit sensor [1].

Property Implementation for TLimit is set to Synthesized

nvidia-gpu: add TLimit sensor properties

Add support for DMTF Redfish properties ReadingBasis and Implementation
for GPU TLimit sensor [1].

Property Implementation for TLimit is set to Synthesized because the GPU
incorporates intelligent logic that determines the temperature delta
from the first thermal management software slowdown event. TLimit is
derived from other reported GPU sensors, such as HBM, Tavg, and others.

DBus Interface definition -
https://gerrit.openbmc.org/c/openbmc/phosphor-dbus-interfaces/+/81658

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 -s -k -u 'root:0penBmc' https://10.137.203.137/redfish/v1/Chassis/NVIDIA_GB200_1/Sensors/temperature_NVIDIA_GB200_GPU_0_TEMP_1
{
"@odata.id": "/redfish/v1/Chassis/NVIDIA_GB200_1/Sensors/temperature_NVIDIA_GB200_GPU_0_TEMP_1",
"@odata.type": "#Sensor.v1_2_0.Sensor",
"Description": "Thermal Limit(TLIMIT) Temperature is the distance in deg C from the GPU temperature to the first throttle limit.",
"Id": "temperature_NVIDIA_GB200_GPU_0_TEMP_1",
"Implementation": "Synthesized",
"Name": "NVIDIA GB200 GPU 0 TEMP 1",
"Reading": 56.59375,
"ReadingBasis": "Headroom",
"ReadingRangeMax": 127.0,
"ReadingRangeMin": -128.0,
"ReadingType": "Temperature",
"ReadingUnits": "Cel",
"Status": {
"Health": "OK",
"State": "Enabled"
}
}%

root@gb200nvl-obmc:~# busctl introspect xyz.openbmc_project.GpuSensor /xyz/openbmc_project/sensors/temperature/NVIDIA_GB200_GPU_0_TEMP_1
NAME TYPE SIGNATURE RESULT/VALUE FLAGS
org.freedesktop.DBus.Introspectable interface - - -
.Introspect method - s -
org.freedesktop.DBus.Peer interface - - -
.GetMachineId method - s -
.Ping method - - -
org.freedesktop.DBus.Properties interface - - -
.Get method ss v -
.GetAll method s a{sv} -
.Set method ssv - -
.PropertiesChanged signal sa{sv}as - -
xyz.openbmc_project.Association.Definitions interface - - -
.Associations property a(sss) 1 "chassis" "all_sensors" "/xyz/openb... emits-change
xyz.openbmc_project.Inventory.Item interface - - -
.PrettyName property s "Thermal Limit(TLIMIT) Temperature is... emits-change
xyz.openbmc_project.Sensor.Type interface - - -
.Implementation property s "xyz.openbmc_project.Sensor.Type.Impl... emits-change
.ReadingBasis property s "xyz.openbmc_project.Sensor.Type.Read... emits-change
xyz.openbmc_project.Sensor.Value interface - - -
.MaxValue property d 127 emits-change
.MinValue property d -128 emits-change
.Unit property s "xyz.openbmc_project.Sensor.Value.Uni... emits-change
.Value property d 56.6836 emits-change writable
xyz.openbmc_project.Sensor.ValueMutability interface - - -
.Mutable property b true emits-change
xyz.openbmc_project.State.Decorator.Availability interface - - -
.Available property b true emits-change writable
xyz.openbmc_project.State.Decorator.OperationalStatus interface - - -
.Functional property b true emits-change
```

[1] : https://redfish.dmtf.org/schemas/v1/Sensor.v1_11_0.yaml

Change-Id: I1a16ced44c563794d561d26232a5e5fba041b875
Signed-off-by: Harshit Aghera <haghera@nvidia.com>

show more ...

nvidia-gpu: Fix thresholds for GPU_TEMP_1

Fixes thresholds for GPU_TEMP_1 to be upper critical, warning,
shutdown. Rather than lower critical, et al.

Change-Id: I580766288f3d27a48c75f00ea1dab13f028

nvidia-gpu: Fix thresholds for GPU_TEMP_1

Fixes thresholds for GPU_TEMP_1 to be upper critical, warning,
shutdown. Rather than lower critical, et al.

Change-Id: I580766288f3d27a48c75f00ea1dab13f0284bed6
Signed-off-by: Marc Olberding <molberding@nvidia.com>

show more ...

nvidia-gpu: Fix a number of object lifetime issues

Moves all subsensors and objects treated as shared_ptrs
to be using shared_from_this. This way, if there's an object lifetime
issue we don't segfau

nvidia-gpu: Fix a number of object lifetime issues

Moves all subsensors and objects treated as shared_ptrs
to be using shared_from_this. This way, if there's an object lifetime
issue we don't segfault.

Also separates construction and asio init for NvidiaSmaDevice
so that when we bind to this, its valid after we leave the ctor

Change-Id: I8e3115bc276d2e0eaac0b1dc9a9d2c46e6751d4b
Signed-off-by: Marc Olberding <molberding@nvidia.com>

show more ...

nvidia-gpu: NvidiaGpuDevice fix use after free

Fixes use after free for NvidiaGpuThresholds.
Moves the storage used for communication to be part
of the NvidiaGpuDevice class instead of ephemerally
p

nvidia-gpu: NvidiaGpuDevice fix use after free

Fixes use after free for NvidiaGpuThresholds.
Moves the storage used for communication to be part
of the NvidiaGpuDevice class instead of ephemerally
passed around through free functions

Also makes NvidiaGpuDevice inherit from
std::enable_shared_from_this

Testing: Issue found previous was coredumps
on nvl32-obmc. Asan discovered it was a use
after free in the shared pointer in ThermalLimits

Afterwards, no core dumps or issues reported by asan.
Ran on an nvl32-obmc model with 8 GPU's

Change-Id: I61b606f3a129499089718e7ec804926db5f22c64
Signed-off-by: Marc Olberding <molberding@nvidia.com>

show more ...

nvidia-gpu: deferred init for NvidiaGpuDevice

Adds deferred init for NvidiaGpuDevice, so that when we bind
to this, the this pointer is valid, i.e. after construction
is completed

Change-Id: I24a53

nvidia-gpu: deferred init for NvidiaGpuDevice

Adds deferred init for NvidiaGpuDevice, so that when we bind
to this, the this pointer is valid, i.e. after construction
is completed

Change-Id: I24a53d2ab9be1a2a4431368414a154b48347d2a2
Signed-off-by: Marc Olberding <molberding@nvidia.com>

show more ...

nvidia-gpu: Fix up buffering in MctpRequester

This change does a lot, for better or worse
1. Change MctpRequester to hold both buffers for send and receive
2. This requires changing the callback str

nvidia-gpu: Fix up buffering in MctpRequester

This change does a lot, for better or worse
1. Change MctpRequester to hold both buffers for send and receive
2. This requires changing the callback structure, so the reach is far
3. Changes error reporting to be through std::error_code
4. Collapses the QueuingRequeuster and Requeuster to be MctpRequeuster
5. Doing 4 gets rid of a level indirection and an extra unordered_map
6. Adds proper iid support, which is made significantly easier by 4/5
7. Fixes issues around expiry timer's where we would cancel the timer
for a given request whenever a new packet would come in to be sent.
This could cause lockup if a packet truly did time out and an
interleaved packet finished sending. This moves each queue
to have its own timer.

This fixes an issue where we were receiving buffers in from clients
and then binding them to receive_calls without ensuring that they
are the correct message, thus when receive was called, it was called
with the last bound buffer to async_receive_from. This would cause a
number of issues, ranging from incorrect device discovery results
to core dumps as well as incorrect sensor readings.

This change moves the receive and send buffers to be owned by
the MctpRequester, and a non-owning view is provided via
callback to the client. All existing clients just decode in place
given that buffer.

Tested: loaded onto nvl32-obmc. Correct number of sensors showed up
and the readings were nominal

Change-Id: I67c843691ca79e9fcccfa16df6d611918f25f6ca
Signed-off-by: Marc Olberding <molberding@nvidia.com>

show more ...

nvidia-gpu: add Power Sensor PeakReading Property

Add support for Sensor Properties PeakReading and PeakRedingTime.

Current Limitation -
The ResetMetrics action is currently not supported for Redfi

nvidia-gpu: add Power Sensor PeakReading Property

Add support for Sensor Properties PeakReading and PeakRedingTime.

Current Limitation -
The ResetMetrics action is currently not supported for Redfish URIs in
bmcweb. As a result, the ability to clear PeakReading values for GPU
Power Sensors has not been implemented.

Future Consideration -
If ResetMetrics action support is added to bmcweb in the future, the
corresponding functionality will also need to be implemented in the
dbus-sensor application to ensure full compatibility.

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

```
root@gb200nvl-obmc:~# busctl introspect xyz.openbmc_project.GpuSensor /xyz/openbmc_project/sensors/power/NVIDIA_GB200_GPU_0_Power_0
NAME TYPE SIGNATURE RESULT/VALUE FLAGS
org.freedesktop.DBus.Introspectable interface - - -
.Introspect method - s -
org.freedesktop.DBus.Peer interface - - -
.GetMachineId method - s -
.Ping method - - -
org.freedesktop.DBus.Properties interface - - -
.Get method ss v -
.GetAll method s a{sv} -
.Set method ssv - -
.PropertiesChanged signal sa{sv}as - -
xyz.openbmc_project.Association.Definitions interface - - -
.Associations property a(sss) 1 "chassis" "all_sensors" "/xyz/openb... emits-change
xyz.openbmc_project.Sensor.Value interface - - -
.MaxValue property d 5000 emits-change
.MinValue property d 0 emits-change
.Unit property s "xyz.openbmc_project.Sensor.Value.Uni... emits-change
.Value property d 29.194 emits-change writable
xyz.openbmc_project.Sensor.ValueMutability interface - - -
.Mutable property b true emits-change
xyz.openbmc_project.State.Decorator.Availability interface - - -
.Available property b true emits-change writable
xyz.openbmc_project.State.Decorator.OperationalStatus interface - - -
.Functional property b true emits-change
xyz.openbmc_project.Telemetry.Report interface - - -
.Readings property (ta(ssdt)) 0 1 "PeakReading" "" 80.933 0 emits-change
```

Change-Id: I0a4f7eb0a5db688f32bf80954839140da9bb7e2a
Signed-off-by: Harshit Aghera <haghera@nvidia.com>

show more ...

Fix tidy build

This appears to be something tidy is wrong about. The suggestion of
adding math to the struct initializers appears to not compile.

Move the calculation of hysteresisTrigger and hyst

Fix tidy build

This appears to be something tidy is wrong about. The suggestion of
adding math to the struct initializers appears to not compile.

Move the calculation of hysteresisTrigger and hysteresisPublish into the
constructor body itself to avoid the warning.

Change-Id: I833fd12966c69c0e081692d6d40ba0cf1805ead1
Signed-off-by: Ed Tanous <etanous@nvidia.com>

show more ...

Move Nvidia gpu tests

These tests got caught in the refactor. Move these tests to the correct
location.

Change-Id: Ie8ec10e154d60cb4f24e1f45be36240863438f87
Signed-off-by: Ed Tanous <etanous@nvidi

Move Nvidia gpu tests

These tests got caught in the refactor. Move these tests to the correct
location.

Change-Id: Ie8ec10e154d60cb4f24e1f45be36240863438f87
Signed-off-by: Ed Tanous <etanous@nvidia.com>

show more ...

Remove main

Unit tests don't build if main is enabled.

Change-Id: I4c7210b2a72032d6e15729b5ab5e4201739dd602
Signed-off-by: Ed Tanous <etanous@nvidia.com>