tree: 7b319e3f74bf7f8e21e390ae5ef2d62009938a6c [path history] [tgz]
  1. Device.errors.yaml
  2. Device.metadata.yaml
  3. GPIO.errors.yaml
  4. GPIO.metadata.yaml
  5. IIC.errors.yaml
  6. IIC.metadata.yaml
  7. Inventory.errors.yaml
  8. Inventory.metadata.yaml
  9. IPMISensor.errors.yaml
  10. IPMISensor.metadata.yaml
  11. README.md
xyz/openbmc_project/Error/Callout/README.md

Introduction

A callout is typically an indication of a faulty hardware component in a system. In openbmc, a callout is defined as any other error, via a YAML file. An example would be xyz.openbmc_project.Error.Callout.IIC, to indicate an IIC callout.

The goal is to have applications post callouts using hardware terminology which is familiar to them, such as a sysfs entry, or an IIC address. It would be up to the openbmc error handling component to map such a callout to actual field replacable units (FRUs) on the system.

Architecture and usage

An openbmc error has associated metadata, the same is true for a callout. Such metadata would be defined in the callout YAML interface. Here is an example (for xyz.openbmc_project.Error.Callout.IIC) :

- name: IIC
  meta:
    - str: "CALLOUT_IIC_BUS=%s"
      type: string
    - str: "CALLOUT_IIC_ADDR=%hu"
      type: uint16

An application wanting to add an IIC callout will have to provide values for the metadata fields above. These fields will also let the error handling component figure out that this is in fact an IIC callout.

A callout is typically associated with an error log. For eg, xyz.openbmc_project.Error.Foo may want to add an IIC callout. This is indicated in Foo's YAML interface as follows :

- name: Foo
  description: this is the error Foo
  inherits:
    - xyz.openbmc_project.Error.Callout.IIC

The way this inheritance will be implemented is that, Foo's metadata will include Callout.IIC's as well, so an application wanting to add an IIC callout will have to provide values for Foo and IIC metadata. Like mentioned before, due to the presence of the Callout.IIC metadata, the error handling component can figure out that the error Foo includes an IIC callout.

Currently defined callout interfaces in turn inherit xyz.openbmc_project.Error.Callout.Device, which has metadata common to callouts :

- name: Device
  meta:
    - str: "CALLOUT_ERRNO=%d"
      type: int32
    - str: "CALLOUT_DEVICE_PATH=%s"
      type: string

This way, say an application wants to express an IIC callout in terms of a device path, for lack of IIC information. The application can add the callout metadata fields for both Callout.Device and Callout.IIC, but provide values only for Callout.Callout. That way the error handling component can still decipher this as an IIC callout.

Creation of a callout

This section talks about creation of a callout, once callout related metadata is already in the journal.

Taking an example of a generic device callout here, but this would be the flow in general :

  • An application commits an error that has associated callout metadata. This will cause the error-log server to create a d-bus object for the error.

  • The error-log server will detect that callout metadata is present, will extract the same and hand it over to a sub-module which will map callout metadata to one or more inventory object paths, and will create an association between the error object and the inventory object(s). The mapping from callout metadata to inventory objects is mostly done via the aid of code generated by the system MRW parsers.

  • Generated code : consider a case where an application wants to callout an eeprom on the BMC planar, via a device path, such as /sys/devices/platform/ahb/ahb:apb/1e78a000.i2c/i2c-11/i2c-11/11-0051/eeprom. This would have to be mapped to the BMC planar as the FRU to be called out. MRW parser(s) could be written which, for every device in the IIC subsystem, can provide a corresponding inventory object path. The error-log server in this case has to, by looking at the device path, determine that the device is on an IIC bus, and make use of the code generated to map the device to inventory objects. Similar MRW parsers could be written for other device subsystems.