On the BMC, PELs are created from the standard event logs provided by phosphor-logging using a message registry that provides the PEL related fields. The message registry is a JSON file.
A component ID is a 2 byte value of the form 0xYY00 used in a PEL to:
BD SRCs.Component IDs are specified in the message registry either as the upper byte of the SRC reason code field for BD SRCs, or in the standalone ComponentID field.
Component IDs will be unique on a per-repository basis for errors unique to that repository. When the same errors are created by multiple repositories, those errors will all share the same component ID. The master list of component IDs is here. That file can used by PEL parsers to display a name for the component ID. The 'O' in the name is the creator ID value for BMC created PELs.
The message registry schema is here, and the message registry itself is here. The schema will be validated either during a bitbake build or during CI, or eventually possibly both.
In the message registry, there are fields for specifying:
This is the key into the message registry, and is the Message property of the OpenBMC event log that the PEL is being created from.
"Name": "xyz.openbmc_project.Power.Fault"
This field is part of the PEL User Header section, and is used to specify the subsystem pertaining to the error. It is an enumeration that maps to the actual PEL value. If the subsystem isn't known ahead of time, it can be passed in at the time of PEL creation using the 'PEL_SUBSYSTEM' AdditionalData field. In this case, 'Subsystem' isn't required, though 'PossibleSubsystems' is.
"Subsystem": "power_supply"
This field is used by scripts that build documentation from the message registry to know which subsystems are possible for an error when it can't be hardcoded using the 'Subsystem' field. It is mutually exclusive with the 'Subsystem' field.
"PossibleSubsystems": ["memory", "processor"]
This field is part of the PEL User Header section, and is used to specify the PEL severity. It is an optional field, if it isn't specified, then the severity of the OpenBMC event log will be converted into a PEL severity value.
It can either be the plain severity value, or an array of severity values that are based on system type, where an entry without a system type will match anything unless another entry has a matching system type.
"Severity": "unrecoverable"
Severity": [ { "System": "system1", "SevValue": "recovered" }, { "Severity": "unrecoverable" } ]
The above example shows that on system 'system1' the severity will be recovered, and on every other system it will be unrecoverable.
This is an optional field and is used to override the Severity field when a specific manufacturing isolation mode is enabled. It has the same format as Severity.
"MfgSeverity": "unrecoverable"
This field is part of the PEL User Header section, and is used to specify the event scope, as defined by the PEL spec. It is optional and defaults to "entire platform".
"EventScope": "entire_platform"
This field is part of the PEL User Header section, and is used to specify the event type, as defined by the PEL spec. It is optional and defaults to "not applicable" for non-informational logs, and "misc_information_only" for informational ones.
"EventType": "na"
This field is part of the PEL User Header section, and is used to specify the PEL action flags, as defined by the PEL spec. It is an array of enumerations.
The action flags can usually be deduced from other PEL fields, such as the severity or if there are any callouts. As such, this is an optional field and if not supplied the code will fill them in based on those fields.
In fact, even if supplied here, the code may still modify them to ensure they are correct. The rules used for this are here.
"ActionFlags": ["service_action", "report", "call_home"]
This is an optional field and is used to override the Action Flags field when a specific manufacturing isolation mode is enabled.
"MfgActionFlags": ["service_action", "report", "call_home"]
This is the component ID of the PEL creator, in the form 0xYY00. For BD SRCs, this is an optional field and if not present the value will be taken from the upper byte of the reason code. If present for BD SRCs, then this byte must match the upper byte of the reason code.
"ComponentID": "0x5500"
This specifies the type of SRC to create. The type is the first 2 characters of the 8 character ASCII string field of the PEL. The allowed types are BD, for the standard OpenBMC error, and 11, for power related errors. It is optional and if not specified will default to BD.
Note: The ASCII string for BD SRCs looks like: BDBBCCCC, where:
For 11 SRCs, it looks like: 1100RRRR, where RRRR is the SRC reason code.
"Type": "11"
This is the 4 character value in the latter half of the SRC ASCII string. It is treated as a 2 byte hex value, such as 0x5678. For BD SRCs, the first byte is the same as the first byte of the component ID field in the Private Header section that represents the creator's component ID.
"ReasonCode": "0x5544"
The symptom ID is in the Extended User Header section and is defined in the PEL spec as the unique event signature string. It always starts with the ASCII string. This field in the message registry allows one to choose which SRC words to use in addition to the ASCII string field to form the symptom ID. All words are separated by underscores. If not specified, the code will choose a default format, which may depend on the SRC type.
For example: ["SRCWord3", "SRCWord9"] would be: <ASCII_STRING>_<SRCWord3>_<SRCWord9>, which could look like: B181320_00000050_49000000.
"SymptomIDFields": ["SRCWord3", "SRCWord9"]
In a PEL, these SRC words are free format and can be filled in by the user as desired. On the BMC, the source of these words is the AdditionalData fields in the event log. The message registry provides a way for the log creator to specify which AdditionalData property field to get the data from, and also to define what the SRC word means for use by parsers. If not specified, these SRC words will be set to zero in the PEL.
"Words6to9": { "6": { "description": "Failing unit number", "AdditionalDataPropSource": "PS_NUM" } }
Bit 6 in hex word 5 of the SRC means that one or more called out resources have been deconfigured, and this flag can be used to set that bit. The only other way to set it is by indicating it when passing in the callouts via JSON.
This is looked at by the software that creates the periodic PELs that indicate a system is running with deconfigured hardware.
"DeconfigFlag": true
The documentation fields are used by PEL parsers to display a human readable description of a PEL. They are also the source for the Redfish event log messages.
This field is used by the BMC's PEL parser as the description of the error log. It will also be used in Redfish event logs. It supports argument substitution using the %1, %2, etc placeholders allowing any of the SRC user data words 6 - 9 to be displayed as part of the message. If the placeholders are used, then the MessageArgSources property must be present to say which SRC words to use for each placeholder.
"Message": "Processor %1 had %2 errors"
This optional field is required when the Message field contains the %X placeholder arguments. It is an array that says which SRC words to get the placeholders from. In the example below, SRC word 6 would be used for %1, and SRC word 7 for %2.
"MessageArgSources": [ "SRCWord6", "SRCWord7" ]
A short description of the error. This is required by the Redfish schema to generate a Redfish message entry, but is not used in Redfish or PEL output.
"Description": "A power fault"
This is an optional free format text field for keeping any notes for the registry entry, as comments are not allowed in JSON. It is an array of strings for easier readability of long fields.
"Notes": [ "This entry is for every type of power fault.", "There is probably a hardware failure." ]
The callout fields allow one to specify the PEL callouts (either a hardware FRU, a symbolic FRU, or a maintenance procedure) in the entry for a particular error. These callouts can vary based on system type, as well as a user specified AdditionalData property field. Callouts will be added to the PEL in the order they are listed in the JSON. If a callout is passed into the error, say with CALLOUT_INVENTORY_PATH, then that callout will be added to the PEL before the callouts in the registry.
There is room for up to 10 callouts in a PEL.
"Callouts": [ { "System": "system1", "CalloutList": [ { "Priority": "high", "LocCode": "P1-C1" }, { "Priority": "low", "LocCode": "P1" } ] }, { "CalloutList": [ { "Priority": "high", "Procedure": "SVCDOCS" } ] } ]
The above example shows that on system 'system1', the FRU at location P1-C1 will be called out with a priority of high, and the FRU at P1 with a priority of low. On every other system, the maintenance procedure SVCDOCS is called out.
"CalloutsUsingAD": { "ADName": "PROC_NUM", "CalloutsWithTheirADValues": [ { "ADValue": "0", "Callouts": [ { "CalloutList": [ { "Priority": "high", "LocCode": "P1-C5" } ] } ] }, { "ADValue": "1", "Callouts": [ { "CalloutList": [ { "Priority": "high", "LocCode": "P1-C6" } ] } ] } ] }
This example shows that the callouts were selected based on the 'PROC_NUM' AdditionalData field. When PROC_NUM was 0, the FRU at P1-C5 was called out. When it was 1, P1-C6 was called out. Note that the same 'Callouts' array is used as in the previous example, so these callouts can also depend on the system type.
If it's desired to use a different set of callouts when there isn't a match on the AdditionalData field, one can use CalloutsWhenNoADMatch. In the following example, the 'air_mover' callout will be added if 'PROC_NUM' isn't 0. 'CalloutsWhenNoADMatch' has the same schema as the 'Callouts' section.
"CalloutsUsingAD": { "ADName": "PROC_NUM", "CalloutsWithTheirADValues": [ { "ADValue": "0", "Callouts": [ { "CalloutList": [ { "Priority": "high", "LocCode": "P1-C5" } ] } ] }, ], "CalloutsWhenNoADMatch": [ { "CalloutList": [ { "Priority": "high", "SymbolicFRU": "air_mover" } ] } ] }
This field can be used to modify the failing component type field in the callout when the default doesn't fit:
{ "Priority": "high", "Procedure": "FIXIT22" "CalloutType": "config_procedure" }
The defaults are:
A special case is when one wants to use a symbolic FRU callout with a trusted location code, but the location code to use isn't known until runtime. This means it can't be specified using the 'LocCode' key in the registry.
In this case, one should use the 'SymbolicFRUTrusted' key along with the 'UseInventoryLocCode' key, and then pass in the inventory item that has the desired location code using the 'CALLOUT_INVENTORY_PATH' entry inside of the AdditionalData property. The code will then look up the location code for that passed in inventory FRU and place it in the symbolic FRU callout. The normal FRU callout with that inventory item will not be created. The symbolic FRU must be the first callout in the registry for this to work.
{ "Priority": "high", "SymbolicFRUTrusted": "AIR_MOVR", "UseInventoryLocCode": true }
The PEL daemon can be told to capture pieces of the journal in PEL UserData sections. This could be useful for debugging problems where a BMC dump which would also contain the journal isn't available.
The 'JournalCapture' field has two formats, one that will create one UserData section with the previous N lines of the journal, and another that can capture any number of journal snippets based on the journal's SYSLOG_IDENTIFIER field.
"JournalCapture": { "NumLines": 30 }
"JournalCapture": { "Sections": [ { "SyslogID": "phosphor-bmc-state-manager", "NumLines": 20 }, { "SyslogID": "phosphor-log-manager", "NumLines": 15 } ] }
The first example will capture the previous 30 lines from the journal into a single UserData section.
The second example will create two UserData sections, the first with the most recent 20 lines from phosphor-bmc-state-manager, and the second with 15 lines from phosphor-log-manager.
If a UserData section would make the PEL exceed its maximum size of 16KB, it will be dropped.
The general process for adding new entries to the message registry is:
Update message_registry.json to add the new errors.
If a new component ID is used (usually the first byte of the SRC reason code), document it in O_component_ids.json.
Validate the file. It must be valid JSON and obey the schema. The validate_registry.py script in extensions/openpower-pels/registry/tools will validate both, though it requires the python-jsonschema package to do the schema validation. This script is also run to validate the message registry as part of CI testing.
./tools/validate_registry.py -s schema/schema.json -r message_registry.json
One can test what PELs are generated from these new entries without writing any code to create the corresponding event logs:
Copy the modified message_registry.json into /etc/phosphor-logging/ on the BMC. That directory may need to be created.
Use busctl to call the Create method to create an event log corresponding to the message registry entry under test.
busctl call xyz.openbmc_project.Logging /xyz/openbmc_project/logging \ xyz.openbmc_project.Logging.Create Create ssa{ss} \ xyz.openbmc_project.Common.Error.Timeout \ xyz.openbmc_project.Logging.Entry.Level.Error 1 "TIMEOUT_IN_MSEC" "5"
Check the PEL that was created using peltool.
When finished, delete the file from /etc/phosphor-logging/.