[fm-discuss] Re: Re: FMRI Reference Documentation?

[Keith M Wesolowski]

On Tue, Sep 26, 2006 at 12:40:20PM -0700, Robert Cibrario wrote:

> Interesting. From the PCI dictionary I see the following codes (see below) - What precisely is transmitted then in the trap?
> rc
> 
>      FMDICT: name=PCI version=1 maxkey=3 dictid=0x5043
>       
>      defect.io.pci.driver=0
>      fault.io.pci.bus=1
>      defect.io.pci.driver fault.io.pci.bus=2
>      fault.io.pci.device=3
>      defect.io.pci.driver fault.io.pci.device=4
>      fault.io.pci.bus fault.io.pci.device=5
>      defect.io.pci.driver fault.io.pci.bus fault.io.pci.device=6
>      fault.io.pci.device-interr=7
>      fault.io.pci.bus-linkerr fault.io.pci.device-interr=8
>      fault.io.pci.device-invreq fault.io.pci.device-noresp=9
>      fault.io.pci.device-interr fault.io.pci.device-invreq=10
>      fault.io.pci.device-noresp=11

The trap itself contains only three fields: the diagnosis UUID, the
fault code, and a reference URL (by default,
http://www.sun.com/msg/<code>).  There is substantial additional
information available in the MIB, which is not enabled by default.
You can find the entire MIB, which defines the traps, 4 tables, and a
scalar, in /etc/sma/snmp/mibs/SUN-FM-MIB.mib.

As for the information from the dictionaries in usr/src/cmd/fm/dicts,
the fault classes you list above (defect.io.pci.driver, for example)
will appear as the sunFmFaultEventClass field; each diagnosis will
have one or more corresponding sunFmFaultEventEntry objects.  However,
what you want to look at in this case are the corresponding *.po
files, which contain (msgid, msgstr) pairs for each code.  The IDs are
of the form <code>.<keyword>; the strings are dumped into the syslog
and could conceivably be made available via the MIB as well.

Currently, we only use the global "snmp-url" property to assign the
base URL (to which the code will be appended).  None of the per-code
data is made available by the MIB, though it is logged.  See
usr/src/cmd/fm/modules/common/{syslog-msgs,snmp-trapgen}.  This is the
same data available at the URL encoded within the trap.

So the RFE here, which you should file at bugs.opensolaris.org, is to
include the code-specific information from the dictionary in the MIB,
in the sunFmProblemEntry objects.  Would that address your concerns,
or is the stability of the codes still an issue as well?

-- 
Keith M Wesolowski		"Sir, we're surrounded!" 
FishWorks			"Excellent; we can attack in any direction!"