[fm-discuss] PCI errors?

[Richard Elling]

Matty wrote:
> After rebooting one of my V880s this morning, I noticed that the fault
> manager detected some issues with a PCI bus [1][2][3][4]. Are there
> any documents available that describe how to interpret the fmdump
> output? Also, is there a reason fmadm doesn't actually list the bus /
> motherboard as being faulty?
>   

Nit: the SCSI controller on the MB is identified as possibly being
faulty:

   13%  fault.io.pci.device

        Problem in: hc:///motherboard=0/hostbridge=0/pcibus=1/pcidev=1/pcifn=0
           Affects: dev:////pci@8,700000/scsi@1
               FRU: hc:///component=MB


Note that all of the suspect parts share the same PCI bus (pcibus=1)

One of the problems with busses, like PCI (and unlike PCI-Express)
is that if a device hangs the bus, then it may be difficult to isolate.
Hence, the message ID information (PCI-8000-42) shows that we
suspect a number of devices, but can't exactly tell which one is defective.
    http://www.sun.com/msg/PCI-8000-42

I can't speak for the V880 diagnosis engine. But the system does support
hot-plug of the PCI cards.  As such, each card slot has a set of diagnosis
LEDs which show the state of the card.  You might take a peek.
    http://docs.sun.com/app/docs/doc/806-6592-11

 -- richard