On Sat, Aug 12, 2006 at 01:24:15PM -0700, Chris Stromsoe wrote: > I have two cards that detect as 0007 in the particular slot and one that > detects as 000f. Moving a 0007 to a different slot detects it as 000f. > The slots are on a 64-bit pci riser card, so it's possible that there is a > bad connection on one of the fingers. I'll try the boards in a different > machien and see it makes any difference. Definitely sounds like dodgy connections. > >I don't like the look of this. If you're getting ffffffff back from > >those particular registers, the chip has clearly gone fatal. But I use > >875 controllers *all the time*, so there's something odd going on here. > >I don't know what to suggest at this point, I'm afraid. > > Whenever the cards are detected as 000f they have given me that error at > boot. The cards came from Sun and are running Sun firmware if that makes > any difference at all. Up until a week ago they were all functioning in a > couple of Sun E450 running Solaris 9 without any visible problems. Actually, there's no firmware on the cards; the firmware is built into the driver and uploaded to the onboard ram at initialisation time. If there's dodgy connectors, then the card probably detected a PCI parity error and went fatal. I think it'd be useful to leverage the PCI error recovery framework to at least display a useful error message to the user on systems with little or no support for error *recovery*. - : send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
