Re: SB1K crash kernel panic - memory or CPU ? Andrew Tyson wrote:
> Hi,
>
> My SB1K ( dual 750MHz/4GB running S9 08/03 patched late last year) just
> crashed for no apparent reason. I have appended the relevant
> /var/adm/messages log at the end of the post. I tried using adb which
> came up with the following;
>
> Jan 26 00:07:48 horus SUNW,UltraSPARC-III: [ID 326222 kern.warning]
> WARNING: [AF
> T1] Timeout (TO) Event detected by CPU0 Privileged Data Access at TL=0,
> errID 0x
> 00000c74.79225e80
> Jan 26 00:07:48 horus AFSR 0x00001000<TO>.00000000 AFAR
> 0x000007f8.00610900
> Jan 26 00:07:48 horus Fault_PC 0x134061c
> Jan 26 00:07:48 horus unix: [ID 836849 kern.notice]
> Jan 26 00:07:48 horus ^Mpanic[cpu0]/thread=2a10007dd40:
> Jan 26 00:07:48 horus unix: [ID 855895 kern.notice] [AFT1] errID
> 0x00000c74.7922
> 5e80 TO Error(s)
This is a TO (timeout) error from the Safari databus from an access to
an address that claimed to be mapped (ie, some device claims to have
implemented that physical address range) but for which data did not return
within the time expected. The fault address 0x000007f8.00610900 I think looks
like a device address rather than real cacheable memory, but I don't know
which device that would map to. Chances are that there's a bad or marginal
PCI card, I'd guess.
Incidentally Solaris 10 makes all this hugely more elegant.
Gavin |