Panic caused by bad memory?

[Charles Sprickman]

On Wed, 25 Oct 2006, John Baldwin wrote:

> On Wednesday 25 October 2006 02:28, Charles Sprickman wrote:
>> On Tue, 24 Oct 2006 perryh at pluto.rain.com wrote:
>>
>>>> I can't get a kernel dump since it fails like this each time:
>>>>
>>>> dumping to dev #da/0x20001, offset 2097152
>>>> dump 1024 1023 1022 1021 Aborting dump due to I/O error.
>>>> status == 0xb, scsi status == 0x0
>>>> failed, reason: i/o error
>>>
>>> Bad memory seems unlikely to cause an I/O error trying to write the
>>> dump to the swap partition.  I'd guess a dicey drive -- and bad
>>> swap space could also account for the original crash.  You might
>>> be able to get a backup by booting single user, provided nothing
>>> activates the (presumably bad) swap partition.
>>
>> Just for the record, this box is running an Adaptec raid controller (2005S
>> - ZCR card) and swap is coming off a mirrored array.
>>
>> Coincidentally, I have a utility box where it had bad blocks on the swap
>> partition (but no others) - what I saw there is that the box would just
>> hang and spit out a bunch of "swap_pager timeout" messages to the console.
>> Quick and dirty remote fix while waiting for a drive?  Run file-backed
>> swap on /usr. :)
>>
>> Let's pretend for a minute it's not the drive that's the root cause...
>> Not saying it isn't - we're none too thrilled with these Adaptec RAID
>> controllers...  Do those memory addresses in the panic message point
>> towards bad memory if they are always the same?
>
> No, they are virtual addresses.  Having the same EIP means you are crashing in
> the same place.  Did you recently kldunload a module before it crashed?

Same place == same code?  The only change on this box was a massive 
portupgrade which included apache, php, mysql, postgres and most of the 
additional gnu tools.

There is one module that someone set to load on boot, and that's the 
linuxolator.  I have disabled that in rc.conf for now and we'll see what 
happens after the next panic.

We also have a few sticks of RAM on order now...

Thanks,

Charles

> -- 
> John Baldwin
> _______________________________________________
> freebsd-hackers at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-hackers
> To unsubscribe, send any mail to "freebsd-hackers-unsubscribe at freebsd.org"
>