[KLUG Members] Page fault questions.

[Adam Tauno Williams]

>I've got a box that I had to reboot this morning because of a page fault.  

Yikes.

>Here's the system:
>ASUS motherboard
>Athlon 1.1 GHz
>768 MB RAM
>SCSI RAID (on a dpt decade raid card :( ) with (I believe) 4 disks.
>So: postmaster 'crashed' as a result of a page fault in the system.  A PG 
>process was still left 'running' although you couldn't kill it. So I 
>reboot.  fsck complains, it gets fscked and fixed, comes back up. (and 
>before I rebooted it, I tried to fsck a partition after I unmounted it - 
>fsck seg faulted, page fault).  

Hrmm... that is a good one.  Does dmesg say anything?  Can you "dd if=/dev/sd??
of=/dev/null bs=512"?

>Syslog doesn't show any drive or swap errors (Are swap errors logged to 
>the syslog?) so I'm guessing right now that it's RAM.  The RAID card isn't 

That would be my first guess.

>barking, so I know it's not a failed drive.  Could my swap partition have 
>become corrupted?  Should I, late at night, do a  swapoff and then 

Sure, but after a reboot it doesn't matter what is in your swap partition.  The
kernel should only read from blocks it has already written pages to.

>re-format the swap partition? Run a badblocks check on it?  I need to know 

If fsck page faults I'm guessing that badblock will too.

>what this is, but I don't have these problems very often, especially with 
>this hardware configuration (specific SCSI card, etc).  My coworker is 
>pretty much sure it's the SCSI disks, but of course, doesn't know which 

Then why doesn't the RAID controller complain?  Seems unlikely to be a disk to me.

>one, and says that a swap error won't go to the syslog (i.e. a bad block, 
>or something). Yet, it can log the page fault, so...  

It will go to dmesg certainly.