[lug] kernel paging request errors

[D. Stimits]

rm at mamma.varadinet.de wrote:
> 
> On Tue, Mar 27, 2001 at 07:17:54AM -0700, D. Stimits wrote:
> > dan radom wrote:
> > >
> > > hi,
> > >
> > > every 10 minutes i get the following error in my logs...
> > >
> > > Mar 26 15:50:01 mars kernel: Unable to handle kernel paging request at virtual address 5760c297
> >
> > Does the address "5760c297" consistently appear? Is it always that
> > address? If and only if it is that address at all times, do you have
> > more than one DIMM or other memory type that you can swap locations with
> > (if so, physically swap memory and see if the address changes or the
> > machine has other problems that then show up)?
> >
> 
> Hmm, the error messages  is from the file fault.c in the architecture
> dependend parts of your kernel sources <arch>/mm/fault.c in the
> function 'do_page_fault(...' This is a central function for the
> memory manaegment. The comment here is:
> 
> /* ops. The kernel tried to access some bad page. We'll have to
>  * terminate things with extreme prejudice.
>  *
>  * First we check if it was the bootup rw-test, though..
>  */
> 
> which of course can mean a lot of things. It _could_ mean bad RAM
> but i also could mean _non_existing_ RAM. I had a similar problem
> recently with a server that had plenty of RAM. Unfortunately the
> super-new BIOS couldn't handle it. The box would boot up with a
> random amount of memory available. Somtimes the whole RAM was there,
> sometimes only 55Mb ... Adding an expicit mem=521M to the kernel
> startup parameters fixed it. What does 'free' show?
> BTW, the address in the error report won't be the same (exept in a
> very rare coincidence); as the message states, this is a _virtual_
> address, not the hardware address.
> 
>  Ralf
> 

In the case of sometimes finding ram but otherwise not, there is also a
possibility that the ram is marginal at the given operating speed.
Possibly setting the BIOS to an extra wait state when accessing ram
could give a clue; if a longer wait state makes it find the ram, or if
memory errors go away after using longer wait, then it could just be
clocking too fast.

D. Stimits, stimits at idcomm.com