Subject: Re: segmap alloc failed
To: Greg Wohletz <greg@duke.CS.UNLV.EDU>
From: Michael L. Hitch <mhitch@lightning.oscs.montana.edu>
List: port-pmax
Date: 07/20/1997 12:20:56
On Jul 18,  3:03pm, Greg Wohletz wrote:
> I can try this patch, but I'm still curious as to what type of events can
> cause this memory shortfall to occur.  The system doesn't do anything other
> than run innd.  Innd is typically consumes about 20meg of vm space, the
> system has 48meg of ram.  he panic's aren't happening at a time associated
> with any particular cron even (like expire).  Typically when the system is
> running vmstat reports around 20meg free ram, It rarely has to do any paging.

  This problem will occur when the `vm_page_queue_free' list is empty.  I'm
not nearly familiar enough with the VM system to know what conditions can
cause this.  Something must be requesting a lot of physical memory if your
system normally reports 20MB free memory.

  I've been running with a modified version of pmap.c that reports when
the segment table or segment map can't be obtained, and have been finally
able to reproduce this occurance.  The only way I've been able to get the
segmap allocation to fail is by running multiple makes - usually a make -j?
at the top of the source tree and a parallel make -j? in the kernel tree
(the -j? values were between 8 and 16 if I remember correctly).  My 40MB
system is currently reporting about 22MB free memory sitting idle, but
when I run the multiple makes, it runs much lower.

Michael

-- 
Michael L. Hitch			mhitch@montana.edu
Computer Consultant
Information Technology Center
Montana State University	Bozeman, MT	USA