On Wed, Aug 04, 1999 at 09:28:49AM -0400, D'Arcy J.M. Cain wrote:
> Same thing I see in the same address neighbourhood.  How much memory and
> swap do you have?

32MB RAM, swap:
Device      1K-blocks     Used    Avail Capacity  Priority
/dev/sd2b      100700     1412    99288     1%    0
/dev/sd1b      100700     1464    99236     1%    0
/dev/sd0b       51480        4    51476     0%    1
Total          252880     2880   250000     1%

> 
> > System is a p100 with 100Mbs DEC 21140A ethernet, running www, ftp and sup
> > services.
> > www traffic is moderate, but ftp traffic has increased a lot in the last
> > few days, especially due to what looks like a squid cache which seems
> > to mirror the whole NetBSD 1.4 tree with several concurent connections.
> 
> Mine runs www, Radius and PostgreSQL.  I suspect that the actual bug has
> nothing to do with the actual programs running but rather with the amount
> of activity.
> 
> I wrote a program to allocate lots of memory and write something in it
> and watched with top -osize while it chewed up 127M of memory.  At that
> point it starting failing to allocate but before that it accessed swap
> so my original thought that there was something wrong with swap functions
> seems to be incorrect.  I am really at a loss.
> 

If bill's theory is correct (and I fear it is), spl being lowered at
the wrong time can cause various kind of failures, including probably
corrupting some VM internal datas. 

--
Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
--