Matthias Scheler writes:
> On Tue, Oct 11, 2005 at 10:09:34PM +1000, Daniel Carosone wrote:
> > Hm. With others posting me-too's, I thought a counter-point would be
> > in order.
> 
> You probabily failed to provide one. :-)
> 
> > I've updated several machines since that date, mostly on the weekend;
> > a Xen dom0 and a couple of domU's, a new machine, and my laptop.  No
> > problems seen.
> 
> I think my system is suffering from two problems:
> 
> 1.) Panics caused by the page finger print code.
> 
>     This has been fixed on the 7th when this code was backed out. So if
>     you updated during the weekend you won't have seen this problem
>     because none of your systems use that code.
> 
> 2.) Panics caused by tmpfs.
> 
>     The panic which I've posted this morning might be related to tmpfs.
>     Thomas Klausner who is not seeing panics any more isn't using it.
>     Peter Postma's and my system both use it and crash frequently.

tmpfs uses 2K of kernel memory per file.  With lots of files in a 
tmpfs, it's very easy to exhaust kernel memory... (Bumping NKMEMPAGES 
does help, but it doesn't solve the problem)

Memory statistics by type                           Type  Kern
         Type  InUse MemUse HighUse  Limit Requests Limit Limit Size(s)
      UVM aobj120000240000K 240000K240000K   182391   11     0  2048

(tmpfs allocates one UVM aobj per file, assuming VMSWAP is enabled 
(which it is by default))

Later...

Greg Oster