Excerpts from internet.computing.netbsd.port-i386: 30-Sep-96 Re:
Problems with 1.2/i386  by Jason Thorpe@nas.nasa.go 
>  William O Ferry <WOFerry+@CMU.EDU> wrote:
>  
[LOTS O' savecore foo deleted]
>  
> YOW!
>  
> That's ... odd... when did this creep in?  Did it correspond to any
> other changes in your system?  I certainly never saw this on my
> sparcs or hp300s... and my i486 box is solid as a rock.
>  
> You say below you recently added memory ... did these appear when you
> added the memory?

    The crashes would seem to have appeared around the time I did 3
things at once:
    Go from 16 to 32MB (2 to 4 8MB SIMMs)
    Update to XFree86 3.1.2F (and start running xdm)
    sup and `make build` to NetBSD 1.2_BETA

So far, I've tried running without X, but it still likes to panic if I
log out on console, so I doubt it's X...  =)  Last week, I added another
64MB, which required that I remove 2 of the 8MB SIMMs.  I removed the
two that I had added around when the whole mess began.  The machine
still crashed.  Thinking that it might be bad SIMM sockets, I removed
the original 16 and moved the new 64 into the upper two SIMM slots,
leaving the lower two empty (NetBSD was only seeing 64 anyway, though
the BIOS reported all 80).  And even with the original sockets in use
and new SIMMs, it took it less than a day to panic, in the middle of an
X session (nothing related to logouts at all).  My next deal is to go
back to the original 16MB configuration (ouch!).  If all else fails,
I'll try dropping back to 1.1 or such.  Unfortunately I can't drop back
to an earlier XFree86, thanks to their wonderful beta expiration deal
(namely the fact that they don't have any overlap between when one beta
expires and they release the next one).  But I don't think it's X, it
just panics more often when X is running (so I guess it is memory / disk
/ load related).  The search continues...

> usage is:
>  
> options         EXTMEM_SIZE=<size of extended memory in Kbytes>
>  
> If you compile a kernel with this option, then remove memory, you'll
> have to recompile the kernel to reflect that.

    Thanks, but I think I'll pass that up and just sell the extra
memory, I don't want to have to switch kernels over it...  =)

>  >     Also, is the cache size determined solely based on the amount of
>  > memory?  I've heard that systems like Linux basically use all available
>  > memory as cache, and was wondering how NetBSD handles it.
>  
> Define "cache".  Buffer cache?  NetBSD/i386 uses 10% of the first N
> meg (don't know that number off-hand) and 5% of the rest, _I think_.
> NetBSD/hp300 just uses 10% of RAM (because, as the comment says, HP's
> tend to run long on memory and short on disk speed :-).

    Yeah, I guess disk cache.  A friend of mine was bragging about Linux
and how it uses all (or at least) most available space for cache.  He
said that it also accounts for a large part of the performance
difference between OS/2 and WinNT.  I was just curious as to how NetBSD
does it.

    Thanks for the help.  I'd appreciate any further thoughts.  I'll get
back on any findings I have as I keep changing my setup.

                                                          Will Ferry

--------------------------------------------------------------------------
William O Ferry <woferry@CMU.EDU> | finger:  woferry@WarpDrive.RES.CMU.EDU
talk:  finger for online status   | http://warpdrive.res.cmu.edu/~woferry/
--------------------------------------------------------------------------