Subject: Re: Update time!
To: Andy R <quadreverb@yahoo.com>
From: Josh Tolbert <hemi@scoundrelz.net>
List: port-dreamcast
Date: 01/06/2003 21:42:08
Hi Andy,

On Mon, 6 Jan 2003, Andy R wrote:

> --- Josh Tolbert <hemi@scoundrelz.net> wrote:
> > Hey folks,
> > 	It's been quite a while since I have played with my
> > Dreamcast, so
> > I dusted it off and installed the old stuff I had
> > built. Seeing how the
> > old system worked, I realized it would probably be a
> > wise idea to
> > incorporate some of the suggestions others had made
> > to the swap method,
> > config setup and the like, so I decided to
> > re-cross-compile the whole
> > shebang from scratch. Since a lot changed
> > (especially in the way of kernel
> > configuration), I decided to write new docs on the
> > whole deal, too...So,
> > without further ado, I present
> >
> > http://www.puresimplicity.net/~hemi/dreamcast/
> >
> > 	I decided to not even bother releasing files to
> > support burning a
> > kernel on CD, since Marcus' IP Upload Slave method
> > works so well. No CD-Rs
> > were harmed during this experiment; I used my old IP
> > Upload Slave CD-R.
> >
> > 	Other changes include using a file (/.swap) for
> > swap instead of a
> > file mounted from another NFS share, a truly
> > cross-compiled world instead
> > of a hacked half-broken world and files pirated from
> > another machine and
> > better documentation.
> >
> > 	In the next few days I'm going to stick up some
> > differently-configured kernels (and their config
> > files) so others can have
> > a place to start from if they decide to build their
> > own kernels.
> >
> > 	puresimplicity.net is a box completely under my
> > control; I moved
> > the stuff there so I could keep track of how much
> > bandwidth the machine is
> > using and the like.
> >
> > 	If you have any suggestions or comments, please let
> > me know.
>
> Josh, great work!
>
> Ok, I'll let you know...
>
> Way back when (about 5 months ago in port-dreamcast
> terms), you had written this:
>
> >>>
> I worked around the problem by creating a 64M swap
> file.
> >>>
>
> This is from this post:
> http://mail-index.netbsd.org/port-dreamcast/2002/08/31/0011.html
>
> Your site still shows you preparing a 32 meg swap
> file. Was there any value added to using a 64 meg
> swapfile? I just made one after you had said that and
> didn't ask any questions. I know it's not the best
> thing to use more swap than you're supposed to (cuts
> into real memory), but I don't like crashes either.
>

	Hrm, good question. I haven't thought about the problem since I
stopped running the old build a few months ago. :) Since I've had the
machine running the new world stuff I haven't had a single OoM problem,
though. I think the errors I was seeing were a combination of running the
BBA at 100BaseTX, trying to compile irssi (which reboots the machines
during the POSIX threads build) and just not paying enough attention.
Since I'm now running the BBA at 10BaseT by default in the new build I
haven't had any more problems, but the POSIX threads causing a reboot
still exists. It's definitely something to note in the documentation,
though. I'll update the docs to suggest a 64M swap file if the machines
runs out of memory and swap with a 32M file.

> A quickie flowchart of the "booting events" that take
> place may help one grasp the process better. I just
> started to learn netbooting when you put this stuff
> up, and I'm still not very comfortable with it.
>

	I'm still not very comfortable with it, either. I have a fair idea
of what's going on, but generally I just keep plugging away at it till it
works. The guide would have been out a day earlier if I hadn't got
frustrated with the NFS export due to me forgetting to turn on rpcbind in
/etc/rc.conf on the NetBSD PC and went to bed. :)

	As far as a timeline, how much detail are you looking for? I could
see starting with the Dreamcast powered on and waiting at the IP Upload
Slave, but not really before that since Marcus' site covers how the IP
Upload Slave works in detail. Would something graphical be worth it, ir
would a text explanation be adequate?

> Andy
>

Thanks,
Josh