QEMU emulation

[Hans Bezemer]

On Saturday 15 September 2012, you wrote:
Hi Delbert!

> Looks like you have something quite useful already.  I looked
> at using QEMU shortly after my NN arrived about two years
> ago.  Looked too complicated for potential return.  For
> someone who claims to not "know" much about the process,
> you have done very well.
I have been playing to get the Raspberry Pi going and wonderdd whether I could 
apply the same thing to the NanoNote. I must admit I never thought it would 
work and the first steps indicated that much. Think about it: taking a kernel 
from a random Debian MIPS Malta distribution and use it to boot a rootfs that 
was never made to be bootstrapped this way.

But after I got past "init" (note I used "init=/bin/sh" the first few times 
just to get a prompt) I started to solve problem by problem - simply by 
googling it and applying the solution. Then I started to reinstate things, 
first "init", then "rcS", etc. IOW, one simple step at the time.

Note I had to fsck.ext2 the thing after each crash - which I did with a 
fast-booting Debian ARM distribution I'd used to develop the RasPi emulation. 
For comparison I kept an untouched rootfs, cleanly untarred. After a certain 
moment I started to add functionality, mostly networking stuff. As others 
have said: I could most likely do that from the prompt of my host, but it's 
awkward and error prone. Better safe than sorry.

After having SSH things became much easier, because I could bring the thing 
down in a controlled way when it crashed - very handy while testing the right 
setting for MPLayer ;-)

There are only a handful of things that don't work: two games, the "listener" 
stuff - most are related to missing sound, the rest with the non-working file 
selection program. You could fix that with a ported "dialog" and a bit of 
shell scripting.

Note that two years ago this would probably not have been possible. QEMU has 
developed quite a bit lately, even with the recent v1.2.0. E.g. I never got 
Coherent to work - now it does.

> Right now I'm immersed in several other software development
> tasks but look forward to a summary of the steps you took
> to get where you are now.
My last steps have been to erradicate the "setfont2" entries in 
the /usr/share/gmenu2x entries. I think most of the rest is pretty much 
condensed in this file:

NOTE THAT THE FILE BELOW HAS NOT BEEN PROPERLY TESTED YET!
*** USE AT YOUR OWN RISK ***

---8<---
#!/bin/sh
set -x

PROGNAME=`basename $0`
ROOTDIR=$1

if [ -z $ROOTDIR ]; then
   echo "Cannot apply changes to your root directory!"
   exit 1
fi

# Make devices.

mkdir /dev/pts
mknod -m 0666 "$ROOTDIR/ptmx"         c 5 2
mknod -m 0600 "$ROOTDIR/dev/console"  c 5 1
mknod -m 0660 "$ROOTDIR/dev/full"     c 1 7
mknod -m 0640 "$ROOTDIR/dev/kmem"     c 1 2
mknod -m 0660 "$ROOTDIR/dev/loop0"    b 7 0
mknod -m 0640 "$ROOTDIR/dev/mem"      c 1 1
mknod -m 0666 "$ROOTDIR/dev/null"     c 1 3
mknod -m 0640 "$ROOTDIR/dev/port"     c 1 4
mknod -m 0666 "$ROOTDIR/dev/random"   c 1 8
mknod -m 0660 "$ROOTDIR/dev/tty"      c 5 0
mknod -m 0666 "$ROOTDIR/dev/urandom"  c 1 9
mknod -m 0666 "$ROOTDIR/dev/zero"     c 1 5
mknod -m 0666 "$ROOTDIR/dev/fb0"      c 29 0
mknod -m 0666 "$ROOTDIR/dev/sda"      b 8 0
mknod -m 0666 "$ROOTDIR/dev/sda1"     b 8 1

ln -s /proc/kcore "$ROOTDIR/dev/core"
ln -s /dev/ram1 "$ROOTDIR/dev/ram"
ln -s /dev/sda1 "$ROOTDIR/dev/root"

for N in `seq 0 5` ; do
    mknod -m 0660 "$ROOTDIR/dev/tty$N" c 4 $N
done

for N in `seq 0 15` ; do
    mknod -m 0660 "$ROOTDIR/dev/ram$N" b 1 $N
done

mkdir "$ROOTDIR/sys/kernel"
mkdir "$ROOTDIR/sys/kernel/debug"

cat << 'EOF' > "$ROOTDIR/etc/inittab"
::sysinit:/etc/init.d/rcS S boot
::shutdown:/etc/init.d/rcS K stop
::askfirst:/bin/sh /usr/bin/gmenu2x
EOF

cat << 'EOF' > "$ROOTDIR/etc/init.d/qemu"
#!/bin/sh
mount -t proc none /proc
mount -t devpts devpts /dev/pts

if [ -f /var/run/dbus.pid ]; then
   rm /var/run/dbus.pid
fi

setfont /usr/share/kbd/consolefonts/kernel-6x11-font
EOF

cd "$ROOTDIR/etc/rc.d"
ln -s ../init.d/qemu S09qemu
rm K01ben-nanonote
rm S01ben-nanonote
rm S93triggerhappy
cd -

cat << 'EOF' > "$ROOTDIR/etc/config/network"
# Copyright (C) 2006 OpenWrt.org

config interface loopback
        option ifname   lo
        option proto    static
        option ipaddr   127.0.0.1
        option netmask  255.0.0.0

config interface lan
        option ifname   eth0
        option proto    static
        option ipaddr   192.168.254.101
        option netmask  255.255.255.0
        option gateway  192.168.254.100
        option dns      "8.8.8.8 8.8.4.4"
        option hostname BenNanoNote
EOF

cat << 'EOF' > "$ROOTDIR/etc/uci-defaults/01-optimization-for-ben-nanonote"
#!/bin/sh

# no need in NanoNote
/etc/init.d/atd disable
/etc/init.d/led disable
/etc/init.d/telnet disable

# generate fontconfig cache
/usr/bin/fc-cache

exit 0
EOF

cat << 'EOF' > "$ROOTDIR/etc/uci-defaults/99-set-time"
#!/bin/sh 
date --set 201207072133
EOF

cat << EOF > "$ROODIR/usr/bin/backgammon.sh"
#!/bin/sh
setfont /usr/share/kbd/consolefonts/kernel-6x11-font
exec /usr/bin/backgammon
EOF
---8<---

> An acid test of the system would be to compile the OpenWRT
> kernel under the system and get it running:)  May not return
> the effort but would be an interesting challenge.
A better kernel, geared to the NN would be best, I think. E.g. with an 
embedded es1370, ac97 (which is closest to the NN) or Intel sound driver 
(hda). I'm sure someone here can make one. I'm not too much into kernel 
compilation. Never did it, don't know what QEMU requires in that respect.

> Thanks for your work and enthusiasm--I can sense it in
> your E-mails!
Well, it was fun to do. And with my NN misbehaving I needed a replacement to 
test my 4tH compiler. I've compiled it in the blind for the last few releases 
and I don't like to release stuff I can't test. For the record: I've tested 
the latest 4tH BNN release and it worked out of the box, so I've worried for 
nothing.

Personally, I feel that every H/W manufacturer should release and maintain 
something like this. Yes, I know: "It's only $25, $35, $99" or whatever your 
price is, but developers (especially FOSS) won't buy a device for every darn 
port they want to support.

.. and of course, I have a weak spot for my little thingy. It felt good to see 
the "root at BenNanoNote" prompt again ;-)

Hans Bezemer
-- 
Absolutely no trees were killed to produce this sig. Well, OK, we had to tie 
one up and torture it. That's it.

Visit our website! http://thebeez.home.xs4all.nl/4tH/

*** Home of the 4tH compiler! ***