Revisiting the 486sx issue

[llondel]

I've been trying to put together a system that will run on a 486sx. I've had a good read through the previous stuff on this topic at http://forum.tinycorelinux.net/index.php?topic=5462.0 and have a system that runs happily on an FPU-equipped system(*), but if I try to boot it with the 'no387' kernel argument it hangs immediately after loading the initrd and saying 'Ready'.

I've definitely recompiled the kernel and all the modules using the patched kernel at http://distro.ibiblio.org/pub/linux/distributions/tinycorelinux/3.x/release/src/kernel/linux-2.6.33.3-patched.tbz2 and started with the config in http://distro.ibiblio.org/pub/linux/distributions/tinycorelinux/3.x/release/src/kernel/config-2.6.33.3-tinycore where I just told it to include the maths emulation.

I extracted the 3.5 ISO and did an rm -rf /lib/modules on it to blow away the previous stuff, and then installed the new modules in place. I put the new bzImage file into /boot.

So what am I missing? (note that I did also try this with 3.4 and had exactly the same problem)

(*) I'm still waiting for the 486sx system to arrive. It's based on a Vortex86SX.

[curaga]

That should be enough, so try to find out where things are breaking. Boot with the "debug" bootcode to see the kernel messages.

[danielibarnes]

Some tips I found:
1) Try adding "nofxsr" boot parameter also
2) Make sure your kernel does not have SMP or CONFIG_X86_GENERIC enabled

[llondel]

I managed to get it to boot with no387 this evening. 

I turned all the knobs and used all the tips.

Is there a quick way of compiling only those modules found in the release base image, and to compress them as in the base image? I guess it might all be done with a post-compile script to strip out the extras into extensions rather than just compile the base set in the first place.

Anyway, thanks for the tips so far.

[danielibarnes]

I turned all the knobs and used all the tips

Can you please elaborate? I'd like to know exactly what made it work. I am interested in the
MicroClient JrSX, but I'd like to know the precise steps to get it to work.

Here is how I compile the modules. It can give you a start:

tar xjf linux-2.6.33.3-patched.tbz2
cd linux-2.6.33.3
make mrproper
# Copy custom config here
cp config-2.6.33.3-tinycore .config
make oldconfig < /dev/null
for m in $(cd /lib/modules/$(uname -r)/kernel; find . -name '*ko.gz'|sed 's:^./::;s:.gz$::;/^misc/d'; cd - >/dev/null); do
MSRC=$m
# Set your destination path here
MDST=/usr/local/lib/modules/$(uname -r)/kernel/$m.gz
make $MSRC || exit 1
mkdir -p $(dirname $MDST)
gzip -2 < $MSRC > $MDST || exit 1
advdef -qz4 $MDST
done

There are 288 modules, so it will take a while. I don't know if it's any faster than building everything and copying over only what you need. I omitted misc above because the ramzswap.ko is actually in drivers/staging/ramzswap, but I suspect it is built from newer source. Not sure what to do there. Maybe someone can help.

[llondel]

Can you please elaborate? I'd like to know exactly what made it work. I am interested in the
MicroClient JrSX, but I'd like to know the precise steps to get it to work.

That's my final target as well. I lost track of the .config and I think I overwrote the winning answer late last night  so I'll have another go and be more careful and write it up. I left it doing a clean compile with what I think was the right answer and went to bed.

One thing I did notice is that my test system, an Aspire One, didn't like having the agp stuff present and that was frustrating me a bit because even when telling it to use text mode, it would still hang. Deleting those drivers made it behave.

[llondel]

diff config-2.6.33.3-tinycore 486sx-config

Code: [Select]

Deleted as it didn't work - see later post

Note that this hasn't yet been tested on the target machine, but it boots on the Aspire One with text, nofxsr and no387 specified. My MicroClient JrSX will hopefully arrive tomorrow.

[curaga]

Unified diffs in the future please (diff -u), they're more readable


Butting in on the MicroClient, it's my impression that one can get better hw for less by grabbing a thin client off ebay. They're similar size, too.

[danielibarnes]

it's my impression that one can get better hw for less by grabbing a thin client off ebay.

Yes, that's true. Shipping for the MicroClient JrSX to the US is more than 60% of its cost. That makes ebay a significantly less expensive option.

This Thin Clients site is full of useful information, including references to Tiny Core.

[gerald_clark]

Also, the eBox-2300 is about the same price, and is a full 486 version of the Vortex86.
It runs TC V2 and V3.

[llondel]

Unified diffs in the future please (diff -u), they're more readable 

I can post one of them as well if you want 

As for cost, this is ultimately for something being sold as part of a bigger system, so the hardware needs to be FCC/CE approved and available in quantity. At that point the shipping cost becomes less of an issue.

The eBox looks interesting as an alternative, although now I appear to have a bootable SX kernel, I'll probably stick with the original unit. It does help to have a plan B in case purchasing have problems though.

[danielibarnes]

Also, the eBox-2300 is about the same price, and is a full 486 version of the Vortex86.
It runs TC V2 and V3.

I did some research. The eBox is the same as the NorhTec MicroClient. They both get their systems from DMP Electronics. WDL Systems (North Carolina) is a distributor of the eBox. It looks perfect for Tiny Core.

[gerald_clark]

I got my eBox-2300 from WDL.
Everything in the unit works fine under TC and MC.
The eBox-3300 looks good, but it uses a non standard CF controller, so you cannot access the Compact Flash.

[llondel]

I now have a target system. Unfortunately it doesn't boot the code that was built from the above config. It loads the kernel and initrd OK, unpacks the kernel and then hangs.

I need a smiley with a thinking cap.

[Rich]

                       / \
Here you go.