current state:
shithub.us/rminnich/nix main branch
I made one change to nix/386/etherigbe.c, if CLS is 0 or 0xff, I just set
it to 64 and continue. Works fine.
I run NIX, and once the VM starts up:
bind -a '#l0' /net
bind -a '#I' /net
ipconfig
listen1 -tv1 tcp!*!23 /boot/rc -m /boot/rcmain -i

then on another terminal:
nc 127.0.0.1 2323
ps aux
Await             1    0:00   5:39       13K 20       rc
Wakeme            2    0:00   5:39       10K 53370    genrandom
Wakeme            3    0:00   5:39       13K 0        conskbdq
Wakeme            4    0:00   5:39       13K 0        conskbdq
Wakeme            5    0:00   5:39       13K 0        alarm
Wakeme           10    0:00   5:32       13K 0        #l0lproc
Wakeme           11    0:00   5:32       13K 10       #l0rproc
Wakeme           13    0:00   5:32       13K 0        rxmitproc
Wakeme           15    0:00   5:32       13K 10       etherread4
Wakeme           16    0:00   5:32       13K 0        recvarpproc
Wakeme           17    0:00   5:32       13K 0        etherread6
Sleep            18    0:00   5:32       13K 0        ipconfig
Wakeme           20    0:00   5:25       13K 0        #I0tcpack
Await            38    0:00   0:08       13K 0        rc
Pread            40    0:00   0:00       13K 0        ps

so we can get in, next step is drawterm.

Then I can do the execac test.

I still get occasional panics on cclose, maybe the later commits resolve
this.

It's interesting looking around the 15y ago stuff.
/sys/src/nix/root/{suck,blow}.c are tests for the IBM blue gene torus.
You can really see the transition we were making from blue gene back to
amd64 for the NIX work. The blue gene kernel started with the original k8
work DOE sponsored starting in the 2005 timeframe. This is pretty clearly a
kernel caught mid change.

If you drawterm in, you're going to need to go with the old protocol on
17010, this code base is that old.

On Mon, Sep 7, 2026 at 8:51 PM ron minnich <[email protected]> wrote:

> I forget to mention:
> sys/src/nix/k8/9k8cpufs[.elf] are working kernels, you can try them before
> you try a build.
> the namespace the script sets up is mandatory.
> you need to be on the branch named main
>
>
> On Mon, Sep 7, 2026 at 8:47 PM ron minnich <[email protected]> wrote:
>
>> Kim took a look, created 5 easy patches, and we're back on the air.
>>
>> shithub.us:rminnich/nix
>>
>> see patches/RON_NOTES, which is an rc script.
>>
>> to run with qemu:
>> qemu-system-x86_64 -machine pc,acpi=off,hpet=off -cpu Opteron_G3 -smp
>> 4,sockets=4,cores=1,threads=1 -m 1024M -kernel 9k8cpufs.elf  [-nographic]
>>
>> Please, someone, tell me the options to get a network up.
>>
>> note what we'd set up here: the cpufs image includes all the bits and
>> pieces you need
>> to start a cpu server (see /boot). So it's basically a small standalone
>> image that functions as a cpu server, nothing more .
>>
>> To run on an ac, use execac.
>>
>> There's some stuff we did that I think is missing. We had, in my memory,
>> modified rc so that it would let you easily start commands on an ac, with
>> some syntax like [2]date or something. I can't find that.
>>
>> Questions to me, Kim has done their best but has little time to continue.
>>
>> Also, this is a change against the very first commit. All the subsequent
>> changes need to be added.
>>
>> ron
>>

------------------------------------------
9fans: 9fans
Permalink: 
https://9fans.topicbox.com/groups/9fans/Ta8d90b97cb2cb844-M635975bcb3570e3b87dbf8e9
Delivery options: https://9fans.topicbox.com/groups/9fans/subscription

Reply via email to