Difficulty with netbooting - le0: lost carrier when bringing down the netbsd kernel

Hi folks!

For literally the first time in at least a couple decades, I'm trying to power up and play with my 9000/362 (I think that would be the designation based on what I have). Because I don't have an existing system to bootstrap an OS onto a disk with, I'm trying the netboot route. Plus, it seemed like an interesting exercise I haven't done since the early 2000s.

I tried following the instructions here: https://www.netbsd.org/docs/network/netboot/intro.hp300.html - specifically the steps listed at the bottom (rbootd, rarpd, etc). I successfully followed those instructions many years ago but I think that SYS_UBOOT has changed enough that they are well out of date.

TL;DR: I'm getting an le0: lot carrier when it's trying to bring down the kernel.

---

A bit of hardware background. My machine has the thinnet NIC on the System Interface board, and an AUI interface in the expansion chassis. Because immediate access to AUI networking was a little easier (just grab a tranceiver and I'm ready to go), I decided to use the AUI. I switched the select codes on both ethernet devices so that the AUI had the lowest number (not sure if this is necessary). Time to start the netboot setup.

rbootd installation went smoothly enough. After the hp was turned on, a little while later both the rbootd debug output & the behavior on the hp console indicated success.

rarpd was where the success stopped. I never got a rarp request. At this point I turned to my favorite overeager coding buddy, chatGPT for assistance, and surprisingly, it had some good info.

Main useful observation: the console message was saying that the SYS_UBOOT bootstrapper was using bootp, and that rarpd was no longer in use. Interesting (if true). Time to switch to a bootp process. Rather than suggesting traditional bootp, my little coding buddy suggested I try something called dnsmasq as a more modern alternative. Sure, let's try that out.

I download and configure dnsmasq. This took a bit of poking, but I eventually got it set up in a way that seemed sane. On the 300, the bootstrap would start up, and then start asking for a variety of kernels (netbsd, netbsd.gz, etc) but never actually got any data. It just loops over and over through the series of kernel names. Using tcpdump on the tftp port, I see no requests from the hp. Coding buddy suggests that the SYS_UBOOT might insist on using the thinnet NIC, regardless of Select Code or how it brought down the bootstrap. At that point I break for the night.

This morning, I dug up some thinnet hardware & cabling, and switched everything over to use that NIC. Now, I actually see requests for a kernel showing up in the dnsmasq logs:

dnsmasq-dhcp: 326617420 bootfile name: netbsd

dnsmasq-dhcp: 326617420 next server: 192.168.2.227

dnsmasq-dhcp: 326617420 sent size: 1 option: 53 message-type 5

dnsmasq-dhcp: 326617420 sent size: 4 option: 54 server-identifier 192.168.2.227

dnsmasq-dhcp: 326617420 sent size: 4 option: 1 netmask 255.255.0.0

dnsmasq-dhcp: 326617420 sent size: 4 option: 28 broadcast 192.168.255.255

dnsmasq-dhcp: 326617420 sent size: 4 option: 3 router 192.168.1.1

and on the HP console I just see a sequence of errors:

le0: lost carrier

bootp: no reply

boot: client IP address: 0.0.0.0

root: addr=0.0.0.0 path=

le0: lost carrier

arpwhohas: no response for 0.0.0.0

And that’s where I’m stuck. I suspect those messages indicate physical level errors; e.g. the onboard bnc wiring, 10baseT hub I’m using that has a 10base2 connector on it is faulty, or more likely, the rg58 cabling/tees/terminators have decayed or oxidized over the decades and are no longer reliable. I’m perhaps a little surprised that it’s so reliably unreliable, but maybe that’s the difference between bootp and a protocol that requires a sustained carrier running on marginal cabling.

Any insight or ideas would be appreciated! I feel like I’m really close and I’d like to push past this. In particular:

- Ideas on what I might be doing wrong? I know I’m straying from the standard path, but out of what seems like necessity. I could try to go back to traditional bootp, but that doesn’t seem like it would make a difference. It’s the tftp part that’s failing.

- Ideas on how to do this all over the AUI, which seems like it might be a more reliable connection.

- Have you used the hp300 netboot instructions lately as published on the NetBSD site, and been successful, or found another path?

Thanks for any assistance – it would be nice to get this machine up and running.

-mike