Re: Net-installing SS20 & NFS woes (?)

[Michael <macallan%netbsd.org@localhost>]

Hello,

On Sun, 9 Aug 2020 12:14:20 +0200
Romain Dolbeau <romain%dolbeau.org@localhost> wrote:

> I'm trying to netinstall a new SS20 I just got. But I can't get the
> bootloader to download the kernel.

Hmm, I netboot all sorts of machines, including two different SS20.
From a NetBSD host though.

> I removed a problem DIMM (stuck bits), ran the diagnostics, the
> hardware seems OK. However, 'boot net' fails complaining about the
> carrier. Doing 'boot net-tpe' works (... to a point).
> See a screenshot at
> <http://www.dolbeau.name/dolbeau/files/IMG_20200809_114117608_50p.jpg>.
> 
> The host serving all the files & services is a Linux Debian 9
> (oldstable) on x86-64. I don't have NetBSD-running hardware on that
> network (vacations :-) ).
> /etc/exports is "/ss20pinstall
> 192.168.0.62/24(rw,no_root_squash,no_subtree_check,insecure)" (it was
> originally /32 but I wanted to check with another machine).
> 
> I've set up dhcp/bootparamd/atftp/rarpd to an extent that things seem
> to work, see the screenshot above: the PROM download the boot.net
> file, and from daemon.log:
> 
> atftpd[21297]: Serving C0A8003E.SUN4M to 192.168.0.62:47521
> (...)
> rarpd[21243]: RARP request from 08:00:20:76:3e:08 on eth0
> rarpd[21243]: RARP response to 08:00:20:76:3e:08 192.168.0.62 on eth0
> rpc.bootparamd: whoami got question for 192.168.0.62
> rpc.bootparamd: This is host ss20p
> rpc.bootparamd: Returning ss20p.dolbeau.name   (none)    192.168.0.200
> rpc.bootparamd: getfile got question for "ss20p.dolbeau.name" and file "root"
> rpc.bootparamd: returning server:192.168.0.200 path:/ss20pinstall
> address: 192.168.0.200
> rpc.mountd[21828]: authenticated mount request from 192.168.0.62:1021
> for /ss20pinstall (/ss20pinstall)
> 
> The /ss20pinstall is properly exported and can be mounted from e.g. a
> OSX host. I've also tried limiting the host to NFSv2 in
> /etc/defaults/nfs-kernel-server:
> 
> RPCMOUNTDOPTS="--manage-gids"
> #RPCMOUNTDOPTS="--nfs-version 2 --no-nfs-version 3 --no-nfs-version 4"
> 
> But I get the same error either way: "rpc_call: error = 2" then
> "Cannot load netbsd: error=72", as can be seen in the screenshot.

Hmm, 2 is ENOENT, 72 is EBADRPC. So I guess the bootloader does have
trouble talking to your nfs server.
 
> I also tried both a compressed and uncompressed file 'netbsd' (which
> is netbsd-GENERIC) in was it was a size issue.
> 
> After the bootloader asks about the device, if I hit return twice
> (once for device, once for 'boot' - presumably the name of the file),
> I get only "Cannot load netbsd: error=60".
> Trying again, the error is now 23. All subsequent attempts are error
> 23 as well. If I exit to the PROM and do boot net-tpe again, the prom
> reset the system and I'm back to error 72.
> 
> I'm guessing there's something wrong with my NFS setup or the network
> hardware, but I can't figure it out.
> 
> Any suggestion welcome.

Just a few thoughts, I didn't touch Linux in more than a decade.
- our bootloader speaks DHCP, no need to mess with bootparamd.
  Shouldn't hurt though, so no need to change it, but if you want to
  netboot other machines you'll need DHCP at some point.
- I'd use a packet sniffer to see what exactly the client tries to do
  that fails. As in, is it the mount request? Or loading the kernel? Or
  something entirely different I'm missing?
- you may need something like -maproot=0:0 in /etc/exports for the
  client to be able to access files owned by root. No idea what UID the
  bootloader uses, it might be root...

have fun
Michael