> tessier pulls down SYS_UBOOT no prob, grabs its bootparams, and
> starts to load the kernel (stock). Watching a tcpdump, the
> kernel transfer stop at about the 108th 'block.'
> 
> 19:35:21.585206 tessier.108 > 3jane.nfs: 104 read [|nfs]
> 19:35:21.586522 3jane.nfs > tessier.108: reply ok 1124 read [|nfs]
> 19:35:29.566069 tessier.108 > 3jane.nfs: 104 read [|nfs]
> 19:35:29.567436 3jane.nfs > tessier.108: reply ok 1124 read [|nfs]
> 
> Then SYS_UBOOT complains
> 
> boot: Input/output error.

One lucky time, it actually loaded the kernel all the way. It proceded
to boot, and everything worked okay. I have been unable to repeat
the boot.

SYS_UBOOT uses NFS/udp to xfer the kernel, right? It seemed after
the kernel booted (watching tcpdump on the 2 hosts), NFS/tcp
worked MUCH faster. I _think_ what I'm seeing is tessier
making 4 read requests on the same block, then halting. If the
nfs server (under almost zero load) responds with a packet before
the 4th packet, everything is cool.

The my netbooting SPARC ELC boots get the kernel a LOT faster 
(hp is 25mhz, ELC is 33mhz). The little spinning bar just zooms.

I'm wondering if my le0 interface is failing. I checked the jumper
on the inside, and have tried 10baseT and coax. The only thing
thats really different about the 380 is last time, it had 32mb,
this time only 16mb of RAM.

Is there a way I can check the le0 interface using rbootd? Possibly
put together some sort of package that it can download and run
a hardware check?

        thanx in advance,

                Nathan