I would like to add to Peter's answer to Nate,

I had non of the problems mentioned netbooting from a  Debian 1.3.x
Gnu/Linux systems running the 2.0.29 linux kernel.

Thanks,
Lazaro






Nathan Gelbard wrote:
>We didnt have enough disk space on the net-boot server to place copies
>of the root-fs for each net-boot system, so we added a Quantum 425s
>to our Linux box, and specified in bootparams that root be mounted
>off of there. This wouldnt be a problem, except that the Linux
>nfs code is somehow broken, and we cant seem to mount ANYTHING from
>Linux -> NetBSD (or even Linux -> Linux) read/write. We're running
>Slackware 3.4
Linux->Linux shouldn't be a problem. For Linux->NetBSD you have to
tell the NetBSD system to use 1K packets [because it can't cope with
the larger (fragmented) packets everybody else uses, unless somebody's
fixed the NFS code since I encountered the problem...] From my
netbooting HPs from Linux sketchy HOWTO:
7) If your HP sticks halfway through boot trying to write to the NFS
system:
this is a symptom of using Linux as an NFS server. Linux NFS can't cope
with 8K write requests, which are fragmented across several packets. The
solution is to remount the root filesystem to use 1K blocks, by putting
the following line near the top of the HP's /etc/rc :
# Very first thing to do is remount root with sensible values
# for using Linux as an NFS server:
mount -t nfs -a -o -I=1024,-r=1024,-w=1024 dram:/export/hp/root/ /
Note that this isn't a real solution, since swap is still using 4K writes.
(ie as soon as it starts to swap, it'll fall over!)
Suggested fix is to build a kernel with:
options NFS_BOOT_RWSIZE=1024

Hope this helps...
Peter Maydell