SUMMARY: NFS across subnets

From: Fred Hoare (fred@dip1.ee.uct.ac.za)
Date: Mon Apr 05 1993 - 11:06:11 CDT


The original problem that I had was the following:
-----
The problem that I am having is that the files mounted across the
subnet are becoming corrupted when seen on the client machine. Most of
the files are executables which die giving error messages such as
invalid command, segmentation fault and bus error. They are not
corrupted for the machines on the local subnet. Copying the files from
the NFS mounted partition to the local partition and then doing a diff
of the files shows that there are differences. Two copies of the same
file also show differences. Using ftp to copy across the files works
and the files are not corrupted.

---
So far I have got three replies and the problem is now solved.
szgyula@cipher.pha.jhu.edu (Gyula Szokoly x4120)replied saying that he
had mounted a filesystem from the US to Hungary without any problems.
ccdw@kudu.ru.ac.za (Dave Wilson) said that they had had the same
problem when using a PC as a bridge, but that it had gone away when
they used a "commercial" bridge. Barry Margolin <barmar@Think.COM>
supplied the answer that solved the problem for me

> Do you have UDP checksums enabled on the client and server? They're turned > off by default. It sounds like your router is corrupting the packets. FTP > transfers are always checksummed (since it uses TCP, in which checksums > aren't optional). > > To enable UDP checksums, edit /sys/netinet/in_proto.c and change the value > for udp_cksum to 1. Then rebuild your kernel. > > You can also change it in the running kernel with: > > # echo "_udp_cksum/W 1" | adb -w /vmunix /dev/kmem

Thanks to these people for the prompt replies, and to anyone else who may respond after this summary goes out.

-- Frederick Hoare email: fred@dip1.ee.uct.ac.za Image Processing Laboratory Department of Electrical Engineering University of Cape Town, South Africa



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:07:41 CDT