NFS data corruption

[Dmitry Marakasov]

Hi!

I'm using -CURRENT on my new desktop box (due to hardware support issues
with 7.x), and I've experienced reproducible data corruptions with NFS.

Server: 7.0-RELEASE amd64. 
Clients: 7.0-RELEASE i386 [no problems]
         CURRENT i386 [corruption]

The problem showed itself as checksum mismatch when building a first
port on a new box with distfiles/ mounted via NFS (using amd). For
now I have two cases of corruption - in both cases it is single
difference of one 128 byte block with file offsets 0x65F872 and
0x61A072.

Seems like the corruption only appears once in either server lifetime
or mount, as I've checked md5 for all 8GB of distfiles and only had
single corruption, i.e. it shows itself once after the start of the
client (in first 10-20MB's) and then everything is OK (I was able to
build bunch of ports without problems).

Is there any way I can help to diagnose and fix the problem?

Here's amd.map file which is used to mount this filesystem:

---
/defaults	type:=nfs;rhost:=hive;opts:=rw,nosuid,noexec,-3,-i,-s,-T
*		rfs:=/pool/${key}
---

Here's uname -a:
FreeBSD hades.panopticon 8.0-CURRENT FreeBSD 8.0-CURRENT #0: Fri Jan 23 03:39:16 UTC 2009
root at chrysalis.panopticon:/mnt/usr/obj/mnt/usr/src/sys/HADES  i386

-- 
Dmitry Marakasov   .   55B5 0596 FF1E 8D84 5F56  9510 D35A 80DD F9D2 F77D
amdmi3 at amdmi3.ru  ..:  jabber: amdmi3 at jabber.ru    http://www.amdmi3.ru