Panic in NFS (current/amd64)

[Vincent <10.50%free.fr@localhost>]

Hi folks,

since quite a lot of time, as I mentioned it before, I have problem with NFS
exchanges between my NetBSD 4.0/i386 server and my NetBSD-current/amd64 client
machine. Sometimes the machine hangs forever on the NFS transfer (but I can
still use it), but sometimes the kernel crashes altogether. Up to now, I hadn't
been able to get a debug message, but I finally managed to see what ddb was
saying.
Here it is (I omit the leading f's in the 64-bit addresses):

uvm_fault (0xf...80655240, 0xffff8001439db000, 1) -> e
fatal page fault in supervisor mode
trap type 6 code 0 np 0xf...f80330495 cs8 rflags 10202 cr2 0xffff8001439db000
cpl 4 rsp 0xffff8000463ad1c8
kernel: page fault trap, code=0
Stopped in pid 0.55 (system) @ netbsd:cpu_in_cksum+0xa5

I was unable to backtrace anything else, because ddb itself faulted before
spitting out any other useful info.

That remembers me of an old thread about checksum computation being bogus.

Cheers
Vincent