>> Now, this is my 1.4T kernel
> Some mbuf-leaks were fixed in the NFS code since then.

I'm inclined to doubt that's relevant, because the SPARC was not
affected and i386 with ex0 was not affected, while i386 with vr0 was.
Unless there's something different in the way vr0 and ex0 receive
packets on i386, and le0 on sparc, that tickles the NFS-code bug, I
suppose; that seems unlikely but not impossible to me.

If I ever got the point of putting something newer on my NFS server, I
may be able to test whether I still have problems with vr0 then.  But
for the moment, I'm using post-1.4T only on machines that 1.4T doesn't
support and one test-case i386 machine.

> See kern/13807, kern/14209, kern/18924.

13807 is not relevant; IPsec is not in use, and the problem is on the
server, not the client.

18924 is also not relevant (IPsec again).

14209 very well might be relevant; I've never tried loopback NFS, so I
don't know whether my problem exists in that case.

Thanks for the pointers!

/~\ The ASCII				der Mouse
\ / Ribbon Campaign
 X  Against HTML	       mouse@rodents.montreal.qc.ca
/ \ Email!	     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B