>It seems that there is a problem with mapped nfs files, since some nfs buffers
>in the buffer cache are only partially valid (the b_validoff and b_validend).
>If a file that has been recently written is read back, you can get garbage
>out of the non-valid regions of the buffers. This only shows up for NQNFS and
>NFS v3, since all buffers are invalidated upon close for NFS v2, after a file
>is written.

   There is a mechanism in our merged cache code for dealing with partially
valid pages that should be usable for this purpose. I'll defer to John Dyson
as he is the author of the merged cache.

>ps: If there is a particular person that I should be emailing w.r.t. nfs/vfs
>    stuff in each 4.4BSD-Lite descendant group, just let me know and I will
>    use that instead of the bugs mailing list.

   In the FreeBSD group that would be John Dyson.

-DG