> so, there's inconsistency between ffs and nfs.
> ffs truncates the file to original size when error.  but nfs doesn't.
> maybe we should define what VOP_WRITE should do on error.

Hmm, I guess it may return error and truncate it or may return number
of bytes successfully written (but it is not sure how it is useful).

But currently, it may leave data written with returning an error if
part of source region isn't mapped correctly, right?

BTW, is ffs truncation too early?

enami.