> To me, the proposed change seems to be heading in the wrong direction.
> We should aim to support more sharing of valid, non-mbuf kernel data
> (e.g., data loaned from struct bufs) with subsystems which expect
> mbufs chains; not less sharing.
> 
> From that perspective, the test for ``can I modify the kernel storage
> backing this mbuf in-place?''needs to be stricter (less inclusive),
> not more inclusive, since such `sharing' may involve non-mbuf
> references to the same underlying storage.

my point is, there's no reason to make all external storage read-only.
what you need is M_EXT_READONLY i mentioned in another mail (or M_EXT_RW).
i didn't include it in the patch just because currently there's no code
which need it in tree (afaik).

YAMAMOTO Takashi