cvs commit: src/sys/alpha/alpha vm_machdep.csrc/sys/amd64/amd64 vm_machdep.c src/sys/i386/i386 vm_machdep.c src/sys/ia64/ia64 vm_machdep.c src/sys/powerpc/powerpc vm_machdep.c

[Bruce Evans]

On Sat, 27 Dec 2003, Sam Leffler wrote:

> On Saturday 27 December 2003 06:03 pm, Sam Leffler wrote:
> > On Saturday 27 December 2003 05:56 pm, Mike Silbersack wrote:
> > > On Sat, 27 Dec 2003, Sam Leffler wrote:
> > > > >   Move the declaration of sfbufspeak and sfbufsused to mbuf.h,
> > > > >   and use imax instead of max, as sfbufspeak and sfbufsused
> > > > >   are signed.
> > > >
> > > > These have nothing to do with mbufs; why move the decls to mbuf.h?

Oops.  I moved them there first because there was already an sfbufs
declaration there (only 1 before).

> > > Because I can't think of anywhere else sendfile related values would fit.
> > > So far our only use of sfbufs in the system is to attach them to mbufs
> > > and send them out via sockets.  FWIW, I'm about to overload mbstat with
> > > some sendfile statistics as well, because I can't think of another place
> > > that they would fit well.  I'm open to ideas, as long as they don't
> > > involve creating yet another header file.
> >
> > Looks like there's an sf_buf.h.  If that's not to your liking perhaps
> > systm.h or kernel.h.  Please don't pollute mbuf.h.
>
> Looks like this stuff had already landed in subr_mbuf.c so I guess mbuf.h
> makes sense.  Sure wish this stuff had been kept separate; it doesn't make
> sense to me that it's being tossed in to the mbuf code.

sf_buf.h is certainly better for the declarations.

subr_mbuf.c only has sysctls and stubs now.  The implementation is now
mainly in ${MACHINE_ARCH}/${MACHINE_ARCH}/vm_machdep.c, so a new .c
file for sfbufs would be even emptier than it would have been originally.

sfbufs are also mixed with mbufs in netstat (mbuf.c and netstat -m).

Bruce