On Fri, Jan 24, 2014 at 1:55 AM, Ben Pfaff <b...@nicira.com> wrote:
> On Thu, Jan 23, 2014 at 04:45:34PM -0800, Luigi Rizzo wrote:
> > On Thu, Jan 23, 2014 at 4:43 PM, Ben Pfaff <b...@nicira.com> wrote:
> >
> > > On Fri, Jan 24, 2014 at 01:29:32AM +0100, Daniele Di Proietto wrote:
> > > > I'll just attach the patch (just for a quick preview, it is not ready
> > > jet)
> > > > that builds the userspace components for the kernel datapath in
> FreeBSD.
> > > We
> > > > provide some linux functionalities (netlink and epoll) through
> wrappers
> > > > (not included here) so that LINUX_DATAPATH is true. What we may need
> is a
> > > > better way to understand which features are present to exclude some
> > > blocks
> > > > inside LINUX_DATAPATH.
> > >
> > > From reading this, it looks like the BSD kernel datapath port also
> > > uses Netlink as the communication channel to the kernel. Is that
> > > correct?
> > >
> >
> > correct. daniele implemented netlink generic sockets
> > for FreeBSD as a kernel module as part of the work.
>
> Hi Ben,
we have been revising this change today:
> OK, that explains why there are interesting boundaries being crossed
> here.
>
> How about this:
>
> - Replace LINUX_DATAPATH by NETLINK_DATAPATH, since that seems
> to describe the relationship better.
>
yes we believe we should go for it.
>
> - Instead of (LINUX_DATAPATH && !HAVE_IF_DL), write just
> __linux__. In the cases I looked at, the interface didn't
> seem too likely to be of interest outside Linux.
>
For this one there are only two things to keep in mind:
1. lib/automake.mk has no access to __linux__
so we should define some other variable in the autoconf files.
Any suggestion ?
2. our lib/automake.mk which compiles NETLINK_DATAPATH
on FreeBSD has a block as follows (indentation for readability):
if NETLINK_DATAPATH
lib_libopenvswitch_la_SOURCES += \
lib/dpif-linux.c \
... # other common files
if HAVE_IF_DL
else
lib_libopenvswitch_la_SOURCES += \
lib/netdev-linux.c \
... # other linux-specific files
endif
endif
our lib/netdev.c has a similar block in netdev_initialize()
So moving these inner blocks outside would probably make
the structure less readable.
None of these is unsurmountable so we can do either way.
Let us know which way you prefer.
cheers
luigi/daniele
> - Where individual features seem to be sensibly testable, do
> that.
>
> - If the FreeBSD kernel datapath port makes it to upstream
> FreeBSD, or otherwise achieves some kind of widespread
> adoption, then we might eventually want to rename dpif-linux
> to dpif-netlink.
>
--
-----------------------------------------+-------------------------------
Prof. Luigi RIZZO, ri...@iet.unipi.it . Dip. di Ing. dell'Informazione
http://www.iet.unipi.it/~luigi/ . Universita` di Pisa
TEL +39-050-2211611 . via Diotisalvi 2
Mobile +39-338-6809875 . 56122 PISA (Italy)
-----------------------------------------+-------------------------------
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
