On Tue, Sep 11, 2012 at 10:24:14PM -0700, Ben Pfaff wrote: > On Tue, Sep 11, 2012 at 01:30:05PM +0900, Simon Horman wrote: > > From: Isaku Yamahata <yamah...@valinux.co.jp> > > > > Signed-off-by: Isaku Yamahata <yamah...@valinux.co.jp> > > Signed-off-by: Simon Horman <ho...@verge.net.au> > > There are two reasons that a field is currently marked read-only: > > 1. It doesn't make sense, in my opinion, to modify the field. > For example, modifying the Ethernet type means that you'd > lose all the L3 and L4 information and is probably a > mistake. (I don't remember ever seeing a request for that, > actually, even though people ask for all kinds of crazy > things.) > > 2. We haven't implemented setting that field in the kernel > module yet, e.g. IPv6 label. > > I can only guess why oxm_writable is sometimes different in this patch, > but I'd guess: > > a. Some fields are marked read-only because they're from NXM, > not OXM. I don't see any benefit to that; it doesn't hurt to > allow it. (Also, the point of OXM/NXM is extensibility. I > don't think that disallowing it fits that intent.) > > b. Fields in category 2 above are marked writable. That won't > be too useful. (But if you know of some hardware provider > implementation that can write to these fields, I'm happy to > mark those fields writable across the board.) > > So I'd rather just use the 'writable' member instead of introducing a > new one.
Sure, I don't have any objections to that at this time. I'll alter the patch-set accordingly. _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
