On Fri, Aug 16, 2013 at 3:35 PM, Jesse Gross <je...@nicira.com> wrote:
> On Fri, Aug 16, 2013 at 3:28 PM, Pravin Shelar <pshe...@nicira.com> wrote: > > > > > > > > On Fri, Aug 16, 2013 at 3:24 PM, Jesse Gross <je...@nicira.com> wrote: > >> > >> On Fri, Aug 16, 2013 at 3:00 PM, Pravin B Shelar <pshe...@nicira.com> > >> wrote: > >> > net_namespace-device can get registered after module init, e.g. vxlan > >> > registers name-space-device on port add. On kernel without namespace > >> > support __net_init is defined as __init which cause panic on vxlan > port > >> > add. Following patch fixes it. > >> > >> The fix looks basically right to me but I think the commit message is > >> a little misleading - on new kernels __net_init is defined as __init > >> when namespace support is compiled out and otherwise it is a no-op. > >> Therefore, on old kernels the backport is actually correct since there > >> are never any namespaces (and it fact this was reported on 2.6.32, > >> which doesn't use the backport). > >> > >> The problem is that we shouldn't be using these annotations in places > >> where the namespace devices can be registered after module init. I > >> believe that we actually need to keep the __net_init backported > >> definition for old kernels, since we still have one proper use in > >> datapath.c. > > > > > > right, thats what this patch does, it removes annotations. > > I am not sure what is missing. > > I think this part will make it not compile on old kernels: > > -#define __net_init __init > -#define __net_exit __exit > #endif /* 2.6.24 */ > ok, I will send updated patch.
_______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
