Re: problem with vr0 (enabling polling)

[Gary Allan <dragonfly@xxxxxxxxxxxxxxx>]

George Georgalis wrote:
> On Wed, Jan 26, 2005 at 07:37:02PM +0000, Gary Allan wrote:
>
> > I've noticed that polling support for the vr device can not be 
> > controlled or displayed from ifconfig (Unlike in FreeBSD). Adding the 
> > following seems to enable that support. No guarantees on correctness but 
> > maybe someone else can take a look.
> >
> > cd /usr/src/sys/dev/netif/vr/
> > diff -ruN if_vr.c if_vr.c.new
> > --- if_vr.c     2005-01-26 19:19:42.000000000 +0000
> > +++ if_vr.c.new 2005-01-26 19:26:04.000000000 +0000
> > @@ -880,6 +880,10 @@
> >       ifp->if_init = vr_init;
> >       ifp->if_baudrate = 10000000;
> >       ifp->if_snd.ifq_maxlen = VR_TX_LIST_CNT - 1;
> > +#ifdef DEVICE_POLLING
> > +       ifp->if_capabilities |= IFCAP_POLLING;
> > +#endif
> > +       ifp->if_capenable = ifp->if_capabilities;
> >
> >       /*
> >        * Do MII setup.
> > @@ -1690,6 +1694,11 @@
> >               mii = device_get_softc(sc->vr_miibus);
> >               error = ifmedia_ioctl(ifp, ifr, &mii->mii_media, command);
> >               break;
> > +       case SIOCSIFCAP:
> > +                ifp->if_capenable &= ~IFCAP_POLLING;
> > +                ifp->if_capenable |= ifr->ifr_reqcap & IFCAP_POLLING;
> > +                break;
> > +
> >       default:
> >               error = EINVAL;
> >               break;
> >
> >
> > After the modifications with DEVICE_POLLING kernel support and 
> > kern.polling.enable=1.
>
>
> Neat, now I can see that polling is really on...
>
> # ifconfig vr0 -polling && ifconfig vr0
> vr0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
>         inet6 fe80::230:1bff:feb4:99a8%vr0 prefixlen 64 scopeid 0x3 
>         inet 0.0.0.0 netmask 0xff000000 broadcast 255.255.255.255
>         ether 00:30:1b:b4:99:a8
>         media: Ethernet none (10baseT/UTP)
>         status: no carrier
> # ifconfig vr0 polling && ifconfig vr0
> vr0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
>         options=40<polling>
>         inet6 fe80::230:1bff:feb4:99a8%vr0 prefixlen 64 scopeid 0x3 
>         inet 0.0.0.0 netmask 0xff000000 broadcast 255.255.255.255
>         ether 00:30:1b:b4:99:a8
>         media: Ethernet none (10baseT/UTP)
>         status: no carrier
>
> And I know exactly what people mean by vr0 code stinks, what a mess.
>
> BTW - would it be reasonable to teach patch that tabs and spaces are
> both whitespace?
>
> // George

I think this could be a bug.

The vr device has polling support I'm not sure that it can actually be 
enabled. I think this patch (or similar) is needed to set 
ifp->if_capabilities |= IFCAP_POLLING; and set/clear the IFCAP_POLLING 
flag in ifp->if_capenable. As it stands I can't see how 
ifp->if_capenable can ever have this option enabled.

Also the code only tests against ifp.if_flags & IFF_POLLING and again I 
can't see where this flag is set/cleared. Other similar code (rl) tests 
against ifp->if_capenable & IFCAP_POLLING and ifp.if_flags & IFF_POLLING 
but only modifies the if_capenable flags. Can someone explain how 
if_capenable and if_flags operate?

Regards

G.Allan