[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20110421231434.GD7888@rere.qmqm.pl>
Date: Fri, 22 Apr 2011 01:14:34 +0200
From: Michał Mirosław <mirq-linux@...e.qmqm.pl>
To: Ben Hutchings <bhutchings@...arflare.com>
Cc: Vladislav Zolotarov <vladz@...adcom.com>,
Eric Dumazet <eric.dumazet@...il.com>,
"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
Eilon Greenstein <eilong@...adcom.com>
Subject: Re: [PATCH v5] net: bnx2x: convert to hw_features
On Thu, Apr 21, 2011 at 11:52:22PM +0100, Ben Hutchings wrote:
> On Fri, 2011-04-22 at 00:15 +0200, Michał Mirosław wrote:
> > On Thu, Apr 21, 2011 at 08:19:19PM +0100, Ben Hutchings wrote:
> > > /* Transfer changeable features to wanted_features and enable
> > > * software offloads (GSO and GRO).
> > > */
> > > dev->hw_features |= NETIF_F_SOFT_FEATURES;
> > > dev->features |= NETIF_F_SOFT_FEATURES;
> > > dev->wanted_features = dev->features & dev->hw_features;
> > >
> > > This doesn't work correctly for features that are always enabled, like
> > > NETIF_F_HW_VLAN_RX in bnx2x, which are set in dev->features but not in
> > > dev->hw_features.
> >
> > > The name 'hw_features' really wasn't a good choice - the obvious meaning
> > > and the meaning assumed by this code is 'hardware-supported features'
> > > and not 'hardware-supported features that can be toggled'. And since we
> > > add NETIF_F_SOFT_FEATURES, it really only means 'features that can be
> > > toggled'.
> >
> > I won't argue about hw_features name - I just couldn't find a better one.
> > Comment in include/linux/netdevice.h clearly explains the purpose of this
> > field.
> >
> > wanted_features is supposed to be limited by hw_features (so that it's always
> > true that (hw_features & wanted_features) == wanted_features). If you have
> > an idea how to make that more clear, I'd be happy to update descriptions.
>
> Then the computation of 'changed' in __ethtool_set_flags() is wrong:
>
> /* allow changing only bits set in hw_features */
> changed = (data ^ dev->wanted_features) & flags_dup_features;
> if (changed & ~dev->hw_features)
> return (changed & dev->hw_features) ? -EINVAL : -EOPNOTSUPP;
Yes! This doesn't take account of features enabled but not togglable.
> You need to add something like:
>
> /* Features that are requested to be on, are already on, and cannot
> * be changed, have not changed.
> */
> changes &= ~(data & dev->features & ~dev->hw_features);
>
> It seems like there ought to be a way to simplify that, though!
Maybe something I just sent will do.
Best Regards,
Michał Mirosław
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists