Re: [PATCH v2 net] bonding: add xdp_features support

[Lorenzo Bianconi <lorenzo@kernel.org>]

[-- Attachment #1: Type: text/plain, Size: 3473 bytes --]

> On Mon, 2023-05-01 at 15:33 +0200, Daniel Borkmann wrote:
> > On 4/30/23 12:02 PM, Lorenzo Bianconi wrote:
> > > Introduce xdp_features support for bonding driver according to the slave
> > > devices attached to the master one. xdp_features is required whenever we
> > > want to xdp_redirect traffic into a bond device and then into selected
> > > slaves attached to it.
> > > 
> > > Fixes: 66c0e13ad236 ("drivers: net: turn on XDP features")
> > > Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
> > 
> > Please also keep Jussi in Cc for bonding + XDP reviews [added here].
> 
> Perhaps worth adding such info to the maintainer file for future
> memory?
> 
> > > ---
> > > Change since v1:
> > > - remove bpf self-test patch from the series
> > 
> > Given you targeted net tree, was this patch run against BPF CI locally from
> > your side to avoid breakage again?
> > 
> > Thanks,
> > Daniel
> > 
> > > ---
> > >   drivers/net/bonding/bond_main.c    | 48 ++++++++++++++++++++++++++++++
> > >   drivers/net/bonding/bond_options.c |  2 ++
> > >   include/net/bonding.h              |  1 +
> > >   3 files changed, 51 insertions(+)
> > > 
> > > diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
> > > index 710548dbd0c1..c98121b426a4 100644
> > > --- a/drivers/net/bonding/bond_main.c
> > > +++ b/drivers/net/bonding/bond_main.c
> > > @@ -1789,6 +1789,45 @@ static void bond_ether_setup(struct net_device *bond_dev)
> > >   	bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING;
> > >   }
> > >   
> > > +void bond_xdp_set_features(struct net_device *bond_dev)
> > > +{
> > > +	struct bonding *bond = netdev_priv(bond_dev);
> > > +	xdp_features_t val = NETDEV_XDP_ACT_MASK;
> > > +	struct list_head *iter;
> > > +	struct slave *slave;
> > > +
> > > +	ASSERT_RTNL();
> > > +
> > > +	if (!bond_xdp_check(bond)) {
> > > +		xdp_clear_features_flag(bond_dev);
> > > +		return;
> > > +	}
> > > +
> > > +	bond_for_each_slave(bond, slave, iter) {
> > > +		struct net_device *dev = slave->dev;
> > > +
> > > +		if (!(dev->xdp_features & NETDEV_XDP_ACT_BASIC)) {
> > > +			xdp_clear_features_flag(bond_dev);
> > > +			return;
> > > +		}
> > > +
> > > +		if (!(dev->xdp_features & NETDEV_XDP_ACT_REDIRECT))
> > > +			val &= ~NETDEV_XDP_ACT_REDIRECT;
> > > +		if (!(dev->xdp_features & NETDEV_XDP_ACT_NDO_XMIT))
> > > +			val &= ~NETDEV_XDP_ACT_NDO_XMIT;
> > > +		if (!(dev->xdp_features & NETDEV_XDP_ACT_XSK_ZEROCOPY))
> > > +			val &= ~NETDEV_XDP_ACT_XSK_ZEROCOPY;
> > > +		if (!(dev->xdp_features & NETDEV_XDP_ACT_HW_OFFLOAD))
> > > +			val &= ~NETDEV_XDP_ACT_HW_OFFLOAD;
> > > +		if (!(dev->xdp_features & NETDEV_XDP_ACT_RX_SG))
> > > +			val &= ~NETDEV_XDP_ACT_RX_SG;
> > > +		if (!(dev->xdp_features & NETDEV_XDP_ACT_NDO_XMIT_SG))
> > > +			val &= ~NETDEV_XDP_ACT_NDO_XMIT_SG;
> 
> Can we expect NETDEV_XDP_ACT_MASK changing in the future (e.g. new
> features to be added)? If so the above code will break silently, as the
> new features will be unconditionally enabled. What about adding a
> BUILD_BUG() to catch such situation? 

I used NETDEV_XDP_ACT_MASK here in order to enable all the XDP features when
we do not have any slave device attache to the bond one. If we add a new
feature to netdev_xdp_act in the future I would say it is fine we inherit it
here otherwise we will need to explicitly add it.

Regards,
Lorenzo

> 
> > 
> Cheers,
> 
> Paolo
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]