On Wed, 21 Aug 2024 04:56:29 +0000 Mina Almasry wrote: > When net devices propagate xdp configurations to slave devices, or when > core propagates xdp configuration to a device, we will need to perform > a memory provider check to ensure we're not binding xdp to a device > using unreadable netmem. > > Currently ->ndo_bpf calls are all over the place. Adding checks to all > these places would not be ideal. > > Refactor all the ->ndo_bpf calls into one place where we can add this > check in the future. > > Suggested-by: Jakub Kicinski <kuba@xxxxxxxxxx> > Signed-off-by: Mina Almasry <almasrymina@xxxxxxxxxx> > diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c > index f9633a6f8571..73f9416c6c1b 100644 > --- a/drivers/net/bonding/bond_main.c > +++ b/drivers/net/bonding/bond_main.c > @@ -2258,7 +2258,7 @@ int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev, > goto err_sysfs_del; > } > > - res = slave_dev->netdev_ops->ndo_bpf(slave_dev, &xdp); > + res = dev_xdp_propagate(slave_dev, &xdp); I was hoping we can fold the "is there any program present already" but I'm not sure if that check itself isn't buggy... so let's leave that part to someone else. Hangbin, would you be willing to take a look at testing (and fixing) the XDP program propagation? I did a naive test of adding a bond and veth under it, I attached an XDP prog to bond, and nothing happened on the veth. Maybe I'm misreading but I expected the XDP prog to show up on the veth. > diff --git a/drivers/net/hyperv/netvsc_bpf.c b/drivers/net/hyperv/netvsc_bpf.c > index 4a9522689fa4..e01c5997a551 100644 > --- a/drivers/net/hyperv/netvsc_bpf.c > +++ b/drivers/net/hyperv/netvsc_bpf.c > @@ -183,7 +183,7 @@ int netvsc_vf_setxdp(struct net_device *vf_netdev, struct bpf_prog *prog) > xdp.command = XDP_SETUP_PROG; > xdp.prog = prog; > > - ret = vf_netdev->netdev_ops->ndo_bpf(vf_netdev, &xdp); > + ret = dev_xdp_propagate(vf_netdev, &xdp); Again, the driver itself appears rather questionable but we can leave it be :) > @@ -130,7 +130,7 @@ static int bpf_map_offload_ndo(struct bpf_offloaded_map *offmap, > /* Caller must make sure netdev is valid */ > netdev = offmap->netdev; > > - return netdev->netdev_ops->ndo_bpf(netdev, &data); > + return dev_xdp_propagate(netdev, &data); This is not propagation, it's an offload call, let's not convert it > diff --git a/net/xdp/xsk_buff_pool.c b/net/xdp/xsk_buff_pool.c > index c0e0204b9630..f44d68c8d75d 100644 > --- a/net/xdp/xsk_buff_pool.c > +++ b/net/xdp/xsk_buff_pool.c > @@ -149,7 +149,7 @@ static void xp_disable_drv_zc(struct xsk_buff_pool *pool) > bpf.xsk.pool = NULL; > bpf.xsk.queue_id = pool->queue_id; > > - err = pool->netdev->netdev_ops->ndo_bpf(pool->netdev, &bpf); > + err = dev_xdp_propagate(pool->netdev, &bpf); > > if (err) > WARN(1, "Failed to disable zero-copy!\n"); > @@ -215,7 +215,7 @@ int xp_assign_dev(struct xsk_buff_pool *pool, > bpf.xsk.pool = pool; > bpf.xsk.queue_id = queue_id; > > - err = netdev->netdev_ops->ndo_bpf(netdev, &bpf); > + err = dev_xdp_propagate(netdev, &bpf); > if (err) > goto err_unreg_pool; > That's also not xdp propagation. If you're not doing so already in your series - you should look at queue state here directly and check if it has MP installed. Conversely you should look at rxq->pool when binding MP. But as I said, that's not part of the XDP refactor, just as part of your series. So in case my ramblings were confusing - code LG, but ditch the net/xdp/xsk_buff_pool.c and kernel/bpf/offload.c changes.
