Re: [PATCH for-rc] RDMA/core: Fix ENODEV error for iWARP test over vlan

[Leon Romanovsky <leonro@nvidia.com>]

On Thu, Nov 21, 2024 at 08:52:16PM +0100, Zhu Yanjun wrote:
> 在 2024/11/19 7:18, Anumula Murali Mohan Reddy 写道:
> > If traffic is over vlan, cma_validate_port() fails to match
> > net_device ifindex with bound_if_index and results in ENODEV error.
> > As iWARP gid table is static, it contains entry corresponding  to
> > only one net device which is either real netdev or vlan netdev for
> > cases like  siw attached to a vlan interface.
> > This patch fixes the issue by assigning bound_if_index with net
> > device index, if real net device obtained from bound if index matches
> > with net device retrieved from gid table
> > 
> > Fixes: f8ef1be816bf ("RDMA/cma: Avoid GID lookups on iWARP devices")
> > Link: https://lore.kernel.org/all/ZzNgdrjo1kSCGbRz@chelsio.com/
> > Signed-off-by: Anumula Murali Mohan Reddy <anumula@chelsio.com>
> > Signed-off-by: Potnuri Bharat Teja <bharat@chelsio.com>
> > ---
> >   drivers/infiniband/core/cma.c | 17 ++++++++++++++++-
> >   1 file changed, 16 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c
> > index 64ace0b968f0..97657e1810d8 100644
> > --- a/drivers/infiniband/core/cma.c
> > +++ b/drivers/infiniband/core/cma.c
> > @@ -689,7 +689,7 @@ cma_validate_port(struct ib_device *device, u32 port,
> >   	const struct ib_gid_attr *sgid_attr = ERR_PTR(-ENODEV);
> >   	int bound_if_index = dev_addr->bound_dev_if;
> >   	int dev_type = dev_addr->dev_type;
> > -	struct net_device *ndev = NULL;
> > +	struct net_device *ndev = NULL, *pdev = NULL;
> 
> In the original source code, the local variables lay out in the form of
> Reverse Christmas Tree.
> But the new code breaks the rule of Reverse Christmas Tree.

There is no such rule outside of netdev world. I suggest to everyone in
RDMA to stick to that practice as sometimes we need to send the code to
netdev too and it is much easier to keep their coding style from the
beginning.

However, for pure RDMA patches, we don't care.

Thanks


> 
> Zhu Yanjun
> 
> >   	if (!rdma_dev_access_netns(device, id_priv->id.route.addr.dev_addr.net))
> >   		goto out;
> > @@ -714,6 +714,21 @@ cma_validate_port(struct ib_device *device, u32 port,
> >   		rcu_read_lock();
> >   		ndev = rcu_dereference(sgid_attr->ndev);
> > +		if (ndev->ifindex != bound_if_index) {
> > +			pdev = dev_get_by_index_rcu(dev_addr->net, bound_if_index);
> > +			if (pdev) {
> > +				if (is_vlan_dev(pdev)) {
> > +					pdev = vlan_dev_real_dev(pdev);
> > +					if (ndev->ifindex == pdev->ifindex)
> > +						bound_if_index = pdev->ifindex;
> > +				}
> > +				if (is_vlan_dev(ndev)) {
> > +					pdev = vlan_dev_real_dev(ndev);
> > +					if (bound_if_index == pdev->ifindex)
> > +						bound_if_index = ndev->ifindex;
> > +				}
> > +			}
> > +		}
> >   		if (!net_eq(dev_net(ndev), dev_addr->net) ||
> >   		    ndev->ifindex != bound_if_index) {
> >   			rdma_put_gid_attr(sgid_attr);
> 
>