Re: linux-next: manual merge of the rdma tree with Linus' tree

[Jason Gunthorpe <jgg@nvidia.com>]

On Wed, Jan 15, 2025 at 08:27:21AM +1100, Stephen Rothwell wrote:
> Hi Jason,
> 
> On Tue, 14 Jan 2025 16:48:28 -0400 Jason Gunthorpe <jgg@nvidia.com> wrote:
> >
> > I think we need to retain the ib_get_curr_port_state() call:
> 
> Why? 

That is how the new system is supposed to work.. Drivers are supposed
to call ib_get_curr_port_state() not open code it. This matches what
is in the for-next tree:

        attr->state = ib_get_curr_port_state(sdev->netdev);
        attr->phys_state = attr->state == IB_PORT_ACTIVE ?
                IB_PORT_PHYS_STATE_LINK_UP : IB_PORT_PHYS_STATE_DISABLED;

> It is no longer used to determine the attr->phys-state value and
> then attr->state is set again just below.  

Ah, missed that, it should be deleted also, and the phys_state should
use the other hunk too:

--- drivers/infiniband/sw/siw/siw_verbs.c       2025-01-14 16:37:02.023738738 -0400
+++ /home/jgg/oss/testing-k.o/drivers/infiniband/sw/siw/siw_verbs.c     2025-01-14 20:38:42.611302948 -0400
@@ -189,10 +189,9 @@
        attr->max_msg_sz = -1;
        attr->max_mtu = ib_mtu_int_to_enum(ndev->max_mtu);
        attr->active_mtu = ib_mtu_int_to_enum(READ_ONCE(ndev->mtu));
-       attr->phys_state = (netif_running(ndev) && netif_carrier_ok(ndev)) ?
+       attr->state = ib_get_curr_port_state(ndev);
+       attr->phys_state = attr->state == IB_PORT_ACTIVE ?
                IB_PORT_PHYS_STATE_LINK_UP : IB_PORT_PHYS_STATE_DISABLED;
-       attr->state = attr->phys_state == IB_PORT_PHYS_STATE_LINK_UP ?
-               IB_PORT_ACTIVE : IB_PORT_DOWN;
        attr->port_cap_flags = IB_PORT_CM_SUP | IB_PORT_DEVICE_MGMT_SUP;
        /*
         * All zero

Jason