On Wed, Apr 10, 2019 at 03:28:10PM +0300, Leon Romanovsky wrote: > On Wed, Apr 10, 2019 at 12:12:41PM +0000, Jason Gunthorpe wrote: > > On Wed, Apr 10, 2019 at 02:59:52PM +0300, Leon Romanovsky wrote: > > > From: Leon Romanovsky <leonro@xxxxxxxxxxxx> > > > > > > There is no need to print parent, because it is > > > part of standard sysfs anyway. > > > > > > [leonro@server ~]$ l /sys/class/infiniband/rxe0/ > > > lrwxrwxrwx 1 root root 0 Apr 10 11:53 device -> ../../../virtio10 > > > > > > Signed-off-by: Leon Romanovsky <leonro@xxxxxxxxxxxx> > > > drivers/infiniband/sw/rxe/rxe_verbs.c | 21 --------------------- > > > 1 file changed, 21 deletions(-) > > > > I think this breaks the udev rules if I recall Steve's debugging. > > It works for me. > > [leonro@server ~]$ lsmod| grep rxe > [leonro@server ~]$ sudo rdma link add rxe_eth0 type rxe netdev eth0 > [ 98.363500] rdma_rxe: loaded > [ 98.462058] infiniband rxe_eth0: set active > [ 98.462421] infiniband rxe_eth0: added eth0 > [leonro@server ~]$ lsmod| grep rxe > rdma_rxe 344064 0 > ip6_udp_tunnel 20480 1 rdma_rxe > udp_tunnel 24576 1 rdma_rxe > ib_uverbs 360448 3 rdma_rxe,rdma_ucm,mlx5_ib > ib_core 839680 11 > rdma_cm,ib_ipoib,rdma_rxe,rpcrdma,iw_cm,ib_iser,ib_umad,rdma_ucm,ib_uverbs,mlx5_ib,ib_cm > [leonro@server ~]$ ls /sys/class/infiniband/rxe_eth0/ > device fw_ver node_desc node_guid node_type ports power subsystem > sys_image_guid uevent > [leonro@server ~]$ sudo rdma link del rxe_eth0 > [leonro@server ~]$ lsmod| grep rxe > rdma_rxe 344064 0 > ip6_udp_tunnel 20480 1 rdma_rxe > udp_tunnel 24576 1 rdma_rxe > ib_uverbs 360448 3 rdma_rxe,rdma_ucm,mlx5_ib > ib_core 839680 11 > rdma_cm,ib_ipoib,rdma_rxe,rpcrdma,iw_cm,ib_iser,ib_umad,rdma_ucm,ib_uverbs,mlx5_ib,ib_cm That isn't enough, dump the udev database for rxe_eth0 and show it is still marked with the right protocol Jason
