Yes, I was testing on top of commits in https://github.com/jgunthorpe/linux/commits/device_locking_cleanup I have tested the latest (with the updated patch) on branch wip/jgg-for-next Thanks, Parvi > On Feb 18, 2019, at 8:38 PM, Jason Gunthorpe <jgg@xxxxxxxx> wrote: > > On Sat, Feb 09, 2019 at 09:28:30AM -0800, Parvi Kaustubhi wrote: >> There is a dead lock in usnic ib_register and netdev_notify >> path. >> >> usnic_ib_discover_pf() >> | mutex_lock(&usnic_ib_ibdev_list_lock); >> | usnic_ib_device_add(); >> | ib_register_device() >> | usnic_ib_query_port() >> | mutex_lock(&us_ibdev->usdev_lock); >> | ib_get_eth_speed() >> | rtnl_lock() >> >> order of lock: &usnic_ib_ibdev_list_lock -> usdev_lock -> rtnl_lock >> >> rtnl_lock() >> | usnic_ib_netdevice_event() >> | mutex_lock(&usnic_ib_ibdev_list_lock); >> >> order of lock: rtnl_lock -> &usnic_ib_ibdev_list_lock >> >> Solution is to use ib_device_get_by_netdev() to lookup ib_dev while >> handling netdev/ inet events. >> >> Signed-off-by: Parvi Kaustubhi <pkaustub@xxxxxxxxx> >> Reviewed-by: Govindarajulu Varadarajan <gvaradar@xxxxxxxxx> >> Reviewed-by: Tanmay Inamdar <tinamdar@xxxxxxxxx> >> Changelog: >> >> v3->v4 >> * Added missing ib_device_put* calls. >> * Fixed wonky control flow >> >> v2->v3: >> * Jason: drivers should not hold any locks while calling ib_unregister_device() >> * Jason: use https://github.com/jgunthorpe/linux/commits/device_locking_cleanup >> >> v1->v2: >> * Have notifier blocks in usnic_ib_dev instead of using workqueue to defer >> event handling. >> drivers/infiniband/hw/usnic/usnic_ib_main.c | 30 ++++++++++++++--------------- >> 1 file changed, 14 insertions(+), 16 deletions(-) > > I added the hunks to use setnedev from my as-yet-unsent patch and > applies this to for-next. I assume you tested on top of that. You > should probably test the final result. > > Thanks, > Jason
