]> git.proxmox.com Git - mirror_ubuntu-kernels.git/commit
net: thunderx: use spin_lock_bh in nicvf_set_rx_mode_task()
authorXin Long <lucien.xin@gmail.com>
Tue, 4 Aug 2020 07:02:30 +0000 (15:02 +0800)
committerDavid S. Miller <davem@davemloft.net>
Tue, 4 Aug 2020 20:03:36 +0000 (13:03 -0700)
commitbab9693a9a8c6dd19f670408ec1e78e12a320682
treee9bdca142d32f2ee95dc4c9fa733c59c9a88cfec
parent2ac24d6d685cb1c0e0fbb189dc607b05720fb5d1
net: thunderx: use spin_lock_bh in nicvf_set_rx_mode_task()

A dead lock was triggered on thunderx driver:

        CPU0                    CPU1
        ----                    ----
   [01] lock(&(&nic->rx_mode_wq_lock)->rlock);
                           [11] lock(&(&mc->mca_lock)->rlock);
                           [12] lock(&(&nic->rx_mode_wq_lock)->rlock);
   [02] <Interrupt> lock(&(&mc->mca_lock)->rlock);

The path for each is:

  [01] worker_thread() -> process_one_work() -> nicvf_set_rx_mode_task()
  [02] mld_ifc_timer_expire()
  [11] ipv6_add_dev() -> ipv6_dev_mc_inc() -> igmp6_group_added() ->
  [12] dev_mc_add() -> __dev_set_rx_mode() -> nicvf_set_rx_mode()

To fix it, it needs to disable bh on [1], so that the timer on [2]
wouldn't be triggered until rx_mode_wq_lock is released. So change
to use spin_lock_bh() instead of spin_lock().

Thanks to Paolo for helping with this.

v1->v2:
  - post to netdev.

Reported-by: Rafael P. <rparrazo@redhat.com>
Tested-by: Dean Nelson <dnelson@redhat.com>
Fixes: 469998c861fa ("net: thunderx: prevent concurrent data re-writing by nicvf_set_rx_mode")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/cavium/thunder/nicvf_main.c