]> git.baikalelectronics.ru Git - kernel.git/commitdiff
net/mlx5: Fix cleanup of bridge delayed work
authorShay Drory <shayd@nvidia.com>
Mon, 13 Sep 2021 12:48:11 +0000 (15:48 +0300)
committerSaeed Mahameed <saeedm@nvidia.com>
Tue, 12 Oct 2021 20:51:45 +0000 (13:51 -0700)
Currently, bridge cleanup is calling to cancel_delayed_work(). When this
function is finished, there is a chance that the delayed work is still
running. Also, the delayed work is queueing itself.
As a result, we might execute the delayed work after the bridge cleanup
have finished and hit a null-ptr oops[1].

Fix it by using cancel_delayed_work_sync(), which is waiting until the
work is done and will cancel the queue work.

[1]
[ 8202.143043 ] BUG: kernel NULL pointer dereference, address: 0000000000000000
[ 8202.144438 ] #PF: supervisor write access in kernel mode
[ 8202.145476 ] #PF: error_code(0x0002) - not-present page
[ 8202.146520 ] PGD 0 P4D 0
[ 8202.147126 ] Oops: 0002 [#1] SMP NOPTI
[ 8202.147899 ] CPU: 1 PID: 0 Comm: swapper/1 Not tainted 5.14.0-rc6_for_upstream_min_debug_2021_08_25_16_06 #1
[ 8202.149741 ] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
[ 8202.151908 ] RIP: 0010:_raw_spin_lock+0xc/0x20
[ 8202.156234 ] RSP: 0018:ffff88846f885ea0 EFLAGS: 00010046
[ 8202.157289 ] RAX: 0000000000000000 RBX: ffff88846f880000 RCX: 0000000000000000
[ 8202.158731 ] RDX: 0000000000000001 RSI: ffff8881004000c8 RDI: 0000000000000000
[ 8202.160177 ] RBP: ffff8881fe684978 R08: ffff888100140000 R09: ffffffff824455b8
[ 8202.161569 ] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000001
[ 8202.163004 ] R13: 0000000000000012 R14: 0000000000000200 R15: ffff88812992d000
[ 8202.164018 ] FS:  0000000000000000(0000) GS:ffff88846f880000(0000) knlGS:0000000000000000
[ 8202.164960 ] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 8202.165634 ] CR2: 0000000000000000 CR3: 0000000108cac004 CR4: 0000000000370ea0
[ 8202.166450 ] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 8202.167807 ] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 8202.168852 ] Call Trace:
[ 8202.169421 ]  <IRQ>
[ 8202.169792 ]  __queue_work+0xf2/0x3d0
[ 8202.170481 ]  ? queue_work_node+0x40/0x40
[ 8202.171270 ]  call_timer_fn+0x2b/0x100
[ 8202.171932 ]  __run_timers.part.0+0x152/0x220
[ 8202.172717 ]  ? __hrtimer_run_queues+0x171/0x290
[ 8202.173526 ]  ? kvm_clock_get_cycles+0xd/0x10
[ 8202.174232 ]  ? ktime_get+0x35/0x90
[ 8202.174943 ]  run_timer_softirq+0x26/0x50
[ 8202.175745 ]  __do_softirq+0xc7/0x271
[ 8202.176373 ]  irq_exit_rcu+0x93/0xb0
[ 8202.176983 ]  sysvec_apic_timer_interrupt+0x72/0x90
[ 8202.177755 ]  </IRQ>
[ 8202.178245 ]  asm_sysvec_apic_timer_interrupt+0x12/0x20

Fixes: c636a0f0f3f0 ("net/mlx5: Bridge, dynamic entry ageing")
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Maor Gottlieb <maorg@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
drivers/net/ethernet/mellanox/mlx5/core/en/rep/bridge.c

index b5ddaa82755f170eae6c6a17d70861f9013f23b7..c6d2f8c78db71ab2342cbb368de174831724e102 100644 (file)
@@ -475,9 +475,6 @@ void mlx5e_rep_bridge_init(struct mlx5e_priv *priv)
                esw_warn(mdev, "Failed to allocate bridge offloads workqueue\n");
                goto err_alloc_wq;
        }
-       INIT_DELAYED_WORK(&br_offloads->update_work, mlx5_esw_bridge_update_work);
-       queue_delayed_work(br_offloads->wq, &br_offloads->update_work,
-                          msecs_to_jiffies(MLX5_ESW_BRIDGE_UPDATE_INTERVAL));
 
        br_offloads->nb.notifier_call = mlx5_esw_bridge_switchdev_event;
        err = register_switchdev_notifier(&br_offloads->nb);
@@ -500,6 +497,9 @@ void mlx5e_rep_bridge_init(struct mlx5e_priv *priv)
                         err);
                goto err_register_netdev;
        }
+       INIT_DELAYED_WORK(&br_offloads->update_work, mlx5_esw_bridge_update_work);
+       queue_delayed_work(br_offloads->wq, &br_offloads->update_work,
+                          msecs_to_jiffies(MLX5_ESW_BRIDGE_UPDATE_INTERVAL));
        return;
 
 err_register_netdev:
@@ -523,10 +523,10 @@ void mlx5e_rep_bridge_cleanup(struct mlx5e_priv *priv)
        if (!br_offloads)
                return;
 
+       cancel_delayed_work_sync(&br_offloads->update_work);
        unregister_netdevice_notifier(&br_offloads->netdev_nb);
        unregister_switchdev_blocking_notifier(&br_offloads->nb_blk);
        unregister_switchdev_notifier(&br_offloads->nb);
-       cancel_delayed_work(&br_offloads->update_work);
        destroy_workqueue(br_offloads->wq);
        rtnl_lock();
        mlx5_esw_bridge_cleanup(esw);