From f4aebbfb56ed0c186adbeb2799df836da50f78e3 Mon Sep 17 00:00:00 2001 From: Aya Levin Date: Wed, 20 May 2020 10:37:42 +0300 Subject: net/mlx5e: Fix CPU mapping after function reload to avoid aRFS RX crash After function reload, CPU mapping used by aRFS RX is broken, leading to a kernel panic. Fix by moving initialization of rx_cpu_rmap from netdev_init to netdev_attach. IRQ table is re-allocated on mlx5_load, but netdev is not re-initialize. Trace of the panic: [ 22.055672] general protection fault, probably for non-canonical address 0x785634120000ff1c: 0000 [#1] SMP PTI [ 22.065010] CPU: 4 PID: 0 Comm: swapper/4 Not tainted 5.7.0-rc2-for-upstream-perf-2020-04-21_16-34-03-31 #1 [ 22.067967] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.12.1-0-ga5cab58e9a3f-prebuilt.qemu.org 04/01/2014 [ 22.071174] RIP: 0010:get_rps_cpu+0x267/0x300 [ 22.075692] RSP: 0018:ffffc90000244d60 EFLAGS: 00010202 [ 22.076888] RAX: ffff888459b0e400 RBX: 0000000000000000 RCX:0000000000000007 [ 22.078364] RDX: 0000000000008884 RSI: ffff888467cb5b00 RDI:0000000000000000 [ 22.079815] RBP: 00000000ff342b27 R08: 0000000000000007 R09:0000000000000003 [ 22.081289] R10: ffffffffffffffff R11: 00000000000070cc R12:ffff888454900000 [ 22.082767] R13: ffffc90000e5a950 R14: ffffc90000244dc0 R15:0000000000000007 [ 22.084190] FS: 0000000000000000(0000) GS:ffff88846fc80000(0000)knlGS:0000000000000000 [ 22.086161] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 22.087427] CR2: ffffffffffffffff CR3: 0000000464426003 CR4:0000000000760ee0 [ 22.088888] DR0: 0000000000000000 DR1: 0000000000000000 DR2:0000000000000000 [ 22.090336] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:0000000000000400 [ 22.091764] PKRU: 55555554 [ 22.092618] Call Trace: [ 22.093442] [ 22.094211] ? kvm_clock_get_cycles+0xd/0x10 [ 22.095272] netif_receive_skb_list_internal+0x258/0x2a0 [ 22.096460] gro_normal_list.part.137+0x19/0x40 [ 22.097547] napi_complete_done+0xc6/0x110 [ 22.098685] mlx5e_napi_poll+0x190/0x670 [mlx5_core] [ 22.099859] net_rx_action+0x2a0/0x400 [ 22.100848] __do_softirq+0xd8/0x2a8 [ 22.101829] irq_exit+0xa5/0xb0 [ 22.102750] do_IRQ+0x52/0xd0 [ 22.103654] common_interrupt+0xf/0xf [ 22.104641] Fixes: 4383cfcc65e7 ("net/mlx5: Add devlink reload") Signed-off-by: Aya Levin Reviewed-by: Eran Ben Elisha Signed-off-by: Saeed Mahameed --- drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'drivers') diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c index 888e38b21c3d..081f15074cac 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c @@ -5118,6 +5118,10 @@ static int mlx5e_init_nic_rx(struct mlx5e_priv *priv) if (err) goto err_destroy_flow_steering; +#ifdef CONFIG_MLX5_EN_ARFS + priv->netdev->rx_cpu_rmap = mlx5_eq_table_get_rmap(priv->mdev); +#endif + return 0; err_destroy_flow_steering: @@ -5289,10 +5293,6 @@ int mlx5e_netdev_init(struct net_device *netdev, /* netdev init */ netif_carrier_off(netdev); -#ifdef CONFIG_MLX5_EN_ARFS - netdev->rx_cpu_rmap = mlx5_eq_table_get_rmap(mdev); -#endif - return 0; err_free_cpumask: -- cgit v1.2.3-59-g8ed1b