Re: [bug report] net/mlx5e: Simulate missing IPsec TX limits hardware functionality

[Leon Romanovsky <leon@kernel.org>]

On Tue, May 30, 2023 at 02:26:19PM +0300, Dan Carpenter wrote:
> Hello Leon Romanovsky,
> 
> The patch b2f7b01d36a9: "net/mlx5e: Simulate missing IPsec TX limits
> hardware functionality" from Mar 30, 2023, leads to the following
> Smatch static checker warning:
> 
> 	drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec.c:68 mlx5e_ipsec_handle_tx_limit()
> 	warn: sleeping in atomic context
> 
> drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec.c
>     57 static void mlx5e_ipsec_handle_tx_limit(struct work_struct *_work)
>     58 {
>     59         struct mlx5e_ipsec_dwork *dwork =
>     60                 container_of(_work, struct mlx5e_ipsec_dwork, dwork.work);
>     61         struct mlx5e_ipsec_sa_entry *sa_entry = dwork->sa_entry;
>     62         struct xfrm_state *x = sa_entry->x;
>     63 
>     64         spin_lock(&x->lock);
> 
> Holding a spinlock.
> 
>     65         xfrm_state_check_expire(x);
>     66         if (x->km.state == XFRM_STATE_EXPIRED) {
>     67                 sa_entry->attrs.drop = true;
> --> 68                 mlx5e_accel_ipsec_fs_modify(sa_entry);
> 
> This function call will do some GFP_KERNEL allocations so it sleeps.

Thanks Dan, I have a fix in my queue and will submit it.

Thanks

> 
>     69         }
>     70         spin_unlock(&x->lock);
>     71 
>     72         if (sa_entry->attrs.drop)
>     73                 return;
>     74 
>     75         queue_delayed_work(sa_entry->ipsec->wq, &dwork->dwork,
>     76                            MLX5_IPSEC_RESCHED);
>     77 }
> 
> regards,
> dan carpenter