答复: [PATCH][net-next] nfp: flower: use GFP_KERNEL in idr_preload

[Li,Rongqing]

> 
> From: Li RongQing <lirongqing@xxxxxxxxx>
> 
> idr_preload() is called in a sleepable context before acquiring a spinlock. Change
> the allocation flag from GFP_ATOMIC to GFP_KERNEL to allow direct reclaim
> and improve allocation success rate under memory pressure.
> 
> Signed-off-by: Li RongQing <lirongqing@xxxxxxxxx>
> ---
>  drivers/net/ethernet/netronome/nfp/flower/main.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/netronome/nfp/flower/main.c
> b/drivers/net/ethernet/netronome/nfp/flower/main.c
> index 83eaa5a..8f232c6 100644
> --- a/drivers/net/ethernet/netronome/nfp/flower/main.c
> +++ b/drivers/net/ethernet/netronome/nfp/flower/main.c
> @@ -63,7 +63,7 @@ nfp_flower_get_internal_port_id(struct nfp_app *app,
> struct net_device *netdev)
>  	if (id > 0)
>  		return id;
> 
> -	idr_preload(GFP_ATOMIC);
> +	idr_preload(GFP_KERNEL);
>  	spin_lock_bh(&priv->internal_ports.lock);
>  	id = idr_alloc(&priv->internal_ports.port_ids, netdev,
>  		       NFP_MIN_INT_PORT_ID, NFP_MAX_INT_PORT_ID,
> GFP_ATOMIC);
> --
> 2.9.4


Sorry, please drop this patch,;

nfp_flower_get_internal_port_id() maybe be called in system_highpri_wq workqueue, so GFP_ATOMIC is right

[Li,Rongqing]