On Tue, 6 May 2025 11:35:28 +0300 Gur Stavi wrote:
> + page = page_pool_alloc_frag(page_pool, &page_offset, buf_len,
> + GFP_ATOMIC | __GFP_COMP);
No need for __GFP_COMP. page pool will add it if order > 0
> +static struct sk_buff *hinic3_fetch_rx_buffer(struct hinic3_rxq *rxq,
> + u32 pkt_len)
> +{
> + struct net_device *netdev = rxq->netdev;
> + struct sk_buff *skb;
> + u32 sge_num;
> +
> + skb = netdev_alloc_skb_ip_align(netdev, HINIC3_RX_HDR_SIZE);
Looks like this is only called from NAPI so you can use napi_alloc_skb()
it has an object cache so it is measurably faster
