[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20181210.114403.1590402440504388038.davem@davemloft.net>
Date: Mon, 10 Dec 2018 11:44:03 -0800 (PST)
From: David Miller <davem@...emloft.net>
To: horms+renesas@...ge.net.au
Cc: sergei.shtylyov@...entembedded.com, magnus.damm@...il.com,
netdev@...r.kernel.org, linux-renesas-soc@...r.kernel.org
Subject: Re: [PATCH/RFT net] ravb: expand rx descriptor data to accommodate
hw checksum
From: Simon Horman <horms+renesas@...ge.net.au>
Date: Mon, 10 Dec 2018 09:59:17 +0100
> +#define RAVB_CSUM_LEN 2
> +
...
> priv->rx_buf_sz = (ndev->mtu <= 1492 ? PKT_BUF_SZ : ndev->mtu) +
> - ETH_HLEN + VLAN_HLEN;
> + ETH_HLEN + VLAN_HLEN + RAVB_CSUM_LEN;
...
> + if (unlikely(skb->len < RAVB_CSUM_LEN))
...
> - hw_csum = skb_tail_pointer(skb) - 2;
> + hw_csum = skb_tail_pointer(skb) - RAVB_CSUM_LEN;
...
> - skb_trim(skb, skb->len - 2);
> + skb_trim(skb, skb->len - RAVB_CSUM_LEN);
Unlike Sergei, I think this macro define should be kept in the fix.
It is absolutely crucial for anyone reading this code to understand
what this value is all about.
People reading the code aren't able to go automatically back to a
commit to learn what this value means, and even if they could they
shouldn't have to do so for a bunch of magic '2' constants placed all
over.
Even in the most fundamental way, the macro is required to satisfy
the "no magic constants" rule for kernel code.
Powered by blists - more mailing lists