Re: [PATCH v2] ethernet: call __skb_pull() in eth_type_trans()

[Changli Gao <xiaosuo@gmail.com>]

On Tue, May 4, 2010 at 3:54 AM, David Miller <davem@davemloft.net> wrote:
> From: Changli Gao <xiaosuo@gmail.com>
> Date: Mon,  3 May 2010 22:12:52 +0800
>
>> @@ -162,7 +162,10 @@ __be16 eth_type_trans(struct sk_buff *skb, struct net_device *dev)
>>
>>       skb->dev = dev;
>>       skb_reset_mac_header(skb);
>> -     skb_pull_inline(skb, ETH_HLEN);
>> +     if (unlikely(skb->len < ETH_ZLEN))
>> +             dev_warn(&dev->dev, "too small ethernet packet: %u bytes\n",
>> +                      skb->len);
>> +     __skb_pull(skb, ETH_HLEN);
>>       eth = eth_hdr(skb);
>
> And now it's even more expensive than skb_pull_inline() :-)
>
> Really, things are fine as-is.
>

It seems no callers pass eth_type_trans() a packet, whose length is
less than ETH_HLEN. It means that skb_pull() always returns non-NULL.
And if skb_pull() returns NULL, the later memory dereferences must be
invalid. So, we can safely call __skb_pull() instead of skb_pull().
And If the current code works, there is no reason the new code without
the check(skb->len < ETH_HLEN) doesn't work.

As Eric mentioned above, GRE only assures the length of the packets
passed to eth_type_trans() isn't less than ETH_HLEN, we should check
skb->len before we dereference skb->data.

        rawp = skb->data;

        /*
         *      This is a magic hack to spot IPX packets. Older Novell breaks
         *      the protocol design and runs IPX over 802.3 without an 802.2 LLC
         *      layer. We look for FFFF which isn't a used 802.2 SSAP/DSAP. This
         *      won't work for fault tolerant netware but does for the rest.
         */
        if (*(unsigned short *)rawp == 0xFFFF)
                return htons(ETH_P_802_3);


For performance, how about inlining eth_type_trans(). Because its main
users are NIC drivers, and there aren't likely many kinds of NICs at
the same time, inlining it won't increases the size of the kernel
image much.

-- 
Regards，
Changli Gao(xiaosuo@gmail.com)