Re: [PATCH] cls_u32: use skb_copy_bits() to dereference data safely
From: Changli Gao <hidden>
Date: 2010-06-02 13:43:37
From: Changli Gao <hidden>
Date: 2010-06-02 13:43:37
On Wed, Jun 2, 2010 at 9:36 PM, jamal [off-list ref] wrote:
On Wed, 2010-06-02 at 21:14 +0800, Changli Gao wrote:quoted
Maybe skb_header_pointer() is lighter.A little worse than skb_copy_bits(). In any case, this change is going to hurt.
Why? it is an inline function, and in most cases, there isn't any function call.
Dave, can we assume the upper layers(qdiscs in this case) are responsible for any linearizing? Changli, if you have time - can you also audit tcf_pedit() since it follows TheLinuxWay(tm).
Yea, pedit has the same issue. Besides this issue, they should use get_unaligned() instead. -- Regards, Changli Gao(xiaosuo@gmail.com)