Re: [PATCH net-next] net: Check frag_lists first to prevent data out of order

[Weilong Chen <chenweilong@huawei.com>]

在 2015/8/28 12:37, Eric Dumazet 写道:
> On Fri, 2015-08-28 at 11:35 +0800, Weilong Chen wrote:
>> Thanks for reply.
>>
>>> On Wed, 2015-08-26 at 19:12 -0700, Eric Dumazet wrote:
>>>> On Thu, 2015-08-27 at 08:56 +0800, chenweilong@huawei.com wrote:
>>>>> From: Weilong Chen <chenweilong@huawei.com>
>>>>>
>>>>> When try to merge several skbs to prior one, if the frag_list is
>>>>> used and the the last one is a small packet, once the condition
>>>>> "len <= skb_tailroom(to)" is satisfied, we will get a wrong
>>>>> packet!
>>>>> This patch just check frag_lists before the condtion to prevent
>>>>> this from happening.
>>>>>
>>>>> Signed-off-by: Weilong Chen <chenweilong@huawei.com>
>>>>> ---
>>>>>    net/core/skbuff.c | 6 +++---
>>>>>    1 file changed, 3 insertions(+), 3 deletions(-)
>>>>>
>>>>> diff --git a/net/core/skbuff.c b/net/core/skbuff.c
>>>>> index 8a725cc..d08edcb 100644
>>>>> --- a/net/core/skbuff.c
>>>>> +++ b/net/core/skbuff.c
>>>>> @@ -4133,6 +4133,9 @@ bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
>>>>>    	if (skb_cloned(to))
>>>>>    		return false;
>>>>>
>>>>> +	if (skb_has_frag_list(to) || skb_has_frag_list(from))
>>>>> +		return false;
>>>>> +
>>>>>    	if (len <= skb_tailroom(to)) {
>>>>>    		if (len)
>>>>>    			BUG_ON(skb_copy_bits(from, 0, skb_put(to, len), len));
>>>>> @@ -4140,9 +4143,6 @@ bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
>>>>>    		return true;
>>>>>    	}
>>>>>
>>>>> -	if (skb_has_frag_list(to) || skb_has_frag_list(from))
>>>>> -		return false;
>>>>> -
>>>>>    	if (skb_headlen(from) != 0) {
>>>>>    		struct page *page;
>>>>>    		unsigned int offset;
>>>>
>>>> Sigh.
>>>>
>>>> No idea what problem you tried to solve.
>>>>
>>>> This patch is not needed.
>>>>
>>>> If (len <= skb_tailroom()), then it is obviously correct to copy_bits()
>>>> the bytes.
>>>>
>>>> Hints :
>>>>
>>>> - If @to has a fraglist, then skb_tailroom(to) is 0 so the copy can not
>>>> be done.
>>
>> How to make sure it?
>> In my test, @to has a fraglist, but skb_tailroom(to) is not 0!
>> The test is about tipc, the function tipc_buf_append will merge 3 skbs
>> to one:
>> packet 1: len = 1420 skb_tailroom = 190
>> packet 2: len = 1420
>> 	packet 2 will be add to 1' fraglist, but not update skb_tailroom
>> packet 3: len = 60
>> 	Here's the error!
>
> Then fix tipc, or make sure you read and understood the code.
>
> By definition, if one skb has something in fraglist, it is non linear.
>
> If you read skb_tailroom(), you'll see this function returns 0 if skb is
> non linear.

Yes you'er right, when packet 2 is add to 1' fraglist, tipc should

update the packet 1' 'len' and 'data_len'. In my kernel(v3.10), it

doesn't do that, which was corrected in commit 5074a (v3.16).

Thanks for your patience

>
> include/linux/skbuff.h-/**
> include/linux/skbuff.h: *       skb_tailroom - bytes at buffer end
> include/linux/skbuff.h- *       @skb: buffer to check
> include/linux/skbuff.h- *
> include/linux/skbuff.h- *       Return the number of bytes of free space at the tail of an sk_buff
> include/linux/skbuff.h- */
> include/linux/skbuff.h:static inline int skb_tailroom(const struct sk_buff *skb)
> include/linux/skbuff.h-{
> include/linux/skbuff.h- return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
> include/linux/skbuff.h-}
>
>
>
>
> .
>