mptcp.lists.linux.dev archive mirror
 help / color / mirror / Atom feed
From: Jianguo Wu <wujianguo106@163.com>
To: Paolo Abeni <pabeni@redhat.com>,
	Mat Martineau <mathew.j.martineau@linux.intel.com>
Cc: mptcp@lists.linux.dev, fw@strlen.de
Subject: Re: [PATCH v5 4/4] mptcp: avoid processing packet if a subflow reset
Date: Fri, 25 Jun 2021 08:51:42 +0800	[thread overview]
Message-ID: <e5de2c05-ba90-1045-183f-0beb813059fb@163.com> (raw)
In-Reply-To: <7c26d28d7249536615bf945b09d683f81eb06be0.camel@redhat.com>



On 2021/6/24 18:02, Paolo Abeni wrote:
> On Thu, 2021-06-24 at 09:57 +0800, Jianguo Wu wrote:
>>
>> On 2021/6/23 17:48, Paolo Abeni wrote:
>>> On Tue, 2021-06-22 at 17:00 -0700, Mat Martineau wrote:
>>>> On Mon, 21 Jun 2021, Jianguo Wu wrote:
>>>>
>>>>> Hi Mat,
>>>>>
>>>>> On 2021/6/19 8:19, Mat Martineau wrote:
>>>>>> On Wed, 16 Jun 2021, wujianguo106@163.com wrote:
>>>>>>
>>>>>>> From: Jianguo Wu <wujianguo@chinatelecom.cn>
>>>>>>>
>>>>>>> If check_fully_established() causes a subflow reset, it should not
>>>>>>> continue to process the packet in tcp_data_queue().
>>>>>>>
>>>>>>> setting:
>>>>>>>     TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq;
>>>>>>>
>>>>>>> so that the following check will drop the pkt in
>>>>>>> tcp_data_queue():
>>>>>>>  if (TCP_SKB_CB(skb)->seq == TCP_SKB_CB(skb)->end_seq) {
>>>>>>>     __kfree_skb(skb);
>>>>>>>     return;
>>>>>>>  }
>>>>>>>
>>>>>>> Fixes: d582484726c4 ("mptcp: fix fallback for MP_JOIN subflows")
>>>>>>> Signed-off-by: Jianguo Wu <wujianguo@chinatelecom.cn>
>>>>>>> ---
>>>>>>> net/mptcp/options.c | 6 ++++++
>>>>>>> 1 file changed, 6 insertions(+)
>>>>>>>
>>>>>>> diff --git a/net/mptcp/options.c b/net/mptcp/options.c
>>>>>>> index 1aec01686c1a..be435c5421cd 100644
>>>>>>> --- a/net/mptcp/options.c
>>>>>>> +++ b/net/mptcp/options.c
>>>>>>> @@ -926,6 +926,12 @@ static bool check_fully_established(struct mptcp_sock *msk, struct sock *ssk,
>>>>>>>     return true;
>>>>>>>
>>>>>>> reset:
>>>>>>> +    /* If a subflow is reset, the packet should not continue to be
>>>>>>> +     * processed in tcp_data_queue(), so setting: end_seq = seq,
>>>>>>> +     * then tcp_data_queue() will drop the packet.
>>>>>>> +     */
>>>>>>> +    TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq;
>>>>>>> +
>>>>>>
>>>>>> This does have the desired effect when mptcp_incoming_options() is 
>>>>>> called from tcp_data_queue(), but mptcp_incoming_options() is also 
>>>>>> called from tcp_reset() and tcp_rcv_state_process(). The other callers 
>>>>>> appear to tolerate the sequence number modification.
>>>>>>
>>>>>> I think it would be clearer to either add a return value or output 
>>>>>> parameter to mptcp_incoming_options() to explicitly tell the caller 
>>>>>> that a reset has been sent and tcp_done() called. Then it would be 
>>>>>> clearer in tcp_data_queue() that the packet is being discarded due to 
>>>>>> mptcp header content.
>>>>>>
>>>>>
>>>>> If a reset has been sent and tcp_done() called in 
>>>>> check_fully_established(), the sk_state will be TCP_CLOSE, how about 
>>>>> just do (sk_state == TCP_CLOSE) check in tcp_data_queue() as it did in 
>>>>> the V1 of this patch?
>>>>
>>>> Oh, I see now that Paolo suggested the the end_seq assignment in order to 
>>>> only modify MPTCP code.
>>>>
>>>> I still think it's better to make it clear that we're discarding a packet 
>>>> due to the mptcp headers - using the existing sequence check (intended to 
>>>> detect acks) in tcp_data_queue() seems sneaky to me.
>>>>
>>>> Something like
>>>>
>>>> if (sk_is_mptcp(sk) && !mptcp_incoming_options(sk, skb)) {
>>>>  	__kfree_skb(skb);
>>>>  	return;
>>>> }
>>>>
>>>> seems both compact and clear. Does that seem ok Paolo?
>>>
>>> Uhmmm... we need to touch every mptcp_incoming_options() call site, and
>>> in tcp_reset() the above chunk looks a bit strange to me. Probably we
>>> could just ignore the mptcp_incoming_options() return value there. 
>>>
>>> Otherwise no big objections - not sure about upstream ;)
>>>
>>
>> Hi Mat and Paolo,
>>
>> If you both agree, I will send a new version that mptcp_incoming_options() add a return value, and only check return value in tcp_data_queue(),
>> ignore the return value in other call site.
> 
> Even the hook in tcp_rcv_state_process() can be followed by
> tcp_data_queue().
> 
> I *think* it's better ignoring the return value of
> mptcp_incoming_options() only in tcp_reset(), adding there a comment -
> something alike "mptcp can't tell us to ignore reset pkts".
> 
> Cheers,
> 
> Paolo
> 
> p.s. I'm sorry for the long, difficult and somewhat on/off review
> process. This change is indeed tricky. Don't despair, it looks like
> it's near to an happy ending!
> 

Thanks for your review and suggestions! I will send a new version.

> 


  parent reply	other threads:[~2021-06-25  1:07 UTC|newest]

Thread overview: 24+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-06-16 10:49 [PATCH v5 0/4] Fix some mptcp syncookie process bugs wujianguo106
2021-06-16 10:49 ` [PATCH v5 1/4] mptcp: fix warning in __skb_flow_dissect() when do syn cookie for subflow join wujianguo106
2021-06-18 22:40   ` Mat Martineau
2021-06-21  6:14     ` Jianguo Wu
2021-06-21 10:09       ` Jianguo Wu
2021-06-22 23:38         ` Mat Martineau
2021-06-16 10:49 ` [PATCH v5 2/4] mptcp: remove redundant req destruct in subflow_check_req() wujianguo106
2021-06-16 10:49 ` [PATCH v5 3/4] mptcp: fix syncookie process if mptcp can not_accept new subflow wujianguo106
2021-06-18 23:19   ` Mat Martineau
2021-06-21  6:24     ` Jianguo Wu
2021-06-24  2:08       ` Jianguo Wu
2021-06-24 22:36         ` Mat Martineau
2021-06-16 10:49 ` [PATCH v5 4/4] mptcp: avoid processing packet if a subflow reset wujianguo106
2021-06-19  0:19   ` Mat Martineau
2021-06-21  6:35     ` Jianguo Wu
2021-06-23  0:00       ` Mat Martineau
2021-06-23  9:48         ` Paolo Abeni
2021-06-24  1:57           ` Jianguo Wu
2021-06-24 10:02             ` Paolo Abeni
2021-06-24 22:38               ` Mat Martineau
2021-06-25  0:51               ` Jianguo Wu [this message]
2021-06-19  0:26 ` [PATCH v5 0/4] Fix some mptcp syncookie process bugs Mat Martineau
2021-06-21  6:39   ` Jianguo Wu
2021-06-23  0:07     ` Mat Martineau

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=e5de2c05-ba90-1045-183f-0beb813059fb@163.com \
    --to=wujianguo106@163.com \
    --cc=fw@strlen.de \
    --cc=mathew.j.martineau@linux.intel.com \
    --cc=mptcp@lists.linux.dev \
    --cc=pabeni@redhat.com \
    --subject='Re: [PATCH v5 4/4] mptcp: avoid processing packet if a subflow reset' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).