netdev - Re: [PATCH net-next v2] mptcp: be careful on MPTCP-level ack.

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <d05ac8b9-3522-e4fc-d3ce-4bea74a6dfbf@gmail.com>
Date:   Wed, 2 Dec 2020 17:30:14 +0100
From:   Eric Dumazet <eric.dumazet@...il.com>
To:     Paolo Abeni <pabeni@...hat.com>, netdev@...r.kernel.org
Cc:     Jakub Kicinski <kuba@...nel.org>, mptcp@...ts.01.org
Subject: Re: [PATCH net-next v2] mptcp: be careful on MPTCP-level ack.



On 12/2/20 5:10 PM, Eric Dumazet wrote:
> 
> 
> On 12/2/20 4:37 PM, Paolo Abeni wrote:
>> On Wed, 2020-12-02 at 14:18 +0100, Eric Dumazet wrote:
>>>
>>> On 11/24/20 10:51 PM, Paolo Abeni wrote:
>>>> We can enter the main mptcp_recvmsg() loop even when
>>>> no subflows are connected. As note by Eric, that would
>>>> result in a divide by zero oops on ack generation.
>>>>
>>>> Address the issue by checking the subflow status before
>>>> sending the ack.
>>>>
>>>> Additionally protect mptcp_recvmsg() against invocation
>>>> with weird socket states.
>>>>
>>>> v1 -> v2:
>>>>  - removed unneeded inline keyword - Jakub
>>>>
>>>> Reported-and-suggested-by: Eric Dumazet <eric.dumazet@...il.com>
>>>> Fixes: ea4ca586b16f ("mptcp: refine MPTCP-level ack scheduling")
>>>> Signed-off-by: Paolo Abeni <pabeni@...hat.com>
>>>> ---
>>>>  net/mptcp/protocol.c | 67 ++++++++++++++++++++++++++++++++------------
>>>>  1 file changed, 49 insertions(+), 18 deletions(-)
>>>>
>>>
>>> Looking at mptcp recvmsg(), it seems that a read(fd, ..., 0) will
>>> trigger an infinite loop if there is available data in receive queue ?
>>
>> Thank you for looking into this!
>>
>> I can't reproduce the issue with the following packetdrill ?!?
>>
>> +0.0  connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
>> +0.1   > S 0:0(0) <mss 1460,sackOK,TS val 100 ecr 0,nop,wscale 8,mpcapable v1 fflags[flag_h] nokey>
>> +0.1   < S. 0:0(0) ack 1 win 65535 <mss 1460,sackOK,TS val 700 ecr 100,nop,wscaale 8,mpcapable v1 flags[flag_h] key[skey=2] >
>> +0.1  > . 1:1(0) ack 1 <nop, nop, TS val 100 ecr 700,mpcapable v1 flags[flag_h]] key[ckey,skey]>
>> +0.1 fcntl(3, F_SETFL, O_RDWR) = 0
>> +0.1   < .  1:201(200) ack 1 win 225 <dss dack8=1 dsn8=1 ssn=1 dll=200 nocs,  nop, nop>
>> +0.1   > .  1:1(0) ack 201 <nop, nop, TS val 100 ecr 700, dss dack8=201 dll=00 nocs>
>> +0.1 read(3, ..., 0) = 0
>>
>> The main recvmsg() loop is interrupted by the following check:
>>
>>                 if (copied >= target)
>>                         break;
> 
> @copied should be 0, and @target should be 1
> 
> Are you sure the above condition is triggering ?
> 
> Maybe read(fd, ..., 0) does not reach recvmsg() at all.

Yes, sock_read_iter() has a shortcut :

if (!iov_iter_count(to))    /* Match SYS5 behaviour */
     res = sock_recvmsg(sock, &msg, msg.msg_flags);

but recvmsg() does not have such check, or maybe I have not looked at the right place.

> 
> You could try recvmsg() or recvmmsg(), 
> 
>>
>> I guess we could loop while the msk has available rcv space and some
>> subflow is feeding new data. If so, I think moving:
>>
>> 	if (skb_queue_empty(&msk->receive_queue) &&
>>                     __mptcp_move_skbs(msk, len - copied))
>>                         continue;
>>
>> after the above check should address the issue, and will make the
>> common case faster. Let me test the above - unless I underlooked
>> something relevant!
>>
>> Thanks,
>>
>> Paolo
>>