netdev - Re: (2) FW: [Resource Leak] Suggesting patch for tcp

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] [day] [month] [year] [list]

Message-ID: <d54f7e53-cb84-b4f6-f16e-74ca64bc2796@gmail.com>
Date:   Wed, 28 Nov 2018 06:40:40 -0800
From:   Eric Dumazet <eric.dumazet@...il.com>
To:     soukjin.bae@...sung.com, Eric Dumazet <eric.dumazet@...il.com>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        "lorenzo@...gle.com" <lorenzo@...gle.com>
Cc:     유금환 <geumhwan.yu@...sung.com>
Subject: Re: (2) FW: [Resource Leak] Suggesting patch for tcp_close



On 11/27/2018 10:17 PM, 배석진 wrote:
>>> we saw hundreds of not closed tcp session with FIN_WAIT1 and LAST_ACK.
>>
>> These sessions should have a timer, and eventually disappear.
> 
> FIN_WAIT2 and TIME_WAIT have a timer.
> but FIN_WAIT1 and LAST_ACK are have too?

Sure. Otherwise we have a more serious bug than tcp_abort().


> 
>> Do you have a test to demonstrate the issue ?
>>  
>> I know Lorenzo wrote tests, so presumably new tests are needed.
> 
> yes, i had a test myself, and attached the result.
> step is easy, with android device.
> 
> 1. remove simcard (no mobile online)
> 2. connect wifi and check tcp sessions
> 3. disconnect wifi and check again.
> 4. repeat step 2~3
> 

A real test is something that can be automated.

Like a packetdrill test.

> as i said, gms make this problem often at port 443 or 5228. 
> i've confirm session owner is them by pid, when session was ESTABLISH state.
> and i found this at speedguide.net
> 
> - Port(s)  : 5228
> - Protocol : tcp,udp
> - Service  : android
> - Details  : Port 5228 is used by the Google Playstore (Android market). 
>              Google talk also uses ports 443, 5222 and 5228. 
>              Google Chrome user settings sync (facorites, history, passwords) uses port 5228. 
> - Related ports : 443 5222
> 
> 
> below is report we have received.
> i omitted full list shortly. its list up to 900 line.
> 
> ------ NETSTAT (netstat -npWae) ------
> Active Internet connections (established and servers)
> Proto Recv-Q Send-Q Local Address               Foreign Address             State       User  Inode   PID/Program Name
> tcp   0      0      127.0.0.1:5037              0.0.0.0:*                   LISTEN      0     617046  2768/adbd
> tcp   0      1      192.168.0.112:53828         211.231.100.211:443         FIN_WAIT1   0     0       -
> tcp   0      1      192.168.0.112:33908         27.0.236.141:443            FIN_WAIT1   0     0       -
> tcp   0      1      192.168.0.112:58480         183.111.43.14:443           LAST_ACK    0     0       -
> tcp   0      1      192.168.0.112:56558         203.133.167.207:443         FIN_WAIT1   0     0       -
> ...                                                                                                  
> tcp6  0     32      ::ffff:192.168.0.112:50072  ::ffff:74.125.203.188:5228  FIN_WAIT1   0     0       -
> tcp6  0      1      ::ffff:192.168.1.5:59744    ::ffff:216.58.197.238:443   FIN_WAIT1   0     0       -
> tcp6  0     32      ::ffff:192.168.0.112:40512  ::ffff:108.177.125.188:5228 FIN_WAIT1   0     0       -
> tcp6  0     32      ::ffff:192.168.0.112:46560  ::ffff:108.177.97.188:5228  FIN_WAIT1   0     0       -
> 
> 
>> BTW your patch title is confusing.
>> You are changing tcp_abort(), not tcp_close()
> 
> yes right, i meant tcp session close as title.
> but my finger put the underbar himself :o
>