[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20181128084909epcms1p64e84e47e798fcf949f58838d9aabab14@epcms1p6>
Date: Wed, 28 Nov 2018 17:49:09 +0900
From: 배석진 <soukjin.bae@...sung.com>
To: Lorenzo Colitti <lorenzo@...gle.com>
CC: Eric Dumazet <eric.dumazet@...il.com>,
"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
유금환 <geumhwan.yu@...sung.com>,
Maciej Żenczykowski <zenczykowski@...il.com>,
배석진 <soukjin.bae@...sung.com>
Subject: RE:(2) (2) FW: [Resource Leak] Suggesting patch for tcp_close
> What harm is caused by these stale sessions? I thought that was the
> intended behaviour.
>
our system stability guys concern about this.
when its count grows up too much, whether it can be harm to system or not.
> If you look at the original design discussions that led to the
> SOCK_DESTROY and tcp_abort patch, the goal of SOCK_DESTROY was not
> primarily to clear TCP state, but primarily to unblock applications
> that were stuck in blocking socket operations such as read(), write()
> and connect. That is the reason why it only calls tcp_done if the
> SOCK_DEAD flag is not set. See in particular
> https://www.spinics.net/lists/netdev/msg352716.html , where opposition
> was voiced to being able to close sockets in TIME_WAIT_STATE. That
> said, I don't have a strong opinion on this: whatever works for Eric
> works for me.
>
ok, i got your saying about purpose of SOCK_DESTROY.
maybe we have two options.
add a new point that can managing those tcp sessions.
or, just let it them.
thanks!
Powered by blists - more mailing lists