netdev - Re: orphan sockets

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]

Date:	Fri, 8 Jul 2011 19:28:27 +0300
From:	Daniel Baluta <daniel.baluta@...il.com>
To:	Denys Fedoryshchenko <denys@...p.net.lb>
Cc:	netdev@...r.kernel.org
Subject: Re: orphan sockets

On Fri, Jul 8, 2011 at 6:13 PM, Denys Fedoryshchenko <denys@...p.net.lb> wrote:
> Hi
>
> Just after digging orphan sockets problem, that was causing "Address already
> in use" for bind(),i found that in documentation is mentioned:
>
> tcp_orphan_retries - INTEGER
>        This value influences the timeout of a locally closed TCP connection,
>        when RTO retransmissions remain unacknowledged.
>        See tcp_retries2 for more details.
>
>        The default value is 7.
>        If your machine is a loaded WEB server,
>        you should think about lowering this value, such sockets
>        may consume significant resources. Cf. tcp_max_orphans.

This should be 8. tcp(7) got it wright.

> But all servers i have, i notice that  tcp_orphan_retries = 0 by default, i
> check in code, and found that:
>        if (retries == 0 && alive)
>                retries = 8;
> is a bit confusing, that tcp_orphan_retries = 0, is in fact = 8, or to be
> more exact is 8 if socket has rto < RTO_MAX. But nothing like 7 mentioned in
> documentation, i guess it is wrong?. My english too bad to edit that, but
> maybe someone will take a look :-)

I guess 0 means, let the operating system to choose the default value for this.
Basically, what we should do here is replace 7 with 8 in
Documentation/networking/ip-sysctl.txt
I can try to make a patch for this if you want.

thanks,
Daniel.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html