netdev - Re: TCP slowdown in 2.6.25?

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: <1209466284.6904.36.camel@gentoo-jocke.transmode.se>
Date:	Tue, 29 Apr 2008 12:51:24 +0200
From:	Joakim Tjernlund <joakim.tjernlund@...nsmode.se>
To:	Ilpo Järvinen <ilpo.jarvinen@...sinki.fi>
Cc:	Netdev <netdev@...r.kernel.org>
Subject: Re: TCP slowdown in 2.6.25?


On Mon, 2008-04-28 at 18:48 +0300, Ilpo Järvinen wrote:
> On Mon, 28 Apr 2008, Joakim Tjernlund wrote:
> 
> > On Mon, 2008-04-28 at 17:42 +0300, Ilpo Järvinen wrote:
> > > On Mon, 28 Apr 2008, Joakim Tjernlund wrote:
> > > > On Mon, 2008-04-28 at 14:12 +0200, Joakim Tjernlund wrote:
> > > > > After upgrading from 2.6.23 to 2.6.25 I am seeing a big slow down on
> > > > > our old www server. Stracing shows lots of EAGAIN but i not sure
> > > > > these are relevant. I got no idea where to start looking so
> > > > > I would appreciate some pointers. Perhaps this is a know issue?
> > > > > 
> > > > > CPU is a PowerPC, mpc8321, network driver ucc_geth
> > > > 
> > > > Some info:
> > > > Even ftp transfers takes forever, but only in one direction: from the
> > > > node with kernel 2.6.25 to some other host. Strace show me that
> > > > write(2) hangs for long periods before succeeding to transmit 
> > > > the data.
> 
> > > To some specific or "any" other host? How long is "long"?
> > 
> > Any host. A 1304632 bytes file takes about 110 seconds to transfer
> > (12 Kbytes/sec)
> >
> > A quick capture with wireshark indicates that some sent packages from
> > the node never reaches the host on the network.
> 
> Thanks, this is the important bit of info, TCP flow control limit seems 
> just a consequence of that... ...Now we'll just have to figure out what is 
> causing those losses.
> 
> Is there some middlebox on the route to that host?

Problem solved, was a bug in the ucc_geth driver. See the patch I just
sent. Thanks for your help.

 Jocke
--
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