From mboxrd@z Thu Jan 1 00:00:00 1970 From: Hannes Frederic Sowa Subject: Re: Linux 3.6.x kernel crash in tcp_slow_start / bictcp_cong_avoid with wfica Date: Wed, 23 Jan 2013 23:00:39 +0100 Message-ID: <20130123220039.GB16641@order.stressinduktion.org> References: <20130123161238.GE8912@reaktio.net> <20130123214445.GA16641@order.stressinduktion.org> <20130123215151.GF8912@reaktio.net> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: netdev@vger.kernel.org To: Pasi =?utf-8?B?S8Okcmtrw6RpbmVu?= Return-path: Received: from order.stressinduktion.org ([87.106.68.36]:47291 "EHLO order.stressinduktion.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752273Ab3AWWAk (ORCPT ); Wed, 23 Jan 2013 17:00:40 -0500 Content-Disposition: inline In-Reply-To: <20130123215151.GF8912@reaktio.net> Sender: netdev-owner@vger.kernel.org List-ID: On Wed, Jan 23, 2013 at 11:51:51PM +0200, Pasi K=C3=A4rkk=C3=A4inen wro= te: > On Wed, Jan 23, 2013 at 10:44:45PM +0100, Hannes Frederic Sowa wrote: > > On Wed, Jan 23, 2013 at 06:12:38PM +0200, Pasi K=C3=A4rkk=C3=A4inen= wrote: > > > My laptop running Fedora 17 randomly crashes when using wfica (Ci= trix Receiver / ICA Client)=20 > > > proprietary closed source client for accessing remote desktop ser= ver.=20 > > > wfica is running as a normal user, so it shouldn't be able to cau= se a kernel crash.=20 > > >=20 > > > Crashes seem to happen most often when I use either wlan- or 3G m= obile data Internet connection.=20 > > > These kernel crashes happen ramdomly, often 1-2 times a week.=20 > > > What usually happens is that I notice the Internet connection has= died,=20 > > > and when I check the kernel dmesg I'm seeing the traceback.=20 > > > At this point I'm still able to use the gnome desktop for maybe 1= -5 minutes,=20 > > > but applications start to fail one by one, and finally everything= just halts=20 > > > and I need to power cycle the laptop.=20 > >=20 > > Would it be possible to capture a strace -e network, so we could se= e which > > setsockopt combinations are being applied to the socket. Is capturi= ng ss -ioem > > in such moments still possible? > >=20 >=20 > I don't think I'm able to strace or anything at that point..=20 > It took me at least 10 tries (=3Dcrashes) before I was able to "dmesg= > log.txt" successfully ..=20 > (usually writing/redirecting to a file fails because the disk/fs is a= lready unavailable..) The strace can be done without the system going downhill. Instead of -e network, "-e socket,connect,setsockopt,getsockopt,shutdown,close" can be used to reduce the output.