From mboxrd@z Thu Jan 1 00:00:00 1970 From: =?ISO-8859-1?Q?BERTRAND_Jo=EBl?= Date: Sat, 10 Mar 2007 08:38:20 +0000 Subject: Re: Dead lock with 2.6.20 Message-Id: <45F26E7C.2040801@systella.fr> List-Id: References: <45F1C172.60103@systella.fr> In-Reply-To: <45F1C172.60103@systella.fr> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable To: sparclinux@vger.kernel.org Daniel Smolik a =E9crit : > BERTRAND Jo=EBl napsal(a): >> Daniel Smolik a =E9crit : >> >>> BERTRAND Jo=EBl napsal(a): >>> >>>> Hello, >>>> >>>> I have seen a very strange deadlock on a UltraSPARC 2=20 >>>> workstation (2 UltraSPARC-II/296 MHz, 2048 MB, framebuffer Creator3D=20 >>>> on UPA slot, 2 73 GB SCA disks that form a raid1 array). >>>> >>>> Hardware sounds good and I have tested it with SUNvts during two=20 >>>> days without any failure. >>>> >>>> This workstation is a mail server and runs with Debian/testing,=20 >>>> kernel 2.6.20. It's a NIS and NFSv3/TCP clients. Its uptime was more=20 >>>> than 9 days and it hangs without any message in the log, without=20 >>>> panic and without oops. When it hangs, the queue of sendmail was=20 >>>> about 3 GB due to a failure of a internet peering. I have rebooted=20 >>>> the workstation (power button :-( ). You can find my .config at >>>> >>>> http://rayleigh.systella.fr/~bertrand/config-sun4u-2.6.20.gz >>>> >>>> I don't know how debug this bug... >>> >>> Hi, >>> do you have attached console ? >> >> >> No... I don't have any console. >> >>> Do you try STOP-A ? >> >> >> Stop+A doesn't help. I have tried, of course ;-) >> >>> I have similar problem on E450 tomorrow. After each second reboot=20 >>> system hang and I can't STOP-A. Only power-off helps. System hang=20 >>> after udev insert sg module. I disable udev and this solve my problem. >> >> >> Do you think that /etc/init.d/udev stop can solve this trouble ? > I turn off udev now and it helps. Test if have or not inserted sg module. Hello, Without sg module and without udev, 2.6.20 hangs with a dead lock on=20 heavy disk IO (maybe a bug in sbus or esp support because I have a=20 U60/SMP that run with the same configuration without any trouble). I=20 have tried a 2.6.18, same bug. Regards, JKB