Bug#777735: systemd: Disk corruption on reboot initiated through systemd

[Peter Colberg]

On Fri, Feb 13, 2015 at 08:02:09AM +0100, Michael Biebl wrote:
> Am 13.02.2015 um 07:44 schrieb Michael Biebl:
> > I went on and looked what the watchdog package does. Interestingly, it
> > clamps any timeout value to 254s, thus not hitting this issue.
> > Apparently, i6300esb is able to deal with such a timeout.
> > 
> > Peter, I therefore recommend you set
> > 
> >  ShutdownWatchdogSec=254s
> > 
> > in /etc/systemd/system.conf as well for the time being. This should be
> > sufficient to workaround this issue for this buggy driver.
> > Setting 256s, does indeed trigger the issue, which leads to the
> > conclusion, that the driver is using a char to hold a time value (0-255).
> 
> Actually, after a few more experiments, it turns out that for i6300esb,
> the safe max value in my qemu VM is 549s. Setting it to 550s triggers
> the reset.

In the linux-3.16.7-ckt4 source the range of the timeout value for
the i6300esb is 1s < timeout < 2*1023s. In theory it should work
with the default shutdown timeout of 10min.

For the ib700 the range is 0s <= timeout <= 30s, hence the error.

Peter