IDE write ordering ?

Lamar Owen lowen at pari.edu
Tue Dec 2 18:31:12 UTC 2003


On Monday 01 December 2003 06:36 pm, Alan Cox wrote:
> On Mon, Dec 01, 2003 at 09:33:17AM +0100, David Balazic wrote:
> > So my question is , how can one test if the journalled FS on his disk is
> > actually working
> > as it should or is it just waste of time and giving a false sense of
> > safety

> Its extremely hard to do because you need fairly complex tools to ensure
> outstanding I/O during a sudden power off and out of order drive writeback
> etc

What you need is a way of reading the AC mains coming into the power supply, 
or readings of the DC rails.  If the rail drops below a threshold value, it's 
time to button up.  With typical power supply output capacitors and/or low 
loading on the rails, you might have up to half a second to react.  One 
wishes that the power supply could send a 'I'm about to croak!' signal to the 
motherboard that the kernel could key off of, and it would be even nicer if 
the PS had a flywheel capacitor set on all outputs before crowbarring (the 
voltages have to be sequenced during powerfail as well as powerup).

I remember the intelligence behind the old Seagate ST251 stepper motor 
actuator MFM hard drive.  It auto parked at poweroff, by using the spindle 
motor as a generator, driving the stepper to park.  With servos that's not a 
problem due to the spring tension that is a normal part of the voice coil 
mechanism.  That couldn't be done with a stepper.

So it can be done if there is hardware support.  If the MB allows monitoring 
of the DC rails, one could do such, no?
-- 
Lamar Owen
Director of Information Technology
Pisgah Astronomical Research Institute
1 PARI Drive
Rosman, NC  28772
(828)862-5554
www.pari.edu





More information about the fedora-test-list mailing list