[OLPC-AU] Critical issue - testing required

[Sridhar Dhanapalan]

Hey all,

We at OLPC Australia have come across a critical issue, and we'd
really appreciate some assistance from the community in testing the
fix.

It seems that some XOs suspended and placed in charging racks are
waking each other up. This has led to overheating in some units - to
the point where the plastic warps and the screens become damaged.

OLPC have come up with a solution - a new version of olpc-powerd:

  http://dev.laptop.org/~pgf/rpms/olpc-powerd-29-1.fc11.i586.rpm
  md5sum 285753a1a54999b7dc58508020b1f184

It has been tested on an XO-1.5, but it has not had wider community testing.

We would be very grateful if we could have some help in testing it on
two different builds:

  * on our latest XO-AU build (most important for us)
  * on OLPC's os852 (helpful for OLPC)

XO-AU is based on OLPC's os85x, and is what we intend to use in
Australian schools. For more information:

  http://lists.laptop.org/pipermail/olpc-au/2010-October/000087.html
  http://lists.laptop.org/pipermail/olpc-au/2010-October/000101.html

Feel free to ask me if you require further information to facilitate testing.


Now for some further detail on the problem:

---------- Forwarded message ----------
From: James Cameron
Date: 14 October 2010 08:55


G'day Sridhar,

Thank you for your problem report.  We have discovered a potential cause
of the problem.  Here is a summary followed by a detailed write-up.

Summary.

Laptops may resume from suspend when placed in the rack, and therefore
they should be turned off before being placed in the rack.  We plan to
issue a software change that will prevent this.  Heat is generated
during charging, and combined with the heat of operation, with the
laptop insulated by adjacent laptops may cause the case to melt.

Details.

1.  Magnet

The XO-1 and XO-1.5 laptops have a magnet under the front left, about
20mm from the left edge, 35mm from the front edge, and about 93mm from
the centre of the touchpad.  You can often locate it using a USB key;
dragging the metal plug in that corner.  You can feel that it has two
attracting lobes; the North and South poles are in a horizontal axis.
The field is weaker on the underside of the laptop, but not absent.

2.  Sensors

This magnet is sensed by two locations on the main board; behind the
microphone, and behind the camera.  The first sensor detects lid
closure.  The second sensor detects ebook mode.  The sensors detect
field strength and polarity.

The magnet has to be strong enough to reach through to the main board in
those two positions, over several millimetres.

However, this makes it strong enough to occasionally cancel the field of
a magnet from an immediately adjacent laptop.  This cancelling may occur
during movement, or once the laptops are in position.

3.  Software

The software that manages suspend and resume in response to lid close
and open can miss a lid close event if it occurs within a second or so
of a lid close event.  You can test this by closing the lid, wait for
the power LED to go out, then open the latches, lift the lid a few
millimetres, and close it quickly within about a second.  The laptop
may wake and stay awake.

4.  Scenario

Here is what may have happened.

The child closed the laptop.  It suspended.  It was inserted in
the rack.  As it passed close to another laptop, it woke, or the other
laptop woke, or both.

Charging and operation of the laptop continued.  Cooling of the back
face of the laptop would have been hindered by the presence of the other
laptop.  The small amount of airflow would have caused the area near the
hinge, near the bottom of the screen, and either side to become hot.

The heat would have built until the battery sensor reaches 55C, after
which charging would be stopped, but the laptop CPU would continue to be
active.  Heat would have built up more slowly, and the CPU would have
throttled at 100C.

5.  Workarounds

The level 1 workaround we suggest is that the children should turn off
the laptop before putting them in the rack.  The teacher should inspect
the laptops in the rack, and make sure the power LED is off, and not
blinking.

It isn't enough to check only for power LED on, since the contents of
the rack may move slightly during charging, due to thermal expansion,
changing the alignment of the magnetic fields and triggering the
sensors.  The teacher should check for a blinking power LED.

The level 2 workaround we are scheduling is a change to require power
button press to wake up the laptop, ignoring the lid open event.  We
will supply this as a software patch.





Sridhar Dhanapalan
Technical Manager
One Laptop per Child Australia
M: +61 425 239 701
E: sridhar at laptop.org.au
A: G.P.O. Box 731
     Sydney, NSW 2001
W: www.laptop.org.au

Be Part of the Equation....

Show your support for OLPC Australia by texting the word LAPTOP to 044
SUPPORT (0447 877 678)*
For every 100 messages of support received, Telstra will donate an
additional XO laptop to a remote child - up to 500 XOs**

*Standard SMS charges apply
**Full terms and conditions at Ideas for Good