#10680 HIGH Gen2: XO-1.75 A2 has CPU faults when simultaneously using WLAN and SD

[Zarro Boogs per Child]

#10680: XO-1.75 A2 has CPU faults when simultaneously using WLAN and SD
---------------------------+------------------------------------------------
 Reporter:  wad            |                 Owner:  wad     
     Type:  defect         |                Status:  new     
 Priority:  high           |             Milestone:  Gen2    
Component:  hardware       |               Version:  1.75-A2 
 Keywords:  XO-1.75, XO-3  |           Next_action:  diagnose
 Verified:  0              |   Deployment_affected:          
Blockedby:                 |              Blocking:  3       
---------------------------+------------------------------------------------
 Some XO-1.75 A2 motherboards that otherwise work perfectly well (running
 memtester and SD read/write tests simultaneously for days) will show CPU
 faults when running the WLAN while simultaneously exercising memory and
 SD.

 These boards will run WLAN tests for long periods of time without CPU
 faults in the absence of heavy memory and SD activity.

 Interestingly, the CPU faults are all in processes associated with the
 network activity.  memtester and the SD tests continue operating fine, but
 either Python, Sugar, or X will crash with a CPU fault (unimplemented
 instruction, segmentation fault).

 Examples of motherboards showing this problem are boards 2, 10, and 17.
 These boards are operating at Vmain voltages of 1.25 and 1.35V.  They have
 EC firmware 4.0.0.6, OFW Q4A13, and OS3 installed.

 A laptop is prepared for the tests using the following steps.  Note that
 many of the laptops around 1CC have had steps 1 and 2 done.

 1) Install the memory test program, memtester.  An appropriate rpm for OS3
 may be found at:
 http://dev.laptop.org/~wad/memtester-4.1.2-1.fc12.armv5tel.rpm

 2) Install the SD wear leveling test program.  This is a provided as a
 directory which needs to be installed as /nand, available from:
 http://dev.laptop.org/~wad/sd_test.zip
 You will need to create a /logs/nand directory for the logs resulting from
 this test.

 3) Enable access to sugar by providing the olpc user with a password
 (using the passwd command).  The next time you try to log into the display
 on the laptop (as olpc), you will be prompted for a password.  Before
 entering it, select Sugar (instead of Gnome) on the tiny popup menu at the
 bottom of the screen.  This only has to be done once.

 4) Log into sugar, and insert an SD card or USB stick containing the
 ShowPhotos-5.activity (available at
 http://dev.laptop.org/~wad/ShowPhotos-5.xo ).  Select the SD card (from
 the Frame) and click on the ShowPhotos activity to launch it.  Once
 launched, you can remove the USB/SD device.

 You can stop the ShowPhotos activity from the Pippy Activity symbol (a
 snake in a diamond) in the frame, either by stopping it from the pull-down
 menu from the frame, or by selecting it then pressing the stop icon.   I
 make no claims as to the suitability of the ShowPhotos activity for
 anything, but it does sit there generating network activity without using
 the SD subsystem.

 To run the tests:

 A) Boot the laptop all the way into Sugar, and start the ShowPhotos
 activity

 I do the following two in individual virtual console windows, but you
 could probably use Terminal:

 B) As root, launch memtester 100M

 C) As root, run /nand/test.sh

 The two symptoms seen are:

 S1) On python error, the screen switches to the ShowPhotos activity
 console (text on white background), which shows both a network error and
 an ALSA error (both innocuous) and ends with various messages related to
 CPU faults.

 S2) On Sugar/X error, the screen returns to the login screen.

 Related symptoms may be:

 S3) Stopping the sequencing through the images --- a single image stays on
 the screen.

 S4) Displaying an error message about being unable to load an image.

-- 
Ticket URL: <http://dev.laptop.org/ticket/10680>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system