Olpc-runin-tests

From OLPC
Revision as of 22:06, 30 January 2012 by Quozl (talk | contribs)
Jump to: navigation, search

olpc-runin-tests is a collection of tests for the XO-1.5, XO-1.75 and XO-3 that are run during late stages of laptop production. It might also be useful for in-field diagnostics. It is designed to test the hardware as well as the software installation, but is not designed to cover all software functions.

By default, the tests are run in a loop for several hours. If any test fails at any point, the system is immediately shut down.

Logs are written in /runin/logdir. fail.log is the single log which you can quickly check for success or failure; the file is empty on success.

For developers

Source:

Current maintainer: James Cameron

Questions and contributions should be sent to the devel mailing list.

How to test

At the firmware ok prompt, set the TS manufacturing tag to RUNIN:

ok delete-tag TS
ok add-tag TS RUNIN

Boot the system as normal, and instead of the normal boot routine, the runin tests will kick in.

When done, use the ok prompt to reset the TS tag to SHIP:

ok delete-tag TS
ok add-tag TS SHIP

Original design document

Note: this is not fully reflective of the current implementation, but is enough to give you a feel.

 Early in the boot (rc.d S00- or some such) check the TS mfg flag for run in

 if !run-in then exit
    run-in =  [ `cat /OFW/mfg-data/TS` == "RUNIN" ] (or what ever bash function we have to do this)

 Start run-in test:
  
  remove /bootpart/runin/fail.log
  remove /bootpart/runin/olpc.fth
  
  uncompress fs_hash.md5.whatever
 
  Start X with our set of tests in xinitrc.

  Grab start time.
  Start tests:

    - Set the charger pwm to 0% so it runs off of battery as much as possible
      Tell EC not to charge battery. Run test for n hours on battery. 
      After 4-n hours re-enable full power and tell EC to charge battery
      
    - Run md5deep to compare the all the files vs the manifest.
      run over and over.
	? running multiple md5deeps on different parts of the fs
	for a different type of stress.
    
    - Run cpuburn with floating point
    
    - Run memtester
    
    - Run camera with output to the screen an possbily to file

    - Run play audio (file of silence)

    - Check cputemp

    - ? Turn wireless on and off by reloading firmware?
    - ? Suspend resume?
    - ? do some r and r stuff to work the video engine harder ?	

    - Reboot tests [Not quite sure how to do with other long running tests]

  Wait time < 4hours:
    watch /tmp/fail.log for any items to show up
    if present the write fail.log and olpc.fth to /bootpart/runin and reboot

  Check battery to make sure it returned to fully charged

 Clean up droppings

 Re-run md5deep to verify the files are still pristine.
 Copy Run-In result file into /bootpart/runin
  • Each test runs in its own process and checks for pass fail and appends to /tmp/fail.log
  • If any test fails it outputs a fail report into a 'fail.log'. Top level script watchs for items in /tmp/fail.log and if found aborts the runin early. Then copies the /tmp/fail.log and rundone.fth to /bootpart/runin.