Bug 5336

Summary: 'init' command does not work as intended
Product: [Runtime] General Runtime Reporter: Stoicescu Cornel <corneliux.stoicescu>
Component: General RuntimeAssignee: Aníbal Limón <anibal.limon>
Status: RESOLVED FIXED QA Contact:
Severity: normal    
Priority: Medium CC: anibal.limon, dvhart, Qi.Chen
Version: 1.5   
Target Milestone: 1.8 M4   
Hardware: x86   
OS: Multiple   
Whiteboard:
OS type for building Yocto: --- Type of Regression: ---
Verified: Documentation change: No (bug/feature does not impact docs)

Description Stoicescu Cornel 2013-10-10 14:40:54 UTC 
When in runlevel X and issuing command 'init Y', the system should cleanly migrate to runlevel Y. This behaviour does not happen in most cases using Yocto images.

Some examples of bad behavior:
1) In any image(testes core-image-sato, core-image-lsb-sdk) issuing command 'init 1' will result in a technical hang

2) In core-image-lsb-sdk images, from the default runlevel 5, issuing the sequence of commands: 'init 3' 'init 5' 'init 3' 'init 5' will result in an oprofile-server:1054 error(could not bind to the server. Address already in use).

3) In core-image-sato-sdk images issuing 'init 3' command from the GUI terminal is completely ignored.


NOTE: 'init 0' and 'init 6' commands correctlr shutdown and restart the system(during my tests)
Comment 1 Saul Wold 2014-05-12 17:49:30 UTC 
Is this still an issue and if so, can you verify if it's just a real HW (which 
machines) and/or Qemu?
Comment 2 Stoicescu Cornel 2014-05-27 07:15:34 UTC 
We will test this on meta-intel HW targets today. Core BSPs will be tested in the next few days as well and QEMU targets on the first weekly test. I will get as much feedback and post the results here.
Comment 3 Stoicescu Cornel 2014-06-04 09:40:56 UTC 
Verified on meta-intel 1.6 release candidates and issues happen as described.
Comment 4 Stephen K Jolley 2014-06-18 17:24:03 UTC 
Info Supplied, moved back out of NEEDINFO per RP.
Comment 5 Stoicescu Cornel 2014-07-16 10:08:43 UTC 
Also tested qemux86-64 with poky master:a5531a2b8983318b99c119a87b78a92cf84160bb and the issue is the same.
Comment 6 Aníbal Limón 2014-11-04 20:51:39 UTC 
Stoicescu,

I'm doing some debugging with qemux86-64/core-image-sato using master branch, i have some concerns.

1) core-image-sato init 1, i reproduce this but it seems that not hang only didn't display shell in the monitor output, i used serial and vm it's running.

$ runqemu qemux86-64 core-image-sato serial

This is the issue?

2) core-image-lsb sdk init 3 - init 5, i can't reproduce the problem do you have another information?

3) core-image-sato-sdk init 3, gives me the same problem that 1).
Comment 7 Stoicescu Cornel 2014-11-11 08:34:44 UTC 
Lucian, can you please help investigate the current status of the issues presented in this bug? Try to reproduce steps 1, 2, 3 from the first post.
Comment 8 Lucian Musat 2014-12-17 11:48:44 UTC 
Tried to reproduce this on 1.8 M1 (b813bdebb36501500e86fea5f7e15b4b15ea0902) Lenovo X220 laptop genericx86-64 and this is what happened:

1. Core-image-sato-sdk image hangs on init 1, 3 and 5. Works for init 0 and 6.
2. Core-image-lsb-sdk ignores init 3 and 5. Actually it tries the first time but fails with: 
"INIT: Switching to runlevel 3
cups: started scheduler
umount: /mnt/.psplash: not mounted"
and returns to console. After that init 3 and 5 are ignored.
Init 1 hangs at "INIT: no more processes left in this runlevel"
Init 0 and 6 works.