Bug 639393

Summary: startpar fails for no apparent reason
Product: [openSUSE] openSUSE 11.3 Reporter: Rob Fortune <rob.fortune>
Component: BasesystemAssignee: E-mail List <bnc-team-screening>
Status: VERIFIED FIXED QA Contact: E-mail List <qa-bugs>
Severity: Minor    
Priority: P5 - None    
Version: Final   
Target Milestone: ---   
Hardware: Other   
OS: openSUSE 11.3   
Whiteboard:
Found By: --- Services Priority:
Business Priority: Blocker: ---
Marketing QA Status: --- IT Deployment: ---
Attachments: Log of running /etc/init.d/rc to runlevel 3

Description Rob Fortune 2010-09-14 22:45:08 UTC
User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US) AppleWebKit/534.7 SUSE/7.0.513.0 (KHTML, like Gecko) Chrome/7.0.513.0 Safari/534.7

I have an openvz based VPS. Anyway, I deleted their distro and put OpenSUSE on it of course. The init scripts fail for two reasons, but I will report only one of them here.


Reproducible: Always

Steps to Reproduce:
1. Get a openvz container with suse in it
2. Boot it
3.
Actual Results:  
Fails to boot with

tcgetattr: Inappropriate ioctl for device 



Expected Results:  
Completed boot

I checked out startpar from obs:

osc co 'openSUSE:11.3:Update' sysvinit


--- startpar.c~ 2009-08-20 12:11:03.000000000 +0100
+++ startpar.c  2010-09-14 23:31:28.541357764 +0100
@@ -893,7 +893,7 @@
   if (tcgetattr(0, &tio))
     {
       perror("tcgetattr");
-      exit(1);
+      //exit(1);
     }
   if (ioctl(0, TIOCGWINSZ, &wz) == 0)
     wzok = 1;

Commenting out the exit line if that fails seems to have no effect and makes OpenSUSE sysvinit scripts work in the openvz containers, with one other change I will make a separate bug report for.

It does spew other errors about those tc* kernel interfaces but again they're just spam and seem to have no effect, the system runs fine.
Comment 1 Rob Fortune 2010-09-14 23:28:22 UTC
And you don't see the bootup messages on most containers anyway.
Comment 2 Dr. Werner Fink 2010-09-15 06:45:04 UTC
Why does the openvz container not allow to get information about
the used terminal? This is IMHO not a bug of startpar but more a
bug of the environment.  Beside this what message do you see?
Comment 3 Rob Fortune 2010-09-16 09:31:03 UTC
Created attachment 389914 [details]
Log of running /etc/init.d/rc to runlevel 3
Comment 4 Rob Fortune 2010-09-16 09:34:12 UTC
I never suggested it was a startpar bug, suse wasn't designed to be run inside such a restrictive little box. You can't change the kernel, are denied access to loopback devices and to fuse. This is the kind of setup you get with a low-end VPS.

If you want more info, why not grab an account yourself. They're $5 a month and I think you have a trial period or some such but I don't guarantee it. The first month comes will fully managed support and the admins are very responsive to questions, I'm sure they would be very happy to answer questions you have which makes suse play more nicely in their world.

http://myvirpus.com/aff.php?aff=303
Comment 5 Rob Fortune 2010-09-16 09:36:29 UTC
They're good little VPS as it goes, not very reliable ($5) but you get a large amount of bandwidth and diskspace and a decent amount of ram, far beyond any other provider I know of in that price bracket.
Comment 6 Rob Fortune 2010-09-16 09:38:02 UTC
Sorry for the colours in the PS4 trace, you might want to download that locally and look at it with less. The colours are helpful here but they make reading it online a mess.
Comment 7 Rob Fortune 2010-09-20 14:32:25 UTC
The thing is, I agree completely that it's an overzealous restriction on rights, but startpar doesn't fail when it can't complete the terminal info calls as demonstrated by the fact if I just comment out the exit it boots successfully. So it's more like a soft rather than hard error.
Comment 8 Rob Fortune 2010-09-20 14:34:26 UTC
My suggestion is that you don't exit but issue a warning and continue to try to boot, there doesn't seem to be any sense in aborting and leaving you without the system booted correctly and it does seem to work when it can't get the info anyway.
Comment 9 Dr. Werner Fink 2010-09-28 14:29:19 UTC
Just changed that for the next release