Bug 114425

Summary: irqbalance installed and started on non smp system
Product: [openSUSE] SUSE LINUX 10.0 Reporter: Danny Al-Gaaf <dalgaaf>
Component: InstallationAssignee: Jiri Srain <jsrain>
Status: RESOLVED FIXED QA Contact: Klaus Kämpf <kkaempf>
Severity: Critical    
Priority: P5 - None CC: anders, jsrain, ma
Version: Beta 4   
Target Milestone: ---   
Hardware: Other   
OS: All   
Whiteboard:
Found By: Other Services Priority:
Business Priority: Blocker: ---
Marketing QA Status: --- IT Deployment: ---
Attachments: yast logs

Description Danny Al-Gaaf 2005-08-31 14:51:32 UTC
Fresh installed Beta4 install and starts irqbalance on a non smp machine (ACER 
Travelmate C300XMi laptop).
Comment 1 Danny Al-Gaaf 2005-08-31 14:52:07 UTC
*** Bug 114424 has been marked as a duplicate of this bug. ***
Comment 2 Danny Al-Gaaf 2005-08-31 14:52:25 UTC
Created attachment 48329 [details]
yast logs
Comment 3 Lukas Ocilka 2005-09-01 15:43:38 UTC
Michael, I'm not really sure but aren't you responsible for that?
Comment 4 Klaus Kämpf 2005-09-01 15:54:46 UTC
I didn't find irqbalance mentioned within yast. This must be triggered by some 
other source. 
Comment 5 Klaus Kämpf 2005-09-01 15:59:16 UTC
Its part of the normal package selection  
  
Comment 6 Anders Norrbring 2005-09-02 14:13:15 UTC
irq_balance is both installed and activated on beta 4.

Running a VMware virtual setup, single Opteron CPU in the VM.
Comment 7 Danny Al-Gaaf 2005-09-02 16:48:59 UTC
> Its part of the normal package selection 

But it should not. For what need I this on a normal x86 single CPU system? It is 
started but fails. If this is part of the normal package selection, then remove 
it from there and install only if a multiprocessor system is detected!
Comment 8 Anders Norrbring 2005-09-02 16:57:36 UTC
Well, not exactly "normal"...

To specify it precisely, the complete system is like this:

Host server: Dual Opteron
Host O/S: SLES9

VMware GSX server 3.2

The virtual machine that I install SuSE 10 on is under this host, and to the
guest (virtual mahine) the hardware platform looks like a single CPU (Opteron)
machine.

So you're right about that irq_balance should not be enabled. I understand that
it's installed since it's a base package, but should only be enabled on
multi-CPU platforms, which VMware GSX cannot emulate.
Comment 9 Danny Al-Gaaf 2005-09-02 18:16:29 UTC
> So you're right about that irq_balance should not be enabled. I understand 
> that it's installed since it's a base package, but should only be enabled on
> multi-CPU platforms, which VMware GSX cannot emulate.

No, it should not be installed. Not to start is only a workaround. This packages 
sould never be installed on a system with single processor by default or one of 
the package selections. We check so many other things while installation e.g. if 
the machine is a toshiba laptop ... so why not check if this is a multiprocessor 
system? IMO this was also not installed and started on Beta3. Why now?
Comment 10 Anders Norrbring 2005-09-02 18:22:06 UTC
Alright.. My mistake about the install.

I didn't run beta 3, I've been away for a couple of days, so I went from 2 to 4.
 Now that you mention it, I didn't notice it on beta 2, but that wasn't run in
VMware here, I ran that natively on a single opteron box.
Comment 11 Andreas Jaeger 2005-09-05 11:28:12 UTC
Stefan, let's take it out of the selections again for 10.0.

irqbalance is needed by kernel-smp on x86-64 systems - and therefore let's not
penalize everybody else.

please assign back to ma as enhancement for 10.1.
Comment 12 Klaus Kämpf 2005-09-05 11:48:06 UTC
should be fairly easy to 'fix' this in YaST. 
We already have this code: 
        // add numactl on x86_64 with SMP 
        if (Arch::has_smp () && Arch::x86_64 ()) 
        { 
            packages = add (packages, "numactl"); 
        } 
 
adding another 
packages = add (packages, "irqbalance"); 
seems rather uncritical ... 
Comment 13 Stefan Dirsch 2005-09-05 12:29:31 UTC
Reassigning to ma ...
Comment 14 Jiri Srain 2005-09-05 12:33:11 UTC
This is not Michael's, but mine. 
 
Andreas, do you want to add the suggested piece of code? To which 
architectures (limit it to only AMD64, or for all archs)? 
Comment 15 Jiri Srain 2005-09-05 12:34:21 UTC
Andreas, please, check comments #12 and #14...  
Comment 16 Andreas Jaeger 2005-09-05 14:25:39 UTC
Yes, please add for x86-64 and SMP - as numactl above.
Comment 17 Jiri Srain 2005-09-06 11:21:41 UTC
Done in SVN, submitting to STABLE.