Re: linux-next: Tree for November 22 (kvm)

[Zachary Amsden]

On 11/30/2010 02:17 AM, Glauber Costa wrote:
> On Mon, 2010-11-29 at 08:08 -1000, Zachary Amsden wrote:
>    
> > On 11/29/2010 07:52 AM, Randy Dunlap wrote:
> >      
> > > On 11/29/10 09:47, Zachary Amsden wrote:
> > >
> > >        
> > > > On 11/29/2010 06:35 AM, Avi Kivity wrote:
> > > >
> > > >          
> > > > > On 11/29/2010 06:33 PM, Randy Dunlap wrote:
> > > > >
> > > > >            
> > > > > > On Mon, 22 Nov 2010 13:26:27 -0800 Randy Dunlap wrote:
> > > > > >
> > > > > >
> > > > > >              
> > > > > > >    On Mon, 22 Nov 2010 13:49:11 +1100 Stephen Rothwell wrote:
> > > > > > >
> > > > > > >
> > > > > > >                
> > > > > > > >    Hi all,
> > > > > > > >
> > > > > > > >    Changes since 20101119:
> > > > > > > >
> > > > > > > >                  
> > > > > > >    kvm.c:(.init.text+0x11f49): undefined reference to
> > > > > > >
> > > > > > >                
> > > > > > `kvm_register_clock'
> > > > > >
> > > > > >              
> > > > > > >    when CONFIG_KVM_CLOCK is not enabled.
> > > > > > >
> > > > > > >                
> > > > > > BUild error still present in linux-next-2010-NOV-29.
> > > > > >
> > > > > >
> > > > > >              
> > > > > Glauber, Zach?
> > > > >
> > > > >
> > > > >            
> > > > I can only speculate this reference is being called from smpboot without
> > > > CONFIG guarding?
> > > >
> > > >          
> > > Sorry, looks like I dropped the first line of the error messages:
> > >
> > > arch/x86/built-in.o: In function `kvm_smp_prepare_boot_cpu':
> > > kvm.c:(.init.text+0xad38): undefined reference to `kvm_register_clock'
> > >
> > > from arch/x86/kernel/kvm.c:
> > >
> > > #ifdef CONFIG_SMP
> > > static void __init kvm_smp_prepare_boot_cpu(void)
> > > {
> > > 	WARN_ON(kvm_register_clock("primary cpu clock"));
> > > 	kvm_guest_cpu_init();
> > > 	native_smp_prepare_boot_cpu();
> > > }
> > >
> > > so it looks like you are correct...
> > >
> > >        
> > Looks like this is the appropriate fix:
> >
> > #ifdef CONFIG_SMP
> > static void __init kvm_smp_prepare_boot_cpu(void)
> > {
> > #ifdef CONFIG_KVM_CLOCK
> > 	WARN_ON(kvm_register_clock("primary cpu clock"));
> > #endif
> > 	kvm_guest_cpu_init();
> > 	native_smp_prepare_boot_cpu();
> > }
> >
> >
> > The SMP code is still buggy as well, wrt printk timing, in that it
> > doesn't get called early enough, correct?  Has anyone thought of a good
> > solution to that problem?
> >
> > Basically the problem is CPU-1 will get CPU-0's per-cpu areas copied
> > over, and these are not valid for CPU-1.  If the clocksource is used on
> > CPU-1 before kvm clock gets setup, it can go backwards, wreaking havoc,
> > causing panic, etc.
> >
> > What is the best test to guard against this?  Perhaps we should keep the
> > CPU number in the per-cpu data and test against it?
> >      
> Can we identify precisely when it happens? If we can, we can try to
> force a hypervisor exit or re-register right after the data is copied
> over. This will force the per-cpu structure to get updated with good
> values.
>    

It's as if we need a new type of per-cpu data, per_cpu_zeroed, which is 
always cleared prior to starting on the AP.

I imagine other code paths may need this as well.

Zach
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/