[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] crash in nvmx_vcpu_destroy



On Thu, Feb 21, Tim Deegan wrote:

> At 11:01 +0000 on 21 Feb (1361444479), Tim Deegan wrote:
> > (Cc'ing the vmx maintainers)
> > 
> > At 15:58 +0100 on 20 Feb (1361375903), Olaf Hering wrote:
> > > while doing "while xm migrate --live domU localhost;do sleep 1;done" I
> > > just got the crash shown below. And it can be reproduced.
> > > 
> > > The guest has 2 vcpus and 512mb, it runs pvops 3.7.9
> > 
> > Anything interesting printed before the crash?  My best guess by code
> > inspection is that nvmx->launched_list never got initialized, because of
> > some failure in vcpu init.
> > 
> > Also, if you have the xen-syms for this image, can you extract a
> > file/line-number for the crashing %rip (ffff82c4c01dd197)?
> > I'd expect it to be vvmx.c:150 or thereabouts.
> > 
> > And thirdly, can you try the attached patch?
> 
> Oops - not sure what I tested before , but that one doesn't even
> compile!  Try this instead.

This patch fixes the crash for me. Thanks.

Olaf

> diff --git a/xen/arch/x86/hvm/vmx/vvmx.c b/xen/arch/x86/hvm/vmx/vvmx.c
> index 4f3f94d..5d00ff7 100644
> --- a/xen/arch/x86/hvm/vmx/vvmx.c
> +++ b/xen/arch/x86/hvm/vmx/vvmx.c
> @@ -147,10 +147,13 @@ void nvmx_vcpu_destroy(struct vcpu *v)
>          nvcpu->nv_n2vmcx = NULL;
>      }
>  
> -    list_for_each_entry_safe(item, n, &nvmx->launched_list, node)
> +    if ( nvmx->launched_list.next )
>      {
> -        list_del(&item->node);
> -        xfree(item);
> +        list_for_each_entry_safe(item, n, &nvmx->launched_list, node)
> +        {
> +            list_del(&item->node);
> +            xfree(item);
> +        }
>      }
>  
>      if ( v->arch.hvm_vmx.vmread_bitmap )


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.