[PATCH] arch/x86: Fix kdump on x86 with physically hotadded CPUs

Jiri Olsa jolsa at redhat.com
Thu Oct 6 23:49:27 PDT 2016


On Thu, Oct 06, 2016 at 11:25:43AM -0400, Prarit Bhargava wrote:
> 
> 
> On 10/05/2016 12:14 PM, Jiri Olsa wrote:
> > On Tue, Oct 04, 2016 at 04:38:16PM +0200, Thomas Gleixner wrote:
> >> On Tue, 4 Oct 2016, Prarit Bhargava wrote:
> >>> On 10/04/2016 06:58 AM, Thomas Gleixner wrote:
> >>>> While it is the right thing to initialize the package map in that case, it
> >>>> still papers over a robustness issue in the uncore code, which needs to be
> >>>> fixed first.
> >>>
> >>> I will include a separate patch with an error check for pkg == 0xffff in the
> >>> uncore code.
> >>
> >> 0xffff? That won't help. The id returned is -1 if the entry is not
> >> initialized. And aside of that just patching that particular place is not
> >> helping as the uncore code and also rapl is relying on the package map
> >> being populated.
> >>
> >> So we need a sanity check in the initialization code which prevents any of
> >> this being executed.
> > 
> > I still need to test this, but how about something like this?
> > 
> > thanks,
> > jirka
> > 
> > 
> > ---
> > diff --git a/arch/x86/events/intel/rapl.c b/arch/x86/events/intel/rapl.c
> > index 28865938aadf..61d087a2f25d 100644
> > --- a/arch/x86/events/intel/rapl.c
> > +++ b/arch/x86/events/intel/rapl.c
> > @@ -598,8 +598,13 @@ static int rapl_cpu_online(unsigned int cpu)
> >  
> >  static int rapl_cpu_prepare(unsigned int cpu)
> >  {
> > -	struct rapl_pmu *pmu = cpu_to_rapl_pmu(cpu);
> > +	struct rapl_pmu *pmu;
> > +	int pkg = topology_logical_package_id(cpu);
> > +
> > +	if (WARN_ON(pkg == -1))
> > +		return -EINVAL;
> >  
> > +	pmu = cpu_to_rapl_pmu(cpu);
> >  	if (pmu)
> >  		return 0;
> 
> I thought about doing this but it seems like every time some driver uses
> topology_logical_package_id() the driver would have to replicate the error
> checking code.

hm, unless we guarantee topology_logical_package_id always
returns sane values I dont see another way

jirka



More information about the kexec mailing list