Re: [BUG?] Interrupts enabled after xen_acpi_processor_resume+0x0/0x34 [xen_acpi_processor]

[Stanislaw Gruszka <sgruszka@redhat.com>]

(Cc: correct Rafael email)

On Tue, Jan 28, 2014 at 09:24:57PM -0800, David Rientjes wrote:
> On Tue, 28 Jan 2014, Konrad Rzeszutek Wilk wrote:
> 
> > diff --git a/drivers/xen/xen-acpi-processor.c b/drivers/xen/xen-acpi-processor.c
> > index 7231859..7602229 100644
> > --- a/drivers/xen/xen-acpi-processor.c
> > +++ b/drivers/xen/xen-acpi-processor.c
> > @@ -46,7 +46,7 @@ module_param_named(off, no_hypercall, int, 0400);
> >   */
> >  static unsigned int nr_acpi_bits;
> >  /* Mutex to protect the acpi_ids_done - for CPU hotplug use. */
> > -static DEFINE_MUTEX(acpi_ids_mutex);
> > +static DEFINE_SPINLOCK(acpi_ids_lock);
> >  /* Which ACPI ID we have processed from 'struct acpi_processor'. */
> >  static unsigned long *acpi_ids_done;
> >  /* Which ACPI ID exist in the SSDT/DSDT processor definitions. */
> > @@ -68,7 +68,7 @@ static int push_cxx_to_hypervisor(struct acpi_processor *_pr)
> >  	int ret = 0;
> >  
> >  	dst_cx_states = kcalloc(_pr->power.count,
> > -				sizeof(struct xen_processor_cx), GFP_KERNEL);
> > +				sizeof(struct xen_processor_cx), GFP_ATOMIC);
> >  	if (!dst_cx_states)
> >  		return -ENOMEM;
> >  
> > @@ -149,7 +149,7 @@ xen_copy_pss_data(struct acpi_processor *_pr,
> >  		     sizeof(struct acpi_processor_px));
> >  
> >  	dst_states = kcalloc(_pr->performance->state_count,
> > -			     sizeof(struct xen_processor_px), GFP_KERNEL);
> > +			     sizeof(struct xen_processor_px), GFP_ATOMIC);
> >  	if (!dst_states)
> >  		return ERR_PTR(-ENOMEM);
> >  
> > @@ -275,9 +275,9 @@ static int upload_pm_data(struct acpi_processor *_pr)
> >  {
> >  	int err = 0;
> >  
> > -	mutex_lock(&acpi_ids_mutex);
> > +	spin_lock(&acpi_ids_lock);
> >  	if (__test_and_set_bit(_pr->acpi_id, acpi_ids_done)) {
> > -		mutex_unlock(&acpi_ids_mutex);
> > +		spin_unlock(&acpi_ids_lock);
> >  		return -EBUSY;
> >  	}
> >  	if (_pr->flags.power)
> > @@ -286,7 +286,7 @@ static int upload_pm_data(struct acpi_processor *_pr)
> >  	if (_pr->performance && _pr->performance->states)
> >  		err |= push_pxx_to_hypervisor(_pr);
> >  
> > -	mutex_unlock(&acpi_ids_mutex);
> > +	spin_unlock(&acpi_ids_lock);
> >  	return err;
> >  }
> >  static unsigned int __init get_max_acpi_id(void)
> 
> Looks incomplete, what about the kzalloc() in 
> xen_upload_processor_pm_data() and kcalloc()s in check_acpi_ids()?

Indeed and additionally from check_acpi_ids() we call
acpi_walk_namespace(), which also take mutexes. Hence unfortunately
making xen_upload_processor_pm_data() atomic is not easy, but possibly
can be done by saving some data in memory after initialization.

Or perhaps this problem can be solved differently, by not using 
yscore_ops->resume(), but some other resume callback from core, which
allow to sleep. Than can require registering dummy device or sysfs
class, but maybe there are simpler solutions.

Thanks
Stanislaw