[PATCH v7 04/16] ACPI: processor: Move checks and availability of acpi_processor earlier

[Jonathan Cameron]

On Mon, 22 Apr 2024 20:56:55 +0200
"Rafael J. Wysocki" <rafael at kernel.org> wrote:

> On Thu, Apr 18, 2024 at 3:56 PM Jonathan Cameron
> <Jonathan.Cameron at huawei.com> wrote:
> >
> > Make the per_cpu(processors, cpu) entries available earlier so that
> > they are available in arch_register_cpu() as ARM64 will need access
> > to the acpi_handle to distinguish between acpi_processor_add()
> > and earlier registration attempts (which will fail as _STA cannot
> > be checked).
> >
> > Reorder the remove flow to clear this per_cpu() after
> > arch_unregister_cpu() has completed, allowing it to be used in
> > there as well.
> >
> > Note that on x86 for the CPU hotplug case, the pr->id prior to
> > acpi_map_cpu() may be invalid. Thus the per_cpu() structures
> > must be initialized after that call or after checking the ID
> > is valid (not hotplug path).
> >
> > Signed-off-by: Jonathan Cameron <Jonathan.Cameron at huawei.com>
> > ---
> > v7: Swap order with acpi_unmap_cpu() in acpi_processor_remove()
> >     to keep it in reverse order of the setup path. (thanks Salil)
> >     Fix an issue with placement of CONFIG_ACPI_HOTPLUG_CPU guards.
> > v6: As per discussion in v5 thread, don't use the cpu->dev and
> >     make this data available earlier by moving the assignment checks
> >     int acpi_processor_get_info().
> > ---
> >  drivers/acpi/acpi_processor.c | 78 +++++++++++++++++++++--------------
> >  1 file changed, 46 insertions(+), 32 deletions(-)
> >
> > diff --git a/drivers/acpi/acpi_processor.c b/drivers/acpi/acpi_processor.c
> > index ba0a6f0ac841..ac7ddb30f10e 100644
> > --- a/drivers/acpi/acpi_processor.c
> > +++ b/drivers/acpi/acpi_processor.c
> > @@ -183,8 +183,36 @@ static void __init acpi_pcc_cpufreq_init(void) {}
> >  #endif /* CONFIG_X86 */
> >
> >  /* Initialization */
> > +static DEFINE_PER_CPU(void *, processor_device_array);
> > +
> > +static void acpi_processor_set_per_cpu(struct acpi_processor *pr,
> > +                                      struct acpi_device *device)
> > +{
> > +       BUG_ON(pr->id >= nr_cpu_ids);
> > +       /*
> > +        * Buggy BIOS check.
> > +        * ACPI id of processors can be reported wrongly by the BIOS.
> > +        * Don't trust it blindly
> > +        */
> > +       if (per_cpu(processor_device_array, pr->id) != NULL &&
> > +           per_cpu(processor_device_array, pr->id) != device) {
> > +               dev_warn(&device->dev,
> > +                        "BIOS reported wrong ACPI id %d for the processor\n",
> > +                        pr->id);
> > +               /* Give up, but do not abort the namespace scan. */
> > +               return;  
> 
> In this case the caller should make acpi_pricessor_add() return 0, I
> think, because otherwise it will attempt to acpi_bind_one() "pr" to
> "device" which will confuse things.
> 
> So I would make this return false to indicate that.
> 
> Or just fold it into the caller and do the error handling there.

The bios bug mentioned in reply to patch 14 (DSDT entries for non existent CPUs
that have no _STA entries) showed me that we need to know if this succeeded
(I'd not read this at that point).

I'll make it return a bool to say this succeeded and in both call sites
return 0 if not to deal with the bios bug here.  Making sure not to clear
the per_cpu() structures unless this we get past that call.  If we do
and arch_register_cpu() fails we need to clear these two IDs.

Doing so means that acpi_processor_hotadd_init() is side effect free and
hence we can return in acpi_processor_get_info() which avoids the
need to clear pointers when we don't have a valid pr->id to do it with.

So fully agree we need to bail out properly if this fails.

Jonathan