Re: [PATCH v11 8/9] PCI: Split PME state selection into a local static function

[Andy Shevchenko]

On Thu, Aug 10, 2023 at 11:29:49AM -0500, Limonciello, Mario wrote:
> On 8/10/2023 11:21 AM, Andy Shevchenko wrote:
> > On Wed, Aug 09, 2023 at 01:54:52PM -0500, Mario Limonciello wrote:

...

> > > +static inline pci_power_t pci_get_wake_pme_state(struct pci_dev *dev)
> > > +{
> > > +	pci_power_t state = PCI_D3hot;
> > > +
> > > +	while (state && !(dev->pme_support & (1 << state)))
> > > +		state--;
> > > +
> > > +	return state;
> > 
> > Sparse won't be happy about this code (with CF=-D__CHECK_ENDIAN__).
> > 
> > Basically it's something like
> > 
> > 	return (__force pci_power_t)fls(dev->pme_support & GENMASK(PCI_D3hot, 0));
> > 
> > (but double check and test the logic).
> > 
> > > +}
> > 
> > ...
> > 
> > Yeah, I see that is the existing code, perhaps amend it first?
> > 
> 
> Are you sure?

Yes.

Just the original code

drivers/pci/pci.c:2711:60: warning: restricted pci_power_t degrades to integer
drivers/pci/pci.c:2712:30: warning: restricted pci_power_t degrades to integer

                /*
                 * Find the deepest state from which the device can generate
                 * PME#.
                 */
2711 ==>         while (state && !(dev->pme_support & (1 << state)))
2712 ==>                state--;

How is yours different?

> I actually double checked the sparse output using this
> command before I sent it.
> 
> make W=1 C=1 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__' drivers/pci/pci.o
> 
> I didn't see anything related to the line numbers for this function.

Just in case...

$ gcc --version
gcc (Debian 12.3.0-5) 12.3.0

$ sparse --version
0.6.4 (Debian: 0.6.4-3)

$ make --version
GNU Make 4.3
Built for x86_64-pc-linux-gnu

-- 
With Best Regards,
Andy Shevchenko