Re: [PATCH v9 3/3] PCI/ACPI: Use device constraints to decide PCI target state fallback policy

[Andy Shevchenko]

On Thu, Aug 03, 2023 at 11:37:10PM -0500, Mario Limonciello wrote:
> On 8/3/23 23:30, Andy Shevchenko wrote:
> > On Thu, Aug 03, 2023 at 08:02:29PM -0500, Mario Limonciello wrote:

...

> > > +	pci_dbg(dev, "ACPI device constraint: %d\n", constraint);
> > 
> > Does it make sense before the below check? Why can we be interested in the
> > _exact_ negative values? (Note that non-printing is already a sign that either
> > we don't call this or have negative constraint.)
> 
> There are two different negative values that can come up:
> -ENODEV or -EINVAL.  Both were interesting while coming up with this series
> because they mean something different about why a constraint wasn't
> selected.
> 
> -ENODEV means the constraint wasn't found.
> -EINVAL means the constraint was found but something is wrong with the table
> parser or the table itself.  I found the table parser wasn't working
> correctly originaly thanks to this.
> 
> Maybe now that I've got it all working you're right and this should go
> after the error checking.

Or maybe moved to the acpi_get_lps0_constraint().

-- 
With Best Regards,
Andy Shevchenko