Re: Please revert "ACPICA: AML interpreter: add region addresses in global list during initialization"

[Jean Delvare <jdelvare@suse.de>]

Hi Erik,

On Tue, 20 Nov 2018 18:30:22 +0000, Schmauss, Erik wrote:
> > -----Original Message-----
> > From: Greg KH [mailto:greg@kroah.com]
> > Sent: Tuesday, November 20, 2018 1:12 AM
> > To: Jean Delvare <jdelvare@suse.de>
> > Cc: stable@vger.kernel.org; Guenter Roeck <linux@roeck-us.net>;
> > Schmauss, Erik <erik.schmauss@intel.com>; Wysocki, Rafael J
> > <rafael.j.wysocki@intel.com>
> > Subject: Re: Please revert "ACPICA: AML interpreter: add region addresses in
> > global list during initialization"
> > 
> > On Tue, Nov 20, 2018 at 10:03:59AM +0100, Jean Delvare wrote:  
> > > On Tue, 20 Nov 2018 09:54:19 +0100, Greg KH wrote:  
> > > > Ok, I'll go revert this, but shouldn't it also be reverted in
> > > > Linus's tree as well?  
> > >  
> > > No. As I understand it (with my limited knowledge of ACPICA), the
> > > change itself is correct. The problem is that it will detect resource
> > > conflicts which were unnoticed before, and that will prevent drivers
> > > from loading. Some of them may be addressed with driver fixes or new
> > > drivers. Others are false positives (due to bogus BIOS) which users
> > > will have to work around with acpi_resource_conflicts=lax. We have
> > > been through this before, nothing new really, but it takes years to
> > > address such problems. This just can't be done in stable kernel series.  
> 
> I would like to give you more context.
> 
> There was a fairly complicated change that occurred in 4.17 and we
> caused a regression by forgetting to add region addresses in a global list
> during operation region initialization. We found the regression when bug
> reporters tried to boot their macbook pro and asus laptop and saw that
> there was a difference in behavior when drivers are being loaded

Commit 4abb951b73ff0a8a979113ef185651aa3c8da19b has no Fixes tag. Which
commit introduced the regression? Can you point me to the associated
bug reports?

> So what I am trying to say is that we have been emitting these errors for a
> while before we caused the regression. The goal with this patch is to keep
> the behavior the same as kernels older than 4.17 where warnings are
> printed to dmesg due to resource conflicts.

Fine with me for upstream, but I still need to be convinced that it
belongs to stable series. For now, the only 2 related bugs I know of are
#200011 (which is NOT fixed by commit
4abb951b73ff0a8a979113ef185651aa3c8da19b) and #201721 (which is caused
by that commit). 1 vs 0, revert wins. If you want me to change my mind,
you must provide additional data points proving that commit
4abb951b73ff0a8a979113ef185651aa3c8da19b solves more functional
regressions than it causes.

Thanks,
-- 
Jean Delvare
SUSE L3 Support