Re: Regression on Dell XPS13 (was: [char-misc for 4.10-rc4 V2] mei: bus: enable OS version only for SPT and newer)

[Greg KH]

On Sat, Jan 21, 2017 at 12:49:50PM +0100, Rafael J. Wysocki wrote:
> On Sat, Jan 21, 2017 at 10:11 AM, Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> > On Fri, Jan 20, 2017 at 11:11:45PM +0000, Mario.Limonciello@xxxxxxxx wrote:
> >> Greg,
> >>
> >> > -----Original Message-----
> >> > From: Darren Hart [mailto:dvhart@xxxxxxxxxxxxx]
> >> > Sent: Tuesday, January 17, 2017 5:34 PM
> >> > To: Limonciello, Mario <Mario_Limonciello@xxxxxxxx>
> >> > Cc: gregkh@xxxxxxxxxxxxxxxxxxx; rafael.j.wysocki@xxxxxxxxx;
> >> > pmenzel@xxxxxxxxxxxxx; linux@xxxxxxxxxxxxx; tomas.winkler@xxxxxxxxx;
> >> > jan@xxxxxxxxxx; alexander.usyskin@xxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> >> > yu.c.chen@xxxxxxxxx; tomi.p.sarvela@xxxxxxxxx; daniel@xxxxxxxxx;
> >> > len.brown@xxxxxxxxx; linux-pm@xxxxxxxxxxxxxxx
> >> > Subject: Re: Regression on Dell XPS13 (was: [char-misc for 4.10-rc4 V2] mei:
> >> > bus: enable OS version only for SPT and newer)
> >> >
> >> > On Tue, Jan 17, 2017 at 06:38:43PM +0000, Mario.Limonciello@xxxxxxxx
> >> > wrote:
> >> > > > -----Original Message-----
> >> > > > From: Greg KH [mailto:gregkh@xxxxxxxxxxxxxxxxxxx]
> >> > > > Sent: Tuesday, January 17, 2017 12:24 PM
> >> > > > To: Limonciello, Mario <Mario_Limonciello@xxxxxxxx>
> >> > > > Cc: pmenzel@xxxxxxxxxxxxx; rafael.j.wysocki@xxxxxxxxx;
> >> > > > linux@xxxxxxxxxxxxx; tomas.winkler@xxxxxxxxx; jan@xxxxxxxxxx;
> >> > > > alexander.usyskin@xxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> >> > > > yu.c.chen@xxxxxxxxx; tomi.p.sarvela@xxxxxxxxx; daniel@xxxxxxxxx;
> >> > > > len.brown@xxxxxxxxx; linux-pm@xxxxxxxxxxxxxxx
> >> > > > Subject: Re: Regression on Dell XPS13 (was: [char-misc for 4.10-rc4 V2] mei:
> >> > > > bus: enable OS version only for SPT and newer)
> >> > > >
> >> > > > On Tue, Jan 17, 2017 at 04:57:49PM +0000, Mario.Limonciello@xxxxxxxx
> >> > > > wrote:
> >> > > > > So in the <6s scenario, the intel-hid driver is responsible to
> >> > > > > receive the ACPI event and process accordingly.  The maintainer
> >> > > > > has a patch ready for the intel-hid portion of this work, but it's
> >> > > > > currently being reviewed by Intel to ensure it can be legally submitted
> >> > into the kernel.
> >> > > >
> >> > > > Who at Intel do I need to go kick to make this mythical legal review
> >> > > > happen faster so we can see the code?
> >> > > >
> >> > > > Len and Rafael, what is going on here?
> >> > > >
> >> > >
> >> > > Len and Darren are both in the loop on the discussion around this patch.
> >> > > I don't know if they'll have any (public) comments they can add on the
> >> > > matter yet however.
> >> >
> >> > Thanks Mario. Yes, there isn't much to say here in public other than to confirm
> >> > we are keenly aware of the problem and have been actively working on fixing
> >> > it, both for this instance, and the deeper systematic failure that resulted in this
> >> > situation. No amount of kicking will expedite the process at this point, but
> >> > should we feel the need, we'll reach out.
> >> >
> >>
> >> The approval has come through and the patch has been submitted.
> >> http://www.spinics.net/lists/platform-driver-x86/msg10286.html
> >
> > Looks like it needs some work :)
> >
> > And why isn't it tagged to go to the 4.10-stable kernel if it really
> > does fix some systems?
> >
> >> Note: this is only half of the fix, the second half needs the ACPI subsystem to
> >> not be frozen to be able to receive this event.
> >
> > Where is that change?
> >
> > I'm still worried about 4.10-final, is that going to be broken for these
> > types of systems?
> 
> Nope.

Great!

> I'm going to revert commit 08b98d329165 (PM / sleep / ACPI: Use the
> ACPI_FADT_LOW_POWER_S0 flag) which introduced the problem in the first
> place (yes, the bisection result it misleading).  In fact, the revert
> is in my linux-next branch already, but I'm sort of out in the woods
> now and the actual pull request will be sent next week.
> 
> BTW, this is a usual process.  We had tried to change the default
> behaviour for certain class of systems and it didn't work, so we need
> to take a step back, fix the problems that were exposed and try again
> later.
> 
> It has been confused quite a bit, though, which is kind of worrisome.

Not a problem at all, and it's fine that it happened this way, I just
wanted to make sure the regression didn't hit a -final release.

Have a great vacation/trip.

thanks,

greg k-h