New changes have appeared in the meantime, in particular one partly undoing
what we still haven't merged (patch 1 here).

1: add 'preferred_cstates' module argument
2: add core C6 optimization for SPR
3: add AlderLake support
4: disable IBRS during long idle
5: make SPR C1 and C1E be independent

Jan

Henry,

On 18.08.2022 15:02, Jan Beulich wrote:
> New changes have appeared in the meantime, in particular one partly undoing
> what we still haven't merged (patch 1 here).
> 
> 1: add 'preferred_cstates' module argument
> 2: add core C6 optimization for SPR
> 3: add AlderLake support
> 4: disable IBRS during long idle
> 5: make SPR C1 and C1E be independent

strictly speaking patches 3-5 are late submissions. Patch 5, however,
actually corrects patch 1, and I'd prefer to keep things in the order
in which they were put in for Linux. Whether we actually want patch 4
is to be determined; if not that one should be easy to leave out. In
any event I'd like to ask for you to consider granting an exception
on these last three patches.

Thanks, Jan

Hi Jan,

> -----Original Message-----
> From: Jan Beulich <jbeulich@suse.com>
> Subject: Re: [PATCH v3 0/5] x86/mwait-idle: (remaining) SPR + (new) ADL
> support
> 
> Henry,
> 
> On 18.08.2022 15:02, Jan Beulich wrote:
> > New changes have appeared in the meantime, in particular one partly
> undoing
> > what we still haven't merged (patch 1 here).
> >
> > 1: add 'preferred_cstates' module argument
> > 2: add core C6 optimization for SPR
> > 3: add AlderLake support
> > 4: disable IBRS during long idle
> > 5: make SPR C1 and C1E be independent
> 
> strictly speaking patches 3-5 are late submissions. Patch 5, however,
> actually corrects patch 1, and I'd prefer to keep things in the order
> in which they were put in for Linux. Whether we actually want patch 4
> is to be determined; if not that one should be easy to leave out. In
> any event I'd like to ask for you to consider granting an exception
> on these last three patches.

Thank you for informing this.

Let me add this series to my list first. The feature freeze date is
Fri Sep 2, 2022 and code freeze date is Fri Sep 23, 2022. My personal
feeling would be as long as (1) this series can get a good shape to go in
before the code freeze and (2) other x86 maintainers don't have
objection to accept this series, it is fine with me. Also if x86 people want
this series in the release and it won't cause the delay of release, I don't
think there would be any reason to refuse this series :))) So I guess please
just keep these 3 patches for now and let's see what the discussion in
this series will be.

Kind regards,
Henry

> 
> Thanks, Jan

Henry,

On 18.08.2022 15:02, Jan Beulich wrote:
> New changes have appeared in the meantime, in particular one partly undoing
> what we still haven't merged (patch 1 here).
> 
> 1: add 'preferred_cstates' module argument
> 2: add core C6 optimization for SPR
> 3: add AlderLake support
> 4: disable IBRS during long idle
> 5: make SPR C1 and C1E be independent

just now acks have completed coming in for this series of ports of earlier
Linux commits. This all targets newer hardware which of course we'd like
to properly support with 4.17. Unfortunately the series was kind of lost
with all the security work which had been ongoing. Therefore I'd like to
ask (in part also motivated by Andrew to do so) that you consider still
allowing in this series. In the (I think) not very likely case of problems
(first and foremost because of the changes coming from Linux, rather than
being homegrown) people always have the option of simply turning off the
use of this driver on their systems.

Thanks, Jan

Hi Jan,

> -----Original Message-----
> From: Jan Beulich <jbeulich@suse.com>
> Subject: [4.17?] Re: [PATCH v3 0/5] x86/mwait-idle: (remaining) SPR + (new)
> ADL support
> 
> Henry,
> 
> On 18.08.2022 15:02, Jan Beulich wrote:
> > New changes have appeared in the meantime, in particular one partly
> undoing
> > what we still haven't merged (patch 1 here).
> >
> > 1: add 'preferred_cstates' module argument
> > 2: add core C6 optimization for SPR
> > 3: add AlderLake support
> > 4: disable IBRS during long idle
> > 5: make SPR C1 and C1E be independent
> 
> just now acks have completed coming in for this series of ports of earlier
> Linux commits. This all targets newer hardware which of course we'd like
> to properly support with 4.17. Unfortunately the series was kind of lost
> with all the security work which had been ongoing. Therefore I'd like to
> ask (in part also motivated by Andrew to do so) that you consider still
> allowing in this series. In the (I think) not very likely case of problems
> (first and foremost because of the changes coming from Linux, rather than
> being homegrown) people always have the option of simply turning off the
> use of this driver on their systems.

I just checked the x86 gitlab board for the list and saw this series is on it. So
yes, please add my release ack. For now we still have 2 weeks before the release
and I think that would be a reasonable time for bugs to emerge.

To set expectations I am personally be more conservative in between the RC3
to RC4 week, because the fear of last minute bugs.

Kind regards,
Henry

> 
> Thanks, Jan