kSummit proposal - The Linux kernel in ChromeOS

kSummit proposal - The Linux kernel in ChromeOS

[Alex Levin]

Hi Linux friends,

Jesse Barnes, Chrome OS baseOS (Firmware+Kernel) lead, and myself
would like to present the current state of affairs of the Linux kernel
on ChromeOS and the challenges we face, how we solve them and get your
feedback.  We can also talk about how our efforts can help upstream
development, for example by running experiments in the field to
compare approaches to a specific problem or area.

Shipping ChromeOS to millions of users that span across hundreds of
different platforms, multiple active kernel versions and across many
different SoC architectures, introduces interesting challenges:

- Testing the upstream RC on as many platforms as we can as early as we can.
- Updating the Linux kernel on existing platforms (millions of users at a time).
- Managing technical “debt” and keeping the ChromeOS kernel as close
as possible to the upstream kernel.
- Current pain points in dealing with upstream.

We feel 45-60 minutes would be enough and will allow a discussion.

Thanks a lot in advance,

Alex Levin,
ChromeOS platform tech lead.

Re: kSummit proposal - The Linux kernel in ChromeOS

[Linus Walleij]

On Tue, Aug 24, 2021 at 3:06 AM Alex Levin <levinale@google.com> wrote:

> - Testing the upstream RC on as many platforms as we can as early as we can.
> - Updating the Linux kernel on existing platforms (millions of users at a time).
> - Managing technical “debt” and keeping the ChromeOS kernel as close
> as possible to the upstream kernel.
> - Current pain points in dealing with upstream.
>
> We feel 45-60 minutes would be enough and will allow a discussion.

I think this is really interesting.

Also explain what this stuff is: drivers/platform/chrome/* and why it is
so necessarily different and can't live in existing subsystems?

Also: your experience with using ACPI and when firmware can be
fixed and when it can not, and when/if the kernel has to accommodate
quirks rather than fixing the firmware.

Yours,
Linus Walleij

Re: kSummit proposal - The Linux kernel in ChromeOS

[Laurent Pinchart]

On Fri, Aug 27, 2021 at 12:07:56AM +0200, Linus Walleij wrote:
> On Tue, Aug 24, 2021 at 3:06 AM Alex Levin wrote:
> 
> > - Testing the upstream RC on as many platforms as we can as early as we can.
> > - Updating the Linux kernel on existing platforms (millions of users at a time).
> > - Managing technical “debt” and keeping the ChromeOS kernel as close
> > as possible to the upstream kernel.
> > - Current pain points in dealing with upstream.
> >
> > We feel 45-60 minutes would be enough and will allow a discussion.
> 
> I think this is really interesting.
> 
> Also explain what this stuff is: drivers/platform/chrome/* and why it is
> so necessarily different and can't live in existing subsystems?
> 
> Also: your experience with using ACPI and when firmware can be
> fixed and when it can not, and when/if the kernel has to accommodate
> quirks rather than fixing the firmware.

Could this also include lessons learnt related to convincing and helping
SoC vendors to upstream drivers ?

-- 
Regards,

Laurent Pinchart

Re: kSummit proposal - The Linux kernel in ChromeOS

[Alex Levin]

Thanks for the feedback Linus and Laurent.

We are willing to extend this to 90 minutes and cover the
opportunities and challenges we have WRT to SoC manufacturers
upstreaming and how ChromeOS is influencing upstreaming.

I wasn't sure we will be accepted as a ksummit so it seems that the
track for now is a MC, but that might change.

> Also: your experience with using ACPI and when firmware can be
> fixed and when it can not, and when/if the kernel has to accommodate
> quirks rather than fixing the firmware.

We can definitely touch on this topic. It might be too narrow of a
topic to discuss widely (I would gladly have a virtual cup of coffee),
but we will include in our talk the FW challenges we see and how they
impact kernel deployment, development and uprev. It goes a tad above
(or beyond?) quirks.

> Also explain what this stuff is: drivers/platform/chrome/* and why it is
> so necessarily different and can't live in existing subsystems?

This is our cros_ec (Chrome OS embedded controller) driver


On Thu, Aug 26, 2021 at 3:27 PM Laurent Pinchart
<laurent.pinchart@ideasonboard.com> wrote:
>
> On Fri, Aug 27, 2021 at 12:07:56AM +0200, Linus Walleij wrote:
> > On Tue, Aug 24, 2021 at 3:06 AM Alex Levin wrote:
> >
> > > - Testing the upstream RC on as many platforms as we can as early as we can.
> > > - Updating the Linux kernel on existing platforms (millions of users at a time).
> > > - Managing technical “debt” and keeping the ChromeOS kernel as close
> > > as possible to the upstream kernel.
> > > - Current pain points in dealing with upstream.
> > >
> > > We feel 45-60 minutes would be enough and will allow a discussion.
> >
> > I think this is really interesting.
> >
> > Also explain what this stuff is: drivers/platform/chrome/* and why it is
> > so necessarily different and can't live in existing subsystems?
> >
> > Also: your experience with using ACPI and when firmware can be
> > fixed and when it can not, and when/if the kernel has to accommodate
> > quirks rather than fixing the firmware.
>
> Could this also include lessons learnt related to convincing and helping
> SoC vendors to upstream drivers ?
>
> --
> Regards,
>
> Laurent Pinchart



-- 
-Alex

Re: kSummit proposal - The Linux kernel in ChromeOS

[Serge E. Hallyn]

On Fri, Aug 27, 2021 at 01:27:10AM +0300, Laurent Pinchart wrote:
> On Fri, Aug 27, 2021 at 12:07:56AM +0200, Linus Walleij wrote:
> > On Tue, Aug 24, 2021 at 3:06 AM Alex Levin wrote:
> > 
> > > - Testing the upstream RC on as many platforms as we can as early as we can.
> > > - Updating the Linux kernel on existing platforms (millions of users at a time).
> > > - Managing technical “debt” and keeping the ChromeOS kernel as close
> > > as possible to the upstream kernel.
> > > - Current pain points in dealing with upstream.
> > >
> > > We feel 45-60 minutes would be enough and will allow a discussion.
> > 
> > I think this is really interesting.
> > 
> > Also explain what this stuff is: drivers/platform/chrome/* and why it is
> > so necessarily different and can't live in existing subsystems?
> > 
> > Also: your experience with using ACPI and when firmware can be
> > fixed and when it can not, and when/if the kernel has to accommodate
> > quirks rather than fixing the firmware.
> 
> Could this also include lessons learnt related to convincing and helping
> SoC vendors to upstream drivers ?

That sounds awesome.

Re: kSummit proposal - The Linux kernel in ChromeOS

[Linus Walleij]

On Fri, Aug 27, 2021 at 3:01 AM Alex Levin <levinale@google.com> wrote:

> > Also: your experience with using ACPI and when firmware can be
> > fixed and when it can not, and when/if the kernel has to accommodate
> > quirks rather than fixing the firmware.
>
> We can definitely touch on this topic. It might be too narrow of a
> topic to discuss widely (I would gladly have a virtual cup of coffee),

This has been and remains a really big and recurring problem
for drivers across the whole board of devices using ACPI, so I bet
more people are interested.

Yours,
Linus Walleij

Re: kSummit proposal - The Linux kernel in ChromeOS

[Theodore Ts'o]

On Thu, Aug 26, 2021 at 06:01:04PM -0700, Alex Levin wrote:
> Thanks for the feedback Linus and Laurent.
> 
> We are willing to extend this to 90 minutes and cover the
> opportunities and challenges we have WRT to SoC manufacturers
> upstreaming and how ChromeOS is influencing upstreaming.
> 
> I wasn't sure we will be accepted as a ksummit so it seems that the
> track for now is a MC, but that might change.

Kernel Summit sessions are only 45 minutes, but we do have plenty of
time for BOF's.

What MC track were you proposing this for?

					- Ted

Re: kSummit proposal - The Linux kernel in ChromeOS

[Jonathan Cameron]

On Fri, 27 Aug 2021 22:22:15 +0200
Linus Walleij <linus.walleij@linaro.org> wrote:

> On Fri, Aug 27, 2021 at 3:01 AM Alex Levin <levinale@google.com> wrote:
> 
> > > Also: your experience with using ACPI and when firmware can be
> > > fixed and when it can not, and when/if the kernel has to accommodate
> > > quirks rather than fixing the firmware.  
> >
> > We can definitely touch on this topic. It might be too narrow of a
> > topic to discuss widely (I would gladly have a virtual cup of coffee),  
> 
> This has been and remains a really big and recurring problem
> for drivers across the whole board of devices using ACPI, so I bet
> more people are interested.
> 
> Yours,
> Linus Walleij
> 
Indeed. I'm interested (in this and the proposal in general!)

Jonathan

Re: kSummit proposal - The Linux kernel in ChromeOS

[Theodore Ts'o]

On Thu, Aug 26, 2021 at 06:01:04PM -0700, Alex Levin wrote:
> Thanks for the feedback Linus and Laurent.
> 
> We are willing to extend this to 90 minutes and cover the
> opportunities and challenges we have WRT to SoC manufacturers
> upstreaming and how ChromeOS is influencing upstreaming.
> 
> I wasn't sure we will be accepted as a ksummit so it seems that the
> track for now is a MC, but that might change.

I've added the Performance and Scalability MC runners to this thread.

After taking a closer look at the proposal submitted to the
Performance MC:

   https://linuxplumbersconf.org/event/11/abstracts/1042/

... I noted that (a) the proposal wasn't about scalability in the
traditional sense (e.g., how well does the kernel work on CPU's with
192 cores) but rather scalability of the development process, and (b)
the talk was only scheduled for 20 minutes.

So I talked to Alex off-line, and we agreed that best approach was
probably to transfer his session transferred to the kernel summit
track, where he would be able to get a 45 minute slot --- assuming, of
course, that the Performance MC runners would be OK with this change.

If people would like to talk more than what can fit in a 45 minute
slot, there will be BOF session slots on Friday that people can
request for more conversation.  There will also be "hack rooms" that
are available, which will not be live-streamed nor permanently
recorded.

Dave, Daniel, Pasha --- does this seem reasonable to you?

Thanks,

						- Ted

Re: kSummit proposal - The Linux kernel in ChromeOS

[Davidlohr Bueso]

(resending in plain text, sorry).

This is fine by me, and we had noted that this proposal was not about
performance in the conventional way, so being in ksummit track makes
sense.

Thanks,
Davidlohr


On Mon, Aug 30, 2021 at 1:44 PM Davidlohr Bueso <dave.bueso@gmail.com> wrote:
>
> This is fine by me, and we had noted that this proposal was not about performance in the conventional way, so being in ksummit track makes sense.
>
> Thanks,
> Davidlohr
>
> On Mon, Aug 30, 2021 at 12:33 PM Theodore Ts'o <tytso@mit.edu> wrote:
>>
>> On Thu, Aug 26, 2021 at 06:01:04PM -0700, Alex Levin wrote:
>> > Thanks for the feedback Linus and Laurent.
>> >
>> > We are willing to extend this to 90 minutes and cover the
>> > opportunities and challenges we have WRT to SoC manufacturers
>> > upstreaming and how ChromeOS is influencing upstreaming.
>> >
>> > I wasn't sure we will be accepted as a ksummit so it seems that the
>> > track for now is a MC, but that might change.
>>
>> I've added the Performance and Scalability MC runners to this thread.
>>
>> After taking a closer look at the proposal submitted to the
>> Performance MC:
>>
>>    https://linuxplumbersconf.org/event/11/abstracts/1042/
>>
>> ... I noted that (a) the proposal wasn't about scalability in the
>> traditional sense (e.g., how well does the kernel work on CPU's with
>> 192 cores) but rather scalability of the development process, and (b)
>> the talk was only scheduled for 20 minutes.
>>
>> So I talked to Alex off-line, and we agreed that best approach was
>> probably to transfer his session transferred to the kernel summit
>> track, where he would be able to get a 45 minute slot --- assuming, of
>> course, that the Performance MC runners would be OK with this change.
>>
>> If people would like to talk more than what can fit in a 45 minute
>> slot, there will be BOF session slots on Friday that people can
>> request for more conversation.  There will also be "hack rooms" that
>> are available, which will not be live-streamed nor permanently
>> recorded.
>>
>> Dave, Daniel, Pasha --- does this seem reasonable to you?
>>
>> Thanks,
>>
>>                                                 - Ted

Re: kSummit proposal - The Linux kernel in ChromeOS

[Daniel Jordan]

On Mon, Aug 30, 2021 at 01:49:43PM -0700, Davidlohr Bueso wrote:
> This is fine by me, and we had noted that this proposal was not about
> performance in the conventional way, so being in ksummit track makes
> sense.

Agreed, fine by me too.

Re: kSummit proposal - The Linux kernel in ChromeOS

[Pavel Tatashin]

Sounds good, to me.

On Mon, Aug 30, 2021 at 5:05 PM Daniel Jordan
<daniel.m.jordan@oracle.com> wrote:
>
> On Mon, Aug 30, 2021 at 01:49:43PM -0700, Davidlohr Bueso wrote:
> > This is fine by me, and we had noted that this proposal was not about
> > performance in the conventional way, so being in ksummit track makes
> > sense.
>
> Agreed, fine by me too.