[Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Jason Cooper]

All,

Over the past couple of years, the ARM sub-arch maintainers have been
doing a great amount of work migrating to devicetree.

In conjunction with this effort, we've been slowly migrating the
existing legacy platforms (board files) over to devicetree.  For the
small part the mvebu maintainers are responsible for, we can see the
light at the end of the tunnel now. :-)

I imagine several other sub-arches are in a similar situation.  I think
it would be useful for us all to sit down, compare notes, and pool our
resources/experiences to coordinate a final push.

So, I'm proposing a session where each sub-arch gives a brief run-down
of the status of the legacy board conversion, and wraps up with a todo
list.  After all of the sub-arches have given their status (5 - 10
minutes each?), we hash out helping each other with the final pieces.

Also on the agenda, discussing the fate of the legacy sub-arches that
haven't even begun conversion to devicetree.

The intended outcome would be to have a sorted list of remaining
conversions, and who is pitching in to help out for each item.  Also, a
sorted list of unconverted sub-arches, broken down into 'deprecate',
'convert', and 'seek info'.

On a final note, I'd like to include a hobbyist voice or two in here.
Since a lot of the legacy platforms are supported by embedded Debian,
OpenWRT, and similar, getting an idea of which ones we can deprecate
would save quite a bit of work.

I haven't heard word of an ARM mini-summit yet.  If that happens,
obviously this topic would be most appropriate there.

Nominations:

Jason Cooper		(auto-nominated)
Andrew Lunn
Gregory Clement
Sebastian Hesselbarth	(auto-nominated)
Thomas Petazzoni	(auto-nominated)

I can think of several names from the ARM Linux mailinglist, but since I
can't create a conclusive list, I'll leave it up to the individuals to
pipe up if they are interested.

thx,

Jason.

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Jonathan Corbet]

On Sat, 10 May 2014 23:00:09 -0400
Jason Cooper <jason@lakedaemon.net> wrote:

> So, I'm proposing a session where each sub-arch gives a brief run-down
> of the status of the legacy board conversion, and wraps up with a todo
> list.  After all of the sub-arches have given their status (5 - 10
> minutes each?), we hash out helping each other with the final pieces.

This *really* looks like an ARM minisummit topic to me; hopefully one
of those is in the works?

jon

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Alexandre Belloni]

Hi,

On 11/05/2014 at 08:37:29 -0400, Jonathan Corbet wrote :
> On Sat, 10 May 2014 23:00:09 -0400
> Jason Cooper <jason@lakedaemon.net> wrote:
> 
> > So, I'm proposing a session where each sub-arch gives a brief run-down
> > of the status of the legacy board conversion, and wraps up with a todo
> > list.  After all of the sub-arches have given their status (5 - 10
> > minutes each?), we hash out helping each other with the final pieces.
> 
> This *really* looks like an ARM minisummit topic to me; hopefully one
> of those is in the works?
> 

It was supposed to take place at ELC, see:
http://lists.infradead.org/pipermail/linux-arm-kernel/2014-April/245388.html

I'm not sure we should try to organize a new ARM mini summit so soon
after that failed attempt.

Regards,

-- 
Alexandre Belloni, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Olof Johansson]

On Sun, May 11, 2014 at 5:37 AM, Jonathan Corbet <corbet@lwn.net> wrote:
> On Sat, 10 May 2014 23:00:09 -0400
> Jason Cooper <jason@lakedaemon.net> wrote:
>
>> So, I'm proposing a session where each sub-arch gives a brief run-down
>> of the status of the legacy board conversion, and wraps up with a todo
>> list.  After all of the sub-arches have given their status (5 - 10
>> minutes each?), we hash out helping each other with the final pieces.
>
> This *really* looks like an ARM minisummit topic to me;

Agreed.

> hopefully one of those is in the works?

Not this year.

Last year we had a two-day workshop, and most of it ended up centered
around DT. This year, that topic has overlap with other architectures
so it's likely better handled as shared technical track topic.

There doesn't seem to be a whole lot of topics needing to be covered,
at least not based on the call for proposals at ELC (see reply from
Alexandre). So, we'll take a break this year and come back and
organize one next year if there's more material again.

I suppose it's a positive sign; things are running relatively smoothly
right now so there's less stuff to iron out in person, or at least to
iron out together with a roomful of attendees.


-Olof

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Grant Likely]

On Sun, 11 May 2014 08:37:29 -0400, Jonathan Corbet <corbet@lwn.net> wrote:
> On Sat, 10 May 2014 23:00:09 -0400
> Jason Cooper <jason@lakedaemon.net> wrote:
> 
> > So, I'm proposing a session where each sub-arch gives a brief run-down
> > of the status of the legacy board conversion, and wraps up with a todo
> > list.  After all of the sub-arches have given their status (5 - 10
> > minutes each?), we hash out helping each other with the final pieces.
> 
> This *really* looks like an ARM minisummit topic to me; hopefully one
> of those is in the works?

It does, doesn't it? I wouldn't want this as a main ksummit topic.

The problem with ARM minisummits these days is it is very easy to
devolve into a nothing-but-dt meeting with a bunch of people sitting
around looking either annoyed or bored. We weren't able to pull enough
topics together when we tried to do an ARM minisummit at the ELC.

Instead of a traditional ARM minisummit, perhaps we should do an ARM
platforms minisprint instead. Light on any kind of presentations, but
have the right people in the room to try and knock out some of the
legacy backlog (which is kind of what Jason described)

g.

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Jason Cooper]

On Thu, May 15, 2014 at 01:15:46PM +0100, Grant Likely wrote:
> On Sun, 11 May 2014 08:37:29 -0400, Jonathan Corbet <corbet@lwn.net> wrote:
> > On Sat, 10 May 2014 23:00:09 -0400
> > Jason Cooper <jason@lakedaemon.net> wrote:
> > 
> > > So, I'm proposing a session where each sub-arch gives a brief run-down
> > > of the status of the legacy board conversion, and wraps up with a todo
> > > list.  After all of the sub-arches have given their status (5 - 10
> > > minutes each?), we hash out helping each other with the final pieces.
> > 
> > This *really* looks like an ARM minisummit topic to me; hopefully one
> > of those is in the works?
> 
> It does, doesn't it? I wouldn't want this as a main ksummit topic.
> 
> The problem with ARM minisummits these days is it is very easy to
> devolve into a nothing-but-dt meeting with a bunch of people sitting
> around looking either annoyed or bored. We weren't able to pull enough
> topics together when we tried to do an ARM minisummit at the ELC.
> 
> Instead of a traditional ARM minisummit, perhaps we should do an ARM
> platforms minisprint instead. Light on any kind of presentations, but
> have the right people in the room to try and knock out some of the
> legacy backlog (which is kind of what Jason described)

Agreed.  I think that would be an effective use of time and resources.

thx,

Jason.

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Rob Herring]

On Thu, May 15, 2014 at 7:15 AM, Grant Likely <grant.likely@secretlab.ca> wrote:
> On Sun, 11 May 2014 08:37:29 -0400, Jonathan Corbet <corbet@lwn.net> wrote:
>> On Sat, 10 May 2014 23:00:09 -0400
>> Jason Cooper <jason@lakedaemon.net> wrote:
>>
>> > So, I'm proposing a session where each sub-arch gives a brief run-down
>> > of the status of the legacy board conversion, and wraps up with a todo
>> > list.  After all of the sub-arches have given their status (5 - 10
>> > minutes each?), we hash out helping each other with the final pieces.
>>
>> This *really* looks like an ARM minisummit topic to me; hopefully one
>> of those is in the works?
>
> It does, doesn't it? I wouldn't want this as a main ksummit topic.
>
> The problem with ARM minisummits these days is it is very easy to
> devolve into a nothing-but-dt meeting with a bunch of people sitting
> around looking either annoyed or bored. We weren't able to pull enough
> topics together when we tried to do an ARM minisummit at the ELC.
>
> Instead of a traditional ARM minisummit, perhaps we should do an ARM
> platforms minisprint instead. Light on any kind of presentations, but
> have the right people in the room to try and knock out some of the
> legacy backlog (which is kind of what Jason described)

Isn't a large part of the backlog cases of we need DT bindings for X?
There are cases like moving platforms to common clk, but is there
anything to discuss for those? Most of those cases need bodies to work
on them. It seems like the rest of the todo lists could become an all
DT discussion. I'm not saying it shouldn't happen, just pointing out
where I think sprint discussions will go.

I do think having todo lists would be valuable. I would like to see
more than a one off list and have a living document. Linus W had a
great spreadsheet on this for multi-platform enablement. Perhaps this
would spur people to do some of the clean-ups (or it will just bit rot
:( ).

Rob

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Jason Cooper]

On Thu, May 15, 2014 at 01:50:02PM -0500, Rob Herring wrote:
> On Thu, May 15, 2014 at 7:15 AM, Grant Likely <grant.likely@secretlab.ca> wrote:
> > On Sun, 11 May 2014 08:37:29 -0400, Jonathan Corbet <corbet@lwn.net> wrote:
> >> On Sat, 10 May 2014 23:00:09 -0400
> >> Jason Cooper <jason@lakedaemon.net> wrote:
> >>
> >> > So, I'm proposing a session where each sub-arch gives a brief run-down
> >> > of the status of the legacy board conversion, and wraps up with a todo
> >> > list.  After all of the sub-arches have given their status (5 - 10
> >> > minutes each?), we hash out helping each other with the final pieces.
> >>
> >> This *really* looks like an ARM minisummit topic to me; hopefully one
> >> of those is in the works?
> >
> > It does, doesn't it? I wouldn't want this as a main ksummit topic.
> >
> > The problem with ARM minisummits these days is it is very easy to
> > devolve into a nothing-but-dt meeting with a bunch of people sitting
> > around looking either annoyed or bored. We weren't able to pull enough
> > topics together when we tried to do an ARM minisummit at the ELC.
> >
> > Instead of a traditional ARM minisummit, perhaps we should do an ARM
> > platforms minisprint instead. Light on any kind of presentations, but
> > have the right people in the room to try and knock out some of the
> > legacy backlog (which is kind of what Jason described)
> 
> Isn't a large part of the backlog cases of we need DT bindings for X?
> There are cases like moving platforms to common clk, but is there
> anything to discuss for those? Most of those cases need bodies to work
> on them. It seems like the rest of the todo lists could become an all
> DT discussion. I'm not saying it shouldn't happen, just pointing out
> where I think sprint discussions will go.

As gregkh mentioned for the staging tree:

$ ls arch/arm/*/TODO

would be helpful for folks who aren't familiar with an SoC, but can
easily generate patches.  imho, this is work for *after* the proposed
discussion.  We don't currently have a good idea which directories to
deprecate, convert, or leave to bit rot.  Once we do, I think the TODO
list would be helpful.

My main reason for raising this topic was the proximity to LinuxCon.
There's a better chance of getting some distro representation to give us
valuable feedback: "This SoC is quiet, no patches, but is actively used"
and "We thought the kernel guys had a reason for keeping that one."

eg: I know ixp4xx had an active community around it at one time for the
NSLU2, and the Gateworks boards used that SoC.  Are there people still
running it?  Well, my Dad is, but I can fix that if needed. ;-)

thx,

Jason.

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Olof Johansson]

Hi,

On Thu, May 15, 2014 at 12:42 PM, Jason Cooper <jason@lakedaemon.net> wrote:
> On Thu, May 15, 2014 at 01:50:02PM -0500, Rob Herring wrote:
>> On Thu, May 15, 2014 at 7:15 AM, Grant Likely <grant.likely@secretlab.ca> wrote:
>> > On Sun, 11 May 2014 08:37:29 -0400, Jonathan Corbet <corbet@lwn.net> wrote:
>> >> On Sat, 10 May 2014 23:00:09 -0400
>> >> Jason Cooper <jason@lakedaemon.net> wrote:
>> >>
>> >> > So, I'm proposing a session where each sub-arch gives a brief run-down
>> >> > of the status of the legacy board conversion, and wraps up with a todo
>> >> > list.  After all of the sub-arches have given their status (5 - 10
>> >> > minutes each?), we hash out helping each other with the final pieces.
>> >>
>> >> This *really* looks like an ARM minisummit topic to me; hopefully one
>> >> of those is in the works?
>> >
>> > It does, doesn't it? I wouldn't want this as a main ksummit topic.
>> >
>> > The problem with ARM minisummits these days is it is very easy to
>> > devolve into a nothing-but-dt meeting with a bunch of people sitting
>> > around looking either annoyed or bored. We weren't able to pull enough
>> > topics together when we tried to do an ARM minisummit at the ELC.
>> >
>> > Instead of a traditional ARM minisummit, perhaps we should do an ARM
>> > platforms minisprint instead. Light on any kind of presentations, but
>> > have the right people in the room to try and knock out some of the
>> > legacy backlog (which is kind of what Jason described)
>>
>> Isn't a large part of the backlog cases of we need DT bindings for X?
>> There are cases like moving platforms to common clk, but is there
>> anything to discuss for those? Most of those cases need bodies to work
>> on them. It seems like the rest of the todo lists could become an all
>> DT discussion. I'm not saying it shouldn't happen, just pointing out
>> where I think sprint discussions will go.
>
> As gregkh mentioned for the staging tree:
>
> $ ls arch/arm/*/TODO
>
> would be helpful for folks who aren't familiar with an SoC, but can
> easily generate patches.  imho, this is work for *after* the proposed
> discussion.  We don't currently have a good idea which directories to
> deprecate, convert, or leave to bit rot.  Once we do, I think the TODO
> list would be helpful.

Anyone should feel free to add a TODO file and list in their machine
directory at any given time. I don't think there's need for discussion
before that happens. There's also the elinux.org wikis if you want to
keep it out of the kernel -- keeping the TODO around for old kernel
versions might not be very useful so having it out of tree might be
just as fine.

> My main reason for raising this topic was the proximity to LinuxCon.
> There's a better chance of getting some distro representation to give us
> valuable feedback: "This SoC is quiet, no patches, but is actively used"
> and "We thought the kernel guys had a reason for keeping that one."

Funny, I would have thought ELC to have been a much more suitable
venue for that, especially since the older ARM platforms tend to be
embedded, and not generic compute ones. :)

> eg: I know ixp4xx had an active community around it at one time for the
> NSLU2, and the Gateworks boards used that SoC.  Are there people still
> running it?  Well, my Dad is, but I can fix that if needed. ;-)

I suspect it'll be hard to get that answered at LinuxCon, nor at an
ARM maintainer summit since the fringe platforms tend to not be
represented there either.


-Olof

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Jason Cooper]

On Thu, May 15, 2014 at 08:30:10PM -0700, Olof Johansson wrote:
> Hi,
> 
> On Thu, May 15, 2014 at 12:42 PM, Jason Cooper <jason@lakedaemon.net> wrote:
> > On Thu, May 15, 2014 at 01:50:02PM -0500, Rob Herring wrote:
> >> On Thu, May 15, 2014 at 7:15 AM, Grant Likely <grant.likely@secretlab.ca> wrote:
> >> > On Sun, 11 May 2014 08:37:29 -0400, Jonathan Corbet <corbet@lwn.net> wrote:
> >> >> On Sat, 10 May 2014 23:00:09 -0400
> >> >> Jason Cooper <jason@lakedaemon.net> wrote:
> >> >>
> >> >> > So, I'm proposing a session where each sub-arch gives a brief run-down
> >> >> > of the status of the legacy board conversion, and wraps up with a todo
> >> >> > list.  After all of the sub-arches have given their status (5 - 10
> >> >> > minutes each?), we hash out helping each other with the final pieces.
> >> >>
> >> >> This *really* looks like an ARM minisummit topic to me; hopefully one
> >> >> of those is in the works?
> >> >
> >> > It does, doesn't it? I wouldn't want this as a main ksummit topic.
> >> >
> >> > The problem with ARM minisummits these days is it is very easy to
> >> > devolve into a nothing-but-dt meeting with a bunch of people sitting
> >> > around looking either annoyed or bored. We weren't able to pull enough
> >> > topics together when we tried to do an ARM minisummit at the ELC.
> >> >
> >> > Instead of a traditional ARM minisummit, perhaps we should do an ARM
> >> > platforms minisprint instead. Light on any kind of presentations, but
> >> > have the right people in the room to try and knock out some of the
> >> > legacy backlog (which is kind of what Jason described)
> >>
> >> Isn't a large part of the backlog cases of we need DT bindings for X?
> >> There are cases like moving platforms to common clk, but is there
> >> anything to discuss for those? Most of those cases need bodies to work
> >> on them. It seems like the rest of the todo lists could become an all
> >> DT discussion. I'm not saying it shouldn't happen, just pointing out
> >> where I think sprint discussions will go.
> >
> > As gregkh mentioned for the staging tree:
> >
> > $ ls arch/arm/*/TODO
> >
> > would be helpful for folks who aren't familiar with an SoC, but can
> > easily generate patches.  imho, this is work for *after* the proposed
> > discussion.  We don't currently have a good idea which directories to
> > deprecate, convert, or leave to bit rot.  Once we do, I think the TODO
> > list would be helpful.
> 
> Anyone should feel free to add a TODO file and list in their machine
> directory at any given time. I don't think there's need for discussion
> before that happens. 

If there's a maintainer adding a todo file, then by default, that keeps
it out of bitrot and deprecation categories :)

> There's also the elinux.org wikis if you want to
> keep it out of the kernel -- keeping the TODO around for old kernel
> versions might not be very useful so having it out of tree might be
> just as fine.
> 
> > My main reason for raising this topic was the proximity to LinuxCon.
> > There's a better chance of getting some distro representation to give us
> > valuable feedback: "This SoC is quiet, no patches, but is actively used"
> > and "We thought the kernel guys had a reason for keeping that one."
> 
> Funny, I would have thought ELC to have been a much more suitable
> venue for that, especially since the older ARM platforms tend to be
> embedded, and not generic compute ones. :)

I agree, but I couldn't make it there this year :(

Also, when I initially submitted this proposal I didn't know if there
was going to be a ARM mini-summit or not.  If not, I thought this might
be worth discussing amongst the ARM folks for an hour or so.

> > eg: I know ixp4xx had an active community around it at one time for the
> > NSLU2, and the Gateworks boards used that SoC.  Are there people still
> > running it?  Well, my Dad is, but I can fix that if needed. ;-)
> 
> I suspect it'll be hard to get that answered at LinuxCon, nor at an
> ARM maintainer summit since the fringe platforms tend to not be
> represented there either.

Ok, I assumed *everybody* in the Linux ecosystem went to LinuxCon.  I've
never actually been to one, but that museum was really crowded in
Edinburgh.  I figured that _must_ be everybody. :)

Joking aside, I concede your point.  It looks like the last two mvebu
legacy platforms (dove, mv78xx0) might get merged into mach-mvebu during
the next cycle.  After that, we can reach out to the Debian and OpenWRT
guys and see what legacy platforms are in use or rotting.  We can go
from there.

thx,

Jason.

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Ben Hutchings]

[-- Attachment #1: Type: text/plain, Size: 1127 bytes --]

On Fri, 2014-05-16 at 00:10 -0400, Jason Cooper wrote:
[...]
> Ok, I assumed *everybody* in the Linux ecosystem went to LinuxCon.  I've
> never actually been to one, but that museum was really crowded in
> Edinburgh.  I figured that _must_ be everybody. :)

I find it rather less useful than Plumbers'.

> Joking aside, I concede your point.  It looks like the last two mvebu
> legacy platforms (dove, mv78xx0) might get merged into mach-mvebu during
> the next cycle.  After that, we can reach out to the Debian and OpenWRT
> guys and see what legacy platforms are in use or rotting.  We can go
> from there.

I just removed ixp4xx from Debian as the kernel image wouldn't fit in
the NSLU2's kernel partition any more (even after disabling quite a few
features that are enabled in most other configurations).  iop32x was
removed for the same reason, a while ago.

Our 'legacy' (which I take to mean pre-v7) configurations are now
kirkwood, mv78xx0, orion5x (all about to converge to mvebu) and
versatile.

Ben.

-- 
Ben Hutchings
If you seem to know what you are doing, you'll be given more to do.

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 828 bytes --]

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Jason Cooper]

On Fri, May 16, 2014 at 07:31:51PM +0100, Ben Hutchings wrote:
> On Fri, 2014-05-16 at 00:10 -0400, Jason Cooper wrote:
> > Joking aside, I concede your point.  It looks like the last two mvebu
> > legacy platforms (dove, mv78xx0) might get merged into mach-mvebu during
> > the next cycle.  After that, we can reach out to the Debian and OpenWRT
> > guys and see what legacy platforms are in use or rotting.  We can go
> > from there.
> 
> I just removed ixp4xx from Debian as the kernel image wouldn't fit in
> the NSLU2's kernel partition any more (even after disabling quite a few
> features that are enabled in most other configurations).  iop32x was
> removed for the same reason, a while ago.

Hmmm, I have two NSLU2's, they might make a good platform for the
tinification/IoT work.  Small flash, small RAM, slow.

> Our 'legacy' (which I take to mean pre-v7) configurations are now
> kirkwood, mv78xx0, orion5x (all about to converge to mvebu) and
> versatile.

I'm not familiar with versatile, the others I've seen a few times. ;-)

Does Debian see any user activity with mv78xx0?  I know your build
system used to have boards with that SoC family, but are there any other
users?

thx,

Jason.

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Ben Hutchings]

[-- Attachment #1: Type: text/plain, Size: 1856 bytes --]

On Fri, 2014-05-16 at 14:42 -0400, Jason Cooper wrote:
> On Fri, May 16, 2014 at 07:31:51PM +0100, Ben Hutchings wrote:
> > On Fri, 2014-05-16 at 00:10 -0400, Jason Cooper wrote:
> > > Joking aside, I concede your point.  It looks like the last two mvebu
> > > legacy platforms (dove, mv78xx0) might get merged into mach-mvebu during
> > > the next cycle.  After that, we can reach out to the Debian and OpenWRT
> > > guys and see what legacy platforms are in use or rotting.  We can go
> > > from there.
> > 
> > I just removed ixp4xx from Debian as the kernel image wouldn't fit in
> > the NSLU2's kernel partition any more (even after disabling quite a few
> > features that are enabled in most other configurations).  iop32x was
> > removed for the same reason, a while ago.
> 
> Hmmm, I have two NSLU2's, they might make a good platform for the
> tinification/IoT work.  Small flash, small RAM, slow.

Yes, whereas Debian is a general purpose distribution where we don't
attempt to guess which applications will be used on which platforms.

> > Our 'legacy' (which I take to mean pre-v7) configurations are now
> > kirkwood, mv78xx0, orion5x (all about to converge to mvebu) and
> > versatile.
> 
> I'm not familiar with versatile, the others I've seen a few times. ;-)

Versatile is one of ARM's own development boards, but more importantly
it's something QEMU can emulate.

> Does Debian see any user activity with mv78xx0?  I know your build
> system used to have boards with that SoC family, but are there any other
> users?

There is still one build server (ancina.debian.org) using that board,
but I don't know that there are any other users.  The kernel package was
only added to support those build servers.

Ben.

-- 
Ben Hutchings
If you seem to know what you are doing, you'll be given more to do.

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 828 bytes --]

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Arnd Bergmann]

On Friday 16 May 2014 23:20:42 Ben Hutchings wrote:
> On Fri, 2014-05-16 at 14:42 -0400, Jason Cooper wrote:
> > On Fri, May 16, 2014 at 07:31:51PM +0100, Ben Hutchings wrote:
> > > Our 'legacy' (which I take to mean pre-v7) configurations are now
> > > kirkwood, mv78xx0, orion5x (all about to converge to mvebu) and
> > > versatile.
> > 
> > I'm not familiar with versatile, the others I've seen a few times. 
> 
> Versatile is one of ARM's own development boards, but more importantly
> it's something QEMU can emulate.

We should really have this in multiplatform soon, I think the last issue
for versatile is common clock support, but Linus Walleij may have more
details about what remains here.

We should also make it possible to enable ARCH_VIRT for pre-ARMv7 to
get qemu users to move over to that.

	Arnd

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Peter Maydell]

On 19 May 2014 13:20, Arnd Bergmann <arnd@arndb.de> wrote:
> On Friday 16 May 2014 23:20:42 Ben Hutchings wrote:
>> Versatile is one of ARM's own development boards, but more importantly
>> it's something QEMU can emulate.
>
> We should really have this in multiplatform soon, I think the last issue
> for versatile is common clock support, but Linus Walleij may have more
> details about what remains here.
>
> We should also make it possible to enable ARCH_VIRT for pre-ARMv7 to
> get qemu users to move over to that.

I think the main reason many people still use QEMU's versatilepb
model is that it has PCI. If you want people to move off that then
getting those "generic device-tree driven PCI" patches into the
kernel would probably be a good start :-)  At that point we can
reasonably easily add PCI support to the QEMU 'virt' board.

thanks
-- PMM

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Rob Herring]

On Mon, May 19, 2014 at 7:20 AM, Arnd Bergmann <arnd@arndb.de> wrote:
> On Friday 16 May 2014 23:20:42 Ben Hutchings wrote:
>> On Fri, 2014-05-16 at 14:42 -0400, Jason Cooper wrote:
>> > On Fri, May 16, 2014 at 07:31:51PM +0100, Ben Hutchings wrote:
>> > > Our 'legacy' (which I take to mean pre-v7) configurations are now
>> > > kirkwood, mv78xx0, orion5x (all about to converge to mvebu) and
>> > > versatile.
>> >
>> > I'm not familiar with versatile, the others I've seen a few times.
>>
>> Versatile is one of ARM's own development boards, but more importantly
>> it's something QEMU can emulate.
>
> We should really have this in multiplatform soon, I think the last issue
> for versatile is common clock support, but Linus Walleij may have more
> details about what remains here.

While not necessarily the same thing, but does move versatile in the
right direction, I've got patches in progress which fully move
versatile over to DT. The main issues there are:

common clock
irq cntlr init from DT
Move PCI to DT and platform driver

I'd like to do this all in one step to avoid doing things like non-DT
common clock init. It's blocked on the common DT PCI code and my free
time.

There are then a few bits related to mach/plat headers needed move
versatile to multiplatform. I think these issues are mostly the same
changes that have been done or are needed on integrator which is what
Linus has been working on.

Rob

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Arnd Bergmann]

On Monday 19 May 2014 18:06:02 Rob Herring wrote:
> > On Friday 16 May 2014 23:20:42 Ben Hutchings wrote:
> >> On Fri, 2014-05-16 at 14:42 -0400, Jason Cooper wrote:
> >> > On Fri, May 16, 2014 at 07:31:51PM +0100, Ben Hutchings wrote:
> >> > > Our 'legacy' (which I take to mean pre-v7) configurations are now
> >> > > kirkwood, mv78xx0, orion5x (all about to converge to mvebu) and
> >> > > versatile.
> >> >
> >> > I'm not familiar with versatile, the others I've seen a few times.
> >>
> >> Versatile is one of ARM's own development boards, but more importantly
> >> it's something QEMU can emulate.
> >
> > We should really have this in multiplatform soon, I think the last issue
> > for versatile is common clock support, but Linus Walleij may have more
> > details about what remains here.
> 
> While not necessarily the same thing, but does move versatile in the
> right direction, I've got patches in progress which fully move
> versatile over to DT. The main issues there are:
> 
> common clock
> irq cntlr init from DT
> Move PCI to DT and platform driver
> 
> I'd like to do this all in one step to avoid doing things like non-DT
> common clock init. 

Sounds good, yes.

> It's blocked on the common DT PCI code and my free
> time.
> 
> There are then a few bits related to mach/plat headers needed move
> versatile to multiplatform. I think these issues are mostly the same
> changes that have been done or are needed on integrator which is what
> Linus has been working on.

Ah, right. I see these two includes from mach-versatile to plat/*.h headers:

arch/arm/mach-versatile/core.c:#include <plat/clcd.h>
arch/arm/mach-versatile/core.c:#include <plat/sched_clock.h>

We can deal with those by adding

ccflags-y := -I$(srctree)/arch/arm/plat-versatile/include

to mach-versatile/Makefile, or we can rearrange the code. The opposite
direction would be harder, but I don't see any remaining dependencies
there.

	Arnd

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Rob Herring]

On Tue, May 20, 2014 at 3:08 AM, Arnd Bergmann <arnd@arndb.de> wrote:
> On Monday 19 May 2014 18:06:02 Rob Herring wrote:
>> > On Friday 16 May 2014 23:20:42 Ben Hutchings wrote:
>> >> On Fri, 2014-05-16 at 14:42 -0400, Jason Cooper wrote:
>> >> > On Fri, May 16, 2014 at 07:31:51PM +0100, Ben Hutchings wrote:
>> >> > > Our 'legacy' (which I take to mean pre-v7) configurations are now
>> >> > > kirkwood, mv78xx0, orion5x (all about to converge to mvebu) and
>> >> > > versatile.
>> >> >
>> >> > I'm not familiar with versatile, the others I've seen a few times.
>> >>
>> >> Versatile is one of ARM's own development boards, but more importantly
>> >> it's something QEMU can emulate.
>> >
>> > We should really have this in multiplatform soon, I think the last issue
>> > for versatile is common clock support, but Linus Walleij may have more
>> > details about what remains here.
>>
>> While not necessarily the same thing, but does move versatile in the
>> right direction, I've got patches in progress which fully move
>> versatile over to DT. The main issues there are:
>>
>> common clock
>> irq cntlr init from DT
>> Move PCI to DT and platform driver
>>
>> I'd like to do this all in one step to avoid doing things like non-DT
>> common clock init.
>
> Sounds good, yes.
>
>> It's blocked on the common DT PCI code and my free
>> time.
>>
>> There are then a few bits related to mach/plat headers needed move
>> versatile to multiplatform. I think these issues are mostly the same
>> changes that have been done or are needed on integrator which is what
>> Linus has been working on.
>
> Ah, right. I see these two includes from mach-versatile to plat/*.h headers:
>
> arch/arm/mach-versatile/core.c:#include <plat/clcd.h>
> arch/arm/mach-versatile/core.c:#include <plat/sched_clock.h>
>
> We can deal with those by adding
>
> ccflags-y := -I$(srctree)/arch/arm/plat-versatile/include
>
> to mach-versatile/Makefile, or we can rearrange the code. The opposite
> direction would be harder, but I don't see any remaining dependencies
> there.

It looked like more work when I started this, but thanks to Linus's
work on plat dir, it is pretty straight-forward to finish the
conversion. Here's a branch that enables multi-platform for versatile:

git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux.git
versatile-multiplatform

So it is PCI and LEDs that have to be sorted out first.

Rob

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Linus Walleij]

On Tue, May 20, 2014 at 10:08 AM, Arnd Bergmann <arnd@arndb.de> wrote:

> Ah, right. I see these two includes from mach-versatile to plat/*.h headers:
>
> arch/arm/mach-versatile/core.c:#include <plat/clcd.h>

I have a patch for this one.

> arch/arm/mach-versatile/core.c:#include <plat/sched_clock.h>

In the Integrator I solved this by inlining that code into the
machine. After the introduction of sched_clock_register()
this becomes very few lines of code really.

Yours,
Linus Walleij

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Arnd Bergmann]

On Friday 23 May 2014 15:48:17 Linus Walleij wrote:
> On Tue, May 20, 2014 at 10:08 AM, Arnd Bergmann <arnd@arndb.de> wrote:
> 
> > Ah, right. I see these two includes from mach-versatile to plat/*.h headers:
> >
> > arch/arm/mach-versatile/core.c:#include <plat/clcd.h>
> 
> I have a patch for this one.
> 
> > arch/arm/mach-versatile/core.c:#include <plat/sched_clock.h>
> 
> In the Integrator I solved this by inlining that code into the
> machine. After the introduction of sched_clock_register()
> this becomes very few lines of code really.

Ah, very good.

	Arnd

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Linus Walleij]

On Tue, May 20, 2014 at 1:06 AM, Rob Herring <robh@kernel.org> wrote:
> On Mon, May 19, 2014 at 7:20 AM, Arnd Bergmann <arnd@arndb.de> wrote:

>> We should really have this in multiplatform soon, I think the last issue
>> for versatile is common clock support, but Linus Walleij may have more
>> details about what remains here.
>
> While not necessarily the same thing, but does move versatile in the
> right direction, I've got patches in progress which fully move
> versatile over to DT.

Doing this two-stage rocket:
Legacy -> DT -> multiplatform

Is usually the smoothest. That is what I did on U300, U8500
and will do on the Integrator.

> I'd like to do this all in one step to avoid doing things like non-DT
> common clock init. It's blocked on the common DT PCI code and my free
> time.

I agree. And invariably this will confuse a lot of QEMU users
until they figure out how to boot their new image with a device
tree. But that is business as usual I guess.

> There are then a few bits related to mach/plat headers needed move
> versatile to multiplatform. I think these issues are mostly the same
> changes that have been done or are needed on integrator which is what
> Linus has been working on.

I'm onto these as part of both Integrator and RealView modernization.

Yours,
Linus Walleij

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Linus Walleij]

On Thu, May 15, 2014 at 8:50 PM, Rob Herring <robh@kernel.org> wrote:

> I do think having todo lists would be valuable. I would like to see
> more than a one off list and have a living document. Linus W had a
> great spreadsheet on this for multi-platform enablement. Perhaps this
> would spur people to do some of the clean-ups (or it will just bit rot
> :( ).

It's still there:
https://docs.google.com/spreadsheet/ccc?key=0Aj_PQh-9xwkMdHRHYzRzOEhyMEt5b3UtOFIwcHRpWEE

However I don't think there is much to discuss really. It's just work,
we know how to do it.

The real problem is that of somebody doing the work, and when maintainers
don't want to do the work, how to proceed by leaving the platform as-is
(legacy), delete it (lacking active maintainer) or have some custodian
like me acquire the hardware and fix it up myself.

Arnd and I have discussed this and we basically think that the StrongARM
stuff (RISC PC, Netwinder, SA1100-based handhelds etc) are legacy
systems while all ARMv4T and later are target for modernization if
and only if an active maintainer is available to pursue the work.

Yours,
Linus Walleij

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Arnd Bergmann]

On Friday 23 May 2014 15:40:15 Linus Walleij wrote:
> On Thu, May 15, 2014 at 8:50 PM, Rob Herring <robh@kernel.org> wrote:
> 
> > I do think having todo lists would be valuable. I would like to see
> > more than a one off list and have a living document. Linus W had a
> > great spreadsheet on this for multi-platform enablement. Perhaps this
> > would spur people to do some of the clean-ups (or it will just bit rot
> >  ).
> 
> It's still there:
> https://docs.google.com/spreadsheet/ccc?key=0Aj_PQh-9xwkMdHRHYzRzOEhyMEt5b3UtOFIwcHRpWEE
> 
> However I don't think there is much to discuss really. It's just work,
> we know how to do it.
> 
> The real problem is that of somebody doing the work, and when maintainers
> don't want to do the work, how to proceed by leaving the platform as-is
> (legacy), delete it (lacking active maintainer) or have some custodian
> like me acquire the hardware and fix it up myself.
> 
> Arnd and I have discussed this and we basically think that the StrongARM
> stuff (RISC PC, Netwinder, SA1100-based handhelds etc) are legacy
> systems while all ARMv4T and later are target for modernization if
> and only if an active maintainer is available to pursue the work.

I was also including the XScale based platforms (ixp, iop, pxa) in the
same category as StrongARM, although PXA sticks out a bit as there is new
development going on for that platform.

	Arnd

Re: [Ksummit-discuss] [TECH TOPIC] ARM legacy board DT conversion finalization

[Olof Johansson]

On Thu, May 15, 2014 at 5:15 AM, Grant Likely <grant.likely@secretlab.ca> wrote:
> On Sun, 11 May 2014 08:37:29 -0400, Jonathan Corbet <corbet@lwn.net> wrote:
>> On Sat, 10 May 2014 23:00:09 -0400
>> Jason Cooper <jason@lakedaemon.net> wrote:
>>
>> > So, I'm proposing a session where each sub-arch gives a brief run-down
>> > of the status of the legacy board conversion, and wraps up with a todo
>> > list.  After all of the sub-arches have given their status (5 - 10
>> > minutes each?), we hash out helping each other with the final pieces.
>>
>> This *really* looks like an ARM minisummit topic to me; hopefully one
>> of those is in the works?
>
> It does, doesn't it? I wouldn't want this as a main ksummit topic.
>
> The problem with ARM minisummits these days is it is very easy to
> devolve into a nothing-but-dt meeting with a bunch of people sitting
> around looking either annoyed or bored. We weren't able to pull enough
> topics together when we tried to do an ARM minisummit at the ELC.
>
> Instead of a traditional ARM minisummit, perhaps we should do an ARM
> platforms minisprint instead. Light on any kind of presentations, but
> have the right people in the room to try and knock out some of the
> legacy backlog (which is kind of what Jason described)

I'm not entirely sold on the practicality of this. So many of the
legacy platforms need some random piece of hardware to test on, and
having people drag that with them to Chicago seems awkward.

But I'm willing to entertain the thought and see where we end up. We
would need a rough todo list and a list of people that would need to
be in the room for the thing to be productive.

(I wonder if it would make more sense to colocate something like it
with the Linaro Connect in California this fall instead).


-Olof