Renaming MACH_TYPE_KINETIS_K60 into MACH_TYPE_KINETIS

Renaming MACH_TYPE_KINETIS_K60 into MACH_TYPE_KINETIS

[Alexander Potashev]

Hi,

About three weeks ago I registered a machine type for the Freescale
Kinetis K60 microcontroller. ( http://www.arm.linux.org.uk/developer/machines/list.php?id=3896 )

But later I realized that there should be a single machine type for all
Freescale Kinetis MCUs (K10 through K70), because those MCUs are
software-compatible. I tried to rename "kinetis_k60" to just "kinetis"
at the page mentioned above, but the names of the CONFIG_* macro and the
identifier name did not change.

I really do not want to register another machine type since it would be
a mess having two machine types - one for Kinetis and another for
Kinetis K60. Is it possible to change CONFIG_MACH_KINETIS_K60 to
CONFIG_MACH_KINETIS and MACH_TYPE_KINETIS_K60 to MACH_TYPE_KINETIS in
the database at http://www.arm.linux.org.uk/developer/machines/ ?

I hope, no one has really started using this machine type yet, so the
change is going to be transparent for everyone.


-- 
Regards,
Alexander

Renaming MACH_TYPE_KINETIS_K60 into MACH_TYPE_KINETIS

[Marek Vasut]

> Hi,
> 
> About three weeks ago I registered a machine type for the Freescale
> Kinetis K60 microcontroller. (
> http://www.arm.linux.org.uk/developer/machines/list.php?id=3896 )
> 
> But later I realized that there should be a single machine type for all
> Freescale Kinetis MCUs (K10 through K70), because those MCUs are
> software-compatible. I tried to rename "kinetis_k60" to just "kinetis"
> at the page mentioned above, but the names of the CONFIG_* macro and the
> identifier name did not change.
> 
> I really do not want to register another machine type since it would be
> a mess having two machine types - one for Kinetis and another for
> Kinetis K60. Is it possible to change CONFIG_MACH_KINETIS_K60 to
> CONFIG_MACH_KINETIS and MACH_TYPE_KINETIS_K60 to MACH_TYPE_KINETIS in
> the database at http://www.arm.linux.org.uk/developer/machines/ ?
> 
> I hope, no one has really started using this machine type yet, so the
> change is going to be transparent for everyone.

If you use FDT, you won't need machine ID at all.

M

Renaming MACH_TYPE_KINETIS_K60 into MACH_TYPE_KINETIS

[Alexander Potashev]

On Wed, 2011-12-28 at 11:00 +0100, Marek Vasut wrote:
> If you use FDT, you won't need machine ID at all. 

Hi Marek,

You are right, but we have not used FDT for any of our ARM Linux ports
yet (consistency matters.) If we switch to FDT now, a lot of code will
need to be rewritten.

That is why I would prefer to stay with machine type-based
identification.


-- 
Regards,
Alexander

Renaming MACH_TYPE_KINETIS_K60 into MACH_TYPE_KINETIS

[Marek Vasut]

> On Wed, 2011-12-28 at 11:00 +0100, Marek Vasut wrote:
> > If you use FDT, you won't need machine ID at all.
> 
> Hi Marek,
> 
> You are right, but we have not used FDT for any of our ARM Linux ports
> yet (consistency matters.) If we switch to FDT now, a lot of code will
> need to be rewritten.
> 
> That is why I would prefer to stay with machine type-based
> identification.

Well ... you probably do, let's see about the others ;-)

Renaming MACH_TYPE_KINETIS_K60 into MACH_TYPE_KINETIS

[Rob Herring]

On 12/28/2011 04:51 AM, Alexander Potashev wrote:
> On Wed, 2011-12-28 at 11:00 +0100, Marek Vasut wrote:
>> If you use FDT, you won't need machine ID at all. 
> 
> Hi Marek,
> 
> You are right, but we have not used FDT for any of our ARM Linux ports
> yet (consistency matters.) If we switch to FDT now, a lot of code will
> need to be rewritten.

Not if you are only talking about the machine identification. You need a
minimal dts file with a compatible string for the board and add this
string to your machine desc.

DT is a requirement for adding any new ARM platform.

Rob

Renaming MACH_TYPE_KINETIS_K60 into MACH_TYPE_KINETIS

[Russell King - ARM Linux]

On Wed, Dec 28, 2011 at 01:51:22PM +0400, Alexander Potashev wrote:
> Hi,
> 
> About three weeks ago I registered a machine type for the Freescale
> Kinetis K60 microcontroller. ( http://www.arm.linux.org.uk/developer/machines/list.php?id=3896 )
> 
> But later I realized that there should be a single machine type for all
> Freescale Kinetis MCUs (K10 through K70), because those MCUs are
> software-compatible. I tried to rename "kinetis_k60" to just "kinetis"
> at the page mentioned above, but the names of the CONFIG_* macro and the
> identifier name did not change.
> 
> I really do not want to register another machine type since it would be
> a mess having two machine types - one for Kinetis and another for
> Kinetis K60. Is it possible to change CONFIG_MACH_KINETIS_K60 to
> CONFIG_MACH_KINETIS and MACH_TYPE_KINETIS_K60 to MACH_TYPE_KINETIS in
> the database at http://www.arm.linux.org.uk/developer/machines/ ?

I notice that you sent me a very similar email on 24th December, which I
haven't responded to yet: given that it's the Christmas period, and
people take holidays, I think you're being a little hasty over your
resending.

In any case, yes, I can change it.  However, as has already been
pointed out, we're moving over to using device tree as the primary form
of platform differentiation.  With device tree, machine type numbers
are completely meaningless (and aren't actually used by the kernel.)

While I will fix the entry, I suspect that Arnd will now refuse to
merge any new SoC support for mainline which is not using the device
tree.

Renaming MACH_TYPE_KINETIS_K60 into MACH_TYPE_KINETIS

[Alexander Potashev]

Hi Russell,

On 3 ?????? 2012 11:03:37 Russell King - ARM Linux wrote:
> I notice that you sent me a very similar email on 24th December, which I
> haven't responded to yet: given that it's the Christmas period, and
> people take holidays, I think you're being a little hasty over your
> resending.

The reason why I have re-sent the message to the mailing list was that I 
feeled like this discussion should be public, not just that I was bored 
waiting for the response.

Anyway, sorry for the noise.

> In any case, yes, I can change it.  However, as has already been
> pointed out, we're moving over to using device tree as the primary form
> of platform differentiation.  With device tree, machine type numbers
> are completely meaningless (and aren't actually used by the kernel.)

Thanks in advance for fixing the entry!

> While I will fix the entry, I suspect that Arnd will now refuse to
> merge any new SoC support for mainline which is not using the device
> tree.

Linux kernel development in our company is still based on the 2.6.33 version 
of the kernel. Most of our code is not going to be sent upstream. We will 
probably think about FDT when upgrading to a newer version of the kernel.


-- 
Regards,
Alexander

Renaming MACH_TYPE_KINETIS_K60 into MACH_TYPE_KINETIS

[Russell King - ARM Linux]

On Tue, Jan 03, 2012 at 03:21:03PM +0400, Alexander Potashev wrote:
> Hi Russell,
> 
> On 3 ?????? 2012 11:03:37 Russell King - ARM Linux wrote:
> > I notice that you sent me a very similar email on 24th December, which I
> > haven't responded to yet: given that it's the Christmas period, and
> > people take holidays, I think you're being a little hasty over your
> > resending.
> 
> The reason why I have re-sent the message to the mailing list was that I 
> feeled like this discussion should be public, not just that I was bored 
> waiting for the response.
> 
> Anyway, sorry for the noise.
> 
> > In any case, yes, I can change it.  However, as has already been
> > pointed out, we're moving over to using device tree as the primary form
> > of platform differentiation.  With device tree, machine type numbers
> > are completely meaningless (and aren't actually used by the kernel.)
> 
> Thanks in advance for fixing the entry!
> 
> > While I will fix the entry, I suspect that Arnd will now refuse to
> > merge any new SoC support for mainline which is not using the device
> > tree.
> 
> Linux kernel development in our company is still based on the 2.6.33 version 
> of the kernel. Most of our code is not going to be sent upstream. We will 
> probably think about FDT when upgrading to a newer version of the kernel.

I think it may help your case to point out that this is a uclinux platform,
being Cortex M4 based, rather than a mmu-ful platform.  While much of the
DT progress has been focused on mmu-ful platforms, much of this should just
work for DT with uclinux.

However, I suspect there may be issues such as how to provide the kernel
with a DT blob which would need to be addressed which would make this
non-trivial.

There's also the issue whether DT on uclinux makes sense as these
platforms are normally highly focussed and highly embedded (and it
really doesn't make that much sense to have a single kernel image
running on all.)

Renaming MACH_TYPE_KINETIS_K60 into MACH_TYPE_KINETIS

[Alexander Potashev]

On 3 ?????? 2012 15:32:03 Russell King - ARM Linux wrote:
> There's also the issue whether DT on uclinux makes sense as these
> platforms are normally highly focussed and highly embedded (and it
> really doesn't make that much sense to have a single kernel image
> running on all.)

Right. Our workflow is to have a custom kernel image for every particular 
embedded system and we are happy with it.