Re: SHMobile Compatibility String Inconsistencies

Re: SHMobile Compatibility String Inconsistencies

[David Gibson]

[-- Attachment #1: Type: text/plain, Size: 2014 bytes --]

On Fri, Aug 23, 2013 at 01:31:31PM +0200, Laurent Pinchart wrote:
> Hi David,
> 
> On Friday 23 August 2013 12:11:11 David Gibson wrote:
> > On Thu, Aug 22, 2013 at 02:46:40PM +0900, Simon Horman wrote:
> > > Hi Laurent, Hi Guennadi, Hi All,
> > > 
> > > Olof has brought to my attention that there is some inconsistency
> > > in the way that compatibility strings for SHMobile are named and he
> > > has asked us to clean things up for v3.12.
> > > 
> > > Looking through arch/arm/boot/dts/ I see that we have:
> > > 
> > > 1. {gpio,pfc}-r8aXXXX and;
> > > 2. r8aXXXX-sdhi
> > > 
> > > The inconsistency that Olof has asked us to resolve is that we
> > > should either use r8aXXXX- or -r8aXXXX. Not both.
> > > 
> > > It seems to me that neither option is inherently better than the other
> > > so we should just choose the path of least resistance to make things
> > > consistent.
> > > 
> > > Laurent, Guennadi, do you have any opinions on if it would
> > > be easier to change the GPIO and PFC compatibility strings;
> > > or to change the SDHI compatibility strings?
> > > 
> > > Ideally I would like you to come to some sort of consensus and send
> > > patches.
> > 
> > So, by all means clean this up in the dts.
> > 
> > BUT, in keeping with the recent discussions on improving the DT
> > process, the corresponding drivers must continue to recognize both
> > forms, so that old DTs will still work correctly.
> 
> Given the early state of DT support in arm/mach-shmobile, I'm pretty sure we 
> have no DT-based systems in the wild. The old compatibility string could in my 
> opinion just be dropped.

Hm, well, ok.  Just remember to always err on the side of broad
compatiblity.  We've become sloppy with incompatible DT updates and
its caused us a bunch of grief.

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: Type: application/pgp-signature, Size: 836 bytes --]

Re: SHMobile Compatibility String Inconsistencies

[Guennadi Liakhovetski]

On Sat, 24 Aug 2013, Simon Horman wrote:

> On Fri, Aug 23, 2013 at 01:31:31PM +0200, Laurent Pinchart wrote:
> > Hi David,
> > 
> > On Friday 23 August 2013 12:11:11 David Gibson wrote:
> > > On Thu, Aug 22, 2013 at 02:46:40PM +0900, Simon Horman wrote:
> > > > Hi Laurent, Hi Guennadi, Hi All,
> > > > 
> > > > Olof has brought to my attention that there is some inconsistency
> > > > in the way that compatibility strings for SHMobile are named and he
> > > > has asked us to clean things up for v3.12.
> > > > 
> > > > Looking through arch/arm/boot/dts/ I see that we have:
> > > > 
> > > > 1. {gpio,pfc}-r8aXXXX and;

Add shdma-<soc> to the above

> > > > 2. r8aXXXX-sdhi
> > > > 
> > > > The inconsistency that Olof has asked us to resolve is that we
> > > > should either use r8aXXXX- or -r8aXXXX. Not both.

Given the 3:1 score the choice seems rather simple to me. So, if we do
have to make those consistent, let's change SDHI.

> > > > It seems to me that neither option is inherently better than the other
> > > > so we should just choose the path of least resistance to make things
> > > > consistent.
> > > > 
> > > > Laurent, Guennadi, do you have any opinions on if it would
> > > > be easier to change the GPIO and PFC compatibility strings;
> > > > or to change the SDHI compatibility strings?
> > > > 
> > > > Ideally I would like you to come to some sort of consensus and send
> > > > patches.
> > > 
> > > So, by all means clean this up in the dts.
> > > 
> > > BUT, in keeping with the recent discussions on improving the DT
> > > process, the corresponding drivers must continue to recognize both
> > > forms, so that old DTs will still work correctly.
> > 
> > Given the early state of DT support in arm/mach-shmobile, I'm pretty sure we 
> > have no DT-based systems in the wild. The old compatibility string could in my 
> > opinion just be dropped.
> 
> I tend to agree, though I don't mind either way.

So, what path should we choose? I see 2 SDHI users in your current "devel" 
brunch: sh73a0 (kzm9g) and r8a73a4 (ape6evm). Should we make a single 
patch, changing the driver and both users and push it via ARM with Chris' 
ack or shall we add new compats, switch .dtsi's, remove old compats - over 
2 kernel versions? Can we still get anything for this into 3.12 or is it 
too late? Would it be a "fix" enough for -rc2 / late -rc1?

Thanks
Guennadi
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/

Re: SHMobile Compatibility String Inconsistencies

[Laurent Pinchart]

Hi Guennadi,

On Monday 26 August 2013 18:08:52 Guennadi Liakhovetski wrote:
> On Sat, 24 Aug 2013, Simon Horman wrote:
> > On Fri, Aug 23, 2013 at 01:31:31PM +0200, Laurent Pinchart wrote:
> > > On Friday 23 August 2013 12:11:11 David Gibson wrote:
> > > > On Thu, Aug 22, 2013 at 02:46:40PM +0900, Simon Horman wrote:
> > > > > Hi Laurent, Hi Guennadi, Hi All,
> > > > > 
> > > > > Olof has brought to my attention that there is some inconsistency
> > > > > in the way that compatibility strings for SHMobile are named and he
> > > > > has asked us to clean things up for v3.12.
> > > > > 
> > > > > Looking through arch/arm/boot/dts/ I see that we have:
> > > > > 
> > > > > 1. {gpio,pfc}-r8aXXXX and;
> 
> Add shdma-<soc> to the above
> 
> > > > > 2. r8aXXXX-sdhi
> > > > > 
> > > > > The inconsistency that Olof has asked us to resolve is that we
> > > > > should either use r8aXXXX- or -r8aXXXX. Not both.
> 
> Given the 3:1 score the choice seems rather simple to me. So, if we do
> have to make those consistent, let's change SDHI.

Agreed.

> > > > > It seems to me that neither option is inherently better than the
> > > > > other so we should just choose the path of least resistance to make
> > > > > things consistent.
> > > > > 
> > > > > Laurent, Guennadi, do you have any opinions on if it would
> > > > > be easier to change the GPIO and PFC compatibility strings;
> > > > > or to change the SDHI compatibility strings?
> > > > > 
> > > > > Ideally I would like you to come to some sort of consensus and send
> > > > > patches.
> > > > 
> > > > So, by all means clean this up in the dts.
> > > > 
> > > > BUT, in keeping with the recent discussions on improving the DT
> > > > process, the corresponding drivers must continue to recognize both
> > > > forms, so that old DTs will still work correctly.
> > > 
> > > Given the early state of DT support in arm/mach-shmobile, I'm pretty
> > > sure we have no DT-based systems in the wild. The old compatibility
> > > string could in my opinion just be dropped.
> > 
> > I tend to agree, though I don't mind either way.
> 
> So, what path should we choose? I see 2 SDHI users in your current "devel"
> brunch: sh73a0 (kzm9g) and r8a73a4 (ape6evm). Should we make a single
> patch, changing the driver and both users and push it via ARM with Chris'
> ack or shall we add new compats, switch .dtsi's, remove old compats - over
> 2 kernel versions?

I'm fine with both, but in the latter case I don't see a need to spread it 
over two kernel versions.

> Can we still get anything for this into 3.12 or is it too late? Would it be
> a "fix" enough for -rc2 / late -rc1?

-- 
Regards,

Laurent Pinchart

Re: SHMobile Compatibility String Inconsistencies

[Guennadi Liakhovetski]

Hi Laurent,

On Tue, 27 Aug 2013, Laurent Pinchart wrote:

> Hi Guennadi,
> 
> On Monday 26 August 2013 18:08:52 Guennadi Liakhovetski wrote:
> > On Sat, 24 Aug 2013, Simon Horman wrote:
> > > On Fri, Aug 23, 2013 at 01:31:31PM +0200, Laurent Pinchart wrote:
> > > > On Friday 23 August 2013 12:11:11 David Gibson wrote:
> > > > > On Thu, Aug 22, 2013 at 02:46:40PM +0900, Simon Horman wrote:
> > > > > > Hi Laurent, Hi Guennadi, Hi All,
> > > > > > 
> > > > > > Olof has brought to my attention that there is some inconsistency
> > > > > > in the way that compatibility strings for SHMobile are named and he
> > > > > > has asked us to clean things up for v3.12.
> > > > > > 
> > > > > > Looking through arch/arm/boot/dts/ I see that we have:
> > > > > > 
> > > > > > 1. {gpio,pfc}-r8aXXXX and;
> > 
> > Add shdma-<soc> to the above
> > 
> > > > > > 2. r8aXXXX-sdhi
> > > > > > 
> > > > > > The inconsistency that Olof has asked us to resolve is that we
> > > > > > should either use r8aXXXX- or -r8aXXXX. Not both.
> > 
> > Given the 3:1 score the choice seems rather simple to me. So, if we do
> > have to make those consistent, let's change SDHI.
> 
> Agreed.
> 
> > > > > > It seems to me that neither option is inherently better than the
> > > > > > other so we should just choose the path of least resistance to make
> > > > > > things consistent.
> > > > > > 
> > > > > > Laurent, Guennadi, do you have any opinions on if it would
> > > > > > be easier to change the GPIO and PFC compatibility strings;
> > > > > > or to change the SDHI compatibility strings?
> > > > > > 
> > > > > > Ideally I would like you to come to some sort of consensus and send
> > > > > > patches.
> > > > > 
> > > > > So, by all means clean this up in the dts.
> > > > > 
> > > > > BUT, in keeping with the recent discussions on improving the DT
> > > > > process, the corresponding drivers must continue to recognize both
> > > > > forms, so that old DTs will still work correctly.
> > > > 
> > > > Given the early state of DT support in arm/mach-shmobile, I'm pretty
> > > > sure we have no DT-based systems in the wild. The old compatibility
> > > > string could in my opinion just be dropped.
> > > 
> > > I tend to agree, though I don't mind either way.
> > 
> > So, what path should we choose? I see 2 SDHI users in your current "devel"
> > brunch: sh73a0 (kzm9g) and r8a73a4 (ape6evm). Should we make a single
> > patch, changing the driver and both users and push it via ARM with Chris'
> > ack or shall we add new compats, switch .dtsi's, remove old compats - over
> > 2 kernel versions?
> 
> I'm fine with both, but in the latter case I don't see a need to spread it 
> over two kernel versions.

The latter only makes sense if we want to push patches via separate trees, 
i.e.
DMA
ARM
DMA
and doing this within 1 kernel release would be difficult. Whereas if we 
want to push all 3 patches via 1 tree, then I'd say just doing 1 patch 
would be easier.

Thanks
Guennadi

> > Can we still get anything for this into 3.12 or is it too late? Would it be
> > a "fix" enough for -rc2 / late -rc1?
> 
> -- 
> Regards,
> 
> Laurent Pinchart

---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/

Re: SHMobile Compatibility String Inconsistencies

[Olof Johansson]

Hi,

On Thu, Aug 22, 2013 at 5:19 PM, Simon Horman <horms@verge.net.au> wrote:

> My understanding from Olof is that it is fine to just make
> an SH-mobile-wide decision.

Sorry for being vague on this -- you really should try to join in on
the same conventions as the other platforms (if one gets established).

My general opinion is aligned with Stephen Warren's in this case, but
it's up to the devicetree-maintainers to make final calls.


-Olof