Message-ID: <20140521211105.GD24341@quad.lixom.net>
Date:	Wed, 21 May 2014 14:11:05 -0700
From:	Olof Johansson <olof@...om.net>
To:	Nicolas Ferre <nicolas.ferre@...el.com>
Cc:	Alexandre Belloni <alexandre.belloni@...e-electrons.com>,
	Boris BREZILLON <boris.brezillon@...e-electrons.com>,
	Arnd Bergmann <arnd@...db.de>, arm@...nel.org,
	Linux Kernel list <linux-kernel@...r.kernel.org>,
	linux-arm-kernel <linux-arm-kernel@...ts.infradead.org>,
	Jean-Christophe PLAGNIOL-VILLARD <plagnioj@...osoft.com>,
	Wenyou Yang <wenyou.yang@...el.com>,
	Ludovic Desroches <ludovic.desroches@...el.com>
Subject: Re: [GIT PULL] at91: DT for 3.16 #2

On Wed, May 21, 2014 at 12:17:09PM +0200, Nicolas Ferre wrote:
> On 20/05/2014 18:47, Olof Johansson :
> > Hi,
> > 
> > On Tue, May 20, 2014 at 05:19:24PM +0200, Nicolas Ferre wrote:
> >> On 20/05/2014 07:50, Olof Johansson :
> >>> On Wed, May 14, 2014 at 11:19:22AM +0200, Nicolas Ferre wrote:
> >>>> Arnd, Olof, Kevin,
> >>>>
> >>>> More DT material for AT91. Some fixes that apply on what was merged for 3.15
> >>>> but that are not very critical.
> >>>> The other patches are feature additions to old or very recent product/board:
> >>>> at91sam9261 or sama5d3 Xplained.
> >>>>
> >>>> Thanks, best regards,
> >>>>
> >>>> The following changes since commit 27a96a0364787d2b41d2a72d08143d95263e1b07:
> >>>>
> >>>>   ARM: at91: sama5d3: clock for ssc from rk pin (2014-04-18 22:43:44 +0200)
> >>>>
> >>>> are available in the git repository at:
> >>>>
> >>>>   git://github.com/at91linux/linux-at91.git tags/at91-dt2
> >>>>
> >>>> for you to fetch changes up to a93f9c88b7701d1c4c3b22d39d64a408f000a6ef:
> >>>>
> >>>>   ARM: at91/dt: at91-sama5d3_xplained: add the regulator device node (2014-05-12 16:48:54 +0200)
> >>>
> >>> Merged, but:
> >>>
> >>>>  arch/arm/boot/dts/at91-sama5d3_xplained.dts |  62 +++++++++++++++
> >>>>  arch/arm/boot/dts/at91sam9261.dtsi          | 114 ++++++++++++++++++++++++++--
> >>>>  arch/arm/boot/dts/at91sam9rl.dtsi           |   7 +-
> >>>>  arch/arm/boot/dts/sama5d3.dtsi              |  78 +++++++++++++++++++
> >>>
> >>> Grmbl. I remember being somewhat annoyed that you didn't use at91 prefix
> >>> when you introduced the sama5d3 dtsi files, but please don't start using
> >>> it on a random board like this, especially when other boards just use
> >>> the sama5d3_<board>.dts format.
> >>
> >> Well, I don't understand completely. Since our discussion during 3.10
> >> merge window ([GIT PULL] at91: DT changes for 3.10 #2), I try to conform
> >> to this rule:
> >>
> >> 1/ all pre-3.10 and 3.10 device tree file names stay unchanged
> >> -> sama5d3.dti (SoC)
> >> -> sama5d35ek.dts (board)
> >>
> >> 2/ all *SoC* DT files conform to their marking:
> >> at91sam9263.dtsi
> >> at91sam9rl.dtsi
> >> sama5d3.dtsi, sama5d36.dtsi
> >> sama5d4.dtsi, sama5d46.dtsi (maybe in the future, who knows...)
> >>
> >> 3/ all post-3.10 *boards* have the "at91-" prefix, whether they are
> >> populated with sam9 or sama5:
> >> at91-ariag25.dts (since 3.10, using a at91sam9g25)
> >> at91-qil_a9260.dts (since 3.14, using at91sam9260)
> >> at91-sama5d3_xplained.dts (since 3.14, using sama5d36)
> >>
> >> The rule for AT91 has never been to prefix the board DT filename with
> >> the name of the SoC or SoC family.
> > 
> > So, going back and looking at the discussion from a year ago, I think the
> > disconnect was in what consistency we were looking for. Yes, we would have
> > preferred to prefix the sama5d3* dts/dtsis with at91, and you even
> > offered to do it. ;-) But I think even more important for sanity is
> > to stay consistent with how we handle all platforms, which is that the
> > board dts files are prefixed with the SoC name.
> > 
> > In the past, we've had cases where this didn't happen, but these days we have
> > tried to be very consistent on it. I.e. omap3*, exynos<##>*, etc.
> 
> Okay, but once again, I tried to deal with the existing files, not break
> any user's habit, before any convention had been clearly established,
> and now... we are reaching a deadlock having to re-consider again our DT
> filenames?

I don't think we're at a deadlock here. We're trying to figure out the best way
forward.

> 
> > So, if you have at91- as a prefix, have the SoC as the second component. But
> > that gets awkward too, so I would juts use the current SoC dtsi as the prefix
> > at91sam9263-<boardname>.dts, or sama5d35_<boardname>.dts.
> > 
> >>> Care to fix this up in time for 3.16 merge window?
> >>
> >> Well, I do not know what to fix as the files were already present in
> >> mainline before this kernel revision and that I am a little bit
> >> reluctant to change file names after they are merged in mainline.
> >>
> >> Now, can we keep the current policy described above (somehow weird, I
> >> admit) for future SoCs and boards?
> > 
> > It is unfortunate that I didn't catch this for 3.14 so that name has been
> > there in a release. I guess the least disruptive thing for now would be
> > to change over to use the SoC dtsi prefix for any new board files from
> > here on out, and treat at91-sama5d3_xplained as a one-time thing.
> 
> That is not only about sama5d3, what about sam9-based boards? Several of
> them already have this "at91-" prefix (and not a soc prefix).
> 
> So this will introduce another temporary naming convention for these
> files which are actually used by people: everyone will be puzzled.

Yes, everyone is already puzzled though.

> Okay, there are two ways to escape this situation:
> 1/ change nothing and conform to what I stated above. It is more like
>    finding a rationale to an existing situation than a real neat
>    policy, but hey, "nobody's perfect".
> 2/ change *everything* in AT91 DT file naming with a schema that
>    we will have to validate with care (<soc family> or <soc>,
>    "_" or "-", what about boards with modules, etc.).
>    That will certainly disturb many of our users without a real benefit.

I don't think it's worth renaming everything at this time.

> This directory is flat, the board names are chosen by companies and
> people that we do not control, a user tend to like finding his preferred
> board dtb file unchanged from a kernel revision to another...
> Well all this lead me to think that we don't have to loose too much time
> thinking about a new strict convention for this file naming or changing
> all this once again just for the sake of it.
> 
> Other SoC maintainers beautifully designed from the beginning the naming
> scheme of their DT files, fine. AT91 did not and forgive me but when
> opening arch/arm/boot/dts/Makefile file and seeing some file names, I'm
> not ashamed. Moreover, now that I said to everybody since 3.10 to prefix
> their *board* name with "at91-", I have to say something else, I don't
> think it is worth it.

I don't agree with everything above, but it's not worth arguing for the
sake of arguing. :) I think we can tweak what you're doing now and get
things to work well by merging new dts files with at91-<soc>-board.dts
as the name. As mentioned, don't worry about the existing files. This
shouldn't be a significiant change to what you've been telling people
since 3.10 to cause much confusion.


-Olof
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists