[GIT,PULL,(RESEND)] ARM: uniphier: UniPhier fixes for v4.12
mbox

Message ID CAK7LNARbSkbEOhs7QMd+6VV+KWbkjK4LR9rmQ=YBWqPY+8_XPQ@mail.gmail.com
State New
Headers show

Pull-request

git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-uniphier.git

Message

Masahiro Yamada May 22, 2017, 2:03 a.m. UTC
Hi Arnd, Olof,

Last week, I thought I found a problem in my pull request,
but it turned out my misunderstanding.

So, my pull request I sent is OK.
I am re-sending this request mail for clarification.

Here are some ARM UniPhier SoC fixes for v4.12.
Due to the eMMC driver update, PHY parameters must be specified
via DT to make the eMMC controller work again.
The rest is DTC warning fixes reported when building DT with W=1.
Please pull!


The following changes since commit 2ea659a9ef488125eb46da6eb571de5eae5c43f6:

  Linux 4.12-rc1 (2017-05-13 13:19:49 -0700)

are available in the git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-uniphier.git
tags/uniphier-fixes-v4.12

for you to fetch changes up to 96abf9ded7954d9e6f3a7775353d9550b1283db9:

  arm64: dts: uniphier: fix simple-bus unit address format error
(2017-05-14 11:43:46 +0900)

----------------------------------------------------------------
UniPhier ARM SoC fixes for v4.12

- add necessary DT properties to get eMMC work again
- fix DTC warnings when compiled with W=1

----------------------------------------------------------------
Masahiro Yamada (4):
      arm64: dts: uniphier: add input-delay properties to Cadence eMMC node
      arm64: dts: uniphier: add cdns, phy-dll-delay-sdclk(-hsmmc) for eMMC
      ARM: dts: uniphier: fix simple-bus unit address format error
      arm64: dts: uniphier: fix simple-bus unit address format error

Viresh Kumar (2):
      ARM: uniphier: Use - instead of @ for DT OPP entries
      arm64: uniphier: Use - instead of @ for DT OPP entries

 arch/arm/boot/dts/uniphier-ld4.dtsi              |  2 +-
 arch/arm/boot/dts/uniphier-pro4.dtsi             |  2 +-
 arch/arm/boot/dts/uniphier-pro5.dtsi             | 34 ++++++++++-----------
 arch/arm/boot/dts/uniphier-pxs2.dtsi             | 18 +++++------
 arch/arm/boot/dts/uniphier-sld3.dtsi             |  2 +-
 arch/arm/boot/dts/uniphier-sld8.dtsi             |  2 +-
 arch/arm/boot/dts/uniphier-support-card.dtsi     |  4 +--
 arch/arm64/boot/dts/socionext/uniphier-ld11.dtsi | 21 ++++++++-----
 arch/arm64/boot/dts/socionext/uniphier-ld20.dtsi | 39 +++++++++++++-----------
 9 files changed, 67 insertions(+), 57 deletions(-)

Comments

Olof Johansson May 25, 2017, 11:59 p.m. UTC | #1
On Mon, May 22, 2017 at 11:03:49AM +0900, Masahiro Yamada wrote:
> Hi Arnd, Olof,
> 
> Last week, I thought I found a problem in my pull request,
> but it turned out my misunderstanding.
> 
> So, my pull request I sent is OK.
> I am re-sending this request mail for clarification.
> 
> Here are some ARM UniPhier SoC fixes for v4.12.
> Due to the eMMC driver update, PHY parameters must be specified
> via DT to make the eMMC controller work again.

That's not good, that's a regression in behavior. The driver should
have a fallback when the properties are not present to avoid regressions
like these.

Please work with the driver maintainer to either get them to revert
their change that regresses you, or promptly fix it appropriately with
fallback values.

Also, the driver changes were commited back in March, with plenty of time
to land DT changes well before the merge window. Waiting until the merge
window is closed is not appropriate -- this should have come in earlier.


-Olof
Masahiro Yamada May 26, 2017, 1:42 a.m. UTC | #2
Hi Olof,

2017-05-26 8:59 GMT+09:00 Olof Johansson <olof@lixom.net>:
> On Mon, May 22, 2017 at 11:03:49AM +0900, Masahiro Yamada wrote:
>> Hi Arnd, Olof,
>>
>> Last week, I thought I found a problem in my pull request,
>> but it turned out my misunderstanding.
>>
>> So, my pull request I sent is OK.
>> I am re-sending this request mail for clarification.
>>
>> Here are some ARM UniPhier SoC fixes for v4.12.
>> Due to the eMMC driver update, PHY parameters must be specified
>> via DT to make the eMMC controller work again.
>
> That's not good, that's a regression in behavior. The driver should
> have a fallback when the properties are not present to avoid regressions
> like these.

Right.  The care for old DT is missing from the driver change.
Theoretically, it is a regression if Linux 4.12 kernel image is used with
DT from Linux 4.11 tree.

However, all our customer products are based on Linux 4.4 (or older)
so I am sure there is no practical problem.
Because the eMMC driver was not upstreamed at that time,
products kernels are locally patched as needed.

I am upstreaming patches to create a better code base for future products, but
Linux 4.11 and 4.12 are not LTS, so never used at least on socionext products.


> Please work with the driver maintainer to either get them to revert
> their change that regresses you, or promptly fix it appropriately with
> fallback values.

You are right, but this is a different issue.


If you are unhappy with pulling this as a fix
(because it is not DT but the driver that should be fixed),
I can include the DT changes in usual PRs for v4.13-rc1.


As I said above, I admit this is not fatal for us
because socionext kernel developers are co-located
and know the upstream status.



> Also, the driver changes were commited back in March, with plenty of time
> to land DT changes well before the merge window. Waiting until the merge
> window is closed is not appropriate -- this should have come in earlier.

You are right.
But, v4.11-rc4 had already passed and I had already sent the pull request
before the driver change was merged.

I generally try to send my PRs around -rc4
because "sorry, this is too late" is scary.
I should have sent the second pull request.
If you do not mind some after -rc4, I will do late PRs next time.


So, what should I do about this?

If you do not want to pull this, I will move the patches
to for-v4.13 branch, which will be sent around -rc4.