vexpress issues in next-20121029

vexpress issues in next-20121029

[Stephen Warren]

Pawel,

I see two issues on vexpress in next-20121029:

1) At compile-time: v2m_timer_init() and v2m_dt_timer_init() reference
functions vexpress_clk_init() and vexpress_clk_of_init() respectively,
which don't appear anywhere in the source tree.

2) With those calls commented out, I find that
vexpress_sysreg_init_leds() is device_initcall, and so executes even
when not running on vexpress HW. This crashes on Tegra (which I have
converted to single-zImage locally).

These appear to come from the following commits:

0891642 ARM: vexpress: Start using new Versatile Express infrastructure
568e5e4 mfd: Versatile Express system registers driver

vexpress issues in next-20121029

[Pawel Moll]

Hi Stephen,

On Mon, 2012-10-29 at 20:24 +0000, Stephen Warren wrote:
> 1) At compile-time: v2m_timer_init() and v2m_dt_timer_init() reference
> functions vexpress_clk_init() and vexpress_clk_of_init() respectively,
> which don't appear anywhere in the source tree.

There was a glitch between clk-next and arm-soc - it should be fine
starting with next-20121030.

> 2) With those calls commented out, I find that
> vexpress_sysreg_init_leds() is device_initcall, and so executes even
> when not running on vexpress HW. This crashes on Tegra (which I have
> converted to single-zImage locally).

Uh, how stupid of me! I have to finally get used to the new
multi-platform-kernel reality... Sorry about it - a quick workaround
would be:

8<----------------------------
diff --git a/drivers/mfd/vexpress-sysreg.c
b/drivers/mfd/vexpress-sysreg.c
index 059d6b1..99e89a8 100644
--- a/drivers/mfd/vexpress-sysreg.c
+++ b/drivers/mfd/vexpress-sysreg.c
@@ -474,7 +474,6 @@ static int __init vexpress_sysreg_init(void)
 }
 core_initcall(vexpress_sysreg_init);
 
-
 #if defined(CONFIG_LEDS_CLASS)
 
 struct vexpress_sysreg_led {
@@ -526,6 +525,9 @@ static int __init vexpress_sysreg_init_leds(void)
 	struct vexpress_sysreg_led *led;
 	int i;
 
+	if (!vexpress_sysreg_base)
+		return 0;
+
 	/* Clear all user LEDs */
 	writel(0, vexpress_sysreg_base + SYS_LED);
 
8<----------------------------

while I'm improving the situation by better means.

Sorry about this and thanks for letting me know!

Pawe?

vexpress issues in next-20121029

[Russell King - ARM Linux]

On Tue, Oct 30, 2012 at 04:16:59PM +0000, Pawel Moll wrote:
> Hi Stephen,
> 
> On Mon, 2012-10-29 at 20:24 +0000, Stephen Warren wrote:
> > 1) At compile-time: v2m_timer_init() and v2m_dt_timer_init() reference
> > functions vexpress_clk_init() and vexpress_clk_of_init() respectively,
> > which don't appear anywhere in the source tree.
> 
> There was a glitch between clk-next and arm-soc - it should be fine
> starting with next-20121030.

The problem is still there - my builds of my tree plus arm-soc are
continuing to fail with:

arch/arm/mach-vexpress/built-in.o: In function `v2m_timer_init':
reset.c:(.init.text+0xe0): undefined reference to `vexpress_clk_init'
arch/arm/mach-vexpress/built-in.o: In function `v2m_dt_timer_init':
reset.c:(.init.text+0x114): undefined reference to `vexpress_clk_of_init'

My guess is you have a dependency between the clk-next tree and arm-soc
which you haven't told the arm-soc people about.

vexpress issues in next-20121029

[Russell King - ARM Linux]

On Mon, Nov 05, 2012 at 09:35:32AM +0000, Russell King - ARM Linux wrote:
> On Tue, Oct 30, 2012 at 04:16:59PM +0000, Pawel Moll wrote:
> > Hi Stephen,
> > 
> > On Mon, 2012-10-29 at 20:24 +0000, Stephen Warren wrote:
> > > 1) At compile-time: v2m_timer_init() and v2m_dt_timer_init() reference
> > > functions vexpress_clk_init() and vexpress_clk_of_init() respectively,
> > > which don't appear anywhere in the source tree.
> > 
> > There was a glitch between clk-next and arm-soc - it should be fine
> > starting with next-20121030.
> 
> The problem is still there - my builds of my tree plus arm-soc are
> continuing to fail with:
> 
> arch/arm/mach-vexpress/built-in.o: In function `v2m_timer_init':
> reset.c:(.init.text+0xe0): undefined reference to `vexpress_clk_init'
> arch/arm/mach-vexpress/built-in.o: In function `v2m_dt_timer_init':
> reset.c:(.init.text+0x114): undefined reference to `vexpress_clk_of_init'
> 
> My guess is you have a dependency between the clk-next tree and arm-soc
> which you haven't told the arm-soc people about.

Oh, and the above seems to affect all my kautobuilds for any ARM Ltd
development platform - it's not just vexpress which is affected by this
anymore.

vexpress issues in next-20121029

[Pawel Moll]

On Mon, 2012-11-05 at 09:45 +0000, Russell King - ARM Linux wrote:
> On Mon, Nov 05, 2012 at 09:35:32AM +0000, Russell King - ARM Linux wrote:
> > On Tue, Oct 30, 2012 at 04:16:59PM +0000, Pawel Moll wrote:
> > > There was a glitch between clk-next and arm-soc - it should be fine
> > > starting with next-20121030.
> > 
> > The problem is still there - my builds of my tree plus arm-soc are
> > continuing to fail with:
> > 
> > arch/arm/mach-vexpress/built-in.o: In function `v2m_timer_init':
> > reset.c:(.init.text+0xe0): undefined reference to `vexpress_clk_init'
> > arch/arm/mach-vexpress/built-in.o: In function `v2m_dt_timer_init':
> > reset.c:(.init.text+0x114): undefined reference to `vexpress_clk_of_init'
> > 
> > My guess is you have a dependency between the clk-next tree and arm-soc
> > which you haven't told the arm-soc people about.
> 
> Oh, and the above seems to affect all my kautobuilds for any ARM Ltd
> development platform - it's not just vexpress which is affected by this
> anymore.

I've just successfully built defconfigs for vexpress, versatile and
realview with next-20121105 so I guess Arnd simply didn't have time last
week (the Connect event) to sort out the arm-soc tree...

Arnd, will you pull the vexpress-clk-soc (containing, as you suggested,
the soc stuff rebased on top of the clk branch) or do you want me to do
something else?

Thanks!

Pawe?

vexpress issues in next-20121029

[Arnd Bergmann]

On Monday 05 November 2012, Pawel Moll wrote:
> On Mon, 2012-11-05 at 09:45 +0000, Russell King - ARM Linux wrote:
> > On Mon, Nov 05, 2012 at 09:35:32AM +0000, Russell King - ARM Linux wrote:
> > > On Tue, Oct 30, 2012 at 04:16:59PM +0000, Pawel Moll wrote:
> > > > There was a glitch between clk-next and arm-soc - it should be fine
> > > > starting with next-20121030.
> > > 
> > > The problem is still there - my builds of my tree plus arm-soc are
> > > continuing to fail with:
> > > 
> > > arch/arm/mach-vexpress/built-in.o: In function `v2m_timer_init':
> > > reset.c:(.init.text+0xe0): undefined reference to `vexpress_clk_init'
> > > arch/arm/mach-vexpress/built-in.o: In function `v2m_dt_timer_init':
> > > reset.c:(.init.text+0x114): undefined reference to `vexpress_clk_of_init'
> > > 
> > > My guess is you have a dependency between the clk-next tree and arm-soc
> > > which you haven't told the arm-soc people about.
> > 
> > Oh, and the above seems to affect all my kautobuilds for any ARM Ltd
> > development platform - it's not just vexpress which is affected by this
> > anymore.
> 
> I've just successfully built defconfigs for vexpress, versatile and
> realview with next-20121105 so I guess Arnd simply didn't have time last
> week (the Connect event) to sort out the arm-soc tree...
> 
> Arnd, will you pull the vexpress-clk-soc (containing, as you suggested,
> the soc stuff rebased on top of the clk branch) or do you want me to do
> something else?

Sorry for not getting back to this earlier. It was indeed a very busy
week and I did not manage to pull in any branches.

I've just tried pulling in your branch again, but it appears unchanged:
The patches are still based on 807e45b328, which is a different commit
from bcd6f569e874 that is in Mike's tree. Please do as I asked you before
and rebase on top of the commit that you sent him, and make sure that
this is a commit that Mike never rebases.

	Arnd

vexpress issues in next-20121029

[Pawel Moll]

On Mon, 2012-11-05 at 16:47 +0000, Arnd Bergmann wrote:
> I've just tried pulling in your branch again, but it appears unchanged:
> The patches are still based on 807e45b328, which is a different commit
> from bcd6f569e874 that is in Mike's tree. Please do as I asked you before
> and rebase on top of the commit that you sent him,

Ok, I think I got the idea now. Sorry about not catching up straight
away...

The following changes since commit bcd6f569e87471d7f104bd9497f0b516a3b12e32:

  clk: Common clocks implementation for Versatile Express (2012-10-29 11:08:03 -0700)

are available in the git repository at:

  git://git.linaro.org/people/pawelmoll/linux.git vexpress-clk-soc

for you to fetch changes up to 433683a66401adb0150792e725cc4f631c94de46:

  ARM: vexpress: Remove motherboard dependencies in the DTS files (2012-11-05 17:09:52 +0000)

>  and make sure that this is a commit that Mike never rebases.

Uh. Mike, is your clk-next subject to rebases?

Pawe?

vexpress issues in next-20121029

[Arnd Bergmann]

On Monday 05 November 2012, Pawel Moll wrote:
> On Mon, 2012-11-05 at 16:47 +0000, Arnd Bergmann wrote:
> > I've just tried pulling in your branch again, but it appears unchanged:
> > The patches are still based on 807e45b328, which is a different commit
> > from bcd6f569e874 that is in Mike's tree. Please do as I asked you before
> > and rebase on top of the commit that you sent him,
> 
> Ok, I think I got the idea now. Sorry about not catching up straight
> away...
> 
> The following changes since commit bcd6f569e87471d7f104bd9497f0b516a3b12e32:
> 
>   clk: Common clocks implementation for Versatile Express (2012-10-29 11:08:03 -0700)
> 
> are available in the git repository at:
> 
>   git://git.linaro.org/people/pawelmoll/linux.git vexpress-clk-soc
> 
> for you to fetch changes up to 433683a66401adb0150792e725cc4f631c94de46:

Ok, thanks!

I've put it into the next/soc2 branch now, separate from the earlier next/soc
branch, since we now have a dependency on another branch.


>   ARM: vexpress: Remove motherboard dependencies in the DTS files (2012-11-05 17:09:52 +0000)
> 
> >  and make sure that this is a commit that Mike never rebases.
> 
> Uh. Mike, is your clk-next subject to rebases?

On a related note, there are other patches below this one now:

  clk: Common clocks implementation for Versatile Express
  clk: Versatile Express clock generators ("osc") driver
  CLK: clk-twl6040: Initial clock driver for OMAP4+ McPDM fclk clock
  clk: fix return value check in sirfsoc_of_clk_init()
  clk: fix return value check in of_fixed_clk_setup()
  clk: ux500: Update sdmmc clock to 100MHz for u8500
  clk: ux500: Support prcmu ape opp voltage clock
  mfd: dbx500: Export prmcu_request_ape_opp_100_voltage
  clk: Don't return negative numbers for unsigned values with !clk
  clk: Fix documentation typos
  clk: Document .is_enabled op
  clk: SPEAr: Vco-pll: Fix compilation warning

Mike, do you prefer us to wait for those to make it into v3.8-rc1 before
pushing the patches from Pawel, or can we just send the entire branch
along with the other changes?

FWIW, there is a way to avoid dependencies like this if the patches that
are required in two branches are based directly on an -rc release and
then merged into the two maintainer trees, rather than having one maintainer
tree pull in (part of) the history of another one.

	Arnd

vexpress issues in next-20121029

[Pawel Moll]

On Mon, 2012-11-05 at 17:52 +0000, Arnd Bergmann wrote:
> Mike, do you prefer us to wait for those to make it into v3.8-rc1 before
> pushing the patches from Pawel, or can we just send the entire branch
> along with the other changes?

Or, maybe, you prefer to remove the vexpress patches from clk-next
completely (if it is to be rebased) and let them go only through
arm-soc? (just asking)

Pawe?

vexpress issues in next-20121029

[Stephen Warren]

On 10/29/2012 02:24 PM, Stephen Warren wrote:
> Pawel,
> 
> I see two issues on vexpress in next-20121029:
...
> 2) With those calls commented out, I find that
> vexpress_sysreg_init_leds() is device_initcall, and so executes even
> when not running on vexpress HW. This crashes on Tegra (which I have
> converted to single-zImage locally).

This issue still appears to be present in next-20121107.

vexpress issues in next-20121029

[Pawel Moll]

On Wed, 2012-11-07 at 20:36 +0000, Stephen Warren wrote:
> On 10/29/2012 02:24 PM, Stephen Warren wrote:
> > 2) With those calls commented out, I find that
> > vexpress_sysreg_init_leds() is device_initcall, and so executes even
> > when not running on vexpress HW. This crashes on Tegra (which I have
> > converted to single-zImage locally).
> 
> This issue still appears to be present in next-20121107.

Yes, sorry about that - last week I've posted patches adding a generic
memory mapped leds driver and using it instead, but haven't managed to
get any comments from the maintainers. I've pinged them today and if I
hear nothing till tomorrow, I'll remove the offending code from
vexpress-sysreg.c completely and wait will this stuff till next cycle.

Pawe?