diff mbox

ARM: imx6: fix SMP compilation again

Message ID 16557676.B1fW0k49Wz@wuerfel
State New
Headers show

Commit Message

Arnd Bergmann July 22, 2014, 10:26 a.m. UTC 
On Tuesday 22 July 2014 16:50:56 Shawn Guo wrote:
> On Tue, Jul 22, 2014 at 10:30:50AM +0200, Arnd Bergmann wrote:
> > My earlier patch 1fc593feaf8e ("ARM: imx: build i.MX6 functions
> > only when needed") fixed a problem with building an i.MX5 kernel,
> > but missed the corner case of building a kernel for i.MX5 with
> > SMP enabled. It is an extremely rare case that has only now
> > happened after many thousands of randconfig kernel builds.
> > 
> > The error message I get is
> > arch/arm/mach-imx/built-in.o: In function `v7_secondary_startup':
> > :(.text+0x5124): undefined reference to `v7_invalidate_l1'
> 
> I cannot reproduce it, and unsure how it happens.
> 
> The v7_invalidate_l1() is defined in arch/arm/mm/cache-v7.S, and should
> be available for i.MX5 build.
> 

Ah, you are right, I got the text wrong here, and I think I now understand
why it didn't happen earlier. Updated patch below (same patch, new
text).

8<-------------
Subject: [PATCH] ARM: imx6: fix SMP compilation again

My earlier patch 1fc593feaf8e ("ARM: imx: build i.MX6 functions
only when needed") fixed a problem with building an i.MX5 kernel,
but now the problem has returned for the case where we allow
ARMv6K SMP builds in multiplatform. With CONFIG_CPU_V7 disabled,
but i.MX3 and SMP enabled, we get this build error:

arch/arm/mach-imx/built-in.o: In function `v7_secondary_startup':
:(.text+0x5124): undefined reference to `v7_invalidate_l1'

This puts the code inside of an "ifdef CONFIG_SMP" to hopefully
do the right thing in all configurations.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>

Comments

Shawn Guo July 22, 2014, 1:48 p.m. UTC | #1 
On Tue, Jul 22, 2014 at 12:26:31PM +0200, Arnd Bergmann wrote:
> Ah, you are right, I got the text wrong here, and I think I now understand
> why it didn't happen earlier. Updated patch below (same patch, new
> text).
> 
> 8<-------------
> Subject: [PATCH] ARM: imx6: fix SMP compilation again
> 
> My earlier patch 1fc593feaf8e ("ARM: imx: build i.MX6 functions
> only when needed") fixed a problem with building an i.MX5 kernel,
> but now the problem has returned for the case where we allow
> ARMv6K SMP builds in multiplatform. With CONFIG_CPU_V7 disabled,
> but i.MX3 and SMP enabled, we get this build error:

Maybe I missed something, but I have problem to get the following three
conditions meet in a single kernel configuration.

 a) CONFIG_CPU_V7 disabled
 b) i.MX3 enabled
 c) SMP enabled

When I get a) and b) in the config, I have no way to get c).  And it
seems that the only way to get a) and c) at the same time is to enable
MACH_REALVIEW_PB11MP, which is not part of multi-platform support.  So
i.MX3 cannot be enabled in there.

I tried both mainline and -next tree.  I really need some help to
reproduce the error first.

> 
> arch/arm/mach-imx/built-in.o: In function `v7_secondary_startup':
> :(.text+0x5124): undefined reference to `v7_invalidate_l1'
> 
> This puts the code inside of an "ifdef CONFIG_SMP" to hopefully

The code says "ifdef CONFIG_SOC_IMX6"?

Shawn

> do the right thing in all configurations.
> 
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
> 
> diff --git a/arch/arm/mach-imx/Makefile b/arch/arm/mach-imx/Makefile
> index ac88599ca080..23c02932bf84 100644
> --- a/arch/arm/mach-imx/Makefile
> +++ b/arch/arm/mach-imx/Makefile
> @@ -93,9 +93,11 @@ obj-$(CONFIG_HAVE_IMX_ANATOP) += anatop.o
>  obj-$(CONFIG_HAVE_IMX_GPC) += gpc.o
>  obj-$(CONFIG_HAVE_IMX_MMDC) += mmdc.o
>  obj-$(CONFIG_HAVE_IMX_SRC) += src.o
> +ifdef CONFIG_SOC_IMX6
>  AFLAGS_headsmp.o :=-Wa,-march=armv7-a
>  obj-$(CONFIG_SMP) += headsmp.o platsmp.o
>  obj-$(CONFIG_HOTPLUG_CPU) += hotplug.o
> +endif
>  obj-$(CONFIG_SOC_IMX6Q) += clk-imx6q.o mach-imx6q.o
>  obj-$(CONFIG_SOC_IMX6SL) += clk-imx6sl.o mach-imx6sl.o
>  obj-$(CONFIG_SOC_IMX6SX) += clk-imx6sx.o mach-imx6sx.o
>
Uwe Kleine-König July 22, 2014, 2:31 p.m. UTC | #2 
On Tue, Jul 22, 2014 at 09:48:16PM +0800, Shawn Guo wrote:
> On Tue, Jul 22, 2014 at 12:26:31PM +0200, Arnd Bergmann wrote:
> > Ah, you are right, I got the text wrong here, and I think I now understand
> > why it didn't happen earlier. Updated patch below (same patch, new
> > text).
> > 
> > 8<-------------
> > Subject: [PATCH] ARM: imx6: fix SMP compilation again
> > 
> > My earlier patch 1fc593feaf8e ("ARM: imx: build i.MX6 functions
> > only when needed") fixed a problem with building an i.MX5 kernel,
> > but now the problem has returned for the case where we allow
> > ARMv6K SMP builds in multiplatform. With CONFIG_CPU_V7 disabled,
> > but i.MX3 and SMP enabled, we get this build error:
> 
> Maybe I missed something, but I have problem to get the following three
> conditions meet in a single kernel configuration.
> 
>  a) CONFIG_CPU_V7 disabled
>  b) i.MX3 enabled
>  c) SMP enabled
> 
> When I get a) and b) in the config, I have no way to get c).  And it
> seems that the only way to get a) and c) at the same time is to enable
> MACH_REALVIEW_PB11MP, which is not part of multi-platform support.  So
> i.MX3 cannot be enabled in there.
> 
> I tried both mainline and -next tree.  I really need some help to
> reproduce the error first.
Without even looking at the patch, I think the key here is this part of
Arnd's commit log:

	for the case where we allow ARMv6K SMP builds in multiplatform
	...

Best regards
Uwe
Arnd Bergmann July 22, 2014, 2:37 p.m. UTC | #3 
On Tuesday 22 July 2014 21:48:16 Shawn Guo wrote:
> On Tue, Jul 22, 2014 at 12:26:31PM +0200, Arnd Bergmann wrote:
> > Ah, you are right, I got the text wrong here, and I think I now understand
> > why it didn't happen earlier. Updated patch below (same patch, new
> > text).
> > 
> > 8<-------------
> > Subject: [PATCH] ARM: imx6: fix SMP compilation again
> > 
> > My earlier patch 1fc593feaf8e ("ARM: imx: build i.MX6 functions
> > only when needed") fixed a problem with building an i.MX5 kernel,
> > but now the problem has returned for the case where we allow
> > ARMv6K SMP builds in multiplatform. With CONFIG_CPU_V7 disabled,
> > but i.MX3 and SMP enabled, we get this build error:
> 
> Maybe I missed something, but I have problem to get the following three
> conditions meet in a single kernel configuration.
> 
>  a) CONFIG_CPU_V7 disabled
>  b) i.MX3 enabled
>  c) SMP enabled
> 
> When I get a) and b) in the config, I have no way to get c).  And it
> seems that the only way to get a) and c) at the same time is to enable
> MACH_REALVIEW_PB11MP, which is not part of multi-platform support.  So
> i.MX3 cannot be enabled in there.

I have a patch series to make ARCH_REALVIEW multiplatform-enabled,
and I guess the condition can only be hit with these patches, or any
other that selects HAS_SMP.

> I tried both mainline and -next tree.  I really need some help to
> reproduce the error first.

My branch is at git.kernel.org/pub/scm/linux/kernel/git/arnd/playground.git
in the randconfig-next branch. Sorry for the inconvenience.

> > arch/arm/mach-imx/built-in.o: In function `v7_secondary_startup':
> > :(.text+0x5124): undefined reference to `v7_invalidate_l1'
> > 
> > This puts the code inside of an "ifdef CONFIG_SMP" to hopefully
> 
> The code says "ifdef CONFIG_SOC_IMX6"?

It seems I'm having a bad day. I'll fix it up.

	Arnd
Shawn Guo July 22, 2014, 2:52 p.m. UTC | #4 
On Tue, Jul 22, 2014 at 04:37:55PM +0200, Arnd Bergmann wrote:
> On Tuesday 22 July 2014 21:48:16 Shawn Guo wrote:
> > I tried both mainline and -next tree.  I really need some help to
> > reproduce the error first.
> 
> My branch is at git.kernel.org/pub/scm/linux/kernel/git/arnd/playground.git
> in the randconfig-next branch. Sorry for the inconvenience.

Okay, I can see the error on your branch with the fixing patch reverted.

> 
> > > arch/arm/mach-imx/built-in.o: In function `v7_secondary_startup':
> > > :(.text+0x5124): undefined reference to `v7_invalidate_l1'
> > > 
> > > This puts the code inside of an "ifdef CONFIG_SMP" to hopefully
> > 
> > The code says "ifdef CONFIG_SOC_IMX6"?
> 
> It seems I'm having a bad day. I'll fix it up.

With that fixed up,

Acked-by: Shawn Guo <shawn.guo@freescale.com>

Or let me know if you expect me to handle the patch.

Shawn
Arnd Bergmann July 22, 2014, 6:52 p.m. UTC | #5 
On Tuesday 22 July 2014 22:52:57 Shawn Guo wrote:
> On Tue, Jul 22, 2014 at 04:37:55PM +0200, Arnd Bergmann wrote:
> > On Tuesday 22 July 2014 21:48:16 Shawn Guo wrote:
> > > I tried both mainline and -next tree.  I really need some help to
> > > reproduce the error first.
> > 
> > My branch is at git.kernel.org/pub/scm/linux/kernel/git/arnd/playground.git
> > in the randconfig-next branch. Sorry for the inconvenience.
> 
> Okay, I can see the error on your branch with the fixing patch reverted.
> 
> > 
> > > > arch/arm/mach-imx/built-in.o: In function `v7_secondary_startup':
> > > > :(.text+0x5124): undefined reference to `v7_invalidate_l1'
> > > > 
> > > > This puts the code inside of an "ifdef CONFIG_SMP" to hopefully
> > > 
> > > The code says "ifdef CONFIG_SOC_IMX6"?
> > 
> > It seems I'm having a bad day. I'll fix it up.
> 
> With that fixed up,
> 
> Acked-by: Shawn Guo <shawn.guo@freescale.com>
> 
> Or let me know if you expect me to handle the patch.
> 

If you don't mind, just put it into your tree. I'm currently in the
middle of going through my older patches and don't want to lose this one.

	Arnd
Shawn Guo July 23, 2014, 5:02 a.m. UTC | #6 
On Tue, Jul 22, 2014 at 08:52:58PM +0200, Arnd Bergmann wrote:
> On Tuesday 22 July 2014 22:52:57 Shawn Guo wrote:
> > On Tue, Jul 22, 2014 at 04:37:55PM +0200, Arnd Bergmann wrote:
> > > On Tuesday 22 July 2014 21:48:16 Shawn Guo wrote:
> > > > I tried both mainline and -next tree.  I really need some help to
> > > > reproduce the error first.
> > > 
> > > My branch is at git.kernel.org/pub/scm/linux/kernel/git/arnd/playground.git
> > > in the randconfig-next branch. Sorry for the inconvenience.
> > 
> > Okay, I can see the error on your branch with the fixing patch reverted.
> > 
> > > 
> > > > > arch/arm/mach-imx/built-in.o: In function `v7_secondary_startup':
> > > > > :(.text+0x5124): undefined reference to `v7_invalidate_l1'
> > > > > 
> > > > > This puts the code inside of an "ifdef CONFIG_SMP" to hopefully
> > > > 
> > > > The code says "ifdef CONFIG_SOC_IMX6"?
> > > 
> > > It seems I'm having a bad day. I'll fix it up.
> > 
> > With that fixed up,
> > 
> > Acked-by: Shawn Guo <shawn.guo@freescale.com>
> > 
> > Or let me know if you expect me to handle the patch.
> > 
> 
> If you don't mind, just put it into your tree. I'm currently in the
> middle of going through my older patches and don't want to lose this one.

Okay, applied after fixing up the commit log.

Shawn
diff mbox

Patch

diff --git a/arch/arm/mach-imx/Makefile b/arch/arm/mach-imx/Makefile
index ac88599ca080..23c02932bf84 100644
--- a/arch/arm/mach-imx/Makefile
+++ b/arch/arm/mach-imx/Makefile
@@ -93,9 +93,11 @@  obj-$(CONFIG_HAVE_IMX_ANATOP) += anatop.o
 obj-$(CONFIG_HAVE_IMX_GPC) += gpc.o
 obj-$(CONFIG_HAVE_IMX_MMDC) += mmdc.o
 obj-$(CONFIG_HAVE_IMX_SRC) += src.o
+ifdef CONFIG_SOC_IMX6
 AFLAGS_headsmp.o :=-Wa,-march=armv7-a
 obj-$(CONFIG_SMP) += headsmp.o platsmp.o
 obj-$(CONFIG_HOTPLUG_CPU) += hotplug.o
+endif
 obj-$(CONFIG_SOC_IMX6Q) += clk-imx6q.o mach-imx6q.o
 obj-$(CONFIG_SOC_IMX6SL) += clk-imx6sl.o mach-imx6sl.o
 obj-$(CONFIG_SOC_IMX6SX) += clk-imx6sx.o mach-imx6sx.o