[PATCH] gpio: syscon: reduce message level when direction reg offset not in dt

[PATCH] gpio: syscon: reduce message level when direction reg offset not in dt

[grygorii.strashko]

From: Grygorii Strashko <grygorii.strashko@linaro.org>

Now GPIO syscon driver produces bunch of warnings during the
boot of Kesytone 2 SoCs:
 gpio-syscon soc:keystone_dsp_gpio@02620240: can't read the dir register offset!
 gpio-syscon soc:keystone_dsp_gpio@2620244: can't read the dir register offset!

This message unintentionally was added using dev_err(), but its
actual log level is debug, because third cell of "ti,syscon-dev" is
optional.

Hence change it to dev_dbg() as it should be.

This patch fixes commit:
 5a3e3f8 ("gpio: syscon: retriave syscon node and regs offsets from dt")

Reported-by: Russell King <linux@arm.linux.org.uk>
Signed-off-by: Grygorii Strashko <grygorii.strashko@linaro.org>
---
 drivers/gpio/gpio-syscon.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpio/gpio-syscon.c b/drivers/gpio/gpio-syscon.c
index 257e298..045a952 100644
--- a/drivers/gpio/gpio-syscon.c
+++ b/drivers/gpio/gpio-syscon.c
@@ -219,7 +219,7 @@ static int syscon_gpio_probe(struct platform_device *pdev)
 		ret = of_property_read_u32_index(np, "gpio,syscon-dev", 2,
 						 &priv->dir_reg_offset);
 		if (ret)
-			dev_err(dev, "can't read the dir register offset!\n");
+			dev_dbg(dev, "can't read the dir register offset!\n");
 
 		priv->dir_reg_offset <<= 3;
 	}
-- 
1.9.1

Re: [PATCH] gpio: syscon: reduce message level when direction reg offset not in dt

[Murali Karicheri]

On 03/24/2015 02:42 PM, grygorii.strashko@linaro.org wrote:
> From: Grygorii Strashko<grygorii.strashko@linaro.org>
>
> Now GPIO syscon driver produces bunch of warnings during the
> boot of Kesytone 2 SoCs:
>   gpio-syscon soc:keystone_dsp_gpio@02620240: can't read the dir register offset!
>   gpio-syscon soc:keystone_dsp_gpio@2620244: can't read the dir register offset!
>
> This message unintentionally was added using dev_err(), but its
> actual log level is debug, because third cell of "ti,syscon-dev" is
> optional.
>
> Hence change it to dev_dbg() as it should be.
>
> This patch fixes commit:
>   5a3e3f8 ("gpio: syscon: retriave syscon node and regs offsets from dt")
>
> Reported-by: Russell King<linux@arm.linux.org.uk>
> Signed-off-by: Grygorii Strashko<grygorii.strashko@linaro.org>
> ---
>   drivers/gpio/gpio-syscon.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/gpio/gpio-syscon.c b/drivers/gpio/gpio-syscon.c
> index 257e298..045a952 100644
> --- a/drivers/gpio/gpio-syscon.c
> +++ b/drivers/gpio/gpio-syscon.c
> @@ -219,7 +219,7 @@ static int syscon_gpio_probe(struct platform_device *pdev)
>   		ret = of_property_read_u32_index(np, "gpio,syscon-dev", 2,
>   						&priv->dir_reg_offset);
>   		if (ret)
> -			dev_err(dev, "can't read the dir register offset!\n");
> +			dev_dbg(dev, "can't read the dir register offset!\n");
>
>   		priv->dir_reg_offset<<= 3;
>   	}

Tested this on K2HK EVM and found the error log is not appearing any more.

Tested-by: Murali Karicheri <m-karicheri2@ti.com>

-- 
Murali Karicheri
Linux Kernel, Texas Instruments

Re: [PATCH] gpio: syscon: reduce message level when direction reg offset not in dt

[santosh shilimkar]

On 3/24/2015 11:42 AM, grygorii.strashko@linaro.org wrote:
> From: Grygorii Strashko <grygorii.strashko@linaro.org>
>
> Now GPIO syscon driver produces bunch of warnings during the
> boot of Kesytone 2 SoCs:
>   gpio-syscon soc:keystone_dsp_gpio@02620240: can't read the dir register offset!
>   gpio-syscon soc:keystone_dsp_gpio@2620244: can't read the dir register offset!
>
> This message unintentionally was added using dev_err(), but its
> actual log level is debug, because third cell of "ti,syscon-dev" is
> optional.
>
> Hence change it to dev_dbg() as it should be.
>
> This patch fixes commit:
>   5a3e3f8 ("gpio: syscon: retriave syscon node and regs offsets from dt")
>
> Reported-by: Russell King <linux@arm.linux.org.uk>
> Signed-off-by: Grygorii Strashko <grygorii.strashko@linaro.org>
> ---
Acked-by: Santosh Shilimkar <ssantosh@kernel.org>

Re: [PATCH] gpio: syscon: reduce message level when direction reg offset not in dt

[Linus Walleij]

On Tue, Mar 24, 2015 at 7:42 PM,  <grygorii.strashko@linaro.org> wrote:

> From: Grygorii Strashko <grygorii.strashko@linaro.org>
>
> Now GPIO syscon driver produces bunch of warnings during the
> boot of Kesytone 2 SoCs:
>  gpio-syscon soc:keystone_dsp_gpio@02620240: can't read the dir register offset!
>  gpio-syscon soc:keystone_dsp_gpio@2620244: can't read the dir register offset!
>
> This message unintentionally was added using dev_err(), but its
> actual log level is debug, because third cell of "ti,syscon-dev" is
> optional.
>
> Hence change it to dev_dbg() as it should be.
>
> This patch fixes commit:
>  5a3e3f8 ("gpio: syscon: retriave syscon node and regs offsets from dt")
>
> Reported-by: Russell King <linux@arm.linux.org.uk>
> Signed-off-by: Grygorii Strashko <grygorii.strashko@linaro.org>

OK sorry for missing this when asking about the strange messages.
Patch applied for fixes with these test tags and ack.

Yours,
Linus Walleij

[PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in gcc 4.7.3 (was: Re: Possible regression in gcc 4.7.3 next-20150323 due to "ARM, arm64: kvm: get rid of the bounce page")

[Geert Uytterhoeven]

	Hi Kevin,

On Tue, 31 Mar 2015, Kevin Hilman wrote:
> Ard Biesheuvel <ard.biesheuvel@linaro.org> writes:
> Nope, that branch is already part of linux-next, and linux-next still
> fails to compile for 20+ defconfigs[1]
> 
> > Could you elaborate on the issue please? What is the error you are
> > getting, and can you confirm that is is caused by ld choking on the
> > linker script? If not, this is another error than the one we have been
> > trying to fix
> 
> It's definitely not linker script related.
> 
> Using "arm-linux-gnueabi-gcc (Ubuntu/Linaro 4.7.3-12ubuntu1) 4.7.3",
> here's the error when building for multi_v7_defconfig (full log
> available[2]):
> 
> ../mm/migrate.c: In function 'migrate_pages':
> ../mm/migrate.c:1148:1: internal compiler error: in push_minipool_fix, at config/arm/arm.c:13101
> Please submit a full bug report,
> with preprocessed source if appropriate.
> See <file:///usr/share/doc/gcc-4.7/README.Bugs> for instructions.
> Preprocessed source stored into /tmp/ccO1Nz1m.out file, please attach
> this to your bugreport.
> make[2]: *** [mm/migrate.o] Error 1
> make[2]: Target `__build' not remade because of errors.
> make[1]: *** [mm] Error 2
> 
> build bisect points to commit 21f992084aeb[3], but that doesn't revert
> cleanly so I haven't got any further than that yet.

I installed gcc-arm-linux-gnueabi (4:4.7.2-1 from Ubuntu 14.04 LTS) and could
reproduce the ICE. I came up with the workaround below.
Does this work for you?

>From 7ebe83316eaf1952e55a76754ce7a5832e461b8c Mon Sep 17 00:00:00 2001
From: Geert Uytterhoeven <geert+renesas@glider.be>
Date: Wed, 1 Apr 2015 11:22:51 +0200
Subject: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in
 gcc 4.7.3
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

With gcc version 4.7.3 (Ubuntu/Linaro 4.7.3-12ubuntu1) :

    mm/migrate.c: In function ‘migrate_pages’:
    mm/migrate.c:1148:1: internal compiler error: in push_minipool_fix, at config/arm/arm.c:13500
    Please submit a full bug report,
    with preprocessed source if appropriate.
    See <file:///usr/share/doc/gcc-4.7/README.Bugs> for instructions.
    Preprocessed source stored into /tmp/ccPoM1tr.out file, please attach this to your bugreport.
    make[1]: *** [mm/migrate.o] Error 1
    make: *** [mm/migrate.o] Error 2

Mark unmap_and_move() (which is used in a single place only) "noinline"
to work around this compiler bug.

Reported-by: Kevin Hilman <khilman@kernel.org>
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
---
 mm/migrate.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/mm/migrate.c b/mm/migrate.c
index 114602a68111d809..98f8574456c2010c 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -904,9 +904,10 @@ out:
  * Obtain the lock on page, remove all ptes and migrate the page
  * to the newly allocated page in newpage.
  */
-static int unmap_and_move(new_page_t get_new_page, free_page_t put_new_page,
-			unsigned long private, struct page *page, int force,
-			enum migrate_mode mode)
+static noinline int unmap_and_move(new_page_t get_new_page,
+				   free_page_t put_new_page,
+				   unsigned long private, struct page *page,
+				   int force, enum migrate_mode mode)
 {
 	int rc = 0;
 	int *result = NULL;
-- 
1.9.1


Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds

Re: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in gcc 4.7.3

[Marc Zyngier]

On 01/04/15 10:37, Geert Uytterhoeven wrote:
> 	Hi Kevin,
> 
> On Tue, 31 Mar 2015, Kevin Hilman wrote:
>> Ard Biesheuvel <ard.biesheuvel@linaro.org> writes:
>> Nope, that branch is already part of linux-next, and linux-next still
>> fails to compile for 20+ defconfigs[1]
>>
>>> Could you elaborate on the issue please? What is the error you are
>>> getting, and can you confirm that is is caused by ld choking on the
>>> linker script? If not, this is another error than the one we have been
>>> trying to fix
>>
>> It's definitely not linker script related.
>>
>> Using "arm-linux-gnueabi-gcc (Ubuntu/Linaro 4.7.3-12ubuntu1) 4.7.3",
>> here's the error when building for multi_v7_defconfig (full log
>> available[2]):
>>
>> ../mm/migrate.c: In function 'migrate_pages':
>> ../mm/migrate.c:1148:1: internal compiler error: in push_minipool_fix, at config/arm/arm.c:13101
>> Please submit a full bug report,
>> with preprocessed source if appropriate.
>> See <file:///usr/share/doc/gcc-4.7/README.Bugs> for instructions.
>> Preprocessed source stored into /tmp/ccO1Nz1m.out file, please attach
>> this to your bugreport.
>> make[2]: *** [mm/migrate.o] Error 1
>> make[2]: Target `__build' not remade because of errors.
>> make[1]: *** [mm] Error 2
>>
>> build bisect points to commit 21f992084aeb[3], but that doesn't revert
>> cleanly so I haven't got any further than that yet.
> 
> I installed gcc-arm-linux-gnueabi (4:4.7.2-1 from Ubuntu 14.04 LTS) and could
> reproduce the ICE. I came up with the workaround below.
> Does this work for you?
> 
> From 7ebe83316eaf1952e55a76754ce7a5832e461b8c Mon Sep 17 00:00:00 2001
> From: Geert Uytterhoeven <geert+renesas@glider.be>
> Date: Wed, 1 Apr 2015 11:22:51 +0200
> Subject: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in
>  gcc 4.7.3
> MIME-Version: 1.0
> Content-Type: text/plain; charset=UTF-8
> Content-Transfer-Encoding: 8bit
> 
> With gcc version 4.7.3 (Ubuntu/Linaro 4.7.3-12ubuntu1) :
> 
>     mm/migrate.c: In function ‘migrate_pages’:
>     mm/migrate.c:1148:1: internal compiler error: in push_minipool_fix, at config/arm/arm.c:13500
>     Please submit a full bug report,
>     with preprocessed source if appropriate.
>     See <file:///usr/share/doc/gcc-4.7/README.Bugs> for instructions.
>     Preprocessed source stored into /tmp/ccPoM1tr.out file, please attach this to your bugreport.
>     make[1]: *** [mm/migrate.o] Error 1
>     make: *** [mm/migrate.o] Error 2
> 
> Mark unmap_and_move() (which is used in a single place only) "noinline"
> to work around this compiler bug.
> 
> Reported-by: Kevin Hilman <khilman@kernel.org>
> Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> ---
>  mm/migrate.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
> 
> diff --git a/mm/migrate.c b/mm/migrate.c
> index 114602a68111d809..98f8574456c2010c 100644
> --- a/mm/migrate.c
> +++ b/mm/migrate.c
> @@ -904,9 +904,10 @@ out:
>   * Obtain the lock on page, remove all ptes and migrate the page
>   * to the newly allocated page in newpage.
>   */
> -static int unmap_and_move(new_page_t get_new_page, free_page_t put_new_page,
> -			unsigned long private, struct page *page, int force,
> -			enum migrate_mode mode)
> +static noinline int unmap_and_move(new_page_t get_new_page,
> +				   free_page_t put_new_page,
> +				   unsigned long private, struct page *page,
> +				   int force, enum migrate_mode mode)
>  {
>  	int rc = 0;
>  	int *result = NULL;
> 

Ouch. That's really ugly. And on 32bit ARM, we end-up spilling half of
the parameters on the stack, which is not going to help performance
either (not that this would be useful on 32bit ARM anyway...).

Any chance you could make this dependent on some compiler detection
mechanism?

Thanks,

	M.
-- 
Jazz is not dead. It just smells funny...

Re: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in gcc 4.7.3

[Andrew Morton]

On Wed, 01 Apr 2015 10:47:49 +0100 Marc Zyngier <marc.zyngier@arm.com> wrote:

> > -static int unmap_and_move(new_page_t get_new_page, free_page_t put_new_page,
> > -			unsigned long private, struct page *page, int force,
> > -			enum migrate_mode mode)
> > +static noinline int unmap_and_move(new_page_t get_new_page,
> > +				   free_page_t put_new_page,
> > +				   unsigned long private, struct page *page,
> > +				   int force, enum migrate_mode mode)
> >  {
> >  	int rc = 0;
> >  	int *result = NULL;
> > 
> 
> Ouch. That's really ugly. And on 32bit ARM, we end-up spilling half of
> the parameters on the stack, which is not going to help performance
> either (not that this would be useful on 32bit ARM anyway...).
> 
> Any chance you could make this dependent on some compiler detection
> mechanism?

With my arm compiler (gcc-4.4.4) the patch makes no difference -
unmap_and_move() isn't being inlined anyway.

How does this look?

Kevin, could you please retest?  I might have fat-fingered something...

--- a/mm/migrate.c~mm-migrate-mark-unmap_and_move-noinline-to-avoid-ice-in-gcc-473-fix
+++ a/mm/migrate.c
@@ -901,10 +901,20 @@ out:
 }
 
 /*
+ * gcc-4.7.3 on arm gets an ICE when inlining unmap_and_move().  Work around
+ * it.
+ */
+#if GCC_VERSION == 40703 && defined(CONFIG_ARM)
+#define ICE_noinline noinline
+#else
+#define ICE_noinline
+#endif
+
+/*
  * Obtain the lock on page, remove all ptes and migrate the page
  * to the newly allocated page in newpage.
  */
-static noinline int unmap_and_move(new_page_t get_new_page,
+static ICE_noinline int unmap_and_move(new_page_t get_new_page,
 				   free_page_t put_new_page,
 				   unsigned long private, struct page *page,
 				   int force, enum migrate_mode mode)
_

Re: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in gcc 4.7.3

[Kevin Hilman]

Andrew Morton <akpm@linux-foundation.org> writes:

> On Wed, 01 Apr 2015 10:47:49 +0100 Marc Zyngier <marc.zyngier@arm.com> wrote:
>
>> > -static int unmap_and_move(new_page_t get_new_page, free_page_t put_new_page,
>> > -			unsigned long private, struct page *page, int force,
>> > -			enum migrate_mode mode)
>> > +static noinline int unmap_and_move(new_page_t get_new_page,
>> > +				   free_page_t put_new_page,
>> > +				   unsigned long private, struct page *page,
>> > +				   int force, enum migrate_mode mode)
>> >  {
>> >  	int rc = 0;
>> >  	int *result = NULL;
>> > 
>> 
>> Ouch. That's really ugly. And on 32bit ARM, we end-up spilling half of
>> the parameters on the stack, which is not going to help performance
>> either (not that this would be useful on 32bit ARM anyway...).
>> 
>> Any chance you could make this dependent on some compiler detection
>> mechanism?
>
> With my arm compiler (gcc-4.4.4) the patch makes no difference -
> unmap_and_move() isn't being inlined anyway.
>
> How does this look?
>
> Kevin, could you please retest?  I might have fat-fingered something...

Your patch on top of Geert's still compiles fine for me with gcc-4.7.3.
However, I'm not sure how specific we can be on the versions.  

/me goes to test a few more compilers...   OK...

ICE: 4.7.1, 4.7.3, 4.8.3
OK: 4.6.3, 4.9.2, 4.9.3

The diff below[2] on top of yours compiles fine here and at least covers
the compilers I *know* to trigger the ICE.

Kevin


[1]
diff --git a/mm/migrate.c b/mm/migrate.c
index 25fd7f6291de..6e15ae3248e0 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -901,10 +901,10 @@ out:
 }
 
 /*
- * gcc-4.7.3 on arm gets an ICE when inlining unmap_and_move().  Work around
+ * gcc 4.7 and 4.8 on arm gets an ICE when inlining unmap_and_move().  Work around
  * it.
  */
-#if GCC_VERSION == 40703 && defined(CONFIG_ARM)
+#if (GCC_VERSION >= 40700 && GCC_VERSION < 40900) && defined(CONFIG_ARM)
 #define ICE_noinline noinline
 #else
 #define ICE_noinline

Re: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in gcc 4.7.3

[Russell King - ARM Linux]

On Wed, Apr 01, 2015 at 02:54:59PM -0700, Kevin Hilman wrote:
> Your patch on top of Geert's still compiles fine for me with gcc-4.7.3.
> However, I'm not sure how specific we can be on the versions.  
> 
> /me goes to test a few more compilers...   OK...
> 
> ICE: 4.7.1, 4.7.3, 4.8.3
> OK: 4.6.3, 4.9.2, 4.9.3
> 
> The diff below[2] on top of yours compiles fine here and at least covers
> the compilers I *know* to trigger the ICE.

Interesting.  I'm using stock gcc 4.7.4 here, though I'm not building
-next (only mainline + my tree + arm-soc) and it hasn't shown a problem
yet.

I think we need to ask the question: is the bug in stock GCC or Linaro
GCC?  If it's not in stock GCC, then it's a GCC vendor problem :)

-- 
FTTC broadband for 0.8mile line: currently at 10.5Mbps down 400kbps up
according to speedtest.net.

Re: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in gcc 4.7.3

[Geert Uytterhoeven]

Hi Russell,

On Wed, Apr 1, 2015 at 11:59 PM, Russell King - ARM Linux
<linux@arm.linux.org.uk> wrote:
> On Wed, Apr 01, 2015 at 02:54:59PM -0700, Kevin Hilman wrote:
>> Your patch on top of Geert's still compiles fine for me with gcc-4.7.3.
>> However, I'm not sure how specific we can be on the versions.
>>
>> /me goes to test a few more compilers...   OK...
>>
>> ICE: 4.7.1, 4.7.3, 4.8.3
>> OK: 4.6.3, 4.9.2, 4.9.3
>>
>> The diff below[2] on top of yours compiles fine here and at least covers
>> the compilers I *know* to trigger the ICE.
>
> Interesting.  I'm using stock gcc 4.7.4 here, though I'm not building
> -next (only mainline + my tree + arm-soc) and it hasn't shown a problem
> yet.

Mainline doesn't fail.

> I think we need to ask the question: is the bug in stock GCC or Linaro
> GCC?  If it's not in stock GCC, then it's a GCC vendor problem :)

Can you please try -next (e.g. next-20150320)?

make bockw_defconfig
make mm/migrate.o

Thanks!

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in gcc 4.7.3

[Lina Iyer]

On Wed, Apr 01 2015 at 15:57 -0600, Kevin Hilman wrote:
>Andrew Morton <akpm@linux-foundation.org> writes:
>
>> On Wed, 01 Apr 2015 10:47:49 +0100 Marc Zyngier <marc.zyngier@arm.com> wrote:
>>
>>> > -static int unmap_and_move(new_page_t get_new_page, free_page_t put_new_page,
>>> > -			unsigned long private, struct page *page, int force,
>>> > -			enum migrate_mode mode)
>>> > +static noinline int unmap_and_move(new_page_t get_new_page,
>>> > +				   free_page_t put_new_page,
>>> > +				   unsigned long private, struct page *page,
>>> > +				   int force, enum migrate_mode mode)
>>> >  {
>>> >  	int rc = 0;
>>> >  	int *result = NULL;
>>> >
>>>
>>> Ouch. That's really ugly. And on 32bit ARM, we end-up spilling half of
>>> the parameters on the stack, which is not going to help performance
>>> either (not that this would be useful on 32bit ARM anyway...).
>>>
>>> Any chance you could make this dependent on some compiler detection
>>> mechanism?
>>
>> With my arm compiler (gcc-4.4.4) the patch makes no difference -
>> unmap_and_move() isn't being inlined anyway.
>>
>> How does this look?
>>
>> Kevin, could you please retest?  I might have fat-fingered something...
>
>Your patch on top of Geert's still compiles fine for me with gcc-4.7.3.
>However, I'm not sure how specific we can be on the versions.
>
>/me goes to test a few more compilers...   OK...
>
>ICE: 4.7.1, 4.7.3, 4.8.3
>OK: 4.6.3, 4.9.2, 4.9.3
>
>The diff below[2] on top of yours compiles fine here and at least covers
>the compilers I *know* to trigger the ICE.

I see ICE on 
arm-linux-gnueabi-gcc (Ubuntu/Linaro 4.7.4-2ubuntu1) 4.7.4

>
>Kevin
>
>
>[1]
>diff --git a/mm/migrate.c b/mm/migrate.c
>index 25fd7f6291de..6e15ae3248e0 100644
>--- a/mm/migrate.c
>+++ b/mm/migrate.c
>@@ -901,10 +901,10 @@ out:
> }
>
> /*
>- * gcc-4.7.3 on arm gets an ICE when inlining unmap_and_move().  Work around
>+ * gcc 4.7 and 4.8 on arm gets an ICE when inlining unmap_and_move().  Work around
>  * it.
>  */
>-#if GCC_VERSION == 40703 && defined(CONFIG_ARM)
>+#if (GCC_VERSION >= 40700 && GCC_VERSION < 40900) && defined(CONFIG_ARM)
> #define ICE_noinline noinline
> #else
> #define ICE_noinline
>
>_______________________________________________
>linux-arm-kernel mailing list
>linux-arm-kernel@lists.infradead.org
>http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in gcc 4.7.3

[Kevin Hilman]

On Thu, Apr 2, 2015 at 12:12 PM, Lina Iyer <lina.iyer@linaro.org> wrote:
> On Wed, Apr 01 2015 at 15:57 -0600, Kevin Hilman wrote:
>>
>> Andrew Morton <akpm@linux-foundation.org> writes:
>>
>>> On Wed, 01 Apr 2015 10:47:49 +0100 Marc Zyngier <marc.zyngier@arm.com>
>>> wrote:
>>>
>>>> > -static int unmap_and_move(new_page_t get_new_page, free_page_t
>>>> > put_new_page,
>>>> > -                     unsigned long private, struct page *page, int
>>>> > force,
>>>> > -                     enum migrate_mode mode)
>>>> > +static noinline int unmap_and_move(new_page_t get_new_page,
>>>> > +                                free_page_t put_new_page,
>>>> > +                                unsigned long private, struct page
>>>> > *page,
>>>> > +                                int force, enum migrate_mode mode)
>>>> >  {
>>>> >       int rc = 0;
>>>> >       int *result = NULL;
>>>> >
>>>>
>>>> Ouch. That's really ugly. And on 32bit ARM, we end-up spilling half of
>>>> the parameters on the stack, which is not going to help performance
>>>> either (not that this would be useful on 32bit ARM anyway...).
>>>>
>>>> Any chance you could make this dependent on some compiler detection
>>>> mechanism?
>>>
>>>
>>> With my arm compiler (gcc-4.4.4) the patch makes no difference -
>>> unmap_and_move() isn't being inlined anyway.
>>>
>>> How does this look?
>>>
>>> Kevin, could you please retest?  I might have fat-fingered something...
>>
>>
>> Your patch on top of Geert's still compiles fine for me with gcc-4.7.3.
>> However, I'm not sure how specific we can be on the versions.
>>
>> /me goes to test a few more compilers...   OK...
>>
>> ICE: 4.7.1, 4.7.3, 4.8.3
>> OK: 4.6.3, 4.9.2, 4.9.3
>>
>> The diff below[2] on top of yours compiles fine here and at least covers
>> the compilers I *know* to trigger the ICE.
>
>
> I see ICE on arm-linux-gnueabi-gcc (Ubuntu/Linaro 4.7.4-2ubuntu1) 4.7.4
>

Thanks for checking.  I'm assuming my patch fixes it for your since
that should catch any 4.7.x compiler.

Kevin

Re: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in gcc 4.7.3

[Lina Iyer]

On Thu, Apr 02 2015 at 15:12 -0600, Kevin Hilman wrote:
>On Thu, Apr 2, 2015 at 12:12 PM, Lina Iyer <lina.iyer@linaro.org> wrote:
>> On Wed, Apr 01 2015 at 15:57 -0600, Kevin Hilman wrote:
>>>
>>> Andrew Morton <akpm@linux-foundation.org> writes:
>>>
>>>> On Wed, 01 Apr 2015 10:47:49 +0100 Marc Zyngier <marc.zyngier@arm.com>
>>>> wrote:
>>>>
>>>>> > -static int unmap_and_move(new_page_t get_new_page, free_page_t
>>>>> > put_new_page,
>>>>> > -                     unsigned long private, struct page *page, int
>>>>> > force,
>>>>> > -                     enum migrate_mode mode)
>>>>> > +static noinline int unmap_and_move(new_page_t get_new_page,
>>>>> > +                                free_page_t put_new_page,
>>>>> > +                                unsigned long private, struct page
>>>>> > *page,
>>>>> > +                                int force, enum migrate_mode mode)
>>>>> >  {
>>>>> >       int rc = 0;
>>>>> >       int *result = NULL;
>>>>> >
>>>>>
>>>>> Ouch. That's really ugly. And on 32bit ARM, we end-up spilling half of
>>>>> the parameters on the stack, which is not going to help performance
>>>>> either (not that this would be useful on 32bit ARM anyway...).
>>>>>
>>>>> Any chance you could make this dependent on some compiler detection
>>>>> mechanism?
>>>>
>>>>
>>>> With my arm compiler (gcc-4.4.4) the patch makes no difference -
>>>> unmap_and_move() isn't being inlined anyway.
>>>>
>>>> How does this look?
>>>>
>>>> Kevin, could you please retest?  I might have fat-fingered something...
>>>
>>>
>>> Your patch on top of Geert's still compiles fine for me with gcc-4.7.3.
>>> However, I'm not sure how specific we can be on the versions.
>>>
>>> /me goes to test a few more compilers...   OK...
>>>
>>> ICE: 4.7.1, 4.7.3, 4.8.3
>>> OK: 4.6.3, 4.9.2, 4.9.3
>>>
>>> The diff below[2] on top of yours compiles fine here and at least covers
>>> the compilers I *know* to trigger the ICE.
>>
>>
>> I see ICE on arm-linux-gnueabi-gcc (Ubuntu/Linaro 4.7.4-2ubuntu1) 4.7.4
>>
>
>Thanks for checking.  I'm assuming my patch fixes it for your since
>that should catch any 4.7.x compiler.

Yes, thank you. This fixes it on 4.7.4
>
>Kevin

Re: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in gcc 4.7.3

[Kevin Hilman]

Hi Andrew,

On Wed, Apr 1, 2015 at 2:54 PM, Kevin Hilman <khilman@kernel.org> wrote:
> Andrew Morton <akpm@linux-foundation.org> writes:
>
>> On Wed, 01 Apr 2015 10:47:49 +0100 Marc Zyngier <marc.zyngier@arm.com> wrote:
>>
>>> > -static int unmap_and_move(new_page_t get_new_page, free_page_t put_new_page,
>>> > -                  unsigned long private, struct page *page, int force,
>>> > -                  enum migrate_mode mode)
>>> > +static noinline int unmap_and_move(new_page_t get_new_page,
>>> > +                             free_page_t put_new_page,
>>> > +                             unsigned long private, struct page *page,
>>> > +                             int force, enum migrate_mode mode)
>>> >  {
>>> >    int rc = 0;
>>> >    int *result = NULL;
>>> >
>>>
>>> Ouch. That's really ugly. And on 32bit ARM, we end-up spilling half of
>>> the parameters on the stack, which is not going to help performance
>>> either (not that this would be useful on 32bit ARM anyway...).
>>>
>>> Any chance you could make this dependent on some compiler detection
>>> mechanism?
>>
>> With my arm compiler (gcc-4.4.4) the patch makes no difference -
>> unmap_and_move() isn't being inlined anyway.
>>
>> How does this look?
>>
>> Kevin, could you please retest?  I might have fat-fingered something...
>
> Your patch on top of Geert's still compiles fine for me with gcc-4.7.3.
> However, I'm not sure how specific we can be on the versions.
>
> /me goes to test a few more compilers...   OK...
>
> ICE: 4.7.1, 4.7.3, 4.8.3
> OK: 4.6.3, 4.9.2, 4.9.3
>
> The diff below[2] on top of yours compiles fine here and at least covers
> the compilers I *know* to trigger the ICE.

I see my fix in your mmots since last Thurs (4/2), but it's not in
mmotm (last updated today) so today's linux-next still has the ICE for
anything other than gcc-4.7.3.   Just checking to see when you plan to
update mmotm.

Thanks,

Kevin

Re: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in gcc 4.7.3

[Andrew Morton]

On Tue, 7 Apr 2015 10:57:52 -0700 Kevin Hilman <khilman@kernel.org> wrote:

> > The diff below[2] on top of yours compiles fine here and at least covers
> > the compilers I *know* to trigger the ICE.
> 
> I see my fix in your mmots since last Thurs (4/2), but it's not in
> mmotm (last updated today) so today's linux-next still has the ICE for
> anything other than gcc-4.7.3.   Just checking to see when you plan to
> update mmotm.

It should all be there today?

Re: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in gcc 4.7.3

[Kevin Hilman]

Andrew Morton <akpm@linux-foundation.org> writes:

> On Tue, 7 Apr 2015 10:57:52 -0700 Kevin Hilman <khilman@kernel.org> wrote:
>
>> > The diff below[2] on top of yours compiles fine here and at least covers
>> > the compilers I *know* to trigger the ICE.
>> 
>> I see my fix in your mmots since last Thurs (4/2), but it's not in
>> mmotm (last updated today) so today's linux-next still has the ICE for
>> anything other than gcc-4.7.3.   Just checking to see when you plan to
>> update mmotm.
>
> It should all be there today?

Nope.  

In mmotm, only the original patch plus your first fix is there:

$ curl -sO http://www.ozlabs.org/~akpm/mmotm/broken-out.tar.gz
$ tar -tavf broken-out.tar.gz |grep gcc-473
-rw-r----- akpm/eng       1838 2015-04-01 14:41 broken-out/mm-migrate-mark-unmap_and_move-noinline-to-avoid-ice-in-gcc-473.patch
-rw-r----- akpm/eng       1309 2015-04-01 14:41 broken-out/mm-migrate-mark-unmap_and_move-noinline-to-avoid-ice-in-gcc-473-fix.patch

but in mmots, the additional ptch from me, plus another comment fixup
from you are also there:

$ curl -sO http://www.ozlabs.org/~akpm/mmots/broken-out.tar.gz
$ tar -tavf broken-out.tar.gz |grep gcc-473
-rw-r----- akpm/eng       1882 2015-04-06 16:24 broken-out/mm-migrate-mark-unmap_and_move-noinline-to-avoid-ice-in-gcc-473.patch
-rw-r----- akpm/eng       1271 2015-04-06 16:24 broken-out/mm-migrate-mark-unmap_and_move-noinline-to-avoid-ice-in-gcc-473-fix.patch
-rw-r----- akpm/eng       1382 2015-04-06 16:24 broken-out/mm-migrate-mark-unmap_and_move-noinline-to-avoid-ice-in-gcc-473-fix-fix.patch
-rw-r----- akpm/eng        968 2015-04-06 16:24 broken-out/mm-migrate-mark-unmap_and_move-noinline-to-avoid-ice-in-gcc-473-fix-fix-fix.patch


Kevin

Re: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in gcc 4.7.3

[Andrew Morton]

On Tue, 07 Apr 2015 15:41:32 -0700 Kevin Hilman <khilman@kernel.org> wrote:

> Andrew Morton <akpm@linux-foundation.org> writes:
> 
> > On Tue, 7 Apr 2015 10:57:52 -0700 Kevin Hilman <khilman@kernel.org> wrote:
> >
> >> > The diff below[2] on top of yours compiles fine here and at least covers
> >> > the compilers I *know* to trigger the ICE.
> >> 
> >> I see my fix in your mmots since last Thurs (4/2), but it's not in
> >> mmotm (last updated today) so today's linux-next still has the ICE for
> >> anything other than gcc-4.7.3.   Just checking to see when you plan to
> >> update mmotm.
> >
> > It should all be there today?
> 
> Nope.  

huh, I swear I did an mmotm yesterday.

Let me see if I can sort out the watchdog mess and produce something
releasable...

Re: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in gcc 4.7.3

[Kevin Hilman]

Andrew Morton <akpm@linux-foundation.org> writes:

> On Tue, 07 Apr 2015 15:41:32 -0700 Kevin Hilman <khilman@kernel.org> wrote:
>
>> Andrew Morton <akpm@linux-foundation.org> writes:
>> 
>> > On Tue, 7 Apr 2015 10:57:52 -0700 Kevin Hilman <khilman@kernel.org> wrote:
>> >
>> >> > The diff below[2] on top of yours compiles fine here and at least covers
>> >> > the compilers I *know* to trigger the ICE.
>> >> 
>> >> I see my fix in your mmots since last Thurs (4/2), but it's not in
>> >> mmotm (last updated today) so today's linux-next still has the ICE for
>> >> anything other than gcc-4.7.3.   Just checking to see when you plan to
>> >> update mmotm.
>> >
>> > It should all be there today?
>> 
>> Nope.  
>
> huh, I swear I did an mmotm yesterday.

Well, based on the timestamp of the mmotm dir on ozlabs, it appears you
did.  That's why I was confused why the gcc-473 patches from mmots aren't
there.

> Let me see if I can sort out the watchdog mess and produce something
> releasable...

OK, thanks.

Kevin

Re: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in gcc 4.7.3

[Andrew Morton]

On Tue, 07 Apr 2015 16:27:44 -0700 Kevin Hilman <khilman@kernel.org> wrote:

> >> > It should all be there today?
> >> 
> >> Nope.  
> >
> > huh, I swear I did an mmotm yesterday.
> 
> Well, based on the timestamp of the mmotm dir on ozlabs, it appears you
> did.  That's why I was confused why the gcc-473 patches from mmots aren't
> there.

Things look a bit better now.

Re: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in gcc 4.7.3

[Kevin Hilman]

Andrew Morton <akpm@linux-foundation.org> writes:

> On Tue, 07 Apr 2015 16:27:44 -0700 Kevin Hilman <khilman@kernel.org> wrote:
>
>> >> > It should all be there today?
>> >> 
>> >> Nope.  
>> >
>> > huh, I swear I did an mmotm yesterday.
>> 
>> Well, based on the timestamp of the mmotm dir on ozlabs, it appears you
>> did.  That's why I was confused why the gcc-473 patches from mmots aren't
>> there.
>
> Things look a bit better now.

Yup, I can confirm all 4 patches are there now.  Things should be in
good shape for the next -next.

Thanks,

Kevin

Re: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in gcc 4.7.3

[Kevin Hilman]

Geert Uytterhoeven <geert@linux-m68k.org> writes:

[...]

>> build bisect points to commit 21f992084aeb[3], but that doesn't revert
>> cleanly so I haven't got any further than that yet.
>
> I installed gcc-arm-linux-gnueabi (4:4.7.2-1 from Ubuntu 14.04 LTS) and could
> reproduce the ICE. I came up with the workaround below.

Awesome, thanks!

> Does this work for you?

Yes, that patch works well and fixes the regression. Build results for
all the defconfigs here:

   http://kernelci.org/build/khilman/kernel/v4.0-rc6-8294-g2ef3958cc27e/

and the remaining issues arent' realted to this ICE.

> From 7ebe83316eaf1952e55a76754ce7a5832e461b8c Mon Sep 17 00:00:00 2001
> From: Geert Uytterhoeven <geert+renesas@glider.be>
> Date: Wed, 1 Apr 2015 11:22:51 +0200
> Subject: [PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in
>  gcc 4.7.3
> MIME-Version: 1.0
> Content-Type: text/plain; charset=UTF-8
> Content-Transfer-Encoding: 8bit
>
> With gcc version 4.7.3 (Ubuntu/Linaro 4.7.3-12ubuntu1) :
>
>     mm/migrate.c: In function ‘migrate_pages’:
>     mm/migrate.c:1148:1: internal compiler error: in push_minipool_fix, at config/arm/arm.c:13500
>     Please submit a full bug report,
>     with preprocessed source if appropriate.
>     See <file:///usr/share/doc/gcc-4.7/README.Bugs> for instructions.
>     Preprocessed source stored into /tmp/ccPoM1tr.out file, please attach this to your bugreport.
>     make[1]: *** [mm/migrate.o] Error 1
>     make: *** [mm/migrate.o] Error 2
>
> Mark unmap_and_move() (which is used in a single place only) "noinline"
> to work around this compiler bug.
>
> Reported-by: Kevin Hilman <khilman@kernel.org>
> Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>

Tested-by: Kevin Hilman <khilman@linaro.org>

> ---
>  mm/migrate.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
>
> diff --git a/mm/migrate.c b/mm/migrate.c
> index 114602a68111d809..98f8574456c2010c 100644
> --- a/mm/migrate.c
> +++ b/mm/migrate.c
> @@ -904,9 +904,10 @@ out:
>   * Obtain the lock on page, remove all ptes and migrate the page
>   * to the newly allocated page in newpage.
>   */
> -static int unmap_and_move(new_page_t get_new_page, free_page_t put_new_page,
> -			unsigned long private, struct page *page, int force,
> -			enum migrate_mode mode)
> +static noinline int unmap_and_move(new_page_t get_new_page,
> +				   free_page_t put_new_page,
> +				   unsigned long private, struct page *page,
> +				   int force, enum migrate_mode mode)
>  {
>  	int rc = 0;
>  	int *result = NULL;