Patchwork [01/10] Makefile: Add option CONFIG_DISABLE_GCC_AUTOMATIC_INLINING

login
register
mail settings
Submitter Ezequiel Garcia
Date Sept. 8, 2012, 8:47 p.m.
Message ID <1347137279-17568-1-git-send-email-elezegarcia@gmail.com>
Download mbox | patch
Permalink /patch/1427211/
State New, archived
Headers show

Comments

Ezequiel Garcia - Sept. 8, 2012, 8:47 p.m.
As its name suggest this option prevents gcc from auto inlining
small functions. This is very important if one wants to obtain
traces with accurate call sites.

Without this option, gcc will collapse some small functions,
even when not marked as 'inline' thus making impossible to
correlate the trace caller address to the real function it belongs.

Of course, the resultant kernel is slower and slightly smaller,
but that's not an issue if the focus is on tracing accuracy.

Cc: Michal Marek <mmarek@suse.cz>
Signed-off-by: Ezequiel Garcia <elezegarcia@gmail.com>
---
 Makefile          |    4 ++++
 lib/Kconfig.debug |   11 +++++++++++
 2 files changed, 15 insertions(+), 0 deletions(-)
Sam Ravnborg - Sept. 8, 2012, 9:43 p.m.
On Sat, Sep 08, 2012 at 05:47:50PM -0300, Ezequiel Garcia wrote:
> As its name suggest this option prevents gcc from auto inlining
> small functions. This is very important if one wants to obtain
> traces with accurate call sites.
> 
> Without this option, gcc will collapse some small functions,
> even when not marked as 'inline' thus making impossible to
> correlate the trace caller address to the real function it belongs.
> 
> Of course, the resultant kernel is slower and slightly smaller,
> but that's not an issue if the focus is on tracing accuracy.
> 
> Cc: Michal Marek <mmarek@suse.cz>
> Signed-off-by: Ezequiel Garcia <elezegarcia@gmail.com>
> ---
>  Makefile          |    4 ++++
>  lib/Kconfig.debug |   11 +++++++++++
>  2 files changed, 15 insertions(+), 0 deletions(-)
> 
> diff --git a/Makefile b/Makefile
> index ddf5be9..df6045a 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -561,6 +561,10 @@ else
>  KBUILD_CFLAGS	+= -O2
>  endif
>  
> +ifdef CONFIG_DISABLE_GCC_AUTOMATIC_INLINING
> +KBUILD_CFLAGS	+= -fno-inline-small-functions
> +endif
> +
>  include $(srctree)/arch/$(SRCARCH)/Makefile
>  
>  ifdef CONFIG_READABLE_ASM
> diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
> index 2403a63..c8fd50f 100644
> --- a/lib/Kconfig.debug
> +++ b/lib/Kconfig.debug
> @@ -1265,6 +1265,17 @@ config LATENCYTOP
>  source mm/Kconfig.debug
>  source kernel/trace/Kconfig
>  
> +config DISABLE_GCC_AUTOMATIC_INLINING
> +	bool "Disable gcc automatic inlining"

Could we please call this option for:
config CC_DISABLE_AUTOMATIC_INLINING

We have at least a few other options following that naming scheme.
And today we have no options named *GCC_*

	Sam
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
David Rientjes - Sept. 9, 2012, 9:25 p.m.
On Sat, 8 Sep 2012, Ezequiel Garcia wrote:

> diff --git a/Makefile b/Makefile
> index ddf5be9..df6045a 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -561,6 +561,10 @@ else
>  KBUILD_CFLAGS	+= -O2
>  endif
>  
> +ifdef CONFIG_DISABLE_GCC_AUTOMATIC_INLINING
> +KBUILD_CFLAGS	+= -fno-inline-small-functions

This isn't the only option that controls automatic inlining of functions, 
see indirect-inlining, inline-functions, and inline-functions-called-once.

> +endif
> +
>  include $(srctree)/arch/$(SRCARCH)/Makefile
>  
>  ifdef CONFIG_READABLE_ASM
> diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
> index 2403a63..c8fd50f 100644
> --- a/lib/Kconfig.debug
> +++ b/lib/Kconfig.debug
> @@ -1265,6 +1265,17 @@ config LATENCYTOP
>  source mm/Kconfig.debug
>  source kernel/trace/Kconfig
>  
> +config DISABLE_GCC_AUTOMATIC_INLINING
> +	bool "Disable gcc automatic inlining"
> +	depends on TRACING
> +	help
> +	  This option tells gcc he's not allowed to inline functions automatically,
> +	  when they are not marked as 'inline'.
> +	  In turn, this enables to trace an event with an accurate call site.
> +	  Of course, the resultant kernel is slower and slightly smaller.
> +
> +	  Select this option only if you want to trace call sites accurately.
> +
>  config PROVIDE_OHCI1394_DMA_INIT
>  	bool "Remote debugging over FireWire early on boot"
>  	depends on PCI && X86
> -- 
> 1.7.8.6
> 
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo@kvack.org.  For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
Ezequiel Garcia - Sept. 13, 2012, 12:30 a.m.
Hi,

On Sun, Sep 9, 2012 at 6:25 PM, David Rientjes <rientjes@google.com> wrote:
> On Sat, 8 Sep 2012, Ezequiel Garcia wrote:
>
>> diff --git a/Makefile b/Makefile
>> index ddf5be9..df6045a 100644
>> --- a/Makefile
>> +++ b/Makefile
>> @@ -561,6 +561,10 @@ else
>>  KBUILD_CFLAGS        += -O2
>>  endif
>>
>> +ifdef CONFIG_DISABLE_GCC_AUTOMATIC_INLINING
>> +KBUILD_CFLAGS        += -fno-inline-small-functions
>
> This isn't the only option that controls automatic inlining of functions,
> see indirect-inlining, inline-functions, and inline-functions-called-once.
>

I'll check about this gcc options and re-send, renamed as:
CONFIG_DISABLE_CC_AUTOMATIC_INLINING

Thanks for both your feedback,
Ezequiel.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
Michal Marek - Sept. 13, 2012, 7:17 a.m.
Dne 13.9.2012 02:30, Ezequiel Garcia napsal(a):
> Hi,
> 
> On Sun, Sep 9, 2012 at 6:25 PM, David Rientjes <rientjes@google.com> wrote:
>> On Sat, 8 Sep 2012, Ezequiel Garcia wrote:
>>
>>> diff --git a/Makefile b/Makefile
>>> index ddf5be9..df6045a 100644
>>> --- a/Makefile
>>> +++ b/Makefile
>>> @@ -561,6 +561,10 @@ else
>>>  KBUILD_CFLAGS        += -O2
>>>  endif
>>>
>>> +ifdef CONFIG_DISABLE_GCC_AUTOMATIC_INLINING
>>> +KBUILD_CFLAGS        += -fno-inline-small-functions
>>
>> This isn't the only option that controls automatic inlining of functions,
>> see indirect-inlining, inline-functions, and inline-functions-called-once.
>>
> 
> I'll check about this gcc options and re-send, renamed as:
> CONFIG_DISABLE_CC_AUTOMATIC_INLINING

Please name it CONFIG_CC_<something>

Michal

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
Ezequiel Garcia - Sept. 13, 2012, 9:16 a.m.
On Thu, Sep 13, 2012 at 4:17 AM, Michal Marek <mmarek@suse.cz> wrote:
> Dne 13.9.2012 02:30, Ezequiel Garcia napsal(a):
>> Hi,
>>
>> On Sun, Sep 9, 2012 at 6:25 PM, David Rientjes <rientjes@google.com> wrote:
>>> On Sat, 8 Sep 2012, Ezequiel Garcia wrote:
>>>
>>>> diff --git a/Makefile b/Makefile
>>>> index ddf5be9..df6045a 100644
>>>> --- a/Makefile
>>>> +++ b/Makefile
>>>> @@ -561,6 +561,10 @@ else
>>>>  KBUILD_CFLAGS        += -O2
>>>>  endif
>>>>
>>>> +ifdef CONFIG_DISABLE_GCC_AUTOMATIC_INLINING
>>>> +KBUILD_CFLAGS        += -fno-inline-small-functions
>>>
>>> This isn't the only option that controls automatic inlining of functions,
>>> see indirect-inlining, inline-functions, and inline-functions-called-once.
>>>
>>
>> I'll check about this gcc options and re-send, renamed as:
>> CONFIG_DISABLE_CC_AUTOMATIC_INLINING
>
> Please name it CONFIG_CC_<something>
>

Yeah, that's what I meant, sorry.

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

Patch

diff --git a/Makefile b/Makefile
index ddf5be9..df6045a 100644
--- a/Makefile
+++ b/Makefile
@@ -561,6 +561,10 @@  else
 KBUILD_CFLAGS	+= -O2
 endif
 
+ifdef CONFIG_DISABLE_GCC_AUTOMATIC_INLINING
+KBUILD_CFLAGS	+= -fno-inline-small-functions
+endif
+
 include $(srctree)/arch/$(SRCARCH)/Makefile
 
 ifdef CONFIG_READABLE_ASM
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 2403a63..c8fd50f 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1265,6 +1265,17 @@  config LATENCYTOP
 source mm/Kconfig.debug
 source kernel/trace/Kconfig
 
+config DISABLE_GCC_AUTOMATIC_INLINING
+	bool "Disable gcc automatic inlining"
+	depends on TRACING
+	help
+	  This option tells gcc he's not allowed to inline functions automatically,
+	  when they are not marked as 'inline'.
+	  In turn, this enables to trace an event with an accurate call site.
+	  Of course, the resultant kernel is slower and slightly smaller.
+
+	  Select this option only if you want to trace call sites accurately.
+
 config PROVIDE_OHCI1394_DMA_INIT
 	bool "Remote debugging over FireWire early on boot"
 	depends on PCI && X86