diff mbox

kbuild: disable Android-specific compiler features

Message ID 20160206004438.GA21780@www.outflux.net (mailing list archive)
State New, archived
Headers show

Commit Message

Kees Cook Feb. 6, 2016, 12:44 a.m. UTC
The Android compilers enable some non-standard features by default. While
most Android build systems inject the needed "-mno-android" option via
KCFLAGS, it happens too late (at least on x86_64), since KBUILD_CFLAGS
gains KCFLAGS after running (and failing) many cc-option tests. (For
example, the stack-protector tests happen after arch-specific
KBUILD_CFLAGS are added but before the external KCFLAGS are added.) As
such, we should notice this option and immediately turn it on as the
first cc-option test we run.

Signed-off-by: Kees Cook <keescook@chromium.org>
---
 Makefile | 4 ++++
 1 file changed, 4 insertions(+)

Comments

Kees Cook Feb. 17, 2016, 11:03 p.m. UTC | #1
On Fri, Feb 5, 2016 at 4:44 PM, Kees Cook <keescook@chromium.org> wrote:
> The Android compilers enable some non-standard features by default. While
> most Android build systems inject the needed "-mno-android" option via
> KCFLAGS, it happens too late (at least on x86_64), since KBUILD_CFLAGS
> gains KCFLAGS after running (and failing) many cc-option tests. (For
> example, the stack-protector tests happen after arch-specific
> KBUILD_CFLAGS are added but before the external KCFLAGS are added.) As
> such, we should notice this option and immediately turn it on as the
> first cc-option test we run.
>
> Signed-off-by: Kees Cook <keescook@chromium.org>

Michal, is this okay? I'd like to get it into -next soon if possible.

Thanks!

-Kees

> ---
>  Makefile | 4 ++++
>  1 file changed, 4 insertions(+)
>
> diff --git a/Makefile b/Makefile
> index 6c1a3c247988..126c98b582bb 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -393,6 +393,10 @@ KBUILD_CFLAGS   := -Wall -Wundef -Wstrict-prototypes -Wno-trigraphs \
>                    -Wno-format-security \
>                    -std=gnu89
>
> +# We must turn off the Android-specific compiler options as early as possible
> +# otherwise cc-option calls below may erroneously fail.
> +KBUILD_CFLAGS  += $(call cc-option,-mno-android,)
> +
>  KBUILD_AFLAGS_KERNEL :=
>  KBUILD_CFLAGS_KERNEL :=
>  KBUILD_AFLAGS   := -D__ASSEMBLY__
> --
> 2.6.3
>
>
> --
> Kees Cook
> Chrome OS & Brillo Security
Michal Marek Feb. 18, 2016, 9:36 a.m. UTC | #2
Dne 18.2.2016 v 00:03 Kees Cook napsal(a):
> On Fri, Feb 5, 2016 at 4:44 PM, Kees Cook <keescook@chromium.org> wrote:
>> The Android compilers enable some non-standard features by default. While
>> most Android build systems inject the needed "-mno-android" option via
>> KCFLAGS, it happens too late (at least on x86_64), since KBUILD_CFLAGS
>> gains KCFLAGS after running (and failing) many cc-option tests. (For
>> example, the stack-protector tests happen after arch-specific
>> KBUILD_CFLAGS are added but before the external KCFLAGS are added.) As
>> such, we should notice this option and immediately turn it on as the
>> first cc-option test we run.
>>
>> Signed-off-by: Kees Cook <keescook@chromium.org>
> 
> Michal, is this okay? I'd like to get it into -next soon if possible.

Is there a way to detect these compilers and are there some canonical
versions of these? I do not think it's a good idea to add workaround for
each random gcc fork.

Michal
--
To unsubscribe from this list: send the line "unsubscribe linux-kbuild" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Kees Cook Feb. 18, 2016, 5:54 p.m. UTC | #3
On Thu, Feb 18, 2016 at 1:36 AM, Michal Marek <mmarek@suse.com> wrote:
> Dne 18.2.2016 v 00:03 Kees Cook napsal(a):
>> On Fri, Feb 5, 2016 at 4:44 PM, Kees Cook <keescook@chromium.org> wrote:
>>> The Android compilers enable some non-standard features by default. While
>>> most Android build systems inject the needed "-mno-android" option via
>>> KCFLAGS, it happens too late (at least on x86_64), since KBUILD_CFLAGS
>>> gains KCFLAGS after running (and failing) many cc-option tests. (For
>>> example, the stack-protector tests happen after arch-specific
>>> KBUILD_CFLAGS are added but before the external KCFLAGS are added.) As
>>> such, we should notice this option and immediately turn it on as the
>>> first cc-option test we run.
>>>
>>> Signed-off-by: Kees Cook <keescook@chromium.org>
>>
>> Michal, is this okay? I'd like to get it into -next soon if possible.
>
> Is there a way to detect these compilers and are there some canonical
> versions of these? I do not think it's a good idea to add workaround for
> each random gcc fork.

It's detectable by -mno-android being a valid option. It's not a
random gcc fork, as it's part of the Android SDK gcc prebuilts that
all Android vendors and hobbiests use to build their kernels.

-Kees
diff mbox

Patch

diff --git a/Makefile b/Makefile
index 6c1a3c247988..126c98b582bb 100644
--- a/Makefile
+++ b/Makefile
@@ -393,6 +393,10 @@  KBUILD_CFLAGS   := -Wall -Wundef -Wstrict-prototypes -Wno-trigraphs \
 		   -Wno-format-security \
 		   -std=gnu89
 
+# We must turn off the Android-specific compiler options as early as possible
+# otherwise cc-option calls below may erroneously fail.
+KBUILD_CFLAGS	+= $(call cc-option,-mno-android,)
+
 KBUILD_AFLAGS_KERNEL :=
 KBUILD_CFLAGS_KERNEL :=
 KBUILD_AFLAGS   := -D__ASSEMBLY__