| Message ID | 1507064302-3972-1-git-send-email-linux@roeck-us.net (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi, On Tue, Oct 3, 2017 at 1:58 PM, Guenter Roeck <linux@roeck-us.net> wrote: > If LD or AS is specified on the command line using "make LD=<my-ld>", > the endianness flag will not be added. Depending on the toolchain used, > this can result in endianness mismatch errors, such as > > aarch64-cros-linux-gnu/binutils-bin/2.27.0/ld.bfd.real: > arch/arm64/crypto/.tmp_aes-ce-cipher.o: > compiled for a big endian system and target is little endian > > The problem was observed when using the Chrome OS build system (which sets > LD on the command line) for an arm64:allmodconfig build. > > Use the override flag for both LD and AS to ensure that the endianness > flag is always set. > > Cc: Douglas Anderson <dianders@chromium.org> > Signed-off-by: Guenter Roeck <linux@roeck-us.net> > --- > arch/arm64/Makefile | 8 ++++---- > 1 file changed, 4 insertions(+), 4 deletions(-) This seems right to me. My understanding of the way things work is that if you want a different linker or assembler you pass it on the command line. That means you need "override" like this if you want to append to it. As a self-proclaimed non-expert on the kernel build system I'm not sure it's worth much, but in any case: Reviewed-by: Douglas Anderson <dianders@chromium.org>
On Tue, Oct 03, 2017 at 01:58:22PM -0700, Guenter Roeck wrote: > If LD or AS is specified on the command line using "make LD=<my-ld>", > the endianness flag will not be added. Depending on the toolchain used, > this can result in endianness mismatch errors, such as > > aarch64-cros-linux-gnu/binutils-bin/2.27.0/ld.bfd.real: > arch/arm64/crypto/.tmp_aes-ce-cipher.o: > compiled for a big endian system and target is little endian > > The problem was observed when using the Chrome OS build system (which sets > LD on the command line) for an arm64:allmodconfig build. > > Use the override flag for both LD and AS to ensure that the endianness > flag is always set. > > Cc: Douglas Anderson <dianders@chromium.org> > Signed-off-by: Guenter Roeck <linux@roeck-us.net> Please ignore this patch for now; it doesn't help with my problem. Not yet sure what happens, but it appears that "-EB" doesn't make it into the LD variable used by cmd_make_builtin in scripts/Makefile.build. I'll have to figure out what is going on. Thanks, Guenter > --- > arch/arm64/Makefile | 8 ++++---- > 1 file changed, 4 insertions(+), 4 deletions(-) > > diff --git a/arch/arm64/Makefile b/arch/arm64/Makefile > index 939b310913cf..d780d00bea9c 100644 > --- a/arch/arm64/Makefile > +++ b/arch/arm64/Makefile > @@ -56,15 +56,15 @@ KBUILD_AFLAGS += $(call cc-option,-mabi=lp64) > ifeq ($(CONFIG_CPU_BIG_ENDIAN), y) > KBUILD_CPPFLAGS += -mbig-endian > CHECKFLAGS += -D__AARCH64EB__ > -AS += -EB > -LD += -EB > +override AS += -EB > +override LD += -EB > LDFLAGS += -maarch64linuxb > UTS_MACHINE := aarch64_be > else > KBUILD_CPPFLAGS += -mlittle-endian > CHECKFLAGS += -D__AARCH64EL__ > -AS += -EL > -LD += -EL > +override AS += -EL > +override LD += -EL > LDFLAGS += -maarch64linux > UTS_MACHINE := aarch64 > endif > -- > 2.7.4 >
On Wed, Oct 04, 2017 at 11:31:08AM -0700, Guenter Roeck wrote: > On Tue, Oct 03, 2017 at 01:58:22PM -0700, Guenter Roeck wrote: > > If LD or AS is specified on the command line using "make LD=<my-ld>", > > the endianness flag will not be added. Depending on the toolchain used, > > this can result in endianness mismatch errors, such as > > > > aarch64-cros-linux-gnu/binutils-bin/2.27.0/ld.bfd.real: > > arch/arm64/crypto/.tmp_aes-ce-cipher.o: > > compiled for a big endian system and target is little endian > > > > The problem was observed when using the Chrome OS build system (which sets > > LD on the command line) for an arm64:allmodconfig build. > > > > Use the override flag for both LD and AS to ensure that the endianness > > flag is always set. > > > > Cc: Douglas Anderson <dianders@chromium.org> > > Signed-off-by: Guenter Roeck <linux@roeck-us.net> > > Please ignore this patch for now; it doesn't help with my problem. Not yet > sure what happens, but it appears that "-EB" doesn't make it into the LD > variable used by cmd_make_builtin in scripts/Makefile.build. I'll have > to figure out what is going on. > Yes, the problem is that the gcc in the Chrome OS toolchain doesn't tell the linker which endianness to use (ie it doesn't translate -mbig-endian into a linker flag). Please forget this patch, it doesn't solve anything. Sorry for the noise. Guenter
diff --git a/arch/arm64/Makefile b/arch/arm64/Makefile index 939b310913cf..d780d00bea9c 100644 --- a/arch/arm64/Makefile +++ b/arch/arm64/Makefile @@ -56,15 +56,15 @@ KBUILD_AFLAGS += $(call cc-option,-mabi=lp64) ifeq ($(CONFIG_CPU_BIG_ENDIAN), y) KBUILD_CPPFLAGS += -mbig-endian CHECKFLAGS += -D__AARCH64EB__ -AS += -EB -LD += -EB +override AS += -EB +override LD += -EB LDFLAGS += -maarch64linuxb UTS_MACHINE := aarch64_be else KBUILD_CPPFLAGS += -mlittle-endian CHECKFLAGS += -D__AARCH64EL__ -AS += -EL -LD += -EL +override AS += -EL +override LD += -EL LDFLAGS += -maarch64linux UTS_MACHINE := aarch64 endif
If LD or AS is specified on the command line using "make LD=<my-ld>", the endianness flag will not be added. Depending on the toolchain used, this can result in endianness mismatch errors, such as aarch64-cros-linux-gnu/binutils-bin/2.27.0/ld.bfd.real: arch/arm64/crypto/.tmp_aes-ce-cipher.o: compiled for a big endian system and target is little endian The problem was observed when using the Chrome OS build system (which sets LD on the command line) for an arm64:allmodconfig build. Use the override flag for both LD and AS to ensure that the endianness flag is always set. Cc: Douglas Anderson <dianders@chromium.org> Signed-off-by: Guenter Roeck <linux@roeck-us.net> --- arch/arm64/Makefile | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)