[PATCH] x86/build: Fix stack alignment for clang

From: Matthias Kaehlcke
Date: Tue Jul 11 2017 - 17:30:18 EST


Commit d77698df39a5 ("x86/build: Specify stack alignment for clang")
intended to use the same stack alignment for clang as with gcc. The
two compilers use different options to configure the stack alignment
(gcc: -mpreferred-stack-boundary=n, clang: -mstack-alignment=n).
The above commit assumes that the clang option uses the same parameter
type as gcc, i.e. that the alignment is specified as 2^n. However clang
interprets the value of this option literally to use an alignment of n,
in consequence the stack remains misaligned.

Change the values used with -mstack-alignment to be the actual alignment
instead of a power of two.

Signed-off-by: Matthias Kaehlcke <mka@xxxxxxxxxxxx>
---
Apparently I had really bad luck when testing d77698df39a5. It fixed the
crash I was investigating and the stack in the debugged function looked
good with the patch, so I didn't had any doubt that -mstack-alignment=3
(for x86-64) was the correct solution :(

If anyone has suggestions for macro magic to encapsulate the different
option name and paramater type I'm happy to adapt the patch. I
experimented a bit but didn't find a good solution.

arch/x86/Makefile | 15 ++++-----------
1 file changed, 4 insertions(+), 11 deletions(-)

diff --git a/arch/x86/Makefile b/arch/x86/Makefile
index 1e902f926be3..9c6f11ebf288 100644
--- a/arch/x86/Makefile
+++ b/arch/x86/Makefile
@@ -11,14 +11,6 @@ else
KBUILD_DEFCONFIG := $(ARCH)_defconfig
endif

-# For gcc stack alignment is specified with -mpreferred-stack-boundary,
-# clang has the option -mstack-alignment for that purpose.
-ifneq ($(call cc-option, -mpreferred-stack-boundary=4),)
- cc_stack_align_opt := -mpreferred-stack-boundary
-else ifneq ($(call cc-option, -mstack-alignment=4),)
- cc_stack_align_opt := -mstack-alignment
-endif
-
# How to compile the 16-bit code. Note we always compile for -march=i386;
# that way we can complain to the user if the CPU is insufficient.
#
@@ -36,7 +28,8 @@ REALMODE_CFLAGS := $(M16_CFLAGS) -g -Os -D__KERNEL__ \

REALMODE_CFLAGS += $(call __cc-option, $(CC), $(REALMODE_CFLAGS), -ffreestanding)
REALMODE_CFLAGS += $(call __cc-option, $(CC), $(REALMODE_CFLAGS), -fno-stack-protector)
-REALMODE_CFLAGS += $(call __cc-option, $(CC), $(REALMODE_CFLAGS), $(cc_stack_align_opt)=2)
+REALMODE_CFLAGS += $(call __cc-option, $(CC), $(REALMODE_CFLAGS),\
+ -mpreferred-stack-boundary=2,-mstack-alignment=4)
export REALMODE_CFLAGS

# BITS is used as extension for files which are available in a 32 bit
@@ -76,7 +69,7 @@ ifeq ($(CONFIG_X86_32),y)
# Align the stack to the register width instead of using the default
# alignment of 16 bytes. This reduces stack usage and the number of
# alignment instructions.
- KBUILD_CFLAGS += $(call cc-option,$(cc_stack_align_opt)=2)
+ KBUILD_CFLAGS += $(call cc-option,-mpreferred-stack-boundary=2,-mstack-alignment=4)

# Disable unit-at-a-time mode on pre-gcc-4.0 compilers, it makes gcc use
# a lot more stack due to the lack of sharing of stacklots:
@@ -115,7 +108,7 @@ else
# default alignment which keep the stack *mis*aligned.
# Furthermore an alignment to the register width reduces stack usage
# and the number of alignment instructions.
- KBUILD_CFLAGS += $(call cc-option,$(cc_stack_align_opt)=3)
+ KBUILD_CFLAGS += $(call cc-option,-mpreferred-stack-boundary=3,-mstack-alignment=8)

# Use -mskip-rax-setup if supported.
KBUILD_CFLAGS += $(call cc-option,-mskip-rax-setup)
--
2.13.2.725.g09c95d1e9-goog