drm/amdgpu: fix stack alignment ABI mismatch for Clang
authorNick Desaulniers <ndesaulniers@google.com>
Wed, 16 Oct 2019 23:02:07 +0000 (16:02 -0700)
committerAlex Deucher <alexander.deucher@amd.com>
Wed, 30 Oct 2019 15:56:20 +0000 (11:56 -0400)
The x86 kernel is compiled with an 8B stack alignment via
`-mpreferred-stack-boundary=3` for GCC since 3.6-rc1 via
commit d9b0cde91c60 ("x86-64, gcc: Use -mpreferred-stack-boundary=3 if supported")
or `-mstack-alignment=8` for Clang. Parts of the AMDGPU driver are
compiled with 16B stack alignment.

Generally, the stack alignment is part of the ABI. Linking together two
different translation units with differing stack alignment is dangerous,
particularly when the translation unit with the smaller stack alignment
makes calls into the translation unit with the larger stack alignment.
While 8B aligned stacks are sometimes also 16B aligned, they are not
always.

Multiple users have reported General Protection Faults (GPF) when using
the AMDGPU driver compiled with Clang. Clang is placing objects in stack
slots assuming the stack is 16B aligned, and selecting instructions that
require 16B aligned memory operands.

At runtime, syscall handlers with 8B aligned stack call into code that
assumes 16B stack alignment.  When the stack is a multiple of 8B but not
16B, these instructions result in a GPF.

Remove the code that added compatibility between the differing compiler
flags, as it will result in runtime GPFs when built with Clang. Cleanups
for GCC will be sent in later patches in the series.

Link: https://github.com/ClangBuiltLinux/linux/issues/735
Debugged-by: Yuxuan Shui <yshuiv7@gmail.com>
Reported-by: Shirish S <shirish.s@amd.com>
Reported-by: Yuxuan Shui <yshuiv7@gmail.com>
Suggested-by: Andrew Cooper <andrew.cooper3@citrix.com>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drivers/gpu/drm/amd/display/dc/calcs/Makefile
drivers/gpu/drm/amd/display/dc/dcn20/Makefile
drivers/gpu/drm/amd/display/dc/dcn21/Makefile
drivers/gpu/drm/amd/display/dc/dml/Makefile
drivers/gpu/drm/amd/display/dc/dsc/Makefile

index 985633c..4b1a8a0 100644 (file)
 # It calculates Bandwidth and Watermarks values for HW programming
 #
 
-ifneq ($(call cc-option, -mpreferred-stack-boundary=4),)
-       cc_stack_align := -mpreferred-stack-boundary=4
-else ifneq ($(call cc-option, -mstack-alignment=16),)
-       cc_stack_align := -mstack-alignment=16
-endif
+calcs_ccflags := -mhard-float -msse
 
-calcs_ccflags := -mhard-float -msse $(cc_stack_align)
+ifdef CONFIG_CC_IS_GCC
+calcs_ccflags += -mpreferred-stack-boundary=4
+endif
 
 ifdef CONFIG_CC_IS_CLANG
 calcs_ccflags += -msse2
index ddb8d56..5fe3eb8 100644 (file)
@@ -10,13 +10,11 @@ ifdef CONFIG_DRM_AMD_DC_DSC_SUPPORT
 DCN20 += dcn20_dsc.o
 endif
 
-ifneq ($(call cc-option, -mpreferred-stack-boundary=4),)
-       cc_stack_align := -mpreferred-stack-boundary=4
-else ifneq ($(call cc-option, -mstack-alignment=16),)
-       cc_stack_align := -mstack-alignment=16
-endif
+CFLAGS_$(AMDDALPATH)/dc/dcn20/dcn20_resource.o := -mhard-float -msse
 
-CFLAGS_$(AMDDALPATH)/dc/dcn20/dcn20_resource.o := -mhard-float -msse $(cc_stack_align)
+ifdef CONFIG_CC_IS_GCC
+CFLAGS_$(AMDDALPATH)/dc/dcn20/dcn20_resource.o += -mpreferred-stack-boundary=4
+endif
 
 ifdef CONFIG_CC_IS_CLANG
 CFLAGS_$(AMDDALPATH)/dc/dcn20/dcn20_resource.o += -msse2
index ef673bf..7057e20 100644 (file)
@@ -3,13 +3,11 @@
 
 DCN21 = dcn21_hubp.o dcn21_hubbub.o dcn21_resource.o
 
-ifneq ($(call cc-option, -mpreferred-stack-boundary=4),)
-       cc_stack_align := -mpreferred-stack-boundary=4
-else ifneq ($(call cc-option, -mstack-alignment=16),)
-       cc_stack_align := -mstack-alignment=16
-endif
+CFLAGS_$(AMDDALPATH)/dc/dcn21/dcn21_resource.o := -mhard-float -msse
 
-CFLAGS_$(AMDDALPATH)/dc/dcn21/dcn21_resource.o := -mhard-float -msse $(cc_stack_align)
+ifdef CONFIG_CC_IS_GCC
+CFLAGS_$(AMDDALPATH)/dc/dcn21/dcn21_resource.o += -mpreferred-stack-boundary=4
+endif
 
 ifdef CONFIG_CC_IS_CLANG
 CFLAGS_$(AMDDALPATH)/dc/dcn21/dcn21_resource.o += -msse2
index 5b2a65b..1bd6e30 100644 (file)
 # It provides the general basic services required by other DAL
 # subcomponents.
 
-ifneq ($(call cc-option, -mpreferred-stack-boundary=4),)
-       cc_stack_align := -mpreferred-stack-boundary=4
-else ifneq ($(call cc-option, -mstack-alignment=16),)
-       cc_stack_align := -mstack-alignment=16
-endif
+dml_ccflags := -mhard-float -msse
 
-dml_ccflags := -mhard-float -msse $(cc_stack_align)
+ifdef CONFIG_CC_IS_GCC
+dml_ccflags += -mpreferred-stack-boundary=4
+endif
 
 ifdef CONFIG_CC_IS_CLANG
 dml_ccflags += -msse2
index b456cd2..932c305 100644 (file)
@@ -1,13 +1,11 @@
 #
 # Makefile for the 'dsc' sub-component of DAL.
 
-ifneq ($(call cc-option, -mpreferred-stack-boundary=4),)
-       cc_stack_align := -mpreferred-stack-boundary=4
-else ifneq ($(call cc-option, -mstack-alignment=16),)
-       cc_stack_align := -mstack-alignment=16
-endif
+dsc_ccflags := -mhard-float -msse
 
-dsc_ccflags := -mhard-float -msse $(cc_stack_align)
+ifdef CONFIG_CC_IS_GCC
+dsc_ccflags += -mpreferred-stack-boundary=4
+endif
 
 ifdef CONFIG_CC_IS_CLANG
 dsc_ccflags += -msse2