x86-64 SIMD: Support CET if C compiler enables it

- Detect at configure time, via the __CET__ C preprocessor macro, whether the C compiler will include either indirect branch tracking (IBT) or shadow stack support, and define a NASM macro (__CET__) if so. - Modify the x86-64 SIMD code so that it includes appropriate endbr64 instructions (to support IBT) and an appropriate .note.gnu.property section (to support both IBT and shadow stack) when __CET__ is defined. Closes #350
libjpeg-turbo · Feb 29, 2024 · 3202feb · 3202feb
1 parent 1335547
commit 3202feb
Show file tree

Hide file tree

Showing 31 changed files with 81 additions and 1 deletion.
diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
@@ -149,7 +149,7 @@ jobs:
           mkdir build
           pushd build
           cmake -G"Unix Makefiles" -DWITH_JPEG8=1 \
-            -DCMAKE_C_FLAGS='--std=gnu90 -Wall -Werror -Wextra -Wpedantic -pedantic-errors -Wdouble-promotion -Wformat-overflow=2 -Wformat-security -Wformat-signedness -Wformat-truncation=2 -Wformat-y2k -Wmissing-include-dirs -Wshift-overflow=2 -Wswitch-bool -Wno-unused-parameter -Wuninitialized -Wstrict-overflow=2 -Wstringop-overflow=4 -Wstringop-truncation -Wduplicated-branches -Wduplicated-cond -Wdeclaration-after-statement -Wshadow -Wunsafe-loop-optimizations -Wundef -Wcast-align -Wno-clobbered -Wjump-misses-init -Wno-sign-compare -Wlogical-op -Waggregate-return -Wstrict-prototypes -Wold-style-definition -Wmissing-prototypes -Wmissing-declarations -Wpacked -Wredundant-decls -Wnested-externs -Winline -Wno-long-long -Wdisabled-optimization -Wno-overlength-strings' \
+            -DCMAKE_C_FLAGS='--std=gnu90 -Wall -Werror -Wextra -Wpedantic -pedantic-errors -Wdouble-promotion -Wformat-overflow=2 -Wformat-security -Wformat-signedness -Wformat-truncation=2 -Wformat-y2k -Wmissing-include-dirs -Wshift-overflow=2 -Wswitch-bool -Wno-unused-parameter -Wuninitialized -Wstrict-overflow=2 -Wstringop-overflow=4 -Wstringop-truncation -Wduplicated-branches -Wduplicated-cond -Wdeclaration-after-statement -Wshadow -Wunsafe-loop-optimizations -Wundef -Wcast-align -Wno-clobbered -Wjump-misses-init -Wno-sign-compare -Wlogical-op -Waggregate-return -Wstrict-prototypes -Wold-style-definition -Wmissing-prototypes -Wmissing-declarations -Wpacked -Wredundant-decls -Wnested-externs -Winline -Wno-long-long -Wdisabled-optimization -Wno-overlength-strings -fcf-protection' \
             ..
           export NUMCPUS=`grep -c '^processor' /proc/cpuinfo`
           make -j$NUMCPUS --load-average=$NUMCPUS

diff --git a/BUILDING.md b/BUILDING.md
@@ -14,6 +14,8 @@ Build Requirements
   (if building x86 or x86-64 SIMD extensions)
   * If using NASM, 2.13 or later is required.
   * If using Yasm, 1.2.0 or later is required.
+  * NASM 2.15 or later is required if building libjpeg-turbo with Intel
+    Control-flow Enforcement Technology (CET) support.
   * If building on macOS, NASM or Yasm can be obtained from
     [MacPorts](http://www.macports.org/) or [Homebrew](http://brew.sh/).
      - NOTE: Currently, if it is desirable to hide the SIMD function symbols in

diff --git a/ChangeLog.md b/ChangeLog.md
@@ -8,6 +8,10 @@ libjpeg-turbo components to depend on the Visual C++ run-time DLL when built
 with Visual C++ and CMake 3.15 or later, regardless of value of the
 `WITH_CRT_DLL` CMake variable.
 
+2. The x86-64 SIMD extensions now include support for Intel Control-flow
+Enforcement Technology (CET), which is enabled automatically if CET is enabled
+in the C compiler.
+
 
 3.0.2
 =====

diff --git a/simd/CMakeLists.txt b/simd/CMakeLists.txt
@@ -96,6 +96,18 @@ if(NOT WIN32 AND (CMAKE_POSITION_INDEPENDENT_CODE OR ENABLE_SHARED))
   set(CMAKE_ASM_NASM_FLAGS "${CMAKE_ASM_NASM_FLAGS} -DPIC")
 endif()
 
+if(CPU_TYPE STREQUAL "x86_64" AND CMAKE_ASM_NASM_OBJECT_FORMAT MATCHES "^elf")
+  check_c_source_compiles("
+    #if (__CET__ & 3) == 0
+    #error \"CET not enabled\"
+    #endif
+    int main(void) { return 0; }" HAVE_CET)
+
+  if(HAVE_CET)
+    set(CMAKE_ASM_NASM_FLAGS "${CMAKE_ASM_NASM_FLAGS} -D__CET__")
+  endif()
+endif()
+
 string(TOUPPER ${CMAKE_BUILD_TYPE} CMAKE_BUILD_TYPE_UC)
 set(EFFECTIVE_ASM_NASM_FLAGS "${CMAKE_ASM_NASM_FLAGS} ${CMAKE_ASM_NASM_FLAGS_${CMAKE_BUILD_TYPE_UC}}")
 message(STATUS "CMAKE_ASM_NASM_FLAGS = ${EFFECTIVE_ASM_NASM_FLAGS}")

diff --git a/simd/nasm/jsimdext.inc b/simd/nasm/jsimdext.inc
@@ -76,6 +76,14 @@
 ; mark stack as non-executable
 section .note.GNU-stack noalloc noexec nowrite progbits
 
+%ifdef __CET__
+%ifdef __x86_64__
+section .note.gnu.property note alloc noexec align=8
+    dd 0x00000004, 0x00000010, 0x00000005, 0x00554e47
+    dd 0xc0000002, 0x00000004, 0x00000003, 0x00000000
+%endif
+%endif
+
 ; -- segment definition --
 ;
 %ifdef __x86_64__
@@ -513,6 +521,19 @@ const_base:
 
 %endif
 
+%ifdef __CET__
+
+%imacro ENDBR64 0
+    dd 0xfa1e0ff3
+%endmacro
+
+%else
+
+%imacro ENDBR64 0
+%endmacro
+
+%endif
+
 ; --------------------------------------------------------------------------
 ;  Defines picked up from the C headers
 ;

diff --git a/simd/x86_64/jccolext-avx2.asm b/simd/x86_64/jccolext-avx2.asm
@@ -41,6 +41,7 @@
     GLOBAL_FUNCTION(jsimd_rgb_ycc_convert_avx2)
 
 EXTN(jsimd_rgb_ycc_convert_avx2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     push        r15

diff --git a/simd/x86_64/jccolext-sse2.asm b/simd/x86_64/jccolext-sse2.asm
@@ -40,6 +40,7 @@
     GLOBAL_FUNCTION(jsimd_rgb_ycc_convert_sse2)
 
 EXTN(jsimd_rgb_ycc_convert_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     push        r15

diff --git a/simd/x86_64/jcgryext-avx2.asm b/simd/x86_64/jcgryext-avx2.asm
@@ -41,6 +41,7 @@
     GLOBAL_FUNCTION(jsimd_rgb_gray_convert_avx2)
 
 EXTN(jsimd_rgb_gray_convert_avx2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     push        r15

diff --git a/simd/x86_64/jcgryext-sse2.asm b/simd/x86_64/jcgryext-sse2.asm
@@ -40,6 +40,7 @@
     GLOBAL_FUNCTION(jsimd_rgb_gray_convert_sse2)
 
 EXTN(jsimd_rgb_gray_convert_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     push        r15

diff --git a/simd/x86_64/jchuff-sse2.asm b/simd/x86_64/jchuff-sse2.asm
@@ -261,6 +261,7 @@ times 1 << 15 db 16
     GLOBAL_FUNCTION(jsimd_huff_encode_one_block_sse2)
 
 EXTN(jsimd_huff_encode_one_block_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
 

diff --git a/simd/x86_64/jcphuff-sse2.asm b/simd/x86_64/jcphuff-sse2.asm
@@ -283,6 +283,7 @@
     GLOBAL_FUNCTION(jsimd_encode_mcu_AC_first_prepare_sse2)
 
 EXTN(jsimd_encode_mcu_AC_first_prepare_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     and         rsp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
@@ -446,6 +447,7 @@ EXTN(jsimd_encode_mcu_AC_first_prepare_sse2):
     GLOBAL_FUNCTION(jsimd_encode_mcu_AC_refine_prepare_sse2)
 
 EXTN(jsimd_encode_mcu_AC_refine_prepare_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     and         rsp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits

diff --git a/simd/x86_64/jcsample-avx2.asm b/simd/x86_64/jcsample-avx2.asm
@@ -44,6 +44,7 @@
     GLOBAL_FUNCTION(jsimd_h2v1_downsample_avx2)
 
 EXTN(jsimd_h2v1_downsample_avx2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     COLLECT_ARGS 6
@@ -205,6 +206,7 @@ EXTN(jsimd_h2v1_downsample_avx2):
     GLOBAL_FUNCTION(jsimd_h2v2_downsample_avx2)
 
 EXTN(jsimd_h2v2_downsample_avx2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     COLLECT_ARGS 6

diff --git a/simd/x86_64/jcsample-sse2.asm b/simd/x86_64/jcsample-sse2.asm
@@ -43,6 +43,7 @@
     GLOBAL_FUNCTION(jsimd_h2v1_downsample_sse2)
 
 EXTN(jsimd_h2v1_downsample_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     COLLECT_ARGS 6
@@ -187,6 +188,7 @@ EXTN(jsimd_h2v1_downsample_sse2):
     GLOBAL_FUNCTION(jsimd_h2v2_downsample_sse2)
 
 EXTN(jsimd_h2v2_downsample_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     COLLECT_ARGS 6

diff --git a/simd/x86_64/jdcolext-avx2.asm b/simd/x86_64/jdcolext-avx2.asm
@@ -42,6 +42,7 @@
     GLOBAL_FUNCTION(jsimd_ycc_rgb_convert_avx2)
 
 EXTN(jsimd_ycc_rgb_convert_avx2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     push        r15

diff --git a/simd/x86_64/jdcolext-sse2.asm b/simd/x86_64/jdcolext-sse2.asm
@@ -41,6 +41,7 @@
     GLOBAL_FUNCTION(jsimd_ycc_rgb_convert_sse2)
 
 EXTN(jsimd_ycc_rgb_convert_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     push        r15

diff --git a/simd/x86_64/jdmrgext-avx2.asm b/simd/x86_64/jdmrgext-avx2.asm
@@ -42,6 +42,7 @@
     GLOBAL_FUNCTION(jsimd_h2v1_merged_upsample_avx2)
 
 EXTN(jsimd_h2v1_merged_upsample_avx2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     push        r15
@@ -506,6 +507,7 @@ EXTN(jsimd_h2v1_merged_upsample_avx2):
     GLOBAL_FUNCTION(jsimd_h2v2_merged_upsample_avx2)
 
 EXTN(jsimd_h2v2_merged_upsample_avx2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     COLLECT_ARGS 4

diff --git a/simd/x86_64/jdmrgext-sse2.asm b/simd/x86_64/jdmrgext-sse2.asm
@@ -41,6 +41,7 @@
     GLOBAL_FUNCTION(jsimd_h2v1_merged_upsample_sse2)
 
 EXTN(jsimd_h2v1_merged_upsample_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     push        r15
@@ -448,6 +449,7 @@ EXTN(jsimd_h2v1_merged_upsample_sse2):
     GLOBAL_FUNCTION(jsimd_h2v2_merged_upsample_sse2)
 
 EXTN(jsimd_h2v2_merged_upsample_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     COLLECT_ARGS 4

diff --git a/simd/x86_64/jdsample-avx2.asm b/simd/x86_64/jdsample-avx2.asm
@@ -62,6 +62,7 @@ PW_EIGHT times 16 dw 8
     GLOBAL_FUNCTION(jsimd_h2v1_fancy_upsample_avx2)
 
 EXTN(jsimd_h2v1_fancy_upsample_avx2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     PUSH_XMM    3
@@ -215,6 +216,7 @@ EXTN(jsimd_h2v1_fancy_upsample_avx2):
     GLOBAL_FUNCTION(jsimd_h2v2_fancy_upsample_avx2)
 
 EXTN(jsimd_h2v2_fancy_upsample_avx2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     push        r15
@@ -524,6 +526,7 @@ EXTN(jsimd_h2v2_fancy_upsample_avx2):
     GLOBAL_FUNCTION(jsimd_h2v1_upsample_avx2)
 
 EXTN(jsimd_h2v1_upsample_avx2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     COLLECT_ARGS 4
@@ -612,6 +615,7 @@ EXTN(jsimd_h2v1_upsample_avx2):
     GLOBAL_FUNCTION(jsimd_h2v2_upsample_avx2)
 
 EXTN(jsimd_h2v2_upsample_avx2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     COLLECT_ARGS 4

diff --git a/simd/x86_64/jdsample-sse2.asm b/simd/x86_64/jdsample-sse2.asm
@@ -61,6 +61,7 @@ PW_EIGHT times 8 dw 8
     GLOBAL_FUNCTION(jsimd_h2v1_fancy_upsample_sse2)
 
 EXTN(jsimd_h2v1_fancy_upsample_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     COLLECT_ARGS 4
@@ -202,6 +203,7 @@ EXTN(jsimd_h2v1_fancy_upsample_sse2):
     GLOBAL_FUNCTION(jsimd_h2v2_fancy_upsample_sse2)
 
 EXTN(jsimd_h2v2_fancy_upsample_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     push        r15
@@ -497,6 +499,7 @@ EXTN(jsimd_h2v2_fancy_upsample_sse2):
     GLOBAL_FUNCTION(jsimd_h2v1_upsample_sse2)
 
 EXTN(jsimd_h2v1_upsample_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     COLLECT_ARGS 4
@@ -583,6 +586,7 @@ EXTN(jsimd_h2v1_upsample_sse2):
     GLOBAL_FUNCTION(jsimd_h2v2_upsample_sse2)
 
 EXTN(jsimd_h2v2_upsample_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     COLLECT_ARGS 4

diff --git a/simd/x86_64/jfdctflt-sse.asm b/simd/x86_64/jfdctflt-sse.asm
@@ -66,6 +66,7 @@ PD_1_306 times 4 dd 1.306562964876376527856643
     GLOBAL_FUNCTION(jsimd_fdct_float_sse)
 
 EXTN(jsimd_fdct_float_sse):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     push        r15

diff --git a/simd/x86_64/jfdctfst-sse2.asm b/simd/x86_64/jfdctfst-sse2.asm
@@ -81,6 +81,7 @@ PW_F1306 times 8 dw F_1_306 << CONST_SHIFT
     GLOBAL_FUNCTION(jsimd_fdct_ifast_sse2)
 
 EXTN(jsimd_fdct_ifast_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     push        r15

diff --git a/simd/x86_64/jfdctint-avx2.asm b/simd/x86_64/jfdctint-avx2.asm
@@ -260,6 +260,7 @@ PW_1_NEG1                  times 8  dw  1
     GLOBAL_FUNCTION(jsimd_fdct_islow_avx2)
 
 EXTN(jsimd_fdct_islow_avx2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     COLLECT_ARGS 1

diff --git a/simd/x86_64/jfdctint-sse2.asm b/simd/x86_64/jfdctint-sse2.asm
@@ -102,6 +102,7 @@ PW_DESCALE_P2X times 8 dw  1 << (PASS1_BITS - 1)
     GLOBAL_FUNCTION(jsimd_fdct_islow_sse2)
 
 EXTN(jsimd_fdct_islow_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     push        r15

diff --git a/simd/x86_64/jidctflt-sse2.asm b/simd/x86_64/jidctflt-sse2.asm
@@ -76,6 +76,7 @@ PB_CENTERJSAMP  times 16 db  CENTERJSAMPLE
     GLOBAL_FUNCTION(jsimd_idct_float_sse2)
 
 EXTN(jsimd_idct_float_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     push        r15

diff --git a/simd/x86_64/jidctfst-sse2.asm b/simd/x86_64/jidctfst-sse2.asm
@@ -95,6 +95,7 @@ PB_CENTERJSAMP times 16 db  CENTERJSAMPLE
     GLOBAL_FUNCTION(jsimd_idct_ifast_sse2)
 
 EXTN(jsimd_idct_ifast_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     push        r15

diff --git a/simd/x86_64/jidctint-avx2.asm b/simd/x86_64/jidctint-avx2.asm
@@ -282,6 +282,7 @@ PW_1_NEG1                  times 8  dw  1
     GLOBAL_FUNCTION(jsimd_idct_islow_avx2)
 
 EXTN(jsimd_idct_islow_avx2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp                     ; rbp = aligned rbp
     PUSH_XMM    4

diff --git a/simd/x86_64/jidctint-sse2.asm b/simd/x86_64/jidctint-sse2.asm
@@ -108,6 +108,7 @@ PB_CENTERJSAMP times 16 db  CENTERJSAMPLE
     GLOBAL_FUNCTION(jsimd_idct_islow_sse2)
 
 EXTN(jsimd_idct_islow_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     push        r15

diff --git a/simd/x86_64/jidctred-sse2.asm b/simd/x86_64/jidctred-sse2.asm
@@ -116,6 +116,7 @@ PB_CENTERJSAMP  times 16 db  CENTERJSAMPLE
     GLOBAL_FUNCTION(jsimd_idct_4x4_sse2)
 
 EXTN(jsimd_idct_4x4_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     push        r15
@@ -413,6 +414,7 @@ EXTN(jsimd_idct_4x4_sse2):
     GLOBAL_FUNCTION(jsimd_idct_2x2_sse2)
 
 EXTN(jsimd_idct_2x2_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     COLLECT_ARGS 4

diff --git a/simd/x86_64/jquantf-sse2.asm b/simd/x86_64/jquantf-sse2.asm
@@ -37,6 +37,7 @@
     GLOBAL_FUNCTION(jsimd_convsamp_float_sse2)
 
 EXTN(jsimd_convsamp_float_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     COLLECT_ARGS 3
@@ -109,6 +110,7 @@ EXTN(jsimd_convsamp_float_sse2):
     GLOBAL_FUNCTION(jsimd_quantize_float_sse2)
 
 EXTN(jsimd_quantize_float_sse2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     COLLECT_ARGS 3

diff --git a/simd/x86_64/jquanti-avx2.asm b/simd/x86_64/jquanti-avx2.asm
@@ -38,6 +38,7 @@
     GLOBAL_FUNCTION(jsimd_convsamp_avx2)
 
 EXTN(jsimd_convsamp_avx2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     COLLECT_ARGS 3
@@ -115,6 +116,7 @@ EXTN(jsimd_convsamp_avx2):
     GLOBAL_FUNCTION(jsimd_quantize_avx2)
 
 EXTN(jsimd_quantize_avx2):
+    ENDBR64
     push        rbp
     mov         rbp, rsp
     COLLECT_ARGS 3