Index: compiler-rt/CMakeLists.txt
===================================================================
--- compiler-rt/CMakeLists.txt
+++ compiler-rt/CMakeLists.txt
@@ -49,7 +49,8 @@
 mark_as_advanced(COMPILER_RT_BUILD_ORC)
 option(COMPILER_RT_BUILD_GWP_ASAN "Build GWP-ASan, and link it into SCUDO" ON)
 mark_as_advanced(COMPILER_RT_BUILD_GWP_ASAN)
-
+option(COMPILER_RT_ENABLE_CET "Build Compiler RT with CET enabled" OFF)
+mark_as_advanced(COMPILER_RT_ENABLE_CET)
 set(COMPILER_RT_ASAN_SHADOW_SCALE ""
     CACHE STRING "Override the shadow scale to be used in ASan runtime")
 
@@ -69,6 +70,17 @@
       -D${COMPILER_RT_ASAN_SHADOW_SCALE_DEFINITION})
 endif()
 
+if (COMPILER_RT_ENABLE_CET)
+  if (MSVC)
+    message(FATAL_ERROR "compiler-rt CET support is not available for MSVC!")
+  endif()
+
+  check_c_compiler_flag(-fcf-protection=full COMPILER_RT_HAS_FCF_PROTECTION_FLAG)
+  if(NOT COMPILER_RT_HAS_FCF_PROTECTION_FLAG)
+    message(FATAL_ERROR "Compiler used to build compiler-rt doesn't support CET!")
+  endif()
+endif()
+
 if(FUCHSIA)
   set(COMPILER_RT_HWASAN_WITH_INTERCEPTORS_DEFAULT OFF)
 else()
Index: compiler-rt/lib/builtins/CMakeLists.txt
===================================================================
--- compiler-rt/lib/builtins/CMakeLists.txt
+++ compiler-rt/lib/builtins/CMakeLists.txt
@@ -674,6 +674,8 @@
 else ()
   set(BUILTIN_CFLAGS "")
 
+  append_list_if(COMPILER_RT_ENABLE_CET -fcf-protection=full BUILTIN_CFLAGS)
+
   append_list_if(COMPILER_RT_HAS_FLOAT16 -DCOMPILER_RT_HAS_FLOAT16 BUILTIN_CFLAGS)
 
   append_list_if(COMPILER_RT_HAS_STD_C11_FLAG -std=c11 BUILTIN_CFLAGS)
Index: compiler-rt/lib/builtins/assembly.h
===================================================================
--- compiler-rt/lib/builtins/assembly.h
+++ compiler-rt/lib/builtins/assembly.h
@@ -14,6 +14,14 @@
 #ifndef COMPILERRT_ASSEMBLY_H
 #define COMPILERRT_ASSEMBLY_H
 
+#if (defined(__i386__) || defined(__x86_64__)) && defined(__linux__) &&        \
+    defined(__CET__)
+#include <cet.h>
+#define _COMPILER_RT_ENDBR _CET_ENDBR
+#else
+#define _COMPILER_RT_ENDBR
+#endif
+
 #if defined(__APPLE__) && defined(__aarch64__)
 #define SEPARATOR %%
 #else
Index: compiler-rt/lib/builtins/i386/ashldi3.S
===================================================================
--- compiler-rt/lib/builtins/i386/ashldi3.S
+++ compiler-rt/lib/builtins/i386/ashldi3.S
@@ -19,6 +19,7 @@
 .text
 .balign 4
 DEFINE_COMPILERRT_FUNCTION(__ashldi3)
+	_COMPILER_RT_ENDBR
 	movd	  12(%esp),		%xmm2	// Load count
 #ifndef TRUST_CALLERS_USE_64_BIT_STORES
 	movd	   4(%esp),		%xmm0
@@ -39,6 +40,7 @@
 .text
 .balign 4
 DEFINE_COMPILERRT_FUNCTION(__ashldi3)
+	_COMPILER_RT_ENDBR
 	movl	  12(%esp),		%ecx	// Load count
 	movl	   8(%esp),		%edx	// Load high
 	movl	   4(%esp),		%eax	// Load low
Index: compiler-rt/lib/builtins/i386/ashrdi3.S
===================================================================
--- compiler-rt/lib/builtins/i386/ashrdi3.S
+++ compiler-rt/lib/builtins/i386/ashrdi3.S
@@ -12,6 +12,7 @@
 .text
 .balign 4
 DEFINE_COMPILERRT_FUNCTION(__ashrdi3)
+	_COMPILER_RT_ENDBR
 	movd	  12(%esp),		%xmm2	// Load count
 	movl	   8(%esp),		%eax
 #ifndef TRUST_CALLERS_USE_64_BIT_STORES
@@ -49,6 +50,7 @@
 .text
 .balign 4
 DEFINE_COMPILERRT_FUNCTION(__ashrdi3)
+	_COMPILER_RT_ENDBR
 	movl	  12(%esp),		%ecx	// Load count
 	movl	   8(%esp),		%edx	// Load high
 	movl	   4(%esp),		%eax	// Load low
Index: compiler-rt/lib/builtins/i386/divdi3.S
===================================================================
--- compiler-rt/lib/builtins/i386/divdi3.S
+++ compiler-rt/lib/builtins/i386/divdi3.S
@@ -27,6 +27,7 @@
 // value, then restoring the correct sign at the end of the computation.  This could
 // certainly be improved upon.
 
+	_COMPILER_RT_ENDBR
 	pushl		%esi
 	movl	 20(%esp),			%edx	// high word of b
 	movl	 16(%esp),			%eax	// low word of b
Index: compiler-rt/lib/builtins/i386/floatdidf.S
===================================================================
--- compiler-rt/lib/builtins/i386/floatdidf.S
+++ compiler-rt/lib/builtins/i386/floatdidf.S
@@ -23,6 +23,7 @@
 .text
 .balign 4
 DEFINE_COMPILERRT_FUNCTION(__floatdidf)
+	_COMPILER_RT_ENDBR
 	cvtsi2sd	8(%esp),			%xmm1
 	movss		4(%esp),			%xmm0 // low 32 bits of a
 	calll		0f
Index: compiler-rt/lib/builtins/i386/floatdisf.S
===================================================================
--- compiler-rt/lib/builtins/i386/floatdisf.S
+++ compiler-rt/lib/builtins/i386/floatdisf.S
@@ -18,6 +18,7 @@
 .text
 .balign 4
 DEFINE_COMPILERRT_FUNCTION(__floatdisf)
+	_COMPILER_RT_ENDBR
 #ifndef TRUST_CALLERS_USE_64_BIT_STORES
 	movd		4(%esp),	%xmm0
 	movd		8(%esp),	%xmm1
Index: compiler-rt/lib/builtins/i386/floatdixf.S
===================================================================
--- compiler-rt/lib/builtins/i386/floatdixf.S
+++ compiler-rt/lib/builtins/i386/floatdixf.S
@@ -18,6 +18,7 @@
 .text
 .balign 4
 DEFINE_COMPILERRT_FUNCTION(__floatdixf)
+	_COMPILER_RT_ENDBR
 #ifndef TRUST_CALLERS_USE_64_BIT_STORES
 	movd		4(%esp),	%xmm0
 	movd		8(%esp),	%xmm1
Index: compiler-rt/lib/builtins/i386/floatundidf.S
===================================================================
--- compiler-rt/lib/builtins/i386/floatundidf.S
+++ compiler-rt/lib/builtins/i386/floatundidf.S
@@ -35,6 +35,7 @@
 .text
 .balign 4
 DEFINE_COMPILERRT_FUNCTION(__floatundidf)
+	_COMPILER_RT_ENDBR
 	movss	8(%esp),						%xmm1 // high 32 bits of a
 	movss	4(%esp),						%xmm0 // low 32 bits of a
 	calll	0f
Index: compiler-rt/lib/builtins/i386/floatundisf.S
===================================================================
--- compiler-rt/lib/builtins/i386/floatundisf.S
+++ compiler-rt/lib/builtins/i386/floatundisf.S
@@ -75,6 +75,7 @@
 .text
 .balign 4
 DEFINE_COMPILERRT_FUNCTION(__floatundisf)
+	_COMPILER_RT_ENDBR
 	movl		8(%esp),		%eax
 	movd		8(%esp),		%xmm1
 	movd		4(%esp),		%xmm0
Index: compiler-rt/lib/builtins/i386/floatundixf.S
===================================================================
--- compiler-rt/lib/builtins/i386/floatundixf.S
+++ compiler-rt/lib/builtins/i386/floatundixf.S
@@ -27,6 +27,7 @@
 .text
 .balign 4
 DEFINE_COMPILERRT_FUNCTION(__floatundixf)
+	_COMPILER_RT_ENDBR
 	calll	0f
 0:	popl	%eax
 	movss	8(%esp),			%xmm0	// hi 32 bits of input
Index: compiler-rt/lib/builtins/i386/lshrdi3.S
===================================================================
--- compiler-rt/lib/builtins/i386/lshrdi3.S
+++ compiler-rt/lib/builtins/i386/lshrdi3.S
@@ -19,6 +19,7 @@
 .text
 .balign 4
 DEFINE_COMPILERRT_FUNCTION(__lshrdi3)
+	_COMPILER_RT_ENDBR
 	movd	  12(%esp),		%xmm2	// Load count
 #ifndef TRUST_CALLERS_USE_64_BIT_STORES
 	movd	   4(%esp),		%xmm0
@@ -39,6 +40,7 @@
 .text
 .balign 4
 DEFINE_COMPILERRT_FUNCTION(__lshrdi3)
+	_COMPILER_RT_ENDBR
 	movl	  12(%esp),		%ecx	// Load count
 	movl	   8(%esp),		%edx	// Load high
 	movl	   4(%esp),		%eax	// Load low
Index: compiler-rt/lib/builtins/i386/moddi3.S
===================================================================
--- compiler-rt/lib/builtins/i386/moddi3.S
+++ compiler-rt/lib/builtins/i386/moddi3.S
@@ -27,6 +27,7 @@
 // This is currently implemented by wrapping the unsigned modulus up in an absolute
 // value.  This could certainly be improved upon.
 
+	_COMPILER_RT_ENDBR
 	pushl		%esi
 	movl	 20(%esp),			%edx	// high word of b
 	movl	 16(%esp),			%eax	// low word of b
Index: compiler-rt/lib/builtins/i386/muldi3.S
===================================================================
--- compiler-rt/lib/builtins/i386/muldi3.S
+++ compiler-rt/lib/builtins/i386/muldi3.S
@@ -11,6 +11,7 @@
 .text
 .balign 4
 DEFINE_COMPILERRT_FUNCTION(__muldi3)
+	_COMPILER_RT_ENDBR
 	pushl	%ebx
 	movl  16(%esp),		%eax	// b.lo
 	movl  12(%esp),		%ecx	// a.hi
Index: compiler-rt/lib/builtins/i386/udivdi3.S
===================================================================
--- compiler-rt/lib/builtins/i386/udivdi3.S
+++ compiler-rt/lib/builtins/i386/udivdi3.S
@@ -23,6 +23,7 @@
 .balign 4
 DEFINE_COMPILERRT_FUNCTION(__udivdi3)
 
+        _COMPILER_RT_ENDBR
 	pushl		%ebx
 	movl	 20(%esp),			%ebx	// Find the index i of the leading bit in b.
 	bsrl		%ebx,			%ecx	// If the high word of b is zero, jump to
Index: compiler-rt/lib/builtins/i386/umoddi3.S
===================================================================
--- compiler-rt/lib/builtins/i386/umoddi3.S
+++ compiler-rt/lib/builtins/i386/umoddi3.S
@@ -23,7 +23,7 @@
 .text
 .balign 4
 DEFINE_COMPILERRT_FUNCTION(__umoddi3)
-
+	_COMPILER_RT_ENDBR
 	pushl		%ebx
 	movl	 20(%esp),			%ebx	// Find the index i of the leading bit in b.
 	bsrl		%ebx,			%ecx	// If the high word of b is zero, jump to
Index: compiler-rt/lib/builtins/x86_64/floatundidf.S
===================================================================
--- compiler-rt/lib/builtins/x86_64/floatundidf.S
+++ compiler-rt/lib/builtins/x86_64/floatundidf.S
@@ -35,6 +35,7 @@
 .text
 .balign 4
 DEFINE_COMPILERRT_FUNCTION(__floatundidf)
+	_COMPILER_RT_ENDBR
 	movd	%edi,							%xmm0 // low 32 bits of a
 	shrq	$32,							%rdi  // high 32 bits of a
 	orq		REL_ADDR(twop84),				%rdi  // 0x1p84 + a_hi (no rounding occurs)
Index: compiler-rt/lib/builtins/x86_64/floatundisf.S
===================================================================
--- compiler-rt/lib/builtins/x86_64/floatundisf.S
+++ compiler-rt/lib/builtins/x86_64/floatundisf.S
@@ -19,6 +19,7 @@
 .text
 .balign 4
 DEFINE_COMPILERRT_FUNCTION(__floatundisf)
+	_COMPILER_RT_ENDBR
 	movq		$1,			%rsi
 	testq		%rdi,		%rdi
 	js			1f
Index: compiler-rt/lib/builtins/x86_64/floatundixf.S
===================================================================
--- compiler-rt/lib/builtins/x86_64/floatundixf.S
+++ compiler-rt/lib/builtins/x86_64/floatundixf.S
@@ -20,6 +20,7 @@
 
 	.balign 4
 DEFINE_COMPILERRT_FUNCTION(__floatundixf)
+	_COMPILER_RT_ENDBR
 	movq	%rdi,	 -8(%rsp)
 	fildq	-8(%rsp)
 	test	%rdi,		%rdi
Index: compiler-rt/test/builtins/CMakeLists.txt
===================================================================
--- compiler-rt/test/builtins/CMakeLists.txt
+++ compiler-rt/test/builtins/CMakeLists.txt
@@ -49,6 +49,11 @@
     string(REPLACE ";" " " BUILTINS_TEST_TARGET_CFLAGS "${BUILTINS_TEST_TARGET_CFLAGS}")
   endif()
 
+  if (${arch} MATCHES "i386|x86_64" AND COMPILER_RT_ENABLE_CET)
+    list(APPEND BUILTINS_TEST_TARGET_CFLAGS -fcf-protection=full)
+    string(REPLACE ";" " " BUILTINS_TEST_TARGET_CFLAGS "${BUILTINS_TEST_TARGET_CFLAGS}")
+  endif()
+
   # Compute builtins available in library and add them as lit features.
   if(APPLE)
     # TODO: Support other Apple platforms.