ffi 1.11.1 → 1.15.5

data/ext/ffi_c/libffi/src/powerpc/linux64.S

@@ -36,8 +36,10 @@
 	.cfi_startproc
 # if _CALL_ELF == 2
 ffi_call_LINUX64:
+#  ifndef __PCREL__
 	addis	%r2, %r12, .TOC.-ffi_call_LINUX64@ha
 	addi	%r2, %r2, .TOC.-ffi_call_LINUX64@l
+#  endif
 	.localentry ffi_call_LINUX64, . - ffi_call_LINUX64
 # else
 	.section	".opd","aw"
@@ -89,9 +91,15 @@ ffi_call_LINUX64:
 	/* Call ffi_prep_args64.  */
 	mr	%r4, %r1
 # if defined _CALL_LINUX || _CALL_ELF == 2
+#  ifdef __PCREL__
+	bl	ffi_prep_args64@notoc
+#  else
 	bl	ffi_prep_args64
+	nop
+#  endif
 # else
 	bl	.ffi_prep_args64
+	nop
 # endif
 
 # if _CALL_ELF == 2
@@ -101,40 +109,70 @@ ffi_call_LINUX64:
 	ld	%r2, 8(%r29)
 # endif
 	/* Now do the call.  */
-	/* Set up cr1 with bits 4-7 of the flags.  */
-	mtcrf	0x40, %r31
+	/* Set up cr1 with bits 3-7 of the flags.  */
+	mtcrf	0xc0, %r31
 
 	/* Get the address to call into CTR.  */
 	mtctr	%r12
 	/* Load all those argument registers.  */
-	ld	%r3, -32-(8*8)(%r28)
-	ld	%r4, -32-(7*8)(%r28)
-	ld	%r5, -32-(6*8)(%r28)
-	ld	%r6, -32-(5*8)(%r28)
+	addi	%r29, %r28, -32-(8*8)
+	ld	%r3,  (0*8)(%r29)
+	ld	%r4,  (1*8)(%r29)
+	ld	%r5,  (2*8)(%r29)
+	ld	%r6,  (3*8)(%r29)
 	bf-	5, 1f
-	ld	%r7, -32-(4*8)(%r28)
-	ld	%r8, -32-(3*8)(%r28)
-	ld	%r9, -32-(2*8)(%r28)
-	ld	%r10, -32-(1*8)(%r28)
+	ld	%r7,  (4*8)(%r29)
+	ld	%r8,  (5*8)(%r29)
+	ld	%r9,  (6*8)(%r29)
+	ld	%r10, (7*8)(%r29)
 1:
 
 	/* Load all the FP registers.  */
 	bf-	6, 2f
-	lfd	%f1, -32-(21*8)(%r28)
-	lfd	%f2, -32-(20*8)(%r28)
-	lfd	%f3, -32-(19*8)(%r28)
-	lfd	%f4, -32-(18*8)(%r28)
-	lfd	%f5, -32-(17*8)(%r28)
-	lfd	%f6, -32-(16*8)(%r28)
-	lfd	%f7, -32-(15*8)(%r28)
-	lfd	%f8, -32-(14*8)(%r28)
-	lfd	%f9, -32-(13*8)(%r28)
-	lfd	%f10, -32-(12*8)(%r28)
-	lfd	%f11, -32-(11*8)(%r28)
-	lfd	%f12, -32-(10*8)(%r28)
-	lfd	%f13, -32-(9*8)(%r28)
+	addi	%r29, %r29, -(14*8)
+	lfd	%f1,  ( 1*8)(%r29)
+	lfd	%f2,  ( 2*8)(%r29)
+	lfd	%f3,  ( 3*8)(%r29)
+	lfd	%f4,  ( 4*8)(%r29)
+	lfd	%f5,  ( 5*8)(%r29)
+	lfd	%f6,  ( 6*8)(%r29)
+	lfd	%f7,  ( 7*8)(%r29)
+	lfd	%f8,  ( 8*8)(%r29)
+	lfd	%f9,  ( 9*8)(%r29)
+	lfd	%f10, (10*8)(%r29)
+	lfd	%f11, (11*8)(%r29)
+	lfd	%f12, (12*8)(%r29)
+	lfd	%f13, (13*8)(%r29)
 2:
 
+	/* Load all the vector registers.  */
+	bf-	3, 3f
+	addi	%r29, %r29, -16
+	lvx	%v13, 0, %r29
+	addi	%r29, %r29, -16
+	lvx	%v12, 0, %r29
+	addi	%r29, %r29, -16
+	lvx	%v11, 0, %r29
+	addi	%r29, %r29, -16
+	lvx	%v10, 0, %r29
+	addi	%r29, %r29, -16
+	lvx	%v9,  0, %r29
+	addi	%r29, %r29, -16
+	lvx	%v8,  0, %r29
+	addi	%r29, %r29, -16
+	lvx	%v7,  0, %r29
+	addi	%r29, %r29, -16
+	lvx	%v6,  0, %r29
+	addi	%r29, %r29, -16
+	lvx	%v5,  0, %r29
+	addi	%r29, %r29, -16
+	lvx	%v4,  0, %r29
+	addi	%r29, %r29, -16
+	lvx	%v3,  0, %r29
+	addi	%r29, %r29, -16
+	lvx	%v2,  0, %r29
+3:
+
 	/* Make the call.  */
 	ld	%r11, 8(%r28)
 	bctrl
@@ -152,6 +190,7 @@ ffi_call_LINUX64:
 	bt	31, .Lstruct_return_value
 	bt	30, .Ldone_return_value
 	bt	29, .Lfp_return_value
+	bt	28, .Lvec_return_value
 	std	%r3, 0(%r30)
 	/* Fall through...  */
 
@@ -167,12 +206,16 @@ ffi_call_LINUX64:
 	ld	%r31, -8(%r1)
 	blr
 
+.Lvec_return_value:
+	stvx	%v2, 0, %r30
+	b	.Ldone_return_value
+
 .Lfp_return_value:
 	.cfi_def_cfa_register 28
-	bf	28, .Lfloat_return_value
-	stfd	%f1, 0(%r30)
 	mtcrf	0x02, %r31 /* cr6  */
-	bf	27, .Ldone_return_value
+	bf	27, .Lfloat_return_value
+	stfd	%f1, 0(%r30)
+	bf	26, .Ldone_return_value
 	stfd	%f2, 8(%r30)
 	b	.Ldone_return_value
 .Lfloat_return_value:
@@ -180,8 +223,9 @@ ffi_call_LINUX64:
 	b	.Ldone_return_value
 
 .Lstruct_return_value:
-	bf	29, .Lsmall_struct
-	bf	28, .Lfloat_homog_return_value
+	bf	29, .Lvec_homog_or_small_struct
+	mtcrf	0x02, %r31 /* cr6  */
+	bf	27, .Lfloat_homog_return_value
 	stfd	%f1, 0(%r30)
 	stfd	%f2, 8(%r30)
 	stfd	%f3, 16(%r30)
@@ -203,6 +247,25 @@ ffi_call_LINUX64:
 	stfs	%f8, 28(%r30)
 	b	.Ldone_return_value
 
+.Lvec_homog_or_small_struct:
+	bf	28, .Lsmall_struct
+	stvx	%v2, 0, %r30
+	addi	%r30, %r30, 16
+	stvx	%v3, 0, %r30
+	addi	%r30, %r30, 16
+	stvx	%v4, 0, %r30
+	addi	%r30, %r30, 16
+	stvx	%v5, 0, %r30
+	addi	%r30, %r30, 16
+	stvx	%v6, 0, %r30
+	addi	%r30, %r30, 16
+	stvx	%v7, 0, %r30
+	addi	%r30, %r30, 16
+	stvx	%v8, 0, %r30
+	addi	%r30, %r30, 16
+	stvx	%v9, 0, %r30
+	b	.Ldone_return_value
+
 .Lsmall_struct:
 	std	%r3, 0(%r30)
 	std	%r4, 8(%r30)

data/ext/ffi_c/libffi/src/powerpc/linux64_closure.S

@@ -37,8 +37,10 @@
 	.cfi_startproc
 # if _CALL_ELF == 2
 ffi_closure_LINUX64:
+#  ifndef __PCREL__
 	addis	%r2, %r12, .TOC.-ffi_closure_LINUX64@ha
 	addi	%r2, %r2, .TOC.-ffi_closure_LINUX64@l
+#  endif
 	.localentry ffi_closure_LINUX64, . - ffi_closure_LINUX64
 # else
 	.section        ".opd","aw"
@@ -61,9 +63,15 @@ ffi_closure_LINUX64:
 # endif
 
 # if _CALL_ELF == 2
-#  32 byte special reg save area + 64 byte parm save area
-#  + 64 byte retval area + 13*8 fpr save area + round to 16
-#  define STACKFRAME 272
+#  ifdef __VEC__
+#   32 byte special reg save area + 64 byte parm save area
+#   + 128 byte retval area + 13*8 fpr save area + 12*16 vec save area + round to 16
+#   define STACKFRAME 528
+#  else
+#   32 byte special reg save area + 64 byte parm save area
+#   + 64 byte retval area + 13*8 fpr save area + round to 16
+#   define STACKFRAME 272
+#  endif
 #  define PARMSAVE 32
 #  define RETVAL PARMSAVE+64
 # else
@@ -143,9 +151,38 @@ ffi_closure_LINUX64:
 	stfd	%f12, -104+(11*8)(%r1)
 	stfd	%f13, -104+(12*8)(%r1)
 
-	# load up the pointer to the saved fpr registers */
+	# load up the pointer to the saved fpr registers
 	addi	%r8, %r1, -104
 
+# ifdef __VEC__
+	# load up the pointer to the saved vector registers
+	# 8 bytes padding for 16-byte alignment at -112(%r1)
+	addi	%r9, %r8, -24
+	stvx	%v13, 0, %r9
+	addi	%r9, %r9, -16
+	stvx	%v12, 0, %r9
+	addi	%r9, %r9, -16
+	stvx	%v11, 0, %r9
+	addi	%r9, %r9, -16
+	stvx	%v10, 0, %r9
+	addi	%r9, %r9, -16
+	stvx	%v9, 0, %r9
+	addi	%r9, %r9, -16
+	stvx	%v8, 0, %r9
+	addi	%r9, %r9, -16
+	stvx	%v7, 0, %r9
+	addi	%r9, %r9, -16
+	stvx	%v6, 0, %r9
+	addi	%r9, %r9, -16
+	stvx	%v5, 0, %r9
+	addi	%r9, %r9, -16
+	stvx	%v4, 0, %r9
+	addi	%r9, %r9, -16
+	stvx	%v3, 0, %r9
+	addi	%r9, %r9, -16
+	stvx	%v2, 0, %r9
+# endif
+
 	# load up the pointer to the result storage
 	addi	%r6, %r1, -STACKFRAME+RETVAL
 
@@ -155,11 +192,19 @@ ffi_closure_LINUX64:
 
 	# make the call
 # if defined _CALL_LINUX || _CALL_ELF == 2
+#  ifdef __PCREL__
+	bl ffi_closure_helper_LINUX64@notoc
+.Lret:
+#  else
 	bl ffi_closure_helper_LINUX64
+.Lret:
+	nop
+#  endif
 # else
 	bl .ffi_closure_helper_LINUX64
-# endif
 .Lret:
+	nop
+# endif
 
 	# now r3 contains the return type
 	# so use it to look up in a table
@@ -313,6 +358,16 @@ ffi_closure_LINUX64:
 	.cfi_def_cfa_offset 0
 	blr
 	.cfi_def_cfa_offset STACKFRAME
+# case FFI_V2_TYPE_VECTOR
+	addi %r3, %r1, RETVAL
+	lvx %v2, 0, %r3
+	mtlr %r0
+	b .Lfinish
+# case FFI_V2_TYPE_VECTOR_HOMOG
+	addi %r3, %r1, RETVAL
+	lvx %v2, 0, %r3
+	addi %r3, %r3, 16
+	b .Lmorevector
 # case FFI_V2_TYPE_FLOAT_HOMOG
 	lfs %f1, RETVAL+0(%r1)
 	lfs %f2, RETVAL+4(%r1)
@@ -332,6 +387,25 @@ ffi_closure_LINUX64:
 	.cfi_def_cfa_offset 0
 	blr
 	.cfi_def_cfa_offset STACKFRAME
+.Lmorevector:
+	lvx %v3, 0, %r3
+	addi %r3, %r3, 16
+	lvx %v4, 0, %r3
+	addi %r3, %r3, 16
+	lvx %v5, 0, %r3
+	mtlr %r0
+	addi %r3, %r3, 16
+	lvx %v6, 0, %r3
+	addi %r3, %r3, 16
+	lvx %v7, 0, %r3
+	addi %r3, %r3, 16
+	lvx %v8, 0, %r3
+	addi %r3, %r3, 16
+	lvx %v9, 0, %r3
+	addi %r1, %r1, STACKFRAME
+	.cfi_def_cfa_offset 0
+	blr
+	.cfi_def_cfa_offset STACKFRAME
 .Lmorefloat:
 	lfs %f4, RETVAL+12(%r1)
 	mtlr %r0
@@ -396,8 +470,10 @@ ffi_closure_LINUX64:
 	.cfi_startproc
 # if _CALL_ELF == 2
 ffi_go_closure_linux64:
+#  ifndef __PCREL__
 	addis	%r2, %r12, .TOC.-ffi_go_closure_linux64@ha
 	addi	%r2, %r2, .TOC.-ffi_go_closure_linux64@l
+#  endif
 	.localentry ffi_go_closure_linux64, . - ffi_go_closure_linux64
 # else
 	.section        ".opd","aw"

data/ext/ffi_c/libffi/src/powerpc/sysv.S

@@ -104,17 +104,16 @@ ENTRY(ffi_call_SYSV)
 	bctrl
 
 	/* Now, deal with the return value.  */
-	mtcrf	0x01,%r31 /* cr7  */
+	mtcrf	0x03,%r31 /* cr6-cr7  */
 	bt-	31,L(small_struct_return_value)
 	bt-	30,L(done_return_value)
 #ifndef __NO_FPRS__
 	bt-	29,L(fp_return_value)
 #endif
 	stw	%r3,0(%r30)
-	bf+	28,L(done_return_value)
+	bf+	27,L(done_return_value)
 	stw	%r4,4(%r30)
-	mtcrf	0x02,%r31 /* cr6  */
-	bf	27,L(done_return_value)
+	bf	26,L(done_return_value)
 	stw     %r5,8(%r30)
 	stw	%r6,12(%r30)
 	/* Fall through...  */
@@ -145,10 +144,9 @@ L(done_return_value):
 #ifndef __NO_FPRS__
 L(fp_return_value):
 	.cfi_restore_state
-	bf	28,L(float_return_value)
+	bf	27,L(float_return_value)
 	stfd	%f1,0(%r30)
-	mtcrf   0x02,%r31 /* cr6  */
-	bf	27,L(done_return_value)
+	bf	26,L(done_return_value)
 	stfd	%f2,8(%r30)
 	b	L(done_return_value)
 L(float_return_value):

data/ext/ffi_c/libffi/src/prep_cif.c

@@ -129,7 +129,9 @@ ffi_status FFI_HIDDEN ffi_prep_cif_core(ffi_cif *cif, ffi_abi abi,
   cif->rtype = rtype;
 
   cif->flags = 0;
-
+#if (defined(_M_ARM64) || defined(__aarch64__)) && defined(_WIN32)
+  cif->is_variadic = isvariadic;
+#endif
 #if HAVE_LONG_DOUBLE_VARIANT
   ffi_prep_types (abi);
 #endif
@@ -199,7 +201,7 @@ ffi_status FFI_HIDDEN ffi_prep_cif_core(ffi_cif *cif, ffi_abi abi,
 	    bytes = 6*4;
 #endif
 
-	  bytes += STACK_ARG_SIZE((*ptr)->size);
+	  bytes += (unsigned int)STACK_ARG_SIZE((*ptr)->size);
 	}
 #endif
     }

data/ext/ffi_c/libffi/src/x86/ffi.c

@@ -29,7 +29,7 @@
    DEALINGS IN THE SOFTWARE.
    ----------------------------------------------------------------------- */
 
-#ifndef __x86_64__
+#if defined(__i386__) || defined(_M_IX86)
 #include <ffi.h>
 #include <ffi_common.h>
 #include <stdint.h>
@@ -51,6 +51,13 @@
 # define __declspec(x)  __attribute__((x))
 #endif
 
+#if defined(_MSC_VER) && defined(_M_IX86)
+/* Stack is not 16-byte aligned on Windows.  */
+#define STACK_ALIGN(bytes) (bytes)
+#else
+#define STACK_ALIGN(bytes) FFI_ALIGN (bytes, 16)
+#endif
+
 /* Perform machine dependent cif processing.  */
 ffi_status FFI_HIDDEN
 ffi_prep_cif_machdep(ffi_cif *cif)
@@ -177,7 +184,7 @@ ffi_prep_cif_machdep(ffi_cif *cif)
       bytes = FFI_ALIGN (bytes, t->alignment);
       bytes += FFI_ALIGN (t->size, FFI_SIZEOF_ARG);
     }
-  cif->bytes = FFI_ALIGN (bytes, FFI_SIZEOF_ARG);
+  cif->bytes = bytes;
 
   return FFI_OK;
 }
@@ -285,7 +292,7 @@ ffi_call_int (ffi_cif *cif, void (*fn)(void), void *rvalue,
 	}
     }
 
-  bytes = cif->bytes;
+  bytes = STACK_ALIGN (cif->bytes);
   stack = alloca(bytes + sizeof(*frame) + rsize);
   argp = (dir < 0 ? stack + bytes : stack);
   frame = (struct call_frame *)(stack + bytes);
@@ -346,7 +353,7 @@ ffi_call_int (ffi_cif *cif, void (*fn)(void), void *rvalue,
 	  size_t align = FFI_SIZEOF_ARG;
 
 	  /* Issue 434: For thiscall and fastcall, if the paramter passed
-	     as 64-bit integer or struct, all following integer paramters
+	     as 64-bit integer or struct, all following integer parameters
 	     will be passed on stack.  */
 	  if ((cabi == FFI_THISCALL || cabi == FFI_FASTCALL)
 	      && (t == FFI_TYPE_SINT64
@@ -390,12 +397,14 @@ ffi_call (ffi_cif *cif, void (*fn)(void), void *rvalue, void **avalue)
   ffi_call_int (cif, fn, rvalue, avalue, NULL);
 }
 
+#ifdef FFI_GO_CLOSURES
 void
 ffi_call_go (ffi_cif *cif, void (*fn)(void), void *rvalue,
 	     void **avalue, void *closure)
 {
   ffi_call_int (cif, fn, rvalue, avalue, closure);
 }
+#endif
 
 /** private members **/
 
@@ -429,7 +438,7 @@ ffi_closure_inner (struct closure_frame *frame, char *stack)
   rvalue = frame->rettemp;
   pabi = &abi_params[cabi];
   dir = pabi->dir;
-  argp = (dir < 0 ? stack + cif->bytes : stack);
+  argp = (dir < 0 ? stack + STACK_ALIGN (cif->bytes) : stack);
 
   switch (flags)
     {
@@ -485,7 +494,7 @@ ffi_closure_inner (struct closure_frame *frame, char *stack)
 	    align = 16;
 
 	  /* Issue 434: For thiscall and fastcall, if the paramter passed
-	     as 64-bit integer or struct, all following integer paramters
+	     as 64-bit integer or struct, all following integer parameters
 	     will be passed on stack.  */
 	  if ((cabi == FFI_THISCALL || cabi == FFI_FASTCALL)
 	      && (t == FFI_TYPE_SINT64
@@ -550,13 +559,16 @@ ffi_prep_closure_loc (ffi_closure* closure,
       return FFI_BAD_ABI;
     }
 
+  /* endbr32.  */
+  *(UINT32 *) tramp = 0xfb1e0ff3;
+
   /* movl or pushl immediate.  */
-  tramp[0] = op;
-  *(void **)(tramp + 1) = codeloc;
+  tramp[4] = op;
+  *(void **)(tramp + 5) = codeloc;
 
   /* jmp dest */
-  tramp[5] = 0xe9;
-  *(unsigned *)(tramp + 6) = (unsigned)dest - ((unsigned)codeloc + 10);
+  tramp[9] = 0xe9;
+  *(unsigned *)(tramp + 10) = (unsigned)dest - ((unsigned)codeloc + 14);
 
   closure->cif = cif;
   closure->fun = fun;
@@ -565,6 +577,8 @@ ffi_prep_closure_loc (ffi_closure* closure,
   return FFI_OK;
 }
 
+#ifdef FFI_GO_CLOSURES
+
 void FFI_HIDDEN ffi_go_closure_EAX(void);
 void FFI_HIDDEN ffi_go_closure_ECX(void);
 void FFI_HIDDEN ffi_go_closure_STDCALL(void);
@@ -601,6 +615,8 @@ ffi_prep_go_closure (ffi_go_closure* closure, ffi_cif* cif,
   return FFI_OK;
 }
 
+#endif /* FFI_GO_CLOSURES */
+
 /* ------- Native raw API support -------------------------------- */
 
 #if !FFI_NO_RAW_API
@@ -693,7 +709,7 @@ ffi_raw_call(ffi_cif *cif, void (*fn)(void), void *rvalue, ffi_raw *avalue)
 	}
     }
 
-  bytes = cif->bytes;
+  bytes = STACK_ALIGN (cif->bytes);
   argp = stack =
       (void *)((uintptr_t)alloca(bytes + sizeof(*frame) + rsize + 15) & ~16);
   frame = (struct call_frame *)(stack + bytes);
@@ -751,4 +767,4 @@ ffi_raw_call(ffi_cif *cif, void (*fn)(void), void *rvalue, ffi_raw *avalue)
   ffi_call_i386 (frame, stack);
 }
 #endif /* !FFI_NO_RAW_API */
-#endif /* !__x86_64__ */
+#endif /* __i386__ */

data/ext/ffi_c/libffi/src/x86/ffi64.c

@@ -282,7 +282,7 @@ classify_argument (ffi_type *type, enum x86_64_reg_class classes[],
 
 	    /* The X86_64_SSEUP_CLASS should be always preceded by
 	       X86_64_SSE_CLASS or X86_64_SSEUP_CLASS.  */
-	    if (classes[i] == X86_64_SSEUP_CLASS
+	    if (i > 1 && classes[i] == X86_64_SSEUP_CLASS
 		&& classes[i - 1] != X86_64_SSE_CLASS
 		&& classes[i - 1] != X86_64_SSEUP_CLASS)
 	      {
@@ -293,7 +293,7 @@ classify_argument (ffi_type *type, enum x86_64_reg_class classes[],
 
 	    /*  If X86_64_X87UP_CLASS isn't preceded by X86_64_X87_CLASS,
 		everything should be passed in memory.  */
-	    if (classes[i] == X86_64_X87UP_CLASS
+	    if (i > 1 && classes[i] == X86_64_X87UP_CLASS
 		&& (classes[i - 1] != X86_64_X87_CLASS))
 	      {
 		/* The first one should never be X86_64_X87UP_CLASS.  */
@@ -394,7 +394,7 @@ extern ffi_status
 ffi_prep_cif_machdep_efi64(ffi_cif *cif);
 #endif
 
-ffi_status
+ffi_status FFI_HIDDEN
 ffi_prep_cif_machdep (ffi_cif *cif)
 {
   int gprcount, ssecount, i, avn, ngpr, nsse;
@@ -451,9 +451,11 @@ ffi_prep_cif_machdep (ffi_cif *cif)
     case FFI_TYPE_DOUBLE:
       flags = UNIX64_RET_XMM64;
       break;
+#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE
     case FFI_TYPE_LONGDOUBLE:
       flags = UNIX64_RET_X87;
       break;
+#endif
     case FFI_TYPE_STRUCT:
       n = examine_argument (cif->rtype, classes, 1, &ngpr, &nsse);
       if (n == 0)
@@ -686,6 +688,8 @@ ffi_call (ffi_cif *cif, void (*fn)(void), void *rvalue, void **avalue)
   ffi_call_int (cif, fn, rvalue, avalue, NULL);
 }
 
+#ifdef FFI_GO_CLOSURES
+
 #ifndef __ILP32__
 extern void
 ffi_call_go_efi64(ffi_cif *cif, void (*fn)(void), void *rvalue,
@@ -706,6 +710,7 @@ ffi_call_go (ffi_cif *cif, void (*fn)(void), void *rvalue,
   ffi_call_int (cif, fn, rvalue, avalue, closure);
 }
 
+#endif /* FFI_GO_CLOSURES */
 
 extern void ffi_closure_unix64(void) FFI_HIDDEN;
 extern void ffi_closure_unix64_sse(void) FFI_HIDDEN;
@@ -726,13 +731,15 @@ ffi_prep_closure_loc (ffi_closure* closure,
 		      void *user_data,
 		      void *codeloc)
 {
-  static const unsigned char trampoline[16] = {
-    /* leaq  -0x7(%rip),%r10   # 0x0  */
-    0x4c, 0x8d, 0x15, 0xf9, 0xff, 0xff, 0xff,
-    /* jmpq  *0x3(%rip)        # 0x10 */
-    0xff, 0x25, 0x03, 0x00, 0x00, 0x00,
-    /* nopl  (%rax) */
-    0x0f, 0x1f, 0x00
+  static const unsigned char trampoline[24] = {
+    /* endbr64 */
+    0xf3, 0x0f, 0x1e, 0xfa,
+    /* leaq  -0xb(%rip),%r10   # 0x0  */
+    0x4c, 0x8d, 0x15, 0xf5, 0xff, 0xff, 0xff,
+    /* jmpq  *0x7(%rip)        # 0x18 */
+    0xff, 0x25, 0x07, 0x00, 0x00, 0x00,
+    /* nopl  0(%rax) */
+    0x0f, 0x1f, 0x80, 0x00, 0x00, 0x00, 0x00
   };
   void (*dest)(void);
   char *tramp = closure->tramp;
@@ -750,7 +757,7 @@ ffi_prep_closure_loc (ffi_closure* closure,
     dest = ffi_closure_unix64;
 
   memcpy (tramp, trampoline, sizeof(trampoline));
-  *(UINT64 *)(tramp + 16) = (uintptr_t)dest;
+  *(UINT64 *)(tramp + sizeof (trampoline)) = (uintptr_t)dest;
 
   closure->cif = cif;
   closure->fun = fun;
@@ -852,6 +859,8 @@ ffi_closure_unix64_inner(ffi_cif *cif,
   return flags;
 }
 
+#ifdef FFI_GO_CLOSURES
+
 extern void ffi_go_closure_unix64(void) FFI_HIDDEN;
 extern void ffi_go_closure_unix64_sse(void) FFI_HIDDEN;
 
@@ -881,4 +890,6 @@ ffi_prep_go_closure (ffi_go_closure* closure, ffi_cif* cif,
   return FFI_OK;
 }
 
+#endif /* FFI_GO_CLOSURES */
+
 #endif /* __x86_64__ */

data/ext/ffi_c/libffi/src/x86/ffitarget.h

@@ -136,12 +136,25 @@ typedef enum ffi_abi {
 
 #if defined (X86_64) || defined(X86_WIN64) \
     || (defined (__x86_64__) && defined (X86_DARWIN))
-# define FFI_TRAMPOLINE_SIZE 24
+/* 4 bytes of ENDBR64 + 7 bytes of LEA + 6 bytes of JMP + 7 bytes of NOP
+   + 8 bytes of pointer.  */
+# define FFI_TRAMPOLINE_SIZE 32
 # define FFI_NATIVE_RAW_API 0
 #else
-# define FFI_TRAMPOLINE_SIZE 12
+/* 4 bytes of ENDBR32 + 5 bytes of MOV + 5 bytes of JMP + 2 unused
+   bytes.  */
+# define FFI_TRAMPOLINE_SIZE 16
 # define FFI_NATIVE_RAW_API 1  /* x86 has native raw api support */
 #endif
 
+#if !defined(GENERATE_LIBFFI_MAP) && defined(__ASSEMBLER__) \
+    && defined(__CET__)
+# include <cet.h>
+# define _CET_NOTRACK notrack
+#else
+# define _CET_ENDBR
+# define _CET_NOTRACK
+#endif
+
 #endif