ffi 1.16.3 → 1.17.2

data/ext/ffi_c/libffi/src/aarch64/ffi.c

@@ -63,7 +63,7 @@ struct call_context
 #if FFI_EXEC_TRAMPOLINE_TABLE
 
 #ifdef __MACH__
-#ifdef HAVE_PTRAUTH
+#ifdef HAVE_ARM64E_PTRAUTH
 #include <ptrauth.h>
 #endif
 #include <mach/vm_param.h>
@@ -386,51 +386,64 @@ extend_hfa_type (void *dest, void *src, int h)
   ssize_t f = h - AARCH64_RET_S4;
   void *x0;
 
+#define BTI_J "hint #36"
   asm volatile (
 	"adr	%0, 0f\n"
 "	add	%0, %0, %1\n"
 "	br	%0\n"
-"0:	ldp	s16, s17, [%3]\n"	/* S4 */
+"0:	"BTI_J"\n"			/* S4 */
+"	ldp	s16, s17, [%3]\n"
 "	ldp	s18, s19, [%3, #8]\n"
 "	b	4f\n"
-"	ldp	s16, s17, [%3]\n"	/* S3 */
+"	"BTI_J"\n"			/* S3 */
+"	ldp	s16, s17, [%3]\n"
 "	ldr	s18, [%3, #8]\n"
 "	b	3f\n"
-"	ldp	s16, s17, [%3]\n"	/* S2 */
+"	"BTI_J"\n"			/* S2 */
+"	ldp	s16, s17, [%3]\n"
 "	b	2f\n"
 "	nop\n"
-"	ldr	s16, [%3]\n"		/* S1 */
+"	"BTI_J"\n"			/* S1 */
+"	ldr	s16, [%3]\n"
 "	b	1f\n"
 "	nop\n"
-"	ldp	d16, d17, [%3]\n"	/* D4 */
+"	"BTI_J"\n"			/* D4 */
+"	ldp	d16, d17, [%3]\n"
 "	ldp	d18, d19, [%3, #16]\n"
 "	b	4f\n"
-"	ldp	d16, d17, [%3]\n"	/* D3 */
+"	"BTI_J"\n"			/* D3 */
+"	ldp	d16, d17, [%3]\n"
 "	ldr	d18, [%3, #16]\n"
 "	b	3f\n"
-"	ldp	d16, d17, [%3]\n"	/* D2 */
+"	"BTI_J"\n"			/* D2 */
+"	ldp	d16, d17, [%3]\n"
 "	b	2f\n"
 "	nop\n"
-"	ldr	d16, [%3]\n"		/* D1 */
+"	"BTI_J"\n"			/* D1 */
+"	ldr	d16, [%3]\n"
 "	b	1f\n"
 "	nop\n"
-"	ldp	q16, q17, [%3]\n"	/* Q4 */
+"	"BTI_J"\n"			/* Q4 */
+"	ldp	q16, q17, [%3]\n"
 "	ldp	q18, q19, [%3, #32]\n"
 "	b	4f\n"
-"	ldp	q16, q17, [%3]\n"	/* Q3 */
+"	"BTI_J"\n"			/* Q3 */
+"	ldp	q16, q17, [%3]\n"
 "	ldr	q18, [%3, #32]\n"
 "	b	3f\n"
-"	ldp	q16, q17, [%3]\n"	/* Q2 */
+"	"BTI_J"\n"			/* Q2 */
+"	ldp	q16, q17, [%3]\n"
 "	b	2f\n"
 "	nop\n"
-"	ldr	q16, [%3]\n"		/* Q1 */
+"	"BTI_J"\n"			/* Q1 */
+"	ldr	q16, [%3]\n"
 "	b	1f\n"
 "4:	str	q19, [%2, #48]\n"
 "3:	str	q18, [%2, #32]\n"
 "2:	str	q17, [%2, #16]\n"
 "1:	str	q16, [%2]"
     : "=&r"(x0)
-    : "r"(f * 12), "r"(dest), "r"(src)
+    : "r"(f * 16), "r"(dest), "r"(src)
     : "memory", "v16", "v17", "v18", "v19");
 }
 #endif
@@ -632,7 +645,10 @@ extern void ffi_call_SYSV (struct call_context *context, void *frame,
 			   void *closure) FFI_HIDDEN;
 
 /* Call a function with the provided arguments and capture the return
-   value.  */
+   value.
+   n.b. ffi_call_SYSV will steal the alloca'd `stack` variable here for use
+   _as its own stack_ - so we need to compile this function without ASAN */
+FFI_ASAN_NO_SANITIZE
 static void
 ffi_call_int (ffi_cif *cif, void (*fn)(void), void *orig_rvalue,
 	      void **avalue, void *closure)
@@ -669,7 +685,7 @@ ffi_call_int (ffi_cif *cif, void (*fn)(void), void *orig_rvalue,
   else if (flags & AARCH64_RET_NEED_COPY)
     rsize = 16;
 
-  /* Allocate consectutive stack for everything we'll need.
+  /* Allocate consecutive stack for everything we'll need.
      The frame uses 40 bytes for: lr, fp, rvalue, flags, sp */
   context = alloca (sizeof(struct call_context) + stack_bytes + 40 + rsize);
   stack = context + 1;
@@ -864,7 +880,7 @@ ffi_prep_closure_loc (ffi_closure *closure,
 
 #if FFI_EXEC_TRAMPOLINE_TABLE
 # ifdef __MACH__
-#  ifdef HAVE_PTRAUTH
+#  ifdef HAVE_ARM64E_PTRAUTH
   codeloc = ptrauth_auth_data(codeloc, ptrauth_key_function_pointer, 0);
 #  endif
   void **config = (void **)((uint8_t *)codeloc - PAGE_MAX_SIZE);

data/ext/ffi_c/libffi/src/aarch64/ffitarget.h

@@ -83,8 +83,8 @@ typedef enum ffi_abi
 
 #if defined (__APPLE__)
 #define FFI_EXTRA_CIF_FIELDS unsigned aarch64_nfixedargs
-#elif !defined(_WIN32)
-/* iOS and Windows reserve x18 for the system.  Disable Go closures until
+#elif !defined(_WIN32) && !defined(__ANDROID__)
+/* iOS, Windows and Android reserve x18 for the system.  Disable Go closures until
    a new static chain is chosen.  */
 #define FFI_GO_CLOSURES 1
 #endif

data/ext/ffi_c/libffi/src/aarch64/internal.h

@@ -81,20 +81,66 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.  */
 /* Helpers for writing assembly compatible with arm ptr auth */
 #ifdef LIBFFI_ASM
 
-#ifdef HAVE_PTRAUTH
-#define SIGN_LR pacibsp
-#define SIGN_LR_WITH_REG(x) pacib lr, x
-#define AUTH_LR_AND_RET retab
-#define AUTH_LR_WITH_REG(x) autib lr, x
-#define BRANCH_AND_LINK_TO_REG blraaz
-#define BRANCH_TO_REG braaz
-#else
-#define SIGN_LR
-#define SIGN_LR_WITH_REG(x)
-#define AUTH_LR_AND_RET ret
-#define AUTH_LR_WITH_REG(x)
-#define BRANCH_AND_LINK_TO_REG blr
-#define BRANCH_TO_REG br
-#endif
-
-#endif
+  #if defined(HAVE_ARM64E_PTRAUTH)
+  /* ARM64E ABI For Darwin */
+  #define SIGN_LR pacibsp
+  #define SIGN_LR_WITH_REG(x) pacib lr, x
+  #define AUTH_LR_AND_RET retab
+  #define AUTH_LR_WITH_REG(x) autib lr, x
+  #define BRANCH_AND_LINK_TO_REG blraaz
+  #define SIGN_LR_LINUX_ONLY
+  #define BRANCH_TO_REG braaz
+  #define PAC_CFI_WINDOW_SAVE
+  #define GNU_PROPERTY_AARCH64_POINTER_AUTH 0
+  /* Linux PAC Support */
+  #elif defined(__ARM_FEATURE_PAC_DEFAULT)
+    #define GNU_PROPERTY_AARCH64_POINTER_AUTH (1 << 1)
+    #define PAC_CFI_WINDOW_SAVE cfi_window_save
+    #define TMP_REG x9
+    #define BRANCH_TO_REG br
+    #define BRANCH_AND_LINK_TO_REG blr
+	#define SIGN_LR_LINUX_ONLY SIGN_LR
+    /* Which key to sign with? */
+    #if (__ARM_FEATURE_PAC_DEFAULT & 1) == 1
+      /* Signed with A-key */
+      #define SIGN_LR            hint #25  /* paciasp */
+      #define AUTH_LR            hint #29  /* autiasp */
+    #else
+      /* Signed with B-key */
+      #define SIGN_LR            hint #27  /* pacibsp */
+      #define AUTH_LR            hint #31  /* autibsp */
+    #endif /* __ARM_FEATURE_PAC_DEFAULT */
+    #define AUTH_LR_WITH_REG(x) _auth_lr_with_reg x
+.macro _auth_lr_with_reg modifier
+    mov TMP_REG, sp
+    mov sp, \modifier
+    AUTH_LR
+    mov sp, TMP_REG
+.endm
+  #define SIGN_LR_WITH_REG(x) _sign_lr_with_reg x
+.macro _sign_lr_with_reg modifier
+    mov TMP_REG, sp
+    mov sp, \modifier
+    SIGN_LR
+    mov sp, TMP_REG
+.endm
+  #define AUTH_LR_AND_RET _auth_lr_and_ret modifier
+.macro _auth_lr_and_ret modifier
+    AUTH_LR
+    ret
+.endm
+  #undef TMP_REG
+
+  /* No Pointer Auth */
+  #else
+    #define SIGN_LR
+    #define SIGN_LR_WITH_REG(x)
+    #define AUTH_LR_AND_RET ret
+    #define AUTH_LR_WITH_REG(x)
+    #define BRANCH_AND_LINK_TO_REG blr
+    #define SIGN_LR_LINUX_ONLY
+    #define BRANCH_TO_REG br
+    #define PAC_CFI_WINDOW_SAVE
+    #define GNU_PROPERTY_AARCH64_POINTER_AUTH 0
+  #endif /* HAVE_ARM64E_PTRAUTH */
+#endif /* LIBFFI_ASM */

data/ext/ffi_c/libffi/src/aarch64/sysv.S

@@ -64,6 +64,13 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.  */
 #define PTR_SIZE	8
 #endif
 
+#define BTI_C hint #34
+#define BTI_J hint #36
+/*
+ * The ELF Notes section needs to indicate if BTI is supported, as the first ELF loaded that doesn't
+ * declare this support disables it for memory region containing the loaded library.
+ */
+# define GNU_PROPERTY_AARCH64_BTI (1 << 0)         /* Has Branch Target Identification */
 	.text
 	.align 4
 
@@ -82,29 +89,30 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.  */
    x5 closure
 */
 
-	cfi_startproc
 CNAME(ffi_call_SYSV):
-	/* Sign the lr with x1 since that is where it will be stored */
+	cfi_startproc
+	BTI_C
+	PAC_CFI_WINDOW_SAVE
+	/* Sign the lr with x1 since that is the CFA which is the modifer used in auth instructions */
 	SIGN_LR_WITH_REG(x1)
 
-	/* Use a stack frame allocated by our caller.  */
-#if defined(HAVE_PTRAUTH) && defined(__APPLE__)
+#if defined(HAVE_ARM64E_PTRAUTH) && defined(__APPLE__)
 	/* darwin's libunwind assumes that the cfa is the sp and that's the data
 	 * used to sign the lr.  In order to allow unwinding through this
 	 * function it is necessary to point the cfa at the signing register.
 	 */
 	cfi_def_cfa(x1, 0);
-#else
-	cfi_def_cfa(x1, 40);
 #endif
+	/* Use a stack frame allocated by our caller.  */
 	stp	x29, x30, [x1]
+	cfi_def_cfa_register(x1)
+	cfi_rel_offset (x29, 0)
+	cfi_rel_offset (x30, 8)
 	mov	x9, sp
 	str	x9, [x1, #32]
 	mov	x29, x1
-	mov	sp, x0
 	cfi_def_cfa_register(x29)
-	cfi_rel_offset (x29, 0)
-	cfi_rel_offset (x30, 8)
+	mov	sp, x0
 
 	mov	x9, x2			/* save fn */
 	mov	x8, x3			/* install structure return */
@@ -144,78 +152,142 @@ CNAME(ffi_call_SYSV):
 	/* Save the return value as directed.  */
 	adr	x5, 0f
 	and	w4, w4, #AARCH64_RET_MASK
-	add	x5, x5, x4, lsl #3
+	add	x5, x5, x4, lsl #4
 	br	x5
 
-	/* Note that each table entry is 2 insns, and thus 8 bytes.
+	/* Note that each table entry is 4 insns, and thus 16 bytes.
 	   For integer data, note that we're storing into ffi_arg
 	   and therefore we want to extend to 64 bits; these types
 	   have two consecutive entries allocated for them.  */
 	.align	4
-0:	b 99f				/* VOID */
+0:	BTI_J				/* VOID */
+	b 99f
+	nop 
 	nop
-1:	str	x0, [x3]		/* INT64 */
+1:	BTI_J				/* INT64 */
+	str	x0, [x3]
 	b 99f
-2:	stp	x0, x1, [x3]		/* INT128 */
+	nop
+2:	BTI_J				/* INT128 */
+	stp	x0, x1, [x3]
 	b 99f
+	nop
 3:	brk	#1000			/* UNUSED */
 	b 99f
+	nop
+	nop
 4:	brk	#1000			/* UNUSED */
 	b 99f
+	nop
+	nop
 5:	brk	#1000			/* UNUSED */
 	b 99f
+	nop
+	nop
 6:	brk	#1000			/* UNUSED */
 	b 99f
+	nop
+	nop
 7:	brk	#1000			/* UNUSED */
 	b 99f
-8:	st4	{ v0.s, v1.s, v2.s, v3.s }[0], [x3]	/* S4 */
+	nop
+	nop
+8:	BTI_J				/* S4 */
+	st4	{ v0.s, v1.s, v2.s, v3.s }[0], [x3]
 	b 99f
-9:	st3	{ v0.s, v1.s, v2.s }[0], [x3]	/* S3 */
+	nop
+9:	BTI_J				/* S3 */
+	st3	{ v0.s, v1.s, v2.s }[0], [x3]
 	b 99f
-10:	stp	s0, s1, [x3]		/* S2 */
+	nop
+10:	BTI_J				/* S2 */
+	stp	s0, s1, [x3]
 	b 99f
-11:	str	s0, [x3]		/* S1 */
+	nop
+11:	BTI_J
+	str	s0, [x3]		/* S1 */
 	b 99f
-12:	st4	{ v0.d, v1.d, v2.d, v3.d }[0], [x3]	/* D4 */
+	nop
+12:	BTI_J				/* D4 */
+	st4	{ v0.d, v1.d, v2.d, v3.d }[0], [x3]
 	b 99f
-13:	st3	{ v0.d, v1.d, v2.d }[0], [x3]	/* D3 */
+	nop
+13:	BTI_J				/* D3 */
+	st3	{ v0.d, v1.d, v2.d }[0], [x3]
 	b 99f
-14:	stp	d0, d1, [x3]		/* D2 */
+	nop
+14:	BTI_J				/* D2 */
+	stp	d0, d1, [x3]
 	b 99f
-15:	str	d0, [x3]		/* D1 */
+	nop
+15:	BTI_J				/* D1 */
+	str	d0, [x3]
 	b 99f
-16:	str	q3, [x3, #48]		/* Q4 */
 	nop
-17:	str	q2, [x3, #32]		/* Q3 */
+16:	BTI_J				/* Q4 */
+	str	q3, [x3, #48]
+	nop
+	nop
+17:	BTI_J				/* Q3 */
+	str	q2, [x3, #32]
 	nop
-18:	stp	q0, q1, [x3]		/* Q2 */
+	nop
+18:	BTI_J				/* Q2 */
+	stp	q0, q1, [x3]
 	b 99f
-19:	str	q0, [x3]		/* Q1 */
+	nop
+19:	BTI_J				/* Q1 */
+	str	q0, [x3]
 	b 99f
-20:	uxtb	w0, w0			/* UINT8 */
+	nop
+20:	BTI_J				/* UINT8 */
+	uxtb	w0, w0
 	str	x0, [x3]
+	nop
 21:	b 99f				/* reserved */
 	nop
-22:	uxth	w0, w0			/* UINT16 */
+	nop
+	nop
+22:	BTI_J				/* UINT16 */
+	uxth	w0, w0
 	str	x0, [x3]
+	nop
 23:	b 99f				/* reserved */
 	nop
-24:	mov	w0, w0			/* UINT32 */
+	nop
+	nop
+24:	BTI_J				/* UINT32 */
+	mov	w0, w0
 	str	x0, [x3]
+	nop
 25:	b 99f				/* reserved */
 	nop
-26:	sxtb	x0, w0			/* SINT8 */
+	nop
+	nop
+26:	BTI_J				/* SINT8 */
+	sxtb	x0, w0
 	str	x0, [x3]
+	nop
 27:	b 99f				/* reserved */
 	nop
-28:	sxth	x0, w0			/* SINT16 */
+	nop
+	nop
+28:	BTI_J				/* SINT16 */
+	sxth	x0, w0
 	str	x0, [x3]
+	nop
 29:	b 99f				/* reserved */
 	nop
-30:	sxtw	x0, w0			/* SINT32 */
+	nop
+	nop
+30:	BTI_J				/* SINT32 */
+	sxtw	x0, w0
 	str	x0, [x3]
+	nop
 31:	b 99f				/* reserved */
 	nop
+	nop
+	nop
 
 	/* Return now that result has been populated. */
 99:
@@ -252,7 +324,9 @@ CNAME(ffi_call_SYSV):
 	.align 4
 CNAME(ffi_closure_SYSV_V):
 	cfi_startproc
+	BTI_C
 	SIGN_LR
+	PAC_CFI_WINDOW_SAVE
 	stp     x29, x30, [sp, #-ffi_closure_SYSV_FS]!
 	cfi_adjust_cfa_offset (ffi_closure_SYSV_FS)
 	cfi_rel_offset (x29, 0)
@@ -274,9 +348,11 @@ CNAME(ffi_closure_SYSV_V):
 #endif
 
 	.align	4
-	cfi_startproc
 CNAME(ffi_closure_SYSV):
+	cfi_startproc
+	BTI_C
 	SIGN_LR
+	PAC_CFI_WINDOW_SAVE
 	stp     x29, x30, [sp, #-ffi_closure_SYSV_FS]!
 	cfi_adjust_cfa_offset (ffi_closure_SYSV_FS)
 	cfi_rel_offset (x29, 0)
@@ -305,73 +381,135 @@ L(do_closure):
 	/* Load the return value as directed.  */
 	adr	x1, 0f
 	and	w0, w0, #AARCH64_RET_MASK
-	add	x1, x1, x0, lsl #3
+	add	x1, x1, x0, lsl #4
 	add	x3, sp, #16+CALL_CONTEXT_SIZE
 	br	x1
 
-	/* Note that each table entry is 2 insns, and thus 8 bytes.  */
+	/* Note that each table entry is 4 insns, and thus 16 bytes.  */
 	.align	4
-0:	b	99f			/* VOID */
+0:	BTI_J				/* VOID */
+	b	99f
+	nop
 	nop
-1:	ldr	x0, [x3]		/* INT64 */
+1:	BTI_J				/* INT64 */
+	ldr	x0, [x3]
 	b	99f
-2:	ldp	x0, x1, [x3]		/* INT128 */
+	nop
+2:	BTI_J				/* INT128 */
+	ldp	x0, x1, [x3]
 	b	99f
+	nop
 3:	brk	#1000			/* UNUSED */
 	nop
+	nop
+	nop
 4:	brk	#1000			/* UNUSED */
 	nop
+	nop
+	nop
 5:	brk	#1000			/* UNUSED */
 	nop
+	nop
+	nop
 6:	brk	#1000			/* UNUSED */
 	nop
+	nop
+	nop
 7:	brk	#1000			/* UNUSED */
 	nop
-8:	ldr	s3, [x3, #12]		/* S4 */
 	nop
-9:	ldr	s2, [x3, #8]		/* S3 */
 	nop
-10:	ldp	s0, s1, [x3]		/* S2 */
+8:	BTI_J				/* S4 */
+	ldr	s3, [x3, #12]
+	nop
+	nop
+9:	BTI_J				/* S3 */
+	ldr	s2, [x3, #8]
+	nop
+	nop
+10:	BTI_J				/* S2 */
+	ldp	s0, s1, [x3]
 	b	99f
-11:	ldr	s0, [x3]		/* S1 */
+	nop
+11:	BTI_J				/* S1 */
+	ldr	s0, [x3]
 	b	99f
-12:	ldr	d3, [x3, #24]		/* D4 */
 	nop
-13:	ldr	d2, [x3, #16]		/* D3 */
+12:	BTI_J				/* D4 */
+	ldr	d3, [x3, #24]
+	nop
+	nop
+13:	BTI_J				/* D3 */
+	ldr	d2, [x3, #16]
 	nop
-14:	ldp	d0, d1, [x3]		/* D2 */
+	nop
+14:	BTI_J				/* D2 */
+	ldp	d0, d1, [x3]
 	b	99f
-15:	ldr	d0, [x3]		/* D1 */
+	nop
+15:	BTI_J				/* D1 */
+	ldr	d0, [x3]
 	b	99f
-16:	ldr	q3, [x3, #48]		/* Q4 */
 	nop
-17:	ldr	q2, [x3, #32]		/* Q3 */
+16:	BTI_J				/* Q4 */
+	ldr	q3, [x3, #48]
+	nop
+	nop
+17:	BTI_J				/* Q3 */
+	ldr	q2, [x3, #32]
 	nop
-18:	ldp	q0, q1, [x3]		/* Q2 */
+	nop
+18:	BTI_J				/* Q2 */
+	ldp	q0, q1, [x3]
 	b	99f
-19:	ldr	q0, [x3]		/* Q1 */
+	nop
+19:	BTI_J				/* Q1 */
+	ldr	q0, [x3]
 	b	99f
-20:	ldrb	w0, [x3, #BE(7)]	/* UINT8 */
+	nop
+20:	BTI_J				/* UINT8 */
+	ldrb	w0, [x3, #BE(7)]
 	b	99f
+	nop
 21:	brk	#1000			/* reserved */
 	nop
-22:	ldrh	w0, [x3, #BE(6)]	/* UINT16 */
+	nop
+	nop
+22:	BTI_J				/* UINT16 */
+	ldrh	w0, [x3, #BE(6)]
 	b	99f
+	nop
 23:	brk	#1000			/* reserved */
 	nop
-24:	ldr	w0, [x3, #BE(4)]	/* UINT32 */
+	nop
+	nop
+24:	BTI_J				/* UINT32 */
+	ldr	w0, [x3, #BE(4)]
 	b	99f
+	nop
 25:	brk	#1000			/* reserved */
 	nop
-26:	ldrsb	x0, [x3, #BE(7)]	/* SINT8 */
+	nop
+	nop
+26:	BTI_J				/* SINT8 */
+	ldrsb	x0, [x3, #BE(7)]
 	b	99f
+	nop
 27:	brk	#1000			/* reserved */
 	nop
-28:	ldrsh	x0, [x3, #BE(6)]	/* SINT16 */
+	nop
+	nop
+28:	BTI_J				/* SINT16 */
+	ldrsh	x0, [x3, #BE(6)]
 	b	99f
+	nop
 29:	brk	#1000			/* reserved */
 	nop
-30:	ldrsw	x0, [x3, #BE(4)]	/* SINT32 */
+	nop
+	nop
+30:	BTI_J				/* SINT32 */
+	ldrsw	x0, [x3, #BE(4)]
+	nop
 	nop
 31:					/* reserved */
 99:	ldp     x29, x30, [sp], #ffi_closure_SYSV_FS
@@ -391,6 +529,7 @@ L(do_closure):
 #if defined(FFI_EXEC_STATIC_TRAMP)
 	.align 4
 CNAME(ffi_closure_SYSV_V_alt):
+	BTI_C
 	/* See the comments above trampoline_code_table. */
 	ldr	x17, [sp, #8]			/* Load closure in x17 */
 	add	sp, sp, #16			/* Restore the stack */
@@ -405,6 +544,7 @@ CNAME(ffi_closure_SYSV_V_alt):
 
 	.align 4
 CNAME(ffi_closure_SYSV_alt):
+	BTI_C
 	/* See the comments above trampoline_code_table. */
 	ldr	x17, [sp, #8]			/* Load closure in x17 */
 	add	sp, sp, #16			/* Restore the stack */
@@ -485,6 +625,7 @@ CNAME(ffi_closure_trampoline_table_page):
 	.align 4
 CNAME(ffi_go_closure_SYSV_V):
 	cfi_startproc
+	BTI_C
 	stp     x29, x30, [sp, #-ffi_closure_SYSV_FS]!
 	cfi_adjust_cfa_offset (ffi_closure_SYSV_FS)
 	cfi_rel_offset (x29, 0)
@@ -506,8 +647,11 @@ CNAME(ffi_go_closure_SYSV_V):
 #endif
 
 	.align	4
-	cfi_startproc
 CNAME(ffi_go_closure_SYSV):
+	cfi_startproc
+	BTI_C
+	SIGN_LR_LINUX_ONLY
+	PAC_CFI_WINDOW_SAVE
 	stp     x29, x30, [sp, #-ffi_closure_SYSV_FS]!
 	cfi_adjust_cfa_offset (ffi_closure_SYSV_FS)
 	cfi_rel_offset (x29, 0)
@@ -539,5 +683,17 @@ CNAME(ffi_go_closure_SYSV):
 
 #if defined __ELF__ && defined __linux__
 	.section .note.GNU-stack,"",%progbits
+
+	.pushsection .note.gnu.property, "a";
+	.balign 8;
+	.long 4;
+	.long 0x10;
+	.long 0x5;
+	.asciz "GNU";
+	.long 0xc0000000; /* GNU_PROPERTY_AARCH64_FEATURE_1_AND */
+	.long 4;
+	.long GNU_PROPERTY_AARCH64_BTI | GNU_PROPERTY_AARCH64_POINTER_AUTH;
+	.long 0;
+	.popsection;
 #endif
 

data/ext/ffi_c/libffi/src/arc/arcompact.S

@@ -39,14 +39,14 @@
 #define LARG ldl
 #define SARG stl
 #define ADDPTR addl
-#define MOVPTR movl_s
+#define MOVPTR movl
 #else
 #define PTRS 4
 #define FLTS 4
 #define LARG ld
 #define SARG st
 #define ADDPTR add
-#define MOVPTR mov_s
+#define MOVPTR mov
 #endif
 
 #define FRAME_LEN (8 * PTRS + 16)