ffi 1.15.4 → 1.16.3

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

@@ -40,6 +40,12 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.  */
 #endif
 #endif
 
+#ifdef __APPLE__
+# define L(X)     CONCAT1(L, X)
+#else
+# define L(X)     CONCAT1(.L, X)
+#endif
+
 #ifdef __AARCH64EB__
 # define BE(X)	X
 #else
@@ -56,14 +62,6 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.  */
 #define PTR_SIZE	4
 #else
 #define PTR_SIZE	8
-#endif
-
-#if FFI_EXEC_TRAMPOLINE_TABLE && defined(__MACH__) && defined(HAVE_PTRAUTH)
-# define BR(r)  braaz r
-# define BLR(r) blraaz r
-#else
-# define BR(r)  br r
-# define BLR(r) blr r
 #endif
 
 	.text
@@ -86,9 +84,22 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.  */
 
 	cfi_startproc
 CNAME(ffi_call_SYSV):
+	/* Sign the lr with x1 since that is where it will be stored */
+	SIGN_LR_WITH_REG(x1)
+
 	/* Use a stack frame allocated by our caller.  */
-	cfi_def_cfa(x1, 32);
+#if defined(HAVE_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
 	stp	x29, x30, [x1]
+	mov	x9, sp
+	str	x9, [x1, #32]
 	mov	x29, x1
 	mov	sp, x0
 	cfi_def_cfa_register(x29)
@@ -119,13 +130,15 @@ CNAME(ffi_call_SYSV):
 	/* Deallocate the context, leaving the stacked arguments.  */
 	add	sp, sp, #CALL_CONTEXT_SIZE
 
-	BLR(x9)				/* call fn */
+	BRANCH_AND_LINK_TO_REG     x9			/* call fn */
 
 	ldp	x3, x4, [x29, #16]	/* reload rvalue and flags */
 
 	/* Partially deconstruct the stack frame.  */
-	mov     sp, x29
+	ldr	x9, [x29, #32]
+	mov	sp, x9
 	cfi_def_cfa_register (sp)
+	mov	x2, x29			/* Preserve for auth */
 	ldp     x29, x30, [x29]
 
 	/* Save the return value as directed.  */
@@ -139,71 +152,76 @@ CNAME(ffi_call_SYSV):
 	   and therefore we want to extend to 64 bits; these types
 	   have two consecutive entries allocated for them.  */
 	.align	4
-0:	ret				/* VOID */
+0:	b 99f				/* VOID */
 	nop
 1:	str	x0, [x3]		/* INT64 */
-	ret
+	b 99f
 2:	stp	x0, x1, [x3]		/* INT128 */
-	ret
+	b 99f
 3:	brk	#1000			/* UNUSED */
-	ret
+	b 99f
 4:	brk	#1000			/* UNUSED */
-	ret
+	b 99f
 5:	brk	#1000			/* UNUSED */
-	ret
+	b 99f
 6:	brk	#1000			/* UNUSED */
-	ret
+	b 99f
 7:	brk	#1000			/* UNUSED */
-	ret
+	b 99f
 8:	st4	{ v0.s, v1.s, v2.s, v3.s }[0], [x3]	/* S4 */
-	ret
+	b 99f
 9:	st3	{ v0.s, v1.s, v2.s }[0], [x3]	/* S3 */
-	ret
+	b 99f
 10:	stp	s0, s1, [x3]		/* S2 */
-	ret
+	b 99f
 11:	str	s0, [x3]		/* S1 */
-	ret
+	b 99f
 12:	st4	{ v0.d, v1.d, v2.d, v3.d }[0], [x3]	/* D4 */
-	ret
+	b 99f
 13:	st3	{ v0.d, v1.d, v2.d }[0], [x3]	/* D3 */
-	ret
+	b 99f
 14:	stp	d0, d1, [x3]		/* D2 */
-	ret
+	b 99f
 15:	str	d0, [x3]		/* D1 */
-	ret
+	b 99f
 16:	str	q3, [x3, #48]		/* Q4 */
 	nop
 17:	str	q2, [x3, #32]		/* Q3 */
 	nop
 18:	stp	q0, q1, [x3]		/* Q2 */
-	ret
+	b 99f
 19:	str	q0, [x3]		/* Q1 */
-	ret
+	b 99f
 20:	uxtb	w0, w0			/* UINT8 */
 	str	x0, [x3]
-21:	ret				/* reserved */
+21:	b 99f				/* reserved */
 	nop
 22:	uxth	w0, w0			/* UINT16 */
 	str	x0, [x3]
-23:	ret				/* reserved */
+23:	b 99f				/* reserved */
 	nop
 24:	mov	w0, w0			/* UINT32 */
 	str	x0, [x3]
-25:	ret				/* reserved */
+25:	b 99f				/* reserved */
 	nop
 26:	sxtb	x0, w0			/* SINT8 */
 	str	x0, [x3]
-27:	ret				/* reserved */
+27:	b 99f				/* reserved */
 	nop
 28:	sxth	x0, w0			/* SINT16 */
 	str	x0, [x3]
-29:	ret				/* reserved */
+29:	b 99f				/* reserved */
 	nop
 30:	sxtw	x0, w0			/* SINT32 */
 	str	x0, [x3]
-31:	ret				/* reserved */
+31:	b 99f				/* reserved */
 	nop
 
+	/* Return now that result has been populated. */
+99:
+	AUTH_LR_WITH_REG(x2)
+	ret
+
 	cfi_endproc
 
 	.globl	CNAME(ffi_call_SYSV)
@@ -213,6 +231,8 @@ CNAME(ffi_call_SYSV):
 	.size CNAME(ffi_call_SYSV), .-CNAME(ffi_call_SYSV)
 #endif
 
+#if FFI_CLOSURES
+
 /* ffi_closure_SYSV
 
    Closure invocation glue. This is the low level code invoked directly by
@@ -232,6 +252,7 @@ CNAME(ffi_call_SYSV):
 	.align 4
 CNAME(ffi_closure_SYSV_V):
 	cfi_startproc
+	SIGN_LR
 	stp     x29, x30, [sp, #-ffi_closure_SYSV_FS]!
 	cfi_adjust_cfa_offset (ffi_closure_SYSV_FS)
 	cfi_rel_offset (x29, 0)
@@ -255,6 +276,7 @@ CNAME(ffi_closure_SYSV_V):
 	.align	4
 	cfi_startproc
 CNAME(ffi_closure_SYSV):
+	SIGN_LR
 	stp     x29, x30, [sp, #-ffi_closure_SYSV_FS]!
 	cfi_adjust_cfa_offset (ffi_closure_SYSV_FS)
 	cfi_rel_offset (x29, 0)
@@ -271,7 +293,9 @@ CNAME(ffi_closure_SYSV):
 	/* Load ffi_closure_inner arguments.  */
 	ldp	PTR_REG(0), PTR_REG(1), [x17, #FFI_TRAMPOLINE_CLOSURE_OFFSET]	/* load cif, fn */
 	ldr	PTR_REG(2), [x17, #FFI_TRAMPOLINE_CLOSURE_OFFSET+PTR_SIZE*2]	/* load user_data */
-.Ldo_closure:
+#ifdef FFI_GO_CLOSURES
+L(do_closure):
+#endif
 	add	x3, sp, #16				/* load context */
 	add	x4, sp, #ffi_closure_SYSV_FS		/* load stack */
 	add	x5, sp, #16+CALL_CONTEXT_SIZE		/* load rvalue */
@@ -279,9 +303,6 @@ CNAME(ffi_closure_SYSV):
 	bl      CNAME(ffi_closure_SYSV_inner)
 
 	/* Load the return value as directed.  */
-#if FFI_EXEC_TRAMPOLINE_TABLE && defined(__MACH__) && defined(HAVE_PTRAUTH)
-	autiza	x1
-#endif
 	adr	x1, 0f
 	and	w0, w0, #AARCH64_RET_MASK
 	add	x1, x1, x0, lsl #3
@@ -357,7 +378,7 @@ CNAME(ffi_closure_SYSV):
 	cfi_adjust_cfa_offset (-ffi_closure_SYSV_FS)
 	cfi_restore (x29)
 	cfi_restore (x30)
-	ret
+	AUTH_LR_AND_RET
 	cfi_endproc
 
 	.globl	CNAME(ffi_closure_SYSV)
@@ -367,6 +388,76 @@ CNAME(ffi_closure_SYSV):
 	.size	CNAME(ffi_closure_SYSV), . - CNAME(ffi_closure_SYSV)
 #endif
 
+#if defined(FFI_EXEC_STATIC_TRAMP)
+	.align 4
+CNAME(ffi_closure_SYSV_V_alt):
+	/* See the comments above trampoline_code_table. */
+	ldr	x17, [sp, #8]			/* Load closure in x17 */
+	add	sp, sp, #16			/* Restore the stack */
+	b	CNAME(ffi_closure_SYSV_V)
+
+	.globl	CNAME(ffi_closure_SYSV_V_alt)
+	FFI_HIDDEN(CNAME(ffi_closure_SYSV_V_alt))
+#ifdef __ELF__
+	.type	CNAME(ffi_closure_SYSV_V_alt), #function
+	.size	CNAME(ffi_closure_SYSV_V_alt), . - CNAME(ffi_closure_SYSV_V_alt)
+#endif
+
+	.align 4
+CNAME(ffi_closure_SYSV_alt):
+	/* See the comments above trampoline_code_table. */
+	ldr	x17, [sp, #8]			/* Load closure in x17 */
+	add	sp, sp, #16			/* Restore the stack */
+	b	CNAME(ffi_closure_SYSV)
+
+	.globl	CNAME(ffi_closure_SYSV_alt)
+	FFI_HIDDEN(CNAME(ffi_closure_SYSV_alt))
+#ifdef __ELF__
+	.type	CNAME(ffi_closure_SYSV_alt), #function
+	.size	CNAME(ffi_closure_SYSV_alt), . - CNAME(ffi_closure_SYSV_alt)
+#endif
+
+/*
+ * Below is the definition of the trampoline code table. Each element in
+ * the code table is a trampoline.
+ */
+/*
+ * The trampoline uses register x17. It saves the original value of x17 on
+ * the stack.
+ *
+ * The trampoline has two parameters - target code to jump to and data for
+ * the target code. The trampoline extracts the parameters from its parameter
+ * block (see tramp_table_map()). The trampoline saves the data address on
+ * the stack. Finally, it jumps to the target code.
+ *
+ * The target code can choose to:
+ *
+ * - restore the value of x17
+ * - load the data address in a register
+ * - restore the stack pointer to what it was when the trampoline was invoked.
+ */
+	.align	AARCH64_TRAMP_MAP_SHIFT
+CNAME(trampoline_code_table):
+	.rept	AARCH64_TRAMP_MAP_SIZE / AARCH64_TRAMP_SIZE
+	sub	sp, sp, #16		/* Make space on the stack */
+	str	x17, [sp]		/* Save x17 on stack */
+	adr	x17, #16376		/* Get data address */
+	ldr	x17, [x17]		/* Copy data into x17 */
+	str	x17, [sp, #8]		/* Save data on stack */
+	adr	x17, #16372		/* Get code address */
+	ldr	x17, [x17]		/* Load code address into x17 */
+	br	x17			/* Jump to code */
+	.endr
+
+	.globl CNAME(trampoline_code_table)
+	FFI_HIDDEN(CNAME(trampoline_code_table))
+#ifdef __ELF__
+	.type	CNAME(trampoline_code_table), #function
+	.size	CNAME(trampoline_code_table), . - CNAME(trampoline_code_table)
+#endif
+	.align	AARCH64_TRAMP_MAP_SHIFT
+#endif /* FFI_EXEC_STATIC_TRAMP */
+
 #if FFI_EXEC_TRAMPOLINE_TABLE
 
 #ifdef __MACH__
@@ -376,7 +467,7 @@ CNAME(ffi_closure_trampoline_table_page):
     .rept PAGE_MAX_SIZE / FFI_TRAMPOLINE_SIZE
     adr x16, -PAGE_MAX_SIZE
     ldp x17, x16, [x16]
-    BR(x16)
+    br x16
 	nop		/* each entry in the trampoline config page is 2*sizeof(void*) so the trampoline itself cannot be smaller than 16 bytes */
     .endr
 
@@ -433,7 +524,7 @@ CNAME(ffi_go_closure_SYSV):
 	/* Load ffi_closure_inner arguments.  */
 	ldp	PTR_REG(0), PTR_REG(1), [x18, #PTR_SIZE]/* load cif, fn */
 	mov	x2, x18					/* load user_data */
-	b	.Ldo_closure
+	b	L(do_closure)
 	cfi_endproc
 
 	.globl	CNAME(ffi_go_closure_SYSV)
@@ -443,6 +534,7 @@ CNAME(ffi_go_closure_SYSV):
 	.size	CNAME(ffi_go_closure_SYSV), . - CNAME(ffi_go_closure_SYSV)
 #endif
 #endif /* FFI_GO_CLOSURES */
+#endif /* FFI_CLOSURES */
 #endif /* __arm64__ */
 
 #if defined __ELF__ && defined __linux__

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

@@ -33,103 +33,178 @@
 #define ENTRY(x) .globl CNAME(x)` .type CNAME(x),%function` CNAME(x):
 #endif
 
+#if __SIZEOF_POINTER__ == 8
+#define PTRS 8
+#define FLTS 8
+#define LARG ldl
+#define SARG stl
+#define ADDPTR addl
+#define MOVPTR movl_s
+#else
+#define PTRS 4
+#define FLTS 4
+#define LARG ld
+#define SARG st
+#define ADDPTR add
+#define MOVPTR mov_s
+#endif
+
+#define FRAME_LEN (8 * PTRS + 16)
+
 .text
 
-        /* R0:   ffi_prep_args */
-        /* R1:   &ecif */
-        /* R2:   cif->bytes */
-        /* R3:   fig->flags */
-        /* R4:   ecif.rvalue */
-        /* R5:   fn */
-ENTRY(ffi_call_ARCompact)
+ENTRY(ffi_call_asm)
+        .cfi_startproc
+
         /* Save registers.  */
-        st.a       fp, [sp, -4]        /* fp + 20, fp */
-        push_s     blink               /* fp + 16, blink */
-        st.a       r4, [sp, -4]        /* fp + 12, ecif.rvalue */
-        push_s     r3                  /* fp +  8, fig->flags */
-        st.a       r5, [sp, -4]        /* fp +  4, fn */
-        push_s     r2                  /* fp +  0, cif->bytes */
-        mov        fp, sp
-
-        /* Make room for all of the new args.  */
-        sub        sp, sp, r2
-
-        /* Place all of the ffi_prep_args in position.  */
-        /* ffi_prep_args(char *stack, extended_cif *ecif) */
-        /* R1 already set.  */
-
-        /* And call.  */
-        jl_s.d     [r0]
-        mov_s      r0, sp
-
-        ld.ab      r12, [fp, 4]        /* cif->bytes */
-        ld.ab      r11, [fp, 4]        /* fn */
-
-        /* Move first 8 parameters in registers...  */
-        ld_s       r0, [sp]
-        ld_s       r1, [sp, 4]
-        ld_s       r2, [sp, 8]
-        ld_s       r3, [sp, 12]
-        ld         r4, [sp, 16]
-        ld         r5, [sp, 20]
-        ld         r6, [sp, 24]
-        ld         r7, [sp, 28]
-
-        /* ...and adjust the stack.  */
-        min        r12, r12, 32
+        .cfi_def_cfa r1, FRAME_LEN
+        SARG    fp, [r1, FRAME_LEN - 2*PTRS]
+        .cfi_offset fp, -2*PTRS
+        SARG    blink, [r1, FRAME_LEN - 1*PTRS]
+        .cfi_offset blink, -1*PTRS
+        ADDPTR  fp, r1, FRAME_LEN
+        MOVPTR  sp, r0
+        .cfi_def_cfa fp, 0
+
+        /* Load arguments.  */
+        MOVPTR  r11, r2      /* fn */
+        MOVPTR  r12, r3      /* closure */
+
+        /* Save arguments.  */
+        LARG    r0, [fp, -FRAME_LEN+0*PTRS]
+        LARG    r1, [fp, -FRAME_LEN+1*PTRS]
+        LARG    r2, [fp, -FRAME_LEN+2*PTRS]
+        LARG    r3, [fp, -FRAME_LEN+3*PTRS]
+        LARG    r4, [fp, -FRAME_LEN+4*PTRS]
+        LARG    r5, [fp, -FRAME_LEN+5*PTRS]
+        LARG    r6, [fp, -FRAME_LEN+6*PTRS]
+        LARG    r7, [fp, -FRAME_LEN+7*PTRS]
 
         /* Call the function.  */
-        jl.d       [r11]
-        add        sp, sp, r12 
-
-        mov        sp, fp        
-        pop_s      r3        /* fig->flags, return type */
-        pop_s      r2        /* ecif.rvalue, pointer for return value */
-
-        /* If the return value pointer is NULL, assume no return value.  */
-        breq.d     r2, 0, epilogue
-        pop_s      blink
-
-        /* Return INT.  */
-        brne       r3, FFI_TYPE_INT, return_double
-        b.d        epilogue
-        st_s       r0, [r2]        
-
-return_double:
-        brne       r3, FFI_TYPE_DOUBLE, epilogue
-        st_s       r0, [r2]        
-        st_s       r1, [r2,4]
-
-epilogue:
-        j_s.d      [blink]
-        ld.ab      fp, [sp, 4]
-
-ENTRY(ffi_closure_ARCompact)
-        st.a       r0, [sp, -32]
-        st_s       r1, [sp, 4]
-        st_s       r2, [sp, 8]
-        st_s       r3, [sp, 12]
-        st         r4, [sp, 16]
-        st         r5, [sp, 20]
-        st         r6, [sp, 24]
-        st         r7, [sp, 28]
-
-        /* pointer to arguments */
-        mov_s      r2, sp
-
-        /* return value goes here */
-        sub        sp, sp, 8
-        mov_s      r1, sp
-
-        push_s     blink
+        jl         [r11] 
+
+        /* Save return value (r0/r1)  */
+        SARG    r0, [fp, -FRAME_LEN+0*PTRS]
+        SARG    r1, [fp, -FRAME_LEN+1*PTRS]
+
+        /* Restore and return.  */
+        add     sp, fp, -FRAME_LEN
+        .cfi_def_cfa sp, FRAME_LEN
+        LARG    blink, [fp, -1*PTRS]
+        .cfi_restore blink
+        LARG    fp, [fp, -2*PTRS]
+        .cfi_restore fp
+        j_s     [blink]
+        .cfi_endproc
+        .size   ffi_call_asm, .-ffi_call_asm
+
+/*
+  ffi_closure_asm. Expects address of the passed-in ffi_closure in r8.
+  void ffi_closure_inner (ffi_cif *cif,
+		          void (*fun) (ffi_cif *, void *, void **, void *),
+		          void *user_data,
+		          size_t *stackargs, struct call_context *regargs)
+*/
+
+ENTRY(ffi_closure_asm)
+        .cfi_startproc
+
+        ADDPTR  sp, sp, -FRAME_LEN
+        .cfi_def_cfa_offset FRAME_LEN
+
+        /* Make a frame.  */
+        SARG    fp, [sp, FRAME_LEN-2*PTRS]
+        .cfi_offset fp, -2*PTRS
+        SARG    blink, [sp, FRAME_LEN-1*PTRS]
+        .cfi_offset blink, -1*PTRS
+        ADDPTR  fp, sp, FRAME_LEN
+
+        /* Save arguments.  */
+        SARG    r0, [sp, 0*PTRS]
+        SARG    r1, [sp, 1*PTRS]
+        SARG    r2, [sp, 2*PTRS]
+        SARG    r3, [sp, 3*PTRS]
+        SARG    r4, [sp, 4*PTRS]
+        SARG    r5, [sp, 5*PTRS]
+        SARG    r6, [sp, 6*PTRS]
+        SARG    r7, [sp, 7*PTRS]
+
+        /* Enter C.  */
+        LARG    r0, [r8, FFI_TRAMPOLINE_SIZE+0*PTRS]
+        LARG    r1, [r8, FFI_TRAMPOLINE_SIZE+1*PTRS]
+        LARG    r2, [r8, FFI_TRAMPOLINE_SIZE+2*PTRS]
+        ADDPTR  r3, sp, FRAME_LEN
+        MOVPTR  r4, sp
+
+        /* Call the C code.  */
+        bl      ffi_closure_inner
         
-        bl.d       ffi_closure_inner_ARCompact
-        mov_s      r0, r8                /* codeloc, set by trampoline */
-
-        pop_s      blink
-
-        /* set return value to r1:r0 */
-        pop_s      r0
-        pop_s      r1
-        j_s.d      [blink]
-        add_s      sp, sp, 32
+        /* Return values.  */
+        LARG    r0, [sp, 0*PTRS]
+        LARG    r1, [sp, 1*PTRS]
+
+        /* Restore and return.  */
+        LARG    blink, [sp, FRAME_LEN-1*PTRS]
+        .cfi_restore blink
+        LARG    fp, [sp, FRAME_LEN-2*PTRS]
+        .cfi_restore fp
+        ADDPTR  sp, sp, FRAME_LEN
+        .cfi_def_cfa_offset 0
+        j_s     [blink]
+        .cfi_endproc
+        .size ffi_closure_asm, .-ffi_closure_asm
+
+/*
+  ffi_go_closure_asm.  Expects address of the passed-in ffi_go_closure in r12.
+  void ffi_closure_inner (ffi_cif *cif,
+		          void (*fun) (ffi_cif *, void *, void **, void *),
+		          void *user_data,
+		          size_t *stackargs, struct call_context *regargs)
+*/
+
+ENTRY(ffi_go_closure_asm)
+        .cfi_startproc
+
+        ADDPTR  sp, sp, -FRAME_LEN
+        .cfi_def_cfa_offset FRAME_LEN
+
+        /* make a frame */
+        SARG    fp, [sp, FRAME_LEN-2*PTRS]
+        .cfi_offset fp, -2*PTRS
+        SARG    blink, [sp, FRAME_LEN-1*PTRS]
+        .cfi_offset blink, -1*PTRS
+        ADDPTR  fp, sp, FRAME_LEN
+
+        /* save arguments */
+        SARG    r0, [sp, 0*PTRS]
+        SARG    r1, [sp, 1*PTRS]
+        SARG    r2, [sp, 2*PTRS]
+        SARG    r3, [sp, 3*PTRS]
+        SARG    r4, [sp, 4*PTRS]
+        SARG    r5, [sp, 5*PTRS]
+        SARG    r6, [sp, 6*PTRS]
+        SARG    r7, [sp, 7*PTRS]
+
+        /* enter C */
+        LARG    r0, [r12, 1*PTRS]
+        LARG    r1, [r12, 2*PTRS]
+        MOVPTR  r2, r12
+        ADDPTR  r3, sp, FRAME_LEN
+        MOVPTR  r4, sp
+
+        bl      ffi_closure_inner
+
+        /* Return values. */
+        LARG    r0, [sp, 0*PTRS]
+        LARG    r1, [sp, 1*PTRS]
+
+
+        LARG    blink, [sp, FRAME_LEN-1*PTRS]
+        .cfi_restore blink
+        LARG    fp, [sp, FRAME_LEN-2*PTRS]
+        .cfi_restore fp
+        ADDPTR  sp, sp, FRAME_LEN
+        .cfi_def_cfa_offset 0
+        j_s     [blink]
+        .cfi_endproc
+        .size ffi_go_closure_asm, .-ffi_go_closure_asm