aboutsummaryrefslogtreecommitdiffstats
path: root/contrib/restricted/libffi/src/aarch64/ffi.c
blob: 1ebf43c192afe8eadc90fffd600602f30956cef7 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
/* Copyright (c) 2009, 2010, 2011, 2012 ARM Ltd.

Permission is hereby granted, free of charge, to any person obtaining
a copy of this software and associated documentation files (the
``Software''), to deal in the Software without restriction, including
without limitation the rights to use, copy, modify, merge, publish,
distribute, sublicense, and/or sell copies of the Software, and to
permit persons to whom the Software is furnished to do so, subject to
the following conditions:

The above copyright notice and this permission notice shall be
included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED ``AS IS'', WITHOUT WARRANTY OF ANY KIND,
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.  */

#if defined(__aarch64__) || defined(__arm64__)|| defined (_M_ARM64)
#include <stdio.h>
#include <stdlib.h>
#include <stdint.h>
#include <fficonfig.h>
#include <ffi.h>
#include <ffi_common.h>
#include "internal.h"
#ifdef _M_ARM64
#include <windows.h> /* FlushInstructionCache */
#endif

/* Force FFI_TYPE_LONGDOUBLE to be different than FFI_TYPE_DOUBLE;
   all further uses in this file will refer to the 128-bit type.  */
#if FFI_TYPE_DOUBLE != FFI_TYPE_LONGDOUBLE
# if FFI_TYPE_LONGDOUBLE != 4
#  error FFI_TYPE_LONGDOUBLE out of date
# endif
#else
# undef FFI_TYPE_LONGDOUBLE
# define FFI_TYPE_LONGDOUBLE 4
#endif

union _d
{
  UINT64 d;
  UINT32 s[2];
};

struct _v
{
  union _d d[2] __attribute__((aligned(16)));
};

struct call_context
{
  struct _v v[N_V_ARG_REG];
  UINT64 x[N_X_ARG_REG];
};

#if FFI_EXEC_TRAMPOLINE_TABLE

#ifdef __MACH__
#include <mach/vm_param.h>
#endif

#else

#if defined (__clang__) && defined (__APPLE__)
extern void sys_icache_invalidate (void *start, size_t len);
#endif

static inline void
ffi_clear_cache (void *start, void *end)
{
#if defined (__clang__) && defined (__APPLE__)
  sys_icache_invalidate (start, (char *)end - (char *)start);
#elif defined (__GNUC__)
  __builtin___clear_cache (start, end);
#elif defined (_M_ARM64)
  FlushInstructionCache(GetCurrentProcess(), start, (char*)end - (char*)start);
#else
#error "Missing builtin to flush instruction cache"
#endif
}

#endif

/* A subroutine of is_vfp_type.  Given a structure type, return the type code
   of the first non-structure element.  Recurse for structure elements.
   Return -1 if the structure is in fact empty, i.e. no nested elements.  */

static int
is_hfa0 (const ffi_type *ty)
{
  ffi_type **elements = ty->elements;
  int i, ret = -1;

  if (elements != NULL)
    for (i = 0; elements[i]; ++i)
      {
        ret = elements[i]->type;
        if (ret == FFI_TYPE_STRUCT || ret == FFI_TYPE_COMPLEX)
          {
            ret = is_hfa0 (elements[i]);
            if (ret < 0)
              continue;
          }
        break;
      }

  return ret;
}

/* A subroutine of is_vfp_type.  Given a structure type, return true if all
   of the non-structure elements are the same as CANDIDATE.  */

static int
is_hfa1 (const ffi_type *ty, int candidate)
{
  ffi_type **elements = ty->elements;
  int i;

  if (elements != NULL)
    for (i = 0; elements[i]; ++i)
      {
        int t = elements[i]->type;
        if (t == FFI_TYPE_STRUCT || t == FFI_TYPE_COMPLEX)
          {
            if (!is_hfa1 (elements[i], candidate))
              return 0;
          }
        else if (t != candidate)
          return 0;
      }

  return 1;
}

/* Determine if TY may be allocated to the FP registers.  This is both an
   fp scalar type as well as an homogenous floating point aggregate (HFA).
   That is, a structure consisting of 1 to 4 members of all the same type,
   where that type is an fp scalar.

   Returns non-zero iff TY is an HFA.  The result is the AARCH64_RET_*
   constant for the type.  */

static int
is_vfp_type (const ffi_type *ty)
{
  ffi_type **elements;
  int candidate, i;
  size_t size, ele_count;

  /* Quickest tests first.  */
  candidate = ty->type;
  switch (candidate)
    {
    default:
      return 0;
    case FFI_TYPE_FLOAT:
    case FFI_TYPE_DOUBLE:
    case FFI_TYPE_LONGDOUBLE:
      ele_count = 1;
      goto done;
    case FFI_TYPE_COMPLEX:
      candidate = ty->elements[0]->type;
      switch (candidate)
	{
	case FFI_TYPE_FLOAT:
	case FFI_TYPE_DOUBLE:
	case FFI_TYPE_LONGDOUBLE:
	  ele_count = 2;
	  goto done;
	}
      return 0;
    case FFI_TYPE_STRUCT:
      break;
    }

  /* No HFA types are smaller than 4 bytes, or larger than 64 bytes.  */
  size = ty->size;
  if (size < 4 || size > 64)
    return 0;

  /* Find the type of the first non-structure member.  */
  elements = ty->elements;
  candidate = elements[0]->type;
  if (candidate == FFI_TYPE_STRUCT || candidate == FFI_TYPE_COMPLEX)
    {
      for (i = 0; ; ++i)
        {
          candidate = is_hfa0 (elements[i]);
          if (candidate >= 0)
            break;
        }
    }

  /* If the first member is not a floating point type, it's not an HFA.
     Also quickly re-check the size of the structure.  */
  switch (candidate)
    {
    case FFI_TYPE_FLOAT:
      ele_count = size / sizeof(float);
      if (size != ele_count * sizeof(float))
        return 0;
      break;
    case FFI_TYPE_DOUBLE:
      ele_count = size / sizeof(double);
      if (size != ele_count * sizeof(double))
        return 0;
      break;
    case FFI_TYPE_LONGDOUBLE:
      ele_count = size / sizeof(long double);
      if (size != ele_count * sizeof(long double))
        return 0;
      break;
    default:
      return 0;
    }
  if (ele_count > 4)
    return 0;

  /* Finally, make sure that all scalar elements are the same type.  */
  for (i = 0; elements[i]; ++i)
    {
      int t = elements[i]->type;
      if (t == FFI_TYPE_STRUCT || t == FFI_TYPE_COMPLEX)
        {
          if (!is_hfa1 (elements[i], candidate))
            return 0;
        }
      else if (t != candidate)
        return 0;
    }

  /* All tests succeeded.  Encode the result.  */
 done:
  return candidate * 4 + (4 - (int)ele_count);
}

/* Representation of the procedure call argument marshalling
   state.

   The terse state variable names match the names used in the AARCH64
   PCS. */

struct arg_state
{
  unsigned ngrn;                /* Next general-purpose register number. */
  unsigned nsrn;                /* Next vector register number. */
  size_t nsaa;                  /* Next stack offset. */

#if defined (__APPLE__)
  unsigned allocating_variadic;
#endif
};

/* Initialize a procedure call argument marshalling state.  */
static void
arg_init (struct arg_state *state)
{
  state->ngrn = 0;
  state->nsrn = 0;
  state->nsaa = 0;
#if defined (__APPLE__)
  state->allocating_variadic = 0;
#endif
}

/* Allocate an aligned slot on the stack and return a pointer to it.  */
static void *
allocate_to_stack (struct arg_state *state, void *stack,
		   size_t alignment, size_t size)
{
  size_t nsaa = state->nsaa;

  /* Round up the NSAA to the larger of 8 or the natural
     alignment of the argument's type.  */
#if defined (__APPLE__)
  if (state->allocating_variadic && alignment < 8)
    alignment = 8;
#else
  if (alignment < 8)
    alignment = 8;
#endif
    
  nsaa = FFI_ALIGN (nsaa, alignment);
  state->nsaa = nsaa + size;

  return (char *)stack + nsaa;
}

static ffi_arg
extend_integer_type (void *source, int type)
{
  switch (type)
    {
    case FFI_TYPE_UINT8:
      return *(UINT8 *) source;
    case FFI_TYPE_SINT8:
      return *(SINT8 *) source;
    case FFI_TYPE_UINT16:
      return *(UINT16 *) source;
    case FFI_TYPE_SINT16:
      return *(SINT16 *) source;
    case FFI_TYPE_UINT32:
      return *(UINT32 *) source;
    case FFI_TYPE_INT:
    case FFI_TYPE_SINT32:
      return *(SINT32 *) source;
    case FFI_TYPE_UINT64:
    case FFI_TYPE_SINT64:
      return *(UINT64 *) source;
      break;
    case FFI_TYPE_POINTER:
      return *(uintptr_t *) source;
    default:
      abort();
    }
}

#if defined(_MSC_VER)
void extend_hfa_type (void *dest, void *src, int h);
#else
static void
extend_hfa_type (void *dest, void *src, int h)
{
  ssize_t f = h - AARCH64_RET_S4;
  void *x0;

  asm volatile (
	"adr	%0, 0f\n"
"	add	%0, %0, %1\n"
"	br	%0\n"
"0:	ldp	s16, s17, [%3]\n"	/* S4 */
"	ldp	s18, s19, [%3, #8]\n"
"	b	4f\n"
"	ldp	s16, s17, [%3]\n"	/* S3 */
"	ldr	s18, [%3, #8]\n"
"	b	3f\n"
"	ldp	s16, s17, [%3]\n"	/* S2 */
"	b	2f\n"
"	nop\n"
"	ldr	s16, [%3]\n"		/* S1 */
"	b	1f\n"
"	nop\n"
"	ldp	d16, d17, [%3]\n"	/* D4 */
"	ldp	d18, d19, [%3, #16]\n"
"	b	4f\n"
"	ldp	d16, d17, [%3]\n"	/* D3 */
"	ldr	d18, [%3, #16]\n"
"	b	3f\n"
"	ldp	d16, d17, [%3]\n"	/* D2 */
"	b	2f\n"
"	nop\n"
"	ldr	d16, [%3]\n"		/* D1 */
"	b	1f\n"
"	nop\n"
"	ldp	q16, q17, [%3]\n"	/* Q4 */
"	ldp	q18, q19, [%3, #32]\n"
"	b	4f\n"
"	ldp	q16, q17, [%3]\n"	/* Q3 */
"	ldr	q18, [%3, #32]\n"
"	b	3f\n"
"	ldp	q16, q17, [%3]\n"	/* Q2 */
"	b	2f\n"
"	nop\n"
"	ldr	q16, [%3]\n"		/* Q1 */
"	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)
    : "memory", "v16", "v17", "v18", "v19");
}
#endif

#if defined(_MSC_VER)
void* compress_hfa_type (void *dest, void *src, int h);
#else
static void *
compress_hfa_type (void *dest, void *reg, int h)
{
  switch (h)
    {
    case AARCH64_RET_S1:
      if (dest == reg)
	{
#ifdef __AARCH64EB__
	  dest += 12;
#endif
	}
      else
	*(float *)dest = *(float *)reg;
      break;
    case AARCH64_RET_S2:
      asm ("ldp q16, q17, [%1]\n\t"
	   "st2 { v16.s, v17.s }[0], [%0]"
	   : : "r"(dest), "r"(reg) : "memory", "v16", "v17");
      break;
    case AARCH64_RET_S3:
      asm ("ldp q16, q17, [%1]\n\t"
	   "ldr q18, [%1, #32]\n\t"
	   "st3 { v16.s, v17.s, v18.s }[0], [%0]"
	   : : "r"(dest), "r"(reg) : "memory", "v16", "v17", "v18");
      break;
    case AARCH64_RET_S4:
      asm ("ldp q16, q17, [%1]\n\t"
	   "ldp q18, q19, [%1, #32]\n\t"
	   "st4 { v16.s, v17.s, v18.s, v19.s }[0], [%0]"
	   : : "r"(dest), "r"(reg) : "memory", "v16", "v17", "v18", "v19");
      break;

    case AARCH64_RET_D1:
      if (dest == reg)
	{
#ifdef __AARCH64EB__
	  dest += 8;
#endif
	}
      else
	*(double *)dest = *(double *)reg;
      break;
    case AARCH64_RET_D2:
      asm ("ldp q16, q17, [%1]\n\t"
	   "st2 { v16.d, v17.d }[0], [%0]"
	   : : "r"(dest), "r"(reg) : "memory", "v16", "v17");
      break;
    case AARCH64_RET_D3:
      asm ("ldp q16, q17, [%1]\n\t"
	   "ldr q18, [%1, #32]\n\t"
	   "st3 { v16.d, v17.d, v18.d }[0], [%0]"
	   : : "r"(dest), "r"(reg) : "memory", "v16", "v17", "v18");
      break;
    case AARCH64_RET_D4:
      asm ("ldp q16, q17, [%1]\n\t"
	   "ldp q18, q19, [%1, #32]\n\t"
	   "st4 { v16.d, v17.d, v18.d, v19.d }[0], [%0]"
	   : : "r"(dest), "r"(reg) : "memory", "v16", "v17", "v18", "v19");
      break;

    default:
      if (dest != reg)
	return memcpy (dest, reg, 16 * (4 - (h & 3)));
      break;
    }
  return dest;
}
#endif

/* Either allocate an appropriate register for the argument type, or if
   none are available, allocate a stack slot and return a pointer
   to the allocated space.  */

static void *
allocate_int_to_reg_or_stack (struct call_context *context,
			      struct arg_state *state,
			      void *stack, size_t size)
{
  if (state->ngrn < N_X_ARG_REG)
    return &context->x[state->ngrn++];

  state->ngrn = N_X_ARG_REG;
  return allocate_to_stack (state, stack, size, size);
}

ffi_status FFI_HIDDEN
ffi_prep_cif_machdep (ffi_cif *cif)
{
  ffi_type *rtype = cif->rtype;
  size_t bytes = cif->bytes;
  int flags, i, n;

  switch (rtype->type)
    {
    case FFI_TYPE_VOID:
      flags = AARCH64_RET_VOID;
      break;
    case FFI_TYPE_UINT8:
      flags = AARCH64_RET_UINT8;
      break;
    case FFI_TYPE_UINT16:
      flags = AARCH64_RET_UINT16;
      break;
    case FFI_TYPE_UINT32:
      flags = AARCH64_RET_UINT32;
      break;
    case FFI_TYPE_SINT8:
      flags = AARCH64_RET_SINT8;
      break;
    case FFI_TYPE_SINT16:
      flags = AARCH64_RET_SINT16;
      break;
    case FFI_TYPE_INT:
    case FFI_TYPE_SINT32:
      flags = AARCH64_RET_SINT32;
      break;
    case FFI_TYPE_SINT64:
    case FFI_TYPE_UINT64:
      flags = AARCH64_RET_INT64;
      break;
    case FFI_TYPE_POINTER:
      flags = (sizeof(void *) == 4 ? AARCH64_RET_UINT32 : AARCH64_RET_INT64);
      break;

    case FFI_TYPE_FLOAT:
    case FFI_TYPE_DOUBLE:
    case FFI_TYPE_LONGDOUBLE:
    case FFI_TYPE_STRUCT:
    case FFI_TYPE_COMPLEX:
      flags = is_vfp_type (rtype);
      if (flags == 0)
	{
	  size_t s = rtype->size;
	  if (s > 16)
	    {
	      flags = AARCH64_RET_VOID | AARCH64_RET_IN_MEM;
	      bytes += 8;
	    }
	  else if (s == 16)
	    flags = AARCH64_RET_INT128;
	  else if (s == 8)
	    flags = AARCH64_RET_INT64;
	  else
	    flags = AARCH64_RET_INT128 | AARCH64_RET_NEED_COPY;
	}
      break;

    default:
      abort();
    }

  for (i = 0, n = cif->nargs; i < n; i++)
    if (is_vfp_type (cif->arg_types[i]))
      {
	flags |= AARCH64_FLAG_ARG_V;
	break;
      }

  /* Round the stack up to a multiple of the stack alignment requirement. */
  cif->bytes = (unsigned) FFI_ALIGN(bytes, 16);
  cif->flags = flags;
#if defined (__APPLE__)
  cif->aarch64_nfixedargs = 0;
#endif

  return FFI_OK;
}

#if defined (__APPLE__)
/* Perform Apple-specific cif processing for variadic calls */
ffi_status FFI_HIDDEN
ffi_prep_cif_machdep_var(ffi_cif *cif, unsigned int nfixedargs,
			 unsigned int ntotalargs)
{
  ffi_status status = ffi_prep_cif_machdep (cif);
  cif->aarch64_nfixedargs = nfixedargs;
  return status;
}
#endif /* __APPLE__ */

extern void ffi_call_SYSV (struct call_context *context, void *frame,
			   void (*fn)(void), void *rvalue, int flags,
			   void *closure) FFI_HIDDEN;

/* Call a function with the provided arguments and capture the return
   value.  */
static void
ffi_call_int (ffi_cif *cif, void (*fn)(void), void *orig_rvalue,
	      void **avalue, void *closure)
{
  struct call_context *context;
  void *stack, *frame, *rvalue;
  struct arg_state state;
  size_t stack_bytes, rtype_size, rsize;
  int i, nargs, flags;
  ffi_type *rtype;

  flags = cif->flags;
  rtype = cif->rtype;
  rtype_size = rtype->size;
  stack_bytes = cif->bytes;

  /* If the target function returns a structure via hidden pointer,
     then we cannot allow a null rvalue.  Otherwise, mash a null
     rvalue to void return type.  */
  rsize = 0;
  if (flags & AARCH64_RET_IN_MEM)
    {
      if (orig_rvalue == NULL)
	rsize = rtype_size;
    }
  else if (orig_rvalue == NULL)
    flags &= AARCH64_FLAG_ARG_V;
  else if (flags & AARCH64_RET_NEED_COPY)
    rsize = 16;

  /* Allocate consectutive stack for everything we'll need.  */
  context = alloca (sizeof(struct call_context) + stack_bytes + 32 + rsize);
  stack = context + 1;
  frame = (void*)((uintptr_t)stack + (uintptr_t)stack_bytes);
  rvalue = (rsize ? (void*)((uintptr_t)frame + 32) : orig_rvalue);

  arg_init (&state);
  for (i = 0, nargs = cif->nargs; i < nargs; i++)
    {
      ffi_type *ty = cif->arg_types[i];
      size_t s = ty->size;
      void *a = avalue[i];
      int h, t;

      t = ty->type;
      switch (t)
	{
	case FFI_TYPE_VOID:
	  FFI_ASSERT (0);
	  break;

	/* If the argument is a basic type the argument is allocated to an
	   appropriate register, or if none are available, to the stack.  */
	case FFI_TYPE_INT:
	case FFI_TYPE_UINT8:
	case FFI_TYPE_SINT8:
	case FFI_TYPE_UINT16:
	case FFI_TYPE_SINT16:
	case FFI_TYPE_UINT32:
	case FFI_TYPE_SINT32:
	case FFI_TYPE_UINT64:
	case FFI_TYPE_SINT64:
	case FFI_TYPE_POINTER:
	do_pointer:
	  {
	    ffi_arg ext = extend_integer_type (a, t);
	    if (state.ngrn < N_X_ARG_REG)
	      context->x[state.ngrn++] = ext;
	    else
	      {
		void *d = allocate_to_stack (&state, stack, ty->alignment, s);
		state.ngrn = N_X_ARG_REG;
		/* Note that the default abi extends each argument
		   to a full 64-bit slot, while the iOS abi allocates
		   only enough space. */
#ifdef __APPLE__
		memcpy(d, a, s);
#else
		*(ffi_arg *)d = ext;
#endif
	      }
	  }
	  break;

	case FFI_TYPE_FLOAT:
	case FFI_TYPE_DOUBLE:
	case FFI_TYPE_LONGDOUBLE:
	case FFI_TYPE_STRUCT:
	case FFI_TYPE_COMPLEX:
	  {
	    void *dest;

	    h = is_vfp_type (ty);
	    if (h)
	      {
		int elems = 4 - (h & 3);
#ifdef _M_ARM64 /* for handling armasm calling convention */
                if (cif->is_variadic)
                  {
                    if (state.ngrn + elems <= N_X_ARG_REG)
                      {
                        dest = &context->x[state.ngrn];
                        state.ngrn += elems;
                        extend_hfa_type(dest, a, h);
                        break;
                      }
                    state.nsrn = N_X_ARG_REG;
                    dest = allocate_to_stack(&state, stack, ty->alignment, s);
                  }
                else
                  {
#endif /* for handling armasm calling convention */
	        if (state.nsrn + elems <= N_V_ARG_REG)
		  {
		    dest = &context->v[state.nsrn];
		    state.nsrn += elems;
		    extend_hfa_type (dest, a, h);
		    break;
		  }
		state.nsrn = N_V_ARG_REG;
		dest = allocate_to_stack (&state, stack, ty->alignment, s);
#ifdef _M_ARM64 /* for handling armasm calling convention */
	      }
#endif /* for handling armasm calling convention */
	      }
	    else if (s > 16)
	      {
		/* If the argument is a composite type that is larger than 16
		   bytes, then the argument has been copied to memory, and
		   the argument is replaced by a pointer to the copy.  */
		a = &avalue[i];
		t = FFI_TYPE_POINTER;
		s = sizeof (void *);
		goto do_pointer;
	      }
	    else
	      {
		size_t n = (s + 7) / 8;
		if (state.ngrn + n <= N_X_ARG_REG)
		  {
		    /* If the argument is a composite type and the size in
		       double-words is not more than the number of available
		       X registers, then the argument is copied into
		       consecutive X registers.  */
		    dest = &context->x[state.ngrn];
                    state.ngrn += (unsigned int)n;
		  }
		else
		  {
		    /* Otherwise, there are insufficient X registers. Further
		       X register allocations are prevented, the NSAA is
		       adjusted and the argument is copied to memory at the
		       adjusted NSAA.  */
		    state.ngrn = N_X_ARG_REG;
		    dest = allocate_to_stack (&state, stack, ty->alignment, s);
		  }
		}
	      memcpy (dest, a, s);
	    }
	  break;

	default:
	  abort();
	}

#if defined (__APPLE__)
      if (i + 1 == cif->aarch64_nfixedargs)
	{
	  state.ngrn = N_X_ARG_REG;
	  state.nsrn = N_V_ARG_REG;
	  state.allocating_variadic = 1;
	}
#endif
    }

  ffi_call_SYSV (context, frame, fn, rvalue, flags, closure);

  if (flags & AARCH64_RET_NEED_COPY)
    memcpy (orig_rvalue, rvalue, rtype_size);
}

void
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 /* FFI_GO_CLOSURES */

/* Build a trampoline.  */

extern void ffi_closure_SYSV (void) FFI_HIDDEN;
extern void ffi_closure_SYSV_V (void) FFI_HIDDEN;

ffi_status
ffi_prep_closure_loc (ffi_closure *closure,
                      ffi_cif* cif,
                      void (*fun)(ffi_cif*,void*,void**,void*),
                      void *user_data,
                      void *codeloc)
{
  if (cif->abi != FFI_SYSV)
    return FFI_BAD_ABI;

  void (*start)(void);
  
  if (cif->flags & AARCH64_FLAG_ARG_V)
    start = ffi_closure_SYSV_V;
  else
    start = ffi_closure_SYSV;

#if FFI_EXEC_TRAMPOLINE_TABLE
#ifdef __MACH__
  void **config = (void **)((uint8_t *)codeloc - PAGE_MAX_SIZE);
  config[0] = closure;
  config[1] = start;
#endif
#else
  static const unsigned char trampoline[16] = {
    0x90, 0x00, 0x00, 0x58,	/* ldr	x16, tramp+16	*/
    0xf1, 0xff, 0xff, 0x10,	/* adr	x17, tramp+0	*/
    0x00, 0x02, 0x1f, 0xd6	/* br	x16		*/
  };
  char *tramp = closure->tramp;
  
  memcpy (tramp, trampoline, sizeof(trampoline));
  
  *(UINT64 *)(tramp + 16) = (uintptr_t)start;

  ffi_clear_cache(tramp, tramp + FFI_TRAMPOLINE_SIZE);

  /* Also flush the cache for code mapping.  */
#ifdef _M_ARM64
  // Not using dlmalloc.c for Windows ARM64 builds
  // so calling ffi_data_to_code_pointer() isn't necessary
  unsigned char *tramp_code = tramp;
  #else
  unsigned char *tramp_code = ffi_data_to_code_pointer (tramp);
  #endif
  ffi_clear_cache (tramp_code, tramp_code + FFI_TRAMPOLINE_SIZE);
#endif

  closure->cif = cif;
  closure->fun = fun;
  closure->user_data = user_data;

  return FFI_OK;
}

#ifdef FFI_GO_CLOSURES
extern void ffi_go_closure_SYSV (void) FFI_HIDDEN;
extern void ffi_go_closure_SYSV_V (void) FFI_HIDDEN;

ffi_status
ffi_prep_go_closure (ffi_go_closure *closure, ffi_cif* cif,
                     void (*fun)(ffi_cif*,void*,void**,void*))
{
  void (*start)(void);

  if (cif->abi != FFI_SYSV)
    return FFI_BAD_ABI;

  if (cif->flags & AARCH64_FLAG_ARG_V)
    start = ffi_go_closure_SYSV_V;
  else
    start = ffi_go_closure_SYSV;

  closure->tramp = start;
  closure->cif = cif;
  closure->fun = fun;

  return FFI_OK;
}
#endif /* FFI_GO_CLOSURES */

/* Primary handler to setup and invoke a function within a closure.

   A closure when invoked enters via the assembler wrapper
   ffi_closure_SYSV(). The wrapper allocates a call context on the
   stack, saves the interesting registers (from the perspective of
   the calling convention) into the context then passes control to
   ffi_closure_SYSV_inner() passing the saved context and a pointer to
   the stack at the point ffi_closure_SYSV() was invoked.

   On the return path the assembler wrapper will reload call context
   registers.

   ffi_closure_SYSV_inner() marshalls the call context into ffi value
   descriptors, invokes the wrapped function, then marshalls the return
   value back into the call context.  */

int FFI_HIDDEN
ffi_closure_SYSV_inner (ffi_cif *cif,
			void (*fun)(ffi_cif*,void*,void**,void*),
			void *user_data,
			struct call_context *context,
			void *stack, void *rvalue, void *struct_rvalue)
{
  void **avalue = (void**) alloca (cif->nargs * sizeof (void*));
  int i, h, nargs, flags;
  struct arg_state state;

  arg_init (&state);

  for (i = 0, nargs = cif->nargs; i < nargs; i++)
    {
      ffi_type *ty = cif->arg_types[i];
      int t = ty->type;
      size_t n, s = ty->size;

      switch (t)
	{
	case FFI_TYPE_VOID:
	  FFI_ASSERT (0);
	  break;

	case FFI_TYPE_INT:
	case FFI_TYPE_UINT8:
	case FFI_TYPE_SINT8:
	case FFI_TYPE_UINT16:
	case FFI_TYPE_SINT16:
	case FFI_TYPE_UINT32:
	case FFI_TYPE_SINT32:
	case FFI_TYPE_UINT64:
	case FFI_TYPE_SINT64:
	case FFI_TYPE_POINTER:
	  avalue[i] = allocate_int_to_reg_or_stack (context, &state, stack, s);
	  break;

	case FFI_TYPE_FLOAT:
	case FFI_TYPE_DOUBLE:
	case FFI_TYPE_LONGDOUBLE:
	case FFI_TYPE_STRUCT:
	case FFI_TYPE_COMPLEX:
	  h = is_vfp_type (ty);
	  if (h)
	    {
	      n = 4 - (h & 3);
#ifdef _M_ARM64  /* for handling armasm calling convention */
              if (cif->is_variadic)
                {
                  if (state.ngrn + n <= N_X_ARG_REG)
                    {
                      void *reg = &context->x[state.ngrn];
                      state.ngrn += (unsigned int)n;
    
                      /* Eeek! We need a pointer to the structure, however the
                       homogeneous float elements are being passed in individual
                       registers, therefore for float and double the structure
                       is not represented as a contiguous sequence of bytes in
                       our saved register context.  We don't need the original
                       contents of the register storage, so we reformat the
                       structure into the same memory.  */
                      avalue[i] = compress_hfa_type(reg, reg, h);
                    }
                  else
                    {
                      state.ngrn = N_X_ARG_REG;
                      state.nsrn = N_V_ARG_REG;
                      avalue[i] = allocate_to_stack(&state, stack,
                             ty->alignment, s);
                    }
                }
              else
                {
#endif  /* for handling armasm calling convention */
                  if (state.nsrn + n <= N_V_ARG_REG)
                    {
                      void *reg = &context->v[state.nsrn];
                      state.nsrn += (unsigned int)n;
                      avalue[i] = compress_hfa_type(reg, reg, h);
                    }
                  else
                    {
                      state.nsrn = N_V_ARG_REG;
                      avalue[i] = allocate_to_stack(&state, stack,
                                                   ty->alignment, s);
                    }
#ifdef _M_ARM64  /* for handling armasm calling convention */
                }
#endif  /* for handling armasm calling convention */
            }
          else if (s > 16)
            {
              /* Replace Composite type of size greater than 16 with a
                  pointer.  */
              avalue[i] = *(void **)
              allocate_int_to_reg_or_stack (context, &state, stack,
                                         sizeof (void *));
            }
          else
            {
              n = (s + 7) / 8;
              if (state.ngrn + n <= N_X_ARG_REG)
                {
                  avalue[i] = &context->x[state.ngrn];
                  state.ngrn += (unsigned int)n;
                }
              else
                {
                  state.ngrn = N_X_ARG_REG;
                  avalue[i] = allocate_to_stack(&state, stack,
                                           ty->alignment, s);
                }
            }
          break;

        default:
          abort();
      }

#if defined (__APPLE__)
      if (i + 1 == cif->aarch64_nfixedargs)
	{
	  state.ngrn = N_X_ARG_REG;
	  state.nsrn = N_V_ARG_REG;
	  state.allocating_variadic = 1;
	}
#endif
    }

  flags = cif->flags;
  if (flags & AARCH64_RET_IN_MEM)
    rvalue = struct_rvalue;

  fun (cif, rvalue, avalue, user_data);

  return flags;
}

#endif /* (__aarch64__) || defined(__arm64__)|| defined (_M_ARM64)*/