New RISC-V port (#281) * Add RISC-V support This patch adds support for the RISC-V architecture (https://riscv.org). This patch has been tested using QEMU user-mode emulation and GCC 7.2.0 in the following configurations: * -march=rv32imac -mabi=ilp32 * -march=rv32g -mabi=ilp32d * -march=rv64imac -mabi=lp64 * -march=rv64g -mabi=lp64d The ABI currently can be found at https://github.com/riscv/riscv-elf-psabi-doc/blob/master/riscv-elf.md . * Add RISC-V to README * RISC-V: fix configure.host
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
diff --git a/Makefile.am b/Makefile.am
index 7145a6c..fad1bdd 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -109,6 +109,7 @@ noinst_HEADERS = \
src/or1k/ffitarget.h \
src/pa/ffitarget.h \
src/powerpc/ffitarget.h src/powerpc/asm.h src/powerpc/ffi_powerpc.h \
+ src/riscv/ffitarget.h \
src/s390/ffitarget.h src/s390/internal.h \
src/sh/ffitarget.h \
src/sh64/ffitarget.h \
@@ -144,6 +145,7 @@ EXTRA_libffi_la_SOURCES = \
src/powerpc/linux64_closure.S src/powerpc/ppc_closure.S \
src/powerpc/aix.S src/powerpc/darwin.S src/powerpc/aix_closure.S \
src/powerpc/darwin_closure.S src/powerpc/ffi_darwin.c \
+ src/riscv/ffi.c src/riscv/sysv.S \
src/s390/ffi.c src/s390/sysv.S \
src/sh/ffi.c src/sh/sysv.S \
src/sh64/ffi.c src/sh64/sysv.S \
diff --git a/README b/README
index 8ebe74d..63e2e97 100644
--- a/README
+++ b/README
@@ -84,6 +84,8 @@ tested:
| PowerPC 64-bit | FreeBSD | GCC |
| PowerPC 64-bit | Linux ELFv1 | GCC |
| PowerPC 64-bit | Linux ELFv2 | GCC |
+| RISC-V 32-bit | Linux | GCC |
+| RISC-V 64-bit | Linux | GCC |
| S390 | Linux | GCC |
| S390X | Linux | GCC |
| SPARC | Linux | GCC |
@@ -188,6 +190,7 @@ See the git log for details at http://github.com/libffi/libffi.
4.0 TBD
New API in support of GO closures.
+ Add RISC-V support.
3.2.1 Nov-12-14
Build fix for non-iOS AArch64 targets.
diff --git a/configure.host b/configure.host
index 34e83f7..7634c3a 100644
--- a/configure.host
+++ b/configure.host
@@ -206,6 +206,11 @@ case "${host}" in
TARGET=POWERPC; TARGETDIR=powerpc
;;
+ riscv*-*)
+ TARGET=RISCV; TARGETDIR=riscv
+ SOURCES="ffi.c sysv.S"
+ ;;
+
s390-*-* | s390x-*-*)
TARGET=S390; TARGETDIR=s390
SOURCES="ffi.c sysv.S"
diff --git a/include/ffi_common.h b/include/ffi_common.h
index d1d9fd8..7fb0a64 100644
--- a/include/ffi_common.h
+++ b/include/ffi_common.h
@@ -74,7 +74,9 @@ void ffi_type_test(ffi_type *a, char *file, int line);
#define FFI_ASSERT_VALID_TYPE(x)
#endif
+/* v cast to size_t and aligned up to a multiple of a */
#define FFI_ALIGN(v, a) (((((size_t) (v))-1) | ((a)-1))+1)
+/* v cast to size_t and aligned down to a multiple of a */
#define ALIGN_DOWN(v, a) (((size_t) (v)) & -a)
/* Perform machine dependent cif processing */
diff --git a/src/riscv/ffi.c b/src/riscv/ffi.c
new file mode 100644
index 0000000..b744fdd
--- /dev/null
+++ b/src/riscv/ffi.c
@@ -0,0 +1,445 @@
+/* -----------------------------------------------------------------------
+ ffi.c - Copyright (c) 2015 Michael Knyszek <mknyszek@berkeley.edu>
+ 2015 Andrew Waterman <waterman@cs.berkeley.edu>
+ 2018 Stef O'Rear <sorear2@gmail.com>
+ Based on MIPS N32/64 port
+
+ RISC-V Foreign Function Interface
+
+ 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.
+ ----------------------------------------------------------------------- */
+
+#include <ffi.h>
+#include <ffi_common.h>
+
+#include <stdlib.h>
+#include <stdint.h>
+
+#if __riscv_float_abi_double
+#define ABI_FLEN 64
+#define ABI_FLOAT double
+#elif __riscv_float_abi_single
+#define ABI_FLEN 32
+#define ABI_FLOAT float
+#endif
+
+#define NARGREG 8
+#define STKALIGN 16
+#define MAXCOPYARG (2 * sizeof(double))
+
+typedef struct call_context
+{
+#if ABI_FLEN
+ ABI_FLOAT fa[8];
+#endif
+ size_t a[8];
+ /* used by the assembly code to in-place construct its own stack frame */
+ char frame[16];
+} call_context;
+
+typedef struct call_builder
+{
+ call_context *aregs;
+ int used_integer;
+ int used_float;
+ size_t *used_stack;
+} call_builder;
+
+/* integer (not pointer) less than ABI XLEN */
+/* FFI_TYPE_INT does not appear to be used */
+#if __SIZEOF_POINTER__ == 8
+#define IS_INT(type) ((type) >= FFI_TYPE_UINT8 && (type) <= FFI_TYPE_SINT64)
+#else
+#define IS_INT(type) ((type) >= FFI_TYPE_UINT8 && (type) <= FFI_TYPE_SINT32)
+#endif
+
+#if ABI_FLEN
+typedef struct {
+ char as_elements, type1, offset2, type2;
+} float_struct_info;
+
+#if ABI_FLEN >= 64
+#define IS_FLOAT(type) ((type) >= FFI_TYPE_FLOAT && (type) <= FFI_TYPE_DOUBLE)
+#else
+#define IS_FLOAT(type) ((type) == FFI_TYPE_FLOAT)
+#endif
+
+static ffi_type **flatten_struct(ffi_type *in, ffi_type **out, ffi_type **out_end) {
+ int i;
+ if (out == out_end) return out;
+ if (in->type != FFI_TYPE_STRUCT) {
+ *(out++) = in;
+ } else {
+ for (i = 0; in->elements[i]; i++)
+ out = flatten_struct(in->elements[i], out, out_end);
+ }
+ return out;
+}
+
+/* Structs with at most two fields after flattening, one of which is of
+ floating point type, are passed in multiple registers if sufficient
+ registers are available. */
+static float_struct_info struct_passed_as_elements(call_builder *cb, ffi_type *top) {
+ float_struct_info ret = {0, 0, 0, 0};
+ ffi_type *fields[3];
+ int num_floats, num_ints;
+ int num_fields = flatten_struct(top, fields, fields + 3) - fields;
+
+ if (num_fields == 1) {
+ if (IS_FLOAT(fields[0]->type)) {
+ ret.as_elements = 1;
+ ret.type1 = fields[0]->type;
+ }
+ } else if (num_fields == 2) {
+ num_floats = IS_FLOAT(fields[0]->type) + IS_FLOAT(fields[1]->type);
+ num_ints = IS_INT(fields[0]->type) + IS_INT(fields[1]->type);
+ if (num_floats == 0 || num_floats + num_ints != 2)
+ return ret;
+ if (cb->used_float + num_floats > NARGREG || cb->used_integer + (2 - num_floats) > NARGREG)
+ return ret;
+ if (!IS_FLOAT(fields[0]->type) && !IS_FLOAT(fields[1]->type))
+ return ret;
+
+ ret.type1 = fields[0]->type;
+ ret.type2 = fields[1]->type;
+ ret.offset2 = FFI_ALIGN(fields[0]->size, fields[1]->alignment);
+ ret.as_elements = 1;
+ }
+
+ return ret;
+}
+#endif
+
+/* allocates a single register, float register, or XLEN-sized stack slot to a datum */
+static void marshal_atom(call_builder *cb, int type, void *data) {
+ size_t value = 0;
+ switch (type) {
+ case FFI_TYPE_UINT8: value = *(uint8_t *)data; break;
+ case FFI_TYPE_SINT8: value = *(int8_t *)data; break;
+ case FFI_TYPE_UINT16: value = *(uint16_t *)data; break;
+ case FFI_TYPE_SINT16: value = *(int16_t *)data; break;
+ /* 32-bit quantities are always sign-extended in the ABI */
+ case FFI_TYPE_UINT32: value = *(int32_t *)data; break;
+ case FFI_TYPE_SINT32: value = *(int32_t *)data; break;
+#if __SIZEOF_POINTER__ == 8
+ case FFI_TYPE_UINT64: value = *(uint64_t *)data; break;
+ case FFI_TYPE_SINT64: value = *(int64_t *)data; break;
+#endif
+ case FFI_TYPE_POINTER: value = *(size_t *)data; break;
+
+ /* float values may be recoded in an implementation-defined way
+ by hardware conforming to 2.1 or earlier, so use asm to
+ reinterpret floats as doubles */
+#if ABI_FLEN >= 32
+ case FFI_TYPE_FLOAT:
+ asm("" : "=f"(cb->aregs->fa[cb->used_float++]) : "0"(*(float *)data));
+ return;
+#endif
+#if ABI_FLEN >= 64
+ case FFI_TYPE_DOUBLE:
+ asm("" : "=f"(cb->aregs->fa[cb->used_float++]) : "0"(*(double *)data));
+ return;
+#endif
+ default: FFI_ASSERT(0); break;
+ }
+
+ if (cb->used_integer == NARGREG) {
+ *cb->used_stack++ = value;
+ } else {
+ cb->aregs->a[cb->used_integer++] = value;
+ }
+}
+
+static void unmarshal_atom(call_builder *cb, int type, void *data) {
+ size_t value;
+ switch (type) {
+#if ABI_FLEN >= 32
+ case FFI_TYPE_FLOAT:
+ asm("" : "=f"(*(float *)data) : "0"(cb->aregs->fa[cb->used_float++]));
+ return;
+#endif
+#if ABI_FLEN >= 64
+ case FFI_TYPE_DOUBLE:
+ asm("" : "=f"(*(double *)data) : "0"(cb->aregs->fa[cb->used_float++]));
+ return;
+#endif
+ }
+
+ if (cb->used_integer == NARGREG) {
+ value = *cb->used_stack++;
+ } else {
+ value = cb->aregs->a[cb->used_integer++];
+ }
+
+ switch (type) {
+ case FFI_TYPE_UINT8: *(uint8_t *)data = value; break;
+ case FFI_TYPE_SINT8: *(uint8_t *)data = value; break;
+ case FFI_TYPE_UINT16: *(uint16_t *)data = value; break;
+ case FFI_TYPE_SINT16: *(uint16_t *)data = value; break;
+ case FFI_TYPE_UINT32: *(uint32_t *)data = value; break;
+ case FFI_TYPE_SINT32: *(uint32_t *)data = value; break;
+#if __SIZEOF_POINTER__ == 8
+ case FFI_TYPE_UINT64: *(uint64_t *)data = value; break;
+ case FFI_TYPE_SINT64: *(uint64_t *)data = value; break;
+#endif
+ case FFI_TYPE_POINTER: *(size_t *)data = value; break;
+ default: FFI_ASSERT(0); break;
+ }
+}
+
+/* adds an argument to a call, or a not by reference return value */
+static void marshal(call_builder *cb, ffi_type *type, int var, void *data) {
+ size_t realign[2];
+
+#if ABI_FLEN
+ if (!var && type->type == FFI_TYPE_STRUCT) {
+ float_struct_info fsi = struct_passed_as_elements(cb, type);
+ if (fsi.as_elements) {
+ marshal_atom(cb, fsi.type1, data);
+ if (fsi.offset2)
+ marshal_atom(cb, fsi.type2, ((char*)data) + fsi.offset2);
+ return;
+ }
+ }
+
+ if (!var && cb->used_float < NARGREG && IS_FLOAT(type->type)) {
+ marshal_atom(cb, type->type, data);
+ return;
+ }
+#endif
+
+ if (type->size > 2 * __SIZEOF_POINTER__) {
+ /* pass by reference */
+ marshal_atom(cb, FFI_TYPE_POINTER, &data);
+ } else if (IS_INT(type->type) || type->type == FFI_TYPE_POINTER) {
+ marshal_atom(cb, type->type, data);
+ } else {
+ /* overlong integers, soft-float floats, and structs without special
+ float handling are treated identically from this point on */
+
+ /* variadics are aligned even in registers */
+ if (type->alignment > __SIZEOF_POINTER__) {
+ if (var)
+ cb->used_integer = FFI_ALIGN(cb->used_integer, 2);
+ cb->used_stack = (size_t *)FFI_ALIGN(cb->used_stack, 2*__SIZEOF_POINTER__);
+ }
+
+ memcpy(realign, data, type->size);
+ if (type->size > 0)
+ marshal_atom(cb, FFI_TYPE_POINTER, realign);
+ if (type->size > __SIZEOF_POINTER__)
+ marshal_atom(cb, FFI_TYPE_POINTER, realign + 1);
+ }
+}
+
+/* for arguments passed by reference returns the pointer, otherwise the arg is copied (up to MAXCOPYARG bytes) */
+static void *unmarshal(call_builder *cb, ffi_type *type, int var, void *data) {
+ size_t realign[2];
+ void *pointer;
+
+#if ABI_FLEN
+ if (!var && type->type == FFI_TYPE_STRUCT) {
+ float_struct_info fsi = struct_passed_as_elements(cb, type);
+ if (fsi.as_elements) {
+ unmarshal_atom(cb, fsi.type1, data);
+ if (fsi.offset2)
+ unmarshal_atom(cb, fsi.type2, ((char*)data) + fsi.offset2);
+ return data;
+ }
+ }
+
+ if (!var && cb->used_float < NARGREG && IS_FLOAT(type->type)) {
+ unmarshal_atom(cb, type->type, data);
+ return data;
+ }
+#endif
+
+ if (type->size > 2 * __SIZEOF_POINTER__) {
+ /* pass by reference */
+ unmarshal_atom(cb, FFI_TYPE_POINTER, (char*)&pointer);
+ return pointer;
+ } else if (IS_INT(type->type) || type->type == FFI_TYPE_POINTER) {
+ unmarshal_atom(cb, type->type, data);
+ return data;
+ } else {
+ /* overlong integers, soft-float floats, and structs without special
+ float handling are treated identically from this point on */
+
+ /* variadics are aligned even in registers */
+ if (type->alignment > __SIZEOF_POINTER__) {
+ if (var)
+ cb->used_integer = FFI_ALIGN(cb->used_integer, 2);
+ cb->used_stack = (size_t *)FFI_ALIGN(cb->used_stack, 2*__SIZEOF_POINTER__);
+ }
+
+ if (type->size > 0)
+ unmarshal_atom(cb, FFI_TYPE_POINTER, realign);
+ if (type->size > __SIZEOF_POINTER__)
+ unmarshal_atom(cb, FFI_TYPE_POINTER, realign + 1);
+ memcpy(data, realign, type->size);
+ return data;
+ }
+}
+
+static int passed_by_ref(call_builder *cb, ffi_type *type, int var) {
+#if ABI_FLEN
+ if (!var && type->type == FFI_TYPE_STRUCT) {
+ float_struct_info fsi = struct_passed_as_elements(cb, type);
+ if (fsi.as_elements) return 0;
+ }
+#endif
+
+ return type->size > 2 * __SIZEOF_POINTER__;
+}
+
+/* Perform machine dependent cif processing */
+ffi_status ffi_prep_cif_machdep(ffi_cif *cif) {
+ cif->riscv_nfixedargs = cif->nargs;
+ return FFI_OK;
+}
+
+/* Perform machine dependent cif processing when we have a variadic function */
+
+ffi_status ffi_prep_cif_machdep_var(ffi_cif *cif, unsigned int nfixedargs, unsigned int ntotalargs) {
+ cif->riscv_nfixedargs = nfixedargs;
+ return FFI_OK;
+}
+
+/* Low level routine for calling functions */
+extern void ffi_call_asm(void *stack, struct call_context *regs, void (*fn)(void)) FFI_HIDDEN;
+
+void ffi_call(ffi_cif *cif, void (*fn)(void), void *rvalue, void **avalue)
+{
+ /* this is a conservative estimate, assuming a complex return value and
+ that all remaining arguments are long long / __int128 */
+ size_t arg_bytes = cif->nargs <= 3 ? 0 :
+ FFI_ALIGN(2 * sizeof(size_t) * (cif->nargs - 3), STKALIGN);
+ size_t rval_bytes = 0;
+ if (rvalue == NULL && cif->rtype->size > 2*__SIZEOF_POINTER__)
+ rval_bytes = FFI_ALIGN(cif->rtype->size, STKALIGN);
+ size_t alloc_size = arg_bytes + rval_bytes + sizeof(call_context);
+
+ /* the assembly code will deallocate all stack data at lower addresses
+ than the argument region, so we need to allocate the frame and the
+ return value after the arguments in a single allocation */
+ size_t alloc_base;
+ /* Argument region must be 16-byte aligned */
+ if (_Alignof(max_align_t) >= STKALIGN) {
+ /* since sizeof long double is normally 16, the compiler will
+ guarantee alloca alignment to at least that much */
+ alloc_base = (size_t)alloca(alloc_size);
+ } else {
+ alloc_base = FFI_ALIGN(alloca(alloc_size + STKALIGN - 1), STKALIGN);
+ }
+
+ if (rval_bytes)
+ rvalue = (void*)(alloc_base + arg_bytes);
+
+ call_builder cb;
+ cb.used_float = cb.used_integer = 0;
+ cb.aregs = (call_context*)(alloc_base + arg_bytes + rval_bytes);
+ cb.used_stack = (void*)alloc_base;
+
+ int return_by_ref = passed_by_ref(&cb, cif->rtype, 0);
+ if (return_by_ref)
+ marshal(&cb, &ffi_type_pointer, 0, &rvalue);
+
+ int i;
+ for (i = 0; i < cif->nargs; i++)
+ marshal(&cb, cif->arg_types[i], i >= cif->riscv_nfixedargs, avalue[i]);
+
+ ffi_call_asm((void*)alloc_base, cb.aregs, fn);
+
+ cb.used_float = cb.used_integer = 0;
+ if (!return_by_ref && rvalue)
+ unmarshal(&cb, cif->rtype, 0, rvalue);
+}
+
+extern void ffi_closure_asm(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)
+{
+ uint32_t *tramp = (uint32_t *) &closure->tramp[0];
+ uint64_t fn = (uint64_t) (uintptr_t) ffi_closure_asm;
+
+ if (cif->abi <= FFI_FIRST_ABI || cif->abi >= FFI_LAST_ABI)
+ return FFI_BAD_ABI;
+
+ /* we will call ffi_closure_inner with codeloc, not closure, but as long
+ as the memory is readable it should work */
+
+ tramp[0] = 0x00000317; /* auipc t1, 0 (i.e. t0 <- codeloc) */
+#if __SIZEOF_POINTER__ == 8
+ tramp[1] = 0x01033383; /* ld t2, 16(t1) */
+#else
+ tramp[1] = 0x01032383; /* lw t2, 16(t1) */
+#endif
+ tramp[2] = 0x00038067; /* jr t2 */
+ tramp[3] = 0x00000013; /* nop */
+ tramp[4] = fn;
+ tramp[5] = fn >> 32;
+
+ closure->cif = cif;
+ closure->fun = fun;
+ closure->user_data = user_data;
+
+ __builtin___clear_cache(codeloc, codeloc + FFI_TRAMPOLINE_SIZE);
+
+ return FFI_OK;
+}
+
+/* Called by the assembly code with aregs pointing to saved argument registers
+ and stack pointing to the stacked arguments. Return values passed in
+ registers will be reloaded from aregs. */
+void FFI_HIDDEN ffi_closure_inner(size_t *stack, call_context *aregs, ffi_closure *closure) {
+ ffi_cif *cif = closure->cif;
+ void **avalue = alloca(cif->nargs * sizeof(void*));
+ /* storage for arguments which will be copied by unmarshal(). We could
+ theoretically avoid the copies in many cases and use at most 128 bytes
+ of memory, but allocating disjoint storage for each argument is
+ simpler. */
+ char *astorage = alloca(cif->nargs * MAXCOPYARG);
+ void *rvalue;
+ call_builder cb;
+ int return_by_ref;
+ int i;
+
+ cb.aregs = aregs;
+ cb.used_integer = cb.used_float = 0;
+ cb.used_stack = stack;
+
+ return_by_ref = passed_by_ref(&cb, cif->rtype, 0);
+ if (return_by_ref)
+ unmarshal(&cb, &ffi_type_pointer, 0, &rvalue);
+ else
+ rvalue = alloca(cif->rtype->size);
+
+ for (i = 0; i < cif->nargs; i++)
+ avalue[i] = unmarshal(&cb, cif->arg_types[i],
+ i >= cif->riscv_nfixedargs, astorage + i*MAXCOPYARG);
+
+ (closure->fun)(cif, rvalue, avalue, closure->user_data);
+
+ if (!return_by_ref && cif->rtype->type != FFI_TYPE_VOID) {
+ cb.used_integer = cb.used_float = 0;
+ marshal(&cb, cif->rtype, 0, rvalue);
+ }
+}
diff --git a/src/riscv/ffitarget.h b/src/riscv/ffitarget.h
new file mode 100644
index 0000000..fcaa899
--- /dev/null
+++ b/src/riscv/ffitarget.h
@@ -0,0 +1,68 @@
+/* -----------------------------------------------------------------*-C-*-
+ ffitarget.h - 2014 Michael Knyszek
+
+ Target configuration macros for RISC-V.
+
+ 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.
+
+ ----------------------------------------------------------------------- */
+
+#ifndef LIBFFI_TARGET_H
+#define LIBFFI_TARGET_H
+
+#ifndef LIBFFI_H
+#error "Please do not include ffitarget.h directly into your source. Use ffi.h instead."
+#endif
+
+#ifndef __riscv
+#error "libffi was configured for a RISC-V target but this does not appear to be a RISC-V compiler."
+#endif
+
+#ifndef LIBFFI_ASM
+
+typedef unsigned long ffi_arg;
+typedef signed long ffi_sarg;
+
+/* FFI_UNUSED_NN and riscv_unused are to maintain ABI compatibility with a
+ distributed Berkeley patch from 2014, and can be removed at SONAME bump */
+typedef enum ffi_abi {
+ FFI_FIRST_ABI = 0,
+ FFI_SYSV,
+ FFI_UNUSED_1,
+ FFI_UNUSED_2,
+ FFI_UNUSED_3,
+ FFI_LAST_ABI,
+
+ FFI_DEFAULT_ABI = FFI_SYSV
+} ffi_abi;
+
+#endif /* LIBFFI_ASM */
+
+/* ---- Definitions for closures ----------------------------------------- */
+
+#define FFI_CLOSURES 1
+#define FFI_TRAMPOLINE_SIZE 24
+#define FFI_NATIVE_RAW_API 0
+#define FFI_EXTRA_CIF_FIELDS unsigned riscv_nfixedargs; unsigned riscv_unused;
+#define FFI_TARGET_SPECIFIC_VARIADIC
+
+#endif
+
diff --git a/src/riscv/sysv.S b/src/riscv/sysv.S
new file mode 100644
index 0000000..2d09865
--- /dev/null
+++ b/src/riscv/sysv.S
@@ -0,0 +1,214 @@
+/* -----------------------------------------------------------------------
+ ffi.c - Copyright (c) 2015 Michael Knyszek <mknyszek@berkeley.edu>
+ 2015 Andrew Waterman <waterman@cs.berkeley.edu>
+ 2018 Stef O'Rear <sorear2@gmail.com>
+
+ RISC-V Foreign Function Interface
+
+ 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.
+ ----------------------------------------------------------------------- */
+
+#define LIBFFI_ASM
+#include <fficonfig.h>
+#include <ffi.h>
+
+/* Define aliases so that we can handle all ABIs uniformly */
+
+#if __SIZEOF_POINTER__ == 8
+#define PTRS 8
+#define LARG ld
+#define SARG sd
+#else
+#define PTRS 4
+#define LARG lw
+#define SARG sw
+#endif
+
+#if __riscv_float_abi_double
+#define FLTS 8
+#define FLARG fld
+#define FSARG fsd
+#elif __riscv_float_abi_single
+#define FLTS 4
+#define FLARG flw
+#define FSARG fsw
+#else
+#define FLTS 0
+#endif
+
+#define fp s0
+
+ .text
+ .globl ffi_call_asm
+ .type ffi_call_asm, @function
+ .hidden ffi_call_asm
+/*
+ struct call_context {
+ floatreg fa[8];
+ intreg a[8];
+ intreg pad[rv32 ? 2 : 0];
+ intreg save_fp, save_ra;
+ }
+ void ffi_call_asm(size_t *stackargs, struct call_context *regargs,
+ void (*fn)(void));
+*/
+
+#define FRAME_LEN (8 * FLTS + 8 * PTRS + 16)
+
+ffi_call_asm:
+ .cfi_startproc
+
+ /*
+ We are NOT going to set up an ordinary stack frame. In order to pass
+ the stacked args to the called function, we adjust our stack pointer to
+ a0, which is in the _caller's_ alloca area. We establish our own stack
+ frame at the end of the call_context.
+
+ Anything below the arguments will be freed at this point, although we
+ preserve the call_context so that it can be read back in the caller.
+ */
+
+ .cfi_def_cfa 11, FRAME_LEN # interim CFA based on a1
+ SARG fp, FRAME_LEN - 2*PTRS(a1)
+ .cfi_offset 8, -2*PTRS
+ SARG ra, FRAME_LEN - 1*PTRS(a1)
+ .cfi_offset 1, -1*PTRS
+
+ addi fp, a1, FRAME_LEN
+ mv sp, a0
+ .cfi_def_cfa 8, 0 # our frame is fully set up
+
+ # Load arguments
+ mv t1, a2
+
+#if FLTS
+ FLARG fa0, -FRAME_LEN+0*FLTS(fp)
+ FLARG fa1, -FRAME_LEN+1*FLTS(fp)
+ FLARG fa2, -FRAME_LEN+2*FLTS(fp)
+ FLARG fa3, -FRAME_LEN+3*FLTS(fp)
+ FLARG fa4, -FRAME_LEN+4*FLTS(fp)
+ FLARG fa5, -FRAME_LEN+5*FLTS(fp)
+ FLARG fa6, -FRAME_LEN+6*FLTS(fp)
+ FLARG fa7, -FRAME_LEN+7*FLTS(fp)
+#endif
+
+ LARG a0, -FRAME_LEN+8*FLTS+0*PTRS(fp)
+ LARG a1, -FRAME_LEN+8*FLTS+1*PTRS(fp)
+ LARG a2, -FRAME_LEN+8*FLTS+2*PTRS(fp)
+ LARG a3, -FRAME_LEN+8*FLTS+3*PTRS(fp)
+ LARG a4, -FRAME_LEN+8*FLTS+4*PTRS(fp)
+ LARG a5, -FRAME_LEN+8*FLTS+5*PTRS(fp)
+ LARG a6, -FRAME_LEN+8*FLTS+6*PTRS(fp)
+ LARG a7, -FRAME_LEN+8*FLTS+7*PTRS(fp)
+
+ /* Call */
+ jalr t1
+
+ /* Save return values - only a0/a1 (fa0/fa1) are used */
+#if FLTS
+ FSARG fa0, -FRAME_LEN+0*FLTS(fp)
+ FSARG fa1, -FRAME_LEN+1*FLTS(fp)
+#endif
+
+ SARG a0, -FRAME_LEN+8*FLTS+0*PTRS(fp)
+ SARG a1, -FRAME_LEN+8*FLTS+1*PTRS(fp)
+
+ /* Restore and return */
+ addi sp, fp, -FRAME_LEN
+ .cfi_def_cfa 2, FRAME_LEN
+ LARG ra, -1*PTRS(fp)
+ .cfi_restore 1
+ LARG fp, -2*PTRS(fp)
+ .cfi_restore 8
+ ret
+ .cfi_endproc
+ .size ffi_call_asm, .-ffi_call_asm
+
+
+/*
+ ffi_closure_asm. Expects address of the passed-in ffi_closure in t1.
+ void ffi_closure_inner(size_t *stackargs, struct call_context *regargs,
+ ffi_closure *closure);
+*/
+
+ .globl ffi_closure_asm
+ .hidden ffi_closure_asm
+ .type ffi_closure_asm, @function
+ffi_closure_asm:
+ .cfi_startproc
+
+ addi sp, sp, -FRAME_LEN
+ .cfi_def_cfa_offset FRAME_LEN
+
+ /* make a frame */
+ SARG fp, FRAME_LEN - 2*PTRS(sp)
+ .cfi_offset 8, -2*PTRS
+ SARG ra, FRAME_LEN - 1*PTRS(sp)
+ .cfi_offset 1, -1*PTRS
+ addi fp, sp, FRAME_LEN
+
+ /* save arguments */
+#if FLTS
+ FSARG fa0, 0*FLTS(sp)
+ FSARG fa1, 1*FLTS(sp)
+ FSARG fa2, 2*FLTS(sp)
+ FSARG fa3, 3*FLTS(sp)
+ FSARG fa4, 4*FLTS(sp)
+ FSARG fa5, 5*FLTS(sp)
+ FSARG fa6, 6*FLTS(sp)
+ FSARG fa7, 7*FLTS(sp)
+#endif
+
+ SARG a0, 8*FLTS+0*PTRS(sp)
+ SARG a1, 8*FLTS+1*PTRS(sp)
+ SARG a2, 8*FLTS+2*PTRS(sp)
+ SARG a3, 8*FLTS+3*PTRS(sp)
+ SARG a4, 8*FLTS+4*PTRS(sp)
+ SARG a5, 8*FLTS+5*PTRS(sp)
+ SARG a6, 8*FLTS+6*PTRS(sp)
+ SARG a7, 8*FLTS+7*PTRS(sp)
+
+ /* enter C */
+ addi a0, sp, FRAME_LEN
+ mv a1, sp
+ mv a2, t1
+
+ call ffi_closure_inner
+
+ /* return values */
+#if FLTS
+ FLARG fa0, 0*FLTS(sp)
+ FLARG fa1, 1*FLTS(sp)
+#endif
+
+ LARG a0, 8*FLTS+0*PTRS(sp)
+ LARG a1, 8*FLTS+1*PTRS(sp)
+
+ /* restore and return */
+ LARG ra, FRAME_LEN-1*PTRS(sp)
+ .cfi_restore 1
+ LARG fp, FRAME_LEN-2*PTRS(sp)
+ .cfi_restore 8
+ addi sp, sp, FRAME_LEN
+ .cfi_def_cfa_offset 0
+ ret
+ .cfi_endproc
+ .size ffi_closure_asm, .-ffi_closure_asm