NASM 2.08rc9
[nasm/avx512.git] / output / outmacho64.c
blobba785faab1d204eec77dda79f7238eab580b01bc
1 /* ----------------------------------------------------------------------- *
2 *
3 * Copyright 1996-2009 The NASM Authors - All Rights Reserved
4 * See the file AUTHORS included with the NASM distribution for
5 * the specific copyright holders.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following
9 * conditions are met:
11 * * Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * * Redistributions in binary form must reproduce the above
14 * copyright notice, this list of conditions and the following
15 * disclaimer in the documentation and/or other materials provided
16 * with the distribution.
18 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
19 * CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
20 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
21 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
22 * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
23 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
24 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
25 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
26 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
29 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
30 * EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32 * ----------------------------------------------------------------------- */
35 * outmacho64.c output routines for the Netwide Assembler to produce
36 * NeXTstep/OpenStep/Rhapsody/Darwin/MacOS X (x86_64) object files
39 /* Most of this file is, like Mach-O itself, based on a.out. For more
40 * guidelines see outaout.c. */
42 #include "compiler.h"
44 #include <stdio.h>
45 #include <stdlib.h>
46 #include <string.h>
47 #include <ctype.h>
48 #include <inttypes.h>
50 #include "nasm.h"
51 #include "nasmlib.h"
52 #include "saa.h"
53 #include "raa.h"
54 #include "output/outform.h"
55 #include "output/outlib.h"
57 #if defined(OF_MACHO64)
59 /* Mach-O in-file header structure sizes */
60 #define MACHO_HEADER64_SIZE (32)
61 #define MACHO_SEGCMD64_SIZE (72)
62 #define MACHO_SECTCMD64_SIZE (80)
63 #define MACHO_SYMCMD_SIZE (24)
64 #define MACHO_NLIST64_SIZE (16)
65 #define MACHO_RELINFO64_SIZE (8)
67 /* Mach-O file header values */
68 #define MH_MAGIC_64 (0xfeedfacf)
69 #define CPU_TYPE_X86_64 (0x01000007) /* x86-64 platform */
70 #define CPU_SUBTYPE_I386_ALL (3) /* all-x86 compatible */
71 #define MH_OBJECT (0x1) /* object file */
73 #define LC_SEGMENT_64 (0x19) /* segment load command */
74 #define LC_SYMTAB (0x2) /* symbol table load command */
76 #define VM_PROT_NONE (0x00)
77 #define VM_PROT_READ (0x01)
78 #define VM_PROT_WRITE (0x02)
79 #define VM_PROT_EXECUTE (0x04)
81 #define VM_PROT_DEFAULT (VM_PROT_READ | VM_PROT_WRITE | VM_PROT_EXECUTE)
82 #define VM_PROT_ALL (VM_PROT_READ | VM_PROT_WRITE | VM_PROT_EXECUTE)
84 struct section {
85 /* nasm internal data */
86 struct section *next;
87 struct SAA *data;
88 int32_t index;
89 struct reloc *relocs;
90 int align;
92 /* data that goes into the file */
93 char sectname[16]; /* what this section is called */
94 char segname[16]; /* segment this section will be in */
95 uint64_t addr; /* in-memory address (subject to alignment) */
96 uint64_t size; /* in-memory and -file size */
97 uint32_t nreloc; /* relocation entry count */
98 uint32_t flags; /* type and attributes (masked) */
99 uint32_t extreloc; /* external relocations */
102 #define SECTION_TYPE 0x000000ff /* section type mask */
104 #define S_REGULAR (0x0) /* standard section */
105 #define S_ZEROFILL (0x1) /* zerofill, in-memory only */
107 #define SECTION_ATTRIBUTES_SYS 0x00ffff00 /* system setable attributes */
108 #define S_ATTR_SOME_INSTRUCTIONS 0x00000400 /* section contains some
109 machine instructions */
110 #define S_ATTR_EXT_RELOC 0x00000200 /* section has external
111 relocation entries */
112 #define S_ATTR_LOC_RELOC 0x00000100 /* section has local
113 relocation entries */
114 #define S_ATTR_PURE_INSTRUCTIONS 0x80000000 /* section uses pure
115 machine instructions */
117 static struct sectmap {
118 const char *nasmsect;
119 const char *segname;
120 const char *sectname;
121 const int32_t flags;
122 } sectmap[] = {
123 {".text", "__TEXT", "__text", S_REGULAR|S_ATTR_SOME_INSTRUCTIONS|S_ATTR_PURE_INSTRUCTIONS},
124 {".data", "__DATA", "__data", S_REGULAR},
125 {".rodata", "__DATA", "__const", S_REGULAR},
126 {".bss", "__DATA", "__bss", S_ZEROFILL},
127 {NULL, NULL, NULL, 0}
130 struct reloc {
131 /* nasm internal data */
132 struct reloc *next;
134 /* data that goes into the file */
135 int32_t addr; /* op's offset in section */
136 uint32_t snum:24, /* contains symbol index if
137 ** ext otherwise in-file
138 ** section number */
139 pcrel:1, /* relative relocation */
140 length:2, /* 0=byte, 1=word, 2=int32_t, 3=int64_t */
141 ext:1, /* external symbol referenced */
142 type:4; /* reloc type */
145 #define R_ABS 0 /* absolute relocation */
146 #define R_SCATTERED 0x80000000 /* reloc entry is scattered if
147 ** highest bit == 1 */
149 struct symbol {
150 /* nasm internal data */
151 struct symbol *next; /* next symbol in the list */
152 char *name; /* name of this symbol */
153 int32_t initial_snum; /* symbol number used above in
154 reloc */
155 int32_t snum; /* true snum for reloc */
157 /* data that goes into the file */
158 uint32_t strx; /* string table index */
159 uint8_t type; /* symbol type */
160 uint8_t sect; /* NO_SECT or section number */
161 uint16_t desc; /* for stab debugging, 0 for us */
162 uint64_t value; /* offset of symbol in section */
165 /* symbol type bits */
166 #define N_EXT 0x01 /* global or external symbol */
168 #define N_UNDF 0x0 /* undefined symbol | n_sect == */
169 #define N_ABS 0x2 /* absolute symbol | NO_SECT */
170 #define N_SECT 0xe /* defined symbol, n_sect holds
171 ** section number */
173 #define N_TYPE 0x0e /* type bit mask */
175 #define DEFAULT_SECTION_ALIGNMENT 0 /* byte (i.e. no) alignment */
177 /* special section number values */
178 #define NO_SECT 0 /* no section, invalid */
179 #define MAX_SECT 255 /* maximum number of sections */
181 static struct section *sects, **sectstail;
182 static struct symbol *syms, **symstail;
183 static uint32_t nsyms;
185 /* These variables are set by macho_layout_symbols() to organize
186 the symbol table and string table in order the dynamic linker
187 expects. They are then used in macho_write() to put out the
188 symbols and strings in that order.
190 The order of the symbol table is:
191 local symbols
192 defined external symbols (sorted by name)
193 undefined external symbols (sorted by name)
195 The order of the string table is:
196 strings for external symbols
197 strings for local symbols
199 static uint32_t ilocalsym = 0;
200 static uint32_t iextdefsym = 0;
201 static uint32_t iundefsym = 0;
202 static uint32_t nlocalsym;
203 static uint32_t nextdefsym;
204 static uint32_t nundefsym;
205 static struct symbol **extdefsyms = NULL;
206 static struct symbol **undefsyms = NULL;
208 static struct RAA *extsyms;
209 static struct SAA *strs;
210 static uint32_t strslen;
212 extern struct ofmt of_macho64;
214 /* Global file information. This should be cleaned up into either
215 a structure or as function arguments. */
216 uint32_t head_ncmds64 = 0;
217 uint32_t head_sizeofcmds64 = 0;
218 uint64_t seg_filesize64 = 0;
219 uint64_t seg_vmsize64 = 0;
220 uint32_t seg_nsects64 = 0;
221 uint64_t rel_padcnt64 = 0;
224 #define xstrncpy(xdst, xsrc) \
225 memset(xdst, '\0', sizeof(xdst)); /* zero out whole buffer */ \
226 strncpy(xdst, xsrc, sizeof(xdst)); /* copy over string */ \
227 xdst[sizeof(xdst) - 1] = '\0'; /* proper null-termination */
229 #define alignint32_t(x) \
230 ALIGN(x, sizeof(int32_t)) /* align x to int32_t boundary */
232 #define alignint64_t(x) \
233 ALIGN(x, sizeof(int64_t)) /* align x to int64_t boundary */
235 static void debug_reloc (struct reloc *);
236 static void debug_section_relocs (struct section *) _unused;
238 static int exact_log2 (uint32_t align)
240 if (align == 0) {
241 return 0;
242 } else if (align & (align-1)) {
243 return -1; /* Not a power of 2 */
244 } else {
245 #ifdef HAVE_GNUC_4
246 return __builtin_ctzl (align);
247 #else
248 uint32_t result = 0;
250 /* We know exactly one bit is set at this point. */
251 if (align & 0xffff0000)
252 result |= 16;
253 if (align & 0xff00ff00)
254 result |= 8;
255 if (align & 0xf0f0f0f0)
256 result |= 4;
257 if (align & 0xcccccccc)
258 result |= 2;
259 if (align & 0xaaaaaaaa)
260 result |= 1;
262 return result;
263 #endif
267 static struct section *get_section_by_name(const char *segname,
268 const char *sectname)
270 struct section *s;
272 for (s = sects; s != NULL; s = s->next)
273 if (!strcmp(s->segname, segname) && !strcmp(s->sectname, sectname))
274 break;
276 return s;
279 static struct section *get_section_by_index(const int32_t index)
281 struct section *s;
283 for (s = sects; s != NULL; s = s->next)
284 if (index == s->index)
285 break;
287 return s;
290 static int32_t get_section_index_by_name(const char *segname,
291 const char *sectname)
293 struct section *s;
295 for (s = sects; s != NULL; s = s->next)
296 if (!strcmp(s->segname, segname) && !strcmp(s->sectname, sectname))
297 return s->index;
299 return -1;
302 static char *get_section_name_by_index(const int32_t index)
304 struct section *s;
306 for (s = sects; s != NULL; s = s->next)
307 if (index == s->index)
308 return s->sectname;
310 return NULL;
313 static uint8_t get_section_fileindex_by_index(const int32_t index)
315 struct section *s;
316 uint8_t i = 1;
318 for (s = sects; s != NULL && i < MAX_SECT; s = s->next, ++i)
319 if (index == s->index)
320 return i;
322 if (i == MAX_SECT)
323 nasm_error(ERR_WARNING,
324 "too many sections (>255) - clipped by fileindex");
326 return NO_SECT;
329 static struct symbol *get_closest_section_symbol_by_offset(uint8_t fileindex, int64_t offset)
331 struct symbol *sym;
333 for (sym = syms; sym != NULL; sym = sym->next) {
334 if ((sym->sect != NO_SECT) &&
335 (sym->sect == fileindex) &&
336 ((int64_t)sym->value >= offset))
337 return sym;
340 return NULL;
345 * Special section numbers which are used to define Mach-O special
346 * symbols, which can be used with WRT to provide PIC relocation
347 * types.
349 static int32_t macho_gotpcrel_sect;
351 static void macho_init(void)
353 char zero = 0;
355 maxbits = 64;
357 sects = NULL;
358 sectstail = &sects;
360 syms = NULL;
361 symstail = &syms;
362 nsyms = 0;
363 nlocalsym = 0;
364 nextdefsym = 0;
365 nundefsym = 0;
367 extsyms = raa_init();
368 strs = saa_init(1L);
370 /* string table starts with a zero byte - don't ask why */
371 saa_wbytes(strs, &zero, sizeof(char));
372 strslen = 1;
374 /* add special symbol for ..gotpcrel */
375 macho_gotpcrel_sect = seg_alloc();
376 macho_gotpcrel_sect++;
377 define_label("..gotpcrel", macho_gotpcrel_sect, 0L, NULL, false, false);
380 static void sect_write(struct section *sect,
381 const uint8_t *data, uint32_t len)
383 saa_wbytes(sect->data, data, len);
384 sect->size += len;
387 static int32_t add_reloc(struct section *sect, int32_t section,
388 int pcrel, int bytes, int64_t reloff)
390 struct reloc *r;
391 struct symbol *sym;
392 int32_t fi;
393 int32_t adjustment = 0;
395 /* NeXT as puts relocs in reversed order (address-wise) into the
396 ** files, so we do the same, doesn't seem to make much of a
397 ** difference either way */
398 r = nasm_malloc(sizeof(struct reloc));
399 r->next = sect->relocs;
400 sect->relocs = r;
402 /* the current end of the section will be the symbol's address for
403 ** now, might have to be fixed by macho_fixup_relocs() later on. make
404 ** sure we don't make the symbol scattered by setting the highest
405 ** bit by accident */
406 r->addr = sect->size & ~R_SCATTERED;
407 r->ext = 1;
408 r->pcrel = (pcrel ? 1 : 0);
410 /* match byte count 1, 2, 4, 8 to length codes 0, 1, 2, 3 respectively */
411 switch(bytes){
412 case 1:
413 r->length = 0;
414 break;
415 case 2:
416 r->length = 1;
417 break;
418 case 4:
419 r->length = 2;
420 break;
421 case 8:
422 r->length = 3;
423 break;
424 default:
425 break;
428 /* set default relocation values */
429 r->type = 0; // X86_64_RELOC_UNSIGNED
430 r->snum = R_ABS; // Absolute Symbol (indicates no relocation)
432 /* absolute relocation */
433 if (pcrel == 0) {
435 /* intra-section */
436 if (section == NO_SEG) {
437 // r->snum = R_ABS; // Set above
439 /* inter-section */
440 } else {
441 fi = get_section_fileindex_by_index(section);
443 /* external */
444 if (fi == NO_SECT) {
445 r->snum = raa_read(extsyms, section);
447 /* local */
448 } else {
449 sym = get_closest_section_symbol_by_offset(fi, reloff);
450 r->snum = sym->initial_snum;
451 adjustment = sym->value;
455 /* relative relocation */
456 } else if (pcrel == 1) {
458 /* intra-section */
459 if (section == NO_SEG) {
460 r->type = 1; // X86_64_RELOC_SIGNED
462 /* inter-section */
463 } else {
464 r->type = 2; // X86_64_RELOC_BRANCH
465 fi = get_section_fileindex_by_index(section);
467 /* external */
468 if (fi == NO_SECT) {
469 sect->extreloc = 1;
470 r->snum = raa_read(extsyms, section);
472 /* local */
473 } else {
474 sym = get_closest_section_symbol_by_offset(fi, reloff);
475 r->snum = sym->initial_snum;
476 adjustment = sym->value;
480 /* subtractor */
481 } else if (pcrel == 2) {
482 r->pcrel = 0;
483 r->type = 5; // X86_64_RELOC_SUBTRACTOR
485 /* gotpcrel */
486 } else if (pcrel == 3) {
487 r->type = 4; // X86_64_RELOC_GOT
488 r->snum = macho_gotpcrel_sect;
490 /* gotpcrel MOVQ load */
491 } else if (pcrel == 4) {
492 r->type = 3; // X86_64_RELOC_GOT_LOAD
493 r->snum = macho_gotpcrel_sect;
496 ++sect->nreloc;
498 return adjustment;
501 static void macho_output(int32_t secto, const void *data,
502 enum out_type type, uint64_t size,
503 int32_t section, int32_t wrt)
505 struct section *s, *sbss;
506 int64_t addr;
507 uint8_t mydata[16], *p, gotload;
509 if (secto == NO_SEG) {
510 if (type != OUT_RESERVE)
511 nasm_error(ERR_NONFATAL, "attempt to assemble code in "
512 "[ABSOLUTE] space");
514 return;
517 s = get_section_by_index(secto);
519 if (s == NULL) {
520 nasm_error(ERR_WARNING, "attempt to assemble code in"
521 " section %d: defaulting to `.text'", secto);
522 s = get_section_by_name("__TEXT", "__text");
524 /* should never happen */
525 if (s == NULL)
526 nasm_error(ERR_PANIC, "text section not found");
529 sbss = get_section_by_name("__DATA", "__bss");
531 if (s == sbss && type != OUT_RESERVE) {
532 nasm_error(ERR_WARNING, "attempt to initialize memory in the"
533 " BSS section: ignored");
534 s->size += realsize(type, size);
535 return;
538 switch (type) {
539 case OUT_RESERVE:
540 if (s != sbss) {
541 nasm_error(ERR_WARNING, "uninitialized space declared in"
542 " %s section: zeroing",
543 get_section_name_by_index(secto));
545 sect_write(s, NULL, size);
546 } else
547 s->size += size;
549 break;
551 case OUT_RAWDATA:
552 if (section != NO_SEG)
553 nasm_error(ERR_PANIC, "OUT_RAWDATA with other than NO_SEG");
555 sect_write(s, data, size);
556 break;
558 case OUT_ADDRESS:
559 addr = *(int64_t *)data;
560 if (section != NO_SEG) {
561 if (section % 2) {
562 nasm_error(ERR_NONFATAL, "Mach-O format does not support"
563 " section base references");
564 } else {
565 if (wrt == NO_SEG) {
566 if (size < 8) {
567 nasm_error(ERR_NONFATAL, "Mach-O 64-bit format does not support"
568 " 32-bit absolute addresses");
570 Seemingly, Mach-O's X86_64_RELOC_SUBTRACTOR would require
571 pre-determined knowledge of where the image base would be,
572 making it impractical for use in intermediate object files
574 } else {
575 addr -= add_reloc(s, section, 0, size, addr); // X86_64_RELOC_UNSIGNED
577 } else {
578 nasm_error(ERR_NONFATAL, "Mach-O format does not support"
579 " this use of WRT");
584 p = mydata;
585 WRITEADDR(p, addr, size);
586 sect_write(s, mydata, size);
587 break;
589 case OUT_REL2ADR:
590 p = mydata;
591 WRITESHORT(p, *(int64_t *)data);
593 if (section == secto)
594 nasm_error(ERR_PANIC, "intra-section OUT_REL2ADR");
596 if (section == NO_SEG) {
597 /* Do nothing */
598 } else if (section % 2) {
599 nasm_error(ERR_NONFATAL, "Mach-O format does not support"
600 " section base references");
601 } else {
602 nasm_error(ERR_NONFATAL, "Unsupported non-32-bit"
603 " Macho-O relocation [2]");
606 sect_write(s, mydata, 2L);
607 break;
609 case OUT_REL4ADR:
610 p = mydata;
611 WRITELONG(p, *(int64_t *)data);
613 if (section == secto)
614 nasm_error(ERR_PANIC, "intra-section OUT_REL4ADR");
616 if (section != NO_SEG && section % 2) {
617 nasm_error(ERR_NONFATAL, "Mach-O format does not support"
618 " section base references");
619 } else {
620 if (wrt == NO_SEG) {
621 *mydata -= add_reloc(s, section, 1, 4, (int64_t)*mydata); // X86_64_RELOC_SIGNED/BRANCH
622 } else if (wrt == macho_gotpcrel_sect) {
623 if (s->data->datalen > 1) {
624 saa_fread(s->data, s->data->datalen-2, &gotload, 1); // Retrieve Instruction Opcode
625 } else {
626 gotload = 0;
628 if (gotload == 0x8B) { // Check for MOVQ Opcode
629 *mydata -= add_reloc(s, section, 4, 4, (int64_t)*mydata); // X86_64_GOT_LOAD (MOVQ load)
630 } else {
631 *mydata -= add_reloc(s, section, 3, 4, (int64_t)*mydata); // X86_64_GOT
633 } else {
634 nasm_error(ERR_NONFATAL, "Mach-O format does not support"
635 " this use of WRT");
636 wrt = NO_SEG; /* we can at least _try_ to continue */
640 sect_write(s, mydata, 4L);
641 break;
643 default:
644 nasm_error(ERR_PANIC, "unknown output type?");
645 break;
649 static int32_t macho_section(char *name, int pass, int *bits)
651 int32_t index, originalIndex;
652 char *sectionAttributes;
653 struct sectmap *sm;
654 struct section *s;
656 (void)pass;
658 /* Default to 64 bits. */
659 if (!name) {
660 *bits = 64;
661 name = ".text";
662 sectionAttributes = NULL;
663 } else {
664 sectionAttributes = name;
665 name = nasm_strsep(&sectionAttributes, " \t");
668 for (sm = sectmap; sm->nasmsect != NULL; ++sm) {
669 /* make lookup into section name translation table */
670 if (!strcmp(name, sm->nasmsect)) {
671 char *currentAttribute;
673 /* try to find section with that name */
674 originalIndex = index = get_section_index_by_name(sm->segname,
675 sm->sectname);
677 /* create it if it doesn't exist yet */
678 if (index == -1) {
679 s = *sectstail = nasm_malloc(sizeof(struct section));
680 s->next = NULL;
681 sectstail = &s->next;
683 s->data = saa_init(1L);
684 s->index = seg_alloc();
685 s->relocs = NULL;
686 s->align = -1;
688 xstrncpy(s->segname, sm->segname);
689 xstrncpy(s->sectname, sm->sectname);
690 s->size = 0;
691 s->nreloc = 0;
692 s->flags = sm->flags;
694 index = s->index;
695 } else {
696 s = get_section_by_index(index);
699 while ((NULL != sectionAttributes)
700 && (currentAttribute = nasm_strsep(&sectionAttributes, " \t"))) {
701 if (0 != *currentAttribute) {
702 if (!nasm_strnicmp("align=", currentAttribute, 6)) {
703 char *end;
704 int newAlignment, value;
706 value = strtoul(currentAttribute + 6, (char**)&end, 0);
707 newAlignment = exact_log2(value);
709 if (0 != *end) {
710 nasm_error(ERR_PANIC,
711 "unknown or missing alignment value \"%s\" "
712 "specified for section \"%s\"",
713 currentAttribute + 6,
714 name);
715 return NO_SEG;
716 } else if (0 > newAlignment) {
717 nasm_error(ERR_PANIC,
718 "alignment of %d (for section \"%s\") is not "
719 "a power of two",
720 value,
721 name);
722 return NO_SEG;
725 if ((-1 != originalIndex)
726 && (s->align != newAlignment)
727 && (s->align != -1)) {
728 nasm_error(ERR_PANIC,
729 "section \"%s\" has already been specified "
730 "with alignment %d, conflicts with new "
731 "alignment of %d",
732 name,
733 (1 << s->align),
734 value);
735 return NO_SEG;
738 s->align = newAlignment;
739 } else if (!nasm_stricmp("data", currentAttribute)) {
740 /* Do nothing; 'data' is implicit */
741 } else {
742 nasm_error(ERR_PANIC,
743 "unknown section attribute %s for section %s",
744 currentAttribute,
745 name);
746 return NO_SEG;
751 return index;
755 nasm_error(ERR_PANIC, "invalid section name %s", name);
756 return NO_SEG;
759 static void macho_symdef(char *name, int32_t section, int64_t offset,
760 int is_global, char *special)
762 struct symbol *sym;
764 if (special) {
765 nasm_error(ERR_NONFATAL, "The Mach-O output format does "
766 "not support any special symbol types");
767 return;
770 if (is_global == 3) {
771 nasm_error(ERR_NONFATAL, "The Mach-O format does not "
772 "(yet) support forward reference fixups.");
773 return;
776 if (name[0] == '.' && name[1] == '.' && name[2] != '@') {
778 * This is a NASM special symbol. We never allow it into
779 * the Macho-O symbol table, even if it's a valid one. If it
780 * _isn't_ a valid one, we should barf immediately.
782 if (strcmp(name, "..gotpcrel"))
783 nasm_error(ERR_NONFATAL, "unrecognized special symbol `%s'", name);
784 return;
787 sym = *symstail = nasm_malloc(sizeof(struct symbol));
788 sym->next = NULL;
789 symstail = &sym->next;
791 sym->name = name;
792 sym->strx = strslen;
793 sym->type = 0;
794 sym->desc = 0;
795 sym->value = offset;
796 sym->initial_snum = -1;
798 /* external and common symbols get N_EXT */
799 if (is_global != 0) {
800 sym->type |= N_EXT;
803 if (section == NO_SEG) {
804 /* symbols in no section get absolute */
805 sym->type |= N_ABS;
806 sym->sect = NO_SECT;
807 } else {
808 sym->type |= N_SECT;
810 /* get the in-file index of the section the symbol was defined in */
811 sym->sect = get_section_fileindex_by_index(section);
813 /* track the initially allocated symbol number for use in future fix-ups */
814 sym->initial_snum = nsyms;
816 if (sym->sect == NO_SECT) {
818 /* remember symbol number of references to external
819 ** symbols, this works because every external symbol gets
820 ** its own section number allocated internally by nasm and
821 ** can so be used as a key */
822 extsyms = raa_write(extsyms, section, nsyms);
824 switch (is_global) {
825 case 1:
826 case 2:
827 /* there isn't actually a difference between global
828 ** and common symbols, both even have their size in
829 ** sym->value */
830 sym->type = N_EXT;
831 break;
833 default:
834 /* give an error on unfound section if it's not an
835 ** external or common symbol (assemble_file() does a
836 ** seg_alloc() on every call for them) */
837 nasm_error(ERR_PANIC, "in-file index for section %d not found",
838 section);
842 ++nsyms;
845 static int32_t macho_segbase(int32_t section)
847 return section;
850 static void macho_filename(char *inname, char *outname)
852 standard_extension(inname, outname, ".o");
855 extern macros_t macho_stdmac[];
857 /* Comparison function for qsort symbol layout. */
858 static int layout_compare (const struct symbol **s1,
859 const struct symbol **s2)
861 return (strcmp ((*s1)->name, (*s2)->name));
864 /* The native assembler does a few things in a similar function
866 * Remove temporary labels
867 * Sort symbols according to local, external, undefined (by name)
868 * Order the string table
870 We do not remove temporary labels right now.
872 numsyms is the total number of symbols we have. strtabsize is the
873 number entries in the string table. */
875 static void macho_layout_symbols (uint32_t *numsyms,
876 uint32_t *strtabsize)
878 struct symbol *sym, **symp;
879 uint32_t i,j;
881 *numsyms = 0;
882 *strtabsize = sizeof (char);
884 symp = &syms;
886 while ((sym = *symp)) {
887 /* Undefined symbols are now external. */
888 if (sym->type == N_UNDF)
889 sym->type |= N_EXT;
891 if ((sym->type & N_EXT) == 0) {
892 sym->snum = *numsyms;
893 *numsyms = *numsyms + 1;
894 nlocalsym++;
896 else {
897 if ((sym->type & N_TYPE) != N_UNDF) {
898 nextdefsym++;
899 } else {
900 nundefsym++;
903 /* If we handle debug info we'll want
904 to check for it here instead of just
905 adding the symbol to the string table. */
906 sym->strx = *strtabsize;
907 saa_wbytes (strs, sym->name, (int32_t)(strlen(sym->name) + 1));
908 *strtabsize += strlen(sym->name) + 1;
910 symp = &(sym->next);
913 /* Next, sort the symbols. Most of this code is a direct translation from
914 the Apple cctools symbol layout. We need to keep compatibility with that. */
915 /* Set the indexes for symbol groups into the symbol table */
916 ilocalsym = 0;
917 iextdefsym = nlocalsym;
918 iundefsym = nlocalsym + nextdefsym;
920 /* allocate arrays for sorting externals by name */
921 extdefsyms = nasm_malloc(nextdefsym * sizeof(struct symbol *));
922 undefsyms = nasm_malloc(nundefsym * sizeof(struct symbol *));
924 i = 0;
925 j = 0;
927 symp = &syms;
929 while ((sym = *symp)) {
931 if((sym->type & N_EXT) == 0) {
932 sym->strx = *strtabsize;
933 saa_wbytes (strs, sym->name, (int32_t)(strlen (sym->name) + 1));
934 *strtabsize += strlen(sym->name) + 1;
936 else {
937 if((sym->type & N_TYPE) != N_UNDF) {
938 extdefsyms[i++] = sym;
939 } else {
940 undefsyms[j++] = sym;
943 symp = &(sym->next);
946 qsort(extdefsyms, nextdefsym, sizeof(struct symbol *),
947 (int (*)(const void *, const void *))layout_compare);
948 qsort(undefsyms, nundefsym, sizeof(struct symbol *),
949 (int (*)(const void *, const void *))layout_compare);
951 for(i = 0; i < nextdefsym; i++) {
952 extdefsyms[i]->snum = *numsyms;
953 *numsyms += 1;
955 for(j = 0; j < nundefsym; j++) {
956 undefsyms[j]->snum = *numsyms;
957 *numsyms += 1;
961 /* Calculate some values we'll need for writing later. */
963 static void macho_calculate_sizes (void)
965 struct section *s;
967 /* count sections and calculate in-memory and in-file offsets */
968 for (s = sects; s != NULL; s = s->next) {
969 uint64_t pad = 0;
971 /* zerofill sections aren't actually written to the file */
972 if ((s->flags & SECTION_TYPE) != S_ZEROFILL)
973 seg_filesize64 += s->size;
975 /* recalculate segment address based on alignment and vm size */
976 s->addr = seg_vmsize64;
977 /* we need section alignment to calculate final section address */
978 if (s->align == -1)
979 s->align = DEFAULT_SECTION_ALIGNMENT;
980 if(s->align) {
981 uint64_t newaddr = ALIGN(s->addr, 1 << s->align);
982 pad = newaddr - s->addr;
983 s->addr = newaddr;
986 seg_vmsize64 += s->size + pad;
987 ++seg_nsects64;
990 /* calculate size of all headers, load commands and sections to
991 ** get a pointer to the start of all the raw data */
992 if (seg_nsects64 > 0) {
993 ++head_ncmds64;
994 head_sizeofcmds64 +=
995 MACHO_SEGCMD64_SIZE + seg_nsects64 * MACHO_SECTCMD64_SIZE;
998 if (nsyms > 0) {
999 ++head_ncmds64;
1000 head_sizeofcmds64 += MACHO_SYMCMD_SIZE;
1004 /* Write out the header information for the file. */
1006 static void macho_write_header (void)
1008 fwriteint32_t(MH_MAGIC_64, ofile); /* magic */
1009 fwriteint32_t(CPU_TYPE_X86_64, ofile); /* CPU type */
1010 fwriteint32_t(CPU_SUBTYPE_I386_ALL, ofile); /* CPU subtype */
1011 fwriteint32_t(MH_OBJECT, ofile); /* Mach-O file type */
1012 fwriteint32_t(head_ncmds64, ofile); /* number of load commands */
1013 fwriteint32_t(head_sizeofcmds64, ofile); /* size of load commands */
1014 fwriteint32_t(0, ofile); /* no flags */
1015 fwriteint32_t(0, ofile); /* reserved for future use */
1018 /* Write out the segment load command at offset. */
1020 static uint32_t macho_write_segment (uint64_t offset)
1022 uint64_t rel_base = alignint64_t (offset + seg_filesize64);
1023 uint32_t s_reloff = 0;
1024 struct section *s;
1026 fwriteint32_t(LC_SEGMENT_64, ofile); /* cmd == LC_SEGMENT_64 */
1028 /* size of load command including section load commands */
1029 fwriteint32_t(MACHO_SEGCMD64_SIZE + seg_nsects64 *
1030 MACHO_SECTCMD64_SIZE, ofile);
1032 /* in an MH_OBJECT file all sections are in one unnamed (name
1033 ** all zeros) segment */
1034 fwritezero(16, ofile);
1035 fwriteint64_t(0, ofile); /* in-memory offset */
1036 fwriteint64_t(seg_vmsize64, ofile); /* in-memory size */
1037 fwriteint64_t(offset, ofile); /* in-file offset to data */
1038 fwriteint64_t(seg_filesize64, ofile); /* in-file size */
1039 fwriteint32_t(VM_PROT_DEFAULT, ofile); /* maximum vm protection */
1040 fwriteint32_t(VM_PROT_DEFAULT, ofile); /* initial vm protection */
1041 fwriteint32_t(seg_nsects64, ofile); /* number of sections */
1042 fwriteint32_t(0, ofile); /* no flags */
1044 /* emit section headers */
1045 for (s = sects; s != NULL; s = s->next) {
1046 fwrite(s->sectname, sizeof(s->sectname), 1, ofile);
1047 fwrite(s->segname, sizeof(s->segname), 1, ofile);
1048 fwriteint64_t(s->addr, ofile);
1049 fwriteint64_t(s->size, ofile);
1051 /* dummy data for zerofill sections or proper values */
1052 if ((s->flags & SECTION_TYPE) != S_ZEROFILL) {
1053 fwriteint32_t(offset, ofile);
1054 /* Write out section alignment, as a power of two.
1055 e.g. 32-bit word alignment would be 2 (2^2 = 4). */
1056 if (s->align == -1)
1057 s->align = DEFAULT_SECTION_ALIGNMENT;
1058 fwriteint32_t(s->align, ofile);
1059 /* To be compatible with cctools as we emit
1060 a zero reloff if we have no relocations. */
1061 fwriteint32_t(s->nreloc ? rel_base + s_reloff : 0, ofile);
1062 fwriteint32_t(s->nreloc, ofile);
1064 offset += s->size;
1065 s_reloff += s->nreloc * MACHO_RELINFO64_SIZE;
1066 } else {
1067 fwriteint32_t(0, ofile);
1068 fwriteint32_t(0, ofile);
1069 fwriteint32_t(0, ofile);
1070 fwriteint32_t(0, ofile);
1073 if (s->nreloc) {
1074 s->flags |= S_ATTR_LOC_RELOC;
1075 if (s->extreloc)
1076 s->flags |= S_ATTR_EXT_RELOC;
1079 fwriteint32_t(s->flags, ofile); /* flags */
1080 fwriteint32_t(0, ofile); /* reserved */
1081 fwriteint32_t(0, ofile); /* reserved */
1083 fwriteint32_t(0, ofile); /* align */
1086 rel_padcnt64 = rel_base - offset;
1087 offset = rel_base + s_reloff;
1089 return offset;
1092 /* For a given chain of relocs r, write out the entire relocation
1093 chain to the object file. */
1095 static void macho_write_relocs (struct reloc *r)
1097 while (r) {
1098 uint32_t word2;
1100 fwriteint32_t(r->addr, ofile); /* reloc offset */
1102 word2 = r->snum;
1103 word2 |= r->pcrel << 24;
1104 word2 |= r->length << 25;
1105 word2 |= r->ext << 27;
1106 word2 |= r->type << 28;
1107 fwriteint32_t(word2, ofile); /* reloc data */
1108 r = r->next;
1112 /* Write out the section data. */
1113 static void macho_write_section (void)
1115 struct section *s, *s2;
1116 struct reloc *r;
1117 uint8_t fi, *p, *q, blk[8];
1118 int32_t len;
1119 int64_t l;
1121 for (s = sects; s != NULL; s = s->next) {
1122 if ((s->flags & SECTION_TYPE) == S_ZEROFILL)
1123 continue;
1125 /* no padding needs to be done to the sections */
1127 /* Like a.out Mach-O references things in the data or bss
1128 * sections by addresses which are actually relative to the
1129 * start of the _text_ section, in the _file_. See outaout.c
1130 * for more information. */
1131 saa_rewind(s->data);
1132 for (r = s->relocs; r != NULL; r = r->next) {
1133 len = (int32_t)r->length << 1;
1134 if(len > 4) len = 8;
1135 saa_fread(s->data, r->addr, blk, len);
1136 p = q = blk;
1137 l = *p++;
1139 /* get offset based on relocation type */
1140 if (r->length > 0) {
1141 l += ((int64_t)*p++) << 8;
1143 if (r->length > 1) {
1144 l += ((int64_t)*p++) << 16;
1145 l += ((int64_t)*p++) << 24;
1148 if (r->length > 2) {
1149 l += ((int64_t)*p++) << 32;
1150 l += ((int64_t)*p++) << 40;
1151 l += ((int64_t)*p++) << 48;
1152 l += ((int64_t)*p++) << 56;
1158 /* If the relocation is internal add to the current section
1159 offset. Otherwise the only value we need is the symbol
1160 offset which we already have. The linker takes care
1161 of the rest of the address. */
1162 if (!r->ext) {
1163 /* generate final address by section address and offset */
1164 for (s2 = sects, fi = 1;
1165 s2 != NULL; s2 = s2->next, fi++) {
1166 if (fi == r->snum) {
1167 l += s2->addr;
1168 break;
1173 /* write new offset back */
1174 if (r->length == 3)
1175 WRITEDLONG(q, l);
1176 else if (r->length == 2)
1177 WRITELONG(q, l);
1178 else if (r->length == 1)
1179 WRITESHORT(q, l);
1180 else
1181 *q++ = l & 0xFF;
1183 saa_fwrite(s->data, r->addr, blk, len);
1186 /* dump the section data to file */
1187 saa_fpwrite(s->data, ofile);
1190 /* pad last section up to reloc entries on int64_t boundary */
1191 fwritezero(rel_padcnt64, ofile);
1193 /* emit relocation entries */
1194 for (s = sects; s != NULL; s = s->next)
1195 macho_write_relocs (s->relocs);
1198 /* Write out the symbol table. We should already have sorted this
1199 before now. */
1200 static void macho_write_symtab (void)
1202 struct symbol *sym;
1203 struct section *s;
1204 int64_t fi;
1205 uint64_t i;
1207 /* we don't need to pad here since MACHO_RELINFO_SIZE == 8 */
1209 for (sym = syms; sym != NULL; sym = sym->next) {
1210 if ((sym->type & N_EXT) == 0) {
1211 fwriteint32_t(sym->strx, ofile); /* string table entry number */
1212 fwrite(&sym->type, 1, 1, ofile); /* symbol type */
1213 fwrite(&sym->sect, 1, 1, ofile); /* section */
1214 fwriteint16_t(sym->desc, ofile); /* description */
1216 /* Fix up the symbol value now that we know the final section
1217 sizes. */
1218 if (((sym->type & N_TYPE) == N_SECT) && (sym->sect != NO_SECT)) {
1219 for (s = sects, fi = 1; s != NULL; s = s->next, fi++) {
1220 if (fi == sym->sect) {
1221 sym->value += s->addr;
1222 break;
1227 fwriteint64_t(sym->value, ofile); /* value (i.e. offset) */
1231 for (i = 0; i < nextdefsym; i++) {
1232 sym = extdefsyms[i];
1233 fwriteint32_t(sym->strx, ofile);
1234 fwrite(&sym->type, 1, 1, ofile); /* symbol type */
1235 fwrite(&sym->sect, 1, 1, ofile); /* section */
1236 fwriteint16_t(sym->desc, ofile); /* description */
1238 /* Fix up the symbol value now that we know the final section
1239 sizes. */
1240 if (((sym->type & N_TYPE) == N_SECT) && (sym->sect != NO_SECT)) {
1241 for (s = sects, fi = 1;
1242 s != NULL && fi < sym->sect; s = s->next, ++fi)
1243 sym->value += s->size;
1246 fwriteint64_t(sym->value, ofile); /* value (i.e. offset) */
1249 for (i = 0; i < nundefsym; i++) {
1250 sym = undefsyms[i];
1251 fwriteint32_t(sym->strx, ofile);
1252 fwrite(&sym->type, 1, 1, ofile); /* symbol type */
1253 fwrite(&sym->sect, 1, 1, ofile); /* section */
1254 fwriteint16_t(sym->desc, ofile); /* description */
1256 // Fix up the symbol value now that we know the final section sizes.
1257 if (((sym->type & N_TYPE) == N_SECT) && (sym->sect != NO_SECT)) {
1258 for (s = sects, fi = 1;
1259 s != NULL && fi < sym->sect; s = s->next, ++fi)
1260 sym->value += s->size;
1263 fwriteint64_t(sym->value, ofile); // value (i.e. offset)
1268 /* Fixup the snum in the relocation entries, we should be
1269 doing this only for externally referenced symbols. */
1270 static void macho_fixup_relocs (struct reloc *r)
1272 struct symbol *sym;
1274 while (r != NULL) {
1275 if (r->ext) {
1276 for (sym = syms; sym != NULL; sym = sym->next) {
1277 if (sym->initial_snum == r->snum) {
1278 r->snum = sym->snum;
1279 break;
1283 r = r->next;
1287 /* Write out the object file. */
1289 static void macho_write (void)
1291 uint64_t offset = 0;
1293 /* mach-o object file structure:
1295 ** mach header
1296 ** uint32_t magic
1297 ** int cpu type
1298 ** int cpu subtype
1299 ** uint32_t mach file type
1300 ** uint32_t number of load commands
1301 ** uint32_t size of all load commands
1302 ** (includes section struct size of segment command)
1303 ** uint32_t flags
1305 ** segment command
1306 ** uint32_t command type == LC_SEGMENT_64
1307 ** uint32_t size of load command
1308 ** (including section load commands)
1309 ** char[16] segment name
1310 ** uint64_t in-memory offset
1311 ** uint64_t in-memory size
1312 ** uint64_t in-file offset to data area
1313 ** uint64_t in-file size
1314 ** (in-memory size excluding zerofill sections)
1315 ** int maximum vm protection
1316 ** int initial vm protection
1317 ** uint32_t number of sections
1318 ** uint32_t flags
1320 ** section commands
1321 ** char[16] section name
1322 ** char[16] segment name
1323 ** uint64_t in-memory offset
1324 ** uint64_t in-memory size
1325 ** uint32_t in-file offset
1326 ** uint32_t alignment
1327 ** (irrelevant in MH_OBJECT)
1328 ** uint32_t in-file offset of relocation entires
1329 ** uint32_t number of relocations
1330 ** uint32_t flags
1331 ** uint32_t reserved
1332 ** uint32_t reserved
1334 ** symbol table command
1335 ** uint32_t command type == LC_SYMTAB
1336 ** uint32_t size of load command
1337 ** uint32_t symbol table offset
1338 ** uint32_t number of symbol table entries
1339 ** uint32_t string table offset
1340 ** uint32_t string table size
1342 ** raw section data
1344 ** padding to int64_t boundary
1346 ** relocation data (struct reloc)
1347 ** int32_t offset
1348 ** uint data (symbolnum, pcrel, length, extern, type)
1350 ** symbol table data (struct nlist)
1351 ** int32_t string table entry number
1352 ** uint8_t type
1353 ** (extern, absolute, defined in section)
1354 ** uint8_t section
1355 ** (0 for global symbols, section number of definition (>= 1, <=
1356 ** 254) for local symbols, size of variable for common symbols
1357 ** [type == extern])
1358 ** int16_t description
1359 ** (for stab debugging format)
1360 ** uint64_t value (i.e. file offset) of symbol or stab offset
1362 ** string table data
1363 ** list of null-terminated strings
1366 /* Emit the Mach-O header. */
1367 macho_write_header();
1369 offset = MACHO_HEADER64_SIZE + head_sizeofcmds64;
1371 /* emit the segment load command */
1372 if (seg_nsects64 > 0)
1373 offset = macho_write_segment (offset);
1374 else
1375 nasm_error(ERR_WARNING, "no sections?");
1377 if (nsyms > 0) {
1378 /* write out symbol command */
1379 fwriteint32_t(LC_SYMTAB, ofile); /* cmd == LC_SYMTAB */
1380 fwriteint32_t(MACHO_SYMCMD_SIZE, ofile); /* size of load command */
1381 fwriteint32_t(offset, ofile); /* symbol table offset */
1382 fwriteint32_t(nsyms, ofile); /* number of symbol
1383 ** table entries */
1385 offset += nsyms * MACHO_NLIST64_SIZE;
1386 fwriteint32_t(offset, ofile); /* string table offset */
1387 fwriteint32_t(strslen, ofile); /* string table size */
1390 /* emit section data */
1391 if (seg_nsects64 > 0)
1392 macho_write_section ();
1394 /* emit symbol table if we have symbols */
1395 if (nsyms > 0)
1396 macho_write_symtab ();
1398 /* we don't need to pad here since MACHO_NLIST64_SIZE == 16 */
1400 /* emit string table */
1401 saa_fpwrite(strs, ofile);
1403 /* We do quite a bit here, starting with finalizing all of the data
1404 for the object file, writing, and then freeing all of the data from
1405 the file. */
1407 static void macho_cleanup(int debuginfo)
1409 struct section *s;
1410 struct reloc *r;
1411 struct symbol *sym;
1413 (void)debuginfo;
1415 /* Sort all symbols. */
1416 macho_layout_symbols (&nsyms, &strslen);
1418 /* Fixup relocation entries */
1419 for (s = sects; s != NULL; s = s->next) {
1420 macho_fixup_relocs (s->relocs);
1423 /* First calculate and finalize needed values. */
1424 macho_calculate_sizes();
1425 macho_write();
1427 /* free up everything */
1428 while (sects->next) {
1429 s = sects;
1430 sects = sects->next;
1432 saa_free(s->data);
1433 while (s->relocs != NULL) {
1434 r = s->relocs;
1435 s->relocs = s->relocs->next;
1436 nasm_free(r);
1439 nasm_free(s);
1442 saa_free(strs);
1443 raa_free(extsyms);
1445 if (syms) {
1446 while (syms->next) {
1447 sym = syms;
1448 syms = syms->next;
1450 nasm_free (sym);
1455 /* Debugging routines. */
1456 static void debug_reloc (struct reloc *r)
1458 fprintf (stdout, "reloc:\n");
1459 fprintf (stdout, "\taddr: %"PRId32"\n", r->addr);
1460 fprintf (stdout, "\tsnum: %d\n", r->snum);
1461 fprintf (stdout, "\tpcrel: %d\n", r->pcrel);
1462 fprintf (stdout, "\tlength: %d\n", r->length);
1463 fprintf (stdout, "\text: %d\n", r->ext);
1464 fprintf (stdout, "\ttype: %d\n", r->type);
1467 static void debug_section_relocs (struct section *s)
1469 struct reloc *r = s->relocs;
1471 fprintf (stdout, "relocs for section %s:\n\n", s->sectname);
1473 while (r != NULL) {
1474 debug_reloc (r);
1475 r = r->next;
1479 struct ofmt of_macho64 = {
1480 "NeXTstep/OpenStep/Rhapsody/Darwin/MacOS X (x86_64) object files",
1481 "macho64",
1483 null_debug_arr,
1484 &null_debug_form,
1485 macho_stdmac,
1486 macho_init,
1487 null_setinfo,
1488 macho_output,
1489 macho_symdef,
1490 macho_section,
1491 macho_segbase,
1492 null_directive,
1493 macho_filename,
1494 macho_cleanup
1497 #endif
1500 * Local Variables:
1501 * mode:c
1502 * c-basic-offset:4
1503 * End:
1505 * end of file */