5 * This work is licensed under the terms of 3-clause BSD license.
6 * See COPYING file in the top-level directory.
9 #define NO_OBSOLETE_FUNCS
14 #include <kernwin.hpp>
23 #define IS_START(w, y) !strncmp(w, y, strlen(y))
24 #define ARRAY_SIZE(x) (sizeof(x) / sizeof(x[0]))
26 static char **name_cache;
27 static size_t name_cache_size;
29 // non-local branch targets
30 static ea_t *nonlocal_bt;
31 static int nonlocal_bt_alloc;
32 static int nonlocal_bt_cnt;
34 //--------------------------------------------------------------------------
35 static int idaapi init(void)
40 //--------------------------------------------------------------------------
41 static void idaapi term(void)
45 if (nonlocal_bt != NULL) {
49 nonlocal_bt_alloc = 0;
51 if (name_cache != NULL) {
52 for (i = 0; i < name_cache_size; i++)
60 //--------------------------------------------------------------------------
62 static const char *reserved_names[] = {
75 static int is_name_reserved(const char *name)
78 for (i = 0; i < ARRAY_SIZE(reserved_names); i++)
79 if (strcasecmp(name, reserved_names[i]) == 0)
85 static int nonlocal_bt_cmp(const void *p1, const void *p2)
87 const ea_t *e1 = (const ea_t *)p1, *e2 = (const ea_t *)p2;
91 static void nonlocal_add(ea_t ea)
93 if (nonlocal_bt_cnt >= nonlocal_bt_alloc) {
94 nonlocal_bt_alloc += nonlocal_bt_alloc * 2 + 64;
95 nonlocal_bt = (ea_t *)realloc(nonlocal_bt,
96 nonlocal_bt_alloc * sizeof(nonlocal_bt[0]));
97 if (nonlocal_bt == NULL) {
102 nonlocal_bt[nonlocal_bt_cnt++] = ea;
105 // is instruction a (un)conditional jump (not call)?
106 static int is_insn_jmp(uint16 itype)
108 return itype == NN_jmp || (NN_ja <= itype && itype <= NN_jz);
111 static void do_def_line(char *buf, size_t buf_size, const char *line,
112 ea_t ea, func_t *func)
114 char func_name[256] = "<nf>";
121 tag_remove(line, buf, buf_size); // remove color codes
127 memmove(buf, buf + 9, len - 9 + 1); // rm address
130 while (*p && *p != ' ' && *p != ':')
133 ea_ret = (ea_t *)bsearch(&ea, nonlocal_bt, nonlocal_bt_cnt,
134 sizeof(nonlocal_bt[0]), nonlocal_bt_cmp);
135 global_label = (ea_ret != NULL);
138 get_func_name(ea, func_name, sizeof(func_name));
139 is_libfunc = func->flags & FUNC_LIB;
141 for (i = 0; i < get_item_size(ea); i++) {
143 if (xb.first_to(ea + i, XREF_DATA)) {
144 if (!is_libfunc && xb.type == dr_O)
145 msg("%x: offset xref in %s\n", ea, func_name);
152 msg("no trailing blank in '%s'\n", buf);
159 static int name_cache_cmp(const void *p1, const void *p2)
161 // masm ignores case, so do we
162 return stricmp(*(char * const *)p1, *(char * const *)p2);
165 static void rebuild_name_cache(void)
170 // build a sorted name cache
171 newsize = get_nlist_size();
172 if (newsize > name_cache_size) {
173 tmp = realloc(name_cache, newsize * sizeof(name_cache[0]));
175 msg("OOM for name cache\n");
178 name_cache = (char **)tmp;
180 for (i = 0; i < name_cache_size; i++)
182 for (i = 0; i < newsize; i++)
183 name_cache[i] = strdup(get_nlist_name(i));
185 name_cache_size = newsize;
186 qsort(name_cache, name_cache_size, sizeof(name_cache[0]),
190 static void my_rename(ea_t ea, char *name)
196 qsnprintf(buf, sizeof(buf), "%s", name);
199 pp = (char **)bsearch(&p, name_cache, name_cache_size,
200 sizeof(name_cache[0]), name_cache_cmp);
204 qsnprintf(buf, sizeof(buf), "%s_g%d", name, n);
210 msg("rename failure? '%s'\n", name);
212 do_name_anyway(ea, buf);
213 rebuild_name_cache();
216 static void make_align(ea_t ea)
221 tmp_ea = next_head(ea, inf.maxEA);
222 if ((tmp_ea & 0x03) == 0) {
223 n = calc_max_align(tmp_ea);
224 if (n > 4) // masm doesn't like more..
226 msg("%x: align %d\n", ea, 1 << n);
227 do_unknown(ea, DOUNK_SIMPLE);
228 doAlign(ea, tmp_ea - ea, n);
232 static void idaapi run(int /*arg*/)
234 // isEnabled(ea) // address belongs to disassembly
235 // ea_t ea = get_screen_ea();
237 // foo = DecodeInstruction(ScreenEA());
245 ea_t ui_ea_block = 0, ea_size;
246 ea_t tmp_ea, target_ea;
257 // get rid of structs, masm doesn't understand them
258 idx = get_first_struc_idx();
259 while (idx != BADNODE) {
260 tid_t tid = get_struc_by_idx(idx);
261 struc_t *struc = get_struc(tid);
262 get_struc_name(tid, buf, sizeof(buf));
263 msg("removing struct '%s'\n", buf);
264 //del_struc_members(struc, 0, get_max_offset(struc));
267 idx = get_first_struc_idx();
270 rebuild_name_cache();
272 // 1st pass: walk through all funcs
277 func_tail_iterator_t fti(func);
279 msg("%x: func_tail_iterator_t main failed\n", ea);
282 const area_t &f_area = fti.chunk();
285 // rename global syms which conflict with frame member names
286 frame = get_frame(func);
289 for (m = 0; m < (int)frame->memqty; m++)
291 ret = get_member_name(frame->members[m].id, buf, sizeof(buf));
293 msg("%x: member has no name?\n", ea);
296 if (buf[0] == ' ') // what's this?
298 if (IS_START(buf, "arg_") || IS_START(buf, "var_"))
301 // check for dupe names
303 for (m1 = 0; m1 < m; m1++) {
304 get_member_name(frame->members[m1].id, buf2, sizeof(buf2));
305 if (stricmp(buf, buf2) == 0)
309 if (is_name_reserved(buf) || dupe) {
310 msg("%x: renaming '%s'\n", ea, buf);
311 qstrncat(buf, "_", sizeof(buf));
312 ret = set_member_name(frame, frame->members[m].soff, buf);
314 msg("%x: renaming failed\n", ea);
320 pp = (char **)bsearch(&p, name_cache, name_cache_size,
321 sizeof(name_cache[0]), name_cache_cmp);
325 tmp_ea = get_name_ea(BADADDR, *pp);
326 msg("%x: renaming '%s' because of '%s' at %x\n",
327 tmp_ea, *pp, buf, ea);
328 my_rename(tmp_ea, *pp);
332 // detect tailcalls to next func with 'jmp $+5' (offset 0)
333 if (f_area.endEA - f_area.startEA >= 5
334 && decode_insn(f_area.endEA - 5) && cmd.itype == NN_jmp
335 && cmd.Operands[0].type == o_near
336 && cmd.Operands[0].addr == f_area.endEA
337 && get_name(BADADDR, f_area.endEA, buf, sizeof(buf))
338 && get_cmt(f_area.endEA - 5, false, buf2, sizeof(buf2)) <= 0)
340 qsnprintf(buf2, sizeof(buf2), "sctpatch: jmp %s", buf);
341 set_cmt(f_area.endEA - 5, buf2, false);
344 func = get_next_func(ea);
347 // 2nd pass over whole .text and .(ro)data segments
348 for (ea = inf.minEA; ea != BADADDR; ea = next_head(ea, inf.maxEA))
350 segment_t *seg = getseg(ea);
353 if (seg->type == SEG_XTRN)
355 if (seg->type != SEG_CODE && seg->type != SEG_DATA)
358 ea_flags = get_flags_novalue(ea);
360 if (isCode(ea_flags))
362 if (!decode_insn(ea)) {
363 msg("%x: decode_insn() failed\n", ea);
367 // masm doesn't understand IDA's float/xmm types
368 if (cmd.itype == NN_fld || cmd.itype == NN_fst
369 || cmd.itype == NN_movapd || cmd.itype == NN_movlpd)
371 for (o = 0; o < UA_MAXOP; o++) {
372 if (cmd.Operands[o].type == o_void)
375 if (cmd.Operands[o].type == o_mem) {
376 tmp_ea = cmd.Operands[o].addr;
377 flags_t tmp_flg = get_flags_novalue(tmp_ea);
379 if (isDouble(tmp_flg))
381 get_name(ea, tmp_ea, buf, sizeof(buf));
382 msg("%x: converting dbl %x '%s'\n", ea, tmp_ea, buf);
385 if (isOwrd(tmp_flg) || isYwrd(tmp_flg) || isTbyt(tmp_flg))
387 get_name(ea, tmp_ea, buf, sizeof(buf));
388 msg("%x: undefining lrg %x '%s'\n", ea, tmp_ea, buf);
389 do_unknown(tmp_ea, DOUNK_EXPAND);
394 else if (cmd.itype == NN_lea) {
395 // detect code alignment
396 if (cmd.Operands[0].reg == cmd.Operands[1].reg
397 && cmd.Operands[1].type == o_displ
398 && cmd.Operands[1].addr == 0)
403 else if (!isDefArg1(ea_flags)
404 && cmd.Operands[1].type == o_mem // why o_mem?
405 && cmd.Operands[1].dtyp == dt_dword)
407 if (inf.minEA <= cmd.Operands[1].addr
408 && cmd.Operands[1].addr < inf.maxEA)
410 // lea to segments, like ds:58D6A8h[edx*8]
411 msg("%x: lea offset to %x\n", ea, cmd.Operands[1].addr);
412 op_offset(ea, 1, REF_OFF32);
416 // ds:0[eax*8] -> [eax*8+0]
417 msg("%x: dropping ds: for %x\n", ea, cmd.Operands[1].addr);
422 else if (cmd.itype == NN_mov && cmd.segpref == 0x1e // 2e?
423 && cmd.Operands[0].type == o_reg
424 && cmd.Operands[1].type == o_reg
425 && cmd.Operands[0].dtyp == cmd.Operands[1].dtyp
426 && cmd.Operands[0].reg == cmd.Operands[1].reg)
428 // db 2Eh; mov eax, eax
432 // find non-local branches
433 if (is_insn_jmp(cmd.itype) && cmd.Operands[0].type == o_near)
435 target_ea = cmd.Operands[0].addr;
437 nonlocal_add(target_ea);
439 ret = get_func_chunknum(func, target_ea);
441 // a jump to another func or chunk
442 // check if it lands on func start
443 if (!isFunc(get_flags_novalue(target_ea)))
444 nonlocal_add(target_ea);
451 ea_size = get_item_size(ea);
453 if (func == NULL && isOff0(ea_flags)) {
454 for (tmp_ea = 0; tmp_ea < ea_size; tmp_ea += 4)
455 nonlocal_add(get_long(ea + tmp_ea));
458 // IDA vs masm float/mmx/xmm type incompatibility
459 if (isDouble(ea_flags))
461 msg("%x: converting double\n", ea);
464 else if (isTbyt(ea_flags) || isPackReal(ea_flags))
468 else if (isOwrd(ea_flags)) {
470 get_name(BADADDR, ea, buf, sizeof(buf));
471 if (IS_START(buf, "xmm"))
474 // masm doesn't understand IDA's unicode
475 else if (isASCII(ea_flags) && ea_size >= 4
476 && (get_long(ea) & 0xff00ff00) == 0) // lame..
480 // masm doesn't understand large aligns
481 else if (isAlign(ea_flags) && ea_size >= 0x10)
483 msg("%x: undefining align %d\n", ea, ea_size);
484 do_unknown(ea, DOUNK_EXPAND);
489 get_name(BADADDR, ea, buf, sizeof(buf));
490 msg("%x: undefining '%s'\n", ea, buf);
491 do_unknown(ea, DOUNK_EXPAND);
496 // check namelist for reserved names and
497 // matching names with different case (nasm ignores case)
498 n = get_nlist_size();
499 for (i = 0; i < n; i++) {
500 ea = get_nlist_ea(i);
501 name = get_nlist_name(i);
503 msg("%x: null name?\n", ea);
506 qsnprintf(buf, sizeof(buf), "%s", name);
508 int need_rename = is_name_reserved(name);
511 pp = (char **)bsearch(&p, name_cache, name_cache_size,
512 sizeof(name_cache[0]), name_cache_cmp);
514 if (pp > name_cache && stricmp(pp[-1], pp[0]) == 0)
516 else if (pp < name_cache + name_cache_size - 1
517 && stricmp(pp[0], pp[1]) == 0)
524 // rename vars with '?@' (funcs are ok)
526 ea_flags = get_flags_novalue(ea);
527 if (!isCode(ea_flags) && strpbrk(name, "?@"))
530 if (need_rename || change_qat) {
531 msg("%x: renaming name '%s'\n", ea, name);
534 for (p = buf; *p != 0; p++) {
535 if (*p == '?' || *p == '@') {
536 qsnprintf(buf2, sizeof(buf2), "%02x", (unsigned char)*p);
537 memmove(p + 1, p, strlen(p) + 1);
547 if (nonlocal_bt_cnt > 1) {
548 qsort(nonlocal_bt, nonlocal_bt_cnt,
549 sizeof(nonlocal_bt[0]), nonlocal_bt_cmp);
552 char *fname = askfile_c(1, NULL, "Save asm file");
555 fout = qfopen(fname, "w");
557 msg("couldn't open '%s'\n", fname);
561 show_wait_box("Saving..");
563 // deal with the beginning
565 int flags = 0; // calc_default_idaplace_flags();
566 linearray_t ln(&flags);
571 n = ln.get_linecnt();
572 for (i = 0; i < n - 1; i++) {
573 do_def_line(buf, sizeof(buf), ln.down(), ea, NULL);
574 if (strstr(buf, "include"))
578 qfprintf(fout, "%s\n", buf);
579 p = strstr(buf, ".mmx");
581 memcpy(p, ".xmm", 4);
583 qfprintf(fout, "%s\n", buf);
586 p = strstr(buf, ".model");
588 qstrncpy(p, "include imports.inc", sizeof(buf) - (p - buf));
590 qfprintf(fout, "\n%s\n", buf);
599 int drop_large = 0, do_rva = 0, set_scale = 0, jmp_near = 0;
600 int word_imm = 0, dword_imm = 0, do_pushf = 0, do_nops = 0;
602 if ((ea >> 14) != ui_ea_block) {
603 ui_ea_block = ea >> 14;
611 segment_t *seg = getseg(ea);
612 if (!seg || (seg->type != SEG_CODE && seg->type != SEG_DATA))
615 ea_flags = get_flags_novalue(ea);
616 if (isCode(ea_flags))
618 if (!decode_insn(ea))
621 if (is_insn_jmp(cmd.itype) && cmd.Operands[0].type == o_near
622 && cmd.Operands[0].dtyp == dt_dword)
626 else if ((cmd.itype == NN_pushf || cmd.itype == NN_popf)
632 for (o = 0; o < UA_MAXOP; o++) {
633 const op_t &opr = cmd.Operands[o];
634 if (opr.type == o_void)
638 if (opr.type == o_mem && opr.specval_shorts.high == 0x21)
640 if (opr.hasSIB && x86_scale(opr) == 0
641 && x86_index(opr) != INDEX_NONE)
645 // annoying alignment variant..
646 if (opr.type == o_imm && opr.dtyp == dt_dword
647 && (opr.value < 0x80 || opr.value > 0xffffff80)
648 && cmd.size >= opr.offb + 4)
650 if (get_long(ea + opr.offb) == opr.value)
653 else if (opr.type == o_imm && opr.dtyp == dt_word
654 && (opr.value < 0x80 || opr.value > 0xff80)
655 && cmd.size >= opr.offb + 2)
657 if (get_word(ea + opr.offb) == (ushort)opr.value)
660 else if (opr.type == o_displ && opr.addr == 0
661 && opr.offb != 0 && opr.hasSIB && opr.sib == 0x24)
663 // uses [esp+0] with 0 encoded into op
669 if (isOff0(ea_flags))
674 n = ln.get_linecnt();
675 for (i = pl.lnnum; i < n; i++) {
676 do_def_line(buf, sizeof(buf), ln.down(), ea, func);
679 for (fw = buf; *fw != 0 && *fw == ' '; )
684 p = strstr(fw, "large ");
686 memmove(p, p + 6, strlen(p + 6) + 1);
689 p = strstr(fw, " rva ");
692 memmove(p + 4 + 3, p + 4, strlen(p + 4) + 1);
693 memcpy(p + 1, "offset", 6);
699 if (p != NULL && p[1] == 'e') {
701 // scale is 1, must specify it explicitly so that
702 // masm chooses the right scaled reg
703 memmove(p + 2, p, strlen(p) + 1);
709 if (fw != buf && fw[0] == 'j')
711 while (p && *p != ' ')
713 while (p && *p == ' ')
716 memmove(p + 9, p, strlen(p) + 1);
717 memcpy(p, "near ptr ", 9);
722 p = strstr(fw, ", ");
723 if (p != NULL && '0' <= p[2] && p[2] <= '9') {
725 memmove(p + 9, p, strlen(p) + 1);
726 memcpy(p, "word ptr ", 9);
729 else if (dword_imm) {
730 p = strstr(fw, ", ");
731 if (p != NULL && '0' <= p[2] && p[2] <= '9') {
733 memmove(p + 10, p, strlen(p) + 1);
734 memcpy(p, "dword ptr ", 10);
738 p = strstr(fw, "pushf");
740 p = strstr(fw, "popf");
742 p = strchr(p, 'f') + 1;
743 memmove(p + 1, p, strlen(p) + 1);
748 if (fw[0] == 'a' && IS_START(fw, "assume cs")) {
749 // "assume cs" causes problems with ext syms
750 memmove(fw + 1, fw, strlen(fw) + 1);
753 else if (fw[0] == 'e' && IS_START(fw, "end") && fw[3] == ' ') {
755 qfprintf(fout, "include public.inc\n\n");
762 qfprintf(fout, "%s\n", buf);
765 while (do_nops-- > 0)
766 qfprintf(fout, " nop ; adj\n");
768 // note: next_head skips some undefined stuff
769 ea = next_not_tail(ea); // correct?
784 msg("%d lines saved.\n", fout_line);
787 //--------------------------------------------------------------------------
789 static const char comment[] = "Generate disassembly for nasm";
790 static const char help[] = "Generate asm file\n";
791 static const char wanted_name[] = "Save asm";
792 static const char wanted_hotkey[] = "Shift-S";
794 //--------------------------------------------------------------------------
796 // PLUGIN DESCRIPTION BLOCK
798 //--------------------------------------------------------------------------
801 IDP_INTERFACE_VERSION,
804 term, // terminate. this pointer may be NULL.
805 run, // invoke plugin
806 comment, // long comment about the plugin
807 // it could appear in the status line
809 help, // multiline help about the plugin
810 wanted_name, // the preferred short name of the plugin
811 wanted_hotkey // the preferred hotkey to run the plugin
814 // vim:ts=2:shiftwidth=2:expandtab