1 /*
2 * Copyright 2008-2009 Sun Microsystems, Inc. All Rights Reserved.
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4 *
5 * This code is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 only, as
7 * published by the Free Software Foundation.
8 *
9 * This code is distributed in the hope that it will be useful, but WITHOUT
10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
12 * version 2 for more details (a copy is included in the LICENSE file that
13 * accompanied this code).
14 *
15 * You should have received a copy of the GNU General Public License version
16 * 2 along with this work; if not, write to the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
18 *
19 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
20 * CA 95054 USA or visit www.sun.com if you need additional information or
21 * have any questions.
22 *
23 */
24
25 /* hsdis.c -- dump a range of addresses as native instructions
26 This implements the plugin protocol required by the
27 HotSpot PrintAssembly option.
28 */
29
30 #include "hsdis.h"
31 #include <libiberty.h>
32 #include <bfd.h>
33 #include <dis-asm.h>
34 #include <inttypes.h>
35 #include <string.h>
36 #include <errno.h>
37
38 #ifndef bool
39 #define bool int
40 #define true 1
41 #define false 0
42 #endif /*bool*/
43
44 /* short names for stuff in hsdis.h */
45 typedef decode_instructions_event_callback_ftype event_callback_t;
46 typedef decode_instructions_printf_callback_ftype printf_callback_t;
47
48 /* disassemble_info.application_data object */
49 struct hsdis_app_data {
50 uintptr_t start_va, end_va; /* virtual address of data */
51 /* the arguments to decode_instructions */
52 uintptr_t start; uintptr_t end;
53 event_callback_t event_callback; void* event_stream;
54 printf_callback_t printf_callback; void* printf_stream;
55 bool losing;
56
57 /* the architecture being disassembled */
58 const char* arch_name;
59 const bfd_arch_info_type* arch_info;
60
61 /* the disassembler we are going to use: */
62 disassembler_ftype dfn;
63 struct disassemble_info dinfo; /* the actual struct! */
64
65 char mach_option[64];
66 char insn_options[256];
67 };
68
69 #define DECL_APP_DATA(dinfo) \
70 struct hsdis_app_data* app_data = (struct hsdis_app_data*) (dinfo)->application_data
71
72 #define DECL_EVENT_CALLBACK(app_data) \
73 event_callback_t event_callback = (app_data)->event_callback; \
74 void* event_stream = (app_data)->event_stream
75
76 #define DECL_PRINTF_CALLBACK(app_data) \
77 printf_callback_t printf_callback = (app_data)->printf_callback; \
78 void* printf_stream = (app_data)->printf_stream
79
80
81 static void print_help(struct hsdis_app_data* app_data,
82 const char* msg, const char* arg);
83 static void setup_app_data(struct hsdis_app_data* app_data,
84 const char* options);
85 static const char* format_insn_close(const char* close,
86 disassemble_info* dinfo,
87 char* buf, size_t bufsize);
88
89 void*
90 #ifdef DLL_ENTRY
91 DLL_ENTRY
92 #endif
93 decode_instructions_virtual(void* start_va,
94 void* start_pv, void* end_pv,
95 event_callback_t event_callback_arg, void* event_stream_arg,
96 printf_callback_t printf_callback_arg, void* printf_stream_arg,
97 const char* options) {
98 struct hsdis_app_data app_data;
99 memset(&app_data, 0, sizeof(app_data));
100 app_data.start = (uintptr_t) start_pv;
101 app_data.end = (uintptr_t) end_pv;
102 app_data.start_va = (uintptr_t) start_va;
103 app_data.end_va = app_data.start_va + app_data.end - app_data.start;
104 app_data.event_callback = event_callback_arg;
105 app_data.event_stream = event_stream_arg;
106 app_data.printf_callback = printf_callback_arg;
107 app_data.printf_stream = printf_stream_arg;
108
109 setup_app_data(&app_data, options);
110 char buf[128];
111
112 {
113 /* now reload everything from app_data: */
114 DECL_EVENT_CALLBACK(&app_data);
115 DECL_PRINTF_CALLBACK(&app_data);
116 uintptr_t start = app_data.start_va;
117 uintptr_t end = app_data.end_va;
118 uintptr_t p = start;
119
120 (*event_callback)(event_stream, "insns", (void*)start);
121
122 (*event_callback)(event_stream, "mach name='%s'",
123 (void*) app_data.arch_info->printable_name);
124 if (app_data.dinfo.bytes_per_line != 0) {
125 (*event_callback)(event_stream, "format bytes-per-line='%p'/",
126 (void*)(intptr_t) app_data.dinfo.bytes_per_line);
127 }
128
129 while (p < end && !app_data.losing) {
130 (*event_callback)(event_stream, "insn", (void*) p);
131
132 /* reset certain state, so we can read it with confidence */
133 app_data.dinfo.insn_info_valid = 0;
134 app_data.dinfo.branch_delay_insns = 0;
135 app_data.dinfo.data_size = 0;
136 app_data.dinfo.insn_type = 0;
137
138 int size = (*app_data.dfn)((bfd_vma) p, &app_data.dinfo);
139
140 if (size > 0) p += size;
141 else app_data.losing = true;
142
143 const char* insn_close = format_insn_close("/insn", &app_data.dinfo,
144 buf, sizeof(buf));
145 (*event_callback)(event_stream, insn_close, (void*) p);
146
147 /* follow each complete insn by a nice newline */
148 (*printf_callback)(printf_stream, "\n");
149 }
150
151 (*event_callback)(event_stream, "/insns", (void*) p);
152 return (void*) p;
153 }
154 }
155
156 void*
157 #ifdef DLL_ENTRY
158 DLL_ENTRY
159 #endif
160 decode_instructions(void* start_pv, void* end_pv,
161 event_callback_t event_callback_arg, void* event_stream_arg,
162 printf_callback_t printf_callback_arg, void* printf_stream_arg,
163 const char* options) {
164 return decode_instructions_virtual(start_pv, start_pv, end_pv,
165 event_callback_arg, event_stream_arg,
166 printf_callback_arg, printf_stream_arg,
167 options);
168 }
169
170 /* take the address of the function, for luck, and also test the typedef: */
171 const decode_instructions_ftype decode_instructions_address = &decode_instructions;
172
173 static const char* format_insn_close(const char* close,
174 disassemble_info* dinfo,
175 char* buf, size_t bufsize) {
176 if (!dinfo->insn_info_valid)
177 return close;
178 enum dis_insn_type itype = dinfo->insn_type;
179 int dsize = dinfo->data_size, delays = dinfo->branch_delay_insns;
180 if ((itype == dis_nonbranch && (dsize | delays) == 0)
181 || (strlen(close) + 3*20 > bufsize))
182 return close;
183
184 const char* type = "unknown";
185 switch (itype) {
186 case dis_nonbranch: type = NULL; break;
187 case dis_branch: type = "branch"; break;
188 case dis_condbranch: type = "condbranch"; break;
189 case dis_jsr: type = "jsr"; break;
190 case dis_condjsr: type = "condjsr"; break;
191 case dis_dref: type = "dref"; break;
192 case dis_dref2: type = "dref2"; break;
193 }
194
195 strcpy(buf, close);
196 char* p = buf;
197 if (type) sprintf(p += strlen(p), " type='%s'", type);
198 if (dsize) sprintf(p += strlen(p), " dsize='%d'", dsize);
199 if (delays) sprintf(p += strlen(p), " delay='%d'", delays);
200 return buf;
201 }
202
203 /* handler functions */
204
205 static int
206 hsdis_read_memory_func(bfd_vma memaddr,
207 bfd_byte* myaddr,
208 unsigned int length,
209 struct disassemble_info* dinfo) {
210 DECL_APP_DATA(dinfo);
211 // convert the virtual address membar into an address within memory buffer
212 uintptr_t memaddr_p = ((uintptr_t) memaddr) - app_data->start_va + app_data->start;
213 if (memaddr_p + length > app_data->end) {
214 /* read is out of bounds */
215 return EIO;
216 } else {
217 memcpy(myaddr, (bfd_byte*) memaddr_p, length);
218 return 0;
219 }
220 }
221
222 static void
223 hsdis_print_address_func(bfd_vma vma, struct disassemble_info* dinfo) {
224 /* the actual value to print: */
225 void* addr_value = (void*) (uintptr_t) vma;
226 DECL_APP_DATA(dinfo);
227 DECL_EVENT_CALLBACK(app_data);
228
229 /* issue the event: */
230 void* result =
231 (*event_callback)(event_stream, "addr/", addr_value);
232 if (result == NULL) {
233 /* event declined */
234 generic_print_address(vma, dinfo);
235 }
236 }
237
238
239 /* configuration */
240
241 static void set_optional_callbacks(struct hsdis_app_data* app_data);
242 static void parse_caller_options(struct hsdis_app_data* app_data,
243 const char* caller_options);
244 static const char* native_arch_name();
245 static enum bfd_endian native_endian();
246 static const bfd_arch_info_type* find_arch_info(const char* arch_nane);
247 static bfd* get_native_bfd(const bfd_arch_info_type* arch_info,
248 /* to avoid malloc: */
249 bfd* empty_bfd, bfd_target* empty_xvec);
250 static void init_disassemble_info_from_bfd(struct disassemble_info* dinfo,
251 void *stream,
252 fprintf_ftype fprintf_func,
253 bfd* bfd,
254 char* disassembler_options);
255 static void parse_fake_insn(disassembler_ftype dfn,
256 struct disassemble_info* dinfo);
257
258 static void setup_app_data(struct hsdis_app_data* app_data,
259 const char* caller_options) {
260 /* Make reasonable defaults for null callbacks.
261 A non-null stream for a null callback is assumed to be a FILE* for output.
262 Events are rendered as XML.
263 */
264 set_optional_callbacks(app_data);
265
266 /* Look into caller_options for anything interesting. */
267 if (caller_options != NULL)
268 parse_caller_options(app_data, caller_options);
269
270 /* Discover which architecture we are going to disassemble. */
271 app_data->arch_name = &app_data->mach_option[0];
272 if (app_data->arch_name[0] == '\0')
273 app_data->arch_name = native_arch_name();
274 app_data->arch_info = find_arch_info(app_data->arch_name);
275
276 /* Make a fake bfd to hold the arch. and byteorder info. */
277 struct {
278 bfd_target empty_xvec;
279 bfd empty_bfd;
280 } buf;
281 bfd* native_bfd = get_native_bfd(app_data->arch_info,
282 /* to avoid malloc: */
283 &buf.empty_bfd, &buf.empty_xvec);
284 init_disassemble_info_from_bfd(&app_data->dinfo,
285 app_data->printf_stream,
286 app_data->printf_callback,
287 native_bfd,
288 app_data->insn_options);
289
290 /* Finish linking together the various callback blocks. */
291 app_data->dinfo.application_data = (void*) app_data;
292 app_data->dfn = disassembler(native_bfd);
293 app_data->dinfo.print_address_func = hsdis_print_address_func;
294 app_data->dinfo.read_memory_func = hsdis_read_memory_func;
295
296 if (app_data->dfn == NULL) {
297 const char* bad = app_data->arch_name;
298 static bool complained;
299 if (bad == &app_data->mach_option[0])
300 print_help(app_data, "bad mach=%s", bad);
301 else if (!complained)
302 print_help(app_data, "bad native mach=%s; please port hsdis to this platform", bad);
303 complained = true;
304 /* must bail out */
305 app_data->losing = true;
306 return;
307 }
308
309 parse_fake_insn(app_data->dfn, &app_data->dinfo);
310 }
311
312
313 /* ignore all events, return a null */
314 static void* null_event_callback(void* ignore_stream, const char* ignore_event, void* arg) {
315 return NULL;
316 }
317
318 /* print all events as XML markup */
319 static void* xml_event_callback(void* stream, const char* event, void* arg) {
320 FILE* fp = (FILE*) stream;
321 #define NS_PFX "dis:"
322 if (event[0] != '/') {
323 /* issue the tag, with or without a formatted argument */
324 fprintf(fp, "<"NS_PFX);
325 fprintf(fp, event, arg);
326 fprintf(fp, ">");
327 } else {
328 ++event; /* skip slash */
329 const char* argp = strchr(event, ' ');
330 if (argp == NULL) {
331 /* no arguments; just issue the closing tag */
332 fprintf(fp, "</"NS_PFX"%s>", event);
333 } else {
334 /* split out the closing attributes as <dis:foo_done attr='val'/> */
335 int event_prefix = (argp - event);
336 fprintf(fp, "<"NS_PFX"%.*s_done", event_prefix, event);
337 fprintf(fp, argp, arg);
338 fprintf(fp, "/></"NS_PFX"%.*s>", event_prefix, event);
339 }
340 }
341 return NULL;
342 }
343
344 static void set_optional_callbacks(struct hsdis_app_data* app_data) {
345 if (app_data->printf_callback == NULL) {
346 int (*fprintf_callback)(FILE*, const char*, ...) = &fprintf;
347 FILE* fprintf_stream = stdout;
348 app_data->printf_callback = (printf_callback_t) fprintf_callback;
349 if (app_data->printf_stream == NULL)
350 app_data->printf_stream = (void*) fprintf_stream;
351 }
352 if (app_data->event_callback == NULL) {
353 if (app_data->event_stream == NULL)
354 app_data->event_callback = &null_event_callback;
355 else
356 app_data->event_callback = &xml_event_callback;
357 }
358
359 }
360
361 static void parse_caller_options(struct hsdis_app_data* app_data, const char* caller_options) {
362 char* iop_base = app_data->insn_options;
363 char* iop_limit = iop_base + sizeof(app_data->insn_options) - 1;
364 char* iop = iop_base;
365 const char* p;
366 for (p = caller_options; p != NULL; ) {
367 const char* q = strchr(p, ',');
368 size_t plen = (q == NULL) ? strlen(p) : ((q++) - p);
369 if (plen == 4 && strncmp(p, "help", plen) == 0) {
370 print_help(app_data, NULL, NULL);
371 } else if (plen >= 5 && strncmp(p, "mach=", 5) == 0) {
372 char* mach_option = app_data->mach_option;
373 size_t mach_size = sizeof(app_data->mach_option);
374 mach_size -= 1; /*leave room for the null*/
375 if (plen > mach_size) plen = mach_size;
376 strncpy(mach_option, p, plen);
377 mach_option[plen] = '\0';
378 } else if (plen > 6 && strncmp(p, "hsdis-", 6)) {
379 // do not pass these to the next level
380 } else {
381 /* just copy it; {i386,sparc}-dis.c might like to see it */
382 if (iop > iop_base && iop < iop_limit) (*iop++) = ',';
383 if (iop + plen > iop_limit)
384 plen = iop_limit - iop;
385 strncpy(iop, p, plen);
386 iop += plen;
387 }
388 p = q;
389 }
390 }
391
392 static void print_help(struct hsdis_app_data* app_data,
393 const char* msg, const char* arg) {
394 DECL_PRINTF_CALLBACK(app_data);
395 if (msg != NULL) {
396 (*printf_callback)(printf_stream, "hsdis: ");
397 (*printf_callback)(printf_stream, msg, arg);
398 (*printf_callback)(printf_stream, "\n");
399 }
400 (*printf_callback)(printf_stream, "hsdis output options:\n");
401 if (printf_callback == (printf_callback_t) &fprintf)
402 disassembler_usage((FILE*) printf_stream);
403 else
404 disassembler_usage(stderr); /* better than nothing */
405 (*printf_callback)(printf_stream, " mach=<arch> select disassembly mode\n");
406 #if defined(LIBARCH_i386) || defined(LIBARCH_amd64)
407 (*printf_callback)(printf_stream, " mach=i386 select 32-bit mode\n");
408 (*printf_callback)(printf_stream, " mach=x86-64 select 64-bit mode\n");
409 (*printf_callback)(printf_stream, " suffix always print instruction suffix\n");
410 #endif
411 (*printf_callback)(printf_stream, " help print this message\n");
412 }
413
414
415 /* low-level bfd and arch stuff that binutils doesn't do for us */
416
417 static const bfd_arch_info_type* find_arch_info(const char* arch_name) {
418 const bfd_arch_info_type* arch_info = bfd_scan_arch(arch_name);
419 if (arch_info == NULL) {
420 extern const bfd_arch_info_type bfd_default_arch_struct;
421 arch_info = &bfd_default_arch_struct;
422 }
423 return arch_info;
424 }
425
426 static const char* native_arch_name() {
427 const char* res = NULL;
428 #ifdef LIBARCH_i386
429 res = "i386";
430 #endif
431 #ifdef LIBARCH_amd64
432 res = "i386:x86-64";
433 #endif
434 #ifdef LIBARCH_sparc
435 res = "sparc:v8plusb";
436 #endif
437 #ifdef LIBARCH_sparcv9
438 res = "sparc:v9b";
439 #endif
440 if (res == NULL)
441 res = "architecture not set in Makefile!";
442 return res;
443 }
444
445 static enum bfd_endian native_endian() {
446 int32_t endian_test = 'x';
447 if (*(const char*) &endian_test == 'x')
448 return BFD_ENDIAN_LITTLE;
449 else
450 return BFD_ENDIAN_BIG;
451 }
452
453 static bfd* get_native_bfd(const bfd_arch_info_type* arch_info,
454 bfd* empty_bfd, bfd_target* empty_xvec) {
455 memset(empty_bfd, 0, sizeof(*empty_bfd));
456 memset(empty_xvec, 0, sizeof(*empty_xvec));
457 empty_xvec->flavour = bfd_target_unknown_flavour;
458 empty_xvec->byteorder = native_endian();
459 empty_bfd->xvec = empty_xvec;
460 empty_bfd->arch_info = arch_info;
461 return empty_bfd;
462 }
463
464 static int read_zero_data_only(bfd_vma ignore_p,
465 bfd_byte* myaddr, unsigned int length,
466 struct disassemble_info *ignore_info) {
467 memset(myaddr, 0, length);
468 return 0;
469 }
470 static int print_to_dev_null(void* ignore_stream, const char* ignore_format, ...) {
471 return 0;
472 }
473
474 /* Prime the pump by running the selected disassembler on a null input.
475 This forces the machine-specific disassembler to divulge invariant
476 information like bytes_per_line.
477 */
478 static void parse_fake_insn(disassembler_ftype dfn,
479 struct disassemble_info* dinfo) {
480 typedef int (*read_memory_ftype)
481 (bfd_vma memaddr, bfd_byte *myaddr, unsigned int length,
482 struct disassemble_info *info);
483 read_memory_ftype read_memory_func = dinfo->read_memory_func;
484 fprintf_ftype fprintf_func = dinfo->fprintf_func;
485
486 dinfo->read_memory_func = &read_zero_data_only;
487 dinfo->fprintf_func = &print_to_dev_null;
488 (*dfn)(0, dinfo);
489
490 // put it back:
491 dinfo->read_memory_func = read_memory_func;
492 dinfo->fprintf_func = fprintf_func;
493 }
494
495 static void init_disassemble_info_from_bfd(struct disassemble_info* dinfo,
496 void *stream,
497 fprintf_ftype fprintf_func,
498 bfd* abfd,
499 char* disassembler_options) {
500 init_disassemble_info(dinfo, stream, fprintf_func);
501
502 dinfo->flavour = bfd_get_flavour(abfd);
503 dinfo->arch = bfd_get_arch(abfd);
504 dinfo->mach = bfd_get_mach(abfd);
505 dinfo->disassembler_options = disassembler_options;
506 dinfo->octets_per_byte = bfd_octets_per_byte (abfd);
507 dinfo->skip_zeroes = sizeof(void*) * 2;
508 dinfo->skip_zeroes_at_end = sizeof(void*)-1;
509 dinfo->disassembler_needs_relocs = FALSE;
510
511 if (bfd_big_endian(abfd))
512 dinfo->display_endian = dinfo->endian = BFD_ENDIAN_BIG;
513 else if (bfd_little_endian(abfd))
514 dinfo->display_endian = dinfo->endian = BFD_ENDIAN_LITTLE;
515 else
516 dinfo->endian = native_endian();
517
518 disassemble_init_for_target(dinfo);
519 }