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 <libiberty.h>
31 #include <bfd.h>
32 #include <dis-asm.h>
33 #include <inttypes.h>
34 #include <string.h>
35 #include <errno.h>
36 #include "hsdis.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 /* virtual address of data */
51 uintptr_t start_va, end_va;
52 /* the instructions to be decoded */
53 unsigned char* buffer;
54 uintptr_t length;
55 event_callback_t event_callback; void* event_stream;
56 printf_callback_t printf_callback; void* printf_stream;
57 bool losing;
58 bool do_newline;
59
60 /* the architecture being disassembled */
61 const char* arch_name;
62 const bfd_arch_info_type* arch_info;
63
64 /* the disassembler we are going to use: */
65 disassembler_ftype dfn;
66 struct disassemble_info dinfo; /* the actual struct! */
67
68 char mach_option[64];
69 char insn_options[256];
70 };
71
72 static void* decode(struct hsdis_app_data* app_data, const char* options);
73
74 #define DECL_APP_DATA(dinfo) \
75 struct hsdis_app_data* app_data = (struct hsdis_app_data*) (dinfo)->application_data
76
77 #define DECL_EVENT_CALLBACK(app_data) \
78 event_callback_t event_callback = (app_data)->event_callback; \
79 void* event_stream = (app_data)->event_stream
80
81 #define DECL_PRINTF_CALLBACK(app_data) \
82 printf_callback_t printf_callback = (app_data)->printf_callback; \
83 void* printf_stream = (app_data)->printf_stream
84
85
86 static void print_help(struct hsdis_app_data* app_data,
87 const char* msg, const char* arg);
88 static void setup_app_data(struct hsdis_app_data* app_data,
89 const char* options);
90 static const char* format_insn_close(const char* close,
91 disassemble_info* dinfo,
92 char* buf, size_t bufsize);
93
94 /* This is the compatability interface for older version of hotspot */
95 void*
96 #ifdef DLL_ENTRY
97 DLL_ENTRY
98 #endif
99 decode_instructions(void* start_pv, void* end_pv,
100 event_callback_t event_callback_arg, void* event_stream_arg,
101 printf_callback_t printf_callback_arg, void* printf_stream_arg,
102 const char* options) {
103 struct hsdis_app_data app_data;
104 memset(&app_data, 0, sizeof(app_data));
105 app_data.buffer = (unsigned char*) start_pv;
106 app_data.length = (uintptr_t)end_pv - (uintptr_t)start_pv;
107 app_data.start_va = (uintptr_t) start_pv;
108 app_data.end_va = app_data.start_va + app_data.length;
109 app_data.event_callback = event_callback_arg;
110 app_data.event_stream = event_stream_arg;
111 app_data.printf_callback = printf_callback_arg;
112 app_data.printf_stream = printf_stream_arg;
113 app_data.do_newline = true;
114
115 return decode(&app_data, options);
116 }
117
118 void*
119 #ifdef DLL_ENTRY
120 DLL_ENTRY
121 #endif
122 decode_instructions_virtual(uintptr_t start_va, uintptr_t end_va,
123 unsigned char* buffer, uintptr_t length,
124 event_callback_t event_callback_arg, void* event_stream_arg,
125 printf_callback_t printf_callback_arg, void* printf_stream_arg,
126 const char* options) {
127 struct hsdis_app_data app_data;
128 memset(&app_data, 0, sizeof(app_data));
129 app_data.start_va = start_va;
130 app_data.end_va = end_va;
131 app_data.buffer = buffer;
132 app_data.length = length;
133 app_data.event_callback = event_callback_arg;
134 app_data.event_stream = event_stream_arg;
135 app_data.printf_callback = printf_callback_arg;
136 app_data.printf_stream = printf_stream_arg;
137 app_data.do_newline = false;
138
139 return decode(&app_data, options);
140 }
141
142 static void* decode(struct hsdis_app_data* app_data, const char* options) {
143 setup_app_data(app_data, options);
144 char buf[128];
145
146 {
147 /* now reload everything from app_data: */
148 DECL_EVENT_CALLBACK(app_data);
149 DECL_PRINTF_CALLBACK(app_data);
150 uintptr_t start = app_data->start_va;
151 uintptr_t end = app_data->end_va;
152 uintptr_t p = start;
153
154 (*event_callback)(event_stream, "insns", (void*)start);
155
156 (*event_callback)(event_stream, "mach name='%s'",
157 (void*) app_data->arch_info->printable_name);
158 if (app_data->dinfo.bytes_per_line != 0) {
159 (*event_callback)(event_stream, "format bytes-per-line='%p'/",
160 (void*)(intptr_t) app_data->dinfo.bytes_per_line);
161 }
162
163 while (p < end && !app_data->losing) {
164 (*event_callback)(event_stream, "insn", (void*) p);
165
166 /* reset certain state, so we can read it with confidence */
167 app_data->dinfo.insn_info_valid = 0;
168 app_data->dinfo.branch_delay_insns = 0;
169 app_data->dinfo.data_size = 0;
170 app_data->dinfo.insn_type = 0;
171
172 int size = (*app_data->dfn)((bfd_vma) p, &app_data->dinfo);
173
174 if (size > 0) p += size;
175 else app_data->losing = true;
176
177 if (!app_data->losing) {
178 const char* insn_close = format_insn_close("/insn", &app_data->dinfo,
179 buf, sizeof(buf));
180 (*event_callback)(event_stream, insn_close, (void*) p) != NULL;
181
182 if (app_data->do_newline) {
183 /* follow each complete insn by a nice newline */
184 (*printf_callback)(printf_stream, "\n");
185 }
186 }
187 }
188
189 (*event_callback)(event_stream, "/insns", (void*) p);
190 return (void*) p;
191 }
192 }
193
194 /* take the address of the function, for luck, and also test the typedef: */
195 const decode_instructions_ftype decode_instructions_address = &decode_instructions_virtual;
196
197 static const char* format_insn_close(const char* close,
198 disassemble_info* dinfo,
199 char* buf, size_t bufsize) {
200 if (!dinfo->insn_info_valid)
201 return close;
202 enum dis_insn_type itype = dinfo->insn_type;
203 int dsize = dinfo->data_size, delays = dinfo->branch_delay_insns;
204 if ((itype == dis_nonbranch && (dsize | delays) == 0)
205 || (strlen(close) + 3*20 > bufsize))
206 return close;
207
208 const char* type = "unknown";
209 switch (itype) {
210 case dis_nonbranch: type = NULL; break;
211 case dis_branch: type = "branch"; break;
212 case dis_condbranch: type = "condbranch"; break;
213 case dis_jsr: type = "jsr"; break;
214 case dis_condjsr: type = "condjsr"; break;
215 case dis_dref: type = "dref"; break;
216 case dis_dref2: type = "dref2"; break;
217 }
218
219 strcpy(buf, close);
220 char* p = buf;
221 if (type) sprintf(p += strlen(p), " type='%s'", type);
222 if (dsize) sprintf(p += strlen(p), " dsize='%d'", dsize);
223 if (delays) sprintf(p += strlen(p), " delay='%d'", delays);
224 return buf;
225 }
226
227 /* handler functions */
228
229 static int
230 hsdis_read_memory_func(bfd_vma memaddr,
231 bfd_byte* myaddr,
232 unsigned int length,
233 struct disassemble_info* dinfo) {
234 DECL_APP_DATA(dinfo);
235 /* convert the virtual address memaddr into an address within memory buffer */
236 uintptr_t offset = ((uintptr_t) memaddr) - app_data->start_va;
237 if (offset + length > app_data->length) {
238 /* read is out of bounds */
239 return EIO;
240 } else {
241 memcpy(myaddr, (bfd_byte*) (app_data->buffer + offset), length);
242 return 0;
243 }
244 }
245
246 static void
247 hsdis_print_address_func(bfd_vma vma, struct disassemble_info* dinfo) {
248 /* the actual value to print: */
249 void* addr_value = (void*) (uintptr_t) vma;
250 DECL_APP_DATA(dinfo);
251 DECL_EVENT_CALLBACK(app_data);
252
253 /* issue the event: */
254 void* result =
255 (*event_callback)(event_stream, "addr/", addr_value);
256 if (result == NULL) {
257 /* event declined */
258 generic_print_address(vma, dinfo);
259 }
260 }
261
262
263 /* configuration */
264
265 static void set_optional_callbacks(struct hsdis_app_data* app_data);
266 static void parse_caller_options(struct hsdis_app_data* app_data,
267 const char* caller_options);
268 static const char* native_arch_name();
269 static enum bfd_endian native_endian();
270 static const bfd_arch_info_type* find_arch_info(const char* arch_nane);
271 static bfd* get_native_bfd(const bfd_arch_info_type* arch_info,
272 /* to avoid malloc: */
273 bfd* empty_bfd, bfd_target* empty_xvec);
274 static void init_disassemble_info_from_bfd(struct disassemble_info* dinfo,
275 void *stream,
276 fprintf_ftype fprintf_func,
277 bfd* bfd,
278 char* disassembler_options);
279 static void parse_fake_insn(disassembler_ftype dfn,
280 struct disassemble_info* dinfo);
281
282 static void setup_app_data(struct hsdis_app_data* app_data,
283 const char* caller_options) {
284 /* Make reasonable defaults for null callbacks.
285 A non-null stream for a null callback is assumed to be a FILE* for output.
286 Events are rendered as XML.
287 */
288 set_optional_callbacks(app_data);
289
290 /* Look into caller_options for anything interesting. */
291 if (caller_options != NULL)
292 parse_caller_options(app_data, caller_options);
293
294 /* Discover which architecture we are going to disassemble. */
295 app_data->arch_name = &app_data->mach_option[0];
296 if (app_data->arch_name[0] == '\0')
297 app_data->arch_name = native_arch_name();
298 app_data->arch_info = find_arch_info(app_data->arch_name);
299
300 /* Make a fake bfd to hold the arch. and byteorder info. */
301 struct {
302 bfd_target empty_xvec;
303 bfd empty_bfd;
304 } buf;
305 bfd* native_bfd = get_native_bfd(app_data->arch_info,
306 /* to avoid malloc: */
307 &buf.empty_bfd, &buf.empty_xvec);
308 init_disassemble_info_from_bfd(&app_data->dinfo,
309 app_data->printf_stream,
310 app_data->printf_callback,
311 native_bfd,
312 app_data->insn_options);
313
314 /* Finish linking together the various callback blocks. */
315 app_data->dinfo.application_data = (void*) app_data;
316 app_data->dfn = disassembler(native_bfd);
317 app_data->dinfo.print_address_func = hsdis_print_address_func;
318 app_data->dinfo.read_memory_func = hsdis_read_memory_func;
319
320 if (app_data->dfn == NULL) {
321 const char* bad = app_data->arch_name;
322 static bool complained;
323 if (bad == &app_data->mach_option[0])
324 print_help(app_data, "bad mach=%s", bad);
325 else if (!complained)
326 print_help(app_data, "bad native mach=%s; please port hsdis to this platform", bad);
327 complained = true;
328 /* must bail out */
329 app_data->losing = true;
330 return;
331 }
332
333 parse_fake_insn(app_data->dfn, &app_data->dinfo);
334 }
335
336
337 /* ignore all events, return a null */
338 static void* null_event_callback(void* ignore_stream, const char* ignore_event, void* arg) {
339 return NULL;
340 }
341
342 /* print all events as XML markup */
343 static void* xml_event_callback(void* stream, const char* event, void* arg) {
344 FILE* fp = (FILE*) stream;
345 #define NS_PFX "dis:"
346 if (event[0] != '/') {
347 /* issue the tag, with or without a formatted argument */
348 fprintf(fp, "<"NS_PFX);
349 fprintf(fp, event, arg);
350 fprintf(fp, ">");
351 } else {
352 ++event; /* skip slash */
353 const char* argp = strchr(event, ' ');
354 if (argp == NULL) {
355 /* no arguments; just issue the closing tag */
356 fprintf(fp, "</"NS_PFX"%s>", event);
357 } else {
358 /* split out the closing attributes as <dis:foo_done attr='val'/> */
359 int event_prefix = (argp - event);
360 fprintf(fp, "<"NS_PFX"%.*s_done", event_prefix, event);
361 fprintf(fp, argp, arg);
362 fprintf(fp, "/></"NS_PFX"%.*s>", event_prefix, event);
363 }
364 }
365 return NULL;
366 }
367
368 static void set_optional_callbacks(struct hsdis_app_data* app_data) {
369 if (app_data->printf_callback == NULL) {
370 int (*fprintf_callback)(FILE*, const char*, ...) = &fprintf;
371 FILE* fprintf_stream = stdout;
372 app_data->printf_callback = (printf_callback_t) fprintf_callback;
373 if (app_data->printf_stream == NULL)
374 app_data->printf_stream = (void*) fprintf_stream;
375 }
376 if (app_data->event_callback == NULL) {
377 if (app_data->event_stream == NULL)
378 app_data->event_callback = &null_event_callback;
379 else
380 app_data->event_callback = &xml_event_callback;
381 }
382
383 }
384
385 static void parse_caller_options(struct hsdis_app_data* app_data, const char* caller_options) {
386 char* iop_base = app_data->insn_options;
387 char* iop_limit = iop_base + sizeof(app_data->insn_options) - 1;
388 char* iop = iop_base;
389 const char* p;
390 for (p = caller_options; p != NULL; ) {
391 const char* q = strchr(p, ',');
392 size_t plen = (q == NULL) ? strlen(p) : ((q++) - p);
393 if (plen == 4 && strncmp(p, "help", plen) == 0) {
394 print_help(app_data, NULL, NULL);
395 } else if (plen >= 5 && strncmp(p, "mach=", 5) == 0) {
396 char* mach_option = app_data->mach_option;
397 size_t mach_size = sizeof(app_data->mach_option);
398 mach_size -= 1; /*leave room for the null*/
399 if (plen > mach_size) plen = mach_size;
400 strncpy(mach_option, p, plen);
401 mach_option[plen] = '\0';
402 } else if (plen > 6 && strncmp(p, "hsdis-", 6)) {
403 /* do not pass these to the next level */
404 } else {
405 /* just copy it; {i386,sparc}-dis.c might like to see it */
406 if (iop > iop_base && iop < iop_limit) (*iop++) = ',';
407 if (iop + plen > iop_limit)
408 plen = iop_limit - iop;
409 strncpy(iop, p, plen);
410 iop += plen;
411 }
412 p = q;
413 }
414 }
415
416 static void print_help(struct hsdis_app_data* app_data,
417 const char* msg, const char* arg) {
418 DECL_PRINTF_CALLBACK(app_data);
419 if (msg != NULL) {
420 (*printf_callback)(printf_stream, "hsdis: ");
421 (*printf_callback)(printf_stream, msg, arg);
422 (*printf_callback)(printf_stream, "\n");
423 }
424 (*printf_callback)(printf_stream, "hsdis output options:\n");
425 if (printf_callback == (printf_callback_t) &fprintf)
426 disassembler_usage((FILE*) printf_stream);
427 else
428 disassembler_usage(stderr); /* better than nothing */
429 (*printf_callback)(printf_stream, " mach=<arch> select disassembly mode\n");
430 #if defined(LIBARCH_i386) || defined(LIBARCH_amd64)
431 (*printf_callback)(printf_stream, " mach=i386 select 32-bit mode\n");
432 (*printf_callback)(printf_stream, " mach=x86-64 select 64-bit mode\n");
433 (*printf_callback)(printf_stream, " suffix always print instruction suffix\n");
434 #endif
435 (*printf_callback)(printf_stream, " help print this message\n");
436 }
437
438
439 /* low-level bfd and arch stuff that binutils doesn't do for us */
440
441 static const bfd_arch_info_type* find_arch_info(const char* arch_name) {
442 const bfd_arch_info_type* arch_info = bfd_scan_arch(arch_name);
443 if (arch_info == NULL) {
444 extern const bfd_arch_info_type bfd_default_arch_struct;
445 arch_info = &bfd_default_arch_struct;
446 }
447 return arch_info;
448 }
449
450 static const char* native_arch_name() {
451 const char* res = NULL;
452 #ifdef LIBARCH_i386
453 res = "i386";
454 #endif
455 #ifdef LIBARCH_amd64
456 res = "i386:x86-64";
457 #endif
458 #ifdef LIBARCH_sparc
459 res = "sparc:v8plusb";
460 #endif
461 #ifdef LIBARCH_sparcv9
462 res = "sparc:v9b";
463 #endif
464 if (res == NULL)
465 res = "architecture not set in Makefile!";
466 return res;
467 }
468
469 static enum bfd_endian native_endian() {
470 int32_t endian_test = 'x';
471 if (*(const char*) &endian_test == 'x')
472 return BFD_ENDIAN_LITTLE;
473 else
474 return BFD_ENDIAN_BIG;
475 }
476
477 static bfd* get_native_bfd(const bfd_arch_info_type* arch_info,
478 bfd* empty_bfd, bfd_target* empty_xvec) {
479 memset(empty_bfd, 0, sizeof(*empty_bfd));
480 memset(empty_xvec, 0, sizeof(*empty_xvec));
481 empty_xvec->flavour = bfd_target_unknown_flavour;
482 empty_xvec->byteorder = native_endian();
483 empty_bfd->xvec = empty_xvec;
484 empty_bfd->arch_info = arch_info;
485 return empty_bfd;
486 }
487
488 static int read_zero_data_only(bfd_vma ignore_p,
489 bfd_byte* myaddr, unsigned int length,
490 struct disassemble_info *ignore_info) {
491 memset(myaddr, 0, length);
492 return 0;
493 }
494 static int print_to_dev_null(void* ignore_stream, const char* ignore_format, ...) {
495 return 0;
496 }
497
498 /* Prime the pump by running the selected disassembler on a null input.
499 This forces the machine-specific disassembler to divulge invariant
500 information like bytes_per_line.
501 */
502 static void parse_fake_insn(disassembler_ftype dfn,
503 struct disassemble_info* dinfo) {
504 typedef int (*read_memory_ftype)
505 (bfd_vma memaddr, bfd_byte *myaddr, unsigned int length,
506 struct disassemble_info *info);
507 read_memory_ftype read_memory_func = dinfo->read_memory_func;
508 fprintf_ftype fprintf_func = dinfo->fprintf_func;
509
510 dinfo->read_memory_func = &read_zero_data_only;
511 dinfo->fprintf_func = &print_to_dev_null;
512 (*dfn)(0, dinfo);
513
514 /* put it back */
515 dinfo->read_memory_func = read_memory_func;
516 dinfo->fprintf_func = fprintf_func;
517 }
518
519 static void init_disassemble_info_from_bfd(struct disassemble_info* dinfo,
520 void *stream,
521 fprintf_ftype fprintf_func,
522 bfd* abfd,
523 char* disassembler_options) {
524 init_disassemble_info(dinfo, stream, fprintf_func);
525
526 dinfo->flavour = bfd_get_flavour(abfd);
527 dinfo->arch = bfd_get_arch(abfd);
528 dinfo->mach = bfd_get_mach(abfd);
529 dinfo->disassembler_options = disassembler_options;
530 dinfo->octets_per_byte = bfd_octets_per_byte (abfd);
531 dinfo->skip_zeroes = sizeof(void*) * 2;
532 dinfo->skip_zeroes_at_end = sizeof(void*)-1;
533 dinfo->disassembler_needs_relocs = FALSE;
534
535 if (bfd_big_endian(abfd))
536 dinfo->display_endian = dinfo->endian = BFD_ENDIAN_BIG;
537 else if (bfd_little_endian(abfd))
538 dinfo->display_endian = dinfo->endian = BFD_ENDIAN_LITTLE;
539 else
540 dinfo->endian = native_endian();
541
542 disassemble_init_for_target(dinfo);
543 }