1 /*
2 * Copyright (c) 2008, 2010, Oracle and/or its affiliates. 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 Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20 * or visit www.oracle.com if you need additional information or have any
21 * questions.
22 *
23 */
24
25 #include "precompiled.hpp"
26
27 /* hsdis.c -- dump a range of addresses as native instructions
28 This implements the plugin protocol required by the
29 HotSpot PrintAssembly option.
30 */
31
32 #include "hsdis.h"
33
34 #include <sysdep.h>
35 #include <libiberty.h>
36 #include <bfd.h>
37 #include <dis-asm.h>
38 #include <inttypes.h>
39
40 #ifndef bool
41 #define bool int
42 #define true 1
43 #define false 0
44 #endif /*bool*/
45
46 /* short names for stuff in hsdis.h */
47 typedef decode_instructions_event_callback_ftype event_callback_t;
48 typedef decode_instructions_printf_callback_ftype printf_callback_t;
49
50 /* disassemble_info.application_data object */
51 struct hsdis_app_data {
52 /* the arguments to decode_instructions */
53 uintptr_t start; uintptr_t end;
54 event_callback_t event_callback; void* event_stream;
55 printf_callback_t printf_callback; void* printf_stream;
56 bool losing;
57
58 /* the architecture being disassembled */
59 const char* arch_name;
60 const bfd_arch_info_type* arch_info;
61
62 /* the disassembler we are going to use: */
63 disassembler_ftype dfn;
64 struct disassemble_info dinfo; /* the actual struct! */
65
66 char mach_option[64];
67 char insn_options[256];
68 };
69
70 #define DECL_APP_DATA(dinfo) \
71 struct hsdis_app_data* app_data = (struct hsdis_app_data*) (dinfo)->application_data
72
73 #define DECL_EVENT_CALLBACK(app_data) \
74 event_callback_t event_callback = (app_data)->event_callback; \
75 void* event_stream = (app_data)->event_stream
76
77 #define DECL_PRINTF_CALLBACK(app_data) \
78 printf_callback_t printf_callback = (app_data)->printf_callback; \
79 void* printf_stream = (app_data)->printf_stream
80
81
82 static void print_help(struct hsdis_app_data* app_data,
83 const char* msg, const char* arg);
84 static void setup_app_data(struct hsdis_app_data* app_data,
85 const char* options);
86 static const char* format_insn_close(const char* close,
87 disassemble_info* dinfo,
88 char* buf, size_t bufsize);
89
90 void*
91 #ifdef DLL_ENTRY
92 DLL_ENTRY
93 #endif
94 decode_instructions(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.event_callback = event_callback_arg;
103 app_data.event_stream = event_stream_arg;
104 app_data.printf_callback = printf_callback_arg;
105 app_data.printf_stream = printf_stream_arg;
106
107 setup_app_data(&app_data, options);
108 char buf[128];
109
110 {
111 /* now reload everything from app_data: */
112 DECL_EVENT_CALLBACK(&app_data);
113 DECL_PRINTF_CALLBACK(&app_data);
114 uintptr_t start = app_data.start;
115 uintptr_t end = app_data.end;
116 uintptr_t p = start;
117
118 (*event_callback)(event_stream, "insns", (void*)start);
119
120 (*event_callback)(event_stream, "mach name='%s'",
121 (void*) app_data.arch_info->printable_name);
122 if (app_data.dinfo.bytes_per_line != 0) {
123 (*event_callback)(event_stream, "format bytes-per-line='%p'/",
124 (void*)(intptr_t) app_data.dinfo.bytes_per_line);
125 }
126
127 while (p < end && !app_data.losing) {
128 (*event_callback)(event_stream, "insn", (void*) p);
129
130 /* reset certain state, so we can read it with confidence */
131 app_data.dinfo.insn_info_valid = 0;
132 app_data.dinfo.branch_delay_insns = 0;
133 app_data.dinfo.data_size = 0;
134 app_data.dinfo.insn_type = 0;
135
136 int size = (*app_data.dfn)((bfd_vma) p, &app_data.dinfo);
137
138 if (size > 0) p += size;
139 else app_data.losing = true;
140
141 const char* insn_close = format_insn_close("/insn", &app_data.dinfo,
142 buf, sizeof(buf));
143 (*event_callback)(event_stream, insn_close, (void*) p);
144
145 /* follow each complete insn by a nice newline */
146 (*printf_callback)(printf_stream, "\n");
147 }
148
149 (*event_callback)(event_stream, "/insns", (void*) p);
150 return (void*) p;
151 }
152 }
153
154 /* take the address of the function, for luck, and also test the typedef: */
155 const decode_instructions_ftype decode_instructions_address = &decode_instructions;
156
157 static const char* format_insn_close(const char* close,
158 disassemble_info* dinfo,
159 char* buf, size_t bufsize) {
160 if (!dinfo->insn_info_valid)
161 return close;
162 enum dis_insn_type itype = dinfo->insn_type;
163 int dsize = dinfo->data_size, delays = dinfo->branch_delay_insns;
164 if ((itype == dis_nonbranch && (dsize | delays) == 0)
165 || (strlen(close) + 3*20 > bufsize))
166 return close;
167
168 const char* type = "unknown";
169 switch (itype) {
170 case dis_nonbranch: type = NULL; break;
171 case dis_branch: type = "branch"; break;
172 case dis_condbranch: type = "condbranch"; break;
173 case dis_jsr: type = "jsr"; break;
174 case dis_condjsr: type = "condjsr"; break;
175 case dis_dref: type = "dref"; break;
176 case dis_dref2: type = "dref2"; break;
177 }
178
179 strcpy(buf, close);
180 char* p = buf;
181 if (type) sprintf(p += strlen(p), " type='%s'", type);
182 if (dsize) sprintf(p += strlen(p), " dsize='%d'", dsize);
183 if (delays) sprintf(p += strlen(p), " delay='%d'", delays);
184 return buf;
185 }
186
187 /* handler functions */
188
189 static int
190 hsdis_read_memory_func(bfd_vma memaddr,
191 bfd_byte* myaddr,
192 unsigned int length,
193 struct disassemble_info* dinfo) {
194 uintptr_t memaddr_p = (uintptr_t) memaddr;
195 DECL_APP_DATA(dinfo);
196 if (memaddr_p + length > app_data->end) {
197 /* read is out of bounds */
198 return EIO;
199 } else {
200 memcpy(myaddr, (bfd_byte*) memaddr_p, length);
201 return 0;
202 }
203 }
204
205 static void
206 hsdis_print_address_func(bfd_vma vma, struct disassemble_info* dinfo) {
207 /* the actual value to print: */
208 void* addr_value = (void*) (uintptr_t) vma;
209 DECL_APP_DATA(dinfo);
210 DECL_EVENT_CALLBACK(app_data);
211
212 /* issue the event: */
213 void* result =
214 (*event_callback)(event_stream, "addr/", addr_value);
215 if (result == NULL) {
216 /* event declined */
217 generic_print_address(vma, dinfo);
218 }
219 }
220
221
222 /* configuration */
223
224 static void set_optional_callbacks(struct hsdis_app_data* app_data);
225 static void parse_caller_options(struct hsdis_app_data* app_data,
226 const char* caller_options);
227 static const char* native_arch_name();
228 static enum bfd_endian native_endian();
229 static const bfd_arch_info_type* find_arch_info(const char* arch_nane);
230 static bfd* get_native_bfd(const bfd_arch_info_type* arch_info,
231 /* to avoid malloc: */
232 bfd* empty_bfd, bfd_target* empty_xvec);
233 static void init_disassemble_info_from_bfd(struct disassemble_info* dinfo,
234 void *stream,
235 fprintf_ftype fprintf_func,
236 bfd* bfd,
237 char* disassembler_options);
238 static void parse_fake_insn(disassembler_ftype dfn,
239 struct disassemble_info* dinfo);
240
241 static void setup_app_data(struct hsdis_app_data* app_data,
242 const char* caller_options) {
243 /* Make reasonable defaults for null callbacks.
244 A non-null stream for a null callback is assumed to be a FILE* for output.
245 Events are rendered as XML.
246 */
247 set_optional_callbacks(app_data);
248
249 /* Look into caller_options for anything interesting. */
250 if (caller_options != NULL)
251 parse_caller_options(app_data, caller_options);
252
253 /* Discover which architecture we are going to disassemble. */
254 app_data->arch_name = &app_data->mach_option[0];
255 if (app_data->arch_name[0] == '\0')
256 app_data->arch_name = native_arch_name();
257 app_data->arch_info = find_arch_info(app_data->arch_name);
258
259 /* Make a fake bfd to hold the arch. and byteorder info. */
260 struct {
261 bfd_target empty_xvec;
262 bfd empty_bfd;
263 } buf;
264 bfd* native_bfd = get_native_bfd(app_data->arch_info,
265 /* to avoid malloc: */
266 &buf.empty_bfd, &buf.empty_xvec);
267 init_disassemble_info_from_bfd(&app_data->dinfo,
268 app_data->printf_stream,
269 app_data->printf_callback,
270 native_bfd,
271 app_data->insn_options);
272
273 /* Finish linking together the various callback blocks. */
274 app_data->dinfo.application_data = (void*) app_data;
275 app_data->dfn = disassembler(native_bfd);
276 app_data->dinfo.print_address_func = hsdis_print_address_func;
277 app_data->dinfo.read_memory_func = hsdis_read_memory_func;
278
279 if (app_data->dfn == NULL) {
280 const char* bad = app_data->arch_name;
281 static bool complained;
282 if (bad == &app_data->mach_option[0])
283 print_help(app_data, "bad mach=%s", bad);
284 else if (!complained)
285 print_help(app_data, "bad native mach=%s; please port hsdis to this platform", bad);
286 complained = true;
287 /* must bail out */
288 app_data->losing = true;
289 return;
290 }
291
292 parse_fake_insn(app_data->dfn, &app_data->dinfo);
293 }
294
295
296 /* ignore all events, return a null */
297 static void* null_event_callback(void* ignore_stream, const char* ignore_event, void* arg) {
298 return NULL;
299 }
300
301 /* print all events as XML markup */
302 static void* xml_event_callback(void* stream, const char* event, void* arg) {
303 FILE* fp = (FILE*) stream;
304 #define NS_PFX "dis:"
305 if (event[0] != '/') {
306 /* issue the tag, with or without a formatted argument */
307 fprintf(fp, "<"NS_PFX);
308 fprintf(fp, event, arg);
309 fprintf(fp, ">");
310 } else {
311 ++event; /* skip slash */
312 const char* argp = strchr(event, ' ');
313 if (argp == NULL) {
314 /* no arguments; just issue the closing tag */
315 fprintf(fp, "</"NS_PFX"%s>", event);
316 } else {
317 /* split out the closing attributes as <dis:foo_done attr='val'/> */
318 int event_prefix = (argp - event);
319 fprintf(fp, "<"NS_PFX"%.*s_done", event_prefix, event);
320 fprintf(fp, argp, arg);
321 fprintf(fp, "/></"NS_PFX"%.*s>", event_prefix, event);
322 }
323 }
324 return NULL;
325 }
326
327 static void set_optional_callbacks(struct hsdis_app_data* app_data) {
328 if (app_data->printf_callback == NULL) {
329 int (*fprintf_callback)(FILE*, const char*, ...) = &fprintf;
330 FILE* fprintf_stream = stdout;
331 app_data->printf_callback = (printf_callback_t) fprintf_callback;
332 if (app_data->printf_stream == NULL)
333 app_data->printf_stream = (void*) fprintf_stream;
334 }
335 if (app_data->event_callback == NULL) {
336 if (app_data->event_stream == NULL)
337 app_data->event_callback = &null_event_callback;
338 else
339 app_data->event_callback = &xml_event_callback;
340 }
341
342 }
343
344 static void parse_caller_options(struct hsdis_app_data* app_data, const char* caller_options) {
345 char* iop_base = app_data->insn_options;
346 char* iop_limit = iop_base + sizeof(app_data->insn_options) - 1;
347 char* iop = iop_base;
348 const char* p;
349 for (p = caller_options; p != NULL; ) {
350 const char* q = strchr(p, ',');
351 size_t plen = (q == NULL) ? strlen(p) : ((q++) - p);
352 if (plen == 4 && strncmp(p, "help", plen) == 0) {
353 print_help(app_data, NULL, NULL);
354 } else if (plen >= 5 && strncmp(p, "mach=", 5) == 0) {
355 char* mach_option = app_data->mach_option;
356 size_t mach_size = sizeof(app_data->mach_option);
357 mach_size -= 1; /*leave room for the null*/
358 if (plen > mach_size) plen = mach_size;
359 strncpy(mach_option, p, plen);
360 mach_option[plen] = '\0';
361 } else if (plen > 6 && strncmp(p, "hsdis-", 6)) {
362 // do not pass these to the next level
363 } else {
364 /* just copy it; {i386,sparc}-dis.c might like to see it */
365 if (iop > iop_base && iop < iop_limit) (*iop++) = ',';
366 if (iop + plen > iop_limit)
367 plen = iop_limit - iop;
368 strncpy(iop, p, plen);
369 iop += plen;
370 }
371 p = q;
372 }
373 }
374
375 static void print_help(struct hsdis_app_data* app_data,
376 const char* msg, const char* arg) {
377 DECL_PRINTF_CALLBACK(app_data);
378 if (msg != NULL) {
379 (*printf_callback)(printf_stream, "hsdis: ");
380 (*printf_callback)(printf_stream, msg, arg);
381 (*printf_callback)(printf_stream, "\n");
382 }
383 (*printf_callback)(printf_stream, "hsdis output options:\n");
384 if (printf_callback == (printf_callback_t) &fprintf)
385 disassembler_usage((FILE*) printf_stream);
386 else
387 disassembler_usage(stderr); /* better than nothing */
388 (*printf_callback)(printf_stream, " mach=<arch> select disassembly mode\n");
389 #if defined(LIBARCH_i386) || defined(LIBARCH_amd64)
390 (*printf_callback)(printf_stream, " mach=i386 select 32-bit mode\n");
391 (*printf_callback)(printf_stream, " mach=x86-64 select 64-bit mode\n");
392 (*printf_callback)(printf_stream, " suffix always print instruction suffix\n");
393 #endif
394 (*printf_callback)(printf_stream, " help print this message\n");
395 }
396
397
398 /* low-level bfd and arch stuff that binutils doesn't do for us */
399
400 static const bfd_arch_info_type* find_arch_info(const char* arch_name) {
401 const bfd_arch_info_type* arch_info = bfd_scan_arch(arch_name);
402 if (arch_info == NULL) {
403 extern const bfd_arch_info_type bfd_default_arch_struct;
404 arch_info = &bfd_default_arch_struct;
405 }
406 return arch_info;
407 }
408
409 static const char* native_arch_name() {
410 const char* res = NULL;
411 #ifdef LIBARCH_i386
412 res = "i386";
413 #endif
414 #ifdef LIBARCH_amd64
415 res = "i386:x86-64";
416 #endif
417 #ifdef LIBARCH_sparc
418 res = "sparc:v8plusb";
419 #endif
420 #ifdef LIBARCH_sparcv9
421 res = "sparc:v9b";
422 #endif
423 if (res == NULL)
424 res = "architecture not set in Makefile!";
425 return res;
426 }
427
428 static enum bfd_endian native_endian() {
429 int32_t endian_test = 'x';
430 if (*(const char*) &endian_test == 'x')
431 return BFD_ENDIAN_LITTLE;
432 else
433 return BFD_ENDIAN_BIG;
434 }
435
436 static bfd* get_native_bfd(const bfd_arch_info_type* arch_info,
437 bfd* empty_bfd, bfd_target* empty_xvec) {
438 memset(empty_bfd, 0, sizeof(*empty_bfd));
439 memset(empty_xvec, 0, sizeof(*empty_xvec));
440 empty_xvec->flavour = bfd_target_unknown_flavour;
441 empty_xvec->byteorder = native_endian();
442 empty_bfd->xvec = empty_xvec;
443 empty_bfd->arch_info = arch_info;
444 return empty_bfd;
445 }
446
447 static int read_zero_data_only(bfd_vma ignore_p,
448 bfd_byte* myaddr, unsigned int length,
449 struct disassemble_info *ignore_info) {
450 memset(myaddr, 0, length);
451 return 0;
452 }
453 static int print_to_dev_null(void* ignore_stream, const char* ignore_format, ...) {
454 return 0;
455 }
456
457 /* Prime the pump by running the selected disassembler on a null input.
458 This forces the machine-specific disassembler to divulge invariant
459 information like bytes_per_line.
460 */
461 static void parse_fake_insn(disassembler_ftype dfn,
462 struct disassemble_info* dinfo) {
463 typedef int (*read_memory_ftype)
464 (bfd_vma memaddr, bfd_byte *myaddr, unsigned int length,
465 struct disassemble_info *info);
466 read_memory_ftype read_memory_func = dinfo->read_memory_func;
467 fprintf_ftype fprintf_func = dinfo->fprintf_func;
468
469 dinfo->read_memory_func = &read_zero_data_only;
470 dinfo->fprintf_func = &print_to_dev_null;
471 (*dfn)(0, dinfo);
472
473 // put it back:
474 dinfo->read_memory_func = read_memory_func;
475 dinfo->fprintf_func = fprintf_func;
476 }
477
478 static void init_disassemble_info_from_bfd(struct disassemble_info* dinfo,
479 void *stream,
480 fprintf_ftype fprintf_func,
481 bfd* abfd,
482 char* disassembler_options) {
483 init_disassemble_info(dinfo, stream, fprintf_func);
484
485 dinfo->flavour = bfd_get_flavour(abfd);
486 dinfo->arch = bfd_get_arch(abfd);
487 dinfo->mach = bfd_get_mach(abfd);
488 dinfo->disassembler_options = disassembler_options;
489 dinfo->octets_per_byte = bfd_octets_per_byte (abfd);
490 dinfo->skip_zeroes = sizeof(void*) * 2;
491 dinfo->skip_zeroes_at_end = sizeof(void*)-1;
492 dinfo->disassembler_needs_relocs = FALSE;
493
494 if (bfd_big_endian(abfd))
495 dinfo->display_endian = dinfo->endian = BFD_ENDIAN_BIG;
496 else if (bfd_little_endian(abfd))
497 dinfo->display_endian = dinfo->endian = BFD_ENDIAN_LITTLE;
498 else
499 dinfo->endian = native_endian();
500
501 disassemble_init_for_target(dinfo);
502 }