e61ec0d4bf3554112cb3101f9a556ccf119b3aea
[binutils-gdb.git] / gdb / solib-darwin.c
1 /* Handle Darwin shared libraries for GDB, the GNU Debugger.
2
3 Copyright (C) 2009-2022 Free Software Foundation, Inc.
4
5 This file is part of GDB.
6
7 This program is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 3 of the License, or
10 (at your option) any later version.
11
12 This program is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
16
17 You should have received a copy of the GNU General Public License
18 along with this program. If not, see <http://www.gnu.org/licenses/>. */
19
20 #include "defs.h"
21
22 #include "symtab.h"
23 #include "bfd.h"
24 #include "symfile.h"
25 #include "objfiles.h"
26 #include "gdbcore.h"
27 #include "target.h"
28 #include "inferior.h"
29 #include "regcache.h"
30 #include "gdbthread.h"
31 #include "gdb_bfd.h"
32
33 #include "solist.h"
34 #include "solib.h"
35 #include "solib-svr4.h"
36
37 #include "bfd-target.h"
38 #include "elf-bfd.h"
39 #include "exec.h"
40 #include "auxv.h"
41 #include "mach-o.h"
42 #include "mach-o/external.h"
43
44 struct gdb_dyld_image_info
45 {
46 /* Base address (which corresponds to the Mach-O header). */
47 CORE_ADDR mach_header;
48 /* Image file path. */
49 CORE_ADDR file_path;
50 /* st.m_time of image file. */
51 unsigned long mtime;
52 };
53
54 /* Content of inferior dyld_all_image_infos structure.
55 See /usr/include/mach-o/dyld_images.h for the documentation. */
56 struct gdb_dyld_all_image_infos
57 {
58 /* Version (1). */
59 unsigned int version;
60 /* Number of images. */
61 unsigned int count;
62 /* Image description. */
63 CORE_ADDR info;
64 /* Notifier (function called when a library is added or removed). */
65 CORE_ADDR notifier;
66 };
67
68 /* Current all_image_infos version. */
69 #define DYLD_VERSION_MIN 1
70 #define DYLD_VERSION_MAX 15
71
72 /* Per PSPACE specific data. */
73 struct darwin_info
74 {
75 /* Address of structure dyld_all_image_infos in inferior. */
76 CORE_ADDR all_image_addr = 0;
77
78 /* Gdb copy of dyld_all_info_infos. */
79 struct gdb_dyld_all_image_infos all_image {};
80 };
81
82 /* Per-program-space data key. */
83 static const registry<program_space>::key<darwin_info>
84 solib_darwin_pspace_data;
85
86 /* Get the current darwin data. If none is found yet, add it now. This
87 function always returns a valid object. */
88
89 static struct darwin_info *
90 get_darwin_info (void)
91 {
92 struct darwin_info *info;
93
94 info = solib_darwin_pspace_data.get (current_program_space);
95 if (info != NULL)
96 return info;
97
98 return solib_darwin_pspace_data.emplace (current_program_space);
99 }
100
101 /* Return non-zero if the version in dyld_all_image is known. */
102
103 static int
104 darwin_dyld_version_ok (const struct darwin_info *info)
105 {
106 return info->all_image.version >= DYLD_VERSION_MIN
107 && info->all_image.version <= DYLD_VERSION_MAX;
108 }
109
110 /* Read dyld_all_image from inferior. */
111
112 static void
113 darwin_load_image_infos (struct darwin_info *info)
114 {
115 gdb_byte buf[24];
116 enum bfd_endian byte_order = gdbarch_byte_order (target_gdbarch ());
117 struct type *ptr_type = builtin_type (target_gdbarch ())->builtin_data_ptr;
118 int len;
119
120 /* If the structure address is not known, don't continue. */
121 if (info->all_image_addr == 0)
122 return;
123
124 /* The structure has 4 fields: version (4 bytes), count (4 bytes),
125 info (pointer) and notifier (pointer). */
126 len = 4 + 4 + 2 * TYPE_LENGTH (ptr_type);
127 gdb_assert (len <= sizeof (buf));
128 memset (&info->all_image, 0, sizeof (info->all_image));
129
130 /* Read structure raw bytes from target. */
131 if (target_read_memory (info->all_image_addr, buf, len))
132 return;
133
134 /* Extract the fields. */
135 info->all_image.version = extract_unsigned_integer (buf, 4, byte_order);
136 if (!darwin_dyld_version_ok (info))
137 return;
138
139 info->all_image.count = extract_unsigned_integer (buf + 4, 4, byte_order);
140 info->all_image.info = extract_typed_address (buf + 8, ptr_type);
141 info->all_image.notifier = extract_typed_address
142 (buf + 8 + TYPE_LENGTH (ptr_type), ptr_type);
143 }
144
145 /* Link map info to include in an allocated so_list entry. */
146
147 struct lm_info_darwin : public lm_info_base
148 {
149 /* The target location of lm. */
150 CORE_ADDR lm_addr = 0;
151 };
152
153 /* Lookup the value for a specific symbol. */
154
155 static CORE_ADDR
156 lookup_symbol_from_bfd (bfd *abfd, const char *symname)
157 {
158 long storage_needed;
159 asymbol **symbol_table;
160 unsigned int number_of_symbols;
161 unsigned int i;
162 CORE_ADDR symaddr = 0;
163
164 storage_needed = bfd_get_symtab_upper_bound (abfd);
165
166 if (storage_needed <= 0)
167 return 0;
168
169 symbol_table = (asymbol **) xmalloc (storage_needed);
170 number_of_symbols = bfd_canonicalize_symtab (abfd, symbol_table);
171
172 for (i = 0; i < number_of_symbols; i++)
173 {
174 asymbol *sym = symbol_table[i];
175
176 if (strcmp (sym->name, symname) == 0
177 && (sym->section->flags & (SEC_CODE | SEC_DATA)) != 0)
178 {
179 /* BFD symbols are section relative. */
180 symaddr = sym->value + sym->section->vma;
181 break;
182 }
183 }
184 xfree (symbol_table);
185
186 return symaddr;
187 }
188
189 /* Return program interpreter string. */
190
191 static char *
192 find_program_interpreter (void)
193 {
194 char *buf = NULL;
195
196 /* If we have an current exec_bfd, get the interpreter from the load
197 commands. */
198 if (current_program_space->exec_bfd ())
199 {
200 bfd_mach_o_load_command *cmd;
201
202 if (bfd_mach_o_lookup_command (current_program_space->exec_bfd (),
203 BFD_MACH_O_LC_LOAD_DYLINKER, &cmd) == 1)
204 return cmd->command.dylinker.name_str;
205 }
206
207 /* If we didn't find it, read from memory.
208 FIXME: todo. */
209 return buf;
210 }
211
212 /* Not used. I don't see how the main symbol file can be found: the
213 interpreter name is needed and it is known from the executable file.
214 Note that darwin-nat.c implements pid_to_exec_file. */
215
216 static int
217 open_symbol_file_object (int from_tty)
218 {
219 return 0;
220 }
221
222 /* Build a list of currently loaded shared objects. See solib-svr4.c. */
223
224 static struct so_list *
225 darwin_current_sos (void)
226 {
227 struct type *ptr_type = builtin_type (target_gdbarch ())->builtin_data_ptr;
228 enum bfd_endian byte_order = type_byte_order (ptr_type);
229 int ptr_len = TYPE_LENGTH (ptr_type);
230 unsigned int image_info_size;
231 struct so_list *head = NULL;
232 struct so_list *tail = NULL;
233 int i;
234 struct darwin_info *info = get_darwin_info ();
235
236 /* Be sure image infos are loaded. */
237 darwin_load_image_infos (info);
238
239 if (!darwin_dyld_version_ok (info))
240 return NULL;
241
242 image_info_size = ptr_len * 3;
243
244 /* Read infos for each solib.
245 The first entry was rumored to be the executable itself, but this is not
246 true when a large number of shared libraries are used (table expanded ?).
247 We now check all entries, but discard executable images. */
248 for (i = 0; i < info->all_image.count; i++)
249 {
250 CORE_ADDR iinfo = info->all_image.info + i * image_info_size;
251 gdb_byte buf[image_info_size];
252 CORE_ADDR load_addr;
253 CORE_ADDR path_addr;
254 struct mach_o_header_external hdr;
255 unsigned long hdr_val;
256
257 /* Read image info from inferior. */
258 if (target_read_memory (iinfo, buf, image_info_size))
259 break;
260
261 load_addr = extract_typed_address (buf, ptr_type);
262 path_addr = extract_typed_address (buf + ptr_len, ptr_type);
263
264 /* Read Mach-O header from memory. */
265 if (target_read_memory (load_addr, (gdb_byte *) &hdr, sizeof (hdr) - 4))
266 break;
267 /* Discard wrong magic numbers. Shouldn't happen. */
268 hdr_val = extract_unsigned_integer
269 (hdr.magic, sizeof (hdr.magic), byte_order);
270 if (hdr_val != BFD_MACH_O_MH_MAGIC && hdr_val != BFD_MACH_O_MH_MAGIC_64)
271 continue;
272 /* Discard executable. Should happen only once. */
273 hdr_val = extract_unsigned_integer
274 (hdr.filetype, sizeof (hdr.filetype), byte_order);
275 if (hdr_val == BFD_MACH_O_MH_EXECUTE)
276 continue;
277
278 gdb::unique_xmalloc_ptr<char> file_path
279 = target_read_string (path_addr, SO_NAME_MAX_PATH_SIZE - 1);
280 if (file_path == nullptr)
281 break;
282
283 /* Create and fill the new so_list element. */
284 gdb::unique_xmalloc_ptr<struct so_list> newobj (XCNEW (struct so_list));
285
286 lm_info_darwin *li = new lm_info_darwin;
287 newobj->lm_info = li;
288
289 strncpy (newobj->so_name, file_path.get (), SO_NAME_MAX_PATH_SIZE - 1);
290 newobj->so_name[SO_NAME_MAX_PATH_SIZE - 1] = '\0';
291 strcpy (newobj->so_original_name, newobj->so_name);
292 li->lm_addr = load_addr;
293
294 if (head == NULL)
295 head = newobj.get ();
296 else
297 tail->next = newobj.get ();
298 tail = newobj.release ();
299 }
300
301 return head;
302 }
303
304 /* Check LOAD_ADDR points to a Mach-O executable header. Return LOAD_ADDR
305 in case of success, 0 in case of failure. */
306
307 static CORE_ADDR
308 darwin_validate_exec_header (CORE_ADDR load_addr)
309 {
310 enum bfd_endian byte_order = gdbarch_byte_order (target_gdbarch ());
311 struct mach_o_header_external hdr;
312 unsigned long hdr_val;
313
314 /* Read Mach-O header from memory. */
315 if (target_read_memory (load_addr, (gdb_byte *) &hdr, sizeof (hdr) - 4))
316 return 0;
317
318 /* Discard wrong magic numbers. Shouldn't happen. */
319 hdr_val = extract_unsigned_integer
320 (hdr.magic, sizeof (hdr.magic), byte_order);
321 if (hdr_val != BFD_MACH_O_MH_MAGIC && hdr_val != BFD_MACH_O_MH_MAGIC_64)
322 return 0;
323
324 /* Check executable. */
325 hdr_val = extract_unsigned_integer
326 (hdr.filetype, sizeof (hdr.filetype), byte_order);
327 if (hdr_val == BFD_MACH_O_MH_EXECUTE)
328 return load_addr;
329
330 return 0;
331 }
332
333 /* Get the load address of the executable using dyld list of images.
334 We assume that the dyld info are correct (which is wrong if the target
335 is stopped at the first instruction). */
336
337 static CORE_ADDR
338 darwin_read_exec_load_addr_from_dyld (struct darwin_info *info)
339 {
340 struct type *ptr_type = builtin_type (target_gdbarch ())->builtin_data_ptr;
341 int ptr_len = TYPE_LENGTH (ptr_type);
342 unsigned int image_info_size = ptr_len * 3;
343 int i;
344
345 /* Read infos for each solib. One of them should be the executable. */
346 for (i = 0; i < info->all_image.count; i++)
347 {
348 CORE_ADDR iinfo = info->all_image.info + i * image_info_size;
349 gdb_byte buf[image_info_size];
350 CORE_ADDR load_addr;
351
352 /* Read image info from inferior. */
353 if (target_read_memory (iinfo, buf, image_info_size))
354 break;
355
356 load_addr = extract_typed_address (buf, ptr_type);
357 if (darwin_validate_exec_header (load_addr) == load_addr)
358 return load_addr;
359 }
360
361 return 0;
362 }
363
364 /* Get the load address of the executable when the PC is at the dyld
365 entry point using parameter passed by the kernel (at SP). */
366
367 static CORE_ADDR
368 darwin_read_exec_load_addr_at_init (struct darwin_info *info)
369 {
370 struct gdbarch *gdbarch = target_gdbarch ();
371 enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);
372 int addr_size = gdbarch_addr_bit (gdbarch) / 8;
373 ULONGEST load_ptr_addr;
374 ULONGEST load_addr;
375 gdb_byte buf[8];
376
377 /* Get SP. */
378 if (regcache_cooked_read_unsigned (get_current_regcache (),
379 gdbarch_sp_regnum (gdbarch),
380 &load_ptr_addr) != REG_VALID)
381 return 0;
382
383 /* Read value at SP (image load address). */
384 if (target_read_memory (load_ptr_addr, buf, addr_size))
385 return 0;
386
387 load_addr = extract_unsigned_integer (buf, addr_size, byte_order);
388
389 return darwin_validate_exec_header (load_addr);
390 }
391
392 /* Return 1 if PC lies in the dynamic symbol resolution code of the
393 run time loader. */
394
395 static int
396 darwin_in_dynsym_resolve_code (CORE_ADDR pc)
397 {
398 return 0;
399 }
400
401 /* A wrapper for bfd_mach_o_fat_extract that handles reference
402 counting properly. This will either return NULL, or return a new
403 reference to a BFD. */
404
405 static gdb_bfd_ref_ptr
406 gdb_bfd_mach_o_fat_extract (bfd *abfd, bfd_format format,
407 const bfd_arch_info_type *arch)
408 {
409 bfd *result = bfd_mach_o_fat_extract (abfd, format, arch);
410
411 if (result == NULL)
412 return NULL;
413
414 if (result == abfd)
415 gdb_bfd_ref (result);
416 else
417 gdb_bfd_mark_parent (result, abfd);
418
419 return gdb_bfd_ref_ptr (result);
420 }
421
422 /* Return the BFD for the program interpreter. */
423
424 static gdb_bfd_ref_ptr
425 darwin_get_dyld_bfd ()
426 {
427 char *interp_name;
428
429 /* This method doesn't work with an attached process. */
430 if (current_inferior ()->attach_flag)
431 return NULL;
432
433 /* Find the program interpreter. */
434 interp_name = find_program_interpreter ();
435 if (!interp_name)
436 return NULL;
437
438 /* Create a bfd for the interpreter. */
439 gdb_bfd_ref_ptr dyld_bfd (gdb_bfd_open (interp_name, gnutarget));
440 if (dyld_bfd != NULL)
441 {
442 gdb_bfd_ref_ptr sub
443 (gdb_bfd_mach_o_fat_extract (dyld_bfd.get (), bfd_object,
444 gdbarch_bfd_arch_info (target_gdbarch ())));
445 dyld_bfd = sub;
446 }
447 return dyld_bfd;
448 }
449
450 /* Extract dyld_all_image_addr when the process was just created, assuming the
451 current PC is at the entry of the dynamic linker. */
452
453 static void
454 darwin_solib_get_all_image_info_addr_at_init (struct darwin_info *info)
455 {
456 CORE_ADDR load_addr = 0;
457 gdb_bfd_ref_ptr dyld_bfd = darwin_get_dyld_bfd ();
458
459 if (dyld_bfd == NULL)
460 return;
461
462 /* We find the dynamic linker's base address by examining
463 the current pc (which should point at the entry point for the
464 dynamic linker) and subtracting the offset of the entry point. */
465 load_addr = (regcache_read_pc (get_current_regcache ())
466 - bfd_get_start_address (dyld_bfd.get ()));
467
468 /* Now try to set a breakpoint in the dynamic linker. */
469 info->all_image_addr =
470 lookup_symbol_from_bfd (dyld_bfd.get (), "_dyld_all_image_infos");
471
472 if (info->all_image_addr == 0)
473 return;
474
475 info->all_image_addr += load_addr;
476 }
477
478 /* Extract dyld_all_image_addr reading it from
479 TARGET_OBJECT_DARWIN_DYLD_INFO. */
480
481 static void
482 darwin_solib_read_all_image_info_addr (struct darwin_info *info)
483 {
484 gdb_byte buf[8];
485 LONGEST len;
486 struct type *ptr_type = builtin_type (target_gdbarch ())->builtin_data_ptr;
487
488 /* Sanity check. */
489 if (TYPE_LENGTH (ptr_type) > sizeof (buf))
490 return;
491
492 len = target_read (current_inferior ()->top_target (),
493 TARGET_OBJECT_DARWIN_DYLD_INFO,
494 NULL, buf, 0, TYPE_LENGTH (ptr_type));
495 if (len <= 0)
496 return;
497
498 /* The use of BIG endian is intended, as BUF is a raw stream of bytes. This
499 makes the support of remote protocol easier. */
500 info->all_image_addr = extract_unsigned_integer (buf, len, BFD_ENDIAN_BIG);
501 }
502
503 /* Shared library startup support. See documentation in solib-svr4.c. */
504
505 static void
506 darwin_solib_create_inferior_hook (int from_tty)
507 {
508 /* Everything below only makes sense if we have a running inferior. */
509 if (!target_has_execution ())
510 return;
511
512 struct darwin_info *info = get_darwin_info ();
513 CORE_ADDR load_addr;
514
515 info->all_image_addr = 0;
516
517 darwin_solib_read_all_image_info_addr (info);
518
519 if (info->all_image_addr == 0)
520 darwin_solib_get_all_image_info_addr_at_init (info);
521
522 if (info->all_image_addr == 0)
523 return;
524
525 darwin_load_image_infos (info);
526
527 if (!darwin_dyld_version_ok (info))
528 {
529 warning (_("unhandled dyld version (%d)"), info->all_image.version);
530 return;
531 }
532
533 if (info->all_image.count != 0)
534 {
535 /* Possible relocate the main executable (PIE). */
536 load_addr = darwin_read_exec_load_addr_from_dyld (info);
537 }
538 else
539 {
540 /* Possible issue:
541 Do not break on the notifier if dyld is not initialized (deduced from
542 count == 0). In that case, dyld hasn't relocated itself and the
543 notifier may point to a wrong address. */
544
545 load_addr = darwin_read_exec_load_addr_at_init (info);
546 }
547
548 if (load_addr != 0 && current_program_space->symfile_object_file != NULL)
549 {
550 CORE_ADDR vmaddr;
551
552 /* Find the base address of the executable. */
553 vmaddr = bfd_mach_o_get_base_address (current_program_space->exec_bfd ());
554
555 /* Relocate. */
556 if (vmaddr != load_addr)
557 objfile_rebase (current_program_space->symfile_object_file,
558 load_addr - vmaddr);
559 }
560
561 /* Set solib notifier (to reload list of shared libraries). */
562 CORE_ADDR notifier = info->all_image.notifier;
563
564 if (info->all_image.count == 0)
565 {
566 /* Dyld hasn't yet relocated itself, so the notifier address may
567 be incorrect (as it has to be relocated). */
568 CORE_ADDR start
569 = bfd_get_start_address (current_program_space->exec_bfd ());
570 if (start == 0)
571 notifier = 0;
572 else
573 {
574 gdb_bfd_ref_ptr dyld_bfd = darwin_get_dyld_bfd ();
575 if (dyld_bfd != NULL)
576 {
577 CORE_ADDR dyld_bfd_start_address;
578 CORE_ADDR dyld_relocated_base_address;
579 CORE_ADDR pc;
580
581 dyld_bfd_start_address = bfd_get_start_address (dyld_bfd.get());
582
583 /* We find the dynamic linker's base address by examining
584 the current pc (which should point at the entry point
585 for the dynamic linker) and subtracting the offset of
586 the entry point. */
587
588 pc = regcache_read_pc (get_current_regcache ());
589 dyld_relocated_base_address = pc - dyld_bfd_start_address;
590
591 /* We get the proper notifier relocated address by
592 adding the dyld relocated base address to the current
593 notifier offset value. */
594
595 notifier += dyld_relocated_base_address;
596 }
597 }
598 }
599
600 /* Add the breakpoint which is hit by dyld when the list of solib is
601 modified. */
602 if (notifier != 0)
603 create_solib_event_breakpoint (target_gdbarch (), notifier);
604 }
605
606 static void
607 darwin_clear_solib (void)
608 {
609 struct darwin_info *info = get_darwin_info ();
610
611 info->all_image_addr = 0;
612 info->all_image.version = 0;
613 }
614
615 static void
616 darwin_free_so (struct so_list *so)
617 {
618 lm_info_darwin *li = (lm_info_darwin *) so->lm_info;
619
620 delete li;
621 }
622
623 /* The section table is built from bfd sections using bfd VMAs.
624 Relocate these VMAs according to solib info. */
625
626 static void
627 darwin_relocate_section_addresses (struct so_list *so,
628 struct target_section *sec)
629 {
630 lm_info_darwin *li = (lm_info_darwin *) so->lm_info;
631
632 sec->addr += li->lm_addr;
633 sec->endaddr += li->lm_addr;
634
635 /* Best effort to set addr_high/addr_low. This is used only by
636 'info sharedlibary'. */
637 if (so->addr_high == 0)
638 {
639 so->addr_low = sec->addr;
640 so->addr_high = sec->endaddr;
641 }
642 if (sec->endaddr > so->addr_high)
643 so->addr_high = sec->endaddr;
644 if (sec->addr < so->addr_low)
645 so->addr_low = sec->addr;
646 }
647 \f
648 static gdb_bfd_ref_ptr
649 darwin_bfd_open (const char *pathname)
650 {
651 int found_file;
652
653 /* Search for shared library file. */
654 gdb::unique_xmalloc_ptr<char> found_pathname
655 = solib_find (pathname, &found_file);
656 if (found_pathname == NULL)
657 perror_with_name (pathname);
658
659 /* Open bfd for shared library. */
660 gdb_bfd_ref_ptr abfd (solib_bfd_fopen (found_pathname.get (), found_file));
661
662 gdb_bfd_ref_ptr res
663 (gdb_bfd_mach_o_fat_extract (abfd.get (), bfd_object,
664 gdbarch_bfd_arch_info (target_gdbarch ())));
665 if (res == NULL)
666 error (_("`%s': not a shared-library: %s"),
667 bfd_get_filename (abfd.get ()), bfd_errmsg (bfd_get_error ()));
668
669 /* The current filename for fat-binary BFDs is a name generated
670 by BFD, usually a string containing the name of the architecture.
671 Reset its value to the actual filename. */
672 bfd_set_filename (res.get (), pathname);
673
674 return res;
675 }
676
677 struct target_so_ops darwin_so_ops;
678
679 void _initialize_darwin_solib ();
680 void
681 _initialize_darwin_solib ()
682 {
683 darwin_so_ops.relocate_section_addresses = darwin_relocate_section_addresses;
684 darwin_so_ops.free_so = darwin_free_so;
685 darwin_so_ops.clear_solib = darwin_clear_solib;
686 darwin_so_ops.solib_create_inferior_hook = darwin_solib_create_inferior_hook;
687 darwin_so_ops.current_sos = darwin_current_sos;
688 darwin_so_ops.open_symbol_file_object = open_symbol_file_object;
689 darwin_so_ops.in_dynsym_resolve_code = darwin_in_dynsym_resolve_code;
690 darwin_so_ops.bfd_open = darwin_bfd_open;
691 }