/* Support for generating PDB CodeView debugging files. Copyright (C) 2022-2023 Free Software Foundation, Inc. This file is part of the GNU Binutils. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, MA 02110-1301, USA. */ #include "pdb.h" #include "bfdlink.h" #include "ld.h" #include "ldmain.h" #include "ldmisc.h" #include "libbfd.h" #include "libiberty.h" #include "coff/i386.h" #include "coff/external.h" #include "coff/internal.h" #include "coff/pe.h" #include "libcoff.h" #include struct public { struct public *next; uint32_t offset; uint32_t hash; unsigned int index; uint16_t section; uint32_t address; }; struct string { struct string *next; uint32_t hash; uint32_t offset; uint32_t source_file_offset; size_t len; char s[]; }; struct string_table { struct string *strings_head; struct string *strings_tail; uint32_t strings_len; htab_t hashmap; }; struct mod_source_files { uint16_t files_count; struct string **files; }; struct source_files_info { uint16_t mod_count; struct mod_source_files *mods; }; struct type_entry { struct type_entry *next; uint32_t index; uint32_t cv_hash; bool has_udt_src_line; uint8_t data[]; }; struct types { htab_t hashmap; uint32_t num_types; struct type_entry *first; struct type_entry *last; }; struct global { struct global *next; uint32_t offset; uint32_t hash; uint32_t refcount; unsigned int index; uint8_t data[]; }; struct globals { uint32_t num_entries; struct global *first; struct global *last; htab_t hashmap; }; struct in_sc { asection *s; uint16_t sect_num; uint16_t mod_index; }; static const uint32_t crc_table[] = { 0x00000000, 0x77073096, 0xee0e612c, 0x990951ba, 0x076dc419, 0x706af48f, 0xe963a535, 0x9e6495a3, 0x0edb8832, 0x79dcb8a4, 0xe0d5e91e, 0x97d2d988, 0x09b64c2b, 0x7eb17cbd, 0xe7b82d07, 0x90bf1d91, 0x1db71064, 0x6ab020f2, 0xf3b97148, 0x84be41de, 0x1adad47d, 0x6ddde4eb, 0xf4d4b551, 0x83d385c7, 0x136c9856, 0x646ba8c0, 0xfd62f97a, 0x8a65c9ec, 0x14015c4f, 0x63066cd9, 0xfa0f3d63, 0x8d080df5, 0x3b6e20c8, 0x4c69105e, 0xd56041e4, 0xa2677172, 0x3c03e4d1, 0x4b04d447, 0xd20d85fd, 0xa50ab56b, 0x35b5a8fa, 0x42b2986c, 0xdbbbc9d6, 0xacbcf940, 0x32d86ce3, 0x45df5c75, 0xdcd60dcf, 0xabd13d59, 0x26d930ac, 0x51de003a, 0xc8d75180, 0xbfd06116, 0x21b4f4b5, 0x56b3c423, 0xcfba9599, 0xb8bda50f, 0x2802b89e, 0x5f058808, 0xc60cd9b2, 0xb10be924, 0x2f6f7c87, 0x58684c11, 0xc1611dab, 0xb6662d3d, 0x76dc4190, 0x01db7106, 0x98d220bc, 0xefd5102a, 0x71b18589, 0x06b6b51f, 0x9fbfe4a5, 0xe8b8d433, 0x7807c9a2, 0x0f00f934, 0x9609a88e, 0xe10e9818, 0x7f6a0dbb, 0x086d3d2d, 0x91646c97, 0xe6635c01, 0x6b6b51f4, 0x1c6c6162, 0x856530d8, 0xf262004e, 0x6c0695ed, 0x1b01a57b, 0x8208f4c1, 0xf50fc457, 0x65b0d9c6, 0x12b7e950, 0x8bbeb8ea, 0xfcb9887c, 0x62dd1ddf, 0x15da2d49, 0x8cd37cf3, 0xfbd44c65, 0x4db26158, 0x3ab551ce, 0xa3bc0074, 0xd4bb30e2, 0x4adfa541, 0x3dd895d7, 0xa4d1c46d, 0xd3d6f4fb, 0x4369e96a, 0x346ed9fc, 0xad678846, 0xda60b8d0, 0x44042d73, 0x33031de5, 0xaa0a4c5f, 0xdd0d7cc9, 0x5005713c, 0x270241aa, 0xbe0b1010, 0xc90c2086, 0x5768b525, 0x206f85b3, 0xb966d409, 0xce61e49f, 0x5edef90e, 0x29d9c998, 0xb0d09822, 0xc7d7a8b4, 0x59b33d17, 0x2eb40d81, 0xb7bd5c3b, 0xc0ba6cad, 0xedb88320, 0x9abfb3b6, 0x03b6e20c, 0x74b1d29a, 0xead54739, 0x9dd277af, 0x04db2615, 0x73dc1683, 0xe3630b12, 0x94643b84, 0x0d6d6a3e, 0x7a6a5aa8, 0xe40ecf0b, 0x9309ff9d, 0x0a00ae27, 0x7d079eb1, 0xf00f9344, 0x8708a3d2, 0x1e01f268, 0x6906c2fe, 0xf762575d, 0x806567cb, 0x196c3671, 0x6e6b06e7, 0xfed41b76, 0x89d32be0, 0x10da7a5a, 0x67dd4acc, 0xf9b9df6f, 0x8ebeeff9, 0x17b7be43, 0x60b08ed5, 0xd6d6a3e8, 0xa1d1937e, 0x38d8c2c4, 0x4fdff252, 0xd1bb67f1, 0xa6bc5767, 0x3fb506dd, 0x48b2364b, 0xd80d2bda, 0xaf0a1b4c, 0x36034af6, 0x41047a60, 0xdf60efc3, 0xa867df55, 0x316e8eef, 0x4669be79, 0xcb61b38c, 0xbc66831a, 0x256fd2a0, 0x5268e236, 0xcc0c7795, 0xbb0b4703, 0x220216b9, 0x5505262f, 0xc5ba3bbe, 0xb2bd0b28, 0x2bb45a92, 0x5cb36a04, 0xc2d7ffa7, 0xb5d0cf31, 0x2cd99e8b, 0x5bdeae1d, 0x9b64c2b0, 0xec63f226, 0x756aa39c, 0x026d930a, 0x9c0906a9, 0xeb0e363f, 0x72076785, 0x05005713, 0x95bf4a82, 0xe2b87a14, 0x7bb12bae, 0x0cb61b38, 0x92d28e9b, 0xe5d5be0d, 0x7cdcefb7, 0x0bdbdf21, 0x86d3d2d4, 0xf1d4e242, 0x68ddb3f8, 0x1fda836e, 0x81be16cd, 0xf6b9265b, 0x6fb077e1, 0x18b74777, 0x88085ae6, 0xff0f6a70, 0x66063bca, 0x11010b5c, 0x8f659eff, 0xf862ae69, 0x616bffd3, 0x166ccf45, 0xa00ae278, 0xd70dd2ee, 0x4e048354, 0x3903b3c2, 0xa7672661, 0xd06016f7, 0x4969474d, 0x3e6e77db, 0xaed16a4a, 0xd9d65adc, 0x40df0b66, 0x37d83bf0, 0xa9bcae53, 0xdebb9ec5, 0x47b2cf7f, 0x30b5ffe9, 0xbdbdf21c, 0xcabac28a, 0x53b39330, 0x24b4a3a6, 0xbad03605, 0xcdd70693, 0x54de5729, 0x23d967bf, 0xb3667a2e, 0xc4614ab8, 0x5d681b02, 0x2a6f2b94, 0xb40bbe37, 0xc30c8ea1, 0x5a05df1b, 0x2d02ef8d }; /* Add a new stream to the PDB archive, and return its BFD. */ static bfd * add_stream (bfd *pdb, const char *name, uint16_t *stream_num) { bfd *stream; uint16_t num; stream = bfd_create (name ? name : "", pdb); if (!stream) return NULL; if (!bfd_make_writable (stream)) { bfd_close (stream); return false; } if (!pdb->archive_head) { bfd_set_archive_head (pdb, stream); num = 0; } else { bfd *b = pdb->archive_head; num = 1; while (b->archive_next) { num++; b = b->archive_next; } b->archive_next = stream; } if (stream_num) *stream_num = num; return stream; } /* Stream 0 ought to be a copy of the MSF directory from the last time the PDB file was written. Because we don't do incremental writes this isn't applicable to us, but we fill it with a dummy value so as not to confuse radare. */ static bool create_old_directory_stream (bfd *pdb) { bfd *stream; char buf[sizeof (uint32_t)]; stream = add_stream (pdb, NULL, NULL); if (!stream) return false; bfd_putl32 (0, buf); return bfd_bwrite (buf, sizeof (uint32_t), stream) == sizeof (uint32_t); } /* Calculate the hash of a given string. */ static uint32_t calc_hash (const char *data, size_t len) { uint32_t hash = 0; while (len >= 4) { hash ^= data[0]; hash ^= data[1] << 8; hash ^= data[2] << 16; hash ^= data[3] << 24; data += 4; len -= 4; } if (len >= 2) { hash ^= data[0]; hash ^= data[1] << 8; data += 2; len -= 2; } if (len != 0) hash ^= *data; hash |= 0x20202020; hash ^= (hash >> 11); return hash ^ (hash >> 16); } /* Stream 1 is the PDB info stream - see https://llvm.org/docs/PDB/PdbStream.html. */ static bool populate_info_stream (bfd *pdb, bfd *info_stream, const unsigned char *guid) { bool ret = false; struct pdb_stream_70 h; uint32_t num_entries, num_buckets; uint32_t names_length, stream_num; char int_buf[sizeof (uint32_t)]; struct hash_entry { uint32_t offset; uint32_t value; }; struct hash_entry **buckets = NULL; /* Write header. */ bfd_putl32 (PDB_STREAM_VERSION_VC70, &h.version); bfd_putl32 (time (NULL), &h.signature); bfd_putl32 (1, &h.age); bfd_putl32 (bfd_getb32 (guid), h.guid); bfd_putl16 (bfd_getb16 (&guid[4]), &h.guid[4]); bfd_putl16 (bfd_getb16 (&guid[6]), &h.guid[6]); memcpy (&h.guid[8], &guid[8], 8); if (bfd_bwrite (&h, sizeof (h), info_stream) != sizeof (h)) return false; /* Write hash list of named streams. This is a "rollover" hash, i.e. if a bucket is filled an entry gets placed in the next free slot. */ num_entries = 0; for (bfd *b = pdb->archive_head; b; b = b->archive_next) { if (strcmp (b->filename, "")) num_entries++; } num_buckets = num_entries * 2; names_length = 0; stream_num = 0; if (num_buckets > 0) { buckets = xmalloc (sizeof (struct hash_entry *) * num_buckets); memset (buckets, 0, sizeof (struct hash_entry *) * num_buckets); for (bfd *b = pdb->archive_head; b; b = b->archive_next) { if (strcmp (b->filename, "")) { size_t len = strlen (b->filename); uint32_t hash = (uint16_t) calc_hash (b->filename, len); uint32_t bucket_num = hash % num_buckets; while (buckets[bucket_num]) { bucket_num++; if (bucket_num == num_buckets) bucket_num = 0; } buckets[bucket_num] = xmalloc (sizeof (struct hash_entry)); buckets[bucket_num]->offset = names_length; buckets[bucket_num]->value = stream_num; names_length += len + 1; } stream_num++; } } /* Write the strings list - the hash keys are indexes into this. */ bfd_putl32 (names_length, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), info_stream) != sizeof (uint32_t)) goto end; for (bfd *b = pdb->archive_head; b; b = b->archive_next) { if (!strcmp (b->filename, "")) continue; size_t len = strlen (b->filename) + 1; if (bfd_bwrite (b->filename, len, info_stream) != len) goto end; } /* Write the number of entries and buckets. */ bfd_putl32 (num_entries, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), info_stream) != sizeof (uint32_t)) goto end; bfd_putl32 (num_buckets, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), info_stream) != sizeof (uint32_t)) goto end; /* Write the present bitmap. */ bfd_putl32 ((num_buckets + 31) / 32, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), info_stream) != sizeof (uint32_t)) goto end; for (unsigned int i = 0; i < num_buckets; i += 32) { uint32_t v = 0; for (unsigned int j = 0; j < 32; j++) { if (i + j >= num_buckets) break; if (buckets[i + j]) v |= 1 << j; } bfd_putl32 (v, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), info_stream) != sizeof (uint32_t)) goto end; } /* Write the (empty) deleted bitmap. */ bfd_putl32 (0, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), info_stream) != sizeof (uint32_t)) goto end; /* Write the buckets. */ for (unsigned int i = 0; i < num_buckets; i++) { if (buckets[i]) { bfd_putl32 (buckets[i]->offset, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), info_stream) != sizeof (uint32_t)) goto end; bfd_putl32 (buckets[i]->value, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), info_stream) != sizeof (uint32_t)) goto end; } } bfd_putl32 (0, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), info_stream) != sizeof (uint32_t)) goto end; bfd_putl32 (PDB_STREAM_VERSION_VC140, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), info_stream) != sizeof (uint32_t)) goto end; ret = true; end: for (unsigned int i = 0; i < num_buckets; i++) { if (buckets[i]) free (buckets[i]); } free (buckets); return ret; } /* Calculate the CRC32 used for type hashes. */ static uint32_t crc32 (const uint8_t *data, size_t len) { uint32_t crc = 0; while (len > 0) { crc = (crc >> 8) ^ crc_table[(crc & 0xff) ^ *data]; data++; len--; } return crc; } /* Stream 2 is the type information (TPI) stream, and stream 4 is the ID information (IPI) stream. They differ only in which records go in which stream. */ static bool populate_type_stream (bfd *pdb, bfd *stream, struct types *types) { struct pdb_tpi_stream_header h; struct type_entry *e; uint32_t len = 0, index_offset_len, off; struct bfd *hash_stream = NULL; uint16_t hash_stream_index; static const uint32_t index_skip = 0x2000; e = types->first; index_offset_len = 0; while (e) { uint32_t old_len = len; len += sizeof (uint16_t) + bfd_getl16 (e->data); if (old_len == 0 || old_len / index_skip != len / index_skip) index_offset_len += sizeof (uint32_t) * 2; e = e->next; } /* Each type stream also has a stream which holds the hash value for each type, along with a skip list to speed up searching. */ hash_stream = add_stream (pdb, "", &hash_stream_index); if (!hash_stream) return false; bfd_putl32 (TPI_STREAM_VERSION_80, &h.version); bfd_putl32 (sizeof (h), &h.header_size); bfd_putl32 (TPI_FIRST_INDEX, &h.type_index_begin); bfd_putl32 (TPI_FIRST_INDEX + types->num_types, &h.type_index_end); bfd_putl32 (len, &h.type_record_bytes); bfd_putl16 (hash_stream_index, &h.hash_stream_index); bfd_putl16 (0xffff, &h.hash_aux_stream_index); bfd_putl32 (sizeof (uint32_t), &h.hash_key_size); bfd_putl32 (NUM_TPI_HASH_BUCKETS, &h.num_hash_buckets); bfd_putl32 (0, &h.hash_value_buffer_offset); bfd_putl32 (types->num_types * sizeof (uint32_t), &h.hash_value_buffer_length); bfd_putl32 (types->num_types * sizeof (uint32_t), &h.index_offset_buffer_offset); bfd_putl32 (index_offset_len, &h.index_offset_buffer_length); bfd_putl32 ((types->num_types * sizeof (uint32_t)) + index_offset_len, &h.hash_adj_buffer_offset); bfd_putl32 (0, &h.hash_adj_buffer_length); if (bfd_bwrite (&h, sizeof (h), stream) != sizeof (h)) return false; /* Write the type definitions into the main stream, and the hashes into the hash stream. The hashes have already been calculated in handle_type. */ e = types->first; while (e) { uint8_t buf[sizeof (uint32_t)]; uint16_t size; size = bfd_getl16 (e->data); if (bfd_bwrite (e->data, size + sizeof (uint16_t), stream) != size + sizeof (uint16_t)) return false; bfd_putl32 (e->cv_hash % NUM_TPI_HASH_BUCKETS, buf); if (bfd_bwrite (buf, sizeof (uint32_t), hash_stream) != sizeof (uint32_t)) return false; e = e->next; } /* Write the index offsets, i.e. the skip list, into the hash stream. We copy MSVC here by writing a new entry for every 8192 bytes. */ e = types->first; off = 0; while (e) { uint32_t old_off = off; uint16_t size = bfd_getl16 (e->data); off += size + sizeof (uint16_t); if (old_off == 0 || old_off / index_skip != len / index_skip) { uint8_t buf[sizeof (uint32_t)]; bfd_putl32 (TPI_FIRST_INDEX + e->index, buf); if (bfd_bwrite (buf, sizeof (uint32_t), hash_stream) != sizeof (uint32_t)) return false; bfd_putl32 (old_off, buf); if (bfd_bwrite (buf, sizeof (uint32_t), hash_stream) != sizeof (uint32_t)) return false; } e = e->next; } return true; } /* Return the PE architecture number for the image. */ static uint16_t get_arch_number (bfd *abfd) { switch (abfd->arch_info->arch) { case bfd_arch_i386: if (abfd->arch_info->mach & bfd_mach_x86_64) return IMAGE_FILE_MACHINE_AMD64; else return IMAGE_FILE_MACHINE_I386; case bfd_arch_aarch64: return IMAGE_FILE_MACHINE_ARM64; default: return 0; } } /* Validate the DEBUG_S_FILECHKSMS entry within a module's .debug$S section, and copy it to the module's symbol stream. */ static bool copy_filechksms (uint8_t *data, uint32_t size, char *string_table, struct string_table *strings, uint8_t *out, struct mod_source_files *mod_source) { uint8_t *orig_data = data; uint32_t orig_size = size; uint16_t num_files = 0; struct string **strptr; bfd_putl32 (DEBUG_S_FILECHKSMS, out); out += sizeof (uint32_t); bfd_putl32 (size, out); out += sizeof (uint32_t); /* Calculate the number of files, and check for any overflows. */ while (size > 0) { struct file_checksum *fc = (struct file_checksum *) data; uint8_t padding; size_t len; if (size < sizeof (struct file_checksum)) { bfd_set_error (bfd_error_bad_value); return false; } len = sizeof (struct file_checksum) + fc->checksum_length; if (size < len) { bfd_set_error (bfd_error_bad_value); return false; } data += len; size -= len; if (len % sizeof (uint32_t)) padding = sizeof (uint32_t) - (len % sizeof (uint32_t)); else padding = 0; if (size < padding) { bfd_set_error (bfd_error_bad_value); return false; } num_files++; data += padding; size -= padding; } /* Add the files to mod_source, so that they'll appear in the source info substream. */ strptr = NULL; if (num_files > 0) { uint16_t new_count = num_files + mod_source->files_count; mod_source->files = xrealloc (mod_source->files, sizeof (struct string *) * new_count); strptr = mod_source->files + mod_source->files_count; mod_source->files_count += num_files; } /* Actually copy the data. */ data = orig_data; size = orig_size; while (size > 0) { struct file_checksum *fc = (struct file_checksum *) data; uint32_t string_off; uint8_t padding; size_t len; struct string *str = NULL; string_off = bfd_getl32 (&fc->file_id); len = sizeof (struct file_checksum) + fc->checksum_length; if (len % sizeof (uint32_t)) padding = sizeof (uint32_t) - (len % sizeof (uint32_t)); else padding = 0; /* Remap the "file ID", i.e. the offset in the module's string table, so it points to the right place in the main string table. */ if (string_table) { char *fn = string_table + string_off; size_t fn_len = strlen (fn); uint32_t hash = calc_hash (fn, fn_len); void **slot; slot = htab_find_slot_with_hash (strings->hashmap, fn, hash, NO_INSERT); if (slot) str = (struct string *) *slot; } *strptr = str; strptr++; bfd_putl32 (str ? str->offset : 0, &fc->file_id); memcpy (out, data, len + padding); data += len + padding; size -= len + padding; out += len + padding; } return true; } /* Add a string to the strings table, if it's not already there. Returns its offset within the string table. */ static uint32_t add_string (char *str, size_t len, struct string_table *strings) { uint32_t hash = calc_hash (str, len); struct string *s; void **slot; slot = htab_find_slot_with_hash (strings->hashmap, str, hash, INSERT); if (!*slot) { *slot = xmalloc (offsetof (struct string, s) + len); s = (struct string *) *slot; s->next = NULL; s->hash = hash; s->offset = strings->strings_len; s->source_file_offset = 0xffffffff; s->len = len; memcpy (s->s, str, len); if (strings->strings_tail) strings->strings_tail->next = s; else strings->strings_head = s; strings->strings_tail = s; strings->strings_len += len + 1; } else { s = (struct string *) *slot; } return s->offset; } /* Return the hash of an entry in the string table. */ static hashval_t hash_string_table_entry (const void *p) { const struct string *s = (const struct string *) p; return s->hash; } /* Compare an entry in the string table with a string. */ static int eq_string_table_entry (const void *a, const void *b) { const struct string *s1 = (const struct string *) a; const char *s2 = (const char *) b; size_t s2_len = strlen (s2); if (s2_len != s1->len) return 0; return memcmp (s1->s, s2, s2_len) == 0; } /* Parse the string table within the .debug$S section. */ static void parse_string_table (bfd_byte *data, size_t size, struct string_table *strings) { while (true) { size_t len = strnlen ((char *) data, size); add_string ((char *) data, len, strings); data += len + 1; if (size <= len + 1) break; size -= len + 1; } } /* Remap a type reference within a CodeView symbol. */ static bool remap_symbol_type (void *data, struct type_entry **map, uint32_t num_types) { uint32_t type = bfd_getl32 (data); /* Ignore builtin types (those with IDs below 0x1000). */ if (type < TPI_FIRST_INDEX) return true; if (type >= TPI_FIRST_INDEX + num_types) { einfo (_("%P: CodeView symbol references out of range type %v\n"), type); return false; } type = TPI_FIRST_INDEX + map[type - TPI_FIRST_INDEX]->index; bfd_putl32 (type, data); return true; } /* Add an entry into the globals stream. If it already exists, increase the refcount. */ static bool add_globals_ref (struct globals *glob, bfd *sym_rec_stream, const char *name, size_t name_len, uint8_t *data, size_t len) { void **slot; uint32_t hash; struct global *g; slot = htab_find_slot_with_hash (glob->hashmap, data, iterative_hash (data, len, 0), INSERT); if (*slot) { g = *slot; g->refcount++; return true; } *slot = xmalloc (offsetof (struct global, data) + len); hash = crc32 ((const uint8_t *) name, name_len); hash %= NUM_GLOBALS_HASH_BUCKETS; g = *slot; g->next = NULL; g->offset = bfd_tell (sym_rec_stream); g->hash = hash; g->refcount = 1; memcpy (g->data, data, len + 1); glob->num_entries++; if (glob->last) glob->last->next = g; else glob->first = g; glob->last = g; return bfd_bwrite (data, len, sym_rec_stream) == len; } /* Find the end of the current scope within symbols data. */ static uint8_t * find_end_of_scope (uint8_t *data, uint32_t size) { unsigned int scope_level = 1; uint16_t len; len = bfd_getl16 (data) + sizeof (uint16_t); data += len; size -= len; while (true) { uint16_t type; if (size < sizeof (uint32_t)) return NULL; len = bfd_getl16 (data) + sizeof (uint16_t); type = bfd_getl16 (data + sizeof (uint16_t)); if (size < len) return NULL; switch (type) { case S_GPROC32: case S_LPROC32: case S_BLOCK32: case S_INLINESITE: case S_THUNK32: scope_level++; break; case S_END: case S_PROC_ID_END: case S_INLINESITE_END: scope_level--; if (scope_level == 0) return data; break; } data += len; size -= len; } } /* Return the size of an extended value parameter, as used in LF_ENUMERATE etc. */ static unsigned int extended_value_len (uint16_t type) { switch (type) { case LF_CHAR: return 1; case LF_SHORT: case LF_USHORT: return 2; case LF_LONG: case LF_ULONG: return 4; case LF_QUADWORD: case LF_UQUADWORD: return 8; } return 0; } /* Parse the symbols in a .debug$S section, and copy them to the module's symbol stream. */ static bool parse_symbols (uint8_t *data, uint32_t size, uint8_t **buf, struct type_entry **map, uint32_t num_types, bfd *sym_rec_stream, struct globals *glob, uint16_t mod_num) { uint8_t *orig_buf = *buf; unsigned int scope_level = 0; uint8_t *scope = NULL; while (size >= sizeof (uint16_t)) { uint16_t len, type; len = bfd_getl16 (data) + sizeof (uint16_t); if (len > size) { bfd_set_error (bfd_error_bad_value); return false; } type = bfd_getl16 (data + sizeof (uint16_t)); switch (type) { case S_LDATA32: case S_GDATA32: case S_LTHREAD32: case S_GTHREAD32: { struct datasym *d = (struct datasym *) data; size_t name_len; if (len < offsetof (struct datasym, name)) { einfo (_("%P: warning: truncated CodeView record" " S_LDATA32/S_GDATA32/S_LTHREAD32/S_GTHREAD32\n")); bfd_set_error (bfd_error_bad_value); return false; } if (scope_level == 0) { uint16_t section = bfd_getl16 (&d->section); if (section == 0) /* GC'd, ignore */ break; } name_len = strnlen (d->name, len - offsetof (struct datasym, name)); if (name_len == len - offsetof (struct datasym, name)) { einfo (_("%P: warning: name for S_LDATA32/S_GDATA32/" "S_LTHREAD32/S_GTHREAD32 has no terminating" " zero\n")); bfd_set_error (bfd_error_bad_value); return false; } if (!remap_symbol_type (&d->type, map, num_types)) { bfd_set_error (bfd_error_bad_value); return false; } /* If S_LDATA32 or S_LTHREAD32, copy into module symbols. */ if (type == S_LDATA32 || type == S_LTHREAD32) { memcpy (*buf, d, len); *buf += len; } /* S_LDATA32 and S_LTHREAD32 only go in globals if not in function scope. */ if (type == S_GDATA32 || type == S_GTHREAD32 || scope_level == 0) { if (!add_globals_ref (glob, sym_rec_stream, d->name, name_len, data, len)) return false; } break; } case S_GPROC32: case S_LPROC32: case S_GPROC32_ID: case S_LPROC32_ID: { struct procsym *proc = (struct procsym *) data; size_t name_len; uint16_t section; uint32_t end; uint8_t *endptr; size_t ref_size, padding; struct refsym *ref; if (len < offsetof (struct procsym, name)) { einfo (_("%P: warning: truncated CodeView record" " S_GPROC32/S_LPROC32\n")); bfd_set_error (bfd_error_bad_value); return false; } section = bfd_getl16 (&proc->section); endptr = find_end_of_scope (data, size); if (!endptr) { einfo (_("%P: warning: could not find end of" " S_GPROC32/S_LPROC32 record\n")); bfd_set_error (bfd_error_bad_value); return false; } if (section == 0) /* skip if GC'd */ { /* Skip to after S_END. */ size -= endptr - data; data = endptr; len = bfd_getl16 (data) + sizeof (uint16_t); data += len; size -= len; continue; } name_len = strnlen (proc->name, len - offsetof (struct procsym, name)); if (name_len == len - offsetof (struct procsym, name)) { einfo (_("%P: warning: name for S_GPROC32/S_LPROC32 has no" " terminating zero\n")); bfd_set_error (bfd_error_bad_value); return false; } if (type == S_GPROC32_ID || type == S_LPROC32_ID) { /* Transform into S_GPROC32 / S_LPROC32. */ uint32_t t_idx = bfd_getl32 (&proc->type); struct type_entry *t; uint16_t t_type; if (t_idx < TPI_FIRST_INDEX || t_idx >= TPI_FIRST_INDEX + num_types) { einfo (_("%P: CodeView symbol references out of range" " type %v\n"), type); bfd_set_error (bfd_error_bad_value); return false; } t = map[t_idx - TPI_FIRST_INDEX]; t_type = bfd_getl16 (t->data + sizeof (uint16_t)); switch (t_type) { case LF_FUNC_ID: { struct lf_func_id *t_data = (struct lf_func_id *) t->data; /* Replace proc->type with function type. */ memcpy (&proc->type, &t_data->function_type, sizeof (uint32_t)); break; } case LF_MFUNC_ID: { struct lf_mfunc_id *t_data = (struct lf_mfunc_id *) t->data; /* Replace proc->type with function type. */ memcpy (&proc->type, &t_data->function_type, sizeof (uint32_t)); break; } default: einfo (_("%P: CodeView S_GPROC32_ID/S_LPROC32_ID symbol" " referenced unknown type as ID\n")); bfd_set_error (bfd_error_bad_value); return false; } /* Change record type. */ if (type == S_GPROC32_ID) bfd_putl32 (S_GPROC32, &proc->kind); else bfd_putl32 (S_LPROC32, &proc->kind); } else { if (!remap_symbol_type (&proc->type, map, num_types)) { bfd_set_error (bfd_error_bad_value); return false; } } end = *buf - orig_buf + sizeof (uint32_t) + endptr - data; bfd_putl32 (end, &proc->end); /* Add S_PROCREF / S_LPROCREF to globals stream. */ ref_size = offsetof (struct refsym, name) + name_len + 1; if (ref_size % sizeof (uint32_t)) padding = sizeof (uint32_t) - (ref_size % sizeof (uint32_t)); else padding = 0; ref = xmalloc (ref_size + padding); bfd_putl16 (ref_size + padding - sizeof (uint16_t), &ref->size); bfd_putl16 (type == S_GPROC32 || type == S_GPROC32_ID ? S_PROCREF : S_LPROCREF, &ref->kind); bfd_putl32 (0, &ref->sum_name); bfd_putl32 (*buf - orig_buf + sizeof (uint32_t), &ref->symbol_offset); bfd_putl16 (mod_num + 1, &ref->mod); memcpy (ref->name, proc->name, name_len + 1); memset (ref->name + name_len + 1, 0, padding); if (!add_globals_ref (glob, sym_rec_stream, proc->name, name_len, (uint8_t *) ref, ref_size + padding)) { free (ref); return false; } free (ref); scope = *buf; memcpy (*buf, proc, len); *buf += len; scope_level++; break; } case S_UDT: { struct udtsym *udt = (struct udtsym *) data; size_t name_len; if (len < offsetof (struct udtsym, name)) { einfo (_("%P: warning: truncated CodeView record" " S_UDT\n")); bfd_set_error (bfd_error_bad_value); return false; } name_len = strnlen (udt->name, len - offsetof (struct udtsym, name)); if (name_len == len - offsetof (struct udtsym, name)) { einfo (_("%P: warning: name for S_UDT has no" " terminating zero\n")); bfd_set_error (bfd_error_bad_value); return false; } if (!remap_symbol_type (&udt->type, map, num_types)) { bfd_set_error (bfd_error_bad_value); return false; } /* S_UDT goes in the symbols stream if within a procedure, otherwise it goes in the globals stream. */ if (scope_level == 0) { if (!add_globals_ref (glob, sym_rec_stream, udt->name, name_len, data, len)) return false; } else { memcpy (*buf, udt, len); *buf += len; } break; } case S_CONSTANT: { struct constsym *c = (struct constsym *) data; size_t name_len, rec_size; uint16_t val; if (len < offsetof (struct constsym, name)) { einfo (_("%P: warning: truncated CodeView record" " S_CONSTANT\n")); bfd_set_error (bfd_error_bad_value); return false; } rec_size = offsetof (struct constsym, name); val = bfd_getl16 (&c->value); /* If val >= 0x8000, actual value follows. */ if (val >= 0x8000) { unsigned int param_len = extended_value_len (val); if (param_len == 0) { einfo (_("%P: warning: unhandled type %v within" " S_CONSTANT\n"), val); bfd_set_error (bfd_error_bad_value); return false; } rec_size += param_len; } name_len = strnlen ((const char *) data + rec_size, len - rec_size); if (name_len == len - rec_size) { einfo (_("%P: warning: name for S_CONSTANT has no" " terminating zero\n")); bfd_set_error (bfd_error_bad_value); return false; } if (!remap_symbol_type (&c->type, map, num_types)) { bfd_set_error (bfd_error_bad_value); return false; } if (!add_globals_ref (glob, sym_rec_stream, (const char *) data + rec_size, name_len, data, len)) return false; break; } case S_END: case S_INLINESITE_END: case S_PROC_ID_END: memcpy (*buf, data, len); if (type == S_PROC_ID_END) /* transform to S_END */ bfd_putl16 (S_END, *buf + sizeof (uint16_t)); /* Reset scope variable back to the address of the previous scope start. */ if (scope) { uint32_t parent; uint16_t scope_start_type = bfd_getl16 (scope + sizeof (uint16_t)); switch (scope_start_type) { case S_GPROC32: case S_LPROC32: parent = bfd_getl32 (scope + offsetof (struct procsym, parent)); break; case S_BLOCK32: parent = bfd_getl32 (scope + offsetof (struct blocksym, parent)); break; case S_INLINESITE: parent = bfd_getl32 (scope + offsetof (struct inline_site, parent)); break; case S_THUNK32: parent = bfd_getl32 (scope + offsetof (struct thunk, parent)); break; default: einfo (_("%P: warning: unexpected CodeView scope start" " record %v\n"), scope_start_type); bfd_set_error (bfd_error_bad_value); return false; } if (parent == 0) scope = NULL; else scope = orig_buf + parent - sizeof (uint32_t); } *buf += len; scope_level--; break; case S_BUILDINFO: { struct buildinfosym *bi = (struct buildinfosym *) data; if (len < sizeof (struct buildinfosym)) { einfo (_("%P: warning: truncated CodeView record" " S_BUILDINFO\n")); bfd_set_error (bfd_error_bad_value); return false; } if (!remap_symbol_type (&bi->type, map, num_types)) { bfd_set_error (bfd_error_bad_value); return false; } memcpy (*buf, data, len); *buf += len; break; } case S_BLOCK32: { struct blocksym *bl = (struct blocksym *) data; uint8_t *endptr; uint32_t end; if (len < offsetof (struct blocksym, name)) { einfo (_("%P: warning: truncated CodeView record" " S_BLOCK32\n")); bfd_set_error (bfd_error_bad_value); return false; } bfd_putl32 (scope - orig_buf + sizeof (uint32_t), &bl->parent); endptr = find_end_of_scope (data, size); if (!endptr) { einfo (_("%P: warning: could not find end of" " S_BLOCK32 record\n")); bfd_set_error (bfd_error_bad_value); return false; } end = *buf - orig_buf + sizeof (uint32_t) + endptr - data; bfd_putl32 (end, &bl->end); scope = *buf; memcpy (*buf, data, len); *buf += len; scope_level++; break; } case S_BPREL32: { struct bprelsym *bp = (struct bprelsym *) data; if (len < offsetof (struct bprelsym, name)) { einfo (_("%P: warning: truncated CodeView record" " S_BPREL32\n")); bfd_set_error (bfd_error_bad_value); return false; } if (!remap_symbol_type (&bp->type, map, num_types)) { bfd_set_error (bfd_error_bad_value); return false; } memcpy (*buf, data, len); *buf += len; break; } case S_REGISTER: { struct regsym *reg = (struct regsym *) data; if (len < offsetof (struct regsym, name)) { einfo (_("%P: warning: truncated CodeView record" " S_REGISTER\n")); bfd_set_error (bfd_error_bad_value); return false; } if (!remap_symbol_type (®->type, map, num_types)) { bfd_set_error (bfd_error_bad_value); return false; } memcpy (*buf, data, len); *buf += len; break; } case S_REGREL32: { struct regrel *rr = (struct regrel *) data; if (len < offsetof (struct regrel, name)) { einfo (_("%P: warning: truncated CodeView record" " S_REGREL32\n")); bfd_set_error (bfd_error_bad_value); return false; } if (!remap_symbol_type (&rr->type, map, num_types)) { bfd_set_error (bfd_error_bad_value); return false; } memcpy (*buf, data, len); *buf += len; break; } case S_LOCAL: { struct localsym *l = (struct localsym *) data; if (len < offsetof (struct localsym, name)) { einfo (_("%P: warning: truncated CodeView record" " S_LOCAL\n")); bfd_set_error (bfd_error_bad_value); return false; } if (!remap_symbol_type (&l->type, map, num_types)) { bfd_set_error (bfd_error_bad_value); return false; } memcpy (*buf, data, len); *buf += len; break; } case S_INLINESITE: { struct inline_site *is = (struct inline_site *) data; uint8_t *endptr; uint32_t end; if (len < offsetof (struct inline_site, binary_annotations)) { einfo (_("%P: warning: truncated CodeView record" " S_INLINESITE\n")); bfd_set_error (bfd_error_bad_value); return false; } bfd_putl32 (scope - orig_buf + sizeof (uint32_t), &is->parent); endptr = find_end_of_scope (data, size); if (!endptr) { einfo (_("%P: warning: could not find end of" " S_INLINESITE record\n")); bfd_set_error (bfd_error_bad_value); return false; } end = *buf - orig_buf + sizeof (uint32_t) + endptr - data; bfd_putl32 (end, &is->end); if (!remap_symbol_type (&is->inlinee, map, num_types)) { bfd_set_error (bfd_error_bad_value); return false; } scope = *buf; memcpy (*buf, data, len); *buf += len; scope_level++; break; } case S_THUNK32: { struct thunk *th = (struct thunk *) data; uint8_t *endptr; uint32_t end; if (len < offsetof (struct thunk, name)) { einfo (_("%P: warning: truncated CodeView record" " S_THUNK32\n")); bfd_set_error (bfd_error_bad_value); return false; } bfd_putl32 (scope - orig_buf + sizeof (uint32_t), &th->parent); endptr = find_end_of_scope (data, size); if (!endptr) { einfo (_("%P: warning: could not find end of" " S_THUNK32 record\n")); bfd_set_error (bfd_error_bad_value); return false; } end = *buf - orig_buf + sizeof (uint32_t) + endptr - data; bfd_putl32 (end, &th->end); scope = *buf; memcpy (*buf, data, len); *buf += len; scope_level++; break; } case S_HEAPALLOCSITE: { struct heap_alloc_site *has = (struct heap_alloc_site *) data; if (len < sizeof (struct heap_alloc_site)) { einfo (_("%P: warning: truncated CodeView record" " S_HEAPALLOCSITE\n")); bfd_set_error (bfd_error_bad_value); return false; } if (!remap_symbol_type (&has->type, map, num_types)) { bfd_set_error (bfd_error_bad_value); return false; } memcpy (*buf, data, len); *buf += len; break; } case S_OBJNAME: /* just copy */ case S_COMPILE3: case S_UNAMESPACE: case S_FRAMEPROC: case S_FRAMECOOKIE: case S_LABEL32: case S_DEFRANGE_REGISTER_REL: case S_DEFRANGE_FRAMEPOINTER_REL: case S_DEFRANGE_SUBFIELD_REGISTER: case S_DEFRANGE_FRAMEPOINTER_REL_FULL_SCOPE: case S_DEFRANGE_REGISTER: memcpy (*buf, data, len); *buf += len; break; default: einfo (_("%P: warning: unrecognized CodeView record %v\n"), type); bfd_set_error (bfd_error_bad_value); return false; } data += len; size -= len; } return true; } /* For a given symbol subsection, work out how much space to allocate in the result module stream. This is different because we don't copy certain symbols, such as S_CONSTANT, and we skip over any procedures or data that have been GC'd out. */ static bool calculate_symbols_size (uint8_t *data, uint32_t size, uint32_t *sym_size) { unsigned int scope_level = 0; while (size >= sizeof (uint32_t)) { uint16_t len = bfd_getl16 (data) + sizeof (uint16_t); uint16_t type = bfd_getl16 (data + sizeof (uint16_t)); switch (type) { case S_LDATA32: case S_LTHREAD32: { struct datasym *d = (struct datasym *) data; uint16_t section; if (len < offsetof (struct datasym, name)) { einfo (_("%P: warning: truncated CodeView record" " S_LDATA32/S_LTHREAD32\n")); return false; } section = bfd_getl16 (&d->section); /* copy if not GC'd or within function */ if (scope_level != 0 || section != 0) *sym_size += len; } case S_GDATA32: case S_GTHREAD32: case S_CONSTANT: /* Not copied into symbols stream. */ break; case S_GPROC32: case S_LPROC32: case S_GPROC32_ID: case S_LPROC32_ID: { struct procsym *proc = (struct procsym *) data; uint16_t section; if (len < offsetof (struct procsym, name)) { einfo (_("%P: warning: truncated CodeView record" " S_GPROC32/S_LPROC32\n")); return false; } section = bfd_getl16 (&proc->section); if (section != 0) { *sym_size += len; } else { uint8_t *endptr = find_end_of_scope (data, size); if (!endptr) { einfo (_("%P: warning: could not find end of" " S_GPROC32/S_LPROC32 record\n")); return false; } /* Skip to after S_END. */ size -= endptr - data; data = endptr; len = bfd_getl16 (data) + sizeof (uint16_t); data += len; size -= len; continue; } scope_level++; break; } case S_UDT: if (scope_level != 0) /* only goes in symbols if local */ *sym_size += len; break; case S_BLOCK32: /* always copied */ case S_INLINESITE: case S_THUNK32: *sym_size += len; scope_level++; break; case S_END: /* always copied */ case S_PROC_ID_END: case S_INLINESITE_END: *sym_size += len; scope_level--; break; case S_OBJNAME: /* always copied */ case S_COMPILE3: case S_UNAMESPACE: case S_FRAMEPROC: case S_FRAMECOOKIE: case S_LABEL32: case S_BUILDINFO: case S_BPREL32: case S_REGISTER: case S_REGREL32: case S_LOCAL: case S_DEFRANGE_REGISTER_REL: case S_DEFRANGE_FRAMEPOINTER_REL: case S_DEFRANGE_SUBFIELD_REGISTER: case S_DEFRANGE_FRAMEPOINTER_REL_FULL_SCOPE: case S_DEFRANGE_REGISTER: case S_HEAPALLOCSITE: *sym_size += len; break; default: einfo (_("%P: warning: unrecognized CodeView record %v\n"), type); return false; } data += len; size -= len; } return true; } /* Parse the .debug$S section within an object file. */ static bool handle_debugs_section (asection *s, bfd *mod, struct string_table *strings, uint8_t **dataptr, uint32_t *sizeptr, struct mod_source_files *mod_source, bfd *abfd, uint8_t **syms, uint32_t *sym_byte_size, struct type_entry **map, uint32_t num_types, bfd *sym_rec_stream, struct globals *glob, uint16_t mod_num) { bfd_byte *data = NULL; size_t off; uint32_t c13_size = 0; char *string_table = NULL; uint8_t *buf, *bufptr, *symbuf, *symbufptr; uint32_t sym_size = 0; if (!bfd_get_full_section_contents (mod, s, &data)) return false; if (!data) return false; /* Resolve relocations. Addresses are stored within the .debug$S section as a .secidx, .secrel32 pair. */ if (s->flags & SEC_RELOC) { struct internal_reloc *relocs; struct internal_syment *symbols; asection **sectlist; unsigned int syment_count; int sect_num; struct external_syment *ext; syment_count = obj_raw_syment_count (mod); relocs = _bfd_coff_read_internal_relocs (mod, s, false, NULL, true, NULL); symbols = xmalloc (sizeof (struct internal_syment) * syment_count); sectlist = xmalloc (sizeof (asection *) * syment_count); ext = (struct external_syment *) (coff_data (mod)->external_syms); for (unsigned int i = 0; i < syment_count; i++) { bfd_coff_swap_sym_in (mod, &ext[i], &symbols[i]); } sect_num = 1; for (asection *sect = mod->sections; sect; sect = sect->next) { for (unsigned int i = 0; i < syment_count; i++) { if (symbols[i].n_scnum == sect_num) sectlist[i] = sect; } sect_num++; } if (!bfd_coff_relocate_section (abfd, coff_data (abfd)->link_info, mod, s, data, relocs, symbols, sectlist)) { free (sectlist); free (symbols); free (data); return false; } free (sectlist); free (symbols); } if (bfd_getl32 (data) != CV_SIGNATURE_C13) { free (data); return true; } off = sizeof (uint32_t); /* calculate size */ while (off + sizeof (uint32_t) <= s->size) { uint32_t type, size; type = bfd_getl32 (data + off); off += sizeof (uint32_t); if (off + sizeof (uint32_t) > s->size) { free (data); bfd_set_error (bfd_error_bad_value); return false; } size = bfd_getl32 (data + off); off += sizeof (uint32_t); if (off + size > s->size) { free (data); bfd_set_error (bfd_error_bad_value); return false; } switch (type) { case DEBUG_S_FILECHKSMS: c13_size += sizeof (uint32_t) + sizeof (uint32_t) + size; if (c13_size % sizeof (uint32_t)) c13_size += sizeof (uint32_t) - (c13_size % sizeof (uint32_t)); break; case DEBUG_S_STRINGTABLE: parse_string_table (data + off, size, strings); string_table = (char *) data + off; break; case DEBUG_S_LINES: { uint16_t sect; if (size < sizeof (uint32_t) + sizeof (uint16_t)) { free (data); bfd_set_error (bfd_error_bad_value); return false; } sect = bfd_getl16 (data + off + sizeof (uint32_t)); /* Skip GC'd symbols. */ if (sect != 0) { c13_size += sizeof (uint32_t) + sizeof (uint32_t) + size; if (c13_size % sizeof (uint32_t)) c13_size += sizeof (uint32_t) - (c13_size % sizeof (uint32_t)); } break; } case DEBUG_S_SYMBOLS: if (!calculate_symbols_size (data + off, size, &sym_size)) { free (data); bfd_set_error (bfd_error_bad_value); return false; } break; } off += size; if (off % sizeof (uint32_t)) off += sizeof (uint32_t) - (off % sizeof (uint32_t)); } if (sym_size % sizeof (uint32_t)) sym_size += sizeof (uint32_t) - (sym_size % sizeof (uint32_t)); if (c13_size == 0 && sym_size == 0) { free (data); return true; } /* copy data */ buf = NULL; if (c13_size != 0) buf = xmalloc (c13_size); bufptr = buf; symbuf = NULL; if (sym_size != 0) symbuf = xmalloc (sym_size); symbufptr = symbuf; off = sizeof (uint32_t); while (off + sizeof (uint32_t) <= s->size) { uint32_t type, size; type = bfd_getl32 (data + off); off += sizeof (uint32_t); size = bfd_getl32 (data + off); off += sizeof (uint32_t); switch (type) { case DEBUG_S_FILECHKSMS: if (!copy_filechksms (data + off, size, string_table, strings, bufptr, mod_source)) { free (data); free (symbuf); return false; } bufptr += sizeof (uint32_t) + sizeof (uint32_t) + size; break; case DEBUG_S_LINES: { uint16_t sect; sect = bfd_getl16 (data + off + sizeof (uint32_t)); /* Skip if GC'd. */ if (sect != 0) { bfd_putl32 (type, bufptr); bufptr += sizeof (uint32_t); bfd_putl32 (size, bufptr); bufptr += sizeof (uint32_t); memcpy (bufptr, data + off, size); bufptr += size; } break; } case DEBUG_S_SYMBOLS: if (!parse_symbols (data + off, size, &symbufptr, map, num_types, sym_rec_stream, glob, mod_num)) { free (data); free (symbuf); return false; } break; } off += size; if (off % sizeof (uint32_t)) off += sizeof (uint32_t) - (off % sizeof (uint32_t)); } free (data); if (buf) { if (*dataptr) { /* Append the C13 info to what's already there, if the module has multiple .debug$S sections. */ *dataptr = xrealloc (*dataptr, *sizeptr + c13_size); memcpy (*dataptr + *sizeptr, buf, c13_size); free (buf); } else { *dataptr = buf; } *sizeptr += c13_size; } if (symbuf) { if (*syms) { *syms = xrealloc (*syms, *sym_byte_size + sym_size); memcpy (*syms + *sym_byte_size, symbuf, sym_size); free (symbuf); } else { *syms = symbuf; } *sym_byte_size += sym_size; } return true; } /* Remap the type number stored in data from the per-module numbering to that of the deduplicated output list. */ static bool remap_type (void *data, struct type_entry **map, uint32_t type_num, uint32_t num_types) { uint32_t type = bfd_getl32 (data); /* Ignore builtin types (those with IDs below 0x1000). */ if (type < TPI_FIRST_INDEX) return true; if (type >= TPI_FIRST_INDEX + type_num) { einfo (_("%P: CodeView type %v references other type %v not yet " "declared\n"), TPI_FIRST_INDEX + type_num, type); return false; } if (type >= TPI_FIRST_INDEX + num_types) { einfo (_("%P: CodeView type %v references out of range type %v\n"), TPI_FIRST_INDEX + type_num, type); return false; } type = TPI_FIRST_INDEX + map[type - TPI_FIRST_INDEX]->index; bfd_putl32 (type, data); return true; } /* Determines whether the name of a struct, class, or union counts as "anonymous". Non-anonymous types have a hash based on just the name, rather than the whole structure. */ static bool is_name_anonymous (char *name, size_t len) { static const char tag1[] = ""; static const char tag2[] = "__unnamed"; static const char tag3[] = "::"; static const char tag4[] = "::__unnamed"; if (len == sizeof (tag1) - 1 && !memcmp (name, tag1, sizeof (tag1) - 1)) return true; if (len == sizeof (tag2) - 1 && !memcmp (name, tag2, sizeof (tag2) - 1)) return true; if (len >= sizeof (tag3) - 1 && !memcmp (name + len - sizeof (tag3) + 1, tag3, sizeof (tag3) - 1)) return true; if (len >= sizeof (tag4) - 1 && !memcmp (name + len - sizeof (tag4) + 1, tag4, sizeof (tag4) - 1)) return true; return false; } /* Handle LF_UDT_SRC_LINE type entries, which are a special case. These give the source file and line number for each user-defined type that is declared. We parse these and emit instead an LF_UDT_MOD_SRC_LINE entry, which also includes the module number. */ static bool handle_udt_src_line (uint8_t *data, uint16_t size, struct type_entry **map, uint32_t type_num, uint32_t num_types, struct types *ids, uint16_t mod_num, struct string_table *strings) { struct lf_udt_src_line *usl = (struct lf_udt_src_line *) data; uint32_t orig_type, source_file_type; void **slot; hashval_t hash; struct type_entry *e, *type_e, *str_e; struct lf_udt_mod_src_line *umsl; struct lf_string_id *str; uint32_t source_file_offset; if (size < sizeof (struct lf_udt_src_line)) { einfo (_("%P: warning: truncated CodeView type record" " LF_UDT_SRC_LINE\n")); return false; } /* Check if LF_UDT_MOD_SRC_LINE already present for type, and return. */ orig_type = bfd_getl32 (&usl->type); if (orig_type < TPI_FIRST_INDEX || orig_type >= TPI_FIRST_INDEX + num_types || !map[orig_type - TPI_FIRST_INDEX]) { einfo (_("%P: warning: CodeView type record LF_UDT_SRC_LINE" " referred to unknown type %v\n"), orig_type); return false; } type_e = map[orig_type - TPI_FIRST_INDEX]; /* Skip if type already declared in other module. */ if (type_e->has_udt_src_line) return true; if (!remap_type (&usl->type, map, type_num, num_types)) return false; /* Extract string from source_file_type. */ source_file_type = bfd_getl32 (&usl->source_file_type); if (source_file_type < TPI_FIRST_INDEX || source_file_type >= TPI_FIRST_INDEX + num_types || !map[source_file_type - TPI_FIRST_INDEX]) { einfo (_("%P: warning: CodeView type record LF_UDT_SRC_LINE" " referred to unknown string %v\n"), source_file_type); return false; } str_e = map[source_file_type - TPI_FIRST_INDEX]; if (bfd_getl16 (str_e->data + sizeof (uint16_t)) != LF_STRING_ID) { einfo (_("%P: warning: CodeView type record LF_UDT_SRC_LINE" " pointed to unexpected record type\n")); return false; } str = (struct lf_string_id *) str_e->data; /* Add string to string table. */ source_file_offset = add_string (str->string, strlen (str->string), strings); /* Add LF_UDT_MOD_SRC_LINE entry. */ size = sizeof (struct lf_udt_mod_src_line); e = xmalloc (offsetof (struct type_entry, data) + size); e->next = NULL; e->index = ids->num_types; e->has_udt_src_line = false; /* LF_UDT_MOD_SRC_LINE use calc_hash on the type number, rather than the crc32 used for type hashes elsewhere. */ e->cv_hash = calc_hash ((char *) &usl->type, sizeof (uint32_t)); type_e->has_udt_src_line = true; umsl = (struct lf_udt_mod_src_line *) e->data; bfd_putl16 (size - sizeof (uint16_t), &umsl->size); bfd_putl16 (LF_UDT_MOD_SRC_LINE, &umsl->kind); memcpy (&umsl->type, &usl->type, sizeof (uint32_t)); bfd_putl32 (source_file_offset, &umsl->source_file_string); memcpy (&umsl->line_no, &usl->line_no, sizeof (uint32_t)); bfd_putl16 (mod_num + 1, &umsl->module_no); hash = iterative_hash (e->data, size, 0); slot = htab_find_slot_with_hash (ids->hashmap, data, hash, INSERT); if (!slot) { free (e); return false; } if (*slot) { free (e); einfo (_("%P: warning: duplicate CodeView type record " "LF_UDT_MOD_SRC_LINE\n")); return false; } *slot = e; if (ids->last) ids->last->next = e; else ids->first = e; ids->last = e; map[type_num] = e; ids->num_types++; return true; } /* Parse a type definition in the .debug$T section. We remap the numbers of any referenced types, and if the type is not a duplicate of one already seen add it to types (for TPI types) or ids (for IPI types). */ static bool handle_type (uint8_t *data, struct type_entry **map, uint32_t type_num, uint32_t num_types, struct types *types, struct types *ids, uint16_t mod_num, struct string_table *strings) { uint16_t size, type; void **slot; hashval_t hash; bool other_hash = false; uint32_t cv_hash; struct types *t; bool ipi = false; size = bfd_getl16 (data) + sizeof (uint16_t); type = bfd_getl16 (data + sizeof (uint16_t)); switch (type) { case LF_MODIFIER: { struct lf_modifier *mod = (struct lf_modifier *) data; if (size < offsetof (struct lf_modifier, modifier)) { einfo (_("%P: warning: truncated CodeView type record " "LF_MODIFIER\n")); return false; } if (!remap_type (&mod->base_type, map, type_num, num_types)) return false; break; } case LF_POINTER: { struct lf_pointer *ptr = (struct lf_pointer *) data; if (size < offsetof (struct lf_pointer, attributes)) { einfo (_("%P: warning: truncated CodeView type record" " LF_POINTER\n")); return false; } if (!remap_type (&ptr->base_type, map, type_num, num_types)) return false; break; } case LF_PROCEDURE: { struct lf_procedure *proc = (struct lf_procedure *) data; if (size < sizeof (struct lf_procedure)) { einfo (_("%P: warning: truncated CodeView type record" " LF_PROCEDURE\n")); return false; } if (!remap_type (&proc->return_type, map, type_num, num_types)) return false; if (!remap_type (&proc->arglist, map, type_num, num_types)) return false; break; } case LF_MFUNCTION: { struct lf_mfunction *func = (struct lf_mfunction *) data; if (size < sizeof (struct lf_procedure)) { einfo (_("%P: warning: truncated CodeView type record" " LF_MFUNCTION\n")); return false; } if (!remap_type (&func->return_type, map, type_num, num_types)) return false; if (!remap_type (&func->containing_class_type, map, type_num, num_types)) return false; if (!remap_type (&func->this_type, map, type_num, num_types)) return false; if (!remap_type (&func->arglist, map, type_num, num_types)) return false; break; } case LF_ARGLIST: { uint32_t num_entries; struct lf_arglist *al = (struct lf_arglist *) data; if (size < offsetof (struct lf_arglist, args)) { einfo (_("%P: warning: truncated CodeView type record" " LF_ARGLIST\n")); return false; } num_entries = bfd_getl32 (&al->num_entries); if (size < offsetof (struct lf_arglist, args) + (num_entries * sizeof (uint32_t))) { einfo (_("%P: warning: truncated CodeView type record" " LF_ARGLIST\n")); return false; } for (uint32_t i = 0; i < num_entries; i++) { if (!remap_type (&al->args[i], map, type_num, num_types)) return false; } break; } case LF_FIELDLIST: { uint16_t left = size - sizeof (uint16_t) - sizeof (uint16_t); uint8_t *ptr = data + sizeof (uint16_t) + sizeof (uint16_t); while (left > 0) { uint16_t subtype; if (left < sizeof (uint16_t)) { einfo (_("%P: warning: truncated CodeView type record" " LF_FIELDLIST\n")); return false; } subtype = bfd_getl16 (ptr); switch (subtype) { case LF_MEMBER: { struct lf_member *mem = (struct lf_member *) ptr; uint16_t offset; size_t name_len, subtype_len; if (left < offsetof (struct lf_member, name)) { einfo (_("%P: warning: truncated CodeView type record" " LF_MEMBER\n")); return false; } if (!remap_type (&mem->type, map, type_num, num_types)) return false; subtype_len = offsetof (struct lf_member, name); offset = bfd_getl16 (&mem->offset); /* If offset >= 0x8000, actual value follows. */ if (offset >= 0x8000) { unsigned int param_len = extended_value_len (offset); if (param_len == 0) { einfo (_("%P: warning: unhandled type %v within" " LF_MEMBER\n"), offset); return false; } subtype_len += param_len; if (left < subtype_len) { einfo (_("%P: warning: truncated CodeView type record" " LF_MEMBER\n")); return false; } } name_len = strnlen ((char *) mem + subtype_len, left - subtype_len); if (name_len == left - offsetof (struct lf_member, name)) { einfo (_("%P: warning: name for LF_MEMBER has no" " terminating zero\n")); return false; } name_len++; subtype_len += name_len; if (subtype_len % 4 != 0) subtype_len += 4 - (subtype_len % 4); if (left < subtype_len) { einfo (_("%P: warning: truncated CodeView type record" " LF_FIELDLIST\n")); return false; } ptr += subtype_len; left -= subtype_len; break; } case LF_ENUMERATE: { struct lf_enumerate *en = (struct lf_enumerate *) ptr; size_t name_len, subtype_len; uint16_t val; if (left < offsetof (struct lf_enumerate, name)) { einfo (_("%P: warning: truncated CodeView type record" " LF_ENUMERATE\n")); return false; } subtype_len = offsetof (struct lf_enumerate, name); val = bfd_getl16 (&en->value); /* If val >= 0x8000, the actual value immediately follows. */ if (val >= 0x8000) { unsigned int param_len = extended_value_len (val); if (param_len == 0) { einfo (_("%P: warning: unhandled type %v within" " LF_ENUMERATE\n"), val); return false; } if (left < subtype_len + param_len) { einfo (_("%P: warning: truncated CodeView type" " record LF_ENUMERATE\n")); return false; } subtype_len += param_len; } name_len = strnlen ((char *) ptr + subtype_len, left - subtype_len); if (name_len == left - offsetof (struct lf_enumerate, name)) { einfo (_("%P: warning: name for LF_ENUMERATE has no" " terminating zero\n")); return false; } name_len++; subtype_len += name_len; if (subtype_len % 4 != 0) subtype_len += 4 - (subtype_len % 4); if (left < subtype_len) { einfo (_("%P: warning: truncated CodeView type record" " LF_ENUMERATE\n")); return false; } ptr += subtype_len; left -= subtype_len; break; } case LF_INDEX: { struct lf_index *ind = (struct lf_index *) ptr; if (left < sizeof (struct lf_index)) { einfo (_("%P: warning: truncated CodeView type record" " LF_INDEX\n")); return false; } if (!remap_type (&ind->index, map, type_num, num_types)) return false; ptr += sizeof (struct lf_index); left -= sizeof (struct lf_index); break; } case LF_ONEMETHOD: { struct lf_onemethod *meth = (struct lf_onemethod *) ptr; size_t name_len, subtype_len; if (left < offsetof (struct lf_onemethod, name)) { einfo (_("%P: warning: truncated CodeView type record" " LF_ONEMETHOD\n")); return false; } if (!remap_type (&meth->method_type, map, type_num, num_types)) return false; name_len = strnlen (meth->name, left - offsetof (struct lf_onemethod, name)); if (name_len == left - offsetof (struct lf_onemethod, name)) { einfo (_("%P: warning: name for LF_ONEMETHOD has no" " terminating zero\n")); return false; } name_len++; subtype_len = offsetof (struct lf_onemethod, name) + name_len; if (subtype_len % 4 != 0) subtype_len += 4 - (subtype_len % 4); if (left < subtype_len) { einfo (_("%P: warning: truncated CodeView type record" " LF_FIELDLIST\n")); return false; } ptr += subtype_len; left -= subtype_len; break; } case LF_METHOD: { struct lf_method *meth = (struct lf_method *) ptr; size_t name_len, subtype_len; if (left < offsetof (struct lf_method, name)) { einfo (_("%P: warning: truncated CodeView type record" " LF_METHOD\n")); return false; } if (!remap_type (&meth->method_list, map, type_num, num_types)) return false; name_len = strnlen (meth->name, left - offsetof (struct lf_method, name)); if (name_len == left - offsetof (struct lf_method, name)) { einfo (_("%P: warning: name for LF_METHOD has no" " terminating zero\n")); return false; } name_len++; subtype_len = offsetof (struct lf_method, name) + name_len; if (subtype_len % 4 != 0) subtype_len += 4 - (subtype_len % 4); if (left < subtype_len) { einfo (_("%P: warning: truncated CodeView type record" " LF_FIELDLIST\n")); return false; } ptr += subtype_len; left -= subtype_len; break; } case LF_BCLASS: { struct lf_bclass *bc = (struct lf_bclass *) ptr; size_t subtype_len; uint16_t offset; if (left < sizeof (struct lf_bclass)) { einfo (_("%P: warning: truncated CodeView type record" " LF_BCLASS\n")); return false; } if (!remap_type (&bc->base_class_type, map, type_num, num_types)) return false; subtype_len = sizeof (struct lf_bclass); offset = bfd_getl16 (&bc->offset); /* If offset >= 0x8000, actual value follows. */ if (offset >= 0x8000) { unsigned int param_len = extended_value_len (offset); if (param_len == 0) { einfo (_("%P: warning: unhandled type %v within" " LF_BCLASS\n"), offset); return false; } subtype_len += param_len; if (left < subtype_len) { einfo (_("%P: warning: truncated CodeView type record" " LF_BCLASS\n")); return false; } } if (subtype_len % 4 != 0) subtype_len += 4 - (subtype_len % 4); if (left < subtype_len) { einfo (_("%P: warning: truncated CodeView type record" " LF_BCLASS\n")); return false; } ptr += subtype_len; left -= subtype_len; break; } case LF_VFUNCTAB: { struct lf_vfunctab *vft = (struct lf_vfunctab *) ptr; if (left < sizeof (struct lf_vfunctab)) { einfo (_("%P: warning: truncated CodeView type record" " LF_VFUNCTAB\n")); return false; } if (!remap_type (&vft->type, map, type_num, num_types)) return false; ptr += sizeof (struct lf_vfunctab); left -= sizeof (struct lf_vfunctab); break; } case LF_VBCLASS: case LF_IVBCLASS: { struct lf_vbclass *vbc = (struct lf_vbclass *) ptr; size_t subtype_len; uint16_t offset; if (left < sizeof (struct lf_vbclass)) { einfo (_("%P: warning: truncated CodeView type record" " LF_VBCLASS/LF_IVBCLASS\n")); return false; } if (!remap_type (&vbc->base_class_type, map, type_num, num_types)) return false; if (!remap_type (&vbc->virtual_base_pointer_type, map, type_num, num_types)) return false; subtype_len = offsetof (struct lf_vbclass, virtual_base_vbtable_offset); offset = bfd_getl16 (&vbc->virtual_base_pointer_offset); /* If offset >= 0x8000, actual value follows. */ if (offset >= 0x8000) { unsigned int param_len = extended_value_len (offset); if (param_len == 0) { einfo (_("%P: warning: unhandled type %v within" " LF_VBCLASS/LF_IVBCLASS\n"), offset); return false; } subtype_len += param_len; if (left < subtype_len) { einfo (_("%P: warning: truncated CodeView type record" " LF_VBCLASS/LF_IVBCLASS\n")); return false; } } offset = bfd_getl16 ((char *)vbc + subtype_len); subtype_len += sizeof (uint16_t); /* If offset >= 0x8000, actual value follows. */ if (offset >= 0x8000) { unsigned int param_len = extended_value_len (offset); if (param_len == 0) { einfo (_("%P: warning: unhandled type %v within" " LF_VBCLASS/LF_IVBCLASS\n"), offset); return false; } subtype_len += param_len; if (left < subtype_len) { einfo (_("%P: warning: truncated CodeView type record" " LF_VBCLASS/LF_IVBCLASS\n")); return false; } } if (subtype_len % 4 != 0) subtype_len += 4 - (subtype_len % 4); if (left < subtype_len) { einfo (_("%P: warning: truncated CodeView type record" " LF_VBCLASS/LF_IVBCLASS\n")); return false; } ptr += subtype_len; left -= subtype_len; break; } case LF_STMEMBER: { struct lf_static_member *st = (struct lf_static_member *) ptr; size_t name_len, subtype_len; if (left < offsetof (struct lf_static_member, name)) { einfo (_("%P: warning: truncated CodeView type record" " LF_STMEMBER\n")); return false; } if (!remap_type (&st->type, map, type_num, num_types)) return false; name_len = strnlen (st->name, left - offsetof (struct lf_static_member, name)); if (name_len == left - offsetof (struct lf_static_member, name)) { einfo (_("%P: warning: name for LF_STMEMBER has no" " terminating zero\n")); return false; } name_len++; subtype_len = offsetof (struct lf_static_member, name) + name_len; if (subtype_len % 4 != 0) subtype_len += 4 - (subtype_len % 4); if (left < subtype_len) { einfo (_("%P: warning: truncated CodeView type record" " LF_FIELDLIST\n")); return false; } ptr += subtype_len; left -= subtype_len; break; } case LF_NESTTYPE: { struct lf_nest_type *nest = (struct lf_nest_type *) ptr; size_t name_len, subtype_len; if (left < offsetof (struct lf_nest_type, name)) { einfo (_("%P: warning: truncated CodeView type record" " LF_NESTTYPE\n")); return false; } if (!remap_type (&nest->type, map, type_num, num_types)) return false; name_len = strnlen (nest->name, left - offsetof (struct lf_nest_type, name)); if (name_len == left - offsetof (struct lf_nest_type, name)) { einfo (_("%P: warning: name for LF_NESTTYPE has no" " terminating zero\n")); return false; } name_len++; subtype_len = offsetof (struct lf_nest_type, name) + name_len; if (subtype_len % 4 != 0) subtype_len += 4 - (subtype_len % 4); if (left < subtype_len) { einfo (_("%P: warning: truncated CodeView type record" " LF_FIELDLIST\n")); return false; } ptr += subtype_len; left -= subtype_len; break; } default: einfo (_("%P: warning: unrecognized CodeView subtype %v\n"), subtype); return false; } } break; } case LF_BITFIELD: { struct lf_bitfield *bf = (struct lf_bitfield *) data; if (size < offsetof (struct lf_bitfield, length)) { einfo (_("%P: warning: truncated CodeView type record" " LF_BITFIELD\n")); return false; } if (!remap_type (&bf->base_type, map, type_num, num_types)) return false; break; } case LF_METHODLIST: { struct lf_methodlist *ml = (struct lf_methodlist *) data; unsigned int num_entries; if (size < offsetof (struct lf_methodlist, entries)) { einfo (_("%P: warning: truncated CodeView type record" " LF_METHODLIST\n")); return false; } if ((size - offsetof (struct lf_methodlist, entries)) % sizeof (struct lf_methodlist_entry)) { einfo (_("%P: warning: malformed CodeView type record" " LF_METHODLIST\n")); return false; } num_entries = (size - offsetof (struct lf_methodlist, entries)) / sizeof (struct lf_methodlist_entry); for (unsigned int i = 0; i < num_entries; i++) { if (!remap_type (&ml->entries[i].method_type, map, type_num, num_types)) return false; } break; } case LF_ARRAY: { struct lf_array *arr = (struct lf_array *) data; if (size < offsetof (struct lf_array, length_in_bytes)) { einfo (_("%P: warning: truncated CodeView type record" " LF_ARRAY\n")); return false; } if (!remap_type (&arr->element_type, map, type_num, num_types)) return false; if (!remap_type (&arr->index_type, map, type_num, num_types)) return false; break; } case LF_CLASS: case LF_STRUCTURE: { struct lf_class *cl = (struct lf_class *) data; uint16_t prop, num_bytes; size_t name_len, name_off; if (size < offsetof (struct lf_class, name)) { einfo (_("%P: warning: truncated CodeView type record" " LF_CLASS/LF_STRUCTURE\n")); return false; } if (!remap_type (&cl->field_list, map, type_num, num_types)) return false; if (!remap_type (&cl->derived_from, map, type_num, num_types)) return false; if (!remap_type (&cl->vshape, map, type_num, num_types)) return false; name_off = offsetof (struct lf_class, name); num_bytes = bfd_getl16 (&cl->length); /* If num_bytes >= 0x8000, actual value follows. */ if (num_bytes >= 0x8000) { unsigned int param_len = extended_value_len (num_bytes); if (param_len == 0) { einfo (_("%P: warning: unhandled type %v within" " LF_CLASS/LF_STRUCTURE\n"), num_bytes); return false; } name_off += param_len; if (size < name_off) { einfo (_("%P: warning: truncated CodeView type record" " LF_CLASS/LF_STRUCTURE\n")); return false; } } name_len = strnlen ((char *) cl + name_off, size - name_off); if (name_len == size - name_off) { einfo (_("%P: warning: name for LF_CLASS/LF_STRUCTURE has no" " terminating zero\n")); return false; } prop = bfd_getl16 (&cl->properties); if (prop & CV_PROP_HAS_UNIQUE_NAME) { /* Structure has another name following first one. */ size_t len = name_off + name_len + 1; size_t unique_name_len; unique_name_len = strnlen ((char *) cl + name_off + name_len + 1, size - len); if (unique_name_len == size - len) { einfo (_("%P: warning: unique name for LF_CLASS/LF_STRUCTURE" " has no terminating zero\n")); return false; } } if (!(prop & (CV_PROP_FORWARD_REF | CV_PROP_SCOPED)) && !is_name_anonymous ((char *) cl + name_off, name_len)) { other_hash = true; cv_hash = crc32 ((uint8_t *) cl + name_off, name_len); } break; } case LF_UNION: { struct lf_union *un = (struct lf_union *) data; uint16_t prop, num_bytes; size_t name_len, name_off; if (size < offsetof (struct lf_union, name)) { einfo (_("%P: warning: truncated CodeView type record" " LF_UNION\n")); return false; } if (!remap_type (&un->field_list, map, type_num, num_types)) return false; name_off = offsetof (struct lf_union, name); num_bytes = bfd_getl16 (&un->length); /* If num_bytes >= 0x8000, actual value follows. */ if (num_bytes >= 0x8000) { unsigned int param_len = extended_value_len (num_bytes); if (param_len == 0) { einfo (_("%P: warning: unhandled type %v within" " LF_UNION\n"), num_bytes); return false; } name_off += param_len; if (size < name_off) { einfo (_("%P: warning: truncated CodeView type record" " LF_UNION\n")); return false; } } name_len = strnlen ((char *) un + name_off, size - name_off); if (name_len == size - name_off) { einfo (_("%P: warning: name for LF_UNION has no" " terminating zero\n")); return false; } prop = bfd_getl16 (&un->properties); if (prop & CV_PROP_HAS_UNIQUE_NAME) { /* Structure has another name following first one. */ size_t len = name_off + name_len + 1; size_t unique_name_len; unique_name_len = strnlen ((char *) un + name_off + name_len + 1, size - len); if (unique_name_len == size - len) { einfo (_("%P: warning: unique name for LF_UNION has" " no terminating zero\n")); return false; } } if (!(prop & (CV_PROP_FORWARD_REF | CV_PROP_SCOPED)) && !is_name_anonymous ((char *) un + name_off, name_len)) { other_hash = true; cv_hash = crc32 ((uint8_t *) un + name_off, name_len); } break; } case LF_ENUM: { struct lf_enum *en = (struct lf_enum *) data; uint16_t prop; size_t name_len; if (size < offsetof (struct lf_enum, name)) { einfo (_("%P: warning: truncated CodeView type record" " LF_ENUM\n")); return false; } if (!remap_type (&en->underlying_type, map, type_num, num_types)) return false; if (!remap_type (&en->field_list, map, type_num, num_types)) return false; name_len = strnlen (en->name, size - offsetof (struct lf_enum, name)); if (name_len == size - offsetof (struct lf_enum, name)) { einfo (_("%P: warning: name for LF_ENUM has no" " terminating zero\n")); return false; } prop = bfd_getl16 (&en->properties); if (prop & CV_PROP_HAS_UNIQUE_NAME) { /* Structure has another name following first one. */ size_t len = offsetof (struct lf_enum, name) + name_len + 1; size_t unique_name_len; unique_name_len = strnlen (en->name + name_len + 1, size - len); if (unique_name_len == size - len) { einfo (_("%P: warning: unique name for LF_ENUM has" " no terminating zero\n")); return false; } } break; } case LF_VTSHAPE: /* Does not reference any types, nothing to be done. */ break; case LF_VFTABLE: { struct lf_vftable *vft = (struct lf_vftable *) data; if (size < offsetof (struct lf_vftable, names)) { einfo (_("%P: warning: truncated CodeView type record" " LF_VFTABLE\n")); return false; } if (!remap_type (&vft->type, map, type_num, num_types)) return false; if (!remap_type (&vft->base_vftable, map, type_num, num_types)) return false; break; } case LF_STRING_ID: { struct lf_string_id *str = (struct lf_string_id *) data; size_t string_len; if (size < offsetof (struct lf_string_id, string)) { einfo (_("%P: warning: truncated CodeView type record" " LF_STRING_ID\n")); return false; } if (!remap_type (&str->substring, map, type_num, num_types)) return false; string_len = strnlen (str->string, size - offsetof (struct lf_string_id, string)); if (string_len == size - offsetof (struct lf_string_id, string)) { einfo (_("%P: warning: string for LF_STRING_ID has no" " terminating zero\n")); return false; } ipi = true; break; } case LF_SUBSTR_LIST: { uint32_t num_entries; struct lf_arglist *ssl = (struct lf_arglist *) data; if (size < offsetof (struct lf_arglist, args)) { einfo (_("%P: warning: truncated CodeView type record" " LF_SUBSTR_LIST\n")); return false; } num_entries = bfd_getl32 (&ssl->num_entries); if (size < offsetof (struct lf_arglist, args) + (num_entries * sizeof (uint32_t))) { einfo (_("%P: warning: truncated CodeView type record" " LF_SUBSTR_LIST\n")); return false; } for (uint32_t i = 0; i < num_entries; i++) { if (!remap_type (&ssl->args[i], map, type_num, num_types)) return false; } ipi = true; break; } case LF_BUILDINFO: { uint16_t num_entries; struct lf_build_info *bi = (struct lf_build_info *) data; if (size < offsetof (struct lf_build_info, strings)) { einfo (_("%P: warning: truncated CodeView type record" " LF_BUILDINFO\n")); return false; } num_entries = bfd_getl16 (&bi->count); if (size < offsetof (struct lf_build_info, strings) + (num_entries * sizeof (uint32_t))) { einfo (_("%P: warning: truncated CodeView type record" " LF_BUILDINFO\n")); return false; } for (uint32_t i = 0; i < num_entries; i++) { if (!remap_type (&bi->strings[i], map, type_num, num_types)) return false; } ipi = true; break; } case LF_FUNC_ID: { struct lf_func_id *func = (struct lf_func_id *) data; size_t name_len; if (size < offsetof (struct lf_func_id, name)) { einfo (_("%P: warning: truncated CodeView type record" " LF_FUNC_ID\n")); return false; } if (!remap_type (&func->parent_scope, map, type_num, num_types)) return false; if (!remap_type (&func->function_type, map, type_num, num_types)) return false; name_len = strnlen (func->name, size - offsetof (struct lf_func_id, name)); if (name_len == size - offsetof (struct lf_func_id, name)) { einfo (_("%P: warning: string for LF_FUNC_ID has no" " terminating zero\n")); return false; } ipi = true; break; } case LF_MFUNC_ID: { struct lf_mfunc_id *mfunc = (struct lf_mfunc_id *) data; size_t name_len; if (size < offsetof (struct lf_mfunc_id, name)) { einfo (_("%P: warning: truncated CodeView type record" " LF_MFUNC_ID\n")); return false; } if (!remap_type (&mfunc->parent_type, map, type_num, num_types)) return false; if (!remap_type (&mfunc->function_type, map, type_num, num_types)) return false; name_len = strnlen (mfunc->name, size - offsetof (struct lf_mfunc_id, name)); if (name_len == size - offsetof (struct lf_mfunc_id, name)) { einfo (_("%P: warning: string for LF_MFUNC_ID has no" " terminating zero\n")); return false; } ipi = true; break; } case LF_UDT_SRC_LINE: return handle_udt_src_line (data, size, map, type_num, num_types, ids, mod_num, strings); default: einfo (_("%P: warning: unrecognized CodeView type %v\n"), type); return false; } hash = iterative_hash (data, size, 0); t = ipi ? ids : types; slot = htab_find_slot_with_hash (t->hashmap, data, hash, INSERT); if (!slot) return false; if (!*slot) /* new entry */ { struct type_entry *e; *slot = xmalloc (offsetof (struct type_entry, data) + size); e = (struct type_entry *) *slot; e->next = NULL; e->index = t->num_types; if (other_hash) e->cv_hash = cv_hash; else e->cv_hash = crc32 (data, size); e->has_udt_src_line = false; memcpy (e->data, data, size); if (t->last) t->last->next = e; else t->first = e; t->last = e; map[type_num] = e; t->num_types++; } else /* duplicate */ { map[type_num] = (struct type_entry *) *slot; } return true; } /* Parse the .debug$T section of a module, and pass any type definitions found to handle_type. */ static bool handle_debugt_section (asection *s, bfd *mod, struct types *types, struct types *ids, uint16_t mod_num, struct string_table *strings, struct type_entry ***map, uint32_t *num_types) { bfd_byte *data = NULL; size_t off; uint32_t type_num; if (!bfd_get_full_section_contents (mod, s, &data)) return false; if (!data) return false; if (bfd_getl32 (data) != CV_SIGNATURE_C13) { free (data); return true; } off = sizeof (uint32_t); while (off + sizeof (uint16_t) <= s->size) { uint16_t size; size = bfd_getl16 (data + off); off += sizeof (uint16_t); if (size + off > s->size || size <= sizeof (uint16_t)) { free (data); bfd_set_error (bfd_error_bad_value); return false; } (*num_types)++; off += size; } if (*num_types == 0) { free (data); return true; } *map = xcalloc (*num_types, sizeof (struct type_entry *)); off = sizeof (uint32_t); type_num = 0; while (off + sizeof (uint16_t) <= s->size) { uint16_t size; size = bfd_getl16 (data + off); if (!handle_type (data + off, *map, type_num, *num_types, types, ids, mod_num, strings)) { free (data); free (*map); bfd_set_error (bfd_error_bad_value); return false; } off += sizeof (uint16_t) + size; type_num++; } free (data); return true; } /* Return the CodeView constant for the selected architecture. */ static uint16_t target_processor (bfd *abfd) { switch (abfd->arch_info->arch) { case bfd_arch_i386: if (abfd->arch_info->mach & bfd_mach_x86_64) return CV_CFL_X64; else return CV_CFL_80386; case bfd_arch_aarch64: return CV_CFL_ARM64; default: return 0; } } /* Create the symbols that go in "* Linker *", the dummy module created for the linker itself. */ static bool create_linker_symbols (bfd *abfd, uint8_t **syms, uint32_t *sym_byte_size, const char *pdb_name) { uint8_t *ptr; struct objname *name; struct compile3 *comp; struct envblock *env; size_t padding1, padding2, env_size; char *cwdval, *exeval, *pdbval; /* extra NUL for padding */ static const char linker_fn[] = "* Linker *\0"; static const char linker_name[] = "GNU LD " VERSION; static const char cwd[] = "cwd"; static const char exe[] = "exe"; static const char pdb[] = "pdb"; cwdval = getcwd (NULL, 0); if (!cwdval) { einfo (_("%P: warning: unable to get working directory\n")); return false; } exeval = lrealpath (program_name); if (!exeval) { einfo (_("%P: warning: unable to get program name\n")); free (cwdval); return false; } pdbval = lrealpath (pdb_name); if (!pdbval) { einfo (_("%P: warning: unable to get full path to PDB\n")); free (exeval); free (cwdval); return false; } *sym_byte_size += offsetof (struct objname, name) + sizeof (linker_fn); *sym_byte_size += offsetof (struct compile3, compiler) + sizeof (linker_name); if (*sym_byte_size % 4) padding1 = 4 - (*sym_byte_size % 4); else padding1 = 0; *sym_byte_size += padding1; env_size = offsetof (struct envblock, strings); env_size += sizeof (cwd); env_size += strlen (cwdval) + 1; env_size += sizeof (exe); env_size += strlen (exeval) + 1; env_size += sizeof (pdb); env_size += strlen (pdbval) + 1; if (env_size % 4) padding2 = 4 - (env_size % 4); else padding2 = 0; env_size += padding2; *sym_byte_size += env_size; *syms = xmalloc (*sym_byte_size); ptr = *syms; /* Write S_OBJNAME */ name = (struct objname *) ptr; bfd_putl16 (offsetof (struct objname, name) + sizeof (linker_fn) - sizeof (uint16_t), &name->size); bfd_putl16 (S_OBJNAME, &name->kind); bfd_putl32 (0, &name->signature); memcpy (name->name, linker_fn, sizeof (linker_fn)); ptr += offsetof (struct objname, name) + sizeof (linker_fn); /* Write S_COMPILE3 */ comp = (struct compile3 *) ptr; bfd_putl16 (offsetof (struct compile3, compiler) + sizeof (linker_name) + padding1 - sizeof (uint16_t), &comp->size); bfd_putl16 (S_COMPILE3, &comp->kind); bfd_putl32 (CV_CFL_LINK, &comp->flags); bfd_putl16 (target_processor (abfd), &comp->machine); bfd_putl16 (0, &comp->frontend_major); bfd_putl16 (0, &comp->frontend_minor); bfd_putl16 (0, &comp->frontend_build); bfd_putl16 (0, &comp->frontend_qfe); bfd_putl16 (0, &comp->backend_major); bfd_putl16 (0, &comp->backend_minor); bfd_putl16 (0, &comp->backend_build); bfd_putl16 (0, &comp->backend_qfe); memcpy (comp->compiler, linker_name, sizeof (linker_name)); memset (comp->compiler + sizeof (linker_name), 0, padding1); ptr += offsetof (struct compile3, compiler) + sizeof (linker_name) + padding1; /* Write S_ENVBLOCK */ env = (struct envblock *) ptr; bfd_putl16 (env_size - sizeof (uint16_t), &env->size); bfd_putl16 (S_ENVBLOCK, &env->kind); env->flags = 0; ptr += offsetof (struct envblock, strings); memcpy (ptr, cwd, sizeof (cwd)); ptr += sizeof (cwd); memcpy (ptr, cwdval, strlen (cwdval) + 1); ptr += strlen (cwdval) + 1; memcpy (ptr, exe, sizeof (exe)); ptr += sizeof (exe); memcpy (ptr, exeval, strlen (exeval) + 1); ptr += strlen (exeval) + 1; memcpy (ptr, pdb, sizeof (pdb)); ptr += sizeof (pdb); memcpy (ptr, pdbval, strlen (pdbval) + 1); ptr += strlen (pdbval) + 1; /* Microsoft's LINK also includes "cmd", the command-line options passed to the linker, but unfortunately we don't have access to argc and argv at this stage. */ memset (ptr, 0, padding2); free (pdbval); free (exeval); free (cwdval); return true; } /* Populate the module stream, which consists of the transformed .debug$S data for each object file. */ static bool populate_module_stream (bfd *stream, bfd *mod, uint32_t *sym_byte_size, struct string_table *strings, uint32_t *c13_info_size, struct mod_source_files *mod_source, bfd *abfd, struct types *types, struct types *ids, uint16_t mod_num, bfd *sym_rec_stream, struct globals *glob, const char *pdb_name) { uint8_t int_buf[sizeof (uint32_t)]; uint8_t *c13_info = NULL; uint8_t *syms = NULL; *sym_byte_size = 0; *c13_info_size = 0; if (!strcmp (bfd_get_filename (mod), "dll stuff")) { if (!create_linker_symbols (mod, &syms, sym_byte_size, pdb_name)) return false; } else { struct type_entry **map = NULL; uint32_t num_types = 0; /* Process .debug$T section. */ for (asection *s = mod->sections; s; s = s->next) { if (!strcmp (s->name, ".debug$T") && s->size >= sizeof (uint32_t)) { if (!handle_debugt_section (s, mod, types, ids, mod_num, strings, &map, &num_types)) { free (mod_source->files); return false; } break; } } /* Process .debug$S section(s). */ for (asection *s = mod->sections; s; s = s->next) { if (!strcmp (s->name, ".debug$S") && s->size >= sizeof (uint32_t)) { if (!handle_debugs_section (s, mod, strings, &c13_info, c13_info_size, mod_source, abfd, &syms, sym_byte_size, map, num_types, sym_rec_stream, glob, mod_num)) { free (c13_info); free (syms); free (mod_source->files); free (map); return false; } } } free (map); } /* Write the signature. */ bfd_putl32 (CV_SIGNATURE_C13, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), stream) != sizeof (uint32_t)) { free (c13_info); free (syms); return false; } if (syms) { if (bfd_bwrite (syms, *sym_byte_size, stream) != *sym_byte_size) { free (c13_info); free (syms); return false; } free (syms); } if (c13_info) { if (bfd_bwrite (c13_info, *c13_info_size, stream) != *c13_info_size) { free (c13_info); return false; } free (c13_info); } /* Write the global refs size. */ bfd_putl32 (0, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), stream) != sizeof (uint32_t)) return false; return true; } /* Create the module info substream within the DBI. */ static bool create_module_info_substream (bfd *abfd, bfd *pdb, void **data, uint32_t *size, struct string_table *strings, struct source_files_info *source, struct types *types, struct types *ids, bfd *sym_rec_stream, struct globals *glob, const char *pdb_name) { uint8_t *ptr; unsigned int mod_num; static const char linker_fn[] = "* Linker *"; *size = 0; for (bfd *in = coff_data (abfd)->link_info->input_bfds; in; in = in->link.next) { size_t len = sizeof (struct module_info); if (!strcmp (bfd_get_filename (in), "dll stuff")) { len += sizeof (linker_fn); /* Object name. */ len++; /* Empty module name. */ } else if (in->my_archive) { char *name = lrealpath (bfd_get_filename (in)); len += strlen (name) + 1; /* Object name. */ free (name); name = lrealpath (bfd_get_filename (in->my_archive)); len += strlen (name) + 1; /* Archive name. */ free (name); } else { char *name = lrealpath (bfd_get_filename (in)); size_t name_len = strlen (name) + 1; len += name_len; /* Object name. */ len += name_len; /* And again as the archive name. */ free (name); } if (len % 4) len += 4 - (len % 4); *size += len; source->mod_count++; } *data = xmalloc (*size); ptr = *data; source->mods = xmalloc (source->mod_count * sizeof (struct mod_source_files)); memset (source->mods, 0, source->mod_count * sizeof (struct mod_source_files)); mod_num = 0; for (bfd *in = coff_data (abfd)->link_info->input_bfds; in; in = in->link.next) { struct module_info *mod = (struct module_info *) ptr; uint16_t stream_num; bfd *stream; uint32_t sym_byte_size, c13_info_size; uint8_t *start = ptr; stream = add_stream (pdb, NULL, &stream_num); if (!stream) { for (unsigned int i = 0; i < source->mod_count; i++) { free (source->mods[i].files); } free (source->mods); free (*data); return false; } if (!populate_module_stream (stream, in, &sym_byte_size, strings, &c13_info_size, &source->mods[mod_num], abfd, types, ids, mod_num, sym_rec_stream, glob, pdb_name)) { for (unsigned int i = 0; i < source->mod_count; i++) { free (source->mods[i].files); } free (source->mods); free (*data); return false; } bfd_putl32 (0, &mod->unused1); /* These are dummy values - MSVC copies the first section contribution entry here, but doesn't seem to use it for anything. */ bfd_putl16 (0xffff, &mod->sc.section); bfd_putl16 (0, &mod->sc.padding1); bfd_putl32 (0, &mod->sc.offset); bfd_putl32 (0xffffffff, &mod->sc.size); bfd_putl32 (0, &mod->sc.characteristics); bfd_putl16 (0xffff, &mod->sc.module_index); bfd_putl16 (0, &mod->sc.padding2); bfd_putl32 (0, &mod->sc.data_crc); bfd_putl32 (0, &mod->sc.reloc_crc); bfd_putl16 (0, &mod->flags); bfd_putl16 (stream_num, &mod->module_sym_stream); bfd_putl32 (sizeof (uint32_t) + sym_byte_size, &mod->sym_byte_size); bfd_putl32 (0, &mod->c11_byte_size); bfd_putl32 (c13_info_size, &mod->c13_byte_size); bfd_putl16 (0, &mod->source_file_count); bfd_putl16 (0, &mod->padding); bfd_putl32 (0, &mod->unused2); bfd_putl32 (0, &mod->source_file_name_index); bfd_putl32 (0, &mod->pdb_file_path_name_index); ptr += sizeof (struct module_info); if (!strcmp (bfd_get_filename (in), "dll stuff")) { /* Object name. */ memcpy (ptr, linker_fn, sizeof (linker_fn)); ptr += sizeof (linker_fn); /* Empty module name. */ *ptr = 0; ptr++; } else if (in->my_archive) { char *name = lrealpath (bfd_get_filename (in)); size_t name_len = strlen (name) + 1; /* Object name. */ memcpy (ptr, name, name_len); ptr += name_len; free (name); name = lrealpath (bfd_get_filename (in->my_archive)); name_len = strlen (name) + 1; /* Archive name. */ memcpy (ptr, name, name_len); ptr += name_len; free (name); } else { char *name = lrealpath (bfd_get_filename (in)); size_t name_len = strlen (name) + 1; /* Object name. */ memcpy (ptr, name, name_len); ptr += name_len; /* Object name again as archive name. */ memcpy (ptr, name, name_len); ptr += name_len; free (name); } /* Pad to next four-byte boundary. */ if ((ptr - start) % 4) { memset (ptr, 0, 4 - ((ptr - start) % 4)); ptr += 4 - ((ptr - start) % 4); } mod_num++; } return true; } /* Return the index of a given output section. */ static uint16_t find_section_number (bfd *abfd, asection *sect) { uint16_t i = 1; for (asection *s = abfd->sections; s; s = s->next) { if (s == sect) return i; /* Empty sections aren't output. */ if (s->size != 0) i++; } return 0; } /* Used as parameter to qsort, to sort section contributions by section and offset. */ static int section_contribs_compare (const void *p1, const void *p2) { const struct in_sc *sc1 = p1; const struct in_sc *sc2 = p2; if (sc1->sect_num < sc2->sect_num) return -1; if (sc1->sect_num > sc2->sect_num) return 1; if (sc1->s->output_offset < sc2->s->output_offset) return -1; if (sc1->s->output_offset > sc2->s->output_offset) return 1; return 0; } /* Create the substream which maps addresses in the image file to locations in the original object files. */ static bool create_section_contrib_substream (bfd *abfd, void **data, uint32_t *size) { unsigned int num_sc = 0; struct section_contribution *sc; uint16_t mod_index; char *sect_flags; file_ptr offset; struct in_sc *sc_in, *sc2; uint32_t *ptr; for (bfd *in = coff_data (abfd)->link_info->input_bfds; in; in = in->link.next) { for (asection *s = in->sections; s; s = s->next) { if (s->size == 0 || discarded_section (s)) continue; num_sc++; } } *size = sizeof (uint32_t) + (num_sc * sizeof (struct section_contribution)); *data = xmalloc (*size); bfd_putl32 (SECTION_CONTRIB_VERSION_60, *data); /* Read characteristics of outputted sections. */ sect_flags = xmalloc (sizeof (uint32_t) * abfd->section_count); offset = bfd_coff_filhsz (abfd) + bfd_coff_aoutsz (abfd); offset += offsetof (struct external_scnhdr, s_flags); for (unsigned int i = 0; i < abfd->section_count; i++) { bfd_seek (abfd, offset, SEEK_SET); if (bfd_bread (sect_flags + (i * sizeof (uint32_t)), sizeof (uint32_t), abfd) != sizeof (uint32_t)) { free (*data); free (sect_flags); return false; } offset += sizeof (struct external_scnhdr); } /* Microsoft's DIA expects section contributions to be sorted by section number and offset, otherwise it will be unable to resolve line numbers. */ sc_in = xmalloc (num_sc * sizeof (* sc_in)); sc2 = sc_in; mod_index = 0; for (bfd *in = coff_data (abfd)->link_info->input_bfds; in; in = in->link.next) { for (asection *s = in->sections; s; s = s->next) { if (s->size == 0 || discarded_section (s)) continue; sc2->s = s; sc2->sect_num = find_section_number (abfd, s->output_section); sc2->mod_index = mod_index; sc2++; } mod_index++; } qsort (sc_in, num_sc, sizeof (* sc_in), section_contribs_compare); ptr = *data; sc = (struct section_contribution *) (ptr + 1); /* Skip the version word. */ for (unsigned int i = 0; i < num_sc; i++) { memcpy (&sc->characteristics, sect_flags + ((sc_in[i].sect_num - 1) * sizeof (uint32_t)), sizeof (uint32_t)); bfd_putl16 (sc_in[i].sect_num, &sc->section); bfd_putl16 (0, &sc->padding1); bfd_putl32 (sc_in[i].s->output_offset, &sc->offset); bfd_putl32 (sc_in[i].s->size, &sc->size); bfd_putl16 (sc_in[i].mod_index, &sc->module_index); bfd_putl16 (0, &sc->padding2); bfd_putl32 (0, &sc->data_crc); bfd_putl32 (0, &sc->reloc_crc); sc++; } free (sc_in); free (sect_flags); return true; } /* The source info substream lives within the DBI stream, and lists the source files for each object file (i.e. it's derived from the DEBUG_S_FILECHKSMS parts of the .debug$S sections). This is a bit superfluous, as the filenames are also available in the C13 parts of the module streams, but MSVC relies on it to work properly. */ static void create_source_info_substream (void **data, uint32_t *size, struct source_files_info *source) { uint16_t dedupe_source_files_count = 0; uint16_t source_files_count = 0; uint32_t strings_len = 0; uint8_t *ptr; /* Loop through the source files, marking unique filenames. The pointers here are for entries in the main string table, and so have already been deduplicated. */ for (uint16_t i = 0; i < source->mod_count; i++) { for (uint16_t j = 0; j < source->mods[i].files_count; j++) { if (source->mods[i].files[j]) { if (source->mods[i].files[j]->source_file_offset == 0xffffffff) { source->mods[i].files[j]->source_file_offset = strings_len; strings_len += source->mods[i].files[j]->len + 1; dedupe_source_files_count++; } source_files_count++; } } } *size = sizeof (uint16_t) + sizeof (uint16_t); *size += (sizeof (uint16_t) + sizeof (uint16_t)) * source->mod_count; *size += sizeof (uint32_t) * source_files_count; *size += strings_len; *data = xmalloc (*size); ptr = (uint8_t *) *data; /* Write header (module count and source file count). */ bfd_putl16 (source->mod_count, ptr); ptr += sizeof (uint16_t); bfd_putl16 (dedupe_source_files_count, ptr); ptr += sizeof (uint16_t); /* Write "ModIndices". As the LLVM documentation puts it, "this array is present, but does not appear to be useful". */ for (uint16_t i = 0; i < source->mod_count; i++) { bfd_putl16 (i, ptr); ptr += sizeof (uint16_t); } /* Write source file count for each module. */ for (uint16_t i = 0; i < source->mod_count; i++) { bfd_putl16 (source->mods[i].files_count, ptr); ptr += sizeof (uint16_t); } /* For each module, write the offsets within the string table for each source file. */ for (uint16_t i = 0; i < source->mod_count; i++) { for (uint16_t j = 0; j < source->mods[i].files_count; j++) { if (source->mods[i].files[j]) { bfd_putl32 (source->mods[i].files[j]->source_file_offset, ptr); ptr += sizeof (uint32_t); } } } /* Write the string table. We set source_file_offset to a dummy value for each entry we write, so we don't write duplicate filenames. */ for (uint16_t i = 0; i < source->mod_count; i++) { for (uint16_t j = 0; j < source->mods[i].files_count; j++) { if (source->mods[i].files[j] && source->mods[i].files[j]->source_file_offset != 0xffffffff) { memcpy (ptr, source->mods[i].files[j]->s, source->mods[i].files[j]->len); ptr += source->mods[i].files[j]->len; *ptr = 0; ptr++; source->mods[i].files[j]->source_file_offset = 0xffffffff; } } } } /* Used as parameter to qsort, to sort globals by hash. */ static int global_compare_hash (const void *s1, const void *s2) { const struct global *g1 = *(const struct global **) s1; const struct global *g2 = *(const struct global **) s2; if (g1->hash < g2->hash) return -1; if (g1->hash > g2->hash) return 1; return 0; } /* Create the globals stream, which contains the unmangled symbol names. */ static bool create_globals_stream (bfd *pdb, struct globals *glob, uint16_t *stream_num) { bfd *stream; struct globals_hash_header h; uint32_t buckets_size, filled_buckets = 0; struct global **sorted = NULL; bool ret = false; struct global *buckets[NUM_GLOBALS_HASH_BUCKETS]; char int_buf[sizeof (uint32_t)]; stream = add_stream (pdb, NULL, stream_num); if (!stream) return false; memset (buckets, 0, sizeof (buckets)); if (glob->num_entries > 0) { struct global *g; /* Create an array of pointers, sorted by hash value. */ sorted = xmalloc (sizeof (struct global *) * glob->num_entries); g = glob->first; for (unsigned int i = 0; i < glob->num_entries; i++) { sorted[i] = g; g = g->next; } qsort (sorted, glob->num_entries, sizeof (struct global *), global_compare_hash); /* Populate the buckets. */ for (unsigned int i = 0; i < glob->num_entries; i++) { if (!buckets[sorted[i]->hash]) { buckets[sorted[i]->hash] = sorted[i]; filled_buckets++; } sorted[i]->index = i; } } buckets_size = NUM_GLOBALS_HASH_BUCKETS / 8; buckets_size += sizeof (uint32_t); buckets_size += filled_buckets * sizeof (uint32_t); bfd_putl32 (GLOBALS_HASH_SIGNATURE, &h.signature); bfd_putl32 (GLOBALS_HASH_VERSION_70, &h.version); bfd_putl32 (glob->num_entries * sizeof (struct hash_record), &h.entries_size); bfd_putl32 (buckets_size, &h.buckets_size); if (bfd_bwrite (&h, sizeof (h), stream) != sizeof (h)) return false; /* Write hash entries, sorted by hash. */ for (unsigned int i = 0; i < glob->num_entries; i++) { struct hash_record hr; bfd_putl32 (sorted[i]->offset + 1, &hr.offset); bfd_putl32 (sorted[i]->refcount, &hr.reference); if (bfd_bwrite (&hr, sizeof (hr), stream) != sizeof (hr)) goto end; } /* Write the bitmap for filled and unfilled buckets. */ for (unsigned int i = 0; i < NUM_GLOBALS_HASH_BUCKETS; i += 8) { uint8_t v = 0; for (unsigned int j = 0; j < 8; j++) { if (buckets[i + j]) v |= 1 << j; } if (bfd_bwrite (&v, sizeof (v), stream) != sizeof (v)) goto end; } /* Add a 4-byte gap. */ bfd_putl32 (0, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), stream) != sizeof (uint32_t)) goto end; /* Write the bucket offsets. */ for (unsigned int i = 0; i < NUM_GLOBALS_HASH_BUCKETS; i++) { if (buckets[i]) { /* 0xc is size of internal hash_record structure in Microsoft's parser. */ bfd_putl32 (buckets[i]->index * 0xc, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), stream) != sizeof (uint32_t)) goto end; } } ret = true; end: free (sorted); return ret; } /* Hash an entry in the globals list. */ static hashval_t hash_global_entry (const void *p) { const struct global *g = (const struct global *) p; uint16_t len = bfd_getl16 (g->data); return iterative_hash (g->data, len, 0); } /* Compare an entry in the globals list with a symbol. */ static int eq_global_entry (const void *a, const void *b) { const struct global *g = (const struct global *) a; uint16_t len1, len2; len1 = bfd_getl16 (g->data) + sizeof (uint16_t); len2 = bfd_getl16 (b) + sizeof (uint16_t); if (len1 != len2) return 0; return !memcmp (g->data, b, len1); } /* Stream 4 is the debug information (DBI) stream. */ static bool populate_dbi_stream (bfd *stream, bfd *abfd, bfd *pdb, uint16_t section_header_stream_num, uint16_t sym_rec_stream_num, uint16_t publics_stream_num, struct string_table *strings, struct types *types, struct types *ids, bfd *sym_rec_stream, const char *pdb_name) { struct pdb_dbi_stream_header h; struct optional_dbg_header opt; void *mod_info, *sc, *source_info; uint32_t mod_info_size, sc_size, source_info_size; struct source_files_info source; struct globals glob; uint16_t globals_stream_num; source.mod_count = 0; source.mods = NULL; glob.num_entries = 0; glob.first = NULL; glob.last = NULL; glob.hashmap = htab_create_alloc (0, hash_global_entry, eq_global_entry, free, xcalloc, free); if (!create_module_info_substream (abfd, pdb, &mod_info, &mod_info_size, strings, &source, types, ids, sym_rec_stream, &glob, pdb_name)) { htab_delete (glob.hashmap); return false; } if (!create_globals_stream (pdb, &glob, &globals_stream_num)) { htab_delete (glob.hashmap); for (unsigned int i = 0; i < source.mod_count; i++) { free (source.mods[i].files); } free (source.mods); free (mod_info); return false; } htab_delete (glob.hashmap); if (!create_section_contrib_substream (abfd, &sc, &sc_size)) { for (unsigned int i = 0; i < source.mod_count; i++) { free (source.mods[i].files); } free (source.mods); free (mod_info); return false; } create_source_info_substream (&source_info, &source_info_size, &source); for (unsigned int i = 0; i < source.mod_count; i++) { free (source.mods[i].files); } free (source.mods); bfd_putl32 (0xffffffff, &h.version_signature); bfd_putl32 (DBI_STREAM_VERSION_70, &h.version_header); bfd_putl32 (1, &h.age); bfd_putl16 (globals_stream_num, &h.global_stream_index); bfd_putl16 (0x8e1d, &h.build_number); // MSVC 14.29 bfd_putl16 (publics_stream_num, &h.public_stream_index); bfd_putl16 (0, &h.pdb_dll_version); bfd_putl16 (sym_rec_stream_num, &h.sym_record_stream); bfd_putl16 (0, &h.pdb_dll_rbld); bfd_putl32 (mod_info_size, &h.mod_info_size); bfd_putl32 (sc_size, &h.section_contribution_size); bfd_putl32 (0, &h.section_map_size); bfd_putl32 (source_info_size, &h.source_info_size); bfd_putl32 (0, &h.type_server_map_size); bfd_putl32 (0, &h.mfc_type_server_index); bfd_putl32 (sizeof (opt), &h.optional_dbg_header_size); bfd_putl32 (0, &h.ec_substream_size); bfd_putl16 (0, &h.flags); bfd_putl16 (get_arch_number (abfd), &h.machine); bfd_putl32 (0, &h.padding); if (bfd_bwrite (&h, sizeof (h), stream) != sizeof (h)) { free (source_info); free (sc); free (mod_info); return false; } if (bfd_bwrite (mod_info, mod_info_size, stream) != mod_info_size) { free (source_info); free (sc); free (mod_info); return false; } free (mod_info); if (bfd_bwrite (sc, sc_size, stream) != sc_size) { free (source_info); free (sc); return false; } free (sc); if (bfd_bwrite (source_info, source_info_size, stream) != source_info_size) { free (source_info); return false; } free (source_info); bfd_putl16 (0xffff, &opt.fpo_stream); bfd_putl16 (0xffff, &opt.exception_stream); bfd_putl16 (0xffff, &opt.fixup_stream); bfd_putl16 (0xffff, &opt.omap_to_src_stream); bfd_putl16 (0xffff, &opt.omap_from_src_stream); bfd_putl16 (section_header_stream_num, &opt.section_header_stream); bfd_putl16 (0xffff, &opt.token_map_stream); bfd_putl16 (0xffff, &opt.xdata_stream); bfd_putl16 (0xffff, &opt.pdata_stream); bfd_putl16 (0xffff, &opt.new_fpo_stream); bfd_putl16 (0xffff, &opt.orig_section_header_stream); if (bfd_bwrite (&opt, sizeof (opt), stream) != sizeof (opt)) return false; return true; } /* Used as parameter to qsort, to sort publics by hash. */ static int public_compare_hash (const void *s1, const void *s2) { const struct public *p1 = *(const struct public **) s1; const struct public *p2 = *(const struct public **) s2; if (p1->hash < p2->hash) return -1; if (p1->hash > p2->hash) return 1; return 0; } /* Used as parameter to qsort, to sort publics by address. */ static int public_compare_addr (const void *s1, const void *s2) { const struct public *p1 = *(const struct public **) s1; const struct public *p2 = *(const struct public **) s2; if (p1->section < p2->section) return -1; if (p1->section > p2->section) return 1; if (p1->address < p2->address) return -1; if (p1->address > p2->address) return 1; return 0; } /* The publics stream is a hash map of S_PUB32 records, which are stored in the symbol record stream. Each S_PUB32 entry represents a symbol from the point of view of the linker: a section index, an offset within the section, and a mangled name. Compare with S_GDATA32 and S_GPROC32, which are the same thing but generated by the compiler. */ static bool populate_publics_stream (bfd *stream, bfd *abfd, bfd *sym_rec_stream) { struct publics_header header; struct globals_hash_header hash_header; const unsigned int num_buckets = 4096; unsigned int num_entries = 0, filled_buckets = 0; unsigned int buckets_size, sym_hash_size; char int_buf[sizeof (uint32_t)]; struct public *publics_head = NULL, *publics_tail = NULL; struct public **buckets; struct public **sorted = NULL; bool ret = false; buckets = xmalloc (sizeof (struct public *) * num_buckets); memset (buckets, 0, sizeof (struct public *) * num_buckets); /* Loop through the global symbols in our input files, and write S_PUB32 records in the symbol record stream for those that make it into the final image. */ for (bfd *in = coff_data (abfd)->link_info->input_bfds; in; in = in->link.next) { if (!in->outsymbols) continue; for (unsigned int i = 0; i < in->symcount; i++) { struct bfd_symbol *sym = in->outsymbols[i]; if (sym->flags & BSF_GLOBAL) { struct pubsym ps; uint16_t record_length; const char *name = sym->name; size_t name_len = strlen (name); struct public *p = xmalloc (sizeof (struct public)); unsigned int padding = 0; uint16_t section; uint32_t flags = 0; section = find_section_number (abfd, sym->section->output_section); if (section == 0) continue; p->next = NULL; p->offset = bfd_tell (sym_rec_stream); p->hash = calc_hash (name, name_len) % num_buckets; p->section = section; p->address = sym->section->output_offset + sym->value; record_length = sizeof (struct pubsym) + name_len + 1; if (record_length % 4) padding = 4 - (record_length % 4); /* Assume that all global symbols in executable sections are functions. */ if (sym->section->flags & SEC_CODE) flags = PUBSYM_FUNCTION; bfd_putl16 (record_length + padding - sizeof (uint16_t), &ps.record_length); bfd_putl16 (S_PUB32, &ps.record_type); bfd_putl32 (flags, &ps.flags); bfd_putl32 (p->address, &ps.offset); bfd_putl16 (p->section, &ps.section); if (bfd_bwrite (&ps, sizeof (struct pubsym), sym_rec_stream) != sizeof (struct pubsym)) goto end; if (bfd_bwrite (name, name_len + 1, sym_rec_stream) != name_len + 1) goto end; for (unsigned int j = 0; j < padding; j++) { uint8_t b = 0; if (bfd_bwrite (&b, sizeof (uint8_t), sym_rec_stream) != sizeof (uint8_t)) goto end; } if (!publics_head) publics_head = p; else publics_tail->next = p; publics_tail = p; num_entries++; } } } if (num_entries > 0) { /* Create an array of pointers, sorted by hash value. */ sorted = xmalloc (sizeof (struct public *) * num_entries); struct public *p = publics_head; for (unsigned int i = 0; i < num_entries; i++) { sorted[i] = p; p = p->next; } qsort (sorted, num_entries, sizeof (struct public *), public_compare_hash); /* Populate the buckets. */ for (unsigned int i = 0; i < num_entries; i++) { if (!buckets[sorted[i]->hash]) { buckets[sorted[i]->hash] = sorted[i]; filled_buckets++; } sorted[i]->index = i; } } buckets_size = num_buckets / 8; buckets_size += sizeof (uint32_t); buckets_size += filled_buckets * sizeof (uint32_t); sym_hash_size = sizeof (hash_header); sym_hash_size += num_entries * sizeof (struct hash_record); sym_hash_size += buckets_size; /* Output the publics header. */ bfd_putl32 (sym_hash_size, &header.sym_hash_size); bfd_putl32 (num_entries * sizeof (uint32_t), &header.addr_map_size); bfd_putl32 (0, &header.num_thunks); bfd_putl32 (0, &header.thunks_size); bfd_putl32 (0, &header.thunk_table); bfd_putl32 (0, &header.thunk_table_offset); bfd_putl32 (0, &header.num_sects); if (bfd_bwrite (&header, sizeof (header), stream) != sizeof (header)) goto end; /* Output the global hash header. */ bfd_putl32 (GLOBALS_HASH_SIGNATURE, &hash_header.signature); bfd_putl32 (GLOBALS_HASH_VERSION_70, &hash_header.version); bfd_putl32 (num_entries * sizeof (struct hash_record), &hash_header.entries_size); bfd_putl32 (buckets_size, &hash_header.buckets_size); if (bfd_bwrite (&hash_header, sizeof (hash_header), stream) != sizeof (hash_header)) goto end; /* Write the entries in hash order. */ for (unsigned int i = 0; i < num_entries; i++) { struct hash_record hr; bfd_putl32 (sorted[i]->offset + 1, &hr.offset); bfd_putl32 (1, &hr.reference); if (bfd_bwrite (&hr, sizeof (hr), stream) != sizeof (hr)) goto end; } /* Write the bitmap for filled and unfilled buckets. */ for (unsigned int i = 0; i < num_buckets; i += 8) { uint8_t v = 0; for (unsigned int j = 0; j < 8; j++) { if (buckets[i + j]) v |= 1 << j; } if (bfd_bwrite (&v, sizeof (v), stream) != sizeof (v)) goto end; } /* Add a 4-byte gap. */ bfd_putl32 (0, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), stream) != sizeof (uint32_t)) goto end; /* Write the bucket offsets. */ for (unsigned int i = 0; i < num_buckets; i++) { if (buckets[i]) { /* 0xc is size of internal hash_record structure in Microsoft's parser. */ bfd_putl32 (buckets[i]->index * 0xc, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), stream) != sizeof (uint32_t)) goto end; } } /* Write the address map: offsets into the symbol record stream of S_PUB32 records, ordered by address. */ if (num_entries > 0) { qsort (sorted, num_entries, sizeof (struct public *), public_compare_addr); for (unsigned int i = 0; i < num_entries; i++) { bfd_putl32 (sorted[i]->offset, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), stream) != sizeof (uint32_t)) goto end; } } ret = true; end: free (buckets); while (publics_head) { struct public *p = publics_head->next; free (publics_head); publics_head = p; } free (sorted); return ret; } /* The section header stream contains a copy of the section headers from the PE file, in the same format. */ static bool create_section_header_stream (bfd *pdb, bfd *abfd, uint16_t *num) { bfd *stream; unsigned int section_count; file_ptr scn_base; size_t len; char *buf; stream = add_stream (pdb, NULL, num); if (!stream) return false; section_count = abfd->section_count; /* Empty sections aren't output. */ for (asection *sect = abfd->sections; sect; sect = sect->next) { if (sect->size == 0) section_count--; } if (section_count == 0) return true; /* Copy section table from output - it's already been written at this point. */ scn_base = bfd_coff_filhsz (abfd) + bfd_coff_aoutsz (abfd); bfd_seek (abfd, scn_base, SEEK_SET); len = section_count * sizeof (struct external_scnhdr); buf = xmalloc (len); if (bfd_bread (buf, len, abfd) != len) { free (buf); return false; } if (bfd_bwrite (buf, len, stream) != len) { free (buf); return false; } free (buf); return true; } /* Populate the "/names" named stream, which contains the string table. */ static bool populate_names_stream (bfd *stream, struct string_table *strings) { char int_buf[sizeof (uint32_t)]; struct string_table_header h; uint32_t num_strings = 0, num_buckets; struct string **buckets; bfd_putl32 (STRING_TABLE_SIGNATURE, &h.signature); bfd_putl32 (STRING_TABLE_VERSION, &h.version); if (bfd_bwrite (&h, sizeof (h), stream) != sizeof (h)) return false; bfd_putl32 (strings->strings_len, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), stream) != sizeof (uint32_t)) return false; int_buf[0] = 0; if (bfd_bwrite (int_buf, 1, stream) != 1) return false; for (struct string *s = strings->strings_head; s; s = s->next) { if (bfd_bwrite (s->s, s->len, stream) != s->len) return false; if (bfd_bwrite (int_buf, 1, stream) != 1) return false; num_strings++; } num_buckets = num_strings * 2; buckets = xmalloc (sizeof (struct string *) * num_buckets); memset (buckets, 0, sizeof (struct string *) * num_buckets); for (struct string *s = strings->strings_head; s; s = s->next) { uint32_t bucket_num = s->hash % num_buckets; while (buckets[bucket_num]) { bucket_num++; if (bucket_num == num_buckets) bucket_num = 0; } buckets[bucket_num] = s; } bfd_putl32 (num_buckets, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), stream) != sizeof (uint32_t)) { free (buckets); return false; } for (unsigned int i = 0; i < num_buckets; i++) { if (buckets[i]) bfd_putl32 (buckets[i]->offset, int_buf); else bfd_putl32 (0, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), stream) != sizeof (uint32_t)) { free (buckets); return false; } } free (buckets); bfd_putl32 (num_strings, int_buf); if (bfd_bwrite (int_buf, sizeof (uint32_t), stream) != sizeof (uint32_t)) return false; return true; } /* Calculate the hash of a type_entry. */ static hashval_t hash_type_entry (const void *p) { const struct type_entry *e = (const struct type_entry *) p; uint16_t size = bfd_getl16 (e->data) + sizeof (uint16_t); return iterative_hash (e->data, size, 0); } /* Compare a type_entry with a type. */ static int eq_type_entry (const void *a, const void *b) { const struct type_entry *e = (const struct type_entry *) a; uint16_t size_a = bfd_getl16 (e->data); uint16_t size_b = bfd_getl16 (b); if (size_a != size_b) return 0; return memcmp (e->data + sizeof (uint16_t), (const uint8_t *) b + sizeof (uint16_t), size_a) == 0; } /* Create a PDB debugging file for the PE image file abfd with the build ID guid, stored at pdb_name. */ bool create_pdb_file (bfd *abfd, const char *pdb_name, const unsigned char *guid) { bfd *pdb; bool ret = false; bfd *info_stream, *dbi_stream, *names_stream, *sym_rec_stream, *publics_stream, *tpi_stream, *ipi_stream; uint16_t section_header_stream_num, sym_rec_stream_num, publics_stream_num; struct string_table strings; struct types types, ids; pdb = bfd_openw (pdb_name, "pdb"); if (!pdb) { einfo (_("%P: warning: cannot create PDB file: %E\n")); return false; } strings.strings_head = NULL; strings.strings_tail = NULL; strings.strings_len = 1; strings.hashmap = htab_create_alloc (0, hash_string_table_entry, eq_string_table_entry, free, xcalloc, free); bfd_set_format (pdb, bfd_archive); if (!create_old_directory_stream (pdb)) { einfo (_("%P: warning: cannot create old directory stream " "in PDB file: %E\n")); goto end; } info_stream = add_stream (pdb, NULL, NULL); if (!info_stream) { einfo (_("%P: warning: cannot create info stream " "in PDB file: %E\n")); goto end; } tpi_stream = add_stream (pdb, NULL, NULL); if (!tpi_stream) { einfo (_("%P: warning: cannot create TPI stream " "in PDB file: %E\n")); goto end; } dbi_stream = add_stream (pdb, NULL, NULL); if (!dbi_stream) { einfo (_("%P: warning: cannot create DBI stream " "in PDB file: %E\n")); goto end; } ipi_stream = add_stream (pdb, NULL, NULL); if (!ipi_stream) { einfo (_("%P: warning: cannot create IPI stream " "in PDB file: %E\n")); goto end; } names_stream = add_stream (pdb, "/names", NULL); if (!names_stream) { einfo (_("%P: warning: cannot create /names stream " "in PDB file: %E\n")); goto end; } sym_rec_stream = add_stream (pdb, NULL, &sym_rec_stream_num); if (!sym_rec_stream) { einfo (_("%P: warning: cannot create symbol record stream " "in PDB file: %E\n")); goto end; } publics_stream = add_stream (pdb, NULL, &publics_stream_num); if (!publics_stream) { einfo (_("%P: warning: cannot create publics stream " "in PDB file: %E\n")); goto end; } if (!create_section_header_stream (pdb, abfd, §ion_header_stream_num)) { einfo (_("%P: warning: cannot create section header stream " "in PDB file: %E\n")); goto end; } types.num_types = 0; types.hashmap = htab_create_alloc (0, hash_type_entry, eq_type_entry, free, xcalloc, free); types.first = types.last = NULL; ids.num_types = 0; ids.hashmap = htab_create_alloc (0, hash_type_entry, eq_type_entry, free, xcalloc, free); ids.first = ids.last = NULL; if (!populate_dbi_stream (dbi_stream, abfd, pdb, section_header_stream_num, sym_rec_stream_num, publics_stream_num, &strings, &types, &ids, sym_rec_stream, pdb_name)) { einfo (_("%P: warning: cannot populate DBI stream " "in PDB file: %E\n")); htab_delete (types.hashmap); htab_delete (ids.hashmap); goto end; } if (!populate_type_stream (pdb, tpi_stream, &types)) { einfo (_("%P: warning: cannot populate TPI stream " "in PDB file: %E\n")); htab_delete (types.hashmap); htab_delete (ids.hashmap); goto end; } htab_delete (types.hashmap); if (!populate_type_stream (pdb, ipi_stream, &ids)) { einfo (_("%P: warning: cannot populate IPI stream " "in PDB file: %E\n")); htab_delete (ids.hashmap); goto end; } htab_delete (ids.hashmap); add_string ("", 0, &strings); if (!populate_names_stream (names_stream, &strings)) { einfo (_("%P: warning: cannot populate names stream " "in PDB file: %E\n")); goto end; } if (!populate_publics_stream (publics_stream, abfd, sym_rec_stream)) { einfo (_("%P: warning: cannot populate publics stream " "in PDB file: %E\n")); goto end; } if (!populate_info_stream (pdb, info_stream, guid)) { einfo (_("%P: warning: cannot populate info stream " "in PDB file: %E\n")); goto end; } ret = true; end: bfd_close (pdb); htab_delete (strings.hashmap); return ret; }