1 /*

     2  * Copyright (c) 2003, 2013, Oracle and/or its affiliates. All rights reserved.

     3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.

     4  *

     5  * This code is free software; you can redistribute it and/or modify it

     6  * under the terms of the GNU General Public License version 2 only, as

     7  * published by the Free Software Foundation.

     8  *

     9  * This code is distributed in the hope that it will be useful, but WITHOUT

    10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or

    11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License

    12  * version 2 for more details (a copy is included in the LICENSE file that

    13  * accompanied this code).

    14  *

    15  * You should have received a copy of the GNU General Public License version

    16  * 2 along with this work; if not, write to the Free Software Foundation,

    17  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.

    18  *

    19  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA

    20  * or visit www.oracle.com if you need additional information or have any

    21  * questions.

    22  *

    23  */

    24 

    25 #include <unistd.h>

    26 #include <search.h>

    27 #include <stdlib.h>

    28 #include <string.h>

    29 #include <db.h>

    30 #include <fcntl.h>

    31 

    32 #include "libproc_impl.h"

    33 #include "symtab.h"

    34 #ifndef __APPLE__

    35 #include "salibelf.h"

    36 #endif // __APPLE__

    37 

    38 

    39 // ----------------------------------------------------

    40 // functions for symbol lookups

    41 // ----------------------------------------------------

    42 

    43 typedef struct symtab_symbol {

    44   char *name;                // name like __ZThread_...

    45   uintptr_t offset;          // to loaded address

    46   uintptr_t size;            // size strlen

    47 } symtab_symbol;

    48 

    49 typedef struct symtab {

    50   char *strs;                // all symbols "__symbol1__'\0'__symbol2__...."

    51   size_t num_symbols;

    52   DB* hash_table;

    53   symtab_symbol* symbols;

    54 } symtab_t;

    55 

    56 #ifdef __APPLE__

    57 

    58 void build_search_table(symtab_t *symtab) {

    59   int i;

    60   for (i = 0; i < symtab->num_symbols; i++) {

    61     DBT key, value;

    62     key.data = symtab->symbols[i].name;

    63     key.size = strlen(key.data) + 1;

    64     value.data = &(symtab->symbols[i]);

    65     value.size = sizeof(symtab_symbol);

    66     (*symtab->hash_table->put)(symtab->hash_table, &key, &value, 0);

    67 

    68     // check result

    69     if (is_debug()) {

    70       DBT rkey, rvalue;

    71       char* tmp = (char *)malloc(strlen(symtab->symbols[i].name) + 1);

    72       strcpy(tmp, symtab->symbols[i].name);

    73       rkey.data = tmp;

    74       rkey.size = strlen(tmp) + 1;

    75       (*symtab->hash_table->get)(symtab->hash_table, &rkey, &rvalue, 0);

    76       // we may get a copy back so compare contents

    77       symtab_symbol *res = (symtab_symbol *)rvalue.data;

    78       if (strcmp(res->name, symtab->symbols[i].name)  ||

    79           res->offset != symtab->symbols[i].offset    ||

    80           res->size != symtab->symbols[i].size) {

    81         print_debug("error to get hash_table value!\n");

    82       }

    83       free(tmp);

    84     }

    85   }

    86 }

    87 

    88 // read symbol table from given fd.

    89 struct symtab* build_symtab(int fd) {

    90   symtab_t* symtab = NULL;

    91   int i;

    92   mach_header_64 header;

    93   off_t image_start;

    94 

    95   if (!get_arch_off(fd, CPU_TYPE_X86_64, &image_start)) {

    96     print_debug("failed in get fat header\n");

    97     return NULL;

    98   }

    99   lseek(fd, image_start, SEEK_SET);

   100   if (read(fd, (void *)&header, sizeof(mach_header_64)) != sizeof(mach_header_64)) {

   101     print_debug("reading header failed!\n");

   102     return NULL;

   103   }

   104   // header

   105   if (header.magic != MH_MAGIC_64) {

   106     print_debug("not a valid .dylib file\n");

   107     return NULL;

   108   }

   109 

   110   load_command lcmd;

   111   symtab_command symtabcmd;

   112   nlist_64 lentry;

   113 

   114   bool lcsymtab_exist = false;

   115 

   116   long filepos = ltell(fd);

   117   for (i = 0; i < header.ncmds; i++) {

   118     lseek(fd, filepos, SEEK_SET);

   119     if (read(fd, (void *)&lcmd, sizeof(load_command)) != sizeof(load_command)) {

   120       print_debug("read load_command failed for file\n");

   121       return NULL;

   122     }

   123     filepos += lcmd.cmdsize;  // next command position

   124     if (lcmd.cmd == LC_SYMTAB) {

   125       lseek(fd, -sizeof(load_command), SEEK_CUR);

   126       lcsymtab_exist = true;

   127       break;

   128     }

   129   }

   130   if (!lcsymtab_exist) {

   131     print_debug("No symtab command found!\n");

   132     return NULL;

   133   }

   134   if (read(fd, (void *)&symtabcmd, sizeof(symtab_command)) != sizeof(symtab_command)) {

   135     print_debug("read symtab_command failed for file");

   136     return NULL;

   137   }

   138   symtab = (symtab_t *)malloc(sizeof(symtab_t));

   139   if (symtab == NULL) {

   140     print_debug("out of memory: allocating symtab\n");

   141     return NULL;

   142   }

   143 

   144   // create hash table, we use berkeley db to

   145   // manipulate the hash table.

   146   symtab->hash_table = dbopen(NULL, O_CREAT | O_RDWR, 0600, DB_HASH, NULL);

   147   if (symtab->hash_table == NULL)

   148     goto quit;

   149 

   150   symtab->num_symbols = symtabcmd.nsyms;

   151   symtab->symbols = (symtab_symbol *)malloc(sizeof(symtab_symbol) * symtab->num_symbols);

   152   symtab->strs    = (char *)malloc(sizeof(char) * symtabcmd.strsize);

   153   if (symtab->symbols == NULL || symtab->strs == NULL) {

   154      print_debug("out of memory: allocating symtab.symbol or symtab.strs\n");

   155      goto quit;

   156   }

   157   lseek(fd, image_start + symtabcmd.symoff, SEEK_SET);

   158   for (i = 0; i < symtab->num_symbols; i++) {

   159     if (read(fd, (void *)&lentry, sizeof(nlist_64)) != sizeof(nlist_64)) {

   160       print_debug("read nlist_64 failed at %i\n", i);

   161       goto quit;

   162     }

   163     symtab->symbols[i].offset = lentry.n_value;

   164     symtab->symbols[i].size  = lentry.n_un.n_strx;        // index

   165   }

   166 

   167   // string table

   168   lseek(fd, image_start + symtabcmd.stroff, SEEK_SET);

   169   int size = read(fd, (void *)(symtab->strs), symtabcmd.strsize * sizeof(char));

   170   if (size != symtabcmd.strsize * sizeof(char)) {

   171      print_debug("reading string table failed\n");

   172      goto quit;

   173   }

   174 

   175   for (i = 0; i < symtab->num_symbols; i++) {

   176     symtab->symbols[i].name = symtab->strs + symtab->symbols[i].size;

   177     if (i > 0) {

   178       // fix size

   179       symtab->symbols[i - 1].size = symtab->symbols[i].size - symtab->symbols[i - 1].size;

   180       print_debug("%s size = %d\n", symtab->symbols[i - 1].name, symtab->symbols[i - 1].size);

   181 

   182     }

   183 

   184     if (i == symtab->num_symbols - 1) {

   185       // last index

   186       symtab->symbols[i].size =

   187             symtabcmd.strsize - symtab->symbols[i].size;

   188       print_debug("%s size = %d\n", symtab->symbols[i].name, symtab->symbols[i].size);

   189     }

   190   }

   191 

   192   // build a hashtable for fast query

   193   build_search_table(symtab);

   194   return symtab;

   195 quit:

   196   if (symtab) destroy_symtab(symtab);

   197   return NULL;

   198 }

   199 

   200 #else // __APPLE__

   201 

   202 struct elf_section {

   203   ELF_SHDR   *c_shdr;

   204   void       *c_data;

   205 };

   206 

   207 // read symbol table from given fd.

   208 struct symtab* build_symtab(int fd) {

   209   ELF_EHDR ehdr;

   210   struct symtab* symtab = NULL;

   211 

   212   // Reading of elf header

   213   struct elf_section *scn_cache = NULL;

   214   int cnt = 0;

   215   ELF_SHDR* shbuf = NULL;

   216   ELF_SHDR* cursct = NULL;

   217   ELF_PHDR* phbuf = NULL;

   218   int symtab_found = 0;

   219   int dynsym_found = 0;

   220   uint32_t symsection = SHT_SYMTAB;

   221 

   222   uintptr_t baseaddr = (uintptr_t)-1;

   223 

   224   lseek(fd, (off_t)0L, SEEK_SET);

   225   if (! read_elf_header(fd, &ehdr)) {

   226     // not an elf

   227     return NULL;

   228   }

   229 

   230   // read ELF header

   231   if ((shbuf = read_section_header_table(fd, &ehdr)) == NULL) {

   232     goto quit;

   233   }

   234 

   235   baseaddr = find_base_address(fd, &ehdr);

   236 

   237   scn_cache = calloc(ehdr.e_shnum, sizeof(*scn_cache));

   238   if (scn_cache == NULL) {

   239     goto quit;

   240   }

   241 

   242   for (cursct = shbuf, cnt = 0; cnt < ehdr.e_shnum; cnt++) {

   243     scn_cache[cnt].c_shdr = cursct;

   244     if (cursct->sh_type == SHT_SYMTAB ||

   245         cursct->sh_type == SHT_STRTAB ||

   246         cursct->sh_type == SHT_DYNSYM) {

   247       if ( (scn_cache[cnt].c_data = read_section_data(fd, &ehdr, cursct)) == NULL) {

   248          goto quit;

   249       }

   250     }

   251 

   252     if (cursct->sh_type == SHT_SYMTAB)

   253        symtab_found++;

   254 

   255     if (cursct->sh_type == SHT_DYNSYM)

   256        dynsym_found++;

   257 

   258     cursct++;

   259   }

   260 

   261   if (!symtab_found && dynsym_found)

   262      symsection = SHT_DYNSYM;

   263 

   264   for (cnt = 1; cnt < ehdr.e_shnum; cnt++) {

   265     ELF_SHDR *shdr = scn_cache[cnt].c_shdr;

   266 

   267     if (shdr->sh_type == symsection) {

   268       ELF_SYM  *syms;

   269       int j, n;

   270       size_t size;

   271 

   272       // FIXME: there could be multiple data buffers associated with the

   273       // same ELF section. Here we can handle only one buffer. See man page

   274       // for elf_getdata on Solaris.

   275 

   276       // guarantee(symtab == NULL, "multiple symtab");

   277       symtab = calloc(1, sizeof(*symtab));

   278       if (symtab == NULL) {

   279          goto quit;

   280       }

   281       // the symbol table

   282       syms = (ELF_SYM *)scn_cache[cnt].c_data;

   283 

   284       // number of symbols

   285       n = shdr->sh_size / shdr->sh_entsize;

   286 

   287       // create hash table, we use berkeley db to

   288       // manipulate the hash table.

   289       symtab->hash_table = dbopen(NULL, O_CREAT | O_RDWR, 0600, DB_HASH, NULL);

   290       // guarantee(symtab->hash_table, "unexpected failure: dbopen");

   291       if (symtab->hash_table == NULL)

   292         goto bad;

   293 

   294       // shdr->sh_link points to the section that contains the actual strings

   295       // for symbol names. the st_name field in ELF_SYM is just the

   296       // string table index. we make a copy of the string table so the

   297       // strings will not be destroyed by elf_end.

   298       size = scn_cache[shdr->sh_link].c_shdr->sh_size;

   299       symtab->strs = malloc(size);

   300       if (symtab->strs == NULL)

   301         goto bad;

   302       memcpy(symtab->strs, scn_cache[shdr->sh_link].c_data, size);

   303 

   304       // allocate memory for storing symbol offset and size;

   305       symtab->num_symbols = n;

   306       symtab->symbols = calloc(n , sizeof(*symtab->symbols));

   307       if (symtab->symbols == NULL)

   308         goto bad;

   309 

   310       // copy symbols info our symtab and enter them info the hash table

   311       for (j = 0; j < n; j++, syms++) {

   312         DBT key, value;

   313         char *sym_name = symtab->strs + syms->st_name;

   314 

   315         // skip non-object and non-function symbols

   316         int st_type = ELF_ST_TYPE(syms->st_info);

   317         if ( st_type != STT_FUNC && st_type != STT_OBJECT)

   318            continue;

   319         // skip empty strings and undefined symbols

   320         if (*sym_name == '\0' || syms->st_shndx == SHN_UNDEF) continue;

   321 

   322         symtab->symbols[j].name   = sym_name;

   323         symtab->symbols[j].offset = syms->st_value - baseaddr;

   324         symtab->symbols[j].size   = syms->st_size;

   325 

   326         key.data = sym_name;

   327         key.size = strlen(sym_name) + 1;

   328         value.data = &(symtab->symbols[j]);

   329         value.size = sizeof(symtab_symbol);

   330         (*symtab->hash_table->put)(symtab->hash_table, &key, &value, 0);

   331       }

   332     }

   333   }

   334   goto quit;

   335 

   336 bad:

   337   destroy_symtab(symtab);

   338   symtab = NULL;

   339 

   340 quit:

   341   if (shbuf) free(shbuf);

   342   if (phbuf) free(phbuf);

   343   if (scn_cache) {

   344     for (cnt = 0; cnt < ehdr.e_shnum; cnt++) {

   345       if (scn_cache[cnt].c_data != NULL) {

   346         free(scn_cache[cnt].c_data);

   347       }

   348     }

   349     free(scn_cache);

   350   }

   351   return symtab;

   352 }

   353 

   354 #endif // __APPLE__

   355 

   356 void destroy_symtab(symtab_t* symtab) {

   357   if (!symtab) return;

   358   free(symtab->strs);

   359   free(symtab->symbols);

   360   free(symtab);

   361 }

   362 

   363 uintptr_t search_symbol(struct symtab* symtab, uintptr_t base, const char *sym_name, int *sym_size) {

   364   DBT key, value;

   365   int ret;

   366 

   367   // library does not have symbol table

   368   if (!symtab || !symtab->hash_table) {

   369      return 0;

   370   }

   371 

   372   key.data = (char*)(uintptr_t)sym_name;

   373   key.size = strlen(sym_name) + 1;

   374   ret = (*symtab->hash_table->get)(symtab->hash_table, &key, &value, 0);

   375   if (ret == 0) {

   376     symtab_symbol *sym = value.data;

   377     uintptr_t rslt = (uintptr_t) ((char*)base + sym->offset);

   378     if (sym_size) *sym_size = sym->size;

   379     return rslt;

   380   }

   381 

   382   return 0;

   383 }

   384 

   385 const char* nearest_symbol(struct symtab* symtab, uintptr_t offset,

   386                            uintptr_t* poffset) {

   387   int n = 0;

   388   if (!symtab) return NULL;

   389   for (; n < symtab->num_symbols; n++) {

   390     symtab_symbol* sym = &(symtab->symbols[n]);

   391     if (sym->name != NULL &&

   392       offset >= sym->offset && offset < sym->offset + sym->size) {

   393       if (poffset) *poffset = (offset - sym->offset);

   394       return sym->name;

   395     }

   396   }

   397   return NULL;

   398 }