EXTERNAL_HEADERS/mach-o/loader.h

   1 /*
   2  * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_LICENSE_HEADER_START@
   5  *
   6  * The contents of this file constitute Original Code as defined in and
   7  * are subject to the Apple Public Source License Version 1.1 (the
   8  * "License").  You may not use this file except in compliance with the
   9  * License.  Please obtain a copy of the License at
  10  * http://www.apple.com/publicsource and read it before using this file.
  11  *
  12  * This Original Code and all software distributed under the License are
  13  * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  14  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  15  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  16  * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT.  Please see the
  17  * License for the specific language governing rights and limitations
  18  * under the License.
  19  *
  20  * @APPLE_LICENSE_HEADER_END@
  21  */
  22 #ifndef _MACHO_LOADER_H_
  23 #define _MACHO_LOADER_H_
  24
  25 /*
  26  * This file describes the format of mach object files.
  27  */
  28
  29 /*
  30  * <mach/machine.h> is needed here for the cpu_type_t and cpu_subtype_t types
  31  * and contains the constants for the possible values of these types.
  32  */
  33 #include <mach/machine.h>
  34
  35 /*
  36  * <mach/vm_prot.h> is needed here for the vm_prot_t type and contains the
  37  * constants that are or'ed together for the possible values of this type.
  38  */
  39 #include <mach/vm_prot.h>
  40
  41 /*
  42  * <machine/thread_status.h> is expected to define the flavors of the thread
  43  * states and the structures of those flavors for each machine.
  44  */
  45 #include <mach/machine/thread_status.h>
  46 #include <architecture/byte_order.h>
  47
  48 /*
  49  * The mach header appears at the very beginning of the object file.
  50  */
  51 struct mach_header {
  52         unsigned long   magic;          /* mach magic number identifier */
  53         cpu_type_t      cputype;        /* cpu specifier */
  54         cpu_subtype_t   cpusubtype;     /* machine specifier */
  55         unsigned long   filetype;       /* type of file */
  56         unsigned long   ncmds;          /* number of load commands */
  57         unsigned long   sizeofcmds;     /* the size of all the load commands */
  58         unsigned long   flags;          /* flags */
  59 };
  60
  61 /* Constant for the magic field of the mach_header */
  62 #define MH_MAGIC        0xfeedface      /* the mach magic number */
  63 #define MH_CIGAM        NXSwapInt(MH_MAGIC)
  64
  65 /*
  66  * The layout of the file depends on the filetype.  For all but the MH_OBJECT
  67  * file type the segments are padded out and aligned on a segment alignment
  68  * boundary for efficient demand pageing.  The MH_EXECUTE, MH_FVMLIB, MH_DYLIB,
  69  * MH_DYLINKER and MH_BUNDLE file types also have the headers included as part
  70  * of their first segment.
  71  *
  72  * The file type MH_OBJECT is a compact format intended as output of the
  73  * assembler and input (and possibly output) of the link editor (the .o
  74  * format).  All sections are in one unnamed segment with no segment padding.
  75  * This format is used as an executable format when the file is so small the
  76  * segment padding greatly increases it's size.
  77  *
  78  * The file type MH_PRELOAD is an executable format intended for things that
  79  * not executed under the kernel (proms, stand alones, kernels, etc).  The
  80  * format can be executed under the kernel but may demand paged it and not
  81  * preload it before execution.
  82  *
  83  * A core file is in MH_CORE format and can be any in an arbritray legal
  84  * Mach-O file.
  85  *
  86  * Constants for the filetype field of the mach_header
  87  */
  88 #define MH_OBJECT       0x1             /* relocatable object file */
  89 #define MH_EXECUTE      0x2             /* demand paged executable file */
  90 #define MH_FVMLIB       0x3             /* fixed VM shared library file */
  91 #define MH_CORE         0x4             /* core file */
  92 #define MH_PRELOAD      0x5             /* preloaded executable file */
  93 #define MH_DYLIB        0x6             /* dynamicly bound shared library file*/
  94 #define MH_DYLINKER     0x7             /* dynamic link editor */
  95 #define MH_BUNDLE       0x8             /* dynamicly bound bundle file */
  96
  97 /* Constants for the flags field of the mach_header */
  98 #define MH_NOUNDEFS     0x1             /* the object file has no undefined
  99                                            references, can be executed */
 100 #define MH_INCRLINK     0x2             /* the object file is the output of an
 101                                            incremental link against a base file
 102                                            and can't be link edited again */
 103 #define MH_DYLDLINK     0x4             /* the object file is input for the
 104                                            dynamic linker and can't be staticly
 105                                            link edited again */
 106 #define MH_BINDATLOAD   0x8             /* the object file's undefined
 107                                            references are bound by the dynamic
 108                                            linker when loaded. */
 109 #define MH_PREBOUND     0x10            /* the file has it's dynamic undefined
 110                                            references prebound. */
 111
 112 /*
 113  * The load commands directly follow the mach_header.  The total size of all
 114  * of the commands is given by the sizeofcmds field in the mach_header.  All
 115  * load commands must have as their first two fields cmd and cmdsize.  The cmd
 116  * field is filled in with a constant for that command type.  Each command type
 117  * has a structure specifically for it.  The cmdsize field is the size in bytes
 118  * of the particular load command structure plus anything that follows it that
 119  * is a part of the load command (i.e. section structures, strings, etc.).  To
 120  * advance to the next load command the cmdsize can be added to the offset or
 121  * pointer of the current load command.  The cmdsize MUST be a multiple of
 122  * sizeof(long) (this is forever the maximum alignment of any load commands).
 123  * The padded bytes must be zero.  All tables in the object file must also
 124  * follow these rules so the file can be memory mapped.  Otherwise the pointers
 125  * to these tables will not work well or at all on some machines.  With all
 126  * padding zeroed like objects will compare byte for byte.
 127  */
 128 struct load_command {
 129         unsigned long cmd;              /* type of load command */
 130         unsigned long cmdsize;          /* total size of command in bytes */
 131 };
 132
 133 /* Constants for the cmd field of all load commands, the type */
 134 #define LC_SEGMENT      0x1     /* segment of this file to be mapped */
 135 #define LC_SYMTAB       0x2     /* link-edit stab symbol table info */
 136 #define LC_SYMSEG       0x3     /* link-edit gdb symbol table info (obsolete) */
 137 #define LC_THREAD       0x4     /* thread */
 138 #define LC_UNIXTHREAD   0x5     /* unix thread (includes a stack) */
 139 #define LC_LOADFVMLIB   0x6     /* load a specified fixed VM shared library */
 140 #define LC_IDFVMLIB     0x7     /* fixed VM shared library identification */
 141 #define LC_IDENT        0x8     /* object identification info (obsolete) */
 142 #define LC_FVMFILE      0x9     /* fixed VM file inclusion (internal use) */
 143 #define LC_PREPAGE      0xa     /* prepage command (internal use) */
 144 #define LC_DYSYMTAB     0xb     /* dynamic link-edit symbol table info */
 145 #define LC_LOAD_DYLIB   0xc     /* load a dynamicly linked shared library */
 146 #define LC_ID_DYLIB     0xd     /* dynamicly linked shared lib identification */
 147 #define LC_LOAD_DYLINKER 0xe    /* load a dynamic linker */
 148 #define LC_ID_DYLINKER  0xf     /* dynamic linker identification */
 149 #define LC_PREBOUND_DYLIB 0x10  /* modules prebound for a dynamicly */
 150                                 /*  linked shared library */
 151
 152 /*
 153  * A variable length string in a load command is represented by an lc_str
 154  * union.  The strings are stored just after the load command structure and
 155  * the offset is from the start of the load command structure.  The size
 156  * of the string is reflected in the cmdsize field of the load command.
 157  * Once again any padded bytes to bring the cmdsize field to a multiple
 158  * of sizeof(long) must be zero.
 159  */
 160 union lc_str {
 161         unsigned long   offset; /* offset to the string */
 162         char            *ptr;   /* pointer to the string */
 163 };
 164
 165 /*
 166  * The segment load command indicates that a part of this file is to be
 167  * mapped into the task's address space.  The size of this segment in memory,
 168  * vmsize, maybe equal to or larger than the amount to map from this file,
 169  * filesize.  The file is mapped starting at fileoff to the beginning of
 170  * the segment in memory, vmaddr.  The rest of the memory of the segment,
 171  * if any, is allocated zero fill on demand.  The segment's maximum virtual
 172  * memory protection and initial virtual memory protection are specified
 173  * by the maxprot and initprot fields.  If the segment has sections then the
 174  * section structures directly follow the segment command and their size is
 175  * reflected in cmdsize.
 176  */
 177 struct segment_command {
 178         unsigned long   cmd;            /* LC_SEGMENT */
 179         unsigned long   cmdsize;        /* includes sizeof section structs */
 180         char            segname[16];    /* segment name */
 181         unsigned long   vmaddr;         /* memory address of this segment */
 182         unsigned long   vmsize;         /* memory size of this segment */
 183         unsigned long   fileoff;        /* file offset of this segment */
 184         unsigned long   filesize;       /* amount to map from the file */
 185         vm_prot_t       maxprot;        /* maximum VM protection */
 186         vm_prot_t       initprot;       /* initial VM protection */
 187         unsigned long   nsects;         /* number of sections in segment */
 188         unsigned long   flags;          /* flags */
 189 };
 190
 191 /* Constants for the flags field of the segment_command */
 192 #define SG_HIGHVM       0x1     /* the file contents for this segment is for
 193                                    the high part of the VM space, the low part
 194                                    is zero filled (for stacks in core files) */
 195 #define SG_FVMLIB       0x2     /* this segment is the VM that is allocated by
 196                                    a fixed VM library, for overlap checking in
 197                                    the link editor */
 198 #define SG_NORELOC      0x4     /* this segment has nothing that was relocated
 199                                    in it and nothing relocated to it, that is
 200                                    it maybe safely replaced without relocation*/
 201
 202 /*
 203  * A segment is made up of zero or more sections.  Non-MH_OBJECT files have
 204  * all of their segments with the proper sections in each, and padded to the
 205  * specified segment alignment when produced by the link editor.  The first
 206  * segment of a MH_EXECUTE and MH_FVMLIB format file contains the mach_header
 207  * and load commands of the object file before it's first section.  The zero
 208  * fill sections are always last in their segment (in all formats).  This
 209  * allows the zeroed segment padding to be mapped into memory where zero fill
 210  * sections might be.
 211  *
 212  * The MH_OBJECT format has all of it's sections in one segment for
 213  * compactness.  There is no padding to a specified segment boundary and the
 214  * mach_header and load commands are not part of the segment.
 215  *
 216  * Sections with the same section name, sectname, going into the same segment,
 217  * segname, are combined by the link editor.  The resulting section is aligned
 218  * to the maximum alignment of the combined sections and is the new section's
 219  * alignment.  The combined sections are aligned to their original alignment in
 220  * the combined section.  Any padded bytes to get the specified alignment are
 221  * zeroed.
 222  *
 223  * The format of the relocation entries referenced by the reloff and nreloc
 224  * fields of the section structure for mach object files is described in the
 225  * header file <reloc.h>.
 226  */
 227 struct section {
 228         char            sectname[16];   /* name of this section */
 229         char            segname[16];    /* segment this section goes in */
 230         unsigned long   addr;           /* memory address of this section */
 231         unsigned long   size;           /* size in bytes of this section */
 232         unsigned long   offset;         /* file offset of this section */
 233         unsigned long   align;          /* section alignment (power of 2) */
 234         unsigned long   reloff;         /* file offset of relocation entries */
 235         unsigned long   nreloc;         /* number of relocation entries */
 236         unsigned long   flags;          /* flags (section type and attributes)*/
 237         unsigned long   reserved1;      /* reserved */
 238         unsigned long   reserved2;      /* reserved */
 239 };
 240
 241 /*
 242  * The flags field of a section structure is separated into two parts a section
 243  * type and section attributes.  The section types are mutually exclusive (it
 244  * can only have one type) but the section attributes are not (it may have more
 245  * than one attribute).
 246  */
 247 #define SECTION_TYPE             0x000000ff     /* 256 section types */
 248 #define SECTION_ATTRIBUTES       0xffffff00     /*  24 section attributes */
 249
 250 /* Constants for the type of a section */
 251 #define S_REGULAR               0x0     /* regular section */
 252 #define S_ZEROFILL              0x1     /* zero fill on demand section */
 253 #define S_CSTRING_LITERALS      0x2     /* section with only literal C strings*/
 254 #define S_4BYTE_LITERALS        0x3     /* section with only 4 byte literals */
 255 #define S_8BYTE_LITERALS        0x4     /* section with only 8 byte literals */
 256 #define S_LITERAL_POINTERS      0x5     /* section with only pointers to */
 257                                         /*  literals */
 258 /*
 259  * For the two types of symbol pointers sections and the symbol stubs section
 260  * they have indirect symbol table entries.  For each of the entries in the
 261  * section the indirect symbol table entries, in corresponding order in the
 262  * indirect symbol table, start at the index stored in the reserved1 field
 263  * of the section structure.  Since the indirect symbol table entries
 264  * correspond to the entries in the section the number of indirect symbol table
 265  * entries is inferred from the size of the section divided by the size of the
 266  * entries in the section.  For symbol pointers sections the size of the entries
 267  * in the section is 4 bytes and for symbol stubs sections the byte size of the
 268  * stubs is stored in the reserved2 field of the section structure.
 269  */
 270 #define S_NON_LAZY_SYMBOL_POINTERS      0x6     /* section with only non-lazy
 271                                                    symbol pointers */
 272 #define S_LAZY_SYMBOL_POINTERS          0x7     /* section with only lazy symbol
 273                                                    pointers */
 274 #define S_SYMBOL_STUBS                  0x8     /* section with only symbol
 275                                                    stubs, byte size of stub in
 276                                                    the reserved2 field */
 277 #define S_MOD_INIT_FUNC_POINTERS        0x9     /* section with only function
 278                                                    pointers for initialization*/
 279 /*
 280  * Constants for the section attributes part of the flags field of a section
 281  * structure.
 282  */
 283 #define SECTION_ATTRIBUTES_USR   0xff000000     /* User setable attributes */
 284 #define S_ATTR_PURE_INSTRUCTIONS 0x80000000     /* section contains only true
 285                                                    machine instructions */
 286 #define SECTION_ATTRIBUTES_SYS   0x00ffff00     /* system setable attributes */
 287 #define S_ATTR_SOME_INSTRUCTIONS 0x00000400     /* section contains some
 288                                                    machine instructions */
 289 #define S_ATTR_EXT_RELOC         0x00000200     /* section has external
 290                                                    relocation entries */
 291 #define S_ATTR_LOC_RELOC         0x00000100     /* section has local
 292                                                    relocation entries */
 293
 294
 295 /*
 296  * The names of segments and sections in them are mostly meaningless to the
 297  * link-editor.  But there are few things to support traditional UNIX
 298  * executables that require the link-editor and assembler to use some names
 299  * agreed upon by convention.
 300  *
 301  * The initial protection of the "__TEXT" segment has write protection turned
 302  * off (not writeable).
 303  *
 304  * The link-editor will allocate common symbols at the end of the "__common"
 305  * section in the "__DATA" segment.  It will create the section and segment
 306  * if needed.
 307  */
 308
 309 /* The currently known segment names and the section names in those segments */
 310
 311 #define SEG_PAGEZERO    "__PAGEZERO"    /* the pagezero segment which has no */
 312                                         /* protections and catches NULL */
 313                                         /* references for MH_EXECUTE files */
 314
 315
 316 #define SEG_TEXT        "__TEXT"        /* the tradition UNIX text segment */
 317 #define SECT_TEXT       "__text"        /* the real text part of the text */
 318                                         /* section no headers, and no padding */
 319 #define SECT_FVMLIB_INIT0 "__fvmlib_init0"      /* the fvmlib initialization */
 320                                                 /*  section */
 321 #define SECT_FVMLIB_INIT1 "__fvmlib_init1"      /* the section following the */
 322                                                 /*  fvmlib initialization */
 323                                                 /*  section */
 324
 325 #define SEG_DATA        "__DATA"        /* the tradition UNIX data segment */
 326 #define SECT_DATA       "__data"        /* the real initialized data section */
 327                                         /* no padding, no bss overlap */
 328 #define SECT_BSS        "__bss"         /* the real uninitialized data section*/
 329                                         /* no padding */
 330 #define SECT_COMMON     "__common"      /* the section common symbols are */
 331                                         /* allocated in by the link editor */
 332
 333 #define SEG_OBJC        "__OBJC"        /* objective-C runtime segment */
 334 #define SECT_OBJC_SYMBOLS "__symbol_table"      /* symbol table */
 335 #define SECT_OBJC_MODULES "__module_info"       /* module information */
 336 #define SECT_OBJC_STRINGS "__selector_strs"     /* string table */
 337 #define SECT_OBJC_REFS "__selector_refs"        /* string table */
 338
 339 #define SEG_ICON         "__ICON"       /* the NeXT icon segment */
 340 #define SECT_ICON_HEADER "__header"     /* the icon headers */
 341 #define SECT_ICON_TIFF   "__tiff"       /* the icons in tiff format */
 342
 343 #define SEG_LINKEDIT    "__LINKEDIT"    /* the segment containing all structs */
 344                                         /* created and maintained by the link */
 345                                         /* editor.  Created with -seglinkedit */
 346                                         /* option to ld(1) for MH_EXECUTE and */
 347                                         /* FVMLIB file types only */
 348
 349 #define SEG_UNIXSTACK   "__UNIXSTACK"   /* the unix stack segment */
 350
 351 /*
 352  * Fixed virtual memory shared libraries are identified by two things.  The
 353  * target pathname (the name of the library as found for execution), and the
 354  * minor version number.  The address of where the headers are loaded is in
 355  * header_addr.
 356  */
 357 struct fvmlib {
 358         union lc_str    name;           /* library's target pathname */
 359         unsigned long   minor_version;  /* library's minor version number */
 360         unsigned long   header_addr;    /* library's header address */
 361 };
 362
 363 /*
 364  * A fixed virtual shared library (filetype == MH_FVMLIB in the mach header)
 365  * contains a fvmlib_command (cmd == LC_IDFVMLIB) to identify the library.
 366  * An object that uses a fixed virtual shared library also contains a
 367  * fvmlib_command (cmd == LC_LOADFVMLIB) for each library it uses.
 368  */
 369 struct fvmlib_command {
 370         unsigned long   cmd;            /* LC_IDFVMLIB or LC_LOADFVMLIB */
 371         unsigned long   cmdsize;        /* includes pathname string */
 372         struct fvmlib   fvmlib;         /* the library identification */
 373 };
 374
 375 /*
 376  * Dynamicly linked shared libraries are identified by two things.  The
 377  * pathname (the name of the library as found for execution), and the
 378  * compatibility version number.  The pathname must match and the compatibility
 379  * number in the user of the library must be greater than or equal to the
 380  * library being used.  The time stamp is used to record the time a library was
 381  * built and copied into user so it can be use to determined if the library used
 382  * at runtime is exactly the same as used to built the program.
 383  */
 384 struct dylib {
 385     union lc_str  name;                 /* library's path name */
 386     unsigned long timestamp;            /* library's build time stamp */
 387     unsigned long current_version;      /* library's current version number */
 388     unsigned long compatibility_version;/* library's compatibility vers number*/
 389 };
 390
 391 /*
 392  * A dynamicly linked shared library (filetype == MH_DYLIB in the mach header)
 393  * contains a dylib_command (cmd == LC_ID_DYLIB) to identify the library.
 394  * An object that uses a dynamicly linked shared library also contains a
 395  * dylib_command (cmd == LC_LOAD_DYLIB) for each library it uses.
 396  */
 397 struct dylib_command {
 398         unsigned long   cmd;            /* LC_ID_DYLIB or LC_LOAD_DYLIB */
 399         unsigned long   cmdsize;        /* includes pathname string */
 400         struct dylib    dylib;          /* the library identification */
 401 };
 402
 403 /*
 404  * A program (filetype == MH_EXECUTE) or bundle (filetype == MH_BUNDLE) that is
 405  * prebound to it's dynamic libraries has one of these for each library that
 406  * the static linker used in prebinding.  It contains a bit vector for the
 407  * modules in the library.  The bits indicate which modules are bound (1) and
 408  * which are not (0) from the library.  The bit for module 0 is the low bit
 409  * of the first byte.  So the bit for the Nth module is:
 410  * (linked_modules[N/8] >> N%8) & 1
 411  */
 412 struct prebound_dylib_command {
 413         unsigned long   cmd;            /* LC_PREBOUND_DYLIB */
 414         unsigned long   cmdsize;        /* includes strings */
 415         union lc_str    name;           /* library's path name */
 416         unsigned long   nmodules;       /* number of modules in library */
 417         union lc_str    linked_modules; /* bit vector of linked modules */
 418 };
 419
 420 /*
 421  * A program that uses a dynamic linker contains a dylinker_command to identify
 422  * the name of the dynamic linker (LC_LOAD_DYLINKER).  And a dynamic linker
 423  * contains a dylinker_command to identify the dynamic linker (LC_ID_DYLINKER).
 424  * A file can have at most one of these.
 425  */
 426 struct dylinker_command {
 427         unsigned long   cmd;            /* LC_ID_DYLINKER or LC_LOAD_DYLINKER */
 428         unsigned long   cmdsize;        /* includes pathname string */
 429         union lc_str    name;           /* dynamic linker's path name */
 430 };
 431
 432 /*
 433  * Thread commands contain machine-specific data structures suitable for
 434  * use in the thread state primitives.  The machine specific data structures
 435  * follow the struct thread_command as follows.
 436  * Each flavor of machine specific data structure is preceded by an unsigned
 437  * long constant for the flavor of that data structure, an unsigned long
 438  * that is the count of longs of the size of the state data structure and then
 439  * the state data structure follows.  This triple may be repeated for many
 440  * flavors.  The constants for the flavors, counts and state data structure
 441  * definitions are expected to be in the header file <machine/thread_status.h>.
 442  * These machine specific data structures sizes must be multiples of
 443  * sizeof(long).  The cmdsize reflects the total size of the thread_command
 444  * and all of the sizes of the constants for the flavors, counts and state
 445  * data structures.
 446  *
 447  * For executable objects that are unix processes there will be one
 448  * thread_command (cmd == LC_UNIXTHREAD) created for it by the link-editor.
 449  * This is the same as a LC_THREAD, except that a stack is automatically
 450  * created (based on the shell's limit for the stack size).  Command arguments
 451  * and environment variables are copied onto that stack.
 452  */
 453 struct thread_command {
 454         unsigned long   cmd;            /* LC_THREAD or  LC_UNIXTHREAD */
 455         unsigned long   cmdsize;        /* total size of this command */
 456         /* unsigned long flavor            flavor of thread state */
 457         /* unsigned long count             count of longs in thread state */
 458         /* struct XXX_thread_state state   thread state for this flavor */
 459         /* ... */
 460 };
 461
 462 /*
 463  * The symtab_command contains the offsets and sizes of the link-edit 4.3BSD
 464  * "stab" style symbol table information as described in the header files
 465  * <nlist.h> and <stab.h>.
 466  */
 467 struct symtab_command {
 468         unsigned long   cmd;            /* LC_SYMTAB */
 469         unsigned long   cmdsize;        /* sizeof(struct symtab_command) */
 470         unsigned long   symoff;         /* symbol table offset */
 471         unsigned long   nsyms;          /* number of symbol table entries */
 472         unsigned long   stroff;         /* string table offset */
 473         unsigned long   strsize;        /* string table size in bytes */
 474 };
 475
 476 /*
 477  * This is the second set of the symbolic information which is used to support
 478  * the data structures for the dynamicly link editor.
 479  *
 480  * The original set of symbolic information in the symtab_command which contains
 481  * the symbol and string tables must also be present when this load command is
 482  * present.  When this load command is present the symbol table is organized
 483  * into three groups of symbols:
 484  *      local symbols (static and debugging symbols) - grouped by module
 485  *      defined external symbols - grouped by module (sorted by name if not lib)
 486  *      undefined external symbols (sorted by name)
 487  * In this load command there are offsets and counts to each of the three groups
 488  * of symbols.
 489  *
 490  * This load command contains a the offsets and sizes of the following new
 491  * symbolic information tables:
 492  *      table of contents
 493  *      module table
 494  *      reference symbol table
 495  *      indirect symbol table
 496  * The first three tables above (the table of contents, module table and
 497  * reference symbol table) are only present if the file is a dynamicly linked
 498  * shared library.  For executable and object modules, which are files
 499  * containing only one module, the information that would be in these three
 500  * tables is determined as follows:
 501  *      table of contents - the defined external symbols are sorted by name
 502  *      module table - the file contains only one module so everything in the
 503  *                     file is part of the module.
 504  *      reference symbol table - is the defined and undefined external symbols
 505  *
 506  * For dynamicly linked shared library files this load command also contains
 507  * offsets and sizes to the pool of relocation entries for all sections
 508  * separated into two groups:
 509  *      external relocation entries
 510  *      local relocation entries
 511  * For executable and object modules the relocation entries continue to hang
 512  * off the section structures.
 513  */
 514 struct dysymtab_command {
 515     unsigned long cmd;          /* LC_DYSYMTAB */
 516     unsigned long cmdsize;      /* sizeof(struct dysymtab_command) */
 517
 518     /*
 519      * The symbols indicated by symoff and nsyms of the LC_SYMTAB load command
 520      * are grouped into the following three groups:
 521      *    local symbols (further grouped by the module they are from)
 522      *    defined external symbols (further grouped by the module they are from)
 523      *    undefined symbols
 524      *
 525      * The local symbols are used only for debugging.  The dynamic binding
 526      * process may have to use them to indicate to the debugger the local
 527      * symbols for a module that is being bound.
 528      *
 529      * The last two groups are used by the dynamic binding process to do the
 530      * binding (indirectly through the module table and the reference symbol
 531      * table when this is a dynamicly linked shared library file).
 532      */
 533     unsigned long ilocalsym;    /* index to local symbols */
 534     unsigned long nlocalsym;    /* number of local symbols */
 535
 536     unsigned long iextdefsym;   /* index to externally defined symbols */
 537     unsigned long nextdefsym;   /* number of externally defined symbols */
 538
 539     unsigned long iundefsym;    /* index to undefined symbols */
 540     unsigned long nundefsym;    /* number of undefined symbols */
 541
 542     /*
 543      * For the for the dynamic binding process to find which module a symbol
 544      * is defined in the table of contents is used (analogous to the ranlib
 545      * structure in an archive) which maps defined external symbols to modules
 546      * they are defined in.  This exists only in a dynamicly linked shared
 547      * library file.  For executable and object modules the defined external
 548      * symbols are sorted by name and is use as the table of contents.
 549      */
 550     unsigned long tocoff;       /* file offset to table of contents */
 551     unsigned long ntoc;         /* number of entries in table of contents */
 552
 553     /*
 554      * To support dynamic binding of "modules" (whole object files) the symbol
 555      * table must reflect the modules that the file was created from.  This is
 556      * done by having a module table that has indexes and counts into the merged
 557      * tables for each module.  The module structure that these two entries
 558      * refer to is described below.  This exists only in a dynamicly linked
 559      * shared library file.  For executable and object modules the file only
 560      * contains one module so everything in the file belongs to the module.
 561      */
 562     unsigned long modtaboff;    /* file offset to module table */
 563     unsigned long nmodtab;      /* number of module table entries */
 564
 565     /*
 566      * To support dynamic module binding the module structure for each module
 567      * indicates the external references (defined and undefined) each module
 568      * makes.  For each module there is an offset and a count into the
 569      * reference symbol table for the symbols that the module references.
 570      * This exists only in a dynamicly linked shared library file.  For
 571      * executable and object modules the defined external symbols and the
 572      * undefined external symbols indicates the external references.
 573      */
 574     unsigned long extrefsymoff;  /* offset to referenced symbol table */
 575     unsigned long nextrefsyms;   /* number of referenced symbol table entries */
 576
 577     /*
 578      * The sections that contain "symbol pointers" and "routine stubs" have
 579      * indexes and (implied counts based on the size of the section and fixed
 580      * size of the entry) into the "indirect symbol" table for each pointer
 581      * and stub.  For every section of these two types the index into the
 582      * indirect symbol table is stored in the section header in the field
 583      * reserved1.  An indirect symbol table entry is simply a 32bit index into
 584      * the symbol table to the symbol that the pointer or stub is referring to.
 585      * The indirect symbol table is ordered to match the entries in the section.
 586      */
 587     unsigned long indirectsymoff; /* file offset to the indirect symbol table */
 588     unsigned long nindirectsyms;  /* number of indirect symbol table entries */
 589
 590     /*
 591      * To support relocating an individual module in a library file quickly the
 592      * external relocation entries for each module in the library need to be
 593      * accessed efficiently.  Since the relocation entries can't be accessed
 594      * through the section headers for a library file they are separated into
 595      * groups of local and external entries further grouped by module.  In this
 596      * case the presents of this load command who's extreloff, nextrel,
 597      * locreloff and nlocrel fields are non-zero indicates that the relocation
 598      * entries of non-merged sections are not referenced through the section
 599      * structures (and the reloff and nreloc fields in the section headers are
 600      * set to zero).
 601      *
 602      * Since the relocation entries are not accessed through the section headers
 603      * this requires the r_address field to be something other than a section
 604      * offset to identify the item to be relocated.  In this case r_address is
 605      * set to the offset from the vmaddr of the first LC_SEGMENT command.
 606      *
 607      * The relocation entries are grouped by module and the module table
 608      * entries have indexes and counts into them for the group of external
 609      * relocation entries for that the module.
 610      *
 611      * For sections that are merged across modules there must not be any
 612      * remaining external relocation entries for them (for merged sections
 613      * remaining relocation entries must be local).
 614      */
 615     unsigned long extreloff;    /* offset to external relocation entries */
 616     unsigned long nextrel;      /* number of external relocation entries */
 617
 618     /*
 619      * All the local relocation entries are grouped together (they are not
 620      * grouped by their module since they are only used if the object is moved
 621      * from it staticly link edited address).
 622      */
 623     unsigned long locreloff;    /* offset to local relocation entries */
 624     unsigned long nlocrel;      /* number of local relocation entries */
 625
 626 };
 627
 628 /*
 629  * An indirect symbol table entry is simply a 32bit index into the symbol table
 630  * to the symbol that the pointer or stub is refering to.  Unless it is for a
 631  * non-lazy symbol pointer section for a defined symbol which strip(1) as
 632  * removed.  In which case it has the value INDIRECT_SYMBOL_LOCAL.  If the
 633  * symbol was also absolute INDIRECT_SYMBOL_ABS is or'ed with that.
 634  */
 635 #define INDIRECT_SYMBOL_LOCAL   0x80000000
 636 #define INDIRECT_SYMBOL_ABS     0x40000000
 637
 638
 639 /* a table of contents entry */
 640 struct dylib_table_of_contents {
 641     unsigned long symbol_index; /* the defined external symbol
 642                                    (index into the symbol table) */
 643     unsigned long module_index; /* index into the module table this symbol
 644                                    is defined in */
 645 };
 646
 647 /* a module table entry */
 648 struct dylib_module {
 649     unsigned long module_name;  /* the module name (index into string table) */
 650
 651     unsigned long iextdefsym;   /* index into externally defined symbols */
 652     unsigned long nextdefsym;   /* number of externally defined symbols */
 653     unsigned long irefsym;              /* index into reference symbol table */
 654     unsigned long nrefsym;      /* number of reference symbol table entries */
 655     unsigned long ilocalsym;    /* index into symbols for local symbols */
 656     unsigned long nlocalsym;    /* number of local symbols */
 657
 658     unsigned long iextrel;      /* index into external relocation entries */
 659     unsigned long nextrel;      /* number of external relocation entries */
 660
 661     unsigned long iinit;        /* index into the init section */
 662     unsigned long ninit;        /* number of init section entries */
 663
 664     unsigned long               /* for this module address of the start of */
 665         objc_module_info_addr;  /*  the (__OBJC,__module_info) section */
 666     unsigned long               /* for this module size of */
 667         objc_module_info_size;  /*  the (__OBJC,__module_info) section */
 668 };
 669
 670 /*
 671  * The entries in the reference symbol table are used when loading the module
 672  * (both by the static and dynamic link editors) and if the module is unloaded
 673  * or replaced.  Therefore all external symbols (defined and undefined) are
 674  * listed in the module's reference table.  The flags describe the type of
 675  * reference that is being made.  The constants for the flags are defined in
 676  * <mach-o/nlist.h> as they are also used for symbol table entries.
 677  */
 678 struct dylib_reference {
 679     unsigned long isym:24,      /* index into the symbol table */
 680                   flags:8;      /* flags to indicate the type of reference */
 681 };
 682
 683 /*
 684  * The symseg_command contains the offset and size of the GNU style
 685  * symbol table information as described in the header file <symseg.h>.
 686  * The symbol roots of the symbol segments must also be aligned properly
 687  * in the file.  So the requirement of keeping the offsets aligned to a
 688  * multiple of a sizeof(long) translates to the length field of the symbol
 689  * roots also being a multiple of a long.  Also the padding must again be
 690  * zeroed. (THIS IS OBSOLETE and no longer supported).
 691  */
 692 struct symseg_command {
 693         unsigned long   cmd;            /* LC_SYMSEG */
 694         unsigned long   cmdsize;        /* sizeof(struct symseg_command) */
 695         unsigned long   offset;         /* symbol segment offset */
 696         unsigned long   size;           /* symbol segment size in bytes */
 697 };
 698
 699 /*
 700  * The ident_command contains a free format string table following the
 701  * ident_command structure.  The strings are null terminated and the size of
 702  * the command is padded out with zero bytes to a multiple of sizeof(long).
 703  * (THIS IS OBSOLETE and no longer supported).
 704  */
 705 struct ident_command {
 706         unsigned long cmd;              /* LC_IDENT */
 707         unsigned long cmdsize;          /* strings that follow this command */
 708 };
 709
 710 /*
 711  * The fvmfile_command contains a reference to a file to be loaded at the
 712  * specified virtual address.  (Presently, this command is reserved for NeXT
 713  * internal use.  The kernel ignores this command when loading a program into
 714  * memory).
 715  */
 716 struct fvmfile_command {
 717         unsigned long cmd;              /* LC_FVMFILE */
 718         unsigned long cmdsize;          /* includes pathname string */
 719         union lc_str    name;           /* files pathname */
 720         unsigned long   header_addr;    /* files virtual address */
 721 };
 722
 723 #endif /* _MACHO_LOADER_H_ */