osfmk/mach-o/loader.h

   1 /*
   2  * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
   5  *
   6  * This file contains Original Code and/or Modifications of Original Code
   7  * as defined in and that are subject to the Apple Public Source License
   8  * Version 2.0 (the 'License'). You may not use this file except in
   9  * compliance with the License. The rights granted to you under the License
  10  * may not be used to create, or enable the creation or redistribution of,
  11  * unlawful or unlicensed copies of an Apple operating system, or to
  12  * circumvent, violate, or enable the circumvention or violation of, any
  13  * terms of an Apple operating system software license agreement.
  14  *
  15  * Please obtain a copy of the License at
  16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
  17  *
  18  * The Original Code and all software distributed under the License are
  19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  23  * Please see the License for the specific language governing rights and
  24  * limitations under the License.
  25  *
  26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  27  */
  28 #ifndef _MACHO_LOADER_H_
  29 #define _MACHO_LOADER_H_
  30
  31 /*
  32  * This file describes the format of mach object files.
  33  *
  34  * NOTE:        This header is used for manipulationg 32 bit mach object
  35  *              withing a 32 bit mach_kernel for the purpose of dealing
  36  *              with linking loadable kernel modules.
  37  */
  38
  39 /*
  40  * <mach/machine.h> is needed here for the cpu_type_t and cpu_subtype_t types
  41  * and contains the constants for the possible values of these types.
  42  */
  43 #include <mach/machine.h>
  44
  45 /*
  46  * <mach/vm_prot.h> is needed here for the vm_prot_t type and contains the
  47  * constants that are or'ed together for the possible values of this type.
  48  */
  49 #include <mach/vm_prot.h>
  50
  51 /*
  52  * <machine/thread_status.h> is expected to define the flavors of the thread
  53  * states and the structures of those flavors for each machine.
  54  */
  55 #include <mach/machine/thread_status.h>
  56
  57 /*
  58  * The mach header appears at the very beginning of the object file.
  59  */
  60 struct mach_header {
  61         unsigned long   magic;          /* mach magic number identifier */
  62         cpu_type_t      cputype;        /* cpu specifier */
  63         cpu_subtype_t   cpusubtype;     /* machine specifier */
  64         unsigned long   filetype;       /* type of file */
  65         unsigned long   ncmds;          /* number of load commands */
  66         unsigned long   sizeofcmds;     /* the size of all the load commands */
  67         unsigned long   flags;          /* flags */
  68 };
  69
  70 /* Constant for the magic field of the mach_header */
  71 #define MH_MAGIC        0xfeedface      /* the mach magic number */
  72 #define MH_CIGAM        0xcefaedfe
  73
  74 /*
  75  * The layout of the file depends on the filetype.  For all but the MH_OBJECT
  76  * file type the segments are padded out and aligned on a segment alignment
  77  * boundary for efficient demand pageing.  The MH_EXECUTE, MH_FVMLIB, MH_DYLIB,
  78  * MH_DYLINKER and MH_BUNDLE file types also have the headers included as part
  79  * of their first segment.
  80  *
  81  * The file type MH_OBJECT is a compact format intended as output of the
  82  * assembler and input (and possibly output) of the link editor (the .o
  83  * format).  All sections are in one unnamed segment with no segment padding.
  84  * This format is used as an executable format when the file is so small the
  85  * segment padding greatly increases it's size.
  86  *
  87  * The file type MH_PRELOAD is an executable format intended for things that
  88  * not executed under the kernel (proms, stand alones, kernels, etc).  The
  89  * format can be executed under the kernel but may demand paged it and not
  90  * preload it before execution.
  91  *
  92  * A core file is in MH_CORE format and can be any in an arbritray legal
  93  * Mach-O file.
  94  *
  95  * Constants for the filetype field of the mach_header
  96  */
  97 #define MH_OBJECT       0x1             /* relocatable object file */
  98 #define MH_EXECUTE      0x2             /* demand paged executable file */
  99 #define MH_FVMLIB       0x3             /* fixed VM shared library file */
 100 #define MH_CORE         0x4             /* core file */
 101 #define MH_PRELOAD      0x5             /* preloaded executable file */
 102 #define MH_DYLIB        0x6             /* dynamicly bound shared library file*/
 103 #define MH_DYLINKER     0x7             /* dynamic link editor */
 104 #define MH_BUNDLE       0x8             /* dynamicly bound bundle file */
 105
 106 /* Constants for the flags field of the mach_header */
 107 #define MH_NOUNDEFS     0x1             /* the object file has no undefined
 108                                            references, can be executed */
 109 #define MH_INCRLINK     0x2             /* the object file is the output of an
 110                                            incremental link against a base file
 111                                            and can't be link edited again */
 112 #define MH_DYLDLINK     0x4             /* the object file is input for the
 113                                            dynamic linker and can't be staticly
 114                                            link edited again */
 115 #define MH_BINDATLOAD   0x8             /* the object file's undefined
 116                                            references are bound by the dynamic
 117                                            linker when loaded. */
 118 #define MH_PREBOUND     0x10            /* the file has it's dynamic undefined
 119                                            references prebound. */
 120
 121 /*
 122  * The load commands directly follow the mach_header.  The total size of all
 123  * of the commands is given by the sizeofcmds field in the mach_header.  All
 124  * load commands must have as their first two fields cmd and cmdsize.  The cmd
 125  * field is filled in with a constant for that command type.  Each command type
 126  * has a structure specifically for it.  The cmdsize field is the size in bytes
 127  * of the particular load command structure plus anything that follows it that
 128  * is a part of the load command (i.e. section structures, strings, etc.).  To
 129  * advance to the next load command the cmdsize can be added to the offset or
 130  * pointer of the current load command.  The cmdsize MUST be a multiple of
 131  * sizeof(long) (this is forever the maximum alignment of any load commands).
 132  * The padded bytes must be zero.  All tables in the object file must also
 133  * follow these rules so the file can be memory mapped.  Otherwise the pointers
 134  * to these tables will not work well or at all on some machines.  With all
 135  * padding zeroed like objects will compare byte for byte.
 136  */
 137 struct load_command {
 138         unsigned long cmd;              /* type of load command */
 139         unsigned long cmdsize;          /* total size of command in bytes */
 140 };
 141
 142 /* Constants for the cmd field of all load commands, the type */
 143 #define LC_SEGMENT      0x1     /* segment of this file to be mapped */
 144 #define LC_SYMTAB       0x2     /* link-edit stab symbol table info */
 145 #define LC_SYMSEG       0x3     /* link-edit gdb symbol table info (obsolete) */
 146 #define LC_THREAD       0x4     /* thread */
 147 #define LC_UNIXTHREAD   0x5     /* unix thread (includes a stack) */
 148 #define LC_LOADFVMLIB   0x6     /* load a specified fixed VM shared library */
 149 #define LC_IDFVMLIB     0x7     /* fixed VM shared library identification */
 150 #define LC_IDENT        0x8     /* object identification info (obsolete) */
 151 #define LC_FVMFILE      0x9     /* fixed VM file inclusion (internal use) */
 152 #define LC_PREPAGE      0xa     /* prepage command (internal use) */
 153 #define LC_DYSYMTAB     0xb     /* dynamic link-edit symbol table info */
 154 #define LC_LOAD_DYLIB   0xc     /* load a dynamicly linked shared library */
 155 #define LC_ID_DYLIB     0xd     /* dynamicly linked shared lib identification */
 156 #define LC_LOAD_DYLINKER 0xe    /* load a dynamic linker */
 157 #define LC_ID_DYLINKER  0xf     /* dynamic linker identification */
 158 #define LC_PREBOUND_DYLIB 0x10  /* modules prebound for a dynamicly */
 159                                 /*  linked shared library */
 160
 161 #define LC_UUID         0x1b    /* the uuid */
 162
 163 /*
 164  * A variable length string in a load command is represented by an lc_str
 165  * union.  The strings are stored just after the load command structure and
 166  * the offset is from the start of the load command structure.  The size
 167  * of the string is reflected in the cmdsize field of the load command.
 168  * Once again any padded bytes to bring the cmdsize field to a multiple
 169  * of sizeof(long) must be zero.
 170  */
 171 union lc_str {
 172         unsigned long   offset; /* offset to the string */
 173         char            *ptr;   /* pointer to the string */
 174 };
 175
 176 /*
 177  * The segment load command indicates that a part of this file is to be
 178  * mapped into the task's address space.  The size of this segment in memory,
 179  * vmsize, maybe equal to or larger than the amount to map from this file,
 180  * filesize.  The file is mapped starting at fileoff to the beginning of
 181  * the segment in memory, vmaddr.  The rest of the memory of the segment,
 182  * if any, is allocated zero fill on demand.  The segment's maximum virtual
 183  * memory protection and initial virtual memory protection are specified
 184  * by the maxprot and initprot fields.  If the segment has sections then the
 185  * section structures directly follow the segment command and their size is
 186  * reflected in cmdsize.
 187  */
 188 struct segment_command {
 189         unsigned long   cmd;            /* LC_SEGMENT */
 190         unsigned long   cmdsize;        /* includes sizeof section structs */
 191         char            segname[16];    /* segment name */
 192         unsigned long   vmaddr;         /* memory address of this segment */
 193         unsigned long   vmsize;         /* memory size of this segment */
 194         unsigned long   fileoff;        /* file offset of this segment */
 195         unsigned long   filesize;       /* amount to map from the file */
 196         vm_prot_t       maxprot;        /* maximum VM protection */
 197         vm_prot_t       initprot;       /* initial VM protection */
 198         unsigned long   nsects;         /* number of sections in segment */
 199         unsigned long   flags;          /* flags */
 200 };
 201
 202 /* Constants for the flags field of the segment_command */
 203 #define SG_HIGHVM       0x1     /* the file contents for this segment is for
 204                                    the high part of the VM space, the low part
 205                                    is zero filled (for stacks in core files) */
 206 #define SG_FVMLIB       0x2     /* this segment is the VM that is allocated by
 207                                    a fixed VM library, for overlap checking in
 208                                    the link editor */
 209 #define SG_NORELOC      0x4     /* this segment has nothing that was relocated
 210                                    in it and nothing relocated to it, that is
 211                                    it maybe safely replaced without relocation*/
 212
 213 /*
 214  * A segment is made up of zero or more sections.  Non-MH_OBJECT files have
 215  * all of their segments with the proper sections in each, and padded to the
 216  * specified segment alignment when produced by the link editor.  The first
 217  * segment of a MH_EXECUTE and MH_FVMLIB format file contains the mach_header
 218  * and load commands of the object file before it's first section.  The zero
 219  * fill sections are always last in their segment (in all formats).  This
 220  * allows the zeroed segment padding to be mapped into memory where zero fill
 221  * sections might be.
 222  *
 223  * The MH_OBJECT format has all of it's sections in one segment for
 224  * compactness.  There is no padding to a specified segment boundary and the
 225  * mach_header and load commands are not part of the segment.
 226  *
 227  * Sections with the same section name, sectname, going into the same segment,
 228  * segname, are combined by the link editor.  The resulting section is aligned
 229  * to the maximum alignment of the combined sections and is the new section's
 230  * alignment.  The combined sections are aligned to their original alignment in
 231  * the combined section.  Any padded bytes to get the specified alignment are
 232  * zeroed.
 233  *
 234  * The format of the relocation entries referenced by the reloff and nreloc
 235  * fields of the section structure for mach object files is described in the
 236  * header file <reloc.h>.
 237  */
 238 struct section {
 239         char            sectname[16];   /* name of this section */
 240         char            segname[16];    /* segment this section goes in */
 241         unsigned long   addr;           /* memory address of this section */
 242         unsigned long   size;           /* size in bytes of this section */
 243         unsigned long   offset;         /* file offset of this section */
 244         unsigned long   align;          /* section alignment (power of 2) */
 245         unsigned long   reloff;         /* file offset of relocation entries */
 246         unsigned long   nreloc;         /* number of relocation entries */
 247         unsigned long   flags;          /* flags (section type and attributes)*/
 248         unsigned long   reserved1;      /* reserved */
 249         unsigned long   reserved2;      /* reserved */
 250 };
 251
 252 /*
 253  * The flags field of a section structure is separated into two parts a section
 254  * type and section attributes.  The section types are mutually exclusive (it
 255  * can only have one type) but the section attributes are not (it may have more
 256  * than one attribute).
 257  */
 258 #define SECTION_TYPE             0x000000ff     /* 256 section types */
 259 #define SECTION_ATTRIBUTES       0xffffff00     /*  24 section attributes */
 260
 261 /* Constants for the type of a section */
 262 #define S_REGULAR               0x0     /* regular section */
 263 #define S_ZEROFILL              0x1     /* zero fill on demand section */
 264 #define S_CSTRING_LITERALS      0x2     /* section with only literal C strings*/
 265 #define S_4BYTE_LITERALS        0x3     /* section with only 4 byte literals */
 266 #define S_8BYTE_LITERALS        0x4     /* section with only 8 byte literals */
 267 #define S_LITERAL_POINTERS      0x5     /* section with only pointers to */
 268                                         /*  literals */
 269 /*
 270  * For the two types of symbol pointers sections and the symbol stubs section
 271  * they have indirect symbol table entries.  For each of the entries in the
 272  * section the indirect symbol table entries, in corresponding order in the
 273  * indirect symbol table, start at the index stored in the reserved1 field
 274  * of the section structure.  Since the indirect symbol table entries
 275  * correspond to the entries in the section the number of indirect symbol table
 276  * entries is inferred from the size of the section divided by the size of the
 277  * entries in the section.  For symbol pointers sections the size of the entries
 278  * in the section is 4 bytes and for symbol stubs sections the byte size of the
 279  * stubs is stored in the reserved2 field of the section structure.
 280  */
 281 #define S_NON_LAZY_SYMBOL_POINTERS      0x6     /* section with only non-lazy
 282                                                    symbol pointers */
 283 #define S_LAZY_SYMBOL_POINTERS          0x7     /* section with only lazy symbol
 284                                                    pointers */
 285 #define S_SYMBOL_STUBS                  0x8     /* section with only symbol
 286                                                    stubs, byte size of stub in
 287                                                    the reserved2 field */
 288 #define S_MOD_INIT_FUNC_POINTERS        0x9     /* section with only function
 289                                                    pointers for initialization*/
 290 /*
 291  * Constants for the section attributes part of the flags field of a section
 292  * structure.
 293  */
 294 #define SECTION_ATTRIBUTES_USR   0xff000000     /* User setable attributes */
 295 #define S_ATTR_PURE_INSTRUCTIONS 0x80000000     /* section contains only true
 296                                                    machine instructions */
 297 #define SECTION_ATTRIBUTES_SYS   0x00ffff00     /* system setable attributes */
 298 #define S_ATTR_SOME_INSTRUCTIONS 0x00000400     /* section contains some
 299                                                    machine instructions */
 300 #define S_ATTR_EXT_RELOC         0x00000200     /* section has external
 301                                                    relocation entries */
 302 #define S_ATTR_LOC_RELOC         0x00000100     /* section has local
 303                                                    relocation entries */
 304
 305
 306 /*
 307  * The names of segments and sections in them are mostly meaningless to the
 308  * link-editor.  But there are few things to support traditional UNIX
 309  * executables that require the link-editor and assembler to use some names
 310  * agreed upon by convention.
 311  *
 312  * The initial protection of the "__TEXT" segment has write protection turned
 313  * off (not writeable).
 314  *
 315  * The link-editor will allocate common symbols at the end of the "__common"
 316  * section in the "__DATA" segment.  It will create the section and segment
 317  * if needed.
 318  */
 319
 320 /* The currently known segment names and the section names in those segments */
 321
 322 #define SEG_PAGEZERO    "__PAGEZERO"    /* the pagezero segment which has no */
 323                                         /* protections and catches NULL */
 324                                         /* references for MH_EXECUTE files */
 325
 326
 327 #define SEG_TEXT        "__TEXT"        /* the tradition UNIX text segment */
 328 #define SECT_TEXT       "__text"        /* the real text part of the text */
 329                                         /* section no headers, and no padding */
 330 #define SECT_FVMLIB_INIT0 "__fvmlib_init0"      /* the fvmlib initialization */
 331                                                 /*  section */
 332 #define SECT_FVMLIB_INIT1 "__fvmlib_init1"      /* the section following the */
 333                                                 /*  fvmlib initialization */
 334                                                 /*  section */
 335
 336 #define SEG_DATA        "__DATA"        /* the tradition UNIX data segment */
 337 #define SECT_DATA       "__data"        /* the real initialized data section */
 338                                         /* no padding, no bss overlap */
 339 #define SECT_BSS        "__bss"         /* the real uninitialized data section*/
 340                                         /* no padding */
 341 #define SECT_COMMON     "__common"      /* the section common symbols are */
 342                                         /* allocated in by the link editor */
 343
 344 #define SEG_OBJC        "__OBJC"        /* objective-C runtime segment */
 345 #define SECT_OBJC_SYMBOLS "__symbol_table"      /* symbol table */
 346 #define SECT_OBJC_MODULES "__module_info"       /* module information */
 347 #define SECT_OBJC_STRINGS "__selector_strs"     /* string table */
 348 #define SECT_OBJC_REFS "__selector_refs"        /* string table */
 349
 350 #define SEG_ICON         "__ICON"       /* the NeXT icon segment */
 351 #define SECT_ICON_HEADER "__header"     /* the icon headers */
 352 #define SECT_ICON_TIFF   "__tiff"       /* the icons in tiff format */
 353
 354 #define SEG_LINKEDIT    "__LINKEDIT"    /* the segment containing all structs */
 355                                         /* created and maintained by the link */
 356                                         /* editor.  Created with -seglinkedit */
 357                                         /* option to ld(1) for MH_EXECUTE and */
 358                                         /* FVMLIB file types only */
 359
 360 #define SEG_UNIXSTACK   "__UNIXSTACK"   /* the unix stack segment */
 361
 362 /*
 363  * Fixed virtual memory shared libraries are identified by two things.  The
 364  * target pathname (the name of the library as found for execution), and the
 365  * minor version number.  The address of where the headers are loaded is in
 366  * header_addr.
 367  */
 368 struct fvmlib {
 369         union lc_str    name;           /* library's target pathname */
 370         unsigned long   minor_version;  /* library's minor version number */
 371         unsigned long   header_addr;    /* library's header address */
 372 };
 373
 374 /*
 375  * A fixed virtual shared library (filetype == MH_FVMLIB in the mach header)
 376  * contains a fvmlib_command (cmd == LC_IDFVMLIB) to identify the library.
 377  * An object that uses a fixed virtual shared library also contains a
 378  * fvmlib_command (cmd == LC_LOADFVMLIB) for each library it uses.
 379  */
 380 struct fvmlib_command {
 381         unsigned long   cmd;            /* LC_IDFVMLIB or LC_LOADFVMLIB */
 382         unsigned long   cmdsize;        /* includes pathname string */
 383         struct fvmlib   fvmlib;         /* the library identification */
 384 };
 385
 386 /*
 387  * Dynamicly linked shared libraries are identified by two things.  The
 388  * pathname (the name of the library as found for execution), and the
 389  * compatibility version number.  The pathname must match and the compatibility
 390  * number in the user of the library must be greater than or equal to the
 391  * library being used.  The time stamp is used to record the time a library was
 392  * built and copied into user so it can be use to determined if the library used
 393  * at runtime is exactly the same as used to built the program.
 394  */
 395 struct dylib {
 396     union lc_str  name;                 /* library's path name */
 397     unsigned long timestamp;            /* library's build time stamp */
 398     unsigned long current_version;      /* library's current version number */
 399     unsigned long compatibility_version;/* library's compatibility vers number*/
 400 };
 401
 402 /*
 403  * A dynamicly linked shared library (filetype == MH_DYLIB in the mach header)
 404  * contains a dylib_command (cmd == LC_ID_DYLIB) to identify the library.
 405  * An object that uses a dynamicly linked shared library also contains a
 406  * dylib_command (cmd == LC_LOAD_DYLIB) for each library it uses.
 407  */
 408 struct dylib_command {
 409         unsigned long   cmd;            /* LC_ID_DYLIB or LC_LOAD_DYLIB */
 410         unsigned long   cmdsize;        /* includes pathname string */
 411         struct dylib    dylib;          /* the library identification */
 412 };
 413
 414 /*
 415  * A program (filetype == MH_EXECUTE) or bundle (filetype == MH_BUNDLE) that is
 416  * prebound to it's dynamic libraries has one of these for each library that
 417  * the static linker used in prebinding.  It contains a bit vector for the
 418  * modules in the library.  The bits indicate which modules are bound (1) and
 419  * which are not (0) from the library.  The bit for module 0 is the low bit
 420  * of the first byte.  So the bit for the Nth module is:
 421  * (linked_modules[N/8] >> N%8) & 1
 422  */
 423 struct prebound_dylib_command {
 424         unsigned long   cmd;            /* LC_PREBOUND_DYLIB */
 425         unsigned long   cmdsize;        /* includes strings */
 426         union lc_str    name;           /* library's path name */
 427         unsigned long   nmodules;       /* number of modules in library */
 428         union lc_str    linked_modules; /* bit vector of linked modules */
 429 };
 430
 431 /*
 432  * A program that uses a dynamic linker contains a dylinker_command to identify
 433  * the name of the dynamic linker (LC_LOAD_DYLINKER).  And a dynamic linker
 434  * contains a dylinker_command to identify the dynamic linker (LC_ID_DYLINKER).
 435  * A file can have at most one of these.
 436  */
 437 struct dylinker_command {
 438         unsigned long   cmd;            /* LC_ID_DYLINKER or LC_LOAD_DYLINKER */
 439         unsigned long   cmdsize;        /* includes pathname string */
 440         union lc_str    name;           /* dynamic linker's path name */
 441 };
 442
 443 /*
 444  * Thread commands contain machine-specific data structures suitable for
 445  * use in the thread state primitives.  The machine specific data structures
 446  * follow the struct thread_command as follows.
 447  * Each flavor of machine specific data structure is preceded by an unsigned
 448  * long constant for the flavor of that data structure, an unsigned long
 449  * that is the count of longs of the size of the state data structure and then
 450  * the state data structure follows.  This triple may be repeated for many
 451  * flavors.  The constants for the flavors, counts and state data structure
 452  * definitions are expected to be in the header file <machine/thread_status.h>.
 453  * These machine specific data structures sizes must be multiples of
 454  * sizeof(long).  The cmdsize reflects the total size of the thread_command
 455  * and all of the sizes of the constants for the flavors, counts and state
 456  * data structures.
 457  *
 458  * For executable objects that are unix processes there will be one
 459  * thread_command (cmd == LC_UNIXTHREAD) created for it by the link-editor.
 460  * This is the same as a LC_THREAD, except that a stack is automatically
 461  * created (based on the shell's limit for the stack size).  Command arguments
 462  * and environment variables are copied onto that stack.
 463  */
 464 struct thread_command {
 465         unsigned long   cmd;            /* LC_THREAD or  LC_UNIXTHREAD */
 466         unsigned long   cmdsize;        /* total size of this command */
 467         /* unsigned long flavor            flavor of thread state */
 468         /* unsigned long count             count of longs in thread state */
 469         /* struct XXX_thread_state state   thread state for this flavor */
 470         /* ... */
 471 };
 472
 473 /*
 474  * The symtab_command contains the offsets and sizes of the link-edit 4.3BSD
 475  * "stab" style symbol table information as described in the header files
 476  * <nlist.h> and <stab.h>.
 477  */
 478 struct symtab_command {
 479         unsigned long   cmd;            /* LC_SYMTAB */
 480         unsigned long   cmdsize;        /* sizeof(struct symtab_command) */
 481         unsigned long   symoff;         /* symbol table offset */
 482         unsigned long   nsyms;          /* number of symbol table entries */
 483         unsigned long   stroff;         /* string table offset */
 484         unsigned long   strsize;        /* string table size in bytes */
 485 };
 486
 487 /*
 488  * This is the second set of the symbolic information which is used to support
 489  * the data structures for the dynamicly link editor.
 490  *
 491  * The original set of symbolic information in the symtab_command which contains
 492  * the symbol and string tables must also be present when this load command is
 493  * present.  When this load command is present the symbol table is organized
 494  * into three groups of symbols:
 495  *      local symbols (static and debugging symbols) - grouped by module
 496  *      defined external symbols - grouped by module (sorted by name if not lib)
 497  *      undefined external symbols (sorted by name)
 498  * In this load command there are offsets and counts to each of the three groups
 499  * of symbols.
 500  *
 501  * This load command contains a the offsets and sizes of the following new
 502  * symbolic information tables:
 503  *      table of contents
 504  *      module table
 505  *      reference symbol table
 506  *      indirect symbol table
 507  * The first three tables above (the table of contents, module table and
 508  * reference symbol table) are only present if the file is a dynamicly linked
 509  * shared library.  For executable and object modules, which are files
 510  * containing only one module, the information that would be in these three
 511  * tables is determined as follows:
 512  *      table of contents - the defined external symbols are sorted by name
 513  *      module table - the file contains only one module so everything in the
 514  *                     file is part of the module.
 515  *      reference symbol table - is the defined and undefined external symbols
 516  *
 517  * For dynamicly linked shared library files this load command also contains
 518  * offsets and sizes to the pool of relocation entries for all sections
 519  * separated into two groups:
 520  *      external relocation entries
 521  *      local relocation entries
 522  * For executable and object modules the relocation entries continue to hang
 523  * off the section structures.
 524  */
 525 struct dysymtab_command {
 526     unsigned long cmd;          /* LC_DYSYMTAB */
 527     unsigned long cmdsize;      /* sizeof(struct dysymtab_command) */
 528
 529     /*
 530      * The symbols indicated by symoff and nsyms of the LC_SYMTAB load command
 531      * are grouped into the following three groups:
 532      *    local symbols (further grouped by the module they are from)
 533      *    defined external symbols (further grouped by the module they are from)
 534      *    undefined symbols
 535      *
 536      * The local symbols are used only for debugging.  The dynamic binding
 537      * process may have to use them to indicate to the debugger the local
 538      * symbols for a module that is being bound.
 539      *
 540      * The last two groups are used by the dynamic binding process to do the
 541      * binding (indirectly through the module table and the reference symbol
 542      * table when this is a dynamicly linked shared library file).
 543      */
 544     unsigned long ilocalsym;    /* index to local symbols */
 545     unsigned long nlocalsym;    /* number of local symbols */
 546
 547     unsigned long iextdefsym;   /* index to externally defined symbols */
 548     unsigned long nextdefsym;   /* number of externally defined symbols */
 549
 550     unsigned long iundefsym;    /* index to undefined symbols */
 551     unsigned long nundefsym;    /* number of undefined symbols */
 552
 553     /*
 554      * For the for the dynamic binding process to find which module a symbol
 555      * is defined in the table of contents is used (analogous to the ranlib
 556      * structure in an archive) which maps defined external symbols to modules
 557      * they are defined in.  This exists only in a dynamicly linked shared
 558      * library file.  For executable and object modules the defined external
 559      * symbols are sorted by name and is use as the table of contents.
 560      */
 561     unsigned long tocoff;       /* file offset to table of contents */
 562     unsigned long ntoc;         /* number of entries in table of contents */
 563
 564     /*
 565      * To support dynamic binding of "modules" (whole object files) the symbol
 566      * table must reflect the modules that the file was created from.  This is
 567      * done by having a module table that has indexes and counts into the merged
 568      * tables for each module.  The module structure that these two entries
 569      * refer to is described below.  This exists only in a dynamicly linked
 570      * shared library file.  For executable and object modules the file only
 571      * contains one module so everything in the file belongs to the module.
 572      */
 573     unsigned long modtaboff;    /* file offset to module table */
 574     unsigned long nmodtab;      /* number of module table entries */
 575
 576     /*
 577      * To support dynamic module binding the module structure for each module
 578      * indicates the external references (defined and undefined) each module
 579      * makes.  For each module there is an offset and a count into the
 580      * reference symbol table for the symbols that the module references.
 581      * This exists only in a dynamicly linked shared library file.  For
 582      * executable and object modules the defined external symbols and the
 583      * undefined external symbols indicates the external references.
 584      */
 585     unsigned long extrefsymoff;  /* offset to referenced symbol table */
 586     unsigned long nextrefsyms;   /* number of referenced symbol table entries */
 587
 588     /*
 589      * The sections that contain "symbol pointers" and "routine stubs" have
 590      * indexes and (implied counts based on the size of the section and fixed
 591      * size of the entry) into the "indirect symbol" table for each pointer
 592      * and stub.  For every section of these two types the index into the
 593      * indirect symbol table is stored in the section header in the field
 594      * reserved1.  An indirect symbol table entry is simply a 32bit index into
 595      * the symbol table to the symbol that the pointer or stub is referring to.
 596      * The indirect symbol table is ordered to match the entries in the section.
 597      */
 598     unsigned long indirectsymoff; /* file offset to the indirect symbol table */
 599     unsigned long nindirectsyms;  /* number of indirect symbol table entries */
 600
 601     /*
 602      * To support relocating an individual module in a library file quickly the
 603      * external relocation entries for each module in the library need to be
 604      * accessed efficiently.  Since the relocation entries can't be accessed
 605      * through the section headers for a library file they are separated into
 606      * groups of local and external entries further grouped by module.  In this
 607      * case the presents of this load command who's extreloff, nextrel,
 608      * locreloff and nlocrel fields are non-zero indicates that the relocation
 609      * entries of non-merged sections are not referenced through the section
 610      * structures (and the reloff and nreloc fields in the section headers are
 611      * set to zero).
 612      *
 613      * Since the relocation entries are not accessed through the section headers
 614      * this requires the r_address field to be something other than a section
 615      * offset to identify the item to be relocated.  In this case r_address is
 616      * set to the offset from the vmaddr of the first LC_SEGMENT command.
 617      *
 618      * The relocation entries are grouped by module and the module table
 619      * entries have indexes and counts into them for the group of external
 620      * relocation entries for that the module.
 621      *
 622      * For sections that are merged across modules there must not be any
 623      * remaining external relocation entries for them (for merged sections
 624      * remaining relocation entries must be local).
 625      */
 626     unsigned long extreloff;    /* offset to external relocation entries */
 627     unsigned long nextrel;      /* number of external relocation entries */
 628
 629     /*
 630      * All the local relocation entries are grouped together (they are not
 631      * grouped by their module since they are only used if the object is moved
 632      * from it staticly link edited address).
 633      */
 634     unsigned long locreloff;    /* offset to local relocation entries */
 635     unsigned long nlocrel;      /* number of local relocation entries */
 636
 637 };
 638
 639 /*
 640  * An indirect symbol table entry is simply a 32bit index into the symbol table
 641  * to the symbol that the pointer or stub is refering to.  Unless it is for a
 642  * non-lazy symbol pointer section for a defined symbol which strip(1) as
 643  * removed.  In which case it has the value INDIRECT_SYMBOL_LOCAL.  If the
 644  * symbol was also absolute INDIRECT_SYMBOL_ABS is or'ed with that.
 645  */
 646 #define INDIRECT_SYMBOL_LOCAL   0x80000000
 647 #define INDIRECT_SYMBOL_ABS     0x40000000
 648
 649
 650 /* a table of contents entry */
 651 struct dylib_table_of_contents {
 652     unsigned long symbol_index; /* the defined external symbol
 653                                    (index into the symbol table) */
 654     unsigned long module_index; /* index into the module table this symbol
 655                                    is defined in */
 656 };
 657
 658 /* a module table entry */
 659 struct dylib_module {
 660     unsigned long module_name;  /* the module name (index into string table) */
 661
 662     unsigned long iextdefsym;   /* index into externally defined symbols */
 663     unsigned long nextdefsym;   /* number of externally defined symbols */
 664     unsigned long irefsym;              /* index into reference symbol table */
 665     unsigned long nrefsym;      /* number of reference symbol table entries */
 666     unsigned long ilocalsym;    /* index into symbols for local symbols */
 667     unsigned long nlocalsym;    /* number of local symbols */
 668
 669     unsigned long iextrel;      /* index into external relocation entries */
 670     unsigned long nextrel;      /* number of external relocation entries */
 671
 672     unsigned long iinit;        /* index into the init section */
 673     unsigned long ninit;        /* number of init section entries */
 674
 675     unsigned long               /* for this module address of the start of */
 676         objc_module_info_addr;  /*  the (__OBJC,__module_info) section */
 677     unsigned long               /* for this module size of */
 678         objc_module_info_size;  /*  the (__OBJC,__module_info) section */
 679 };
 680
 681 /*
 682  * The entries in the reference symbol table are used when loading the module
 683  * (both by the static and dynamic link editors) and if the module is unloaded
 684  * or replaced.  Therefore all external symbols (defined and undefined) are
 685  * listed in the module's reference table.  The flags describe the type of
 686  * reference that is being made.  The constants for the flags are defined in
 687  * <mach-o/nlist.h> as they are also used for symbol table entries.
 688  */
 689 struct dylib_reference {
 690     unsigned long isym:24,      /* index into the symbol table */
 691                   flags:8;      /* flags to indicate the type of reference */
 692 };
 693
 694 /*
 695  * The uuid load command contains a single 128-bit unique random number that
 696  * identifies an object produced by the static link editor.
 697  */
 698 struct uuid_command {
 699     unsigned long       cmd;            /* LC_UUID */
 700     unsigned long       cmdsize;        /* sizeof(struct uuid_command) */
 701     unsigned char       uuid[16];       /* the 128-bit uuid */
 702 };
 703
 704 /*
 705  * The symseg_command contains the offset and size of the GNU style
 706  * symbol table information as described in the header file <symseg.h>.
 707  * The symbol roots of the symbol segments must also be aligned properly
 708  * in the file.  So the requirement of keeping the offsets aligned to a
 709  * multiple of a sizeof(long) translates to the length field of the symbol
 710  * roots also being a multiple of a long.  Also the padding must again be
 711  * zeroed. (THIS IS OBSOLETE and no longer supported).
 712  */
 713 struct symseg_command {
 714         unsigned long   cmd;            /* LC_SYMSEG */
 715         unsigned long   cmdsize;        /* sizeof(struct symseg_command) */
 716         unsigned long   offset;         /* symbol segment offset */
 717         unsigned long   size;           /* symbol segment size in bytes */
 718 };
 719
 720 /*
 721  * The ident_command contains a free format string table following the
 722  * ident_command structure.  The strings are null terminated and the size of
 723  * the command is padded out with zero bytes to a multiple of sizeof(long).
 724  * (THIS IS OBSOLETE and no longer supported).
 725  */
 726 struct ident_command {
 727         unsigned long cmd;              /* LC_IDENT */
 728         unsigned long cmdsize;          /* strings that follow this command */
 729 };
 730
 731 /*
 732  * The fvmfile_command contains a reference to a file to be loaded at the
 733  * specified virtual address.  (Presently, this command is reserved for NeXT
 734  * internal use.  The kernel ignores this command when loading a program into
 735  * memory).
 736  */
 737 struct fvmfile_command {
 738         unsigned long cmd;              /* LC_FVMFILE */
 739         unsigned long cmdsize;          /* includes pathname string */
 740         union lc_str    name;           /* files pathname */
 741         unsigned long   header_addr;    /* files virtual address */
 742 };
 743
 744 #endif /*_MACHO_LOADER_H_*/