1 /* 2 * Copyright (c) 1982 Regents of the University of California. 3 * All rights reserved. The Berkeley software License Agreement 4 * specifies the terms and conditions for redistribution. 5 * 6 * @(#)as.h 5.1 (Berkeley) 04/24/85 7 */ 8 9 #define reg register 10 11 #include <sys/types.h> 12 #include <a.out.h> 13 #include <stab.h> 14 15 #define readonly 16 #define NINST 300 17 18 #define NEXP 20 /* max number of expr. terms per instruction */ 19 #define NARG 6 /* max number of args per instruction */ 20 #define NHASH 1103 /* hash table is dynamically extended */ 21 #define TNAMESIZE 32 /* maximum length of temporary file names */ 22 #define NLOC 4 /* number of location ctrs */ 23 /* 24 * Sizes for character buffers. 25 * what size #define name comments 26 * 27 * name assembly NCPName 28 * name save STRPOOLDALLOP 29 * 30 * -name saving is a simple first fit 31 */ 32 #ifndef STRPOOLDALLOP 33 # define STRPOOLDALLOP 8192 34 #endif not STRPOOLDALLOP 35 36 #define NCPName NCPS 37 #ifndef NCPS 38 # undef NCPName 39 # define NCPName 4096 40 #endif not NCPS 41 /* 42 * Check sizes, and compiler error if sizes botch 43 */ 44 #if STRPOOLDALLOP < NCPName 45 $$$botch with definition sizes 46 #endif test botches 47 /* 48 * Symbol types 49 */ 50 #define XUNDEF 0x0 51 #define XABS 0x2 52 #define XTEXT 0x4 53 #define XDATA 0x6 54 #define XBSS 0x8 55 56 #define XXTRN 0x1 57 #define XTYPE 0x1E 58 59 #define XFORW 0x20 /* Was forward-referenced when undefined */ 60 61 #define ERR (-1) 62 #define NBPW 32 /* Bits per word */ 63 64 #define AMASK 017 65 66 /* 67 * Actual argument syntax types 68 */ 69 #define AREG 1 /* %r */ 70 #define ABASE 2 /* (%r) */ 71 #define ADECR 3 /* -(%r) */ 72 #define AINCR 4 /* (%r)+ */ 73 #define ADISP 5 /* expr(%r) */ 74 #define AEXP 6 /* expr */ 75 #define AIMM 7 /* $ expr */ 76 #define ASTAR 8 /* * */ 77 #define AINDX 16 /* [%r] */ 78 /* 79 * Definitions for the things found in ``instrs'' 80 */ 81 #define INSTTAB 1 82 #include "instrs.h" 83 84 /* 85 * Tells outrel what it is relocating 86 * RELOC_PCREL is an implicit argument to outrel; it is or'ed in 87 * with a TYPX 88 */ 89 #define RELOC_PCREL (1<<TYPLG) 90 /* 91 * reference types for loader 92 */ 93 #define PCREL 1 94 #define LEN1 2 95 #define LEN2 4 96 #define LEN4 6 97 #define LEN8 8 98 #define LEN16 10 99 100 extern int reflen[]; /* {LEN*+PCREL} ==> number of bytes */ 101 extern int lgreflen[]; /* {LEN*+PCREL} ==> lg number of bytes */ 102 extern int len124[]; /* {1,2,4,8,16} ==> {LEN1, LEN2, LEN4, LEN8} */ 103 extern char mod124[]; /* {1,2,4,8,16} ==> {bits to construct operands */ 104 extern int type_124[]; /* {1,2,4,8,16} ==> {TYPB,TYPW,TYPL,TYPQ,TYPO} */ 105 extern int ty_NORELOC[]; /* {TYPB..TYPH} ==> {1 if relocation not OK */ 106 extern int ty_float[]; /* {TYPB..TYPH} ==> {1 if floating number */ 107 extern int ty_LEN[]; /* {TYPB..TYPH} ==> {LEN1..LEN16} */ 108 extern int ty_nbyte[]; /* {TYPB..TYPH} ==> {1,2,4,8,16} */ 109 extern int ty_nlg[]; /* {TYPB..TYPH} ==> lg{1,2,4,8,16} */ 110 extern char *ty_string[]; /* {TYPB..TYPH} ==> printable */ 111 112 #define TMPC 7 113 #define HW 0x1 114 #define FW 0x3 115 #define DW 0x7 116 #define OW 0xF 117 118 #define round(x,y) (((x)+(y)) & ~(y)) 119 120 #define STABTYPS 0340 121 #define STABFLAG 0200 122 123 /* 124 * Follows are the definitions for the symbol table tags, which are 125 * all unsigned characters.. 126 * High value tags are generated by the asembler for internal 127 * use. 128 * Low valued tags are the parser coded tokens the scanner returns. 129 * There are several pertinant bounds in this ordering: 130 * a) Symbols greater than JXQUESTIONABLE 131 * are used by the jxxx bumper, indicating that 132 * the symbol table entry is a jxxx entry 133 * that has yet to be bumped. 134 * b) Symbols greater than IGNOREBOUND are not 135 * bequeathed to the loader; they are truly 136 * for assembler internal use only. 137 * c) Symbols greater than OKTOBUMP represent 138 * indices into the program text that should 139 * be changed in preceeding jumps or aligns 140 * must get turned into their long form. 141 */ 142 143 #define TAGMASK 0xFF 144 145 # define JXACTIVE 0xFF /*jxxx size unknown*/ 146 # define JXNOTYET 0xFE /*jxxx size known, but not yet expanded*/ 147 # define JXALIGN 0xFD /*align jxxx entry*/ 148 # define JXINACTIVE 0xFC /*jxxx size known and expanded*/ 149 150 #define JXQUESTIONABLE 0xFB 151 152 # define JXTUNNEL 0xFA /*jxxx that jumps to another*/ 153 # define OBSOLETE 0xF9 /*erroneously entered symbol*/ 154 155 #define IGNOREBOUND 0xF8 /*symbols greater than this are ignored*/ 156 # define STABFLOATING 0xF7 157 # define LABELID 0xF6 158 159 #define OKTOBUMP 0xF5 160 # define STABFIXED 0xF4 161 162 /* 163 * astoks.h contains reserved word codings the parser should 164 * know about 165 */ 166 #include "astoks.h" 167 168 /* 169 * The structure for one symbol table entry. 170 * Symbol table entries are used for both user defined symbols, 171 * and symbol slots generated to create the jxxx jump from 172 * slots. 173 * Caution: the instructions are stored in a shorter version 174 * of the struct symtab, using all fields in sym_nm and 175 * tag. The fields used in sym_nm are carefully redeclared 176 * in struct Instab and struct instab (see below). 177 * If struct nlist gets changed, then Instab and instab may 178 * have to be changed. 179 */ 180 181 struct symtab{ 182 struct nlist s_nm; 183 u_char s_tag; /* assembler tag */ 184 u_char s_ptype; /* if tag == NAME */ 185 u_char s_jxoveralign; /* if a JXXX, jumped over align */ 186 short s_index; /* which segment */ 187 struct symtab *s_dest; /* if JXXX, where going to */ 188 #ifdef DEBUG 189 short s_jxline; /* source line of the jump from */ 190 #endif 191 }; 192 /* 193 * Redefinitions of the fields in symtab for 194 * use when the symbol table entry marks a jxxx instruction. 195 */ 196 #define s_jxbump s_ptype /* tag == JX..., how far to expand */ 197 #define s_jxfear s_desc /* how far needs to be bumped */ 198 /* 199 * Redefinitions of fields in the struct nlist for symbols so that 200 * one saves typing, and so that they conform 201 * with the old naming conventions. 202 */ 203 #define s_name s_nm.n_un.n_name 204 #define i_name s_name 205 #define FETCHNAME(sp) (((struct strdesc *)(sp)->s_name)->sd_string) 206 #define STRLEN(sp) (((struct strdesc *)(sp)->s_name)->sd_strlen) 207 #define STROFF(sp) (((struct strdesc *)(sp)->s_name)->sd_stroff) 208 #define STRPLACE(sp) (((struct strdesc *)(sp)->s_name)->sd_place) 209 #define s_nmx s_nm.n_un.n_strx /* string table index */ 210 #define s_type s_nm.n_type /* type of the symbol */ 211 #define s_other s_nm.n_other /* other information for sdb */ 212 #define s_desc s_nm.n_desc /* type descriptor */ 213 #define s_value s_nm.n_value /* value of the symbol, or sdb delta */ 214 215 struct instab{ 216 struct nlist s_nm; /* instruction name, type (opcode) */ 217 u_char s_tag; 218 u_char s_eopcode; 219 char s_pad[2]; /* round to 20 bytes */ 220 }; 221 typedef struct instab *Iptr; 222 /* 223 * The fields nm.n_desc and nm.n_value total 6 bytes; this is 224 * just enough for the 6 bytes describing the argument types. 225 * We use a macro to define access to these 6 bytes, assuming that 226 * they are allocated adjacently. 227 * IF THE FORMAT OF STRUCT nlist CHANGES, THESE MAY HAVE TO BE CHANGED. 228 * 229 * Instab is cleverly declared to look very much like the combination of 230 * a struct symtab and a struct nlist. 231 */ 232 /* 233 * With the 1981 VAX architecture reference manual, 234 * DEC defined and named two byte opcodes. 235 * In addition, DEC defined four new one byte instructions for 236 * queue manipulation. 237 * The assembler was patched in 1982 to reflect this change. 238 * 239 * The two byte opcodes are preceded with an escape byte 240 * (usually an ESCD) and an opcode byte. 241 * For one byte opcodes, the opcode is called the primary opcode. 242 * For two byte opcodes, the second opcode is called the primary opcode. 243 * 244 * We store the primary opcode in I_popcode, 245 * and the escape opcode in I_eopcode. 246 * 247 * For one byte opcodes in the basic arhitecture, 248 * I_eopcode is CORE 249 * For one byte opcodes in the new architecture definition, 250 * I_eopcode is NEW 251 * For the two byte opcodes, I_eopcode is the escape byte. 252 * 253 * The assembler checks if a NEW or two byte opcode is used, 254 * and issues a warning diagnostic. 255 */ 256 /* 257 * For upward compatability reasons, we can't have the two opcodes 258 * forming an operator specifier byte(s) be physically adjacent 259 * in the instruction table. 260 * We define a structure and a constructor that is used in 261 * the instruction generator. 262 */ 263 struct Opcode{ 264 u_char Op_eopcode; 265 u_char Op_popcode; 266 }; 267 268 #define BADPOINT 0xAAAAAAAA 269 /* 270 * See if a structured opcode is bad 271 */ 272 #define ITABCHECK(o) ((itab[o.Op_eopcode] != (Iptr*)BADPOINT) && (itab[o.Op_eopcode][o.Op_popcode] != (Iptr)BADPOINT)) 273 /* 274 * Index the itab by a structured opcode 275 */ 276 #define ITABFETCH(o) itab[o.Op_eopcode][o.Op_popcode] 277 278 struct Instab{ 279 char *I_name; 280 u_char I_popcode; /* basic op code */ 281 char I_nargs; 282 char I_args[6]; 283 u_char I_s_tag; 284 u_char I_eopcode; 285 char I_pad[2]; /* round to 20 bytes */ 286 }; 287 /* 288 * Redefinitions of fields in the struct nlist for instructions so that 289 * one saves typing, and conforms to the old naming conventions 290 */ 291 #define i_popcode s_nm.n_type /* use the same field as symtab.type */ 292 #define i_eopcode s_eopcode 293 #define i_nargs s_nm.n_other /* number of arguments */ 294 #define fetcharg(ptr, n) ((struct Instab *)ptr)->I_args[n] 295 296 struct arg { /*one argument to an instruction*/ 297 char a_atype; 298 char a_areg1; 299 char a_areg2; 300 char a_dispsize; /*usually d124, unless have B^, etc*/ 301 struct exp *a_xp; 302 }; 303 /* 304 * Definitions for numbers and expressions. 305 */ 306 #include "asnumber.h" 307 struct exp { 308 Bignum e_number; /* 128 bits of #, plus tag */ 309 char e_xtype; 310 char e_xloc; 311 struct symtab *e_xname; 312 }; 313 #define e_xvalue e_number.num_num.numIl_int.Il_long 314 315 #define MINLIT 0 316 #define MAXLIT 63 317 318 #define MINBYTE -128 319 #define MAXBYTE 127 320 #define MINUBYTE 0 321 #define MAXUBYTE 255 322 323 #define MINWORD -32768 324 #define MAXWORD 32767 325 #define MINUWORD 0 326 #define MAXUWORD 65535 327 328 #define ISLIT(x) (((x) >= MINLIT) && ((x) <= MAXLIT)) 329 #define ISBYTE(x) (((x) >= MINBYTE) && ((x) <= MAXBYTE)) 330 #define ISUBYTE(x) (((x) >= MINUBYTE) && ((x) <= MAXUBYTE)) 331 #define ISWORD(x) (((x) >= MINWORD) && ((x) <= MAXWORD)) 332 #define ISUWORD(x) (((x) >= MINUWORD) && ((x) <= MAXUWORD)) 333 /* 334 * Definitions for strings. 335 * 336 * Strings are stored in the string pool; see strsave(str, length) 337 * Strings are known by their length and values. 338 * A string pointer points to the beginning of the value bytes; 339 * 340 * If this structure is changed, change insts also. 341 */ 342 struct strdesc{ 343 int sd_stroff; /* offset into string file */ 344 short sd_place; /* where string is */ 345 u_short sd_strlen; /* string length */ 346 char sd_string[1]; /* the string itself, flexible length */ 347 }; 348 /* 349 * Where a string can be. If these are changed, also change instrs. 350 */ 351 #define STR_FILE 0x1 352 #define STR_CORE 0x2 353 #define STR_BOTH 0x3 354 355 struct strdesc *savestr(); 356 357 /* 358 * Global variables 359 */ 360 extern struct arg arglist[NARG]; /*building operands in instructions*/ 361 extern struct exp explist[NEXP]; /*building up a list of expressions*/ 362 extern struct exp *xp; /*current free expression*/ 363 /* 364 * Communication between the scanner and the jxxx handlers. 365 * lastnam: the last name seen on the input 366 * lastjxxx: pointer to the last symbol table entry for 367 * a jump from 368 */ 369 extern struct symtab *lastnam; 370 extern struct symtab *lastjxxx; 371 /* 372 * Lgensym is used to make up funny names for local labels. 373 * lgensym[i] is the current funny number to put after 374 * references to if, lgensym[i]-1 is for ib. 375 * genref[i] is set when the label is referenced before 376 * it is defined (i.e. 2f) so that we can be sure these 377 * labels are always defined to avoid weird diagnostics 378 * from the loader later. 379 */ 380 extern int lgensym[10]; 381 extern char genref[10]; 382 383 extern struct exp *dotp; /* the current dot location */ 384 extern int loctr; 385 386 extern struct exec hdr; /* a.out header */ 387 extern u_long tsize; /* total text size */ 388 extern u_long dsize; /* total data size */ 389 extern u_long trsize; /* total text relocation size */ 390 extern u_long drsize; /* total data relocation size */ 391 extern u_long datbase; /* base of the data segment */ 392 /* 393 * Bitoff and bitfield keep track of the packing into 394 * bytes mandated by the expression syntax <expr> ':' <expr> 395 */ 396 extern int bitoff; 397 extern long bitfield; 398 399 /* 400 * The lexical analyzer builds up symbols in yytext. Lookup 401 * expects its argument in this buffer 402 */ 403 extern char yytext[NCPName+2]; /* text buffer for lexical */ 404 /* 405 * Variables to manage the input assembler source file 406 */ 407 extern int lineno; /*the line number*/ 408 extern char *dotsname; /*the name of the as source*/ 409 410 extern FILE *tokfile; /* temp token communication*/ 411 extern FILE *strfile; /* temp string file*/ 412 extern char tokfilename[TNAMESIZE]; /* token file name */ 413 extern char strfilename[TNAMESIZE]; /* string file name */ 414 extern int strfilepos; /* position in string file */ 415 416 extern int passno; /* 1 or 2 */ 417 418 extern int anyerrs; /*errors as'ing arguments*/ 419 extern int anywarnings; /*warnings as'ing arguments*/ 420 extern int silent; /*don't mention the errors*/ 421 extern int savelabels; /*save labels in a.out*/ 422 extern int orgwarn; /* questionable origin ? */ 423 extern int useVM; /*use virtual memory temp file*/ 424 extern int jxxxJUMP; /*use jmp instead of brw for jxxx */ 425 extern int readonlydata; /*initialized data into text space*/ 426 extern int nGHnumbers; /* GH numbers used */ 427 extern int nGHopcodes; /* GH opcodes used */ 428 extern int nnewopcodes; /* new opcodes used */ 429 #ifdef DEBUG 430 extern int debug; 431 extern int toktrace; 432 #endif 433 /* 434 * Information about the instructions 435 */ 436 extern struct instab **itab[NINST]; /*maps opcodes to instructions*/ 437 extern readonly struct Instab instab[]; 438 439 extern int curlen; /*current literal storage size*/ 440 extern int d124; /*current pointer storage size*/ 441 extern int maxalign; /*maximum .align allowed*/ 442 443 struct symtab **lookup(); /*argument in yytext*/ 444 struct symtab *symalloc(); 445 446 char *Calloc(); 447 char *ClearCalloc(); 448 449 #define outb(val) {dotp->e_xvalue++; if (passno==2) bputc((val), (txtfil));} 450 451 #define outs(cp, lg) dotp->e_xvalue += (lg); if (passno == 2) bwrite((cp), (lg), (txtfil)) 452 453 #define Outb(o) outb(o) 454 /* 455 * Most of the time, the argument to flushfield is a power of two constant, 456 * the calculations involving it can be optimized to shifts. 457 */ 458 #define flushfield(n) if (bitoff != 0) Flushfield( ( (bitoff+n-1) /n ) * n) 459 460 /* 461 * The biobuf structure and associated routines are used to write 462 * into one file at several places concurrently. Calling bopen 463 * with a biobuf structure sets it up to write ``biofd'' starting 464 * at the specified offset. You can then use ``bwrite'' and/or ``bputc'' 465 * to stuff characters in the stream, much like ``fwrite'' and ``fputc''. 466 * Calling bflush drains all the buffers and MUST be done before exit. 467 */ 468 struct biobuf { 469 short b_nleft; /* Number free spaces left in b_buf */ 470 /* Initialize to be less than BUFSIZ initially, to boundary align in file */ 471 char *b_ptr; /* Next place to stuff characters */ 472 char *b_buf; /* Pointer to the buffer */ 473 off_t b_off; /* Current file offset */ 474 struct biobuf *b_link; /* Link in chain for bflush() */ 475 }; 476 #define bputc(c,b) ((b)->b_nleft ? (--(b)->b_nleft, *(b)->b_ptr++ = (c)) \ 477 : bflushc(b, c)) 478 #define BFILE struct biobuf 479 480 extern BFILE *biobufs; /* head of the block I/O buffer chain */ 481 extern int biofd; /* file descriptor for block I/O file */ 482 extern int biobufsize; /* optimal block size for I/O */ 483 extern off_t boffset; /* physical position in logical file */ 484 485 /* 486 * For each of the named .text .data segments 487 * (introduced by .text <expr>), we maintain 488 * the current value of the dot, and the BFILE where 489 * the information for each of the segments is placed 490 * during the second pass. 491 */ 492 extern struct exp usedot[NLOC + NLOC]; 493 extern BFILE *usefile[NLOC + NLOC]; 494 extern BFILE *txtfil;/* file for text and data: into usefile */ 495 /* 496 * Relocation information for each segment is accumulated 497 * seperately from the others. Writing the relocation 498 * information is logically viewed as writing to one 499 * relocation saving file for each segment; physically 500 * we have a bunch of buffers allocated internally that 501 * contain the relocation information. 502 */ 503 struct relbufdesc *rusefile[NLOC + NLOC]; 504 struct relbufdesc *relfil; 505