1 /* 2 * Copyright (c) 1982 Regents of the University of California 3 * @(#)as.h 4.19 08/11/83 4 */ 5 #define reg register 6 7 #include <sys/types.h> 8 #include <a.out.h> 9 #include <stab.h> 10 11 #define readonly 12 #define NINST 300 13 14 #define NEXP 20 /* max number of expr. terms per instruction */ 15 #define NARG 6 /* max number of args per instruction */ 16 #define NHASH 1103 /* hash table is dynamically extended */ 17 #define TNAMESIZE 32 /* maximum length of temporary file names */ 18 #define NLOC 4 /* number of location ctrs */ 19 /* 20 * Sizes for character buffers. 21 * what size #define name comments 22 * 23 * name assembly NCPName 24 * name save STRPOOLDALLOP 25 * 26 * -name saving is a simple first fit 27 */ 28 #ifndef STRPOOLDALLOP 29 # define STRPOOLDALLOP 8192 30 #endif not STRPOOLDALLOP 31 32 #define NCPName NCPS 33 #ifndef NCPS 34 # undef NCPName 35 # define NCPName 4096 36 #endif not NCPS 37 /* 38 * Check sizes, and compiler error if sizes botch 39 */ 40 #if STRPOOLDALLOP < NCPName 41 $$$botch with definition sizes 42 #endif test botches 43 /* 44 * Symbol types 45 */ 46 #define XUNDEF 0x0 47 #define XABS 0x2 48 #define XTEXT 0x4 49 #define XDATA 0x6 50 #define XBSS 0x8 51 52 #define XXTRN 0x1 53 #define XTYPE 0x1E 54 55 #define XFORW 0x20 /* Was forward-referenced when undefined */ 56 57 #define ERR (-1) 58 #define NBPW 32 /* Bits per word */ 59 60 #define AMASK 017 61 62 /* 63 * Actual argument syntax types 64 */ 65 #define AREG 1 /* %r */ 66 #define ABASE 2 /* (%r) */ 67 #define ADECR 3 /* -(%r) */ 68 #define AINCR 4 /* (%r)+ */ 69 #define ADISP 5 /* expr(%r) */ 70 #define AEXP 6 /* expr */ 71 #define AIMM 7 /* $ expr */ 72 #define ASTAR 8 /* * */ 73 #define AINDX 16 /* [%r] */ 74 /* 75 * Definitions for the things found in ``instrs'' 76 */ 77 #define INSTTAB 1 78 #include "instrs.h" 79 80 /* 81 * Tells outrel what it is relocating 82 * RELOC_PCREL is an implicit argument to outrel; it is or'ed in 83 * with a TYPX 84 */ 85 #define RELOC_PCREL (1<<TYPLG) 86 /* 87 * reference types for loader 88 */ 89 #define PCREL 1 90 #define LEN1 2 91 #define LEN2 4 92 #define LEN4 6 93 #define LEN8 8 94 #define LEN16 10 95 96 extern int reflen[]; /* {LEN*+PCREL} ==> number of bytes */ 97 extern int lgreflen[]; /* {LEN*+PCREL} ==> lg number of bytes */ 98 extern int len124[]; /* {1,2,4,8,16} ==> {LEN1, LEN2, LEN4, LEN8} */ 99 extern char mod124[]; /* {1,2,4,8,16} ==> {bits to construct operands */ 100 extern int type_124[]; /* {1,2,4,8,16} ==> {TYPB,TYPW,TYPL,TYPQ,TYPO} */ 101 extern int ty_NORELOC[]; /* {TYPB..TYPH} ==> {1 if relocation not OK */ 102 extern int ty_float[]; /* {TYPB..TYPH} ==> {1 if floating number */ 103 extern int ty_LEN[]; /* {TYPB..TYPH} ==> {LEN1..LEN16} */ 104 extern int ty_nbyte[]; /* {TYPB..TYPH} ==> {1,2,4,8,16} */ 105 extern int ty_nlg[]; /* {TYPB..TYPH} ==> lg{1,2,4,8,16} */ 106 extern char *ty_string[]; /* {TYPB..TYPH} ==> printable */ 107 108 #define TMPC 7 109 #define HW 0x1 110 #define FW 0x3 111 #define DW 0x7 112 #define OW 0xF 113 114 #define round(x,y) (((x)+(y)) & ~(y)) 115 116 #define STABTYPS 0340 117 #define STABFLAG 0200 118 119 /* 120 * Follows are the definitions for the symbol table tags, which are 121 * all unsigned characters.. 122 * High value tags are generated by the asembler for internal 123 * use. 124 * Low valued tags are the parser coded tokens the scanner returns. 125 * There are several pertinant bounds in this ordering: 126 * a) Symbols greater than JXQUESTIONABLE 127 * are used by the jxxx bumper, indicating that 128 * the symbol table entry is a jxxx entry 129 * that has yet to be bumped. 130 * b) Symbols greater than IGNOREBOUND are not 131 * bequeathed to the loader; they are truly 132 * for assembler internal use only. 133 * c) Symbols greater than OKTOBUMP represent 134 * indices into the program text that should 135 * be changed in preceeding jumps or aligns 136 * must get turned into their long form. 137 */ 138 139 #define TAGMASK 0xFF 140 141 # define JXACTIVE 0xFF /*jxxx size unknown*/ 142 # define JXNOTYET 0xFE /*jxxx size known, but not yet expanded*/ 143 # define JXALIGN 0xFD /*align jxxx entry*/ 144 # define JXINACTIVE 0xFC /*jxxx size known and expanded*/ 145 146 #define JXQUESTIONABLE 0xFB 147 148 # define JXTUNNEL 0xFA /*jxxx that jumps to another*/ 149 # define OBSOLETE 0xF9 /*erroneously entered symbol*/ 150 151 #define IGNOREBOUND 0xF8 /*symbols greater than this are ignored*/ 152 # define STABFLOATING 0xF7 153 # define LABELID 0xF6 154 155 #define OKTOBUMP 0xF5 156 # define STABFIXED 0xF4 157 158 /* 159 * astoks.h contains reserved word codings the parser should 160 * know about 161 */ 162 #include "astoks.h" 163 164 /* 165 * The structure for one symbol table entry. 166 * Symbol table entries are used for both user defined symbols, 167 * and symbol slots generated to create the jxxx jump from 168 * slots. 169 * Caution: the instructions are stored in a shorter version 170 * of the struct symtab, using all fields in sym_nm and 171 * tag. The fields used in sym_nm are carefully redeclared 172 * in struct Instab and struct instab (see below). 173 * If struct nlist gets changed, then Instab and instab may 174 * have to be changed. 175 */ 176 177 struct symtab{ 178 struct nlist s_nm; 179 u_char s_tag; /* assembler tag */ 180 u_char s_ptype; /* if tag == NAME */ 181 u_char s_jxoveralign; /* if a JXXX, jumped over align */ 182 short s_index; /* which segment */ 183 struct symtab *s_dest; /* if JXXX, where going to */ 184 #ifdef DJXXX 185 short s_jxline; /* source line of the jump from */ 186 #endif 187 }; 188 /* 189 * Redefinitions of the fields in symtab for 190 * use when the symbol table entry marks a jxxx instruction. 191 */ 192 #define s_jxbump s_ptype /* tag == JX..., how far to expand */ 193 #define s_jxfear s_desc /* how far needs to be bumped */ 194 /* 195 * Redefinitions of fields in the struct nlist for symbols so that 196 * one saves typing, and so that they conform 197 * with the old naming conventions. 198 */ 199 #define s_name s_nm.n_un.n_name 200 #define i_name s_name 201 #define FETCHNAME(sp) (((struct strdesc *)(sp)->s_name)->sd_string) 202 #define STRLEN(sp) (((struct strdesc *)(sp)->s_name)->sd_strlen) 203 #define STROFF(sp) (((struct strdesc *)(sp)->s_name)->sd_stroff) 204 #define STRPLACE(sp) (((struct strdesc *)(sp)->s_name)->sd_place) 205 #define s_nmx s_nm.n_un.n_strx /* string table index */ 206 #define s_type s_nm.n_type /* type of the symbol */ 207 #define s_other s_nm.n_other /* other information for sdb */ 208 #define s_desc s_nm.n_desc /* type descriptor */ 209 #define s_value s_nm.n_value /* value of the symbol, or sdb delta */ 210 211 struct instab{ 212 struct nlist s_nm; /* instruction name, type (opcode) */ 213 u_char s_tag; 214 u_char s_eopcode; 215 char s_pad[2]; /* round to 20 bytes */ 216 }; 217 typedef struct instab *Iptr; 218 /* 219 * The fields nm.n_desc and nm.n_value total 6 bytes; this is 220 * just enough for the 6 bytes describing the argument types. 221 * We use a macro to define access to these 6 bytes, assuming that 222 * they are allocated adjacently. 223 * IF THE FORMAT OF STRUCT nlist CHANGES, THESE MAY HAVE TO BE CHANGED. 224 * 225 * Instab is cleverly declared to look very much like the combination of 226 * a struct symtab and a struct nlist. 227 */ 228 /* 229 * With the 1981 VAX architecture reference manual, 230 * DEC defined and named two byte opcodes. 231 * In addition, DEC defined four new one byte instructions for 232 * queue manipulation. 233 * The assembler was patched in 1982 to reflect this change. 234 * 235 * The two byte opcodes are preceded with an escape byte 236 * (usually an ESCD) and an opcode byte. 237 * For one byte opcodes, the opcode is called the primary opcode. 238 * For two byte opcodes, the second opcode is called the primary opcode. 239 * 240 * We store the primary opcode in I_popcode, 241 * and the escape opcode in I_eopcode. 242 * 243 * For one byte opcodes in the basic arhitecture, 244 * I_eopcode is CORE 245 * For one byte opcodes in the new architecture definition, 246 * I_eopcode is NEW 247 * For the two byte opcodes, I_eopcode is the escape byte. 248 * 249 * The assembler checks if a NEW or two byte opcode is used, 250 * and issues a warning diagnostic. 251 */ 252 /* 253 * For upward compatability reasons, we can't have the two opcodes 254 * forming an operator specifier byte(s) be physically adjacent 255 * in the instruction table. 256 * We define a structure and a constructor that is used in 257 * the instruction generator. 258 */ 259 struct Opcode{ 260 u_char Op_eopcode; 261 u_char Op_popcode; 262 }; 263 264 #define BADPOINT 0xAAAAAAAA 265 /* 266 * See if a structured opcode is bad 267 */ 268 #define ITABCHECK(o) ((itab[o.Op_eopcode] != (Iptr*)BADPOINT) && (itab[o.Op_eopcode][o.Op_popcode] != (Iptr)BADPOINT)) 269 /* 270 * Index the itab by a structured opcode 271 */ 272 #define ITABFETCH(o) itab[o.Op_eopcode][o.Op_popcode] 273 274 struct Instab{ 275 char *I_name; 276 u_char I_popcode; /* basic op code */ 277 char I_nargs; 278 char I_args[6]; 279 u_char I_s_tag; 280 u_char I_eopcode; 281 char I_pad[2]; /* round to 20 bytes */ 282 }; 283 /* 284 * Redefinitions of fields in the struct nlist for instructions so that 285 * one saves typing, and conforms to the old naming conventions 286 */ 287 #define i_popcode s_nm.n_type /* use the same field as symtab.type */ 288 #define i_eopcode s_eopcode 289 #define i_nargs s_nm.n_other /* number of arguments */ 290 #define fetcharg(ptr, n) ((struct Instab *)ptr)->I_args[n] 291 292 struct arg { /*one argument to an instruction*/ 293 char a_atype; 294 char a_areg1; 295 char a_areg2; 296 char a_dispsize; /*usually d124, unless have B^, etc*/ 297 struct exp *a_xp; 298 }; 299 /* 300 * Definitions for numbers and expressions. 301 */ 302 #include "asnumber.h" 303 struct exp { 304 Bignum e_number; /* 128 bits of #, plus tag */ 305 char e_xtype; 306 char e_xloc; 307 struct symtab *e_xname; 308 }; 309 #define e_xvalue e_number.num_num.numIl_int.Il_long 310 311 #define MINLIT 0 312 #define MAXLIT 63 313 314 #define MINBYTE -128 315 #define MAXBYTE 127 316 #define MINUBYTE 0 317 #define MAXUBYTE 255 318 319 #define MINWORD -32768 320 #define MAXWORD 32767 321 #define MINUWORD 0 322 #define MAXUWORD 65535 323 324 #define ISLIT(x) (((x) >= MINLIT) && ((x) <= MAXLIT)) 325 #define ISBYTE(x) (((x) >= MINBYTE) && ((x) <= MAXBYTE)) 326 #define ISUBYTE(x) (((x) >= MINUBYTE) && ((x) <= MAXUBYTE)) 327 #define ISWORD(x) (((x) >= MINWORD) && ((x) <= MAXWORD)) 328 #define ISUWORD(x) (((x) >= MINUWORD) && ((x) <= MAXUWORD)) 329 /* 330 * Definitions for strings. 331 * 332 * Strings are stored in the string pool; see strsave(str, length) 333 * Strings are known by their length and values. 334 * A string pointer points to the beginning of the value bytes; 335 * 336 * If this structure is changed, change insts also. 337 */ 338 struct strdesc{ 339 int sd_stroff; /* offset into string file */ 340 short sd_place; /* where string is */ 341 u_short sd_strlen; /* string length */ 342 char sd_string[1]; /* the string itself, flexible length */ 343 }; 344 /* 345 * Where a string can be. If these are changed, also change instrs. 346 */ 347 #define STR_FILE 0x1 348 #define STR_CORE 0x2 349 #define STR_BOTH 0x3 350 351 struct strdesc *savestr(); 352 353 /* 354 * Global variables 355 */ 356 extern struct arg arglist[NARG]; /*building operands in instructions*/ 357 extern struct exp explist[NEXP]; /*building up a list of expressions*/ 358 extern struct exp *xp; /*current free expression*/ 359 /* 360 * Communication between the scanner and the jxxx handlers. 361 * lastnam: the last name seen on the input 362 * lastjxxx: pointer to the last symbol table entry for 363 * a jump from 364 */ 365 extern struct symtab *lastnam; 366 extern struct symtab *lastjxxx; 367 /* 368 * Lgensym is used to make up funny names for local labels. 369 * lgensym[i] is the current funny number to put after 370 * references to if, lgensym[i]-1 is for ib. 371 * genref[i] is set when the label is referenced before 372 * it is defined (i.e. 2f) so that we can be sure these 373 * labels are always defined to avoid weird diagnostics 374 * from the loader later. 375 */ 376 extern int lgensym[10]; 377 extern char genref[10]; 378 379 extern struct exp *dotp; /* the current dot location */ 380 extern int loctr; 381 382 extern struct exec hdr; /* a.out header */ 383 extern u_long tsize; /* total text size */ 384 extern u_long dsize; /* total data size */ 385 extern u_long trsize; /* total text relocation size */ 386 extern u_long drsize; /* total data relocation size */ 387 extern u_long datbase; /* base of the data segment */ 388 /* 389 * Bitoff and bitfield keep track of the packing into 390 * bytes mandated by the expression syntax <expr> ':' <expr> 391 */ 392 extern int bitoff; 393 extern long bitfield; 394 395 /* 396 * The lexical analyzer builds up symbols in yytext. Lookup 397 * expects its argument in this buffer 398 */ 399 extern char yytext[NCPName+2]; /* text buffer for lexical */ 400 /* 401 * Variables to manage the input assembler source file 402 */ 403 extern int lineno; /*the line number*/ 404 extern char *dotsname; /*the name of the as source*/ 405 406 extern FILE *tokfile; /* temp token communication*/ 407 extern FILE *strfile; /* temp string file*/ 408 extern char tokfilename[TNAMESIZE]; /* token file name */ 409 extern char strfilename[TNAMESIZE]; /* string file name */ 410 extern int strfilepos; /* position in string file */ 411 412 extern int passno; /* 1 or 2 */ 413 414 extern int anyerrs; /*errors as'ing arguments*/ 415 extern int anywarnings; /*warnings as'ing arguments*/ 416 extern int silent; /*don't mention the errors*/ 417 extern int savelabels; /*save labels in a.out*/ 418 extern int orgwarn; /* questionable origin ? */ 419 extern int useVM; /*use virtual memory temp file*/ 420 extern int jxxxJUMP; /*use jmp instead of brw for jxxx */ 421 extern int readonlydata; /*initialized data into text space*/ 422 extern int nGHnumbers; /* GH numbers used */ 423 extern int nGHopcodes; /* GH opcodes used */ 424 extern int nnewopcodes; /* new opcodes used */ 425 #ifdef DEBUG 426 extern int debug; 427 extern int toktrace; 428 #endif 429 /* 430 * Information about the instructions 431 */ 432 extern struct instab **itab[NINST]; /*maps opcodes to instructions*/ 433 extern readonly struct Instab instab[]; 434 435 extern int curlen; /*current literal storage size*/ 436 extern int d124; /*current pointer storage size*/ 437 438 struct symtab **lookup(); /*argument in yytext*/ 439 struct symtab *symalloc(); 440 441 char *Calloc(); 442 char *ClearCalloc(); 443 444 #define outb(val) {dotp->e_xvalue++; if (passno==2) bputc((val), (txtfil));} 445 446 #define outs(cp, lg) dotp->e_xvalue += (lg); if (passno == 2) bwrite((cp), (lg), (txtfil)) 447 448 #define Outb(o) outb(o) 449 /* 450 * Most of the time, the argument to flushfield is a power of two constant, 451 * the calculations involving it can be optimized to shifts. 452 */ 453 #define flushfield(n) if (bitoff != 0) Flushfield( ( (bitoff+n-1) /n ) * n) 454 455 /* 456 * The biobuf structure and associated routines are used to write 457 * into one file at several places concurrently. Calling bopen 458 * with a biobuf structure sets it up to write ``biofd'' starting 459 * at the specified offset. You can then use ``bwrite'' and/or ``bputc'' 460 * to stuff characters in the stream, much like ``fwrite'' and ``fputc''. 461 * Calling bflush drains all the buffers and MUST be done before exit. 462 */ 463 struct biobuf { 464 short b_nleft; /* Number free spaces left in b_buf */ 465 /* Initialize to be less than BUFSIZ initially, to boundary align in file */ 466 char *b_ptr; /* Next place to stuff characters */ 467 char b_buf[BUFSIZ]; /* The buffer itself */ 468 off_t b_off; /* Current file offset */ 469 struct biobuf *b_link; /* Link in chain for bflush() */ 470 }; 471 #define bputc(c,b) ((b)->b_nleft ? (--(b)->b_nleft, *(b)->b_ptr++ = (c)) \ 472 : bflushc(b, c)) 473 #define BFILE struct biobuf 474 475 extern BFILE *biobufs; /* head of the block I/O buffer chain */ 476 extern int biofd; /* file descriptor for block I/O file */ 477 extern off_t boffset; /* physical position in logical file */ 478 479 /* 480 * For each of the named .text .data segments 481 * (introduced by .text <expr>), we maintain 482 * the current value of the dot, and the BFILE where 483 * the information for each of the segments is placed 484 * during the second pass. 485 */ 486 extern struct exp usedot[NLOC + NLOC]; 487 extern BFILE *usefile[NLOC + NLOC]; 488 extern BFILE *txtfil;/* file for text and data: into usefile */ 489 /* 490 * Relocation information for each segment is accumulated 491 * seperately from the others. Writing the relocation 492 * information is logically viewed as writing to one 493 * relocation saving file for each segment; physically 494 * we have a bunch of buffers allocated internally that 495 * contain the relocation information. 496 */ 497 struct relbufdesc *rusefile[NLOC + NLOC]; 498 struct relbufdesc *relfil; 499