| 1 | /* Copyright (c) 1980 Regents of the University of California */ |
| 2 | /* "@(#)as.h 4.8 %G%" */ |
| 3 | #ifdef VMS |
| 4 | # define vax 1 |
| 5 | # define VAX 1 |
| 6 | #endif VMS |
| 7 | |
| 8 | #include <sys/types.h> |
| 9 | #ifdef UNIX |
| 10 | |
| 11 | #ifdef FLEXNAMES |
| 12 | # include <a.out.h> |
| 13 | # include <stab.h> |
| 14 | #else not FLEXNAMES |
| 15 | # define ONLIST |
| 16 | # include "a.out.h" |
| 17 | # include <stab.h> |
| 18 | #endif FLEXNAMES |
| 19 | |
| 20 | #endif UNIX |
| 21 | #ifdef VMS |
| 22 | |
| 23 | #ifdef UNIXDEVEL |
| 24 | # include <a.out.h> |
| 25 | #else not UNIXDEVEL |
| 26 | # include <aout.h> |
| 27 | #endif not UNIXDEVEL |
| 28 | |
| 29 | #endif VMS |
| 30 | |
| 31 | #define readonly |
| 32 | #define NINST 300 |
| 33 | |
| 34 | #define NEXP 20 /* max number of expr. terms per instruction */ |
| 35 | #define NARG 6 /* max number of args per instruction */ |
| 36 | #define NHASH 1103 /* hash table is dynamically extended */ |
| 37 | #define TNAMESIZE 32 /* maximum length of temporary file names */ |
| 38 | #define NLOC 4 /* number of location ctrs */ |
| 39 | |
| 40 | #ifdef UNIX |
| 41 | # ifndef FLEXNAMES |
| 42 | # ifndef NCPS |
| 43 | # define NCPS 8 /* number of characters per symbol*/ |
| 44 | # endif |
| 45 | # else |
| 46 | # ifdef NCPS |
| 47 | # undef NCPS |
| 48 | # endif |
| 49 | # define NCPS BUFSIZ /* needed to allocate yytext */ |
| 50 | # endif |
| 51 | # endif UNIX |
| 52 | |
| 53 | # ifdef VMS |
| 54 | # ifdef NCPS |
| 55 | # undef NCPS |
| 56 | # endif NCPS |
| 57 | # define NCPS 15 |
| 58 | # endif VMS |
| 59 | |
| 60 | /* |
| 61 | * Symbol types |
| 62 | */ |
| 63 | #define XUNDEF 0x0 |
| 64 | #define XABS 0x2 |
| 65 | #define XTEXT 0x4 |
| 66 | #define XDATA 0x6 |
| 67 | #define XBSS 0x8 |
| 68 | |
| 69 | #define XXTRN 0x1 |
| 70 | #define XTYPE 0x1E |
| 71 | |
| 72 | #define XFORW 0x20 /* Was forward-referenced when undefined */ |
| 73 | |
| 74 | #define ERR (-1) |
| 75 | #define NBPW 32 /* Bits per word */ |
| 76 | |
| 77 | #define AMASK 017 |
| 78 | |
| 79 | /* |
| 80 | * Actual argument syntax types |
| 81 | */ |
| 82 | #define AREG 1 /* %r */ |
| 83 | #define ABASE 2 /* (%r) */ |
| 84 | #define ADECR 3 /* -(%r) */ |
| 85 | #define AINCR 4 /* (%r)+ */ |
| 86 | #define ADISP 5 /* expr(%r) */ |
| 87 | #define AEXP 6 /* expr */ |
| 88 | #define AIMM 7 /* $ expr */ |
| 89 | #define ASTAR 8 /* * */ |
| 90 | #define AINDX 16 /* [%r] */ |
| 91 | |
| 92 | /* |
| 93 | * Argument access types used to test validity of operands to operators |
| 94 | */ |
| 95 | #define ACCR (1<<3) /* read */ |
| 96 | #define ACCW (2<<3) /* write */ |
| 97 | #define ACCB (4<<3) /* branch displacement */ |
| 98 | #define ACCA (8<<3) /* address only */ |
| 99 | #define ACCM (ACCR | ACCW) /* modify */ |
| 100 | #define ACCI (ACCB | ACCR) /* XFC code */ |
| 101 | |
| 102 | #define ACCESSMASK (ACCA | ACCR | ACCW | ACCB) /* the mask */ |
| 103 | |
| 104 | /* |
| 105 | * Argument data types |
| 106 | * Also used to tell outrel what it is relocating |
| 107 | * (possibly in combination with RELOC_PCREL and TYPNONE) |
| 108 | */ |
| 109 | #define TYPB 0 /* byte */ |
| 110 | #define TYPW 1 /* word */ |
| 111 | #define TYPL 2 /* long */ |
| 112 | #define TYPQ 3 /* quad */ |
| 113 | #define TYPF 4 /* floating */ |
| 114 | #define TYPD 5 /* double floating */ |
| 115 | #define TYPNONE 6 /* when nothing */ |
| 116 | #define RELOC_PCREL 8 /* implicit argument to outrel; ==> PCREL */ |
| 117 | |
| 118 | #define TYPMASK 7 |
| 119 | |
| 120 | /* |
| 121 | * reference types for loader |
| 122 | */ |
| 123 | #define PCREL 1 |
| 124 | #define LEN1 2 |
| 125 | #define LEN2 4 |
| 126 | #define LEN4 6 |
| 127 | #define LEN8 8 |
| 128 | |
| 129 | extern int reflen[]; /* {LEN*+PCREL} ==> number of bytes */ |
| 130 | extern int lgreflen[]; /* {LEN*+PCREL} ==> lg number of bytes */ |
| 131 | extern int len124[]; /* {1,2,4,8} ==> {LEN1, LEN2, LEN4, LEN8} */ |
| 132 | extern char mod124[]; /* {1,2,4,8} ==> {bits to construct operands */ |
| 133 | extern int type_124[]; /* {1,2,4,8} ==> {TYPB, TYPW, TYPL, TYPQ} */ |
| 134 | extern int ty_NORELOC[]; /* {TYPB..TYPD} ==> {1 if relocation not OK */ |
| 135 | extern int ty_LEN[]; /* {TYPB..TYPD} ==> {LEN1..LEN8} */ |
| 136 | extern int ty_nbyte[]; /* {TYPB..TYPD} ==> {1,2,4,8} */ |
| 137 | extern int ty_nlg[]; /* {TYPB..TYPD} ==> lg{1,2,4,8} */ |
| 138 | |
| 139 | #define TMPC 7 |
| 140 | #define HW 01 |
| 141 | #define FW 03 |
| 142 | #define DW 07 |
| 143 | |
| 144 | #ifdef UNIX |
| 145 | # include <pagsiz.h> |
| 146 | #endif UNIX |
| 147 | |
| 148 | #ifdef VMS |
| 149 | # define PAGRND 0x1FFL |
| 150 | #endif VMS |
| 151 | |
| 152 | #define round(x,y) (((x)+(y)) & ~(y)) |
| 153 | |
| 154 | #define STABTYPS 0340 |
| 155 | #define STABFLAG 0200 |
| 156 | |
| 157 | /* |
| 158 | * Follows are the definitions for the symbol table tags, which are |
| 159 | * all unsigned characters.. |
| 160 | * High value tags are generated by the asembler for internal |
| 161 | * use. |
| 162 | * Low valued tags are the parser coded tokens the scanner returns. |
| 163 | * There are several pertinant bounds in this ordering: |
| 164 | * a) Symbols greater than JXQUESTIONABLE |
| 165 | * are used by the jxxx bumper, indicating that |
| 166 | * the symbol table entry is a jxxx entry |
| 167 | * that has yet to be bumped. |
| 168 | * b) Symbols greater than IGNOREBOUND are not |
| 169 | * bequeathed to the loader; they are truly |
| 170 | * for assembler internal use only. |
| 171 | * c) Symbols greater than OKTOBUMP represent |
| 172 | * indices into the program text that should |
| 173 | * be changed in preceeding jumps or aligns |
| 174 | * must get turned into their long form. |
| 175 | */ |
| 176 | |
| 177 | #define TAGMASK 0xFF |
| 178 | |
| 179 | # define JXACTIVE 0xFF /*jxxx size unknown*/ |
| 180 | # define JXNOTYET 0xFE /*jxxx size known, but not yet expanded*/ |
| 181 | # define JXALIGN 0xFD /*align jxxx entry*/ |
| 182 | # define JXINACTIVE 0xFC /*jxxx size known and expanded*/ |
| 183 | |
| 184 | #define JXQUESTIONABLE 0xFB |
| 185 | |
| 186 | # define JXTUNNEL 0xFA /*jxxx that jumps to another*/ |
| 187 | # define OBSOLETE 0xF9 /*erroneously entered symbol*/ |
| 188 | |
| 189 | #define IGNOREBOUND 0xF8 /*symbols greater than this are ignored*/ |
| 190 | # define STABFLOATING 0xF7 |
| 191 | # define LABELID 0xF6 |
| 192 | |
| 193 | #define OKTOBUMP 0xF5 |
| 194 | # define STABFIXED 0xF4 |
| 195 | |
| 196 | /* |
| 197 | * astoks.h contains reserved word codings the parser should |
| 198 | * know about |
| 199 | */ |
| 200 | #include "astoks.h" |
| 201 | |
| 202 | /* |
| 203 | * The structure for one symbol table entry. |
| 204 | * Symbol table entries are used for both user defined symbols, |
| 205 | * and symbol slots generated to create the jxxx jump from |
| 206 | * slots. |
| 207 | * Caution: the instructions are stored in a shorter version |
| 208 | * of the struct symtab, using all fields in sym_nm and |
| 209 | * tag. The fields used in sym_nm are carefully redeclared |
| 210 | * in struct Instab and struct instab (see below). |
| 211 | * If struct nlist gets changed, then Instab and instab may |
| 212 | * have to be changed. |
| 213 | */ |
| 214 | |
| 215 | struct symtab{ |
| 216 | struct nlist s_nm; |
| 217 | u_char s_tag; /* assembler tag */ |
| 218 | u_char s_ptype; /* if tag == NAME */ |
| 219 | u_char s_jxoveralign; /* if a JXXX, jumped over align */ |
| 220 | short s_index; /* which segment */ |
| 221 | struct symtab *s_dest; /* if JXXX, where going to */ |
| 222 | #ifdef DJXXX |
| 223 | short s_jxline; /* source line of the jump from */ |
| 224 | #endif |
| 225 | }; |
| 226 | /* |
| 227 | * Redefinitions of the fields in symtab for |
| 228 | * use when the symbol table entry marks a jxxx instruction. |
| 229 | */ |
| 230 | #define s_jxbump s_ptype /* tag == JX..., how far to expand */ |
| 231 | #define s_jxfear s_desc /* how far needs to be bumped */ |
| 232 | /* |
| 233 | * Redefinitions of fields in the struct nlist for symbols so that |
| 234 | * one saves typing, and so that they conform |
| 235 | * with the old naming conventions. |
| 236 | */ |
| 237 | #ifdef FLEXNAMES |
| 238 | #define s_name s_nm.n_un.n_name /* name pointer */ |
| 239 | #define s_nmx s_nm.n_un.n_strx /* string table index */ |
| 240 | #else not FLEXNAMES |
| 241 | #define s_name s_nm.n_name |
| 242 | #endif |
| 243 | #define s_type s_nm.n_type /* type of the symbol */ |
| 244 | #define s_other s_nm.n_other /* other information for sdb */ |
| 245 | #define s_desc s_nm.n_desc /* type descriptor */ |
| 246 | #define s_value s_nm.n_value /* value of the symbol, or sdb delta */ |
| 247 | |
| 248 | struct instab{ |
| 249 | struct nlist s_nm; /* instruction name, type (opcode) */ |
| 250 | u_char s_tag; |
| 251 | char s_pad[3]; /* round to 20 bytes */ |
| 252 | }; |
| 253 | /* |
| 254 | * The fields nm.n_desc and nm.n_value total 6 bytes; this is |
| 255 | * just enough for the 6 bytes describing the argument types. |
| 256 | * We use a macro to define access to these 6 bytes, assuming that |
| 257 | * they are allocated adjacently. |
| 258 | * IF THE FORMAT OF STRUCT nlist CHANGES, THESE MAY HAVE TO BE CHANGED. |
| 259 | * |
| 260 | * Instab is cleverly declared to look very much like the combination of |
| 261 | * a struct symtab and a struct nlist. |
| 262 | */ |
| 263 | struct Instab{ |
| 264 | #ifdef FLEXNAMES |
| 265 | char *I_name; |
| 266 | #else not FLEXNAMES |
| 267 | char I_name[NCPS]; |
| 268 | #endif |
| 269 | u_char I_opcode; |
| 270 | char I_nargs; |
| 271 | char I_args[6]; |
| 272 | u_char I_s_tag; |
| 273 | char I_pad[3]; /* round to 20 bytes */ |
| 274 | }; |
| 275 | /* |
| 276 | * Redefinitions of fields in the struct nlist for instructions so that |
| 277 | * one saves typing, and conforms to the old naming conventions |
| 278 | */ |
| 279 | #define i_opcode s_nm.n_type /* use the same field as symtab.type */ |
| 280 | #define i_nargs s_nm.n_other /* number of arguments */ |
| 281 | #define fetcharg(ptr, n) ((struct Instab *)ptr)->I_args[n] |
| 282 | |
| 283 | struct arg { /*one argument to an instruction*/ |
| 284 | char a_atype; |
| 285 | char a_areg1; |
| 286 | char a_areg2; |
| 287 | char a_dispsize; /*usually d124, unless have B^, etc*/ |
| 288 | struct exp *a_xp; |
| 289 | }; |
| 290 | |
| 291 | struct exp { |
| 292 | long e_xvalue; /* MUST be the first field (look at union Double) */ |
| 293 | long e_yvalue; /* MUST be second field; least sig word of a double */ |
| 294 | char e_xtype; |
| 295 | char e_xloc; |
| 296 | struct symtab *e_xname; |
| 297 | }; |
| 298 | |
| 299 | #define doub_MSW e_xvalue |
| 300 | #define doub_LSW e_yvalue |
| 301 | |
| 302 | union Double { |
| 303 | struct{ |
| 304 | long doub_MSW; |
| 305 | long doub_LSW; |
| 306 | } dis_dvalue; |
| 307 | double dvalue; |
| 308 | }; |
| 309 | |
| 310 | struct Quad { |
| 311 | long quad_low_long; |
| 312 | long quad_high_long; |
| 313 | }; |
| 314 | |
| 315 | /* |
| 316 | * Magic layout macros |
| 317 | */ |
| 318 | #define MINBYTE -128 |
| 319 | #define MAXBYTE 127 |
| 320 | #define MINWORD -32768 |
| 321 | #define MAXWORD 32767 |
| 322 | |
| 323 | #define LITFLTMASK 0x000043F0 /*really magic*/ |
| 324 | /* |
| 325 | * Is the floating point double word in xp a |
| 326 | * short literal floating point number? |
| 327 | */ |
| 328 | #define slitflt(xp) \ |
| 329 | ( (xp->doub_LSW == 0) \ |
| 330 | && ( (xp->doub_MSW & LITFLTMASK) \ |
| 331 | == xp->doub_MSW) ) |
| 332 | /* |
| 333 | * If it is a slitflt, then extract the 6 interesting bits |
| 334 | */ |
| 335 | #define extlitflt(xp) \ |
| 336 | xp->doub_MSW >> 4 |
| 337 | |
| 338 | extern struct arg arglist[NARG]; /*building operands in instructions*/ |
| 339 | extern struct exp explist[NEXP]; /*building up a list of expressions*/ |
| 340 | extern struct exp *xp; /*current free expression*/ |
| 341 | /* |
| 342 | * Communication between the scanner and the jxxx handlers. |
| 343 | * lastnam: the last name seen on the input |
| 344 | * lastjxxx: pointer to the last symbol table entry for |
| 345 | * a jump from |
| 346 | */ |
| 347 | extern struct symtab *lastnam; |
| 348 | extern struct symtab *lastjxxx; |
| 349 | |
| 350 | #ifdef VMS |
| 351 | extern char *vms_obj_ptr; /* object buffer pointer */ |
| 352 | extern char sobuf[]; /* object buffer */ |
| 353 | extern int objfil; /* VMS object file descriptor */ |
| 354 | #endif VMS |
| 355 | |
| 356 | /* |
| 357 | * Lgensym is used to make up funny names for local labels. |
| 358 | * lgensym[i] is the current funny number to put after |
| 359 | * references to if, lgensym[i]-1 is for ib. |
| 360 | * genref[i] is set when the label is referenced before |
| 361 | * it is defined (i.e. 2f) so that we can be sure these |
| 362 | * labels are always defined to avoid weird diagnostics |
| 363 | * from the loader later. |
| 364 | */ |
| 365 | extern int lgensym[10]; |
| 366 | extern char genref[10]; |
| 367 | |
| 368 | extern char tmpn1[TNAMESIZE]; /* Interpass temporary */ |
| 369 | extern struct exp *dotp; /* the current dot location */ |
| 370 | extern int loctr; |
| 371 | |
| 372 | extern struct exec hdr; /* a.out header */ |
| 373 | extern u_long tsize; /* total text size */ |
| 374 | extern u_long dsize; /* total data size */ |
| 375 | extern u_long trsize; /* total text relocation size */ |
| 376 | extern u_long drsize; /* total data relocation size */ |
| 377 | extern u_long datbase; /* base of the data segment */ |
| 378 | /* |
| 379 | * Bitoff and bitfield keep track of the packing into |
| 380 | * bytes mandated by the expression syntax <expr> ':' <expr> |
| 381 | */ |
| 382 | extern int bitoff; |
| 383 | extern long bitfield; |
| 384 | |
| 385 | /* |
| 386 | * The lexical analyzer builds up symbols in yytext. Lookup |
| 387 | * expects its argument in this buffer |
| 388 | */ |
| 389 | extern char yytext[NCPS+2]; /* text buffer for lexical */ |
| 390 | /* |
| 391 | * Variables to manage the input assembler source file |
| 392 | */ |
| 393 | extern int lineno; /*the line number*/ |
| 394 | extern char *dotsname; /*the name of the as source*/ |
| 395 | |
| 396 | extern FILE *tmpfil; /* interpass communication*/ |
| 397 | |
| 398 | extern int passno; /* 1 or 2 */ |
| 399 | |
| 400 | extern int anyerrs; /*errors assembling arguments*/ |
| 401 | extern int silent; /*don't mention the errors*/ |
| 402 | extern int savelabels; /*save labels in a.out*/ |
| 403 | extern int orgwarn; /* questionable origin ? */ |
| 404 | extern int useVM; /*use virtual memory temp file*/ |
| 405 | extern int jxxxJUMP; /*use jmp instead of brw for jxxx */ |
| 406 | extern int readonlydata; /*initialized data into text space*/ |
| 407 | #ifdef DEBUG |
| 408 | extern int debug; |
| 409 | extern int toktrace; |
| 410 | #endif |
| 411 | /* |
| 412 | * Information about the instructions |
| 413 | */ |
| 414 | extern struct instab *itab[NINST]; /*maps opcodes to instructions*/ |
| 415 | extern readonly struct Instab instab[]; |
| 416 | |
| 417 | extern int curlen; /*current literal storage size*/ |
| 418 | extern int d124; /*current pointer storage size*/ |
| 419 | |
| 420 | struct symtab **lookup(); /*argument in yytext*/ |
| 421 | struct symtab *symalloc(); |
| 422 | |
| 423 | #define outb(val) {dotp->e_xvalue++; if (passno==2) bputc((val), (txtfil));} |
| 424 | |
| 425 | #define outs(cp, lg) dotp->e_xvalue += (lg); if (passno == 2) bwrite((cp), (lg), (txtfil)) |
| 426 | |
| 427 | /* |
| 428 | * Most of the time, the argument to flushfield is a power of two constant, |
| 429 | * the calculations involving it can be optimized to shifts. |
| 430 | */ |
| 431 | #define flushfield(n) if (bitoff != 0) Flushfield( ( (bitoff+n-1) /n ) * n) |
| 432 | |
| 433 | /* |
| 434 | * The biobuf structure and associated routines are used to write |
| 435 | * into one file at several places concurrently. Calling bopen |
| 436 | * with a biobuf structure sets it up to write ``biofd'' starting |
| 437 | * at the specified offset. You can then use ``bwrite'' and/or ``bputc'' |
| 438 | * to stuff characters in the stream, much like ``fwrite'' and ``fputc''. |
| 439 | * Calling bflush drains all the buffers and MUST be done before exit. |
| 440 | */ |
| 441 | struct biobuf { |
| 442 | short b_nleft; /* Number free spaces left in b_buf */ |
| 443 | /* Initialize to be less than BUFSIZ initially, to boundary align in file */ |
| 444 | char *b_ptr; /* Next place to stuff characters */ |
| 445 | char b_buf[BUFSIZ]; /* The buffer itself */ |
| 446 | off_t b_off; /* Current file offset */ |
| 447 | struct biobuf *b_link; /* Link in chain for bflush() */ |
| 448 | }; |
| 449 | #define bputc(c,b) ((b)->b_nleft ? (--(b)->b_nleft, *(b)->b_ptr++ = (c)) \ |
| 450 | : bflushc(b, c)) |
| 451 | #define BFILE struct biobuf |
| 452 | |
| 453 | extern BFILE *biobufs; /* head of the block I/O buffer chain */ |
| 454 | extern int biofd; /* file descriptor for block I/O file */ |
| 455 | extern off_t boffset; /* physical position in logical file */ |
| 456 | |
| 457 | /* |
| 458 | * For each of the named .text .data segments |
| 459 | * (introduced by .text <expr>), we maintain |
| 460 | * the current value of the dot, and the BFILE where |
| 461 | * the information for each of the segments is placed |
| 462 | * during the second pass. |
| 463 | */ |
| 464 | extern struct exp usedot[NLOC + NLOC]; |
| 465 | extern BFILE *usefile[NLOC + NLOC]; |
| 466 | extern BFILE *txtfil;/* file for text and data: into usefile */ |
| 467 | /* |
| 468 | * Relocation information for each segment is accumulated |
| 469 | * seperately from the others. Writing the relocation |
| 470 | * information is logically viewed as writing to one |
| 471 | * relocation saving file for each segment; physically |
| 472 | * we have a bunch of buffers allocated internally that |
| 473 | * contain the relocation information. |
| 474 | */ |
| 475 | struct relbufdesc *rusefile[NLOC + NLOC]; |
| 476 | struct relbufdesc *relfil; |