optimize.c revision 98530
1/* 2 * Copyright (c) 1988, 1989, 1990, 1991, 1993, 1994, 1995, 1996 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that: (1) source code distributions 7 * retain the above copyright notice and this paragraph in its entirety, (2) 8 * distributions including binary code include the above copyright notice and 9 * this paragraph in its entirety in the documentation or other materials 10 * provided with the distribution, and (3) all advertising materials mentioning 11 * features or use of this software display the following acknowledgement: 12 * ``This product includes software developed by the University of California, 13 * Lawrence Berkeley Laboratory and its contributors.'' Neither the name of 14 * the University nor the names of its contributors may be used to endorse 15 * or promote products derived from this software without specific prior 16 * written permission. 17 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED 18 * WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF 19 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. 20 * 21 * Optimization module for tcpdump intermediate representation. 22 */ 23#ifndef lint 24static const char rcsid[] = 25 "@(#) $Header: /tcpdump/master/libpcap/optimize.c,v 1.69 2001/11/12 21:57:06 fenner Exp $ (LBL)"; 26#endif 27 28#ifdef HAVE_CONFIG_H 29#include "config.h" 30#endif 31 32#include <sys/types.h> 33#include <sys/time.h> 34 35#include <stdio.h> 36#include <stdlib.h> 37#include <memory.h> 38 39#include <errno.h> 40 41#include "pcap-int.h" 42 43#include "gencode.h" 44 45#ifdef HAVE_OS_PROTO_H 46#include "os-proto.h" 47#endif 48 49#ifdef BDEBUG 50extern int dflag; 51#endif 52 53#define A_ATOM BPF_MEMWORDS 54#define X_ATOM (BPF_MEMWORDS+1) 55 56#define NOP -1 57 58/* 59 * This define is used to represent *both* the accumulator and 60 * x register in use-def computations. 61 * Currently, the use-def code assumes only one definition per instruction. 62 */ 63#define AX_ATOM N_ATOMS 64 65/* 66 * A flag to indicate that further optimization is needed. 67 * Iterative passes are continued until a given pass yields no 68 * branch movement. 69 */ 70static int done; 71 72/* 73 * A block is marked if only if its mark equals the current mark. 74 * Rather than traverse the code array, marking each item, 'cur_mark' is 75 * incremented. This automatically makes each element unmarked. 76 */ 77static int cur_mark; 78#define isMarked(p) ((p)->mark == cur_mark) 79#define unMarkAll() cur_mark += 1 80#define Mark(p) ((p)->mark = cur_mark) 81 82static void opt_init(struct block *); 83static void opt_cleanup(void); 84 85static void make_marks(struct block *); 86static void mark_code(struct block *); 87 88static void intern_blocks(struct block *); 89 90static int eq_slist(struct slist *, struct slist *); 91 92static void find_levels_r(struct block *); 93 94static void find_levels(struct block *); 95static void find_dom(struct block *); 96static void propedom(struct edge *); 97static void find_edom(struct block *); 98static void find_closure(struct block *); 99static int atomuse(struct stmt *); 100static int atomdef(struct stmt *); 101static void compute_local_ud(struct block *); 102static void find_ud(struct block *); 103static void init_val(void); 104static int F(int, int, int); 105static inline void vstore(struct stmt *, int *, int, int); 106static void opt_blk(struct block *, int); 107static int use_conflict(struct block *, struct block *); 108static void opt_j(struct edge *); 109static void or_pullup(struct block *); 110static void and_pullup(struct block *); 111static void opt_blks(struct block *, int); 112static inline void link_inedge(struct edge *, struct block *); 113static void find_inedges(struct block *); 114static void opt_root(struct block **); 115static void opt_loop(struct block *, int); 116static void fold_op(struct stmt *, int, int); 117static inline struct slist *this_op(struct slist *); 118static void opt_not(struct block *); 119static void opt_peep(struct block *); 120static void opt_stmt(struct stmt *, int[], int); 121static void deadstmt(struct stmt *, struct stmt *[]); 122static void opt_deadstores(struct block *); 123static void opt_blk(struct block *, int); 124static int use_conflict(struct block *, struct block *); 125static void opt_j(struct edge *); 126static struct block *fold_edge(struct block *, struct edge *); 127static inline int eq_blk(struct block *, struct block *); 128static int slength(struct slist *); 129static int count_blocks(struct block *); 130static void number_blks_r(struct block *); 131static int count_stmts(struct block *); 132static int convert_code_r(struct block *); 133#ifdef BDEBUG 134static void opt_dump(struct block *); 135#endif 136 137static int n_blocks; 138struct block **blocks; 139static int n_edges; 140struct edge **edges; 141 142/* 143 * A bit vector set representation of the dominators. 144 * We round up the set size to the next power of two. 145 */ 146static int nodewords; 147static int edgewords; 148struct block **levels; 149bpf_u_int32 *space; 150#define BITS_PER_WORD (8*sizeof(bpf_u_int32)) 151/* 152 * True if a is in uset {p} 153 */ 154#define SET_MEMBER(p, a) \ 155((p)[(unsigned)(a) / BITS_PER_WORD] & (1 << ((unsigned)(a) % BITS_PER_WORD))) 156 157/* 158 * Add 'a' to uset p. 159 */ 160#define SET_INSERT(p, a) \ 161(p)[(unsigned)(a) / BITS_PER_WORD] |= (1 << ((unsigned)(a) % BITS_PER_WORD)) 162 163/* 164 * Delete 'a' from uset p. 165 */ 166#define SET_DELETE(p, a) \ 167(p)[(unsigned)(a) / BITS_PER_WORD] &= ~(1 << ((unsigned)(a) % BITS_PER_WORD)) 168 169/* 170 * a := a intersect b 171 */ 172#define SET_INTERSECT(a, b, n)\ 173{\ 174 register bpf_u_int32 *_x = a, *_y = b;\ 175 register int _n = n;\ 176 while (--_n >= 0) *_x++ &= *_y++;\ 177} 178 179/* 180 * a := a - b 181 */ 182#define SET_SUBTRACT(a, b, n)\ 183{\ 184 register bpf_u_int32 *_x = a, *_y = b;\ 185 register int _n = n;\ 186 while (--_n >= 0) *_x++ &=~ *_y++;\ 187} 188 189/* 190 * a := a union b 191 */ 192#define SET_UNION(a, b, n)\ 193{\ 194 register bpf_u_int32 *_x = a, *_y = b;\ 195 register int _n = n;\ 196 while (--_n >= 0) *_x++ |= *_y++;\ 197} 198 199static uset all_dom_sets; 200static uset all_closure_sets; 201static uset all_edge_sets; 202 203#ifndef MAX 204#define MAX(a,b) ((a)>(b)?(a):(b)) 205#endif 206 207static void 208find_levels_r(b) 209 struct block *b; 210{ 211 int level; 212 213 if (isMarked(b)) 214 return; 215 216 Mark(b); 217 b->link = 0; 218 219 if (JT(b)) { 220 find_levels_r(JT(b)); 221 find_levels_r(JF(b)); 222 level = MAX(JT(b)->level, JF(b)->level) + 1; 223 } else 224 level = 0; 225 b->level = level; 226 b->link = levels[level]; 227 levels[level] = b; 228} 229 230/* 231 * Level graph. The levels go from 0 at the leaves to 232 * N_LEVELS at the root. The levels[] array points to the 233 * first node of the level list, whose elements are linked 234 * with the 'link' field of the struct block. 235 */ 236static void 237find_levels(root) 238 struct block *root; 239{ 240 memset((char *)levels, 0, n_blocks * sizeof(*levels)); 241 unMarkAll(); 242 find_levels_r(root); 243} 244 245/* 246 * Find dominator relationships. 247 * Assumes graph has been leveled. 248 */ 249static void 250find_dom(root) 251 struct block *root; 252{ 253 int i; 254 struct block *b; 255 bpf_u_int32 *x; 256 257 /* 258 * Initialize sets to contain all nodes. 259 */ 260 x = all_dom_sets; 261 i = n_blocks * nodewords; 262 while (--i >= 0) 263 *x++ = ~0; 264 /* Root starts off empty. */ 265 for (i = nodewords; --i >= 0;) 266 root->dom[i] = 0; 267 268 /* root->level is the highest level no found. */ 269 for (i = root->level; i >= 0; --i) { 270 for (b = levels[i]; b; b = b->link) { 271 SET_INSERT(b->dom, b->id); 272 if (JT(b) == 0) 273 continue; 274 SET_INTERSECT(JT(b)->dom, b->dom, nodewords); 275 SET_INTERSECT(JF(b)->dom, b->dom, nodewords); 276 } 277 } 278} 279 280static void 281propedom(ep) 282 struct edge *ep; 283{ 284 SET_INSERT(ep->edom, ep->id); 285 if (ep->succ) { 286 SET_INTERSECT(ep->succ->et.edom, ep->edom, edgewords); 287 SET_INTERSECT(ep->succ->ef.edom, ep->edom, edgewords); 288 } 289} 290 291/* 292 * Compute edge dominators. 293 * Assumes graph has been leveled and predecessors established. 294 */ 295static void 296find_edom(root) 297 struct block *root; 298{ 299 int i; 300 uset x; 301 struct block *b; 302 303 x = all_edge_sets; 304 for (i = n_edges * edgewords; --i >= 0; ) 305 x[i] = ~0; 306 307 /* root->level is the highest level no found. */ 308 memset(root->et.edom, 0, edgewords * sizeof(*(uset)0)); 309 memset(root->ef.edom, 0, edgewords * sizeof(*(uset)0)); 310 for (i = root->level; i >= 0; --i) { 311 for (b = levels[i]; b != 0; b = b->link) { 312 propedom(&b->et); 313 propedom(&b->ef); 314 } 315 } 316} 317 318/* 319 * Find the backwards transitive closure of the flow graph. These sets 320 * are backwards in the sense that we find the set of nodes that reach 321 * a given node, not the set of nodes that can be reached by a node. 322 * 323 * Assumes graph has been leveled. 324 */ 325static void 326find_closure(root) 327 struct block *root; 328{ 329 int i; 330 struct block *b; 331 332 /* 333 * Initialize sets to contain no nodes. 334 */ 335 memset((char *)all_closure_sets, 0, 336 n_blocks * nodewords * sizeof(*all_closure_sets)); 337 338 /* root->level is the highest level no found. */ 339 for (i = root->level; i >= 0; --i) { 340 for (b = levels[i]; b; b = b->link) { 341 SET_INSERT(b->closure, b->id); 342 if (JT(b) == 0) 343 continue; 344 SET_UNION(JT(b)->closure, b->closure, nodewords); 345 SET_UNION(JF(b)->closure, b->closure, nodewords); 346 } 347 } 348} 349 350/* 351 * Return the register number that is used by s. If A and X are both 352 * used, return AX_ATOM. If no register is used, return -1. 353 * 354 * The implementation should probably change to an array access. 355 */ 356static int 357atomuse(s) 358 struct stmt *s; 359{ 360 register int c = s->code; 361 362 if (c == NOP) 363 return -1; 364 365 switch (BPF_CLASS(c)) { 366 367 case BPF_RET: 368 return (BPF_RVAL(c) == BPF_A) ? A_ATOM : 369 (BPF_RVAL(c) == BPF_X) ? X_ATOM : -1; 370 371 case BPF_LD: 372 case BPF_LDX: 373 return (BPF_MODE(c) == BPF_IND) ? X_ATOM : 374 (BPF_MODE(c) == BPF_MEM) ? s->k : -1; 375 376 case BPF_ST: 377 return A_ATOM; 378 379 case BPF_STX: 380 return X_ATOM; 381 382 case BPF_JMP: 383 case BPF_ALU: 384 if (BPF_SRC(c) == BPF_X) 385 return AX_ATOM; 386 return A_ATOM; 387 388 case BPF_MISC: 389 return BPF_MISCOP(c) == BPF_TXA ? X_ATOM : A_ATOM; 390 } 391 abort(); 392 /* NOTREACHED */ 393} 394 395/* 396 * Return the register number that is defined by 's'. We assume that 397 * a single stmt cannot define more than one register. If no register 398 * is defined, return -1. 399 * 400 * The implementation should probably change to an array access. 401 */ 402static int 403atomdef(s) 404 struct stmt *s; 405{ 406 if (s->code == NOP) 407 return -1; 408 409 switch (BPF_CLASS(s->code)) { 410 411 case BPF_LD: 412 case BPF_ALU: 413 return A_ATOM; 414 415 case BPF_LDX: 416 return X_ATOM; 417 418 case BPF_ST: 419 case BPF_STX: 420 return s->k; 421 422 case BPF_MISC: 423 return BPF_MISCOP(s->code) == BPF_TAX ? X_ATOM : A_ATOM; 424 } 425 return -1; 426} 427 428static void 429compute_local_ud(b) 430 struct block *b; 431{ 432 struct slist *s; 433 atomset def = 0, use = 0, kill = 0; 434 int atom; 435 436 for (s = b->stmts; s; s = s->next) { 437 if (s->s.code == NOP) 438 continue; 439 atom = atomuse(&s->s); 440 if (atom >= 0) { 441 if (atom == AX_ATOM) { 442 if (!ATOMELEM(def, X_ATOM)) 443 use |= ATOMMASK(X_ATOM); 444 if (!ATOMELEM(def, A_ATOM)) 445 use |= ATOMMASK(A_ATOM); 446 } 447 else if (atom < N_ATOMS) { 448 if (!ATOMELEM(def, atom)) 449 use |= ATOMMASK(atom); 450 } 451 else 452 abort(); 453 } 454 atom = atomdef(&s->s); 455 if (atom >= 0) { 456 if (!ATOMELEM(use, atom)) 457 kill |= ATOMMASK(atom); 458 def |= ATOMMASK(atom); 459 } 460 } 461 if (!ATOMELEM(def, A_ATOM) && BPF_CLASS(b->s.code) == BPF_JMP) 462 use |= ATOMMASK(A_ATOM); 463 464 b->def = def; 465 b->kill = kill; 466 b->in_use = use; 467} 468 469/* 470 * Assume graph is already leveled. 471 */ 472static void 473find_ud(root) 474 struct block *root; 475{ 476 int i, maxlevel; 477 struct block *p; 478 479 /* 480 * root->level is the highest level no found; 481 * count down from there. 482 */ 483 maxlevel = root->level; 484 for (i = maxlevel; i >= 0; --i) 485 for (p = levels[i]; p; p = p->link) { 486 compute_local_ud(p); 487 p->out_use = 0; 488 } 489 490 for (i = 1; i <= maxlevel; ++i) { 491 for (p = levels[i]; p; p = p->link) { 492 p->out_use |= JT(p)->in_use | JF(p)->in_use; 493 p->in_use |= p->out_use &~ p->kill; 494 } 495 } 496} 497 498/* 499 * These data structures are used in a Cocke and Shwarz style 500 * value numbering scheme. Since the flowgraph is acyclic, 501 * exit values can be propagated from a node's predecessors 502 * provided it is uniquely defined. 503 */ 504struct valnode { 505 int code; 506 int v0, v1; 507 int val; 508 struct valnode *next; 509}; 510 511#define MODULUS 213 512static struct valnode *hashtbl[MODULUS]; 513static int curval; 514static int maxval; 515 516/* Integer constants mapped with the load immediate opcode. */ 517#define K(i) F(BPF_LD|BPF_IMM|BPF_W, i, 0L) 518 519struct vmapinfo { 520 int is_const; 521 bpf_int32 const_val; 522}; 523 524struct vmapinfo *vmap; 525struct valnode *vnode_base; 526struct valnode *next_vnode; 527 528static void 529init_val() 530{ 531 curval = 0; 532 next_vnode = vnode_base; 533 memset((char *)vmap, 0, maxval * sizeof(*vmap)); 534 memset((char *)hashtbl, 0, sizeof hashtbl); 535} 536 537/* Because we really don't have an IR, this stuff is a little messy. */ 538static int 539F(code, v0, v1) 540 int code; 541 int v0, v1; 542{ 543 u_int hash; 544 int val; 545 struct valnode *p; 546 547 hash = (u_int)code ^ (v0 << 4) ^ (v1 << 8); 548 hash %= MODULUS; 549 550 for (p = hashtbl[hash]; p; p = p->next) 551 if (p->code == code && p->v0 == v0 && p->v1 == v1) 552 return p->val; 553 554 val = ++curval; 555 if (BPF_MODE(code) == BPF_IMM && 556 (BPF_CLASS(code) == BPF_LD || BPF_CLASS(code) == BPF_LDX)) { 557 vmap[val].const_val = v0; 558 vmap[val].is_const = 1; 559 } 560 p = next_vnode++; 561 p->val = val; 562 p->code = code; 563 p->v0 = v0; 564 p->v1 = v1; 565 p->next = hashtbl[hash]; 566 hashtbl[hash] = p; 567 568 return val; 569} 570 571static inline void 572vstore(s, valp, newval, alter) 573 struct stmt *s; 574 int *valp; 575 int newval; 576 int alter; 577{ 578 if (alter && *valp == newval) 579 s->code = NOP; 580 else 581 *valp = newval; 582} 583 584static void 585fold_op(s, v0, v1) 586 struct stmt *s; 587 int v0, v1; 588{ 589 bpf_int32 a, b; 590 591 a = vmap[v0].const_val; 592 b = vmap[v1].const_val; 593 594 switch (BPF_OP(s->code)) { 595 case BPF_ADD: 596 a += b; 597 break; 598 599 case BPF_SUB: 600 a -= b; 601 break; 602 603 case BPF_MUL: 604 a *= b; 605 break; 606 607 case BPF_DIV: 608 if (b == 0) 609 bpf_error("division by zero"); 610 a /= b; 611 break; 612 613 case BPF_AND: 614 a &= b; 615 break; 616 617 case BPF_OR: 618 a |= b; 619 break; 620 621 case BPF_LSH: 622 a <<= b; 623 break; 624 625 case BPF_RSH: 626 a >>= b; 627 break; 628 629 case BPF_NEG: 630 a = -a; 631 break; 632 633 default: 634 abort(); 635 } 636 s->k = a; 637 s->code = BPF_LD|BPF_IMM; 638 done = 0; 639} 640 641static inline struct slist * 642this_op(s) 643 struct slist *s; 644{ 645 while (s != 0 && s->s.code == NOP) 646 s = s->next; 647 return s; 648} 649 650static void 651opt_not(b) 652 struct block *b; 653{ 654 struct block *tmp = JT(b); 655 656 JT(b) = JF(b); 657 JF(b) = tmp; 658} 659 660static void 661opt_peep(b) 662 struct block *b; 663{ 664 struct slist *s; 665 struct slist *next, *last; 666 int val; 667 668 s = b->stmts; 669 if (s == 0) 670 return; 671 672 last = s; 673 while (1) { 674 s = this_op(s); 675 if (s == 0) 676 break; 677 next = this_op(s->next); 678 if (next == 0) 679 break; 680 last = next; 681 682 /* 683 * st M[k] --> st M[k] 684 * ldx M[k] tax 685 */ 686 if (s->s.code == BPF_ST && 687 next->s.code == (BPF_LDX|BPF_MEM) && 688 s->s.k == next->s.k) { 689 done = 0; 690 next->s.code = BPF_MISC|BPF_TAX; 691 } 692 /* 693 * ld #k --> ldx #k 694 * tax txa 695 */ 696 if (s->s.code == (BPF_LD|BPF_IMM) && 697 next->s.code == (BPF_MISC|BPF_TAX)) { 698 s->s.code = BPF_LDX|BPF_IMM; 699 next->s.code = BPF_MISC|BPF_TXA; 700 done = 0; 701 } 702 /* 703 * This is an ugly special case, but it happens 704 * when you say tcp[k] or udp[k] where k is a constant. 705 */ 706 if (s->s.code == (BPF_LD|BPF_IMM)) { 707 struct slist *add, *tax, *ild; 708 709 /* 710 * Check that X isn't used on exit from this 711 * block (which the optimizer might cause). 712 * We know the code generator won't generate 713 * any local dependencies. 714 */ 715 if (ATOMELEM(b->out_use, X_ATOM)) 716 break; 717 718 if (next->s.code != (BPF_LDX|BPF_MSH|BPF_B)) 719 add = next; 720 else 721 add = this_op(next->next); 722 if (add == 0 || add->s.code != (BPF_ALU|BPF_ADD|BPF_X)) 723 break; 724 725 tax = this_op(add->next); 726 if (tax == 0 || tax->s.code != (BPF_MISC|BPF_TAX)) 727 break; 728 729 ild = this_op(tax->next); 730 if (ild == 0 || BPF_CLASS(ild->s.code) != BPF_LD || 731 BPF_MODE(ild->s.code) != BPF_IND) 732 break; 733 /* 734 * XXX We need to check that X is not 735 * subsequently used. We know we can eliminate the 736 * accumulator modifications since it is defined 737 * by the last stmt of this sequence. 738 * 739 * We want to turn this sequence: 740 * 741 * (004) ldi #0x2 {s} 742 * (005) ldxms [14] {next} -- optional 743 * (006) addx {add} 744 * (007) tax {tax} 745 * (008) ild [x+0] {ild} 746 * 747 * into this sequence: 748 * 749 * (004) nop 750 * (005) ldxms [14] 751 * (006) nop 752 * (007) nop 753 * (008) ild [x+2] 754 * 755 */ 756 ild->s.k += s->s.k; 757 s->s.code = NOP; 758 add->s.code = NOP; 759 tax->s.code = NOP; 760 done = 0; 761 } 762 s = next; 763 } 764 /* 765 * If we have a subtract to do a comparison, and the X register 766 * is a known constant, we can merge this value into the 767 * comparison. 768 */ 769 if (last->s.code == (BPF_ALU|BPF_SUB|BPF_X) && 770 !ATOMELEM(b->out_use, A_ATOM)) { 771 val = b->val[X_ATOM]; 772 if (vmap[val].is_const) { 773 int op; 774 775 b->s.k += vmap[val].const_val; 776 op = BPF_OP(b->s.code); 777 if (op == BPF_JGT || op == BPF_JGE) { 778 struct block *t = JT(b); 779 JT(b) = JF(b); 780 JF(b) = t; 781 b->s.k += 0x80000000; 782 } 783 last->s.code = NOP; 784 done = 0; 785 } else if (b->s.k == 0) { 786 /* 787 * sub x -> nop 788 * j #0 j x 789 */ 790 last->s.code = NOP; 791 b->s.code = BPF_CLASS(b->s.code) | BPF_OP(b->s.code) | 792 BPF_X; 793 done = 0; 794 } 795 } 796 /* 797 * Likewise, a constant subtract can be simplified. 798 */ 799 else if (last->s.code == (BPF_ALU|BPF_SUB|BPF_K) && 800 !ATOMELEM(b->out_use, A_ATOM)) { 801 int op; 802 803 b->s.k += last->s.k; 804 last->s.code = NOP; 805 op = BPF_OP(b->s.code); 806 if (op == BPF_JGT || op == BPF_JGE) { 807 struct block *t = JT(b); 808 JT(b) = JF(b); 809 JF(b) = t; 810 b->s.k += 0x80000000; 811 } 812 done = 0; 813 } 814 /* 815 * and #k nop 816 * jeq #0 -> jset #k 817 */ 818 if (last->s.code == (BPF_ALU|BPF_AND|BPF_K) && 819 !ATOMELEM(b->out_use, A_ATOM) && b->s.k == 0) { 820 b->s.k = last->s.k; 821 b->s.code = BPF_JMP|BPF_K|BPF_JSET; 822 last->s.code = NOP; 823 done = 0; 824 opt_not(b); 825 } 826 /* 827 * If the accumulator is a known constant, we can compute the 828 * comparison result. 829 */ 830 val = b->val[A_ATOM]; 831 if (vmap[val].is_const && BPF_SRC(b->s.code) == BPF_K) { 832 bpf_int32 v = vmap[val].const_val; 833 switch (BPF_OP(b->s.code)) { 834 835 case BPF_JEQ: 836 v = v == b->s.k; 837 break; 838 839 case BPF_JGT: 840 v = (unsigned)v > b->s.k; 841 break; 842 843 case BPF_JGE: 844 v = (unsigned)v >= b->s.k; 845 break; 846 847 case BPF_JSET: 848 v &= b->s.k; 849 break; 850 851 default: 852 abort(); 853 } 854 if (JF(b) != JT(b)) 855 done = 0; 856 if (v) 857 JF(b) = JT(b); 858 else 859 JT(b) = JF(b); 860 } 861} 862 863/* 864 * Compute the symbolic value of expression of 's', and update 865 * anything it defines in the value table 'val'. If 'alter' is true, 866 * do various optimizations. This code would be cleaner if symbolic 867 * evaluation and code transformations weren't folded together. 868 */ 869static void 870opt_stmt(s, val, alter) 871 struct stmt *s; 872 int val[]; 873 int alter; 874{ 875 int op; 876 int v; 877 878 switch (s->code) { 879 880 case BPF_LD|BPF_ABS|BPF_W: 881 case BPF_LD|BPF_ABS|BPF_H: 882 case BPF_LD|BPF_ABS|BPF_B: 883 v = F(s->code, s->k, 0L); 884 vstore(s, &val[A_ATOM], v, alter); 885 break; 886 887 case BPF_LD|BPF_IND|BPF_W: 888 case BPF_LD|BPF_IND|BPF_H: 889 case BPF_LD|BPF_IND|BPF_B: 890 v = val[X_ATOM]; 891 if (alter && vmap[v].is_const) { 892 s->code = BPF_LD|BPF_ABS|BPF_SIZE(s->code); 893 s->k += vmap[v].const_val; 894 v = F(s->code, s->k, 0L); 895 done = 0; 896 } 897 else 898 v = F(s->code, s->k, v); 899 vstore(s, &val[A_ATOM], v, alter); 900 break; 901 902 case BPF_LD|BPF_LEN: 903 v = F(s->code, 0L, 0L); 904 vstore(s, &val[A_ATOM], v, alter); 905 break; 906 907 case BPF_LD|BPF_IMM: 908 v = K(s->k); 909 vstore(s, &val[A_ATOM], v, alter); 910 break; 911 912 case BPF_LDX|BPF_IMM: 913 v = K(s->k); 914 vstore(s, &val[X_ATOM], v, alter); 915 break; 916 917 case BPF_LDX|BPF_MSH|BPF_B: 918 v = F(s->code, s->k, 0L); 919 vstore(s, &val[X_ATOM], v, alter); 920 break; 921 922 case BPF_ALU|BPF_NEG: 923 if (alter && vmap[val[A_ATOM]].is_const) { 924 s->code = BPF_LD|BPF_IMM; 925 s->k = -vmap[val[A_ATOM]].const_val; 926 val[A_ATOM] = K(s->k); 927 } 928 else 929 val[A_ATOM] = F(s->code, val[A_ATOM], 0L); 930 break; 931 932 case BPF_ALU|BPF_ADD|BPF_K: 933 case BPF_ALU|BPF_SUB|BPF_K: 934 case BPF_ALU|BPF_MUL|BPF_K: 935 case BPF_ALU|BPF_DIV|BPF_K: 936 case BPF_ALU|BPF_AND|BPF_K: 937 case BPF_ALU|BPF_OR|BPF_K: 938 case BPF_ALU|BPF_LSH|BPF_K: 939 case BPF_ALU|BPF_RSH|BPF_K: 940 op = BPF_OP(s->code); 941 if (alter) { 942 if (s->k == 0) { 943 /* don't optimize away "sub #0" 944 * as it may be needed later to 945 * fixup the generated math code */ 946 if (op == BPF_ADD || 947 op == BPF_LSH || op == BPF_RSH || 948 op == BPF_OR) { 949 s->code = NOP; 950 break; 951 } 952 if (op == BPF_MUL || op == BPF_AND) { 953 s->code = BPF_LD|BPF_IMM; 954 val[A_ATOM] = K(s->k); 955 break; 956 } 957 } 958 if (vmap[val[A_ATOM]].is_const) { 959 fold_op(s, val[A_ATOM], K(s->k)); 960 val[A_ATOM] = K(s->k); 961 break; 962 } 963 } 964 val[A_ATOM] = F(s->code, val[A_ATOM], K(s->k)); 965 break; 966 967 case BPF_ALU|BPF_ADD|BPF_X: 968 case BPF_ALU|BPF_SUB|BPF_X: 969 case BPF_ALU|BPF_MUL|BPF_X: 970 case BPF_ALU|BPF_DIV|BPF_X: 971 case BPF_ALU|BPF_AND|BPF_X: 972 case BPF_ALU|BPF_OR|BPF_X: 973 case BPF_ALU|BPF_LSH|BPF_X: 974 case BPF_ALU|BPF_RSH|BPF_X: 975 op = BPF_OP(s->code); 976 if (alter && vmap[val[X_ATOM]].is_const) { 977 if (vmap[val[A_ATOM]].is_const) { 978 fold_op(s, val[A_ATOM], val[X_ATOM]); 979 val[A_ATOM] = K(s->k); 980 } 981 else { 982 s->code = BPF_ALU|BPF_K|op; 983 s->k = vmap[val[X_ATOM]].const_val; 984 done = 0; 985 val[A_ATOM] = 986 F(s->code, val[A_ATOM], K(s->k)); 987 } 988 break; 989 } 990 /* 991 * Check if we're doing something to an accumulator 992 * that is 0, and simplify. This may not seem like 993 * much of a simplification but it could open up further 994 * optimizations. 995 * XXX We could also check for mul by 1, and -1, etc. 996 */ 997 if (alter && vmap[val[A_ATOM]].is_const 998 && vmap[val[A_ATOM]].const_val == 0) { 999 if (op == BPF_ADD || op == BPF_OR || 1000 op == BPF_LSH || op == BPF_RSH || op == BPF_SUB) { 1001 s->code = BPF_MISC|BPF_TXA; 1002 vstore(s, &val[A_ATOM], val[X_ATOM], alter); 1003 break; 1004 } 1005 else if (op == BPF_MUL || op == BPF_DIV || 1006 op == BPF_AND) { 1007 s->code = BPF_LD|BPF_IMM; 1008 s->k = 0; 1009 vstore(s, &val[A_ATOM], K(s->k), alter); 1010 break; 1011 } 1012 else if (op == BPF_NEG) { 1013 s->code = NOP; 1014 break; 1015 } 1016 } 1017 val[A_ATOM] = F(s->code, val[A_ATOM], val[X_ATOM]); 1018 break; 1019 1020 case BPF_MISC|BPF_TXA: 1021 vstore(s, &val[A_ATOM], val[X_ATOM], alter); 1022 break; 1023 1024 case BPF_LD|BPF_MEM: 1025 v = val[s->k]; 1026 if (alter && vmap[v].is_const) { 1027 s->code = BPF_LD|BPF_IMM; 1028 s->k = vmap[v].const_val; 1029 done = 0; 1030 } 1031 vstore(s, &val[A_ATOM], v, alter); 1032 break; 1033 1034 case BPF_MISC|BPF_TAX: 1035 vstore(s, &val[X_ATOM], val[A_ATOM], alter); 1036 break; 1037 1038 case BPF_LDX|BPF_MEM: 1039 v = val[s->k]; 1040 if (alter && vmap[v].is_const) { 1041 s->code = BPF_LDX|BPF_IMM; 1042 s->k = vmap[v].const_val; 1043 done = 0; 1044 } 1045 vstore(s, &val[X_ATOM], v, alter); 1046 break; 1047 1048 case BPF_ST: 1049 vstore(s, &val[s->k], val[A_ATOM], alter); 1050 break; 1051 1052 case BPF_STX: 1053 vstore(s, &val[s->k], val[X_ATOM], alter); 1054 break; 1055 } 1056} 1057 1058static void 1059deadstmt(s, last) 1060 register struct stmt *s; 1061 register struct stmt *last[]; 1062{ 1063 register int atom; 1064 1065 atom = atomuse(s); 1066 if (atom >= 0) { 1067 if (atom == AX_ATOM) { 1068 last[X_ATOM] = 0; 1069 last[A_ATOM] = 0; 1070 } 1071 else 1072 last[atom] = 0; 1073 } 1074 atom = atomdef(s); 1075 if (atom >= 0) { 1076 if (last[atom]) { 1077 done = 0; 1078 last[atom]->code = NOP; 1079 } 1080 last[atom] = s; 1081 } 1082} 1083 1084static void 1085opt_deadstores(b) 1086 register struct block *b; 1087{ 1088 register struct slist *s; 1089 register int atom; 1090 struct stmt *last[N_ATOMS]; 1091 1092 memset((char *)last, 0, sizeof last); 1093 1094 for (s = b->stmts; s != 0; s = s->next) 1095 deadstmt(&s->s, last); 1096 deadstmt(&b->s, last); 1097 1098 for (atom = 0; atom < N_ATOMS; ++atom) 1099 if (last[atom] && !ATOMELEM(b->out_use, atom)) { 1100 last[atom]->code = NOP; 1101 done = 0; 1102 } 1103} 1104 1105static void 1106opt_blk(b, do_stmts) 1107 struct block *b; 1108 int do_stmts; 1109{ 1110 struct slist *s; 1111 struct edge *p; 1112 int i; 1113 bpf_int32 aval; 1114 1115#if 0 1116 for (s = b->stmts; s && s->next; s = s->next) 1117 if (BPF_CLASS(s->s.code) == BPF_JMP) { 1118 do_stmts = 0; 1119 break; 1120 } 1121#endif 1122 1123 /* 1124 * Initialize the atom values. 1125 * If we have no predecessors, everything is undefined. 1126 * Otherwise, we inherent our values from our predecessors. 1127 * If any register has an ambiguous value (i.e. control paths are 1128 * merging) give it the undefined value of 0. 1129 */ 1130 p = b->in_edges; 1131 if (p == 0) 1132 memset((char *)b->val, 0, sizeof(b->val)); 1133 else { 1134 memcpy((char *)b->val, (char *)p->pred->val, sizeof(b->val)); 1135 while ((p = p->next) != NULL) { 1136 for (i = 0; i < N_ATOMS; ++i) 1137 if (b->val[i] != p->pred->val[i]) 1138 b->val[i] = 0; 1139 } 1140 } 1141 aval = b->val[A_ATOM]; 1142 for (s = b->stmts; s; s = s->next) 1143 opt_stmt(&s->s, b->val, do_stmts); 1144 1145 /* 1146 * This is a special case: if we don't use anything from this 1147 * block, and we load the accumulator with value that is 1148 * already there, or if this block is a return, 1149 * eliminate all the statements. 1150 */ 1151 if (do_stmts && 1152 ((b->out_use == 0 && aval != 0 &&b->val[A_ATOM] == aval) || 1153 BPF_CLASS(b->s.code) == BPF_RET)) { 1154 if (b->stmts != 0) { 1155 b->stmts = 0; 1156 done = 0; 1157 } 1158 } else { 1159 opt_peep(b); 1160 opt_deadstores(b); 1161 } 1162 /* 1163 * Set up values for branch optimizer. 1164 */ 1165 if (BPF_SRC(b->s.code) == BPF_K) 1166 b->oval = K(b->s.k); 1167 else 1168 b->oval = b->val[X_ATOM]; 1169 b->et.code = b->s.code; 1170 b->ef.code = -b->s.code; 1171} 1172 1173/* 1174 * Return true if any register that is used on exit from 'succ', has 1175 * an exit value that is different from the corresponding exit value 1176 * from 'b'. 1177 */ 1178static int 1179use_conflict(b, succ) 1180 struct block *b, *succ; 1181{ 1182 int atom; 1183 atomset use = succ->out_use; 1184 1185 if (use == 0) 1186 return 0; 1187 1188 for (atom = 0; atom < N_ATOMS; ++atom) 1189 if (ATOMELEM(use, atom)) 1190 if (b->val[atom] != succ->val[atom]) 1191 return 1; 1192 return 0; 1193} 1194 1195static struct block * 1196fold_edge(child, ep) 1197 struct block *child; 1198 struct edge *ep; 1199{ 1200 int sense; 1201 int aval0, aval1, oval0, oval1; 1202 int code = ep->code; 1203 1204 if (code < 0) { 1205 code = -code; 1206 sense = 0; 1207 } else 1208 sense = 1; 1209 1210 if (child->s.code != code) 1211 return 0; 1212 1213 aval0 = child->val[A_ATOM]; 1214 oval0 = child->oval; 1215 aval1 = ep->pred->val[A_ATOM]; 1216 oval1 = ep->pred->oval; 1217 1218 if (aval0 != aval1) 1219 return 0; 1220 1221 if (oval0 == oval1) 1222 /* 1223 * The operands are identical, so the 1224 * result is true if a true branch was 1225 * taken to get here, otherwise false. 1226 */ 1227 return sense ? JT(child) : JF(child); 1228 1229 if (sense && code == (BPF_JMP|BPF_JEQ|BPF_K)) 1230 /* 1231 * At this point, we only know the comparison if we 1232 * came down the true branch, and it was an equality 1233 * comparison with a constant. We rely on the fact that 1234 * distinct constants have distinct value numbers. 1235 */ 1236 return JF(child); 1237 1238 return 0; 1239} 1240 1241static void 1242opt_j(ep) 1243 struct edge *ep; 1244{ 1245 register int i, k; 1246 register struct block *target; 1247 1248 if (JT(ep->succ) == 0) 1249 return; 1250 1251 if (JT(ep->succ) == JF(ep->succ)) { 1252 /* 1253 * Common branch targets can be eliminated, provided 1254 * there is no data dependency. 1255 */ 1256 if (!use_conflict(ep->pred, ep->succ->et.succ)) { 1257 done = 0; 1258 ep->succ = JT(ep->succ); 1259 } 1260 } 1261 /* 1262 * For each edge dominator that matches the successor of this 1263 * edge, promote the edge successor to the its grandchild. 1264 * 1265 * XXX We violate the set abstraction here in favor a reasonably 1266 * efficient loop. 1267 */ 1268 top: 1269 for (i = 0; i < edgewords; ++i) { 1270 register bpf_u_int32 x = ep->edom[i]; 1271 1272 while (x != 0) { 1273 k = ffs(x) - 1; 1274 x &=~ (1 << k); 1275 k += i * BITS_PER_WORD; 1276 1277 target = fold_edge(ep->succ, edges[k]); 1278 /* 1279 * Check that there is no data dependency between 1280 * nodes that will be violated if we move the edge. 1281 */ 1282 if (target != 0 && !use_conflict(ep->pred, target)) { 1283 done = 0; 1284 ep->succ = target; 1285 if (JT(target) != 0) 1286 /* 1287 * Start over unless we hit a leaf. 1288 */ 1289 goto top; 1290 return; 1291 } 1292 } 1293 } 1294} 1295 1296 1297static void 1298or_pullup(b) 1299 struct block *b; 1300{ 1301 int val, at_top; 1302 struct block *pull; 1303 struct block **diffp, **samep; 1304 struct edge *ep; 1305 1306 ep = b->in_edges; 1307 if (ep == 0) 1308 return; 1309 1310 /* 1311 * Make sure each predecessor loads the same value. 1312 * XXX why? 1313 */ 1314 val = ep->pred->val[A_ATOM]; 1315 for (ep = ep->next; ep != 0; ep = ep->next) 1316 if (val != ep->pred->val[A_ATOM]) 1317 return; 1318 1319 if (JT(b->in_edges->pred) == b) 1320 diffp = &JT(b->in_edges->pred); 1321 else 1322 diffp = &JF(b->in_edges->pred); 1323 1324 at_top = 1; 1325 while (1) { 1326 if (*diffp == 0) 1327 return; 1328 1329 if (JT(*diffp) != JT(b)) 1330 return; 1331 1332 if (!SET_MEMBER((*diffp)->dom, b->id)) 1333 return; 1334 1335 if ((*diffp)->val[A_ATOM] != val) 1336 break; 1337 1338 diffp = &JF(*diffp); 1339 at_top = 0; 1340 } 1341 samep = &JF(*diffp); 1342 while (1) { 1343 if (*samep == 0) 1344 return; 1345 1346 if (JT(*samep) != JT(b)) 1347 return; 1348 1349 if (!SET_MEMBER((*samep)->dom, b->id)) 1350 return; 1351 1352 if ((*samep)->val[A_ATOM] == val) 1353 break; 1354 1355 /* XXX Need to check that there are no data dependencies 1356 between dp0 and dp1. Currently, the code generator 1357 will not produce such dependencies. */ 1358 samep = &JF(*samep); 1359 } 1360#ifdef notdef 1361 /* XXX This doesn't cover everything. */ 1362 for (i = 0; i < N_ATOMS; ++i) 1363 if ((*samep)->val[i] != pred->val[i]) 1364 return; 1365#endif 1366 /* Pull up the node. */ 1367 pull = *samep; 1368 *samep = JF(pull); 1369 JF(pull) = *diffp; 1370 1371 /* 1372 * At the top of the chain, each predecessor needs to point at the 1373 * pulled up node. Inside the chain, there is only one predecessor 1374 * to worry about. 1375 */ 1376 if (at_top) { 1377 for (ep = b->in_edges; ep != 0; ep = ep->next) { 1378 if (JT(ep->pred) == b) 1379 JT(ep->pred) = pull; 1380 else 1381 JF(ep->pred) = pull; 1382 } 1383 } 1384 else 1385 *diffp = pull; 1386 1387 done = 0; 1388} 1389 1390static void 1391and_pullup(b) 1392 struct block *b; 1393{ 1394 int val, at_top; 1395 struct block *pull; 1396 struct block **diffp, **samep; 1397 struct edge *ep; 1398 1399 ep = b->in_edges; 1400 if (ep == 0) 1401 return; 1402 1403 /* 1404 * Make sure each predecessor loads the same value. 1405 */ 1406 val = ep->pred->val[A_ATOM]; 1407 for (ep = ep->next; ep != 0; ep = ep->next) 1408 if (val != ep->pred->val[A_ATOM]) 1409 return; 1410 1411 if (JT(b->in_edges->pred) == b) 1412 diffp = &JT(b->in_edges->pred); 1413 else 1414 diffp = &JF(b->in_edges->pred); 1415 1416 at_top = 1; 1417 while (1) { 1418 if (*diffp == 0) 1419 return; 1420 1421 if (JF(*diffp) != JF(b)) 1422 return; 1423 1424 if (!SET_MEMBER((*diffp)->dom, b->id)) 1425 return; 1426 1427 if ((*diffp)->val[A_ATOM] != val) 1428 break; 1429 1430 diffp = &JT(*diffp); 1431 at_top = 0; 1432 } 1433 samep = &JT(*diffp); 1434 while (1) { 1435 if (*samep == 0) 1436 return; 1437 1438 if (JF(*samep) != JF(b)) 1439 return; 1440 1441 if (!SET_MEMBER((*samep)->dom, b->id)) 1442 return; 1443 1444 if ((*samep)->val[A_ATOM] == val) 1445 break; 1446 1447 /* XXX Need to check that there are no data dependencies 1448 between diffp and samep. Currently, the code generator 1449 will not produce such dependencies. */ 1450 samep = &JT(*samep); 1451 } 1452#ifdef notdef 1453 /* XXX This doesn't cover everything. */ 1454 for (i = 0; i < N_ATOMS; ++i) 1455 if ((*samep)->val[i] != pred->val[i]) 1456 return; 1457#endif 1458 /* Pull up the node. */ 1459 pull = *samep; 1460 *samep = JT(pull); 1461 JT(pull) = *diffp; 1462 1463 /* 1464 * At the top of the chain, each predecessor needs to point at the 1465 * pulled up node. Inside the chain, there is only one predecessor 1466 * to worry about. 1467 */ 1468 if (at_top) { 1469 for (ep = b->in_edges; ep != 0; ep = ep->next) { 1470 if (JT(ep->pred) == b) 1471 JT(ep->pred) = pull; 1472 else 1473 JF(ep->pred) = pull; 1474 } 1475 } 1476 else 1477 *diffp = pull; 1478 1479 done = 0; 1480} 1481 1482static void 1483opt_blks(root, do_stmts) 1484 struct block *root; 1485 int do_stmts; 1486{ 1487 int i, maxlevel; 1488 struct block *p; 1489 1490 init_val(); 1491 maxlevel = root->level; 1492 1493 find_inedges(root); 1494 for (i = maxlevel; i >= 0; --i) 1495 for (p = levels[i]; p; p = p->link) 1496 opt_blk(p, do_stmts); 1497 1498 if (do_stmts) 1499 /* 1500 * No point trying to move branches; it can't possibly 1501 * make a difference at this point. 1502 */ 1503 return; 1504 1505 for (i = 1; i <= maxlevel; ++i) { 1506 for (p = levels[i]; p; p = p->link) { 1507 opt_j(&p->et); 1508 opt_j(&p->ef); 1509 } 1510 } 1511 1512 find_inedges(root); 1513 for (i = 1; i <= maxlevel; ++i) { 1514 for (p = levels[i]; p; p = p->link) { 1515 or_pullup(p); 1516 and_pullup(p); 1517 } 1518 } 1519} 1520 1521static inline void 1522link_inedge(parent, child) 1523 struct edge *parent; 1524 struct block *child; 1525{ 1526 parent->next = child->in_edges; 1527 child->in_edges = parent; 1528} 1529 1530static void 1531find_inedges(root) 1532 struct block *root; 1533{ 1534 int i; 1535 struct block *b; 1536 1537 for (i = 0; i < n_blocks; ++i) 1538 blocks[i]->in_edges = 0; 1539 1540 /* 1541 * Traverse the graph, adding each edge to the predecessor 1542 * list of its successors. Skip the leaves (i.e. level 0). 1543 */ 1544 for (i = root->level; i > 0; --i) { 1545 for (b = levels[i]; b != 0; b = b->link) { 1546 link_inedge(&b->et, JT(b)); 1547 link_inedge(&b->ef, JF(b)); 1548 } 1549 } 1550} 1551 1552static void 1553opt_root(b) 1554 struct block **b; 1555{ 1556 struct slist *tmp, *s; 1557 1558 s = (*b)->stmts; 1559 (*b)->stmts = 0; 1560 while (BPF_CLASS((*b)->s.code) == BPF_JMP && JT(*b) == JF(*b)) 1561 *b = JT(*b); 1562 1563 tmp = (*b)->stmts; 1564 if (tmp != 0) 1565 sappend(s, tmp); 1566 (*b)->stmts = s; 1567 1568 /* 1569 * If the root node is a return, then there is no 1570 * point executing any statements (since the bpf machine 1571 * has no side effects). 1572 */ 1573 if (BPF_CLASS((*b)->s.code) == BPF_RET) 1574 (*b)->stmts = 0; 1575} 1576 1577static void 1578opt_loop(root, do_stmts) 1579 struct block *root; 1580 int do_stmts; 1581{ 1582 1583#ifdef BDEBUG 1584 if (dflag > 1) { 1585 printf("opt_loop(root, %d) begin\n", do_stmts); 1586 opt_dump(root); 1587 } 1588#endif 1589 do { 1590 done = 1; 1591 find_levels(root); 1592 find_dom(root); 1593 find_closure(root); 1594 find_ud(root); 1595 find_edom(root); 1596 opt_blks(root, do_stmts); 1597#ifdef BDEBUG 1598 if (dflag > 1) { 1599 printf("opt_loop(root, %d) bottom, done=%d\n", do_stmts, done); 1600 opt_dump(root); 1601 } 1602#endif 1603 } while (!done); 1604} 1605 1606/* 1607 * Optimize the filter code in its dag representation. 1608 */ 1609void 1610bpf_optimize(rootp) 1611 struct block **rootp; 1612{ 1613 struct block *root; 1614 1615 root = *rootp; 1616 1617 opt_init(root); 1618 opt_loop(root, 0); 1619 opt_loop(root, 1); 1620 intern_blocks(root); 1621#ifdef BDEBUG 1622 if (dflag > 1) { 1623 printf("after intern_blocks()\n"); 1624 opt_dump(root); 1625 } 1626#endif 1627 opt_root(rootp); 1628#ifdef BDEBUG 1629 if (dflag > 1) { 1630 printf("after opt_root()\n"); 1631 opt_dump(root); 1632 } 1633#endif 1634 opt_cleanup(); 1635} 1636 1637static void 1638make_marks(p) 1639 struct block *p; 1640{ 1641 if (!isMarked(p)) { 1642 Mark(p); 1643 if (BPF_CLASS(p->s.code) != BPF_RET) { 1644 make_marks(JT(p)); 1645 make_marks(JF(p)); 1646 } 1647 } 1648} 1649 1650/* 1651 * Mark code array such that isMarked(i) is true 1652 * only for nodes that are alive. 1653 */ 1654static void 1655mark_code(p) 1656 struct block *p; 1657{ 1658 cur_mark += 1; 1659 make_marks(p); 1660} 1661 1662/* 1663 * True iff the two stmt lists load the same value from the packet into 1664 * the accumulator. 1665 */ 1666static int 1667eq_slist(x, y) 1668 struct slist *x, *y; 1669{ 1670 while (1) { 1671 while (x && x->s.code == NOP) 1672 x = x->next; 1673 while (y && y->s.code == NOP) 1674 y = y->next; 1675 if (x == 0) 1676 return y == 0; 1677 if (y == 0) 1678 return x == 0; 1679 if (x->s.code != y->s.code || x->s.k != y->s.k) 1680 return 0; 1681 x = x->next; 1682 y = y->next; 1683 } 1684} 1685 1686static inline int 1687eq_blk(b0, b1) 1688 struct block *b0, *b1; 1689{ 1690 if (b0->s.code == b1->s.code && 1691 b0->s.k == b1->s.k && 1692 b0->et.succ == b1->et.succ && 1693 b0->ef.succ == b1->ef.succ) 1694 return eq_slist(b0->stmts, b1->stmts); 1695 return 0; 1696} 1697 1698static void 1699intern_blocks(root) 1700 struct block *root; 1701{ 1702 struct block *p; 1703 int i, j; 1704 int done; 1705 top: 1706 done = 1; 1707 for (i = 0; i < n_blocks; ++i) 1708 blocks[i]->link = 0; 1709 1710 mark_code(root); 1711 1712 for (i = n_blocks - 1; --i >= 0; ) { 1713 if (!isMarked(blocks[i])) 1714 continue; 1715 for (j = i + 1; j < n_blocks; ++j) { 1716 if (!isMarked(blocks[j])) 1717 continue; 1718 if (eq_blk(blocks[i], blocks[j])) { 1719 blocks[i]->link = blocks[j]->link ? 1720 blocks[j]->link : blocks[j]; 1721 break; 1722 } 1723 } 1724 } 1725 for (i = 0; i < n_blocks; ++i) { 1726 p = blocks[i]; 1727 if (JT(p) == 0) 1728 continue; 1729 if (JT(p)->link) { 1730 done = 0; 1731 JT(p) = JT(p)->link; 1732 } 1733 if (JF(p)->link) { 1734 done = 0; 1735 JF(p) = JF(p)->link; 1736 } 1737 } 1738 if (!done) 1739 goto top; 1740} 1741 1742static void 1743opt_cleanup() 1744{ 1745 free((void *)vnode_base); 1746 free((void *)vmap); 1747 free((void *)edges); 1748 free((void *)space); 1749 free((void *)levels); 1750 free((void *)blocks); 1751} 1752 1753/* 1754 * Return the number of stmts in 's'. 1755 */ 1756static int 1757slength(s) 1758 struct slist *s; 1759{ 1760 int n = 0; 1761 1762 for (; s; s = s->next) 1763 if (s->s.code != NOP) 1764 ++n; 1765 return n; 1766} 1767 1768/* 1769 * Return the number of nodes reachable by 'p'. 1770 * All nodes should be initially unmarked. 1771 */ 1772static int 1773count_blocks(p) 1774 struct block *p; 1775{ 1776 if (p == 0 || isMarked(p)) 1777 return 0; 1778 Mark(p); 1779 return count_blocks(JT(p)) + count_blocks(JF(p)) + 1; 1780} 1781 1782/* 1783 * Do a depth first search on the flow graph, numbering the 1784 * the basic blocks, and entering them into the 'blocks' array.` 1785 */ 1786static void 1787number_blks_r(p) 1788 struct block *p; 1789{ 1790 int n; 1791 1792 if (p == 0 || isMarked(p)) 1793 return; 1794 1795 Mark(p); 1796 n = n_blocks++; 1797 p->id = n; 1798 blocks[n] = p; 1799 1800 number_blks_r(JT(p)); 1801 number_blks_r(JF(p)); 1802} 1803 1804/* 1805 * Return the number of stmts in the flowgraph reachable by 'p'. 1806 * The nodes should be unmarked before calling. 1807 * 1808 * Note that "stmts" means "instructions", and that this includes 1809 * 1810 * side-effect statements in 'p' (slength(p->stmts)); 1811 * 1812 * statements in the true branch from 'p' (count_stmts(JT(p))); 1813 * 1814 * statements in the false branch from 'p' (count_stmts(JF(p))); 1815 * 1816 * the conditional jump itself (1); 1817 * 1818 * an extra long jump if the true branch requires it (p->longjt); 1819 * 1820 * an extra long jump if the false branch requires it (p->longjf). 1821 */ 1822static int 1823count_stmts(p) 1824 struct block *p; 1825{ 1826 int n; 1827 1828 if (p == 0 || isMarked(p)) 1829 return 0; 1830 Mark(p); 1831 n = count_stmts(JT(p)) + count_stmts(JF(p)); 1832 return slength(p->stmts) + n + 1 + p->longjt + p->longjf; 1833} 1834 1835/* 1836 * Allocate memory. All allocation is done before optimization 1837 * is begun. A linear bound on the size of all data structures is computed 1838 * from the total number of blocks and/or statements. 1839 */ 1840static void 1841opt_init(root) 1842 struct block *root; 1843{ 1844 bpf_u_int32 *p; 1845 int i, n, max_stmts; 1846 1847 /* 1848 * First, count the blocks, so we can malloc an array to map 1849 * block number to block. Then, put the blocks into the array. 1850 */ 1851 unMarkAll(); 1852 n = count_blocks(root); 1853 blocks = (struct block **)malloc(n * sizeof(*blocks)); 1854 unMarkAll(); 1855 n_blocks = 0; 1856 number_blks_r(root); 1857 1858 n_edges = 2 * n_blocks; 1859 edges = (struct edge **)malloc(n_edges * sizeof(*edges)); 1860 1861 /* 1862 * The number of levels is bounded by the number of nodes. 1863 */ 1864 levels = (struct block **)malloc(n_blocks * sizeof(*levels)); 1865 1866 edgewords = n_edges / (8 * sizeof(bpf_u_int32)) + 1; 1867 nodewords = n_blocks / (8 * sizeof(bpf_u_int32)) + 1; 1868 1869 /* XXX */ 1870 space = (bpf_u_int32 *)malloc(2 * n_blocks * nodewords * sizeof(*space) 1871 + n_edges * edgewords * sizeof(*space)); 1872 p = space; 1873 all_dom_sets = p; 1874 for (i = 0; i < n; ++i) { 1875 blocks[i]->dom = p; 1876 p += nodewords; 1877 } 1878 all_closure_sets = p; 1879 for (i = 0; i < n; ++i) { 1880 blocks[i]->closure = p; 1881 p += nodewords; 1882 } 1883 all_edge_sets = p; 1884 for (i = 0; i < n; ++i) { 1885 register struct block *b = blocks[i]; 1886 1887 b->et.edom = p; 1888 p += edgewords; 1889 b->ef.edom = p; 1890 p += edgewords; 1891 b->et.id = i; 1892 edges[i] = &b->et; 1893 b->ef.id = n_blocks + i; 1894 edges[n_blocks + i] = &b->ef; 1895 b->et.pred = b; 1896 b->ef.pred = b; 1897 } 1898 max_stmts = 0; 1899 for (i = 0; i < n; ++i) 1900 max_stmts += slength(blocks[i]->stmts) + 1; 1901 /* 1902 * We allocate at most 3 value numbers per statement, 1903 * so this is an upper bound on the number of valnodes 1904 * we'll need. 1905 */ 1906 maxval = 3 * max_stmts; 1907 vmap = (struct vmapinfo *)malloc(maxval * sizeof(*vmap)); 1908 vnode_base = (struct valnode *)malloc(maxval * sizeof(*vnode_base)); 1909} 1910 1911/* 1912 * Some pointers used to convert the basic block form of the code, 1913 * into the array form that BPF requires. 'fstart' will point to 1914 * the malloc'd array while 'ftail' is used during the recursive traversal. 1915 */ 1916static struct bpf_insn *fstart; 1917static struct bpf_insn *ftail; 1918 1919#ifdef BDEBUG 1920int bids[1000]; 1921#endif 1922 1923/* 1924 * Returns true if successful. Returns false if a branch has 1925 * an offset that is too large. If so, we have marked that 1926 * branch so that on a subsequent iteration, it will be treated 1927 * properly. 1928 */ 1929static int 1930convert_code_r(p) 1931 struct block *p; 1932{ 1933 struct bpf_insn *dst; 1934 struct slist *src; 1935 int slen; 1936 u_int off; 1937 int extrajmps; /* number of extra jumps inserted */ 1938 struct slist **offset = NULL; 1939 1940 if (p == 0 || isMarked(p)) 1941 return (1); 1942 Mark(p); 1943 1944 if (convert_code_r(JF(p)) == 0) 1945 return (0); 1946 if (convert_code_r(JT(p)) == 0) 1947 return (0); 1948 1949 slen = slength(p->stmts); 1950 dst = ftail -= (slen + 1 + p->longjt + p->longjf); 1951 /* inflate length by any extra jumps */ 1952 1953 p->offset = dst - fstart; 1954 1955 /* generate offset[] for convenience */ 1956 if (slen) { 1957 offset = (struct slist **)calloc(sizeof(struct slist *), slen); 1958 if (!offset) { 1959 bpf_error("not enough core"); 1960 /*NOTREACHED*/ 1961 } 1962 } 1963 src = p->stmts; 1964 for (off = 0; off < slen && src; off++) { 1965#if 0 1966 printf("off=%d src=%x\n", off, src); 1967#endif 1968 offset[off] = src; 1969 src = src->next; 1970 } 1971 1972 off = 0; 1973 for (src = p->stmts; src; src = src->next) { 1974 if (src->s.code == NOP) 1975 continue; 1976 dst->code = (u_short)src->s.code; 1977 dst->k = src->s.k; 1978 1979 /* fill block-local relative jump */ 1980 if (BPF_CLASS(src->s.code) != BPF_JMP || src->s.code == (BPF_JMP|BPF_JA)) { 1981#if 0 1982 if (src->s.jt || src->s.jf) { 1983 bpf_error("illegal jmp destination"); 1984 /*NOTREACHED*/ 1985 } 1986#endif 1987 goto filled; 1988 } 1989 if (off == slen - 2) /*???*/ 1990 goto filled; 1991 1992 { 1993 int i; 1994 int jt, jf; 1995 char *ljerr = "%s for block-local relative jump: off=%d"; 1996 1997#if 0 1998 printf("code=%x off=%d %x %x\n", src->s.code, 1999 off, src->s.jt, src->s.jf); 2000#endif 2001 2002 if (!src->s.jt || !src->s.jf) { 2003 bpf_error(ljerr, "no jmp destination", off); 2004 /*NOTREACHED*/ 2005 } 2006 2007 jt = jf = 0; 2008 for (i = 0; i < slen; i++) { 2009 if (offset[i] == src->s.jt) { 2010 if (jt) { 2011 bpf_error(ljerr, "multiple matches", off); 2012 /*NOTREACHED*/ 2013 } 2014 2015 dst->jt = i - off - 1; 2016 jt++; 2017 } 2018 if (offset[i] == src->s.jf) { 2019 if (jf) { 2020 bpf_error(ljerr, "multiple matches", off); 2021 /*NOTREACHED*/ 2022 } 2023 dst->jf = i - off - 1; 2024 jf++; 2025 } 2026 } 2027 if (!jt || !jf) { 2028 bpf_error(ljerr, "no destination found", off); 2029 /*NOTREACHED*/ 2030 } 2031 } 2032filled: 2033 ++dst; 2034 ++off; 2035 } 2036 if (offset) 2037 free(offset); 2038 2039#ifdef BDEBUG 2040 bids[dst - fstart] = p->id + 1; 2041#endif 2042 dst->code = (u_short)p->s.code; 2043 dst->k = p->s.k; 2044 if (JT(p)) { 2045 extrajmps = 0; 2046 off = JT(p)->offset - (p->offset + slen) - 1; 2047 if (off >= 256) { 2048 /* offset too large for branch, must add a jump */ 2049 if (p->longjt == 0) { 2050 /* mark this instruction and retry */ 2051 p->longjt++; 2052 return(0); 2053 } 2054 /* branch if T to following jump */ 2055 dst->jt = extrajmps; 2056 extrajmps++; 2057 dst[extrajmps].code = BPF_JMP|BPF_JA; 2058 dst[extrajmps].k = off - extrajmps; 2059 } 2060 else 2061 dst->jt = off; 2062 off = JF(p)->offset - (p->offset + slen) - 1; 2063 if (off >= 256) { 2064 /* offset too large for branch, must add a jump */ 2065 if (p->longjf == 0) { 2066 /* mark this instruction and retry */ 2067 p->longjf++; 2068 return(0); 2069 } 2070 /* branch if F to following jump */ 2071 /* if two jumps are inserted, F goes to second one */ 2072 dst->jf = extrajmps; 2073 extrajmps++; 2074 dst[extrajmps].code = BPF_JMP|BPF_JA; 2075 dst[extrajmps].k = off - extrajmps; 2076 } 2077 else 2078 dst->jf = off; 2079 } 2080 return (1); 2081} 2082 2083 2084/* 2085 * Convert flowgraph intermediate representation to the 2086 * BPF array representation. Set *lenp to the number of instructions. 2087 */ 2088struct bpf_insn * 2089icode_to_fcode(root, lenp) 2090 struct block *root; 2091 int *lenp; 2092{ 2093 int n; 2094 struct bpf_insn *fp; 2095 2096 /* 2097 * Loop doing convert_code_r() until no branches remain 2098 * with too-large offsets. 2099 */ 2100 while (1) { 2101 unMarkAll(); 2102 n = *lenp = count_stmts(root); 2103 2104 fp = (struct bpf_insn *)malloc(sizeof(*fp) * n); 2105 memset((char *)fp, 0, sizeof(*fp) * n); 2106 fstart = fp; 2107 ftail = fp + n; 2108 2109 unMarkAll(); 2110 if (convert_code_r(root)) 2111 break; 2112 free(fp); 2113 } 2114 2115 return fp; 2116} 2117 2118/* 2119 * Make a copy of a BPF program and put it in the "fcode" member of 2120 * a "pcap_t". 2121 * 2122 * If we fail to allocate memory for the copy, fill in the "errbuf" 2123 * member of the "pcap_t" with an error message, and return -1; 2124 * otherwise, return 0. 2125 */ 2126int 2127install_bpf_program(pcap_t *p, struct bpf_program *fp) 2128{ 2129 size_t prog_size; 2130 2131 /* 2132 * Free up any already installed program. 2133 */ 2134 pcap_freecode(&p->fcode); 2135 2136 prog_size = sizeof(*fp->bf_insns) * fp->bf_len; 2137 p->fcode.bf_len = fp->bf_len; 2138 p->fcode.bf_insns = (struct bpf_insn *)malloc(prog_size); 2139 if (p->fcode.bf_insns == NULL) { 2140 snprintf(p->errbuf, sizeof(p->errbuf), 2141 "malloc: %s", pcap_strerror(errno)); 2142 return (-1); 2143 } 2144 memcpy(p->fcode.bf_insns, fp->bf_insns, prog_size); 2145 return (0); 2146} 2147 2148#ifdef BDEBUG 2149static void 2150opt_dump(root) 2151 struct block *root; 2152{ 2153 struct bpf_program f; 2154 2155 memset(bids, 0, sizeof bids); 2156 f.bf_insns = icode_to_fcode(root, &f.bf_len); 2157 bpf_dump(&f, 1); 2158 putchar('\n'); 2159 free((char *)f.bf_insns); 2160} 2161#endif 2162