2 * IRC - Internet Relay Chat, common/match.c
3 * Copyright (C) 1990 Jarkko Oikarinen
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 1, or (at your option)
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write to the Free Software
17 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
24 #include "ircd_chattr.h"
25 #include "ircd_string.h"
26 #include "ircd_snprintf.h"
31 * Written by Run (carlo@runaway.xs4all.nl), 25-10-96
34 * From: Carlo Wood <carlo@runaway.xs4all.nl>
35 * Message-Id: <199609021026.MAA02393@runaway.xs4all.nl>
36 * Subject: [C-Com] Analysis for `mmatch' (was: gline4 problem)
37 * To: coder-com@mail.undernet.org (coder committee)
38 * Date: Mon, 2 Sep 1996 12:26:01 +0200 (MET DST)
40 * We need a new function `mmatch(const char *old_mask, const char *new_mask)'
41 * which returns `true' likewise the current `match' (start with copying it),
42 * but which treats '*' and '?' in `new_mask' differently (not "\*" and "\?" !)
43 * as follows: a '*' in `new_mask' does not match a '?' in `old_mask' and
44 * a '?' in `new_mask' does not match a '\?' in `old_mask'.
45 * And ofcourse... a '*' in `new_mask' does not match a '\*' in `old_mask'...
46 * And last but not least, '\?' and '\*' in `new_mask' now become one character.
49 int mmatch(const char *old_mask, const char *new_mask)
51 const char *m = old_mask;
52 const char *n = new_mask;
73 for (m--; (m > old_mask) && (*m == '?'); m--)
75 if ((*m == '*') && (m > old_mask) && (m[-1] != '\\'))
81 /* Added to `mmatch' : Because '\?' and '\*' now is one character: */
82 if ((*na == '\\') && ((na[1] == '*') || (na[1] == '?')))
93 if ((*m == '\\') && ((m[1] == '*') || (m[1] == '?')))
101 /* Added to `mmatch' : Because '\?' and '\*' now is one character: */
102 if ((*n == '\\') && ((n[1] == '*') || (n[1] == '?')))
111 * This `if' has been changed compared to match() to do the following:
113 * old (m) new (n) boolean expression
114 * * any (*m == '*' && !mq) ||
115 * ? any except '*' (*m == '?' && !mq && (*n != '*' || nq)) ||
116 * any except * or ? same as m (!((*m == '*' || *m == '?') && !mq) &&
117 * ToLower(*m) == ToLower(*n) &&
118 * !((mq && !nq) || (!mq && nq)))
120 * Here `any' also includes \* and \? !
122 * After reworking the boolean expressions, we get:
123 * (Optimized to use boolean shortcircuits, with most frequently occuring
124 * cases upfront (which took 2 hours!)).
126 if ((*m == '*' && !mq) ||
127 ((!mq || nq) && ToLower(*m) == ToLower(*n)) ||
128 (*m == '?' && !mq && (*n != '*' || nq)))
141 /* Added to `mmatch' : Because '\?' and '\*' now is one character: */
142 if ((*na == '\\') && ((na[1] == '*') || (na[1] == '?')))
151 * Compare if a given string (name) matches the given
152 * mask (which can contain wild cards: '*' - match any
153 * number of chars, '?' - match any single character.
158 * Originally by Douglas A Lewis (dalewis@acsu.buffalo.edu)
159 * Rewritten by Timothy Vogelsang (netski), net@astrolink.org
162 int match(const char *mask, const char *name)
164 const char *m = mask, *n = name;
165 const char *m_tmp = mask, *n_tmp = name;
171 while (*m == '*') /* clean up any additional wildcards */
178 if (*m == '\\') /* next wildcard is disregarded */
183 return 0; /* match */
185 for (m--; (m > mask) && (*m == '?'); m--);
188 if (*m == '*' && (m > mask))
189 return 0; /* match */
198 while (*m == '*') /* clean up any additional wildcards */
203 if (ToLower(*m) != ToLower(*n) && *m != '?') {
205 return 1; /* failure! */
218 return 1; /* no match! */
223 * Collapse a pattern string into minimal components.
224 * This particular version is "in place", so that it changes the pattern
225 * which is to be reduced to a "minimal" size.
227 * (C) Carlo Wood - 6 Oct 1998
228 * Speedup rewrite by Andrea Cocito, December 1998.
229 * Note that this new optimized alghoritm can *only* work in place.
232 char *collapse(char *mask)
242 if ((*m == '*') && ((m[1] == '*') || (m[1] == '?')))
251 if (star && (*m != '?'))
257 if ((*m == '\\') && ((m[1] == '*') || (m[1] == '?')))
266 if ((*m == '\\') && ((m[1] == '*') || (m[1] == '?')))
276 ***************** Nemesi's matchcomp() / matchexec() **************
279 /* These functions allow the use of "compiled" masks, you compile a mask
280 * by means of matchcomp() that gets the plain text mask as input and writes
281 * its result in the memory locations addressed by the 3 parameters:
282 * - *cmask will contain the text of the compiled mask
283 * - *minlen will contain the lenght of the shortest string that can match
285 * - *charset will contain the minimal set of chars needed to match the mask
286 * You can pass NULL as *charset and it will be simply not returned, but you
287 * MUST pass valid pointers for *minlen and *cmask (wich must be big enough
288 * to contain the compiled mask text that is in the worst case as long as the
289 * text of the mask itself in plaintext format) and the return value of
290 * matchcomp() will be the number of chars actually written there (excluded
291 * the trailing zero). cmask can be == mask, matchcomp() can work in place.
292 * The {cmask, minlen} couple of values make the real compiled mask and
293 * need to be passed to the functions that use the compiled mask, if you pass
294 * the wrong minlen or something wrong in cmask to one of these expect a
295 * coredump. This means that when you record a compiled mask you must store
296 * *both* these values.
297 * Once compiled the mask can be used to match a string by means of
298 * matchexec(), it can be printed back to human-readable format by means
299 * of sprintmatch() or it can be compared to another compiled mask by means
300 * of mmexec() that will tell if it completely overrides that mask (a lot like
301 * what mmatch() does for plain text masks).
302 * You can gain a lot of speed in many situations avoiding to matchexec() when:
303 * - The maximum lenght of the field you are about to match() the mask to is
304 * shorter than minlen, in example when matching abc*def*ghil with a nick:
305 * It just cannot match since a nick is at most 9 chars long and the mask
306 * needs at least 10 chars (10 will be the value returned in minlen).
307 * - The charset allowed for the field you are about to match to doesn't
308 * "contain" the charset returned by matchcomp(), in example when you
309 * have *.* as mask it makes no sense to try to match it against a nick
310 * because, again, a nick can't contain a '.', you can check this with
311 * a simple (charset & NTL_IRCNK) in this case.
312 * - As a special case, since compiled masks are forced to lowercase,
313 * it would make no sense to use the NTL_LOWER and NTL_UPPER on a compiled
314 * mask, thus they are reused as follows: if the NTL_LOWER bit of charset
315 * is set it means that the mask contains only non-wilds chars (i.e. you can
316 * use strCasecmp() to match it or a direct hash lookup), if the NTL_UPPER
317 * bit is set it means that it contains only wild chars (and you can
318 * match it with strlen(field)>=minlen).
319 * Do these optimizations ONLY when the data you are about to pass to
320 * matchexec() are *known* to be invalid in advance, using strChattr()
321 * or strlen() on the text would be slower than calling matchexec() directly
323 * Internally a compiled mask contain in the *cmask area the text of
324 * the plain text form of the mask itself with applied the following hacks:
325 * - All characters are forced to lowercase (so that uppercase letters and
326 * specifically the symbols 'A' and 'Z' are reserved for special use)
327 * - All non-escaped stars '*' are replaced by the letter 'Z'
328 * - All non-escaped question marks '?' are replaced by the letter 'A'
329 * - All escape characters are removed, the wilds escaped by them are
330 * then passed by without the escape since they don't collide anymore
331 * with the real wilds (encoded as A/Z)
332 * - Finally the part of the mask that follows the last asterisk is
333 * reversed (byte order mirroring) and moved right after the first
335 * After all this a mask like: Head*CHUNK1*chu\*nK2*ch??k3*TaIl
336 * .... becomes: headZliatZchunk1Zchu*nk2ZchAAk3
337 * This can still be printed on a console, more or less understood by an
338 * human and handled with the usual str*() library functions.
339 * When you store somewhere the compiled mask you can avoid storing the
340 * textform of it since it can be "decompiled" by means of sprintmatch(),
341 * but at that time the following things are changed in the mask:
342 * - All chars have been forced to lowercase.
343 * - The mask is collapsed.
344 * The balance point of using compiled masks in terms of CPU is when you expect
345 * to use matchexec() instead of match() at least 20 times on the same mask
346 * or when you expect to use mmexec() instead of mmatch() 3 times.
352 * Compiles a mask into a form suitable for using in matchexec().
355 int matchcomp(char *cmask, int *minlen, int *charset, const char *mask)
357 const char *m = mask;
362 int l1, l2, lmin, loop, sign;
367 int chset2 = (NTL_LOWER | NTL_UPPER);
379 chset2 &= ~NTL_LOWER;
382 if ((*m == '?') || (*m == '*'))
390 chset2 &= ~NTL_LOWER;
395 chset &= IRCD_CharAttrTab[*b++ - CHAR_MIN];
396 chset2 &= ~NTL_UPPER;
400 *charset = (chset | chset2);
411 for (x1 = ls + 1, x2 = (b - 1); x1 < x2; x1++, x2--)
420 while ((lmin = (l1 < l2) ? l1 : l2))
423 for (loop = 0; loop < lmin; loop++)
431 l1 -= (sign < 0) ? 0 : lmin;
432 l2 -= (sign > 0) ? 0 : lmin;
445 * Executes a match with a mask previosuly compiled with matchcomp()
446 * Note 1: If the mask isn't correctly produced by matchcomp() I will core
447 * Note 2: 'min' MUST be the value returned by matchcomp on that mask,
448 * or.... I will core even faster :-)
449 * Note 3: This piece of code is not intended to be nice but efficient.
452 int matchexec(const char *string, const char *cmask, int minlen)
454 const char *s = string - 1;
455 const char *b = cmask - 1;
461 while ((ToLower(*++s) == *++b) && *s);
463 return ((*b != '\0') && ((*b++ != 'Z') || (*b != '\0')));
474 if ((trash = (s - string - minlen)) < 0)
478 while ((ToLower(*--s) == *++b) && *b && (ToLower(*--s) == *++b) && *b
479 && (ToLower(*--s) == *++b) && *b && (ToLower(*--s) == *++b) && *b);
492 while ((ToLower(*++s) != ch))
498 while ((ToLower(*++s) == *++b) && *b);
521 * Prints the human readable version of *cmask into *mask, (decompiles
523 * The area pointed by *mask MUST be big enough (the mask might be up to
524 * twice the size of its compiled form if it's made all of \? or \*, and
525 * this function can NOT work in place since it might enflate the mask)
526 * The printed mask is not identical to the one that was compiled to cmask,
527 * infact it is 1) forced to all lowercase, 2) collapsed, both things
528 * are supposed to NOT change it's meaning.
529 * It returns the number of chars actually written to *mask;
532 int matchdecomp(char *mask, const char *cmask)
535 const char *rcm = cmask;
536 const char *begtail, *endtail;
544 for (; (*rcm != 'Z'); rcm++, rtb++)
546 if ((*rcm == '?') || (*rcm == '*'))
548 if (!((*rtb = ((*rcm == 'A') ? '?' : *rcm))))
555 while (*rcm && (*rcm != 'Z'))
580 for (rcm = endtail; (--rcm) > begtail; *rtb++ = ((*rcm == 'A') ? '?' : *rcm))
581 if ((*rcm == '?') || (*rcm == '*'))
590 * Checks if a wider compiled mask (wcm/wminlen) completely overrides
591 * a more restrict one (rcm/rminlen), basically what mmatch() does for
592 * non-compiled masks, returns 0 if the override is true (like mmatch()).
593 * "the wider overrides the restrict" means that any string that matches
594 * the restrict one _will_ also match the wider one, always.
595 * In this we behave differently from mmatch() because in example we return
596 * true for " a?*cd overrides a*bcd " for wich the override happens for how
597 * we literally defined it, here mmatch() would have returned false.
598 * The original concepts and the base alghoritm are copied from mmatch()
599 * written by Run (Carlo Wood), this function is written by
600 * Nemesi (Andrea Cocito)
603 int mmexec(const char *wcm, int wminlen, const char *rcm, int rminlen)
605 const char *w, *r, *br, *bw, *rx, *rz;
608 /* First of all rm must have enough non-stars to 'contain' wm */
609 if ((trash = rminlen - wminlen) < 0)
615 /* Let's start the game, remember that '*' is mapped to 'Z', '?'
616 is mapped to 'A' and that head?*??*?chunk*???*tail becomes
617 headAAAAZliatAAAZchunk for compiled masks */
619 /* Match the head of wm with the head of rm */
620 for (; (*r) && (*r != 'Z') && ((*w == *r) || (*w == 'A')); r++, w++);
622 while (*w == 'A') /* Eat extra '?' before '*' in wm if got '*' in rm */
624 if (*w != 'Z') /* head1<any>.. can't match head2<any>.. */
625 return ((*w) || (*r)) ? 1 : 0; /* and head<nul> matches only head<nul> */
627 return 0; /* headZ<nul> matches head<anything> */
629 /* Does rm have any stars in it ? let's check */
630 for (rx = r; *r && (*r != 'Z'); r++);
633 /* rm has no stars and thus isn't a mask but it's just a flat
634 string: special handling occurs here, note that eat must be 0 here */
639 for (; r--, (*w) && ((*w == *r) || (*w == 'A')); w++);
640 if (*w != 'Z') /* headZliat1<any> fails on head<any>2tail */
641 return (*w) ? 1 : 0; /* but headZliat<nul> matches head<any>tail */
644 /* match the chunks */
646 { /* This loop can't break but only return */
648 for (bw = w++; (*w != *rx); rx++) /* Seek the 1st char of the chunk */
649 if (--trash < 0) /* See if we can trash one more char of rm */
650 return 1; /* If not we can only fail of course */
651 for (r = ++rx, w++; (*w) && ((*w == *r) || (*w == 'A')); r++, w++);
652 if (!*w) /* Did last loop match the rest of chunk ? */
653 return 0; /* ... Yes, end of wm, matched ! */
655 { /* ... No, hitted non-star */
656 w = bw; /* Rollback at beginning of chunk */
657 if (--trash < 0) /* Trashed the char where this try started */
658 return 1; /* if we can't trash more chars fail */
662 rx = r; /* Successfully matched a chunk, move rx */
663 } /* and go on with the next one */
667 /* rm has at least one '*' and thus is a 'real' mask */
668 rz = r++; /* rx = unused of head, rz = beg-tail */
670 /* Match the tail of wm (if any) against the tail of rm */
673 for (; (*w) && (*r != 'Z') && ((*w == *r) || (*w == 'A')); w++, r++);
674 if (*r == 'Z') /* extra '?' before tail are fluff, just flush 'em */
677 if (*w != 'Z') /* We aren't matching a chunk, can't rollback */
681 /* Match the chunks of wm against what remains of the head of rm */
685 for (bw++; (rx < rz) && (*bw != *rx); rx++) /* Seek the first */
686 if (--trash < 0) /* waste some trash reserve */
688 if (!(rx < rz)) /* head finished */
690 for (bw++, (br = ++rx);
691 (br < rz) && (*bw) && ((*bw == *br) || (*bw == 'A')); br++, bw++);
692 if (!(br < rz)) /* Note that we didn't use any 'eat' char yet, if */
693 while (*bw == 'A') /* there were eat-en chars the head would be over */
694 bw++, eat++; /* Happens only at end of head, and eat is still 0 */
701 { /* If we failed because we got the end of head */
702 trash -= (br - rx); /* it makes no sense to rollback, just trash */
703 if (--trash < 0) /* all the rest of the head wich isn't long */
704 return 1; /* enough for this chunk and go out of this */
705 break; /* loop, then we try with the chunks of rm */
717 /* Match the unused chunks of wm against the chunks of rm */
719 for (; *r && (*r != 'Z'); r++);
726 while (eat && *r) /* the '?' we had eated make us skip as many chars */
727 if (*r++ != 'Z') /* here, but can't skip stars or trailing zero */
729 for (bw++; (*r) && (*bw != *r); r++)
730 if ((*r != 'Z') && (--trash < 0))
734 for ((br = ++r), bw++;
735 (*br) && (*br != 'Z') && ((*bw == *br) || (*bw == 'A')); br++, bw++);
744 if ((!*br) || (*r == 'Z'))
745 { /* If we hit the end of rm or a star in it */
746 trash -= (br - r); /* makes no sense to rollback within this */
747 if (trash < 0) /* same chunk of br, skip it all and then */
748 return 1; /* either rollback or break this loop if */
749 if (!*br) /* it was the end of rm */
764 /* match the remaining chunks of wm against what remains of the tail of rm */
765 r = rz - eat - 1; /* can't have <nul> or 'Z'within the tail, so just move r */
769 for (bw++; (*bw != *r); r--)
774 for ((br = --r), bw++;
775 (*bw) && (br >= rx) && ((*bw == *br) || (*bw == 'A')); br--, bw++);
791 return 1; /* Auch... something left out ? Fail */
797 #include <sys/socket.h>
798 #include <netinet/in.h>
799 #include <arpa/inet.h>
801 static int ipmask_parse_ipv4(const char *in, struct in_addr *out)
808 class = sscanf(in, "%d.%d.%d.%d/%d", &ad[0], &ad[1], &ad[2], &ad[3], &bits);
811 ircd_snprintf(0, ipname, sizeof(ipname), "%d.%d.%d.%d", ad[0], ad[1], ad[2], ad[3]);
812 out->s_addr = inet_addr(ipname);
816 int check_if_ipmask(const char *mask)
821 for (p = mask; *p; ++p)
822 if (*p != '*' && *p != '?' && *p != '.' && *p != '/')
832 int ipmask_parse(const char *in, struct irc_in_addr *mask, unsigned char *bits_ptr)
838 if (check_if_ipmask(in)) {
839 bits = ipmask_parse_ipv4(in, &ipv4);
840 mask->in6_16[0] = mask->in6_16[1] = mask->in6_16[2] = 0;
841 mask->in6_16[3] = mask->in6_16[4] = mask->in6_16[5] = 0;
842 memcpy(&mask->in6_16[6], &ipv4.s_addr, sizeof(ipv4.s_addr));
845 if (!(p = strchr(in, '/')))
849 if (!ircd_aton(mask, in)) {
865 int ipmask_check(const struct irc_in_addr *addr, const struct irc_in_addr *mask, unsigned char bits)
869 for (k = 0; k < 8; k++) {
871 return (addr->in6_16[k] & ((unsigned char) (0xffff << (16-bits)))) == mask->in6_16[k];
872 if (addr->in6_16[k] != mask->in6_16[k])