/* parse.c Common parser code for dhcpd and dhclient. */ /* * Copyright (c) 1995, 1996, 1997, 1998 The Internet Software Consortium. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of The Internet Software Consortium nor the names * of its contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE INTERNET SOFTWARE CONSORTIUM AND * CONTRIBUTORS ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE * DISCLAIMED. IN NO EVENT SHALL THE INTERNET SOFTWARE CONSORTIUM OR * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * This software has been written for the Internet Software Consortium * by Ted Lemon in cooperation with Vixie * Enterprises. To learn more about the Internet Software Consortium, * see ``http://www.vix.com/isc''. To learn more about Vixie * Enterprises, see ``http://www.vix.com''. */ #include "dhcpd.h" #include "dhctoken.h" /* Skip to the semicolon ending the current statement. If we encounter braces, the matching closing brace terminates the statement. If we encounter a right brace but haven't encountered a left brace, return leaving the brace in the token buffer for the caller. If we see a semicolon and haven't seen a left brace, return. This lets us skip over: statement; statement foo bar { } statement foo bar { statement { } } statement} ...et cetera. */ void skip_to_semi (cfile) FILE *cfile; { int token; char *val; int brace_count = 0; do { token = peek_token (&val, cfile); if (token == RBRACE) { if (brace_count) { token = next_token (&val, cfile); if (!--brace_count) return; } else return; } else if (token == LBRACE) { brace_count++; } else if (token == SEMI && !brace_count) { token = next_token (&val, cfile); return; } else if (token == EOL) { /* EOL only happens when parsing /etc/resolv.conf, and we treat it like a semicolon because the resolv.conf file is line-oriented. */ token = next_token (&val, cfile); return; } token = next_token (&val, cfile); } while (token != EOF); } int parse_semi (cfile) FILE *cfile; { int token; char *val; token = next_token (&val, cfile); if (token != SEMI) { parse_warn ("semicolon expected."); skip_to_semi (cfile); return 0; } return 1; } /* string-parameter :== STRING SEMI */ char *parse_string (cfile) FILE *cfile; { char *val; int token; char *s; token = next_token (&val, cfile); if (token != STRING) { parse_warn ("filename must be a string"); skip_to_semi (cfile); return (char *)0; } s = (char *)malloc (strlen (val) + 1); if (!s) error ("no memory for string %s.", val); strlcpy (s, val, strlen(val) + 1); if (!parse_semi (cfile)) return (char *)0; return s; } /* hostname :== identifier | hostname DOT identifier */ char *parse_host_name (cfile) FILE *cfile; { char *val; int token; int len = 0; char *s; char *t; pair c = (pair)0; /* Read a dotted hostname... */ do { /* Read a token, which should be an identifier. */ token = next_token (&val, cfile); if (!is_identifier (token) && token != NUMBER) { parse_warn ("expecting an identifier in hostname"); skip_to_semi (cfile); return (char *)0; } /* Store this identifier... */ if (!(s = (char *)malloc (strlen (val) + 1))) error ("can't allocate temp space for hostname."); strlcpy (s, val, strlen(val) + 1); c = cons ((caddr_t)s, c); len += strlen (s) + 1; /* Look for a dot; if it's there, keep going, otherwise we're done. */ token = peek_token (&val, cfile); if (token == DOT) token = next_token (&val, cfile); } while (token == DOT); /* Assemble the hostname together into a string. */ if (!(s = (char *)malloc (len))) error ("can't allocate space for hostname."); t = s + len; *--t = 0; while (c) { pair cdr = c -> cdr; int l = strlen ((char *)(c -> car)); t -= l; memcpy (t, (char *)(c -> car), l); /* Free up temp space. */ free (c -> car); free (c); c = cdr; if (t != s) *--t = '.'; } return s; } int parse_ip_addr (cfile, addr) FILE *cfile; struct iaddr *addr; { addr -> len = 4; if (parse_numeric_aggregate (cfile, addr -> iabuf, &addr -> len, DOT, 10, 8)) return 1; return 0; } /* hardware-parameter :== HARDWARE ETHERNET csns SEMI csns :== NUMBER | csns COLON NUMBER */ void parse_hardware_param (cfile, hardware) FILE *cfile; struct hardware *hardware; { char *val; int token; int hlen; unsigned char *t; token = next_token (&val, cfile); switch (token) { case ETHERNET: hardware -> htype = HTYPE_ETHER; break; case TOKEN_RING: hardware -> htype = HTYPE_IEEE802; break; case FDDI: hardware -> htype = HTYPE_FDDI; break; default: parse_warn ("expecting a network hardware type"); skip_to_semi (cfile); return; } /* Parse the hardware address information. Technically, it would make a lot of sense to restrict the length of the data we'll accept here to the length of a particular hardware address type. Unfortunately, there are some broken clients out there that put bogus data in the chaddr buffer, and we accept that data in the lease file rather than simply failing on such clients. Yuck. */ hlen = 0; t = parse_numeric_aggregate (cfile, (unsigned char *)0, &hlen, COLON, 16, 8); if (!t) return; if (hlen > sizeof hardware -> haddr) { free (t); parse_warn ("hardware address too long"); } else { hardware -> hlen = hlen; memcpy ((unsigned char *)&hardware -> haddr [0], t, hardware -> hlen); if (hlen < sizeof hardware -> haddr) memset (&hardware -> haddr [hlen], 0, (sizeof hardware -> haddr) - hlen); free (t); } token = next_token (&val, cfile); if (token != SEMI) { parse_warn ("expecting semicolon."); skip_to_semi (cfile); } } /* lease-time :== NUMBER SEMI */ void parse_lease_time (cfile, timep) FILE *cfile; TIME *timep; { char *val; int token; token = next_token (&val, cfile); if (token != NUMBER) { parse_warn ("Expecting numeric lease time"); skip_to_semi (cfile); return; } convert_num ((unsigned char *)timep, val, 10, 32); /* Unswap the number - convert_num returns stuff in NBO. */ *timep = ntohl (*timep); /* XXX */ parse_semi (cfile); } /* No BNF for numeric aggregates - that's defined by the caller. What this function does is to parse a sequence of numbers separated by the token specified in separator. If max is zero, any number of numbers will be parsed; otherwise, exactly max numbers are expected. Base and size tell us how to internalize the numbers once they've been tokenized. */ unsigned char *parse_numeric_aggregate (cfile, buf, max, separator, base, size) FILE *cfile; unsigned char *buf; int *max; int separator; int base; int size; { char *val; int token; unsigned char *bufp = buf, *s = NULL; char *t; int count = 0; pair c = (pair)0; if (!bufp && *max) { bufp = (unsigned char *)malloc (*max * size / 8); if (!bufp) error ("can't allocate space for numeric aggregate"); } else s = bufp; do { if (count) { token = peek_token (&val, cfile); if (token != separator) { if (!*max) break; if (token != RBRACE && token != LBRACE) token = next_token (&val, cfile); parse_warn ("too few numbers."); if (token != SEMI) skip_to_semi (cfile); return (unsigned char *)0; } token = next_token (&val, cfile); } token = next_token (&val, cfile); if (token == EOF) { parse_warn ("unexpected end of file"); break; } /* Allow NUMBER_OR_NAME if base is 16. */ if (token != NUMBER && (base != 16 || token != NUMBER_OR_NAME)) { parse_warn ("expecting numeric value."); skip_to_semi (cfile); return (unsigned char *)0; } /* If we can, convert the number now; otherwise, build a linked list of all the numbers. */ if (s) { convert_num (s, val, base, size); s += size / 8; } else { t = (char *)malloc (strlen (val) + 1); if (!t) error ("no temp space for number."); strlcpy (t, val, strlen(val)+1); c = cons (t, c); } } while (++count != *max); /* If we had to cons up a list, convert it now. */ if (c) { bufp = (unsigned char *)malloc (count * size / 8); if (!bufp) error ("can't allocate space for numeric aggregate."); s = bufp + count - size / 8; *max = count; } while (c) { pair cdr = c -> cdr; convert_num (s, (char *)(c -> car), base, size); s -= size / 8; /* Free up temp space. */ free (c -> car); free (c); c = cdr; } return bufp; } void convert_num (buf, str, base, size) unsigned char *buf; char *str; int base; int size; { char *ptr = str; int negative = 0; u_int32_t val = 0; int tval; int max; if (*ptr == '-') { negative = 1; ++ptr; } /* If base wasn't specified, figure it out from the data. */ if (!base) { if (ptr [0] == '0') { if (ptr [1] == 'x') { base = 16; ptr += 2; } else if (isascii (ptr [1]) && isdigit (ptr [1])) { base = 8; ptr += 1; } else { base = 10; } } else { base = 10; } } do { tval = *ptr++; /* XXX assumes ASCII... */ if (tval >= 'a') tval = tval - 'a' + 10; else if (tval >= 'A') tval = tval - 'A' + 10; else if (tval >= '0') tval -= '0'; else { warn ("Bogus number: %s.", str); break; } if (tval >= base) { warn ("Bogus number: %s: digit %d not in base %d", str, tval, base); break; } val = val * base + tval; } while (*ptr); if (negative) max = (1 << (size - 1)); else max = (1 << (size - 1)) + ((1 << (size - 1)) - 1); if (val > max) { switch (base) { case 8: warn ("value %s%o exceeds max (%d) for precision.", negative ? "-" : "", val, max); break; case 16: warn ("value %s%x exceeds max (%d) for precision.", negative ? "-" : "", val, max); break; default: warn ("value %s%u exceeds max (%d) for precision.", negative ? "-" : "", val, max); break; } } if (negative) { switch (size) { case 8: *buf = -(unsigned long)val; break; case 16: putShort (buf, -(unsigned long)val); break; case 32: putLong (buf, -(unsigned long)val); break; default: warn ("Unexpected integer size: %d", size); break; } } else { switch (size) { case 8: *buf = (u_int8_t)val; break; case 16: putUShort (buf, (u_int16_t)val); break; case 32: putULong (buf, val); break; default: warn ("Unexpected integer size: %d", size); break; } } } /* date :== NUMBER NUMBER SLASH NUMBER SLASH NUMBER NUMBER COLON NUMBER COLON NUMBER SEMI Dates are always in GMT; first number is day of week; next is year/month/day; next is hours:minutes:seconds on a 24-hour clock. */ TIME parse_date (cfile) FILE *cfile; { struct tm tm; int guess; char *val; int token; static int months [11] = { 31, 59, 90, 120, 151, 181, 212, 243, 273, 304, 334 }; /* Day of week... */ token = next_token (&val, cfile); if (token != NUMBER) { parse_warn ("numeric day of week expected."); if (token != SEMI) skip_to_semi (cfile); return (TIME)0; } tm.tm_wday = atoi (val); /* Year... */ token = next_token (&val, cfile); if (token != NUMBER) { parse_warn ("numeric year expected."); if (token != SEMI) skip_to_semi (cfile); return (TIME)0; } tm.tm_year = atoi (val); if (tm.tm_year > 1900) tm.tm_year -= 1900; /* Slash separating year from month... */ token = next_token (&val, cfile); if (token != SLASH) { parse_warn ("expected slash separating year from month."); if (token != SEMI) skip_to_semi (cfile); return (TIME)0; } /* Month... */ token = next_token (&val, cfile); if (token != NUMBER) { parse_warn ("numeric month expected."); if (token != SEMI) skip_to_semi (cfile); return (TIME)0; } tm.tm_mon = atoi (val) - 1; /* Slash separating month from day... */ token = next_token (&val, cfile); if (token != SLASH) { parse_warn ("expected slash separating month from day."); if (token != SEMI) skip_to_semi (cfile); return (TIME)0; } /* Month... */ token = next_token (&val, cfile); if (token != NUMBER) { parse_warn ("numeric day of month expected."); if (token != SEMI) skip_to_semi (cfile); return (TIME)0; } tm.tm_mday = atoi (val); /* Hour... */ token = next_token (&val, cfile); if (token != NUMBER) { parse_warn ("numeric hour expected."); if (token != SEMI) skip_to_semi (cfile); return (TIME)0; } tm.tm_hour = atoi (val); /* Colon separating hour from minute... */ token = next_token (&val, cfile); if (token != COLON) { parse_warn ("expected colon separating hour from minute."); if (token != SEMI) skip_to_semi (cfile); return (TIME)0; } /* Minute... */ token = next_token (&val, cfile); if (token != NUMBER) { parse_warn ("numeric minute expected."); if (token != SEMI) skip_to_semi (cfile); return (TIME)0; } tm.tm_min = atoi (val); /* Colon separating minute from second... */ token = next_token (&val, cfile); if (token != COLON) { parse_warn ("expected colon separating hour from minute."); if (token != SEMI) skip_to_semi (cfile); return (TIME)0; } /* Minute... */ token = next_token (&val, cfile); if (token != NUMBER) { parse_warn ("numeric minute expected."); if (token != SEMI) skip_to_semi (cfile); return (TIME)0; } tm.tm_sec = atoi (val); tm.tm_isdst = 0; /* XXX */ /* We assume that mktime does not use tm_yday. */ tm.tm_yday = 0; /* Make sure the date ends in a semicolon... */ token = next_token (&val, cfile); if (token != SEMI) { parse_warn ("semicolon expected."); skip_to_semi (cfile); return 0; } /* Guess the time value... */ guess = ((((((365 * (tm.tm_year - 70) + /* Days in years since '70 */ (tm.tm_year - 69) / 4 + /* Leap days since '70 */ (tm.tm_mon /* Days in months this year */ ? months [tm.tm_mon - 1] : 0) + (tm.tm_mon > 1 && /* Leap day this year */ !((tm.tm_year - 72) & 3)) + tm.tm_mday - 1) * 24) + /* Day of month */ tm.tm_hour) * 60) + tm.tm_min) * 60) + tm.tm_sec; /* This guess could be wrong because of leap seconds or other weirdness we don't know about that the system does. For now, we're just going to accept the guess, but at some point it might be nice to do a successive approximation here to get an exact value. Even if the error is small, if the server is restarted frequently (and thus the lease database is reread), the error could accumulate into something significant. */ return guess; }