/* conflex.c Lexical scanner for dhcpd config file... */ /* * Copyright (c) 2004-2005 by Internet Systems Consortium, Inc. ("ISC") * Copyright (c) 1995-2003 by Internet Software Consortium * * Permission to use, copy, modify, and distribute this software for any * purpose with or without fee is hereby granted, provided that the above * copyright notice and this permission notice appear in all copies. * * THE SOFTWARE IS PROVIDED "AS IS" AND ISC DISCLAIMS ALL WARRANTIES * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL ISC BE LIABLE FOR * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT * OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. * * Internet Systems Consortium, Inc. * 950 Charter Street * Redwood City, CA 94063 * * http://www.isc.org/ * * This software has been written for Internet Systems Consortium * by Ted Lemon in cooperation with Vixie Enterprises and Nominum, Inc. * To learn more about Internet Systems Consortium, see * ``http://www.isc.org/''. To learn more about Vixie Enterprises, * see ``http://www.vix.com''. To learn more about Nominum, Inc., see * ``http://www.nominum.com''. */ #ifndef lint static char copyright[] = "$Id: conflex.c,v 1.6 2010/01/17 22:42:36 wiz Exp $ Copyright (c) 2004-2005 Internet Systems Consortium. All rights reserved.\n"; #endif /* not lint */ #include "dhcpd.h" #include static int get_char PROTO ((struct parse *)); static enum dhcp_token get_token PROTO ((struct parse *)); static void skip_to_eol PROTO ((struct parse *)); static enum dhcp_token read_string PROTO ((struct parse *)); static enum dhcp_token read_number PROTO ((int, struct parse *)); static enum dhcp_token read_num_or_name PROTO ((int, struct parse *)); static enum dhcp_token intern PROTO ((char *, enum dhcp_token)); isc_result_t new_parse (cfile, file, inbuf, buflen, name, eolp) struct parse **cfile; int file; char *inbuf; unsigned buflen; const char *name; int eolp; { struct parse *tmp; tmp = dmalloc (sizeof (struct parse), MDL); if (!tmp) return ISC_R_NOMEMORY; memset (tmp, 0, sizeof *tmp); tmp -> token = 0; tmp -> tlname = name; tmp -> lpos = tmp -> line = 1; tmp -> cur_line = tmp -> line1; tmp -> prev_line = tmp -> line2; tmp -> token_line = tmp -> cur_line; tmp -> cur_line [0] = tmp -> prev_line [0] = 0; tmp -> warnings_occurred = 0; tmp -> file = file; tmp -> eol_token = eolp; tmp -> bufix = 0; tmp -> buflen = buflen; if (inbuf) { tmp -> bufsiz = 0; tmp -> inbuf = inbuf; } else { tmp -> inbuf = dmalloc (8192, MDL); if (!tmp -> inbuf) { dfree (tmp, MDL); return ISC_R_NOMEMORY; } tmp -> bufsiz = 8192; } *cfile = tmp; return ISC_R_SUCCESS; } isc_result_t end_parse (cfile) struct parse **cfile; { if ((*cfile) -> bufsiz) dfree ((*cfile) -> inbuf, MDL); dfree (*cfile, MDL); *cfile = (struct parse *)0; return ISC_R_SUCCESS; } static int get_char (cfile) struct parse *cfile; { /* My kingdom for WITH... */ int c; if (cfile -> bufix == cfile -> buflen) { if (cfile -> file != -1) { cfile -> buflen = read (cfile -> file, cfile -> inbuf, cfile -> bufsiz); if (cfile -> buflen == 0) { c = EOF; cfile -> bufix = 0; } else if (cfile -> buflen < 0) { c = EOF; cfile -> bufix = cfile -> buflen = 0; } else { c = cfile -> inbuf [0]; cfile -> bufix = 1; } } else c = EOF; } else { c = cfile -> inbuf [cfile -> bufix]; cfile -> bufix++; } if (!cfile -> ugflag) { if (c == EOL) { if (cfile -> cur_line == cfile -> line1) { cfile -> cur_line = cfile -> line2; cfile -> prev_line = cfile -> line1; } else { cfile -> cur_line = cfile -> line1; cfile -> prev_line = cfile -> line2; } cfile -> line++; cfile -> lpos = 1; cfile -> cur_line [0] = 0; } else if (c != EOF) { if (cfile -> lpos <= 80) { cfile -> cur_line [cfile -> lpos - 1] = c; cfile -> cur_line [cfile -> lpos] = 0; } cfile -> lpos++; } } else cfile -> ugflag = 0; return c; } static enum dhcp_token get_token (cfile) struct parse *cfile; { int c; enum dhcp_token ttok; static char tb [2]; int l, p, u; do { l = cfile -> line; p = cfile -> lpos; u = cfile -> ugflag; c = get_char (cfile); #ifdef OLD_LEXER if (c == '\n' && p == 1 && !u && cfile -> comment_index < sizeof cfile -> comments) cfile -> comments [cfile -> comment_index++] = '\n'; #endif if (!(c == '\n' && cfile -> eol_token) && isascii (c) && isspace (c)) continue; if (c == '#') { #ifdef OLD_LEXER if (cfile -> comment_index < sizeof cfile -> comments) cfile -> comments [cfile -> comment_index++] = '#'; #endif skip_to_eol (cfile); continue; } if (c == '"') { cfile -> lexline = l; cfile -> lexchar = p; ttok = read_string (cfile); break; } if ((isascii (c) && isdigit (c)) || c == '-') { cfile -> lexline = l; cfile -> lexchar = p; ttok = read_number (c, cfile); break; } else if (isascii (c) && isalpha (c)) { cfile -> lexline = l; cfile -> lexchar = p; ttok = read_num_or_name (c, cfile); break; } else if (c == EOF) { ttok = END_OF_FILE; cfile -> tlen = 0; break; } else { cfile -> lexline = l; cfile -> lexchar = p; tb [0] = c; tb [1] = 0; cfile -> tval = tb; cfile -> tlen = 1; ttok = c; break; } } while (1); return ttok; } enum dhcp_token next_token (rval, rlen, cfile) const char **rval; unsigned *rlen; struct parse *cfile; { int rv; if (cfile -> token) { if (cfile -> lexline != cfile -> tline) cfile -> token_line = cfile -> cur_line; cfile -> lexchar = cfile -> tlpos; cfile -> lexline = cfile -> tline; rv = cfile -> token; cfile -> token = 0; } else { rv = get_token (cfile); cfile -> token_line = cfile -> cur_line; } if (rval) *rval = cfile -> tval; if (rlen) *rlen = cfile -> tlen; #ifdef DEBUG_TOKENS fprintf (stderr, "%s:%d ", cfile -> tval, rv); #endif return rv; } enum dhcp_token peek_token (rval, rlen, cfile) const char **rval; unsigned int *rlen; struct parse *cfile; { int x; if (!cfile -> token) { cfile -> tlpos = cfile -> lexchar; cfile -> tline = cfile -> lexline; cfile -> token = get_token (cfile); if (cfile -> lexline != cfile -> tline) cfile -> token_line = cfile -> prev_line; x = cfile -> lexchar; cfile -> lexchar = cfile -> tlpos; cfile -> tlpos = x; x = cfile -> lexline; cfile -> lexline = cfile -> tline; cfile -> tline = x; } if (rval) *rval = cfile -> tval; if (rlen) *rlen = cfile -> tlen; #ifdef DEBUG_TOKENS fprintf (stderr, "(%s:%d) ", cfile -> tval, cfile -> token); #endif return cfile -> token; } static void skip_to_eol (cfile) struct parse *cfile; { int c; do { c = get_char (cfile); if (c == EOF) return; #ifdef OLD_LEXER if (cfile -> comment_index < sizeof (cfile -> comments)) comments [cfile -> comment_index++] = c; #endif if (c == EOL) { return; } } while (1); } static enum dhcp_token read_string (cfile) struct parse *cfile; { int i; int bs = 0; int c; int value = 0; int hex = 0; value = 0; /* XXXGCC -Wuninitialized */ hex = 0; /* XXXGCC -Wuninitialized */ for (i = 0; i < sizeof cfile -> tokbuf; i++) { again: c = get_char (cfile); if (c == EOF) { parse_warn (cfile, "eof in string constant"); break; } if (bs == 1) { switch (c) { case 't': cfile -> tokbuf [i] = '\t'; break; case 'r': cfile -> tokbuf [i] = '\r'; break; case 'n': cfile -> tokbuf [i] = '\n'; break; case 'b': cfile -> tokbuf [i] = '\b'; break; case '0': case '1': case '2': case '3': hex = 0; value = c - '0'; ++bs; goto again; case 'x': hex = 1; value = 0; ++bs; goto again; default: cfile -> tokbuf [i] = c; bs = 0; break; } bs = 0; } else if (bs > 1) { if (hex) { if (c >= '0' && c <= '9') { value = value * 16 + (c - '0'); } else if (c >= 'a' && c <= 'f') { value = value * 16 + (c - 'a' + 10); } else if (c >= 'A' && c <= 'F') { value = value * 16 + (c - 'A' + 10); } else { parse_warn (cfile, "invalid hex digit: %x", c); bs = 0; continue; } if (++bs == 4) { cfile -> tokbuf [i] = value; bs = 0; } else goto again; } else { if (c >= '0' && c <= '9') { value = value * 8 + (c - '0'); } else { if (value != 0) { parse_warn (cfile, "invalid octal digit %x", c); continue; } else cfile -> tokbuf [i] = 0; bs = 0; } if (++bs == 4) { cfile -> tokbuf [i] = value; bs = 0; } else goto again; } } else if (c == '\\') { bs = 1; goto again; } else if (c == '"') break; else cfile -> tokbuf [i] = c; } /* Normally, I'd feel guilty about this, but we're talking about strings that'll fit in a DHCP packet here... */ if (i == sizeof cfile -> tokbuf) { parse_warn (cfile, "string constant larger than internal buffer"); --i; } cfile -> tokbuf [i] = 0; cfile -> tlen = i; cfile -> tval = cfile -> tokbuf; return STRING; } static enum dhcp_token read_number (c, cfile) int c; struct parse *cfile; { #ifdef OLD_LEXER int seenx = 0; #endif int i = 0; int token = NUMBER; cfile -> tokbuf [i++] = c; for (; i < sizeof cfile -> tokbuf; i++) { c = get_char (cfile); #ifndef OLD_LEXER /* Promote NUMBER -> NUMBER_OR_NAME -> NAME, never demote. * Except in the case of '0x' syntax hex, which gets called * a NAME at '0x', and returned to NUMBER_OR_NAME once it's * verified to be at least 0xf or less. */ switch(isascii(c) ? token : BREAK) { case NUMBER: if(isdigit(c)) break; /* FALLTHROUGH */ case NUMBER_OR_NAME: if(isxdigit(c)) { token = NUMBER_OR_NAME; break; } /* FALLTHROUGH */ case NAME: if((i == 2) && isxdigit(c) && (cfile->tokbuf[0] == '0') && ((cfile->tokbuf[1] == 'x') || (cfile->tokbuf[1] == 'X'))) { token = NUMBER_OR_NAME; break; } else if(((c == '-') || (c == '_') || isalnum(c))) { token = NAME; break; } /* FALLTHROUGH */ case BREAK: /* At this point c is either EOF or part of the next * token. If not EOF, rewind the file one byte so * the next token is read from there. */ if(c != EOF) { cfile->bufix--; cfile->ugflag = 1; } goto end_read; default: log_fatal("read_number():%s:%d: impossible case", MDL); } #else /* OLD_LEXER */ if (!seenx && (c == 'x')) { seenx = 1; } else if (!isascii (c) || !isxdigit (c)) { if (c != EOF) { cfile -> bufix--; cfile -> ugflag = 1; } break; } #endif /* OLD_LEXER */ cfile -> tokbuf [i] = c; } if (i == sizeof cfile -> tokbuf) { parse_warn (cfile, "numeric token larger than internal buffer"); --i; } end_read: cfile -> tokbuf [i] = 0; cfile -> tlen = i; cfile -> tval = cfile -> tokbuf; return token; } static enum dhcp_token read_num_or_name (c, cfile) int c; struct parse *cfile; { int i = 0; enum dhcp_token rv = NUMBER_OR_NAME; cfile -> tokbuf [i++] = c; for (; i < sizeof cfile -> tokbuf; i++) { c = get_char (cfile); if (!isascii (c) || (c != '-' && c != '_' && !isalnum (c))) { if (c != EOF) { cfile -> bufix--; cfile -> ugflag = 1; } break; } if (!isxdigit (c)) rv = NAME; cfile -> tokbuf [i] = c; } if (i == sizeof cfile -> tokbuf) { parse_warn (cfile, "token larger than internal buffer"); --i; } cfile -> tokbuf [i] = 0; cfile -> tlen = i; cfile -> tval = cfile -> tokbuf; return intern (cfile -> tval, rv); } static enum dhcp_token intern (atom, dfv) char *atom; enum dhcp_token dfv; { if (!isascii (atom [0])) return dfv; switch (tolower ((unsigned char)atom [0])) { case '-': if (atom [1] == 0) return MINUS; break; case 'a': if (!strncasecmp (atom + 1, "uth", 3)) { if (!strncasecmp (atom + 3, "uthenticat", 10)) { if (!strcasecmp (atom + 13, "ed")) return AUTHENTICATED; if (!strcasecmp (atom + 13, "ion")) return AUTHENTICATION; break; } if (!strcasecmp (atom + 1, "uthoritative")) return AUTHORITATIVE; break; } if (!strcasecmp (atom + 1, "nd")) return AND; if (!strcasecmp (atom + 1, "ppend")) return APPEND; if (!strcasecmp (atom + 1, "llow")) return ALLOW; if (!strcasecmp (atom + 1, "lias")) return ALIAS; if (!strcasecmp (atom + 1, "lgorithm")) return ALGORITHM; if (!strcasecmp (atom + 1, "bandoned")) return TOKEN_ABANDONED; if (!strcasecmp (atom + 1, "dd")) return TOKEN_ADD; if (!strcasecmp (atom + 1, "ll")) return ALL; if (!strcasecmp (atom + 1, "t")) return AT; if (!strcasecmp (atom + 1, "rray")) return ARRAY; if (!strcasecmp (atom + 1, "ddress")) return ADDRESS; if (!strcasecmp (atom + 1, "ctive")) return TOKEN_ACTIVE; break; case 'b': if (!strcasecmp (atom + 1, "ackup")) return TOKEN_BACKUP; if (!strcasecmp (atom + 1, "ootp")) return TOKEN_BOOTP; if (!strcasecmp (atom + 1, "inding")) return BINDING; if (!strcasecmp (atom + 1, "inary-to-ascii")) return BINARY_TO_ASCII; if (!strcasecmp (atom + 1, "ackoff-cutoff")) return BACKOFF_CUTOFF; if (!strcasecmp (atom + 1, "ooting")) return BOOTING; if (!strcasecmp (atom + 1, "oot-unknown-clients")) return BOOT_UNKNOWN_CLIENTS; if (!strcasecmp (atom + 1, "reak")) return BREAK; if (!strcasecmp (atom + 1, "illing")) return BILLING; if (!strcasecmp (atom + 1, "oolean")) return BOOLEAN; if (!strcasecmp (atom + 1, "alance")) return BALANCE; if (!strcasecmp (atom + 1, "ound")) return BOUND; break; case 'c': if (!strcasecmp (atom + 1, "ase")) return CASE; if (!strcasecmp (atom + 1, "ommit")) return COMMIT; if (!strcasecmp (atom + 1, "ode")) return CODE; if (!strcasecmp (atom + 1, "onfig-option")) return CONFIG_OPTION; if (!strcasecmp (atom + 1, "heck")) return CHECK; if (!strcasecmp (atom + 1, "lass")) return CLASS; if (!strcasecmp (atom + 1, "lose")) return TOKEN_CLOSE; if (!strcasecmp (atom + 1, "reate")) return TOKEN_CREATE; if (!strcasecmp (atom + 1, "iaddr")) return CIADDR; if (!strncasecmp (atom + 1, "lient", 5)) { if (!strcasecmp (atom + 6, "-identifier")) return CLIENT_IDENTIFIER; if (!strcasecmp (atom + 6, "-hostname")) return CLIENT_HOSTNAME; if (!strcasecmp (atom + 6, "-state")) return CLIENT_STATE; if (!strcasecmp (atom + 6, "-updates")) return CLIENT_UPDATES; if (!strcasecmp (atom + 6, "s")) return CLIENTS; } if (!strcasecmp (atom + 1, "oncat")) return CONCAT; if (!strcasecmp (atom + 1, "onnect")) return CONNECT; if (!strcasecmp (atom + 1, "ommunications-interrupted")) return COMMUNICATIONS_INTERRUPTED; if (!strcasecmp (atom + 1, "ltt")) return CLTT; break; case 'd': if (!strcasecmp (atom + 1, "ns-update")) return DNS_UPDATE; if (!strcasecmp (atom + 1, "ns-delete")) return DNS_DELETE; if (!strcasecmp (atom + 1, "omain")) return DOMAIN; if (!strcasecmp (atom + 1, "omain-name")) return DOMAIN_NAME; if (!strcasecmp (atom + 1, "o-forward-update")) return DO_FORWARD_UPDATE; if (!strcasecmp (atom + 1, "ebug")) return TOKEN_DEBUG; if (!strcasecmp (atom + 1, "eny")) return DENY; if (!strcasecmp (atom + 1, "eleted")) return TOKEN_DELETED; if (!strcasecmp (atom + 1, "elete")) return TOKEN_DELETE; if (!strncasecmp (atom + 1, "efault", 6)) { if (!atom [7]) return DEFAULT; if (!strcasecmp (atom + 7, "-lease-time")) return DEFAULT_LEASE_TIME; break; } if (!strncasecmp (atom + 1, "ynamic", 6)) { if (!atom [7]) return DYNAMIC; if (!strncasecmp (atom + 7, "-bootp", 6)) { if (!atom [13]) return DYNAMIC_BOOTP; if (!strcasecmp (atom + 13, "-lease-cutoff")) return DYNAMIC_BOOTP_LEASE_CUTOFF; if (!strcasecmp (atom + 13, "-lease-length")) return DYNAMIC_BOOTP_LEASE_LENGTH; break; } } if (!strcasecmp (atom + 1, "uplicates")) return DUPLICATES; if (!strcasecmp (atom + 1, "eclines")) return DECLINES; if (!strncasecmp (atom + 1, "efine", 5)) { if (!strcasecmp (atom + 6, "d")) return DEFINED; if (!atom [6]) return DEFINE; } break; case 'e': if (tolower ((unsigned char)atom [1]) == 'x') { if (!strcasecmp (atom + 2, "tract-int")) return EXTRACT_INT; if (!strcasecmp (atom + 2, "ists")) return EXISTS; if (!strcasecmp (atom + 2, "piry")) return EXPIRY; if (!strcasecmp (atom + 2, "pire")) return EXPIRE; if (!strcasecmp (atom + 2, "pired")) return TOKEN_EXPIRED; } if (!strcasecmp (atom + 1, "ncode-int")) return ENCODE_INT; if (!strcasecmp (atom + 1, "thernet")) return ETHERNET; if (!strcasecmp (atom + 1, "nds")) return ENDS; if (!strncasecmp (atom + 1, "ls", 2)) { if (!strcasecmp (atom + 3, "e")) return ELSE; if (!strcasecmp (atom + 3, "if")) return ELSIF; break; } if (!strcasecmp (atom + 1, "rror")) return ERROR; if (!strcasecmp (atom + 1, "val")) return EVAL; if (!strcasecmp (atom + 1, "ncapsulate")) return ENCAPSULATE; break; case 'f': if (!strcasecmp (atom + 1, "atal")) return FATAL; if (!strcasecmp (atom + 1, "ilename")) return FILENAME; if (!strcasecmp (atom + 1, "ixed-address")) return FIXED_ADDR; if (!strcasecmp (atom + 1, "ddi")) return FDDI; if (!strcasecmp (atom + 1, "ormerr")) return NS_FORMERR; if (!strcasecmp (atom + 1, "unction")) return FUNCTION; if (!strcasecmp (atom + 1, "ailover")) return FAILOVER; if (!strcasecmp (atom + 1, "ree")) return TOKEN_FREE; break; case 'g': if (!strcasecmp (atom + 1, "iaddr")) return GIADDR; if (!strcasecmp (atom + 1, "roup")) return GROUP; if (!strcasecmp (atom + 1, "et-lease-hostnames")) return GET_LEASE_HOSTNAMES; break; case 'h': if (!strcasecmp (atom + 1, "ba")) return HBA; if (!strcasecmp (atom + 1, "ost")) return HOST; if (!strcasecmp (atom + 1, "ost-decl-name")) return HOST_DECL_NAME; if (!strcasecmp (atom + 1, "ardware")) return HARDWARE; if (!strcasecmp (atom + 1, "ostname")) return HOSTNAME; if (!strcasecmp (atom + 1, "elp")) return TOKEN_HELP; break; case 'i': if (!strcasecmp (atom + 1, "nclude")) return INCLUDE; if (!strcasecmp (atom + 1, "nteger")) return INTEGER; if (!strcasecmp (atom + 1, "nfinite")) return INFINITE; if (!strcasecmp (atom + 1, "nfo")) return INFO; if (!strcasecmp (atom + 1, "p-address")) return IP_ADDRESS; if (!strcasecmp (atom + 1, "nitial-interval")) return INITIAL_INTERVAL; if (!strcasecmp (atom + 1, "nterface")) return INTERFACE; if (!strcasecmp (atom + 1, "dentifier")) return IDENTIFIER; if (!strcasecmp (atom + 1, "f")) return IF; if (!strcasecmp (atom + 1, "s")) return IS; if (!strcasecmp (atom + 1, "gnore")) return IGNORE; break; case 'k': if (!strncasecmp (atom + 1, "nown", 4)) { if (!strcasecmp (atom + 5, "-clients")) return KNOWN_CLIENTS; if (!atom[5]) return KNOWN; break; } if (!strcasecmp (atom + 1, "ey")) return KEY; break; case 'l': if (!strcasecmp (atom + 1, "ease")) return LEASE; if (!strcasecmp (atom + 1, "eased-address")) return LEASED_ADDRESS; if (!strcasecmp (atom + 1, "ease-time")) return LEASE_TIME; if (!strcasecmp (atom + 1, "imit")) return LIMIT; if (!strcasecmp (atom + 1, "et")) return LET; if (!strcasecmp (atom + 1, "oad")) return LOAD; if (!strcasecmp (atom + 1, "og")) return LOG; break; case 'm': if (!strncasecmp (atom + 1, "ax", 2)) { if (!atom [3]) return TOKEN_MAX; if (!strcasecmp (atom + 3, "-lease-time")) return MAX_LEASE_TIME; if (!strcasecmp (atom + 3, "-transmit-idle")) return MAX_TRANSMIT_IDLE; if (!strcasecmp (atom + 3, "-response-delay")) return MAX_RESPONSE_DELAY; if (!strcasecmp (atom + 3, "-unacked-updates")) return MAX_UNACKED_UPDATES; } if (!strncasecmp (atom + 1, "in-", 3)) { if (!strcasecmp (atom + 4, "lease-time")) return MIN_LEASE_TIME; if (!strcasecmp (atom + 4, "secs")) return MIN_SECS; break; } if (!strncasecmp (atom + 1, "edi", 3)) { if (!strcasecmp (atom + 4, "a")) return MEDIA; if (!strcasecmp (atom + 4, "um")) return MEDIUM; break; } if (!strcasecmp (atom + 1, "atch")) return MATCH; if (!strcasecmp (atom + 1, "embers")) return MEMBERS; if (!strcasecmp (atom + 1, "y")) return MY; if (!strcasecmp (atom + 1, "clt")) return MCLT; break; case 'n': if (!strcasecmp (atom + 1, "ormal")) return NORMAL; if (!strcasecmp (atom + 1, "ameserver")) return NAMESERVER; if (!strcasecmp (atom + 1, "etmask")) return NETMASK; if (!strcasecmp (atom + 1, "ever")) return NEVER; if (!strcasecmp (atom + 1, "ext-server")) return NEXT_SERVER; if (!strcasecmp (atom + 1, "ot")) return TOKEN_NOT; if (!strcasecmp (atom + 1, "o")) return NO; if (!strcasecmp (atom + 1, "s-update")) return NS_UPDATE; if (!strcasecmp (atom + 1, "oerror")) return NS_NOERROR; if (!strcasecmp (atom + 1, "otauth")) return NS_NOTAUTH; if (!strcasecmp (atom + 1, "otimp")) return NS_NOTIMP; if (!strcasecmp (atom + 1, "otzone")) return NS_NOTZONE; if (!strcasecmp (atom + 1, "xdomain")) return NS_NXDOMAIN; if (!strcasecmp (atom + 1, "xrrset")) return NS_NXRRSET; if (!strcasecmp (atom + 1, "ull")) return TOKEN_NULL; if (!strcasecmp (atom + 1, "ext")) return TOKEN_NEXT; if (!strcasecmp (atom + 1, "ew")) return TOKEN_NEW; break; case 'o': if (!strcasecmp (atom + 1, "mapi")) return OMAPI; if (!strcasecmp (atom + 1, "r")) return OR; if (!strcasecmp (atom + 1, "n")) return ON; if (!strcasecmp (atom + 1, "pen")) return TOKEN_OPEN; if (!strcasecmp (atom + 1, "ption")) return OPTION; if (!strcasecmp (atom + 1, "ne-lease-per-client")) return ONE_LEASE_PER_CLIENT; if (!strcasecmp (atom + 1, "f")) return OF; if (!strcasecmp (atom + 1, "wner")) return OWNER; break; case 'p': if (!strcasecmp (atom + 1, "repend")) return PREPEND; if (!strcasecmp (atom + 1, "acket")) return PACKET; if (!strcasecmp (atom + 1, "ool")) return POOL; if (!strcasecmp (atom + 1, "seudo")) return PSEUDO; if (!strcasecmp (atom + 1, "eer")) return PEER; if (!strcasecmp (atom + 1, "rimary")) return PRIMARY; if (!strncasecmp (atom + 1, "artner", 6)) { if (!atom [7]) return PARTNER; if (!strcasecmp (atom + 7, "-down")) return PARTNER_DOWN; } if (!strcasecmp (atom + 1, "ort")) return PORT; if (!strcasecmp (atom + 1, "otential-conflict")) return POTENTIAL_CONFLICT; if (!strcasecmp (atom + 1, "ick-first-value") || !strcasecmp (atom + 1, "ick")) return PICK; if (!strcasecmp (atom + 1, "aused")) return PAUSED; break; case 'r': if (!strcasecmp (atom + 1, "esolution-interrupted")) return RESOLUTION_INTERRUPTED; if (!strcasecmp (atom + 1, "ange")) return RANGE; if (!strcasecmp (atom + 1, "ecover")) return RECOVER; if (!strcasecmp (atom + 1, "ecover-done")) return RECOVER_DONE; if (!strcasecmp (atom + 1, "ecover-wait")) return RECOVER_WAIT; if (!strcasecmp (atom + 1, "econtact-interval")) return RECONTACT_INTERVAL; if (!strcasecmp (atom + 1, "equest")) return REQUEST; if (!strcasecmp (atom + 1, "equire")) return REQUIRE; if (!strcasecmp (atom + 1, "equire")) return REQUIRE; if (!strcasecmp (atom + 1, "etry")) return RETRY; if (!strcasecmp (atom + 1, "eturn")) return RETURN; if (!strcasecmp (atom + 1, "enew")) return RENEW; if (!strcasecmp (atom + 1, "ebind")) return REBIND; if (!strcasecmp (atom + 1, "eboot")) return REBOOT; if (!strcasecmp (atom + 1, "eject")) return REJECT; if (!strcasecmp (atom + 1, "everse")) return REVERSE; if (!strcasecmp (atom + 1, "elease")) return RELEASE; if (!strcasecmp (atom + 1, "efused")) return NS_REFUSED; if (!strcasecmp (atom + 1, "eleased")) return TOKEN_RELEASED; if (!strcasecmp (atom + 1, "eset")) return TOKEN_RESET; if (!strcasecmp (atom + 1, "eserved")) return TOKEN_RESERVED; if (!strcasecmp (atom + 1, "emove")) return REMOVE; if (!strcasecmp (atom + 1, "efresh")) return REFRESH; break; case 's': if (!strcasecmp (atom + 1, "tate")) return STATE; if (!strcasecmp (atom + 1, "ecret")) return SECRET; if (!strcasecmp (atom + 1, "ervfail")) return NS_SERVFAIL; if (!strcasecmp (atom + 1, "witch")) return SWITCH; if (!strcasecmp (atom + 1, "igned")) return SIGNED; if (!strcasecmp (atom + 1, "tring")) return STRING_TOKEN; if (!strcasecmp (atom + 1, "uffix")) return SUFFIX; if (!strcasecmp (atom + 1, "earch")) return SEARCH; if (!strcasecmp (atom + 1, "tarts")) return STARTS; if (!strcasecmp (atom + 1, "iaddr")) return SIADDR; if (!strcasecmp (atom + 1, "hared-network")) return SHARED_NETWORK; if (!strcasecmp (atom + 1, "econdary")) return SECONDARY; if (!strcasecmp (atom + 1, "erver-name")) return SERVER_NAME; if (!strcasecmp (atom + 1, "erver-identifier")) return SERVER_IDENTIFIER; if (!strcasecmp (atom + 1, "erver")) return SERVER; if (!strcasecmp (atom + 1, "elect-timeout")) return SELECT_TIMEOUT; if (!strcasecmp (atom + 1, "elect")) return SELECT; if (!strcasecmp (atom + 1, "end")) return SEND; if (!strcasecmp (atom + 1, "cript")) return SCRIPT; if (!strcasecmp (atom + 1, "upersede")) return SUPERSEDE; if (!strncasecmp (atom + 1, "ub", 2)) { if (!strcasecmp (atom + 3, "string")) return SUBSTRING; if (!strcasecmp (atom + 3, "net")) return SUBNET; if (!strcasecmp (atom + 3, "class")) return SUBCLASS; break; } if (!strcasecmp (atom + 1, "pawn")) return SPAWN; if (!strcasecmp (atom + 1, "pace")) return SPACE; if (!strcasecmp (atom + 1, "tatic")) return STATIC; if (!strcasecmp (atom + 1, "plit")) return SPLIT; if (!strcasecmp (atom + 1, "et")) return TOKEN_SET; if (!strcasecmp (atom + 1, "econds")) return SECONDS; if (!strcasecmp (atom + 1, "hutdown")) return SHUTDOWN; if (!strcasecmp (atom + 1, "tartup")) return STARTUP; break; case 't': if (!strcasecmp (atom + 1, "imestamp")) return TIMESTAMP; if (!strcasecmp (atom + 1, "imeout")) return TIMEOUT; if (!strcasecmp (atom + 1, "oken-ring")) return TOKEN_RING; if (!strcasecmp (atom + 1, "ext")) return TEXT; if (!strcasecmp (atom + 1, "stp")) return TSTP; if (!strcasecmp (atom + 1, "sfp")) return TSFP; if (!strcasecmp (atom + 1, "ransmission")) return TRANSMISSION; break; case 'u': if (!strcasecmp (atom + 1, "nset")) return UNSET; if (!strcasecmp (atom + 1, "nsigned")) return UNSIGNED; if (!strcasecmp (atom + 1, "id")) return UID; if (!strncasecmp (atom + 1, "se", 2)) { if (!strcasecmp (atom + 3, "r-class")) return USER_CLASS; if (!strcasecmp (atom + 3, "-host-decl-names")) return USE_HOST_DECL_NAMES; if (!strcasecmp (atom + 3, "-lease-addr-for-default-route")) return USE_LEASE_ADDR_FOR_DEFAULT_ROUTE; break; } if (!strncasecmp (atom + 1, "nknown", 6)) { if (!strcasecmp (atom + 7, "-clients")) return UNKNOWN_CLIENTS; if (!strcasecmp (atom + 7, "-state")) return UNKNOWN_STATE; if (!atom [7]) return UNKNOWN; break; } if (!strcasecmp (atom + 1, "nauthenticated")) return AUTHENTICATED; if (!strcasecmp (atom + 1, "pdated-dns-rr")) return UPDATED_DNS_RR; if (!strcasecmp (atom + 1, "pdate")) return UPDATE; break; case 'v': if (!strcasecmp (atom + 1, "endor-class")) return VENDOR_CLASS; if (!strcasecmp (atom + 1, "endor")) return VENDOR; break; case 'w': if (!strcasecmp (atom + 1, "ith")) return WITH; break; case 'y': if (!strcasecmp (atom + 1, "iaddr")) return YIADDR; if (!strcasecmp (atom + 1, "xdomain")) return NS_YXDOMAIN; if (!strcasecmp (atom + 1, "xrrset")) return NS_YXRRSET; break; case 'z': if (!strcasecmp (atom + 1, "one")) return ZONE; break; } return dfv; }