xref: /freebsd/contrib/ntp/ntpd/keyword-gen.c (revision 788ca347b816afd83b2885e0c79aeeb88649b2ab)
1 /*
2  * keyword-gen.c -- generate keyword scanner finite state machine and
3  *		    keyword_text array.
4  *
5  * This program is run to generate ntp_keyword.h
6  * After making a change here, two output files should be committed at
7  * the same time as keyword-gen.c:
8  *	ntp_keyword.h
9  *	keyword-gen-utd
10  *
11  * keyword-gen-utd is a sentinel used by Makefile.am to avoid compiling
12  * keyword_gen.c and generating ntp_keyword.h if the input keyword-gen.c
13  * has not changed.  This is not solely an optimization, it also breaks
14  * a dependency chain that otherwise would cause programs to be compiled
15  * when running "make dist" or "make distdir".  We want these to package
16  * the existing source without building anything but a tarball.  See
17  * [Bug 1470].
18  */
19 #include <config.h>
20 #include <stdio.h>
21 #include <stdlib.h>
22 #include <time.h>
23 
24 #include <ntp_stdlib.h>
25 #include <ntp_config.h>
26 #include <lib_strbuf.h>
27 #include "ntp_scanner.h"
28 #include "ntp_parser.h"
29 
30 
31 /* Define a structure to hold a (keyword, token) pair */
32 struct key_tok {
33 	char *	key;		/* Keyword */
34 	u_short	token;		/* Associated Token */
35 	follby	followedby;	/* nonzero indicates the next token(s)
36 				   forced to be string(s) */
37 };
38 
39 struct key_tok ntp_keywords[] = {
40 { "...",		T_Ellipsis,		FOLLBY_TOKEN },
41 { "allpeers",		T_Allpeers,		FOLLBY_TOKEN },
42 { "automax",		T_Automax,		FOLLBY_TOKEN },
43 { "broadcast",		T_Broadcast,		FOLLBY_STRING },
44 { "broadcastclient",	T_Broadcastclient,	FOLLBY_TOKEN },
45 { "broadcastdelay",	T_Broadcastdelay,	FOLLBY_TOKEN },
46 { "ctl",		T_Ctl,			FOLLBY_TOKEN },
47 { "disable",		T_Disable,		FOLLBY_TOKEN },
48 { "driftfile",		T_Driftfile,		FOLLBY_STRING },
49 { "enable",		T_Enable,		FOLLBY_TOKEN },
50 { "end",		T_End,			FOLLBY_TOKEN },
51 { "filegen",		T_Filegen,		FOLLBY_TOKEN },
52 { "fudge",		T_Fudge,		FOLLBY_STRING },
53 { "io",			T_Io,			FOLLBY_TOKEN },
54 { "includefile",	T_Includefile,		FOLLBY_STRING },
55 { "leapfile",		T_Leapfile,		FOLLBY_STRING },
56 { "logconfig",		T_Logconfig,		FOLLBY_STRINGS_TO_EOC },
57 { "logfile",		T_Logfile,		FOLLBY_STRING },
58 { "manycastclient",	T_Manycastclient,	FOLLBY_STRING },
59 { "manycastserver",	T_Manycastserver,	FOLLBY_STRINGS_TO_EOC },
60 { "mem",		T_Mem,			FOLLBY_TOKEN },
61 { "multicastclient",	T_Multicastclient,	FOLLBY_STRINGS_TO_EOC },
62 { "peer",		T_Peer,			FOLLBY_STRING },
63 { "phone",		T_Phone,		FOLLBY_STRINGS_TO_EOC },
64 { "pidfile",		T_Pidfile,		FOLLBY_STRING },
65 { "pool",		T_Pool,			FOLLBY_STRING },
66 { "discard",		T_Discard,		FOLLBY_TOKEN },
67 { "reset",		T_Reset,		FOLLBY_TOKEN },
68 { "restrict",		T_Restrict,		FOLLBY_TOKEN },
69 { "rlimit",		T_Rlimit,		FOLLBY_TOKEN },
70 { "server",		T_Server,		FOLLBY_STRING },
71 { "setvar",		T_Setvar,		FOLLBY_STRING },
72 { "statistics",		T_Statistics,		FOLLBY_TOKEN },
73 { "statsdir",		T_Statsdir,		FOLLBY_STRING },
74 { "sys",		T_Sys,			FOLLBY_TOKEN },
75 { "tick",		T_Tick,			FOLLBY_TOKEN },
76 { "timer",		T_Timer,		FOLLBY_TOKEN },
77 { "tinker",		T_Tinker,		FOLLBY_TOKEN },
78 { "tos",		T_Tos,			FOLLBY_TOKEN },
79 { "trap",		T_Trap,			FOLLBY_STRING },
80 { "unconfig",		T_Unconfig,		FOLLBY_STRING },
81 { "unpeer",		T_Unpeer,		FOLLBY_STRING },
82 /* authentication_command */
83 { "controlkey",		T_ControlKey,		FOLLBY_TOKEN },
84 { "crypto",		T_Crypto,		FOLLBY_TOKEN },
85 { "keys",		T_Keys,			FOLLBY_STRING },
86 { "keysdir",		T_Keysdir,		FOLLBY_STRING },
87 { "ntpsigndsocket",	T_NtpSignDsocket,	FOLLBY_STRING },
88 { "requestkey",		T_Requestkey,		FOLLBY_TOKEN },
89 { "revoke",		T_Revoke,		FOLLBY_TOKEN },
90 { "trustedkey",		T_Trustedkey,		FOLLBY_TOKEN },
91 /* IPv4/IPv6 protocol override flag */
92 { "-4",			T_Ipv4_flag,		FOLLBY_TOKEN },
93 { "-6",			T_Ipv6_flag,		FOLLBY_TOKEN },
94 /* option */
95 { "autokey",		T_Autokey,		FOLLBY_TOKEN },
96 { "burst",		T_Burst,		FOLLBY_TOKEN },
97 { "iburst",		T_Iburst,		FOLLBY_TOKEN },
98 { "key",		T_Key,			FOLLBY_TOKEN },
99 { "maxpoll",		T_Maxpoll,		FOLLBY_TOKEN },
100 { "mdnstries",		T_Mdnstries,		FOLLBY_TOKEN },
101 { "minpoll",		T_Minpoll,		FOLLBY_TOKEN },
102 { "mode",		T_Mode,			FOLLBY_TOKEN },
103 { "noselect",		T_Noselect,		FOLLBY_TOKEN },
104 { "preempt",		T_Preempt,		FOLLBY_TOKEN },
105 { "true",		T_True,			FOLLBY_TOKEN },
106 { "prefer",		T_Prefer,		FOLLBY_TOKEN },
107 { "ttl",		T_Ttl,			FOLLBY_TOKEN },
108 { "version",		T_Version,		FOLLBY_TOKEN },
109 { "xleave",		T_Xleave,		FOLLBY_TOKEN },
110 /* crypto_command */
111 { "host",		T_Host,			FOLLBY_STRING },
112 { "ident",		T_Ident,		FOLLBY_STRING },
113 { "pw",			T_Pw,			FOLLBY_STRING },
114 { "randfile",		T_Randfile,		FOLLBY_STRING },
115 { "digest",		T_Digest,		FOLLBY_STRING },
116 /*** MONITORING COMMANDS ***/
117 /* stat */
118 { "clockstats",		T_Clockstats,		FOLLBY_TOKEN },
119 { "cryptostats",	T_Cryptostats,		FOLLBY_TOKEN },
120 { "loopstats",		T_Loopstats,		FOLLBY_TOKEN },
121 { "peerstats",		T_Peerstats,		FOLLBY_TOKEN },
122 { "rawstats",		T_Rawstats,		FOLLBY_TOKEN },
123 { "sysstats", 		T_Sysstats,		FOLLBY_TOKEN },
124 { "protostats",		T_Protostats,		FOLLBY_TOKEN },
125 { "timingstats",	T_Timingstats,		FOLLBY_TOKEN },
126 /* filegen_option */
127 { "file",		T_File,			FOLLBY_STRING },
128 { "link",		T_Link,			FOLLBY_TOKEN },
129 { "nolink",		T_Nolink,		FOLLBY_TOKEN },
130 { "type",		T_Type,			FOLLBY_TOKEN },
131 /* filegen_type */
132 { "age",		T_Age,			FOLLBY_TOKEN },
133 { "day",		T_Day,			FOLLBY_TOKEN },
134 { "month",		T_Month,		FOLLBY_TOKEN },
135 { "none",		T_None,			FOLLBY_TOKEN },
136 { "pid",		T_Pid,			FOLLBY_TOKEN },
137 { "week",		T_Week,			FOLLBY_TOKEN },
138 { "year",		T_Year,			FOLLBY_TOKEN },
139 /*** ORPHAN MODE COMMANDS ***/
140 /* tos_option */
141 { "minclock",		T_Minclock,		FOLLBY_TOKEN },
142 { "maxclock",		T_Maxclock,		FOLLBY_TOKEN },
143 { "minsane",		T_Minsane,		FOLLBY_TOKEN },
144 { "floor",		T_Floor,		FOLLBY_TOKEN },
145 { "ceiling",		T_Ceiling,		FOLLBY_TOKEN },
146 { "cohort",		T_Cohort,		FOLLBY_TOKEN },
147 { "mindist",		T_Mindist,		FOLLBY_TOKEN },
148 { "maxdist",		T_Maxdist,		FOLLBY_TOKEN },
149 { "beacon",		T_Beacon,		FOLLBY_TOKEN },
150 { "orphan",		T_Orphan,		FOLLBY_TOKEN },
151 { "orphanwait",		T_Orphanwait,		FOLLBY_TOKEN },
152 { "nonvolatile",	T_Nonvolatile,		FOLLBY_TOKEN },
153 /* access_control_flag */
154 { "default",		T_Default,		FOLLBY_TOKEN },
155 { "source",		T_Source,		FOLLBY_TOKEN },
156 { "flake",		T_Flake,		FOLLBY_TOKEN },
157 { "ignore",		T_Ignore,		FOLLBY_TOKEN },
158 { "limited",		T_Limited,		FOLLBY_TOKEN },
159 { "mssntp",		T_Mssntp,		FOLLBY_TOKEN },
160 { "kod",		T_Kod,			FOLLBY_TOKEN },
161 { "lowpriotrap",	T_Lowpriotrap,		FOLLBY_TOKEN },
162 { "mask",		T_Mask,			FOLLBY_TOKEN },
163 { "nomodify",		T_Nomodify,		FOLLBY_TOKEN },
164 { "nomrulist",		T_Nomrulist,		FOLLBY_TOKEN },
165 { "nopeer",		T_Nopeer,		FOLLBY_TOKEN },
166 { "noquery",		T_Noquery,		FOLLBY_TOKEN },
167 { "noserve",		T_Noserve,		FOLLBY_TOKEN },
168 { "notrap",		T_Notrap,		FOLLBY_TOKEN },
169 { "notrust",		T_Notrust,		FOLLBY_TOKEN },
170 { "ntpport",		T_Ntpport,		FOLLBY_TOKEN },
171 /* discard_option */
172 { "average",		T_Average,		FOLLBY_TOKEN },
173 { "minimum",		T_Minimum,		FOLLBY_TOKEN },
174 { "monitor",		T_Monitor,		FOLLBY_TOKEN },
175 /* mru_option */
176 { "incalloc",		T_Incalloc,		FOLLBY_TOKEN },
177 { "incmem",		T_Incmem,		FOLLBY_TOKEN },
178 { "initalloc",		T_Initalloc,		FOLLBY_TOKEN },
179 { "initmem",		T_Initmem,		FOLLBY_TOKEN },
180 { "mindepth",		T_Mindepth,		FOLLBY_TOKEN },
181 { "maxage",		T_Maxage,		FOLLBY_TOKEN },
182 { "maxdepth",		T_Maxdepth,		FOLLBY_TOKEN },
183 { "maxmem",		T_Maxmem,		FOLLBY_TOKEN },
184 { "mru",		T_Mru,			FOLLBY_TOKEN },
185 /* fudge_factor */
186 { "abbrev",		T_Abbrev,		FOLLBY_STRING },
187 { "flag1",		T_Flag1,		FOLLBY_TOKEN },
188 { "flag2",		T_Flag2,		FOLLBY_TOKEN },
189 { "flag3",		T_Flag3,		FOLLBY_TOKEN },
190 { "flag4",		T_Flag4,		FOLLBY_TOKEN },
191 { "refid",		T_Refid,		FOLLBY_STRING },
192 { "stratum",		T_Stratum,		FOLLBY_TOKEN },
193 { "time1",		T_Time1,		FOLLBY_TOKEN },
194 { "time2",		T_Time2,		FOLLBY_TOKEN },
195 /* system_option */
196 { "auth",		T_Auth,			FOLLBY_TOKEN },
197 { "bclient",		T_Bclient,		FOLLBY_TOKEN },
198 { "calibrate",		T_Calibrate,		FOLLBY_TOKEN },
199 { "kernel",		T_Kernel,		FOLLBY_TOKEN },
200 { "ntp",		T_Ntp,			FOLLBY_TOKEN },
201 { "mode7",		T_Mode7,		FOLLBY_TOKEN },
202 { "stats",		T_Stats,		FOLLBY_TOKEN },
203 /* rlimit_option */
204 { "memlock",		T_Memlock,		FOLLBY_TOKEN },
205 { "stacksize",		T_Stacksize,		FOLLBY_TOKEN },
206 { "filenum",		T_Filenum,		FOLLBY_TOKEN },
207 /* tinker_option */
208 { "step",		T_Step,			FOLLBY_TOKEN },
209 { "panic",		T_Panic,		FOLLBY_TOKEN },
210 { "dispersion",		T_Dispersion,		FOLLBY_TOKEN },
211 { "stepout",		T_Stepout,		FOLLBY_TOKEN },
212 { "allan",		T_Allan,		FOLLBY_TOKEN },
213 { "huffpuff",		T_Huffpuff,		FOLLBY_TOKEN },
214 { "freq",		T_Freq,			FOLLBY_TOKEN },
215 /* miscellaneous_command */
216 { "port",		T_Port,			FOLLBY_TOKEN },
217 { "interface",		T_Interface,		FOLLBY_TOKEN },
218 { "saveconfigdir",	T_Saveconfigdir,	FOLLBY_STRING },
219 /* interface_command (ignore and interface already defined) */
220 { "nic",		T_Nic,			FOLLBY_TOKEN },
221 { "all",		T_All,			FOLLBY_TOKEN },
222 { "ipv4",		T_Ipv4,			FOLLBY_TOKEN },
223 { "ipv6",		T_Ipv6,			FOLLBY_TOKEN },
224 { "wildcard",		T_Wildcard,		FOLLBY_TOKEN },
225 { "listen",		T_Listen,		FOLLBY_TOKEN },
226 { "drop",		T_Drop,			FOLLBY_TOKEN },
227 /* simulator commands */
228 { "simulate",		T_Simulate,		FOLLBY_TOKEN },
229 { "simulation_duration",T_Sim_Duration,		FOLLBY_TOKEN },
230 { "beep_delay",		T_Beep_Delay,		FOLLBY_TOKEN },
231 { "duration",		T_Duration,		FOLLBY_TOKEN },
232 { "server_offset",	T_Server_Offset,	FOLLBY_TOKEN },
233 { "freq_offset",	T_Freq_Offset,		FOLLBY_TOKEN },
234 { "wander",		T_Wander,		FOLLBY_TOKEN },
235 { "jitter",		T_Jitter,		FOLLBY_TOKEN },
236 { "prop_delay",		T_Prop_Delay,		FOLLBY_TOKEN },
237 { "proc_delay",		T_Proc_Delay,		FOLLBY_TOKEN },
238 };
239 
240 typedef struct big_scan_state_tag {
241 	char	ch;		/* Character this state matches on */
242 	char	followedby;	/* Forces next token(s) to T_String */
243 	u_short	finishes_token;	/* nonzero ID if last keyword char */
244 	u_short	match_next_s;	/* next state to check matching ch */
245 	u_short	other_next_s;	/* next state to check if not ch */
246 } big_scan_state;
247 
248 /*
249  * Note: to increase MAXSTATES beyond 2048, be aware it is currently
250  * crammed into 11 bits in scan_state form.  Raising to 4096 would be
251  * relatively easy by storing the followedby value in a separate
252  * array with one entry per token, and shrinking the char value to
253  * 7 bits to free a bit for accepting/non-accepting.  More than 4096
254  * states will require expanding scan_state beyond 32 bits each.
255  */
256 #define MAXSTATES	2048
257 #define MAX_TOK_LEN	63
258 
259 const char *	current_keyword;/* for error reporting */
260 big_scan_state	sst[MAXSTATES];	/* scanner FSM state entries */
261 u_short		sst_highwater;	/* next entry index to consider */
262 char *		symb[1024];	/* map token ID to symbolic name */
263 
264 /* for libntp */
265 const char *	progname = "keyword-gen";
266 
267 int		main			(int, char **);
268 static void	generate_preamble	(void);
269 static void	generate_fsm		(void);
270 static void	generate_token_text	(void);
271 static u_short	create_keyword_scanner	(void);
272 static u_short	create_scan_states	(char *, u_short, follby, u_short);
273 int		compare_key_tok_id	(const void *, const void *);
274 int		compare_key_tok_text	(const void *, const void *);
275 void		populate_symb		(char *);
276 const char *	symbname		(u_short);
277 
278 
279 int main(int argc, char **argv)
280 {
281 	if (argc < 2) {
282 		fprintf(stderr, "Usage:\n%s t_header.h\n", argv[0]);
283 		exit(1);
284 	}
285 	debug = 1;
286 
287 	populate_symb(argv[1]);
288 
289 	generate_preamble();
290 	generate_token_text();
291 	generate_fsm();
292 
293 	return 0;
294 }
295 
296 
297 static void
298 generate_preamble(void)
299 {
300 	time_t now;
301 	char timestamp[128];
302 	char preamble[] =
303 "/*\n"
304 " * ntp_keyword.h\n"
305 " * \n"
306 " * NOTE: edit this file with caution, it is generated by keyword-gen.c\n"
307 " *\t Generated %s UTC	  diff_ignore_line\n"
308 " *\n"
309 " */\n"
310 "#include \"ntp_scanner.h\"\n"
311 "#include \"ntp_parser.h\"\n"
312 "\n";
313 
314 	time(&now);
315 	if (!strftime(timestamp, sizeof(timestamp),
316 		      "%Y-%m-%d %H:%M:%S", gmtime(&now)))
317 		timestamp[0] = '\0';
318 
319 	printf(preamble, timestamp);
320 }
321 
322 
323 static void
324 generate_fsm(void)
325 {
326 	char rprefix[MAX_TOK_LEN + 1];
327 	char prefix[MAX_TOK_LEN + 1];
328 	char token_id_comment[16 + MAX_TOK_LEN + 1];
329 	size_t prefix_len;
330 	char *p;
331 	char *r;
332 	u_short initial_state;
333 	u_short this_state;
334 	u_short prev_state;
335 	u_short state;
336 	u_short i;
337 	u_short token;
338 
339 	/*
340 	 * Sort ntp_keywords in alphabetical keyword order.  This is
341 	 * not necessary, but minimizes nonfunctional changes in the
342 	 * generated finite state machine when keywords are modified.
343 	 */
344 	qsort(ntp_keywords, COUNTOF(ntp_keywords),
345 	      sizeof(ntp_keywords[0]), compare_key_tok_text);
346 
347 	/*
348 	 * To save space, reserve the state array entry matching each
349 	 * token number for its terminal state, so the token identifier
350 	 * does not need to be stored in each state, but can be
351 	 * recovered trivially.  To mark the entry reserved,
352 	 * finishes_token is nonzero.
353 	 */
354 
355 	for (i = 0; i < COUNTOF(ntp_keywords); i++) {
356 		token = ntp_keywords[i].token;
357 		if (1 > token || token >= COUNTOF(sst)) {
358 			fprintf(stderr,
359 				"keyword-gen sst[%u] too small "
360 				"for keyword '%s' id %d\n",
361 				(int)COUNTOF(sst),
362 				ntp_keywords[i].key,
363 				token);
364 			exit(4);
365 		}
366 		sst[token].finishes_token = token;
367 	}
368 
369 	initial_state = create_keyword_scanner();
370 
371 	fprintf(stderr,
372 		"%d keywords consumed %d states of %d max.\n",
373 		(int)COUNTOF(ntp_keywords),
374 		sst_highwater - 1,
375 		(int)COUNTOF(sst) - 1);
376 
377 	printf("#define SCANNER_INIT_S %d\n\n", initial_state);
378 
379 	printf("const scan_state sst[%d] = {\n"
380 	       "/*SS_T( ch,\tf-by, match, other ),\t\t\t\t */\n"
381 	       "  0,\t\t\t\t      /* %5d %-17s */\n",
382 	       sst_highwater,
383 	       0, "");
384 
385 	for (i = 1; i < sst_highwater; i++) {
386 
387 		/* verify fields will fit */
388 		if (sst[i].followedby & ~0x3) {
389 			fprintf(stderr,
390 				"keyword-gen internal error "
391 				"sst[%d].followedby %d too big\n",
392 				i, sst[i].followedby);
393 			exit(7);
394 		}
395 
396 		if (sst_highwater <= sst[i].match_next_s
397 		    || sst[i].match_next_s & ~0x7ff) {
398 			fprintf(stderr,
399 				"keyword-gen internal error "
400 				"sst[%d].match_next_s %d too big\n",
401 				i, sst[i].match_next_s);
402 			exit(8);
403 		}
404 
405 		if (sst_highwater <= sst[i].other_next_s
406 		    || sst[i].other_next_s & ~0x7ff) {
407 			fprintf(stderr,
408 				"keyword-gen internal error "
409 				"sst[%d].other_next_s %d too big\n",
410 				i, sst[i].other_next_s);
411 			exit(9);
412 		}
413 
414 		if (sst[i].finishes_token) {
415 			snprintf(token_id_comment,
416 				 sizeof(token_id_comment), "%5d %-17s",
417 				 i, symbname(sst[i].finishes_token));
418 			if (i != sst[i].finishes_token) {
419 				fprintf(stderr,
420 					"keyword-gen internal error "
421 					"entry %d finishes token %d\n",
422 					i, sst[i].finishes_token);
423 				exit(5);
424 			}
425 		} else {
426 		/*
427 		 * Determine the keyword prefix that leads to this
428 		 * state.  This is expensive but keyword-gen is run
429 		 * only when it changes.  Distributing keyword-gen-utd
430 		 * achieves that, which is why it must be committed
431 		 * at the same time as keyword-gen.c and ntp_keyword.h.
432 		 *
433 		 * Scan the state array iteratively looking for a state
434 		 * which leads to the current one, collecting matching
435 		 * characters along the way.  There is only one such
436 		 * path back to the starting state given the way our
437 		 * scanner state machine is built and the practice of
438 		 * using the spelling of the keyword as its T_* token
439 		 * identifier, which results in never having two
440 		 * spellings result in the same T_* value.
441 		 */
442 			prefix_len = 0;
443 			prev_state = 0;
444 			this_state = i;
445 			do {
446 				for (state = 1; state < sst_highwater; state++)
447 					if (sst[state].other_next_s == this_state) {
448 						this_state = state;
449 						break;
450 					} else if (sst[state].match_next_s == this_state) {
451 						this_state = state;
452 						rprefix[prefix_len] = sst[state].ch;
453 						prefix_len++;
454 						break;
455 					}
456 			} while (this_state != initial_state);
457 
458 			if (prefix_len) {
459 				/* reverse rprefix into prefix */
460 				p = prefix + prefix_len;
461 				r = rprefix;
462 				while (r < rprefix + prefix_len)
463 					*--p = *r++;
464 			}
465 			prefix[prefix_len] = '\0';
466 
467 			snprintf(token_id_comment,
468 				 sizeof(token_id_comment), "%5d %-17s",
469 				 i, (initial_state == i)
470 					? "[initial state]"
471 					: prefix);
472 		}
473 
474 		printf("  S_ST( '%c',\t%d,    %5u, %5u )%s /* %s */\n",
475 		       sst[i].ch,
476 		       sst[i].followedby,
477 		       sst[i].match_next_s,
478 		       sst[i].other_next_s,
479 		       (i + 1 < sst_highwater)
480 			   ? ","
481 			   : " ",
482 		       token_id_comment);
483 	}
484 
485 	printf("};\n\n");
486 }
487 
488 
489 /* Define a function to create the states of the scanner. This function
490  * is used by the create_keyword_scanner function below.
491  *
492  * This function takes a suffix of a keyword, the token to be returned on
493  * recognizing the complete keyword, and any pre-existing state that exists
494  * for some other keyword that has the same prefix as the current one.
495  */
496 static u_short
497 create_scan_states(
498 	char *	text,
499 	u_short	token,
500 	follby	followedby,
501 	u_short	prev_state
502 	)
503 {
504 	u_short my_state;
505 	u_short return_state;
506 	u_short prev_char_s;
507 	u_short curr_char_s;
508 
509 	return_state = prev_state;
510 	curr_char_s = prev_state;
511 	prev_char_s = 0;
512 
513 	/* Find the correct position to insert the state.
514 	 * All states should be in alphabetical order
515 	 */
516 	while (curr_char_s && (text[0] < sst[curr_char_s].ch)) {
517 		prev_char_s = curr_char_s;
518 		curr_char_s = sst[curr_char_s].other_next_s;
519 	}
520 
521 	/*
522 	 * Check if a previously seen keyword has the same prefix as
523 	 * the current keyword.  If so, simply use the state for that
524 	 * keyword as my_state, otherwise, allocate a new state.
525 	 */
526 	if (curr_char_s && (text[0] == sst[curr_char_s].ch)) {
527 		my_state = curr_char_s;
528 		if ('\0' == text[1]) {
529 			fprintf(stderr,
530 				"Duplicate entries for keyword '%s' in"
531 				" keyword_gen.c ntp_keywords[].\n",
532 				current_keyword);
533 			exit(2);
534 		}
535 	} else {
536 		do
537 			my_state = sst_highwater++;
538 		while (my_state < COUNTOF(sst)
539 		       && sst[my_state].finishes_token);
540 		if (my_state >= COUNTOF(sst)) {
541 			fprintf(stderr,
542 				"fatal, keyword scanner state array "
543 				"sst[%d] is too small, modify\n"
544 				"keyword-gen.c to increase.\n",
545 				(int)COUNTOF(sst));
546 			exit(3);
547 		}
548 		/* Store the next character of the keyword */
549 		sst[my_state].ch = text[0];
550 		sst[my_state].other_next_s = curr_char_s;
551 		sst[my_state].followedby = FOLLBY_NON_ACCEPTING;
552 
553 		if (prev_char_s)
554 			sst[prev_char_s].other_next_s = my_state;
555 		else
556 			return_state = my_state;
557 	}
558 
559 	/* Check if the next character is '\0'.
560 	 * If yes, we are done with the recognition and this is an accepting
561 	 * state.
562 	 * If not, we need to continue scanning
563 	 */
564 	if ('\0' == text[1]) {
565 		sst[my_state].finishes_token = (u_short)token;
566 		sst[my_state].followedby = (char)followedby;
567 
568 		if (sst[token].finishes_token != (u_short)token) {
569 			fprintf(stderr,
570 				"fatal, sst[%d] not reserved for %s.\n",
571 				token, symbname(token));
572 			exit(6);
573 		}
574 		/* relocate so token id is sst[] index */
575 		if (my_state != token) {
576 			sst[token] = sst[my_state];
577 			ZERO(sst[my_state]);
578 			do
579 				sst_highwater--;
580 			while (sst[sst_highwater].finishes_token);
581 			my_state = token;
582 			if (prev_char_s)
583 				sst[prev_char_s].other_next_s = my_state;
584 			else
585 				return_state = my_state;
586 		}
587 	} else
588 		sst[my_state].match_next_s =
589 		    create_scan_states(
590 			&text[1],
591 			token,
592 			followedby,
593 			sst[my_state].match_next_s);
594 
595 	return return_state;
596 }
597 
598 
599 /* Define a function that takes a list of (keyword, token) values and
600  * creates a keywords scanner out of it.
601  */
602 
603 static u_short
604 create_keyword_scanner(void)
605 {
606 	u_short scanner;
607 	u_short i;
608 
609 	sst_highwater = 1;	/* index 0 invalid, unused */
610 	scanner = 0;
611 
612 	for (i = 0; i < COUNTOF(ntp_keywords); i++) {
613 		current_keyword = ntp_keywords[i].key;
614 		scanner =
615 		    create_scan_states(
616 			ntp_keywords[i].key,
617 			ntp_keywords[i].token,
618 			ntp_keywords[i].followedby,
619 			scanner);
620 	}
621 
622 	return scanner;
623 }
624 
625 
626 static void
627 generate_token_text(void)
628 {
629 	u_short lowest_id;
630 	u_short highest_id;
631 	u_short id_count;
632 	u_short id;
633 	u_short i;
634 
635 	/* sort ntp_keywords in token ID order */
636 	qsort(ntp_keywords, COUNTOF(ntp_keywords),
637 	      sizeof(ntp_keywords[0]), compare_key_tok_id);
638 
639 	lowest_id = ntp_keywords[0].token;
640 	highest_id = ntp_keywords[COUNTOF(ntp_keywords) - 1].token;
641 	id_count = highest_id - lowest_id + 1;
642 
643 	printf("#define LOWEST_KEYWORD_ID %d\n\n", lowest_id);
644 
645 	printf("const char * const keyword_text[%d] = {", id_count);
646 
647 	id = lowest_id;
648 	i = 0;
649 	while (i < COUNTOF(ntp_keywords)) {
650 		while (id < ntp_keywords[i].token) {
651 			printf(",\n\t/* %-5d %5d %20s */\tNULL",
652 			       id - lowest_id, id, symbname(id));
653 			id++;
654 		}
655 		if (i > 0)
656 			printf(",");
657 		printf("\n\t/* %-5d %5d %20s */\t\"%s\"",
658 		       id - lowest_id, id, symbname(id),
659 		       ntp_keywords[i].key);
660 		i++;
661 		id++;
662 	}
663 
664 	printf("\n};\n\n");
665 }
666 
667 
668 int
669 compare_key_tok_id(
670 	const void *a1,
671 	const void *a2
672 	)
673 {
674 	const struct key_tok *p1 = a1;
675 	const struct key_tok *p2 = a2;
676 
677 	if (p1->token == p2->token)
678 		return 0;
679 
680 	if (p1->token < p2->token)
681 		return -1;
682 	else
683 		return 1;
684 }
685 
686 
687 int
688 compare_key_tok_text(
689 	const void *a1,
690 	const void *a2
691 	)
692 {
693 	const struct key_tok *p1 = a1;
694 	const struct key_tok *p2 = a2;
695 
696 	return strcmp(p1->key, p2->key);
697 }
698 
699 
700 /*
701  * populate_symb() - populate symb[] lookup array with symbolic token
702  *		     names such that symb[T_Age] == "T_Age", etc.
703  */
704 void
705 populate_symb(
706 	char *header_file
707 	)
708 {
709 	FILE *	yh;
710 	char	line[2 * MAX_TOK_LEN];
711 	char	name[2 * MAX_TOK_LEN];
712 	int	token;
713 
714 	yh = fopen(header_file, "r");
715 	if (NULL == yh) {
716 		perror("unable to open yacc/bison header file");
717 		exit(4);
718 	}
719 
720 	while (NULL != fgets(line, sizeof(line), yh))
721 		if (2 == sscanf(line, "#define %s %d", name, &token)
722 		    && 'T' == name[0] && '_' == name[1] && token >= 0
723 		    && token < COUNTOF(symb)) {
724 
725 			symb[token] = estrdup(name);
726 			if (strlen(name) > MAX_TOK_LEN) {
727 				fprintf(stderr,
728 					"MAX_TOK_LEN %d too small for '%s'\n"
729 					"Edit keyword-gen.c to raise.\n",
730 					MAX_TOK_LEN, name);
731 				exit(10);
732 			}
733 		}
734 	fclose(yh);
735 }
736 
737 
738 const char *
739 symbname(
740 	u_short token
741 	)
742 {
743 	char *name;
744 
745 	if (token < COUNTOF(symb) && symb[token] != NULL) {
746 		name = symb[token];
747 	} else {
748 		LIB_GETBUF(name);
749 		snprintf(name, LIB_BUFLENGTH, "%d", token);
750 	}
751 
752 	return name;
753 }
754