xref: /freebsd/contrib/ntp/ntpd/keyword-gen.c (revision 09a53ad8f1318c5daae6cfb19d97f4f6459f0013)
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 { "dscp",		T_Dscp,			FOLLBY_TOKEN },
50 { "enable",		T_Enable,		FOLLBY_TOKEN },
51 { "end",		T_End,			FOLLBY_TOKEN },
52 { "filegen",		T_Filegen,		FOLLBY_TOKEN },
53 { "fudge",		T_Fudge,		FOLLBY_STRING },
54 { "io",			T_Io,			FOLLBY_TOKEN },
55 { "includefile",	T_Includefile,		FOLLBY_STRING },
56 { "leapfile",		T_Leapfile,		FOLLBY_STRING },
57 { "leapsmearinterval",	T_Leapsmearinterval,	FOLLBY_TOKEN },
58 { "logconfig",		T_Logconfig,		FOLLBY_STRINGS_TO_EOC },
59 { "logfile",		T_Logfile,		FOLLBY_STRING },
60 { "manycastclient",	T_Manycastclient,	FOLLBY_STRING },
61 { "manycastserver",	T_Manycastserver,	FOLLBY_STRINGS_TO_EOC },
62 { "mem",		T_Mem,			FOLLBY_TOKEN },
63 { "multicastclient",	T_Multicastclient,	FOLLBY_STRINGS_TO_EOC },
64 { "peer",		T_Peer,			FOLLBY_STRING },
65 { "phone",		T_Phone,		FOLLBY_STRINGS_TO_EOC },
66 { "pidfile",		T_Pidfile,		FOLLBY_STRING },
67 { "pool",		T_Pool,			FOLLBY_STRING },
68 { "discard",		T_Discard,		FOLLBY_TOKEN },
69 { "reset",		T_Reset,		FOLLBY_TOKEN },
70 { "restrict",		T_Restrict,		FOLLBY_TOKEN },
71 { "rlimit",		T_Rlimit,		FOLLBY_TOKEN },
72 { "server",		T_Server,		FOLLBY_STRING },
73 { "setvar",		T_Setvar,		FOLLBY_STRING },
74 { "statistics",		T_Statistics,		FOLLBY_TOKEN },
75 { "statsdir",		T_Statsdir,		FOLLBY_STRING },
76 { "sys",		T_Sys,			FOLLBY_TOKEN },
77 { "tick",		T_Tick,			FOLLBY_TOKEN },
78 { "timer",		T_Timer,		FOLLBY_TOKEN },
79 { "tinker",		T_Tinker,		FOLLBY_TOKEN },
80 { "tos",		T_Tos,			FOLLBY_TOKEN },
81 { "trap",		T_Trap,			FOLLBY_STRING },
82 { "unconfig",		T_Unconfig,		FOLLBY_STRING },
83 { "unpeer",		T_Unpeer,		FOLLBY_STRING },
84 /* authentication_command */
85 { "controlkey",		T_ControlKey,		FOLLBY_TOKEN },
86 { "crypto",		T_Crypto,		FOLLBY_TOKEN },
87 { "keys",		T_Keys,			FOLLBY_STRING },
88 { "keysdir",		T_Keysdir,		FOLLBY_STRING },
89 { "ntpsigndsocket",	T_NtpSignDsocket,	FOLLBY_STRING },
90 { "requestkey",		T_Requestkey,		FOLLBY_TOKEN },
91 { "revoke",		T_Revoke,		FOLLBY_TOKEN },
92 { "trustedkey",		T_Trustedkey,		FOLLBY_TOKEN },
93 /* IPv4/IPv6 protocol override flag */
94 { "-4",			T_Ipv4_flag,		FOLLBY_TOKEN },
95 { "-6",			T_Ipv6_flag,		FOLLBY_TOKEN },
96 /* option */
97 { "autokey",		T_Autokey,		FOLLBY_TOKEN },
98 { "burst",		T_Burst,		FOLLBY_TOKEN },
99 { "iburst",		T_Iburst,		FOLLBY_TOKEN },
100 { "key",		T_Key,			FOLLBY_TOKEN },
101 { "maxpoll",		T_Maxpoll,		FOLLBY_TOKEN },
102 { "mdnstries",		T_Mdnstries,		FOLLBY_TOKEN },
103 { "minpoll",		T_Minpoll,		FOLLBY_TOKEN },
104 { "mode",		T_Mode,			FOLLBY_TOKEN },
105 { "noselect",		T_Noselect,		FOLLBY_TOKEN },
106 { "preempt",		T_Preempt,		FOLLBY_TOKEN },
107 { "true",		T_True,			FOLLBY_TOKEN },
108 { "prefer",		T_Prefer,		FOLLBY_TOKEN },
109 { "ttl",		T_Ttl,			FOLLBY_TOKEN },
110 { "version",		T_Version,		FOLLBY_TOKEN },
111 { "xleave",		T_Xleave,		FOLLBY_TOKEN },
112 /* crypto_command */
113 { "host",		T_Host,			FOLLBY_STRING },
114 { "ident",		T_Ident,		FOLLBY_STRING },
115 { "pw",			T_Pw,			FOLLBY_STRING },
116 { "randfile",		T_Randfile,		FOLLBY_STRING },
117 { "digest",		T_Digest,		FOLLBY_STRING },
118 /*** MONITORING COMMANDS ***/
119 /* stat */
120 { "clockstats",		T_Clockstats,		FOLLBY_TOKEN },
121 { "cryptostats",	T_Cryptostats,		FOLLBY_TOKEN },
122 { "loopstats",		T_Loopstats,		FOLLBY_TOKEN },
123 { "peerstats",		T_Peerstats,		FOLLBY_TOKEN },
124 { "rawstats",		T_Rawstats,		FOLLBY_TOKEN },
125 { "sysstats", 		T_Sysstats,		FOLLBY_TOKEN },
126 { "protostats",		T_Protostats,		FOLLBY_TOKEN },
127 { "timingstats",	T_Timingstats,		FOLLBY_TOKEN },
128 /* filegen_option */
129 { "file",		T_File,			FOLLBY_STRING },
130 { "link",		T_Link,			FOLLBY_TOKEN },
131 { "nolink",		T_Nolink,		FOLLBY_TOKEN },
132 { "type",		T_Type,			FOLLBY_TOKEN },
133 /* filegen_type */
134 { "age",		T_Age,			FOLLBY_TOKEN },
135 { "day",		T_Day,			FOLLBY_TOKEN },
136 { "month",		T_Month,		FOLLBY_TOKEN },
137 { "none",		T_None,			FOLLBY_TOKEN },
138 { "pid",		T_Pid,			FOLLBY_TOKEN },
139 { "week",		T_Week,			FOLLBY_TOKEN },
140 { "year",		T_Year,			FOLLBY_TOKEN },
141 /*** ORPHAN MODE COMMANDS ***/
142 /* tos_option */
143 { "minclock",		T_Minclock,		FOLLBY_TOKEN },
144 { "maxclock",		T_Maxclock,		FOLLBY_TOKEN },
145 { "minsane",		T_Minsane,		FOLLBY_TOKEN },
146 { "floor",		T_Floor,		FOLLBY_TOKEN },
147 { "ceiling",		T_Ceiling,		FOLLBY_TOKEN },
148 { "cohort",		T_Cohort,		FOLLBY_TOKEN },
149 { "mindist",		T_Mindist,		FOLLBY_TOKEN },
150 { "maxdist",		T_Maxdist,		FOLLBY_TOKEN },
151 { "beacon",		T_Beacon,		FOLLBY_TOKEN },
152 { "orphan",		T_Orphan,		FOLLBY_TOKEN },
153 { "orphanwait",		T_Orphanwait,		FOLLBY_TOKEN },
154 { "nonvolatile",	T_Nonvolatile,		FOLLBY_TOKEN },
155 /* access_control_flag */
156 { "default",		T_Default,		FOLLBY_TOKEN },
157 { "source",		T_Source,		FOLLBY_TOKEN },
158 { "flake",		T_Flake,		FOLLBY_TOKEN },
159 { "ignore",		T_Ignore,		FOLLBY_TOKEN },
160 { "limited",		T_Limited,		FOLLBY_TOKEN },
161 { "mssntp",		T_Mssntp,		FOLLBY_TOKEN },
162 { "kod",		T_Kod,			FOLLBY_TOKEN },
163 { "lowpriotrap",	T_Lowpriotrap,		FOLLBY_TOKEN },
164 { "mask",		T_Mask,			FOLLBY_TOKEN },
165 { "nomodify",		T_Nomodify,		FOLLBY_TOKEN },
166 { "nomrulist",		T_Nomrulist,		FOLLBY_TOKEN },
167 { "nopeer",		T_Nopeer,		FOLLBY_TOKEN },
168 { "noquery",		T_Noquery,		FOLLBY_TOKEN },
169 { "noserve",		T_Noserve,		FOLLBY_TOKEN },
170 { "notrap",		T_Notrap,		FOLLBY_TOKEN },
171 { "notrust",		T_Notrust,		FOLLBY_TOKEN },
172 { "ntpport",		T_Ntpport,		FOLLBY_TOKEN },
173 /* discard_option */
174 { "average",		T_Average,		FOLLBY_TOKEN },
175 { "minimum",		T_Minimum,		FOLLBY_TOKEN },
176 { "monitor",		T_Monitor,		FOLLBY_TOKEN },
177 /* mru_option */
178 { "incalloc",		T_Incalloc,		FOLLBY_TOKEN },
179 { "incmem",		T_Incmem,		FOLLBY_TOKEN },
180 { "initalloc",		T_Initalloc,		FOLLBY_TOKEN },
181 { "initmem",		T_Initmem,		FOLLBY_TOKEN },
182 { "mindepth",		T_Mindepth,		FOLLBY_TOKEN },
183 { "maxage",		T_Maxage,		FOLLBY_TOKEN },
184 { "maxdepth",		T_Maxdepth,		FOLLBY_TOKEN },
185 { "maxmem",		T_Maxmem,		FOLLBY_TOKEN },
186 { "mru",		T_Mru,			FOLLBY_TOKEN },
187 /* fudge_factor */
188 { "abbrev",		T_Abbrev,		FOLLBY_STRING },
189 { "flag1",		T_Flag1,		FOLLBY_TOKEN },
190 { "flag2",		T_Flag2,		FOLLBY_TOKEN },
191 { "flag3",		T_Flag3,		FOLLBY_TOKEN },
192 { "flag4",		T_Flag4,		FOLLBY_TOKEN },
193 { "refid",		T_Refid,		FOLLBY_STRING },
194 { "stratum",		T_Stratum,		FOLLBY_TOKEN },
195 { "time1",		T_Time1,		FOLLBY_TOKEN },
196 { "time2",		T_Time2,		FOLLBY_TOKEN },
197 /* system_option */
198 { "auth",		T_Auth,			FOLLBY_TOKEN },
199 { "bclient",		T_Bclient,		FOLLBY_TOKEN },
200 { "calibrate",		T_Calibrate,		FOLLBY_TOKEN },
201 { "kernel",		T_Kernel,		FOLLBY_TOKEN },
202 { "mode7",		T_Mode7,		FOLLBY_TOKEN },
203 { "ntp",		T_Ntp,			FOLLBY_TOKEN },
204 { "peer_clear_digest_early",	T_PCEdigest,	FOLLBY_TOKEN },
205 { "stats",		T_Stats,		FOLLBY_TOKEN },
206 { "unpeer_crypto_early",	T_UEcrypto,	FOLLBY_TOKEN },
207 { "unpeer_crypto_nak_early",	T_UEcryptonak,	FOLLBY_TOKEN },
208 { "unpeer_digest_early",	T_UEdigest,	FOLLBY_TOKEN },
209 /* rlimit_option */
210 { "memlock",		T_Memlock,		FOLLBY_TOKEN },
211 { "stacksize",		T_Stacksize,		FOLLBY_TOKEN },
212 { "filenum",		T_Filenum,		FOLLBY_TOKEN },
213 /* tinker_option */
214 { "step",		T_Step,			FOLLBY_TOKEN },
215 { "stepback",		T_Stepback,		FOLLBY_TOKEN },
216 { "stepfwd",		T_Stepfwd,		FOLLBY_TOKEN },
217 { "panic",		T_Panic,		FOLLBY_TOKEN },
218 { "dispersion",		T_Dispersion,		FOLLBY_TOKEN },
219 { "stepout",		T_Stepout,		FOLLBY_TOKEN },
220 { "allan",		T_Allan,		FOLLBY_TOKEN },
221 { "huffpuff",		T_Huffpuff,		FOLLBY_TOKEN },
222 { "freq",		T_Freq,			FOLLBY_TOKEN },
223 /* miscellaneous_command */
224 { "port",		T_Port,			FOLLBY_TOKEN },
225 { "interface",		T_Interface,		FOLLBY_TOKEN },
226 { "saveconfigdir",	T_Saveconfigdir,	FOLLBY_STRING },
227 /* interface_command (ignore and interface already defined) */
228 { "nic",		T_Nic,			FOLLBY_TOKEN },
229 { "all",		T_All,			FOLLBY_TOKEN },
230 { "ipv4",		T_Ipv4,			FOLLBY_TOKEN },
231 { "ipv6",		T_Ipv6,			FOLLBY_TOKEN },
232 { "wildcard",		T_Wildcard,		FOLLBY_TOKEN },
233 { "listen",		T_Listen,		FOLLBY_TOKEN },
234 { "drop",		T_Drop,			FOLLBY_TOKEN },
235 /* simulator commands */
236 { "simulate",		T_Simulate,		FOLLBY_TOKEN },
237 { "simulation_duration",T_Sim_Duration,		FOLLBY_TOKEN },
238 { "beep_delay",		T_Beep_Delay,		FOLLBY_TOKEN },
239 { "duration",		T_Duration,		FOLLBY_TOKEN },
240 { "server_offset",	T_Server_Offset,	FOLLBY_TOKEN },
241 { "freq_offset",	T_Freq_Offset,		FOLLBY_TOKEN },
242 { "wander",		T_Wander,		FOLLBY_TOKEN },
243 { "jitter",		T_Jitter,		FOLLBY_TOKEN },
244 { "prop_delay",		T_Prop_Delay,		FOLLBY_TOKEN },
245 { "proc_delay",		T_Proc_Delay,		FOLLBY_TOKEN },
246 };
247 
248 typedef struct big_scan_state_tag {
249 	char	ch;		/* Character this state matches on */
250 	char	followedby;	/* Forces next token(s) to T_String */
251 	u_short	finishes_token;	/* nonzero ID if last keyword char */
252 	u_short	match_next_s;	/* next state to check matching ch */
253 	u_short	other_next_s;	/* next state to check if not ch */
254 } big_scan_state;
255 
256 /*
257  * Note: to increase MAXSTATES beyond 2048, be aware it is currently
258  * crammed into 11 bits in scan_state form.  Raising to 4096 would be
259  * relatively easy by storing the followedby value in a separate
260  * array with one entry per token, and shrinking the char value to
261  * 7 bits to free a bit for accepting/non-accepting.  More than 4096
262  * states will require expanding scan_state beyond 32 bits each.
263  */
264 #define MAXSTATES	2048
265 #define MAX_TOK_LEN	63
266 
267 const char *	current_keyword;/* for error reporting */
268 big_scan_state	sst[MAXSTATES];	/* scanner FSM state entries */
269 u_short		sst_highwater;	/* next entry index to consider */
270 char *		symb[1024];	/* map token ID to symbolic name */
271 
272 /* for libntp */
273 const char *	progname = "keyword-gen";
274 
275 int		main			(int, char **);
276 static void	generate_preamble	(void);
277 static void	generate_fsm		(void);
278 static void	generate_token_text	(void);
279 static u_short	create_keyword_scanner	(void);
280 static u_short	create_scan_states	(char *, u_short, follby, u_short);
281 int		compare_key_tok_id	(const void *, const void *);
282 int		compare_key_tok_text	(const void *, const void *);
283 void		populate_symb		(char *);
284 const char *	symbname		(u_short);
285 
286 
287 int main(int argc, char **argv)
288 {
289 	if (argc < 2) {
290 		fprintf(stderr, "Usage:\n%s t_header.h\n", argv[0]);
291 		exit(1);
292 	}
293 	debug = 1;
294 
295 	populate_symb(argv[1]);
296 
297 	generate_preamble();
298 	generate_token_text();
299 	generate_fsm();
300 
301 	return 0;
302 }
303 
304 
305 static void
306 generate_preamble(void)
307 {
308 	time_t now;
309 	char timestamp[128];
310 	char preamble[] =
311 "/*\n"
312 " * ntp_keyword.h\n"
313 " * \n"
314 " * NOTE: edit this file with caution, it is generated by keyword-gen.c\n"
315 " *\t Generated %s UTC	  diff_ignore_line\n"
316 " *\n"
317 " */\n"
318 "#include \"ntp_scanner.h\"\n"
319 "#include \"ntp_parser.h\"\n"
320 "\n";
321 
322 	time(&now);
323 	if (!strftime(timestamp, sizeof(timestamp),
324 		      "%Y-%m-%d %H:%M:%S", gmtime(&now)))
325 		timestamp[0] = '\0';
326 
327 	printf(preamble, timestamp);
328 }
329 
330 
331 static void
332 generate_fsm(void)
333 {
334 	char rprefix[MAX_TOK_LEN + 1];
335 	char prefix[MAX_TOK_LEN + 1];
336 	char token_id_comment[16 + MAX_TOK_LEN + 1];
337 	size_t prefix_len;
338 	char *p;
339 	char *r;
340 	u_short initial_state;
341 	u_short this_state;
342 	u_short state;
343 	u_short i;
344 	u_short token;
345 
346 	/*
347 	 * Sort ntp_keywords in alphabetical keyword order.  This is
348 	 * not necessary, but minimizes nonfunctional changes in the
349 	 * generated finite state machine when keywords are modified.
350 	 */
351 	qsort(ntp_keywords, COUNTOF(ntp_keywords),
352 	      sizeof(ntp_keywords[0]), compare_key_tok_text);
353 
354 	/*
355 	 * To save space, reserve the state array entry matching each
356 	 * token number for its terminal state, so the token identifier
357 	 * does not need to be stored in each state, but can be
358 	 * recovered trivially.  To mark the entry reserved,
359 	 * finishes_token is nonzero.
360 	 */
361 
362 	for (i = 0; i < COUNTOF(ntp_keywords); i++) {
363 		token = ntp_keywords[i].token;
364 		if (1 > token || token >= COUNTOF(sst)) {
365 			fprintf(stderr,
366 				"keyword-gen sst[%u] too small "
367 				"for keyword '%s' id %d\n",
368 				(int)COUNTOF(sst),
369 				ntp_keywords[i].key,
370 				token);
371 			exit(4);
372 		}
373 		sst[token].finishes_token = token;
374 	}
375 
376 	initial_state = create_keyword_scanner();
377 
378 	fprintf(stderr,
379 		"%d keywords consumed %d states of %d max.\n",
380 		(int)COUNTOF(ntp_keywords),
381 		sst_highwater - 1,
382 		(int)COUNTOF(sst) - 1);
383 
384 	printf("#define SCANNER_INIT_S %d\n\n", initial_state);
385 
386 	printf("const scan_state sst[%d] = {\n"
387 	       "/*SS_T( ch,\tf-by, match, other ),\t\t\t\t */\n"
388 	       "  0,\t\t\t\t      /* %5d %-17s */\n",
389 	       sst_highwater,
390 	       0, "");
391 
392 	for (i = 1; i < sst_highwater; i++) {
393 
394 		/* verify fields will fit */
395 		if (sst[i].followedby & ~0x3) {
396 			fprintf(stderr,
397 				"keyword-gen internal error "
398 				"sst[%d].followedby %d too big\n",
399 				i, sst[i].followedby);
400 			exit(7);
401 		}
402 
403 		if (sst_highwater <= sst[i].match_next_s
404 		    || sst[i].match_next_s & ~0x7ff) {
405 			fprintf(stderr,
406 				"keyword-gen internal error "
407 				"sst[%d].match_next_s %d too big\n",
408 				i, sst[i].match_next_s);
409 			exit(8);
410 		}
411 
412 		if (sst_highwater <= sst[i].other_next_s
413 		    || sst[i].other_next_s & ~0x7ff) {
414 			fprintf(stderr,
415 				"keyword-gen internal error "
416 				"sst[%d].other_next_s %d too big\n",
417 				i, sst[i].other_next_s);
418 			exit(9);
419 		}
420 
421 		if (sst[i].finishes_token) {
422 			snprintf(token_id_comment,
423 				 sizeof(token_id_comment), "%5d %-17s",
424 				 i, symbname(sst[i].finishes_token));
425 			if (i != sst[i].finishes_token) {
426 				fprintf(stderr,
427 					"keyword-gen internal error "
428 					"entry %d finishes token %d\n",
429 					i, sst[i].finishes_token);
430 				exit(5);
431 			}
432 		} else {
433 		/*
434 		 * Determine the keyword prefix that leads to this
435 		 * state.  This is expensive but keyword-gen is run
436 		 * only when it changes.  Distributing keyword-gen-utd
437 		 * achieves that, which is why it must be committed
438 		 * at the same time as keyword-gen.c and ntp_keyword.h.
439 		 *
440 		 * Scan the state array iteratively looking for a state
441 		 * which leads to the current one, collecting matching
442 		 * characters along the way.  There is only one such
443 		 * path back to the starting state given the way our
444 		 * scanner state machine is built and the practice of
445 		 * using the spelling of the keyword as its T_* token
446 		 * identifier, which results in never having two
447 		 * spellings result in the same T_* value.
448 		 */
449 			prefix_len = 0;
450 			this_state = i;
451 			do {
452 				for (state = 1; state < sst_highwater; state++)
453 					if (sst[state].other_next_s == this_state) {
454 						this_state = state;
455 						break;
456 					} else if (sst[state].match_next_s == this_state) {
457 						this_state = state;
458 						rprefix[prefix_len] = sst[state].ch;
459 						prefix_len++;
460 						break;
461 					}
462 			} while (this_state != initial_state);
463 
464 			if (prefix_len) {
465 				/* reverse rprefix into prefix */
466 				p = prefix + prefix_len;
467 				r = rprefix;
468 				while (r < rprefix + prefix_len)
469 					*--p = *r++;
470 			}
471 			prefix[prefix_len] = '\0';
472 
473 			snprintf(token_id_comment,
474 				 sizeof(token_id_comment), "%5d %-17s",
475 				 i, (initial_state == i)
476 					? "[initial state]"
477 					: prefix);
478 		}
479 
480 		printf("  S_ST( '%c',\t%d,    %5u, %5u )%s /* %s */\n",
481 		       sst[i].ch,
482 		       sst[i].followedby,
483 		       sst[i].match_next_s,
484 		       sst[i].other_next_s,
485 		       (i + 1 < sst_highwater)
486 			   ? ","
487 			   : " ",
488 		       token_id_comment);
489 	}
490 
491 	printf("};\n\n");
492 }
493 
494 
495 /* Define a function to create the states of the scanner. This function
496  * is used by the create_keyword_scanner function below.
497  *
498  * This function takes a suffix of a keyword, the token to be returned on
499  * recognizing the complete keyword, and any pre-existing state that exists
500  * for some other keyword that has the same prefix as the current one.
501  */
502 static u_short
503 create_scan_states(
504 	char *	text,
505 	u_short	token,
506 	follby	followedby,
507 	u_short	prev_state
508 	)
509 {
510 	u_short my_state;
511 	u_short return_state;
512 	u_short prev_char_s;
513 	u_short curr_char_s;
514 
515 	return_state = prev_state;
516 	curr_char_s = prev_state;
517 	prev_char_s = 0;
518 
519 	/* Find the correct position to insert the state.
520 	 * All states should be in alphabetical order
521 	 */
522 	while (curr_char_s && (text[0] < sst[curr_char_s].ch)) {
523 		prev_char_s = curr_char_s;
524 		curr_char_s = sst[curr_char_s].other_next_s;
525 	}
526 
527 	/*
528 	 * Check if a previously seen keyword has the same prefix as
529 	 * the current keyword.  If so, simply use the state for that
530 	 * keyword as my_state, otherwise, allocate a new state.
531 	 */
532 	if (curr_char_s && (text[0] == sst[curr_char_s].ch)) {
533 		my_state = curr_char_s;
534 		if ('\0' == text[1]) {
535 			fprintf(stderr,
536 				"Duplicate entries for keyword '%s' in"
537 				" keyword_gen.c ntp_keywords[].\n",
538 				current_keyword);
539 			exit(2);
540 		}
541 	} else {
542 		do
543 			my_state = sst_highwater++;
544 		while (my_state < COUNTOF(sst)
545 		       && sst[my_state].finishes_token);
546 		if (my_state >= COUNTOF(sst)) {
547 			fprintf(stderr,
548 				"fatal, keyword scanner state array "
549 				"sst[%d] is too small, modify\n"
550 				"keyword-gen.c to increase.\n",
551 				(int)COUNTOF(sst));
552 			exit(3);
553 		}
554 		/* Store the next character of the keyword */
555 		sst[my_state].ch = text[0];
556 		sst[my_state].other_next_s = curr_char_s;
557 		sst[my_state].followedby = FOLLBY_NON_ACCEPTING;
558 
559 		if (prev_char_s)
560 			sst[prev_char_s].other_next_s = my_state;
561 		else
562 			return_state = my_state;
563 	}
564 
565 	/* Check if the next character is '\0'.
566 	 * If yes, we are done with the recognition and this is an accepting
567 	 * state.
568 	 * If not, we need to continue scanning
569 	 */
570 	if ('\0' == text[1]) {
571 		sst[my_state].finishes_token = (u_short)token;
572 		sst[my_state].followedby = (char)followedby;
573 
574 		if (sst[token].finishes_token != (u_short)token) {
575 			fprintf(stderr,
576 				"fatal, sst[%d] not reserved for %s.\n",
577 				token, symbname(token));
578 			exit(6);
579 		}
580 		/* relocate so token id is sst[] index */
581 		if (my_state != token) {
582 			sst[token] = sst[my_state];
583 			ZERO(sst[my_state]);
584 			do
585 				sst_highwater--;
586 			while (sst[sst_highwater].finishes_token);
587 			my_state = token;
588 			if (prev_char_s)
589 				sst[prev_char_s].other_next_s = my_state;
590 			else
591 				return_state = my_state;
592 		}
593 	} else
594 		sst[my_state].match_next_s =
595 		    create_scan_states(
596 			&text[1],
597 			token,
598 			followedby,
599 			sst[my_state].match_next_s);
600 
601 	return return_state;
602 }
603 
604 
605 /* Define a function that takes a list of (keyword, token) values and
606  * creates a keywords scanner out of it.
607  */
608 
609 static u_short
610 create_keyword_scanner(void)
611 {
612 	u_short scanner;
613 	u_short i;
614 
615 	sst_highwater = 1;	/* index 0 invalid, unused */
616 	scanner = 0;
617 
618 	for (i = 0; i < COUNTOF(ntp_keywords); i++) {
619 		current_keyword = ntp_keywords[i].key;
620 		scanner =
621 		    create_scan_states(
622 			ntp_keywords[i].key,
623 			ntp_keywords[i].token,
624 			ntp_keywords[i].followedby,
625 			scanner);
626 	}
627 
628 	return scanner;
629 }
630 
631 
632 static void
633 generate_token_text(void)
634 {
635 	u_short lowest_id;
636 	u_short highest_id;
637 	u_short id_count;
638 	u_short id;
639 	u_short i;
640 
641 	/* sort ntp_keywords in token ID order */
642 	qsort(ntp_keywords, COUNTOF(ntp_keywords),
643 	      sizeof(ntp_keywords[0]), compare_key_tok_id);
644 
645 	lowest_id = ntp_keywords[0].token;
646 	highest_id = ntp_keywords[COUNTOF(ntp_keywords) - 1].token;
647 	id_count = highest_id - lowest_id + 1;
648 
649 	printf("#define LOWEST_KEYWORD_ID %d\n\n", lowest_id);
650 
651 	printf("const char * const keyword_text[%d] = {", id_count);
652 
653 	id = lowest_id;
654 	i = 0;
655 	while (i < COUNTOF(ntp_keywords)) {
656 		while (id < ntp_keywords[i].token) {
657 			printf(",\n\t/* %-5d %5d %20s */\tNULL",
658 			       id - lowest_id, id, symbname(id));
659 			id++;
660 		}
661 		if (i > 0)
662 			printf(",");
663 		printf("\n\t/* %-5d %5d %20s */\t\"%s\"",
664 		       id - lowest_id, id, symbname(id),
665 		       ntp_keywords[i].key);
666 		i++;
667 		id++;
668 	}
669 
670 	printf("\n};\n\n");
671 }
672 
673 
674 int
675 compare_key_tok_id(
676 	const void *a1,
677 	const void *a2
678 	)
679 {
680 	const struct key_tok *p1 = a1;
681 	const struct key_tok *p2 = a2;
682 
683 	if (p1->token == p2->token)
684 		return 0;
685 
686 	if (p1->token < p2->token)
687 		return -1;
688 	else
689 		return 1;
690 }
691 
692 
693 int
694 compare_key_tok_text(
695 	const void *a1,
696 	const void *a2
697 	)
698 {
699 	const struct key_tok *p1 = a1;
700 	const struct key_tok *p2 = a2;
701 
702 	return strcmp(p1->key, p2->key);
703 }
704 
705 
706 /*
707  * populate_symb() - populate symb[] lookup array with symbolic token
708  *		     names such that symb[T_Age] == "T_Age", etc.
709  */
710 void
711 populate_symb(
712 	char *header_file
713 	)
714 {
715 	FILE *	yh;
716 	char	line[2 * MAX_TOK_LEN];
717 	char	name[2 * MAX_TOK_LEN];
718 	int	token;
719 
720 	yh = fopen(header_file, "r");
721 	if (NULL == yh) {
722 		perror("unable to open yacc/bison header file");
723 		exit(4);
724 	}
725 
726 	while (NULL != fgets(line, sizeof(line), yh))
727 		if (2 == sscanf(line, "#define %s %d", name, &token)
728 		    && 'T' == name[0] && '_' == name[1] && token >= 0
729 		    && token < COUNTOF(symb)) {
730 
731 			symb[token] = estrdup(name);
732 			if (strlen(name) > MAX_TOK_LEN) {
733 				fprintf(stderr,
734 					"MAX_TOK_LEN %d too small for '%s'\n"
735 					"Edit keyword-gen.c to raise.\n",
736 					MAX_TOK_LEN, name);
737 				exit(10);
738 			}
739 		}
740 	fclose(yh);
741 }
742 
743 
744 const char *
745 symbname(
746 	u_short token
747 	)
748 {
749 	char *name;
750 
751 	if (token < COUNTOF(symb) && symb[token] != NULL) {
752 		name = symb[token];
753 	} else {
754 		LIB_GETBUF(name);
755 		snprintf(name, LIB_BUFLENGTH, "%d", token);
756 	}
757 
758 	return name;
759 }
760