xref: /netbsd-src/external/bsd/ntp/dist/ntpd/keyword-gen.c (revision eabc0478de71e4e011a5b4e0392741e01d491794)
1 /*	$NetBSD: keyword-gen.c,v 1.14 2024/08/18 20:47:17 christos Exp $	*/
2 
3 /*
4  * keyword-gen.c -- generate keyword scanner finite state machine and
5  *		    keyword_text array.
6  *
7  * This program is run to generate ntp_keyword.h
8  * After making a change here, two output files should be committed at
9  * the same time as keyword-gen.c:
10  *	ntp_keyword.h
11  *	keyword-gen-utd
12  *
13  * keyword-gen-utd is a sentinel used by Makefile.am to avoid compiling
14  * keyword_gen.c and generating ntp_keyword.h if the input keyword-gen.c
15  * has not changed.  This is not solely an optimization, it also breaks
16  * a dependency chain that otherwise would cause programs to be compiled
17  * when running "make dist" or "make distdir".  We want these to package
18  * the existing source without building anything but a tarball.  See
19  * [Bug 1470].
20  */
21 #include <config.h>
22 #include <stdio.h>
23 #include <stdlib.h>
24 #include <time.h>
25 
26 #include <ntp_stdlib.h>
27 #include <ntp_config.h>
28 #include "ntp_scanner.h"
29 #include "ntp_parser.h"
30 
31 
32 /* Define a structure to hold a (keyword, token) pair */
33 struct key_tok {
34 	char *	key;		/* Keyword */
35 	u_short	token;		/* Associated Token */
36 	follby	followedby;	/* nonzero indicates the next token(s)
37 				   forced to be string(s) */
38 };
39 
40 struct key_tok ntp_keywords[] = {
41 { "...",		T_Ellipsis,		FOLLBY_TOKEN },
42 { "allpeers",		T_Allpeers,		FOLLBY_TOKEN },
43 { "automax",		T_Automax,		FOLLBY_TOKEN },
44 { "broadcast",		T_Broadcast,		FOLLBY_STRING },
45 { "broadcastclient",	T_Broadcastclient,	FOLLBY_TOKEN },
46 { "broadcastdelay",	T_Broadcastdelay,	FOLLBY_TOKEN },
47 { "checkhash",		T_Checkhash,		FOLLBY_TOKEN },
48 { "ctl",		T_Ctl,			FOLLBY_TOKEN },
49 { "delrestrict",	T_Delrestrict,		FOLLBY_TOKEN },
50 { "device",		T_Device,		FOLLBY_STRING },
51 { "disable",		T_Disable,		FOLLBY_TOKEN },
52 { "driftfile",		T_Driftfile,		FOLLBY_STRING },
53 { "dscp",		T_Dscp,			FOLLBY_TOKEN },
54 { "enable",		T_Enable,		FOLLBY_TOKEN },
55 { "end",		T_End,			FOLLBY_TOKEN },
56 { "filegen",		T_Filegen,		FOLLBY_TOKEN },
57 { "fudge",		T_Fudge,		FOLLBY_STRING },
58 { "ignorehash",		T_Ignorehash,		FOLLBY_TOKEN },
59 { "io",			T_Io,			FOLLBY_TOKEN },
60 { "includefile",	T_Includefile,		FOLLBY_STRING },
61 { "leapfile",		T_Leapfile,		FOLLBY_STRING },
62 { "leapsmearinterval",	T_Leapsmearinterval,	FOLLBY_TOKEN },
63 { "logconfig",		T_Logconfig,		FOLLBY_STRINGS_TO_EOC },
64 { "logfile",		T_Logfile,		FOLLBY_STRING },
65 { "manycastclient",	T_Manycastclient,	FOLLBY_STRING },
66 { "manycastserver",	T_Manycastserver,	FOLLBY_STRINGS_TO_EOC },
67 { "mem",		T_Mem,			FOLLBY_TOKEN },
68 { "multicastclient",	T_Multicastclient,	FOLLBY_STRINGS_TO_EOC },
69 { "peer",		T_Peer,			FOLLBY_STRING },
70 { "phone",		T_Phone,		FOLLBY_STRINGS_TO_EOC },
71 { "pidfile",		T_Pidfile,		FOLLBY_STRING },
72 { "pollskewlist",	T_PollSkewList,		FOLLBY_TOKEN },
73 { "pool",		T_Pool,			FOLLBY_STRING },
74 { "discard",		T_Discard,		FOLLBY_TOKEN },
75 { "reset",		T_Reset,		FOLLBY_TOKEN },
76 { "restrict",		T_Restrict,		FOLLBY_TOKEN },
77 { "rlimit",		T_Rlimit,		FOLLBY_TOKEN },
78 { "server",		T_Server,		FOLLBY_STRING },
79 { "serverresponse",	T_Serverresponse,	FOLLBY_TOKEN },
80 { "fuzz",		T_Fuzz,			FOLLBY_TOKEN },
81 { "poll",		T_Poll,			FOLLBY_TOKEN },
82 { "setvar",		T_Setvar,		FOLLBY_STRING },
83 { "statistics",		T_Statistics,		FOLLBY_TOKEN },
84 { "statsdir",		T_Statsdir,		FOLLBY_STRING },
85 { "sys",		T_Sys,			FOLLBY_TOKEN },
86 { "tick",		T_Tick,			FOLLBY_TOKEN },
87 { "timer",		T_Timer,		FOLLBY_TOKEN },
88 { "tinker",		T_Tinker,		FOLLBY_TOKEN },
89 { "tos",		T_Tos,			FOLLBY_TOKEN },
90 { "trap",		T_Trap,			FOLLBY_STRING },
91 { "unconfig",		T_Unconfig,		FOLLBY_STRING },
92 { "unpeer",		T_Unpeer,		FOLLBY_STRING },
93 { "xmtnonce",		T_Xmtnonce,		FOLLBY_TOKEN },
94 /* authentication_command */
95 { "controlkey",		T_ControlKey,		FOLLBY_TOKEN },
96 { "crypto",		T_Crypto,		FOLLBY_TOKEN },
97 { "keys",		T_Keys,			FOLLBY_STRING },
98 { "keysdir",		T_Keysdir,		FOLLBY_STRING },
99 { "ntpsigndsocket",	T_NtpSignDsocket,	FOLLBY_STRING },
100 { "requestkey",		T_Requestkey,		FOLLBY_TOKEN },
101 { "revoke",		T_Revoke,		FOLLBY_TOKEN },
102 { "trustedkey",		T_Trustedkey,		FOLLBY_TOKEN },
103 /* IPv4/IPv6 protocol override flag */
104 { "-4",			T_Ipv4_flag,		FOLLBY_TOKEN },
105 { "-6",			T_Ipv6_flag,		FOLLBY_TOKEN },
106 /* option */
107 { "autokey",		T_Autokey,		FOLLBY_TOKEN },
108 { "burst",		T_Burst,		FOLLBY_TOKEN },
109 { "iburst",		T_Iburst,		FOLLBY_TOKEN },
110 { "key",		T_Key,			FOLLBY_TOKEN },
111 { "maxpoll",		T_Maxpoll,		FOLLBY_TOKEN },
112 { "mdnstries",		T_Mdnstries,		FOLLBY_TOKEN },
113 { "minpoll",		T_Minpoll,		FOLLBY_TOKEN },
114 { "mode",		T_Mode,			FOLLBY_TOKEN },
115 { "noselect",		T_Noselect,		FOLLBY_TOKEN },
116 { "preempt",		T_Preempt,		FOLLBY_TOKEN },
117 { "true",		T_True,			FOLLBY_TOKEN },
118 { "prefer",		T_Prefer,		FOLLBY_TOKEN },
119 { "ttl",		T_Ttl,			FOLLBY_TOKEN },
120 { "version",		T_Version,		FOLLBY_TOKEN },
121 { "xleave",		T_Xleave,		FOLLBY_TOKEN },
122 /* crypto_command */
123 { "host",		T_Host,			FOLLBY_STRING },
124 { "ident",		T_Ident,		FOLLBY_STRING },
125 { "pw",			T_Pw,			FOLLBY_STRING },
126 { "randfile",		T_Randfile,		FOLLBY_STRING },
127 { "digest",		T_Digest,		FOLLBY_STRING },
128 /*** MONITORING COMMANDS ***/
129 /* stat */
130 { "clockstats",		T_Clockstats,		FOLLBY_TOKEN },
131 { "cryptostats",	T_Cryptostats,		FOLLBY_TOKEN },
132 { "loopstats",		T_Loopstats,		FOLLBY_TOKEN },
133 { "peerstats",		T_Peerstats,		FOLLBY_TOKEN },
134 { "rawstats",		T_Rawstats,		FOLLBY_TOKEN },
135 { "sysstats", 		T_Sysstats,		FOLLBY_TOKEN },
136 { "protostats",		T_Protostats,		FOLLBY_TOKEN },
137 { "timingstats",	T_Timingstats,		FOLLBY_TOKEN },
138 /* filegen_option */
139 { "file",		T_File,			FOLLBY_STRING },
140 { "link",		T_Link,			FOLLBY_TOKEN },
141 { "nolink",		T_Nolink,		FOLLBY_TOKEN },
142 { "type",		T_Type,			FOLLBY_TOKEN },
143 /* filegen_type */
144 { "age",		T_Age,			FOLLBY_TOKEN },
145 { "day",		T_Day,			FOLLBY_TOKEN },
146 { "month",		T_Month,		FOLLBY_TOKEN },
147 { "none",		T_None,			FOLLBY_TOKEN },
148 { "pid",		T_Pid,			FOLLBY_TOKEN },
149 { "week",		T_Week,			FOLLBY_TOKEN },
150 { "year",		T_Year,			FOLLBY_TOKEN },
151 /*** ORPHAN MODE COMMANDS ***/
152 /* tos_option */
153 { "minclock",		T_Minclock,		FOLLBY_TOKEN },
154 { "maxclock",		T_Maxclock,		FOLLBY_TOKEN },
155 { "minsane",		T_Minsane,		FOLLBY_TOKEN },
156 { "floor",		T_Floor,		FOLLBY_TOKEN },
157 { "ceiling",		T_Ceiling,		FOLLBY_TOKEN },
158 { "cohort",		T_Cohort,		FOLLBY_TOKEN },
159 { "mindist",		T_Mindist,		FOLLBY_TOKEN },
160 { "maxdist",		T_Maxdist,		FOLLBY_TOKEN },
161 { "bcpollbstep",	T_Bcpollbstep,		FOLLBY_TOKEN },
162 { "beacon",		T_Beacon,		FOLLBY_TOKEN },
163 { "orphan",		T_Orphan,		FOLLBY_TOKEN },
164 { "orphanwait",		T_Orphanwait,		FOLLBY_TOKEN },
165 { "nonvolatile",	T_Nonvolatile,		FOLLBY_TOKEN },
166 { "basedate",		T_Basedate,		FOLLBY_STRING },
167 /* access_control_flag */
168 { "default",		T_Default,		FOLLBY_TOKEN },
169 { "source",		T_Source,		FOLLBY_TOKEN },
170 { "epeer",		T_Epeer,		FOLLBY_TOKEN },
171 { "noepeer",		T_Noepeer,		FOLLBY_TOKEN },
172 { "flake",		T_Flake,		FOLLBY_TOKEN },
173 { "ignore",		T_Ignore,		FOLLBY_TOKEN },
174 { "ippeerlimit",	T_Ippeerlimit,		FOLLBY_TOKEN },
175 { "limited",		T_Limited,		FOLLBY_TOKEN },
176 { "mssntp",		T_Mssntp,		FOLLBY_TOKEN },
177 { "kod",		T_Kod,			FOLLBY_TOKEN },
178 { "lowpriotrap",	T_Lowpriotrap,		FOLLBY_TOKEN },
179 { "mask",		T_Mask,			FOLLBY_TOKEN },
180 { "nomodify",		T_Nomodify,		FOLLBY_TOKEN },
181 { "nomrulist",		T_Nomrulist,		FOLLBY_TOKEN },
182 { "nopeer",		T_Nopeer,		FOLLBY_TOKEN },
183 { "noquery",		T_Noquery,		FOLLBY_TOKEN },
184 { "noserve",		T_Noserve,		FOLLBY_TOKEN },
185 { "notrap",		T_Notrap,		FOLLBY_TOKEN },
186 { "notrust",		T_Notrust,		FOLLBY_TOKEN },
187 { "ntpport",		T_Ntpport,		FOLLBY_TOKEN },
188 /* discard_option */
189 { "average",		T_Average,		FOLLBY_TOKEN },
190 { "minimum",		T_Minimum,		FOLLBY_TOKEN },
191 { "monitor",		T_Monitor,		FOLLBY_TOKEN },
192 /* mru_option */
193 { "incalloc",		T_Incalloc,		FOLLBY_TOKEN },
194 { "incmem",		T_Incmem,		FOLLBY_TOKEN },
195 { "initalloc",		T_Initalloc,		FOLLBY_TOKEN },
196 { "initmem",		T_Initmem,		FOLLBY_TOKEN },
197 { "mindepth",		T_Mindepth,		FOLLBY_TOKEN },
198 { "maxage",		T_Maxage,		FOLLBY_TOKEN },
199 { "maxdepth",		T_Maxdepth,		FOLLBY_TOKEN },
200 { "maxmem",		T_Maxmem,		FOLLBY_TOKEN },
201 { "mru",		T_Mru,			FOLLBY_TOKEN },
202 /* fudge_factor */
203 { "abbrev",		T_Abbrev,		FOLLBY_STRING },
204 { "flag1",		T_Flag1,		FOLLBY_TOKEN },
205 { "flag2",		T_Flag2,		FOLLBY_TOKEN },
206 { "flag3",		T_Flag3,		FOLLBY_TOKEN },
207 { "flag4",		T_Flag4,		FOLLBY_TOKEN },
208 { "refid",		T_Refid,		FOLLBY_STRING },
209 { "stratum",		T_Stratum,		FOLLBY_TOKEN },
210 { "time1",		T_Time1,		FOLLBY_TOKEN },
211 { "time2",		T_Time2,		FOLLBY_TOKEN },
212 { "minjitter",		T_Minjitter,		FOLLBY_TOKEN },
213 /* device spec */
214 { "ppsdata",		T_PpsData,		FOLLBY_STRING },
215 { "timedata",		T_TimeData,		FOLLBY_STRING },
216 /* system_option */
217 { "auth",		T_Auth,			FOLLBY_TOKEN },
218 { "bclient",		T_Bclient,		FOLLBY_TOKEN },
219 { "calibrate",		T_Calibrate,		FOLLBY_TOKEN },
220 { "kernel",		T_Kernel,		FOLLBY_TOKEN },
221 { "mode7",		T_Mode7,		FOLLBY_TOKEN },
222 { "ntp",		T_Ntp,			FOLLBY_TOKEN },
223 { "peer_clear_digest_early",	T_PCEdigest,	FOLLBY_TOKEN },
224 { "stats",		T_Stats,		FOLLBY_TOKEN },
225 { "unpeer_crypto_early",	T_UEcrypto,	FOLLBY_TOKEN },
226 { "unpeer_crypto_nak_early",	T_UEcryptonak,	FOLLBY_TOKEN },
227 { "unpeer_digest_early",	T_UEdigest,	FOLLBY_TOKEN },
228 /* rlimit_option */
229 { "memlock",		T_Memlock,		FOLLBY_TOKEN },
230 { "stacksize",		T_Stacksize,		FOLLBY_TOKEN },
231 { "filenum",		T_Filenum,		FOLLBY_TOKEN },
232 /* tinker_option */
233 { "step",		T_Step,			FOLLBY_TOKEN },
234 { "stepback",		T_Stepback,		FOLLBY_TOKEN },
235 { "stepfwd",		T_Stepfwd,		FOLLBY_TOKEN },
236 { "panic",		T_Panic,		FOLLBY_TOKEN },
237 { "dispersion",		T_Dispersion,		FOLLBY_TOKEN },
238 { "stepout",		T_Stepout,		FOLLBY_TOKEN },
239 { "allan",		T_Allan,		FOLLBY_TOKEN },
240 { "huffpuff",		T_Huffpuff,		FOLLBY_TOKEN },
241 { "freq",		T_Freq,			FOLLBY_TOKEN },
242 /* miscellaneous_command */
243 { "port",		T_Port,			FOLLBY_TOKEN },
244 { "interface",		T_Interface,		FOLLBY_TOKEN },
245 { "saveconfigdir",	T_Saveconfigdir,	FOLLBY_STRING },
246 /* interface_command (ignore and interface already defined) */
247 { "nic",		T_Nic,			FOLLBY_TOKEN },
248 { "all",		T_All,			FOLLBY_TOKEN },
249 { "ipv4",		T_Ipv4,			FOLLBY_TOKEN },
250 { "ipv6",		T_Ipv6,			FOLLBY_TOKEN },
251 { "wildcard",		T_Wildcard,		FOLLBY_TOKEN },
252 { "listen",		T_Listen,		FOLLBY_TOKEN },
253 { "drop",		T_Drop,			FOLLBY_TOKEN },
254 /* simulator commands */
255 { "simulate",		T_Simulate,		FOLLBY_TOKEN },
256 { "simulation_duration",T_Sim_Duration,		FOLLBY_TOKEN },
257 { "beep_delay",		T_Beep_Delay,		FOLLBY_TOKEN },
258 { "duration",		T_Duration,		FOLLBY_TOKEN },
259 { "server_offset",	T_Server_Offset,	FOLLBY_TOKEN },
260 { "freq_offset",	T_Freq_Offset,		FOLLBY_TOKEN },
261 { "wander",		T_Wander,		FOLLBY_TOKEN },
262 { "jitter",		T_Jitter,		FOLLBY_TOKEN },
263 { "prop_delay",		T_Prop_Delay,		FOLLBY_TOKEN },
264 { "proc_delay",		T_Proc_Delay,		FOLLBY_TOKEN },
265 };
266 
267 typedef struct big_scan_state_tag {
268 	char	ch;		/* Character this state matches on */
269 	char	followedby;	/* Forces next token(s) to T_String */
270 	u_short	finishes_token;	/* nonzero ID if last keyword char */
271 	u_short	match_next_s;	/* next state to check matching ch */
272 	u_short	other_next_s;	/* next state to check if not ch */
273 } big_scan_state;
274 
275 /*
276  * Note: to increase MAXSTATES beyond 2048, be aware it is currently
277  * crammed into 11 bits in scan_state form.  Raising to 4096 would be
278  * relatively easy by storing the followedby value in a separate
279  * array with one entry per token, and shrinking the char value to
280  * 7 bits to free a bit for accepting/non-accepting.  More than 4096
281  * states will require expanding scan_state beyond 32 bits each.
282  */
283 #define MAXSTATES	2048
284 #define MAX_TOK_LEN	63
285 
286 const char *	current_keyword;/* for error reporting */
287 big_scan_state	sst[MAXSTATES];	/* scanner FSM state entries */
288 u_short		sst_highwater;	/* next entry index to consider */
289 char *		symb[1024];	/* map token ID to symbolic name */
290 
291 /* for libntp */
292 const char *	progname = "keyword-gen";
293 
294 int		main			(int, char **);
295 static void	generate_preamble	(void);
296 static void	generate_fsm		(void);
297 static void	generate_token_text	(void);
298 static u_short	create_keyword_scanner	(void);
299 static u_short	create_scan_states	(char *, u_short, follby, u_short);
300 int		compare_key_tok_id	(const void *, const void *);
301 int		compare_key_tok_text	(const void *, const void *);
302 void		populate_symb		(char *);
303 const char *	symbname		(u_short);
304 
305 
306 int main(int argc, char **argv)
307 {
308 	if (argc < 2) {
309 		fprintf(stderr, "Usage:\n%s t_header.h\n", argv[0]);
310 		exit(1);
311 	}
312 	debug = 1;
313 
314 	populate_symb(argv[1]);
315 
316 	generate_preamble();
317 	generate_token_text();
318 	generate_fsm();
319 
320 	return 0;
321 }
322 
323 
324 static void
325 generate_preamble(void)
326 {
327 	time_t now;
328 	char timestamp[128];
329 	char preamble[] =
330 "/*\n"
331 " * ntp_keyword.h\n"
332 " * \n"
333 " * NOTE: edit this file with caution, it is generated by keyword-gen.c\n"
334 " *\t Generated %s UTC	  diff_ignore_line\n"
335 " *\n"
336 " */\n"
337 "#include \"ntp_scanner.h\"\n"
338 "#include \"ntp_parser.h\"\n"
339 "\n";
340 
341 	time(&now);
342 	if (!strftime(timestamp, sizeof(timestamp),
343 		      "%Y-%m-%d %H:%M:%S", gmtime(&now)))
344 		timestamp[0] = '\0';
345 
346 	printf(preamble, timestamp);
347 }
348 
349 
350 static void
351 generate_fsm(void)
352 {
353 	char rprefix[MAX_TOK_LEN + 1];
354 	char prefix[MAX_TOK_LEN + 1];
355 	char token_id_comment[16 + MAX_TOK_LEN + 1];
356 	size_t prefix_len;
357 	char *p;
358 	char *r;
359 	u_short initial_state;
360 	u_short this_state;
361 	u_short state;
362 	u_short i;
363 	u_short token;
364 
365 	/*
366 	 * Sort ntp_keywords in alphabetical keyword order.  This is
367 	 * not necessary, but minimizes nonfunctional changes in the
368 	 * generated finite state machine when keywords are modified.
369 	 */
370 	qsort(ntp_keywords, COUNTOF(ntp_keywords),
371 	      sizeof(ntp_keywords[0]), compare_key_tok_text);
372 
373 	/*
374 	 * To save space, reserve the state array entry matching each
375 	 * token number for its terminal state, so the token identifier
376 	 * does not need to be stored in each state, but can be
377 	 * recovered trivially.  To mark the entry reserved,
378 	 * finishes_token is nonzero.
379 	 */
380 
381 	for (i = 0; i < COUNTOF(ntp_keywords); i++) {
382 		token = ntp_keywords[i].token;
383 		if (1 > token || token >= COUNTOF(sst)) {
384 			fprintf(stderr,
385 				"keyword-gen sst[%u] too small "
386 				"for keyword '%s' id %d\n",
387 				(int)COUNTOF(sst),
388 				ntp_keywords[i].key,
389 				token);
390 			exit(4);
391 		}
392 		sst[token].finishes_token = token;
393 	}
394 
395 	initial_state = create_keyword_scanner();
396 
397 	fprintf(stderr,
398 		"%d keywords consumed %d states of %d max.\n",
399 		(int)COUNTOF(ntp_keywords),
400 		sst_highwater - 1,
401 		(int)COUNTOF(sst) - 1);
402 
403 	printf("#define SCANNER_INIT_S %d\n\n", initial_state);
404 
405 	printf("const scan_state sst[%d] = {\n"
406 	       "/*SS_T( ch,\tf-by, match, other ),\t\t\t\t */\n"
407 	       "  0,\t\t\t\t      /* %5d %-17s */\n",
408 	       sst_highwater,
409 	       0, "");
410 
411 	for (i = 1; i < sst_highwater; i++) {
412 
413 		/* verify fields will fit */
414 		if (sst[i].followedby & ~0x3) {
415 			fprintf(stderr,
416 				"keyword-gen internal error "
417 				"sst[%d].followedby %d too big\n",
418 				i, sst[i].followedby);
419 			exit(7);
420 		}
421 
422 		if (sst_highwater <= sst[i].match_next_s
423 		    || sst[i].match_next_s & ~0x7ff) {
424 			fprintf(stderr,
425 				"keyword-gen internal error "
426 				"sst[%d].match_next_s %d too big\n",
427 				i, sst[i].match_next_s);
428 			exit(8);
429 		}
430 
431 		if (sst_highwater <= sst[i].other_next_s
432 		    || sst[i].other_next_s & ~0x7ff) {
433 			fprintf(stderr,
434 				"keyword-gen internal error "
435 				"sst[%d].other_next_s %d too big\n",
436 				i, sst[i].other_next_s);
437 			exit(9);
438 		}
439 
440 		if (sst[i].finishes_token) {
441 			snprintf(token_id_comment,
442 				 sizeof(token_id_comment), "%5d %-17s",
443 				 i, symbname(sst[i].finishes_token));
444 			if (i != sst[i].finishes_token) {
445 				fprintf(stderr,
446 					"keyword-gen internal error "
447 					"entry %d finishes token %d\n",
448 					i, sst[i].finishes_token);
449 				exit(5);
450 			}
451 		} else {
452 		/*
453 		 * Determine the keyword prefix that leads to this
454 		 * state.  This is expensive but keyword-gen is run
455 		 * only when it changes.  Distributing keyword-gen-utd
456 		 * achieves that, which is why it must be committed
457 		 * at the same time as keyword-gen.c and ntp_keyword.h.
458 		 *
459 		 * Scan the state array iteratively looking for a state
460 		 * which leads to the current one, collecting matching
461 		 * characters along the way.  There is only one such
462 		 * path back to the starting state given the way our
463 		 * scanner state machine is built and the practice of
464 		 * using the spelling of the keyword as its T_* token
465 		 * identifier, which results in never having two
466 		 * spellings result in the same T_* value.
467 		 */
468 			prefix_len = 0;
469 			this_state = i;
470 			do {
471 				for (state = 1; state < sst_highwater; state++)
472 					if (sst[state].other_next_s == this_state) {
473 						this_state = state;
474 						break;
475 					} else if (sst[state].match_next_s == this_state) {
476 						this_state = state;
477 						rprefix[prefix_len] = sst[state].ch;
478 						prefix_len++;
479 						break;
480 					}
481 			} while (this_state != initial_state);
482 
483 			if (prefix_len) {
484 				/* reverse rprefix into prefix */
485 				p = prefix + prefix_len;
486 				r = rprefix;
487 				while (r < rprefix + prefix_len)
488 					*--p = *r++;
489 			}
490 			prefix[prefix_len] = '\0';
491 
492 			snprintf(token_id_comment,
493 				 sizeof(token_id_comment), "%5d %-17s",
494 				 i, (initial_state == i)
495 					? "[initial state]"
496 					: prefix);
497 		}
498 
499 		printf("  S_ST( '%c',\t%d,    %5u, %5u )%s /* %s */\n",
500 		       sst[i].ch,
501 		       sst[i].followedby,
502 		       sst[i].match_next_s,
503 		       sst[i].other_next_s,
504 		       (i + 1 < sst_highwater)
505 			   ? ","
506 			   : " ",
507 		       token_id_comment);
508 	}
509 
510 	printf("};\n\n");
511 }
512 
513 
514 /* Define a function to create the states of the scanner. This function
515  * is used by the create_keyword_scanner function below.
516  *
517  * This function takes a suffix of a keyword, the token to be returned on
518  * recognizing the complete keyword, and any pre-existing state that exists
519  * for some other keyword that has the same prefix as the current one.
520  */
521 static u_short
522 create_scan_states(
523 	char *	text,
524 	u_short	token,
525 	follby	followedby,
526 	u_short	prev_state
527 	)
528 {
529 	u_short my_state;
530 	u_short return_state;
531 	u_short prev_char_s;
532 	u_short curr_char_s;
533 
534 	return_state = prev_state;
535 	curr_char_s = prev_state;
536 	prev_char_s = 0;
537 
538 	/* Find the correct position to insert the state.
539 	 * All states should be in alphabetical order
540 	 */
541 	while (curr_char_s && (text[0] < sst[curr_char_s].ch)) {
542 		prev_char_s = curr_char_s;
543 		curr_char_s = sst[curr_char_s].other_next_s;
544 	}
545 
546 	/*
547 	 * Check if a previously seen keyword has the same prefix as
548 	 * the current keyword.  If so, simply use the state for that
549 	 * keyword as my_state, otherwise, allocate a new state.
550 	 */
551 	if (curr_char_s && (text[0] == sst[curr_char_s].ch)) {
552 		my_state = curr_char_s;
553 		if ('\0' == text[1]) {
554 			fprintf(stderr,
555 				"Duplicate entries for keyword '%s' in"
556 				" keyword_gen.c ntp_keywords[].\n",
557 				current_keyword);
558 			exit(2);
559 		}
560 	} else {
561 		do
562 			my_state = sst_highwater++;
563 		while (my_state < COUNTOF(sst)
564 		       && sst[my_state].finishes_token);
565 		if (my_state >= COUNTOF(sst)) {
566 			fprintf(stderr,
567 				"fatal, keyword scanner state array "
568 				"sst[%d] is too small, modify\n"
569 				"keyword-gen.c to increase.\n",
570 				(int)COUNTOF(sst));
571 			exit(3);
572 		}
573 		/* Store the next character of the keyword */
574 		sst[my_state].ch = text[0];
575 		sst[my_state].other_next_s = curr_char_s;
576 		sst[my_state].followedby = FOLLBY_NON_ACCEPTING;
577 
578 		if (prev_char_s)
579 			sst[prev_char_s].other_next_s = my_state;
580 		else
581 			return_state = my_state;
582 	}
583 
584 	/* Check if the next character is '\0'.
585 	 * If yes, we are done with the recognition and this is an accepting
586 	 * state.
587 	 * If not, we need to continue scanning
588 	 */
589 	if ('\0' == text[1]) {
590 		sst[my_state].finishes_token = (u_short)token;
591 		sst[my_state].followedby = (char)followedby;
592 
593 		if (sst[token].finishes_token != (u_short)token) {
594 			fprintf(stderr,
595 				"fatal, sst[%d] not reserved for %s.\n",
596 				token, symbname(token));
597 			exit(6);
598 		}
599 		/* relocate so token id is sst[] index */
600 		if (my_state != token) {
601 			sst[token] = sst[my_state];
602 			ZERO(sst[my_state]);
603 			do
604 				sst_highwater--;
605 			while (sst[sst_highwater].finishes_token);
606 			my_state = token;
607 			if (prev_char_s)
608 				sst[prev_char_s].other_next_s = my_state;
609 			else
610 				return_state = my_state;
611 		}
612 	} else
613 		sst[my_state].match_next_s =
614 		    create_scan_states(
615 			&text[1],
616 			token,
617 			followedby,
618 			sst[my_state].match_next_s);
619 
620 	return return_state;
621 }
622 
623 
624 /* Define a function that takes a list of (keyword, token) values and
625  * creates a keywords scanner out of it.
626  */
627 
628 static u_short
629 create_keyword_scanner(void)
630 {
631 	u_short scanner;
632 	u_short i;
633 
634 	sst_highwater = 1;	/* index 0 invalid, unused */
635 	scanner = 0;
636 
637 	for (i = 0; i < COUNTOF(ntp_keywords); i++) {
638 		current_keyword = ntp_keywords[i].key;
639 		scanner =
640 		    create_scan_states(
641 			ntp_keywords[i].key,
642 			ntp_keywords[i].token,
643 			ntp_keywords[i].followedby,
644 			scanner);
645 	}
646 
647 	return scanner;
648 }
649 
650 
651 static void
652 generate_token_text(void)
653 {
654 	u_short lowest_id;
655 	u_short highest_id;
656 	u_short id_count;
657 	u_short id;
658 	u_short i;
659 
660 	/* sort ntp_keywords in token ID order */
661 	qsort(ntp_keywords, COUNTOF(ntp_keywords),
662 	      sizeof(ntp_keywords[0]), compare_key_tok_id);
663 
664 	lowest_id = ntp_keywords[0].token;
665 	highest_id = ntp_keywords[COUNTOF(ntp_keywords) - 1].token;
666 	id_count = highest_id - lowest_id + 1;
667 
668 	printf("#define LOWEST_KEYWORD_ID %d\n\n", lowest_id);
669 
670 	printf("const char * const keyword_text[%d] = {", id_count);
671 
672 	id = lowest_id;
673 	i = 0;
674 	while (i < COUNTOF(ntp_keywords)) {
675 		while (id < ntp_keywords[i].token) {
676 			printf(",\n\t/* %-5d %5d %20s */\tNULL",
677 			       id - lowest_id, id, symbname(id));
678 			id++;
679 		}
680 		if (i > 0)
681 			printf(",");
682 		printf("\n\t/* %-5d %5d %20s */\t\"%s\"",
683 		       id - lowest_id, id, symbname(id),
684 		       ntp_keywords[i].key);
685 		i++;
686 		id++;
687 	}
688 
689 	printf("\n};\n\n");
690 }
691 
692 
693 int
694 compare_key_tok_id(
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 	if (p1->token == p2->token)
703 		return 0;
704 
705 	if (p1->token < p2->token)
706 		return -1;
707 	else
708 		return 1;
709 }
710 
711 
712 int
713 compare_key_tok_text(
714 	const void *a1,
715 	const void *a2
716 	)
717 {
718 	const struct key_tok *p1 = a1;
719 	const struct key_tok *p2 = a2;
720 
721 	return strcmp(p1->key, p2->key);
722 }
723 
724 
725 /*
726  * populate_symb() - populate symb[] lookup array with symbolic token
727  *		     names such that symb[T_Age] == "T_Age", etc.
728  */
729 void
730 populate_symb(
731 	char *header_file
732 	)
733 {
734 	FILE *	yh;
735 	char	line[2 * MAX_TOK_LEN];
736 	char	name[2 * MAX_TOK_LEN];
737 	int	token;
738 
739 	yh = fopen(header_file, "r");
740 	if (NULL == yh) {
741 		perror("unable to open yacc/bison header file");
742 		exit(4);
743 	}
744 
745 	while (NULL != fgets(line, sizeof(line), yh))
746 		if (2 == sscanf(line, "#define %s %d", name, &token)
747 		    && 'T' == name[0] && '_' == name[1] && token >= 0
748 		    && token < COUNTOF(symb)) {
749 
750 			symb[token] = estrdup(name);
751 			if (strlen(name) > MAX_TOK_LEN) {
752 				fprintf(stderr,
753 					"MAX_TOK_LEN %d too small for '%s'\n"
754 					"Edit keyword-gen.c to raise.\n",
755 					MAX_TOK_LEN, name);
756 				exit(10);
757 			}
758 		}
759 	fclose(yh);
760 }
761 
762 
763 const char *
764 symbname(
765 	u_short token
766 	)
767 {
768 	char *name;
769 
770 	if (token < COUNTOF(symb) && symb[token] != NULL) {
771 		name = symb[token];
772 	} else {
773 		LIB_GETBUF(name);
774 		snprintf(name, LIB_BUFLENGTH, "%d", token);
775 	}
776 
777 	return name;
778 }
779