xref: /freebsd/sys/kern/subr_scanf.c (revision 38f0b757fd84d17d0fc24739a7cda160c4516d81)
1 /*-
2  * Copyright (c) 1990, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * This code is derived from software contributed to Berkeley by
6  * Chris Torek.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 4. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  * From: Id: vfscanf.c,v 1.13 1998/09/25 12:20:27 obrien Exp
33  * From: static char sccsid[] = "@(#)strtol.c	8.1 (Berkeley) 6/4/93";
34  * From: static char sccsid[] = "@(#)strtoul.c	8.1 (Berkeley) 6/4/93";
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/ctype.h>
43 #include <sys/limits.h>
44 
45 /*
46  * Note that stdarg.h and the ANSI style va_start macro is used for both
47  * ANSI and traditional C compilers.
48  */
49 #include <machine/stdarg.h>
50 
51 #define	BUF		32 	/* Maximum length of numeric string. */
52 
53 /*
54  * Flags used during conversion.
55  */
56 #define	LONG		0x01	/* l: long or double */
57 #define	SHORT		0x04	/* h: short */
58 #define	SUPPRESS	0x08	/* suppress assignment */
59 #define	POINTER		0x10	/* weird %p pointer (`fake hex') */
60 #define	NOSKIP		0x20	/* do not skip blanks */
61 #define	QUAD		0x400
62 #define	SHORTSHORT	0x4000	/** hh: char */
63 
64 /*
65  * The following are used in numeric conversions only:
66  * SIGNOK, NDIGITS, DPTOK, and EXPOK are for floating point;
67  * SIGNOK, NDIGITS, PFXOK, and NZDIGITS are for integral.
68  */
69 #define	SIGNOK		0x40	/* +/- is (still) legal */
70 #define	NDIGITS		0x80	/* no digits detected */
71 
72 #define	DPTOK		0x100	/* (float) decimal point is still legal */
73 #define	EXPOK		0x200	/* (float) exponent (e+3, etc) still legal */
74 
75 #define	PFXOK		0x100	/* 0x prefix is (still) legal */
76 #define	NZDIGITS	0x200	/* no zero digits detected */
77 
78 /*
79  * Conversion types.
80  */
81 #define	CT_CHAR		0	/* %c conversion */
82 #define	CT_CCL		1	/* %[...] conversion */
83 #define	CT_STRING	2	/* %s conversion */
84 #define	CT_INT		3	/* integer, i.e., strtoq or strtouq */
85 typedef u_quad_t (*ccfntype)(const char *, char **, int);
86 
87 static const u_char *__sccl(char *, const u_char *);
88 
89 int
90 sscanf(const char *ibuf, const char *fmt, ...)
91 {
92 	va_list ap;
93 	int ret;
94 
95 	va_start(ap, fmt);
96 	ret = vsscanf(ibuf, fmt, ap);
97 	va_end(ap);
98 	return(ret);
99 }
100 
101 int
102 vsscanf(const char *inp, char const *fmt0, va_list ap)
103 {
104 	int inr;
105 	const u_char *fmt = (const u_char *)fmt0;
106 	int c;			/* character from format, or conversion */
107 	size_t width;		/* field width, or 0 */
108 	char *p;		/* points into all kinds of strings */
109 	int n;			/* handy integer */
110 	int flags;		/* flags as defined above */
111 	char *p0;		/* saves original value of p when necessary */
112 	int nassigned;		/* number of fields assigned */
113 	int nconversions;	/* number of conversions */
114 	int nread;		/* number of characters consumed from fp */
115 	int base;		/* base argument to strtoq/strtouq */
116 	ccfntype ccfn;		/* conversion function (strtoq/strtouq) */
117 	char ccltab[256];	/* character class table for %[...] */
118 	char buf[BUF];		/* buffer for numeric conversions */
119 
120 	/* `basefix' is used to avoid `if' tests in the integer scanner */
121 	static short basefix[17] =
122 		{ 10, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16 };
123 
124 	inr = strlen(inp);
125 
126 	nassigned = 0;
127 	nconversions = 0;
128 	nread = 0;
129 	base = 0;		/* XXX just to keep gcc happy */
130 	ccfn = NULL;		/* XXX just to keep gcc happy */
131 	for (;;) {
132 		c = *fmt++;
133 		if (c == 0)
134 			return (nassigned);
135 		if (isspace(c)) {
136 			while (inr > 0 && isspace(*inp))
137 				nread++, inr--, inp++;
138 			continue;
139 		}
140 		if (c != '%')
141 			goto literal;
142 		width = 0;
143 		flags = 0;
144 		/*
145 		 * switch on the format.  continue if done;
146 		 * break once format type is derived.
147 		 */
148 again:		c = *fmt++;
149 		switch (c) {
150 		case '%':
151 literal:
152 			if (inr <= 0)
153 				goto input_failure;
154 			if (*inp != c)
155 				goto match_failure;
156 			inr--, inp++;
157 			nread++;
158 			continue;
159 
160 		case '*':
161 			flags |= SUPPRESS;
162 			goto again;
163 		case 'l':
164 			if (flags & LONG){
165 				flags &= ~LONG;
166 				flags |= QUAD;
167 			} else {
168 				flags |= LONG;
169 			}
170 			goto again;
171 		case 'q':
172 			flags |= QUAD;
173 			goto again;
174 		case 'h':
175 			if (flags & SHORT){
176 				flags &= ~SHORT;
177 				flags |= SHORTSHORT;
178 			} else {
179 				flags |= SHORT;
180 			}
181 			goto again;
182 
183 		case '0': case '1': case '2': case '3': case '4':
184 		case '5': case '6': case '7': case '8': case '9':
185 			width = width * 10 + c - '0';
186 			goto again;
187 
188 		/*
189 		 * Conversions.
190 		 *
191 		 */
192 		case 'd':
193 			c = CT_INT;
194 			ccfn = (ccfntype)strtoq;
195 			base = 10;
196 			break;
197 
198 		case 'i':
199 			c = CT_INT;
200 			ccfn = (ccfntype)strtoq;
201 			base = 0;
202 			break;
203 
204 		case 'o':
205 			c = CT_INT;
206 			ccfn = strtouq;
207 			base = 8;
208 			break;
209 
210 		case 'u':
211 			c = CT_INT;
212 			ccfn = strtouq;
213 			base = 10;
214 			break;
215 
216 		case 'x':
217 			flags |= PFXOK;	/* enable 0x prefixing */
218 			c = CT_INT;
219 			ccfn = strtouq;
220 			base = 16;
221 			break;
222 
223 		case 's':
224 			c = CT_STRING;
225 			break;
226 
227 		case '[':
228 			fmt = __sccl(ccltab, fmt);
229 			flags |= NOSKIP;
230 			c = CT_CCL;
231 			break;
232 
233 		case 'c':
234 			flags |= NOSKIP;
235 			c = CT_CHAR;
236 			break;
237 
238 		case 'p':	/* pointer format is like hex */
239 			flags |= POINTER | PFXOK;
240 			c = CT_INT;
241 			ccfn = strtouq;
242 			base = 16;
243 			break;
244 
245 		case 'n':
246 			nconversions++;
247 			if (flags & SUPPRESS)	/* ??? */
248 				continue;
249 			if (flags & SHORTSHORT)
250 				*va_arg(ap, char *) = nread;
251 			else if (flags & SHORT)
252 				*va_arg(ap, short *) = nread;
253 			else if (flags & LONG)
254 				*va_arg(ap, long *) = nread;
255 			else if (flags & QUAD)
256 				*va_arg(ap, quad_t *) = nread;
257 			else
258 				*va_arg(ap, int *) = nread;
259 			continue;
260 		}
261 
262 		/*
263 		 * We have a conversion that requires input.
264 		 */
265 		if (inr <= 0)
266 			goto input_failure;
267 
268 		/*
269 		 * Consume leading white space, except for formats
270 		 * that suppress this.
271 		 */
272 		if ((flags & NOSKIP) == 0) {
273 			while (isspace(*inp)) {
274 				nread++;
275 				if (--inr > 0)
276 					inp++;
277 				else
278 					goto input_failure;
279 			}
280 			/*
281 			 * Note that there is at least one character in
282 			 * the buffer, so conversions that do not set NOSKIP
283 			 * can no longer result in an input failure.
284 			 */
285 		}
286 
287 		/*
288 		 * Do the conversion.
289 		 */
290 		switch (c) {
291 
292 		case CT_CHAR:
293 			/* scan arbitrary characters (sets NOSKIP) */
294 			if (width == 0)
295 				width = 1;
296 			if (flags & SUPPRESS) {
297 				size_t sum = 0;
298 				for (;;) {
299 					if ((n = inr) < width) {
300 						sum += n;
301 						width -= n;
302 						inp += n;
303 						if (sum == 0)
304 							goto input_failure;
305 						break;
306 					} else {
307 						sum += width;
308 						inr -= width;
309 						inp += width;
310 						break;
311 					}
312 				}
313 				nread += sum;
314 			} else {
315 				bcopy(inp, va_arg(ap, char *), width);
316 				inr -= width;
317 				inp += width;
318 				nread += width;
319 				nassigned++;
320 			}
321 			nconversions++;
322 			break;
323 
324 		case CT_CCL:
325 			/* scan a (nonempty) character class (sets NOSKIP) */
326 			if (width == 0)
327 				width = (size_t)~0;	/* `infinity' */
328 			/* take only those things in the class */
329 			if (flags & SUPPRESS) {
330 				n = 0;
331 				while (ccltab[(unsigned char)*inp]) {
332 					n++, inr--, inp++;
333 					if (--width == 0)
334 						break;
335 					if (inr <= 0) {
336 						if (n == 0)
337 							goto input_failure;
338 						break;
339 					}
340 				}
341 				if (n == 0)
342 					goto match_failure;
343 			} else {
344 				p0 = p = va_arg(ap, char *);
345 				while (ccltab[(unsigned char)*inp]) {
346 					inr--;
347 					*p++ = *inp++;
348 					if (--width == 0)
349 						break;
350 					if (inr <= 0) {
351 						if (p == p0)
352 							goto input_failure;
353 						break;
354 					}
355 				}
356 				n = p - p0;
357 				if (n == 0)
358 					goto match_failure;
359 				*p = 0;
360 				nassigned++;
361 			}
362 			nread += n;
363 			nconversions++;
364 			break;
365 
366 		case CT_STRING:
367 			/* like CCL, but zero-length string OK, & no NOSKIP */
368 			if (width == 0)
369 				width = (size_t)~0;
370 			if (flags & SUPPRESS) {
371 				n = 0;
372 				while (!isspace(*inp)) {
373 					n++, inr--, inp++;
374 					if (--width == 0)
375 						break;
376 					if (inr <= 0)
377 						break;
378 				}
379 				nread += n;
380 			} else {
381 				p0 = p = va_arg(ap, char *);
382 				while (!isspace(*inp)) {
383 					inr--;
384 					*p++ = *inp++;
385 					if (--width == 0)
386 						break;
387 					if (inr <= 0)
388 						break;
389 				}
390 				*p = 0;
391 				nread += p - p0;
392 				nassigned++;
393 			}
394 			nconversions++;
395 			continue;
396 
397 		case CT_INT:
398 			/* scan an integer as if by strtoq/strtouq */
399 #ifdef hardway
400 			if (width == 0 || width > sizeof(buf) - 1)
401 				width = sizeof(buf) - 1;
402 #else
403 			/* size_t is unsigned, hence this optimisation */
404 			if (--width > sizeof(buf) - 2)
405 				width = sizeof(buf) - 2;
406 			width++;
407 #endif
408 			flags |= SIGNOK | NDIGITS | NZDIGITS;
409 			for (p = buf; width; width--) {
410 				c = *inp;
411 				/*
412 				 * Switch on the character; `goto ok'
413 				 * if we accept it as a part of number.
414 				 */
415 				switch (c) {
416 
417 				/*
418 				 * The digit 0 is always legal, but is
419 				 * special.  For %i conversions, if no
420 				 * digits (zero or nonzero) have been
421 				 * scanned (only signs), we will have
422 				 * base==0.  In that case, we should set
423 				 * it to 8 and enable 0x prefixing.
424 				 * Also, if we have not scanned zero digits
425 				 * before this, do not turn off prefixing
426 				 * (someone else will turn it off if we
427 				 * have scanned any nonzero digits).
428 				 */
429 				case '0':
430 					if (base == 0) {
431 						base = 8;
432 						flags |= PFXOK;
433 					}
434 					if (flags & NZDIGITS)
435 					    flags &= ~(SIGNOK|NZDIGITS|NDIGITS);
436 					else
437 					    flags &= ~(SIGNOK|PFXOK|NDIGITS);
438 					goto ok;
439 
440 				/* 1 through 7 always legal */
441 				case '1': case '2': case '3':
442 				case '4': case '5': case '6': case '7':
443 					base = basefix[base];
444 					flags &= ~(SIGNOK | PFXOK | NDIGITS);
445 					goto ok;
446 
447 				/* digits 8 and 9 ok iff decimal or hex */
448 				case '8': case '9':
449 					base = basefix[base];
450 					if (base <= 8)
451 						break;	/* not legal here */
452 					flags &= ~(SIGNOK | PFXOK | NDIGITS);
453 					goto ok;
454 
455 				/* letters ok iff hex */
456 				case 'A': case 'B': case 'C':
457 				case 'D': case 'E': case 'F':
458 				case 'a': case 'b': case 'c':
459 				case 'd': case 'e': case 'f':
460 					/* no need to fix base here */
461 					if (base <= 10)
462 						break;	/* not legal here */
463 					flags &= ~(SIGNOK | PFXOK | NDIGITS);
464 					goto ok;
465 
466 				/* sign ok only as first character */
467 				case '+': case '-':
468 					if (flags & SIGNOK) {
469 						flags &= ~SIGNOK;
470 						goto ok;
471 					}
472 					break;
473 
474 				/* x ok iff flag still set & 2nd char */
475 				case 'x': case 'X':
476 					if (flags & PFXOK && p == buf + 1) {
477 						base = 16;	/* if %i */
478 						flags &= ~PFXOK;
479 						goto ok;
480 					}
481 					break;
482 				}
483 
484 				/*
485 				 * If we got here, c is not a legal character
486 				 * for a number.  Stop accumulating digits.
487 				 */
488 				break;
489 		ok:
490 				/*
491 				 * c is legal: store it and look at the next.
492 				 */
493 				*p++ = c;
494 				if (--inr > 0)
495 					inp++;
496 				else
497 					break;		/* end of input */
498 			}
499 			/*
500 			 * If we had only a sign, it is no good; push
501 			 * back the sign.  If the number ends in `x',
502 			 * it was [sign] '0' 'x', so push back the x
503 			 * and treat it as [sign] '0'.
504 			 */
505 			if (flags & NDIGITS) {
506 				if (p > buf) {
507 					inp--;
508 					inr++;
509 				}
510 				goto match_failure;
511 			}
512 			c = ((u_char *)p)[-1];
513 			if (c == 'x' || c == 'X') {
514 				--p;
515 				inp--;
516 				inr++;
517 			}
518 			if ((flags & SUPPRESS) == 0) {
519 				u_quad_t res;
520 
521 				*p = 0;
522 				res = (*ccfn)(buf, (char **)NULL, base);
523 				if (flags & POINTER)
524 					*va_arg(ap, void **) =
525 						(void *)(uintptr_t)res;
526 				else if (flags & SHORTSHORT)
527 					*va_arg(ap, char *) = res;
528 				else if (flags & SHORT)
529 					*va_arg(ap, short *) = res;
530 				else if (flags & LONG)
531 					*va_arg(ap, long *) = res;
532 				else if (flags & QUAD)
533 					*va_arg(ap, quad_t *) = res;
534 				else
535 					*va_arg(ap, int *) = res;
536 				nassigned++;
537 			}
538 			nread += p - buf;
539 			nconversions++;
540 			break;
541 
542 		}
543 	}
544 input_failure:
545 	return (nconversions != 0 ? nassigned : -1);
546 match_failure:
547 	return (nassigned);
548 }
549 
550 /*
551  * Fill in the given table from the scanset at the given format
552  * (just after `[').  Return a pointer to the character past the
553  * closing `]'.  The table has a 1 wherever characters should be
554  * considered part of the scanset.
555  */
556 static const u_char *
557 __sccl(char *tab, const u_char *fmt)
558 {
559 	int c, n, v;
560 
561 	/* first `clear' the whole table */
562 	c = *fmt++;		/* first char hat => negated scanset */
563 	if (c == '^') {
564 		v = 1;		/* default => accept */
565 		c = *fmt++;	/* get new first char */
566 	} else
567 		v = 0;		/* default => reject */
568 
569 	/* XXX: Will not work if sizeof(tab*) > sizeof(char) */
570 	for (n = 0; n < 256; n++)
571 		     tab[n] = v;	/* memset(tab, v, 256) */
572 
573 	if (c == 0)
574 		return (fmt - 1);/* format ended before closing ] */
575 
576 	/*
577 	 * Now set the entries corresponding to the actual scanset
578 	 * to the opposite of the above.
579 	 *
580 	 * The first character may be ']' (or '-') without being special;
581 	 * the last character may be '-'.
582 	 */
583 	v = 1 - v;
584 	for (;;) {
585 		tab[c] = v;		/* take character c */
586 doswitch:
587 		n = *fmt++;		/* and examine the next */
588 		switch (n) {
589 
590 		case 0:			/* format ended too soon */
591 			return (fmt - 1);
592 
593 		case '-':
594 			/*
595 			 * A scanset of the form
596 			 *	[01+-]
597 			 * is defined as `the digit 0, the digit 1,
598 			 * the character +, the character -', but
599 			 * the effect of a scanset such as
600 			 *	[a-zA-Z0-9]
601 			 * is implementation defined.  The V7 Unix
602 			 * scanf treats `a-z' as `the letters a through
603 			 * z', but treats `a-a' as `the letter a, the
604 			 * character -, and the letter a'.
605 			 *
606 			 * For compatibility, the `-' is not considerd
607 			 * to define a range if the character following
608 			 * it is either a close bracket (required by ANSI)
609 			 * or is not numerically greater than the character
610 			 * we just stored in the table (c).
611 			 */
612 			n = *fmt;
613 			if (n == ']' || n < c) {
614 				c = '-';
615 				break;	/* resume the for(;;) */
616 			}
617 			fmt++;
618 			/* fill in the range */
619 			do {
620 			    tab[++c] = v;
621 			} while (c < n);
622 			c = n;
623 			/*
624 			 * Alas, the V7 Unix scanf also treats formats
625 			 * such as [a-c-e] as `the letters a through e'.
626 			 * This too is permitted by the standard....
627 			 */
628 			goto doswitch;
629 			break;
630 
631 		case ']':		/* end of scanset */
632 			return (fmt);
633 
634 		default:		/* just another character */
635 			c = n;
636 			break;
637 		}
638 	}
639 	/* NOTREACHED */
640 }
641 
642