xref: /freebsd/sys/kern/subr_scanf.c (revision 780fb4a2)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1990, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Chris Torek.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  * From: Id: vfscanf.c,v 1.13 1998/09/25 12:20:27 obrien Exp
35  * From: static char sccsid[] = "@(#)strtol.c	8.1 (Berkeley) 6/4/93";
36  * From: static char sccsid[] = "@(#)strtoul.c	8.1 (Berkeley) 6/4/93";
37  */
38 
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/ctype.h>
45 #include <sys/limits.h>
46 
47 /*
48  * Note that stdarg.h and the ANSI style va_start macro is used for both
49  * ANSI and traditional C compilers.
50  */
51 #include <machine/stdarg.h>
52 
53 #define	BUF		32 	/* Maximum length of numeric string. */
54 
55 /*
56  * Flags used during conversion.
57  */
58 #define	LONG		0x01	/* l: long or double */
59 #define	SHORT		0x04	/* h: short */
60 #define	SUPPRESS	0x08	/* suppress assignment */
61 #define	POINTER		0x10	/* weird %p pointer (`fake hex') */
62 #define	NOSKIP		0x20	/* do not skip blanks */
63 #define	QUAD		0x400
64 #define	SHORTSHORT	0x4000	/** hh: char */
65 
66 /*
67  * The following are used in numeric conversions only:
68  * SIGNOK, NDIGITS, DPTOK, and EXPOK are for floating point;
69  * SIGNOK, NDIGITS, PFXOK, and NZDIGITS are for integral.
70  */
71 #define	SIGNOK		0x40	/* +/- is (still) legal */
72 #define	NDIGITS		0x80	/* no digits detected */
73 
74 #define	DPTOK		0x100	/* (float) decimal point is still legal */
75 #define	EXPOK		0x200	/* (float) exponent (e+3, etc) still legal */
76 
77 #define	PFXOK		0x100	/* 0x prefix is (still) legal */
78 #define	NZDIGITS	0x200	/* no zero digits detected */
79 
80 /*
81  * Conversion types.
82  */
83 #define	CT_CHAR		0	/* %c conversion */
84 #define	CT_CCL		1	/* %[...] conversion */
85 #define	CT_STRING	2	/* %s conversion */
86 #define	CT_INT		3	/* integer, i.e., strtoq or strtouq */
87 typedef u_quad_t (*ccfntype)(const char *, char **, int);
88 
89 static const u_char *__sccl(char *, const u_char *);
90 
91 int
92 sscanf(const char *ibuf, const char *fmt, ...)
93 {
94 	va_list ap;
95 	int ret;
96 
97 	va_start(ap, fmt);
98 	ret = vsscanf(ibuf, fmt, ap);
99 	va_end(ap);
100 	return(ret);
101 }
102 
103 int
104 vsscanf(const char *inp, char const *fmt0, va_list ap)
105 {
106 	int inr;
107 	const u_char *fmt = (const u_char *)fmt0;
108 	int c;			/* character from format, or conversion */
109 	size_t width;		/* field width, or 0 */
110 	char *p;		/* points into all kinds of strings */
111 	int n;			/* handy integer */
112 	int flags;		/* flags as defined above */
113 	char *p0;		/* saves original value of p when necessary */
114 	int nassigned;		/* number of fields assigned */
115 	int nconversions;	/* number of conversions */
116 	int nread;		/* number of characters consumed from fp */
117 	int base;		/* base argument to strtoq/strtouq */
118 	ccfntype ccfn;		/* conversion function (strtoq/strtouq) */
119 	char ccltab[256];	/* character class table for %[...] */
120 	char buf[BUF];		/* buffer for numeric conversions */
121 
122 	/* `basefix' is used to avoid `if' tests in the integer scanner */
123 	static short basefix[17] =
124 		{ 10, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16 };
125 
126 	inr = strlen(inp);
127 
128 	nassigned = 0;
129 	nconversions = 0;
130 	nread = 0;
131 	base = 0;		/* XXX just to keep gcc happy */
132 	ccfn = NULL;		/* XXX just to keep gcc happy */
133 	for (;;) {
134 		c = *fmt++;
135 		if (c == 0)
136 			return (nassigned);
137 		if (isspace(c)) {
138 			while (inr > 0 && isspace(*inp))
139 				nread++, inr--, inp++;
140 			continue;
141 		}
142 		if (c != '%')
143 			goto literal;
144 		width = 0;
145 		flags = 0;
146 		/*
147 		 * switch on the format.  continue if done;
148 		 * break once format type is derived.
149 		 */
150 again:		c = *fmt++;
151 		switch (c) {
152 		case '%':
153 literal:
154 			if (inr <= 0)
155 				goto input_failure;
156 			if (*inp != c)
157 				goto match_failure;
158 			inr--, inp++;
159 			nread++;
160 			continue;
161 
162 		case '*':
163 			flags |= SUPPRESS;
164 			goto again;
165 		case 'l':
166 			if (flags & LONG){
167 				flags &= ~LONG;
168 				flags |= QUAD;
169 			} else {
170 				flags |= LONG;
171 			}
172 			goto again;
173 		case 'q':
174 			flags |= QUAD;
175 			goto again;
176 		case 'h':
177 			if (flags & SHORT){
178 				flags &= ~SHORT;
179 				flags |= SHORTSHORT;
180 			} else {
181 				flags |= SHORT;
182 			}
183 			goto again;
184 
185 		case '0': case '1': case '2': case '3': case '4':
186 		case '5': case '6': case '7': case '8': case '9':
187 			width = width * 10 + c - '0';
188 			goto again;
189 
190 		/*
191 		 * Conversions.
192 		 *
193 		 */
194 		case 'd':
195 			c = CT_INT;
196 			ccfn = (ccfntype)strtoq;
197 			base = 10;
198 			break;
199 
200 		case 'i':
201 			c = CT_INT;
202 			ccfn = (ccfntype)strtoq;
203 			base = 0;
204 			break;
205 
206 		case 'o':
207 			c = CT_INT;
208 			ccfn = strtouq;
209 			base = 8;
210 			break;
211 
212 		case 'u':
213 			c = CT_INT;
214 			ccfn = strtouq;
215 			base = 10;
216 			break;
217 
218 		case 'x':
219 			flags |= PFXOK;	/* enable 0x prefixing */
220 			c = CT_INT;
221 			ccfn = strtouq;
222 			base = 16;
223 			break;
224 
225 		case 's':
226 			c = CT_STRING;
227 			break;
228 
229 		case '[':
230 			fmt = __sccl(ccltab, fmt);
231 			flags |= NOSKIP;
232 			c = CT_CCL;
233 			break;
234 
235 		case 'c':
236 			flags |= NOSKIP;
237 			c = CT_CHAR;
238 			break;
239 
240 		case 'p':	/* pointer format is like hex */
241 			flags |= POINTER | PFXOK;
242 			c = CT_INT;
243 			ccfn = strtouq;
244 			base = 16;
245 			break;
246 
247 		case 'n':
248 			nconversions++;
249 			if (flags & SUPPRESS)	/* ??? */
250 				continue;
251 			if (flags & SHORTSHORT)
252 				*va_arg(ap, char *) = nread;
253 			else if (flags & SHORT)
254 				*va_arg(ap, short *) = nread;
255 			else if (flags & LONG)
256 				*va_arg(ap, long *) = nread;
257 			else if (flags & QUAD)
258 				*va_arg(ap, quad_t *) = nread;
259 			else
260 				*va_arg(ap, int *) = nread;
261 			continue;
262 		}
263 
264 		/*
265 		 * We have a conversion that requires input.
266 		 */
267 		if (inr <= 0)
268 			goto input_failure;
269 
270 		/*
271 		 * Consume leading white space, except for formats
272 		 * that suppress this.
273 		 */
274 		if ((flags & NOSKIP) == 0) {
275 			while (isspace(*inp)) {
276 				nread++;
277 				if (--inr > 0)
278 					inp++;
279 				else
280 					goto input_failure;
281 			}
282 			/*
283 			 * Note that there is at least one character in
284 			 * the buffer, so conversions that do not set NOSKIP
285 			 * can no longer result in an input failure.
286 			 */
287 		}
288 
289 		/*
290 		 * Do the conversion.
291 		 */
292 		switch (c) {
293 
294 		case CT_CHAR:
295 			/* scan arbitrary characters (sets NOSKIP) */
296 			if (width == 0)
297 				width = 1;
298 			if (flags & SUPPRESS) {
299 				size_t sum = 0;
300 				for (;;) {
301 					if ((n = inr) < width) {
302 						sum += n;
303 						width -= n;
304 						inp += n;
305 						if (sum == 0)
306 							goto input_failure;
307 						break;
308 					} else {
309 						sum += width;
310 						inr -= width;
311 						inp += width;
312 						break;
313 					}
314 				}
315 				nread += sum;
316 			} else {
317 				bcopy(inp, va_arg(ap, char *), width);
318 				inr -= width;
319 				inp += width;
320 				nread += width;
321 				nassigned++;
322 			}
323 			nconversions++;
324 			break;
325 
326 		case CT_CCL:
327 			/* scan a (nonempty) character class (sets NOSKIP) */
328 			if (width == 0)
329 				width = (size_t)~0;	/* `infinity' */
330 			/* take only those things in the class */
331 			if (flags & SUPPRESS) {
332 				n = 0;
333 				while (ccltab[(unsigned char)*inp]) {
334 					n++, inr--, inp++;
335 					if (--width == 0)
336 						break;
337 					if (inr <= 0) {
338 						if (n == 0)
339 							goto input_failure;
340 						break;
341 					}
342 				}
343 				if (n == 0)
344 					goto match_failure;
345 			} else {
346 				p0 = p = va_arg(ap, char *);
347 				while (ccltab[(unsigned char)*inp]) {
348 					inr--;
349 					*p++ = *inp++;
350 					if (--width == 0)
351 						break;
352 					if (inr <= 0) {
353 						if (p == p0)
354 							goto input_failure;
355 						break;
356 					}
357 				}
358 				n = p - p0;
359 				if (n == 0)
360 					goto match_failure;
361 				*p = 0;
362 				nassigned++;
363 			}
364 			nread += n;
365 			nconversions++;
366 			break;
367 
368 		case CT_STRING:
369 			/* like CCL, but zero-length string OK, & no NOSKIP */
370 			if (width == 0)
371 				width = (size_t)~0;
372 			if (flags & SUPPRESS) {
373 				n = 0;
374 				while (!isspace(*inp)) {
375 					n++, inr--, inp++;
376 					if (--width == 0)
377 						break;
378 					if (inr <= 0)
379 						break;
380 				}
381 				nread += n;
382 			} else {
383 				p0 = p = va_arg(ap, char *);
384 				while (!isspace(*inp)) {
385 					inr--;
386 					*p++ = *inp++;
387 					if (--width == 0)
388 						break;
389 					if (inr <= 0)
390 						break;
391 				}
392 				*p = 0;
393 				nread += p - p0;
394 				nassigned++;
395 			}
396 			nconversions++;
397 			continue;
398 
399 		case CT_INT:
400 			/* scan an integer as if by strtoq/strtouq */
401 #ifdef hardway
402 			if (width == 0 || width > sizeof(buf) - 1)
403 				width = sizeof(buf) - 1;
404 #else
405 			/* size_t is unsigned, hence this optimisation */
406 			if (--width > sizeof(buf) - 2)
407 				width = sizeof(buf) - 2;
408 			width++;
409 #endif
410 			flags |= SIGNOK | NDIGITS | NZDIGITS;
411 			for (p = buf; width; width--) {
412 				c = *inp;
413 				/*
414 				 * Switch on the character; `goto ok'
415 				 * if we accept it as a part of number.
416 				 */
417 				switch (c) {
418 
419 				/*
420 				 * The digit 0 is always legal, but is
421 				 * special.  For %i conversions, if no
422 				 * digits (zero or nonzero) have been
423 				 * scanned (only signs), we will have
424 				 * base==0.  In that case, we should set
425 				 * it to 8 and enable 0x prefixing.
426 				 * Also, if we have not scanned zero digits
427 				 * before this, do not turn off prefixing
428 				 * (someone else will turn it off if we
429 				 * have scanned any nonzero digits).
430 				 */
431 				case '0':
432 					if (base == 0) {
433 						base = 8;
434 						flags |= PFXOK;
435 					}
436 					if (flags & NZDIGITS)
437 					    flags &= ~(SIGNOK|NZDIGITS|NDIGITS);
438 					else
439 					    flags &= ~(SIGNOK|PFXOK|NDIGITS);
440 					goto ok;
441 
442 				/* 1 through 7 always legal */
443 				case '1': case '2': case '3':
444 				case '4': case '5': case '6': case '7':
445 					base = basefix[base];
446 					flags &= ~(SIGNOK | PFXOK | NDIGITS);
447 					goto ok;
448 
449 				/* digits 8 and 9 ok iff decimal or hex */
450 				case '8': case '9':
451 					base = basefix[base];
452 					if (base <= 8)
453 						break;	/* not legal here */
454 					flags &= ~(SIGNOK | PFXOK | NDIGITS);
455 					goto ok;
456 
457 				/* letters ok iff hex */
458 				case 'A': case 'B': case 'C':
459 				case 'D': case 'E': case 'F':
460 				case 'a': case 'b': case 'c':
461 				case 'd': case 'e': case 'f':
462 					/* no need to fix base here */
463 					if (base <= 10)
464 						break;	/* not legal here */
465 					flags &= ~(SIGNOK | PFXOK | NDIGITS);
466 					goto ok;
467 
468 				/* sign ok only as first character */
469 				case '+': case '-':
470 					if (flags & SIGNOK) {
471 						flags &= ~SIGNOK;
472 						goto ok;
473 					}
474 					break;
475 
476 				/* x ok iff flag still set & 2nd char */
477 				case 'x': case 'X':
478 					if (flags & PFXOK && p == buf + 1) {
479 						base = 16;	/* if %i */
480 						flags &= ~PFXOK;
481 						goto ok;
482 					}
483 					break;
484 				}
485 
486 				/*
487 				 * If we got here, c is not a legal character
488 				 * for a number.  Stop accumulating digits.
489 				 */
490 				break;
491 		ok:
492 				/*
493 				 * c is legal: store it and look at the next.
494 				 */
495 				*p++ = c;
496 				if (--inr > 0)
497 					inp++;
498 				else
499 					break;		/* end of input */
500 			}
501 			/*
502 			 * If we had only a sign, it is no good; push
503 			 * back the sign.  If the number ends in `x',
504 			 * it was [sign] '0' 'x', so push back the x
505 			 * and treat it as [sign] '0'.
506 			 */
507 			if (flags & NDIGITS) {
508 				if (p > buf) {
509 					inp--;
510 					inr++;
511 				}
512 				goto match_failure;
513 			}
514 			c = ((u_char *)p)[-1];
515 			if (c == 'x' || c == 'X') {
516 				--p;
517 				inp--;
518 				inr++;
519 			}
520 			if ((flags & SUPPRESS) == 0) {
521 				u_quad_t res;
522 
523 				*p = 0;
524 				res = (*ccfn)(buf, (char **)NULL, base);
525 				if (flags & POINTER)
526 					*va_arg(ap, void **) =
527 						(void *)(uintptr_t)res;
528 				else if (flags & SHORTSHORT)
529 					*va_arg(ap, char *) = res;
530 				else if (flags & SHORT)
531 					*va_arg(ap, short *) = res;
532 				else if (flags & LONG)
533 					*va_arg(ap, long *) = res;
534 				else if (flags & QUAD)
535 					*va_arg(ap, quad_t *) = res;
536 				else
537 					*va_arg(ap, int *) = res;
538 				nassigned++;
539 			}
540 			nread += p - buf;
541 			nconversions++;
542 			break;
543 
544 		}
545 	}
546 input_failure:
547 	return (nconversions != 0 ? nassigned : -1);
548 match_failure:
549 	return (nassigned);
550 }
551 
552 /*
553  * Fill in the given table from the scanset at the given format
554  * (just after `[').  Return a pointer to the character past the
555  * closing `]'.  The table has a 1 wherever characters should be
556  * considered part of the scanset.
557  */
558 static const u_char *
559 __sccl(char *tab, const u_char *fmt)
560 {
561 	int c, n, v;
562 
563 	/* first `clear' the whole table */
564 	c = *fmt++;		/* first char hat => negated scanset */
565 	if (c == '^') {
566 		v = 1;		/* default => accept */
567 		c = *fmt++;	/* get new first char */
568 	} else
569 		v = 0;		/* default => reject */
570 
571 	/* XXX: Will not work if sizeof(tab*) > sizeof(char) */
572 	for (n = 0; n < 256; n++)
573 		     tab[n] = v;	/* memset(tab, v, 256) */
574 
575 	if (c == 0)
576 		return (fmt - 1);/* format ended before closing ] */
577 
578 	/*
579 	 * Now set the entries corresponding to the actual scanset
580 	 * to the opposite of the above.
581 	 *
582 	 * The first character may be ']' (or '-') without being special;
583 	 * the last character may be '-'.
584 	 */
585 	v = 1 - v;
586 	for (;;) {
587 		tab[c] = v;		/* take character c */
588 doswitch:
589 		n = *fmt++;		/* and examine the next */
590 		switch (n) {
591 
592 		case 0:			/* format ended too soon */
593 			return (fmt - 1);
594 
595 		case '-':
596 			/*
597 			 * A scanset of the form
598 			 *	[01+-]
599 			 * is defined as `the digit 0, the digit 1,
600 			 * the character +, the character -', but
601 			 * the effect of a scanset such as
602 			 *	[a-zA-Z0-9]
603 			 * is implementation defined.  The V7 Unix
604 			 * scanf treats `a-z' as `the letters a through
605 			 * z', but treats `a-a' as `the letter a, the
606 			 * character -, and the letter a'.
607 			 *
608 			 * For compatibility, the `-' is not considered
609 			 * to define a range if the character following
610 			 * it is either a close bracket (required by ANSI)
611 			 * or is not numerically greater than the character
612 			 * we just stored in the table (c).
613 			 */
614 			n = *fmt;
615 			if (n == ']' || n < c) {
616 				c = '-';
617 				break;	/* resume the for(;;) */
618 			}
619 			fmt++;
620 			/* fill in the range */
621 			do {
622 			    tab[++c] = v;
623 			} while (c < n);
624 			c = n;
625 			/*
626 			 * Alas, the V7 Unix scanf also treats formats
627 			 * such as [a-c-e] as `the letters a through e'.
628 			 * This too is permitted by the standard....
629 			 */
630 			goto doswitch;
631 			break;
632 
633 		case ']':		/* end of scanset */
634 			return (fmt);
635 
636 		default:		/* just another character */
637 			c = n;
638 			break;
639 		}
640 	}
641 	/* NOTREACHED */
642 }
643 
644