xref: /openbsd-src/usr.bin/compress/zopen.c (revision 91f110e064cd7c194e59e019b83bb7496c1c84d4)
1 /*	$OpenBSD: zopen.c,v 1.18 2011/09/22 10:41:04 deraadt Exp $	*/
2 /*	$NetBSD: zopen.c,v 1.5 1995/03/26 09:44:53 glass Exp $	*/
3 
4 /*-
5  * Copyright (c) 1985, 1986, 1992, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * This code is derived from software contributed to Berkeley by
9  * Diomidis Spinellis and James A. Woods, derived from original
10  * work by Spencer Thomas and Joseph Orost.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	From: @(#)zopen.c	8.1 (Berkeley) 6/27/93
37  */
38 
39 /*-
40  * fcompress.c - File compression ala IEEE Computer, June 1984.
41  *
42  * Compress authors:
43  *		Spencer W. Thomas	(decvax!utah-cs!thomas)
44  *		Jim McKie		(decvax!mcvax!jim)
45  *		Steve Davies		(decvax!vax135!petsd!peora!srd)
46  *		Ken Turkowski		(decvax!decwrl!turtlevax!ken)
47  *		James A. Woods		(decvax!ihnp4!ames!jaw)
48  *		Joe Orost		(decvax!vax135!petsd!joe)
49  *
50  * Cleaned up and converted to library returning I/O streams by
51  * Diomidis Spinellis <dds@doc.ic.ac.uk>.
52  *
53  * zopen(filename, mode, bits)
54  *	Returns a FILE * that can be used for read or write.  The modes
55  *	supported are only "r" and "w".  Seeking is not allowed.  On
56  *	reading the file is decompressed, on writing it is compressed.
57  *	The output is compatible with compress(1) with 16 bit tables.
58  *	Any file produced by compress(1) can be read.
59  */
60 
61 #include <sys/param.h>
62 #include <sys/stat.h>
63 
64 #include <ctype.h>
65 #include <errno.h>
66 #include <signal.h>
67 #include <stdio.h>
68 #include <stdlib.h>
69 #include <string.h>
70 #include <unistd.h>
71 #include <fcntl.h>
72 #include "compress.h"
73 
74 #define	BITS		16		/* Default bits. */
75 #define	HSIZE		69001		/* 95% occupancy */
76 #define	ZBUFSIZ		8192		/* I/O buffer size */
77 
78 /* A code_int must be able to hold 2**BITS values of type int, and also -1. */
79 typedef long code_int;
80 typedef long count_int;
81 
82 static const u_char z_magic[] =
83 	{'\037', '\235'};		/* 1F 9D */
84 
85 #define	BIT_MASK	0x1f		/* Defines for third byte of header. */
86 #define	BLOCK_MASK	0x80
87 
88 /*
89  * Masks 0x40 and 0x20 are free.  I think 0x20 should mean that there is
90  * a fourth header byte (for expansion).
91  */
92 #define	INIT_BITS 9			/* Initial number of bits/code. */
93 
94 #define	MAXCODE(n_bits)	((1 << (n_bits)) - 1)
95 
96 struct s_zstate {
97 	int zs_fd;			/* File stream for I/O */
98 	char zs_mode;			/* r or w */
99 	enum {
100 		S_START, S_MAGIC, S_MIDDLE, S_EOF
101 	} zs_state;			/* State of computation */
102 	int zs_n_bits;			/* Number of bits/code. */
103 	int zs_maxbits;			/* User settable max # bits/code. */
104 	code_int zs_maxcode;		/* Maximum code, given n_bits. */
105 	code_int zs_maxmaxcode;		/* Should NEVER generate this code. */
106 	count_int zs_htab[HSIZE];
107 	u_short zs_codetab[HSIZE];
108 	code_int zs_hsize;		/* For dynamic table sizing. */
109 	code_int zs_free_ent;		/* First unused entry. */
110 	/*
111 	 * Block compression parameters -- after all codes are used up,
112 	 * and compression rate changes, start over.
113 	 */
114 	int zs_block_compress;
115 	int zs_clear_flg;
116 	long zs_ratio;
117 	count_int zs_checkpoint;
118 	long zs_in_count;		/* Length of input. */
119 	long zs_bytes_out;		/* Length of output. */
120 	long zs_out_count;		/* # of codes output (for debugging).*/
121 	u_char zs_buf[ZBUFSIZ];		/* I/O buffer */
122 	u_char *zs_bp;			/* Current I/O window in the zs_buf */
123 	int zs_offset;			/* Number of bits in the zs_buf */
124 	union {
125 		struct {
126 			long zs_fcode;
127 			code_int zs_ent;
128 			code_int zs_hsize_reg;
129 			int zs_hshift;
130 		} w;			/* Write parameters */
131 		struct {
132 			u_char *zs_stackp, *zs_ebp;
133 			int zs_finchar;
134 			code_int zs_code, zs_oldcode, zs_incode;
135 			int zs_size;
136 		} r;			/* Read parameters */
137 	} u;
138 };
139 
140 /* Definitions to retain old variable names */
141 #define zs_fcode	u.w.zs_fcode
142 #define zs_ent		u.w.zs_ent
143 #define zs_hsize_reg	u.w.zs_hsize_reg
144 #define zs_hshift	u.w.zs_hshift
145 #define zs_stackp	u.r.zs_stackp
146 #define zs_finchar	u.r.zs_finchar
147 #define zs_code		u.r.zs_code
148 #define zs_oldcode	u.r.zs_oldcode
149 #define zs_incode	u.r.zs_incode
150 #define zs_size		u.r.zs_size
151 #define zs_ebp		u.r.zs_ebp
152 
153 /*
154  * To save much memory, we overlay the table used by compress() with those
155  * used by decompress().  The tab_prefix table is the same size and type as
156  * the codetab.  The tab_suffix table needs 2**BITS characters.  We get this
157  * from the beginning of htab.  The output stack uses the rest of htab, and
158  * contains characters.  There is plenty of room for any possible stack
159  * (stack used to be 8000 characters).
160  */
161 
162 #define	htabof(i)	zs->zs_htab[i]
163 #define	codetabof(i)	zs->zs_codetab[i]
164 
165 #define	tab_prefixof(i)	codetabof(i)
166 #define	tab_suffixof(i)	((u_char *)(zs->zs_htab))[i]
167 #define	de_stack	((u_char *)&tab_suffixof(1 << BITS))
168 
169 #define	CHECK_GAP 10000		/* Ratio check interval. */
170 
171 /*
172  * the next two codes should not be changed lightly, as they must not
173  * lie within the contiguous general code space.
174  */
175 #define	FIRST	257		/* First free entry. */
176 #define	CLEAR	256		/* Table clear output code. */
177 
178 static int	cl_block(struct s_zstate *);
179 static void	cl_hash(struct s_zstate *, count_int);
180 static code_int	getcode(struct s_zstate *);
181 static int	output(struct s_zstate *, code_int);
182 
183 /*-
184  * Algorithm from "A Technique for High Performance Data Compression",
185  * Terry A. Welch, IEEE Computer Vol 17, No 6 (June 1984), pp 8-19.
186  *
187  * Algorithm:
188  *	Modified Lempel-Ziv method (LZW).  Basically finds common
189  * substrings and replaces them with a variable size code.  This is
190  * deterministic, and can be done on the fly.  Thus, the decompression
191  * procedure needs no input table, but tracks the way the table was built.
192  */
193 
194 /*-
195  * compress write
196  *
197  * Algorithm:  use open addressing double hashing (no chaining) on the
198  * prefix code / next character combination.  We do a variant of Knuth's
199  * algorithm D (vol. 3, sec. 6.4) along with G. Knott's relatively-prime
200  * secondary probe.  Here, the modular division first probe is gives way
201  * to a faster exclusive-or manipulation.  Also do block compression with
202  * an adaptive reset, whereby the code table is cleared when the compression
203  * ratio decreases, but after the table fills.  The variable-length output
204  * codes are re-sized at this point, and a special CLEAR code is generated
205  * for the decompressor.  Late addition:  construct the table according to
206  * file size for noticeable speed improvement on small files.  Please direct
207  * questions about this implementation to ames!jaw.
208  */
209 int
210 zwrite(void *cookie, const char *wbp, int num)
211 {
212 	code_int i;
213 	int c, disp;
214 	struct s_zstate *zs;
215 	const u_char *bp;
216 	u_char tmp;
217 	int count;
218 
219 	zs = cookie;
220 	count = num;
221 	bp = (u_char *)wbp;
222 	switch (zs->zs_state) {
223 	case S_MAGIC:
224 		return -1;
225 	case S_EOF:
226 		return 0;
227 	case S_START:
228 		zs->zs_state = S_MIDDLE;
229 
230 		zs->zs_maxmaxcode = 1L << zs->zs_maxbits;
231 		if (write(zs->zs_fd, z_magic, sizeof(z_magic)) !=
232 		    sizeof(z_magic))
233 			return (-1);
234 		tmp = (u_char)(zs->zs_maxbits | zs->zs_block_compress);
235 		if (write(zs->zs_fd, &tmp, sizeof(tmp)) != sizeof(tmp))
236 			return (-1);
237 
238 		zs->zs_bp = zs->zs_buf;
239 		zs->zs_offset = 0;
240 		zs->zs_bytes_out = 3;	/* Includes 3-byte header mojo. */
241 		zs->zs_out_count = 0;
242 		zs->zs_clear_flg = 0;
243 		zs->zs_ratio = 0;
244 		zs->zs_in_count = 1;
245 		zs->zs_checkpoint = CHECK_GAP;
246 		zs->zs_maxcode = MAXCODE(zs->zs_n_bits = INIT_BITS);
247 		zs->zs_free_ent = ((zs->zs_block_compress) ? FIRST : 256);
248 
249 		zs->zs_ent = *bp++;
250 		--count;
251 
252 		zs->zs_hshift = 0;
253 		for (zs->zs_fcode = (long)zs->zs_hsize; zs->zs_fcode < 65536L;
254 		    zs->zs_fcode *= 2L)
255 			zs->zs_hshift++;
256 		/* Set hash code range bound. */
257 		zs->zs_hshift = 8 - zs->zs_hshift;
258 
259 		zs->zs_hsize_reg = zs->zs_hsize;
260 		/* Clear hash table. */
261 		cl_hash(zs, (count_int)zs->zs_hsize_reg);
262 
263 	case S_MIDDLE:
264 		for (i = 0; count-- > 0;) {
265 			c = *bp++;
266 			zs->zs_in_count++;
267 			zs->zs_fcode = (long)(((long)c << zs->zs_maxbits) +
268 			    zs->zs_ent);
269 			/* Xor hashing. */
270 			i = ((c << zs->zs_hshift) ^ zs->zs_ent);
271 
272 			if (htabof(i) == zs->zs_fcode) {
273 				zs->zs_ent = codetabof(i);
274 				continue;
275 			} else if ((long)htabof(i) < 0)	/* Empty slot. */
276 				goto nomatch;
277 			/* Secondary hash (after G. Knott). */
278 			disp = zs->zs_hsize_reg - i;
279 			if (i == 0)
280 			disp = 1;
281 probe:			if ((i -= disp) < 0)
282 				i += zs->zs_hsize_reg;
283 
284 			if (htabof(i) == zs->zs_fcode) {
285 				zs->zs_ent = codetabof(i);
286 				continue;
287 			}
288 			if ((long)htabof(i) >= 0)
289 				goto probe;
290 nomatch:		if (output(zs, (code_int) zs->zs_ent) == -1)
291 				return (-1);
292 			zs->zs_out_count++;
293 			zs->zs_ent = c;
294 			if (zs->zs_free_ent < zs->zs_maxmaxcode) {
295 				/* code -> hashtable */
296 				codetabof(i) = zs->zs_free_ent++;
297 				htabof(i) = zs->zs_fcode;
298 			} else if ((count_int)zs->zs_in_count >=
299 			    zs->zs_checkpoint && zs->zs_block_compress) {
300 				if (cl_block(zs) == -1)
301 					return (-1);
302 			}
303 		}
304 	}
305 	return (num);
306 }
307 
308 int
309 z_close(void *cookie, struct z_info *info, const char *name, struct stat *sb)
310 {
311 	struct s_zstate *zs;
312 	int rval;
313 
314 	zs = cookie;
315 	if (zs->zs_mode == 'w') {		/* Put out the final code. */
316 		if (output(zs, (code_int) zs->zs_ent) == -1) {
317 			(void)close(zs->zs_fd);
318 			free(zs);
319 			return (-1);
320 		}
321 		zs->zs_out_count++;
322 		if (output(zs, (code_int) - 1) == -1) {
323 			(void)close(zs->zs_fd);
324 			free(zs);
325 			return (-1);
326 		}
327 	}
328 
329 	if (info != NULL) {
330 		info->mtime = 0;
331 		info->crc = (u_int32_t)-1;
332 		info->hlen = 0;
333 		info->total_in = (off_t)zs->zs_in_count;
334 		info->total_out = (off_t)zs->zs_bytes_out;
335 	}
336 
337 #ifndef SAVECORE
338 	setfile(name, zs->zs_fd, sb);
339 #endif
340 	rval = close(zs->zs_fd);
341 	free(zs);
342 	return (rval);
343 }
344 
345 static int
346 zclose(void *cookie)
347 {
348 	return z_close(cookie, NULL, NULL, NULL);
349 }
350 
351 /*-
352  * Output the given code.
353  * Inputs:
354  *	code:	A n_bits-bit integer.  If == -1, then EOF.  This assumes
355  *		that n_bits =< (long)wordsize - 1.
356  * Outputs:
357  *	Outputs code to the file.
358  * Assumptions:
359  *	Chars are 8 bits long.
360  * Algorithm:
361  *	Maintain a BITS character long buffer (so that 8 codes will
362  * fit in it exactly).  Use the VAX insv instruction to insert each
363  * code in turn.  When the buffer fills up empty it and start over.
364  */
365 
366 static const u_char lmask[9] =
367 	{0xff, 0xfe, 0xfc, 0xf8, 0xf0, 0xe0, 0xc0, 0x80, 0x00};
368 static const u_char rmask[9] =
369 	{0x00, 0x01, 0x03, 0x07, 0x0f, 0x1f, 0x3f, 0x7f, 0xff};
370 
371 static int
372 output(struct s_zstate *zs, code_int ocode)
373 {
374 	int bits;
375 
376 	if (ocode >= 0) {
377 		int r_off;
378 		u_char *bp;
379 
380 		/* Get to the first byte. */
381 		bp = zs->zs_bp + (zs->zs_offset >> 3);
382 		r_off = zs->zs_offset & 7;
383 		bits = zs->zs_n_bits;
384 
385 		/*
386 		 * Since ocode is always >= 8 bits, only need to mask the first
387 		 * hunk on the left.
388 		 */
389 		*bp = (*bp & rmask[r_off]) | ((ocode << r_off) & lmask[r_off]);
390 		bp++;
391 		bits -= (8 - r_off);
392 		ocode >>= 8 - r_off;
393 		/* Get any 8 bit parts in the middle (<=1 for up to 16 bits) */
394 		if (bits >= 8) {
395 			*bp++ = ocode;
396 			ocode >>= 8;
397 			bits -= 8;
398 		}
399 		/* Last bits. */
400 		if (bits)
401 			*bp = ocode;
402 		zs->zs_offset += zs->zs_n_bits;
403 		if (zs->zs_offset == (zs->zs_n_bits << 3)) {
404 			zs->zs_bp += zs->zs_n_bits;
405 			zs->zs_offset = 0;
406 		}
407 		/*
408 		 * If the next entry is going to be too big for the ocode size,
409 		 * then increase it, if possible.
410 		 */
411 		if (zs->zs_free_ent > zs->zs_maxcode ||
412 		    (zs->zs_clear_flg > 0)) {
413 			/*
414 			 * Write the whole buffer, because the input side won't
415 			 * discover the size increase until after it has read it
416 			 */
417 			if (zs->zs_offset > 0) {
418 				zs->zs_bp += zs->zs_n_bits;
419 				zs->zs_offset = 0;
420 			}
421 
422 			if (zs->zs_clear_flg) {
423 				zs->zs_maxcode =
424 					MAXCODE(zs->zs_n_bits = INIT_BITS);
425 				zs->zs_clear_flg = 0;
426 			} else {
427 				zs->zs_n_bits++;
428 				if (zs->zs_n_bits == zs->zs_maxbits)
429 					zs->zs_maxcode = zs->zs_maxmaxcode;
430 				else
431 					zs->zs_maxcode =
432 					    MAXCODE(zs->zs_n_bits);
433 			}
434 		}
435 
436 		if (zs->zs_bp + zs->zs_n_bits > &zs->zs_buf[ZBUFSIZ]) {
437 			bits = zs->zs_bp - zs->zs_buf;
438 			if (write(zs->zs_fd, zs->zs_buf, bits) != bits)
439 				return (-1);
440 			zs->zs_bytes_out += bits;
441 			if (zs->zs_offset > 0)
442 				fprintf (stderr, "zs_offset != 0\n");
443 			zs->zs_bp = zs->zs_buf;
444 		}
445 	} else {
446 		/* At EOF, write the rest of the buffer. */
447 		if (zs->zs_offset > 0)
448 			zs->zs_bp += (zs->zs_offset + 7) / 8;
449 		if (zs->zs_bp > zs->zs_buf) {
450 			bits = zs->zs_bp - zs->zs_buf;
451 			if (write(zs->zs_fd, zs->zs_buf, bits) != bits)
452 				return (-1);
453 			zs->zs_bytes_out += bits;
454 		}
455 		zs->zs_offset = 0;
456 		zs->zs_bp = zs->zs_buf;
457 	}
458 	return (0);
459 }
460 
461 /*
462  * Decompress read.  This routine adapts to the codes in the file building
463  * the "string" table on-the-fly; requiring no table to be stored in the
464  * compressed file.  The tables used herein are shared with those of the
465  * compress() routine.  See the definitions above.
466  */
467 int
468 zread(void *cookie, char *rbp, int num)
469 {
470 	u_int count;
471 	struct s_zstate *zs;
472 	u_char *bp, header[3];
473 
474 	if (num == 0)
475 		return (0);
476 
477 	zs = cookie;
478 	count = num;
479 	bp = (u_char *)rbp;
480 	switch (zs->zs_state) {
481 	case S_START:
482 		zs->zs_state = S_MIDDLE;
483 		zs->zs_bp = zs->zs_buf;
484 		header[0] = header[1] = header[2] = '\0';
485 		read(zs->zs_fd, header, sizeof(header));
486 		break;
487 	case S_MAGIC:
488 		zs->zs_state = S_MIDDLE;
489 		zs->zs_bp = zs->zs_buf;
490 		header[0] = z_magic[0];
491 		header[1] = z_magic[1];
492 		header[2] = '\0';
493 		read(zs->zs_fd, &header[2], 1);
494 		break;
495 	case S_MIDDLE:
496 		goto middle;
497 	case S_EOF:
498 		goto eof;
499 	}
500 
501 	/* Check the magic number */
502 	if (header[0] != z_magic[0] || header[1] != z_magic[1]) {
503 		errno = EFTYPE;
504 		return (-1);
505 	}
506 	zs->zs_maxbits = header[2];	/* Set -b from file. */
507 	zs->zs_in_count += sizeof(header);
508 	zs->zs_block_compress = zs->zs_maxbits & BLOCK_MASK;
509 	zs->zs_maxbits &= BIT_MASK;
510 	zs->zs_maxmaxcode = 1L << zs->zs_maxbits;
511 	if (zs->zs_maxbits > BITS) {
512 		errno = EFTYPE;
513 		return (-1);
514 	}
515 	/* As above, initialize the first 256 entries in the table. */
516 	zs->zs_maxcode = MAXCODE(zs->zs_n_bits = INIT_BITS);
517 	for (zs->zs_code = 255; zs->zs_code >= 0; zs->zs_code--) {
518 		tab_prefixof(zs->zs_code) = 0;
519 		tab_suffixof(zs->zs_code) = (u_char) zs->zs_code;
520 	}
521 	zs->zs_free_ent = zs->zs_block_compress ? FIRST : 256;
522 
523 	zs->zs_finchar = zs->zs_oldcode = getcode(zs);
524 	if (zs->zs_oldcode == -1)	/* EOF already? */
525 		return (0);	/* Get out of here */
526 
527 	/* First code must be 8 bits = char. */
528 	*bp++ = (u_char)zs->zs_finchar;
529 	count--;
530 	zs->zs_stackp = de_stack;
531 
532 	while ((zs->zs_code = getcode(zs)) > -1) {
533 
534 		if ((zs->zs_code == CLEAR) && zs->zs_block_compress) {
535 			for (zs->zs_code = 255; zs->zs_code >= 0;
536 			    zs->zs_code--)
537 				tab_prefixof(zs->zs_code) = 0;
538 			zs->zs_clear_flg = 1;
539 			zs->zs_free_ent = FIRST - 1;
540 			if ((zs->zs_code = getcode(zs)) == -1)	/* O, untimely death! */
541 				break;
542 		}
543 		zs->zs_incode = zs->zs_code;
544 
545 		/* Special case for KwKwK string. */
546 		if (zs->zs_code >= zs->zs_free_ent) {
547 			*zs->zs_stackp++ = zs->zs_finchar;
548 			zs->zs_code = zs->zs_oldcode;
549 		}
550 
551 		/* Generate output characters in reverse order. */
552 		while (zs->zs_code >= 256) {
553 			/*
554 			 * Bad input file may cause zs_stackp to overflow
555 			 * zs_htab; check here and abort decompression,
556 			 * that's better than dumping core.
557 			 */
558 			if (zs->zs_stackp >= (u_char *)&zs->zs_htab[HSIZE]) {
559 				errno = EINVAL;
560 				return (-1);
561 			}
562 			*zs->zs_stackp++ = tab_suffixof(zs->zs_code);
563 			zs->zs_code = tab_prefixof(zs->zs_code);
564 		}
565 		*zs->zs_stackp++ = zs->zs_finchar = tab_suffixof(zs->zs_code);
566 
567 		/* And put them out in forward order.  */
568 middle:		do {
569 			if (count-- == 0) {
570 				zs->zs_bytes_out += num;
571 				return (num);
572 			}
573 			*bp++ = *--zs->zs_stackp;
574 		} while (zs->zs_stackp > de_stack);
575 
576 		/* Generate the new entry. */
577 		if ((zs->zs_code = zs->zs_free_ent) < zs->zs_maxmaxcode) {
578 			tab_prefixof(zs->zs_code) = (u_short) zs->zs_oldcode;
579 			tab_suffixof(zs->zs_code) = zs->zs_finchar;
580 			zs->zs_free_ent = zs->zs_code + 1;
581 		}
582 
583 		/* Remember previous code. */
584 		zs->zs_oldcode = zs->zs_incode;
585 	}
586 	zs->zs_state = S_EOF;
587 	zs->zs_bytes_out += num - count;
588 eof:	return (num - count);
589 }
590 
591 /*-
592  * Read one code from the standard input.  If EOF, return -1.
593  * Inputs:
594  *	stdin
595  * Outputs:
596  *	code or -1 is returned.
597  */
598 static code_int
599 getcode(struct s_zstate *zs)
600 {
601 	code_int gcode;
602 	int r_off, bits;
603 	u_char *bp;
604 
605 	if (zs->zs_clear_flg > 0 || zs->zs_offset >= zs->zs_size ||
606 	    zs->zs_free_ent > zs->zs_maxcode) {
607 
608 		zs->zs_bp += zs->zs_n_bits;
609 		/*
610 		 * If the next entry will be too big for the current gcode
611 		 * size, then we must increase the size.  This implies reading
612 		 * a new buffer full, too.
613 		 */
614 		if (zs->zs_free_ent > zs->zs_maxcode) {
615 			zs->zs_n_bits++;
616 			if (zs->zs_n_bits == zs->zs_maxbits) {
617 				/* Won't get any bigger now. */
618 				zs->zs_maxcode = zs->zs_maxmaxcode;
619 			} else
620 				zs->zs_maxcode = MAXCODE(zs->zs_n_bits);
621 		}
622 		if (zs->zs_clear_flg > 0) {
623 			zs->zs_maxcode = MAXCODE(zs->zs_n_bits = INIT_BITS);
624 			zs->zs_clear_flg = 0;
625 		}
626 
627 		/* fill the buffer up to the neck */
628 		if (zs->zs_bp + zs->zs_n_bits > zs->zs_ebp) {
629 			for (bp = zs->zs_buf; zs->zs_bp < zs->zs_ebp;
630 				*bp++ = *zs->zs_bp++);
631 			if ((bits = read(zs->zs_fd, bp, ZBUFSIZ -
632 			    (bp - zs->zs_buf))) < 0)
633 				return -1;
634 			zs->zs_in_count += bits;
635 			zs->zs_bp = zs->zs_buf;
636 			zs->zs_ebp = bp + bits;
637 		}
638 		zs->zs_offset = 0;
639 		zs->zs_size = MIN(zs->zs_n_bits, zs->zs_ebp - zs->zs_bp);
640 		if (zs->zs_size == 0)
641 			return -1;
642 		/* Round size down to integral number of codes. */
643 		zs->zs_size = (zs->zs_size << 3) - (zs->zs_n_bits - 1);
644 	}
645 
646 	bp = zs->zs_bp;
647 	r_off = zs->zs_offset;
648 	bits = zs->zs_n_bits;
649 
650 	/* Get to the first byte. */
651 	bp += (r_off >> 3);
652 	r_off &= 7;
653 
654 	/* Get first part (low order bits). */
655 	gcode = (*bp++ >> r_off);
656 	bits -= (8 - r_off);
657 	r_off = 8 - r_off;	/* Now, roffset into gcode word. */
658 
659 	/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
660 	if (bits >= 8) {
661 		gcode |= *bp++ << r_off;
662 		r_off += 8;
663 		bits -= 8;
664 	}
665 
666 	/* High order bits. */
667 	gcode |= (*bp & rmask[bits]) << r_off;
668 	zs->zs_offset += zs->zs_n_bits;
669 
670 	return (gcode);
671 }
672 
673 /* Table clear for block compress. */
674 static int
675 cl_block(struct s_zstate *zs)
676 {
677 	long rat;
678 
679 	zs->zs_checkpoint = zs->zs_in_count + CHECK_GAP;
680 
681 	if (zs->zs_in_count > 0x007fffff) {	/* Shift will overflow. */
682 		rat = zs->zs_bytes_out >> 8;
683 		if (rat == 0)		/* Don't divide by zero. */
684 			rat = 0x7fffffff;
685 		else
686 			rat = zs->zs_in_count / rat;
687 	} else {
688 		/* 8 fractional bits. */
689 		rat = (zs->zs_in_count << 8) / zs->zs_bytes_out;
690 	}
691 	if (rat > zs->zs_ratio)
692 		zs->zs_ratio = rat;
693 	else {
694 		zs->zs_ratio = 0;
695 		cl_hash(zs, (count_int) zs->zs_hsize);
696 		zs->zs_free_ent = FIRST;
697 		zs->zs_clear_flg = 1;
698 		if (output(zs, (code_int) CLEAR) == -1)
699 			return (-1);
700 	}
701 	return (0);
702 }
703 
704 /* Reset code table. */
705 static void
706 cl_hash(struct s_zstate *zs, count_int cl_hsize)
707 {
708 	count_int *htab_p;
709 	long i, m1;
710 
711 	m1 = -1;
712 	htab_p = zs->zs_htab + cl_hsize;
713 	i = cl_hsize - 16;
714 	do {			/* Might use Sys V memset(3) here. */
715 		*(htab_p - 16) = m1;
716 		*(htab_p - 15) = m1;
717 		*(htab_p - 14) = m1;
718 		*(htab_p - 13) = m1;
719 		*(htab_p - 12) = m1;
720 		*(htab_p - 11) = m1;
721 		*(htab_p - 10) = m1;
722 		*(htab_p - 9) = m1;
723 		*(htab_p - 8) = m1;
724 		*(htab_p - 7) = m1;
725 		*(htab_p - 6) = m1;
726 		*(htab_p - 5) = m1;
727 		*(htab_p - 4) = m1;
728 		*(htab_p - 3) = m1;
729 		*(htab_p - 2) = m1;
730 		*(htab_p - 1) = m1;
731 		htab_p -= 16;
732 	} while ((i -= 16) >= 0);
733 	for (i += 16; i > 0; i--)
734 		*--htab_p = m1;
735 }
736 
737 FILE *
738 zopen(const char *name, const char *mode, int bits)
739 {
740 	int fd;
741 	void *cookie;
742 	if ((fd = open(name, (*mode=='r'? O_RDONLY:O_WRONLY|O_CREAT),
743 	    S_IRUSR|S_IWUSR|S_IRGRP|S_IROTH)) == -1)
744 		return NULL;
745 	if ((cookie = z_open(fd, mode, NULL, bits, 0, 0)) == NULL) {
746 		close(fd);
747 		return NULL;
748 	}
749 	return funopen(cookie, (*mode == 'r'?zread:NULL),
750 	    (*mode == 'w'?zwrite:NULL), NULL, zclose);
751 }
752 
753 void *
754 z_open(int fd, const char *mode, char *name, int bits,
755     u_int32_t mtime, int gotmagic)
756 {
757 	struct s_zstate *zs;
758 
759 	if ((mode[0] != 'r' && mode[0] != 'w') || mode[1] != '\0' ||
760 	    bits < 0 || bits > BITS) {
761 		errno = EINVAL;
762 		return (NULL);
763 	}
764 
765 	if ((zs = calloc(1, sizeof(struct s_zstate))) == NULL)
766 		return (NULL);
767 
768 	/* User settable max # bits/code. */
769 	zs->zs_maxbits = bits ? bits : BITS;
770 	/* Should NEVER generate this code. */
771 	zs->zs_maxmaxcode = 1 << zs->zs_maxbits;
772 	zs->zs_hsize = HSIZE;		/* For dynamic table sizing. */
773 	zs->zs_free_ent = 0;		/* First unused entry. */
774 	zs->zs_block_compress = BLOCK_MASK;
775 	zs->zs_clear_flg = 0;
776 	zs->zs_ratio = 0;
777 	zs->zs_checkpoint = CHECK_GAP;
778 	zs->zs_in_count = 0;		/* Length of input. */
779 	zs->zs_out_count = 0;		/* # of codes output (for debugging).*/
780 	zs->zs_state = gotmagic ? S_MAGIC : S_START;
781 	zs->zs_offset = 0;
782 	zs->zs_size = 0;
783 	zs->zs_mode = mode[0];
784 	zs->zs_bp = zs->zs_ebp = zs->zs_buf;
785 
786 	zs->zs_fd = fd;
787 	return zs;
788 }
789