xref: /openbsd/usr.bin/compress/zopen.c (revision 5af055cd)
1 /*	$OpenBSD: zopen.c,v 1.20 2015/02/01 11:50:23 tobias 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/stat.h>
62 
63 #include <ctype.h>
64 #include <errno.h>
65 #include <signal.h>
66 #include <stdio.h>
67 #include <stdlib.h>
68 #include <string.h>
69 #include <unistd.h>
70 #include <fcntl.h>
71 #include "compress.h"
72 
73 #define MINIMUM(a, b)	(((a) < (b)) ? (a) : (b))
74 
75 #define	BITS		16		/* Default bits. */
76 #define	HSIZE		69001		/* 95% occupancy */
77 #define	ZBUFSIZ		8192		/* I/O buffer size */
78 
79 /* A code_int must be able to hold 2**BITS values of type int, and also -1. */
80 typedef long code_int;
81 typedef long count_int;
82 
83 static const u_char z_magic[] =
84 	{'\037', '\235'};		/* 1F 9D */
85 
86 #define	BIT_MASK	0x1f		/* Defines for third byte of header. */
87 #define	BLOCK_MASK	0x80
88 
89 /*
90  * Masks 0x40 and 0x20 are free.  I think 0x20 should mean that there is
91  * a fourth header byte (for expansion).
92  */
93 #define	INIT_BITS 9			/* Initial number of bits/code. */
94 
95 #define	MAXCODE(n_bits)	((1 << (n_bits)) - 1)
96 
97 struct s_zstate {
98 	int zs_fd;			/* File stream for I/O */
99 	char zs_mode;			/* r or w */
100 	enum {
101 		S_START, S_MAGIC, S_MIDDLE, S_EOF
102 	} zs_state;			/* State of computation */
103 	int zs_n_bits;			/* Number of bits/code. */
104 	int zs_maxbits;			/* User settable max # bits/code. */
105 	code_int zs_maxcode;		/* Maximum code, given n_bits. */
106 	code_int zs_maxmaxcode;		/* Should NEVER generate this code. */
107 	count_int zs_htab[HSIZE];
108 	u_short zs_codetab[HSIZE];
109 	code_int zs_hsize;		/* For dynamic table sizing. */
110 	code_int zs_free_ent;		/* First unused entry. */
111 	/*
112 	 * Block compression parameters -- after all codes are used up,
113 	 * and compression rate changes, start over.
114 	 */
115 	int zs_block_compress;
116 	int zs_clear_flg;
117 	long zs_ratio;
118 	count_int zs_checkpoint;
119 	long zs_in_count;		/* Length of input. */
120 	long zs_bytes_out;		/* Length of output. */
121 	long zs_out_count;		/* # of codes output (for debugging).*/
122 	u_char zs_buf[ZBUFSIZ];		/* I/O buffer */
123 	u_char *zs_bp;			/* Current I/O window in the zs_buf */
124 	int zs_offset;			/* Number of bits in the zs_buf */
125 	union {
126 		struct {
127 			long zs_fcode;
128 			code_int zs_ent;
129 			code_int zs_hsize_reg;
130 			int zs_hshift;
131 		} w;			/* Write parameters */
132 		struct {
133 			u_char *zs_stackp, *zs_ebp;
134 			int zs_finchar;
135 			code_int zs_code, zs_oldcode, zs_incode;
136 			int zs_size;
137 		} r;			/* Read parameters */
138 	} u;
139 };
140 
141 /* Definitions to retain old variable names */
142 #define zs_fcode	u.w.zs_fcode
143 #define zs_ent		u.w.zs_ent
144 #define zs_hsize_reg	u.w.zs_hsize_reg
145 #define zs_hshift	u.w.zs_hshift
146 #define zs_stackp	u.r.zs_stackp
147 #define zs_finchar	u.r.zs_finchar
148 #define zs_code		u.r.zs_code
149 #define zs_oldcode	u.r.zs_oldcode
150 #define zs_incode	u.r.zs_incode
151 #define zs_size		u.r.zs_size
152 #define zs_ebp		u.r.zs_ebp
153 
154 /*
155  * To save much memory, we overlay the table used by compress() with those
156  * used by decompress().  The tab_prefix table is the same size and type as
157  * the codetab.  The tab_suffix table needs 2**BITS characters.  We get this
158  * from the beginning of htab.  The output stack uses the rest of htab, and
159  * contains characters.  There is plenty of room for any possible stack
160  * (stack used to be 8000 characters).
161  */
162 
163 #define	htabof(i)	zs->zs_htab[i]
164 #define	codetabof(i)	zs->zs_codetab[i]
165 
166 #define	tab_prefixof(i)	codetabof(i)
167 #define	tab_suffixof(i)	((u_char *)(zs->zs_htab))[i]
168 #define	de_stack	((u_char *)&tab_suffixof(1 << BITS))
169 
170 #define	CHECK_GAP 10000		/* Ratio check interval. */
171 
172 /*
173  * the next two codes should not be changed lightly, as they must not
174  * lie within the contiguous general code space.
175  */
176 #define	FIRST	257		/* First free entry. */
177 #define	CLEAR	256		/* Table clear output code. */
178 
179 static int	cl_block(struct s_zstate *);
180 static void	cl_hash(struct s_zstate *, count_int);
181 static code_int	getcode(struct s_zstate *);
182 static int	output(struct s_zstate *, code_int);
183 
184 /*-
185  * Algorithm from "A Technique for High Performance Data Compression",
186  * Terry A. Welch, IEEE Computer Vol 17, No 6 (June 1984), pp 8-19.
187  *
188  * Algorithm:
189  *	Modified Lempel-Ziv method (LZW).  Basically finds common
190  * substrings and replaces them with a variable size code.  This is
191  * deterministic, and can be done on the fly.  Thus, the decompression
192  * procedure needs no input table, but tracks the way the table was built.
193  */
194 
195 /*-
196  * compress write
197  *
198  * Algorithm:  use open addressing double hashing (no chaining) on the
199  * prefix code / next character combination.  We do a variant of Knuth's
200  * algorithm D (vol. 3, sec. 6.4) along with G. Knott's relatively-prime
201  * secondary probe.  Here, the modular division first probe is gives way
202  * to a faster exclusive-or manipulation.  Also do block compression with
203  * an adaptive reset, whereby the code table is cleared when the compression
204  * ratio decreases, but after the table fills.  The variable-length output
205  * codes are re-sized at this point, and a special CLEAR code is generated
206  * for the decompressor.  Late addition:  construct the table according to
207  * file size for noticeable speed improvement on small files.  Please direct
208  * questions about this implementation to ames!jaw.
209  */
210 int
211 zwrite(void *cookie, const char *wbp, int num)
212 {
213 	code_int i;
214 	int c, disp;
215 	struct s_zstate *zs;
216 	const u_char *bp;
217 	u_char tmp;
218 	int count;
219 
220 	zs = cookie;
221 	count = num;
222 	bp = (u_char *)wbp;
223 	switch (zs->zs_state) {
224 	case S_MAGIC:
225 		return -1;
226 	case S_EOF:
227 		return 0;
228 	case S_START:
229 		zs->zs_state = S_MIDDLE;
230 
231 		zs->zs_maxmaxcode = 1L << zs->zs_maxbits;
232 		if (write(zs->zs_fd, z_magic, sizeof(z_magic)) !=
233 		    sizeof(z_magic))
234 			return (-1);
235 		tmp = (u_char)(zs->zs_maxbits | zs->zs_block_compress);
236 		if (write(zs->zs_fd, &tmp, sizeof(tmp)) != sizeof(tmp))
237 			return (-1);
238 
239 		zs->zs_bp = zs->zs_buf;
240 		zs->zs_offset = 0;
241 		zs->zs_bytes_out = 3;	/* Includes 3-byte header mojo. */
242 		zs->zs_out_count = 0;
243 		zs->zs_clear_flg = 0;
244 		zs->zs_ratio = 0;
245 		zs->zs_in_count = 1;
246 		zs->zs_checkpoint = CHECK_GAP;
247 		zs->zs_maxcode = MAXCODE(zs->zs_n_bits = INIT_BITS);
248 		zs->zs_free_ent = ((zs->zs_block_compress) ? FIRST : 256);
249 
250 		zs->zs_ent = *bp++;
251 		--count;
252 
253 		zs->zs_hshift = 0;
254 		for (zs->zs_fcode = (long)zs->zs_hsize; zs->zs_fcode < 65536L;
255 		    zs->zs_fcode *= 2L)
256 			zs->zs_hshift++;
257 		/* Set hash code range bound. */
258 		zs->zs_hshift = 8 - zs->zs_hshift;
259 
260 		zs->zs_hsize_reg = zs->zs_hsize;
261 		/* Clear hash table. */
262 		cl_hash(zs, (count_int)zs->zs_hsize_reg);
263 
264 	case S_MIDDLE:
265 		for (i = 0; count-- > 0;) {
266 			c = *bp++;
267 			zs->zs_in_count++;
268 			zs->zs_fcode = (long)(((long)c << zs->zs_maxbits) +
269 			    zs->zs_ent);
270 			/* Xor hashing. */
271 			i = ((c << zs->zs_hshift) ^ zs->zs_ent);
272 
273 			if (htabof(i) == zs->zs_fcode) {
274 				zs->zs_ent = codetabof(i);
275 				continue;
276 			} else if ((long)htabof(i) < 0)	/* Empty slot. */
277 				goto nomatch;
278 			/* Secondary hash (after G. Knott). */
279 			disp = zs->zs_hsize_reg - i;
280 			if (i == 0)
281 				disp = 1;
282 probe:			if ((i -= disp) < 0)
283 				i += zs->zs_hsize_reg;
284 
285 			if (htabof(i) == zs->zs_fcode) {
286 				zs->zs_ent = codetabof(i);
287 				continue;
288 			}
289 			if ((long)htabof(i) >= 0)
290 				goto probe;
291 nomatch:		if (output(zs, (code_int) zs->zs_ent) == -1)
292 				return (-1);
293 			zs->zs_out_count++;
294 			zs->zs_ent = c;
295 			if (zs->zs_free_ent < zs->zs_maxmaxcode) {
296 				/* code -> hashtable */
297 				codetabof(i) = zs->zs_free_ent++;
298 				htabof(i) = zs->zs_fcode;
299 			} else if ((count_int)zs->zs_in_count >=
300 			    zs->zs_checkpoint && zs->zs_block_compress) {
301 				if (cl_block(zs) == -1)
302 					return (-1);
303 			}
304 		}
305 	}
306 	return (num);
307 }
308 
309 int
310 z_close(void *cookie, struct z_info *info, const char *name, struct stat *sb)
311 {
312 	struct s_zstate *zs;
313 	int rval;
314 
315 	zs = cookie;
316 	if (zs->zs_mode == 'w') {		/* Put out the final code. */
317 		if (output(zs, (code_int) zs->zs_ent) == -1) {
318 			(void)close(zs->zs_fd);
319 			free(zs);
320 			return (-1);
321 		}
322 		zs->zs_out_count++;
323 		if (output(zs, (code_int) - 1) == -1) {
324 			(void)close(zs->zs_fd);
325 			free(zs);
326 			return (-1);
327 		}
328 	}
329 
330 	if (info != NULL) {
331 		info->mtime = 0;
332 		info->crc = (u_int32_t)-1;
333 		info->hlen = 0;
334 		info->total_in = (off_t)zs->zs_in_count;
335 		info->total_out = (off_t)zs->zs_bytes_out;
336 	}
337 
338 #ifndef SAVECORE
339 	setfile(name, zs->zs_fd, sb);
340 #endif
341 	rval = close(zs->zs_fd);
342 	free(zs);
343 	return (rval);
344 }
345 
346 static int
347 zclose(void *cookie)
348 {
349 	return z_close(cookie, NULL, NULL, NULL);
350 }
351 
352 /*-
353  * Output the given code.
354  * Inputs:
355  *	code:	A n_bits-bit integer.  If == -1, then EOF.  This assumes
356  *		that n_bits =< (long)wordsize - 1.
357  * Outputs:
358  *	Outputs code to the file.
359  * Assumptions:
360  *	Chars are 8 bits long.
361  * Algorithm:
362  *	Maintain a BITS character long buffer (so that 8 codes will
363  * fit in it exactly).  Use the VAX insv instruction to insert each
364  * code in turn.  When the buffer fills up empty it and start over.
365  */
366 
367 static const u_char lmask[9] =
368 	{0xff, 0xfe, 0xfc, 0xf8, 0xf0, 0xe0, 0xc0, 0x80, 0x00};
369 static const u_char rmask[9] =
370 	{0x00, 0x01, 0x03, 0x07, 0x0f, 0x1f, 0x3f, 0x7f, 0xff};
371 
372 static int
373 output(struct s_zstate *zs, code_int ocode)
374 {
375 	int bits;
376 
377 	if (ocode >= 0) {
378 		int r_off;
379 		u_char *bp;
380 
381 		/* Get to the first byte. */
382 		bp = zs->zs_bp + (zs->zs_offset >> 3);
383 		r_off = zs->zs_offset & 7;
384 		bits = zs->zs_n_bits;
385 
386 		/*
387 		 * Since ocode is always >= 8 bits, only need to mask the first
388 		 * hunk on the left.
389 		 */
390 		*bp = (*bp & rmask[r_off]) | ((ocode << r_off) & lmask[r_off]);
391 		bp++;
392 		bits -= (8 - r_off);
393 		ocode >>= 8 - r_off;
394 		/* Get any 8 bit parts in the middle (<=1 for up to 16 bits) */
395 		if (bits >= 8) {
396 			*bp++ = ocode;
397 			ocode >>= 8;
398 			bits -= 8;
399 		}
400 		/* Last bits. */
401 		if (bits)
402 			*bp = ocode;
403 		zs->zs_offset += zs->zs_n_bits;
404 		if (zs->zs_offset == (zs->zs_n_bits << 3)) {
405 			zs->zs_bp += zs->zs_n_bits;
406 			zs->zs_offset = 0;
407 		}
408 		/*
409 		 * If the next entry is going to be too big for the ocode size,
410 		 * then increase it, if possible.
411 		 */
412 		if (zs->zs_free_ent > zs->zs_maxcode ||
413 		    (zs->zs_clear_flg > 0)) {
414 			/*
415 			 * Write the whole buffer, because the input side won't
416 			 * discover the size increase until after it has read it
417 			 */
418 			if (zs->zs_offset > 0) {
419 				zs->zs_bp += zs->zs_n_bits;
420 				zs->zs_offset = 0;
421 			}
422 
423 			if (zs->zs_clear_flg) {
424 				zs->zs_maxcode =
425 					MAXCODE(zs->zs_n_bits = INIT_BITS);
426 				zs->zs_clear_flg = 0;
427 			} else {
428 				zs->zs_n_bits++;
429 				if (zs->zs_n_bits == zs->zs_maxbits)
430 					zs->zs_maxcode = zs->zs_maxmaxcode;
431 				else
432 					zs->zs_maxcode =
433 					    MAXCODE(zs->zs_n_bits);
434 			}
435 		}
436 
437 		if (zs->zs_bp + zs->zs_n_bits > &zs->zs_buf[ZBUFSIZ]) {
438 			bits = zs->zs_bp - zs->zs_buf;
439 			if (write(zs->zs_fd, zs->zs_buf, bits) != bits)
440 				return (-1);
441 			zs->zs_bytes_out += bits;
442 			if (zs->zs_offset > 0)
443 				fprintf (stderr, "zs_offset != 0\n");
444 			zs->zs_bp = zs->zs_buf;
445 		}
446 	} else {
447 		/* At EOF, write the rest of the buffer. */
448 		if (zs->zs_offset > 0)
449 			zs->zs_bp += (zs->zs_offset + 7) / 8;
450 		if (zs->zs_bp > zs->zs_buf) {
451 			bits = zs->zs_bp - zs->zs_buf;
452 			if (write(zs->zs_fd, zs->zs_buf, bits) != bits)
453 				return (-1);
454 			zs->zs_bytes_out += bits;
455 		}
456 		zs->zs_offset = 0;
457 		zs->zs_bp = zs->zs_buf;
458 	}
459 	return (0);
460 }
461 
462 /*
463  * Decompress read.  This routine adapts to the codes in the file building
464  * the "string" table on-the-fly; requiring no table to be stored in the
465  * compressed file.  The tables used herein are shared with those of the
466  * compress() routine.  See the definitions above.
467  */
468 int
469 zread(void *cookie, char *rbp, int num)
470 {
471 	u_int count;
472 	struct s_zstate *zs;
473 	u_char *bp, header[3];
474 
475 	if (num == 0)
476 		return (0);
477 
478 	zs = cookie;
479 	count = num;
480 	bp = (u_char *)rbp;
481 	switch (zs->zs_state) {
482 	case S_START:
483 		zs->zs_state = S_MIDDLE;
484 		zs->zs_bp = zs->zs_buf;
485 		header[0] = header[1] = header[2] = '\0';
486 		read(zs->zs_fd, header, sizeof(header));
487 		break;
488 	case S_MAGIC:
489 		zs->zs_state = S_MIDDLE;
490 		zs->zs_bp = zs->zs_buf;
491 		header[0] = z_magic[0];
492 		header[1] = z_magic[1];
493 		header[2] = '\0';
494 		read(zs->zs_fd, &header[2], 1);
495 		break;
496 	case S_MIDDLE:
497 		goto middle;
498 	case S_EOF:
499 		goto eof;
500 	}
501 
502 	/* Check the magic number */
503 	if (header[0] != z_magic[0] || header[1] != z_magic[1]) {
504 		errno = EFTYPE;
505 		return (-1);
506 	}
507 	zs->zs_maxbits = header[2];	/* Set -b from file. */
508 	zs->zs_in_count += sizeof(header);
509 	zs->zs_block_compress = zs->zs_maxbits & BLOCK_MASK;
510 	zs->zs_maxbits &= BIT_MASK;
511 	zs->zs_maxmaxcode = 1L << zs->zs_maxbits;
512 	if (zs->zs_maxbits > BITS) {
513 		errno = EFTYPE;
514 		return (-1);
515 	}
516 	/* As above, initialize the first 256 entries in the table. */
517 	zs->zs_maxcode = MAXCODE(zs->zs_n_bits = INIT_BITS);
518 	for (zs->zs_code = 255; zs->zs_code >= 0; zs->zs_code--) {
519 		tab_prefixof(zs->zs_code) = 0;
520 		tab_suffixof(zs->zs_code) = (u_char) zs->zs_code;
521 	}
522 	zs->zs_free_ent = zs->zs_block_compress ? FIRST : 256;
523 
524 	zs->zs_finchar = zs->zs_oldcode = getcode(zs);
525 	if (zs->zs_oldcode == -1)	/* EOF already? */
526 		return (0);	/* Get out of here */
527 
528 	/* First code must be 8 bits = char. */
529 	*bp++ = (u_char)zs->zs_finchar;
530 	count--;
531 	zs->zs_stackp = de_stack;
532 
533 	while ((zs->zs_code = getcode(zs)) > -1) {
534 
535 		if ((zs->zs_code == CLEAR) && zs->zs_block_compress) {
536 			for (zs->zs_code = 255; zs->zs_code >= 0;
537 			    zs->zs_code--)
538 				tab_prefixof(zs->zs_code) = 0;
539 			zs->zs_clear_flg = 1;
540 			zs->zs_free_ent = FIRST - 1;
541 			if ((zs->zs_code = getcode(zs)) == -1)	/* O, untimely death! */
542 				break;
543 		}
544 		zs->zs_incode = zs->zs_code;
545 
546 		/* Special case for KwKwK string. */
547 		if (zs->zs_code >= zs->zs_free_ent) {
548 			*zs->zs_stackp++ = zs->zs_finchar;
549 			zs->zs_code = zs->zs_oldcode;
550 		}
551 
552 		/* Generate output characters in reverse order. */
553 		while (zs->zs_code >= 256) {
554 			/*
555 			 * Bad input file may cause zs_stackp to overflow
556 			 * zs_htab; check here and abort decompression,
557 			 * that's better than dumping core.
558 			 */
559 			if (zs->zs_stackp >= (u_char *)&zs->zs_htab[HSIZE]) {
560 				errno = EINVAL;
561 				return (-1);
562 			}
563 			*zs->zs_stackp++ = tab_suffixof(zs->zs_code);
564 			zs->zs_code = tab_prefixof(zs->zs_code);
565 		}
566 		*zs->zs_stackp++ = zs->zs_finchar = tab_suffixof(zs->zs_code);
567 
568 		/* And put them out in forward order.  */
569 middle:		do {
570 			if (count-- == 0) {
571 				zs->zs_bytes_out += num;
572 				return (num);
573 			}
574 			*bp++ = *--zs->zs_stackp;
575 		} while (zs->zs_stackp > de_stack);
576 
577 		/* Generate the new entry. */
578 		if ((zs->zs_code = zs->zs_free_ent) < zs->zs_maxmaxcode) {
579 			tab_prefixof(zs->zs_code) = (u_short) zs->zs_oldcode;
580 			tab_suffixof(zs->zs_code) = zs->zs_finchar;
581 			zs->zs_free_ent = zs->zs_code + 1;
582 		}
583 
584 		/* Remember previous code. */
585 		zs->zs_oldcode = zs->zs_incode;
586 	}
587 	zs->zs_state = S_EOF;
588 	zs->zs_bytes_out += num - count;
589 eof:	return (num - count);
590 }
591 
592 /*-
593  * Read one code from the standard input.  If EOF, return -1.
594  * Inputs:
595  *	stdin
596  * Outputs:
597  *	code or -1 is returned.
598  */
599 static code_int
600 getcode(struct s_zstate *zs)
601 {
602 	code_int gcode;
603 	int r_off, bits;
604 	u_char *bp;
605 
606 	if (zs->zs_clear_flg > 0 || zs->zs_offset >= zs->zs_size ||
607 	    zs->zs_free_ent > zs->zs_maxcode) {
608 
609 		zs->zs_bp += zs->zs_n_bits;
610 		/*
611 		 * If the next entry will be too big for the current gcode
612 		 * size, then we must increase the size.  This implies reading
613 		 * a new buffer full, too.
614 		 */
615 		if (zs->zs_free_ent > zs->zs_maxcode) {
616 			zs->zs_n_bits++;
617 			if (zs->zs_n_bits == zs->zs_maxbits) {
618 				/* Won't get any bigger now. */
619 				zs->zs_maxcode = zs->zs_maxmaxcode;
620 			} else
621 				zs->zs_maxcode = MAXCODE(zs->zs_n_bits);
622 		}
623 		if (zs->zs_clear_flg > 0) {
624 			zs->zs_maxcode = MAXCODE(zs->zs_n_bits = INIT_BITS);
625 			zs->zs_clear_flg = 0;
626 		}
627 
628 		/* fill the buffer up to the neck */
629 		if (zs->zs_bp + zs->zs_n_bits > zs->zs_ebp) {
630 			for (bp = zs->zs_buf; zs->zs_bp < zs->zs_ebp;
631 				*bp++ = *zs->zs_bp++);
632 			if ((bits = read(zs->zs_fd, bp, ZBUFSIZ -
633 			    (bp - zs->zs_buf))) < 0)
634 				return -1;
635 			zs->zs_in_count += bits;
636 			zs->zs_bp = zs->zs_buf;
637 			zs->zs_ebp = bp + bits;
638 		}
639 		zs->zs_offset = 0;
640 		zs->zs_size = MINIMUM(zs->zs_n_bits, zs->zs_ebp - zs->zs_bp);
641 		if (zs->zs_size == 0)
642 			return -1;
643 		/* Round size down to integral number of codes. */
644 		zs->zs_size = (zs->zs_size << 3) - (zs->zs_n_bits - 1);
645 	}
646 
647 	bp = zs->zs_bp;
648 	r_off = zs->zs_offset;
649 	bits = zs->zs_n_bits;
650 
651 	/* Get to the first byte. */
652 	bp += (r_off >> 3);
653 	r_off &= 7;
654 
655 	/* Get first part (low order bits). */
656 	gcode = (*bp++ >> r_off);
657 	bits -= (8 - r_off);
658 	r_off = 8 - r_off;	/* Now, roffset into gcode word. */
659 
660 	/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
661 	if (bits >= 8) {
662 		gcode |= *bp++ << r_off;
663 		r_off += 8;
664 		bits -= 8;
665 	}
666 
667 	/* High order bits. */
668 	gcode |= (*bp & rmask[bits]) << r_off;
669 	zs->zs_offset += zs->zs_n_bits;
670 
671 	return (gcode);
672 }
673 
674 /* Table clear for block compress. */
675 static int
676 cl_block(struct s_zstate *zs)
677 {
678 	long rat;
679 
680 	zs->zs_checkpoint = zs->zs_in_count + CHECK_GAP;
681 
682 	if (zs->zs_in_count > 0x007fffff) {	/* Shift will overflow. */
683 		rat = zs->zs_bytes_out >> 8;
684 		if (rat == 0)		/* Don't divide by zero. */
685 			rat = 0x7fffffff;
686 		else
687 			rat = zs->zs_in_count / rat;
688 	} else {
689 		/* 8 fractional bits. */
690 		rat = (zs->zs_in_count << 8) / zs->zs_bytes_out;
691 	}
692 	if (rat > zs->zs_ratio)
693 		zs->zs_ratio = rat;
694 	else {
695 		zs->zs_ratio = 0;
696 		cl_hash(zs, (count_int) zs->zs_hsize);
697 		zs->zs_free_ent = FIRST;
698 		zs->zs_clear_flg = 1;
699 		if (output(zs, (code_int) CLEAR) == -1)
700 			return (-1);
701 	}
702 	return (0);
703 }
704 
705 /* Reset code table. */
706 static void
707 cl_hash(struct s_zstate *zs, count_int cl_hsize)
708 {
709 	count_int *htab_p;
710 	long i, m1;
711 
712 	m1 = -1;
713 	htab_p = zs->zs_htab + cl_hsize;
714 	i = cl_hsize - 16;
715 	do {			/* Might use Sys V memset(3) here. */
716 		*(htab_p - 16) = m1;
717 		*(htab_p - 15) = m1;
718 		*(htab_p - 14) = m1;
719 		*(htab_p - 13) = m1;
720 		*(htab_p - 12) = m1;
721 		*(htab_p - 11) = m1;
722 		*(htab_p - 10) = m1;
723 		*(htab_p - 9) = m1;
724 		*(htab_p - 8) = m1;
725 		*(htab_p - 7) = m1;
726 		*(htab_p - 6) = m1;
727 		*(htab_p - 5) = m1;
728 		*(htab_p - 4) = m1;
729 		*(htab_p - 3) = m1;
730 		*(htab_p - 2) = m1;
731 		*(htab_p - 1) = m1;
732 		htab_p -= 16;
733 	} while ((i -= 16) >= 0);
734 	for (i += 16; i > 0; i--)
735 		*--htab_p = m1;
736 }
737 
738 FILE *
739 zopen(const char *name, const char *mode, int bits)
740 {
741 	FILE *fp;
742 	int fd;
743 	void *cookie;
744 	if ((fd = open(name, (*mode=='r'? O_RDONLY:O_WRONLY|O_CREAT),
745 	    S_IRUSR|S_IWUSR|S_IRGRP|S_IROTH)) == -1)
746 		return NULL;
747 	if ((cookie = z_open(fd, mode, NULL, bits, 0, 0)) == NULL) {
748 		close(fd);
749 		return NULL;
750 	}
751 	if ((fp = funopen(cookie, (*mode == 'r'?zread:NULL),
752 	    (*mode == 'w'?zwrite:NULL), NULL, zclose)) == NULL) {
753 		close(fd);
754 		free(cookie);
755 		return NULL;
756 	}
757 	return fp;
758 }
759 
760 void *
761 z_open(int fd, const char *mode, char *name, int bits,
762     u_int32_t mtime, int gotmagic)
763 {
764 	struct s_zstate *zs;
765 
766 	if ((mode[0] != 'r' && mode[0] != 'w') || mode[1] != '\0' ||
767 	    bits < 0 || bits > BITS) {
768 		errno = EINVAL;
769 		return (NULL);
770 	}
771 
772 	if ((zs = calloc(1, sizeof(struct s_zstate))) == NULL)
773 		return (NULL);
774 
775 	/* User settable max # bits/code. */
776 	zs->zs_maxbits = bits ? bits : BITS;
777 	/* Should NEVER generate this code. */
778 	zs->zs_maxmaxcode = 1 << zs->zs_maxbits;
779 	zs->zs_hsize = HSIZE;		/* For dynamic table sizing. */
780 	zs->zs_free_ent = 0;		/* First unused entry. */
781 	zs->zs_block_compress = BLOCK_MASK;
782 	zs->zs_clear_flg = 0;
783 	zs->zs_ratio = 0;
784 	zs->zs_checkpoint = CHECK_GAP;
785 	zs->zs_in_count = 0;		/* Length of input. */
786 	zs->zs_out_count = 0;		/* # of codes output (for debugging).*/
787 	zs->zs_state = gotmagic ? S_MAGIC : S_START;
788 	zs->zs_offset = 0;
789 	zs->zs_size = 0;
790 	zs->zs_mode = mode[0];
791 	zs->zs_bp = zs->zs_ebp = zs->zs_buf;
792 
793 	zs->zs_fd = fd;
794 	return zs;
795 }
796