1 /*
  2  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  3  *
  4  * This code is free software; you can redistribute it and/or modify it
  5  * under the terms of the GNU General Public License version 2 only, as
  6  * published by the Free Software Foundation.  Oracle designates this
  7  * particular file as subject to the "Classpath" exception as provided
  8  * by Oracle in the LICENSE file that accompanied this code.
  9  *
 10  * This code is distributed in the hope that it will be useful, but WITHOUT
 11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
 13  * version 2 for more details (a copy is included in the LICENSE file that
 14  * accompanied this code).
 15  *
 16  * You should have received a copy of the GNU General Public License version
 17  * 2 along with this work; if not, write to the Free Software Foundation,
 18  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
 19  *
 20  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 21  * or visit www.oracle.com if you need additional information or have any
 22  * questions.
 23  */
 24 
 25 /* inftrees.c -- generate Huffman trees for efficient decoding
 26  * Copyright (C) 1995-2024 Mark Adler
 27  * For conditions of distribution and use, see copyright notice in zlib.h
 28  */
 29 
 30 #include "zutil.h"
 31 #include "inftrees.h"
 32 
 33 #define MAXBITS 15
 34 
 35 const char inflate_copyright[] =
 36    " inflate 1.3.1 Copyright 1995-2024 Mark Adler ";
 37 /*
 38   If you use the zlib library in a product, an acknowledgment is welcome
 39   in the documentation of your product. If for some reason you cannot
 40   include such an acknowledgment, I would appreciate that you keep this
 41   copyright string in the executable of your product.
 42  */
 43 
 44 /*
 45    Build a set of tables to decode the provided canonical Huffman code.
 46    The code lengths are lens[0..codes-1].  The result starts at *table,
 47    whose indices are 0..2^bits-1.  work is a writable array of at least
 48    lens shorts, which is used as a work area.  type is the type of code
 49    to be generated, CODES, LENS, or DISTS.  On return, zero is success,
 50    -1 is an invalid code, and +1 means that ENOUGH isn't enough.  table
 51    on return points to the next available entry's address.  bits is the
 52    requested root table index bits, and on return it is the actual root
 53    table index bits.  It will differ if the request is greater than the
 54    longest code or if it is less than the shortest code.
 55  */
 56 int ZLIB_INTERNAL inflate_table(codetype type, unsigned short FAR *lens,
 57                                 unsigned codes, code FAR * FAR *table,
 58                                 unsigned FAR *bits, unsigned short FAR *work) {
 59     unsigned len;               /* a code's length in bits */
 60     unsigned sym;               /* index of code symbols */
 61     unsigned min, max;          /* minimum and maximum code lengths */
 62     unsigned root;              /* number of index bits for root table */
 63     unsigned curr;              /* number of index bits for current table */
 64     unsigned drop;              /* code bits to drop for sub-table */
 65     int left;                   /* number of prefix codes available */
 66     unsigned used;              /* code entries in table used */
 67     unsigned huff;              /* Huffman code */
 68     unsigned incr;              /* for incrementing code, index */
 69     unsigned fill;              /* index for replicating entries */
 70     unsigned low;               /* low bits for current root entry */
 71     unsigned mask;              /* mask for low root bits */
 72     code here;                  /* table entry for duplication */
 73     code FAR *next;             /* next available space in table */
 74     const unsigned short FAR *base;     /* base value table to use */
 75     const unsigned short FAR *extra;    /* extra bits table to use */
 76     unsigned match;             /* use base and extra for symbol >= match */
 77     unsigned short count[MAXBITS+1];    /* number of codes of each length */
 78     unsigned short offs[MAXBITS+1];     /* offsets in table for each length */
 79     static const unsigned short lbase[31] = { /* Length codes 257..285 base */
 80         3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31,
 81         35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0};
 82     static const unsigned short lext[31] = { /* Length codes 257..285 extra */
 83         16, 16, 16, 16, 16, 16, 16, 16, 17, 17, 17, 17, 18, 18, 18, 18,
 84         19, 19, 19, 19, 20, 20, 20, 20, 21, 21, 21, 21, 16, 203, 77};
 85     static const unsigned short dbase[32] = { /* Distance codes 0..29 base */
 86         1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
 87         257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
 88         8193, 12289, 16385, 24577, 0, 0};
 89     static const unsigned short dext[32] = { /* Distance codes 0..29 extra */
 90         16, 16, 16, 16, 17, 17, 18, 18, 19, 19, 20, 20, 21, 21, 22, 22,
 91         23, 23, 24, 24, 25, 25, 26, 26, 27, 27,
 92         28, 28, 29, 29, 64, 64};
 93 
 94     /*
 95        Process a set of code lengths to create a canonical Huffman code.  The
 96        code lengths are lens[0..codes-1].  Each length corresponds to the
 97        symbols 0..codes-1.  The Huffman code is generated by first sorting the
 98        symbols by length from short to long, and retaining the symbol order
 99        for codes with equal lengths.  Then the code starts with all zero bits
100        for the first code of the shortest length, and the codes are integer
101        increments for the same length, and zeros are appended as the length
102        increases.  For the deflate format, these bits are stored backwards
103        from their more natural integer increment ordering, and so when the
104        decoding tables are built in the large loop below, the integer codes
105        are incremented backwards.
106 
107        This routine assumes, but does not check, that all of the entries in
108        lens[] are in the range 0..MAXBITS.  The caller must assure this.
109        1..MAXBITS is interpreted as that code length.  zero means that that
110        symbol does not occur in this code.
111 
112        The codes are sorted by computing a count of codes for each length,
113        creating from that a table of starting indices for each length in the
114        sorted table, and then entering the symbols in order in the sorted
115        table.  The sorted table is work[], with that space being provided by
116        the caller.
117 
118        The length counts are used for other purposes as well, i.e. finding
119        the minimum and maximum length codes, determining if there are any
120        codes at all, checking for a valid set of lengths, and looking ahead
121        at length counts to determine sub-table sizes when building the
122        decoding tables.
123      */
124 
125     /* accumulate lengths for codes (assumes lens[] all in 0..MAXBITS) */
126     for (len = 0; len <= MAXBITS; len++)
127         count[len] = 0;
128     for (sym = 0; sym < codes; sym++)
129         count[lens[sym]]++;
130 
131     /* bound code lengths, force root to be within code lengths */
132     root = *bits;
133     for (max = MAXBITS; max >= 1; max--)
134         if (count[max] != 0) break;
135     if (root > max) root = max;
136     if (max == 0) {                     /* no symbols to code at all */
137         here.op = (unsigned char)64;    /* invalid code marker */
138         here.bits = (unsigned char)1;
139         here.val = (unsigned short)0;
140         *(*table)++ = here;             /* make a table to force an error */
141         *(*table)++ = here;
142         *bits = 1;
143         return 0;     /* no symbols, but wait for decoding to report error */
144     }
145     for (min = 1; min < max; min++)
146         if (count[min] != 0) break;
147     if (root < min) root = min;
148 
149     /* check for an over-subscribed or incomplete set of lengths */
150     left = 1;
151     for (len = 1; len <= MAXBITS; len++) {
152         left <<= 1;
153         left -= count[len];
154         if (left < 0) return -1;        /* over-subscribed */
155     }
156     if (left > 0 && (type == CODES || max != 1))
157         return -1;                      /* incomplete set */
158 
159     /* generate offsets into symbol table for each length for sorting */
160     offs[1] = 0;
161     for (len = 1; len < MAXBITS; len++)
162         offs[len + 1] = offs[len] + count[len];
163 
164     /* sort symbols by length, by symbol order within each length */
165     for (sym = 0; sym < codes; sym++)
166         if (lens[sym] != 0) work[offs[lens[sym]]++] = (unsigned short)sym;
167 
168     /*
169        Create and fill in decoding tables.  In this loop, the table being
170        filled is at next and has curr index bits.  The code being used is huff
171        with length len.  That code is converted to an index by dropping drop
172        bits off of the bottom.  For codes where len is less than drop + curr,
173        those top drop + curr - len bits are incremented through all values to
174        fill the table with replicated entries.
175 
176        root is the number of index bits for the root table.  When len exceeds
177        root, sub-tables are created pointed to by the root entry with an index
178        of the low root bits of huff.  This is saved in low to check for when a
179        new sub-table should be started.  drop is zero when the root table is
180        being filled, and drop is root when sub-tables are being filled.
181 
182        When a new sub-table is needed, it is necessary to look ahead in the
183        code lengths to determine what size sub-table is needed.  The length
184        counts are used for this, and so count[] is decremented as codes are
185        entered in the tables.
186 
187        used keeps track of how many table entries have been allocated from the
188        provided *table space.  It is checked for LENS and DIST tables against
189        the constants ENOUGH_LENS and ENOUGH_DISTS to guard against changes in
190        the initial root table size constants.  See the comments in inftrees.h
191        for more information.
192 
193        sym increments through all symbols, and the loop terminates when
194        all codes of length max, i.e. all codes, have been processed.  This
195        routine permits incomplete codes, so another loop after this one fills
196        in the rest of the decoding tables with invalid code markers.
197      */
198 
199     /* set up for code type */
200     switch (type) {
201     case CODES:
202         base = extra = work;    /* dummy value--not used */
203         match = 20;
204         break;
205     case LENS:
206         base = lbase;
207         extra = lext;
208         match = 257;
209         break;
210     default:    /* DISTS */
211         base = dbase;
212         extra = dext;
213         match = 0;
214     }
215 
216     /* initialize state for loop */
217     huff = 0;                   /* starting code */
218     sym = 0;                    /* starting code symbol */
219     len = min;                  /* starting code length */
220     next = *table;              /* current table to fill in */
221     curr = root;                /* current table index bits */
222     drop = 0;                   /* current bits to drop from code for index */
223     low = (unsigned)(-1);       /* trigger new sub-table when len > root */
224     used = 1U << root;          /* use root table entries */
225     mask = used - 1;            /* mask for comparing low */
226 
227     /* check available table space */
228     if ((type == LENS && used > ENOUGH_LENS) ||
229         (type == DISTS && used > ENOUGH_DISTS))
230         return 1;
231 
232     /* process all codes and make table entries */
233     for (;;) {
234         /* create table entry */
235         here.bits = (unsigned char)(len - drop);
236         if (work[sym] + 1U < match) {
237             here.op = (unsigned char)0;
238             here.val = work[sym];
239         }
240         else if (work[sym] >= match) {
241             here.op = (unsigned char)(extra[work[sym] - match]);
242             here.val = base[work[sym] - match];
243         }
244         else {
245             here.op = (unsigned char)(32 + 64);         /* end of block */
246             here.val = 0;
247         }
248 
249         /* replicate for those indices with low len bits equal to huff */
250         incr = 1U << (len - drop);
251         fill = 1U << curr;
252         min = fill;                 /* save offset to next table */
253         do {
254             fill -= incr;
255             next[(huff >> drop) + fill] = here;
256         } while (fill != 0);
257 
258         /* backwards increment the len-bit code huff */
259         incr = 1U << (len - 1);
260         while (huff & incr)
261             incr >>= 1;
262         if (incr != 0) {
263             huff &= incr - 1;
264             huff += incr;
265         }
266         else
267             huff = 0;
268 
269         /* go to next symbol, update count, len */
270         sym++;
271         if (--(count[len]) == 0) {
272             if (len == max) break;
273             len = lens[work[sym]];
274         }
275 
276         /* create new sub-table if needed */
277         if (len > root && (huff & mask) != low) {
278             /* if first time, transition to sub-tables */
279             if (drop == 0)
280                 drop = root;
281 
282             /* increment past last table */
283             next += min;            /* here min is 1 << curr */
284 
285             /* determine length of next table */
286             curr = len - drop;
287             left = (int)(1 << curr);
288             while (curr + drop < max) {
289                 left -= count[curr + drop];
290                 if (left <= 0) break;
291                 curr++;
292                 left <<= 1;
293             }
294 
295             /* check for enough space */
296             used += 1U << curr;
297             if ((type == LENS && used > ENOUGH_LENS) ||
298                 (type == DISTS && used > ENOUGH_DISTS))
299                 return 1;
300 
301             /* point entry in root table to sub-table */
302             low = huff & mask;
303             (*table)[low].op = (unsigned char)curr;
304             (*table)[low].bits = (unsigned char)root;
305             (*table)[low].val = (unsigned short)(next - *table);
306         }
307     }
308 
309     /* fill in remaining table entry if code is incomplete (guaranteed to have
310        at most one remaining entry, since if the code is incomplete, the
311        maximum code length that was allowed to get this far is one bit) */
312     if (huff != 0) {
313         here.op = (unsigned char)64;            /* invalid code marker */
314         here.bits = (unsigned char)(len - drop);
315         here.val = (unsigned short)0;
316         next[huff] = here;
317     }
318 
319     /* set return parameters */
320     *table += used;
321     *bits = root;
322     return 0;
323 }