1 /*
2  * Copyright 2019-2023 The OpenSSL Project Authors. All Rights Reserved.
3  *
4  * Licensed under the Apache License 2.0 (the "License").  You may not use
5  * this file except in compliance with the License.  You can obtain a copy
6  * in the file LICENSE in the source distribution or at
7  * https://www.openssl.org/source/license.html
8  */
9 
10 #include <stddef.h>
11 #include <stdio.h>
12 #include <openssl/e_os2.h>
13 #include "crypto/punycode.h"
14 #include "internal/common.h" /* for HAS_PREFIX */
15 #include "internal/packet.h" /* for WPACKET */
16 
17 static const unsigned int base = 36;
18 static const unsigned int tmin = 1;
19 static const unsigned int tmax = 26;
20 static const unsigned int skew = 38;
21 static const unsigned int damp = 700;
22 static const unsigned int initial_bias = 72;
23 static const unsigned int initial_n = 0x80;
24 static const unsigned int maxint = 0xFFFFFFFF;
25 static const char delimiter = '-';
26 
27 #define LABEL_BUF_SIZE 512
28 
29 /*-
30  * Pseudocode:
31  *
32  * function adapt(delta,numpoints,firsttime):
33  *  if firsttime then let delta = delta div damp
34  *  else let delta = delta div 2
35  *  let delta = delta + (delta div numpoints)
36  *  let k = 0
37  *  while delta > ((base - tmin) * tmax) div 2 do begin
38  *    let delta = delta div (base - tmin)
39  *    let k = k + base
40  *  end
41  *  return k + (((base - tmin + 1) * delta) div (delta + skew))
42  */
43 
adapt(unsigned int delta,unsigned int numpoints,unsigned int firsttime)44 static int adapt(unsigned int delta, unsigned int numpoints,
45                  unsigned int firsttime)
46 {
47     unsigned int k = 0;
48 
49     delta = (firsttime) ? delta / damp : delta / 2;
50     delta = delta + delta / numpoints;
51 
52     while (delta > ((base - tmin) * tmax) / 2) {
53         delta = delta / (base - tmin);
54         k = k + base;
55     }
56 
57     return k + (((base - tmin + 1) * delta) / (delta + skew));
58 }
59 
is_basic(unsigned int a)60 static ossl_inline int is_basic(unsigned int a)
61 {
62     return (a < 0x80) ? 1 : 0;
63 }
64 
65 /*-
66  * code points    digit-values
67  * ------------   ----------------------
68  * 41..5A (A-Z) =  0 to 25, respectively
69  * 61..7A (a-z) =  0 to 25, respectively
70  * 30..39 (0-9) = 26 to 35, respectively
71  */
digit_decoded(const unsigned char a)72 static ossl_inline int digit_decoded(const unsigned char a)
73 {
74     if (a >= 0x41 && a <= 0x5A)
75         return a - 0x41;
76 
77     if (a >= 0x61 && a <= 0x7A)
78         return a - 0x61;
79 
80     if (a >= 0x30 && a <= 0x39)
81         return a - 0x30 + 26;
82 
83     return -1;
84 }
85 
86 /*-
87  * Pseudocode:
88  *
89  * function ossl_punycode_decode
90  *  let n = initial_n
91  *  let i = 0
92  *  let bias = initial_bias
93  *  let output = an empty string indexed from 0
94  *  consume all code points before the last delimiter (if there is one)
95  *    and copy them to output, fail on any non-basic code point
96  *  if more than zero code points were consumed then consume one more
97  *    (which will be the last delimiter)
98  *  while the input is not exhausted do begin
99  *    let oldi = i
100  *    let w = 1
101  *    for k = base to infinity in steps of base do begin
102  *      consume a code point, or fail if there was none to consume
103  *      let digit = the code point's digit-value, fail if it has none
104  *      let i = i + digit * w, fail on overflow
105  *      let t = tmin if k <= bias {+ tmin}, or
106  *              tmax if k >= bias + tmax, or k - bias otherwise
107  *      if digit < t then break
108  *      let w = w * (base - t), fail on overflow
109  *    end
110  *    let bias = adapt(i - oldi, length(output) + 1, test oldi is 0?)
111  *    let n = n + i div (length(output) + 1), fail on overflow
112  *    let i = i mod (length(output) + 1)
113  *    {if n is a basic code point then fail}
114  *    insert n into output at position i
115  *    increment i
116  *  end
117  */
118 
ossl_punycode_decode(const char * pEncoded,const size_t enc_len,unsigned int * pDecoded,unsigned int * pout_length)119 int ossl_punycode_decode(const char *pEncoded, const size_t enc_len,
120                          unsigned int *pDecoded, unsigned int *pout_length)
121 {
122     unsigned int n = initial_n;
123     unsigned int i = 0;
124     unsigned int bias = initial_bias;
125     unsigned int processed_in = 0;
126     unsigned int written_out = 0;
127     unsigned int max_out = *pout_length;
128     unsigned int basic_count = 0;
129     unsigned int loop;
130 
131     if (enc_len >= UINT_MAX)
132         return 0;
133     for (loop = 0; loop < (unsigned int)enc_len; loop++) {
134         if (pEncoded[loop] == delimiter)
135             basic_count = loop;
136     }
137 
138     if (basic_count > 0) {
139         if (basic_count > max_out)
140             return 0;
141 
142         for (loop = 0; loop < basic_count; loop++) {
143             if (is_basic(pEncoded[loop]) == 0)
144                 return 0;
145 
146             pDecoded[loop] = pEncoded[loop];
147             written_out++;
148         }
149         processed_in = basic_count + 1;
150     }
151 
152     for (loop = processed_in; loop < (unsigned int)enc_len;) {
153         unsigned int oldi = i;
154         unsigned int w = 1;
155         unsigned int k, t;
156         int digit;
157 
158         for (k = base;; k += base) {
159             if (loop >= enc_len)
160                 return 0;
161 
162             digit = digit_decoded(pEncoded[loop]);
163             loop++;
164 
165             if (digit < 0)
166                 return 0;
167             if ((unsigned int)digit > (maxint - i) / w)
168                 return 0;
169 
170             i = i + digit * w;
171             t = (k <= bias) ? tmin : (k >= bias + tmax) ? tmax : k - bias;
172 
173             if ((unsigned int)digit < t)
174                 break;
175 
176             if (w > maxint / (base - t))
177                 return 0;
178             w = w * (base - t);
179         }
180 
181         bias = adapt(i - oldi, written_out + 1, (oldi == 0));
182         if (i / (written_out + 1) > maxint - n)
183             return 0;
184         n = n + i / (written_out + 1);
185         i %= (written_out + 1);
186 
187         if (written_out >= max_out)
188             return 0;
189 
190         memmove(pDecoded + i + 1, pDecoded + i,
191                 (written_out - i) * sizeof(*pDecoded));
192         pDecoded[i] = n;
193         i++;
194         written_out++;
195     }
196 
197     *pout_length = written_out;
198     return 1;
199 }
200 
201 /*
202  * Encode a code point using UTF-8
203  * return number of bytes on success, 0 on failure
204  * (also produces U+FFFD, which uses 3 bytes on failure)
205  */
codepoint2utf8(unsigned char * out,unsigned long utf)206 static int codepoint2utf8(unsigned char *out, unsigned long utf)
207 {
208     if (utf <= 0x7F) {
209         /* Plain ASCII */
210         out[0] = (unsigned char)utf;
211         out[1] = 0;
212         return 1;
213     } else if (utf <= 0x07FF) {
214         /* 2-byte unicode */
215         out[0] = (unsigned char)(((utf >> 6) & 0x1F) | 0xC0);
216         out[1] = (unsigned char)(((utf >> 0) & 0x3F) | 0x80);
217         out[2] = 0;
218         return 2;
219     } else if (utf <= 0xFFFF) {
220         /* 3-byte unicode */
221         out[0] = (unsigned char)(((utf >> 12) & 0x0F) | 0xE0);
222         out[1] = (unsigned char)(((utf >> 6) & 0x3F) | 0x80);
223         out[2] = (unsigned char)(((utf >> 0) & 0x3F) | 0x80);
224         out[3] = 0;
225         return 3;
226     } else if (utf <= 0x10FFFF) {
227         /* 4-byte unicode */
228         out[0] = (unsigned char)(((utf >> 18) & 0x07) | 0xF0);
229         out[1] = (unsigned char)(((utf >> 12) & 0x3F) | 0x80);
230         out[2] = (unsigned char)(((utf >> 6) & 0x3F) | 0x80);
231         out[3] = (unsigned char)(((utf >> 0) & 0x3F) | 0x80);
232         out[4] = 0;
233         return 4;
234     } else {
235         /* error - use replacement character */
236         out[0] = (unsigned char)0xEF;
237         out[1] = (unsigned char)0xBF;
238         out[2] = (unsigned char)0xBD;
239         out[3] = 0;
240         return 0;
241     }
242 }
243 
244 /*-
245  * Return values:
246  * 1 - ok
247  * 0 - ok but buf was too short
248  * -1 - bad string passed or other error
249  */
250 
ossl_a2ulabel(const char * in,char * out,size_t outlen)251 int ossl_a2ulabel(const char *in, char *out, size_t outlen)
252 {
253     /*-
254      * Domain name has some parts consisting of ASCII chars joined with dot.
255      * If a part is shorter than 5 chars, it becomes U-label as is.
256      * If it does not start with xn--,    it becomes U-label as is.
257      * Otherwise we try to decode it.
258      */
259     const char *inptr = in;
260     int result = 1;
261     unsigned int i;
262     unsigned int buf[LABEL_BUF_SIZE];      /* It's a hostname */
263     WPACKET pkt;
264 
265     /* Internal API, so should not fail */
266     if (!ossl_assert(out != NULL))
267         return -1;
268 
269     if (!WPACKET_init_static_len(&pkt, (unsigned char *)out, outlen, 0))
270         return -1;
271 
272     while (1) {
273         char *tmpptr = strchr(inptr, '.');
274         size_t delta = tmpptr != NULL ? (size_t)(tmpptr - inptr) : strlen(inptr);
275 
276         if (!HAS_PREFIX(inptr, "xn--")) {
277             if (!WPACKET_memcpy(&pkt, inptr, delta))
278                 result = 0;
279         } else {
280             unsigned int bufsize = LABEL_BUF_SIZE;
281 
282             if (ossl_punycode_decode(inptr + 4, delta - 4, buf, &bufsize) <= 0) {
283                 result = -1;
284                 goto end;
285             }
286 
287             for (i = 0; i < bufsize; i++) {
288                 unsigned char seed[6];
289                 size_t utfsize = codepoint2utf8(seed, buf[i]);
290 
291                 if (utfsize == 0) {
292                     result = -1;
293                     goto end;
294                 }
295 
296                 if (!WPACKET_memcpy(&pkt, seed, utfsize))
297                     result = 0;
298             }
299         }
300 
301         if (tmpptr == NULL)
302             break;
303 
304         if (!WPACKET_put_bytes_u8(&pkt, '.'))
305             result = 0;
306 
307         inptr = tmpptr + 1;
308     }
309 
310     if (!WPACKET_put_bytes_u8(&pkt, '\0'))
311         result = 0;
312  end:
313     WPACKET_cleanup(&pkt);
314     return result;
315 }
316