blob: c1325d8be947ce6a9ba13dc88be75854dde3403b
1 | /* vi: set sw=4 ts=4: */ |
2 | |
3 | /* RFC1035 domain compression routines (C) 2007 Gabriel Somlo <somlo at cmu.edu> |
4 | * |
5 | * Loosely based on the isc-dhcpd implementation by dhankins@isc.org |
6 | * |
7 | * Licensed under GPLv2 or later, see file LICENSE in this source tree. |
8 | */ |
9 | #ifdef DNS_COMPR_TESTING |
10 | # define FAST_FUNC /* nothing */ |
11 | # define xmalloc malloc |
12 | # include <stdlib.h> |
13 | # include <stdint.h> |
14 | # include <string.h> |
15 | # include <stdio.h> |
16 | #else |
17 | # include "common.h" |
18 | #endif |
19 | |
20 | #define NS_MAXDNAME 1025 /* max domain name length */ |
21 | #define NS_MAXCDNAME 255 /* max compressed domain name length */ |
22 | #define NS_MAXLABEL 63 /* max label length */ |
23 | #define NS_MAXDNSRCH 6 /* max domains in search path */ |
24 | #define NS_CMPRSFLGS 0xc0 /* name compression pointer flag */ |
25 | |
26 | |
27 | /* Expand a RFC1035-compressed list of domain names "cstr", of length "clen"; |
28 | * returns a newly allocated string containing the space-separated domains, |
29 | * prefixed with the contents of string pre, or NULL if an error occurs. |
30 | */ |
31 | char* FAST_FUNC dname_dec(const uint8_t *cstr, int clen, const char *pre) |
32 | { |
33 | char *ret = ret; /* for compiler */ |
34 | char *dst = NULL; |
35 | |
36 | /* We make two passes over the cstr string. First, we compute |
37 | * how long the resulting string would be. Then we allocate a |
38 | * new buffer of the required length, and fill it in with the |
39 | * expanded content. The advantage of this approach is not |
40 | * having to deal with requiring callers to supply their own |
41 | * buffer, then having to check if it's sufficiently large, etc. |
42 | */ |
43 | while (1) { |
44 | /* note: "return NULL" below are leak-safe since |
45 | * dst isn't yet allocated */ |
46 | const uint8_t *c; |
47 | unsigned crtpos, retpos, depth, len; |
48 | |
49 | crtpos = retpos = depth = len = 0; |
50 | while (crtpos < clen) { |
51 | c = cstr + crtpos; |
52 | |
53 | if ((*c & NS_CMPRSFLGS) == NS_CMPRSFLGS) { |
54 | /* pointer */ |
55 | if (crtpos + 2 > clen) /* no offset to jump to? abort */ |
56 | return NULL; |
57 | if (retpos == 0) /* toplevel? save return spot */ |
58 | retpos = crtpos + 2; |
59 | depth++; |
60 | crtpos = ((c[0] & 0x3f) << 8) | c[1]; /* jump */ |
61 | } else if (*c) { |
62 | /* label */ |
63 | if (crtpos + *c + 1 > clen) /* label too long? abort */ |
64 | return NULL; |
65 | if (dst) |
66 | memcpy(dst + len, c + 1, *c); |
67 | len += *c + 1; |
68 | crtpos += *c + 1; |
69 | if (dst) |
70 | dst[len - 1] = '.'; |
71 | } else { |
72 | /* NUL: end of current domain name */ |
73 | if (retpos == 0) { |
74 | /* toplevel? keep going */ |
75 | crtpos++; |
76 | } else { |
77 | /* return to toplevel saved spot */ |
78 | crtpos = retpos; |
79 | retpos = depth = 0; |
80 | } |
81 | if (dst) |
82 | dst[len - 1] = ' '; |
83 | } |
84 | |
85 | if (depth > NS_MAXDNSRCH /* too many jumps? abort, it's a loop */ |
86 | || len > NS_MAXDNAME * NS_MAXDNSRCH /* result too long? abort */ |
87 | ) { |
88 | return NULL; |
89 | } |
90 | } |
91 | |
92 | if (!len) /* expanded string has 0 length? abort */ |
93 | return NULL; |
94 | |
95 | if (!dst) { /* first pass? */ |
96 | /* allocate dst buffer and copy pre */ |
97 | unsigned plen = strlen(pre); |
98 | ret = dst = xmalloc(plen + len); |
99 | memcpy(dst, pre, plen); |
100 | dst += plen; |
101 | } else { |
102 | dst[len - 1] = '\0'; |
103 | break; |
104 | } |
105 | } |
106 | |
107 | return ret; |
108 | } |
109 | |
110 | /* Convert a domain name (src) from human-readable "foo.blah.com" format into |
111 | * RFC1035 encoding "\003foo\004blah\003com\000". Return allocated string, or |
112 | * NULL if an error occurs. |
113 | */ |
114 | static uint8_t *convert_dname(const char *src) |
115 | { |
116 | uint8_t c, *res, *lenptr, *dst; |
117 | int len; |
118 | |
119 | res = xmalloc(strlen(src) + 2); |
120 | dst = lenptr = res; |
121 | dst++; |
122 | |
123 | for (;;) { |
124 | c = (uint8_t)*src++; |
125 | if (c == '.' || c == '\0') { /* end of label */ |
126 | len = dst - lenptr - 1; |
127 | /* label too long, too short, or two '.'s in a row? abort */ |
128 | if (len > NS_MAXLABEL || len == 0 || (c == '.' && *src == '.')) { |
129 | free(res); |
130 | return NULL; |
131 | } |
132 | *lenptr = len; |
133 | if (c == '\0' || *src == '\0') /* "" or ".": end of src */ |
134 | break; |
135 | lenptr = dst++; |
136 | continue; |
137 | } |
138 | if (c >= 'A' && c <= 'Z') /* uppercase? convert to lower */ |
139 | c += ('a' - 'A'); |
140 | *dst++ = c; |
141 | } |
142 | |
143 | if (dst - res >= NS_MAXCDNAME) { /* dname too long? abort */ |
144 | free(res); |
145 | return NULL; |
146 | } |
147 | |
148 | *dst = 0; |
149 | return res; |
150 | } |
151 | |
152 | /* Returns the offset within cstr at which dname can be found, or -1 */ |
153 | static int find_offset(const uint8_t *cstr, int clen, const uint8_t *dname) |
154 | { |
155 | const uint8_t *c, *d; |
156 | int off; |
157 | |
158 | /* find all labels in cstr */ |
159 | off = 0; |
160 | while (off < clen) { |
161 | c = cstr + off; |
162 | |
163 | if ((*c & NS_CMPRSFLGS) == NS_CMPRSFLGS) { /* pointer, skip */ |
164 | off += 2; |
165 | continue; |
166 | } |
167 | if (*c) { /* label, try matching dname */ |
168 | d = dname; |
169 | while (1) { |
170 | unsigned len1 = *c + 1; |
171 | if (memcmp(c, d, len1) != 0) |
172 | break; |
173 | if (len1 == 1) /* at terminating NUL - match, return offset */ |
174 | return off; |
175 | d += len1; |
176 | c += len1; |
177 | if ((*c & NS_CMPRSFLGS) == NS_CMPRSFLGS) /* pointer, jump */ |
178 | c = cstr + (((c[0] & 0x3f) << 8) | c[1]); |
179 | } |
180 | off += cstr[off] + 1; |
181 | continue; |
182 | } |
183 | /* NUL, skip */ |
184 | off++; |
185 | } |
186 | |
187 | return -1; |
188 | } |
189 | |
190 | /* Computes string to be appended to cstr so that src would be added to |
191 | * the compression (best case, it's a 2-byte pointer to some offset within |
192 | * cstr; worst case, it's all of src, converted to <4>host<3>com<0> format). |
193 | * The computed string is returned directly; its length is returned via retlen; |
194 | * NULL and 0, respectively, are returned if an error occurs. |
195 | */ |
196 | uint8_t* FAST_FUNC dname_enc(const uint8_t *cstr, int clen, const char *src, int *retlen) |
197 | { |
198 | uint8_t *d, *dname; |
199 | int off; |
200 | |
201 | dname = convert_dname(src); |
202 | if (dname == NULL) { |
203 | *retlen = 0; |
204 | return NULL; |
205 | } |
206 | |
207 | d = dname; |
208 | while (*d) { |
209 | if (cstr) { |
210 | off = find_offset(cstr, clen, d); |
211 | if (off >= 0) { /* found a match, add pointer and return */ |
212 | *d++ = NS_CMPRSFLGS | (off >> 8); |
213 | *d = off; |
214 | break; |
215 | } |
216 | } |
217 | d += *d + 1; |
218 | } |
219 | |
220 | *retlen = d - dname + 1; |
221 | return dname; |
222 | } |
223 | |
224 | #ifdef DNS_COMPR_TESTING |
225 | /* gcc -Wall -DDNS_COMPR_TESTING domain_codec.c -o domain_codec && ./domain_codec */ |
226 | int main(int argc, char **argv) |
227 | { |
228 | int len; |
229 | uint8_t *encoded; |
230 | |
231 | #define DNAME_DEC(encoded,pre) dname_dec((uint8_t*)(encoded), sizeof(encoded), (pre)) |
232 | printf("'%s'\n", DNAME_DEC("\4host\3com\0", "test1:")); |
233 | printf("test2:'%s'\n", DNAME_DEC("\4host\3com\0\4host\3com\0", "")); |
234 | printf("test3:'%s'\n", DNAME_DEC("\4host\3com\0\xC0\0", "")); |
235 | printf("test4:'%s'\n", DNAME_DEC("\4host\3com\0\xC0\5", "")); |
236 | printf("test5:'%s'\n", DNAME_DEC("\4host\3com\0\xC0\5\1z\xC0\xA", "")); |
237 | |
238 | #define DNAME_ENC(cache,source,lenp) dname_enc((uint8_t*)(cache), sizeof(cache), (source), (lenp)) |
239 | encoded = dname_enc(NULL, 0, "test.net", &len); |
240 | printf("test6:'%s' len:%d\n", dname_dec(encoded, len, ""), len); |
241 | encoded = DNAME_ENC("\3net\0", "test.net", &len); |
242 | printf("test7:'%s' len:%d\n", dname_dec(encoded, len, ""), len); |
243 | encoded = DNAME_ENC("\4test\3net\0", "test.net", &len); |
244 | printf("test8:'%s' len:%d\n", dname_dec(encoded, len, ""), len); |
245 | return 0; |
246 | } |
247 | #endif |
248 |