xref: /illumos-gate/usr/src/common/smbsrv/smb_netbios_util.c (revision 581cede61ac9c14d8d4ea452562a567189eead78)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #pragma ident	"%Z%%M%	%I%	%E% SMI"
27 
28 #ifdef _KERNEL
29 #include <sys/types.h>
30 #include <sys/sunddi.h>
31 #else
32 #include <string.h>
33 #endif
34 #include <smbsrv/string.h>
35 #include <smbsrv/ctype.h>
36 #include <smbsrv/netbios.h>
37 
38 static int domainname_is_valid(char *domain_name);
39 
40 /*
41  * Routines than support name compression.
42  *
43  *   The NetBIOS name representation in all NetBIOS packets (for NAME,
44  *   SESSION, and DATAGRAM services) is defined in the Domain Name
45  *   Service RFC 883[3] as "compressed" name messages.  This format is
46  *   called "second-level encoding" in the section entitled
47  *   "Representation of NetBIOS Names" in the Concepts and Methods
48  *   document.
49  *
50  *   For ease of description, the first two paragraphs from page 31,
51  *   the section titled "Domain name representation and compression",
52  *   of RFC 883 are replicated here:
53  *
54  *        Domain names messages are expressed in terms of a sequence
55  *        of labels.  Each label is represented as a one octet length
56  *        field followed by that number of octets.  Since every domain
57  *        name ends with the null label of the root, a compressed
58  *        domain name is terminated by a length byte of zero.  The
59  *        high order two bits of the length field must be zero, and
60  *        the remaining six bits of the length field limit the label
61  *        to 63 octets or less.
62  *
63  *        To simplify implementations, the total length of label
64  *        octets and label length octets that make up a domain name is
65  *        restricted to 255 octets or less.
66  *
67  *   The following is the uncompressed representation of the NetBIOS name
68  *   "FRED ", which is the 4 ASCII characters, F, R, E, D, followed by 12
69  *   space characters (0x20).  This name has the SCOPE_ID: "NETBIOS.COM"
70  *
71  *           EGFCEFEECACACACACACACACACACACACA.NETBIOS.COM
72  *
73  *   This uncompressed representation of names is called "first-level
74  *   encoding" in the section entitled "Representation of NetBIOS Names"
75  *   in the Concepts and Methods document.
76  *
77  *   The following is a pictographic representation of the compressed
78  *   representation of the previous uncompressed Domain Name
79  *   representation.
80  *
81  *                        1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 3 3
82  *    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
83  *   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
84  *   |      0x20     |    E (0x45)   |    G (0x47)   |    F (0x46)   |
85  *   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
86  *   |    C (0x43)   |    E (0x45)   |    F (0x46)   |    E (0x45)   |
87  *   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
88  *   |    E (0x45)   |    C (0x43)   |    A (0x41)   |    C (0x43)   |
89  *   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
90  *   |    A (0x41)   |    C (0x43)   |    A (0x41)   |    C (0x43)   |
91  *   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
92  *   |    A (0x41)   |    C (0x43)   |    A (0x41)   |    C (0x43)   |
93  *   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
94  *   |    A (0x41)   |    C (0x43)   |    A (0x41)   |    C (0x43)   |
95  *   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
96  *   |    A (0x41)   |    C (0x43)   |    A (0x41)   |    C (0x43)   |
97  *   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
98  *   |    A (0x41)   |    C (0x43)   |    A (0x41)   |    C (0x43)   |
99  *   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
100  *   |    A (0X41)   |      0x07     |    N (0x4E)   |    E (0x45)   |
101  *   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
102  *   |    T (0x54)   |    B (0x42)   |    I (0x49)   |    O (0x4F)   |
103  *   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
104  *   |    S (0x53)   |      0x03     |    C (0x43)   |    O (0x4F)   |
105  *   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
106  *   |    M (0x4D)   |      0x00     |
107  *   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
108  *
109  *   Each section of a domain name is called a label [7 (page 31)].  A
110  *   label can be a maximum of 63 bytes.  The first byte of a label in
111  *   compressed representation is the number of bytes in the label.  For
112  *   the above example, the first 0x20 is the number of bytes in the
113  *   left-most label, EGFCEFEECACACACACACACACACACACACA, of the domain
114  *   name.  The bytes following the label length count are the characters
115  *   of the label.  The following labels are in sequence after the first
116  *   label, which is the encoded NetBIOS name, until a zero (0x00) length
117  *   count.  The zero length count represents the root label, which is
118  *   always null.
119  *
120  *   A label length count is actually a 6-bit field in the label length
121  *   field.  The most significant 2 bits of the field, bits 7 and 6, are
122  *   flags allowing an escape from the above compressed representation.
123  *   If bits 7 and 6 are both set (11), the following 14 bits are an
124  *   offset pointer into the full message to the actual label string from
125  *   another domain name that belongs in this name.  This label pointer
126  *   allows for a further compression of a domain name in a packet.
127  *
128  *   NetBIOS implementations can only use label string pointers in Name
129  *   Service packets.  They cannot be used in Session or Datagram Service
130  *   packets.
131  *
132  *   The other two possible values for bits 7 and 6 (01 and 10) of a label
133  *   length field are reserved for future use by RFC 883[2 (page 32)].
134  *
135  *   Note that the first octet of a compressed name must contain one of
136  *   the following bit patterns.  (An "x" indicates a bit whose value may
137  *   be either 0 or 1.):
138  *
139  *           00100000 -  Netbios name, length must be 32 (decimal)
140  *           11xxxxxx -  Label string pointer
141  *           10xxxxxx -  Reserved
142  *           01xxxxxx -  Reserved
143  */
144 
145 /*
146  * netbios_first_level_name_encode
147  *
148  * Put test description here.
149  *
150  * Inputs:
151  *	char *	in	-> Name to encode
152  *	char *	out	-> Buffer to encode into.
153  *	int	length	-> # of bytes to encode.
154  *
155  * Returns:
156  *	Nothing
157  */
158 int
159 netbios_first_level_name_encode(unsigned char *name, unsigned char *scope,
160     unsigned char *out, int max_out)
161 {
162 	unsigned char	ch, len;
163 	unsigned char	 *in;
164 	unsigned char	 *lp;
165 	unsigned char	 *op = out;
166 
167 	if (max_out < 0x21)
168 		return (-1);
169 
170 	in = name;
171 	*op++ = 0x20;
172 	for (len = 0; len < NETBIOS_NAME_SZ; len++) {
173 		ch = *in++;
174 		*op++ = 'A' + ((ch >> 4) & 0xF);
175 		*op++ = 'A' + ((ch) & 0xF);
176 	}
177 
178 	max_out -= 0x21;
179 
180 	in = scope;
181 	len = 0;
182 	lp = op++;
183 	while (((ch = *in++) != 0) && (max_out-- > 1)) {
184 		if (ch == 0) {
185 			if ((*lp = len) != 0)
186 				*op++ = 0;
187 			break;
188 		}
189 		if (ch == '.') {
190 			*lp = len;
191 			lp = op++;
192 			len = 0;
193 		} else {
194 			*op++ = ch;
195 			len++;
196 		}
197 	}
198 	*lp = len;
199 	if (len != 0)
200 		*op = 0;
201 
202 	/*LINTED E_PTRDIFF_OVERFLOW*/
203 	return (op - out);
204 }
205 
206 /*
207  * smb_first_level_name_decode
208  *
209  * The null terminated string "in" is the name to decode. The output
210  * is placed in the name_entry structure "name".
211  *
212  * The scope field is a series of length designated labels as described
213  * in the "Domain name representation and compression" section of RFC883.
214  * The two high order two bits of the length field must be zero, the
215  * remaining six bits contain the field length. The total length of the
216  * domain name is restricted to 255 octets but note that the trailing
217  * root label and its dot are not printed. When converting the labels,
218  * the length fields are replaced by dots.
219  *
220  * Returns the number of bytes scanned or -1 to indicate an error.
221  */
222 int
223 netbios_first_level_name_decode(char *in, char *name, char *scope)
224 {
225 	unsigned int	length, bytes;
226 	char		c1, c2;
227 	char		*cp;
228 	char		*out;
229 
230 	cp = in;
231 
232 	if ((length = *cp++) != 0x20) {
233 		return (-1);
234 	}
235 
236 	out = name;
237 	while (length > 0) {
238 		c1 = *cp++;
239 		c2 = *cp++;
240 
241 		if ('A' <= c1 && c1 <= 'P' && 'A' <= c2 && c2 <= 'P') {
242 			c1 -= 'A';
243 			c2 -= 'A';
244 			*out++ = (c1 << 4) | (c2);
245 		} else {
246 			return (-1);		/* conversion error */
247 		}
248 		length -= 2;
249 	}
250 
251 	out = scope;
252 	bytes = 0;
253 	for (length = *cp++; length != 0; length = *cp++) {
254 		if ((length & 0xc0) != 0x00) {
255 			/*
256 			 * This is a pointer or a reserved field. If it's
257 			 * a pointer (16-bits) we have to skip the next byte.
258 			 */
259 			if ((length & 0xc0) == 0xc0) {
260 				cp++;
261 				continue;
262 			}
263 		}
264 
265 		/*
266 		 * Replace the length with a '.', except for the first one.
267 		 */
268 		if (out != scope) {
269 			*out++ = '.';
270 			bytes++;
271 		}
272 
273 		while (length-- > 0) {
274 			if (bytes++ >= (NETBIOS_DOMAIN_NAME_MAX - 1)) {
275 				return (-1);
276 			}
277 			*out++ = *cp++;
278 		}
279 	}
280 	*out = 0;
281 
282 	/*
283 	 * We are supposed to preserve all 8-bits of the domain name
284 	 * but due to the single byte representation in the name cache
285 	 * and UTF-8 encoding everywhere else, we restrict domain names
286 	 * to Appendix 1 - Domain Name Syntax Specification in RFC883.
287 	 */
288 	if (domainname_is_valid(scope))	{
289 		(void) utf8_strupr(scope);
290 		/*LINTED E_PTRDIFF_OVERFLOW*/
291 		return (cp - in);
292 	}
293 
294 	scope[0] = '\0';
295 	return (-1);
296 }
297 
298 /*
299  * smb_netbios_name_isvalid
300  *
301  * This function is provided to be used by session service
302  * which runs in kernel in order to hide name_entry definition.
303  *
304  * It returns the decoded name in the provided buffer as 'out'
305  * if it's not null.
306  *
307  * Returns 0 if decode fails, 1 if it succeeds.
308  */
309 int
310 netbios_name_isvalid(char *in, char *out)
311 {
312 	char name[NETBIOS_NAME_SZ];
313 	char scope[NETBIOS_DOMAIN_NAME_MAX];
314 
315 	if (netbios_first_level_name_decode(in, name, scope) < 0)
316 		return (0);
317 
318 	if (out)
319 		(void) strlcpy(out, name, NETBIOS_NAME_SZ);
320 
321 	return (1);
322 }
323 
324 /*
325  * Characters that we allow in DNS domain names, in addition to
326  * alphanumeric characters. This is not quite consistent with
327  * RFC883. This is global so that it can be patched if there is
328  * a need to change the valid characters in the field.
329  */
330 unsigned char *dns_allowed = (unsigned char *)"-_";
331 
332 /*
333  * dns_is_allowed
334  *
335  * Check the dns_allowed characters and return true (1) if the character
336  * is in the table. Otherwise return false (0).
337  */
338 static int
339 dns_is_allowed(unsigned char c)
340 {
341 	unsigned char *p = dns_allowed;
342 
343 	while (*p) {
344 		if (c == *p++)
345 			return (1);
346 	}
347 
348 	return (0);
349 }
350 
351 
352 /*
353  * domainname_is_valid
354  *
355  * Check the specified domain name for mostly compliance with RFC883
356  * Appendix 1. Names may contain alphanumeric characters, hyphens,
357  * underscores and dots. The first character after a dot must be an
358  * alphabetic character. RFC883 doesn't mention underscores but we
359  * allow it due to common use, and we don't check that labels end
360  * with an alphanumeric character.
361  *
362  * Returns true (1) if the name is valid. Otherwise returns false (0).
363  */
364 static int
365 domainname_is_valid(char *domain_name)
366 {
367 	char *name;
368 	int first_char = 1;
369 
370 	if (domain_name == 0)
371 		return (0);
372 
373 	for (name = domain_name; *name != 0; ++name) {
374 		if (*name == '.') {
375 			first_char = 1;
376 			continue;
377 		}
378 
379 		if (first_char)	{
380 			if (mts_isalpha_ascii(*name) == 0)
381 				return (0);
382 
383 			first_char = 0;
384 			continue;
385 		}
386 
387 		if (mts_isalnum_ascii(*name) || dns_is_allowed(*name))
388 			continue;
389 
390 		return (0);
391 	}
392 
393 	return (1);
394 }
395