1 /*
2 * Copyright 2014-2022 The GmSSL Project. All Rights Reserved.
3 *
4 * Licensed under the Apache License, Version 2.0 (the License); you may
5 * not use this file except in compliance with the License.
6 *
7 * http://www.apache.org/licenses/LICENSE-2.0
8 */
9
10
11 #include <stdio.h>
12 #include <stdlib.h>
13 #include <string.h>
14 #include <limits.h>
15 #include <assert.h>
16 #include <gmssl/base64.h>
17 #include <gmssl/error.h>
18
19 static unsigned char conv_ascii2bin(unsigned char a);
20 #define conv_bin2ascii(a) (data_bin2ascii[(a)&0x3f])
21
22
23 /*-
24 * 64 char lines
25 * pad input with 0
26 * left over chars are set to =
27 * 1 byte => xx==
28 * 2 bytes => xxx=
29 * 3 bytes => xxxx
30 */
31 #define BIN_PER_LINE (64/4*3)
32 #define CHUNKS_PER_LINE (64/4)
33 #define CHAR_PER_LINE (64+1)
34
35 static const unsigned char data_bin2ascii[65] = "ABCDEFGHIJKLMNOPQRSTUVWXYZ\
36 abcdefghijklmnopqrstuvwxyz0123456789+/";
37
38 /*-
39 * 0xF0 is a EOLN
40 * 0xF1 is ignore but next needs to be 0xF0 (for \r\n processing).
41 * 0xF2 is EOF
42 * 0xE0 is ignore at start of line.
43 * 0xFF is error
44 */
45
46 #define B64_EOLN 0xF0
47 #define B64_CR 0xF1
48 #define B64_EOF 0xF2
49 #define B64_WS 0xE0
50 #define B64_ERROR 0xFF
51 #define B64_NOT_BASE64(a) (((a)|0x13) == 0xF3)
52 #define B64_BASE64(a) (!B64_NOT_BASE64(a))
53
54 static const unsigned char data_ascii2bin[128] = {
55 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
56 0xFF, 0xE0, 0xF0, 0xFF, 0xFF, 0xF1, 0xFF, 0xFF,
57 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
58 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
59 0xE0, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
60 0xFF, 0xFF, 0xFF, 0x3E, 0xFF, 0xF2, 0xFF, 0x3F,
61 0x34, 0x35, 0x36, 0x37, 0x38, 0x39, 0x3A, 0x3B,
62 0x3C, 0x3D, 0xFF, 0xFF, 0xFF, 0x00, 0xFF, 0xFF,
63 0xFF, 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06,
64 0x07, 0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E,
65 0x0F, 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16,
66 0x17, 0x18, 0x19, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
67 0xFF, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F, 0x20,
68 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, 0x28,
69 0x29, 0x2A, 0x2B, 0x2C, 0x2D, 0x2E, 0x2F, 0x30,
70 0x31, 0x32, 0x33, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
71 };
72
conv_ascii2bin(unsigned char a)73 static unsigned char conv_ascii2bin(unsigned char a)
74 {
75 if (a & 0x80)
76 return B64_ERROR;
77 return data_ascii2bin[a];
78 }
79
80
base64_ctx_num(BASE64_CTX * ctx)81 int base64_ctx_num(BASE64_CTX *ctx)
82 {
83 return ctx->num;
84 }
85
base64_encode_init(BASE64_CTX * ctx)86 void base64_encode_init(BASE64_CTX *ctx)
87 {
88 ctx->length = 48;
89 ctx->num = 0;
90 ctx->line_num = 0;
91 }
92
base64_encode_update(BASE64_CTX * ctx,const uint8_t * in,int inl,uint8_t * out,int * outl)93 int base64_encode_update(BASE64_CTX *ctx, const uint8_t *in, int inl, uint8_t *out, int *outl)
94 {
95 int i, j;
96 size_t total = 0;
97
98 *outl = 0;
99 if (inl <= 0)
100 return 0;
101 assert(ctx->length <= (int)sizeof(ctx->enc_data));
102 if (ctx->length - ctx->num > inl) {
103 memcpy(&(ctx->enc_data[ctx->num]), in, inl);
104 ctx->num += inl;
105 return 1;
106 }
107 if (ctx->num != 0) {
108 i = ctx->length - ctx->num;
109 memcpy(&(ctx->enc_data[ctx->num]), in, i);
110 in += i;
111 inl -= i;
112 j = base64_encode_block(out, ctx->enc_data, ctx->length);
113 ctx->num = 0;
114 out += j;
115 *(out++) = '\n';
116 *out = '\0';
117 total = j + 1;
118 }
119 while (inl >= ctx->length && total <= INT_MAX) {
120 j = base64_encode_block(out, in, ctx->length);
121 in += ctx->length;
122 inl -= ctx->length;
123 out += j;
124 *(out++) = '\n';
125 *out = '\0';
126 total += j + 1;
127 }
128 if (total > INT_MAX) {
129 /* Too much output data! */
130 *outl = 0;
131 return 0;
132 }
133 if (inl != 0)
134 memcpy(&(ctx->enc_data[0]), in, inl);
135 ctx->num = inl;
136 *outl = total;
137
138 return 1;
139 }
140
base64_encode_finish(BASE64_CTX * ctx,uint8_t * out,int * outl)141 void base64_encode_finish(BASE64_CTX *ctx, uint8_t *out, int *outl)
142 {
143 unsigned int ret = 0;
144
145 if (ctx->num != 0) {
146 ret = base64_encode_block(out, ctx->enc_data, ctx->num);
147 out[ret++] = '\n';
148 out[ret] = '\0';
149 ctx->num = 0;
150 }
151 *outl = ret;
152 }
153
base64_encode_block(unsigned char * t,const unsigned char * f,int dlen)154 int base64_encode_block(unsigned char *t, const unsigned char *f, int dlen)
155 {
156 int i, ret = 0;
157 unsigned long l;
158
159 for (i = dlen; i > 0; i -= 3) {
160 if (i >= 3) {
161 l = (((unsigned long)f[0]) << 16L) |
162 (((unsigned long)f[1]) << 8L) | f[2];
163 *(t++) = conv_bin2ascii(l >> 18L);
164 *(t++) = conv_bin2ascii(l >> 12L);
165 *(t++) = conv_bin2ascii(l >> 6L);
166 *(t++) = conv_bin2ascii(l);
167 } else {
168 l = ((unsigned long)f[0]) << 16L;
169 if (i == 2)
170 l |= ((unsigned long)f[1] << 8L);
171
172 *(t++) = conv_bin2ascii(l >> 18L);
173 *(t++) = conv_bin2ascii(l >> 12L);
174 *(t++) = (i == 1) ? '=' : conv_bin2ascii(l >> 6L);
175 *(t++) = '=';
176 }
177 ret += 4;
178 f += 3;
179 }
180
181 *t = '\0';
182 return (ret);
183 }
184
base64_decode_init(BASE64_CTX * ctx)185 void base64_decode_init(BASE64_CTX *ctx)
186 {
187 /* Only ctx->num is used during decoding. */
188 ctx->num = 0;
189 ctx->length = 0;
190 ctx->line_num = 0;
191 ctx->expect_nl = 0;
192 }
193
194 /*-
195 * -1 for error
196 * 0 for last line
197 * 1 for full line
198 *
199 * Note: even though base64_decode_update attempts to detect and report end of
200 * content, the context doesn't currently remember it and will accept more data
201 * in the next call. Therefore, the caller is responsible for checking and
202 * rejecting a 0 return value in the middle of content.
203 *
204 * Note: even though base64_decode_update has historically tried to detect end of
205 * content based on line length, this has never worked properly. Therefore,
206 * we now return 0 when one of the following is true:
207 * - Padding or B64_EOF was detected and the last block is complete.
208 * - Input has zero-length.
209 * -1 is returned if:
210 * - Invalid characters are detected.
211 * - There is extra trailing padding, or data after padding.
212 * - B64_EOF is detected after an incomplete base64 block.
213 */
base64_decode_update(BASE64_CTX * ctx,const uint8_t * in,int inl,uint8_t * out,int * outl)214 int base64_decode_update(BASE64_CTX *ctx, const uint8_t *in, int inl, uint8_t *out, int *outl)
215 {
216 int seof = 0, eof = 0, rv = -1, ret = 0, i, v, tmp, n, decoded_len;
217 unsigned char *d;
218
219 n = ctx->num;
220 d = ctx->enc_data;
221
222 if (n > 0 && d[n - 1] == '=') {
223 eof++;
224 if (n > 1 && d[n - 2] == '=')
225 eof++;
226 }
227
228 /* Legacy behaviour: an empty input chunk signals end of input. */
229 if (inl == 0) {
230 rv = 0;
231 goto end;
232 }
233
234 for (i = 0; i < inl; i++) {
235 tmp = *(in++);
236 v = conv_ascii2bin(tmp);
237 if (v == B64_ERROR) {
238 rv = -1;
239 error_print();
240 goto end;
241 }
242
243 if (tmp == '=') {
244 eof++;
245 } else if (eof > 0 && B64_BASE64(v)) {
246 /* More data after padding. */
247 rv = -1;
248 error_print();
249 goto end;
250 }
251
252 if (eof > 2) {
253 rv = -1;
254 error_print();
255 goto end;
256 }
257
258 if (v == B64_EOF) {
259 seof = 1;
260 goto tail;
261 }
262
263 /* Only save valid base64 characters. */
264 if (B64_BASE64(v)) {
265 if (n >= 64) {
266 /*
267 * We increment n once per loop, and empty the buffer as soon as
268 * we reach 64 characters, so this can only happen if someone's
269 * manually messed with the ctx. Refuse to write any more data.
270 */
271 rv = -1;
272 error_print();
273 goto end;
274 }
275 assert(n < (int)sizeof(ctx->enc_data));
276 d[n++] = tmp;
277 }
278
279 if (n == 64) {
280 decoded_len = base64_decode_block(out, d, n);
281 n = 0;
282 if (decoded_len < 0 || eof > decoded_len) {
283 rv = -1;
284 goto end;
285 }
286 ret += decoded_len - eof;
287 out += decoded_len - eof;
288 }
289 }
290
291 /*
292 * Legacy behaviour: if the current line is a full base64-block (i.e., has
293 * 0 mod 4 base64 characters), it is processed immediately. We keep this
294 * behaviour as applications may not be calling base64_decode_final properly.
295 */
296 tail:
297 if (n > 0) {
298 if ((n & 3) == 0) {
299 decoded_len = base64_decode_block(out, d, n);
300 n = 0;
301 if (decoded_len < 0 || eof > decoded_len) {
302 error_print();
303 rv = -1;
304 goto end;
305 }
306 ret += (decoded_len - eof);
307 } else if (seof) {
308 /* EOF in the middle of a base64 block. */
309 error_print();
310 rv = -1;
311 goto end;
312 }
313 }
314
315 rv = seof || (n == 0 && eof) ? 0 : 1;
316 end:
317 /* Legacy behaviour. This should probably rather be zeroed on error. */
318 *outl = ret;
319 ctx->num = n;
320 return (rv);
321 }
322
base64_decode_block(unsigned char * t,const unsigned char * f,int n)323 int base64_decode_block(unsigned char *t, const unsigned char *f, int n)
324 {
325 int i, ret = 0, a, b, c, d;
326 unsigned long l;
327
328 /* trim white space from the start of the line. */
329 while ((conv_ascii2bin(*f) == B64_WS) && (n > 0)) {
330 f++;
331 n--;
332 }
333
334 /*
335 * strip off stuff at the end of the line ascii2bin values B64_WS,
336 * B64_EOLN, B64_EOLN and B64_EOF
337 */
338 while ((n > 3) && (B64_NOT_BASE64(conv_ascii2bin(f[n - 1]))))
339 n--;
340
341 if (n % 4 != 0)
342 return (-1);
343
344 for (i = 0; i < n; i += 4) {
345 a = conv_ascii2bin(*(f++));
346 b = conv_ascii2bin(*(f++));
347 c = conv_ascii2bin(*(f++));
348 d = conv_ascii2bin(*(f++));
349 if ((a & 0x80) || (b & 0x80) || (c & 0x80) || (d & 0x80))
350 return (-1);
351 l = ((((unsigned long)a) << 18L) |
352 (((unsigned long)b) << 12L) |
353 (((unsigned long)c) << 6L) | (((unsigned long)d)));
354 *(t++) = (unsigned char)(l >> 16L) & 0xff;
355 *(t++) = (unsigned char)(l >> 8L) & 0xff;
356 *(t++) = (unsigned char)(l) & 0xff;
357 ret += 3;
358 }
359 return (ret);
360 }
361
base64_decode_finish(BASE64_CTX * ctx,uint8_t * out,int * outl)362 int base64_decode_finish(BASE64_CTX *ctx, uint8_t *out, int *outl)
363 {
364 int i;
365
366 *outl = 0;
367 if (ctx->num != 0) {
368 i = base64_decode_block(out, ctx->enc_data, ctx->num);
369 if (i < 0) {
370 error_print();
371 return (-1);
372 }
373 ctx->num = 0;
374 *outl = i;
375 return (1);
376 } else
377 return (1);
378 }
379