Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 1 | #ifndef _SPOP_FUNCTIONS_H |
| 2 | #define _SPOP_FUNCTIONS_H |
| 3 | |
Willy Tarreau | a1bd1fa | 2019-03-29 17:26:33 +0100 | [diff] [blame] | 4 | #include <inttypes.h> |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 5 | #include <string.h> |
| 6 | #include <spoe_types.h> |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 7 | |
| 8 | |
| 9 | #ifndef MIN |
| 10 | #define MIN(a, b) (((a) < (b)) ? (a) : (b)) |
| 11 | #endif |
| 12 | |
| 13 | #ifndef MAX |
| 14 | #define MAX(a, b) (((a) > (b)) ? (a) : (b)) |
| 15 | #endif |
| 16 | |
| 17 | |
| 18 | /* Encode the integer <i> into a varint (variable-length integer). The encoded |
| 19 | * value is copied in <*buf>. Here is the encoding format: |
| 20 | * |
| 21 | * 0 <= X < 240 : 1 byte (7.875 bits) [ XXXX XXXX ] |
| 22 | * 240 <= X < 2288 : 2 bytes (11 bits) [ 1111 XXXX ] [ 0XXX XXXX ] |
| 23 | * 2288 <= X < 264432 : 3 bytes (18 bits) [ 1111 XXXX ] [ 1XXX XXXX ] [ 0XXX XXXX ] |
| 24 | * 264432 <= X < 33818864 : 4 bytes (25 bits) [ 1111 XXXX ] [ 1XXX XXXX ]*2 [ 0XXX XXXX ] |
| 25 | * 33818864 <= X < 4328786160 : 5 bytes (32 bits) [ 1111 XXXX ] [ 1XXX XXXX ]*3 [ 0XXX XXXX ] |
| 26 | * ... |
| 27 | * |
| 28 | * On success, it returns the number of written bytes and <*buf> is moved after |
| 29 | * the encoded value. Otherwise, it returns -1. */ |
| 30 | static inline int |
| 31 | encode_varint(uint64_t i, char **buf, char *end) |
| 32 | { |
| 33 | unsigned char *p = (unsigned char *)*buf; |
| 34 | int r; |
| 35 | |
| 36 | if (p >= (unsigned char *)end) |
| 37 | return -1; |
| 38 | |
| 39 | if (i < 240) { |
| 40 | *p++ = i; |
| 41 | *buf = (char *)p; |
| 42 | return 1; |
| 43 | } |
| 44 | |
| 45 | *p++ = (unsigned char)i | 240; |
| 46 | i = (i - 240) >> 4; |
| 47 | while (i >= 128) { |
| 48 | if (p >= (unsigned char *)end) |
| 49 | return -1; |
| 50 | *p++ = (unsigned char)i | 128; |
| 51 | i = (i - 128) >> 7; |
| 52 | } |
| 53 | |
| 54 | if (p >= (unsigned char *)end) |
| 55 | return -1; |
| 56 | *p++ = (unsigned char)i; |
| 57 | |
| 58 | r = ((char *)p - *buf); |
| 59 | *buf = (char *)p; |
| 60 | return r; |
| 61 | } |
| 62 | |
| 63 | /* Decode a varint from <*buf> and save the decoded value in <*i>. See |
| 64 | * 'spoe_encode_varint' for details about varint. |
| 65 | * On success, it returns the number of read bytes and <*buf> is moved after the |
| 66 | * varint. Otherwise, it returns -1. */ |
| 67 | static inline int |
| 68 | decode_varint(char **buf, char *end, uint64_t *i) |
| 69 | { |
| 70 | unsigned char *p = (unsigned char *)*buf; |
| 71 | int r; |
| 72 | |
| 73 | if (p >= (unsigned char *)end) |
| 74 | return -1; |
| 75 | |
| 76 | *i = *p++; |
| 77 | if (*i < 240) { |
| 78 | *buf = (char *)p; |
| 79 | return 1; |
| 80 | } |
| 81 | |
| 82 | r = 4; |
| 83 | do { |
| 84 | if (p >= (unsigned char *)end) |
| 85 | return -1; |
| 86 | *i += (uint64_t)*p << r; |
| 87 | r += 7; |
| 88 | } while (*p++ >= 128); |
| 89 | |
| 90 | r = ((char *)p - *buf); |
| 91 | *buf = (char *)p; |
| 92 | return r; |
| 93 | } |
| 94 | |
| 95 | /* Encode a buffer. Its length <len> is encoded as a varint, followed by a copy |
| 96 | * of <str>. It must have enough space in <*buf> to encode the buffer, else an |
| 97 | * error is triggered. |
| 98 | * On success, it returns <len> and <*buf> is moved after the encoded value. If |
| 99 | * an error occurred, it returns -1. */ |
| 100 | static inline int |
| 101 | spoe_encode_buffer(const char *str, size_t len, char **buf, char *end) |
| 102 | { |
| 103 | char *p = *buf; |
| 104 | int ret; |
| 105 | |
| 106 | if (p >= end) |
| 107 | return -1; |
| 108 | |
| 109 | if (!len) { |
| 110 | *p++ = 0; |
| 111 | *buf = p; |
| 112 | return 0; |
| 113 | } |
| 114 | |
| 115 | ret = encode_varint(len, &p, end); |
| 116 | if (ret == -1 || p + len > end) |
| 117 | return -1; |
| 118 | |
| 119 | memcpy(p, str, len); |
| 120 | *buf = p + len; |
| 121 | return len; |
| 122 | } |
| 123 | |
| 124 | /* Encode a buffer, possibly partially. It does the same thing than |
| 125 | * 'spoe_encode_buffer', but if there is not enough space, it does not fail. |
| 126 | * On success, it returns the number of copied bytes and <*buf> is moved after |
Joseph Herlant | ebe14bb | 2018-11-09 18:36:35 -0800 | [diff] [blame] | 127 | * the encoded value. If an error occurred, it returns -1. */ |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 128 | static inline int |
| 129 | spoe_encode_frag_buffer(const char *str, size_t len, char **buf, char *end) |
| 130 | { |
| 131 | char *p = *buf; |
| 132 | int ret; |
| 133 | |
| 134 | if (p >= end) |
| 135 | return -1; |
| 136 | |
| 137 | if (!len) { |
| 138 | *p++ = 0; |
| 139 | *buf = p; |
| 140 | return 0; |
| 141 | } |
| 142 | |
| 143 | ret = encode_varint(len, &p, end); |
| 144 | if (ret == -1 || p >= end) |
| 145 | return -1; |
| 146 | |
| 147 | ret = (p+len < end) ? len : (end - p); |
| 148 | memcpy(p, str, ret); |
| 149 | *buf = p + ret; |
| 150 | return ret; |
| 151 | } |
| 152 | |
| 153 | /* Decode a buffer. The buffer length is decoded and saved in <*len>. <*str> |
| 154 | * points on the first byte of the buffer. |
| 155 | * On success, it returns the buffer length and <*buf> is moved after the |
| 156 | * encoded buffer. Otherwise, it returns -1. */ |
| 157 | static inline int |
| 158 | spoe_decode_buffer(char **buf, char *end, char **str, uint64_t *len) |
| 159 | { |
| 160 | char *p = *buf; |
| 161 | uint64_t sz; |
| 162 | int ret; |
| 163 | |
| 164 | *str = NULL; |
| 165 | *len = 0; |
| 166 | |
| 167 | ret = decode_varint(&p, end, &sz); |
| 168 | if (ret == -1 || p + sz > end) |
| 169 | return -1; |
| 170 | |
| 171 | *str = p; |
| 172 | *len = sz; |
| 173 | *buf = p + sz; |
| 174 | return sz; |
| 175 | } |
| 176 | |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 177 | /* Encode a typed data using value in <data> and type <type>. On success, it |
| 178 | * returns the number of copied bytes and <*buf> is moved after the encoded |
Joseph Herlant | ebe14bb | 2018-11-09 18:36:35 -0800 | [diff] [blame] | 179 | * value. If an error occurred, it returns -1. |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 180 | * |
| 181 | * If the value is too big to be encoded, depending on its type, then encoding |
| 182 | * failed or the value is partially encoded. Only strings and binaries can be |
| 183 | * partially encoded. In this case, the offset <*off> is updated to known how |
| 184 | * many bytes has been encoded. If <*off> is zero at the end, it means that all |
| 185 | * data has been encoded. */ |
| 186 | static inline int |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 187 | spoe_encode_data(union spoe_data *data, enum spoe_data_type type, unsigned int *off, char **buf, char *end) |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 188 | { |
| 189 | char *p = *buf; |
| 190 | int ret; |
| 191 | |
| 192 | if (p >= end) |
| 193 | return -1; |
| 194 | |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 195 | if (data == NULL) { |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 196 | *p++ = SPOE_DATA_T_NULL; |
| 197 | goto end; |
| 198 | } |
| 199 | |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 200 | *p++ = type; |
| 201 | switch (type) { |
| 202 | case SPOE_DATA_T_BOOL: |
| 203 | p[-1] |= (data->boolean ? SPOE_DATA_FL_TRUE : SPOE_DATA_FL_FALSE); |
| 204 | break; |
| 205 | |
| 206 | case SPOE_DATA_T_INT32: |
| 207 | if (encode_varint(data->int32, &p, end) == -1) |
| 208 | return -1; |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 209 | break; |
| 210 | |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 211 | case SPOE_DATA_T_UINT32: |
| 212 | if (encode_varint(data->uint32, &p, end) == -1) |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 213 | return -1; |
| 214 | break; |
| 215 | |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 216 | case SPOE_DATA_T_INT64: |
| 217 | if (encode_varint(data->int64, &p, end) == -1) |
| 218 | return -1; |
| 219 | break; |
| 220 | |
| 221 | case SPOE_DATA_T_UINT64: |
| 222 | if (encode_varint(data->uint64, &p, end) == -1) |
| 223 | return -1; |
| 224 | break; |
| 225 | |
| 226 | case SPOE_DATA_T_IPV4: |
| 227 | if (p + 4 > end) |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 228 | return -1; |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 229 | memcpy(p, &data->ipv4, 4); |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 230 | p += 4; |
| 231 | break; |
| 232 | |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 233 | case SPOE_DATA_T_IPV6: |
| 234 | if (p + 16 > end) |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 235 | return -1; |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 236 | memcpy(p, &data->ipv6, 16); |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 237 | p += 16; |
| 238 | break; |
| 239 | |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 240 | case SPOE_DATA_T_STR: |
| 241 | case SPOE_DATA_T_BIN: { |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 242 | /* Here, we need to know if the sample has already been |
| 243 | * partially encoded. If yes, we only need to encode the |
| 244 | * remaining, <*off> reprensenting the number of bytes |
| 245 | * already encoded. */ |
| 246 | if (!*off) { |
| 247 | /* First evaluation of the sample : encode the |
| 248 | * type (string or binary), the buffer length |
| 249 | * (as a varint) and at least 1 byte of the |
| 250 | * buffer. */ |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 251 | ret = spoe_encode_frag_buffer(data->chk.ptr, data->chk.len, &p, end); |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 252 | if (ret == -1) |
| 253 | return -1; |
| 254 | } |
| 255 | else { |
| 256 | /* The sample has been fragmented, encode remaining data */ |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 257 | ret = MIN(data->chk.len - *off, end - p); |
| 258 | memcpy(p, data->chk.ptr + *off, ret); |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 259 | p += ret; |
| 260 | } |
| 261 | /* Now update <*off> */ |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 262 | if (ret + *off != data->chk.len) |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 263 | *off += ret; |
| 264 | else |
| 265 | *off = 0; |
| 266 | break; |
| 267 | } |
| 268 | /* |
| 269 | case SMP_T_METH: { |
| 270 | char *m; |
| 271 | size_t len; |
| 272 | |
| 273 | *p++ = SPOE_DATA_T_STR; |
| 274 | switch (smp->data.u.meth.meth) { |
| 275 | case HTTP_METH_OPTIONS: m = "OPTIONS"; len = 7; break; |
| 276 | case HTTP_METH_GET : m = "GET"; len = 3; break; |
| 277 | case HTTP_METH_HEAD : m = "HEAD"; len = 4; break; |
| 278 | case HTTP_METH_POST : m = "POST"; len = 4; break; |
| 279 | case HTTP_METH_PUT : m = "PUT"; len = 3; break; |
| 280 | case HTTP_METH_DELETE : m = "DELETE"; len = 6; break; |
| 281 | case HTTP_METH_TRACE : m = "TRACE"; len = 5; break; |
| 282 | case HTTP_METH_CONNECT: m = "CONNECT"; len = 7; break; |
| 283 | |
| 284 | default : |
| 285 | m = smp->data.u.meth.str.str; |
| 286 | len = smp->data.u.meth.str.len; |
| 287 | } |
| 288 | if (spoe_encode_buffer(m, len, &p, end) == -1) |
| 289 | return -1; |
| 290 | break; |
| 291 | } |
| 292 | */ |
| 293 | |
| 294 | default: |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 295 | /* send type NULL for unknown types */ |
| 296 | p[-1] = SPOE_DATA_T_NULL; |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 297 | break; |
| 298 | } |
| 299 | |
| 300 | end: |
| 301 | ret = (p - *buf); |
| 302 | *buf = p; |
| 303 | return ret; |
| 304 | } |
| 305 | |
| 306 | /* Skip a typed data. If an error occurred, -1 is returned, otherwise the number |
| 307 | * of skipped bytes is returned and the <*buf> is moved after skipped data. |
| 308 | * |
| 309 | * A types data is composed of a type (1 byte) and corresponding data: |
| 310 | * - boolean: non additional data (0 bytes) |
| 311 | * - integers: a variable-length integer (see decode_varint) |
| 312 | * - ipv4: 4 bytes |
| 313 | * - ipv6: 16 bytes |
| 314 | * - binary and string: a buffer prefixed by its size, a variable-length |
| 315 | * integer (see spoe_decode_buffer) */ |
| 316 | static inline int |
| 317 | spoe_skip_data(char **buf, char *end) |
| 318 | { |
| 319 | char *str, *p = *buf; |
| 320 | int type, ret; |
| 321 | uint64_t v, sz; |
| 322 | |
| 323 | if (p >= end) |
| 324 | return -1; |
| 325 | |
| 326 | type = *p++; |
| 327 | switch (type & SPOE_DATA_T_MASK) { |
| 328 | case SPOE_DATA_T_BOOL: |
| 329 | break; |
| 330 | case SPOE_DATA_T_INT32: |
| 331 | case SPOE_DATA_T_INT64: |
| 332 | case SPOE_DATA_T_UINT32: |
| 333 | case SPOE_DATA_T_UINT64: |
| 334 | if (decode_varint(&p, end, &v) == -1) |
| 335 | return -1; |
| 336 | break; |
| 337 | case SPOE_DATA_T_IPV4: |
| 338 | if (p+4 > end) |
| 339 | return -1; |
| 340 | p += 4; |
| 341 | break; |
| 342 | case SPOE_DATA_T_IPV6: |
| 343 | if (p+16 > end) |
| 344 | return -1; |
| 345 | p += 16; |
| 346 | break; |
| 347 | case SPOE_DATA_T_STR: |
| 348 | case SPOE_DATA_T_BIN: |
| 349 | /* All the buffer must be skipped */ |
| 350 | if (spoe_decode_buffer(&p, end, &str, &sz) == -1) |
| 351 | return -1; |
| 352 | break; |
| 353 | } |
| 354 | |
| 355 | ret = (p - *buf); |
| 356 | *buf = p; |
| 357 | return ret; |
| 358 | } |
| 359 | |
| 360 | /* Decode a typed data and fill <smp>. If an error occurred, -1 is returned, |
| 361 | * otherwise the number of read bytes is returned and <*buf> is moved after the |
| 362 | * decoded data. See spoe_skip_data for details. */ |
| 363 | static inline int |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 364 | spoe_decode_data(char **buf, char *end, union spoe_data *data, enum spoe_data_type *type) |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 365 | { |
| 366 | char *str, *p = *buf; |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 367 | int v, r = 0; |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 368 | uint64_t sz; |
| 369 | |
| 370 | if (p >= end) |
| 371 | return -1; |
| 372 | |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 373 | v = *p++; |
| 374 | *type = v & SPOE_DATA_T_MASK; |
| 375 | |
| 376 | switch (*type) { |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 377 | case SPOE_DATA_T_BOOL: |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 378 | data->boolean = ((v & SPOE_DATA_FL_MASK) == SPOE_DATA_FL_TRUE); |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 379 | break; |
| 380 | case SPOE_DATA_T_INT32: |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 381 | if (decode_varint(&p, end, &sz) == -1) |
| 382 | return -1; |
| 383 | data->int32 = sz; |
| 384 | break; |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 385 | case SPOE_DATA_T_INT64: |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 386 | if (decode_varint(&p, end, &sz) == -1) |
| 387 | return -1; |
| 388 | data->int64 = sz; |
| 389 | break; |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 390 | case SPOE_DATA_T_UINT32: |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 391 | if (decode_varint(&p, end, &sz) == -1) |
| 392 | return -1; |
| 393 | data->uint32 = sz; |
| 394 | break; |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 395 | case SPOE_DATA_T_UINT64: |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 396 | if (decode_varint(&p, end, &sz) == -1) |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 397 | return -1; |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 398 | data->uint64 = sz; |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 399 | break; |
| 400 | case SPOE_DATA_T_IPV4: |
| 401 | if (p+4 > end) |
| 402 | return -1; |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 403 | memcpy(&data->ipv4, p, 4); |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 404 | p += 4; |
| 405 | break; |
| 406 | case SPOE_DATA_T_IPV6: |
| 407 | if (p+16 > end) |
| 408 | return -1; |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 409 | memcpy(&data->ipv6, p, 16); |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 410 | p += 16; |
| 411 | break; |
| 412 | case SPOE_DATA_T_STR: |
| 413 | case SPOE_DATA_T_BIN: |
| 414 | /* All the buffer must be decoded */ |
| 415 | if (spoe_decode_buffer(&p, end, &str, &sz) == -1) |
| 416 | return -1; |
Willy Tarreau | 75f4246 | 2017-11-14 15:01:22 +0100 | [diff] [blame] | 417 | data->chk.ptr = str; |
| 418 | data->chk.len = sz; |
| 419 | break; |
| 420 | default: /* SPOE_DATA_T_NULL, unknown */ |
Eric Salama | 8a9c6c2 | 2017-11-10 11:02:23 +0100 | [diff] [blame] | 421 | break; |
| 422 | } |
| 423 | |
| 424 | r = (p - *buf); |
| 425 | *buf = p; |
| 426 | return r; |
| 427 | } |
| 428 | |
| 429 | |
| 430 | #endif |