The Tcpdump Group git mirrors - tcpdump/blob - print-resp.c

   1 /*
   2  * Copyright (c) 2015 The TCPDUMP project
   3  * All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  *
  14  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  15  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  16  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  17  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
  18  * COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  19  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
  20  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  21  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
  22  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  23  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  24  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  25  * POSSIBILITY OF SUCH DAMAGE.
  26  *
  27  * Initial contribution by Andrew Darqui (andrew.darqui@gmail.com).
  28  */
  29
  30 /* \summary: REdis Serialization Protocol (RESP) printer */
  31
  32 #ifdef HAVE_CONFIG_H
  33 #include <config.h>
  34 #endif
  35
  36 #include "netdissect-stdinc.h"
  37 #include "netdissect.h"
  38 #include <limits.h>
  39
  40 #include "extract.h"
  41
  42
  43 /*
  44  * For information regarding RESP, see: https://redis.io/topics/protocol
  45  */
  46
  47 #define RESP_SIMPLE_STRING    '+'
  48 #define RESP_ERROR            '-'
  49 #define RESP_INTEGER          ':'
  50 #define RESP_BULK_STRING      '$'
  51 #define RESP_ARRAY            '*'
  52
  53 #define resp_print_empty(ndo)            ND_PRINT(" empty")
  54 #define resp_print_null(ndo)             ND_PRINT(" null")
  55 #define resp_print_length_too_large(ndo) ND_PRINT(" length too large")
  56 #define resp_print_length_negative(ndo)  ND_PRINT(" length negative and not -1")
  57 #define resp_print_invalid(ndo)          ND_PRINT(" invalid")
  58
  59 static int resp_parse(netdissect_options *, const u_char *, int);
  60 static int resp_print_string_error_integer(netdissect_options *, const u_char *, int);
  61 static int resp_print_simple_string(netdissect_options *, const u_char *, int);
  62 static int resp_print_integer(netdissect_options *, const u_char *, int);
  63 static int resp_print_error(netdissect_options *, const u_char *, int);
  64 static int resp_print_bulk_string(netdissect_options *, const u_char *, int);
  65 static int resp_print_bulk_array(netdissect_options *, const u_char *, int);
  66 static int resp_print_inline(netdissect_options *, const u_char *, int);
  67 static int resp_get_length(netdissect_options *, const u_char *, int, const u_char **);
  68
  69 #define LCHECK2(_tot_len, _len) \
  70     {                           \
  71         if (_tot_len < _len)    \
  72             goto trunc;         \
  73     }
  74
  75 #define LCHECK(_tot_len) LCHECK2(_tot_len, 1)
  76
  77 /*
  78  * FIND_CRLF:
  79  * Attempts to move our 'ptr' forward until a \r\n is found,
  80  * while also making sure we don't exceed the buffer '_len'
  81  * or go past the end of the captured data.
  82  * If we exceed or go past the end of the captured data,
  83  * jump to trunc.
  84  */
  85 #define FIND_CRLF(_ptr, _len)                   \
  86     for (;;) {                                  \
  87         LCHECK2(_len, 2);                       \
  88         ND_TCHECK_2(_ptr);                      \
  89         if (GET_U_1(_ptr) == '\r' &&            \
  90             GET_U_1(_ptr+1) == '\n')            \
  91             break;                              \
  92         _ptr++;                                 \
  93         _len--;                                 \
  94     }
  95
  96 /*
  97  * CONSUME_CRLF
  98  * Consume a CRLF that we've just found.
  99  */
 100 #define CONSUME_CRLF(_ptr, _len) \
 101     _ptr += 2;                   \
 102     _len -= 2;
 103
 104 /*
 105  * FIND_CR_OR_LF
 106  * Attempts to move our '_ptr' forward until a \r or \n is found,
 107  * while also making sure we don't exceed the buffer '_len'
 108  * or go past the end of the captured data.
 109  * If we exceed or go past the end of the captured data,
 110  * jump to trunc.
 111  */
 112 #define FIND_CR_OR_LF(_ptr, _len)           \
 113     for (;;) {                              \
 114         LCHECK(_len);                       \
 115         if (GET_U_1(_ptr) == '\r' ||        \
 116             GET_U_1(_ptr) == '\n')          \
 117             break;                          \
 118         _ptr++;                             \
 119         _len--;                             \
 120     }
 121
 122 /*
 123  * CONSUME_CR_OR_LF
 124  * Consume all consecutive \r and \n bytes.
 125  * If we exceed '_len' or go past the end of the captured data,
 126  * jump to trunc.
 127  */
 128 #define CONSUME_CR_OR_LF(_ptr, _len)             \
 129     {                                            \
 130         int _found_cr_or_lf = 0;                 \
 131         for (;;) {                               \
 132             /*                                   \
 133              * Have we hit the end of data?      \
 134              */                                  \
 135             if (_len == 0 || !ND_TTEST_1(_ptr)) {\
 136                 /*                               \
 137                  * Yes.  Have we seen a \r       \
 138                  * or \n?                        \
 139                  */                              \
 140                 if (_found_cr_or_lf) {           \
 141                     /*                           \
 142                      * Yes.  Just stop.          \
 143                      */                          \
 144                     break;                       \
 145                 }                                \
 146                 /*                               \
 147                  * No.  We ran out of packet.    \
 148                  */                              \
 149                 goto trunc;                      \
 150             }                                    \
 151             if (GET_U_1(_ptr) != '\r' &&         \
 152                 GET_U_1(_ptr) != '\n')           \
 153                 break;                           \
 154             _found_cr_or_lf = 1;                 \
 155             _ptr++;                              \
 156             _len--;                              \
 157         }                                        \
 158     }
 159
 160 /*
 161  * SKIP_OPCODE
 162  * Skip over the opcode character.
 163  * The opcode has already been fetched, so we know it's there, and don't
 164  * need to do any checks.
 165  */
 166 #define SKIP_OPCODE(_ptr, _tot_len) \
 167     _ptr++;                         \
 168     _tot_len--;
 169
 170 /*
 171  * GET_LENGTH
 172  * Get a bulk string or array length.
 173  */
 174 #define GET_LENGTH(_ndo, _tot_len, _ptr, _len)                \
 175     {                                                         \
 176         const u_char *_endp;                                  \
 177         _len = resp_get_length(_ndo, _ptr, _tot_len, &_endp); \
 178         _tot_len -= (_endp - _ptr);                           \
 179         _ptr = _endp;                                         \
 180     }
 181
 182 /*
 183  * TEST_RET_LEN
 184  * If ret_len is < 0, jump to the trunc tag which returns (-1)
 185  * and 'bubbles up' to printing tstr. Otherwise, return ret_len.
 186  *
 187  * Note that using this macro with a semicolon at the end emits a warning from
 188  * Sun C about an unreachable statement (the semicolon is the statement).
 189  */
 190 #define TEST_RET_LEN(rl) \
 191     if (rl < 0) { goto trunc; } else { return rl; }
 192
 193 /*
 194  * TEST_RET_LEN_NORETURN
 195  * If ret_len is < 0, jump to the trunc tag which returns (-1)
 196  * and 'bubbles up' to printing tstr. Otherwise, continue onward.
 197  */
 198 #define TEST_RET_LEN_NORETURN(rl) \
 199     if (rl < 0) { goto trunc; }
 200
 201 /*
 202  * RESP_PRINT_SEGMENT
 203  * Prints a segment in the form of: ' "<stuff>"\n"
 204  * Assumes the data has already been verified as present.
 205  */
 206 #define RESP_PRINT_SEGMENT(_ndo, _bp, _len)            \
 207     ND_PRINT(" \"");                                   \
 208     if (nd_printn(_ndo, _bp, _len, _ndo->ndo_snapend)) \
 209         goto trunc;                                    \
 210     fn_print_char(_ndo, '"');
 211
 212 void
 213 resp_print(netdissect_options *ndo, const u_char *bp, u_int length)
 214 {
 215     int ret_len = 0;
 216
 217     ndo->ndo_protocol = "resp";
 218
 219     ND_PRINT(": RESP");
 220     while (length != 0) {
 221         /*
 222          * This block supports redis pipelining.
 223          * For example, multiple operations can be pipelined within the same string:
 224          * "*2\r\n\$4\r\nINCR\r\n\$1\r\nz\r\n*2\r\n\$4\r\nINCR\r\n\$1\r\nz\r\n*2\r\n\$4\r\nINCR\r\n\$1\r\nz\r\n"
 225          * or
 226          * "PING\r\nPING\r\nPING\r\n"
 227          * In order to handle this case, we must try and parse 'bp' until
 228          * 'length' bytes have been processed or we reach a trunc condition.
 229          */
 230         ret_len = resp_parse(ndo, bp, length);
 231         TEST_RET_LEN_NORETURN(ret_len);
 232         bp += ret_len;
 233         length -= ret_len;
 234     }
 235
 236     return;
 237
 238 trunc:
 239     nd_print_trunc(ndo);
 240 }
 241
 242 static int
 243 resp_parse(netdissect_options *ndo, const u_char *bp, int length)
 244 {
 245     u_char op;
 246     int ret_len;
 247
 248     LCHECK2(length, 1);
 249     op = GET_U_1(bp);
 250
 251     /* bp now points to the op, so these routines must skip it */
 252     switch(op) {
 253         case RESP_SIMPLE_STRING:  ret_len = resp_print_simple_string(ndo, bp, length);   break;
 254         case RESP_INTEGER:        ret_len = resp_print_integer(ndo, bp, length);         break;
 255         case RESP_ERROR:          ret_len = resp_print_error(ndo, bp, length);           break;
 256         case RESP_BULK_STRING:    ret_len = resp_print_bulk_string(ndo, bp, length);     break;
 257         case RESP_ARRAY:          ret_len = resp_print_bulk_array(ndo, bp, length);      break;
 258         default:                  ret_len = resp_print_inline(ndo, bp, length);          break;
 259     }
 260
 261     /*
 262      * This gives up with a "truncated" indicator for all errors,
 263      * including invalid packet errors; that's what we want, as
 264      * we have to give up on further parsing in that case.
 265      */
 266     TEST_RET_LEN(ret_len) // without a semicolon
 267
 268 trunc:
 269     return (-1);
 270 }
 271
 272 static int
 273 resp_print_simple_string(netdissect_options *ndo, const u_char *bp, int length) {
 274     return resp_print_string_error_integer(ndo, bp, length);
 275 }
 276
 277 static int
 278 resp_print_integer(netdissect_options *ndo, const u_char *bp, int length) {
 279     return resp_print_string_error_integer(ndo, bp, length);
 280 }
 281
 282 static int
 283 resp_print_error(netdissect_options *ndo, const u_char *bp, int length) {
 284     return resp_print_string_error_integer(ndo, bp, length);
 285 }
 286
 287 static int
 288 resp_print_string_error_integer(netdissect_options *ndo, const u_char *bp, int length) {
 289     int length_cur = length, len, ret_len;
 290     const u_char *bp_ptr;
 291
 292     /* bp points to the op; skip it */
 293     SKIP_OPCODE(bp, length_cur);
 294     bp_ptr = bp;
 295
 296     /*
 297      * bp now prints past the (+-;) opcode, so it's pointing to the first
 298      * character of the string (which could be numeric).
 299      * +OK\r\n
 300      * -ERR ...\r\n
 301      * :02912309\r\n
 302      *
 303      * Find the \r\n with FIND_CRLF().
 304      */
 305     FIND_CRLF(bp_ptr, length_cur);
 306
 307     /*
 308      * bp_ptr points to the \r\n, so bp_ptr - bp is the length of text
 309      * preceding the \r\n.  That includes the opcode, so don't print
 310      * that.
 311      */
 312     len = ND_BYTES_BETWEEN(bp, bp_ptr);
 313     RESP_PRINT_SEGMENT(ndo, bp, len);
 314     ret_len = 1 /*<opcode>*/ + len /*<string>*/ + 2 /*<CRLF>*/;
 315
 316     TEST_RET_LEN(ret_len) // without a semicolon
 317
 318 trunc:
 319     return (-1);
 320 }
 321
 322 static int
 323 resp_print_bulk_string(netdissect_options *ndo, const u_char *bp, int length) {
 324     int length_cur = length, string_len;
 325
 326     /* bp points to the op; skip it */
 327     SKIP_OPCODE(bp, length_cur);
 328
 329     /* <length>\r\n */
 330     GET_LENGTH(ndo, length_cur, bp, string_len);
 331
 332     if (string_len >= 0) {
 333         /* Byte string of length string_len, starting at bp */
 334         if (string_len == 0)
 335             resp_print_empty(ndo);
 336         else {
 337             LCHECK2(length_cur, string_len);
 338             ND_TCHECK_LEN(bp, string_len);
 339             RESP_PRINT_SEGMENT(ndo, bp, string_len);
 340             bp += string_len;
 341             length_cur -= string_len;
 342         }
 343
 344         /*
 345          * Find the \r\n at the end of the string and skip past it.
 346          * XXX - report an error if the \r\n isn't immediately after
 347          * the item?
 348          */
 349         FIND_CRLF(bp, length_cur);
 350         CONSUME_CRLF(bp, length_cur);
 351     } else {
 352         /* null, truncated, or invalid for some reason */
 353         switch(string_len) {
 354             case (-1):  resp_print_null(ndo);             break;
 355             case (-2):  goto trunc;
 356             case (-3):  resp_print_length_too_large(ndo); break;
 357             case (-4):  resp_print_length_negative(ndo);  break;
 358             default:    resp_print_invalid(ndo);          break;
 359         }
 360     }
 361
 362     return (length - length_cur);
 363
 364 trunc:
 365     return (-1);
 366 }
 367
 368 static int
 369 resp_print_bulk_array(netdissect_options *ndo, const u_char *bp, int length) {
 370     u_int length_cur = length;
 371     int array_len, i, ret_len;
 372
 373     /* bp points to the op; skip it */
 374     SKIP_OPCODE(bp, length_cur);
 375
 376     /* <array_length>\r\n */
 377     GET_LENGTH(ndo, length_cur, bp, array_len);
 378
 379     if (array_len > 0) {
 380         /* non empty array */
 381         for (i = 0; i < array_len; i++) {
 382             ret_len = resp_parse(ndo, bp, length_cur);
 383
 384             TEST_RET_LEN_NORETURN(ret_len);
 385
 386             bp += ret_len;
 387             length_cur -= ret_len;
 388         }
 389     } else {
 390         /* empty, null, truncated, or invalid */
 391         switch(array_len) {
 392             case 0:     resp_print_empty(ndo);            break;
 393             case (-1):  resp_print_null(ndo);             break;
 394             case (-2):  goto trunc;
 395             case (-3):  resp_print_length_too_large(ndo); break;
 396             case (-4):  resp_print_length_negative(ndo);  break;
 397             default:    resp_print_invalid(ndo);          break;
 398         }
 399     }
 400
 401     return (length - length_cur);
 402
 403 trunc:
 404     return (-1);
 405 }
 406
 407 static int
 408 resp_print_inline(netdissect_options *ndo, const u_char *bp, int length) {
 409     int length_cur = length;
 410     int len;
 411     const u_char *bp_ptr;
 412
 413     /*
 414      * Inline commands are simply 'strings' followed by \r or \n or both.
 415      * Redis will do its best to split/parse these strings.
 416      * This feature of redis is implemented to support the ability of
 417      * command parsing from telnet/nc sessions etc.
 418      *
 419      * <string><\r||\n||\r\n...>
 420      */
 421
 422     /*
 423      * Skip forward past any leading \r, \n, or \r\n.
 424      */
 425     CONSUME_CR_OR_LF(bp, length_cur);
 426     bp_ptr = bp;
 427
 428     /*
 429      * Scan forward looking for \r or \n.
 430      */
 431     FIND_CR_OR_LF(bp_ptr, length_cur);
 432
 433     /*
 434      * Found it; bp_ptr points to the \r or \n, so bp_ptr - bp is the
 435      * Length of the line text that precedes it.  Print it.
 436      */
 437     len = ND_BYTES_BETWEEN(bp, bp_ptr);
 438     RESP_PRINT_SEGMENT(ndo, bp, len);
 439
 440     /*
 441      * Skip forward past the \r, \n, or \r\n.
 442      */
 443     CONSUME_CR_OR_LF(bp_ptr, length_cur);
 444
 445     /*
 446      * Return the number of bytes we processed.
 447      */
 448     return (length - length_cur);
 449
 450 trunc:
 451     return (-1);
 452 }
 453
 454 static int
 455 resp_get_length(netdissect_options *ndo, const u_char *bp, int len, const u_char **endp)
 456 {
 457     int result;
 458     u_char c;
 459     int saw_digit;
 460     int neg;
 461     int too_large;
 462
 463     if (len == 0)
 464         goto trunc;
 465     too_large = 0;
 466     neg = 0;
 467     if (GET_U_1(bp) == '-') {
 468         neg = 1;
 469         bp++;
 470         len--;
 471     }
 472     result = 0;
 473     saw_digit = 0;
 474
 475     for (;;) {
 476         if (len == 0)
 477             goto trunc;
 478         c = GET_U_1(bp);
 479         if (!(c >= '0' && c <= '9')) {
 480             if (!saw_digit) {
 481                 bp++;
 482                 goto invalid;
 483             }
 484             break;
 485         }
 486         c -= '0';
 487         if (result > (INT_MAX / 10)) {
 488             /* This will overflow an int when we multiply it by 10. */
 489             too_large = 1;
 490         } else {
 491             result *= 10;
 492             if (result == ((INT_MAX / 10) * 10) && c > (INT_MAX % 10)) {
 493                 /* This will overflow an int when we add c */
 494                 too_large = 1;
 495             } else
 496                 result += c;
 497         }
 498         bp++;
 499         len--;
 500         saw_digit = 1;
 501     }
 502
 503     /*
 504      * OK, we found a non-digit character.  It should be a \r, followed
 505      * by a \n.
 506      */
 507     if (GET_U_1(bp) != '\r') {
 508         bp++;
 509         goto invalid;
 510     }
 511     bp++;
 512     len--;
 513     if (len == 0)
 514         goto trunc;
 515     if (GET_U_1(bp) != '\n') {
 516         bp++;
 517         goto invalid;
 518     }
 519     bp++;
 520     len--;
 521     *endp = bp;
 522     if (neg) {
 523         /* -1 means "null", anything else is invalid */
 524         if (too_large || result != 1)
 525             return (-4);
 526         result = -1;
 527     }
 528     return (too_large ? -3 : result);
 529
 530 trunc:
 531     *endp = bp;
 532     return (-2);
 533
 534 invalid:
 535     *endp = bp;
 536     return (-5);
 537 }