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
59static int resp_parse(netdissect_options *, const u_char *, int);
60static int resp_print_string_error_integer(netdissect_options *, const u_char *, int);
61static int resp_print_simple_string(netdissect_options *, const u_char *, int);
62static int resp_print_integer(netdissect_options *, const u_char *, int);
63static int resp_print_error(netdissect_options *, const u_char *, int);
64static int resp_print_bulk_string(netdissect_options *, const u_char *, int);
65static int resp_print_bulk_array(netdissect_options *, const u_char *, int);
66static int resp_print_inline(netdissect_options *, const u_char *, int);
67static 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#define TEST_RET_LEN(rl) \
188    if (rl < 0) { goto trunc; } else { return rl; }
189
190/*
191 * TEST_RET_LEN_NORETURN
192 * If ret_len is < 0, jump to the trunc tag which returns (-1)
193 * and 'bubbles up' to printing tstr. Otherwise, continue onward.
194 */
195#define TEST_RET_LEN_NORETURN(rl) \
196    if (rl < 0) { goto trunc; }
197
198/*
199 * RESP_PRINT_SEGMENT
200 * Prints a segment in the form of: ' "<stuff>"\n"
201 * Assumes the data has already been verified as present.
202 */
203#define RESP_PRINT_SEGMENT(_ndo, _bp, _len)            \
204    ND_PRINT(" \"");                                   \
205    if (nd_printn(_ndo, _bp, _len, _ndo->ndo_snapend)) \
206        goto trunc;                                    \
207    fn_print_char(_ndo, '"');
208
209void
210resp_print(netdissect_options *ndo, const u_char *bp, u_int length)
211{
212    int ret_len = 0;
213
214    ndo->ndo_protocol = "resp";
215
216    ND_PRINT(": RESP");
217    while (length > 0) {
218        /*
219         * This block supports redis pipelining.
220         * For example, multiple operations can be pipelined within the same string:
221         * "*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"
222         * or
223         * "PING\r\nPING\r\nPING\r\n"
224         * In order to handle this case, we must try and parse 'bp' until
225         * 'length' bytes have been processed or we reach a trunc condition.
226         */
227        ret_len = resp_parse(ndo, bp, length);
228        TEST_RET_LEN_NORETURN(ret_len);
229        bp += ret_len;
230        length -= ret_len;
231    }
232
233    return;
234
235trunc:
236    nd_print_trunc(ndo);
237}
238
239static int
240resp_parse(netdissect_options *ndo, const u_char *bp, int length)
241{
242    u_char op;
243    int ret_len;
244
245    LCHECK2(length, 1);
246    op = GET_U_1(bp);
247
248    /* bp now points to the op, so these routines must skip it */
249    switch(op) {
250        case RESP_SIMPLE_STRING:  ret_len = resp_print_simple_string(ndo, bp, length);   break;
251        case RESP_INTEGER:        ret_len = resp_print_integer(ndo, bp, length);         break;
252        case RESP_ERROR:          ret_len = resp_print_error(ndo, bp, length);           break;
253        case RESP_BULK_STRING:    ret_len = resp_print_bulk_string(ndo, bp, length);     break;
254        case RESP_ARRAY:          ret_len = resp_print_bulk_array(ndo, bp, length);      break;
255        default:                  ret_len = resp_print_inline(ndo, bp, length);          break;
256    }
257
258    /*
259     * This gives up with a "truncated" indicator for all errors,
260     * including invalid packet errors; that's what we want, as
261     * we have to give up on further parsing in that case.
262     */
263    TEST_RET_LEN(ret_len);
264
265trunc:
266    return (-1);
267}
268
269static int
270resp_print_simple_string(netdissect_options *ndo, const u_char *bp, int length) {
271    return resp_print_string_error_integer(ndo, bp, length);
272}
273
274static int
275resp_print_integer(netdissect_options *ndo, const u_char *bp, int length) {
276    return resp_print_string_error_integer(ndo, bp, length);
277}
278
279static int
280resp_print_error(netdissect_options *ndo, const u_char *bp, int length) {
281    return resp_print_string_error_integer(ndo, bp, length);
282}
283
284static int
285resp_print_string_error_integer(netdissect_options *ndo, const u_char *bp, int length) {
286    int length_cur = length, len, ret_len;
287    const u_char *bp_ptr;
288
289    /* bp points to the op; skip it */
290    SKIP_OPCODE(bp, length_cur);
291    bp_ptr = bp;
292
293    /*
294     * bp now prints past the (+-;) opcode, so it's pointing to the first
295     * character of the string (which could be numeric).
296     * +OK\r\n
297     * -ERR ...\r\n
298     * :02912309\r\n
299     *
300     * Find the \r\n with FIND_CRLF().
301     */
302    FIND_CRLF(bp_ptr, length_cur);
303
304    /*
305     * bp_ptr points to the \r\n, so bp_ptr - bp is the length of text
306     * preceding the \r\n.  That includes the opcode, so don't print
307     * that.
308     */
309    len = ND_BYTES_BETWEEN(bp_ptr, bp);
310    RESP_PRINT_SEGMENT(ndo, bp, len);
311    ret_len = 1 /*<opcode>*/ + len /*<string>*/ + 2 /*<CRLF>*/;
312
313    TEST_RET_LEN(ret_len);
314
315trunc:
316    return (-1);
317}
318
319static int
320resp_print_bulk_string(netdissect_options *ndo, const u_char *bp, int length) {
321    int length_cur = length, string_len;
322
323    /* bp points to the op; skip it */
324    SKIP_OPCODE(bp, length_cur);
325
326    /* <length>\r\n */
327    GET_LENGTH(ndo, length_cur, bp, string_len);
328
329    if (string_len >= 0) {
330        /* Byte string of length string_len, starting at bp */
331        if (string_len == 0)
332            resp_print_empty(ndo);
333        else {
334            LCHECK2(length_cur, string_len);
335            ND_TCHECK_LEN(bp, string_len);
336            RESP_PRINT_SEGMENT(ndo, bp, string_len);
337            bp += string_len;
338            length_cur -= string_len;
339        }
340
341        /*
342         * Find the \r\n at the end of the string and skip past it.
343         * XXX - report an error if the \r\n isn't immediately after
344         * the item?
345         */
346        FIND_CRLF(bp, length_cur);
347        CONSUME_CRLF(bp, length_cur);
348    } else {
349        /* null, truncated, or invalid for some reason */
350        switch(string_len) {
351            case (-1):  resp_print_null(ndo);             break;
352            case (-2):  goto trunc;
353            case (-3):  resp_print_length_too_large(ndo); break;
354            case (-4):  resp_print_length_negative(ndo);  break;
355            default:    resp_print_invalid(ndo);          break;
356        }
357    }
358
359    return (length - length_cur);
360
361trunc:
362    return (-1);
363}
364
365static int
366resp_print_bulk_array(netdissect_options *ndo, const u_char *bp, int length) {
367    u_int length_cur = length;
368    int array_len, i, ret_len;
369
370    /* bp points to the op; skip it */
371    SKIP_OPCODE(bp, length_cur);
372
373    /* <array_length>\r\n */
374    GET_LENGTH(ndo, length_cur, bp, array_len);
375
376    if (array_len > 0) {
377        /* non empty array */
378        for (i = 0; i < array_len; i++) {
379            ret_len = resp_parse(ndo, bp, length_cur);
380
381            TEST_RET_LEN_NORETURN(ret_len);
382
383            bp += ret_len;
384            length_cur -= ret_len;
385        }
386    } else {
387        /* empty, null, truncated, or invalid */
388        switch(array_len) {
389            case 0:     resp_print_empty(ndo);            break;
390            case (-1):  resp_print_null(ndo);             break;
391            case (-2):  goto trunc;
392            case (-3):  resp_print_length_too_large(ndo); break;
393            case (-4):  resp_print_length_negative(ndo);  break;
394            default:    resp_print_invalid(ndo);          break;
395        }
396    }
397
398    return (length - length_cur);
399
400trunc:
401    return (-1);
402}
403
404static int
405resp_print_inline(netdissect_options *ndo, const u_char *bp, int length) {
406    int length_cur = length;
407    int len;
408    const u_char *bp_ptr;
409
410    /*
411     * Inline commands are simply 'strings' followed by \r or \n or both.
412     * Redis will do its best to split/parse these strings.
413     * This feature of redis is implemented to support the ability of
414     * command parsing from telnet/nc sessions etc.
415     *
416     * <string><\r||\n||\r\n...>
417     */
418
419    /*
420     * Skip forward past any leading \r, \n, or \r\n.
421     */
422    CONSUME_CR_OR_LF(bp, length_cur);
423    bp_ptr = bp;
424
425    /*
426     * Scan forward looking for \r or \n.
427     */
428    FIND_CR_OR_LF(bp_ptr, length_cur);
429
430    /*
431     * Found it; bp_ptr points to the \r or \n, so bp_ptr - bp is the
432     * Length of the line text that precedes it.  Print it.
433     */
434    len = ND_BYTES_BETWEEN(bp_ptr, bp);
435    RESP_PRINT_SEGMENT(ndo, bp, len);
436
437    /*
438     * Skip forward past the \r, \n, or \r\n.
439     */
440    CONSUME_CR_OR_LF(bp_ptr, length_cur);
441
442    /*
443     * Return the number of bytes we processed.
444     */
445    return (length - length_cur);
446
447trunc:
448    return (-1);
449}
450
451static int
452resp_get_length(netdissect_options *ndo, const u_char *bp, int len, const u_char **endp)
453{
454    int result;
455    u_char c;
456    int saw_digit;
457    int neg;
458    int too_large;
459
460    if (len == 0)
461        goto trunc;
462    too_large = 0;
463    neg = 0;
464    if (GET_U_1(bp) == '-') {
465        neg = 1;
466        bp++;
467        len--;
468    }
469    result = 0;
470    saw_digit = 0;
471
472    for (;;) {
473        if (len == 0)
474            goto trunc;
475        c = GET_U_1(bp);
476        if (!(c >= '0' && c <= '9')) {
477            if (!saw_digit) {
478                bp++;
479                goto invalid;
480            }
481            break;
482        }
483        c -= '0';
484        if (result > (INT_MAX / 10)) {
485            /* This will overflow an int when we multiply it by 10. */
486            too_large = 1;
487        } else {
488            result *= 10;
489            if (result == ((INT_MAX / 10) * 10) && c > (INT_MAX % 10)) {
490                /* This will overflow an int when we add c */
491                too_large = 1;
492            } else
493                result += c;
494        }
495        bp++;
496        len--;
497        saw_digit = 1;
498    }
499
500    /*
501     * OK, we found a non-digit character.  It should be a \r, followed
502     * by a \n.
503     */
504    if (GET_U_1(bp) != '\r') {
505        bp++;
506        goto invalid;
507    }
508    bp++;
509    len--;
510    if (len == 0)
511        goto trunc;
512    if (GET_U_1(bp) != '\n') {
513        bp++;
514        goto invalid;
515    }
516    bp++;
517    len--;
518    *endp = bp;
519    if (neg) {
520        /* -1 means "null", anything else is invalid */
521        if (too_large || result != 1)
522            return (-4);
523        result = -1;
524    }
525    return (too_large ? -3 : result);
526
527trunc:
528    *endp = bp;
529    return (-2);
530
531invalid:
532    *endp = bp;
533    return (-5);
534}
535