/src/mosquitto/deps/picohttpparser/picohttpparser.c

Source (jump to first uncovered line)
/*
 * Copyright (c) 2009-2014 Kazuho Oku, Tokuhiro Matsuno, Daisuke Murase,
 *                         Shigeo Mitsunari
 *
 * The software is licensed under either the MIT License (below) or the Perl
 * license.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this software and associated documentation files (the "Software"), to
 * deal in the Software without restriction, including without limitation the
 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
 * sell copies of the Software, and to permit persons to whom the Software is
 * furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
 * IN THE SOFTWARE.
 */

#include <assert.h>
#include <stddef.h>
#include <string.h>
#ifdef __SSE4_2__
#ifdef _MSC_VER
#include <nmmintrin.h>
#else
#include <x86intrin.h>
#endif
#endif
#include "picohttpparser.h"

#if __GNUC__ >= 3
#define likely(x) __builtin_expect(!!(x), 1)
#define unlikely(x) __builtin_expect(!!(x), 0)
#else
#define likely(x) (x)
#define unlikely(x) (x)
#endif

#ifdef _MSC_VER
#define ALIGNED(n) _declspec(align(n))
#else
#define ALIGNED(n) __attribute__((aligned(n)))
#endif

#define IS_PRINTABLE_ASCII(c) ((unsigned char)(c)-040u < 0137u)

#define CHECK_EOF()                                                                                                                \
    if (buf == buf_end) {                                                                                                          \
        *ret = -2;                                                                                                                 \
        return NULL;                                                                                                               \
    }

#define EXPECT_CHAR_NO_CHECK(ch)                                                                                                   \
    if (*buf++ != ch) {                                                                                                            \
        *ret = -1;                                                                                                                 \
        return NULL;                                                                                                               \
    }

#define EXPECT_CHAR(ch)                                                                                                            \
    CHECK_EOF();                                                                                                                   \
    EXPECT_CHAR_NO_CHECK(ch);

#define ADVANCE_TOKEN(tok, toklen)                                                                                                 \
    do {                                                                                                                           \
        const char *tok_start = buf;                                                                                               \
        static const char ALIGNED(16) ranges2[16] = "\000\040\177\177";                                                            \
        int found2;                                                                                                                \
        buf = findchar_fast(buf, buf_end, ranges2, 4, &found2);                                                                    \
        if (!found2) {                                                                                                             \
            CHECK_EOF();                                                                                                           \
        }                                                                                                                          \
        while (1) {                                                                                                                \
            if (*buf == ' ') {                                                                                                     \
                break;                                                                                                             \
            } else if (unlikely(!IS_PRINTABLE_ASCII(*buf))) {                                                                      \
                if ((unsigned char)*buf < '\040' || *buf == '\177') {                                                              \
                    *ret = -1;                                                                                                     \
                    return NULL;                                                                                                   \
                }                                                                                                                  \
            }                                                                                                                      \
            ++buf;                                                                                                                 \
            CHECK_EOF();                                                                                                           \
        }                                                                                                                          \
        tok = tok_start;                                                                                                           \
        toklen = (size_t)(buf - tok_start);                                                                                        \
    } while (0)

static const char *token_char_map = "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"
                                    "\0\1\0\1\1\1\1\1\0\0\1\1\0\1\1\0\1\1\1\1\1\1\1\1\1\1\0\0\0\0\0\0"
                                    "\0\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\0\0\0\1\1"
                                    "\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\0\1\0\1\0"
                                    "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"
                                    "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"
                                    "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"
                                    "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0";

static const char *findchar_fast(const char *buf, const char *buf_end, const char *ranges, size_t ranges_size, int *found)
{
    *found = 0;
#if __SSE4_2__
    if (likely(buf_end - buf >= 16)) {
        __m128i ranges16 = _mm_loadu_si128((const __m128i *)ranges);

        size_t left = (buf_end - buf) & ~15;
        do {
            __m128i b16 = _mm_loadu_si128((const __m128i *)buf);
            int r = _mm_cmpestri(ranges16, ranges_size, b16, 16, _SIDD_LEAST_SIGNIFICANT | _SIDD_CMP_RANGES | _SIDD_UBYTE_OPS);
            if (unlikely(r != 16)) {
                buf += r;
                *found = 1;
                break;
            }
            buf += 16;
            left -= 16;
        } while (likely(left != 0));
    }
#else
    /* suppress unused parameter warning */
    (void)buf_end;
    (void)ranges;
    (void)ranges_size;
#endif
    return buf;
}

static const char *get_token_to_eol(const char *buf, const char *buf_end, const char **token, size_t *token_len, int *ret)
{
    const char *token_start = buf;

#ifdef __SSE4_2__
    static const char ALIGNED(16) ranges1[16] = "\0\010"    /* allow HT */
                                                "\012\037"  /* allow SP and up to but not including DEL */
                                                "\177\177"; /* allow chars w. MSB set */
    int found;
    buf = findchar_fast(buf, buf_end, ranges1, 6, &found);
    if (found)
        goto FOUND_CTL;
#else
    /* find non-printable char within the next 8 bytes, this is the hottest code; manually inlined */
    while (likely(buf_end - buf >= 8)) {
#define DOIT()                                                                                                                     \
    do {                                                                                                                           \
        if (unlikely(!IS_PRINTABLE_ASCII(*buf)))                                                                                   \
            goto NonPrintable;                                                                                                     \
        ++buf;                                                                                                                     \
    } while (0)
        DOIT();
        DOIT();
        DOIT();
        DOIT();
        DOIT();
        DOIT();
        DOIT();
        DOIT();
#undef DOIT
        continue;
    NonPrintable:
        if ((likely((unsigned char)*buf < '\040') && likely(*buf != '\011')) || unlikely(*buf == '\177')) {
            goto FOUND_CTL;
        }
        ++buf;
    }
#endif
    for (;; ++buf) {
        CHECK_EOF();
        if (unlikely(!IS_PRINTABLE_ASCII(*buf))) {
            if ((likely((unsigned char)*buf < '\040') && likely(*buf != '\011')) || unlikely(*buf == '\177')) {
                goto FOUND_CTL;
            }
        }
    }
FOUND_CTL:
    if (likely(*buf == '\015')) {
        ++buf;
        EXPECT_CHAR('\012');
        *token_len = (size_t)(buf - 2 - token_start);
    } else if (*buf == '\012') {
      *token_len = (size_t)(buf - token_start);
        ++buf;
    } else {
        *ret = -1;
        return NULL;
    }
    *token = token_start;

    return buf;
}

static const char *is_complete(const char *buf, const char *buf_end, size_t last_len, int *ret)
{
    int ret_cnt = 0;
    buf = last_len < 3 ? buf : buf + last_len - 3;

    while (1) {
        CHECK_EOF();
        if (*buf == '\015') {
            ++buf;
            CHECK_EOF();
            EXPECT_CHAR('\012');
            ++ret_cnt;
        } else if (*buf == '\012') {
            ++buf;
            ++ret_cnt;
        } else {
            ++buf;
            ret_cnt = 0;
        }
        if (ret_cnt == 2) {
            return buf;
        }
    }
}

#define PARSE_INT(valp_, mul_)                                                                                                     \
    if (*buf < '0' || '9' < *buf) {                                                                                                \
        buf++;                                                                                                                     \
        *ret = -1;                                                                                                                 \
        return NULL;                                                                                                               \
    }                                                                                                                              \
    *(valp_) = (mul_) * (*buf++ - '0');

#define PARSE_INT_3(valp_)                                                                                                         \
    do {                                                                                                                           \
        int res_ = 0;                                                                                                              \
        PARSE_INT(&res_, 100)                                                                                                      \
        *valp_ = res_;                                                                                                             \
        PARSE_INT(&res_, 10)                                                                                                       \
        *valp_ += res_;                                                                                                            \
        PARSE_INT(&res_, 1)                                                                                                        \
        *valp_ += res_;                                                                                                            \
    } while (0)

/* returned pointer is always within [buf, buf_end), or null */
static const char *parse_token(const char *buf, const char *buf_end, const char **token, size_t *token_len, char next_char,
                               int *ret)
{
    /* We use pcmpestri to detect non-token characters. This instruction can take no more than eight character ranges (8*2*8=128
     * bits that is the size of a SSE register). Due to this restriction, characters `|` and `~` are handled in the slow loop. */
    static const char ALIGNED(16) ranges[] = "\x00 "  /* control chars and up to SP */
                                             "\"\""   /* 0x22 */
                                             "()"     /* 0x28,0x29 */
                                             ",,"     /* 0x2c */
                                             "//"     /* 0x2f */
                                             ":@"     /* 0x3a-0x40 */
                                             "[]"     /* 0x5b-0x5d */
                                             "{\xff"; /* 0x7b-0xff */
    const char *buf_start = buf;
    int found;
    buf = findchar_fast(buf, buf_end, ranges, sizeof(ranges) - 1, &found);
    if (!found) {
        CHECK_EOF();
    }
    while (1) {
        if (*buf == next_char) {
            break;
        } else if (!token_char_map[(unsigned char)*buf]) {
            *ret = -1;
            return NULL;
        }
        ++buf;
        CHECK_EOF();
    }
    *token = buf_start;
    *token_len = (size_t)(buf - buf_start);
    return buf;
}

/* returned pointer is always within [buf, buf_end), or null */
static const char *parse_http_version(const char *buf, const char *buf_end, int *minor_version, int *ret)
{
    /* we want at least [HTTP/1.<two chars>] to try to parse */
    if (buf_end - buf < 9) {
        *ret = -2;
        return NULL;
    }
    EXPECT_CHAR_NO_CHECK('H');
    EXPECT_CHAR_NO_CHECK('T');
    EXPECT_CHAR_NO_CHECK('T');
    EXPECT_CHAR_NO_CHECK('P');
    EXPECT_CHAR_NO_CHECK('/');
    EXPECT_CHAR_NO_CHECK('1');
    EXPECT_CHAR_NO_CHECK('.');
    PARSE_INT(minor_version, 1);
    return buf;
}

static const char *parse_headers(const char *buf, const char *buf_end, struct phr_header *headers, size_t *num_headers,
                                 size_t max_headers, int *ret)
{
    for (;; ++*num_headers) {
        CHECK_EOF();
        if (*buf == '\015') {
            ++buf;
            EXPECT_CHAR('\012');
            break;
        } else if (*buf == '\012') {
            ++buf;
            break;
        }
        if (*num_headers == max_headers) {
            *ret = -1;
            return NULL;
        }
        if (!(*num_headers != 0 && (*buf == ' ' || *buf == '\t'))) {
            /* parsing name, but do not discard SP before colon, see
             * http://www.mozilla.org/security/announce/2006/mfsa2006-33.html */
            if ((buf = parse_token(buf, buf_end, &headers[*num_headers].name, &headers[*num_headers].name_len, ':', ret)) == NULL) {
                return NULL;
            }
            if (headers[*num_headers].name_len == 0) {
                *ret = -1;
                return NULL;
            }
            ++buf;
            for (;; ++buf) {
                CHECK_EOF();
                if (!(*buf == ' ' || *buf == '\t')) {
                    break;
                }
            }
        } else {
            headers[*num_headers].name = NULL;
            headers[*num_headers].name_len = 0;
        }
        const char *value;
        size_t value_len;
        if ((buf = get_token_to_eol(buf, buf_end, &value, &value_len, ret)) == NULL) {
            return NULL;
        }
        /* remove trailing SPs and HTABs */
        const char *value_end = value + value_len;
        for (; value_end != value; --value_end) {
            const char c = *(value_end - 1);
            if (!(c == ' ' || c == '\t')) {
                break;
            }
        }
        headers[*num_headers].value = value;
        headers[*num_headers].value_len = (size_t)(value_end - value);
    }
    return buf;
}

static const char *parse_request(const char *buf, const char *buf_end, const char **method, size_t *method_len, const char **path,
                                 size_t *path_len, int *minor_version, struct phr_header *headers, size_t *num_headers,
                                 size_t max_headers, int *ret)
{
    /* skip first empty line (some clients add CRLF after POST content) */
    CHECK_EOF();
    if (*buf == '\015') {
        ++buf;
        EXPECT_CHAR('\012');
    } else if (*buf == '\012') {
        ++buf;
    }

    /* parse request line */
    if ((buf = parse_token(buf, buf_end, method, method_len, ' ', ret)) == NULL) {
        return NULL;
    }
    do {
        ++buf;
        CHECK_EOF();
    } while (*buf == ' ');
    ADVANCE_TOKEN(*path, *path_len);
    do {
        ++buf;
        CHECK_EOF();
    } while (*buf == ' ');
    if (*method_len == 0 || *path_len == 0) {
        *ret = -1;
        return NULL;
    }
    if ((buf = parse_http_version(buf, buf_end, minor_version, ret)) == NULL) {
        return NULL;
    }
    if (*buf == '\015') {
        ++buf;
        EXPECT_CHAR('\012');
    } else if (*buf == '\012') {
        ++buf;
    } else {
        *ret = -1;
        return NULL;
    }

    return parse_headers(buf, buf_end, headers, num_headers, max_headers, ret);
}

int phr_parse_request(const char *buf_start, size_t len, const char **method, size_t *method_len, const char **path,
                      size_t *path_len, int *minor_version, struct phr_header *headers, size_t *num_headers, size_t last_len)
{
    const char *buf = buf_start, *buf_end = buf_start + len;
    size_t max_headers = *num_headers;
    int r;

    *method = NULL;
    *method_len = 0;
    *path = NULL;
    *path_len = 0;
    *minor_version = -1;
    *num_headers = 0;

    /* if last_len != 0, check if the request is complete (a fast countermeasure
       againt slowloris */
    if (last_len != 0 && is_complete(buf, buf_end, last_len, &r) == NULL) {
        return r;
    }

    if ((buf = parse_request(buf, buf_end, method, method_len, path, path_len, minor_version, headers, num_headers, max_headers,
                             &r)) == NULL) {
        return r;
    }

    return (int)(buf - buf_start);
}

static const char *parse_response(const char *buf, const char *buf_end, int *minor_version, int *status, const char **msg,
                                  size_t *msg_len, struct phr_header *headers, size_t *num_headers, size_t max_headers, int *ret)
{
    /* parse "HTTP/1.x" */
    if ((buf = parse_http_version(buf, buf_end, minor_version, ret)) == NULL) {
        return NULL;
    }
    /* skip space */
    if (*buf != ' ') {
        *ret = -1;
        return NULL;
    }
    do {
        ++buf;
        CHECK_EOF();
    } while (*buf == ' ');
    /* parse status code, we want at least [:digit:][:digit:][:digit:]<other char> to try to parse */
    if (buf_end - buf < 4) {
        *ret = -2;
        return NULL;
    }
    PARSE_INT_3(status);

    /* get message including preceding space */
    if ((buf = get_token_to_eol(buf, buf_end, msg, msg_len, ret)) == NULL) {
        return NULL;
    }
    if (*msg_len == 0) {
        /* ok */
    } else if (**msg == ' ') {
        /* Remove preceding space. Successful return from `get_token_to_eol` guarantees that we would hit something other than SP
         * before running past the end of the given buffer. */
        do {
            ++*msg;
            --*msg_len;
        } while (**msg == ' ');
    } else {
        /* garbage found after status code */
        *ret = -1;
        return NULL;
    }

    return parse_headers(buf, buf_end, headers, num_headers, max_headers, ret);
}

int phr_parse_response(const char *buf_start, size_t len, int *minor_version, int *status, const char **msg, size_t *msg_len,
                       struct phr_header *headers, size_t *num_headers, size_t last_len)
{
    const char *buf = buf_start, *buf_end = buf + len;
    size_t max_headers = *num_headers;
    int r;

    *minor_version = -1;
    *status = 0;
    *msg = NULL;
    *msg_len = 0;
    *num_headers = 0;

    /* if last_len != 0, check if the response is complete (a fast countermeasure
       against slowloris */
    if (last_len != 0 && is_complete(buf, buf_end, last_len, &r) == NULL) {
        return r;
    }

    if ((buf = parse_response(buf, buf_end, minor_version, status, msg, msg_len, headers, num_headers, max_headers, &r)) == NULL) {
        return r;
    }

    return (int)(buf - buf_start);
}

int phr_parse_headers(const char *buf_start, size_t len, struct phr_header *headers, size_t *num_headers, size_t last_len)
{
    const char *buf = buf_start, *buf_end = buf + len;
    size_t max_headers = *num_headers;
    int r;

    *num_headers = 0;

    /* if last_len != 0, check if the response is complete (a fast countermeasure
       against slowloris */
    if (last_len != 0 && is_complete(buf, buf_end, last_len, &r) == NULL) {
        return r;
    }

    if ((buf = parse_headers(buf, buf_end, headers, num_headers, max_headers, &r)) == NULL) {
        return r;
    }

    return (int)(buf - buf_start);
}

enum {
    CHUNKED_IN_CHUNK_SIZE,
    CHUNKED_IN_CHUNK_EXT,
    CHUNKED_IN_CHUNK_DATA,
    CHUNKED_IN_CHUNK_CRLF,
    CHUNKED_IN_TRAILERS_LINE_HEAD,
    CHUNKED_IN_TRAILERS_LINE_MIDDLE
};

static int decode_hex(int ch)
{
    if ('0' <= ch && ch <= '9') {
        return ch - '0';
    } else if ('A' <= ch && ch <= 'F') {
        return ch - 'A' + 0xa;
    } else if ('a' <= ch && ch <= 'f') {
        return ch - 'a' + 0xa;
    } else {
        return -1;
    }
}

ssize_t phr_decode_chunked(struct phr_chunked_decoder *decoder, char *buf, size_t *_bufsz)
{
    size_t dst = 0, src = 0, bufsz = *_bufsz;
    ssize_t ret = -2; /* incomplete */

    while (1) {
        switch (decoder->_state) {
        case CHUNKED_IN_CHUNK_SIZE:
            for (;; ++src) {
                int v;
                if (src == bufsz)
                    goto Exit;
                if ((v = decode_hex(buf[src])) == -1) {
                    if (decoder->_hex_count == 0) {
                        ret = -1;
                        goto Exit;
                    }
                    break;
                }
                if (decoder->_hex_count == sizeof(size_t) * 2) {
                    ret = -1;
                    goto Exit;
                }
                decoder->bytes_left_in_chunk = decoder->bytes_left_in_chunk * 16 + (size_t)v;
                ++decoder->_hex_count;
            }
            decoder->_hex_count = 0;
            decoder->_state = CHUNKED_IN_CHUNK_EXT;
        /* fallthru */
        case CHUNKED_IN_CHUNK_EXT:
            /* RFC 7230 A.2 "Line folding in chunk extensions is disallowed" */
            for (;; ++src) {
                if (src == bufsz)
                    goto Exit;
                if (buf[src] == '\012')
                    break;
            }
            ++src;
            if (decoder->bytes_left_in_chunk == 0) {
                if (decoder->consume_trailer) {
                    decoder->_state = CHUNKED_IN_TRAILERS_LINE_HEAD;
                    break;
                } else {
                    goto Complete;
                }
            }
            decoder->_state = CHUNKED_IN_CHUNK_DATA;
        /* fallthru */
        case CHUNKED_IN_CHUNK_DATA: {
            size_t avail = bufsz - src;
            if (avail < decoder->bytes_left_in_chunk) {
                if (dst != src)
                    memmove(buf + dst, buf + src, avail);
                src += avail;
                dst += avail;
                decoder->bytes_left_in_chunk -= avail;
                goto Exit;
            }
            if (dst != src)
                memmove(buf + dst, buf + src, decoder->bytes_left_in_chunk);
            src += decoder->bytes_left_in_chunk;
            dst += decoder->bytes_left_in_chunk;
            decoder->bytes_left_in_chunk = 0;
            decoder->_state = CHUNKED_IN_CHUNK_CRLF;
        }
        /* fallthru */
        case CHUNKED_IN_CHUNK_CRLF:
            for (;; ++src) {
                if (src == bufsz)
                    goto Exit;
                if (buf[src] != '\015')
                    break;
            }
            if (buf[src] != '\012') {
                ret = -1;
                goto Exit;
            }
            ++src;
            decoder->_state = CHUNKED_IN_CHUNK_SIZE;
            break;
        case CHUNKED_IN_TRAILERS_LINE_HEAD:
            for (;; ++src) {
                if (src == bufsz)
                    goto Exit;
                if (buf[src] != '\015')
                    break;
            }
            if (buf[src++] == '\012')
                goto Complete;
            decoder->_state = CHUNKED_IN_TRAILERS_LINE_MIDDLE;
        /* fallthru */
        case CHUNKED_IN_TRAILERS_LINE_MIDDLE:
            for (;; ++src) {
                if (src == bufsz)
                    goto Exit;
                if (buf[src] == '\012')
                    break;
            }
            ++src;
            decoder->_state = CHUNKED_IN_TRAILERS_LINE_HEAD;
            break;
        default:
            assert(!"decoder is corrupt");
        }
    }

Complete:
    ret = (ssize_t)(bufsz - src);
Exit:
    if (dst != src)
        memmove(buf + dst, buf + src, bufsz - src);
    *_bufsz = dst;
    return ret;
}

int phr_decode_chunked_is_in_data(struct phr_chunked_decoder *decoder)
{
    return decoder->_state == CHUNKED_IN_CHUNK_DATA;
}

#undef CHECK_EOF
#undef EXPECT_CHAR
#undef ADVANCE_TOKEN

Coverage Report

Created: 2025-08-26 06:48

Line	Count	Source (jump to first uncovered line)
1		/*
2		* Copyright (c) 2009-2014 Kazuho Oku, Tokuhiro Matsuno, Daisuke Murase,
3		* Shigeo Mitsunari
4		*
5		* The software is licensed under either the MIT License (below) or the Perl
6		* license.
7		*
8		* Permission is hereby granted, free of charge, to any person obtaining a copy
9		* of this software and associated documentation files (the "Software"), to
10		* deal in the Software without restriction, including without limitation the
11		* rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
12		* sell copies of the Software, and to permit persons to whom the Software is
13		* furnished to do so, subject to the following conditions:
14		*
15		* The above copyright notice and this permission notice shall be included in
16		* all copies or substantial portions of the Software.
17		*
18		* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
19		* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
20		* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
21		* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
22		* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
23		* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
24		* IN THE SOFTWARE.
25		*/
26
27		#include <assert.h>
28		#include <stddef.h>
29		#include <string.h>
30		#ifdef __SSE4_2__
31		#ifdef _MSC_VER
32		#include <nmmintrin.h>
33		#else
34		#include <x86intrin.h>
35		#endif
36		#endif
37		#include "picohttpparser.h"
38
39		#if __GNUC__ >= 3
40	0	#define likely(x) __builtin_expect(!!(x), 1)
41	0	#define unlikely(x) __builtin_expect(!!(x), 0)
42		#else
43		#define likely(x) (x)
44		#define unlikely(x) (x)
45		#endif
46
47		#ifdef _MSC_VER
48		#define ALIGNED(n) _declspec(align(n))
49		#else
50		#define ALIGNED(n) __attribute__((aligned(n)))
51		#endif
52
53		#define IS_PRINTABLE_ASCII(c) ((unsigned char)(c)-040u < 0137u)
54
55		#define CHECK_EOF() \
56	0	if (buf == buf_end) { \
57	0	*ret = -2; \
58	0	return NULL; \
59	0	}
60
61		#define EXPECT_CHAR_NO_CHECK(ch) \
62	0	if (*buf++ != ch) { \
63	0	*ret = -1; \
64	0	return NULL; \
65	0	}
66
67		#define EXPECT_CHAR(ch) \
68	0	CHECK_EOF(); \
69	0	EXPECT_CHAR_NO_CHECK(ch);
70
71		#define ADVANCE_TOKEN(tok, toklen) \
72	0	do { \
73	0	const char *tok_start = buf; \
74	0	static const char ALIGNED(16) ranges2[16] = "\000\040\177\177"; \
75	0	int found2; \
76	0	buf = findchar_fast(buf, buf_end, ranges2, 4, &found2); \
77	0	if (!found2) { \
78	0	CHECK_EOF(); \
79	0	} \
80	0	while (1) { \
81	0	if (*buf == ' ') { \
82	0	break; \
83	0	} else if (unlikely(!IS_PRINTABLE_ASCII(*buf))) { \
84	0	if ((unsigned char)buf < '\040' \|\| buf == '\177') { \
85	0	*ret = -1; \
86	0	return NULL; \
87	0	} \
88	0	} \
89	0	++buf; \
90	0	CHECK_EOF(); \
91	0	} \
92	0	tok = tok_start; \
93	0	toklen = (size_t)(buf - tok_start); \
94	0	} while (0)
95
96		static const char *token_char_map = "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"
97		"\0\1\0\1\1\1\1\1\0\0\1\1\0\1\1\0\1\1\1\1\1\1\1\1\1\1\0\0\0\0\0\0"
98		"\0\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\0\0\0\1\1"
99		"\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\0\1\0\1\0"
100		"\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"
101		"\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"
102		"\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"
103		"\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0";
104
105		static const char findchar_fast(const char buf, const char buf_end, const char ranges, size_t ranges_size, int *found)
106	0	{
107	0	*found = 0;
108		#if __SSE4_2__
109		if (likely(buf_end - buf >= 16)) {
110		__m128i ranges16 = _mm_loadu_si128((const __m128i *)ranges);
111
112		size_t left = (buf_end - buf) & ~15;
113		do {
114		__m128i b16 = _mm_loadu_si128((const __m128i *)buf);
115		int r = _mm_cmpestri(ranges16, ranges_size, b16, 16, _SIDD_LEAST_SIGNIFICANT \| _SIDD_CMP_RANGES \| _SIDD_UBYTE_OPS);
116		if (unlikely(r != 16)) {
117		buf += r;
118		*found = 1;
119		break;
120		}
121		buf += 16;
122		left -= 16;
123		} while (likely(left != 0));
124		}
125		#else
126		/* suppress unused parameter warning */
127	0	(void)buf_end;
128	0	(void)ranges;
129	0	(void)ranges_size;
130	0	#endif
131	0	return buf;
132	0	}
133
134		static const char get_token_to_eol(const char buf, const char buf_end, const char token, size_t token_len, int *ret)
135	0	{
136	0	const char *token_start = buf;
137
138		#ifdef __SSE4_2__
139		static const char ALIGNED(16) ranges1[16] = "\0\010" /* allow HT */
140		"\012\037" /* allow SP and up to but not including DEL */
141		"\177\177"; /* allow chars w. MSB set */
142		int found;
143		buf = findchar_fast(buf, buf_end, ranges1, 6, &found);
144		if (found)
145		goto FOUND_CTL;
146		#else
147		/* find non-printable char within the next 8 bytes, this is the hottest code; manually inlined */
148	0	while (likely(buf_end - buf >= 8)) {
149	0	#define DOIT() \
150	0	do { \
151	0	if (unlikely(!IS_PRINTABLE_ASCII(*buf))) \
152	0	goto NonPrintable; \
153	0	++buf; \
154	0	} while (0)
155	0	DOIT();
156	0	DOIT();
157	0	DOIT();
158	0	DOIT();
159	0	DOIT();
160	0	DOIT();
161	0	DOIT();
162	0	DOIT();
163	0	#undef DOIT
164	0	continue;
165	0	NonPrintable:
166	0	if ((likely((unsigned char)buf < '\040') && likely(buf != '\011')) \|\| unlikely(*buf == '\177')) {
167	0	goto FOUND_CTL;
168	0	}
169	0	++buf;
170	0	}
171	0	#endif
172	0	for (;; ++buf) {
173	0	CHECK_EOF();
174	0	if (unlikely(!IS_PRINTABLE_ASCII(*buf))) {
175	0	if ((likely((unsigned char)buf < '\040') && likely(buf != '\011')) \|\| unlikely(*buf == '\177')) {
176	0	goto FOUND_CTL;
177	0	}
178	0	}
179	0	}
180	0	FOUND_CTL:
181	0	if (likely(*buf == '\015')) {
182	0	++buf;
183	0	EXPECT_CHAR('\012');
184	0	*token_len = (size_t)(buf - 2 - token_start);
185	0	} else if (*buf == '\012') {
186	0	*token_len = (size_t)(buf - token_start);
187	0	++buf;
188	0	} else {
189	0	*ret = -1;
190	0	return NULL;
191	0	}
192	0	*token = token_start;
193
194	0	return buf;
195	0	}
196
197		static const char is_complete(const char buf, const char buf_end, size_t last_len, int ret)
198	0	{
199	0	int ret_cnt = 0;
200	0	buf = last_len < 3 ? buf : buf + last_len - 3;
201
202	0	while (1) {
203	0	CHECK_EOF();
204	0	if (*buf == '\015') {
205	0	++buf;
206	0	CHECK_EOF();
207	0	EXPECT_CHAR('\012');
208	0	++ret_cnt;
209	0	} else if (*buf == '\012') {
210	0	++buf;
211	0	++ret_cnt;
212	0	} else {
213	0	++buf;
214	0	ret_cnt = 0;
215	0	}
216	0	if (ret_cnt == 2) {
217	0	return buf;
218	0	}
219	0	}
220	0	}
221
222		#define PARSE_INT(valp_, mul_) \
223	0	if (buf < '0' \|\| '9' < buf) { \
224	0	buf++; \
225	0	*ret = -1; \
226	0	return NULL; \
227	0	} \
228	0	(valp_) = (mul_) (*buf++ - '0');
229
230		#define PARSE_INT_3(valp_) \
231	0	do { \
232	0	int res_ = 0; \
233	0	PARSE_INT(&res_, 100) \
234	0	*valp_ = res_; \
235	0	PARSE_INT(&res_, 10) \
236	0	*valp_ += res_; \
237	0	PARSE_INT(&res_, 1) \
238	0	*valp_ += res_; \
239	0	} while (0)
240
241		/* returned pointer is always within [buf, buf_end), or null */
242		static const char parse_token(const char buf, const char buf_end, const char token, size_t token_len, char next_char,
243		int *ret)
244	0	{
245		/* We use pcmpestri to detect non-token characters. This instruction can take no more than eight character ranges (828=128
246		* bits that is the size of a SSE register). Due to this restriction, characters `\|` and `~` are handled in the slow loop. */
247	0	static const char ALIGNED(16) ranges[] = "\x00 " /* control chars and up to SP */
248	0	"\"\"" /* 0x22 */
249	0	"()" /* 0x28,0x29 */
250	0	",," /* 0x2c */
251	0	"//" /* 0x2f */
252	0	":@" /* 0x3a-0x40 */
253	0	"[]" /* 0x5b-0x5d */
254	0	"{\xff"; /* 0x7b-0xff */
255	0	const char *buf_start = buf;
256	0	int found;
257	0	buf = findchar_fast(buf, buf_end, ranges, sizeof(ranges) - 1, &found);
258	0	if (!found) {
259	0	CHECK_EOF();
260	0	}
261	0	while (1) {
262	0	if (*buf == next_char) {
263	0	break;
264	0	} else if (!token_char_map[(unsigned char)*buf]) {
265	0	*ret = -1;
266	0	return NULL;
267	0	}
268	0	++buf;
269	0	CHECK_EOF();
270	0	}
271	0	*token = buf_start;
272	0	*token_len = (size_t)(buf - buf_start);
273	0	return buf;
274	0	}
275
276		/* returned pointer is always within [buf, buf_end), or null */
277		static const char parse_http_version(const char buf, const char buf_end, int minor_version, int *ret)
278	0	{
279		/* we want at least [HTTP/1.<two chars>] to try to parse */
280	0	if (buf_end - buf < 9) {
281	0	*ret = -2;
282	0	return NULL;
283	0	}
284	0	EXPECT_CHAR_NO_CHECK('H');
285	0	EXPECT_CHAR_NO_CHECK('T');
286	0	EXPECT_CHAR_NO_CHECK('T');
287	0	EXPECT_CHAR_NO_CHECK('P');
288	0	EXPECT_CHAR_NO_CHECK('/');
289	0	EXPECT_CHAR_NO_CHECK('1');
290	0	EXPECT_CHAR_NO_CHECK('.');
291	0	PARSE_INT(minor_version, 1);
292	0	return buf;
293	0	}
294
295		static const char parse_headers(const char buf, const char buf_end, struct phr_header headers, size_t *num_headers,
296		size_t max_headers, int *ret)
297	0	{
298	0	for (;; ++*num_headers) {
299	0	CHECK_EOF();
300	0	if (*buf == '\015') {
301	0	++buf;
302	0	EXPECT_CHAR('\012');
303	0	break;
304	0	} else if (*buf == '\012') {
305	0	++buf;
306	0	break;
307	0	}
308	0	if (*num_headers == max_headers) {
309	0	*ret = -1;
310	0	return NULL;
311	0	}
312	0	if (!(num_headers != 0 && (buf == ' ' \|\| *buf == '\t'))) {
313		/* parsing name, but do not discard SP before colon, see
314		* http://www.mozilla.org/security/announce/2006/mfsa2006-33.html */
315	0	if ((buf = parse_token(buf, buf_end, &headers[num_headers].name, &headers[num_headers].name_len, ':', ret)) == NULL) {
316	0	return NULL;
317	0	}
318	0	if (headers[*num_headers].name_len == 0) {
319	0	*ret = -1;
320	0	return NULL;
321	0	}
322	0	++buf;
323	0	for (;; ++buf) {
324	0	CHECK_EOF();
325	0	if (!(buf == ' ' \|\| buf == '\t')) {
326	0	break;
327	0	}
328	0	}
329	0	} else {
330	0	headers[*num_headers].name = NULL;
331	0	headers[*num_headers].name_len = 0;
332	0	}
333	0	const char *value;
334	0	size_t value_len;
335	0	if ((buf = get_token_to_eol(buf, buf_end, &value, &value_len, ret)) == NULL) {
336	0	return NULL;
337	0	}
338		/* remove trailing SPs and HTABs */
339	0	const char *value_end = value + value_len;
340	0	for (; value_end != value; --value_end) {
341	0	const char c = *(value_end - 1);
342	0	if (!(c == ' ' \|\| c == '\t')) {
343	0	break;
344	0	}
345	0	}
346	0	headers[*num_headers].value = value;
347	0	headers[*num_headers].value_len = (size_t)(value_end - value);
348	0	}
349	0	return buf;
350	0	}
351
352		static const char parse_request(const char buf, const char buf_end, const char method, size_t method_len, const char **path,
353		size_t path_len, int minor_version, struct phr_header headers, size_t num_headers,
354		size_t max_headers, int *ret)
355	0	{
356		/* skip first empty line (some clients add CRLF after POST content) */
357	0	CHECK_EOF();
358	0	if (*buf == '\015') {
359	0	++buf;
360	0	EXPECT_CHAR('\012');
361	0	} else if (*buf == '\012') {
362	0	++buf;
363	0	}
364
365		/* parse request line */
366	0	if ((buf = parse_token(buf, buf_end, method, method_len, ' ', ret)) == NULL) {
367	0	return NULL;
368	0	}
369	0	do {
370	0	++buf;
371	0	CHECK_EOF();
372	0	} while (*buf == ' ');
373	0	ADVANCE_TOKEN(path, path_len);
374	0	do {
375	0	++buf;
376	0	CHECK_EOF();
377	0	} while (*buf == ' ');
378	0	if (method_len == 0 \|\| path_len == 0) {
379	0	*ret = -1;
380	0	return NULL;
381	0	}
382	0	if ((buf = parse_http_version(buf, buf_end, minor_version, ret)) == NULL) {
383	0	return NULL;
384	0	}
385	0	if (*buf == '\015') {
386	0	++buf;
387	0	EXPECT_CHAR('\012');
388	0	} else if (*buf == '\012') {
389	0	++buf;
390	0	} else {
391	0	*ret = -1;
392	0	return NULL;
393	0	}
394
395	0	return parse_headers(buf, buf_end, headers, num_headers, max_headers, ret);
396	0	}
397
398		int phr_parse_request(const char buf_start, size_t len, const char method, size_t method_len, const char **path,
399		size_t path_len, int minor_version, struct phr_header headers, size_t num_headers, size_t last_len)
400	0	{
401	0	const char buf = buf_start, buf_end = buf_start + len;
402	0	size_t max_headers = *num_headers;
403	0	int r;
404
405	0	*method = NULL;
406	0	*method_len = 0;
407	0	*path = NULL;
408	0	*path_len = 0;
409	0	*minor_version = -1;
410	0	*num_headers = 0;
411
412		/* if last_len != 0, check if the request is complete (a fast countermeasure
413		againt slowloris */
414	0	if (last_len != 0 && is_complete(buf, buf_end, last_len, &r) == NULL) {
415	0	return r;
416	0	}
417
418	0	if ((buf = parse_request(buf, buf_end, method, method_len, path, path_len, minor_version, headers, num_headers, max_headers,
419	0	&r)) == NULL) {
420	0	return r;
421	0	}
422
423	0	return (int)(buf - buf_start);
424	0	}
425
426		static const char parse_response(const char buf, const char buf_end, int minor_version, int status, const char *msg,
427		size_t msg_len, struct phr_header headers, size_t num_headers, size_t max_headers, int ret)
428	0	{
429		/* parse "HTTP/1.x" */
430	0	if ((buf = parse_http_version(buf, buf_end, minor_version, ret)) == NULL) {
431	0	return NULL;
432	0	}
433		/* skip space */
434	0	if (*buf != ' ') {
435	0	*ret = -1;
436	0	return NULL;
437	0	}
438	0	do {
439	0	++buf;
440	0	CHECK_EOF();
441	0	} while (*buf == ' ');
442		/* parse status code, we want at least [:digit:][:digit:][:digit:]<other char> to try to parse */
443	0	if (buf_end - buf < 4) {
444	0	*ret = -2;
445	0	return NULL;
446	0	}
447	0	PARSE_INT_3(status);
448
449		/* get message including preceding space */
450	0	if ((buf = get_token_to_eol(buf, buf_end, msg, msg_len, ret)) == NULL) {
451	0	return NULL;
452	0	}
453	0	if (*msg_len == 0) {
454		/* ok */
455	0	} else if (**msg == ' ') {
456		/* Remove preceding space. Successful return from `get_token_to_eol` guarantees that we would hit something other than SP
457		* before running past the end of the given buffer. */
458	0	do {
459	0	++*msg;
460	0	--*msg_len;
461	0	} while (**msg == ' ');
462	0	} else {
463		/* garbage found after status code */
464	0	*ret = -1;
465	0	return NULL;
466	0	}
467
468	0	return parse_headers(buf, buf_end, headers, num_headers, max_headers, ret);
469	0	}
470
471		int phr_parse_response(const char buf_start, size_t len, int minor_version, int status, const char msg, size_t msg_len,
472		struct phr_header headers, size_t num_headers, size_t last_len)
473	0	{
474	0	const char buf = buf_start, buf_end = buf + len;
475	0	size_t max_headers = *num_headers;
476	0	int r;
477
478	0	*minor_version = -1;
479	0	*status = 0;
480	0	*msg = NULL;
481	0	*msg_len = 0;
482	0	*num_headers = 0;
483
484		/* if last_len != 0, check if the response is complete (a fast countermeasure
485		against slowloris */
486	0	if (last_len != 0 && is_complete(buf, buf_end, last_len, &r) == NULL) {
487	0	return r;
488	0	}
489
490	0	if ((buf = parse_response(buf, buf_end, minor_version, status, msg, msg_len, headers, num_headers, max_headers, &r)) == NULL) {
491	0	return r;
492	0	}
493
494	0	return (int)(buf - buf_start);
495	0	}
496
497		int phr_parse_headers(const char buf_start, size_t len, struct phr_header headers, size_t *num_headers, size_t last_len)
498	0	{
499	0	const char buf = buf_start, buf_end = buf + len;
500	0	size_t max_headers = *num_headers;
501	0	int r;
502
503	0	*num_headers = 0;
504
505		/* if last_len != 0, check if the response is complete (a fast countermeasure
506		against slowloris */
507	0	if (last_len != 0 && is_complete(buf, buf_end, last_len, &r) == NULL) {
508	0	return r;
509	0	}
510
511	0	if ((buf = parse_headers(buf, buf_end, headers, num_headers, max_headers, &r)) == NULL) {
512	0	return r;
513	0	}
514
515	0	return (int)(buf - buf_start);
516	0	}
517
518		enum {
519		CHUNKED_IN_CHUNK_SIZE,
520		CHUNKED_IN_CHUNK_EXT,
521		CHUNKED_IN_CHUNK_DATA,
522		CHUNKED_IN_CHUNK_CRLF,
523		CHUNKED_IN_TRAILERS_LINE_HEAD,
524		CHUNKED_IN_TRAILERS_LINE_MIDDLE
525		};
526
527		static int decode_hex(int ch)
528	0	{
529	0	if ('0' <= ch && ch <= '9') {
530	0	return ch - '0';
531	0	} else if ('A' <= ch && ch <= 'F') {
532	0	return ch - 'A' + 0xa;
533	0	} else if ('a' <= ch && ch <= 'f') {
534	0	return ch - 'a' + 0xa;
535	0	} else {
536	0	return -1;
537	0	}
538	0	}
539
540		ssize_t phr_decode_chunked(struct phr_chunked_decoder decoder, char buf, size_t *_bufsz)
541	0	{
542	0	size_t dst = 0, src = 0, bufsz = *_bufsz;
543	0	ssize_t ret = -2; /* incomplete */
544
545	0	while (1) {
546	0	switch (decoder->_state) {
547	0	case CHUNKED_IN_CHUNK_SIZE:
548	0	for (;; ++src) {
549	0	int v;
550	0	if (src == bufsz)
551	0	goto Exit;
552	0	if ((v = decode_hex(buf[src])) == -1) {
553	0	if (decoder->_hex_count == 0) {
554	0	ret = -1;
555	0	goto Exit;
556	0	}
557	0	break;
558	0	}
559	0	if (decoder->_hex_count == sizeof(size_t) * 2) {
560	0	ret = -1;
561	0	goto Exit;
562	0	}
563	0	decoder->bytes_left_in_chunk = decoder->bytes_left_in_chunk * 16 + (size_t)v;
564	0	++decoder->_hex_count;
565	0	}
566	0	decoder->_hex_count = 0;
567	0	decoder->_state = CHUNKED_IN_CHUNK_EXT;
568		/* fallthru */
569	0	case CHUNKED_IN_CHUNK_EXT:
570		/* RFC 7230 A.2 "Line folding in chunk extensions is disallowed" */
571	0	for (;; ++src) {
572	0	if (src == bufsz)
573	0	goto Exit;
574	0	if (buf[src] == '\012')
575	0	break;
576	0	}
577	0	++src;
578	0	if (decoder->bytes_left_in_chunk == 0) {
579	0	if (decoder->consume_trailer) {
580	0	decoder->_state = CHUNKED_IN_TRAILERS_LINE_HEAD;
581	0	break;
582	0	} else {
583	0	goto Complete;
584	0	}
585	0	}
586	0	decoder->_state = CHUNKED_IN_CHUNK_DATA;
587		/* fallthru */
588	0	case CHUNKED_IN_CHUNK_DATA: {
589	0	size_t avail = bufsz - src;
590	0	if (avail < decoder->bytes_left_in_chunk) {
591	0	if (dst != src)
592	0	memmove(buf + dst, buf + src, avail);
593	0	src += avail;
594	0	dst += avail;
595	0	decoder->bytes_left_in_chunk -= avail;
596	0	goto Exit;
597	0	}
598	0	if (dst != src)
599	0	memmove(buf + dst, buf + src, decoder->bytes_left_in_chunk);
600	0	src += decoder->bytes_left_in_chunk;
601	0	dst += decoder->bytes_left_in_chunk;
602	0	decoder->bytes_left_in_chunk = 0;
603	0	decoder->_state = CHUNKED_IN_CHUNK_CRLF;
604	0	}
605		/* fallthru */
606	0	case CHUNKED_IN_CHUNK_CRLF:
607	0	for (;; ++src) {
608	0	if (src == bufsz)
609	0	goto Exit;
610	0	if (buf[src] != '\015')
611	0	break;
612	0	}
613	0	if (buf[src] != '\012') {
614	0	ret = -1;
615	0	goto Exit;
616	0	}
617	0	++src;
618	0	decoder->_state = CHUNKED_IN_CHUNK_SIZE;
619	0	break;
620	0	case CHUNKED_IN_TRAILERS_LINE_HEAD:
621	0	for (;; ++src) {
622	0	if (src == bufsz)
623	0	goto Exit;
624	0	if (buf[src] != '\015')
625	0	break;
626	0	}
627	0	if (buf[src++] == '\012')
628	0	goto Complete;
629	0	decoder->_state = CHUNKED_IN_TRAILERS_LINE_MIDDLE;
630		/* fallthru */
631	0	case CHUNKED_IN_TRAILERS_LINE_MIDDLE:
632	0	for (;; ++src) {
633	0	if (src == bufsz)
634	0	goto Exit;
635	0	if (buf[src] == '\012')
636	0	break;
637	0	}
638	0	++src;
639	0	decoder->_state = CHUNKED_IN_TRAILERS_LINE_HEAD;
640	0	break;
641	0	default:
642	0	assert(!"decoder is corrupt");
643	0	}
644	0	}
645
646	0	Complete:
647	0	ret = (ssize_t)(bufsz - src);
648	0	Exit:
649	0	if (dst != src)
650	0	memmove(buf + dst, buf + src, bufsz - src);
651	0	*_bufsz = dst;
652	0	return ret;
653	0	}
654
655		int phr_decode_chunked_is_in_data(struct phr_chunked_decoder *decoder)
656	0	{
657	0	return decoder->_state == CHUNKED_IN_CHUNK_DATA;
658	0	}
659
660		#undef CHECK_EOF
661		#undef EXPECT_CHAR
662		#undef ADVANCE_TOKEN