diff options
author | Riccardo Spagni <ric@spagni.net> | 2014-10-05 23:44:31 +0200 |
---|---|---|
committer | Riccardo Spagni <ric@spagni.net> | 2014-10-05 23:44:31 +0200 |
commit | 9ef094b356b4da7542c3cab898dac7e135b76903 (patch) | |
tree | 99b5876712b0b1551fc042fe75447b998e4b0fc1 /external/unbound/ldns/parseutil.c | |
parent | split mnemonic printout over 3 lines (diff) | |
download | monero-9ef094b356b4da7542c3cab898dac7e135b76903.tar.xz |
added unbound to external deps
Diffstat (limited to 'external/unbound/ldns/parseutil.c')
-rw-r--r-- | external/unbound/ldns/parseutil.c | 726 |
1 files changed, 726 insertions, 0 deletions
diff --git a/external/unbound/ldns/parseutil.c b/external/unbound/ldns/parseutil.c new file mode 100644 index 000000000..55e3a5b1a --- /dev/null +++ b/external/unbound/ldns/parseutil.c @@ -0,0 +1,726 @@ +/* + * parseutil.c - parse utilities for string and wire conversion + * + * (c) NLnet Labs, 2004-2006 + * + * See the file LICENSE for the license + */ +/** + * \file + * + * Utility functions for parsing, base32(DNS variant) and base64 encoding + * and decoding, Hex, Time units, Escape codes. + */ + +#include "config.h" +#include "ldns/parseutil.h" +#include <sys/time.h> +#include <time.h> +#include <ctype.h> + +sldns_lookup_table * +sldns_lookup_by_name(sldns_lookup_table *table, const char *name) +{ + while (table->name != NULL) { + if (strcasecmp(name, table->name) == 0) + return table; + table++; + } + return NULL; +} + +sldns_lookup_table * +sldns_lookup_by_id(sldns_lookup_table *table, int id) +{ + while (table->name != NULL) { + if (table->id == id) + return table; + table++; + } + return NULL; +} + +/* Number of days per month (except for February in leap years). */ +static const int mdays[] = { + 31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31 +}; + +#define LDNS_MOD(x,y) (((x) % (y) < 0) ? ((x) % (y) + (y)) : ((x) % (y))) +#define LDNS_DIV(x,y) (((x) % (y) < 0) ? ((x) / (y) - 1 ) : ((x) / (y))) + +static int +is_leap_year(int year) +{ + return LDNS_MOD(year, 4) == 0 && (LDNS_MOD(year, 100) != 0 + || LDNS_MOD(year, 400) == 0); +} + +static int +leap_days(int y1, int y2) +{ + --y1; + --y2; + return (LDNS_DIV(y2, 4) - LDNS_DIV(y1, 4)) - + (LDNS_DIV(y2, 100) - LDNS_DIV(y1, 100)) + + (LDNS_DIV(y2, 400) - LDNS_DIV(y1, 400)); +} + +/* + * Code adapted from Python 2.4.1 sources (Lib/calendar.py). + */ +time_t +sldns_mktime_from_utc(const struct tm *tm) +{ + int year = 1900 + tm->tm_year; + time_t days = 365 * ((time_t) year - 1970) + leap_days(1970, year); + time_t hours; + time_t minutes; + time_t seconds; + int i; + + for (i = 0; i < tm->tm_mon; ++i) { + days += mdays[i]; + } + if (tm->tm_mon > 1 && is_leap_year(year)) { + ++days; + } + days += tm->tm_mday - 1; + + hours = days * 24 + tm->tm_hour; + minutes = hours * 60 + tm->tm_min; + seconds = minutes * 60 + tm->tm_sec; + + return seconds; +} + +#if SIZEOF_TIME_T <= 4 + +static void +sldns_year_and_yday_from_days_since_epoch(int64_t days, struct tm *result) +{ + int year = 1970; + int new_year; + + while (days < 0 || days >= (int64_t) (is_leap_year(year) ? 366 : 365)) { + new_year = year + (int) LDNS_DIV(days, 365); + days -= (new_year - year) * 365; + days -= leap_days(year, new_year); + year = new_year; + } + result->tm_year = year; + result->tm_yday = (int) days; +} + +/* Number of days per month in a leap year. */ +static const int leap_year_mdays[] = { + 31, 29, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31 +}; + +static void +sldns_mon_and_mday_from_year_and_yday(struct tm *result) +{ + int idays = result->tm_yday; + const int *mon_lengths = is_leap_year(result->tm_year) ? + leap_year_mdays : mdays; + + result->tm_mon = 0; + while (idays >= mon_lengths[result->tm_mon]) { + idays -= mon_lengths[result->tm_mon++]; + } + result->tm_mday = idays + 1; +} + +static void +sldns_wday_from_year_and_yday(struct tm *result) +{ + result->tm_wday = 4 /* 1-1-1970 was a thursday */ + + LDNS_MOD((result->tm_year - 1970), 7) * LDNS_MOD(365, 7) + + leap_days(1970, result->tm_year) + + result->tm_yday; + result->tm_wday = LDNS_MOD(result->tm_wday, 7); + if (result->tm_wday < 0) { + result->tm_wday += 7; + } +} + +static struct tm * +sldns_gmtime64_r(int64_t clock, struct tm *result) +{ + result->tm_isdst = 0; + result->tm_sec = (int) LDNS_MOD(clock, 60); + clock = LDNS_DIV(clock, 60); + result->tm_min = (int) LDNS_MOD(clock, 60); + clock = LDNS_DIV(clock, 60); + result->tm_hour = (int) LDNS_MOD(clock, 24); + clock = LDNS_DIV(clock, 24); + + sldns_year_and_yday_from_days_since_epoch(clock, result); + sldns_mon_and_mday_from_year_and_yday(result); + sldns_wday_from_year_and_yday(result); + result->tm_year -= 1900; + + return result; +} + +#endif /* SIZEOF_TIME_T <= 4 */ + +static int64_t +sldns_serial_arithmitics_time(int32_t time, time_t now) +{ + int32_t offset = time - (int32_t) now; + return (int64_t) now + offset; +} + +struct tm * +sldns_serial_arithmitics_gmtime_r(int32_t time, time_t now, struct tm *result) +{ +#if SIZEOF_TIME_T <= 4 + int64_t secs_since_epoch = sldns_serial_arithmitics_time(time, now); + return sldns_gmtime64_r(secs_since_epoch, result); +#else + time_t secs_since_epoch = sldns_serial_arithmitics_time(time, now); + return gmtime_r(&secs_since_epoch, result); +#endif +} + +int +sldns_hexdigit_to_int(char ch) +{ + switch (ch) { + case '0': return 0; + case '1': return 1; + case '2': return 2; + case '3': return 3; + case '4': return 4; + case '5': return 5; + case '6': return 6; + case '7': return 7; + case '8': return 8; + case '9': return 9; + case 'a': case 'A': return 10; + case 'b': case 'B': return 11; + case 'c': case 'C': return 12; + case 'd': case 'D': return 13; + case 'e': case 'E': return 14; + case 'f': case 'F': return 15; + default: + return -1; + } +} + +uint32_t +sldns_str2period(const char *nptr, const char **endptr) +{ + int sign = 0; + uint32_t i = 0; + uint32_t seconds = 0; + + for(*endptr = nptr; **endptr; (*endptr)++) { + switch (**endptr) { + case ' ': + case '\t': + break; + case '-': + if(sign == 0) { + sign = -1; + } else { + return seconds; + } + break; + case '+': + if(sign == 0) { + sign = 1; + } else { + return seconds; + } + break; + case 's': + case 'S': + seconds += i; + i = 0; + break; + case 'm': + case 'M': + seconds += i * 60; + i = 0; + break; + case 'h': + case 'H': + seconds += i * 60 * 60; + i = 0; + break; + case 'd': + case 'D': + seconds += i * 60 * 60 * 24; + i = 0; + break; + case 'w': + case 'W': + seconds += i * 60 * 60 * 24 * 7; + i = 0; + break; + case '0': + case '1': + case '2': + case '3': + case '4': + case '5': + case '6': + case '7': + case '8': + case '9': + i *= 10; + i += (**endptr - '0'); + break; + default: + seconds += i; + /* disregard signedness */ + return seconds; + } + } + seconds += i; + /* disregard signedness */ + return seconds; +} + +int +sldns_parse_escape(uint8_t *ch_p, const char** str_p) +{ + uint16_t val; + + if ((*str_p)[0] && isdigit((*str_p)[0]) && + (*str_p)[1] && isdigit((*str_p)[1]) && + (*str_p)[2] && isdigit((*str_p)[2])) { + + val = (uint16_t)(((*str_p)[0] - '0') * 100 + + ((*str_p)[1] - '0') * 10 + + ((*str_p)[2] - '0')); + + if (val > 255) { + goto error; + } + *ch_p = (uint8_t)val; + *str_p += 3; + return 1; + + } else if ((*str_p)[0] && !isdigit((*str_p)[0])) { + + *ch_p = (uint8_t)*(*str_p)++; + return 1; + } +error: + *str_p = NULL; + return 0; /* LDNS_WIREPARSE_ERR_SYNTAX_BAD_ESCAPE */ +} + +/** parse one character, with escape codes */ +int +sldns_parse_char(uint8_t *ch_p, const char** str_p) +{ + switch (**str_p) { + + case '\0': return 0; + + case '\\': *str_p += 1; + return sldns_parse_escape(ch_p, str_p); + + default: *ch_p = (uint8_t)*(*str_p)++; + return 1; + } +} + +size_t sldns_b32_ntop_calculate_size(size_t src_data_length) +{ + return src_data_length == 0 ? 0 : ((src_data_length - 1) / 5 + 1) * 8; +} + +size_t sldns_b32_ntop_calculate_size_no_padding(size_t src_data_length) +{ + return ((src_data_length + 3) * 8 / 5) - 4; +} + +static int +sldns_b32_ntop_base(const uint8_t* src, size_t src_sz, char* dst, size_t dst_sz, + int extended_hex, int add_padding) +{ + size_t ret_sz; + const char* b32 = extended_hex ? "0123456789abcdefghijklmnopqrstuv" + : "abcdefghijklmnopqrstuvwxyz234567"; + + size_t c = 0; /* c is used to carry partial base32 character over + * byte boundaries for sizes with a remainder. + * (i.e. src_sz % 5 != 0) + */ + + ret_sz = add_padding ? sldns_b32_ntop_calculate_size(src_sz) + : sldns_b32_ntop_calculate_size_no_padding(src_sz); + + /* Do we have enough space? */ + if (dst_sz < ret_sz + 1) + return -1; + + /* We know the size; terminate the string */ + dst[ret_sz] = '\0'; + + /* First process all chunks of five */ + while (src_sz >= 5) { + /* 00000... ........ ........ ........ ........ */ + dst[0] = b32[(src[0] ) >> 3]; + + /* .....111 11...... ........ ........ ........ */ + dst[1] = b32[(src[0] & 0x07) << 2 | src[1] >> 6]; + + /* ........ ..22222. ........ ........ ........ */ + dst[2] = b32[(src[1] & 0x3e) >> 1]; + + /* ........ .......3 3333.... ........ ........ */ + dst[3] = b32[(src[1] & 0x01) << 4 | src[2] >> 4]; + + /* ........ ........ ....4444 4....... ........ */ + dst[4] = b32[(src[2] & 0x0f) << 1 | src[3] >> 7]; + + /* ........ ........ ........ .55555.. ........ */ + dst[5] = b32[(src[3] & 0x7c) >> 2]; + + /* ........ ........ ........ ......66 666..... */ + dst[6] = b32[(src[3] & 0x03) << 3 | src[4] >> 5]; + + /* ........ ........ ........ ........ ...77777 */ + dst[7] = b32[(src[4] & 0x1f) ]; + + src_sz -= 5; + src += 5; + dst += 8; + } + /* Process what remains */ + switch (src_sz) { + case 4: /* ........ ........ ........ ......66 666..... */ + dst[6] = b32[(src[3] & 0x03) << 3]; + + /* ........ ........ ........ .55555.. ........ */ + dst[5] = b32[(src[3] & 0x7c) >> 2]; + + /* ........ ........ ....4444 4....... ........ */ + c = src[3] >> 7 ; + case 3: dst[4] = b32[(src[2] & 0x0f) << 1 | c]; + + /* ........ .......3 3333.... ........ ........ */ + c = src[2] >> 4 ; + case 2: dst[3] = b32[(src[1] & 0x01) << 4 | c]; + + /* ........ ..22222. ........ ........ ........ */ + dst[2] = b32[(src[1] & 0x3e) >> 1]; + + /* .....111 11...... ........ ........ ........ */ + c = src[1] >> 6 ; + case 1: dst[1] = b32[(src[0] & 0x07) << 2 | c]; + + /* 00000... ........ ........ ........ ........ */ + dst[0] = b32[ src[0] >> 3]; + } + /* Add padding */ + if (add_padding) { + switch (src_sz) { + case 1: dst[2] = '='; + dst[3] = '='; + case 2: dst[4] = '='; + case 3: dst[5] = '='; + dst[6] = '='; + case 4: dst[7] = '='; + } + } + return (int)ret_sz; +} + +int +sldns_b32_ntop(const uint8_t* src, size_t src_sz, char* dst, size_t dst_sz) +{ + return sldns_b32_ntop_base(src, src_sz, dst, dst_sz, 0, 1); +} + +int +sldns_b32_ntop_extended_hex(const uint8_t* src, size_t src_sz, + char* dst, size_t dst_sz) +{ + return sldns_b32_ntop_base(src, src_sz, dst, dst_sz, 1, 1); +} + +size_t sldns_b32_pton_calculate_size(size_t src_text_length) +{ + return src_text_length * 5 / 8; +} + +static int +sldns_b32_pton_base(const char* src, size_t src_sz, uint8_t* dst, size_t dst_sz, + int extended_hex, int check_padding) +{ + size_t i = 0; + char ch = '\0'; + uint8_t buf[8]; + uint8_t* start = dst; + + while (src_sz) { + /* Collect 8 characters in buf (if possible) */ + for (i = 0; i < 8; i++) { + + do { + ch = *src++; + --src_sz; + + } while (isspace(ch) && src_sz > 0); + + if (ch == '=' || ch == '\0') + break; + + else if (extended_hex) + + if (ch >= '0' && ch <= '9') + buf[i] = (uint8_t)ch - '0'; + else if (ch >= 'a' && ch <= 'v') + buf[i] = (uint8_t)ch - 'a' + 10; + else if (ch >= 'A' && ch <= 'V') + buf[i] = (uint8_t)ch - 'A' + 10; + else + return -1; + + else if (ch >= 'a' && ch <= 'z') + buf[i] = (uint8_t)ch - 'a'; + else if (ch >= 'A' && ch <= 'Z') + buf[i] = (uint8_t)ch - 'A'; + else if (ch >= '2' && ch <= '7') + buf[i] = (uint8_t)ch - '2' + 26; + else + return -1; + } + /* Less that 8 characters. We're done. */ + if (i < 8) + break; + + /* Enough space available at the destination? */ + if (dst_sz < 5) + return -1; + + /* 00000... ........ ........ ........ ........ */ + /* .....111 11...... ........ ........ ........ */ + dst[0] = buf[0] << 3 | buf[1] >> 2; + + /* .....111 11...... ........ ........ ........ */ + /* ........ ..22222. ........ ........ ........ */ + /* ........ .......3 3333.... ........ ........ */ + dst[1] = buf[1] << 6 | buf[2] << 1 | buf[3] >> 4; + + /* ........ .......3 3333.... ........ ........ */ + /* ........ ........ ....4444 4....... ........ */ + dst[2] = buf[3] << 4 | buf[4] >> 1; + + /* ........ ........ ....4444 4....... ........ */ + /* ........ ........ ........ .55555.. ........ */ + /* ........ ........ ........ ......66 666..... */ + dst[3] = buf[4] << 7 | buf[5] << 2 | buf[6] >> 3; + + /* ........ ........ ........ ......66 666..... */ + /* ........ ........ ........ ........ ...77777 */ + dst[4] = buf[6] << 5 | buf[7]; + + dst += 5; + dst_sz -= 5; + } + /* Not ending on a eight byte boundary? */ + if (i > 0 && i < 8) { + + /* Enough space available at the destination? */ + if (dst_sz < (i + 1) / 2) + return -1; + + switch (i) { + case 7: /* ........ ........ ........ ......66 666..... */ + /* ........ ........ ........ .55555.. ........ */ + /* ........ ........ ....4444 4....... ........ */ + dst[3] = buf[4] << 7 | buf[5] << 2 | buf[6] >> 3; + + case 5: /* ........ ........ ....4444 4....... ........ */ + /* ........ .......3 3333.... ........ ........ */ + dst[2] = buf[3] << 4 | buf[4] >> 1; + + case 4: /* ........ .......3 3333.... ........ ........ */ + /* ........ ..22222. ........ ........ ........ */ + /* .....111 11...... ........ ........ ........ */ + dst[1] = buf[1] << 6 | buf[2] << 1 | buf[3] >> 4; + + case 2: /* .....111 11...... ........ ........ ........ */ + /* 00000... ........ ........ ........ ........ */ + dst[0] = buf[0] << 3 | buf[1] >> 2; + + break; + + default: + return -1; + } + dst += (i + 1) / 2; + + if (check_padding) { + /* Check remaining padding characters */ + if (ch != '=') + return -1; + + /* One down, 8 - i - 1 more to come... */ + for (i = 8 - i - 1; i > 0; i--) { + + do { + if (src_sz == 0) + return -1; + ch = *src++; + src_sz--; + + } while (isspace(ch)); + + if (ch != '=') + return -1; + } + } + } + return dst - start; +} + +int +sldns_b32_pton(const char* src, size_t src_sz, uint8_t* dst, size_t dst_sz) +{ + return sldns_b32_pton_base(src, src_sz, dst, dst_sz, 0, 1); +} + +int +sldns_b32_pton_extended_hex(const char* src, size_t src_sz, + uint8_t* dst, size_t dst_sz) +{ + return sldns_b32_pton_base(src, src_sz, dst, dst_sz, 1, 1); +} + +size_t sldns_b64_ntop_calculate_size(size_t srcsize) +{ + return ((((srcsize + 2) / 3) * 4) + 1); +} + +/* RFC 1521, section 5.2. + * + * The encoding process represents 24-bit groups of input bits as output + * strings of 4 encoded characters. Proceeding from left to right, a + * 24-bit input group is formed by concatenating 3 8-bit input groups. + * These 24 bits are then treated as 4 concatenated 6-bit groups, each + * of which is translated into a single digit in the base64 alphabet. + * + * This routine does not insert spaces or linebreaks after 76 characters. + */ +int sldns_b64_ntop(uint8_t const *src, size_t srclength, + char *target, size_t targsize) +{ + const char* b64 = + "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/"; + const char pad64 = '='; + size_t i = 0, o = 0; + if(targsize < sldns_b64_ntop_calculate_size(srclength)) + return -1; + /* whole chunks: xxxxxxyy yyyyzzzz zzwwwwww */ + while(i+3 <= srclength) { + if(o+4 > targsize) return -1; + target[o] = b64[src[i] >> 2]; + target[o+1] = b64[ ((src[i]&0x03)<<4) | (src[i+1]>>4) ]; + target[o+2] = b64[ ((src[i+1]&0x0f)<<2) | (src[i+2]>>6) ]; + target[o+3] = b64[ (src[i+2]&0x3f) ]; + i += 3; + o += 4; + } + /* remainder */ + switch(srclength - i) { + case 2: + /* two at end, converted into A B C = */ + target[o] = b64[src[i] >> 2]; + target[o+1] = b64[ ((src[i]&0x03)<<4) | (src[i+1]>>4) ]; + target[o+2] = b64[ ((src[i+1]&0x0f)<<2) ]; + target[o+3] = pad64; + i += 2; + o += 4; + break; + case 1: + /* one at end, converted into A B = = */ + target[o] = b64[src[i] >> 2]; + target[o+1] = b64[ ((src[i]&0x03)<<4) ]; + target[o+2] = pad64; + target[o+3] = pad64; + i += 1; + o += 4; + break; + case 0: + default: + /* nothing */ + break; + } + /* assert: i == srclength */ + if(o+1 > targsize) return -1; + target[o] = 0; + return (int)o; +} + +size_t sldns_b64_pton_calculate_size(size_t srcsize) +{ + return (((((srcsize + 3) / 4) * 3)) + 1); +} + +int sldns_b64_pton(char const *src, uint8_t *target, size_t targsize) +{ + const uint8_t pad64 = 64; /* is 64th in the b64 array */ + const char* s = src; + uint8_t in[4]; + size_t o = 0, incount = 0; + + while(*s) { + /* skip any character that is not base64 */ + /* conceptually we do: + const char* b64 = pad'=' is appended to array + "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/="; + const char* d = strchr(b64, *s++); + and use d-b64; + */ + char d = *s++; + if(d <= 'Z' && d >= 'A') + d -= 'A'; + else if(d <= 'z' && d >= 'a') + d = d - 'a' + 26; + else if(d <= '9' && d >= '0') + d = d - '0' + 52; + else if(d == '+') + d = 62; + else if(d == '/') + d = 63; + else if(d == '=') + d = 64; + else continue; + in[incount++] = (uint8_t)d; + if(incount != 4) + continue; + /* process whole block of 4 characters into 3 output bytes */ + if(in[3] == pad64 && in[2] == pad64) { /* A B = = */ + if(o+1 > targsize) + return -1; + target[o] = (in[0]<<2) | ((in[1]&0x30)>>4); + o += 1; + break; /* we are done */ + } else if(in[3] == pad64) { /* A B C = */ + if(o+2 > targsize) + return -1; + target[o] = (in[0]<<2) | ((in[1]&0x30)>>4); + target[o+1]= ((in[1]&0x0f)<<4) | ((in[2]&0x3c)>>2); + o += 2; + break; /* we are done */ + } else { + if(o+3 > targsize) + return -1; + /* write xxxxxxyy yyyyzzzz zzwwwwww */ + target[o] = (in[0]<<2) | ((in[1]&0x30)>>4); + target[o+1]= ((in[1]&0x0f)<<4) | ((in[2]&0x3c)>>2); + target[o+2]= ((in[2]&0x03)<<6) | in[3]; + o += 3; + } + incount = 0; + } + return (int)o; +} |