#include "mongoose.h" #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_internal.h" #endif /* * Copyright (c) 2014 Cesanta Software Limited * All rights reserved */ #ifndef CS_MONGOOSE_SRC_INTERNAL_H_ #define CS_MONGOOSE_SRC_INTERNAL_H_ /* Amalgamated: #include "common/mg_mem.h" */ #ifndef MBUF_REALLOC #define MBUF_REALLOC MG_REALLOC #endif #ifndef MBUF_FREE #define MBUF_FREE MG_FREE #endif #define MG_SET_PTRPTR(_ptr, _v) \ do { \ if (_ptr) *(_ptr) = _v; \ } while (0) #ifndef MG_INTERNAL #define MG_INTERNAL static #endif #ifdef PICOTCP #define NO_LIBC #define MG_DISABLE_PFS #endif /* Amalgamated: #include "common/cs_dbg.h" */ /* Amalgamated: #include "mg_http.h" */ /* Amalgamated: #include "mg_net.h" */ #define MG_CTL_MSG_MESSAGE_SIZE 8192 /* internals that need to be accessible in unit tests */ MG_INTERNAL struct mg_connection *mg_do_connect(struct mg_connection *nc, int proto, union socket_address *sa); MG_INTERNAL int mg_parse_address(const char *str, union socket_address *sa, int *proto, char *host, size_t host_len); MG_INTERNAL void mg_call(struct mg_connection *nc, mg_event_handler_t ev_handler, void *user_data, int ev, void *ev_data); void mg_forward(struct mg_connection *from, struct mg_connection *to); MG_INTERNAL void mg_add_conn(struct mg_mgr *mgr, struct mg_connection *c); MG_INTERNAL void mg_remove_conn(struct mg_connection *c); MG_INTERNAL struct mg_connection *mg_create_connection( struct mg_mgr *mgr, mg_event_handler_t callback, struct mg_add_sock_opts opts); #ifdef _WIN32 /* Retur value is the same as for MultiByteToWideChar. */ int to_wchar(const char *path, wchar_t *wbuf, size_t wbuf_len); #endif struct ctl_msg { mg_event_handler_t callback; char message[MG_CTL_MSG_MESSAGE_SIZE]; }; #if MG_ENABLE_MQTT struct mg_mqtt_message; #define MG_MQTT_ERROR_INCOMPLETE_MSG -1 #define MG_MQTT_ERROR_MALFORMED_MSG -2 MG_INTERNAL int parse_mqtt(struct mbuf *io, struct mg_mqtt_message *mm); #endif /* Forward declarations for testing. */ extern void *(*test_malloc)(size_t size); extern void *(*test_calloc)(size_t count, size_t size); #ifndef MIN #define MIN(a, b) ((a) < (b) ? (a) : (b)) #endif #if MG_ENABLE_HTTP struct mg_serve_http_opts; /* * Reassemble the content of the buffer (buf, blen) which should be * in the HTTP chunked encoding, by collapsing data chunks to the * beginning of the buffer. * * If chunks get reassembled, modify hm->body to point to the reassembled * body and fire MG_EV_HTTP_CHUNK event. If handler sets MG_F_DELETE_CHUNK * in nc->flags, delete reassembled body from the mbuf. * * Return reassembled body size. */ MG_INTERNAL size_t mg_handle_chunked(struct mg_connection *nc, struct http_message *hm, char *buf, size_t blen); #if MG_ENABLE_FILESYSTEM MG_INTERNAL int mg_uri_to_local_path(struct http_message *hm, const struct mg_serve_http_opts *opts, char **local_path, struct mg_str *remainder); MG_INTERNAL time_t mg_parse_date_string(const char *datetime); MG_INTERNAL int mg_is_not_modified(struct http_message *hm, cs_stat_t *st); #endif #if MG_ENABLE_HTTP_CGI MG_INTERNAL void mg_handle_cgi(struct mg_connection *nc, const char *prog, const struct mg_str *path_info, const struct http_message *hm, const struct mg_serve_http_opts *opts); struct mg_http_proto_data_cgi; MG_INTERNAL void mg_http_free_proto_data_cgi(struct mg_http_proto_data_cgi *d); #endif #if MG_ENABLE_HTTP_SSI MG_INTERNAL void mg_handle_ssi_request(struct mg_connection *nc, struct http_message *hm, const char *path, const struct mg_serve_http_opts *opts); #endif #if MG_ENABLE_HTTP_WEBDAV MG_INTERNAL int mg_is_dav_request(const struct mg_str *s); MG_INTERNAL void mg_handle_propfind(struct mg_connection *nc, const char *path, cs_stat_t *stp, struct http_message *hm, struct mg_serve_http_opts *opts); MG_INTERNAL void mg_handle_lock(struct mg_connection *nc, const char *path); MG_INTERNAL void mg_handle_mkcol(struct mg_connection *nc, const char *path, struct http_message *hm); MG_INTERNAL void mg_handle_move(struct mg_connection *c, const struct mg_serve_http_opts *opts, const char *path, struct http_message *hm); MG_INTERNAL void mg_handle_delete(struct mg_connection *nc, const struct mg_serve_http_opts *opts, const char *path); MG_INTERNAL void mg_handle_put(struct mg_connection *nc, const char *path, struct http_message *hm); #endif #if MG_ENABLE_HTTP_WEBSOCKET MG_INTERNAL void mg_ws_handler(struct mg_connection *nc, int ev, void *ev_data MG_UD_ARG(void *user_data)); MG_INTERNAL void mg_ws_handshake(struct mg_connection *nc, const struct mg_str *key, struct http_message *); #endif #endif /* MG_ENABLE_HTTP */ MG_INTERNAL int mg_get_errno(void); MG_INTERNAL void mg_close_conn(struct mg_connection *conn); #if MG_ENABLE_SNTP MG_INTERNAL int mg_sntp_parse_reply(const char *buf, int len, struct mg_sntp_message *msg); #endif #endif /* CS_MONGOOSE_SRC_INTERNAL_H_ */ #ifdef MG_MODULE_LINES #line 1 "common/mg_mem.h" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #ifndef CS_COMMON_MG_MEM_H_ #define CS_COMMON_MG_MEM_H_ #ifdef __cplusplus extern "C" { #endif #ifndef MG_MALLOC #define MG_MALLOC malloc #endif #ifndef MG_CALLOC #define MG_CALLOC calloc #endif #ifndef MG_REALLOC #define MG_REALLOC realloc #endif #ifndef MG_FREE #define MG_FREE free #endif #ifdef __cplusplus } #endif #endif /* CS_COMMON_MG_MEM_H_ */ #ifdef MG_MODULE_LINES #line 1 "common/cs_base64.c" #endif /* * Copyright (c) 2014 Cesanta Software Limited * All rights reserved */ #ifndef EXCLUDE_COMMON /* Amalgamated: #include "common/cs_base64.h" */ #include /* Amalgamated: #include "common/cs_dbg.h" */ /* ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/ */ #define NUM_UPPERCASES ('Z' - 'A' + 1) #define NUM_LETTERS (NUM_UPPERCASES * 2) #define NUM_DIGITS ('9' - '0' + 1) /* * Emit a base64 code char. * * Doesn't use memory, thus it's safe to use to safely dump memory in crashdumps */ static void cs_base64_emit_code(struct cs_base64_ctx *ctx, int v) { if (v < NUM_UPPERCASES) { ctx->b64_putc(v + 'A', ctx->user_data); } else if (v < (NUM_LETTERS)) { ctx->b64_putc(v - NUM_UPPERCASES + 'a', ctx->user_data); } else if (v < (NUM_LETTERS + NUM_DIGITS)) { ctx->b64_putc(v - NUM_LETTERS + '0', ctx->user_data); } else { ctx->b64_putc(v - NUM_LETTERS - NUM_DIGITS == 0 ? '+' : '/', ctx->user_data); } } static void cs_base64_emit_chunk(struct cs_base64_ctx *ctx) { int a, b, c; a = ctx->chunk[0]; b = ctx->chunk[1]; c = ctx->chunk[2]; cs_base64_emit_code(ctx, a >> 2); cs_base64_emit_code(ctx, ((a & 3) << 4) | (b >> 4)); if (ctx->chunk_size > 1) { cs_base64_emit_code(ctx, (b & 15) << 2 | (c >> 6)); } if (ctx->chunk_size > 2) { cs_base64_emit_code(ctx, c & 63); } } void cs_base64_init(struct cs_base64_ctx *ctx, cs_base64_putc_t b64_putc, void *user_data) { ctx->chunk_size = 0; ctx->b64_putc = b64_putc; ctx->user_data = user_data; } void cs_base64_update(struct cs_base64_ctx *ctx, const char *str, size_t len) { const unsigned char *src = (const unsigned char *) str; size_t i; for (i = 0; i < len; i++) { ctx->chunk[ctx->chunk_size++] = src[i]; if (ctx->chunk_size == 3) { cs_base64_emit_chunk(ctx); ctx->chunk_size = 0; } } } void cs_base64_finish(struct cs_base64_ctx *ctx) { if (ctx->chunk_size > 0) { int i; memset(&ctx->chunk[ctx->chunk_size], 0, 3 - ctx->chunk_size); cs_base64_emit_chunk(ctx); for (i = 0; i < (3 - ctx->chunk_size); i++) { ctx->b64_putc('=', ctx->user_data); } } } #define BASE64_ENCODE_BODY \ static const char *b64 = \ "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/"; \ int i, j, a, b, c; \ \ for (i = j = 0; i < src_len; i += 3) { \ a = src[i]; \ b = i + 1 >= src_len ? 0 : src[i + 1]; \ c = i + 2 >= src_len ? 0 : src[i + 2]; \ \ BASE64_OUT(b64[a >> 2]); \ BASE64_OUT(b64[((a & 3) << 4) | (b >> 4)]); \ if (i + 1 < src_len) { \ BASE64_OUT(b64[(b & 15) << 2 | (c >> 6)]); \ } \ if (i + 2 < src_len) { \ BASE64_OUT(b64[c & 63]); \ } \ } \ \ while (j % 4 != 0) { \ BASE64_OUT('='); \ } \ BASE64_FLUSH() #define BASE64_OUT(ch) \ do { \ dst[j++] = (ch); \ } while (0) #define BASE64_FLUSH() \ do { \ dst[j++] = '\0'; \ } while (0) void cs_base64_encode(const unsigned char *src, int src_len, char *dst) { BASE64_ENCODE_BODY; } #undef BASE64_OUT #undef BASE64_FLUSH #if CS_ENABLE_STDIO #define BASE64_OUT(ch) \ do { \ fprintf(f, "%c", (ch)); \ j++; \ } while (0) #define BASE64_FLUSH() void cs_fprint_base64(FILE *f, const unsigned char *src, int src_len) { BASE64_ENCODE_BODY; } #undef BASE64_OUT #undef BASE64_FLUSH #endif /* CS_ENABLE_STDIO */ /* Convert one byte of encoded base64 input stream to 6-bit chunk */ static unsigned char from_b64(unsigned char ch) { /* Inverse lookup map */ static const unsigned char tab[128] = { 255, 255, 255, 255, 255, 255, 255, 255, /* 0 */ 255, 255, 255, 255, 255, 255, 255, 255, /* 8 */ 255, 255, 255, 255, 255, 255, 255, 255, /* 16 */ 255, 255, 255, 255, 255, 255, 255, 255, /* 24 */ 255, 255, 255, 255, 255, 255, 255, 255, /* 32 */ 255, 255, 255, 62, 255, 255, 255, 63, /* 40 */ 52, 53, 54, 55, 56, 57, 58, 59, /* 48 */ 60, 61, 255, 255, 255, 200, 255, 255, /* 56 '=' is 200, on index 61 */ 255, 0, 1, 2, 3, 4, 5, 6, /* 64 */ 7, 8, 9, 10, 11, 12, 13, 14, /* 72 */ 15, 16, 17, 18, 19, 20, 21, 22, /* 80 */ 23, 24, 25, 255, 255, 255, 255, 255, /* 88 */ 255, 26, 27, 28, 29, 30, 31, 32, /* 96 */ 33, 34, 35, 36, 37, 38, 39, 40, /* 104 */ 41, 42, 43, 44, 45, 46, 47, 48, /* 112 */ 49, 50, 51, 255, 255, 255, 255, 255, /* 120 */ }; return tab[ch & 127]; } int cs_base64_decode(const unsigned char *s, int len, char *dst, int *dec_len) { unsigned char a, b, c, d; int orig_len = len; char *orig_dst = dst; while (len >= 4 && (a = from_b64(s[0])) != 255 && (b = from_b64(s[1])) != 255 && (c = from_b64(s[2])) != 255 && (d = from_b64(s[3])) != 255) { s += 4; len -= 4; if (a == 200 || b == 200) break; /* '=' can't be there */ *dst++ = a << 2 | b >> 4; if (c == 200) break; *dst++ = b << 4 | c >> 2; if (d == 200) break; *dst++ = c << 6 | d; } *dst = 0; if (dec_len != NULL) *dec_len = (dst - orig_dst); return orig_len - len; } #endif /* EXCLUDE_COMMON */ #ifdef MG_MODULE_LINES #line 1 "common/cs_dbg.h" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #ifndef CS_COMMON_CS_DBG_H_ #define CS_COMMON_CS_DBG_H_ /* Amalgamated: #include "common/platform.h" */ #if CS_ENABLE_STDIO #include #endif #ifndef CS_ENABLE_DEBUG #define CS_ENABLE_DEBUG 0 #endif #ifndef CS_LOG_ENABLE_TS_DIFF #define CS_LOG_ENABLE_TS_DIFF 0 #endif #ifdef __cplusplus extern "C" { #endif /* __cplusplus */ /* * Log level; `LL_INFO` is the default. Use `cs_log_set_level()` to change it. */ enum cs_log_level { LL_NONE = -1, LL_ERROR = 0, LL_WARN = 1, LL_INFO = 2, LL_DEBUG = 3, LL_VERBOSE_DEBUG = 4, _LL_MIN = -2, _LL_MAX = 5, }; /* * Set max log level to print; messages with the level above the given one will * not be printed. */ void cs_log_set_level(enum cs_log_level level); /* * Set log filter. NULL (a default) logs everything. * Otherwise, function name and file name will be tested against the given * pattern, and only matching messages will be printed. * * For the pattern syntax, refer to `mg_match_prefix()` in `str_util.h`. * * Example: * ```c * void foo(void) { * LOG(LL_INFO, ("hello from foo")); * } * * void bar(void) { * LOG(LL_INFO, ("hello from bar")); * } * * void test(void) { * cs_log_set_filter(NULL); * foo(); * bar(); * * cs_log_set_filter("f*"); * foo(); * bar(); // Will NOT print anything * * cs_log_set_filter("bar"); * foo(); // Will NOT print anything * bar(); * } * ``` */ void cs_log_set_filter(const char *pattern); /* * Helper function which prints message prefix with the given `level`, function * name `func` and `filename`. If message should be printed (accordingly to the * current log level and filter), prints the prefix and returns 1, otherwise * returns 0. * * Clients should typically just use `LOG()` macro. */ int cs_log_print_prefix(enum cs_log_level level, const char *func, const char *filename); extern enum cs_log_level cs_log_threshold; #if CS_ENABLE_STDIO /* * Set file to write logs into. If `NULL`, logs go to `stderr`. */ void cs_log_set_file(FILE *file); /* * Prints log to the current log file, appends "\n" in the end and flushes the * stream. */ void cs_log_printf(const char *fmt, ...) #ifdef __GNUC__ __attribute__((format(printf, 1, 2))) #endif ; /* * Format and print message `x` with the given level `l`. Example: * * ```c * LOG(LL_INFO, ("my info message: %d", 123)); * LOG(LL_DEBUG, ("my debug message: %d", 123)); * ``` */ #define LOG(l, x) \ do { \ if (cs_log_print_prefix(l, __func__, __FILE__)) cs_log_printf x; \ } while (0) #ifndef CS_NDEBUG /* * Shortcut for `LOG(LL_VERBOSE_DEBUG, (...))` */ #define DBG(x) LOG(LL_VERBOSE_DEBUG, x) #else /* NDEBUG */ #define DBG(x) #endif #else /* CS_ENABLE_STDIO */ #define LOG(l, x) #define DBG(x) #endif #ifdef __cplusplus } #endif /* __cplusplus */ #endif /* CS_COMMON_CS_DBG_H_ */ #ifdef MG_MODULE_LINES #line 1 "common/cs_dbg.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ /* Amalgamated: #include "common/cs_dbg.h" */ #include #include #include /* Amalgamated: #include "common/cs_time.h" */ /* Amalgamated: #include "common/str_util.h" */ enum cs_log_level cs_log_threshold WEAK = #if CS_ENABLE_DEBUG LL_VERBOSE_DEBUG; #else LL_ERROR; #endif static char *s_filter_pattern = NULL; static size_t s_filter_pattern_len; void cs_log_set_filter(const char *pattern) WEAK; #if CS_ENABLE_STDIO FILE *cs_log_file WEAK = NULL; #if CS_LOG_ENABLE_TS_DIFF double cs_log_ts WEAK; #endif enum cs_log_level cs_log_cur_msg_level WEAK = LL_NONE; void cs_log_set_filter(const char *pattern) { free(s_filter_pattern); if (pattern != NULL) { s_filter_pattern = strdup(pattern); s_filter_pattern_len = strlen(pattern); } else { s_filter_pattern = NULL; s_filter_pattern_len = 0; } } int cs_log_print_prefix(enum cs_log_level, const char *, const char *) WEAK; int cs_log_print_prefix(enum cs_log_level level, const char *func, const char *filename) { char prefix[21]; if (level > cs_log_threshold) return 0; if (s_filter_pattern != NULL && mg_match_prefix(s_filter_pattern, s_filter_pattern_len, func) == 0 && mg_match_prefix(s_filter_pattern, s_filter_pattern_len, filename) == 0) { return 0; } strncpy(prefix, func, 20); prefix[20] = '\0'; if (cs_log_file == NULL) cs_log_file = stderr; cs_log_cur_msg_level = level; fprintf(cs_log_file, "%-20s ", prefix); #if CS_LOG_ENABLE_TS_DIFF { double now = cs_time(); fprintf(cs_log_file, "%7u ", (unsigned int) ((now - cs_log_ts) * 1000000)); cs_log_ts = now; } #endif return 1; } void cs_log_printf(const char *fmt, ...) WEAK; void cs_log_printf(const char *fmt, ...) { va_list ap; va_start(ap, fmt); vfprintf(cs_log_file, fmt, ap); va_end(ap); fputc('\n', cs_log_file); fflush(cs_log_file); cs_log_cur_msg_level = LL_NONE; } void cs_log_set_file(FILE *file) WEAK; void cs_log_set_file(FILE *file) { cs_log_file = file; } #else void cs_log_set_filter(const char *pattern) { (void) pattern; } #endif /* CS_ENABLE_STDIO */ void cs_log_set_level(enum cs_log_level level) WEAK; void cs_log_set_level(enum cs_log_level level) { cs_log_threshold = level; #if CS_LOG_ENABLE_TS_DIFF && CS_ENABLE_STDIO cs_log_ts = cs_time(); #endif } #ifdef MG_MODULE_LINES #line 1 "common/cs_dirent.h" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #ifndef CS_COMMON_CS_DIRENT_H_ #define CS_COMMON_CS_DIRENT_H_ #include /* Amalgamated: #include "common/platform.h" */ #ifdef __cplusplus extern "C" { #endif /* __cplusplus */ #ifdef CS_DEFINE_DIRENT typedef struct { int dummy; } DIR; struct dirent { int d_ino; #ifdef _WIN32 char d_name[MAX_PATH]; #else /* TODO(rojer): Use PATH_MAX but make sure it's sane on every platform */ char d_name[256]; #endif }; DIR *opendir(const char *dir_name); int closedir(DIR *dir); struct dirent *readdir(DIR *dir); #endif /* CS_DEFINE_DIRENT */ #ifdef __cplusplus } #endif /* __cplusplus */ #endif /* CS_COMMON_CS_DIRENT_H_ */ #ifdef MG_MODULE_LINES #line 1 "common/cs_dirent.c" #endif /* * Copyright (c) 2015 Cesanta Software Limited * All rights reserved */ #ifndef EXCLUDE_COMMON /* Amalgamated: #include "common/mg_mem.h" */ /* Amalgamated: #include "common/cs_dirent.h" */ /* * This file contains POSIX opendir/closedir/readdir API implementation * for systems which do not natively support it (e.g. Windows). */ #ifdef _WIN32 struct win32_dir { DIR d; HANDLE handle; WIN32_FIND_DATAW info; struct dirent result; }; DIR *opendir(const char *name) { struct win32_dir *dir = NULL; wchar_t wpath[MAX_PATH]; DWORD attrs; if (name == NULL) { SetLastError(ERROR_BAD_ARGUMENTS); } else if ((dir = (struct win32_dir *) MG_MALLOC(sizeof(*dir))) == NULL) { SetLastError(ERROR_NOT_ENOUGH_MEMORY); } else { to_wchar(name, wpath, ARRAY_SIZE(wpath)); attrs = GetFileAttributesW(wpath); if (attrs != 0xFFFFFFFF && (attrs & FILE_ATTRIBUTE_DIRECTORY)) { (void) wcscat(wpath, L"\\*"); dir->handle = FindFirstFileW(wpath, &dir->info); dir->result.d_name[0] = '\0'; } else { MG_FREE(dir); dir = NULL; } } return (DIR *) dir; } int closedir(DIR *d) { struct win32_dir *dir = (struct win32_dir *) d; int result = 0; if (dir != NULL) { if (dir->handle != INVALID_HANDLE_VALUE) result = FindClose(dir->handle) ? 0 : -1; MG_FREE(dir); } else { result = -1; SetLastError(ERROR_BAD_ARGUMENTS); } return result; } struct dirent *readdir(DIR *d) { struct win32_dir *dir = (struct win32_dir *) d; struct dirent *result = NULL; if (dir) { memset(&dir->result, 0, sizeof(dir->result)); if (dir->handle != INVALID_HANDLE_VALUE) { result = &dir->result; (void) WideCharToMultiByte(CP_UTF8, 0, dir->info.cFileName, -1, result->d_name, sizeof(result->d_name), NULL, NULL); if (!FindNextFileW(dir->handle, &dir->info)) { (void) FindClose(dir->handle); dir->handle = INVALID_HANDLE_VALUE; } } else { SetLastError(ERROR_FILE_NOT_FOUND); } } else { SetLastError(ERROR_BAD_ARGUMENTS); } return result; } #endif #endif /* EXCLUDE_COMMON */ /* ISO C requires a translation unit to contain at least one declaration */ typedef int cs_dirent_dummy; #ifdef MG_MODULE_LINES #line 1 "common/cs_time.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ /* Amalgamated: #include "common/cs_time.h" */ #ifndef _WIN32 #include /* * There is no sys/time.h on ARMCC. */ #if !(defined(__ARMCC_VERSION) || defined(__ICCARM__)) && \ !defined(__TI_COMPILER_VERSION__) && \ (!defined(CS_PLATFORM) || CS_PLATFORM != CS_P_NXP_LPC) #include #endif #else #include #endif double cs_time(void) WEAK; double cs_time(void) { double now; #ifndef _WIN32 struct timeval tv; if (gettimeofday(&tv, NULL /* tz */) != 0) return 0; now = (double) tv.tv_sec + (((double) tv.tv_usec) / 1000000.0); #else SYSTEMTIME sysnow; FILETIME ftime; GetLocalTime(&sysnow); SystemTimeToFileTime(&sysnow, &ftime); /* * 1. VC 6.0 doesn't support conversion uint64 -> double, so, using int64 * This should not cause a problems in this (21th) century * 2. Windows FILETIME is a number of 100-nanosecond intervals since January * 1, 1601 while time_t is a number of _seconds_ since January 1, 1970 UTC, * thus, we need to convert to seconds and adjust amount (subtract 11644473600 * seconds) */ now = (double) (((int64_t) ftime.dwLowDateTime + ((int64_t) ftime.dwHighDateTime << 32)) / 10000000.0) - 11644473600; #endif /* _WIN32 */ return now; } double cs_timegm(const struct tm *tm) { /* Month-to-day offset for non-leap-years. */ static const int month_day[12] = {0, 31, 59, 90, 120, 151, 181, 212, 243, 273, 304, 334}; /* Most of the calculation is easy; leap years are the main difficulty. */ int month = tm->tm_mon % 12; int year = tm->tm_year + tm->tm_mon / 12; int year_for_leap; int64_t rt; if (month < 0) { /* Negative values % 12 are still negative. */ month += 12; --year; } /* This is the number of Februaries since 1900. */ year_for_leap = (month > 1) ? year + 1 : year; rt = tm->tm_sec /* Seconds */ + 60 * (tm->tm_min /* Minute = 60 seconds */ + 60 * (tm->tm_hour /* Hour = 60 minutes */ + 24 * (month_day[month] + tm->tm_mday - 1 /* Day = 24 hours */ + 365 * (year - 70) /* Year = 365 days */ + (year_for_leap - 69) / 4 /* Every 4 years is leap... */ - (year_for_leap - 1) / 100 /* Except centuries... */ + (year_for_leap + 299) / 400))); /* Except 400s. */ return rt < 0 ? -1 : (double) rt; } #ifdef MG_MODULE_LINES #line 1 "common/cs_endian.h" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #ifndef CS_COMMON_CS_ENDIAN_H_ #define CS_COMMON_CS_ENDIAN_H_ #ifdef __cplusplus extern "C" { #endif /* * clang with std=-c99 uses __LITTLE_ENDIAN, by default * while for ex, RTOS gcc - LITTLE_ENDIAN, by default * it depends on __USE_BSD, but let's have everything */ #if !defined(BYTE_ORDER) && defined(__BYTE_ORDER) #define BYTE_ORDER __BYTE_ORDER #ifndef LITTLE_ENDIAN #define LITTLE_ENDIAN __LITTLE_ENDIAN #endif /* LITTLE_ENDIAN */ #ifndef BIG_ENDIAN #define BIG_ENDIAN __LITTLE_ENDIAN #endif /* BIG_ENDIAN */ #endif /* BYTE_ORDER */ #ifdef __cplusplus } #endif #endif /* CS_COMMON_CS_ENDIAN_H_ */ #ifdef MG_MODULE_LINES #line 1 "common/cs_md5.c" #endif /* * This code implements the MD5 message-digest algorithm. * The algorithm is due to Ron Rivest. This code was * written by Colin Plumb in 1993, no copyright is claimed. * This code is in the public domain; do with it what you wish. * * Equivalent code is available from RSA Data Security, Inc. * This code has been tested against that, and is equivalent, * except that you don't need to include two pages of legalese * with every copy. * * To compute the message digest of a chunk of bytes, declare an * MD5Context structure, pass it to MD5Init, call MD5Update as * needed on buffers full of bytes, and then call MD5Final, which * will fill a supplied 16-byte array with the digest. */ /* Amalgamated: #include "common/cs_md5.h" */ /* Amalgamated: #include "common/str_util.h" */ #if !defined(EXCLUDE_COMMON) #if !CS_DISABLE_MD5 /* Amalgamated: #include "common/cs_endian.h" */ static void byteReverse(unsigned char *buf, unsigned longs) { /* Forrest: MD5 expect LITTLE_ENDIAN, swap if BIG_ENDIAN */ #if BYTE_ORDER == BIG_ENDIAN do { uint32_t t = (uint32_t)((unsigned) buf[3] << 8 | buf[2]) << 16 | ((unsigned) buf[1] << 8 | buf[0]); *(uint32_t *) buf = t; buf += 4; } while (--longs); #else (void) buf; (void) longs; #endif } #define F1(x, y, z) (z ^ (x & (y ^ z))) #define F2(x, y, z) F1(z, x, y) #define F3(x, y, z) (x ^ y ^ z) #define F4(x, y, z) (y ^ (x | ~z)) #define MD5STEP(f, w, x, y, z, data, s) \ (w += f(x, y, z) + data, w = w << s | w >> (32 - s), w += x) /* * Start MD5 accumulation. Set bit count to 0 and buffer to mysterious * initialization constants. */ void cs_md5_init(cs_md5_ctx *ctx) { ctx->buf[0] = 0x67452301; ctx->buf[1] = 0xefcdab89; ctx->buf[2] = 0x98badcfe; ctx->buf[3] = 0x10325476; ctx->bits[0] = 0; ctx->bits[1] = 0; } static void cs_md5_transform(uint32_t buf[4], uint32_t const in[16]) { register uint32_t a, b, c, d; a = buf[0]; b = buf[1]; c = buf[2]; d = buf[3]; MD5STEP(F1, a, b, c, d, in[0] + 0xd76aa478, 7); MD5STEP(F1, d, a, b, c, in[1] + 0xe8c7b756, 12); MD5STEP(F1, c, d, a, b, in[2] + 0x242070db, 17); MD5STEP(F1, b, c, d, a, in[3] + 0xc1bdceee, 22); MD5STEP(F1, a, b, c, d, in[4] + 0xf57c0faf, 7); MD5STEP(F1, d, a, b, c, in[5] + 0x4787c62a, 12); MD5STEP(F1, c, d, a, b, in[6] + 0xa8304613, 17); MD5STEP(F1, b, c, d, a, in[7] + 0xfd469501, 22); MD5STEP(F1, a, b, c, d, in[8] + 0x698098d8, 7); MD5STEP(F1, d, a, b, c, in[9] + 0x8b44f7af, 12); MD5STEP(F1, c, d, a, b, in[10] + 0xffff5bb1, 17); MD5STEP(F1, b, c, d, a, in[11] + 0x895cd7be, 22); MD5STEP(F1, a, b, c, d, in[12] + 0x6b901122, 7); MD5STEP(F1, d, a, b, c, in[13] + 0xfd987193, 12); MD5STEP(F1, c, d, a, b, in[14] + 0xa679438e, 17); MD5STEP(F1, b, c, d, a, in[15] + 0x49b40821, 22); MD5STEP(F2, a, b, c, d, in[1] + 0xf61e2562, 5); MD5STEP(F2, d, a, b, c, in[6] + 0xc040b340, 9); MD5STEP(F2, c, d, a, b, in[11] + 0x265e5a51, 14); MD5STEP(F2, b, c, d, a, in[0] + 0xe9b6c7aa, 20); MD5STEP(F2, a, b, c, d, in[5] + 0xd62f105d, 5); MD5STEP(F2, d, a, b, c, in[10] + 0x02441453, 9); MD5STEP(F2, c, d, a, b, in[15] + 0xd8a1e681, 14); MD5STEP(F2, b, c, d, a, in[4] + 0xe7d3fbc8, 20); MD5STEP(F2, a, b, c, d, in[9] + 0x21e1cde6, 5); MD5STEP(F2, d, a, b, c, in[14] + 0xc33707d6, 9); MD5STEP(F2, c, d, a, b, in[3] + 0xf4d50d87, 14); MD5STEP(F2, b, c, d, a, in[8] + 0x455a14ed, 20); MD5STEP(F2, a, b, c, d, in[13] + 0xa9e3e905, 5); MD5STEP(F2, d, a, b, c, in[2] + 0xfcefa3f8, 9); MD5STEP(F2, c, d, a, b, in[7] + 0x676f02d9, 14); MD5STEP(F2, b, c, d, a, in[12] + 0x8d2a4c8a, 20); MD5STEP(F3, a, b, c, d, in[5] + 0xfffa3942, 4); MD5STEP(F3, d, a, b, c, in[8] + 0x8771f681, 11); MD5STEP(F3, c, d, a, b, in[11] + 0x6d9d6122, 16); MD5STEP(F3, b, c, d, a, in[14] + 0xfde5380c, 23); MD5STEP(F3, a, b, c, d, in[1] + 0xa4beea44, 4); MD5STEP(F3, d, a, b, c, in[4] + 0x4bdecfa9, 11); MD5STEP(F3, c, d, a, b, in[7] + 0xf6bb4b60, 16); MD5STEP(F3, b, c, d, a, in[10] + 0xbebfbc70, 23); MD5STEP(F3, a, b, c, d, in[13] + 0x289b7ec6, 4); MD5STEP(F3, d, a, b, c, in[0] + 0xeaa127fa, 11); MD5STEP(F3, c, d, a, b, in[3] + 0xd4ef3085, 16); MD5STEP(F3, b, c, d, a, in[6] + 0x04881d05, 23); MD5STEP(F3, a, b, c, d, in[9] + 0xd9d4d039, 4); MD5STEP(F3, d, a, b, c, in[12] + 0xe6db99e5, 11); MD5STEP(F3, c, d, a, b, in[15] + 0x1fa27cf8, 16); MD5STEP(F3, b, c, d, a, in[2] + 0xc4ac5665, 23); MD5STEP(F4, a, b, c, d, in[0] + 0xf4292244, 6); MD5STEP(F4, d, a, b, c, in[7] + 0x432aff97, 10); MD5STEP(F4, c, d, a, b, in[14] + 0xab9423a7, 15); MD5STEP(F4, b, c, d, a, in[5] + 0xfc93a039, 21); MD5STEP(F4, a, b, c, d, in[12] + 0x655b59c3, 6); MD5STEP(F4, d, a, b, c, in[3] + 0x8f0ccc92, 10); MD5STEP(F4, c, d, a, b, in[10] + 0xffeff47d, 15); MD5STEP(F4, b, c, d, a, in[1] + 0x85845dd1, 21); MD5STEP(F4, a, b, c, d, in[8] + 0x6fa87e4f, 6); MD5STEP(F4, d, a, b, c, in[15] + 0xfe2ce6e0, 10); MD5STEP(F4, c, d, a, b, in[6] + 0xa3014314, 15); MD5STEP(F4, b, c, d, a, in[13] + 0x4e0811a1, 21); MD5STEP(F4, a, b, c, d, in[4] + 0xf7537e82, 6); MD5STEP(F4, d, a, b, c, in[11] + 0xbd3af235, 10); MD5STEP(F4, c, d, a, b, in[2] + 0x2ad7d2bb, 15); MD5STEP(F4, b, c, d, a, in[9] + 0xeb86d391, 21); buf[0] += a; buf[1] += b; buf[2] += c; buf[3] += d; } void cs_md5_update(cs_md5_ctx *ctx, const unsigned char *buf, size_t len) { uint32_t t; t = ctx->bits[0]; if ((ctx->bits[0] = t + ((uint32_t) len << 3)) < t) ctx->bits[1]++; ctx->bits[1] += (uint32_t) len >> 29; t = (t >> 3) & 0x3f; if (t) { unsigned char *p = (unsigned char *) ctx->in + t; t = 64 - t; if (len < t) { memcpy(p, buf, len); return; } memcpy(p, buf, t); byteReverse(ctx->in, 16); cs_md5_transform(ctx->buf, (uint32_t *) ctx->in); buf += t; len -= t; } while (len >= 64) { memcpy(ctx->in, buf, 64); byteReverse(ctx->in, 16); cs_md5_transform(ctx->buf, (uint32_t *) ctx->in); buf += 64; len -= 64; } memcpy(ctx->in, buf, len); } void cs_md5_final(unsigned char digest[16], cs_md5_ctx *ctx) { unsigned count; unsigned char *p; uint32_t *a; count = (ctx->bits[0] >> 3) & 0x3F; p = ctx->in + count; *p++ = 0x80; count = 64 - 1 - count; if (count < 8) { memset(p, 0, count); byteReverse(ctx->in, 16); cs_md5_transform(ctx->buf, (uint32_t *) ctx->in); memset(ctx->in, 0, 56); } else { memset(p, 0, count - 8); } byteReverse(ctx->in, 14); a = (uint32_t *) ctx->in; a[14] = ctx->bits[0]; a[15] = ctx->bits[1]; cs_md5_transform(ctx->buf, (uint32_t *) ctx->in); byteReverse((unsigned char *) ctx->buf, 4); memcpy(digest, ctx->buf, 16); memset((char *) ctx, 0, sizeof(*ctx)); } #endif /* CS_DISABLE_MD5 */ #endif /* EXCLUDE_COMMON */ #ifdef MG_MODULE_LINES #line 1 "common/cs_sha1.c" #endif /* Copyright(c) By Steve Reid */ /* 100% Public Domain */ /* Amalgamated: #include "common/cs_sha1.h" */ #if !CS_DISABLE_SHA1 && !defined(EXCLUDE_COMMON) /* Amalgamated: #include "common/cs_endian.h" */ #define SHA1HANDSOFF #if defined(__sun) /* Amalgamated: #include "common/solarisfixes.h" */ #endif union char64long16 { unsigned char c[64]; uint32_t l[16]; }; #define rol(value, bits) (((value) << (bits)) | ((value) >> (32 - (bits)))) static uint32_t blk0(union char64long16 *block, int i) { /* Forrest: SHA expect BIG_ENDIAN, swap if LITTLE_ENDIAN */ #if BYTE_ORDER == LITTLE_ENDIAN block->l[i] = (rol(block->l[i], 24) & 0xFF00FF00) | (rol(block->l[i], 8) & 0x00FF00FF); #endif return block->l[i]; } /* Avoid redefine warning (ARM /usr/include/sys/ucontext.h define R0~R4) */ #undef blk #undef R0 #undef R1 #undef R2 #undef R3 #undef R4 #define blk(i) \ (block->l[i & 15] = rol(block->l[(i + 13) & 15] ^ block->l[(i + 8) & 15] ^ \ block->l[(i + 2) & 15] ^ block->l[i & 15], \ 1)) #define R0(v, w, x, y, z, i) \ z += ((w & (x ^ y)) ^ y) + blk0(block, i) + 0x5A827999 + rol(v, 5); \ w = rol(w, 30); #define R1(v, w, x, y, z, i) \ z += ((w & (x ^ y)) ^ y) + blk(i) + 0x5A827999 + rol(v, 5); \ w = rol(w, 30); #define R2(v, w, x, y, z, i) \ z += (w ^ x ^ y) + blk(i) + 0x6ED9EBA1 + rol(v, 5); \ w = rol(w, 30); #define R3(v, w, x, y, z, i) \ z += (((w | x) & y) | (w & x)) + blk(i) + 0x8F1BBCDC + rol(v, 5); \ w = rol(w, 30); #define R4(v, w, x, y, z, i) \ z += (w ^ x ^ y) + blk(i) + 0xCA62C1D6 + rol(v, 5); \ w = rol(w, 30); void cs_sha1_transform(uint32_t state[5], const unsigned char buffer[64]) { uint32_t a, b, c, d, e; union char64long16 block[1]; memcpy(block, buffer, 64); a = state[0]; b = state[1]; c = state[2]; d = state[3]; e = state[4]; R0(a, b, c, d, e, 0); R0(e, a, b, c, d, 1); R0(d, e, a, b, c, 2); R0(c, d, e, a, b, 3); R0(b, c, d, e, a, 4); R0(a, b, c, d, e, 5); R0(e, a, b, c, d, 6); R0(d, e, a, b, c, 7); R0(c, d, e, a, b, 8); R0(b, c, d, e, a, 9); R0(a, b, c, d, e, 10); R0(e, a, b, c, d, 11); R0(d, e, a, b, c, 12); R0(c, d, e, a, b, 13); R0(b, c, d, e, a, 14); R0(a, b, c, d, e, 15); R1(e, a, b, c, d, 16); R1(d, e, a, b, c, 17); R1(c, d, e, a, b, 18); R1(b, c, d, e, a, 19); R2(a, b, c, d, e, 20); R2(e, a, b, c, d, 21); R2(d, e, a, b, c, 22); R2(c, d, e, a, b, 23); R2(b, c, d, e, a, 24); R2(a, b, c, d, e, 25); R2(e, a, b, c, d, 26); R2(d, e, a, b, c, 27); R2(c, d, e, a, b, 28); R2(b, c, d, e, a, 29); R2(a, b, c, d, e, 30); R2(e, a, b, c, d, 31); R2(d, e, a, b, c, 32); R2(c, d, e, a, b, 33); R2(b, c, d, e, a, 34); R2(a, b, c, d, e, 35); R2(e, a, b, c, d, 36); R2(d, e, a, b, c, 37); R2(c, d, e, a, b, 38); R2(b, c, d, e, a, 39); R3(a, b, c, d, e, 40); R3(e, a, b, c, d, 41); R3(d, e, a, b, c, 42); R3(c, d, e, a, b, 43); R3(b, c, d, e, a, 44); R3(a, b, c, d, e, 45); R3(e, a, b, c, d, 46); R3(d, e, a, b, c, 47); R3(c, d, e, a, b, 48); R3(b, c, d, e, a, 49); R3(a, b, c, d, e, 50); R3(e, a, b, c, d, 51); R3(d, e, a, b, c, 52); R3(c, d, e, a, b, 53); R3(b, c, d, e, a, 54); R3(a, b, c, d, e, 55); R3(e, a, b, c, d, 56); R3(d, e, a, b, c, 57); R3(c, d, e, a, b, 58); R3(b, c, d, e, a, 59); R4(a, b, c, d, e, 60); R4(e, a, b, c, d, 61); R4(d, e, a, b, c, 62); R4(c, d, e, a, b, 63); R4(b, c, d, e, a, 64); R4(a, b, c, d, e, 65); R4(e, a, b, c, d, 66); R4(d, e, a, b, c, 67); R4(c, d, e, a, b, 68); R4(b, c, d, e, a, 69); R4(a, b, c, d, e, 70); R4(e, a, b, c, d, 71); R4(d, e, a, b, c, 72); R4(c, d, e, a, b, 73); R4(b, c, d, e, a, 74); R4(a, b, c, d, e, 75); R4(e, a, b, c, d, 76); R4(d, e, a, b, c, 77); R4(c, d, e, a, b, 78); R4(b, c, d, e, a, 79); state[0] += a; state[1] += b; state[2] += c; state[3] += d; state[4] += e; /* Erase working structures. The order of operations is important, * used to ensure that compiler doesn't optimize those out. */ memset(block, 0, sizeof(block)); a = b = c = d = e = 0; (void) a; (void) b; (void) c; (void) d; (void) e; } void cs_sha1_init(cs_sha1_ctx *context) { context->state[0] = 0x67452301; context->state[1] = 0xEFCDAB89; context->state[2] = 0x98BADCFE; context->state[3] = 0x10325476; context->state[4] = 0xC3D2E1F0; context->count[0] = context->count[1] = 0; } void cs_sha1_update(cs_sha1_ctx *context, const unsigned char *data, uint32_t len) { uint32_t i, j; j = context->count[0]; if ((context->count[0] += len << 3) < j) context->count[1]++; context->count[1] += (len >> 29); j = (j >> 3) & 63; if ((j + len) > 63) { memcpy(&context->buffer[j], data, (i = 64 - j)); cs_sha1_transform(context->state, context->buffer); for (; i + 63 < len; i += 64) { cs_sha1_transform(context->state, &data[i]); } j = 0; } else i = 0; memcpy(&context->buffer[j], &data[i], len - i); } void cs_sha1_final(unsigned char digest[20], cs_sha1_ctx *context) { unsigned i; unsigned char finalcount[8], c; for (i = 0; i < 8; i++) { finalcount[i] = (unsigned char) ((context->count[(i >= 4 ? 0 : 1)] >> ((3 - (i & 3)) * 8)) & 255); } c = 0200; cs_sha1_update(context, &c, 1); while ((context->count[0] & 504) != 448) { c = 0000; cs_sha1_update(context, &c, 1); } cs_sha1_update(context, finalcount, 8); for (i = 0; i < 20; i++) { digest[i] = (unsigned char) ((context->state[i >> 2] >> ((3 - (i & 3)) * 8)) & 255); } memset(context, '\0', sizeof(*context)); memset(&finalcount, '\0', sizeof(finalcount)); } void cs_hmac_sha1(const unsigned char *key, size_t keylen, const unsigned char *data, size_t datalen, unsigned char out[20]) { cs_sha1_ctx ctx; unsigned char buf1[64], buf2[64], tmp_key[20], i; if (keylen > sizeof(buf1)) { cs_sha1_init(&ctx); cs_sha1_update(&ctx, key, keylen); cs_sha1_final(tmp_key, &ctx); key = tmp_key; keylen = sizeof(tmp_key); } memset(buf1, 0, sizeof(buf1)); memset(buf2, 0, sizeof(buf2)); memcpy(buf1, key, keylen); memcpy(buf2, key, keylen); for (i = 0; i < sizeof(buf1); i++) { buf1[i] ^= 0x36; buf2[i] ^= 0x5c; } cs_sha1_init(&ctx); cs_sha1_update(&ctx, buf1, sizeof(buf1)); cs_sha1_update(&ctx, data, datalen); cs_sha1_final(out, &ctx); cs_sha1_init(&ctx); cs_sha1_update(&ctx, buf2, sizeof(buf2)); cs_sha1_update(&ctx, out, 20); cs_sha1_final(out, &ctx); } #endif /* EXCLUDE_COMMON */ #ifdef MG_MODULE_LINES #line 1 "common/mbuf.c" #endif /* * Copyright (c) 2014 Cesanta Software Limited * All rights reserved */ #ifndef EXCLUDE_COMMON #include #include /* Amalgamated: #include "common/mbuf.h" */ #ifndef MBUF_REALLOC #define MBUF_REALLOC realloc #endif #ifndef MBUF_FREE #define MBUF_FREE free #endif void mbuf_init(struct mbuf *mbuf, size_t initial_size) WEAK; void mbuf_init(struct mbuf *mbuf, size_t initial_size) { mbuf->len = mbuf->size = 0; mbuf->buf = NULL; mbuf_resize(mbuf, initial_size); } void mbuf_free(struct mbuf *mbuf) WEAK; void mbuf_free(struct mbuf *mbuf) { if (mbuf->buf != NULL) { MBUF_FREE(mbuf->buf); mbuf_init(mbuf, 0); } } void mbuf_resize(struct mbuf *a, size_t new_size) WEAK; void mbuf_resize(struct mbuf *a, size_t new_size) { if (new_size > a->size || (new_size < a->size && new_size >= a->len)) { char *buf = (char *) MBUF_REALLOC(a->buf, new_size); /* * In case realloc fails, there's not much we can do, except keep things as * they are. Note that NULL is a valid return value from realloc when * size == 0, but that is covered too. */ if (buf == NULL && new_size != 0) return; a->buf = buf; a->size = new_size; } } void mbuf_trim(struct mbuf *mbuf) WEAK; void mbuf_trim(struct mbuf *mbuf) { mbuf_resize(mbuf, mbuf->len); } size_t mbuf_insert(struct mbuf *a, size_t off, const void *buf, size_t) WEAK; size_t mbuf_insert(struct mbuf *a, size_t off, const void *buf, size_t len) { char *p = NULL; assert(a != NULL); assert(a->len <= a->size); assert(off <= a->len); /* check overflow */ if (~(size_t) 0 - (size_t) a->buf < len) return 0; if (a->len + len <= a->size) { memmove(a->buf + off + len, a->buf + off, a->len - off); if (buf != NULL) { memcpy(a->buf + off, buf, len); } a->len += len; } else { size_t new_size = (size_t)((a->len + len) * MBUF_SIZE_MULTIPLIER); if ((p = (char *) MBUF_REALLOC(a->buf, new_size)) != NULL) { a->buf = p; memmove(a->buf + off + len, a->buf + off, a->len - off); if (buf != NULL) memcpy(a->buf + off, buf, len); a->len += len; a->size = new_size; } else { len = 0; } } return len; } size_t mbuf_append(struct mbuf *a, const void *buf, size_t len) WEAK; size_t mbuf_append(struct mbuf *a, const void *buf, size_t len) { return mbuf_insert(a, a->len, buf, len); } void mbuf_remove(struct mbuf *mb, size_t n) WEAK; void mbuf_remove(struct mbuf *mb, size_t n) { if (n > 0 && n <= mb->len) { memmove(mb->buf, mb->buf + n, mb->len - n); mb->len -= n; } } #endif /* EXCLUDE_COMMON */ #ifdef MG_MODULE_LINES #line 1 "common/mg_str.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ /* Amalgamated: #include "common/mg_mem.h" */ /* Amalgamated: #include "common/mg_str.h" */ #include #include int mg_ncasecmp(const char *s1, const char *s2, size_t len) WEAK; struct mg_str mg_mk_str(const char *s) WEAK; struct mg_str mg_mk_str(const char *s) { struct mg_str ret = {s, 0}; if (s != NULL) ret.len = strlen(s); return ret; } struct mg_str mg_mk_str_n(const char *s, size_t len) WEAK; struct mg_str mg_mk_str_n(const char *s, size_t len) { struct mg_str ret = {s, len}; return ret; } int mg_vcmp(const struct mg_str *str1, const char *str2) WEAK; int mg_vcmp(const struct mg_str *str1, const char *str2) { size_t n2 = strlen(str2), n1 = str1->len; int r = strncmp(str1->p, str2, (n1 < n2) ? n1 : n2); if (r == 0) { return n1 - n2; } return r; } int mg_vcasecmp(const struct mg_str *str1, const char *str2) WEAK; int mg_vcasecmp(const struct mg_str *str1, const char *str2) { size_t n2 = strlen(str2), n1 = str1->len; int r = mg_ncasecmp(str1->p, str2, (n1 < n2) ? n1 : n2); if (r == 0) { return n1 - n2; } return r; } static struct mg_str mg_strdup_common(const struct mg_str s, int nul_terminate) { struct mg_str r = {NULL, 0}; if (s.len > 0 && s.p != NULL) { char *sc = (char *) MG_MALLOC(s.len + (nul_terminate ? 1 : 0)); if (sc != NULL) { memcpy(sc, s.p, s.len); if (nul_terminate) sc[s.len] = '\0'; r.p = sc; r.len = s.len; } } return r; } struct mg_str mg_strdup(const struct mg_str s) WEAK; struct mg_str mg_strdup(const struct mg_str s) { return mg_strdup_common(s, 0 /* NUL-terminate */); } struct mg_str mg_strdup_nul(const struct mg_str s) WEAK; struct mg_str mg_strdup_nul(const struct mg_str s) { return mg_strdup_common(s, 1 /* NUL-terminate */); } const char *mg_strchr(const struct mg_str s, int c) WEAK; const char *mg_strchr(const struct mg_str s, int c) { size_t i; for (i = 0; i < s.len; i++) { if (s.p[i] == c) return &s.p[i]; } return NULL; } int mg_strcmp(const struct mg_str str1, const struct mg_str str2) WEAK; int mg_strcmp(const struct mg_str str1, const struct mg_str str2) { size_t i = 0; while (i < str1.len && i < str2.len) { if (str1.p[i] < str2.p[i]) return -1; if (str1.p[i] > str2.p[i]) return 1; i++; } if (i < str1.len) return 1; if (i < str2.len) return -1; return 0; } int mg_strncmp(const struct mg_str, const struct mg_str, size_t n) WEAK; int mg_strncmp(const struct mg_str str1, const struct mg_str str2, size_t n) { struct mg_str s1 = str1; struct mg_str s2 = str2; if (s1.len > n) { s1.len = n; } if (s2.len > n) { s2.len = n; } return mg_strcmp(s1, s2); } const char *mg_strstr(const struct mg_str haystack, const struct mg_str needle) WEAK; const char *mg_strstr(const struct mg_str haystack, const struct mg_str needle) { size_t i; if (needle.len > haystack.len) return NULL; for (i = 0; i <= haystack.len - needle.len; i++) { if (memcmp(haystack.p + i, needle.p, needle.len) == 0) { return haystack.p + i; } } return NULL; } #ifdef MG_MODULE_LINES #line 1 "common/str_util.c" #endif /* * Copyright (c) 2015 Cesanta Software Limited * All rights reserved */ #ifndef EXCLUDE_COMMON /* Amalgamated: #include "common/str_util.h" */ /* Amalgamated: #include "common/mg_mem.h" */ /* Amalgamated: #include "common/platform.h" */ #ifndef C_DISABLE_BUILTIN_SNPRINTF #define C_DISABLE_BUILTIN_SNPRINTF 0 #endif /* Amalgamated: #include "common/mg_mem.h" */ size_t c_strnlen(const char *s, size_t maxlen) WEAK; size_t c_strnlen(const char *s, size_t maxlen) { size_t l = 0; for (; l < maxlen && s[l] != '\0'; l++) { } return l; } #define C_SNPRINTF_APPEND_CHAR(ch) \ do { \ if (i < (int) buf_size) buf[i] = ch; \ i++; \ } while (0) #define C_SNPRINTF_FLAG_ZERO 1 #if C_DISABLE_BUILTIN_SNPRINTF int c_vsnprintf(char *buf, size_t buf_size, const char *fmt, va_list ap) WEAK; int c_vsnprintf(char *buf, size_t buf_size, const char *fmt, va_list ap) { return vsnprintf(buf, buf_size, fmt, ap); } #else static int c_itoa(char *buf, size_t buf_size, int64_t num, int base, int flags, int field_width) { char tmp[40]; int i = 0, k = 0, neg = 0; if (num < 0) { neg++; num = -num; } /* Print into temporary buffer - in reverse order */ do { int rem = num % base; if (rem < 10) { tmp[k++] = '0' + rem; } else { tmp[k++] = 'a' + (rem - 10); } num /= base; } while (num > 0); /* Zero padding */ if (flags && C_SNPRINTF_FLAG_ZERO) { while (k < field_width && k < (int) sizeof(tmp) - 1) { tmp[k++] = '0'; } } /* And sign */ if (neg) { tmp[k++] = '-'; } /* Now output */ while (--k >= 0) { C_SNPRINTF_APPEND_CHAR(tmp[k]); } return i; } int c_vsnprintf(char *buf, size_t buf_size, const char *fmt, va_list ap) WEAK; int c_vsnprintf(char *buf, size_t buf_size, const char *fmt, va_list ap) { int ch, i = 0, len_mod, flags, precision, field_width; while ((ch = *fmt++) != '\0') { if (ch != '%') { C_SNPRINTF_APPEND_CHAR(ch); } else { /* * Conversion specification: * zero or more flags (one of: # 0 - + ') * an optional minimum field width (digits) * an optional precision (. followed by digits, or *) * an optional length modifier (one of: hh h l ll L q j z t) * conversion specifier (one of: d i o u x X e E f F g G a A c s p n) */ flags = field_width = precision = len_mod = 0; /* Flags. only zero-pad flag is supported. */ if (*fmt == '0') { flags |= C_SNPRINTF_FLAG_ZERO; } /* Field width */ while (*fmt >= '0' && *fmt <= '9') { field_width *= 10; field_width += *fmt++ - '0'; } /* Dynamic field width */ if (*fmt == '*') { field_width = va_arg(ap, int); fmt++; } /* Precision */ if (*fmt == '.') { fmt++; if (*fmt == '*') { precision = va_arg(ap, int); fmt++; } else { while (*fmt >= '0' && *fmt <= '9') { precision *= 10; precision += *fmt++ - '0'; } } } /* Length modifier */ switch (*fmt) { case 'h': case 'l': case 'L': case 'I': case 'q': case 'j': case 'z': case 't': len_mod = *fmt++; if (*fmt == 'h') { len_mod = 'H'; fmt++; } if (*fmt == 'l') { len_mod = 'q'; fmt++; } break; } ch = *fmt++; if (ch == 's') { const char *s = va_arg(ap, const char *); /* Always fetch parameter */ int j; int pad = field_width - (precision >= 0 ? c_strnlen(s, precision) : 0); for (j = 0; j < pad; j++) { C_SNPRINTF_APPEND_CHAR(' '); } /* `s` may be NULL in case of %.*s */ if (s != NULL) { /* Ignore negative and 0 precisions */ for (j = 0; (precision <= 0 || j < precision) && s[j] != '\0'; j++) { C_SNPRINTF_APPEND_CHAR(s[j]); } } } else if (ch == 'c') { ch = va_arg(ap, int); /* Always fetch parameter */ C_SNPRINTF_APPEND_CHAR(ch); } else if (ch == 'd' && len_mod == 0) { i += c_itoa(buf + i, buf_size - i, va_arg(ap, int), 10, flags, field_width); } else if (ch == 'd' && len_mod == 'l') { i += c_itoa(buf + i, buf_size - i, va_arg(ap, long), 10, flags, field_width); #ifdef SSIZE_MAX } else if (ch == 'd' && len_mod == 'z') { i += c_itoa(buf + i, buf_size - i, va_arg(ap, ssize_t), 10, flags, field_width); #endif } else if (ch == 'd' && len_mod == 'q') { i += c_itoa(buf + i, buf_size - i, va_arg(ap, int64_t), 10, flags, field_width); } else if ((ch == 'x' || ch == 'u') && len_mod == 0) { i += c_itoa(buf + i, buf_size - i, va_arg(ap, unsigned), ch == 'x' ? 16 : 10, flags, field_width); } else if ((ch == 'x' || ch == 'u') && len_mod == 'l') { i += c_itoa(buf + i, buf_size - i, va_arg(ap, unsigned long), ch == 'x' ? 16 : 10, flags, field_width); } else if ((ch == 'x' || ch == 'u') && len_mod == 'z') { i += c_itoa(buf + i, buf_size - i, va_arg(ap, size_t), ch == 'x' ? 16 : 10, flags, field_width); } else if (ch == 'p') { unsigned long num = (unsigned long) (uintptr_t) va_arg(ap, void *); C_SNPRINTF_APPEND_CHAR('0'); C_SNPRINTF_APPEND_CHAR('x'); i += c_itoa(buf + i, buf_size - i, num, 16, flags, 0); } else { #ifndef NO_LIBC /* * TODO(lsm): abort is not nice in a library, remove it * Also, ESP8266 SDK doesn't have it */ abort(); #endif } } } /* Zero-terminate the result */ if (buf_size > 0) { buf[i < (int) buf_size ? i : (int) buf_size - 1] = '\0'; } return i; } #endif int c_snprintf(char *buf, size_t buf_size, const char *fmt, ...) WEAK; int c_snprintf(char *buf, size_t buf_size, const char *fmt, ...) { int result; va_list ap; va_start(ap, fmt); result = c_vsnprintf(buf, buf_size, fmt, ap); va_end(ap); return result; } #ifdef _WIN32 int to_wchar(const char *path, wchar_t *wbuf, size_t wbuf_len) { int ret; char buf[MAX_PATH * 2], buf2[MAX_PATH * 2], *p; strncpy(buf, path, sizeof(buf)); buf[sizeof(buf) - 1] = '\0'; /* Trim trailing slashes. Leave backslash for paths like "X:\" */ p = buf + strlen(buf) - 1; while (p > buf && p[-1] != ':' && (p[0] == '\\' || p[0] == '/')) *p-- = '\0'; memset(wbuf, 0, wbuf_len * sizeof(wchar_t)); ret = MultiByteToWideChar(CP_UTF8, 0, buf, -1, wbuf, (int) wbuf_len); /* * Convert back to Unicode. If doubly-converted string does not match the * original, something is fishy, reject. */ WideCharToMultiByte(CP_UTF8, 0, wbuf, (int) wbuf_len, buf2, sizeof(buf2), NULL, NULL); if (strcmp(buf, buf2) != 0) { wbuf[0] = L'\0'; ret = 0; } return ret; } #endif /* _WIN32 */ /* The simplest O(mn) algorithm. Better implementation are GPLed */ const char *c_strnstr(const char *s, const char *find, size_t slen) WEAK; const char *c_strnstr(const char *s, const char *find, size_t slen) { size_t find_length = strlen(find); size_t i; for (i = 0; i < slen; i++) { if (i + find_length > slen) { return NULL; } if (strncmp(&s[i], find, find_length) == 0) { return &s[i]; } } return NULL; } #if CS_ENABLE_STRDUP char *strdup(const char *src) WEAK; char *strdup(const char *src) { size_t len = strlen(src) + 1; char *ret = MG_MALLOC(len); if (ret != NULL) { strcpy(ret, src); } return ret; } #endif void cs_to_hex(char *to, const unsigned char *p, size_t len) WEAK; void cs_to_hex(char *to, const unsigned char *p, size_t len) { static const char *hex = "0123456789abcdef"; for (; len--; p++) { *to++ = hex[p[0] >> 4]; *to++ = hex[p[0] & 0x0f]; } *to = '\0'; } static int fourbit(int ch) { if (ch >= '0' && ch <= '9') { return ch - '0'; } else if (ch >= 'a' && ch <= 'f') { return ch - 'a' + 10; } else if (ch >= 'A' && ch <= 'F') { return ch - 'A' + 10; } return 0; } void cs_from_hex(char *to, const char *p, size_t len) WEAK; void cs_from_hex(char *to, const char *p, size_t len) { size_t i; for (i = 0; i < len; i += 2) { *to++ = (fourbit(p[i]) << 4) + fourbit(p[i + 1]); } *to = '\0'; } #if CS_ENABLE_TO64 int64_t cs_to64(const char *s) WEAK; int64_t cs_to64(const char *s) { int64_t result = 0; int64_t neg = 1; while (*s && isspace((unsigned char) *s)) s++; if (*s == '-') { neg = -1; s++; } while (isdigit((unsigned char) *s)) { result *= 10; result += (*s - '0'); s++; } return result * neg; } #endif static int str_util_lowercase(const char *s) { return tolower(*(const unsigned char *) s); } int mg_ncasecmp(const char *s1, const char *s2, size_t len) WEAK; int mg_ncasecmp(const char *s1, const char *s2, size_t len) { int diff = 0; if (len > 0) do { diff = str_util_lowercase(s1++) - str_util_lowercase(s2++); } while (diff == 0 && s1[-1] != '\0' && --len > 0); return diff; } int mg_casecmp(const char *s1, const char *s2) WEAK; int mg_casecmp(const char *s1, const char *s2) { return mg_ncasecmp(s1, s2, (size_t) ~0); } int mg_asprintf(char **buf, size_t size, const char *fmt, ...) WEAK; int mg_asprintf(char **buf, size_t size, const char *fmt, ...) { int ret; va_list ap; va_start(ap, fmt); ret = mg_avprintf(buf, size, fmt, ap); va_end(ap); return ret; } int mg_avprintf(char **buf, size_t size, const char *fmt, va_list ap) WEAK; int mg_avprintf(char **buf, size_t size, const char *fmt, va_list ap) { va_list ap_copy; int len; va_copy(ap_copy, ap); len = vsnprintf(*buf, size, fmt, ap_copy); va_end(ap_copy); if (len < 0) { /* eCos and Windows are not standard-compliant and return -1 when * the buffer is too small. Keep allocating larger buffers until we * succeed or out of memory. */ *buf = NULL; /* LCOV_EXCL_START */ while (len < 0) { MG_FREE(*buf); if (size == 0) { size = 5; } size *= 2; if ((*buf = (char *) MG_MALLOC(size)) == NULL) { len = -1; break; } va_copy(ap_copy, ap); len = vsnprintf(*buf, size - 1, fmt, ap_copy); va_end(ap_copy); } /* * Microsoft version of vsnprintf() is not always null-terminated, so put * the terminator manually */ (*buf)[len] = 0; /* LCOV_EXCL_STOP */ } else if (len >= (int) size) { /* Standard-compliant code path. Allocate a buffer that is large enough. */ if ((*buf = (char *) MG_MALLOC(len + 1)) == NULL) { len = -1; /* LCOV_EXCL_LINE */ } else { /* LCOV_EXCL_LINE */ va_copy(ap_copy, ap); len = vsnprintf(*buf, len + 1, fmt, ap_copy); va_end(ap_copy); } } return len; } const char *mg_next_comma_list_entry(const char *, struct mg_str *, struct mg_str *) WEAK; const char *mg_next_comma_list_entry(const char *list, struct mg_str *val, struct mg_str *eq_val) { struct mg_str ret = mg_next_comma_list_entry_n(mg_mk_str(list), val, eq_val); return ret.p; } struct mg_str mg_next_comma_list_entry_n(struct mg_str list, struct mg_str *val, struct mg_str *eq_val) WEAK; struct mg_str mg_next_comma_list_entry_n(struct mg_str list, struct mg_str *val, struct mg_str *eq_val) { if (list.len == 0) { /* End of the list */ list = mg_mk_str(NULL); } else { const char *chr = NULL; *val = list; if ((chr = mg_strchr(*val, ',')) != NULL) { /* Comma found. Store length and shift the list ptr */ val->len = chr - val->p; chr++; list.len -= (chr - list.p); list.p = chr; } else { /* This value is the last one */ list = mg_mk_str_n(list.p + list.len, 0); } if (eq_val != NULL) { /* Value has form "x=y", adjust pointers and lengths */ /* so that val points to "x", and eq_val points to "y". */ eq_val->len = 0; eq_val->p = (const char *) memchr(val->p, '=', val->len); if (eq_val->p != NULL) { eq_val->p++; /* Skip over '=' character */ eq_val->len = val->p + val->len - eq_val->p; val->len = (eq_val->p - val->p) - 1; } } } return list; } size_t mg_match_prefix_n(const struct mg_str, const struct mg_str) WEAK; size_t mg_match_prefix_n(const struct mg_str pattern, const struct mg_str str) { const char *or_str; size_t res = 0, len = 0, i = 0, j = 0; if ((or_str = (const char *) memchr(pattern.p, '|', pattern.len)) != NULL || (or_str = (const char *) memchr(pattern.p, ',', pattern.len)) != NULL) { struct mg_str pstr = {pattern.p, (size_t)(or_str - pattern.p)}; res = mg_match_prefix_n(pstr, str); if (res > 0) return res; pstr.p = or_str + 1; pstr.len = (pattern.p + pattern.len) - (or_str + 1); return mg_match_prefix_n(pstr, str); } for (; i < pattern.len && j < str.len; i++, j++) { if (pattern.p[i] == '?') { continue; } else if (pattern.p[i] == '*') { i++; if (i < pattern.len && pattern.p[i] == '*') { i++; len = str.len - j; } else { len = 0; while (j + len < str.len && str.p[j + len] != '/') len++; } if (i == pattern.len || (pattern.p[i] == '$' && i == pattern.len - 1)) return j + len; do { const struct mg_str pstr = {pattern.p + i, pattern.len - i}; const struct mg_str sstr = {str.p + j + len, str.len - j - len}; res = mg_match_prefix_n(pstr, sstr); } while (res == 0 && len != 0 && len-- > 0); return res == 0 ? 0 : j + res + len; } else if (str_util_lowercase(&pattern.p[i]) != str_util_lowercase(&str.p[j])) { break; } } if (i < pattern.len && pattern.p[i] == '$') { return j == str.len ? str.len : 0; } return i == pattern.len ? j : 0; } size_t mg_match_prefix(const char *, int, const char *) WEAK; size_t mg_match_prefix(const char *pattern, int pattern_len, const char *str) { const struct mg_str pstr = {pattern, (size_t) pattern_len}; struct mg_str s = {str, 0}; if (str != NULL) s.len = strlen(str); return mg_match_prefix_n(pstr, s); } #endif /* EXCLUDE_COMMON */ #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_net.c" #endif /* * Copyright (c) 2014 Cesanta Software Limited * All rights reserved * * This software is dual-licensed: you can redistribute it and/or modify * it under the terms of the GNU General Public License version 2 as * published by the Free Software Foundation. For the terms of this * license, see . * * You are free to use this software under the terms of the GNU General * Public License, but WITHOUT ANY WARRANTY; without even the implied * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. * See the GNU General Public License for more details. * * Alternatively, you can license this software under a commercial * license, as set out in . */ /* Amalgamated: #include "common/cs_time.h" */ /* Amalgamated: #include "mg_dns.h" */ /* Amalgamated: #include "mg_internal.h" */ /* Amalgamated: #include "mg_resolv.h" */ /* Amalgamated: #include "mg_util.h" */ #define MG_MAX_HOST_LEN 200 #define MG_COPY_COMMON_CONNECTION_OPTIONS(dst, src) \ memcpy(dst, src, sizeof(*dst)); /* Which flags can be pre-set by the user at connection creation time. */ #define _MG_ALLOWED_CONNECT_FLAGS_MASK \ (MG_F_USER_1 | MG_F_USER_2 | MG_F_USER_3 | MG_F_USER_4 | MG_F_USER_5 | \ MG_F_USER_6 | MG_F_WEBSOCKET_NO_DEFRAG | MG_F_ENABLE_BROADCAST) /* Which flags should be modifiable by user's callbacks. */ #define _MG_CALLBACK_MODIFIABLE_FLAGS_MASK \ (MG_F_USER_1 | MG_F_USER_2 | MG_F_USER_3 | MG_F_USER_4 | MG_F_USER_5 | \ MG_F_USER_6 | MG_F_WEBSOCKET_NO_DEFRAG | MG_F_SEND_AND_CLOSE | \ MG_F_CLOSE_IMMEDIATELY | MG_F_IS_WEBSOCKET | MG_F_DELETE_CHUNK) #ifndef intptr_t #define intptr_t long #endif MG_INTERNAL void mg_add_conn(struct mg_mgr *mgr, struct mg_connection *c) { DBG(("%p %p", mgr, c)); c->mgr = mgr; c->next = mgr->active_connections; mgr->active_connections = c; c->prev = NULL; if (c->next != NULL) c->next->prev = c; if (c->sock != INVALID_SOCKET) { c->iface->vtable->add_conn(c); } } MG_INTERNAL void mg_remove_conn(struct mg_connection *conn) { if (conn->prev == NULL) conn->mgr->active_connections = conn->next; if (conn->prev) conn->prev->next = conn->next; if (conn->next) conn->next->prev = conn->prev; conn->prev = conn->next = NULL; conn->iface->vtable->remove_conn(conn); } MG_INTERNAL void mg_call(struct mg_connection *nc, mg_event_handler_t ev_handler, void *user_data, int ev, void *ev_data) { static int nesting_level = 0; nesting_level++; if (ev_handler == NULL) { /* * If protocol handler is specified, call it. Otherwise, call user-specified * event handler. */ ev_handler = nc->proto_handler ? nc->proto_handler : nc->handler; } if (ev != MG_EV_POLL) { DBG(("%p %s ev=%d ev_data=%p flags=%lu rmbl=%d smbl=%d", nc, ev_handler == nc->handler ? "user" : "proto", ev, ev_data, nc->flags, (int) nc->recv_mbuf.len, (int) nc->send_mbuf.len)); } #if !defined(NO_LIBC) && MG_ENABLE_HEXDUMP if (nc->mgr->hexdump_file != NULL && ev != MG_EV_POLL && ev != MG_EV_RECV && ev != MG_EV_SEND /* handled separately */) { mg_hexdump_connection(nc, nc->mgr->hexdump_file, NULL, 0, ev); } #endif if (ev_handler != NULL) { unsigned long flags_before = nc->flags; size_t recv_mbuf_before = nc->recv_mbuf.len, recved; ev_handler(nc, ev, ev_data MG_UD_ARG(user_data)); recved = (recv_mbuf_before - nc->recv_mbuf.len); /* Prevent user handler from fiddling with system flags. */ if (ev_handler == nc->handler && nc->flags != flags_before) { nc->flags = (flags_before & ~_MG_CALLBACK_MODIFIABLE_FLAGS_MASK) | (nc->flags & _MG_CALLBACK_MODIFIABLE_FLAGS_MASK); } /* It's important to not double-count recved bytes, and since mg_call can be * called recursively (e.g. proto_handler invokes user handler), we keep * track of recursion and only report received bytes at the top level. */ if (nesting_level == 1 && recved > 0 && !(nc->flags & MG_F_UDP)) { nc->iface->vtable->recved(nc, recved); } } if (ev != MG_EV_POLL) { DBG(("%p after %s flags=%lu rmbl=%d smbl=%d", nc, ev_handler == nc->handler ? "user" : "proto", nc->flags, (int) nc->recv_mbuf.len, (int) nc->send_mbuf.len)); } nesting_level--; #if !MG_ENABLE_CALLBACK_USERDATA (void) user_data; #endif } void mg_if_timer(struct mg_connection *c, double now) { if (c->ev_timer_time > 0 && now >= c->ev_timer_time) { double old_value = c->ev_timer_time; c->ev_timer_time = 0; mg_call(c, NULL, c->user_data, MG_EV_TIMER, &old_value); } } void mg_if_poll(struct mg_connection *nc, time_t now) { if (!(nc->flags & MG_F_SSL) || (nc->flags & MG_F_SSL_HANDSHAKE_DONE)) { mg_call(nc, NULL, nc->user_data, MG_EV_POLL, &now); } } void mg_destroy_conn(struct mg_connection *conn, int destroy_if) { if (destroy_if) conn->iface->vtable->destroy_conn(conn); if (conn->proto_data != NULL && conn->proto_data_destructor != NULL) { conn->proto_data_destructor(conn->proto_data); } #if MG_ENABLE_SSL mg_ssl_if_conn_free(conn); #endif mbuf_free(&conn->recv_mbuf); mbuf_free(&conn->send_mbuf); memset(conn, 0, sizeof(*conn)); MG_FREE(conn); } void mg_close_conn(struct mg_connection *conn) { DBG(("%p %lu %d", conn, conn->flags, conn->sock)); #if MG_ENABLE_SSL if (conn->flags & MG_F_SSL_HANDSHAKE_DONE) { mg_ssl_if_conn_close_notify(conn); } #endif mg_remove_conn(conn); conn->iface->vtable->destroy_conn(conn); mg_call(conn, NULL, conn->user_data, MG_EV_CLOSE, NULL); mg_destroy_conn(conn, 0 /* destroy_if */); } void mg_mgr_init(struct mg_mgr *m, void *user_data) { struct mg_mgr_init_opts opts; memset(&opts, 0, sizeof(opts)); mg_mgr_init_opt(m, user_data, opts); } void mg_mgr_init_opt(struct mg_mgr *m, void *user_data, struct mg_mgr_init_opts opts) { memset(m, 0, sizeof(*m)); #if MG_ENABLE_BROADCAST m->ctl[0] = m->ctl[1] = INVALID_SOCKET; #endif m->user_data = user_data; #ifdef _WIN32 { WSADATA data; WSAStartup(MAKEWORD(2, 2), &data); } #elif defined(__unix__) /* Ignore SIGPIPE signal, so if client cancels the request, it * won't kill the whole process. */ signal(SIGPIPE, SIG_IGN); #endif #if MG_ENABLE_SSL { static int init_done; if (!init_done) { mg_ssl_if_init(); init_done++; } } #endif { int i; if (opts.num_ifaces == 0) { opts.num_ifaces = mg_num_ifaces; opts.ifaces = mg_ifaces; } if (opts.main_iface != NULL) { opts.ifaces[MG_MAIN_IFACE] = opts.main_iface; } m->num_ifaces = opts.num_ifaces; m->ifaces = (struct mg_iface **) MG_MALLOC(sizeof(*m->ifaces) * opts.num_ifaces); for (i = 0; i < mg_num_ifaces; i++) { m->ifaces[i] = mg_if_create_iface(opts.ifaces[i], m); m->ifaces[i]->vtable->init(m->ifaces[i]); } } if (opts.nameserver != NULL) { m->nameserver = strdup(opts.nameserver); } DBG(("==================================")); DBG(("init mgr=%p", m)); } void mg_mgr_free(struct mg_mgr *m) { struct mg_connection *conn, *tmp_conn; DBG(("%p", m)); if (m == NULL) return; /* Do one last poll, see https://github.com/cesanta/mongoose/issues/286 */ mg_mgr_poll(m, 0); #if MG_ENABLE_BROADCAST if (m->ctl[0] != INVALID_SOCKET) closesocket(m->ctl[0]); if (m->ctl[1] != INVALID_SOCKET) closesocket(m->ctl[1]); m->ctl[0] = m->ctl[1] = INVALID_SOCKET; #endif for (conn = m->active_connections; conn != NULL; conn = tmp_conn) { tmp_conn = conn->next; mg_close_conn(conn); } { int i; for (i = 0; i < m->num_ifaces; i++) { m->ifaces[i]->vtable->free(m->ifaces[i]); MG_FREE(m->ifaces[i]); } MG_FREE(m->ifaces); } MG_FREE((char *) m->nameserver); } time_t mg_mgr_poll(struct mg_mgr *m, int timeout_ms) { int i; time_t now = 0; /* oh GCC, seriously ? */ if (m->num_ifaces == 0) { LOG(LL_ERROR, ("cannot poll: no interfaces")); return 0; } for (i = 0; i < m->num_ifaces; i++) { now = m->ifaces[i]->vtable->poll(m->ifaces[i], timeout_ms); } return now; } int mg_vprintf(struct mg_connection *nc, const char *fmt, va_list ap) { char mem[MG_VPRINTF_BUFFER_SIZE], *buf = mem; int len; if ((len = mg_avprintf(&buf, sizeof(mem), fmt, ap)) > 0) { mg_send(nc, buf, len); } if (buf != mem && buf != NULL) { MG_FREE(buf); /* LCOV_EXCL_LINE */ } /* LCOV_EXCL_LINE */ return len; } int mg_printf(struct mg_connection *conn, const char *fmt, ...) { int len; va_list ap; va_start(ap, fmt); len = mg_vprintf(conn, fmt, ap); va_end(ap); return len; } #if MG_ENABLE_SYNC_RESOLVER /* TODO(lsm): use non-blocking resolver */ static int mg_resolve2(const char *host, struct in_addr *ina) { #if MG_ENABLE_GETADDRINFO int rv = 0; struct addrinfo hints, *servinfo, *p; struct sockaddr_in *h = NULL; memset(&hints, 0, sizeof hints); hints.ai_family = AF_INET; hints.ai_socktype = SOCK_STREAM; if ((rv = getaddrinfo(host, NULL, NULL, &servinfo)) != 0) { DBG(("getaddrinfo(%s) failed: %s", host, strerror(mg_get_errno()))); return 0; } for (p = servinfo; p != NULL; p = p->ai_next) { memcpy(&h, &p->ai_addr, sizeof(struct sockaddr_in *)); memcpy(ina, &h->sin_addr, sizeof(ina)); } freeaddrinfo(servinfo); return 1; #else struct hostent *he; if ((he = gethostbyname(host)) == NULL) { DBG(("gethostbyname(%s) failed: %s", host, strerror(mg_get_errno()))); } else { memcpy(ina, he->h_addr_list[0], sizeof(*ina)); return 1; } return 0; #endif /* MG_ENABLE_GETADDRINFO */ } int mg_resolve(const char *host, char *buf, size_t n) { struct in_addr ad; return mg_resolve2(host, &ad) ? snprintf(buf, n, "%s", inet_ntoa(ad)) : 0; } #endif /* MG_ENABLE_SYNC_RESOLVER */ MG_INTERNAL struct mg_connection *mg_create_connection_base( struct mg_mgr *mgr, mg_event_handler_t callback, struct mg_add_sock_opts opts) { struct mg_connection *conn; if ((conn = (struct mg_connection *) MG_CALLOC(1, sizeof(*conn))) != NULL) { conn->sock = INVALID_SOCKET; conn->handler = callback; conn->mgr = mgr; conn->last_io_time = (time_t) mg_time(); conn->iface = (opts.iface != NULL ? opts.iface : mgr->ifaces[MG_MAIN_IFACE]); conn->flags = opts.flags & _MG_ALLOWED_CONNECT_FLAGS_MASK; conn->user_data = opts.user_data; /* * SIZE_MAX is defined as a long long constant in * system headers on some platforms and so it * doesn't compile with pedantic ansi flags. */ conn->recv_mbuf_limit = ~0; } else { MG_SET_PTRPTR(opts.error_string, "failed to create connection"); } return conn; } MG_INTERNAL struct mg_connection *mg_create_connection( struct mg_mgr *mgr, mg_event_handler_t callback, struct mg_add_sock_opts opts) { struct mg_connection *conn = mg_create_connection_base(mgr, callback, opts); if (conn != NULL && !conn->iface->vtable->create_conn(conn)) { MG_FREE(conn); conn = NULL; } if (conn == NULL) { MG_SET_PTRPTR(opts.error_string, "failed to init connection"); } return conn; } /* * Address format: [PROTO://][HOST]:PORT * * HOST could be IPv4/IPv6 address or a host name. * `host` is a destination buffer to hold parsed HOST part. Should be at least * MG_MAX_HOST_LEN bytes long. * `proto` is a returned socket type, either SOCK_STREAM or SOCK_DGRAM * * Return: * -1 on parse error * 0 if HOST needs DNS lookup * >0 length of the address string */ MG_INTERNAL int mg_parse_address(const char *str, union socket_address *sa, int *proto, char *host, size_t host_len) { unsigned int a, b, c, d, port = 0; int ch, len = 0; #if MG_ENABLE_IPV6 char buf[100]; #endif /* * MacOS needs that. If we do not zero it, subsequent bind() will fail. * Also, all-zeroes in the socket address means binding to all addresses * for both IPv4 and IPv6 (INADDR_ANY and IN6ADDR_ANY_INIT). */ memset(sa, 0, sizeof(*sa)); sa->sin.sin_family = AF_INET; *proto = SOCK_STREAM; if (strncmp(str, "udp://", 6) == 0) { str += 6; *proto = SOCK_DGRAM; } else if (strncmp(str, "tcp://", 6) == 0) { str += 6; } if (sscanf(str, "%u.%u.%u.%u:%u%n", &a, &b, &c, &d, &port, &len) == 5) { /* Bind to a specific IPv4 address, e.g. 192.168.1.5:8080 */ sa->sin.sin_addr.s_addr = htonl(((uint32_t) a << 24) | ((uint32_t) b << 16) | c << 8 | d); sa->sin.sin_port = htons((uint16_t) port); #if MG_ENABLE_IPV6 } else if (sscanf(str, "[%99[^]]]:%u%n", buf, &port, &len) == 2 && inet_pton(AF_INET6, buf, &sa->sin6.sin6_addr)) { /* IPv6 address, e.g. [3ffe:2a00:100:7031::1]:8080 */ sa->sin6.sin6_family = AF_INET6; sa->sin.sin_port = htons((uint16_t) port); #endif #if MG_ENABLE_ASYNC_RESOLVER } else if (strlen(str) < host_len && sscanf(str, "%[^ :]:%u%n", host, &port, &len) == 2) { sa->sin.sin_port = htons((uint16_t) port); if (mg_resolve_from_hosts_file(host, sa) != 0) { /* * if resolving from hosts file failed and the host * we are trying to resolve is `localhost` - we should * try to resolve it using `gethostbyname` and do not try * to resolve it via DNS server if gethostbyname has failed too */ if (mg_ncasecmp(host, "localhost", 9) != 0) { return 0; } #if MG_ENABLE_SYNC_RESOLVER if (!mg_resolve2(host, &sa->sin.sin_addr)) { return -1; } #else return -1; #endif } #endif } else if (sscanf(str, ":%u%n", &port, &len) == 1 || sscanf(str, "%u%n", &port, &len) == 1) { /* If only port is specified, bind to IPv4, INADDR_ANY */ sa->sin.sin_port = htons((uint16_t) port); } else { return -1; } /* Required for MG_ENABLE_ASYNC_RESOLVER=0 */ (void) host; (void) host_len; ch = str[len]; /* Character that follows the address */ return port < 0xffffUL && (ch == '\0' || ch == ',' || isspace(ch)) ? len : -1; } struct mg_connection *mg_if_accept_new_conn(struct mg_connection *lc) { struct mg_add_sock_opts opts; struct mg_connection *nc; memset(&opts, 0, sizeof(opts)); nc = mg_create_connection(lc->mgr, lc->handler, opts); if (nc == NULL) return NULL; nc->listener = lc; nc->proto_handler = lc->proto_handler; nc->user_data = lc->user_data; nc->recv_mbuf_limit = lc->recv_mbuf_limit; nc->iface = lc->iface; if (lc->flags & MG_F_SSL) nc->flags |= MG_F_SSL; mg_add_conn(nc->mgr, nc); DBG(("%p %p %d %d", lc, nc, nc->sock, (int) nc->flags)); return nc; } void mg_if_accept_tcp_cb(struct mg_connection *nc, union socket_address *sa, size_t sa_len) { (void) sa_len; nc->sa = *sa; mg_call(nc, NULL, nc->user_data, MG_EV_ACCEPT, &nc->sa); } void mg_send(struct mg_connection *nc, const void *buf, int len) { nc->last_io_time = (time_t) mg_time(); if (nc->flags & MG_F_UDP) { nc->iface->vtable->udp_send(nc, buf, len); } else { nc->iface->vtable->tcp_send(nc, buf, len); } } void mg_if_sent_cb(struct mg_connection *nc, int num_sent) { DBG(("%p %d", nc, num_sent)); #if !defined(NO_LIBC) && MG_ENABLE_HEXDUMP if (nc->mgr && nc->mgr->hexdump_file != NULL) { char *buf = nc->send_mbuf.buf; mg_hexdump_connection(nc, nc->mgr->hexdump_file, buf, num_sent, MG_EV_SEND); } #endif if (num_sent < 0) { nc->flags |= MG_F_CLOSE_IMMEDIATELY; } else { mbuf_remove(&nc->send_mbuf, num_sent); mbuf_trim(&nc->send_mbuf); } mg_call(nc, NULL, nc->user_data, MG_EV_SEND, &num_sent); } MG_INTERNAL void mg_recv_common(struct mg_connection *nc, void *buf, int len, int own) { DBG(("%p %d %u", nc, len, (unsigned int) nc->recv_mbuf.len)); #if !defined(NO_LIBC) && MG_ENABLE_HEXDUMP if (nc->mgr && nc->mgr->hexdump_file != NULL) { mg_hexdump_connection(nc, nc->mgr->hexdump_file, buf, len, MG_EV_RECV); } #endif if (nc->flags & MG_F_CLOSE_IMMEDIATELY) { DBG(("%p discarded %d bytes", nc, len)); /* * This connection will not survive next poll. Do not deliver events, * send data to /dev/null without acking. */ if (own) { MG_FREE(buf); } return; } nc->last_io_time = (time_t) mg_time(); if (!own) { mbuf_append(&nc->recv_mbuf, buf, len); } else if (nc->recv_mbuf.len == 0) { /* Adopt buf as recv_mbuf's backing store. */ mbuf_free(&nc->recv_mbuf); nc->recv_mbuf.buf = (char *) buf; nc->recv_mbuf.size = nc->recv_mbuf.len = len; } else { mbuf_append(&nc->recv_mbuf, buf, len); MG_FREE(buf); } mg_call(nc, NULL, nc->user_data, MG_EV_RECV, &len); } void mg_if_recv_tcp_cb(struct mg_connection *nc, void *buf, int len, int own) { mg_recv_common(nc, buf, len, own); } void mg_if_recv_udp_cb(struct mg_connection *nc, void *buf, int len, union socket_address *sa, size_t sa_len) { assert(nc->flags & MG_F_UDP); DBG(("%p %u", nc, (unsigned int) len)); if (nc->flags & MG_F_LISTENING) { struct mg_connection *lc = nc; /* * Do we have an existing connection for this source? * This is very inefficient for long connection lists. */ for (nc = mg_next(lc->mgr, NULL); nc != NULL; nc = mg_next(lc->mgr, nc)) { if (memcmp(&nc->sa.sa, &sa->sa, sa_len) == 0 && nc->listener == lc) { break; } } if (nc == NULL) { struct mg_add_sock_opts opts; memset(&opts, 0, sizeof(opts)); /* Create fake connection w/out sock initialization */ nc = mg_create_connection_base(lc->mgr, lc->handler, opts); if (nc != NULL) { nc->sock = lc->sock; nc->listener = lc; nc->sa = *sa; nc->proto_handler = lc->proto_handler; nc->user_data = lc->user_data; nc->recv_mbuf_limit = lc->recv_mbuf_limit; nc->flags = MG_F_UDP; /* * Long-lived UDP "connections" i.e. interactions that involve more * than one request and response are rare, most are transactional: * response is sent and the "connection" is closed. Or - should be. * But users (including ourselves) tend to forget about that part, * because UDP is connectionless and one does not think about * processing a UDP request as handling a connection that needs to be * closed. Thus, we begin with SEND_AND_CLOSE flag set, which should * be a reasonable default for most use cases, but it is possible to * turn it off the connection should be kept alive after processing. */ nc->flags |= MG_F_SEND_AND_CLOSE; mg_add_conn(lc->mgr, nc); mg_call(nc, NULL, nc->user_data, MG_EV_ACCEPT, &nc->sa); } else { DBG(("OOM")); /* No return here, we still need to drop on the floor */ } } } if (nc != NULL) { mg_recv_common(nc, buf, len, 1); } else { /* Drop on the floor. */ MG_FREE(buf); } } /* * Schedules an async connect for a resolved address and proto. * Called from two places: `mg_connect_opt()` and from async resolver. * When called from the async resolver, it must trigger `MG_EV_CONNECT` event * with a failure flag to indicate connection failure. */ MG_INTERNAL struct mg_connection *mg_do_connect(struct mg_connection *nc, int proto, union socket_address *sa) { DBG(("%p %s://%s:%hu", nc, proto == SOCK_DGRAM ? "udp" : "tcp", inet_ntoa(sa->sin.sin_addr), ntohs(sa->sin.sin_port))); nc->flags |= MG_F_CONNECTING; if (proto == SOCK_DGRAM) { nc->iface->vtable->connect_udp(nc); } else { nc->iface->vtable->connect_tcp(nc, sa); } mg_add_conn(nc->mgr, nc); return nc; } void mg_if_connect_cb(struct mg_connection *nc, int err) { DBG(("%p connect, err=%d", nc, err)); nc->flags &= ~MG_F_CONNECTING; if (err != 0) { nc->flags |= MG_F_CLOSE_IMMEDIATELY; } mg_call(nc, NULL, nc->user_data, MG_EV_CONNECT, &err); } #if MG_ENABLE_ASYNC_RESOLVER /* * Callback for the async resolver on mg_connect_opt() call. * Main task of this function is to trigger MG_EV_CONNECT event with * either failure (and dealloc the connection) * or success (and proceed with connect() */ static void resolve_cb(struct mg_dns_message *msg, void *data, enum mg_resolve_err e) { struct mg_connection *nc = (struct mg_connection *) data; int i; int failure = -1; nc->flags &= ~MG_F_RESOLVING; if (msg != NULL) { /* * Take the first DNS A answer and run... */ for (i = 0; i < msg->num_answers; i++) { if (msg->answers[i].rtype == MG_DNS_A_RECORD) { /* * Async resolver guarantees that there is at least one answer. * TODO(lsm): handle IPv6 answers too */ mg_dns_parse_record_data(msg, &msg->answers[i], &nc->sa.sin.sin_addr, 4); mg_do_connect(nc, nc->flags & MG_F_UDP ? SOCK_DGRAM : SOCK_STREAM, &nc->sa); return; } } } if (e == MG_RESOLVE_TIMEOUT) { double now = mg_time(); mg_call(nc, NULL, nc->user_data, MG_EV_TIMER, &now); } /* * If we get there was no MG_DNS_A_RECORD in the answer */ mg_call(nc, NULL, nc->user_data, MG_EV_CONNECT, &failure); mg_call(nc, NULL, nc->user_data, MG_EV_CLOSE, NULL); mg_destroy_conn(nc, 1 /* destroy_if */); } #endif struct mg_connection *mg_connect(struct mg_mgr *mgr, const char *address, MG_CB(mg_event_handler_t callback, void *user_data)) { struct mg_connect_opts opts; memset(&opts, 0, sizeof(opts)); return mg_connect_opt(mgr, address, MG_CB(callback, user_data), opts); } struct mg_connection *mg_connect_opt(struct mg_mgr *mgr, const char *address, MG_CB(mg_event_handler_t callback, void *user_data), struct mg_connect_opts opts) { struct mg_connection *nc = NULL; int proto, rc; struct mg_add_sock_opts add_sock_opts; char host[MG_MAX_HOST_LEN]; MG_COPY_COMMON_CONNECTION_OPTIONS(&add_sock_opts, &opts); if ((nc = mg_create_connection(mgr, callback, add_sock_opts)) == NULL) { return NULL; } if ((rc = mg_parse_address(address, &nc->sa, &proto, host, sizeof(host))) < 0) { /* Address is malformed */ MG_SET_PTRPTR(opts.error_string, "cannot parse address"); mg_destroy_conn(nc, 1 /* destroy_if */); return NULL; } nc->flags |= opts.flags & _MG_ALLOWED_CONNECT_FLAGS_MASK; nc->flags |= (proto == SOCK_DGRAM) ? MG_F_UDP : 0; #if MG_ENABLE_CALLBACK_USERDATA nc->user_data = user_data; #else nc->user_data = opts.user_data; #endif #if MG_ENABLE_SSL DBG(("%p %s %s,%s,%s", nc, address, (opts.ssl_cert ? opts.ssl_cert : "-"), (opts.ssl_key ? opts.ssl_key : "-"), (opts.ssl_ca_cert ? opts.ssl_ca_cert : "-"))); if (opts.ssl_cert != NULL || opts.ssl_ca_cert != NULL || opts.ssl_psk_identity != NULL) { const char *err_msg = NULL; struct mg_ssl_if_conn_params params; if (nc->flags & MG_F_UDP) { MG_SET_PTRPTR(opts.error_string, "SSL for UDP is not supported"); mg_destroy_conn(nc, 1 /* destroy_if */); return NULL; } memset(¶ms, 0, sizeof(params)); params.cert = opts.ssl_cert; params.key = opts.ssl_key; params.ca_cert = opts.ssl_ca_cert; params.cipher_suites = opts.ssl_cipher_suites; params.psk_identity = opts.ssl_psk_identity; params.psk_key = opts.ssl_psk_key; if (opts.ssl_ca_cert != NULL) { if (opts.ssl_server_name != NULL) { if (strcmp(opts.ssl_server_name, "*") != 0) { params.server_name = opts.ssl_server_name; } } else if (rc == 0) { /* If it's a DNS name, use host. */ params.server_name = host; } } if (mg_ssl_if_conn_init(nc, ¶ms, &err_msg) != MG_SSL_OK) { MG_SET_PTRPTR(opts.error_string, err_msg); mg_destroy_conn(nc, 1 /* destroy_if */); return NULL; } nc->flags |= MG_F_SSL; } #endif /* MG_ENABLE_SSL */ if (rc == 0) { #if MG_ENABLE_ASYNC_RESOLVER /* * DNS resolution is required for host. * mg_parse_address() fills port in nc->sa, which we pass to resolve_cb() */ struct mg_connection *dns_conn = NULL; struct mg_resolve_async_opts o; memset(&o, 0, sizeof(o)); o.dns_conn = &dns_conn; o.nameserver = opts.nameserver; if (mg_resolve_async_opt(nc->mgr, host, MG_DNS_A_RECORD, resolve_cb, nc, o) != 0) { MG_SET_PTRPTR(opts.error_string, "cannot schedule DNS lookup"); mg_destroy_conn(nc, 1 /* destroy_if */); return NULL; } nc->priv_2 = dns_conn; nc->flags |= MG_F_RESOLVING; return nc; #else MG_SET_PTRPTR(opts.error_string, "Resolver is disabled"); mg_destroy_conn(nc, 1 /* destroy_if */); return NULL; #endif } else { /* Address is parsed and resolved to IP. proceed with connect() */ return mg_do_connect(nc, proto, &nc->sa); } } struct mg_connection *mg_bind(struct mg_mgr *srv, const char *address, MG_CB(mg_event_handler_t event_handler, void *user_data)) { struct mg_bind_opts opts; memset(&opts, 0, sizeof(opts)); return mg_bind_opt(srv, address, MG_CB(event_handler, user_data), opts); } struct mg_connection *mg_bind_opt(struct mg_mgr *mgr, const char *address, MG_CB(mg_event_handler_t callback, void *user_data), struct mg_bind_opts opts) { union socket_address sa; struct mg_connection *nc = NULL; int proto, rc; struct mg_add_sock_opts add_sock_opts; char host[MG_MAX_HOST_LEN]; #if MG_ENABLE_CALLBACK_USERDATA opts.user_data = user_data; #endif if (callback == NULL) { MG_SET_PTRPTR(opts.error_string, "handler is required"); return NULL; } MG_COPY_COMMON_CONNECTION_OPTIONS(&add_sock_opts, &opts); if (mg_parse_address(address, &sa, &proto, host, sizeof(host)) <= 0) { MG_SET_PTRPTR(opts.error_string, "cannot parse address"); return NULL; } nc = mg_create_connection(mgr, callback, add_sock_opts); if (nc == NULL) { return NULL; } nc->sa = sa; nc->flags |= MG_F_LISTENING; if (proto == SOCK_DGRAM) nc->flags |= MG_F_UDP; #if MG_ENABLE_SSL DBG(("%p %s %s,%s,%s", nc, address, (opts.ssl_cert ? opts.ssl_cert : "-"), (opts.ssl_key ? opts.ssl_key : "-"), (opts.ssl_ca_cert ? opts.ssl_ca_cert : "-"))); if (opts.ssl_cert != NULL || opts.ssl_ca_cert != NULL) { const char *err_msg = NULL; struct mg_ssl_if_conn_params params; if (nc->flags & MG_F_UDP) { MG_SET_PTRPTR(opts.error_string, "SSL for UDP is not supported"); mg_destroy_conn(nc, 1 /* destroy_if */); return NULL; } memset(¶ms, 0, sizeof(params)); params.cert = opts.ssl_cert; params.key = opts.ssl_key; params.ca_cert = opts.ssl_ca_cert; params.cipher_suites = opts.ssl_cipher_suites; if (mg_ssl_if_conn_init(nc, ¶ms, &err_msg) != MG_SSL_OK) { MG_SET_PTRPTR(opts.error_string, err_msg); mg_destroy_conn(nc, 1 /* destroy_if */); return NULL; } nc->flags |= MG_F_SSL; } #endif /* MG_ENABLE_SSL */ if (nc->flags & MG_F_UDP) { rc = nc->iface->vtable->listen_udp(nc, &nc->sa); } else { rc = nc->iface->vtable->listen_tcp(nc, &nc->sa); } if (rc != 0) { DBG(("Failed to open listener: %d", rc)); MG_SET_PTRPTR(opts.error_string, "failed to open listener"); mg_destroy_conn(nc, 1 /* destroy_if */); return NULL; } mg_add_conn(nc->mgr, nc); return nc; } struct mg_connection *mg_next(struct mg_mgr *s, struct mg_connection *conn) { return conn == NULL ? s->active_connections : conn->next; } #if MG_ENABLE_BROADCAST void mg_broadcast(struct mg_mgr *mgr, mg_event_handler_t cb, void *data, size_t len) { struct ctl_msg ctl_msg; /* * Mongoose manager has a socketpair, `struct mg_mgr::ctl`, * where `mg_broadcast()` pushes the message. * `mg_mgr_poll()` wakes up, reads a message from the socket pair, and calls * specified callback for each connection. Thus the callback function executes * in event manager thread. */ if (mgr->ctl[0] != INVALID_SOCKET && data != NULL && len < sizeof(ctl_msg.message)) { size_t dummy; ctl_msg.callback = cb; memcpy(ctl_msg.message, data, len); dummy = MG_SEND_FUNC(mgr->ctl[0], (char *) &ctl_msg, offsetof(struct ctl_msg, message) + len, 0); dummy = MG_RECV_FUNC(mgr->ctl[0], (char *) &len, 1, 0); (void) dummy; /* https://gcc.gnu.org/bugzilla/show_bug.cgi?id=25509 */ } } #endif /* MG_ENABLE_BROADCAST */ static int isbyte(int n) { return n >= 0 && n <= 255; } static int parse_net(const char *spec, uint32_t *net, uint32_t *mask) { int n, a, b, c, d, slash = 32, len = 0; if ((sscanf(spec, "%d.%d.%d.%d/%d%n", &a, &b, &c, &d, &slash, &n) == 5 || sscanf(spec, "%d.%d.%d.%d%n", &a, &b, &c, &d, &n) == 4) && isbyte(a) && isbyte(b) && isbyte(c) && isbyte(d) && slash >= 0 && slash < 33) { len = n; *net = ((uint32_t) a << 24) | ((uint32_t) b << 16) | ((uint32_t) c << 8) | d; *mask = slash ? 0xffffffffU << (32 - slash) : 0; } return len; } int mg_check_ip_acl(const char *acl, uint32_t remote_ip) { int allowed, flag; uint32_t net, mask; struct mg_str vec; /* If any ACL is set, deny by default */ allowed = (acl == NULL || *acl == '\0') ? '+' : '-'; while ((acl = mg_next_comma_list_entry(acl, &vec, NULL)) != NULL) { flag = vec.p[0]; if ((flag != '+' && flag != '-') || parse_net(&vec.p[1], &net, &mask) == 0) { return -1; } if (net == (remote_ip & mask)) { allowed = flag; } } DBG(("%08x %c", (unsigned int) remote_ip, allowed)); return allowed == '+'; } /* Move data from one connection to another */ void mg_forward(struct mg_connection *from, struct mg_connection *to) { mg_send(to, from->recv_mbuf.buf, from->recv_mbuf.len); mbuf_remove(&from->recv_mbuf, from->recv_mbuf.len); } double mg_set_timer(struct mg_connection *c, double timestamp) { double result = c->ev_timer_time; c->ev_timer_time = timestamp; /* * If this connection is resolving, it's not in the list of active * connections, so not processed yet. It has a DNS resolver connection * linked to it. Set up a timer for the DNS connection. */ DBG(("%p %p %d -> %lu", c, c->priv_2, (c->flags & MG_F_RESOLVING ? 1 : 0), (unsigned long) timestamp)); if ((c->flags & MG_F_RESOLVING) && c->priv_2 != NULL) { ((struct mg_connection *) c->priv_2)->ev_timer_time = timestamp; } return result; } void mg_sock_set(struct mg_connection *nc, sock_t sock) { if (sock != INVALID_SOCKET) { nc->iface->vtable->sock_set(nc, sock); } } void mg_if_get_conn_addr(struct mg_connection *nc, int remote, union socket_address *sa) { nc->iface->vtable->get_conn_addr(nc, remote, sa); } struct mg_connection *mg_add_sock_opt(struct mg_mgr *s, sock_t sock, MG_CB(mg_event_handler_t callback, void *user_data), struct mg_add_sock_opts opts) { struct mg_connection *nc =NULL;//wyf #if MG_ENABLE_CALLBACK_USERDATA opts.user_data = user_data; #endif nc = mg_create_connection_base(s, callback, opts);//wyf if (nc != NULL) { mg_sock_set(nc, sock); mg_add_conn(nc->mgr, nc); } return nc; } struct mg_connection *mg_add_sock(struct mg_mgr *s, sock_t sock, MG_CB(mg_event_handler_t callback, void *user_data)) { struct mg_add_sock_opts opts; memset(&opts, 0, sizeof(opts)); return mg_add_sock_opt(s, sock, MG_CB(callback, user_data), opts); } double mg_time(void) { return cs_time(); } #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_net_if_socket.h" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #ifndef CS_MONGOOSE_SRC_NET_IF_SOCKET_H_ #define CS_MONGOOSE_SRC_NET_IF_SOCKET_H_ /* Amalgamated: #include "mg_net_if.h" */ #ifdef __cplusplus extern "C" { #endif /* __cplusplus */ #ifndef MG_ENABLE_NET_IF_SOCKET #define MG_ENABLE_NET_IF_SOCKET MG_NET_IF == MG_NET_IF_SOCKET #endif extern const struct mg_iface_vtable mg_socket_iface_vtable; #ifdef __cplusplus } #endif /* __cplusplus */ #endif /* CS_MONGOOSE_SRC_NET_IF_SOCKET_H_ */ #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_net_if_socks.h" #endif /* * Copyright (c) 2014-2017 Cesanta Software Limited * All rights reserved */ #ifndef CS_MONGOOSE_SRC_NET_IF_SOCKS_H_ #define CS_MONGOOSE_SRC_NET_IF_SOCKS_H_ #if MG_ENABLE_SOCKS /* Amalgamated: #include "mg_net_if.h" */ #ifdef __cplusplus extern "C" { #endif /* __cplusplus */ extern const struct mg_iface_vtable mg_socks_iface_vtable; #ifdef __cplusplus } #endif /* __cplusplus */ #endif /* MG_ENABLE_SOCKS */ #endif /* CS_MONGOOSE_SRC_NET_IF_SOCKS_H_ */ #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_net_if.c" #endif /* Amalgamated: #include "mg_net_if.h" */ /* Amalgamated: #include "mg_internal.h" */ /* Amalgamated: #include "mg_net_if_socket.h" */ extern const struct mg_iface_vtable mg_default_iface_vtable; const struct mg_iface_vtable *mg_ifaces[] = { &mg_default_iface_vtable, }; int mg_num_ifaces = (int) (sizeof(mg_ifaces) / sizeof(mg_ifaces[0])); struct mg_iface *mg_if_create_iface(const struct mg_iface_vtable *vtable, struct mg_mgr *mgr) { struct mg_iface *iface = (struct mg_iface *) MG_CALLOC(1, sizeof(*iface)); iface->mgr = mgr; iface->data = NULL; iface->vtable = vtable; return iface; } struct mg_iface *mg_find_iface(struct mg_mgr *mgr, const struct mg_iface_vtable *vtable, struct mg_iface *from) { int i = 0; if (from != NULL) { for (i = 0; i < mgr->num_ifaces; i++) { if (mgr->ifaces[i] == from) { i++; break; } } } for (; i < mgr->num_ifaces; i++) { if (mgr->ifaces[i]->vtable == vtable) { return mgr->ifaces[i]; } } return NULL; } #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_net_if_socket.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #if MG_ENABLE_NET_IF_SOCKET /* Amalgamated: #include "mg_net_if_socket.h" */ /* Amalgamated: #include "mg_internal.h" */ /* Amalgamated: #include "mg_util.h" */ #define MG_TCP_RECV_BUFFER_SIZE 1024 #define MG_UDP_RECV_BUFFER_SIZE 1500 static sock_t mg_open_listening_socket(union socket_address *sa, int type, int proto); #if MG_ENABLE_SSL static void mg_ssl_begin(struct mg_connection *nc); #endif void mg_set_non_blocking_mode(sock_t sock) { #ifdef _WIN32 unsigned long on = 1; ioctlsocket(sock, FIONBIO, &on); #else int flags = fcntl(sock, F_GETFL, 0); fcntl(sock, F_SETFL, flags | O_NONBLOCK); #endif } static int mg_is_error(void) { int err = mg_get_errno(); return err != EINPROGRESS && err != EWOULDBLOCK #ifndef WINCE && err != EAGAIN && err != EINTR #endif #ifdef _WIN32 && WSAGetLastError() != WSAEINTR && WSAGetLastError() != WSAEWOULDBLOCK #endif ; } void mg_socket_if_connect_tcp(struct mg_connection *nc, const union socket_address *sa) { int rc, proto = 0; nc->sock = socket(AF_INET, SOCK_STREAM, proto); if (nc->sock == INVALID_SOCKET) { nc->err = mg_get_errno() ? mg_get_errno() : 1; return; } #if !defined(MG_ESP8266) mg_set_non_blocking_mode(nc->sock); #endif rc = connect(nc->sock, &sa->sa, sizeof(sa->sin)); nc->err = rc < 0 && mg_is_error() ? mg_get_errno() : 0; DBG(("%p sock %d rc %d errno %d err %d", nc, nc->sock, rc, mg_get_errno(), nc->err)); } void mg_socket_if_connect_udp(struct mg_connection *nc) { nc->sock = socket(AF_INET, SOCK_DGRAM, 0); if (nc->sock == INVALID_SOCKET) { nc->err = mg_get_errno() ? mg_get_errno() : 1; return; } if (nc->flags & MG_F_ENABLE_BROADCAST) { int optval = 1; if (setsockopt(nc->sock, SOL_SOCKET, SO_BROADCAST, (const char *) &optval, sizeof(optval)) < 0) { nc->err = mg_get_errno() ? mg_get_errno() : 1; return; } } nc->err = 0; } int mg_socket_if_listen_tcp(struct mg_connection *nc, union socket_address *sa) { int proto = 0; sock_t sock = mg_open_listening_socket(sa, SOCK_STREAM, proto); if (sock == INVALID_SOCKET) { return (mg_get_errno() ? mg_get_errno() : 1); } mg_sock_set(nc, sock); return 0; } int mg_socket_if_listen_udp(struct mg_connection *nc, union socket_address *sa) { sock_t sock = mg_open_listening_socket(sa, SOCK_DGRAM, 0); if (sock == INVALID_SOCKET) return (mg_get_errno() ? mg_get_errno() : 1); mg_sock_set(nc, sock); return 0; } void mg_socket_if_tcp_send(struct mg_connection *nc, const void *buf, size_t len) { mbuf_append(&nc->send_mbuf, buf, len); } void mg_socket_if_udp_send(struct mg_connection *nc, const void *buf, size_t len) { mbuf_append(&nc->send_mbuf, buf, len); } void mg_socket_if_recved(struct mg_connection *nc, size_t len) { (void) nc; (void) len; } int mg_socket_if_create_conn(struct mg_connection *nc) { (void) nc; return 1; } void mg_socket_if_destroy_conn(struct mg_connection *nc) { if (nc->sock == INVALID_SOCKET) return; if (!(nc->flags & MG_F_UDP)) { closesocket(nc->sock); } else { /* Only close outgoing UDP sockets or listeners. */ if (nc->listener == NULL) closesocket(nc->sock); } nc->sock = INVALID_SOCKET; } static int mg_accept_conn(struct mg_connection *lc) { struct mg_connection *nc; union socket_address sa; socklen_t sa_len = sizeof(sa); /* NOTE(lsm): on Windows, sock is always > FD_SETSIZE */ sock_t sock = accept(lc->sock, &sa.sa, &sa_len); if (sock == INVALID_SOCKET) { if (mg_is_error()) DBG(("%p: failed to accept: %d", lc, mg_get_errno())); return 0; } nc = mg_if_accept_new_conn(lc); if (nc == NULL) { closesocket(sock); return 0; } DBG(("%p conn from %s:%d", nc, inet_ntoa(sa.sin.sin_addr), ntohs(sa.sin.sin_port))); mg_sock_set(nc, sock); #if MG_ENABLE_SSL if (lc->flags & MG_F_SSL) { if (mg_ssl_if_conn_accept(nc, lc) != MG_SSL_OK) mg_close_conn(nc); } else #endif { mg_if_accept_tcp_cb(nc, &sa, sa_len); } return 1; } /* 'sa' must be an initialized address to bind to */ static sock_t mg_open_listening_socket(union socket_address *sa, int type, int proto) { socklen_t sa_len = (sa->sa.sa_family == AF_INET) ? sizeof(sa->sin) : sizeof(sa->sin6); sock_t sock = INVALID_SOCKET; #if !MG_LWIP int on = 1; #endif if ((sock = socket(sa->sa.sa_family, type, proto)) != INVALID_SOCKET && #if !MG_LWIP /* LWIP doesn't support either */ #if defined(_WIN32) && defined(SO_EXCLUSIVEADDRUSE) && !defined(WINCE) /* "Using SO_REUSEADDR and SO_EXCLUSIVEADDRUSE" http://goo.gl/RmrFTm */ !setsockopt(sock, SOL_SOCKET, SO_EXCLUSIVEADDRUSE, (void *) &on, sizeof(on)) && #endif #if !defined(_WIN32) || !defined(SO_EXCLUSIVEADDRUSE) /* * SO_RESUSEADDR is not enabled on Windows because the semantics of * SO_REUSEADDR on UNIX and Windows is different. On Windows, * SO_REUSEADDR allows to bind a socket to a port without error even if * the port is already open by another program. This is not the behavior * SO_REUSEADDR was designed for, and leads to hard-to-track failure * scenarios. Therefore, SO_REUSEADDR was disabled on Windows unless * SO_EXCLUSIVEADDRUSE is supported and set on a socket. */ !setsockopt(sock, SOL_SOCKET, SO_REUSEADDR, (void *) &on, sizeof(on)) && #endif #endif /* !MG_LWIP */ !bind(sock, &sa->sa, sa_len) && (type == SOCK_DGRAM || listen(sock, SOMAXCONN) == 0)) { #if !MG_LWIP mg_set_non_blocking_mode(sock); /* In case port was set to 0, get the real port number */ (void) getsockname(sock, &sa->sa, &sa_len); #endif } else if (sock != INVALID_SOCKET) { closesocket(sock); sock = INVALID_SOCKET; } return sock; } static void mg_write_to_socket(struct mg_connection *nc) { struct mbuf *io = &nc->send_mbuf; int n = 0; #if MG_LWIP /* With LWIP we don't know if the socket is ready */ if (io->len == 0) return; #endif assert(io->len > 0); if (nc->flags & MG_F_UDP) { int n = sendto(nc->sock, io->buf, io->len, 0, &nc->sa.sa, sizeof(nc->sa.sin)); DBG(("%p %d %d %d %s:%hu", nc, nc->sock, n, mg_get_errno(), inet_ntoa(nc->sa.sin.sin_addr), ntohs(nc->sa.sin.sin_port))); mg_if_sent_cb(nc, n); return; } #if MG_ENABLE_SSL if (nc->flags & MG_F_SSL) { if (nc->flags & MG_F_SSL_HANDSHAKE_DONE) { n = mg_ssl_if_write(nc, io->buf, io->len); DBG(("%p %d bytes -> %d (SSL)", nc, n, nc->sock)); if (n < 0) { if (n != MG_SSL_WANT_READ && n != MG_SSL_WANT_WRITE) { nc->flags |= MG_F_CLOSE_IMMEDIATELY; } return; } else { /* Successful SSL operation, clear off SSL wait flags */ nc->flags &= ~(MG_F_WANT_READ | MG_F_WANT_WRITE); } } else { mg_ssl_begin(nc); return; } } else #endif { n = (int) MG_SEND_FUNC(nc->sock, io->buf, io->len, 0); DBG(("%p %d bytes -> %d", nc, n, nc->sock)); } mg_if_sent_cb(nc, n); } MG_INTERNAL size_t recv_avail_size(struct mg_connection *conn, size_t max) { size_t avail; if (conn->recv_mbuf_limit < conn->recv_mbuf.len) return 0; avail = conn->recv_mbuf_limit - conn->recv_mbuf.len; return avail > max ? max : avail; } static void mg_handle_tcp_read(struct mg_connection *conn) { int n = 0; char *buf = (char *) MG_MALLOC(MG_TCP_RECV_BUFFER_SIZE); if (buf == NULL) { DBG(("OOM")); return; } #if MG_ENABLE_SSL if (conn->flags & MG_F_SSL) { if (conn->flags & MG_F_SSL_HANDSHAKE_DONE) { /* SSL library may have more bytes ready to read than we ask to read. * Therefore, read in a loop until we read everything. Without the loop, * we skip to the next select() cycle which can just timeout. */ while ((n = mg_ssl_if_read(conn, buf, MG_TCP_RECV_BUFFER_SIZE)) > 0) { DBG(("%p %d bytes <- %d (SSL)", conn, n, conn->sock)); mg_if_recv_tcp_cb(conn, buf, n, 1 /* own */); buf = NULL; if (conn->flags & MG_F_CLOSE_IMMEDIATELY) break; /* buf has been freed, we need a new one. */ buf = (char *) MG_MALLOC(MG_TCP_RECV_BUFFER_SIZE); if (buf == NULL) break; } MG_FREE(buf); if (n < 0 && n != MG_SSL_WANT_READ) conn->flags |= MG_F_CLOSE_IMMEDIATELY; } else { MG_FREE(buf); mg_ssl_begin(conn); return; } } else #endif { n = (int) MG_RECV_FUNC(conn->sock, buf, recv_avail_size(conn, MG_TCP_RECV_BUFFER_SIZE), 0); DBG(("%p %d bytes (PLAIN) <- %d", conn, n, conn->sock)); if (n > 0) { mg_if_recv_tcp_cb(conn, buf, n, 1 /* own */); } else { MG_FREE(buf); } if (n == 0) { /* Orderly shutdown of the socket, try flushing output. */ conn->flags |= MG_F_SEND_AND_CLOSE; } else if (n < 0 && mg_is_error()) { conn->flags |= MG_F_CLOSE_IMMEDIATELY; } } } static int mg_recvfrom(struct mg_connection *nc, union socket_address *sa, socklen_t *sa_len, char **buf) { int n; *buf = (char *) MG_MALLOC(MG_UDP_RECV_BUFFER_SIZE); if (*buf == NULL) { DBG(("Out of memory")); return -ENOMEM; } n = recvfrom(nc->sock, *buf, MG_UDP_RECV_BUFFER_SIZE, 0, &sa->sa, sa_len); if (n <= 0) { DBG(("%p recvfrom: %s", nc, strerror(mg_get_errno()))); MG_FREE(*buf); } return n; } static void mg_handle_udp_read(struct mg_connection *nc) { char *buf = NULL; union socket_address sa; socklen_t sa_len = sizeof(sa); int n = mg_recvfrom(nc, &sa, &sa_len, &buf); DBG(("%p %d bytes from %s:%d", nc, n, inet_ntoa(nc->sa.sin.sin_addr), ntohs(nc->sa.sin.sin_port))); mg_if_recv_udp_cb(nc, buf, n, &sa, sa_len); } #if MG_ENABLE_SSL static void mg_ssl_begin(struct mg_connection *nc) { int server_side = (nc->listener != NULL); enum mg_ssl_if_result res = mg_ssl_if_handshake(nc); DBG(("%p %d res %d", nc, server_side, res)); if (res == MG_SSL_OK) { nc->flags |= MG_F_SSL_HANDSHAKE_DONE; nc->flags &= ~(MG_F_WANT_READ | MG_F_WANT_WRITE); if (server_side) { union socket_address sa; socklen_t sa_len = sizeof(sa); (void) getpeername(nc->sock, &sa.sa, &sa_len); mg_if_accept_tcp_cb(nc, &sa, sa_len); } else { mg_if_connect_cb(nc, 0); } } else if (res != MG_SSL_WANT_READ && res != MG_SSL_WANT_WRITE) { if (!server_side) { mg_if_connect_cb(nc, res); } nc->flags |= MG_F_CLOSE_IMMEDIATELY; } } #endif /* MG_ENABLE_SSL */ #define _MG_F_FD_CAN_READ 1 #define _MG_F_FD_CAN_WRITE 1 << 1 #define _MG_F_FD_ERROR 1 << 2 void mg_mgr_handle_conn(struct mg_connection *nc, int fd_flags, double now) { int worth_logging = fd_flags != 0 || (nc->flags & (MG_F_WANT_READ | MG_F_WANT_WRITE)); if (worth_logging) { DBG(("%p fd=%d fd_flags=%d nc_flags=%lu rmbl=%d smbl=%d", nc, nc->sock, fd_flags, nc->flags, (int) nc->recv_mbuf.len, (int) nc->send_mbuf.len)); } if (nc->flags & MG_F_CONNECTING) { if (fd_flags != 0) { int err = 0; #if !defined(MG_ESP8266) if (!(nc->flags & MG_F_UDP)) { socklen_t len = sizeof(err); int ret = getsockopt(nc->sock, SOL_SOCKET, SO_ERROR, (char *) &err, &len); if (ret != 0) { err = 1; } else if (err == EAGAIN || err == EWOULDBLOCK) { err = 0; } } #else /* * On ESP8266 we use blocking connect. */ err = nc->err; #endif #if MG_ENABLE_SSL if ((nc->flags & MG_F_SSL) && err == 0) { mg_ssl_begin(nc); } else { mg_if_connect_cb(nc, err); } #else mg_if_connect_cb(nc, err); #endif } else if (nc->err != 0) { mg_if_connect_cb(nc, nc->err); } } if (fd_flags & _MG_F_FD_CAN_READ) { if (nc->flags & MG_F_UDP) { mg_handle_udp_read(nc); } else { if (nc->flags & MG_F_LISTENING) { /* * We're not looping here, and accepting just one connection at * a time. The reason is that eCos does not respect non-blocking * flag on a listening socket and hangs in a loop. */ mg_accept_conn(nc); } else { mg_handle_tcp_read(nc); } } } if (!(nc->flags & MG_F_CLOSE_IMMEDIATELY)) { if ((fd_flags & _MG_F_FD_CAN_WRITE) && nc->send_mbuf.len > 0) { mg_write_to_socket(nc); } mg_if_poll(nc, (time_t) now); mg_if_timer(nc, now); } if (worth_logging) { DBG(("%p after fd=%d nc_flags=%lu rmbl=%d smbl=%d", nc, nc->sock, nc->flags, (int) nc->recv_mbuf.len, (int) nc->send_mbuf.len)); } } #if MG_ENABLE_BROADCAST static void mg_mgr_handle_ctl_sock(struct mg_mgr *mgr) { struct ctl_msg ctl_msg; int len = (int) MG_RECV_FUNC(mgr->ctl[1], (char *) &ctl_msg, sizeof(ctl_msg), 0); size_t dummy = MG_SEND_FUNC(mgr->ctl[1], ctl_msg.message, 1, 0); DBG(("read %d from ctl socket", len)); (void) dummy; /* https://gcc.gnu.org/bugzilla/show_bug.cgi?id=25509 */ if (len >= (int) sizeof(ctl_msg.callback) && ctl_msg.callback != NULL) { struct mg_connection *nc; for (nc = mg_next(mgr, NULL); nc != NULL; nc = mg_next(mgr, nc)) { ctl_msg.callback(nc, MG_EV_POLL, ctl_msg.message MG_UD_ARG(nc->user_data)); } } } #endif /* Associate a socket to a connection. */ void mg_socket_if_sock_set(struct mg_connection *nc, sock_t sock) { mg_set_non_blocking_mode(sock); mg_set_close_on_exec(sock); nc->sock = sock; DBG(("%p %d", nc, sock)); } void mg_socket_if_init(struct mg_iface *iface) { (void) iface; DBG(("%p using select()", iface->mgr)); #if MG_ENABLE_BROADCAST mg_socketpair(iface->mgr->ctl, SOCK_DGRAM); #endif } void mg_socket_if_free(struct mg_iface *iface) { (void) iface; } void mg_socket_if_add_conn(struct mg_connection *nc) { (void) nc; } void mg_socket_if_remove_conn(struct mg_connection *nc) { (void) nc; } void mg_add_to_set(sock_t sock, fd_set *set, sock_t *max_fd) { if (sock != INVALID_SOCKET #ifdef __unix__ && sock < (sock_t) FD_SETSIZE #endif ) { FD_SET(sock, set); if (*max_fd == INVALID_SOCKET || sock > *max_fd) { *max_fd = sock; } } } time_t mg_socket_if_poll(struct mg_iface *iface, int timeout_ms) { struct mg_mgr *mgr = iface->mgr; double now = mg_time(); double min_timer; struct mg_connection *nc, *tmp; struct timeval tv; fd_set read_set, write_set, err_set; sock_t max_fd = INVALID_SOCKET; int num_fds, num_ev, num_timers = 0; #ifdef __unix__ int try_dup = 1; #endif FD_ZERO(&read_set); FD_ZERO(&write_set); FD_ZERO(&err_set); #if MG_ENABLE_BROADCAST mg_add_to_set(mgr->ctl[1], &read_set, &max_fd); #endif /* * Note: it is ok to have connections with sock == INVALID_SOCKET in the list, * e.g. timer-only "connections". */ min_timer = 0; for (nc = mgr->active_connections, num_fds = 0; nc != NULL; nc = tmp) { tmp = nc->next; if (nc->sock != INVALID_SOCKET) { num_fds++; #ifdef __unix__ /* A hack to make sure all our file descriptos fit into FD_SETSIZE. */ if (nc->sock >= (sock_t) FD_SETSIZE && try_dup) { int new_sock = dup(nc->sock); if (new_sock >= 0) { if (new_sock < (sock_t) FD_SETSIZE) { closesocket(nc->sock); DBG(("new sock %d -> %d", nc->sock, new_sock)); nc->sock = new_sock; } else { closesocket(new_sock); DBG(("new sock is still larger than FD_SETSIZE, disregard")); try_dup = 0; } } else { try_dup = 0; } } #endif if (!(nc->flags & MG_F_WANT_WRITE) && nc->recv_mbuf.len < nc->recv_mbuf_limit && (!(nc->flags & MG_F_UDP) || nc->listener == NULL)) { mg_add_to_set(nc->sock, &read_set, &max_fd); } if (((nc->flags & MG_F_CONNECTING) && !(nc->flags & MG_F_WANT_READ)) || (nc->send_mbuf.len > 0 && !(nc->flags & MG_F_CONNECTING))) { mg_add_to_set(nc->sock, &write_set, &max_fd); mg_add_to_set(nc->sock, &err_set, &max_fd); } } if (nc->ev_timer_time > 0) { if (num_timers == 0 || nc->ev_timer_time < min_timer) { min_timer = nc->ev_timer_time; } num_timers++; } } /* * If there is a timer to be fired earlier than the requested timeout, * adjust the timeout. */ if (num_timers > 0) { double timer_timeout_ms = (min_timer - mg_time()) * 1000 + 1 /* rounding */; if (timer_timeout_ms < timeout_ms) { timeout_ms = (int) timer_timeout_ms; } } if (timeout_ms < 0) timeout_ms = 0; tv.tv_sec = timeout_ms / 1000; tv.tv_usec = (timeout_ms % 1000) * 1000; num_ev = select((int) max_fd + 1, &read_set, &write_set, &err_set, &tv); now = mg_time(); #if 0 DBG(("select @ %ld num_ev=%d of %d, timeout=%d", (long) now, num_ev, num_fds, timeout_ms)); #endif #if MG_ENABLE_BROADCAST if (num_ev > 0 && mgr->ctl[1] != INVALID_SOCKET && FD_ISSET(mgr->ctl[1], &read_set)) { mg_mgr_handle_ctl_sock(mgr); } #endif for (nc = mgr->active_connections; nc != NULL; nc = tmp) { int fd_flags = 0; if (nc->sock != INVALID_SOCKET) { if (num_ev > 0) { fd_flags = (FD_ISSET(nc->sock, &read_set) && (!(nc->flags & MG_F_UDP) || nc->listener == NULL) ? _MG_F_FD_CAN_READ : 0) | (FD_ISSET(nc->sock, &write_set) ? _MG_F_FD_CAN_WRITE : 0) | (FD_ISSET(nc->sock, &err_set) ? _MG_F_FD_ERROR : 0); } #if MG_LWIP /* With LWIP socket emulation layer, we don't get write events for UDP */ if ((nc->flags & MG_F_UDP) && nc->listener == NULL) { fd_flags |= _MG_F_FD_CAN_WRITE; } #endif } tmp = nc->next; mg_mgr_handle_conn(nc, fd_flags, now); } for (nc = mgr->active_connections; nc != NULL; nc = tmp) { tmp = nc->next; if ((nc->flags & MG_F_CLOSE_IMMEDIATELY) || (nc->send_mbuf.len == 0 && (nc->flags & MG_F_SEND_AND_CLOSE))) { mg_close_conn(nc); } } return (time_t) now; } #if MG_ENABLE_BROADCAST MG_INTERNAL void mg_socketpair_close(sock_t *sock) { while (1) { if (closesocket(*sock) == -1 && errno == EINTR) continue; break; } *sock = INVALID_SOCKET; } MG_INTERNAL sock_t mg_socketpair_accept(sock_t sock, union socket_address *sa, socklen_t sa_len) { sock_t rc; while (1) { if ((rc = accept(sock, &sa->sa, &sa_len)) == INVALID_SOCKET && errno == EINTR) continue; break; } return rc; } int mg_socketpair(sock_t sp[2], int sock_type) { union socket_address sa; sock_t sock; socklen_t len = sizeof(sa.sin); int ret = 0; sock = sp[0] = sp[1] = INVALID_SOCKET; (void) memset(&sa, 0, sizeof(sa)); sa.sin.sin_family = AF_INET; sa.sin.sin_port = htons(0); sa.sin.sin_addr.s_addr = htonl(0x7f000001); /* 127.0.0.1 */ if ((sock = socket(AF_INET, sock_type, 0)) == INVALID_SOCKET) { } else if (bind(sock, &sa.sa, len) != 0) { } else if (sock_type == SOCK_STREAM && listen(sock, 1) != 0) { } else if (getsockname(sock, &sa.sa, &len) != 0) { } else if ((sp[0] = socket(AF_INET, sock_type, 0)) == INVALID_SOCKET) { } else if (connect(sp[0], &sa.sa, len) != 0) { } else if (sock_type == SOCK_DGRAM && (getsockname(sp[0], &sa.sa, &len) != 0 || connect(sock, &sa.sa, len) != 0)) { } else if ((sp[1] = (sock_type == SOCK_DGRAM ? sock : mg_socketpair_accept( sock, &sa, len))) == INVALID_SOCKET) { } else { mg_set_close_on_exec(sp[0]); mg_set_close_on_exec(sp[1]); if (sock_type == SOCK_STREAM) mg_socketpair_close(&sock); ret = 1; } if (!ret) { if (sp[0] != INVALID_SOCKET) mg_socketpair_close(&sp[0]); if (sp[1] != INVALID_SOCKET) mg_socketpair_close(&sp[1]); if (sock != INVALID_SOCKET) mg_socketpair_close(&sock); } return ret; } #endif /* MG_ENABLE_BROADCAST */ static void mg_sock_get_addr(sock_t sock, int remote, union socket_address *sa) { socklen_t slen = sizeof(*sa); memset(sa, 0, slen); if (remote) { getpeername(sock, &sa->sa, &slen); } else { getsockname(sock, &sa->sa, &slen); } } void mg_sock_to_str(sock_t sock, char *buf, size_t len, int flags) { union socket_address sa; mg_sock_get_addr(sock, flags & MG_SOCK_STRINGIFY_REMOTE, &sa); mg_sock_addr_to_str(&sa, buf, len, flags); } void mg_socket_if_get_conn_addr(struct mg_connection *nc, int remote, union socket_address *sa) { if ((nc->flags & MG_F_UDP) && remote) { memcpy(sa, &nc->sa, sizeof(*sa)); return; } mg_sock_get_addr(nc->sock, remote, sa); } /* clang-format off */ #define MG_SOCKET_IFACE_VTABLE \ { \ mg_socket_if_init, \ mg_socket_if_free, \ mg_socket_if_add_conn, \ mg_socket_if_remove_conn, \ mg_socket_if_poll, \ mg_socket_if_listen_tcp, \ mg_socket_if_listen_udp, \ mg_socket_if_connect_tcp, \ mg_socket_if_connect_udp, \ mg_socket_if_tcp_send, \ mg_socket_if_udp_send, \ mg_socket_if_recved, \ mg_socket_if_create_conn, \ mg_socket_if_destroy_conn, \ mg_socket_if_sock_set, \ mg_socket_if_get_conn_addr, \ } /* clang-format on */ const struct mg_iface_vtable mg_socket_iface_vtable = MG_SOCKET_IFACE_VTABLE; #if MG_NET_IF == MG_NET_IF_SOCKET const struct mg_iface_vtable mg_default_iface_vtable = MG_SOCKET_IFACE_VTABLE; #endif #endif /* MG_ENABLE_NET_IF_SOCKET */ #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_net_if_socks.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #if MG_ENABLE_SOCKS struct socksdata { char *proxy_addr; /* HOST:PORT of the socks5 proxy server */ struct mg_connection *s; /* Respective connection to the server */ struct mg_connection *c; /* Connection to the client */ struct mbuf tmp; /* Temporary buffer for sent data */ }; static void socks_if_disband(struct socksdata *d) { LOG(LL_DEBUG, ("disbanding proxy %p %p", d->c, d->s)); if (d->c) d->c->flags |= MG_F_SEND_AND_CLOSE; if (d->s) d->s->flags |= MG_F_SEND_AND_CLOSE; d->c = d->s = NULL; } static void socks_if_handler(struct mg_connection *c, int ev, void *ev_data) { struct socksdata *d = (struct socksdata *) c->user_data; if (ev == MG_EV_CONNECT) { int res = *(int *) ev_data; if (res == 0) { /* Send handshake to the proxy server */ unsigned char buf[] = {MG_SOCKS_VERSION, 1, MG_SOCKS_HANDSHAKE_NOAUTH}; mg_send(d->s, buf, sizeof(buf)); LOG(LL_DEBUG, ("Sent handshake to %s", d->proxy_addr)); } else { LOG(LL_ERROR, ("Cannot connect to %s: %d", d->proxy_addr, res)); d->c->flags |= MG_F_CLOSE_IMMEDIATELY; } } else if (ev == MG_EV_CLOSE) { socks_if_disband(d); } else if (ev == MG_EV_RECV) { /* Handle handshake reply */ if (!(c->flags & MG_SOCKS_HANDSHAKE_DONE)) { /* TODO(lsm): process IPv6 too */ unsigned char buf[10] = {MG_SOCKS_VERSION, MG_SOCKS_CMD_CONNECT, 0, MG_SOCKS_ADDR_IPV4}; if (c->recv_mbuf.len < 2) return; if ((unsigned char) c->recv_mbuf.buf[1] == MG_SOCKS_HANDSHAKE_FAILURE) { LOG(LL_ERROR, ("Server kicked us out")); socks_if_disband(d); return; } mbuf_remove(&c->recv_mbuf, 2); c->flags |= MG_SOCKS_HANDSHAKE_DONE; /* Send connect request */ memcpy(buf + 4, &d->c->sa.sin.sin_addr, 4); memcpy(buf + 8, &d->c->sa.sin.sin_port, 2); mg_send(c, buf, sizeof(buf)); } /* Process connect request */ if ((c->flags & MG_SOCKS_HANDSHAKE_DONE) && !(c->flags & MG_SOCKS_CONNECT_DONE)) { if (c->recv_mbuf.len < 10) return; if (c->recv_mbuf.buf[1] != MG_SOCKS_SUCCESS) { LOG(LL_ERROR, ("Socks connection error: %d", c->recv_mbuf.buf[1])); socks_if_disband(d); return; } mbuf_remove(&c->recv_mbuf, 10); c->flags |= MG_SOCKS_CONNECT_DONE; /* Connected. Move sent data from client, if any, to server */ if (d->s && d->c) { mbuf_append(&d->s->send_mbuf, d->tmp.buf, d->tmp.len); mbuf_free(&d->tmp); } } /* All flags are set, we're in relay mode */ if ((c->flags & MG_SOCKS_CONNECT_DONE) && d->c && d->s) { mbuf_append(&d->c->recv_mbuf, d->s->recv_mbuf.buf, d->s->recv_mbuf.len); mbuf_remove(&d->s->recv_mbuf, d->s->recv_mbuf.len); } } } static void mg_socks_if_connect_tcp(struct mg_connection *c, const union socket_address *sa) { struct socksdata *d = (struct socksdata *) c->iface->data; d->c = c; d->s = mg_connect(c->mgr, d->proxy_addr, socks_if_handler); d->s->user_data = d; LOG(LL_DEBUG, ("%p %s", c, d->proxy_addr)); (void) sa; } static void mg_socks_if_connect_udp(struct mg_connection *c) { (void) c; } static int mg_socks_if_listen_tcp(struct mg_connection *c, union socket_address *sa) { (void) c; (void) sa; return 0; } static int mg_socks_if_listen_udp(struct mg_connection *c, union socket_address *sa) { (void) c; (void) sa; return -1; } static void mg_socks_if_tcp_send(struct mg_connection *c, const void *buf, size_t len) { struct socksdata *d = (struct socksdata *) c->iface->data; LOG(LL_DEBUG, ("%p -> %p %d %d", c, buf, (int) len, (int) c->send_mbuf.len)); if (d && d->s && d->s->flags & MG_SOCKS_CONNECT_DONE) { mbuf_append(&d->s->send_mbuf, d->tmp.buf, d->tmp.len); mbuf_append(&d->s->send_mbuf, buf, len); mbuf_free(&d->tmp); } else { mbuf_append(&d->tmp, buf, len); } } static void mg_socks_if_udp_send(struct mg_connection *c, const void *buf, size_t len) { (void) c; (void) buf; (void) len; } static void mg_socks_if_recved(struct mg_connection *c, size_t len) { (void) c; (void) len; } static int mg_socks_if_create_conn(struct mg_connection *c) { (void) c; return 1; } static void mg_socks_if_destroy_conn(struct mg_connection *c) { c->iface->vtable->free(c->iface); MG_FREE(c->iface); c->iface = NULL; LOG(LL_DEBUG, ("%p", c)); } static void mg_socks_if_sock_set(struct mg_connection *c, sock_t sock) { (void) c; (void) sock; } static void mg_socks_if_init(struct mg_iface *iface) { (void) iface; } static void mg_socks_if_free(struct mg_iface *iface) { struct socksdata *d = (struct socksdata *) iface->data; LOG(LL_DEBUG, ("%p", iface)); if (d != NULL) { socks_if_disband(d); mbuf_free(&d->tmp); MG_FREE(d->proxy_addr); MG_FREE(d); iface->data = NULL; } } static void mg_socks_if_add_conn(struct mg_connection *c) { c->sock = INVALID_SOCKET; } static void mg_socks_if_remove_conn(struct mg_connection *c) { (void) c; } static time_t mg_socks_if_poll(struct mg_iface *iface, int timeout_ms) { LOG(LL_DEBUG, ("%p", iface)); (void) iface; (void) timeout_ms; return (time_t) cs_time(); } static void mg_socks_if_get_conn_addr(struct mg_connection *c, int remote, union socket_address *sa) { LOG(LL_DEBUG, ("%p", c)); (void) c; (void) remote; (void) sa; } const struct mg_iface_vtable mg_socks_iface_vtable = { mg_socks_if_init, mg_socks_if_free, mg_socks_if_add_conn, mg_socks_if_remove_conn, mg_socks_if_poll, mg_socks_if_listen_tcp, mg_socks_if_listen_udp, mg_socks_if_connect_tcp, mg_socks_if_connect_udp, mg_socks_if_tcp_send, mg_socks_if_udp_send, mg_socks_if_recved, mg_socks_if_create_conn, mg_socks_if_destroy_conn, mg_socks_if_sock_set, mg_socks_if_get_conn_addr, }; struct mg_iface *mg_socks_mk_iface(struct mg_mgr *mgr, const char *proxy_addr) { struct mg_iface *iface = mg_if_create_iface(&mg_socks_iface_vtable, mgr); iface->data = MG_CALLOC(1, sizeof(struct socksdata)); ((struct socksdata *) iface->data)->proxy_addr = strdup(proxy_addr); return iface; } #endif #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_ssl_if_openssl.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #if MG_ENABLE_SSL && MG_SSL_IF == MG_SSL_IF_OPENSSL #ifdef __APPLE__ #pragma GCC diagnostic ignored "-Wdeprecated-declarations" #endif #include struct mg_ssl_if_ctx { SSL *ssl; SSL_CTX *ssl_ctx; struct mbuf psk; size_t identity_len; }; void mg_ssl_if_init() { SSL_library_init(); } enum mg_ssl_if_result mg_ssl_if_conn_accept(struct mg_connection *nc, struct mg_connection *lc) { struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) MG_CALLOC(1, sizeof(*ctx)); struct mg_ssl_if_ctx *lc_ctx = (struct mg_ssl_if_ctx *) lc->ssl_if_data; nc->ssl_if_data = ctx; if (ctx == NULL || lc_ctx == NULL) return MG_SSL_ERROR; ctx->ssl_ctx = lc_ctx->ssl_ctx; if ((ctx->ssl = SSL_new(ctx->ssl_ctx)) == NULL) { return MG_SSL_ERROR; } return MG_SSL_OK; } static enum mg_ssl_if_result mg_use_cert(SSL_CTX *ctx, const char *cert, const char *key, const char **err_msg); static enum mg_ssl_if_result mg_use_ca_cert(SSL_CTX *ctx, const char *cert); static enum mg_ssl_if_result mg_set_cipher_list(SSL_CTX *ctx, const char *cl); static enum mg_ssl_if_result mg_ssl_if_ossl_set_psk(struct mg_ssl_if_ctx *ctx, const char *identity, const char *key_str); enum mg_ssl_if_result mg_ssl_if_conn_init( struct mg_connection *nc, const struct mg_ssl_if_conn_params *params, const char **err_msg) { struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) MG_CALLOC(1, sizeof(*ctx)); DBG(("%p %s,%s,%s", nc, (params->cert ? params->cert : ""), (params->key ? params->key : ""), (params->ca_cert ? params->ca_cert : ""))); if (ctx == NULL) { MG_SET_PTRPTR(err_msg, "Out of memory"); return MG_SSL_ERROR; } nc->ssl_if_data = ctx; if (nc->flags & MG_F_LISTENING) { ctx->ssl_ctx = SSL_CTX_new(SSLv23_server_method()); } else { ctx->ssl_ctx = SSL_CTX_new(SSLv23_client_method()); } if (ctx->ssl_ctx == NULL) { MG_SET_PTRPTR(err_msg, "Failed to create SSL context"); return MG_SSL_ERROR; } #ifndef KR_VERSION /* Disable deprecated protocols. */ SSL_CTX_set_options(ctx->ssl_ctx, SSL_OP_NO_SSLv2); SSL_CTX_set_options(ctx->ssl_ctx, SSL_OP_NO_SSLv3); SSL_CTX_set_options(ctx->ssl_ctx, SSL_OP_NO_TLSv1); #ifdef MG_SSL_OPENSSL_NO_COMPRESSION SSL_CTX_set_options(ctx->ssl_ctx, SSL_OP_NO_COMPRESSION); #endif #ifdef MG_SSL_OPENSSL_CIPHER_SERVER_PREFERENCE SSL_CTX_set_options(ctx->ssl_ctx, SSL_OP_CIPHER_SERVER_PREFERENCE); #endif #else /* Krypton only supports TLSv1.2 anyway. */ #endif if (params->cert != NULL && mg_use_cert(ctx->ssl_ctx, params->cert, params->key, err_msg) != MG_SSL_OK) { return MG_SSL_ERROR; } if (params->ca_cert != NULL && mg_use_ca_cert(ctx->ssl_ctx, params->ca_cert) != MG_SSL_OK) { MG_SET_PTRPTR(err_msg, "Invalid SSL CA cert"); return MG_SSL_ERROR; } if (params->server_name != NULL) { #ifdef KR_VERSION SSL_CTX_kr_set_verify_name(ctx->ssl_ctx, params->server_name); #else /* TODO(rojer): Implement server name verification on OpenSSL. */ #endif } if (mg_set_cipher_list(ctx->ssl_ctx, params->cipher_suites) != MG_SSL_OK) { MG_SET_PTRPTR(err_msg, "Invalid cipher suite list"); return MG_SSL_ERROR; } mbuf_init(&ctx->psk, 0); if (mg_ssl_if_ossl_set_psk(ctx, params->psk_identity, params->psk_key) != MG_SSL_OK) { MG_SET_PTRPTR(err_msg, "Invalid PSK settings"); return MG_SSL_ERROR; } if (!(nc->flags & MG_F_LISTENING) && (ctx->ssl = SSL_new(ctx->ssl_ctx)) == NULL) { MG_SET_PTRPTR(err_msg, "Failed to create SSL session"); return MG_SSL_ERROR; } nc->flags |= MG_F_SSL; return MG_SSL_OK; } static enum mg_ssl_if_result mg_ssl_if_ssl_err(struct mg_connection *nc, int res) { struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) nc->ssl_if_data; int err = SSL_get_error(ctx->ssl, res); if (err == SSL_ERROR_WANT_READ) return MG_SSL_WANT_READ; if (err == SSL_ERROR_WANT_WRITE) return MG_SSL_WANT_WRITE; DBG(("%p %p SSL error: %d %d", nc, ctx->ssl_ctx, res, err)); nc->err = err; return MG_SSL_ERROR; } enum mg_ssl_if_result mg_ssl_if_handshake(struct mg_connection *nc) { struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) nc->ssl_if_data; int server_side = (nc->listener != NULL); int res; /* If descriptor is not yet set, do it now. */ if (SSL_get_fd(ctx->ssl) < 0) { if (SSL_set_fd(ctx->ssl, nc->sock) != 1) return MG_SSL_ERROR; } res = server_side ? SSL_accept(ctx->ssl) : SSL_connect(ctx->ssl); if (res != 1) return mg_ssl_if_ssl_err(nc, res); return MG_SSL_OK; } int mg_ssl_if_read(struct mg_connection *nc, void *buf, size_t buf_size) { struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) nc->ssl_if_data; int n = SSL_read(ctx->ssl, buf, buf_size); DBG(("%p %d -> %d", nc, (int) buf_size, n)); if (n < 0) return mg_ssl_if_ssl_err(nc, n); if (n == 0) nc->flags |= MG_F_CLOSE_IMMEDIATELY; return n; } int mg_ssl_if_write(struct mg_connection *nc, const void *data, size_t len) { struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) nc->ssl_if_data; int n = SSL_write(ctx->ssl, data, len); DBG(("%p %d -> %d", nc, (int) len, n)); if (n <= 0) return mg_ssl_if_ssl_err(nc, n); return n; } void mg_ssl_if_conn_close_notify(struct mg_connection *nc) { struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) nc->ssl_if_data; if (ctx == NULL) return; SSL_shutdown(ctx->ssl); } void mg_ssl_if_conn_free(struct mg_connection *nc) { struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) nc->ssl_if_data; if (ctx == NULL) return; nc->ssl_if_data = NULL; if (ctx->ssl != NULL) SSL_free(ctx->ssl); if (ctx->ssl_ctx != NULL && nc->listener == NULL) SSL_CTX_free(ctx->ssl_ctx); mbuf_free(&ctx->psk); memset(ctx, 0, sizeof(*ctx)); MG_FREE(ctx); } /* * Cipher suite options used for TLS negotiation. * https://wiki.mozilla.org/Security/Server_Side_TLS#Recommended_configurations */ static const char mg_s_cipher_list[] = #if defined(MG_SSL_CRYPTO_MODERN) "ECDHE-ECDSA-AES256-GCM-SHA384:ECDHE-RSA-AES256-GCM-SHA384:" "ECDHE-ECDSA-AES128-GCM-SHA256:ECDHE-RSA-AES128-GCM-SHA256:" "DHE-RSA-AES128-GCM-SHA256:DHE-DSS-AES128-GCM-SHA256:kEDH+AESGCM:" "ECDHE-RSA-AES128-SHA256:ECDHE-ECDSA-AES128-SHA256:ECDHE-RSA-AES128-SHA:" "ECDHE-ECDSA-AES128-SHA:ECDHE-RSA-AES256-SHA384:ECDHE-ECDSA-AES256-SHA384:" "ECDHE-RSA-AES256-SHA:ECDHE-ECDSA-AES256-SHA:DHE-RSA-AES128-SHA256:" "DHE-RSA-AES128-SHA:DHE-DSS-AES128-SHA256:DHE-RSA-AES256-SHA256:" "DHE-DSS-AES256-SHA:DHE-RSA-AES256-SHA:" "!aNULL:!eNULL:!EXPORT:!DES:!RC4:!3DES:!MD5:!PSK" #elif defined(MG_SSL_CRYPTO_OLD) "ECDHE-RSA-AES128-GCM-SHA256:ECDHE-ECDSA-AES128-GCM-SHA256:" "ECDHE-RSA-AES256-GCM-SHA384:ECDHE-ECDSA-AES256-GCM-SHA384:" "DHE-RSA-AES128-GCM-SHA256:DHE-DSS-AES128-GCM-SHA256:kEDH+AESGCM:" "ECDHE-RSA-AES128-SHA256:ECDHE-ECDSA-AES128-SHA256:ECDHE-RSA-AES128-SHA:" "ECDHE-ECDSA-AES128-SHA:ECDHE-RSA-AES256-SHA384:ECDHE-ECDSA-AES256-SHA384:" "ECDHE-RSA-AES256-SHA:ECDHE-ECDSA-AES256-SHA:DHE-RSA-AES128-SHA256:" "DHE-RSA-AES128-SHA:DHE-DSS-AES128-SHA256:DHE-RSA-AES256-SHA256:" "DHE-DSS-AES256-SHA:DHE-RSA-AES256-SHA:ECDHE-RSA-DES-CBC3-SHA:" "ECDHE-ECDSA-DES-CBC3-SHA:AES128-GCM-SHA256:AES256-GCM-SHA384:" "AES128-SHA256:AES256-SHA256:AES128-SHA:AES256-SHA:AES:DES-CBC3-SHA:" "HIGH:!aNULL:!eNULL:!EXPORT:!DES:!RC4:!MD5:!PSK:!aECDH:" "!EDH-DSS-DES-CBC3-SHA:!EDH-RSA-DES-CBC3-SHA:!KRB5-DES-CBC3-SHA" #else /* Default - intermediate. */ "ECDHE-RSA-AES128-GCM-SHA256:ECDHE-ECDSA-AES128-GCM-SHA256:" "ECDHE-RSA-AES256-GCM-SHA384:ECDHE-ECDSA-AES256-GCM-SHA384:" "DHE-RSA-AES128-GCM-SHA256:DHE-DSS-AES128-GCM-SHA256:kEDH+AESGCM:" "ECDHE-RSA-AES128-SHA256:ECDHE-ECDSA-AES128-SHA256:ECDHE-RSA-AES128-SHA:" "ECDHE-ECDSA-AES128-SHA:ECDHE-RSA-AES256-SHA384:ECDHE-ECDSA-AES256-SHA384:" "ECDHE-RSA-AES256-SHA:ECDHE-ECDSA-AES256-SHA:DHE-RSA-AES128-SHA256:" "DHE-RSA-AES128-SHA:DHE-DSS-AES128-SHA256:DHE-RSA-AES256-SHA256:" "DHE-DSS-AES256-SHA:DHE-RSA-AES256-SHA:AES128-GCM-SHA256:AES256-GCM-SHA384:" "AES128-SHA256:AES256-SHA256:AES128-SHA:AES256-SHA:AES:CAMELLIA:" "DES-CBC3-SHA:!aNULL:!eNULL:!EXPORT:!DES:!RC4:!MD5:!PSK:!aECDH:" "!EDH-DSS-DES-CBC3-SHA:!EDH-RSA-DES-CBC3-SHA:!KRB5-DES-CBC3-SHA" #endif ; /* * Default DH params for PFS cipher negotiation. This is a 2048-bit group. * Will be used if none are provided by the user in the certificate file. */ #if !MG_DISABLE_PFS && !defined(KR_VERSION) static const char mg_s_default_dh_params[] = "\ -----BEGIN DH PARAMETERS-----\n\ MIIBCAKCAQEAlvbgD/qh9znWIlGFcV0zdltD7rq8FeShIqIhkQ0C7hYFThrBvF2E\n\ Z9bmgaP+sfQwGpVlv9mtaWjvERbu6mEG7JTkgmVUJrUt/wiRzwTaCXBqZkdUO8Tq\n\ +E6VOEQAilstG90ikN1Tfo+K6+X68XkRUIlgawBTKuvKVwBhuvlqTGerOtnXWnrt\n\ ym//hd3cd5PBYGBix0i7oR4xdghvfR2WLVu0LgdThTBb6XP7gLd19cQ1JuBtAajZ\n\ wMuPn7qlUkEFDIkAZy59/Hue/H2Q2vU/JsvVhHWCQBL4F1ofEAt50il6ZxR1QfFK\n\ 9VGKDC4oOgm9DlxwwBoC2FjqmvQlqVV3kwIBAg==\n\ -----END DH PARAMETERS-----\n"; #endif static enum mg_ssl_if_result mg_use_ca_cert(SSL_CTX *ctx, const char *cert) { if (cert == NULL || strcmp(cert, "*") == 0) { return MG_SSL_OK; } SSL_CTX_set_verify(ctx, SSL_VERIFY_PEER | SSL_VERIFY_FAIL_IF_NO_PEER_CERT, 0); return SSL_CTX_load_verify_locations(ctx, cert, NULL) == 1 ? MG_SSL_OK : MG_SSL_ERROR; } static enum mg_ssl_if_result mg_use_cert(SSL_CTX *ctx, const char *cert, const char *key, const char **err_msg) { if (key == NULL) key = cert; if (cert == NULL || cert[0] == '\0' || key == NULL || key[0] == '\0') { return MG_SSL_OK; } else if (SSL_CTX_use_certificate_file(ctx, cert, 1) == 0) { MG_SET_PTRPTR(err_msg, "Invalid SSL cert"); return MG_SSL_ERROR; } else if (SSL_CTX_use_PrivateKey_file(ctx, key, 1) == 0) { MG_SET_PTRPTR(err_msg, "Invalid SSL key"); return MG_SSL_ERROR; } else if (SSL_CTX_use_certificate_chain_file(ctx, cert) == 0) { MG_SET_PTRPTR(err_msg, "Invalid CA bundle"); return MG_SSL_ERROR; } else { SSL_CTX_set_mode(ctx, SSL_MODE_ACCEPT_MOVING_WRITE_BUFFER); #if !MG_DISABLE_PFS && !defined(KR_VERSION) BIO *bio = NULL; DH *dh = NULL; /* Try to read DH parameters from the cert/key file. */ bio = BIO_new_file(cert, "r"); if (bio != NULL) { dh = PEM_read_bio_DHparams(bio, NULL, NULL, NULL); BIO_free(bio); } /* * If there are no DH params in the file, fall back to hard-coded ones. * Not ideal, but better than nothing. */ if (dh == NULL) { bio = BIO_new_mem_buf((void *) mg_s_default_dh_params, -1); dh = PEM_read_bio_DHparams(bio, NULL, NULL, NULL); BIO_free(bio); } if (dh != NULL) { SSL_CTX_set_tmp_dh(ctx, dh); SSL_CTX_set_options(ctx, SSL_OP_SINGLE_DH_USE); DH_free(dh); } #if OPENSSL_VERSION_NUMBER > 0x10002000L SSL_CTX_set_ecdh_auto(ctx, 1); #endif #endif } return MG_SSL_OK; } static enum mg_ssl_if_result mg_set_cipher_list(SSL_CTX *ctx, const char *cl) { return (SSL_CTX_set_cipher_list(ctx, cl ? cl : mg_s_cipher_list) == 1 ? MG_SSL_OK : MG_SSL_ERROR); } #ifndef KR_VERSION static unsigned int mg_ssl_if_ossl_psk_cb(SSL *ssl, const char *hint, char *identity, unsigned int max_identity_len, unsigned char *psk, unsigned int max_psk_len) { struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) SSL_CTX_get_app_data(SSL_get_SSL_CTX(ssl)); size_t key_len = ctx->psk.len - ctx->identity_len - 1; DBG(("hint: '%s'", (hint ? hint : ""))); if (ctx->identity_len + 1 > max_identity_len) { DBG(("identity too long")); return 0; } if (key_len > max_psk_len) { DBG(("key too long")); return 0; } memcpy(identity, ctx->psk.buf, ctx->identity_len + 1); memcpy(psk, ctx->psk.buf + ctx->identity_len + 1, key_len); (void) ssl; return key_len; } static enum mg_ssl_if_result mg_ssl_if_ossl_set_psk(struct mg_ssl_if_ctx *ctx, const char *identity, const char *key_str) { unsigned char key[32]; size_t key_len; size_t i = 0; if (identity == NULL && key_str == NULL) return MG_SSL_OK; if (identity == NULL || key_str == NULL) return MG_SSL_ERROR; key_len = strlen(key_str); if (key_len != 32 && key_len != 64) return MG_SSL_ERROR; memset(key, 0, sizeof(key)); key_len = 0; for (i = 0; key_str[i] != '\0'; i++) { unsigned char c; char hc = tolower((int) key_str[i]); if (hc >= '0' && hc <= '9') { c = hc - '0'; } else if (hc >= 'a' && hc <= 'f') { c = hc - 'a' + 0xa; } else { return MG_SSL_ERROR; } key_len = i / 2; key[key_len] <<= 4; key[key_len] |= c; } key_len++; DBG(("identity = '%s', key = (%u)", identity, (unsigned int) key_len)); ctx->identity_len = strlen(identity); mbuf_append(&ctx->psk, identity, ctx->identity_len + 1); mbuf_append(&ctx->psk, key, key_len); SSL_CTX_set_psk_client_callback(ctx->ssl_ctx, mg_ssl_if_ossl_psk_cb); SSL_CTX_set_app_data(ctx->ssl_ctx, ctx); return MG_SSL_OK; } #else static enum mg_ssl_if_result mg_ssl_if_ossl_set_psk(struct mg_ssl_if_ctx *ctx, const char *identity, const char *key_str) { (void) ctx; (void) identity; (void) key_str; /* Krypton does not support PSK. */ return MG_SSL_ERROR; } #endif /* defined(KR_VERSION) */ const char *mg_set_ssl(struct mg_connection *nc, const char *cert, const char *ca_cert) { const char *err_msg = NULL; struct mg_ssl_if_conn_params params; memset(¶ms, 0, sizeof(params)); params.cert = cert; params.ca_cert = ca_cert; if (mg_ssl_if_conn_init(nc, ¶ms, &err_msg) != MG_SSL_OK) { return err_msg; } return NULL; } #endif /* MG_ENABLE_SSL && MG_SSL_IF == MG_SSL_IF_OPENSSL */ #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_ssl_if_mbedtls.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #if MG_ENABLE_SSL && MG_SSL_IF == MG_SSL_IF_MBEDTLS #include #include #include #include #include static void mg_ssl_mbed_log(void *ctx, int level, const char *file, int line, const char *str) { enum cs_log_level cs_level; switch (level) { case 1: cs_level = LL_ERROR; break; case 2: case 3: cs_level = LL_DEBUG; break; default: cs_level = LL_VERBOSE_DEBUG; } /* mbedTLS passes strings with \n at the end, strip it. */ LOG(cs_level, ("%p %.*s", ctx, (int) (strlen(str) - 1), str)); (void) file; (void) line; } struct mg_ssl_if_ctx { mbedtls_ssl_config *conf; mbedtls_ssl_context *ssl; mbedtls_x509_crt *cert; mbedtls_pk_context *key; mbedtls_x509_crt *ca_cert; struct mbuf cipher_suites; }; /* Must be provided by the platform. ctx is struct mg_connection. */ extern int mg_ssl_if_mbed_random(void *ctx, unsigned char *buf, size_t len); void mg_ssl_if_init() { } enum mg_ssl_if_result mg_ssl_if_conn_accept(struct mg_connection *nc, struct mg_connection *lc) { struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) MG_CALLOC(1, sizeof(*ctx)); struct mg_ssl_if_ctx *lc_ctx = (struct mg_ssl_if_ctx *) lc->ssl_if_data; nc->ssl_if_data = ctx; if (ctx == NULL || lc_ctx == NULL) return MG_SSL_ERROR; ctx->ssl = (mbedtls_ssl_context *) MG_CALLOC(1, sizeof(*ctx->ssl)); if (mbedtls_ssl_setup(ctx->ssl, lc_ctx->conf) != 0) { return MG_SSL_ERROR; } return MG_SSL_OK; } static enum mg_ssl_if_result mg_use_cert(struct mg_ssl_if_ctx *ctx, const char *cert, const char *key, const char **err_msg); static enum mg_ssl_if_result mg_use_ca_cert(struct mg_ssl_if_ctx *ctx, const char *cert); static enum mg_ssl_if_result mg_set_cipher_list(struct mg_ssl_if_ctx *ctx, const char *ciphers); static enum mg_ssl_if_result mg_ssl_if_mbed_set_psk(struct mg_ssl_if_ctx *ctx, const char *identity, const char *key); enum mg_ssl_if_result mg_ssl_if_conn_init( struct mg_connection *nc, const struct mg_ssl_if_conn_params *params, const char **err_msg) { struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) MG_CALLOC(1, sizeof(*ctx)); DBG(("%p %s,%s,%s", nc, (params->cert ? params->cert : ""), (params->key ? params->key : ""), (params->ca_cert ? params->ca_cert : ""))); if (ctx == NULL) { MG_SET_PTRPTR(err_msg, "Out of memory"); return MG_SSL_ERROR; } nc->ssl_if_data = ctx; ctx->conf = (mbedtls_ssl_config *) MG_CALLOC(1, sizeof(*ctx->conf)); mbuf_init(&ctx->cipher_suites, 0); mbedtls_ssl_config_init(ctx->conf); mbedtls_ssl_conf_dbg(ctx->conf, mg_ssl_mbed_log, nc); if (mbedtls_ssl_config_defaults( ctx->conf, (nc->flags & MG_F_LISTENING ? MBEDTLS_SSL_IS_SERVER : MBEDTLS_SSL_IS_CLIENT), MBEDTLS_SSL_TRANSPORT_STREAM, MBEDTLS_SSL_PRESET_DEFAULT) != 0) { MG_SET_PTRPTR(err_msg, "Failed to init SSL config"); return MG_SSL_ERROR; } /* TLS 1.2 and up */ mbedtls_ssl_conf_min_version(ctx->conf, MBEDTLS_SSL_MAJOR_VERSION_3, MBEDTLS_SSL_MINOR_VERSION_3); mbedtls_ssl_conf_rng(ctx->conf, mg_ssl_if_mbed_random, nc); if (params->cert != NULL && mg_use_cert(ctx, params->cert, params->key, err_msg) != MG_SSL_OK) { return MG_SSL_ERROR; } if (params->ca_cert != NULL && mg_use_ca_cert(ctx, params->ca_cert) != MG_SSL_OK) { MG_SET_PTRPTR(err_msg, "Invalid SSL CA cert"); return MG_SSL_ERROR; } if (mg_set_cipher_list(ctx, params->cipher_suites) != MG_SSL_OK) { MG_SET_PTRPTR(err_msg, "Invalid cipher suite list"); return MG_SSL_ERROR; } if (mg_ssl_if_mbed_set_psk(ctx, params->psk_identity, params->psk_key) != MG_SSL_OK) { MG_SET_PTRPTR(err_msg, "Invalid PSK settings"); return MG_SSL_ERROR; } if (!(nc->flags & MG_F_LISTENING)) { ctx->ssl = (mbedtls_ssl_context *) MG_CALLOC(1, sizeof(*ctx->ssl)); mbedtls_ssl_init(ctx->ssl); if (mbedtls_ssl_setup(ctx->ssl, ctx->conf) != 0) { MG_SET_PTRPTR(err_msg, "Failed to create SSL session"); return MG_SSL_ERROR; } if (params->server_name != NULL && mbedtls_ssl_set_hostname(ctx->ssl, params->server_name) != 0) { return MG_SSL_ERROR; } } #ifdef MG_SSL_IF_MBEDTLS_MAX_FRAG_LEN if (mbedtls_ssl_conf_max_frag_len(ctx->conf, #if MG_SSL_IF_MBEDTLS_MAX_FRAG_LEN == 512 MBEDTLS_SSL_MAX_FRAG_LEN_512 #elif MG_SSL_IF_MBEDTLS_MAX_FRAG_LEN == 1024 MBEDTLS_SSL_MAX_FRAG_LEN_1024 #elif MG_SSL_IF_MBEDTLS_MAX_FRAG_LEN == 2048 MBEDTLS_SSL_MAX_FRAG_LEN_2048 #elif MG_SSL_IF_MBEDTLS_MAX_FRAG_LEN == 4096 MBEDTLS_SSL_MAX_FRAG_LEN_4096 #else #error Invalid MG_SSL_IF_MBEDTLS_MAX_FRAG_LEN #endif ) != 0) { return MG_SSL_ERROR; } #endif nc->flags |= MG_F_SSL; return MG_SSL_OK; } #if MG_NET_IF == MG_NET_IF_LWIP_LOW_LEVEL int ssl_socket_send(void *ctx, const unsigned char *buf, size_t len); int ssl_socket_recv(void *ctx, unsigned char *buf, size_t len); #else static int ssl_socket_send(void *ctx, const unsigned char *buf, size_t len) { struct mg_connection *nc = (struct mg_connection *) ctx; int n = (int) MG_SEND_FUNC(nc->sock, buf, len, 0); LOG(LL_DEBUG, ("%p %d -> %d", nc, (int) len, n)); if (n >= 0) return n; n = mg_get_errno(); return ((n == EAGAIN || n == EINPROGRESS) ? MBEDTLS_ERR_SSL_WANT_WRITE : -1); } static int ssl_socket_recv(void *ctx, unsigned char *buf, size_t len) { struct mg_connection *nc = (struct mg_connection *) ctx; int n = (int) MG_RECV_FUNC(nc->sock, buf, len, 0); LOG(LL_DEBUG, ("%p %d <- %d", nc, (int) len, n)); if (n >= 0) return n; n = mg_get_errno(); return ((n == EAGAIN || n == EINPROGRESS) ? MBEDTLS_ERR_SSL_WANT_READ : -1); } #endif static enum mg_ssl_if_result mg_ssl_if_mbed_err(struct mg_connection *nc, int ret) { if (ret == MBEDTLS_ERR_SSL_WANT_READ) return MG_SSL_WANT_READ; if (ret == MBEDTLS_ERR_SSL_WANT_WRITE) return MG_SSL_WANT_WRITE; if (ret != MBEDTLS_ERR_SSL_PEER_CLOSE_NOTIFY) { /* CLOSE_NOTIFY = Normal shutdown */ LOG(LL_ERROR, ("%p SSL error: %d", nc, ret)); } nc->err = ret; nc->flags |= MG_F_CLOSE_IMMEDIATELY; return MG_SSL_ERROR; } static void mg_ssl_if_mbed_free_certs_and_keys(struct mg_ssl_if_ctx *ctx) { if (ctx->cert != NULL) { mbedtls_x509_crt_free(ctx->cert); MG_FREE(ctx->cert); ctx->cert = NULL; mbedtls_pk_free(ctx->key); MG_FREE(ctx->key); ctx->key = NULL; } if (ctx->ca_cert != NULL) { mbedtls_ssl_conf_ca_chain(ctx->conf, NULL, NULL); #ifdef MBEDTLS_X509_CA_CHAIN_ON_DISK if (ctx->ca_cert->ca_chain_file != NULL) { MG_FREE((void *) ctx->ca_cert->ca_chain_file); ctx->ca_cert->ca_chain_file = NULL; } #endif mbedtls_x509_crt_free(ctx->ca_cert); MG_FREE(ctx->ca_cert); ctx->ca_cert = NULL; } } enum mg_ssl_if_result mg_ssl_if_handshake(struct mg_connection *nc) { struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) nc->ssl_if_data; int err; /* If bio is not yet set, do it now. */ if (ctx->ssl->p_bio == NULL) { mbedtls_ssl_set_bio(ctx->ssl, nc, ssl_socket_send, ssl_socket_recv, NULL); } err = mbedtls_ssl_handshake(ctx->ssl); if (err != 0) return mg_ssl_if_mbed_err(nc, err); #ifdef MG_SSL_IF_MBEDTLS_FREE_CERTS /* * Free the peer certificate, we don't need it after handshake. * Note that this effectively disables renegotiation. */ mbedtls_x509_crt_free(ctx->ssl->session->peer_cert); mbedtls_free(ctx->ssl->session->peer_cert); ctx->ssl->session->peer_cert = NULL; /* On a client connection we can also free our own and CA certs. */ if (nc->listener == NULL) { if (ctx->conf->key_cert != NULL) { /* Note that this assumes one key_cert entry, which matches our init. */ MG_FREE(ctx->conf->key_cert); ctx->conf->key_cert = NULL; } mbedtls_ssl_conf_ca_chain(ctx->conf, NULL, NULL); mg_ssl_if_mbed_free_certs_and_keys(ctx); } #endif return MG_SSL_OK; } int mg_ssl_if_read(struct mg_connection *nc, void *buf, size_t buf_size) { struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) nc->ssl_if_data; int n = mbedtls_ssl_read(ctx->ssl, (unsigned char *) buf, buf_size); DBG(("%p %d -> %d", nc, (int) buf_size, n)); if (n < 0) return mg_ssl_if_mbed_err(nc, n); if (n == 0) nc->flags |= MG_F_CLOSE_IMMEDIATELY; return n; } int mg_ssl_if_write(struct mg_connection *nc, const void *data, size_t len) { struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) nc->ssl_if_data; int n = mbedtls_ssl_write(ctx->ssl, (const unsigned char *) data, len); DBG(("%p %d -> %d", nc, (int) len, n)); if (n < 0) return mg_ssl_if_mbed_err(nc, n); return n; } void mg_ssl_if_conn_close_notify(struct mg_connection *nc) { struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) nc->ssl_if_data; if (ctx == NULL) return; mbedtls_ssl_close_notify(ctx->ssl); } void mg_ssl_if_conn_free(struct mg_connection *nc) { struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) nc->ssl_if_data; if (ctx == NULL) return; nc->ssl_if_data = NULL; if (ctx->ssl != NULL) { mbedtls_ssl_free(ctx->ssl); MG_FREE(ctx->ssl); } mg_ssl_if_mbed_free_certs_and_keys(ctx); if (ctx->conf != NULL) { mbedtls_ssl_config_free(ctx->conf); MG_FREE(ctx->conf); } mbuf_free(&ctx->cipher_suites); memset(ctx, 0, sizeof(*ctx)); MG_FREE(ctx); } static enum mg_ssl_if_result mg_use_ca_cert(struct mg_ssl_if_ctx *ctx, const char *ca_cert) { if (ca_cert == NULL || strcmp(ca_cert, "*") == 0) { mbedtls_ssl_conf_authmode(ctx->conf, MBEDTLS_SSL_VERIFY_NONE); return MG_SSL_OK; } ctx->ca_cert = (mbedtls_x509_crt *) MG_CALLOC(1, sizeof(*ctx->ca_cert)); mbedtls_x509_crt_init(ctx->ca_cert); #ifdef MBEDTLS_X509_CA_CHAIN_ON_DISK ca_cert = strdup(ca_cert); if (mbedtls_x509_crt_set_ca_chain_file(ctx->ca_cert, ca_cert) != 0) { return MG_SSL_ERROR; } #else if (mbedtls_x509_crt_parse_file(ctx->ca_cert, ca_cert) != 0) { return MG_SSL_ERROR; } #endif mbedtls_ssl_conf_ca_chain(ctx->conf, ctx->ca_cert, NULL); mbedtls_ssl_conf_authmode(ctx->conf, MBEDTLS_SSL_VERIFY_REQUIRED); return MG_SSL_OK; } static enum mg_ssl_if_result mg_use_cert(struct mg_ssl_if_ctx *ctx, const char *cert, const char *key, const char **err_msg) { if (key == NULL) key = cert; if (cert == NULL || cert[0] == '\0' || key == NULL || key[0] == '\0') { return MG_SSL_OK; } ctx->cert = (mbedtls_x509_crt *) MG_CALLOC(1, sizeof(*ctx->cert)); mbedtls_x509_crt_init(ctx->cert); ctx->key = (mbedtls_pk_context *) MG_CALLOC(1, sizeof(*ctx->key)); mbedtls_pk_init(ctx->key); if (mbedtls_x509_crt_parse_file(ctx->cert, cert) != 0) { MG_SET_PTRPTR(err_msg, "Invalid SSL cert"); return MG_SSL_ERROR; } if (mbedtls_pk_parse_keyfile(ctx->key, key, NULL) != 0) { MG_SET_PTRPTR(err_msg, "Invalid SSL key"); return MG_SSL_ERROR; } if (mbedtls_ssl_conf_own_cert(ctx->conf, ctx->cert, ctx->key) != 0) { MG_SET_PTRPTR(err_msg, "Invalid SSL key or cert"); return MG_SSL_ERROR; } return MG_SSL_OK; } static const int mg_s_cipher_list[] = { #if CS_PLATFORM != CS_P_ESP8266 MBEDTLS_TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256, MBEDTLS_TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA256, MBEDTLS_TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256, MBEDTLS_TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA256, MBEDTLS_TLS_DHE_RSA_WITH_AES_128_GCM_SHA256, MBEDTLS_TLS_DHE_RSA_WITH_AES_128_CBC_SHA256, MBEDTLS_TLS_ECDH_ECDSA_WITH_AES_128_GCM_SHA256, MBEDTLS_TLS_ECDH_ECDSA_WITH_AES_128_CBC_SHA256, MBEDTLS_TLS_ECDH_ECDSA_WITH_AES_128_CBC_SHA, MBEDTLS_TLS_ECDH_RSA_WITH_AES_128_GCM_SHA256, MBEDTLS_TLS_ECDH_RSA_WITH_AES_128_CBC_SHA256, MBEDTLS_TLS_ECDH_RSA_WITH_AES_128_CBC_SHA, MBEDTLS_TLS_RSA_WITH_AES_128_GCM_SHA256, MBEDTLS_TLS_RSA_WITH_AES_128_CBC_SHA256, MBEDTLS_TLS_RSA_WITH_AES_128_CBC_SHA, #else /* * ECDHE is way too slow on ESP8266 w/o cryptochip, this sometimes results * in WiFi STA deauths. Use weaker but faster cipher suites. Sad but true. * Disable DHE completely because it's just hopelessly slow. */ MBEDTLS_TLS_RSA_WITH_AES_128_GCM_SHA256, MBEDTLS_TLS_RSA_WITH_AES_128_CBC_SHA256, MBEDTLS_TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256, MBEDTLS_TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA256, MBEDTLS_TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256, MBEDTLS_TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA256, MBEDTLS_TLS_ECDH_ECDSA_WITH_AES_128_GCM_SHA256, MBEDTLS_TLS_ECDH_ECDSA_WITH_AES_128_CBC_SHA256, MBEDTLS_TLS_ECDH_ECDSA_WITH_AES_128_CBC_SHA, MBEDTLS_TLS_ECDH_RSA_WITH_AES_128_GCM_SHA256, MBEDTLS_TLS_ECDH_RSA_WITH_AES_128_CBC_SHA256, MBEDTLS_TLS_ECDH_RSA_WITH_AES_128_CBC_SHA, MBEDTLS_TLS_RSA_WITH_AES_128_CBC_SHA, #endif /* CS_PLATFORM != CS_P_ESP8266 */ 0, }; /* * Ciphers can be specified as a colon-separated list of cipher suite names. * These can be found in * https://github.com/ARMmbed/mbedtls/blob/development/library/ssl_ciphersuites.c#L267 * E.g.: TLS-ECDHE-ECDSA-WITH-AES-128-GCM-SHA256:TLS-DHE-RSA-WITH-AES-256-CCM */ static enum mg_ssl_if_result mg_set_cipher_list(struct mg_ssl_if_ctx *ctx, const char *ciphers) { if (ciphers != NULL) { int l, id; const char *s = ciphers, *e; char tmp[50]; while (s != NULL) { e = strchr(s, ':'); l = (e != NULL ? (e - s) : (int) strlen(s)); strncpy(tmp, s, l); tmp[l] = '\0'; id = mbedtls_ssl_get_ciphersuite_id(tmp); DBG(("%s -> %04x", tmp, id)); if (id != 0) { mbuf_append(&ctx->cipher_suites, &id, sizeof(id)); } s = (e != NULL ? e + 1 : NULL); } if (ctx->cipher_suites.len == 0) return MG_SSL_ERROR; id = 0; mbuf_append(&ctx->cipher_suites, &id, sizeof(id)); mbuf_trim(&ctx->cipher_suites); mbedtls_ssl_conf_ciphersuites(ctx->conf, (const int *) ctx->cipher_suites.buf); } else { mbedtls_ssl_conf_ciphersuites(ctx->conf, mg_s_cipher_list); } return MG_SSL_OK; } static enum mg_ssl_if_result mg_ssl_if_mbed_set_psk(struct mg_ssl_if_ctx *ctx, const char *identity, const char *key_str) { unsigned char key[32]; size_t key_len; if (identity == NULL && key_str == NULL) return MG_SSL_OK; if (identity == NULL || key_str == NULL) return MG_SSL_ERROR; key_len = strlen(key_str); if (key_len != 32 && key_len != 64) return MG_SSL_ERROR; size_t i = 0; memset(key, 0, sizeof(key)); key_len = 0; for (i = 0; key_str[i] != '\0'; i++) { unsigned char c; char hc = tolower((int) key_str[i]); if (hc >= '0' && hc <= '9') { c = hc - '0'; } else if (hc >= 'a' && hc <= 'f') { c = hc - 'a' + 0xa; } else { return MG_SSL_ERROR; } key_len = i / 2; key[key_len] <<= 4; key[key_len] |= c; } key_len++; DBG(("identity = '%s', key = (%u)", identity, (unsigned int) key_len)); /* mbedTLS makes copies of psk and identity. */ if (mbedtls_ssl_conf_psk(ctx->conf, (const unsigned char *) key, key_len, (const unsigned char *) identity, strlen(identity)) != 0) { return MG_SSL_ERROR; } return MG_SSL_OK; } const char *mg_set_ssl(struct mg_connection *nc, const char *cert, const char *ca_cert) { const char *err_msg = NULL; struct mg_ssl_if_conn_params params; memset(¶ms, 0, sizeof(params)); params.cert = cert; params.ca_cert = ca_cert; if (mg_ssl_if_conn_init(nc, ¶ms, &err_msg) != MG_SSL_OK) { return err_msg; } return NULL; } /* Lazy RNG. Warning: it would be a bad idea to do this in production! */ #ifdef MG_SSL_MBED_DUMMY_RANDOM int mg_ssl_if_mbed_random(void *ctx, unsigned char *buf, size_t len) { (void) ctx; while (len--) *buf++ = rand(); return 0; } #endif #endif /* MG_ENABLE_SSL && MG_SSL_IF == MG_SSL_IF_MBEDTLS */ #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_uri.c" #endif /* * Copyright (c) 2014 Cesanta Software Limited * All rights reserved */ /* Amalgamated: #include "mg_internal.h" */ /* Amalgamated: #include "mg_uri.h" */ /* * scan string until encountering one of `seps`, keeping track of component * boundaries in `res`. * * `p` will point to the char after the separator or it will be `end`. */ static void parse_uri_component(const char **p, const char *end, const char *seps, struct mg_str *res) { const char *q; res->p = *p; for (; *p < end; (*p)++) { for (q = seps; *q != '\0'; q++) { if (**p == *q) break; } if (*q != '\0') break; } res->len = (*p) - res->p; if (*p < end) (*p)++; } int mg_parse_uri(const struct mg_str uri, struct mg_str *scheme, struct mg_str *user_info, struct mg_str *host, unsigned int *port, struct mg_str *path, struct mg_str *query, struct mg_str *fragment) { struct mg_str rscheme = {0, 0}, ruser_info = {0, 0}, rhost = {0, 0}, rpath = {0, 0}, rquery = {0, 0}, rfragment = {0, 0}; unsigned int rport = 0; enum { P_START, P_SCHEME_OR_PORT, P_USER_INFO, P_HOST, P_PORT, P_REST } state = P_START; const char *p = uri.p, *end = p + uri.len; while (p < end) { switch (state) { case P_START: /* * expecting on of: * - `scheme://xxxx` * - `xxxx:port` * - `[a:b:c]:port` * - `xxxx/path` */ if (*p == '[') { state = P_HOST; break; } for (; p < end; p++) { if (*p == ':') { state = P_SCHEME_OR_PORT; break; } else if (*p == '/') { state = P_REST; break; } } if (state == P_START || state == P_REST) { rhost.p = uri.p; rhost.len = p - uri.p; } break; case P_SCHEME_OR_PORT: if (end - p >= 3 && strncmp(p, "://", 3) == 0) { rscheme.p = uri.p; rscheme.len = p - uri.p; state = P_USER_INFO; p += 3; } else { rhost.p = uri.p; rhost.len = p - uri.p; state = P_PORT; } break; case P_USER_INFO: ruser_info.p = p; for (; p < end; p++) { if (*p == '@' || *p == '[' || *p == '/') { break; } } if (p == end || *p == '/' || *p == '[') { /* backtrack and parse as host */ p = ruser_info.p; } ruser_info.len = p - ruser_info.p; state = P_HOST; break; case P_HOST: if (*p == '@') p++; rhost.p = p; if (*p == '[') { int found = 0; for (; !found && p < end; p++) { found = (*p == ']'); } if (!found) return -1; } else { for (; p < end; p++) { if (*p == ':' || *p == '/') break; } } rhost.len = p - rhost.p; if (p < end) { if (*p == ':') { state = P_PORT; break; } else if (*p == '/') { state = P_REST; break; } } break; case P_PORT: p++; for (; p < end; p++) { if (*p == '/') { state = P_REST; break; } rport *= 10; rport += *p - '0'; } break; case P_REST: /* `p` points to separator. `path` includes the separator */ parse_uri_component(&p, end, "?#", &rpath); if (p < end && *(p - 1) == '?') { parse_uri_component(&p, end, "#", &rquery); } parse_uri_component(&p, end, "", &rfragment); break; } } if (scheme != 0) *scheme = rscheme; if (user_info != 0) *user_info = ruser_info; if (host != 0) *host = rhost; if (port != 0) *port = rport; if (path != 0) *path = rpath; if (query != 0) *query = rquery; if (fragment != 0) *fragment = rfragment; return 0; } /* Normalize the URI path. Remove/resolve "." and "..". */ int mg_normalize_uri_path(const struct mg_str *in, struct mg_str *out) { const char *s = in->p, *se = s + in->len; char *cp = (char *) out->p, *d; if (in->len == 0 || *s != '/') { out->len = 0; return 0; } d = cp; while (s < se) { const char *next = s; struct mg_str component; parse_uri_component(&next, se, "/", &component); if (mg_vcmp(&component, ".") == 0) { /* Yum. */ } else if (mg_vcmp(&component, "..") == 0) { /* Backtrack to previous slash. */ if (d > cp + 1 && *(d - 1) == '/') d--; while (d > cp && *(d - 1) != '/') d--; } else { memmove(d, s, next - s); d += next - s; } s = next; } if (d == cp) *d++ = '/'; out->p = cp; out->len = d - cp; return 1; } int mg_assemble_uri(const struct mg_str *scheme, const struct mg_str *user_info, const struct mg_str *host, unsigned int port, const struct mg_str *path, const struct mg_str *query, const struct mg_str *fragment, int normalize_path, struct mg_str *uri) { int result = -1; struct mbuf out; mbuf_init(&out, 0); if (scheme != NULL && scheme->len > 0) { mbuf_append(&out, scheme->p, scheme->len); mbuf_append(&out, "://", 3); } if (user_info != NULL && user_info->len > 0) { mbuf_append(&out, user_info->p, user_info->len); mbuf_append(&out, "@", 1); } if (host != NULL && host->len > 0) { mbuf_append(&out, host->p, host->len); } if (port != 0) { char port_str[20]; int port_str_len = sprintf(port_str, ":%u", port); mbuf_append(&out, port_str, port_str_len); } if (path != NULL && path->len > 0) { if (normalize_path) { struct mg_str npath = mg_strdup(*path); if (npath.len != path->len) goto out; if (!mg_normalize_uri_path(path, &npath)) { free((void *) npath.p); goto out; } mbuf_append(&out, npath.p, npath.len); free((void *) npath.p); } else { mbuf_append(&out, path->p, path->len); } } else if (normalize_path) { mbuf_append(&out, "/", 1); } if (query != NULL && query->len > 0) { mbuf_append(&out, "?", 1); mbuf_append(&out, query->p, query->len); } if (fragment != NULL && fragment->len > 0) { mbuf_append(&out, "#", 1); mbuf_append(&out, fragment->p, fragment->len); } result = 0; out: if (result == 0) { uri->p = out.buf; uri->len = out.len; } else { mbuf_free(&out); uri->p = NULL; uri->len = 0; } return result; } #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_http.c" #endif /* * Copyright (c) 2014 Cesanta Software Limited * All rights reserved */ #if MG_ENABLE_HTTP /* Amalgamated: #include "common/cs_md5.h" */ /* Amalgamated: #include "mg_internal.h" */ /* Amalgamated: #include "mg_util.h" */ /* altbuf {{{ */ /* * Alternate buffer: fills the client-provided buffer with data; and if it's * not large enough, allocates another buffer (via mbuf), similar to asprintf. */ struct altbuf { struct mbuf m; char *user_buf; size_t len; size_t user_buf_size; }; /* * Initializes altbuf; `buf`, `buf_size` is the client-provided buffer. */ MG_INTERNAL void altbuf_init(struct altbuf *ab, char *buf, size_t buf_size) { mbuf_init(&ab->m, 0); ab->user_buf = buf; ab->user_buf_size = buf_size; ab->len = 0; } /* * Appends a single char to the altbuf. */ MG_INTERNAL void altbuf_append(struct altbuf *ab, char c) { if (ab->len < ab->user_buf_size) { /* The data fits into the original buffer */ ab->user_buf[ab->len++] = c; } else { /* The data can't fit into the original buffer, so write it to mbuf. */ /* * First of all, see if that's the first byte which overflows the original * buffer: if so, copy the existing data from there to a newly allocated * mbuf. */ if (ab->len > 0 && ab->m.len == 0) { mbuf_append(&ab->m, ab->user_buf, ab->len); } mbuf_append(&ab->m, &c, 1); ab->len = ab->m.len; } } /* * Resets any data previously appended to altbuf. */ MG_INTERNAL void altbuf_reset(struct altbuf *ab) { mbuf_free(&ab->m); ab->len = 0; } /* * Returns whether the additional buffer was allocated (and thus the data * is in the mbuf, not the client-provided buffer) */ MG_INTERNAL int altbuf_reallocated(struct altbuf *ab) { return ab->len > ab->user_buf_size; } /* * Returns the actual buffer with data, either the client-provided or a newly * allocated one. If `trim` is non-zero, mbuf-backed buffer is trimmed first. */ MG_INTERNAL char *altbuf_get_buf(struct altbuf *ab, int trim) { if (altbuf_reallocated(ab)) { if (trim) { mbuf_trim(&ab->m); } return ab->m.buf; } else { return ab->user_buf; } } /* }}} */ static const char *mg_version_header = "Mongoose/" MG_VERSION; enum mg_http_proto_data_type { DATA_NONE, DATA_FILE, DATA_PUT }; struct mg_http_proto_data_file { FILE *fp; /* Opened file. */ int64_t cl; /* Content-Length. How many bytes to send. */ int64_t sent; /* How many bytes have been already sent. */ int keepalive; /* Keep connection open after sending. */ enum mg_http_proto_data_type type; }; #if MG_ENABLE_HTTP_CGI struct mg_http_proto_data_cgi { struct mg_connection *cgi_nc; }; #endif struct mg_http_proto_data_chuncked { int64_t body_len; /* How many bytes of chunked body was reassembled. */ }; struct mg_http_endpoint { struct mg_http_endpoint *next; struct mg_str uri_pattern; /* owned */ char *auth_domain; /* owned */ char *auth_file; /* owned */ mg_event_handler_t handler; #if MG_ENABLE_CALLBACK_USERDATA void *user_data; #endif }; enum mg_http_multipart_stream_state { MPS_BEGIN, MPS_WAITING_FOR_BOUNDARY, MPS_WAITING_FOR_CHUNK, MPS_GOT_CHUNK, MPS_GOT_BOUNDARY, MPS_FINALIZE, MPS_FINISHED }; struct mg_http_multipart_stream { const char *boundary; int boundary_len; const char *var_name; const char *file_name; void *user_data; int prev_io_len; enum mg_http_multipart_stream_state state; int processing_part; }; struct mg_reverse_proxy_data { struct mg_connection *linked_conn; }; struct mg_ws_proto_data { /* * Defragmented size of the frame so far. * * First byte of nc->recv_mbuf.buf is an op, the rest of the data is * defragmented data. */ size_t reass_len; }; struct mg_http_proto_data { #if MG_ENABLE_FILESYSTEM struct mg_http_proto_data_file file; #endif #if MG_ENABLE_HTTP_CGI struct mg_http_proto_data_cgi cgi; #endif #if MG_ENABLE_HTTP_STREAMING_MULTIPART struct mg_http_multipart_stream mp_stream; #endif #if MG_ENABLE_HTTP_WEBSOCKET struct mg_ws_proto_data ws_data; #endif struct mg_http_proto_data_chuncked chunk; struct mg_http_endpoint *endpoints; mg_event_handler_t endpoint_handler; struct mg_reverse_proxy_data reverse_proxy_data; size_t rcvd; /* How many bytes we have received. */ }; static void mg_http_conn_destructor(void *proto_data); struct mg_connection *mg_connect_http_base( struct mg_mgr *mgr, MG_CB(mg_event_handler_t ev_handler, void *user_data), struct mg_connect_opts opts, const char *scheme1, const char *scheme2, const char *scheme_ssl1, const char *scheme_ssl2, const char *url, struct mg_str *path, struct mg_str *user_info, struct mg_str *host); static struct mg_http_proto_data *mg_http_get_proto_data( struct mg_connection *c) { if (c->proto_data == NULL) { c->proto_data = MG_CALLOC(1, sizeof(struct mg_http_proto_data)); c->proto_data_destructor = mg_http_conn_destructor; } return (struct mg_http_proto_data *) c->proto_data; } #if MG_ENABLE_HTTP_STREAMING_MULTIPART static void mg_http_free_proto_data_mp_stream( struct mg_http_multipart_stream *mp) { MG_FREE((void *) mp->boundary); MG_FREE((void *) mp->var_name); MG_FREE((void *) mp->file_name); memset(mp, 0, sizeof(*mp)); } #endif #if MG_ENABLE_FILESYSTEM static void mg_http_free_proto_data_file(struct mg_http_proto_data_file *d) { if (d != NULL) { if (d->fp != NULL) { fclose(d->fp); } memset(d, 0, sizeof(struct mg_http_proto_data_file)); } } #endif static void mg_http_free_proto_data_endpoints(struct mg_http_endpoint **ep) { struct mg_http_endpoint *current = *ep; while (current != NULL) { struct mg_http_endpoint *tmp = current->next; MG_FREE((void *) current->uri_pattern.p); MG_FREE((void *) current->auth_domain); MG_FREE((void *) current->auth_file); MG_FREE(current); current = tmp; } ep = NULL; } static void mg_http_free_reverse_proxy_data(struct mg_reverse_proxy_data *rpd) { if (rpd->linked_conn != NULL) { /* * Connection has linked one, we have to unlink & close it * since _this_ connection is going to die and * it doesn't make sense to keep another one */ struct mg_http_proto_data *pd = mg_http_get_proto_data(rpd->linked_conn); if (pd->reverse_proxy_data.linked_conn != NULL) { pd->reverse_proxy_data.linked_conn->flags |= MG_F_SEND_AND_CLOSE; pd->reverse_proxy_data.linked_conn = NULL; } rpd->linked_conn = NULL; } } static void mg_http_conn_destructor(void *proto_data) { struct mg_http_proto_data *pd = (struct mg_http_proto_data *) proto_data; #if MG_ENABLE_FILESYSTEM mg_http_free_proto_data_file(&pd->file); #endif #if MG_ENABLE_HTTP_CGI mg_http_free_proto_data_cgi(&pd->cgi); #endif #if MG_ENABLE_HTTP_STREAMING_MULTIPART mg_http_free_proto_data_mp_stream(&pd->mp_stream); #endif mg_http_free_proto_data_endpoints(&pd->endpoints); mg_http_free_reverse_proxy_data(&pd->reverse_proxy_data); MG_FREE(proto_data); } #if MG_ENABLE_FILESYSTEM #define MIME_ENTRY(_ext, _type) \ { _ext, sizeof(_ext) - 1, _type } static const struct { const char *extension; size_t ext_len; const char *mime_type; } mg_static_builtin_mime_types[] = { MIME_ENTRY("html", "text/html"), MIME_ENTRY("html", "text/html"), MIME_ENTRY("htm", "text/html"), MIME_ENTRY("shtm", "text/html"), MIME_ENTRY("shtml", "text/html"), MIME_ENTRY("css", "text/css"), MIME_ENTRY("js", "application/x-javascript"), MIME_ENTRY("ico", "image/x-icon"), MIME_ENTRY("gif", "image/gif"), MIME_ENTRY("jpg", "image/jpeg"), MIME_ENTRY("jpeg", "image/jpeg"), MIME_ENTRY("png", "image/png"), MIME_ENTRY("svg", "image/svg+xml"), MIME_ENTRY("txt", "text/plain"), MIME_ENTRY("torrent", "application/x-bittorrent"), MIME_ENTRY("wav", "audio/x-wav"), MIME_ENTRY("mp3", "audio/x-mp3"), MIME_ENTRY("mid", "audio/mid"), MIME_ENTRY("m3u", "audio/x-mpegurl"), MIME_ENTRY("ogg", "application/ogg"), MIME_ENTRY("ram", "audio/x-pn-realaudio"), MIME_ENTRY("xml", "text/xml"), MIME_ENTRY("ttf", "application/x-font-ttf"), MIME_ENTRY("json", "application/json"), MIME_ENTRY("xslt", "application/xml"), MIME_ENTRY("xsl", "application/xml"), MIME_ENTRY("ra", "audio/x-pn-realaudio"), MIME_ENTRY("doc", "application/msword"), MIME_ENTRY("exe", "application/octet-stream"), MIME_ENTRY("zip", "application/x-zip-compressed"), MIME_ENTRY("xls", "application/excel"), MIME_ENTRY("tgz", "application/x-tar-gz"), MIME_ENTRY("tar", "application/x-tar"), MIME_ENTRY("gz", "application/x-gunzip"), MIME_ENTRY("arj", "application/x-arj-compressed"), MIME_ENTRY("rar", "application/x-rar-compressed"), MIME_ENTRY("rtf", "application/rtf"), MIME_ENTRY("pdf", "application/pdf"), MIME_ENTRY("swf", "application/x-shockwave-flash"), MIME_ENTRY("mpg", "video/mpeg"), MIME_ENTRY("webm", "video/webm"), MIME_ENTRY("mpeg", "video/mpeg"), MIME_ENTRY("mov", "video/quicktime"), MIME_ENTRY("mp4", "video/mp4"), MIME_ENTRY("m4v", "video/x-m4v"), MIME_ENTRY("asf", "video/x-ms-asf"), MIME_ENTRY("avi", "video/x-msvideo"), MIME_ENTRY("bmp", "image/bmp"), {NULL, 0, NULL}}; static struct mg_str mg_get_mime_type(const char *path, const char *dflt, const struct mg_serve_http_opts *opts) { const char *ext, *overrides; size_t i, path_len; struct mg_str r, k, v; path_len = strlen(path); overrides = opts->custom_mime_types; while ((overrides = mg_next_comma_list_entry(overrides, &k, &v)) != NULL) { ext = path + (path_len - k.len); if (path_len > k.len && mg_vcasecmp(&k, ext) == 0) { return v; } } for (i = 0; mg_static_builtin_mime_types[i].extension != NULL; i++) { ext = path + (path_len - mg_static_builtin_mime_types[i].ext_len); if (path_len > mg_static_builtin_mime_types[i].ext_len && ext[-1] == '.' && mg_casecmp(ext, mg_static_builtin_mime_types[i].extension) == 0) { r.p = mg_static_builtin_mime_types[i].mime_type; r.len = strlen(r.p); return r; } } r.p = dflt; r.len = strlen(r.p); return r; } #endif /* * Check whether full request is buffered. Return: * -1 if request is malformed * 0 if request is not yet fully buffered * >0 actual request length, including last \r\n\r\n */ static int mg_http_get_request_len(const char *s, int buf_len) { const unsigned char *buf = (unsigned char *) s; int i; for (i = 0; i < buf_len; i++) { if (!isprint(buf[i]) && buf[i] != '\r' && buf[i] != '\n' && buf[i] < 128) { return -1; } else if (buf[i] == '\n' && i + 1 < buf_len && buf[i + 1] == '\n') { return i + 2; } else if (buf[i] == '\n' && i + 2 < buf_len && buf[i + 1] == '\r' && buf[i + 2] == '\n') { return i + 3; } } return 0; } static const char *mg_http_parse_headers(const char *s, const char *end, int len, struct http_message *req) { int i = 0; while (i < (int) ARRAY_SIZE(req->header_names) - 1) { struct mg_str *k = &req->header_names[i], *v = &req->header_values[i]; s = mg_skip(s, end, ": ", k); s = mg_skip(s, end, "\r\n", v); while (v->len > 0 && v->p[v->len - 1] == ' ') { v->len--; /* Trim trailing spaces in header value */ } /* * If header value is empty - skip it and go to next (if any). * NOTE: Do not add it to headers_values because such addition changes API * behaviour */ if (k->len != 0 && v->len == 0) { continue; } if (k->len == 0 || v->len == 0) { k->p = v->p = NULL; k->len = v->len = 0; break; } if (!mg_ncasecmp(k->p, "Content-Length", 14)) { req->body.len = (size_t) to64(v->p); req->message.len = len + req->body.len; } i++; } return s; } int mg_parse_http(const char *s, int n, struct http_message *hm, int is_req) { const char *end, *qs; int len = mg_http_get_request_len(s, n); if (len <= 0) return len; memset(hm, 0, sizeof(*hm)); hm->message.p = s; hm->body.p = s + len; hm->message.len = hm->body.len = (size_t) ~0; end = s + len; /* Request is fully buffered. Skip leading whitespaces. */ while (s < end && isspace(*(unsigned char *) s)) s++; if (is_req) { /* Parse request line: method, URI, proto */ s = mg_skip(s, end, " ", &hm->method); s = mg_skip(s, end, " ", &hm->uri); s = mg_skip(s, end, "\r\n", &hm->proto); if (hm->uri.p <= hm->method.p || hm->proto.p <= hm->uri.p) return -1; /* If URI contains '?' character, initialize query_string */ if ((qs = (char *) memchr(hm->uri.p, '?', hm->uri.len)) != NULL) { hm->query_string.p = qs + 1; hm->query_string.len = &hm->uri.p[hm->uri.len] - (qs + 1); hm->uri.len = qs - hm->uri.p; } } else { s = mg_skip(s, end, " ", &hm->proto); if (end - s < 4 || s[3] != ' ') return -1; hm->resp_code = atoi(s); if (hm->resp_code < 100 || hm->resp_code >= 600) return -1; s += 4; s = mg_skip(s, end, "\r\n", &hm->resp_status_msg); } s = mg_http_parse_headers(s, end, len, hm); /* * mg_parse_http() is used to parse both HTTP requests and HTTP * responses. If HTTP response does not have Content-Length set, then * body is read until socket is closed, i.e. body.len is infinite (~0). * * For HTTP requests though, according to * http://tools.ietf.org/html/rfc7231#section-8.1.3, * only POST and PUT methods have defined body semantics. * Therefore, if Content-Length is not specified and methods are * not one of PUT or POST, set body length to 0. * * So, * if it is HTTP request, and Content-Length is not set, * and method is not (PUT or POST) then reset body length to zero. */ if (hm->body.len == (size_t) ~0 && is_req && mg_vcasecmp(&hm->method, "PUT") != 0 && mg_vcasecmp(&hm->method, "POST") != 0) { hm->body.len = 0; hm->message.len = len; } return len; } struct mg_str *mg_get_http_header(struct http_message *hm, const char *name) { size_t i, len = strlen(name); for (i = 0; hm->header_names[i].len > 0; i++) { struct mg_str *h = &hm->header_names[i], *v = &hm->header_values[i]; if (h->p != NULL && h->len == len && !mg_ncasecmp(h->p, name, len)) return v; } return NULL; } #if MG_ENABLE_FILESYSTEM static void mg_http_transfer_file_data(struct mg_connection *nc) { struct mg_http_proto_data *pd = mg_http_get_proto_data(nc); char buf[MG_MAX_HTTP_SEND_MBUF]; size_t n = 0, to_read = 0, left = (size_t)(pd->file.cl - pd->file.sent); if (pd->file.type == DATA_FILE) { struct mbuf *io = &nc->send_mbuf; if (io->len >= MG_MAX_HTTP_SEND_MBUF) { to_read = 0; } else { to_read = MG_MAX_HTTP_SEND_MBUF - io->len; } if (to_read > left) { to_read = left; } if (to_read > 0) { n = mg_fread(buf, 1, to_read, pd->file.fp); if (n > 0) { mg_send(nc, buf, n); pd->file.sent += n; DBG(("%p sent %d (total %d)", nc, (int) n, (int) pd->file.sent)); } } else { /* Rate-limited */ } if (pd->file.sent >= pd->file.cl) { LOG(LL_DEBUG, ("%p done, %d bytes", nc, (int) pd->file.sent)); if (!pd->file.keepalive) nc->flags |= MG_F_SEND_AND_CLOSE; mg_http_free_proto_data_file(&pd->file); } } else if (pd->file.type == DATA_PUT) { struct mbuf *io = &nc->recv_mbuf; size_t to_write = left <= 0 ? 0 : left < io->len ? (size_t) left : io->len; size_t n = mg_fwrite(io->buf, 1, to_write, pd->file.fp); if (n > 0) { mbuf_remove(io, n); pd->file.sent += n; } if (n == 0 || pd->file.sent >= pd->file.cl) { if (!pd->file.keepalive) nc->flags |= MG_F_SEND_AND_CLOSE; mg_http_free_proto_data_file(&pd->file); } } #if MG_ENABLE_HTTP_CGI else if (pd->cgi.cgi_nc != NULL) { /* This is POST data that needs to be forwarded to the CGI process */ if (pd->cgi.cgi_nc != NULL) { mg_forward(nc, pd->cgi.cgi_nc); } else { nc->flags |= MG_F_SEND_AND_CLOSE; } } #endif } #endif /* MG_ENABLE_FILESYSTEM */ /* * Parse chunked-encoded buffer. Return 0 if the buffer is not encoded, or * if it's incomplete. If the chunk is fully buffered, return total number of * bytes in a chunk, and store data in `data`, `data_len`. */ static size_t mg_http_parse_chunk(char *buf, size_t len, char **chunk_data, size_t *chunk_len) { unsigned char *s = (unsigned char *) buf; size_t n = 0; /* scanned chunk length */ size_t i = 0; /* index in s */ /* Scan chunk length. That should be a hexadecimal number. */ while (i < len && isxdigit(s[i])) { n *= 16; n += (s[i] >= '0' && s[i] <= '9') ? s[i] - '0' : tolower(s[i]) - 'a' + 10; i++; } /* Skip new line */ if (i == 0 || i + 2 > len || s[i] != '\r' || s[i + 1] != '\n') { return 0; } i += 2; /* Record where the data is */ *chunk_data = (char *) s + i; *chunk_len = n; /* Skip data */ i += n; /* Skip new line */ if (i == 0 || i + 2 > len || s[i] != '\r' || s[i + 1] != '\n') { return 0; } return i + 2; } MG_INTERNAL size_t mg_handle_chunked(struct mg_connection *nc, struct http_message *hm, char *buf, size_t blen) { struct mg_http_proto_data *pd = mg_http_get_proto_data(nc); char *data; size_t i, n, data_len, body_len, zero_chunk_received = 0; /* Find out piece of received data that is not yet reassembled */ body_len = (size_t) pd->chunk.body_len; assert(blen >= body_len); /* Traverse all fully buffered chunks */ for (i = body_len; (n = mg_http_parse_chunk(buf + i, blen - i, &data, &data_len)) > 0; i += n) { /* Collapse chunk data to the rest of HTTP body */ memmove(buf + body_len, data, data_len); body_len += data_len; hm->body.len = body_len; if (data_len == 0) { zero_chunk_received = 1; i += n; break; } } if (i > body_len) { /* Shift unparsed content to the parsed body */ assert(i <= blen); memmove(buf + body_len, buf + i, blen - i); memset(buf + body_len + blen - i, 0, i - body_len); nc->recv_mbuf.len -= i - body_len; pd->chunk.body_len = body_len; /* Send MG_EV_HTTP_CHUNK event */ nc->flags &= ~MG_F_DELETE_CHUNK; mg_call(nc, nc->handler, nc->user_data, MG_EV_HTTP_CHUNK, hm); /* Delete processed data if user set MG_F_DELETE_CHUNK flag */ if (nc->flags & MG_F_DELETE_CHUNK) { memset(buf, 0, body_len); memmove(buf, buf + body_len, blen - i); nc->recv_mbuf.len -= body_len; hm->body.len = 0; pd->chunk.body_len = 0; } if (zero_chunk_received) { /* Total message size is len(body) + len(headers) */ hm->message.len = (size_t) pd->chunk.body_len + blen - i + (hm->body.p - hm->message.p); } } return body_len; } struct mg_http_endpoint *mg_http_get_endpoint_handler(struct mg_connection *nc, struct mg_str *uri_path) { struct mg_http_proto_data *pd; struct mg_http_endpoint *ret = NULL; int matched, matched_max = 0; struct mg_http_endpoint *ep; if (nc == NULL) { return NULL; } pd = mg_http_get_proto_data(nc); ep = pd->endpoints; while (ep != NULL) { if ((matched = mg_match_prefix_n(ep->uri_pattern, *uri_path)) > 0) { if (matched > matched_max) { /* Looking for the longest suitable handler */ ret = ep; matched_max = matched; } } ep = ep->next; } return ret; } #if MG_ENABLE_HTTP_STREAMING_MULTIPART static void mg_http_multipart_continue(struct mg_connection *nc); static void mg_http_multipart_begin(struct mg_connection *nc, struct http_message *hm, int req_len); #endif static void mg_http_call_endpoint_handler(struct mg_connection *nc, int ev, struct http_message *hm); static void deliver_chunk(struct mg_connection *c, struct http_message *hm, int req_len) { /* Incomplete message received. Send MG_EV_HTTP_CHUNK event */ hm->body.len = c->recv_mbuf.len - req_len; c->flags &= ~MG_F_DELETE_CHUNK; mg_call(c, c->handler, c->user_data, MG_EV_HTTP_CHUNK, hm); /* Delete processed data if user set MG_F_DELETE_CHUNK flag */ if (c->flags & MG_F_DELETE_CHUNK) c->recv_mbuf.len = req_len; } /* * lx106 compiler has a bug (TODO(mkm) report and insert tracking bug here) * If a big structure is declared in a big function, lx106 gcc will make it * even bigger (round up to 4k, from 700 bytes of actual size). */ #ifdef __xtensa__ static void mg_http_handler2(struct mg_connection *nc, int ev, void *ev_data MG_UD_ARG(void *user_data), struct http_message *hm) __attribute__((noinline)); void mg_http_handler(struct mg_connection *nc, int ev, void *ev_data MG_UD_ARG(void *user_data)) { struct http_message hm; mg_http_handler2(nc, ev, ev_data MG_UD_ARG(user_data), &hm); } static void mg_http_handler2(struct mg_connection *nc, int ev, void *ev_data MG_UD_ARG(void *user_data), struct http_message *hm) { #else /* !__XTENSA__ */ void mg_http_handler(struct mg_connection *nc, int ev, void *ev_data MG_UD_ARG(void *user_data)) { struct http_message shm, *hm = &shm; #endif /* __XTENSA__ */ struct mg_http_proto_data *pd = mg_http_get_proto_data(nc); struct mbuf *io = &nc->recv_mbuf; int req_len; const int is_req = (nc->listener != NULL); #if MG_ENABLE_HTTP_WEBSOCKET struct mg_str *vec; #endif if (ev == MG_EV_CLOSE) { #if MG_ENABLE_HTTP_CGI /* Close associated CGI forwarder connection */ if (pd->cgi.cgi_nc != NULL) { pd->cgi.cgi_nc->user_data = NULL; pd->cgi.cgi_nc->flags |= MG_F_CLOSE_IMMEDIATELY; } #endif #if MG_ENABLE_HTTP_STREAMING_MULTIPART if (pd->mp_stream.boundary != NULL) { /* * Multipart message is in progress, but connection is closed. * Finish part and request with an error flag. */ struct mg_http_multipart_part mp; memset(&mp, 0, sizeof(mp)); mp.status = -1; mp.var_name = pd->mp_stream.var_name; mp.file_name = pd->mp_stream.file_name; mg_call(nc, (pd->endpoint_handler ? pd->endpoint_handler : nc->handler), nc->user_data, MG_EV_HTTP_PART_END, &mp); mp.var_name = NULL; mp.file_name = NULL; mg_call(nc, (pd->endpoint_handler ? pd->endpoint_handler : nc->handler), nc->user_data, MG_EV_HTTP_MULTIPART_REQUEST_END, &mp); } else #endif if (io->len > 0 && (req_len = mg_parse_http(io->buf, io->len, hm, is_req)) > 0) { /* * For HTTP messages without Content-Length, always send HTTP message * before MG_EV_CLOSE message. */ int ev2 = is_req ? MG_EV_HTTP_REQUEST : MG_EV_HTTP_REPLY; hm->message.len = io->len; hm->body.len = io->buf + io->len - hm->body.p; deliver_chunk(nc, hm, req_len); mg_http_call_endpoint_handler(nc, ev2, hm); } pd->rcvd = 0; } #if MG_ENABLE_FILESYSTEM if (pd->file.fp != NULL) { mg_http_transfer_file_data(nc); } #endif mg_call(nc, nc->handler, nc->user_data, ev, ev_data); if (ev == MG_EV_RECV) { struct mg_str *s; pd->rcvd += *(int *) ev_data; #if MG_ENABLE_HTTP_STREAMING_MULTIPART if (pd->mp_stream.boundary != NULL) { mg_http_multipart_continue(nc); return; } #endif /* MG_ENABLE_HTTP_STREAMING_MULTIPART */ req_len = mg_parse_http(io->buf, io->len, hm, is_req); if (req_len > 0 && (s = mg_get_http_header(hm, "Transfer-Encoding")) != NULL && mg_vcasecmp(s, "chunked") == 0) { mg_handle_chunked(nc, hm, io->buf + req_len, io->len - req_len); } #if MG_ENABLE_HTTP_STREAMING_MULTIPART if (req_len > 0 && (s = mg_get_http_header(hm, "Content-Type")) != NULL && s->len >= 9 && strncmp(s->p, "multipart", 9) == 0) { mg_http_multipart_begin(nc, hm, req_len); mg_http_multipart_continue(nc); return; } #endif /* MG_ENABLE_HTTP_STREAMING_MULTIPART */ /* TODO(alashkin): refactor this ifelseifelseifelseifelse */ if ((req_len < 0 || (req_len == 0 && io->len >= MG_MAX_HTTP_REQUEST_SIZE))) { DBG(("invalid request")); nc->flags |= MG_F_CLOSE_IMMEDIATELY; } else if (req_len == 0) { /* Do nothing, request is not yet fully buffered */ } #if MG_ENABLE_HTTP_WEBSOCKET else if (nc->listener == NULL && mg_get_http_header(hm, "Sec-WebSocket-Accept")) { /* We're websocket client, got handshake response from server. */ /* TODO(lsm): check the validity of accept Sec-WebSocket-Accept */ mbuf_remove(io, req_len); nc->proto_handler = mg_ws_handler; nc->flags |= MG_F_IS_WEBSOCKET; mg_call(nc, nc->handler, nc->user_data, MG_EV_WEBSOCKET_HANDSHAKE_DONE, NULL); mg_ws_handler(nc, MG_EV_RECV, ev_data MG_UD_ARG(user_data)); } else if (nc->listener != NULL && (vec = mg_get_http_header(hm, "Sec-WebSocket-Key")) != NULL) { struct mg_http_endpoint *ep; /* This is a websocket request. Switch protocol handlers. */ mbuf_remove(io, req_len); nc->proto_handler = mg_ws_handler; nc->flags |= MG_F_IS_WEBSOCKET; /* * If we have a handler set up with mg_register_http_endpoint(), * deliver subsequent websocket events to this handler after the * protocol switch. */ ep = mg_http_get_endpoint_handler(nc->listener, &hm->uri); if (ep != NULL) { nc->handler = ep->handler; #if MG_ENABLE_CALLBACK_USERDATA nc->user_data = ep->user_data; #endif } /* Send handshake */ mg_call(nc, nc->handler, nc->user_data, MG_EV_WEBSOCKET_HANDSHAKE_REQUEST, hm); if (!(nc->flags & (MG_F_CLOSE_IMMEDIATELY | MG_F_SEND_AND_CLOSE))) { if (nc->send_mbuf.len == 0) { mg_ws_handshake(nc, vec, hm); } mg_call(nc, nc->handler, nc->user_data, MG_EV_WEBSOCKET_HANDSHAKE_DONE, NULL); mg_ws_handler(nc, MG_EV_RECV, ev_data MG_UD_ARG(user_data)); } } #endif /* MG_ENABLE_HTTP_WEBSOCKET */ else if (hm->message.len > pd->rcvd) { /* Not yet received all HTTP body, deliver MG_EV_HTTP_CHUNK */ deliver_chunk(nc, hm, req_len); if (nc->recv_mbuf_limit > 0 && nc->recv_mbuf.len >= nc->recv_mbuf_limit) { LOG(LL_ERROR, ("%p recv buffer (%lu bytes) exceeds the limit " "%lu bytes, and not drained, closing", nc, (unsigned long) nc->recv_mbuf.len, (unsigned long) nc->recv_mbuf_limit)); nc->flags |= MG_F_CLOSE_IMMEDIATELY; } } else { /* We did receive all HTTP body. */ int trigger_ev = nc->listener ? MG_EV_HTTP_REQUEST : MG_EV_HTTP_REPLY; char addr[32]; mg_sock_addr_to_str(&nc->sa, addr, sizeof(addr), MG_SOCK_STRINGIFY_IP | MG_SOCK_STRINGIFY_PORT); DBG(("%p %s %.*s %.*s", nc, addr, (int) hm->method.len, hm->method.p, (int) hm->uri.len, hm->uri.p)); deliver_chunk(nc, hm, req_len); /* Whole HTTP message is fully buffered, call event handler */ mg_http_call_endpoint_handler(nc, trigger_ev, hm); mbuf_remove(io, hm->message.len); pd->rcvd = 0; } } } static size_t mg_get_line_len(const char *buf, size_t buf_len) { size_t len = 0; while (len < buf_len && buf[len] != '\n') len++; return len == buf_len ? 0 : len + 1; } #if MG_ENABLE_HTTP_STREAMING_MULTIPART static void mg_http_multipart_begin(struct mg_connection *nc, struct http_message *hm, int req_len) { struct mg_http_proto_data *pd = mg_http_get_proto_data(nc); struct mg_str *ct; struct mbuf *io = &nc->recv_mbuf; char boundary_buf[100]; char *boundary = boundary_buf; int boundary_len; ct = mg_get_http_header(hm, "Content-Type"); if (ct == NULL) { /* We need more data - or it isn't multipart mesage */ goto exit_mp; } /* Content-type should start with "multipart" */ if (ct->len < 9 || strncmp(ct->p, "multipart", 9) != 0) { goto exit_mp; } boundary_len = mg_http_parse_header2(ct, "boundary", &boundary, sizeof(boundary_buf)); if (boundary_len == 0) { /* * Content type is multipart, but there is no boundary, * probably malformed request */ nc->flags = MG_F_CLOSE_IMMEDIATELY; DBG(("invalid request")); goto exit_mp; } /* If we reach this place - that is multipart request */ if (pd->mp_stream.boundary != NULL) { /* * Another streaming request was in progress, * looks like protocol error */ nc->flags |= MG_F_CLOSE_IMMEDIATELY; } else { struct mg_http_endpoint *ep = NULL; pd->mp_stream.state = MPS_BEGIN; pd->mp_stream.boundary = strdup(boundary); pd->mp_stream.boundary_len = strlen(boundary); pd->mp_stream.var_name = pd->mp_stream.file_name = NULL; pd->endpoint_handler = nc->handler; ep = mg_http_get_endpoint_handler(nc->listener, &hm->uri); if (ep != NULL) { pd->endpoint_handler = ep->handler; } mg_http_call_endpoint_handler(nc, MG_EV_HTTP_MULTIPART_REQUEST, hm); mbuf_remove(io, req_len); } exit_mp: if (boundary != boundary_buf) MG_FREE(boundary); } #define CONTENT_DISPOSITION "Content-Disposition: " static void mg_http_multipart_call_handler(struct mg_connection *c, int ev, const char *data, size_t data_len) { struct mg_http_multipart_part mp; struct mg_http_proto_data *pd = mg_http_get_proto_data(c); memset(&mp, 0, sizeof(mp)); mp.var_name = pd->mp_stream.var_name; mp.file_name = pd->mp_stream.file_name; mp.user_data = pd->mp_stream.user_data; mp.data.p = data; mp.data.len = data_len; mg_call(c, pd->endpoint_handler, c->user_data, ev, &mp); pd->mp_stream.user_data = mp.user_data; } static int mg_http_multipart_got_chunk(struct mg_connection *c) { struct mg_http_proto_data *pd = mg_http_get_proto_data(c); struct mbuf *io = &c->recv_mbuf; mg_http_multipart_call_handler(c, MG_EV_HTTP_PART_DATA, io->buf, pd->mp_stream.prev_io_len); mbuf_remove(io, pd->mp_stream.prev_io_len); pd->mp_stream.prev_io_len = 0; pd->mp_stream.state = MPS_WAITING_FOR_CHUNK; return 0; } static int mg_http_multipart_finalize(struct mg_connection *c) { struct mg_http_proto_data *pd = mg_http_get_proto_data(c); mg_http_multipart_call_handler(c, MG_EV_HTTP_PART_END, NULL, 0); MG_FREE((void *) pd->mp_stream.file_name); pd->mp_stream.file_name = NULL; MG_FREE((void *) pd->mp_stream.var_name); pd->mp_stream.var_name = NULL; mg_http_multipart_call_handler(c, MG_EV_HTTP_MULTIPART_REQUEST_END, NULL, 0); mg_http_free_proto_data_mp_stream(&pd->mp_stream); pd->mp_stream.state = MPS_FINISHED; return 1; } static int mg_http_multipart_wait_for_boundary(struct mg_connection *c) { const char *boundary; struct mbuf *io = &c->recv_mbuf; struct mg_http_proto_data *pd = mg_http_get_proto_data(c); if (pd->mp_stream.boundary == NULL) { pd->mp_stream.state = MPS_FINALIZE; DBG(("Invalid request: boundary not initialized")); return 0; } if ((int) io->len < pd->mp_stream.boundary_len + 2) { return 0; } boundary = c_strnstr(io->buf, pd->mp_stream.boundary, io->len); if (boundary != NULL) { const char *boundary_end = (boundary + pd->mp_stream.boundary_len); if (io->len - (boundary_end - io->buf) < 4) { return 0; } if (strncmp(boundary_end, "--\r\n", 4) == 0) { pd->mp_stream.state = MPS_FINALIZE; mbuf_remove(io, (boundary_end - io->buf) + 4); } else { pd->mp_stream.state = MPS_GOT_BOUNDARY; } } else { return 0; } return 1; } static void mg_http_parse_header_internal(struct mg_str *hdr, const char *var_name, struct altbuf *ab); static int mg_http_multipart_process_boundary(struct mg_connection *c) { int data_size; const char *boundary, *block_begin; struct mbuf *io = &c->recv_mbuf; struct mg_http_proto_data *pd = mg_http_get_proto_data(c); struct altbuf ab_file_name, ab_var_name; int line_len; boundary = c_strnstr(io->buf, pd->mp_stream.boundary, io->len); block_begin = boundary + pd->mp_stream.boundary_len + 2; data_size = io->len - (block_begin - io->buf); altbuf_init(&ab_file_name, NULL, 0); altbuf_init(&ab_var_name, NULL, 0); while (data_size > 0 && (line_len = mg_get_line_len(block_begin, data_size)) != 0) { if (line_len > (int) sizeof(CONTENT_DISPOSITION) && mg_ncasecmp(block_begin, CONTENT_DISPOSITION, sizeof(CONTENT_DISPOSITION) - 1) == 0) { struct mg_str header; header.p = block_begin + sizeof(CONTENT_DISPOSITION) - 1; header.len = line_len - sizeof(CONTENT_DISPOSITION) - 1; altbuf_reset(&ab_var_name); mg_http_parse_header_internal(&header, "name", &ab_var_name); altbuf_reset(&ab_file_name); mg_http_parse_header_internal(&header, "filename", &ab_file_name); block_begin += line_len; data_size -= line_len; continue; } if (line_len == 2 && mg_ncasecmp(block_begin, "\r\n", 2) == 0) { mbuf_remove(io, block_begin - io->buf + 2); if (pd->mp_stream.processing_part != 0) { mg_http_multipart_call_handler(c, MG_EV_HTTP_PART_END, NULL, 0); } /* Reserve 2 bytes for "\r\n" in file_name and var_name */ altbuf_append(&ab_file_name, '\0'); altbuf_append(&ab_file_name, '\0'); altbuf_append(&ab_var_name, '\0'); altbuf_append(&ab_var_name, '\0'); MG_FREE((void *) pd->mp_stream.file_name); pd->mp_stream.file_name = altbuf_get_buf(&ab_file_name, 1 /* trim */); MG_FREE((void *) pd->mp_stream.var_name); pd->mp_stream.var_name = altbuf_get_buf(&ab_var_name, 1 /* trim */); mg_http_multipart_call_handler(c, MG_EV_HTTP_PART_BEGIN, NULL, 0); pd->mp_stream.state = MPS_WAITING_FOR_CHUNK; pd->mp_stream.processing_part++; return 1; } block_begin += line_len; } pd->mp_stream.state = MPS_WAITING_FOR_BOUNDARY; altbuf_reset(&ab_var_name); altbuf_reset(&ab_file_name); return 0; } static int mg_http_multipart_continue_wait_for_chunk(struct mg_connection *c) { struct mg_http_proto_data *pd = mg_http_get_proto_data(c); struct mbuf *io = &c->recv_mbuf; const char *boundary; if ((int) io->len < pd->mp_stream.boundary_len + 6 /* \r\n, --, -- */) { return 0; } boundary = c_strnstr(io->buf, pd->mp_stream.boundary, io->len); if (boundary == NULL && pd->mp_stream.prev_io_len == 0) { pd->mp_stream.prev_io_len = io->len; return 0; } else if (boundary == NULL && (int) io->len > pd->mp_stream.prev_io_len + pd->mp_stream.boundary_len + 4) { pd->mp_stream.state = MPS_GOT_CHUNK; return 1; } else if (boundary != NULL) { int data_size = (boundary - io->buf - 4); mg_http_multipart_call_handler(c, MG_EV_HTTP_PART_DATA, io->buf, data_size); mbuf_remove(io, (boundary - io->buf)); pd->mp_stream.prev_io_len = 0; pd->mp_stream.state = MPS_WAITING_FOR_BOUNDARY; return 1; } else { return 0; } } static void mg_http_multipart_continue(struct mg_connection *c) { struct mg_http_proto_data *pd = mg_http_get_proto_data(c); while (1) { switch (pd->mp_stream.state) { case MPS_BEGIN: { pd->mp_stream.state = MPS_WAITING_FOR_BOUNDARY; break; } case MPS_WAITING_FOR_BOUNDARY: { if (mg_http_multipart_wait_for_boundary(c) == 0) { return; } break; } case MPS_GOT_BOUNDARY: { if (mg_http_multipart_process_boundary(c) == 0) { return; } break; } case MPS_WAITING_FOR_CHUNK: { if (mg_http_multipart_continue_wait_for_chunk(c) == 0) { return; } break; } case MPS_GOT_CHUNK: { if (mg_http_multipart_got_chunk(c) == 0) { return; } break; } case MPS_FINALIZE: { if (mg_http_multipart_finalize(c) == 0) { return; } break; } case MPS_FINISHED: { mbuf_remove(&c->recv_mbuf, c->recv_mbuf.len); return; } } } } struct file_upload_state { char *lfn; size_t num_recd; FILE *fp; }; #endif /* MG_ENABLE_HTTP_STREAMING_MULTIPART */ void mg_set_protocol_http_websocket(struct mg_connection *nc) { nc->proto_handler = mg_http_handler; } const char *mg_status_message(int status_code) { switch (status_code) { case 206: return "Partial Content"; case 301: return "Moved"; case 302: return "Found"; case 400: return "Bad Request"; case 401: return "Unauthorized"; case 403: return "Forbidden"; case 404: return "Not Found"; case 416: return "Requested Range Not Satisfiable"; case 418: return "I'm a teapot"; case 500: return "Internal Server Error"; case 502: return "Bad Gateway"; case 503: return "Service Unavailable"; #if MG_ENABLE_EXTRA_ERRORS_DESC case 100: return "Continue"; case 101: return "Switching Protocols"; case 102: return "Processing"; case 200: return "OK"; case 201: return "Created"; case 202: return "Accepted"; case 203: return "Non-Authoritative Information"; case 204: return "No Content"; case 205: return "Reset Content"; case 207: return "Multi-Status"; case 208: return "Already Reported"; case 226: return "IM Used"; case 300: return "Multiple Choices"; case 303: return "See Other"; case 304: return "Not Modified"; case 305: return "Use Proxy"; case 306: return "Switch Proxy"; case 307: return "Temporary Redirect"; case 308: return "Permanent Redirect"; case 402: return "Payment Required"; case 405: return "Method Not Allowed"; case 406: return "Not Acceptable"; case 407: return "Proxy Authentication Required"; case 408: return "Request Timeout"; case 409: return "Conflict"; case 410: return "Gone"; case 411: return "Length Required"; case 412: return "Precondition Failed"; case 413: return "Payload Too Large"; case 414: return "URI Too Long"; case 415: return "Unsupported Media Type"; case 417: return "Expectation Failed"; case 422: return "Unprocessable Entity"; case 423: return "Locked"; case 424: return "Failed Dependency"; case 426: return "Upgrade Required"; case 428: return "Precondition Required"; case 429: return "Too Many Requests"; case 431: return "Request Header Fields Too Large"; case 451: return "Unavailable For Legal Reasons"; case 501: return "Not Implemented"; case 504: return "Gateway Timeout"; case 505: return "HTTP Version Not Supported"; case 506: return "Variant Also Negotiates"; case 507: return "Insufficient Storage"; case 508: return "Loop Detected"; case 510: return "Not Extended"; case 511: return "Network Authentication Required"; #endif /* MG_ENABLE_EXTRA_ERRORS_DESC */ default: return "OK"; } } void mg_send_response_line_s(struct mg_connection *nc, int status_code, const struct mg_str extra_headers) { mg_printf(nc, "HTTP/1.1 %d %s\r\nServer: %s\r\n", status_code, mg_status_message(status_code), mg_version_header); if (extra_headers.len > 0) { mg_printf(nc, "%.*s\r\n", (int) extra_headers.len, extra_headers.p); } } void mg_send_response_line(struct mg_connection *nc, int status_code, const char *extra_headers) { mg_send_response_line_s(nc, status_code, mg_mk_str(extra_headers)); } void mg_http_send_redirect(struct mg_connection *nc, int status_code, const struct mg_str location, const struct mg_str extra_headers) { char bbody[100], *pbody = bbody; int bl = mg_asprintf(&pbody, sizeof(bbody), "

Moved here.\r\n", (int) location.len, location.p); char bhead[150], *phead = bhead; mg_asprintf(&phead, sizeof(bhead), "Location: %.*s\r\n" "Content-Type: text/html\r\n" "Content-Length: %d\r\n" "Cache-Control: no-cache\r\n" "%.*s%s", (int) location.len, location.p, bl, (int) extra_headers.len, extra_headers.p, (extra_headers.len > 0 ? "\r\n" : "")); mg_send_response_line(nc, status_code, phead); if (phead != bhead) MG_FREE(phead); mg_send(nc, pbody, bl); if (pbody != bbody) MG_FREE(pbody); } void mg_send_head(struct mg_connection *c, int status_code, int64_t content_length, const char *extra_headers) { mg_send_response_line(c, status_code, extra_headers); if (content_length < 0) { mg_printf(c, "%s", "Transfer-Encoding: chunked\r\n"); } else { mg_printf(c, "Content-Length: %" INT64_FMT "\r\n", content_length); } mg_send(c, "\r\n", 2); } void mg_http_send_error(struct mg_connection *nc, int code, const char *reason) { if (!reason) reason = mg_status_message(code); LOG(LL_DEBUG, ("%p %d %s", nc, code, reason)); mg_send_head(nc, code, strlen(reason), "Content-Type: text/plain\r\nConnection: close"); mg_send(nc, reason, strlen(reason)); nc->flags |= MG_F_SEND_AND_CLOSE; } #if MG_ENABLE_FILESYSTEM static void mg_http_construct_etag(char *buf, size_t buf_len, const cs_stat_t *st) { snprintf(buf, buf_len, "\"%lx.%" INT64_FMT "\"", (unsigned long) st->st_mtime, (int64_t) st->st_size); } #ifndef WINCE static void mg_gmt_time_string(char *buf, size_t buf_len, time_t *t) { strftime(buf, buf_len, "%a, %d %b %Y %H:%M:%S GMT", gmtime(t)); } #else /* Look wince_lib.c for WindowsCE implementation */ static void mg_gmt_time_string(char *buf, size_t buf_len, time_t *t); #endif static int mg_http_parse_range_header(const struct mg_str *header, int64_t *a, int64_t *b) { /* * There is no snscanf. Headers are not guaranteed to be NUL-terminated, * so we have this. Ugh. */ int result; char *p = (char *) MG_MALLOC(header->len + 1); if (p == NULL) return 0; memcpy(p, header->p, header->len); p[header->len] = '\0'; result = sscanf(p, "bytes=%" INT64_FMT "-%" INT64_FMT, a, b); MG_FREE(p); return result; } void mg_http_serve_file(struct mg_connection *nc, struct http_message *hm, const char *path, const struct mg_str mime_type, const struct mg_str extra_headers) { struct mg_http_proto_data *pd = mg_http_get_proto_data(nc); cs_stat_t st; LOG(LL_DEBUG, ("%p [%s] %.*s", nc, path, (int) mime_type.len, mime_type.p)); if (mg_stat(path, &st) != 0 || (pd->file.fp = mg_fopen(path, "rb")) == NULL) { int code, err = mg_get_errno(); switch (err) { case EACCES: code = 403; break; case ENOENT: code = 404; break; default: code = 500; }; mg_http_send_error(nc, code, "Open failed"); } else { char etag[50], current_time[50], last_modified[50], range[70]; time_t t = (time_t) mg_time(); int64_t r1 = 0, r2 = 0, cl = st.st_size; struct mg_str *range_hdr = mg_get_http_header(hm, "Range"); int n, status_code = 200; /* Handle Range header */ range[0] = '\0'; if (range_hdr != NULL && (n = mg_http_parse_range_header(range_hdr, &r1, &r2)) > 0 && r1 >= 0 && r2 >= 0) { /* If range is specified like "400-", set second limit to content len */ if (n == 1) { r2 = cl - 1; } if (r1 > r2 || r2 >= cl) { status_code = 416; cl = 0; snprintf(range, sizeof(range), "Content-Range: bytes */%" INT64_FMT "\r\n", (int64_t) st.st_size); } else { status_code = 206; cl = r2 - r1 + 1; snprintf(range, sizeof(range), "Content-Range: bytes %" INT64_FMT "-%" INT64_FMT "/%" INT64_FMT "\r\n", r1, r1 + cl - 1, (int64_t) st.st_size); #if _FILE_OFFSET_BITS == 64 || _POSIX_C_SOURCE >= 200112L || \ _XOPEN_SOURCE >= 600 fseeko(pd->file.fp, r1, SEEK_SET); #else fseek(pd->file.fp, (long) r1, SEEK_SET); #endif } } #if !MG_DISABLE_HTTP_KEEP_ALIVE { struct mg_str *conn_hdr = mg_get_http_header(hm, "Connection"); if (conn_hdr != NULL) { pd->file.keepalive = (mg_vcasecmp(conn_hdr, "keep-alive") == 0); } else { pd->file.keepalive = (mg_vcmp(&hm->proto, "HTTP/1.1") == 0); } } #endif mg_http_construct_etag(etag, sizeof(etag), &st); mg_gmt_time_string(current_time, sizeof(current_time), &t); mg_gmt_time_string(last_modified, sizeof(last_modified), &st.st_mtime); /* * Content length casted to size_t because: * 1) that's the maximum buffer size anyway * 2) ESP8266 RTOS SDK newlib vprintf cannot contain a 64bit arg at non-last * position * TODO(mkm): fix ESP8266 RTOS SDK */ mg_send_response_line_s(nc, status_code, extra_headers); mg_printf(nc, "Date: %s\r\n" "Last-Modified: %s\r\n" "Accept-Ranges: bytes\r\n" "Content-Type: %.*s\r\n" "Connection: %s\r\n" "Content-Length: %" SIZE_T_FMT "\r\n" "%sEtag: %s\r\n\r\n", current_time, last_modified, (int) mime_type.len, mime_type.p, (pd->file.keepalive ? "keep-alive" : "close"), (size_t) cl, range, etag); pd->file.cl = cl; pd->file.type = DATA_FILE; mg_http_transfer_file_data(nc); } } static void mg_http_serve_file2(struct mg_connection *nc, const char *path, struct http_message *hm, struct mg_serve_http_opts *opts) { #if MG_ENABLE_HTTP_SSI if (mg_match_prefix(opts->ssi_pattern, strlen(opts->ssi_pattern), path) > 0) { mg_handle_ssi_request(nc, hm, path, opts); return; } #endif mg_http_serve_file(nc, hm, path, mg_get_mime_type(path, "text/plain", opts), mg_mk_str(opts->extra_headers)); } #endif int mg_url_decode(const char *src, int src_len, char *dst, int dst_len, int is_form_url_encoded) { int i, j, a, b; #define HEXTOI(x) (isdigit(x) ? x - '0' : x - 'W') for (i = j = 0; i < src_len && j < dst_len - 1; i++, j++) { if (src[i] == '%') { if (i < src_len - 2 && isxdigit(*(const unsigned char *) (src + i + 1)) && isxdigit(*(const unsigned char *) (src + i + 2))) { a = tolower(*(const unsigned char *) (src + i + 1)); b = tolower(*(const unsigned char *) (src + i + 2)); dst[j] = (char) ((HEXTOI(a) << 4) | HEXTOI(b)); i += 2; } else { return -1; } } else if (is_form_url_encoded && src[i] == '+') { dst[j] = ' '; } else { dst[j] = src[i]; } } dst[j] = '\0'; /* Null-terminate the destination */ return i >= src_len ? j : -1; } int mg_get_http_var(const struct mg_str *buf, const char *name, char *dst, size_t dst_len) { const char *p, *e, *s; size_t name_len; int len; /* * According to the documentation function returns negative * value in case of error. For debug purposes it returns: * -1 - src is wrong (NUUL) * -2 - dst is wrong (NULL) * -3 - failed to decode url or dst is to small * -4 - name does not exist */ if (dst == NULL || dst_len == 0) { len = -2; } else if (buf->p == NULL || name == NULL || buf->len == 0) { len = -1; dst[0] = '\0'; } else { name_len = strlen(name); e = buf->p + buf->len; len = -4; dst[0] = '\0'; for (p = buf->p; p + name_len < e; p++) { if ((p == buf->p || p[-1] == '&') && p[name_len] == '=' && !mg_ncasecmp(name, p, name_len)) { p += name_len + 1; s = (const char *) memchr(p, '&', (size_t)(e - p)); if (s == NULL) { s = e; } len = mg_url_decode(p, (size_t)(s - p), dst, dst_len, 1); /* -1 means: failed to decode or dst is too small */ if (len == -1) { len = -3; } break; } } } return len; } void mg_send_http_chunk(struct mg_connection *nc, const char *buf, size_t len) { char chunk_size[50]; int n; n = snprintf(chunk_size, sizeof(chunk_size), "%lX\r\n", (unsigned long) len); mg_send(nc, chunk_size, n); mg_send(nc, buf, len); mg_send(nc, "\r\n", 2); } void mg_printf_http_chunk(struct mg_connection *nc, const char *fmt, ...) { char mem[MG_VPRINTF_BUFFER_SIZE], *buf = mem; int len; va_list ap; va_start(ap, fmt); len = mg_avprintf(&buf, sizeof(mem), fmt, ap); va_end(ap); if (len >= 0) { mg_send_http_chunk(nc, buf, len); } /* LCOV_EXCL_START */ if (buf != mem && buf != NULL) { MG_FREE(buf); } /* LCOV_EXCL_STOP */ } void mg_printf_html_escape(struct mg_connection *nc, const char *fmt, ...) { char mem[MG_VPRINTF_BUFFER_SIZE], *buf = mem; int i, j, len; va_list ap; va_start(ap, fmt); len = mg_avprintf(&buf, sizeof(mem), fmt, ap); va_end(ap); if (len >= 0) { for (i = j = 0; i < len; i++) { if (buf[i] == '<' || buf[i] == '>') { mg_send(nc, buf + j, i - j); mg_send(nc, buf[i] == '<' ? "<" : ">", 4); j = i + 1; } } mg_send(nc, buf + j, i - j); } /* LCOV_EXCL_START */ if (buf != mem && buf != NULL) { MG_FREE(buf); } /* LCOV_EXCL_STOP */ } static void mg_http_parse_header_internal(struct mg_str *hdr, const char *var_name, struct altbuf *ab) { int ch = ' ', ch1 = ',', n = strlen(var_name); const char *p, *end = hdr ? hdr->p + hdr->len : NULL, *s = NULL; /* Find where variable starts */ for (s = hdr->p; s != NULL && s + n < end; s++) { if ((s == hdr->p || s[-1] == ch || s[-1] == ch1 || s[-1] == ';') && s[n] == '=' && !strncmp(s, var_name, n)) break; } if (s != NULL && &s[n + 1] < end) { s += n + 1; if (*s == '"' || *s == '\'') { ch = ch1 = *s++; } p = s; while (p < end && p[0] != ch && p[0] != ch1) { if (ch != ' ' && p[0] == '\\' && p[1] == ch) p++; altbuf_append(ab, *p++); } if (ch != ' ' && *p != ch) { altbuf_reset(ab); } } /* If there is some data, append a NUL. */ if (ab->len > 0) { altbuf_append(ab, '\0'); } } int mg_http_parse_header2(struct mg_str *hdr, const char *var_name, char **buf, size_t buf_size) { struct altbuf ab; altbuf_init(&ab, *buf, buf_size); if (hdr == NULL) return 0; if (*buf != NULL && buf_size > 0) *buf[0] = '\0'; mg_http_parse_header_internal(hdr, var_name, &ab); /* * Get a (trimmed) buffer, and return a len without a NUL byte which might * have been added. */ *buf = altbuf_get_buf(&ab, 1 /* trim */); return ab.len > 0 ? ab.len - 1 : 0; } int mg_http_parse_header(struct mg_str *hdr, const char *var_name, char *buf, size_t buf_size) { char *buf2 = buf; int len = mg_http_parse_header2(hdr, var_name, &buf2, buf_size); if (buf2 != buf) { /* Buffer was not enough and was reallocated: free it and just return 0 */ MG_FREE(buf2); return 0; } return len; } int mg_get_http_basic_auth(struct http_message *hm, char *user, size_t user_len, char *pass, size_t pass_len) { struct mg_str *hdr = mg_get_http_header(hm, "Authorization"); if (hdr == NULL) return -1; return mg_parse_http_basic_auth(hdr, user, user_len, pass, pass_len); } int mg_parse_http_basic_auth(struct mg_str *hdr, char *user, size_t user_len, char *pass, size_t pass_len) { char *buf = NULL; char fmt[64]; int res = 0; if (mg_strncmp(*hdr, mg_mk_str("Basic "), 6) != 0) return -1; buf = (char *) MG_MALLOC(hdr->len); cs_base64_decode((unsigned char *) hdr->p + 6, hdr->len, buf, NULL); /* e.g. "%123[^:]:%321[^\n]" */ snprintf(fmt, sizeof(fmt), "%%%" SIZE_T_FMT "[^:]:%%%" SIZE_T_FMT "[^\n]", user_len - 1, pass_len - 1); if (sscanf(buf, fmt, user, pass) == 0) { res = -1; } MG_FREE(buf); return res; } #if MG_ENABLE_FILESYSTEM static int mg_is_file_hidden(const char *path, const struct mg_serve_http_opts *opts, int exclude_specials) { const char *p1 = opts->per_directory_auth_file; const char *p2 = opts->hidden_file_pattern; /* Strip directory path from the file name */ const char *pdir = strrchr(path, DIRSEP); if (pdir != NULL) { path = pdir + 1; } return (exclude_specials && (!strcmp(path, ".") || !strcmp(path, ".."))) || (p1 != NULL && mg_match_prefix(p1, strlen(p1), path) == strlen(p1)) || (p2 != NULL && mg_match_prefix(p2, strlen(p2), path) > 0); } #if !MG_DISABLE_HTTP_DIGEST_AUTH #ifndef MG_EXT_MD5 void mg_hash_md5_v(size_t num_msgs, const uint8_t *msgs[], const size_t *msg_lens, uint8_t *digest) { size_t i; cs_md5_ctx md5_ctx; cs_md5_init(&md5_ctx); for (i = 0; i < num_msgs; i++) { cs_md5_update(&md5_ctx, msgs[i], msg_lens[i]); } cs_md5_final(digest, &md5_ctx); } #else extern void mg_hash_md5_v(size_t num_msgs, const uint8_t *msgs[], const size_t *msg_lens, uint8_t *digest); #endif void cs_md5(char buf[33], ...) { unsigned char hash[16]; const uint8_t *msgs[20], *p; size_t msg_lens[20]; size_t num_msgs = 0; va_list ap; va_start(ap, buf); while ((p = va_arg(ap, const unsigned char *) ) != NULL) { msgs[num_msgs] = p; msg_lens[num_msgs] = va_arg(ap, size_t); num_msgs++; } va_end(ap); mg_hash_md5_v(num_msgs, msgs, msg_lens, hash); cs_to_hex(buf, hash, sizeof(hash)); } static void mg_mkmd5resp(const char *method, size_t method_len, const char *uri, size_t uri_len, const char *ha1, size_t ha1_len, const char *nonce, size_t nonce_len, const char *nc, size_t nc_len, const char *cnonce, size_t cnonce_len, const char *qop, size_t qop_len, char *resp) { static const char colon[] = ":"; static const size_t one = 1; char ha2[33]; cs_md5(ha2, method, method_len, colon, one, uri, uri_len, NULL); cs_md5(resp, ha1, ha1_len, colon, one, nonce, nonce_len, colon, one, nc, nc_len, colon, one, cnonce, cnonce_len, colon, one, qop, qop_len, colon, one, ha2, sizeof(ha2) - 1, NULL); } int mg_http_create_digest_auth_header(char *buf, size_t buf_len, const char *method, const char *uri, const char *auth_domain, const char *user, const char *passwd, const char *nonce) { static const char colon[] = ":", qop[] = "auth"; static const size_t one = 1; char ha1[33], resp[33], cnonce[40]; snprintf(cnonce, sizeof(cnonce), "%lx", (unsigned long) mg_time()); cs_md5(ha1, user, (size_t) strlen(user), colon, one, auth_domain, (size_t) strlen(auth_domain), colon, one, passwd, (size_t) strlen(passwd), NULL); mg_mkmd5resp(method, strlen(method), uri, strlen(uri), ha1, sizeof(ha1) - 1, nonce, strlen(nonce), "1", one, cnonce, strlen(cnonce), qop, sizeof(qop) - 1, resp); return snprintf(buf, buf_len, "Authorization: Digest username=\"%s\"," "realm=\"%s\",uri=\"%s\",qop=%s,nc=1,cnonce=%s," "nonce=%s,response=%s\r\n", user, auth_domain, uri, qop, cnonce, nonce, resp); } /* * Check for authentication timeout. * Clients send time stamp encoded in nonce. Make sure it is not too old, * to prevent replay attacks. * Assumption: nonce is a hexadecimal number of seconds since 1970. */ static int mg_check_nonce(const char *nonce) { unsigned long now = (unsigned long) mg_time(); unsigned long val = (unsigned long) strtoul(nonce, NULL, 16); return (now >= val) && (now - val < 60 * 60); } int mg_http_check_digest_auth(struct http_message *hm, const char *auth_domain, FILE *fp) { int ret = 0; struct mg_str *hdr; char username_buf[50], cnonce_buf[64], response_buf[40], uri_buf[200], qop_buf[20], nc_buf[20], nonce_buf[16]; char *username = username_buf, *cnonce = cnonce_buf, *response = response_buf, *uri = uri_buf, *qop = qop_buf, *nc = nc_buf, *nonce = nonce_buf; /* Parse "Authorization:" header, fail fast on parse error */ if (hm == NULL || fp == NULL || (hdr = mg_get_http_header(hm, "Authorization")) == NULL || mg_http_parse_header2(hdr, "username", &username, sizeof(username_buf)) == 0 || mg_http_parse_header2(hdr, "cnonce", &cnonce, sizeof(cnonce_buf)) == 0 || mg_http_parse_header2(hdr, "response", &response, sizeof(response_buf)) == 0 || mg_http_parse_header2(hdr, "uri", &uri, sizeof(uri_buf)) == 0 || mg_http_parse_header2(hdr, "qop", &qop, sizeof(qop_buf)) == 0 || mg_http_parse_header2(hdr, "nc", &nc, sizeof(nc_buf)) == 0 || mg_http_parse_header2(hdr, "nonce", &nonce, sizeof(nonce_buf)) == 0 || mg_check_nonce(nonce) == 0) { ret = 0; goto clean; } /* NOTE(lsm): due to a bug in MSIE, we do not compare URIs */ ret = mg_check_digest_auth( hm->method, mg_mk_str_n( hm->uri.p, hm->uri.len + (hm->query_string.len ? hm->query_string.len + 1 : 0)), mg_mk_str(username), mg_mk_str(cnonce), mg_mk_str(response), mg_mk_str(qop), mg_mk_str(nc), mg_mk_str(nonce), mg_mk_str(auth_domain), fp); clean: if (username != username_buf) MG_FREE(username); if (cnonce != cnonce_buf) MG_FREE(cnonce); if (response != response_buf) MG_FREE(response); if (uri != uri_buf) MG_FREE(uri); if (qop != qop_buf) MG_FREE(qop); if (nc != nc_buf) MG_FREE(nc); if (nonce != nonce_buf) MG_FREE(nonce); return ret; } int mg_check_digest_auth(struct mg_str method, struct mg_str uri, struct mg_str username, struct mg_str cnonce, struct mg_str response, struct mg_str qop, struct mg_str nc, struct mg_str nonce, struct mg_str auth_domain, FILE *fp) { char buf[128], f_user[sizeof(buf)], f_ha1[sizeof(buf)], f_domain[sizeof(buf)]; char expected_response[33]; /* * Read passwords file line by line. If should have htdigest format, * i.e. each line should be a colon-separated sequence: * USER_NAME:DOMAIN_NAME:HA1_HASH_OF_USER_DOMAIN_AND_PASSWORD */ while (fgets(buf, sizeof(buf), fp) != NULL) { if (sscanf(buf, "%[^:]:%[^:]:%s", f_user, f_domain, f_ha1) == 3 && mg_vcmp(&username, f_user) == 0 && mg_vcmp(&auth_domain, f_domain) == 0) { /* Username and domain matched, check the password */ mg_mkmd5resp(method.p, method.len, uri.p, uri.len, f_ha1, strlen(f_ha1), nonce.p, nonce.len, nc.p, nc.len, cnonce.p, cnonce.len, qop.p, qop.len, expected_response); LOG(LL_DEBUG, ("%.*s %s %.*s %s", (int) username.len, username.p, f_domain, (int) response.len, response.p, expected_response)); return mg_ncasecmp(response.p, expected_response, response.len) == 0; } } /* None of the entries in the passwords file matched - return failure */ return 0; } int mg_http_is_authorized(struct http_message *hm, struct mg_str path, const char *domain, const char *passwords_file, int flags) { char buf[MG_MAX_PATH]; const char *p; FILE *fp; int authorized = 1; if (domain != NULL && passwords_file != NULL) { if (flags & MG_AUTH_FLAG_IS_GLOBAL_PASS_FILE) { fp = mg_fopen(passwords_file, "r"); } else if (flags & MG_AUTH_FLAG_IS_DIRECTORY) { snprintf(buf, sizeof(buf), "%.*s%c%s", (int) path.len, path.p, DIRSEP, passwords_file); fp = mg_fopen(buf, "r"); } else { p = strrchr(path.p, DIRSEP); if (p == NULL) p = path.p; snprintf(buf, sizeof(buf), "%.*s%c%s", (int) (p - path.p), path.p, DIRSEP, passwords_file); fp = mg_fopen(buf, "r"); } if (fp != NULL) { authorized = mg_http_check_digest_auth(hm, domain, fp); fclose(fp); } else if (!(flags & MG_AUTH_FLAG_ALLOW_MISSING_FILE)) { authorized = 0; } } LOG(LL_DEBUG, ("%.*s %s %x %d", (int) path.len, path.p, passwords_file ? passwords_file : "", flags, authorized)); return authorized; } #else int mg_http_is_authorized(struct http_message *hm, const struct mg_str path, const char *domain, const char *passwords_file, int flags) { (void) hm; (void) path; (void) domain; (void) passwords_file; (void) flags; return 1; } #endif #if MG_ENABLE_DIRECTORY_LISTING static void mg_escape(const char *src, char *dst, size_t dst_len) { size_t n = 0; while (*src != '\0' && n + 5 < dst_len) { unsigned char ch = *(unsigned char *) src++; if (ch == '<') { n += snprintf(dst + n, dst_len - n, "%s", "<"); } else { dst[n++] = ch; } } dst[n] = '\0'; } static void mg_print_dir_entry(struct mg_connection *nc, const char *file_name, cs_stat_t *stp) { char size[64], mod[64], path[MG_MAX_PATH]; int64_t fsize = stp->st_size; int is_dir = S_ISDIR(stp->st_mode); const char *slash = is_dir ? "/" : ""; struct mg_str href; if (is_dir) { snprintf(size, sizeof(size), "%s", "[DIRECTORY]"); } else { /* * We use (double) cast below because MSVC 6 compiler cannot * convert unsigned __int64 to double. */ if (fsize < 1024) { snprintf(size, sizeof(size), "%d", (int) fsize); } else if (fsize < 0x100000) { snprintf(size, sizeof(size), "%.1fk", (double) fsize / 1024.0); } else if (fsize < 0x40000000) { snprintf(size, sizeof(size), "%.1fM", (double) fsize / 1048576); } else { snprintf(size, sizeof(size), "%.1fG", (double) fsize / 1073741824); } } strftime(mod, sizeof(mod), "%d-%b-%Y %H:%M", localtime(&stp->st_mtime)); mg_escape(file_name, path, sizeof(path)); href = mg_url_encode(mg_mk_str(file_name)); mg_printf_http_chunk(nc, "%s%s" "%s%s\n", href.p, slash, path, slash, mod, is_dir ? -1 : fsize, size); free((void *) href.p); } static void mg_scan_directory(struct mg_connection *nc, const char *dir, const struct mg_serve_http_opts *opts, void (*func)(struct mg_connection *, const char *, cs_stat_t *)) { char path[MG_MAX_PATH]; cs_stat_t st; struct dirent *dp; DIR *dirp; LOG(LL_DEBUG, ("%p [%s]", nc, dir)); if ((dirp = (opendir(dir))) != NULL) { while ((dp = readdir(dirp)) != NULL) { /* Do not show current dir and hidden files */ if (mg_is_file_hidden((const char *) dp->d_name, opts, 1)) { continue; } snprintf(path, sizeof(path), "%s/%s", dir, dp->d_name); if (mg_stat(path, &st) == 0) { func(nc, (const char *) dp->d_name, &st); } } closedir(dirp); } else { LOG(LL_DEBUG, ("%p opendir(%s) -> %d", nc, dir, mg_get_errno())); } } static void mg_send_directory_listing(struct mg_connection *nc, const char *dir, struct http_message *hm, struct mg_serve_http_opts *opts) { static const char *sort_js_code = ""; mg_send_response_line(nc, 200, opts->extra_headers); mg_printf(nc, "%s: %s\r\n%s: %s\r\n\r\n", "Transfer-Encoding", "chunked", "Content-Type", "text/html; charset=utf-8"); mg_printf_http_chunk( nc, "Index of %.*s%s%s" "\n" "

Index of %.*s

\n" "" "\n" "\n" "", (int) hm->uri.len, hm->uri.p, sort_js_code, sort_js_code2, (int) hm->uri.len, hm->uri.p); mg_scan_directory(nc, dir, opts, mg_print_dir_entry); mg_printf_http_chunk(nc, "\n" "
Name" "Modified" "Size


\n" "
%s
\n" "", mg_version_header); mg_send_http_chunk(nc, "", 0); /* TODO(rojer): Remove when cesanta/dev/issues/197 is fixed. */ nc->flags |= MG_F_SEND_AND_CLOSE; } #endif /* MG_ENABLE_DIRECTORY_LISTING */ /* * Given a directory path, find one of the files specified in the * comma-separated list of index files `list`. * First found index file wins. If an index file is found, then gets * appended to the `path`, stat-ed, and result of `stat()` passed to `stp`. * If index file is not found, then `path` and `stp` remain unchanged. */ MG_INTERNAL void mg_find_index_file(const char *path, const char *list, char **index_file, cs_stat_t *stp) { struct mg_str vec; size_t path_len = strlen(path); int found = 0; *index_file = NULL; /* Traverse index files list. For each entry, append it to the given */ /* path and see if the file exists. If it exists, break the loop */ while ((list = mg_next_comma_list_entry(list, &vec, NULL)) != NULL) { cs_stat_t st; size_t len = path_len + 1 + vec.len + 1; *index_file = (char *) MG_REALLOC(*index_file, len); if (*index_file == NULL) break; snprintf(*index_file, len, "%s%c%.*s", path, DIRSEP, (int) vec.len, vec.p); /* Does it exist? Is it a file? */ if (mg_stat(*index_file, &st) == 0 && S_ISREG(st.st_mode)) { /* Yes it does, break the loop */ *stp = st; found = 1; break; } } if (!found) { MG_FREE(*index_file); *index_file = NULL; } LOG(LL_DEBUG, ("[%s] [%s]", path, (*index_file ? *index_file : ""))); } #if MG_ENABLE_HTTP_URL_REWRITES static int mg_http_send_port_based_redirect( struct mg_connection *c, struct http_message *hm, const struct mg_serve_http_opts *opts) { const char *rewrites = opts->url_rewrites; struct mg_str a, b; char local_port[20] = {'%'}; mg_conn_addr_to_str(c, local_port + 1, sizeof(local_port) - 1, MG_SOCK_STRINGIFY_PORT); while ((rewrites = mg_next_comma_list_entry(rewrites, &a, &b)) != NULL) { if (mg_vcmp(&a, local_port) == 0) { mg_send_response_line(c, 301, NULL); mg_printf(c, "Content-Length: 0\r\nLocation: %.*s%.*s\r\n\r\n", (int) b.len, b.p, (int) (hm->proto.p - hm->uri.p - 1), hm->uri.p); return 1; } } return 0; } static void mg_reverse_proxy_handler(struct mg_connection *nc, int ev, void *ev_data MG_UD_ARG(void *user_data)) { struct http_message *hm = (struct http_message *) ev_data; struct mg_http_proto_data *pd = mg_http_get_proto_data(nc); if (pd == NULL || pd->reverse_proxy_data.linked_conn == NULL) { DBG(("%p: upstream closed", nc)); return; } switch (ev) { case MG_EV_CONNECT: if (*(int *) ev_data != 0) { mg_http_send_error(pd->reverse_proxy_data.linked_conn, 502, NULL); } break; /* TODO(mkm): handle streaming */ case MG_EV_HTTP_REPLY: mg_send(pd->reverse_proxy_data.linked_conn, hm->message.p, hm->message.len); pd->reverse_proxy_data.linked_conn->flags |= MG_F_SEND_AND_CLOSE; nc->flags |= MG_F_CLOSE_IMMEDIATELY; break; case MG_EV_CLOSE: pd->reverse_proxy_data.linked_conn->flags |= MG_F_SEND_AND_CLOSE; break; } #if MG_ENABLE_CALLBACK_USERDATA (void) user_data; #endif } void mg_http_reverse_proxy(struct mg_connection *nc, const struct http_message *hm, struct mg_str mount, struct mg_str upstream) { struct mg_connection *be; char burl[256], *purl = burl; int i; const char *error; struct mg_connect_opts opts; struct mg_str path = MG_NULL_STR, user_info = MG_NULL_STR, host = MG_NULL_STR; memset(&opts, 0, sizeof(opts)); opts.error_string = &error; mg_asprintf(&purl, sizeof(burl), "%.*s%.*s", (int) upstream.len, upstream.p, (int) (hm->uri.len - mount.len), hm->uri.p + mount.len); be = mg_connect_http_base(nc->mgr, MG_CB(mg_reverse_proxy_handler, NULL), opts, "http", NULL, "https", NULL, purl, &path, &user_info, &host); LOG(LL_DEBUG, ("Proxying %.*s to %s (rule: %.*s)", (int) hm->uri.len, hm->uri.p, purl, (int) mount.len, mount.p)); if (be == NULL) { LOG(LL_ERROR, ("Error connecting to %s: %s", purl, error)); mg_http_send_error(nc, 502, NULL); goto cleanup; } /* link connections to each other, they must live and die together */ mg_http_get_proto_data(be)->reverse_proxy_data.linked_conn = nc; mg_http_get_proto_data(nc)->reverse_proxy_data.linked_conn = be; /* send request upstream */ mg_printf(be, "%.*s %.*s HTTP/1.1\r\n", (int) hm->method.len, hm->method.p, (int) path.len, path.p); mg_printf(be, "Host: %.*s\r\n", (int) host.len, host.p); for (i = 0; i < MG_MAX_HTTP_HEADERS && hm->header_names[i].len > 0; i++) { struct mg_str hn = hm->header_names[i]; struct mg_str hv = hm->header_values[i]; /* we rewrite the host header */ if (mg_vcasecmp(&hn, "Host") == 0) continue; /* * Don't pass chunked transfer encoding to the client because hm->body is * already dechunked when we arrive here. */ if (mg_vcasecmp(&hn, "Transfer-encoding") == 0 && mg_vcasecmp(&hv, "chunked") == 0) { mg_printf(be, "Content-Length: %" SIZE_T_FMT "\r\n", hm->body.len); continue; } /* We don't support proxying Expect: 100-continue. */ if (mg_vcasecmp(&hn, "Expect") == 0 && mg_vcasecmp(&hv, "100-continue") == 0) { continue; } mg_printf(be, "%.*s: %.*s\r\n", (int) hn.len, hn.p, (int) hv.len, hv.p); } mg_send(be, "\r\n", 2); mg_send(be, hm->body.p, hm->body.len); cleanup: if (purl != burl) MG_FREE(purl); } static int mg_http_handle_forwarding(struct mg_connection *nc, struct http_message *hm, const struct mg_serve_http_opts *opts) { const char *rewrites = opts->url_rewrites; struct mg_str a, b; struct mg_str p1 = MG_MK_STR("http://"), p2 = MG_MK_STR("https://"); while ((rewrites = mg_next_comma_list_entry(rewrites, &a, &b)) != NULL) { if (mg_strncmp(a, hm->uri, a.len) == 0) { if (mg_strncmp(b, p1, p1.len) == 0 || mg_strncmp(b, p2, p2.len) == 0) { mg_http_reverse_proxy(nc, hm, a, b); return 1; } } } return 0; } #endif /* MG_ENABLE_FILESYSTEM */ MG_INTERNAL int mg_uri_to_local_path(struct http_message *hm, const struct mg_serve_http_opts *opts, char **local_path, struct mg_str *remainder) { int ok = 1; const char *cp = hm->uri.p, *cp_end = hm->uri.p + hm->uri.len; struct mg_str root = {NULL, 0}; const char *file_uri_start = cp; *local_path = NULL; remainder->p = NULL; remainder->len = 0; { /* 1. Determine which root to use. */ #if MG_ENABLE_HTTP_URL_REWRITES const char *rewrites = opts->url_rewrites; #else const char *rewrites = ""; #endif struct mg_str *hh = mg_get_http_header(hm, "Host"); struct mg_str a, b; /* Check rewrites first. */ while ((rewrites = mg_next_comma_list_entry(rewrites, &a, &b)) != NULL) { if (a.len > 1 && a.p[0] == '@') { /* Host rewrite. */ if (hh != NULL && hh->len == a.len - 1 && mg_ncasecmp(a.p + 1, hh->p, a.len - 1) == 0) { root = b; break; } } else { /* Regular rewrite, URI=directory */ size_t match_len = mg_match_prefix_n(a, hm->uri); if (match_len > 0) { file_uri_start = hm->uri.p + match_len; if (*file_uri_start == '/' || file_uri_start == cp_end) { /* Match ended at component boundary, ok. */ } else if (*(file_uri_start - 1) == '/') { /* Pattern ends with '/', backtrack. */ file_uri_start--; } else { /* No match: must fall on the component boundary. */ continue; } root = b; break; } } } /* If no rewrite rules matched, use DAV or regular document root. */ if (root.p == NULL) { #if MG_ENABLE_HTTP_WEBDAV if (opts->dav_document_root != NULL && mg_is_dav_request(&hm->method)) { root.p = opts->dav_document_root; root.len = strlen(opts->dav_document_root); } else #endif { root.p = opts->document_root; root.len = strlen(opts->document_root); } } assert(root.p != NULL && root.len > 0); } { /* 2. Find where in the canonical URI path the local path ends. */ const char *u = file_uri_start + 1; char *lp = (char *) MG_MALLOC(root.len + hm->uri.len + 1); char *lp_end = lp + root.len + hm->uri.len + 1; char *p = lp, *ps; int exists = 1; if (lp == NULL) { ok = 0; goto out; } memcpy(p, root.p, root.len); p += root.len; if (*(p - 1) == DIRSEP) p--; *p = '\0'; ps = p; /* Chop off URI path components one by one and build local path. */ while (u <= cp_end) { const char *next = u; struct mg_str component; if (exists) { cs_stat_t st; exists = (mg_stat(lp, &st) == 0); if (exists && S_ISREG(st.st_mode)) { /* We found the terminal, the rest of the URI (if any) is path_info. */ if (*(u - 1) == '/') u--; break; } } if (u >= cp_end) break; parse_uri_component((const char **) &next, cp_end, "/", &component); if (component.len > 0) { int len; memmove(p + 1, component.p, component.len); len = mg_url_decode(p + 1, component.len, p + 1, lp_end - p - 1, 0); if (len <= 0) { ok = 0; break; } component.p = p + 1; component.len = len; if (mg_vcmp(&component, ".") == 0) { /* Yum. */ } else if (mg_vcmp(&component, "..") == 0) { while (p > ps && *p != DIRSEP) p--; *p = '\0'; } else { size_t i; #ifdef _WIN32 /* On Windows, make sure it's valid Unicode (no funny stuff). */ wchar_t buf[MG_MAX_PATH * 2]; if (to_wchar(component.p, buf, MG_MAX_PATH) == 0) { DBG(("[%.*s] smells funny", (int) component.len, component.p)); ok = 0; break; } #endif *p++ = DIRSEP; /* No NULs and DIRSEPs in the component (percent-encoded). */ for (i = 0; i < component.len; i++, p++) { if (*p == '\0' || *p == DIRSEP #ifdef _WIN32 /* On Windows, "/" is also accepted, so check for that too. */ || *p == '/' #endif ) { ok = 0; break; } } } } u = next; } if (ok) { *local_path = lp; if (u > cp_end) u = cp_end; remainder->p = u; remainder->len = cp_end - u; } else { MG_FREE(lp); } } out: LOG(LL_DEBUG, ("'%.*s' -> '%s' + '%.*s'", (int) hm->uri.len, hm->uri.p, *local_path ? *local_path : "", (int) remainder->len, remainder->p)); return ok; } static int mg_get_month_index(const char *s) { static const char *month_names[] = {"Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"}; size_t i; for (i = 0; i < ARRAY_SIZE(month_names); i++) if (!strcmp(s, month_names[i])) return (int) i; return -1; } static int mg_num_leap_years(int year) { return year / 4 - year / 100 + year / 400; } /* Parse UTC date-time string, and return the corresponding time_t value. */ MG_INTERNAL time_t mg_parse_date_string(const char *datetime) { static const unsigned short days_before_month[] = { 0, 31, 59, 90, 120, 151, 181, 212, 243, 273, 304, 334}; char month_str[32]; int second, minute, hour, day, month, year, leap_days, days; time_t result = (time_t) 0; if (((sscanf(datetime, "%d/%3s/%d %d:%d:%d", &day, month_str, &year, &hour, &minute, &second) == 6) || (sscanf(datetime, "%d %3s %d %d:%d:%d", &day, month_str, &year, &hour, &minute, &second) == 6) || (sscanf(datetime, "%*3s, %d %3s %d %d:%d:%d", &day, month_str, &year, &hour, &minute, &second) == 6) || (sscanf(datetime, "%d-%3s-%d %d:%d:%d", &day, month_str, &year, &hour, &minute, &second) == 6)) && year > 1970 && (month = mg_get_month_index(month_str)) != -1) { leap_days = mg_num_leap_years(year) - mg_num_leap_years(1970); year -= 1970; days = year * 365 + days_before_month[month] + (day - 1) + leap_days; result = days * 24 * 3600 + hour * 3600 + minute * 60 + second; } return result; } MG_INTERNAL int mg_is_not_modified(struct http_message *hm, cs_stat_t *st) { struct mg_str *hdr; if ((hdr = mg_get_http_header(hm, "If-None-Match")) != NULL) { char etag[64]; mg_http_construct_etag(etag, sizeof(etag), st); return mg_vcasecmp(hdr, etag) == 0; } else if ((hdr = mg_get_http_header(hm, "If-Modified-Since")) != NULL) { return st->st_mtime <= mg_parse_date_string(hdr->p); } else { return 0; } } void mg_http_send_digest_auth_request(struct mg_connection *c, const char *domain) { mg_printf(c, "HTTP/1.1 401 Unauthorized\r\n" "WWW-Authenticate: Digest qop=\"auth\", " "realm=\"%s\", nonce=\"%lx\"\r\n" "Content-Length: 0\r\n\r\n", domain, (unsigned long) mg_time()); } static void mg_http_send_options(struct mg_connection *nc) { mg_printf(nc, "%s", "HTTP/1.1 200 OK\r\nAllow: GET, POST, HEAD, CONNECT, OPTIONS" #if MG_ENABLE_HTTP_WEBDAV ", MKCOL, PUT, DELETE, PROPFIND, MOVE\r\nDAV: 1,2" #endif "\r\n\r\n"); nc->flags |= MG_F_SEND_AND_CLOSE; } static int mg_is_creation_request(const struct http_message *hm) { return mg_vcmp(&hm->method, "MKCOL") == 0 || mg_vcmp(&hm->method, "PUT") == 0; } MG_INTERNAL void mg_send_http_file(struct mg_connection *nc, char *path, const struct mg_str *path_info, struct http_message *hm, struct mg_serve_http_opts *opts) { int exists, is_directory, is_cgi; #if MG_ENABLE_HTTP_WEBDAV int is_dav = mg_is_dav_request(&hm->method); #else int is_dav = 0; #endif char *index_file = NULL; cs_stat_t st; exists = (mg_stat(path, &st) == 0); is_directory = exists && S_ISDIR(st.st_mode); if (is_directory) mg_find_index_file(path, opts->index_files, &index_file, &st); is_cgi = (mg_match_prefix(opts->cgi_file_pattern, strlen(opts->cgi_file_pattern), index_file ? index_file : path) > 0); LOG(LL_DEBUG, ("%p %.*s [%s] exists=%d is_dir=%d is_dav=%d is_cgi=%d index=%s", nc, (int) hm->method.len, hm->method.p, path, exists, is_directory, is_dav, is_cgi, index_file ? index_file : "")); if (is_directory && hm->uri.p[hm->uri.len - 1] != '/' && !is_dav) { mg_printf(nc, "HTTP/1.1 301 Moved\r\nLocation: %.*s/\r\n" "Content-Length: 0\r\n\r\n", (int) hm->uri.len, hm->uri.p); MG_FREE(index_file); return; } /* If we have path_info, the only way to handle it is CGI. */ if (path_info->len > 0 && !is_cgi) { mg_http_send_error(nc, 501, NULL); MG_FREE(index_file); return; } if (is_dav && opts->dav_document_root == NULL) { mg_http_send_error(nc, 501, NULL); } else if (!mg_http_is_authorized( hm, mg_mk_str(path), opts->auth_domain, opts->global_auth_file, ((is_directory ? MG_AUTH_FLAG_IS_DIRECTORY : 0) | MG_AUTH_FLAG_IS_GLOBAL_PASS_FILE | MG_AUTH_FLAG_ALLOW_MISSING_FILE)) || !mg_http_is_authorized( hm, mg_mk_str(path), opts->auth_domain, opts->per_directory_auth_file, ((is_directory ? MG_AUTH_FLAG_IS_DIRECTORY : 0) | MG_AUTH_FLAG_ALLOW_MISSING_FILE))) { mg_http_send_digest_auth_request(nc, opts->auth_domain); } else if (is_cgi) { #if MG_ENABLE_HTTP_CGI mg_handle_cgi(nc, index_file ? index_file : path, path_info, hm, opts); #else mg_http_send_error(nc, 501, NULL); #endif /* MG_ENABLE_HTTP_CGI */ } else if ((!exists || mg_is_file_hidden(path, opts, 0 /* specials are ok */)) && !mg_is_creation_request(hm)) { mg_http_send_error(nc, 404, NULL); #if MG_ENABLE_HTTP_WEBDAV } else if (!mg_vcmp(&hm->method, "PROPFIND")) { mg_handle_propfind(nc, path, &st, hm, opts); #if !MG_DISABLE_DAV_AUTH } else if (is_dav && (opts->dav_auth_file == NULL || (strcmp(opts->dav_auth_file, "-") != 0 && !mg_http_is_authorized( hm, mg_mk_str(path), opts->auth_domain, opts->dav_auth_file, ((is_directory ? MG_AUTH_FLAG_IS_DIRECTORY : 0) | MG_AUTH_FLAG_IS_GLOBAL_PASS_FILE | MG_AUTH_FLAG_ALLOW_MISSING_FILE))))) { mg_http_send_digest_auth_request(nc, opts->auth_domain); #endif } else if (!mg_vcmp(&hm->method, "MKCOL")) { mg_handle_mkcol(nc, path, hm); } else if (!mg_vcmp(&hm->method, "DELETE")) { mg_handle_delete(nc, opts, path); } else if (!mg_vcmp(&hm->method, "PUT")) { mg_handle_put(nc, path, hm); } else if (!mg_vcmp(&hm->method, "MOVE")) { mg_handle_move(nc, opts, path, hm); #if MG_ENABLE_FAKE_DAVLOCK } else if (!mg_vcmp(&hm->method, "LOCK")) { mg_handle_lock(nc, path); #endif #endif /* MG_ENABLE_HTTP_WEBDAV */ } else if (!mg_vcmp(&hm->method, "OPTIONS")) { mg_http_send_options(nc); } else if (is_directory && index_file == NULL) { #if MG_ENABLE_DIRECTORY_LISTING if (strcmp(opts->enable_directory_listing, "yes") == 0) { mg_send_directory_listing(nc, path, hm, opts); } else { mg_http_send_error(nc, 403, NULL); } #else mg_http_send_error(nc, 501, NULL); #endif } else if (mg_is_not_modified(hm, &st)) { mg_http_send_error(nc, 304, "Not Modified"); } else { mg_http_serve_file2(nc, index_file ? index_file : path, hm, opts); } MG_FREE(index_file); } void mg_serve_http(struct mg_connection *nc, struct http_message *hm, struct mg_serve_http_opts opts) { char *path = NULL; struct mg_str *hdr, path_info; uint32_t remote_ip = ntohl(*(uint32_t *) &nc->sa.sin.sin_addr); if (mg_check_ip_acl(opts.ip_acl, remote_ip) != 1) { /* Not allowed to connect */ mg_http_send_error(nc, 403, NULL); nc->flags |= MG_F_SEND_AND_CLOSE; return; } #if MG_ENABLE_HTTP_URL_REWRITES if (mg_http_handle_forwarding(nc, hm, &opts)) { return; } if (mg_http_send_port_based_redirect(nc, hm, &opts)) { return; } #endif if (opts.document_root == NULL) { opts.document_root = "."; } if (opts.per_directory_auth_file == NULL) { opts.per_directory_auth_file = ".htpasswd"; } if (opts.enable_directory_listing == NULL) { opts.enable_directory_listing = "yes"; } if (opts.cgi_file_pattern == NULL) { opts.cgi_file_pattern = "**.cgi$|**.php$"; } if (opts.ssi_pattern == NULL) { opts.ssi_pattern = "**.shtml$|**.shtm$"; } if (opts.index_files == NULL) { opts.index_files = "index.html,index.htm,index.shtml,index.cgi,index.php"; } /* Normalize path - resolve "." and ".." (in-place). */ if (!mg_normalize_uri_path(&hm->uri, &hm->uri)) { mg_http_send_error(nc, 400, NULL); return; } if (mg_uri_to_local_path(hm, &opts, &path, &path_info) == 0) { mg_http_send_error(nc, 404, NULL); return; } mg_send_http_file(nc, path, &path_info, hm, &opts); MG_FREE(path); path = NULL; /* Close connection for non-keep-alive requests */ if (mg_vcmp(&hm->proto, "HTTP/1.1") != 0 || ((hdr = mg_get_http_header(hm, "Connection")) != NULL && mg_vcmp(hdr, "keep-alive") != 0)) { #if 0 nc->flags |= MG_F_SEND_AND_CLOSE; #endif } } #if MG_ENABLE_HTTP_STREAMING_MULTIPART void mg_file_upload_handler(struct mg_connection *nc, int ev, void *ev_data, mg_fu_fname_fn local_name_fn MG_UD_ARG(void *user_data)) { switch (ev) { case MG_EV_HTTP_PART_BEGIN: { struct mg_http_multipart_part *mp = (struct mg_http_multipart_part *) ev_data; struct file_upload_state *fus = (struct file_upload_state *) MG_CALLOC(1, sizeof(*fus)); struct mg_str lfn = local_name_fn(nc, mg_mk_str(mp->file_name)); mp->user_data = NULL; if (lfn.p == NULL || lfn.len == 0) { LOG(LL_ERROR, ("%p Not allowed to upload %s", nc, mp->file_name)); mg_printf(nc, "HTTP/1.1 403 Not Allowed\r\n" "Content-Type: text/plain\r\n" "Connection: close\r\n\r\n" "Not allowed to upload %s\r\n", mp->file_name); nc->flags |= MG_F_SEND_AND_CLOSE; return; } fus->lfn = (char *) MG_MALLOC(lfn.len + 1); memcpy(fus->lfn, lfn.p, lfn.len); fus->lfn[lfn.len] = '\0'; if (lfn.p != mp->file_name) MG_FREE((char *) lfn.p); LOG(LL_DEBUG, ("%p Receiving file %s -> %s", nc, mp->file_name, fus->lfn)); fus->fp = mg_fopen(fus->lfn, "w"); if (fus->fp == NULL) { mg_printf(nc, "HTTP/1.1 500 Internal Server Error\r\n" "Content-Type: text/plain\r\n" "Connection: close\r\n\r\n"); LOG(LL_ERROR, ("Failed to open %s: %d\n", fus->lfn, mg_get_errno())); mg_printf(nc, "Failed to open %s: %d\n", fus->lfn, mg_get_errno()); /* Do not close the connection just yet, discard remainder of the data. * This is because at the time of writing some browsers (Chrome) fail to * render response before all the data is sent. */ } mp->user_data = (void *) fus; break; } case MG_EV_HTTP_PART_DATA: { struct mg_http_multipart_part *mp = (struct mg_http_multipart_part *) ev_data; struct file_upload_state *fus = (struct file_upload_state *) mp->user_data; if (fus == NULL || fus->fp == NULL) break; if (mg_fwrite(mp->data.p, 1, mp->data.len, fus->fp) != mp->data.len) { LOG(LL_ERROR, ("Failed to write to %s: %d, wrote %d", fus->lfn, mg_get_errno(), (int) fus->num_recd)); if (mg_get_errno() == ENOSPC #ifdef SPIFFS_ERR_FULL || mg_get_errno() == SPIFFS_ERR_FULL #endif ) { mg_printf(nc, "HTTP/1.1 413 Payload Too Large\r\n" "Content-Type: text/plain\r\n" "Connection: close\r\n\r\n"); mg_printf(nc, "Failed to write to %s: no space left; wrote %d\r\n", fus->lfn, (int) fus->num_recd); } else { mg_printf(nc, "HTTP/1.1 500 Internal Server Error\r\n" "Content-Type: text/plain\r\n" "Connection: close\r\n\r\n"); mg_printf(nc, "Failed to write to %s: %d, wrote %d", mp->file_name, mg_get_errno(), (int) fus->num_recd); } fclose(fus->fp); remove(fus->lfn); fus->fp = NULL; /* Do not close the connection just yet, discard remainder of the data. * This is because at the time of writing some browsers (Chrome) fail to * render response before all the data is sent. */ return; } fus->num_recd += mp->data.len; LOG(LL_DEBUG, ("%p rec'd %d bytes, %d total", nc, (int) mp->data.len, (int) fus->num_recd)); break; } case MG_EV_HTTP_PART_END: { struct mg_http_multipart_part *mp = (struct mg_http_multipart_part *) ev_data; struct file_upload_state *fus = (struct file_upload_state *) mp->user_data; if (fus == NULL) break; if (mp->status >= 0 && fus->fp != NULL) { LOG(LL_DEBUG, ("%p Uploaded %s (%s), %d bytes", nc, mp->file_name, fus->lfn, (int) fus->num_recd)); mg_printf(nc, "HTTP/1.1 200 OK\r\n" "Content-Type: text/plain\r\n" "Connection: close\r\n\r\n" "Ok, %s - %d bytes.\r\n", mp->file_name, (int) fus->num_recd); } else { LOG(LL_ERROR, ("Failed to store %s (%s)", mp->file_name, fus->lfn)); /* * mp->status < 0 means connection was terminated, so no reason to send * HTTP reply */ } if (fus->fp != NULL) fclose(fus->fp); MG_FREE(fus->lfn); MG_FREE(fus); mp->user_data = NULL; nc->flags |= MG_F_SEND_AND_CLOSE; break; } } #if MG_ENABLE_CALLBACK_USERDATA (void) user_data; #endif } #endif /* MG_ENABLE_HTTP_STREAMING_MULTIPART */ #endif /* MG_ENABLE_FILESYSTEM */ struct mg_connection *mg_connect_http_base( struct mg_mgr *mgr, MG_CB(mg_event_handler_t ev_handler, void *user_data), struct mg_connect_opts opts, const char *scheme1, const char *scheme2, const char *scheme_ssl1, const char *scheme_ssl2, const char *url, struct mg_str *path, struct mg_str *user_info, struct mg_str *host) { struct mg_connection *nc = NULL; unsigned int port_i = 0; int use_ssl = 0; struct mg_str scheme, query, fragment; char conn_addr_buf[2]; char *conn_addr = conn_addr_buf; if (mg_parse_uri(mg_mk_str(url), &scheme, user_info, host, &port_i, path, &query, &fragment) != 0) { MG_SET_PTRPTR(opts.error_string, "cannot parse url"); goto out; } /* If query is present, do not strip it. Pass to the caller. */ if (query.len > 0) path->len += query.len + 1; if (scheme.len == 0 || mg_vcmp(&scheme, scheme1) == 0 || (scheme2 != NULL && mg_vcmp(&scheme, scheme2) == 0)) { use_ssl = 0; if (port_i == 0) port_i = 80; } else if (mg_vcmp(&scheme, scheme_ssl1) == 0 || (scheme2 != NULL && mg_vcmp(&scheme, scheme_ssl2) == 0)) { use_ssl = 1; if (port_i == 0) port_i = 443; } else { goto out; } mg_asprintf(&conn_addr, sizeof(conn_addr_buf), "tcp://%.*s:%u", (int) host->len, host->p, port_i); if (conn_addr == NULL) goto out; LOG(LL_DEBUG, ("%s use_ssl? %d %s", url, use_ssl, conn_addr)); if (use_ssl) { #if MG_ENABLE_SSL /* * Schema requires SSL, but no SSL parameters were provided in opts. * In order to maintain backward compatibility, use a faux-SSL with no * verification. */ if (opts.ssl_ca_cert == NULL) { opts.ssl_ca_cert = "*"; } #else MG_SET_PTRPTR(opts.error_string, "ssl is disabled"); goto out; #endif } if ((nc = mg_connect_opt(mgr, conn_addr, MG_CB(ev_handler, user_data), opts)) != NULL) { mg_set_protocol_http_websocket(nc); } out: if (conn_addr != NULL && conn_addr != conn_addr_buf) MG_FREE(conn_addr); return nc; } struct mg_connection *mg_connect_http_opt( struct mg_mgr *mgr, MG_CB(mg_event_handler_t ev_handler, void *user_data), struct mg_connect_opts opts, const char *url, const char *extra_headers, const char *post_data) { struct mg_str user = MG_NULL_STR, null_str = MG_NULL_STR; struct mg_str host = MG_NULL_STR, path = MG_NULL_STR; struct mbuf auth; struct mg_connection *nc = mg_connect_http_base(mgr, MG_CB(ev_handler, user_data), opts, "http", NULL, "https", NULL, url, &path, &user, &host); if (nc == NULL) { return NULL; } mbuf_init(&auth, 0); if (user.len > 0) { mg_basic_auth_header(user, null_str, &auth); } if (post_data == NULL) post_data = ""; if (extra_headers == NULL) extra_headers = ""; if (path.len == 0) path = mg_mk_str("/"); if (host.len == 0) host = mg_mk_str(""); // mg_printf(nc, "%s %.*s HTTP/1.1\r\nHost: %.*s\r\nContent-Length: %" SIZE_T_FMT // "\r\n%.*s%s\r\n%s", mg_printf(nc, "%s %.*s HTTP/1.1\r\nHost: %.*s\r\nContent-Length: %d\r\n%.*s%s\r\n%s", (post_data[0] == '\0' ? "GET" : "POST"), (int) path.len, path.p, (int) (path.p - host.p), host.p, strlen(post_data), (int) auth.len, (auth.buf == NULL ? "" : auth.buf), extra_headers, post_data); mbuf_free(&auth); return nc; } ///*WYF add*************************************/ struct mg_connection *mg_connect_http_optDEL( struct mg_mgr *mgr, MG_CB(mg_event_handler_t ev_handler, void *user_data), struct mg_connect_opts opts, const char *url, const char *extra_headers, const char *post_data) { struct mg_str user = MG_NULL_STR, null_str = MG_NULL_STR; struct mg_str host = MG_NULL_STR, path = MG_NULL_STR; struct mbuf auth; struct mg_connection *nc = mg_connect_http_base(mgr, MG_CB(ev_handler, user_data), opts, "http", NULL, "https", NULL, url, &path, &user, &host); if (nc == NULL) { return NULL; } mbuf_init(&auth, 0); if (user.len > 0) { mg_basic_auth_header(user, null_str, &auth); } if (post_data == NULL) post_data = ""; if (extra_headers == NULL) extra_headers = ""; if (path.len == 0) path = mg_mk_str("/"); if (host.len == 0) host = mg_mk_str(""); // mg_printf(nc, "%s %.*s HTTP/1.1\r\nHost: %.*s\r\nContent-Length: %" SIZE_T_FMT // "\r\n%.*s%s\r\n%s", // mg_printf(nc, "%s %.*s HTTP/1.1\r\nHost: %.*s\r\nContent-Length: %" SIZE_T_FMT // "\r\n%.*s%s\r\n%s", mg_printf(nc, "%s %.*s HTTP/1.1\r\nHost: %.*s\r\nContent-Length: %d\r\n%.*s%s\r\n%s", ("DELETE"), (int) path.len, path.p, (int) (path.p - host.p), host.p, strlen(post_data), (int) auth.len, (auth.buf == NULL ? "" : auth.buf), extra_headers, post_data); mbuf_free(&auth); return nc; } struct mg_connection *mg_connect_httpDEL( struct mg_mgr *mgr, MG_CB(mg_event_handler_t ev_handler, void *user_data), const char *url, const char *extra_headers, const char *post_data) { struct mg_connect_opts opts; memset(&opts, 0, sizeof(opts)); return mg_connect_http_optDEL(mgr, MG_CB(ev_handler, user_data), opts, url, extra_headers, post_data); } //**WYF add*************************************/ struct mg_connection *mg_connect_http( struct mg_mgr *mgr, MG_CB(mg_event_handler_t ev_handler, void *user_data), const char *url, const char *extra_headers, const char *post_data) { struct mg_connect_opts opts; memset(&opts, 0, sizeof(opts)); return mg_connect_http_opt(mgr, MG_CB(ev_handler, user_data), opts, url, extra_headers, post_data); } size_t mg_parse_multipart(const char *buf, size_t buf_len, char *var_name, size_t var_name_len, char *file_name, size_t file_name_len, const char **data, size_t *data_len) { static const char cd[] = "Content-Disposition: "; size_t hl, bl, n, ll, pos, cdl = sizeof(cd) - 1; int shl; if (buf == NULL || buf_len <= 0) return 0; if ((shl = mg_http_get_request_len(buf, buf_len)) <= 0) return 0; hl = shl; if (buf[0] != '-' || buf[1] != '-' || buf[2] == '\n') return 0; /* Get boundary length */ bl = mg_get_line_len(buf, buf_len); /* Loop through headers, fetch variable name and file name */ var_name[0] = file_name[0] = '\0'; for (n = bl; (ll = mg_get_line_len(buf + n, hl - n)) > 0; n += ll) { if (mg_ncasecmp(cd, buf + n, cdl) == 0) { struct mg_str header; header.p = buf + n + cdl; header.len = ll - (cdl + 2); { char *var_name2 = var_name; mg_http_parse_header2(&header, "name", &var_name2, var_name_len); /* TODO: handle reallocated buffer correctly */ if (var_name2 != var_name) { MG_FREE(var_name2); var_name[0] = '\0'; } } { char *file_name2 = file_name; mg_http_parse_header2(&header, "filename", &file_name2, file_name_len); /* TODO: handle reallocated buffer correctly */ if (file_name2 != file_name) { MG_FREE(file_name2); file_name[0] = '\0'; } } } } /* Scan through the body, search for terminating boundary */ for (pos = hl; pos + (bl - 2) < buf_len; pos++) { if (buf[pos] == '-' && !strncmp(buf, &buf[pos], bl - 2)) { if (data_len != NULL) *data_len = (pos - 2) - hl; if (data != NULL) *data = buf + hl; return pos; } } return 0; } void mg_register_http_endpoint_opt(struct mg_connection *nc, const char *uri_path, mg_event_handler_t handler, struct mg_http_endpoint_opts opts) { struct mg_http_proto_data *pd = NULL; struct mg_http_endpoint *new_ep = NULL; if (nc == NULL) return; new_ep = (struct mg_http_endpoint *) MG_CALLOC(1, sizeof(*new_ep)); if (new_ep == NULL) return; pd = mg_http_get_proto_data(nc); new_ep->uri_pattern = mg_strdup(mg_mk_str(uri_path)); if (opts.auth_domain != NULL && opts.auth_file != NULL) { new_ep->auth_domain = strdup(opts.auth_domain); new_ep->auth_file = strdup(opts.auth_file); } new_ep->handler = handler; #if MG_ENABLE_CALLBACK_USERDATA new_ep->user_data = opts.user_data; #endif new_ep->next = pd->endpoints; pd->endpoints = new_ep; } static void mg_http_call_endpoint_handler(struct mg_connection *nc, int ev, struct http_message *hm) { struct mg_http_proto_data *pd = mg_http_get_proto_data(nc); void *user_data = nc->user_data; if (ev == MG_EV_HTTP_REQUEST #if MG_ENABLE_HTTP_STREAMING_MULTIPART || ev == MG_EV_HTTP_MULTIPART_REQUEST #endif ) { struct mg_http_endpoint *ep = mg_http_get_endpoint_handler(nc->listener, &hm->uri); if (ep != NULL) { #if MG_ENABLE_FILESYSTEM && !MG_DISABLE_HTTP_DIGEST_AUTH if (!mg_http_is_authorized(hm, hm->uri, ep->auth_domain, ep->auth_file, MG_AUTH_FLAG_IS_GLOBAL_PASS_FILE)) { mg_http_send_digest_auth_request(nc, ep->auth_domain); return; } #endif pd->endpoint_handler = ep->handler; #if MG_ENABLE_CALLBACK_USERDATA user_data = ep->user_data; #endif } } mg_call(nc, pd->endpoint_handler ? pd->endpoint_handler : nc->handler, user_data, ev, hm); } void mg_register_http_endpoint(struct mg_connection *nc, const char *uri_path, MG_CB(mg_event_handler_t handler, void *user_data)) { struct mg_http_endpoint_opts opts; memset(&opts, 0, sizeof(opts)); #if MG_ENABLE_CALLBACK_USERDATA opts.user_data = user_data; #endif mg_register_http_endpoint_opt(nc, uri_path, handler, opts); } #endif /* MG_ENABLE_HTTP */ #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_http_cgi.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #ifndef _WIN32 #include #endif #if MG_ENABLE_HTTP && MG_ENABLE_HTTP_CGI #ifndef MG_MAX_CGI_ENVIR_VARS #define MG_MAX_CGI_ENVIR_VARS 64 #endif #ifndef MG_ENV_EXPORT_TO_CGI #define MG_ENV_EXPORT_TO_CGI "MONGOOSE_CGI" #endif #define MG_F_HTTP_CGI_PARSE_HEADERS MG_F_USER_1 /* * This structure helps to create an environment for the spawned CGI program. * Environment is an array of "VARIABLE=VALUE\0" ASCIIZ strings, * last element must be NULL. * However, on Windows there is a requirement that all these VARIABLE=VALUE\0 * strings must reside in a contiguous buffer. The end of the buffer is * marked by two '\0' characters. * We satisfy both worlds: we create an envp array (which is vars), all * entries are actually pointers inside buf. */ struct mg_cgi_env_block { struct mg_connection *nc; char buf[MG_CGI_ENVIRONMENT_SIZE]; /* Environment buffer */ const char *vars[MG_MAX_CGI_ENVIR_VARS]; /* char *envp[] */ int len; /* Space taken */ int nvars; /* Number of variables in envp[] */ }; #ifdef _WIN32 struct mg_threadparam { sock_t s; HANDLE hPipe; }; static int mg_wait_until_ready(sock_t sock, int for_read) { fd_set set; FD_ZERO(&set); FD_SET(sock, &set); return select(sock + 1, for_read ? &set : 0, for_read ? 0 : &set, 0, 0) == 1; } static void *mg_push_to_stdin(void *arg) { struct mg_threadparam *tp = (struct mg_threadparam *) arg; int n, sent, stop = 0; DWORD k; char buf[BUFSIZ]; while (!stop && mg_wait_until_ready(tp->s, 1) && (n = recv(tp->s, buf, sizeof(buf), 0)) > 0) { if (n == -1 && GetLastError() == WSAEWOULDBLOCK) continue; for (sent = 0; !stop && sent < n; sent += k) { if (!WriteFile(tp->hPipe, buf + sent, n - sent, &k, 0)) stop = 1; } } DBG(("%s", "FORWARED EVERYTHING TO CGI")); CloseHandle(tp->hPipe); MG_FREE(tp); return NULL; } static void *mg_pull_from_stdout(void *arg) { struct mg_threadparam *tp = (struct mg_threadparam *) arg; int k = 0, stop = 0; DWORD n, sent; char buf[BUFSIZ]; while (!stop && ReadFile(tp->hPipe, buf, sizeof(buf), &n, NULL)) { for (sent = 0; !stop && sent < n; sent += k) { if (mg_wait_until_ready(tp->s, 0) && (k = send(tp->s, buf + sent, n - sent, 0)) <= 0) stop = 1; } } DBG(("%s", "EOF FROM CGI")); CloseHandle(tp->hPipe); shutdown(tp->s, 2); // Without this, IO thread may get truncated data closesocket(tp->s); MG_FREE(tp); return NULL; } static void mg_spawn_stdio_thread(sock_t sock, HANDLE hPipe, void *(*func)(void *)) { struct mg_threadparam *tp = (struct mg_threadparam *) MG_MALLOC(sizeof(*tp)); if (tp != NULL) { tp->s = sock; tp->hPipe = hPipe; mg_start_thread(func, tp); } } static void mg_abs_path(const char *utf8_path, char *abs_path, size_t len) { wchar_t buf[MG_MAX_PATH], buf2[MG_MAX_PATH]; to_wchar(utf8_path, buf, ARRAY_SIZE(buf)); GetFullPathNameW(buf, ARRAY_SIZE(buf2), buf2, NULL); WideCharToMultiByte(CP_UTF8, 0, buf2, wcslen(buf2) + 1, abs_path, len, 0, 0); } static int mg_start_process(const char *interp, const char *cmd, const char *env, const char *envp[], const char *dir, sock_t sock) { STARTUPINFOW si; PROCESS_INFORMATION pi; HANDLE a[2], b[2], me = GetCurrentProcess(); wchar_t wcmd[MG_MAX_PATH], full_dir[MG_MAX_PATH]; char buf[MG_MAX_PATH], buf2[MG_MAX_PATH], buf5[MG_MAX_PATH], buf4[MG_MAX_PATH], cmdline[MG_MAX_PATH]; DWORD flags = DUPLICATE_CLOSE_SOURCE | DUPLICATE_SAME_ACCESS; FILE *fp; memset(&si, 0, sizeof(si)); memset(&pi, 0, sizeof(pi)); si.cb = sizeof(si); si.dwFlags = STARTF_USESTDHANDLES | STARTF_USESHOWWINDOW; si.wShowWindow = SW_HIDE; si.hStdError = GetStdHandle(STD_ERROR_HANDLE); CreatePipe(&a[0], &a[1], NULL, 0); CreatePipe(&b[0], &b[1], NULL, 0); DuplicateHandle(me, a[0], me, &si.hStdInput, 0, TRUE, flags); DuplicateHandle(me, b[1], me, &si.hStdOutput, 0, TRUE, flags); if (interp == NULL && (fp = mg_fopen(cmd, "r")) != NULL) { buf[0] = buf[1] = '\0'; fgets(buf, sizeof(buf), fp); buf[sizeof(buf) - 1] = '\0'; if (buf[0] == '#' && buf[1] == '!') { interp = buf + 2; /* Trim leading spaces: https://github.com/cesanta/mongoose/issues/489 */ while (*interp != '\0' && isspace(*(unsigned char *) interp)) { interp++; } } fclose(fp); } snprintf(buf, sizeof(buf), "%s/%s", dir, cmd); mg_abs_path(buf, buf2, ARRAY_SIZE(buf2)); mg_abs_path(dir, buf5, ARRAY_SIZE(buf5)); to_wchar(dir, full_dir, ARRAY_SIZE(full_dir)); if (interp != NULL) { mg_abs_path(interp, buf4, ARRAY_SIZE(buf4)); snprintf(cmdline, sizeof(cmdline), "%s \"%s\"", buf4, buf2); } else { snprintf(cmdline, sizeof(cmdline), "\"%s\"", buf2); } to_wchar(cmdline, wcmd, ARRAY_SIZE(wcmd)); if (CreateProcessW(NULL, wcmd, NULL, NULL, TRUE, CREATE_NEW_PROCESS_GROUP, (void *) env, full_dir, &si, &pi) != 0) { mg_spawn_stdio_thread(sock, a[1], mg_push_to_stdin); mg_spawn_stdio_thread(sock, b[0], mg_pull_from_stdout); CloseHandle(si.hStdOutput); CloseHandle(si.hStdInput); CloseHandle(pi.hThread); CloseHandle(pi.hProcess); } else { CloseHandle(a[1]); CloseHandle(b[0]); closesocket(sock); } DBG(("CGI command: [%ls] -> %p", wcmd, pi.hProcess)); /* Not closing a[0] and b[1] because we've used DUPLICATE_CLOSE_SOURCE */ (void) envp; return (pi.hProcess != NULL); } #else static int mg_start_process(const char *interp, const char *cmd, const char *env, const char *envp[], const char *dir, sock_t sock) { char buf[500]; pid_t pid = fork(); (void) env; if (pid == 0) { /* * In Linux `chdir` declared with `warn_unused_result` attribute * To shutup compiler we have yo use result in some way */ int tmp = chdir(dir); (void) tmp; (void) dup2(sock, 0); (void) dup2(sock, 1); closesocket(sock); /* * After exec, all signal handlers are restored to their default values, * with one exception of SIGCHLD. According to POSIX.1-2001 and Linux's * implementation, SIGCHLD's handler will leave unchanged after exec * if it was set to be ignored. Restore it to default action. */ signal(SIGCHLD, SIG_DFL); if (interp == NULL) { execle(cmd, cmd, (char *) 0, envp); /* (char *) 0 to squash warning */ } else { execle(interp, interp, cmd, (char *) 0, envp); } snprintf(buf, sizeof(buf), "Status: 500\r\n\r\n" "500 Server Error: %s%s%s: %s", interp == NULL ? "" : interp, interp == NULL ? "" : " ", cmd, strerror(errno)); send(1, buf, strlen(buf), 0); _exit(EXIT_FAILURE); /* exec call failed */ } return (pid != 0); } #endif /* _WIN32 */ /* * Append VARIABLE=VALUE\0 string to the buffer, and add a respective * pointer into the vars array. */ static char *mg_addenv(struct mg_cgi_env_block *block, const char *fmt, ...) { int n, space; char *added = block->buf + block->len; va_list ap; /* Calculate how much space is left in the buffer */ space = sizeof(block->buf) - (block->len + 2); if (space > 0) { /* Copy VARIABLE=VALUE\0 string into the free space */ va_start(ap, fmt); n = vsnprintf(added, (size_t) space, fmt, ap); va_end(ap); /* Make sure we do not overflow buffer and the envp array */ if (n > 0 && n + 1 < space && block->nvars < (int) ARRAY_SIZE(block->vars) - 2) { /* Append a pointer to the added string into the envp array */ block->vars[block->nvars++] = added; /* Bump up used length counter. Include \0 terminator */ block->len += n + 1; } } return added; } static void mg_addenv2(struct mg_cgi_env_block *blk, const char *name) { const char *s; if ((s = getenv(name)) != NULL) mg_addenv(blk, "%s=%s", name, s); } static void mg_prepare_cgi_environment(struct mg_connection *nc, const char *prog, const struct mg_str *path_info, const struct http_message *hm, const struct mg_serve_http_opts *opts, struct mg_cgi_env_block *blk) { const char *s; struct mg_str *h; char *p; size_t i; char buf[100]; size_t path_info_len = path_info != NULL ? path_info->len : 0; blk->len = blk->nvars = 0; blk->nc = nc; if ((s = getenv("SERVER_NAME")) != NULL) { mg_addenv(blk, "SERVER_NAME=%s", s); } else { mg_sock_to_str(nc->sock, buf, sizeof(buf), 3); mg_addenv(blk, "SERVER_NAME=%s", buf); } mg_addenv(blk, "SERVER_ROOT=%s", opts->document_root); mg_addenv(blk, "DOCUMENT_ROOT=%s", opts->document_root); mg_addenv(blk, "SERVER_SOFTWARE=%s/%s", "Mongoose", MG_VERSION); /* Prepare the environment block */ mg_addenv(blk, "%s", "GATEWAY_INTERFACE=CGI/1.1"); mg_addenv(blk, "%s", "SERVER_PROTOCOL=HTTP/1.1"); mg_addenv(blk, "%s", "REDIRECT_STATUS=200"); /* For PHP */ mg_addenv(blk, "REQUEST_METHOD=%.*s", (int) hm->method.len, hm->method.p); mg_addenv(blk, "REQUEST_URI=%.*s%s%.*s", (int) hm->uri.len, hm->uri.p, hm->query_string.len == 0 ? "" : "?", (int) hm->query_string.len, hm->query_string.p); mg_conn_addr_to_str(nc, buf, sizeof(buf), MG_SOCK_STRINGIFY_REMOTE | MG_SOCK_STRINGIFY_IP); mg_addenv(blk, "REMOTE_ADDR=%s", buf); mg_conn_addr_to_str(nc, buf, sizeof(buf), MG_SOCK_STRINGIFY_PORT); mg_addenv(blk, "SERVER_PORT=%s", buf); s = hm->uri.p + hm->uri.len - path_info_len - 1; if (*s == '/') { const char *base_name = strrchr(prog, DIRSEP); mg_addenv(blk, "SCRIPT_NAME=%.*s/%s", (int) (s - hm->uri.p), hm->uri.p, (base_name != NULL ? base_name + 1 : prog)); } else { mg_addenv(blk, "SCRIPT_NAME=%.*s", (int) (s - hm->uri.p + 1), hm->uri.p); } mg_addenv(blk, "SCRIPT_FILENAME=%s", prog); if (path_info != NULL && path_info->len > 0) { mg_addenv(blk, "PATH_INFO=%.*s", (int) path_info->len, path_info->p); /* Not really translated... */ mg_addenv(blk, "PATH_TRANSLATED=%.*s", (int) path_info->len, path_info->p); } #if MG_ENABLE_SSL mg_addenv(blk, "HTTPS=%s", (nc->flags & MG_F_SSL ? "on" : "off")); #else mg_addenv(blk, "HTTPS=off"); #endif if ((h = mg_get_http_header((struct http_message *) hm, "Content-Type")) != NULL) { mg_addenv(blk, "CONTENT_TYPE=%.*s", (int) h->len, h->p); } if (hm->query_string.len > 0) { mg_addenv(blk, "QUERY_STRING=%.*s", (int) hm->query_string.len, hm->query_string.p); } if ((h = mg_get_http_header((struct http_message *) hm, "Content-Length")) != NULL) { mg_addenv(blk, "CONTENT_LENGTH=%.*s", (int) h->len, h->p); } mg_addenv2(blk, "PATH"); mg_addenv2(blk, "TMP"); mg_addenv2(blk, "TEMP"); mg_addenv2(blk, "TMPDIR"); mg_addenv2(blk, "PERLLIB"); mg_addenv2(blk, MG_ENV_EXPORT_TO_CGI); #ifdef _WIN32 mg_addenv2(blk, "COMSPEC"); mg_addenv2(blk, "SYSTEMROOT"); mg_addenv2(blk, "SystemDrive"); mg_addenv2(blk, "ProgramFiles"); mg_addenv2(blk, "ProgramFiles(x86)"); mg_addenv2(blk, "CommonProgramFiles(x86)"); #else mg_addenv2(blk, "LD_LIBRARY_PATH"); #endif /* _WIN32 */ /* Add all headers as HTTP_* variables */ for (i = 0; hm->header_names[i].len > 0; i++) { p = mg_addenv(blk, "HTTP_%.*s=%.*s", (int) hm->header_names[i].len, hm->header_names[i].p, (int) hm->header_values[i].len, hm->header_values[i].p); /* Convert variable name into uppercase, and change - to _ */ for (; *p != '=' && *p != '\0'; p++) { if (*p == '-') *p = '_'; *p = (char) toupper(*(unsigned char *) p); } } blk->vars[blk->nvars++] = NULL; blk->buf[blk->len++] = '\0'; } static void mg_cgi_ev_handler(struct mg_connection *cgi_nc, int ev, void *ev_data MG_UD_ARG(void *user_data)) { #if !MG_ENABLE_CALLBACK_USERDATA void *user_data = cgi_nc->user_data; #endif struct mg_connection *nc = (struct mg_connection *) user_data; (void) ev_data; if (nc == NULL) { /* The corresponding network connection was closed. */ cgi_nc->flags |= MG_F_CLOSE_IMMEDIATELY; return; } switch (ev) { case MG_EV_RECV: /* * CGI script does not output reply line, like "HTTP/1.1 CODE XXXXX\n" * It outputs headers, then body. Headers might include "Status" * header, which changes CODE, and it might include "Location" header * which changes CODE to 302. * * Therefore we do not send the output from the CGI script to the user * until all CGI headers are received. * * Here we parse the output from the CGI script, and if all headers has * been received, send appropriate reply line, and forward all * received headers to the client. */ if (nc->flags & MG_F_HTTP_CGI_PARSE_HEADERS) { struct mbuf *io = &cgi_nc->recv_mbuf; int len = mg_http_get_request_len(io->buf, io->len); if (len == 0) break; if (len < 0 || io->len > MG_MAX_HTTP_REQUEST_SIZE) { cgi_nc->flags |= MG_F_CLOSE_IMMEDIATELY; mg_http_send_error(nc, 500, "Bad headers"); } else { struct http_message hm; struct mg_str *h; mg_http_parse_headers(io->buf, io->buf + io->len, io->len, &hm); if (mg_get_http_header(&hm, "Location") != NULL) { mg_printf(nc, "%s", "HTTP/1.1 302 Moved\r\n"); } else if ((h = mg_get_http_header(&hm, "Status")) != NULL) { mg_printf(nc, "HTTP/1.1 %.*s\r\n", (int) h->len, h->p); } else { mg_printf(nc, "%s", "HTTP/1.1 200 OK\r\n"); } } nc->flags &= ~MG_F_HTTP_CGI_PARSE_HEADERS; } if (!(nc->flags & MG_F_HTTP_CGI_PARSE_HEADERS)) { mg_forward(cgi_nc, nc); } break; case MG_EV_CLOSE: DBG(("%p CLOSE", cgi_nc)); mg_http_free_proto_data_cgi(&mg_http_get_proto_data(nc)->cgi); nc->flags |= MG_F_SEND_AND_CLOSE; break; } } MG_INTERNAL void mg_handle_cgi(struct mg_connection *nc, const char *prog, const struct mg_str *path_info, const struct http_message *hm, const struct mg_serve_http_opts *opts) { struct mg_cgi_env_block blk; char dir[MG_MAX_PATH]; const char *p; sock_t fds[2]; DBG(("%p [%s]", nc, prog)); mg_prepare_cgi_environment(nc, prog, path_info, hm, opts, &blk); /* * CGI must be executed in its own directory. 'dir' must point to the * directory containing executable program, 'p' must point to the * executable program name relative to 'dir'. */ if ((p = strrchr(prog, DIRSEP)) == NULL) { snprintf(dir, sizeof(dir), "%s", "."); } else { snprintf(dir, sizeof(dir), "%.*s", (int) (p - prog), prog); prog = p + 1; } if (!mg_socketpair(fds, SOCK_STREAM)) { nc->flags |= MG_F_CLOSE_IMMEDIATELY; return; } #ifndef _WIN32 struct sigaction sa; sigemptyset(&sa.sa_mask); sa.sa_handler = SIG_IGN; sa.sa_flags = 0; sigaction(SIGCHLD, &sa, NULL); #endif if (mg_start_process(opts->cgi_interpreter, prog, blk.buf, blk.vars, dir, fds[1]) != 0) { size_t n = nc->recv_mbuf.len - (hm->message.len - hm->body.len); struct mg_connection *cgi_nc = mg_add_sock(nc->mgr, fds[0], mg_cgi_ev_handler MG_UD_ARG(nc)); struct mg_http_proto_data *cgi_pd = mg_http_get_proto_data(nc); cgi_pd->cgi.cgi_nc = cgi_nc; #if !MG_ENABLE_CALLBACK_USERDATA cgi_pd->cgi.cgi_nc->user_data = nc; #endif nc->flags |= MG_F_HTTP_CGI_PARSE_HEADERS; /* Push POST data to the CGI */ if (n > 0 && n < nc->recv_mbuf.len) { mg_send(cgi_pd->cgi.cgi_nc, hm->body.p, n); } mbuf_remove(&nc->recv_mbuf, nc->recv_mbuf.len); } else { closesocket(fds[0]); mg_http_send_error(nc, 500, "CGI failure"); } #ifndef _WIN32 closesocket(fds[1]); /* On Windows, CGI stdio thread closes that socket */ #endif } MG_INTERNAL void mg_http_free_proto_data_cgi(struct mg_http_proto_data_cgi *d) { if (d == NULL) return; if (d->cgi_nc != NULL) { d->cgi_nc->flags |= MG_F_CLOSE_IMMEDIATELY; d->cgi_nc->user_data = NULL; } memset(d, 0, sizeof(*d)); } #endif /* MG_ENABLE_HTTP && MG_ENABLE_HTTP_CGI */ #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_http_ssi.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #if MG_ENABLE_HTTP && MG_ENABLE_HTTP_SSI && MG_ENABLE_FILESYSTEM static void mg_send_ssi_file(struct mg_connection *nc, struct http_message *hm, const char *path, FILE *fp, int include_level, const struct mg_serve_http_opts *opts); static void mg_send_file_data(struct mg_connection *nc, FILE *fp) { char buf[BUFSIZ]; size_t n; while ((n = mg_fread(buf, 1, sizeof(buf), fp)) > 0) { mg_send(nc, buf, n); } } static void mg_do_ssi_include(struct mg_connection *nc, struct http_message *hm, const char *ssi, char *tag, int include_level, const struct mg_serve_http_opts *opts) { char file_name[MG_MAX_PATH], path[MG_MAX_PATH], *p; FILE *fp; /* * sscanf() is safe here, since send_ssi_file() also uses buffer * of size MG_BUF_LEN to get the tag. So strlen(tag) is always < MG_BUF_LEN. */ if (sscanf(tag, " virtual=\"%[^\"]\"", file_name) == 1) { /* File name is relative to the webserver root */ snprintf(path, sizeof(path), "%s/%s", opts->document_root, file_name); } else if (sscanf(tag, " abspath=\"%[^\"]\"", file_name) == 1) { /* * File name is relative to the webserver working directory * or it is absolute system path */ snprintf(path, sizeof(path), "%s", file_name); } else if (sscanf(tag, " file=\"%[^\"]\"", file_name) == 1 || sscanf(tag, " \"%[^\"]\"", file_name) == 1) { /* File name is relative to the currect document */ snprintf(path, sizeof(path), "%s", ssi); if ((p = strrchr(path, DIRSEP)) != NULL) { p[1] = '\0'; } snprintf(path + strlen(path), sizeof(path) - strlen(path), "%s", file_name); } else { mg_printf(nc, "Bad SSI #include: [%s]", tag); return; } if ((fp = mg_fopen(path, "rb")) == NULL) { mg_printf(nc, "SSI include error: mg_fopen(%s): %s", path, strerror(mg_get_errno())); } else { mg_set_close_on_exec((sock_t) fileno(fp)); if (mg_match_prefix(opts->ssi_pattern, strlen(opts->ssi_pattern), path) > 0) { mg_send_ssi_file(nc, hm, path, fp, include_level + 1, opts); } else { mg_send_file_data(nc, fp); } fclose(fp); } } #if MG_ENABLE_HTTP_SSI_EXEC static void do_ssi_exec(struct mg_connection *nc, char *tag) { char cmd[BUFSIZ]; FILE *fp; if (sscanf(tag, " \"%[^\"]\"", cmd) != 1) { mg_printf(nc, "Bad SSI #exec: [%s]", tag); } else if ((fp = popen(cmd, "r")) == NULL) { mg_printf(nc, "Cannot SSI #exec: [%s]: %s", cmd, strerror(mg_get_errno())); } else { mg_send_file_data(nc, fp); pclose(fp); } } #endif /* MG_ENABLE_HTTP_SSI_EXEC */ /* * SSI directive has the following format: * */ static void mg_send_ssi_file(struct mg_connection *nc, struct http_message *hm, const char *path, FILE *fp, int include_level, const struct mg_serve_http_opts *opts) { static const struct mg_str btag = MG_MK_STR(" */ buf[i--] = '\0'; while (i > 0 && buf[i] == ' ') { buf[i--] = '\0'; } /* Handle known SSI directives */ if (strncmp(p, d_include.p, d_include.len) == 0) { mg_do_ssi_include(nc, hm, path, p + d_include.len + 1, include_level, opts); } else if (strncmp(p, d_call.p, d_call.len) == 0) { struct mg_ssi_call_ctx cctx; memset(&cctx, 0, sizeof(cctx)); cctx.req = hm; cctx.file = mg_mk_str(path); cctx.arg = mg_mk_str(p + d_call.len + 1); mg_call(nc, NULL, nc->user_data, MG_EV_SSI_CALL, (void *) cctx.arg.p); /* NUL added above */ mg_call(nc, NULL, nc->user_data, MG_EV_SSI_CALL_CTX, &cctx); #if MG_ENABLE_HTTP_SSI_EXEC } else if (strncmp(p, d_exec.p, d_exec.len) == 0) { do_ssi_exec(nc, p + d_exec.len + 1); #endif } else { /* Silently ignore unknown SSI directive. */ } len = 0; } else if (ch == '<') { in_ssi_tag = 1; if (len > 0) { mg_send(nc, buf, (size_t) len); } len = 0; buf[len++] = ch & 0xff; } else if (in_ssi_tag) { if (len == (int) btag.len && strncmp(buf, btag.p, btag.len) != 0) { /* Not an SSI tag */ in_ssi_tag = 0; } else if (len == (int) sizeof(buf) - 2) { mg_printf(nc, "%s: SSI tag is too large", path); len = 0; } buf[len++] = ch & 0xff; } else { buf[len++] = ch & 0xff; if (len == (int) sizeof(buf)) { mg_send(nc, buf, (size_t) len); len = 0; } } } /* Send the rest of buffered data */ if (len > 0) { mg_send(nc, buf, (size_t) len); } } MG_INTERNAL void mg_handle_ssi_request(struct mg_connection *nc, struct http_message *hm, const char *path, const struct mg_serve_http_opts *opts) { FILE *fp; struct mg_str mime_type; DBG(("%p %s", nc, path)); if ((fp = mg_fopen(path, "rb")) == NULL) { mg_http_send_error(nc, 404, NULL); } else { mg_set_close_on_exec((sock_t) fileno(fp)); mime_type = mg_get_mime_type(path, "text/plain", opts); mg_send_response_line(nc, 200, opts->extra_headers); mg_printf(nc, "Content-Type: %.*s\r\n" "Connection: close\r\n\r\n", (int) mime_type.len, mime_type.p); mg_send_ssi_file(nc, hm, path, fp, 0, opts); fclose(fp); nc->flags |= MG_F_SEND_AND_CLOSE; } } #endif /* MG_ENABLE_HTTP_SSI && MG_ENABLE_HTTP && MG_ENABLE_FILESYSTEM */ #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_http_webdav.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #if MG_ENABLE_HTTP && MG_ENABLE_HTTP_WEBDAV MG_INTERNAL int mg_is_dav_request(const struct mg_str *s) { static const char *methods[] = { "PUT", "DELETE", "MKCOL", "PROPFIND", "MOVE" #if MG_ENABLE_FAKE_DAVLOCK , "LOCK", "UNLOCK" #endif }; size_t i; for (i = 0; i < ARRAY_SIZE(methods); i++) { if (mg_vcmp(s, methods[i]) == 0) { return 1; } } return 0; } static int mg_mkdir(const char *path, uint32_t mode) { #ifndef _WIN32 return mkdir(path, mode); #else (void) mode; return _mkdir(path); #endif } static void mg_print_props(struct mg_connection *nc, const char *name, cs_stat_t *stp) { char mtime[64]; time_t t = stp->st_mtime; /* store in local variable for NDK compile */ struct mg_str name_esc = mg_url_encode(mg_mk_str(name)); mg_gmt_time_string(mtime, sizeof(mtime), &t); mg_printf(nc, "" "%s" "" "" "%s" "%" INT64_FMT "" "%s" "" "HTTP/1.1 200 OK" "" "\n", name_esc.p, S_ISDIR(stp->st_mode) ? "" : "", (int64_t) stp->st_size, mtime); free((void *) name_esc.p); } MG_INTERNAL void mg_handle_propfind(struct mg_connection *nc, const char *path, cs_stat_t *stp, struct http_message *hm, struct mg_serve_http_opts *opts) { static const char header[] = "HTTP/1.1 207 Multi-Status\r\n" "Connection: close\r\n" "Content-Type: text/xml; charset=utf-8\r\n\r\n" "" "\n"; static const char footer[] = "\n"; const struct mg_str *depth = mg_get_http_header(hm, "Depth"); /* Print properties for the requested resource itself */ if (S_ISDIR(stp->st_mode) && strcmp(opts->enable_directory_listing, "yes") != 0) { mg_printf(nc, "%s", "HTTP/1.1 403 Directory Listing Denied\r\n\r\n"); } else { char uri[MG_MAX_PATH]; mg_send(nc, header, sizeof(header) - 1); snprintf(uri, sizeof(uri), "%.*s", (int) hm->uri.len, hm->uri.p); mg_print_props(nc, uri, stp); if (S_ISDIR(stp->st_mode) && (depth == NULL || mg_vcmp(depth, "0") != 0)) { mg_scan_directory(nc, path, opts, mg_print_props); } mg_send(nc, footer, sizeof(footer) - 1); nc->flags |= MG_F_SEND_AND_CLOSE; } } #if MG_ENABLE_FAKE_DAVLOCK /* * Windows explorer (probably there are another WebDav clients like it) * requires LOCK support in webdav. W/out this, it still works, but fails * to save file: shows error message and offers "Save As". * "Save as" works, but this message is very annoying. * This is fake lock, which doesn't lock something, just returns LOCK token, * UNLOCK always answers "OK". * With this fake LOCK Windows Explorer looks happy and saves file. * NOTE: that is not DAV LOCK imlementation, it is just a way to shut up * Windows native DAV client. This is why FAKE LOCK is not enabed by default */ MG_INTERNAL void mg_handle_lock(struct mg_connection *nc, const char *path) { static const char *reply = "HTTP/1.1 207 Multi-Status\r\n" "Connection: close\r\n" "Content-Type: text/xml; charset=utf-8\r\n\r\n" "" "\n" "\n" "\n" "\n" "\n" "opaquelocktoken:%s%u" "" "" "\n" "" "\n"; mg_printf(nc, reply, path, (unsigned int) mg_time()); nc->flags |= MG_F_SEND_AND_CLOSE; } #endif MG_INTERNAL void mg_handle_mkcol(struct mg_connection *nc, const char *path, struct http_message *hm) { int status_code = 500; if (hm->body.len != (size_t) ~0 && hm->body.len > 0) { status_code = 415; } else if (!mg_mkdir(path, 0755)) { status_code = 201; } else if (errno == EEXIST) { status_code = 405; } else if (errno == EACCES) { status_code = 403; } else if (errno == ENOENT) { status_code = 409; } else { status_code = 500; } mg_http_send_error(nc, status_code, NULL); } static int mg_remove_directory(const struct mg_serve_http_opts *opts, const char *dir) { char path[MG_MAX_PATH]; struct dirent *dp; cs_stat_t st; DIR *dirp; if ((dirp = opendir(dir)) == NULL) return 0; while ((dp = readdir(dirp)) != NULL) { if (mg_is_file_hidden((const char *) dp->d_name, opts, 1)) { continue; } snprintf(path, sizeof(path), "%s%c%s", dir, '/', dp->d_name); mg_stat(path, &st); if (S_ISDIR(st.st_mode)) { mg_remove_directory(opts, path); } else { remove(path); } } closedir(dirp); rmdir(dir); return 1; } MG_INTERNAL void mg_handle_move(struct mg_connection *c, const struct mg_serve_http_opts *opts, const char *path, struct http_message *hm) { const struct mg_str *dest = mg_get_http_header(hm, "Destination"); if (dest == NULL) { mg_http_send_error(c, 411, NULL); } else { const char *p = (char *) memchr(dest->p, '/', dest->len); if (p != NULL && p[1] == '/' && (p = (char *) memchr(p + 2, '/', dest->p + dest->len - p)) != NULL) { char buf[MG_MAX_PATH]; snprintf(buf, sizeof(buf), "%s%.*s", opts->dav_document_root, (int) (dest->p + dest->len - p), p); if (rename(path, buf) == 0) { mg_http_send_error(c, 200, NULL); } else { mg_http_send_error(c, 418, NULL); } } else { mg_http_send_error(c, 500, NULL); } } } MG_INTERNAL void mg_handle_delete(struct mg_connection *nc, const struct mg_serve_http_opts *opts, const char *path) { cs_stat_t st; if (mg_stat(path, &st) != 0) { mg_http_send_error(nc, 404, NULL); } else if (S_ISDIR(st.st_mode)) { mg_remove_directory(opts, path); mg_http_send_error(nc, 204, NULL); } else if (remove(path) == 0) { mg_http_send_error(nc, 204, NULL); } else { mg_http_send_error(nc, 423, NULL); } } /* Return -1 on error, 1 on success. */ static int mg_create_itermediate_directories(const char *path) { const char *s; /* Create intermediate directories if they do not exist */ for (s = path + 1; *s != '\0'; s++) { if (*s == '/') { char buf[MG_MAX_PATH]; cs_stat_t st; snprintf(buf, sizeof(buf), "%.*s", (int) (s - path), path); buf[sizeof(buf) - 1] = '\0'; if (mg_stat(buf, &st) != 0 && mg_mkdir(buf, 0755) != 0) { return -1; } } } return 1; } MG_INTERNAL void mg_handle_put(struct mg_connection *nc, const char *path, struct http_message *hm) { struct mg_http_proto_data *pd = mg_http_get_proto_data(nc); cs_stat_t st; const struct mg_str *cl_hdr = mg_get_http_header(hm, "Content-Length"); int rc, status_code = mg_stat(path, &st) == 0 ? 200 : 201; mg_http_free_proto_data_file(&pd->file); if ((rc = mg_create_itermediate_directories(path)) == 0) { mg_printf(nc, "HTTP/1.1 %d OK\r\nContent-Length: 0\r\n\r\n", status_code); } else if (rc == -1) { mg_http_send_error(nc, 500, NULL); } else if (cl_hdr == NULL) { mg_http_send_error(nc, 411, NULL); } else if ((pd->file.fp = mg_fopen(path, "w+b")) == NULL) { mg_http_send_error(nc, 500, NULL); } else { const struct mg_str *range_hdr = mg_get_http_header(hm, "Content-Range"); int64_t r1 = 0, r2 = 0; pd->file.type = DATA_PUT; mg_set_close_on_exec((sock_t) fileno(pd->file.fp)); pd->file.cl = to64(cl_hdr->p); if (range_hdr != NULL && mg_http_parse_range_header(range_hdr, &r1, &r2) > 0) { status_code = 206; fseeko(pd->file.fp, r1, SEEK_SET); pd->file.cl = r2 > r1 ? r2 - r1 + 1 : pd->file.cl - r1; } mg_printf(nc, "HTTP/1.1 %d OK\r\nContent-Length: 0\r\n\r\n", status_code); /* Remove HTTP request from the mbuf, leave only payload */ mbuf_remove(&nc->recv_mbuf, hm->message.len - hm->body.len); mg_http_transfer_file_data(nc); } } #endif /* MG_ENABLE_HTTP && MG_ENABLE_HTTP_WEBDAV */ #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_http_websocket.c" #endif /* * Copyright (c) 2014 Cesanta Software Limited * All rights reserved */ #if MG_ENABLE_HTTP && MG_ENABLE_HTTP_WEBSOCKET /* Amalgamated: #include "common/cs_sha1.h" */ #ifndef MG_WEBSOCKET_PING_INTERVAL_SECONDS #define MG_WEBSOCKET_PING_INTERVAL_SECONDS 5 #endif #define FLAGS_MASK_FIN (1 << 7) #define FLAGS_MASK_OP 0x0f static int mg_is_ws_fragment(unsigned char flags) { return (flags & FLAGS_MASK_FIN) == 0 || (flags & FLAGS_MASK_OP) == WEBSOCKET_OP_CONTINUE; } static int mg_is_ws_first_fragment(unsigned char flags) { return (flags & FLAGS_MASK_FIN) == 0 && (flags & FLAGS_MASK_OP) != WEBSOCKET_OP_CONTINUE; } static int mg_is_ws_control_frame(unsigned char flags) { unsigned char op = (flags & FLAGS_MASK_OP); return op == WEBSOCKET_OP_CLOSE || op == WEBSOCKET_OP_PING || op == WEBSOCKET_OP_PONG; } static void mg_handle_incoming_websocket_frame(struct mg_connection *nc, struct websocket_message *wsm) { if (wsm->flags & 0x8) { mg_call(nc, nc->handler, nc->user_data, MG_EV_WEBSOCKET_CONTROL_FRAME, wsm); } else { mg_call(nc, nc->handler, nc->user_data, MG_EV_WEBSOCKET_FRAME, wsm); } } static struct mg_ws_proto_data *mg_ws_get_proto_data(struct mg_connection *nc) { struct mg_http_proto_data *htd = mg_http_get_proto_data(nc); return (htd != NULL ? &htd->ws_data : NULL); } /* * Sends a Close websocket frame with the given data, and closes the underlying * connection. If `len` is ~0, strlen(data) is used. */ static void mg_ws_close(struct mg_connection *nc, const void *data, size_t len) { if ((int) len == ~0) { len = strlen((const char *) data); } mg_send_websocket_frame(nc, WEBSOCKET_OP_CLOSE, data, len); nc->flags |= MG_F_SEND_AND_CLOSE; } static int mg_deliver_websocket_data(struct mg_connection *nc) { /* Using unsigned char *, cause of integer arithmetic below */ uint64_t i, data_len = 0, frame_len = 0, new_data_len = nc->recv_mbuf.len, len, mask_len = 0, header_len = 0; struct mg_ws_proto_data *wsd = mg_ws_get_proto_data(nc); unsigned char *new_data = (unsigned char *) nc->recv_mbuf.buf, *e = (unsigned char *) nc->recv_mbuf.buf + nc->recv_mbuf.len; uint8_t flags; int ok, reass; if (wsd->reass_len > 0) { /* * We already have some previously received data which we need to * reassemble and deliver to the client code when we get the final * fragment. * * NOTE: it doesn't mean that the current message must be a continuation: * it might be a control frame (Close, Ping or Pong), which should be * handled without breaking the fragmented message. */ size_t existing_len = wsd->reass_len; assert(new_data_len >= existing_len); new_data += existing_len; new_data_len -= existing_len; } flags = new_data[0]; reass = new_data_len > 0 && mg_is_ws_fragment(flags) && !(nc->flags & MG_F_WEBSOCKET_NO_DEFRAG); if (reass && mg_is_ws_control_frame(flags)) { /* * Control frames can't be fragmented, so if we encounter fragmented * control frame, close connection immediately. */ mg_ws_close(nc, "fragmented control frames are illegal", ~0); return 0; } else if (new_data_len > 0 && !reass && !mg_is_ws_control_frame(flags) && wsd->reass_len > 0) { /* * When in the middle of a fragmented message, only the continuations * and control frames are allowed. */ mg_ws_close(nc, "non-continuation in the middle of a fragmented message", ~0); return 0; } if (new_data_len >= 2) { len = new_data[1] & 0x7f; mask_len = new_data[1] & FLAGS_MASK_FIN ? 4 : 0; if (len < 126 && new_data_len >= mask_len) { data_len = len; header_len = 2 + mask_len; } else if (len == 126 && new_data_len >= 4 + mask_len) { header_len = 4 + mask_len; data_len = ntohs(*(uint16_t *) &new_data[2]); } else if (new_data_len >= 10 + mask_len) { header_len = 10 + mask_len; data_len = (((uint64_t) ntohl(*(uint32_t *) &new_data[2])) << 32) + ntohl(*(uint32_t *) &new_data[6]); } } frame_len = header_len + data_len; ok = (frame_len > 0 && frame_len <= new_data_len); /* Check for overflow */ if (frame_len < header_len || frame_len < data_len) { ok = 0; mg_ws_close(nc, "overflowed message", ~0); } if (ok) { size_t cleanup_len = 0; struct websocket_message wsm; wsm.size = (size_t) data_len; wsm.data = new_data + header_len; wsm.flags = flags; /* Apply mask if necessary */ if (mask_len > 0) { for (i = 0; i < data_len; i++) { new_data[i + header_len] ^= (new_data + header_len - mask_len)[i % 4]; } } if (reass) { /* This is a message fragment */ if (mg_is_ws_first_fragment(flags)) { /* * On the first fragmented frame, skip the first byte (op) and also * reset size to 1 (op), it'll be incremented with the data len below. */ new_data += 1; wsd->reass_len = 1 /* op */; } /* Append this frame to the reassembled buffer */ memmove(new_data, wsm.data, e - wsm.data); wsd->reass_len += wsm.size; nc->recv_mbuf.len -= wsm.data - new_data; if (flags & FLAGS_MASK_FIN) { /* On last fragmented frame - call user handler and remove data */ wsm.flags = FLAGS_MASK_FIN | nc->recv_mbuf.buf[0]; wsm.data = (unsigned char *) nc->recv_mbuf.buf + 1 /* op */; wsm.size = wsd->reass_len - 1 /* op */; cleanup_len = wsd->reass_len; wsd->reass_len = 0; /* Pass reassembled message to the client code. */ mg_handle_incoming_websocket_frame(nc, &wsm); mbuf_remove(&nc->recv_mbuf, cleanup_len); /* Cleanup frame */ } } else { /* * This is a complete message, not a fragment. It might happen in between * of a fragmented message (in this case, WebSocket protocol requires * current message to be a control frame). */ cleanup_len = (size_t) frame_len; /* First of all, check if we need to react on a control frame. */ switch (flags & FLAGS_MASK_OP) { case WEBSOCKET_OP_PING: mg_send_websocket_frame(nc, WEBSOCKET_OP_PONG, wsm.data, wsm.size); break; case WEBSOCKET_OP_CLOSE: mg_ws_close(nc, wsm.data, wsm.size); break; } /* Pass received message to the client code. */ mg_handle_incoming_websocket_frame(nc, &wsm); /* Cleanup frame */ memmove(nc->recv_mbuf.buf + wsd->reass_len, nc->recv_mbuf.buf + wsd->reass_len + cleanup_len, nc->recv_mbuf.len - wsd->reass_len - cleanup_len); nc->recv_mbuf.len -= cleanup_len; } } return ok; } struct ws_mask_ctx { size_t pos; /* zero means unmasked */ uint32_t mask; }; static uint32_t mg_ws_random_mask(void) { uint32_t mask; /* * The spec requires WS client to generate hard to * guess mask keys. From RFC6455, Section 5.3: * * The unpredictability of the masking key is essential to prevent * authors of malicious applications from selecting the bytes that appear on * the wire. * * Hence this feature is essential when the actual end user of this API * is untrusted code that wouldn't have access to a lower level net API * anyway (e.g. web browsers). Hence this feature is low prio for most * mongoose use cases and thus can be disabled, e.g. when porting to a platform * that lacks rand(). */ #if MG_DISABLE_WS_RANDOM_MASK mask = 0xefbeadde; /* generated with a random number generator, I swear */ #else if (sizeof(long) >= 4) { mask = (uint32_t) rand(); } else if (sizeof(long) == 2) { mask = (uint32_t) rand() << 16 | (uint32_t) rand(); } #endif return mask; } static void mg_send_ws_header(struct mg_connection *nc, int op, size_t len, struct ws_mask_ctx *ctx) { int header_len; unsigned char header[10]; header[0] = (op & WEBSOCKET_DONT_FIN ? 0x0 : FLAGS_MASK_FIN) | (op & FLAGS_MASK_OP); if (len < 126) { header[1] = (unsigned char) len; header_len = 2; } else if (len < 65535) { uint16_t tmp = htons((uint16_t) len); header[1] = 126; memcpy(&header[2], &tmp, sizeof(tmp)); header_len = 4; } else { uint32_t tmp; header[1] = 127; tmp = htonl((uint32_t)((uint64_t) len >> 32)); memcpy(&header[2], &tmp, sizeof(tmp)); tmp = htonl((uint32_t)(len & 0xffffffff)); memcpy(&header[6], &tmp, sizeof(tmp)); header_len = 10; } /* client connections enable masking */ if (nc->listener == NULL) { header[1] |= 1 << 7; /* set masking flag */ mg_send(nc, header, header_len); ctx->mask = mg_ws_random_mask(); mg_send(nc, &ctx->mask, sizeof(ctx->mask)); ctx->pos = nc->send_mbuf.len; } else { mg_send(nc, header, header_len); ctx->pos = 0; } } static void mg_ws_mask_frame(struct mbuf *mbuf, struct ws_mask_ctx *ctx) { size_t i; if (ctx->pos == 0) return; for (i = 0; i < (mbuf->len - ctx->pos); i++) { mbuf->buf[ctx->pos + i] ^= ((char *) &ctx->mask)[i % 4]; } } void mg_send_websocket_frame(struct mg_connection *nc, int op, const void *data, size_t len) { struct ws_mask_ctx ctx; DBG(("%p %d %d", nc, op, (int) len)); mg_send_ws_header(nc, op, len, &ctx); mg_send(nc, data, len); mg_ws_mask_frame(&nc->send_mbuf, &ctx); if (op == WEBSOCKET_OP_CLOSE) { nc->flags |= MG_F_SEND_AND_CLOSE; } } void mg_send_websocket_framev(struct mg_connection *nc, int op, const struct mg_str *strv, int strvcnt) { struct ws_mask_ctx ctx; int i; int len = 0; for (i = 0; i < strvcnt; i++) { len += strv[i].len; } mg_send_ws_header(nc, op, len, &ctx); for (i = 0; i < strvcnt; i++) { mg_send(nc, strv[i].p, strv[i].len); } mg_ws_mask_frame(&nc->send_mbuf, &ctx); if (op == WEBSOCKET_OP_CLOSE) { nc->flags |= MG_F_SEND_AND_CLOSE; } } void mg_printf_websocket_frame(struct mg_connection *nc, int op, const char *fmt, ...) { char mem[MG_VPRINTF_BUFFER_SIZE], *buf = mem; va_list ap; int len; va_start(ap, fmt); if ((len = mg_avprintf(&buf, sizeof(mem), fmt, ap)) > 0) { mg_send_websocket_frame(nc, op, buf, len); } va_end(ap); if (buf != mem && buf != NULL) { MG_FREE(buf); } } MG_INTERNAL void mg_ws_handler(struct mg_connection *nc, int ev, void *ev_data MG_UD_ARG(void *user_data)) { mg_call(nc, nc->handler, nc->user_data, ev, ev_data); switch (ev) { case MG_EV_RECV: do { } while (mg_deliver_websocket_data(nc)); break; case MG_EV_POLL: /* Ping idle websocket connections */ { time_t now = *(time_t *) ev_data; if (nc->flags & MG_F_IS_WEBSOCKET && now > nc->last_io_time + MG_WEBSOCKET_PING_INTERVAL_SECONDS) { mg_send_websocket_frame(nc, WEBSOCKET_OP_PING, "", 0); } } break; default: break; } #if MG_ENABLE_CALLBACK_USERDATA (void) user_data; #endif } #ifndef MG_EXT_SHA1 void mg_hash_sha1_v(size_t num_msgs, const uint8_t *msgs[], const size_t *msg_lens, uint8_t *digest) { size_t i; cs_sha1_ctx sha_ctx; cs_sha1_init(&sha_ctx); for (i = 0; i < num_msgs; i++) { cs_sha1_update(&sha_ctx, msgs[i], msg_lens[i]); } cs_sha1_final(digest, &sha_ctx); } #else extern void mg_hash_sha1_v(size_t num_msgs, const uint8_t *msgs[], const size_t *msg_lens, uint8_t *digest); #endif MG_INTERNAL void mg_ws_handshake(struct mg_connection *nc, const struct mg_str *key, struct http_message *hm) { static const char *magic = "258EAFA5-E914-47DA-95CA-C5AB0DC85B11"; const uint8_t *msgs[2] = {(const uint8_t *) key->p, (const uint8_t *) magic}; const size_t msg_lens[2] = {key->len, 36}; unsigned char sha[20]; char b64_sha[30]; struct mg_str *s; mg_hash_sha1_v(2, msgs, msg_lens, sha); mg_base64_encode(sha, sizeof(sha), b64_sha); mg_printf(nc, "%s", "HTTP/1.1 101 Switching Protocols\r\n" "Upgrade: websocket\r\n" "Connection: Upgrade\r\n"); s = mg_get_http_header(hm, "Sec-WebSocket-Protocol"); if (s != NULL) { mg_printf(nc, "Sec-WebSocket-Protocol: %.*s\r\n", (int) s->len, s->p); } mg_printf(nc, "Sec-WebSocket-Accept: %s%s", b64_sha, "\r\n\r\n"); DBG(("%p %.*s %s", nc, (int) key->len, key->p, b64_sha)); } void mg_send_websocket_handshake2(struct mg_connection *nc, const char *path, const char *host, const char *protocol, const char *extra_headers) { mg_send_websocket_handshake3(nc, path, host, protocol, extra_headers, NULL, NULL); } void mg_send_websocket_handshake3(struct mg_connection *nc, const char *path, const char *host, const char *protocol, const char *extra_headers, const char *user, const char *pass) { mg_send_websocket_handshake3v(nc, mg_mk_str(path), mg_mk_str(host), mg_mk_str(protocol), mg_mk_str(extra_headers), mg_mk_str(user), mg_mk_str(pass)); } void mg_send_websocket_handshake3v(struct mg_connection *nc, const struct mg_str path, const struct mg_str host, const struct mg_str protocol, const struct mg_str extra_headers, const struct mg_str user, const struct mg_str pass) { struct mbuf auth; char key[25]; uint32_t nonce[4]; nonce[0] = mg_ws_random_mask(); nonce[1] = mg_ws_random_mask(); nonce[2] = mg_ws_random_mask(); nonce[3] = mg_ws_random_mask(); mg_base64_encode((unsigned char *) &nonce, sizeof(nonce), key); mbuf_init(&auth, 0); if (user.len > 0) { mg_basic_auth_header(user, pass, &auth); } /* * NOTE: the (auth.buf == NULL ? "" : auth.buf) is because cc3200 libc is * broken: it doesn't like zero length to be passed to %.*s * i.e. sprintf("f%.*so", (int)0, NULL), yields `f\0o`. * because it handles NULL specially (and incorrectly). */ mg_printf(nc, "GET %.*s HTTP/1.1\r\n" "Upgrade: websocket\r\n" "Connection: Upgrade\r\n" "%.*s" "Sec-WebSocket-Version: 13\r\n" "Sec-WebSocket-Key: %s\r\n", (int) path.len, path.p, (int) auth.len, (auth.buf == NULL ? "" : auth.buf), key); /* TODO(mkm): take default hostname from http proto data if host == NULL */ if (host.len > 0) { int host_len = (int) (path.p - host.p); /* Account for possible :PORT */ mg_printf(nc, "Host: %.*s\r\n", host_len, host.p); } if (protocol.len > 0) { mg_printf(nc, "Sec-WebSocket-Protocol: %.*s\r\n", (int) protocol.len, protocol.p); } if (extra_headers.len > 0) { mg_printf(nc, "%.*s", (int) extra_headers.len, extra_headers.p); } mg_printf(nc, "\r\n"); mbuf_free(&auth); } void mg_send_websocket_handshake(struct mg_connection *nc, const char *path, const char *extra_headers) { struct mg_str null_str = MG_NULL_STR; mg_send_websocket_handshake3v( nc, mg_mk_str(path), null_str /* host */, null_str /* protocol */, mg_mk_str(extra_headers), null_str /* user */, null_str /* pass */); } struct mg_connection *mg_connect_ws_opt( struct mg_mgr *mgr, MG_CB(mg_event_handler_t ev_handler, void *user_data), struct mg_connect_opts opts, const char *url, const char *protocol, const char *extra_headers) { struct mg_str null_str = MG_NULL_STR; struct mg_str host = MG_NULL_STR, path = MG_NULL_STR, user_info = MG_NULL_STR; struct mg_connection *nc = mg_connect_http_base(mgr, MG_CB(ev_handler, user_data), opts, "http", "ws", "https", "wss", url, &path, &user_info, &host); if (nc != NULL) { mg_send_websocket_handshake3v(nc, path, host, mg_mk_str(protocol), mg_mk_str(extra_headers), user_info, null_str); } return nc; } struct mg_connection *mg_connect_ws( struct mg_mgr *mgr, MG_CB(mg_event_handler_t ev_handler, void *user_data), const char *url, const char *protocol, const char *extra_headers) { struct mg_connect_opts opts; memset(&opts, 0, sizeof(opts)); return mg_connect_ws_opt(mgr, MG_CB(ev_handler, user_data), opts, url, protocol, extra_headers); } #endif /* MG_ENABLE_HTTP && MG_ENABLE_HTTP_WEBSOCKET */ #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_util.c" #endif /* * Copyright (c) 2014 Cesanta Software Limited * All rights reserved */ /* Amalgamated: #include "common/cs_base64.h" */ /* Amalgamated: #include "mg_internal.h" */ /* Amalgamated: #include "mg_util.h" */ /* For platforms with limited libc */ #ifndef MAX #define MAX(a, b) ((a) > (b) ? (a) : (b)) #endif const char *mg_skip(const char *s, const char *end, const char *delims, struct mg_str *v) { v->p = s; while (s < end && strchr(delims, *(unsigned char *) s) == NULL) s++; v->len = s - v->p; while (s < end && strchr(delims, *(unsigned char *) s) != NULL) s++; return s; } #if MG_ENABLE_FILESYSTEM && !defined(MG_USER_FILE_FUNCTIONS) int mg_stat(const char *path, cs_stat_t *st) { #ifdef _WIN32 wchar_t wpath[MG_MAX_PATH]; to_wchar(path, wpath, ARRAY_SIZE(wpath)); DBG(("[%ls] -> %d", wpath, _wstati64(wpath, st))); return _wstati64(wpath, st); #else return stat(path, st); #endif } FILE *mg_fopen(const char *path, const char *mode) { #ifdef _WIN32 wchar_t wpath[MG_MAX_PATH], wmode[10]; to_wchar(path, wpath, ARRAY_SIZE(wpath)); to_wchar(mode, wmode, ARRAY_SIZE(wmode)); return _wfopen(wpath, wmode); #else return fopen(path, mode); #endif } int mg_open(const char *path, int flag, int mode) { /* LCOV_EXCL_LINE */ #if defined(_WIN32) && !defined(WINCE) wchar_t wpath[MG_MAX_PATH]; to_wchar(path, wpath, ARRAY_SIZE(wpath)); return _wopen(wpath, flag, mode); #else return open(path, flag, mode); /* LCOV_EXCL_LINE */ #endif } size_t mg_fread(void *ptr, size_t size, size_t count, FILE *f) { return fread(ptr, size, count, f); } size_t mg_fwrite(const void *ptr, size_t size, size_t count, FILE *f) { return fwrite(ptr, size, count, f); } #endif void mg_base64_encode(const unsigned char *src, int src_len, char *dst) { cs_base64_encode(src, src_len, dst); } int mg_base64_decode(const unsigned char *s, int len, char *dst) { return cs_base64_decode(s, len, dst, NULL); } #if MG_ENABLE_THREADS void *mg_start_thread(void *(*f)(void *), void *p) { #ifdef WINCE return (void *) CreateThread(NULL, 0, (LPTHREAD_START_ROUTINE) f, p, 0, NULL); #elif defined(_WIN32) return (void *) _beginthread((void(__cdecl *) (void *) ) f, 0, p); #else pthread_t thread_id = (pthread_t) 0; pthread_attr_t attr; (void) pthread_attr_init(&attr); (void) pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED); #if defined(MG_STACK_SIZE) && MG_STACK_SIZE > 1 (void) pthread_attr_setstacksize(&attr, MG_STACK_SIZE); #endif pthread_create(&thread_id, &attr, f, p); pthread_attr_destroy(&attr); return (void *) thread_id; #endif } #endif /* MG_ENABLE_THREADS */ /* Set close-on-exec bit for a given socket. */ void mg_set_close_on_exec(sock_t sock) { #if defined(_WIN32) && !defined(WINCE) (void) SetHandleInformation((HANDLE) sock, HANDLE_FLAG_INHERIT, 0); #elif defined(__unix__) fcntl(sock, F_SETFD, FD_CLOEXEC); #else (void) sock; #endif } int mg_sock_addr_to_str(const union socket_address *sa, char *buf, size_t len, int flags) { int is_v6; if (buf == NULL || len <= 0) return 0; memset(buf, 0, len); #if MG_ENABLE_IPV6 is_v6 = sa->sa.sa_family == AF_INET6; #else is_v6 = 0; #endif if (flags & MG_SOCK_STRINGIFY_IP) { #if MG_ENABLE_IPV6 const void *addr = NULL; char *start = buf; socklen_t capacity = len; if (!is_v6) { addr = &sa->sin.sin_addr; } else { addr = (void *) &sa->sin6.sin6_addr; if (flags & MG_SOCK_STRINGIFY_PORT) { *buf = '['; start++; capacity--; } } if (inet_ntop(sa->sa.sa_family, addr, start, capacity) == NULL) { goto cleanup; } #elif defined(_WIN32) || MG_LWIP || (MG_NET_IF == MG_NET_IF_PIC32) /* Only Windoze Vista (and newer) have inet_ntop() */ char *addr_str = inet_ntoa(sa->sin.sin_addr); if (addr_str != NULL) { strncpy(buf, inet_ntoa(sa->sin.sin_addr), len - 1); } else { goto cleanup; } #else if (inet_ntop(AF_INET, (void *) &sa->sin.sin_addr, buf, len - 1) == NULL) { goto cleanup; } #endif } if (flags & MG_SOCK_STRINGIFY_PORT) { int port = ntohs(sa->sin.sin_port); if (flags & MG_SOCK_STRINGIFY_IP) { int buf_len = strlen(buf); snprintf(buf + buf_len, len - (buf_len + 1), "%s:%d", (is_v6 ? "]" : ""), port); } else { snprintf(buf, len, "%d", port); } } return strlen(buf); cleanup: *buf = '\0'; return 0; } int mg_conn_addr_to_str(struct mg_connection *nc, char *buf, size_t len, int flags) { union socket_address sa; memset(&sa, 0, sizeof(sa)); mg_if_get_conn_addr(nc, flags & MG_SOCK_STRINGIFY_REMOTE, &sa); return mg_sock_addr_to_str(&sa, buf, len, flags); } #if MG_ENABLE_HEXDUMP static int mg_hexdump_n(const void *buf, int len, char *dst, int dst_len, int offset) { const unsigned char *p = (const unsigned char *) buf; char ascii[17] = ""; int i, idx, n = 0; for (i = 0; i < len; i++) { idx = i % 16; if (idx == 0) { if (i > 0) n += snprintf(dst + n, MAX(dst_len - n, 0), " %s\n", ascii); n += snprintf(dst + n, MAX(dst_len - n, 0), "%04x ", i + offset); } if (dst_len - n < 0) { return n; } n += snprintf(dst + n, MAX(dst_len - n, 0), " %02x", p[i]); ascii[idx] = p[i] < 0x20 || p[i] > 0x7e ? '.' : p[i]; ascii[idx + 1] = '\0'; } while (i++ % 16) n += snprintf(dst + n, MAX(dst_len - n, 0), "%s", " "); n += snprintf(dst + n, MAX(dst_len - n, 0), " %s\n", ascii); return n; } int mg_hexdump(const void *buf, int len, char *dst, int dst_len) { return mg_hexdump_n(buf, len, dst, dst_len, 0); } void mg_hexdumpf(FILE *fp, const void *buf, int len) { char tmp[80]; int offset = 0, n; while (len > 0) { n = (len < 16 ? len : 16); mg_hexdump_n(((const char *) buf) + offset, n, tmp, sizeof(tmp), offset); fputs(tmp, fp); offset += n; len -= n; } } void mg_hexdump_connection(struct mg_connection *nc, const char *path, const void *buf, int num_bytes, int ev) { FILE *fp = NULL; char src[60], dst[60]; const char *tag = NULL; switch (ev) { case MG_EV_RECV: tag = "<-"; break; case MG_EV_SEND: tag = "->"; break; case MG_EV_ACCEPT: tag = " 0) { mg_hexdumpf(fp, buf, num_bytes); } if (fp != stdout && fp != stderr) fclose(fp); } #endif int mg_is_big_endian(void) { static const int n = 1; /* TODO(mkm) use compiletime check with 4-byte char literal */ return ((char *) &n)[0] == 0; } DO_NOT_WARN_UNUSED MG_INTERNAL int mg_get_errno(void) { #ifndef WINCE return errno; #else /* TODO(alashkin): translate error codes? */ return GetLastError(); #endif } void mg_mbuf_append_base64_putc(char ch, void *user_data) { struct mbuf *mbuf = (struct mbuf *) user_data; mbuf_append(mbuf, &ch, sizeof(ch)); } void mg_mbuf_append_base64(struct mbuf *mbuf, const void *data, size_t len) { struct cs_base64_ctx ctx; cs_base64_init(&ctx, mg_mbuf_append_base64_putc, mbuf); cs_base64_update(&ctx, (const char *) data, len); cs_base64_finish(&ctx); } void mg_basic_auth_header(const struct mg_str user, const struct mg_str pass, struct mbuf *buf) { const char *header_prefix = "Authorization: Basic "; const char *header_suffix = "\r\n"; struct cs_base64_ctx ctx; cs_base64_init(&ctx, mg_mbuf_append_base64_putc, buf); mbuf_append(buf, header_prefix, strlen(header_prefix)); cs_base64_update(&ctx, user.p, user.len); if (pass.len > 0) { cs_base64_update(&ctx, ":", 1); cs_base64_update(&ctx, pass.p, pass.len); } cs_base64_finish(&ctx); mbuf_append(buf, header_suffix, strlen(header_suffix)); } struct mg_str mg_url_encode(const struct mg_str src) { static const char *dont_escape = "._-$,;~()/"; static const char *hex = "0123456789abcdef"; size_t i = 0; struct mbuf mb; mbuf_init(&mb, src.len); for (i = 0; i < src.len; i++) { const unsigned char c = *((const unsigned char *) src.p + i); if (isalnum(c) || strchr(dont_escape, c) != NULL) { mbuf_append(&mb, &c, 1); } else { mbuf_append(&mb, "%", 1); mbuf_append(&mb, &hex[c >> 4], 1); mbuf_append(&mb, &hex[c & 15], 1); } } mbuf_append(&mb, "", 1); mbuf_trim(&mb); return mg_mk_str_n(mb.buf, mb.len - 1); } #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_mqtt.c" #endif /* * Copyright (c) 2014 Cesanta Software Limited * All rights reserved */ #if MG_ENABLE_MQTT #include /* Amalgamated: #include "mg_internal.h" */ /* Amalgamated: #include "mg_mqtt.h" */ static uint16_t getu16(const char *p) { const uint8_t *up = (const uint8_t *) p; return (up[0] << 8) + up[1]; } static const char *scanto(const char *p, struct mg_str *s) { s->len = getu16(p); s->p = p + 2; return s->p + s->len; } MG_INTERNAL int parse_mqtt(struct mbuf *io, struct mg_mqtt_message *mm) { uint8_t header; size_t len = 0, len_len = 0; const char *p, *end; unsigned char lc = 0; int cmd; if (io->len < 2) return MG_MQTT_ERROR_INCOMPLETE_MSG; header = io->buf[0]; cmd = header >> 4; /* decode mqtt variable length */ len = len_len = 0; p = io->buf + 1; while ((size_t)(p - io->buf) < io->len) { lc = *((const unsigned char *) p++); len += (lc & 0x7f) << 7 * len_len; len_len++; if (!(lc & 0x80)) break; if (len_len > 4) return MG_MQTT_ERROR_MALFORMED_MSG; } end = p + len; if (lc & 0x80 || len > (io->len - (p - io->buf))) { return MG_MQTT_ERROR_INCOMPLETE_MSG; } mm->cmd = cmd; mm->qos = MG_MQTT_GET_QOS(header); switch (cmd) { case MG_MQTT_CMD_CONNECT: { p = scanto(p, &mm->protocol_name); if (p > end - 4) return MG_MQTT_ERROR_MALFORMED_MSG; mm->protocol_version = *(uint8_t *) p++; mm->connect_flags = *(uint8_t *) p++; mm->keep_alive_timer = getu16(p); p += 2; if (p >= end) return MG_MQTT_ERROR_MALFORMED_MSG; p = scanto(p, &mm->client_id); if (p > end) return MG_MQTT_ERROR_MALFORMED_MSG; if (mm->connect_flags & MG_MQTT_HAS_WILL) { if (p >= end) return MG_MQTT_ERROR_MALFORMED_MSG; p = scanto(p, &mm->will_topic); } if (mm->connect_flags & MG_MQTT_HAS_WILL) { if (p >= end) return MG_MQTT_ERROR_MALFORMED_MSG; p = scanto(p, &mm->will_message); } if (mm->connect_flags & MG_MQTT_HAS_USER_NAME) { if (p >= end) return MG_MQTT_ERROR_MALFORMED_MSG; p = scanto(p, &mm->user_name); } if (mm->connect_flags & MG_MQTT_HAS_PASSWORD) { if (p >= end) return MG_MQTT_ERROR_MALFORMED_MSG; p = scanto(p, &mm->password); } if (p != end) return MG_MQTT_ERROR_MALFORMED_MSG; LOG(LL_DEBUG, ("%d %2x %d proto [%.*s] client_id [%.*s] will_topic [%.*s] " "will_msg [%.*s] user_name [%.*s] password [%.*s]", (int) len, (int) mm->connect_flags, (int) mm->keep_alive_timer, (int) mm->protocol_name.len, mm->protocol_name.p, (int) mm->client_id.len, mm->client_id.p, (int) mm->will_topic.len, mm->will_topic.p, (int) mm->will_message.len, mm->will_message.p, (int) mm->user_name.len, mm->user_name.p, (int) mm->password.len, mm->password.p)); break; } case MG_MQTT_CMD_CONNACK: if (end - p < 2) return MG_MQTT_ERROR_MALFORMED_MSG; mm->connack_ret_code = p[1]; break; case MG_MQTT_CMD_PUBACK: case MG_MQTT_CMD_PUBREC: case MG_MQTT_CMD_PUBREL: case MG_MQTT_CMD_PUBCOMP: case MG_MQTT_CMD_SUBACK: mm->message_id = getu16(p); break; case MG_MQTT_CMD_PUBLISH: { p = scanto(p, &mm->topic); if (p > end) return MG_MQTT_ERROR_MALFORMED_MSG; if (mm->qos > 0) { if (end - p < 2) return MG_MQTT_ERROR_MALFORMED_MSG; mm->message_id = getu16(p); p += 2; } mm->payload.p = p; mm->payload.len = end - p; break; } case MG_MQTT_CMD_SUBSCRIBE: if (end - p < 2) return MG_MQTT_ERROR_MALFORMED_MSG; mm->message_id = getu16(p); p += 2; /* * topic expressions are left in the payload and can be parsed with * `mg_mqtt_next_subscribe_topic` */ mm->payload.p = p; mm->payload.len = end - p; break; default: /* Unhandled command */ break; } mm->len = end - io->buf; return mm->len; } static void mqtt_handler(struct mg_connection *nc, int ev, void *ev_data MG_UD_ARG(void *user_data)) { struct mbuf *io = &nc->recv_mbuf; struct mg_mqtt_message mm; memset(&mm, 0, sizeof(mm)); nc->handler(nc, ev, ev_data MG_UD_ARG(user_data)); switch (ev) { case MG_EV_ACCEPT: if (nc->proto_data == NULL) mg_set_protocol_mqtt(nc); break; case MG_EV_RECV: { /* There can be multiple messages in the buffer, process them all. */ while (1) { int len = parse_mqtt(io, &mm); if (len < 0) { if (len == MG_MQTT_ERROR_MALFORMED_MSG) { /* Protocol error. */ nc->flags |= MG_F_CLOSE_IMMEDIATELY; } else if (len == MG_MQTT_ERROR_INCOMPLETE_MSG) { /* Not fully buffered, let's check if we have a chance to get more * data later */ if (nc->recv_mbuf_limit > 0 && nc->recv_mbuf.len >= nc->recv_mbuf_limit) { LOG(LL_ERROR, ("%p recv buffer (%lu bytes) exceeds the limit " "%lu bytes, and not drained, closing", nc, (unsigned long) nc->recv_mbuf.len, (unsigned long) nc->recv_mbuf_limit)); nc->flags |= MG_F_CLOSE_IMMEDIATELY; } } else { /* Should never be here */ LOG(LL_ERROR, ("%p invalid len: %d, closing", nc, len)); nc->flags |= MG_F_CLOSE_IMMEDIATELY; } break; } nc->handler(nc, MG_MQTT_EVENT_BASE + mm.cmd, &mm MG_UD_ARG(user_data)); mbuf_remove(io, len); } break; } case MG_EV_POLL: { struct mg_mqtt_proto_data *pd = (struct mg_mqtt_proto_data *) nc->proto_data; double now = mg_time(); if (pd->keep_alive > 0 && pd->last_control_time > 0 && (now - pd->last_control_time) > pd->keep_alive) { LOG(LL_DEBUG, ("Send PINGREQ")); mg_mqtt_ping(nc); } break; } } } static void mg_mqtt_proto_data_destructor(void *proto_data) { MG_FREE(proto_data); } int mg_mqtt_match_topic_expression(struct mg_str exp, struct mg_str topic) { /* TODO(mkm): implement real matching */ if (memchr(exp.p, '#', exp.len)) { /* exp `foo/#` will become `foo/` */ exp.len -= 1; /* * topic should be longer than the expression: e.g. topic `foo/bar` does * match `foo/#`, but neither `foo` nor `foo/` do. */ if (topic.len <= exp.len) { return 0; } /* Truncate topic so that it'll pass the next length check */ topic.len = exp.len; } if (topic.len != exp.len) { return 0; } return strncmp(topic.p, exp.p, exp.len) == 0; } int mg_mqtt_vmatch_topic_expression(const char *exp, struct mg_str topic) { return mg_mqtt_match_topic_expression(mg_mk_str(exp), topic); } void mg_set_protocol_mqtt(struct mg_connection *nc) { nc->proto_handler = mqtt_handler; nc->proto_data = MG_CALLOC(1, sizeof(struct mg_mqtt_proto_data)); nc->proto_data_destructor = mg_mqtt_proto_data_destructor; } static void mg_mqtt_prepend_header(struct mg_connection *nc, uint8_t cmd, uint8_t flags, size_t len) { struct mg_mqtt_proto_data *pd = (struct mg_mqtt_proto_data *) nc->proto_data; size_t off = nc->send_mbuf.len - len; uint8_t header = cmd << 4 | (uint8_t) flags; uint8_t buf[1 + sizeof(size_t)]; uint8_t *vlen = &buf[1]; assert(nc->send_mbuf.len >= len); buf[0] = header; /* mqtt variable length encoding */ do { *vlen = len % 0x80; len /= 0x80; if (len > 0) *vlen |= 0x80; vlen++; } while (len > 0); mbuf_insert(&nc->send_mbuf, off, buf, vlen - buf); pd->last_control_time = mg_time(); } void mg_send_mqtt_handshake(struct mg_connection *nc, const char *client_id) { static struct mg_send_mqtt_handshake_opts opts; mg_send_mqtt_handshake_opt(nc, client_id, opts); } void mg_send_mqtt_handshake_opt(struct mg_connection *nc, const char *client_id, struct mg_send_mqtt_handshake_opts opts) { uint16_t hlen, nlen, rem_len = 0; struct mg_mqtt_proto_data *pd = (struct mg_mqtt_proto_data *) nc->proto_data; mg_send(nc, "\00\04MQTT\04", 7); rem_len += 7; if (opts.user_name != NULL) { opts.flags |= MG_MQTT_HAS_USER_NAME; } if (opts.password != NULL) { opts.flags |= MG_MQTT_HAS_PASSWORD; } if (opts.will_topic != NULL && opts.will_message != NULL) { opts.flags |= MG_MQTT_HAS_WILL; } if (opts.keep_alive == 0) { opts.keep_alive = 60; } mg_send(nc, &opts.flags, 1); rem_len += 1; nlen = htons(opts.keep_alive); mg_send(nc, &nlen, 2); rem_len += 2; hlen = strlen(client_id); nlen = htons((uint16_t) hlen); mg_send(nc, &nlen, 2); mg_send(nc, client_id, hlen); rem_len += 2 + hlen; if (opts.flags & MG_MQTT_HAS_WILL) { hlen = strlen(opts.will_topic); nlen = htons((uint16_t) hlen); mg_send(nc, &nlen, 2); mg_send(nc, opts.will_topic, hlen); rem_len += 2 + hlen; hlen = strlen(opts.will_message); nlen = htons((uint16_t) hlen); mg_send(nc, &nlen, 2); mg_send(nc, opts.will_message, hlen); rem_len += 2 + hlen; } if (opts.flags & MG_MQTT_HAS_USER_NAME) { hlen = strlen(opts.user_name); nlen = htons((uint16_t) hlen); mg_send(nc, &nlen, 2); mg_send(nc, opts.user_name, hlen); rem_len += 2 + hlen; } if (opts.flags & MG_MQTT_HAS_PASSWORD) { hlen = strlen(opts.password); nlen = htons((uint16_t) hlen); mg_send(nc, &nlen, 2); mg_send(nc, opts.password, hlen); rem_len += 2 + hlen; } mg_mqtt_prepend_header(nc, MG_MQTT_CMD_CONNECT, 0, rem_len); if (pd != NULL) { pd->keep_alive = opts.keep_alive; } } void mg_mqtt_publish(struct mg_connection *nc, const char *topic, uint16_t message_id, int flags, const void *data, size_t len) { size_t old_len = nc->send_mbuf.len; uint16_t topic_len = htons((uint16_t) strlen(topic)); uint16_t message_id_net = htons(message_id); mg_send(nc, &topic_len, 2); mg_send(nc, topic, strlen(topic)); if (MG_MQTT_GET_QOS(flags) > 0) { mg_send(nc, &message_id_net, 2); } mg_send(nc, data, len); mg_mqtt_prepend_header(nc, MG_MQTT_CMD_PUBLISH, flags, nc->send_mbuf.len - old_len); } void mg_mqtt_subscribe(struct mg_connection *nc, const struct mg_mqtt_topic_expression *topics, size_t topics_len, uint16_t message_id) { size_t old_len = nc->send_mbuf.len; uint16_t message_id_n = htons(message_id); size_t i; mg_send(nc, (char *) &message_id_n, 2); for (i = 0; i < topics_len; i++) { uint16_t topic_len_n = htons((uint16_t) strlen(topics[i].topic)); mg_send(nc, &topic_len_n, 2); mg_send(nc, topics[i].topic, strlen(topics[i].topic)); mg_send(nc, &topics[i].qos, 1); } mg_mqtt_prepend_header(nc, MG_MQTT_CMD_SUBSCRIBE, MG_MQTT_QOS(1), nc->send_mbuf.len - old_len); } int mg_mqtt_next_subscribe_topic(struct mg_mqtt_message *msg, struct mg_str *topic, uint8_t *qos, int pos) { unsigned char *buf = (unsigned char *) msg->payload.p + pos; int new_pos; if ((size_t) pos >= msg->payload.len) return -1; topic->len = buf[0] << 8 | buf[1]; topic->p = (char *) buf + 2; new_pos = pos + 2 + topic->len + 1; if ((size_t) new_pos > msg->payload.len) return -1; *qos = buf[2 + topic->len]; return new_pos; } void mg_mqtt_unsubscribe(struct mg_connection *nc, char **topics, size_t topics_len, uint16_t message_id) { size_t old_len = nc->send_mbuf.len; uint16_t message_id_n = htons(message_id); size_t i; mg_send(nc, (char *) &message_id_n, 2); for (i = 0; i < topics_len; i++) { uint16_t topic_len_n = htons((uint16_t) strlen(topics[i])); mg_send(nc, &topic_len_n, 2); mg_send(nc, topics[i], strlen(topics[i])); } mg_mqtt_prepend_header(nc, MG_MQTT_CMD_UNSUBSCRIBE, MG_MQTT_QOS(1), nc->send_mbuf.len - old_len); } void mg_mqtt_connack(struct mg_connection *nc, uint8_t return_code) { uint8_t unused = 0; mg_send(nc, &unused, 1); mg_send(nc, &return_code, 1); mg_mqtt_prepend_header(nc, MG_MQTT_CMD_CONNACK, 0, 2); } /* * Sends a command which contains only a `message_id` and a QoS level of 1. * * Helper function. */ static void mg_send_mqtt_short_command(struct mg_connection *nc, uint8_t cmd, uint16_t message_id) { uint16_t message_id_net = htons(message_id); uint8_t flags = (cmd == MG_MQTT_CMD_PUBREL ? 2 : 0); mg_send(nc, &message_id_net, 2); mg_mqtt_prepend_header(nc, cmd, flags, 2 /* len */); } void mg_mqtt_puback(struct mg_connection *nc, uint16_t message_id) { mg_send_mqtt_short_command(nc, MG_MQTT_CMD_PUBACK, message_id); } void mg_mqtt_pubrec(struct mg_connection *nc, uint16_t message_id) { mg_send_mqtt_short_command(nc, MG_MQTT_CMD_PUBREC, message_id); } void mg_mqtt_pubrel(struct mg_connection *nc, uint16_t message_id) { mg_send_mqtt_short_command(nc, MG_MQTT_CMD_PUBREL, message_id); } void mg_mqtt_pubcomp(struct mg_connection *nc, uint16_t message_id) { mg_send_mqtt_short_command(nc, MG_MQTT_CMD_PUBCOMP, message_id); } void mg_mqtt_suback(struct mg_connection *nc, uint8_t *qoss, size_t qoss_len, uint16_t message_id) { size_t i; uint16_t message_id_net = htons(message_id); mg_send(nc, &message_id_net, 2); for (i = 0; i < qoss_len; i++) { mg_send(nc, &qoss[i], 1); } mg_mqtt_prepend_header(nc, MG_MQTT_CMD_SUBACK, MG_MQTT_QOS(1), 2 + qoss_len); } void mg_mqtt_unsuback(struct mg_connection *nc, uint16_t message_id) { mg_send_mqtt_short_command(nc, MG_MQTT_CMD_UNSUBACK, message_id); } void mg_mqtt_ping(struct mg_connection *nc) { mg_mqtt_prepend_header(nc, MG_MQTT_CMD_PINGREQ, 0, 0); } void mg_mqtt_pong(struct mg_connection *nc) { mg_mqtt_prepend_header(nc, MG_MQTT_CMD_PINGRESP, 0, 0); } void mg_mqtt_disconnect(struct mg_connection *nc) { mg_mqtt_prepend_header(nc, MG_MQTT_CMD_DISCONNECT, 0, 0); } #endif /* MG_ENABLE_MQTT */ #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_mqtt_server.c" #endif /* * Copyright (c) 2014 Cesanta Software Limited * All rights reserved */ /* Amalgamated: #include "mg_internal.h" */ /* Amalgamated: #include "mg_mqtt_server.h" */ #if MG_ENABLE_MQTT_BROKER static void mg_mqtt_session_init(struct mg_mqtt_broker *brk, struct mg_mqtt_session *s, struct mg_connection *nc) { s->brk = brk; s->subscriptions = NULL; s->num_subscriptions = 0; s->nc = nc; } static void mg_mqtt_add_session(struct mg_mqtt_session *s) { LIST_INSERT_HEAD(&s->brk->sessions, s, link); } static void mg_mqtt_remove_session(struct mg_mqtt_session *s) { LIST_REMOVE(s, link); } static void mg_mqtt_destroy_session(struct mg_mqtt_session *s) { size_t i; for (i = 0; i < s->num_subscriptions; i++) { MG_FREE((void *) s->subscriptions[i].topic); } MG_FREE(s->subscriptions); MG_FREE(s); } static void mg_mqtt_close_session(struct mg_mqtt_session *s) { mg_mqtt_remove_session(s); mg_mqtt_destroy_session(s); } void mg_mqtt_broker_init(struct mg_mqtt_broker *brk, void *user_data) { LIST_INIT(&brk->sessions); brk->user_data = user_data; } static void mg_mqtt_broker_handle_connect(struct mg_mqtt_broker *brk, struct mg_connection *nc) { struct mg_mqtt_session *s = (struct mg_mqtt_session *) MG_CALLOC(1, sizeof *s); if (s == NULL) { /* LCOV_EXCL_START */ mg_mqtt_connack(nc, MG_EV_MQTT_CONNACK_SERVER_UNAVAILABLE); return; /* LCOV_EXCL_STOP */ } /* TODO(mkm): check header (magic and version) */ mg_mqtt_session_init(brk, s, nc); nc->priv_2 = s; mg_mqtt_add_session(s); mg_mqtt_connack(nc, MG_EV_MQTT_CONNACK_ACCEPTED); } static void mg_mqtt_broker_handle_subscribe(struct mg_connection *nc, struct mg_mqtt_message *msg) { struct mg_mqtt_session *ss = (struct mg_mqtt_session *) nc->priv_2; uint8_t qoss[MG_MQTT_MAX_SESSION_SUBSCRIPTIONS]; size_t num_subs = 0; struct mg_str topic; uint8_t qos; int pos; struct mg_mqtt_topic_expression *te; for (pos = 0; (pos = mg_mqtt_next_subscribe_topic(msg, &topic, &qos, pos)) != -1;) { if (num_subs >= sizeof(MG_MQTT_MAX_SESSION_SUBSCRIPTIONS) || (ss->num_subscriptions + num_subs >= MG_MQTT_MAX_SESSION_SUBSCRIPTIONS)) { nc->flags |= MG_F_CLOSE_IMMEDIATELY; return; } qoss[num_subs++] = qos; } if (num_subs > 0) { te = (struct mg_mqtt_topic_expression *) MG_REALLOC( ss->subscriptions, sizeof(*ss->subscriptions) * (ss->num_subscriptions + num_subs)); if (te == NULL) { nc->flags |= MG_F_CLOSE_IMMEDIATELY; return; } ss->subscriptions = te; for (pos = 0; pos < (int) msg->payload.len && (pos = mg_mqtt_next_subscribe_topic(msg, &topic, &qos, pos)) != -1; ss->num_subscriptions++) { te = &ss->subscriptions[ss->num_subscriptions]; te->topic = (char *) MG_MALLOC(topic.len + 1); te->qos = qos; memcpy((char *) te->topic, topic.p, topic.len); ((char *) te->topic)[topic.len] = '\0'; } } if (pos == (int) msg->payload.len) { mg_mqtt_suback(nc, qoss, num_subs, msg->message_id); } else { /* We did not fully parse the payload, something must be wrong. */ nc->flags |= MG_F_CLOSE_IMMEDIATELY; } } static void mg_mqtt_broker_handle_publish(struct mg_mqtt_broker *brk, struct mg_mqtt_message *msg) { struct mg_mqtt_session *s; size_t i; for (s = mg_mqtt_next(brk, NULL); s != NULL; s = mg_mqtt_next(brk, s)) { for (i = 0; i < s->num_subscriptions; i++) { if (mg_mqtt_vmatch_topic_expression(s->subscriptions[i].topic, msg->topic)) { char buf[100], *p = buf; mg_asprintf(&p, sizeof(buf), "%.*s", (int) msg->topic.len, msg->topic.p); if (p == NULL) { return; } mg_mqtt_publish(s->nc, p, 0, 0, msg->payload.p, msg->payload.len); if (p != buf) { MG_FREE(p); } break; } } } } void mg_mqtt_broker(struct mg_connection *nc, int ev, void *data) { struct mg_mqtt_message *msg = (struct mg_mqtt_message *) data; struct mg_mqtt_broker *brk; if (nc->listener) { brk = (struct mg_mqtt_broker *) nc->listener->priv_2; } else { brk = (struct mg_mqtt_broker *) nc->priv_2; } switch (ev) { case MG_EV_ACCEPT: if (nc->proto_data == NULL) mg_set_protocol_mqtt(nc); nc->priv_2 = NULL; /* Clear up the inherited pointer to broker */ break; case MG_EV_MQTT_CONNECT: if (nc->priv_2 == NULL) { mg_mqtt_broker_handle_connect(brk, nc); } else { /* Repeated CONNECT */ nc->flags |= MG_F_CLOSE_IMMEDIATELY; } break; case MG_EV_MQTT_SUBSCRIBE: if (nc->priv_2 != NULL) { mg_mqtt_broker_handle_subscribe(nc, msg); } else { /* Subscribe before CONNECT */ nc->flags |= MG_F_CLOSE_IMMEDIATELY; } break; case MG_EV_MQTT_PUBLISH: if (nc->priv_2 != NULL) { mg_mqtt_broker_handle_publish(brk, msg); } else { /* Publish before CONNECT */ nc->flags |= MG_F_CLOSE_IMMEDIATELY; } break; case MG_EV_CLOSE: if (nc->listener && nc->priv_2 != NULL) { mg_mqtt_close_session((struct mg_mqtt_session *) nc->priv_2); } break; } } struct mg_mqtt_session *mg_mqtt_next(struct mg_mqtt_broker *brk, struct mg_mqtt_session *s) { return s == NULL ? LIST_FIRST(&brk->sessions) : LIST_NEXT(s, link); } #endif /* MG_ENABLE_MQTT_BROKER */ #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_dns.c" #endif /* * Copyright (c) 2014 Cesanta Software Limited * All rights reserved */ #if MG_ENABLE_DNS /* Amalgamated: #include "mg_internal.h" */ /* Amalgamated: #include "mg_dns.h" */ static int mg_dns_tid = 0xa0; struct mg_dns_header { uint16_t transaction_id; uint16_t flags; uint16_t num_questions; uint16_t num_answers; uint16_t num_authority_prs; uint16_t num_other_prs; }; struct mg_dns_resource_record *mg_dns_next_record( struct mg_dns_message *msg, int query, struct mg_dns_resource_record *prev) { struct mg_dns_resource_record *rr; for (rr = (prev == NULL ? msg->answers : prev + 1); rr - msg->answers < msg->num_answers; rr++) { if (rr->rtype == query) { return rr; } } return NULL; } int mg_dns_parse_record_data(struct mg_dns_message *msg, struct mg_dns_resource_record *rr, void *data, size_t data_len) { switch (rr->rtype) { case MG_DNS_A_RECORD: if (data_len < sizeof(struct in_addr)) { return -1; } if (rr->rdata.p + data_len > msg->pkt.p + msg->pkt.len) { return -1; } memcpy(data, rr->rdata.p, data_len); return 0; #if MG_ENABLE_IPV6 case MG_DNS_AAAA_RECORD: if (data_len < sizeof(struct in6_addr)) { return -1; /* LCOV_EXCL_LINE */ } memcpy(data, rr->rdata.p, data_len); return 0; #endif case MG_DNS_CNAME_RECORD: mg_dns_uncompress_name(msg, &rr->rdata, (char *) data, data_len); return 0; } return -1; } int mg_dns_insert_header(struct mbuf *io, size_t pos, struct mg_dns_message *msg) { struct mg_dns_header header; memset(&header, 0, sizeof(header)); header.transaction_id = msg->transaction_id; header.flags = htons(msg->flags); header.num_questions = htons(msg->num_questions); header.num_answers = htons(msg->num_answers); return mbuf_insert(io, pos, &header, sizeof(header)); } int mg_dns_copy_questions(struct mbuf *io, struct mg_dns_message *msg) { unsigned char *begin, *end; struct mg_dns_resource_record *last_q; if (msg->num_questions <= 0) return 0; begin = (unsigned char *) msg->pkt.p + sizeof(struct mg_dns_header); last_q = &msg->questions[msg->num_questions - 1]; end = (unsigned char *) last_q->name.p + last_q->name.len + 4; return mbuf_append(io, begin, end - begin); } int mg_dns_encode_name(struct mbuf *io, const char *name, size_t len) { const char *s; unsigned char n; size_t pos = io->len; do { if ((s = strchr(name, '.')) == NULL) { s = name + len; } if (s - name > 127) { return -1; /* TODO(mkm) cover */ } n = s - name; /* chunk length */ mbuf_append(io, &n, 1); /* send length */ mbuf_append(io, name, n); if (*s == '.') { n++; } name += n; len -= n; } while (*s != '\0'); mbuf_append(io, "\0", 1); /* Mark end of host name */ return io->len - pos; } int mg_dns_encode_record(struct mbuf *io, struct mg_dns_resource_record *rr, const char *name, size_t nlen, const void *rdata, size_t rlen) { size_t pos = io->len; uint16_t u16; uint32_t u32; if (rr->kind == MG_DNS_INVALID_RECORD) { return -1; /* LCOV_EXCL_LINE */ } if (mg_dns_encode_name(io, name, nlen) == -1) { return -1; } u16 = htons(rr->rtype); mbuf_append(io, &u16, 2); u16 = htons(rr->rclass); mbuf_append(io, &u16, 2); if (rr->kind == MG_DNS_ANSWER) { u32 = htonl(rr->ttl); mbuf_append(io, &u32, 4); if (rr->rtype == MG_DNS_CNAME_RECORD) { int clen; /* fill size after encoding */ size_t off = io->len; mbuf_append(io, &u16, 2); if ((clen = mg_dns_encode_name(io, (const char *) rdata, rlen)) == -1) { return -1; } u16 = clen; io->buf[off] = u16 >> 8; io->buf[off + 1] = u16 & 0xff; } else { u16 = htons((uint16_t) rlen); mbuf_append(io, &u16, 2); mbuf_append(io, rdata, rlen); } } return io->len - pos; } void mg_send_dns_query(struct mg_connection *nc, const char *name, int query_type) { struct mg_dns_message *msg = (struct mg_dns_message *) MG_CALLOC(1, sizeof(*msg)); struct mbuf pkt; struct mg_dns_resource_record *rr = &msg->questions[0]; DBG(("%s %d", name, query_type)); mbuf_init(&pkt, 64 /* Start small, it'll grow as needed. */); msg->transaction_id = ++mg_dns_tid; msg->flags = 0x100; msg->num_questions = 1; mg_dns_insert_header(&pkt, 0, msg); rr->rtype = query_type; rr->rclass = 1; /* Class: inet */ rr->kind = MG_DNS_QUESTION; if (mg_dns_encode_record(&pkt, rr, name, strlen(name), NULL, 0) == -1) { /* TODO(mkm): return an error code */ goto cleanup; /* LCOV_EXCL_LINE */ } /* TCP DNS requires messages to be prefixed with len */ if (!(nc->flags & MG_F_UDP)) { uint16_t len = htons((uint16_t) pkt.len); mbuf_insert(&pkt, 0, &len, 2); } mg_send(nc, pkt.buf, pkt.len); mbuf_free(&pkt); cleanup: MG_FREE(msg); } static unsigned char *mg_parse_dns_resource_record( unsigned char *data, unsigned char *end, struct mg_dns_resource_record *rr, int reply) { unsigned char *name = data; int chunk_len, data_len; while (data < end && (chunk_len = *data)) { if (((unsigned char *) data)[0] & 0xc0) { data += 1; break; } data += chunk_len + 1; } if (data > end - 5) { return NULL; } rr->name.p = (char *) name; rr->name.len = data - name + 1; data++; rr->rtype = data[0] << 8 | data[1]; data += 2; rr->rclass = data[0] << 8 | data[1]; data += 2; rr->kind = reply ? MG_DNS_ANSWER : MG_DNS_QUESTION; if (reply) { if (data >= end - 6) { return NULL; } rr->ttl = (uint32_t) data[0] << 24 | (uint32_t) data[1] << 16 | data[2] << 8 | data[3]; data += 4; data_len = *data << 8 | *(data + 1); data += 2; rr->rdata.p = (char *) data; rr->rdata.len = data_len; data += data_len; } return data; } int mg_parse_dns(const char *buf, int len, struct mg_dns_message *msg) { struct mg_dns_header *header = (struct mg_dns_header *) buf; unsigned char *data = (unsigned char *) buf + sizeof(*header); unsigned char *end = (unsigned char *) buf + len; int i; memset(msg, 0, sizeof(*msg)); msg->pkt.p = buf; msg->pkt.len = len; if (len < (int) sizeof(*header)) return -1; msg->transaction_id = header->transaction_id; msg->flags = ntohs(header->flags); msg->num_questions = ntohs(header->num_questions); if (msg->num_questions > (int) ARRAY_SIZE(msg->questions)) { msg->num_questions = (int) ARRAY_SIZE(msg->questions); } msg->num_answers = ntohs(header->num_answers); if (msg->num_answers > (int) ARRAY_SIZE(msg->answers)) { msg->num_answers = (int) ARRAY_SIZE(msg->answers); } for (i = 0; i < msg->num_questions; i++) { data = mg_parse_dns_resource_record(data, end, &msg->questions[i], 0); if (data == NULL) return -1; } for (i = 0; i < msg->num_answers; i++) { data = mg_parse_dns_resource_record(data, end, &msg->answers[i], 1); if (data == NULL) return -1; } return 0; } size_t mg_dns_uncompress_name(struct mg_dns_message *msg, struct mg_str *name, char *dst, int dst_len) { int chunk_len, num_ptrs = 0; char *old_dst = dst; const unsigned char *data = (unsigned char *) name->p; const unsigned char *end = (unsigned char *) msg->pkt.p + msg->pkt.len; if (data >= end) { return 0; } while ((chunk_len = *data++)) { int leeway = dst_len - (dst - old_dst); if (data >= end) { return 0; } if ((chunk_len & 0xc0) == 0xc0) { uint16_t off = (data[-1] & (~0xc0)) << 8 | data[0]; if (off >= msg->pkt.len) { return 0; } /* Basic circular loop avoidance: allow up to 16 pointer hops. */ if (++num_ptrs > 15) { return 0; } data = (unsigned char *) msg->pkt.p + off; continue; } if (chunk_len > 63) { return 0; } if (chunk_len > leeway) { chunk_len = leeway; } if (data + chunk_len >= end) { return 0; } memcpy(dst, data, chunk_len); data += chunk_len; dst += chunk_len; leeway -= chunk_len; if (leeway == 0) { return dst - old_dst; } *dst++ = '.'; } if (dst != old_dst) { *--dst = 0; } return dst - old_dst; } static void dns_handler(struct mg_connection *nc, int ev, void *ev_data MG_UD_ARG(void *user_data)) { struct mbuf *io = &nc->recv_mbuf; struct mg_dns_message msg; /* Pass low-level events to the user handler */ nc->handler(nc, ev, ev_data MG_UD_ARG(user_data)); switch (ev) { case MG_EV_RECV: if (!(nc->flags & MG_F_UDP)) { mbuf_remove(&nc->recv_mbuf, 2); } if (mg_parse_dns(nc->recv_mbuf.buf, nc->recv_mbuf.len, &msg) == -1) { /* reply + recursion allowed + format error */ memset(&msg, 0, sizeof(msg)); msg.flags = 0x8081; mg_dns_insert_header(io, 0, &msg); if (!(nc->flags & MG_F_UDP)) { uint16_t len = htons((uint16_t) io->len); mbuf_insert(io, 0, &len, 2); } mg_send(nc, io->buf, io->len); } else { /* Call user handler with parsed message */ nc->handler(nc, MG_DNS_MESSAGE, &msg MG_UD_ARG(user_data)); } mbuf_remove(io, io->len); break; } } void mg_set_protocol_dns(struct mg_connection *nc) { nc->proto_handler = dns_handler; } #endif /* MG_ENABLE_DNS */ #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_dns_server.c" #endif /* * Copyright (c) 2014 Cesanta Software Limited * All rights reserved */ #if MG_ENABLE_DNS_SERVER /* Amalgamated: #include "mg_internal.h" */ /* Amalgamated: #include "dns-server.h" */ struct mg_dns_reply mg_dns_create_reply(struct mbuf *io, struct mg_dns_message *msg) { struct mg_dns_reply rep; rep.msg = msg; rep.io = io; rep.start = io->len; /* reply + recursion allowed */ msg->flags |= 0x8080; mg_dns_copy_questions(io, msg); msg->num_answers = 0; return rep; } void mg_dns_send_reply(struct mg_connection *nc, struct mg_dns_reply *r) { size_t sent = r->io->len - r->start; mg_dns_insert_header(r->io, r->start, r->msg); if (!(nc->flags & MG_F_UDP)) { uint16_t len = htons((uint16_t) sent); mbuf_insert(r->io, r->start, &len, 2); } if (&nc->send_mbuf != r->io) { mg_send(nc, r->io->buf + r->start, r->io->len - r->start); r->io->len = r->start; } } int mg_dns_reply_record(struct mg_dns_reply *reply, struct mg_dns_resource_record *question, const char *name, int rtype, int ttl, const void *rdata, size_t rdata_len) { struct mg_dns_message *msg = (struct mg_dns_message *) reply->msg; char rname[512]; struct mg_dns_resource_record *ans = &msg->answers[msg->num_answers]; if (msg->num_answers >= MG_MAX_DNS_ANSWERS) { return -1; /* LCOV_EXCL_LINE */ } if (name == NULL) { name = rname; rname[511] = 0; mg_dns_uncompress_name(msg, &question->name, rname, sizeof(rname) - 1); } *ans = *question; ans->kind = MG_DNS_ANSWER; ans->rtype = rtype; ans->ttl = ttl; if (mg_dns_encode_record(reply->io, ans, name, strlen(name), rdata, rdata_len) == -1) { return -1; /* LCOV_EXCL_LINE */ }; msg->num_answers++; return 0; } #endif /* MG_ENABLE_DNS_SERVER */ #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_resolv.c" #endif /* * Copyright (c) 2014 Cesanta Software Limited * All rights reserved */ #if MG_ENABLE_ASYNC_RESOLVER /* Amalgamated: #include "mg_internal.h" */ /* Amalgamated: #include "mg_resolv.h" */ #ifndef MG_DEFAULT_NAMESERVER #define MG_DEFAULT_NAMESERVER "8.8.8.8" #endif struct mg_resolve_async_request { char name[1024]; int query; mg_resolve_callback_t callback; void *data; time_t timeout; int max_retries; enum mg_resolve_err err; /* state */ time_t last_time; int retries; }; /* * Find what nameserver to use. * * Return 0 if OK, -1 if error */ static int mg_get_ip_address_of_nameserver(char *name, size_t name_len) { int ret = -1; #ifdef _WIN32 int i; LONG err; HKEY hKey, hSub; wchar_t subkey[512], value[128], *key = L"SYSTEM\\ControlSet001\\Services\\Tcpip\\Parameters\\Interfaces"; if ((err = RegOpenKeyExW(HKEY_LOCAL_MACHINE, key, 0, KEY_READ, &hKey)) != ERROR_SUCCESS) { fprintf(stderr, "cannot open reg key %S: %ld\n", key, err); ret = -1; } else { for (ret = -1, i = 0; 1; i++) { DWORD subkey_size = sizeof(subkey), type, len = sizeof(value); if (RegEnumKeyExW(hKey, i, subkey, &subkey_size, NULL, NULL, NULL, NULL) != ERROR_SUCCESS) { break; } if (RegOpenKeyExW(hKey, subkey, 0, KEY_READ, &hSub) == ERROR_SUCCESS && ((RegQueryValueExW(hSub, L"NameServer", 0, &type, (void *) value, &len) == ERROR_SUCCESS && value[0] != '\0') || (RegQueryValueExW(hSub, L"DhcpNameServer", 0, &type, (void *) value, &len) == ERROR_SUCCESS && value[0] != '\0'))) { /* * See https://github.com/cesanta/mongoose/issues/176 * The value taken from the registry can be empty, a single * IP address, or multiple IP addresses separated by comma. * If it's empty, check the next interface. * If it's multiple IP addresses, take the first one. */ wchar_t *comma = wcschr(value, ','); if (comma != NULL) { *comma = '\0'; } /* %S will convert wchar_t -> char */ snprintf(name, name_len, "%S", value); ret = 0; RegCloseKey(hSub); break; } } RegCloseKey(hKey); } #elif MG_ENABLE_FILESYSTEM && defined(MG_RESOLV_CONF_FILE_NAME) FILE *fp; char line[512]; if ((fp = mg_fopen(MG_RESOLV_CONF_FILE_NAME, "r")) == NULL) { ret = -1; } else { /* Try to figure out what nameserver to use */ for (ret = -1; fgets(line, sizeof(line), fp) != NULL;) { unsigned int a, b, c, d; if (sscanf(line, "nameserver %u.%u.%u.%u", &a, &b, &c, &d) == 4) { snprintf(name, name_len, "%u.%u.%u.%u", a, b, c, d); ret = 0; break; } } (void) fclose(fp); } #else snprintf(name, name_len, "%s", MG_DEFAULT_NAMESERVER); #endif /* _WIN32 */ return ret; } int mg_resolve_from_hosts_file(const char *name, union socket_address *usa) { #if MG_ENABLE_FILESYSTEM && defined(MG_HOSTS_FILE_NAME) /* TODO(mkm) cache /etc/hosts */ FILE *fp; char line[1024]; char *p; char alias[256]; unsigned int a, b, c, d; int len = 0; if ((fp = mg_fopen(MG_HOSTS_FILE_NAME, "r")) == NULL) { return -1; } for (; fgets(line, sizeof(line), fp) != NULL;) { if (line[0] == '#') continue; if (sscanf(line, "%u.%u.%u.%u%n", &a, &b, &c, &d, &len) == 0) { /* TODO(mkm): handle ipv6 */ continue; } for (p = line + len; sscanf(p, "%s%n", alias, &len) == 1; p += len) { if (strcmp(alias, name) == 0) { usa->sin.sin_addr.s_addr = htonl(a << 24 | b << 16 | c << 8 | d); fclose(fp); return 0; } } } fclose(fp); #else (void) name; (void) usa; #endif return -1; } static void mg_resolve_async_eh(struct mg_connection *nc, int ev, void *data MG_UD_ARG(void *user_data)) { time_t now = (time_t) mg_time(); struct mg_resolve_async_request *req; struct mg_dns_message *msg; int first = 0; #if !MG_ENABLE_CALLBACK_USERDATA void *user_data = nc->user_data; #endif if (ev != MG_EV_POLL) DBG(("ev=%d user_data=%p", ev, user_data)); req = (struct mg_resolve_async_request *) user_data; if (req == NULL) { return; } switch (ev) { case MG_EV_CONNECT: /* don't depend on timer not being at epoch for sending out first req */ first = 1; /* fallthrough */ case MG_EV_POLL: if (req->retries > req->max_retries) { req->err = MG_RESOLVE_EXCEEDED_RETRY_COUNT; nc->flags |= MG_F_CLOSE_IMMEDIATELY; break; } if (first || now - req->last_time >= req->timeout) { mg_send_dns_query(nc, req->name, req->query); req->last_time = now; req->retries++; } break; case MG_EV_RECV: msg = (struct mg_dns_message *) MG_MALLOC(sizeof(*msg)); if (mg_parse_dns(nc->recv_mbuf.buf, *(int *) data, msg) == 0 && msg->num_answers > 0) { req->callback(msg, req->data, MG_RESOLVE_OK); nc->user_data = NULL; MG_FREE(req); } else { req->err = MG_RESOLVE_NO_ANSWERS; } MG_FREE(msg); nc->flags |= MG_F_CLOSE_IMMEDIATELY; break; case MG_EV_SEND: /* * If a send error occurs, prevent closing of the connection by the core. * We will retry after timeout. */ nc->flags &= ~MG_F_CLOSE_IMMEDIATELY; mbuf_remove(&nc->send_mbuf, nc->send_mbuf.len); break; case MG_EV_TIMER: req->err = MG_RESOLVE_TIMEOUT; nc->flags |= MG_F_CLOSE_IMMEDIATELY; break; case MG_EV_CLOSE: /* If we got here with request still not done, fire an error callback. */ if (req != NULL) { char addr[32]; mg_sock_addr_to_str(&nc->sa, addr, sizeof(addr), MG_SOCK_STRINGIFY_IP); #ifdef MG_LOG_DNS_FAILURES LOG(LL_ERROR, ("Failed to resolve '%s', server %s", req->name, addr)); #endif req->callback(NULL, req->data, req->err); nc->user_data = NULL; MG_FREE(req); } break; } } int mg_resolve_async(struct mg_mgr *mgr, const char *name, int query, mg_resolve_callback_t cb, void *data) { struct mg_resolve_async_opts opts; memset(&opts, 0, sizeof(opts)); return mg_resolve_async_opt(mgr, name, query, cb, data, opts); } int mg_resolve_async_opt(struct mg_mgr *mgr, const char *name, int query, mg_resolve_callback_t cb, void *data, struct mg_resolve_async_opts opts) { struct mg_resolve_async_request *req; struct mg_connection *dns_nc; const char *nameserver = opts.nameserver; char dns_server_buff[17], nameserver_url[26]; if (nameserver == NULL) { nameserver = mgr->nameserver; } DBG(("%s %d %p", name, query, opts.dns_conn)); /* resolve with DNS */ req = (struct mg_resolve_async_request *) MG_CALLOC(1, sizeof(*req)); if (req == NULL) { return -1; } strncpy(req->name, name, sizeof(req->name)); req->name[sizeof(req->name) - 1] = '\0'; req->query = query; req->callback = cb; req->data = data; /* TODO(mkm): parse defaults out of resolve.conf */ req->max_retries = opts.max_retries ? opts.max_retries : 2; req->timeout = opts.timeout ? opts.timeout : 5; /* Lazily initialize dns server */ if (nameserver == NULL) { if (mg_get_ip_address_of_nameserver(dns_server_buff, sizeof(dns_server_buff)) != -1) { nameserver = dns_server_buff; } else { nameserver = MG_DEFAULT_NAMESERVER; } } snprintf(nameserver_url, sizeof(nameserver_url), "udp://%s:53", nameserver); dns_nc = mg_connect(mgr, nameserver_url, MG_CB(mg_resolve_async_eh, NULL)); if (dns_nc == NULL) { MG_FREE(req); return -1; } dns_nc->user_data = req; if (opts.dns_conn != NULL) { *opts.dns_conn = dns_nc; } return 0; } void mg_set_nameserver(struct mg_mgr *mgr, const char *nameserver) { MG_FREE((char *) mgr->nameserver); mgr->nameserver = NULL; if (nameserver != NULL) { mgr->nameserver = strdup(nameserver); } } #endif /* MG_ENABLE_ASYNC_RESOLVER */ #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_coap.c" #endif /* * Copyright (c) 2015 Cesanta Software Limited * All rights reserved * This software is dual-licensed: you can redistribute it and/or modify * it under the terms of the GNU General Public License version 2 as * published by the Free Software Foundation. For the terms of this * license, see . * * You are free to use this software under the terms of the GNU General * Public License, but WITHOUT ANY WARRANTY; without even the implied * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. * See the GNU General Public License for more details. * * Alternatively, you can license this software under a commercial * license, as set out in . */ /* Amalgamated: #include "mg_internal.h" */ /* Amalgamated: #include "mg_coap.h" */ #if MG_ENABLE_COAP void mg_coap_free_options(struct mg_coap_message *cm) { while (cm->options != NULL) { struct mg_coap_option *next = cm->options->next; MG_FREE(cm->options); cm->options = next; } } struct mg_coap_option *mg_coap_add_option(struct mg_coap_message *cm, uint32_t number, char *value, size_t len) { struct mg_coap_option *new_option = (struct mg_coap_option *) MG_CALLOC(1, sizeof(*new_option)); new_option->number = number; new_option->value.p = value; new_option->value.len = len; if (cm->options == NULL) { cm->options = cm->optiomg_tail = new_option; } else { /* * A very simple attention to help clients to compose options: * CoAP wants to see options ASC ordered. * Could be change by using sort in coap_compose */ if (cm->optiomg_tail->number <= new_option->number) { /* if option is already ordered just add it */ cm->optiomg_tail = cm->optiomg_tail->next = new_option; } else { /* looking for appropriate position */ struct mg_coap_option *current_opt = cm->options; struct mg_coap_option *prev_opt = 0; while (current_opt != NULL) { if (current_opt->number > new_option->number) { break; } prev_opt = current_opt; current_opt = current_opt->next; } if (prev_opt != NULL) { prev_opt->next = new_option; new_option->next = current_opt; } else { /* insert new_option to the beginning */ new_option->next = cm->options; cm->options = new_option; } } } return new_option; } /* * Fills CoAP header in mg_coap_message. * * Helper function. */ static char *coap_parse_header(char *ptr, struct mbuf *io, struct mg_coap_message *cm) { if (io->len < sizeof(uint32_t)) { cm->flags |= MG_COAP_NOT_ENOUGH_DATA; return NULL; } /* * Version (Ver): 2-bit unsigned integer. Indicates the CoAP version * number. Implementations of this specification MUST set this field * to 1 (01 binary). Other values are reserved for future versions. * Messages with unknown version numbers MUST be silently ignored. */ if (((uint8_t) *ptr >> 6) != 1) { cm->flags |= MG_COAP_IGNORE; return NULL; } /* * Type (T): 2-bit unsigned integer. Indicates if this message is of * type Confirmable (0), Non-confirmable (1), Acknowledgement (2), or * Reset (3). */ cm->msg_type = ((uint8_t) *ptr & 0x30) >> 4; cm->flags |= MG_COAP_MSG_TYPE_FIELD; /* * Token Length (TKL): 4-bit unsigned integer. Indicates the length of * the variable-length Token field (0-8 bytes). Lengths 9-15 are * reserved, MUST NOT be sent, and MUST be processed as a message * format error. */ cm->token.len = *ptr & 0x0F; if (cm->token.len > 8) { cm->flags |= MG_COAP_FORMAT_ERROR; return NULL; } ptr++; /* * Code: 8-bit unsigned integer, split into a 3-bit class (most * significant bits) and a 5-bit detail (least significant bits) */ cm->code_class = (uint8_t) *ptr >> 5; cm->code_detail = *ptr & 0x1F; cm->flags |= (MG_COAP_CODE_CLASS_FIELD | MG_COAP_CODE_DETAIL_FIELD); ptr++; /* Message ID: 16-bit unsigned integer in network byte order. */ cm->msg_id = (uint8_t) *ptr << 8 | (uint8_t) * (ptr + 1); cm->flags |= MG_COAP_MSG_ID_FIELD; ptr += 2; return ptr; } /* * Fills token information in mg_coap_message. * * Helper function. */ static char *coap_get_token(char *ptr, struct mbuf *io, struct mg_coap_message *cm) { if (cm->token.len != 0) { if (ptr + cm->token.len > io->buf + io->len) { cm->flags |= MG_COAP_NOT_ENOUGH_DATA; return NULL; } else { cm->token.p = ptr; ptr += cm->token.len; cm->flags |= MG_COAP_TOKEN_FIELD; } } return ptr; } /* * Returns Option Delta or Length. * * Helper function. */ static int coap_get_ext_opt(char *ptr, struct mbuf *io, uint16_t *opt_info) { int ret = 0; if (*opt_info == 13) { /* * 13: An 8-bit unsigned integer follows the initial byte and * indicates the Option Delta/Length minus 13. */ if (ptr < io->buf + io->len) { *opt_info = (uint8_t) *ptr + 13; ret = sizeof(uint8_t); } else { ret = -1; /* LCOV_EXCL_LINE */ } } else if (*opt_info == 14) { /* * 14: A 16-bit unsigned integer in network byte order follows the * initial byte and indicates the Option Delta/Length minus 269. */ if (ptr + sizeof(uint8_t) < io->buf + io->len) { *opt_info = ((uint8_t) *ptr << 8 | (uint8_t) * (ptr + 1)) + 269; ret = sizeof(uint16_t); } else { ret = -1; /* LCOV_EXCL_LINE */ } } return ret; } /* * Fills options in mg_coap_message. * * Helper function. * * General options format: * +---------------+---------------+ * | Option Delta | Option Length | 1 byte * +---------------+---------------+ * \ Option Delta (extended) \ 0-2 bytes * +-------------------------------+ * / Option Length (extended) \ 0-2 bytes * +-------------------------------+ * \ Option Value \ 0 or more bytes * +-------------------------------+ */ static char *coap_get_options(char *ptr, struct mbuf *io, struct mg_coap_message *cm) { uint16_t prev_opt = 0; if (ptr == io->buf + io->len) { /* end of packet, ok */ return NULL; } /* 0xFF is payload marker */ while (ptr < io->buf + io->len && (uint8_t) *ptr != 0xFF) { uint16_t option_delta, option_lenght; int optinfo_len; /* Option Delta: 4-bit unsigned integer */ option_delta = ((uint8_t) *ptr & 0xF0) >> 4; /* Option Length: 4-bit unsigned integer */ option_lenght = *ptr & 0x0F; if (option_delta == 15 || option_lenght == 15) { /* * 15: Reserved for future use. If the field is set to this value, * it MUST be processed as a message format error */ cm->flags |= MG_COAP_FORMAT_ERROR; break; } ptr++; /* check for extended option delta */ optinfo_len = coap_get_ext_opt(ptr, io, &option_delta); if (optinfo_len == -1) { cm->flags |= MG_COAP_NOT_ENOUGH_DATA; /* LCOV_EXCL_LINE */ break; /* LCOV_EXCL_LINE */ } ptr += optinfo_len; /* check or extended option lenght */ optinfo_len = coap_get_ext_opt(ptr, io, &option_lenght); if (optinfo_len == -1) { cm->flags |= MG_COAP_NOT_ENOUGH_DATA; /* LCOV_EXCL_LINE */ break; /* LCOV_EXCL_LINE */ } ptr += optinfo_len; /* * Instead of specifying the Option Number directly, the instances MUST * appear in order of their Option Numbers and a delta encoding is used * between them. */ option_delta += prev_opt; mg_coap_add_option(cm, option_delta, ptr, option_lenght); prev_opt = option_delta; if (ptr + option_lenght > io->buf + io->len) { cm->flags |= MG_COAP_NOT_ENOUGH_DATA; /* LCOV_EXCL_LINE */ break; /* LCOV_EXCL_LINE */ } ptr += option_lenght; } if ((cm->flags & MG_COAP_ERROR) != 0) { mg_coap_free_options(cm); return NULL; } cm->flags |= MG_COAP_OPTIOMG_FIELD; if (ptr == io->buf + io->len) { /* end of packet, ok */ return NULL; } ptr++; return ptr; } uint32_t mg_coap_parse(struct mbuf *io, struct mg_coap_message *cm) { char *ptr; memset(cm, 0, sizeof(*cm)); if ((ptr = coap_parse_header(io->buf, io, cm)) == NULL) { return cm->flags; } if ((ptr = coap_get_token(ptr, io, cm)) == NULL) { return cm->flags; } if ((ptr = coap_get_options(ptr, io, cm)) == NULL) { return cm->flags; } /* the rest is payload */ cm->payload.len = io->len - (ptr - io->buf); if (cm->payload.len != 0) { cm->payload.p = ptr; cm->flags |= MG_COAP_PAYLOAD_FIELD; } return cm->flags; } /* * Calculates extended size of given Opt Number/Length in coap message. * * Helper function. */ static size_t coap_get_ext_opt_size(uint32_t value) { int ret = 0; if (value >= 13 && value <= 0xFF + 13) { ret = sizeof(uint8_t); } else if (value > 0xFF + 13 && value <= 0xFFFF + 269) { ret = sizeof(uint16_t); } return ret; } /* * Splits given Opt Number/Length into base and ext values. * * Helper function. */ static int coap_split_opt(uint32_t value, uint8_t *base, uint16_t *ext) { int ret = 0; if (value < 13) { *base = value; } else if (value >= 13 && value <= 0xFF + 13) { *base = 13; *ext = value - 13; ret = sizeof(uint8_t); } else if (value > 0xFF + 13 && value <= 0xFFFF + 269) { *base = 14; *ext = value - 269; ret = sizeof(uint16_t); } return ret; } /* * Puts uint16_t (in network order) into given char stream. * * Helper function. */ static char *coap_add_uint16(char *ptr, uint16_t val) { *ptr = val >> 8; ptr++; *ptr = val & 0x00FF; ptr++; return ptr; } /* * Puts extended value of Opt Number/Length into given char stream. * * Helper function. */ static char *coap_add_opt_info(char *ptr, uint16_t val, size_t len) { if (len == sizeof(uint8_t)) { *ptr = (char) val; ptr++; } else if (len == sizeof(uint16_t)) { ptr = coap_add_uint16(ptr, val); } return ptr; } /* * Verifies given mg_coap_message and calculates message size for it. * * Helper function. */ static uint32_t coap_calculate_packet_size(struct mg_coap_message *cm, size_t *len) { struct mg_coap_option *opt; uint32_t prev_opt_number; *len = 4; /* header */ if (cm->msg_type > MG_COAP_MSG_MAX) { return MG_COAP_ERROR | MG_COAP_MSG_TYPE_FIELD; } if (cm->token.len > 8) { return MG_COAP_ERROR | MG_COAP_TOKEN_FIELD; } if (cm->code_class > 7) { return MG_COAP_ERROR | MG_COAP_CODE_CLASS_FIELD; } if (cm->code_detail > 31) { return MG_COAP_ERROR | MG_COAP_CODE_DETAIL_FIELD; } *len += cm->token.len; if (cm->payload.len != 0) { *len += cm->payload.len + 1; /* ... + 1; add payload marker */ } opt = cm->options; prev_opt_number = 0; while (opt != NULL) { *len += 1; /* basic delta/length */ *len += coap_get_ext_opt_size(opt->number - prev_opt_number); *len += coap_get_ext_opt_size((uint32_t) opt->value.len); /* * Current implementation performs check if * option_number > previous option_number and produces an error * TODO(alashkin): write design doc with limitations * May be resorting is more suitable solution. */ if ((opt->next != NULL && opt->number > opt->next->number) || opt->value.len > 0xFFFF + 269 || opt->number - prev_opt_number > 0xFFFF + 269) { return MG_COAP_ERROR | MG_COAP_OPTIOMG_FIELD; } *len += opt->value.len; prev_opt_number = opt->number; opt = opt->next; } return 0; } uint32_t mg_coap_compose(struct mg_coap_message *cm, struct mbuf *io) { struct mg_coap_option *opt; uint32_t res, prev_opt_number; size_t prev_io_len, packet_size; char *ptr; res = coap_calculate_packet_size(cm, &packet_size); if (res != 0) { return res; } /* saving previous lenght to handle non-empty mbuf */ prev_io_len = io->len; if (mbuf_append(io, NULL, packet_size) == 0) return MG_COAP_ERROR; ptr = io->buf + prev_io_len; /* * since cm is verified, it is possible to use bits shift operator * without additional zeroing of unused bits */ /* ver: 2 bits, msg_type: 2 bits, toklen: 4 bits */ *ptr = (1 << 6) | (cm->msg_type << 4) | (uint8_t)(cm->token.len); ptr++; /* code class: 3 bits, code detail: 5 bits */ *ptr = (cm->code_class << 5) | (cm->code_detail); ptr++; ptr = coap_add_uint16(ptr, cm->msg_id); if (cm->token.len != 0) { memcpy(ptr, cm->token.p, cm->token.len); ptr += cm->token.len; } opt = cm->options; prev_opt_number = 0; while (opt != NULL) { uint8_t delta_base = 0, length_base = 0; uint16_t delta_ext = 0, length_ext = 0; size_t opt_delta_len = coap_split_opt(opt->number - prev_opt_number, &delta_base, &delta_ext); size_t opt_lenght_len = coap_split_opt((uint32_t) opt->value.len, &length_base, &length_ext); *ptr = (delta_base << 4) | length_base; ptr++; ptr = coap_add_opt_info(ptr, delta_ext, opt_delta_len); ptr = coap_add_opt_info(ptr, length_ext, opt_lenght_len); if (opt->value.len != 0) { memcpy(ptr, opt->value.p, opt->value.len); ptr += opt->value.len; } prev_opt_number = opt->number; opt = opt->next; } if (cm->payload.len != 0) { *ptr = (char) -1; ptr++; memcpy(ptr, cm->payload.p, cm->payload.len); } return 0; } uint32_t mg_coap_send_message(struct mg_connection *nc, struct mg_coap_message *cm) { struct mbuf packet_out; uint32_t compose_res; mbuf_init(&packet_out, 0); compose_res = mg_coap_compose(cm, &packet_out); if (compose_res != 0) { return compose_res; /* LCOV_EXCL_LINE */ } mg_send(nc, packet_out.buf, (int) packet_out.len); mbuf_free(&packet_out); return 0; } uint32_t mg_coap_send_ack(struct mg_connection *nc, uint16_t msg_id) { struct mg_coap_message cm; memset(&cm, 0, sizeof(cm)); cm.msg_type = MG_COAP_MSG_ACK; cm.msg_id = msg_id; return mg_coap_send_message(nc, &cm); } static void coap_handler(struct mg_connection *nc, int ev, void *ev_data MG_UD_ARG(void *user_data)) { struct mbuf *io = &nc->recv_mbuf; struct mg_coap_message cm; uint32_t parse_res; memset(&cm, 0, sizeof(cm)); nc->handler(nc, ev, ev_data MG_UD_ARG(user_data)); switch (ev) { case MG_EV_RECV: parse_res = mg_coap_parse(io, &cm); if ((parse_res & MG_COAP_IGNORE) == 0) { if ((cm.flags & MG_COAP_NOT_ENOUGH_DATA) != 0) { /* * Since we support UDP only * MG_COAP_NOT_ENOUGH_DATA == MG_COAP_FORMAT_ERROR */ cm.flags |= MG_COAP_FORMAT_ERROR; /* LCOV_EXCL_LINE */ } /* LCOV_EXCL_LINE */ nc->handler(nc, MG_COAP_EVENT_BASE + cm.msg_type, &cm MG_UD_ARG(user_data)); } mg_coap_free_options(&cm); mbuf_remove(io, io->len); break; } } /* * Attach built-in CoAP event handler to the given connection. * * The user-defined event handler will receive following extra events: * * - MG_EV_COAP_CON * - MG_EV_COAP_NOC * - MG_EV_COAP_ACK * - MG_EV_COAP_RST */ int mg_set_protocol_coap(struct mg_connection *nc) { /* supports UDP only */ if ((nc->flags & MG_F_UDP) == 0) { return -1; } nc->proto_handler = coap_handler; return 0; } #endif /* MG_ENABLE_COAP */ #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_sntp.c" #endif /* * Copyright (c) 2016 Cesanta Software Limited * All rights reserved */ /* Amalgamated: #include "mg_internal.h" */ /* Amalgamated: #include "mg_sntp.h" */ /* Amalgamated: #include "mg_util.h" */ #if MG_ENABLE_SNTP #define SNTP_TIME_OFFSET 2208988800 #ifndef SNTP_TIMEOUT #define SNTP_TIMEOUT 10 #endif #ifndef SNTP_ATTEMPTS #define SNTP_ATTEMPTS 3 #endif static uint64_t mg_get_sec(uint64_t val) { return (val & 0xFFFFFFFF00000000) >> 32; } static uint64_t mg_get_usec(uint64_t val) { uint64_t tmp = (val & 0x00000000FFFFFFFF); tmp *= 1000000; tmp >>= 32; return tmp; } static void mg_ntp_to_tv(uint64_t val, struct timeval *tv) { uint64_t tmp; tmp = mg_get_sec(val); tmp -= SNTP_TIME_OFFSET; tv->tv_sec = tmp; tv->tv_usec = mg_get_usec(val); } static void mg_get_ntp_ts(const char *ntp, uint64_t *val) { uint32_t tmp; memcpy(&tmp, ntp, sizeof(tmp)); tmp = ntohl(tmp); *val = (uint64_t) tmp << 32; memcpy(&tmp, ntp + 4, sizeof(tmp)); tmp = ntohl(tmp); *val |= tmp; } void mg_sntp_send_request(struct mg_connection *c) { uint8_t buf[48] = {0}; /* * header - 8 bit: * LI (2 bit) - 3 (not in sync), VN (3 bit) - 4 (version), * mode (3 bit) - 3 (client) */ buf[0] = (3 << 6) | (4 << 3) | 3; /* * Next fields should be empty in client request * stratum, 8 bit * poll interval, 8 bit * rrecision, 8 bit * root delay, 32 bit * root dispersion, 32 bit * ref id, 32 bit * ref timestamp, 64 bit * originate Timestamp, 64 bit * receive Timestamp, 64 bit */ /* * convert time to sntp format (sntp starts from 00:00:00 01.01.1900) * according to rfc868 it is 2208988800L sec * this information is used to correct roundtrip delay * but if local clock is absolutely broken (and doesn't work even * as simple timer), it is better to disable it */ #ifndef MG_SNTP_NO_DELAY_CORRECTION uint32_t sec; sec = htonl((uint32_t)(mg_time() + SNTP_TIME_OFFSET)); memcpy(&buf[40], &sec, sizeof(sec)); #endif mg_send(c, buf, sizeof(buf)); } #ifndef MG_SNTP_NO_DELAY_CORRECTION static uint64_t mg_calculate_delay(uint64_t t1, uint64_t t2, uint64_t t3) { /* roundloop delay = (T4 - T1) - (T3 - T2) */ uint64_t d1 = ((mg_time() + SNTP_TIME_OFFSET) * 1000000) - (mg_get_sec(t1) * 1000000 + mg_get_usec(t1)); uint64_t d2 = (mg_get_sec(t3) * 1000000 + mg_get_usec(t3)) - (mg_get_sec(t2) * 1000000 + mg_get_usec(t2)); return (d1 > d2) ? d1 - d2 : 0; } #endif MG_INTERNAL int mg_sntp_parse_reply(const char *buf, int len, struct mg_sntp_message *msg) { uint8_t hdr; uint64_t trsm_ts_T3, delay = 0; int mode; struct timeval tv; if (len < 48) { return -1; } hdr = buf[0]; if ((hdr & 0x38) >> 3 != 4) { /* Wrong version */ return -1; } mode = hdr & 0x7; if (mode != 4 && mode != 5) { /* Not a server reply */ return -1; } memset(msg, 0, sizeof(*msg)); msg->kiss_of_death = (buf[1] == 0); /* Server asks to not send requests */ mg_get_ntp_ts(&buf[40], &trsm_ts_T3); #ifndef MG_SNTP_NO_DELAY_CORRECTION { uint64_t orig_ts_T1, recv_ts_T2; mg_get_ntp_ts(&buf[24], &orig_ts_T1); mg_get_ntp_ts(&buf[32], &recv_ts_T2); delay = mg_calculate_delay(orig_ts_T1, recv_ts_T2, trsm_ts_T3); } #endif mg_ntp_to_tv(trsm_ts_T3, &tv); msg->time = (double) tv.tv_sec + (((double) tv.tv_usec + delay) / 1000000.0); return 0; } static void mg_sntp_handler(struct mg_connection *c, int ev, void *ev_data MG_UD_ARG(void *user_data)) { struct mbuf *io = &c->recv_mbuf; struct mg_sntp_message msg; c->handler(c, ev, ev_data MG_UD_ARG(user_data)); switch (ev) { case MG_EV_RECV: { if (mg_sntp_parse_reply(io->buf, io->len, &msg) < 0) { DBG(("Invalid SNTP packet received (%d)", (int) io->len)); c->handler(c, MG_SNTP_MALFORMED_REPLY, NULL MG_UD_ARG(user_data)); } else { c->handler(c, MG_SNTP_REPLY, (void *) &msg MG_UD_ARG(user_data)); } mbuf_remove(io, io->len); break; } } } int mg_set_protocol_sntp(struct mg_connection *c) { if ((c->flags & MG_F_UDP) == 0) { return -1; } c->proto_handler = mg_sntp_handler; return 0; } struct mg_connection *mg_sntp_connect(struct mg_mgr *mgr, MG_CB(mg_event_handler_t event_handler, void *user_data), const char *sntp_server_name) { struct mg_connection *c = NULL; char url[100], *p_url = url; const char *proto = "", *port = "", *tmp; /* If port is not specified, use default (123) */ tmp = strchr(sntp_server_name, ':'); if (tmp != NULL && *(tmp + 1) == '/') { tmp = strchr(tmp + 1, ':'); } if (tmp == NULL) { port = ":123"; } /* Add udp:// if needed */ if (strncmp(sntp_server_name, "udp://", 6) != 0) { proto = "udp://"; } mg_asprintf(&p_url, sizeof(url), "%s%s%s", proto, sntp_server_name, port); c = mg_connect(mgr, p_url, event_handler MG_UD_ARG(user_data)); if (c == NULL) { goto cleanup; } mg_set_protocol_sntp(c); cleanup: if (p_url != url) { MG_FREE(p_url); } return c; } struct sntp_data { mg_event_handler_t hander; int count; }; static void mg_sntp_util_ev_handler(struct mg_connection *c, int ev, void *ev_data MG_UD_ARG(void *user_data)) { #if !MG_ENABLE_CALLBACK_USERDATA void *user_data = c->user_data; #endif struct sntp_data *sd = (struct sntp_data *) user_data; switch (ev) { case MG_EV_CONNECT: if (*(int *) ev_data != 0) { mg_call(c, sd->hander, c->user_data, MG_SNTP_FAILED, NULL); break; } /* fallthrough */ case MG_EV_TIMER: if (sd->count <= SNTP_ATTEMPTS) { mg_sntp_send_request(c); mg_set_timer(c, mg_time() + 10); sd->count++; } else { mg_call(c, sd->hander, c->user_data, MG_SNTP_FAILED, NULL); c->flags |= MG_F_CLOSE_IMMEDIATELY; } break; case MG_SNTP_MALFORMED_REPLY: mg_call(c, sd->hander, c->user_data, MG_SNTP_FAILED, NULL); c->flags |= MG_F_CLOSE_IMMEDIATELY; break; case MG_SNTP_REPLY: mg_call(c, sd->hander, c->user_data, MG_SNTP_REPLY, ev_data); c->flags |= MG_F_CLOSE_IMMEDIATELY; break; case MG_EV_CLOSE: MG_FREE(user_data); c->user_data = NULL; break; } } struct mg_connection *mg_sntp_get_time(struct mg_mgr *mgr, mg_event_handler_t event_handler, const char *sntp_server_name) { struct mg_connection *c; struct sntp_data *sd = (struct sntp_data *) MG_CALLOC(1, sizeof(*sd)); if (sd == NULL) { return NULL; } c = mg_sntp_connect(mgr, MG_CB(mg_sntp_util_ev_handler, sd), sntp_server_name); if (c == NULL) { MG_FREE(sd); return NULL; } sd->hander = event_handler; #if !MG_ENABLE_CALLBACK_USERDATA c->user_data = sd; #endif return c; } #endif /* MG_ENABLE_SNTP */ #ifdef MG_MODULE_LINES #line 1 "mongoose/src/mg_socks.c" #endif /* * Copyright (c) 2017 Cesanta Software Limited * All rights reserved */ #if MG_ENABLE_SOCKS /* Amalgamated: #include "mg_socks.h" */ /* Amalgamated: #include "mg_internal.h" */ /* * https://www.ietf.org/rfc/rfc1928.txt paragraph 3, handle client handshake * * +----+----------+----------+ * |VER | NMETHODS | METHODS | * +----+----------+----------+ * | 1 | 1 | 1 to 255 | * +----+----------+----------+ */ static void mg_socks5_handshake(struct mg_connection *c) { struct mbuf *r = &c->recv_mbuf; if (r->buf[0] != MG_SOCKS_VERSION) { c->flags |= MG_F_CLOSE_IMMEDIATELY; } else if (r->len > 2 && (size_t) r->buf[1] + 2 <= r->len) { /* https://www.ietf.org/rfc/rfc1928.txt paragraph 3 */ unsigned char reply[2] = {MG_SOCKS_VERSION, MG_SOCKS_HANDSHAKE_FAILURE}; int i; for (i = 2; i < r->buf[1] + 2; i++) { /* TODO(lsm): support other auth methods */ if (r->buf[i] == MG_SOCKS_HANDSHAKE_NOAUTH) reply[1] = r->buf[i]; } mbuf_remove(r, 2 + r->buf[1]); mg_send(c, reply, sizeof(reply)); c->flags |= MG_SOCKS_HANDSHAKE_DONE; /* Mark handshake done */ } } static void disband(struct mg_connection *c) { struct mg_connection *c2 = (struct mg_connection *) c->user_data; if (c2 != NULL) { c2->flags |= MG_F_SEND_AND_CLOSE; c2->user_data = NULL; } c->flags |= MG_F_SEND_AND_CLOSE; c->user_data = NULL; } static void relay_data(struct mg_connection *c) { struct mg_connection *c2 = (struct mg_connection *) c->user_data; if (c2 != NULL) { mg_send(c2, c->recv_mbuf.buf, c->recv_mbuf.len); mbuf_remove(&c->recv_mbuf, c->recv_mbuf.len); } else { c->flags |= MG_F_SEND_AND_CLOSE; } } static void serv_ev_handler(struct mg_connection *c, int ev, void *ev_data) { if (ev == MG_EV_CLOSE) { disband(c); } else if (ev == MG_EV_RECV) { relay_data(c); } else if (ev == MG_EV_CONNECT) { int res = *(int *) ev_data; if (res != 0) LOG(LL_ERROR, ("connect error: %d", res)); } } static void mg_socks5_connect(struct mg_connection *c, const char *addr) { struct mg_connection *serv = mg_connect(c->mgr, addr, serv_ev_handler); serv->user_data = c; c->user_data = serv; } /* * Request, https://www.ietf.org/rfc/rfc1928.txt paragraph 4 * * +----+-----+-------+------+----------+----------+ * |VER | CMD | RSV | ATYP | DST.ADDR | DST.PORT | * +----+-----+-------+------+----------+----------+ * | 1 | 1 | X'00' | 1 | Variable | 2 | * +----+-----+-------+------+----------+----------+ */ static void mg_socks5_handle_request(struct mg_connection *c) { struct mbuf *r = &c->recv_mbuf; unsigned char *p = (unsigned char *) r->buf; unsigned char addr_len = 4, reply = MG_SOCKS_SUCCESS; int ver, cmd, atyp; char addr[300]; if (r->len < 8) return; /* return if not fully buffered. min DST.ADDR is 2 */ ver = p[0]; cmd = p[1]; atyp = p[3]; /* TODO(lsm): support other commands */ if (ver != MG_SOCKS_VERSION || cmd != MG_SOCKS_CMD_CONNECT) { reply = MG_SOCKS_CMD_NOT_SUPPORTED; } else if (atyp == MG_SOCKS_ADDR_IPV4) { addr_len = 4; if (r->len < (size_t) addr_len + 6) return; /* return if not buffered */ snprintf(addr, sizeof(addr), "%d.%d.%d.%d:%d", p[4], p[5], p[6], p[7], p[8] << 8 | p[9]); mg_socks5_connect(c, addr); } else if (atyp == MG_SOCKS_ADDR_IPV6) { addr_len = 16; if (r->len < (size_t) addr_len + 6) return; /* return if not buffered */ snprintf(addr, sizeof(addr), "[%x:%x:%x:%x:%x:%x:%x:%x]:%d", p[4] << 8 | p[5], p[6] << 8 | p[7], p[8] << 8 | p[9], p[10] << 8 | p[11], p[12] << 8 | p[13], p[14] << 8 | p[15], p[16] << 8 | p[17], p[18] << 8 | p[19], p[20] << 8 | p[21]); mg_socks5_connect(c, addr); } else if (atyp == MG_SOCKS_ADDR_DOMAIN) { addr_len = p[4] + 1; if (r->len < (size_t) addr_len + 6) return; /* return if not buffered */ snprintf(addr, sizeof(addr), "%.*s:%d", p[4], p + 5, p[4 + addr_len] << 8 | p[4 + addr_len + 1]); mg_socks5_connect(c, addr); } else { reply = MG_SOCKS_ADDR_NOT_SUPPORTED; } /* * Reply, https://www.ietf.org/rfc/rfc1928.txt paragraph 5 * * +----+-----+-------+------+----------+----------+ * |VER | REP | RSV | ATYP | BND.ADDR | BND.PORT | * +----+-----+-------+------+----------+----------+ * | 1 | 1 | X'00' | 1 | Variable | 2 | * +----+-----+-------+------+----------+----------+ */ { unsigned char buf[] = {MG_SOCKS_VERSION, reply, 0}; mg_send(c, buf, sizeof(buf)); } mg_send(c, r->buf + 3, addr_len + 1 + 2); mbuf_remove(r, 6 + addr_len); /* Remove request from the input stream */ c->flags |= MG_SOCKS_CONNECT_DONE; /* Mark ourselves as connected */ } static void socks_handler(struct mg_connection *c, int ev, void *ev_data) { if (ev == MG_EV_RECV) { if (!(c->flags & MG_SOCKS_HANDSHAKE_DONE)) mg_socks5_handshake(c); if (c->flags & MG_SOCKS_HANDSHAKE_DONE && !(c->flags & MG_SOCKS_CONNECT_DONE)) { mg_socks5_handle_request(c); } if (c->flags & MG_SOCKS_CONNECT_DONE) relay_data(c); } else if (ev == MG_EV_CLOSE) { disband(c); } (void) ev_data; } void mg_set_protocol_socks(struct mg_connection *c) { c->proto_handler = socks_handler; } #endif #ifdef MG_MODULE_LINES #line 1 "common/platforms/cc3200/cc3200_libc.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #if CS_PLATFORM == CS_P_CC3200 /* Amalgamated: #include "common/mg_mem.h" */ #include #include #ifndef __TI_COMPILER_VERSION__ #include #include #include #include #endif #include #include #include #include #include #include #include #define CONSOLE_UART UARTA0_BASE #ifdef __TI_COMPILER_VERSION__ int asprintf(char **strp, const char *fmt, ...) { va_list ap; int len; *strp = MG_MALLOC(BUFSIZ); if (*strp == NULL) return -1; va_start(ap, fmt); len = vsnprintf(*strp, BUFSIZ, fmt, ap); va_end(ap); if (len > 0) { *strp = MG_REALLOC(*strp, len + 1); if (*strp == NULL) return -1; } if (len >= BUFSIZ) { va_start(ap, fmt); len = vsnprintf(*strp, len + 1, fmt, ap); va_end(ap); } return len; } #if MG_TI_NO_HOST_INTERFACE time_t HOSTtime() { struct timeval tp; gettimeofday(&tp, NULL); return tp.tv_sec; } #endif #endif /* __TI_COMPILER_VERSION__ */ void fprint_str(FILE *fp, const char *str) { while (*str != '\0') { if (*str == '\n') MAP_UARTCharPut(CONSOLE_UART, '\r'); MAP_UARTCharPut(CONSOLE_UART, *str++); } } void _exit(int status) { fprint_str(stderr, "_exit\n"); /* cause an unaligned access exception, that will drop you into gdb */ *(int *) 1 = status; while (1) ; /* avoid gcc warning because stdlib abort() has noreturn attribute */ } void _not_implemented(const char *what) { fprint_str(stderr, what); fprint_str(stderr, " is not implemented\n"); _exit(42); } int _kill(int pid, int sig) { (void) pid; (void) sig; _not_implemented("_kill"); return -1; } int _getpid() { fprint_str(stderr, "_getpid is not implemented\n"); return 42; } int _isatty(int fd) { /* 0, 1 and 2 are TTYs. */ return fd < 2; } #endif /* CS_PLATFORM == CS_P_CC3200 */ #ifdef MG_MODULE_LINES #line 1 "common/platforms/msp432/msp432_libc.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #if CS_PLATFORM == CS_P_MSP432 #include #include int gettimeofday(struct timeval *tp, void *tzp) { uint32_t ticks = Clock_getTicks(); tp->tv_sec = ticks / 1000; tp->tv_usec = (ticks % 1000) * 1000; return 0; } #endif /* CS_PLATFORM == CS_P_MSP432 */ #ifdef MG_MODULE_LINES #line 1 "common/platforms/nrf5/nrf5_libc.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #if (CS_PLATFORM == CS_P_NRF51 || CS_PLATFORM == CS_P_NRF52) && \ defined(__ARMCC_VERSION) int gettimeofday(struct timeval *tp, void *tzp) { /* TODO */ tp->tv_sec = 0; tp->tv_usec = 0; return 0; } #endif #ifdef MG_MODULE_LINES #line 1 "common/platforms/simplelink/sl_fs_slfs.h" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #ifndef CS_COMMON_PLATFORMS_SIMPLELINK_SL_FS_SLFS_H_ #define CS_COMMON_PLATFORMS_SIMPLELINK_SL_FS_SLFS_H_ #if defined(MG_FS_SLFS) #include #ifndef __TI_COMPILER_VERSION__ #include #include #endif #define MAX_OPEN_SLFS_FILES 8 /* Indirect libc interface - same functions, different names. */ int fs_slfs_open(const char *pathname, int flags, mode_t mode); int fs_slfs_close(int fd); ssize_t fs_slfs_read(int fd, void *buf, size_t count); ssize_t fs_slfs_write(int fd, const void *buf, size_t count); int fs_slfs_stat(const char *pathname, struct stat *s); int fs_slfs_fstat(int fd, struct stat *s); off_t fs_slfs_lseek(int fd, off_t offset, int whence); int fs_slfs_unlink(const char *filename); int fs_slfs_rename(const char *from, const char *to); void fs_slfs_set_new_file_size(const char *name, size_t size); #endif /* defined(MG_FS_SLFS) */ #endif /* CS_COMMON_PLATFORMS_SIMPLELINK_SL_FS_SLFS_H_ */ #ifdef MG_MODULE_LINES #line 1 "common/platforms/simplelink/sl_fs_slfs.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ /* Standard libc interface to TI SimpleLink FS. */ #if defined(MG_FS_SLFS) || defined(CC3200_FS_SLFS) /* Amalgamated: #include "common/platforms/simplelink/sl_fs_slfs.h" */ #include #if CS_PLATFORM == CS_P_CC3200 #include #endif /* Amalgamated: #include "common/cs_dbg.h" */ /* Amalgamated: #include "common/mg_mem.h" */ #if SL_MAJOR_VERSION_NUM < 2 int slfs_open(const unsigned char *fname, uint32_t flags) { _i32 fh; _i32 r = sl_FsOpen(fname, flags, NULL /* token */, &fh); return (r < 0 ? r : fh); } #else /* SL_MAJOR_VERSION_NUM >= 2 */ int slfs_open(const unsigned char *fname, uint32_t flags) { return sl_FsOpen(fname, flags, NULL /* token */); } #endif /* From sl_fs.c */ int set_errno(int e); const char *drop_dir(const char *fname, bool *is_slfs); /* * With SLFS, you have to pre-declare max file size. Yes. Really. * 64K should be enough for everyone. Right? */ #ifndef FS_SLFS_MAX_FILE_SIZE #define FS_SLFS_MAX_FILE_SIZE (64 * 1024) #endif struct sl_file_size_hint { char *name; size_t size; }; struct sl_fd_info { _i32 fh; _off_t pos; size_t size; }; static struct sl_fd_info s_sl_fds[MAX_OPEN_SLFS_FILES]; static struct sl_file_size_hint s_sl_file_size_hints[MAX_OPEN_SLFS_FILES]; static int sl_fs_to_errno(_i32 r) { DBG(("SL error: %d", (int) r)); switch (r) { case SL_FS_OK: return 0; case SL_ERROR_FS_FILE_NAME_EXIST: return EEXIST; case SL_ERROR_FS_WRONG_FILE_NAME: return EINVAL; case SL_ERROR_FS_NO_AVAILABLE_NV_INDEX: case SL_ERROR_FS_NOT_ENOUGH_STORAGE_SPACE: return ENOSPC; case SL_ERROR_FS_FAILED_TO_ALLOCATE_MEM: return ENOMEM; case SL_ERROR_FS_FILE_NOT_EXISTS: return ENOENT; case SL_ERROR_FS_NOT_SUPPORTED: return ENOTSUP; } return ENXIO; } int fs_slfs_open(const char *pathname, int flags, mode_t mode) { int fd; for (fd = 0; fd < MAX_OPEN_SLFS_FILES; fd++) { if (s_sl_fds[fd].fh <= 0) break; } if (fd >= MAX_OPEN_SLFS_FILES) return set_errno(ENOMEM); struct sl_fd_info *fi = &s_sl_fds[fd]; /* * Apply path manipulations again, in case we got here directly * (via TI libc's "add_device"). */ pathname = drop_dir(pathname, NULL); _u32 am = 0; fi->size = (size_t) -1; int rw = (flags & 3); size_t new_size = FS_SLFS_MAX_FILE_SIZE; if (rw == O_RDONLY) { SlFsFileInfo_t sl_fi; _i32 r = sl_FsGetInfo((const _u8 *) pathname, 0, &sl_fi); if (r == SL_FS_OK) { fi->size = SL_FI_FILE_SIZE(sl_fi); } am = SL_FS_READ; } else { if (!(flags & O_TRUNC) || (flags & O_APPEND)) { // FailFS files cannot be opened for append and will be truncated // when opened for write. return set_errno(ENOTSUP); } if (flags & O_CREAT) { size_t i; for (i = 0; i < MAX_OPEN_SLFS_FILES; i++) { if (s_sl_file_size_hints[i].name != NULL && strcmp(s_sl_file_size_hints[i].name, pathname) == 0) { new_size = s_sl_file_size_hints[i].size; MG_FREE(s_sl_file_size_hints[i].name); s_sl_file_size_hints[i].name = NULL; break; } } am = FS_MODE_OPEN_CREATE(new_size, 0); } else { am = SL_FS_WRITE; } } fi->fh = slfs_open((_u8 *) pathname, am); LOG(LL_DEBUG, ("sl_FsOpen(%s, 0x%x) sz %u = %d", pathname, (int) am, (unsigned int) new_size, (int) fi->fh)); int r; if (fi->fh >= 0) { fi->pos = 0; r = fd; } else { r = set_errno(sl_fs_to_errno(fi->fh)); } return r; } int fs_slfs_close(int fd) { struct sl_fd_info *fi = &s_sl_fds[fd]; if (fi->fh <= 0) return set_errno(EBADF); _i32 r = sl_FsClose(fi->fh, NULL, NULL, 0); LOG(LL_DEBUG, ("sl_FsClose(%d) = %d", (int) fi->fh, (int) r)); s_sl_fds[fd].fh = -1; return set_errno(sl_fs_to_errno(r)); } ssize_t fs_slfs_read(int fd, void *buf, size_t count) { struct sl_fd_info *fi = &s_sl_fds[fd]; if (fi->fh <= 0) return set_errno(EBADF); /* Simulate EOF. sl_FsRead @ file_size return SL_FS_ERR_OFFSET_OUT_OF_RANGE. */ if (fi->pos == fi->size) return 0; _i32 r = sl_FsRead(fi->fh, fi->pos, buf, count); DBG(("sl_FsRead(%d, %d, %d) = %d", (int) fi->fh, (int) fi->pos, (int) count, (int) r)); if (r >= 0) { fi->pos += r; return r; } return set_errno(sl_fs_to_errno(r)); } ssize_t fs_slfs_write(int fd, const void *buf, size_t count) { struct sl_fd_info *fi = &s_sl_fds[fd]; if (fi->fh <= 0) return set_errno(EBADF); _i32 r = sl_FsWrite(fi->fh, fi->pos, (_u8 *) buf, count); DBG(("sl_FsWrite(%d, %d, %d) = %d", (int) fi->fh, (int) fi->pos, (int) count, (int) r)); if (r >= 0) { fi->pos += r; return r; } return set_errno(sl_fs_to_errno(r)); } int fs_slfs_stat(const char *pathname, struct stat *s) { SlFsFileInfo_t sl_fi; /* * Apply path manipulations again, in case we got here directly * (via TI libc's "add_device"). */ pathname = drop_dir(pathname, NULL); _i32 r = sl_FsGetInfo((const _u8 *) pathname, 0, &sl_fi); if (r == SL_FS_OK) { s->st_mode = S_IFREG | 0666; s->st_nlink = 1; s->st_size = SL_FI_FILE_SIZE(sl_fi); return 0; } return set_errno(sl_fs_to_errno(r)); } int fs_slfs_fstat(int fd, struct stat *s) { struct sl_fd_info *fi = &s_sl_fds[fd]; if (fi->fh <= 0) return set_errno(EBADF); s->st_mode = 0666; s->st_mode = S_IFREG | 0666; s->st_nlink = 1; s->st_size = fi->size; return 0; } off_t fs_slfs_lseek(int fd, off_t offset, int whence) { if (s_sl_fds[fd].fh <= 0) return set_errno(EBADF); switch (whence) { case SEEK_SET: s_sl_fds[fd].pos = offset; break; case SEEK_CUR: s_sl_fds[fd].pos += offset; break; case SEEK_END: return set_errno(ENOTSUP); } return 0; } int fs_slfs_unlink(const char *pathname) { /* * Apply path manipulations again, in case we got here directly * (via TI libc's "add_device"). */ pathname = drop_dir(pathname, NULL); return set_errno(sl_fs_to_errno(sl_FsDel((const _u8 *) pathname, 0))); } int fs_slfs_rename(const char *from, const char *to) { return set_errno(ENOTSUP); } void fs_slfs_set_new_file_size(const char *name, size_t size) { int i; for (i = 0; i < MAX_OPEN_SLFS_FILES; i++) { if (s_sl_file_size_hints[i].name == NULL) { DBG(("File size hint: %s %d", name, (int) size)); s_sl_file_size_hints[i].name = strdup(name); s_sl_file_size_hints[i].size = size; break; } } } #endif /* defined(MG_FS_SLFS) || defined(CC3200_FS_SLFS) */ #ifdef MG_MODULE_LINES #line 1 "common/platforms/simplelink/sl_fs.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #if MG_NET_IF == MG_NET_IF_SIMPLELINK && \ (defined(MG_FS_SLFS) || defined(MG_FS_SPIFFS)) int set_errno(int e) { errno = e; return (e == 0 ? 0 : -1); } const char *drop_dir(const char *fname, bool *is_slfs) { if (is_slfs != NULL) { *is_slfs = (strncmp(fname, "SL:", 3) == 0); if (*is_slfs) fname += 3; } /* Drop "./", if any */ if (fname[0] == '.' && fname[1] == '/') { fname += 2; } /* * Drop / if it is the only one in the path. * This allows use of /pretend/directories but serves /file.txt as normal. */ if (fname[0] == '/' && strchr(fname + 1, '/') == NULL) { fname++; } return fname; } #if !defined(MG_FS_NO_VFS) #include #include #include #include #include #ifdef __TI_COMPILER_VERSION__ #include #endif /* Amalgamated: #include "common/cs_dbg.h" */ /* Amalgamated: #include "common/platform.h" */ #ifdef CC3200_FS_SPIFFS /* Amalgamated: #include "cc3200_fs_spiffs.h" */ #endif #ifdef MG_FS_SLFS /* Amalgamated: #include "sl_fs_slfs.h" */ #endif #define NUM_SYS_FDS 3 #define SPIFFS_FD_BASE 10 #define SLFS_FD_BASE 100 #if !defined(MG_UART_CHAR_PUT) && !defined(MG_UART_WRITE) #if CS_PLATFORM == CS_P_CC3200 #include #include #include #include #include #define MG_UART_CHAR_PUT(fd, c) MAP_UARTCharPut(UARTA0_BASE, c); #else #define MG_UART_WRITE(fd, buf, len) #endif /* CS_PLATFORM == CS_P_CC3200 */ #endif /* !MG_UART_CHAR_PUT */ enum fd_type { FD_INVALID, FD_SYS, #ifdef CC3200_FS_SPIFFS FD_SPIFFS, #endif #ifdef MG_FS_SLFS FD_SLFS #endif }; static int fd_type(int fd) { if (fd >= 0 && fd < NUM_SYS_FDS) return FD_SYS; #ifdef CC3200_FS_SPIFFS if (fd >= SPIFFS_FD_BASE && fd < SPIFFS_FD_BASE + MAX_OPEN_SPIFFS_FILES) { return FD_SPIFFS; } #endif #ifdef MG_FS_SLFS if (fd >= SLFS_FD_BASE && fd < SLFS_FD_BASE + MAX_OPEN_SLFS_FILES) { return FD_SLFS; } #endif return FD_INVALID; } #if MG_TI_NO_HOST_INTERFACE int open(const char *pathname, unsigned flags, int mode) { #else int _open(const char *pathname, int flags, mode_t mode) { #endif int fd = -1; bool is_sl; const char *fname = drop_dir(pathname, &is_sl); if (is_sl) { #ifdef MG_FS_SLFS fd = fs_slfs_open(fname, flags, mode); if (fd >= 0) fd += SLFS_FD_BASE; #endif } else { #ifdef CC3200_FS_SPIFFS fd = fs_spiffs_open(fname, flags, mode); if (fd >= 0) fd += SPIFFS_FD_BASE; #endif } LOG(LL_DEBUG, ("open(%s, 0x%x) = %d, fname = %s", pathname, flags, fd, fname)); return fd; } int _stat(const char *pathname, struct stat *st) { int res = -1; bool is_sl; const char *fname = drop_dir(pathname, &is_sl); memset(st, 0, sizeof(*st)); /* Simulate statting the root directory. */ if (fname[0] == '\0' || strcmp(fname, ".") == 0) { st->st_ino = 0; st->st_mode = S_IFDIR | 0777; st->st_nlink = 1; st->st_size = 0; return 0; } if (is_sl) { #ifdef MG_FS_SLFS res = fs_slfs_stat(fname, st); #endif } else { #ifdef CC3200_FS_SPIFFS res = fs_spiffs_stat(fname, st); #endif } LOG(LL_DEBUG, ("stat(%s) = %d; fname = %s", pathname, res, fname)); return res; } #if MG_TI_NO_HOST_INTERFACE int close(int fd) { #else int _close(int fd) { #endif int r = -1; switch (fd_type(fd)) { case FD_INVALID: r = set_errno(EBADF); break; case FD_SYS: r = set_errno(EACCES); break; #ifdef CC3200_FS_SPIFFS case FD_SPIFFS: r = fs_spiffs_close(fd - SPIFFS_FD_BASE); break; #endif #ifdef MG_FS_SLFS case FD_SLFS: r = fs_slfs_close(fd - SLFS_FD_BASE); break; #endif } DBG(("close(%d) = %d", fd, r)); return r; } #if MG_TI_NO_HOST_INTERFACE off_t lseek(int fd, off_t offset, int whence) { #else off_t _lseek(int fd, off_t offset, int whence) { #endif int r = -1; switch (fd_type(fd)) { case FD_INVALID: r = set_errno(EBADF); break; case FD_SYS: r = set_errno(ESPIPE); break; #ifdef CC3200_FS_SPIFFS case FD_SPIFFS: r = fs_spiffs_lseek(fd - SPIFFS_FD_BASE, offset, whence); break; #endif #ifdef MG_FS_SLFS case FD_SLFS: r = fs_slfs_lseek(fd - SLFS_FD_BASE, offset, whence); break; #endif } DBG(("lseek(%d, %d, %d) = %d", fd, (int) offset, whence, r)); return r; } int _fstat(int fd, struct stat *s) { int r = -1; memset(s, 0, sizeof(*s)); switch (fd_type(fd)) { case FD_INVALID: r = set_errno(EBADF); break; case FD_SYS: { /* Create barely passable stats for STD{IN,OUT,ERR}. */ memset(s, 0, sizeof(*s)); s->st_ino = fd; s->st_mode = S_IFCHR | 0666; r = 0; break; } #ifdef CC3200_FS_SPIFFS case FD_SPIFFS: r = fs_spiffs_fstat(fd - SPIFFS_FD_BASE, s); break; #endif #ifdef MG_FS_SLFS case FD_SLFS: r = fs_slfs_fstat(fd - SLFS_FD_BASE, s); break; #endif } DBG(("fstat(%d) = %d", fd, r)); return r; } #if MG_TI_NO_HOST_INTERFACE int read(int fd, char *buf, unsigned count) { #else ssize_t _read(int fd, void *buf, size_t count) { #endif int r = -1; switch (fd_type(fd)) { case FD_INVALID: r = set_errno(EBADF); break; case FD_SYS: { if (fd != 0) { r = set_errno(EACCES); break; } /* Should we allow reading from stdin = uart? */ r = set_errno(ENOTSUP); break; } #ifdef CC3200_FS_SPIFFS case FD_SPIFFS: r = fs_spiffs_read(fd - SPIFFS_FD_BASE, buf, count); break; #endif #ifdef MG_FS_SLFS case FD_SLFS: r = fs_slfs_read(fd - SLFS_FD_BASE, buf, count); break; #endif } DBG(("read(%d, %u) = %d", fd, count, r)); return r; } #if MG_TI_NO_HOST_INTERFACE int write(int fd, const char *buf, unsigned count) { #else ssize_t _write(int fd, const void *buf, size_t count) { #endif int r = -1; switch (fd_type(fd)) { case FD_INVALID: r = set_errno(EBADF); break; case FD_SYS: { if (fd == 0) { r = set_errno(EACCES); break; } #ifdef MG_UART_WRITE MG_UART_WRITE(fd, buf, count); #elif defined(MG_UART_CHAR_PUT) { size_t i; for (i = 0; i < count; i++) { const char c = ((const char *) buf)[i]; if (c == '\n') MG_UART_CHAR_PUT(fd, '\r'); MG_UART_CHAR_PUT(fd, c); } } #endif r = count; break; } #ifdef CC3200_FS_SPIFFS case FD_SPIFFS: r = fs_spiffs_write(fd - SPIFFS_FD_BASE, buf, count); break; #endif #ifdef MG_FS_SLFS case FD_SLFS: r = fs_slfs_write(fd - SLFS_FD_BASE, buf, count); break; #endif } return r; } /* * On Newlib we override rename directly too, because the default * implementation using _link and _unlink doesn't work for us. */ #if MG_TI_NO_HOST_INTERFACE || defined(_NEWLIB_VERSION) int rename(const char *frompath, const char *topath) { int r = -1; bool is_sl_from, is_sl_to; const char *from = drop_dir(frompath, &is_sl_from); const char *to = drop_dir(topath, &is_sl_to); if (is_sl_from || is_sl_to) { set_errno(ENOTSUP); } else { #ifdef CC3200_FS_SPIFFS r = fs_spiffs_rename(from, to); #endif } DBG(("rename(%s, %s) = %d", from, to, r)); return r; } #endif /* MG_TI_NO_HOST_INTERFACE || defined(_NEWLIB_VERSION) */ #if MG_TI_NO_HOST_INTERFACE int unlink(const char *pathname) { #else int _unlink(const char *pathname) { #endif int r = -1; bool is_sl; const char *fname = drop_dir(pathname, &is_sl); if (is_sl) { #ifdef MG_FS_SLFS r = fs_slfs_unlink(fname); #endif } else { #ifdef CC3200_FS_SPIFFS r = fs_spiffs_unlink(fname); #endif } DBG(("unlink(%s) = %d, fname = %s", pathname, r, fname)); return r; } #ifdef CC3200_FS_SPIFFS /* FailFS does not support listing files. */ DIR *opendir(const char *dir_name) { DIR *r = NULL; bool is_sl; drop_dir(dir_name, &is_sl); if (is_sl) { r = NULL; set_errno(ENOTSUP); } else { r = fs_spiffs_opendir(dir_name); } DBG(("opendir(%s) = %p", dir_name, r)); return r; } struct dirent *readdir(DIR *dir) { struct dirent *res = fs_spiffs_readdir(dir); DBG(("readdir(%p) = %p", dir, res)); return res; } int closedir(DIR *dir) { int res = fs_spiffs_closedir(dir); DBG(("closedir(%p) = %d", dir, res)); return res; } int rmdir(const char *path) { return fs_spiffs_rmdir(path); } int mkdir(const char *path, mode_t mode) { (void) path; (void) mode; /* for spiffs supports only root dir, which comes from mongoose as '.' */ return (strlen(path) == 1 && *path == '.') ? 0 : ENOTDIR; } #endif int sl_fs_init(void) { int ret = 1; #ifdef __TI_COMPILER_VERSION__ #ifdef MG_FS_SLFS #pragma diag_push #pragma diag_suppress 169 /* Nothing we can do about the prototype mismatch. \ */ ret = (add_device("SL", _MSA, fs_slfs_open, fs_slfs_close, fs_slfs_read, fs_slfs_write, fs_slfs_lseek, fs_slfs_unlink, fs_slfs_rename) == 0); #pragma diag_pop #endif #endif return ret; } #endif /* !defined(MG_FS_NO_VFS) */ #endif /* MG_NET_IF == MG_NET_IF_SIMPLELINK && (defined(MG_FS_SLFS) || \ defined(MG_FS_SPIFFS)) */ #ifdef MG_MODULE_LINES #line 1 "common/platforms/simplelink/sl_socket.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #if MG_NET_IF == MG_NET_IF_SIMPLELINK #include #include /* Amalgamated: #include "common/platform.h" */ const char *inet_ntop(int af, const void *src, char *dst, socklen_t size) { int res; struct in_addr *in = (struct in_addr *) src; if (af != AF_INET) { errno = ENOTSUP; return NULL; } res = snprintf(dst, size, "%lu.%lu.%lu.%lu", SL_IPV4_BYTE(in->s_addr, 0), SL_IPV4_BYTE(in->s_addr, 1), SL_IPV4_BYTE(in->s_addr, 2), SL_IPV4_BYTE(in->s_addr, 3)); return res > 0 ? dst : NULL; } char *inet_ntoa(struct in_addr n) { static char a[16]; return (char *) inet_ntop(AF_INET, &n, a, sizeof(a)); } int inet_pton(int af, const char *src, void *dst) { uint32_t a0, a1, a2, a3; uint8_t *db = (uint8_t *) dst; if (af != AF_INET) { errno = ENOTSUP; return 0; } if (sscanf(src, "%lu.%lu.%lu.%lu", &a0, &a1, &a2, &a3) != 4) { return 0; } *db = a3; *(db + 1) = a2; *(db + 2) = a1; *(db + 3) = a0; return 1; } #endif /* MG_NET_IF == MG_NET_IF_SIMPLELINK */ #ifdef MG_MODULE_LINES #line 1 "common/platforms/simplelink/sl_mg_task.c" #endif #if MG_NET_IF == MG_NET_IF_SIMPLELINK && !defined(MG_SIMPLELINK_NO_OSI) /* Amalgamated: #include "mg_task.h" */ #include enum mg_q_msg_type { MG_Q_MSG_CB, }; struct mg_q_msg { enum mg_q_msg_type type; void (*cb)(struct mg_mgr *mgr, void *arg); void *arg; }; static OsiMsgQ_t s_mg_q; static void mg_task(void *arg); bool mg_start_task(int priority, int stack_size, mg_init_cb mg_init) { if (osi_MsgQCreate(&s_mg_q, "MG", sizeof(struct mg_q_msg), 16) != OSI_OK) { return false; } if (osi_TaskCreate(mg_task, (const signed char *) "MG", stack_size, (void *) mg_init, priority, NULL) != OSI_OK) { return false; } return true; } static void mg_task(void *arg) { struct mg_mgr mgr; mg_init_cb mg_init = (mg_init_cb) arg; mg_mgr_init(&mgr, NULL); mg_init(&mgr); while (1) { struct mg_q_msg msg; mg_mgr_poll(&mgr, 1); if (osi_MsgQRead(&s_mg_q, &msg, 1) != OSI_OK) continue; switch (msg.type) { case MG_Q_MSG_CB: { msg.cb(&mgr, msg.arg); } } } } void mg_run_in_task(void (*cb)(struct mg_mgr *mgr, void *arg), void *cb_arg) { struct mg_q_msg msg = {MG_Q_MSG_CB, cb, cb_arg}; osi_MsgQWrite(&s_mg_q, &msg, OSI_NO_WAIT); } #endif /* MG_NET_IF == MG_NET_IF_SIMPLELINK && !defined(MG_SIMPLELINK_NO_OSI) \ */ #ifdef MG_MODULE_LINES #line 1 "common/platforms/simplelink/sl_net_if.h" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #ifndef CS_COMMON_PLATFORMS_SIMPLELINK_SL_NET_IF_H_ #define CS_COMMON_PLATFORMS_SIMPLELINK_SL_NET_IF_H_ /* Amalgamated: #include "mongoose/src/net_if.h" */ #ifdef __cplusplus extern "C" { #endif /* __cplusplus */ #ifndef MG_ENABLE_NET_IF_SIMPLELINK #define MG_ENABLE_NET_IF_SIMPLELINK MG_NET_IF == MG_NET_IF_SIMPLELINK #endif extern const struct mg_iface_vtable mg_simplelink_iface_vtable; #ifdef __cplusplus } #endif /* __cplusplus */ #endif /* CS_COMMON_PLATFORMS_SIMPLELINK_SL_NET_IF_H_ */ #ifdef MG_MODULE_LINES #line 1 "common/platforms/simplelink/sl_net_if.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ /* Amalgamated: #include "common/platforms/simplelink/sl_net_if.h" */ #if MG_ENABLE_NET_IF_SIMPLELINK /* Amalgamated: #include "mongoose/src/internal.h" */ /* Amalgamated: #include "mongoose/src/util.h" */ #define MG_TCP_RECV_BUFFER_SIZE 1024 #define MG_UDP_RECV_BUFFER_SIZE 1500 static sock_t mg_open_listening_socket(struct mg_connection *nc, union socket_address *sa, int type, int proto); void mg_set_non_blocking_mode(sock_t sock) { SlSockNonblocking_t opt; #if SL_MAJOR_VERSION_NUM < 2 opt.NonblockingEnabled = 1; #else opt.NonBlockingEnabled = 1; #endif sl_SetSockOpt(sock, SL_SOL_SOCKET, SL_SO_NONBLOCKING, &opt, sizeof(opt)); } static int mg_is_error(int n) { return (n < 0 && n != SL_ERROR_BSD_EALREADY && n != SL_ERROR_BSD_EAGAIN); } void mg_sl_if_connect_tcp(struct mg_connection *nc, const union socket_address *sa) { int proto = 0; if (nc->flags & MG_F_SSL) proto = SL_SEC_SOCKET; sock_t sock = sl_Socket(AF_INET, SOCK_STREAM, proto); if (sock < 0) { nc->err = sock; goto out; } mg_sock_set(nc, sock); #if MG_ENABLE_SSL nc->err = sl_set_ssl_opts(sock, nc); if (nc->err != 0) goto out; #endif nc->err = sl_Connect(sock, &sa->sa, sizeof(sa->sin)); out: DBG(("%p to %s:%d sock %d %d err %d", nc, inet_ntoa(sa->sin.sin_addr), ntohs(sa->sin.sin_port), nc->sock, proto, nc->err)); } void mg_sl_if_connect_udp(struct mg_connection *nc) { sock_t sock = sl_Socket(AF_INET, SOCK_DGRAM, 0); if (sock < 0) { nc->err = sock; return; } mg_sock_set(nc, sock); nc->err = 0; } int mg_sl_if_listen_tcp(struct mg_connection *nc, union socket_address *sa) { int proto = 0; if (nc->flags & MG_F_SSL) proto = SL_SEC_SOCKET; sock_t sock = mg_open_listening_socket(nc, sa, SOCK_STREAM, proto); if (sock < 0) return sock; mg_sock_set(nc, sock); return 0; } int mg_sl_if_listen_udp(struct mg_connection *nc, union socket_address *sa) { sock_t sock = mg_open_listening_socket(nc, sa, SOCK_DGRAM, 0); if (sock == INVALID_SOCKET) return (errno ? errno : 1); mg_sock_set(nc, sock); return 0; } void mg_sl_if_tcp_send(struct mg_connection *nc, const void *buf, size_t len) { mbuf_append(&nc->send_mbuf, buf, len); } void mg_sl_if_udp_send(struct mg_connection *nc, const void *buf, size_t len) { mbuf_append(&nc->send_mbuf, buf, len); } void mg_sl_if_recved(struct mg_connection *nc, size_t len) { (void) nc; (void) len; } int mg_sl_if_create_conn(struct mg_connection *nc) { (void) nc; return 1; } void mg_sl_if_destroy_conn(struct mg_connection *nc) { if (nc->sock == INVALID_SOCKET) return; /* For UDP, only close outgoing sockets or listeners. */ if (!(nc->flags & MG_F_UDP) || nc->listener == NULL) { sl_Close(nc->sock); } nc->sock = INVALID_SOCKET; } static int mg_accept_conn(struct mg_connection *lc) { struct mg_connection *nc; union socket_address sa; socklen_t sa_len = sizeof(sa); sock_t sock = sl_Accept(lc->sock, &sa.sa, &sa_len); if (sock < 0) { DBG(("%p: failed to accept: %d", lc, sock)); return 0; } nc = mg_if_accept_new_conn(lc); if (nc == NULL) { sl_Close(sock); return 0; } DBG(("%p conn from %s:%d", nc, inet_ntoa(sa.sin.sin_addr), ntohs(sa.sin.sin_port))); mg_sock_set(nc, sock); if (nc->flags & MG_F_SSL) nc->flags |= MG_F_SSL_HANDSHAKE_DONE; mg_if_accept_tcp_cb(nc, &sa, sa_len); return 1; } /* 'sa' must be an initialized address to bind to */ static sock_t mg_open_listening_socket(struct mg_connection *nc, union socket_address *sa, int type, int proto) { int r; socklen_t sa_len = (sa->sa.sa_family == AF_INET) ? sizeof(sa->sin) : sizeof(sa->sin6); sock_t sock = sl_Socket(sa->sa.sa_family, type, proto); if (sock < 0) return sock; if ((r = sl_Bind(sock, &sa->sa, sa_len)) < 0) goto clean; if (type != SOCK_DGRAM) { #if MG_ENABLE_SSL if ((r = sl_set_ssl_opts(sock, nc)) < 0) goto clean; #endif if ((r = sl_Listen(sock, SOMAXCONN)) < 0) goto clean; } mg_set_non_blocking_mode(sock); clean: if (r < 0) { sl_Close(sock); sock = r; } return sock; } static void mg_write_to_socket(struct mg_connection *nc) { struct mbuf *io = &nc->send_mbuf; int n = 0; if (nc->flags & MG_F_UDP) { n = sl_SendTo(nc->sock, io->buf, io->len, 0, &nc->sa.sa, sizeof(nc->sa.sin)); DBG(("%p %d %d %d %s:%hu", nc, nc->sock, n, errno, inet_ntoa(nc->sa.sin.sin_addr), ntohs(nc->sa.sin.sin_port))); } else { n = (int) sl_Send(nc->sock, io->buf, io->len, 0); DBG(("%p %d bytes -> %d", nc, n, nc->sock)); } if (n > 0) { mg_if_sent_cb(nc, n); } else if (n < 0 && mg_is_error(n)) { /* Something went wrong, drop the connection. */ nc->flags |= MG_F_CLOSE_IMMEDIATELY; } } MG_INTERNAL size_t recv_avail_size(struct mg_connection *conn, size_t max) { size_t avail; if (conn->recv_mbuf_limit < conn->recv_mbuf.len) return 0; avail = conn->recv_mbuf_limit - conn->recv_mbuf.len; return avail > max ? max : avail; } static void mg_handle_tcp_read(struct mg_connection *conn) { int n = 0; char *buf = (char *) MG_MALLOC(MG_TCP_RECV_BUFFER_SIZE); if (buf == NULL) { DBG(("OOM")); return; } n = (int) sl_Recv(conn->sock, buf, recv_avail_size(conn, MG_TCP_RECV_BUFFER_SIZE), 0); DBG(("%p %d bytes <- %d", conn, n, conn->sock)); if (n > 0) { mg_if_recv_tcp_cb(conn, buf, n, 1 /* own */); } else { MG_FREE(buf); } if (n == 0) { /* Orderly shutdown of the socket, try flushing output. */ conn->flags |= MG_F_SEND_AND_CLOSE; } else if (mg_is_error(n)) { conn->flags |= MG_F_CLOSE_IMMEDIATELY; } } static void mg_handle_udp_read(struct mg_connection *nc) { char *buf = (char *) MG_MALLOC(MG_UDP_RECV_BUFFER_SIZE); if (buf == NULL) return; union socket_address sa; socklen_t sa_len = sizeof(sa); int n = sl_RecvFrom(nc->sock, buf, MG_UDP_RECV_BUFFER_SIZE, 0, (SlSockAddr_t *) &sa, &sa_len); DBG(("%p %d bytes from %s:%d", nc, n, inet_ntoa(nc->sa.sin.sin_addr), ntohs(nc->sa.sin.sin_port))); if (n > 0) { mg_if_recv_udp_cb(nc, buf, n, &sa, sa_len); } else { MG_FREE(buf); } } #define _MG_F_FD_CAN_READ 1 #define _MG_F_FD_CAN_WRITE 1 << 1 #define _MG_F_FD_ERROR 1 << 2 void mg_mgr_handle_conn(struct mg_connection *nc, int fd_flags, double now) { DBG(("%p fd=%d fd_flags=%d nc_flags=%lu rmbl=%d smbl=%d", nc, nc->sock, fd_flags, nc->flags, (int) nc->recv_mbuf.len, (int) nc->send_mbuf.len)); if (nc->flags & MG_F_CONNECTING) { if (nc->flags & MG_F_UDP || nc->err != SL_ERROR_BSD_EALREADY) { mg_if_connect_cb(nc, nc->err); } else { /* In SimpleLink, to get status of non-blocking connect() we need to wait * until socket is writable and repeat the call to sl_Connect again, * which will now return the real status. */ if (fd_flags & _MG_F_FD_CAN_WRITE) { nc->err = sl_Connect(nc->sock, &nc->sa.sa, sizeof(nc->sa.sin)); DBG(("%p conn res=%d", nc, nc->err)); if (nc->err == SL_ERROR_BSD_ESECSNOVERIFY || /* TODO(rojer): Provide API to set the date for verification. */ nc->err == SL_ERROR_BSD_ESECDATEERROR #if SL_MAJOR_VERSION_NUM >= 2 /* Per SWRU455, this error does not mean verification failed, * it only means that the cert used is not present in the trusted * root CA catalog. Which is perfectly fine. */ || nc->err == SL_ERROR_BSD_ESECUNKNOWNROOTCA #endif ) { nc->err = 0; } if (nc->flags & MG_F_SSL && nc->err == 0) { nc->flags |= MG_F_SSL_HANDSHAKE_DONE; } mg_if_connect_cb(nc, nc->err); } } /* Ignore read/write in further processing, we've handled it. */ fd_flags &= ~(_MG_F_FD_CAN_READ | _MG_F_FD_CAN_WRITE); } if (fd_flags & _MG_F_FD_CAN_READ) { if (nc->flags & MG_F_UDP) { mg_handle_udp_read(nc); } else { if (nc->flags & MG_F_LISTENING) { mg_accept_conn(nc); } else { mg_handle_tcp_read(nc); } } } if (!(nc->flags & MG_F_CLOSE_IMMEDIATELY)) { if ((fd_flags & _MG_F_FD_CAN_WRITE) && nc->send_mbuf.len > 0) { mg_write_to_socket(nc); } if (!(fd_flags & (_MG_F_FD_CAN_READ | _MG_F_FD_CAN_WRITE))) { mg_if_poll(nc, now); } mg_if_timer(nc, now); } DBG(("%p after fd=%d nc_flags=%lu rmbl=%d smbl=%d", nc, nc->sock, nc->flags, (int) nc->recv_mbuf.len, (int) nc->send_mbuf.len)); } /* Associate a socket to a connection. */ void mg_sl_if_sock_set(struct mg_connection *nc, sock_t sock) { mg_set_non_blocking_mode(sock); nc->sock = sock; DBG(("%p %d", nc, sock)); } void mg_sl_if_init(struct mg_iface *iface) { (void) iface; DBG(("%p using sl_Select()", iface->mgr)); } void mg_sl_if_free(struct mg_iface *iface) { (void) iface; } void mg_sl_if_add_conn(struct mg_connection *nc) { (void) nc; } void mg_sl_if_remove_conn(struct mg_connection *nc) { (void) nc; } time_t mg_sl_if_poll(struct mg_iface *iface, int timeout_ms) { struct mg_mgr *mgr = iface->mgr; double now = mg_time(); double min_timer; struct mg_connection *nc, *tmp; struct SlTimeval_t tv; SlFdSet_t read_set, write_set, err_set; sock_t max_fd = INVALID_SOCKET; int num_fds, num_ev = 0, num_timers = 0; SL_SOCKET_FD_ZERO(&read_set); SL_SOCKET_FD_ZERO(&write_set); SL_SOCKET_FD_ZERO(&err_set); /* * Note: it is ok to have connections with sock == INVALID_SOCKET in the list, * e.g. timer-only "connections". */ min_timer = 0; for (nc = mgr->active_connections, num_fds = 0; nc != NULL; nc = tmp) { tmp = nc->next; if (nc->sock != INVALID_SOCKET) { num_fds++; if (!(nc->flags & MG_F_WANT_WRITE) && nc->recv_mbuf.len < nc->recv_mbuf_limit && (!(nc->flags & MG_F_UDP) || nc->listener == NULL)) { SL_SOCKET_FD_SET(nc->sock, &read_set); if (max_fd == INVALID_SOCKET || nc->sock > max_fd) max_fd = nc->sock; } if (((nc->flags & MG_F_CONNECTING) && !(nc->flags & MG_F_WANT_READ)) || (nc->send_mbuf.len > 0 && !(nc->flags & MG_F_CONNECTING))) { SL_SOCKET_FD_SET(nc->sock, &write_set); SL_SOCKET_FD_SET(nc->sock, &err_set); if (max_fd == INVALID_SOCKET || nc->sock > max_fd) max_fd = nc->sock; } } if (nc->ev_timer_time > 0) { if (num_timers == 0 || nc->ev_timer_time < min_timer) { min_timer = nc->ev_timer_time; } num_timers++; } } /* * If there is a timer to be fired earlier than the requested timeout, * adjust the timeout. */ if (num_timers > 0) { double timer_timeout_ms = (min_timer - mg_time()) * 1000 + 1 /* rounding */; if (timer_timeout_ms < timeout_ms) { timeout_ms = timer_timeout_ms; } } if (timeout_ms < 0) timeout_ms = 0; tv.tv_sec = timeout_ms / 1000; tv.tv_usec = (timeout_ms % 1000) * 1000; if (num_fds > 0) { num_ev = sl_Select((int) max_fd + 1, &read_set, &write_set, &err_set, &tv); } now = mg_time(); DBG(("sl_Select @ %ld num_ev=%d of %d, timeout=%d", (long) now, num_ev, num_fds, timeout_ms)); for (nc = mgr->active_connections; nc != NULL; nc = tmp) { int fd_flags = 0; if (nc->sock != INVALID_SOCKET) { if (num_ev > 0) { fd_flags = (SL_SOCKET_FD_ISSET(nc->sock, &read_set) && (!(nc->flags & MG_F_UDP) || nc->listener == NULL) ? _MG_F_FD_CAN_READ : 0) | (SL_SOCKET_FD_ISSET(nc->sock, &write_set) ? _MG_F_FD_CAN_WRITE : 0) | (SL_SOCKET_FD_ISSET(nc->sock, &err_set) ? _MG_F_FD_ERROR : 0); } /* SimpleLink does not report UDP sockets as writable. */ if (nc->flags & MG_F_UDP && nc->send_mbuf.len > 0) { fd_flags |= _MG_F_FD_CAN_WRITE; } } tmp = nc->next; mg_mgr_handle_conn(nc, fd_flags, now); } for (nc = mgr->active_connections; nc != NULL; nc = tmp) { tmp = nc->next; if ((nc->flags & MG_F_CLOSE_IMMEDIATELY) || (nc->send_mbuf.len == 0 && (nc->flags & MG_F_SEND_AND_CLOSE))) { mg_close_conn(nc); } } return now; } void mg_sl_if_get_conn_addr(struct mg_connection *nc, int remote, union socket_address *sa) { /* SimpleLink does not provide a way to get socket's peer address after * accept or connect. Address should have been preserved in the connection, * so we do our best here by using it. */ if (remote) memcpy(sa, &nc->sa, sizeof(*sa)); } void sl_restart_cb(struct mg_mgr *mgr) { /* * SimpleLink has been restarted, meaning all sockets have been invalidated. * We try our best - we'll restart the listeners, but for outgoing * connections we have no option but to terminate. */ struct mg_connection *nc; for (nc = mg_next(mgr, NULL); nc != NULL; nc = mg_next(mgr, nc)) { if (nc->sock == INVALID_SOCKET) continue; /* Could be a timer */ if (nc->flags & MG_F_LISTENING) { DBG(("restarting %p %s:%d", nc, inet_ntoa(nc->sa.sin.sin_addr), ntohs(nc->sa.sin.sin_port))); int res = (nc->flags & MG_F_UDP ? mg_sl_if_listen_udp(nc, &nc->sa) : mg_sl_if_listen_tcp(nc, &nc->sa)); if (res == 0) continue; /* Well, we tried and failed. Fall through to closing. */ } nc->sock = INVALID_SOCKET; DBG(("terminating %p %s:%d", nc, inet_ntoa(nc->sa.sin.sin_addr), ntohs(nc->sa.sin.sin_port))); /* TODO(rojer): Outgoing UDP? */ nc->flags |= MG_F_CLOSE_IMMEDIATELY; } } /* clang-format off */ #define MG_SL_IFACE_VTABLE \ { \ mg_sl_if_init, \ mg_sl_if_free, \ mg_sl_if_add_conn, \ mg_sl_if_remove_conn, \ mg_sl_if_poll, \ mg_sl_if_listen_tcp, \ mg_sl_if_listen_udp, \ mg_sl_if_connect_tcp, \ mg_sl_if_connect_udp, \ mg_sl_if_tcp_send, \ mg_sl_if_udp_send, \ mg_sl_if_recved, \ mg_sl_if_create_conn, \ mg_sl_if_destroy_conn, \ mg_sl_if_sock_set, \ mg_sl_if_get_conn_addr, \ } /* clang-format on */ const struct mg_iface_vtable mg_simplelink_iface_vtable = MG_SL_IFACE_VTABLE; #if MG_NET_IF == MG_NET_IF_SIMPLELINK const struct mg_iface_vtable mg_default_iface_vtable = MG_SL_IFACE_VTABLE; #endif #endif /* MG_ENABLE_NET_IF_SIMPLELINK */ #ifdef MG_MODULE_LINES #line 1 "common/platforms/simplelink/sl_ssl_if.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #if MG_ENABLE_SSL && MG_SSL_IF == MG_SSL_IF_SIMPLELINK /* Amalgamated: #include "common/mg_mem.h" */ #ifndef MG_SSL_IF_SIMPLELINK_SLFS_PREFIX #define MG_SSL_IF_SIMPLELINK_SLFS_PREFIX "SL:" #endif #define MG_SSL_IF_SIMPLELINK_SLFS_PREFIX_LEN \ (sizeof(MG_SSL_IF_SIMPLELINK_SLFS_PREFIX) - 1) struct mg_ssl_if_ctx { char *ssl_cert; char *ssl_key; char *ssl_ca_cert; char *ssl_server_name; }; void mg_ssl_if_init() { } enum mg_ssl_if_result mg_ssl_if_conn_init( struct mg_connection *nc, const struct mg_ssl_if_conn_params *params, const char **err_msg) { struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) MG_CALLOC(1, sizeof(*ctx)); if (ctx == NULL) { MG_SET_PTRPTR(err_msg, "Out of memory"); return MG_SSL_ERROR; } nc->ssl_if_data = ctx; if (params->cert != NULL || params->key != NULL) { if (params->cert != NULL && params->key != NULL) { ctx->ssl_cert = strdup(params->cert); ctx->ssl_key = strdup(params->key); } else { MG_SET_PTRPTR(err_msg, "Both cert and key are required."); return MG_SSL_ERROR; } } if (params->ca_cert != NULL && strcmp(params->ca_cert, "*") != 0) { ctx->ssl_ca_cert = strdup(params->ca_cert); } /* TODO(rojer): cipher_suites. */ if (params->server_name != NULL) { ctx->ssl_server_name = strdup(params->server_name); } return MG_SSL_OK; } void mg_ssl_if_conn_close_notify(struct mg_connection *nc) { /* Nothing to do */ (void) nc; } void mg_ssl_if_conn_free(struct mg_connection *nc) { struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) nc->ssl_if_data; if (ctx == NULL) return; nc->ssl_if_data = NULL; MG_FREE(ctx->ssl_cert); MG_FREE(ctx->ssl_key); MG_FREE(ctx->ssl_ca_cert); MG_FREE(ctx->ssl_server_name); memset(ctx, 0, sizeof(*ctx)); MG_FREE(ctx); } bool pem_to_der(const char *pem_file, const char *der_file) { bool ret = false; FILE *pf = NULL, *df = NULL; bool writing = false; pf = fopen(pem_file, "r"); if (pf == NULL) goto clean; remove(der_file); fs_slfs_set_new_file_size(der_file + MG_SSL_IF_SIMPLELINK_SLFS_PREFIX_LEN, 2048); df = fopen(der_file, "w"); if (df == NULL) goto clean; while (1) { char pem_buf[70]; char der_buf[48]; if (!fgets(pem_buf, sizeof(pem_buf), pf)) break; if (writing) { if (strstr(pem_buf, "-----END ") != NULL) { ret = true; break; } int l = 0; while (!isspace((unsigned int) pem_buf[l])) l++; int der_len = 0; cs_base64_decode((const unsigned char *) pem_buf, sizeof(pem_buf), der_buf, &der_len); if (der_len <= 0) break; if (fwrite(der_buf, 1, der_len, df) != der_len) break; } else if (strstr(pem_buf, "-----BEGIN ") != NULL) { writing = true; } } clean: if (pf != NULL) fclose(pf); if (df != NULL) { fclose(df); if (!ret) remove(der_file); } return ret; } #if MG_ENABLE_FILESYSTEM && defined(MG_FS_SLFS) /* If the file's extension is .pem, convert it to DER format and put on SLFS. */ static char *sl_pem2der(const char *pem_file) { const char *pem_ext = strstr(pem_file, ".pem"); if (pem_ext == NULL || *(pem_ext + 4) != '\0') { return strdup(pem_file); } char *der_file = NULL; /* DER file must be located on SLFS, add prefix. */ int l = mg_asprintf(&der_file, 0, MG_SSL_IF_SIMPLELINK_SLFS_PREFIX "%.*s.der", (int) (pem_ext - pem_file), pem_file); if (der_file == NULL) return NULL; bool result = false; cs_stat_t st; if (mg_stat(der_file, &st) != 0) { result = pem_to_der(pem_file, der_file); LOG(LL_DEBUG, ("%s -> %s = %d", pem_file, der_file, result)); } else { /* File exists, assume it's already been converted. */ result = true; } if (result) { /* Strip the SL: prefix we added since NWP does not expect it. */ memmove(der_file, der_file + MG_SSL_IF_SIMPLELINK_SLFS_PREFIX_LEN, l - 2 /* including \0 */); } else { MG_FREE(der_file); der_file = NULL; } return der_file; } #else static char *sl_pem2der(const char *pem_file) { return strdup(pem_file); } #endif int sl_set_ssl_opts(int sock, struct mg_connection *nc) { int err; const struct mg_ssl_if_ctx *ctx = (struct mg_ssl_if_ctx *) nc->ssl_if_data; DBG(("%p ssl ctx: %p", nc, ctx)); if (ctx != NULL) { DBG(("%p %s,%s,%s,%s", nc, (ctx->ssl_cert ? ctx->ssl_cert : "-"), (ctx->ssl_key ? ctx->ssl_cert : "-"), (ctx->ssl_ca_cert ? ctx->ssl_ca_cert : "-"), (ctx->ssl_server_name ? ctx->ssl_server_name : "-"))); if (ctx->ssl_cert != NULL && ctx->ssl_key != NULL) { char *ssl_cert = sl_pem2der(ctx->ssl_cert); char *ssl_key = sl_pem2der(ctx->ssl_key); if (ssl_cert != NULL && ssl_key != NULL) { err = sl_SetSockOpt(sock, SL_SOL_SOCKET, SL_SO_SECURE_FILES_CERTIFICATE_FILE_NAME, ssl_cert, strlen(ssl_cert)); LOG(LL_INFO, ("CERTIFICATE_FILE_NAME %s -> %d", ssl_cert, err)); err = sl_SetSockOpt(sock, SL_SOL_SOCKET, SL_SO_SECURE_FILES_PRIVATE_KEY_FILE_NAME, ssl_key, strlen(ssl_key)); LOG(LL_INFO, ("PRIVATE_KEY_FILE_NAME %s -> %d", ssl_key, err)); } else { err = -1; } MG_FREE(ssl_cert); MG_FREE(ssl_key); if (err != 0) return err; } if (ctx->ssl_ca_cert != NULL) { if (ctx->ssl_ca_cert[0] != '\0') { char *ssl_ca_cert = sl_pem2der(ctx->ssl_ca_cert); if (ssl_ca_cert != NULL) { err = sl_SetSockOpt(sock, SL_SOL_SOCKET, SL_SO_SECURE_FILES_CA_FILE_NAME, ssl_ca_cert, strlen(ssl_ca_cert)); LOG(LL_INFO, ("CA_FILE_NAME %s -> %d", ssl_ca_cert, err)); } else { err = -1; } MG_FREE(ssl_ca_cert); if (err != 0) return err; } } if (ctx->ssl_server_name != NULL) { err = sl_SetSockOpt(sock, SL_SOL_SOCKET, SL_SO_SECURE_DOMAIN_NAME_VERIFICATION, ctx->ssl_server_name, strlen(ctx->ssl_server_name)); DBG(("DOMAIN_NAME_VERIFICATION %s -> %d", ctx->ssl_server_name, err)); /* Domain name verificationw as added in a NWP service pack, older * versions return SL_ERROR_BSD_ENOPROTOOPT. There isn't much we can do * about it, * so we ignore the error. */ if (err != 0 && err != SL_ERROR_BSD_ENOPROTOOPT) return err; } } return 0; } #endif /* MG_ENABLE_SSL && MG_SSL_IF == MG_SSL_IF_SIMPLELINK */ #ifdef MG_MODULE_LINES #line 1 "common/platforms/lwip/mg_lwip_net_if.h" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #ifndef CS_COMMON_PLATFORMS_LWIP_MG_NET_IF_LWIP_H_ #define CS_COMMON_PLATFORMS_LWIP_MG_NET_IF_LWIP_H_ #ifndef MG_ENABLE_NET_IF_LWIP_LOW_LEVEL #define MG_ENABLE_NET_IF_LWIP_LOW_LEVEL MG_NET_IF == MG_NET_IF_LWIP_LOW_LEVEL #endif #if MG_ENABLE_NET_IF_LWIP_LOW_LEVEL #include extern const struct mg_iface_vtable mg_lwip_iface_vtable; struct mg_lwip_conn_state { struct mg_connection *nc; struct mg_connection *lc; union { struct tcp_pcb *tcp; struct udp_pcb *udp; } pcb; err_t err; size_t num_sent; /* Number of acknowledged bytes to be reported to the core */ struct pbuf *rx_chain; /* Chain of incoming data segments. */ size_t rx_offset; /* Offset within the first pbuf (if partially consumed) */ /* Last SSL write size, for retries. */ int last_ssl_write_size; /* Whether MG_SIG_RECV is already pending for this connection */ int recv_pending : 1; /* Whether the connection is about to close, just `rx_chain` needs to drain */ int draining_rx_chain : 1; }; enum mg_sig_type { MG_SIG_CONNECT_RESULT = 1, MG_SIG_RECV = 2, MG_SIG_CLOSE_CONN = 3, MG_SIG_TOMBSTONE = 4, MG_SIG_ACCEPT = 5, }; void mg_lwip_post_signal(enum mg_sig_type sig, struct mg_connection *nc); /* To be implemented by the platform. */ void mg_lwip_mgr_schedule_poll(struct mg_mgr *mgr); #endif /* MG_ENABLE_NET_IF_LWIP_LOW_LEVEL */ #endif /* CS_COMMON_PLATFORMS_LWIP_MG_NET_IF_LWIP_H_ */ #ifdef MG_MODULE_LINES #line 1 "common/platforms/lwip/mg_lwip_net_if.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #if MG_ENABLE_NET_IF_LWIP_LOW_LEVEL /* Amalgamated: #include "common/mg_mem.h" */ #include #include #include #include #if ((LWIP_VERSION_MAJOR << 8) | LWIP_VERSION_MINOR) >= 0x0105 #include /* For tcp_seg */ #else #include #endif #include /* Amalgamated: #include "common/cs_dbg.h" */ /* * Newest versions of LWIP have ip_2_ip4, older have ipX_2_ip, * even older have nothing. */ #ifndef ip_2_ip4 #ifdef ipX_2_ip #define ip_2_ip4(addr) ipX_2_ip(addr) #else #define ip_2_ip4(addr) (addr) #endif #endif /* * Depending on whether Mongoose is compiled with ipv6 support, use right * lwip functions */ #if MG_ENABLE_IPV6 #define TCP_NEW tcp_new_ip6 #define TCP_BIND tcp_bind_ip6 #define UDP_BIND udp_bind_ip6 #define IPADDR_NTOA(x) ip6addr_ntoa((const ip6_addr_t *)(x)) #define SET_ADDR(dst, src) \ memcpy((dst)->sin6.sin6_addr.s6_addr, (src)->ip6.addr, \ sizeof((dst)->sin6.sin6_addr.s6_addr)) #else #define TCP_NEW tcp_new #define TCP_BIND tcp_bind #define UDP_BIND udp_bind #define IPADDR_NTOA ipaddr_ntoa #define SET_ADDR(dst, src) (dst)->sin.sin_addr.s_addr = ip_2_ip4(src)->addr #endif #if NO_SYS #define tcpip_callback(fn, arg) (fn)(arg) typedef void (*tcpip_callback_fn)(void *arg); #endif void mg_lwip_ssl_do_hs(struct mg_connection *nc); void mg_lwip_ssl_send(struct mg_connection *nc); void mg_lwip_ssl_recv(struct mg_connection *nc); void mg_lwip_if_init(struct mg_iface *iface); void mg_lwip_if_free(struct mg_iface *iface); void mg_lwip_if_add_conn(struct mg_connection *nc); void mg_lwip_if_remove_conn(struct mg_connection *nc); time_t mg_lwip_if_poll(struct mg_iface *iface, int timeout_ms); #if defined(RTOS_SDK) || defined(ESP_PLATFORM) extern void mgos_lock(); extern void mgos_unlock(); #else #define mgos_lock() #define mgos_unlock() #endif static void mg_lwip_recv_common(struct mg_connection *nc, struct pbuf *p); #if LWIP_TCP_KEEPALIVE void mg_lwip_set_keepalive_params(struct mg_connection *nc, int idle, int interval, int count) { if (nc->sock == INVALID_SOCKET || nc->flags & MG_F_UDP) { return; } struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; struct tcp_pcb *tpcb = cs->pcb.tcp; if (idle > 0 && interval > 0 && count > 0) { tpcb->keep_idle = idle * 1000; tpcb->keep_intvl = interval * 1000; tpcb->keep_cnt = count; tpcb->so_options |= SOF_KEEPALIVE; } else { tpcb->so_options &= ~SOF_KEEPALIVE; } } #elif !defined(MG_NO_LWIP_TCP_KEEPALIVE) #warning LWIP TCP keepalive is disabled. Please consider enabling it. #endif /* LWIP_TCP_KEEPALIVE */ static err_t mg_lwip_tcp_conn_cb(void *arg, struct tcp_pcb *tpcb, err_t err) { struct mg_connection *nc = (struct mg_connection *) arg; DBG(("%p connect to %s:%u = %d", nc, IPADDR_NTOA(ipX_2_ip(&tpcb->remote_ip)), tpcb->remote_port, err)); if (nc == NULL) { tcp_abort(tpcb); return ERR_ARG; } struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; cs->err = err; #if LWIP_TCP_KEEPALIVE if (err == 0) mg_lwip_set_keepalive_params(nc, 60, 10, 6); #endif mg_lwip_post_signal(MG_SIG_CONNECT_RESULT, nc); return ERR_OK; } static void mg_lwip_tcp_error_cb(void *arg, err_t err) { struct mg_connection *nc = (struct mg_connection *) arg; DBG(("%p conn error %d", nc, err)); if (nc == NULL) return; struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; cs->pcb.tcp = NULL; /* Has already been deallocated */ if (nc->flags & MG_F_CONNECTING) { cs->err = err; mg_lwip_post_signal(MG_SIG_CONNECT_RESULT, nc); } else { mg_lwip_post_signal(MG_SIG_CLOSE_CONN, nc); } } static err_t mg_lwip_tcp_recv_cb(void *arg, struct tcp_pcb *tpcb, struct pbuf *p, err_t err) { struct mg_connection *nc = (struct mg_connection *) arg; DBG(("%p %p %u %d", nc, tpcb, (p != NULL ? p->tot_len : 0), err)); if (p == NULL) { if (nc != NULL && !(nc->flags & MG_F_CLOSE_IMMEDIATELY)) { struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; if (cs->rx_chain != NULL) { /* * rx_chain still contains non-consumed data, don't close the * connection */ cs->draining_rx_chain = 1; } else { mg_lwip_post_signal(MG_SIG_CLOSE_CONN, nc); } } else { /* Tombstoned connection, do nothing. */ } return ERR_OK; } else if (nc == NULL) { tcp_abort(tpcb); return ERR_ARG; } struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; /* * If we get a chain of more than one segment at once, we need to bump * refcount on the subsequent bufs to make them independent. */ if (p->next != NULL) { struct pbuf *q = p->next; for (; q != NULL; q = q->next) pbuf_ref(q); } mgos_lock(); if (cs->rx_chain == NULL) { cs->rx_offset = 0; } else if (pbuf_clen(cs->rx_chain) >= 4) { /* ESP SDK has a limited pool of 5 pbufs. We must not hog them all or RX * will be completely blocked. We already have at least 4 in the chain, * this one is, so we have to make a copy and release this one. */ struct pbuf *np = pbuf_alloc(PBUF_RAW, p->tot_len, PBUF_RAM); if (np != NULL) { pbuf_copy(np, p); pbuf_free(p); p = np; } } mgos_unlock(); mg_lwip_recv_common(nc, p); return ERR_OK; } static void mg_lwip_consume_rx_chain_tcp(struct mg_connection *nc) { struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; if (cs->rx_chain == NULL) return; #if MG_ENABLE_SSL if (nc->flags & MG_F_SSL) { if (nc->flags & MG_F_SSL_HANDSHAKE_DONE) { mg_lwip_ssl_recv(nc); } else { mg_lwip_ssl_do_hs(nc); } return; } #endif mgos_lock(); while (cs->rx_chain != NULL && nc->recv_mbuf.len < nc->recv_mbuf_limit) { struct pbuf *seg = cs->rx_chain; size_t seg_len = (seg->len - cs->rx_offset); size_t buf_avail = (nc->recv_mbuf_limit - nc->recv_mbuf.len); size_t len = MIN(seg_len, buf_avail); char *data = (char *) MG_MALLOC(len); if (data == NULL) { mgos_unlock(); DBG(("OOM")); return; } pbuf_copy_partial(seg, data, len, cs->rx_offset); cs->rx_offset += len; if (cs->rx_offset == cs->rx_chain->len) { cs->rx_chain = pbuf_dechain(cs->rx_chain); pbuf_free(seg); cs->rx_offset = 0; } mgos_unlock(); mg_if_recv_tcp_cb(nc, data, len, 1 /* own */); mgos_lock(); } mgos_unlock(); } static void mg_lwip_handle_recv_tcp(struct mg_connection *nc) { mg_lwip_consume_rx_chain_tcp(nc); if (nc->send_mbuf.len > 0) { mg_lwip_mgr_schedule_poll(nc->mgr); } } static err_t mg_lwip_tcp_sent_cb(void *arg, struct tcp_pcb *tpcb, u16_t num_sent) { struct mg_connection *nc = (struct mg_connection *) arg; DBG(("%p %p %u %p %p", nc, tpcb, num_sent, tpcb->unsent, tpcb->unacked)); if (nc == NULL) return ERR_OK; if ((nc->flags & MG_F_SEND_AND_CLOSE) && !(nc->flags & MG_F_WANT_WRITE) && nc->send_mbuf.len == 0 && tpcb->unsent == NULL && tpcb->unacked == NULL) { mg_lwip_post_signal(MG_SIG_CLOSE_CONN, nc); } return ERR_OK; } struct mg_lwip_if_connect_tcp_ctx { struct mg_connection *nc; const union socket_address *sa; }; static void mg_lwip_if_connect_tcp_tcpip(void *arg) { struct mg_lwip_if_connect_tcp_ctx *ctx = (struct mg_lwip_if_connect_tcp_ctx *) arg; struct mg_connection *nc = ctx->nc; const union socket_address *sa = ctx->sa; struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; struct tcp_pcb *tpcb = TCP_NEW(); cs->pcb.tcp = tpcb; ip_addr_t *ip = (ip_addr_t *) &sa->sin.sin_addr.s_addr; u16_t port = ntohs(sa->sin.sin_port); tcp_arg(tpcb, nc); tcp_err(tpcb, mg_lwip_tcp_error_cb); tcp_sent(tpcb, mg_lwip_tcp_sent_cb); tcp_recv(tpcb, mg_lwip_tcp_recv_cb); cs->err = TCP_BIND(tpcb, IP_ADDR_ANY, 0 /* any port */); DBG(("%p tcp_bind = %d", nc, cs->err)); if (cs->err != ERR_OK) { mg_lwip_post_signal(MG_SIG_CONNECT_RESULT, nc); return; } cs->err = tcp_connect(tpcb, ip, port, mg_lwip_tcp_conn_cb); DBG(("%p tcp_connect %p = %d", nc, tpcb, cs->err)); if (cs->err != ERR_OK) { mg_lwip_post_signal(MG_SIG_CONNECT_RESULT, nc); return; } } void mg_lwip_if_connect_tcp(struct mg_connection *nc, const union socket_address *sa) { struct mg_lwip_if_connect_tcp_ctx ctx = {.nc = nc, .sa = sa}; tcpip_callback(mg_lwip_if_connect_tcp_tcpip, &ctx); } /* * Lwip included in the SDKs for nRF5x chips has different type for the * callback of `udp_recv()` */ #if ((LWIP_VERSION_MAJOR << 8) | LWIP_VERSION_MINOR) >= 0x0105 static void mg_lwip_udp_recv_cb(void *arg, struct udp_pcb *pcb, struct pbuf *p, const ip_addr_t *addr, u16_t port) #else static void mg_lwip_udp_recv_cb(void *arg, struct udp_pcb *pcb, struct pbuf *p, ip_addr_t *addr, u16_t port) #endif { struct mg_connection *nc = (struct mg_connection *) arg; DBG(("%p %s:%u %p %u %u", nc, IPADDR_NTOA(addr), port, p, p->ref, p->len)); /* Put address in a separate pbuf and tack it onto the packet. */ struct pbuf *sap = pbuf_alloc(PBUF_RAW, sizeof(union socket_address), PBUF_RAM); if (sap == NULL) { pbuf_free(p); return; } union socket_address *sa = (union socket_address *) sap->payload; #if ((LWIP_VERSION_MAJOR << 8) | LWIP_VERSION_MINOR) >= 0x0105 sa->sin.sin_addr.s_addr = ip_2_ip4(addr)->addr; #else sa->sin.sin_addr.s_addr = addr->addr; #endif sa->sin.sin_port = htons(port); /* Logic in the recv handler requires that there be exactly one data pbuf. */ p = pbuf_coalesce(p, PBUF_RAW); pbuf_chain(sap, p); mg_lwip_recv_common(nc, sap); (void) pcb; } static void mg_lwip_recv_common(struct mg_connection *nc, struct pbuf *p) { struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; mgos_lock(); if (cs->rx_chain == NULL) { cs->rx_chain = p; } else { pbuf_chain(cs->rx_chain, p); } if (!cs->recv_pending) { cs->recv_pending = 1; mg_lwip_post_signal(MG_SIG_RECV, nc); } mgos_unlock(); } static void mg_lwip_handle_recv_udp(struct mg_connection *nc) { struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; /* * For UDP, RX chain consists of interleaved address and packet bufs: * Address pbuf followed by exactly one data pbuf (recv_cb took care of that). */ while (cs->rx_chain != NULL) { struct pbuf *sap = cs->rx_chain; struct pbuf *p = sap->next; cs->rx_chain = pbuf_dechain(p); size_t data_len = p->len; char *data = (char *) MG_MALLOC(data_len); if (data != NULL) { pbuf_copy_partial(p, data, data_len, 0); pbuf_free(p); mg_if_recv_udp_cb(nc, data, data_len, (union socket_address *) sap->payload, sap->len); pbuf_free(sap); } else { pbuf_free(p); pbuf_free(sap); } } } static void mg_lwip_if_connect_udp_tcpip(void *arg) { struct mg_connection *nc = (struct mg_connection *) arg; struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; struct udp_pcb *upcb = udp_new(); cs->err = UDP_BIND(upcb, IP_ADDR_ANY, 0 /* any port */); DBG(("%p udp_bind %p = %d", nc, upcb, cs->err)); if (cs->err == ERR_OK) { udp_recv(upcb, mg_lwip_udp_recv_cb, nc); cs->pcb.udp = upcb; } else { udp_remove(upcb); } mg_lwip_post_signal(MG_SIG_CONNECT_RESULT, nc); } void mg_lwip_if_connect_udp(struct mg_connection *nc) { tcpip_callback(mg_lwip_if_connect_udp_tcpip, nc); } void mg_lwip_accept_conn(struct mg_connection *nc, struct tcp_pcb *tpcb) { union socket_address sa; SET_ADDR(&sa, &tpcb->remote_ip); sa.sin.sin_port = htons(tpcb->remote_port); mg_if_accept_tcp_cb(nc, &sa, sizeof(sa.sin)); } static void tcp_close_tcpip(void *arg) { tcp_close((struct tcp_pcb *) arg); } void mg_lwip_handle_accept(struct mg_connection *nc) { struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; if (cs->pcb.tcp == NULL) return; #if MG_ENABLE_SSL if (cs->lc->flags & MG_F_SSL) { if (mg_ssl_if_conn_accept(nc, cs->lc) != MG_SSL_OK) { LOG(LL_ERROR, ("SSL error")); tcpip_callback(tcp_close_tcpip, cs->pcb.tcp); } } else #endif { mg_lwip_accept_conn(nc, cs->pcb.tcp); } } static err_t mg_lwip_accept_cb(void *arg, struct tcp_pcb *newtpcb, err_t err) { struct mg_connection *lc = (struct mg_connection *) arg, *nc; struct mg_lwip_conn_state *lcs, *cs; struct tcp_pcb_listen *lpcb; LOG(LL_DEBUG, ("%p conn %p from %s:%u", lc, newtpcb, IPADDR_NTOA(ipX_2_ip(&newtpcb->remote_ip)), newtpcb->remote_port)); if (lc == NULL) { tcp_abort(newtpcb); return ERR_ABRT; } lcs = (struct mg_lwip_conn_state *) lc->sock; lpcb = (struct tcp_pcb_listen *) lcs->pcb.tcp; #if TCP_LISTEN_BACKLOG tcp_accepted(lpcb); #endif nc = mg_if_accept_new_conn(lc); if (nc == NULL) { tcp_abort(newtpcb); return ERR_ABRT; } cs = (struct mg_lwip_conn_state *) nc->sock; cs->lc = lc; cs->pcb.tcp = newtpcb; /* We need to set up callbacks before returning because data may start * arriving immediately. */ tcp_arg(newtpcb, nc); tcp_err(newtpcb, mg_lwip_tcp_error_cb); tcp_sent(newtpcb, mg_lwip_tcp_sent_cb); tcp_recv(newtpcb, mg_lwip_tcp_recv_cb); #if LWIP_TCP_KEEPALIVE mg_lwip_set_keepalive_params(nc, 60, 10, 6); #endif mg_lwip_post_signal(MG_SIG_ACCEPT, nc); (void) err; (void) lpcb; return ERR_OK; } struct mg_lwip_if_listen_ctx { struct mg_connection *nc; union socket_address *sa; int ret; }; static void mg_lwip_if_listen_tcp_tcpip(void *arg) { struct mg_lwip_if_listen_ctx *ctx = (struct mg_lwip_if_listen_ctx *) arg; struct mg_connection *nc = ctx->nc; union socket_address *sa = ctx->sa; struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; struct tcp_pcb *tpcb = TCP_NEW(); ip_addr_t *ip = (ip_addr_t *) &sa->sin.sin_addr.s_addr; u16_t port = ntohs(sa->sin.sin_port); cs->err = TCP_BIND(tpcb, ip, port); DBG(("%p tcp_bind(%s:%u) = %d", nc, IPADDR_NTOA(ip), port, cs->err)); if (cs->err != ERR_OK) { tcp_close(tpcb); ctx->ret = -1; return; } tcp_arg(tpcb, nc); tpcb = tcp_listen(tpcb); cs->pcb.tcp = tpcb; tcp_accept(tpcb, mg_lwip_accept_cb); ctx->ret = 0; } int mg_lwip_if_listen_tcp(struct mg_connection *nc, union socket_address *sa) { struct mg_lwip_if_listen_ctx ctx = {.nc = nc, .sa = sa}; tcpip_callback(mg_lwip_if_listen_tcp_tcpip, &ctx); return ctx.ret; } static void mg_lwip_if_listen_udp_tcpip(void *arg) { struct mg_lwip_if_listen_ctx *ctx = (struct mg_lwip_if_listen_ctx *) arg; struct mg_connection *nc = ctx->nc; union socket_address *sa = ctx->sa; struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; struct udp_pcb *upcb = udp_new(); ip_addr_t *ip = (ip_addr_t *) &sa->sin.sin_addr.s_addr; u16_t port = ntohs(sa->sin.sin_port); cs->err = UDP_BIND(upcb, ip, port); DBG(("%p udb_bind(%s:%u) = %d", nc, IPADDR_NTOA(ip), port, cs->err)); if (cs->err != ERR_OK) { udp_remove(upcb); ctx->ret = -1; } else { udp_recv(upcb, mg_lwip_udp_recv_cb, nc); cs->pcb.udp = upcb; ctx->ret = 0; } } int mg_lwip_if_listen_udp(struct mg_connection *nc, union socket_address *sa) { struct mg_lwip_if_listen_ctx ctx = {.nc = nc, .sa = sa}; tcpip_callback(mg_lwip_if_listen_udp_tcpip, &ctx); return ctx.ret; } struct mg_lwip_tcp_write_ctx { struct mg_connection *nc; const void *data; uint16_t len; int ret; }; static void tcp_output_tcpip(void *arg) { tcp_output((struct tcp_pcb *) arg); } static void mg_lwip_tcp_write_tcpip(void *arg) { struct mg_lwip_tcp_write_ctx *ctx = (struct mg_lwip_tcp_write_ctx *) arg; struct mg_connection *nc = ctx->nc; struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; struct tcp_pcb *tpcb = cs->pcb.tcp; size_t len = MIN(tpcb->mss, MIN(ctx->len, tpcb->snd_buf)); size_t unsent, unacked; if (len == 0) { DBG(("%p no buf avail %u %u %p %p", tpcb, tpcb->snd_buf, tpcb->snd_queuelen, tpcb->unsent, tpcb->unacked)); tcpip_callback(tcp_output_tcpip, tpcb); ctx->ret = 0; return; } unsent = (tpcb->unsent != NULL ? tpcb->unsent->len : 0); unacked = (tpcb->unacked != NULL ? tpcb->unacked->len : 0); /* * On ESP8266 we only allow one TCP segment in flight at any given time. * This may increase latency and reduce efficiency of tcp windowing, * but memory is scarce and precious on that platform so we do this to * reduce footprint. */ #if CS_PLATFORM == CS_P_ESP8266 if (unacked > 0) { ctx->ret = 0; return; } len = MIN(len, (TCP_MSS - unsent)); #endif cs->err = tcp_write(tpcb, ctx->data, len, TCP_WRITE_FLAG_COPY); unsent = (tpcb->unsent != NULL ? tpcb->unsent->len : 0); unacked = (tpcb->unacked != NULL ? tpcb->unacked->len : 0); DBG(("%p tcp_write %u = %d, %u %u", tpcb, len, cs->err, unsent, unacked)); if (cs->err != ERR_OK) { /* * We ignore ERR_MEM because memory will be freed up when the data is sent * and we'll retry. */ ctx->ret = (cs->err == ERR_MEM ? 0 : -1); return; } ctx->ret = len; } static int mg_lwip_tcp_write(struct mg_connection *nc, const void *data, uint16_t len) { struct mg_lwip_tcp_write_ctx ctx = {.nc = nc, .data = data, .len = len}; struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; struct tcp_pcb *tpcb = cs->pcb.tcp; if (tpcb == NULL) { return -1; } tcpip_callback(mg_lwip_tcp_write_tcpip, &ctx); return ctx.ret; } struct udp_sendto_ctx { struct udp_pcb *upcb; struct pbuf *p; ip_addr_t *ip; uint16_t port; int ret; }; static void udp_sendto_tcpip(void *arg) { struct udp_sendto_ctx *ctx = (struct udp_sendto_ctx *) arg; ctx->ret = udp_sendto(ctx->upcb, ctx->p, ctx->ip, ctx->port); } static int mg_lwip_udp_send(struct mg_connection *nc, const void *data, uint16_t len) { struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; if (cs->pcb.udp == NULL) { /* * In case of UDP, this usually means, what * async DNS resolve is still in progress and connection * is not ready yet */ DBG(("%p socket is not connected", nc)); return -1; } struct udp_pcb *upcb = cs->pcb.udp; struct pbuf *p = pbuf_alloc(PBUF_TRANSPORT, len, PBUF_RAM); #if defined(LWIP_IPV4) && LWIP_IPV4 && defined(LWIP_IPV6) && LWIP_IPV6 ip_addr_t ip = {.u_addr.ip4.addr = nc->sa.sin.sin_addr.s_addr, .type = 0}; #else ip_addr_t ip = {.addr = nc->sa.sin.sin_addr.s_addr}; #endif u16_t port = ntohs(nc->sa.sin.sin_port); if (p == NULL) { DBG(("OOM")); return 0; } memcpy(p->payload, data, len); struct udp_sendto_ctx ctx = {.upcb = upcb, .p = p, .ip = &ip, .port = port}; tcpip_callback(udp_sendto_tcpip, &ctx); cs->err = ctx.ret; pbuf_free(p); return (cs->err == ERR_OK ? len : -1); } static void mg_lwip_send_more(struct mg_connection *nc) { int num_sent = 0; if (nc->sock == INVALID_SOCKET) return; if (nc->flags & MG_F_UDP) { num_sent = mg_lwip_udp_send(nc, nc->send_mbuf.buf, nc->send_mbuf.len); DBG(("%p mg_lwip_udp_send %u = %d", nc, nc->send_mbuf.len, num_sent)); } else { num_sent = mg_lwip_tcp_write(nc, nc->send_mbuf.buf, nc->send_mbuf.len); DBG(("%p mg_lwip_tcp_write %u = %d", nc, nc->send_mbuf.len, num_sent)); } if (num_sent == 0) return; if (num_sent > 0) { mg_if_sent_cb(nc, num_sent); } else { mg_lwip_post_signal(MG_SIG_CLOSE_CONN, nc); } } void mg_lwip_if_tcp_send(struct mg_connection *nc, const void *buf, size_t len) { mbuf_append(&nc->send_mbuf, buf, len); mg_lwip_mgr_schedule_poll(nc->mgr); } void mg_lwip_if_udp_send(struct mg_connection *nc, const void *buf, size_t len) { mbuf_append(&nc->send_mbuf, buf, len); mg_lwip_mgr_schedule_poll(nc->mgr); } struct tcp_recved_ctx { struct tcp_pcb *tpcb; size_t len; }; void tcp_recved_tcpip(void *arg) { struct tcp_recved_ctx *ctx = (struct tcp_recved_ctx *) arg; tcp_recved(ctx->tpcb, ctx->len); } void mg_lwip_if_recved(struct mg_connection *nc, size_t len) { if (nc->flags & MG_F_UDP) return; struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; if (nc->sock == INVALID_SOCKET || cs->pcb.tcp == NULL) { DBG(("%p invalid socket", nc)); return; } DBG(("%p %p %u %u", nc, cs->pcb.tcp, len, (cs->rx_chain ? cs->rx_chain->tot_len : 0))); struct tcp_recved_ctx ctx = {.tpcb = cs->pcb.tcp, .len = len}; #if MG_ENABLE_SSL if (!(nc->flags & MG_F_SSL)) { tcpip_callback(tcp_recved_tcpip, &ctx); } else { /* Currently SSL acknowledges data immediately. * TODO(rojer): Find a way to propagate mg_lwip_if_recved. */ } #else tcpip_callback(tcp_recved_tcpip, &ctx); #endif mbuf_trim(&nc->recv_mbuf); } int mg_lwip_if_create_conn(struct mg_connection *nc) { struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) MG_CALLOC(1, sizeof(*cs)); if (cs == NULL) return 0; cs->nc = nc; nc->sock = (intptr_t) cs; return 1; } static void udp_remove_tcpip(void *arg) { udp_remove((struct udp_pcb *) arg); } void mg_lwip_if_destroy_conn(struct mg_connection *nc) { if (nc->sock == INVALID_SOCKET) return; struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; if (!(nc->flags & MG_F_UDP)) { struct tcp_pcb *tpcb = cs->pcb.tcp; if (tpcb != NULL) { tcp_arg(tpcb, NULL); DBG(("%p tcp_close %p", nc, tpcb)); tcp_arg(tpcb, NULL); tcpip_callback(tcp_close_tcpip, tpcb); } while (cs->rx_chain != NULL) { struct pbuf *seg = cs->rx_chain; cs->rx_chain = pbuf_dechain(cs->rx_chain); pbuf_free(seg); } memset(cs, 0, sizeof(*cs)); MG_FREE(cs); } else if (nc->listener == NULL) { /* Only close outgoing UDP pcb or listeners. */ struct udp_pcb *upcb = cs->pcb.udp; if (upcb != NULL) { DBG(("%p udp_remove %p", nc, upcb)); tcpip_callback(udp_remove_tcpip, upcb); } memset(cs, 0, sizeof(*cs)); MG_FREE(cs); } nc->sock = INVALID_SOCKET; } void mg_lwip_if_get_conn_addr(struct mg_connection *nc, int remote, union socket_address *sa) { memset(sa, 0, sizeof(*sa)); if (nc == NULL || nc->sock == INVALID_SOCKET) return; struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; if (nc->flags & MG_F_UDP) { struct udp_pcb *upcb = cs->pcb.udp; if (remote) { memcpy(sa, &nc->sa, sizeof(*sa)); } else if (upcb != NULL) { sa->sin.sin_port = htons(upcb->local_port); SET_ADDR(sa, &upcb->local_ip); } } else { struct tcp_pcb *tpcb = cs->pcb.tcp; if (remote) { memcpy(sa, &nc->sa, sizeof(*sa)); } else if (tpcb != NULL) { sa->sin.sin_port = htons(tpcb->local_port); SET_ADDR(sa, &tpcb->local_ip); } } } void mg_lwip_if_sock_set(struct mg_connection *nc, sock_t sock) { nc->sock = sock; } /* clang-format off */ #define MG_LWIP_IFACE_VTABLE \ { \ mg_lwip_if_init, \ mg_lwip_if_free, \ mg_lwip_if_add_conn, \ mg_lwip_if_remove_conn, \ mg_lwip_if_poll, \ mg_lwip_if_listen_tcp, \ mg_lwip_if_listen_udp, \ mg_lwip_if_connect_tcp, \ mg_lwip_if_connect_udp, \ mg_lwip_if_tcp_send, \ mg_lwip_if_udp_send, \ mg_lwip_if_recved, \ mg_lwip_if_create_conn, \ mg_lwip_if_destroy_conn, \ mg_lwip_if_sock_set, \ mg_lwip_if_get_conn_addr, \ } /* clang-format on */ const struct mg_iface_vtable mg_lwip_iface_vtable = MG_LWIP_IFACE_VTABLE; #if MG_NET_IF == MG_NET_IF_LWIP_LOW_LEVEL const struct mg_iface_vtable mg_default_iface_vtable = MG_LWIP_IFACE_VTABLE; #endif #endif /* MG_ENABLE_NET_IF_LWIP_LOW_LEVEL */ #ifdef MG_MODULE_LINES #line 1 "common/platforms/lwip/mg_lwip_ev_mgr.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #if MG_NET_IF == MG_NET_IF_LWIP_LOW_LEVEL #ifndef MG_SIG_QUEUE_LEN #define MG_SIG_QUEUE_LEN 32 #endif struct mg_ev_mgr_lwip_signal { int sig; struct mg_connection *nc; }; struct mg_ev_mgr_lwip_data { struct mg_ev_mgr_lwip_signal sig_queue[MG_SIG_QUEUE_LEN]; int sig_queue_len; int start_index; }; void mg_lwip_post_signal(enum mg_sig_type sig, struct mg_connection *nc) { struct mg_ev_mgr_lwip_data *md = (struct mg_ev_mgr_lwip_data *) nc->iface->data; mgos_lock(); if (md->sig_queue_len >= MG_SIG_QUEUE_LEN) { mgos_unlock(); return; } int end_index = (md->start_index + md->sig_queue_len) % MG_SIG_QUEUE_LEN; md->sig_queue[end_index].sig = sig; md->sig_queue[end_index].nc = nc; md->sig_queue_len++; mg_lwip_mgr_schedule_poll(nc->mgr); mgos_unlock(); } void mg_ev_mgr_lwip_process_signals(struct mg_mgr *mgr) { struct mg_ev_mgr_lwip_data *md = (struct mg_ev_mgr_lwip_data *) mgr->ifaces[MG_MAIN_IFACE]->data; while (md->sig_queue_len > 0) { mgos_lock(); int sig = md->sig_queue[md->start_index].sig; struct mg_connection *nc = md->sig_queue[md->start_index].nc; struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; md->start_index = (md->start_index + 1) % MG_SIG_QUEUE_LEN; md->sig_queue_len--; mgos_unlock(); if (nc->iface == NULL || nc->mgr == NULL) continue; switch (sig) { case MG_SIG_CONNECT_RESULT: { #if MG_ENABLE_SSL if (cs->err == 0 && (nc->flags & MG_F_SSL) && !(nc->flags & MG_F_SSL_HANDSHAKE_DONE)) { mg_lwip_ssl_do_hs(nc); } else #endif { mg_if_connect_cb(nc, cs->err); } break; } case MG_SIG_CLOSE_CONN: { nc->flags |= MG_F_SEND_AND_CLOSE; mg_close_conn(nc); break; } case MG_SIG_RECV: { cs->recv_pending = 0; if (nc->flags & MG_F_UDP) { mg_lwip_handle_recv_udp(nc); } else { mg_lwip_handle_recv_tcp(nc); } break; } case MG_SIG_TOMBSTONE: { break; } case MG_SIG_ACCEPT: { mg_lwip_handle_accept(nc); break; } } } } void mg_lwip_if_init(struct mg_iface *iface) { LOG(LL_INFO, ("%p Mongoose init", iface)); iface->data = MG_CALLOC(1, sizeof(struct mg_ev_mgr_lwip_data)); } void mg_lwip_if_free(struct mg_iface *iface) { MG_FREE(iface->data); iface->data = NULL; } void mg_lwip_if_add_conn(struct mg_connection *nc) { (void) nc; } void mg_lwip_if_remove_conn(struct mg_connection *nc) { struct mg_ev_mgr_lwip_data *md = (struct mg_ev_mgr_lwip_data *) nc->iface->data; /* Walk the queue and null-out further signals for this conn. */ for (int i = 0; i < MG_SIG_QUEUE_LEN; i++) { if (md->sig_queue[i].nc == nc) { md->sig_queue[i].sig = MG_SIG_TOMBSTONE; } } } time_t mg_lwip_if_poll(struct mg_iface *iface, int timeout_ms) { struct mg_mgr *mgr = iface->mgr; int n = 0; double now = mg_time(); struct mg_connection *nc, *tmp; double min_timer = 0; int num_timers = 0; #if 0 DBG(("begin poll @%u", (unsigned int) (now * 1000))); #endif mg_ev_mgr_lwip_process_signals(mgr); for (nc = mgr->active_connections; nc != NULL; nc = tmp) { struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; tmp = nc->next; n++; if ((nc->flags & MG_F_CLOSE_IMMEDIATELY) || ((nc->flags & MG_F_SEND_AND_CLOSE) && (nc->flags & MG_F_UDP) && (nc->send_mbuf.len == 0))) { mg_close_conn(nc); continue; } mg_if_poll(nc, now); mg_if_timer(nc, now); #if MG_ENABLE_SSL if ((nc->flags & MG_F_SSL) && cs != NULL && cs->pcb.tcp != NULL && cs->pcb.tcp->state == ESTABLISHED) { if (((nc->flags & MG_F_WANT_WRITE) || ((nc->send_mbuf.len > 0) && (nc->flags & MG_F_SSL_HANDSHAKE_DONE))) && cs->pcb.tcp->snd_buf > 0) { /* Can write more. */ if (nc->flags & MG_F_SSL_HANDSHAKE_DONE) { if (!(nc->flags & MG_F_CONNECTING)) mg_lwip_ssl_send(nc); } else { mg_lwip_ssl_do_hs(nc); } } if (cs->rx_chain != NULL || (nc->flags & MG_F_WANT_READ)) { if (nc->flags & MG_F_SSL_HANDSHAKE_DONE) { if (!(nc->flags & MG_F_CONNECTING)) mg_lwip_ssl_recv(nc); } else { mg_lwip_ssl_do_hs(nc); } } } else #endif /* MG_ENABLE_SSL */ { if (nc->send_mbuf.len > 0 && !(nc->flags & MG_F_CONNECTING)) { mg_lwip_send_more(nc); } } if (nc->sock != INVALID_SOCKET && !(nc->flags & (MG_F_UDP | MG_F_LISTENING)) && cs->pcb.tcp != NULL && cs->pcb.tcp->unsent != NULL) { tcpip_callback(tcp_output_tcpip, cs->pcb.tcp); } if (nc->ev_timer_time > 0) { if (num_timers == 0 || nc->ev_timer_time < min_timer) { min_timer = nc->ev_timer_time; } num_timers++; } if (nc->sock != INVALID_SOCKET) { /* Try to consume data from cs->rx_chain */ mg_lwip_consume_rx_chain_tcp(nc); /* * If the connection is about to close, and rx_chain is finally empty, * send the MG_SIG_CLOSE_CONN signal */ if (cs->draining_rx_chain && cs->rx_chain == NULL) { mg_lwip_post_signal(MG_SIG_CLOSE_CONN, nc); } } } #if 0 DBG(("end poll @%u, %d conns, %d timers (min %u), next in %d ms", (unsigned int) (now * 1000), n, num_timers, (unsigned int) (min_timer * 1000), timeout_ms)); #endif (void) timeout_ms; return now; } uint32_t mg_lwip_get_poll_delay_ms(struct mg_mgr *mgr) { struct mg_connection *nc; double now; double min_timer = 0; int num_timers = 0; mg_ev_mgr_lwip_process_signals(mgr); for (nc = mg_next(mgr, NULL); nc != NULL; nc = mg_next(mgr, nc)) { struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; if (nc->ev_timer_time > 0) { if (num_timers == 0 || nc->ev_timer_time < min_timer) { min_timer = nc->ev_timer_time; } num_timers++; } if (nc->send_mbuf.len > 0 #if MG_ENABLE_SSL || (nc->flags & MG_F_WANT_WRITE) #endif ) { int can_send = 0; /* We have stuff to send, but can we? */ if (nc->flags & MG_F_UDP) { /* UDP is always ready for sending. */ can_send = (cs->pcb.udp != NULL); } else { can_send = (cs->pcb.tcp != NULL && cs->pcb.tcp->snd_buf > 0); } /* We want and can send, request a poll immediately. */ if (can_send) return 0; } } uint32_t timeout_ms = ~0; now = mg_time(); if (num_timers > 0) { /* If we have a timer that is past due, do a poll ASAP. */ if (min_timer < now) return 0; double timer_timeout_ms = (min_timer - now) * 1000 + 1 /* rounding */; if (timer_timeout_ms < timeout_ms) { timeout_ms = timer_timeout_ms; } } return timeout_ms; } #endif /* MG_NET_IF == MG_NET_IF_LWIP_LOW_LEVEL */ #ifdef MG_MODULE_LINES #line 1 "common/platforms/lwip/mg_lwip_ssl_if.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #if MG_ENABLE_SSL && MG_NET_IF == MG_NET_IF_LWIP_LOW_LEVEL /* Amalgamated: #include "common/mg_mem.h" */ /* Amalgamated: #include "common/cs_dbg.h" */ #include #include #ifndef MG_LWIP_SSL_IO_SIZE #define MG_LWIP_SSL_IO_SIZE 1024 #endif #ifndef MIN #define MIN(a, b) ((a) < (b) ? (a) : (b)) #endif void mg_lwip_ssl_do_hs(struct mg_connection *nc) { struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; int server_side = (nc->listener != NULL); enum mg_ssl_if_result res; if (nc->flags & MG_F_CLOSE_IMMEDIATELY) return; res = mg_ssl_if_handshake(nc); DBG(("%p %lu %d %d", nc, nc->flags, server_side, res)); if (res != MG_SSL_OK) { if (res == MG_SSL_WANT_WRITE) { nc->flags |= MG_F_WANT_WRITE; cs->err = 0; } else if (res == MG_SSL_WANT_READ) { /* * Nothing to do in particular, we are callback-driven. * What we definitely do not need anymore is SSL reading (nothing left). */ nc->flags &= ~MG_F_WANT_READ; cs->err = 0; } else { cs->err = res; if (server_side) { mg_lwip_post_signal(MG_SIG_CLOSE_CONN, nc); } else { mg_lwip_post_signal(MG_SIG_CONNECT_RESULT, nc); } } } else { cs->err = 0; nc->flags &= ~MG_F_WANT_WRITE; /* * Handshake is done. Schedule a read immediately to consume app data * which may already be waiting. */ nc->flags |= (MG_F_SSL_HANDSHAKE_DONE | MG_F_WANT_READ); if (server_side) { mg_lwip_accept_conn(nc, cs->pcb.tcp); } else { mg_lwip_post_signal(MG_SIG_CONNECT_RESULT, nc); } } } void mg_lwip_ssl_send(struct mg_connection *nc) { if (nc->sock == INVALID_SOCKET) { DBG(("%p invalid socket", nc)); return; } struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; /* It's ok if the buffer is empty. Return value of 0 may also be valid. */ int len = cs->last_ssl_write_size; if (len == 0) { len = MIN(MG_LWIP_SSL_IO_SIZE, nc->send_mbuf.len); } int ret = mg_ssl_if_write(nc, nc->send_mbuf.buf, len); DBG(("%p SSL_write %u = %d", nc, len, ret)); if (ret > 0) { mg_if_sent_cb(nc, ret); cs->last_ssl_write_size = 0; } else if (ret < 0) { /* This is tricky. We must remember the exact data we were sending to retry * exactly the same send next time. */ cs->last_ssl_write_size = len; } if (ret == len) { nc->flags &= ~MG_F_WANT_WRITE; } else if (ret == MG_SSL_WANT_WRITE) { nc->flags |= MG_F_WANT_WRITE; } else { mg_lwip_post_signal(MG_SIG_CLOSE_CONN, nc); } } void mg_lwip_ssl_recv(struct mg_connection *nc) { struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; /* Don't deliver data before connect callback */ if (nc->flags & MG_F_CONNECTING) return; while (nc->recv_mbuf.len < nc->recv_mbuf_limit) { char *buf = (char *) MG_MALLOC(MG_LWIP_SSL_IO_SIZE); if (buf == NULL) return; int ret = mg_ssl_if_read(nc, buf, MG_LWIP_SSL_IO_SIZE); DBG(("%p %p SSL_read %u = %d", nc, cs->rx_chain, MG_LWIP_SSL_IO_SIZE, ret)); if (ret <= 0) { MG_FREE(buf); if (ret == MG_SSL_WANT_WRITE) { nc->flags |= MG_F_WANT_WRITE; return; } else if (ret == MG_SSL_WANT_READ) { /* * Nothing to do in particular, we are callback-driven. * What we definitely do not need anymore is SSL reading (nothing left). */ nc->flags &= ~MG_F_WANT_READ; cs->err = 0; return; } else { mg_lwip_post_signal(MG_SIG_CLOSE_CONN, nc); return; } } else { mg_if_recv_tcp_cb(nc, buf, ret, 1 /* own */); } } } #ifdef KR_VERSION ssize_t kr_send(int fd, const void *buf, size_t len) { struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) fd; int ret = mg_lwip_tcp_write(cs->nc, buf, len); DBG(("%p mg_lwip_tcp_write %u = %d", cs->nc, len, ret)); if (ret == 0) ret = KR_IO_WOULDBLOCK; return ret; } ssize_t kr_recv(int fd, void *buf, size_t len) { struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) fd; struct pbuf *seg = cs->rx_chain; if (seg == NULL) { DBG(("%u - nothing to read", len)); return KR_IO_WOULDBLOCK; } size_t seg_len = (seg->len - cs->rx_offset); DBG(("%u %u %u %u", len, cs->rx_chain->len, seg_len, cs->rx_chain->tot_len)); len = MIN(len, seg_len); pbuf_copy_partial(seg, buf, len, cs->rx_offset); cs->rx_offset += len; tcp_recved(cs->pcb.tcp, len); if (cs->rx_offset == cs->rx_chain->len) { cs->rx_chain = pbuf_dechain(cs->rx_chain); pbuf_free(seg); cs->rx_offset = 0; } return len; } #elif MG_SSL_IF == MG_SSL_IF_MBEDTLS int ssl_socket_send(void *ctx, const unsigned char *buf, size_t len) { struct mg_connection *nc = (struct mg_connection *) ctx; struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; int ret = mg_lwip_tcp_write(cs->nc, buf, len); if (ret == 0) ret = MBEDTLS_ERR_SSL_WANT_WRITE; LOG(LL_DEBUG, ("%p %d -> %d", nc, len, ret)); return ret; } int ssl_socket_recv(void *ctx, unsigned char *buf, size_t len) { struct mg_connection *nc = (struct mg_connection *) ctx; struct mg_lwip_conn_state *cs = (struct mg_lwip_conn_state *) nc->sock; struct pbuf *seg = cs->rx_chain; if (seg == NULL) { DBG(("%u - nothing to read", len)); return MBEDTLS_ERR_SSL_WANT_READ; } size_t seg_len = (seg->len - cs->rx_offset); DBG(("%u %u %u %u", len, cs->rx_chain->len, seg_len, cs->rx_chain->tot_len)); mgos_lock(); len = MIN(len, seg_len); pbuf_copy_partial(seg, buf, len, cs->rx_offset); cs->rx_offset += len; /* TCP PCB may be NULL if connection has already been closed * but we still have data to deliver to SSL. */ if (cs->pcb.tcp != NULL) tcp_recved(cs->pcb.tcp, len); if (cs->rx_offset == cs->rx_chain->len) { cs->rx_chain = pbuf_dechain(cs->rx_chain); pbuf_free(seg); cs->rx_offset = 0; } mgos_unlock(); LOG(LL_DEBUG, ("%p <- %d", nc, (int) len)); return len; } #endif #endif /* MG_ENABLE_SSL && MG_NET_IF == MG_NET_IF_LWIP_LOW_LEVEL */ #ifdef MG_MODULE_LINES #line 1 "common/platforms/wince/wince_libc.c" #endif /* * Copyright (c) 2016 Cesanta Software Limited * All rights reserved */ #ifdef WINCE const char *strerror(int err) { /* * TODO(alashkin): there is no strerror on WinCE; * look for similar wce_xxxx function */ static char buf[10]; snprintf(buf, sizeof(buf), "%d", err); return buf; } int open(const char *filename, int oflag, int pmode) { /* * TODO(alashkin): mg_open function is not used in mongoose * but exists in documentation as utility function * Shall we delete it at all or implement for WinCE as well? */ DebugBreak(); return 0; /* for compiler */ } int _wstati64(const wchar_t *path, cs_stat_t *st) { DWORD fa = GetFileAttributesW(path); if (fa == INVALID_FILE_ATTRIBUTES) { return -1; } memset(st, 0, sizeof(*st)); if ((fa & FILE_ATTRIBUTE_DIRECTORY) == 0) { HANDLE h; FILETIME ftime; st->st_mode |= _S_IFREG; h = CreateFileW(path, GENERIC_READ, 0, NULL, OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL, NULL); if (h == INVALID_HANDLE_VALUE) { return -1; } st->st_size = GetFileSize(h, NULL); GetFileTime(h, NULL, NULL, &ftime); st->st_mtime = (uint32_t)((((uint64_t) ftime.dwLowDateTime + ((uint64_t) ftime.dwHighDateTime << 32)) / 10000000.0) - 11644473600); CloseHandle(h); } else { st->st_mode |= _S_IFDIR; } return 0; } /* Windows CE doesn't have neither gmtime nor strftime */ static void mg_gmt_time_string(char *buf, size_t buf_len, time_t *t) { FILETIME ft; SYSTEMTIME systime; if (t != NULL) { uint64_t filetime = (*t + 11644473600) * 10000000; ft.dwLowDateTime = filetime & 0xFFFFFFFF; ft.dwHighDateTime = (filetime & 0xFFFFFFFF00000000) >> 32; FileTimeToSystemTime(&ft, &systime); } else { GetSystemTime(&systime); } /* There is no PRIu16 in WinCE SDK */ snprintf(buf, buf_len, "%d.%d.%d %d:%d:%d GMT", (int) systime.wYear, (int) systime.wMonth, (int) systime.wDay, (int) systime.wHour, (int) systime.wMinute, (int) systime.wSecond); } #endif #ifdef MG_MODULE_LINES #line 1 "common/platforms/pic32/pic32_net_if.h" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #ifndef CS_COMMON_PLATFORMS_PIC32_NET_IF_H_ #define CS_COMMON_PLATFORMS_PIC32_NET_IF_H_ /* Amalgamated: #include "mongoose/src/net_if.h" */ #ifdef __cplusplus extern "C" { #endif /* __cplusplus */ #ifndef MG_ENABLE_NET_IF_PIC32 #define MG_ENABLE_NET_IF_PIC32 MG_NET_IF == MG_NET_IF_PIC32 #endif extern const struct mg_iface_vtable mg_pic32_iface_vtable; #ifdef __cplusplus } #endif /* __cplusplus */ #endif /* CS_COMMON_PLATFORMS_PIC32_NET_IF_H_ */ #ifdef MG_MODULE_LINES #line 1 "common/platforms/pic32/pic32_net_if.c" #endif /* * Copyright (c) 2014-2016 Cesanta Software Limited * All rights reserved */ #if MG_ENABLE_NET_IF_PIC32 int mg_pic32_if_create_conn(struct mg_connection *nc) { (void) nc; return 1; } void mg_pic32_if_recved(struct mg_connection *nc, size_t len) { (void) nc; (void) len; } void mg_pic32_if_add_conn(struct mg_connection *nc) { (void) nc; } void mg_pic32_if_init(struct mg_iface *iface) { (void) iface; (void) mg_get_errno(); /* Shutup compiler */ } void mg_pic32_if_free(struct mg_iface *iface) { (void) iface; } void mg_pic32_if_remove_conn(struct mg_connection *nc) { (void) nc; } void mg_pic32_if_destroy_conn(struct mg_connection *nc) { if (nc->sock == INVALID_SOCKET) return; /* For UDP, only close outgoing sockets or listeners. */ if (!(nc->flags & MG_F_UDP)) { /* Close TCP */ TCPIP_TCP_Close((TCP_SOCKET) nc->sock); } else if (nc->listener == NULL) { /* Only close outgoing UDP or listeners. */ TCPIP_UDP_Close((UDP_SOCKET) nc->sock); } nc->sock = INVALID_SOCKET; } int mg_pic32_if_listen_udp(struct mg_connection *nc, union socket_address *sa) { nc->sock = TCPIP_UDP_ServerOpen( sa->sin.sin_family == AF_INET ? IP_ADDRESS_TYPE_IPV4 : IP_ADDRESS_TYPE_IPV6, ntohs(sa->sin.sin_port), sa->sin.sin_addr.s_addr == 0 ? 0 : (IP_MULTI_ADDRESS *) &sa->sin); if (nc->sock == INVALID_SOCKET) { return -1; } return 0; } void mg_pic32_if_udp_send(struct mg_connection *nc, const void *buf, size_t len) { mbuf_append(&nc->send_mbuf, buf, len); } void mg_pic32_if_tcp_send(struct mg_connection *nc, const void *buf, size_t len) { mbuf_append(&nc->send_mbuf, buf, len); } int mg_pic32_if_listen_tcp(struct mg_connection *nc, union socket_address *sa) { nc->sock = TCPIP_TCP_ServerOpen( sa->sin.sin_family == AF_INET ? IP_ADDRESS_TYPE_IPV4 : IP_ADDRESS_TYPE_IPV6, ntohs(sa->sin.sin_port), sa->sin.sin_addr.s_addr == 0 ? 0 : (IP_MULTI_ADDRESS *) &sa->sin); memcpy(&nc->sa, sa, sizeof(*sa)); if (nc->sock == INVALID_SOCKET) { return -1; } return 0; } static int mg_accept_conn(struct mg_connection *lc) { struct mg_connection *nc; TCP_SOCKET_INFO si; union socket_address sa; nc = mg_if_accept_new_conn(lc); if (nc == NULL) { return 0; } nc->sock = lc->sock; nc->flags &= ~MG_F_LISTENING; if (!TCPIP_TCP_SocketInfoGet((TCP_SOCKET) nc->sock, &si)) { return 0; } if (si.addressType == IP_ADDRESS_TYPE_IPV4) { sa.sin.sin_family = AF_INET; sa.sin.sin_port = htons(si.remotePort); sa.sin.sin_addr.s_addr = si.remoteIPaddress.v4Add.Val; } else { /* TODO(alashkin): do something with _potential_ IPv6 */ memset(&sa, 0, sizeof(sa)); } mg_if_accept_tcp_cb(nc, (union socket_address *) &sa, sizeof(sa)); return mg_pic32_if_listen_tcp(lc, &lc->sa) >= 0; } char *inet_ntoa(struct in_addr in) { static char addr[17]; snprintf(addr, sizeof(addr), "%d.%d.%d.%d", (int) in.S_un.S_un_b.s_b1, (int) in.S_un.S_un_b.s_b2, (int) in.S_un.S_un_b.s_b3, (int) in.S_un.S_un_b.s_b4); return addr; } static void mg_handle_send(struct mg_connection *nc) { uint16_t bytes_written = 0; if (nc->flags & MG_F_UDP) { if (!TCPIP_UDP_RemoteBind( (UDP_SOCKET) nc->sock, nc->sa.sin.sin_family == AF_INET ? IP_ADDRESS_TYPE_IPV4 : IP_ADDRESS_TYPE_IPV6, ntohs(nc->sa.sin.sin_port), (IP_MULTI_ADDRESS *) &nc->sa.sin)) { nc->flags |= MG_F_CLOSE_IMMEDIATELY; return; } bytes_written = TCPIP_UDP_TxPutIsReady((UDP_SOCKET) nc->sock, 0); if (bytes_written >= nc->send_mbuf.len) { if (TCPIP_UDP_ArrayPut((UDP_SOCKET) nc->sock, (uint8_t *) nc->send_mbuf.buf, nc->send_mbuf.len) != nc->send_mbuf.len) { nc->flags |= MG_F_CLOSE_IMMEDIATELY; bytes_written = 0; } } } else { bytes_written = TCPIP_TCP_FifoTxFreeGet((TCP_SOCKET) nc->sock); if (bytes_written != 0) { if (bytes_written > nc->send_mbuf.len) { bytes_written = nc->send_mbuf.len; } if (TCPIP_TCP_ArrayPut((TCP_SOCKET) nc->sock, (uint8_t *) nc->send_mbuf.buf, bytes_written) != bytes_written) { nc->flags |= MG_F_CLOSE_IMMEDIATELY; bytes_written = 0; } } } mg_if_sent_cb(nc, bytes_written); } static void mg_handle_recv(struct mg_connection *nc) { uint16_t bytes_read = 0; uint8_t *buf = NULL; if (nc->flags & MG_F_UDP) { bytes_read = TCPIP_UDP_GetIsReady((UDP_SOCKET) nc->sock); if (bytes_read != 0 && (nc->recv_mbuf_limit == -1 || nc->recv_mbuf.len + bytes_read < nc->recv_mbuf_limit)) { buf = (uint8_t *) MG_MALLOC(bytes_read); if (TCPIP_UDP_ArrayGet((UDP_SOCKET) nc->sock, buf, bytes_read) != bytes_read) { nc->flags |= MG_F_CLOSE_IMMEDIATELY; bytes_read = 0; MG_FREE(buf); } } } else { bytes_read = TCPIP_TCP_GetIsReady((TCP_SOCKET) nc->sock); if (bytes_read != 0) { if (nc->recv_mbuf_limit != -1 && nc->recv_mbuf_limit - nc->recv_mbuf.len > bytes_read) { bytes_read = nc->recv_mbuf_limit - nc->recv_mbuf.len; } buf = (uint8_t *) MG_MALLOC(bytes_read); if (TCPIP_TCP_ArrayGet((TCP_SOCKET) nc->sock, buf, bytes_read) != bytes_read) { nc->flags |= MG_F_CLOSE_IMMEDIATELY; MG_FREE(buf); bytes_read = 0; } } } if (bytes_read != 0) { mg_if_recv_tcp_cb(nc, buf, bytes_read, 1 /* own */); } } time_t mg_pic32_if_poll(struct mg_iface *iface, int timeout_ms) { struct mg_mgr *mgr = iface->mgr; double now = mg_time(); struct mg_connection *nc, *tmp; for (nc = mgr->active_connections; nc != NULL; nc = tmp) { tmp = nc->next; if (nc->flags & MG_F_CONNECTING) { /* processing connections */ if (nc->flags & MG_F_UDP || TCPIP_TCP_IsConnected((TCP_SOCKET) nc->sock)) { mg_if_connect_cb(nc, 0); } } else if (nc->flags & MG_F_LISTENING) { if (TCPIP_TCP_IsConnected((TCP_SOCKET) nc->sock)) { /* accept new connections */ mg_accept_conn(nc); } } else { if (nc->send_mbuf.len != 0) { mg_handle_send(nc); } if (nc->recv_mbuf_limit == -1 || nc->recv_mbuf.len < nc->recv_mbuf_limit) { mg_handle_recv(nc); } } } for (nc = mgr->active_connections; nc != NULL; nc = tmp) { tmp = nc->next; if ((nc->flags & MG_F_CLOSE_IMMEDIATELY) || (nc->send_mbuf.len == 0 && (nc->flags & MG_F_SEND_AND_CLOSE))) { mg_close_conn(nc); } } return now; } void mg_pic32_if_sock_set(struct mg_connection *nc, sock_t sock) { nc->sock = sock; } void mg_pic32_if_get_conn_addr(struct mg_connection *nc, int remote, union socket_address *sa) { /* TODO(alaskin): not implemented yet */ } void mg_pic32_if_connect_tcp(struct mg_connection *nc, const union socket_address *sa) { nc->sock = TCPIP_TCP_ClientOpen( sa->sin.sin_family == AF_INET ? IP_ADDRESS_TYPE_IPV4 : IP_ADDRESS_TYPE_IPV6, ntohs(sa->sin.sin_port), (IP_MULTI_ADDRESS *) &sa->sin); nc->err = (nc->sock == INVALID_SOCKET) ? -1 : 0; } void mg_pic32_if_connect_udp(struct mg_connection *nc) { nc->sock = TCPIP_UDP_ClientOpen(IP_ADDRESS_TYPE_ANY, 0, NULL); nc->err = (nc->sock == INVALID_SOCKET) ? -1 : 0; } /* clang-format off */ #define MG_PIC32_IFACE_VTABLE \ { \ mg_pic32_if_init, \ mg_pic32_if_free, \ mg_pic32_if_add_conn, \ mg_pic32_if_remove_conn, \ mg_pic32_if_poll, \ mg_pic32_if_listen_tcp, \ mg_pic32_if_listen_udp, \ mg_pic32_if_connect_tcp, \ mg_pic32_if_connect_udp, \ mg_pic32_if_tcp_send, \ mg_pic32_if_udp_send, \ mg_pic32_if_recved, \ mg_pic32_if_create_conn, \ mg_pic32_if_destroy_conn, \ mg_pic32_if_sock_set, \ mg_pic32_if_get_conn_addr, \ } /* clang-format on */ const struct mg_iface_vtable mg_pic32_iface_vtable = MG_PIC32_IFACE_VTABLE; #if MG_NET_IF == MG_NET_IF_PIC32 const struct mg_iface_vtable mg_default_iface_vtable = MG_PIC32_IFACE_VTABLE; #endif #endif /* MG_ENABLE_NET_IF_PIC32 */ #ifdef MG_MODULE_LINES #line 1 "common/platforms/windows/windows_direct.c" #endif /* * Copyright (c) 2017 Cesanta Software Limited * All rights reserved */ #ifdef _WIN32 int rmdir(const char *dirname) { return _rmdir(dirname); } unsigned int sleep(unsigned int seconds) { Sleep(seconds * 1000); return 0; } #endif /* _WIN32 */